{ "best_metric": null, "best_model_checkpoint": null, "epoch": 8.0, "eval_steps": 500, "global_step": 43624, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 1.5278838808250573e-08, "loss": 1.2951, "step": 1 }, { "epoch": 0.0, "learning_rate": 3.0557677616501146e-08, "loss": 1.1738, "step": 2 }, { "epoch": 0.0, "learning_rate": 4.5836516424751726e-08, "loss": 1.2578, "step": 3 }, { "epoch": 0.0, "learning_rate": 6.111535523300229e-08, "loss": 1.2314, "step": 4 }, { "epoch": 0.0, "learning_rate": 7.639419404125288e-08, "loss": 1.2841, "step": 5 }, { "epoch": 0.0, "learning_rate": 9.167303284950345e-08, "loss": 1.2511, "step": 6 }, { "epoch": 0.0, "learning_rate": 1.0695187165775401e-07, "loss": 1.2549, "step": 7 }, { "epoch": 0.0, "learning_rate": 1.2223071046600458e-07, "loss": 1.2959, "step": 8 }, { "epoch": 0.0, "learning_rate": 1.3750954927425516e-07, "loss": 1.3513, "step": 9 }, { "epoch": 0.0, "learning_rate": 1.5278838808250576e-07, "loss": 1.2275, "step": 10 }, { "epoch": 0.0, "learning_rate": 1.680672268907563e-07, "loss": 1.3077, "step": 11 }, { "epoch": 0.0, "learning_rate": 1.833460656990069e-07, "loss": 1.1737, "step": 12 }, { "epoch": 0.0, "learning_rate": 1.9862490450725748e-07, "loss": 1.279, "step": 13 }, { "epoch": 0.0, "learning_rate": 2.1390374331550802e-07, "loss": 1.2623, "step": 14 }, { "epoch": 0.0, "learning_rate": 2.2918258212375862e-07, "loss": 1.262, "step": 15 }, { "epoch": 0.0, "learning_rate": 2.4446142093200917e-07, "loss": 1.232, "step": 16 }, { "epoch": 0.0, "learning_rate": 2.597402597402598e-07, "loss": 1.2254, "step": 17 }, { "epoch": 0.0, "learning_rate": 2.750190985485103e-07, "loss": 1.2176, "step": 18 }, { "epoch": 0.0, "learning_rate": 2.902979373567609e-07, "loss": 1.2508, "step": 19 }, { "epoch": 0.0, "learning_rate": 3.055767761650115e-07, "loss": 1.2598, "step": 20 }, { "epoch": 0.0, "learning_rate": 3.208556149732621e-07, "loss": 1.2163, "step": 21 }, { "epoch": 0.0, "learning_rate": 3.361344537815126e-07, "loss": 1.1838, "step": 22 }, { "epoch": 0.0, "learning_rate": 3.5141329258976323e-07, "loss": 1.2465, "step": 23 }, { "epoch": 0.0, "learning_rate": 3.666921313980138e-07, "loss": 1.2495, "step": 24 }, { "epoch": 0.0, "learning_rate": 3.819709702062643e-07, "loss": 1.1798, "step": 25 }, { "epoch": 0.0, "learning_rate": 3.9724980901451495e-07, "loss": 1.1345, "step": 26 }, { "epoch": 0.0, "learning_rate": 4.125286478227655e-07, "loss": 1.171, "step": 27 }, { "epoch": 0.01, "learning_rate": 4.2780748663101604e-07, "loss": 1.18, "step": 28 }, { "epoch": 0.01, "learning_rate": 4.4308632543926667e-07, "loss": 1.1516, "step": 29 }, { "epoch": 0.01, "learning_rate": 4.5836516424751724e-07, "loss": 1.156, "step": 30 }, { "epoch": 0.01, "learning_rate": 4.7364400305576776e-07, "loss": 1.1613, "step": 31 }, { "epoch": 0.01, "learning_rate": 4.889228418640183e-07, "loss": 1.1912, "step": 32 }, { "epoch": 0.01, "learning_rate": 5.042016806722689e-07, "loss": 1.1251, "step": 33 }, { "epoch": 0.01, "learning_rate": 5.194805194805196e-07, "loss": 1.0314, "step": 34 }, { "epoch": 0.01, "learning_rate": 5.347593582887701e-07, "loss": 1.1108, "step": 35 }, { "epoch": 0.01, "learning_rate": 5.500381970970206e-07, "loss": 1.0259, "step": 36 }, { "epoch": 0.01, "learning_rate": 5.653170359052713e-07, "loss": 1.042, "step": 37 }, { "epoch": 0.01, "learning_rate": 5.805958747135218e-07, "loss": 1.0327, "step": 38 }, { "epoch": 0.01, "learning_rate": 5.958747135217723e-07, "loss": 1.0519, "step": 39 }, { "epoch": 0.01, "learning_rate": 6.11153552330023e-07, "loss": 1.0861, "step": 40 }, { "epoch": 0.01, "learning_rate": 6.264323911382736e-07, "loss": 0.9968, "step": 41 }, { "epoch": 0.01, "learning_rate": 6.417112299465242e-07, "loss": 0.9551, "step": 42 }, { "epoch": 0.01, "learning_rate": 6.569900687547746e-07, "loss": 1.0212, "step": 43 }, { "epoch": 0.01, "learning_rate": 6.722689075630252e-07, "loss": 0.9783, "step": 44 }, { "epoch": 0.01, "learning_rate": 6.875477463712758e-07, "loss": 0.9769, "step": 45 }, { "epoch": 0.01, "learning_rate": 7.028265851795265e-07, "loss": 0.8886, "step": 46 }, { "epoch": 0.01, "learning_rate": 7.18105423987777e-07, "loss": 0.8542, "step": 47 }, { "epoch": 0.01, "learning_rate": 7.333842627960276e-07, "loss": 0.8128, "step": 48 }, { "epoch": 0.01, "learning_rate": 7.486631016042781e-07, "loss": 0.7922, "step": 49 }, { "epoch": 0.01, "learning_rate": 7.639419404125286e-07, "loss": 0.7489, "step": 50 }, { "epoch": 0.01, "learning_rate": 7.792207792207792e-07, "loss": 0.7519, "step": 51 }, { "epoch": 0.01, "learning_rate": 7.944996180290299e-07, "loss": 0.7737, "step": 52 }, { "epoch": 0.01, "learning_rate": 8.097784568372805e-07, "loss": 0.7548, "step": 53 }, { "epoch": 0.01, "learning_rate": 8.25057295645531e-07, "loss": 0.6927, "step": 54 }, { "epoch": 0.01, "learning_rate": 8.403361344537816e-07, "loss": 0.7285, "step": 55 }, { "epoch": 0.01, "learning_rate": 8.556149732620321e-07, "loss": 0.7535, "step": 56 }, { "epoch": 0.01, "learning_rate": 8.708938120702827e-07, "loss": 0.7114, "step": 57 }, { "epoch": 0.01, "learning_rate": 8.861726508785333e-07, "loss": 0.6959, "step": 58 }, { "epoch": 0.01, "learning_rate": 9.014514896867839e-07, "loss": 0.7235, "step": 59 }, { "epoch": 0.01, "learning_rate": 9.167303284950345e-07, "loss": 0.6971, "step": 60 }, { "epoch": 0.01, "learning_rate": 9.320091673032851e-07, "loss": 0.6759, "step": 61 }, { "epoch": 0.01, "learning_rate": 9.472880061115355e-07, "loss": 0.6724, "step": 62 }, { "epoch": 0.01, "learning_rate": 9.625668449197862e-07, "loss": 0.6529, "step": 63 }, { "epoch": 0.01, "learning_rate": 9.778456837280367e-07, "loss": 0.6314, "step": 64 }, { "epoch": 0.01, "learning_rate": 9.931245225362874e-07, "loss": 0.6438, "step": 65 }, { "epoch": 0.01, "learning_rate": 1.0084033613445378e-06, "loss": 0.63, "step": 66 }, { "epoch": 0.01, "learning_rate": 1.0236822001527885e-06, "loss": 0.6225, "step": 67 }, { "epoch": 0.01, "learning_rate": 1.0389610389610392e-06, "loss": 0.6169, "step": 68 }, { "epoch": 0.01, "learning_rate": 1.0542398777692896e-06, "loss": 0.5914, "step": 69 }, { "epoch": 0.01, "learning_rate": 1.0695187165775401e-06, "loss": 0.6013, "step": 70 }, { "epoch": 0.01, "learning_rate": 1.0847975553857908e-06, "loss": 0.5777, "step": 71 }, { "epoch": 0.01, "learning_rate": 1.1000763941940413e-06, "loss": 0.5802, "step": 72 }, { "epoch": 0.01, "learning_rate": 1.115355233002292e-06, "loss": 0.5733, "step": 73 }, { "epoch": 0.01, "learning_rate": 1.1306340718105426e-06, "loss": 0.5699, "step": 74 }, { "epoch": 0.01, "learning_rate": 1.145912910618793e-06, "loss": 0.5547, "step": 75 }, { "epoch": 0.01, "learning_rate": 1.1611917494270435e-06, "loss": 0.5546, "step": 76 }, { "epoch": 0.01, "learning_rate": 1.1764705882352942e-06, "loss": 0.5191, "step": 77 }, { "epoch": 0.01, "learning_rate": 1.1917494270435447e-06, "loss": 0.5163, "step": 78 }, { "epoch": 0.01, "learning_rate": 1.2070282658517954e-06, "loss": 0.5435, "step": 79 }, { "epoch": 0.01, "learning_rate": 1.222307104660046e-06, "loss": 0.5346, "step": 80 }, { "epoch": 0.01, "learning_rate": 1.2375859434682965e-06, "loss": 0.5251, "step": 81 }, { "epoch": 0.02, "learning_rate": 1.2528647822765472e-06, "loss": 0.5245, "step": 82 }, { "epoch": 0.02, "learning_rate": 1.2681436210847979e-06, "loss": 0.5151, "step": 83 }, { "epoch": 0.02, "learning_rate": 1.2834224598930483e-06, "loss": 0.5251, "step": 84 }, { "epoch": 0.02, "learning_rate": 1.2987012987012986e-06, "loss": 0.5072, "step": 85 }, { "epoch": 0.02, "learning_rate": 1.3139801375095493e-06, "loss": 0.5062, "step": 86 }, { "epoch": 0.02, "learning_rate": 1.3292589763178e-06, "loss": 0.5041, "step": 87 }, { "epoch": 0.02, "learning_rate": 1.3445378151260504e-06, "loss": 0.5138, "step": 88 }, { "epoch": 0.02, "learning_rate": 1.359816653934301e-06, "loss": 0.5254, "step": 89 }, { "epoch": 0.02, "learning_rate": 1.3750954927425516e-06, "loss": 0.4805, "step": 90 }, { "epoch": 0.02, "learning_rate": 1.3903743315508022e-06, "loss": 0.5014, "step": 91 }, { "epoch": 0.02, "learning_rate": 1.405653170359053e-06, "loss": 0.4878, "step": 92 }, { "epoch": 0.02, "learning_rate": 1.4209320091673034e-06, "loss": 0.4977, "step": 93 }, { "epoch": 0.02, "learning_rate": 1.436210847975554e-06, "loss": 0.4771, "step": 94 }, { "epoch": 0.02, "learning_rate": 1.4514896867838045e-06, "loss": 0.467, "step": 95 }, { "epoch": 0.02, "learning_rate": 1.4667685255920552e-06, "loss": 0.4853, "step": 96 }, { "epoch": 0.02, "learning_rate": 1.482047364400306e-06, "loss": 0.498, "step": 97 }, { "epoch": 0.02, "learning_rate": 1.4973262032085562e-06, "loss": 0.4807, "step": 98 }, { "epoch": 0.02, "learning_rate": 1.5126050420168068e-06, "loss": 0.4838, "step": 99 }, { "epoch": 0.02, "learning_rate": 1.5278838808250573e-06, "loss": 0.4582, "step": 100 }, { "epoch": 0.02, "learning_rate": 1.543162719633308e-06, "loss": 0.4744, "step": 101 }, { "epoch": 0.02, "learning_rate": 1.5584415584415584e-06, "loss": 0.4814, "step": 102 }, { "epoch": 0.02, "learning_rate": 1.5737203972498091e-06, "loss": 0.4703, "step": 103 }, { "epoch": 0.02, "learning_rate": 1.5889992360580598e-06, "loss": 0.4799, "step": 104 }, { "epoch": 0.02, "learning_rate": 1.6042780748663103e-06, "loss": 0.4612, "step": 105 }, { "epoch": 0.02, "learning_rate": 1.619556913674561e-06, "loss": 0.4404, "step": 106 }, { "epoch": 0.02, "learning_rate": 1.6348357524828114e-06, "loss": 0.4556, "step": 107 }, { "epoch": 0.02, "learning_rate": 1.650114591291062e-06, "loss": 0.4447, "step": 108 }, { "epoch": 0.02, "learning_rate": 1.6653934300993128e-06, "loss": 0.4492, "step": 109 }, { "epoch": 0.02, "learning_rate": 1.6806722689075632e-06, "loss": 0.4543, "step": 110 }, { "epoch": 0.02, "learning_rate": 1.6959511077158137e-06, "loss": 0.4546, "step": 111 }, { "epoch": 0.02, "learning_rate": 1.7112299465240642e-06, "loss": 0.4363, "step": 112 }, { "epoch": 0.02, "learning_rate": 1.7265087853323149e-06, "loss": 0.4311, "step": 113 }, { "epoch": 0.02, "learning_rate": 1.7417876241405653e-06, "loss": 0.4365, "step": 114 }, { "epoch": 0.02, "learning_rate": 1.757066462948816e-06, "loss": 0.4347, "step": 115 }, { "epoch": 0.02, "learning_rate": 1.7723453017570667e-06, "loss": 0.4236, "step": 116 }, { "epoch": 0.02, "learning_rate": 1.7876241405653171e-06, "loss": 0.4076, "step": 117 }, { "epoch": 0.02, "learning_rate": 1.8029029793735678e-06, "loss": 0.4123, "step": 118 }, { "epoch": 0.02, "learning_rate": 1.8181818181818183e-06, "loss": 0.4026, "step": 119 }, { "epoch": 0.02, "learning_rate": 1.833460656990069e-06, "loss": 0.3852, "step": 120 }, { "epoch": 0.02, "learning_rate": 1.8487394957983196e-06, "loss": 0.3839, "step": 121 }, { "epoch": 0.02, "learning_rate": 1.8640183346065701e-06, "loss": 0.3847, "step": 122 }, { "epoch": 0.02, "learning_rate": 1.8792971734148208e-06, "loss": 0.3779, "step": 123 }, { "epoch": 0.02, "learning_rate": 1.894576012223071e-06, "loss": 0.3952, "step": 124 }, { "epoch": 0.02, "learning_rate": 1.9098548510313215e-06, "loss": 0.3677, "step": 125 }, { "epoch": 0.02, "learning_rate": 1.9251336898395724e-06, "loss": 0.3702, "step": 126 }, { "epoch": 0.02, "learning_rate": 1.940412528647823e-06, "loss": 0.3637, "step": 127 }, { "epoch": 0.02, "learning_rate": 1.9556913674560733e-06, "loss": 0.3644, "step": 128 }, { "epoch": 0.02, "learning_rate": 1.9709702062643242e-06, "loss": 0.3653, "step": 129 }, { "epoch": 0.02, "learning_rate": 1.9862490450725747e-06, "loss": 0.3626, "step": 130 }, { "epoch": 0.02, "learning_rate": 2.001527883880825e-06, "loss": 0.3637, "step": 131 }, { "epoch": 0.02, "learning_rate": 2.0168067226890756e-06, "loss": 0.3298, "step": 132 }, { "epoch": 0.02, "learning_rate": 2.0320855614973265e-06, "loss": 0.3229, "step": 133 }, { "epoch": 0.02, "learning_rate": 2.047364400305577e-06, "loss": 0.353, "step": 134 }, { "epoch": 0.02, "learning_rate": 2.0626432391138275e-06, "loss": 0.3311, "step": 135 }, { "epoch": 0.02, "learning_rate": 2.0779220779220784e-06, "loss": 0.3241, "step": 136 }, { "epoch": 0.03, "learning_rate": 2.0932009167303284e-06, "loss": 0.3552, "step": 137 }, { "epoch": 0.03, "learning_rate": 2.1084797555385793e-06, "loss": 0.3306, "step": 138 }, { "epoch": 0.03, "learning_rate": 2.1237585943468298e-06, "loss": 0.3229, "step": 139 }, { "epoch": 0.03, "learning_rate": 2.1390374331550802e-06, "loss": 0.3094, "step": 140 }, { "epoch": 0.03, "learning_rate": 2.154316271963331e-06, "loss": 0.3097, "step": 141 }, { "epoch": 0.03, "learning_rate": 2.1695951107715816e-06, "loss": 0.3197, "step": 142 }, { "epoch": 0.03, "learning_rate": 2.184873949579832e-06, "loss": 0.3162, "step": 143 }, { "epoch": 0.03, "learning_rate": 2.2001527883880825e-06, "loss": 0.3015, "step": 144 }, { "epoch": 0.03, "learning_rate": 2.2154316271963334e-06, "loss": 0.3018, "step": 145 }, { "epoch": 0.03, "learning_rate": 2.230710466004584e-06, "loss": 0.3108, "step": 146 }, { "epoch": 0.03, "learning_rate": 2.2459893048128343e-06, "loss": 0.299, "step": 147 }, { "epoch": 0.03, "learning_rate": 2.2612681436210852e-06, "loss": 0.3048, "step": 148 }, { "epoch": 0.03, "learning_rate": 2.2765469824293357e-06, "loss": 0.2907, "step": 149 }, { "epoch": 0.03, "learning_rate": 2.291825821237586e-06, "loss": 0.3075, "step": 150 }, { "epoch": 0.03, "learning_rate": 2.3071046600458366e-06, "loss": 0.2957, "step": 151 }, { "epoch": 0.03, "learning_rate": 2.322383498854087e-06, "loss": 0.2843, "step": 152 }, { "epoch": 0.03, "learning_rate": 2.337662337662338e-06, "loss": 0.2856, "step": 153 }, { "epoch": 0.03, "learning_rate": 2.3529411764705885e-06, "loss": 0.2922, "step": 154 }, { "epoch": 0.03, "learning_rate": 2.368220015278839e-06, "loss": 0.2735, "step": 155 }, { "epoch": 0.03, "learning_rate": 2.3834988540870894e-06, "loss": 0.2767, "step": 156 }, { "epoch": 0.03, "learning_rate": 2.3987776928953403e-06, "loss": 0.2787, "step": 157 }, { "epoch": 0.03, "learning_rate": 2.4140565317035907e-06, "loss": 0.2687, "step": 158 }, { "epoch": 0.03, "learning_rate": 2.4293353705118412e-06, "loss": 0.2796, "step": 159 }, { "epoch": 0.03, "learning_rate": 2.444614209320092e-06, "loss": 0.2719, "step": 160 }, { "epoch": 0.03, "learning_rate": 2.4598930481283426e-06, "loss": 0.2856, "step": 161 }, { "epoch": 0.03, "learning_rate": 2.475171886936593e-06, "loss": 0.2726, "step": 162 }, { "epoch": 0.03, "learning_rate": 2.4904507257448435e-06, "loss": 0.2802, "step": 163 }, { "epoch": 0.03, "learning_rate": 2.5057295645530944e-06, "loss": 0.2695, "step": 164 }, { "epoch": 0.03, "learning_rate": 2.521008403361345e-06, "loss": 0.2669, "step": 165 }, { "epoch": 0.03, "learning_rate": 2.5362872421695958e-06, "loss": 0.2681, "step": 166 }, { "epoch": 0.03, "learning_rate": 2.551566080977846e-06, "loss": 0.256, "step": 167 }, { "epoch": 0.03, "learning_rate": 2.5668449197860967e-06, "loss": 0.2699, "step": 168 }, { "epoch": 0.03, "learning_rate": 2.582123758594347e-06, "loss": 0.2568, "step": 169 }, { "epoch": 0.03, "learning_rate": 2.597402597402597e-06, "loss": 0.2663, "step": 170 }, { "epoch": 0.03, "learning_rate": 2.612681436210848e-06, "loss": 0.2591, "step": 171 }, { "epoch": 0.03, "learning_rate": 2.6279602750190986e-06, "loss": 0.2588, "step": 172 }, { "epoch": 0.03, "learning_rate": 2.6432391138273494e-06, "loss": 0.2517, "step": 173 }, { "epoch": 0.03, "learning_rate": 2.6585179526356e-06, "loss": 0.2642, "step": 174 }, { "epoch": 0.03, "learning_rate": 2.673796791443851e-06, "loss": 0.2381, "step": 175 }, { "epoch": 0.03, "learning_rate": 2.689075630252101e-06, "loss": 0.2661, "step": 176 }, { "epoch": 0.03, "learning_rate": 2.7043544690603517e-06, "loss": 0.2578, "step": 177 }, { "epoch": 0.03, "learning_rate": 2.719633307868602e-06, "loss": 0.248, "step": 178 }, { "epoch": 0.03, "learning_rate": 2.734912146676853e-06, "loss": 0.2376, "step": 179 }, { "epoch": 0.03, "learning_rate": 2.750190985485103e-06, "loss": 0.2582, "step": 180 }, { "epoch": 0.03, "learning_rate": 2.765469824293354e-06, "loss": 0.2397, "step": 181 }, { "epoch": 0.03, "learning_rate": 2.7807486631016045e-06, "loss": 0.2524, "step": 182 }, { "epoch": 0.03, "learning_rate": 2.796027501909855e-06, "loss": 0.2342, "step": 183 }, { "epoch": 0.03, "learning_rate": 2.811306340718106e-06, "loss": 0.2366, "step": 184 }, { "epoch": 0.03, "learning_rate": 2.826585179526356e-06, "loss": 0.2386, "step": 185 }, { "epoch": 0.03, "learning_rate": 2.841864018334607e-06, "loss": 0.2256, "step": 186 }, { "epoch": 0.03, "learning_rate": 2.8571428571428573e-06, "loss": 0.2501, "step": 187 }, { "epoch": 0.03, "learning_rate": 2.872421695951108e-06, "loss": 0.224, "step": 188 }, { "epoch": 0.03, "learning_rate": 2.8877005347593586e-06, "loss": 0.246, "step": 189 }, { "epoch": 0.03, "learning_rate": 2.902979373567609e-06, "loss": 0.2199, "step": 190 }, { "epoch": 0.04, "learning_rate": 2.9182582123758595e-06, "loss": 0.2373, "step": 191 }, { "epoch": 0.04, "learning_rate": 2.9335370511841104e-06, "loss": 0.2499, "step": 192 }, { "epoch": 0.04, "learning_rate": 2.948815889992361e-06, "loss": 0.224, "step": 193 }, { "epoch": 0.04, "learning_rate": 2.964094728800612e-06, "loss": 0.2466, "step": 194 }, { "epoch": 0.04, "learning_rate": 2.979373567608862e-06, "loss": 0.2356, "step": 195 }, { "epoch": 0.04, "learning_rate": 2.9946524064171123e-06, "loss": 0.2269, "step": 196 }, { "epoch": 0.04, "learning_rate": 3.009931245225363e-06, "loss": 0.2306, "step": 197 }, { "epoch": 0.04, "learning_rate": 3.0252100840336137e-06, "loss": 0.2395, "step": 198 }, { "epoch": 0.04, "learning_rate": 3.0404889228418646e-06, "loss": 0.2197, "step": 199 }, { "epoch": 0.04, "learning_rate": 3.0557677616501146e-06, "loss": 0.2281, "step": 200 }, { "epoch": 0.04, "learning_rate": 3.0710466004583655e-06, "loss": 0.241, "step": 201 }, { "epoch": 0.04, "learning_rate": 3.086325439266616e-06, "loss": 0.2229, "step": 202 }, { "epoch": 0.04, "learning_rate": 3.101604278074867e-06, "loss": 0.218, "step": 203 }, { "epoch": 0.04, "learning_rate": 3.116883116883117e-06, "loss": 0.2141, "step": 204 }, { "epoch": 0.04, "learning_rate": 3.1321619556913678e-06, "loss": 0.2247, "step": 205 }, { "epoch": 0.04, "learning_rate": 3.1474407944996182e-06, "loss": 0.2161, "step": 206 }, { "epoch": 0.04, "learning_rate": 3.162719633307869e-06, "loss": 0.2282, "step": 207 }, { "epoch": 0.04, "learning_rate": 3.1779984721161196e-06, "loss": 0.2143, "step": 208 }, { "epoch": 0.04, "learning_rate": 3.1932773109243696e-06, "loss": 0.2157, "step": 209 }, { "epoch": 0.04, "learning_rate": 3.2085561497326205e-06, "loss": 0.2245, "step": 210 }, { "epoch": 0.04, "learning_rate": 3.223834988540871e-06, "loss": 0.2205, "step": 211 }, { "epoch": 0.04, "learning_rate": 3.239113827349122e-06, "loss": 0.2265, "step": 212 }, { "epoch": 0.04, "learning_rate": 3.254392666157372e-06, "loss": 0.2279, "step": 213 }, { "epoch": 0.04, "learning_rate": 3.269671504965623e-06, "loss": 0.213, "step": 214 }, { "epoch": 0.04, "learning_rate": 3.2849503437738733e-06, "loss": 0.222, "step": 215 }, { "epoch": 0.04, "learning_rate": 3.300229182582124e-06, "loss": 0.2225, "step": 216 }, { "epoch": 0.04, "learning_rate": 3.3155080213903747e-06, "loss": 0.2072, "step": 217 }, { "epoch": 0.04, "learning_rate": 3.3307868601986255e-06, "loss": 0.2057, "step": 218 }, { "epoch": 0.04, "learning_rate": 3.3460656990068756e-06, "loss": 0.2225, "step": 219 }, { "epoch": 0.04, "learning_rate": 3.3613445378151265e-06, "loss": 0.1999, "step": 220 }, { "epoch": 0.04, "learning_rate": 3.376623376623377e-06, "loss": 0.2227, "step": 221 }, { "epoch": 0.04, "learning_rate": 3.3919022154316274e-06, "loss": 0.207, "step": 222 }, { "epoch": 0.04, "learning_rate": 3.4071810542398783e-06, "loss": 0.218, "step": 223 }, { "epoch": 0.04, "learning_rate": 3.4224598930481284e-06, "loss": 0.2064, "step": 224 }, { "epoch": 0.04, "learning_rate": 3.4377387318563792e-06, "loss": 0.2077, "step": 225 }, { "epoch": 0.04, "learning_rate": 3.4530175706646297e-06, "loss": 0.2125, "step": 226 }, { "epoch": 0.04, "learning_rate": 3.4682964094728806e-06, "loss": 0.2283, "step": 227 }, { "epoch": 0.04, "learning_rate": 3.4835752482811306e-06, "loss": 0.2188, "step": 228 }, { "epoch": 0.04, "learning_rate": 3.4988540870893815e-06, "loss": 0.2165, "step": 229 }, { "epoch": 0.04, "learning_rate": 3.514132925897632e-06, "loss": 0.2194, "step": 230 }, { "epoch": 0.04, "learning_rate": 3.529411764705883e-06, "loss": 0.2018, "step": 231 }, { "epoch": 0.04, "learning_rate": 3.5446906035141334e-06, "loss": 0.2091, "step": 232 }, { "epoch": 0.04, "learning_rate": 3.5599694423223842e-06, "loss": 0.2033, "step": 233 }, { "epoch": 0.04, "learning_rate": 3.5752482811306343e-06, "loss": 0.1961, "step": 234 }, { "epoch": 0.04, "learning_rate": 3.5905271199388848e-06, "loss": 0.1967, "step": 235 }, { "epoch": 0.04, "learning_rate": 3.6058059587471356e-06, "loss": 0.2104, "step": 236 }, { "epoch": 0.04, "learning_rate": 3.6210847975553857e-06, "loss": 0.2088, "step": 237 }, { "epoch": 0.04, "learning_rate": 3.6363636363636366e-06, "loss": 0.2072, "step": 238 }, { "epoch": 0.04, "learning_rate": 3.651642475171887e-06, "loss": 0.206, "step": 239 }, { "epoch": 0.04, "learning_rate": 3.666921313980138e-06, "loss": 0.217, "step": 240 }, { "epoch": 0.04, "learning_rate": 3.6822001527883884e-06, "loss": 0.2047, "step": 241 }, { "epoch": 0.04, "learning_rate": 3.6974789915966393e-06, "loss": 0.2206, "step": 242 }, { "epoch": 0.04, "learning_rate": 3.7127578304048893e-06, "loss": 0.1968, "step": 243 }, { "epoch": 0.04, "learning_rate": 3.7280366692131402e-06, "loss": 0.2052, "step": 244 }, { "epoch": 0.04, "learning_rate": 3.7433155080213907e-06, "loss": 0.2111, "step": 245 }, { "epoch": 0.05, "learning_rate": 3.7585943468296416e-06, "loss": 0.2123, "step": 246 }, { "epoch": 0.05, "learning_rate": 3.7738731856378916e-06, "loss": 0.2005, "step": 247 }, { "epoch": 0.05, "learning_rate": 3.789152024446142e-06, "loss": 0.2003, "step": 248 }, { "epoch": 0.05, "learning_rate": 3.804430863254393e-06, "loss": 0.201, "step": 249 }, { "epoch": 0.05, "learning_rate": 3.819709702062643e-06, "loss": 0.1975, "step": 250 }, { "epoch": 0.05, "learning_rate": 3.834988540870894e-06, "loss": 0.2003, "step": 251 }, { "epoch": 0.05, "learning_rate": 3.850267379679145e-06, "loss": 0.2032, "step": 252 }, { "epoch": 0.05, "learning_rate": 3.865546218487396e-06, "loss": 0.2188, "step": 253 }, { "epoch": 0.05, "learning_rate": 3.880825057295646e-06, "loss": 0.203, "step": 254 }, { "epoch": 0.05, "learning_rate": 3.896103896103897e-06, "loss": 0.1873, "step": 255 }, { "epoch": 0.05, "learning_rate": 3.911382734912147e-06, "loss": 0.1825, "step": 256 }, { "epoch": 0.05, "learning_rate": 3.926661573720398e-06, "loss": 0.2073, "step": 257 }, { "epoch": 0.05, "learning_rate": 3.9419404125286485e-06, "loss": 0.1892, "step": 258 }, { "epoch": 0.05, "learning_rate": 3.957219251336899e-06, "loss": 0.2137, "step": 259 }, { "epoch": 0.05, "learning_rate": 3.972498090145149e-06, "loss": 0.2022, "step": 260 }, { "epoch": 0.05, "learning_rate": 3.9877769289533994e-06, "loss": 0.2049, "step": 261 }, { "epoch": 0.05, "learning_rate": 4.00305576776165e-06, "loss": 0.2016, "step": 262 }, { "epoch": 0.05, "learning_rate": 4.018334606569901e-06, "loss": 0.2067, "step": 263 }, { "epoch": 0.05, "learning_rate": 4.033613445378151e-06, "loss": 0.1869, "step": 264 }, { "epoch": 0.05, "learning_rate": 4.048892284186402e-06, "loss": 0.1936, "step": 265 }, { "epoch": 0.05, "learning_rate": 4.064171122994653e-06, "loss": 0.182, "step": 266 }, { "epoch": 0.05, "learning_rate": 4.079449961802903e-06, "loss": 0.2005, "step": 267 }, { "epoch": 0.05, "learning_rate": 4.094728800611154e-06, "loss": 0.1835, "step": 268 }, { "epoch": 0.05, "learning_rate": 4.110007639419404e-06, "loss": 0.2017, "step": 269 }, { "epoch": 0.05, "learning_rate": 4.125286478227655e-06, "loss": 0.1773, "step": 270 }, { "epoch": 0.05, "learning_rate": 4.140565317035906e-06, "loss": 0.1928, "step": 271 }, { "epoch": 0.05, "learning_rate": 4.155844155844157e-06, "loss": 0.1847, "step": 272 }, { "epoch": 0.05, "learning_rate": 4.171122994652407e-06, "loss": 0.1976, "step": 273 }, { "epoch": 0.05, "learning_rate": 4.186401833460657e-06, "loss": 0.1939, "step": 274 }, { "epoch": 0.05, "learning_rate": 4.201680672268908e-06, "loss": 0.1818, "step": 275 }, { "epoch": 0.05, "learning_rate": 4.2169595110771586e-06, "loss": 0.1851, "step": 276 }, { "epoch": 0.05, "learning_rate": 4.2322383498854095e-06, "loss": 0.1973, "step": 277 }, { "epoch": 0.05, "learning_rate": 4.2475171886936595e-06, "loss": 0.1879, "step": 278 }, { "epoch": 0.05, "learning_rate": 4.26279602750191e-06, "loss": 0.1794, "step": 279 }, { "epoch": 0.05, "learning_rate": 4.2780748663101604e-06, "loss": 0.1929, "step": 280 }, { "epoch": 0.05, "learning_rate": 4.293353705118411e-06, "loss": 0.1845, "step": 281 }, { "epoch": 0.05, "learning_rate": 4.308632543926662e-06, "loss": 0.2069, "step": 282 }, { "epoch": 0.05, "learning_rate": 4.323911382734913e-06, "loss": 0.1973, "step": 283 }, { "epoch": 0.05, "learning_rate": 4.339190221543163e-06, "loss": 0.1765, "step": 284 }, { "epoch": 0.05, "learning_rate": 4.354469060351414e-06, "loss": 0.1993, "step": 285 }, { "epoch": 0.05, "learning_rate": 4.369747899159664e-06, "loss": 0.1982, "step": 286 }, { "epoch": 0.05, "learning_rate": 4.385026737967915e-06, "loss": 0.1956, "step": 287 }, { "epoch": 0.05, "learning_rate": 4.400305576776165e-06, "loss": 0.1955, "step": 288 }, { "epoch": 0.05, "learning_rate": 4.415584415584416e-06, "loss": 0.1825, "step": 289 }, { "epoch": 0.05, "learning_rate": 4.430863254392667e-06, "loss": 0.1959, "step": 290 }, { "epoch": 0.05, "learning_rate": 4.446142093200917e-06, "loss": 0.1996, "step": 291 }, { "epoch": 0.05, "learning_rate": 4.461420932009168e-06, "loss": 0.1925, "step": 292 }, { "epoch": 0.05, "learning_rate": 4.476699770817418e-06, "loss": 0.1865, "step": 293 }, { "epoch": 0.05, "learning_rate": 4.491978609625669e-06, "loss": 0.1949, "step": 294 }, { "epoch": 0.05, "learning_rate": 4.5072574484339196e-06, "loss": 0.1895, "step": 295 }, { "epoch": 0.05, "learning_rate": 4.5225362872421705e-06, "loss": 0.2005, "step": 296 }, { "epoch": 0.05, "learning_rate": 4.5378151260504205e-06, "loss": 0.193, "step": 297 }, { "epoch": 0.05, "learning_rate": 4.553093964858671e-06, "loss": 0.2009, "step": 298 }, { "epoch": 0.05, "learning_rate": 4.5683728036669214e-06, "loss": 0.1783, "step": 299 }, { "epoch": 0.06, "learning_rate": 4.583651642475172e-06, "loss": 0.2059, "step": 300 }, { "epoch": 0.06, "learning_rate": 4.598930481283423e-06, "loss": 0.206, "step": 301 }, { "epoch": 0.06, "learning_rate": 4.614209320091673e-06, "loss": 0.2031, "step": 302 }, { "epoch": 0.06, "learning_rate": 4.629488158899924e-06, "loss": 0.1959, "step": 303 }, { "epoch": 0.06, "learning_rate": 4.644766997708174e-06, "loss": 0.1855, "step": 304 }, { "epoch": 0.06, "learning_rate": 4.660045836516425e-06, "loss": 0.1953, "step": 305 }, { "epoch": 0.06, "learning_rate": 4.675324675324676e-06, "loss": 0.1957, "step": 306 }, { "epoch": 0.06, "learning_rate": 4.690603514132927e-06, "loss": 0.1939, "step": 307 }, { "epoch": 0.06, "learning_rate": 4.705882352941177e-06, "loss": 0.1958, "step": 308 }, { "epoch": 0.06, "learning_rate": 4.721161191749428e-06, "loss": 0.1981, "step": 309 }, { "epoch": 0.06, "learning_rate": 4.736440030557678e-06, "loss": 0.1955, "step": 310 }, { "epoch": 0.06, "learning_rate": 4.751718869365929e-06, "loss": 0.1867, "step": 311 }, { "epoch": 0.06, "learning_rate": 4.766997708174179e-06, "loss": 0.171, "step": 312 }, { "epoch": 0.06, "learning_rate": 4.78227654698243e-06, "loss": 0.1865, "step": 313 }, { "epoch": 0.06, "learning_rate": 4.7975553857906806e-06, "loss": 0.19, "step": 314 }, { "epoch": 0.06, "learning_rate": 4.812834224598931e-06, "loss": 0.1995, "step": 315 }, { "epoch": 0.06, "learning_rate": 4.8281130634071815e-06, "loss": 0.1916, "step": 316 }, { "epoch": 0.06, "learning_rate": 4.8433919022154315e-06, "loss": 0.1997, "step": 317 }, { "epoch": 0.06, "learning_rate": 4.8586707410236824e-06, "loss": 0.1888, "step": 318 }, { "epoch": 0.06, "learning_rate": 4.873949579831933e-06, "loss": 0.1951, "step": 319 }, { "epoch": 0.06, "learning_rate": 4.889228418640184e-06, "loss": 0.1943, "step": 320 }, { "epoch": 0.06, "learning_rate": 4.904507257448434e-06, "loss": 0.1812, "step": 321 }, { "epoch": 0.06, "learning_rate": 4.919786096256685e-06, "loss": 0.1863, "step": 322 }, { "epoch": 0.06, "learning_rate": 4.935064935064935e-06, "loss": 0.1802, "step": 323 }, { "epoch": 0.06, "learning_rate": 4.950343773873186e-06, "loss": 0.1913, "step": 324 }, { "epoch": 0.06, "learning_rate": 4.965622612681437e-06, "loss": 0.1854, "step": 325 }, { "epoch": 0.06, "learning_rate": 4.980901451489687e-06, "loss": 0.1935, "step": 326 }, { "epoch": 0.06, "learning_rate": 4.996180290297938e-06, "loss": 0.1924, "step": 327 }, { "epoch": 0.06, "learning_rate": 5.011459129106189e-06, "loss": 0.1845, "step": 328 }, { "epoch": 0.06, "learning_rate": 5.026737967914439e-06, "loss": 0.1966, "step": 329 }, { "epoch": 0.06, "learning_rate": 5.04201680672269e-06, "loss": 0.1792, "step": 330 }, { "epoch": 0.06, "learning_rate": 5.05729564553094e-06, "loss": 0.1853, "step": 331 }, { "epoch": 0.06, "learning_rate": 5.0725744843391915e-06, "loss": 0.1803, "step": 332 }, { "epoch": 0.06, "learning_rate": 5.0878533231474415e-06, "loss": 0.1802, "step": 333 }, { "epoch": 0.06, "learning_rate": 5.103132161955692e-06, "loss": 0.1837, "step": 334 }, { "epoch": 0.06, "learning_rate": 5.118411000763942e-06, "loss": 0.1762, "step": 335 }, { "epoch": 0.06, "learning_rate": 5.133689839572193e-06, "loss": 0.1805, "step": 336 }, { "epoch": 0.06, "learning_rate": 5.148968678380443e-06, "loss": 0.1872, "step": 337 }, { "epoch": 0.06, "learning_rate": 5.164247517188694e-06, "loss": 0.1934, "step": 338 }, { "epoch": 0.06, "learning_rate": 5.179526355996944e-06, "loss": 0.1775, "step": 339 }, { "epoch": 0.06, "learning_rate": 5.194805194805194e-06, "loss": 0.1882, "step": 340 }, { "epoch": 0.06, "learning_rate": 5.210084033613446e-06, "loss": 0.1807, "step": 341 }, { "epoch": 0.06, "learning_rate": 5.225362872421696e-06, "loss": 0.1925, "step": 342 }, { "epoch": 0.06, "learning_rate": 5.240641711229947e-06, "loss": 0.1858, "step": 343 }, { "epoch": 0.06, "learning_rate": 5.255920550038197e-06, "loss": 0.1716, "step": 344 }, { "epoch": 0.06, "learning_rate": 5.271199388846449e-06, "loss": 0.1779, "step": 345 }, { "epoch": 0.06, "learning_rate": 5.286478227654699e-06, "loss": 0.1705, "step": 346 }, { "epoch": 0.06, "learning_rate": 5.301757066462949e-06, "loss": 0.1777, "step": 347 }, { "epoch": 0.06, "learning_rate": 5.3170359052712e-06, "loss": 0.1923, "step": 348 }, { "epoch": 0.06, "learning_rate": 5.332314744079451e-06, "loss": 0.1831, "step": 349 }, { "epoch": 0.06, "learning_rate": 5.347593582887702e-06, "loss": 0.1667, "step": 350 }, { "epoch": 0.06, "learning_rate": 5.362872421695952e-06, "loss": 0.1804, "step": 351 }, { "epoch": 0.06, "learning_rate": 5.378151260504202e-06, "loss": 0.1891, "step": 352 }, { "epoch": 0.06, "learning_rate": 5.393430099312453e-06, "loss": 0.1745, "step": 353 }, { "epoch": 0.06, "learning_rate": 5.4087089381207035e-06, "loss": 0.1579, "step": 354 }, { "epoch": 0.07, "learning_rate": 5.423987776928954e-06, "loss": 0.1627, "step": 355 }, { "epoch": 0.07, "learning_rate": 5.439266615737204e-06, "loss": 0.1708, "step": 356 }, { "epoch": 0.07, "learning_rate": 5.4545454545454545e-06, "loss": 0.1876, "step": 357 }, { "epoch": 0.07, "learning_rate": 5.469824293353706e-06, "loss": 0.1843, "step": 358 }, { "epoch": 0.07, "learning_rate": 5.485103132161956e-06, "loss": 0.1782, "step": 359 }, { "epoch": 0.07, "learning_rate": 5.500381970970206e-06, "loss": 0.1578, "step": 360 }, { "epoch": 0.07, "learning_rate": 5.515660809778457e-06, "loss": 0.1892, "step": 361 }, { "epoch": 0.07, "learning_rate": 5.530939648586708e-06, "loss": 0.1731, "step": 362 }, { "epoch": 0.07, "learning_rate": 5.546218487394959e-06, "loss": 0.1715, "step": 363 }, { "epoch": 0.07, "learning_rate": 5.561497326203209e-06, "loss": 0.185, "step": 364 }, { "epoch": 0.07, "learning_rate": 5.576776165011459e-06, "loss": 0.1711, "step": 365 }, { "epoch": 0.07, "learning_rate": 5.59205500381971e-06, "loss": 0.17, "step": 366 }, { "epoch": 0.07, "learning_rate": 5.607333842627961e-06, "loss": 0.177, "step": 367 }, { "epoch": 0.07, "learning_rate": 5.622612681436212e-06, "loss": 0.1796, "step": 368 }, { "epoch": 0.07, "learning_rate": 5.637891520244462e-06, "loss": 0.1617, "step": 369 }, { "epoch": 0.07, "learning_rate": 5.653170359052712e-06, "loss": 0.1577, "step": 370 }, { "epoch": 0.07, "learning_rate": 5.6684491978609635e-06, "loss": 0.1777, "step": 371 }, { "epoch": 0.07, "learning_rate": 5.683728036669214e-06, "loss": 0.1767, "step": 372 }, { "epoch": 0.07, "learning_rate": 5.6990068754774645e-06, "loss": 0.1731, "step": 373 }, { "epoch": 0.07, "learning_rate": 5.7142857142857145e-06, "loss": 0.1891, "step": 374 }, { "epoch": 0.07, "learning_rate": 5.729564553093966e-06, "loss": 0.188, "step": 375 }, { "epoch": 0.07, "learning_rate": 5.744843391902216e-06, "loss": 0.1728, "step": 376 }, { "epoch": 0.07, "learning_rate": 5.760122230710466e-06, "loss": 0.1641, "step": 377 }, { "epoch": 0.07, "learning_rate": 5.775401069518717e-06, "loss": 0.1647, "step": 378 }, { "epoch": 0.07, "learning_rate": 5.790679908326967e-06, "loss": 0.1711, "step": 379 }, { "epoch": 0.07, "learning_rate": 5.805958747135218e-06, "loss": 0.1817, "step": 380 }, { "epoch": 0.07, "learning_rate": 5.821237585943469e-06, "loss": 0.1831, "step": 381 }, { "epoch": 0.07, "learning_rate": 5.836516424751719e-06, "loss": 0.1702, "step": 382 }, { "epoch": 0.07, "learning_rate": 5.851795263559969e-06, "loss": 0.1828, "step": 383 }, { "epoch": 0.07, "learning_rate": 5.867074102368221e-06, "loss": 0.1636, "step": 384 }, { "epoch": 0.07, "learning_rate": 5.882352941176471e-06, "loss": 0.1798, "step": 385 }, { "epoch": 0.07, "learning_rate": 5.897631779984722e-06, "loss": 0.1836, "step": 386 }, { "epoch": 0.07, "learning_rate": 5.912910618792972e-06, "loss": 0.1589, "step": 387 }, { "epoch": 0.07, "learning_rate": 5.928189457601224e-06, "loss": 0.1616, "step": 388 }, { "epoch": 0.07, "learning_rate": 5.943468296409474e-06, "loss": 0.1721, "step": 389 }, { "epoch": 0.07, "learning_rate": 5.958747135217724e-06, "loss": 0.1788, "step": 390 }, { "epoch": 0.07, "learning_rate": 5.9740259740259746e-06, "loss": 0.1582, "step": 391 }, { "epoch": 0.07, "learning_rate": 5.989304812834225e-06, "loss": 0.1747, "step": 392 }, { "epoch": 0.07, "learning_rate": 6.004583651642476e-06, "loss": 0.1719, "step": 393 }, { "epoch": 0.07, "learning_rate": 6.019862490450726e-06, "loss": 0.17, "step": 394 }, { "epoch": 0.07, "learning_rate": 6.0351413292589764e-06, "loss": 0.1645, "step": 395 }, { "epoch": 0.07, "learning_rate": 6.050420168067227e-06, "loss": 0.1812, "step": 396 }, { "epoch": 0.07, "learning_rate": 6.065699006875478e-06, "loss": 0.1674, "step": 397 }, { "epoch": 0.07, "learning_rate": 6.080977845683729e-06, "loss": 0.1638, "step": 398 }, { "epoch": 0.07, "learning_rate": 6.096256684491979e-06, "loss": 0.1737, "step": 399 }, { "epoch": 0.07, "learning_rate": 6.111535523300229e-06, "loss": 0.171, "step": 400 }, { "epoch": 0.07, "learning_rate": 6.126814362108481e-06, "loss": 0.1697, "step": 401 }, { "epoch": 0.07, "learning_rate": 6.142093200916731e-06, "loss": 0.1753, "step": 402 }, { "epoch": 0.07, "learning_rate": 6.157372039724981e-06, "loss": 0.1731, "step": 403 }, { "epoch": 0.07, "learning_rate": 6.172650878533232e-06, "loss": 0.1713, "step": 404 }, { "epoch": 0.07, "learning_rate": 6.187929717341482e-06, "loss": 0.1722, "step": 405 }, { "epoch": 0.07, "learning_rate": 6.203208556149734e-06, "loss": 0.1646, "step": 406 }, { "epoch": 0.07, "learning_rate": 6.218487394957984e-06, "loss": 0.175, "step": 407 }, { "epoch": 0.07, "learning_rate": 6.233766233766234e-06, "loss": 0.1837, "step": 408 }, { "epoch": 0.08, "learning_rate": 6.249045072574485e-06, "loss": 0.1668, "step": 409 }, { "epoch": 0.08, "learning_rate": 6.2643239113827356e-06, "loss": 0.1779, "step": 410 }, { "epoch": 0.08, "learning_rate": 6.2796027501909865e-06, "loss": 0.1713, "step": 411 }, { "epoch": 0.08, "learning_rate": 6.2948815889992365e-06, "loss": 0.1627, "step": 412 }, { "epoch": 0.08, "learning_rate": 6.3101604278074865e-06, "loss": 0.169, "step": 413 }, { "epoch": 0.08, "learning_rate": 6.325439266615738e-06, "loss": 0.1813, "step": 414 }, { "epoch": 0.08, "learning_rate": 6.340718105423988e-06, "loss": 0.1918, "step": 415 }, { "epoch": 0.08, "learning_rate": 6.355996944232239e-06, "loss": 0.1812, "step": 416 }, { "epoch": 0.08, "learning_rate": 6.371275783040489e-06, "loss": 0.162, "step": 417 }, { "epoch": 0.08, "learning_rate": 6.386554621848739e-06, "loss": 0.1659, "step": 418 }, { "epoch": 0.08, "learning_rate": 6.401833460656991e-06, "loss": 0.1607, "step": 419 }, { "epoch": 0.08, "learning_rate": 6.417112299465241e-06, "loss": 0.1739, "step": 420 }, { "epoch": 0.08, "learning_rate": 6.432391138273492e-06, "loss": 0.1634, "step": 421 }, { "epoch": 0.08, "learning_rate": 6.447669977081742e-06, "loss": 0.176, "step": 422 }, { "epoch": 0.08, "learning_rate": 6.462948815889994e-06, "loss": 0.1786, "step": 423 }, { "epoch": 0.08, "learning_rate": 6.478227654698244e-06, "loss": 0.1806, "step": 424 }, { "epoch": 0.08, "learning_rate": 6.493506493506494e-06, "loss": 0.1545, "step": 425 }, { "epoch": 0.08, "learning_rate": 6.508785332314744e-06, "loss": 0.1673, "step": 426 }, { "epoch": 0.08, "learning_rate": 6.524064171122996e-06, "loss": 0.1533, "step": 427 }, { "epoch": 0.08, "learning_rate": 6.539343009931246e-06, "loss": 0.1734, "step": 428 }, { "epoch": 0.08, "learning_rate": 6.5546218487394966e-06, "loss": 0.1586, "step": 429 }, { "epoch": 0.08, "learning_rate": 6.569900687547747e-06, "loss": 0.1636, "step": 430 }, { "epoch": 0.08, "learning_rate": 6.585179526355997e-06, "loss": 0.1677, "step": 431 }, { "epoch": 0.08, "learning_rate": 6.600458365164248e-06, "loss": 0.1587, "step": 432 }, { "epoch": 0.08, "learning_rate": 6.6157372039724984e-06, "loss": 0.1721, "step": 433 }, { "epoch": 0.08, "learning_rate": 6.631016042780749e-06, "loss": 0.1708, "step": 434 }, { "epoch": 0.08, "learning_rate": 6.646294881588999e-06, "loss": 0.1762, "step": 435 }, { "epoch": 0.08, "learning_rate": 6.661573720397251e-06, "loss": 0.171, "step": 436 }, { "epoch": 0.08, "learning_rate": 6.676852559205501e-06, "loss": 0.1695, "step": 437 }, { "epoch": 0.08, "learning_rate": 6.692131398013751e-06, "loss": 0.172, "step": 438 }, { "epoch": 0.08, "learning_rate": 6.707410236822002e-06, "loss": 0.1682, "step": 439 }, { "epoch": 0.08, "learning_rate": 6.722689075630253e-06, "loss": 0.1569, "step": 440 }, { "epoch": 0.08, "learning_rate": 6.737967914438504e-06, "loss": 0.1656, "step": 441 }, { "epoch": 0.08, "learning_rate": 6.753246753246754e-06, "loss": 0.1722, "step": 442 }, { "epoch": 0.08, "learning_rate": 6.768525592055004e-06, "loss": 0.1825, "step": 443 }, { "epoch": 0.08, "learning_rate": 6.783804430863255e-06, "loss": 0.1693, "step": 444 }, { "epoch": 0.08, "learning_rate": 6.799083269671506e-06, "loss": 0.16, "step": 445 }, { "epoch": 0.08, "learning_rate": 6.814362108479757e-06, "loss": 0.1682, "step": 446 }, { "epoch": 0.08, "learning_rate": 6.829640947288007e-06, "loss": 0.1628, "step": 447 }, { "epoch": 0.08, "learning_rate": 6.844919786096257e-06, "loss": 0.1698, "step": 448 }, { "epoch": 0.08, "learning_rate": 6.8601986249045084e-06, "loss": 0.1621, "step": 449 }, { "epoch": 0.08, "learning_rate": 6.8754774637127585e-06, "loss": 0.1794, "step": 450 }, { "epoch": 0.08, "learning_rate": 6.8907563025210085e-06, "loss": 0.1712, "step": 451 }, { "epoch": 0.08, "learning_rate": 6.906035141329259e-06, "loss": 0.154, "step": 452 }, { "epoch": 0.08, "learning_rate": 6.92131398013751e-06, "loss": 0.1657, "step": 453 }, { "epoch": 0.08, "learning_rate": 6.936592818945761e-06, "loss": 0.1618, "step": 454 }, { "epoch": 0.08, "learning_rate": 6.951871657754011e-06, "loss": 0.1507, "step": 455 }, { "epoch": 0.08, "learning_rate": 6.967150496562261e-06, "loss": 0.1699, "step": 456 }, { "epoch": 0.08, "learning_rate": 6.982429335370512e-06, "loss": 0.1657, "step": 457 }, { "epoch": 0.08, "learning_rate": 6.997708174178763e-06, "loss": 0.1583, "step": 458 }, { "epoch": 0.08, "learning_rate": 7.012987012987014e-06, "loss": 0.17, "step": 459 }, { "epoch": 0.08, "learning_rate": 7.028265851795264e-06, "loss": 0.1704, "step": 460 }, { "epoch": 0.08, "learning_rate": 7.043544690603514e-06, "loss": 0.1553, "step": 461 }, { "epoch": 0.08, "learning_rate": 7.058823529411766e-06, "loss": 0.1492, "step": 462 }, { "epoch": 0.08, "learning_rate": 7.074102368220016e-06, "loss": 0.1652, "step": 463 }, { "epoch": 0.09, "learning_rate": 7.089381207028267e-06, "loss": 0.1541, "step": 464 }, { "epoch": 0.09, "learning_rate": 7.104660045836517e-06, "loss": 0.1535, "step": 465 }, { "epoch": 0.09, "learning_rate": 7.1199388846447685e-06, "loss": 0.1764, "step": 466 }, { "epoch": 0.09, "learning_rate": 7.1352177234530185e-06, "loss": 0.162, "step": 467 }, { "epoch": 0.09, "learning_rate": 7.150496562261269e-06, "loss": 0.1629, "step": 468 }, { "epoch": 0.09, "learning_rate": 7.1657754010695195e-06, "loss": 0.1665, "step": 469 }, { "epoch": 0.09, "learning_rate": 7.1810542398777695e-06, "loss": 0.1688, "step": 470 }, { "epoch": 0.09, "learning_rate": 7.19633307868602e-06, "loss": 0.1658, "step": 471 }, { "epoch": 0.09, "learning_rate": 7.211611917494271e-06, "loss": 0.1662, "step": 472 }, { "epoch": 0.09, "learning_rate": 7.226890756302521e-06, "loss": 0.16, "step": 473 }, { "epoch": 0.09, "learning_rate": 7.242169595110771e-06, "loss": 0.1568, "step": 474 }, { "epoch": 0.09, "learning_rate": 7.257448433919023e-06, "loss": 0.1644, "step": 475 }, { "epoch": 0.09, "learning_rate": 7.272727272727273e-06, "loss": 0.163, "step": 476 }, { "epoch": 0.09, "learning_rate": 7.288006111535524e-06, "loss": 0.1551, "step": 477 }, { "epoch": 0.09, "learning_rate": 7.303284950343774e-06, "loss": 0.1493, "step": 478 }, { "epoch": 0.09, "learning_rate": 7.318563789152026e-06, "loss": 0.1623, "step": 479 }, { "epoch": 0.09, "learning_rate": 7.333842627960276e-06, "loss": 0.1624, "step": 480 }, { "epoch": 0.09, "learning_rate": 7.349121466768526e-06, "loss": 0.1528, "step": 481 }, { "epoch": 0.09, "learning_rate": 7.364400305576777e-06, "loss": 0.166, "step": 482 }, { "epoch": 0.09, "learning_rate": 7.379679144385027e-06, "loss": 0.1598, "step": 483 }, { "epoch": 0.09, "learning_rate": 7.394957983193279e-06, "loss": 0.1499, "step": 484 }, { "epoch": 0.09, "learning_rate": 7.410236822001529e-06, "loss": 0.166, "step": 485 }, { "epoch": 0.09, "learning_rate": 7.425515660809779e-06, "loss": 0.1524, "step": 486 }, { "epoch": 0.09, "learning_rate": 7.44079449961803e-06, "loss": 0.1483, "step": 487 }, { "epoch": 0.09, "learning_rate": 7.4560733384262805e-06, "loss": 0.1712, "step": 488 }, { "epoch": 0.09, "learning_rate": 7.471352177234531e-06, "loss": 0.1683, "step": 489 }, { "epoch": 0.09, "learning_rate": 7.486631016042781e-06, "loss": 0.1499, "step": 490 }, { "epoch": 0.09, "learning_rate": 7.5019098548510314e-06, "loss": 0.1613, "step": 491 }, { "epoch": 0.09, "learning_rate": 7.517188693659283e-06, "loss": 0.1471, "step": 492 }, { "epoch": 0.09, "learning_rate": 7.532467532467533e-06, "loss": 0.176, "step": 493 }, { "epoch": 0.09, "learning_rate": 7.547746371275783e-06, "loss": 0.168, "step": 494 }, { "epoch": 0.09, "learning_rate": 7.563025210084034e-06, "loss": 0.1657, "step": 495 }, { "epoch": 0.09, "learning_rate": 7.578304048892284e-06, "loss": 0.1686, "step": 496 }, { "epoch": 0.09, "learning_rate": 7.593582887700536e-06, "loss": 0.1702, "step": 497 }, { "epoch": 0.09, "learning_rate": 7.608861726508786e-06, "loss": 0.1643, "step": 498 }, { "epoch": 0.09, "learning_rate": 7.624140565317036e-06, "loss": 0.1649, "step": 499 }, { "epoch": 0.09, "learning_rate": 7.639419404125286e-06, "loss": 0.1551, "step": 500 }, { "epoch": 0.09, "learning_rate": 7.654698242933538e-06, "loss": 0.1585, "step": 501 }, { "epoch": 0.09, "learning_rate": 7.669977081741788e-06, "loss": 0.1656, "step": 502 }, { "epoch": 0.09, "learning_rate": 7.68525592055004e-06, "loss": 0.1597, "step": 503 }, { "epoch": 0.09, "learning_rate": 7.70053475935829e-06, "loss": 0.1543, "step": 504 }, { "epoch": 0.09, "learning_rate": 7.71581359816654e-06, "loss": 0.1516, "step": 505 }, { "epoch": 0.09, "learning_rate": 7.731092436974791e-06, "loss": 0.1675, "step": 506 }, { "epoch": 0.09, "learning_rate": 7.746371275783041e-06, "loss": 0.1505, "step": 507 }, { "epoch": 0.09, "learning_rate": 7.761650114591292e-06, "loss": 0.1621, "step": 508 }, { "epoch": 0.09, "learning_rate": 7.776928953399542e-06, "loss": 0.1543, "step": 509 }, { "epoch": 0.09, "learning_rate": 7.792207792207793e-06, "loss": 0.16, "step": 510 }, { "epoch": 0.09, "learning_rate": 7.807486631016043e-06, "loss": 0.1612, "step": 511 }, { "epoch": 0.09, "learning_rate": 7.822765469824293e-06, "loss": 0.1595, "step": 512 }, { "epoch": 0.09, "learning_rate": 7.838044308632543e-06, "loss": 0.1507, "step": 513 }, { "epoch": 0.09, "learning_rate": 7.853323147440795e-06, "loss": 0.1608, "step": 514 }, { "epoch": 0.09, "learning_rate": 7.868601986249045e-06, "loss": 0.1665, "step": 515 }, { "epoch": 0.09, "learning_rate": 7.883880825057297e-06, "loss": 0.1701, "step": 516 }, { "epoch": 0.09, "learning_rate": 7.899159663865547e-06, "loss": 0.1528, "step": 517 }, { "epoch": 0.09, "learning_rate": 7.914438502673799e-06, "loss": 0.1674, "step": 518 }, { "epoch": 0.1, "learning_rate": 7.929717341482049e-06, "loss": 0.1567, "step": 519 }, { "epoch": 0.1, "learning_rate": 7.944996180290299e-06, "loss": 0.1621, "step": 520 }, { "epoch": 0.1, "learning_rate": 7.960275019098549e-06, "loss": 0.1565, "step": 521 }, { "epoch": 0.1, "learning_rate": 7.975553857906799e-06, "loss": 0.1412, "step": 522 }, { "epoch": 0.1, "learning_rate": 7.99083269671505e-06, "loss": 0.1529, "step": 523 }, { "epoch": 0.1, "learning_rate": 8.0061115355233e-06, "loss": 0.1494, "step": 524 }, { "epoch": 0.1, "learning_rate": 8.02139037433155e-06, "loss": 0.1645, "step": 525 }, { "epoch": 0.1, "learning_rate": 8.036669213139802e-06, "loss": 0.1617, "step": 526 }, { "epoch": 0.1, "learning_rate": 8.051948051948052e-06, "loss": 0.1599, "step": 527 }, { "epoch": 0.1, "learning_rate": 8.067226890756303e-06, "loss": 0.156, "step": 528 }, { "epoch": 0.1, "learning_rate": 8.082505729564554e-06, "loss": 0.1512, "step": 529 }, { "epoch": 0.1, "learning_rate": 8.097784568372804e-06, "loss": 0.1407, "step": 530 }, { "epoch": 0.1, "learning_rate": 8.113063407181056e-06, "loss": 0.1605, "step": 531 }, { "epoch": 0.1, "learning_rate": 8.128342245989306e-06, "loss": 0.1652, "step": 532 }, { "epoch": 0.1, "learning_rate": 8.143621084797556e-06, "loss": 0.1575, "step": 533 }, { "epoch": 0.1, "learning_rate": 8.158899923605806e-06, "loss": 0.1711, "step": 534 }, { "epoch": 0.1, "learning_rate": 8.174178762414056e-06, "loss": 0.1483, "step": 535 }, { "epoch": 0.1, "learning_rate": 8.189457601222308e-06, "loss": 0.1531, "step": 536 }, { "epoch": 0.1, "learning_rate": 8.204736440030558e-06, "loss": 0.1535, "step": 537 }, { "epoch": 0.1, "learning_rate": 8.220015278838808e-06, "loss": 0.1679, "step": 538 }, { "epoch": 0.1, "learning_rate": 8.23529411764706e-06, "loss": 0.1542, "step": 539 }, { "epoch": 0.1, "learning_rate": 8.25057295645531e-06, "loss": 0.139, "step": 540 }, { "epoch": 0.1, "learning_rate": 8.265851795263562e-06, "loss": 0.1572, "step": 541 }, { "epoch": 0.1, "learning_rate": 8.281130634071812e-06, "loss": 0.1676, "step": 542 }, { "epoch": 0.1, "learning_rate": 8.296409472880062e-06, "loss": 0.1547, "step": 543 }, { "epoch": 0.1, "learning_rate": 8.311688311688313e-06, "loss": 0.1649, "step": 544 }, { "epoch": 0.1, "learning_rate": 8.326967150496563e-06, "loss": 0.1416, "step": 545 }, { "epoch": 0.1, "learning_rate": 8.342245989304813e-06, "loss": 0.1748, "step": 546 }, { "epoch": 0.1, "learning_rate": 8.357524828113064e-06, "loss": 0.169, "step": 547 }, { "epoch": 0.1, "learning_rate": 8.372803666921314e-06, "loss": 0.1598, "step": 548 }, { "epoch": 0.1, "learning_rate": 8.388082505729565e-06, "loss": 0.16, "step": 549 }, { "epoch": 0.1, "learning_rate": 8.403361344537815e-06, "loss": 0.1553, "step": 550 }, { "epoch": 0.1, "learning_rate": 8.418640183346067e-06, "loss": 0.1512, "step": 551 }, { "epoch": 0.1, "learning_rate": 8.433919022154317e-06, "loss": 0.1474, "step": 552 }, { "epoch": 0.1, "learning_rate": 8.449197860962567e-06, "loss": 0.1628, "step": 553 }, { "epoch": 0.1, "learning_rate": 8.464476699770819e-06, "loss": 0.1598, "step": 554 }, { "epoch": 0.1, "learning_rate": 8.479755538579069e-06, "loss": 0.16, "step": 555 }, { "epoch": 0.1, "learning_rate": 8.495034377387319e-06, "loss": 0.1364, "step": 556 }, { "epoch": 0.1, "learning_rate": 8.51031321619557e-06, "loss": 0.155, "step": 557 }, { "epoch": 0.1, "learning_rate": 8.52559205500382e-06, "loss": 0.1502, "step": 558 }, { "epoch": 0.1, "learning_rate": 8.54087089381207e-06, "loss": 0.1554, "step": 559 }, { "epoch": 0.1, "learning_rate": 8.556149732620321e-06, "loss": 0.1479, "step": 560 }, { "epoch": 0.1, "learning_rate": 8.571428571428571e-06, "loss": 0.158, "step": 561 }, { "epoch": 0.1, "learning_rate": 8.586707410236823e-06, "loss": 0.1519, "step": 562 }, { "epoch": 0.1, "learning_rate": 8.601986249045073e-06, "loss": 0.1542, "step": 563 }, { "epoch": 0.1, "learning_rate": 8.617265087853324e-06, "loss": 0.1496, "step": 564 }, { "epoch": 0.1, "learning_rate": 8.632543926661574e-06, "loss": 0.1608, "step": 565 }, { "epoch": 0.1, "learning_rate": 8.647822765469826e-06, "loss": 0.1515, "step": 566 }, { "epoch": 0.1, "learning_rate": 8.663101604278076e-06, "loss": 0.1621, "step": 567 }, { "epoch": 0.1, "learning_rate": 8.678380443086326e-06, "loss": 0.1538, "step": 568 }, { "epoch": 0.1, "learning_rate": 8.693659281894576e-06, "loss": 0.1565, "step": 569 }, { "epoch": 0.1, "learning_rate": 8.708938120702828e-06, "loss": 0.1603, "step": 570 }, { "epoch": 0.1, "learning_rate": 8.724216959511078e-06, "loss": 0.1591, "step": 571 }, { "epoch": 0.1, "learning_rate": 8.739495798319328e-06, "loss": 0.1596, "step": 572 }, { "epoch": 0.11, "learning_rate": 8.754774637127578e-06, "loss": 0.1481, "step": 573 }, { "epoch": 0.11, "learning_rate": 8.77005347593583e-06, "loss": 0.1514, "step": 574 }, { "epoch": 0.11, "learning_rate": 8.78533231474408e-06, "loss": 0.1553, "step": 575 }, { "epoch": 0.11, "learning_rate": 8.80061115355233e-06, "loss": 0.1544, "step": 576 }, { "epoch": 0.11, "learning_rate": 8.815889992360582e-06, "loss": 0.1566, "step": 577 }, { "epoch": 0.11, "learning_rate": 8.831168831168832e-06, "loss": 0.1558, "step": 578 }, { "epoch": 0.11, "learning_rate": 8.846447669977084e-06, "loss": 0.157, "step": 579 }, { "epoch": 0.11, "learning_rate": 8.861726508785334e-06, "loss": 0.1505, "step": 580 }, { "epoch": 0.11, "learning_rate": 8.877005347593584e-06, "loss": 0.142, "step": 581 }, { "epoch": 0.11, "learning_rate": 8.892284186401834e-06, "loss": 0.1675, "step": 582 }, { "epoch": 0.11, "learning_rate": 8.907563025210085e-06, "loss": 0.1631, "step": 583 }, { "epoch": 0.11, "learning_rate": 8.922841864018335e-06, "loss": 0.1501, "step": 584 }, { "epoch": 0.11, "learning_rate": 8.938120702826586e-06, "loss": 0.1577, "step": 585 }, { "epoch": 0.11, "learning_rate": 8.953399541634836e-06, "loss": 0.1402, "step": 586 }, { "epoch": 0.11, "learning_rate": 8.968678380443087e-06, "loss": 0.1433, "step": 587 }, { "epoch": 0.11, "learning_rate": 8.983957219251337e-06, "loss": 0.1564, "step": 588 }, { "epoch": 0.11, "learning_rate": 8.999236058059589e-06, "loss": 0.1592, "step": 589 }, { "epoch": 0.11, "learning_rate": 9.014514896867839e-06, "loss": 0.1652, "step": 590 }, { "epoch": 0.11, "learning_rate": 9.02979373567609e-06, "loss": 0.1495, "step": 591 }, { "epoch": 0.11, "learning_rate": 9.045072574484341e-06, "loss": 0.1656, "step": 592 }, { "epoch": 0.11, "learning_rate": 9.060351413292591e-06, "loss": 0.1568, "step": 593 }, { "epoch": 0.11, "learning_rate": 9.075630252100841e-06, "loss": 0.1614, "step": 594 }, { "epoch": 0.11, "learning_rate": 9.090909090909091e-06, "loss": 0.1285, "step": 595 }, { "epoch": 0.11, "learning_rate": 9.106187929717343e-06, "loss": 0.1614, "step": 596 }, { "epoch": 0.11, "learning_rate": 9.121466768525593e-06, "loss": 0.1535, "step": 597 }, { "epoch": 0.11, "learning_rate": 9.136745607333843e-06, "loss": 0.161, "step": 598 }, { "epoch": 0.11, "learning_rate": 9.152024446142093e-06, "loss": 0.1513, "step": 599 }, { "epoch": 0.11, "learning_rate": 9.167303284950345e-06, "loss": 0.1622, "step": 600 }, { "epoch": 0.11, "learning_rate": 9.182582123758595e-06, "loss": 0.15, "step": 601 }, { "epoch": 0.11, "learning_rate": 9.197860962566846e-06, "loss": 0.1561, "step": 602 }, { "epoch": 0.11, "learning_rate": 9.213139801375096e-06, "loss": 0.1326, "step": 603 }, { "epoch": 0.11, "learning_rate": 9.228418640183347e-06, "loss": 0.1525, "step": 604 }, { "epoch": 0.11, "learning_rate": 9.243697478991598e-06, "loss": 0.1443, "step": 605 }, { "epoch": 0.11, "learning_rate": 9.258976317799848e-06, "loss": 0.1636, "step": 606 }, { "epoch": 0.11, "learning_rate": 9.274255156608098e-06, "loss": 0.146, "step": 607 }, { "epoch": 0.11, "learning_rate": 9.289533995416348e-06, "loss": 0.1685, "step": 608 }, { "epoch": 0.11, "learning_rate": 9.3048128342246e-06, "loss": 0.1699, "step": 609 }, { "epoch": 0.11, "learning_rate": 9.32009167303285e-06, "loss": 0.156, "step": 610 }, { "epoch": 0.11, "learning_rate": 9.3353705118411e-06, "loss": 0.1449, "step": 611 }, { "epoch": 0.11, "learning_rate": 9.350649350649352e-06, "loss": 0.1539, "step": 612 }, { "epoch": 0.11, "learning_rate": 9.365928189457602e-06, "loss": 0.1516, "step": 613 }, { "epoch": 0.11, "learning_rate": 9.381207028265854e-06, "loss": 0.1456, "step": 614 }, { "epoch": 0.11, "learning_rate": 9.396485867074104e-06, "loss": 0.146, "step": 615 }, { "epoch": 0.11, "learning_rate": 9.411764705882354e-06, "loss": 0.1454, "step": 616 }, { "epoch": 0.11, "learning_rate": 9.427043544690604e-06, "loss": 0.1563, "step": 617 }, { "epoch": 0.11, "learning_rate": 9.442322383498856e-06, "loss": 0.1551, "step": 618 }, { "epoch": 0.11, "learning_rate": 9.457601222307106e-06, "loss": 0.1584, "step": 619 }, { "epoch": 0.11, "learning_rate": 9.472880061115356e-06, "loss": 0.1542, "step": 620 }, { "epoch": 0.11, "learning_rate": 9.488158899923606e-06, "loss": 0.1391, "step": 621 }, { "epoch": 0.11, "learning_rate": 9.503437738731857e-06, "loss": 0.1364, "step": 622 }, { "epoch": 0.11, "learning_rate": 9.518716577540108e-06, "loss": 0.1631, "step": 623 }, { "epoch": 0.11, "learning_rate": 9.533995416348358e-06, "loss": 0.1597, "step": 624 }, { "epoch": 0.11, "learning_rate": 9.54927425515661e-06, "loss": 0.1611, "step": 625 }, { "epoch": 0.11, "learning_rate": 9.56455309396486e-06, "loss": 0.1573, "step": 626 }, { "epoch": 0.11, "learning_rate": 9.579831932773111e-06, "loss": 0.1552, "step": 627 }, { "epoch": 0.12, "learning_rate": 9.595110771581361e-06, "loss": 0.1455, "step": 628 }, { "epoch": 0.12, "learning_rate": 9.610389610389611e-06, "loss": 0.1498, "step": 629 }, { "epoch": 0.12, "learning_rate": 9.625668449197861e-06, "loss": 0.1356, "step": 630 }, { "epoch": 0.12, "learning_rate": 9.640947288006113e-06, "loss": 0.1506, "step": 631 }, { "epoch": 0.12, "learning_rate": 9.656226126814363e-06, "loss": 0.1532, "step": 632 }, { "epoch": 0.12, "learning_rate": 9.671504965622613e-06, "loss": 0.1659, "step": 633 }, { "epoch": 0.12, "learning_rate": 9.686783804430863e-06, "loss": 0.1569, "step": 634 }, { "epoch": 0.12, "learning_rate": 9.702062643239115e-06, "loss": 0.1558, "step": 635 }, { "epoch": 0.12, "learning_rate": 9.717341482047365e-06, "loss": 0.151, "step": 636 }, { "epoch": 0.12, "learning_rate": 9.732620320855617e-06, "loss": 0.152, "step": 637 }, { "epoch": 0.12, "learning_rate": 9.747899159663867e-06, "loss": 0.1613, "step": 638 }, { "epoch": 0.12, "learning_rate": 9.763177998472117e-06, "loss": 0.1459, "step": 639 }, { "epoch": 0.12, "learning_rate": 9.778456837280368e-06, "loss": 0.1552, "step": 640 }, { "epoch": 0.12, "learning_rate": 9.793735676088618e-06, "loss": 0.1421, "step": 641 }, { "epoch": 0.12, "learning_rate": 9.809014514896868e-06, "loss": 0.1493, "step": 642 }, { "epoch": 0.12, "learning_rate": 9.824293353705119e-06, "loss": 0.1563, "step": 643 }, { "epoch": 0.12, "learning_rate": 9.83957219251337e-06, "loss": 0.1361, "step": 644 }, { "epoch": 0.12, "learning_rate": 9.85485103132162e-06, "loss": 0.152, "step": 645 }, { "epoch": 0.12, "learning_rate": 9.87012987012987e-06, "loss": 0.1468, "step": 646 }, { "epoch": 0.12, "learning_rate": 9.88540870893812e-06, "loss": 0.1392, "step": 647 }, { "epoch": 0.12, "learning_rate": 9.900687547746372e-06, "loss": 0.1639, "step": 648 }, { "epoch": 0.12, "learning_rate": 9.915966386554622e-06, "loss": 0.1606, "step": 649 }, { "epoch": 0.12, "learning_rate": 9.931245225362874e-06, "loss": 0.1505, "step": 650 }, { "epoch": 0.12, "learning_rate": 9.946524064171124e-06, "loss": 0.1473, "step": 651 }, { "epoch": 0.12, "learning_rate": 9.961802902979374e-06, "loss": 0.164, "step": 652 }, { "epoch": 0.12, "learning_rate": 9.977081741787626e-06, "loss": 0.1537, "step": 653 }, { "epoch": 0.12, "learning_rate": 9.992360580595876e-06, "loss": 0.1527, "step": 654 }, { "epoch": 0.12, "learning_rate": 1.0007639419404128e-05, "loss": 0.1545, "step": 655 }, { "epoch": 0.12, "learning_rate": 1.0022918258212378e-05, "loss": 0.1492, "step": 656 }, { "epoch": 0.12, "learning_rate": 1.0038197097020628e-05, "loss": 0.1685, "step": 657 }, { "epoch": 0.12, "learning_rate": 1.0053475935828878e-05, "loss": 0.1634, "step": 658 }, { "epoch": 0.12, "learning_rate": 1.0068754774637128e-05, "loss": 0.1446, "step": 659 }, { "epoch": 0.12, "learning_rate": 1.008403361344538e-05, "loss": 0.1446, "step": 660 }, { "epoch": 0.12, "learning_rate": 1.009931245225363e-05, "loss": 0.1531, "step": 661 }, { "epoch": 0.12, "learning_rate": 1.011459129106188e-05, "loss": 0.1406, "step": 662 }, { "epoch": 0.12, "learning_rate": 1.012987012987013e-05, "loss": 0.1513, "step": 663 }, { "epoch": 0.12, "learning_rate": 1.0145148968678383e-05, "loss": 0.1474, "step": 664 }, { "epoch": 0.12, "learning_rate": 1.0160427807486633e-05, "loss": 0.1547, "step": 665 }, { "epoch": 0.12, "learning_rate": 1.0175706646294883e-05, "loss": 0.1654, "step": 666 }, { "epoch": 0.12, "learning_rate": 1.0190985485103133e-05, "loss": 0.1603, "step": 667 }, { "epoch": 0.12, "learning_rate": 1.0206264323911383e-05, "loss": 0.1464, "step": 668 }, { "epoch": 0.12, "learning_rate": 1.0221543162719633e-05, "loss": 0.1457, "step": 669 }, { "epoch": 0.12, "learning_rate": 1.0236822001527883e-05, "loss": 0.15, "step": 670 }, { "epoch": 0.12, "learning_rate": 1.0252100840336135e-05, "loss": 0.1448, "step": 671 }, { "epoch": 0.12, "learning_rate": 1.0267379679144387e-05, "loss": 0.1493, "step": 672 }, { "epoch": 0.12, "learning_rate": 1.0282658517952637e-05, "loss": 0.1468, "step": 673 }, { "epoch": 0.12, "learning_rate": 1.0297937356760887e-05, "loss": 0.1339, "step": 674 }, { "epoch": 0.12, "learning_rate": 1.0313216195569139e-05, "loss": 0.1488, "step": 675 }, { "epoch": 0.12, "learning_rate": 1.0328495034377389e-05, "loss": 0.1597, "step": 676 }, { "epoch": 0.12, "learning_rate": 1.0343773873185639e-05, "loss": 0.1411, "step": 677 }, { "epoch": 0.12, "learning_rate": 1.0359052711993889e-05, "loss": 0.1614, "step": 678 }, { "epoch": 0.12, "learning_rate": 1.0374331550802139e-05, "loss": 0.1571, "step": 679 }, { "epoch": 0.12, "learning_rate": 1.0389610389610389e-05, "loss": 0.1423, "step": 680 }, { "epoch": 0.12, "learning_rate": 1.0404889228418642e-05, "loss": 0.1548, "step": 681 }, { "epoch": 0.13, "learning_rate": 1.0420168067226892e-05, "loss": 0.1436, "step": 682 }, { "epoch": 0.13, "learning_rate": 1.0435446906035142e-05, "loss": 0.1428, "step": 683 }, { "epoch": 0.13, "learning_rate": 1.0450725744843392e-05, "loss": 0.1424, "step": 684 }, { "epoch": 0.13, "learning_rate": 1.0466004583651644e-05, "loss": 0.1477, "step": 685 }, { "epoch": 0.13, "learning_rate": 1.0481283422459894e-05, "loss": 0.158, "step": 686 }, { "epoch": 0.13, "learning_rate": 1.0496562261268144e-05, "loss": 0.1381, "step": 687 }, { "epoch": 0.13, "learning_rate": 1.0511841100076394e-05, "loss": 0.1583, "step": 688 }, { "epoch": 0.13, "learning_rate": 1.0527119938884644e-05, "loss": 0.1423, "step": 689 }, { "epoch": 0.13, "learning_rate": 1.0542398777692898e-05, "loss": 0.1511, "step": 690 }, { "epoch": 0.13, "learning_rate": 1.0557677616501148e-05, "loss": 0.1635, "step": 691 }, { "epoch": 0.13, "learning_rate": 1.0572956455309398e-05, "loss": 0.1488, "step": 692 }, { "epoch": 0.13, "learning_rate": 1.0588235294117648e-05, "loss": 0.1604, "step": 693 }, { "epoch": 0.13, "learning_rate": 1.0603514132925898e-05, "loss": 0.1435, "step": 694 }, { "epoch": 0.13, "learning_rate": 1.0618792971734148e-05, "loss": 0.1524, "step": 695 }, { "epoch": 0.13, "learning_rate": 1.06340718105424e-05, "loss": 0.1593, "step": 696 }, { "epoch": 0.13, "learning_rate": 1.064935064935065e-05, "loss": 0.1623, "step": 697 }, { "epoch": 0.13, "learning_rate": 1.0664629488158901e-05, "loss": 0.1602, "step": 698 }, { "epoch": 0.13, "learning_rate": 1.0679908326967151e-05, "loss": 0.1602, "step": 699 }, { "epoch": 0.13, "learning_rate": 1.0695187165775403e-05, "loss": 0.1349, "step": 700 }, { "epoch": 0.13, "learning_rate": 1.0710466004583653e-05, "loss": 0.1508, "step": 701 }, { "epoch": 0.13, "learning_rate": 1.0725744843391903e-05, "loss": 0.1625, "step": 702 }, { "epoch": 0.13, "learning_rate": 1.0741023682200153e-05, "loss": 0.1657, "step": 703 }, { "epoch": 0.13, "learning_rate": 1.0756302521008403e-05, "loss": 0.1483, "step": 704 }, { "epoch": 0.13, "learning_rate": 1.0771581359816653e-05, "loss": 0.1379, "step": 705 }, { "epoch": 0.13, "learning_rate": 1.0786860198624905e-05, "loss": 0.1479, "step": 706 }, { "epoch": 0.13, "learning_rate": 1.0802139037433157e-05, "loss": 0.1583, "step": 707 }, { "epoch": 0.13, "learning_rate": 1.0817417876241407e-05, "loss": 0.158, "step": 708 }, { "epoch": 0.13, "learning_rate": 1.0832696715049657e-05, "loss": 0.1483, "step": 709 }, { "epoch": 0.13, "learning_rate": 1.0847975553857909e-05, "loss": 0.1621, "step": 710 }, { "epoch": 0.13, "learning_rate": 1.0863254392666159e-05, "loss": 0.1382, "step": 711 }, { "epoch": 0.13, "learning_rate": 1.0878533231474409e-05, "loss": 0.1366, "step": 712 }, { "epoch": 0.13, "learning_rate": 1.0893812070282659e-05, "loss": 0.1545, "step": 713 }, { "epoch": 0.13, "learning_rate": 1.0909090909090909e-05, "loss": 0.1452, "step": 714 }, { "epoch": 0.13, "learning_rate": 1.0924369747899159e-05, "loss": 0.1469, "step": 715 }, { "epoch": 0.13, "learning_rate": 1.0939648586707412e-05, "loss": 0.1569, "step": 716 }, { "epoch": 0.13, "learning_rate": 1.0954927425515662e-05, "loss": 0.1569, "step": 717 }, { "epoch": 0.13, "learning_rate": 1.0970206264323912e-05, "loss": 0.1415, "step": 718 }, { "epoch": 0.13, "learning_rate": 1.0985485103132163e-05, "loss": 0.1579, "step": 719 }, { "epoch": 0.13, "learning_rate": 1.1000763941940413e-05, "loss": 0.1472, "step": 720 }, { "epoch": 0.13, "learning_rate": 1.1016042780748664e-05, "loss": 0.1461, "step": 721 }, { "epoch": 0.13, "learning_rate": 1.1031321619556914e-05, "loss": 0.1486, "step": 722 }, { "epoch": 0.13, "learning_rate": 1.1046600458365164e-05, "loss": 0.1447, "step": 723 }, { "epoch": 0.13, "learning_rate": 1.1061879297173416e-05, "loss": 0.1458, "step": 724 }, { "epoch": 0.13, "learning_rate": 1.1077158135981668e-05, "loss": 0.157, "step": 725 }, { "epoch": 0.13, "learning_rate": 1.1092436974789918e-05, "loss": 0.1401, "step": 726 }, { "epoch": 0.13, "learning_rate": 1.1107715813598168e-05, "loss": 0.1548, "step": 727 }, { "epoch": 0.13, "learning_rate": 1.1122994652406418e-05, "loss": 0.1443, "step": 728 }, { "epoch": 0.13, "learning_rate": 1.1138273491214668e-05, "loss": 0.142, "step": 729 }, { "epoch": 0.13, "learning_rate": 1.1153552330022918e-05, "loss": 0.1456, "step": 730 }, { "epoch": 0.13, "learning_rate": 1.116883116883117e-05, "loss": 0.1472, "step": 731 }, { "epoch": 0.13, "learning_rate": 1.118411000763942e-05, "loss": 0.1436, "step": 732 }, { "epoch": 0.13, "learning_rate": 1.1199388846447672e-05, "loss": 0.1527, "step": 733 }, { "epoch": 0.13, "learning_rate": 1.1214667685255922e-05, "loss": 0.1428, "step": 734 }, { "epoch": 0.13, "learning_rate": 1.1229946524064172e-05, "loss": 0.1405, "step": 735 }, { "epoch": 0.13, "learning_rate": 1.1245225362872423e-05, "loss": 0.1437, "step": 736 }, { "epoch": 0.14, "learning_rate": 1.1260504201680673e-05, "loss": 0.1445, "step": 737 }, { "epoch": 0.14, "learning_rate": 1.1275783040488924e-05, "loss": 0.1566, "step": 738 }, { "epoch": 0.14, "learning_rate": 1.1291061879297174e-05, "loss": 0.1416, "step": 739 }, { "epoch": 0.14, "learning_rate": 1.1306340718105424e-05, "loss": 0.1469, "step": 740 }, { "epoch": 0.14, "learning_rate": 1.1321619556913674e-05, "loss": 0.1478, "step": 741 }, { "epoch": 0.14, "learning_rate": 1.1336898395721927e-05, "loss": 0.1335, "step": 742 }, { "epoch": 0.14, "learning_rate": 1.1352177234530177e-05, "loss": 0.1511, "step": 743 }, { "epoch": 0.14, "learning_rate": 1.1367456073338427e-05, "loss": 0.149, "step": 744 }, { "epoch": 0.14, "learning_rate": 1.1382734912146677e-05, "loss": 0.1333, "step": 745 }, { "epoch": 0.14, "learning_rate": 1.1398013750954929e-05, "loss": 0.1414, "step": 746 }, { "epoch": 0.14, "learning_rate": 1.1413292589763179e-05, "loss": 0.1397, "step": 747 }, { "epoch": 0.14, "learning_rate": 1.1428571428571429e-05, "loss": 0.1491, "step": 748 }, { "epoch": 0.14, "learning_rate": 1.1443850267379679e-05, "loss": 0.1513, "step": 749 }, { "epoch": 0.14, "learning_rate": 1.1459129106187932e-05, "loss": 0.1423, "step": 750 }, { "epoch": 0.14, "learning_rate": 1.1474407944996183e-05, "loss": 0.1489, "step": 751 }, { "epoch": 0.14, "learning_rate": 1.1489686783804433e-05, "loss": 0.1529, "step": 752 }, { "epoch": 0.14, "learning_rate": 1.1504965622612683e-05, "loss": 0.1581, "step": 753 }, { "epoch": 0.14, "learning_rate": 1.1520244461420933e-05, "loss": 0.1497, "step": 754 }, { "epoch": 0.14, "learning_rate": 1.1535523300229183e-05, "loss": 0.1434, "step": 755 }, { "epoch": 0.14, "learning_rate": 1.1550802139037434e-05, "loss": 0.1466, "step": 756 }, { "epoch": 0.14, "learning_rate": 1.1566080977845684e-05, "loss": 0.1575, "step": 757 }, { "epoch": 0.14, "learning_rate": 1.1581359816653935e-05, "loss": 0.1545, "step": 758 }, { "epoch": 0.14, "learning_rate": 1.1596638655462186e-05, "loss": 0.1456, "step": 759 }, { "epoch": 0.14, "learning_rate": 1.1611917494270436e-05, "loss": 0.1648, "step": 760 }, { "epoch": 0.14, "learning_rate": 1.1627196333078688e-05, "loss": 0.1476, "step": 761 }, { "epoch": 0.14, "learning_rate": 1.1642475171886938e-05, "loss": 0.1451, "step": 762 }, { "epoch": 0.14, "learning_rate": 1.1657754010695188e-05, "loss": 0.1568, "step": 763 }, { "epoch": 0.14, "learning_rate": 1.1673032849503438e-05, "loss": 0.1518, "step": 764 }, { "epoch": 0.14, "learning_rate": 1.1688311688311688e-05, "loss": 0.1574, "step": 765 }, { "epoch": 0.14, "learning_rate": 1.1703590527119938e-05, "loss": 0.1434, "step": 766 }, { "epoch": 0.14, "learning_rate": 1.171886936592819e-05, "loss": 0.1327, "step": 767 }, { "epoch": 0.14, "learning_rate": 1.1734148204736442e-05, "loss": 0.1474, "step": 768 }, { "epoch": 0.14, "learning_rate": 1.1749427043544692e-05, "loss": 0.1478, "step": 769 }, { "epoch": 0.14, "learning_rate": 1.1764705882352942e-05, "loss": 0.1517, "step": 770 }, { "epoch": 0.14, "learning_rate": 1.1779984721161194e-05, "loss": 0.134, "step": 771 }, { "epoch": 0.14, "learning_rate": 1.1795263559969444e-05, "loss": 0.1471, "step": 772 }, { "epoch": 0.14, "learning_rate": 1.1810542398777694e-05, "loss": 0.1295, "step": 773 }, { "epoch": 0.14, "learning_rate": 1.1825821237585944e-05, "loss": 0.1366, "step": 774 }, { "epoch": 0.14, "learning_rate": 1.1841100076394194e-05, "loss": 0.1449, "step": 775 }, { "epoch": 0.14, "learning_rate": 1.1856378915202447e-05, "loss": 0.1461, "step": 776 }, { "epoch": 0.14, "learning_rate": 1.1871657754010697e-05, "loss": 0.1454, "step": 777 }, { "epoch": 0.14, "learning_rate": 1.1886936592818947e-05, "loss": 0.1557, "step": 778 }, { "epoch": 0.14, "learning_rate": 1.1902215431627197e-05, "loss": 0.129, "step": 779 }, { "epoch": 0.14, "learning_rate": 1.1917494270435447e-05, "loss": 0.1396, "step": 780 }, { "epoch": 0.14, "learning_rate": 1.1932773109243699e-05, "loss": 0.1579, "step": 781 }, { "epoch": 0.14, "learning_rate": 1.1948051948051949e-05, "loss": 0.1457, "step": 782 }, { "epoch": 0.14, "learning_rate": 1.19633307868602e-05, "loss": 0.1369, "step": 783 }, { "epoch": 0.14, "learning_rate": 1.197860962566845e-05, "loss": 0.1419, "step": 784 }, { "epoch": 0.14, "learning_rate": 1.1993888464476701e-05, "loss": 0.1418, "step": 785 }, { "epoch": 0.14, "learning_rate": 1.2009167303284953e-05, "loss": 0.148, "step": 786 }, { "epoch": 0.14, "learning_rate": 1.2024446142093203e-05, "loss": 0.1371, "step": 787 }, { "epoch": 0.14, "learning_rate": 1.2039724980901453e-05, "loss": 0.1502, "step": 788 }, { "epoch": 0.14, "learning_rate": 1.2055003819709703e-05, "loss": 0.1478, "step": 789 }, { "epoch": 0.14, "learning_rate": 1.2070282658517953e-05, "loss": 0.1452, "step": 790 }, { "epoch": 0.15, "learning_rate": 1.2085561497326203e-05, "loss": 0.1457, "step": 791 }, { "epoch": 0.15, "learning_rate": 1.2100840336134455e-05, "loss": 0.1427, "step": 792 }, { "epoch": 0.15, "learning_rate": 1.2116119174942705e-05, "loss": 0.1422, "step": 793 }, { "epoch": 0.15, "learning_rate": 1.2131398013750956e-05, "loss": 0.1395, "step": 794 }, { "epoch": 0.15, "learning_rate": 1.2146676852559206e-05, "loss": 0.155, "step": 795 }, { "epoch": 0.15, "learning_rate": 1.2161955691367458e-05, "loss": 0.1544, "step": 796 }, { "epoch": 0.15, "learning_rate": 1.2177234530175708e-05, "loss": 0.1365, "step": 797 }, { "epoch": 0.15, "learning_rate": 1.2192513368983958e-05, "loss": 0.1438, "step": 798 }, { "epoch": 0.15, "learning_rate": 1.2207792207792208e-05, "loss": 0.1378, "step": 799 }, { "epoch": 0.15, "learning_rate": 1.2223071046600458e-05, "loss": 0.1474, "step": 800 }, { "epoch": 0.15, "learning_rate": 1.2238349885408708e-05, "loss": 0.1459, "step": 801 }, { "epoch": 0.15, "learning_rate": 1.2253628724216962e-05, "loss": 0.1585, "step": 802 }, { "epoch": 0.15, "learning_rate": 1.2268907563025212e-05, "loss": 0.1559, "step": 803 }, { "epoch": 0.15, "learning_rate": 1.2284186401833462e-05, "loss": 0.142, "step": 804 }, { "epoch": 0.15, "learning_rate": 1.2299465240641712e-05, "loss": 0.1436, "step": 805 }, { "epoch": 0.15, "learning_rate": 1.2314744079449962e-05, "loss": 0.1549, "step": 806 }, { "epoch": 0.15, "learning_rate": 1.2330022918258214e-05, "loss": 0.1324, "step": 807 }, { "epoch": 0.15, "learning_rate": 1.2345301757066464e-05, "loss": 0.1512, "step": 808 }, { "epoch": 0.15, "learning_rate": 1.2360580595874714e-05, "loss": 0.1433, "step": 809 }, { "epoch": 0.15, "learning_rate": 1.2375859434682964e-05, "loss": 0.1461, "step": 810 }, { "epoch": 0.15, "learning_rate": 1.2391138273491217e-05, "loss": 0.147, "step": 811 }, { "epoch": 0.15, "learning_rate": 1.2406417112299467e-05, "loss": 0.1573, "step": 812 }, { "epoch": 0.15, "learning_rate": 1.2421695951107717e-05, "loss": 0.1482, "step": 813 }, { "epoch": 0.15, "learning_rate": 1.2436974789915967e-05, "loss": 0.1395, "step": 814 }, { "epoch": 0.15, "learning_rate": 1.2452253628724218e-05, "loss": 0.1428, "step": 815 }, { "epoch": 0.15, "learning_rate": 1.2467532467532468e-05, "loss": 0.1478, "step": 816 }, { "epoch": 0.15, "learning_rate": 1.248281130634072e-05, "loss": 0.1504, "step": 817 }, { "epoch": 0.15, "learning_rate": 1.249809014514897e-05, "loss": 0.1383, "step": 818 }, { "epoch": 0.15, "learning_rate": 1.251336898395722e-05, "loss": 0.1512, "step": 819 }, { "epoch": 0.15, "learning_rate": 1.2528647822765471e-05, "loss": 0.16, "step": 820 }, { "epoch": 0.15, "learning_rate": 1.2543926661573723e-05, "loss": 0.1451, "step": 821 }, { "epoch": 0.15, "learning_rate": 1.2559205500381973e-05, "loss": 0.1385, "step": 822 }, { "epoch": 0.15, "learning_rate": 1.2574484339190223e-05, "loss": 0.1554, "step": 823 }, { "epoch": 0.15, "learning_rate": 1.2589763177998473e-05, "loss": 0.1378, "step": 824 }, { "epoch": 0.15, "learning_rate": 1.2605042016806723e-05, "loss": 0.1332, "step": 825 }, { "epoch": 0.15, "learning_rate": 1.2620320855614973e-05, "loss": 0.139, "step": 826 }, { "epoch": 0.15, "learning_rate": 1.2635599694423225e-05, "loss": 0.1609, "step": 827 }, { "epoch": 0.15, "learning_rate": 1.2650878533231477e-05, "loss": 0.1534, "step": 828 }, { "epoch": 0.15, "learning_rate": 1.2666157372039727e-05, "loss": 0.1448, "step": 829 }, { "epoch": 0.15, "learning_rate": 1.2681436210847977e-05, "loss": 0.1547, "step": 830 }, { "epoch": 0.15, "learning_rate": 1.2696715049656227e-05, "loss": 0.1339, "step": 831 }, { "epoch": 0.15, "learning_rate": 1.2711993888464478e-05, "loss": 0.1485, "step": 832 }, { "epoch": 0.15, "learning_rate": 1.2727272727272728e-05, "loss": 0.1498, "step": 833 }, { "epoch": 0.15, "learning_rate": 1.2742551566080979e-05, "loss": 0.1556, "step": 834 }, { "epoch": 0.15, "learning_rate": 1.2757830404889229e-05, "loss": 0.134, "step": 835 }, { "epoch": 0.15, "learning_rate": 1.2773109243697479e-05, "loss": 0.1415, "step": 836 }, { "epoch": 0.15, "learning_rate": 1.2788388082505732e-05, "loss": 0.1378, "step": 837 }, { "epoch": 0.15, "learning_rate": 1.2803666921313982e-05, "loss": 0.1402, "step": 838 }, { "epoch": 0.15, "learning_rate": 1.2818945760122232e-05, "loss": 0.1372, "step": 839 }, { "epoch": 0.15, "learning_rate": 1.2834224598930482e-05, "loss": 0.1525, "step": 840 }, { "epoch": 0.15, "learning_rate": 1.2849503437738732e-05, "loss": 0.1495, "step": 841 }, { "epoch": 0.15, "learning_rate": 1.2864782276546984e-05, "loss": 0.1453, "step": 842 }, { "epoch": 0.15, "learning_rate": 1.2880061115355234e-05, "loss": 0.1475, "step": 843 }, { "epoch": 0.15, "learning_rate": 1.2895339954163484e-05, "loss": 0.1517, "step": 844 }, { "epoch": 0.15, "learning_rate": 1.2910618792971734e-05, "loss": 0.1516, "step": 845 }, { "epoch": 0.16, "learning_rate": 1.2925897631779988e-05, "loss": 0.1591, "step": 846 }, { "epoch": 0.16, "learning_rate": 1.2941176470588238e-05, "loss": 0.1572, "step": 847 }, { "epoch": 0.16, "learning_rate": 1.2956455309396488e-05, "loss": 0.1322, "step": 848 }, { "epoch": 0.16, "learning_rate": 1.2971734148204738e-05, "loss": 0.1551, "step": 849 }, { "epoch": 0.16, "learning_rate": 1.2987012987012988e-05, "loss": 0.1564, "step": 850 }, { "epoch": 0.16, "learning_rate": 1.3002291825821238e-05, "loss": 0.1498, "step": 851 }, { "epoch": 0.16, "learning_rate": 1.3017570664629488e-05, "loss": 0.1486, "step": 852 }, { "epoch": 0.16, "learning_rate": 1.303284950343774e-05, "loss": 0.1366, "step": 853 }, { "epoch": 0.16, "learning_rate": 1.3048128342245991e-05, "loss": 0.1406, "step": 854 }, { "epoch": 0.16, "learning_rate": 1.3063407181054241e-05, "loss": 0.1471, "step": 855 }, { "epoch": 0.16, "learning_rate": 1.3078686019862491e-05, "loss": 0.1464, "step": 856 }, { "epoch": 0.16, "learning_rate": 1.3093964858670743e-05, "loss": 0.1593, "step": 857 }, { "epoch": 0.16, "learning_rate": 1.3109243697478993e-05, "loss": 0.1394, "step": 858 }, { "epoch": 0.16, "learning_rate": 1.3124522536287243e-05, "loss": 0.1425, "step": 859 }, { "epoch": 0.16, "learning_rate": 1.3139801375095493e-05, "loss": 0.1501, "step": 860 }, { "epoch": 0.16, "learning_rate": 1.3155080213903743e-05, "loss": 0.1456, "step": 861 }, { "epoch": 0.16, "learning_rate": 1.3170359052711993e-05, "loss": 0.1337, "step": 862 }, { "epoch": 0.16, "learning_rate": 1.3185637891520247e-05, "loss": 0.1481, "step": 863 }, { "epoch": 0.16, "learning_rate": 1.3200916730328497e-05, "loss": 0.1592, "step": 864 }, { "epoch": 0.16, "learning_rate": 1.3216195569136747e-05, "loss": 0.1465, "step": 865 }, { "epoch": 0.16, "learning_rate": 1.3231474407944997e-05, "loss": 0.1396, "step": 866 }, { "epoch": 0.16, "learning_rate": 1.3246753246753249e-05, "loss": 0.1369, "step": 867 }, { "epoch": 0.16, "learning_rate": 1.3262032085561499e-05, "loss": 0.1342, "step": 868 }, { "epoch": 0.16, "learning_rate": 1.3277310924369749e-05, "loss": 0.1543, "step": 869 }, { "epoch": 0.16, "learning_rate": 1.3292589763177999e-05, "loss": 0.1352, "step": 870 }, { "epoch": 0.16, "learning_rate": 1.3307868601986249e-05, "loss": 0.1382, "step": 871 }, { "epoch": 0.16, "learning_rate": 1.3323147440794502e-05, "loss": 0.1472, "step": 872 }, { "epoch": 0.16, "learning_rate": 1.3338426279602752e-05, "loss": 0.1592, "step": 873 }, { "epoch": 0.16, "learning_rate": 1.3353705118411002e-05, "loss": 0.1533, "step": 874 }, { "epoch": 0.16, "learning_rate": 1.3368983957219252e-05, "loss": 0.1466, "step": 875 }, { "epoch": 0.16, "learning_rate": 1.3384262796027502e-05, "loss": 0.1703, "step": 876 }, { "epoch": 0.16, "learning_rate": 1.3399541634835752e-05, "loss": 0.1457, "step": 877 }, { "epoch": 0.16, "learning_rate": 1.3414820473644004e-05, "loss": 0.1352, "step": 878 }, { "epoch": 0.16, "learning_rate": 1.3430099312452254e-05, "loss": 0.1553, "step": 879 }, { "epoch": 0.16, "learning_rate": 1.3445378151260506e-05, "loss": 0.1623, "step": 880 }, { "epoch": 0.16, "learning_rate": 1.3460656990068756e-05, "loss": 0.1367, "step": 881 }, { "epoch": 0.16, "learning_rate": 1.3475935828877008e-05, "loss": 0.1435, "step": 882 }, { "epoch": 0.16, "learning_rate": 1.3491214667685258e-05, "loss": 0.154, "step": 883 }, { "epoch": 0.16, "learning_rate": 1.3506493506493508e-05, "loss": 0.1344, "step": 884 }, { "epoch": 0.16, "learning_rate": 1.3521772345301758e-05, "loss": 0.1435, "step": 885 }, { "epoch": 0.16, "learning_rate": 1.3537051184110008e-05, "loss": 0.1397, "step": 886 }, { "epoch": 0.16, "learning_rate": 1.3552330022918258e-05, "loss": 0.1468, "step": 887 }, { "epoch": 0.16, "learning_rate": 1.356760886172651e-05, "loss": 0.1596, "step": 888 }, { "epoch": 0.16, "learning_rate": 1.3582887700534761e-05, "loss": 0.1496, "step": 889 }, { "epoch": 0.16, "learning_rate": 1.3598166539343011e-05, "loss": 0.1402, "step": 890 }, { "epoch": 0.16, "learning_rate": 1.3613445378151261e-05, "loss": 0.134, "step": 891 }, { "epoch": 0.16, "learning_rate": 1.3628724216959513e-05, "loss": 0.1444, "step": 892 }, { "epoch": 0.16, "learning_rate": 1.3644003055767763e-05, "loss": 0.1393, "step": 893 }, { "epoch": 0.16, "learning_rate": 1.3659281894576013e-05, "loss": 0.1492, "step": 894 }, { "epoch": 0.16, "learning_rate": 1.3674560733384263e-05, "loss": 0.1377, "step": 895 }, { "epoch": 0.16, "learning_rate": 1.3689839572192513e-05, "loss": 0.1476, "step": 896 }, { "epoch": 0.16, "learning_rate": 1.3705118411000763e-05, "loss": 0.1356, "step": 897 }, { "epoch": 0.16, "learning_rate": 1.3720397249809017e-05, "loss": 0.1359, "step": 898 }, { "epoch": 0.16, "learning_rate": 1.3735676088617267e-05, "loss": 0.133, "step": 899 }, { "epoch": 0.17, "learning_rate": 1.3750954927425517e-05, "loss": 0.1348, "step": 900 }, { "epoch": 0.17, "learning_rate": 1.3766233766233767e-05, "loss": 0.1377, "step": 901 }, { "epoch": 0.17, "learning_rate": 1.3781512605042017e-05, "loss": 0.1384, "step": 902 }, { "epoch": 0.17, "learning_rate": 1.3796791443850269e-05, "loss": 0.1453, "step": 903 }, { "epoch": 0.17, "learning_rate": 1.3812070282658519e-05, "loss": 0.1479, "step": 904 }, { "epoch": 0.17, "learning_rate": 1.3827349121466769e-05, "loss": 0.1459, "step": 905 }, { "epoch": 0.17, "learning_rate": 1.384262796027502e-05, "loss": 0.1443, "step": 906 }, { "epoch": 0.17, "learning_rate": 1.3857906799083272e-05, "loss": 0.1404, "step": 907 }, { "epoch": 0.17, "learning_rate": 1.3873185637891522e-05, "loss": 0.1485, "step": 908 }, { "epoch": 0.17, "learning_rate": 1.3888464476699772e-05, "loss": 0.1381, "step": 909 }, { "epoch": 0.17, "learning_rate": 1.3903743315508022e-05, "loss": 0.1506, "step": 910 }, { "epoch": 0.17, "learning_rate": 1.3919022154316273e-05, "loss": 0.1529, "step": 911 }, { "epoch": 0.17, "learning_rate": 1.3934300993124523e-05, "loss": 0.136, "step": 912 }, { "epoch": 0.17, "learning_rate": 1.3949579831932774e-05, "loss": 0.1415, "step": 913 }, { "epoch": 0.17, "learning_rate": 1.3964858670741024e-05, "loss": 0.1337, "step": 914 }, { "epoch": 0.17, "learning_rate": 1.3980137509549276e-05, "loss": 0.1443, "step": 915 }, { "epoch": 0.17, "learning_rate": 1.3995416348357526e-05, "loss": 0.1514, "step": 916 }, { "epoch": 0.17, "learning_rate": 1.4010695187165778e-05, "loss": 0.1448, "step": 917 }, { "epoch": 0.17, "learning_rate": 1.4025974025974028e-05, "loss": 0.14, "step": 918 }, { "epoch": 0.17, "learning_rate": 1.4041252864782278e-05, "loss": 0.1412, "step": 919 }, { "epoch": 0.17, "learning_rate": 1.4056531703590528e-05, "loss": 0.1407, "step": 920 }, { "epoch": 0.17, "learning_rate": 1.4071810542398778e-05, "loss": 0.1403, "step": 921 }, { "epoch": 0.17, "learning_rate": 1.4087089381207028e-05, "loss": 0.119, "step": 922 }, { "epoch": 0.17, "learning_rate": 1.4102368220015278e-05, "loss": 0.147, "step": 923 }, { "epoch": 0.17, "learning_rate": 1.4117647058823532e-05, "loss": 0.1447, "step": 924 }, { "epoch": 0.17, "learning_rate": 1.4132925897631782e-05, "loss": 0.136, "step": 925 }, { "epoch": 0.17, "learning_rate": 1.4148204736440032e-05, "loss": 0.1423, "step": 926 }, { "epoch": 0.17, "learning_rate": 1.4163483575248282e-05, "loss": 0.1512, "step": 927 }, { "epoch": 0.17, "learning_rate": 1.4178762414056533e-05, "loss": 0.1407, "step": 928 }, { "epoch": 0.17, "learning_rate": 1.4194041252864783e-05, "loss": 0.1562, "step": 929 }, { "epoch": 0.17, "learning_rate": 1.4209320091673034e-05, "loss": 0.1581, "step": 930 }, { "epoch": 0.17, "learning_rate": 1.4224598930481284e-05, "loss": 0.149, "step": 931 }, { "epoch": 0.17, "learning_rate": 1.4239877769289537e-05, "loss": 0.1389, "step": 932 }, { "epoch": 0.17, "learning_rate": 1.4255156608097787e-05, "loss": 0.1349, "step": 933 }, { "epoch": 0.17, "learning_rate": 1.4270435446906037e-05, "loss": 0.1361, "step": 934 }, { "epoch": 0.17, "learning_rate": 1.4285714285714287e-05, "loss": 0.1275, "step": 935 }, { "epoch": 0.17, "learning_rate": 1.4300993124522537e-05, "loss": 0.1342, "step": 936 }, { "epoch": 0.17, "learning_rate": 1.4316271963330787e-05, "loss": 0.1214, "step": 937 }, { "epoch": 0.17, "learning_rate": 1.4331550802139039e-05, "loss": 0.135, "step": 938 }, { "epoch": 0.17, "learning_rate": 1.4346829640947289e-05, "loss": 0.1356, "step": 939 }, { "epoch": 0.17, "learning_rate": 1.4362108479755539e-05, "loss": 0.1249, "step": 940 }, { "epoch": 0.17, "learning_rate": 1.437738731856379e-05, "loss": 0.15, "step": 941 }, { "epoch": 0.17, "learning_rate": 1.439266615737204e-05, "loss": 0.1304, "step": 942 }, { "epoch": 0.17, "learning_rate": 1.4407944996180293e-05, "loss": 0.1481, "step": 943 }, { "epoch": 0.17, "learning_rate": 1.4423223834988543e-05, "loss": 0.16, "step": 944 }, { "epoch": 0.17, "learning_rate": 1.4438502673796793e-05, "loss": 0.1368, "step": 945 }, { "epoch": 0.17, "learning_rate": 1.4453781512605043e-05, "loss": 0.1363, "step": 946 }, { "epoch": 0.17, "learning_rate": 1.4469060351413293e-05, "loss": 0.1315, "step": 947 }, { "epoch": 0.17, "learning_rate": 1.4484339190221543e-05, "loss": 0.1364, "step": 948 }, { "epoch": 0.17, "learning_rate": 1.4499618029029795e-05, "loss": 0.1479, "step": 949 }, { "epoch": 0.17, "learning_rate": 1.4514896867838046e-05, "loss": 0.1352, "step": 950 }, { "epoch": 0.17, "learning_rate": 1.4530175706646296e-05, "loss": 0.1323, "step": 951 }, { "epoch": 0.17, "learning_rate": 1.4545454545454546e-05, "loss": 0.1306, "step": 952 }, { "epoch": 0.17, "learning_rate": 1.4560733384262798e-05, "loss": 0.1326, "step": 953 }, { "epoch": 0.17, "learning_rate": 1.4576012223071048e-05, "loss": 0.121, "step": 954 }, { "epoch": 0.18, "learning_rate": 1.4591291061879298e-05, "loss": 0.1368, "step": 955 }, { "epoch": 0.18, "learning_rate": 1.4606569900687548e-05, "loss": 0.1519, "step": 956 }, { "epoch": 0.18, "learning_rate": 1.4621848739495798e-05, "loss": 0.1286, "step": 957 }, { "epoch": 0.18, "learning_rate": 1.4637127578304052e-05, "loss": 0.1473, "step": 958 }, { "epoch": 0.18, "learning_rate": 1.4652406417112302e-05, "loss": 0.1331, "step": 959 }, { "epoch": 0.18, "learning_rate": 1.4667685255920552e-05, "loss": 0.1503, "step": 960 }, { "epoch": 0.18, "learning_rate": 1.4682964094728802e-05, "loss": 0.1541, "step": 961 }, { "epoch": 0.18, "learning_rate": 1.4698242933537052e-05, "loss": 0.1258, "step": 962 }, { "epoch": 0.18, "learning_rate": 1.4713521772345304e-05, "loss": 0.1279, "step": 963 }, { "epoch": 0.18, "learning_rate": 1.4728800611153554e-05, "loss": 0.1347, "step": 964 }, { "epoch": 0.18, "learning_rate": 1.4744079449961804e-05, "loss": 0.1493, "step": 965 }, { "epoch": 0.18, "learning_rate": 1.4759358288770054e-05, "loss": 0.1562, "step": 966 }, { "epoch": 0.18, "learning_rate": 1.4774637127578305e-05, "loss": 0.1429, "step": 967 }, { "epoch": 0.18, "learning_rate": 1.4789915966386557e-05, "loss": 0.1533, "step": 968 }, { "epoch": 0.18, "learning_rate": 1.4805194805194807e-05, "loss": 0.144, "step": 969 }, { "epoch": 0.18, "learning_rate": 1.4820473644003057e-05, "loss": 0.1416, "step": 970 }, { "epoch": 0.18, "learning_rate": 1.4835752482811307e-05, "loss": 0.1448, "step": 971 }, { "epoch": 0.18, "learning_rate": 1.4851031321619557e-05, "loss": 0.1478, "step": 972 }, { "epoch": 0.18, "learning_rate": 1.4866310160427807e-05, "loss": 0.1504, "step": 973 }, { "epoch": 0.18, "learning_rate": 1.488158899923606e-05, "loss": 0.1284, "step": 974 }, { "epoch": 0.18, "learning_rate": 1.489686783804431e-05, "loss": 0.142, "step": 975 }, { "epoch": 0.18, "learning_rate": 1.4912146676852561e-05, "loss": 0.1361, "step": 976 }, { "epoch": 0.18, "learning_rate": 1.4927425515660811e-05, "loss": 0.1307, "step": 977 }, { "epoch": 0.18, "learning_rate": 1.4942704354469063e-05, "loss": 0.1421, "step": 978 }, { "epoch": 0.18, "learning_rate": 1.4957983193277313e-05, "loss": 0.1484, "step": 979 }, { "epoch": 0.18, "learning_rate": 1.4973262032085563e-05, "loss": 0.1448, "step": 980 }, { "epoch": 0.18, "learning_rate": 1.4988540870893813e-05, "loss": 0.1487, "step": 981 }, { "epoch": 0.18, "learning_rate": 1.5003819709702063e-05, "loss": 0.1292, "step": 982 }, { "epoch": 0.18, "learning_rate": 1.5019098548510313e-05, "loss": 0.1445, "step": 983 }, { "epoch": 0.18, "learning_rate": 1.5034377387318566e-05, "loss": 0.1303, "step": 984 }, { "epoch": 0.18, "learning_rate": 1.5049656226126816e-05, "loss": 0.1508, "step": 985 }, { "epoch": 0.18, "learning_rate": 1.5064935064935066e-05, "loss": 0.1489, "step": 986 }, { "epoch": 0.18, "learning_rate": 1.5080213903743316e-05, "loss": 0.1481, "step": 987 }, { "epoch": 0.18, "learning_rate": 1.5095492742551567e-05, "loss": 0.1404, "step": 988 }, { "epoch": 0.18, "learning_rate": 1.5110771581359818e-05, "loss": 0.1381, "step": 989 }, { "epoch": 0.18, "learning_rate": 1.5126050420168068e-05, "loss": 0.1349, "step": 990 }, { "epoch": 0.18, "learning_rate": 1.5141329258976318e-05, "loss": 0.1322, "step": 991 }, { "epoch": 0.18, "learning_rate": 1.5156608097784568e-05, "loss": 0.1364, "step": 992 }, { "epoch": 0.18, "learning_rate": 1.5171886936592822e-05, "loss": 0.1391, "step": 993 }, { "epoch": 0.18, "learning_rate": 1.5187165775401072e-05, "loss": 0.1334, "step": 994 }, { "epoch": 0.18, "learning_rate": 1.5202444614209322e-05, "loss": 0.1397, "step": 995 }, { "epoch": 0.18, "learning_rate": 1.5217723453017572e-05, "loss": 0.1399, "step": 996 }, { "epoch": 0.18, "learning_rate": 1.5233002291825822e-05, "loss": 0.1464, "step": 997 }, { "epoch": 0.18, "learning_rate": 1.5248281130634072e-05, "loss": 0.1423, "step": 998 }, { "epoch": 0.18, "learning_rate": 1.5263559969442322e-05, "loss": 0.1335, "step": 999 }, { "epoch": 0.18, "learning_rate": 1.5278838808250572e-05, "loss": 0.1272, "step": 1000 }, { "epoch": 0.18, "learning_rate": 1.5294117647058822e-05, "loss": 0.1367, "step": 1001 }, { "epoch": 0.18, "learning_rate": 1.5309396485867076e-05, "loss": 0.1361, "step": 1002 }, { "epoch": 0.18, "learning_rate": 1.5324675324675326e-05, "loss": 0.1366, "step": 1003 }, { "epoch": 0.18, "learning_rate": 1.5339954163483576e-05, "loss": 0.1415, "step": 1004 }, { "epoch": 0.18, "learning_rate": 1.5355233002291826e-05, "loss": 0.134, "step": 1005 }, { "epoch": 0.18, "learning_rate": 1.537051184110008e-05, "loss": 0.1514, "step": 1006 }, { "epoch": 0.18, "learning_rate": 1.538579067990833e-05, "loss": 0.1296, "step": 1007 }, { "epoch": 0.18, "learning_rate": 1.540106951871658e-05, "loss": 0.1254, "step": 1008 }, { "epoch": 0.19, "learning_rate": 1.541634835752483e-05, "loss": 0.1368, "step": 1009 }, { "epoch": 0.19, "learning_rate": 1.543162719633308e-05, "loss": 0.1531, "step": 1010 }, { "epoch": 0.19, "learning_rate": 1.5446906035141333e-05, "loss": 0.1201, "step": 1011 }, { "epoch": 0.19, "learning_rate": 1.5462184873949583e-05, "loss": 0.144, "step": 1012 }, { "epoch": 0.19, "learning_rate": 1.5477463712757833e-05, "loss": 0.1465, "step": 1013 }, { "epoch": 0.19, "learning_rate": 1.5492742551566083e-05, "loss": 0.139, "step": 1014 }, { "epoch": 0.19, "learning_rate": 1.5508021390374333e-05, "loss": 0.1461, "step": 1015 }, { "epoch": 0.19, "learning_rate": 1.5523300229182583e-05, "loss": 0.1401, "step": 1016 }, { "epoch": 0.19, "learning_rate": 1.5538579067990833e-05, "loss": 0.1286, "step": 1017 }, { "epoch": 0.19, "learning_rate": 1.5553857906799083e-05, "loss": 0.1523, "step": 1018 }, { "epoch": 0.19, "learning_rate": 1.5569136745607337e-05, "loss": 0.1371, "step": 1019 }, { "epoch": 0.19, "learning_rate": 1.5584415584415587e-05, "loss": 0.147, "step": 1020 }, { "epoch": 0.19, "learning_rate": 1.5599694423223837e-05, "loss": 0.1547, "step": 1021 }, { "epoch": 0.19, "learning_rate": 1.5614973262032087e-05, "loss": 0.1258, "step": 1022 }, { "epoch": 0.19, "learning_rate": 1.5630252100840337e-05, "loss": 0.1284, "step": 1023 }, { "epoch": 0.19, "learning_rate": 1.5645530939648587e-05, "loss": 0.1476, "step": 1024 }, { "epoch": 0.19, "learning_rate": 1.5660809778456837e-05, "loss": 0.1456, "step": 1025 }, { "epoch": 0.19, "learning_rate": 1.5676088617265087e-05, "loss": 0.1448, "step": 1026 }, { "epoch": 0.19, "learning_rate": 1.569136745607334e-05, "loss": 0.1558, "step": 1027 }, { "epoch": 0.19, "learning_rate": 1.570664629488159e-05, "loss": 0.1429, "step": 1028 }, { "epoch": 0.19, "learning_rate": 1.572192513368984e-05, "loss": 0.1518, "step": 1029 }, { "epoch": 0.19, "learning_rate": 1.573720397249809e-05, "loss": 0.131, "step": 1030 }, { "epoch": 0.19, "learning_rate": 1.5752482811306344e-05, "loss": 0.1394, "step": 1031 }, { "epoch": 0.19, "learning_rate": 1.5767761650114594e-05, "loss": 0.1345, "step": 1032 }, { "epoch": 0.19, "learning_rate": 1.5783040488922844e-05, "loss": 0.1446, "step": 1033 }, { "epoch": 0.19, "learning_rate": 1.5798319327731094e-05, "loss": 0.1414, "step": 1034 }, { "epoch": 0.19, "learning_rate": 1.5813598166539344e-05, "loss": 0.1617, "step": 1035 }, { "epoch": 0.19, "learning_rate": 1.5828877005347597e-05, "loss": 0.1368, "step": 1036 }, { "epoch": 0.19, "learning_rate": 1.5844155844155847e-05, "loss": 0.1427, "step": 1037 }, { "epoch": 0.19, "learning_rate": 1.5859434682964098e-05, "loss": 0.1495, "step": 1038 }, { "epoch": 0.19, "learning_rate": 1.5874713521772348e-05, "loss": 0.1564, "step": 1039 }, { "epoch": 0.19, "learning_rate": 1.5889992360580598e-05, "loss": 0.1517, "step": 1040 }, { "epoch": 0.19, "learning_rate": 1.5905271199388848e-05, "loss": 0.1395, "step": 1041 }, { "epoch": 0.19, "learning_rate": 1.5920550038197098e-05, "loss": 0.1469, "step": 1042 }, { "epoch": 0.19, "learning_rate": 1.5935828877005348e-05, "loss": 0.1369, "step": 1043 }, { "epoch": 0.19, "learning_rate": 1.5951107715813598e-05, "loss": 0.1394, "step": 1044 }, { "epoch": 0.19, "learning_rate": 1.596638655462185e-05, "loss": 0.1353, "step": 1045 }, { "epoch": 0.19, "learning_rate": 1.59816653934301e-05, "loss": 0.145, "step": 1046 }, { "epoch": 0.19, "learning_rate": 1.599694423223835e-05, "loss": 0.1348, "step": 1047 }, { "epoch": 0.19, "learning_rate": 1.60122230710466e-05, "loss": 0.1374, "step": 1048 }, { "epoch": 0.19, "learning_rate": 1.602750190985485e-05, "loss": 0.1298, "step": 1049 }, { "epoch": 0.19, "learning_rate": 1.60427807486631e-05, "loss": 0.1516, "step": 1050 }, { "epoch": 0.19, "learning_rate": 1.605805958747135e-05, "loss": 0.1511, "step": 1051 }, { "epoch": 0.19, "learning_rate": 1.6073338426279605e-05, "loss": 0.146, "step": 1052 }, { "epoch": 0.19, "learning_rate": 1.6088617265087855e-05, "loss": 0.1347, "step": 1053 }, { "epoch": 0.19, "learning_rate": 1.6103896103896105e-05, "loss": 0.1408, "step": 1054 }, { "epoch": 0.19, "learning_rate": 1.6119174942704355e-05, "loss": 0.1416, "step": 1055 }, { "epoch": 0.19, "learning_rate": 1.6134453781512605e-05, "loss": 0.1478, "step": 1056 }, { "epoch": 0.19, "learning_rate": 1.614973262032086e-05, "loss": 0.1428, "step": 1057 }, { "epoch": 0.19, "learning_rate": 1.616501145912911e-05, "loss": 0.1472, "step": 1058 }, { "epoch": 0.19, "learning_rate": 1.618029029793736e-05, "loss": 0.1375, "step": 1059 }, { "epoch": 0.19, "learning_rate": 1.619556913674561e-05, "loss": 0.1592, "step": 1060 }, { "epoch": 0.19, "learning_rate": 1.621084797555386e-05, "loss": 0.1475, "step": 1061 }, { "epoch": 0.19, "learning_rate": 1.6226126814362112e-05, "loss": 0.1392, "step": 1062 }, { "epoch": 0.19, "learning_rate": 1.6241405653170362e-05, "loss": 0.1509, "step": 1063 }, { "epoch": 0.2, "learning_rate": 1.6256684491978612e-05, "loss": 0.1378, "step": 1064 }, { "epoch": 0.2, "learning_rate": 1.6271963330786862e-05, "loss": 0.1317, "step": 1065 }, { "epoch": 0.2, "learning_rate": 1.6287242169595112e-05, "loss": 0.1481, "step": 1066 }, { "epoch": 0.2, "learning_rate": 1.6302521008403362e-05, "loss": 0.1551, "step": 1067 }, { "epoch": 0.2, "learning_rate": 1.6317799847211612e-05, "loss": 0.1469, "step": 1068 }, { "epoch": 0.2, "learning_rate": 1.6333078686019862e-05, "loss": 0.1343, "step": 1069 }, { "epoch": 0.2, "learning_rate": 1.6348357524828112e-05, "loss": 0.1331, "step": 1070 }, { "epoch": 0.2, "learning_rate": 1.6363636363636366e-05, "loss": 0.136, "step": 1071 }, { "epoch": 0.2, "learning_rate": 1.6378915202444616e-05, "loss": 0.1386, "step": 1072 }, { "epoch": 0.2, "learning_rate": 1.6394194041252866e-05, "loss": 0.1483, "step": 1073 }, { "epoch": 0.2, "learning_rate": 1.6409472880061116e-05, "loss": 0.1435, "step": 1074 }, { "epoch": 0.2, "learning_rate": 1.6424751718869366e-05, "loss": 0.1395, "step": 1075 }, { "epoch": 0.2, "learning_rate": 1.6440030557677616e-05, "loss": 0.1369, "step": 1076 }, { "epoch": 0.2, "learning_rate": 1.645530939648587e-05, "loss": 0.1333, "step": 1077 }, { "epoch": 0.2, "learning_rate": 1.647058823529412e-05, "loss": 0.1462, "step": 1078 }, { "epoch": 0.2, "learning_rate": 1.648586707410237e-05, "loss": 0.1393, "step": 1079 }, { "epoch": 0.2, "learning_rate": 1.650114591291062e-05, "loss": 0.1439, "step": 1080 }, { "epoch": 0.2, "learning_rate": 1.651642475171887e-05, "loss": 0.1471, "step": 1081 }, { "epoch": 0.2, "learning_rate": 1.6531703590527123e-05, "loss": 0.1371, "step": 1082 }, { "epoch": 0.2, "learning_rate": 1.6546982429335373e-05, "loss": 0.1484, "step": 1083 }, { "epoch": 0.2, "learning_rate": 1.6562261268143623e-05, "loss": 0.1478, "step": 1084 }, { "epoch": 0.2, "learning_rate": 1.6577540106951873e-05, "loss": 0.1355, "step": 1085 }, { "epoch": 0.2, "learning_rate": 1.6592818945760123e-05, "loss": 0.1332, "step": 1086 }, { "epoch": 0.2, "learning_rate": 1.6608097784568373e-05, "loss": 0.1251, "step": 1087 }, { "epoch": 0.2, "learning_rate": 1.6623376623376627e-05, "loss": 0.1362, "step": 1088 }, { "epoch": 0.2, "learning_rate": 1.6638655462184877e-05, "loss": 0.1391, "step": 1089 }, { "epoch": 0.2, "learning_rate": 1.6653934300993127e-05, "loss": 0.1298, "step": 1090 }, { "epoch": 0.2, "learning_rate": 1.6669213139801377e-05, "loss": 0.1459, "step": 1091 }, { "epoch": 0.2, "learning_rate": 1.6684491978609627e-05, "loss": 0.1463, "step": 1092 }, { "epoch": 0.2, "learning_rate": 1.6699770817417877e-05, "loss": 0.1427, "step": 1093 }, { "epoch": 0.2, "learning_rate": 1.6715049656226127e-05, "loss": 0.1504, "step": 1094 }, { "epoch": 0.2, "learning_rate": 1.6730328495034377e-05, "loss": 0.1478, "step": 1095 }, { "epoch": 0.2, "learning_rate": 1.6745607333842627e-05, "loss": 0.1407, "step": 1096 }, { "epoch": 0.2, "learning_rate": 1.676088617265088e-05, "loss": 0.1464, "step": 1097 }, { "epoch": 0.2, "learning_rate": 1.677616501145913e-05, "loss": 0.126, "step": 1098 }, { "epoch": 0.2, "learning_rate": 1.679144385026738e-05, "loss": 0.1336, "step": 1099 }, { "epoch": 0.2, "learning_rate": 1.680672268907563e-05, "loss": 0.1334, "step": 1100 }, { "epoch": 0.2, "learning_rate": 1.682200152788388e-05, "loss": 0.1372, "step": 1101 }, { "epoch": 0.2, "learning_rate": 1.6837280366692134e-05, "loss": 0.1399, "step": 1102 }, { "epoch": 0.2, "learning_rate": 1.6852559205500384e-05, "loss": 0.1479, "step": 1103 }, { "epoch": 0.2, "learning_rate": 1.6867838044308634e-05, "loss": 0.1413, "step": 1104 }, { "epoch": 0.2, "learning_rate": 1.6883116883116884e-05, "loss": 0.1548, "step": 1105 }, { "epoch": 0.2, "learning_rate": 1.6898395721925134e-05, "loss": 0.1544, "step": 1106 }, { "epoch": 0.2, "learning_rate": 1.6913674560733388e-05, "loss": 0.1427, "step": 1107 }, { "epoch": 0.2, "learning_rate": 1.6928953399541638e-05, "loss": 0.1408, "step": 1108 }, { "epoch": 0.2, "learning_rate": 1.6944232238349888e-05, "loss": 0.1495, "step": 1109 }, { "epoch": 0.2, "learning_rate": 1.6959511077158138e-05, "loss": 0.1406, "step": 1110 }, { "epoch": 0.2, "learning_rate": 1.6974789915966388e-05, "loss": 0.1473, "step": 1111 }, { "epoch": 0.2, "learning_rate": 1.6990068754774638e-05, "loss": 0.1415, "step": 1112 }, { "epoch": 0.2, "learning_rate": 1.7005347593582888e-05, "loss": 0.1385, "step": 1113 }, { "epoch": 0.2, "learning_rate": 1.702062643239114e-05, "loss": 0.1406, "step": 1114 }, { "epoch": 0.2, "learning_rate": 1.703590527119939e-05, "loss": 0.154, "step": 1115 }, { "epoch": 0.2, "learning_rate": 1.705118411000764e-05, "loss": 0.1323, "step": 1116 }, { "epoch": 0.2, "learning_rate": 1.706646294881589e-05, "loss": 0.1399, "step": 1117 }, { "epoch": 0.21, "learning_rate": 1.708174178762414e-05, "loss": 0.1479, "step": 1118 }, { "epoch": 0.21, "learning_rate": 1.7097020626432392e-05, "loss": 0.1366, "step": 1119 }, { "epoch": 0.21, "learning_rate": 1.7112299465240642e-05, "loss": 0.1238, "step": 1120 }, { "epoch": 0.21, "learning_rate": 1.7127578304048892e-05, "loss": 0.1523, "step": 1121 }, { "epoch": 0.21, "learning_rate": 1.7142857142857142e-05, "loss": 0.1534, "step": 1122 }, { "epoch": 0.21, "learning_rate": 1.7158135981665395e-05, "loss": 0.1378, "step": 1123 }, { "epoch": 0.21, "learning_rate": 1.7173414820473645e-05, "loss": 0.1266, "step": 1124 }, { "epoch": 0.21, "learning_rate": 1.7188693659281895e-05, "loss": 0.1382, "step": 1125 }, { "epoch": 0.21, "learning_rate": 1.7203972498090145e-05, "loss": 0.1436, "step": 1126 }, { "epoch": 0.21, "learning_rate": 1.7219251336898395e-05, "loss": 0.1439, "step": 1127 }, { "epoch": 0.21, "learning_rate": 1.723453017570665e-05, "loss": 0.1366, "step": 1128 }, { "epoch": 0.21, "learning_rate": 1.72498090145149e-05, "loss": 0.145, "step": 1129 }, { "epoch": 0.21, "learning_rate": 1.726508785332315e-05, "loss": 0.1216, "step": 1130 }, { "epoch": 0.21, "learning_rate": 1.72803666921314e-05, "loss": 0.1473, "step": 1131 }, { "epoch": 0.21, "learning_rate": 1.7295645530939652e-05, "loss": 0.1381, "step": 1132 }, { "epoch": 0.21, "learning_rate": 1.7310924369747902e-05, "loss": 0.1394, "step": 1133 }, { "epoch": 0.21, "learning_rate": 1.7326203208556153e-05, "loss": 0.1496, "step": 1134 }, { "epoch": 0.21, "learning_rate": 1.7341482047364403e-05, "loss": 0.1256, "step": 1135 }, { "epoch": 0.21, "learning_rate": 1.7356760886172653e-05, "loss": 0.1297, "step": 1136 }, { "epoch": 0.21, "learning_rate": 1.7372039724980903e-05, "loss": 0.1356, "step": 1137 }, { "epoch": 0.21, "learning_rate": 1.7387318563789153e-05, "loss": 0.1246, "step": 1138 }, { "epoch": 0.21, "learning_rate": 1.7402597402597403e-05, "loss": 0.1268, "step": 1139 }, { "epoch": 0.21, "learning_rate": 1.7417876241405656e-05, "loss": 0.1423, "step": 1140 }, { "epoch": 0.21, "learning_rate": 1.7433155080213906e-05, "loss": 0.1305, "step": 1141 }, { "epoch": 0.21, "learning_rate": 1.7448433919022156e-05, "loss": 0.1357, "step": 1142 }, { "epoch": 0.21, "learning_rate": 1.7463712757830406e-05, "loss": 0.1403, "step": 1143 }, { "epoch": 0.21, "learning_rate": 1.7478991596638656e-05, "loss": 0.1341, "step": 1144 }, { "epoch": 0.21, "learning_rate": 1.7494270435446906e-05, "loss": 0.1372, "step": 1145 }, { "epoch": 0.21, "learning_rate": 1.7509549274255156e-05, "loss": 0.1295, "step": 1146 }, { "epoch": 0.21, "learning_rate": 1.7524828113063406e-05, "loss": 0.1399, "step": 1147 }, { "epoch": 0.21, "learning_rate": 1.754010695187166e-05, "loss": 0.1319, "step": 1148 }, { "epoch": 0.21, "learning_rate": 1.755538579067991e-05, "loss": 0.1376, "step": 1149 }, { "epoch": 0.21, "learning_rate": 1.757066462948816e-05, "loss": 0.1342, "step": 1150 }, { "epoch": 0.21, "learning_rate": 1.758594346829641e-05, "loss": 0.1353, "step": 1151 }, { "epoch": 0.21, "learning_rate": 1.760122230710466e-05, "loss": 0.1466, "step": 1152 }, { "epoch": 0.21, "learning_rate": 1.7616501145912914e-05, "loss": 0.1326, "step": 1153 }, { "epoch": 0.21, "learning_rate": 1.7631779984721164e-05, "loss": 0.141, "step": 1154 }, { "epoch": 0.21, "learning_rate": 1.7647058823529414e-05, "loss": 0.1307, "step": 1155 }, { "epoch": 0.21, "learning_rate": 1.7662337662337664e-05, "loss": 0.1282, "step": 1156 }, { "epoch": 0.21, "learning_rate": 1.7677616501145914e-05, "loss": 0.1437, "step": 1157 }, { "epoch": 0.21, "learning_rate": 1.7692895339954167e-05, "loss": 0.1254, "step": 1158 }, { "epoch": 0.21, "learning_rate": 1.7708174178762417e-05, "loss": 0.1549, "step": 1159 }, { "epoch": 0.21, "learning_rate": 1.7723453017570667e-05, "loss": 0.1316, "step": 1160 }, { "epoch": 0.21, "learning_rate": 1.7738731856378917e-05, "loss": 0.1431, "step": 1161 }, { "epoch": 0.21, "learning_rate": 1.7754010695187167e-05, "loss": 0.1481, "step": 1162 }, { "epoch": 0.21, "learning_rate": 1.7769289533995417e-05, "loss": 0.1323, "step": 1163 }, { "epoch": 0.21, "learning_rate": 1.7784568372803667e-05, "loss": 0.1448, "step": 1164 }, { "epoch": 0.21, "learning_rate": 1.7799847211611917e-05, "loss": 0.1533, "step": 1165 }, { "epoch": 0.21, "learning_rate": 1.781512605042017e-05, "loss": 0.1482, "step": 1166 }, { "epoch": 0.21, "learning_rate": 1.783040488922842e-05, "loss": 0.1424, "step": 1167 }, { "epoch": 0.21, "learning_rate": 1.784568372803667e-05, "loss": 0.1294, "step": 1168 }, { "epoch": 0.21, "learning_rate": 1.786096256684492e-05, "loss": 0.1422, "step": 1169 }, { "epoch": 0.21, "learning_rate": 1.787624140565317e-05, "loss": 0.1489, "step": 1170 }, { "epoch": 0.21, "learning_rate": 1.789152024446142e-05, "loss": 0.1277, "step": 1171 }, { "epoch": 0.21, "learning_rate": 1.790679908326967e-05, "loss": 0.1298, "step": 1172 }, { "epoch": 0.22, "learning_rate": 1.792207792207792e-05, "loss": 0.1423, "step": 1173 }, { "epoch": 0.22, "learning_rate": 1.7937356760886175e-05, "loss": 0.1458, "step": 1174 }, { "epoch": 0.22, "learning_rate": 1.7952635599694425e-05, "loss": 0.144, "step": 1175 }, { "epoch": 0.22, "learning_rate": 1.7967914438502675e-05, "loss": 0.1444, "step": 1176 }, { "epoch": 0.22, "learning_rate": 1.7983193277310925e-05, "loss": 0.1383, "step": 1177 }, { "epoch": 0.22, "learning_rate": 1.7998472116119178e-05, "loss": 0.141, "step": 1178 }, { "epoch": 0.22, "learning_rate": 1.8013750954927428e-05, "loss": 0.1354, "step": 1179 }, { "epoch": 0.22, "learning_rate": 1.8029029793735678e-05, "loss": 0.1304, "step": 1180 }, { "epoch": 0.22, "learning_rate": 1.8044308632543928e-05, "loss": 0.1439, "step": 1181 }, { "epoch": 0.22, "learning_rate": 1.805958747135218e-05, "loss": 0.1413, "step": 1182 }, { "epoch": 0.22, "learning_rate": 1.807486631016043e-05, "loss": 0.1359, "step": 1183 }, { "epoch": 0.22, "learning_rate": 1.8090145148968682e-05, "loss": 0.133, "step": 1184 }, { "epoch": 0.22, "learning_rate": 1.8105423987776932e-05, "loss": 0.1404, "step": 1185 }, { "epoch": 0.22, "learning_rate": 1.8120702826585182e-05, "loss": 0.1269, "step": 1186 }, { "epoch": 0.22, "learning_rate": 1.8135981665393432e-05, "loss": 0.1358, "step": 1187 }, { "epoch": 0.22, "learning_rate": 1.8151260504201682e-05, "loss": 0.1318, "step": 1188 }, { "epoch": 0.22, "learning_rate": 1.8166539343009932e-05, "loss": 0.1352, "step": 1189 }, { "epoch": 0.22, "learning_rate": 1.8181818181818182e-05, "loss": 0.1482, "step": 1190 }, { "epoch": 0.22, "learning_rate": 1.8197097020626432e-05, "loss": 0.1252, "step": 1191 }, { "epoch": 0.22, "learning_rate": 1.8212375859434686e-05, "loss": 0.1336, "step": 1192 }, { "epoch": 0.22, "learning_rate": 1.8227654698242936e-05, "loss": 0.1442, "step": 1193 }, { "epoch": 0.22, "learning_rate": 1.8242933537051186e-05, "loss": 0.1448, "step": 1194 }, { "epoch": 0.22, "learning_rate": 1.8258212375859436e-05, "loss": 0.1338, "step": 1195 }, { "epoch": 0.22, "learning_rate": 1.8273491214667686e-05, "loss": 0.1395, "step": 1196 }, { "epoch": 0.22, "learning_rate": 1.8288770053475936e-05, "loss": 0.1365, "step": 1197 }, { "epoch": 0.22, "learning_rate": 1.8304048892284186e-05, "loss": 0.1483, "step": 1198 }, { "epoch": 0.22, "learning_rate": 1.831932773109244e-05, "loss": 0.1407, "step": 1199 }, { "epoch": 0.22, "learning_rate": 1.833460656990069e-05, "loss": 0.1335, "step": 1200 }, { "epoch": 0.22, "learning_rate": 1.834988540870894e-05, "loss": 0.1423, "step": 1201 }, { "epoch": 0.22, "learning_rate": 1.836516424751719e-05, "loss": 0.1339, "step": 1202 }, { "epoch": 0.22, "learning_rate": 1.8380443086325443e-05, "loss": 0.1593, "step": 1203 }, { "epoch": 0.22, "learning_rate": 1.8395721925133693e-05, "loss": 0.1324, "step": 1204 }, { "epoch": 0.22, "learning_rate": 1.8411000763941943e-05, "loss": 0.1301, "step": 1205 }, { "epoch": 0.22, "learning_rate": 1.8426279602750193e-05, "loss": 0.1326, "step": 1206 }, { "epoch": 0.22, "learning_rate": 1.8441558441558443e-05, "loss": 0.1375, "step": 1207 }, { "epoch": 0.22, "learning_rate": 1.8456837280366693e-05, "loss": 0.1364, "step": 1208 }, { "epoch": 0.22, "learning_rate": 1.8472116119174943e-05, "loss": 0.1559, "step": 1209 }, { "epoch": 0.22, "learning_rate": 1.8487394957983196e-05, "loss": 0.1287, "step": 1210 }, { "epoch": 0.22, "learning_rate": 1.8502673796791447e-05, "loss": 0.1354, "step": 1211 }, { "epoch": 0.22, "learning_rate": 1.8517952635599697e-05, "loss": 0.1548, "step": 1212 }, { "epoch": 0.22, "learning_rate": 1.8533231474407947e-05, "loss": 0.1339, "step": 1213 }, { "epoch": 0.22, "learning_rate": 1.8548510313216197e-05, "loss": 0.1413, "step": 1214 }, { "epoch": 0.22, "learning_rate": 1.8563789152024447e-05, "loss": 0.1336, "step": 1215 }, { "epoch": 0.22, "learning_rate": 1.8579067990832697e-05, "loss": 0.1442, "step": 1216 }, { "epoch": 0.22, "learning_rate": 1.8594346829640947e-05, "loss": 0.1342, "step": 1217 }, { "epoch": 0.22, "learning_rate": 1.86096256684492e-05, "loss": 0.1295, "step": 1218 }, { "epoch": 0.22, "learning_rate": 1.862490450725745e-05, "loss": 0.1363, "step": 1219 }, { "epoch": 0.22, "learning_rate": 1.86401833460657e-05, "loss": 0.1321, "step": 1220 }, { "epoch": 0.22, "learning_rate": 1.865546218487395e-05, "loss": 0.14, "step": 1221 }, { "epoch": 0.22, "learning_rate": 1.86707410236822e-05, "loss": 0.1309, "step": 1222 }, { "epoch": 0.22, "learning_rate": 1.868601986249045e-05, "loss": 0.1314, "step": 1223 }, { "epoch": 0.22, "learning_rate": 1.8701298701298704e-05, "loss": 0.1331, "step": 1224 }, { "epoch": 0.22, "learning_rate": 1.8716577540106954e-05, "loss": 0.1367, "step": 1225 }, { "epoch": 0.22, "learning_rate": 1.8731856378915204e-05, "loss": 0.1439, "step": 1226 }, { "epoch": 0.23, "learning_rate": 1.8747135217723454e-05, "loss": 0.1425, "step": 1227 }, { "epoch": 0.23, "learning_rate": 1.8762414056531707e-05, "loss": 0.1371, "step": 1228 }, { "epoch": 0.23, "learning_rate": 1.8777692895339957e-05, "loss": 0.1365, "step": 1229 }, { "epoch": 0.23, "learning_rate": 1.8792971734148208e-05, "loss": 0.1355, "step": 1230 }, { "epoch": 0.23, "learning_rate": 1.8808250572956458e-05, "loss": 0.1328, "step": 1231 }, { "epoch": 0.23, "learning_rate": 1.8823529411764708e-05, "loss": 0.1256, "step": 1232 }, { "epoch": 0.23, "learning_rate": 1.8838808250572958e-05, "loss": 0.1363, "step": 1233 }, { "epoch": 0.23, "learning_rate": 1.8854087089381208e-05, "loss": 0.152, "step": 1234 }, { "epoch": 0.23, "learning_rate": 1.8869365928189458e-05, "loss": 0.1401, "step": 1235 }, { "epoch": 0.23, "learning_rate": 1.888464476699771e-05, "loss": 0.1386, "step": 1236 }, { "epoch": 0.23, "learning_rate": 1.889992360580596e-05, "loss": 0.1394, "step": 1237 }, { "epoch": 0.23, "learning_rate": 1.891520244461421e-05, "loss": 0.1298, "step": 1238 }, { "epoch": 0.23, "learning_rate": 1.893048128342246e-05, "loss": 0.1554, "step": 1239 }, { "epoch": 0.23, "learning_rate": 1.894576012223071e-05, "loss": 0.1322, "step": 1240 }, { "epoch": 0.23, "learning_rate": 1.896103896103896e-05, "loss": 0.1285, "step": 1241 }, { "epoch": 0.23, "learning_rate": 1.897631779984721e-05, "loss": 0.1392, "step": 1242 }, { "epoch": 0.23, "learning_rate": 1.899159663865546e-05, "loss": 0.1334, "step": 1243 }, { "epoch": 0.23, "learning_rate": 1.9006875477463715e-05, "loss": 0.1303, "step": 1244 }, { "epoch": 0.23, "learning_rate": 1.9022154316271965e-05, "loss": 0.1255, "step": 1245 }, { "epoch": 0.23, "learning_rate": 1.9037433155080215e-05, "loss": 0.1358, "step": 1246 }, { "epoch": 0.23, "learning_rate": 1.9052711993888465e-05, "loss": 0.1306, "step": 1247 }, { "epoch": 0.23, "learning_rate": 1.9067990832696715e-05, "loss": 0.1421, "step": 1248 }, { "epoch": 0.23, "learning_rate": 1.908326967150497e-05, "loss": 0.1298, "step": 1249 }, { "epoch": 0.23, "learning_rate": 1.909854851031322e-05, "loss": 0.1348, "step": 1250 }, { "epoch": 0.23, "learning_rate": 1.911382734912147e-05, "loss": 0.1325, "step": 1251 }, { "epoch": 0.23, "learning_rate": 1.912910618792972e-05, "loss": 0.1455, "step": 1252 }, { "epoch": 0.23, "learning_rate": 1.9144385026737972e-05, "loss": 0.1298, "step": 1253 }, { "epoch": 0.23, "learning_rate": 1.9159663865546222e-05, "loss": 0.1376, "step": 1254 }, { "epoch": 0.23, "learning_rate": 1.9174942704354472e-05, "loss": 0.1179, "step": 1255 }, { "epoch": 0.23, "learning_rate": 1.9190221543162722e-05, "loss": 0.1231, "step": 1256 }, { "epoch": 0.23, "learning_rate": 1.9205500381970972e-05, "loss": 0.1186, "step": 1257 }, { "epoch": 0.23, "learning_rate": 1.9220779220779222e-05, "loss": 0.1299, "step": 1258 }, { "epoch": 0.23, "learning_rate": 1.9236058059587472e-05, "loss": 0.148, "step": 1259 }, { "epoch": 0.23, "learning_rate": 1.9251336898395722e-05, "loss": 0.1309, "step": 1260 }, { "epoch": 0.23, "learning_rate": 1.9266615737203972e-05, "loss": 0.1362, "step": 1261 }, { "epoch": 0.23, "learning_rate": 1.9281894576012226e-05, "loss": 0.147, "step": 1262 }, { "epoch": 0.23, "learning_rate": 1.9297173414820476e-05, "loss": 0.1342, "step": 1263 }, { "epoch": 0.23, "learning_rate": 1.9312452253628726e-05, "loss": 0.1501, "step": 1264 }, { "epoch": 0.23, "learning_rate": 1.9327731092436976e-05, "loss": 0.1394, "step": 1265 }, { "epoch": 0.23, "learning_rate": 1.9343009931245226e-05, "loss": 0.1281, "step": 1266 }, { "epoch": 0.23, "learning_rate": 1.9358288770053476e-05, "loss": 0.1265, "step": 1267 }, { "epoch": 0.23, "learning_rate": 1.9373567608861726e-05, "loss": 0.1178, "step": 1268 }, { "epoch": 0.23, "learning_rate": 1.9388846447669976e-05, "loss": 0.1255, "step": 1269 }, { "epoch": 0.23, "learning_rate": 1.940412528647823e-05, "loss": 0.1393, "step": 1270 }, { "epoch": 0.23, "learning_rate": 1.941940412528648e-05, "loss": 0.1304, "step": 1271 }, { "epoch": 0.23, "learning_rate": 1.943468296409473e-05, "loss": 0.1392, "step": 1272 }, { "epoch": 0.23, "learning_rate": 1.944996180290298e-05, "loss": 0.1487, "step": 1273 }, { "epoch": 0.23, "learning_rate": 1.9465240641711233e-05, "loss": 0.1234, "step": 1274 }, { "epoch": 0.23, "learning_rate": 1.9480519480519483e-05, "loss": 0.1341, "step": 1275 }, { "epoch": 0.23, "learning_rate": 1.9495798319327733e-05, "loss": 0.1429, "step": 1276 }, { "epoch": 0.23, "learning_rate": 1.9511077158135983e-05, "loss": 0.1468, "step": 1277 }, { "epoch": 0.23, "learning_rate": 1.9526355996944233e-05, "loss": 0.1416, "step": 1278 }, { "epoch": 0.23, "learning_rate": 1.9541634835752487e-05, "loss": 0.1432, "step": 1279 }, { "epoch": 0.23, "learning_rate": 1.9556913674560737e-05, "loss": 0.1433, "step": 1280 }, { "epoch": 0.23, "learning_rate": 1.9572192513368987e-05, "loss": 0.1217, "step": 1281 }, { "epoch": 0.24, "learning_rate": 1.9587471352177237e-05, "loss": 0.1418, "step": 1282 }, { "epoch": 0.24, "learning_rate": 1.9602750190985487e-05, "loss": 0.139, "step": 1283 }, { "epoch": 0.24, "learning_rate": 1.9618029029793737e-05, "loss": 0.1319, "step": 1284 }, { "epoch": 0.24, "learning_rate": 1.9633307868601987e-05, "loss": 0.1099, "step": 1285 }, { "epoch": 0.24, "learning_rate": 1.9648586707410237e-05, "loss": 0.1455, "step": 1286 }, { "epoch": 0.24, "learning_rate": 1.9663865546218487e-05, "loss": 0.1377, "step": 1287 }, { "epoch": 0.24, "learning_rate": 1.967914438502674e-05, "loss": 0.1436, "step": 1288 }, { "epoch": 0.24, "learning_rate": 1.969442322383499e-05, "loss": 0.1244, "step": 1289 }, { "epoch": 0.24, "learning_rate": 1.970970206264324e-05, "loss": 0.1471, "step": 1290 }, { "epoch": 0.24, "learning_rate": 1.972498090145149e-05, "loss": 0.1177, "step": 1291 }, { "epoch": 0.24, "learning_rate": 1.974025974025974e-05, "loss": 0.1276, "step": 1292 }, { "epoch": 0.24, "learning_rate": 1.975553857906799e-05, "loss": 0.132, "step": 1293 }, { "epoch": 0.24, "learning_rate": 1.977081741787624e-05, "loss": 0.1402, "step": 1294 }, { "epoch": 0.24, "learning_rate": 1.9786096256684494e-05, "loss": 0.1307, "step": 1295 }, { "epoch": 0.24, "learning_rate": 1.9801375095492744e-05, "loss": 0.121, "step": 1296 }, { "epoch": 0.24, "learning_rate": 1.9816653934300994e-05, "loss": 0.137, "step": 1297 }, { "epoch": 0.24, "learning_rate": 1.9831932773109244e-05, "loss": 0.136, "step": 1298 }, { "epoch": 0.24, "learning_rate": 1.9847211611917498e-05, "loss": 0.1484, "step": 1299 }, { "epoch": 0.24, "learning_rate": 1.9862490450725748e-05, "loss": 0.1285, "step": 1300 }, { "epoch": 0.24, "learning_rate": 1.9877769289533998e-05, "loss": 0.1326, "step": 1301 }, { "epoch": 0.24, "learning_rate": 1.9893048128342248e-05, "loss": 0.1207, "step": 1302 }, { "epoch": 0.24, "learning_rate": 1.9908326967150498e-05, "loss": 0.1292, "step": 1303 }, { "epoch": 0.24, "learning_rate": 1.9923605805958748e-05, "loss": 0.1284, "step": 1304 }, { "epoch": 0.24, "learning_rate": 1.9938884644767e-05, "loss": 0.1385, "step": 1305 }, { "epoch": 0.24, "learning_rate": 1.995416348357525e-05, "loss": 0.1329, "step": 1306 }, { "epoch": 0.24, "learning_rate": 1.99694423223835e-05, "loss": 0.1317, "step": 1307 }, { "epoch": 0.24, "learning_rate": 1.998472116119175e-05, "loss": 0.1286, "step": 1308 }, { "epoch": 0.24, "learning_rate": 2e-05, "loss": 0.1343, "step": 1309 }, { "epoch": 0.24, "learning_rate": 1.9999999972439884e-05, "loss": 0.1232, "step": 1310 }, { "epoch": 0.24, "learning_rate": 1.9999999889759534e-05, "loss": 0.1303, "step": 1311 }, { "epoch": 0.24, "learning_rate": 1.9999999751958946e-05, "loss": 0.1364, "step": 1312 }, { "epoch": 0.24, "learning_rate": 1.999999955903813e-05, "loss": 0.1325, "step": 1313 }, { "epoch": 0.24, "learning_rate": 1.999999931099708e-05, "loss": 0.1425, "step": 1314 }, { "epoch": 0.24, "learning_rate": 1.9999999007835802e-05, "loss": 0.1359, "step": 1315 }, { "epoch": 0.24, "learning_rate": 1.999999864955429e-05, "loss": 0.1283, "step": 1316 }, { "epoch": 0.24, "learning_rate": 1.9999998236152555e-05, "loss": 0.1322, "step": 1317 }, { "epoch": 0.24, "learning_rate": 1.9999997767630598e-05, "loss": 0.1282, "step": 1318 }, { "epoch": 0.24, "learning_rate": 1.9999997243988413e-05, "loss": 0.1343, "step": 1319 }, { "epoch": 0.24, "learning_rate": 1.9999996665226013e-05, "loss": 0.1393, "step": 1320 }, { "epoch": 0.24, "learning_rate": 1.9999996031343394e-05, "loss": 0.1284, "step": 1321 }, { "epoch": 0.24, "learning_rate": 1.9999995342340564e-05, "loss": 0.1329, "step": 1322 }, { "epoch": 0.24, "learning_rate": 1.9999994598217526e-05, "loss": 0.1314, "step": 1323 }, { "epoch": 0.24, "learning_rate": 1.9999993798974282e-05, "loss": 0.1414, "step": 1324 }, { "epoch": 0.24, "learning_rate": 1.999999294461084e-05, "loss": 0.1336, "step": 1325 }, { "epoch": 0.24, "learning_rate": 1.99999920351272e-05, "loss": 0.1302, "step": 1326 }, { "epoch": 0.24, "learning_rate": 1.9999991070523374e-05, "loss": 0.1415, "step": 1327 }, { "epoch": 0.24, "learning_rate": 1.9999990050799362e-05, "loss": 0.1293, "step": 1328 }, { "epoch": 0.24, "learning_rate": 1.9999988975955165e-05, "loss": 0.1296, "step": 1329 }, { "epoch": 0.24, "learning_rate": 1.99999878459908e-05, "loss": 0.1461, "step": 1330 }, { "epoch": 0.24, "learning_rate": 1.9999986660906266e-05, "loss": 0.1397, "step": 1331 }, { "epoch": 0.24, "learning_rate": 1.999998542070157e-05, "loss": 0.1346, "step": 1332 }, { "epoch": 0.24, "learning_rate": 1.9999984125376724e-05, "loss": 0.1502, "step": 1333 }, { "epoch": 0.24, "learning_rate": 1.9999982774931726e-05, "loss": 0.1277, "step": 1334 }, { "epoch": 0.24, "learning_rate": 1.9999981369366592e-05, "loss": 0.1321, "step": 1335 }, { "epoch": 0.25, "learning_rate": 1.9999979908681327e-05, "loss": 0.137, "step": 1336 }, { "epoch": 0.25, "learning_rate": 1.999997839287594e-05, "loss": 0.1499, "step": 1337 }, { "epoch": 0.25, "learning_rate": 1.9999976821950434e-05, "loss": 0.1343, "step": 1338 }, { "epoch": 0.25, "learning_rate": 1.999997519590482e-05, "loss": 0.1308, "step": 1339 }, { "epoch": 0.25, "learning_rate": 1.999997351473911e-05, "loss": 0.1323, "step": 1340 }, { "epoch": 0.25, "learning_rate": 1.9999971778453317e-05, "loss": 0.1263, "step": 1341 }, { "epoch": 0.25, "learning_rate": 1.999996998704744e-05, "loss": 0.121, "step": 1342 }, { "epoch": 0.25, "learning_rate": 1.9999968140521493e-05, "loss": 0.1254, "step": 1343 }, { "epoch": 0.25, "learning_rate": 1.999996623887549e-05, "loss": 0.1227, "step": 1344 }, { "epoch": 0.25, "learning_rate": 1.9999964282109436e-05, "loss": 0.1303, "step": 1345 }, { "epoch": 0.25, "learning_rate": 1.999996227022335e-05, "loss": 0.1467, "step": 1346 }, { "epoch": 0.25, "learning_rate": 1.9999960203217233e-05, "loss": 0.132, "step": 1347 }, { "epoch": 0.25, "learning_rate": 1.9999958081091102e-05, "loss": 0.1256, "step": 1348 }, { "epoch": 0.25, "learning_rate": 1.9999955903844967e-05, "loss": 0.1299, "step": 1349 }, { "epoch": 0.25, "learning_rate": 1.9999953671478843e-05, "loss": 0.1407, "step": 1350 }, { "epoch": 0.25, "learning_rate": 1.999995138399274e-05, "loss": 0.1473, "step": 1351 }, { "epoch": 0.25, "learning_rate": 1.999994904138667e-05, "loss": 0.1461, "step": 1352 }, { "epoch": 0.25, "learning_rate": 1.9999946643660646e-05, "loss": 0.1293, "step": 1353 }, { "epoch": 0.25, "learning_rate": 1.9999944190814685e-05, "loss": 0.1325, "step": 1354 }, { "epoch": 0.25, "learning_rate": 1.9999941682848794e-05, "loss": 0.1311, "step": 1355 }, { "epoch": 0.25, "learning_rate": 1.999993911976299e-05, "loss": 0.1339, "step": 1356 }, { "epoch": 0.25, "learning_rate": 1.9999936501557288e-05, "loss": 0.142, "step": 1357 }, { "epoch": 0.25, "learning_rate": 1.99999338282317e-05, "loss": 0.1445, "step": 1358 }, { "epoch": 0.25, "learning_rate": 1.9999931099786247e-05, "loss": 0.128, "step": 1359 }, { "epoch": 0.25, "learning_rate": 1.9999928316220938e-05, "loss": 0.1359, "step": 1360 }, { "epoch": 0.25, "learning_rate": 1.999992547753579e-05, "loss": 0.1349, "step": 1361 }, { "epoch": 0.25, "learning_rate": 1.9999922583730816e-05, "loss": 0.1342, "step": 1362 }, { "epoch": 0.25, "learning_rate": 1.9999919634806037e-05, "loss": 0.1269, "step": 1363 }, { "epoch": 0.25, "learning_rate": 1.9999916630761463e-05, "loss": 0.1302, "step": 1364 }, { "epoch": 0.25, "learning_rate": 1.999991357159712e-05, "loss": 0.1306, "step": 1365 }, { "epoch": 0.25, "learning_rate": 1.9999910457313012e-05, "loss": 0.1344, "step": 1366 }, { "epoch": 0.25, "learning_rate": 1.9999907287909166e-05, "loss": 0.1228, "step": 1367 }, { "epoch": 0.25, "learning_rate": 1.99999040633856e-05, "loss": 0.1304, "step": 1368 }, { "epoch": 0.25, "learning_rate": 1.9999900783742325e-05, "loss": 0.1268, "step": 1369 }, { "epoch": 0.25, "learning_rate": 1.9999897448979365e-05, "loss": 0.1457, "step": 1370 }, { "epoch": 0.25, "learning_rate": 1.9999894059096735e-05, "loss": 0.1332, "step": 1371 }, { "epoch": 0.25, "learning_rate": 1.9999890614094456e-05, "loss": 0.1228, "step": 1372 }, { "epoch": 0.25, "learning_rate": 1.999988711397254e-05, "loss": 0.1408, "step": 1373 }, { "epoch": 0.25, "learning_rate": 1.9999883558731018e-05, "loss": 0.1326, "step": 1374 }, { "epoch": 0.25, "learning_rate": 1.9999879948369903e-05, "loss": 0.132, "step": 1375 }, { "epoch": 0.25, "learning_rate": 1.9999876282889213e-05, "loss": 0.126, "step": 1376 }, { "epoch": 0.25, "learning_rate": 1.9999872562288973e-05, "loss": 0.1277, "step": 1377 }, { "epoch": 0.25, "learning_rate": 1.99998687865692e-05, "loss": 0.1355, "step": 1378 }, { "epoch": 0.25, "learning_rate": 1.9999864955729917e-05, "loss": 0.1379, "step": 1379 }, { "epoch": 0.25, "learning_rate": 1.9999861069771142e-05, "loss": 0.1337, "step": 1380 }, { "epoch": 0.25, "learning_rate": 1.99998571286929e-05, "loss": 0.1434, "step": 1381 }, { "epoch": 0.25, "learning_rate": 1.999985313249521e-05, "loss": 0.1336, "step": 1382 }, { "epoch": 0.25, "learning_rate": 1.9999849081178095e-05, "loss": 0.1341, "step": 1383 }, { "epoch": 0.25, "learning_rate": 1.999984497474158e-05, "loss": 0.1431, "step": 1384 }, { "epoch": 0.25, "learning_rate": 1.9999840813185684e-05, "loss": 0.1285, "step": 1385 }, { "epoch": 0.25, "learning_rate": 1.999983659651043e-05, "loss": 0.1233, "step": 1386 }, { "epoch": 0.25, "learning_rate": 1.9999832324715846e-05, "loss": 0.1302, "step": 1387 }, { "epoch": 0.25, "learning_rate": 1.9999827997801947e-05, "loss": 0.1282, "step": 1388 }, { "epoch": 0.25, "learning_rate": 1.9999823615768766e-05, "loss": 0.1408, "step": 1389 }, { "epoch": 0.25, "learning_rate": 1.999981917861632e-05, "loss": 0.1329, "step": 1390 }, { "epoch": 0.26, "learning_rate": 1.9999814686344637e-05, "loss": 0.1269, "step": 1391 }, { "epoch": 0.26, "learning_rate": 1.9999810138953744e-05, "loss": 0.1302, "step": 1392 }, { "epoch": 0.26, "learning_rate": 1.999980553644366e-05, "loss": 0.1304, "step": 1393 }, { "epoch": 0.26, "learning_rate": 1.9999800878814414e-05, "loss": 0.1386, "step": 1394 }, { "epoch": 0.26, "learning_rate": 1.9999796166066034e-05, "loss": 0.143, "step": 1395 }, { "epoch": 0.26, "learning_rate": 1.999979139819854e-05, "loss": 0.1237, "step": 1396 }, { "epoch": 0.26, "learning_rate": 1.9999786575211963e-05, "loss": 0.1398, "step": 1397 }, { "epoch": 0.26, "learning_rate": 1.999978169710633e-05, "loss": 0.1424, "step": 1398 }, { "epoch": 0.26, "learning_rate": 1.9999776763881663e-05, "loss": 0.1338, "step": 1399 }, { "epoch": 0.26, "learning_rate": 1.999977177553799e-05, "loss": 0.1229, "step": 1400 }, { "epoch": 0.26, "learning_rate": 1.9999766732075345e-05, "loss": 0.144, "step": 1401 }, { "epoch": 0.26, "learning_rate": 1.9999761633493754e-05, "loss": 0.1577, "step": 1402 }, { "epoch": 0.26, "learning_rate": 1.999975647979324e-05, "loss": 0.1386, "step": 1403 }, { "epoch": 0.26, "learning_rate": 1.999975127097383e-05, "loss": 0.1283, "step": 1404 }, { "epoch": 0.26, "learning_rate": 1.9999746007035557e-05, "loss": 0.1387, "step": 1405 }, { "epoch": 0.26, "learning_rate": 1.9999740687978456e-05, "loss": 0.1303, "step": 1406 }, { "epoch": 0.26, "learning_rate": 1.9999735313802546e-05, "loss": 0.1298, "step": 1407 }, { "epoch": 0.26, "learning_rate": 1.999972988450786e-05, "loss": 0.139, "step": 1408 }, { "epoch": 0.26, "learning_rate": 1.999972440009443e-05, "loss": 0.1438, "step": 1409 }, { "epoch": 0.26, "learning_rate": 1.9999718860562286e-05, "loss": 0.1339, "step": 1410 }, { "epoch": 0.26, "learning_rate": 1.9999713265911452e-05, "loss": 0.1272, "step": 1411 }, { "epoch": 0.26, "learning_rate": 1.999970761614197e-05, "loss": 0.1459, "step": 1412 }, { "epoch": 0.26, "learning_rate": 1.9999701911253868e-05, "loss": 0.1278, "step": 1413 }, { "epoch": 0.26, "learning_rate": 1.999969615124717e-05, "loss": 0.1294, "step": 1414 }, { "epoch": 0.26, "learning_rate": 1.999969033612191e-05, "loss": 0.1373, "step": 1415 }, { "epoch": 0.26, "learning_rate": 1.9999684465878127e-05, "loss": 0.1439, "step": 1416 }, { "epoch": 0.26, "learning_rate": 1.9999678540515845e-05, "loss": 0.1423, "step": 1417 }, { "epoch": 0.26, "learning_rate": 1.9999672560035102e-05, "loss": 0.1278, "step": 1418 }, { "epoch": 0.26, "learning_rate": 1.999966652443593e-05, "loss": 0.124, "step": 1419 }, { "epoch": 0.26, "learning_rate": 1.9999660433718367e-05, "loss": 0.1317, "step": 1420 }, { "epoch": 0.26, "learning_rate": 1.9999654287882435e-05, "loss": 0.129, "step": 1421 }, { "epoch": 0.26, "learning_rate": 1.9999648086928175e-05, "loss": 0.1353, "step": 1422 }, { "epoch": 0.26, "learning_rate": 1.9999641830855624e-05, "loss": 0.1418, "step": 1423 }, { "epoch": 0.26, "learning_rate": 1.999963551966481e-05, "loss": 0.1329, "step": 1424 }, { "epoch": 0.26, "learning_rate": 1.999962915335577e-05, "loss": 0.1259, "step": 1425 }, { "epoch": 0.26, "learning_rate": 1.9999622731928543e-05, "loss": 0.1291, "step": 1426 }, { "epoch": 0.26, "learning_rate": 1.999961625538316e-05, "loss": 0.1335, "step": 1427 }, { "epoch": 0.26, "learning_rate": 1.9999609723719657e-05, "loss": 0.1293, "step": 1428 }, { "epoch": 0.26, "learning_rate": 1.999960313693807e-05, "loss": 0.1429, "step": 1429 }, { "epoch": 0.26, "learning_rate": 1.9999596495038438e-05, "loss": 0.1202, "step": 1430 }, { "epoch": 0.26, "learning_rate": 1.9999589798020796e-05, "loss": 0.1372, "step": 1431 }, { "epoch": 0.26, "learning_rate": 1.9999583045885182e-05, "loss": 0.1405, "step": 1432 }, { "epoch": 0.26, "learning_rate": 1.999957623863163e-05, "loss": 0.1366, "step": 1433 }, { "epoch": 0.26, "learning_rate": 1.999956937626018e-05, "loss": 0.1229, "step": 1434 }, { "epoch": 0.26, "learning_rate": 1.9999562458770866e-05, "loss": 0.1335, "step": 1435 }, { "epoch": 0.26, "learning_rate": 1.9999555486163733e-05, "loss": 0.1333, "step": 1436 }, { "epoch": 0.26, "learning_rate": 1.9999548458438817e-05, "loss": 0.1347, "step": 1437 }, { "epoch": 0.26, "learning_rate": 1.9999541375596153e-05, "loss": 0.1413, "step": 1438 }, { "epoch": 0.26, "learning_rate": 1.999953423763578e-05, "loss": 0.1265, "step": 1439 }, { "epoch": 0.26, "learning_rate": 1.9999527044557746e-05, "loss": 0.1297, "step": 1440 }, { "epoch": 0.26, "learning_rate": 1.9999519796362084e-05, "loss": 0.1288, "step": 1441 }, { "epoch": 0.26, "learning_rate": 1.9999512493048832e-05, "loss": 0.131, "step": 1442 }, { "epoch": 0.26, "learning_rate": 1.9999505134618036e-05, "loss": 0.1343, "step": 1443 }, { "epoch": 0.26, "learning_rate": 1.999949772106973e-05, "loss": 0.1319, "step": 1444 }, { "epoch": 0.26, "learning_rate": 1.9999490252403963e-05, "loss": 0.1371, "step": 1445 }, { "epoch": 0.27, "learning_rate": 1.9999482728620765e-05, "loss": 0.1358, "step": 1446 }, { "epoch": 0.27, "learning_rate": 1.9999475149720187e-05, "loss": 0.1295, "step": 1447 }, { "epoch": 0.27, "learning_rate": 1.9999467515702268e-05, "loss": 0.1323, "step": 1448 }, { "epoch": 0.27, "learning_rate": 1.999945982656705e-05, "loss": 0.132, "step": 1449 }, { "epoch": 0.27, "learning_rate": 1.9999452082314573e-05, "loss": 0.1333, "step": 1450 }, { "epoch": 0.27, "learning_rate": 1.9999444282944885e-05, "loss": 0.1315, "step": 1451 }, { "epoch": 0.27, "learning_rate": 1.9999436428458025e-05, "loss": 0.1417, "step": 1452 }, { "epoch": 0.27, "learning_rate": 1.9999428518854037e-05, "loss": 0.1421, "step": 1453 }, { "epoch": 0.27, "learning_rate": 1.999942055413296e-05, "loss": 0.1434, "step": 1454 }, { "epoch": 0.27, "learning_rate": 1.9999412534294847e-05, "loss": 0.1356, "step": 1455 }, { "epoch": 0.27, "learning_rate": 1.999940445933974e-05, "loss": 0.1434, "step": 1456 }, { "epoch": 0.27, "learning_rate": 1.999939632926768e-05, "loss": 0.1387, "step": 1457 }, { "epoch": 0.27, "learning_rate": 1.9999388144078713e-05, "loss": 0.1291, "step": 1458 }, { "epoch": 0.27, "learning_rate": 1.9999379903772885e-05, "loss": 0.1304, "step": 1459 }, { "epoch": 0.27, "learning_rate": 1.9999371608350238e-05, "loss": 0.1457, "step": 1460 }, { "epoch": 0.27, "learning_rate": 1.999936325781082e-05, "loss": 0.1519, "step": 1461 }, { "epoch": 0.27, "learning_rate": 1.999935485215468e-05, "loss": 0.1313, "step": 1462 }, { "epoch": 0.27, "learning_rate": 1.9999346391381862e-05, "loss": 0.1363, "step": 1463 }, { "epoch": 0.27, "learning_rate": 1.9999337875492412e-05, "loss": 0.1327, "step": 1464 }, { "epoch": 0.27, "learning_rate": 1.9999329304486375e-05, "loss": 0.1409, "step": 1465 }, { "epoch": 0.27, "learning_rate": 1.9999320678363803e-05, "loss": 0.1268, "step": 1466 }, { "epoch": 0.27, "learning_rate": 1.999931199712474e-05, "loss": 0.1165, "step": 1467 }, { "epoch": 0.27, "learning_rate": 1.9999303260769235e-05, "loss": 0.1319, "step": 1468 }, { "epoch": 0.27, "learning_rate": 1.9999294469297338e-05, "loss": 0.1231, "step": 1469 }, { "epoch": 0.27, "learning_rate": 1.999928562270909e-05, "loss": 0.1269, "step": 1470 }, { "epoch": 0.27, "learning_rate": 1.9999276721004556e-05, "loss": 0.1336, "step": 1471 }, { "epoch": 0.27, "learning_rate": 1.9999267764183767e-05, "loss": 0.1367, "step": 1472 }, { "epoch": 0.27, "learning_rate": 1.999925875224678e-05, "loss": 0.1303, "step": 1473 }, { "epoch": 0.27, "learning_rate": 1.9999249685193646e-05, "loss": 0.1324, "step": 1474 }, { "epoch": 0.27, "learning_rate": 1.9999240563024412e-05, "loss": 0.1387, "step": 1475 }, { "epoch": 0.27, "learning_rate": 1.9999231385739134e-05, "loss": 0.1245, "step": 1476 }, { "epoch": 0.27, "learning_rate": 1.9999222153337853e-05, "loss": 0.1315, "step": 1477 }, { "epoch": 0.27, "learning_rate": 1.999921286582063e-05, "loss": 0.1208, "step": 1478 }, { "epoch": 0.27, "learning_rate": 1.9999203523187506e-05, "loss": 0.1208, "step": 1479 }, { "epoch": 0.27, "learning_rate": 1.9999194125438542e-05, "loss": 0.1284, "step": 1480 }, { "epoch": 0.27, "learning_rate": 1.9999184672573785e-05, "loss": 0.1276, "step": 1481 }, { "epoch": 0.27, "learning_rate": 1.9999175164593286e-05, "loss": 0.1316, "step": 1482 }, { "epoch": 0.27, "learning_rate": 1.99991656014971e-05, "loss": 0.1258, "step": 1483 }, { "epoch": 0.27, "learning_rate": 1.9999155983285278e-05, "loss": 0.1292, "step": 1484 }, { "epoch": 0.27, "learning_rate": 1.999914630995788e-05, "loss": 0.1266, "step": 1485 }, { "epoch": 0.27, "learning_rate": 1.9999136581514947e-05, "loss": 0.1305, "step": 1486 }, { "epoch": 0.27, "learning_rate": 1.999912679795654e-05, "loss": 0.1355, "step": 1487 }, { "epoch": 0.27, "learning_rate": 1.9999116959282712e-05, "loss": 0.1238, "step": 1488 }, { "epoch": 0.27, "learning_rate": 1.9999107065493518e-05, "loss": 0.1296, "step": 1489 }, { "epoch": 0.27, "learning_rate": 1.999909711658901e-05, "loss": 0.1316, "step": 1490 }, { "epoch": 0.27, "learning_rate": 1.9999087112569246e-05, "loss": 0.1253, "step": 1491 }, { "epoch": 0.27, "learning_rate": 1.999907705343428e-05, "loss": 0.1351, "step": 1492 }, { "epoch": 0.27, "learning_rate": 1.9999066939184165e-05, "loss": 0.1303, "step": 1493 }, { "epoch": 0.27, "learning_rate": 1.999905676981896e-05, "loss": 0.1273, "step": 1494 }, { "epoch": 0.27, "learning_rate": 1.999904654533872e-05, "loss": 0.1255, "step": 1495 }, { "epoch": 0.27, "learning_rate": 1.99990362657435e-05, "loss": 0.141, "step": 1496 }, { "epoch": 0.27, "learning_rate": 1.999902593103336e-05, "loss": 0.1218, "step": 1497 }, { "epoch": 0.27, "learning_rate": 1.9999015541208355e-05, "loss": 0.1232, "step": 1498 }, { "epoch": 0.27, "learning_rate": 1.999900509626854e-05, "loss": 0.1433, "step": 1499 }, { "epoch": 0.28, "learning_rate": 1.9998994596213976e-05, "loss": 0.1416, "step": 1500 }, { "epoch": 0.28, "learning_rate": 1.999898404104472e-05, "loss": 0.125, "step": 1501 }, { "epoch": 0.28, "learning_rate": 1.9998973430760827e-05, "loss": 0.1366, "step": 1502 }, { "epoch": 0.28, "learning_rate": 1.999896276536236e-05, "loss": 0.1457, "step": 1503 }, { "epoch": 0.28, "learning_rate": 1.9998952044849376e-05, "loss": 0.1353, "step": 1504 }, { "epoch": 0.28, "learning_rate": 1.9998941269221933e-05, "loss": 0.1365, "step": 1505 }, { "epoch": 0.28, "learning_rate": 1.999893043848009e-05, "loss": 0.1325, "step": 1506 }, { "epoch": 0.28, "learning_rate": 1.9998919552623913e-05, "loss": 0.1302, "step": 1507 }, { "epoch": 0.28, "learning_rate": 1.9998908611653456e-05, "loss": 0.1423, "step": 1508 }, { "epoch": 0.28, "learning_rate": 1.9998897615568778e-05, "loss": 0.136, "step": 1509 }, { "epoch": 0.28, "learning_rate": 1.9998886564369943e-05, "loss": 0.1185, "step": 1510 }, { "epoch": 0.28, "learning_rate": 1.999887545805701e-05, "loss": 0.1276, "step": 1511 }, { "epoch": 0.28, "learning_rate": 1.9998864296630044e-05, "loss": 0.1215, "step": 1512 }, { "epoch": 0.28, "learning_rate": 1.9998853080089104e-05, "loss": 0.1298, "step": 1513 }, { "epoch": 0.28, "learning_rate": 1.999884180843425e-05, "loss": 0.1181, "step": 1514 }, { "epoch": 0.28, "learning_rate": 1.9998830481665544e-05, "loss": 0.1293, "step": 1515 }, { "epoch": 0.28, "learning_rate": 1.999881909978305e-05, "loss": 0.126, "step": 1516 }, { "epoch": 0.28, "learning_rate": 1.9998807662786833e-05, "loss": 0.1491, "step": 1517 }, { "epoch": 0.28, "learning_rate": 1.9998796170676954e-05, "loss": 0.1285, "step": 1518 }, { "epoch": 0.28, "learning_rate": 1.9998784623453477e-05, "loss": 0.1332, "step": 1519 }, { "epoch": 0.28, "learning_rate": 1.9998773021116462e-05, "loss": 0.1221, "step": 1520 }, { "epoch": 0.28, "learning_rate": 1.999876136366598e-05, "loss": 0.1468, "step": 1521 }, { "epoch": 0.28, "learning_rate": 1.9998749651102086e-05, "loss": 0.1287, "step": 1522 }, { "epoch": 0.28, "learning_rate": 1.9998737883424854e-05, "loss": 0.1329, "step": 1523 }, { "epoch": 0.28, "learning_rate": 1.9998726060634342e-05, "loss": 0.1242, "step": 1524 }, { "epoch": 0.28, "learning_rate": 1.9998714182730618e-05, "loss": 0.1251, "step": 1525 }, { "epoch": 0.28, "learning_rate": 1.9998702249713747e-05, "loss": 0.1405, "step": 1526 }, { "epoch": 0.28, "learning_rate": 1.9998690261583796e-05, "loss": 0.1395, "step": 1527 }, { "epoch": 0.28, "learning_rate": 1.9998678218340833e-05, "loss": 0.1307, "step": 1528 }, { "epoch": 0.28, "learning_rate": 1.9998666119984916e-05, "loss": 0.1279, "step": 1529 }, { "epoch": 0.28, "learning_rate": 1.999865396651612e-05, "loss": 0.1278, "step": 1530 }, { "epoch": 0.28, "learning_rate": 1.9998641757934508e-05, "loss": 0.1403, "step": 1531 }, { "epoch": 0.28, "learning_rate": 1.9998629494240148e-05, "loss": 0.1333, "step": 1532 }, { "epoch": 0.28, "learning_rate": 1.999861717543311e-05, "loss": 0.114, "step": 1533 }, { "epoch": 0.28, "learning_rate": 1.999860480151346e-05, "loss": 0.1261, "step": 1534 }, { "epoch": 0.28, "learning_rate": 1.999859237248126e-05, "loss": 0.1448, "step": 1535 }, { "epoch": 0.28, "learning_rate": 1.9998579888336592e-05, "loss": 0.1325, "step": 1536 }, { "epoch": 0.28, "learning_rate": 1.9998567349079515e-05, "loss": 0.1305, "step": 1537 }, { "epoch": 0.28, "learning_rate": 1.99985547547101e-05, "loss": 0.1418, "step": 1538 }, { "epoch": 0.28, "learning_rate": 1.9998542105228414e-05, "loss": 0.1369, "step": 1539 }, { "epoch": 0.28, "learning_rate": 1.9998529400634537e-05, "loss": 0.1264, "step": 1540 }, { "epoch": 0.28, "learning_rate": 1.9998516640928525e-05, "loss": 0.1197, "step": 1541 }, { "epoch": 0.28, "learning_rate": 1.999850382611046e-05, "loss": 0.1233, "step": 1542 }, { "epoch": 0.28, "learning_rate": 1.9998490956180404e-05, "loss": 0.1391, "step": 1543 }, { "epoch": 0.28, "learning_rate": 1.9998478031138434e-05, "loss": 0.126, "step": 1544 }, { "epoch": 0.28, "learning_rate": 1.9998465050984617e-05, "loss": 0.1209, "step": 1545 }, { "epoch": 0.28, "learning_rate": 1.9998452015719028e-05, "loss": 0.1173, "step": 1546 }, { "epoch": 0.28, "learning_rate": 1.9998438925341735e-05, "loss": 0.1318, "step": 1547 }, { "epoch": 0.28, "learning_rate": 1.9998425779852816e-05, "loss": 0.1356, "step": 1548 }, { "epoch": 0.28, "learning_rate": 1.999841257925234e-05, "loss": 0.1329, "step": 1549 }, { "epoch": 0.28, "learning_rate": 1.9998399323540374e-05, "loss": 0.1332, "step": 1550 }, { "epoch": 0.28, "learning_rate": 1.9998386012717e-05, "loss": 0.1263, "step": 1551 }, { "epoch": 0.28, "learning_rate": 1.9998372646782288e-05, "loss": 0.1279, "step": 1552 }, { "epoch": 0.28, "learning_rate": 1.999835922573631e-05, "loss": 0.1248, "step": 1553 }, { "epoch": 0.28, "learning_rate": 1.9998345749579145e-05, "loss": 0.1194, "step": 1554 }, { "epoch": 0.29, "learning_rate": 1.9998332218310862e-05, "loss": 0.1435, "step": 1555 }, { "epoch": 0.29, "learning_rate": 1.999831863193154e-05, "loss": 0.1368, "step": 1556 }, { "epoch": 0.29, "learning_rate": 1.9998304990441245e-05, "loss": 0.1393, "step": 1557 }, { "epoch": 0.29, "learning_rate": 1.9998291293840066e-05, "loss": 0.1251, "step": 1558 }, { "epoch": 0.29, "learning_rate": 1.9998277542128068e-05, "loss": 0.126, "step": 1559 }, { "epoch": 0.29, "learning_rate": 1.999826373530533e-05, "loss": 0.1305, "step": 1560 }, { "epoch": 0.29, "learning_rate": 1.999824987337193e-05, "loss": 0.1313, "step": 1561 }, { "epoch": 0.29, "learning_rate": 1.999823595632794e-05, "loss": 0.1303, "step": 1562 }, { "epoch": 0.29, "learning_rate": 1.999822198417344e-05, "loss": 0.1238, "step": 1563 }, { "epoch": 0.29, "learning_rate": 1.9998207956908503e-05, "loss": 0.1373, "step": 1564 }, { "epoch": 0.29, "learning_rate": 1.9998193874533216e-05, "loss": 0.138, "step": 1565 }, { "epoch": 0.29, "learning_rate": 1.9998179737047647e-05, "loss": 0.1207, "step": 1566 }, { "epoch": 0.29, "learning_rate": 1.9998165544451876e-05, "loss": 0.1346, "step": 1567 }, { "epoch": 0.29, "learning_rate": 1.9998151296745983e-05, "loss": 0.1234, "step": 1568 }, { "epoch": 0.29, "learning_rate": 1.9998136993930045e-05, "loss": 0.1382, "step": 1569 }, { "epoch": 0.29, "learning_rate": 1.9998122636004146e-05, "loss": 0.1446, "step": 1570 }, { "epoch": 0.29, "learning_rate": 1.9998108222968355e-05, "loss": 0.1204, "step": 1571 }, { "epoch": 0.29, "learning_rate": 1.999809375482276e-05, "loss": 0.1355, "step": 1572 }, { "epoch": 0.29, "learning_rate": 1.9998079231567437e-05, "loss": 0.1248, "step": 1573 }, { "epoch": 0.29, "learning_rate": 1.999806465320247e-05, "loss": 0.1297, "step": 1574 }, { "epoch": 0.29, "learning_rate": 1.9998050019727933e-05, "loss": 0.1147, "step": 1575 }, { "epoch": 0.29, "learning_rate": 1.999803533114391e-05, "loss": 0.1198, "step": 1576 }, { "epoch": 0.29, "learning_rate": 1.9998020587450483e-05, "loss": 0.1251, "step": 1577 }, { "epoch": 0.29, "learning_rate": 1.9998005788647733e-05, "loss": 0.1293, "step": 1578 }, { "epoch": 0.29, "learning_rate": 1.999799093473574e-05, "loss": 0.1295, "step": 1579 }, { "epoch": 0.29, "learning_rate": 1.9997976025714587e-05, "loss": 0.1209, "step": 1580 }, { "epoch": 0.29, "learning_rate": 1.9997961061584358e-05, "loss": 0.1181, "step": 1581 }, { "epoch": 0.29, "learning_rate": 1.9997946042345128e-05, "loss": 0.1236, "step": 1582 }, { "epoch": 0.29, "learning_rate": 1.999793096799699e-05, "loss": 0.1382, "step": 1583 }, { "epoch": 0.29, "learning_rate": 1.9997915838540024e-05, "loss": 0.1198, "step": 1584 }, { "epoch": 0.29, "learning_rate": 1.9997900653974308e-05, "loss": 0.1231, "step": 1585 }, { "epoch": 0.29, "learning_rate": 1.9997885414299928e-05, "loss": 0.1244, "step": 1586 }, { "epoch": 0.29, "learning_rate": 1.999787011951697e-05, "loss": 0.1311, "step": 1587 }, { "epoch": 0.29, "learning_rate": 1.999785476962552e-05, "loss": 0.115, "step": 1588 }, { "epoch": 0.29, "learning_rate": 1.999783936462566e-05, "loss": 0.1369, "step": 1589 }, { "epoch": 0.29, "learning_rate": 1.999782390451747e-05, "loss": 0.129, "step": 1590 }, { "epoch": 0.29, "learning_rate": 1.999780838930105e-05, "loss": 0.1206, "step": 1591 }, { "epoch": 0.29, "learning_rate": 1.999779281897647e-05, "loss": 0.1294, "step": 1592 }, { "epoch": 0.29, "learning_rate": 1.9997777193543824e-05, "loss": 0.1279, "step": 1593 }, { "epoch": 0.29, "learning_rate": 1.9997761513003194e-05, "loss": 0.1297, "step": 1594 }, { "epoch": 0.29, "learning_rate": 1.9997745777354668e-05, "loss": 0.1132, "step": 1595 }, { "epoch": 0.29, "learning_rate": 1.9997729986598335e-05, "loss": 0.1352, "step": 1596 }, { "epoch": 0.29, "learning_rate": 1.9997714140734275e-05, "loss": 0.1391, "step": 1597 }, { "epoch": 0.29, "learning_rate": 1.9997698239762587e-05, "loss": 0.1273, "step": 1598 }, { "epoch": 0.29, "learning_rate": 1.999768228368335e-05, "loss": 0.1298, "step": 1599 }, { "epoch": 0.29, "learning_rate": 1.9997666272496652e-05, "loss": 0.1239, "step": 1600 }, { "epoch": 0.29, "learning_rate": 1.9997650206202585e-05, "loss": 0.135, "step": 1601 }, { "epoch": 0.29, "learning_rate": 1.9997634084801235e-05, "loss": 0.1293, "step": 1602 }, { "epoch": 0.29, "learning_rate": 1.9997617908292694e-05, "loss": 0.1308, "step": 1603 }, { "epoch": 0.29, "learning_rate": 1.9997601676677046e-05, "loss": 0.116, "step": 1604 }, { "epoch": 0.29, "learning_rate": 1.9997585389954384e-05, "loss": 0.1472, "step": 1605 }, { "epoch": 0.29, "learning_rate": 1.99975690481248e-05, "loss": 0.1169, "step": 1606 }, { "epoch": 0.29, "learning_rate": 1.9997552651188378e-05, "loss": 0.1418, "step": 1607 }, { "epoch": 0.29, "learning_rate": 1.999753619914521e-05, "loss": 0.1322, "step": 1608 }, { "epoch": 0.3, "learning_rate": 1.9997519691995392e-05, "loss": 0.129, "step": 1609 }, { "epoch": 0.3, "learning_rate": 1.9997503129739008e-05, "loss": 0.1294, "step": 1610 }, { "epoch": 0.3, "learning_rate": 1.9997486512376157e-05, "loss": 0.1274, "step": 1611 }, { "epoch": 0.3, "learning_rate": 1.999746983990692e-05, "loss": 0.1304, "step": 1612 }, { "epoch": 0.3, "learning_rate": 1.9997453112331403e-05, "loss": 0.125, "step": 1613 }, { "epoch": 0.3, "learning_rate": 1.9997436329649683e-05, "loss": 0.1411, "step": 1614 }, { "epoch": 0.3, "learning_rate": 1.9997419491861866e-05, "loss": 0.1313, "step": 1615 }, { "epoch": 0.3, "learning_rate": 1.9997402598968034e-05, "loss": 0.1311, "step": 1616 }, { "epoch": 0.3, "learning_rate": 1.9997385650968284e-05, "loss": 0.123, "step": 1617 }, { "epoch": 0.3, "learning_rate": 1.9997368647862713e-05, "loss": 0.148, "step": 1618 }, { "epoch": 0.3, "learning_rate": 1.999735158965141e-05, "loss": 0.1292, "step": 1619 }, { "epoch": 0.3, "learning_rate": 1.9997334476334472e-05, "loss": 0.127, "step": 1620 }, { "epoch": 0.3, "learning_rate": 1.999731730791199e-05, "loss": 0.1358, "step": 1621 }, { "epoch": 0.3, "learning_rate": 1.999730008438406e-05, "loss": 0.1226, "step": 1622 }, { "epoch": 0.3, "learning_rate": 1.999728280575078e-05, "loss": 0.1299, "step": 1623 }, { "epoch": 0.3, "learning_rate": 1.9997265472012247e-05, "loss": 0.1364, "step": 1624 }, { "epoch": 0.3, "learning_rate": 1.999724808316855e-05, "loss": 0.1309, "step": 1625 }, { "epoch": 0.3, "learning_rate": 1.9997230639219784e-05, "loss": 0.1373, "step": 1626 }, { "epoch": 0.3, "learning_rate": 1.999721314016605e-05, "loss": 0.138, "step": 1627 }, { "epoch": 0.3, "learning_rate": 1.9997195586007444e-05, "loss": 0.1194, "step": 1628 }, { "epoch": 0.3, "learning_rate": 1.9997177976744062e-05, "loss": 0.1438, "step": 1629 }, { "epoch": 0.3, "learning_rate": 1.9997160312376e-05, "loss": 0.1353, "step": 1630 }, { "epoch": 0.3, "learning_rate": 1.9997142592903355e-05, "loss": 0.1187, "step": 1631 }, { "epoch": 0.3, "learning_rate": 1.9997124818326226e-05, "loss": 0.1268, "step": 1632 }, { "epoch": 0.3, "learning_rate": 1.999710698864471e-05, "loss": 0.1245, "step": 1633 }, { "epoch": 0.3, "learning_rate": 1.999708910385891e-05, "loss": 0.1303, "step": 1634 }, { "epoch": 0.3, "learning_rate": 1.999707116396892e-05, "loss": 0.1398, "step": 1635 }, { "epoch": 0.3, "learning_rate": 1.9997053168974837e-05, "loss": 0.118, "step": 1636 }, { "epoch": 0.3, "learning_rate": 1.9997035118876762e-05, "loss": 0.1286, "step": 1637 }, { "epoch": 0.3, "learning_rate": 1.9997017013674796e-05, "loss": 0.1164, "step": 1638 }, { "epoch": 0.3, "learning_rate": 1.999699885336904e-05, "loss": 0.1262, "step": 1639 }, { "epoch": 0.3, "learning_rate": 1.9996980637959592e-05, "loss": 0.1255, "step": 1640 }, { "epoch": 0.3, "learning_rate": 1.999696236744655e-05, "loss": 0.1216, "step": 1641 }, { "epoch": 0.3, "learning_rate": 1.999694404183002e-05, "loss": 0.1233, "step": 1642 }, { "epoch": 0.3, "learning_rate": 1.99969256611101e-05, "loss": 0.121, "step": 1643 }, { "epoch": 0.3, "learning_rate": 1.999690722528689e-05, "loss": 0.1247, "step": 1644 }, { "epoch": 0.3, "learning_rate": 1.9996888734360495e-05, "loss": 0.1283, "step": 1645 }, { "epoch": 0.3, "learning_rate": 1.9996870188331013e-05, "loss": 0.1308, "step": 1646 }, { "epoch": 0.3, "learning_rate": 1.9996851587198548e-05, "loss": 0.1294, "step": 1647 }, { "epoch": 0.3, "learning_rate": 1.9996832930963205e-05, "loss": 0.1272, "step": 1648 }, { "epoch": 0.3, "learning_rate": 1.9996814219625082e-05, "loss": 0.1499, "step": 1649 }, { "epoch": 0.3, "learning_rate": 1.9996795453184288e-05, "loss": 0.13, "step": 1650 }, { "epoch": 0.3, "learning_rate": 1.999677663164092e-05, "loss": 0.1293, "step": 1651 }, { "epoch": 0.3, "learning_rate": 1.9996757754995086e-05, "loss": 0.1316, "step": 1652 }, { "epoch": 0.3, "learning_rate": 1.9996738823246892e-05, "loss": 0.1331, "step": 1653 }, { "epoch": 0.3, "learning_rate": 1.9996719836396436e-05, "loss": 0.1132, "step": 1654 }, { "epoch": 0.3, "learning_rate": 1.9996700794443827e-05, "loss": 0.1282, "step": 1655 }, { "epoch": 0.3, "learning_rate": 1.999668169738917e-05, "loss": 0.1283, "step": 1656 }, { "epoch": 0.3, "learning_rate": 1.9996662545232568e-05, "loss": 0.1236, "step": 1657 }, { "epoch": 0.3, "learning_rate": 1.999664333797413e-05, "loss": 0.1333, "step": 1658 }, { "epoch": 0.3, "learning_rate": 1.9996624075613957e-05, "loss": 0.1337, "step": 1659 }, { "epoch": 0.3, "learning_rate": 1.999660475815216e-05, "loss": 0.1237, "step": 1660 }, { "epoch": 0.3, "learning_rate": 1.9996585385588842e-05, "loss": 0.1136, "step": 1661 }, { "epoch": 0.3, "learning_rate": 1.999656595792411e-05, "loss": 0.1275, "step": 1662 }, { "epoch": 0.3, "learning_rate": 1.9996546475158073e-05, "loss": 0.1316, "step": 1663 }, { "epoch": 0.31, "learning_rate": 1.999652693729084e-05, "loss": 0.1338, "step": 1664 }, { "epoch": 0.31, "learning_rate": 1.9996507344322515e-05, "loss": 0.1252, "step": 1665 }, { "epoch": 0.31, "learning_rate": 1.9996487696253207e-05, "loss": 0.126, "step": 1666 }, { "epoch": 0.31, "learning_rate": 1.9996467993083024e-05, "loss": 0.1427, "step": 1667 }, { "epoch": 0.31, "learning_rate": 1.9996448234812077e-05, "loss": 0.1352, "step": 1668 }, { "epoch": 0.31, "learning_rate": 1.999642842144047e-05, "loss": 0.1352, "step": 1669 }, { "epoch": 0.31, "learning_rate": 1.999640855296832e-05, "loss": 0.1315, "step": 1670 }, { "epoch": 0.31, "learning_rate": 1.999638862939573e-05, "loss": 0.1269, "step": 1671 }, { "epoch": 0.31, "learning_rate": 1.999636865072281e-05, "loss": 0.1192, "step": 1672 }, { "epoch": 0.31, "learning_rate": 1.9996348616949673e-05, "loss": 0.1275, "step": 1673 }, { "epoch": 0.31, "learning_rate": 1.9996328528076427e-05, "loss": 0.1363, "step": 1674 }, { "epoch": 0.31, "learning_rate": 1.9996308384103186e-05, "loss": 0.1351, "step": 1675 }, { "epoch": 0.31, "learning_rate": 1.999628818503006e-05, "loss": 0.1367, "step": 1676 }, { "epoch": 0.31, "learning_rate": 1.9996267930857154e-05, "loss": 0.1348, "step": 1677 }, { "epoch": 0.31, "learning_rate": 1.999624762158459e-05, "loss": 0.1278, "step": 1678 }, { "epoch": 0.31, "learning_rate": 1.999622725721247e-05, "loss": 0.1299, "step": 1679 }, { "epoch": 0.31, "learning_rate": 1.9996206837740918e-05, "loss": 0.1262, "step": 1680 }, { "epoch": 0.31, "learning_rate": 1.9996186363170037e-05, "loss": 0.1367, "step": 1681 }, { "epoch": 0.31, "learning_rate": 1.999616583349994e-05, "loss": 0.1386, "step": 1682 }, { "epoch": 0.31, "learning_rate": 1.9996145248730743e-05, "loss": 0.116, "step": 1683 }, { "epoch": 0.31, "learning_rate": 1.9996124608862562e-05, "loss": 0.1366, "step": 1684 }, { "epoch": 0.31, "learning_rate": 1.9996103913895507e-05, "loss": 0.1495, "step": 1685 }, { "epoch": 0.31, "learning_rate": 1.999608316382969e-05, "loss": 0.1331, "step": 1686 }, { "epoch": 0.31, "learning_rate": 1.9996062358665235e-05, "loss": 0.1193, "step": 1687 }, { "epoch": 0.31, "learning_rate": 1.9996041498402245e-05, "loss": 0.131, "step": 1688 }, { "epoch": 0.31, "learning_rate": 1.999602058304084e-05, "loss": 0.1212, "step": 1689 }, { "epoch": 0.31, "learning_rate": 1.9995999612581136e-05, "loss": 0.1369, "step": 1690 }, { "epoch": 0.31, "learning_rate": 1.999597858702325e-05, "loss": 0.1317, "step": 1691 }, { "epoch": 0.31, "learning_rate": 1.9995957506367292e-05, "loss": 0.1382, "step": 1692 }, { "epoch": 0.31, "learning_rate": 1.9995936370613386e-05, "loss": 0.1289, "step": 1693 }, { "epoch": 0.31, "learning_rate": 1.9995915179761646e-05, "loss": 0.1288, "step": 1694 }, { "epoch": 0.31, "learning_rate": 1.999589393381218e-05, "loss": 0.1236, "step": 1695 }, { "epoch": 0.31, "learning_rate": 1.9995872632765117e-05, "loss": 0.1118, "step": 1696 }, { "epoch": 0.31, "learning_rate": 1.9995851276620573e-05, "loss": 0.1275, "step": 1697 }, { "epoch": 0.31, "learning_rate": 1.9995829865378656e-05, "loss": 0.1278, "step": 1698 }, { "epoch": 0.31, "learning_rate": 1.9995808399039496e-05, "loss": 0.1304, "step": 1699 }, { "epoch": 0.31, "learning_rate": 1.9995786877603204e-05, "loss": 0.1284, "step": 1700 }, { "epoch": 0.31, "learning_rate": 1.99957653010699e-05, "loss": 0.1257, "step": 1701 }, { "epoch": 0.31, "learning_rate": 1.9995743669439703e-05, "loss": 0.1429, "step": 1702 }, { "epoch": 0.31, "learning_rate": 1.9995721982712734e-05, "loss": 0.1115, "step": 1703 }, { "epoch": 0.31, "learning_rate": 1.9995700240889112e-05, "loss": 0.1367, "step": 1704 }, { "epoch": 0.31, "learning_rate": 1.9995678443968954e-05, "loss": 0.1206, "step": 1705 }, { "epoch": 0.31, "learning_rate": 1.9995656591952382e-05, "loss": 0.1272, "step": 1706 }, { "epoch": 0.31, "learning_rate": 1.999563468483952e-05, "loss": 0.1165, "step": 1707 }, { "epoch": 0.31, "learning_rate": 1.999561272263048e-05, "loss": 0.1368, "step": 1708 }, { "epoch": 0.31, "learning_rate": 1.9995590705325394e-05, "loss": 0.1189, "step": 1709 }, { "epoch": 0.31, "learning_rate": 1.9995568632924375e-05, "loss": 0.1373, "step": 1710 }, { "epoch": 0.31, "learning_rate": 1.999554650542755e-05, "loss": 0.1131, "step": 1711 }, { "epoch": 0.31, "learning_rate": 1.9995524322835035e-05, "loss": 0.1233, "step": 1712 }, { "epoch": 0.31, "learning_rate": 1.9995502085146958e-05, "loss": 0.1185, "step": 1713 }, { "epoch": 0.31, "learning_rate": 1.999547979236344e-05, "loss": 0.1376, "step": 1714 }, { "epoch": 0.31, "learning_rate": 1.99954574444846e-05, "loss": 0.1159, "step": 1715 }, { "epoch": 0.31, "learning_rate": 1.9995435041510568e-05, "loss": 0.1188, "step": 1716 }, { "epoch": 0.31, "learning_rate": 1.999541258344146e-05, "loss": 0.1344, "step": 1717 }, { "epoch": 0.32, "learning_rate": 1.9995390070277406e-05, "loss": 0.1258, "step": 1718 }, { "epoch": 0.32, "learning_rate": 1.9995367502018526e-05, "loss": 0.1414, "step": 1719 }, { "epoch": 0.32, "learning_rate": 1.999534487866495e-05, "loss": 0.1134, "step": 1720 }, { "epoch": 0.32, "learning_rate": 1.9995322200216798e-05, "loss": 0.1147, "step": 1721 }, { "epoch": 0.32, "learning_rate": 1.9995299466674194e-05, "loss": 0.125, "step": 1722 }, { "epoch": 0.32, "learning_rate": 1.9995276678037266e-05, "loss": 0.1246, "step": 1723 }, { "epoch": 0.32, "learning_rate": 1.9995253834306138e-05, "loss": 0.1287, "step": 1724 }, { "epoch": 0.32, "learning_rate": 1.9995230935480937e-05, "loss": 0.1309, "step": 1725 }, { "epoch": 0.32, "learning_rate": 1.999520798156179e-05, "loss": 0.1248, "step": 1726 }, { "epoch": 0.32, "learning_rate": 1.999518497254882e-05, "loss": 0.127, "step": 1727 }, { "epoch": 0.32, "learning_rate": 1.999516190844216e-05, "loss": 0.1402, "step": 1728 }, { "epoch": 0.32, "learning_rate": 1.999513878924193e-05, "loss": 0.123, "step": 1729 }, { "epoch": 0.32, "learning_rate": 1.9995115614948263e-05, "loss": 0.1183, "step": 1730 }, { "epoch": 0.32, "learning_rate": 1.9995092385561282e-05, "loss": 0.1393, "step": 1731 }, { "epoch": 0.32, "learning_rate": 1.999506910108112e-05, "loss": 0.1491, "step": 1732 }, { "epoch": 0.32, "learning_rate": 1.9995045761507902e-05, "loss": 0.1271, "step": 1733 }, { "epoch": 0.32, "learning_rate": 1.9995022366841758e-05, "loss": 0.1164, "step": 1734 }, { "epoch": 0.32, "learning_rate": 1.9994998917082818e-05, "loss": 0.1322, "step": 1735 }, { "epoch": 0.32, "learning_rate": 1.999497541223121e-05, "loss": 0.1361, "step": 1736 }, { "epoch": 0.32, "learning_rate": 1.999495185228706e-05, "loss": 0.1297, "step": 1737 }, { "epoch": 0.32, "learning_rate": 1.99949282372505e-05, "loss": 0.122, "step": 1738 }, { "epoch": 0.32, "learning_rate": 1.9994904567121665e-05, "loss": 0.119, "step": 1739 }, { "epoch": 0.32, "learning_rate": 1.9994880841900682e-05, "loss": 0.1404, "step": 1740 }, { "epoch": 0.32, "learning_rate": 1.999485706158768e-05, "loss": 0.1443, "step": 1741 }, { "epoch": 0.32, "learning_rate": 1.9994833226182795e-05, "loss": 0.1377, "step": 1742 }, { "epoch": 0.32, "learning_rate": 1.9994809335686152e-05, "loss": 0.1301, "step": 1743 }, { "epoch": 0.32, "learning_rate": 1.9994785390097882e-05, "loss": 0.1304, "step": 1744 }, { "epoch": 0.32, "learning_rate": 1.999476138941813e-05, "loss": 0.109, "step": 1745 }, { "epoch": 0.32, "learning_rate": 1.999473733364701e-05, "loss": 0.126, "step": 1746 }, { "epoch": 0.32, "learning_rate": 1.999471322278467e-05, "loss": 0.1157, "step": 1747 }, { "epoch": 0.32, "learning_rate": 1.9994689056831235e-05, "loss": 0.1328, "step": 1748 }, { "epoch": 0.32, "learning_rate": 1.999466483578684e-05, "loss": 0.1326, "step": 1749 }, { "epoch": 0.32, "learning_rate": 1.999464055965162e-05, "loss": 0.1324, "step": 1750 }, { "epoch": 0.32, "learning_rate": 1.9994616228425704e-05, "loss": 0.1258, "step": 1751 }, { "epoch": 0.32, "learning_rate": 1.9994591842109225e-05, "loss": 0.1228, "step": 1752 }, { "epoch": 0.32, "learning_rate": 1.999456740070233e-05, "loss": 0.1326, "step": 1753 }, { "epoch": 0.32, "learning_rate": 1.999454290420514e-05, "loss": 0.1251, "step": 1754 }, { "epoch": 0.32, "learning_rate": 1.99945183526178e-05, "loss": 0.1251, "step": 1755 }, { "epoch": 0.32, "learning_rate": 1.9994493745940436e-05, "loss": 0.1369, "step": 1756 }, { "epoch": 0.32, "learning_rate": 1.999446908417319e-05, "loss": 0.1311, "step": 1757 }, { "epoch": 0.32, "learning_rate": 1.9994444367316193e-05, "loss": 0.1243, "step": 1758 }, { "epoch": 0.32, "learning_rate": 1.999441959536959e-05, "loss": 0.1322, "step": 1759 }, { "epoch": 0.32, "learning_rate": 1.999439476833351e-05, "loss": 0.1207, "step": 1760 }, { "epoch": 0.32, "learning_rate": 1.999436988620809e-05, "loss": 0.144, "step": 1761 }, { "epoch": 0.32, "learning_rate": 1.9994344948993474e-05, "loss": 0.1229, "step": 1762 }, { "epoch": 0.32, "learning_rate": 1.999431995668979e-05, "loss": 0.1211, "step": 1763 }, { "epoch": 0.32, "learning_rate": 1.999429490929718e-05, "loss": 0.1396, "step": 1764 }, { "epoch": 0.32, "learning_rate": 1.9994269806815783e-05, "loss": 0.1142, "step": 1765 }, { "epoch": 0.32, "learning_rate": 1.999424464924574e-05, "loss": 0.1341, "step": 1766 }, { "epoch": 0.32, "learning_rate": 1.9994219436587182e-05, "loss": 0.1368, "step": 1767 }, { "epoch": 0.32, "learning_rate": 1.9994194168840256e-05, "loss": 0.1029, "step": 1768 }, { "epoch": 0.32, "learning_rate": 1.9994168846005095e-05, "loss": 0.1389, "step": 1769 }, { "epoch": 0.32, "learning_rate": 1.9994143468081845e-05, "loss": 0.1102, "step": 1770 }, { "epoch": 0.32, "learning_rate": 1.999411803507064e-05, "loss": 0.1254, "step": 1771 }, { "epoch": 0.32, "learning_rate": 1.999409254697162e-05, "loss": 0.1198, "step": 1772 }, { "epoch": 0.33, "learning_rate": 1.9994067003784928e-05, "loss": 0.1297, "step": 1773 }, { "epoch": 0.33, "learning_rate": 1.9994041405510705e-05, "loss": 0.1299, "step": 1774 }, { "epoch": 0.33, "learning_rate": 1.9994015752149095e-05, "loss": 0.1258, "step": 1775 }, { "epoch": 0.33, "learning_rate": 1.999399004370023e-05, "loss": 0.1241, "step": 1776 }, { "epoch": 0.33, "learning_rate": 1.9993964280164265e-05, "loss": 0.1256, "step": 1777 }, { "epoch": 0.33, "learning_rate": 1.999393846154133e-05, "loss": 0.1257, "step": 1778 }, { "epoch": 0.33, "learning_rate": 1.9993912587831573e-05, "loss": 0.1278, "step": 1779 }, { "epoch": 0.33, "learning_rate": 1.9993886659035137e-05, "loss": 0.13, "step": 1780 }, { "epoch": 0.33, "learning_rate": 1.999386067515216e-05, "loss": 0.1255, "step": 1781 }, { "epoch": 0.33, "learning_rate": 1.999383463618279e-05, "loss": 0.1233, "step": 1782 }, { "epoch": 0.33, "learning_rate": 1.999380854212717e-05, "loss": 0.1286, "step": 1783 }, { "epoch": 0.33, "learning_rate": 1.9993782392985444e-05, "loss": 0.1047, "step": 1784 }, { "epoch": 0.33, "learning_rate": 1.9993756188757756e-05, "loss": 0.1312, "step": 1785 }, { "epoch": 0.33, "learning_rate": 1.9993729929444246e-05, "loss": 0.1177, "step": 1786 }, { "epoch": 0.33, "learning_rate": 1.999370361504507e-05, "loss": 0.1284, "step": 1787 }, { "epoch": 0.33, "learning_rate": 1.9993677245560356e-05, "loss": 0.118, "step": 1788 }, { "epoch": 0.33, "learning_rate": 1.9993650820990263e-05, "loss": 0.1221, "step": 1789 }, { "epoch": 0.33, "learning_rate": 1.9993624341334935e-05, "loss": 0.1201, "step": 1790 }, { "epoch": 0.33, "learning_rate": 1.9993597806594517e-05, "loss": 0.1389, "step": 1791 }, { "epoch": 0.33, "learning_rate": 1.9993571216769147e-05, "loss": 0.1225, "step": 1792 }, { "epoch": 0.33, "learning_rate": 1.999354457185898e-05, "loss": 0.1366, "step": 1793 }, { "epoch": 0.33, "learning_rate": 1.9993517871864166e-05, "loss": 0.1151, "step": 1794 }, { "epoch": 0.33, "learning_rate": 1.9993491116784843e-05, "loss": 0.1302, "step": 1795 }, { "epoch": 0.33, "learning_rate": 1.9993464306621166e-05, "loss": 0.1348, "step": 1796 }, { "epoch": 0.33, "learning_rate": 1.9993437441373275e-05, "loss": 0.1218, "step": 1797 }, { "epoch": 0.33, "learning_rate": 1.9993410521041328e-05, "loss": 0.1348, "step": 1798 }, { "epoch": 0.33, "learning_rate": 1.9993383545625466e-05, "loss": 0.1209, "step": 1799 }, { "epoch": 0.33, "learning_rate": 1.999335651512584e-05, "loss": 0.1187, "step": 1800 }, { "epoch": 0.33, "learning_rate": 1.9993329429542594e-05, "loss": 0.1251, "step": 1801 }, { "epoch": 0.33, "learning_rate": 1.999330228887589e-05, "loss": 0.1247, "step": 1802 }, { "epoch": 0.33, "learning_rate": 1.9993275093125867e-05, "loss": 0.1254, "step": 1803 }, { "epoch": 0.33, "learning_rate": 1.9993247842292672e-05, "loss": 0.1284, "step": 1804 }, { "epoch": 0.33, "learning_rate": 1.9993220536376467e-05, "loss": 0.1273, "step": 1805 }, { "epoch": 0.33, "learning_rate": 1.9993193175377394e-05, "loss": 0.1286, "step": 1806 }, { "epoch": 0.33, "learning_rate": 1.9993165759295605e-05, "loss": 0.1336, "step": 1807 }, { "epoch": 0.33, "learning_rate": 1.9993138288131255e-05, "loss": 0.1245, "step": 1808 }, { "epoch": 0.33, "learning_rate": 1.999311076188449e-05, "loss": 0.1198, "step": 1809 }, { "epoch": 0.33, "learning_rate": 1.9993083180555468e-05, "loss": 0.133, "step": 1810 }, { "epoch": 0.33, "learning_rate": 1.9993055544144337e-05, "loss": 0.1321, "step": 1811 }, { "epoch": 0.33, "learning_rate": 1.9993027852651246e-05, "loss": 0.1338, "step": 1812 }, { "epoch": 0.33, "learning_rate": 1.999300010607635e-05, "loss": 0.1283, "step": 1813 }, { "epoch": 0.33, "learning_rate": 1.9992972304419813e-05, "loss": 0.1136, "step": 1814 }, { "epoch": 0.33, "learning_rate": 1.999294444768177e-05, "loss": 0.1239, "step": 1815 }, { "epoch": 0.33, "learning_rate": 1.9992916535862387e-05, "loss": 0.1396, "step": 1816 }, { "epoch": 0.33, "learning_rate": 1.999288856896181e-05, "loss": 0.1241, "step": 1817 }, { "epoch": 0.33, "learning_rate": 1.9992860546980196e-05, "loss": 0.1188, "step": 1818 }, { "epoch": 0.33, "learning_rate": 1.9992832469917708e-05, "loss": 0.1292, "step": 1819 }, { "epoch": 0.33, "learning_rate": 1.9992804337774486e-05, "loss": 0.1259, "step": 1820 }, { "epoch": 0.33, "learning_rate": 1.9992776150550698e-05, "loss": 0.1245, "step": 1821 }, { "epoch": 0.33, "learning_rate": 1.999274790824649e-05, "loss": 0.1287, "step": 1822 }, { "epoch": 0.33, "learning_rate": 1.9992719610862023e-05, "loss": 0.124, "step": 1823 }, { "epoch": 0.33, "learning_rate": 1.999269125839745e-05, "loss": 0.1231, "step": 1824 }, { "epoch": 0.33, "learning_rate": 1.9992662850852932e-05, "loss": 0.1259, "step": 1825 }, { "epoch": 0.33, "learning_rate": 1.999263438822862e-05, "loss": 0.1277, "step": 1826 }, { "epoch": 0.34, "learning_rate": 1.9992605870524674e-05, "loss": 0.1166, "step": 1827 }, { "epoch": 0.34, "learning_rate": 1.999257729774125e-05, "loss": 0.1229, "step": 1828 }, { "epoch": 0.34, "learning_rate": 1.9992548669878503e-05, "loss": 0.1156, "step": 1829 }, { "epoch": 0.34, "learning_rate": 1.9992519986936597e-05, "loss": 0.1254, "step": 1830 }, { "epoch": 0.34, "learning_rate": 1.9992491248915686e-05, "loss": 0.1304, "step": 1831 }, { "epoch": 0.34, "learning_rate": 1.999246245581593e-05, "loss": 0.127, "step": 1832 }, { "epoch": 0.34, "learning_rate": 1.999243360763748e-05, "loss": 0.1188, "step": 1833 }, { "epoch": 0.34, "learning_rate": 1.9992404704380513e-05, "loss": 0.1166, "step": 1834 }, { "epoch": 0.34, "learning_rate": 1.999237574604517e-05, "loss": 0.1185, "step": 1835 }, { "epoch": 0.34, "learning_rate": 1.9992346732631618e-05, "loss": 0.1163, "step": 1836 }, { "epoch": 0.34, "learning_rate": 1.999231766414002e-05, "loss": 0.1333, "step": 1837 }, { "epoch": 0.34, "learning_rate": 1.9992288540570534e-05, "loss": 0.1267, "step": 1838 }, { "epoch": 0.34, "learning_rate": 1.9992259361923314e-05, "loss": 0.1251, "step": 1839 }, { "epoch": 0.34, "learning_rate": 1.9992230128198528e-05, "loss": 0.1296, "step": 1840 }, { "epoch": 0.34, "learning_rate": 1.9992200839396337e-05, "loss": 0.1274, "step": 1841 }, { "epoch": 0.34, "learning_rate": 1.9992171495516902e-05, "loss": 0.1257, "step": 1842 }, { "epoch": 0.34, "learning_rate": 1.999214209656038e-05, "loss": 0.1205, "step": 1843 }, { "epoch": 0.34, "learning_rate": 1.999211264252694e-05, "loss": 0.1257, "step": 1844 }, { "epoch": 0.34, "learning_rate": 1.999208313341674e-05, "loss": 0.1194, "step": 1845 }, { "epoch": 0.34, "learning_rate": 1.9992053569229945e-05, "loss": 0.1097, "step": 1846 }, { "epoch": 0.34, "learning_rate": 1.999202394996671e-05, "loss": 0.1235, "step": 1847 }, { "epoch": 0.34, "learning_rate": 1.9991994275627213e-05, "loss": 0.1228, "step": 1848 }, { "epoch": 0.34, "learning_rate": 1.9991964546211604e-05, "loss": 0.1342, "step": 1849 }, { "epoch": 0.34, "learning_rate": 1.9991934761720057e-05, "loss": 0.123, "step": 1850 }, { "epoch": 0.34, "learning_rate": 1.9991904922152727e-05, "loss": 0.1264, "step": 1851 }, { "epoch": 0.34, "learning_rate": 1.9991875027509786e-05, "loss": 0.1214, "step": 1852 }, { "epoch": 0.34, "learning_rate": 1.9991845077791394e-05, "loss": 0.1388, "step": 1853 }, { "epoch": 0.34, "learning_rate": 1.9991815072997718e-05, "loss": 0.1174, "step": 1854 }, { "epoch": 0.34, "learning_rate": 1.9991785013128922e-05, "loss": 0.1197, "step": 1855 }, { "epoch": 0.34, "learning_rate": 1.9991754898185177e-05, "loss": 0.1175, "step": 1856 }, { "epoch": 0.34, "learning_rate": 1.9991724728166645e-05, "loss": 0.1309, "step": 1857 }, { "epoch": 0.34, "learning_rate": 1.9991694503073488e-05, "loss": 0.1253, "step": 1858 }, { "epoch": 0.34, "learning_rate": 1.999166422290588e-05, "loss": 0.1293, "step": 1859 }, { "epoch": 0.34, "learning_rate": 1.9991633887663983e-05, "loss": 0.1237, "step": 1860 }, { "epoch": 0.34, "learning_rate": 1.9991603497347967e-05, "loss": 0.1209, "step": 1861 }, { "epoch": 0.34, "learning_rate": 1.9991573051957997e-05, "loss": 0.1038, "step": 1862 }, { "epoch": 0.34, "learning_rate": 1.9991542551494243e-05, "loss": 0.1297, "step": 1863 }, { "epoch": 0.34, "learning_rate": 1.9991511995956874e-05, "loss": 0.1426, "step": 1864 }, { "epoch": 0.34, "learning_rate": 1.9991481385346057e-05, "loss": 0.1292, "step": 1865 }, { "epoch": 0.34, "learning_rate": 1.9991450719661957e-05, "loss": 0.1249, "step": 1866 }, { "epoch": 0.34, "learning_rate": 1.999141999890475e-05, "loss": 0.1303, "step": 1867 }, { "epoch": 0.34, "learning_rate": 1.9991389223074597e-05, "loss": 0.1215, "step": 1868 }, { "epoch": 0.34, "learning_rate": 1.999135839217168e-05, "loss": 0.1214, "step": 1869 }, { "epoch": 0.34, "learning_rate": 1.9991327506196156e-05, "loss": 0.1143, "step": 1870 }, { "epoch": 0.34, "learning_rate": 1.99912965651482e-05, "loss": 0.1317, "step": 1871 }, { "epoch": 0.34, "learning_rate": 1.9991265569027986e-05, "loss": 0.1258, "step": 1872 }, { "epoch": 0.34, "learning_rate": 1.9991234517835682e-05, "loss": 0.1332, "step": 1873 }, { "epoch": 0.34, "learning_rate": 1.9991203411571457e-05, "loss": 0.1131, "step": 1874 }, { "epoch": 0.34, "learning_rate": 1.999117225023549e-05, "loss": 0.118, "step": 1875 }, { "epoch": 0.34, "learning_rate": 1.9991141033827943e-05, "loss": 0.1285, "step": 1876 }, { "epoch": 0.34, "learning_rate": 1.9991109762348992e-05, "loss": 0.1159, "step": 1877 }, { "epoch": 0.34, "learning_rate": 1.9991078435798813e-05, "loss": 0.1198, "step": 1878 }, { "epoch": 0.34, "learning_rate": 1.9991047054177573e-05, "loss": 0.1228, "step": 1879 }, { "epoch": 0.34, "learning_rate": 1.9991015617485446e-05, "loss": 0.1336, "step": 1880 }, { "epoch": 0.34, "learning_rate": 1.9990984125722612e-05, "loss": 0.1384, "step": 1881 }, { "epoch": 0.35, "learning_rate": 1.999095257888924e-05, "loss": 0.1355, "step": 1882 }, { "epoch": 0.35, "learning_rate": 1.9990920976985498e-05, "loss": 0.1213, "step": 1883 }, { "epoch": 0.35, "learning_rate": 1.9990889320011567e-05, "loss": 0.1288, "step": 1884 }, { "epoch": 0.35, "learning_rate": 1.999085760796762e-05, "loss": 0.1151, "step": 1885 }, { "epoch": 0.35, "learning_rate": 1.9990825840853835e-05, "loss": 0.1298, "step": 1886 }, { "epoch": 0.35, "learning_rate": 1.999079401867038e-05, "loss": 0.1143, "step": 1887 }, { "epoch": 0.35, "learning_rate": 1.9990762141417434e-05, "loss": 0.1371, "step": 1888 }, { "epoch": 0.35, "learning_rate": 1.9990730209095176e-05, "loss": 0.1331, "step": 1889 }, { "epoch": 0.35, "learning_rate": 1.999069822170378e-05, "loss": 0.1302, "step": 1890 }, { "epoch": 0.35, "learning_rate": 1.9990666179243415e-05, "loss": 0.13, "step": 1891 }, { "epoch": 0.35, "learning_rate": 1.9990634081714272e-05, "loss": 0.1323, "step": 1892 }, { "epoch": 0.35, "learning_rate": 1.9990601929116512e-05, "loss": 0.1292, "step": 1893 }, { "epoch": 0.35, "learning_rate": 1.9990569721450324e-05, "loss": 0.1062, "step": 1894 }, { "epoch": 0.35, "learning_rate": 1.9990537458715882e-05, "loss": 0.1259, "step": 1895 }, { "epoch": 0.35, "learning_rate": 1.9990505140913365e-05, "loss": 0.132, "step": 1896 }, { "epoch": 0.35, "learning_rate": 1.999047276804295e-05, "loss": 0.1113, "step": 1897 }, { "epoch": 0.35, "learning_rate": 1.999044034010481e-05, "loss": 0.1211, "step": 1898 }, { "epoch": 0.35, "learning_rate": 1.999040785709913e-05, "loss": 0.1089, "step": 1899 }, { "epoch": 0.35, "learning_rate": 1.9990375319026092e-05, "loss": 0.122, "step": 1900 }, { "epoch": 0.35, "learning_rate": 1.999034272588587e-05, "loss": 0.1164, "step": 1901 }, { "epoch": 0.35, "learning_rate": 1.9990310077678642e-05, "loss": 0.119, "step": 1902 }, { "epoch": 0.35, "learning_rate": 1.9990277374404593e-05, "loss": 0.1163, "step": 1903 }, { "epoch": 0.35, "learning_rate": 1.9990244616063902e-05, "loss": 0.1332, "step": 1904 }, { "epoch": 0.35, "learning_rate": 1.9990211802656746e-05, "loss": 0.1155, "step": 1905 }, { "epoch": 0.35, "learning_rate": 1.9990178934183307e-05, "loss": 0.1138, "step": 1906 }, { "epoch": 0.35, "learning_rate": 1.9990146010643768e-05, "loss": 0.1271, "step": 1907 }, { "epoch": 0.35, "learning_rate": 1.9990113032038316e-05, "loss": 0.1228, "step": 1908 }, { "epoch": 0.35, "learning_rate": 1.9990079998367123e-05, "loss": 0.1162, "step": 1909 }, { "epoch": 0.35, "learning_rate": 1.9990046909630375e-05, "loss": 0.1227, "step": 1910 }, { "epoch": 0.35, "learning_rate": 1.9990013765828253e-05, "loss": 0.1231, "step": 1911 }, { "epoch": 0.35, "learning_rate": 1.9989980566960946e-05, "loss": 0.1243, "step": 1912 }, { "epoch": 0.35, "learning_rate": 1.9989947313028627e-05, "loss": 0.1251, "step": 1913 }, { "epoch": 0.35, "learning_rate": 1.998991400403149e-05, "loss": 0.1335, "step": 1914 }, { "epoch": 0.35, "learning_rate": 1.9989880639969704e-05, "loss": 0.1129, "step": 1915 }, { "epoch": 0.35, "learning_rate": 1.9989847220843466e-05, "loss": 0.125, "step": 1916 }, { "epoch": 0.35, "learning_rate": 1.998981374665296e-05, "loss": 0.1136, "step": 1917 }, { "epoch": 0.35, "learning_rate": 1.9989780217398363e-05, "loss": 0.1348, "step": 1918 }, { "epoch": 0.35, "learning_rate": 1.9989746633079864e-05, "loss": 0.1438, "step": 1919 }, { "epoch": 0.35, "learning_rate": 1.9989712993697646e-05, "loss": 0.1172, "step": 1920 }, { "epoch": 0.35, "learning_rate": 1.9989679299251897e-05, "loss": 0.119, "step": 1921 }, { "epoch": 0.35, "learning_rate": 1.9989645549742803e-05, "loss": 0.1284, "step": 1922 }, { "epoch": 0.35, "learning_rate": 1.998961174517055e-05, "loss": 0.1188, "step": 1923 }, { "epoch": 0.35, "learning_rate": 1.998957788553532e-05, "loss": 0.1236, "step": 1924 }, { "epoch": 0.35, "learning_rate": 1.9989543970837305e-05, "loss": 0.1176, "step": 1925 }, { "epoch": 0.35, "learning_rate": 1.9989510001076687e-05, "loss": 0.1202, "step": 1926 }, { "epoch": 0.35, "learning_rate": 1.9989475976253655e-05, "loss": 0.1253, "step": 1927 }, { "epoch": 0.35, "learning_rate": 1.9989441896368403e-05, "loss": 0.1243, "step": 1928 }, { "epoch": 0.35, "learning_rate": 1.998940776142111e-05, "loss": 0.1179, "step": 1929 }, { "epoch": 0.35, "learning_rate": 1.9989373571411965e-05, "loss": 0.1116, "step": 1930 }, { "epoch": 0.35, "learning_rate": 1.9989339326341163e-05, "loss": 0.1281, "step": 1931 }, { "epoch": 0.35, "learning_rate": 1.9989305026208886e-05, "loss": 0.136, "step": 1932 }, { "epoch": 0.35, "learning_rate": 1.9989270671015326e-05, "loss": 0.1368, "step": 1933 }, { "epoch": 0.35, "learning_rate": 1.9989236260760672e-05, "loss": 0.1294, "step": 1934 }, { "epoch": 0.35, "learning_rate": 1.9989201795445115e-05, "loss": 0.1324, "step": 1935 }, { "epoch": 0.36, "learning_rate": 1.9989167275068842e-05, "loss": 0.1324, "step": 1936 }, { "epoch": 0.36, "learning_rate": 1.9989132699632046e-05, "loss": 0.1175, "step": 1937 }, { "epoch": 0.36, "learning_rate": 1.9989098069134917e-05, "loss": 0.1306, "step": 1938 }, { "epoch": 0.36, "learning_rate": 1.9989063383577644e-05, "loss": 0.1221, "step": 1939 }, { "epoch": 0.36, "learning_rate": 1.998902864296042e-05, "loss": 0.1274, "step": 1940 }, { "epoch": 0.36, "learning_rate": 1.9988993847283434e-05, "loss": 0.1289, "step": 1941 }, { "epoch": 0.36, "learning_rate": 1.9988958996546883e-05, "loss": 0.127, "step": 1942 }, { "epoch": 0.36, "learning_rate": 1.9988924090750957e-05, "loss": 0.1231, "step": 1943 }, { "epoch": 0.36, "learning_rate": 1.9988889129895844e-05, "loss": 0.1269, "step": 1944 }, { "epoch": 0.36, "learning_rate": 1.9988854113981743e-05, "loss": 0.1165, "step": 1945 }, { "epoch": 0.36, "learning_rate": 1.998881904300884e-05, "loss": 0.1116, "step": 1946 }, { "epoch": 0.36, "learning_rate": 1.9988783916977338e-05, "loss": 0.1144, "step": 1947 }, { "epoch": 0.36, "learning_rate": 1.9988748735887424e-05, "loss": 0.1348, "step": 1948 }, { "epoch": 0.36, "learning_rate": 1.998871349973929e-05, "loss": 0.1374, "step": 1949 }, { "epoch": 0.36, "learning_rate": 1.9988678208533132e-05, "loss": 0.1138, "step": 1950 }, { "epoch": 0.36, "learning_rate": 1.998864286226915e-05, "loss": 0.127, "step": 1951 }, { "epoch": 0.36, "learning_rate": 1.998860746094753e-05, "loss": 0.1183, "step": 1952 }, { "epoch": 0.36, "learning_rate": 1.9988572004568476e-05, "loss": 0.1273, "step": 1953 }, { "epoch": 0.36, "learning_rate": 1.9988536493132176e-05, "loss": 0.1122, "step": 1954 }, { "epoch": 0.36, "learning_rate": 1.9988500926638832e-05, "loss": 0.1246, "step": 1955 }, { "epoch": 0.36, "learning_rate": 1.9988465305088634e-05, "loss": 0.1236, "step": 1956 }, { "epoch": 0.36, "learning_rate": 1.9988429628481783e-05, "loss": 0.1217, "step": 1957 }, { "epoch": 0.36, "learning_rate": 1.998839389681847e-05, "loss": 0.1228, "step": 1958 }, { "epoch": 0.36, "learning_rate": 1.9988358110098903e-05, "loss": 0.1356, "step": 1959 }, { "epoch": 0.36, "learning_rate": 1.998832226832327e-05, "loss": 0.1125, "step": 1960 }, { "epoch": 0.36, "learning_rate": 1.998828637149177e-05, "loss": 0.1261, "step": 1961 }, { "epoch": 0.36, "learning_rate": 1.9988250419604597e-05, "loss": 0.1432, "step": 1962 }, { "epoch": 0.36, "learning_rate": 1.9988214412661958e-05, "loss": 0.1329, "step": 1963 }, { "epoch": 0.36, "learning_rate": 1.9988178350664043e-05, "loss": 0.1249, "step": 1964 }, { "epoch": 0.36, "learning_rate": 1.998814223361106e-05, "loss": 0.1097, "step": 1965 }, { "epoch": 0.36, "learning_rate": 1.99881060615032e-05, "loss": 0.1148, "step": 1966 }, { "epoch": 0.36, "learning_rate": 1.9988069834340668e-05, "loss": 0.1191, "step": 1967 }, { "epoch": 0.36, "learning_rate": 1.998803355212366e-05, "loss": 0.1162, "step": 1968 }, { "epoch": 0.36, "learning_rate": 1.9987997214852374e-05, "loss": 0.1174, "step": 1969 }, { "epoch": 0.36, "learning_rate": 1.9987960822527017e-05, "loss": 0.1216, "step": 1970 }, { "epoch": 0.36, "learning_rate": 1.9987924375147787e-05, "loss": 0.1361, "step": 1971 }, { "epoch": 0.36, "learning_rate": 1.998788787271488e-05, "loss": 0.1132, "step": 1972 }, { "epoch": 0.36, "learning_rate": 1.9987851315228503e-05, "loss": 0.1255, "step": 1973 }, { "epoch": 0.36, "learning_rate": 1.9987814702688853e-05, "loss": 0.1101, "step": 1974 }, { "epoch": 0.36, "learning_rate": 1.998777803509614e-05, "loss": 0.1256, "step": 1975 }, { "epoch": 0.36, "learning_rate": 1.9987741312450556e-05, "loss": 0.1232, "step": 1976 }, { "epoch": 0.36, "learning_rate": 1.998770453475231e-05, "loss": 0.1281, "step": 1977 }, { "epoch": 0.36, "learning_rate": 1.99876677020016e-05, "loss": 0.1337, "step": 1978 }, { "epoch": 0.36, "learning_rate": 1.9987630814198637e-05, "loss": 0.1322, "step": 1979 }, { "epoch": 0.36, "learning_rate": 1.9987593871343615e-05, "loss": 0.1211, "step": 1980 }, { "epoch": 0.36, "learning_rate": 1.998755687343674e-05, "loss": 0.1222, "step": 1981 }, { "epoch": 0.36, "learning_rate": 1.998751982047822e-05, "loss": 0.1159, "step": 1982 }, { "epoch": 0.36, "learning_rate": 1.9987482712468257e-05, "loss": 0.124, "step": 1983 }, { "epoch": 0.36, "learning_rate": 1.9987445549407055e-05, "loss": 0.1216, "step": 1984 }, { "epoch": 0.36, "learning_rate": 1.9987408331294816e-05, "loss": 0.1266, "step": 1985 }, { "epoch": 0.36, "learning_rate": 1.9987371058131754e-05, "loss": 0.12, "step": 1986 }, { "epoch": 0.36, "learning_rate": 1.9987333729918067e-05, "loss": 0.1245, "step": 1987 }, { "epoch": 0.36, "learning_rate": 1.998729634665396e-05, "loss": 0.1152, "step": 1988 }, { "epoch": 0.36, "learning_rate": 1.9987258908339643e-05, "loss": 0.1114, "step": 1989 }, { "epoch": 0.36, "learning_rate": 1.998722141497532e-05, "loss": 0.111, "step": 1990 }, { "epoch": 0.37, "learning_rate": 1.9987183866561198e-05, "loss": 0.1221, "step": 1991 }, { "epoch": 0.37, "learning_rate": 1.9987146263097486e-05, "loss": 0.1105, "step": 1992 }, { "epoch": 0.37, "learning_rate": 1.998710860458439e-05, "loss": 0.1284, "step": 1993 }, { "epoch": 0.37, "learning_rate": 1.998707089102212e-05, "loss": 0.1182, "step": 1994 }, { "epoch": 0.37, "learning_rate": 1.9987033122410878e-05, "loss": 0.1294, "step": 1995 }, { "epoch": 0.37, "learning_rate": 1.9986995298750875e-05, "loss": 0.109, "step": 1996 }, { "epoch": 0.37, "learning_rate": 1.998695742004232e-05, "loss": 0.1148, "step": 1997 }, { "epoch": 0.37, "learning_rate": 1.9986919486285422e-05, "loss": 0.1289, "step": 1998 }, { "epoch": 0.37, "learning_rate": 1.998688149748039e-05, "loss": 0.114, "step": 1999 }, { "epoch": 0.37, "learning_rate": 1.9986843453627435e-05, "loss": 0.1262, "step": 2000 }, { "epoch": 0.37, "learning_rate": 1.9986805354726763e-05, "loss": 0.1079, "step": 2001 }, { "epoch": 0.37, "learning_rate": 1.998676720077859e-05, "loss": 0.1186, "step": 2002 }, { "epoch": 0.37, "learning_rate": 1.9986728991783115e-05, "loss": 0.1276, "step": 2003 }, { "epoch": 0.37, "learning_rate": 1.998669072774056e-05, "loss": 0.129, "step": 2004 }, { "epoch": 0.37, "learning_rate": 1.9986652408651132e-05, "loss": 0.1281, "step": 2005 }, { "epoch": 0.37, "learning_rate": 1.9986614034515042e-05, "loss": 0.1253, "step": 2006 }, { "epoch": 0.37, "learning_rate": 1.99865756053325e-05, "loss": 0.1272, "step": 2007 }, { "epoch": 0.37, "learning_rate": 1.998653712110372e-05, "loss": 0.1181, "step": 2008 }, { "epoch": 0.37, "learning_rate": 1.9986498581828914e-05, "loss": 0.123, "step": 2009 }, { "epoch": 0.37, "learning_rate": 1.998645998750829e-05, "loss": 0.1163, "step": 2010 }, { "epoch": 0.37, "learning_rate": 1.998642133814207e-05, "loss": 0.1306, "step": 2011 }, { "epoch": 0.37, "learning_rate": 1.9986382633730457e-05, "loss": 0.142, "step": 2012 }, { "epoch": 0.37, "learning_rate": 1.9986343874273672e-05, "loss": 0.1192, "step": 2013 }, { "epoch": 0.37, "learning_rate": 1.9986305059771926e-05, "loss": 0.128, "step": 2014 }, { "epoch": 0.37, "learning_rate": 1.9986266190225435e-05, "loss": 0.1122, "step": 2015 }, { "epoch": 0.37, "learning_rate": 1.9986227265634407e-05, "loss": 0.1146, "step": 2016 }, { "epoch": 0.37, "learning_rate": 1.9986188285999057e-05, "loss": 0.1327, "step": 2017 }, { "epoch": 0.37, "learning_rate": 1.9986149251319608e-05, "loss": 0.1179, "step": 2018 }, { "epoch": 0.37, "learning_rate": 1.998611016159627e-05, "loss": 0.1243, "step": 2019 }, { "epoch": 0.37, "learning_rate": 1.998607101682926e-05, "loss": 0.1361, "step": 2020 }, { "epoch": 0.37, "learning_rate": 1.998603181701879e-05, "loss": 0.1237, "step": 2021 }, { "epoch": 0.37, "learning_rate": 1.9985992562165077e-05, "loss": 0.1186, "step": 2022 }, { "epoch": 0.37, "learning_rate": 1.9985953252268343e-05, "loss": 0.1178, "step": 2023 }, { "epoch": 0.37, "learning_rate": 1.9985913887328802e-05, "loss": 0.119, "step": 2024 }, { "epoch": 0.37, "learning_rate": 1.9985874467346665e-05, "loss": 0.1268, "step": 2025 }, { "epoch": 0.37, "learning_rate": 1.9985834992322158e-05, "loss": 0.1206, "step": 2026 }, { "epoch": 0.37, "learning_rate": 1.9985795462255492e-05, "loss": 0.1141, "step": 2027 }, { "epoch": 0.37, "learning_rate": 1.9985755877146887e-05, "loss": 0.1317, "step": 2028 }, { "epoch": 0.37, "learning_rate": 1.998571623699656e-05, "loss": 0.1192, "step": 2029 }, { "epoch": 0.37, "learning_rate": 1.9985676541804735e-05, "loss": 0.1232, "step": 2030 }, { "epoch": 0.37, "learning_rate": 1.9985636791571624e-05, "loss": 0.1178, "step": 2031 }, { "epoch": 0.37, "learning_rate": 1.9985596986297456e-05, "loss": 0.1263, "step": 2032 }, { "epoch": 0.37, "learning_rate": 1.998555712598244e-05, "loss": 0.1224, "step": 2033 }, { "epoch": 0.37, "learning_rate": 1.9985517210626794e-05, "loss": 0.1191, "step": 2034 }, { "epoch": 0.37, "learning_rate": 1.9985477240230745e-05, "loss": 0.1178, "step": 2035 }, { "epoch": 0.37, "learning_rate": 1.9985437214794516e-05, "loss": 0.1274, "step": 2036 }, { "epoch": 0.37, "learning_rate": 1.998539713431832e-05, "loss": 0.1263, "step": 2037 }, { "epoch": 0.37, "learning_rate": 1.9985356998802383e-05, "loss": 0.1304, "step": 2038 }, { "epoch": 0.37, "learning_rate": 1.9985316808246923e-05, "loss": 0.1242, "step": 2039 }, { "epoch": 0.37, "learning_rate": 1.998527656265216e-05, "loss": 0.1265, "step": 2040 }, { "epoch": 0.37, "learning_rate": 1.998523626201832e-05, "loss": 0.1223, "step": 2041 }, { "epoch": 0.37, "learning_rate": 1.9985195906345626e-05, "loss": 0.1217, "step": 2042 }, { "epoch": 0.37, "learning_rate": 1.99851554956343e-05, "loss": 0.1215, "step": 2043 }, { "epoch": 0.37, "learning_rate": 1.9985115029884556e-05, "loss": 0.1255, "step": 2044 }, { "epoch": 0.38, "learning_rate": 1.9985074509096632e-05, "loss": 0.1358, "step": 2045 }, { "epoch": 0.38, "learning_rate": 1.9985033933270737e-05, "loss": 0.1238, "step": 2046 }, { "epoch": 0.38, "learning_rate": 1.9984993302407103e-05, "loss": 0.135, "step": 2047 }, { "epoch": 0.38, "learning_rate": 1.9984952616505952e-05, "loss": 0.1311, "step": 2048 }, { "epoch": 0.38, "learning_rate": 1.998491187556751e-05, "loss": 0.128, "step": 2049 }, { "epoch": 0.38, "learning_rate": 1.9984871079592e-05, "loss": 0.1231, "step": 2050 }, { "epoch": 0.38, "learning_rate": 1.9984830228579645e-05, "loss": 0.1199, "step": 2051 }, { "epoch": 0.38, "learning_rate": 1.9984789322530673e-05, "loss": 0.1212, "step": 2052 }, { "epoch": 0.38, "learning_rate": 1.9984748361445306e-05, "loss": 0.1267, "step": 2053 }, { "epoch": 0.38, "learning_rate": 1.9984707345323773e-05, "loss": 0.1389, "step": 2054 }, { "epoch": 0.38, "learning_rate": 1.9984666274166302e-05, "loss": 0.1187, "step": 2055 }, { "epoch": 0.38, "learning_rate": 1.9984625147973115e-05, "loss": 0.115, "step": 2056 }, { "epoch": 0.38, "learning_rate": 1.998458396674444e-05, "loss": 0.1242, "step": 2057 }, { "epoch": 0.38, "learning_rate": 1.9984542730480502e-05, "loss": 0.1256, "step": 2058 }, { "epoch": 0.38, "learning_rate": 1.9984501439181532e-05, "loss": 0.114, "step": 2059 }, { "epoch": 0.38, "learning_rate": 1.9984460092847756e-05, "loss": 0.1321, "step": 2060 }, { "epoch": 0.38, "learning_rate": 1.9984418691479406e-05, "loss": 0.1157, "step": 2061 }, { "epoch": 0.38, "learning_rate": 1.9984377235076703e-05, "loss": 0.1262, "step": 2062 }, { "epoch": 0.38, "learning_rate": 1.9984335723639878e-05, "loss": 0.1253, "step": 2063 }, { "epoch": 0.38, "learning_rate": 1.998429415716916e-05, "loss": 0.1277, "step": 2064 }, { "epoch": 0.38, "learning_rate": 1.9984252535664784e-05, "loss": 0.1356, "step": 2065 }, { "epoch": 0.38, "learning_rate": 1.9984210859126967e-05, "loss": 0.1278, "step": 2066 }, { "epoch": 0.38, "learning_rate": 1.998416912755595e-05, "loss": 0.1256, "step": 2067 }, { "epoch": 0.38, "learning_rate": 1.998412734095196e-05, "loss": 0.1215, "step": 2068 }, { "epoch": 0.38, "learning_rate": 1.9984085499315222e-05, "loss": 0.1345, "step": 2069 }, { "epoch": 0.38, "learning_rate": 1.9984043602645973e-05, "loss": 0.1222, "step": 2070 }, { "epoch": 0.38, "learning_rate": 1.9984001650944444e-05, "loss": 0.1277, "step": 2071 }, { "epoch": 0.38, "learning_rate": 1.998395964421086e-05, "loss": 0.1195, "step": 2072 }, { "epoch": 0.38, "learning_rate": 1.998391758244546e-05, "loss": 0.1236, "step": 2073 }, { "epoch": 0.38, "learning_rate": 1.998387546564847e-05, "loss": 0.1319, "step": 2074 }, { "epoch": 0.38, "learning_rate": 1.9983833293820124e-05, "loss": 0.132, "step": 2075 }, { "epoch": 0.38, "learning_rate": 1.9983791066960657e-05, "loss": 0.1145, "step": 2076 }, { "epoch": 0.38, "learning_rate": 1.99837487850703e-05, "loss": 0.1248, "step": 2077 }, { "epoch": 0.38, "learning_rate": 1.9983706448149282e-05, "loss": 0.1278, "step": 2078 }, { "epoch": 0.38, "learning_rate": 1.9983664056197846e-05, "loss": 0.1202, "step": 2079 }, { "epoch": 0.38, "learning_rate": 1.9983621609216217e-05, "loss": 0.12, "step": 2080 }, { "epoch": 0.38, "learning_rate": 1.998357910720463e-05, "loss": 0.1284, "step": 2081 }, { "epoch": 0.38, "learning_rate": 1.9983536550163323e-05, "loss": 0.1133, "step": 2082 }, { "epoch": 0.38, "learning_rate": 1.9983493938092526e-05, "loss": 0.1208, "step": 2083 }, { "epoch": 0.38, "learning_rate": 1.998345127099248e-05, "loss": 0.1227, "step": 2084 }, { "epoch": 0.38, "learning_rate": 1.9983408548863415e-05, "loss": 0.1217, "step": 2085 }, { "epoch": 0.38, "learning_rate": 1.998336577170557e-05, "loss": 0.1155, "step": 2086 }, { "epoch": 0.38, "learning_rate": 1.9983322939519176e-05, "loss": 0.135, "step": 2087 }, { "epoch": 0.38, "learning_rate": 1.9983280052304475e-05, "loss": 0.111, "step": 2088 }, { "epoch": 0.38, "learning_rate": 1.9983237110061696e-05, "loss": 0.1163, "step": 2089 }, { "epoch": 0.38, "learning_rate": 1.9983194112791086e-05, "loss": 0.1331, "step": 2090 }, { "epoch": 0.38, "learning_rate": 1.9983151060492875e-05, "loss": 0.1146, "step": 2091 }, { "epoch": 0.38, "learning_rate": 1.99831079531673e-05, "loss": 0.1212, "step": 2092 }, { "epoch": 0.38, "learning_rate": 1.9983064790814598e-05, "loss": 0.114, "step": 2093 }, { "epoch": 0.38, "learning_rate": 1.9983021573435013e-05, "loss": 0.1196, "step": 2094 }, { "epoch": 0.38, "learning_rate": 1.998297830102878e-05, "loss": 0.1226, "step": 2095 }, { "epoch": 0.38, "learning_rate": 1.998293497359613e-05, "loss": 0.1148, "step": 2096 }, { "epoch": 0.38, "learning_rate": 1.998289159113731e-05, "loss": 0.1304, "step": 2097 }, { "epoch": 0.38, "learning_rate": 1.9982848153652562e-05, "loss": 0.114, "step": 2098 }, { "epoch": 0.38, "learning_rate": 1.998280466114212e-05, "loss": 0.1359, "step": 2099 }, { "epoch": 0.39, "learning_rate": 1.9982761113606225e-05, "loss": 0.109, "step": 2100 }, { "epoch": 0.39, "learning_rate": 1.9982717511045115e-05, "loss": 0.1027, "step": 2101 }, { "epoch": 0.39, "learning_rate": 1.9982673853459034e-05, "loss": 0.1216, "step": 2102 }, { "epoch": 0.39, "learning_rate": 1.998263014084822e-05, "loss": 0.1378, "step": 2103 }, { "epoch": 0.39, "learning_rate": 1.9982586373212914e-05, "loss": 0.1276, "step": 2104 }, { "epoch": 0.39, "learning_rate": 1.998254255055336e-05, "loss": 0.1166, "step": 2105 }, { "epoch": 0.39, "learning_rate": 1.9982498672869795e-05, "loss": 0.1137, "step": 2106 }, { "epoch": 0.39, "learning_rate": 1.9982454740162466e-05, "loss": 0.108, "step": 2107 }, { "epoch": 0.39, "learning_rate": 1.9982410752431613e-05, "loss": 0.1271, "step": 2108 }, { "epoch": 0.39, "learning_rate": 1.9982366709677475e-05, "loss": 0.123, "step": 2109 }, { "epoch": 0.39, "learning_rate": 1.99823226119003e-05, "loss": 0.1243, "step": 2110 }, { "epoch": 0.39, "learning_rate": 1.9982278459100328e-05, "loss": 0.1237, "step": 2111 }, { "epoch": 0.39, "learning_rate": 1.9982234251277805e-05, "loss": 0.1206, "step": 2112 }, { "epoch": 0.39, "learning_rate": 1.998218998843297e-05, "loss": 0.1227, "step": 2113 }, { "epoch": 0.39, "learning_rate": 1.998214567056607e-05, "loss": 0.1194, "step": 2114 }, { "epoch": 0.39, "learning_rate": 1.998210129767735e-05, "loss": 0.1138, "step": 2115 }, { "epoch": 0.39, "learning_rate": 1.998205686976706e-05, "loss": 0.114, "step": 2116 }, { "epoch": 0.39, "learning_rate": 1.9982012386835433e-05, "loss": 0.1232, "step": 2117 }, { "epoch": 0.39, "learning_rate": 1.998196784888272e-05, "loss": 0.1075, "step": 2118 }, { "epoch": 0.39, "learning_rate": 1.9981923255909167e-05, "loss": 0.1181, "step": 2119 }, { "epoch": 0.39, "learning_rate": 1.9981878607915017e-05, "loss": 0.1237, "step": 2120 }, { "epoch": 0.39, "learning_rate": 1.998183390490052e-05, "loss": 0.126, "step": 2121 }, { "epoch": 0.39, "learning_rate": 1.998178914686592e-05, "loss": 0.1201, "step": 2122 }, { "epoch": 0.39, "learning_rate": 1.9981744333811462e-05, "loss": 0.1371, "step": 2123 }, { "epoch": 0.39, "learning_rate": 1.99816994657374e-05, "loss": 0.1189, "step": 2124 }, { "epoch": 0.39, "learning_rate": 1.9981654542643976e-05, "loss": 0.1118, "step": 2125 }, { "epoch": 0.39, "learning_rate": 1.998160956453144e-05, "loss": 0.1171, "step": 2126 }, { "epoch": 0.39, "learning_rate": 1.998156453140003e-05, "loss": 0.1167, "step": 2127 }, { "epoch": 0.39, "learning_rate": 1.998151944325001e-05, "loss": 0.1251, "step": 2128 }, { "epoch": 0.39, "learning_rate": 1.998147430008162e-05, "loss": 0.1186, "step": 2129 }, { "epoch": 0.39, "learning_rate": 1.998142910189511e-05, "loss": 0.1337, "step": 2130 }, { "epoch": 0.39, "learning_rate": 1.9981383848690728e-05, "loss": 0.1308, "step": 2131 }, { "epoch": 0.39, "learning_rate": 1.9981338540468724e-05, "loss": 0.119, "step": 2132 }, { "epoch": 0.39, "learning_rate": 1.9981293177229347e-05, "loss": 0.1322, "step": 2133 }, { "epoch": 0.39, "learning_rate": 1.998124775897285e-05, "loss": 0.1241, "step": 2134 }, { "epoch": 0.39, "learning_rate": 1.998120228569948e-05, "loss": 0.1407, "step": 2135 }, { "epoch": 0.39, "learning_rate": 1.998115675740949e-05, "loss": 0.1152, "step": 2136 }, { "epoch": 0.39, "learning_rate": 1.998111117410313e-05, "loss": 0.1216, "step": 2137 }, { "epoch": 0.39, "learning_rate": 1.9981065535780654e-05, "loss": 0.1178, "step": 2138 }, { "epoch": 0.39, "learning_rate": 1.998101984244231e-05, "loss": 0.1181, "step": 2139 }, { "epoch": 0.39, "learning_rate": 1.9980974094088345e-05, "loss": 0.1077, "step": 2140 }, { "epoch": 0.39, "learning_rate": 1.998092829071902e-05, "loss": 0.121, "step": 2141 }, { "epoch": 0.39, "learning_rate": 1.998088243233459e-05, "loss": 0.1222, "step": 2142 }, { "epoch": 0.39, "learning_rate": 1.99808365189353e-05, "loss": 0.1268, "step": 2143 }, { "epoch": 0.39, "learning_rate": 1.99807905505214e-05, "loss": 0.1375, "step": 2144 }, { "epoch": 0.39, "learning_rate": 1.9980744527093154e-05, "loss": 0.12, "step": 2145 }, { "epoch": 0.39, "learning_rate": 1.9980698448650805e-05, "loss": 0.1069, "step": 2146 }, { "epoch": 0.39, "learning_rate": 1.9980652315194616e-05, "loss": 0.1114, "step": 2147 }, { "epoch": 0.39, "learning_rate": 1.9980606126724836e-05, "loss": 0.12, "step": 2148 }, { "epoch": 0.39, "learning_rate": 1.9980559883241723e-05, "loss": 0.1205, "step": 2149 }, { "epoch": 0.39, "learning_rate": 1.998051358474553e-05, "loss": 0.1143, "step": 2150 }, { "epoch": 0.39, "learning_rate": 1.9980467231236514e-05, "loss": 0.1314, "step": 2151 }, { "epoch": 0.39, "learning_rate": 1.9980420822714927e-05, "loss": 0.1171, "step": 2152 }, { "epoch": 0.39, "learning_rate": 1.9980374359181026e-05, "loss": 0.122, "step": 2153 }, { "epoch": 0.4, "learning_rate": 1.9980327840635066e-05, "loss": 0.1148, "step": 2154 }, { "epoch": 0.4, "learning_rate": 1.998028126707731e-05, "loss": 0.1249, "step": 2155 }, { "epoch": 0.4, "learning_rate": 1.9980234638508003e-05, "loss": 0.1282, "step": 2156 }, { "epoch": 0.4, "learning_rate": 1.9980187954927415e-05, "loss": 0.1209, "step": 2157 }, { "epoch": 0.4, "learning_rate": 1.9980141216335793e-05, "loss": 0.1178, "step": 2158 }, { "epoch": 0.4, "learning_rate": 1.9980094422733404e-05, "loss": 0.1179, "step": 2159 }, { "epoch": 0.4, "learning_rate": 1.9980047574120497e-05, "loss": 0.1267, "step": 2160 }, { "epoch": 0.4, "learning_rate": 1.9980000670497334e-05, "loss": 0.1421, "step": 2161 }, { "epoch": 0.4, "learning_rate": 1.997995371186417e-05, "loss": 0.1104, "step": 2162 }, { "epoch": 0.4, "learning_rate": 1.9979906698221274e-05, "loss": 0.1214, "step": 2163 }, { "epoch": 0.4, "learning_rate": 1.9979859629568896e-05, "loss": 0.1223, "step": 2164 }, { "epoch": 0.4, "learning_rate": 1.9979812505907298e-05, "loss": 0.1323, "step": 2165 }, { "epoch": 0.4, "learning_rate": 1.997976532723674e-05, "loss": 0.1243, "step": 2166 }, { "epoch": 0.4, "learning_rate": 1.997971809355748e-05, "loss": 0.1183, "step": 2167 }, { "epoch": 0.4, "learning_rate": 1.997967080486978e-05, "loss": 0.1215, "step": 2168 }, { "epoch": 0.4, "learning_rate": 1.9979623461173902e-05, "loss": 0.1452, "step": 2169 }, { "epoch": 0.4, "learning_rate": 1.9979576062470108e-05, "loss": 0.1179, "step": 2170 }, { "epoch": 0.4, "learning_rate": 1.9979528608758653e-05, "loss": 0.117, "step": 2171 }, { "epoch": 0.4, "learning_rate": 1.9979481100039804e-05, "loss": 0.1266, "step": 2172 }, { "epoch": 0.4, "learning_rate": 1.9979433536313823e-05, "loss": 0.1219, "step": 2173 }, { "epoch": 0.4, "learning_rate": 1.997938591758097e-05, "loss": 0.134, "step": 2174 }, { "epoch": 0.4, "learning_rate": 1.9979338243841505e-05, "loss": 0.1185, "step": 2175 }, { "epoch": 0.4, "learning_rate": 1.9979290515095697e-05, "loss": 0.123, "step": 2176 }, { "epoch": 0.4, "learning_rate": 1.9979242731343803e-05, "loss": 0.1183, "step": 2177 }, { "epoch": 0.4, "learning_rate": 1.997919489258609e-05, "loss": 0.109, "step": 2178 }, { "epoch": 0.4, "learning_rate": 1.9979146998822823e-05, "loss": 0.1126, "step": 2179 }, { "epoch": 0.4, "learning_rate": 1.9979099050054264e-05, "loss": 0.117, "step": 2180 }, { "epoch": 0.4, "learning_rate": 1.9979051046280677e-05, "loss": 0.1255, "step": 2181 }, { "epoch": 0.4, "learning_rate": 1.997900298750232e-05, "loss": 0.1199, "step": 2182 }, { "epoch": 0.4, "learning_rate": 1.9978954873719476e-05, "loss": 0.1131, "step": 2183 }, { "epoch": 0.4, "learning_rate": 1.9978906704932392e-05, "loss": 0.1213, "step": 2184 }, { "epoch": 0.4, "learning_rate": 1.9978858481141343e-05, "loss": 0.1196, "step": 2185 }, { "epoch": 0.4, "learning_rate": 1.997881020234659e-05, "loss": 0.1205, "step": 2186 }, { "epoch": 0.4, "learning_rate": 1.9978761868548403e-05, "loss": 0.1258, "step": 2187 }, { "epoch": 0.4, "learning_rate": 1.997871347974705e-05, "loss": 0.1233, "step": 2188 }, { "epoch": 0.4, "learning_rate": 1.997866503594279e-05, "loss": 0.1201, "step": 2189 }, { "epoch": 0.4, "learning_rate": 1.9978616537135897e-05, "loss": 0.135, "step": 2190 }, { "epoch": 0.4, "learning_rate": 1.9978567983326632e-05, "loss": 0.1121, "step": 2191 }, { "epoch": 0.4, "learning_rate": 1.997851937451527e-05, "loss": 0.1277, "step": 2192 }, { "epoch": 0.4, "learning_rate": 1.9978470710702073e-05, "loss": 0.1186, "step": 2193 }, { "epoch": 0.4, "learning_rate": 1.9978421991887313e-05, "loss": 0.1309, "step": 2194 }, { "epoch": 0.4, "learning_rate": 1.997837321807126e-05, "loss": 0.1286, "step": 2195 }, { "epoch": 0.4, "learning_rate": 1.9978324389254174e-05, "loss": 0.1185, "step": 2196 }, { "epoch": 0.4, "learning_rate": 1.9978275505436337e-05, "loss": 0.1215, "step": 2197 }, { "epoch": 0.4, "learning_rate": 1.9978226566618007e-05, "loss": 0.1199, "step": 2198 }, { "epoch": 0.4, "learning_rate": 1.997817757279946e-05, "loss": 0.1194, "step": 2199 }, { "epoch": 0.4, "learning_rate": 1.9978128523980965e-05, "loss": 0.1206, "step": 2200 }, { "epoch": 0.4, "learning_rate": 1.997807942016279e-05, "loss": 0.1257, "step": 2201 }, { "epoch": 0.4, "learning_rate": 1.997803026134521e-05, "loss": 0.1197, "step": 2202 }, { "epoch": 0.4, "learning_rate": 1.997798104752849e-05, "loss": 0.134, "step": 2203 }, { "epoch": 0.4, "learning_rate": 1.9977931778712905e-05, "loss": 0.1194, "step": 2204 }, { "epoch": 0.4, "learning_rate": 1.997788245489873e-05, "loss": 0.1273, "step": 2205 }, { "epoch": 0.4, "learning_rate": 1.997783307608623e-05, "loss": 0.1061, "step": 2206 }, { "epoch": 0.4, "learning_rate": 1.9977783642275682e-05, "loss": 0.1042, "step": 2207 }, { "epoch": 0.4, "learning_rate": 1.9977734153467354e-05, "loss": 0.1255, "step": 2208 }, { "epoch": 0.41, "learning_rate": 1.9977684609661525e-05, "loss": 0.1036, "step": 2209 }, { "epoch": 0.41, "learning_rate": 1.9977635010858463e-05, "loss": 0.1202, "step": 2210 }, { "epoch": 0.41, "learning_rate": 1.9977585357058444e-05, "loss": 0.1136, "step": 2211 }, { "epoch": 0.41, "learning_rate": 1.997753564826174e-05, "loss": 0.1278, "step": 2212 }, { "epoch": 0.41, "learning_rate": 1.9977485884468628e-05, "loss": 0.1209, "step": 2213 }, { "epoch": 0.41, "learning_rate": 1.9977436065679377e-05, "loss": 0.1312, "step": 2214 }, { "epoch": 0.41, "learning_rate": 1.9977386191894265e-05, "loss": 0.1252, "step": 2215 }, { "epoch": 0.41, "learning_rate": 1.9977336263113568e-05, "loss": 0.112, "step": 2216 }, { "epoch": 0.41, "learning_rate": 1.997728627933756e-05, "loss": 0.1337, "step": 2217 }, { "epoch": 0.41, "learning_rate": 1.9977236240566513e-05, "loss": 0.13, "step": 2218 }, { "epoch": 0.41, "learning_rate": 1.9977186146800707e-05, "loss": 0.1262, "step": 2219 }, { "epoch": 0.41, "learning_rate": 1.997713599804042e-05, "loss": 0.1164, "step": 2220 }, { "epoch": 0.41, "learning_rate": 1.9977085794285927e-05, "loss": 0.1259, "step": 2221 }, { "epoch": 0.41, "learning_rate": 1.99770355355375e-05, "loss": 0.1099, "step": 2222 }, { "epoch": 0.41, "learning_rate": 1.997698522179542e-05, "loss": 0.1162, "step": 2223 }, { "epoch": 0.41, "learning_rate": 1.9976934853059964e-05, "loss": 0.1342, "step": 2224 }, { "epoch": 0.41, "learning_rate": 1.9976884429331407e-05, "loss": 0.1224, "step": 2225 }, { "epoch": 0.41, "learning_rate": 1.9976833950610032e-05, "loss": 0.1414, "step": 2226 }, { "epoch": 0.41, "learning_rate": 1.9976783416896117e-05, "loss": 0.1225, "step": 2227 }, { "epoch": 0.41, "learning_rate": 1.9976732828189932e-05, "loss": 0.1189, "step": 2228 }, { "epoch": 0.41, "learning_rate": 1.9976682184491764e-05, "loss": 0.1313, "step": 2229 }, { "epoch": 0.41, "learning_rate": 1.9976631485801892e-05, "loss": 0.1105, "step": 2230 }, { "epoch": 0.41, "learning_rate": 1.997658073212059e-05, "loss": 0.1161, "step": 2231 }, { "epoch": 0.41, "learning_rate": 1.9976529923448143e-05, "loss": 0.123, "step": 2232 }, { "epoch": 0.41, "learning_rate": 1.9976479059784832e-05, "loss": 0.1166, "step": 2233 }, { "epoch": 0.41, "learning_rate": 1.9976428141130936e-05, "loss": 0.1171, "step": 2234 }, { "epoch": 0.41, "learning_rate": 1.997637716748673e-05, "loss": 0.1226, "step": 2235 }, { "epoch": 0.41, "learning_rate": 1.99763261388525e-05, "loss": 0.1161, "step": 2236 }, { "epoch": 0.41, "learning_rate": 1.9976275055228526e-05, "loss": 0.1113, "step": 2237 }, { "epoch": 0.41, "learning_rate": 1.997622391661509e-05, "loss": 0.1256, "step": 2238 }, { "epoch": 0.41, "learning_rate": 1.997617272301248e-05, "loss": 0.1302, "step": 2239 }, { "epoch": 0.41, "learning_rate": 1.9976121474420966e-05, "loss": 0.1143, "step": 2240 }, { "epoch": 0.41, "learning_rate": 1.9976070170840838e-05, "loss": 0.1195, "step": 2241 }, { "epoch": 0.41, "learning_rate": 1.9976018812272378e-05, "loss": 0.1246, "step": 2242 }, { "epoch": 0.41, "learning_rate": 1.9975967398715867e-05, "loss": 0.1255, "step": 2243 }, { "epoch": 0.41, "learning_rate": 1.9975915930171593e-05, "loss": 0.1345, "step": 2244 }, { "epoch": 0.41, "learning_rate": 1.9975864406639835e-05, "loss": 0.1205, "step": 2245 }, { "epoch": 0.41, "learning_rate": 1.997581282812088e-05, "loss": 0.1222, "step": 2246 }, { "epoch": 0.41, "learning_rate": 1.9975761194615008e-05, "loss": 0.1211, "step": 2247 }, { "epoch": 0.41, "learning_rate": 1.997570950612251e-05, "loss": 0.1219, "step": 2248 }, { "epoch": 0.41, "learning_rate": 1.9975657762643666e-05, "loss": 0.1144, "step": 2249 }, { "epoch": 0.41, "learning_rate": 1.9975605964178762e-05, "loss": 0.1158, "step": 2250 }, { "epoch": 0.41, "learning_rate": 1.9975554110728084e-05, "loss": 0.1155, "step": 2251 }, { "epoch": 0.41, "learning_rate": 1.997550220229192e-05, "loss": 0.1242, "step": 2252 }, { "epoch": 0.41, "learning_rate": 1.997545023887055e-05, "loss": 0.1281, "step": 2253 }, { "epoch": 0.41, "learning_rate": 1.9975398220464268e-05, "loss": 0.1198, "step": 2254 }, { "epoch": 0.41, "learning_rate": 1.9975346147073357e-05, "loss": 0.1172, "step": 2255 }, { "epoch": 0.41, "learning_rate": 1.9975294018698105e-05, "loss": 0.1535, "step": 2256 }, { "epoch": 0.41, "learning_rate": 1.9975241835338797e-05, "loss": 0.1223, "step": 2257 }, { "epoch": 0.41, "learning_rate": 1.9975189596995724e-05, "loss": 0.1092, "step": 2258 }, { "epoch": 0.41, "learning_rate": 1.997513730366917e-05, "loss": 0.1213, "step": 2259 }, { "epoch": 0.41, "learning_rate": 1.997508495535943e-05, "loss": 0.129, "step": 2260 }, { "epoch": 0.41, "learning_rate": 1.9975032552066783e-05, "loss": 0.121, "step": 2261 }, { "epoch": 0.41, "learning_rate": 1.9974980093791524e-05, "loss": 0.1294, "step": 2262 }, { "epoch": 0.42, "learning_rate": 1.9974927580533944e-05, "loss": 0.1153, "step": 2263 }, { "epoch": 0.42, "learning_rate": 1.9974875012294327e-05, "loss": 0.1271, "step": 2264 }, { "epoch": 0.42, "learning_rate": 1.9974822389072964e-05, "loss": 0.119, "step": 2265 }, { "epoch": 0.42, "learning_rate": 1.9974769710870147e-05, "loss": 0.1295, "step": 2266 }, { "epoch": 0.42, "learning_rate": 1.9974716977686167e-05, "loss": 0.1117, "step": 2267 }, { "epoch": 0.42, "learning_rate": 1.9974664189521313e-05, "loss": 0.1158, "step": 2268 }, { "epoch": 0.42, "learning_rate": 1.997461134637588e-05, "loss": 0.1259, "step": 2269 }, { "epoch": 0.42, "learning_rate": 1.997455844825015e-05, "loss": 0.1053, "step": 2270 }, { "epoch": 0.42, "learning_rate": 1.997450549514442e-05, "loss": 0.1266, "step": 2271 }, { "epoch": 0.42, "learning_rate": 1.9974452487058987e-05, "loss": 0.1104, "step": 2272 }, { "epoch": 0.42, "learning_rate": 1.9974399423994133e-05, "loss": 0.1248, "step": 2273 }, { "epoch": 0.42, "learning_rate": 1.9974346305950156e-05, "loss": 0.1116, "step": 2274 }, { "epoch": 0.42, "learning_rate": 1.9974293132927354e-05, "loss": 0.114, "step": 2275 }, { "epoch": 0.42, "learning_rate": 1.997423990492601e-05, "loss": 0.1147, "step": 2276 }, { "epoch": 0.42, "learning_rate": 1.997418662194642e-05, "loss": 0.1289, "step": 2277 }, { "epoch": 0.42, "learning_rate": 1.9974133283988886e-05, "loss": 0.1145, "step": 2278 }, { "epoch": 0.42, "learning_rate": 1.997407989105369e-05, "loss": 0.1245, "step": 2279 }, { "epoch": 0.42, "learning_rate": 1.9974026443141136e-05, "loss": 0.1166, "step": 2280 }, { "epoch": 0.42, "learning_rate": 1.9973972940251514e-05, "loss": 0.1248, "step": 2281 }, { "epoch": 0.42, "learning_rate": 1.9973919382385117e-05, "loss": 0.132, "step": 2282 }, { "epoch": 0.42, "learning_rate": 1.9973865769542243e-05, "loss": 0.1195, "step": 2283 }, { "epoch": 0.42, "learning_rate": 1.997381210172319e-05, "loss": 0.1192, "step": 2284 }, { "epoch": 0.42, "learning_rate": 1.9973758378928246e-05, "loss": 0.1155, "step": 2285 }, { "epoch": 0.42, "learning_rate": 1.9973704601157717e-05, "loss": 0.1243, "step": 2286 }, { "epoch": 0.42, "learning_rate": 1.9973650768411893e-05, "loss": 0.1089, "step": 2287 }, { "epoch": 0.42, "learning_rate": 1.9973596880691072e-05, "loss": 0.1284, "step": 2288 }, { "epoch": 0.42, "learning_rate": 1.997354293799555e-05, "loss": 0.1209, "step": 2289 }, { "epoch": 0.42, "learning_rate": 1.997348894032563e-05, "loss": 0.1277, "step": 2290 }, { "epoch": 0.42, "learning_rate": 1.99734348876816e-05, "loss": 0.1317, "step": 2291 }, { "epoch": 0.42, "learning_rate": 1.9973380780063767e-05, "loss": 0.1205, "step": 2292 }, { "epoch": 0.42, "learning_rate": 1.9973326617472428e-05, "loss": 0.1278, "step": 2293 }, { "epoch": 0.42, "learning_rate": 1.9973272399907874e-05, "loss": 0.1137, "step": 2294 }, { "epoch": 0.42, "learning_rate": 1.9973218127370407e-05, "loss": 0.1265, "step": 2295 }, { "epoch": 0.42, "learning_rate": 1.9973163799860333e-05, "loss": 0.1225, "step": 2296 }, { "epoch": 0.42, "learning_rate": 1.9973109417377946e-05, "loss": 0.1299, "step": 2297 }, { "epoch": 0.42, "learning_rate": 1.9973054979923544e-05, "loss": 0.1241, "step": 2298 }, { "epoch": 0.42, "learning_rate": 1.9973000487497433e-05, "loss": 0.1376, "step": 2299 }, { "epoch": 0.42, "learning_rate": 1.9972945940099902e-05, "loss": 0.1221, "step": 2300 }, { "epoch": 0.42, "learning_rate": 1.997289133773127e-05, "loss": 0.1228, "step": 2301 }, { "epoch": 0.42, "learning_rate": 1.9972836680391817e-05, "loss": 0.1052, "step": 2302 }, { "epoch": 0.42, "learning_rate": 1.997278196808186e-05, "loss": 0.1217, "step": 2303 }, { "epoch": 0.42, "learning_rate": 1.9972727200801696e-05, "loss": 0.1324, "step": 2304 }, { "epoch": 0.42, "learning_rate": 1.9972672378551624e-05, "loss": 0.1134, "step": 2305 }, { "epoch": 0.42, "learning_rate": 1.997261750133195e-05, "loss": 0.1302, "step": 2306 }, { "epoch": 0.42, "learning_rate": 1.9972562569142974e-05, "loss": 0.1379, "step": 2307 }, { "epoch": 0.42, "learning_rate": 1.9972507581984996e-05, "loss": 0.1245, "step": 2308 }, { "epoch": 0.42, "learning_rate": 1.9972452539858324e-05, "loss": 0.1234, "step": 2309 }, { "epoch": 0.42, "learning_rate": 1.997239744276326e-05, "loss": 0.1152, "step": 2310 }, { "epoch": 0.42, "learning_rate": 1.9972342290700112e-05, "loss": 0.1287, "step": 2311 }, { "epoch": 0.42, "learning_rate": 1.997228708366918e-05, "loss": 0.1312, "step": 2312 }, { "epoch": 0.42, "learning_rate": 1.997223182167076e-05, "loss": 0.1181, "step": 2313 }, { "epoch": 0.42, "learning_rate": 1.997217650470517e-05, "loss": 0.1137, "step": 2314 }, { "epoch": 0.42, "learning_rate": 1.997212113277271e-05, "loss": 0.1083, "step": 2315 }, { "epoch": 0.42, "learning_rate": 1.9972065705873686e-05, "loss": 0.1211, "step": 2316 }, { "epoch": 0.42, "learning_rate": 1.9972010224008395e-05, "loss": 0.1254, "step": 2317 }, { "epoch": 0.43, "learning_rate": 1.9971954687177154e-05, "loss": 0.1174, "step": 2318 }, { "epoch": 0.43, "learning_rate": 1.997189909538027e-05, "loss": 0.1187, "step": 2319 }, { "epoch": 0.43, "learning_rate": 1.997184344861804e-05, "loss": 0.117, "step": 2320 }, { "epoch": 0.43, "learning_rate": 1.9971787746890775e-05, "loss": 0.1223, "step": 2321 }, { "epoch": 0.43, "learning_rate": 1.997173199019878e-05, "loss": 0.1164, "step": 2322 }, { "epoch": 0.43, "learning_rate": 1.997167617854237e-05, "loss": 0.1212, "step": 2323 }, { "epoch": 0.43, "learning_rate": 1.9971620311921843e-05, "loss": 0.1085, "step": 2324 }, { "epoch": 0.43, "learning_rate": 1.9971564390337514e-05, "loss": 0.1298, "step": 2325 }, { "epoch": 0.43, "learning_rate": 1.9971508413789688e-05, "loss": 0.1146, "step": 2326 }, { "epoch": 0.43, "learning_rate": 1.9971452382278677e-05, "loss": 0.1246, "step": 2327 }, { "epoch": 0.43, "learning_rate": 1.997139629580478e-05, "loss": 0.1134, "step": 2328 }, { "epoch": 0.43, "learning_rate": 1.9971340154368318e-05, "loss": 0.1195, "step": 2329 }, { "epoch": 0.43, "learning_rate": 1.9971283957969594e-05, "loss": 0.1165, "step": 2330 }, { "epoch": 0.43, "learning_rate": 1.9971227706608918e-05, "loss": 0.1138, "step": 2331 }, { "epoch": 0.43, "learning_rate": 1.9971171400286602e-05, "loss": 0.127, "step": 2332 }, { "epoch": 0.43, "learning_rate": 1.9971115039002957e-05, "loss": 0.1205, "step": 2333 }, { "epoch": 0.43, "learning_rate": 1.9971058622758292e-05, "loss": 0.114, "step": 2334 }, { "epoch": 0.43, "learning_rate": 1.9971002151552915e-05, "loss": 0.1214, "step": 2335 }, { "epoch": 0.43, "learning_rate": 1.9970945625387144e-05, "loss": 0.1119, "step": 2336 }, { "epoch": 0.43, "learning_rate": 1.997088904426129e-05, "loss": 0.1081, "step": 2337 }, { "epoch": 0.43, "learning_rate": 1.9970832408175656e-05, "loss": 0.1259, "step": 2338 }, { "epoch": 0.43, "learning_rate": 1.9970775717130562e-05, "loss": 0.1186, "step": 2339 }, { "epoch": 0.43, "learning_rate": 1.997071897112632e-05, "loss": 0.1149, "step": 2340 }, { "epoch": 0.43, "learning_rate": 1.997066217016324e-05, "loss": 0.1265, "step": 2341 }, { "epoch": 0.43, "learning_rate": 1.997060531424164e-05, "loss": 0.1239, "step": 2342 }, { "epoch": 0.43, "learning_rate": 1.9970548403361825e-05, "loss": 0.1117, "step": 2343 }, { "epoch": 0.43, "learning_rate": 1.9970491437524115e-05, "loss": 0.1264, "step": 2344 }, { "epoch": 0.43, "learning_rate": 1.9970434416728827e-05, "loss": 0.1166, "step": 2345 }, { "epoch": 0.43, "learning_rate": 1.9970377340976265e-05, "loss": 0.1187, "step": 2346 }, { "epoch": 0.43, "learning_rate": 1.9970320210266753e-05, "loss": 0.1335, "step": 2347 }, { "epoch": 0.43, "learning_rate": 1.9970263024600598e-05, "loss": 0.1282, "step": 2348 }, { "epoch": 0.43, "learning_rate": 1.9970205783978125e-05, "loss": 0.1238, "step": 2349 }, { "epoch": 0.43, "learning_rate": 1.997014848839964e-05, "loss": 0.1284, "step": 2350 }, { "epoch": 0.43, "learning_rate": 1.9970091137865467e-05, "loss": 0.1303, "step": 2351 }, { "epoch": 0.43, "learning_rate": 1.9970033732375915e-05, "loss": 0.1227, "step": 2352 }, { "epoch": 0.43, "learning_rate": 1.9969976271931303e-05, "loss": 0.1105, "step": 2353 }, { "epoch": 0.43, "learning_rate": 1.996991875653195e-05, "loss": 0.1234, "step": 2354 }, { "epoch": 0.43, "learning_rate": 1.9969861186178174e-05, "loss": 0.1144, "step": 2355 }, { "epoch": 0.43, "learning_rate": 1.9969803560870284e-05, "loss": 0.1146, "step": 2356 }, { "epoch": 0.43, "learning_rate": 1.9969745880608608e-05, "loss": 0.1048, "step": 2357 }, { "epoch": 0.43, "learning_rate": 1.9969688145393456e-05, "loss": 0.1257, "step": 2358 }, { "epoch": 0.43, "learning_rate": 1.996963035522515e-05, "loss": 0.1246, "step": 2359 }, { "epoch": 0.43, "learning_rate": 1.9969572510104006e-05, "loss": 0.1205, "step": 2360 }, { "epoch": 0.43, "learning_rate": 1.9969514610030348e-05, "loss": 0.1174, "step": 2361 }, { "epoch": 0.43, "learning_rate": 1.9969456655004492e-05, "loss": 0.1082, "step": 2362 }, { "epoch": 0.43, "learning_rate": 1.9969398645026757e-05, "loss": 0.1256, "step": 2363 }, { "epoch": 0.43, "learning_rate": 1.996934058009746e-05, "loss": 0.128, "step": 2364 }, { "epoch": 0.43, "learning_rate": 1.9969282460216927e-05, "loss": 0.1166, "step": 2365 }, { "epoch": 0.43, "learning_rate": 1.9969224285385475e-05, "loss": 0.1113, "step": 2366 }, { "epoch": 0.43, "learning_rate": 1.9969166055603426e-05, "loss": 0.1228, "step": 2367 }, { "epoch": 0.43, "learning_rate": 1.99691077708711e-05, "loss": 0.1077, "step": 2368 }, { "epoch": 0.43, "learning_rate": 1.9969049431188816e-05, "loss": 0.1236, "step": 2369 }, { "epoch": 0.43, "learning_rate": 1.99689910365569e-05, "loss": 0.1196, "step": 2370 }, { "epoch": 0.43, "learning_rate": 1.996893258697567e-05, "loss": 0.1238, "step": 2371 }, { "epoch": 0.43, "learning_rate": 1.9968874082445454e-05, "loss": 0.1276, "step": 2372 }, { "epoch": 0.44, "learning_rate": 1.9968815522966567e-05, "loss": 0.1218, "step": 2373 }, { "epoch": 0.44, "learning_rate": 1.9968756908539334e-05, "loss": 0.0976, "step": 2374 }, { "epoch": 0.44, "learning_rate": 1.9968698239164085e-05, "loss": 0.1126, "step": 2375 }, { "epoch": 0.44, "learning_rate": 1.9968639514841134e-05, "loss": 0.1276, "step": 2376 }, { "epoch": 0.44, "learning_rate": 1.996858073557081e-05, "loss": 0.1246, "step": 2377 }, { "epoch": 0.44, "learning_rate": 1.9968521901353432e-05, "loss": 0.1301, "step": 2378 }, { "epoch": 0.44, "learning_rate": 1.9968463012189332e-05, "loss": 0.1302, "step": 2379 }, { "epoch": 0.44, "learning_rate": 1.9968404068078827e-05, "loss": 0.117, "step": 2380 }, { "epoch": 0.44, "learning_rate": 1.9968345069022248e-05, "loss": 0.1131, "step": 2381 }, { "epoch": 0.44, "learning_rate": 1.996828601501992e-05, "loss": 0.1111, "step": 2382 }, { "epoch": 0.44, "learning_rate": 1.9968226906072158e-05, "loss": 0.1189, "step": 2383 }, { "epoch": 0.44, "learning_rate": 1.99681677421793e-05, "loss": 0.1166, "step": 2384 }, { "epoch": 0.44, "learning_rate": 1.996810852334167e-05, "loss": 0.1143, "step": 2385 }, { "epoch": 0.44, "learning_rate": 1.9968049249559586e-05, "loss": 0.1163, "step": 2386 }, { "epoch": 0.44, "learning_rate": 1.996798992083339e-05, "loss": 0.1262, "step": 2387 }, { "epoch": 0.44, "learning_rate": 1.9967930537163393e-05, "loss": 0.1203, "step": 2388 }, { "epoch": 0.44, "learning_rate": 1.9967871098549932e-05, "loss": 0.112, "step": 2389 }, { "epoch": 0.44, "learning_rate": 1.996781160499333e-05, "loss": 0.1231, "step": 2390 }, { "epoch": 0.44, "learning_rate": 1.996775205649392e-05, "loss": 0.1163, "step": 2391 }, { "epoch": 0.44, "learning_rate": 1.9967692453052025e-05, "loss": 0.1142, "step": 2392 }, { "epoch": 0.44, "learning_rate": 1.9967632794667974e-05, "loss": 0.1134, "step": 2393 }, { "epoch": 0.44, "learning_rate": 1.9967573081342103e-05, "loss": 0.1086, "step": 2394 }, { "epoch": 0.44, "learning_rate": 1.996751331307473e-05, "loss": 0.1271, "step": 2395 }, { "epoch": 0.44, "learning_rate": 1.9967453489866194e-05, "loss": 0.1204, "step": 2396 }, { "epoch": 0.44, "learning_rate": 1.996739361171682e-05, "loss": 0.1204, "step": 2397 }, { "epoch": 0.44, "learning_rate": 1.9967333678626935e-05, "loss": 0.1299, "step": 2398 }, { "epoch": 0.44, "learning_rate": 1.996727369059688e-05, "loss": 0.1282, "step": 2399 }, { "epoch": 0.44, "learning_rate": 1.9967213647626975e-05, "loss": 0.1224, "step": 2400 }, { "epoch": 0.44, "learning_rate": 1.996715354971755e-05, "loss": 0.1112, "step": 2401 }, { "epoch": 0.44, "learning_rate": 1.996709339686895e-05, "loss": 0.1329, "step": 2402 }, { "epoch": 0.44, "learning_rate": 1.9967033189081495e-05, "loss": 0.1165, "step": 2403 }, { "epoch": 0.44, "learning_rate": 1.9966972926355517e-05, "loss": 0.1197, "step": 2404 }, { "epoch": 0.44, "learning_rate": 1.9966912608691354e-05, "loss": 0.1215, "step": 2405 }, { "epoch": 0.44, "learning_rate": 1.9966852236089335e-05, "loss": 0.1259, "step": 2406 }, { "epoch": 0.44, "learning_rate": 1.996679180854979e-05, "loss": 0.1159, "step": 2407 }, { "epoch": 0.44, "learning_rate": 1.9966731326073056e-05, "loss": 0.1177, "step": 2408 }, { "epoch": 0.44, "learning_rate": 1.996667078865947e-05, "loss": 0.1209, "step": 2409 }, { "epoch": 0.44, "learning_rate": 1.9966610196309352e-05, "loss": 0.1183, "step": 2410 }, { "epoch": 0.44, "learning_rate": 1.9966549549023053e-05, "loss": 0.1226, "step": 2411 }, { "epoch": 0.44, "learning_rate": 1.9966488846800894e-05, "loss": 0.1181, "step": 2412 }, { "epoch": 0.44, "learning_rate": 1.996642808964322e-05, "loss": 0.112, "step": 2413 }, { "epoch": 0.44, "learning_rate": 1.9966367277550358e-05, "loss": 0.1105, "step": 2414 }, { "epoch": 0.44, "learning_rate": 1.9966306410522648e-05, "loss": 0.1293, "step": 2415 }, { "epoch": 0.44, "learning_rate": 1.996624548856042e-05, "loss": 0.1383, "step": 2416 }, { "epoch": 0.44, "learning_rate": 1.9966184511664016e-05, "loss": 0.1168, "step": 2417 }, { "epoch": 0.44, "learning_rate": 1.996612347983377e-05, "loss": 0.1202, "step": 2418 }, { "epoch": 0.44, "learning_rate": 1.9966062393070014e-05, "loss": 0.1202, "step": 2419 }, { "epoch": 0.44, "learning_rate": 1.9966001251373095e-05, "loss": 0.1262, "step": 2420 }, { "epoch": 0.44, "learning_rate": 1.996594005474334e-05, "loss": 0.1209, "step": 2421 }, { "epoch": 0.44, "learning_rate": 1.996587880318109e-05, "loss": 0.1194, "step": 2422 }, { "epoch": 0.44, "learning_rate": 1.996581749668668e-05, "loss": 0.1203, "step": 2423 }, { "epoch": 0.44, "learning_rate": 1.9965756135260456e-05, "loss": 0.1113, "step": 2424 }, { "epoch": 0.44, "learning_rate": 1.9965694718902745e-05, "loss": 0.1282, "step": 2425 }, { "epoch": 0.44, "learning_rate": 1.9965633247613895e-05, "loss": 0.1141, "step": 2426 }, { "epoch": 0.45, "learning_rate": 1.9965571721394242e-05, "loss": 0.1336, "step": 2427 }, { "epoch": 0.45, "learning_rate": 1.996551014024412e-05, "loss": 0.1149, "step": 2428 }, { "epoch": 0.45, "learning_rate": 1.9965448504163878e-05, "loss": 0.1145, "step": 2429 }, { "epoch": 0.45, "learning_rate": 1.9965386813153847e-05, "loss": 0.1192, "step": 2430 }, { "epoch": 0.45, "learning_rate": 1.996532506721437e-05, "loss": 0.1223, "step": 2431 }, { "epoch": 0.45, "learning_rate": 1.996526326634579e-05, "loss": 0.1277, "step": 2432 }, { "epoch": 0.45, "learning_rate": 1.9965201410548443e-05, "loss": 0.1302, "step": 2433 }, { "epoch": 0.45, "learning_rate": 1.9965139499822673e-05, "loss": 0.1155, "step": 2434 }, { "epoch": 0.45, "learning_rate": 1.9965077534168823e-05, "loss": 0.1234, "step": 2435 }, { "epoch": 0.45, "learning_rate": 1.9965015513587227e-05, "loss": 0.1238, "step": 2436 }, { "epoch": 0.45, "learning_rate": 1.9964953438078235e-05, "loss": 0.1175, "step": 2437 }, { "epoch": 0.45, "learning_rate": 1.9964891307642188e-05, "loss": 0.11, "step": 2438 }, { "epoch": 0.45, "learning_rate": 1.9964829122279426e-05, "loss": 0.1296, "step": 2439 }, { "epoch": 0.45, "learning_rate": 1.996476688199029e-05, "loss": 0.1166, "step": 2440 }, { "epoch": 0.45, "learning_rate": 1.9964704586775128e-05, "loss": 0.1122, "step": 2441 }, { "epoch": 0.45, "learning_rate": 1.996464223663428e-05, "loss": 0.1304, "step": 2442 }, { "epoch": 0.45, "learning_rate": 1.996457983156809e-05, "loss": 0.1235, "step": 2443 }, { "epoch": 0.45, "learning_rate": 1.9964517371576903e-05, "loss": 0.1142, "step": 2444 }, { "epoch": 0.45, "learning_rate": 1.996445485666106e-05, "loss": 0.1245, "step": 2445 }, { "epoch": 0.45, "learning_rate": 1.9964392286820914e-05, "loss": 0.121, "step": 2446 }, { "epoch": 0.45, "learning_rate": 1.9964329662056802e-05, "loss": 0.1174, "step": 2447 }, { "epoch": 0.45, "learning_rate": 1.996426698236907e-05, "loss": 0.1227, "step": 2448 }, { "epoch": 0.45, "learning_rate": 1.9964204247758064e-05, "loss": 0.1204, "step": 2449 }, { "epoch": 0.45, "learning_rate": 1.9964141458224136e-05, "loss": 0.1239, "step": 2450 }, { "epoch": 0.45, "learning_rate": 1.996407861376762e-05, "loss": 0.1205, "step": 2451 }, { "epoch": 0.45, "learning_rate": 1.9964015714388877e-05, "loss": 0.1076, "step": 2452 }, { "epoch": 0.45, "learning_rate": 1.996395276008824e-05, "loss": 0.1179, "step": 2453 }, { "epoch": 0.45, "learning_rate": 1.9963889750866065e-05, "loss": 0.1241, "step": 2454 }, { "epoch": 0.45, "learning_rate": 1.9963826686722693e-05, "loss": 0.1169, "step": 2455 }, { "epoch": 0.45, "learning_rate": 1.9963763567658475e-05, "loss": 0.1138, "step": 2456 }, { "epoch": 0.45, "learning_rate": 1.996370039367376e-05, "loss": 0.1306, "step": 2457 }, { "epoch": 0.45, "learning_rate": 1.9963637164768895e-05, "loss": 0.1175, "step": 2458 }, { "epoch": 0.45, "learning_rate": 1.9963573880944232e-05, "loss": 0.1305, "step": 2459 }, { "epoch": 0.45, "learning_rate": 1.996351054220011e-05, "loss": 0.1154, "step": 2460 }, { "epoch": 0.45, "learning_rate": 1.9963447148536888e-05, "loss": 0.1266, "step": 2461 }, { "epoch": 0.45, "learning_rate": 1.9963383699954915e-05, "loss": 0.1301, "step": 2462 }, { "epoch": 0.45, "learning_rate": 1.9963320196454534e-05, "loss": 0.1222, "step": 2463 }, { "epoch": 0.45, "learning_rate": 1.99632566380361e-05, "loss": 0.1169, "step": 2464 }, { "epoch": 0.45, "learning_rate": 1.996319302469996e-05, "loss": 0.127, "step": 2465 }, { "epoch": 0.45, "learning_rate": 1.9963129356446468e-05, "loss": 0.1205, "step": 2466 }, { "epoch": 0.45, "learning_rate": 1.9963065633275973e-05, "loss": 0.1224, "step": 2467 }, { "epoch": 0.45, "learning_rate": 1.996300185518883e-05, "loss": 0.1057, "step": 2468 }, { "epoch": 0.45, "learning_rate": 1.996293802218538e-05, "loss": 0.135, "step": 2469 }, { "epoch": 0.45, "learning_rate": 1.9962874134265988e-05, "loss": 0.1377, "step": 2470 }, { "epoch": 0.45, "learning_rate": 1.9962810191431e-05, "loss": 0.113, "step": 2471 }, { "epoch": 0.45, "learning_rate": 1.9962746193680765e-05, "loss": 0.1317, "step": 2472 }, { "epoch": 0.45, "learning_rate": 1.9962682141015643e-05, "loss": 0.1195, "step": 2473 }, { "epoch": 0.45, "learning_rate": 1.9962618033435985e-05, "loss": 0.1207, "step": 2474 }, { "epoch": 0.45, "learning_rate": 1.9962553870942142e-05, "loss": 0.1201, "step": 2475 }, { "epoch": 0.45, "learning_rate": 1.9962489653534468e-05, "loss": 0.1274, "step": 2476 }, { "epoch": 0.45, "learning_rate": 1.9962425381213316e-05, "loss": 0.1119, "step": 2477 }, { "epoch": 0.45, "learning_rate": 1.9962361053979042e-05, "loss": 0.1228, "step": 2478 }, { "epoch": 0.45, "learning_rate": 1.9962296671832e-05, "loss": 0.1013, "step": 2479 }, { "epoch": 0.45, "learning_rate": 1.996223223477255e-05, "loss": 0.1105, "step": 2480 }, { "epoch": 0.45, "learning_rate": 1.996216774280104e-05, "loss": 0.1256, "step": 2481 }, { "epoch": 0.46, "learning_rate": 1.9962103195917824e-05, "loss": 0.1173, "step": 2482 }, { "epoch": 0.46, "learning_rate": 1.996203859412327e-05, "loss": 0.1171, "step": 2483 }, { "epoch": 0.46, "learning_rate": 1.9961973937417718e-05, "loss": 0.1231, "step": 2484 }, { "epoch": 0.46, "learning_rate": 1.9961909225801536e-05, "loss": 0.1194, "step": 2485 }, { "epoch": 0.46, "learning_rate": 1.9961844459275077e-05, "loss": 0.127, "step": 2486 }, { "epoch": 0.46, "learning_rate": 1.9961779637838698e-05, "loss": 0.1325, "step": 2487 }, { "epoch": 0.46, "learning_rate": 1.9961714761492757e-05, "loss": 0.1317, "step": 2488 }, { "epoch": 0.46, "learning_rate": 1.9961649830237608e-05, "loss": 0.1266, "step": 2489 }, { "epoch": 0.46, "learning_rate": 1.9961584844073612e-05, "loss": 0.114, "step": 2490 }, { "epoch": 0.46, "learning_rate": 1.996151980300113e-05, "loss": 0.1061, "step": 2491 }, { "epoch": 0.46, "learning_rate": 1.9961454707020515e-05, "loss": 0.1147, "step": 2492 }, { "epoch": 0.46, "learning_rate": 1.9961389556132128e-05, "loss": 0.1266, "step": 2493 }, { "epoch": 0.46, "learning_rate": 1.996132435033633e-05, "loss": 0.1211, "step": 2494 }, { "epoch": 0.46, "learning_rate": 1.996125908963348e-05, "loss": 0.1276, "step": 2495 }, { "epoch": 0.46, "learning_rate": 1.9961193774023934e-05, "loss": 0.125, "step": 2496 }, { "epoch": 0.46, "learning_rate": 1.9961128403508055e-05, "loss": 0.1058, "step": 2497 }, { "epoch": 0.46, "learning_rate": 1.9961062978086204e-05, "loss": 0.1308, "step": 2498 }, { "epoch": 0.46, "learning_rate": 1.996099749775874e-05, "loss": 0.1258, "step": 2499 }, { "epoch": 0.46, "learning_rate": 1.9960931962526024e-05, "loss": 0.1256, "step": 2500 }, { "epoch": 0.46, "learning_rate": 1.9960866372388417e-05, "loss": 0.1196, "step": 2501 }, { "epoch": 0.46, "learning_rate": 1.9960800727346284e-05, "loss": 0.1372, "step": 2502 }, { "epoch": 0.46, "learning_rate": 1.9960735027399984e-05, "loss": 0.1206, "step": 2503 }, { "epoch": 0.46, "learning_rate": 1.9960669272549875e-05, "loss": 0.1183, "step": 2504 }, { "epoch": 0.46, "learning_rate": 1.9960603462796327e-05, "loss": 0.1192, "step": 2505 }, { "epoch": 0.46, "learning_rate": 1.9960537598139696e-05, "loss": 0.1201, "step": 2506 }, { "epoch": 0.46, "learning_rate": 1.996047167858035e-05, "loss": 0.1262, "step": 2507 }, { "epoch": 0.46, "learning_rate": 1.996040570411865e-05, "loss": 0.1169, "step": 2508 }, { "epoch": 0.46, "learning_rate": 1.9960339674754966e-05, "loss": 0.1285, "step": 2509 }, { "epoch": 0.46, "learning_rate": 1.996027359048965e-05, "loss": 0.1172, "step": 2510 }, { "epoch": 0.46, "learning_rate": 1.9960207451323076e-05, "loss": 0.1081, "step": 2511 }, { "epoch": 0.46, "learning_rate": 1.9960141257255603e-05, "loss": 0.1132, "step": 2512 }, { "epoch": 0.46, "learning_rate": 1.99600750082876e-05, "loss": 0.1303, "step": 2513 }, { "epoch": 0.46, "learning_rate": 1.996000870441943e-05, "loss": 0.1102, "step": 2514 }, { "epoch": 0.46, "learning_rate": 1.995994234565146e-05, "loss": 0.1082, "step": 2515 }, { "epoch": 0.46, "learning_rate": 1.995987593198405e-05, "loss": 0.1048, "step": 2516 }, { "epoch": 0.46, "learning_rate": 1.9959809463417573e-05, "loss": 0.1201, "step": 2517 }, { "epoch": 0.46, "learning_rate": 1.9959742939952393e-05, "loss": 0.1118, "step": 2518 }, { "epoch": 0.46, "learning_rate": 1.9959676361588874e-05, "loss": 0.1246, "step": 2519 }, { "epoch": 0.46, "learning_rate": 1.995960972832739e-05, "loss": 0.1104, "step": 2520 }, { "epoch": 0.46, "learning_rate": 1.99595430401683e-05, "loss": 0.1307, "step": 2521 }, { "epoch": 0.46, "learning_rate": 1.9959476297111977e-05, "loss": 0.1006, "step": 2522 }, { "epoch": 0.46, "learning_rate": 1.995940949915879e-05, "loss": 0.1258, "step": 2523 }, { "epoch": 0.46, "learning_rate": 1.9959342646309102e-05, "loss": 0.1275, "step": 2524 }, { "epoch": 0.46, "learning_rate": 1.9959275738563286e-05, "loss": 0.1252, "step": 2525 }, { "epoch": 0.46, "learning_rate": 1.9959208775921704e-05, "loss": 0.1259, "step": 2526 }, { "epoch": 0.46, "learning_rate": 1.995914175838473e-05, "loss": 0.1149, "step": 2527 }, { "epoch": 0.46, "learning_rate": 1.9959074685952738e-05, "loss": 0.1228, "step": 2528 }, { "epoch": 0.46, "learning_rate": 1.9959007558626095e-05, "loss": 0.1195, "step": 2529 }, { "epoch": 0.46, "learning_rate": 1.9958940376405164e-05, "loss": 0.1124, "step": 2530 }, { "epoch": 0.46, "learning_rate": 1.995887313929032e-05, "loss": 0.1275, "step": 2531 }, { "epoch": 0.46, "learning_rate": 1.9958805847281937e-05, "loss": 0.1094, "step": 2532 }, { "epoch": 0.46, "learning_rate": 1.995873850038038e-05, "loss": 0.1188, "step": 2533 }, { "epoch": 0.46, "learning_rate": 1.9958671098586026e-05, "loss": 0.1197, "step": 2534 }, { "epoch": 0.46, "learning_rate": 1.995860364189924e-05, "loss": 0.1216, "step": 2535 }, { "epoch": 0.47, "learning_rate": 1.9958536130320397e-05, "loss": 0.1109, "step": 2536 }, { "epoch": 0.47, "learning_rate": 1.9958468563849875e-05, "loss": 0.1155, "step": 2537 }, { "epoch": 0.47, "learning_rate": 1.995840094248804e-05, "loss": 0.1048, "step": 2538 }, { "epoch": 0.47, "learning_rate": 1.995833326623526e-05, "loss": 0.1242, "step": 2539 }, { "epoch": 0.47, "learning_rate": 1.995826553509192e-05, "loss": 0.1233, "step": 2540 }, { "epoch": 0.47, "learning_rate": 1.9958197749058384e-05, "loss": 0.1103, "step": 2541 }, { "epoch": 0.47, "learning_rate": 1.9958129908135028e-05, "loss": 0.1304, "step": 2542 }, { "epoch": 0.47, "learning_rate": 1.995806201232223e-05, "loss": 0.1171, "step": 2543 }, { "epoch": 0.47, "learning_rate": 1.9957994061620357e-05, "loss": 0.1202, "step": 2544 }, { "epoch": 0.47, "learning_rate": 1.9957926056029788e-05, "loss": 0.1169, "step": 2545 }, { "epoch": 0.47, "learning_rate": 1.99578579955509e-05, "loss": 0.1165, "step": 2546 }, { "epoch": 0.47, "learning_rate": 1.9957789880184066e-05, "loss": 0.1133, "step": 2547 }, { "epoch": 0.47, "learning_rate": 1.9957721709929654e-05, "loss": 0.1348, "step": 2548 }, { "epoch": 0.47, "learning_rate": 1.9957653484788054e-05, "loss": 0.1184, "step": 2549 }, { "epoch": 0.47, "learning_rate": 1.995758520475963e-05, "loss": 0.1086, "step": 2550 }, { "epoch": 0.47, "learning_rate": 1.9957516869844768e-05, "loss": 0.1168, "step": 2551 }, { "epoch": 0.47, "learning_rate": 1.9957448480043835e-05, "loss": 0.1086, "step": 2552 }, { "epoch": 0.47, "learning_rate": 1.9957380035357214e-05, "loss": 0.1122, "step": 2553 }, { "epoch": 0.47, "learning_rate": 1.9957311535785283e-05, "loss": 0.1294, "step": 2554 }, { "epoch": 0.47, "learning_rate": 1.9957242981328416e-05, "loss": 0.117, "step": 2555 }, { "epoch": 0.47, "learning_rate": 1.9957174371986992e-05, "loss": 0.1272, "step": 2556 }, { "epoch": 0.47, "learning_rate": 1.995710570776139e-05, "loss": 0.1285, "step": 2557 }, { "epoch": 0.47, "learning_rate": 1.995703698865199e-05, "loss": 0.1213, "step": 2558 }, { "epoch": 0.47, "learning_rate": 1.9956968214659163e-05, "loss": 0.1094, "step": 2559 }, { "epoch": 0.47, "learning_rate": 1.99568993857833e-05, "loss": 0.1181, "step": 2560 }, { "epoch": 0.47, "learning_rate": 1.9956830502024773e-05, "loss": 0.1182, "step": 2561 }, { "epoch": 0.47, "learning_rate": 1.9956761563383963e-05, "loss": 0.1332, "step": 2562 }, { "epoch": 0.47, "learning_rate": 1.995669256986125e-05, "loss": 0.1256, "step": 2563 }, { "epoch": 0.47, "learning_rate": 1.995662352145701e-05, "loss": 0.1388, "step": 2564 }, { "epoch": 0.47, "learning_rate": 1.9956554418171634e-05, "loss": 0.1148, "step": 2565 }, { "epoch": 0.47, "learning_rate": 1.9956485260005494e-05, "loss": 0.1186, "step": 2566 }, { "epoch": 0.47, "learning_rate": 1.9956416046958976e-05, "loss": 0.1071, "step": 2567 }, { "epoch": 0.47, "learning_rate": 1.9956346779032458e-05, "loss": 0.118, "step": 2568 }, { "epoch": 0.47, "learning_rate": 1.995627745622632e-05, "loss": 0.1148, "step": 2569 }, { "epoch": 0.47, "learning_rate": 1.995620807854095e-05, "loss": 0.1245, "step": 2570 }, { "epoch": 0.47, "learning_rate": 1.995613864597673e-05, "loss": 0.1294, "step": 2571 }, { "epoch": 0.47, "learning_rate": 1.995606915853404e-05, "loss": 0.1252, "step": 2572 }, { "epoch": 0.47, "learning_rate": 1.995599961621326e-05, "loss": 0.1273, "step": 2573 }, { "epoch": 0.47, "learning_rate": 1.995593001901478e-05, "loss": 0.1093, "step": 2574 }, { "epoch": 0.47, "learning_rate": 1.995586036693898e-05, "loss": 0.1269, "step": 2575 }, { "epoch": 0.47, "learning_rate": 1.995579065998624e-05, "loss": 0.1269, "step": 2576 }, { "epoch": 0.47, "learning_rate": 1.9955720898156956e-05, "loss": 0.1264, "step": 2577 }, { "epoch": 0.47, "learning_rate": 1.9955651081451503e-05, "loss": 0.1129, "step": 2578 }, { "epoch": 0.47, "learning_rate": 1.995558120987027e-05, "loss": 0.1136, "step": 2579 }, { "epoch": 0.47, "learning_rate": 1.9955511283413632e-05, "loss": 0.1153, "step": 2580 }, { "epoch": 0.47, "learning_rate": 1.995544130208199e-05, "loss": 0.1284, "step": 2581 }, { "epoch": 0.47, "learning_rate": 1.995537126587572e-05, "loss": 0.1035, "step": 2582 }, { "epoch": 0.47, "learning_rate": 1.995530117479521e-05, "loss": 0.1217, "step": 2583 }, { "epoch": 0.47, "learning_rate": 1.9955231028840848e-05, "loss": 0.1268, "step": 2584 }, { "epoch": 0.47, "learning_rate": 1.9955160828013018e-05, "loss": 0.1122, "step": 2585 }, { "epoch": 0.47, "learning_rate": 1.995509057231211e-05, "loss": 0.1285, "step": 2586 }, { "epoch": 0.47, "learning_rate": 1.9955020261738506e-05, "loss": 0.1156, "step": 2587 }, { "epoch": 0.47, "learning_rate": 1.99549498962926e-05, "loss": 0.1248, "step": 2588 }, { "epoch": 0.47, "learning_rate": 1.9954879475974776e-05, "loss": 0.1226, "step": 2589 }, { "epoch": 0.47, "learning_rate": 1.9954809000785422e-05, "loss": 0.1188, "step": 2590 }, { "epoch": 0.48, "learning_rate": 1.995473847072493e-05, "loss": 0.1197, "step": 2591 }, { "epoch": 0.48, "learning_rate": 1.9954667885793687e-05, "loss": 0.1172, "step": 2592 }, { "epoch": 0.48, "learning_rate": 1.995459724599208e-05, "loss": 0.1267, "step": 2593 }, { "epoch": 0.48, "learning_rate": 1.99545265513205e-05, "loss": 0.1087, "step": 2594 }, { "epoch": 0.48, "learning_rate": 1.9954455801779335e-05, "loss": 0.1064, "step": 2595 }, { "epoch": 0.48, "learning_rate": 1.9954384997368976e-05, "loss": 0.1202, "step": 2596 }, { "epoch": 0.48, "learning_rate": 1.9954314138089814e-05, "loss": 0.117, "step": 2597 }, { "epoch": 0.48, "learning_rate": 1.9954243223942244e-05, "loss": 0.1289, "step": 2598 }, { "epoch": 0.48, "learning_rate": 1.9954172254926644e-05, "loss": 0.1108, "step": 2599 }, { "epoch": 0.48, "learning_rate": 1.995410123104342e-05, "loss": 0.1225, "step": 2600 }, { "epoch": 0.48, "learning_rate": 1.9954030152292954e-05, "loss": 0.1278, "step": 2601 }, { "epoch": 0.48, "learning_rate": 1.995395901867564e-05, "loss": 0.1223, "step": 2602 }, { "epoch": 0.48, "learning_rate": 1.9953887830191872e-05, "loss": 0.125, "step": 2603 }, { "epoch": 0.48, "learning_rate": 1.9953816586842043e-05, "loss": 0.1104, "step": 2604 }, { "epoch": 0.48, "learning_rate": 1.9953745288626543e-05, "loss": 0.1141, "step": 2605 }, { "epoch": 0.48, "learning_rate": 1.9953673935545764e-05, "loss": 0.1267, "step": 2606 }, { "epoch": 0.48, "learning_rate": 1.99536025276001e-05, "loss": 0.116, "step": 2607 }, { "epoch": 0.48, "learning_rate": 1.995353106478995e-05, "loss": 0.1072, "step": 2608 }, { "epoch": 0.48, "learning_rate": 1.99534595471157e-05, "loss": 0.1286, "step": 2609 }, { "epoch": 0.48, "learning_rate": 1.9953387974577748e-05, "loss": 0.108, "step": 2610 }, { "epoch": 0.48, "learning_rate": 1.995331634717649e-05, "loss": 0.1189, "step": 2611 }, { "epoch": 0.48, "learning_rate": 1.9953244664912316e-05, "loss": 0.1255, "step": 2612 }, { "epoch": 0.48, "learning_rate": 1.9953172927785627e-05, "loss": 0.1061, "step": 2613 }, { "epoch": 0.48, "learning_rate": 1.9953101135796813e-05, "loss": 0.1159, "step": 2614 }, { "epoch": 0.48, "learning_rate": 1.9953029288946275e-05, "loss": 0.1129, "step": 2615 }, { "epoch": 0.48, "learning_rate": 1.9952957387234406e-05, "loss": 0.1094, "step": 2616 }, { "epoch": 0.48, "learning_rate": 1.9952885430661602e-05, "loss": 0.1198, "step": 2617 }, { "epoch": 0.48, "learning_rate": 1.995281341922826e-05, "loss": 0.127, "step": 2618 }, { "epoch": 0.48, "learning_rate": 1.995274135293478e-05, "loss": 0.111, "step": 2619 }, { "epoch": 0.48, "learning_rate": 1.9952669231781555e-05, "loss": 0.1208, "step": 2620 }, { "epoch": 0.48, "learning_rate": 1.9952597055768982e-05, "loss": 0.1096, "step": 2621 }, { "epoch": 0.48, "learning_rate": 1.9952524824897464e-05, "loss": 0.1217, "step": 2622 }, { "epoch": 0.48, "learning_rate": 1.9952452539167392e-05, "loss": 0.123, "step": 2623 }, { "epoch": 0.48, "learning_rate": 1.9952380198579174e-05, "loss": 0.1115, "step": 2624 }, { "epoch": 0.48, "learning_rate": 1.9952307803133196e-05, "loss": 0.1226, "step": 2625 }, { "epoch": 0.48, "learning_rate": 1.995223535282987e-05, "loss": 0.1215, "step": 2626 }, { "epoch": 0.48, "learning_rate": 1.995216284766959e-05, "loss": 0.1254, "step": 2627 }, { "epoch": 0.48, "learning_rate": 1.9952090287652752e-05, "loss": 0.1159, "step": 2628 }, { "epoch": 0.48, "learning_rate": 1.995201767277976e-05, "loss": 0.1096, "step": 2629 }, { "epoch": 0.48, "learning_rate": 1.9951945003051014e-05, "loss": 0.1269, "step": 2630 }, { "epoch": 0.48, "learning_rate": 1.9951872278466914e-05, "loss": 0.119, "step": 2631 }, { "epoch": 0.48, "learning_rate": 1.9951799499027856e-05, "loss": 0.1128, "step": 2632 }, { "epoch": 0.48, "learning_rate": 1.995172666473425e-05, "loss": 0.1081, "step": 2633 }, { "epoch": 0.48, "learning_rate": 1.9951653775586496e-05, "loss": 0.131, "step": 2634 }, { "epoch": 0.48, "learning_rate": 1.9951580831584993e-05, "loss": 0.1196, "step": 2635 }, { "epoch": 0.48, "learning_rate": 1.995150783273014e-05, "loss": 0.1296, "step": 2636 }, { "epoch": 0.48, "learning_rate": 1.9951434779022344e-05, "loss": 0.1187, "step": 2637 }, { "epoch": 0.48, "learning_rate": 1.9951361670462008e-05, "loss": 0.1032, "step": 2638 }, { "epoch": 0.48, "learning_rate": 1.9951288507049532e-05, "loss": 0.0996, "step": 2639 }, { "epoch": 0.48, "learning_rate": 1.995121528878532e-05, "loss": 0.106, "step": 2640 }, { "epoch": 0.48, "learning_rate": 1.995114201566978e-05, "loss": 0.1082, "step": 2641 }, { "epoch": 0.48, "learning_rate": 1.9951068687703307e-05, "loss": 0.103, "step": 2642 }, { "epoch": 0.48, "learning_rate": 1.995099530488631e-05, "loss": 0.1189, "step": 2643 }, { "epoch": 0.48, "learning_rate": 1.9950921867219196e-05, "loss": 0.1044, "step": 2644 }, { "epoch": 0.49, "learning_rate": 1.9950848374702368e-05, "loss": 0.1097, "step": 2645 }, { "epoch": 0.49, "learning_rate": 1.995077482733623e-05, "loss": 0.1313, "step": 2646 }, { "epoch": 0.49, "learning_rate": 1.995070122512119e-05, "loss": 0.1041, "step": 2647 }, { "epoch": 0.49, "learning_rate": 1.9950627568057647e-05, "loss": 0.115, "step": 2648 }, { "epoch": 0.49, "learning_rate": 1.9950553856146015e-05, "loss": 0.1181, "step": 2649 }, { "epoch": 0.49, "learning_rate": 1.9950480089386696e-05, "loss": 0.125, "step": 2650 }, { "epoch": 0.49, "learning_rate": 1.99504062677801e-05, "loss": 0.105, "step": 2651 }, { "epoch": 0.49, "learning_rate": 1.995033239132663e-05, "loss": 0.1055, "step": 2652 }, { "epoch": 0.49, "learning_rate": 1.9950258460026693e-05, "loss": 0.1036, "step": 2653 }, { "epoch": 0.49, "learning_rate": 1.99501844738807e-05, "loss": 0.1192, "step": 2654 }, { "epoch": 0.49, "learning_rate": 1.9950110432889057e-05, "loss": 0.1263, "step": 2655 }, { "epoch": 0.49, "learning_rate": 1.9950036337052173e-05, "loss": 0.114, "step": 2656 }, { "epoch": 0.49, "learning_rate": 1.9949962186370454e-05, "loss": 0.119, "step": 2657 }, { "epoch": 0.49, "learning_rate": 1.994988798084431e-05, "loss": 0.121, "step": 2658 }, { "epoch": 0.49, "learning_rate": 1.9949813720474152e-05, "loss": 0.1208, "step": 2659 }, { "epoch": 0.49, "learning_rate": 1.9949739405260386e-05, "loss": 0.1088, "step": 2660 }, { "epoch": 0.49, "learning_rate": 1.9949665035203425e-05, "loss": 0.1198, "step": 2661 }, { "epoch": 0.49, "learning_rate": 1.9949590610303677e-05, "loss": 0.1158, "step": 2662 }, { "epoch": 0.49, "learning_rate": 1.9949516130561553e-05, "loss": 0.1139, "step": 2663 }, { "epoch": 0.49, "learning_rate": 1.9949441595977462e-05, "loss": 0.1222, "step": 2664 }, { "epoch": 0.49, "learning_rate": 1.9949367006551816e-05, "loss": 0.1087, "step": 2665 }, { "epoch": 0.49, "learning_rate": 1.9949292362285025e-05, "loss": 0.128, "step": 2666 }, { "epoch": 0.49, "learning_rate": 1.99492176631775e-05, "loss": 0.1177, "step": 2667 }, { "epoch": 0.49, "learning_rate": 1.994914290922966e-05, "loss": 0.1298, "step": 2668 }, { "epoch": 0.49, "learning_rate": 1.9949068100441903e-05, "loss": 0.1127, "step": 2669 }, { "epoch": 0.49, "learning_rate": 1.9948993236814656e-05, "loss": 0.112, "step": 2670 }, { "epoch": 0.49, "learning_rate": 1.994891831834832e-05, "loss": 0.122, "step": 2671 }, { "epoch": 0.49, "learning_rate": 1.994884334504332e-05, "loss": 0.1183, "step": 2672 }, { "epoch": 0.49, "learning_rate": 1.9948768316900057e-05, "loss": 0.1062, "step": 2673 }, { "epoch": 0.49, "learning_rate": 1.994869323391895e-05, "loss": 0.1205, "step": 2674 }, { "epoch": 0.49, "learning_rate": 1.9948618096100418e-05, "loss": 0.1171, "step": 2675 }, { "epoch": 0.49, "learning_rate": 1.9948542903444866e-05, "loss": 0.1119, "step": 2676 }, { "epoch": 0.49, "learning_rate": 1.9948467655952713e-05, "loss": 0.1074, "step": 2677 }, { "epoch": 0.49, "learning_rate": 1.9948392353624375e-05, "loss": 0.1069, "step": 2678 }, { "epoch": 0.49, "learning_rate": 1.9948316996460263e-05, "loss": 0.1239, "step": 2679 }, { "epoch": 0.49, "learning_rate": 1.9948241584460796e-05, "loss": 0.1088, "step": 2680 }, { "epoch": 0.49, "learning_rate": 1.9948166117626385e-05, "loss": 0.1279, "step": 2681 }, { "epoch": 0.49, "learning_rate": 1.9948090595957453e-05, "loss": 0.1137, "step": 2682 }, { "epoch": 0.49, "learning_rate": 1.9948015019454413e-05, "loss": 0.1263, "step": 2683 }, { "epoch": 0.49, "learning_rate": 1.9947939388117675e-05, "loss": 0.1317, "step": 2684 }, { "epoch": 0.49, "learning_rate": 1.994786370194767e-05, "loss": 0.1099, "step": 2685 }, { "epoch": 0.49, "learning_rate": 1.99477879609448e-05, "loss": 0.115, "step": 2686 }, { "epoch": 0.49, "learning_rate": 1.9947712165109495e-05, "loss": 0.1188, "step": 2687 }, { "epoch": 0.49, "learning_rate": 1.9947636314442168e-05, "loss": 0.1154, "step": 2688 }, { "epoch": 0.49, "learning_rate": 1.9947560408943234e-05, "loss": 0.1282, "step": 2689 }, { "epoch": 0.49, "learning_rate": 1.9947484448613116e-05, "loss": 0.1025, "step": 2690 }, { "epoch": 0.49, "learning_rate": 1.9947408433452228e-05, "loss": 0.1173, "step": 2691 }, { "epoch": 0.49, "learning_rate": 1.9947332363460993e-05, "loss": 0.1253, "step": 2692 }, { "epoch": 0.49, "learning_rate": 1.9947256238639828e-05, "loss": 0.1106, "step": 2693 }, { "epoch": 0.49, "learning_rate": 1.9947180058989156e-05, "loss": 0.1258, "step": 2694 }, { "epoch": 0.49, "learning_rate": 1.994710382450939e-05, "loss": 0.1122, "step": 2695 }, { "epoch": 0.49, "learning_rate": 1.994702753520096e-05, "loss": 0.1204, "step": 2696 }, { "epoch": 0.49, "learning_rate": 1.9946951191064277e-05, "loss": 0.1152, "step": 2697 }, { "epoch": 0.49, "learning_rate": 1.994687479209977e-05, "loss": 0.1227, "step": 2698 }, { "epoch": 0.49, "learning_rate": 1.994679833830785e-05, "loss": 0.1077, "step": 2699 }, { "epoch": 0.5, "learning_rate": 1.9946721829688952e-05, "loss": 0.1302, "step": 2700 }, { "epoch": 0.5, "learning_rate": 1.9946645266243485e-05, "loss": 0.1338, "step": 2701 }, { "epoch": 0.5, "learning_rate": 1.994656864797188e-05, "loss": 0.1209, "step": 2702 }, { "epoch": 0.5, "learning_rate": 1.9946491974874554e-05, "loss": 0.113, "step": 2703 }, { "epoch": 0.5, "learning_rate": 1.9946415246951928e-05, "loss": 0.1165, "step": 2704 }, { "epoch": 0.5, "learning_rate": 1.994633846420443e-05, "loss": 0.1274, "step": 2705 }, { "epoch": 0.5, "learning_rate": 1.9946261626632483e-05, "loss": 0.1133, "step": 2706 }, { "epoch": 0.5, "learning_rate": 1.994618473423651e-05, "loss": 0.1271, "step": 2707 }, { "epoch": 0.5, "learning_rate": 1.994610778701693e-05, "loss": 0.1106, "step": 2708 }, { "epoch": 0.5, "learning_rate": 1.9946030784974174e-05, "loss": 0.1094, "step": 2709 }, { "epoch": 0.5, "learning_rate": 1.994595372810866e-05, "loss": 0.1249, "step": 2710 }, { "epoch": 0.5, "learning_rate": 1.9945876616420817e-05, "loss": 0.1118, "step": 2711 }, { "epoch": 0.5, "learning_rate": 1.9945799449911073e-05, "loss": 0.1113, "step": 2712 }, { "epoch": 0.5, "learning_rate": 1.994572222857985e-05, "loss": 0.1188, "step": 2713 }, { "epoch": 0.5, "learning_rate": 1.9945644952427568e-05, "loss": 0.1214, "step": 2714 }, { "epoch": 0.5, "learning_rate": 1.9945567621454657e-05, "loss": 0.1197, "step": 2715 }, { "epoch": 0.5, "learning_rate": 1.9945490235661547e-05, "loss": 0.1211, "step": 2716 }, { "epoch": 0.5, "learning_rate": 1.994541279504866e-05, "loss": 0.1087, "step": 2717 }, { "epoch": 0.5, "learning_rate": 1.9945335299616427e-05, "loss": 0.1238, "step": 2718 }, { "epoch": 0.5, "learning_rate": 1.9945257749365274e-05, "loss": 0.1134, "step": 2719 }, { "epoch": 0.5, "learning_rate": 1.9945180144295626e-05, "loss": 0.1231, "step": 2720 }, { "epoch": 0.5, "learning_rate": 1.9945102484407912e-05, "loss": 0.1114, "step": 2721 }, { "epoch": 0.5, "learning_rate": 1.9945024769702558e-05, "loss": 0.1159, "step": 2722 }, { "epoch": 0.5, "learning_rate": 1.9944947000179996e-05, "loss": 0.1224, "step": 2723 }, { "epoch": 0.5, "learning_rate": 1.9944869175840656e-05, "loss": 0.13, "step": 2724 }, { "epoch": 0.5, "learning_rate": 1.994479129668496e-05, "loss": 0.1161, "step": 2725 }, { "epoch": 0.5, "learning_rate": 1.9944713362713343e-05, "loss": 0.1167, "step": 2726 }, { "epoch": 0.5, "learning_rate": 1.9944635373926233e-05, "loss": 0.1137, "step": 2727 }, { "epoch": 0.5, "learning_rate": 1.994455733032406e-05, "loss": 0.1143, "step": 2728 }, { "epoch": 0.5, "learning_rate": 1.9944479231907254e-05, "loss": 0.1201, "step": 2729 }, { "epoch": 0.5, "learning_rate": 1.9944401078676243e-05, "loss": 0.1035, "step": 2730 }, { "epoch": 0.5, "learning_rate": 1.9944322870631466e-05, "loss": 0.1104, "step": 2731 }, { "epoch": 0.5, "learning_rate": 1.994424460777334e-05, "loss": 0.121, "step": 2732 }, { "epoch": 0.5, "learning_rate": 1.9944166290102315e-05, "loss": 0.1299, "step": 2733 }, { "epoch": 0.5, "learning_rate": 1.9944087917618807e-05, "loss": 0.1136, "step": 2734 }, { "epoch": 0.5, "learning_rate": 1.9944009490323252e-05, "loss": 0.1242, "step": 2735 }, { "epoch": 0.5, "learning_rate": 1.9943931008216083e-05, "loss": 0.1223, "step": 2736 }, { "epoch": 0.5, "learning_rate": 1.9943852471297737e-05, "loss": 0.1131, "step": 2737 }, { "epoch": 0.5, "learning_rate": 1.994377387956864e-05, "loss": 0.1093, "step": 2738 }, { "epoch": 0.5, "learning_rate": 1.9943695233029233e-05, "loss": 0.1264, "step": 2739 }, { "epoch": 0.5, "learning_rate": 1.9943616531679942e-05, "loss": 0.1325, "step": 2740 }, { "epoch": 0.5, "learning_rate": 1.9943537775521206e-05, "loss": 0.1303, "step": 2741 }, { "epoch": 0.5, "learning_rate": 1.9943458964553454e-05, "loss": 0.1125, "step": 2742 }, { "epoch": 0.5, "learning_rate": 1.994338009877712e-05, "loss": 0.1153, "step": 2743 }, { "epoch": 0.5, "learning_rate": 1.9943301178192647e-05, "loss": 0.1314, "step": 2744 }, { "epoch": 0.5, "learning_rate": 1.9943222202800463e-05, "loss": 0.1238, "step": 2745 }, { "epoch": 0.5, "learning_rate": 1.9943143172601008e-05, "loss": 0.1138, "step": 2746 }, { "epoch": 0.5, "learning_rate": 1.9943064087594713e-05, "loss": 0.1081, "step": 2747 }, { "epoch": 0.5, "learning_rate": 1.9942984947782012e-05, "loss": 0.1283, "step": 2748 }, { "epoch": 0.5, "learning_rate": 1.9942905753163352e-05, "loss": 0.1216, "step": 2749 }, { "epoch": 0.5, "learning_rate": 1.994282650373916e-05, "loss": 0.1246, "step": 2750 }, { "epoch": 0.5, "learning_rate": 1.9942747199509873e-05, "loss": 0.113, "step": 2751 }, { "epoch": 0.5, "learning_rate": 1.9942667840475934e-05, "loss": 0.1179, "step": 2752 }, { "epoch": 0.5, "learning_rate": 1.9942588426637775e-05, "loss": 0.1202, "step": 2753 }, { "epoch": 0.51, "learning_rate": 1.9942508957995835e-05, "loss": 0.132, "step": 2754 }, { "epoch": 0.51, "learning_rate": 1.9942429434550555e-05, "loss": 0.1254, "step": 2755 }, { "epoch": 0.51, "learning_rate": 1.994234985630237e-05, "loss": 0.107, "step": 2756 }, { "epoch": 0.51, "learning_rate": 1.9942270223251717e-05, "loss": 0.1228, "step": 2757 }, { "epoch": 0.51, "learning_rate": 1.994219053539904e-05, "loss": 0.1202, "step": 2758 }, { "epoch": 0.51, "learning_rate": 1.9942110792744777e-05, "loss": 0.1091, "step": 2759 }, { "epoch": 0.51, "learning_rate": 1.9942030995289363e-05, "loss": 0.1251, "step": 2760 }, { "epoch": 0.51, "learning_rate": 1.9941951143033242e-05, "loss": 0.1065, "step": 2761 }, { "epoch": 0.51, "learning_rate": 1.9941871235976857e-05, "loss": 0.1136, "step": 2762 }, { "epoch": 0.51, "learning_rate": 1.9941791274120643e-05, "loss": 0.1105, "step": 2763 }, { "epoch": 0.51, "learning_rate": 1.9941711257465042e-05, "loss": 0.116, "step": 2764 }, { "epoch": 0.51, "learning_rate": 1.9941631186010496e-05, "loss": 0.1217, "step": 2765 }, { "epoch": 0.51, "learning_rate": 1.9941551059757444e-05, "loss": 0.1083, "step": 2766 }, { "epoch": 0.51, "learning_rate": 1.994147087870633e-05, "loss": 0.1321, "step": 2767 }, { "epoch": 0.51, "learning_rate": 1.9941390642857598e-05, "loss": 0.1085, "step": 2768 }, { "epoch": 0.51, "learning_rate": 1.9941310352211687e-05, "loss": 0.1256, "step": 2769 }, { "epoch": 0.51, "learning_rate": 1.994123000676904e-05, "loss": 0.1181, "step": 2770 }, { "epoch": 0.51, "learning_rate": 1.9941149606530104e-05, "loss": 0.1212, "step": 2771 }, { "epoch": 0.51, "learning_rate": 1.9941069151495315e-05, "loss": 0.123, "step": 2772 }, { "epoch": 0.51, "learning_rate": 1.9940988641665118e-05, "loss": 0.1118, "step": 2773 }, { "epoch": 0.51, "learning_rate": 1.9940908077039963e-05, "loss": 0.13, "step": 2774 }, { "epoch": 0.51, "learning_rate": 1.9940827457620284e-05, "loss": 0.1093, "step": 2775 }, { "epoch": 0.51, "learning_rate": 1.9940746783406535e-05, "loss": 0.1183, "step": 2776 }, { "epoch": 0.51, "learning_rate": 1.9940666054399157e-05, "loss": 0.1188, "step": 2777 }, { "epoch": 0.51, "learning_rate": 1.9940585270598593e-05, "loss": 0.1047, "step": 2778 }, { "epoch": 0.51, "learning_rate": 1.9940504432005293e-05, "loss": 0.1269, "step": 2779 }, { "epoch": 0.51, "learning_rate": 1.99404235386197e-05, "loss": 0.1205, "step": 2780 }, { "epoch": 0.51, "learning_rate": 1.9940342590442257e-05, "loss": 0.1142, "step": 2781 }, { "epoch": 0.51, "learning_rate": 1.9940261587473412e-05, "loss": 0.1075, "step": 2782 }, { "epoch": 0.51, "learning_rate": 1.9940180529713613e-05, "loss": 0.1243, "step": 2783 }, { "epoch": 0.51, "learning_rate": 1.994009941716331e-05, "loss": 0.1175, "step": 2784 }, { "epoch": 0.51, "learning_rate": 1.994001824982294e-05, "loss": 0.1065, "step": 2785 }, { "epoch": 0.51, "learning_rate": 1.993993702769296e-05, "loss": 0.1196, "step": 2786 }, { "epoch": 0.51, "learning_rate": 1.9939855750773815e-05, "loss": 0.1199, "step": 2787 }, { "epoch": 0.51, "learning_rate": 1.9939774419065953e-05, "loss": 0.1206, "step": 2788 }, { "epoch": 0.51, "learning_rate": 1.993969303256982e-05, "loss": 0.1189, "step": 2789 }, { "epoch": 0.51, "learning_rate": 1.993961159128586e-05, "loss": 0.1164, "step": 2790 }, { "epoch": 0.51, "learning_rate": 1.9939530095214536e-05, "loss": 0.1203, "step": 2791 }, { "epoch": 0.51, "learning_rate": 1.9939448544356287e-05, "loss": 0.1112, "step": 2792 }, { "epoch": 0.51, "learning_rate": 1.9939366938711563e-05, "loss": 0.1178, "step": 2793 }, { "epoch": 0.51, "learning_rate": 1.993928527828082e-05, "loss": 0.1084, "step": 2794 }, { "epoch": 0.51, "learning_rate": 1.99392035630645e-05, "loss": 0.1272, "step": 2795 }, { "epoch": 0.51, "learning_rate": 1.9939121793063058e-05, "loss": 0.1176, "step": 2796 }, { "epoch": 0.51, "learning_rate": 1.9939039968276942e-05, "loss": 0.1154, "step": 2797 }, { "epoch": 0.51, "learning_rate": 1.993895808870661e-05, "loss": 0.0962, "step": 2798 }, { "epoch": 0.51, "learning_rate": 1.9938876154352504e-05, "loss": 0.1226, "step": 2799 }, { "epoch": 0.51, "learning_rate": 1.993879416521508e-05, "loss": 0.11, "step": 2800 }, { "epoch": 0.51, "learning_rate": 1.9938712121294788e-05, "loss": 0.1025, "step": 2801 }, { "epoch": 0.51, "learning_rate": 1.9938630022592088e-05, "loss": 0.1046, "step": 2802 }, { "epoch": 0.51, "learning_rate": 1.9938547869107423e-05, "loss": 0.1122, "step": 2803 }, { "epoch": 0.51, "learning_rate": 1.9938465660841248e-05, "loss": 0.1098, "step": 2804 }, { "epoch": 0.51, "learning_rate": 1.993838339779402e-05, "loss": 0.1086, "step": 2805 }, { "epoch": 0.51, "learning_rate": 1.9938301079966186e-05, "loss": 0.1262, "step": 2806 }, { "epoch": 0.51, "learning_rate": 1.9938218707358208e-05, "loss": 0.1168, "step": 2807 }, { "epoch": 0.51, "learning_rate": 1.9938136279970535e-05, "loss": 0.1141, "step": 2808 }, { "epoch": 0.52, "learning_rate": 1.993805379780362e-05, "loss": 0.113, "step": 2809 }, { "epoch": 0.52, "learning_rate": 1.993797126085792e-05, "loss": 0.1184, "step": 2810 }, { "epoch": 0.52, "learning_rate": 1.993788866913389e-05, "loss": 0.1134, "step": 2811 }, { "epoch": 0.52, "learning_rate": 1.9937806022631986e-05, "loss": 0.1088, "step": 2812 }, { "epoch": 0.52, "learning_rate": 1.993772332135266e-05, "loss": 0.1157, "step": 2813 }, { "epoch": 0.52, "learning_rate": 1.9937640565296372e-05, "loss": 0.117, "step": 2814 }, { "epoch": 0.52, "learning_rate": 1.9937557754463575e-05, "loss": 0.1076, "step": 2815 }, { "epoch": 0.52, "learning_rate": 1.9937474888854727e-05, "loss": 0.116, "step": 2816 }, { "epoch": 0.52, "learning_rate": 1.9937391968470284e-05, "loss": 0.1033, "step": 2817 }, { "epoch": 0.52, "learning_rate": 1.9937308993310707e-05, "loss": 0.1255, "step": 2818 }, { "epoch": 0.52, "learning_rate": 1.9937225963376447e-05, "loss": 0.1228, "step": 2819 }, { "epoch": 0.52, "learning_rate": 1.9937142878667965e-05, "loss": 0.1195, "step": 2820 }, { "epoch": 0.52, "learning_rate": 1.993705973918572e-05, "loss": 0.1148, "step": 2821 }, { "epoch": 0.52, "learning_rate": 1.993697654493017e-05, "loss": 0.1226, "step": 2822 }, { "epoch": 0.52, "learning_rate": 1.9936893295901768e-05, "loss": 0.1024, "step": 2823 }, { "epoch": 0.52, "learning_rate": 1.9936809992100983e-05, "loss": 0.129, "step": 2824 }, { "epoch": 0.52, "learning_rate": 1.9936726633528264e-05, "loss": 0.113, "step": 2825 }, { "epoch": 0.52, "learning_rate": 1.9936643220184076e-05, "loss": 0.1011, "step": 2826 }, { "epoch": 0.52, "learning_rate": 1.9936559752068878e-05, "loss": 0.1079, "step": 2827 }, { "epoch": 0.52, "learning_rate": 1.9936476229183133e-05, "loss": 0.1082, "step": 2828 }, { "epoch": 0.52, "learning_rate": 1.9936392651527293e-05, "loss": 0.1101, "step": 2829 }, { "epoch": 0.52, "learning_rate": 1.993630901910183e-05, "loss": 0.1106, "step": 2830 }, { "epoch": 0.52, "learning_rate": 1.9936225331907193e-05, "loss": 0.1211, "step": 2831 }, { "epoch": 0.52, "learning_rate": 1.993614158994385e-05, "loss": 0.1018, "step": 2832 }, { "epoch": 0.52, "learning_rate": 1.9936057793212263e-05, "loss": 0.1136, "step": 2833 }, { "epoch": 0.52, "learning_rate": 1.9935973941712894e-05, "loss": 0.1116, "step": 2834 }, { "epoch": 0.52, "learning_rate": 1.9935890035446204e-05, "loss": 0.1249, "step": 2835 }, { "epoch": 0.52, "learning_rate": 1.9935806074412654e-05, "loss": 0.1085, "step": 2836 }, { "epoch": 0.52, "learning_rate": 1.9935722058612708e-05, "loss": 0.1119, "step": 2837 }, { "epoch": 0.52, "learning_rate": 1.993563798804683e-05, "loss": 0.1123, "step": 2838 }, { "epoch": 0.52, "learning_rate": 1.993555386271548e-05, "loss": 0.1104, "step": 2839 }, { "epoch": 0.52, "learning_rate": 1.993546968261913e-05, "loss": 0.1225, "step": 2840 }, { "epoch": 0.52, "learning_rate": 1.9935385447758233e-05, "loss": 0.1107, "step": 2841 }, { "epoch": 0.52, "learning_rate": 1.993530115813326e-05, "loss": 0.1161, "step": 2842 }, { "epoch": 0.52, "learning_rate": 1.9935216813744677e-05, "loss": 0.1067, "step": 2843 }, { "epoch": 0.52, "learning_rate": 1.993513241459294e-05, "loss": 0.1193, "step": 2844 }, { "epoch": 0.52, "learning_rate": 1.9935047960678528e-05, "loss": 0.1293, "step": 2845 }, { "epoch": 0.52, "learning_rate": 1.9934963452001894e-05, "loss": 0.123, "step": 2846 }, { "epoch": 0.52, "learning_rate": 1.993487888856351e-05, "loss": 0.1273, "step": 2847 }, { "epoch": 0.52, "learning_rate": 1.993479427036384e-05, "loss": 0.1058, "step": 2848 }, { "epoch": 0.52, "learning_rate": 1.993470959740335e-05, "loss": 0.1203, "step": 2849 }, { "epoch": 0.52, "learning_rate": 1.9934624869682513e-05, "loss": 0.1214, "step": 2850 }, { "epoch": 0.52, "learning_rate": 1.9934540087201788e-05, "loss": 0.1141, "step": 2851 }, { "epoch": 0.52, "learning_rate": 1.9934455249961643e-05, "loss": 0.1259, "step": 2852 }, { "epoch": 0.52, "learning_rate": 1.9934370357962552e-05, "loss": 0.1189, "step": 2853 }, { "epoch": 0.52, "learning_rate": 1.9934285411204977e-05, "loss": 0.126, "step": 2854 }, { "epoch": 0.52, "learning_rate": 1.9934200409689385e-05, "loss": 0.1283, "step": 2855 }, { "epoch": 0.52, "learning_rate": 1.993411535341625e-05, "loss": 0.114, "step": 2856 }, { "epoch": 0.52, "learning_rate": 1.993403024238604e-05, "loss": 0.1197, "step": 2857 }, { "epoch": 0.52, "learning_rate": 1.9933945076599224e-05, "loss": 0.1266, "step": 2858 }, { "epoch": 0.52, "learning_rate": 1.9933859856056264e-05, "loss": 0.1183, "step": 2859 }, { "epoch": 0.52, "learning_rate": 1.9933774580757645e-05, "loss": 0.1177, "step": 2860 }, { "epoch": 0.52, "learning_rate": 1.993368925070382e-05, "loss": 0.1091, "step": 2861 }, { "epoch": 0.52, "learning_rate": 1.9933603865895266e-05, "loss": 0.1196, "step": 2862 }, { "epoch": 0.53, "learning_rate": 1.9933518426332463e-05, "loss": 0.1223, "step": 2863 }, { "epoch": 0.53, "learning_rate": 1.9933432932015867e-05, "loss": 0.1218, "step": 2864 }, { "epoch": 0.53, "learning_rate": 1.9933347382945957e-05, "loss": 0.1031, "step": 2865 }, { "epoch": 0.53, "learning_rate": 1.9933261779123207e-05, "loss": 0.1213, "step": 2866 }, { "epoch": 0.53, "learning_rate": 1.993317612054808e-05, "loss": 0.1193, "step": 2867 }, { "epoch": 0.53, "learning_rate": 1.9933090407221057e-05, "loss": 0.1066, "step": 2868 }, { "epoch": 0.53, "learning_rate": 1.9933004639142606e-05, "loss": 0.1201, "step": 2869 }, { "epoch": 0.53, "learning_rate": 1.9932918816313203e-05, "loss": 0.1315, "step": 2870 }, { "epoch": 0.53, "learning_rate": 1.9932832938733314e-05, "loss": 0.121, "step": 2871 }, { "epoch": 0.53, "learning_rate": 1.9932747006403418e-05, "loss": 0.1137, "step": 2872 }, { "epoch": 0.53, "learning_rate": 1.9932661019323993e-05, "loss": 0.1151, "step": 2873 }, { "epoch": 0.53, "learning_rate": 1.9932574977495505e-05, "loss": 0.1134, "step": 2874 }, { "epoch": 0.53, "learning_rate": 1.993248888091843e-05, "loss": 0.1194, "step": 2875 }, { "epoch": 0.53, "learning_rate": 1.9932402729593243e-05, "loss": 0.1109, "step": 2876 }, { "epoch": 0.53, "learning_rate": 1.9932316523520422e-05, "loss": 0.127, "step": 2877 }, { "epoch": 0.53, "learning_rate": 1.9932230262700435e-05, "loss": 0.105, "step": 2878 }, { "epoch": 0.53, "learning_rate": 1.9932143947133766e-05, "loss": 0.1176, "step": 2879 }, { "epoch": 0.53, "learning_rate": 1.9932057576820886e-05, "loss": 0.1174, "step": 2880 }, { "epoch": 0.53, "learning_rate": 1.9931971151762272e-05, "loss": 0.1269, "step": 2881 }, { "epoch": 0.53, "learning_rate": 1.99318846719584e-05, "loss": 0.1142, "step": 2882 }, { "epoch": 0.53, "learning_rate": 1.993179813740975e-05, "loss": 0.1202, "step": 2883 }, { "epoch": 0.53, "learning_rate": 1.993171154811679e-05, "loss": 0.1085, "step": 2884 }, { "epoch": 0.53, "learning_rate": 1.993162490408001e-05, "loss": 0.1227, "step": 2885 }, { "epoch": 0.53, "learning_rate": 1.9931538205299876e-05, "loss": 0.1136, "step": 2886 }, { "epoch": 0.53, "learning_rate": 1.993145145177687e-05, "loss": 0.1215, "step": 2887 }, { "epoch": 0.53, "learning_rate": 1.9931364643511474e-05, "loss": 0.1074, "step": 2888 }, { "epoch": 0.53, "learning_rate": 1.9931277780504165e-05, "loss": 0.1092, "step": 2889 }, { "epoch": 0.53, "learning_rate": 1.9931190862755416e-05, "loss": 0.1266, "step": 2890 }, { "epoch": 0.53, "learning_rate": 1.9931103890265713e-05, "loss": 0.1171, "step": 2891 }, { "epoch": 0.53, "learning_rate": 1.993101686303553e-05, "loss": 0.1115, "step": 2892 }, { "epoch": 0.53, "learning_rate": 1.9930929781065352e-05, "loss": 0.1046, "step": 2893 }, { "epoch": 0.53, "learning_rate": 1.9930842644355656e-05, "loss": 0.1163, "step": 2894 }, { "epoch": 0.53, "learning_rate": 1.993075545290692e-05, "loss": 0.1022, "step": 2895 }, { "epoch": 0.53, "learning_rate": 1.9930668206719632e-05, "loss": 0.1232, "step": 2896 }, { "epoch": 0.53, "learning_rate": 1.9930580905794263e-05, "loss": 0.11, "step": 2897 }, { "epoch": 0.53, "learning_rate": 1.9930493550131303e-05, "loss": 0.1131, "step": 2898 }, { "epoch": 0.53, "learning_rate": 1.993040613973123e-05, "loss": 0.1142, "step": 2899 }, { "epoch": 0.53, "learning_rate": 1.9930318674594526e-05, "loss": 0.1167, "step": 2900 }, { "epoch": 0.53, "learning_rate": 1.993023115472167e-05, "loss": 0.1216, "step": 2901 }, { "epoch": 0.53, "learning_rate": 1.9930143580113152e-05, "loss": 0.1251, "step": 2902 }, { "epoch": 0.53, "learning_rate": 1.9930055950769446e-05, "loss": 0.1248, "step": 2903 }, { "epoch": 0.53, "learning_rate": 1.9929968266691037e-05, "loss": 0.1244, "step": 2904 }, { "epoch": 0.53, "learning_rate": 1.9929880527878413e-05, "loss": 0.0966, "step": 2905 }, { "epoch": 0.53, "learning_rate": 1.9929792734332058e-05, "loss": 0.1287, "step": 2906 }, { "epoch": 0.53, "learning_rate": 1.9929704886052446e-05, "loss": 0.1217, "step": 2907 }, { "epoch": 0.53, "learning_rate": 1.9929616983040076e-05, "loss": 0.1093, "step": 2908 }, { "epoch": 0.53, "learning_rate": 1.9929529025295418e-05, "loss": 0.106, "step": 2909 }, { "epoch": 0.53, "learning_rate": 1.9929441012818964e-05, "loss": 0.1255, "step": 2910 }, { "epoch": 0.53, "learning_rate": 1.9929352945611202e-05, "loss": 0.1238, "step": 2911 }, { "epoch": 0.53, "learning_rate": 1.9929264823672613e-05, "loss": 0.124, "step": 2912 }, { "epoch": 0.53, "learning_rate": 1.992917664700368e-05, "loss": 0.1077, "step": 2913 }, { "epoch": 0.53, "learning_rate": 1.9929088415604897e-05, "loss": 0.108, "step": 2914 }, { "epoch": 0.53, "learning_rate": 1.9929000129476743e-05, "loss": 0.119, "step": 2915 }, { "epoch": 0.53, "learning_rate": 1.992891178861971e-05, "loss": 0.1048, "step": 2916 }, { "epoch": 0.53, "learning_rate": 1.9928823393034283e-05, "loss": 0.1142, "step": 2917 }, { "epoch": 0.54, "learning_rate": 1.9928734942720946e-05, "loss": 0.1209, "step": 2918 }, { "epoch": 0.54, "learning_rate": 1.992864643768019e-05, "loss": 0.1311, "step": 2919 }, { "epoch": 0.54, "learning_rate": 1.9928557877912506e-05, "loss": 0.101, "step": 2920 }, { "epoch": 0.54, "learning_rate": 1.9928469263418376e-05, "loss": 0.1176, "step": 2921 }, { "epoch": 0.54, "learning_rate": 1.9928380594198287e-05, "loss": 0.1211, "step": 2922 }, { "epoch": 0.54, "learning_rate": 1.9928291870252735e-05, "loss": 0.1013, "step": 2923 }, { "epoch": 0.54, "learning_rate": 1.9928203091582207e-05, "loss": 0.1181, "step": 2924 }, { "epoch": 0.54, "learning_rate": 1.992811425818719e-05, "loss": 0.1313, "step": 2925 }, { "epoch": 0.54, "learning_rate": 1.9928025370068174e-05, "loss": 0.1176, "step": 2926 }, { "epoch": 0.54, "learning_rate": 1.992793642722565e-05, "loss": 0.1161, "step": 2927 }, { "epoch": 0.54, "learning_rate": 1.992784742966011e-05, "loss": 0.112, "step": 2928 }, { "epoch": 0.54, "learning_rate": 1.9927758377372037e-05, "loss": 0.1132, "step": 2929 }, { "epoch": 0.54, "learning_rate": 1.9927669270361933e-05, "loss": 0.1127, "step": 2930 }, { "epoch": 0.54, "learning_rate": 1.992758010863028e-05, "loss": 0.1102, "step": 2931 }, { "epoch": 0.54, "learning_rate": 1.9927490892177572e-05, "loss": 0.1147, "step": 2932 }, { "epoch": 0.54, "learning_rate": 1.9927401621004307e-05, "loss": 0.119, "step": 2933 }, { "epoch": 0.54, "learning_rate": 1.9927312295110966e-05, "loss": 0.1258, "step": 2934 }, { "epoch": 0.54, "learning_rate": 1.992722291449805e-05, "loss": 0.1162, "step": 2935 }, { "epoch": 0.54, "learning_rate": 1.9927133479166045e-05, "loss": 0.1209, "step": 2936 }, { "epoch": 0.54, "learning_rate": 1.992704398911545e-05, "loss": 0.1226, "step": 2937 }, { "epoch": 0.54, "learning_rate": 1.9926954444346757e-05, "loss": 0.1116, "step": 2938 }, { "epoch": 0.54, "learning_rate": 1.992686484486046e-05, "loss": 0.1157, "step": 2939 }, { "epoch": 0.54, "learning_rate": 1.992677519065705e-05, "loss": 0.1144, "step": 2940 }, { "epoch": 0.54, "learning_rate": 1.9926685481737018e-05, "loss": 0.1102, "step": 2941 }, { "epoch": 0.54, "learning_rate": 1.9926595718100868e-05, "loss": 0.1102, "step": 2942 }, { "epoch": 0.54, "learning_rate": 1.992650589974909e-05, "loss": 0.1152, "step": 2943 }, { "epoch": 0.54, "learning_rate": 1.9926416026682175e-05, "loss": 0.1108, "step": 2944 }, { "epoch": 0.54, "learning_rate": 1.9926326098900624e-05, "loss": 0.1217, "step": 2945 }, { "epoch": 0.54, "learning_rate": 1.9926236116404937e-05, "loss": 0.119, "step": 2946 }, { "epoch": 0.54, "learning_rate": 1.9926146079195597e-05, "loss": 0.0981, "step": 2947 }, { "epoch": 0.54, "learning_rate": 1.9926055987273108e-05, "loss": 0.113, "step": 2948 }, { "epoch": 0.54, "learning_rate": 1.992596584063797e-05, "loss": 0.1158, "step": 2949 }, { "epoch": 0.54, "learning_rate": 1.992587563929067e-05, "loss": 0.1191, "step": 2950 }, { "epoch": 0.54, "learning_rate": 1.9925785383231716e-05, "loss": 0.1072, "step": 2951 }, { "epoch": 0.54, "learning_rate": 1.9925695072461596e-05, "loss": 0.1262, "step": 2952 }, { "epoch": 0.54, "learning_rate": 1.9925604706980815e-05, "loss": 0.1273, "step": 2953 }, { "epoch": 0.54, "learning_rate": 1.9925514286789868e-05, "loss": 0.121, "step": 2954 }, { "epoch": 0.54, "learning_rate": 1.9925423811889252e-05, "loss": 0.1174, "step": 2955 }, { "epoch": 0.54, "learning_rate": 1.992533328227947e-05, "loss": 0.1204, "step": 2956 }, { "epoch": 0.54, "learning_rate": 1.992524269796102e-05, "loss": 0.1126, "step": 2957 }, { "epoch": 0.54, "learning_rate": 1.9925152058934393e-05, "loss": 0.1296, "step": 2958 }, { "epoch": 0.54, "learning_rate": 1.9925061365200102e-05, "loss": 0.1072, "step": 2959 }, { "epoch": 0.54, "learning_rate": 1.9924970616758636e-05, "loss": 0.121, "step": 2960 }, { "epoch": 0.54, "learning_rate": 1.99248798136105e-05, "loss": 0.1048, "step": 2961 }, { "epoch": 0.54, "learning_rate": 1.99247889557562e-05, "loss": 0.1064, "step": 2962 }, { "epoch": 0.54, "learning_rate": 1.992469804319622e-05, "loss": 0.1281, "step": 2963 }, { "epoch": 0.54, "learning_rate": 1.992460707593108e-05, "loss": 0.1222, "step": 2964 }, { "epoch": 0.54, "learning_rate": 1.992451605396127e-05, "loss": 0.1141, "step": 2965 }, { "epoch": 0.54, "learning_rate": 1.9924424977287296e-05, "loss": 0.1287, "step": 2966 }, { "epoch": 0.54, "learning_rate": 1.992433384590966e-05, "loss": 0.1208, "step": 2967 }, { "epoch": 0.54, "learning_rate": 1.9924242659828864e-05, "loss": 0.1223, "step": 2968 }, { "epoch": 0.54, "learning_rate": 1.9924151419045404e-05, "loss": 0.1115, "step": 2969 }, { "epoch": 0.54, "learning_rate": 1.9924060123559798e-05, "loss": 0.1174, "step": 2970 }, { "epoch": 0.54, "learning_rate": 1.9923968773372533e-05, "loss": 0.1217, "step": 2971 }, { "epoch": 0.55, "learning_rate": 1.992387736848412e-05, "loss": 0.1188, "step": 2972 }, { "epoch": 0.55, "learning_rate": 1.9923785908895066e-05, "loss": 0.1207, "step": 2973 }, { "epoch": 0.55, "learning_rate": 1.992369439460587e-05, "loss": 0.114, "step": 2974 }, { "epoch": 0.55, "learning_rate": 1.992360282561704e-05, "loss": 0.1032, "step": 2975 }, { "epoch": 0.55, "learning_rate": 1.9923511201929077e-05, "loss": 0.1197, "step": 2976 }, { "epoch": 0.55, "learning_rate": 1.9923419523542488e-05, "loss": 0.1187, "step": 2977 }, { "epoch": 0.55, "learning_rate": 1.9923327790457777e-05, "loss": 0.1122, "step": 2978 }, { "epoch": 0.55, "learning_rate": 1.992323600267545e-05, "loss": 0.0994, "step": 2979 }, { "epoch": 0.55, "learning_rate": 1.9923144160196014e-05, "loss": 0.1218, "step": 2980 }, { "epoch": 0.55, "learning_rate": 1.9923052263019977e-05, "loss": 0.1274, "step": 2981 }, { "epoch": 0.55, "learning_rate": 1.9922960311147842e-05, "loss": 0.1285, "step": 2982 }, { "epoch": 0.55, "learning_rate": 1.992286830458012e-05, "loss": 0.1173, "step": 2983 }, { "epoch": 0.55, "learning_rate": 1.992277624331731e-05, "loss": 0.1123, "step": 2984 }, { "epoch": 0.55, "learning_rate": 1.992268412735993e-05, "loss": 0.108, "step": 2985 }, { "epoch": 0.55, "learning_rate": 1.9922591956708478e-05, "loss": 0.1109, "step": 2986 }, { "epoch": 0.55, "learning_rate": 1.9922499731363468e-05, "loss": 0.1188, "step": 2987 }, { "epoch": 0.55, "learning_rate": 1.992240745132541e-05, "loss": 0.1179, "step": 2988 }, { "epoch": 0.55, "learning_rate": 1.992231511659481e-05, "loss": 0.1311, "step": 2989 }, { "epoch": 0.55, "learning_rate": 1.9922222727172175e-05, "loss": 0.1147, "step": 2990 }, { "epoch": 0.55, "learning_rate": 1.9922130283058013e-05, "loss": 0.1193, "step": 2991 }, { "epoch": 0.55, "learning_rate": 1.992203778425284e-05, "loss": 0.1151, "step": 2992 }, { "epoch": 0.55, "learning_rate": 1.992194523075716e-05, "loss": 0.1061, "step": 2993 }, { "epoch": 0.55, "learning_rate": 1.9921852622571488e-05, "loss": 0.1088, "step": 2994 }, { "epoch": 0.55, "learning_rate": 1.9921759959696326e-05, "loss": 0.127, "step": 2995 }, { "epoch": 0.55, "learning_rate": 1.99216672421322e-05, "loss": 0.1063, "step": 2996 }, { "epoch": 0.55, "learning_rate": 1.9921574469879606e-05, "loss": 0.1179, "step": 2997 }, { "epoch": 0.55, "learning_rate": 1.992148164293906e-05, "loss": 0.1182, "step": 2998 }, { "epoch": 0.55, "learning_rate": 1.9921388761311078e-05, "loss": 0.1186, "step": 2999 }, { "epoch": 0.55, "learning_rate": 1.9921295824996167e-05, "loss": 0.1067, "step": 3000 }, { "epoch": 0.55, "learning_rate": 1.992120283399484e-05, "loss": 0.1182, "step": 3001 }, { "epoch": 0.55, "learning_rate": 1.9921109788307614e-05, "loss": 0.1195, "step": 3002 }, { "epoch": 0.55, "learning_rate": 1.9921016687934994e-05, "loss": 0.1108, "step": 3003 }, { "epoch": 0.55, "learning_rate": 1.9920923532877504e-05, "loss": 0.1046, "step": 3004 }, { "epoch": 0.55, "learning_rate": 1.9920830323135645e-05, "loss": 0.1135, "step": 3005 }, { "epoch": 0.55, "learning_rate": 1.992073705870994e-05, "loss": 0.112, "step": 3006 }, { "epoch": 0.55, "learning_rate": 1.9920643739600897e-05, "loss": 0.1129, "step": 3007 }, { "epoch": 0.55, "learning_rate": 1.9920550365809034e-05, "loss": 0.118, "step": 3008 }, { "epoch": 0.55, "learning_rate": 1.992045693733487e-05, "loss": 0.1036, "step": 3009 }, { "epoch": 0.55, "learning_rate": 1.992036345417891e-05, "loss": 0.1202, "step": 3010 }, { "epoch": 0.55, "learning_rate": 1.9920269916341674e-05, "loss": 0.1164, "step": 3011 }, { "epoch": 0.55, "learning_rate": 1.9920176323823678e-05, "loss": 0.1251, "step": 3012 }, { "epoch": 0.55, "learning_rate": 1.992008267662544e-05, "loss": 0.1122, "step": 3013 }, { "epoch": 0.55, "learning_rate": 1.9919988974747473e-05, "loss": 0.1247, "step": 3014 }, { "epoch": 0.55, "learning_rate": 1.9919895218190292e-05, "loss": 0.1228, "step": 3015 }, { "epoch": 0.55, "learning_rate": 1.991980140695442e-05, "loss": 0.1136, "step": 3016 }, { "epoch": 0.55, "learning_rate": 1.9919707541040368e-05, "loss": 0.1125, "step": 3017 }, { "epoch": 0.55, "learning_rate": 1.9919613620448656e-05, "loss": 0.114, "step": 3018 }, { "epoch": 0.55, "learning_rate": 1.9919519645179802e-05, "loss": 0.106, "step": 3019 }, { "epoch": 0.55, "learning_rate": 1.9919425615234324e-05, "loss": 0.1154, "step": 3020 }, { "epoch": 0.55, "learning_rate": 1.9919331530612737e-05, "loss": 0.1132, "step": 3021 }, { "epoch": 0.55, "learning_rate": 1.9919237391315563e-05, "loss": 0.1083, "step": 3022 }, { "epoch": 0.55, "learning_rate": 1.991914319734332e-05, "loss": 0.1206, "step": 3023 }, { "epoch": 0.55, "learning_rate": 1.9919048948696527e-05, "loss": 0.125, "step": 3024 }, { "epoch": 0.55, "learning_rate": 1.9918954645375705e-05, "loss": 0.1073, "step": 3025 }, { "epoch": 0.55, "learning_rate": 1.9918860287381373e-05, "loss": 0.1091, "step": 3026 }, { "epoch": 0.56, "learning_rate": 1.991876587471405e-05, "loss": 0.1142, "step": 3027 }, { "epoch": 0.56, "learning_rate": 1.991867140737426e-05, "loss": 0.129, "step": 3028 }, { "epoch": 0.56, "learning_rate": 1.9918576885362517e-05, "loss": 0.1304, "step": 3029 }, { "epoch": 0.56, "learning_rate": 1.9918482308679345e-05, "loss": 0.1314, "step": 3030 }, { "epoch": 0.56, "learning_rate": 1.9918387677325268e-05, "loss": 0.1285, "step": 3031 }, { "epoch": 0.56, "learning_rate": 1.9918292991300805e-05, "loss": 0.116, "step": 3032 }, { "epoch": 0.56, "learning_rate": 1.991819825060648e-05, "loss": 0.113, "step": 3033 }, { "epoch": 0.56, "learning_rate": 1.9918103455242814e-05, "loss": 0.1062, "step": 3034 }, { "epoch": 0.56, "learning_rate": 1.991800860521033e-05, "loss": 0.1208, "step": 3035 }, { "epoch": 0.56, "learning_rate": 1.991791370050955e-05, "loss": 0.1128, "step": 3036 }, { "epoch": 0.56, "learning_rate": 1.9917818741140995e-05, "loss": 0.1078, "step": 3037 }, { "epoch": 0.56, "learning_rate": 1.991772372710519e-05, "loss": 0.1085, "step": 3038 }, { "epoch": 0.56, "learning_rate": 1.9917628658402664e-05, "loss": 0.1054, "step": 3039 }, { "epoch": 0.56, "learning_rate": 1.9917533535033932e-05, "loss": 0.1077, "step": 3040 }, { "epoch": 0.56, "learning_rate": 1.9917438356999526e-05, "loss": 0.116, "step": 3041 }, { "epoch": 0.56, "learning_rate": 1.9917343124299965e-05, "loss": 0.113, "step": 3042 }, { "epoch": 0.56, "learning_rate": 1.9917247836935777e-05, "loss": 0.1165, "step": 3043 }, { "epoch": 0.56, "learning_rate": 1.991715249490749e-05, "loss": 0.1168, "step": 3044 }, { "epoch": 0.56, "learning_rate": 1.9917057098215624e-05, "loss": 0.1139, "step": 3045 }, { "epoch": 0.56, "learning_rate": 1.9916961646860704e-05, "loss": 0.1218, "step": 3046 }, { "epoch": 0.56, "learning_rate": 1.9916866140843262e-05, "loss": 0.1253, "step": 3047 }, { "epoch": 0.56, "learning_rate": 1.991677058016382e-05, "loss": 0.121, "step": 3048 }, { "epoch": 0.56, "learning_rate": 1.9916674964822905e-05, "loss": 0.1011, "step": 3049 }, { "epoch": 0.56, "learning_rate": 1.9916579294821046e-05, "loss": 0.1166, "step": 3050 }, { "epoch": 0.56, "learning_rate": 1.9916483570158776e-05, "loss": 0.117, "step": 3051 }, { "epoch": 0.56, "learning_rate": 1.991638779083661e-05, "loss": 0.1093, "step": 3052 }, { "epoch": 0.56, "learning_rate": 1.9916291956855082e-05, "loss": 0.132, "step": 3053 }, { "epoch": 0.56, "learning_rate": 1.9916196068214724e-05, "loss": 0.1183, "step": 3054 }, { "epoch": 0.56, "learning_rate": 1.991610012491606e-05, "loss": 0.1045, "step": 3055 }, { "epoch": 0.56, "learning_rate": 1.9916004126959615e-05, "loss": 0.1066, "step": 3056 }, { "epoch": 0.56, "learning_rate": 1.9915908074345927e-05, "loss": 0.1154, "step": 3057 }, { "epoch": 0.56, "learning_rate": 1.9915811967075524e-05, "loss": 0.1092, "step": 3058 }, { "epoch": 0.56, "learning_rate": 1.9915715805148927e-05, "loss": 0.1201, "step": 3059 }, { "epoch": 0.56, "learning_rate": 1.9915619588566676e-05, "loss": 0.1072, "step": 3060 }, { "epoch": 0.56, "learning_rate": 1.99155233173293e-05, "loss": 0.1078, "step": 3061 }, { "epoch": 0.56, "learning_rate": 1.991542699143732e-05, "loss": 0.1247, "step": 3062 }, { "epoch": 0.56, "learning_rate": 1.9915330610891276e-05, "loss": 0.1151, "step": 3063 }, { "epoch": 0.56, "learning_rate": 1.99152341756917e-05, "loss": 0.111, "step": 3064 }, { "epoch": 0.56, "learning_rate": 1.991513768583912e-05, "loss": 0.1075, "step": 3065 }, { "epoch": 0.56, "learning_rate": 1.991504114133407e-05, "loss": 0.1125, "step": 3066 }, { "epoch": 0.56, "learning_rate": 1.991494454217708e-05, "loss": 0.1065, "step": 3067 }, { "epoch": 0.56, "learning_rate": 1.9914847888368683e-05, "loss": 0.1091, "step": 3068 }, { "epoch": 0.56, "learning_rate": 1.9914751179909416e-05, "loss": 0.1034, "step": 3069 }, { "epoch": 0.56, "learning_rate": 1.9914654416799802e-05, "loss": 0.1141, "step": 3070 }, { "epoch": 0.56, "learning_rate": 1.9914557599040386e-05, "loss": 0.1077, "step": 3071 }, { "epoch": 0.56, "learning_rate": 1.9914460726631694e-05, "loss": 0.1203, "step": 3072 }, { "epoch": 0.56, "learning_rate": 1.991436379957426e-05, "loss": 0.121, "step": 3073 }, { "epoch": 0.56, "learning_rate": 1.9914266817868624e-05, "loss": 0.1303, "step": 3074 }, { "epoch": 0.56, "learning_rate": 1.9914169781515316e-05, "loss": 0.1181, "step": 3075 }, { "epoch": 0.56, "learning_rate": 1.991407269051487e-05, "loss": 0.1166, "step": 3076 }, { "epoch": 0.56, "learning_rate": 1.9913975544867826e-05, "loss": 0.1099, "step": 3077 }, { "epoch": 0.56, "learning_rate": 1.9913878344574713e-05, "loss": 0.1152, "step": 3078 }, { "epoch": 0.56, "learning_rate": 1.9913781089636073e-05, "loss": 0.1164, "step": 3079 }, { "epoch": 0.56, "learning_rate": 1.9913683780052437e-05, "loss": 0.1106, "step": 3080 }, { "epoch": 0.57, "learning_rate": 1.9913586415824347e-05, "loss": 0.1009, "step": 3081 }, { "epoch": 0.57, "learning_rate": 1.9913488996952336e-05, "loss": 0.1156, "step": 3082 }, { "epoch": 0.57, "learning_rate": 1.991339152343694e-05, "loss": 0.1111, "step": 3083 }, { "epoch": 0.57, "learning_rate": 1.9913293995278694e-05, "loss": 0.1122, "step": 3084 }, { "epoch": 0.57, "learning_rate": 1.9913196412478144e-05, "loss": 0.1191, "step": 3085 }, { "epoch": 0.57, "learning_rate": 1.9913098775035823e-05, "loss": 0.1, "step": 3086 }, { "epoch": 0.57, "learning_rate": 1.991300108295227e-05, "loss": 0.1212, "step": 3087 }, { "epoch": 0.57, "learning_rate": 1.991290333622802e-05, "loss": 0.1432, "step": 3088 }, { "epoch": 0.57, "learning_rate": 1.9912805534863614e-05, "loss": 0.1156, "step": 3089 }, { "epoch": 0.57, "learning_rate": 1.9912707678859596e-05, "loss": 0.1118, "step": 3090 }, { "epoch": 0.57, "learning_rate": 1.99126097682165e-05, "loss": 0.1027, "step": 3091 }, { "epoch": 0.57, "learning_rate": 1.9912511802934863e-05, "loss": 0.1064, "step": 3092 }, { "epoch": 0.57, "learning_rate": 1.9912413783015232e-05, "loss": 0.1038, "step": 3093 }, { "epoch": 0.57, "learning_rate": 1.9912315708458144e-05, "loss": 0.1076, "step": 3094 }, { "epoch": 0.57, "learning_rate": 1.9912217579264137e-05, "loss": 0.1112, "step": 3095 }, { "epoch": 0.57, "learning_rate": 1.9912119395433754e-05, "loss": 0.1142, "step": 3096 }, { "epoch": 0.57, "learning_rate": 1.991202115696754e-05, "loss": 0.1075, "step": 3097 }, { "epoch": 0.57, "learning_rate": 1.9911922863866035e-05, "loss": 0.1165, "step": 3098 }, { "epoch": 0.57, "learning_rate": 1.9911824516129776e-05, "loss": 0.1101, "step": 3099 }, { "epoch": 0.57, "learning_rate": 1.9911726113759305e-05, "loss": 0.1102, "step": 3100 }, { "epoch": 0.57, "learning_rate": 1.991162765675517e-05, "loss": 0.1209, "step": 3101 }, { "epoch": 0.57, "learning_rate": 1.9911529145117913e-05, "loss": 0.1116, "step": 3102 }, { "epoch": 0.57, "learning_rate": 1.9911430578848075e-05, "loss": 0.1169, "step": 3103 }, { "epoch": 0.57, "learning_rate": 1.9911331957946198e-05, "loss": 0.114, "step": 3104 }, { "epoch": 0.57, "learning_rate": 1.9911233282412826e-05, "loss": 0.1097, "step": 3105 }, { "epoch": 0.57, "learning_rate": 1.9911134552248507e-05, "loss": 0.1084, "step": 3106 }, { "epoch": 0.57, "learning_rate": 1.991103576745378e-05, "loss": 0.1109, "step": 3107 }, { "epoch": 0.57, "learning_rate": 1.991093692802919e-05, "loss": 0.0992, "step": 3108 }, { "epoch": 0.57, "learning_rate": 1.9910838033975288e-05, "loss": 0.1245, "step": 3109 }, { "epoch": 0.57, "learning_rate": 1.991073908529261e-05, "loss": 0.1209, "step": 3110 }, { "epoch": 0.57, "learning_rate": 1.9910640081981708e-05, "loss": 0.1171, "step": 3111 }, { "epoch": 0.57, "learning_rate": 1.9910541024043126e-05, "loss": 0.1147, "step": 3112 }, { "epoch": 0.57, "learning_rate": 1.991044191147741e-05, "loss": 0.1067, "step": 3113 }, { "epoch": 0.57, "learning_rate": 1.9910342744285106e-05, "loss": 0.1093, "step": 3114 }, { "epoch": 0.57, "learning_rate": 1.991024352246676e-05, "loss": 0.1263, "step": 3115 }, { "epoch": 0.57, "learning_rate": 1.991014424602292e-05, "loss": 0.1039, "step": 3116 }, { "epoch": 0.57, "learning_rate": 1.9910044914954132e-05, "loss": 0.1274, "step": 3117 }, { "epoch": 0.57, "learning_rate": 1.9909945529260945e-05, "loss": 0.1048, "step": 3118 }, { "epoch": 0.57, "learning_rate": 1.9909846088943904e-05, "loss": 0.1096, "step": 3119 }, { "epoch": 0.57, "learning_rate": 1.990974659400356e-05, "loss": 0.1196, "step": 3120 }, { "epoch": 0.57, "learning_rate": 1.9909647044440463e-05, "loss": 0.1122, "step": 3121 }, { "epoch": 0.57, "learning_rate": 1.9909547440255158e-05, "loss": 0.1265, "step": 3122 }, { "epoch": 0.57, "learning_rate": 1.9909447781448194e-05, "loss": 0.1155, "step": 3123 }, { "epoch": 0.57, "learning_rate": 1.990934806802012e-05, "loss": 0.1172, "step": 3124 }, { "epoch": 0.57, "learning_rate": 1.9909248299971494e-05, "loss": 0.1108, "step": 3125 }, { "epoch": 0.57, "learning_rate": 1.9909148477302856e-05, "loss": 0.1122, "step": 3126 }, { "epoch": 0.57, "learning_rate": 1.9909048600014758e-05, "loss": 0.1154, "step": 3127 }, { "epoch": 0.57, "learning_rate": 1.9908948668107754e-05, "loss": 0.1273, "step": 3128 }, { "epoch": 0.57, "learning_rate": 1.990884868158239e-05, "loss": 0.1103, "step": 3129 }, { "epoch": 0.57, "learning_rate": 1.9908748640439222e-05, "loss": 0.1137, "step": 3130 }, { "epoch": 0.57, "learning_rate": 1.99086485446788e-05, "loss": 0.109, "step": 3131 }, { "epoch": 0.57, "learning_rate": 1.9908548394301673e-05, "loss": 0.1155, "step": 3132 }, { "epoch": 0.57, "learning_rate": 1.9908448189308396e-05, "loss": 0.125, "step": 3133 }, { "epoch": 0.57, "learning_rate": 1.990834792969952e-05, "loss": 0.1123, "step": 3134 }, { "epoch": 0.57, "learning_rate": 1.99082476154756e-05, "loss": 0.1214, "step": 3135 }, { "epoch": 0.58, "learning_rate": 1.9908147246637185e-05, "loss": 0.1185, "step": 3136 }, { "epoch": 0.58, "learning_rate": 1.990804682318483e-05, "loss": 0.1143, "step": 3137 }, { "epoch": 0.58, "learning_rate": 1.990794634511909e-05, "loss": 0.1061, "step": 3138 }, { "epoch": 0.58, "learning_rate": 1.9907845812440517e-05, "loss": 0.1159, "step": 3139 }, { "epoch": 0.58, "learning_rate": 1.9907745225149667e-05, "loss": 0.095, "step": 3140 }, { "epoch": 0.58, "learning_rate": 1.990764458324709e-05, "loss": 0.1207, "step": 3141 }, { "epoch": 0.58, "learning_rate": 1.990754388673335e-05, "loss": 0.102, "step": 3142 }, { "epoch": 0.58, "learning_rate": 1.9907443135608992e-05, "loss": 0.1085, "step": 3143 }, { "epoch": 0.58, "learning_rate": 1.9907342329874572e-05, "loss": 0.1123, "step": 3144 }, { "epoch": 0.58, "learning_rate": 1.9907241469530654e-05, "loss": 0.1166, "step": 3145 }, { "epoch": 0.58, "learning_rate": 1.9907140554577787e-05, "loss": 0.1192, "step": 3146 }, { "epoch": 0.58, "learning_rate": 1.990703958501653e-05, "loss": 0.1098, "step": 3147 }, { "epoch": 0.58, "learning_rate": 1.990693856084744e-05, "loss": 0.1009, "step": 3148 }, { "epoch": 0.58, "learning_rate": 1.990683748207107e-05, "loss": 0.1086, "step": 3149 }, { "epoch": 0.58, "learning_rate": 1.990673634868798e-05, "loss": 0.1171, "step": 3150 }, { "epoch": 0.58, "learning_rate": 1.9906635160698727e-05, "loss": 0.1069, "step": 3151 }, { "epoch": 0.58, "learning_rate": 1.990653391810387e-05, "loss": 0.119, "step": 3152 }, { "epoch": 0.58, "learning_rate": 1.9906432620903967e-05, "loss": 0.1159, "step": 3153 }, { "epoch": 0.58, "learning_rate": 1.9906331269099573e-05, "loss": 0.1069, "step": 3154 }, { "epoch": 0.58, "learning_rate": 1.9906229862691248e-05, "loss": 0.1126, "step": 3155 }, { "epoch": 0.58, "learning_rate": 1.9906128401679553e-05, "loss": 0.1071, "step": 3156 }, { "epoch": 0.58, "learning_rate": 1.990602688606505e-05, "loss": 0.114, "step": 3157 }, { "epoch": 0.58, "learning_rate": 1.990592531584829e-05, "loss": 0.114, "step": 3158 }, { "epoch": 0.58, "learning_rate": 1.990582369102984e-05, "loss": 0.1191, "step": 3159 }, { "epoch": 0.58, "learning_rate": 1.9905722011610255e-05, "loss": 0.1198, "step": 3160 }, { "epoch": 0.58, "learning_rate": 1.99056202775901e-05, "loss": 0.1167, "step": 3161 }, { "epoch": 0.58, "learning_rate": 1.990551848896994e-05, "loss": 0.1062, "step": 3162 }, { "epoch": 0.58, "learning_rate": 1.990541664575032e-05, "loss": 0.1114, "step": 3163 }, { "epoch": 0.58, "learning_rate": 1.9905314747931816e-05, "loss": 0.1098, "step": 3164 }, { "epoch": 0.58, "learning_rate": 1.9905212795514984e-05, "loss": 0.1053, "step": 3165 }, { "epoch": 0.58, "learning_rate": 1.9905110788500386e-05, "loss": 0.1174, "step": 3166 }, { "epoch": 0.58, "learning_rate": 1.990500872688859e-05, "loss": 0.1151, "step": 3167 }, { "epoch": 0.58, "learning_rate": 1.990490661068015e-05, "loss": 0.1003, "step": 3168 }, { "epoch": 0.58, "learning_rate": 1.9904804439875635e-05, "loss": 0.1269, "step": 3169 }, { "epoch": 0.58, "learning_rate": 1.99047022144756e-05, "loss": 0.1038, "step": 3170 }, { "epoch": 0.58, "learning_rate": 1.990459993448062e-05, "loss": 0.1141, "step": 3171 }, { "epoch": 0.58, "learning_rate": 1.9904497599891255e-05, "loss": 0.119, "step": 3172 }, { "epoch": 0.58, "learning_rate": 1.990439521070806e-05, "loss": 0.1197, "step": 3173 }, { "epoch": 0.58, "learning_rate": 1.9904292766931613e-05, "loss": 0.1153, "step": 3174 }, { "epoch": 0.58, "learning_rate": 1.990419026856247e-05, "loss": 0.1224, "step": 3175 }, { "epoch": 0.58, "learning_rate": 1.99040877156012e-05, "loss": 0.1178, "step": 3176 }, { "epoch": 0.58, "learning_rate": 1.9903985108048363e-05, "loss": 0.1197, "step": 3177 }, { "epoch": 0.58, "learning_rate": 1.990388244590453e-05, "loss": 0.1183, "step": 3178 }, { "epoch": 0.58, "learning_rate": 1.9903779729170264e-05, "loss": 0.1081, "step": 3179 }, { "epoch": 0.58, "learning_rate": 1.9903676957846132e-05, "loss": 0.1103, "step": 3180 }, { "epoch": 0.58, "learning_rate": 1.9903574131932702e-05, "loss": 0.1225, "step": 3181 }, { "epoch": 0.58, "learning_rate": 1.9903471251430537e-05, "loss": 0.1209, "step": 3182 }, { "epoch": 0.58, "learning_rate": 1.990336831634021e-05, "loss": 0.1244, "step": 3183 }, { "epoch": 0.58, "learning_rate": 1.9903265326662282e-05, "loss": 0.1074, "step": 3184 }, { "epoch": 0.58, "learning_rate": 1.9903162282397324e-05, "loss": 0.1077, "step": 3185 }, { "epoch": 0.58, "learning_rate": 1.9903059183545904e-05, "loss": 0.1083, "step": 3186 }, { "epoch": 0.58, "learning_rate": 1.9902956030108588e-05, "loss": 0.0955, "step": 3187 }, { "epoch": 0.58, "learning_rate": 1.990285282208595e-05, "loss": 0.1185, "step": 3188 }, { "epoch": 0.58, "learning_rate": 1.9902749559478552e-05, "loss": 0.1214, "step": 3189 }, { "epoch": 0.58, "learning_rate": 1.990264624228697e-05, "loss": 0.1131, "step": 3190 }, { "epoch": 0.59, "learning_rate": 1.9902542870511765e-05, "loss": 0.1088, "step": 3191 }, { "epoch": 0.59, "learning_rate": 1.9902439444153518e-05, "loss": 0.1087, "step": 3192 }, { "epoch": 0.59, "learning_rate": 1.9902335963212786e-05, "loss": 0.1222, "step": 3193 }, { "epoch": 0.59, "learning_rate": 1.9902232427690152e-05, "loss": 0.106, "step": 3194 }, { "epoch": 0.59, "learning_rate": 1.9902128837586175e-05, "loss": 0.1163, "step": 3195 }, { "epoch": 0.59, "learning_rate": 1.9902025192901435e-05, "loss": 0.1158, "step": 3196 }, { "epoch": 0.59, "learning_rate": 1.99019214936365e-05, "loss": 0.121, "step": 3197 }, { "epoch": 0.59, "learning_rate": 1.9901817739791945e-05, "loss": 0.1274, "step": 3198 }, { "epoch": 0.59, "learning_rate": 1.9901713931368333e-05, "loss": 0.1191, "step": 3199 }, { "epoch": 0.59, "learning_rate": 1.9901610068366245e-05, "loss": 0.1207, "step": 3200 }, { "epoch": 0.59, "learning_rate": 1.990150615078625e-05, "loss": 0.1127, "step": 3201 }, { "epoch": 0.59, "learning_rate": 1.9901402178628923e-05, "loss": 0.115, "step": 3202 }, { "epoch": 0.59, "learning_rate": 1.9901298151894834e-05, "loss": 0.1081, "step": 3203 }, { "epoch": 0.59, "learning_rate": 1.9901194070584555e-05, "loss": 0.1132, "step": 3204 }, { "epoch": 0.59, "learning_rate": 1.9901089934698663e-05, "loss": 0.1149, "step": 3205 }, { "epoch": 0.59, "learning_rate": 1.9900985744237734e-05, "loss": 0.1149, "step": 3206 }, { "epoch": 0.59, "learning_rate": 1.9900881499202336e-05, "loss": 0.1162, "step": 3207 }, { "epoch": 0.59, "learning_rate": 1.990077719959305e-05, "loss": 0.1096, "step": 3208 }, { "epoch": 0.59, "learning_rate": 1.9900672845410446e-05, "loss": 0.1211, "step": 3209 }, { "epoch": 0.59, "learning_rate": 1.9900568436655103e-05, "loss": 0.1195, "step": 3210 }, { "epoch": 0.59, "learning_rate": 1.9900463973327595e-05, "loss": 0.1281, "step": 3211 }, { "epoch": 0.59, "learning_rate": 1.99003594554285e-05, "loss": 0.1191, "step": 3212 }, { "epoch": 0.59, "learning_rate": 1.9900254882958387e-05, "loss": 0.1155, "step": 3213 }, { "epoch": 0.59, "learning_rate": 1.9900150255917838e-05, "loss": 0.1184, "step": 3214 }, { "epoch": 0.59, "learning_rate": 1.990004557430743e-05, "loss": 0.1208, "step": 3215 }, { "epoch": 0.59, "learning_rate": 1.989994083812774e-05, "loss": 0.1155, "step": 3216 }, { "epoch": 0.59, "learning_rate": 1.989983604737934e-05, "loss": 0.1136, "step": 3217 }, { "epoch": 0.59, "learning_rate": 1.9899731202062817e-05, "loss": 0.1175, "step": 3218 }, { "epoch": 0.59, "learning_rate": 1.989962630217874e-05, "loss": 0.115, "step": 3219 }, { "epoch": 0.59, "learning_rate": 1.989952134772769e-05, "loss": 0.1235, "step": 3220 }, { "epoch": 0.59, "learning_rate": 1.9899416338710253e-05, "loss": 0.1188, "step": 3221 }, { "epoch": 0.59, "learning_rate": 1.9899311275126995e-05, "loss": 0.1087, "step": 3222 }, { "epoch": 0.59, "learning_rate": 1.9899206156978503e-05, "loss": 0.1196, "step": 3223 }, { "epoch": 0.59, "learning_rate": 1.9899100984265357e-05, "loss": 0.1176, "step": 3224 }, { "epoch": 0.59, "learning_rate": 1.989899575698813e-05, "loss": 0.111, "step": 3225 }, { "epoch": 0.59, "learning_rate": 1.989889047514741e-05, "loss": 0.1129, "step": 3226 }, { "epoch": 0.59, "learning_rate": 1.9898785138743772e-05, "loss": 0.1193, "step": 3227 }, { "epoch": 0.59, "learning_rate": 1.98986797477778e-05, "loss": 0.1202, "step": 3228 }, { "epoch": 0.59, "learning_rate": 1.9898574302250072e-05, "loss": 0.1217, "step": 3229 }, { "epoch": 0.59, "learning_rate": 1.9898468802161172e-05, "loss": 0.1102, "step": 3230 }, { "epoch": 0.59, "learning_rate": 1.989836324751168e-05, "loss": 0.1251, "step": 3231 }, { "epoch": 0.59, "learning_rate": 1.9898257638302177e-05, "loss": 0.1145, "step": 3232 }, { "epoch": 0.59, "learning_rate": 1.9898151974533244e-05, "loss": 0.116, "step": 3233 }, { "epoch": 0.59, "learning_rate": 1.9898046256205468e-05, "loss": 0.1305, "step": 3234 }, { "epoch": 0.59, "learning_rate": 1.989794048331943e-05, "loss": 0.1173, "step": 3235 }, { "epoch": 0.59, "learning_rate": 1.989783465587571e-05, "loss": 0.1214, "step": 3236 }, { "epoch": 0.59, "learning_rate": 1.9897728773874892e-05, "loss": 0.1035, "step": 3237 }, { "epoch": 0.59, "learning_rate": 1.9897622837317565e-05, "loss": 0.0958, "step": 3238 }, { "epoch": 0.59, "learning_rate": 1.989751684620431e-05, "loss": 0.1129, "step": 3239 }, { "epoch": 0.59, "learning_rate": 1.9897410800535705e-05, "loss": 0.1158, "step": 3240 }, { "epoch": 0.59, "learning_rate": 1.9897304700312344e-05, "loss": 0.1178, "step": 3241 }, { "epoch": 0.59, "learning_rate": 1.9897198545534806e-05, "loss": 0.1132, "step": 3242 }, { "epoch": 0.59, "learning_rate": 1.9897092336203675e-05, "loss": 0.1109, "step": 3243 }, { "epoch": 0.59, "learning_rate": 1.9896986072319545e-05, "loss": 0.1133, "step": 3244 }, { "epoch": 0.6, "learning_rate": 1.989687975388299e-05, "loss": 0.1077, "step": 3245 }, { "epoch": 0.6, "learning_rate": 1.9896773380894604e-05, "loss": 0.1056, "step": 3246 }, { "epoch": 0.6, "learning_rate": 1.9896666953354975e-05, "loss": 0.1209, "step": 3247 }, { "epoch": 0.6, "learning_rate": 1.9896560471264682e-05, "loss": 0.1209, "step": 3248 }, { "epoch": 0.6, "learning_rate": 1.9896453934624316e-05, "loss": 0.1254, "step": 3249 }, { "epoch": 0.6, "learning_rate": 1.9896347343434465e-05, "loss": 0.1048, "step": 3250 }, { "epoch": 0.6, "learning_rate": 1.9896240697695716e-05, "loss": 0.1079, "step": 3251 }, { "epoch": 0.6, "learning_rate": 1.989613399740866e-05, "loss": 0.0968, "step": 3252 }, { "epoch": 0.6, "learning_rate": 1.9896027242573876e-05, "loss": 0.1105, "step": 3253 }, { "epoch": 0.6, "learning_rate": 1.9895920433191957e-05, "loss": 0.1018, "step": 3254 }, { "epoch": 0.6, "learning_rate": 1.9895813569263497e-05, "loss": 0.1189, "step": 3255 }, { "epoch": 0.6, "learning_rate": 1.9895706650789077e-05, "loss": 0.12, "step": 3256 }, { "epoch": 0.6, "learning_rate": 1.9895599677769293e-05, "loss": 0.1224, "step": 3257 }, { "epoch": 0.6, "learning_rate": 1.989549265020473e-05, "loss": 0.1182, "step": 3258 }, { "epoch": 0.6, "learning_rate": 1.989538556809598e-05, "loss": 0.1188, "step": 3259 }, { "epoch": 0.6, "learning_rate": 1.9895278431443634e-05, "loss": 0.1182, "step": 3260 }, { "epoch": 0.6, "learning_rate": 1.989517124024828e-05, "loss": 0.1182, "step": 3261 }, { "epoch": 0.6, "learning_rate": 1.9895063994510512e-05, "loss": 0.1045, "step": 3262 }, { "epoch": 0.6, "learning_rate": 1.989495669423092e-05, "loss": 0.1079, "step": 3263 }, { "epoch": 0.6, "learning_rate": 1.9894849339410092e-05, "loss": 0.1232, "step": 3264 }, { "epoch": 0.6, "learning_rate": 1.9894741930048624e-05, "loss": 0.1176, "step": 3265 }, { "epoch": 0.6, "learning_rate": 1.989463446614711e-05, "loss": 0.1176, "step": 3266 }, { "epoch": 0.6, "learning_rate": 1.9894526947706135e-05, "loss": 0.1119, "step": 3267 }, { "epoch": 0.6, "learning_rate": 1.9894419374726298e-05, "loss": 0.1104, "step": 3268 }, { "epoch": 0.6, "learning_rate": 1.9894311747208187e-05, "loss": 0.1121, "step": 3269 }, { "epoch": 0.6, "learning_rate": 1.98942040651524e-05, "loss": 0.1121, "step": 3270 }, { "epoch": 0.6, "learning_rate": 1.9894096328559525e-05, "loss": 0.1164, "step": 3271 }, { "epoch": 0.6, "learning_rate": 1.989398853743016e-05, "loss": 0.1341, "step": 3272 }, { "epoch": 0.6, "learning_rate": 1.98938806917649e-05, "loss": 0.1172, "step": 3273 }, { "epoch": 0.6, "learning_rate": 1.989377279156434e-05, "loss": 0.1134, "step": 3274 }, { "epoch": 0.6, "learning_rate": 1.989366483682907e-05, "loss": 0.1083, "step": 3275 }, { "epoch": 0.6, "learning_rate": 1.989355682755969e-05, "loss": 0.1055, "step": 3276 }, { "epoch": 0.6, "learning_rate": 1.9893448763756787e-05, "loss": 0.1142, "step": 3277 }, { "epoch": 0.6, "learning_rate": 1.989334064542097e-05, "loss": 0.117, "step": 3278 }, { "epoch": 0.6, "learning_rate": 1.9893232472552822e-05, "loss": 0.1177, "step": 3279 }, { "epoch": 0.6, "learning_rate": 1.9893124245152948e-05, "loss": 0.1338, "step": 3280 }, { "epoch": 0.6, "learning_rate": 1.9893015963221942e-05, "loss": 0.1211, "step": 3281 }, { "epoch": 0.6, "learning_rate": 1.98929076267604e-05, "loss": 0.1126, "step": 3282 }, { "epoch": 0.6, "learning_rate": 1.9892799235768915e-05, "loss": 0.1202, "step": 3283 }, { "epoch": 0.6, "learning_rate": 1.989269079024809e-05, "loss": 0.1066, "step": 3284 }, { "epoch": 0.6, "learning_rate": 1.9892582290198527e-05, "loss": 0.1174, "step": 3285 }, { "epoch": 0.6, "learning_rate": 1.9892473735620816e-05, "loss": 0.1221, "step": 3286 }, { "epoch": 0.6, "learning_rate": 1.9892365126515555e-05, "loss": 0.1112, "step": 3287 }, { "epoch": 0.6, "learning_rate": 1.989225646288335e-05, "loss": 0.1196, "step": 3288 }, { "epoch": 0.6, "learning_rate": 1.9892147744724792e-05, "loss": 0.1117, "step": 3289 }, { "epoch": 0.6, "learning_rate": 1.9892038972040485e-05, "loss": 0.1041, "step": 3290 }, { "epoch": 0.6, "learning_rate": 1.989193014483103e-05, "loss": 0.1139, "step": 3291 }, { "epoch": 0.6, "learning_rate": 1.989182126309702e-05, "loss": 0.1204, "step": 3292 }, { "epoch": 0.6, "learning_rate": 1.9891712326839065e-05, "loss": 0.1158, "step": 3293 }, { "epoch": 0.6, "learning_rate": 1.9891603336057753e-05, "loss": 0.1082, "step": 3294 }, { "epoch": 0.6, "learning_rate": 1.9891494290753696e-05, "loss": 0.1083, "step": 3295 }, { "epoch": 0.6, "learning_rate": 1.989138519092749e-05, "loss": 0.1024, "step": 3296 }, { "epoch": 0.6, "learning_rate": 1.989127603657974e-05, "loss": 0.1173, "step": 3297 }, { "epoch": 0.6, "learning_rate": 1.989116682771104e-05, "loss": 0.1124, "step": 3298 }, { "epoch": 0.6, "learning_rate": 1.9891057564322e-05, "loss": 0.115, "step": 3299 }, { "epoch": 0.61, "learning_rate": 1.9890948246413215e-05, "loss": 0.1067, "step": 3300 }, { "epoch": 0.61, "learning_rate": 1.9890838873985295e-05, "loss": 0.1176, "step": 3301 }, { "epoch": 0.61, "learning_rate": 1.9890729447038838e-05, "loss": 0.1041, "step": 3302 }, { "epoch": 0.61, "learning_rate": 1.989061996557445e-05, "loss": 0.1149, "step": 3303 }, { "epoch": 0.61, "learning_rate": 1.989051042959273e-05, "loss": 0.0959, "step": 3304 }, { "epoch": 0.61, "learning_rate": 1.989040083909429e-05, "loss": 0.1194, "step": 3305 }, { "epoch": 0.61, "learning_rate": 1.9890291194079724e-05, "loss": 0.1098, "step": 3306 }, { "epoch": 0.61, "learning_rate": 1.9890181494549642e-05, "loss": 0.1263, "step": 3307 }, { "epoch": 0.61, "learning_rate": 1.9890071740504648e-05, "loss": 0.11, "step": 3308 }, { "epoch": 0.61, "learning_rate": 1.9889961931945348e-05, "loss": 0.1208, "step": 3309 }, { "epoch": 0.61, "learning_rate": 1.9889852068872347e-05, "loss": 0.1088, "step": 3310 }, { "epoch": 0.61, "learning_rate": 1.988974215128625e-05, "loss": 0.1193, "step": 3311 }, { "epoch": 0.61, "learning_rate": 1.988963217918766e-05, "loss": 0.1066, "step": 3312 }, { "epoch": 0.61, "learning_rate": 1.9889522152577187e-05, "loss": 0.1114, "step": 3313 }, { "epoch": 0.61, "learning_rate": 1.9889412071455438e-05, "loss": 0.1008, "step": 3314 }, { "epoch": 0.61, "learning_rate": 1.9889301935823012e-05, "loss": 0.1145, "step": 3315 }, { "epoch": 0.61, "learning_rate": 1.9889191745680527e-05, "loss": 0.1184, "step": 3316 }, { "epoch": 0.61, "learning_rate": 1.9889081501028588e-05, "loss": 0.1123, "step": 3317 }, { "epoch": 0.61, "learning_rate": 1.9888971201867795e-05, "loss": 0.1286, "step": 3318 }, { "epoch": 0.61, "learning_rate": 1.9888860848198762e-05, "loss": 0.1073, "step": 3319 }, { "epoch": 0.61, "learning_rate": 1.9888750440022098e-05, "loss": 0.0968, "step": 3320 }, { "epoch": 0.61, "learning_rate": 1.988863997733841e-05, "loss": 0.1138, "step": 3321 }, { "epoch": 0.61, "learning_rate": 1.9888529460148304e-05, "loss": 0.1073, "step": 3322 }, { "epoch": 0.61, "learning_rate": 1.9888418888452394e-05, "loss": 0.1212, "step": 3323 }, { "epoch": 0.61, "learning_rate": 1.9888308262251286e-05, "loss": 0.1083, "step": 3324 }, { "epoch": 0.61, "learning_rate": 1.9888197581545593e-05, "loss": 0.118, "step": 3325 }, { "epoch": 0.61, "learning_rate": 1.9888086846335922e-05, "loss": 0.1097, "step": 3326 }, { "epoch": 0.61, "learning_rate": 1.9887976056622886e-05, "loss": 0.1075, "step": 3327 }, { "epoch": 0.61, "learning_rate": 1.9887865212407095e-05, "loss": 0.1224, "step": 3328 }, { "epoch": 0.61, "learning_rate": 1.988775431368916e-05, "loss": 0.11, "step": 3329 }, { "epoch": 0.61, "learning_rate": 1.9887643360469687e-05, "loss": 0.1191, "step": 3330 }, { "epoch": 0.61, "learning_rate": 1.9887532352749293e-05, "loss": 0.1191, "step": 3331 }, { "epoch": 0.61, "learning_rate": 1.988742129052859e-05, "loss": 0.11, "step": 3332 }, { "epoch": 0.61, "learning_rate": 1.9887310173808192e-05, "loss": 0.1061, "step": 3333 }, { "epoch": 0.61, "learning_rate": 1.9887199002588704e-05, "loss": 0.1171, "step": 3334 }, { "epoch": 0.61, "learning_rate": 1.9887087776870748e-05, "loss": 0.1111, "step": 3335 }, { "epoch": 0.61, "learning_rate": 1.988697649665493e-05, "loss": 0.1094, "step": 3336 }, { "epoch": 0.61, "learning_rate": 1.9886865161941866e-05, "loss": 0.1136, "step": 3337 }, { "epoch": 0.61, "learning_rate": 1.988675377273217e-05, "loss": 0.117, "step": 3338 }, { "epoch": 0.61, "learning_rate": 1.9886642329026457e-05, "loss": 0.1132, "step": 3339 }, { "epoch": 0.61, "learning_rate": 1.9886530830825337e-05, "loss": 0.101, "step": 3340 }, { "epoch": 0.61, "learning_rate": 1.988641927812943e-05, "loss": 0.1037, "step": 3341 }, { "epoch": 0.61, "learning_rate": 1.988630767093935e-05, "loss": 0.1115, "step": 3342 }, { "epoch": 0.61, "learning_rate": 1.9886196009255704e-05, "loss": 0.1179, "step": 3343 }, { "epoch": 0.61, "learning_rate": 1.9886084293079118e-05, "loss": 0.1028, "step": 3344 }, { "epoch": 0.61, "learning_rate": 1.9885972522410205e-05, "loss": 0.114, "step": 3345 }, { "epoch": 0.61, "learning_rate": 1.9885860697249576e-05, "loss": 0.1137, "step": 3346 }, { "epoch": 0.61, "learning_rate": 1.9885748817597857e-05, "loss": 0.103, "step": 3347 }, { "epoch": 0.61, "learning_rate": 1.9885636883455652e-05, "loss": 0.1215, "step": 3348 }, { "epoch": 0.61, "learning_rate": 1.988552489482359e-05, "loss": 0.1188, "step": 3349 }, { "epoch": 0.61, "learning_rate": 1.988541285170228e-05, "loss": 0.1207, "step": 3350 }, { "epoch": 0.61, "learning_rate": 1.9885300754092346e-05, "loss": 0.1111, "step": 3351 }, { "epoch": 0.61, "learning_rate": 1.9885188601994404e-05, "loss": 0.1156, "step": 3352 }, { "epoch": 0.61, "learning_rate": 1.9885076395409066e-05, "loss": 0.0971, "step": 3353 }, { "epoch": 0.62, "learning_rate": 1.9884964134336957e-05, "loss": 0.1247, "step": 3354 }, { "epoch": 0.62, "learning_rate": 1.9884851818778695e-05, "loss": 0.1048, "step": 3355 }, { "epoch": 0.62, "learning_rate": 1.9884739448734897e-05, "loss": 0.1192, "step": 3356 }, { "epoch": 0.62, "learning_rate": 1.9884627024206186e-05, "loss": 0.1054, "step": 3357 }, { "epoch": 0.62, "learning_rate": 1.9884514545193174e-05, "loss": 0.1188, "step": 3358 }, { "epoch": 0.62, "learning_rate": 1.988440201169649e-05, "loss": 0.1071, "step": 3359 }, { "epoch": 0.62, "learning_rate": 1.9884289423716748e-05, "loss": 0.116, "step": 3360 }, { "epoch": 0.62, "learning_rate": 1.9884176781254577e-05, "loss": 0.105, "step": 3361 }, { "epoch": 0.62, "learning_rate": 1.988406408431059e-05, "loss": 0.1182, "step": 3362 }, { "epoch": 0.62, "learning_rate": 1.9883951332885406e-05, "loss": 0.1169, "step": 3363 }, { "epoch": 0.62, "learning_rate": 1.9883838526979652e-05, "loss": 0.1129, "step": 3364 }, { "epoch": 0.62, "learning_rate": 1.9883725666593953e-05, "loss": 0.1116, "step": 3365 }, { "epoch": 0.62, "learning_rate": 1.988361275172892e-05, "loss": 0.1053, "step": 3366 }, { "epoch": 0.62, "learning_rate": 1.9883499782385188e-05, "loss": 0.1192, "step": 3367 }, { "epoch": 0.62, "learning_rate": 1.988338675856337e-05, "loss": 0.1041, "step": 3368 }, { "epoch": 0.62, "learning_rate": 1.9883273680264092e-05, "loss": 0.1055, "step": 3369 }, { "epoch": 0.62, "learning_rate": 1.9883160547487982e-05, "loss": 0.122, "step": 3370 }, { "epoch": 0.62, "learning_rate": 1.9883047360235656e-05, "loss": 0.1199, "step": 3371 }, { "epoch": 0.62, "learning_rate": 1.988293411850774e-05, "loss": 0.1036, "step": 3372 }, { "epoch": 0.62, "learning_rate": 1.9882820822304864e-05, "loss": 0.1217, "step": 3373 }, { "epoch": 0.62, "learning_rate": 1.9882707471627648e-05, "loss": 0.1147, "step": 3374 }, { "epoch": 0.62, "learning_rate": 1.9882594066476716e-05, "loss": 0.1033, "step": 3375 }, { "epoch": 0.62, "learning_rate": 1.988248060685269e-05, "loss": 0.1118, "step": 3376 }, { "epoch": 0.62, "learning_rate": 1.9882367092756204e-05, "loss": 0.1238, "step": 3377 }, { "epoch": 0.62, "learning_rate": 1.9882253524187877e-05, "loss": 0.1133, "step": 3378 }, { "epoch": 0.62, "learning_rate": 1.9882139901148338e-05, "loss": 0.1173, "step": 3379 }, { "epoch": 0.62, "learning_rate": 1.9882026223638207e-05, "loss": 0.1088, "step": 3380 }, { "epoch": 0.62, "learning_rate": 1.9881912491658122e-05, "loss": 0.1185, "step": 3381 }, { "epoch": 0.62, "learning_rate": 1.9881798705208703e-05, "loss": 0.1013, "step": 3382 }, { "epoch": 0.62, "learning_rate": 1.988168486429058e-05, "loss": 0.1145, "step": 3383 }, { "epoch": 0.62, "learning_rate": 1.9881570968904377e-05, "loss": 0.1093, "step": 3384 }, { "epoch": 0.62, "learning_rate": 1.988145701905072e-05, "loss": 0.112, "step": 3385 }, { "epoch": 0.62, "learning_rate": 1.9881343014730245e-05, "loss": 0.1139, "step": 3386 }, { "epoch": 0.62, "learning_rate": 1.988122895594357e-05, "loss": 0.1098, "step": 3387 }, { "epoch": 0.62, "learning_rate": 1.9881114842691334e-05, "loss": 0.0988, "step": 3388 }, { "epoch": 0.62, "learning_rate": 1.9881000674974164e-05, "loss": 0.1015, "step": 3389 }, { "epoch": 0.62, "learning_rate": 1.9880886452792683e-05, "loss": 0.109, "step": 3390 }, { "epoch": 0.62, "learning_rate": 1.9880772176147526e-05, "loss": 0.1151, "step": 3391 }, { "epoch": 0.62, "learning_rate": 1.988065784503932e-05, "loss": 0.1105, "step": 3392 }, { "epoch": 0.62, "learning_rate": 1.9880543459468694e-05, "loss": 0.1096, "step": 3393 }, { "epoch": 0.62, "learning_rate": 1.9880429019436285e-05, "loss": 0.1045, "step": 3394 }, { "epoch": 0.62, "learning_rate": 1.9880314524942718e-05, "loss": 0.1141, "step": 3395 }, { "epoch": 0.62, "learning_rate": 1.988019997598863e-05, "loss": 0.123, "step": 3396 }, { "epoch": 0.62, "learning_rate": 1.9880085372574644e-05, "loss": 0.1109, "step": 3397 }, { "epoch": 0.62, "learning_rate": 1.9879970714701394e-05, "loss": 0.1296, "step": 3398 }, { "epoch": 0.62, "learning_rate": 1.9879856002369516e-05, "loss": 0.1152, "step": 3399 }, { "epoch": 0.62, "learning_rate": 1.987974123557964e-05, "loss": 0.1211, "step": 3400 }, { "epoch": 0.62, "learning_rate": 1.98796264143324e-05, "loss": 0.1268, "step": 3401 }, { "epoch": 0.62, "learning_rate": 1.987951153862843e-05, "loss": 0.1064, "step": 3402 }, { "epoch": 0.62, "learning_rate": 1.9879396608468354e-05, "loss": 0.1067, "step": 3403 }, { "epoch": 0.62, "learning_rate": 1.987928162385282e-05, "loss": 0.1261, "step": 3404 }, { "epoch": 0.62, "learning_rate": 1.987916658478245e-05, "loss": 0.1193, "step": 3405 }, { "epoch": 0.62, "learning_rate": 1.9879051491257883e-05, "loss": 0.1159, "step": 3406 }, { "epoch": 0.62, "learning_rate": 1.9878936343279752e-05, "loss": 0.1116, "step": 3407 }, { "epoch": 0.62, "learning_rate": 1.9878821140848695e-05, "loss": 0.1011, "step": 3408 }, { "epoch": 0.63, "learning_rate": 1.9878705883965342e-05, "loss": 0.0983, "step": 3409 }, { "epoch": 0.63, "learning_rate": 1.987859057263033e-05, "loss": 0.1061, "step": 3410 }, { "epoch": 0.63, "learning_rate": 1.9878475206844298e-05, "loss": 0.1123, "step": 3411 }, { "epoch": 0.63, "learning_rate": 1.987835978660788e-05, "loss": 0.1164, "step": 3412 }, { "epoch": 0.63, "learning_rate": 1.9878244311921708e-05, "loss": 0.1101, "step": 3413 }, { "epoch": 0.63, "learning_rate": 1.9878128782786422e-05, "loss": 0.1062, "step": 3414 }, { "epoch": 0.63, "learning_rate": 1.987801319920266e-05, "loss": 0.1158, "step": 3415 }, { "epoch": 0.63, "learning_rate": 1.987789756117106e-05, "loss": 0.1078, "step": 3416 }, { "epoch": 0.63, "learning_rate": 1.9877781868692257e-05, "loss": 0.109, "step": 3417 }, { "epoch": 0.63, "learning_rate": 1.9877666121766886e-05, "loss": 0.1112, "step": 3418 }, { "epoch": 0.63, "learning_rate": 1.987755032039559e-05, "loss": 0.118, "step": 3419 }, { "epoch": 0.63, "learning_rate": 1.9877434464579005e-05, "loss": 0.1137, "step": 3420 }, { "epoch": 0.63, "learning_rate": 1.9877318554317767e-05, "loss": 0.1151, "step": 3421 }, { "epoch": 0.63, "learning_rate": 1.9877202589612522e-05, "loss": 0.1036, "step": 3422 }, { "epoch": 0.63, "learning_rate": 1.9877086570463905e-05, "loss": 0.1146, "step": 3423 }, { "epoch": 0.63, "learning_rate": 1.9876970496872553e-05, "loss": 0.1137, "step": 3424 }, { "epoch": 0.63, "learning_rate": 1.9876854368839108e-05, "loss": 0.1263, "step": 3425 }, { "epoch": 0.63, "learning_rate": 1.987673818636421e-05, "loss": 0.1139, "step": 3426 }, { "epoch": 0.63, "learning_rate": 1.98766219494485e-05, "loss": 0.1105, "step": 3427 }, { "epoch": 0.63, "learning_rate": 1.9876505658092618e-05, "loss": 0.1082, "step": 3428 }, { "epoch": 0.63, "learning_rate": 1.9876389312297207e-05, "loss": 0.1149, "step": 3429 }, { "epoch": 0.63, "learning_rate": 1.9876272912062907e-05, "loss": 0.0953, "step": 3430 }, { "epoch": 0.63, "learning_rate": 1.987615645739036e-05, "loss": 0.1105, "step": 3431 }, { "epoch": 0.63, "learning_rate": 1.9876039948280206e-05, "loss": 0.113, "step": 3432 }, { "epoch": 0.63, "learning_rate": 1.9875923384733088e-05, "loss": 0.1142, "step": 3433 }, { "epoch": 0.63, "learning_rate": 1.9875806766749647e-05, "loss": 0.1124, "step": 3434 }, { "epoch": 0.63, "learning_rate": 1.9875690094330532e-05, "loss": 0.1105, "step": 3435 }, { "epoch": 0.63, "learning_rate": 1.9875573367476377e-05, "loss": 0.1221, "step": 3436 }, { "epoch": 0.63, "learning_rate": 1.9875456586187836e-05, "loss": 0.0994, "step": 3437 }, { "epoch": 0.63, "learning_rate": 1.9875339750465543e-05, "loss": 0.1078, "step": 3438 }, { "epoch": 0.63, "learning_rate": 1.9875222860310147e-05, "loss": 0.1149, "step": 3439 }, { "epoch": 0.63, "learning_rate": 1.987510591572229e-05, "loss": 0.1185, "step": 3440 }, { "epoch": 0.63, "learning_rate": 1.987498891670262e-05, "loss": 0.1262, "step": 3441 }, { "epoch": 0.63, "learning_rate": 1.9874871863251777e-05, "loss": 0.1096, "step": 3442 }, { "epoch": 0.63, "learning_rate": 1.9874754755370412e-05, "loss": 0.1148, "step": 3443 }, { "epoch": 0.63, "learning_rate": 1.9874637593059167e-05, "loss": 0.1006, "step": 3444 }, { "epoch": 0.63, "learning_rate": 1.9874520376318685e-05, "loss": 0.1171, "step": 3445 }, { "epoch": 0.63, "learning_rate": 1.987440310514962e-05, "loss": 0.1129, "step": 3446 }, { "epoch": 0.63, "learning_rate": 1.987428577955261e-05, "loss": 0.1087, "step": 3447 }, { "epoch": 0.63, "learning_rate": 1.9874168399528307e-05, "loss": 0.1152, "step": 3448 }, { "epoch": 0.63, "learning_rate": 1.9874050965077356e-05, "loss": 0.106, "step": 3449 }, { "epoch": 0.63, "learning_rate": 1.9873933476200405e-05, "loss": 0.1193, "step": 3450 }, { "epoch": 0.63, "learning_rate": 1.98738159328981e-05, "loss": 0.117, "step": 3451 }, { "epoch": 0.63, "learning_rate": 1.9873698335171093e-05, "loss": 0.1171, "step": 3452 }, { "epoch": 0.63, "learning_rate": 1.9873580683020027e-05, "loss": 0.1281, "step": 3453 }, { "epoch": 0.63, "learning_rate": 1.9873462976445553e-05, "loss": 0.1106, "step": 3454 }, { "epoch": 0.63, "learning_rate": 1.9873345215448324e-05, "loss": 0.1363, "step": 3455 }, { "epoch": 0.63, "learning_rate": 1.987322740002898e-05, "loss": 0.1242, "step": 3456 }, { "epoch": 0.63, "learning_rate": 1.9873109530188177e-05, "loss": 0.1051, "step": 3457 }, { "epoch": 0.63, "learning_rate": 1.9872991605926563e-05, "loss": 0.1176, "step": 3458 }, { "epoch": 0.63, "learning_rate": 1.9872873627244788e-05, "loss": 0.1106, "step": 3459 }, { "epoch": 0.63, "learning_rate": 1.9872755594143506e-05, "loss": 0.0992, "step": 3460 }, { "epoch": 0.63, "learning_rate": 1.9872637506623357e-05, "loss": 0.1027, "step": 3461 }, { "epoch": 0.63, "learning_rate": 1.9872519364685003e-05, "loss": 0.1121, "step": 3462 }, { "epoch": 0.64, "learning_rate": 1.9872401168329088e-05, "loss": 0.114, "step": 3463 }, { "epoch": 0.64, "learning_rate": 1.987228291755627e-05, "loss": 0.1088, "step": 3464 }, { "epoch": 0.64, "learning_rate": 1.9872164612367196e-05, "loss": 0.1156, "step": 3465 }, { "epoch": 0.64, "learning_rate": 1.9872046252762517e-05, "loss": 0.0961, "step": 3466 }, { "epoch": 0.64, "learning_rate": 1.987192783874289e-05, "loss": 0.1086, "step": 3467 }, { "epoch": 0.64, "learning_rate": 1.9871809370308967e-05, "loss": 0.1157, "step": 3468 }, { "epoch": 0.64, "learning_rate": 1.9871690847461398e-05, "loss": 0.1042, "step": 3469 }, { "epoch": 0.64, "learning_rate": 1.9871572270200835e-05, "loss": 0.1002, "step": 3470 }, { "epoch": 0.64, "learning_rate": 1.9871453638527936e-05, "loss": 0.1028, "step": 3471 }, { "epoch": 0.64, "learning_rate": 1.9871334952443355e-05, "loss": 0.1172, "step": 3472 }, { "epoch": 0.64, "learning_rate": 1.9871216211947746e-05, "loss": 0.1246, "step": 3473 }, { "epoch": 0.64, "learning_rate": 1.987109741704176e-05, "loss": 0.1074, "step": 3474 }, { "epoch": 0.64, "learning_rate": 1.9870978567726052e-05, "loss": 0.1106, "step": 3475 }, { "epoch": 0.64, "learning_rate": 1.987085966400128e-05, "loss": 0.112, "step": 3476 }, { "epoch": 0.64, "learning_rate": 1.98707407058681e-05, "loss": 0.1055, "step": 3477 }, { "epoch": 0.64, "learning_rate": 1.9870621693327164e-05, "loss": 0.1066, "step": 3478 }, { "epoch": 0.64, "learning_rate": 1.9870502626379127e-05, "loss": 0.1198, "step": 3479 }, { "epoch": 0.64, "learning_rate": 1.9870383505024654e-05, "loss": 0.106, "step": 3480 }, { "epoch": 0.64, "learning_rate": 1.9870264329264394e-05, "loss": 0.12, "step": 3481 }, { "epoch": 0.64, "learning_rate": 1.9870145099099004e-05, "loss": 0.1049, "step": 3482 }, { "epoch": 0.64, "learning_rate": 1.987002581452915e-05, "loss": 0.1073, "step": 3483 }, { "epoch": 0.64, "learning_rate": 1.9869906475555475e-05, "loss": 0.1016, "step": 3484 }, { "epoch": 0.64, "learning_rate": 1.9869787082178647e-05, "loss": 0.1261, "step": 3485 }, { "epoch": 0.64, "learning_rate": 1.986966763439932e-05, "loss": 0.1149, "step": 3486 }, { "epoch": 0.64, "learning_rate": 1.986954813221815e-05, "loss": 0.1166, "step": 3487 }, { "epoch": 0.64, "learning_rate": 1.9869428575635808e-05, "loss": 0.1276, "step": 3488 }, { "epoch": 0.64, "learning_rate": 1.9869308964652937e-05, "loss": 0.1196, "step": 3489 }, { "epoch": 0.64, "learning_rate": 1.986918929927021e-05, "loss": 0.1111, "step": 3490 }, { "epoch": 0.64, "learning_rate": 1.9869069579488275e-05, "loss": 0.1279, "step": 3491 }, { "epoch": 0.64, "learning_rate": 1.9868949805307802e-05, "loss": 0.1135, "step": 3492 }, { "epoch": 0.64, "learning_rate": 1.9868829976729444e-05, "loss": 0.1094, "step": 3493 }, { "epoch": 0.64, "learning_rate": 1.986871009375386e-05, "loss": 0.1212, "step": 3494 }, { "epoch": 0.64, "learning_rate": 1.986859015638172e-05, "loss": 0.1201, "step": 3495 }, { "epoch": 0.64, "learning_rate": 1.986847016461368e-05, "loss": 0.1214, "step": 3496 }, { "epoch": 0.64, "learning_rate": 1.9868350118450403e-05, "loss": 0.1082, "step": 3497 }, { "epoch": 0.64, "learning_rate": 1.9868230017892544e-05, "loss": 0.0882, "step": 3498 }, { "epoch": 0.64, "learning_rate": 1.9868109862940773e-05, "loss": 0.1126, "step": 3499 }, { "epoch": 0.64, "learning_rate": 1.986798965359575e-05, "loss": 0.1086, "step": 3500 }, { "epoch": 0.64, "learning_rate": 1.986786938985813e-05, "loss": 0.1031, "step": 3501 }, { "epoch": 0.64, "learning_rate": 1.986774907172859e-05, "loss": 0.0968, "step": 3502 }, { "epoch": 0.64, "learning_rate": 1.9867628699207784e-05, "loss": 0.1048, "step": 3503 }, { "epoch": 0.64, "learning_rate": 1.9867508272296377e-05, "loss": 0.1138, "step": 3504 }, { "epoch": 0.64, "learning_rate": 1.9867387790995033e-05, "loss": 0.1126, "step": 3505 }, { "epoch": 0.64, "learning_rate": 1.9867267255304416e-05, "loss": 0.1102, "step": 3506 }, { "epoch": 0.64, "learning_rate": 1.9867146665225193e-05, "loss": 0.1064, "step": 3507 }, { "epoch": 0.64, "learning_rate": 1.9867026020758026e-05, "loss": 0.1031, "step": 3508 }, { "epoch": 0.64, "learning_rate": 1.9866905321903577e-05, "loss": 0.1011, "step": 3509 }, { "epoch": 0.64, "learning_rate": 1.986678456866252e-05, "loss": 0.1267, "step": 3510 }, { "epoch": 0.64, "learning_rate": 1.986666376103551e-05, "loss": 0.102, "step": 3511 }, { "epoch": 0.64, "learning_rate": 1.986654289902322e-05, "loss": 0.1053, "step": 3512 }, { "epoch": 0.64, "learning_rate": 1.9866421982626318e-05, "loss": 0.115, "step": 3513 }, { "epoch": 0.64, "learning_rate": 1.986630101184546e-05, "loss": 0.1083, "step": 3514 }, { "epoch": 0.64, "learning_rate": 1.9866179986681327e-05, "loss": 0.1118, "step": 3515 }, { "epoch": 0.64, "learning_rate": 1.9866058907134574e-05, "loss": 0.1175, "step": 3516 }, { "epoch": 0.64, "learning_rate": 1.986593777320587e-05, "loss": 0.1265, "step": 3517 }, { "epoch": 0.65, "learning_rate": 1.9865816584895894e-05, "loss": 0.1121, "step": 3518 }, { "epoch": 0.65, "learning_rate": 1.9865695342205303e-05, "loss": 0.1096, "step": 3519 }, { "epoch": 0.65, "learning_rate": 1.9865574045134764e-05, "loss": 0.1116, "step": 3520 }, { "epoch": 0.65, "learning_rate": 1.9865452693684957e-05, "loss": 0.1041, "step": 3521 }, { "epoch": 0.65, "learning_rate": 1.986533128785654e-05, "loss": 0.1038, "step": 3522 }, { "epoch": 0.65, "learning_rate": 1.9865209827650186e-05, "loss": 0.1149, "step": 3523 }, { "epoch": 0.65, "learning_rate": 1.9865088313066563e-05, "loss": 0.1206, "step": 3524 }, { "epoch": 0.65, "learning_rate": 1.986496674410634e-05, "loss": 0.1236, "step": 3525 }, { "epoch": 0.65, "learning_rate": 1.9864845120770194e-05, "loss": 0.1141, "step": 3526 }, { "epoch": 0.65, "learning_rate": 1.9864723443058787e-05, "loss": 0.1146, "step": 3527 }, { "epoch": 0.65, "learning_rate": 1.9864601710972794e-05, "loss": 0.119, "step": 3528 }, { "epoch": 0.65, "learning_rate": 1.986447992451289e-05, "loss": 0.1057, "step": 3529 }, { "epoch": 0.65, "learning_rate": 1.9864358083679734e-05, "loss": 0.1154, "step": 3530 }, { "epoch": 0.65, "learning_rate": 1.986423618847401e-05, "loss": 0.1038, "step": 3531 }, { "epoch": 0.65, "learning_rate": 1.9864114238896386e-05, "loss": 0.1282, "step": 3532 }, { "epoch": 0.65, "learning_rate": 1.9863992234947532e-05, "loss": 0.1076, "step": 3533 }, { "epoch": 0.65, "learning_rate": 1.986387017662812e-05, "loss": 0.1133, "step": 3534 }, { "epoch": 0.65, "learning_rate": 1.9863748063938823e-05, "loss": 0.1229, "step": 3535 }, { "epoch": 0.65, "learning_rate": 1.986362589688032e-05, "loss": 0.1141, "step": 3536 }, { "epoch": 0.65, "learning_rate": 1.9863503675453277e-05, "loss": 0.1121, "step": 3537 }, { "epoch": 0.65, "learning_rate": 1.9863381399658373e-05, "loss": 0.1139, "step": 3538 }, { "epoch": 0.65, "learning_rate": 1.9863259069496276e-05, "loss": 0.1049, "step": 3539 }, { "epoch": 0.65, "learning_rate": 1.9863136684967665e-05, "loss": 0.1184, "step": 3540 }, { "epoch": 0.65, "learning_rate": 1.9863014246073216e-05, "loss": 0.1131, "step": 3541 }, { "epoch": 0.65, "learning_rate": 1.98628917528136e-05, "loss": 0.1065, "step": 3542 }, { "epoch": 0.65, "learning_rate": 1.9862769205189494e-05, "loss": 0.1206, "step": 3543 }, { "epoch": 0.65, "learning_rate": 1.9862646603201572e-05, "loss": 0.1026, "step": 3544 }, { "epoch": 0.65, "learning_rate": 1.9862523946850513e-05, "loss": 0.1088, "step": 3545 }, { "epoch": 0.65, "learning_rate": 1.986240123613699e-05, "loss": 0.1208, "step": 3546 }, { "epoch": 0.65, "learning_rate": 1.986227847106168e-05, "loss": 0.1102, "step": 3547 }, { "epoch": 0.65, "learning_rate": 1.986215565162526e-05, "loss": 0.106, "step": 3548 }, { "epoch": 0.65, "learning_rate": 1.9862032777828405e-05, "loss": 0.102, "step": 3549 }, { "epoch": 0.65, "learning_rate": 1.9861909849671795e-05, "loss": 0.1157, "step": 3550 }, { "epoch": 0.65, "learning_rate": 1.986178686715611e-05, "loss": 0.1274, "step": 3551 }, { "epoch": 0.65, "learning_rate": 1.9861663830282022e-05, "loss": 0.1207, "step": 3552 }, { "epoch": 0.65, "learning_rate": 1.9861540739050215e-05, "loss": 0.1052, "step": 3553 }, { "epoch": 0.65, "learning_rate": 1.9861417593461362e-05, "loss": 0.105, "step": 3554 }, { "epoch": 0.65, "learning_rate": 1.9861294393516145e-05, "loss": 0.1071, "step": 3555 }, { "epoch": 0.65, "learning_rate": 1.986117113921524e-05, "loss": 0.1287, "step": 3556 }, { "epoch": 0.65, "learning_rate": 1.986104783055933e-05, "loss": 0.1038, "step": 3557 }, { "epoch": 0.65, "learning_rate": 1.9860924467549096e-05, "loss": 0.0974, "step": 3558 }, { "epoch": 0.65, "learning_rate": 1.986080105018521e-05, "loss": 0.1144, "step": 3559 }, { "epoch": 0.65, "learning_rate": 1.9860677578468363e-05, "loss": 0.111, "step": 3560 }, { "epoch": 0.65, "learning_rate": 1.9860554052399225e-05, "loss": 0.1251, "step": 3561 }, { "epoch": 0.65, "learning_rate": 1.9860430471978487e-05, "loss": 0.1017, "step": 3562 }, { "epoch": 0.65, "learning_rate": 1.9860306837206823e-05, "loss": 0.1185, "step": 3563 }, { "epoch": 0.65, "learning_rate": 1.9860183148084914e-05, "loss": 0.1092, "step": 3564 }, { "epoch": 0.65, "learning_rate": 1.986005940461345e-05, "loss": 0.1116, "step": 3565 }, { "epoch": 0.65, "learning_rate": 1.9859935606793104e-05, "loss": 0.1059, "step": 3566 }, { "epoch": 0.65, "learning_rate": 1.9859811754624562e-05, "loss": 0.1125, "step": 3567 }, { "epoch": 0.65, "learning_rate": 1.985968784810851e-05, "loss": 0.1179, "step": 3568 }, { "epoch": 0.65, "learning_rate": 1.9859563887245624e-05, "loss": 0.1124, "step": 3569 }, { "epoch": 0.65, "learning_rate": 1.9859439872036592e-05, "loss": 0.1161, "step": 3570 }, { "epoch": 0.65, "learning_rate": 1.9859315802482097e-05, "loss": 0.116, "step": 3571 }, { "epoch": 0.66, "learning_rate": 1.985919167858282e-05, "loss": 0.1002, "step": 3572 }, { "epoch": 0.66, "learning_rate": 1.985906750033945e-05, "loss": 0.1065, "step": 3573 }, { "epoch": 0.66, "learning_rate": 1.9858943267752666e-05, "loss": 0.114, "step": 3574 }, { "epoch": 0.66, "learning_rate": 1.985881898082316e-05, "loss": 0.1043, "step": 3575 }, { "epoch": 0.66, "learning_rate": 1.985869463955161e-05, "loss": 0.1208, "step": 3576 }, { "epoch": 0.66, "learning_rate": 1.9858570243938704e-05, "loss": 0.1199, "step": 3577 }, { "epoch": 0.66, "learning_rate": 1.985844579398513e-05, "loss": 0.1258, "step": 3578 }, { "epoch": 0.66, "learning_rate": 1.985832128969157e-05, "loss": 0.1204, "step": 3579 }, { "epoch": 0.66, "learning_rate": 1.9858196731058712e-05, "loss": 0.1123, "step": 3580 }, { "epoch": 0.66, "learning_rate": 1.9858072118087244e-05, "loss": 0.1133, "step": 3581 }, { "epoch": 0.66, "learning_rate": 1.985794745077785e-05, "loss": 0.1126, "step": 3582 }, { "epoch": 0.66, "learning_rate": 1.985782272913122e-05, "loss": 0.1073, "step": 3583 }, { "epoch": 0.66, "learning_rate": 1.985769795314804e-05, "loss": 0.1244, "step": 3584 }, { "epoch": 0.66, "learning_rate": 1.9857573122828997e-05, "loss": 0.0981, "step": 3585 }, { "epoch": 0.66, "learning_rate": 1.985744823817478e-05, "loss": 0.1098, "step": 3586 }, { "epoch": 0.66, "learning_rate": 1.9857323299186077e-05, "loss": 0.1028, "step": 3587 }, { "epoch": 0.66, "learning_rate": 1.9857198305863582e-05, "loss": 0.1029, "step": 3588 }, { "epoch": 0.66, "learning_rate": 1.985707325820797e-05, "loss": 0.1066, "step": 3589 }, { "epoch": 0.66, "learning_rate": 1.9856948156219948e-05, "loss": 0.1077, "step": 3590 }, { "epoch": 0.66, "learning_rate": 1.985682299990019e-05, "loss": 0.1178, "step": 3591 }, { "epoch": 0.66, "learning_rate": 1.9856697789249397e-05, "loss": 0.0989, "step": 3592 }, { "epoch": 0.66, "learning_rate": 1.9856572524268253e-05, "loss": 0.1158, "step": 3593 }, { "epoch": 0.66, "learning_rate": 1.985644720495745e-05, "loss": 0.1155, "step": 3594 }, { "epoch": 0.66, "learning_rate": 1.9856321831317683e-05, "loss": 0.1255, "step": 3595 }, { "epoch": 0.66, "learning_rate": 1.9856196403349636e-05, "loss": 0.1071, "step": 3596 }, { "epoch": 0.66, "learning_rate": 1.9856070921054004e-05, "loss": 0.1217, "step": 3597 }, { "epoch": 0.66, "learning_rate": 1.9855945384431477e-05, "loss": 0.1111, "step": 3598 }, { "epoch": 0.66, "learning_rate": 1.9855819793482748e-05, "loss": 0.1128, "step": 3599 }, { "epoch": 0.66, "learning_rate": 1.985569414820851e-05, "loss": 0.0997, "step": 3600 }, { "epoch": 0.66, "learning_rate": 1.9855568448609458e-05, "loss": 0.117, "step": 3601 }, { "epoch": 0.66, "learning_rate": 1.985544269468628e-05, "loss": 0.1218, "step": 3602 }, { "epoch": 0.66, "learning_rate": 1.9855316886439668e-05, "loss": 0.1199, "step": 3603 }, { "epoch": 0.66, "learning_rate": 1.985519102387032e-05, "loss": 0.1103, "step": 3604 }, { "epoch": 0.66, "learning_rate": 1.985506510697893e-05, "loss": 0.1147, "step": 3605 }, { "epoch": 0.66, "learning_rate": 1.985493913576619e-05, "loss": 0.1062, "step": 3606 }, { "epoch": 0.66, "learning_rate": 1.9854813110232794e-05, "loss": 0.1062, "step": 3607 }, { "epoch": 0.66, "learning_rate": 1.9854687030379436e-05, "loss": 0.1063, "step": 3608 }, { "epoch": 0.66, "learning_rate": 1.9854560896206813e-05, "loss": 0.11, "step": 3609 }, { "epoch": 0.66, "learning_rate": 1.9854434707715618e-05, "loss": 0.1181, "step": 3610 }, { "epoch": 0.66, "learning_rate": 1.985430846490655e-05, "loss": 0.1211, "step": 3611 }, { "epoch": 0.66, "learning_rate": 1.9854182167780304e-05, "loss": 0.1207, "step": 3612 }, { "epoch": 0.66, "learning_rate": 1.9854055816337575e-05, "loss": 0.1046, "step": 3613 }, { "epoch": 0.66, "learning_rate": 1.9853929410579057e-05, "loss": 0.1052, "step": 3614 }, { "epoch": 0.66, "learning_rate": 1.985380295050545e-05, "loss": 0.1051, "step": 3615 }, { "epoch": 0.66, "learning_rate": 1.985367643611745e-05, "loss": 0.1185, "step": 3616 }, { "epoch": 0.66, "learning_rate": 1.9853549867415758e-05, "loss": 0.1114, "step": 3617 }, { "epoch": 0.66, "learning_rate": 1.9853423244401065e-05, "loss": 0.1077, "step": 3618 }, { "epoch": 0.66, "learning_rate": 1.9853296567074075e-05, "loss": 0.1125, "step": 3619 }, { "epoch": 0.66, "learning_rate": 1.985316983543548e-05, "loss": 0.1058, "step": 3620 }, { "epoch": 0.66, "learning_rate": 1.985304304948598e-05, "loss": 0.1112, "step": 3621 }, { "epoch": 0.66, "learning_rate": 1.9852916209226284e-05, "loss": 0.1061, "step": 3622 }, { "epoch": 0.66, "learning_rate": 1.9852789314657074e-05, "loss": 0.1035, "step": 3623 }, { "epoch": 0.66, "learning_rate": 1.9852662365779066e-05, "loss": 0.109, "step": 3624 }, { "epoch": 0.66, "learning_rate": 1.9852535362592947e-05, "loss": 0.1118, "step": 3625 }, { "epoch": 0.66, "learning_rate": 1.9852408305099423e-05, "loss": 0.1158, "step": 3626 }, { "epoch": 0.67, "learning_rate": 1.9852281193299198e-05, "loss": 0.1103, "step": 3627 }, { "epoch": 0.67, "learning_rate": 1.9852154027192963e-05, "loss": 0.1092, "step": 3628 }, { "epoch": 0.67, "learning_rate": 1.9852026806781427e-05, "loss": 0.1092, "step": 3629 }, { "epoch": 0.67, "learning_rate": 1.9851899532065287e-05, "loss": 0.1273, "step": 3630 }, { "epoch": 0.67, "learning_rate": 1.9851772203045248e-05, "loss": 0.1154, "step": 3631 }, { "epoch": 0.67, "learning_rate": 1.9851644819722008e-05, "loss": 0.1119, "step": 3632 }, { "epoch": 0.67, "learning_rate": 1.985151738209627e-05, "loss": 0.124, "step": 3633 }, { "epoch": 0.67, "learning_rate": 1.9851389890168738e-05, "loss": 0.1075, "step": 3634 }, { "epoch": 0.67, "learning_rate": 1.9851262343940117e-05, "loss": 0.1166, "step": 3635 }, { "epoch": 0.67, "learning_rate": 1.9851134743411104e-05, "loss": 0.1126, "step": 3636 }, { "epoch": 0.67, "learning_rate": 1.9851007088582407e-05, "loss": 0.1045, "step": 3637 }, { "epoch": 0.67, "learning_rate": 1.9850879379454727e-05, "loss": 0.1247, "step": 3638 }, { "epoch": 0.67, "learning_rate": 1.9850751616028772e-05, "loss": 0.108, "step": 3639 }, { "epoch": 0.67, "learning_rate": 1.985062379830524e-05, "loss": 0.1129, "step": 3640 }, { "epoch": 0.67, "learning_rate": 1.985049592628484e-05, "loss": 0.1129, "step": 3641 }, { "epoch": 0.67, "learning_rate": 1.9850367999968275e-05, "loss": 0.102, "step": 3642 }, { "epoch": 0.67, "learning_rate": 1.9850240019356252e-05, "loss": 0.1125, "step": 3643 }, { "epoch": 0.67, "learning_rate": 1.9850111984449472e-05, "loss": 0.116, "step": 3644 }, { "epoch": 0.67, "learning_rate": 1.984998389524865e-05, "loss": 0.1083, "step": 3645 }, { "epoch": 0.67, "learning_rate": 1.9849855751754485e-05, "loss": 0.1089, "step": 3646 }, { "epoch": 0.67, "learning_rate": 1.984972755396768e-05, "loss": 0.1095, "step": 3647 }, { "epoch": 0.67, "learning_rate": 1.984959930188895e-05, "loss": 0.0909, "step": 3648 }, { "epoch": 0.67, "learning_rate": 1.9849470995518993e-05, "loss": 0.1109, "step": 3649 }, { "epoch": 0.67, "learning_rate": 1.9849342634858524e-05, "loss": 0.1114, "step": 3650 }, { "epoch": 0.67, "learning_rate": 1.9849214219908247e-05, "loss": 0.1129, "step": 3651 }, { "epoch": 0.67, "learning_rate": 1.9849085750668874e-05, "loss": 0.1095, "step": 3652 }, { "epoch": 0.67, "learning_rate": 1.9848957227141106e-05, "loss": 0.1138, "step": 3653 }, { "epoch": 0.67, "learning_rate": 1.9848828649325654e-05, "loss": 0.1142, "step": 3654 }, { "epoch": 0.67, "learning_rate": 1.984870001722323e-05, "loss": 0.1177, "step": 3655 }, { "epoch": 0.67, "learning_rate": 1.984857133083454e-05, "loss": 0.107, "step": 3656 }, { "epoch": 0.67, "learning_rate": 1.9848442590160295e-05, "loss": 0.1091, "step": 3657 }, { "epoch": 0.67, "learning_rate": 1.9848313795201205e-05, "loss": 0.1044, "step": 3658 }, { "epoch": 0.67, "learning_rate": 1.9848184945957976e-05, "loss": 0.1156, "step": 3659 }, { "epoch": 0.67, "learning_rate": 1.984805604243132e-05, "loss": 0.0943, "step": 3660 }, { "epoch": 0.67, "learning_rate": 1.984792708462195e-05, "loss": 0.0998, "step": 3661 }, { "epoch": 0.67, "learning_rate": 1.9847798072530575e-05, "loss": 0.0962, "step": 3662 }, { "epoch": 0.67, "learning_rate": 1.9847669006157907e-05, "loss": 0.1176, "step": 3663 }, { "epoch": 0.67, "learning_rate": 1.9847539885504658e-05, "loss": 0.109, "step": 3664 }, { "epoch": 0.67, "learning_rate": 1.9847410710571535e-05, "loss": 0.0931, "step": 3665 }, { "epoch": 0.67, "learning_rate": 1.9847281481359256e-05, "loss": 0.116, "step": 3666 }, { "epoch": 0.67, "learning_rate": 1.984715219786853e-05, "loss": 0.1084, "step": 3667 }, { "epoch": 0.67, "learning_rate": 1.984702286010007e-05, "loss": 0.1003, "step": 3668 }, { "epoch": 0.67, "learning_rate": 1.984689346805459e-05, "loss": 0.1064, "step": 3669 }, { "epoch": 0.67, "learning_rate": 1.98467640217328e-05, "loss": 0.1121, "step": 3670 }, { "epoch": 0.67, "learning_rate": 1.9846634521135422e-05, "loss": 0.1213, "step": 3671 }, { "epoch": 0.67, "learning_rate": 1.984650496626316e-05, "loss": 0.1154, "step": 3672 }, { "epoch": 0.67, "learning_rate": 1.9846375357116733e-05, "loss": 0.1092, "step": 3673 }, { "epoch": 0.67, "learning_rate": 1.984624569369685e-05, "loss": 0.102, "step": 3674 }, { "epoch": 0.67, "learning_rate": 1.9846115976004234e-05, "loss": 0.1113, "step": 3675 }, { "epoch": 0.67, "learning_rate": 1.9845986204039596e-05, "loss": 0.1011, "step": 3676 }, { "epoch": 0.67, "learning_rate": 1.9845856377803653e-05, "loss": 0.1218, "step": 3677 }, { "epoch": 0.67, "learning_rate": 1.9845726497297116e-05, "loss": 0.1077, "step": 3678 }, { "epoch": 0.67, "learning_rate": 1.98455965625207e-05, "loss": 0.1203, "step": 3679 }, { "epoch": 0.67, "learning_rate": 1.9845466573475134e-05, "loss": 0.1133, "step": 3680 }, { "epoch": 0.68, "learning_rate": 1.984533653016112e-05, "loss": 0.1093, "step": 3681 }, { "epoch": 0.68, "learning_rate": 1.984520643257938e-05, "loss": 0.1171, "step": 3682 }, { "epoch": 0.68, "learning_rate": 1.9845076280730634e-05, "loss": 0.1145, "step": 3683 }, { "epoch": 0.68, "learning_rate": 1.9844946074615596e-05, "loss": 0.113, "step": 3684 }, { "epoch": 0.68, "learning_rate": 1.9844815814234986e-05, "loss": 0.1162, "step": 3685 }, { "epoch": 0.68, "learning_rate": 1.984468549958952e-05, "loss": 0.1076, "step": 3686 }, { "epoch": 0.68, "learning_rate": 1.9844555130679914e-05, "loss": 0.11, "step": 3687 }, { "epoch": 0.68, "learning_rate": 1.9844424707506893e-05, "loss": 0.1109, "step": 3688 }, { "epoch": 0.68, "learning_rate": 1.984429423007117e-05, "loss": 0.1156, "step": 3689 }, { "epoch": 0.68, "learning_rate": 1.9844163698373472e-05, "loss": 0.1144, "step": 3690 }, { "epoch": 0.68, "learning_rate": 1.9844033112414505e-05, "loss": 0.1256, "step": 3691 }, { "epoch": 0.68, "learning_rate": 1.9843902472195e-05, "loss": 0.1035, "step": 3692 }, { "epoch": 0.68, "learning_rate": 1.9843771777715676e-05, "loss": 0.1131, "step": 3693 }, { "epoch": 0.68, "learning_rate": 1.984364102897725e-05, "loss": 0.1074, "step": 3694 }, { "epoch": 0.68, "learning_rate": 1.9843510225980443e-05, "loss": 0.1263, "step": 3695 }, { "epoch": 0.68, "learning_rate": 1.9843379368725978e-05, "loss": 0.1078, "step": 3696 }, { "epoch": 0.68, "learning_rate": 1.9843248457214576e-05, "loss": 0.1077, "step": 3697 }, { "epoch": 0.68, "learning_rate": 1.9843117491446957e-05, "loss": 0.14, "step": 3698 }, { "epoch": 0.68, "learning_rate": 1.9842986471423845e-05, "loss": 0.1309, "step": 3699 }, { "epoch": 0.68, "learning_rate": 1.984285539714596e-05, "loss": 0.1095, "step": 3700 }, { "epoch": 0.68, "learning_rate": 1.9842724268614027e-05, "loss": 0.1198, "step": 3701 }, { "epoch": 0.68, "learning_rate": 1.9842593085828764e-05, "loss": 0.1146, "step": 3702 }, { "epoch": 0.68, "learning_rate": 1.9842461848790898e-05, "loss": 0.1021, "step": 3703 }, { "epoch": 0.68, "learning_rate": 1.9842330557501155e-05, "loss": 0.0987, "step": 3704 }, { "epoch": 0.68, "learning_rate": 1.9842199211960253e-05, "loss": 0.1044, "step": 3705 }, { "epoch": 0.68, "learning_rate": 1.9842067812168916e-05, "loss": 0.0953, "step": 3706 }, { "epoch": 0.68, "learning_rate": 1.9841936358127874e-05, "loss": 0.1153, "step": 3707 }, { "epoch": 0.68, "learning_rate": 1.9841804849837847e-05, "loss": 0.1119, "step": 3708 }, { "epoch": 0.68, "learning_rate": 1.9841673287299562e-05, "loss": 0.1094, "step": 3709 }, { "epoch": 0.68, "learning_rate": 1.9841541670513743e-05, "loss": 0.1199, "step": 3710 }, { "epoch": 0.68, "learning_rate": 1.9841409999481115e-05, "loss": 0.117, "step": 3711 }, { "epoch": 0.68, "learning_rate": 1.9841278274202406e-05, "loss": 0.1055, "step": 3712 }, { "epoch": 0.68, "learning_rate": 1.984114649467834e-05, "loss": 0.1094, "step": 3713 }, { "epoch": 0.68, "learning_rate": 1.984101466090964e-05, "loss": 0.1043, "step": 3714 }, { "epoch": 0.68, "learning_rate": 1.984088277289704e-05, "loss": 0.1078, "step": 3715 }, { "epoch": 0.68, "learning_rate": 1.9840750830641265e-05, "loss": 0.1066, "step": 3716 }, { "epoch": 0.68, "learning_rate": 1.984061883414304e-05, "loss": 0.1008, "step": 3717 }, { "epoch": 0.68, "learning_rate": 1.9840486783403093e-05, "loss": 0.1045, "step": 3718 }, { "epoch": 0.68, "learning_rate": 1.9840354678422154e-05, "loss": 0.1217, "step": 3719 }, { "epoch": 0.68, "learning_rate": 1.9840222519200948e-05, "loss": 0.1167, "step": 3720 }, { "epoch": 0.68, "learning_rate": 1.9840090305740204e-05, "loss": 0.1149, "step": 3721 }, { "epoch": 0.68, "learning_rate": 1.983995803804065e-05, "loss": 0.1107, "step": 3722 }, { "epoch": 0.68, "learning_rate": 1.983982571610302e-05, "loss": 0.1054, "step": 3723 }, { "epoch": 0.68, "learning_rate": 1.983969333992804e-05, "loss": 0.1177, "step": 3724 }, { "epoch": 0.68, "learning_rate": 1.983956090951644e-05, "loss": 0.1157, "step": 3725 }, { "epoch": 0.68, "learning_rate": 1.983942842486895e-05, "loss": 0.106, "step": 3726 }, { "epoch": 0.68, "learning_rate": 1.98392958859863e-05, "loss": 0.1242, "step": 3727 }, { "epoch": 0.68, "learning_rate": 1.9839163292869218e-05, "loss": 0.1076, "step": 3728 }, { "epoch": 0.68, "learning_rate": 1.983903064551844e-05, "loss": 0.1157, "step": 3729 }, { "epoch": 0.68, "learning_rate": 1.983889794393469e-05, "loss": 0.1092, "step": 3730 }, { "epoch": 0.68, "learning_rate": 1.983876518811871e-05, "loss": 0.1138, "step": 3731 }, { "epoch": 0.68, "learning_rate": 1.9838632378071222e-05, "loss": 0.1046, "step": 3732 }, { "epoch": 0.68, "learning_rate": 1.9838499513792964e-05, "loss": 0.1166, "step": 3733 }, { "epoch": 0.68, "learning_rate": 1.9838366595284665e-05, "loss": 0.1076, "step": 3734 }, { "epoch": 0.68, "learning_rate": 1.983823362254706e-05, "loss": 0.1192, "step": 3735 }, { "epoch": 0.69, "learning_rate": 1.983810059558088e-05, "loss": 0.1149, "step": 3736 }, { "epoch": 0.69, "learning_rate": 1.983796751438686e-05, "loss": 0.0999, "step": 3737 }, { "epoch": 0.69, "learning_rate": 1.983783437896573e-05, "loss": 0.105, "step": 3738 }, { "epoch": 0.69, "learning_rate": 1.983770118931823e-05, "loss": 0.1183, "step": 3739 }, { "epoch": 0.69, "learning_rate": 1.9837567945445085e-05, "loss": 0.1154, "step": 3740 }, { "epoch": 0.69, "learning_rate": 1.983743464734704e-05, "loss": 0.1129, "step": 3741 }, { "epoch": 0.69, "learning_rate": 1.9837301295024825e-05, "loss": 0.1125, "step": 3742 }, { "epoch": 0.69, "learning_rate": 1.983716788847917e-05, "loss": 0.1133, "step": 3743 }, { "epoch": 0.69, "learning_rate": 1.983703442771082e-05, "loss": 0.1098, "step": 3744 }, { "epoch": 0.69, "learning_rate": 1.9836900912720505e-05, "loss": 0.1182, "step": 3745 }, { "epoch": 0.69, "learning_rate": 1.9836767343508963e-05, "loss": 0.1121, "step": 3746 }, { "epoch": 0.69, "learning_rate": 1.9836633720076924e-05, "loss": 0.1194, "step": 3747 }, { "epoch": 0.69, "learning_rate": 1.983650004242513e-05, "loss": 0.1263, "step": 3748 }, { "epoch": 0.69, "learning_rate": 1.9836366310554322e-05, "loss": 0.1032, "step": 3749 }, { "epoch": 0.69, "learning_rate": 1.9836232524465233e-05, "loss": 0.1231, "step": 3750 }, { "epoch": 0.69, "learning_rate": 1.9836098684158594e-05, "loss": 0.1268, "step": 3751 }, { "epoch": 0.69, "learning_rate": 1.983596478963515e-05, "loss": 0.1214, "step": 3752 }, { "epoch": 0.69, "learning_rate": 1.983583084089564e-05, "loss": 0.1191, "step": 3753 }, { "epoch": 0.69, "learning_rate": 1.98356968379408e-05, "loss": 0.1183, "step": 3754 }, { "epoch": 0.69, "learning_rate": 1.9835562780771365e-05, "loss": 0.1023, "step": 3755 }, { "epoch": 0.69, "learning_rate": 1.9835428669388085e-05, "loss": 0.1098, "step": 3756 }, { "epoch": 0.69, "learning_rate": 1.9835294503791687e-05, "loss": 0.1165, "step": 3757 }, { "epoch": 0.69, "learning_rate": 1.9835160283982913e-05, "loss": 0.0979, "step": 3758 }, { "epoch": 0.69, "learning_rate": 1.9835026009962507e-05, "loss": 0.1088, "step": 3759 }, { "epoch": 0.69, "learning_rate": 1.983489168173121e-05, "loss": 0.1072, "step": 3760 }, { "epoch": 0.69, "learning_rate": 1.9834757299289755e-05, "loss": 0.112, "step": 3761 }, { "epoch": 0.69, "learning_rate": 1.983462286263889e-05, "loss": 0.1057, "step": 3762 }, { "epoch": 0.69, "learning_rate": 1.9834488371779354e-05, "loss": 0.1128, "step": 3763 }, { "epoch": 0.69, "learning_rate": 1.9834353826711887e-05, "loss": 0.1113, "step": 3764 }, { "epoch": 0.69, "learning_rate": 1.9834219227437233e-05, "loss": 0.1088, "step": 3765 }, { "epoch": 0.69, "learning_rate": 1.983408457395613e-05, "loss": 0.1079, "step": 3766 }, { "epoch": 0.69, "learning_rate": 1.983394986626932e-05, "loss": 0.1083, "step": 3767 }, { "epoch": 0.69, "learning_rate": 1.9833815104377553e-05, "loss": 0.1064, "step": 3768 }, { "epoch": 0.69, "learning_rate": 1.9833680288281563e-05, "loss": 0.1082, "step": 3769 }, { "epoch": 0.69, "learning_rate": 1.9833545417982098e-05, "loss": 0.1138, "step": 3770 }, { "epoch": 0.69, "learning_rate": 1.98334104934799e-05, "loss": 0.1068, "step": 3771 }, { "epoch": 0.69, "learning_rate": 1.9833275514775716e-05, "loss": 0.11, "step": 3772 }, { "epoch": 0.69, "learning_rate": 1.9833140481870282e-05, "loss": 0.1129, "step": 3773 }, { "epoch": 0.69, "learning_rate": 1.9833005394764352e-05, "loss": 0.1178, "step": 3774 }, { "epoch": 0.69, "learning_rate": 1.9832870253458664e-05, "loss": 0.1057, "step": 3775 }, { "epoch": 0.69, "learning_rate": 1.9832735057953964e-05, "loss": 0.1188, "step": 3776 }, { "epoch": 0.69, "learning_rate": 1.9832599808251e-05, "loss": 0.1045, "step": 3777 }, { "epoch": 0.69, "learning_rate": 1.983246450435051e-05, "loss": 0.1143, "step": 3778 }, { "epoch": 0.69, "learning_rate": 1.983232914625325e-05, "loss": 0.1069, "step": 3779 }, { "epoch": 0.69, "learning_rate": 1.983219373395996e-05, "loss": 0.1082, "step": 3780 }, { "epoch": 0.69, "learning_rate": 1.983205826747139e-05, "loss": 0.1187, "step": 3781 }, { "epoch": 0.69, "learning_rate": 1.9831922746788284e-05, "loss": 0.1145, "step": 3782 }, { "epoch": 0.69, "learning_rate": 1.9831787171911388e-05, "loss": 0.1109, "step": 3783 }, { "epoch": 0.69, "learning_rate": 1.9831651542841452e-05, "loss": 0.1201, "step": 3784 }, { "epoch": 0.69, "learning_rate": 1.9831515859579222e-05, "loss": 0.1047, "step": 3785 }, { "epoch": 0.69, "learning_rate": 1.9831380122125444e-05, "loss": 0.1133, "step": 3786 }, { "epoch": 0.69, "learning_rate": 1.983124433048087e-05, "loss": 0.1109, "step": 3787 }, { "epoch": 0.69, "learning_rate": 1.9831108484646248e-05, "loss": 0.1103, "step": 3788 }, { "epoch": 0.69, "learning_rate": 1.9830972584622327e-05, "loss": 0.1075, "step": 3789 }, { "epoch": 0.7, "learning_rate": 1.983083663040985e-05, "loss": 0.1016, "step": 3790 }, { "epoch": 0.7, "learning_rate": 1.9830700622009575e-05, "loss": 0.1161, "step": 3791 }, { "epoch": 0.7, "learning_rate": 1.9830564559422247e-05, "loss": 0.1084, "step": 3792 }, { "epoch": 0.7, "learning_rate": 1.9830428442648616e-05, "loss": 0.1103, "step": 3793 }, { "epoch": 0.7, "learning_rate": 1.9830292271689437e-05, "loss": 0.1167, "step": 3794 }, { "epoch": 0.7, "learning_rate": 1.983015604654545e-05, "loss": 0.1087, "step": 3795 }, { "epoch": 0.7, "learning_rate": 1.9830019767217416e-05, "loss": 0.1177, "step": 3796 }, { "epoch": 0.7, "learning_rate": 1.982988343370608e-05, "loss": 0.1045, "step": 3797 }, { "epoch": 0.7, "learning_rate": 1.98297470460122e-05, "loss": 0.1201, "step": 3798 }, { "epoch": 0.7, "learning_rate": 1.9829610604136527e-05, "loss": 0.1128, "step": 3799 }, { "epoch": 0.7, "learning_rate": 1.9829474108079804e-05, "loss": 0.1117, "step": 3800 }, { "epoch": 0.7, "learning_rate": 1.9829337557842795e-05, "loss": 0.1103, "step": 3801 }, { "epoch": 0.7, "learning_rate": 1.982920095342624e-05, "loss": 0.1053, "step": 3802 }, { "epoch": 0.7, "learning_rate": 1.9829064294830903e-05, "loss": 0.1124, "step": 3803 }, { "epoch": 0.7, "learning_rate": 1.9828927582057534e-05, "loss": 0.1174, "step": 3804 }, { "epoch": 0.7, "learning_rate": 1.9828790815106888e-05, "loss": 0.1163, "step": 3805 }, { "epoch": 0.7, "learning_rate": 1.9828653993979713e-05, "loss": 0.1108, "step": 3806 }, { "epoch": 0.7, "learning_rate": 1.9828517118676768e-05, "loss": 0.1101, "step": 3807 }, { "epoch": 0.7, "learning_rate": 1.9828380189198805e-05, "loss": 0.1221, "step": 3808 }, { "epoch": 0.7, "learning_rate": 1.982824320554658e-05, "loss": 0.1165, "step": 3809 }, { "epoch": 0.7, "learning_rate": 1.982810616772085e-05, "loss": 0.1135, "step": 3810 }, { "epoch": 0.7, "learning_rate": 1.982796907572237e-05, "loss": 0.103, "step": 3811 }, { "epoch": 0.7, "learning_rate": 1.982783192955189e-05, "loss": 0.0995, "step": 3812 }, { "epoch": 0.7, "learning_rate": 1.9827694729210173e-05, "loss": 0.1205, "step": 3813 }, { "epoch": 0.7, "learning_rate": 1.982755747469797e-05, "loss": 0.1197, "step": 3814 }, { "epoch": 0.7, "learning_rate": 1.9827420166016043e-05, "loss": 0.1133, "step": 3815 }, { "epoch": 0.7, "learning_rate": 1.9827282803165145e-05, "loss": 0.1118, "step": 3816 }, { "epoch": 0.7, "learning_rate": 1.9827145386146032e-05, "loss": 0.1084, "step": 3817 }, { "epoch": 0.7, "learning_rate": 1.9827007914959468e-05, "loss": 0.1065, "step": 3818 }, { "epoch": 0.7, "learning_rate": 1.98268703896062e-05, "loss": 0.1115, "step": 3819 }, { "epoch": 0.7, "learning_rate": 1.9826732810087e-05, "loss": 0.1205, "step": 3820 }, { "epoch": 0.7, "learning_rate": 1.982659517640261e-05, "loss": 0.1174, "step": 3821 }, { "epoch": 0.7, "learning_rate": 1.98264574885538e-05, "loss": 0.119, "step": 3822 }, { "epoch": 0.7, "learning_rate": 1.9826319746541327e-05, "loss": 0.1166, "step": 3823 }, { "epoch": 0.7, "learning_rate": 1.982618195036595e-05, "loss": 0.1169, "step": 3824 }, { "epoch": 0.7, "learning_rate": 1.9826044100028428e-05, "loss": 0.1121, "step": 3825 }, { "epoch": 0.7, "learning_rate": 1.982590619552952e-05, "loss": 0.1143, "step": 3826 }, { "epoch": 0.7, "learning_rate": 1.9825768236869987e-05, "loss": 0.1159, "step": 3827 }, { "epoch": 0.7, "learning_rate": 1.982563022405059e-05, "loss": 0.1168, "step": 3828 }, { "epoch": 0.7, "learning_rate": 1.982549215707209e-05, "loss": 0.1298, "step": 3829 }, { "epoch": 0.7, "learning_rate": 1.9825354035935243e-05, "loss": 0.1175, "step": 3830 }, { "epoch": 0.7, "learning_rate": 1.9825215860640815e-05, "loss": 0.1172, "step": 3831 }, { "epoch": 0.7, "learning_rate": 1.982507763118957e-05, "loss": 0.0974, "step": 3832 }, { "epoch": 0.7, "learning_rate": 1.9824939347582264e-05, "loss": 0.1086, "step": 3833 }, { "epoch": 0.7, "learning_rate": 1.9824801009819663e-05, "loss": 0.1159, "step": 3834 }, { "epoch": 0.7, "learning_rate": 1.9824662617902532e-05, "loss": 0.0985, "step": 3835 }, { "epoch": 0.7, "learning_rate": 1.9824524171831626e-05, "loss": 0.1165, "step": 3836 }, { "epoch": 0.7, "learning_rate": 1.9824385671607716e-05, "loss": 0.115, "step": 3837 }, { "epoch": 0.7, "learning_rate": 1.982424711723156e-05, "loss": 0.1181, "step": 3838 }, { "epoch": 0.7, "learning_rate": 1.9824108508703923e-05, "loss": 0.1095, "step": 3839 }, { "epoch": 0.7, "learning_rate": 1.982396984602557e-05, "loss": 0.1122, "step": 3840 }, { "epoch": 0.7, "learning_rate": 1.9823831129197268e-05, "loss": 0.1071, "step": 3841 }, { "epoch": 0.7, "learning_rate": 1.9823692358219774e-05, "loss": 0.0972, "step": 3842 }, { "epoch": 0.7, "learning_rate": 1.9823553533093862e-05, "loss": 0.1088, "step": 3843 }, { "epoch": 0.7, "learning_rate": 1.982341465382029e-05, "loss": 0.0878, "step": 3844 }, { "epoch": 0.71, "learning_rate": 1.982327572039983e-05, "loss": 0.0999, "step": 3845 }, { "epoch": 0.71, "learning_rate": 1.982313673283324e-05, "loss": 0.1209, "step": 3846 }, { "epoch": 0.71, "learning_rate": 1.9822997691121292e-05, "loss": 0.1066, "step": 3847 }, { "epoch": 0.71, "learning_rate": 1.9822858595264748e-05, "loss": 0.108, "step": 3848 }, { "epoch": 0.71, "learning_rate": 1.982271944526438e-05, "loss": 0.1159, "step": 3849 }, { "epoch": 0.71, "learning_rate": 1.9822580241120953e-05, "loss": 0.1108, "step": 3850 }, { "epoch": 0.71, "learning_rate": 1.982244098283523e-05, "loss": 0.1049, "step": 3851 }, { "epoch": 0.71, "learning_rate": 1.9822301670407985e-05, "loss": 0.1196, "step": 3852 }, { "epoch": 0.71, "learning_rate": 1.9822162303839986e-05, "loss": 0.111, "step": 3853 }, { "epoch": 0.71, "learning_rate": 1.982202288313199e-05, "loss": 0.1067, "step": 3854 }, { "epoch": 0.71, "learning_rate": 1.9821883408284782e-05, "loss": 0.1102, "step": 3855 }, { "epoch": 0.71, "learning_rate": 1.982174387929912e-05, "loss": 0.1042, "step": 3856 }, { "epoch": 0.71, "learning_rate": 1.9821604296175774e-05, "loss": 0.1094, "step": 3857 }, { "epoch": 0.71, "learning_rate": 1.9821464658915513e-05, "loss": 0.1142, "step": 3858 }, { "epoch": 0.71, "learning_rate": 1.9821324967519113e-05, "loss": 0.1035, "step": 3859 }, { "epoch": 0.71, "learning_rate": 1.982118522198734e-05, "loss": 0.1053, "step": 3860 }, { "epoch": 0.71, "learning_rate": 1.9821045422320963e-05, "loss": 0.1045, "step": 3861 }, { "epoch": 0.71, "learning_rate": 1.982090556852075e-05, "loss": 0.1194, "step": 3862 }, { "epoch": 0.71, "learning_rate": 1.982076566058748e-05, "loss": 0.1226, "step": 3863 }, { "epoch": 0.71, "learning_rate": 1.9820625698521918e-05, "loss": 0.0954, "step": 3864 }, { "epoch": 0.71, "learning_rate": 1.9820485682324836e-05, "loss": 0.0924, "step": 3865 }, { "epoch": 0.71, "learning_rate": 1.982034561199701e-05, "loss": 0.1045, "step": 3866 }, { "epoch": 0.71, "learning_rate": 1.9820205487539206e-05, "loss": 0.1113, "step": 3867 }, { "epoch": 0.71, "learning_rate": 1.98200653089522e-05, "loss": 0.1056, "step": 3868 }, { "epoch": 0.71, "learning_rate": 1.981992507623676e-05, "loss": 0.1064, "step": 3869 }, { "epoch": 0.71, "learning_rate": 1.981978478939367e-05, "loss": 0.1063, "step": 3870 }, { "epoch": 0.71, "learning_rate": 1.981964444842369e-05, "loss": 0.1093, "step": 3871 }, { "epoch": 0.71, "learning_rate": 1.9819504053327605e-05, "loss": 0.1066, "step": 3872 }, { "epoch": 0.71, "learning_rate": 1.9819363604106182e-05, "loss": 0.1195, "step": 3873 }, { "epoch": 0.71, "learning_rate": 1.9819223100760198e-05, "loss": 0.112, "step": 3874 }, { "epoch": 0.71, "learning_rate": 1.981908254329042e-05, "loss": 0.1197, "step": 3875 }, { "epoch": 0.71, "learning_rate": 1.9818941931697634e-05, "loss": 0.101, "step": 3876 }, { "epoch": 0.71, "learning_rate": 1.981880126598261e-05, "loss": 0.0964, "step": 3877 }, { "epoch": 0.71, "learning_rate": 1.9818660546146124e-05, "loss": 0.109, "step": 3878 }, { "epoch": 0.71, "learning_rate": 1.981851977218895e-05, "loss": 0.1158, "step": 3879 }, { "epoch": 0.71, "learning_rate": 1.9818378944111864e-05, "loss": 0.1095, "step": 3880 }, { "epoch": 0.71, "learning_rate": 1.9818238061915642e-05, "loss": 0.1004, "step": 3881 }, { "epoch": 0.71, "learning_rate": 1.9818097125601065e-05, "loss": 0.1071, "step": 3882 }, { "epoch": 0.71, "learning_rate": 1.9817956135168906e-05, "loss": 0.1042, "step": 3883 }, { "epoch": 0.71, "learning_rate": 1.981781509061994e-05, "loss": 0.101, "step": 3884 }, { "epoch": 0.71, "learning_rate": 1.981767399195495e-05, "loss": 0.1046, "step": 3885 }, { "epoch": 0.71, "learning_rate": 1.9817532839174713e-05, "loss": 0.1023, "step": 3886 }, { "epoch": 0.71, "learning_rate": 1.981739163228e-05, "loss": 0.0989, "step": 3887 }, { "epoch": 0.71, "learning_rate": 1.9817250371271597e-05, "loss": 0.1094, "step": 3888 }, { "epoch": 0.71, "learning_rate": 1.981710905615028e-05, "loss": 0.1072, "step": 3889 }, { "epoch": 0.71, "learning_rate": 1.981696768691683e-05, "loss": 0.0996, "step": 3890 }, { "epoch": 0.71, "learning_rate": 1.981682626357202e-05, "loss": 0.1041, "step": 3891 }, { "epoch": 0.71, "learning_rate": 1.9816684786116634e-05, "loss": 0.1024, "step": 3892 }, { "epoch": 0.71, "learning_rate": 1.9816543254551456e-05, "loss": 0.1184, "step": 3893 }, { "epoch": 0.71, "learning_rate": 1.9816401668877258e-05, "loss": 0.1023, "step": 3894 }, { "epoch": 0.71, "learning_rate": 1.9816260029094825e-05, "loss": 0.1164, "step": 3895 }, { "epoch": 0.71, "learning_rate": 1.9816118335204935e-05, "loss": 0.1181, "step": 3896 }, { "epoch": 0.71, "learning_rate": 1.981597658720837e-05, "loss": 0.1189, "step": 3897 }, { "epoch": 0.71, "learning_rate": 1.9815834785105915e-05, "loss": 0.115, "step": 3898 }, { "epoch": 0.72, "learning_rate": 1.981569292889835e-05, "loss": 0.1, "step": 3899 }, { "epoch": 0.72, "learning_rate": 1.981555101858645e-05, "loss": 0.1115, "step": 3900 }, { "epoch": 0.72, "learning_rate": 1.9815409054171008e-05, "loss": 0.1105, "step": 3901 }, { "epoch": 0.72, "learning_rate": 1.9815267035652802e-05, "loss": 0.1102, "step": 3902 }, { "epoch": 0.72, "learning_rate": 1.9815124963032607e-05, "loss": 0.1106, "step": 3903 }, { "epoch": 0.72, "learning_rate": 1.981498283631122e-05, "loss": 0.111, "step": 3904 }, { "epoch": 0.72, "learning_rate": 1.9814840655489414e-05, "loss": 0.1172, "step": 3905 }, { "epoch": 0.72, "learning_rate": 1.981469842056798e-05, "loss": 0.1176, "step": 3906 }, { "epoch": 0.72, "learning_rate": 1.9814556131547698e-05, "loss": 0.1046, "step": 3907 }, { "epoch": 0.72, "learning_rate": 1.981441378842935e-05, "loss": 0.1034, "step": 3908 }, { "epoch": 0.72, "learning_rate": 1.9814271391213725e-05, "loss": 0.117, "step": 3909 }, { "epoch": 0.72, "learning_rate": 1.9814128939901606e-05, "loss": 0.1007, "step": 3910 }, { "epoch": 0.72, "learning_rate": 1.9813986434493775e-05, "loss": 0.099, "step": 3911 }, { "epoch": 0.72, "learning_rate": 1.9813843874991026e-05, "loss": 0.1186, "step": 3912 }, { "epoch": 0.72, "learning_rate": 1.9813701261394136e-05, "loss": 0.1224, "step": 3913 }, { "epoch": 0.72, "learning_rate": 1.9813558593703897e-05, "loss": 0.1135, "step": 3914 }, { "epoch": 0.72, "learning_rate": 1.9813415871921093e-05, "loss": 0.1029, "step": 3915 }, { "epoch": 0.72, "learning_rate": 1.9813273096046508e-05, "loss": 0.1139, "step": 3916 }, { "epoch": 0.72, "learning_rate": 1.9813130266080934e-05, "loss": 0.1108, "step": 3917 }, { "epoch": 0.72, "learning_rate": 1.9812987382025154e-05, "loss": 0.1144, "step": 3918 }, { "epoch": 0.72, "learning_rate": 1.981284444387996e-05, "loss": 0.1135, "step": 3919 }, { "epoch": 0.72, "learning_rate": 1.981270145164614e-05, "loss": 0.1059, "step": 3920 }, { "epoch": 0.72, "learning_rate": 1.9812558405324475e-05, "loss": 0.116, "step": 3921 }, { "epoch": 0.72, "learning_rate": 1.981241530491576e-05, "loss": 0.1119, "step": 3922 }, { "epoch": 0.72, "learning_rate": 1.9812272150420783e-05, "loss": 0.1093, "step": 3923 }, { "epoch": 0.72, "learning_rate": 1.9812128941840328e-05, "loss": 0.1139, "step": 3924 }, { "epoch": 0.72, "learning_rate": 1.9811985679175193e-05, "loss": 0.0981, "step": 3925 }, { "epoch": 0.72, "learning_rate": 1.981184236242616e-05, "loss": 0.1088, "step": 3926 }, { "epoch": 0.72, "learning_rate": 1.9811698991594027e-05, "loss": 0.1046, "step": 3927 }, { "epoch": 0.72, "learning_rate": 1.9811555566679576e-05, "loss": 0.1163, "step": 3928 }, { "epoch": 0.72, "learning_rate": 1.9811412087683598e-05, "loss": 0.1206, "step": 3929 }, { "epoch": 0.72, "learning_rate": 1.981126855460689e-05, "loss": 0.1026, "step": 3930 }, { "epoch": 0.72, "learning_rate": 1.981112496745024e-05, "loss": 0.1238, "step": 3931 }, { "epoch": 0.72, "learning_rate": 1.9810981326214437e-05, "loss": 0.1032, "step": 3932 }, { "epoch": 0.72, "learning_rate": 1.9810837630900278e-05, "loss": 0.128, "step": 3933 }, { "epoch": 0.72, "learning_rate": 1.9810693881508548e-05, "loss": 0.1137, "step": 3934 }, { "epoch": 0.72, "learning_rate": 1.981055007804005e-05, "loss": 0.1126, "step": 3935 }, { "epoch": 0.72, "learning_rate": 1.9810406220495567e-05, "loss": 0.1069, "step": 3936 }, { "epoch": 0.72, "learning_rate": 1.9810262308875892e-05, "loss": 0.1094, "step": 3937 }, { "epoch": 0.72, "learning_rate": 1.9810118343181824e-05, "loss": 0.1182, "step": 3938 }, { "epoch": 0.72, "learning_rate": 1.9809974323414154e-05, "loss": 0.1113, "step": 3939 }, { "epoch": 0.72, "learning_rate": 1.9809830249573677e-05, "loss": 0.1183, "step": 3940 }, { "epoch": 0.72, "learning_rate": 1.980968612166118e-05, "loss": 0.1216, "step": 3941 }, { "epoch": 0.72, "learning_rate": 1.980954193967747e-05, "loss": 0.1118, "step": 3942 }, { "epoch": 0.72, "learning_rate": 1.9809397703623333e-05, "loss": 0.1163, "step": 3943 }, { "epoch": 0.72, "learning_rate": 1.9809253413499565e-05, "loss": 0.1153, "step": 3944 }, { "epoch": 0.72, "learning_rate": 1.980910906930696e-05, "loss": 0.1044, "step": 3945 }, { "epoch": 0.72, "learning_rate": 1.9808964671046322e-05, "loss": 0.0955, "step": 3946 }, { "epoch": 0.72, "learning_rate": 1.980882021871844e-05, "loss": 0.1051, "step": 3947 }, { "epoch": 0.72, "learning_rate": 1.9808675712324108e-05, "loss": 0.121, "step": 3948 }, { "epoch": 0.72, "learning_rate": 1.9808531151864127e-05, "loss": 0.1033, "step": 3949 }, { "epoch": 0.72, "learning_rate": 1.980838653733929e-05, "loss": 0.1043, "step": 3950 }, { "epoch": 0.72, "learning_rate": 1.98082418687504e-05, "loss": 0.1196, "step": 3951 }, { "epoch": 0.72, "learning_rate": 1.980809714609825e-05, "loss": 0.1178, "step": 3952 }, { "epoch": 0.72, "learning_rate": 1.9807952369383638e-05, "loss": 0.1051, "step": 3953 }, { "epoch": 0.73, "learning_rate": 1.9807807538607363e-05, "loss": 0.1055, "step": 3954 }, { "epoch": 0.73, "learning_rate": 1.9807662653770223e-05, "loss": 0.1106, "step": 3955 }, { "epoch": 0.73, "learning_rate": 1.980751771487302e-05, "loss": 0.1001, "step": 3956 }, { "epoch": 0.73, "learning_rate": 1.9807372721916545e-05, "loss": 0.1064, "step": 3957 }, { "epoch": 0.73, "learning_rate": 1.9807227674901605e-05, "loss": 0.1099, "step": 3958 }, { "epoch": 0.73, "learning_rate": 1.9807082573828992e-05, "loss": 0.1211, "step": 3959 }, { "epoch": 0.73, "learning_rate": 1.9806937418699515e-05, "loss": 0.1099, "step": 3960 }, { "epoch": 0.73, "learning_rate": 1.980679220951397e-05, "loss": 0.1156, "step": 3961 }, { "epoch": 0.73, "learning_rate": 1.980664694627315e-05, "loss": 0.0969, "step": 3962 }, { "epoch": 0.73, "learning_rate": 1.9806501628977864e-05, "loss": 0.1198, "step": 3963 }, { "epoch": 0.73, "learning_rate": 1.9806356257628913e-05, "loss": 0.1095, "step": 3964 }, { "epoch": 0.73, "learning_rate": 1.98062108322271e-05, "loss": 0.1035, "step": 3965 }, { "epoch": 0.73, "learning_rate": 1.9806065352773218e-05, "loss": 0.1104, "step": 3966 }, { "epoch": 0.73, "learning_rate": 1.9805919819268076e-05, "loss": 0.1178, "step": 3967 }, { "epoch": 0.73, "learning_rate": 1.9805774231712475e-05, "loss": 0.1129, "step": 3968 }, { "epoch": 0.73, "learning_rate": 1.9805628590107216e-05, "loss": 0.1088, "step": 3969 }, { "epoch": 0.73, "learning_rate": 1.9805482894453104e-05, "loss": 0.1229, "step": 3970 }, { "epoch": 0.73, "learning_rate": 1.9805337144750937e-05, "loss": 0.1056, "step": 3971 }, { "epoch": 0.73, "learning_rate": 1.9805191341001522e-05, "loss": 0.1179, "step": 3972 }, { "epoch": 0.73, "learning_rate": 1.9805045483205666e-05, "loss": 0.1055, "step": 3973 }, { "epoch": 0.73, "learning_rate": 1.980489957136417e-05, "loss": 0.109, "step": 3974 }, { "epoch": 0.73, "learning_rate": 1.9804753605477835e-05, "loss": 0.1147, "step": 3975 }, { "epoch": 0.73, "learning_rate": 1.9804607585547472e-05, "loss": 0.1121, "step": 3976 }, { "epoch": 0.73, "learning_rate": 1.980446151157388e-05, "loss": 0.1115, "step": 3977 }, { "epoch": 0.73, "learning_rate": 1.9804315383557864e-05, "loss": 0.1177, "step": 3978 }, { "epoch": 0.73, "learning_rate": 1.9804169201500235e-05, "loss": 0.1047, "step": 3979 }, { "epoch": 0.73, "learning_rate": 1.9804022965401796e-05, "loss": 0.1113, "step": 3980 }, { "epoch": 0.73, "learning_rate": 1.980387667526335e-05, "loss": 0.113, "step": 3981 }, { "epoch": 0.73, "learning_rate": 1.9803730331085708e-05, "loss": 0.114, "step": 3982 }, { "epoch": 0.73, "learning_rate": 1.9803583932869675e-05, "loss": 0.1191, "step": 3983 }, { "epoch": 0.73, "learning_rate": 1.9803437480616054e-05, "loss": 0.1189, "step": 3984 }, { "epoch": 0.73, "learning_rate": 1.9803290974325658e-05, "loss": 0.1221, "step": 3985 }, { "epoch": 0.73, "learning_rate": 1.9803144413999297e-05, "loss": 0.1042, "step": 3986 }, { "epoch": 0.73, "learning_rate": 1.980299779963777e-05, "loss": 0.0998, "step": 3987 }, { "epoch": 0.73, "learning_rate": 1.980285113124189e-05, "loss": 0.1221, "step": 3988 }, { "epoch": 0.73, "learning_rate": 1.9802704408812463e-05, "loss": 0.1051, "step": 3989 }, { "epoch": 0.73, "learning_rate": 1.9802557632350304e-05, "loss": 0.1197, "step": 3990 }, { "epoch": 0.73, "learning_rate": 1.9802410801856214e-05, "loss": 0.0937, "step": 3991 }, { "epoch": 0.73, "learning_rate": 1.9802263917331007e-05, "loss": 0.1012, "step": 3992 }, { "epoch": 0.73, "learning_rate": 1.9802116978775493e-05, "loss": 0.126, "step": 3993 }, { "epoch": 0.73, "learning_rate": 1.980196998619048e-05, "loss": 0.1138, "step": 3994 }, { "epoch": 0.73, "learning_rate": 1.980182293957678e-05, "loss": 0.1084, "step": 3995 }, { "epoch": 0.73, "learning_rate": 1.98016758389352e-05, "loss": 0.1125, "step": 3996 }, { "epoch": 0.73, "learning_rate": 1.9801528684266554e-05, "loss": 0.118, "step": 3997 }, { "epoch": 0.73, "learning_rate": 1.980138147557165e-05, "loss": 0.0978, "step": 3998 }, { "epoch": 0.73, "learning_rate": 1.9801234212851305e-05, "loss": 0.1122, "step": 3999 }, { "epoch": 0.73, "learning_rate": 1.9801086896106327e-05, "loss": 0.1026, "step": 4000 }, { "epoch": 0.73, "learning_rate": 1.9800939525337527e-05, "loss": 0.1035, "step": 4001 }, { "epoch": 0.73, "learning_rate": 1.980079210054572e-05, "loss": 0.1146, "step": 4002 }, { "epoch": 0.73, "learning_rate": 1.9800644621731716e-05, "loss": 0.1143, "step": 4003 }, { "epoch": 0.73, "learning_rate": 1.980049708889633e-05, "loss": 0.1278, "step": 4004 }, { "epoch": 0.73, "learning_rate": 1.9800349502040377e-05, "loss": 0.1126, "step": 4005 }, { "epoch": 0.73, "learning_rate": 1.9800201861164665e-05, "loss": 0.1073, "step": 4006 }, { "epoch": 0.73, "learning_rate": 1.980005416627001e-05, "loss": 0.1068, "step": 4007 }, { "epoch": 0.74, "learning_rate": 1.979990641735723e-05, "loss": 0.1079, "step": 4008 }, { "epoch": 0.74, "learning_rate": 1.9799758614427134e-05, "loss": 0.0962, "step": 4009 }, { "epoch": 0.74, "learning_rate": 1.9799610757480538e-05, "loss": 0.1089, "step": 4010 }, { "epoch": 0.74, "learning_rate": 1.9799462846518258e-05, "loss": 0.1012, "step": 4011 }, { "epoch": 0.74, "learning_rate": 1.979931488154111e-05, "loss": 0.1109, "step": 4012 }, { "epoch": 0.74, "learning_rate": 1.979916686254991e-05, "loss": 0.1199, "step": 4013 }, { "epoch": 0.74, "learning_rate": 1.9799018789545472e-05, "loss": 0.1176, "step": 4014 }, { "epoch": 0.74, "learning_rate": 1.9798870662528615e-05, "loss": 0.1179, "step": 4015 }, { "epoch": 0.74, "learning_rate": 1.979872248150015e-05, "loss": 0.1201, "step": 4016 }, { "epoch": 0.74, "learning_rate": 1.9798574246460898e-05, "loss": 0.1101, "step": 4017 }, { "epoch": 0.74, "learning_rate": 1.9798425957411672e-05, "loss": 0.1067, "step": 4018 }, { "epoch": 0.74, "learning_rate": 1.97982776143533e-05, "loss": 0.1232, "step": 4019 }, { "epoch": 0.74, "learning_rate": 1.979812921728659e-05, "loss": 0.1122, "step": 4020 }, { "epoch": 0.74, "learning_rate": 1.9797980766212358e-05, "loss": 0.1156, "step": 4021 }, { "epoch": 0.74, "learning_rate": 1.9797832261131428e-05, "loss": 0.115, "step": 4022 }, { "epoch": 0.74, "learning_rate": 1.979768370204462e-05, "loss": 0.1121, "step": 4023 }, { "epoch": 0.74, "learning_rate": 1.9797535088952748e-05, "loss": 0.129, "step": 4024 }, { "epoch": 0.74, "learning_rate": 1.979738642185663e-05, "loss": 0.1022, "step": 4025 }, { "epoch": 0.74, "learning_rate": 1.979723770075709e-05, "loss": 0.1179, "step": 4026 }, { "epoch": 0.74, "learning_rate": 1.979708892565495e-05, "loss": 0.1129, "step": 4027 }, { "epoch": 0.74, "learning_rate": 1.9796940096551022e-05, "loss": 0.105, "step": 4028 }, { "epoch": 0.74, "learning_rate": 1.9796791213446135e-05, "loss": 0.1122, "step": 4029 }, { "epoch": 0.74, "learning_rate": 1.9796642276341102e-05, "loss": 0.0993, "step": 4030 }, { "epoch": 0.74, "learning_rate": 1.979649328523675e-05, "loss": 0.1063, "step": 4031 }, { "epoch": 0.74, "learning_rate": 1.9796344240133894e-05, "loss": 0.1103, "step": 4032 }, { "epoch": 0.74, "learning_rate": 1.9796195141033362e-05, "loss": 0.1112, "step": 4033 }, { "epoch": 0.74, "learning_rate": 1.979604598793597e-05, "loss": 0.1144, "step": 4034 }, { "epoch": 0.74, "learning_rate": 1.9795896780842544e-05, "loss": 0.1137, "step": 4035 }, { "epoch": 0.74, "learning_rate": 1.9795747519753907e-05, "loss": 0.1115, "step": 4036 }, { "epoch": 0.74, "learning_rate": 1.9795598204670877e-05, "loss": 0.1128, "step": 4037 }, { "epoch": 0.74, "learning_rate": 1.9795448835594284e-05, "loss": 0.1068, "step": 4038 }, { "epoch": 0.74, "learning_rate": 1.9795299412524948e-05, "loss": 0.1065, "step": 4039 }, { "epoch": 0.74, "learning_rate": 1.979514993546369e-05, "loss": 0.111, "step": 4040 }, { "epoch": 0.74, "learning_rate": 1.9795000404411337e-05, "loss": 0.1133, "step": 4041 }, { "epoch": 0.74, "learning_rate": 1.9794850819368708e-05, "loss": 0.1184, "step": 4042 }, { "epoch": 0.74, "learning_rate": 1.9794701180336636e-05, "loss": 0.1062, "step": 4043 }, { "epoch": 0.74, "learning_rate": 1.979455148731594e-05, "loss": 0.1008, "step": 4044 }, { "epoch": 0.74, "learning_rate": 1.979440174030745e-05, "loss": 0.1002, "step": 4045 }, { "epoch": 0.74, "learning_rate": 1.9794251939311987e-05, "loss": 0.0967, "step": 4046 }, { "epoch": 0.74, "learning_rate": 1.9794102084330376e-05, "loss": 0.1111, "step": 4047 }, { "epoch": 0.74, "learning_rate": 1.9793952175363446e-05, "loss": 0.1245, "step": 4048 }, { "epoch": 0.74, "learning_rate": 1.979380221241202e-05, "loss": 0.099, "step": 4049 }, { "epoch": 0.74, "learning_rate": 1.979365219547693e-05, "loss": 0.1126, "step": 4050 }, { "epoch": 0.74, "learning_rate": 1.9793502124559e-05, "loss": 0.1167, "step": 4051 }, { "epoch": 0.74, "learning_rate": 1.979335199965905e-05, "loss": 0.1131, "step": 4052 }, { "epoch": 0.74, "learning_rate": 1.979320182077792e-05, "loss": 0.1097, "step": 4053 }, { "epoch": 0.74, "learning_rate": 1.9793051587916432e-05, "loss": 0.1118, "step": 4054 }, { "epoch": 0.74, "learning_rate": 1.9792901301075416e-05, "loss": 0.1105, "step": 4055 }, { "epoch": 0.74, "learning_rate": 1.9792750960255697e-05, "loss": 0.1028, "step": 4056 }, { "epoch": 0.74, "learning_rate": 1.9792600565458105e-05, "loss": 0.1047, "step": 4057 }, { "epoch": 0.74, "learning_rate": 1.9792450116683465e-05, "loss": 0.1088, "step": 4058 }, { "epoch": 0.74, "learning_rate": 1.979229961393262e-05, "loss": 0.0967, "step": 4059 }, { "epoch": 0.74, "learning_rate": 1.979214905720638e-05, "loss": 0.1204, "step": 4060 }, { "epoch": 0.74, "learning_rate": 1.979199844650559e-05, "loss": 0.1045, "step": 4061 }, { "epoch": 0.74, "learning_rate": 1.9791847781831076e-05, "loss": 0.1126, "step": 4062 }, { "epoch": 0.75, "learning_rate": 1.9791697063183664e-05, "loss": 0.1172, "step": 4063 }, { "epoch": 0.75, "learning_rate": 1.979154629056419e-05, "loss": 0.1164, "step": 4064 }, { "epoch": 0.75, "learning_rate": 1.9791395463973486e-05, "loss": 0.1037, "step": 4065 }, { "epoch": 0.75, "learning_rate": 1.979124458341238e-05, "loss": 0.1042, "step": 4066 }, { "epoch": 0.75, "learning_rate": 1.9791093648881704e-05, "loss": 0.0993, "step": 4067 }, { "epoch": 0.75, "learning_rate": 1.9790942660382287e-05, "loss": 0.1135, "step": 4068 }, { "epoch": 0.75, "learning_rate": 1.979079161791497e-05, "loss": 0.1089, "step": 4069 }, { "epoch": 0.75, "learning_rate": 1.9790640521480577e-05, "loss": 0.1109, "step": 4070 }, { "epoch": 0.75, "learning_rate": 1.9790489371079947e-05, "loss": 0.1043, "step": 4071 }, { "epoch": 0.75, "learning_rate": 1.9790338166713905e-05, "loss": 0.11, "step": 4072 }, { "epoch": 0.75, "learning_rate": 1.9790186908383294e-05, "loss": 0.117, "step": 4073 }, { "epoch": 0.75, "learning_rate": 1.979003559608894e-05, "loss": 0.1273, "step": 4074 }, { "epoch": 0.75, "learning_rate": 1.9789884229831685e-05, "loss": 0.1125, "step": 4075 }, { "epoch": 0.75, "learning_rate": 1.9789732809612355e-05, "loss": 0.1091, "step": 4076 }, { "epoch": 0.75, "learning_rate": 1.9789581335431792e-05, "loss": 0.1108, "step": 4077 }, { "epoch": 0.75, "learning_rate": 1.9789429807290825e-05, "loss": 0.1186, "step": 4078 }, { "epoch": 0.75, "learning_rate": 1.9789278225190292e-05, "loss": 0.1091, "step": 4079 }, { "epoch": 0.75, "learning_rate": 1.9789126589131026e-05, "loss": 0.105, "step": 4080 }, { "epoch": 0.75, "learning_rate": 1.9788974899113868e-05, "loss": 0.108, "step": 4081 }, { "epoch": 0.75, "learning_rate": 1.978882315513965e-05, "loss": 0.1181, "step": 4082 }, { "epoch": 0.75, "learning_rate": 1.978867135720921e-05, "loss": 0.1032, "step": 4083 }, { "epoch": 0.75, "learning_rate": 1.978851950532338e-05, "loss": 0.1098, "step": 4084 }, { "epoch": 0.75, "learning_rate": 1.9788367599483006e-05, "loss": 0.1055, "step": 4085 }, { "epoch": 0.75, "learning_rate": 1.9788215639688917e-05, "loss": 0.1104, "step": 4086 }, { "epoch": 0.75, "learning_rate": 1.978806362594196e-05, "loss": 0.1132, "step": 4087 }, { "epoch": 0.75, "learning_rate": 1.978791155824296e-05, "loss": 0.1163, "step": 4088 }, { "epoch": 0.75, "learning_rate": 1.9787759436592763e-05, "loss": 0.0923, "step": 4089 }, { "epoch": 0.75, "learning_rate": 1.9787607260992212e-05, "loss": 0.1198, "step": 4090 }, { "epoch": 0.75, "learning_rate": 1.9787455031442134e-05, "loss": 0.1045, "step": 4091 }, { "epoch": 0.75, "learning_rate": 1.9787302747943382e-05, "loss": 0.108, "step": 4092 }, { "epoch": 0.75, "learning_rate": 1.9787150410496783e-05, "loss": 0.1124, "step": 4093 }, { "epoch": 0.75, "learning_rate": 1.978699801910318e-05, "loss": 0.1154, "step": 4094 }, { "epoch": 0.75, "learning_rate": 1.978684557376342e-05, "loss": 0.1008, "step": 4095 }, { "epoch": 0.75, "learning_rate": 1.9786693074478335e-05, "loss": 0.113, "step": 4096 }, { "epoch": 0.75, "learning_rate": 1.978654052124877e-05, "loss": 0.0999, "step": 4097 }, { "epoch": 0.75, "learning_rate": 1.9786387914075563e-05, "loss": 0.1106, "step": 4098 }, { "epoch": 0.75, "learning_rate": 1.9786235252959555e-05, "loss": 0.1055, "step": 4099 }, { "epoch": 0.75, "learning_rate": 1.978608253790159e-05, "loss": 0.1103, "step": 4100 }, { "epoch": 0.75, "learning_rate": 1.978592976890251e-05, "loss": 0.115, "step": 4101 }, { "epoch": 0.75, "learning_rate": 1.9785776945963154e-05, "loss": 0.1032, "step": 4102 }, { "epoch": 0.75, "learning_rate": 1.9785624069084364e-05, "loss": 0.113, "step": 4103 }, { "epoch": 0.75, "learning_rate": 1.978547113826699e-05, "loss": 0.104, "step": 4104 }, { "epoch": 0.75, "learning_rate": 1.978531815351187e-05, "loss": 0.1018, "step": 4105 }, { "epoch": 0.75, "learning_rate": 1.9785165114819845e-05, "loss": 0.1093, "step": 4106 }, { "epoch": 0.75, "learning_rate": 1.9785012022191762e-05, "loss": 0.1081, "step": 4107 }, { "epoch": 0.75, "learning_rate": 1.978485887562846e-05, "loss": 0.1207, "step": 4108 }, { "epoch": 0.75, "learning_rate": 1.978470567513079e-05, "loss": 0.1234, "step": 4109 }, { "epoch": 0.75, "learning_rate": 1.9784552420699593e-05, "loss": 0.1103, "step": 4110 }, { "epoch": 0.75, "learning_rate": 1.9784399112335712e-05, "loss": 0.1052, "step": 4111 }, { "epoch": 0.75, "learning_rate": 1.9784245750039993e-05, "loss": 0.1086, "step": 4112 }, { "epoch": 0.75, "learning_rate": 1.9784092333813286e-05, "loss": 0.1058, "step": 4113 }, { "epoch": 0.75, "learning_rate": 1.978393886365643e-05, "loss": 0.1192, "step": 4114 }, { "epoch": 0.75, "learning_rate": 1.9783785339570275e-05, "loss": 0.1108, "step": 4115 }, { "epoch": 0.75, "learning_rate": 1.9783631761555667e-05, "loss": 0.1157, "step": 4116 }, { "epoch": 0.75, "learning_rate": 1.978347812961345e-05, "loss": 0.1171, "step": 4117 }, { "epoch": 0.76, "learning_rate": 1.978332444374447e-05, "loss": 0.1044, "step": 4118 }, { "epoch": 0.76, "learning_rate": 1.9783170703949582e-05, "loss": 0.1145, "step": 4119 }, { "epoch": 0.76, "learning_rate": 1.9783016910229622e-05, "loss": 0.1094, "step": 4120 }, { "epoch": 0.76, "learning_rate": 1.9782863062585447e-05, "loss": 0.1078, "step": 4121 }, { "epoch": 0.76, "learning_rate": 1.97827091610179e-05, "loss": 0.1165, "step": 4122 }, { "epoch": 0.76, "learning_rate": 1.978255520552783e-05, "loss": 0.1096, "step": 4123 }, { "epoch": 0.76, "learning_rate": 1.978240119611609e-05, "loss": 0.1104, "step": 4124 }, { "epoch": 0.76, "learning_rate": 1.9782247132783525e-05, "loss": 0.1116, "step": 4125 }, { "epoch": 0.76, "learning_rate": 1.9782093015530982e-05, "loss": 0.1004, "step": 4126 }, { "epoch": 0.76, "learning_rate": 1.9781938844359317e-05, "loss": 0.1116, "step": 4127 }, { "epoch": 0.76, "learning_rate": 1.9781784619269373e-05, "loss": 0.1018, "step": 4128 }, { "epoch": 0.76, "learning_rate": 1.9781630340262004e-05, "loss": 0.1168, "step": 4129 }, { "epoch": 0.76, "learning_rate": 1.9781476007338058e-05, "loss": 0.1045, "step": 4130 }, { "epoch": 0.76, "learning_rate": 1.9781321620498387e-05, "loss": 0.1097, "step": 4131 }, { "epoch": 0.76, "learning_rate": 1.9781167179743847e-05, "loss": 0.1027, "step": 4132 }, { "epoch": 0.76, "learning_rate": 1.9781012685075276e-05, "loss": 0.1074, "step": 4133 }, { "epoch": 0.76, "learning_rate": 1.978085813649354e-05, "loss": 0.1032, "step": 4134 }, { "epoch": 0.76, "learning_rate": 1.9780703533999484e-05, "loss": 0.1086, "step": 4135 }, { "epoch": 0.76, "learning_rate": 1.978054887759396e-05, "loss": 0.1111, "step": 4136 }, { "epoch": 0.76, "learning_rate": 1.978039416727782e-05, "loss": 0.1133, "step": 4137 }, { "epoch": 0.76, "learning_rate": 1.978023940305192e-05, "loss": 0.1132, "step": 4138 }, { "epoch": 0.76, "learning_rate": 1.978008458491711e-05, "loss": 0.111, "step": 4139 }, { "epoch": 0.76, "learning_rate": 1.9779929712874248e-05, "loss": 0.1105, "step": 4140 }, { "epoch": 0.76, "learning_rate": 1.977977478692418e-05, "loss": 0.0983, "step": 4141 }, { "epoch": 0.76, "learning_rate": 1.977961980706777e-05, "loss": 0.1136, "step": 4142 }, { "epoch": 0.76, "learning_rate": 1.9779464773305862e-05, "loss": 0.1057, "step": 4143 }, { "epoch": 0.76, "learning_rate": 1.9779309685639317e-05, "loss": 0.1182, "step": 4144 }, { "epoch": 0.76, "learning_rate": 1.9779154544068984e-05, "loss": 0.1004, "step": 4145 }, { "epoch": 0.76, "learning_rate": 1.9778999348595724e-05, "loss": 0.1071, "step": 4146 }, { "epoch": 0.76, "learning_rate": 1.9778844099220394e-05, "loss": 0.1089, "step": 4147 }, { "epoch": 0.76, "learning_rate": 1.9778688795943845e-05, "loss": 0.102, "step": 4148 }, { "epoch": 0.76, "learning_rate": 1.977853343876693e-05, "loss": 0.1136, "step": 4149 }, { "epoch": 0.76, "learning_rate": 1.9778378027690513e-05, "loss": 0.1111, "step": 4150 }, { "epoch": 0.76, "learning_rate": 1.977822256271545e-05, "loss": 0.1134, "step": 4151 }, { "epoch": 0.76, "learning_rate": 1.9778067043842587e-05, "loss": 0.1008, "step": 4152 }, { "epoch": 0.76, "learning_rate": 1.9777911471072795e-05, "loss": 0.117, "step": 4153 }, { "epoch": 0.76, "learning_rate": 1.9777755844406927e-05, "loss": 0.1089, "step": 4154 }, { "epoch": 0.76, "learning_rate": 1.9777600163845837e-05, "loss": 0.1088, "step": 4155 }, { "epoch": 0.76, "learning_rate": 1.9777444429390388e-05, "loss": 0.096, "step": 4156 }, { "epoch": 0.76, "learning_rate": 1.9777288641041434e-05, "loss": 0.1131, "step": 4157 }, { "epoch": 0.76, "learning_rate": 1.9777132798799837e-05, "loss": 0.0933, "step": 4158 }, { "epoch": 0.76, "learning_rate": 1.9776976902666452e-05, "loss": 0.1055, "step": 4159 }, { "epoch": 0.76, "learning_rate": 1.9776820952642145e-05, "loss": 0.1059, "step": 4160 }, { "epoch": 0.76, "learning_rate": 1.9776664948727772e-05, "loss": 0.094, "step": 4161 }, { "epoch": 0.76, "learning_rate": 1.977650889092419e-05, "loss": 0.1072, "step": 4162 }, { "epoch": 0.76, "learning_rate": 1.9776352779232267e-05, "loss": 0.1176, "step": 4163 }, { "epoch": 0.76, "learning_rate": 1.9776196613652854e-05, "loss": 0.0994, "step": 4164 }, { "epoch": 0.76, "learning_rate": 1.9776040394186817e-05, "loss": 0.0961, "step": 4165 }, { "epoch": 0.76, "learning_rate": 1.9775884120835018e-05, "loss": 0.1041, "step": 4166 }, { "epoch": 0.76, "learning_rate": 1.9775727793598312e-05, "loss": 0.1014, "step": 4167 }, { "epoch": 0.76, "learning_rate": 1.9775571412477573e-05, "loss": 0.1174, "step": 4168 }, { "epoch": 0.76, "learning_rate": 1.9775414977473648e-05, "loss": 0.118, "step": 4169 }, { "epoch": 0.76, "learning_rate": 1.9775258488587414e-05, "loss": 0.1143, "step": 4170 }, { "epoch": 0.76, "learning_rate": 1.977510194581972e-05, "loss": 0.0963, "step": 4171 }, { "epoch": 0.77, "learning_rate": 1.9774945349171435e-05, "loss": 0.1155, "step": 4172 }, { "epoch": 0.77, "learning_rate": 1.9774788698643427e-05, "loss": 0.101, "step": 4173 }, { "epoch": 0.77, "learning_rate": 1.977463199423655e-05, "loss": 0.1054, "step": 4174 }, { "epoch": 0.77, "learning_rate": 1.9774475235951676e-05, "loss": 0.1146, "step": 4175 }, { "epoch": 0.77, "learning_rate": 1.9774318423789666e-05, "loss": 0.093, "step": 4176 }, { "epoch": 0.77, "learning_rate": 1.977416155775138e-05, "loss": 0.1098, "step": 4177 }, { "epoch": 0.77, "learning_rate": 1.9774004637837688e-05, "loss": 0.1118, "step": 4178 }, { "epoch": 0.77, "learning_rate": 1.9773847664049452e-05, "loss": 0.1065, "step": 4179 }, { "epoch": 0.77, "learning_rate": 1.9773690636387542e-05, "loss": 0.1199, "step": 4180 }, { "epoch": 0.77, "learning_rate": 1.9773533554852818e-05, "loss": 0.1174, "step": 4181 }, { "epoch": 0.77, "learning_rate": 1.977337641944615e-05, "loss": 0.0988, "step": 4182 }, { "epoch": 0.77, "learning_rate": 1.9773219230168397e-05, "loss": 0.1096, "step": 4183 }, { "epoch": 0.77, "learning_rate": 1.9773061987020438e-05, "loss": 0.1054, "step": 4184 }, { "epoch": 0.77, "learning_rate": 1.9772904690003128e-05, "loss": 0.109, "step": 4185 }, { "epoch": 0.77, "learning_rate": 1.9772747339117338e-05, "loss": 0.1183, "step": 4186 }, { "epoch": 0.77, "learning_rate": 1.9772589934363936e-05, "loss": 0.1091, "step": 4187 }, { "epoch": 0.77, "learning_rate": 1.977243247574379e-05, "loss": 0.108, "step": 4188 }, { "epoch": 0.77, "learning_rate": 1.9772274963257773e-05, "loss": 0.1083, "step": 4189 }, { "epoch": 0.77, "learning_rate": 1.9772117396906737e-05, "loss": 0.1188, "step": 4190 }, { "epoch": 0.77, "learning_rate": 1.9771959776691568e-05, "loss": 0.1112, "step": 4191 }, { "epoch": 0.77, "learning_rate": 1.9771802102613127e-05, "loss": 0.1224, "step": 4192 }, { "epoch": 0.77, "learning_rate": 1.9771644374672284e-05, "loss": 0.1045, "step": 4193 }, { "epoch": 0.77, "learning_rate": 1.9771486592869907e-05, "loss": 0.1067, "step": 4194 }, { "epoch": 0.77, "learning_rate": 1.9771328757206863e-05, "loss": 0.1061, "step": 4195 }, { "epoch": 0.77, "learning_rate": 1.9771170867684033e-05, "loss": 0.107, "step": 4196 }, { "epoch": 0.77, "learning_rate": 1.9771012924302275e-05, "loss": 0.0953, "step": 4197 }, { "epoch": 0.77, "learning_rate": 1.9770854927062468e-05, "loss": 0.1178, "step": 4198 }, { "epoch": 0.77, "learning_rate": 1.977069687596548e-05, "loss": 0.1145, "step": 4199 }, { "epoch": 0.77, "learning_rate": 1.977053877101218e-05, "loss": 0.1062, "step": 4200 }, { "epoch": 0.77, "learning_rate": 1.9770380612203442e-05, "loss": 0.1158, "step": 4201 }, { "epoch": 0.77, "learning_rate": 1.9770222399540138e-05, "loss": 0.0999, "step": 4202 }, { "epoch": 0.77, "learning_rate": 1.977006413302314e-05, "loss": 0.1059, "step": 4203 }, { "epoch": 0.77, "learning_rate": 1.976990581265332e-05, "loss": 0.1061, "step": 4204 }, { "epoch": 0.77, "learning_rate": 1.9769747438431545e-05, "loss": 0.103, "step": 4205 }, { "epoch": 0.77, "learning_rate": 1.97695890103587e-05, "loss": 0.109, "step": 4206 }, { "epoch": 0.77, "learning_rate": 1.9769430528435646e-05, "loss": 0.1119, "step": 4207 }, { "epoch": 0.77, "learning_rate": 1.9769271992663262e-05, "loss": 0.1071, "step": 4208 }, { "epoch": 0.77, "learning_rate": 1.9769113403042426e-05, "loss": 0.1083, "step": 4209 }, { "epoch": 0.77, "learning_rate": 1.9768954759574004e-05, "loss": 0.1141, "step": 4210 }, { "epoch": 0.77, "learning_rate": 1.976879606225888e-05, "loss": 0.1096, "step": 4211 }, { "epoch": 0.77, "learning_rate": 1.976863731109792e-05, "loss": 0.1011, "step": 4212 }, { "epoch": 0.77, "learning_rate": 1.9768478506092e-05, "loss": 0.1298, "step": 4213 }, { "epoch": 0.77, "learning_rate": 1.9768319647242e-05, "loss": 0.1053, "step": 4214 }, { "epoch": 0.77, "learning_rate": 1.9768160734548793e-05, "loss": 0.1036, "step": 4215 }, { "epoch": 0.77, "learning_rate": 1.9768001768013258e-05, "loss": 0.1164, "step": 4216 }, { "epoch": 0.77, "learning_rate": 1.9767842747636262e-05, "loss": 0.1054, "step": 4217 }, { "epoch": 0.77, "learning_rate": 1.9767683673418695e-05, "loss": 0.1146, "step": 4218 }, { "epoch": 0.77, "learning_rate": 1.976752454536142e-05, "loss": 0.1051, "step": 4219 }, { "epoch": 0.77, "learning_rate": 1.9767365363465325e-05, "loss": 0.1135, "step": 4220 }, { "epoch": 0.77, "learning_rate": 1.9767206127731283e-05, "loss": 0.1126, "step": 4221 }, { "epoch": 0.77, "learning_rate": 1.976704683816017e-05, "loss": 0.1037, "step": 4222 }, { "epoch": 0.77, "learning_rate": 1.9766887494752873e-05, "loss": 0.1086, "step": 4223 }, { "epoch": 0.77, "learning_rate": 1.9766728097510255e-05, "loss": 0.0988, "step": 4224 }, { "epoch": 0.77, "learning_rate": 1.9766568646433205e-05, "loss": 0.1174, "step": 4225 }, { "epoch": 0.77, "learning_rate": 1.97664091415226e-05, "loss": 0.1083, "step": 4226 }, { "epoch": 0.78, "learning_rate": 1.976624958277932e-05, "loss": 0.1055, "step": 4227 }, { "epoch": 0.78, "learning_rate": 1.9766089970204244e-05, "loss": 0.1243, "step": 4228 }, { "epoch": 0.78, "learning_rate": 1.976593030379825e-05, "loss": 0.1129, "step": 4229 }, { "epoch": 0.78, "learning_rate": 1.9765770583562222e-05, "loss": 0.1147, "step": 4230 }, { "epoch": 0.78, "learning_rate": 1.9765610809497035e-05, "loss": 0.1125, "step": 4231 }, { "epoch": 0.78, "learning_rate": 1.976545098160357e-05, "loss": 0.1189, "step": 4232 }, { "epoch": 0.78, "learning_rate": 1.9765291099882715e-05, "loss": 0.1118, "step": 4233 }, { "epoch": 0.78, "learning_rate": 1.9765131164335345e-05, "loss": 0.0916, "step": 4234 }, { "epoch": 0.78, "learning_rate": 1.9764971174962345e-05, "loss": 0.1048, "step": 4235 }, { "epoch": 0.78, "learning_rate": 1.9764811131764592e-05, "loss": 0.0939, "step": 4236 }, { "epoch": 0.78, "learning_rate": 1.9764651034742974e-05, "loss": 0.1111, "step": 4237 }, { "epoch": 0.78, "learning_rate": 1.976449088389837e-05, "loss": 0.0973, "step": 4238 }, { "epoch": 0.78, "learning_rate": 1.976433067923166e-05, "loss": 0.1133, "step": 4239 }, { "epoch": 0.78, "learning_rate": 1.9764170420743736e-05, "loss": 0.1101, "step": 4240 }, { "epoch": 0.78, "learning_rate": 1.976401010843547e-05, "loss": 0.0969, "step": 4241 }, { "epoch": 0.78, "learning_rate": 1.9763849742307758e-05, "loss": 0.1089, "step": 4242 }, { "epoch": 0.78, "learning_rate": 1.9763689322361473e-05, "loss": 0.1252, "step": 4243 }, { "epoch": 0.78, "learning_rate": 1.9763528848597503e-05, "loss": 0.1108, "step": 4244 }, { "epoch": 0.78, "learning_rate": 1.976336832101674e-05, "loss": 0.0974, "step": 4245 }, { "epoch": 0.78, "learning_rate": 1.9763207739620055e-05, "loss": 0.1013, "step": 4246 }, { "epoch": 0.78, "learning_rate": 1.9763047104408343e-05, "loss": 0.1184, "step": 4247 }, { "epoch": 0.78, "learning_rate": 1.9762886415382486e-05, "loss": 0.1034, "step": 4248 }, { "epoch": 0.78, "learning_rate": 1.9762725672543372e-05, "loss": 0.0969, "step": 4249 }, { "epoch": 0.78, "learning_rate": 1.9762564875891883e-05, "loss": 0.1092, "step": 4250 }, { "epoch": 0.78, "learning_rate": 1.9762404025428908e-05, "loss": 0.115, "step": 4251 }, { "epoch": 0.78, "learning_rate": 1.9762243121155333e-05, "loss": 0.1204, "step": 4252 }, { "epoch": 0.78, "learning_rate": 1.976208216307205e-05, "loss": 0.1001, "step": 4253 }, { "epoch": 0.78, "learning_rate": 1.9761921151179937e-05, "loss": 0.1052, "step": 4254 }, { "epoch": 0.78, "learning_rate": 1.9761760085479883e-05, "loss": 0.0983, "step": 4255 }, { "epoch": 0.78, "learning_rate": 1.9761598965972784e-05, "loss": 0.105, "step": 4256 }, { "epoch": 0.78, "learning_rate": 1.9761437792659522e-05, "loss": 0.1087, "step": 4257 }, { "epoch": 0.78, "learning_rate": 1.9761276565540986e-05, "loss": 0.1048, "step": 4258 }, { "epoch": 0.78, "learning_rate": 1.9761115284618063e-05, "loss": 0.1006, "step": 4259 }, { "epoch": 0.78, "learning_rate": 1.9760953949891645e-05, "loss": 0.1092, "step": 4260 }, { "epoch": 0.78, "learning_rate": 1.976079256136262e-05, "loss": 0.1083, "step": 4261 }, { "epoch": 0.78, "learning_rate": 1.976063111903188e-05, "loss": 0.1119, "step": 4262 }, { "epoch": 0.78, "learning_rate": 1.9760469622900308e-05, "loss": 0.1088, "step": 4263 }, { "epoch": 0.78, "learning_rate": 1.9760308072968805e-05, "loss": 0.1044, "step": 4264 }, { "epoch": 0.78, "learning_rate": 1.976014646923825e-05, "loss": 0.1053, "step": 4265 }, { "epoch": 0.78, "learning_rate": 1.975998481170954e-05, "loss": 0.1074, "step": 4266 }, { "epoch": 0.78, "learning_rate": 1.9759823100383565e-05, "loss": 0.0934, "step": 4267 }, { "epoch": 0.78, "learning_rate": 1.9759661335261217e-05, "loss": 0.115, "step": 4268 }, { "epoch": 0.78, "learning_rate": 1.9759499516343387e-05, "loss": 0.0992, "step": 4269 }, { "epoch": 0.78, "learning_rate": 1.975933764363097e-05, "loss": 0.1003, "step": 4270 }, { "epoch": 0.78, "learning_rate": 1.975917571712485e-05, "loss": 0.1115, "step": 4271 }, { "epoch": 0.78, "learning_rate": 1.9759013736825928e-05, "loss": 0.1066, "step": 4272 }, { "epoch": 0.78, "learning_rate": 1.9758851702735092e-05, "loss": 0.1059, "step": 4273 }, { "epoch": 0.78, "learning_rate": 1.975868961485324e-05, "loss": 0.0944, "step": 4274 }, { "epoch": 0.78, "learning_rate": 1.975852747318126e-05, "loss": 0.1038, "step": 4275 }, { "epoch": 0.78, "learning_rate": 1.9758365277720045e-05, "loss": 0.0988, "step": 4276 }, { "epoch": 0.78, "learning_rate": 1.9758203028470497e-05, "loss": 0.1097, "step": 4277 }, { "epoch": 0.78, "learning_rate": 1.9758040725433503e-05, "loss": 0.0986, "step": 4278 }, { "epoch": 0.78, "learning_rate": 1.9757878368609957e-05, "loss": 0.1185, "step": 4279 }, { "epoch": 0.78, "learning_rate": 1.9757715958000762e-05, "loss": 0.1152, "step": 4280 }, { "epoch": 0.79, "learning_rate": 1.9757553493606804e-05, "loss": 0.1129, "step": 4281 }, { "epoch": 0.79, "learning_rate": 1.9757390975428987e-05, "loss": 0.1091, "step": 4282 }, { "epoch": 0.79, "learning_rate": 1.97572284034682e-05, "loss": 0.1148, "step": 4283 }, { "epoch": 0.79, "learning_rate": 1.9757065777725342e-05, "loss": 0.1129, "step": 4284 }, { "epoch": 0.79, "learning_rate": 1.975690309820131e-05, "loss": 0.114, "step": 4285 }, { "epoch": 0.79, "learning_rate": 1.9756740364896992e-05, "loss": 0.1156, "step": 4286 }, { "epoch": 0.79, "learning_rate": 1.97565775778133e-05, "loss": 0.1194, "step": 4287 }, { "epoch": 0.79, "learning_rate": 1.975641473695112e-05, "loss": 0.1072, "step": 4288 }, { "epoch": 0.79, "learning_rate": 1.9756251842311357e-05, "loss": 0.1051, "step": 4289 }, { "epoch": 0.79, "learning_rate": 1.9756088893894905e-05, "loss": 0.1204, "step": 4290 }, { "epoch": 0.79, "learning_rate": 1.975592589170266e-05, "loss": 0.1049, "step": 4291 }, { "epoch": 0.79, "learning_rate": 1.9755762835735524e-05, "loss": 0.1083, "step": 4292 }, { "epoch": 0.79, "learning_rate": 1.9755599725994394e-05, "loss": 0.1185, "step": 4293 }, { "epoch": 0.79, "learning_rate": 1.975543656248017e-05, "loss": 0.1033, "step": 4294 }, { "epoch": 0.79, "learning_rate": 1.9755273345193756e-05, "loss": 0.1041, "step": 4295 }, { "epoch": 0.79, "learning_rate": 1.975511007413604e-05, "loss": 0.1068, "step": 4296 }, { "epoch": 0.79, "learning_rate": 1.975494674930793e-05, "loss": 0.1094, "step": 4297 }, { "epoch": 0.79, "learning_rate": 1.975478337071033e-05, "loss": 0.1139, "step": 4298 }, { "epoch": 0.79, "learning_rate": 1.975461993834413e-05, "loss": 0.1106, "step": 4299 }, { "epoch": 0.79, "learning_rate": 1.9754456452210235e-05, "loss": 0.1148, "step": 4300 }, { "epoch": 0.79, "learning_rate": 1.975429291230955e-05, "loss": 0.1056, "step": 4301 }, { "epoch": 0.79, "learning_rate": 1.9754129318642977e-05, "loss": 0.1099, "step": 4302 }, { "epoch": 0.79, "learning_rate": 1.9753965671211413e-05, "loss": 0.1073, "step": 4303 }, { "epoch": 0.79, "learning_rate": 1.975380197001576e-05, "loss": 0.1053, "step": 4304 }, { "epoch": 0.79, "learning_rate": 1.9753638215056923e-05, "loss": 0.1143, "step": 4305 }, { "epoch": 0.79, "learning_rate": 1.9753474406335805e-05, "loss": 0.1074, "step": 4306 }, { "epoch": 0.79, "learning_rate": 1.9753310543853303e-05, "loss": 0.111, "step": 4307 }, { "epoch": 0.79, "learning_rate": 1.9753146627610328e-05, "loss": 0.1102, "step": 4308 }, { "epoch": 0.79, "learning_rate": 1.975298265760778e-05, "loss": 0.1272, "step": 4309 }, { "epoch": 0.79, "learning_rate": 1.9752818633846566e-05, "loss": 0.1284, "step": 4310 }, { "epoch": 0.79, "learning_rate": 1.9752654556327583e-05, "loss": 0.1021, "step": 4311 }, { "epoch": 0.79, "learning_rate": 1.975249042505174e-05, "loss": 0.1226, "step": 4312 }, { "epoch": 0.79, "learning_rate": 1.9752326240019943e-05, "loss": 0.1158, "step": 4313 }, { "epoch": 0.79, "learning_rate": 1.9752162001233097e-05, "loss": 0.1095, "step": 4314 }, { "epoch": 0.79, "learning_rate": 1.9751997708692103e-05, "loss": 0.1108, "step": 4315 }, { "epoch": 0.79, "learning_rate": 1.975183336239787e-05, "loss": 0.0976, "step": 4316 }, { "epoch": 0.79, "learning_rate": 1.97516689623513e-05, "loss": 0.1005, "step": 4317 }, { "epoch": 0.79, "learning_rate": 1.9751504508553306e-05, "loss": 0.1135, "step": 4318 }, { "epoch": 0.79, "learning_rate": 1.975134000100479e-05, "loss": 0.1066, "step": 4319 }, { "epoch": 0.79, "learning_rate": 1.975117543970666e-05, "loss": 0.1124, "step": 4320 }, { "epoch": 0.79, "learning_rate": 1.975101082465982e-05, "loss": 0.0971, "step": 4321 }, { "epoch": 0.79, "learning_rate": 1.9750846155865182e-05, "loss": 0.1012, "step": 4322 }, { "epoch": 0.79, "learning_rate": 1.975068143332365e-05, "loss": 0.1023, "step": 4323 }, { "epoch": 0.79, "learning_rate": 1.9750516657036133e-05, "loss": 0.1111, "step": 4324 }, { "epoch": 0.79, "learning_rate": 1.975035182700354e-05, "loss": 0.1022, "step": 4325 }, { "epoch": 0.79, "learning_rate": 1.9750186943226784e-05, "loss": 0.1006, "step": 4326 }, { "epoch": 0.79, "learning_rate": 1.9750022005706765e-05, "loss": 0.1038, "step": 4327 }, { "epoch": 0.79, "learning_rate": 1.9749857014444396e-05, "loss": 0.1106, "step": 4328 }, { "epoch": 0.79, "learning_rate": 1.9749691969440586e-05, "loss": 0.1158, "step": 4329 }, { "epoch": 0.79, "learning_rate": 1.9749526870696248e-05, "loss": 0.1063, "step": 4330 }, { "epoch": 0.79, "learning_rate": 1.974936171821229e-05, "loss": 0.1211, "step": 4331 }, { "epoch": 0.79, "learning_rate": 1.974919651198962e-05, "loss": 0.1251, "step": 4332 }, { "epoch": 0.79, "learning_rate": 1.9749031252029153e-05, "loss": 0.1064, "step": 4333 }, { "epoch": 0.79, "learning_rate": 1.9748865938331794e-05, "loss": 0.1153, "step": 4334 }, { "epoch": 0.79, "learning_rate": 1.974870057089846e-05, "loss": 0.0998, "step": 4335 }, { "epoch": 0.8, "learning_rate": 1.9748535149730056e-05, "loss": 0.1161, "step": 4336 }, { "epoch": 0.8, "learning_rate": 1.9748369674827504e-05, "loss": 0.1013, "step": 4337 }, { "epoch": 0.8, "learning_rate": 1.9748204146191706e-05, "loss": 0.1057, "step": 4338 }, { "epoch": 0.8, "learning_rate": 1.9748038563823577e-05, "loss": 0.1042, "step": 4339 }, { "epoch": 0.8, "learning_rate": 1.9747872927724034e-05, "loss": 0.1194, "step": 4340 }, { "epoch": 0.8, "learning_rate": 1.9747707237893987e-05, "loss": 0.1131, "step": 4341 }, { "epoch": 0.8, "learning_rate": 1.9747541494334346e-05, "loss": 0.1035, "step": 4342 }, { "epoch": 0.8, "learning_rate": 1.9747375697046033e-05, "loss": 0.1013, "step": 4343 }, { "epoch": 0.8, "learning_rate": 1.9747209846029953e-05, "loss": 0.1063, "step": 4344 }, { "epoch": 0.8, "learning_rate": 1.9747043941287026e-05, "loss": 0.1175, "step": 4345 }, { "epoch": 0.8, "learning_rate": 1.9746877982818162e-05, "loss": 0.1148, "step": 4346 }, { "epoch": 0.8, "learning_rate": 1.9746711970624282e-05, "loss": 0.1068, "step": 4347 }, { "epoch": 0.8, "learning_rate": 1.9746545904706294e-05, "loss": 0.101, "step": 4348 }, { "epoch": 0.8, "learning_rate": 1.974637978506512e-05, "loss": 0.1016, "step": 4349 }, { "epoch": 0.8, "learning_rate": 1.9746213611701668e-05, "loss": 0.1092, "step": 4350 }, { "epoch": 0.8, "learning_rate": 1.974604738461686e-05, "loss": 0.1031, "step": 4351 }, { "epoch": 0.8, "learning_rate": 1.974588110381161e-05, "loss": 0.1133, "step": 4352 }, { "epoch": 0.8, "learning_rate": 1.9745714769286834e-05, "loss": 0.1048, "step": 4353 }, { "epoch": 0.8, "learning_rate": 1.9745548381043454e-05, "loss": 0.1213, "step": 4354 }, { "epoch": 0.8, "learning_rate": 1.974538193908238e-05, "loss": 0.1179, "step": 4355 }, { "epoch": 0.8, "learning_rate": 1.9745215443404532e-05, "loss": 0.1161, "step": 4356 }, { "epoch": 0.8, "learning_rate": 1.974504889401083e-05, "loss": 0.1122, "step": 4357 }, { "epoch": 0.8, "learning_rate": 1.974488229090219e-05, "loss": 0.1214, "step": 4358 }, { "epoch": 0.8, "learning_rate": 1.974471563407953e-05, "loss": 0.1154, "step": 4359 }, { "epoch": 0.8, "learning_rate": 1.9744548923543768e-05, "loss": 0.1059, "step": 4360 }, { "epoch": 0.8, "learning_rate": 1.9744382159295826e-05, "loss": 0.1227, "step": 4361 }, { "epoch": 0.8, "learning_rate": 1.974421534133662e-05, "loss": 0.1038, "step": 4362 }, { "epoch": 0.8, "learning_rate": 1.974404846966707e-05, "loss": 0.109, "step": 4363 }, { "epoch": 0.8, "learning_rate": 1.97438815442881e-05, "loss": 0.1176, "step": 4364 }, { "epoch": 0.8, "learning_rate": 1.9743714565200624e-05, "loss": 0.114, "step": 4365 }, { "epoch": 0.8, "learning_rate": 1.9743547532405564e-05, "loss": 0.1074, "step": 4366 }, { "epoch": 0.8, "learning_rate": 1.9743380445903842e-05, "loss": 0.1062, "step": 4367 }, { "epoch": 0.8, "learning_rate": 1.9743213305696383e-05, "loss": 0.1083, "step": 4368 }, { "epoch": 0.8, "learning_rate": 1.97430461117841e-05, "loss": 0.1175, "step": 4369 }, { "epoch": 0.8, "learning_rate": 1.974287886416792e-05, "loss": 0.1125, "step": 4370 }, { "epoch": 0.8, "learning_rate": 1.9742711562848762e-05, "loss": 0.116, "step": 4371 }, { "epoch": 0.8, "learning_rate": 1.9742544207827553e-05, "loss": 0.1068, "step": 4372 }, { "epoch": 0.8, "learning_rate": 1.9742376799105208e-05, "loss": 0.1028, "step": 4373 }, { "epoch": 0.8, "learning_rate": 1.9742209336682654e-05, "loss": 0.0997, "step": 4374 }, { "epoch": 0.8, "learning_rate": 1.9742041820560817e-05, "loss": 0.1123, "step": 4375 }, { "epoch": 0.8, "learning_rate": 1.9741874250740614e-05, "loss": 0.1043, "step": 4376 }, { "epoch": 0.8, "learning_rate": 1.9741706627222975e-05, "loss": 0.1102, "step": 4377 }, { "epoch": 0.8, "learning_rate": 1.9741538950008817e-05, "loss": 0.1066, "step": 4378 }, { "epoch": 0.8, "learning_rate": 1.974137121909907e-05, "loss": 0.1202, "step": 4379 }, { "epoch": 0.8, "learning_rate": 1.974120343449466e-05, "loss": 0.1133, "step": 4380 }, { "epoch": 0.8, "learning_rate": 1.9741035596196502e-05, "loss": 0.1109, "step": 4381 }, { "epoch": 0.8, "learning_rate": 1.9740867704205532e-05, "loss": 0.1239, "step": 4382 }, { "epoch": 0.8, "learning_rate": 1.9740699758522673e-05, "loss": 0.109, "step": 4383 }, { "epoch": 0.8, "learning_rate": 1.9740531759148844e-05, "loss": 0.1006, "step": 4384 }, { "epoch": 0.8, "learning_rate": 1.9740363706084978e-05, "loss": 0.1039, "step": 4385 }, { "epoch": 0.8, "learning_rate": 1.9740195599331996e-05, "loss": 0.1047, "step": 4386 }, { "epoch": 0.8, "learning_rate": 1.9740027438890834e-05, "loss": 0.1106, "step": 4387 }, { "epoch": 0.8, "learning_rate": 1.9739859224762407e-05, "loss": 0.0888, "step": 4388 }, { "epoch": 0.8, "learning_rate": 1.9739690956947652e-05, "loss": 0.1066, "step": 4389 }, { "epoch": 0.81, "learning_rate": 1.9739522635447488e-05, "loss": 0.1092, "step": 4390 }, { "epoch": 0.81, "learning_rate": 1.973935426026285e-05, "loss": 0.1048, "step": 4391 }, { "epoch": 0.81, "learning_rate": 1.973918583139466e-05, "loss": 0.1086, "step": 4392 }, { "epoch": 0.81, "learning_rate": 1.9739017348843853e-05, "loss": 0.1212, "step": 4393 }, { "epoch": 0.81, "learning_rate": 1.9738848812611354e-05, "loss": 0.1073, "step": 4394 }, { "epoch": 0.81, "learning_rate": 1.973868022269809e-05, "loss": 0.1142, "step": 4395 }, { "epoch": 0.81, "learning_rate": 1.9738511579104992e-05, "loss": 0.1104, "step": 4396 }, { "epoch": 0.81, "learning_rate": 1.9738342881832993e-05, "loss": 0.1175, "step": 4397 }, { "epoch": 0.81, "learning_rate": 1.973817413088302e-05, "loss": 0.1022, "step": 4398 }, { "epoch": 0.81, "learning_rate": 1.9738005326256e-05, "loss": 0.1038, "step": 4399 }, { "epoch": 0.81, "learning_rate": 1.973783646795287e-05, "loss": 0.1063, "step": 4400 }, { "epoch": 0.81, "learning_rate": 1.9737667555974553e-05, "loss": 0.1095, "step": 4401 }, { "epoch": 0.81, "learning_rate": 1.9737498590321987e-05, "loss": 0.1172, "step": 4402 }, { "epoch": 0.81, "learning_rate": 1.9737329570996098e-05, "loss": 0.1159, "step": 4403 }, { "epoch": 0.81, "learning_rate": 1.9737160497997824e-05, "loss": 0.1073, "step": 4404 }, { "epoch": 0.81, "learning_rate": 1.973699137132809e-05, "loss": 0.1091, "step": 4405 }, { "epoch": 0.81, "learning_rate": 1.9736822190987834e-05, "loss": 0.1037, "step": 4406 }, { "epoch": 0.81, "learning_rate": 1.9736652956977985e-05, "loss": 0.1151, "step": 4407 }, { "epoch": 0.81, "learning_rate": 1.973648366929948e-05, "loss": 0.0931, "step": 4408 }, { "epoch": 0.81, "learning_rate": 1.9736314327953242e-05, "loss": 0.1038, "step": 4409 }, { "epoch": 0.81, "learning_rate": 1.9736144932940217e-05, "loss": 0.099, "step": 4410 }, { "epoch": 0.81, "learning_rate": 1.973597548426133e-05, "loss": 0.1069, "step": 4411 }, { "epoch": 0.81, "learning_rate": 1.973580598191752e-05, "loss": 0.1067, "step": 4412 }, { "epoch": 0.81, "learning_rate": 1.9735636425909717e-05, "loss": 0.1229, "step": 4413 }, { "epoch": 0.81, "learning_rate": 1.9735466816238863e-05, "loss": 0.1056, "step": 4414 }, { "epoch": 0.81, "learning_rate": 1.9735297152905882e-05, "loss": 0.104, "step": 4415 }, { "epoch": 0.81, "learning_rate": 1.9735127435911717e-05, "loss": 0.1077, "step": 4416 }, { "epoch": 0.81, "learning_rate": 1.9734957665257304e-05, "loss": 0.1073, "step": 4417 }, { "epoch": 0.81, "learning_rate": 1.9734787840943574e-05, "loss": 0.1206, "step": 4418 }, { "epoch": 0.81, "learning_rate": 1.9734617962971465e-05, "loss": 0.1065, "step": 4419 }, { "epoch": 0.81, "learning_rate": 1.9734448031341916e-05, "loss": 0.1018, "step": 4420 }, { "epoch": 0.81, "learning_rate": 1.9734278046055856e-05, "loss": 0.1211, "step": 4421 }, { "epoch": 0.81, "learning_rate": 1.9734108007114233e-05, "loss": 0.1142, "step": 4422 }, { "epoch": 0.81, "learning_rate": 1.9733937914517975e-05, "loss": 0.115, "step": 4423 }, { "epoch": 0.81, "learning_rate": 1.9733767768268025e-05, "loss": 0.1135, "step": 4424 }, { "epoch": 0.81, "learning_rate": 1.973359756836532e-05, "loss": 0.1124, "step": 4425 }, { "epoch": 0.81, "learning_rate": 1.9733427314810796e-05, "loss": 0.1202, "step": 4426 }, { "epoch": 0.81, "learning_rate": 1.9733257007605392e-05, "loss": 0.1158, "step": 4427 }, { "epoch": 0.81, "learning_rate": 1.973308664675005e-05, "loss": 0.1304, "step": 4428 }, { "epoch": 0.81, "learning_rate": 1.9732916232245703e-05, "loss": 0.1145, "step": 4429 }, { "epoch": 0.81, "learning_rate": 1.9732745764093292e-05, "loss": 0.1155, "step": 4430 }, { "epoch": 0.81, "learning_rate": 1.9732575242293764e-05, "loss": 0.1086, "step": 4431 }, { "epoch": 0.81, "learning_rate": 1.9732404666848052e-05, "loss": 0.1189, "step": 4432 }, { "epoch": 0.81, "learning_rate": 1.9732234037757094e-05, "loss": 0.0971, "step": 4433 }, { "epoch": 0.81, "learning_rate": 1.9732063355021838e-05, "loss": 0.124, "step": 4434 }, { "epoch": 0.81, "learning_rate": 1.9731892618643217e-05, "loss": 0.1211, "step": 4435 }, { "epoch": 0.81, "learning_rate": 1.9731721828622178e-05, "loss": 0.1129, "step": 4436 }, { "epoch": 0.81, "learning_rate": 1.973155098495966e-05, "loss": 0.0888, "step": 4437 }, { "epoch": 0.81, "learning_rate": 1.9731380087656605e-05, "loss": 0.1124, "step": 4438 }, { "epoch": 0.81, "learning_rate": 1.9731209136713953e-05, "loss": 0.1048, "step": 4439 }, { "epoch": 0.81, "learning_rate": 1.9731038132132652e-05, "loss": 0.1054, "step": 4440 }, { "epoch": 0.81, "learning_rate": 1.9730867073913637e-05, "loss": 0.1048, "step": 4441 }, { "epoch": 0.81, "learning_rate": 1.973069596205786e-05, "loss": 0.1216, "step": 4442 }, { "epoch": 0.81, "learning_rate": 1.9730524796566256e-05, "loss": 0.1058, "step": 4443 }, { "epoch": 0.81, "learning_rate": 1.973035357743977e-05, "loss": 0.1085, "step": 4444 }, { "epoch": 0.82, "learning_rate": 1.9730182304679347e-05, "loss": 0.1129, "step": 4445 }, { "epoch": 0.82, "learning_rate": 1.9730010978285934e-05, "loss": 0.1018, "step": 4446 }, { "epoch": 0.82, "learning_rate": 1.972983959826047e-05, "loss": 0.115, "step": 4447 }, { "epoch": 0.82, "learning_rate": 1.9729668164603903e-05, "loss": 0.1072, "step": 4448 }, { "epoch": 0.82, "learning_rate": 1.9729496677317178e-05, "loss": 0.1075, "step": 4449 }, { "epoch": 0.82, "learning_rate": 1.972932513640124e-05, "loss": 0.1107, "step": 4450 }, { "epoch": 0.82, "learning_rate": 1.972915354185703e-05, "loss": 0.1315, "step": 4451 }, { "epoch": 0.82, "learning_rate": 1.97289818936855e-05, "loss": 0.109, "step": 4452 }, { "epoch": 0.82, "learning_rate": 1.9728810191887597e-05, "loss": 0.1054, "step": 4453 }, { "epoch": 0.82, "learning_rate": 1.972863843646426e-05, "loss": 0.1249, "step": 4454 }, { "epoch": 0.82, "learning_rate": 1.972846662741644e-05, "loss": 0.0994, "step": 4455 }, { "epoch": 0.82, "learning_rate": 1.9728294764745085e-05, "loss": 0.1042, "step": 4456 }, { "epoch": 0.82, "learning_rate": 1.9728122848451143e-05, "loss": 0.1104, "step": 4457 }, { "epoch": 0.82, "learning_rate": 1.9727950878535557e-05, "loss": 0.1084, "step": 4458 }, { "epoch": 0.82, "learning_rate": 1.9727778854999283e-05, "loss": 0.1175, "step": 4459 }, { "epoch": 0.82, "learning_rate": 1.9727606777843258e-05, "loss": 0.115, "step": 4460 }, { "epoch": 0.82, "learning_rate": 1.972743464706844e-05, "loss": 0.1006, "step": 4461 }, { "epoch": 0.82, "learning_rate": 1.9727262462675775e-05, "loss": 0.1077, "step": 4462 }, { "epoch": 0.82, "learning_rate": 1.9727090224666208e-05, "loss": 0.1164, "step": 4463 }, { "epoch": 0.82, "learning_rate": 1.9726917933040695e-05, "loss": 0.0978, "step": 4464 }, { "epoch": 0.82, "learning_rate": 1.972674558780018e-05, "loss": 0.1208, "step": 4465 }, { "epoch": 0.82, "learning_rate": 1.9726573188945618e-05, "loss": 0.0966, "step": 4466 }, { "epoch": 0.82, "learning_rate": 1.9726400736477954e-05, "loss": 0.1035, "step": 4467 }, { "epoch": 0.82, "learning_rate": 1.9726228230398146e-05, "loss": 0.1084, "step": 4468 }, { "epoch": 0.82, "learning_rate": 1.9726055670707133e-05, "loss": 0.1024, "step": 4469 }, { "epoch": 0.82, "learning_rate": 1.972588305740588e-05, "loss": 0.1101, "step": 4470 }, { "epoch": 0.82, "learning_rate": 1.972571039049533e-05, "loss": 0.1, "step": 4471 }, { "epoch": 0.82, "learning_rate": 1.9725537669976435e-05, "loss": 0.112, "step": 4472 }, { "epoch": 0.82, "learning_rate": 1.9725364895850147e-05, "loss": 0.1153, "step": 4473 }, { "epoch": 0.82, "learning_rate": 1.972519206811742e-05, "loss": 0.0973, "step": 4474 }, { "epoch": 0.82, "learning_rate": 1.9725019186779212e-05, "loss": 0.0908, "step": 4475 }, { "epoch": 0.82, "learning_rate": 1.9724846251836465e-05, "loss": 0.1091, "step": 4476 }, { "epoch": 0.82, "learning_rate": 1.972467326329014e-05, "loss": 0.0906, "step": 4477 }, { "epoch": 0.82, "learning_rate": 1.9724500221141184e-05, "loss": 0.1093, "step": 4478 }, { "epoch": 0.82, "learning_rate": 1.972432712539056e-05, "loss": 0.1077, "step": 4479 }, { "epoch": 0.82, "learning_rate": 1.9724153976039218e-05, "loss": 0.0999, "step": 4480 }, { "epoch": 0.82, "learning_rate": 1.9723980773088108e-05, "loss": 0.1122, "step": 4481 }, { "epoch": 0.82, "learning_rate": 1.9723807516538186e-05, "loss": 0.106, "step": 4482 }, { "epoch": 0.82, "learning_rate": 1.9723634206390415e-05, "loss": 0.1157, "step": 4483 }, { "epoch": 0.82, "learning_rate": 1.972346084264574e-05, "loss": 0.1181, "step": 4484 }, { "epoch": 0.82, "learning_rate": 1.9723287425305125e-05, "loss": 0.1099, "step": 4485 }, { "epoch": 0.82, "learning_rate": 1.9723113954369518e-05, "loss": 0.1003, "step": 4486 }, { "epoch": 0.82, "learning_rate": 1.972294042983988e-05, "loss": 0.1056, "step": 4487 }, { "epoch": 0.82, "learning_rate": 1.972276685171717e-05, "loss": 0.0967, "step": 4488 }, { "epoch": 0.82, "learning_rate": 1.9722593220002338e-05, "loss": 0.1036, "step": 4489 }, { "epoch": 0.82, "learning_rate": 1.9722419534696346e-05, "loss": 0.1058, "step": 4490 }, { "epoch": 0.82, "learning_rate": 1.9722245795800145e-05, "loss": 0.0956, "step": 4491 }, { "epoch": 0.82, "learning_rate": 1.9722072003314703e-05, "loss": 0.1123, "step": 4492 }, { "epoch": 0.82, "learning_rate": 1.972189815724097e-05, "loss": 0.1101, "step": 4493 }, { "epoch": 0.82, "learning_rate": 1.9721724257579907e-05, "loss": 0.111, "step": 4494 }, { "epoch": 0.82, "learning_rate": 1.9721550304332474e-05, "loss": 0.1086, "step": 4495 }, { "epoch": 0.82, "learning_rate": 1.9721376297499626e-05, "loss": 0.1014, "step": 4496 }, { "epoch": 0.82, "learning_rate": 1.9721202237082325e-05, "loss": 0.1079, "step": 4497 }, { "epoch": 0.82, "learning_rate": 1.972102812308153e-05, "loss": 0.1082, "step": 4498 }, { "epoch": 0.83, "learning_rate": 1.9720853955498197e-05, "loss": 0.1104, "step": 4499 }, { "epoch": 0.83, "learning_rate": 1.972067973433329e-05, "loss": 0.1077, "step": 4500 }, { "epoch": 0.83, "learning_rate": 1.972050545958777e-05, "loss": 0.1146, "step": 4501 }, { "epoch": 0.83, "learning_rate": 1.9720331131262598e-05, "loss": 0.1123, "step": 4502 }, { "epoch": 0.83, "learning_rate": 1.972015674935873e-05, "loss": 0.1116, "step": 4503 }, { "epoch": 0.83, "learning_rate": 1.9719982313877134e-05, "loss": 0.1195, "step": 4504 }, { "epoch": 0.83, "learning_rate": 1.9719807824818762e-05, "loss": 0.0993, "step": 4505 }, { "epoch": 0.83, "learning_rate": 1.9719633282184585e-05, "loss": 0.1104, "step": 4506 }, { "epoch": 0.83, "learning_rate": 1.971945868597556e-05, "loss": 0.1006, "step": 4507 }, { "epoch": 0.83, "learning_rate": 1.9719284036192653e-05, "loss": 0.106, "step": 4508 }, { "epoch": 0.83, "learning_rate": 1.9719109332836826e-05, "loss": 0.0999, "step": 4509 }, { "epoch": 0.83, "learning_rate": 1.971893457590904e-05, "loss": 0.1115, "step": 4510 }, { "epoch": 0.83, "learning_rate": 1.9718759765410257e-05, "loss": 0.1106, "step": 4511 }, { "epoch": 0.83, "learning_rate": 1.971858490134144e-05, "loss": 0.1141, "step": 4512 }, { "epoch": 0.83, "learning_rate": 1.9718409983703562e-05, "loss": 0.1086, "step": 4513 }, { "epoch": 0.83, "learning_rate": 1.9718235012497574e-05, "loss": 0.1203, "step": 4514 }, { "epoch": 0.83, "learning_rate": 1.971805998772445e-05, "loss": 0.1183, "step": 4515 }, { "epoch": 0.83, "learning_rate": 1.9717884909385156e-05, "loss": 0.1013, "step": 4516 }, { "epoch": 0.83, "learning_rate": 1.9717709777480647e-05, "loss": 0.105, "step": 4517 }, { "epoch": 0.83, "learning_rate": 1.9717534592011895e-05, "loss": 0.1095, "step": 4518 }, { "epoch": 0.83, "learning_rate": 1.9717359352979866e-05, "loss": 0.1113, "step": 4519 }, { "epoch": 0.83, "learning_rate": 1.9717184060385522e-05, "loss": 0.115, "step": 4520 }, { "epoch": 0.83, "learning_rate": 1.9717008714229836e-05, "loss": 0.1084, "step": 4521 }, { "epoch": 0.83, "learning_rate": 1.9716833314513766e-05, "loss": 0.1094, "step": 4522 }, { "epoch": 0.83, "learning_rate": 1.9716657861238286e-05, "loss": 0.1039, "step": 4523 }, { "epoch": 0.83, "learning_rate": 1.971648235440436e-05, "loss": 0.1102, "step": 4524 }, { "epoch": 0.83, "learning_rate": 1.9716306794012953e-05, "loss": 0.1047, "step": 4525 }, { "epoch": 0.83, "learning_rate": 1.971613118006504e-05, "loss": 0.1192, "step": 4526 }, { "epoch": 0.83, "learning_rate": 1.971595551256158e-05, "loss": 0.1073, "step": 4527 }, { "epoch": 0.83, "learning_rate": 1.9715779791503545e-05, "loss": 0.1131, "step": 4528 }, { "epoch": 0.83, "learning_rate": 1.9715604016891906e-05, "loss": 0.1118, "step": 4529 }, { "epoch": 0.83, "learning_rate": 1.9715428188727632e-05, "loss": 0.0938, "step": 4530 }, { "epoch": 0.83, "learning_rate": 1.971525230701169e-05, "loss": 0.1079, "step": 4531 }, { "epoch": 0.83, "learning_rate": 1.971507637174505e-05, "loss": 0.1086, "step": 4532 }, { "epoch": 0.83, "learning_rate": 1.9714900382928674e-05, "loss": 0.1056, "step": 4533 }, { "epoch": 0.83, "learning_rate": 1.9714724340563547e-05, "loss": 0.1021, "step": 4534 }, { "epoch": 0.83, "learning_rate": 1.971454824465063e-05, "loss": 0.104, "step": 4535 }, { "epoch": 0.83, "learning_rate": 1.9714372095190896e-05, "loss": 0.1164, "step": 4536 }, { "epoch": 0.83, "learning_rate": 1.9714195892185314e-05, "loss": 0.1086, "step": 4537 }, { "epoch": 0.83, "learning_rate": 1.971401963563486e-05, "loss": 0.1072, "step": 4538 }, { "epoch": 0.83, "learning_rate": 1.97138433255405e-05, "loss": 0.1158, "step": 4539 }, { "epoch": 0.83, "learning_rate": 1.9713666961903203e-05, "loss": 0.1004, "step": 4540 }, { "epoch": 0.83, "learning_rate": 1.9713490544723954e-05, "loss": 0.1162, "step": 4541 }, { "epoch": 0.83, "learning_rate": 1.9713314074003715e-05, "loss": 0.1174, "step": 4542 }, { "epoch": 0.83, "learning_rate": 1.971313754974346e-05, "loss": 0.1075, "step": 4543 }, { "epoch": 0.83, "learning_rate": 1.9712960971944164e-05, "loss": 0.1021, "step": 4544 }, { "epoch": 0.83, "learning_rate": 1.97127843406068e-05, "loss": 0.1029, "step": 4545 }, { "epoch": 0.83, "learning_rate": 1.971260765573234e-05, "loss": 0.1143, "step": 4546 }, { "epoch": 0.83, "learning_rate": 1.971243091732176e-05, "loss": 0.1015, "step": 4547 }, { "epoch": 0.83, "learning_rate": 1.9712254125376033e-05, "loss": 0.1174, "step": 4548 }, { "epoch": 0.83, "learning_rate": 1.9712077279896133e-05, "loss": 0.1019, "step": 4549 }, { "epoch": 0.83, "learning_rate": 1.971190038088304e-05, "loss": 0.1076, "step": 4550 }, { "epoch": 0.83, "learning_rate": 1.9711723428337717e-05, "loss": 0.0989, "step": 4551 }, { "epoch": 0.83, "learning_rate": 1.971154642226115e-05, "loss": 0.1088, "step": 4552 }, { "epoch": 0.83, "learning_rate": 1.9711369362654315e-05, "loss": 0.1137, "step": 4553 }, { "epoch": 0.84, "learning_rate": 1.971119224951818e-05, "loss": 0.1053, "step": 4554 }, { "epoch": 0.84, "learning_rate": 1.971101508285373e-05, "loss": 0.111, "step": 4555 }, { "epoch": 0.84, "learning_rate": 1.9710837862661934e-05, "loss": 0.1065, "step": 4556 }, { "epoch": 0.84, "learning_rate": 1.971066058894377e-05, "loss": 0.1068, "step": 4557 }, { "epoch": 0.84, "learning_rate": 1.9710483261700225e-05, "loss": 0.1004, "step": 4558 }, { "epoch": 0.84, "learning_rate": 1.971030588093226e-05, "loss": 0.1089, "step": 4559 }, { "epoch": 0.84, "learning_rate": 1.971012844664087e-05, "loss": 0.1193, "step": 4560 }, { "epoch": 0.84, "learning_rate": 1.9709950958827018e-05, "loss": 0.1179, "step": 4561 }, { "epoch": 0.84, "learning_rate": 1.9709773417491693e-05, "loss": 0.103, "step": 4562 }, { "epoch": 0.84, "learning_rate": 1.9709595822635866e-05, "loss": 0.1033, "step": 4563 }, { "epoch": 0.84, "learning_rate": 1.9709418174260523e-05, "loss": 0.1135, "step": 4564 }, { "epoch": 0.84, "learning_rate": 1.9709240472366636e-05, "loss": 0.1098, "step": 4565 }, { "epoch": 0.84, "learning_rate": 1.970906271695519e-05, "loss": 0.1016, "step": 4566 }, { "epoch": 0.84, "learning_rate": 1.9708884908027163e-05, "loss": 0.0951, "step": 4567 }, { "epoch": 0.84, "learning_rate": 1.970870704558353e-05, "loss": 0.1086, "step": 4568 }, { "epoch": 0.84, "learning_rate": 1.9708529129625286e-05, "loss": 0.1034, "step": 4569 }, { "epoch": 0.84, "learning_rate": 1.9708351160153397e-05, "loss": 0.1251, "step": 4570 }, { "epoch": 0.84, "learning_rate": 1.970817313716885e-05, "loss": 0.1058, "step": 4571 }, { "epoch": 0.84, "learning_rate": 1.970799506067262e-05, "loss": 0.1175, "step": 4572 }, { "epoch": 0.84, "learning_rate": 1.97078169306657e-05, "loss": 0.0961, "step": 4573 }, { "epoch": 0.84, "learning_rate": 1.9707638747149062e-05, "loss": 0.1054, "step": 4574 }, { "epoch": 0.84, "learning_rate": 1.9707460510123693e-05, "loss": 0.1053, "step": 4575 }, { "epoch": 0.84, "learning_rate": 1.9707282219590575e-05, "loss": 0.1102, "step": 4576 }, { "epoch": 0.84, "learning_rate": 1.970710387555069e-05, "loss": 0.103, "step": 4577 }, { "epoch": 0.84, "learning_rate": 1.970692547800502e-05, "loss": 0.0944, "step": 4578 }, { "epoch": 0.84, "learning_rate": 1.970674702695455e-05, "loss": 0.0991, "step": 4579 }, { "epoch": 0.84, "learning_rate": 1.9706568522400263e-05, "loss": 0.1113, "step": 4580 }, { "epoch": 0.84, "learning_rate": 1.9706389964343143e-05, "loss": 0.1083, "step": 4581 }, { "epoch": 0.84, "learning_rate": 1.9706211352784172e-05, "loss": 0.1131, "step": 4582 }, { "epoch": 0.84, "learning_rate": 1.970603268772434e-05, "loss": 0.1148, "step": 4583 }, { "epoch": 0.84, "learning_rate": 1.9705853969164627e-05, "loss": 0.1197, "step": 4584 }, { "epoch": 0.84, "learning_rate": 1.970567519710602e-05, "loss": 0.114, "step": 4585 }, { "epoch": 0.84, "learning_rate": 1.9705496371549503e-05, "loss": 0.1003, "step": 4586 }, { "epoch": 0.84, "learning_rate": 1.9705317492496064e-05, "loss": 0.1079, "step": 4587 }, { "epoch": 0.84, "learning_rate": 1.9705138559946687e-05, "loss": 0.1138, "step": 4588 }, { "epoch": 0.84, "learning_rate": 1.9704959573902355e-05, "loss": 0.1207, "step": 4589 }, { "epoch": 0.84, "learning_rate": 1.9704780534364064e-05, "loss": 0.1208, "step": 4590 }, { "epoch": 0.84, "learning_rate": 1.9704601441332792e-05, "loss": 0.1069, "step": 4591 }, { "epoch": 0.84, "learning_rate": 1.970442229480953e-05, "loss": 0.1002, "step": 4592 }, { "epoch": 0.84, "learning_rate": 1.9704243094795267e-05, "loss": 0.1035, "step": 4593 }, { "epoch": 0.84, "learning_rate": 1.9704063841290984e-05, "loss": 0.0867, "step": 4594 }, { "epoch": 0.84, "learning_rate": 1.970388453429768e-05, "loss": 0.1101, "step": 4595 }, { "epoch": 0.84, "learning_rate": 1.9703705173816334e-05, "loss": 0.1187, "step": 4596 }, { "epoch": 0.84, "learning_rate": 1.9703525759847938e-05, "loss": 0.1087, "step": 4597 }, { "epoch": 0.84, "learning_rate": 1.970334629239348e-05, "loss": 0.107, "step": 4598 }, { "epoch": 0.84, "learning_rate": 1.970316677145395e-05, "loss": 0.1082, "step": 4599 }, { "epoch": 0.84, "learning_rate": 1.9702987197030338e-05, "loss": 0.106, "step": 4600 }, { "epoch": 0.84, "learning_rate": 1.9702807569123636e-05, "loss": 0.1107, "step": 4601 }, { "epoch": 0.84, "learning_rate": 1.970262788773483e-05, "loss": 0.1084, "step": 4602 }, { "epoch": 0.84, "learning_rate": 1.970244815286491e-05, "loss": 0.1224, "step": 4603 }, { "epoch": 0.84, "learning_rate": 1.9702268364514868e-05, "loss": 0.1071, "step": 4604 }, { "epoch": 0.84, "learning_rate": 1.9702088522685695e-05, "loss": 0.1048, "step": 4605 }, { "epoch": 0.84, "learning_rate": 1.9701908627378385e-05, "loss": 0.1092, "step": 4606 }, { "epoch": 0.84, "learning_rate": 1.970172867859393e-05, "loss": 0.1104, "step": 4607 }, { "epoch": 0.85, "learning_rate": 1.9701548676333314e-05, "loss": 0.0959, "step": 4608 }, { "epoch": 0.85, "learning_rate": 1.9701368620597537e-05, "loss": 0.1097, "step": 4609 }, { "epoch": 0.85, "learning_rate": 1.9701188511387584e-05, "loss": 0.119, "step": 4610 }, { "epoch": 0.85, "learning_rate": 1.970100834870446e-05, "loss": 0.1149, "step": 4611 }, { "epoch": 0.85, "learning_rate": 1.9700828132549146e-05, "loss": 0.111, "step": 4612 }, { "epoch": 0.85, "learning_rate": 1.9700647862922643e-05, "loss": 0.0993, "step": 4613 }, { "epoch": 0.85, "learning_rate": 1.9700467539825937e-05, "loss": 0.1078, "step": 4614 }, { "epoch": 0.85, "learning_rate": 1.9700287163260033e-05, "loss": 0.1077, "step": 4615 }, { "epoch": 0.85, "learning_rate": 1.9700106733225913e-05, "loss": 0.1193, "step": 4616 }, { "epoch": 0.85, "learning_rate": 1.969992624972458e-05, "loss": 0.1114, "step": 4617 }, { "epoch": 0.85, "learning_rate": 1.969974571275703e-05, "loss": 0.1182, "step": 4618 }, { "epoch": 0.85, "learning_rate": 1.9699565122324247e-05, "loss": 0.0997, "step": 4619 }, { "epoch": 0.85, "learning_rate": 1.969938447842724e-05, "loss": 0.1167, "step": 4620 }, { "epoch": 0.85, "learning_rate": 1.9699203781066995e-05, "loss": 0.1133, "step": 4621 }, { "epoch": 0.85, "learning_rate": 1.9699023030244513e-05, "loss": 0.117, "step": 4622 }, { "epoch": 0.85, "learning_rate": 1.969884222596079e-05, "loss": 0.1141, "step": 4623 }, { "epoch": 0.85, "learning_rate": 1.9698661368216818e-05, "loss": 0.1038, "step": 4624 }, { "epoch": 0.85, "learning_rate": 1.96984804570136e-05, "loss": 0.1187, "step": 4625 }, { "epoch": 0.85, "learning_rate": 1.9698299492352132e-05, "loss": 0.1153, "step": 4626 }, { "epoch": 0.85, "learning_rate": 1.9698118474233404e-05, "loss": 0.099, "step": 4627 }, { "epoch": 0.85, "learning_rate": 1.9697937402658427e-05, "loss": 0.1102, "step": 4628 }, { "epoch": 0.85, "learning_rate": 1.9697756277628187e-05, "loss": 0.1092, "step": 4629 }, { "epoch": 0.85, "learning_rate": 1.969757509914369e-05, "loss": 0.1014, "step": 4630 }, { "epoch": 0.85, "learning_rate": 1.969739386720593e-05, "loss": 0.1046, "step": 4631 }, { "epoch": 0.85, "learning_rate": 1.9697212581815908e-05, "loss": 0.1049, "step": 4632 }, { "epoch": 0.85, "learning_rate": 1.9697031242974622e-05, "loss": 0.1108, "step": 4633 }, { "epoch": 0.85, "learning_rate": 1.9696849850683075e-05, "loss": 0.1041, "step": 4634 }, { "epoch": 0.85, "learning_rate": 1.9696668404942264e-05, "loss": 0.0974, "step": 4635 }, { "epoch": 0.85, "learning_rate": 1.969648690575319e-05, "loss": 0.1155, "step": 4636 }, { "epoch": 0.85, "learning_rate": 1.969630535311685e-05, "loss": 0.1041, "step": 4637 }, { "epoch": 0.85, "learning_rate": 1.969612374703425e-05, "loss": 0.1202, "step": 4638 }, { "epoch": 0.85, "learning_rate": 1.969594208750639e-05, "loss": 0.1066, "step": 4639 }, { "epoch": 0.85, "learning_rate": 1.969576037453427e-05, "loss": 0.1209, "step": 4640 }, { "epoch": 0.85, "learning_rate": 1.969557860811889e-05, "loss": 0.1251, "step": 4641 }, { "epoch": 0.85, "learning_rate": 1.969539678826125e-05, "loss": 0.1058, "step": 4642 }, { "epoch": 0.85, "learning_rate": 1.9695214914962363e-05, "loss": 0.1092, "step": 4643 }, { "epoch": 0.85, "learning_rate": 1.969503298822322e-05, "loss": 0.1123, "step": 4644 }, { "epoch": 0.85, "learning_rate": 1.9694851008044827e-05, "loss": 0.1091, "step": 4645 }, { "epoch": 0.85, "learning_rate": 1.9694668974428192e-05, "loss": 0.1039, "step": 4646 }, { "epoch": 0.85, "learning_rate": 1.9694486887374312e-05, "loss": 0.1155, "step": 4647 }, { "epoch": 0.85, "learning_rate": 1.9694304746884193e-05, "loss": 0.0893, "step": 4648 }, { "epoch": 0.85, "learning_rate": 1.969412255295884e-05, "loss": 0.1219, "step": 4649 }, { "epoch": 0.85, "learning_rate": 1.9693940305599256e-05, "loss": 0.1062, "step": 4650 }, { "epoch": 0.85, "learning_rate": 1.9693758004806447e-05, "loss": 0.0985, "step": 4651 }, { "epoch": 0.85, "learning_rate": 1.9693575650581415e-05, "loss": 0.1085, "step": 4652 }, { "epoch": 0.85, "learning_rate": 1.9693393242925168e-05, "loss": 0.1113, "step": 4653 }, { "epoch": 0.85, "learning_rate": 1.969321078183871e-05, "loss": 0.096, "step": 4654 }, { "epoch": 0.85, "learning_rate": 1.969302826732305e-05, "loss": 0.1177, "step": 4655 }, { "epoch": 0.85, "learning_rate": 1.9692845699379186e-05, "loss": 0.1094, "step": 4656 }, { "epoch": 0.85, "learning_rate": 1.969266307800813e-05, "loss": 0.1143, "step": 4657 }, { "epoch": 0.85, "learning_rate": 1.9692480403210892e-05, "loss": 0.1045, "step": 4658 }, { "epoch": 0.85, "learning_rate": 1.9692297674988473e-05, "loss": 0.1164, "step": 4659 }, { "epoch": 0.85, "learning_rate": 1.969211489334188e-05, "loss": 0.0922, "step": 4660 }, { "epoch": 0.85, "learning_rate": 1.9691932058272124e-05, "loss": 0.1072, "step": 4661 }, { "epoch": 0.85, "learning_rate": 1.9691749169780213e-05, "loss": 0.1077, "step": 4662 }, { "epoch": 0.86, "learning_rate": 1.9691566227867153e-05, "loss": 0.1073, "step": 4663 }, { "epoch": 0.86, "learning_rate": 1.9691383232533955e-05, "loss": 0.1046, "step": 4664 }, { "epoch": 0.86, "learning_rate": 1.969120018378162e-05, "loss": 0.1211, "step": 4665 }, { "epoch": 0.86, "learning_rate": 1.969101708161117e-05, "loss": 0.1058, "step": 4666 }, { "epoch": 0.86, "learning_rate": 1.96908339260236e-05, "loss": 0.1054, "step": 4667 }, { "epoch": 0.86, "learning_rate": 1.9690650717019927e-05, "loss": 0.1069, "step": 4668 }, { "epoch": 0.86, "learning_rate": 1.969046745460116e-05, "loss": 0.1135, "step": 4669 }, { "epoch": 0.86, "learning_rate": 1.9690284138768313e-05, "loss": 0.1028, "step": 4670 }, { "epoch": 0.86, "learning_rate": 1.9690100769522392e-05, "loss": 0.0951, "step": 4671 }, { "epoch": 0.86, "learning_rate": 1.9689917346864406e-05, "loss": 0.1094, "step": 4672 }, { "epoch": 0.86, "learning_rate": 1.968973387079537e-05, "loss": 0.1044, "step": 4673 }, { "epoch": 0.86, "learning_rate": 1.9689550341316295e-05, "loss": 0.1082, "step": 4674 }, { "epoch": 0.86, "learning_rate": 1.968936675842819e-05, "loss": 0.1167, "step": 4675 }, { "epoch": 0.86, "learning_rate": 1.968918312213207e-05, "loss": 0.1086, "step": 4676 }, { "epoch": 0.86, "learning_rate": 1.9688999432428945e-05, "loss": 0.1048, "step": 4677 }, { "epoch": 0.86, "learning_rate": 1.9688815689319825e-05, "loss": 0.1079, "step": 4678 }, { "epoch": 0.86, "learning_rate": 1.968863189280573e-05, "loss": 0.112, "step": 4679 }, { "epoch": 0.86, "learning_rate": 1.968844804288767e-05, "loss": 0.1057, "step": 4680 }, { "epoch": 0.86, "learning_rate": 1.9688264139566656e-05, "loss": 0.1041, "step": 4681 }, { "epoch": 0.86, "learning_rate": 1.96880801828437e-05, "loss": 0.1044, "step": 4682 }, { "epoch": 0.86, "learning_rate": 1.9687896172719824e-05, "loss": 0.0911, "step": 4683 }, { "epoch": 0.86, "learning_rate": 1.9687712109196032e-05, "loss": 0.1085, "step": 4684 }, { "epoch": 0.86, "learning_rate": 1.9687527992273346e-05, "loss": 0.1052, "step": 4685 }, { "epoch": 0.86, "learning_rate": 1.968734382195278e-05, "loss": 0.1036, "step": 4686 }, { "epoch": 0.86, "learning_rate": 1.9687159598235347e-05, "loss": 0.0924, "step": 4687 }, { "epoch": 0.86, "learning_rate": 1.9686975321122067e-05, "loss": 0.1094, "step": 4688 }, { "epoch": 0.86, "learning_rate": 1.9686790990613948e-05, "loss": 0.124, "step": 4689 }, { "epoch": 0.86, "learning_rate": 1.968660660671201e-05, "loss": 0.1126, "step": 4690 }, { "epoch": 0.86, "learning_rate": 1.968642216941727e-05, "loss": 0.1231, "step": 4691 }, { "epoch": 0.86, "learning_rate": 1.9686237678730746e-05, "loss": 0.114, "step": 4692 }, { "epoch": 0.86, "learning_rate": 1.968605313465345e-05, "loss": 0.1024, "step": 4693 }, { "epoch": 0.86, "learning_rate": 1.9685868537186404e-05, "loss": 0.1072, "step": 4694 }, { "epoch": 0.86, "learning_rate": 1.968568388633062e-05, "loss": 0.108, "step": 4695 }, { "epoch": 0.86, "learning_rate": 1.9685499182087124e-05, "loss": 0.1056, "step": 4696 }, { "epoch": 0.86, "learning_rate": 1.968531442445693e-05, "loss": 0.0983, "step": 4697 }, { "epoch": 0.86, "learning_rate": 1.9685129613441056e-05, "loss": 0.0973, "step": 4698 }, { "epoch": 0.86, "learning_rate": 1.9684944749040514e-05, "loss": 0.1102, "step": 4699 }, { "epoch": 0.86, "learning_rate": 1.9684759831256338e-05, "loss": 0.1124, "step": 4700 }, { "epoch": 0.86, "learning_rate": 1.9684574860089537e-05, "loss": 0.1128, "step": 4701 }, { "epoch": 0.86, "learning_rate": 1.968438983554113e-05, "loss": 0.1011, "step": 4702 }, { "epoch": 0.86, "learning_rate": 1.9684204757612138e-05, "loss": 0.1075, "step": 4703 }, { "epoch": 0.86, "learning_rate": 1.9684019626303587e-05, "loss": 0.1099, "step": 4704 }, { "epoch": 0.86, "learning_rate": 1.968383444161649e-05, "loss": 0.1101, "step": 4705 }, { "epoch": 0.86, "learning_rate": 1.968364920355187e-05, "loss": 0.1022, "step": 4706 }, { "epoch": 0.86, "learning_rate": 1.968346391211075e-05, "loss": 0.1024, "step": 4707 }, { "epoch": 0.86, "learning_rate": 1.968327856729415e-05, "loss": 0.1158, "step": 4708 }, { "epoch": 0.86, "learning_rate": 1.9683093169103093e-05, "loss": 0.0906, "step": 4709 }, { "epoch": 0.86, "learning_rate": 1.96829077175386e-05, "loss": 0.1105, "step": 4710 }, { "epoch": 0.86, "learning_rate": 1.9682722212601693e-05, "loss": 0.1034, "step": 4711 }, { "epoch": 0.86, "learning_rate": 1.968253665429339e-05, "loss": 0.1173, "step": 4712 }, { "epoch": 0.86, "learning_rate": 1.968235104261472e-05, "loss": 0.118, "step": 4713 }, { "epoch": 0.86, "learning_rate": 1.9682165377566705e-05, "loss": 0.1135, "step": 4714 }, { "epoch": 0.86, "learning_rate": 1.968197965915037e-05, "loss": 0.1094, "step": 4715 }, { "epoch": 0.86, "learning_rate": 1.9681793887366734e-05, "loss": 0.1232, "step": 4716 }, { "epoch": 0.87, "learning_rate": 1.9681608062216823e-05, "loss": 0.1002, "step": 4717 }, { "epoch": 0.87, "learning_rate": 1.968142218370166e-05, "loss": 0.1213, "step": 4718 }, { "epoch": 0.87, "learning_rate": 1.9681236251822274e-05, "loss": 0.099, "step": 4719 }, { "epoch": 0.87, "learning_rate": 1.9681050266579687e-05, "loss": 0.1132, "step": 4720 }, { "epoch": 0.87, "learning_rate": 1.9680864227974925e-05, "loss": 0.1149, "step": 4721 }, { "epoch": 0.87, "learning_rate": 1.9680678136009008e-05, "loss": 0.111, "step": 4722 }, { "epoch": 0.87, "learning_rate": 1.9680491990682967e-05, "loss": 0.1148, "step": 4723 }, { "epoch": 0.87, "learning_rate": 1.968030579199783e-05, "loss": 0.11, "step": 4724 }, { "epoch": 0.87, "learning_rate": 1.968011953995462e-05, "loss": 0.1219, "step": 4725 }, { "epoch": 0.87, "learning_rate": 1.9679933234554365e-05, "loss": 0.11, "step": 4726 }, { "epoch": 0.87, "learning_rate": 1.967974687579809e-05, "loss": 0.1096, "step": 4727 }, { "epoch": 0.87, "learning_rate": 1.9679560463686826e-05, "loss": 0.1021, "step": 4728 }, { "epoch": 0.87, "learning_rate": 1.9679373998221596e-05, "loss": 0.1104, "step": 4729 }, { "epoch": 0.87, "learning_rate": 1.9679187479403427e-05, "loss": 0.0908, "step": 4730 }, { "epoch": 0.87, "learning_rate": 1.9679000907233353e-05, "loss": 0.1055, "step": 4731 }, { "epoch": 0.87, "learning_rate": 1.9678814281712398e-05, "loss": 0.1164, "step": 4732 }, { "epoch": 0.87, "learning_rate": 1.9678627602841594e-05, "loss": 0.1047, "step": 4733 }, { "epoch": 0.87, "learning_rate": 1.9678440870621965e-05, "loss": 0.1099, "step": 4734 }, { "epoch": 0.87, "learning_rate": 1.9678254085054548e-05, "loss": 0.1089, "step": 4735 }, { "epoch": 0.87, "learning_rate": 1.9678067246140365e-05, "loss": 0.1125, "step": 4736 }, { "epoch": 0.87, "learning_rate": 1.9677880353880447e-05, "loss": 0.1025, "step": 4737 }, { "epoch": 0.87, "learning_rate": 1.9677693408275827e-05, "loss": 0.1097, "step": 4738 }, { "epoch": 0.87, "learning_rate": 1.9677506409327532e-05, "loss": 0.0954, "step": 4739 }, { "epoch": 0.87, "learning_rate": 1.96773193570366e-05, "loss": 0.0907, "step": 4740 }, { "epoch": 0.87, "learning_rate": 1.9677132251404054e-05, "loss": 0.1191, "step": 4741 }, { "epoch": 0.87, "learning_rate": 1.9676945092430926e-05, "loss": 0.1063, "step": 4742 }, { "epoch": 0.87, "learning_rate": 1.9676757880118254e-05, "loss": 0.1024, "step": 4743 }, { "epoch": 0.87, "learning_rate": 1.9676570614467063e-05, "loss": 0.1146, "step": 4744 }, { "epoch": 0.87, "learning_rate": 1.967638329547839e-05, "loss": 0.11, "step": 4745 }, { "epoch": 0.87, "learning_rate": 1.9676195923153263e-05, "loss": 0.1088, "step": 4746 }, { "epoch": 0.87, "learning_rate": 1.967600849749272e-05, "loss": 0.1003, "step": 4747 }, { "epoch": 0.87, "learning_rate": 1.9675821018497793e-05, "loss": 0.1129, "step": 4748 }, { "epoch": 0.87, "learning_rate": 1.967563348616951e-05, "loss": 0.1057, "step": 4749 }, { "epoch": 0.87, "learning_rate": 1.967544590050891e-05, "loss": 0.1156, "step": 4750 }, { "epoch": 0.87, "learning_rate": 1.9675258261517024e-05, "loss": 0.0939, "step": 4751 }, { "epoch": 0.87, "learning_rate": 1.967507056919489e-05, "loss": 0.1227, "step": 4752 }, { "epoch": 0.87, "learning_rate": 1.9674882823543542e-05, "loss": 0.1152, "step": 4753 }, { "epoch": 0.87, "learning_rate": 1.967469502456401e-05, "loss": 0.1161, "step": 4754 }, { "epoch": 0.87, "learning_rate": 1.9674507172257335e-05, "loss": 0.1156, "step": 4755 }, { "epoch": 0.87, "learning_rate": 1.9674319266624546e-05, "loss": 0.1148, "step": 4756 }, { "epoch": 0.87, "learning_rate": 1.9674131307666683e-05, "loss": 0.1217, "step": 4757 }, { "epoch": 0.87, "learning_rate": 1.9673943295384786e-05, "loss": 0.0996, "step": 4758 }, { "epoch": 0.87, "learning_rate": 1.9673755229779884e-05, "loss": 0.1015, "step": 4759 }, { "epoch": 0.87, "learning_rate": 1.9673567110853015e-05, "loss": 0.0959, "step": 4760 }, { "epoch": 0.87, "learning_rate": 1.9673378938605222e-05, "loss": 0.1021, "step": 4761 }, { "epoch": 0.87, "learning_rate": 1.967319071303753e-05, "loss": 0.1278, "step": 4762 }, { "epoch": 0.87, "learning_rate": 1.967300243415099e-05, "loss": 0.1053, "step": 4763 }, { "epoch": 0.87, "learning_rate": 1.9672814101946637e-05, "loss": 0.1175, "step": 4764 }, { "epoch": 0.87, "learning_rate": 1.9672625716425502e-05, "loss": 0.1019, "step": 4765 }, { "epoch": 0.87, "learning_rate": 1.9672437277588624e-05, "loss": 0.1131, "step": 4766 }, { "epoch": 0.87, "learning_rate": 1.9672248785437053e-05, "loss": 0.1211, "step": 4767 }, { "epoch": 0.87, "learning_rate": 1.967206023997181e-05, "loss": 0.1078, "step": 4768 }, { "epoch": 0.87, "learning_rate": 1.9671871641193954e-05, "loss": 0.1085, "step": 4769 }, { "epoch": 0.87, "learning_rate": 1.9671682989104514e-05, "loss": 0.1319, "step": 4770 }, { "epoch": 0.87, "learning_rate": 1.9671494283704527e-05, "loss": 0.1248, "step": 4771 }, { "epoch": 0.88, "learning_rate": 1.967130552499504e-05, "loss": 0.1094, "step": 4772 }, { "epoch": 0.88, "learning_rate": 1.9671116712977084e-05, "loss": 0.1204, "step": 4773 }, { "epoch": 0.88, "learning_rate": 1.9670927847651707e-05, "loss": 0.1255, "step": 4774 }, { "epoch": 0.88, "learning_rate": 1.9670738929019958e-05, "loss": 0.1238, "step": 4775 }, { "epoch": 0.88, "learning_rate": 1.9670549957082864e-05, "loss": 0.1156, "step": 4776 }, { "epoch": 0.88, "learning_rate": 1.9670360931841472e-05, "loss": 0.1376, "step": 4777 }, { "epoch": 0.88, "learning_rate": 1.967017185329682e-05, "loss": 0.1097, "step": 4778 }, { "epoch": 0.88, "learning_rate": 1.966998272144996e-05, "loss": 0.0995, "step": 4779 }, { "epoch": 0.88, "learning_rate": 1.9669793536301926e-05, "loss": 0.1092, "step": 4780 }, { "epoch": 0.88, "learning_rate": 1.9669604297853766e-05, "loss": 0.1053, "step": 4781 }, { "epoch": 0.88, "learning_rate": 1.9669415006106517e-05, "loss": 0.1149, "step": 4782 }, { "epoch": 0.88, "learning_rate": 1.966922566106123e-05, "loss": 0.107, "step": 4783 }, { "epoch": 0.88, "learning_rate": 1.966903626271894e-05, "loss": 0.1025, "step": 4784 }, { "epoch": 0.88, "learning_rate": 1.96688468110807e-05, "loss": 0.1056, "step": 4785 }, { "epoch": 0.88, "learning_rate": 1.9668657306147546e-05, "loss": 0.1188, "step": 4786 }, { "epoch": 0.88, "learning_rate": 1.9668467747920527e-05, "loss": 0.1123, "step": 4787 }, { "epoch": 0.88, "learning_rate": 1.9668278136400693e-05, "loss": 0.108, "step": 4788 }, { "epoch": 0.88, "learning_rate": 1.9668088471589076e-05, "loss": 0.1132, "step": 4789 }, { "epoch": 0.88, "learning_rate": 1.9667898753486733e-05, "loss": 0.104, "step": 4790 }, { "epoch": 0.88, "learning_rate": 1.9667708982094705e-05, "loss": 0.1091, "step": 4791 }, { "epoch": 0.88, "learning_rate": 1.9667519157414036e-05, "loss": 0.1195, "step": 4792 }, { "epoch": 0.88, "learning_rate": 1.9667329279445778e-05, "loss": 0.1014, "step": 4793 }, { "epoch": 0.88, "learning_rate": 1.9667139348190972e-05, "loss": 0.1121, "step": 4794 }, { "epoch": 0.88, "learning_rate": 1.966694936365067e-05, "loss": 0.1001, "step": 4795 }, { "epoch": 0.88, "learning_rate": 1.9666759325825918e-05, "loss": 0.1151, "step": 4796 }, { "epoch": 0.88, "learning_rate": 1.9666569234717757e-05, "loss": 0.1109, "step": 4797 }, { "epoch": 0.88, "learning_rate": 1.9666379090327242e-05, "loss": 0.1987, "step": 4798 }, { "epoch": 0.88, "learning_rate": 1.966618889265542e-05, "loss": 0.1359, "step": 4799 }, { "epoch": 0.88, "learning_rate": 1.966599864170334e-05, "loss": 0.1209, "step": 4800 }, { "epoch": 0.88, "learning_rate": 1.9665808337472047e-05, "loss": 0.132, "step": 4801 }, { "epoch": 0.88, "learning_rate": 1.966561797996259e-05, "loss": 0.122, "step": 4802 }, { "epoch": 0.88, "learning_rate": 1.9665427569176024e-05, "loss": 0.1031, "step": 4803 }, { "epoch": 0.88, "learning_rate": 1.9665237105113394e-05, "loss": 0.1283, "step": 4804 }, { "epoch": 0.88, "learning_rate": 1.966504658777575e-05, "loss": 0.1117, "step": 4805 }, { "epoch": 0.88, "learning_rate": 1.966485601716414e-05, "loss": 0.1189, "step": 4806 }, { "epoch": 0.88, "learning_rate": 1.966466539327962e-05, "loss": 0.1108, "step": 4807 }, { "epoch": 0.88, "learning_rate": 1.9664474716123236e-05, "loss": 0.1077, "step": 4808 }, { "epoch": 0.88, "learning_rate": 1.9664283985696043e-05, "loss": 0.1036, "step": 4809 }, { "epoch": 0.88, "learning_rate": 1.966409320199909e-05, "loss": 0.1231, "step": 4810 }, { "epoch": 0.88, "learning_rate": 1.9663902365033428e-05, "loss": 0.1265, "step": 4811 }, { "epoch": 0.88, "learning_rate": 1.966371147480011e-05, "loss": 0.1175, "step": 4812 }, { "epoch": 0.88, "learning_rate": 1.966352053130019e-05, "loss": 0.1136, "step": 4813 }, { "epoch": 0.88, "learning_rate": 1.966332953453472e-05, "loss": 0.1109, "step": 4814 }, { "epoch": 0.88, "learning_rate": 1.9663138484504745e-05, "loss": 0.0971, "step": 4815 }, { "epoch": 0.88, "learning_rate": 1.9662947381211328e-05, "loss": 0.1249, "step": 4816 }, { "epoch": 0.88, "learning_rate": 1.9662756224655517e-05, "loss": 0.1183, "step": 4817 }, { "epoch": 0.88, "learning_rate": 1.966256501483837e-05, "loss": 0.1148, "step": 4818 }, { "epoch": 0.88, "learning_rate": 1.9662373751760934e-05, "loss": 0.1143, "step": 4819 }, { "epoch": 0.88, "learning_rate": 1.9662182435424273e-05, "loss": 0.1082, "step": 4820 }, { "epoch": 0.88, "learning_rate": 1.966199106582943e-05, "loss": 0.1119, "step": 4821 }, { "epoch": 0.88, "learning_rate": 1.966179964297747e-05, "loss": 0.1044, "step": 4822 }, { "epoch": 0.88, "learning_rate": 1.9661608166869445e-05, "loss": 0.1147, "step": 4823 }, { "epoch": 0.88, "learning_rate": 1.9661416637506407e-05, "loss": 0.1051, "step": 4824 }, { "epoch": 0.88, "learning_rate": 1.9661225054889415e-05, "loss": 0.1093, "step": 4825 }, { "epoch": 0.89, "learning_rate": 1.9661033419019525e-05, "loss": 0.1232, "step": 4826 }, { "epoch": 0.89, "learning_rate": 1.966084172989779e-05, "loss": 0.1088, "step": 4827 }, { "epoch": 0.89, "learning_rate": 1.9660649987525268e-05, "loss": 0.1215, "step": 4828 }, { "epoch": 0.89, "learning_rate": 1.9660458191903023e-05, "loss": 0.1062, "step": 4829 }, { "epoch": 0.89, "learning_rate": 1.96602663430321e-05, "loss": 0.1161, "step": 4830 }, { "epoch": 0.89, "learning_rate": 1.9660074440913565e-05, "loss": 0.1185, "step": 4831 }, { "epoch": 0.89, "learning_rate": 1.9659882485548473e-05, "loss": 0.1189, "step": 4832 }, { "epoch": 0.89, "learning_rate": 1.9659690476937886e-05, "loss": 0.1164, "step": 4833 }, { "epoch": 0.89, "learning_rate": 1.9659498415082853e-05, "loss": 0.1227, "step": 4834 }, { "epoch": 0.89, "learning_rate": 1.9659306299984443e-05, "loss": 0.0995, "step": 4835 }, { "epoch": 0.89, "learning_rate": 1.9659114131643707e-05, "loss": 0.1058, "step": 4836 }, { "epoch": 0.89, "learning_rate": 1.965892191006171e-05, "loss": 0.1087, "step": 4837 }, { "epoch": 0.89, "learning_rate": 1.9658729635239506e-05, "loss": 0.1086, "step": 4838 }, { "epoch": 0.89, "learning_rate": 1.965853730717816e-05, "loss": 0.0938, "step": 4839 }, { "epoch": 0.89, "learning_rate": 1.9658344925878733e-05, "loss": 0.1125, "step": 4840 }, { "epoch": 0.89, "learning_rate": 1.9658152491342277e-05, "loss": 0.1161, "step": 4841 }, { "epoch": 0.89, "learning_rate": 1.9657960003569862e-05, "loss": 0.1175, "step": 4842 }, { "epoch": 0.89, "learning_rate": 1.9657767462562544e-05, "loss": 0.1221, "step": 4843 }, { "epoch": 0.89, "learning_rate": 1.9657574868321388e-05, "loss": 0.1178, "step": 4844 }, { "epoch": 0.89, "learning_rate": 1.965738222084745e-05, "loss": 0.1089, "step": 4845 }, { "epoch": 0.89, "learning_rate": 1.9657189520141796e-05, "loss": 0.1223, "step": 4846 }, { "epoch": 0.89, "learning_rate": 1.9656996766205486e-05, "loss": 0.0928, "step": 4847 }, { "epoch": 0.89, "learning_rate": 1.9656803959039586e-05, "loss": 0.0915, "step": 4848 }, { "epoch": 0.89, "learning_rate": 1.9656611098645153e-05, "loss": 0.1071, "step": 4849 }, { "epoch": 0.89, "learning_rate": 1.9656418185023254e-05, "loss": 0.1047, "step": 4850 }, { "epoch": 0.89, "learning_rate": 1.9656225218174954e-05, "loss": 0.1118, "step": 4851 }, { "epoch": 0.89, "learning_rate": 1.9656032198101315e-05, "loss": 0.1214, "step": 4852 }, { "epoch": 0.89, "learning_rate": 1.9655839124803395e-05, "loss": 0.1038, "step": 4853 }, { "epoch": 0.89, "learning_rate": 1.965564599828227e-05, "loss": 0.1175, "step": 4854 }, { "epoch": 0.89, "learning_rate": 1.9655452818538993e-05, "loss": 0.1079, "step": 4855 }, { "epoch": 0.89, "learning_rate": 1.9655259585574638e-05, "loss": 0.1059, "step": 4856 }, { "epoch": 0.89, "learning_rate": 1.965506629939026e-05, "loss": 0.1193, "step": 4857 }, { "epoch": 0.89, "learning_rate": 1.9654872959986936e-05, "loss": 0.1012, "step": 4858 }, { "epoch": 0.89, "learning_rate": 1.9654679567365727e-05, "loss": 0.1218, "step": 4859 }, { "epoch": 0.89, "learning_rate": 1.9654486121527694e-05, "loss": 0.1029, "step": 4860 }, { "epoch": 0.89, "learning_rate": 1.965429262247391e-05, "loss": 0.1119, "step": 4861 }, { "epoch": 0.89, "learning_rate": 1.9654099070205434e-05, "loss": 0.1052, "step": 4862 }, { "epoch": 0.89, "learning_rate": 1.9653905464723344e-05, "loss": 0.117, "step": 4863 }, { "epoch": 0.89, "learning_rate": 1.9653711806028697e-05, "loss": 0.1103, "step": 4864 }, { "epoch": 0.89, "learning_rate": 1.9653518094122565e-05, "loss": 0.1086, "step": 4865 }, { "epoch": 0.89, "learning_rate": 1.9653324329006017e-05, "loss": 0.0959, "step": 4866 }, { "epoch": 0.89, "learning_rate": 1.965313051068012e-05, "loss": 0.1223, "step": 4867 }, { "epoch": 0.89, "learning_rate": 1.9652936639145937e-05, "loss": 0.0981, "step": 4868 }, { "epoch": 0.89, "learning_rate": 1.9652742714404543e-05, "loss": 0.1038, "step": 4869 }, { "epoch": 0.89, "learning_rate": 1.9652548736457007e-05, "loss": 0.1085, "step": 4870 }, { "epoch": 0.89, "learning_rate": 1.9652354705304395e-05, "loss": 0.1006, "step": 4871 }, { "epoch": 0.89, "learning_rate": 1.9652160620947777e-05, "loss": 0.1147, "step": 4872 }, { "epoch": 0.89, "learning_rate": 1.9651966483388223e-05, "loss": 0.1104, "step": 4873 }, { "epoch": 0.89, "learning_rate": 1.9651772292626804e-05, "loss": 0.1171, "step": 4874 }, { "epoch": 0.89, "learning_rate": 1.965157804866459e-05, "loss": 0.1166, "step": 4875 }, { "epoch": 0.89, "learning_rate": 1.9651383751502654e-05, "loss": 0.098, "step": 4876 }, { "epoch": 0.89, "learning_rate": 1.965118940114206e-05, "loss": 0.1108, "step": 4877 }, { "epoch": 0.89, "learning_rate": 1.965099499758389e-05, "loss": 0.1045, "step": 4878 }, { "epoch": 0.89, "learning_rate": 1.9650800540829204e-05, "loss": 0.1081, "step": 4879 }, { "epoch": 0.89, "learning_rate": 1.9650606030879082e-05, "loss": 0.098, "step": 4880 }, { "epoch": 0.9, "learning_rate": 1.9650411467734593e-05, "loss": 0.1112, "step": 4881 }, { "epoch": 0.9, "learning_rate": 1.965021685139681e-05, "loss": 0.1021, "step": 4882 }, { "epoch": 0.9, "learning_rate": 1.9650022181866807e-05, "loss": 0.1135, "step": 4883 }, { "epoch": 0.9, "learning_rate": 1.9649827459145654e-05, "loss": 0.1126, "step": 4884 }, { "epoch": 0.9, "learning_rate": 1.9649632683234427e-05, "loss": 0.1064, "step": 4885 }, { "epoch": 0.9, "learning_rate": 1.9649437854134197e-05, "loss": 0.1138, "step": 4886 }, { "epoch": 0.9, "learning_rate": 1.964924297184604e-05, "loss": 0.1118, "step": 4887 }, { "epoch": 0.9, "learning_rate": 1.964904803637103e-05, "loss": 0.1086, "step": 4888 }, { "epoch": 0.9, "learning_rate": 1.9648853047710244e-05, "loss": 0.1123, "step": 4889 }, { "epoch": 0.9, "learning_rate": 1.964865800586475e-05, "loss": 0.1067, "step": 4890 }, { "epoch": 0.9, "learning_rate": 1.964846291083563e-05, "loss": 0.108, "step": 4891 }, { "epoch": 0.9, "learning_rate": 1.9648267762623952e-05, "loss": 0.1109, "step": 4892 }, { "epoch": 0.9, "learning_rate": 1.9648072561230796e-05, "loss": 0.1115, "step": 4893 }, { "epoch": 0.9, "learning_rate": 1.9647877306657243e-05, "loss": 0.1232, "step": 4894 }, { "epoch": 0.9, "learning_rate": 1.9647681998904362e-05, "loss": 0.1238, "step": 4895 }, { "epoch": 0.9, "learning_rate": 1.964748663797323e-05, "loss": 0.1023, "step": 4896 }, { "epoch": 0.9, "learning_rate": 1.964729122386493e-05, "loss": 0.1093, "step": 4897 }, { "epoch": 0.9, "learning_rate": 1.9647095756580528e-05, "loss": 0.1053, "step": 4898 }, { "epoch": 0.9, "learning_rate": 1.9646900236121117e-05, "loss": 0.1131, "step": 4899 }, { "epoch": 0.9, "learning_rate": 1.9646704662487762e-05, "loss": 0.111, "step": 4900 }, { "epoch": 0.9, "learning_rate": 1.964650903568154e-05, "loss": 0.1172, "step": 4901 }, { "epoch": 0.9, "learning_rate": 1.9646313355703542e-05, "loss": 0.1089, "step": 4902 }, { "epoch": 0.9, "learning_rate": 1.9646117622554835e-05, "loss": 0.1131, "step": 4903 }, { "epoch": 0.9, "learning_rate": 1.9645921836236498e-05, "loss": 0.1081, "step": 4904 }, { "epoch": 0.9, "learning_rate": 1.9645725996749618e-05, "loss": 0.1017, "step": 4905 }, { "epoch": 0.9, "learning_rate": 1.9645530104095273e-05, "loss": 0.1058, "step": 4906 }, { "epoch": 0.9, "learning_rate": 1.9645334158274536e-05, "loss": 0.1004, "step": 4907 }, { "epoch": 0.9, "learning_rate": 1.9645138159288493e-05, "loss": 0.1124, "step": 4908 }, { "epoch": 0.9, "learning_rate": 1.964494210713822e-05, "loss": 0.1075, "step": 4909 }, { "epoch": 0.9, "learning_rate": 1.9644746001824805e-05, "loss": 0.0927, "step": 4910 }, { "epoch": 0.9, "learning_rate": 1.9644549843349318e-05, "loss": 0.0964, "step": 4911 }, { "epoch": 0.9, "learning_rate": 1.9644353631712848e-05, "loss": 0.0991, "step": 4912 }, { "epoch": 0.9, "learning_rate": 1.9644157366916475e-05, "loss": 0.1036, "step": 4913 }, { "epoch": 0.9, "learning_rate": 1.9643961048961283e-05, "loss": 0.1152, "step": 4914 }, { "epoch": 0.9, "learning_rate": 1.9643764677848353e-05, "loss": 0.1037, "step": 4915 }, { "epoch": 0.9, "learning_rate": 1.9643568253578763e-05, "loss": 0.1133, "step": 4916 }, { "epoch": 0.9, "learning_rate": 1.96433717761536e-05, "loss": 0.0976, "step": 4917 }, { "epoch": 0.9, "learning_rate": 1.9643175245573944e-05, "loss": 0.1016, "step": 4918 }, { "epoch": 0.9, "learning_rate": 1.9642978661840884e-05, "loss": 0.1045, "step": 4919 }, { "epoch": 0.9, "learning_rate": 1.9642782024955497e-05, "loss": 0.1037, "step": 4920 }, { "epoch": 0.9, "learning_rate": 1.964258533491887e-05, "loss": 0.1029, "step": 4921 }, { "epoch": 0.9, "learning_rate": 1.9642388591732085e-05, "loss": 0.1148, "step": 4922 }, { "epoch": 0.9, "learning_rate": 1.9642191795396228e-05, "loss": 0.1091, "step": 4923 }, { "epoch": 0.9, "learning_rate": 1.9641994945912385e-05, "loss": 0.114, "step": 4924 }, { "epoch": 0.9, "learning_rate": 1.9641798043281642e-05, "loss": 0.1091, "step": 4925 }, { "epoch": 0.9, "learning_rate": 1.964160108750508e-05, "loss": 0.114, "step": 4926 }, { "epoch": 0.9, "learning_rate": 1.964140407858379e-05, "loss": 0.1177, "step": 4927 }, { "epoch": 0.9, "learning_rate": 1.964120701651885e-05, "loss": 0.1078, "step": 4928 }, { "epoch": 0.9, "learning_rate": 1.9641009901311356e-05, "loss": 0.1025, "step": 4929 }, { "epoch": 0.9, "learning_rate": 1.9640812732962386e-05, "loss": 0.1139, "step": 4930 }, { "epoch": 0.9, "learning_rate": 1.9640615511473033e-05, "loss": 0.1096, "step": 4931 }, { "epoch": 0.9, "learning_rate": 1.964041823684438e-05, "loss": 0.0919, "step": 4932 }, { "epoch": 0.9, "learning_rate": 1.9640220909077518e-05, "loss": 0.1128, "step": 4933 }, { "epoch": 0.9, "learning_rate": 1.9640023528173528e-05, "loss": 0.1069, "step": 4934 }, { "epoch": 0.91, "learning_rate": 1.9639826094133505e-05, "loss": 0.109, "step": 4935 }, { "epoch": 0.91, "learning_rate": 1.9639628606958535e-05, "loss": 0.108, "step": 4936 }, { "epoch": 0.91, "learning_rate": 1.9639431066649704e-05, "loss": 0.107, "step": 4937 }, { "epoch": 0.91, "learning_rate": 1.963923347320811e-05, "loss": 0.1049, "step": 4938 }, { "epoch": 0.91, "learning_rate": 1.9639035826634827e-05, "loss": 0.1047, "step": 4939 }, { "epoch": 0.91, "learning_rate": 1.9638838126930956e-05, "loss": 0.1057, "step": 4940 }, { "epoch": 0.91, "learning_rate": 1.9638640374097585e-05, "loss": 0.0985, "step": 4941 }, { "epoch": 0.91, "learning_rate": 1.96384425681358e-05, "loss": 0.1061, "step": 4942 }, { "epoch": 0.91, "learning_rate": 1.9638244709046694e-05, "loss": 0.1005, "step": 4943 }, { "epoch": 0.91, "learning_rate": 1.9638046796831358e-05, "loss": 0.1044, "step": 4944 }, { "epoch": 0.91, "learning_rate": 1.9637848831490884e-05, "loss": 0.114, "step": 4945 }, { "epoch": 0.91, "learning_rate": 1.963765081302636e-05, "loss": 0.1062, "step": 4946 }, { "epoch": 0.91, "learning_rate": 1.9637452741438877e-05, "loss": 0.1053, "step": 4947 }, { "epoch": 0.91, "learning_rate": 1.963725461672953e-05, "loss": 0.1082, "step": 4948 }, { "epoch": 0.91, "learning_rate": 1.963705643889941e-05, "loss": 0.1031, "step": 4949 }, { "epoch": 0.91, "learning_rate": 1.963685820794961e-05, "loss": 0.1055, "step": 4950 }, { "epoch": 0.91, "learning_rate": 1.963665992388122e-05, "loss": 0.1063, "step": 4951 }, { "epoch": 0.91, "learning_rate": 1.9636461586695335e-05, "loss": 0.1098, "step": 4952 }, { "epoch": 0.91, "learning_rate": 1.963626319639305e-05, "loss": 0.0993, "step": 4953 }, { "epoch": 0.91, "learning_rate": 1.9636064752975456e-05, "loss": 0.1071, "step": 4954 }, { "epoch": 0.91, "learning_rate": 1.9635866256443646e-05, "loss": 0.1058, "step": 4955 }, { "epoch": 0.91, "learning_rate": 1.9635667706798717e-05, "loss": 0.1001, "step": 4956 }, { "epoch": 0.91, "learning_rate": 1.963546910404176e-05, "loss": 0.1085, "step": 4957 }, { "epoch": 0.91, "learning_rate": 1.9635270448173873e-05, "loss": 0.1051, "step": 4958 }, { "epoch": 0.91, "learning_rate": 1.9635071739196148e-05, "loss": 0.0997, "step": 4959 }, { "epoch": 0.91, "learning_rate": 1.9634872977109684e-05, "loss": 0.1154, "step": 4960 }, { "epoch": 0.91, "learning_rate": 1.9634674161915575e-05, "loss": 0.091, "step": 4961 }, { "epoch": 0.91, "learning_rate": 1.9634475293614913e-05, "loss": 0.1051, "step": 4962 }, { "epoch": 0.91, "learning_rate": 1.9634276372208802e-05, "loss": 0.1101, "step": 4963 }, { "epoch": 0.91, "learning_rate": 1.963407739769833e-05, "loss": 0.108, "step": 4964 }, { "epoch": 0.91, "learning_rate": 1.9633878370084597e-05, "loss": 0.1145, "step": 4965 }, { "epoch": 0.91, "learning_rate": 1.9633679289368702e-05, "loss": 0.1176, "step": 4966 }, { "epoch": 0.91, "learning_rate": 1.9633480155551742e-05, "loss": 0.1026, "step": 4967 }, { "epoch": 0.91, "learning_rate": 1.9633280968634812e-05, "loss": 0.1118, "step": 4968 }, { "epoch": 0.91, "learning_rate": 1.9633081728619013e-05, "loss": 0.1127, "step": 4969 }, { "epoch": 0.91, "learning_rate": 1.9632882435505443e-05, "loss": 0.0909, "step": 4970 }, { "epoch": 0.91, "learning_rate": 1.9632683089295195e-05, "loss": 0.1168, "step": 4971 }, { "epoch": 0.91, "learning_rate": 1.963248368998938e-05, "loss": 0.099, "step": 4972 }, { "epoch": 0.91, "learning_rate": 1.963228423758908e-05, "loss": 0.0976, "step": 4973 }, { "epoch": 0.91, "learning_rate": 1.963208473209541e-05, "loss": 0.1103, "step": 4974 }, { "epoch": 0.91, "learning_rate": 1.963188517350946e-05, "loss": 0.1173, "step": 4975 }, { "epoch": 0.91, "learning_rate": 1.9631685561832335e-05, "loss": 0.1078, "step": 4976 }, { "epoch": 0.91, "learning_rate": 1.9631485897065136e-05, "loss": 0.1052, "step": 4977 }, { "epoch": 0.91, "learning_rate": 1.9631286179208957e-05, "loss": 0.1025, "step": 4978 }, { "epoch": 0.91, "learning_rate": 1.9631086408264904e-05, "loss": 0.113, "step": 4979 }, { "epoch": 0.91, "learning_rate": 1.9630886584234078e-05, "loss": 0.1086, "step": 4980 }, { "epoch": 0.91, "learning_rate": 1.9630686707117575e-05, "loss": 0.1006, "step": 4981 }, { "epoch": 0.91, "learning_rate": 1.9630486776916505e-05, "loss": 0.1122, "step": 4982 }, { "epoch": 0.91, "learning_rate": 1.963028679363197e-05, "loss": 0.1162, "step": 4983 }, { "epoch": 0.91, "learning_rate": 1.963008675726506e-05, "loss": 0.1068, "step": 4984 }, { "epoch": 0.91, "learning_rate": 1.9629886667816895e-05, "loss": 0.1108, "step": 4985 }, { "epoch": 0.91, "learning_rate": 1.9629686525288564e-05, "loss": 0.1146, "step": 4986 }, { "epoch": 0.91, "learning_rate": 1.9629486329681175e-05, "loss": 0.1059, "step": 4987 }, { "epoch": 0.91, "learning_rate": 1.9629286080995835e-05, "loss": 0.1078, "step": 4988 }, { "epoch": 0.91, "learning_rate": 1.9629085779233642e-05, "loss": 0.0958, "step": 4989 }, { "epoch": 0.92, "learning_rate": 1.9628885424395704e-05, "loss": 0.1086, "step": 4990 }, { "epoch": 0.92, "learning_rate": 1.9628685016483122e-05, "loss": 0.1121, "step": 4991 }, { "epoch": 0.92, "learning_rate": 1.9628484555497004e-05, "loss": 0.1142, "step": 4992 }, { "epoch": 0.92, "learning_rate": 1.9628284041438454e-05, "loss": 0.107, "step": 4993 }, { "epoch": 0.92, "learning_rate": 1.9628083474308577e-05, "loss": 0.1014, "step": 4994 }, { "epoch": 0.92, "learning_rate": 1.9627882854108477e-05, "loss": 0.1106, "step": 4995 }, { "epoch": 0.92, "learning_rate": 1.9627682180839263e-05, "loss": 0.109, "step": 4996 }, { "epoch": 0.92, "learning_rate": 1.9627481454502036e-05, "loss": 0.1155, "step": 4997 }, { "epoch": 0.92, "learning_rate": 1.962728067509791e-05, "loss": 0.1035, "step": 4998 }, { "epoch": 0.92, "learning_rate": 1.9627079842627986e-05, "loss": 0.1162, "step": 4999 }, { "epoch": 0.92, "learning_rate": 1.962687895709337e-05, "loss": 0.1176, "step": 5000 }, { "epoch": 0.92, "learning_rate": 1.9626678018495174e-05, "loss": 0.1169, "step": 5001 }, { "epoch": 0.92, "learning_rate": 1.96264770268345e-05, "loss": 0.0984, "step": 5002 }, { "epoch": 0.92, "learning_rate": 1.9626275982112463e-05, "loss": 0.0992, "step": 5003 }, { "epoch": 0.92, "learning_rate": 1.9626074884330163e-05, "loss": 0.1178, "step": 5004 }, { "epoch": 0.92, "learning_rate": 1.9625873733488717e-05, "loss": 0.1198, "step": 5005 }, { "epoch": 0.92, "learning_rate": 1.9625672529589228e-05, "loss": 0.1073, "step": 5006 }, { "epoch": 0.92, "learning_rate": 1.9625471272632805e-05, "loss": 0.0966, "step": 5007 }, { "epoch": 0.92, "learning_rate": 1.9625269962620563e-05, "loss": 0.1041, "step": 5008 }, { "epoch": 0.92, "learning_rate": 1.9625068599553602e-05, "loss": 0.1047, "step": 5009 }, { "epoch": 0.92, "learning_rate": 1.962486718343304e-05, "loss": 0.1089, "step": 5010 }, { "epoch": 0.92, "learning_rate": 1.9624665714259983e-05, "loss": 0.1049, "step": 5011 }, { "epoch": 0.92, "learning_rate": 1.9624464192035545e-05, "loss": 0.1104, "step": 5012 }, { "epoch": 0.92, "learning_rate": 1.9624262616760837e-05, "loss": 0.1199, "step": 5013 }, { "epoch": 0.92, "learning_rate": 1.9624060988436968e-05, "loss": 0.1117, "step": 5014 }, { "epoch": 0.92, "learning_rate": 1.9623859307065044e-05, "loss": 0.1002, "step": 5015 }, { "epoch": 0.92, "learning_rate": 1.9623657572646184e-05, "loss": 0.0975, "step": 5016 }, { "epoch": 0.92, "learning_rate": 1.96234557851815e-05, "loss": 0.1097, "step": 5017 }, { "epoch": 0.92, "learning_rate": 1.96232539446721e-05, "loss": 0.1115, "step": 5018 }, { "epoch": 0.92, "learning_rate": 1.96230520511191e-05, "loss": 0.1133, "step": 5019 }, { "epoch": 0.92, "learning_rate": 1.9622850104523613e-05, "loss": 0.1003, "step": 5020 }, { "epoch": 0.92, "learning_rate": 1.962264810488675e-05, "loss": 0.097, "step": 5021 }, { "epoch": 0.92, "learning_rate": 1.9622446052209626e-05, "loss": 0.1099, "step": 5022 }, { "epoch": 0.92, "learning_rate": 1.9622243946493353e-05, "loss": 0.1053, "step": 5023 }, { "epoch": 0.92, "learning_rate": 1.9622041787739047e-05, "loss": 0.1069, "step": 5024 }, { "epoch": 0.92, "learning_rate": 1.9621839575947817e-05, "loss": 0.1105, "step": 5025 }, { "epoch": 0.92, "learning_rate": 1.9621637311120785e-05, "loss": 0.114, "step": 5026 }, { "epoch": 0.92, "learning_rate": 1.9621434993259062e-05, "loss": 0.109, "step": 5027 }, { "epoch": 0.92, "learning_rate": 1.9621232622363765e-05, "loss": 0.1137, "step": 5028 }, { "epoch": 0.92, "learning_rate": 1.9621030198436007e-05, "loss": 0.1136, "step": 5029 }, { "epoch": 0.92, "learning_rate": 1.9620827721476905e-05, "loss": 0.1029, "step": 5030 }, { "epoch": 0.92, "learning_rate": 1.9620625191487578e-05, "loss": 0.1056, "step": 5031 }, { "epoch": 0.92, "learning_rate": 1.9620422608469136e-05, "loss": 0.1141, "step": 5032 }, { "epoch": 0.92, "learning_rate": 1.96202199724227e-05, "loss": 0.1031, "step": 5033 }, { "epoch": 0.92, "learning_rate": 1.962001728334939e-05, "loss": 0.1052, "step": 5034 }, { "epoch": 0.92, "learning_rate": 1.9619814541250316e-05, "loss": 0.105, "step": 5035 }, { "epoch": 0.92, "learning_rate": 1.9619611746126598e-05, "loss": 0.0965, "step": 5036 }, { "epoch": 0.92, "learning_rate": 1.9619408897979356e-05, "loss": 0.1003, "step": 5037 }, { "epoch": 0.92, "learning_rate": 1.9619205996809704e-05, "loss": 0.1031, "step": 5038 }, { "epoch": 0.92, "learning_rate": 1.9619003042618766e-05, "loss": 0.0848, "step": 5039 }, { "epoch": 0.92, "learning_rate": 1.961880003540766e-05, "loss": 0.0944, "step": 5040 }, { "epoch": 0.92, "learning_rate": 1.96185969751775e-05, "loss": 0.1116, "step": 5041 }, { "epoch": 0.92, "learning_rate": 1.9618393861929407e-05, "loss": 0.115, "step": 5042 }, { "epoch": 0.92, "learning_rate": 1.9618190695664508e-05, "loss": 0.1067, "step": 5043 }, { "epoch": 0.92, "learning_rate": 1.9617987476383908e-05, "loss": 0.0977, "step": 5044 }, { "epoch": 0.93, "learning_rate": 1.961778420408874e-05, "loss": 0.1, "step": 5045 }, { "epoch": 0.93, "learning_rate": 1.9617580878780124e-05, "loss": 0.1002, "step": 5046 }, { "epoch": 0.93, "learning_rate": 1.9617377500459168e-05, "loss": 0.1009, "step": 5047 }, { "epoch": 0.93, "learning_rate": 1.9617174069127007e-05, "loss": 0.0959, "step": 5048 }, { "epoch": 0.93, "learning_rate": 1.961697058478476e-05, "loss": 0.0961, "step": 5049 }, { "epoch": 0.93, "learning_rate": 1.961676704743354e-05, "loss": 0.0849, "step": 5050 }, { "epoch": 0.93, "learning_rate": 1.961656345707448e-05, "loss": 0.1005, "step": 5051 }, { "epoch": 0.93, "learning_rate": 1.9616359813708694e-05, "loss": 0.0945, "step": 5052 }, { "epoch": 0.93, "learning_rate": 1.9616156117337306e-05, "loss": 0.116, "step": 5053 }, { "epoch": 0.93, "learning_rate": 1.961595236796144e-05, "loss": 0.1091, "step": 5054 }, { "epoch": 0.93, "learning_rate": 1.9615748565582226e-05, "loss": 0.1165, "step": 5055 }, { "epoch": 0.93, "learning_rate": 1.9615544710200774e-05, "loss": 0.1059, "step": 5056 }, { "epoch": 0.93, "learning_rate": 1.9615340801818218e-05, "loss": 0.092, "step": 5057 }, { "epoch": 0.93, "learning_rate": 1.9615136840435675e-05, "loss": 0.1098, "step": 5058 }, { "epoch": 0.93, "learning_rate": 1.9614932826054274e-05, "loss": 0.116, "step": 5059 }, { "epoch": 0.93, "learning_rate": 1.961472875867514e-05, "loss": 0.1014, "step": 5060 }, { "epoch": 0.93, "learning_rate": 1.961452463829939e-05, "loss": 0.1073, "step": 5061 }, { "epoch": 0.93, "learning_rate": 1.961432046492816e-05, "loss": 0.1082, "step": 5062 }, { "epoch": 0.93, "learning_rate": 1.961411623856257e-05, "loss": 0.1123, "step": 5063 }, { "epoch": 0.93, "learning_rate": 1.9613911959203744e-05, "loss": 0.0955, "step": 5064 }, { "epoch": 0.93, "learning_rate": 1.961370762685281e-05, "loss": 0.1067, "step": 5065 }, { "epoch": 0.93, "learning_rate": 1.9613503241510895e-05, "loss": 0.0969, "step": 5066 }, { "epoch": 0.93, "learning_rate": 1.9613298803179126e-05, "loss": 0.0994, "step": 5067 }, { "epoch": 0.93, "learning_rate": 1.961309431185863e-05, "loss": 0.109, "step": 5068 }, { "epoch": 0.93, "learning_rate": 1.9612889767550533e-05, "loss": 0.1044, "step": 5069 }, { "epoch": 0.93, "learning_rate": 1.961268517025596e-05, "loss": 0.1075, "step": 5070 }, { "epoch": 0.93, "learning_rate": 1.961248051997604e-05, "loss": 0.1031, "step": 5071 }, { "epoch": 0.93, "learning_rate": 1.9612275816711902e-05, "loss": 0.1133, "step": 5072 }, { "epoch": 0.93, "learning_rate": 1.9612071060464677e-05, "loss": 0.1204, "step": 5073 }, { "epoch": 0.93, "learning_rate": 1.9611866251235493e-05, "loss": 0.1065, "step": 5074 }, { "epoch": 0.93, "learning_rate": 1.9611661389025472e-05, "loss": 0.1065, "step": 5075 }, { "epoch": 0.93, "learning_rate": 1.961145647383575e-05, "loss": 0.1018, "step": 5076 }, { "epoch": 0.93, "learning_rate": 1.9611251505667457e-05, "loss": 0.1107, "step": 5077 }, { "epoch": 0.93, "learning_rate": 1.961104648452172e-05, "loss": 0.1045, "step": 5078 }, { "epoch": 0.93, "learning_rate": 1.9610841410399668e-05, "loss": 0.1027, "step": 5079 }, { "epoch": 0.93, "learning_rate": 1.9610636283302434e-05, "loss": 0.1089, "step": 5080 }, { "epoch": 0.93, "learning_rate": 1.9610431103231147e-05, "loss": 0.1061, "step": 5081 }, { "epoch": 0.93, "learning_rate": 1.9610225870186938e-05, "loss": 0.1122, "step": 5082 }, { "epoch": 0.93, "learning_rate": 1.961002058417094e-05, "loss": 0.1038, "step": 5083 }, { "epoch": 0.93, "learning_rate": 1.9609815245184282e-05, "loss": 0.1185, "step": 5084 }, { "epoch": 0.93, "learning_rate": 1.96096098532281e-05, "loss": 0.1018, "step": 5085 }, { "epoch": 0.93, "learning_rate": 1.960940440830352e-05, "loss": 0.1118, "step": 5086 }, { "epoch": 0.93, "learning_rate": 1.9609198910411682e-05, "loss": 0.1163, "step": 5087 }, { "epoch": 0.93, "learning_rate": 1.960899335955371e-05, "loss": 0.1126, "step": 5088 }, { "epoch": 0.93, "learning_rate": 1.9608787755730746e-05, "loss": 0.1161, "step": 5089 }, { "epoch": 0.93, "learning_rate": 1.9608582098943916e-05, "loss": 0.0989, "step": 5090 }, { "epoch": 0.93, "learning_rate": 1.9608376389194354e-05, "loss": 0.1229, "step": 5091 }, { "epoch": 0.93, "learning_rate": 1.96081706264832e-05, "loss": 0.1099, "step": 5092 }, { "epoch": 0.93, "learning_rate": 1.9607964810811583e-05, "loss": 0.1062, "step": 5093 }, { "epoch": 0.93, "learning_rate": 1.960775894218064e-05, "loss": 0.1078, "step": 5094 }, { "epoch": 0.93, "learning_rate": 1.9607553020591506e-05, "loss": 0.1146, "step": 5095 }, { "epoch": 0.93, "learning_rate": 1.960734704604531e-05, "loss": 0.1043, "step": 5096 }, { "epoch": 0.93, "learning_rate": 1.9607141018543197e-05, "loss": 0.1082, "step": 5097 }, { "epoch": 0.93, "learning_rate": 1.9606934938086292e-05, "loss": 0.098, "step": 5098 }, { "epoch": 0.94, "learning_rate": 1.960672880467574e-05, "loss": 0.1071, "step": 5099 }, { "epoch": 0.94, "learning_rate": 1.9606522618312675e-05, "loss": 0.0873, "step": 5100 }, { "epoch": 0.94, "learning_rate": 1.960631637899823e-05, "loss": 0.1109, "step": 5101 }, { "epoch": 0.94, "learning_rate": 1.9606110086733544e-05, "loss": 0.1038, "step": 5102 }, { "epoch": 0.94, "learning_rate": 1.9605903741519757e-05, "loss": 0.1057, "step": 5103 }, { "epoch": 0.94, "learning_rate": 1.9605697343357998e-05, "loss": 0.1089, "step": 5104 }, { "epoch": 0.94, "learning_rate": 1.9605490892249416e-05, "loss": 0.1099, "step": 5105 }, { "epoch": 0.94, "learning_rate": 1.9605284388195142e-05, "loss": 0.0944, "step": 5106 }, { "epoch": 0.94, "learning_rate": 1.9605077831196315e-05, "loss": 0.1047, "step": 5107 }, { "epoch": 0.94, "learning_rate": 1.9604871221254072e-05, "loss": 0.1061, "step": 5108 }, { "epoch": 0.94, "learning_rate": 1.9604664558369554e-05, "loss": 0.1126, "step": 5109 }, { "epoch": 0.94, "learning_rate": 1.96044578425439e-05, "loss": 0.0996, "step": 5110 }, { "epoch": 0.94, "learning_rate": 1.960425107377825e-05, "loss": 0.0898, "step": 5111 }, { "epoch": 0.94, "learning_rate": 1.9604044252073743e-05, "loss": 0.1097, "step": 5112 }, { "epoch": 0.94, "learning_rate": 1.9603837377431523e-05, "loss": 0.0978, "step": 5113 }, { "epoch": 0.94, "learning_rate": 1.9603630449852722e-05, "loss": 0.108, "step": 5114 }, { "epoch": 0.94, "learning_rate": 1.9603423469338485e-05, "loss": 0.1056, "step": 5115 }, { "epoch": 0.94, "learning_rate": 1.9603216435889954e-05, "loss": 0.1018, "step": 5116 }, { "epoch": 0.94, "learning_rate": 1.9603009349508267e-05, "loss": 0.1109, "step": 5117 }, { "epoch": 0.94, "learning_rate": 1.960280221019457e-05, "loss": 0.1175, "step": 5118 }, { "epoch": 0.94, "learning_rate": 1.960259501795e-05, "loss": 0.1114, "step": 5119 }, { "epoch": 0.94, "learning_rate": 1.9602387772775705e-05, "loss": 0.1023, "step": 5120 }, { "epoch": 0.94, "learning_rate": 1.960218047467282e-05, "loss": 0.1085, "step": 5121 }, { "epoch": 0.94, "learning_rate": 1.9601973123642493e-05, "loss": 0.1116, "step": 5122 }, { "epoch": 0.94, "learning_rate": 1.9601765719685867e-05, "loss": 0.1115, "step": 5123 }, { "epoch": 0.94, "learning_rate": 1.960155826280408e-05, "loss": 0.1146, "step": 5124 }, { "epoch": 0.94, "learning_rate": 1.9601350752998282e-05, "loss": 0.0999, "step": 5125 }, { "epoch": 0.94, "learning_rate": 1.960114319026961e-05, "loss": 0.1139, "step": 5126 }, { "epoch": 0.94, "learning_rate": 1.9600935574619213e-05, "loss": 0.1114, "step": 5127 }, { "epoch": 0.94, "learning_rate": 1.9600727906048237e-05, "loss": 0.1157, "step": 5128 }, { "epoch": 0.94, "learning_rate": 1.960052018455782e-05, "loss": 0.1095, "step": 5129 }, { "epoch": 0.94, "learning_rate": 1.9600312410149113e-05, "loss": 0.1114, "step": 5130 }, { "epoch": 0.94, "learning_rate": 1.960010458282326e-05, "loss": 0.1146, "step": 5131 }, { "epoch": 0.94, "learning_rate": 1.9599896702581406e-05, "loss": 0.1007, "step": 5132 }, { "epoch": 0.94, "learning_rate": 1.959968876942469e-05, "loss": 0.1004, "step": 5133 }, { "epoch": 0.94, "learning_rate": 1.959948078335427e-05, "loss": 0.1164, "step": 5134 }, { "epoch": 0.94, "learning_rate": 1.9599272744371288e-05, "loss": 0.1075, "step": 5135 }, { "epoch": 0.94, "learning_rate": 1.9599064652476887e-05, "loss": 0.1051, "step": 5136 }, { "epoch": 0.94, "learning_rate": 1.959885650767222e-05, "loss": 0.1132, "step": 5137 }, { "epoch": 0.94, "learning_rate": 1.959864830995843e-05, "loss": 0.1179, "step": 5138 }, { "epoch": 0.94, "learning_rate": 1.9598440059336662e-05, "loss": 0.1046, "step": 5139 }, { "epoch": 0.94, "learning_rate": 1.959823175580807e-05, "loss": 0.1099, "step": 5140 }, { "epoch": 0.94, "learning_rate": 1.95980233993738e-05, "loss": 0.1058, "step": 5141 }, { "epoch": 0.94, "learning_rate": 1.9597814990035e-05, "loss": 0.1032, "step": 5142 }, { "epoch": 0.94, "learning_rate": 1.959760652779282e-05, "loss": 0.1076, "step": 5143 }, { "epoch": 0.94, "learning_rate": 1.959739801264841e-05, "loss": 0.1, "step": 5144 }, { "epoch": 0.94, "learning_rate": 1.959718944460291e-05, "loss": 0.1003, "step": 5145 }, { "epoch": 0.94, "learning_rate": 1.959698082365748e-05, "loss": 0.0989, "step": 5146 }, { "epoch": 0.94, "learning_rate": 1.959677214981327e-05, "loss": 0.1046, "step": 5147 }, { "epoch": 0.94, "learning_rate": 1.9596563423071426e-05, "loss": 0.1134, "step": 5148 }, { "epoch": 0.94, "learning_rate": 1.9596354643433096e-05, "loss": 0.1211, "step": 5149 }, { "epoch": 0.94, "learning_rate": 1.959614581089944e-05, "loss": 0.1014, "step": 5150 }, { "epoch": 0.94, "learning_rate": 1.9595936925471598e-05, "loss": 0.0984, "step": 5151 }, { "epoch": 0.94, "learning_rate": 1.959572798715073e-05, "loss": 0.1012, "step": 5152 }, { "epoch": 0.94, "learning_rate": 1.9595518995937987e-05, "loss": 0.1212, "step": 5153 }, { "epoch": 0.95, "learning_rate": 1.9595309951834514e-05, "loss": 0.1047, "step": 5154 }, { "epoch": 0.95, "learning_rate": 1.9595100854841468e-05, "loss": 0.1126, "step": 5155 }, { "epoch": 0.95, "learning_rate": 1.9594891704960002e-05, "loss": 0.1052, "step": 5156 }, { "epoch": 0.95, "learning_rate": 1.959468250219127e-05, "loss": 0.1051, "step": 5157 }, { "epoch": 0.95, "learning_rate": 1.959447324653642e-05, "loss": 0.0974, "step": 5158 }, { "epoch": 0.95, "learning_rate": 1.959426393799661e-05, "loss": 0.1101, "step": 5159 }, { "epoch": 0.95, "learning_rate": 1.9594054576572996e-05, "loss": 0.1097, "step": 5160 }, { "epoch": 0.95, "learning_rate": 1.9593845162266724e-05, "loss": 0.1151, "step": 5161 }, { "epoch": 0.95, "learning_rate": 1.9593635695078954e-05, "loss": 0.1065, "step": 5162 }, { "epoch": 0.95, "learning_rate": 1.959342617501084e-05, "loss": 0.0986, "step": 5163 }, { "epoch": 0.95, "learning_rate": 1.9593216602063532e-05, "loss": 0.1051, "step": 5164 }, { "epoch": 0.95, "learning_rate": 1.9593006976238194e-05, "loss": 0.1096, "step": 5165 }, { "epoch": 0.95, "learning_rate": 1.9592797297535978e-05, "loss": 0.1152, "step": 5166 }, { "epoch": 0.95, "learning_rate": 1.9592587565958037e-05, "loss": 0.1043, "step": 5167 }, { "epoch": 0.95, "learning_rate": 1.9592377781505526e-05, "loss": 0.1083, "step": 5168 }, { "epoch": 0.95, "learning_rate": 1.9592167944179605e-05, "loss": 0.1133, "step": 5169 }, { "epoch": 0.95, "learning_rate": 1.9591958053981433e-05, "loss": 0.1081, "step": 5170 }, { "epoch": 0.95, "learning_rate": 1.959174811091216e-05, "loss": 0.1083, "step": 5171 }, { "epoch": 0.95, "learning_rate": 1.9591538114972946e-05, "loss": 0.0984, "step": 5172 }, { "epoch": 0.95, "learning_rate": 1.9591328066164952e-05, "loss": 0.106, "step": 5173 }, { "epoch": 0.95, "learning_rate": 1.959111796448933e-05, "loss": 0.1189, "step": 5174 }, { "epoch": 0.95, "learning_rate": 1.9590907809947244e-05, "loss": 0.1031, "step": 5175 }, { "epoch": 0.95, "learning_rate": 1.959069760253985e-05, "loss": 0.1134, "step": 5176 }, { "epoch": 0.95, "learning_rate": 1.9590487342268304e-05, "loss": 0.1221, "step": 5177 }, { "epoch": 0.95, "learning_rate": 1.9590277029133768e-05, "loss": 0.0977, "step": 5178 }, { "epoch": 0.95, "learning_rate": 1.95900666631374e-05, "loss": 0.1135, "step": 5179 }, { "epoch": 0.95, "learning_rate": 1.9589856244280358e-05, "loss": 0.1089, "step": 5180 }, { "epoch": 0.95, "learning_rate": 1.9589645772563806e-05, "loss": 0.1149, "step": 5181 }, { "epoch": 0.95, "learning_rate": 1.95894352479889e-05, "loss": 0.1, "step": 5182 }, { "epoch": 0.95, "learning_rate": 1.9589224670556805e-05, "loss": 0.1066, "step": 5183 }, { "epoch": 0.95, "learning_rate": 1.9589014040268678e-05, "loss": 0.1156, "step": 5184 }, { "epoch": 0.95, "learning_rate": 1.958880335712568e-05, "loss": 0.108, "step": 5185 }, { "epoch": 0.95, "learning_rate": 1.958859262112897e-05, "loss": 0.1087, "step": 5186 }, { "epoch": 0.95, "learning_rate": 1.9588381832279722e-05, "loss": 0.0985, "step": 5187 }, { "epoch": 0.95, "learning_rate": 1.958817099057908e-05, "loss": 0.1156, "step": 5188 }, { "epoch": 0.95, "learning_rate": 1.958796009602822e-05, "loss": 0.0905, "step": 5189 }, { "epoch": 0.95, "learning_rate": 1.9587749148628298e-05, "loss": 0.1048, "step": 5190 }, { "epoch": 0.95, "learning_rate": 1.9587538148380476e-05, "loss": 0.102, "step": 5191 }, { "epoch": 0.95, "learning_rate": 1.9587327095285923e-05, "loss": 0.1065, "step": 5192 }, { "epoch": 0.95, "learning_rate": 1.9587115989345794e-05, "loss": 0.1049, "step": 5193 }, { "epoch": 0.95, "learning_rate": 1.958690483056126e-05, "loss": 0.1178, "step": 5194 }, { "epoch": 0.95, "learning_rate": 1.958669361893348e-05, "loss": 0.0959, "step": 5195 }, { "epoch": 0.95, "learning_rate": 1.9586482354463624e-05, "loss": 0.0927, "step": 5196 }, { "epoch": 0.95, "learning_rate": 1.9586271037152847e-05, "loss": 0.0965, "step": 5197 }, { "epoch": 0.95, "learning_rate": 1.9586059667002325e-05, "loss": 0.1008, "step": 5198 }, { "epoch": 0.95, "learning_rate": 1.9585848244013214e-05, "loss": 0.1062, "step": 5199 }, { "epoch": 0.95, "learning_rate": 1.9585636768186685e-05, "loss": 0.1123, "step": 5200 }, { "epoch": 0.95, "learning_rate": 1.95854252395239e-05, "loss": 0.1165, "step": 5201 }, { "epoch": 0.95, "learning_rate": 1.9585213658026028e-05, "loss": 0.1101, "step": 5202 }, { "epoch": 0.95, "learning_rate": 1.958500202369423e-05, "loss": 0.104, "step": 5203 }, { "epoch": 0.95, "learning_rate": 1.958479033652968e-05, "loss": 0.0803, "step": 5204 }, { "epoch": 0.95, "learning_rate": 1.958457859653354e-05, "loss": 0.0896, "step": 5205 }, { "epoch": 0.95, "learning_rate": 1.958436680370698e-05, "loss": 0.1016, "step": 5206 }, { "epoch": 0.95, "learning_rate": 1.9584154958051162e-05, "loss": 0.0884, "step": 5207 }, { "epoch": 0.96, "learning_rate": 1.958394305956726e-05, "loss": 0.0982, "step": 5208 }, { "epoch": 0.96, "learning_rate": 1.958373110825644e-05, "loss": 0.1235, "step": 5209 }, { "epoch": 0.96, "learning_rate": 1.9583519104119867e-05, "loss": 0.1194, "step": 5210 }, { "epoch": 0.96, "learning_rate": 1.9583307047158714e-05, "loss": 0.105, "step": 5211 }, { "epoch": 0.96, "learning_rate": 1.9583094937374147e-05, "loss": 0.1068, "step": 5212 }, { "epoch": 0.96, "learning_rate": 1.9582882774767338e-05, "loss": 0.097, "step": 5213 }, { "epoch": 0.96, "learning_rate": 1.9582670559339453e-05, "loss": 0.1165, "step": 5214 }, { "epoch": 0.96, "learning_rate": 1.9582458291091664e-05, "loss": 0.1079, "step": 5215 }, { "epoch": 0.96, "learning_rate": 1.958224597002514e-05, "loss": 0.1117, "step": 5216 }, { "epoch": 0.96, "learning_rate": 1.9582033596141053e-05, "loss": 0.0963, "step": 5217 }, { "epoch": 0.96, "learning_rate": 1.9581821169440573e-05, "loss": 0.1165, "step": 5218 }, { "epoch": 0.96, "learning_rate": 1.9581608689924864e-05, "loss": 0.1116, "step": 5219 }, { "epoch": 0.96, "learning_rate": 1.958139615759511e-05, "loss": 0.1033, "step": 5220 }, { "epoch": 0.96, "learning_rate": 1.9581183572452474e-05, "loss": 0.1144, "step": 5221 }, { "epoch": 0.96, "learning_rate": 1.958097093449813e-05, "loss": 0.1225, "step": 5222 }, { "epoch": 0.96, "learning_rate": 1.958075824373325e-05, "loss": 0.1065, "step": 5223 }, { "epoch": 0.96, "learning_rate": 1.9580545500159003e-05, "loss": 0.1009, "step": 5224 }, { "epoch": 0.96, "learning_rate": 1.958033270377657e-05, "loss": 0.0983, "step": 5225 }, { "epoch": 0.96, "learning_rate": 1.9580119854587113e-05, "loss": 0.0984, "step": 5226 }, { "epoch": 0.96, "learning_rate": 1.9579906952591814e-05, "loss": 0.1015, "step": 5227 }, { "epoch": 0.96, "learning_rate": 1.9579693997791845e-05, "loss": 0.1152, "step": 5228 }, { "epoch": 0.96, "learning_rate": 1.9579480990188373e-05, "loss": 0.0951, "step": 5229 }, { "epoch": 0.96, "learning_rate": 1.9579267929782585e-05, "loss": 0.1032, "step": 5230 }, { "epoch": 0.96, "learning_rate": 1.957905481657564e-05, "loss": 0.1133, "step": 5231 }, { "epoch": 0.96, "learning_rate": 1.9578841650568726e-05, "loss": 0.1147, "step": 5232 }, { "epoch": 0.96, "learning_rate": 1.957862843176301e-05, "loss": 0.1105, "step": 5233 }, { "epoch": 0.96, "learning_rate": 1.957841516015967e-05, "loss": 0.109, "step": 5234 }, { "epoch": 0.96, "learning_rate": 1.9578201835759882e-05, "loss": 0.1096, "step": 5235 }, { "epoch": 0.96, "learning_rate": 1.957798845856482e-05, "loss": 0.0979, "step": 5236 }, { "epoch": 0.96, "learning_rate": 1.9577775028575662e-05, "loss": 0.1043, "step": 5237 }, { "epoch": 0.96, "learning_rate": 1.957756154579358e-05, "loss": 0.1198, "step": 5238 }, { "epoch": 0.96, "learning_rate": 1.957734801021976e-05, "loss": 0.1101, "step": 5239 }, { "epoch": 0.96, "learning_rate": 1.957713442185537e-05, "loss": 0.1163, "step": 5240 }, { "epoch": 0.96, "learning_rate": 1.957692078070159e-05, "loss": 0.1058, "step": 5241 }, { "epoch": 0.96, "learning_rate": 1.95767070867596e-05, "loss": 0.11, "step": 5242 }, { "epoch": 0.96, "learning_rate": 1.9576493340030578e-05, "loss": 0.0938, "step": 5243 }, { "epoch": 0.96, "learning_rate": 1.9576279540515698e-05, "loss": 0.0955, "step": 5244 }, { "epoch": 0.96, "learning_rate": 1.957606568821614e-05, "loss": 0.1082, "step": 5245 }, { "epoch": 0.96, "learning_rate": 1.9575851783133086e-05, "loss": 0.1101, "step": 5246 }, { "epoch": 0.96, "learning_rate": 1.9575637825267715e-05, "loss": 0.1037, "step": 5247 }, { "epoch": 0.96, "learning_rate": 1.95754238146212e-05, "loss": 0.1034, "step": 5248 }, { "epoch": 0.96, "learning_rate": 1.9575209751194727e-05, "loss": 0.1043, "step": 5249 }, { "epoch": 0.96, "learning_rate": 1.9574995634989468e-05, "loss": 0.0989, "step": 5250 }, { "epoch": 0.96, "learning_rate": 1.9574781466006613e-05, "loss": 0.1185, "step": 5251 }, { "epoch": 0.96, "learning_rate": 1.9574567244247338e-05, "loss": 0.1028, "step": 5252 }, { "epoch": 0.96, "learning_rate": 1.9574352969712824e-05, "loss": 0.112, "step": 5253 }, { "epoch": 0.96, "learning_rate": 1.957413864240425e-05, "loss": 0.0999, "step": 5254 }, { "epoch": 0.96, "learning_rate": 1.9573924262322802e-05, "loss": 0.0975, "step": 5255 }, { "epoch": 0.96, "learning_rate": 1.957370982946966e-05, "loss": 0.1055, "step": 5256 }, { "epoch": 0.96, "learning_rate": 1.9573495343846006e-05, "loss": 0.1098, "step": 5257 }, { "epoch": 0.96, "learning_rate": 1.9573280805453018e-05, "loss": 0.0993, "step": 5258 }, { "epoch": 0.96, "learning_rate": 1.9573066214291883e-05, "loss": 0.1169, "step": 5259 }, { "epoch": 0.96, "learning_rate": 1.957285157036378e-05, "loss": 0.098, "step": 5260 }, { "epoch": 0.96, "learning_rate": 1.95726368736699e-05, "loss": 0.0886, "step": 5261 }, { "epoch": 0.96, "learning_rate": 1.9572422124211418e-05, "loss": 0.0974, "step": 5262 }, { "epoch": 0.97, "learning_rate": 1.9572207321989524e-05, "loss": 0.1075, "step": 5263 }, { "epoch": 0.97, "learning_rate": 1.9571992467005395e-05, "loss": 0.0959, "step": 5264 }, { "epoch": 0.97, "learning_rate": 1.957177755926022e-05, "loss": 0.1023, "step": 5265 }, { "epoch": 0.97, "learning_rate": 1.9571562598755187e-05, "loss": 0.1189, "step": 5266 }, { "epoch": 0.97, "learning_rate": 1.9571347585491473e-05, "loss": 0.1047, "step": 5267 }, { "epoch": 0.97, "learning_rate": 1.9571132519470267e-05, "loss": 0.106, "step": 5268 }, { "epoch": 0.97, "learning_rate": 1.9570917400692757e-05, "loss": 0.1302, "step": 5269 }, { "epoch": 0.97, "learning_rate": 1.957070222916012e-05, "loss": 0.0995, "step": 5270 }, { "epoch": 0.97, "learning_rate": 1.9570487004873553e-05, "loss": 0.1206, "step": 5271 }, { "epoch": 0.97, "learning_rate": 1.9570271727834238e-05, "loss": 0.1147, "step": 5272 }, { "epoch": 0.97, "learning_rate": 1.957005639804336e-05, "loss": 0.0969, "step": 5273 }, { "epoch": 0.97, "learning_rate": 1.9569841015502105e-05, "loss": 0.1126, "step": 5274 }, { "epoch": 0.97, "learning_rate": 1.9569625580211665e-05, "loss": 0.1157, "step": 5275 }, { "epoch": 0.97, "learning_rate": 1.956941009217322e-05, "loss": 0.1038, "step": 5276 }, { "epoch": 0.97, "learning_rate": 1.9569194551387968e-05, "loss": 0.1175, "step": 5277 }, { "epoch": 0.97, "learning_rate": 1.956897895785709e-05, "loss": 0.1014, "step": 5278 }, { "epoch": 0.97, "learning_rate": 1.9568763311581775e-05, "loss": 0.1098, "step": 5279 }, { "epoch": 0.97, "learning_rate": 1.9568547612563212e-05, "loss": 0.1007, "step": 5280 }, { "epoch": 0.97, "learning_rate": 1.956833186080259e-05, "loss": 0.0975, "step": 5281 }, { "epoch": 0.97, "learning_rate": 1.9568116056301098e-05, "loss": 0.104, "step": 5282 }, { "epoch": 0.97, "learning_rate": 1.9567900199059928e-05, "loss": 0.1115, "step": 5283 }, { "epoch": 0.97, "learning_rate": 1.9567684289080268e-05, "loss": 0.1072, "step": 5284 }, { "epoch": 0.97, "learning_rate": 1.9567468326363307e-05, "loss": 0.1103, "step": 5285 }, { "epoch": 0.97, "learning_rate": 1.956725231091024e-05, "loss": 0.1143, "step": 5286 }, { "epoch": 0.97, "learning_rate": 1.9567036242722247e-05, "loss": 0.1047, "step": 5287 }, { "epoch": 0.97, "learning_rate": 1.956682012180053e-05, "loss": 0.1075, "step": 5288 }, { "epoch": 0.97, "learning_rate": 1.9566603948146277e-05, "loss": 0.0967, "step": 5289 }, { "epoch": 0.97, "learning_rate": 1.956638772176068e-05, "loss": 0.1113, "step": 5290 }, { "epoch": 0.97, "learning_rate": 1.9566171442644927e-05, "loss": 0.1036, "step": 5291 }, { "epoch": 0.97, "learning_rate": 1.9565955110800213e-05, "loss": 0.0983, "step": 5292 }, { "epoch": 0.97, "learning_rate": 1.9565738726227735e-05, "loss": 0.0905, "step": 5293 }, { "epoch": 0.97, "learning_rate": 1.9565522288928676e-05, "loss": 0.1203, "step": 5294 }, { "epoch": 0.97, "learning_rate": 1.9565305798904235e-05, "loss": 0.1192, "step": 5295 }, { "epoch": 0.97, "learning_rate": 1.9565089256155604e-05, "loss": 0.0948, "step": 5296 }, { "epoch": 0.97, "learning_rate": 1.9564872660683977e-05, "loss": 0.1079, "step": 5297 }, { "epoch": 0.97, "learning_rate": 1.956465601249055e-05, "loss": 0.1141, "step": 5298 }, { "epoch": 0.97, "learning_rate": 1.9564439311576515e-05, "loss": 0.1075, "step": 5299 }, { "epoch": 0.97, "learning_rate": 1.9564222557943064e-05, "loss": 0.11, "step": 5300 }, { "epoch": 0.97, "learning_rate": 1.9564005751591394e-05, "loss": 0.1096, "step": 5301 }, { "epoch": 0.97, "learning_rate": 1.95637888925227e-05, "loss": 0.1099, "step": 5302 }, { "epoch": 0.97, "learning_rate": 1.956357198073818e-05, "loss": 0.1119, "step": 5303 }, { "epoch": 0.97, "learning_rate": 1.9563355016239028e-05, "loss": 0.1219, "step": 5304 }, { "epoch": 0.97, "learning_rate": 1.9563137999026437e-05, "loss": 0.1135, "step": 5305 }, { "epoch": 0.97, "learning_rate": 1.9562920929101604e-05, "loss": 0.1108, "step": 5306 }, { "epoch": 0.97, "learning_rate": 1.956270380646573e-05, "loss": 0.103, "step": 5307 }, { "epoch": 0.97, "learning_rate": 1.9562486631120007e-05, "loss": 0.1005, "step": 5308 }, { "epoch": 0.97, "learning_rate": 1.9562269403065635e-05, "loss": 0.108, "step": 5309 }, { "epoch": 0.97, "learning_rate": 1.9562052122303808e-05, "loss": 0.1231, "step": 5310 }, { "epoch": 0.97, "learning_rate": 1.9561834788835725e-05, "loss": 0.1035, "step": 5311 }, { "epoch": 0.97, "learning_rate": 1.9561617402662586e-05, "loss": 0.1226, "step": 5312 }, { "epoch": 0.97, "learning_rate": 1.9561399963785586e-05, "loss": 0.1111, "step": 5313 }, { "epoch": 0.97, "learning_rate": 1.9561182472205932e-05, "loss": 0.0937, "step": 5314 }, { "epoch": 0.97, "learning_rate": 1.9560964927924807e-05, "loss": 0.0966, "step": 5315 }, { "epoch": 0.97, "learning_rate": 1.9560747330943425e-05, "loss": 0.0987, "step": 5316 }, { "epoch": 0.98, "learning_rate": 1.956052968126298e-05, "loss": 0.0979, "step": 5317 }, { "epoch": 0.98, "learning_rate": 1.956031197888467e-05, "loss": 0.1194, "step": 5318 }, { "epoch": 0.98, "learning_rate": 1.9560094223809696e-05, "loss": 0.1079, "step": 5319 }, { "epoch": 0.98, "learning_rate": 1.955987641603926e-05, "loss": 0.101, "step": 5320 }, { "epoch": 0.98, "learning_rate": 1.955965855557456e-05, "loss": 0.0939, "step": 5321 }, { "epoch": 0.98, "learning_rate": 1.9559440642416798e-05, "loss": 0.1058, "step": 5322 }, { "epoch": 0.98, "learning_rate": 1.9559222676567174e-05, "loss": 0.0958, "step": 5323 }, { "epoch": 0.98, "learning_rate": 1.955900465802689e-05, "loss": 0.1018, "step": 5324 }, { "epoch": 0.98, "learning_rate": 1.9558786586797153e-05, "loss": 0.0983, "step": 5325 }, { "epoch": 0.98, "learning_rate": 1.955856846287916e-05, "loss": 0.106, "step": 5326 }, { "epoch": 0.98, "learning_rate": 1.955835028627411e-05, "loss": 0.0991, "step": 5327 }, { "epoch": 0.98, "learning_rate": 1.9558132056983212e-05, "loss": 0.1228, "step": 5328 }, { "epoch": 0.98, "learning_rate": 1.9557913775007666e-05, "loss": 0.1076, "step": 5329 }, { "epoch": 0.98, "learning_rate": 1.9557695440348676e-05, "loss": 0.1121, "step": 5330 }, { "epoch": 0.98, "learning_rate": 1.955747705300744e-05, "loss": 0.1054, "step": 5331 }, { "epoch": 0.98, "learning_rate": 1.955725861298517e-05, "loss": 0.088, "step": 5332 }, { "epoch": 0.98, "learning_rate": 1.9557040120283068e-05, "loss": 0.1041, "step": 5333 }, { "epoch": 0.98, "learning_rate": 1.9556821574902335e-05, "loss": 0.0979, "step": 5334 }, { "epoch": 0.98, "learning_rate": 1.9556602976844176e-05, "loss": 0.1051, "step": 5335 }, { "epoch": 0.98, "learning_rate": 1.95563843261098e-05, "loss": 0.1029, "step": 5336 }, { "epoch": 0.98, "learning_rate": 1.955616562270041e-05, "loss": 0.1022, "step": 5337 }, { "epoch": 0.98, "learning_rate": 1.955594686661721e-05, "loss": 0.1051, "step": 5338 }, { "epoch": 0.98, "learning_rate": 1.955572805786141e-05, "loss": 0.1092, "step": 5339 }, { "epoch": 0.98, "learning_rate": 1.955550919643421e-05, "loss": 0.1154, "step": 5340 }, { "epoch": 0.98, "learning_rate": 1.9555290282336822e-05, "loss": 0.0951, "step": 5341 }, { "epoch": 0.98, "learning_rate": 1.9555071315570445e-05, "loss": 0.0936, "step": 5342 }, { "epoch": 0.98, "learning_rate": 1.9554852296136295e-05, "loss": 0.1085, "step": 5343 }, { "epoch": 0.98, "learning_rate": 1.9554633224035574e-05, "loss": 0.1042, "step": 5344 }, { "epoch": 0.98, "learning_rate": 1.9554414099269492e-05, "loss": 0.1064, "step": 5345 }, { "epoch": 0.98, "learning_rate": 1.9554194921839255e-05, "loss": 0.0977, "step": 5346 }, { "epoch": 0.98, "learning_rate": 1.9553975691746073e-05, "loss": 0.0984, "step": 5347 }, { "epoch": 0.98, "learning_rate": 1.9553756408991148e-05, "loss": 0.105, "step": 5348 }, { "epoch": 0.98, "learning_rate": 1.95535370735757e-05, "loss": 0.1138, "step": 5349 }, { "epoch": 0.98, "learning_rate": 1.955331768550093e-05, "loss": 0.0959, "step": 5350 }, { "epoch": 0.98, "learning_rate": 1.9553098244768048e-05, "loss": 0.1072, "step": 5351 }, { "epoch": 0.98, "learning_rate": 1.9552878751378266e-05, "loss": 0.104, "step": 5352 }, { "epoch": 0.98, "learning_rate": 1.9552659205332795e-05, "loss": 0.1076, "step": 5353 }, { "epoch": 0.98, "learning_rate": 1.9552439606632836e-05, "loss": 0.1075, "step": 5354 }, { "epoch": 0.98, "learning_rate": 1.955221995527961e-05, "loss": 0.0927, "step": 5355 }, { "epoch": 0.98, "learning_rate": 1.9552000251274325e-05, "loss": 0.1099, "step": 5356 }, { "epoch": 0.98, "learning_rate": 1.9551780494618188e-05, "loss": 0.1145, "step": 5357 }, { "epoch": 0.98, "learning_rate": 1.9551560685312414e-05, "loss": 0.1106, "step": 5358 }, { "epoch": 0.98, "learning_rate": 1.9551340823358217e-05, "loss": 0.1061, "step": 5359 }, { "epoch": 0.98, "learning_rate": 1.9551120908756802e-05, "loss": 0.1084, "step": 5360 }, { "epoch": 0.98, "learning_rate": 1.9550900941509385e-05, "loss": 0.1111, "step": 5361 }, { "epoch": 0.98, "learning_rate": 1.9550680921617177e-05, "loss": 0.1018, "step": 5362 }, { "epoch": 0.98, "learning_rate": 1.9550460849081397e-05, "loss": 0.1068, "step": 5363 }, { "epoch": 0.98, "learning_rate": 1.9550240723903248e-05, "loss": 0.106, "step": 5364 }, { "epoch": 0.98, "learning_rate": 1.955002054608395e-05, "loss": 0.0957, "step": 5365 }, { "epoch": 0.98, "learning_rate": 1.9549800315624714e-05, "loss": 0.1152, "step": 5366 }, { "epoch": 0.98, "learning_rate": 1.9549580032526755e-05, "loss": 0.1019, "step": 5367 }, { "epoch": 0.98, "learning_rate": 1.954935969679129e-05, "loss": 0.116, "step": 5368 }, { "epoch": 0.98, "learning_rate": 1.9549139308419528e-05, "loss": 0.1049, "step": 5369 }, { "epoch": 0.98, "learning_rate": 1.954891886741269e-05, "loss": 0.1129, "step": 5370 }, { "epoch": 0.98, "learning_rate": 1.9548698373771983e-05, "loss": 0.1049, "step": 5371 }, { "epoch": 0.99, "learning_rate": 1.9548477827498627e-05, "loss": 0.1054, "step": 5372 }, { "epoch": 0.99, "learning_rate": 1.954825722859384e-05, "loss": 0.1037, "step": 5373 }, { "epoch": 0.99, "learning_rate": 1.954803657705884e-05, "loss": 0.0986, "step": 5374 }, { "epoch": 0.99, "learning_rate": 1.9547815872894832e-05, "loss": 0.1229, "step": 5375 }, { "epoch": 0.99, "learning_rate": 1.954759511610304e-05, "loss": 0.1075, "step": 5376 }, { "epoch": 0.99, "learning_rate": 1.9547374306684678e-05, "loss": 0.1021, "step": 5377 }, { "epoch": 0.99, "learning_rate": 1.954715344464097e-05, "loss": 0.0911, "step": 5378 }, { "epoch": 0.99, "learning_rate": 1.9546932529973128e-05, "loss": 0.1111, "step": 5379 }, { "epoch": 0.99, "learning_rate": 1.9546711562682368e-05, "loss": 0.1165, "step": 5380 }, { "epoch": 0.99, "learning_rate": 1.954649054276991e-05, "loss": 0.1167, "step": 5381 }, { "epoch": 0.99, "learning_rate": 1.9546269470236973e-05, "loss": 0.1037, "step": 5382 }, { "epoch": 0.99, "learning_rate": 1.9546048345084777e-05, "loss": 0.104, "step": 5383 }, { "epoch": 0.99, "learning_rate": 1.9545827167314536e-05, "loss": 0.1012, "step": 5384 }, { "epoch": 0.99, "learning_rate": 1.9545605936927475e-05, "loss": 0.1043, "step": 5385 }, { "epoch": 0.99, "learning_rate": 1.9545384653924804e-05, "loss": 0.1095, "step": 5386 }, { "epoch": 0.99, "learning_rate": 1.9545163318307757e-05, "loss": 0.114, "step": 5387 }, { "epoch": 0.99, "learning_rate": 1.9544941930077543e-05, "loss": 0.1023, "step": 5388 }, { "epoch": 0.99, "learning_rate": 1.9544720489235386e-05, "loss": 0.0912, "step": 5389 }, { "epoch": 0.99, "learning_rate": 1.9544498995782507e-05, "loss": 0.1008, "step": 5390 }, { "epoch": 0.99, "learning_rate": 1.9544277449720125e-05, "loss": 0.1127, "step": 5391 }, { "epoch": 0.99, "learning_rate": 1.954405585104946e-05, "loss": 0.1043, "step": 5392 }, { "epoch": 0.99, "learning_rate": 1.9543834199771735e-05, "loss": 0.1039, "step": 5393 }, { "epoch": 0.99, "learning_rate": 1.9543612495888178e-05, "loss": 0.1108, "step": 5394 }, { "epoch": 0.99, "learning_rate": 1.9543390739400002e-05, "loss": 0.1191, "step": 5395 }, { "epoch": 0.99, "learning_rate": 1.954316893030843e-05, "loss": 0.0943, "step": 5396 }, { "epoch": 0.99, "learning_rate": 1.954294706861469e-05, "loss": 0.1094, "step": 5397 }, { "epoch": 0.99, "learning_rate": 1.9542725154320004e-05, "loss": 0.0911, "step": 5398 }, { "epoch": 0.99, "learning_rate": 1.954250318742559e-05, "loss": 0.1099, "step": 5399 }, { "epoch": 0.99, "learning_rate": 1.9542281167932676e-05, "loss": 0.1113, "step": 5400 }, { "epoch": 0.99, "learning_rate": 1.9542059095842484e-05, "loss": 0.1108, "step": 5401 }, { "epoch": 0.99, "learning_rate": 1.9541836971156242e-05, "loss": 0.1183, "step": 5402 }, { "epoch": 0.99, "learning_rate": 1.9541614793875166e-05, "loss": 0.1089, "step": 5403 }, { "epoch": 0.99, "learning_rate": 1.954139256400049e-05, "loss": 0.0975, "step": 5404 }, { "epoch": 0.99, "learning_rate": 1.9541170281533436e-05, "loss": 0.1046, "step": 5405 }, { "epoch": 0.99, "learning_rate": 1.954094794647522e-05, "loss": 0.1071, "step": 5406 }, { "epoch": 0.99, "learning_rate": 1.9540725558827082e-05, "loss": 0.1155, "step": 5407 }, { "epoch": 0.99, "learning_rate": 1.9540503118590242e-05, "loss": 0.1153, "step": 5408 }, { "epoch": 0.99, "learning_rate": 1.954028062576592e-05, "loss": 0.1042, "step": 5409 }, { "epoch": 0.99, "learning_rate": 1.954005808035535e-05, "loss": 0.0994, "step": 5410 }, { "epoch": 0.99, "learning_rate": 1.953983548235976e-05, "loss": 0.1039, "step": 5411 }, { "epoch": 0.99, "learning_rate": 1.953961283178037e-05, "loss": 0.1024, "step": 5412 }, { "epoch": 0.99, "learning_rate": 1.953939012861841e-05, "loss": 0.1144, "step": 5413 }, { "epoch": 0.99, "learning_rate": 1.953916737287511e-05, "loss": 0.1014, "step": 5414 }, { "epoch": 0.99, "learning_rate": 1.9538944564551695e-05, "loss": 0.1011, "step": 5415 }, { "epoch": 0.99, "learning_rate": 1.9538721703649392e-05, "loss": 0.1021, "step": 5416 }, { "epoch": 0.99, "learning_rate": 1.9538498790169434e-05, "loss": 0.0944, "step": 5417 }, { "epoch": 0.99, "learning_rate": 1.9538275824113046e-05, "loss": 0.1103, "step": 5418 }, { "epoch": 0.99, "learning_rate": 1.953805280548146e-05, "loss": 0.1038, "step": 5419 }, { "epoch": 0.99, "learning_rate": 1.95378297342759e-05, "loss": 0.1055, "step": 5420 }, { "epoch": 0.99, "learning_rate": 1.9537606610497605e-05, "loss": 0.1023, "step": 5421 }, { "epoch": 0.99, "learning_rate": 1.9537383434147795e-05, "loss": 0.1124, "step": 5422 }, { "epoch": 0.99, "learning_rate": 1.9537160205227705e-05, "loss": 0.0947, "step": 5423 }, { "epoch": 0.99, "learning_rate": 1.953693692373856e-05, "loss": 0.0965, "step": 5424 }, { "epoch": 0.99, "learning_rate": 1.95367135896816e-05, "loss": 0.0936, "step": 5425 }, { "epoch": 1.0, "learning_rate": 1.953649020305805e-05, "loss": 0.0995, "step": 5426 }, { "epoch": 1.0, "learning_rate": 1.9536266763869143e-05, "loss": 0.1014, "step": 5427 }, { "epoch": 1.0, "learning_rate": 1.9536043272116112e-05, "loss": 0.0989, "step": 5428 }, { "epoch": 1.0, "learning_rate": 1.9535819727800184e-05, "loss": 0.0995, "step": 5429 }, { "epoch": 1.0, "learning_rate": 1.9535596130922595e-05, "loss": 0.1107, "step": 5430 }, { "epoch": 1.0, "learning_rate": 1.953537248148458e-05, "loss": 0.0977, "step": 5431 }, { "epoch": 1.0, "learning_rate": 1.9535148779487365e-05, "loss": 0.0982, "step": 5432 }, { "epoch": 1.0, "learning_rate": 1.9534925024932188e-05, "loss": 0.1078, "step": 5433 }, { "epoch": 1.0, "learning_rate": 1.953470121782028e-05, "loss": 0.1014, "step": 5434 }, { "epoch": 1.0, "learning_rate": 1.9534477358152874e-05, "loss": 0.1193, "step": 5435 }, { "epoch": 1.0, "learning_rate": 1.9534253445931207e-05, "loss": 0.1126, "step": 5436 }, { "epoch": 1.0, "learning_rate": 1.9534029481156512e-05, "loss": 0.1014, "step": 5437 }, { "epoch": 1.0, "learning_rate": 1.9533805463830024e-05, "loss": 0.1038, "step": 5438 }, { "epoch": 1.0, "learning_rate": 1.9533581393952978e-05, "loss": 0.1064, "step": 5439 }, { "epoch": 1.0, "learning_rate": 1.9533357271526604e-05, "loss": 0.1007, "step": 5440 }, { "epoch": 1.0, "learning_rate": 1.9533133096552142e-05, "loss": 0.1065, "step": 5441 }, { "epoch": 1.0, "learning_rate": 1.953290886903083e-05, "loss": 0.1089, "step": 5442 }, { "epoch": 1.0, "learning_rate": 1.95326845889639e-05, "loss": 0.1153, "step": 5443 }, { "epoch": 1.0, "learning_rate": 1.9532460256352587e-05, "loss": 0.1027, "step": 5444 }, { "epoch": 1.0, "learning_rate": 1.953223587119813e-05, "loss": 0.0894, "step": 5445 }, { "epoch": 1.0, "learning_rate": 1.9532011433501768e-05, "loss": 0.0958, "step": 5446 }, { "epoch": 1.0, "learning_rate": 1.9531786943264733e-05, "loss": 0.0938, "step": 5447 }, { "epoch": 1.0, "learning_rate": 1.9531562400488267e-05, "loss": 0.125, "step": 5448 }, { "epoch": 1.0, "learning_rate": 1.9531337805173604e-05, "loss": 0.0988, "step": 5449 }, { "epoch": 1.0, "learning_rate": 1.9531113157321986e-05, "loss": 0.1005, "step": 5450 }, { "epoch": 1.0, "learning_rate": 1.9530888456934648e-05, "loss": 0.098, "step": 5451 }, { "epoch": 1.0, "learning_rate": 1.953066370401283e-05, "loss": 0.1007, "step": 5452 }, { "epoch": 1.0, "learning_rate": 1.953043889855777e-05, "loss": 0.0954, "step": 5453 }, { "epoch": 1.0, "learning_rate": 1.9530214040570706e-05, "loss": 0.0974, "step": 5454 }, { "epoch": 1.0, "learning_rate": 1.952998913005288e-05, "loss": 0.089, "step": 5455 }, { "epoch": 1.0, "learning_rate": 1.9529764167005534e-05, "loss": 0.0991, "step": 5456 }, { "epoch": 1.0, "learning_rate": 1.95295391514299e-05, "loss": 0.1048, "step": 5457 }, { "epoch": 1.0, "learning_rate": 1.9529314083327223e-05, "loss": 0.1008, "step": 5458 }, { "epoch": 1.0, "learning_rate": 1.9529088962698748e-05, "loss": 0.0932, "step": 5459 }, { "epoch": 1.0, "learning_rate": 1.952886378954571e-05, "loss": 0.1011, "step": 5460 }, { "epoch": 1.0, "learning_rate": 1.9528638563869347e-05, "loss": 0.1137, "step": 5461 }, { "epoch": 1.0, "learning_rate": 1.9528413285670908e-05, "loss": 0.109, "step": 5462 }, { "epoch": 1.0, "learning_rate": 1.9528187954951635e-05, "loss": 0.1056, "step": 5463 }, { "epoch": 1.0, "learning_rate": 1.952796257171276e-05, "loss": 0.1071, "step": 5464 }, { "epoch": 1.0, "learning_rate": 1.9527737135955538e-05, "loss": 0.0995, "step": 5465 }, { "epoch": 1.0, "learning_rate": 1.9527511647681204e-05, "loss": 0.0985, "step": 5466 }, { "epoch": 1.0, "learning_rate": 1.9527286106891e-05, "loss": 0.1124, "step": 5467 }, { "epoch": 1.0, "learning_rate": 1.9527060513586175e-05, "loss": 0.1178, "step": 5468 }, { "epoch": 1.0, "learning_rate": 1.952683486776797e-05, "loss": 0.1066, "step": 5469 }, { "epoch": 1.0, "learning_rate": 1.9526609169437623e-05, "loss": 0.1076, "step": 5470 }, { "epoch": 1.0, "learning_rate": 1.9526383418596388e-05, "loss": 0.1148, "step": 5471 }, { "epoch": 1.0, "learning_rate": 1.95261576152455e-05, "loss": 0.111, "step": 5472 }, { "epoch": 1.0, "learning_rate": 1.9525931759386213e-05, "loss": 0.1069, "step": 5473 }, { "epoch": 1.0, "learning_rate": 1.9525705851019764e-05, "loss": 0.1057, "step": 5474 }, { "epoch": 1.0, "learning_rate": 1.95254798901474e-05, "loss": 0.1108, "step": 5475 }, { "epoch": 1.0, "learning_rate": 1.9525253876770372e-05, "loss": 0.0959, "step": 5476 }, { "epoch": 1.0, "learning_rate": 1.9525027810889917e-05, "loss": 0.1026, "step": 5477 }, { "epoch": 1.0, "learning_rate": 1.9524801692507286e-05, "loss": 0.1041, "step": 5478 }, { "epoch": 1.0, "learning_rate": 1.9524575521623727e-05, "loss": 0.1059, "step": 5479 }, { "epoch": 1.0, "learning_rate": 1.9524349298240487e-05, "loss": 0.1168, "step": 5480 }, { "epoch": 1.01, "learning_rate": 1.9524123022358808e-05, "loss": 0.0985, "step": 5481 }, { "epoch": 1.01, "learning_rate": 1.952389669397994e-05, "loss": 0.105, "step": 5482 }, { "epoch": 1.01, "learning_rate": 1.9523670313105128e-05, "loss": 0.0977, "step": 5483 }, { "epoch": 1.01, "learning_rate": 1.9523443879735623e-05, "loss": 0.0947, "step": 5484 }, { "epoch": 1.01, "learning_rate": 1.9523217393872675e-05, "loss": 0.0972, "step": 5485 }, { "epoch": 1.01, "learning_rate": 1.9522990855517526e-05, "loss": 0.1057, "step": 5486 }, { "epoch": 1.01, "learning_rate": 1.952276426467143e-05, "loss": 0.0945, "step": 5487 }, { "epoch": 1.01, "learning_rate": 1.9522537621335634e-05, "loss": 0.1032, "step": 5488 }, { "epoch": 1.01, "learning_rate": 1.952231092551139e-05, "loss": 0.1206, "step": 5489 }, { "epoch": 1.01, "learning_rate": 1.952208417719994e-05, "loss": 0.1084, "step": 5490 }, { "epoch": 1.01, "learning_rate": 1.952185737640254e-05, "loss": 0.093, "step": 5491 }, { "epoch": 1.01, "learning_rate": 1.952163052312044e-05, "loss": 0.0988, "step": 5492 }, { "epoch": 1.01, "learning_rate": 1.9521403617354892e-05, "loss": 0.0951, "step": 5493 }, { "epoch": 1.01, "learning_rate": 1.952117665910714e-05, "loss": 0.1053, "step": 5494 }, { "epoch": 1.01, "learning_rate": 1.9520949648378444e-05, "loss": 0.0958, "step": 5495 }, { "epoch": 1.01, "learning_rate": 1.952072258517005e-05, "loss": 0.1115, "step": 5496 }, { "epoch": 1.01, "learning_rate": 1.9520495469483207e-05, "loss": 0.0963, "step": 5497 }, { "epoch": 1.01, "learning_rate": 1.9520268301319173e-05, "loss": 0.1081, "step": 5498 }, { "epoch": 1.01, "learning_rate": 1.9520041080679192e-05, "loss": 0.1073, "step": 5499 }, { "epoch": 1.01, "learning_rate": 1.9519813807564527e-05, "loss": 0.1074, "step": 5500 }, { "epoch": 1.01, "learning_rate": 1.951958648197642e-05, "loss": 0.0943, "step": 5501 }, { "epoch": 1.01, "learning_rate": 1.9519359103916136e-05, "loss": 0.1032, "step": 5502 }, { "epoch": 1.01, "learning_rate": 1.9519131673384917e-05, "loss": 0.1035, "step": 5503 }, { "epoch": 1.01, "learning_rate": 1.951890419038402e-05, "loss": 0.1077, "step": 5504 }, { "epoch": 1.01, "learning_rate": 1.9518676654914702e-05, "loss": 0.1021, "step": 5505 }, { "epoch": 1.01, "learning_rate": 1.951844906697822e-05, "loss": 0.1058, "step": 5506 }, { "epoch": 1.01, "learning_rate": 1.9518221426575817e-05, "loss": 0.1069, "step": 5507 }, { "epoch": 1.01, "learning_rate": 1.9517993733708754e-05, "loss": 0.0968, "step": 5508 }, { "epoch": 1.01, "learning_rate": 1.951776598837829e-05, "loss": 0.0996, "step": 5509 }, { "epoch": 1.01, "learning_rate": 1.9517538190585677e-05, "loss": 0.1047, "step": 5510 }, { "epoch": 1.01, "learning_rate": 1.951731034033217e-05, "loss": 0.1036, "step": 5511 }, { "epoch": 1.01, "learning_rate": 1.9517082437619027e-05, "loss": 0.1045, "step": 5512 }, { "epoch": 1.01, "learning_rate": 1.9516854482447497e-05, "loss": 0.1007, "step": 5513 }, { "epoch": 1.01, "learning_rate": 1.951662647481885e-05, "loss": 0.1042, "step": 5514 }, { "epoch": 1.01, "learning_rate": 1.951639841473433e-05, "loss": 0.0856, "step": 5515 }, { "epoch": 1.01, "learning_rate": 1.95161703021952e-05, "loss": 0.0914, "step": 5516 }, { "epoch": 1.01, "learning_rate": 1.9515942137202714e-05, "loss": 0.1048, "step": 5517 }, { "epoch": 1.01, "learning_rate": 1.9515713919758138e-05, "loss": 0.1126, "step": 5518 }, { "epoch": 1.01, "learning_rate": 1.9515485649862717e-05, "loss": 0.1204, "step": 5519 }, { "epoch": 1.01, "learning_rate": 1.9515257327517722e-05, "loss": 0.1055, "step": 5520 }, { "epoch": 1.01, "learning_rate": 1.95150289527244e-05, "loss": 0.0956, "step": 5521 }, { "epoch": 1.01, "learning_rate": 1.951480052548402e-05, "loss": 0.0921, "step": 5522 }, { "epoch": 1.01, "learning_rate": 1.9514572045797836e-05, "loss": 0.1026, "step": 5523 }, { "epoch": 1.01, "learning_rate": 1.9514343513667103e-05, "loss": 0.1003, "step": 5524 }, { "epoch": 1.01, "learning_rate": 1.9514114929093094e-05, "loss": 0.0967, "step": 5525 }, { "epoch": 1.01, "learning_rate": 1.9513886292077053e-05, "loss": 0.1023, "step": 5526 }, { "epoch": 1.01, "learning_rate": 1.9513657602620254e-05, "loss": 0.1014, "step": 5527 }, { "epoch": 1.01, "learning_rate": 1.9513428860723946e-05, "loss": 0.1094, "step": 5528 }, { "epoch": 1.01, "learning_rate": 1.95132000663894e-05, "loss": 0.1101, "step": 5529 }, { "epoch": 1.01, "learning_rate": 1.9512971219617867e-05, "loss": 0.1016, "step": 5530 }, { "epoch": 1.01, "learning_rate": 1.951274232041062e-05, "loss": 0.0987, "step": 5531 }, { "epoch": 1.01, "learning_rate": 1.951251336876891e-05, "loss": 0.1072, "step": 5532 }, { "epoch": 1.01, "learning_rate": 1.951228436469401e-05, "loss": 0.0975, "step": 5533 }, { "epoch": 1.01, "learning_rate": 1.9512055308187167e-05, "loss": 0.114, "step": 5534 }, { "epoch": 1.02, "learning_rate": 1.9511826199249654e-05, "loss": 0.1061, "step": 5535 }, { "epoch": 1.02, "learning_rate": 1.9511597037882735e-05, "loss": 0.0958, "step": 5536 }, { "epoch": 1.02, "learning_rate": 1.9511367824087672e-05, "loss": 0.1003, "step": 5537 }, { "epoch": 1.02, "learning_rate": 1.9511138557865722e-05, "loss": 0.0924, "step": 5538 }, { "epoch": 1.02, "learning_rate": 1.9510909239218155e-05, "loss": 0.1054, "step": 5539 }, { "epoch": 1.02, "learning_rate": 1.9510679868146235e-05, "loss": 0.0902, "step": 5540 }, { "epoch": 1.02, "learning_rate": 1.9510450444651225e-05, "loss": 0.1067, "step": 5541 }, { "epoch": 1.02, "learning_rate": 1.9510220968734387e-05, "loss": 0.0928, "step": 5542 }, { "epoch": 1.02, "learning_rate": 1.950999144039699e-05, "loss": 0.0919, "step": 5543 }, { "epoch": 1.02, "learning_rate": 1.9509761859640298e-05, "loss": 0.0983, "step": 5544 }, { "epoch": 1.02, "learning_rate": 1.9509532226465576e-05, "loss": 0.095, "step": 5545 }, { "epoch": 1.02, "learning_rate": 1.9509302540874086e-05, "loss": 0.1114, "step": 5546 }, { "epoch": 1.02, "learning_rate": 1.95090728028671e-05, "loss": 0.1123, "step": 5547 }, { "epoch": 1.02, "learning_rate": 1.950884301244588e-05, "loss": 0.0841, "step": 5548 }, { "epoch": 1.02, "learning_rate": 1.9508613169611697e-05, "loss": 0.1065, "step": 5549 }, { "epoch": 1.02, "learning_rate": 1.9508383274365818e-05, "loss": 0.0969, "step": 5550 }, { "epoch": 1.02, "learning_rate": 1.95081533267095e-05, "loss": 0.1052, "step": 5551 }, { "epoch": 1.02, "learning_rate": 1.9507923326644022e-05, "loss": 0.1012, "step": 5552 }, { "epoch": 1.02, "learning_rate": 1.950769327417065e-05, "loss": 0.0936, "step": 5553 }, { "epoch": 1.02, "learning_rate": 1.950746316929065e-05, "loss": 0.1108, "step": 5554 }, { "epoch": 1.02, "learning_rate": 1.9507233012005288e-05, "loss": 0.1082, "step": 5555 }, { "epoch": 1.02, "learning_rate": 1.9507002802315835e-05, "loss": 0.1112, "step": 5556 }, { "epoch": 1.02, "learning_rate": 1.950677254022356e-05, "loss": 0.1104, "step": 5557 }, { "epoch": 1.02, "learning_rate": 1.9506542225729732e-05, "loss": 0.103, "step": 5558 }, { "epoch": 1.02, "learning_rate": 1.950631185883562e-05, "loss": 0.108, "step": 5559 }, { "epoch": 1.02, "learning_rate": 1.9506081439542497e-05, "loss": 0.103, "step": 5560 }, { "epoch": 1.02, "learning_rate": 1.950585096785163e-05, "loss": 0.0969, "step": 5561 }, { "epoch": 1.02, "learning_rate": 1.9505620443764284e-05, "loss": 0.12, "step": 5562 }, { "epoch": 1.02, "learning_rate": 1.9505389867281742e-05, "loss": 0.1041, "step": 5563 }, { "epoch": 1.02, "learning_rate": 1.9505159238405265e-05, "loss": 0.1022, "step": 5564 }, { "epoch": 1.02, "learning_rate": 1.950492855713613e-05, "loss": 0.1136, "step": 5565 }, { "epoch": 1.02, "learning_rate": 1.9504697823475604e-05, "loss": 0.1071, "step": 5566 }, { "epoch": 1.02, "learning_rate": 1.9504467037424964e-05, "loss": 0.0989, "step": 5567 }, { "epoch": 1.02, "learning_rate": 1.9504236198985475e-05, "loss": 0.1, "step": 5568 }, { "epoch": 1.02, "learning_rate": 1.9504005308158415e-05, "loss": 0.1037, "step": 5569 }, { "epoch": 1.02, "learning_rate": 1.9503774364945057e-05, "loss": 0.0947, "step": 5570 }, { "epoch": 1.02, "learning_rate": 1.950354336934667e-05, "loss": 0.1134, "step": 5571 }, { "epoch": 1.02, "learning_rate": 1.950331232136453e-05, "loss": 0.1057, "step": 5572 }, { "epoch": 1.02, "learning_rate": 1.950308122099991e-05, "loss": 0.1101, "step": 5573 }, { "epoch": 1.02, "learning_rate": 1.9502850068254083e-05, "loss": 0.1036, "step": 5574 }, { "epoch": 1.02, "learning_rate": 1.9502618863128323e-05, "loss": 0.1059, "step": 5575 }, { "epoch": 1.02, "learning_rate": 1.9502387605623908e-05, "loss": 0.1113, "step": 5576 }, { "epoch": 1.02, "learning_rate": 1.9502156295742107e-05, "loss": 0.1011, "step": 5577 }, { "epoch": 1.02, "learning_rate": 1.95019249334842e-05, "loss": 0.1034, "step": 5578 }, { "epoch": 1.02, "learning_rate": 1.950169351885146e-05, "loss": 0.0984, "step": 5579 }, { "epoch": 1.02, "learning_rate": 1.9501462051845158e-05, "loss": 0.1073, "step": 5580 }, { "epoch": 1.02, "learning_rate": 1.950123053246658e-05, "loss": 0.1194, "step": 5581 }, { "epoch": 1.02, "learning_rate": 1.9500998960716996e-05, "loss": 0.0811, "step": 5582 }, { "epoch": 1.02, "learning_rate": 1.950076733659768e-05, "loss": 0.1097, "step": 5583 }, { "epoch": 1.02, "learning_rate": 1.9500535660109916e-05, "loss": 0.1085, "step": 5584 }, { "epoch": 1.02, "learning_rate": 1.9500303931254977e-05, "loss": 0.1047, "step": 5585 }, { "epoch": 1.02, "learning_rate": 1.9500072150034136e-05, "loss": 0.0993, "step": 5586 }, { "epoch": 1.02, "learning_rate": 1.9499840316448675e-05, "loss": 0.1004, "step": 5587 }, { "epoch": 1.02, "learning_rate": 1.9499608430499875e-05, "loss": 0.0838, "step": 5588 }, { "epoch": 1.02, "learning_rate": 1.949937649218901e-05, "loss": 0.1049, "step": 5589 }, { "epoch": 1.03, "learning_rate": 1.9499144501517357e-05, "loss": 0.1071, "step": 5590 }, { "epoch": 1.03, "learning_rate": 1.94989124584862e-05, "loss": 0.1034, "step": 5591 }, { "epoch": 1.03, "learning_rate": 1.9498680363096813e-05, "loss": 0.0884, "step": 5592 }, { "epoch": 1.03, "learning_rate": 1.9498448215350477e-05, "loss": 0.0869, "step": 5593 }, { "epoch": 1.03, "learning_rate": 1.9498216015248473e-05, "loss": 0.096, "step": 5594 }, { "epoch": 1.03, "learning_rate": 1.9497983762792077e-05, "loss": 0.1009, "step": 5595 }, { "epoch": 1.03, "learning_rate": 1.9497751457982577e-05, "loss": 0.1041, "step": 5596 }, { "epoch": 1.03, "learning_rate": 1.9497519100821244e-05, "loss": 0.1038, "step": 5597 }, { "epoch": 1.03, "learning_rate": 1.9497286691309368e-05, "loss": 0.0993, "step": 5598 }, { "epoch": 1.03, "learning_rate": 1.9497054229448223e-05, "loss": 0.1104, "step": 5599 }, { "epoch": 1.03, "learning_rate": 1.9496821715239093e-05, "loss": 0.0895, "step": 5600 }, { "epoch": 1.03, "learning_rate": 1.9496589148683256e-05, "loss": 0.1269, "step": 5601 }, { "epoch": 1.03, "learning_rate": 1.9496356529782e-05, "loss": 0.0938, "step": 5602 }, { "epoch": 1.03, "learning_rate": 1.9496123858536604e-05, "loss": 0.1069, "step": 5603 }, { "epoch": 1.03, "learning_rate": 1.949589113494835e-05, "loss": 0.115, "step": 5604 }, { "epoch": 1.03, "learning_rate": 1.9495658359018522e-05, "loss": 0.1007, "step": 5605 }, { "epoch": 1.03, "learning_rate": 1.9495425530748403e-05, "loss": 0.1038, "step": 5606 }, { "epoch": 1.03, "learning_rate": 1.9495192650139275e-05, "loss": 0.0966, "step": 5607 }, { "epoch": 1.03, "learning_rate": 1.9494959717192422e-05, "loss": 0.1165, "step": 5608 }, { "epoch": 1.03, "learning_rate": 1.949472673190913e-05, "loss": 0.0982, "step": 5609 }, { "epoch": 1.03, "learning_rate": 1.9494493694290683e-05, "loss": 0.1044, "step": 5610 }, { "epoch": 1.03, "learning_rate": 1.9494260604338365e-05, "loss": 0.1069, "step": 5611 }, { "epoch": 1.03, "learning_rate": 1.9494027462053458e-05, "loss": 0.1036, "step": 5612 }, { "epoch": 1.03, "learning_rate": 1.949379426743725e-05, "loss": 0.1071, "step": 5613 }, { "epoch": 1.03, "learning_rate": 1.9493561020491024e-05, "loss": 0.1079, "step": 5614 }, { "epoch": 1.03, "learning_rate": 1.9493327721216068e-05, "loss": 0.1062, "step": 5615 }, { "epoch": 1.03, "learning_rate": 1.9493094369613667e-05, "loss": 0.0971, "step": 5616 }, { "epoch": 1.03, "learning_rate": 1.949286096568511e-05, "loss": 0.0962, "step": 5617 }, { "epoch": 1.03, "learning_rate": 1.949262750943168e-05, "loss": 0.1046, "step": 5618 }, { "epoch": 1.03, "learning_rate": 1.9492394000854663e-05, "loss": 0.1071, "step": 5619 }, { "epoch": 1.03, "learning_rate": 1.9492160439955347e-05, "loss": 0.1045, "step": 5620 }, { "epoch": 1.03, "learning_rate": 1.9491926826735022e-05, "loss": 0.1101, "step": 5621 }, { "epoch": 1.03, "learning_rate": 1.9491693161194977e-05, "loss": 0.1058, "step": 5622 }, { "epoch": 1.03, "learning_rate": 1.9491459443336495e-05, "loss": 0.1128, "step": 5623 }, { "epoch": 1.03, "learning_rate": 1.9491225673160864e-05, "loss": 0.0974, "step": 5624 }, { "epoch": 1.03, "learning_rate": 1.9490991850669374e-05, "loss": 0.0949, "step": 5625 }, { "epoch": 1.03, "learning_rate": 1.9490757975863313e-05, "loss": 0.1034, "step": 5626 }, { "epoch": 1.03, "learning_rate": 1.9490524048743976e-05, "loss": 0.0994, "step": 5627 }, { "epoch": 1.03, "learning_rate": 1.949029006931265e-05, "loss": 0.1085, "step": 5628 }, { "epoch": 1.03, "learning_rate": 1.949005603757062e-05, "loss": 0.1087, "step": 5629 }, { "epoch": 1.03, "learning_rate": 1.948982195351918e-05, "loss": 0.1169, "step": 5630 }, { "epoch": 1.03, "learning_rate": 1.9489587817159615e-05, "loss": 0.1088, "step": 5631 }, { "epoch": 1.03, "learning_rate": 1.9489353628493224e-05, "loss": 0.0963, "step": 5632 }, { "epoch": 1.03, "learning_rate": 1.9489119387521288e-05, "loss": 0.0964, "step": 5633 }, { "epoch": 1.03, "learning_rate": 1.948888509424511e-05, "loss": 0.0894, "step": 5634 }, { "epoch": 1.03, "learning_rate": 1.9488650748665973e-05, "loss": 0.1098, "step": 5635 }, { "epoch": 1.03, "learning_rate": 1.9488416350785168e-05, "loss": 0.1025, "step": 5636 }, { "epoch": 1.03, "learning_rate": 1.9488181900603992e-05, "loss": 0.1077, "step": 5637 }, { "epoch": 1.03, "learning_rate": 1.9487947398123736e-05, "loss": 0.1203, "step": 5638 }, { "epoch": 1.03, "learning_rate": 1.9487712843345694e-05, "loss": 0.1014, "step": 5639 }, { "epoch": 1.03, "learning_rate": 1.9487478236271153e-05, "loss": 0.1106, "step": 5640 }, { "epoch": 1.03, "learning_rate": 1.9487243576901407e-05, "loss": 0.0934, "step": 5641 }, { "epoch": 1.03, "learning_rate": 1.948700886523776e-05, "loss": 0.1137, "step": 5642 }, { "epoch": 1.03, "learning_rate": 1.9486774101281488e-05, "loss": 0.1062, "step": 5643 }, { "epoch": 1.04, "learning_rate": 1.9486539285033902e-05, "loss": 0.0926, "step": 5644 }, { "epoch": 1.04, "learning_rate": 1.9486304416496286e-05, "loss": 0.1161, "step": 5645 }, { "epoch": 1.04, "learning_rate": 1.948606949566994e-05, "loss": 0.1007, "step": 5646 }, { "epoch": 1.04, "learning_rate": 1.9485834522556155e-05, "loss": 0.1038, "step": 5647 }, { "epoch": 1.04, "learning_rate": 1.948559949715623e-05, "loss": 0.1096, "step": 5648 }, { "epoch": 1.04, "learning_rate": 1.9485364419471454e-05, "loss": 0.1096, "step": 5649 }, { "epoch": 1.04, "learning_rate": 1.9485129289503132e-05, "loss": 0.1031, "step": 5650 }, { "epoch": 1.04, "learning_rate": 1.9484894107252556e-05, "loss": 0.1039, "step": 5651 }, { "epoch": 1.04, "learning_rate": 1.948465887272102e-05, "loss": 0.1108, "step": 5652 }, { "epoch": 1.04, "learning_rate": 1.948442358590982e-05, "loss": 0.1056, "step": 5653 }, { "epoch": 1.04, "learning_rate": 1.9484188246820254e-05, "loss": 0.1041, "step": 5654 }, { "epoch": 1.04, "learning_rate": 1.9483952855453624e-05, "loss": 0.1022, "step": 5655 }, { "epoch": 1.04, "learning_rate": 1.9483717411811218e-05, "loss": 0.1001, "step": 5656 }, { "epoch": 1.04, "learning_rate": 1.9483481915894346e-05, "loss": 0.0996, "step": 5657 }, { "epoch": 1.04, "learning_rate": 1.9483246367704293e-05, "loss": 0.1018, "step": 5658 }, { "epoch": 1.04, "learning_rate": 1.948301076724237e-05, "loss": 0.1137, "step": 5659 }, { "epoch": 1.04, "learning_rate": 1.9482775114509867e-05, "loss": 0.1022, "step": 5660 }, { "epoch": 1.04, "learning_rate": 1.9482539409508082e-05, "loss": 0.113, "step": 5661 }, { "epoch": 1.04, "learning_rate": 1.9482303652238324e-05, "loss": 0.1015, "step": 5662 }, { "epoch": 1.04, "learning_rate": 1.9482067842701882e-05, "loss": 0.0976, "step": 5663 }, { "epoch": 1.04, "learning_rate": 1.9481831980900062e-05, "loss": 0.1129, "step": 5664 }, { "epoch": 1.04, "learning_rate": 1.948159606683416e-05, "loss": 0.0916, "step": 5665 }, { "epoch": 1.04, "learning_rate": 1.948136010050548e-05, "loss": 0.0932, "step": 5666 }, { "epoch": 1.04, "learning_rate": 1.9481124081915317e-05, "loss": 0.0963, "step": 5667 }, { "epoch": 1.04, "learning_rate": 1.9480888011064983e-05, "loss": 0.1235, "step": 5668 }, { "epoch": 1.04, "learning_rate": 1.948065188795577e-05, "loss": 0.1114, "step": 5669 }, { "epoch": 1.04, "learning_rate": 1.948041571258898e-05, "loss": 0.0979, "step": 5670 }, { "epoch": 1.04, "learning_rate": 1.9480179484965915e-05, "loss": 0.1073, "step": 5671 }, { "epoch": 1.04, "learning_rate": 1.947994320508788e-05, "loss": 0.1145, "step": 5672 }, { "epoch": 1.04, "learning_rate": 1.9479706872956176e-05, "loss": 0.1033, "step": 5673 }, { "epoch": 1.04, "learning_rate": 1.9479470488572104e-05, "loss": 0.108, "step": 5674 }, { "epoch": 1.04, "learning_rate": 1.947923405193697e-05, "loss": 0.1128, "step": 5675 }, { "epoch": 1.04, "learning_rate": 1.947899756305208e-05, "loss": 0.1137, "step": 5676 }, { "epoch": 1.04, "learning_rate": 1.947876102191873e-05, "loss": 0.1027, "step": 5677 }, { "epoch": 1.04, "learning_rate": 1.9478524428538226e-05, "loss": 0.1051, "step": 5678 }, { "epoch": 1.04, "learning_rate": 1.9478287782911873e-05, "loss": 0.0908, "step": 5679 }, { "epoch": 1.04, "learning_rate": 1.9478051085040978e-05, "loss": 0.106, "step": 5680 }, { "epoch": 1.04, "learning_rate": 1.9477814334926843e-05, "loss": 0.1035, "step": 5681 }, { "epoch": 1.04, "learning_rate": 1.947757753257077e-05, "loss": 0.097, "step": 5682 }, { "epoch": 1.04, "learning_rate": 1.9477340677974072e-05, "loss": 0.1191, "step": 5683 }, { "epoch": 1.04, "learning_rate": 1.9477103771138045e-05, "loss": 0.116, "step": 5684 }, { "epoch": 1.04, "learning_rate": 1.9476866812064005e-05, "loss": 0.1007, "step": 5685 }, { "epoch": 1.04, "learning_rate": 1.9476629800753254e-05, "loss": 0.0984, "step": 5686 }, { "epoch": 1.04, "learning_rate": 1.9476392737207093e-05, "loss": 0.0902, "step": 5687 }, { "epoch": 1.04, "learning_rate": 1.9476155621426835e-05, "loss": 0.1148, "step": 5688 }, { "epoch": 1.04, "learning_rate": 1.9475918453413786e-05, "loss": 0.109, "step": 5689 }, { "epoch": 1.04, "learning_rate": 1.947568123316925e-05, "loss": 0.0871, "step": 5690 }, { "epoch": 1.04, "learning_rate": 1.947544396069454e-05, "loss": 0.1188, "step": 5691 }, { "epoch": 1.04, "learning_rate": 1.947520663599096e-05, "loss": 0.0929, "step": 5692 }, { "epoch": 1.04, "learning_rate": 1.947496925905982e-05, "loss": 0.0984, "step": 5693 }, { "epoch": 1.04, "learning_rate": 1.9474731829902424e-05, "loss": 0.1109, "step": 5694 }, { "epoch": 1.04, "learning_rate": 1.9474494348520087e-05, "loss": 0.1017, "step": 5695 }, { "epoch": 1.04, "learning_rate": 1.9474256814914116e-05, "loss": 0.1023, "step": 5696 }, { "epoch": 1.04, "learning_rate": 1.9474019229085816e-05, "loss": 0.1088, "step": 5697 }, { "epoch": 1.04, "learning_rate": 1.94737815910365e-05, "loss": 0.1008, "step": 5698 }, { "epoch": 1.05, "learning_rate": 1.947354390076748e-05, "loss": 0.0935, "step": 5699 }, { "epoch": 1.05, "learning_rate": 1.9473306158280065e-05, "loss": 0.1069, "step": 5700 }, { "epoch": 1.05, "learning_rate": 1.9473068363575564e-05, "loss": 0.1091, "step": 5701 }, { "epoch": 1.05, "learning_rate": 1.947283051665529e-05, "loss": 0.0945, "step": 5702 }, { "epoch": 1.05, "learning_rate": 1.947259261752055e-05, "loss": 0.0992, "step": 5703 }, { "epoch": 1.05, "learning_rate": 1.9472354666172657e-05, "loss": 0.0983, "step": 5704 }, { "epoch": 1.05, "learning_rate": 1.9472116662612922e-05, "loss": 0.1064, "step": 5705 }, { "epoch": 1.05, "learning_rate": 1.9471878606842663e-05, "loss": 0.1006, "step": 5706 }, { "epoch": 1.05, "learning_rate": 1.9471640498863183e-05, "loss": 0.0926, "step": 5707 }, { "epoch": 1.05, "learning_rate": 1.9471402338675804e-05, "loss": 0.0972, "step": 5708 }, { "epoch": 1.05, "learning_rate": 1.9471164126281827e-05, "loss": 0.1031, "step": 5709 }, { "epoch": 1.05, "learning_rate": 1.9470925861682574e-05, "loss": 0.0944, "step": 5710 }, { "epoch": 1.05, "learning_rate": 1.9470687544879355e-05, "loss": 0.0979, "step": 5711 }, { "epoch": 1.05, "learning_rate": 1.9470449175873486e-05, "loss": 0.1005, "step": 5712 }, { "epoch": 1.05, "learning_rate": 1.9470210754666277e-05, "loss": 0.0893, "step": 5713 }, { "epoch": 1.05, "learning_rate": 1.9469972281259047e-05, "loss": 0.1053, "step": 5714 }, { "epoch": 1.05, "learning_rate": 1.9469733755653103e-05, "loss": 0.1195, "step": 5715 }, { "epoch": 1.05, "learning_rate": 1.946949517784977e-05, "loss": 0.0946, "step": 5716 }, { "epoch": 1.05, "learning_rate": 1.9469256547850356e-05, "loss": 0.099, "step": 5717 }, { "epoch": 1.05, "learning_rate": 1.9469017865656175e-05, "loss": 0.0973, "step": 5718 }, { "epoch": 1.05, "learning_rate": 1.9468779131268553e-05, "loss": 0.118, "step": 5719 }, { "epoch": 1.05, "learning_rate": 1.946854034468879e-05, "loss": 0.1004, "step": 5720 }, { "epoch": 1.05, "learning_rate": 1.9468301505918218e-05, "loss": 0.0981, "step": 5721 }, { "epoch": 1.05, "learning_rate": 1.9468062614958144e-05, "loss": 0.1033, "step": 5722 }, { "epoch": 1.05, "learning_rate": 1.9467823671809887e-05, "loss": 0.11, "step": 5723 }, { "epoch": 1.05, "learning_rate": 1.946758467647476e-05, "loss": 0.1101, "step": 5724 }, { "epoch": 1.05, "learning_rate": 1.946734562895409e-05, "loss": 0.1063, "step": 5725 }, { "epoch": 1.05, "learning_rate": 1.9467106529249186e-05, "loss": 0.1019, "step": 5726 }, { "epoch": 1.05, "learning_rate": 1.9466867377361372e-05, "loss": 0.0948, "step": 5727 }, { "epoch": 1.05, "learning_rate": 1.946662817329196e-05, "loss": 0.0887, "step": 5728 }, { "epoch": 1.05, "learning_rate": 1.9466388917042275e-05, "loss": 0.1062, "step": 5729 }, { "epoch": 1.05, "learning_rate": 1.946614960861363e-05, "loss": 0.1107, "step": 5730 }, { "epoch": 1.05, "learning_rate": 1.9465910248007347e-05, "loss": 0.111, "step": 5731 }, { "epoch": 1.05, "learning_rate": 1.9465670835224742e-05, "loss": 0.0959, "step": 5732 }, { "epoch": 1.05, "learning_rate": 1.9465431370267144e-05, "loss": 0.1001, "step": 5733 }, { "epoch": 1.05, "learning_rate": 1.9465191853135864e-05, "loss": 0.1001, "step": 5734 }, { "epoch": 1.05, "learning_rate": 1.9464952283832222e-05, "loss": 0.1127, "step": 5735 }, { "epoch": 1.05, "learning_rate": 1.946471266235754e-05, "loss": 0.1114, "step": 5736 }, { "epoch": 1.05, "learning_rate": 1.9464472988713146e-05, "loss": 0.0908, "step": 5737 }, { "epoch": 1.05, "learning_rate": 1.9464233262900354e-05, "loss": 0.0995, "step": 5738 }, { "epoch": 1.05, "learning_rate": 1.9463993484920483e-05, "loss": 0.1006, "step": 5739 }, { "epoch": 1.05, "learning_rate": 1.9463753654774865e-05, "loss": 0.079, "step": 5740 }, { "epoch": 1.05, "learning_rate": 1.9463513772464808e-05, "loss": 0.0984, "step": 5741 }, { "epoch": 1.05, "learning_rate": 1.9463273837991643e-05, "loss": 0.1124, "step": 5742 }, { "epoch": 1.05, "learning_rate": 1.946303385135669e-05, "loss": 0.1007, "step": 5743 }, { "epoch": 1.05, "learning_rate": 1.9462793812561278e-05, "loss": 0.105, "step": 5744 }, { "epoch": 1.05, "learning_rate": 1.946255372160672e-05, "loss": 0.0945, "step": 5745 }, { "epoch": 1.05, "learning_rate": 1.9462313578494345e-05, "loss": 0.0917, "step": 5746 }, { "epoch": 1.05, "learning_rate": 1.946207338322548e-05, "loss": 0.1104, "step": 5747 }, { "epoch": 1.05, "learning_rate": 1.946183313580144e-05, "loss": 0.104, "step": 5748 }, { "epoch": 1.05, "learning_rate": 1.946159283622355e-05, "loss": 0.0996, "step": 5749 }, { "epoch": 1.05, "learning_rate": 1.9461352484493144e-05, "loss": 0.1045, "step": 5750 }, { "epoch": 1.05, "learning_rate": 1.9461112080611545e-05, "loss": 0.0959, "step": 5751 }, { "epoch": 1.05, "learning_rate": 1.9460871624580072e-05, "loss": 0.1141, "step": 5752 }, { "epoch": 1.06, "learning_rate": 1.946063111640005e-05, "loss": 0.1024, "step": 5753 }, { "epoch": 1.06, "learning_rate": 1.946039055607281e-05, "loss": 0.1047, "step": 5754 }, { "epoch": 1.06, "learning_rate": 1.9460149943599675e-05, "loss": 0.0978, "step": 5755 }, { "epoch": 1.06, "learning_rate": 1.945990927898197e-05, "loss": 0.1082, "step": 5756 }, { "epoch": 1.06, "learning_rate": 1.9459668562221024e-05, "loss": 0.1152, "step": 5757 }, { "epoch": 1.06, "learning_rate": 1.9459427793318166e-05, "loss": 0.1005, "step": 5758 }, { "epoch": 1.06, "learning_rate": 1.945918697227472e-05, "loss": 0.1148, "step": 5759 }, { "epoch": 1.06, "learning_rate": 1.9458946099092013e-05, "loss": 0.0982, "step": 5760 }, { "epoch": 1.06, "learning_rate": 1.9458705173771375e-05, "loss": 0.0945, "step": 5761 }, { "epoch": 1.06, "learning_rate": 1.9458464196314128e-05, "loss": 0.1142, "step": 5762 }, { "epoch": 1.06, "learning_rate": 1.945822316672161e-05, "loss": 0.0993, "step": 5763 }, { "epoch": 1.06, "learning_rate": 1.945798208499514e-05, "loss": 0.0992, "step": 5764 }, { "epoch": 1.06, "learning_rate": 1.9457740951136056e-05, "loss": 0.099, "step": 5765 }, { "epoch": 1.06, "learning_rate": 1.9457499765145676e-05, "loss": 0.0942, "step": 5766 }, { "epoch": 1.06, "learning_rate": 1.945725852702534e-05, "loss": 0.0999, "step": 5767 }, { "epoch": 1.06, "learning_rate": 1.945701723677637e-05, "loss": 0.1072, "step": 5768 }, { "epoch": 1.06, "learning_rate": 1.9456775894400105e-05, "loss": 0.1007, "step": 5769 }, { "epoch": 1.06, "learning_rate": 1.945653449989787e-05, "loss": 0.106, "step": 5770 }, { "epoch": 1.06, "learning_rate": 1.945629305327099e-05, "loss": 0.1108, "step": 5771 }, { "epoch": 1.06, "learning_rate": 1.9456051554520803e-05, "loss": 0.0901, "step": 5772 }, { "epoch": 1.06, "learning_rate": 1.945581000364864e-05, "loss": 0.1013, "step": 5773 }, { "epoch": 1.06, "learning_rate": 1.945556840065583e-05, "loss": 0.1031, "step": 5774 }, { "epoch": 1.06, "learning_rate": 1.9455326745543703e-05, "loss": 0.0979, "step": 5775 }, { "epoch": 1.06, "learning_rate": 1.9455085038313596e-05, "loss": 0.1095, "step": 5776 }, { "epoch": 1.06, "learning_rate": 1.9454843278966838e-05, "loss": 0.0949, "step": 5777 }, { "epoch": 1.06, "learning_rate": 1.9454601467504763e-05, "loss": 0.1053, "step": 5778 }, { "epoch": 1.06, "learning_rate": 1.94543596039287e-05, "loss": 0.1129, "step": 5779 }, { "epoch": 1.06, "learning_rate": 1.945411768823999e-05, "loss": 0.0925, "step": 5780 }, { "epoch": 1.06, "learning_rate": 1.9453875720439955e-05, "loss": 0.0926, "step": 5781 }, { "epoch": 1.06, "learning_rate": 1.9453633700529942e-05, "loss": 0.0983, "step": 5782 }, { "epoch": 1.06, "learning_rate": 1.945339162851128e-05, "loss": 0.0988, "step": 5783 }, { "epoch": 1.06, "learning_rate": 1.9453149504385297e-05, "loss": 0.105, "step": 5784 }, { "epoch": 1.06, "learning_rate": 1.9452907328153334e-05, "loss": 0.0962, "step": 5785 }, { "epoch": 1.06, "learning_rate": 1.9452665099816723e-05, "loss": 0.1052, "step": 5786 }, { "epoch": 1.06, "learning_rate": 1.9452422819376804e-05, "loss": 0.1073, "step": 5787 }, { "epoch": 1.06, "learning_rate": 1.9452180486834906e-05, "loss": 0.0965, "step": 5788 }, { "epoch": 1.06, "learning_rate": 1.945193810219237e-05, "loss": 0.094, "step": 5789 }, { "epoch": 1.06, "learning_rate": 1.945169566545053e-05, "loss": 0.1002, "step": 5790 }, { "epoch": 1.06, "learning_rate": 1.9451453176610716e-05, "loss": 0.0925, "step": 5791 }, { "epoch": 1.06, "learning_rate": 1.9451210635674278e-05, "loss": 0.1023, "step": 5792 }, { "epoch": 1.06, "learning_rate": 1.945096804264254e-05, "loss": 0.1065, "step": 5793 }, { "epoch": 1.06, "learning_rate": 1.945072539751685e-05, "loss": 0.1097, "step": 5794 }, { "epoch": 1.06, "learning_rate": 1.9450482700298537e-05, "loss": 0.102, "step": 5795 }, { "epoch": 1.06, "learning_rate": 1.945023995098894e-05, "loss": 0.1018, "step": 5796 }, { "epoch": 1.06, "learning_rate": 1.9449997149589403e-05, "loss": 0.0963, "step": 5797 }, { "epoch": 1.06, "learning_rate": 1.9449754296101257e-05, "loss": 0.0932, "step": 5798 }, { "epoch": 1.06, "learning_rate": 1.944951139052585e-05, "loss": 0.1023, "step": 5799 }, { "epoch": 1.06, "learning_rate": 1.9449268432864508e-05, "loss": 0.107, "step": 5800 }, { "epoch": 1.06, "learning_rate": 1.9449025423118583e-05, "loss": 0.0922, "step": 5801 }, { "epoch": 1.06, "learning_rate": 1.94487823612894e-05, "loss": 0.1062, "step": 5802 }, { "epoch": 1.06, "learning_rate": 1.9448539247378315e-05, "loss": 0.1005, "step": 5803 }, { "epoch": 1.06, "learning_rate": 1.9448296081386656e-05, "loss": 0.112, "step": 5804 }, { "epoch": 1.06, "learning_rate": 1.9448052863315773e-05, "loss": 0.0901, "step": 5805 }, { "epoch": 1.06, "learning_rate": 1.9447809593166997e-05, "loss": 0.1114, "step": 5806 }, { "epoch": 1.06, "learning_rate": 1.9447566270941673e-05, "loss": 0.107, "step": 5807 }, { "epoch": 1.07, "learning_rate": 1.9447322896641142e-05, "loss": 0.1031, "step": 5808 }, { "epoch": 1.07, "learning_rate": 1.9447079470266748e-05, "loss": 0.0904, "step": 5809 }, { "epoch": 1.07, "learning_rate": 1.944683599181983e-05, "loss": 0.0945, "step": 5810 }, { "epoch": 1.07, "learning_rate": 1.944659246130173e-05, "loss": 0.1092, "step": 5811 }, { "epoch": 1.07, "learning_rate": 1.944634887871379e-05, "loss": 0.0963, "step": 5812 }, { "epoch": 1.07, "learning_rate": 1.944610524405736e-05, "loss": 0.1055, "step": 5813 }, { "epoch": 1.07, "learning_rate": 1.944586155733377e-05, "loss": 0.1102, "step": 5814 }, { "epoch": 1.07, "learning_rate": 1.9445617818544372e-05, "loss": 0.1036, "step": 5815 }, { "epoch": 1.07, "learning_rate": 1.9445374027690507e-05, "loss": 0.0991, "step": 5816 }, { "epoch": 1.07, "learning_rate": 1.944513018477352e-05, "loss": 0.1095, "step": 5817 }, { "epoch": 1.07, "learning_rate": 1.944488628979475e-05, "loss": 0.1126, "step": 5818 }, { "epoch": 1.07, "learning_rate": 1.944464234275555e-05, "loss": 0.101, "step": 5819 }, { "epoch": 1.07, "learning_rate": 1.944439834365726e-05, "loss": 0.0975, "step": 5820 }, { "epoch": 1.07, "learning_rate": 1.944415429250122e-05, "loss": 0.1065, "step": 5821 }, { "epoch": 1.07, "learning_rate": 1.9443910189288786e-05, "loss": 0.1135, "step": 5822 }, { "epoch": 1.07, "learning_rate": 1.9443666034021297e-05, "loss": 0.1044, "step": 5823 }, { "epoch": 1.07, "learning_rate": 1.9443421826700096e-05, "loss": 0.1025, "step": 5824 }, { "epoch": 1.07, "learning_rate": 1.9443177567326538e-05, "loss": 0.0977, "step": 5825 }, { "epoch": 1.07, "learning_rate": 1.944293325590196e-05, "loss": 0.1039, "step": 5826 }, { "epoch": 1.07, "learning_rate": 1.9442688892427713e-05, "loss": 0.1095, "step": 5827 }, { "epoch": 1.07, "learning_rate": 1.9442444476905146e-05, "loss": 0.0937, "step": 5828 }, { "epoch": 1.07, "learning_rate": 1.9442200009335602e-05, "loss": 0.116, "step": 5829 }, { "epoch": 1.07, "learning_rate": 1.944195548972043e-05, "loss": 0.1004, "step": 5830 }, { "epoch": 1.07, "learning_rate": 1.9441710918060978e-05, "loss": 0.0969, "step": 5831 }, { "epoch": 1.07, "learning_rate": 1.9441466294358597e-05, "loss": 0.0939, "step": 5832 }, { "epoch": 1.07, "learning_rate": 1.9441221618614626e-05, "loss": 0.1048, "step": 5833 }, { "epoch": 1.07, "learning_rate": 1.944097689083043e-05, "loss": 0.0938, "step": 5834 }, { "epoch": 1.07, "learning_rate": 1.944073211100734e-05, "loss": 0.1075, "step": 5835 }, { "epoch": 1.07, "learning_rate": 1.9440487279146717e-05, "loss": 0.1017, "step": 5836 }, { "epoch": 1.07, "learning_rate": 1.9440242395249906e-05, "loss": 0.1007, "step": 5837 }, { "epoch": 1.07, "learning_rate": 1.9439997459318258e-05, "loss": 0.0943, "step": 5838 }, { "epoch": 1.07, "learning_rate": 1.9439752471353125e-05, "loss": 0.0928, "step": 5839 }, { "epoch": 1.07, "learning_rate": 1.9439507431355853e-05, "loss": 0.0906, "step": 5840 }, { "epoch": 1.07, "learning_rate": 1.9439262339327794e-05, "loss": 0.0955, "step": 5841 }, { "epoch": 1.07, "learning_rate": 1.94390171952703e-05, "loss": 0.0921, "step": 5842 }, { "epoch": 1.07, "learning_rate": 1.9438771999184725e-05, "loss": 0.0984, "step": 5843 }, { "epoch": 1.07, "learning_rate": 1.9438526751072417e-05, "loss": 0.0997, "step": 5844 }, { "epoch": 1.07, "learning_rate": 1.9438281450934725e-05, "loss": 0.1013, "step": 5845 }, { "epoch": 1.07, "learning_rate": 1.9438036098773008e-05, "loss": 0.097, "step": 5846 }, { "epoch": 1.07, "learning_rate": 1.943779069458861e-05, "loss": 0.0995, "step": 5847 }, { "epoch": 1.07, "learning_rate": 1.9437545238382896e-05, "loss": 0.0989, "step": 5848 }, { "epoch": 1.07, "learning_rate": 1.9437299730157205e-05, "loss": 0.1139, "step": 5849 }, { "epoch": 1.07, "learning_rate": 1.94370541699129e-05, "loss": 0.1169, "step": 5850 }, { "epoch": 1.07, "learning_rate": 1.9436808557651332e-05, "loss": 0.0902, "step": 5851 }, { "epoch": 1.07, "learning_rate": 1.943656289337385e-05, "loss": 0.1191, "step": 5852 }, { "epoch": 1.07, "learning_rate": 1.9436317177081818e-05, "loss": 0.0976, "step": 5853 }, { "epoch": 1.07, "learning_rate": 1.943607140877658e-05, "loss": 0.0905, "step": 5854 }, { "epoch": 1.07, "learning_rate": 1.9435825588459497e-05, "loss": 0.1055, "step": 5855 }, { "epoch": 1.07, "learning_rate": 1.9435579716131923e-05, "loss": 0.111, "step": 5856 }, { "epoch": 1.07, "learning_rate": 1.9435333791795212e-05, "loss": 0.0975, "step": 5857 }, { "epoch": 1.07, "learning_rate": 1.9435087815450718e-05, "loss": 0.1094, "step": 5858 }, { "epoch": 1.07, "learning_rate": 1.9434841787099804e-05, "loss": 0.1107, "step": 5859 }, { "epoch": 1.07, "learning_rate": 1.9434595706743818e-05, "loss": 0.0868, "step": 5860 }, { "epoch": 1.07, "learning_rate": 1.943434957438412e-05, "loss": 0.1046, "step": 5861 }, { "epoch": 1.08, "learning_rate": 1.9434103390022066e-05, "loss": 0.0952, "step": 5862 }, { "epoch": 1.08, "learning_rate": 1.943385715365901e-05, "loss": 0.0952, "step": 5863 }, { "epoch": 1.08, "learning_rate": 1.9433610865296316e-05, "loss": 0.1045, "step": 5864 }, { "epoch": 1.08, "learning_rate": 1.9433364524935338e-05, "loss": 0.104, "step": 5865 }, { "epoch": 1.08, "learning_rate": 1.9433118132577432e-05, "loss": 0.0982, "step": 5866 }, { "epoch": 1.08, "learning_rate": 1.943287168822396e-05, "loss": 0.105, "step": 5867 }, { "epoch": 1.08, "learning_rate": 1.9432625191876276e-05, "loss": 0.0984, "step": 5868 }, { "epoch": 1.08, "learning_rate": 1.9432378643535742e-05, "loss": 0.0995, "step": 5869 }, { "epoch": 1.08, "learning_rate": 1.9432132043203716e-05, "loss": 0.0971, "step": 5870 }, { "epoch": 1.08, "learning_rate": 1.9431885390881556e-05, "loss": 0.0977, "step": 5871 }, { "epoch": 1.08, "learning_rate": 1.9431638686570625e-05, "loss": 0.1116, "step": 5872 }, { "epoch": 1.08, "learning_rate": 1.9431391930272278e-05, "loss": 0.0967, "step": 5873 }, { "epoch": 1.08, "learning_rate": 1.943114512198788e-05, "loss": 0.1107, "step": 5874 }, { "epoch": 1.08, "learning_rate": 1.9430898261718786e-05, "loss": 0.101, "step": 5875 }, { "epoch": 1.08, "learning_rate": 1.9430651349466366e-05, "loss": 0.104, "step": 5876 }, { "epoch": 1.08, "learning_rate": 1.943040438523197e-05, "loss": 0.1143, "step": 5877 }, { "epoch": 1.08, "learning_rate": 1.9430157369016966e-05, "loss": 0.114, "step": 5878 }, { "epoch": 1.08, "learning_rate": 1.9429910300822714e-05, "loss": 0.1036, "step": 5879 }, { "epoch": 1.08, "learning_rate": 1.9429663180650572e-05, "loss": 0.1, "step": 5880 }, { "epoch": 1.08, "learning_rate": 1.9429416008501905e-05, "loss": 0.1133, "step": 5881 }, { "epoch": 1.08, "learning_rate": 1.942916878437808e-05, "loss": 0.1049, "step": 5882 }, { "epoch": 1.08, "learning_rate": 1.9428921508280454e-05, "loss": 0.1035, "step": 5883 }, { "epoch": 1.08, "learning_rate": 1.9428674180210394e-05, "loss": 0.1128, "step": 5884 }, { "epoch": 1.08, "learning_rate": 1.9428426800169257e-05, "loss": 0.0961, "step": 5885 }, { "epoch": 1.08, "learning_rate": 1.9428179368158412e-05, "loss": 0.1081, "step": 5886 }, { "epoch": 1.08, "learning_rate": 1.9427931884179224e-05, "loss": 0.1074, "step": 5887 }, { "epoch": 1.08, "learning_rate": 1.9427684348233053e-05, "loss": 0.1113, "step": 5888 }, { "epoch": 1.08, "learning_rate": 1.9427436760321264e-05, "loss": 0.1068, "step": 5889 }, { "epoch": 1.08, "learning_rate": 1.9427189120445225e-05, "loss": 0.1172, "step": 5890 }, { "epoch": 1.08, "learning_rate": 1.9426941428606295e-05, "loss": 0.1013, "step": 5891 }, { "epoch": 1.08, "learning_rate": 1.942669368480585e-05, "loss": 0.109, "step": 5892 }, { "epoch": 1.08, "learning_rate": 1.942644588904524e-05, "loss": 0.1087, "step": 5893 }, { "epoch": 1.08, "learning_rate": 1.9426198041325845e-05, "loss": 0.0944, "step": 5894 }, { "epoch": 1.08, "learning_rate": 1.9425950141649025e-05, "loss": 0.1018, "step": 5895 }, { "epoch": 1.08, "learning_rate": 1.9425702190016146e-05, "loss": 0.1074, "step": 5896 }, { "epoch": 1.08, "learning_rate": 1.9425454186428573e-05, "loss": 0.0978, "step": 5897 }, { "epoch": 1.08, "learning_rate": 1.942520613088768e-05, "loss": 0.1141, "step": 5898 }, { "epoch": 1.08, "learning_rate": 1.942495802339483e-05, "loss": 0.094, "step": 5899 }, { "epoch": 1.08, "learning_rate": 1.942470986395139e-05, "loss": 0.0993, "step": 5900 }, { "epoch": 1.08, "learning_rate": 1.9424461652558726e-05, "loss": 0.1103, "step": 5901 }, { "epoch": 1.08, "learning_rate": 1.942421338921821e-05, "loss": 0.101, "step": 5902 }, { "epoch": 1.08, "learning_rate": 1.9423965073931207e-05, "loss": 0.1062, "step": 5903 }, { "epoch": 1.08, "learning_rate": 1.9423716706699092e-05, "loss": 0.1128, "step": 5904 }, { "epoch": 1.08, "learning_rate": 1.9423468287523227e-05, "loss": 0.0918, "step": 5905 }, { "epoch": 1.08, "learning_rate": 1.9423219816404984e-05, "loss": 0.1129, "step": 5906 }, { "epoch": 1.08, "learning_rate": 1.9422971293345732e-05, "loss": 0.1044, "step": 5907 }, { "epoch": 1.08, "learning_rate": 1.9422722718346843e-05, "loss": 0.1009, "step": 5908 }, { "epoch": 1.08, "learning_rate": 1.9422474091409683e-05, "loss": 0.1092, "step": 5909 }, { "epoch": 1.08, "learning_rate": 1.942222541253563e-05, "loss": 0.1056, "step": 5910 }, { "epoch": 1.08, "learning_rate": 1.9421976681726047e-05, "loss": 0.1005, "step": 5911 }, { "epoch": 1.08, "learning_rate": 1.942172789898231e-05, "loss": 0.1073, "step": 5912 }, { "epoch": 1.08, "learning_rate": 1.9421479064305782e-05, "loss": 0.0966, "step": 5913 }, { "epoch": 1.08, "learning_rate": 1.9421230177697844e-05, "loss": 0.1009, "step": 5914 }, { "epoch": 1.08, "learning_rate": 1.9420981239159865e-05, "loss": 0.0981, "step": 5915 }, { "epoch": 1.08, "learning_rate": 1.9420732248693216e-05, "loss": 0.1083, "step": 5916 }, { "epoch": 1.09, "learning_rate": 1.9420483206299274e-05, "loss": 0.1073, "step": 5917 }, { "epoch": 1.09, "learning_rate": 1.9420234111979403e-05, "loss": 0.1072, "step": 5918 }, { "epoch": 1.09, "learning_rate": 1.9419984965734982e-05, "loss": 0.1052, "step": 5919 }, { "epoch": 1.09, "learning_rate": 1.9419735767567385e-05, "loss": 0.1254, "step": 5920 }, { "epoch": 1.09, "learning_rate": 1.9419486517477982e-05, "loss": 0.111, "step": 5921 }, { "epoch": 1.09, "learning_rate": 1.941923721546815e-05, "loss": 0.1035, "step": 5922 }, { "epoch": 1.09, "learning_rate": 1.9418987861539263e-05, "loss": 0.1037, "step": 5923 }, { "epoch": 1.09, "learning_rate": 1.9418738455692692e-05, "loss": 0.0995, "step": 5924 }, { "epoch": 1.09, "learning_rate": 1.9418488997929815e-05, "loss": 0.0918, "step": 5925 }, { "epoch": 1.09, "learning_rate": 1.9418239488252003e-05, "loss": 0.107, "step": 5926 }, { "epoch": 1.09, "learning_rate": 1.9417989926660637e-05, "loss": 0.1053, "step": 5927 }, { "epoch": 1.09, "learning_rate": 1.941774031315709e-05, "loss": 0.1042, "step": 5928 }, { "epoch": 1.09, "learning_rate": 1.9417490647742738e-05, "loss": 0.1025, "step": 5929 }, { "epoch": 1.09, "learning_rate": 1.9417240930418954e-05, "loss": 0.0955, "step": 5930 }, { "epoch": 1.09, "learning_rate": 1.941699116118712e-05, "loss": 0.1087, "step": 5931 }, { "epoch": 1.09, "learning_rate": 1.941674134004861e-05, "loss": 0.1063, "step": 5932 }, { "epoch": 1.09, "learning_rate": 1.94164914670048e-05, "loss": 0.1173, "step": 5933 }, { "epoch": 1.09, "learning_rate": 1.9416241542057072e-05, "loss": 0.1146, "step": 5934 }, { "epoch": 1.09, "learning_rate": 1.9415991565206797e-05, "loss": 0.0841, "step": 5935 }, { "epoch": 1.09, "learning_rate": 1.9415741536455354e-05, "loss": 0.0995, "step": 5936 }, { "epoch": 1.09, "learning_rate": 1.9415491455804126e-05, "loss": 0.1019, "step": 5937 }, { "epoch": 1.09, "learning_rate": 1.9415241323254487e-05, "loss": 0.1065, "step": 5938 }, { "epoch": 1.09, "learning_rate": 1.9414991138807817e-05, "loss": 0.0998, "step": 5939 }, { "epoch": 1.09, "learning_rate": 1.94147409024655e-05, "loss": 0.0986, "step": 5940 }, { "epoch": 1.09, "learning_rate": 1.9414490614228904e-05, "loss": 0.1023, "step": 5941 }, { "epoch": 1.09, "learning_rate": 1.9414240274099415e-05, "loss": 0.0996, "step": 5942 }, { "epoch": 1.09, "learning_rate": 1.9413989882078418e-05, "loss": 0.1021, "step": 5943 }, { "epoch": 1.09, "learning_rate": 1.9413739438167288e-05, "loss": 0.1001, "step": 5944 }, { "epoch": 1.09, "learning_rate": 1.9413488942367404e-05, "loss": 0.1032, "step": 5945 }, { "epoch": 1.09, "learning_rate": 1.9413238394680148e-05, "loss": 0.1032, "step": 5946 }, { "epoch": 1.09, "learning_rate": 1.94129877951069e-05, "loss": 0.1156, "step": 5947 }, { "epoch": 1.09, "learning_rate": 1.9412737143649045e-05, "loss": 0.1047, "step": 5948 }, { "epoch": 1.09, "learning_rate": 1.9412486440307963e-05, "loss": 0.1109, "step": 5949 }, { "epoch": 1.09, "learning_rate": 1.9412235685085034e-05, "loss": 0.0978, "step": 5950 }, { "epoch": 1.09, "learning_rate": 1.9411984877981643e-05, "loss": 0.0828, "step": 5951 }, { "epoch": 1.09, "learning_rate": 1.9411734018999168e-05, "loss": 0.1015, "step": 5952 }, { "epoch": 1.09, "learning_rate": 1.9411483108138994e-05, "loss": 0.1081, "step": 5953 }, { "epoch": 1.09, "learning_rate": 1.9411232145402508e-05, "loss": 0.1176, "step": 5954 }, { "epoch": 1.09, "learning_rate": 1.9410981130791088e-05, "loss": 0.0999, "step": 5955 }, { "epoch": 1.09, "learning_rate": 1.941073006430612e-05, "loss": 0.1048, "step": 5956 }, { "epoch": 1.09, "learning_rate": 1.941047894594899e-05, "loss": 0.0917, "step": 5957 }, { "epoch": 1.09, "learning_rate": 1.9410227775721077e-05, "loss": 0.1075, "step": 5958 }, { "epoch": 1.09, "learning_rate": 1.9409976553623767e-05, "loss": 0.1148, "step": 5959 }, { "epoch": 1.09, "learning_rate": 1.9409725279658444e-05, "loss": 0.0933, "step": 5960 }, { "epoch": 1.09, "learning_rate": 1.94094739538265e-05, "loss": 0.0977, "step": 5961 }, { "epoch": 1.09, "learning_rate": 1.940922257612931e-05, "loss": 0.1058, "step": 5962 }, { "epoch": 1.09, "learning_rate": 1.940897114656827e-05, "loss": 0.1024, "step": 5963 }, { "epoch": 1.09, "learning_rate": 1.9408719665144756e-05, "loss": 0.1157, "step": 5964 }, { "epoch": 1.09, "learning_rate": 1.940846813186016e-05, "loss": 0.0944, "step": 5965 }, { "epoch": 1.09, "learning_rate": 1.940821654671587e-05, "loss": 0.1125, "step": 5966 }, { "epoch": 1.09, "learning_rate": 1.940796490971327e-05, "loss": 0.0963, "step": 5967 }, { "epoch": 1.09, "learning_rate": 1.9407713220853744e-05, "loss": 0.1023, "step": 5968 }, { "epoch": 1.09, "learning_rate": 1.940746148013868e-05, "loss": 0.096, "step": 5969 }, { "epoch": 1.09, "learning_rate": 1.9407209687569475e-05, "loss": 0.0929, "step": 5970 }, { "epoch": 1.09, "learning_rate": 1.9406957843147507e-05, "loss": 0.1006, "step": 5971 }, { "epoch": 1.1, "learning_rate": 1.9406705946874166e-05, "loss": 0.1081, "step": 5972 }, { "epoch": 1.1, "learning_rate": 1.9406453998750842e-05, "loss": 0.1008, "step": 5973 }, { "epoch": 1.1, "learning_rate": 1.9406201998778924e-05, "loss": 0.0999, "step": 5974 }, { "epoch": 1.1, "learning_rate": 1.9405949946959796e-05, "loss": 0.0917, "step": 5975 }, { "epoch": 1.1, "learning_rate": 1.9405697843294857e-05, "loss": 0.0997, "step": 5976 }, { "epoch": 1.1, "learning_rate": 1.940544568778549e-05, "loss": 0.1037, "step": 5977 }, { "epoch": 1.1, "learning_rate": 1.9405193480433082e-05, "loss": 0.0985, "step": 5978 }, { "epoch": 1.1, "learning_rate": 1.940494122123903e-05, "loss": 0.0977, "step": 5979 }, { "epoch": 1.1, "learning_rate": 1.9404688910204723e-05, "loss": 0.1149, "step": 5980 }, { "epoch": 1.1, "learning_rate": 1.9404436547331547e-05, "loss": 0.0901, "step": 5981 }, { "epoch": 1.1, "learning_rate": 1.9404184132620896e-05, "loss": 0.1055, "step": 5982 }, { "epoch": 1.1, "learning_rate": 1.9403931666074167e-05, "loss": 0.1007, "step": 5983 }, { "epoch": 1.1, "learning_rate": 1.9403679147692745e-05, "loss": 0.0925, "step": 5984 }, { "epoch": 1.1, "learning_rate": 1.940342657747802e-05, "loss": 0.0923, "step": 5985 }, { "epoch": 1.1, "learning_rate": 1.9403173955431393e-05, "loss": 0.0939, "step": 5986 }, { "epoch": 1.1, "learning_rate": 1.9402921281554244e-05, "loss": 0.0951, "step": 5987 }, { "epoch": 1.1, "learning_rate": 1.9402668555847976e-05, "loss": 0.0949, "step": 5988 }, { "epoch": 1.1, "learning_rate": 1.9402415778313978e-05, "loss": 0.0955, "step": 5989 }, { "epoch": 1.1, "learning_rate": 1.9402162948953645e-05, "loss": 0.1022, "step": 5990 }, { "epoch": 1.1, "learning_rate": 1.9401910067768367e-05, "loss": 0.1008, "step": 5991 }, { "epoch": 1.1, "learning_rate": 1.9401657134759544e-05, "loss": 0.1039, "step": 5992 }, { "epoch": 1.1, "learning_rate": 1.9401404149928563e-05, "loss": 0.1052, "step": 5993 }, { "epoch": 1.1, "learning_rate": 1.9401151113276822e-05, "loss": 0.0934, "step": 5994 }, { "epoch": 1.1, "learning_rate": 1.940089802480572e-05, "loss": 0.1003, "step": 5995 }, { "epoch": 1.1, "learning_rate": 1.9400644884516648e-05, "loss": 0.0947, "step": 5996 }, { "epoch": 1.1, "learning_rate": 1.9400391692410996e-05, "loss": 0.1006, "step": 5997 }, { "epoch": 1.1, "learning_rate": 1.9400138448490167e-05, "loss": 0.0961, "step": 5998 }, { "epoch": 1.1, "learning_rate": 1.9399885152755558e-05, "loss": 0.1084, "step": 5999 }, { "epoch": 1.1, "learning_rate": 1.939963180520856e-05, "loss": 0.1008, "step": 6000 }, { "epoch": 1.1, "learning_rate": 1.9399378405850567e-05, "loss": 0.0977, "step": 6001 }, { "epoch": 1.1, "learning_rate": 1.9399124954682986e-05, "loss": 0.1023, "step": 6002 }, { "epoch": 1.1, "learning_rate": 1.9398871451707203e-05, "loss": 0.0952, "step": 6003 }, { "epoch": 1.1, "learning_rate": 1.9398617896924622e-05, "loss": 0.1172, "step": 6004 }, { "epoch": 1.1, "learning_rate": 1.939836429033664e-05, "loss": 0.0997, "step": 6005 }, { "epoch": 1.1, "learning_rate": 1.9398110631944655e-05, "loss": 0.1107, "step": 6006 }, { "epoch": 1.1, "learning_rate": 1.939785692175006e-05, "loss": 0.1177, "step": 6007 }, { "epoch": 1.1, "learning_rate": 1.939760315975426e-05, "loss": 0.0933, "step": 6008 }, { "epoch": 1.1, "learning_rate": 1.9397349345958653e-05, "loss": 0.112, "step": 6009 }, { "epoch": 1.1, "learning_rate": 1.9397095480364633e-05, "loss": 0.119, "step": 6010 }, { "epoch": 1.1, "learning_rate": 1.9396841562973604e-05, "loss": 0.0875, "step": 6011 }, { "epoch": 1.1, "learning_rate": 1.9396587593786967e-05, "loss": 0.1008, "step": 6012 }, { "epoch": 1.1, "learning_rate": 1.9396333572806115e-05, "loss": 0.0987, "step": 6013 }, { "epoch": 1.1, "learning_rate": 1.939607950003245e-05, "loss": 0.0969, "step": 6014 }, { "epoch": 1.1, "learning_rate": 1.939582537546738e-05, "loss": 0.1182, "step": 6015 }, { "epoch": 1.1, "learning_rate": 1.93955711991123e-05, "loss": 0.0954, "step": 6016 }, { "epoch": 1.1, "learning_rate": 1.939531697096861e-05, "loss": 0.1098, "step": 6017 }, { "epoch": 1.1, "learning_rate": 1.939506269103771e-05, "loss": 0.0857, "step": 6018 }, { "epoch": 1.1, "learning_rate": 1.939480835932101e-05, "loss": 0.099, "step": 6019 }, { "epoch": 1.1, "learning_rate": 1.9394553975819903e-05, "loss": 0.1026, "step": 6020 }, { "epoch": 1.1, "learning_rate": 1.9394299540535795e-05, "loss": 0.0948, "step": 6021 }, { "epoch": 1.1, "learning_rate": 1.9394045053470087e-05, "loss": 0.0903, "step": 6022 }, { "epoch": 1.1, "learning_rate": 1.939379051462418e-05, "loss": 0.0995, "step": 6023 }, { "epoch": 1.1, "learning_rate": 1.9393535923999486e-05, "loss": 0.0911, "step": 6024 }, { "epoch": 1.1, "learning_rate": 1.9393281281597402e-05, "loss": 0.1055, "step": 6025 }, { "epoch": 1.11, "learning_rate": 1.9393026587419324e-05, "loss": 0.0931, "step": 6026 }, { "epoch": 1.11, "learning_rate": 1.939277184146667e-05, "loss": 0.0988, "step": 6027 }, { "epoch": 1.11, "learning_rate": 1.9392517043740833e-05, "loss": 0.1088, "step": 6028 }, { "epoch": 1.11, "learning_rate": 1.9392262194243225e-05, "loss": 0.1091, "step": 6029 }, { "epoch": 1.11, "learning_rate": 1.9392007292975248e-05, "loss": 0.1029, "step": 6030 }, { "epoch": 1.11, "learning_rate": 1.9391752339938305e-05, "loss": 0.0985, "step": 6031 }, { "epoch": 1.11, "learning_rate": 1.9391497335133804e-05, "loss": 0.1036, "step": 6032 }, { "epoch": 1.11, "learning_rate": 1.939124227856315e-05, "loss": 0.0928, "step": 6033 }, { "epoch": 1.11, "learning_rate": 1.9390987170227746e-05, "loss": 0.109, "step": 6034 }, { "epoch": 1.11, "learning_rate": 1.9390732010129004e-05, "loss": 0.1023, "step": 6035 }, { "epoch": 1.11, "learning_rate": 1.9390476798268326e-05, "loss": 0.1129, "step": 6036 }, { "epoch": 1.11, "learning_rate": 1.939022153464712e-05, "loss": 0.0941, "step": 6037 }, { "epoch": 1.11, "learning_rate": 1.938996621926679e-05, "loss": 0.1116, "step": 6038 }, { "epoch": 1.11, "learning_rate": 1.938971085212875e-05, "loss": 0.1097, "step": 6039 }, { "epoch": 1.11, "learning_rate": 1.9389455433234403e-05, "loss": 0.1069, "step": 6040 }, { "epoch": 1.11, "learning_rate": 1.9389199962585156e-05, "loss": 0.109, "step": 6041 }, { "epoch": 1.11, "learning_rate": 1.9388944440182423e-05, "loss": 0.1078, "step": 6042 }, { "epoch": 1.11, "learning_rate": 1.9388688866027604e-05, "loss": 0.1101, "step": 6043 }, { "epoch": 1.11, "learning_rate": 1.9388433240122113e-05, "loss": 0.1031, "step": 6044 }, { "epoch": 1.11, "learning_rate": 1.938817756246736e-05, "loss": 0.1025, "step": 6045 }, { "epoch": 1.11, "learning_rate": 1.938792183306475e-05, "loss": 0.106, "step": 6046 }, { "epoch": 1.11, "learning_rate": 1.9387666051915697e-05, "loss": 0.1058, "step": 6047 }, { "epoch": 1.11, "learning_rate": 1.9387410219021606e-05, "loss": 0.0981, "step": 6048 }, { "epoch": 1.11, "learning_rate": 1.938715433438389e-05, "loss": 0.0971, "step": 6049 }, { "epoch": 1.11, "learning_rate": 1.938689839800396e-05, "loss": 0.1107, "step": 6050 }, { "epoch": 1.11, "learning_rate": 1.938664240988323e-05, "loss": 0.1119, "step": 6051 }, { "epoch": 1.11, "learning_rate": 1.9386386370023104e-05, "loss": 0.1099, "step": 6052 }, { "epoch": 1.11, "learning_rate": 1.9386130278424998e-05, "loss": 0.0919, "step": 6053 }, { "epoch": 1.11, "learning_rate": 1.9385874135090316e-05, "loss": 0.1094, "step": 6054 }, { "epoch": 1.11, "learning_rate": 1.938561794002048e-05, "loss": 0.1011, "step": 6055 }, { "epoch": 1.11, "learning_rate": 1.9385361693216903e-05, "loss": 0.1015, "step": 6056 }, { "epoch": 1.11, "learning_rate": 1.9385105394680987e-05, "loss": 0.0963, "step": 6057 }, { "epoch": 1.11, "learning_rate": 1.938484904441415e-05, "loss": 0.1023, "step": 6058 }, { "epoch": 1.11, "learning_rate": 1.9384592642417806e-05, "loss": 0.1046, "step": 6059 }, { "epoch": 1.11, "learning_rate": 1.938433618869337e-05, "loss": 0.0973, "step": 6060 }, { "epoch": 1.11, "learning_rate": 1.938407968324225e-05, "loss": 0.0969, "step": 6061 }, { "epoch": 1.11, "learning_rate": 1.9383823126065864e-05, "loss": 0.1089, "step": 6062 }, { "epoch": 1.11, "learning_rate": 1.9383566517165624e-05, "loss": 0.0969, "step": 6063 }, { "epoch": 1.11, "learning_rate": 1.9383309856542948e-05, "loss": 0.0983, "step": 6064 }, { "epoch": 1.11, "learning_rate": 1.9383053144199247e-05, "loss": 0.1111, "step": 6065 }, { "epoch": 1.11, "learning_rate": 1.9382796380135935e-05, "loss": 0.0971, "step": 6066 }, { "epoch": 1.11, "learning_rate": 1.9382539564354434e-05, "loss": 0.1107, "step": 6067 }, { "epoch": 1.11, "learning_rate": 1.938228269685615e-05, "loss": 0.1034, "step": 6068 }, { "epoch": 1.11, "learning_rate": 1.9382025777642507e-05, "loss": 0.0965, "step": 6069 }, { "epoch": 1.11, "learning_rate": 1.9381768806714918e-05, "loss": 0.0965, "step": 6070 }, { "epoch": 1.11, "learning_rate": 1.9381511784074802e-05, "loss": 0.1098, "step": 6071 }, { "epoch": 1.11, "learning_rate": 1.9381254709723568e-05, "loss": 0.1112, "step": 6072 }, { "epoch": 1.11, "learning_rate": 1.938099758366264e-05, "loss": 0.1145, "step": 6073 }, { "epoch": 1.11, "learning_rate": 1.9380740405893433e-05, "loss": 0.0992, "step": 6074 }, { "epoch": 1.11, "learning_rate": 1.9380483176417368e-05, "loss": 0.1119, "step": 6075 }, { "epoch": 1.11, "learning_rate": 1.9380225895235857e-05, "loss": 0.1024, "step": 6076 }, { "epoch": 1.11, "learning_rate": 1.9379968562350324e-05, "loss": 0.1011, "step": 6077 }, { "epoch": 1.11, "learning_rate": 1.9379711177762183e-05, "loss": 0.1098, "step": 6078 }, { "epoch": 1.11, "learning_rate": 1.9379453741472853e-05, "loss": 0.1211, "step": 6079 }, { "epoch": 1.11, "learning_rate": 1.937919625348376e-05, "loss": 0.1044, "step": 6080 }, { "epoch": 1.12, "learning_rate": 1.937893871379631e-05, "loss": 0.098, "step": 6081 }, { "epoch": 1.12, "learning_rate": 1.9378681122411934e-05, "loss": 0.0951, "step": 6082 }, { "epoch": 1.12, "learning_rate": 1.9378423479332045e-05, "loss": 0.0897, "step": 6083 }, { "epoch": 1.12, "learning_rate": 1.9378165784558068e-05, "loss": 0.0857, "step": 6084 }, { "epoch": 1.12, "learning_rate": 1.9377908038091422e-05, "loss": 0.0992, "step": 6085 }, { "epoch": 1.12, "learning_rate": 1.937765023993353e-05, "loss": 0.1067, "step": 6086 }, { "epoch": 1.12, "learning_rate": 1.9377392390085804e-05, "loss": 0.1052, "step": 6087 }, { "epoch": 1.12, "learning_rate": 1.9377134488549675e-05, "loss": 0.1177, "step": 6088 }, { "epoch": 1.12, "learning_rate": 1.937687653532656e-05, "loss": 0.1025, "step": 6089 }, { "epoch": 1.12, "learning_rate": 1.9376618530417882e-05, "loss": 0.1087, "step": 6090 }, { "epoch": 1.12, "learning_rate": 1.9376360473825064e-05, "loss": 0.1061, "step": 6091 }, { "epoch": 1.12, "learning_rate": 1.9376102365549528e-05, "loss": 0.1032, "step": 6092 }, { "epoch": 1.12, "learning_rate": 1.9375844205592696e-05, "loss": 0.1006, "step": 6093 }, { "epoch": 1.12, "learning_rate": 1.937558599395599e-05, "loss": 0.0996, "step": 6094 }, { "epoch": 1.12, "learning_rate": 1.9375327730640833e-05, "loss": 0.106, "step": 6095 }, { "epoch": 1.12, "learning_rate": 1.937506941564865e-05, "loss": 0.0969, "step": 6096 }, { "epoch": 1.12, "learning_rate": 1.9374811048980866e-05, "loss": 0.1145, "step": 6097 }, { "epoch": 1.12, "learning_rate": 1.9374552630638904e-05, "loss": 0.1051, "step": 6098 }, { "epoch": 1.12, "learning_rate": 1.9374294160624188e-05, "loss": 0.1131, "step": 6099 }, { "epoch": 1.12, "learning_rate": 1.9374035638938142e-05, "loss": 0.0986, "step": 6100 }, { "epoch": 1.12, "learning_rate": 1.9373777065582195e-05, "loss": 0.1131, "step": 6101 }, { "epoch": 1.12, "learning_rate": 1.9373518440557766e-05, "loss": 0.096, "step": 6102 }, { "epoch": 1.12, "learning_rate": 1.9373259763866286e-05, "loss": 0.091, "step": 6103 }, { "epoch": 1.12, "learning_rate": 1.9373001035509178e-05, "loss": 0.1195, "step": 6104 }, { "epoch": 1.12, "learning_rate": 1.9372742255487868e-05, "loss": 0.1055, "step": 6105 }, { "epoch": 1.12, "learning_rate": 1.9372483423803782e-05, "loss": 0.1071, "step": 6106 }, { "epoch": 1.12, "learning_rate": 1.937222454045835e-05, "loss": 0.0923, "step": 6107 }, { "epoch": 1.12, "learning_rate": 1.9371965605452996e-05, "loss": 0.1083, "step": 6108 }, { "epoch": 1.12, "learning_rate": 1.9371706618789148e-05, "loss": 0.1108, "step": 6109 }, { "epoch": 1.12, "learning_rate": 1.9371447580468234e-05, "loss": 0.0927, "step": 6110 }, { "epoch": 1.12, "learning_rate": 1.937118849049168e-05, "loss": 0.1092, "step": 6111 }, { "epoch": 1.12, "learning_rate": 1.9370929348860915e-05, "loss": 0.1063, "step": 6112 }, { "epoch": 1.12, "learning_rate": 1.9370670155577373e-05, "loss": 0.1, "step": 6113 }, { "epoch": 1.12, "learning_rate": 1.9370410910642473e-05, "loss": 0.1124, "step": 6114 }, { "epoch": 1.12, "learning_rate": 1.937015161405765e-05, "loss": 0.1002, "step": 6115 }, { "epoch": 1.12, "learning_rate": 1.936989226582433e-05, "loss": 0.1051, "step": 6116 }, { "epoch": 1.12, "learning_rate": 1.9369632865943946e-05, "loss": 0.1081, "step": 6117 }, { "epoch": 1.12, "learning_rate": 1.9369373414417924e-05, "loss": 0.1086, "step": 6118 }, { "epoch": 1.12, "learning_rate": 1.93691139112477e-05, "loss": 0.0988, "step": 6119 }, { "epoch": 1.12, "learning_rate": 1.9368854356434698e-05, "loss": 0.1, "step": 6120 }, { "epoch": 1.12, "learning_rate": 1.936859474998035e-05, "loss": 0.0986, "step": 6121 }, { "epoch": 1.12, "learning_rate": 1.936833509188609e-05, "loss": 0.1016, "step": 6122 }, { "epoch": 1.12, "learning_rate": 1.9368075382153348e-05, "loss": 0.0999, "step": 6123 }, { "epoch": 1.12, "learning_rate": 1.9367815620783554e-05, "loss": 0.1048, "step": 6124 }, { "epoch": 1.12, "learning_rate": 1.936755580777814e-05, "loss": 0.098, "step": 6125 }, { "epoch": 1.12, "learning_rate": 1.9367295943138542e-05, "loss": 0.1048, "step": 6126 }, { "epoch": 1.12, "learning_rate": 1.936703602686619e-05, "loss": 0.0951, "step": 6127 }, { "epoch": 1.12, "learning_rate": 1.9366776058962512e-05, "loss": 0.1077, "step": 6128 }, { "epoch": 1.12, "learning_rate": 1.9366516039428945e-05, "loss": 0.103, "step": 6129 }, { "epoch": 1.12, "learning_rate": 1.9366255968266925e-05, "loss": 0.09, "step": 6130 }, { "epoch": 1.12, "learning_rate": 1.936599584547788e-05, "loss": 0.0998, "step": 6131 }, { "epoch": 1.12, "learning_rate": 1.9365735671063247e-05, "loss": 0.0961, "step": 6132 }, { "epoch": 1.12, "learning_rate": 1.936547544502446e-05, "loss": 0.109, "step": 6133 }, { "epoch": 1.12, "learning_rate": 1.9365215167362956e-05, "loss": 0.1134, "step": 6134 }, { "epoch": 1.13, "learning_rate": 1.9364954838080164e-05, "loss": 0.1085, "step": 6135 }, { "epoch": 1.13, "learning_rate": 1.936469445717752e-05, "loss": 0.1029, "step": 6136 }, { "epoch": 1.13, "learning_rate": 1.9364434024656465e-05, "loss": 0.1121, "step": 6137 }, { "epoch": 1.13, "learning_rate": 1.936417354051843e-05, "loss": 0.1114, "step": 6138 }, { "epoch": 1.13, "learning_rate": 1.9363913004764847e-05, "loss": 0.0945, "step": 6139 }, { "epoch": 1.13, "learning_rate": 1.9363652417397153e-05, "loss": 0.1129, "step": 6140 }, { "epoch": 1.13, "learning_rate": 1.9363391778416797e-05, "loss": 0.0971, "step": 6141 }, { "epoch": 1.13, "learning_rate": 1.9363131087825196e-05, "loss": 0.1054, "step": 6142 }, { "epoch": 1.13, "learning_rate": 1.9362870345623807e-05, "loss": 0.1092, "step": 6143 }, { "epoch": 1.13, "learning_rate": 1.936260955181405e-05, "loss": 0.1005, "step": 6144 }, { "epoch": 1.13, "learning_rate": 1.9362348706397374e-05, "loss": 0.0865, "step": 6145 }, { "epoch": 1.13, "learning_rate": 1.936208780937521e-05, "loss": 0.0961, "step": 6146 }, { "epoch": 1.13, "learning_rate": 1.9361826860749002e-05, "loss": 0.1052, "step": 6147 }, { "epoch": 1.13, "learning_rate": 1.936156586052018e-05, "loss": 0.1126, "step": 6148 }, { "epoch": 1.13, "learning_rate": 1.9361304808690193e-05, "loss": 0.0968, "step": 6149 }, { "epoch": 1.13, "learning_rate": 1.9361043705260473e-05, "loss": 0.0871, "step": 6150 }, { "epoch": 1.13, "learning_rate": 1.9360782550232457e-05, "loss": 0.1167, "step": 6151 }, { "epoch": 1.13, "learning_rate": 1.936052134360759e-05, "loss": 0.0955, "step": 6152 }, { "epoch": 1.13, "learning_rate": 1.9360260085387313e-05, "loss": 0.103, "step": 6153 }, { "epoch": 1.13, "learning_rate": 1.9359998775573062e-05, "loss": 0.1015, "step": 6154 }, { "epoch": 1.13, "learning_rate": 1.935973741416628e-05, "loss": 0.0965, "step": 6155 }, { "epoch": 1.13, "learning_rate": 1.93594760011684e-05, "loss": 0.1185, "step": 6156 }, { "epoch": 1.13, "learning_rate": 1.9359214536580875e-05, "loss": 0.1015, "step": 6157 }, { "epoch": 1.13, "learning_rate": 1.935895302040514e-05, "loss": 0.0954, "step": 6158 }, { "epoch": 1.13, "learning_rate": 1.9358691452642633e-05, "loss": 0.0995, "step": 6159 }, { "epoch": 1.13, "learning_rate": 1.93584298332948e-05, "loss": 0.1253, "step": 6160 }, { "epoch": 1.13, "learning_rate": 1.9358168162363083e-05, "loss": 0.1047, "step": 6161 }, { "epoch": 1.13, "learning_rate": 1.9357906439848925e-05, "loss": 0.109, "step": 6162 }, { "epoch": 1.13, "learning_rate": 1.935764466575377e-05, "loss": 0.1135, "step": 6163 }, { "epoch": 1.13, "learning_rate": 1.9357382840079057e-05, "loss": 0.0942, "step": 6164 }, { "epoch": 1.13, "learning_rate": 1.9357120962826227e-05, "loss": 0.1152, "step": 6165 }, { "epoch": 1.13, "learning_rate": 1.935685903399673e-05, "loss": 0.1045, "step": 6166 }, { "epoch": 1.13, "learning_rate": 1.9356597053592004e-05, "loss": 0.1111, "step": 6167 }, { "epoch": 1.13, "learning_rate": 1.9356335021613498e-05, "loss": 0.109, "step": 6168 }, { "epoch": 1.13, "learning_rate": 1.9356072938062652e-05, "loss": 0.095, "step": 6169 }, { "epoch": 1.13, "learning_rate": 1.9355810802940914e-05, "loss": 0.1034, "step": 6170 }, { "epoch": 1.13, "learning_rate": 1.935554861624973e-05, "loss": 0.1006, "step": 6171 }, { "epoch": 1.13, "learning_rate": 1.9355286377990542e-05, "loss": 0.098, "step": 6172 }, { "epoch": 1.13, "learning_rate": 1.9355024088164796e-05, "loss": 0.1003, "step": 6173 }, { "epoch": 1.13, "learning_rate": 1.9354761746773936e-05, "loss": 0.0995, "step": 6174 }, { "epoch": 1.13, "learning_rate": 1.935449935381941e-05, "loss": 0.1144, "step": 6175 }, { "epoch": 1.13, "learning_rate": 1.935423690930267e-05, "loss": 0.0986, "step": 6176 }, { "epoch": 1.13, "learning_rate": 1.9353974413225152e-05, "loss": 0.1024, "step": 6177 }, { "epoch": 1.13, "learning_rate": 1.935371186558831e-05, "loss": 0.1034, "step": 6178 }, { "epoch": 1.13, "learning_rate": 1.9353449266393587e-05, "loss": 0.1027, "step": 6179 }, { "epoch": 1.13, "learning_rate": 1.9353186615642433e-05, "loss": 0.1064, "step": 6180 }, { "epoch": 1.13, "learning_rate": 1.9352923913336296e-05, "loss": 0.1041, "step": 6181 }, { "epoch": 1.13, "learning_rate": 1.935266115947662e-05, "loss": 0.103, "step": 6182 }, { "epoch": 1.13, "learning_rate": 1.9352398354064863e-05, "loss": 0.0984, "step": 6183 }, { "epoch": 1.13, "learning_rate": 1.9352135497102464e-05, "loss": 0.0943, "step": 6184 }, { "epoch": 1.13, "learning_rate": 1.9351872588590873e-05, "loss": 0.0991, "step": 6185 }, { "epoch": 1.13, "learning_rate": 1.9351609628531546e-05, "loss": 0.1044, "step": 6186 }, { "epoch": 1.13, "learning_rate": 1.9351346616925925e-05, "loss": 0.1085, "step": 6187 }, { "epoch": 1.13, "learning_rate": 1.9351083553775462e-05, "loss": 0.0887, "step": 6188 }, { "epoch": 1.13, "learning_rate": 1.9350820439081608e-05, "loss": 0.1078, "step": 6189 }, { "epoch": 1.14, "learning_rate": 1.935055727284581e-05, "loss": 0.0945, "step": 6190 }, { "epoch": 1.14, "learning_rate": 1.9350294055069522e-05, "loss": 0.0983, "step": 6191 }, { "epoch": 1.14, "learning_rate": 1.9350030785754198e-05, "loss": 0.097, "step": 6192 }, { "epoch": 1.14, "learning_rate": 1.9349767464901282e-05, "loss": 0.0965, "step": 6193 }, { "epoch": 1.14, "learning_rate": 1.934950409251223e-05, "loss": 0.1078, "step": 6194 }, { "epoch": 1.14, "learning_rate": 1.9349240668588493e-05, "loss": 0.1077, "step": 6195 }, { "epoch": 1.14, "learning_rate": 1.934897719313152e-05, "loss": 0.0966, "step": 6196 }, { "epoch": 1.14, "learning_rate": 1.9348713666142765e-05, "loss": 0.1092, "step": 6197 }, { "epoch": 1.14, "learning_rate": 1.9348450087623684e-05, "loss": 0.0975, "step": 6198 }, { "epoch": 1.14, "learning_rate": 1.934818645757573e-05, "loss": 0.0881, "step": 6199 }, { "epoch": 1.14, "learning_rate": 1.9347922776000346e-05, "loss": 0.1165, "step": 6200 }, { "epoch": 1.14, "learning_rate": 1.9347659042898996e-05, "loss": 0.1079, "step": 6201 }, { "epoch": 1.14, "learning_rate": 1.9347395258273133e-05, "loss": 0.1072, "step": 6202 }, { "epoch": 1.14, "learning_rate": 1.9347131422124207e-05, "loss": 0.0911, "step": 6203 }, { "epoch": 1.14, "learning_rate": 1.9346867534453672e-05, "loss": 0.1081, "step": 6204 }, { "epoch": 1.14, "learning_rate": 1.9346603595262982e-05, "loss": 0.1124, "step": 6205 }, { "epoch": 1.14, "learning_rate": 1.9346339604553596e-05, "loss": 0.0939, "step": 6206 }, { "epoch": 1.14, "learning_rate": 1.934607556232697e-05, "loss": 0.1038, "step": 6207 }, { "epoch": 1.14, "learning_rate": 1.9345811468584553e-05, "loss": 0.0927, "step": 6208 }, { "epoch": 1.14, "learning_rate": 1.9345547323327805e-05, "loss": 0.1024, "step": 6209 }, { "epoch": 1.14, "learning_rate": 1.9345283126558184e-05, "loss": 0.0944, "step": 6210 }, { "epoch": 1.14, "learning_rate": 1.934501887827714e-05, "loss": 0.097, "step": 6211 }, { "epoch": 1.14, "learning_rate": 1.9344754578486134e-05, "loss": 0.1084, "step": 6212 }, { "epoch": 1.14, "learning_rate": 1.9344490227186624e-05, "loss": 0.108, "step": 6213 }, { "epoch": 1.14, "learning_rate": 1.9344225824380063e-05, "loss": 0.1068, "step": 6214 }, { "epoch": 1.14, "learning_rate": 1.934396137006791e-05, "loss": 0.1156, "step": 6215 }, { "epoch": 1.14, "learning_rate": 1.9343696864251624e-05, "loss": 0.0967, "step": 6216 }, { "epoch": 1.14, "learning_rate": 1.934343230693266e-05, "loss": 0.1049, "step": 6217 }, { "epoch": 1.14, "learning_rate": 1.934316769811248e-05, "loss": 0.102, "step": 6218 }, { "epoch": 1.14, "learning_rate": 1.9342903037792542e-05, "loss": 0.1088, "step": 6219 }, { "epoch": 1.14, "learning_rate": 1.93426383259743e-05, "loss": 0.1118, "step": 6220 }, { "epoch": 1.14, "learning_rate": 1.934237356265922e-05, "loss": 0.0962, "step": 6221 }, { "epoch": 1.14, "learning_rate": 1.9342108747848754e-05, "loss": 0.1057, "step": 6222 }, { "epoch": 1.14, "learning_rate": 1.9341843881544372e-05, "loss": 0.1101, "step": 6223 }, { "epoch": 1.14, "learning_rate": 1.9341578963747522e-05, "loss": 0.1072, "step": 6224 }, { "epoch": 1.14, "learning_rate": 1.9341313994459675e-05, "loss": 0.0828, "step": 6225 }, { "epoch": 1.14, "learning_rate": 1.9341048973682284e-05, "loss": 0.099, "step": 6226 }, { "epoch": 1.14, "learning_rate": 1.934078390141681e-05, "loss": 0.1039, "step": 6227 }, { "epoch": 1.14, "learning_rate": 1.9340518777664718e-05, "loss": 0.1001, "step": 6228 }, { "epoch": 1.14, "learning_rate": 1.9340253602427467e-05, "loss": 0.0873, "step": 6229 }, { "epoch": 1.14, "learning_rate": 1.933998837570652e-05, "loss": 0.1134, "step": 6230 }, { "epoch": 1.14, "learning_rate": 1.9339723097503338e-05, "loss": 0.0929, "step": 6231 }, { "epoch": 1.14, "learning_rate": 1.9339457767819384e-05, "loss": 0.0925, "step": 6232 }, { "epoch": 1.14, "learning_rate": 1.933919238665612e-05, "loss": 0.1042, "step": 6233 }, { "epoch": 1.14, "learning_rate": 1.933892695401501e-05, "loss": 0.1152, "step": 6234 }, { "epoch": 1.14, "learning_rate": 1.9338661469897513e-05, "loss": 0.0988, "step": 6235 }, { "epoch": 1.14, "learning_rate": 1.9338395934305097e-05, "loss": 0.0856, "step": 6236 }, { "epoch": 1.14, "learning_rate": 1.9338130347239225e-05, "loss": 0.0947, "step": 6237 }, { "epoch": 1.14, "learning_rate": 1.933786470870136e-05, "loss": 0.1042, "step": 6238 }, { "epoch": 1.14, "learning_rate": 1.933759901869296e-05, "loss": 0.0894, "step": 6239 }, { "epoch": 1.14, "learning_rate": 1.9337333277215502e-05, "loss": 0.0862, "step": 6240 }, { "epoch": 1.14, "learning_rate": 1.9337067484270438e-05, "loss": 0.1107, "step": 6241 }, { "epoch": 1.14, "learning_rate": 1.9336801639859243e-05, "loss": 0.1145, "step": 6242 }, { "epoch": 1.14, "learning_rate": 1.933653574398338e-05, "loss": 0.0983, "step": 6243 }, { "epoch": 1.15, "learning_rate": 1.9336269796644314e-05, "loss": 0.0965, "step": 6244 }, { "epoch": 1.15, "learning_rate": 1.9336003797843505e-05, "loss": 0.0874, "step": 6245 }, { "epoch": 1.15, "learning_rate": 1.933573774758243e-05, "loss": 0.0961, "step": 6246 }, { "epoch": 1.15, "learning_rate": 1.9335471645862547e-05, "loss": 0.102, "step": 6247 }, { "epoch": 1.15, "learning_rate": 1.9335205492685324e-05, "loss": 0.094, "step": 6248 }, { "epoch": 1.15, "learning_rate": 1.9334939288052236e-05, "loss": 0.0968, "step": 6249 }, { "epoch": 1.15, "learning_rate": 1.9334673031964736e-05, "loss": 0.1013, "step": 6250 }, { "epoch": 1.15, "learning_rate": 1.9334406724424304e-05, "loss": 0.1126, "step": 6251 }, { "epoch": 1.15, "learning_rate": 1.93341403654324e-05, "loss": 0.0967, "step": 6252 }, { "epoch": 1.15, "learning_rate": 1.93338739549905e-05, "loss": 0.0979, "step": 6253 }, { "epoch": 1.15, "learning_rate": 1.9333607493100067e-05, "loss": 0.1039, "step": 6254 }, { "epoch": 1.15, "learning_rate": 1.933334097976257e-05, "loss": 0.0998, "step": 6255 }, { "epoch": 1.15, "learning_rate": 1.9333074414979478e-05, "loss": 0.1044, "step": 6256 }, { "epoch": 1.15, "learning_rate": 1.9332807798752264e-05, "loss": 0.1011, "step": 6257 }, { "epoch": 1.15, "learning_rate": 1.933254113108239e-05, "loss": 0.0945, "step": 6258 }, { "epoch": 1.15, "learning_rate": 1.9332274411971333e-05, "loss": 0.1008, "step": 6259 }, { "epoch": 1.15, "learning_rate": 1.933200764142056e-05, "loss": 0.0955, "step": 6260 }, { "epoch": 1.15, "learning_rate": 1.9331740819431545e-05, "loss": 0.102, "step": 6261 }, { "epoch": 1.15, "learning_rate": 1.9331473946005754e-05, "loss": 0.0938, "step": 6262 }, { "epoch": 1.15, "learning_rate": 1.933120702114466e-05, "loss": 0.1094, "step": 6263 }, { "epoch": 1.15, "learning_rate": 1.9330940044849735e-05, "loss": 0.1071, "step": 6264 }, { "epoch": 1.15, "learning_rate": 1.933067301712245e-05, "loss": 0.0954, "step": 6265 }, { "epoch": 1.15, "learning_rate": 1.933040593796428e-05, "loss": 0.0921, "step": 6266 }, { "epoch": 1.15, "learning_rate": 1.9330138807376688e-05, "loss": 0.1053, "step": 6267 }, { "epoch": 1.15, "learning_rate": 1.9329871625361155e-05, "loss": 0.0914, "step": 6268 }, { "epoch": 1.15, "learning_rate": 1.932960439191915e-05, "loss": 0.0971, "step": 6269 }, { "epoch": 1.15, "learning_rate": 1.9329337107052147e-05, "loss": 0.097, "step": 6270 }, { "epoch": 1.15, "learning_rate": 1.9329069770761618e-05, "loss": 0.0876, "step": 6271 }, { "epoch": 1.15, "learning_rate": 1.9328802383049043e-05, "loss": 0.0875, "step": 6272 }, { "epoch": 1.15, "learning_rate": 1.9328534943915887e-05, "loss": 0.089, "step": 6273 }, { "epoch": 1.15, "learning_rate": 1.932826745336363e-05, "loss": 0.0982, "step": 6274 }, { "epoch": 1.15, "learning_rate": 1.932799991139374e-05, "loss": 0.1024, "step": 6275 }, { "epoch": 1.15, "learning_rate": 1.93277323180077e-05, "loss": 0.1054, "step": 6276 }, { "epoch": 1.15, "learning_rate": 1.932746467320698e-05, "loss": 0.1054, "step": 6277 }, { "epoch": 1.15, "learning_rate": 1.9327196976993056e-05, "loss": 0.1, "step": 6278 }, { "epoch": 1.15, "learning_rate": 1.9326929229367404e-05, "loss": 0.0961, "step": 6279 }, { "epoch": 1.15, "learning_rate": 1.9326661430331497e-05, "loss": 0.1072, "step": 6280 }, { "epoch": 1.15, "learning_rate": 1.9326393579886815e-05, "loss": 0.0928, "step": 6281 }, { "epoch": 1.15, "learning_rate": 1.9326125678034834e-05, "loss": 0.1135, "step": 6282 }, { "epoch": 1.15, "learning_rate": 1.932585772477703e-05, "loss": 0.099, "step": 6283 }, { "epoch": 1.15, "learning_rate": 1.9325589720114882e-05, "loss": 0.1126, "step": 6284 }, { "epoch": 1.15, "learning_rate": 1.932532166404986e-05, "loss": 0.1039, "step": 6285 }, { "epoch": 1.15, "learning_rate": 1.932505355658345e-05, "loss": 0.1082, "step": 6286 }, { "epoch": 1.15, "learning_rate": 1.9324785397717125e-05, "loss": 0.1056, "step": 6287 }, { "epoch": 1.15, "learning_rate": 1.9324517187452363e-05, "loss": 0.1042, "step": 6288 }, { "epoch": 1.15, "learning_rate": 1.932424892579065e-05, "loss": 0.0949, "step": 6289 }, { "epoch": 1.15, "learning_rate": 1.932398061273345e-05, "loss": 0.0861, "step": 6290 }, { "epoch": 1.15, "learning_rate": 1.9323712248282252e-05, "loss": 0.1088, "step": 6291 }, { "epoch": 1.15, "learning_rate": 1.932344383243854e-05, "loss": 0.1005, "step": 6292 }, { "epoch": 1.15, "learning_rate": 1.932317536520378e-05, "loss": 0.1003, "step": 6293 }, { "epoch": 1.15, "learning_rate": 1.9322906846579467e-05, "loss": 0.0996, "step": 6294 }, { "epoch": 1.15, "learning_rate": 1.9322638276567066e-05, "loss": 0.0932, "step": 6295 }, { "epoch": 1.15, "learning_rate": 1.9322369655168067e-05, "loss": 0.1177, "step": 6296 }, { "epoch": 1.15, "learning_rate": 1.9322100982383946e-05, "loss": 0.1064, "step": 6297 }, { "epoch": 1.15, "learning_rate": 1.9321832258216188e-05, "loss": 0.1012, "step": 6298 }, { "epoch": 1.16, "learning_rate": 1.9321563482666272e-05, "loss": 0.0984, "step": 6299 }, { "epoch": 1.16, "learning_rate": 1.932129465573568e-05, "loss": 0.1033, "step": 6300 }, { "epoch": 1.16, "learning_rate": 1.9321025777425892e-05, "loss": 0.1043, "step": 6301 }, { "epoch": 1.16, "learning_rate": 1.9320756847738393e-05, "loss": 0.1105, "step": 6302 }, { "epoch": 1.16, "learning_rate": 1.9320487866674666e-05, "loss": 0.1023, "step": 6303 }, { "epoch": 1.16, "learning_rate": 1.9320218834236185e-05, "loss": 0.1073, "step": 6304 }, { "epoch": 1.16, "learning_rate": 1.9319949750424444e-05, "loss": 0.1056, "step": 6305 }, { "epoch": 1.16, "learning_rate": 1.931968061524092e-05, "loss": 0.0964, "step": 6306 }, { "epoch": 1.16, "learning_rate": 1.9319411428687102e-05, "loss": 0.1003, "step": 6307 }, { "epoch": 1.16, "learning_rate": 1.9319142190764465e-05, "loss": 0.1012, "step": 6308 }, { "epoch": 1.16, "learning_rate": 1.93188729014745e-05, "loss": 0.1233, "step": 6309 }, { "epoch": 1.16, "learning_rate": 1.9318603560818688e-05, "loss": 0.0893, "step": 6310 }, { "epoch": 1.16, "learning_rate": 1.931833416879852e-05, "loss": 0.1071, "step": 6311 }, { "epoch": 1.16, "learning_rate": 1.931806472541547e-05, "loss": 0.097, "step": 6312 }, { "epoch": 1.16, "learning_rate": 1.9317795230671032e-05, "loss": 0.0999, "step": 6313 }, { "epoch": 1.16, "learning_rate": 1.9317525684566686e-05, "loss": 0.0936, "step": 6314 }, { "epoch": 1.16, "learning_rate": 1.931725608710392e-05, "loss": 0.1016, "step": 6315 }, { "epoch": 1.16, "learning_rate": 1.9316986438284222e-05, "loss": 0.0998, "step": 6316 }, { "epoch": 1.16, "learning_rate": 1.931671673810908e-05, "loss": 0.1, "step": 6317 }, { "epoch": 1.16, "learning_rate": 1.9316446986579968e-05, "loss": 0.1013, "step": 6318 }, { "epoch": 1.16, "learning_rate": 1.931617718369839e-05, "loss": 0.1042, "step": 6319 }, { "epoch": 1.16, "learning_rate": 1.9315907329465823e-05, "loss": 0.0955, "step": 6320 }, { "epoch": 1.16, "learning_rate": 1.9315637423883754e-05, "loss": 0.1116, "step": 6321 }, { "epoch": 1.16, "learning_rate": 1.9315367466953673e-05, "loss": 0.1035, "step": 6322 }, { "epoch": 1.16, "learning_rate": 1.931509745867707e-05, "loss": 0.0937, "step": 6323 }, { "epoch": 1.16, "learning_rate": 1.9314827399055433e-05, "loss": 0.0947, "step": 6324 }, { "epoch": 1.16, "learning_rate": 1.931455728809025e-05, "loss": 0.1031, "step": 6325 }, { "epoch": 1.16, "learning_rate": 1.9314287125783004e-05, "loss": 0.0974, "step": 6326 }, { "epoch": 1.16, "learning_rate": 1.9314016912135194e-05, "loss": 0.0907, "step": 6327 }, { "epoch": 1.16, "learning_rate": 1.9313746647148303e-05, "loss": 0.0973, "step": 6328 }, { "epoch": 1.16, "learning_rate": 1.931347633082382e-05, "loss": 0.1025, "step": 6329 }, { "epoch": 1.16, "learning_rate": 1.9313205963163238e-05, "loss": 0.1064, "step": 6330 }, { "epoch": 1.16, "learning_rate": 1.931293554416805e-05, "loss": 0.1046, "step": 6331 }, { "epoch": 1.16, "learning_rate": 1.931266507383974e-05, "loss": 0.1022, "step": 6332 }, { "epoch": 1.16, "learning_rate": 1.9312394552179804e-05, "loss": 0.0989, "step": 6333 }, { "epoch": 1.16, "learning_rate": 1.931212397918973e-05, "loss": 0.1102, "step": 6334 }, { "epoch": 1.16, "learning_rate": 1.931185335487101e-05, "loss": 0.091, "step": 6335 }, { "epoch": 1.16, "learning_rate": 1.9311582679225137e-05, "loss": 0.1087, "step": 6336 }, { "epoch": 1.16, "learning_rate": 1.9311311952253602e-05, "loss": 0.1091, "step": 6337 }, { "epoch": 1.16, "learning_rate": 1.9311041173957895e-05, "loss": 0.1105, "step": 6338 }, { "epoch": 1.16, "learning_rate": 1.9310770344339516e-05, "loss": 0.0983, "step": 6339 }, { "epoch": 1.16, "learning_rate": 1.9310499463399946e-05, "loss": 0.1024, "step": 6340 }, { "epoch": 1.16, "learning_rate": 1.931022853114069e-05, "loss": 0.0953, "step": 6341 }, { "epoch": 1.16, "learning_rate": 1.9309957547563235e-05, "loss": 0.1121, "step": 6342 }, { "epoch": 1.16, "learning_rate": 1.9309686512669077e-05, "loss": 0.1123, "step": 6343 }, { "epoch": 1.16, "learning_rate": 1.9309415426459706e-05, "loss": 0.1148, "step": 6344 }, { "epoch": 1.16, "learning_rate": 1.930914428893662e-05, "loss": 0.1013, "step": 6345 }, { "epoch": 1.16, "learning_rate": 1.9308873100101314e-05, "loss": 0.1139, "step": 6346 }, { "epoch": 1.16, "learning_rate": 1.930860185995528e-05, "loss": 0.0965, "step": 6347 }, { "epoch": 1.16, "learning_rate": 1.9308330568500015e-05, "loss": 0.1, "step": 6348 }, { "epoch": 1.16, "learning_rate": 1.9308059225737015e-05, "loss": 0.1067, "step": 6349 }, { "epoch": 1.16, "learning_rate": 1.9307787831667772e-05, "loss": 0.1078, "step": 6350 }, { "epoch": 1.16, "learning_rate": 1.9307516386293785e-05, "loss": 0.1058, "step": 6351 }, { "epoch": 1.16, "learning_rate": 1.930724488961655e-05, "loss": 0.1046, "step": 6352 }, { "epoch": 1.17, "learning_rate": 1.9306973341637566e-05, "loss": 0.1032, "step": 6353 }, { "epoch": 1.17, "learning_rate": 1.9306701742358322e-05, "loss": 0.097, "step": 6354 }, { "epoch": 1.17, "learning_rate": 1.930643009178032e-05, "loss": 0.1023, "step": 6355 }, { "epoch": 1.17, "learning_rate": 1.9306158389905065e-05, "loss": 0.1028, "step": 6356 }, { "epoch": 1.17, "learning_rate": 1.930588663673404e-05, "loss": 0.0998, "step": 6357 }, { "epoch": 1.17, "learning_rate": 1.930561483226875e-05, "loss": 0.1001, "step": 6358 }, { "epoch": 1.17, "learning_rate": 1.9305342976510696e-05, "loss": 0.1085, "step": 6359 }, { "epoch": 1.17, "learning_rate": 1.930507106946137e-05, "loss": 0.1034, "step": 6360 }, { "epoch": 1.17, "learning_rate": 1.9304799111122275e-05, "loss": 0.0868, "step": 6361 }, { "epoch": 1.17, "learning_rate": 1.930452710149491e-05, "loss": 0.1091, "step": 6362 }, { "epoch": 1.17, "learning_rate": 1.9304255040580777e-05, "loss": 0.0986, "step": 6363 }, { "epoch": 1.17, "learning_rate": 1.930398292838137e-05, "loss": 0.0915, "step": 6364 }, { "epoch": 1.17, "learning_rate": 1.9303710764898192e-05, "loss": 0.1054, "step": 6365 }, { "epoch": 1.17, "learning_rate": 1.9303438550132737e-05, "loss": 0.1014, "step": 6366 }, { "epoch": 1.17, "learning_rate": 1.9303166284086518e-05, "loss": 0.0977, "step": 6367 }, { "epoch": 1.17, "learning_rate": 1.9302893966761026e-05, "loss": 0.1038, "step": 6368 }, { "epoch": 1.17, "learning_rate": 1.9302621598157762e-05, "loss": 0.1162, "step": 6369 }, { "epoch": 1.17, "learning_rate": 1.930234917827823e-05, "loss": 0.1055, "step": 6370 }, { "epoch": 1.17, "learning_rate": 1.9302076707123935e-05, "loss": 0.097, "step": 6371 }, { "epoch": 1.17, "learning_rate": 1.930180418469637e-05, "loss": 0.0996, "step": 6372 }, { "epoch": 1.17, "learning_rate": 1.9301531610997045e-05, "loss": 0.0947, "step": 6373 }, { "epoch": 1.17, "learning_rate": 1.9301258986027462e-05, "loss": 0.1076, "step": 6374 }, { "epoch": 1.17, "learning_rate": 1.930098630978912e-05, "loss": 0.0966, "step": 6375 }, { "epoch": 1.17, "learning_rate": 1.9300713582283525e-05, "loss": 0.0962, "step": 6376 }, { "epoch": 1.17, "learning_rate": 1.9300440803512176e-05, "loss": 0.0988, "step": 6377 }, { "epoch": 1.17, "learning_rate": 1.930016797347658e-05, "loss": 0.111, "step": 6378 }, { "epoch": 1.17, "learning_rate": 1.929989509217824e-05, "loss": 0.1097, "step": 6379 }, { "epoch": 1.17, "learning_rate": 1.9299622159618664e-05, "loss": 0.1064, "step": 6380 }, { "epoch": 1.17, "learning_rate": 1.9299349175799348e-05, "loss": 0.103, "step": 6381 }, { "epoch": 1.17, "learning_rate": 1.9299076140721806e-05, "loss": 0.1052, "step": 6382 }, { "epoch": 1.17, "learning_rate": 1.9298803054387535e-05, "loss": 0.1064, "step": 6383 }, { "epoch": 1.17, "learning_rate": 1.9298529916798044e-05, "loss": 0.1111, "step": 6384 }, { "epoch": 1.17, "learning_rate": 1.929825672795484e-05, "loss": 0.1043, "step": 6385 }, { "epoch": 1.17, "learning_rate": 1.9297983487859425e-05, "loss": 0.0982, "step": 6386 }, { "epoch": 1.17, "learning_rate": 1.9297710196513308e-05, "loss": 0.0843, "step": 6387 }, { "epoch": 1.17, "learning_rate": 1.9297436853917998e-05, "loss": 0.1106, "step": 6388 }, { "epoch": 1.17, "learning_rate": 1.9297163460074993e-05, "loss": 0.1038, "step": 6389 }, { "epoch": 1.17, "learning_rate": 1.9296890014985806e-05, "loss": 0.1159, "step": 6390 }, { "epoch": 1.17, "learning_rate": 1.9296616518651946e-05, "loss": 0.0958, "step": 6391 }, { "epoch": 1.17, "learning_rate": 1.9296342971074917e-05, "loss": 0.1106, "step": 6392 }, { "epoch": 1.17, "learning_rate": 1.9296069372256227e-05, "loss": 0.098, "step": 6393 }, { "epoch": 1.17, "learning_rate": 1.9295795722197383e-05, "loss": 0.09, "step": 6394 }, { "epoch": 1.17, "learning_rate": 1.92955220208999e-05, "loss": 0.0987, "step": 6395 }, { "epoch": 1.17, "learning_rate": 1.929524826836528e-05, "loss": 0.1079, "step": 6396 }, { "epoch": 1.17, "learning_rate": 1.9294974464595033e-05, "loss": 0.1153, "step": 6397 }, { "epoch": 1.17, "learning_rate": 1.9294700609590666e-05, "loss": 0.0997, "step": 6398 }, { "epoch": 1.17, "learning_rate": 1.9294426703353697e-05, "loss": 0.0948, "step": 6399 }, { "epoch": 1.17, "learning_rate": 1.9294152745885627e-05, "loss": 0.1007, "step": 6400 }, { "epoch": 1.17, "learning_rate": 1.929387873718797e-05, "loss": 0.0972, "step": 6401 }, { "epoch": 1.17, "learning_rate": 1.9293604677262237e-05, "loss": 0.0998, "step": 6402 }, { "epoch": 1.17, "learning_rate": 1.9293330566109936e-05, "loss": 0.1033, "step": 6403 }, { "epoch": 1.17, "learning_rate": 1.929305640373258e-05, "loss": 0.1112, "step": 6404 }, { "epoch": 1.17, "learning_rate": 1.9292782190131677e-05, "loss": 0.1183, "step": 6405 }, { "epoch": 1.17, "learning_rate": 1.9292507925308744e-05, "loss": 0.1043, "step": 6406 }, { "epoch": 1.17, "learning_rate": 1.9292233609265288e-05, "loss": 0.0995, "step": 6407 }, { "epoch": 1.18, "learning_rate": 1.929195924200282e-05, "loss": 0.1008, "step": 6408 }, { "epoch": 1.18, "learning_rate": 1.929168482352286e-05, "loss": 0.099, "step": 6409 }, { "epoch": 1.18, "learning_rate": 1.9291410353826913e-05, "loss": 0.098, "step": 6410 }, { "epoch": 1.18, "learning_rate": 1.9291135832916494e-05, "loss": 0.1067, "step": 6411 }, { "epoch": 1.18, "learning_rate": 1.929086126079312e-05, "loss": 0.1009, "step": 6412 }, { "epoch": 1.18, "learning_rate": 1.9290586637458297e-05, "loss": 0.1053, "step": 6413 }, { "epoch": 1.18, "learning_rate": 1.9290311962913547e-05, "loss": 0.0914, "step": 6414 }, { "epoch": 1.18, "learning_rate": 1.9290037237160377e-05, "loss": 0.0981, "step": 6415 }, { "epoch": 1.18, "learning_rate": 1.9289762460200304e-05, "loss": 0.0994, "step": 6416 }, { "epoch": 1.18, "learning_rate": 1.9289487632034843e-05, "loss": 0.1037, "step": 6417 }, { "epoch": 1.18, "learning_rate": 1.9289212752665506e-05, "loss": 0.0983, "step": 6418 }, { "epoch": 1.18, "learning_rate": 1.9288937822093813e-05, "loss": 0.0999, "step": 6419 }, { "epoch": 1.18, "learning_rate": 1.9288662840321278e-05, "loss": 0.1143, "step": 6420 }, { "epoch": 1.18, "learning_rate": 1.9288387807349417e-05, "loss": 0.1019, "step": 6421 }, { "epoch": 1.18, "learning_rate": 1.928811272317974e-05, "loss": 0.0993, "step": 6422 }, { "epoch": 1.18, "learning_rate": 1.928783758781377e-05, "loss": 0.0874, "step": 6423 }, { "epoch": 1.18, "learning_rate": 1.9287562401253023e-05, "loss": 0.0981, "step": 6424 }, { "epoch": 1.18, "learning_rate": 1.9287287163499014e-05, "loss": 0.1071, "step": 6425 }, { "epoch": 1.18, "learning_rate": 1.928701187455326e-05, "loss": 0.1071, "step": 6426 }, { "epoch": 1.18, "learning_rate": 1.9286736534417282e-05, "loss": 0.0994, "step": 6427 }, { "epoch": 1.18, "learning_rate": 1.9286461143092592e-05, "loss": 0.1035, "step": 6428 }, { "epoch": 1.18, "learning_rate": 1.9286185700580714e-05, "loss": 0.1061, "step": 6429 }, { "epoch": 1.18, "learning_rate": 1.9285910206883157e-05, "loss": 0.0972, "step": 6430 }, { "epoch": 1.18, "learning_rate": 1.928563466200145e-05, "loss": 0.099, "step": 6431 }, { "epoch": 1.18, "learning_rate": 1.9285359065937105e-05, "loss": 0.0988, "step": 6432 }, { "epoch": 1.18, "learning_rate": 1.928508341869164e-05, "loss": 0.1007, "step": 6433 }, { "epoch": 1.18, "learning_rate": 1.9284807720266584e-05, "loss": 0.0976, "step": 6434 }, { "epoch": 1.18, "learning_rate": 1.9284531970663447e-05, "loss": 0.1072, "step": 6435 }, { "epoch": 1.18, "learning_rate": 1.9284256169883754e-05, "loss": 0.1038, "step": 6436 }, { "epoch": 1.18, "learning_rate": 1.928398031792902e-05, "loss": 0.1026, "step": 6437 }, { "epoch": 1.18, "learning_rate": 1.9283704414800773e-05, "loss": 0.1039, "step": 6438 }, { "epoch": 1.18, "learning_rate": 1.928342846050053e-05, "loss": 0.1007, "step": 6439 }, { "epoch": 1.18, "learning_rate": 1.9283152455029807e-05, "loss": 0.098, "step": 6440 }, { "epoch": 1.18, "learning_rate": 1.9282876398390135e-05, "loss": 0.0949, "step": 6441 }, { "epoch": 1.18, "learning_rate": 1.928260029058303e-05, "loss": 0.1056, "step": 6442 }, { "epoch": 1.18, "learning_rate": 1.928232413161001e-05, "loss": 0.0987, "step": 6443 }, { "epoch": 1.18, "learning_rate": 1.928204792147261e-05, "loss": 0.1105, "step": 6444 }, { "epoch": 1.18, "learning_rate": 1.9281771660172336e-05, "loss": 0.0964, "step": 6445 }, { "epoch": 1.18, "learning_rate": 1.9281495347710723e-05, "loss": 0.0898, "step": 6446 }, { "epoch": 1.18, "learning_rate": 1.9281218984089292e-05, "loss": 0.0963, "step": 6447 }, { "epoch": 1.18, "learning_rate": 1.9280942569309565e-05, "loss": 0.0992, "step": 6448 }, { "epoch": 1.18, "learning_rate": 1.928066610337306e-05, "loss": 0.0956, "step": 6449 }, { "epoch": 1.18, "learning_rate": 1.928038958628131e-05, "loss": 0.1077, "step": 6450 }, { "epoch": 1.18, "learning_rate": 1.9280113018035833e-05, "loss": 0.0984, "step": 6451 }, { "epoch": 1.18, "learning_rate": 1.927983639863816e-05, "loss": 0.0816, "step": 6452 }, { "epoch": 1.18, "learning_rate": 1.9279559728089808e-05, "loss": 0.0963, "step": 6453 }, { "epoch": 1.18, "learning_rate": 1.9279283006392304e-05, "loss": 0.084, "step": 6454 }, { "epoch": 1.18, "learning_rate": 1.9279006233547176e-05, "loss": 0.1057, "step": 6455 }, { "epoch": 1.18, "learning_rate": 1.927872940955595e-05, "loss": 0.1043, "step": 6456 }, { "epoch": 1.18, "learning_rate": 1.9278452534420148e-05, "loss": 0.1035, "step": 6457 }, { "epoch": 1.18, "learning_rate": 1.92781756081413e-05, "loss": 0.1022, "step": 6458 }, { "epoch": 1.18, "learning_rate": 1.9277898630720928e-05, "loss": 0.0961, "step": 6459 }, { "epoch": 1.18, "learning_rate": 1.9277621602160565e-05, "loss": 0.1021, "step": 6460 }, { "epoch": 1.18, "learning_rate": 1.9277344522461732e-05, "loss": 0.106, "step": 6461 }, { "epoch": 1.19, "learning_rate": 1.927706739162596e-05, "loss": 0.0995, "step": 6462 }, { "epoch": 1.19, "learning_rate": 1.9276790209654774e-05, "loss": 0.0992, "step": 6463 }, { "epoch": 1.19, "learning_rate": 1.9276512976549703e-05, "loss": 0.1026, "step": 6464 }, { "epoch": 1.19, "learning_rate": 1.9276235692312272e-05, "loss": 0.099, "step": 6465 }, { "epoch": 1.19, "learning_rate": 1.927595835694402e-05, "loss": 0.1074, "step": 6466 }, { "epoch": 1.19, "learning_rate": 1.9275680970446462e-05, "loss": 0.1074, "step": 6467 }, { "epoch": 1.19, "learning_rate": 1.9275403532821136e-05, "loss": 0.0986, "step": 6468 }, { "epoch": 1.19, "learning_rate": 1.9275126044069568e-05, "loss": 0.1194, "step": 6469 }, { "epoch": 1.19, "learning_rate": 1.9274848504193287e-05, "loss": 0.1028, "step": 6470 }, { "epoch": 1.19, "learning_rate": 1.9274570913193825e-05, "loss": 0.098, "step": 6471 }, { "epoch": 1.19, "learning_rate": 1.9274293271072712e-05, "loss": 0.0987, "step": 6472 }, { "epoch": 1.19, "learning_rate": 1.9274015577831477e-05, "loss": 0.0979, "step": 6473 }, { "epoch": 1.19, "learning_rate": 1.9273737833471648e-05, "loss": 0.0982, "step": 6474 }, { "epoch": 1.19, "learning_rate": 1.927346003799476e-05, "loss": 0.0964, "step": 6475 }, { "epoch": 1.19, "learning_rate": 1.9273182191402342e-05, "loss": 0.1074, "step": 6476 }, { "epoch": 1.19, "learning_rate": 1.9272904293695926e-05, "loss": 0.0992, "step": 6477 }, { "epoch": 1.19, "learning_rate": 1.9272626344877047e-05, "loss": 0.1039, "step": 6478 }, { "epoch": 1.19, "learning_rate": 1.9272348344947235e-05, "loss": 0.1237, "step": 6479 }, { "epoch": 1.19, "learning_rate": 1.927207029390802e-05, "loss": 0.1077, "step": 6480 }, { "epoch": 1.19, "learning_rate": 1.9271792191760936e-05, "loss": 0.1017, "step": 6481 }, { "epoch": 1.19, "learning_rate": 1.9271514038507515e-05, "loss": 0.0871, "step": 6482 }, { "epoch": 1.19, "learning_rate": 1.9271235834149293e-05, "loss": 0.081, "step": 6483 }, { "epoch": 1.19, "learning_rate": 1.9270957578687805e-05, "loss": 0.0861, "step": 6484 }, { "epoch": 1.19, "learning_rate": 1.927067927212458e-05, "loss": 0.0972, "step": 6485 }, { "epoch": 1.19, "learning_rate": 1.927040091446115e-05, "loss": 0.1115, "step": 6486 }, { "epoch": 1.19, "learning_rate": 1.9270122505699053e-05, "loss": 0.1056, "step": 6487 }, { "epoch": 1.19, "learning_rate": 1.9269844045839827e-05, "loss": 0.0884, "step": 6488 }, { "epoch": 1.19, "learning_rate": 1.9269565534885003e-05, "loss": 0.1065, "step": 6489 }, { "epoch": 1.19, "learning_rate": 1.9269286972836117e-05, "loss": 0.1085, "step": 6490 }, { "epoch": 1.19, "learning_rate": 1.92690083596947e-05, "loss": 0.099, "step": 6491 }, { "epoch": 1.19, "learning_rate": 1.9268729695462297e-05, "loss": 0.1051, "step": 6492 }, { "epoch": 1.19, "learning_rate": 1.9268450980140435e-05, "loss": 0.114, "step": 6493 }, { "epoch": 1.19, "learning_rate": 1.9268172213730654e-05, "loss": 0.1017, "step": 6494 }, { "epoch": 1.19, "learning_rate": 1.9267893396234494e-05, "loss": 0.1058, "step": 6495 }, { "epoch": 1.19, "learning_rate": 1.926761452765349e-05, "loss": 0.1018, "step": 6496 }, { "epoch": 1.19, "learning_rate": 1.9267335607989174e-05, "loss": 0.1052, "step": 6497 }, { "epoch": 1.19, "learning_rate": 1.9267056637243083e-05, "loss": 0.0865, "step": 6498 }, { "epoch": 1.19, "learning_rate": 1.9266777615416762e-05, "loss": 0.0985, "step": 6499 }, { "epoch": 1.19, "learning_rate": 1.926649854251175e-05, "loss": 0.1088, "step": 6500 }, { "epoch": 1.19, "learning_rate": 1.9266219418529578e-05, "loss": 0.1074, "step": 6501 }, { "epoch": 1.19, "learning_rate": 1.9265940243471785e-05, "loss": 0.0984, "step": 6502 }, { "epoch": 1.19, "learning_rate": 1.9265661017339914e-05, "loss": 0.0981, "step": 6503 }, { "epoch": 1.19, "learning_rate": 1.9265381740135504e-05, "loss": 0.0939, "step": 6504 }, { "epoch": 1.19, "learning_rate": 1.926510241186009e-05, "loss": 0.0891, "step": 6505 }, { "epoch": 1.19, "learning_rate": 1.9264823032515218e-05, "loss": 0.098, "step": 6506 }, { "epoch": 1.19, "learning_rate": 1.9264543602102423e-05, "loss": 0.1107, "step": 6507 }, { "epoch": 1.19, "learning_rate": 1.9264264120623247e-05, "loss": 0.0902, "step": 6508 }, { "epoch": 1.19, "learning_rate": 1.9263984588079227e-05, "loss": 0.0984, "step": 6509 }, { "epoch": 1.19, "learning_rate": 1.9263705004471907e-05, "loss": 0.0956, "step": 6510 }, { "epoch": 1.19, "learning_rate": 1.926342536980283e-05, "loss": 0.1027, "step": 6511 }, { "epoch": 1.19, "learning_rate": 1.9263145684073537e-05, "loss": 0.1111, "step": 6512 }, { "epoch": 1.19, "learning_rate": 1.9262865947285567e-05, "loss": 0.0862, "step": 6513 }, { "epoch": 1.19, "learning_rate": 1.9262586159440463e-05, "loss": 0.1042, "step": 6514 }, { "epoch": 1.19, "learning_rate": 1.9262306320539764e-05, "loss": 0.0978, "step": 6515 }, { "epoch": 1.19, "learning_rate": 1.926202643058502e-05, "loss": 0.1077, "step": 6516 }, { "epoch": 1.2, "learning_rate": 1.9261746489577767e-05, "loss": 0.0985, "step": 6517 }, { "epoch": 1.2, "learning_rate": 1.926146649751955e-05, "loss": 0.1086, "step": 6518 }, { "epoch": 1.2, "learning_rate": 1.9261186454411912e-05, "loss": 0.1129, "step": 6519 }, { "epoch": 1.2, "learning_rate": 1.92609063602564e-05, "loss": 0.0992, "step": 6520 }, { "epoch": 1.2, "learning_rate": 1.9260626215054555e-05, "loss": 0.0983, "step": 6521 }, { "epoch": 1.2, "learning_rate": 1.926034601880792e-05, "loss": 0.0958, "step": 6522 }, { "epoch": 1.2, "learning_rate": 1.926006577151804e-05, "loss": 0.1112, "step": 6523 }, { "epoch": 1.2, "learning_rate": 1.9259785473186462e-05, "loss": 0.1086, "step": 6524 }, { "epoch": 1.2, "learning_rate": 1.9259505123814727e-05, "loss": 0.1038, "step": 6525 }, { "epoch": 1.2, "learning_rate": 1.9259224723404385e-05, "loss": 0.1063, "step": 6526 }, { "epoch": 1.2, "learning_rate": 1.925894427195698e-05, "loss": 0.0969, "step": 6527 }, { "epoch": 1.2, "learning_rate": 1.9258663769474058e-05, "loss": 0.1026, "step": 6528 }, { "epoch": 1.2, "learning_rate": 1.925838321595716e-05, "loss": 0.0944, "step": 6529 }, { "epoch": 1.2, "learning_rate": 1.925810261140784e-05, "loss": 0.0953, "step": 6530 }, { "epoch": 1.2, "learning_rate": 1.9257821955827642e-05, "loss": 0.1084, "step": 6531 }, { "epoch": 1.2, "learning_rate": 1.9257541249218112e-05, "loss": 0.1106, "step": 6532 }, { "epoch": 1.2, "learning_rate": 1.92572604915808e-05, "loss": 0.091, "step": 6533 }, { "epoch": 1.2, "learning_rate": 1.925697968291725e-05, "loss": 0.0952, "step": 6534 }, { "epoch": 1.2, "learning_rate": 1.9256698823229007e-05, "loss": 0.1048, "step": 6535 }, { "epoch": 1.2, "learning_rate": 1.9256417912517627e-05, "loss": 0.1049, "step": 6536 }, { "epoch": 1.2, "learning_rate": 1.9256136950784653e-05, "loss": 0.0947, "step": 6537 }, { "epoch": 1.2, "learning_rate": 1.9255855938031635e-05, "loss": 0.1062, "step": 6538 }, { "epoch": 1.2, "learning_rate": 1.9255574874260125e-05, "loss": 0.1049, "step": 6539 }, { "epoch": 1.2, "learning_rate": 1.9255293759471664e-05, "loss": 0.0995, "step": 6540 }, { "epoch": 1.2, "learning_rate": 1.9255012593667812e-05, "loss": 0.1037, "step": 6541 }, { "epoch": 1.2, "learning_rate": 1.9254731376850114e-05, "loss": 0.1075, "step": 6542 }, { "epoch": 1.2, "learning_rate": 1.9254450109020116e-05, "loss": 0.1213, "step": 6543 }, { "epoch": 1.2, "learning_rate": 1.9254168790179376e-05, "loss": 0.0973, "step": 6544 }, { "epoch": 1.2, "learning_rate": 1.925388742032944e-05, "loss": 0.1026, "step": 6545 }, { "epoch": 1.2, "learning_rate": 1.925360599947186e-05, "loss": 0.1028, "step": 6546 }, { "epoch": 1.2, "learning_rate": 1.9253324527608183e-05, "loss": 0.0918, "step": 6547 }, { "epoch": 1.2, "learning_rate": 1.9253043004739967e-05, "loss": 0.104, "step": 6548 }, { "epoch": 1.2, "learning_rate": 1.9252761430868763e-05, "loss": 0.0976, "step": 6549 }, { "epoch": 1.2, "learning_rate": 1.925247980599612e-05, "loss": 0.0902, "step": 6550 }, { "epoch": 1.2, "learning_rate": 1.9252198130123593e-05, "loss": 0.0931, "step": 6551 }, { "epoch": 1.2, "learning_rate": 1.925191640325273e-05, "loss": 0.1028, "step": 6552 }, { "epoch": 1.2, "learning_rate": 1.925163462538509e-05, "loss": 0.0884, "step": 6553 }, { "epoch": 1.2, "learning_rate": 1.925135279652222e-05, "loss": 0.0932, "step": 6554 }, { "epoch": 1.2, "learning_rate": 1.925107091666568e-05, "loss": 0.0967, "step": 6555 }, { "epoch": 1.2, "learning_rate": 1.9250788985817018e-05, "loss": 0.1151, "step": 6556 }, { "epoch": 1.2, "learning_rate": 1.9250507003977793e-05, "loss": 0.1032, "step": 6557 }, { "epoch": 1.2, "learning_rate": 1.9250224971149552e-05, "loss": 0.1109, "step": 6558 }, { "epoch": 1.2, "learning_rate": 1.924994288733386e-05, "loss": 0.1135, "step": 6559 }, { "epoch": 1.2, "learning_rate": 1.924966075253226e-05, "loss": 0.0939, "step": 6560 }, { "epoch": 1.2, "learning_rate": 1.924937856674632e-05, "loss": 0.0979, "step": 6561 }, { "epoch": 1.2, "learning_rate": 1.9249096329977583e-05, "loss": 0.1088, "step": 6562 }, { "epoch": 1.2, "learning_rate": 1.9248814042227614e-05, "loss": 0.1239, "step": 6563 }, { "epoch": 1.2, "learning_rate": 1.924853170349796e-05, "loss": 0.1018, "step": 6564 }, { "epoch": 1.2, "learning_rate": 1.9248249313790186e-05, "loss": 0.0916, "step": 6565 }, { "epoch": 1.2, "learning_rate": 1.9247966873105848e-05, "loss": 0.1106, "step": 6566 }, { "epoch": 1.2, "learning_rate": 1.9247684381446497e-05, "loss": 0.098, "step": 6567 }, { "epoch": 1.2, "learning_rate": 1.9247401838813693e-05, "loss": 0.1051, "step": 6568 }, { "epoch": 1.2, "learning_rate": 1.9247119245208994e-05, "loss": 0.1003, "step": 6569 }, { "epoch": 1.2, "learning_rate": 1.9246836600633955e-05, "loss": 0.1008, "step": 6570 }, { "epoch": 1.21, "learning_rate": 1.924655390509014e-05, "loss": 0.1058, "step": 6571 }, { "epoch": 1.21, "learning_rate": 1.92462711585791e-05, "loss": 0.102, "step": 6572 }, { "epoch": 1.21, "learning_rate": 1.92459883611024e-05, "loss": 0.0992, "step": 6573 }, { "epoch": 1.21, "learning_rate": 1.9245705512661594e-05, "loss": 0.0896, "step": 6574 }, { "epoch": 1.21, "learning_rate": 1.924542261325824e-05, "loss": 0.0951, "step": 6575 }, { "epoch": 1.21, "learning_rate": 1.92451396628939e-05, "loss": 0.0986, "step": 6576 }, { "epoch": 1.21, "learning_rate": 1.9244856661570136e-05, "loss": 0.0965, "step": 6577 }, { "epoch": 1.21, "learning_rate": 1.9244573609288507e-05, "loss": 0.1076, "step": 6578 }, { "epoch": 1.21, "learning_rate": 1.924429050605057e-05, "loss": 0.0857, "step": 6579 }, { "epoch": 1.21, "learning_rate": 1.9244007351857886e-05, "loss": 0.1074, "step": 6580 }, { "epoch": 1.21, "learning_rate": 1.9243724146712018e-05, "loss": 0.1138, "step": 6581 }, { "epoch": 1.21, "learning_rate": 1.9243440890614526e-05, "loss": 0.1086, "step": 6582 }, { "epoch": 1.21, "learning_rate": 1.9243157583566972e-05, "loss": 0.0959, "step": 6583 }, { "epoch": 1.21, "learning_rate": 1.9242874225570912e-05, "loss": 0.096, "step": 6584 }, { "epoch": 1.21, "learning_rate": 1.924259081662792e-05, "loss": 0.1096, "step": 6585 }, { "epoch": 1.21, "learning_rate": 1.9242307356739547e-05, "loss": 0.1044, "step": 6586 }, { "epoch": 1.21, "learning_rate": 1.9242023845907362e-05, "loss": 0.1045, "step": 6587 }, { "epoch": 1.21, "learning_rate": 1.924174028413292e-05, "loss": 0.1089, "step": 6588 }, { "epoch": 1.21, "learning_rate": 1.9241456671417794e-05, "loss": 0.1119, "step": 6589 }, { "epoch": 1.21, "learning_rate": 1.924117300776354e-05, "loss": 0.0926, "step": 6590 }, { "epoch": 1.21, "learning_rate": 1.9240889293171724e-05, "loss": 0.1098, "step": 6591 }, { "epoch": 1.21, "learning_rate": 1.9240605527643908e-05, "loss": 0.0883, "step": 6592 }, { "epoch": 1.21, "learning_rate": 1.924032171118166e-05, "loss": 0.0953, "step": 6593 }, { "epoch": 1.21, "learning_rate": 1.9240037843786544e-05, "loss": 0.0926, "step": 6594 }, { "epoch": 1.21, "learning_rate": 1.9239753925460117e-05, "loss": 0.1051, "step": 6595 }, { "epoch": 1.21, "learning_rate": 1.9239469956203957e-05, "loss": 0.1084, "step": 6596 }, { "epoch": 1.21, "learning_rate": 1.923918593601962e-05, "loss": 0.1089, "step": 6597 }, { "epoch": 1.21, "learning_rate": 1.923890186490867e-05, "loss": 0.0938, "step": 6598 }, { "epoch": 1.21, "learning_rate": 1.923861774287268e-05, "loss": 0.098, "step": 6599 }, { "epoch": 1.21, "learning_rate": 1.9238333569913212e-05, "loss": 0.0985, "step": 6600 }, { "epoch": 1.21, "learning_rate": 1.9238049346031833e-05, "loss": 0.1157, "step": 6601 }, { "epoch": 1.21, "learning_rate": 1.923776507123011e-05, "loss": 0.0993, "step": 6602 }, { "epoch": 1.21, "learning_rate": 1.923748074550961e-05, "loss": 0.1029, "step": 6603 }, { "epoch": 1.21, "learning_rate": 1.9237196368871897e-05, "loss": 0.0958, "step": 6604 }, { "epoch": 1.21, "learning_rate": 1.923691194131854e-05, "loss": 0.1041, "step": 6605 }, { "epoch": 1.21, "learning_rate": 1.9236627462851114e-05, "loss": 0.0975, "step": 6606 }, { "epoch": 1.21, "learning_rate": 1.9236342933471178e-05, "loss": 0.0951, "step": 6607 }, { "epoch": 1.21, "learning_rate": 1.9236058353180303e-05, "loss": 0.0977, "step": 6608 }, { "epoch": 1.21, "learning_rate": 1.9235773721980055e-05, "loss": 0.1088, "step": 6609 }, { "epoch": 1.21, "learning_rate": 1.923548903987201e-05, "loss": 0.1022, "step": 6610 }, { "epoch": 1.21, "learning_rate": 1.9235204306857732e-05, "loss": 0.0841, "step": 6611 }, { "epoch": 1.21, "learning_rate": 1.9234919522938792e-05, "loss": 0.1071, "step": 6612 }, { "epoch": 1.21, "learning_rate": 1.923463468811676e-05, "loss": 0.0974, "step": 6613 }, { "epoch": 1.21, "learning_rate": 1.92343498023932e-05, "loss": 0.1077, "step": 6614 }, { "epoch": 1.21, "learning_rate": 1.9234064865769693e-05, "loss": 0.1072, "step": 6615 }, { "epoch": 1.21, "learning_rate": 1.9233779878247803e-05, "loss": 0.102, "step": 6616 }, { "epoch": 1.21, "learning_rate": 1.92334948398291e-05, "loss": 0.0998, "step": 6617 }, { "epoch": 1.21, "learning_rate": 1.923320975051516e-05, "loss": 0.1062, "step": 6618 }, { "epoch": 1.21, "learning_rate": 1.923292461030755e-05, "loss": 0.0975, "step": 6619 }, { "epoch": 1.21, "learning_rate": 1.9232639419207844e-05, "loss": 0.1081, "step": 6620 }, { "epoch": 1.21, "learning_rate": 1.9232354177217612e-05, "loss": 0.0985, "step": 6621 }, { "epoch": 1.21, "learning_rate": 1.9232068884338427e-05, "loss": 0.0958, "step": 6622 }, { "epoch": 1.21, "learning_rate": 1.9231783540571864e-05, "loss": 0.0957, "step": 6623 }, { "epoch": 1.21, "learning_rate": 1.9231498145919492e-05, "loss": 0.1007, "step": 6624 }, { "epoch": 1.21, "learning_rate": 1.923121270038289e-05, "loss": 0.0916, "step": 6625 }, { "epoch": 1.22, "learning_rate": 1.9230927203963622e-05, "loss": 0.0975, "step": 6626 }, { "epoch": 1.22, "learning_rate": 1.9230641656663272e-05, "loss": 0.1018, "step": 6627 }, { "epoch": 1.22, "learning_rate": 1.9230356058483404e-05, "loss": 0.1148, "step": 6628 }, { "epoch": 1.22, "learning_rate": 1.92300704094256e-05, "loss": 0.1089, "step": 6629 }, { "epoch": 1.22, "learning_rate": 1.922978470949143e-05, "loss": 0.1088, "step": 6630 }, { "epoch": 1.22, "learning_rate": 1.922949895868247e-05, "loss": 0.114, "step": 6631 }, { "epoch": 1.22, "learning_rate": 1.92292131570003e-05, "loss": 0.1073, "step": 6632 }, { "epoch": 1.22, "learning_rate": 1.922892730444649e-05, "loss": 0.0912, "step": 6633 }, { "epoch": 1.22, "learning_rate": 1.9228641401022612e-05, "loss": 0.0949, "step": 6634 }, { "epoch": 1.22, "learning_rate": 1.922835544673025e-05, "loss": 0.0959, "step": 6635 }, { "epoch": 1.22, "learning_rate": 1.9228069441570976e-05, "loss": 0.1056, "step": 6636 }, { "epoch": 1.22, "learning_rate": 1.9227783385546364e-05, "loss": 0.0982, "step": 6637 }, { "epoch": 1.22, "learning_rate": 1.9227497278657997e-05, "loss": 0.1039, "step": 6638 }, { "epoch": 1.22, "learning_rate": 1.922721112090745e-05, "loss": 0.1044, "step": 6639 }, { "epoch": 1.22, "learning_rate": 1.9226924912296297e-05, "loss": 0.1028, "step": 6640 }, { "epoch": 1.22, "learning_rate": 1.9226638652826115e-05, "loss": 0.1031, "step": 6641 }, { "epoch": 1.22, "learning_rate": 1.922635234249849e-05, "loss": 0.1051, "step": 6642 }, { "epoch": 1.22, "learning_rate": 1.922606598131499e-05, "loss": 0.1019, "step": 6643 }, { "epoch": 1.22, "learning_rate": 1.9225779569277204e-05, "loss": 0.1112, "step": 6644 }, { "epoch": 1.22, "learning_rate": 1.92254931063867e-05, "loss": 0.1051, "step": 6645 }, { "epoch": 1.22, "learning_rate": 1.9225206592645063e-05, "loss": 0.0965, "step": 6646 }, { "epoch": 1.22, "learning_rate": 1.922492002805387e-05, "loss": 0.1003, "step": 6647 }, { "epoch": 1.22, "learning_rate": 1.9224633412614705e-05, "loss": 0.1133, "step": 6648 }, { "epoch": 1.22, "learning_rate": 1.922434674632914e-05, "loss": 0.0887, "step": 6649 }, { "epoch": 1.22, "learning_rate": 1.9224060029198762e-05, "loss": 0.0918, "step": 6650 }, { "epoch": 1.22, "learning_rate": 1.922377326122515e-05, "loss": 0.1032, "step": 6651 }, { "epoch": 1.22, "learning_rate": 1.922348644240988e-05, "loss": 0.1044, "step": 6652 }, { "epoch": 1.22, "learning_rate": 1.9223199572754544e-05, "loss": 0.0988, "step": 6653 }, { "epoch": 1.22, "learning_rate": 1.922291265226071e-05, "loss": 0.095, "step": 6654 }, { "epoch": 1.22, "learning_rate": 1.9222625680929965e-05, "loss": 0.0998, "step": 6655 }, { "epoch": 1.22, "learning_rate": 1.9222338658763894e-05, "loss": 0.0983, "step": 6656 }, { "epoch": 1.22, "learning_rate": 1.9222051585764076e-05, "loss": 0.1019, "step": 6657 }, { "epoch": 1.22, "learning_rate": 1.9221764461932092e-05, "loss": 0.0896, "step": 6658 }, { "epoch": 1.22, "learning_rate": 1.922147728726953e-05, "loss": 0.1108, "step": 6659 }, { "epoch": 1.22, "learning_rate": 1.9221190061777966e-05, "loss": 0.0954, "step": 6660 }, { "epoch": 1.22, "learning_rate": 1.9220902785458985e-05, "loss": 0.1024, "step": 6661 }, { "epoch": 1.22, "learning_rate": 1.9220615458314174e-05, "loss": 0.1047, "step": 6662 }, { "epoch": 1.22, "learning_rate": 1.9220328080345113e-05, "loss": 0.1107, "step": 6663 }, { "epoch": 1.22, "learning_rate": 1.9220040651553388e-05, "loss": 0.0934, "step": 6664 }, { "epoch": 1.22, "learning_rate": 1.9219753171940583e-05, "loss": 0.0833, "step": 6665 }, { "epoch": 1.22, "learning_rate": 1.9219465641508282e-05, "loss": 0.0988, "step": 6666 }, { "epoch": 1.22, "learning_rate": 1.9219178060258073e-05, "loss": 0.1063, "step": 6667 }, { "epoch": 1.22, "learning_rate": 1.9218890428191536e-05, "loss": 0.0906, "step": 6668 }, { "epoch": 1.22, "learning_rate": 1.921860274531026e-05, "loss": 0.0837, "step": 6669 }, { "epoch": 1.22, "learning_rate": 1.921831501161583e-05, "loss": 0.1026, "step": 6670 }, { "epoch": 1.22, "learning_rate": 1.9218027227109833e-05, "loss": 0.1005, "step": 6671 }, { "epoch": 1.22, "learning_rate": 1.921773939179385e-05, "loss": 0.0863, "step": 6672 }, { "epoch": 1.22, "learning_rate": 1.9217451505669477e-05, "loss": 0.1053, "step": 6673 }, { "epoch": 1.22, "learning_rate": 1.9217163568738292e-05, "loss": 0.095, "step": 6674 }, { "epoch": 1.22, "learning_rate": 1.9216875581001885e-05, "loss": 0.0961, "step": 6675 }, { "epoch": 1.22, "learning_rate": 1.9216587542461844e-05, "loss": 0.105, "step": 6676 }, { "epoch": 1.22, "learning_rate": 1.921629945311976e-05, "loss": 0.0933, "step": 6677 }, { "epoch": 1.22, "learning_rate": 1.921601131297721e-05, "loss": 0.0953, "step": 6678 }, { "epoch": 1.22, "learning_rate": 1.9215723122035793e-05, "loss": 0.0957, "step": 6679 }, { "epoch": 1.23, "learning_rate": 1.92154348802971e-05, "loss": 0.1026, "step": 6680 }, { "epoch": 1.23, "learning_rate": 1.921514658776271e-05, "loss": 0.1006, "step": 6681 }, { "epoch": 1.23, "learning_rate": 1.9214858244434213e-05, "loss": 0.0942, "step": 6682 }, { "epoch": 1.23, "learning_rate": 1.9214569850313203e-05, "loss": 0.0906, "step": 6683 }, { "epoch": 1.23, "learning_rate": 1.921428140540127e-05, "loss": 0.0949, "step": 6684 }, { "epoch": 1.23, "learning_rate": 1.9213992909700003e-05, "loss": 0.0977, "step": 6685 }, { "epoch": 1.23, "learning_rate": 1.921370436321099e-05, "loss": 0.0925, "step": 6686 }, { "epoch": 1.23, "learning_rate": 1.921341576593582e-05, "loss": 0.1127, "step": 6687 }, { "epoch": 1.23, "learning_rate": 1.9213127117876088e-05, "loss": 0.0979, "step": 6688 }, { "epoch": 1.23, "learning_rate": 1.9212838419033386e-05, "loss": 0.1104, "step": 6689 }, { "epoch": 1.23, "learning_rate": 1.9212549669409298e-05, "loss": 0.1003, "step": 6690 }, { "epoch": 1.23, "learning_rate": 1.921226086900543e-05, "loss": 0.0962, "step": 6691 }, { "epoch": 1.23, "learning_rate": 1.9211972017823354e-05, "loss": 0.1122, "step": 6692 }, { "epoch": 1.23, "learning_rate": 1.9211683115864676e-05, "loss": 0.1011, "step": 6693 }, { "epoch": 1.23, "learning_rate": 1.9211394163130984e-05, "loss": 0.0978, "step": 6694 }, { "epoch": 1.23, "learning_rate": 1.9211105159623874e-05, "loss": 0.1101, "step": 6695 }, { "epoch": 1.23, "learning_rate": 1.9210816105344935e-05, "loss": 0.1004, "step": 6696 }, { "epoch": 1.23, "learning_rate": 1.9210527000295763e-05, "loss": 0.0929, "step": 6697 }, { "epoch": 1.23, "learning_rate": 1.921023784447795e-05, "loss": 0.0955, "step": 6698 }, { "epoch": 1.23, "learning_rate": 1.9209948637893088e-05, "loss": 0.0931, "step": 6699 }, { "epoch": 1.23, "learning_rate": 1.9209659380542773e-05, "loss": 0.1137, "step": 6700 }, { "epoch": 1.23, "learning_rate": 1.9209370072428602e-05, "loss": 0.0958, "step": 6701 }, { "epoch": 1.23, "learning_rate": 1.920908071355217e-05, "loss": 0.0992, "step": 6702 }, { "epoch": 1.23, "learning_rate": 1.9208791303915063e-05, "loss": 0.1034, "step": 6703 }, { "epoch": 1.23, "learning_rate": 1.920850184351889e-05, "loss": 0.1022, "step": 6704 }, { "epoch": 1.23, "learning_rate": 1.9208212332365232e-05, "loss": 0.0963, "step": 6705 }, { "epoch": 1.23, "learning_rate": 1.9207922770455697e-05, "loss": 0.0983, "step": 6706 }, { "epoch": 1.23, "learning_rate": 1.9207633157791873e-05, "loss": 0.0958, "step": 6707 }, { "epoch": 1.23, "learning_rate": 1.920734349437536e-05, "loss": 0.098, "step": 6708 }, { "epoch": 1.23, "learning_rate": 1.9207053780207753e-05, "loss": 0.1117, "step": 6709 }, { "epoch": 1.23, "learning_rate": 1.920676401529065e-05, "loss": 0.1008, "step": 6710 }, { "epoch": 1.23, "learning_rate": 1.920647419962565e-05, "loss": 0.1042, "step": 6711 }, { "epoch": 1.23, "learning_rate": 1.920618433321435e-05, "loss": 0.1004, "step": 6712 }, { "epoch": 1.23, "learning_rate": 1.9205894416058344e-05, "loss": 0.1007, "step": 6713 }, { "epoch": 1.23, "learning_rate": 1.920560444815923e-05, "loss": 0.0974, "step": 6714 }, { "epoch": 1.23, "learning_rate": 1.9205314429518612e-05, "loss": 0.094, "step": 6715 }, { "epoch": 1.23, "learning_rate": 1.9205024360138083e-05, "loss": 0.1014, "step": 6716 }, { "epoch": 1.23, "learning_rate": 1.9204734240019246e-05, "loss": 0.0968, "step": 6717 }, { "epoch": 1.23, "learning_rate": 1.92044440691637e-05, "loss": 0.0959, "step": 6718 }, { "epoch": 1.23, "learning_rate": 1.920415384757304e-05, "loss": 0.0988, "step": 6719 }, { "epoch": 1.23, "learning_rate": 1.9203863575248863e-05, "loss": 0.1022, "step": 6720 }, { "epoch": 1.23, "learning_rate": 1.9203573252192784e-05, "loss": 0.1023, "step": 6721 }, { "epoch": 1.23, "learning_rate": 1.9203282878406387e-05, "loss": 0.1004, "step": 6722 }, { "epoch": 1.23, "learning_rate": 1.9202992453891282e-05, "loss": 0.1007, "step": 6723 }, { "epoch": 1.23, "learning_rate": 1.9202701978649068e-05, "loss": 0.0951, "step": 6724 }, { "epoch": 1.23, "learning_rate": 1.9202411452681344e-05, "loss": 0.095, "step": 6725 }, { "epoch": 1.23, "learning_rate": 1.9202120875989712e-05, "loss": 0.105, "step": 6726 }, { "epoch": 1.23, "learning_rate": 1.9201830248575772e-05, "loss": 0.0958, "step": 6727 }, { "epoch": 1.23, "learning_rate": 1.920153957044113e-05, "loss": 0.1023, "step": 6728 }, { "epoch": 1.23, "learning_rate": 1.920124884158739e-05, "loss": 0.0983, "step": 6729 }, { "epoch": 1.23, "learning_rate": 1.9200958062016144e-05, "loss": 0.088, "step": 6730 }, { "epoch": 1.23, "learning_rate": 1.9200667231729007e-05, "loss": 0.1028, "step": 6731 }, { "epoch": 1.23, "learning_rate": 1.9200376350727576e-05, "loss": 0.1002, "step": 6732 }, { "epoch": 1.23, "learning_rate": 1.920008541901345e-05, "loss": 0.0954, "step": 6733 }, { "epoch": 1.23, "learning_rate": 1.9199794436588244e-05, "loss": 0.1053, "step": 6734 }, { "epoch": 1.24, "learning_rate": 1.9199503403453553e-05, "loss": 0.1034, "step": 6735 }, { "epoch": 1.24, "learning_rate": 1.9199212319610986e-05, "loss": 0.0934, "step": 6736 }, { "epoch": 1.24, "learning_rate": 1.919892118506214e-05, "loss": 0.095, "step": 6737 }, { "epoch": 1.24, "learning_rate": 1.919862999980863e-05, "loss": 0.1102, "step": 6738 }, { "epoch": 1.24, "learning_rate": 1.919833876385205e-05, "loss": 0.0996, "step": 6739 }, { "epoch": 1.24, "learning_rate": 1.919804747719402e-05, "loss": 0.0857, "step": 6740 }, { "epoch": 1.24, "learning_rate": 1.9197756139836128e-05, "loss": 0.0988, "step": 6741 }, { "epoch": 1.24, "learning_rate": 1.919746475177999e-05, "loss": 0.0912, "step": 6742 }, { "epoch": 1.24, "learning_rate": 1.9197173313027215e-05, "loss": 0.1085, "step": 6743 }, { "epoch": 1.24, "learning_rate": 1.91968818235794e-05, "loss": 0.1059, "step": 6744 }, { "epoch": 1.24, "learning_rate": 1.9196590283438167e-05, "loss": 0.1082, "step": 6745 }, { "epoch": 1.24, "learning_rate": 1.9196298692605104e-05, "loss": 0.1059, "step": 6746 }, { "epoch": 1.24, "learning_rate": 1.919600705108183e-05, "loss": 0.1099, "step": 6747 }, { "epoch": 1.24, "learning_rate": 1.9195715358869948e-05, "loss": 0.0911, "step": 6748 }, { "epoch": 1.24, "learning_rate": 1.919542361597107e-05, "loss": 0.115, "step": 6749 }, { "epoch": 1.24, "learning_rate": 1.9195131822386803e-05, "loss": 0.1031, "step": 6750 }, { "epoch": 1.24, "learning_rate": 1.9194839978118748e-05, "loss": 0.1035, "step": 6751 }, { "epoch": 1.24, "learning_rate": 1.9194548083168524e-05, "loss": 0.1111, "step": 6752 }, { "epoch": 1.24, "learning_rate": 1.9194256137537734e-05, "loss": 0.1047, "step": 6753 }, { "epoch": 1.24, "learning_rate": 1.9193964141227986e-05, "loss": 0.0976, "step": 6754 }, { "epoch": 1.24, "learning_rate": 1.9193672094240898e-05, "loss": 0.1021, "step": 6755 }, { "epoch": 1.24, "learning_rate": 1.9193379996578072e-05, "loss": 0.1096, "step": 6756 }, { "epoch": 1.24, "learning_rate": 1.919308784824112e-05, "loss": 0.1017, "step": 6757 }, { "epoch": 1.24, "learning_rate": 1.9192795649231652e-05, "loss": 0.1173, "step": 6758 }, { "epoch": 1.24, "learning_rate": 1.919250339955128e-05, "loss": 0.0944, "step": 6759 }, { "epoch": 1.24, "learning_rate": 1.919221109920161e-05, "loss": 0.0974, "step": 6760 }, { "epoch": 1.24, "learning_rate": 1.919191874818426e-05, "loss": 0.0957, "step": 6761 }, { "epoch": 1.24, "learning_rate": 1.919162634650084e-05, "loss": 0.0993, "step": 6762 }, { "epoch": 1.24, "learning_rate": 1.9191333894152957e-05, "loss": 0.0938, "step": 6763 }, { "epoch": 1.24, "learning_rate": 1.919104139114223e-05, "loss": 0.0973, "step": 6764 }, { "epoch": 1.24, "learning_rate": 1.9190748837470264e-05, "loss": 0.0978, "step": 6765 }, { "epoch": 1.24, "learning_rate": 1.9190456233138678e-05, "loss": 0.0988, "step": 6766 }, { "epoch": 1.24, "learning_rate": 1.919016357814908e-05, "loss": 0.0959, "step": 6767 }, { "epoch": 1.24, "learning_rate": 1.9189870872503084e-05, "loss": 0.096, "step": 6768 }, { "epoch": 1.24, "learning_rate": 1.918957811620231e-05, "loss": 0.0949, "step": 6769 }, { "epoch": 1.24, "learning_rate": 1.918928530924836e-05, "loss": 0.0992, "step": 6770 }, { "epoch": 1.24, "learning_rate": 1.9188992451642857e-05, "loss": 0.094, "step": 6771 }, { "epoch": 1.24, "learning_rate": 1.918869954338741e-05, "loss": 0.106, "step": 6772 }, { "epoch": 1.24, "learning_rate": 1.9188406584483637e-05, "loss": 0.0927, "step": 6773 }, { "epoch": 1.24, "learning_rate": 1.9188113574933153e-05, "loss": 0.1047, "step": 6774 }, { "epoch": 1.24, "learning_rate": 1.9187820514737573e-05, "loss": 0.1032, "step": 6775 }, { "epoch": 1.24, "learning_rate": 1.9187527403898506e-05, "loss": 0.0992, "step": 6776 }, { "epoch": 1.24, "learning_rate": 1.9187234242417578e-05, "loss": 0.0966, "step": 6777 }, { "epoch": 1.24, "learning_rate": 1.9186941030296396e-05, "loss": 0.1028, "step": 6778 }, { "epoch": 1.24, "learning_rate": 1.9186647767536583e-05, "loss": 0.0987, "step": 6779 }, { "epoch": 1.24, "learning_rate": 1.9186354454139748e-05, "loss": 0.0897, "step": 6780 }, { "epoch": 1.24, "learning_rate": 1.9186061090107517e-05, "loss": 0.0984, "step": 6781 }, { "epoch": 1.24, "learning_rate": 1.91857676754415e-05, "loss": 0.1094, "step": 6782 }, { "epoch": 1.24, "learning_rate": 1.9185474210143314e-05, "loss": 0.1095, "step": 6783 }, { "epoch": 1.24, "learning_rate": 1.918518069421458e-05, "loss": 0.1095, "step": 6784 }, { "epoch": 1.24, "learning_rate": 1.9184887127656916e-05, "loss": 0.1006, "step": 6785 }, { "epoch": 1.24, "learning_rate": 1.9184593510471938e-05, "loss": 0.0998, "step": 6786 }, { "epoch": 1.24, "learning_rate": 1.9184299842661267e-05, "loss": 0.0945, "step": 6787 }, { "epoch": 1.24, "learning_rate": 1.918400612422652e-05, "loss": 0.0931, "step": 6788 }, { "epoch": 1.25, "learning_rate": 1.9183712355169314e-05, "loss": 0.1072, "step": 6789 }, { "epoch": 1.25, "learning_rate": 1.918341853549127e-05, "loss": 0.1003, "step": 6790 }, { "epoch": 1.25, "learning_rate": 1.918312466519401e-05, "loss": 0.0878, "step": 6791 }, { "epoch": 1.25, "learning_rate": 1.9182830744279154e-05, "loss": 0.0996, "step": 6792 }, { "epoch": 1.25, "learning_rate": 1.9182536772748314e-05, "loss": 0.1112, "step": 6793 }, { "epoch": 1.25, "learning_rate": 1.918224275060312e-05, "loss": 0.0914, "step": 6794 }, { "epoch": 1.25, "learning_rate": 1.918194867784519e-05, "loss": 0.1122, "step": 6795 }, { "epoch": 1.25, "learning_rate": 1.918165455447614e-05, "loss": 0.0966, "step": 6796 }, { "epoch": 1.25, "learning_rate": 1.9181360380497594e-05, "loss": 0.0907, "step": 6797 }, { "epoch": 1.25, "learning_rate": 1.9181066155911178e-05, "loss": 0.0994, "step": 6798 }, { "epoch": 1.25, "learning_rate": 1.918077188071851e-05, "loss": 0.1145, "step": 6799 }, { "epoch": 1.25, "learning_rate": 1.9180477554921212e-05, "loss": 0.1068, "step": 6800 }, { "epoch": 1.25, "learning_rate": 1.9180183178520904e-05, "loss": 0.1009, "step": 6801 }, { "epoch": 1.25, "learning_rate": 1.9179888751519215e-05, "loss": 0.1165, "step": 6802 }, { "epoch": 1.25, "learning_rate": 1.9179594273917765e-05, "loss": 0.1069, "step": 6803 }, { "epoch": 1.25, "learning_rate": 1.9179299745718173e-05, "loss": 0.1037, "step": 6804 }, { "epoch": 1.25, "learning_rate": 1.9179005166922067e-05, "loss": 0.1043, "step": 6805 }, { "epoch": 1.25, "learning_rate": 1.917871053753107e-05, "loss": 0.0884, "step": 6806 }, { "epoch": 1.25, "learning_rate": 1.9178415857546805e-05, "loss": 0.0949, "step": 6807 }, { "epoch": 1.25, "learning_rate": 1.9178121126970895e-05, "loss": 0.1089, "step": 6808 }, { "epoch": 1.25, "learning_rate": 1.9177826345804967e-05, "loss": 0.1022, "step": 6809 }, { "epoch": 1.25, "learning_rate": 1.917753151405065e-05, "loss": 0.1125, "step": 6810 }, { "epoch": 1.25, "learning_rate": 1.9177236631709557e-05, "loss": 0.1075, "step": 6811 }, { "epoch": 1.25, "learning_rate": 1.9176941698783325e-05, "loss": 0.1043, "step": 6812 }, { "epoch": 1.25, "learning_rate": 1.9176646715273575e-05, "loss": 0.1055, "step": 6813 }, { "epoch": 1.25, "learning_rate": 1.9176351681181932e-05, "loss": 0.0971, "step": 6814 }, { "epoch": 1.25, "learning_rate": 1.9176056596510025e-05, "loss": 0.0981, "step": 6815 }, { "epoch": 1.25, "learning_rate": 1.917576146125948e-05, "loss": 0.0962, "step": 6816 }, { "epoch": 1.25, "learning_rate": 1.917546627543192e-05, "loss": 0.105, "step": 6817 }, { "epoch": 1.25, "learning_rate": 1.9175171039028974e-05, "loss": 0.0935, "step": 6818 }, { "epoch": 1.25, "learning_rate": 1.9174875752052276e-05, "loss": 0.1099, "step": 6819 }, { "epoch": 1.25, "learning_rate": 1.917458041450344e-05, "loss": 0.0988, "step": 6820 }, { "epoch": 1.25, "learning_rate": 1.9174285026384107e-05, "loss": 0.1081, "step": 6821 }, { "epoch": 1.25, "learning_rate": 1.91739895876959e-05, "loss": 0.1008, "step": 6822 }, { "epoch": 1.25, "learning_rate": 1.9173694098440447e-05, "loss": 0.0952, "step": 6823 }, { "epoch": 1.25, "learning_rate": 1.9173398558619377e-05, "loss": 0.0925, "step": 6824 }, { "epoch": 1.25, "learning_rate": 1.917310296823432e-05, "loss": 0.1029, "step": 6825 }, { "epoch": 1.25, "learning_rate": 1.9172807327286903e-05, "loss": 0.0939, "step": 6826 }, { "epoch": 1.25, "learning_rate": 1.9172511635778756e-05, "loss": 0.1058, "step": 6827 }, { "epoch": 1.25, "learning_rate": 1.9172215893711514e-05, "loss": 0.1101, "step": 6828 }, { "epoch": 1.25, "learning_rate": 1.91719201010868e-05, "loss": 0.1063, "step": 6829 }, { "epoch": 1.25, "learning_rate": 1.917162425790625e-05, "loss": 0.0916, "step": 6830 }, { "epoch": 1.25, "learning_rate": 1.917132836417149e-05, "loss": 0.1156, "step": 6831 }, { "epoch": 1.25, "learning_rate": 1.9171032419884153e-05, "loss": 0.1109, "step": 6832 }, { "epoch": 1.25, "learning_rate": 1.917073642504587e-05, "loss": 0.0989, "step": 6833 }, { "epoch": 1.25, "learning_rate": 1.9170440379658275e-05, "loss": 0.0863, "step": 6834 }, { "epoch": 1.25, "learning_rate": 1.9170144283722995e-05, "loss": 0.1101, "step": 6835 }, { "epoch": 1.25, "learning_rate": 1.9169848137241667e-05, "loss": 0.0906, "step": 6836 }, { "epoch": 1.25, "learning_rate": 1.9169551940215922e-05, "loss": 0.0982, "step": 6837 }, { "epoch": 1.25, "learning_rate": 1.916925569264739e-05, "loss": 0.0944, "step": 6838 }, { "epoch": 1.25, "learning_rate": 1.9168959394537708e-05, "loss": 0.0976, "step": 6839 }, { "epoch": 1.25, "learning_rate": 1.916866304588851e-05, "loss": 0.1148, "step": 6840 }, { "epoch": 1.25, "learning_rate": 1.9168366646701417e-05, "loss": 0.1082, "step": 6841 }, { "epoch": 1.25, "learning_rate": 1.916807019697808e-05, "loss": 0.1003, "step": 6842 }, { "epoch": 1.25, "learning_rate": 1.9167773696720123e-05, "loss": 0.1009, "step": 6843 }, { "epoch": 1.26, "learning_rate": 1.916747714592918e-05, "loss": 0.0909, "step": 6844 }, { "epoch": 1.26, "learning_rate": 1.9167180544606895e-05, "loss": 0.1021, "step": 6845 }, { "epoch": 1.26, "learning_rate": 1.916688389275489e-05, "loss": 0.093, "step": 6846 }, { "epoch": 1.26, "learning_rate": 1.9166587190374807e-05, "loss": 0.0951, "step": 6847 }, { "epoch": 1.26, "learning_rate": 1.916629043746828e-05, "loss": 0.102, "step": 6848 }, { "epoch": 1.26, "learning_rate": 1.9165993634036948e-05, "loss": 0.0961, "step": 6849 }, { "epoch": 1.26, "learning_rate": 1.9165696780082444e-05, "loss": 0.0988, "step": 6850 }, { "epoch": 1.26, "learning_rate": 1.9165399875606403e-05, "loss": 0.1075, "step": 6851 }, { "epoch": 1.26, "learning_rate": 1.9165102920610464e-05, "loss": 0.1015, "step": 6852 }, { "epoch": 1.26, "learning_rate": 1.9164805915096263e-05, "loss": 0.1052, "step": 6853 }, { "epoch": 1.26, "learning_rate": 1.9164508859065438e-05, "loss": 0.1072, "step": 6854 }, { "epoch": 1.26, "learning_rate": 1.9164211752519627e-05, "loss": 0.0964, "step": 6855 }, { "epoch": 1.26, "learning_rate": 1.916391459546046e-05, "loss": 0.1054, "step": 6856 }, { "epoch": 1.26, "learning_rate": 1.9163617387889587e-05, "loss": 0.095, "step": 6857 }, { "epoch": 1.26, "learning_rate": 1.9163320129808637e-05, "loss": 0.1078, "step": 6858 }, { "epoch": 1.26, "learning_rate": 1.9163022821219252e-05, "loss": 0.0971, "step": 6859 }, { "epoch": 1.26, "learning_rate": 1.9162725462123074e-05, "loss": 0.0877, "step": 6860 }, { "epoch": 1.26, "learning_rate": 1.9162428052521736e-05, "loss": 0.0941, "step": 6861 }, { "epoch": 1.26, "learning_rate": 1.916213059241688e-05, "loss": 0.0914, "step": 6862 }, { "epoch": 1.26, "learning_rate": 1.9161833081810142e-05, "loss": 0.0868, "step": 6863 }, { "epoch": 1.26, "learning_rate": 1.9161535520703172e-05, "loss": 0.1044, "step": 6864 }, { "epoch": 1.26, "learning_rate": 1.91612379090976e-05, "loss": 0.1006, "step": 6865 }, { "epoch": 1.26, "learning_rate": 1.916094024699507e-05, "loss": 0.1079, "step": 6866 }, { "epoch": 1.26, "learning_rate": 1.9160642534397222e-05, "loss": 0.1085, "step": 6867 }, { "epoch": 1.26, "learning_rate": 1.9160344771305702e-05, "loss": 0.1021, "step": 6868 }, { "epoch": 1.26, "learning_rate": 1.9160046957722144e-05, "loss": 0.0957, "step": 6869 }, { "epoch": 1.26, "learning_rate": 1.915974909364819e-05, "loss": 0.0996, "step": 6870 }, { "epoch": 1.26, "learning_rate": 1.9159451179085487e-05, "loss": 0.0911, "step": 6871 }, { "epoch": 1.26, "learning_rate": 1.9159153214035674e-05, "loss": 0.0938, "step": 6872 }, { "epoch": 1.26, "learning_rate": 1.9158855198500393e-05, "loss": 0.1114, "step": 6873 }, { "epoch": 1.26, "learning_rate": 1.915855713248129e-05, "loss": 0.095, "step": 6874 }, { "epoch": 1.26, "learning_rate": 1.9158259015980005e-05, "loss": 0.1041, "step": 6875 }, { "epoch": 1.26, "learning_rate": 1.915796084899818e-05, "loss": 0.1006, "step": 6876 }, { "epoch": 1.26, "learning_rate": 1.9157662631537463e-05, "loss": 0.1048, "step": 6877 }, { "epoch": 1.26, "learning_rate": 1.9157364363599493e-05, "loss": 0.102, "step": 6878 }, { "epoch": 1.26, "learning_rate": 1.9157066045185918e-05, "loss": 0.104, "step": 6879 }, { "epoch": 1.26, "learning_rate": 1.915676767629838e-05, "loss": 0.0874, "step": 6880 }, { "epoch": 1.26, "learning_rate": 1.915646925693852e-05, "loss": 0.0908, "step": 6881 }, { "epoch": 1.26, "learning_rate": 1.915617078710799e-05, "loss": 0.0923, "step": 6882 }, { "epoch": 1.26, "learning_rate": 1.9155872266808433e-05, "loss": 0.1032, "step": 6883 }, { "epoch": 1.26, "learning_rate": 1.9155573696041494e-05, "loss": 0.1018, "step": 6884 }, { "epoch": 1.26, "learning_rate": 1.9155275074808816e-05, "loss": 0.0991, "step": 6885 }, { "epoch": 1.26, "learning_rate": 1.915497640311205e-05, "loss": 0.1077, "step": 6886 }, { "epoch": 1.26, "learning_rate": 1.9154677680952834e-05, "loss": 0.0907, "step": 6887 }, { "epoch": 1.26, "learning_rate": 1.9154378908332827e-05, "loss": 0.1136, "step": 6888 }, { "epoch": 1.26, "learning_rate": 1.9154080085253665e-05, "loss": 0.086, "step": 6889 }, { "epoch": 1.26, "learning_rate": 1.9153781211717e-05, "loss": 0.0837, "step": 6890 }, { "epoch": 1.26, "learning_rate": 1.9153482287724478e-05, "loss": 0.1013, "step": 6891 }, { "epoch": 1.26, "learning_rate": 1.915318331327775e-05, "loss": 0.1031, "step": 6892 }, { "epoch": 1.26, "learning_rate": 1.9152884288378456e-05, "loss": 0.1031, "step": 6893 }, { "epoch": 1.26, "learning_rate": 1.9152585213028253e-05, "loss": 0.0881, "step": 6894 }, { "epoch": 1.26, "learning_rate": 1.9152286087228786e-05, "loss": 0.0999, "step": 6895 }, { "epoch": 1.26, "learning_rate": 1.9151986910981703e-05, "loss": 0.1115, "step": 6896 }, { "epoch": 1.26, "learning_rate": 1.9151687684288654e-05, "loss": 0.1132, "step": 6897 }, { "epoch": 1.26, "learning_rate": 1.9151388407151287e-05, "loss": 0.1037, "step": 6898 }, { "epoch": 1.27, "learning_rate": 1.915108907957125e-05, "loss": 0.0965, "step": 6899 }, { "epoch": 1.27, "learning_rate": 1.9150789701550203e-05, "loss": 0.0954, "step": 6900 }, { "epoch": 1.27, "learning_rate": 1.9150490273089787e-05, "loss": 0.0908, "step": 6901 }, { "epoch": 1.27, "learning_rate": 1.9150190794191648e-05, "loss": 0.0993, "step": 6902 }, { "epoch": 1.27, "learning_rate": 1.914989126485745e-05, "loss": 0.112, "step": 6903 }, { "epoch": 1.27, "learning_rate": 1.9149591685088832e-05, "loss": 0.0995, "step": 6904 }, { "epoch": 1.27, "learning_rate": 1.9149292054887455e-05, "loss": 0.0995, "step": 6905 }, { "epoch": 1.27, "learning_rate": 1.914899237425496e-05, "loss": 0.0975, "step": 6906 }, { "epoch": 1.27, "learning_rate": 1.9148692643193008e-05, "loss": 0.1013, "step": 6907 }, { "epoch": 1.27, "learning_rate": 1.914839286170325e-05, "loss": 0.0978, "step": 6908 }, { "epoch": 1.27, "learning_rate": 1.9148093029787335e-05, "loss": 0.0903, "step": 6909 }, { "epoch": 1.27, "learning_rate": 1.9147793147446912e-05, "loss": 0.0952, "step": 6910 }, { "epoch": 1.27, "learning_rate": 1.9147493214683645e-05, "loss": 0.1023, "step": 6911 }, { "epoch": 1.27, "learning_rate": 1.914719323149918e-05, "loss": 0.0913, "step": 6912 }, { "epoch": 1.27, "learning_rate": 1.914689319789517e-05, "loss": 0.1011, "step": 6913 }, { "epoch": 1.27, "learning_rate": 1.9146593113873268e-05, "loss": 0.1009, "step": 6914 }, { "epoch": 1.27, "learning_rate": 1.9146292979435135e-05, "loss": 0.0907, "step": 6915 }, { "epoch": 1.27, "learning_rate": 1.9145992794582418e-05, "loss": 0.0912, "step": 6916 }, { "epoch": 1.27, "learning_rate": 1.9145692559316777e-05, "loss": 0.099, "step": 6917 }, { "epoch": 1.27, "learning_rate": 1.914539227363986e-05, "loss": 0.1153, "step": 6918 }, { "epoch": 1.27, "learning_rate": 1.914509193755333e-05, "loss": 0.1031, "step": 6919 }, { "epoch": 1.27, "learning_rate": 1.9144791551058838e-05, "loss": 0.0929, "step": 6920 }, { "epoch": 1.27, "learning_rate": 1.914449111415804e-05, "loss": 0.0964, "step": 6921 }, { "epoch": 1.27, "learning_rate": 1.9144190626852596e-05, "loss": 0.0957, "step": 6922 }, { "epoch": 1.27, "learning_rate": 1.9143890089144157e-05, "loss": 0.0952, "step": 6923 }, { "epoch": 1.27, "learning_rate": 1.9143589501034382e-05, "loss": 0.1032, "step": 6924 }, { "epoch": 1.27, "learning_rate": 1.9143288862524926e-05, "loss": 0.1079, "step": 6925 }, { "epoch": 1.27, "learning_rate": 1.9142988173617448e-05, "loss": 0.1029, "step": 6926 }, { "epoch": 1.27, "learning_rate": 1.9142687434313606e-05, "loss": 0.0964, "step": 6927 }, { "epoch": 1.27, "learning_rate": 1.9142386644615058e-05, "loss": 0.1081, "step": 6928 }, { "epoch": 1.27, "learning_rate": 1.9142085804523457e-05, "loss": 0.0885, "step": 6929 }, { "epoch": 1.27, "learning_rate": 1.9141784914040467e-05, "loss": 0.1071, "step": 6930 }, { "epoch": 1.27, "learning_rate": 1.9141483973167744e-05, "loss": 0.1083, "step": 6931 }, { "epoch": 1.27, "learning_rate": 1.914118298190695e-05, "loss": 0.093, "step": 6932 }, { "epoch": 1.27, "learning_rate": 1.9140881940259738e-05, "loss": 0.0952, "step": 6933 }, { "epoch": 1.27, "learning_rate": 1.914058084822777e-05, "loss": 0.0917, "step": 6934 }, { "epoch": 1.27, "learning_rate": 1.9140279705812707e-05, "loss": 0.0966, "step": 6935 }, { "epoch": 1.27, "learning_rate": 1.9139978513016207e-05, "loss": 0.0927, "step": 6936 }, { "epoch": 1.27, "learning_rate": 1.9139677269839936e-05, "loss": 0.1064, "step": 6937 }, { "epoch": 1.27, "learning_rate": 1.9139375976285547e-05, "loss": 0.0948, "step": 6938 }, { "epoch": 1.27, "learning_rate": 1.9139074632354703e-05, "loss": 0.1036, "step": 6939 }, { "epoch": 1.27, "learning_rate": 1.9138773238049065e-05, "loss": 0.1, "step": 6940 }, { "epoch": 1.27, "learning_rate": 1.9138471793370293e-05, "loss": 0.106, "step": 6941 }, { "epoch": 1.27, "learning_rate": 1.9138170298320054e-05, "loss": 0.1046, "step": 6942 }, { "epoch": 1.27, "learning_rate": 1.9137868752900005e-05, "loss": 0.1068, "step": 6943 }, { "epoch": 1.27, "learning_rate": 1.913756715711181e-05, "loss": 0.0917, "step": 6944 }, { "epoch": 1.27, "learning_rate": 1.913726551095713e-05, "loss": 0.1053, "step": 6945 }, { "epoch": 1.27, "learning_rate": 1.9136963814437627e-05, "loss": 0.0904, "step": 6946 }, { "epoch": 1.27, "learning_rate": 1.913666206755497e-05, "loss": 0.0941, "step": 6947 }, { "epoch": 1.27, "learning_rate": 1.9136360270310815e-05, "loss": 0.0883, "step": 6948 }, { "epoch": 1.27, "learning_rate": 1.913605842270683e-05, "loss": 0.1115, "step": 6949 }, { "epoch": 1.27, "learning_rate": 1.913575652474467e-05, "loss": 0.1077, "step": 6950 }, { "epoch": 1.27, "learning_rate": 1.913545457642601e-05, "loss": 0.0866, "step": 6951 }, { "epoch": 1.27, "learning_rate": 1.913515257775251e-05, "loss": 0.1023, "step": 6952 }, { "epoch": 1.28, "learning_rate": 1.9134850528725838e-05, "loss": 0.0996, "step": 6953 }, { "epoch": 1.28, "learning_rate": 1.9134548429347654e-05, "loss": 0.1041, "step": 6954 }, { "epoch": 1.28, "learning_rate": 1.9134246279619626e-05, "loss": 0.0973, "step": 6955 }, { "epoch": 1.28, "learning_rate": 1.9133944079543418e-05, "loss": 0.0935, "step": 6956 }, { "epoch": 1.28, "learning_rate": 1.9133641829120692e-05, "loss": 0.1066, "step": 6957 }, { "epoch": 1.28, "learning_rate": 1.9133339528353123e-05, "loss": 0.1107, "step": 6958 }, { "epoch": 1.28, "learning_rate": 1.9133037177242372e-05, "loss": 0.1022, "step": 6959 }, { "epoch": 1.28, "learning_rate": 1.9132734775790104e-05, "loss": 0.1049, "step": 6960 }, { "epoch": 1.28, "learning_rate": 1.913243232399799e-05, "loss": 0.1023, "step": 6961 }, { "epoch": 1.28, "learning_rate": 1.9132129821867695e-05, "loss": 0.0971, "step": 6962 }, { "epoch": 1.28, "learning_rate": 1.9131827269400888e-05, "loss": 0.1028, "step": 6963 }, { "epoch": 1.28, "learning_rate": 1.9131524666599233e-05, "loss": 0.0961, "step": 6964 }, { "epoch": 1.28, "learning_rate": 1.9131222013464403e-05, "loss": 0.097, "step": 6965 }, { "epoch": 1.28, "learning_rate": 1.913091930999806e-05, "loss": 0.1003, "step": 6966 }, { "epoch": 1.28, "learning_rate": 1.9130616556201874e-05, "loss": 0.1094, "step": 6967 }, { "epoch": 1.28, "learning_rate": 1.913031375207752e-05, "loss": 0.1048, "step": 6968 }, { "epoch": 1.28, "learning_rate": 1.9130010897626658e-05, "loss": 0.0964, "step": 6969 }, { "epoch": 1.28, "learning_rate": 1.9129707992850967e-05, "loss": 0.0939, "step": 6970 }, { "epoch": 1.28, "learning_rate": 1.912940503775211e-05, "loss": 0.0943, "step": 6971 }, { "epoch": 1.28, "learning_rate": 1.9129102032331758e-05, "loss": 0.1004, "step": 6972 }, { "epoch": 1.28, "learning_rate": 1.912879897659158e-05, "loss": 0.1062, "step": 6973 }, { "epoch": 1.28, "learning_rate": 1.912849587053325e-05, "loss": 0.113, "step": 6974 }, { "epoch": 1.28, "learning_rate": 1.912819271415844e-05, "loss": 0.1094, "step": 6975 }, { "epoch": 1.28, "learning_rate": 1.9127889507468813e-05, "loss": 0.101, "step": 6976 }, { "epoch": 1.28, "learning_rate": 1.9127586250466045e-05, "loss": 0.1099, "step": 6977 }, { "epoch": 1.28, "learning_rate": 1.9127282943151813e-05, "loss": 0.1077, "step": 6978 }, { "epoch": 1.28, "learning_rate": 1.912697958552778e-05, "loss": 0.1078, "step": 6979 }, { "epoch": 1.28, "learning_rate": 1.912667617759562e-05, "loss": 0.1021, "step": 6980 }, { "epoch": 1.28, "learning_rate": 1.9126372719357012e-05, "loss": 0.1086, "step": 6981 }, { "epoch": 1.28, "learning_rate": 1.912606921081362e-05, "loss": 0.1064, "step": 6982 }, { "epoch": 1.28, "learning_rate": 1.9125765651967124e-05, "loss": 0.1243, "step": 6983 }, { "epoch": 1.28, "learning_rate": 1.912546204281919e-05, "loss": 0.0998, "step": 6984 }, { "epoch": 1.28, "learning_rate": 1.9125158383371496e-05, "loss": 0.1021, "step": 6985 }, { "epoch": 1.28, "learning_rate": 1.9124854673625717e-05, "loss": 0.1091, "step": 6986 }, { "epoch": 1.28, "learning_rate": 1.9124550913583526e-05, "loss": 0.1122, "step": 6987 }, { "epoch": 1.28, "learning_rate": 1.9124247103246595e-05, "loss": 0.1008, "step": 6988 }, { "epoch": 1.28, "learning_rate": 1.91239432426166e-05, "loss": 0.103, "step": 6989 }, { "epoch": 1.28, "learning_rate": 1.9123639331695213e-05, "loss": 0.0967, "step": 6990 }, { "epoch": 1.28, "learning_rate": 1.912333537048412e-05, "loss": 0.0935, "step": 6991 }, { "epoch": 1.28, "learning_rate": 1.9123031358984982e-05, "loss": 0.0935, "step": 6992 }, { "epoch": 1.28, "learning_rate": 1.9122727297199483e-05, "loss": 0.099, "step": 6993 }, { "epoch": 1.28, "learning_rate": 1.91224231851293e-05, "loss": 0.0956, "step": 6994 }, { "epoch": 1.28, "learning_rate": 1.9122119022776106e-05, "loss": 0.1023, "step": 6995 }, { "epoch": 1.28, "learning_rate": 1.9121814810141575e-05, "loss": 0.0988, "step": 6996 }, { "epoch": 1.28, "learning_rate": 1.912151054722739e-05, "loss": 0.1067, "step": 6997 }, { "epoch": 1.28, "learning_rate": 1.9121206234035226e-05, "loss": 0.0911, "step": 6998 }, { "epoch": 1.28, "learning_rate": 1.9120901870566756e-05, "loss": 0.0942, "step": 6999 }, { "epoch": 1.28, "learning_rate": 1.9120597456823665e-05, "loss": 0.1032, "step": 7000 }, { "epoch": 1.28, "learning_rate": 1.9120292992807624e-05, "loss": 0.1013, "step": 7001 }, { "epoch": 1.28, "learning_rate": 1.9119988478520314e-05, "loss": 0.0951, "step": 7002 }, { "epoch": 1.28, "learning_rate": 1.9119683913963417e-05, "loss": 0.0955, "step": 7003 }, { "epoch": 1.28, "learning_rate": 1.9119379299138607e-05, "loss": 0.0937, "step": 7004 }, { "epoch": 1.28, "learning_rate": 1.9119074634047567e-05, "loss": 0.0988, "step": 7005 }, { "epoch": 1.28, "learning_rate": 1.911876991869197e-05, "loss": 0.1026, "step": 7006 }, { "epoch": 1.28, "learning_rate": 1.91184651530735e-05, "loss": 0.0971, "step": 7007 }, { "epoch": 1.29, "learning_rate": 1.9118160337193838e-05, "loss": 0.1005, "step": 7008 }, { "epoch": 1.29, "learning_rate": 1.9117855471054662e-05, "loss": 0.0934, "step": 7009 }, { "epoch": 1.29, "learning_rate": 1.9117550554657653e-05, "loss": 0.0985, "step": 7010 }, { "epoch": 1.29, "learning_rate": 1.9117245588004492e-05, "loss": 0.0991, "step": 7011 }, { "epoch": 1.29, "learning_rate": 1.9116940571096863e-05, "loss": 0.0919, "step": 7012 }, { "epoch": 1.29, "learning_rate": 1.911663550393644e-05, "loss": 0.0982, "step": 7013 }, { "epoch": 1.29, "learning_rate": 1.911633038652491e-05, "loss": 0.098, "step": 7014 }, { "epoch": 1.29, "learning_rate": 1.9116025218863952e-05, "loss": 0.1031, "step": 7015 }, { "epoch": 1.29, "learning_rate": 1.9115720000955252e-05, "loss": 0.1006, "step": 7016 }, { "epoch": 1.29, "learning_rate": 1.9115414732800486e-05, "loss": 0.0912, "step": 7017 }, { "epoch": 1.29, "learning_rate": 1.9115109414401343e-05, "loss": 0.1058, "step": 7018 }, { "epoch": 1.29, "learning_rate": 1.9114804045759504e-05, "loss": 0.0805, "step": 7019 }, { "epoch": 1.29, "learning_rate": 1.911449862687665e-05, "loss": 0.0896, "step": 7020 }, { "epoch": 1.29, "learning_rate": 1.9114193157754464e-05, "loss": 0.1063, "step": 7021 }, { "epoch": 1.29, "learning_rate": 1.9113887638394636e-05, "loss": 0.1084, "step": 7022 }, { "epoch": 1.29, "learning_rate": 1.9113582068798845e-05, "loss": 0.0906, "step": 7023 }, { "epoch": 1.29, "learning_rate": 1.9113276448968773e-05, "loss": 0.1016, "step": 7024 }, { "epoch": 1.29, "learning_rate": 1.911297077890611e-05, "loss": 0.0929, "step": 7025 }, { "epoch": 1.29, "learning_rate": 1.9112665058612537e-05, "loss": 0.1021, "step": 7026 }, { "epoch": 1.29, "learning_rate": 1.9112359288089744e-05, "loss": 0.1021, "step": 7027 }, { "epoch": 1.29, "learning_rate": 1.911205346733941e-05, "loss": 0.1113, "step": 7028 }, { "epoch": 1.29, "learning_rate": 1.9111747596363227e-05, "loss": 0.1008, "step": 7029 }, { "epoch": 1.29, "learning_rate": 1.9111441675162874e-05, "loss": 0.1086, "step": 7030 }, { "epoch": 1.29, "learning_rate": 1.9111135703740045e-05, "loss": 0.1003, "step": 7031 }, { "epoch": 1.29, "learning_rate": 1.9110829682096418e-05, "loss": 0.0982, "step": 7032 }, { "epoch": 1.29, "learning_rate": 1.9110523610233688e-05, "loss": 0.1043, "step": 7033 }, { "epoch": 1.29, "learning_rate": 1.9110217488153536e-05, "loss": 0.11, "step": 7034 }, { "epoch": 1.29, "learning_rate": 1.9109911315857656e-05, "loss": 0.1065, "step": 7035 }, { "epoch": 1.29, "learning_rate": 1.9109605093347725e-05, "loss": 0.0993, "step": 7036 }, { "epoch": 1.29, "learning_rate": 1.9109298820625443e-05, "loss": 0.0965, "step": 7037 }, { "epoch": 1.29, "learning_rate": 1.9108992497692488e-05, "loss": 0.0997, "step": 7038 }, { "epoch": 1.29, "learning_rate": 1.9108686124550554e-05, "loss": 0.1007, "step": 7039 }, { "epoch": 1.29, "learning_rate": 1.9108379701201332e-05, "loss": 0.1073, "step": 7040 }, { "epoch": 1.29, "learning_rate": 1.9108073227646505e-05, "loss": 0.1074, "step": 7041 }, { "epoch": 1.29, "learning_rate": 1.9107766703887764e-05, "loss": 0.0999, "step": 7042 }, { "epoch": 1.29, "learning_rate": 1.9107460129926804e-05, "loss": 0.1026, "step": 7043 }, { "epoch": 1.29, "learning_rate": 1.9107153505765308e-05, "loss": 0.0962, "step": 7044 }, { "epoch": 1.29, "learning_rate": 1.9106846831404966e-05, "loss": 0.104, "step": 7045 }, { "epoch": 1.29, "learning_rate": 1.9106540106847474e-05, "loss": 0.0894, "step": 7046 }, { "epoch": 1.29, "learning_rate": 1.910623333209452e-05, "loss": 0.1, "step": 7047 }, { "epoch": 1.29, "learning_rate": 1.910592650714779e-05, "loss": 0.0936, "step": 7048 }, { "epoch": 1.29, "learning_rate": 1.9105619632008983e-05, "loss": 0.1037, "step": 7049 }, { "epoch": 1.29, "learning_rate": 1.9105312706679785e-05, "loss": 0.1002, "step": 7050 }, { "epoch": 1.29, "learning_rate": 1.9105005731161897e-05, "loss": 0.1028, "step": 7051 }, { "epoch": 1.29, "learning_rate": 1.9104698705456996e-05, "loss": 0.1118, "step": 7052 }, { "epoch": 1.29, "learning_rate": 1.9104391629566787e-05, "loss": 0.0929, "step": 7053 }, { "epoch": 1.29, "learning_rate": 1.9104084503492957e-05, "loss": 0.0934, "step": 7054 }, { "epoch": 1.29, "learning_rate": 1.9103777327237202e-05, "loss": 0.1023, "step": 7055 }, { "epoch": 1.29, "learning_rate": 1.9103470100801206e-05, "loss": 0.0988, "step": 7056 }, { "epoch": 1.29, "learning_rate": 1.9103162824186676e-05, "loss": 0.1152, "step": 7057 }, { "epoch": 1.29, "learning_rate": 1.9102855497395297e-05, "loss": 0.0975, "step": 7058 }, { "epoch": 1.29, "learning_rate": 1.9102548120428766e-05, "loss": 0.0938, "step": 7059 }, { "epoch": 1.29, "learning_rate": 1.9102240693288777e-05, "loss": 0.0898, "step": 7060 }, { "epoch": 1.29, "learning_rate": 1.910193321597702e-05, "loss": 0.1031, "step": 7061 }, { "epoch": 1.3, "learning_rate": 1.91016256884952e-05, "loss": 0.0991, "step": 7062 }, { "epoch": 1.3, "learning_rate": 1.9101318110845004e-05, "loss": 0.0938, "step": 7063 }, { "epoch": 1.3, "learning_rate": 1.9101010483028127e-05, "loss": 0.0989, "step": 7064 }, { "epoch": 1.3, "learning_rate": 1.910070280504627e-05, "loss": 0.1088, "step": 7065 }, { "epoch": 1.3, "learning_rate": 1.9100395076901122e-05, "loss": 0.1099, "step": 7066 }, { "epoch": 1.3, "learning_rate": 1.910008729859439e-05, "loss": 0.0951, "step": 7067 }, { "epoch": 1.3, "learning_rate": 1.9099779470127754e-05, "loss": 0.0935, "step": 7068 }, { "epoch": 1.3, "learning_rate": 1.9099471591502927e-05, "loss": 0.092, "step": 7069 }, { "epoch": 1.3, "learning_rate": 1.9099163662721597e-05, "loss": 0.0886, "step": 7070 }, { "epoch": 1.3, "learning_rate": 1.9098855683785468e-05, "loss": 0.1002, "step": 7071 }, { "epoch": 1.3, "learning_rate": 1.9098547654696226e-05, "loss": 0.0955, "step": 7072 }, { "epoch": 1.3, "learning_rate": 1.9098239575455583e-05, "loss": 0.1024, "step": 7073 }, { "epoch": 1.3, "learning_rate": 1.9097931446065226e-05, "loss": 0.0966, "step": 7074 }, { "epoch": 1.3, "learning_rate": 1.909762326652686e-05, "loss": 0.0982, "step": 7075 }, { "epoch": 1.3, "learning_rate": 1.909731503684218e-05, "loss": 0.106, "step": 7076 }, { "epoch": 1.3, "learning_rate": 1.9097006757012888e-05, "loss": 0.1022, "step": 7077 }, { "epoch": 1.3, "learning_rate": 1.909669842704068e-05, "loss": 0.0992, "step": 7078 }, { "epoch": 1.3, "learning_rate": 1.9096390046927257e-05, "loss": 0.1059, "step": 7079 }, { "epoch": 1.3, "learning_rate": 1.909608161667432e-05, "loss": 0.0926, "step": 7080 }, { "epoch": 1.3, "learning_rate": 1.909577313628357e-05, "loss": 0.1157, "step": 7081 }, { "epoch": 1.3, "learning_rate": 1.9095464605756702e-05, "loss": 0.0971, "step": 7082 }, { "epoch": 1.3, "learning_rate": 1.909515602509542e-05, "loss": 0.0866, "step": 7083 }, { "epoch": 1.3, "learning_rate": 1.9094847394301427e-05, "loss": 0.0939, "step": 7084 }, { "epoch": 1.3, "learning_rate": 1.9094538713376422e-05, "loss": 0.1044, "step": 7085 }, { "epoch": 1.3, "learning_rate": 1.9094229982322107e-05, "loss": 0.0771, "step": 7086 }, { "epoch": 1.3, "learning_rate": 1.909392120114018e-05, "loss": 0.1074, "step": 7087 }, { "epoch": 1.3, "learning_rate": 1.9093612369832348e-05, "loss": 0.092, "step": 7088 }, { "epoch": 1.3, "learning_rate": 1.9093303488400313e-05, "loss": 0.0913, "step": 7089 }, { "epoch": 1.3, "learning_rate": 1.9092994556845774e-05, "loss": 0.1048, "step": 7090 }, { "epoch": 1.3, "learning_rate": 1.909268557517044e-05, "loss": 0.0993, "step": 7091 }, { "epoch": 1.3, "learning_rate": 1.9092376543376004e-05, "loss": 0.0951, "step": 7092 }, { "epoch": 1.3, "learning_rate": 1.909206746146418e-05, "loss": 0.1073, "step": 7093 }, { "epoch": 1.3, "learning_rate": 1.9091758329436664e-05, "loss": 0.0965, "step": 7094 }, { "epoch": 1.3, "learning_rate": 1.9091449147295166e-05, "loss": 0.1175, "step": 7095 }, { "epoch": 1.3, "learning_rate": 1.9091139915041386e-05, "loss": 0.0972, "step": 7096 }, { "epoch": 1.3, "learning_rate": 1.9090830632677032e-05, "loss": 0.0925, "step": 7097 }, { "epoch": 1.3, "learning_rate": 1.9090521300203805e-05, "loss": 0.0917, "step": 7098 }, { "epoch": 1.3, "learning_rate": 1.909021191762341e-05, "loss": 0.0994, "step": 7099 }, { "epoch": 1.3, "learning_rate": 1.9089902484937554e-05, "loss": 0.1006, "step": 7100 }, { "epoch": 1.3, "learning_rate": 1.9089593002147945e-05, "loss": 0.1127, "step": 7101 }, { "epoch": 1.3, "learning_rate": 1.9089283469256284e-05, "loss": 0.0988, "step": 7102 }, { "epoch": 1.3, "learning_rate": 1.908897388626428e-05, "loss": 0.1078, "step": 7103 }, { "epoch": 1.3, "learning_rate": 1.908866425317364e-05, "loss": 0.0996, "step": 7104 }, { "epoch": 1.3, "learning_rate": 1.908835456998607e-05, "loss": 0.0964, "step": 7105 }, { "epoch": 1.3, "learning_rate": 1.9088044836703273e-05, "loss": 0.1027, "step": 7106 }, { "epoch": 1.3, "learning_rate": 1.9087735053326964e-05, "loss": 0.1126, "step": 7107 }, { "epoch": 1.3, "learning_rate": 1.9087425219858844e-05, "loss": 0.1029, "step": 7108 }, { "epoch": 1.3, "learning_rate": 1.908711533630063e-05, "loss": 0.0917, "step": 7109 }, { "epoch": 1.3, "learning_rate": 1.9086805402654013e-05, "loss": 0.1006, "step": 7110 }, { "epoch": 1.3, "learning_rate": 1.9086495418920717e-05, "loss": 0.0983, "step": 7111 }, { "epoch": 1.3, "learning_rate": 1.9086185385102448e-05, "loss": 0.091, "step": 7112 }, { "epoch": 1.3, "learning_rate": 1.908587530120091e-05, "loss": 0.1081, "step": 7113 }, { "epoch": 1.3, "learning_rate": 1.9085565167217812e-05, "loss": 0.1013, "step": 7114 }, { "epoch": 1.3, "learning_rate": 1.9085254983154868e-05, "loss": 0.1088, "step": 7115 }, { "epoch": 1.3, "learning_rate": 1.9084944749013784e-05, "loss": 0.0925, "step": 7116 }, { "epoch": 1.31, "learning_rate": 1.9084634464796275e-05, "loss": 0.1045, "step": 7117 }, { "epoch": 1.31, "learning_rate": 1.9084324130504045e-05, "loss": 0.115, "step": 7118 }, { "epoch": 1.31, "learning_rate": 1.908401374613881e-05, "loss": 0.0901, "step": 7119 }, { "epoch": 1.31, "learning_rate": 1.908370331170228e-05, "loss": 0.0953, "step": 7120 }, { "epoch": 1.31, "learning_rate": 1.908339282719616e-05, "loss": 0.1016, "step": 7121 }, { "epoch": 1.31, "learning_rate": 1.908308229262217e-05, "loss": 0.0877, "step": 7122 }, { "epoch": 1.31, "learning_rate": 1.9082771707982013e-05, "loss": 0.1102, "step": 7123 }, { "epoch": 1.31, "learning_rate": 1.908246107327741e-05, "loss": 0.0919, "step": 7124 }, { "epoch": 1.31, "learning_rate": 1.9082150388510067e-05, "loss": 0.0836, "step": 7125 }, { "epoch": 1.31, "learning_rate": 1.9081839653681696e-05, "loss": 0.0982, "step": 7126 }, { "epoch": 1.31, "learning_rate": 1.9081528868794013e-05, "loss": 0.0952, "step": 7127 }, { "epoch": 1.31, "learning_rate": 1.9081218033848733e-05, "loss": 0.0931, "step": 7128 }, { "epoch": 1.31, "learning_rate": 1.9080907148847566e-05, "loss": 0.0977, "step": 7129 }, { "epoch": 1.31, "learning_rate": 1.908059621379222e-05, "loss": 0.0877, "step": 7130 }, { "epoch": 1.31, "learning_rate": 1.908028522868442e-05, "loss": 0.1093, "step": 7131 }, { "epoch": 1.31, "learning_rate": 1.907997419352588e-05, "loss": 0.1072, "step": 7132 }, { "epoch": 1.31, "learning_rate": 1.9079663108318304e-05, "loss": 0.1166, "step": 7133 }, { "epoch": 1.31, "learning_rate": 1.9079351973063408e-05, "loss": 0.094, "step": 7134 }, { "epoch": 1.31, "learning_rate": 1.9079040787762915e-05, "loss": 0.1004, "step": 7135 }, { "epoch": 1.31, "learning_rate": 1.907872955241854e-05, "loss": 0.1026, "step": 7136 }, { "epoch": 1.31, "learning_rate": 1.907841826703199e-05, "loss": 0.0953, "step": 7137 }, { "epoch": 1.31, "learning_rate": 1.907810693160499e-05, "loss": 0.1028, "step": 7138 }, { "epoch": 1.31, "learning_rate": 1.907779554613925e-05, "loss": 0.1033, "step": 7139 }, { "epoch": 1.31, "learning_rate": 1.9077484110636487e-05, "loss": 0.1103, "step": 7140 }, { "epoch": 1.31, "learning_rate": 1.9077172625098424e-05, "loss": 0.0911, "step": 7141 }, { "epoch": 1.31, "learning_rate": 1.9076861089526768e-05, "loss": 0.0953, "step": 7142 }, { "epoch": 1.31, "learning_rate": 1.907654950392324e-05, "loss": 0.1023, "step": 7143 }, { "epoch": 1.31, "learning_rate": 1.907623786828956e-05, "loss": 0.1076, "step": 7144 }, { "epoch": 1.31, "learning_rate": 1.9075926182627448e-05, "loss": 0.0945, "step": 7145 }, { "epoch": 1.31, "learning_rate": 1.9075614446938614e-05, "loss": 0.0978, "step": 7146 }, { "epoch": 1.31, "learning_rate": 1.907530266122478e-05, "loss": 0.0985, "step": 7147 }, { "epoch": 1.31, "learning_rate": 1.9074990825487667e-05, "loss": 0.1034, "step": 7148 }, { "epoch": 1.31, "learning_rate": 1.9074678939728993e-05, "loss": 0.095, "step": 7149 }, { "epoch": 1.31, "learning_rate": 1.9074367003950475e-05, "loss": 0.0998, "step": 7150 }, { "epoch": 1.31, "learning_rate": 1.9074055018153835e-05, "loss": 0.0989, "step": 7151 }, { "epoch": 1.31, "learning_rate": 1.9073742982340787e-05, "loss": 0.0884, "step": 7152 }, { "epoch": 1.31, "learning_rate": 1.9073430896513056e-05, "loss": 0.1004, "step": 7153 }, { "epoch": 1.31, "learning_rate": 1.9073118760672366e-05, "loss": 0.1048, "step": 7154 }, { "epoch": 1.31, "learning_rate": 1.907280657482043e-05, "loss": 0.1009, "step": 7155 }, { "epoch": 1.31, "learning_rate": 1.9072494338958968e-05, "loss": 0.0953, "step": 7156 }, { "epoch": 1.31, "learning_rate": 1.907218205308971e-05, "loss": 0.103, "step": 7157 }, { "epoch": 1.31, "learning_rate": 1.907186971721437e-05, "loss": 0.1078, "step": 7158 }, { "epoch": 1.31, "learning_rate": 1.907155733133467e-05, "loss": 0.0964, "step": 7159 }, { "epoch": 1.31, "learning_rate": 1.9071244895452334e-05, "loss": 0.1049, "step": 7160 }, { "epoch": 1.31, "learning_rate": 1.9070932409569085e-05, "loss": 0.1071, "step": 7161 }, { "epoch": 1.31, "learning_rate": 1.9070619873686643e-05, "loss": 0.1063, "step": 7162 }, { "epoch": 1.31, "learning_rate": 1.9070307287806734e-05, "loss": 0.1014, "step": 7163 }, { "epoch": 1.31, "learning_rate": 1.9069994651931077e-05, "loss": 0.1072, "step": 7164 }, { "epoch": 1.31, "learning_rate": 1.9069681966061398e-05, "loss": 0.0979, "step": 7165 }, { "epoch": 1.31, "learning_rate": 1.906936923019942e-05, "loss": 0.0997, "step": 7166 }, { "epoch": 1.31, "learning_rate": 1.9069056444346863e-05, "loss": 0.0967, "step": 7167 }, { "epoch": 1.31, "learning_rate": 1.9068743608505454e-05, "loss": 0.1086, "step": 7168 }, { "epoch": 1.31, "learning_rate": 1.9068430722676924e-05, "loss": 0.0967, "step": 7169 }, { "epoch": 1.31, "learning_rate": 1.9068117786862984e-05, "loss": 0.0946, "step": 7170 }, { "epoch": 1.32, "learning_rate": 1.9067804801065372e-05, "loss": 0.1033, "step": 7171 }, { "epoch": 1.32, "learning_rate": 1.9067491765285803e-05, "loss": 0.1069, "step": 7172 }, { "epoch": 1.32, "learning_rate": 1.906717867952601e-05, "loss": 0.1017, "step": 7173 }, { "epoch": 1.32, "learning_rate": 1.9066865543787715e-05, "loss": 0.1034, "step": 7174 }, { "epoch": 1.32, "learning_rate": 1.9066552358072648e-05, "loss": 0.1042, "step": 7175 }, { "epoch": 1.32, "learning_rate": 1.906623912238253e-05, "loss": 0.1041, "step": 7176 }, { "epoch": 1.32, "learning_rate": 1.9065925836719086e-05, "loss": 0.0973, "step": 7177 }, { "epoch": 1.32, "learning_rate": 1.9065612501084046e-05, "loss": 0.1029, "step": 7178 }, { "epoch": 1.32, "learning_rate": 1.906529911547914e-05, "loss": 0.1025, "step": 7179 }, { "epoch": 1.32, "learning_rate": 1.9064985679906093e-05, "loss": 0.1022, "step": 7180 }, { "epoch": 1.32, "learning_rate": 1.9064672194366635e-05, "loss": 0.1069, "step": 7181 }, { "epoch": 1.32, "learning_rate": 1.9064358658862486e-05, "loss": 0.1164, "step": 7182 }, { "epoch": 1.32, "learning_rate": 1.9064045073395387e-05, "loss": 0.1056, "step": 7183 }, { "epoch": 1.32, "learning_rate": 1.9063731437967052e-05, "loss": 0.0988, "step": 7184 }, { "epoch": 1.32, "learning_rate": 1.9063417752579223e-05, "loss": 0.0993, "step": 7185 }, { "epoch": 1.32, "learning_rate": 1.906310401723362e-05, "loss": 0.0917, "step": 7186 }, { "epoch": 1.32, "learning_rate": 1.9062790231931972e-05, "loss": 0.0961, "step": 7187 }, { "epoch": 1.32, "learning_rate": 1.9062476396676016e-05, "loss": 0.0996, "step": 7188 }, { "epoch": 1.32, "learning_rate": 1.906216251146748e-05, "loss": 0.0924, "step": 7189 }, { "epoch": 1.32, "learning_rate": 1.9061848576308087e-05, "loss": 0.1008, "step": 7190 }, { "epoch": 1.32, "learning_rate": 1.9061534591199575e-05, "loss": 0.0943, "step": 7191 }, { "epoch": 1.32, "learning_rate": 1.906122055614367e-05, "loss": 0.1044, "step": 7192 }, { "epoch": 1.32, "learning_rate": 1.9060906471142106e-05, "loss": 0.0962, "step": 7193 }, { "epoch": 1.32, "learning_rate": 1.9060592336196616e-05, "loss": 0.0946, "step": 7194 }, { "epoch": 1.32, "learning_rate": 1.9060278151308925e-05, "loss": 0.1143, "step": 7195 }, { "epoch": 1.32, "learning_rate": 1.905996391648077e-05, "loss": 0.1002, "step": 7196 }, { "epoch": 1.32, "learning_rate": 1.9059649631713882e-05, "loss": 0.1112, "step": 7197 }, { "epoch": 1.32, "learning_rate": 1.9059335297009994e-05, "loss": 0.1002, "step": 7198 }, { "epoch": 1.32, "learning_rate": 1.9059020912370836e-05, "loss": 0.1028, "step": 7199 }, { "epoch": 1.32, "learning_rate": 1.905870647779814e-05, "loss": 0.0975, "step": 7200 }, { "epoch": 1.32, "learning_rate": 1.9058391993293646e-05, "loss": 0.1005, "step": 7201 }, { "epoch": 1.32, "learning_rate": 1.9058077458859083e-05, "loss": 0.0865, "step": 7202 }, { "epoch": 1.32, "learning_rate": 1.9057762874496183e-05, "loss": 0.0996, "step": 7203 }, { "epoch": 1.32, "learning_rate": 1.905744824020668e-05, "loss": 0.0966, "step": 7204 }, { "epoch": 1.32, "learning_rate": 1.9057133555992316e-05, "loss": 0.0926, "step": 7205 }, { "epoch": 1.32, "learning_rate": 1.9056818821854814e-05, "loss": 0.1074, "step": 7206 }, { "epoch": 1.32, "learning_rate": 1.905650403779592e-05, "loss": 0.1084, "step": 7207 }, { "epoch": 1.32, "learning_rate": 1.905618920381736e-05, "loss": 0.093, "step": 7208 }, { "epoch": 1.32, "learning_rate": 1.9055874319920872e-05, "loss": 0.0998, "step": 7209 }, { "epoch": 1.32, "learning_rate": 1.905555938610819e-05, "loss": 0.0918, "step": 7210 }, { "epoch": 1.32, "learning_rate": 1.905524440238106e-05, "loss": 0.103, "step": 7211 }, { "epoch": 1.32, "learning_rate": 1.9054929368741203e-05, "loss": 0.0984, "step": 7212 }, { "epoch": 1.32, "learning_rate": 1.905461428519037e-05, "loss": 0.0981, "step": 7213 }, { "epoch": 1.32, "learning_rate": 1.9054299151730288e-05, "loss": 0.0954, "step": 7214 }, { "epoch": 1.32, "learning_rate": 1.9053983968362696e-05, "loss": 0.0881, "step": 7215 }, { "epoch": 1.32, "learning_rate": 1.9053668735089333e-05, "loss": 0.0979, "step": 7216 }, { "epoch": 1.32, "learning_rate": 1.9053353451911936e-05, "loss": 0.1056, "step": 7217 }, { "epoch": 1.32, "learning_rate": 1.9053038118832244e-05, "loss": 0.1061, "step": 7218 }, { "epoch": 1.32, "learning_rate": 1.905272273585199e-05, "loss": 0.0984, "step": 7219 }, { "epoch": 1.32, "learning_rate": 1.905240730297292e-05, "loss": 0.1016, "step": 7220 }, { "epoch": 1.32, "learning_rate": 1.9052091820196764e-05, "loss": 0.099, "step": 7221 }, { "epoch": 1.32, "learning_rate": 1.9051776287525273e-05, "loss": 0.0912, "step": 7222 }, { "epoch": 1.32, "learning_rate": 1.9051460704960175e-05, "loss": 0.1076, "step": 7223 }, { "epoch": 1.32, "learning_rate": 1.9051145072503216e-05, "loss": 0.0966, "step": 7224 }, { "epoch": 1.32, "learning_rate": 1.905082939015613e-05, "loss": 0.0946, "step": 7225 }, { "epoch": 1.33, "learning_rate": 1.905051365792066e-05, "loss": 0.0924, "step": 7226 }, { "epoch": 1.33, "learning_rate": 1.905019787579855e-05, "loss": 0.0867, "step": 7227 }, { "epoch": 1.33, "learning_rate": 1.9049882043791535e-05, "loss": 0.108, "step": 7228 }, { "epoch": 1.33, "learning_rate": 1.9049566161901357e-05, "loss": 0.0906, "step": 7229 }, { "epoch": 1.33, "learning_rate": 1.9049250230129762e-05, "loss": 0.0944, "step": 7230 }, { "epoch": 1.33, "learning_rate": 1.9048934248478487e-05, "loss": 0.1006, "step": 7231 }, { "epoch": 1.33, "learning_rate": 1.904861821694927e-05, "loss": 0.0942, "step": 7232 }, { "epoch": 1.33, "learning_rate": 1.9048302135543863e-05, "loss": 0.103, "step": 7233 }, { "epoch": 1.33, "learning_rate": 1.9047986004263997e-05, "loss": 0.0928, "step": 7234 }, { "epoch": 1.33, "learning_rate": 1.9047669823111425e-05, "loss": 0.0965, "step": 7235 }, { "epoch": 1.33, "learning_rate": 1.9047353592087883e-05, "loss": 0.1006, "step": 7236 }, { "epoch": 1.33, "learning_rate": 1.9047037311195118e-05, "loss": 0.0987, "step": 7237 }, { "epoch": 1.33, "learning_rate": 1.904672098043487e-05, "loss": 0.1013, "step": 7238 }, { "epoch": 1.33, "learning_rate": 1.9046404599808884e-05, "loss": 0.0854, "step": 7239 }, { "epoch": 1.33, "learning_rate": 1.9046088169318902e-05, "loss": 0.0932, "step": 7240 }, { "epoch": 1.33, "learning_rate": 1.9045771688966673e-05, "loss": 0.0917, "step": 7241 }, { "epoch": 1.33, "learning_rate": 1.9045455158753937e-05, "loss": 0.1072, "step": 7242 }, { "epoch": 1.33, "learning_rate": 1.904513857868244e-05, "loss": 0.0958, "step": 7243 }, { "epoch": 1.33, "learning_rate": 1.9044821948753925e-05, "loss": 0.0968, "step": 7244 }, { "epoch": 1.33, "learning_rate": 1.9044505268970143e-05, "loss": 0.1002, "step": 7245 }, { "epoch": 1.33, "learning_rate": 1.904418853933284e-05, "loss": 0.1064, "step": 7246 }, { "epoch": 1.33, "learning_rate": 1.904387175984375e-05, "loss": 0.1015, "step": 7247 }, { "epoch": 1.33, "learning_rate": 1.904355493050463e-05, "loss": 0.1022, "step": 7248 }, { "epoch": 1.33, "learning_rate": 1.9043238051317225e-05, "loss": 0.1, "step": 7249 }, { "epoch": 1.33, "learning_rate": 1.904292112228328e-05, "loss": 0.1025, "step": 7250 }, { "epoch": 1.33, "learning_rate": 1.904260414340454e-05, "loss": 0.0875, "step": 7251 }, { "epoch": 1.33, "learning_rate": 1.9042287114682755e-05, "loss": 0.0952, "step": 7252 }, { "epoch": 1.33, "learning_rate": 1.904197003611967e-05, "loss": 0.0937, "step": 7253 }, { "epoch": 1.33, "learning_rate": 1.9041652907717037e-05, "loss": 0.1022, "step": 7254 }, { "epoch": 1.33, "learning_rate": 1.90413357294766e-05, "loss": 0.0979, "step": 7255 }, { "epoch": 1.33, "learning_rate": 1.9041018501400108e-05, "loss": 0.0869, "step": 7256 }, { "epoch": 1.33, "learning_rate": 1.904070122348931e-05, "loss": 0.1014, "step": 7257 }, { "epoch": 1.33, "learning_rate": 1.9040383895745956e-05, "loss": 0.1004, "step": 7258 }, { "epoch": 1.33, "learning_rate": 1.9040066518171793e-05, "loss": 0.0986, "step": 7259 }, { "epoch": 1.33, "learning_rate": 1.9039749090768573e-05, "loss": 0.1142, "step": 7260 }, { "epoch": 1.33, "learning_rate": 1.9039431613538047e-05, "loss": 0.1003, "step": 7261 }, { "epoch": 1.33, "learning_rate": 1.903911408648196e-05, "loss": 0.0992, "step": 7262 }, { "epoch": 1.33, "learning_rate": 1.903879650960206e-05, "loss": 0.098, "step": 7263 }, { "epoch": 1.33, "learning_rate": 1.9038478882900104e-05, "loss": 0.0888, "step": 7264 }, { "epoch": 1.33, "learning_rate": 1.9038161206377843e-05, "loss": 0.1095, "step": 7265 }, { "epoch": 1.33, "learning_rate": 1.9037843480037024e-05, "loss": 0.0893, "step": 7266 }, { "epoch": 1.33, "learning_rate": 1.9037525703879402e-05, "loss": 0.1097, "step": 7267 }, { "epoch": 1.33, "learning_rate": 1.9037207877906726e-05, "loss": 0.1062, "step": 7268 }, { "epoch": 1.33, "learning_rate": 1.9036890002120747e-05, "loss": 0.1007, "step": 7269 }, { "epoch": 1.33, "learning_rate": 1.903657207652322e-05, "loss": 0.114, "step": 7270 }, { "epoch": 1.33, "learning_rate": 1.9036254101115893e-05, "loss": 0.092, "step": 7271 }, { "epoch": 1.33, "learning_rate": 1.9035936075900525e-05, "loss": 0.0867, "step": 7272 }, { "epoch": 1.33, "learning_rate": 1.9035618000878865e-05, "loss": 0.1033, "step": 7273 }, { "epoch": 1.33, "learning_rate": 1.9035299876052666e-05, "loss": 0.0975, "step": 7274 }, { "epoch": 1.33, "learning_rate": 1.9034981701423682e-05, "loss": 0.1004, "step": 7275 }, { "epoch": 1.33, "learning_rate": 1.903466347699367e-05, "loss": 0.0972, "step": 7276 }, { "epoch": 1.33, "learning_rate": 1.9034345202764378e-05, "loss": 0.1031, "step": 7277 }, { "epoch": 1.33, "learning_rate": 1.9034026878737565e-05, "loss": 0.1063, "step": 7278 }, { "epoch": 1.33, "learning_rate": 1.9033708504914984e-05, "loss": 0.0978, "step": 7279 }, { "epoch": 1.34, "learning_rate": 1.903339008129839e-05, "loss": 0.1114, "step": 7280 }, { "epoch": 1.34, "learning_rate": 1.9033071607889542e-05, "loss": 0.0909, "step": 7281 }, { "epoch": 1.34, "learning_rate": 1.9032753084690186e-05, "loss": 0.0947, "step": 7282 }, { "epoch": 1.34, "learning_rate": 1.9032434511702087e-05, "loss": 0.0834, "step": 7283 }, { "epoch": 1.34, "learning_rate": 1.9032115888926995e-05, "loss": 0.0977, "step": 7284 }, { "epoch": 1.34, "learning_rate": 1.903179721636667e-05, "loss": 0.0988, "step": 7285 }, { "epoch": 1.34, "learning_rate": 1.9031478494022865e-05, "loss": 0.0928, "step": 7286 }, { "epoch": 1.34, "learning_rate": 1.9031159721897344e-05, "loss": 0.0929, "step": 7287 }, { "epoch": 1.34, "learning_rate": 1.9030840899991853e-05, "loss": 0.1027, "step": 7288 }, { "epoch": 1.34, "learning_rate": 1.903052202830816e-05, "loss": 0.105, "step": 7289 }, { "epoch": 1.34, "learning_rate": 1.9030203106848017e-05, "loss": 0.0995, "step": 7290 }, { "epoch": 1.34, "learning_rate": 1.902988413561318e-05, "loss": 0.103, "step": 7291 }, { "epoch": 1.34, "learning_rate": 1.9029565114605414e-05, "loss": 0.0958, "step": 7292 }, { "epoch": 1.34, "learning_rate": 1.9029246043826467e-05, "loss": 0.1007, "step": 7293 }, { "epoch": 1.34, "learning_rate": 1.902892692327811e-05, "loss": 0.1024, "step": 7294 }, { "epoch": 1.34, "learning_rate": 1.9028607752962097e-05, "loss": 0.0944, "step": 7295 }, { "epoch": 1.34, "learning_rate": 1.9028288532880187e-05, "loss": 0.0851, "step": 7296 }, { "epoch": 1.34, "learning_rate": 1.9027969263034135e-05, "loss": 0.1123, "step": 7297 }, { "epoch": 1.34, "learning_rate": 1.9027649943425705e-05, "loss": 0.1104, "step": 7298 }, { "epoch": 1.34, "learning_rate": 1.9027330574056663e-05, "loss": 0.1051, "step": 7299 }, { "epoch": 1.34, "learning_rate": 1.9027011154928756e-05, "loss": 0.106, "step": 7300 }, { "epoch": 1.34, "learning_rate": 1.9026691686043755e-05, "loss": 0.1053, "step": 7301 }, { "epoch": 1.34, "learning_rate": 1.902637216740342e-05, "loss": 0.1001, "step": 7302 }, { "epoch": 1.34, "learning_rate": 1.9026052599009506e-05, "loss": 0.0916, "step": 7303 }, { "epoch": 1.34, "learning_rate": 1.902573298086378e-05, "loss": 0.0907, "step": 7304 }, { "epoch": 1.34, "learning_rate": 1.9025413312968006e-05, "loss": 0.1041, "step": 7305 }, { "epoch": 1.34, "learning_rate": 1.902509359532394e-05, "loss": 0.0988, "step": 7306 }, { "epoch": 1.34, "learning_rate": 1.9024773827933344e-05, "loss": 0.1014, "step": 7307 }, { "epoch": 1.34, "learning_rate": 1.9024454010797984e-05, "loss": 0.1005, "step": 7308 }, { "epoch": 1.34, "learning_rate": 1.9024134143919625e-05, "loss": 0.1009, "step": 7309 }, { "epoch": 1.34, "learning_rate": 1.9023814227300028e-05, "loss": 0.0892, "step": 7310 }, { "epoch": 1.34, "learning_rate": 1.9023494260940953e-05, "loss": 0.1077, "step": 7311 }, { "epoch": 1.34, "learning_rate": 1.9023174244844167e-05, "loss": 0.1068, "step": 7312 }, { "epoch": 1.34, "learning_rate": 1.9022854179011433e-05, "loss": 0.1016, "step": 7313 }, { "epoch": 1.34, "learning_rate": 1.9022534063444515e-05, "loss": 0.1072, "step": 7314 }, { "epoch": 1.34, "learning_rate": 1.9022213898145176e-05, "loss": 0.1039, "step": 7315 }, { "epoch": 1.34, "learning_rate": 1.9021893683115188e-05, "loss": 0.1137, "step": 7316 }, { "epoch": 1.34, "learning_rate": 1.9021573418356305e-05, "loss": 0.1001, "step": 7317 }, { "epoch": 1.34, "learning_rate": 1.90212531038703e-05, "loss": 0.0998, "step": 7318 }, { "epoch": 1.34, "learning_rate": 1.9020932739658938e-05, "loss": 0.1012, "step": 7319 }, { "epoch": 1.34, "learning_rate": 1.902061232572398e-05, "loss": 0.1016, "step": 7320 }, { "epoch": 1.34, "learning_rate": 1.9020291862067203e-05, "loss": 0.1013, "step": 7321 }, { "epoch": 1.34, "learning_rate": 1.9019971348690357e-05, "loss": 0.1123, "step": 7322 }, { "epoch": 1.34, "learning_rate": 1.9019650785595222e-05, "loss": 0.1008, "step": 7323 }, { "epoch": 1.34, "learning_rate": 1.901933017278356e-05, "loss": 0.098, "step": 7324 }, { "epoch": 1.34, "learning_rate": 1.9019009510257137e-05, "loss": 0.1002, "step": 7325 }, { "epoch": 1.34, "learning_rate": 1.9018688798017727e-05, "loss": 0.1005, "step": 7326 }, { "epoch": 1.34, "learning_rate": 1.9018368036067086e-05, "loss": 0.0995, "step": 7327 }, { "epoch": 1.34, "learning_rate": 1.9018047224407e-05, "loss": 0.0819, "step": 7328 }, { "epoch": 1.34, "learning_rate": 1.9017726363039216e-05, "loss": 0.096, "step": 7329 }, { "epoch": 1.34, "learning_rate": 1.9017405451965517e-05, "loss": 0.0928, "step": 7330 }, { "epoch": 1.34, "learning_rate": 1.9017084491187665e-05, "loss": 0.0931, "step": 7331 }, { "epoch": 1.34, "learning_rate": 1.9016763480707436e-05, "loss": 0.0939, "step": 7332 }, { "epoch": 1.34, "learning_rate": 1.9016442420526594e-05, "loss": 0.1041, "step": 7333 }, { "epoch": 1.34, "learning_rate": 1.9016121310646907e-05, "loss": 0.0994, "step": 7334 }, { "epoch": 1.35, "learning_rate": 1.9015800151070154e-05, "loss": 0.0975, "step": 7335 }, { "epoch": 1.35, "learning_rate": 1.9015478941798097e-05, "loss": 0.1071, "step": 7336 }, { "epoch": 1.35, "learning_rate": 1.9015157682832506e-05, "loss": 0.0956, "step": 7337 }, { "epoch": 1.35, "learning_rate": 1.901483637417516e-05, "loss": 0.1098, "step": 7338 }, { "epoch": 1.35, "learning_rate": 1.901451501582782e-05, "loss": 0.1013, "step": 7339 }, { "epoch": 1.35, "learning_rate": 1.9014193607792265e-05, "loss": 0.0935, "step": 7340 }, { "epoch": 1.35, "learning_rate": 1.901387215007026e-05, "loss": 0.1069, "step": 7341 }, { "epoch": 1.35, "learning_rate": 1.9013550642663586e-05, "loss": 0.0859, "step": 7342 }, { "epoch": 1.35, "learning_rate": 1.9013229085574006e-05, "loss": 0.0975, "step": 7343 }, { "epoch": 1.35, "learning_rate": 1.9012907478803298e-05, "loss": 0.0855, "step": 7344 }, { "epoch": 1.35, "learning_rate": 1.901258582235323e-05, "loss": 0.0993, "step": 7345 }, { "epoch": 1.35, "learning_rate": 1.901226411622558e-05, "loss": 0.098, "step": 7346 }, { "epoch": 1.35, "learning_rate": 1.901194236042212e-05, "loss": 0.0991, "step": 7347 }, { "epoch": 1.35, "learning_rate": 1.9011620554944617e-05, "loss": 0.112, "step": 7348 }, { "epoch": 1.35, "learning_rate": 1.9011298699794857e-05, "loss": 0.1124, "step": 7349 }, { "epoch": 1.35, "learning_rate": 1.9010976794974606e-05, "loss": 0.0939, "step": 7350 }, { "epoch": 1.35, "learning_rate": 1.901065484048564e-05, "loss": 0.1174, "step": 7351 }, { "epoch": 1.35, "learning_rate": 1.901033283632973e-05, "loss": 0.107, "step": 7352 }, { "epoch": 1.35, "learning_rate": 1.9010010782508657e-05, "loss": 0.1028, "step": 7353 }, { "epoch": 1.35, "learning_rate": 1.900968867902419e-05, "loss": 0.109, "step": 7354 }, { "epoch": 1.35, "learning_rate": 1.9009366525878113e-05, "loss": 0.1135, "step": 7355 }, { "epoch": 1.35, "learning_rate": 1.9009044323072197e-05, "loss": 0.1003, "step": 7356 }, { "epoch": 1.35, "learning_rate": 1.9008722070608212e-05, "loss": 0.0931, "step": 7357 }, { "epoch": 1.35, "learning_rate": 1.9008399768487944e-05, "loss": 0.0986, "step": 7358 }, { "epoch": 1.35, "learning_rate": 1.9008077416713165e-05, "loss": 0.0937, "step": 7359 }, { "epoch": 1.35, "learning_rate": 1.9007755015285653e-05, "loss": 0.1026, "step": 7360 }, { "epoch": 1.35, "learning_rate": 1.9007432564207184e-05, "loss": 0.0945, "step": 7361 }, { "epoch": 1.35, "learning_rate": 1.9007110063479533e-05, "loss": 0.0944, "step": 7362 }, { "epoch": 1.35, "learning_rate": 1.9006787513104482e-05, "loss": 0.0988, "step": 7363 }, { "epoch": 1.35, "learning_rate": 1.9006464913083807e-05, "loss": 0.0896, "step": 7364 }, { "epoch": 1.35, "learning_rate": 1.9006142263419287e-05, "loss": 0.0962, "step": 7365 }, { "epoch": 1.35, "learning_rate": 1.90058195641127e-05, "loss": 0.1038, "step": 7366 }, { "epoch": 1.35, "learning_rate": 1.900549681516582e-05, "loss": 0.1198, "step": 7367 }, { "epoch": 1.35, "learning_rate": 1.9005174016580437e-05, "loss": 0.0878, "step": 7368 }, { "epoch": 1.35, "learning_rate": 1.9004851168358323e-05, "loss": 0.0934, "step": 7369 }, { "epoch": 1.35, "learning_rate": 1.9004528270501256e-05, "loss": 0.093, "step": 7370 }, { "epoch": 1.35, "learning_rate": 1.900420532301102e-05, "loss": 0.101, "step": 7371 }, { "epoch": 1.35, "learning_rate": 1.9003882325889394e-05, "loss": 0.1031, "step": 7372 }, { "epoch": 1.35, "learning_rate": 1.9003559279138156e-05, "loss": 0.1016, "step": 7373 }, { "epoch": 1.35, "learning_rate": 1.9003236182759084e-05, "loss": 0.0984, "step": 7374 }, { "epoch": 1.35, "learning_rate": 1.9002913036753972e-05, "loss": 0.1039, "step": 7375 }, { "epoch": 1.35, "learning_rate": 1.9002589841124587e-05, "loss": 0.0933, "step": 7376 }, { "epoch": 1.35, "learning_rate": 1.9002266595872716e-05, "loss": 0.0947, "step": 7377 }, { "epoch": 1.35, "learning_rate": 1.9001943301000144e-05, "loss": 0.0941, "step": 7378 }, { "epoch": 1.35, "learning_rate": 1.9001619956508645e-05, "loss": 0.1043, "step": 7379 }, { "epoch": 1.35, "learning_rate": 1.900129656240001e-05, "loss": 0.1059, "step": 7380 }, { "epoch": 1.35, "learning_rate": 1.9000973118676014e-05, "loss": 0.1018, "step": 7381 }, { "epoch": 1.35, "learning_rate": 1.9000649625338444e-05, "loss": 0.1022, "step": 7382 }, { "epoch": 1.35, "learning_rate": 1.900032608238909e-05, "loss": 0.095, "step": 7383 }, { "epoch": 1.35, "learning_rate": 1.9000002489829717e-05, "loss": 0.1013, "step": 7384 }, { "epoch": 1.35, "learning_rate": 1.8999678847662124e-05, "loss": 0.1071, "step": 7385 }, { "epoch": 1.35, "learning_rate": 1.899935515588809e-05, "loss": 0.0899, "step": 7386 }, { "epoch": 1.35, "learning_rate": 1.89990314145094e-05, "loss": 0.0947, "step": 7387 }, { "epoch": 1.35, "learning_rate": 1.8998707623527838e-05, "loss": 0.0965, "step": 7388 }, { "epoch": 1.36, "learning_rate": 1.8998383782945185e-05, "loss": 0.0981, "step": 7389 }, { "epoch": 1.36, "learning_rate": 1.8998059892763234e-05, "loss": 0.1163, "step": 7390 }, { "epoch": 1.36, "learning_rate": 1.8997735952983767e-05, "loss": 0.1123, "step": 7391 }, { "epoch": 1.36, "learning_rate": 1.8997411963608566e-05, "loss": 0.1021, "step": 7392 }, { "epoch": 1.36, "learning_rate": 1.899708792463942e-05, "loss": 0.1139, "step": 7393 }, { "epoch": 1.36, "learning_rate": 1.8996763836078114e-05, "loss": 0.0933, "step": 7394 }, { "epoch": 1.36, "learning_rate": 1.8996439697926437e-05, "loss": 0.1048, "step": 7395 }, { "epoch": 1.36, "learning_rate": 1.8996115510186173e-05, "loss": 0.096, "step": 7396 }, { "epoch": 1.36, "learning_rate": 1.8995791272859108e-05, "loss": 0.095, "step": 7397 }, { "epoch": 1.36, "learning_rate": 1.899546698594703e-05, "loss": 0.094, "step": 7398 }, { "epoch": 1.36, "learning_rate": 1.899514264945173e-05, "loss": 0.0922, "step": 7399 }, { "epoch": 1.36, "learning_rate": 1.8994818263374994e-05, "loss": 0.1064, "step": 7400 }, { "epoch": 1.36, "learning_rate": 1.8994493827718604e-05, "loss": 0.0847, "step": 7401 }, { "epoch": 1.36, "learning_rate": 1.8994169342484356e-05, "loss": 0.096, "step": 7402 }, { "epoch": 1.36, "learning_rate": 1.8993844807674038e-05, "loss": 0.1049, "step": 7403 }, { "epoch": 1.36, "learning_rate": 1.8993520223289433e-05, "loss": 0.0901, "step": 7404 }, { "epoch": 1.36, "learning_rate": 1.8993195589332337e-05, "loss": 0.1028, "step": 7405 }, { "epoch": 1.36, "learning_rate": 1.8992870905804535e-05, "loss": 0.1027, "step": 7406 }, { "epoch": 1.36, "learning_rate": 1.8992546172707816e-05, "loss": 0.1077, "step": 7407 }, { "epoch": 1.36, "learning_rate": 1.8992221390043973e-05, "loss": 0.0927, "step": 7408 }, { "epoch": 1.36, "learning_rate": 1.8991896557814797e-05, "loss": 0.1078, "step": 7409 }, { "epoch": 1.36, "learning_rate": 1.8991571676022074e-05, "loss": 0.0991, "step": 7410 }, { "epoch": 1.36, "learning_rate": 1.89912467446676e-05, "loss": 0.1029, "step": 7411 }, { "epoch": 1.36, "learning_rate": 1.899092176375316e-05, "loss": 0.1036, "step": 7412 }, { "epoch": 1.36, "learning_rate": 1.8990596733280552e-05, "loss": 0.0988, "step": 7413 }, { "epoch": 1.36, "learning_rate": 1.8990271653251563e-05, "loss": 0.1029, "step": 7414 }, { "epoch": 1.36, "learning_rate": 1.8989946523667988e-05, "loss": 0.0928, "step": 7415 }, { "epoch": 1.36, "learning_rate": 1.8989621344531613e-05, "loss": 0.0934, "step": 7416 }, { "epoch": 1.36, "learning_rate": 1.8989296115844234e-05, "loss": 0.1064, "step": 7417 }, { "epoch": 1.36, "learning_rate": 1.8988970837607648e-05, "loss": 0.0834, "step": 7418 }, { "epoch": 1.36, "learning_rate": 1.8988645509823642e-05, "loss": 0.1001, "step": 7419 }, { "epoch": 1.36, "learning_rate": 1.8988320132494012e-05, "loss": 0.1028, "step": 7420 }, { "epoch": 1.36, "learning_rate": 1.898799470562055e-05, "loss": 0.1054, "step": 7421 }, { "epoch": 1.36, "learning_rate": 1.8987669229205052e-05, "loss": 0.1101, "step": 7422 }, { "epoch": 1.36, "learning_rate": 1.898734370324931e-05, "loss": 0.1002, "step": 7423 }, { "epoch": 1.36, "learning_rate": 1.898701812775512e-05, "loss": 0.0872, "step": 7424 }, { "epoch": 1.36, "learning_rate": 1.8986692502724273e-05, "loss": 0.1069, "step": 7425 }, { "epoch": 1.36, "learning_rate": 1.898636682815857e-05, "loss": 0.0929, "step": 7426 }, { "epoch": 1.36, "learning_rate": 1.89860411040598e-05, "loss": 0.0874, "step": 7427 }, { "epoch": 1.36, "learning_rate": 1.8985715330429763e-05, "loss": 0.0915, "step": 7428 }, { "epoch": 1.36, "learning_rate": 1.8985389507270247e-05, "loss": 0.0967, "step": 7429 }, { "epoch": 1.36, "learning_rate": 1.8985063634583058e-05, "loss": 0.1073, "step": 7430 }, { "epoch": 1.36, "learning_rate": 1.898473771236999e-05, "loss": 0.0983, "step": 7431 }, { "epoch": 1.36, "learning_rate": 1.898441174063283e-05, "loss": 0.105, "step": 7432 }, { "epoch": 1.36, "learning_rate": 1.898408571937339e-05, "loss": 0.1024, "step": 7433 }, { "epoch": 1.36, "learning_rate": 1.8983759648593457e-05, "loss": 0.1027, "step": 7434 }, { "epoch": 1.36, "learning_rate": 1.8983433528294824e-05, "loss": 0.1076, "step": 7435 }, { "epoch": 1.36, "learning_rate": 1.8983107358479302e-05, "loss": 0.0959, "step": 7436 }, { "epoch": 1.36, "learning_rate": 1.8982781139148677e-05, "loss": 0.1014, "step": 7437 }, { "epoch": 1.36, "learning_rate": 1.8982454870304758e-05, "loss": 0.098, "step": 7438 }, { "epoch": 1.36, "learning_rate": 1.8982128551949334e-05, "loss": 0.111, "step": 7439 }, { "epoch": 1.36, "learning_rate": 1.8981802184084207e-05, "loss": 0.094, "step": 7440 }, { "epoch": 1.36, "learning_rate": 1.8981475766711174e-05, "loss": 0.0946, "step": 7441 }, { "epoch": 1.36, "learning_rate": 1.898114929983204e-05, "loss": 0.0904, "step": 7442 }, { "epoch": 1.36, "learning_rate": 1.89808227834486e-05, "loss": 0.1048, "step": 7443 }, { "epoch": 1.37, "learning_rate": 1.898049621756265e-05, "loss": 0.1033, "step": 7444 }, { "epoch": 1.37, "learning_rate": 1.8980169602175997e-05, "loss": 0.1061, "step": 7445 }, { "epoch": 1.37, "learning_rate": 1.897984293729044e-05, "loss": 0.0965, "step": 7446 }, { "epoch": 1.37, "learning_rate": 1.8979516222907776e-05, "loss": 0.1091, "step": 7447 }, { "epoch": 1.37, "learning_rate": 1.8979189459029813e-05, "loss": 0.1003, "step": 7448 }, { "epoch": 1.37, "learning_rate": 1.8978862645658343e-05, "loss": 0.1018, "step": 7449 }, { "epoch": 1.37, "learning_rate": 1.8978535782795172e-05, "loss": 0.1034, "step": 7450 }, { "epoch": 1.37, "learning_rate": 1.89782088704421e-05, "loss": 0.1023, "step": 7451 }, { "epoch": 1.37, "learning_rate": 1.8977881908600934e-05, "loss": 0.1028, "step": 7452 }, { "epoch": 1.37, "learning_rate": 1.8977554897273473e-05, "loss": 0.1073, "step": 7453 }, { "epoch": 1.37, "learning_rate": 1.8977227836461514e-05, "loss": 0.1114, "step": 7454 }, { "epoch": 1.37, "learning_rate": 1.897690072616687e-05, "loss": 0.1104, "step": 7455 }, { "epoch": 1.37, "learning_rate": 1.8976573566391335e-05, "loss": 0.1017, "step": 7456 }, { "epoch": 1.37, "learning_rate": 1.897624635713672e-05, "loss": 0.0938, "step": 7457 }, { "epoch": 1.37, "learning_rate": 1.8975919098404826e-05, "loss": 0.0918, "step": 7458 }, { "epoch": 1.37, "learning_rate": 1.897559179019745e-05, "loss": 0.0917, "step": 7459 }, { "epoch": 1.37, "learning_rate": 1.8975264432516405e-05, "loss": 0.1053, "step": 7460 }, { "epoch": 1.37, "learning_rate": 1.897493702536349e-05, "loss": 0.0954, "step": 7461 }, { "epoch": 1.37, "learning_rate": 1.8974609568740514e-05, "loss": 0.1101, "step": 7462 }, { "epoch": 1.37, "learning_rate": 1.897428206264928e-05, "loss": 0.0891, "step": 7463 }, { "epoch": 1.37, "learning_rate": 1.8973954507091594e-05, "loss": 0.1063, "step": 7464 }, { "epoch": 1.37, "learning_rate": 1.8973626902069257e-05, "loss": 0.1008, "step": 7465 }, { "epoch": 1.37, "learning_rate": 1.8973299247584082e-05, "loss": 0.1009, "step": 7466 }, { "epoch": 1.37, "learning_rate": 1.897297154363787e-05, "loss": 0.1097, "step": 7467 }, { "epoch": 1.37, "learning_rate": 1.8972643790232425e-05, "loss": 0.0924, "step": 7468 }, { "epoch": 1.37, "learning_rate": 1.897231598736956e-05, "loss": 0.1038, "step": 7469 }, { "epoch": 1.37, "learning_rate": 1.8971988135051083e-05, "loss": 0.1064, "step": 7470 }, { "epoch": 1.37, "learning_rate": 1.8971660233278793e-05, "loss": 0.1162, "step": 7471 }, { "epoch": 1.37, "learning_rate": 1.89713322820545e-05, "loss": 0.0959, "step": 7472 }, { "epoch": 1.37, "learning_rate": 1.8971004281380018e-05, "loss": 0.0945, "step": 7473 }, { "epoch": 1.37, "learning_rate": 1.8970676231257144e-05, "loss": 0.1132, "step": 7474 }, { "epoch": 1.37, "learning_rate": 1.89703481316877e-05, "loss": 0.1092, "step": 7475 }, { "epoch": 1.37, "learning_rate": 1.897001998267348e-05, "loss": 0.096, "step": 7476 }, { "epoch": 1.37, "learning_rate": 1.8969691784216304e-05, "loss": 0.0955, "step": 7477 }, { "epoch": 1.37, "learning_rate": 1.8969363536317976e-05, "loss": 0.1038, "step": 7478 }, { "epoch": 1.37, "learning_rate": 1.8969035238980304e-05, "loss": 0.1027, "step": 7479 }, { "epoch": 1.37, "learning_rate": 1.8968706892205106e-05, "loss": 0.0825, "step": 7480 }, { "epoch": 1.37, "learning_rate": 1.896837849599418e-05, "loss": 0.1058, "step": 7481 }, { "epoch": 1.37, "learning_rate": 1.896805005034934e-05, "loss": 0.095, "step": 7482 }, { "epoch": 1.37, "learning_rate": 1.89677215552724e-05, "loss": 0.0878, "step": 7483 }, { "epoch": 1.37, "learning_rate": 1.896739301076517e-05, "loss": 0.1076, "step": 7484 }, { "epoch": 1.37, "learning_rate": 1.896706441682946e-05, "loss": 0.1029, "step": 7485 }, { "epoch": 1.37, "learning_rate": 1.8966735773467077e-05, "loss": 0.0939, "step": 7486 }, { "epoch": 1.37, "learning_rate": 1.8966407080679842e-05, "loss": 0.1002, "step": 7487 }, { "epoch": 1.37, "learning_rate": 1.8966078338469556e-05, "loss": 0.0948, "step": 7488 }, { "epoch": 1.37, "learning_rate": 1.8965749546838038e-05, "loss": 0.1058, "step": 7489 }, { "epoch": 1.37, "learning_rate": 1.8965420705787097e-05, "loss": 0.0928, "step": 7490 }, { "epoch": 1.37, "learning_rate": 1.896509181531855e-05, "loss": 0.1011, "step": 7491 }, { "epoch": 1.37, "learning_rate": 1.8964762875434207e-05, "loss": 0.1085, "step": 7492 }, { "epoch": 1.37, "learning_rate": 1.896443388613588e-05, "loss": 0.0934, "step": 7493 }, { "epoch": 1.37, "learning_rate": 1.8964104847425382e-05, "loss": 0.0982, "step": 7494 }, { "epoch": 1.37, "learning_rate": 1.896377575930453e-05, "loss": 0.1027, "step": 7495 }, { "epoch": 1.37, "learning_rate": 1.8963446621775134e-05, "loss": 0.1095, "step": 7496 }, { "epoch": 1.37, "learning_rate": 1.896311743483901e-05, "loss": 0.0913, "step": 7497 }, { "epoch": 1.38, "learning_rate": 1.8962788198497978e-05, "loss": 0.1157, "step": 7498 }, { "epoch": 1.38, "learning_rate": 1.8962458912753843e-05, "loss": 0.1064, "step": 7499 }, { "epoch": 1.38, "learning_rate": 1.8962129577608425e-05, "loss": 0.0879, "step": 7500 }, { "epoch": 1.38, "learning_rate": 1.8961800193063537e-05, "loss": 0.1029, "step": 7501 }, { "epoch": 1.38, "learning_rate": 1.8961470759121e-05, "loss": 0.0978, "step": 7502 }, { "epoch": 1.38, "learning_rate": 1.8961141275782624e-05, "loss": 0.1069, "step": 7503 }, { "epoch": 1.38, "learning_rate": 1.8960811743050227e-05, "loss": 0.0957, "step": 7504 }, { "epoch": 1.38, "learning_rate": 1.8960482160925625e-05, "loss": 0.0918, "step": 7505 }, { "epoch": 1.38, "learning_rate": 1.896015252941064e-05, "loss": 0.0974, "step": 7506 }, { "epoch": 1.38, "learning_rate": 1.895982284850708e-05, "loss": 0.1044, "step": 7507 }, { "epoch": 1.38, "learning_rate": 1.895949311821677e-05, "loss": 0.0965, "step": 7508 }, { "epoch": 1.38, "learning_rate": 1.895916333854152e-05, "loss": 0.0932, "step": 7509 }, { "epoch": 1.38, "learning_rate": 1.8958833509483154e-05, "loss": 0.1049, "step": 7510 }, { "epoch": 1.38, "learning_rate": 1.895850363104349e-05, "loss": 0.0986, "step": 7511 }, { "epoch": 1.38, "learning_rate": 1.8958173703224338e-05, "loss": 0.1015, "step": 7512 }, { "epoch": 1.38, "learning_rate": 1.8957843726027526e-05, "loss": 0.0947, "step": 7513 }, { "epoch": 1.38, "learning_rate": 1.895751369945487e-05, "loss": 0.0959, "step": 7514 }, { "epoch": 1.38, "learning_rate": 1.895718362350819e-05, "loss": 0.1005, "step": 7515 }, { "epoch": 1.38, "learning_rate": 1.8956853498189302e-05, "loss": 0.0948, "step": 7516 }, { "epoch": 1.38, "learning_rate": 1.895652332350003e-05, "loss": 0.0962, "step": 7517 }, { "epoch": 1.38, "learning_rate": 1.8956193099442187e-05, "loss": 0.0986, "step": 7518 }, { "epoch": 1.38, "learning_rate": 1.89558628260176e-05, "loss": 0.1033, "step": 7519 }, { "epoch": 1.38, "learning_rate": 1.895553250322809e-05, "loss": 0.1036, "step": 7520 }, { "epoch": 1.38, "learning_rate": 1.8955202131075472e-05, "loss": 0.092, "step": 7521 }, { "epoch": 1.38, "learning_rate": 1.895487170956157e-05, "loss": 0.1063, "step": 7522 }, { "epoch": 1.38, "learning_rate": 1.8954541238688208e-05, "loss": 0.0875, "step": 7523 }, { "epoch": 1.38, "learning_rate": 1.8954210718457205e-05, "loss": 0.0923, "step": 7524 }, { "epoch": 1.38, "learning_rate": 1.8953880148870377e-05, "loss": 0.0973, "step": 7525 }, { "epoch": 1.38, "learning_rate": 1.895354952992956e-05, "loss": 0.0838, "step": 7526 }, { "epoch": 1.38, "learning_rate": 1.8953218861636564e-05, "loss": 0.0997, "step": 7527 }, { "epoch": 1.38, "learning_rate": 1.8952888143993212e-05, "loss": 0.1061, "step": 7528 }, { "epoch": 1.38, "learning_rate": 1.895255737700134e-05, "loss": 0.101, "step": 7529 }, { "epoch": 1.38, "learning_rate": 1.8952226560662752e-05, "loss": 0.0798, "step": 7530 }, { "epoch": 1.38, "learning_rate": 1.8951895694979288e-05, "loss": 0.1121, "step": 7531 }, { "epoch": 1.38, "learning_rate": 1.8951564779952764e-05, "loss": 0.1046, "step": 7532 }, { "epoch": 1.38, "learning_rate": 1.8951233815585005e-05, "loss": 0.1, "step": 7533 }, { "epoch": 1.38, "learning_rate": 1.895090280187783e-05, "loss": 0.0984, "step": 7534 }, { "epoch": 1.38, "learning_rate": 1.895057173883308e-05, "loss": 0.0977, "step": 7535 }, { "epoch": 1.38, "learning_rate": 1.895024062645256e-05, "loss": 0.1046, "step": 7536 }, { "epoch": 1.38, "learning_rate": 1.894990946473811e-05, "loss": 0.1058, "step": 7537 }, { "epoch": 1.38, "learning_rate": 1.8949578253691544e-05, "loss": 0.0913, "step": 7538 }, { "epoch": 1.38, "learning_rate": 1.8949246993314694e-05, "loss": 0.0927, "step": 7539 }, { "epoch": 1.38, "learning_rate": 1.8948915683609387e-05, "loss": 0.1002, "step": 7540 }, { "epoch": 1.38, "learning_rate": 1.8948584324577445e-05, "loss": 0.097, "step": 7541 }, { "epoch": 1.38, "learning_rate": 1.89482529162207e-05, "loss": 0.097, "step": 7542 }, { "epoch": 1.38, "learning_rate": 1.8947921458540974e-05, "loss": 0.099, "step": 7543 }, { "epoch": 1.38, "learning_rate": 1.8947589951540094e-05, "loss": 0.0952, "step": 7544 }, { "epoch": 1.38, "learning_rate": 1.894725839521989e-05, "loss": 0.1019, "step": 7545 }, { "epoch": 1.38, "learning_rate": 1.8946926789582186e-05, "loss": 0.1011, "step": 7546 }, { "epoch": 1.38, "learning_rate": 1.8946595134628815e-05, "loss": 0.11, "step": 7547 }, { "epoch": 1.38, "learning_rate": 1.8946263430361598e-05, "loss": 0.0998, "step": 7548 }, { "epoch": 1.38, "learning_rate": 1.8945931676782373e-05, "loss": 0.0983, "step": 7549 }, { "epoch": 1.38, "learning_rate": 1.8945599873892956e-05, "loss": 0.096, "step": 7550 }, { "epoch": 1.38, "learning_rate": 1.894526802169519e-05, "loss": 0.0859, "step": 7551 }, { "epoch": 1.38, "learning_rate": 1.8944936120190893e-05, "loss": 0.1061, "step": 7552 }, { "epoch": 1.39, "learning_rate": 1.8944604169381896e-05, "loss": 0.1031, "step": 7553 }, { "epoch": 1.39, "learning_rate": 1.8944272169270035e-05, "loss": 0.0961, "step": 7554 }, { "epoch": 1.39, "learning_rate": 1.8943940119857137e-05, "loss": 0.0975, "step": 7555 }, { "epoch": 1.39, "learning_rate": 1.894360802114503e-05, "loss": 0.1056, "step": 7556 }, { "epoch": 1.39, "learning_rate": 1.8943275873135545e-05, "loss": 0.093, "step": 7557 }, { "epoch": 1.39, "learning_rate": 1.8942943675830515e-05, "loss": 0.0828, "step": 7558 }, { "epoch": 1.39, "learning_rate": 1.894261142923177e-05, "loss": 0.1067, "step": 7559 }, { "epoch": 1.39, "learning_rate": 1.8942279133341142e-05, "loss": 0.0925, "step": 7560 }, { "epoch": 1.39, "learning_rate": 1.8941946788160463e-05, "loss": 0.0942, "step": 7561 }, { "epoch": 1.39, "learning_rate": 1.894161439369156e-05, "loss": 0.1009, "step": 7562 }, { "epoch": 1.39, "learning_rate": 1.8941281949936272e-05, "loss": 0.0864, "step": 7563 }, { "epoch": 1.39, "learning_rate": 1.8940949456896427e-05, "loss": 0.1041, "step": 7564 }, { "epoch": 1.39, "learning_rate": 1.894061691457386e-05, "loss": 0.0958, "step": 7565 }, { "epoch": 1.39, "learning_rate": 1.8940284322970405e-05, "loss": 0.1045, "step": 7566 }, { "epoch": 1.39, "learning_rate": 1.893995168208789e-05, "loss": 0.1028, "step": 7567 }, { "epoch": 1.39, "learning_rate": 1.893961899192815e-05, "loss": 0.0993, "step": 7568 }, { "epoch": 1.39, "learning_rate": 1.8939286252493025e-05, "loss": 0.0985, "step": 7569 }, { "epoch": 1.39, "learning_rate": 1.8938953463784346e-05, "loss": 0.0952, "step": 7570 }, { "epoch": 1.39, "learning_rate": 1.8938620625803944e-05, "loss": 0.0946, "step": 7571 }, { "epoch": 1.39, "learning_rate": 1.8938287738553654e-05, "loss": 0.1009, "step": 7572 }, { "epoch": 1.39, "learning_rate": 1.8937954802035313e-05, "loss": 0.1038, "step": 7573 }, { "epoch": 1.39, "learning_rate": 1.8937621816250757e-05, "loss": 0.0892, "step": 7574 }, { "epoch": 1.39, "learning_rate": 1.893728878120182e-05, "loss": 0.1115, "step": 7575 }, { "epoch": 1.39, "learning_rate": 1.893695569689034e-05, "loss": 0.1066, "step": 7576 }, { "epoch": 1.39, "learning_rate": 1.8936622563318145e-05, "loss": 0.1129, "step": 7577 }, { "epoch": 1.39, "learning_rate": 1.893628938048708e-05, "loss": 0.1059, "step": 7578 }, { "epoch": 1.39, "learning_rate": 1.8935956148398982e-05, "loss": 0.1004, "step": 7579 }, { "epoch": 1.39, "learning_rate": 1.893562286705568e-05, "loss": 0.0964, "step": 7580 }, { "epoch": 1.39, "learning_rate": 1.8935289536459018e-05, "loss": 0.0985, "step": 7581 }, { "epoch": 1.39, "learning_rate": 1.893495615661083e-05, "loss": 0.088, "step": 7582 }, { "epoch": 1.39, "learning_rate": 1.8934622727512953e-05, "loss": 0.1007, "step": 7583 }, { "epoch": 1.39, "learning_rate": 1.8934289249167226e-05, "loss": 0.0956, "step": 7584 }, { "epoch": 1.39, "learning_rate": 1.893395572157549e-05, "loss": 0.1012, "step": 7585 }, { "epoch": 1.39, "learning_rate": 1.893362214473958e-05, "loss": 0.0991, "step": 7586 }, { "epoch": 1.39, "learning_rate": 1.893328851866133e-05, "loss": 0.0979, "step": 7587 }, { "epoch": 1.39, "learning_rate": 1.893295484334259e-05, "loss": 0.0872, "step": 7588 }, { "epoch": 1.39, "learning_rate": 1.8932621118785197e-05, "loss": 0.0895, "step": 7589 }, { "epoch": 1.39, "learning_rate": 1.893228734499098e-05, "loss": 0.0945, "step": 7590 }, { "epoch": 1.39, "learning_rate": 1.893195352196179e-05, "loss": 0.0857, "step": 7591 }, { "epoch": 1.39, "learning_rate": 1.8931619649699463e-05, "loss": 0.0944, "step": 7592 }, { "epoch": 1.39, "learning_rate": 1.8931285728205834e-05, "loss": 0.0922, "step": 7593 }, { "epoch": 1.39, "learning_rate": 1.8930951757482755e-05, "loss": 0.0965, "step": 7594 }, { "epoch": 1.39, "learning_rate": 1.893061773753206e-05, "loss": 0.0937, "step": 7595 }, { "epoch": 1.39, "learning_rate": 1.8930283668355586e-05, "loss": 0.0993, "step": 7596 }, { "epoch": 1.39, "learning_rate": 1.8929949549955183e-05, "loss": 0.1011, "step": 7597 }, { "epoch": 1.39, "learning_rate": 1.892961538233269e-05, "loss": 0.0889, "step": 7598 }, { "epoch": 1.39, "learning_rate": 1.8929281165489946e-05, "loss": 0.0919, "step": 7599 }, { "epoch": 1.39, "learning_rate": 1.8928946899428796e-05, "loss": 0.0936, "step": 7600 }, { "epoch": 1.39, "learning_rate": 1.892861258415108e-05, "loss": 0.1059, "step": 7601 }, { "epoch": 1.39, "learning_rate": 1.892827821965864e-05, "loss": 0.1022, "step": 7602 }, { "epoch": 1.39, "learning_rate": 1.8927943805953326e-05, "loss": 0.0882, "step": 7603 }, { "epoch": 1.39, "learning_rate": 1.8927609343036976e-05, "loss": 0.0942, "step": 7604 }, { "epoch": 1.39, "learning_rate": 1.8927274830911433e-05, "loss": 0.095, "step": 7605 }, { "epoch": 1.39, "learning_rate": 1.892694026957854e-05, "loss": 0.1003, "step": 7606 }, { "epoch": 1.4, "learning_rate": 1.8926605659040147e-05, "loss": 0.0934, "step": 7607 }, { "epoch": 1.4, "learning_rate": 1.892627099929809e-05, "loss": 0.1069, "step": 7608 }, { "epoch": 1.4, "learning_rate": 1.8925936290354224e-05, "loss": 0.0969, "step": 7609 }, { "epoch": 1.4, "learning_rate": 1.8925601532210384e-05, "loss": 0.0924, "step": 7610 }, { "epoch": 1.4, "learning_rate": 1.892526672486842e-05, "loss": 0.0969, "step": 7611 }, { "epoch": 1.4, "learning_rate": 1.8924931868330175e-05, "loss": 0.0821, "step": 7612 }, { "epoch": 1.4, "learning_rate": 1.89245969625975e-05, "loss": 0.1184, "step": 7613 }, { "epoch": 1.4, "learning_rate": 1.8924262007672235e-05, "loss": 0.101, "step": 7614 }, { "epoch": 1.4, "learning_rate": 1.892392700355623e-05, "loss": 0.0899, "step": 7615 }, { "epoch": 1.4, "learning_rate": 1.892359195025133e-05, "loss": 0.1013, "step": 7616 }, { "epoch": 1.4, "learning_rate": 1.8923256847759378e-05, "loss": 0.0915, "step": 7617 }, { "epoch": 1.4, "learning_rate": 1.892292169608223e-05, "loss": 0.1028, "step": 7618 }, { "epoch": 1.4, "learning_rate": 1.8922586495221732e-05, "loss": 0.1079, "step": 7619 }, { "epoch": 1.4, "learning_rate": 1.892225124517972e-05, "loss": 0.0913, "step": 7620 }, { "epoch": 1.4, "learning_rate": 1.8921915945958058e-05, "loss": 0.1052, "step": 7621 }, { "epoch": 1.4, "learning_rate": 1.892158059755858e-05, "loss": 0.0986, "step": 7622 }, { "epoch": 1.4, "learning_rate": 1.892124519998314e-05, "loss": 0.1016, "step": 7623 }, { "epoch": 1.4, "learning_rate": 1.892090975323359e-05, "loss": 0.1085, "step": 7624 }, { "epoch": 1.4, "learning_rate": 1.892057425731178e-05, "loss": 0.1111, "step": 7625 }, { "epoch": 1.4, "learning_rate": 1.8920238712219552e-05, "loss": 0.102, "step": 7626 }, { "epoch": 1.4, "learning_rate": 1.8919903117958757e-05, "loss": 0.0982, "step": 7627 }, { "epoch": 1.4, "learning_rate": 1.8919567474531253e-05, "loss": 0.0982, "step": 7628 }, { "epoch": 1.4, "learning_rate": 1.891923178193888e-05, "loss": 0.1017, "step": 7629 }, { "epoch": 1.4, "learning_rate": 1.8918896040183492e-05, "loss": 0.0948, "step": 7630 }, { "epoch": 1.4, "learning_rate": 1.8918560249266942e-05, "loss": 0.0933, "step": 7631 }, { "epoch": 1.4, "learning_rate": 1.891822440919108e-05, "loss": 0.0971, "step": 7632 }, { "epoch": 1.4, "learning_rate": 1.8917888519957756e-05, "loss": 0.102, "step": 7633 }, { "epoch": 1.4, "learning_rate": 1.8917552581568815e-05, "loss": 0.097, "step": 7634 }, { "epoch": 1.4, "learning_rate": 1.8917216594026123e-05, "loss": 0.0962, "step": 7635 }, { "epoch": 1.4, "learning_rate": 1.891688055733152e-05, "loss": 0.0885, "step": 7636 }, { "epoch": 1.4, "learning_rate": 1.8916544471486868e-05, "loss": 0.1077, "step": 7637 }, { "epoch": 1.4, "learning_rate": 1.8916208336494005e-05, "loss": 0.0898, "step": 7638 }, { "epoch": 1.4, "learning_rate": 1.89158721523548e-05, "loss": 0.1032, "step": 7639 }, { "epoch": 1.4, "learning_rate": 1.8915535919071098e-05, "loss": 0.0985, "step": 7640 }, { "epoch": 1.4, "learning_rate": 1.891519963664475e-05, "loss": 0.1, "step": 7641 }, { "epoch": 1.4, "learning_rate": 1.8914863305077615e-05, "loss": 0.1004, "step": 7642 }, { "epoch": 1.4, "learning_rate": 1.8914526924371542e-05, "loss": 0.1006, "step": 7643 }, { "epoch": 1.4, "learning_rate": 1.8914190494528394e-05, "loss": 0.1007, "step": 7644 }, { "epoch": 1.4, "learning_rate": 1.891385401555001e-05, "loss": 0.0998, "step": 7645 }, { "epoch": 1.4, "learning_rate": 1.891351748743826e-05, "loss": 0.1014, "step": 7646 }, { "epoch": 1.4, "learning_rate": 1.8913180910194994e-05, "loss": 0.0968, "step": 7647 }, { "epoch": 1.4, "learning_rate": 1.8912844283822062e-05, "loss": 0.1005, "step": 7648 }, { "epoch": 1.4, "learning_rate": 1.8912507608321324e-05, "loss": 0.095, "step": 7649 }, { "epoch": 1.4, "learning_rate": 1.8912170883694637e-05, "loss": 0.1016, "step": 7650 }, { "epoch": 1.4, "learning_rate": 1.8911834109943857e-05, "loss": 0.1006, "step": 7651 }, { "epoch": 1.4, "learning_rate": 1.8911497287070835e-05, "loss": 0.092, "step": 7652 }, { "epoch": 1.4, "learning_rate": 1.8911160415077434e-05, "loss": 0.1032, "step": 7653 }, { "epoch": 1.4, "learning_rate": 1.8910823493965503e-05, "loss": 0.0969, "step": 7654 }, { "epoch": 1.4, "learning_rate": 1.8910486523736907e-05, "loss": 0.1022, "step": 7655 }, { "epoch": 1.4, "learning_rate": 1.89101495043935e-05, "loss": 0.0956, "step": 7656 }, { "epoch": 1.4, "learning_rate": 1.8909812435937144e-05, "loss": 0.1084, "step": 7657 }, { "epoch": 1.4, "learning_rate": 1.890947531836969e-05, "loss": 0.0868, "step": 7658 }, { "epoch": 1.4, "learning_rate": 1.8909138151693e-05, "loss": 0.109, "step": 7659 }, { "epoch": 1.4, "learning_rate": 1.890880093590893e-05, "loss": 0.1139, "step": 7660 }, { "epoch": 1.4, "learning_rate": 1.8908463671019344e-05, "loss": 0.1023, "step": 7661 }, { "epoch": 1.41, "learning_rate": 1.89081263570261e-05, "loss": 0.1133, "step": 7662 }, { "epoch": 1.41, "learning_rate": 1.8907788993931048e-05, "loss": 0.0992, "step": 7663 }, { "epoch": 1.41, "learning_rate": 1.8907451581736052e-05, "loss": 0.095, "step": 7664 }, { "epoch": 1.41, "learning_rate": 1.890711412044298e-05, "loss": 0.1009, "step": 7665 }, { "epoch": 1.41, "learning_rate": 1.890677661005369e-05, "loss": 0.0992, "step": 7666 }, { "epoch": 1.41, "learning_rate": 1.890643905057003e-05, "loss": 0.0959, "step": 7667 }, { "epoch": 1.41, "learning_rate": 1.8906101441993875e-05, "loss": 0.1024, "step": 7668 }, { "epoch": 1.41, "learning_rate": 1.8905763784327078e-05, "loss": 0.1005, "step": 7669 }, { "epoch": 1.41, "learning_rate": 1.8905426077571502e-05, "loss": 0.1034, "step": 7670 }, { "epoch": 1.41, "learning_rate": 1.890508832172901e-05, "loss": 0.0948, "step": 7671 }, { "epoch": 1.41, "learning_rate": 1.8904750516801465e-05, "loss": 0.0881, "step": 7672 }, { "epoch": 1.41, "learning_rate": 1.8904412662790722e-05, "loss": 0.1033, "step": 7673 }, { "epoch": 1.41, "learning_rate": 1.890407475969865e-05, "loss": 0.095, "step": 7674 }, { "epoch": 1.41, "learning_rate": 1.890373680752711e-05, "loss": 0.0892, "step": 7675 }, { "epoch": 1.41, "learning_rate": 1.890339880627796e-05, "loss": 0.1126, "step": 7676 }, { "epoch": 1.41, "learning_rate": 1.8903060755953073e-05, "loss": 0.109, "step": 7677 }, { "epoch": 1.41, "learning_rate": 1.89027226565543e-05, "loss": 0.0885, "step": 7678 }, { "epoch": 1.41, "learning_rate": 1.8902384508083518e-05, "loss": 0.1021, "step": 7679 }, { "epoch": 1.41, "learning_rate": 1.8902046310542582e-05, "loss": 0.0939, "step": 7680 }, { "epoch": 1.41, "learning_rate": 1.8901708063933357e-05, "loss": 0.1128, "step": 7681 }, { "epoch": 1.41, "learning_rate": 1.8901369768257708e-05, "loss": 0.0994, "step": 7682 }, { "epoch": 1.41, "learning_rate": 1.89010314235175e-05, "loss": 0.0967, "step": 7683 }, { "epoch": 1.41, "learning_rate": 1.89006930297146e-05, "loss": 0.0943, "step": 7684 }, { "epoch": 1.41, "learning_rate": 1.890035458685087e-05, "loss": 0.0969, "step": 7685 }, { "epoch": 1.41, "learning_rate": 1.890001609492818e-05, "loss": 0.0978, "step": 7686 }, { "epoch": 1.41, "learning_rate": 1.889967755394839e-05, "loss": 0.0853, "step": 7687 }, { "epoch": 1.41, "learning_rate": 1.8899338963913368e-05, "loss": 0.0903, "step": 7688 }, { "epoch": 1.41, "learning_rate": 1.8899000324824982e-05, "loss": 0.1123, "step": 7689 }, { "epoch": 1.41, "learning_rate": 1.8898661636685097e-05, "loss": 0.0857, "step": 7690 }, { "epoch": 1.41, "learning_rate": 1.8898322899495585e-05, "loss": 0.101, "step": 7691 }, { "epoch": 1.41, "learning_rate": 1.8897984113258304e-05, "loss": 0.106, "step": 7692 }, { "epoch": 1.41, "learning_rate": 1.889764527797513e-05, "loss": 0.0938, "step": 7693 }, { "epoch": 1.41, "learning_rate": 1.8897306393647923e-05, "loss": 0.096, "step": 7694 }, { "epoch": 1.41, "learning_rate": 1.8896967460278556e-05, "loss": 0.096, "step": 7695 }, { "epoch": 1.41, "learning_rate": 1.8896628477868895e-05, "loss": 0.1026, "step": 7696 }, { "epoch": 1.41, "learning_rate": 1.889628944642081e-05, "loss": 0.0879, "step": 7697 }, { "epoch": 1.41, "learning_rate": 1.889595036593617e-05, "loss": 0.1053, "step": 7698 }, { "epoch": 1.41, "learning_rate": 1.8895611236416842e-05, "loss": 0.0993, "step": 7699 }, { "epoch": 1.41, "learning_rate": 1.88952720578647e-05, "loss": 0.1, "step": 7700 }, { "epoch": 1.41, "learning_rate": 1.8894932830281605e-05, "loss": 0.1105, "step": 7701 }, { "epoch": 1.41, "learning_rate": 1.8894593553669437e-05, "loss": 0.0996, "step": 7702 }, { "epoch": 1.41, "learning_rate": 1.889425422803006e-05, "loss": 0.1006, "step": 7703 }, { "epoch": 1.41, "learning_rate": 1.889391485336534e-05, "loss": 0.0957, "step": 7704 }, { "epoch": 1.41, "learning_rate": 1.8893575429677157e-05, "loss": 0.0964, "step": 7705 }, { "epoch": 1.41, "learning_rate": 1.8893235956967375e-05, "loss": 0.1017, "step": 7706 }, { "epoch": 1.41, "learning_rate": 1.889289643523787e-05, "loss": 0.0944, "step": 7707 }, { "epoch": 1.41, "learning_rate": 1.8892556864490515e-05, "loss": 0.0971, "step": 7708 }, { "epoch": 1.41, "learning_rate": 1.8892217244727177e-05, "loss": 0.0975, "step": 7709 }, { "epoch": 1.41, "learning_rate": 1.8891877575949724e-05, "loss": 0.1001, "step": 7710 }, { "epoch": 1.41, "learning_rate": 1.8891537858160038e-05, "loss": 0.1176, "step": 7711 }, { "epoch": 1.41, "learning_rate": 1.8891198091359982e-05, "loss": 0.1013, "step": 7712 }, { "epoch": 1.41, "learning_rate": 1.889085827555144e-05, "loss": 0.0971, "step": 7713 }, { "epoch": 1.41, "learning_rate": 1.8890518410736275e-05, "loss": 0.093, "step": 7714 }, { "epoch": 1.41, "learning_rate": 1.889017849691637e-05, "loss": 0.0976, "step": 7715 }, { "epoch": 1.42, "learning_rate": 1.8889838534093586e-05, "loss": 0.0974, "step": 7716 }, { "epoch": 1.42, "learning_rate": 1.8889498522269806e-05, "loss": 0.0988, "step": 7717 }, { "epoch": 1.42, "learning_rate": 1.8889158461446904e-05, "loss": 0.0989, "step": 7718 }, { "epoch": 1.42, "learning_rate": 1.8888818351626748e-05, "loss": 0.0892, "step": 7719 }, { "epoch": 1.42, "learning_rate": 1.888847819281122e-05, "loss": 0.1053, "step": 7720 }, { "epoch": 1.42, "learning_rate": 1.8888137985002194e-05, "loss": 0.0944, "step": 7721 }, { "epoch": 1.42, "learning_rate": 1.8887797728201537e-05, "loss": 0.1109, "step": 7722 }, { "epoch": 1.42, "learning_rate": 1.8887457422411137e-05, "loss": 0.1095, "step": 7723 }, { "epoch": 1.42, "learning_rate": 1.888711706763286e-05, "loss": 0.0948, "step": 7724 }, { "epoch": 1.42, "learning_rate": 1.8886776663868586e-05, "loss": 0.1112, "step": 7725 }, { "epoch": 1.42, "learning_rate": 1.8886436211120195e-05, "loss": 0.0869, "step": 7726 }, { "epoch": 1.42, "learning_rate": 1.8886095709389553e-05, "loss": 0.1, "step": 7727 }, { "epoch": 1.42, "learning_rate": 1.8885755158678546e-05, "loss": 0.1012, "step": 7728 }, { "epoch": 1.42, "learning_rate": 1.8885414558989052e-05, "loss": 0.103, "step": 7729 }, { "epoch": 1.42, "learning_rate": 1.8885073910322937e-05, "loss": 0.1048, "step": 7730 }, { "epoch": 1.42, "learning_rate": 1.888473321268209e-05, "loss": 0.1101, "step": 7731 }, { "epoch": 1.42, "learning_rate": 1.8884392466068386e-05, "loss": 0.0978, "step": 7732 }, { "epoch": 1.42, "learning_rate": 1.8884051670483705e-05, "loss": 0.096, "step": 7733 }, { "epoch": 1.42, "learning_rate": 1.888371082592992e-05, "loss": 0.1069, "step": 7734 }, { "epoch": 1.42, "learning_rate": 1.8883369932408913e-05, "loss": 0.1078, "step": 7735 }, { "epoch": 1.42, "learning_rate": 1.8883028989922562e-05, "loss": 0.099, "step": 7736 }, { "epoch": 1.42, "learning_rate": 1.8882687998472747e-05, "loss": 0.0953, "step": 7737 }, { "epoch": 1.42, "learning_rate": 1.888234695806135e-05, "loss": 0.0945, "step": 7738 }, { "epoch": 1.42, "learning_rate": 1.888200586869025e-05, "loss": 0.119, "step": 7739 }, { "epoch": 1.42, "learning_rate": 1.8881664730361322e-05, "loss": 0.1202, "step": 7740 }, { "epoch": 1.42, "learning_rate": 1.888132354307645e-05, "loss": 0.0916, "step": 7741 }, { "epoch": 1.42, "learning_rate": 1.8880982306837518e-05, "loss": 0.0939, "step": 7742 }, { "epoch": 1.42, "learning_rate": 1.88806410216464e-05, "loss": 0.0842, "step": 7743 }, { "epoch": 1.42, "learning_rate": 1.888029968750498e-05, "loss": 0.111, "step": 7744 }, { "epoch": 1.42, "learning_rate": 1.887995830441514e-05, "loss": 0.0972, "step": 7745 }, { "epoch": 1.42, "learning_rate": 1.8879616872378767e-05, "loss": 0.0998, "step": 7746 }, { "epoch": 1.42, "learning_rate": 1.8879275391397732e-05, "loss": 0.1083, "step": 7747 }, { "epoch": 1.42, "learning_rate": 1.8878933861473927e-05, "loss": 0.0971, "step": 7748 }, { "epoch": 1.42, "learning_rate": 1.887859228260923e-05, "loss": 0.0852, "step": 7749 }, { "epoch": 1.42, "learning_rate": 1.887825065480552e-05, "loss": 0.0932, "step": 7750 }, { "epoch": 1.42, "learning_rate": 1.8877908978064688e-05, "loss": 0.0975, "step": 7751 }, { "epoch": 1.42, "learning_rate": 1.8877567252388615e-05, "loss": 0.0942, "step": 7752 }, { "epoch": 1.42, "learning_rate": 1.887722547777918e-05, "loss": 0.0811, "step": 7753 }, { "epoch": 1.42, "learning_rate": 1.8876883654238273e-05, "loss": 0.0998, "step": 7754 }, { "epoch": 1.42, "learning_rate": 1.8876541781767775e-05, "loss": 0.1051, "step": 7755 }, { "epoch": 1.42, "learning_rate": 1.8876199860369567e-05, "loss": 0.0988, "step": 7756 }, { "epoch": 1.42, "learning_rate": 1.8875857890045544e-05, "loss": 0.0954, "step": 7757 }, { "epoch": 1.42, "learning_rate": 1.887551587079758e-05, "loss": 0.1064, "step": 7758 }, { "epoch": 1.42, "learning_rate": 1.8875173802627565e-05, "loss": 0.1053, "step": 7759 }, { "epoch": 1.42, "learning_rate": 1.8874831685537384e-05, "loss": 0.0993, "step": 7760 }, { "epoch": 1.42, "learning_rate": 1.8874489519528922e-05, "loss": 0.0946, "step": 7761 }, { "epoch": 1.42, "learning_rate": 1.8874147304604067e-05, "loss": 0.1012, "step": 7762 }, { "epoch": 1.42, "learning_rate": 1.8873805040764703e-05, "loss": 0.0997, "step": 7763 }, { "epoch": 1.42, "learning_rate": 1.887346272801272e-05, "loss": 0.101, "step": 7764 }, { "epoch": 1.42, "learning_rate": 1.8873120366350002e-05, "loss": 0.0985, "step": 7765 }, { "epoch": 1.42, "learning_rate": 1.8872777955778436e-05, "loss": 0.1093, "step": 7766 }, { "epoch": 1.42, "learning_rate": 1.887243549629991e-05, "loss": 0.0991, "step": 7767 }, { "epoch": 1.42, "learning_rate": 1.8872092987916312e-05, "loss": 0.0999, "step": 7768 }, { "epoch": 1.42, "learning_rate": 1.887175043062953e-05, "loss": 0.1072, "step": 7769 }, { "epoch": 1.42, "learning_rate": 1.8871407824441453e-05, "loss": 0.1027, "step": 7770 }, { "epoch": 1.43, "learning_rate": 1.8871065169353966e-05, "loss": 0.0982, "step": 7771 }, { "epoch": 1.43, "learning_rate": 1.8870722465368957e-05, "loss": 0.1043, "step": 7772 }, { "epoch": 1.43, "learning_rate": 1.8870379712488324e-05, "loss": 0.0839, "step": 7773 }, { "epoch": 1.43, "learning_rate": 1.8870036910713947e-05, "loss": 0.0961, "step": 7774 }, { "epoch": 1.43, "learning_rate": 1.8869694060047718e-05, "loss": 0.0887, "step": 7775 }, { "epoch": 1.43, "learning_rate": 1.886935116049153e-05, "loss": 0.0908, "step": 7776 }, { "epoch": 1.43, "learning_rate": 1.886900821204727e-05, "loss": 0.1088, "step": 7777 }, { "epoch": 1.43, "learning_rate": 1.8868665214716827e-05, "loss": 0.1039, "step": 7778 }, { "epoch": 1.43, "learning_rate": 1.8868322168502094e-05, "loss": 0.0893, "step": 7779 }, { "epoch": 1.43, "learning_rate": 1.8867979073404965e-05, "loss": 0.1059, "step": 7780 }, { "epoch": 1.43, "learning_rate": 1.8867635929427326e-05, "loss": 0.0838, "step": 7781 }, { "epoch": 1.43, "learning_rate": 1.886729273657107e-05, "loss": 0.0932, "step": 7782 }, { "epoch": 1.43, "learning_rate": 1.8866949494838086e-05, "loss": 0.0851, "step": 7783 }, { "epoch": 1.43, "learning_rate": 1.8866606204230272e-05, "loss": 0.0926, "step": 7784 }, { "epoch": 1.43, "learning_rate": 1.8866262864749514e-05, "loss": 0.1006, "step": 7785 }, { "epoch": 1.43, "learning_rate": 1.886591947639771e-05, "loss": 0.0892, "step": 7786 }, { "epoch": 1.43, "learning_rate": 1.8865576039176746e-05, "loss": 0.1026, "step": 7787 }, { "epoch": 1.43, "learning_rate": 1.8865232553088524e-05, "loss": 0.0916, "step": 7788 }, { "epoch": 1.43, "learning_rate": 1.8864889018134927e-05, "loss": 0.0897, "step": 7789 }, { "epoch": 1.43, "learning_rate": 1.886454543431786e-05, "loss": 0.1004, "step": 7790 }, { "epoch": 1.43, "learning_rate": 1.886420180163921e-05, "loss": 0.0922, "step": 7791 }, { "epoch": 1.43, "learning_rate": 1.8863858120100872e-05, "loss": 0.1078, "step": 7792 }, { "epoch": 1.43, "learning_rate": 1.886351438970474e-05, "loss": 0.0987, "step": 7793 }, { "epoch": 1.43, "learning_rate": 1.8863170610452702e-05, "loss": 0.1044, "step": 7794 }, { "epoch": 1.43, "learning_rate": 1.8862826782346667e-05, "loss": 0.1067, "step": 7795 }, { "epoch": 1.43, "learning_rate": 1.8862482905388525e-05, "loss": 0.0925, "step": 7796 }, { "epoch": 1.43, "learning_rate": 1.8862138979580164e-05, "loss": 0.0917, "step": 7797 }, { "epoch": 1.43, "learning_rate": 1.886179500492349e-05, "loss": 0.1006, "step": 7798 }, { "epoch": 1.43, "learning_rate": 1.886145098142039e-05, "loss": 0.1019, "step": 7799 }, { "epoch": 1.43, "learning_rate": 1.886110690907277e-05, "loss": 0.0999, "step": 7800 }, { "epoch": 1.43, "learning_rate": 1.8860762787882512e-05, "loss": 0.0973, "step": 7801 }, { "epoch": 1.43, "learning_rate": 1.8860418617851533e-05, "loss": 0.1032, "step": 7802 }, { "epoch": 1.43, "learning_rate": 1.8860074398981712e-05, "loss": 0.1029, "step": 7803 }, { "epoch": 1.43, "learning_rate": 1.8859730131274956e-05, "loss": 0.109, "step": 7804 }, { "epoch": 1.43, "learning_rate": 1.885938581473316e-05, "loss": 0.1148, "step": 7805 }, { "epoch": 1.43, "learning_rate": 1.8859041449358218e-05, "loss": 0.0978, "step": 7806 }, { "epoch": 1.43, "learning_rate": 1.8858697035152035e-05, "loss": 0.0876, "step": 7807 }, { "epoch": 1.43, "learning_rate": 1.8858352572116506e-05, "loss": 0.0868, "step": 7808 }, { "epoch": 1.43, "learning_rate": 1.8858008060253533e-05, "loss": 0.0994, "step": 7809 }, { "epoch": 1.43, "learning_rate": 1.885766349956501e-05, "loss": 0.0994, "step": 7810 }, { "epoch": 1.43, "learning_rate": 1.8857318890052835e-05, "loss": 0.0964, "step": 7811 }, { "epoch": 1.43, "learning_rate": 1.8856974231718915e-05, "loss": 0.1031, "step": 7812 }, { "epoch": 1.43, "learning_rate": 1.8856629524565144e-05, "loss": 0.0885, "step": 7813 }, { "epoch": 1.43, "learning_rate": 1.8856284768593425e-05, "loss": 0.0978, "step": 7814 }, { "epoch": 1.43, "learning_rate": 1.8855939963805655e-05, "loss": 0.0912, "step": 7815 }, { "epoch": 1.43, "learning_rate": 1.8855595110203736e-05, "loss": 0.0916, "step": 7816 }, { "epoch": 1.43, "learning_rate": 1.885525020778957e-05, "loss": 0.097, "step": 7817 }, { "epoch": 1.43, "learning_rate": 1.885490525656506e-05, "loss": 0.1116, "step": 7818 }, { "epoch": 1.43, "learning_rate": 1.8854560256532098e-05, "loss": 0.1009, "step": 7819 }, { "epoch": 1.43, "learning_rate": 1.8854215207692598e-05, "loss": 0.0957, "step": 7820 }, { "epoch": 1.43, "learning_rate": 1.885387011004846e-05, "loss": 0.0904, "step": 7821 }, { "epoch": 1.43, "learning_rate": 1.8853524963601575e-05, "loss": 0.1014, "step": 7822 }, { "epoch": 1.43, "learning_rate": 1.8853179768353856e-05, "loss": 0.0932, "step": 7823 }, { "epoch": 1.43, "learning_rate": 1.88528345243072e-05, "loss": 0.0934, "step": 7824 }, { "epoch": 1.43, "learning_rate": 1.885248923146352e-05, "loss": 0.1039, "step": 7825 }, { "epoch": 1.44, "learning_rate": 1.8852143889824702e-05, "loss": 0.0912, "step": 7826 }, { "epoch": 1.44, "learning_rate": 1.8851798499392668e-05, "loss": 0.101, "step": 7827 }, { "epoch": 1.44, "learning_rate": 1.8851453060169306e-05, "loss": 0.1036, "step": 7828 }, { "epoch": 1.44, "learning_rate": 1.8851107572156535e-05, "loss": 0.1051, "step": 7829 }, { "epoch": 1.44, "learning_rate": 1.8850762035356248e-05, "loss": 0.0863, "step": 7830 }, { "epoch": 1.44, "learning_rate": 1.8850416449770353e-05, "loss": 0.0943, "step": 7831 }, { "epoch": 1.44, "learning_rate": 1.8850070815400755e-05, "loss": 0.1014, "step": 7832 }, { "epoch": 1.44, "learning_rate": 1.884972513224936e-05, "loss": 0.0915, "step": 7833 }, { "epoch": 1.44, "learning_rate": 1.884937940031807e-05, "loss": 0.089, "step": 7834 }, { "epoch": 1.44, "learning_rate": 1.8849033619608798e-05, "loss": 0.0953, "step": 7835 }, { "epoch": 1.44, "learning_rate": 1.8848687790123442e-05, "loss": 0.1065, "step": 7836 }, { "epoch": 1.44, "learning_rate": 1.8848341911863913e-05, "loss": 0.1047, "step": 7837 }, { "epoch": 1.44, "learning_rate": 1.8847995984832115e-05, "loss": 0.1121, "step": 7838 }, { "epoch": 1.44, "learning_rate": 1.884765000902996e-05, "loss": 0.1067, "step": 7839 }, { "epoch": 1.44, "learning_rate": 1.8847303984459347e-05, "loss": 0.0903, "step": 7840 }, { "epoch": 1.44, "learning_rate": 1.8846957911122188e-05, "loss": 0.0948, "step": 7841 }, { "epoch": 1.44, "learning_rate": 1.884661178902039e-05, "loss": 0.0931, "step": 7842 }, { "epoch": 1.44, "learning_rate": 1.884626561815586e-05, "loss": 0.0842, "step": 7843 }, { "epoch": 1.44, "learning_rate": 1.8845919398530507e-05, "loss": 0.0826, "step": 7844 }, { "epoch": 1.44, "learning_rate": 1.8845573130146237e-05, "loss": 0.1056, "step": 7845 }, { "epoch": 1.44, "learning_rate": 1.8845226813004965e-05, "loss": 0.1003, "step": 7846 }, { "epoch": 1.44, "learning_rate": 1.8844880447108593e-05, "loss": 0.0928, "step": 7847 }, { "epoch": 1.44, "learning_rate": 1.8844534032459033e-05, "loss": 0.0885, "step": 7848 }, { "epoch": 1.44, "learning_rate": 1.8844187569058194e-05, "loss": 0.0996, "step": 7849 }, { "epoch": 1.44, "learning_rate": 1.8843841056907986e-05, "loss": 0.0927, "step": 7850 }, { "epoch": 1.44, "learning_rate": 1.884349449601032e-05, "loss": 0.1002, "step": 7851 }, { "epoch": 1.44, "learning_rate": 1.8843147886367102e-05, "loss": 0.1116, "step": 7852 }, { "epoch": 1.44, "learning_rate": 1.8842801227980247e-05, "loss": 0.0895, "step": 7853 }, { "epoch": 1.44, "learning_rate": 1.8842454520851668e-05, "loss": 0.0999, "step": 7854 }, { "epoch": 1.44, "learning_rate": 1.8842107764983267e-05, "loss": 0.0925, "step": 7855 }, { "epoch": 1.44, "learning_rate": 1.8841760960376964e-05, "loss": 0.1015, "step": 7856 }, { "epoch": 1.44, "learning_rate": 1.884141410703467e-05, "loss": 0.0983, "step": 7857 }, { "epoch": 1.44, "learning_rate": 1.884106720495829e-05, "loss": 0.1114, "step": 7858 }, { "epoch": 1.44, "learning_rate": 1.884072025414974e-05, "loss": 0.1, "step": 7859 }, { "epoch": 1.44, "learning_rate": 1.8840373254610934e-05, "loss": 0.1001, "step": 7860 }, { "epoch": 1.44, "learning_rate": 1.8840026206343786e-05, "loss": 0.1011, "step": 7861 }, { "epoch": 1.44, "learning_rate": 1.8839679109350203e-05, "loss": 0.1051, "step": 7862 }, { "epoch": 1.44, "learning_rate": 1.8839331963632107e-05, "loss": 0.0997, "step": 7863 }, { "epoch": 1.44, "learning_rate": 1.8838984769191402e-05, "loss": 0.0982, "step": 7864 }, { "epoch": 1.44, "learning_rate": 1.8838637526030002e-05, "loss": 0.1003, "step": 7865 }, { "epoch": 1.44, "learning_rate": 1.8838290234149832e-05, "loss": 0.0956, "step": 7866 }, { "epoch": 1.44, "learning_rate": 1.8837942893552796e-05, "loss": 0.1008, "step": 7867 }, { "epoch": 1.44, "learning_rate": 1.883759550424081e-05, "loss": 0.109, "step": 7868 }, { "epoch": 1.44, "learning_rate": 1.8837248066215796e-05, "loss": 0.1053, "step": 7869 }, { "epoch": 1.44, "learning_rate": 1.8836900579479657e-05, "loss": 0.0948, "step": 7870 }, { "epoch": 1.44, "learning_rate": 1.883655304403432e-05, "loss": 0.1079, "step": 7871 }, { "epoch": 1.44, "learning_rate": 1.8836205459881697e-05, "loss": 0.0887, "step": 7872 }, { "epoch": 1.44, "learning_rate": 1.8835857827023697e-05, "loss": 0.0894, "step": 7873 }, { "epoch": 1.44, "learning_rate": 1.8835510145462242e-05, "loss": 0.0974, "step": 7874 }, { "epoch": 1.44, "learning_rate": 1.8835162415199252e-05, "loss": 0.0923, "step": 7875 }, { "epoch": 1.44, "learning_rate": 1.8834814636236638e-05, "loss": 0.1015, "step": 7876 }, { "epoch": 1.44, "learning_rate": 1.883446680857632e-05, "loss": 0.1, "step": 7877 }, { "epoch": 1.44, "learning_rate": 1.8834118932220212e-05, "loss": 0.0961, "step": 7878 }, { "epoch": 1.44, "learning_rate": 1.8833771007170232e-05, "loss": 0.1065, "step": 7879 }, { "epoch": 1.45, "learning_rate": 1.8833423033428302e-05, "loss": 0.0891, "step": 7880 }, { "epoch": 1.45, "learning_rate": 1.8833075010996337e-05, "loss": 0.0955, "step": 7881 }, { "epoch": 1.45, "learning_rate": 1.8832726939876257e-05, "loss": 0.1053, "step": 7882 }, { "epoch": 1.45, "learning_rate": 1.8832378820069977e-05, "loss": 0.1017, "step": 7883 }, { "epoch": 1.45, "learning_rate": 1.8832030651579418e-05, "loss": 0.0964, "step": 7884 }, { "epoch": 1.45, "learning_rate": 1.88316824344065e-05, "loss": 0.1111, "step": 7885 }, { "epoch": 1.45, "learning_rate": 1.883133416855314e-05, "loss": 0.0959, "step": 7886 }, { "epoch": 1.45, "learning_rate": 1.883098585402126e-05, "loss": 0.0978, "step": 7887 }, { "epoch": 1.45, "learning_rate": 1.883063749081278e-05, "loss": 0.0874, "step": 7888 }, { "epoch": 1.45, "learning_rate": 1.8830289078929618e-05, "loss": 0.0905, "step": 7889 }, { "epoch": 1.45, "learning_rate": 1.8829940618373696e-05, "loss": 0.1005, "step": 7890 }, { "epoch": 1.45, "learning_rate": 1.8829592109146936e-05, "loss": 0.0929, "step": 7891 }, { "epoch": 1.45, "learning_rate": 1.8829243551251255e-05, "loss": 0.0891, "step": 7892 }, { "epoch": 1.45, "learning_rate": 1.8828894944688577e-05, "loss": 0.1165, "step": 7893 }, { "epoch": 1.45, "learning_rate": 1.8828546289460825e-05, "loss": 0.1012, "step": 7894 }, { "epoch": 1.45, "learning_rate": 1.8828197585569918e-05, "loss": 0.0983, "step": 7895 }, { "epoch": 1.45, "learning_rate": 1.8827848833017782e-05, "loss": 0.1029, "step": 7896 }, { "epoch": 1.45, "learning_rate": 1.882750003180633e-05, "loss": 0.1, "step": 7897 }, { "epoch": 1.45, "learning_rate": 1.8827151181937498e-05, "loss": 0.1066, "step": 7898 }, { "epoch": 1.45, "learning_rate": 1.88268022834132e-05, "loss": 0.1044, "step": 7899 }, { "epoch": 1.45, "learning_rate": 1.8826453336235358e-05, "loss": 0.103, "step": 7900 }, { "epoch": 1.45, "learning_rate": 1.88261043404059e-05, "loss": 0.0929, "step": 7901 }, { "epoch": 1.45, "learning_rate": 1.8825755295926748e-05, "loss": 0.086, "step": 7902 }, { "epoch": 1.45, "learning_rate": 1.8825406202799824e-05, "loss": 0.0902, "step": 7903 }, { "epoch": 1.45, "learning_rate": 1.8825057061027056e-05, "loss": 0.0914, "step": 7904 }, { "epoch": 1.45, "learning_rate": 1.882470787061037e-05, "loss": 0.1039, "step": 7905 }, { "epoch": 1.45, "learning_rate": 1.8824358631551683e-05, "loss": 0.0839, "step": 7906 }, { "epoch": 1.45, "learning_rate": 1.8824009343852926e-05, "loss": 0.0965, "step": 7907 }, { "epoch": 1.45, "learning_rate": 1.8823660007516023e-05, "loss": 0.0977, "step": 7908 }, { "epoch": 1.45, "learning_rate": 1.8823310622542898e-05, "loss": 0.0945, "step": 7909 }, { "epoch": 1.45, "learning_rate": 1.882296118893548e-05, "loss": 0.0879, "step": 7910 }, { "epoch": 1.45, "learning_rate": 1.8822611706695693e-05, "loss": 0.0993, "step": 7911 }, { "epoch": 1.45, "learning_rate": 1.8822262175825463e-05, "loss": 0.1, "step": 7912 }, { "epoch": 1.45, "learning_rate": 1.8821912596326716e-05, "loss": 0.0913, "step": 7913 }, { "epoch": 1.45, "learning_rate": 1.8821562968201382e-05, "loss": 0.0947, "step": 7914 }, { "epoch": 1.45, "learning_rate": 1.8821213291451387e-05, "loss": 0.0848, "step": 7915 }, { "epoch": 1.45, "learning_rate": 1.8820863566078654e-05, "loss": 0.111, "step": 7916 }, { "epoch": 1.45, "learning_rate": 1.8820513792085116e-05, "loss": 0.0906, "step": 7917 }, { "epoch": 1.45, "learning_rate": 1.8820163969472704e-05, "loss": 0.1024, "step": 7918 }, { "epoch": 1.45, "learning_rate": 1.8819814098243337e-05, "loss": 0.0965, "step": 7919 }, { "epoch": 1.45, "learning_rate": 1.881946417839895e-05, "loss": 0.0862, "step": 7920 }, { "epoch": 1.45, "learning_rate": 1.8819114209941465e-05, "loss": 0.0884, "step": 7921 }, { "epoch": 1.45, "learning_rate": 1.881876419287282e-05, "loss": 0.0964, "step": 7922 }, { "epoch": 1.45, "learning_rate": 1.881841412719494e-05, "loss": 0.1018, "step": 7923 }, { "epoch": 1.45, "learning_rate": 1.8818064012909755e-05, "loss": 0.1006, "step": 7924 }, { "epoch": 1.45, "learning_rate": 1.8817713850019196e-05, "loss": 0.1092, "step": 7925 }, { "epoch": 1.45, "learning_rate": 1.881736363852519e-05, "loss": 0.0959, "step": 7926 }, { "epoch": 1.45, "learning_rate": 1.8817013378429667e-05, "loss": 0.1001, "step": 7927 }, { "epoch": 1.45, "learning_rate": 1.8816663069734563e-05, "loss": 0.0942, "step": 7928 }, { "epoch": 1.45, "learning_rate": 1.88163127124418e-05, "loss": 0.0965, "step": 7929 }, { "epoch": 1.45, "learning_rate": 1.881596230655332e-05, "loss": 0.099, "step": 7930 }, { "epoch": 1.45, "learning_rate": 1.881561185207105e-05, "loss": 0.1021, "step": 7931 }, { "epoch": 1.45, "learning_rate": 1.8815261348996916e-05, "loss": 0.1013, "step": 7932 }, { "epoch": 1.45, "learning_rate": 1.8814910797332858e-05, "loss": 0.1059, "step": 7933 }, { "epoch": 1.45, "learning_rate": 1.8814560197080808e-05, "loss": 0.1132, "step": 7934 }, { "epoch": 1.46, "learning_rate": 1.881420954824269e-05, "loss": 0.1127, "step": 7935 }, { "epoch": 1.46, "learning_rate": 1.8813858850820446e-05, "loss": 0.0987, "step": 7936 }, { "epoch": 1.46, "learning_rate": 1.8813508104816005e-05, "loss": 0.0927, "step": 7937 }, { "epoch": 1.46, "learning_rate": 1.88131573102313e-05, "loss": 0.0979, "step": 7938 }, { "epoch": 1.46, "learning_rate": 1.881280646706827e-05, "loss": 0.0803, "step": 7939 }, { "epoch": 1.46, "learning_rate": 1.8812455575328837e-05, "loss": 0.1059, "step": 7940 }, { "epoch": 1.46, "learning_rate": 1.8812104635014946e-05, "loss": 0.1003, "step": 7941 }, { "epoch": 1.46, "learning_rate": 1.8811753646128526e-05, "loss": 0.104, "step": 7942 }, { "epoch": 1.46, "learning_rate": 1.8811402608671513e-05, "loss": 0.1005, "step": 7943 }, { "epoch": 1.46, "learning_rate": 1.8811051522645847e-05, "loss": 0.0962, "step": 7944 }, { "epoch": 1.46, "learning_rate": 1.8810700388053457e-05, "loss": 0.0902, "step": 7945 }, { "epoch": 1.46, "learning_rate": 1.8810349204896276e-05, "loss": 0.1101, "step": 7946 }, { "epoch": 1.46, "learning_rate": 1.8809997973176248e-05, "loss": 0.0934, "step": 7947 }, { "epoch": 1.46, "learning_rate": 1.8809646692895302e-05, "loss": 0.0993, "step": 7948 }, { "epoch": 1.46, "learning_rate": 1.880929536405538e-05, "loss": 0.0938, "step": 7949 }, { "epoch": 1.46, "learning_rate": 1.8808943986658414e-05, "loss": 0.1068, "step": 7950 }, { "epoch": 1.46, "learning_rate": 1.8808592560706344e-05, "loss": 0.0963, "step": 7951 }, { "epoch": 1.46, "learning_rate": 1.8808241086201106e-05, "loss": 0.1028, "step": 7952 }, { "epoch": 1.46, "learning_rate": 1.8807889563144634e-05, "loss": 0.0949, "step": 7953 }, { "epoch": 1.46, "learning_rate": 1.8807537991538866e-05, "loss": 0.1061, "step": 7954 }, { "epoch": 1.46, "learning_rate": 1.8807186371385746e-05, "loss": 0.0993, "step": 7955 }, { "epoch": 1.46, "learning_rate": 1.880683470268721e-05, "loss": 0.1043, "step": 7956 }, { "epoch": 1.46, "learning_rate": 1.880648298544519e-05, "loss": 0.0929, "step": 7957 }, { "epoch": 1.46, "learning_rate": 1.880613121966163e-05, "loss": 0.0962, "step": 7958 }, { "epoch": 1.46, "learning_rate": 1.880577940533847e-05, "loss": 0.0928, "step": 7959 }, { "epoch": 1.46, "learning_rate": 1.880542754247765e-05, "loss": 0.1095, "step": 7960 }, { "epoch": 1.46, "learning_rate": 1.8805075631081107e-05, "loss": 0.1064, "step": 7961 }, { "epoch": 1.46, "learning_rate": 1.8804723671150777e-05, "loss": 0.1018, "step": 7962 }, { "epoch": 1.46, "learning_rate": 1.8804371662688606e-05, "loss": 0.0968, "step": 7963 }, { "epoch": 1.46, "learning_rate": 1.8804019605696534e-05, "loss": 0.0986, "step": 7964 }, { "epoch": 1.46, "learning_rate": 1.8803667500176497e-05, "loss": 0.0977, "step": 7965 }, { "epoch": 1.46, "learning_rate": 1.8803315346130437e-05, "loss": 0.0974, "step": 7966 }, { "epoch": 1.46, "learning_rate": 1.88029631435603e-05, "loss": 0.104, "step": 7967 }, { "epoch": 1.46, "learning_rate": 1.8802610892468024e-05, "loss": 0.1003, "step": 7968 }, { "epoch": 1.46, "learning_rate": 1.8802258592855547e-05, "loss": 0.1012, "step": 7969 }, { "epoch": 1.46, "learning_rate": 1.8801906244724817e-05, "loss": 0.0929, "step": 7970 }, { "epoch": 1.46, "learning_rate": 1.8801553848077774e-05, "loss": 0.0925, "step": 7971 }, { "epoch": 1.46, "learning_rate": 1.880120140291636e-05, "loss": 0.1027, "step": 7972 }, { "epoch": 1.46, "learning_rate": 1.8800848909242516e-05, "loss": 0.1, "step": 7973 }, { "epoch": 1.46, "learning_rate": 1.8800496367058187e-05, "loss": 0.1051, "step": 7974 }, { "epoch": 1.46, "learning_rate": 1.8800143776365314e-05, "loss": 0.0994, "step": 7975 }, { "epoch": 1.46, "learning_rate": 1.8799791137165847e-05, "loss": 0.0957, "step": 7976 }, { "epoch": 1.46, "learning_rate": 1.8799438449461722e-05, "loss": 0.0995, "step": 7977 }, { "epoch": 1.46, "learning_rate": 1.8799085713254888e-05, "loss": 0.0992, "step": 7978 }, { "epoch": 1.46, "learning_rate": 1.8798732928547286e-05, "loss": 0.0954, "step": 7979 }, { "epoch": 1.46, "learning_rate": 1.879838009534086e-05, "loss": 0.1082, "step": 7980 }, { "epoch": 1.46, "learning_rate": 1.879802721363756e-05, "loss": 0.109, "step": 7981 }, { "epoch": 1.46, "learning_rate": 1.879767428343933e-05, "loss": 0.1006, "step": 7982 }, { "epoch": 1.46, "learning_rate": 1.8797321304748107e-05, "loss": 0.0989, "step": 7983 }, { "epoch": 1.46, "learning_rate": 1.8796968277565848e-05, "loss": 0.0938, "step": 7984 }, { "epoch": 1.46, "learning_rate": 1.879661520189449e-05, "loss": 0.1018, "step": 7985 }, { "epoch": 1.46, "learning_rate": 1.8796262077735986e-05, "loss": 0.1032, "step": 7986 }, { "epoch": 1.46, "learning_rate": 1.8795908905092278e-05, "loss": 0.0963, "step": 7987 }, { "epoch": 1.46, "learning_rate": 1.8795555683965316e-05, "loss": 0.0988, "step": 7988 }, { "epoch": 1.47, "learning_rate": 1.879520241435704e-05, "loss": 0.0979, "step": 7989 }, { "epoch": 1.47, "learning_rate": 1.8794849096269407e-05, "loss": 0.099, "step": 7990 }, { "epoch": 1.47, "learning_rate": 1.879449572970436e-05, "loss": 0.0895, "step": 7991 }, { "epoch": 1.47, "learning_rate": 1.879414231466384e-05, "loss": 0.0929, "step": 7992 }, { "epoch": 1.47, "learning_rate": 1.879378885114981e-05, "loss": 0.1047, "step": 7993 }, { "epoch": 1.47, "learning_rate": 1.8793435339164204e-05, "loss": 0.0916, "step": 7994 }, { "epoch": 1.47, "learning_rate": 1.879308177870898e-05, "loss": 0.103, "step": 7995 }, { "epoch": 1.47, "learning_rate": 1.879272816978608e-05, "loss": 0.0936, "step": 7996 }, { "epoch": 1.47, "learning_rate": 1.879237451239746e-05, "loss": 0.0963, "step": 7997 }, { "epoch": 1.47, "learning_rate": 1.879202080654506e-05, "loss": 0.1059, "step": 7998 }, { "epoch": 1.47, "learning_rate": 1.879166705223084e-05, "loss": 0.1008, "step": 7999 }, { "epoch": 1.47, "learning_rate": 1.8791313249456746e-05, "loss": 0.0818, "step": 8000 }, { "epoch": 1.47, "learning_rate": 1.8790959398224725e-05, "loss": 0.1047, "step": 8001 }, { "epoch": 1.47, "learning_rate": 1.8790605498536728e-05, "loss": 0.1008, "step": 8002 }, { "epoch": 1.47, "learning_rate": 1.8790251550394712e-05, "loss": 0.0815, "step": 8003 }, { "epoch": 1.47, "learning_rate": 1.8789897553800623e-05, "loss": 0.1, "step": 8004 }, { "epoch": 1.47, "learning_rate": 1.878954350875641e-05, "loss": 0.1008, "step": 8005 }, { "epoch": 1.47, "learning_rate": 1.8789189415264028e-05, "loss": 0.0833, "step": 8006 }, { "epoch": 1.47, "learning_rate": 1.878883527332543e-05, "loss": 0.1171, "step": 8007 }, { "epoch": 1.47, "learning_rate": 1.8788481082942563e-05, "loss": 0.0889, "step": 8008 }, { "epoch": 1.47, "learning_rate": 1.8788126844117384e-05, "loss": 0.1054, "step": 8009 }, { "epoch": 1.47, "learning_rate": 1.8787772556851847e-05, "loss": 0.0895, "step": 8010 }, { "epoch": 1.47, "learning_rate": 1.8787418221147897e-05, "loss": 0.0949, "step": 8011 }, { "epoch": 1.47, "learning_rate": 1.8787063837007493e-05, "loss": 0.1116, "step": 8012 }, { "epoch": 1.47, "learning_rate": 1.878670940443259e-05, "loss": 0.0966, "step": 8013 }, { "epoch": 1.47, "learning_rate": 1.8786354923425135e-05, "loss": 0.1065, "step": 8014 }, { "epoch": 1.47, "learning_rate": 1.878600039398709e-05, "loss": 0.0957, "step": 8015 }, { "epoch": 1.47, "learning_rate": 1.87856458161204e-05, "loss": 0.1072, "step": 8016 }, { "epoch": 1.47, "learning_rate": 1.878529118982703e-05, "loss": 0.0925, "step": 8017 }, { "epoch": 1.47, "learning_rate": 1.8784936515108923e-05, "loss": 0.1031, "step": 8018 }, { "epoch": 1.47, "learning_rate": 1.8784581791968045e-05, "loss": 0.1048, "step": 8019 }, { "epoch": 1.47, "learning_rate": 1.8784227020406344e-05, "loss": 0.1051, "step": 8020 }, { "epoch": 1.47, "learning_rate": 1.878387220042578e-05, "loss": 0.0932, "step": 8021 }, { "epoch": 1.47, "learning_rate": 1.8783517332028303e-05, "loss": 0.1284, "step": 8022 }, { "epoch": 1.47, "learning_rate": 1.8783162415215875e-05, "loss": 0.0988, "step": 8023 }, { "epoch": 1.47, "learning_rate": 1.8782807449990453e-05, "loss": 0.1075, "step": 8024 }, { "epoch": 1.47, "learning_rate": 1.8782452436353987e-05, "loss": 0.0887, "step": 8025 }, { "epoch": 1.47, "learning_rate": 1.8782097374308436e-05, "loss": 0.098, "step": 8026 }, { "epoch": 1.47, "learning_rate": 1.8781742263855762e-05, "loss": 0.1018, "step": 8027 }, { "epoch": 1.47, "learning_rate": 1.878138710499792e-05, "loss": 0.0924, "step": 8028 }, { "epoch": 1.47, "learning_rate": 1.878103189773686e-05, "loss": 0.1077, "step": 8029 }, { "epoch": 1.47, "learning_rate": 1.878067664207455e-05, "loss": 0.1009, "step": 8030 }, { "epoch": 1.47, "learning_rate": 1.8780321338012947e-05, "loss": 0.0862, "step": 8031 }, { "epoch": 1.47, "learning_rate": 1.8779965985554006e-05, "loss": 0.0917, "step": 8032 }, { "epoch": 1.47, "learning_rate": 1.8779610584699687e-05, "loss": 0.103, "step": 8033 }, { "epoch": 1.47, "learning_rate": 1.8779255135451947e-05, "loss": 0.0985, "step": 8034 }, { "epoch": 1.47, "learning_rate": 1.8778899637812748e-05, "loss": 0.1083, "step": 8035 }, { "epoch": 1.47, "learning_rate": 1.8778544091784047e-05, "loss": 0.0925, "step": 8036 }, { "epoch": 1.47, "learning_rate": 1.8778188497367806e-05, "loss": 0.1017, "step": 8037 }, { "epoch": 1.47, "learning_rate": 1.8777832854565987e-05, "loss": 0.1196, "step": 8038 }, { "epoch": 1.47, "learning_rate": 1.8777477163380544e-05, "loss": 0.1105, "step": 8039 }, { "epoch": 1.47, "learning_rate": 1.877712142381344e-05, "loss": 0.0984, "step": 8040 }, { "epoch": 1.47, "learning_rate": 1.877676563586664e-05, "loss": 0.0848, "step": 8041 }, { "epoch": 1.47, "learning_rate": 1.8776409799542104e-05, "loss": 0.0974, "step": 8042 }, { "epoch": 1.47, "learning_rate": 1.877605391484179e-05, "loss": 0.0905, "step": 8043 }, { "epoch": 1.48, "learning_rate": 1.877569798176766e-05, "loss": 0.1099, "step": 8044 }, { "epoch": 1.48, "learning_rate": 1.877534200032168e-05, "loss": 0.1077, "step": 8045 }, { "epoch": 1.48, "learning_rate": 1.877498597050581e-05, "loss": 0.1013, "step": 8046 }, { "epoch": 1.48, "learning_rate": 1.8774629892322007e-05, "loss": 0.1036, "step": 8047 }, { "epoch": 1.48, "learning_rate": 1.877427376577224e-05, "loss": 0.0902, "step": 8048 }, { "epoch": 1.48, "learning_rate": 1.877391759085847e-05, "loss": 0.1045, "step": 8049 }, { "epoch": 1.48, "learning_rate": 1.8773561367582662e-05, "loss": 0.1028, "step": 8050 }, { "epoch": 1.48, "learning_rate": 1.877320509594678e-05, "loss": 0.1038, "step": 8051 }, { "epoch": 1.48, "learning_rate": 1.8772848775952784e-05, "loss": 0.0918, "step": 8052 }, { "epoch": 1.48, "learning_rate": 1.8772492407602637e-05, "loss": 0.0834, "step": 8053 }, { "epoch": 1.48, "learning_rate": 1.877213599089831e-05, "loss": 0.1075, "step": 8054 }, { "epoch": 1.48, "learning_rate": 1.877177952584176e-05, "loss": 0.1023, "step": 8055 }, { "epoch": 1.48, "learning_rate": 1.8771423012434963e-05, "loss": 0.0959, "step": 8056 }, { "epoch": 1.48, "learning_rate": 1.8771066450679872e-05, "loss": 0.092, "step": 8057 }, { "epoch": 1.48, "learning_rate": 1.877070984057846e-05, "loss": 0.1032, "step": 8058 }, { "epoch": 1.48, "learning_rate": 1.8770353182132684e-05, "loss": 0.0995, "step": 8059 }, { "epoch": 1.48, "learning_rate": 1.876999647534452e-05, "loss": 0.1069, "step": 8060 }, { "epoch": 1.48, "learning_rate": 1.876963972021593e-05, "loss": 0.1008, "step": 8061 }, { "epoch": 1.48, "learning_rate": 1.8769282916748882e-05, "loss": 0.1033, "step": 8062 }, { "epoch": 1.48, "learning_rate": 1.876892606494534e-05, "loss": 0.0886, "step": 8063 }, { "epoch": 1.48, "learning_rate": 1.8768569164807272e-05, "loss": 0.0956, "step": 8064 }, { "epoch": 1.48, "learning_rate": 1.876821221633664e-05, "loss": 0.0944, "step": 8065 }, { "epoch": 1.48, "learning_rate": 1.8767855219535423e-05, "loss": 0.0967, "step": 8066 }, { "epoch": 1.48, "learning_rate": 1.8767498174405582e-05, "loss": 0.0983, "step": 8067 }, { "epoch": 1.48, "learning_rate": 1.8767141080949086e-05, "loss": 0.1044, "step": 8068 }, { "epoch": 1.48, "learning_rate": 1.87667839391679e-05, "loss": 0.1023, "step": 8069 }, { "epoch": 1.48, "learning_rate": 1.8766426749063997e-05, "loss": 0.0894, "step": 8070 }, { "epoch": 1.48, "learning_rate": 1.8766069510639347e-05, "loss": 0.112, "step": 8071 }, { "epoch": 1.48, "learning_rate": 1.8765712223895912e-05, "loss": 0.1054, "step": 8072 }, { "epoch": 1.48, "learning_rate": 1.876535488883567e-05, "loss": 0.0914, "step": 8073 }, { "epoch": 1.48, "learning_rate": 1.876499750546058e-05, "loss": 0.1004, "step": 8074 }, { "epoch": 1.48, "learning_rate": 1.8764640073772625e-05, "loss": 0.1069, "step": 8075 }, { "epoch": 1.48, "learning_rate": 1.8764282593773765e-05, "loss": 0.1034, "step": 8076 }, { "epoch": 1.48, "learning_rate": 1.8763925065465977e-05, "loss": 0.093, "step": 8077 }, { "epoch": 1.48, "learning_rate": 1.876356748885123e-05, "loss": 0.1032, "step": 8078 }, { "epoch": 1.48, "learning_rate": 1.876320986393149e-05, "loss": 0.1035, "step": 8079 }, { "epoch": 1.48, "learning_rate": 1.876285219070873e-05, "loss": 0.0966, "step": 8080 }, { "epoch": 1.48, "learning_rate": 1.8762494469184928e-05, "loss": 0.0955, "step": 8081 }, { "epoch": 1.48, "learning_rate": 1.876213669936205e-05, "loss": 0.0951, "step": 8082 }, { "epoch": 1.48, "learning_rate": 1.8761778881242067e-05, "loss": 0.1098, "step": 8083 }, { "epoch": 1.48, "learning_rate": 1.8761421014826956e-05, "loss": 0.1022, "step": 8084 }, { "epoch": 1.48, "learning_rate": 1.876106310011869e-05, "loss": 0.1018, "step": 8085 }, { "epoch": 1.48, "learning_rate": 1.8760705137119234e-05, "loss": 0.0971, "step": 8086 }, { "epoch": 1.48, "learning_rate": 1.8760347125830565e-05, "loss": 0.1033, "step": 8087 }, { "epoch": 1.48, "learning_rate": 1.875998906625466e-05, "loss": 0.0882, "step": 8088 }, { "epoch": 1.48, "learning_rate": 1.875963095839349e-05, "loss": 0.0921, "step": 8089 }, { "epoch": 1.48, "learning_rate": 1.875927280224903e-05, "loss": 0.1, "step": 8090 }, { "epoch": 1.48, "learning_rate": 1.875891459782325e-05, "loss": 0.0903, "step": 8091 }, { "epoch": 1.48, "learning_rate": 1.875855634511813e-05, "loss": 0.1167, "step": 8092 }, { "epoch": 1.48, "learning_rate": 1.875819804413564e-05, "loss": 0.0975, "step": 8093 }, { "epoch": 1.48, "learning_rate": 1.875783969487776e-05, "loss": 0.0936, "step": 8094 }, { "epoch": 1.48, "learning_rate": 1.875748129734646e-05, "loss": 0.0898, "step": 8095 }, { "epoch": 1.48, "learning_rate": 1.875712285154372e-05, "loss": 0.0888, "step": 8096 }, { "epoch": 1.48, "learning_rate": 1.8756764357471514e-05, "loss": 0.1038, "step": 8097 }, { "epoch": 1.49, "learning_rate": 1.8756405815131815e-05, "loss": 0.094, "step": 8098 }, { "epoch": 1.49, "learning_rate": 1.8756047224526604e-05, "loss": 0.0952, "step": 8099 }, { "epoch": 1.49, "learning_rate": 1.8755688585657856e-05, "loss": 0.0975, "step": 8100 }, { "epoch": 1.49, "learning_rate": 1.8755329898527545e-05, "loss": 0.1001, "step": 8101 }, { "epoch": 1.49, "learning_rate": 1.875497116313765e-05, "loss": 0.0931, "step": 8102 }, { "epoch": 1.49, "learning_rate": 1.8754612379490155e-05, "loss": 0.0839, "step": 8103 }, { "epoch": 1.49, "learning_rate": 1.8754253547587026e-05, "loss": 0.1013, "step": 8104 }, { "epoch": 1.49, "learning_rate": 1.8753894667430247e-05, "loss": 0.0918, "step": 8105 }, { "epoch": 1.49, "learning_rate": 1.8753535739021798e-05, "loss": 0.1158, "step": 8106 }, { "epoch": 1.49, "learning_rate": 1.8753176762363652e-05, "loss": 0.1158, "step": 8107 }, { "epoch": 1.49, "learning_rate": 1.875281773745779e-05, "loss": 0.0974, "step": 8108 }, { "epoch": 1.49, "learning_rate": 1.8752458664306194e-05, "loss": 0.1031, "step": 8109 }, { "epoch": 1.49, "learning_rate": 1.875209954291084e-05, "loss": 0.094, "step": 8110 }, { "epoch": 1.49, "learning_rate": 1.875174037327371e-05, "loss": 0.1074, "step": 8111 }, { "epoch": 1.49, "learning_rate": 1.8751381155396783e-05, "loss": 0.0885, "step": 8112 }, { "epoch": 1.49, "learning_rate": 1.8751021889282034e-05, "loss": 0.1027, "step": 8113 }, { "epoch": 1.49, "learning_rate": 1.8750662574931446e-05, "loss": 0.1033, "step": 8114 }, { "epoch": 1.49, "learning_rate": 1.8750303212347007e-05, "loss": 0.0848, "step": 8115 }, { "epoch": 1.49, "learning_rate": 1.8749943801530685e-05, "loss": 0.0984, "step": 8116 }, { "epoch": 1.49, "learning_rate": 1.8749584342484474e-05, "loss": 0.1006, "step": 8117 }, { "epoch": 1.49, "learning_rate": 1.8749224835210345e-05, "loss": 0.0886, "step": 8118 }, { "epoch": 1.49, "learning_rate": 1.8748865279710283e-05, "loss": 0.0902, "step": 8119 }, { "epoch": 1.49, "learning_rate": 1.8748505675986274e-05, "loss": 0.112, "step": 8120 }, { "epoch": 1.49, "learning_rate": 1.8748146024040296e-05, "loss": 0.0906, "step": 8121 }, { "epoch": 1.49, "learning_rate": 1.8747786323874332e-05, "loss": 0.0935, "step": 8122 }, { "epoch": 1.49, "learning_rate": 1.8747426575490363e-05, "loss": 0.1033, "step": 8123 }, { "epoch": 1.49, "learning_rate": 1.8747066778890376e-05, "loss": 0.0976, "step": 8124 }, { "epoch": 1.49, "learning_rate": 1.874670693407635e-05, "loss": 0.0909, "step": 8125 }, { "epoch": 1.49, "learning_rate": 1.874634704105027e-05, "loss": 0.1025, "step": 8126 }, { "epoch": 1.49, "learning_rate": 1.8745987099814126e-05, "loss": 0.1036, "step": 8127 }, { "epoch": 1.49, "learning_rate": 1.874562711036989e-05, "loss": 0.0982, "step": 8128 }, { "epoch": 1.49, "learning_rate": 1.8745267072719555e-05, "loss": 0.1049, "step": 8129 }, { "epoch": 1.49, "learning_rate": 1.8744906986865104e-05, "loss": 0.093, "step": 8130 }, { "epoch": 1.49, "learning_rate": 1.874454685280852e-05, "loss": 0.0949, "step": 8131 }, { "epoch": 1.49, "learning_rate": 1.8744186670551787e-05, "loss": 0.0989, "step": 8132 }, { "epoch": 1.49, "learning_rate": 1.8743826440096892e-05, "loss": 0.0923, "step": 8133 }, { "epoch": 1.49, "learning_rate": 1.8743466161445823e-05, "loss": 0.0928, "step": 8134 }, { "epoch": 1.49, "learning_rate": 1.8743105834600564e-05, "loss": 0.0895, "step": 8135 }, { "epoch": 1.49, "learning_rate": 1.8742745459563098e-05, "loss": 0.0941, "step": 8136 }, { "epoch": 1.49, "learning_rate": 1.8742385036335417e-05, "loss": 0.1003, "step": 8137 }, { "epoch": 1.49, "learning_rate": 1.8742024564919503e-05, "loss": 0.1039, "step": 8138 }, { "epoch": 1.49, "learning_rate": 1.8741664045317346e-05, "loss": 0.1102, "step": 8139 }, { "epoch": 1.49, "learning_rate": 1.8741303477530932e-05, "loss": 0.0992, "step": 8140 }, { "epoch": 1.49, "learning_rate": 1.8740942861562246e-05, "loss": 0.0958, "step": 8141 }, { "epoch": 1.49, "learning_rate": 1.8740582197413282e-05, "loss": 0.0999, "step": 8142 }, { "epoch": 1.49, "learning_rate": 1.8740221485086023e-05, "loss": 0.0811, "step": 8143 }, { "epoch": 1.49, "learning_rate": 1.8739860724582456e-05, "loss": 0.0975, "step": 8144 }, { "epoch": 1.49, "learning_rate": 1.8739499915904576e-05, "loss": 0.0968, "step": 8145 }, { "epoch": 1.49, "learning_rate": 1.8739139059054366e-05, "loss": 0.0891, "step": 8146 }, { "epoch": 1.49, "learning_rate": 1.8738778154033813e-05, "loss": 0.0958, "step": 8147 }, { "epoch": 1.49, "learning_rate": 1.8738417200844915e-05, "loss": 0.091, "step": 8148 }, { "epoch": 1.49, "learning_rate": 1.8738056199489655e-05, "loss": 0.117, "step": 8149 }, { "epoch": 1.49, "learning_rate": 1.8737695149970024e-05, "loss": 0.1093, "step": 8150 }, { "epoch": 1.49, "learning_rate": 1.8737334052288015e-05, "loss": 0.0986, "step": 8151 }, { "epoch": 1.49, "learning_rate": 1.8736972906445614e-05, "loss": 0.0929, "step": 8152 }, { "epoch": 1.5, "learning_rate": 1.8736611712444812e-05, "loss": 0.1033, "step": 8153 }, { "epoch": 1.5, "learning_rate": 1.8736250470287602e-05, "loss": 0.1073, "step": 8154 }, { "epoch": 1.5, "learning_rate": 1.873588917997598e-05, "loss": 0.1052, "step": 8155 }, { "epoch": 1.5, "learning_rate": 1.8735527841511926e-05, "loss": 0.0817, "step": 8156 }, { "epoch": 1.5, "learning_rate": 1.873516645489744e-05, "loss": 0.1009, "step": 8157 }, { "epoch": 1.5, "learning_rate": 1.873480502013451e-05, "loss": 0.1094, "step": 8158 }, { "epoch": 1.5, "learning_rate": 1.8734443537225133e-05, "loss": 0.0944, "step": 8159 }, { "epoch": 1.5, "learning_rate": 1.87340820061713e-05, "loss": 0.1106, "step": 8160 }, { "epoch": 1.5, "learning_rate": 1.8733720426975e-05, "loss": 0.0951, "step": 8161 }, { "epoch": 1.5, "learning_rate": 1.873335879963823e-05, "loss": 0.1012, "step": 8162 }, { "epoch": 1.5, "learning_rate": 1.873299712416298e-05, "loss": 0.0995, "step": 8163 }, { "epoch": 1.5, "learning_rate": 1.8732635400551245e-05, "loss": 0.0882, "step": 8164 }, { "epoch": 1.5, "learning_rate": 1.8732273628805018e-05, "loss": 0.0975, "step": 8165 }, { "epoch": 1.5, "learning_rate": 1.8731911808926296e-05, "loss": 0.1101, "step": 8166 }, { "epoch": 1.5, "learning_rate": 1.8731549940917072e-05, "loss": 0.1066, "step": 8167 }, { "epoch": 1.5, "learning_rate": 1.8731188024779337e-05, "loss": 0.0991, "step": 8168 }, { "epoch": 1.5, "learning_rate": 1.873082606051509e-05, "loss": 0.0986, "step": 8169 }, { "epoch": 1.5, "learning_rate": 1.873046404812633e-05, "loss": 0.0889, "step": 8170 }, { "epoch": 1.5, "learning_rate": 1.8730101987615044e-05, "loss": 0.0984, "step": 8171 }, { "epoch": 1.5, "learning_rate": 1.872973987898323e-05, "loss": 0.0897, "step": 8172 }, { "epoch": 1.5, "learning_rate": 1.872937772223289e-05, "loss": 0.1047, "step": 8173 }, { "epoch": 1.5, "learning_rate": 1.872901551736601e-05, "loss": 0.1017, "step": 8174 }, { "epoch": 1.5, "learning_rate": 1.8728653264384594e-05, "loss": 0.0984, "step": 8175 }, { "epoch": 1.5, "learning_rate": 1.872829096329064e-05, "loss": 0.109, "step": 8176 }, { "epoch": 1.5, "learning_rate": 1.8727928614086134e-05, "loss": 0.0846, "step": 8177 }, { "epoch": 1.5, "learning_rate": 1.872756621677309e-05, "loss": 0.0967, "step": 8178 }, { "epoch": 1.5, "learning_rate": 1.872720377135349e-05, "loss": 0.0863, "step": 8179 }, { "epoch": 1.5, "learning_rate": 1.872684127782934e-05, "loss": 0.1036, "step": 8180 }, { "epoch": 1.5, "learning_rate": 1.8726478736202638e-05, "loss": 0.0937, "step": 8181 }, { "epoch": 1.5, "learning_rate": 1.872611614647538e-05, "loss": 0.0777, "step": 8182 }, { "epoch": 1.5, "learning_rate": 1.8725753508649567e-05, "loss": 0.1035, "step": 8183 }, { "epoch": 1.5, "learning_rate": 1.8725390822727192e-05, "loss": 0.1062, "step": 8184 }, { "epoch": 1.5, "learning_rate": 1.8725028088710263e-05, "loss": 0.0986, "step": 8185 }, { "epoch": 1.5, "learning_rate": 1.872466530660077e-05, "loss": 0.1051, "step": 8186 }, { "epoch": 1.5, "learning_rate": 1.8724302476400723e-05, "loss": 0.1091, "step": 8187 }, { "epoch": 1.5, "learning_rate": 1.8723939598112114e-05, "loss": 0.0847, "step": 8188 }, { "epoch": 1.5, "learning_rate": 1.8723576671736944e-05, "loss": 0.0912, "step": 8189 }, { "epoch": 1.5, "learning_rate": 1.8723213697277213e-05, "loss": 0.1029, "step": 8190 }, { "epoch": 1.5, "learning_rate": 1.872285067473493e-05, "loss": 0.1055, "step": 8191 }, { "epoch": 1.5, "learning_rate": 1.8722487604112084e-05, "loss": 0.0955, "step": 8192 }, { "epoch": 1.5, "learning_rate": 1.8722124485410686e-05, "loss": 0.1084, "step": 8193 }, { "epoch": 1.5, "learning_rate": 1.872176131863273e-05, "loss": 0.0882, "step": 8194 }, { "epoch": 1.5, "learning_rate": 1.8721398103780223e-05, "loss": 0.1004, "step": 8195 }, { "epoch": 1.5, "learning_rate": 1.872103484085517e-05, "loss": 0.1002, "step": 8196 }, { "epoch": 1.5, "learning_rate": 1.8720671529859562e-05, "loss": 0.0882, "step": 8197 }, { "epoch": 1.5, "learning_rate": 1.872030817079541e-05, "loss": 0.0928, "step": 8198 }, { "epoch": 1.5, "learning_rate": 1.8719944763664716e-05, "loss": 0.1068, "step": 8199 }, { "epoch": 1.5, "learning_rate": 1.871958130846948e-05, "loss": 0.0949, "step": 8200 }, { "epoch": 1.5, "learning_rate": 1.871921780521171e-05, "loss": 0.0868, "step": 8201 }, { "epoch": 1.5, "learning_rate": 1.8718854253893405e-05, "loss": 0.0981, "step": 8202 }, { "epoch": 1.5, "learning_rate": 1.8718490654516575e-05, "loss": 0.1004, "step": 8203 }, { "epoch": 1.5, "learning_rate": 1.8718127007083218e-05, "loss": 0.1015, "step": 8204 }, { "epoch": 1.5, "learning_rate": 1.8717763311595338e-05, "loss": 0.079, "step": 8205 }, { "epoch": 1.5, "learning_rate": 1.8717399568054947e-05, "loss": 0.0921, "step": 8206 }, { "epoch": 1.51, "learning_rate": 1.8717035776464043e-05, "loss": 0.1091, "step": 8207 }, { "epoch": 1.51, "learning_rate": 1.8716671936824634e-05, "loss": 0.1023, "step": 8208 }, { "epoch": 1.51, "learning_rate": 1.8716308049138723e-05, "loss": 0.1044, "step": 8209 }, { "epoch": 1.51, "learning_rate": 1.871594411340832e-05, "loss": 0.0996, "step": 8210 }, { "epoch": 1.51, "learning_rate": 1.871558012963543e-05, "loss": 0.0995, "step": 8211 }, { "epoch": 1.51, "learning_rate": 1.8715216097822057e-05, "loss": 0.1061, "step": 8212 }, { "epoch": 1.51, "learning_rate": 1.871485201797021e-05, "loss": 0.0969, "step": 8213 }, { "epoch": 1.51, "learning_rate": 1.8714487890081892e-05, "loss": 0.1125, "step": 8214 }, { "epoch": 1.51, "learning_rate": 1.8714123714159113e-05, "loss": 0.0941, "step": 8215 }, { "epoch": 1.51, "learning_rate": 1.871375949020388e-05, "loss": 0.1036, "step": 8216 }, { "epoch": 1.51, "learning_rate": 1.87133952182182e-05, "loss": 0.0947, "step": 8217 }, { "epoch": 1.51, "learning_rate": 1.8713030898204082e-05, "loss": 0.0966, "step": 8218 }, { "epoch": 1.51, "learning_rate": 1.8712666530163534e-05, "loss": 0.1084, "step": 8219 }, { "epoch": 1.51, "learning_rate": 1.8712302114098566e-05, "loss": 0.098, "step": 8220 }, { "epoch": 1.51, "learning_rate": 1.871193765001118e-05, "loss": 0.1, "step": 8221 }, { "epoch": 1.51, "learning_rate": 1.8711573137903393e-05, "loss": 0.0966, "step": 8222 }, { "epoch": 1.51, "learning_rate": 1.8711208577777213e-05, "loss": 0.0955, "step": 8223 }, { "epoch": 1.51, "learning_rate": 1.8710843969634642e-05, "loss": 0.1034, "step": 8224 }, { "epoch": 1.51, "learning_rate": 1.8710479313477697e-05, "loss": 0.091, "step": 8225 }, { "epoch": 1.51, "learning_rate": 1.8710114609308388e-05, "loss": 0.095, "step": 8226 }, { "epoch": 1.51, "learning_rate": 1.8709749857128718e-05, "loss": 0.0989, "step": 8227 }, { "epoch": 1.51, "learning_rate": 1.870938505694071e-05, "loss": 0.091, "step": 8228 }, { "epoch": 1.51, "learning_rate": 1.8709020208746362e-05, "loss": 0.1011, "step": 8229 }, { "epoch": 1.51, "learning_rate": 1.870865531254769e-05, "loss": 0.092, "step": 8230 }, { "epoch": 1.51, "learning_rate": 1.8708290368346708e-05, "loss": 0.1124, "step": 8231 }, { "epoch": 1.51, "learning_rate": 1.8707925376145427e-05, "loss": 0.0957, "step": 8232 }, { "epoch": 1.51, "learning_rate": 1.8707560335945855e-05, "loss": 0.098, "step": 8233 }, { "epoch": 1.51, "learning_rate": 1.8707195247750007e-05, "loss": 0.0922, "step": 8234 }, { "epoch": 1.51, "learning_rate": 1.8706830111559898e-05, "loss": 0.1044, "step": 8235 }, { "epoch": 1.51, "learning_rate": 1.8706464927377532e-05, "loss": 0.0914, "step": 8236 }, { "epoch": 1.51, "learning_rate": 1.870609969520493e-05, "loss": 0.1108, "step": 8237 }, { "epoch": 1.51, "learning_rate": 1.8705734415044104e-05, "loss": 0.1069, "step": 8238 }, { "epoch": 1.51, "learning_rate": 1.8705369086897063e-05, "loss": 0.0992, "step": 8239 }, { "epoch": 1.51, "learning_rate": 1.8705003710765823e-05, "loss": 0.0967, "step": 8240 }, { "epoch": 1.51, "learning_rate": 1.87046382866524e-05, "loss": 0.0996, "step": 8241 }, { "epoch": 1.51, "learning_rate": 1.870427281455881e-05, "loss": 0.1024, "step": 8242 }, { "epoch": 1.51, "learning_rate": 1.8703907294487062e-05, "loss": 0.1044, "step": 8243 }, { "epoch": 1.51, "learning_rate": 1.8703541726439173e-05, "loss": 0.0983, "step": 8244 }, { "epoch": 1.51, "learning_rate": 1.8703176110417157e-05, "loss": 0.0827, "step": 8245 }, { "epoch": 1.51, "learning_rate": 1.8702810446423034e-05, "loss": 0.0913, "step": 8246 }, { "epoch": 1.51, "learning_rate": 1.8702444734458812e-05, "loss": 0.0999, "step": 8247 }, { "epoch": 1.51, "learning_rate": 1.870207897452651e-05, "loss": 0.0956, "step": 8248 }, { "epoch": 1.51, "learning_rate": 1.870171316662815e-05, "loss": 0.1044, "step": 8249 }, { "epoch": 1.51, "learning_rate": 1.8701347310765738e-05, "loss": 0.1009, "step": 8250 }, { "epoch": 1.51, "learning_rate": 1.87009814069413e-05, "loss": 0.1059, "step": 8251 }, { "epoch": 1.51, "learning_rate": 1.8700615455156847e-05, "loss": 0.0906, "step": 8252 }, { "epoch": 1.51, "learning_rate": 1.8700249455414394e-05, "loss": 0.105, "step": 8253 }, { "epoch": 1.51, "learning_rate": 1.8699883407715966e-05, "loss": 0.1086, "step": 8254 }, { "epoch": 1.51, "learning_rate": 1.8699517312063576e-05, "loss": 0.0813, "step": 8255 }, { "epoch": 1.51, "learning_rate": 1.8699151168459242e-05, "loss": 0.1049, "step": 8256 }, { "epoch": 1.51, "learning_rate": 1.8698784976904983e-05, "loss": 0.1014, "step": 8257 }, { "epoch": 1.51, "learning_rate": 1.8698418737402818e-05, "loss": 0.0967, "step": 8258 }, { "epoch": 1.51, "learning_rate": 1.869805244995477e-05, "loss": 0.1023, "step": 8259 }, { "epoch": 1.51, "learning_rate": 1.8697686114562847e-05, "loss": 0.0998, "step": 8260 }, { "epoch": 1.51, "learning_rate": 1.8697319731229076e-05, "loss": 0.1021, "step": 8261 }, { "epoch": 1.52, "learning_rate": 1.8696953299955473e-05, "loss": 0.0966, "step": 8262 }, { "epoch": 1.52, "learning_rate": 1.869658682074406e-05, "loss": 0.092, "step": 8263 }, { "epoch": 1.52, "learning_rate": 1.869622029359686e-05, "loss": 0.1035, "step": 8264 }, { "epoch": 1.52, "learning_rate": 1.8695853718515887e-05, "loss": 0.1006, "step": 8265 }, { "epoch": 1.52, "learning_rate": 1.869548709550317e-05, "loss": 0.103, "step": 8266 }, { "epoch": 1.52, "learning_rate": 1.8695120424560718e-05, "loss": 0.1073, "step": 8267 }, { "epoch": 1.52, "learning_rate": 1.869475370569056e-05, "loss": 0.0973, "step": 8268 }, { "epoch": 1.52, "learning_rate": 1.8694386938894717e-05, "loss": 0.1038, "step": 8269 }, { "epoch": 1.52, "learning_rate": 1.869402012417521e-05, "loss": 0.0904, "step": 8270 }, { "epoch": 1.52, "learning_rate": 1.869365326153406e-05, "loss": 0.1058, "step": 8271 }, { "epoch": 1.52, "learning_rate": 1.869328635097329e-05, "loss": 0.0938, "step": 8272 }, { "epoch": 1.52, "learning_rate": 1.869291939249492e-05, "loss": 0.0904, "step": 8273 }, { "epoch": 1.52, "learning_rate": 1.8692552386100975e-05, "loss": 0.0842, "step": 8274 }, { "epoch": 1.52, "learning_rate": 1.8692185331793477e-05, "loss": 0.0981, "step": 8275 }, { "epoch": 1.52, "learning_rate": 1.8691818229574452e-05, "loss": 0.0904, "step": 8276 }, { "epoch": 1.52, "learning_rate": 1.8691451079445922e-05, "loss": 0.0982, "step": 8277 }, { "epoch": 1.52, "learning_rate": 1.869108388140991e-05, "loss": 0.0902, "step": 8278 }, { "epoch": 1.52, "learning_rate": 1.8690716635468435e-05, "loss": 0.1109, "step": 8279 }, { "epoch": 1.52, "learning_rate": 1.869034934162353e-05, "loss": 0.0943, "step": 8280 }, { "epoch": 1.52, "learning_rate": 1.8689981999877217e-05, "loss": 0.0896, "step": 8281 }, { "epoch": 1.52, "learning_rate": 1.8689614610231517e-05, "loss": 0.0944, "step": 8282 }, { "epoch": 1.52, "learning_rate": 1.8689247172688462e-05, "loss": 0.0904, "step": 8283 }, { "epoch": 1.52, "learning_rate": 1.8688879687250067e-05, "loss": 0.1094, "step": 8284 }, { "epoch": 1.52, "learning_rate": 1.8688512153918367e-05, "loss": 0.1015, "step": 8285 }, { "epoch": 1.52, "learning_rate": 1.8688144572695383e-05, "loss": 0.0959, "step": 8286 }, { "epoch": 1.52, "learning_rate": 1.8687776943583145e-05, "loss": 0.0953, "step": 8287 }, { "epoch": 1.52, "learning_rate": 1.8687409266583672e-05, "loss": 0.1002, "step": 8288 }, { "epoch": 1.52, "learning_rate": 1.8687041541699e-05, "loss": 0.0981, "step": 8289 }, { "epoch": 1.52, "learning_rate": 1.8686673768931154e-05, "loss": 0.1023, "step": 8290 }, { "epoch": 1.52, "learning_rate": 1.8686305948282154e-05, "loss": 0.107, "step": 8291 }, { "epoch": 1.52, "learning_rate": 1.8685938079754035e-05, "loss": 0.0887, "step": 8292 }, { "epoch": 1.52, "learning_rate": 1.8685570163348817e-05, "loss": 0.1117, "step": 8293 }, { "epoch": 1.52, "learning_rate": 1.868520219906854e-05, "loss": 0.0898, "step": 8294 }, { "epoch": 1.52, "learning_rate": 1.868483418691522e-05, "loss": 0.0938, "step": 8295 }, { "epoch": 1.52, "learning_rate": 1.868446612689089e-05, "loss": 0.0909, "step": 8296 }, { "epoch": 1.52, "learning_rate": 1.868409801899758e-05, "loss": 0.1006, "step": 8297 }, { "epoch": 1.52, "learning_rate": 1.8683729863237322e-05, "loss": 0.0904, "step": 8298 }, { "epoch": 1.52, "learning_rate": 1.8683361659612137e-05, "loss": 0.0843, "step": 8299 }, { "epoch": 1.52, "learning_rate": 1.868299340812406e-05, "loss": 0.1033, "step": 8300 }, { "epoch": 1.52, "learning_rate": 1.8682625108775122e-05, "loss": 0.091, "step": 8301 }, { "epoch": 1.52, "learning_rate": 1.868225676156735e-05, "loss": 0.084, "step": 8302 }, { "epoch": 1.52, "learning_rate": 1.8681888366502773e-05, "loss": 0.0943, "step": 8303 }, { "epoch": 1.52, "learning_rate": 1.8681519923583427e-05, "loss": 0.0989, "step": 8304 }, { "epoch": 1.52, "learning_rate": 1.8681151432811338e-05, "loss": 0.0951, "step": 8305 }, { "epoch": 1.52, "learning_rate": 1.8680782894188542e-05, "loss": 0.1022, "step": 8306 }, { "epoch": 1.52, "learning_rate": 1.8680414307717064e-05, "loss": 0.0915, "step": 8307 }, { "epoch": 1.52, "learning_rate": 1.868004567339894e-05, "loss": 0.1033, "step": 8308 }, { "epoch": 1.52, "learning_rate": 1.86796769912362e-05, "loss": 0.1027, "step": 8309 }, { "epoch": 1.52, "learning_rate": 1.8679308261230878e-05, "loss": 0.0908, "step": 8310 }, { "epoch": 1.52, "learning_rate": 1.8678939483385005e-05, "loss": 0.1038, "step": 8311 }, { "epoch": 1.52, "learning_rate": 1.8678570657700615e-05, "loss": 0.0955, "step": 8312 }, { "epoch": 1.52, "learning_rate": 1.8678201784179738e-05, "loss": 0.089, "step": 8313 }, { "epoch": 1.52, "learning_rate": 1.8677832862824413e-05, "loss": 0.0914, "step": 8314 }, { "epoch": 1.52, "learning_rate": 1.8677463893636667e-05, "loss": 0.0953, "step": 8315 }, { "epoch": 1.53, "learning_rate": 1.867709487661854e-05, "loss": 0.0938, "step": 8316 }, { "epoch": 1.53, "learning_rate": 1.867672581177206e-05, "loss": 0.0969, "step": 8317 }, { "epoch": 1.53, "learning_rate": 1.8676356699099266e-05, "loss": 0.0979, "step": 8318 }, { "epoch": 1.53, "learning_rate": 1.8675987538602186e-05, "loss": 0.0876, "step": 8319 }, { "epoch": 1.53, "learning_rate": 1.8675618330282863e-05, "loss": 0.0975, "step": 8320 }, { "epoch": 1.53, "learning_rate": 1.8675249074143333e-05, "loss": 0.1033, "step": 8321 }, { "epoch": 1.53, "learning_rate": 1.8674879770185618e-05, "loss": 0.082, "step": 8322 }, { "epoch": 1.53, "learning_rate": 1.8674510418411768e-05, "loss": 0.101, "step": 8323 }, { "epoch": 1.53, "learning_rate": 1.8674141018823813e-05, "loss": 0.0906, "step": 8324 }, { "epoch": 1.53, "learning_rate": 1.8673771571423786e-05, "loss": 0.0932, "step": 8325 }, { "epoch": 1.53, "learning_rate": 1.8673402076213733e-05, "loss": 0.1006, "step": 8326 }, { "epoch": 1.53, "learning_rate": 1.867303253319568e-05, "loss": 0.0998, "step": 8327 }, { "epoch": 1.53, "learning_rate": 1.867266294237167e-05, "loss": 0.102, "step": 8328 }, { "epoch": 1.53, "learning_rate": 1.867229330374374e-05, "loss": 0.1005, "step": 8329 }, { "epoch": 1.53, "learning_rate": 1.8671923617313922e-05, "loss": 0.1026, "step": 8330 }, { "epoch": 1.53, "learning_rate": 1.867155388308426e-05, "loss": 0.0907, "step": 8331 }, { "epoch": 1.53, "learning_rate": 1.867118410105679e-05, "loss": 0.1022, "step": 8332 }, { "epoch": 1.53, "learning_rate": 1.867081427123355e-05, "loss": 0.0955, "step": 8333 }, { "epoch": 1.53, "learning_rate": 1.8670444393616577e-05, "loss": 0.1057, "step": 8334 }, { "epoch": 1.53, "learning_rate": 1.8670074468207916e-05, "loss": 0.0896, "step": 8335 }, { "epoch": 1.53, "learning_rate": 1.8669704495009598e-05, "loss": 0.0876, "step": 8336 }, { "epoch": 1.53, "learning_rate": 1.8669334474023663e-05, "loss": 0.092, "step": 8337 }, { "epoch": 1.53, "learning_rate": 1.8668964405252158e-05, "loss": 0.0938, "step": 8338 }, { "epoch": 1.53, "learning_rate": 1.8668594288697116e-05, "loss": 0.0955, "step": 8339 }, { "epoch": 1.53, "learning_rate": 1.8668224124360582e-05, "loss": 0.0879, "step": 8340 }, { "epoch": 1.53, "learning_rate": 1.866785391224459e-05, "loss": 0.1055, "step": 8341 }, { "epoch": 1.53, "learning_rate": 1.8667483652351187e-05, "loss": 0.0896, "step": 8342 }, { "epoch": 1.53, "learning_rate": 1.8667113344682408e-05, "loss": 0.0937, "step": 8343 }, { "epoch": 1.53, "learning_rate": 1.86667429892403e-05, "loss": 0.1066, "step": 8344 }, { "epoch": 1.53, "learning_rate": 1.86663725860269e-05, "loss": 0.0969, "step": 8345 }, { "epoch": 1.53, "learning_rate": 1.866600213504425e-05, "loss": 0.0936, "step": 8346 }, { "epoch": 1.53, "learning_rate": 1.8665631636294397e-05, "loss": 0.1163, "step": 8347 }, { "epoch": 1.53, "learning_rate": 1.8665261089779376e-05, "loss": 0.0968, "step": 8348 }, { "epoch": 1.53, "learning_rate": 1.8664890495501234e-05, "loss": 0.0864, "step": 8349 }, { "epoch": 1.53, "learning_rate": 1.8664519853462015e-05, "loss": 0.0989, "step": 8350 }, { "epoch": 1.53, "learning_rate": 1.8664149163663757e-05, "loss": 0.1103, "step": 8351 }, { "epoch": 1.53, "learning_rate": 1.8663778426108505e-05, "loss": 0.1005, "step": 8352 }, { "epoch": 1.53, "learning_rate": 1.8663407640798304e-05, "loss": 0.0924, "step": 8353 }, { "epoch": 1.53, "learning_rate": 1.86630368077352e-05, "loss": 0.09, "step": 8354 }, { "epoch": 1.53, "learning_rate": 1.866266592692123e-05, "loss": 0.0946, "step": 8355 }, { "epoch": 1.53, "learning_rate": 1.8662294998358442e-05, "loss": 0.0971, "step": 8356 }, { "epoch": 1.53, "learning_rate": 1.8661924022048885e-05, "loss": 0.096, "step": 8357 }, { "epoch": 1.53, "learning_rate": 1.8661552997994598e-05, "loss": 0.0859, "step": 8358 }, { "epoch": 1.53, "learning_rate": 1.8661181926197627e-05, "loss": 0.1002, "step": 8359 }, { "epoch": 1.53, "learning_rate": 1.866081080666002e-05, "loss": 0.1062, "step": 8360 }, { "epoch": 1.53, "learning_rate": 1.866043963938382e-05, "loss": 0.1139, "step": 8361 }, { "epoch": 1.53, "learning_rate": 1.8660068424371073e-05, "loss": 0.0947, "step": 8362 }, { "epoch": 1.53, "learning_rate": 1.865969716162383e-05, "loss": 0.1037, "step": 8363 }, { "epoch": 1.53, "learning_rate": 1.865932585114413e-05, "loss": 0.0931, "step": 8364 }, { "epoch": 1.53, "learning_rate": 1.865895449293402e-05, "loss": 0.0971, "step": 8365 }, { "epoch": 1.53, "learning_rate": 1.8658583086995557e-05, "loss": 0.0975, "step": 8366 }, { "epoch": 1.53, "learning_rate": 1.8658211633330777e-05, "loss": 0.0923, "step": 8367 }, { "epoch": 1.53, "learning_rate": 1.865784013194173e-05, "loss": 0.0984, "step": 8368 }, { "epoch": 1.53, "learning_rate": 1.865746858283047e-05, "loss": 0.0985, "step": 8369 }, { "epoch": 1.53, "learning_rate": 1.8657096985999036e-05, "loss": 0.1049, "step": 8370 }, { "epoch": 1.54, "learning_rate": 1.8656725341449483e-05, "loss": 0.0962, "step": 8371 }, { "epoch": 1.54, "learning_rate": 1.8656353649183857e-05, "loss": 0.1056, "step": 8372 }, { "epoch": 1.54, "learning_rate": 1.8655981909204207e-05, "loss": 0.0952, "step": 8373 }, { "epoch": 1.54, "learning_rate": 1.8655610121512584e-05, "loss": 0.093, "step": 8374 }, { "epoch": 1.54, "learning_rate": 1.865523828611103e-05, "loss": 0.0922, "step": 8375 }, { "epoch": 1.54, "learning_rate": 1.8654866403001604e-05, "loss": 0.0985, "step": 8376 }, { "epoch": 1.54, "learning_rate": 1.8654494472186352e-05, "loss": 0.107, "step": 8377 }, { "epoch": 1.54, "learning_rate": 1.865412249366732e-05, "loss": 0.0859, "step": 8378 }, { "epoch": 1.54, "learning_rate": 1.8653750467446564e-05, "loss": 0.1052, "step": 8379 }, { "epoch": 1.54, "learning_rate": 1.865337839352613e-05, "loss": 0.0899, "step": 8380 }, { "epoch": 1.54, "learning_rate": 1.8653006271908075e-05, "loss": 0.0932, "step": 8381 }, { "epoch": 1.54, "learning_rate": 1.8652634102594448e-05, "loss": 0.0827, "step": 8382 }, { "epoch": 1.54, "learning_rate": 1.8652261885587296e-05, "loss": 0.0908, "step": 8383 }, { "epoch": 1.54, "learning_rate": 1.8651889620888676e-05, "loss": 0.0965, "step": 8384 }, { "epoch": 1.54, "learning_rate": 1.8651517308500637e-05, "loss": 0.0875, "step": 8385 }, { "epoch": 1.54, "learning_rate": 1.865114494842523e-05, "loss": 0.0887, "step": 8386 }, { "epoch": 1.54, "learning_rate": 1.865077254066451e-05, "loss": 0.0932, "step": 8387 }, { "epoch": 1.54, "learning_rate": 1.865040008522053e-05, "loss": 0.085, "step": 8388 }, { "epoch": 1.54, "learning_rate": 1.8650027582095346e-05, "loss": 0.0887, "step": 8389 }, { "epoch": 1.54, "learning_rate": 1.8649655031291002e-05, "loss": 0.0931, "step": 8390 }, { "epoch": 1.54, "learning_rate": 1.8649282432809555e-05, "loss": 0.0846, "step": 8391 }, { "epoch": 1.54, "learning_rate": 1.8648909786653066e-05, "loss": 0.0933, "step": 8392 }, { "epoch": 1.54, "learning_rate": 1.864853709282358e-05, "loss": 0.0984, "step": 8393 }, { "epoch": 1.54, "learning_rate": 1.8648164351323158e-05, "loss": 0.0985, "step": 8394 }, { "epoch": 1.54, "learning_rate": 1.8647791562153846e-05, "loss": 0.0985, "step": 8395 }, { "epoch": 1.54, "learning_rate": 1.864741872531771e-05, "loss": 0.099, "step": 8396 }, { "epoch": 1.54, "learning_rate": 1.86470458408168e-05, "loss": 0.0983, "step": 8397 }, { "epoch": 1.54, "learning_rate": 1.8646672908653167e-05, "loss": 0.1016, "step": 8398 }, { "epoch": 1.54, "learning_rate": 1.864629992882887e-05, "loss": 0.1015, "step": 8399 }, { "epoch": 1.54, "learning_rate": 1.864592690134597e-05, "loss": 0.1006, "step": 8400 }, { "epoch": 1.54, "learning_rate": 1.8645553826206516e-05, "loss": 0.1034, "step": 8401 }, { "epoch": 1.54, "learning_rate": 1.8645180703412566e-05, "loss": 0.1016, "step": 8402 }, { "epoch": 1.54, "learning_rate": 1.8644807532966175e-05, "loss": 0.1001, "step": 8403 }, { "epoch": 1.54, "learning_rate": 1.8644434314869408e-05, "loss": 0.0937, "step": 8404 }, { "epoch": 1.54, "learning_rate": 1.8644061049124313e-05, "loss": 0.1116, "step": 8405 }, { "epoch": 1.54, "learning_rate": 1.8643687735732947e-05, "loss": 0.1067, "step": 8406 }, { "epoch": 1.54, "learning_rate": 1.8643314374697377e-05, "loss": 0.085, "step": 8407 }, { "epoch": 1.54, "learning_rate": 1.8642940966019653e-05, "loss": 0.0939, "step": 8408 }, { "epoch": 1.54, "learning_rate": 1.8642567509701842e-05, "loss": 0.0959, "step": 8409 }, { "epoch": 1.54, "learning_rate": 1.864219400574599e-05, "loss": 0.1013, "step": 8410 }, { "epoch": 1.54, "learning_rate": 1.8641820454154163e-05, "loss": 0.1043, "step": 8411 }, { "epoch": 1.54, "learning_rate": 1.8641446854928423e-05, "loss": 0.0992, "step": 8412 }, { "epoch": 1.54, "learning_rate": 1.864107320807082e-05, "loss": 0.0931, "step": 8413 }, { "epoch": 1.54, "learning_rate": 1.864069951358342e-05, "loss": 0.1104, "step": 8414 }, { "epoch": 1.54, "learning_rate": 1.8640325771468288e-05, "loss": 0.1003, "step": 8415 }, { "epoch": 1.54, "learning_rate": 1.8639951981727473e-05, "loss": 0.1043, "step": 8416 }, { "epoch": 1.54, "learning_rate": 1.863957814436304e-05, "loss": 0.0901, "step": 8417 }, { "epoch": 1.54, "learning_rate": 1.863920425937705e-05, "loss": 0.1071, "step": 8418 }, { "epoch": 1.54, "learning_rate": 1.8638830326771566e-05, "loss": 0.0978, "step": 8419 }, { "epoch": 1.54, "learning_rate": 1.8638456346548644e-05, "loss": 0.1031, "step": 8420 }, { "epoch": 1.54, "learning_rate": 1.863808231871035e-05, "loss": 0.0887, "step": 8421 }, { "epoch": 1.54, "learning_rate": 1.8637708243258746e-05, "loss": 0.1053, "step": 8422 }, { "epoch": 1.54, "learning_rate": 1.8637334120195888e-05, "loss": 0.098, "step": 8423 }, { "epoch": 1.54, "learning_rate": 1.863695994952384e-05, "loss": 0.0803, "step": 8424 }, { "epoch": 1.55, "learning_rate": 1.863658573124467e-05, "loss": 0.093, "step": 8425 }, { "epoch": 1.55, "learning_rate": 1.863621146536044e-05, "loss": 0.1024, "step": 8426 }, { "epoch": 1.55, "learning_rate": 1.8635837151873207e-05, "loss": 0.1022, "step": 8427 }, { "epoch": 1.55, "learning_rate": 1.8635462790785034e-05, "loss": 0.1078, "step": 8428 }, { "epoch": 1.55, "learning_rate": 1.8635088382097993e-05, "loss": 0.093, "step": 8429 }, { "epoch": 1.55, "learning_rate": 1.8634713925814142e-05, "loss": 0.0985, "step": 8430 }, { "epoch": 1.55, "learning_rate": 1.8634339421935543e-05, "loss": 0.097, "step": 8431 }, { "epoch": 1.55, "learning_rate": 1.8633964870464263e-05, "loss": 0.1026, "step": 8432 }, { "epoch": 1.55, "learning_rate": 1.863359027140237e-05, "loss": 0.0959, "step": 8433 }, { "epoch": 1.55, "learning_rate": 1.863321562475192e-05, "loss": 0.1038, "step": 8434 }, { "epoch": 1.55, "learning_rate": 1.8632840930514985e-05, "loss": 0.099, "step": 8435 }, { "epoch": 1.55, "learning_rate": 1.863246618869363e-05, "loss": 0.0929, "step": 8436 }, { "epoch": 1.55, "learning_rate": 1.863209139928992e-05, "loss": 0.1072, "step": 8437 }, { "epoch": 1.55, "learning_rate": 1.8631716562305916e-05, "loss": 0.1091, "step": 8438 }, { "epoch": 1.55, "learning_rate": 1.863134167774369e-05, "loss": 0.0885, "step": 8439 }, { "epoch": 1.55, "learning_rate": 1.8630966745605308e-05, "loss": 0.0883, "step": 8440 }, { "epoch": 1.55, "learning_rate": 1.8630591765892835e-05, "loss": 0.0847, "step": 8441 }, { "epoch": 1.55, "learning_rate": 1.8630216738608336e-05, "loss": 0.0966, "step": 8442 }, { "epoch": 1.55, "learning_rate": 1.8629841663753884e-05, "loss": 0.0919, "step": 8443 }, { "epoch": 1.55, "learning_rate": 1.8629466541331535e-05, "loss": 0.0875, "step": 8444 }, { "epoch": 1.55, "learning_rate": 1.862909137134337e-05, "loss": 0.0974, "step": 8445 }, { "epoch": 1.55, "learning_rate": 1.862871615379145e-05, "loss": 0.089, "step": 8446 }, { "epoch": 1.55, "learning_rate": 1.8628340888677846e-05, "loss": 0.103, "step": 8447 }, { "epoch": 1.55, "learning_rate": 1.8627965576004623e-05, "loss": 0.1054, "step": 8448 }, { "epoch": 1.55, "learning_rate": 1.862759021577385e-05, "loss": 0.1092, "step": 8449 }, { "epoch": 1.55, "learning_rate": 1.86272148079876e-05, "loss": 0.0938, "step": 8450 }, { "epoch": 1.55, "learning_rate": 1.862683935264794e-05, "loss": 0.1014, "step": 8451 }, { "epoch": 1.55, "learning_rate": 1.8626463849756938e-05, "loss": 0.1039, "step": 8452 }, { "epoch": 1.55, "learning_rate": 1.8626088299316668e-05, "loss": 0.088, "step": 8453 }, { "epoch": 1.55, "learning_rate": 1.8625712701329196e-05, "loss": 0.0989, "step": 8454 }, { "epoch": 1.55, "learning_rate": 1.8625337055796592e-05, "loss": 0.0702, "step": 8455 }, { "epoch": 1.55, "learning_rate": 1.862496136272093e-05, "loss": 0.0991, "step": 8456 }, { "epoch": 1.55, "learning_rate": 1.8624585622104277e-05, "loss": 0.0981, "step": 8457 }, { "epoch": 1.55, "learning_rate": 1.862420983394871e-05, "loss": 0.1057, "step": 8458 }, { "epoch": 1.55, "learning_rate": 1.862383399825629e-05, "loss": 0.091, "step": 8459 }, { "epoch": 1.55, "learning_rate": 1.8623458115029098e-05, "loss": 0.1051, "step": 8460 }, { "epoch": 1.55, "learning_rate": 1.8623082184269204e-05, "loss": 0.0884, "step": 8461 }, { "epoch": 1.55, "learning_rate": 1.8622706205978677e-05, "loss": 0.0915, "step": 8462 }, { "epoch": 1.55, "learning_rate": 1.862233018015959e-05, "loss": 0.1025, "step": 8463 }, { "epoch": 1.55, "learning_rate": 1.8621954106814022e-05, "loss": 0.0973, "step": 8464 }, { "epoch": 1.55, "learning_rate": 1.862157798594404e-05, "loss": 0.1068, "step": 8465 }, { "epoch": 1.55, "learning_rate": 1.862120181755171e-05, "loss": 0.0971, "step": 8466 }, { "epoch": 1.55, "learning_rate": 1.862082560163912e-05, "loss": 0.0844, "step": 8467 }, { "epoch": 1.55, "learning_rate": 1.8620449338208343e-05, "loss": 0.1067, "step": 8468 }, { "epoch": 1.55, "learning_rate": 1.862007302726144e-05, "loss": 0.0845, "step": 8469 }, { "epoch": 1.55, "learning_rate": 1.8619696668800494e-05, "loss": 0.1037, "step": 8470 }, { "epoch": 1.55, "learning_rate": 1.861932026282758e-05, "loss": 0.0991, "step": 8471 }, { "epoch": 1.55, "learning_rate": 1.861894380934477e-05, "loss": 0.0891, "step": 8472 }, { "epoch": 1.55, "learning_rate": 1.8618567308354136e-05, "loss": 0.1002, "step": 8473 }, { "epoch": 1.55, "learning_rate": 1.8618190759857763e-05, "loss": 0.0849, "step": 8474 }, { "epoch": 1.55, "learning_rate": 1.8617814163857715e-05, "loss": 0.0962, "step": 8475 }, { "epoch": 1.55, "learning_rate": 1.8617437520356074e-05, "loss": 0.099, "step": 8476 }, { "epoch": 1.55, "learning_rate": 1.8617060829354918e-05, "loss": 0.0948, "step": 8477 }, { "epoch": 1.55, "learning_rate": 1.8616684090856322e-05, "loss": 0.0995, "step": 8478 }, { "epoch": 1.55, "learning_rate": 1.861630730486236e-05, "loss": 0.1118, "step": 8479 }, { "epoch": 1.56, "learning_rate": 1.861593047137511e-05, "loss": 0.096, "step": 8480 }, { "epoch": 1.56, "learning_rate": 1.8615553590396648e-05, "loss": 0.0924, "step": 8481 }, { "epoch": 1.56, "learning_rate": 1.8615176661929054e-05, "loss": 0.0889, "step": 8482 }, { "epoch": 1.56, "learning_rate": 1.8614799685974403e-05, "loss": 0.092, "step": 8483 }, { "epoch": 1.56, "learning_rate": 1.8614422662534778e-05, "loss": 0.0934, "step": 8484 }, { "epoch": 1.56, "learning_rate": 1.861404559161225e-05, "loss": 0.0929, "step": 8485 }, { "epoch": 1.56, "learning_rate": 1.8613668473208905e-05, "loss": 0.0873, "step": 8486 }, { "epoch": 1.56, "learning_rate": 1.861329130732681e-05, "loss": 0.1009, "step": 8487 }, { "epoch": 1.56, "learning_rate": 1.861291409396806e-05, "loss": 0.0961, "step": 8488 }, { "epoch": 1.56, "learning_rate": 1.8612536833134722e-05, "loss": 0.0915, "step": 8489 }, { "epoch": 1.56, "learning_rate": 1.8612159524828877e-05, "loss": 0.0975, "step": 8490 }, { "epoch": 1.56, "learning_rate": 1.8611782169052612e-05, "loss": 0.0851, "step": 8491 }, { "epoch": 1.56, "learning_rate": 1.8611404765807997e-05, "loss": 0.0958, "step": 8492 }, { "epoch": 1.56, "learning_rate": 1.861102731509712e-05, "loss": 0.0971, "step": 8493 }, { "epoch": 1.56, "learning_rate": 1.8610649816922056e-05, "loss": 0.0899, "step": 8494 }, { "epoch": 1.56, "learning_rate": 1.8610272271284894e-05, "loss": 0.0991, "step": 8495 }, { "epoch": 1.56, "learning_rate": 1.8609894678187707e-05, "loss": 0.0974, "step": 8496 }, { "epoch": 1.56, "learning_rate": 1.8609517037632575e-05, "loss": 0.1038, "step": 8497 }, { "epoch": 1.56, "learning_rate": 1.8609139349621588e-05, "loss": 0.1114, "step": 8498 }, { "epoch": 1.56, "learning_rate": 1.860876161415682e-05, "loss": 0.0873, "step": 8499 }, { "epoch": 1.56, "learning_rate": 1.860838383124036e-05, "loss": 0.0883, "step": 8500 }, { "epoch": 1.56, "learning_rate": 1.8608006000874282e-05, "loss": 0.0968, "step": 8501 }, { "epoch": 1.56, "learning_rate": 1.8607628123060678e-05, "loss": 0.0984, "step": 8502 }, { "epoch": 1.56, "learning_rate": 1.8607250197801625e-05, "loss": 0.0872, "step": 8503 }, { "epoch": 1.56, "learning_rate": 1.8606872225099205e-05, "loss": 0.0924, "step": 8504 }, { "epoch": 1.56, "learning_rate": 1.8606494204955504e-05, "loss": 0.0998, "step": 8505 }, { "epoch": 1.56, "learning_rate": 1.8606116137372605e-05, "loss": 0.1024, "step": 8506 }, { "epoch": 1.56, "learning_rate": 1.8605738022352594e-05, "loss": 0.1031, "step": 8507 }, { "epoch": 1.56, "learning_rate": 1.8605359859897552e-05, "loss": 0.0881, "step": 8508 }, { "epoch": 1.56, "learning_rate": 1.8604981650009567e-05, "loss": 0.0992, "step": 8509 }, { "epoch": 1.56, "learning_rate": 1.8604603392690718e-05, "loss": 0.0911, "step": 8510 }, { "epoch": 1.56, "learning_rate": 1.8604225087943096e-05, "loss": 0.0904, "step": 8511 }, { "epoch": 1.56, "learning_rate": 1.8603846735768785e-05, "loss": 0.1047, "step": 8512 }, { "epoch": 1.56, "learning_rate": 1.8603468336169865e-05, "loss": 0.1104, "step": 8513 }, { "epoch": 1.56, "learning_rate": 1.860308988914843e-05, "loss": 0.097, "step": 8514 }, { "epoch": 1.56, "learning_rate": 1.860271139470656e-05, "loss": 0.0922, "step": 8515 }, { "epoch": 1.56, "learning_rate": 1.860233285284634e-05, "loss": 0.1031, "step": 8516 }, { "epoch": 1.56, "learning_rate": 1.8601954263569867e-05, "loss": 0.1127, "step": 8517 }, { "epoch": 1.56, "learning_rate": 1.8601575626879218e-05, "loss": 0.0974, "step": 8518 }, { "epoch": 1.56, "learning_rate": 1.860119694277648e-05, "loss": 0.0976, "step": 8519 }, { "epoch": 1.56, "learning_rate": 1.8600818211263748e-05, "loss": 0.093, "step": 8520 }, { "epoch": 1.56, "learning_rate": 1.86004394323431e-05, "loss": 0.0953, "step": 8521 }, { "epoch": 1.56, "learning_rate": 1.860006060601663e-05, "loss": 0.0982, "step": 8522 }, { "epoch": 1.56, "learning_rate": 1.8599681732286423e-05, "loss": 0.1004, "step": 8523 }, { "epoch": 1.56, "learning_rate": 1.8599302811154572e-05, "loss": 0.1064, "step": 8524 }, { "epoch": 1.56, "learning_rate": 1.8598923842623157e-05, "loss": 0.0972, "step": 8525 }, { "epoch": 1.56, "learning_rate": 1.859854482669428e-05, "loss": 0.1015, "step": 8526 }, { "epoch": 1.56, "learning_rate": 1.8598165763370018e-05, "loss": 0.089, "step": 8527 }, { "epoch": 1.56, "learning_rate": 1.8597786652652464e-05, "loss": 0.0989, "step": 8528 }, { "epoch": 1.56, "learning_rate": 1.8597407494543708e-05, "loss": 0.101, "step": 8529 }, { "epoch": 1.56, "learning_rate": 1.859702828904584e-05, "loss": 0.0943, "step": 8530 }, { "epoch": 1.56, "learning_rate": 1.8596649036160953e-05, "loss": 0.0894, "step": 8531 }, { "epoch": 1.56, "learning_rate": 1.8596269735891137e-05, "loss": 0.1013, "step": 8532 }, { "epoch": 1.56, "learning_rate": 1.859589038823848e-05, "loss": 0.1025, "step": 8533 }, { "epoch": 1.57, "learning_rate": 1.859551099320507e-05, "loss": 0.1031, "step": 8534 }, { "epoch": 1.57, "learning_rate": 1.8595131550793005e-05, "loss": 0.1094, "step": 8535 }, { "epoch": 1.57, "learning_rate": 1.8594752061004376e-05, "loss": 0.0868, "step": 8536 }, { "epoch": 1.57, "learning_rate": 1.8594372523841268e-05, "loss": 0.0882, "step": 8537 }, { "epoch": 1.57, "learning_rate": 1.8593992939305776e-05, "loss": 0.0958, "step": 8538 }, { "epoch": 1.57, "learning_rate": 1.85936133074e-05, "loss": 0.0902, "step": 8539 }, { "epoch": 1.57, "learning_rate": 1.859323362812602e-05, "loss": 0.1015, "step": 8540 }, { "epoch": 1.57, "learning_rate": 1.8592853901485935e-05, "loss": 0.0931, "step": 8541 }, { "epoch": 1.57, "learning_rate": 1.859247412748184e-05, "loss": 0.0978, "step": 8542 }, { "epoch": 1.57, "learning_rate": 1.8592094306115828e-05, "loss": 0.0944, "step": 8543 }, { "epoch": 1.57, "learning_rate": 1.859171443738999e-05, "loss": 0.1116, "step": 8544 }, { "epoch": 1.57, "learning_rate": 1.859133452130642e-05, "loss": 0.0931, "step": 8545 }, { "epoch": 1.57, "learning_rate": 1.8590954557867215e-05, "loss": 0.0982, "step": 8546 }, { "epoch": 1.57, "learning_rate": 1.8590574547074464e-05, "loss": 0.0931, "step": 8547 }, { "epoch": 1.57, "learning_rate": 1.8590194488930264e-05, "loss": 0.1061, "step": 8548 }, { "epoch": 1.57, "learning_rate": 1.8589814383436712e-05, "loss": 0.099, "step": 8549 }, { "epoch": 1.57, "learning_rate": 1.8589434230595906e-05, "loss": 0.0996, "step": 8550 }, { "epoch": 1.57, "learning_rate": 1.858905403040993e-05, "loss": 0.0866, "step": 8551 }, { "epoch": 1.57, "learning_rate": 1.8588673782880894e-05, "loss": 0.0947, "step": 8552 }, { "epoch": 1.57, "learning_rate": 1.858829348801088e-05, "loss": 0.0881, "step": 8553 }, { "epoch": 1.57, "learning_rate": 1.8587913145801998e-05, "loss": 0.0979, "step": 8554 }, { "epoch": 1.57, "learning_rate": 1.8587532756256333e-05, "loss": 0.0781, "step": 8555 }, { "epoch": 1.57, "learning_rate": 1.858715231937599e-05, "loss": 0.0866, "step": 8556 }, { "epoch": 1.57, "learning_rate": 1.858677183516306e-05, "loss": 0.1003, "step": 8557 }, { "epoch": 1.57, "learning_rate": 1.8586391303619642e-05, "loss": 0.0969, "step": 8558 }, { "epoch": 1.57, "learning_rate": 1.8586010724747833e-05, "loss": 0.1001, "step": 8559 }, { "epoch": 1.57, "learning_rate": 1.8585630098549735e-05, "loss": 0.1011, "step": 8560 }, { "epoch": 1.57, "learning_rate": 1.8585249425027442e-05, "loss": 0.1019, "step": 8561 }, { "epoch": 1.57, "learning_rate": 1.858486870418305e-05, "loss": 0.1009, "step": 8562 }, { "epoch": 1.57, "learning_rate": 1.8584487936018663e-05, "loss": 0.1036, "step": 8563 }, { "epoch": 1.57, "learning_rate": 1.8584107120536376e-05, "loss": 0.1007, "step": 8564 }, { "epoch": 1.57, "learning_rate": 1.858372625773829e-05, "loss": 0.0963, "step": 8565 }, { "epoch": 1.57, "learning_rate": 1.8583345347626504e-05, "loss": 0.1009, "step": 8566 }, { "epoch": 1.57, "learning_rate": 1.858296439020312e-05, "loss": 0.0936, "step": 8567 }, { "epoch": 1.57, "learning_rate": 1.858258338547023e-05, "loss": 0.1041, "step": 8568 }, { "epoch": 1.57, "learning_rate": 1.8582202333429944e-05, "loss": 0.106, "step": 8569 }, { "epoch": 1.57, "learning_rate": 1.8581821234084355e-05, "loss": 0.106, "step": 8570 }, { "epoch": 1.57, "learning_rate": 1.8581440087435566e-05, "loss": 0.1102, "step": 8571 }, { "epoch": 1.57, "learning_rate": 1.858105889348568e-05, "loss": 0.0932, "step": 8572 }, { "epoch": 1.57, "learning_rate": 1.8580677652236796e-05, "loss": 0.0959, "step": 8573 }, { "epoch": 1.57, "learning_rate": 1.8580296363691016e-05, "loss": 0.0973, "step": 8574 }, { "epoch": 1.57, "learning_rate": 1.857991502785044e-05, "loss": 0.1019, "step": 8575 }, { "epoch": 1.57, "learning_rate": 1.857953364471717e-05, "loss": 0.1057, "step": 8576 }, { "epoch": 1.57, "learning_rate": 1.8579152214293314e-05, "loss": 0.0969, "step": 8577 }, { "epoch": 1.57, "learning_rate": 1.8578770736580966e-05, "loss": 0.0927, "step": 8578 }, { "epoch": 1.57, "learning_rate": 1.857838921158224e-05, "loss": 0.1014, "step": 8579 }, { "epoch": 1.57, "learning_rate": 1.857800763929922e-05, "loss": 0.0935, "step": 8580 }, { "epoch": 1.57, "learning_rate": 1.857762601973403e-05, "loss": 0.0985, "step": 8581 }, { "epoch": 1.57, "learning_rate": 1.857724435288876e-05, "loss": 0.0953, "step": 8582 }, { "epoch": 1.57, "learning_rate": 1.857686263876552e-05, "loss": 0.1004, "step": 8583 }, { "epoch": 1.57, "learning_rate": 1.857648087736641e-05, "loss": 0.0934, "step": 8584 }, { "epoch": 1.57, "learning_rate": 1.857609906869354e-05, "loss": 0.1026, "step": 8585 }, { "epoch": 1.57, "learning_rate": 1.857571721274901e-05, "loss": 0.0964, "step": 8586 }, { "epoch": 1.57, "learning_rate": 1.8575335309534923e-05, "loss": 0.1001, "step": 8587 }, { "epoch": 1.57, "learning_rate": 1.857495335905339e-05, "loss": 0.0953, "step": 8588 }, { "epoch": 1.58, "learning_rate": 1.857457136130651e-05, "loss": 0.1055, "step": 8589 }, { "epoch": 1.58, "learning_rate": 1.8574189316296396e-05, "loss": 0.0963, "step": 8590 }, { "epoch": 1.58, "learning_rate": 1.8573807224025148e-05, "loss": 0.1007, "step": 8591 }, { "epoch": 1.58, "learning_rate": 1.8573425084494873e-05, "loss": 0.0902, "step": 8592 }, { "epoch": 1.58, "learning_rate": 1.8573042897707677e-05, "loss": 0.1067, "step": 8593 }, { "epoch": 1.58, "learning_rate": 1.8572660663665672e-05, "loss": 0.0949, "step": 8594 }, { "epoch": 1.58, "learning_rate": 1.8572278382370955e-05, "loss": 0.1135, "step": 8595 }, { "epoch": 1.58, "learning_rate": 1.8571896053825644e-05, "loss": 0.0911, "step": 8596 }, { "epoch": 1.58, "learning_rate": 1.8571513678031835e-05, "loss": 0.0947, "step": 8597 }, { "epoch": 1.58, "learning_rate": 1.8571131254991646e-05, "loss": 0.0965, "step": 8598 }, { "epoch": 1.58, "learning_rate": 1.857074878470718e-05, "loss": 0.093, "step": 8599 }, { "epoch": 1.58, "learning_rate": 1.857036626718055e-05, "loss": 0.095, "step": 8600 }, { "epoch": 1.58, "learning_rate": 1.8569983702413852e-05, "loss": 0.0976, "step": 8601 }, { "epoch": 1.58, "learning_rate": 1.856960109040921e-05, "loss": 0.1027, "step": 8602 }, { "epoch": 1.58, "learning_rate": 1.8569218431168723e-05, "loss": 0.0876, "step": 8603 }, { "epoch": 1.58, "learning_rate": 1.85688357246945e-05, "loss": 0.0974, "step": 8604 }, { "epoch": 1.58, "learning_rate": 1.856845297098866e-05, "loss": 0.0943, "step": 8605 }, { "epoch": 1.58, "learning_rate": 1.8568070170053306e-05, "loss": 0.1016, "step": 8606 }, { "epoch": 1.58, "learning_rate": 1.8567687321890544e-05, "loss": 0.1012, "step": 8607 }, { "epoch": 1.58, "learning_rate": 1.856730442650249e-05, "loss": 0.1093, "step": 8608 }, { "epoch": 1.58, "learning_rate": 1.8566921483891255e-05, "loss": 0.101, "step": 8609 }, { "epoch": 1.58, "learning_rate": 1.856653849405895e-05, "loss": 0.0939, "step": 8610 }, { "epoch": 1.58, "learning_rate": 1.856615545700768e-05, "loss": 0.0996, "step": 8611 }, { "epoch": 1.58, "learning_rate": 1.8565772372739563e-05, "loss": 0.0954, "step": 8612 }, { "epoch": 1.58, "learning_rate": 1.8565389241256705e-05, "loss": 0.1014, "step": 8613 }, { "epoch": 1.58, "learning_rate": 1.8565006062561222e-05, "loss": 0.0949, "step": 8614 }, { "epoch": 1.58, "learning_rate": 1.8564622836655227e-05, "loss": 0.0995, "step": 8615 }, { "epoch": 1.58, "learning_rate": 1.856423956354083e-05, "loss": 0.1032, "step": 8616 }, { "epoch": 1.58, "learning_rate": 1.856385624322014e-05, "loss": 0.0945, "step": 8617 }, { "epoch": 1.58, "learning_rate": 1.8563472875695278e-05, "loss": 0.1088, "step": 8618 }, { "epoch": 1.58, "learning_rate": 1.8563089460968352e-05, "loss": 0.101, "step": 8619 }, { "epoch": 1.58, "learning_rate": 1.8562705999041478e-05, "loss": 0.0916, "step": 8620 }, { "epoch": 1.58, "learning_rate": 1.8562322489916766e-05, "loss": 0.0981, "step": 8621 }, { "epoch": 1.58, "learning_rate": 1.856193893359633e-05, "loss": 0.0865, "step": 8622 }, { "epoch": 1.58, "learning_rate": 1.8561555330082287e-05, "loss": 0.1011, "step": 8623 }, { "epoch": 1.58, "learning_rate": 1.8561171679376753e-05, "loss": 0.1038, "step": 8624 }, { "epoch": 1.58, "learning_rate": 1.8560787981481838e-05, "loss": 0.0979, "step": 8625 }, { "epoch": 1.58, "learning_rate": 1.856040423639966e-05, "loss": 0.0967, "step": 8626 }, { "epoch": 1.58, "learning_rate": 1.856002044413233e-05, "loss": 0.0958, "step": 8627 }, { "epoch": 1.58, "learning_rate": 1.8559636604681974e-05, "loss": 0.1061, "step": 8628 }, { "epoch": 1.58, "learning_rate": 1.8559252718050693e-05, "loss": 0.0942, "step": 8629 }, { "epoch": 1.58, "learning_rate": 1.855886878424061e-05, "loss": 0.1001, "step": 8630 }, { "epoch": 1.58, "learning_rate": 1.8558484803253848e-05, "loss": 0.0871, "step": 8631 }, { "epoch": 1.58, "learning_rate": 1.8558100775092516e-05, "loss": 0.0918, "step": 8632 }, { "epoch": 1.58, "learning_rate": 1.855771669975873e-05, "loss": 0.1034, "step": 8633 }, { "epoch": 1.58, "learning_rate": 1.855733257725461e-05, "loss": 0.0981, "step": 8634 }, { "epoch": 1.58, "learning_rate": 1.855694840758227e-05, "loss": 0.095, "step": 8635 }, { "epoch": 1.58, "learning_rate": 1.855656419074383e-05, "loss": 0.0911, "step": 8636 }, { "epoch": 1.58, "learning_rate": 1.855617992674141e-05, "loss": 0.1035, "step": 8637 }, { "epoch": 1.58, "learning_rate": 1.8555795615577124e-05, "loss": 0.074, "step": 8638 }, { "epoch": 1.58, "learning_rate": 1.8555411257253097e-05, "loss": 0.1156, "step": 8639 }, { "epoch": 1.58, "learning_rate": 1.8555026851771438e-05, "loss": 0.0895, "step": 8640 }, { "epoch": 1.58, "learning_rate": 1.855464239913427e-05, "loss": 0.0945, "step": 8641 }, { "epoch": 1.58, "learning_rate": 1.855425789934372e-05, "loss": 0.0963, "step": 8642 }, { "epoch": 1.58, "learning_rate": 1.855387335240189e-05, "loss": 0.113, "step": 8643 }, { "epoch": 1.59, "learning_rate": 1.8553488758310917e-05, "loss": 0.083, "step": 8644 }, { "epoch": 1.59, "learning_rate": 1.8553104117072907e-05, "loss": 0.0993, "step": 8645 }, { "epoch": 1.59, "learning_rate": 1.8552719428689993e-05, "loss": 0.1053, "step": 8646 }, { "epoch": 1.59, "learning_rate": 1.8552334693164288e-05, "loss": 0.0965, "step": 8647 }, { "epoch": 1.59, "learning_rate": 1.855194991049791e-05, "loss": 0.1019, "step": 8648 }, { "epoch": 1.59, "learning_rate": 1.8551565080692985e-05, "loss": 0.1051, "step": 8649 }, { "epoch": 1.59, "learning_rate": 1.8551180203751636e-05, "loss": 0.0986, "step": 8650 }, { "epoch": 1.59, "learning_rate": 1.8550795279675976e-05, "loss": 0.0956, "step": 8651 }, { "epoch": 1.59, "learning_rate": 1.8550410308468137e-05, "loss": 0.0904, "step": 8652 }, { "epoch": 1.59, "learning_rate": 1.8550025290130237e-05, "loss": 0.0929, "step": 8653 }, { "epoch": 1.59, "learning_rate": 1.8549640224664392e-05, "loss": 0.0921, "step": 8654 }, { "epoch": 1.59, "learning_rate": 1.8549255112072734e-05, "loss": 0.1032, "step": 8655 }, { "epoch": 1.59, "learning_rate": 1.854886995235738e-05, "loss": 0.1102, "step": 8656 }, { "epoch": 1.59, "learning_rate": 1.8548484745520453e-05, "loss": 0.0909, "step": 8657 }, { "epoch": 1.59, "learning_rate": 1.8548099491564082e-05, "loss": 0.0841, "step": 8658 }, { "epoch": 1.59, "learning_rate": 1.8547714190490385e-05, "loss": 0.1077, "step": 8659 }, { "epoch": 1.59, "learning_rate": 1.8547328842301487e-05, "loss": 0.0966, "step": 8660 }, { "epoch": 1.59, "learning_rate": 1.8546943446999508e-05, "loss": 0.0902, "step": 8661 }, { "epoch": 1.59, "learning_rate": 1.854655800458658e-05, "loss": 0.0858, "step": 8662 }, { "epoch": 1.59, "learning_rate": 1.8546172515064827e-05, "loss": 0.091, "step": 8663 }, { "epoch": 1.59, "learning_rate": 1.8545786978436366e-05, "loss": 0.0791, "step": 8664 }, { "epoch": 1.59, "learning_rate": 1.854540139470333e-05, "loss": 0.1102, "step": 8665 }, { "epoch": 1.59, "learning_rate": 1.8545015763867838e-05, "loss": 0.0967, "step": 8666 }, { "epoch": 1.59, "learning_rate": 1.8544630085932022e-05, "loss": 0.0979, "step": 8667 }, { "epoch": 1.59, "learning_rate": 1.8544244360898003e-05, "loss": 0.0953, "step": 8668 }, { "epoch": 1.59, "learning_rate": 1.8543858588767908e-05, "loss": 0.1093, "step": 8669 }, { "epoch": 1.59, "learning_rate": 1.854347276954387e-05, "loss": 0.0874, "step": 8670 }, { "epoch": 1.59, "learning_rate": 1.8543086903228003e-05, "loss": 0.0834, "step": 8671 }, { "epoch": 1.59, "learning_rate": 1.8542700989822446e-05, "loss": 0.0951, "step": 8672 }, { "epoch": 1.59, "learning_rate": 1.8542315029329317e-05, "loss": 0.0812, "step": 8673 }, { "epoch": 1.59, "learning_rate": 1.8541929021750748e-05, "loss": 0.0937, "step": 8674 }, { "epoch": 1.59, "learning_rate": 1.8541542967088866e-05, "loss": 0.0828, "step": 8675 }, { "epoch": 1.59, "learning_rate": 1.85411568653458e-05, "loss": 0.0865, "step": 8676 }, { "epoch": 1.59, "learning_rate": 1.8540770716523678e-05, "loss": 0.0984, "step": 8677 }, { "epoch": 1.59, "learning_rate": 1.854038452062463e-05, "loss": 0.1063, "step": 8678 }, { "epoch": 1.59, "learning_rate": 1.853999827765078e-05, "loss": 0.0994, "step": 8679 }, { "epoch": 1.59, "learning_rate": 1.853961198760426e-05, "loss": 0.1024, "step": 8680 }, { "epoch": 1.59, "learning_rate": 1.8539225650487197e-05, "loss": 0.0917, "step": 8681 }, { "epoch": 1.59, "learning_rate": 1.853883926630172e-05, "loss": 0.0985, "step": 8682 }, { "epoch": 1.59, "learning_rate": 1.8538452835049966e-05, "loss": 0.0869, "step": 8683 }, { "epoch": 1.59, "learning_rate": 1.8538066356734056e-05, "loss": 0.1039, "step": 8684 }, { "epoch": 1.59, "learning_rate": 1.8537679831356125e-05, "loss": 0.0818, "step": 8685 }, { "epoch": 1.59, "learning_rate": 1.8537293258918304e-05, "loss": 0.0918, "step": 8686 }, { "epoch": 1.59, "learning_rate": 1.8536906639422724e-05, "loss": 0.089, "step": 8687 }, { "epoch": 1.59, "learning_rate": 1.8536519972871513e-05, "loss": 0.1066, "step": 8688 }, { "epoch": 1.59, "learning_rate": 1.8536133259266804e-05, "loss": 0.0954, "step": 8689 }, { "epoch": 1.59, "learning_rate": 1.8535746498610727e-05, "loss": 0.0905, "step": 8690 }, { "epoch": 1.59, "learning_rate": 1.8535359690905415e-05, "loss": 0.0957, "step": 8691 }, { "epoch": 1.59, "learning_rate": 1.8534972836153004e-05, "loss": 0.0896, "step": 8692 }, { "epoch": 1.59, "learning_rate": 1.8534585934355623e-05, "loss": 0.1039, "step": 8693 }, { "epoch": 1.59, "learning_rate": 1.85341989855154e-05, "loss": 0.099, "step": 8694 }, { "epoch": 1.59, "learning_rate": 1.8533811989634473e-05, "loss": 0.0923, "step": 8695 }, { "epoch": 1.59, "learning_rate": 1.8533424946714976e-05, "loss": 0.0957, "step": 8696 }, { "epoch": 1.59, "learning_rate": 1.8533037856759045e-05, "loss": 0.0889, "step": 8697 }, { "epoch": 1.6, "learning_rate": 1.8532650719768804e-05, "loss": 0.1045, "step": 8698 }, { "epoch": 1.6, "learning_rate": 1.8532263535746392e-05, "loss": 0.1028, "step": 8699 }, { "epoch": 1.6, "learning_rate": 1.8531876304693945e-05, "loss": 0.1008, "step": 8700 }, { "epoch": 1.6, "learning_rate": 1.8531489026613596e-05, "loss": 0.0934, "step": 8701 }, { "epoch": 1.6, "learning_rate": 1.8531101701507478e-05, "loss": 0.0953, "step": 8702 }, { "epoch": 1.6, "learning_rate": 1.853071432937773e-05, "loss": 0.087, "step": 8703 }, { "epoch": 1.6, "learning_rate": 1.8530326910226483e-05, "loss": 0.1055, "step": 8704 }, { "epoch": 1.6, "learning_rate": 1.8529939444055878e-05, "loss": 0.0877, "step": 8705 }, { "epoch": 1.6, "learning_rate": 1.8529551930868042e-05, "loss": 0.0884, "step": 8706 }, { "epoch": 1.6, "learning_rate": 1.852916437066512e-05, "loss": 0.0954, "step": 8707 }, { "epoch": 1.6, "learning_rate": 1.8528776763449243e-05, "loss": 0.0888, "step": 8708 }, { "epoch": 1.6, "learning_rate": 1.8528389109222544e-05, "loss": 0.0956, "step": 8709 }, { "epoch": 1.6, "learning_rate": 1.852800140798717e-05, "loss": 0.0971, "step": 8710 }, { "epoch": 1.6, "learning_rate": 1.8527613659745253e-05, "loss": 0.0966, "step": 8711 }, { "epoch": 1.6, "learning_rate": 1.8527225864498922e-05, "loss": 0.0895, "step": 8712 }, { "epoch": 1.6, "learning_rate": 1.852683802225033e-05, "loss": 0.1004, "step": 8713 }, { "epoch": 1.6, "learning_rate": 1.8526450133001604e-05, "loss": 0.0979, "step": 8714 }, { "epoch": 1.6, "learning_rate": 1.852606219675489e-05, "loss": 0.0897, "step": 8715 }, { "epoch": 1.6, "learning_rate": 1.8525674213512313e-05, "loss": 0.0928, "step": 8716 }, { "epoch": 1.6, "learning_rate": 1.8525286183276024e-05, "loss": 0.0849, "step": 8717 }, { "epoch": 1.6, "learning_rate": 1.852489810604816e-05, "loss": 0.0995, "step": 8718 }, { "epoch": 1.6, "learning_rate": 1.8524509981830855e-05, "loss": 0.0932, "step": 8719 }, { "epoch": 1.6, "learning_rate": 1.8524121810626253e-05, "loss": 0.0969, "step": 8720 }, { "epoch": 1.6, "learning_rate": 1.852373359243649e-05, "loss": 0.0968, "step": 8721 }, { "epoch": 1.6, "learning_rate": 1.8523345327263708e-05, "loss": 0.0932, "step": 8722 }, { "epoch": 1.6, "learning_rate": 1.852295701511005e-05, "loss": 0.1018, "step": 8723 }, { "epoch": 1.6, "learning_rate": 1.852256865597765e-05, "loss": 0.1074, "step": 8724 }, { "epoch": 1.6, "learning_rate": 1.8522180249868654e-05, "loss": 0.0925, "step": 8725 }, { "epoch": 1.6, "learning_rate": 1.85217917967852e-05, "loss": 0.0988, "step": 8726 }, { "epoch": 1.6, "learning_rate": 1.8521403296729428e-05, "loss": 0.0857, "step": 8727 }, { "epoch": 1.6, "learning_rate": 1.8521014749703484e-05, "loss": 0.0989, "step": 8728 }, { "epoch": 1.6, "learning_rate": 1.8520626155709506e-05, "loss": 0.0959, "step": 8729 }, { "epoch": 1.6, "learning_rate": 1.8520237514749637e-05, "loss": 0.1053, "step": 8730 }, { "epoch": 1.6, "learning_rate": 1.851984882682602e-05, "loss": 0.0967, "step": 8731 }, { "epoch": 1.6, "learning_rate": 1.8519460091940797e-05, "loss": 0.0892, "step": 8732 }, { "epoch": 1.6, "learning_rate": 1.8519071310096107e-05, "loss": 0.099, "step": 8733 }, { "epoch": 1.6, "learning_rate": 1.85186824812941e-05, "loss": 0.094, "step": 8734 }, { "epoch": 1.6, "learning_rate": 1.8518293605536914e-05, "loss": 0.0956, "step": 8735 }, { "epoch": 1.6, "learning_rate": 1.8517904682826695e-05, "loss": 0.0859, "step": 8736 }, { "epoch": 1.6, "learning_rate": 1.8517515713165585e-05, "loss": 0.0957, "step": 8737 }, { "epoch": 1.6, "learning_rate": 1.851712669655573e-05, "loss": 0.1054, "step": 8738 }, { "epoch": 1.6, "learning_rate": 1.851673763299927e-05, "loss": 0.0991, "step": 8739 }, { "epoch": 1.6, "learning_rate": 1.8516348522498354e-05, "loss": 0.1002, "step": 8740 }, { "epoch": 1.6, "learning_rate": 1.851595936505513e-05, "loss": 0.0879, "step": 8741 }, { "epoch": 1.6, "learning_rate": 1.8515570160671732e-05, "loss": 0.092, "step": 8742 }, { "epoch": 1.6, "learning_rate": 1.8515180909350317e-05, "loss": 0.0982, "step": 8743 }, { "epoch": 1.6, "learning_rate": 1.851479161109302e-05, "loss": 0.0846, "step": 8744 }, { "epoch": 1.6, "learning_rate": 1.8514402265901995e-05, "loss": 0.0992, "step": 8745 }, { "epoch": 1.6, "learning_rate": 1.8514012873779387e-05, "loss": 0.0906, "step": 8746 }, { "epoch": 1.6, "learning_rate": 1.851362343472734e-05, "loss": 0.101, "step": 8747 }, { "epoch": 1.6, "learning_rate": 1.8513233948748e-05, "loss": 0.1, "step": 8748 }, { "epoch": 1.6, "learning_rate": 1.8512844415843514e-05, "loss": 0.1116, "step": 8749 }, { "epoch": 1.6, "learning_rate": 1.851245483601603e-05, "loss": 0.1077, "step": 8750 }, { "epoch": 1.6, "learning_rate": 1.85120652092677e-05, "loss": 0.0975, "step": 8751 }, { "epoch": 1.6, "learning_rate": 1.8511675535600663e-05, "loss": 0.1008, "step": 8752 }, { "epoch": 1.61, "learning_rate": 1.851128581501707e-05, "loss": 0.0972, "step": 8753 }, { "epoch": 1.61, "learning_rate": 1.8510896047519075e-05, "loss": 0.096, "step": 8754 }, { "epoch": 1.61, "learning_rate": 1.8510506233108815e-05, "loss": 0.1036, "step": 8755 }, { "epoch": 1.61, "learning_rate": 1.851011637178845e-05, "loss": 0.0888, "step": 8756 }, { "epoch": 1.61, "learning_rate": 1.8509726463560123e-05, "loss": 0.098, "step": 8757 }, { "epoch": 1.61, "learning_rate": 1.8509336508425985e-05, "loss": 0.0977, "step": 8758 }, { "epoch": 1.61, "learning_rate": 1.8508946506388182e-05, "loss": 0.0867, "step": 8759 }, { "epoch": 1.61, "learning_rate": 1.850855645744887e-05, "loss": 0.0861, "step": 8760 }, { "epoch": 1.61, "learning_rate": 1.8508166361610197e-05, "loss": 0.0927, "step": 8761 }, { "epoch": 1.61, "learning_rate": 1.850777621887431e-05, "loss": 0.11, "step": 8762 }, { "epoch": 1.61, "learning_rate": 1.850738602924336e-05, "loss": 0.103, "step": 8763 }, { "epoch": 1.61, "learning_rate": 1.8506995792719498e-05, "loss": 0.0919, "step": 8764 }, { "epoch": 1.61, "learning_rate": 1.8506605509304875e-05, "loss": 0.0939, "step": 8765 }, { "epoch": 1.61, "learning_rate": 1.8506215179001647e-05, "loss": 0.0966, "step": 8766 }, { "epoch": 1.61, "learning_rate": 1.850582480181196e-05, "loss": 0.1035, "step": 8767 }, { "epoch": 1.61, "learning_rate": 1.8505434377737965e-05, "loss": 0.0926, "step": 8768 }, { "epoch": 1.61, "learning_rate": 1.850504390678182e-05, "loss": 0.1039, "step": 8769 }, { "epoch": 1.61, "learning_rate": 1.850465338894567e-05, "loss": 0.096, "step": 8770 }, { "epoch": 1.61, "learning_rate": 1.8504262824231675e-05, "loss": 0.0953, "step": 8771 }, { "epoch": 1.61, "learning_rate": 1.850387221264198e-05, "loss": 0.1039, "step": 8772 }, { "epoch": 1.61, "learning_rate": 1.850348155417875e-05, "loss": 0.077, "step": 8773 }, { "epoch": 1.61, "learning_rate": 1.8503090848844122e-05, "loss": 0.0852, "step": 8774 }, { "epoch": 1.61, "learning_rate": 1.8502700096640264e-05, "loss": 0.1046, "step": 8775 }, { "epoch": 1.61, "learning_rate": 1.850230929756932e-05, "loss": 0.0908, "step": 8776 }, { "epoch": 1.61, "learning_rate": 1.850191845163345e-05, "loss": 0.087, "step": 8777 }, { "epoch": 1.61, "learning_rate": 1.8501527558834804e-05, "loss": 0.0954, "step": 8778 }, { "epoch": 1.61, "learning_rate": 1.850113661917554e-05, "loss": 0.0947, "step": 8779 }, { "epoch": 1.61, "learning_rate": 1.8500745632657816e-05, "loss": 0.0915, "step": 8780 }, { "epoch": 1.61, "learning_rate": 1.8500354599283776e-05, "loss": 0.1031, "step": 8781 }, { "epoch": 1.61, "learning_rate": 1.8499963519055588e-05, "loss": 0.0978, "step": 8782 }, { "epoch": 1.61, "learning_rate": 1.8499572391975398e-05, "loss": 0.0745, "step": 8783 }, { "epoch": 1.61, "learning_rate": 1.8499181218045368e-05, "loss": 0.1142, "step": 8784 }, { "epoch": 1.61, "learning_rate": 1.849878999726765e-05, "loss": 0.0902, "step": 8785 }, { "epoch": 1.61, "learning_rate": 1.8498398729644403e-05, "loss": 0.0907, "step": 8786 }, { "epoch": 1.61, "learning_rate": 1.8498007415177787e-05, "loss": 0.0961, "step": 8787 }, { "epoch": 1.61, "learning_rate": 1.8497616053869948e-05, "loss": 0.0963, "step": 8788 }, { "epoch": 1.61, "learning_rate": 1.8497224645723054e-05, "loss": 0.0944, "step": 8789 }, { "epoch": 1.61, "learning_rate": 1.849683319073926e-05, "loss": 0.0859, "step": 8790 }, { "epoch": 1.61, "learning_rate": 1.8496441688920722e-05, "loss": 0.0937, "step": 8791 }, { "epoch": 1.61, "learning_rate": 1.84960501402696e-05, "loss": 0.1016, "step": 8792 }, { "epoch": 1.61, "learning_rate": 1.8495658544788046e-05, "loss": 0.0957, "step": 8793 }, { "epoch": 1.61, "learning_rate": 1.8495266902478226e-05, "loss": 0.0825, "step": 8794 }, { "epoch": 1.61, "learning_rate": 1.8494875213342293e-05, "loss": 0.0955, "step": 8795 }, { "epoch": 1.61, "learning_rate": 1.8494483477382413e-05, "loss": 0.0838, "step": 8796 }, { "epoch": 1.61, "learning_rate": 1.8494091694600737e-05, "loss": 0.0866, "step": 8797 }, { "epoch": 1.61, "learning_rate": 1.849369986499943e-05, "loss": 0.1039, "step": 8798 }, { "epoch": 1.61, "learning_rate": 1.8493307988580654e-05, "loss": 0.096, "step": 8799 }, { "epoch": 1.61, "learning_rate": 1.849291606534656e-05, "loss": 0.0983, "step": 8800 }, { "epoch": 1.61, "learning_rate": 1.8492524095299316e-05, "loss": 0.1053, "step": 8801 }, { "epoch": 1.61, "learning_rate": 1.849213207844108e-05, "loss": 0.099, "step": 8802 }, { "epoch": 1.61, "learning_rate": 1.8491740014774014e-05, "loss": 0.0803, "step": 8803 }, { "epoch": 1.61, "learning_rate": 1.8491347904300278e-05, "loss": 0.1001, "step": 8804 }, { "epoch": 1.61, "learning_rate": 1.849095574702203e-05, "loss": 0.1009, "step": 8805 }, { "epoch": 1.61, "learning_rate": 1.849056354294144e-05, "loss": 0.1033, "step": 8806 }, { "epoch": 1.62, "learning_rate": 1.8490171292060662e-05, "loss": 0.0917, "step": 8807 }, { "epoch": 1.62, "learning_rate": 1.848977899438186e-05, "loss": 0.1006, "step": 8808 }, { "epoch": 1.62, "learning_rate": 1.84893866499072e-05, "loss": 0.0902, "step": 8809 }, { "epoch": 1.62, "learning_rate": 1.8488994258638843e-05, "loss": 0.0983, "step": 8810 }, { "epoch": 1.62, "learning_rate": 1.8488601820578943e-05, "loss": 0.0991, "step": 8811 }, { "epoch": 1.62, "learning_rate": 1.848820933572968e-05, "loss": 0.0949, "step": 8812 }, { "epoch": 1.62, "learning_rate": 1.8487816804093205e-05, "loss": 0.0922, "step": 8813 }, { "epoch": 1.62, "learning_rate": 1.8487424225671682e-05, "loss": 0.0964, "step": 8814 }, { "epoch": 1.62, "learning_rate": 1.8487031600467283e-05, "loss": 0.0911, "step": 8815 }, { "epoch": 1.62, "learning_rate": 1.8486638928482162e-05, "loss": 0.0897, "step": 8816 }, { "epoch": 1.62, "learning_rate": 1.8486246209718492e-05, "loss": 0.0801, "step": 8817 }, { "epoch": 1.62, "learning_rate": 1.8485853444178433e-05, "loss": 0.1037, "step": 8818 }, { "epoch": 1.62, "learning_rate": 1.848546063186415e-05, "loss": 0.0964, "step": 8819 }, { "epoch": 1.62, "learning_rate": 1.848506777277781e-05, "loss": 0.0933, "step": 8820 }, { "epoch": 1.62, "learning_rate": 1.8484674866921577e-05, "loss": 0.1028, "step": 8821 }, { "epoch": 1.62, "learning_rate": 1.848428191429762e-05, "loss": 0.094, "step": 8822 }, { "epoch": 1.62, "learning_rate": 1.8483888914908098e-05, "loss": 0.0949, "step": 8823 }, { "epoch": 1.62, "learning_rate": 1.8483495868755185e-05, "loss": 0.0971, "step": 8824 }, { "epoch": 1.62, "learning_rate": 1.8483102775841046e-05, "loss": 0.1024, "step": 8825 }, { "epoch": 1.62, "learning_rate": 1.8482709636167842e-05, "loss": 0.1044, "step": 8826 }, { "epoch": 1.62, "learning_rate": 1.8482316449737742e-05, "loss": 0.0921, "step": 8827 }, { "epoch": 1.62, "learning_rate": 1.848192321655292e-05, "loss": 0.0864, "step": 8828 }, { "epoch": 1.62, "learning_rate": 1.8481529936615533e-05, "loss": 0.0895, "step": 8829 }, { "epoch": 1.62, "learning_rate": 1.8481136609927758e-05, "loss": 0.0927, "step": 8830 }, { "epoch": 1.62, "learning_rate": 1.848074323649176e-05, "loss": 0.0952, "step": 8831 }, { "epoch": 1.62, "learning_rate": 1.8480349816309704e-05, "loss": 0.1067, "step": 8832 }, { "epoch": 1.62, "learning_rate": 1.8479956349383763e-05, "loss": 0.0879, "step": 8833 }, { "epoch": 1.62, "learning_rate": 1.8479562835716104e-05, "loss": 0.0961, "step": 8834 }, { "epoch": 1.62, "learning_rate": 1.8479169275308892e-05, "loss": 0.0955, "step": 8835 }, { "epoch": 1.62, "learning_rate": 1.8478775668164307e-05, "loss": 0.0951, "step": 8836 }, { "epoch": 1.62, "learning_rate": 1.847838201428451e-05, "loss": 0.0891, "step": 8837 }, { "epoch": 1.62, "learning_rate": 1.8477988313671668e-05, "loss": 0.0959, "step": 8838 }, { "epoch": 1.62, "learning_rate": 1.8477594566327958e-05, "loss": 0.0944, "step": 8839 }, { "epoch": 1.62, "learning_rate": 1.847720077225555e-05, "loss": 0.0994, "step": 8840 }, { "epoch": 1.62, "learning_rate": 1.847680693145661e-05, "loss": 0.096, "step": 8841 }, { "epoch": 1.62, "learning_rate": 1.8476413043933316e-05, "loss": 0.1057, "step": 8842 }, { "epoch": 1.62, "learning_rate": 1.8476019109687832e-05, "loss": 0.0948, "step": 8843 }, { "epoch": 1.62, "learning_rate": 1.847562512872233e-05, "loss": 0.0838, "step": 8844 }, { "epoch": 1.62, "learning_rate": 1.847523110103899e-05, "loss": 0.1023, "step": 8845 }, { "epoch": 1.62, "learning_rate": 1.8474837026639975e-05, "loss": 0.1013, "step": 8846 }, { "epoch": 1.62, "learning_rate": 1.8474442905527456e-05, "loss": 0.0842, "step": 8847 }, { "epoch": 1.62, "learning_rate": 1.847404873770361e-05, "loss": 0.0896, "step": 8848 }, { "epoch": 1.62, "learning_rate": 1.847365452317061e-05, "loss": 0.0948, "step": 8849 }, { "epoch": 1.62, "learning_rate": 1.8473260261930626e-05, "loss": 0.1058, "step": 8850 }, { "epoch": 1.62, "learning_rate": 1.847286595398584e-05, "loss": 0.0971, "step": 8851 }, { "epoch": 1.62, "learning_rate": 1.8472471599338413e-05, "loss": 0.0956, "step": 8852 }, { "epoch": 1.62, "learning_rate": 1.8472077197990522e-05, "loss": 0.1072, "step": 8853 }, { "epoch": 1.62, "learning_rate": 1.8471682749944346e-05, "loss": 0.1011, "step": 8854 }, { "epoch": 1.62, "learning_rate": 1.8471288255202056e-05, "loss": 0.0968, "step": 8855 }, { "epoch": 1.62, "learning_rate": 1.847089371376583e-05, "loss": 0.0949, "step": 8856 }, { "epoch": 1.62, "learning_rate": 1.847049912563783e-05, "loss": 0.09, "step": 8857 }, { "epoch": 1.62, "learning_rate": 1.847010449082025e-05, "loss": 0.0949, "step": 8858 }, { "epoch": 1.62, "learning_rate": 1.846970980931525e-05, "loss": 0.0896, "step": 8859 }, { "epoch": 1.62, "learning_rate": 1.846931508112501e-05, "loss": 0.0881, "step": 8860 }, { "epoch": 1.62, "learning_rate": 1.8468920306251715e-05, "loss": 0.0913, "step": 8861 }, { "epoch": 1.63, "learning_rate": 1.8468525484697527e-05, "loss": 0.0923, "step": 8862 }, { "epoch": 1.63, "learning_rate": 1.846813061646463e-05, "loss": 0.1026, "step": 8863 }, { "epoch": 1.63, "learning_rate": 1.8467735701555193e-05, "loss": 0.096, "step": 8864 }, { "epoch": 1.63, "learning_rate": 1.8467340739971406e-05, "loss": 0.1005, "step": 8865 }, { "epoch": 1.63, "learning_rate": 1.8466945731715433e-05, "loss": 0.1004, "step": 8866 }, { "epoch": 1.63, "learning_rate": 1.846655067678946e-05, "loss": 0.0881, "step": 8867 }, { "epoch": 1.63, "learning_rate": 1.8466155575195663e-05, "loss": 0.0946, "step": 8868 }, { "epoch": 1.63, "learning_rate": 1.8465760426936212e-05, "loss": 0.085, "step": 8869 }, { "epoch": 1.63, "learning_rate": 1.8465365232013294e-05, "loss": 0.1061, "step": 8870 }, { "epoch": 1.63, "learning_rate": 1.8464969990429088e-05, "loss": 0.1133, "step": 8871 }, { "epoch": 1.63, "learning_rate": 1.8464574702185765e-05, "loss": 0.1097, "step": 8872 }, { "epoch": 1.63, "learning_rate": 1.8464179367285508e-05, "loss": 0.1035, "step": 8873 }, { "epoch": 1.63, "learning_rate": 1.84637839857305e-05, "loss": 0.1045, "step": 8874 }, { "epoch": 1.63, "learning_rate": 1.846338855752291e-05, "loss": 0.0895, "step": 8875 }, { "epoch": 1.63, "learning_rate": 1.846299308266493e-05, "loss": 0.1056, "step": 8876 }, { "epoch": 1.63, "learning_rate": 1.846259756115873e-05, "loss": 0.085, "step": 8877 }, { "epoch": 1.63, "learning_rate": 1.8462201993006496e-05, "loss": 0.0968, "step": 8878 }, { "epoch": 1.63, "learning_rate": 1.8461806378210406e-05, "loss": 0.0974, "step": 8879 }, { "epoch": 1.63, "learning_rate": 1.846141071677264e-05, "loss": 0.0983, "step": 8880 }, { "epoch": 1.63, "learning_rate": 1.8461015008695382e-05, "loss": 0.0965, "step": 8881 }, { "epoch": 1.63, "learning_rate": 1.846061925398081e-05, "loss": 0.0891, "step": 8882 }, { "epoch": 1.63, "learning_rate": 1.8460223452631106e-05, "loss": 0.1039, "step": 8883 }, { "epoch": 1.63, "learning_rate": 1.845982760464845e-05, "loss": 0.1055, "step": 8884 }, { "epoch": 1.63, "learning_rate": 1.845943171003503e-05, "loss": 0.0898, "step": 8885 }, { "epoch": 1.63, "learning_rate": 1.8459035768793022e-05, "loss": 0.0992, "step": 8886 }, { "epoch": 1.63, "learning_rate": 1.8458639780924614e-05, "loss": 0.0971, "step": 8887 }, { "epoch": 1.63, "learning_rate": 1.8458243746431982e-05, "loss": 0.0945, "step": 8888 }, { "epoch": 1.63, "learning_rate": 1.8457847665317312e-05, "loss": 0.0967, "step": 8889 }, { "epoch": 1.63, "learning_rate": 1.845745153758279e-05, "loss": 0.1028, "step": 8890 }, { "epoch": 1.63, "learning_rate": 1.8457055363230594e-05, "loss": 0.1076, "step": 8891 }, { "epoch": 1.63, "learning_rate": 1.845665914226291e-05, "loss": 0.0919, "step": 8892 }, { "epoch": 1.63, "learning_rate": 1.8456262874681926e-05, "loss": 0.1035, "step": 8893 }, { "epoch": 1.63, "learning_rate": 1.8455866560489823e-05, "loss": 0.0904, "step": 8894 }, { "epoch": 1.63, "learning_rate": 1.8455470199688784e-05, "loss": 0.0991, "step": 8895 }, { "epoch": 1.63, "learning_rate": 1.8455073792280997e-05, "loss": 0.0932, "step": 8896 }, { "epoch": 1.63, "learning_rate": 1.8454677338268644e-05, "loss": 0.0906, "step": 8897 }, { "epoch": 1.63, "learning_rate": 1.845428083765391e-05, "loss": 0.1007, "step": 8898 }, { "epoch": 1.63, "learning_rate": 1.8453884290438984e-05, "loss": 0.0864, "step": 8899 }, { "epoch": 1.63, "learning_rate": 1.845348769662605e-05, "loss": 0.0948, "step": 8900 }, { "epoch": 1.63, "learning_rate": 1.8453091056217292e-05, "loss": 0.088, "step": 8901 }, { "epoch": 1.63, "learning_rate": 1.84526943692149e-05, "loss": 0.0843, "step": 8902 }, { "epoch": 1.63, "learning_rate": 1.8452297635621053e-05, "loss": 0.0891, "step": 8903 }, { "epoch": 1.63, "learning_rate": 1.845190085543795e-05, "loss": 0.098, "step": 8904 }, { "epoch": 1.63, "learning_rate": 1.8451504028667768e-05, "loss": 0.0984, "step": 8905 }, { "epoch": 1.63, "learning_rate": 1.84511071553127e-05, "loss": 0.1, "step": 8906 }, { "epoch": 1.63, "learning_rate": 1.845071023537493e-05, "loss": 0.0753, "step": 8907 }, { "epoch": 1.63, "learning_rate": 1.8450313268856644e-05, "loss": 0.0974, "step": 8908 }, { "epoch": 1.63, "learning_rate": 1.8449916255760038e-05, "loss": 0.1018, "step": 8909 }, { "epoch": 1.63, "learning_rate": 1.84495191960873e-05, "loss": 0.1019, "step": 8910 }, { "epoch": 1.63, "learning_rate": 1.8449122089840607e-05, "loss": 0.1004, "step": 8911 }, { "epoch": 1.63, "learning_rate": 1.8448724937022152e-05, "loss": 0.0978, "step": 8912 }, { "epoch": 1.63, "learning_rate": 1.8448327737634134e-05, "loss": 0.094, "step": 8913 }, { "epoch": 1.63, "learning_rate": 1.8447930491678735e-05, "loss": 0.099, "step": 8914 }, { "epoch": 1.63, "learning_rate": 1.8447533199158143e-05, "loss": 0.1096, "step": 8915 }, { "epoch": 1.64, "learning_rate": 1.844713586007455e-05, "loss": 0.0985, "step": 8916 }, { "epoch": 1.64, "learning_rate": 1.8446738474430148e-05, "loss": 0.0809, "step": 8917 }, { "epoch": 1.64, "learning_rate": 1.8446341042227125e-05, "loss": 0.0966, "step": 8918 }, { "epoch": 1.64, "learning_rate": 1.8445943563467672e-05, "loss": 0.0921, "step": 8919 }, { "epoch": 1.64, "learning_rate": 1.844554603815398e-05, "loss": 0.0988, "step": 8920 }, { "epoch": 1.64, "learning_rate": 1.844514846628824e-05, "loss": 0.1007, "step": 8921 }, { "epoch": 1.64, "learning_rate": 1.8444750847872645e-05, "loss": 0.0951, "step": 8922 }, { "epoch": 1.64, "learning_rate": 1.8444353182909386e-05, "loss": 0.096, "step": 8923 }, { "epoch": 1.64, "learning_rate": 1.844395547140065e-05, "loss": 0.0867, "step": 8924 }, { "epoch": 1.64, "learning_rate": 1.8443557713348637e-05, "loss": 0.0967, "step": 8925 }, { "epoch": 1.64, "learning_rate": 1.8443159908755538e-05, "loss": 0.0888, "step": 8926 }, { "epoch": 1.64, "learning_rate": 1.844276205762354e-05, "loss": 0.0913, "step": 8927 }, { "epoch": 1.64, "learning_rate": 1.8442364159954843e-05, "loss": 0.098, "step": 8928 }, { "epoch": 1.64, "learning_rate": 1.8441966215751636e-05, "loss": 0.104, "step": 8929 }, { "epoch": 1.64, "learning_rate": 1.8441568225016114e-05, "loss": 0.1076, "step": 8930 }, { "epoch": 1.64, "learning_rate": 1.8441170187750467e-05, "loss": 0.0876, "step": 8931 }, { "epoch": 1.64, "learning_rate": 1.8440772103956897e-05, "loss": 0.0939, "step": 8932 }, { "epoch": 1.64, "learning_rate": 1.844037397363759e-05, "loss": 0.1043, "step": 8933 }, { "epoch": 1.64, "learning_rate": 1.8439975796794743e-05, "loss": 0.1057, "step": 8934 }, { "epoch": 1.64, "learning_rate": 1.8439577573430557e-05, "loss": 0.0825, "step": 8935 }, { "epoch": 1.64, "learning_rate": 1.8439179303547217e-05, "loss": 0.0908, "step": 8936 }, { "epoch": 1.64, "learning_rate": 1.8438780987146924e-05, "loss": 0.0904, "step": 8937 }, { "epoch": 1.64, "learning_rate": 1.843838262423187e-05, "loss": 0.1082, "step": 8938 }, { "epoch": 1.64, "learning_rate": 1.843798421480426e-05, "loss": 0.0892, "step": 8939 }, { "epoch": 1.64, "learning_rate": 1.8437585758866277e-05, "loss": 0.1041, "step": 8940 }, { "epoch": 1.64, "learning_rate": 1.8437187256420124e-05, "loss": 0.0996, "step": 8941 }, { "epoch": 1.64, "learning_rate": 1.8436788707468e-05, "loss": 0.1029, "step": 8942 }, { "epoch": 1.64, "learning_rate": 1.84363901120121e-05, "loss": 0.0973, "step": 8943 }, { "epoch": 1.64, "learning_rate": 1.8435991470054615e-05, "loss": 0.101, "step": 8944 }, { "epoch": 1.64, "learning_rate": 1.8435592781597753e-05, "loss": 0.0923, "step": 8945 }, { "epoch": 1.64, "learning_rate": 1.8435194046643702e-05, "loss": 0.0948, "step": 8946 }, { "epoch": 1.64, "learning_rate": 1.8434795265194664e-05, "loss": 0.0895, "step": 8947 }, { "epoch": 1.64, "learning_rate": 1.843439643725284e-05, "loss": 0.0921, "step": 8948 }, { "epoch": 1.64, "learning_rate": 1.8433997562820422e-05, "loss": 0.0916, "step": 8949 }, { "epoch": 1.64, "learning_rate": 1.8433598641899613e-05, "loss": 0.1051, "step": 8950 }, { "epoch": 1.64, "learning_rate": 1.843319967449261e-05, "loss": 0.0932, "step": 8951 }, { "epoch": 1.64, "learning_rate": 1.8432800660601616e-05, "loss": 0.1033, "step": 8952 }, { "epoch": 1.64, "learning_rate": 1.8432401600228823e-05, "loss": 0.0906, "step": 8953 }, { "epoch": 1.64, "learning_rate": 1.843200249337644e-05, "loss": 0.1025, "step": 8954 }, { "epoch": 1.64, "learning_rate": 1.8431603340046655e-05, "loss": 0.0899, "step": 8955 }, { "epoch": 1.64, "learning_rate": 1.8431204140241678e-05, "loss": 0.088, "step": 8956 }, { "epoch": 1.64, "learning_rate": 1.8430804893963706e-05, "loss": 0.0878, "step": 8957 }, { "epoch": 1.64, "learning_rate": 1.843040560121494e-05, "loss": 0.0887, "step": 8958 }, { "epoch": 1.64, "learning_rate": 1.8430006261997585e-05, "loss": 0.0943, "step": 8959 }, { "epoch": 1.64, "learning_rate": 1.8429606876313833e-05, "loss": 0.0917, "step": 8960 }, { "epoch": 1.64, "learning_rate": 1.8429207444165892e-05, "loss": 0.0961, "step": 8961 }, { "epoch": 1.64, "learning_rate": 1.8428807965555958e-05, "loss": 0.0929, "step": 8962 }, { "epoch": 1.64, "learning_rate": 1.8428408440486243e-05, "loss": 0.0937, "step": 8963 }, { "epoch": 1.64, "learning_rate": 1.8428008868958943e-05, "loss": 0.1082, "step": 8964 }, { "epoch": 1.64, "learning_rate": 1.8427609250976256e-05, "loss": 0.0844, "step": 8965 }, { "epoch": 1.64, "learning_rate": 1.8427209586540392e-05, "loss": 0.0811, "step": 8966 }, { "epoch": 1.64, "learning_rate": 1.842680987565355e-05, "loss": 0.0909, "step": 8967 }, { "epoch": 1.64, "learning_rate": 1.842641011831794e-05, "loss": 0.0939, "step": 8968 }, { "epoch": 1.64, "learning_rate": 1.8426010314535754e-05, "loss": 0.0768, "step": 8969 }, { "epoch": 1.64, "learning_rate": 1.8425610464309204e-05, "loss": 0.1018, "step": 8970 }, { "epoch": 1.65, "learning_rate": 1.8425210567640493e-05, "loss": 0.0943, "step": 8971 }, { "epoch": 1.65, "learning_rate": 1.8424810624531823e-05, "loss": 0.0942, "step": 8972 }, { "epoch": 1.65, "learning_rate": 1.8424410634985398e-05, "loss": 0.097, "step": 8973 }, { "epoch": 1.65, "learning_rate": 1.8424010599003424e-05, "loss": 0.1022, "step": 8974 }, { "epoch": 1.65, "learning_rate": 1.8423610516588107e-05, "loss": 0.0924, "step": 8975 }, { "epoch": 1.65, "learning_rate": 1.842321038774165e-05, "loss": 0.099, "step": 8976 }, { "epoch": 1.65, "learning_rate": 1.842281021246626e-05, "loss": 0.0972, "step": 8977 }, { "epoch": 1.65, "learning_rate": 1.8422409990764148e-05, "loss": 0.0989, "step": 8978 }, { "epoch": 1.65, "learning_rate": 1.842200972263751e-05, "loss": 0.1002, "step": 8979 }, { "epoch": 1.65, "learning_rate": 1.8421609408088563e-05, "loss": 0.0992, "step": 8980 }, { "epoch": 1.65, "learning_rate": 1.8421209047119498e-05, "loss": 0.0956, "step": 8981 }, { "epoch": 1.65, "learning_rate": 1.8420808639732537e-05, "loss": 0.1035, "step": 8982 }, { "epoch": 1.65, "learning_rate": 1.842040818592988e-05, "loss": 0.0941, "step": 8983 }, { "epoch": 1.65, "learning_rate": 1.842000768571374e-05, "loss": 0.088, "step": 8984 }, { "epoch": 1.65, "learning_rate": 1.8419607139086318e-05, "loss": 0.1013, "step": 8985 }, { "epoch": 1.65, "learning_rate": 1.8419206546049823e-05, "loss": 0.0951, "step": 8986 }, { "epoch": 1.65, "learning_rate": 1.8418805906606466e-05, "loss": 0.0895, "step": 8987 }, { "epoch": 1.65, "learning_rate": 1.841840522075845e-05, "loss": 0.0896, "step": 8988 }, { "epoch": 1.65, "learning_rate": 1.841800448850799e-05, "loss": 0.0992, "step": 8989 }, { "epoch": 1.65, "learning_rate": 1.8417603709857292e-05, "loss": 0.098, "step": 8990 }, { "epoch": 1.65, "learning_rate": 1.8417202884808563e-05, "loss": 0.0892, "step": 8991 }, { "epoch": 1.65, "learning_rate": 1.8416802013364017e-05, "loss": 0.0931, "step": 8992 }, { "epoch": 1.65, "learning_rate": 1.841640109552586e-05, "loss": 0.1021, "step": 8993 }, { "epoch": 1.65, "learning_rate": 1.8416000131296305e-05, "loss": 0.0916, "step": 8994 }, { "epoch": 1.65, "learning_rate": 1.8415599120677557e-05, "loss": 0.1142, "step": 8995 }, { "epoch": 1.65, "learning_rate": 1.841519806367183e-05, "loss": 0.0971, "step": 8996 }, { "epoch": 1.65, "learning_rate": 1.8414796960281335e-05, "loss": 0.0856, "step": 8997 }, { "epoch": 1.65, "learning_rate": 1.8414395810508284e-05, "loss": 0.1066, "step": 8998 }, { "epoch": 1.65, "learning_rate": 1.841399461435488e-05, "loss": 0.0968, "step": 8999 }, { "epoch": 1.65, "learning_rate": 1.8413593371823347e-05, "loss": 0.0878, "step": 9000 }, { "epoch": 1.65, "learning_rate": 1.841319208291589e-05, "loss": 0.0989, "step": 9001 }, { "epoch": 1.65, "learning_rate": 1.8412790747634715e-05, "loss": 0.095, "step": 9002 }, { "epoch": 1.65, "learning_rate": 1.8412389365982045e-05, "loss": 0.0832, "step": 9003 }, { "epoch": 1.65, "learning_rate": 1.841198793796009e-05, "loss": 0.0852, "step": 9004 }, { "epoch": 1.65, "learning_rate": 1.841158646357106e-05, "loss": 0.091, "step": 9005 }, { "epoch": 1.65, "learning_rate": 1.8411184942817165e-05, "loss": 0.1016, "step": 9006 }, { "epoch": 1.65, "learning_rate": 1.8410783375700624e-05, "loss": 0.1114, "step": 9007 }, { "epoch": 1.65, "learning_rate": 1.8410381762223645e-05, "loss": 0.0954, "step": 9008 }, { "epoch": 1.65, "learning_rate": 1.8409980102388445e-05, "loss": 0.0926, "step": 9009 }, { "epoch": 1.65, "learning_rate": 1.840957839619724e-05, "loss": 0.101, "step": 9010 }, { "epoch": 1.65, "learning_rate": 1.840917664365224e-05, "loss": 0.1072, "step": 9011 }, { "epoch": 1.65, "learning_rate": 1.840877484475566e-05, "loss": 0.0907, "step": 9012 }, { "epoch": 1.65, "learning_rate": 1.8408372999509718e-05, "loss": 0.091, "step": 9013 }, { "epoch": 1.65, "learning_rate": 1.840797110791663e-05, "loss": 0.0882, "step": 9014 }, { "epoch": 1.65, "learning_rate": 1.8407569169978603e-05, "loss": 0.1011, "step": 9015 }, { "epoch": 1.65, "learning_rate": 1.840716718569786e-05, "loss": 0.0974, "step": 9016 }, { "epoch": 1.65, "learning_rate": 1.8406765155076617e-05, "loss": 0.1096, "step": 9017 }, { "epoch": 1.65, "learning_rate": 1.8406363078117084e-05, "loss": 0.0795, "step": 9018 }, { "epoch": 1.65, "learning_rate": 1.840596095482148e-05, "loss": 0.0951, "step": 9019 }, { "epoch": 1.65, "learning_rate": 1.8405558785192024e-05, "loss": 0.0945, "step": 9020 }, { "epoch": 1.65, "learning_rate": 1.8405156569230932e-05, "loss": 0.0994, "step": 9021 }, { "epoch": 1.65, "learning_rate": 1.840475430694042e-05, "loss": 0.088, "step": 9022 }, { "epoch": 1.65, "learning_rate": 1.8404351998322707e-05, "loss": 0.0835, "step": 9023 }, { "epoch": 1.65, "learning_rate": 1.8403949643380005e-05, "loss": 0.0885, "step": 9024 }, { "epoch": 1.66, "learning_rate": 1.8403547242114535e-05, "loss": 0.0898, "step": 9025 }, { "epoch": 1.66, "learning_rate": 1.840314479452852e-05, "loss": 0.1035, "step": 9026 }, { "epoch": 1.66, "learning_rate": 1.840274230062417e-05, "loss": 0.0953, "step": 9027 }, { "epoch": 1.66, "learning_rate": 1.8402339760403715e-05, "loss": 0.0867, "step": 9028 }, { "epoch": 1.66, "learning_rate": 1.840193717386936e-05, "loss": 0.1102, "step": 9029 }, { "epoch": 1.66, "learning_rate": 1.840153454102333e-05, "loss": 0.1041, "step": 9030 }, { "epoch": 1.66, "learning_rate": 1.8401131861867846e-05, "loss": 0.1007, "step": 9031 }, { "epoch": 1.66, "learning_rate": 1.840072913640513e-05, "loss": 0.0891, "step": 9032 }, { "epoch": 1.66, "learning_rate": 1.8400326364637392e-05, "loss": 0.1, "step": 9033 }, { "epoch": 1.66, "learning_rate": 1.839992354656686e-05, "loss": 0.1066, "step": 9034 }, { "epoch": 1.66, "learning_rate": 1.8399520682195755e-05, "loss": 0.0961, "step": 9035 }, { "epoch": 1.66, "learning_rate": 1.8399117771526294e-05, "loss": 0.0954, "step": 9036 }, { "epoch": 1.66, "learning_rate": 1.8398714814560698e-05, "loss": 0.1029, "step": 9037 }, { "epoch": 1.66, "learning_rate": 1.8398311811301194e-05, "loss": 0.0863, "step": 9038 }, { "epoch": 1.66, "learning_rate": 1.839790876174999e-05, "loss": 0.0863, "step": 9039 }, { "epoch": 1.66, "learning_rate": 1.8397505665909326e-05, "loss": 0.0894, "step": 9040 }, { "epoch": 1.66, "learning_rate": 1.8397102523781406e-05, "loss": 0.0998, "step": 9041 }, { "epoch": 1.66, "learning_rate": 1.8396699335368465e-05, "loss": 0.1061, "step": 9042 }, { "epoch": 1.66, "learning_rate": 1.8396296100672716e-05, "loss": 0.0982, "step": 9043 }, { "epoch": 1.66, "learning_rate": 1.839589281969639e-05, "loss": 0.0927, "step": 9044 }, { "epoch": 1.66, "learning_rate": 1.8395489492441706e-05, "loss": 0.0894, "step": 9045 }, { "epoch": 1.66, "learning_rate": 1.8395086118910884e-05, "loss": 0.1046, "step": 9046 }, { "epoch": 1.66, "learning_rate": 1.8394682699106154e-05, "loss": 0.1062, "step": 9047 }, { "epoch": 1.66, "learning_rate": 1.8394279233029735e-05, "loss": 0.0908, "step": 9048 }, { "epoch": 1.66, "learning_rate": 1.8393875720683853e-05, "loss": 0.0949, "step": 9049 }, { "epoch": 1.66, "learning_rate": 1.8393472162070728e-05, "loss": 0.0904, "step": 9050 }, { "epoch": 1.66, "learning_rate": 1.8393068557192592e-05, "loss": 0.0903, "step": 9051 }, { "epoch": 1.66, "learning_rate": 1.8392664906051664e-05, "loss": 0.1041, "step": 9052 }, { "epoch": 1.66, "learning_rate": 1.839226120865017e-05, "loss": 0.089, "step": 9053 }, { "epoch": 1.66, "learning_rate": 1.8391857464990336e-05, "loss": 0.1029, "step": 9054 }, { "epoch": 1.66, "learning_rate": 1.839145367507439e-05, "loss": 0.0908, "step": 9055 }, { "epoch": 1.66, "learning_rate": 1.8391049838904553e-05, "loss": 0.0903, "step": 9056 }, { "epoch": 1.66, "learning_rate": 1.8390645956483048e-05, "loss": 0.0987, "step": 9057 }, { "epoch": 1.66, "learning_rate": 1.839024202781211e-05, "loss": 0.0982, "step": 9058 }, { "epoch": 1.66, "learning_rate": 1.838983805289396e-05, "loss": 0.0907, "step": 9059 }, { "epoch": 1.66, "learning_rate": 1.8389434031730827e-05, "loss": 0.0994, "step": 9060 }, { "epoch": 1.66, "learning_rate": 1.838902996432494e-05, "loss": 0.1031, "step": 9061 }, { "epoch": 1.66, "learning_rate": 1.838862585067852e-05, "loss": 0.1046, "step": 9062 }, { "epoch": 1.66, "learning_rate": 1.83882216907938e-05, "loss": 0.0795, "step": 9063 }, { "epoch": 1.66, "learning_rate": 1.8387817484673e-05, "loss": 0.0885, "step": 9064 }, { "epoch": 1.66, "learning_rate": 1.8387413232318357e-05, "loss": 0.0932, "step": 9065 }, { "epoch": 1.66, "learning_rate": 1.8387008933732096e-05, "loss": 0.0983, "step": 9066 }, { "epoch": 1.66, "learning_rate": 1.8386604588916447e-05, "loss": 0.099, "step": 9067 }, { "epoch": 1.66, "learning_rate": 1.8386200197873632e-05, "loss": 0.1106, "step": 9068 }, { "epoch": 1.66, "learning_rate": 1.838579576060589e-05, "loss": 0.0906, "step": 9069 }, { "epoch": 1.66, "learning_rate": 1.8385391277115445e-05, "loss": 0.0981, "step": 9070 }, { "epoch": 1.66, "learning_rate": 1.8384986747404525e-05, "loss": 0.0907, "step": 9071 }, { "epoch": 1.66, "learning_rate": 1.8384582171475363e-05, "loss": 0.0942, "step": 9072 }, { "epoch": 1.66, "learning_rate": 1.8384177549330187e-05, "loss": 0.095, "step": 9073 }, { "epoch": 1.66, "learning_rate": 1.8383772880971228e-05, "loss": 0.0865, "step": 9074 }, { "epoch": 1.66, "learning_rate": 1.8383368166400713e-05, "loss": 0.096, "step": 9075 }, { "epoch": 1.66, "learning_rate": 1.838296340562088e-05, "loss": 0.1046, "step": 9076 }, { "epoch": 1.66, "learning_rate": 1.8382558598633957e-05, "loss": 0.0899, "step": 9077 }, { "epoch": 1.66, "learning_rate": 1.838215374544217e-05, "loss": 0.0984, "step": 9078 }, { "epoch": 1.66, "learning_rate": 1.838174884604776e-05, "loss": 0.1001, "step": 9079 }, { "epoch": 1.67, "learning_rate": 1.8381343900452954e-05, "loss": 0.0997, "step": 9080 }, { "epoch": 1.67, "learning_rate": 1.838093890865998e-05, "loss": 0.0954, "step": 9081 }, { "epoch": 1.67, "learning_rate": 1.8380533870671077e-05, "loss": 0.0962, "step": 9082 }, { "epoch": 1.67, "learning_rate": 1.8380128786488476e-05, "loss": 0.0954, "step": 9083 }, { "epoch": 1.67, "learning_rate": 1.8379723656114407e-05, "loss": 0.106, "step": 9084 }, { "epoch": 1.67, "learning_rate": 1.8379318479551106e-05, "loss": 0.0932, "step": 9085 }, { "epoch": 1.67, "learning_rate": 1.8378913256800806e-05, "loss": 0.0928, "step": 9086 }, { "epoch": 1.67, "learning_rate": 1.837850798786574e-05, "loss": 0.1004, "step": 9087 }, { "epoch": 1.67, "learning_rate": 1.8378102672748137e-05, "loss": 0.0838, "step": 9088 }, { "epoch": 1.67, "learning_rate": 1.837769731145024e-05, "loss": 0.0953, "step": 9089 }, { "epoch": 1.67, "learning_rate": 1.8377291903974276e-05, "loss": 0.0916, "step": 9090 }, { "epoch": 1.67, "learning_rate": 1.8376886450322487e-05, "loss": 0.0981, "step": 9091 }, { "epoch": 1.67, "learning_rate": 1.83764809504971e-05, "loss": 0.0924, "step": 9092 }, { "epoch": 1.67, "learning_rate": 1.8376075404500357e-05, "loss": 0.0949, "step": 9093 }, { "epoch": 1.67, "learning_rate": 1.837566981233449e-05, "loss": 0.1114, "step": 9094 }, { "epoch": 1.67, "learning_rate": 1.837526417400173e-05, "loss": 0.0905, "step": 9095 }, { "epoch": 1.67, "learning_rate": 1.8374858489504322e-05, "loss": 0.0853, "step": 9096 }, { "epoch": 1.67, "learning_rate": 1.83744527588445e-05, "loss": 0.0926, "step": 9097 }, { "epoch": 1.67, "learning_rate": 1.8374046982024493e-05, "loss": 0.0953, "step": 9098 }, { "epoch": 1.67, "learning_rate": 1.8373641159046545e-05, "loss": 0.0862, "step": 9099 }, { "epoch": 1.67, "learning_rate": 1.837323528991289e-05, "loss": 0.1068, "step": 9100 }, { "epoch": 1.67, "learning_rate": 1.8372829374625766e-05, "loss": 0.0934, "step": 9101 }, { "epoch": 1.67, "learning_rate": 1.8372423413187412e-05, "loss": 0.1018, "step": 9102 }, { "epoch": 1.67, "learning_rate": 1.8372017405600063e-05, "loss": 0.0995, "step": 9103 }, { "epoch": 1.67, "learning_rate": 1.837161135186596e-05, "loss": 0.0954, "step": 9104 }, { "epoch": 1.67, "learning_rate": 1.8371205251987336e-05, "loss": 0.0841, "step": 9105 }, { "epoch": 1.67, "learning_rate": 1.837079910596643e-05, "loss": 0.0989, "step": 9106 }, { "epoch": 1.67, "learning_rate": 1.8370392913805487e-05, "loss": 0.0968, "step": 9107 }, { "epoch": 1.67, "learning_rate": 1.8369986675506743e-05, "loss": 0.0846, "step": 9108 }, { "epoch": 1.67, "learning_rate": 1.8369580391072435e-05, "loss": 0.089, "step": 9109 }, { "epoch": 1.67, "learning_rate": 1.8369174060504803e-05, "loss": 0.1021, "step": 9110 }, { "epoch": 1.67, "learning_rate": 1.836876768380609e-05, "loss": 0.091, "step": 9111 }, { "epoch": 1.67, "learning_rate": 1.836836126097853e-05, "loss": 0.1071, "step": 9112 }, { "epoch": 1.67, "learning_rate": 1.836795479202437e-05, "loss": 0.0967, "step": 9113 }, { "epoch": 1.67, "learning_rate": 1.8367548276945846e-05, "loss": 0.1124, "step": 9114 }, { "epoch": 1.67, "learning_rate": 1.83671417157452e-05, "loss": 0.0898, "step": 9115 }, { "epoch": 1.67, "learning_rate": 1.8366735108424673e-05, "loss": 0.1074, "step": 9116 }, { "epoch": 1.67, "learning_rate": 1.8366328454986506e-05, "loss": 0.0988, "step": 9117 }, { "epoch": 1.67, "learning_rate": 1.8365921755432938e-05, "loss": 0.1085, "step": 9118 }, { "epoch": 1.67, "learning_rate": 1.836551500976622e-05, "loss": 0.1024, "step": 9119 }, { "epoch": 1.67, "learning_rate": 1.8365108217988577e-05, "loss": 0.1104, "step": 9120 }, { "epoch": 1.67, "learning_rate": 1.8364701380102267e-05, "loss": 0.1045, "step": 9121 }, { "epoch": 1.67, "learning_rate": 1.8364294496109528e-05, "loss": 0.0847, "step": 9122 }, { "epoch": 1.67, "learning_rate": 1.8363887566012603e-05, "loss": 0.0841, "step": 9123 }, { "epoch": 1.67, "learning_rate": 1.836348058981373e-05, "loss": 0.0905, "step": 9124 }, { "epoch": 1.67, "learning_rate": 1.836307356751516e-05, "loss": 0.1012, "step": 9125 }, { "epoch": 1.67, "learning_rate": 1.836266649911913e-05, "loss": 0.1012, "step": 9126 }, { "epoch": 1.67, "learning_rate": 1.8362259384627886e-05, "loss": 0.0978, "step": 9127 }, { "epoch": 1.67, "learning_rate": 1.8361852224043674e-05, "loss": 0.1039, "step": 9128 }, { "epoch": 1.67, "learning_rate": 1.8361445017368736e-05, "loss": 0.0868, "step": 9129 }, { "epoch": 1.67, "learning_rate": 1.8361037764605315e-05, "loss": 0.0973, "step": 9130 }, { "epoch": 1.67, "learning_rate": 1.836063046575566e-05, "loss": 0.096, "step": 9131 }, { "epoch": 1.67, "learning_rate": 1.8360223120822013e-05, "loss": 0.1088, "step": 9132 }, { "epoch": 1.67, "learning_rate": 1.8359815729806624e-05, "loss": 0.0958, "step": 9133 }, { "epoch": 1.68, "learning_rate": 1.835940829271173e-05, "loss": 0.0867, "step": 9134 }, { "epoch": 1.68, "learning_rate": 1.8359000809539584e-05, "loss": 0.0871, "step": 9135 }, { "epoch": 1.68, "learning_rate": 1.8358593280292433e-05, "loss": 0.095, "step": 9136 }, { "epoch": 1.68, "learning_rate": 1.8358185704972514e-05, "loss": 0.0899, "step": 9137 }, { "epoch": 1.68, "learning_rate": 1.8357778083582083e-05, "loss": 0.0955, "step": 9138 }, { "epoch": 1.68, "learning_rate": 1.835737041612338e-05, "loss": 0.0862, "step": 9139 }, { "epoch": 1.68, "learning_rate": 1.8356962702598658e-05, "loss": 0.0867, "step": 9140 }, { "epoch": 1.68, "learning_rate": 1.8356554943010163e-05, "loss": 0.0954, "step": 9141 }, { "epoch": 1.68, "learning_rate": 1.8356147137360142e-05, "loss": 0.0781, "step": 9142 }, { "epoch": 1.68, "learning_rate": 1.8355739285650838e-05, "loss": 0.1062, "step": 9143 }, { "epoch": 1.68, "learning_rate": 1.8355331387884506e-05, "loss": 0.1008, "step": 9144 }, { "epoch": 1.68, "learning_rate": 1.835492344406339e-05, "loss": 0.0968, "step": 9145 }, { "epoch": 1.68, "learning_rate": 1.8354515454189742e-05, "loss": 0.0907, "step": 9146 }, { "epoch": 1.68, "learning_rate": 1.8354107418265808e-05, "loss": 0.0992, "step": 9147 }, { "epoch": 1.68, "learning_rate": 1.835369933629384e-05, "loss": 0.1023, "step": 9148 }, { "epoch": 1.68, "learning_rate": 1.8353291208276082e-05, "loss": 0.1018, "step": 9149 }, { "epoch": 1.68, "learning_rate": 1.835288303421479e-05, "loss": 0.0821, "step": 9150 }, { "epoch": 1.68, "learning_rate": 1.8352474814112212e-05, "loss": 0.0965, "step": 9151 }, { "epoch": 1.68, "learning_rate": 1.8352066547970595e-05, "loss": 0.0959, "step": 9152 }, { "epoch": 1.68, "learning_rate": 1.835165823579219e-05, "loss": 0.0844, "step": 9153 }, { "epoch": 1.68, "learning_rate": 1.8351249877579255e-05, "loss": 0.1029, "step": 9154 }, { "epoch": 1.68, "learning_rate": 1.835084147333403e-05, "loss": 0.1063, "step": 9155 }, { "epoch": 1.68, "learning_rate": 1.8350433023058774e-05, "loss": 0.0796, "step": 9156 }, { "epoch": 1.68, "learning_rate": 1.8350024526755734e-05, "loss": 0.092, "step": 9157 }, { "epoch": 1.68, "learning_rate": 1.8349615984427162e-05, "loss": 0.0825, "step": 9158 }, { "epoch": 1.68, "learning_rate": 1.8349207396075313e-05, "loss": 0.0949, "step": 9159 }, { "epoch": 1.68, "learning_rate": 1.8348798761702436e-05, "loss": 0.0963, "step": 9160 }, { "epoch": 1.68, "learning_rate": 1.8348390081310783e-05, "loss": 0.1059, "step": 9161 }, { "epoch": 1.68, "learning_rate": 1.8347981354902613e-05, "loss": 0.0891, "step": 9162 }, { "epoch": 1.68, "learning_rate": 1.834757258248017e-05, "loss": 0.078, "step": 9163 }, { "epoch": 1.68, "learning_rate": 1.8347163764045716e-05, "loss": 0.0904, "step": 9164 }, { "epoch": 1.68, "learning_rate": 1.8346754899601495e-05, "loss": 0.0906, "step": 9165 }, { "epoch": 1.68, "learning_rate": 1.834634598914977e-05, "loss": 0.0919, "step": 9166 }, { "epoch": 1.68, "learning_rate": 1.8345937032692783e-05, "loss": 0.1052, "step": 9167 }, { "epoch": 1.68, "learning_rate": 1.8345528030232804e-05, "loss": 0.0918, "step": 9168 }, { "epoch": 1.68, "learning_rate": 1.8345118981772075e-05, "loss": 0.1068, "step": 9169 }, { "epoch": 1.68, "learning_rate": 1.8344709887312856e-05, "loss": 0.1, "step": 9170 }, { "epoch": 1.68, "learning_rate": 1.8344300746857396e-05, "loss": 0.1025, "step": 9171 }, { "epoch": 1.68, "learning_rate": 1.834389156040796e-05, "loss": 0.0976, "step": 9172 }, { "epoch": 1.68, "learning_rate": 1.83434823279668e-05, "loss": 0.0824, "step": 9173 }, { "epoch": 1.68, "learning_rate": 1.834307304953616e-05, "loss": 0.0896, "step": 9174 }, { "epoch": 1.68, "learning_rate": 1.8342663725118315e-05, "loss": 0.0874, "step": 9175 }, { "epoch": 1.68, "learning_rate": 1.834225435471551e-05, "loss": 0.0933, "step": 9176 }, { "epoch": 1.68, "learning_rate": 1.8341844938330002e-05, "loss": 0.0901, "step": 9177 }, { "epoch": 1.68, "learning_rate": 1.834143547596405e-05, "loss": 0.0831, "step": 9178 }, { "epoch": 1.68, "learning_rate": 1.8341025967619913e-05, "loss": 0.0962, "step": 9179 }, { "epoch": 1.68, "learning_rate": 1.8340616413299842e-05, "loss": 0.0846, "step": 9180 }, { "epoch": 1.68, "learning_rate": 1.8340206813006096e-05, "loss": 0.0928, "step": 9181 }, { "epoch": 1.68, "learning_rate": 1.8339797166740938e-05, "loss": 0.1051, "step": 9182 }, { "epoch": 1.68, "learning_rate": 1.8339387474506624e-05, "loss": 0.0911, "step": 9183 }, { "epoch": 1.68, "learning_rate": 1.8338977736305408e-05, "loss": 0.0886, "step": 9184 }, { "epoch": 1.68, "learning_rate": 1.8338567952139552e-05, "loss": 0.0896, "step": 9185 }, { "epoch": 1.68, "learning_rate": 1.8338158122011313e-05, "loss": 0.0788, "step": 9186 }, { "epoch": 1.68, "learning_rate": 1.8337748245922955e-05, "loss": 0.0925, "step": 9187 }, { "epoch": 1.68, "learning_rate": 1.833733832387673e-05, "loss": 0.0982, "step": 9188 }, { "epoch": 1.69, "learning_rate": 1.83369283558749e-05, "loss": 0.0906, "step": 9189 }, { "epoch": 1.69, "learning_rate": 1.8336518341919732e-05, "loss": 0.1025, "step": 9190 }, { "epoch": 1.69, "learning_rate": 1.8336108282013474e-05, "loss": 0.0974, "step": 9191 }, { "epoch": 1.69, "learning_rate": 1.833569817615839e-05, "loss": 0.0978, "step": 9192 }, { "epoch": 1.69, "learning_rate": 1.833528802435675e-05, "loss": 0.0933, "step": 9193 }, { "epoch": 1.69, "learning_rate": 1.83348778266108e-05, "loss": 0.0936, "step": 9194 }, { "epoch": 1.69, "learning_rate": 1.833446758292281e-05, "loss": 0.0963, "step": 9195 }, { "epoch": 1.69, "learning_rate": 1.833405729329504e-05, "loss": 0.0888, "step": 9196 }, { "epoch": 1.69, "learning_rate": 1.8333646957729748e-05, "loss": 0.0993, "step": 9197 }, { "epoch": 1.69, "learning_rate": 1.8333236576229204e-05, "loss": 0.1039, "step": 9198 }, { "epoch": 1.69, "learning_rate": 1.8332826148795662e-05, "loss": 0.0927, "step": 9199 }, { "epoch": 1.69, "learning_rate": 1.833241567543139e-05, "loss": 0.0882, "step": 9200 }, { "epoch": 1.69, "learning_rate": 1.8332005156138642e-05, "loss": 0.1037, "step": 9201 }, { "epoch": 1.69, "learning_rate": 1.833159459091969e-05, "loss": 0.0952, "step": 9202 }, { "epoch": 1.69, "learning_rate": 1.833118397977679e-05, "loss": 0.0904, "step": 9203 }, { "epoch": 1.69, "learning_rate": 1.833077332271221e-05, "loss": 0.087, "step": 9204 }, { "epoch": 1.69, "learning_rate": 1.8330362619728216e-05, "loss": 0.1032, "step": 9205 }, { "epoch": 1.69, "learning_rate": 1.8329951870827067e-05, "loss": 0.0943, "step": 9206 }, { "epoch": 1.69, "learning_rate": 1.832954107601103e-05, "loss": 0.0965, "step": 9207 }, { "epoch": 1.69, "learning_rate": 1.8329130235282364e-05, "loss": 0.1032, "step": 9208 }, { "epoch": 1.69, "learning_rate": 1.8328719348643337e-05, "loss": 0.0894, "step": 9209 }, { "epoch": 1.69, "learning_rate": 1.832830841609621e-05, "loss": 0.0995, "step": 9210 }, { "epoch": 1.69, "learning_rate": 1.8327897437643258e-05, "loss": 0.0885, "step": 9211 }, { "epoch": 1.69, "learning_rate": 1.832748641328674e-05, "loss": 0.0979, "step": 9212 }, { "epoch": 1.69, "learning_rate": 1.832707534302892e-05, "loss": 0.1034, "step": 9213 }, { "epoch": 1.69, "learning_rate": 1.8326664226872063e-05, "loss": 0.0954, "step": 9214 }, { "epoch": 1.69, "learning_rate": 1.8326253064818443e-05, "loss": 0.0948, "step": 9215 }, { "epoch": 1.69, "learning_rate": 1.832584185687032e-05, "loss": 0.0917, "step": 9216 }, { "epoch": 1.69, "learning_rate": 1.832543060302996e-05, "loss": 0.0873, "step": 9217 }, { "epoch": 1.69, "learning_rate": 1.8325019303299632e-05, "loss": 0.0923, "step": 9218 }, { "epoch": 1.69, "learning_rate": 1.8324607957681603e-05, "loss": 0.0874, "step": 9219 }, { "epoch": 1.69, "learning_rate": 1.832419656617814e-05, "loss": 0.1074, "step": 9220 }, { "epoch": 1.69, "learning_rate": 1.8323785128791507e-05, "loss": 0.1058, "step": 9221 }, { "epoch": 1.69, "learning_rate": 1.8323373645523976e-05, "loss": 0.0914, "step": 9222 }, { "epoch": 1.69, "learning_rate": 1.8322962116377817e-05, "loss": 0.0811, "step": 9223 }, { "epoch": 1.69, "learning_rate": 1.8322550541355294e-05, "loss": 0.0913, "step": 9224 }, { "epoch": 1.69, "learning_rate": 1.8322138920458682e-05, "loss": 0.089, "step": 9225 }, { "epoch": 1.69, "learning_rate": 1.832172725369024e-05, "loss": 0.1051, "step": 9226 }, { "epoch": 1.69, "learning_rate": 1.8321315541052244e-05, "loss": 0.0958, "step": 9227 }, { "epoch": 1.69, "learning_rate": 1.8320903782546964e-05, "loss": 0.0922, "step": 9228 }, { "epoch": 1.69, "learning_rate": 1.8320491978176664e-05, "loss": 0.0982, "step": 9229 }, { "epoch": 1.69, "learning_rate": 1.832008012794362e-05, "loss": 0.0868, "step": 9230 }, { "epoch": 1.69, "learning_rate": 1.83196682318501e-05, "loss": 0.0891, "step": 9231 }, { "epoch": 1.69, "learning_rate": 1.8319256289898374e-05, "loss": 0.0957, "step": 9232 }, { "epoch": 1.69, "learning_rate": 1.8318844302090708e-05, "loss": 0.0892, "step": 9233 }, { "epoch": 1.69, "learning_rate": 1.831843226842938e-05, "loss": 0.0889, "step": 9234 }, { "epoch": 1.69, "learning_rate": 1.831802018891666e-05, "loss": 0.088, "step": 9235 }, { "epoch": 1.69, "learning_rate": 1.8317608063554814e-05, "loss": 0.089, "step": 9236 }, { "epoch": 1.69, "learning_rate": 1.831719589234612e-05, "loss": 0.0924, "step": 9237 }, { "epoch": 1.69, "learning_rate": 1.8316783675292847e-05, "loss": 0.1056, "step": 9238 }, { "epoch": 1.69, "learning_rate": 1.831637141239727e-05, "loss": 0.0939, "step": 9239 }, { "epoch": 1.69, "learning_rate": 1.8315959103661654e-05, "loss": 0.0832, "step": 9240 }, { "epoch": 1.69, "learning_rate": 1.831554674908828e-05, "loss": 0.11, "step": 9241 }, { "epoch": 1.69, "learning_rate": 1.8315134348679415e-05, "loss": 0.0931, "step": 9242 }, { "epoch": 1.7, "learning_rate": 1.831472190243734e-05, "loss": 0.11, "step": 9243 }, { "epoch": 1.7, "learning_rate": 1.831430941036432e-05, "loss": 0.1125, "step": 9244 }, { "epoch": 1.7, "learning_rate": 1.8313896872462628e-05, "loss": 0.0883, "step": 9245 }, { "epoch": 1.7, "learning_rate": 1.8313484288734544e-05, "loss": 0.1014, "step": 9246 }, { "epoch": 1.7, "learning_rate": 1.831307165918234e-05, "loss": 0.0868, "step": 9247 }, { "epoch": 1.7, "learning_rate": 1.8312658983808294e-05, "loss": 0.0977, "step": 9248 }, { "epoch": 1.7, "learning_rate": 1.8312246262614674e-05, "loss": 0.0988, "step": 9249 }, { "epoch": 1.7, "learning_rate": 1.8311833495603758e-05, "loss": 0.0944, "step": 9250 }, { "epoch": 1.7, "learning_rate": 1.831142068277782e-05, "loss": 0.0897, "step": 9251 }, { "epoch": 1.7, "learning_rate": 1.8311007824139138e-05, "loss": 0.11, "step": 9252 }, { "epoch": 1.7, "learning_rate": 1.8310594919689984e-05, "loss": 0.0859, "step": 9253 }, { "epoch": 1.7, "learning_rate": 1.8310181969432638e-05, "loss": 0.1016, "step": 9254 }, { "epoch": 1.7, "learning_rate": 1.8309768973369376e-05, "loss": 0.0921, "step": 9255 }, { "epoch": 1.7, "learning_rate": 1.830935593150247e-05, "loss": 0.0946, "step": 9256 }, { "epoch": 1.7, "learning_rate": 1.83089428438342e-05, "loss": 0.0868, "step": 9257 }, { "epoch": 1.7, "learning_rate": 1.8308529710366843e-05, "loss": 0.0984, "step": 9258 }, { "epoch": 1.7, "learning_rate": 1.8308116531102674e-05, "loss": 0.1036, "step": 9259 }, { "epoch": 1.7, "learning_rate": 1.830770330604397e-05, "loss": 0.1039, "step": 9260 }, { "epoch": 1.7, "learning_rate": 1.8307290035193016e-05, "loss": 0.1016, "step": 9261 }, { "epoch": 1.7, "learning_rate": 1.8306876718552083e-05, "loss": 0.1025, "step": 9262 }, { "epoch": 1.7, "learning_rate": 1.830646335612345e-05, "loss": 0.0996, "step": 9263 }, { "epoch": 1.7, "learning_rate": 1.8306049947909395e-05, "loss": 0.0933, "step": 9264 }, { "epoch": 1.7, "learning_rate": 1.83056364939122e-05, "loss": 0.0912, "step": 9265 }, { "epoch": 1.7, "learning_rate": 1.8305222994134143e-05, "loss": 0.1004, "step": 9266 }, { "epoch": 1.7, "learning_rate": 1.83048094485775e-05, "loss": 0.1034, "step": 9267 }, { "epoch": 1.7, "learning_rate": 1.8304395857244556e-05, "loss": 0.0941, "step": 9268 }, { "epoch": 1.7, "learning_rate": 1.8303982220137585e-05, "loss": 0.0958, "step": 9269 }, { "epoch": 1.7, "learning_rate": 1.830356853725887e-05, "loss": 0.0902, "step": 9270 }, { "epoch": 1.7, "learning_rate": 1.830315480861069e-05, "loss": 0.0988, "step": 9271 }, { "epoch": 1.7, "learning_rate": 1.830274103419533e-05, "loss": 0.0983, "step": 9272 }, { "epoch": 1.7, "learning_rate": 1.8302327214015064e-05, "loss": 0.0946, "step": 9273 }, { "epoch": 1.7, "learning_rate": 1.8301913348072175e-05, "loss": 0.1002, "step": 9274 }, { "epoch": 1.7, "learning_rate": 1.8301499436368946e-05, "loss": 0.0976, "step": 9275 }, { "epoch": 1.7, "learning_rate": 1.8301085478907653e-05, "loss": 0.0877, "step": 9276 }, { "epoch": 1.7, "learning_rate": 1.830067147569059e-05, "loss": 0.0863, "step": 9277 }, { "epoch": 1.7, "learning_rate": 1.8300257426720027e-05, "loss": 0.0912, "step": 9278 }, { "epoch": 1.7, "learning_rate": 1.829984333199825e-05, "loss": 0.0841, "step": 9279 }, { "epoch": 1.7, "learning_rate": 1.8299429191527545e-05, "loss": 0.0881, "step": 9280 }, { "epoch": 1.7, "learning_rate": 1.829901500531019e-05, "loss": 0.0952, "step": 9281 }, { "epoch": 1.7, "learning_rate": 1.829860077334847e-05, "loss": 0.0958, "step": 9282 }, { "epoch": 1.7, "learning_rate": 1.8298186495644672e-05, "loss": 0.0844, "step": 9283 }, { "epoch": 1.7, "learning_rate": 1.8297772172201073e-05, "loss": 0.0883, "step": 9284 }, { "epoch": 1.7, "learning_rate": 1.8297357803019958e-05, "loss": 0.0868, "step": 9285 }, { "epoch": 1.7, "learning_rate": 1.829694338810361e-05, "loss": 0.0937, "step": 9286 }, { "epoch": 1.7, "learning_rate": 1.829652892745432e-05, "loss": 0.0896, "step": 9287 }, { "epoch": 1.7, "learning_rate": 1.829611442107437e-05, "loss": 0.0993, "step": 9288 }, { "epoch": 1.7, "learning_rate": 1.8295699868966038e-05, "loss": 0.0901, "step": 9289 }, { "epoch": 1.7, "learning_rate": 1.8295285271131614e-05, "loss": 0.1011, "step": 9290 }, { "epoch": 1.7, "learning_rate": 1.8294870627573386e-05, "loss": 0.0846, "step": 9291 }, { "epoch": 1.7, "learning_rate": 1.829445593829364e-05, "loss": 0.0945, "step": 9292 }, { "epoch": 1.7, "learning_rate": 1.829404120329465e-05, "loss": 0.1014, "step": 9293 }, { "epoch": 1.7, "learning_rate": 1.8293626422578715e-05, "loss": 0.0946, "step": 9294 }, { "epoch": 1.7, "learning_rate": 1.829321159614812e-05, "loss": 0.0924, "step": 9295 }, { "epoch": 1.7, "learning_rate": 1.8292796724005143e-05, "loss": 0.0903, "step": 9296 }, { "epoch": 1.7, "learning_rate": 1.829238180615208e-05, "loss": 0.0848, "step": 9297 }, { "epoch": 1.71, "learning_rate": 1.8291966842591212e-05, "loss": 0.0892, "step": 9298 }, { "epoch": 1.71, "learning_rate": 1.829155183332483e-05, "loss": 0.0952, "step": 9299 }, { "epoch": 1.71, "learning_rate": 1.8291136778355218e-05, "loss": 0.0976, "step": 9300 }, { "epoch": 1.71, "learning_rate": 1.829072167768467e-05, "loss": 0.1053, "step": 9301 }, { "epoch": 1.71, "learning_rate": 1.8290306531315467e-05, "loss": 0.0985, "step": 9302 }, { "epoch": 1.71, "learning_rate": 1.8289891339249897e-05, "loss": 0.0931, "step": 9303 }, { "epoch": 1.71, "learning_rate": 1.8289476101490257e-05, "loss": 0.0911, "step": 9304 }, { "epoch": 1.71, "learning_rate": 1.8289060818038828e-05, "loss": 0.0871, "step": 9305 }, { "epoch": 1.71, "learning_rate": 1.82886454888979e-05, "loss": 0.0843, "step": 9306 }, { "epoch": 1.71, "learning_rate": 1.828823011406977e-05, "loss": 0.1002, "step": 9307 }, { "epoch": 1.71, "learning_rate": 1.8287814693556716e-05, "loss": 0.0919, "step": 9308 }, { "epoch": 1.71, "learning_rate": 1.8287399227361032e-05, "loss": 0.0889, "step": 9309 }, { "epoch": 1.71, "learning_rate": 1.828698371548501e-05, "loss": 0.0956, "step": 9310 }, { "epoch": 1.71, "learning_rate": 1.8286568157930944e-05, "loss": 0.0903, "step": 9311 }, { "epoch": 1.71, "learning_rate": 1.8286152554701115e-05, "loss": 0.103, "step": 9312 }, { "epoch": 1.71, "learning_rate": 1.8285736905797825e-05, "loss": 0.0947, "step": 9313 }, { "epoch": 1.71, "learning_rate": 1.8285321211223356e-05, "loss": 0.1178, "step": 9314 }, { "epoch": 1.71, "learning_rate": 1.828490547098e-05, "loss": 0.0927, "step": 9315 }, { "epoch": 1.71, "learning_rate": 1.8284489685070054e-05, "loss": 0.1048, "step": 9316 }, { "epoch": 1.71, "learning_rate": 1.8284073853495808e-05, "loss": 0.0788, "step": 9317 }, { "epoch": 1.71, "learning_rate": 1.8283657976259548e-05, "loss": 0.0934, "step": 9318 }, { "epoch": 1.71, "learning_rate": 1.8283242053363573e-05, "loss": 0.1005, "step": 9319 }, { "epoch": 1.71, "learning_rate": 1.8282826084810175e-05, "loss": 0.0931, "step": 9320 }, { "epoch": 1.71, "learning_rate": 1.8282410070601643e-05, "loss": 0.087, "step": 9321 }, { "epoch": 1.71, "learning_rate": 1.8281994010740275e-05, "loss": 0.0881, "step": 9322 }, { "epoch": 1.71, "learning_rate": 1.828157790522836e-05, "loss": 0.0959, "step": 9323 }, { "epoch": 1.71, "learning_rate": 1.82811617540682e-05, "loss": 0.0854, "step": 9324 }, { "epoch": 1.71, "learning_rate": 1.8280745557262077e-05, "loss": 0.1074, "step": 9325 }, { "epoch": 1.71, "learning_rate": 1.828032931481229e-05, "loss": 0.1113, "step": 9326 }, { "epoch": 1.71, "learning_rate": 1.8279913026721138e-05, "loss": 0.1051, "step": 9327 }, { "epoch": 1.71, "learning_rate": 1.8279496692990908e-05, "loss": 0.0835, "step": 9328 }, { "epoch": 1.71, "learning_rate": 1.8279080313623897e-05, "loss": 0.0846, "step": 9329 }, { "epoch": 1.71, "learning_rate": 1.8278663888622404e-05, "loss": 0.0901, "step": 9330 }, { "epoch": 1.71, "learning_rate": 1.827824741798872e-05, "loss": 0.0958, "step": 9331 }, { "epoch": 1.71, "learning_rate": 1.8277830901725145e-05, "loss": 0.1054, "step": 9332 }, { "epoch": 1.71, "learning_rate": 1.827741433983397e-05, "loss": 0.0973, "step": 9333 }, { "epoch": 1.71, "learning_rate": 1.8276997732317493e-05, "loss": 0.0945, "step": 9334 }, { "epoch": 1.71, "learning_rate": 1.827658107917801e-05, "loss": 0.0945, "step": 9335 }, { "epoch": 1.71, "learning_rate": 1.827616438041782e-05, "loss": 0.0921, "step": 9336 }, { "epoch": 1.71, "learning_rate": 1.827574763603922e-05, "loss": 0.0927, "step": 9337 }, { "epoch": 1.71, "learning_rate": 1.82753308460445e-05, "loss": 0.078, "step": 9338 }, { "epoch": 1.71, "learning_rate": 1.8274914010435966e-05, "loss": 0.0841, "step": 9339 }, { "epoch": 1.71, "learning_rate": 1.827449712921591e-05, "loss": 0.1002, "step": 9340 }, { "epoch": 1.71, "learning_rate": 1.8274080202386634e-05, "loss": 0.0891, "step": 9341 }, { "epoch": 1.71, "learning_rate": 1.8273663229950434e-05, "loss": 0.098, "step": 9342 }, { "epoch": 1.71, "learning_rate": 1.8273246211909604e-05, "loss": 0.0917, "step": 9343 }, { "epoch": 1.71, "learning_rate": 1.827282914826645e-05, "loss": 0.097, "step": 9344 }, { "epoch": 1.71, "learning_rate": 1.827241203902327e-05, "loss": 0.0974, "step": 9345 }, { "epoch": 1.71, "learning_rate": 1.827199488418236e-05, "loss": 0.1055, "step": 9346 }, { "epoch": 1.71, "learning_rate": 1.827157768374602e-05, "loss": 0.0973, "step": 9347 }, { "epoch": 1.71, "learning_rate": 1.827116043771655e-05, "loss": 0.0902, "step": 9348 }, { "epoch": 1.71, "learning_rate": 1.827074314609625e-05, "loss": 0.0999, "step": 9349 }, { "epoch": 1.71, "learning_rate": 1.827032580888742e-05, "loss": 0.0961, "step": 9350 }, { "epoch": 1.71, "learning_rate": 1.8269908426092363e-05, "loss": 0.0961, "step": 9351 }, { "epoch": 1.72, "learning_rate": 1.826949099771337e-05, "loss": 0.0914, "step": 9352 }, { "epoch": 1.72, "learning_rate": 1.8269073523752757e-05, "loss": 0.0986, "step": 9353 }, { "epoch": 1.72, "learning_rate": 1.826865600421281e-05, "loss": 0.0898, "step": 9354 }, { "epoch": 1.72, "learning_rate": 1.8268238439095842e-05, "loss": 0.0917, "step": 9355 }, { "epoch": 1.72, "learning_rate": 1.8267820828404148e-05, "loss": 0.0984, "step": 9356 }, { "epoch": 1.72, "learning_rate": 1.826740317214003e-05, "loss": 0.0898, "step": 9357 }, { "epoch": 1.72, "learning_rate": 1.8266985470305796e-05, "loss": 0.0959, "step": 9358 }, { "epoch": 1.72, "learning_rate": 1.8266567722903744e-05, "loss": 0.0949, "step": 9359 }, { "epoch": 1.72, "learning_rate": 1.8266149929936177e-05, "loss": 0.0911, "step": 9360 }, { "epoch": 1.72, "learning_rate": 1.8265732091405393e-05, "loss": 0.1033, "step": 9361 }, { "epoch": 1.72, "learning_rate": 1.8265314207313707e-05, "loss": 0.1052, "step": 9362 }, { "epoch": 1.72, "learning_rate": 1.826489627766341e-05, "loss": 0.0941, "step": 9363 }, { "epoch": 1.72, "learning_rate": 1.8264478302456815e-05, "loss": 0.0876, "step": 9364 }, { "epoch": 1.72, "learning_rate": 1.8264060281696223e-05, "loss": 0.0895, "step": 9365 }, { "epoch": 1.72, "learning_rate": 1.8263642215383935e-05, "loss": 0.1069, "step": 9366 }, { "epoch": 1.72, "learning_rate": 1.8263224103522253e-05, "loss": 0.1038, "step": 9367 }, { "epoch": 1.72, "learning_rate": 1.8262805946113492e-05, "loss": 0.0807, "step": 9368 }, { "epoch": 1.72, "learning_rate": 1.826238774315995e-05, "loss": 0.092, "step": 9369 }, { "epoch": 1.72, "learning_rate": 1.8261969494663932e-05, "loss": 0.0973, "step": 9370 }, { "epoch": 1.72, "learning_rate": 1.8261551200627744e-05, "loss": 0.0881, "step": 9371 }, { "epoch": 1.72, "learning_rate": 1.8261132861053695e-05, "loss": 0.0966, "step": 9372 }, { "epoch": 1.72, "learning_rate": 1.8260714475944087e-05, "loss": 0.0878, "step": 9373 }, { "epoch": 1.72, "learning_rate": 1.8260296045301227e-05, "loss": 0.0952, "step": 9374 }, { "epoch": 1.72, "learning_rate": 1.825987756912742e-05, "loss": 0.1041, "step": 9375 }, { "epoch": 1.72, "learning_rate": 1.8259459047424976e-05, "loss": 0.0965, "step": 9376 }, { "epoch": 1.72, "learning_rate": 1.82590404801962e-05, "loss": 0.0936, "step": 9377 }, { "epoch": 1.72, "learning_rate": 1.82586218674434e-05, "loss": 0.0913, "step": 9378 }, { "epoch": 1.72, "learning_rate": 1.8258203209168883e-05, "loss": 0.0881, "step": 9379 }, { "epoch": 1.72, "learning_rate": 1.8257784505374957e-05, "loss": 0.098, "step": 9380 }, { "epoch": 1.72, "learning_rate": 1.8257365756063926e-05, "loss": 0.1019, "step": 9381 }, { "epoch": 1.72, "learning_rate": 1.8256946961238102e-05, "loss": 0.093, "step": 9382 }, { "epoch": 1.72, "learning_rate": 1.8256528120899798e-05, "loss": 0.0844, "step": 9383 }, { "epoch": 1.72, "learning_rate": 1.8256109235051315e-05, "loss": 0.086, "step": 9384 }, { "epoch": 1.72, "learning_rate": 1.825569030369496e-05, "loss": 0.0873, "step": 9385 }, { "epoch": 1.72, "learning_rate": 1.8255271326833052e-05, "loss": 0.0943, "step": 9386 }, { "epoch": 1.72, "learning_rate": 1.8254852304467893e-05, "loss": 0.1022, "step": 9387 }, { "epoch": 1.72, "learning_rate": 1.8254433236601794e-05, "loss": 0.103, "step": 9388 }, { "epoch": 1.72, "learning_rate": 1.8254014123237066e-05, "loss": 0.0833, "step": 9389 }, { "epoch": 1.72, "learning_rate": 1.825359496437602e-05, "loss": 0.0904, "step": 9390 }, { "epoch": 1.72, "learning_rate": 1.8253175760020964e-05, "loss": 0.0961, "step": 9391 }, { "epoch": 1.72, "learning_rate": 1.825275651017421e-05, "loss": 0.0918, "step": 9392 }, { "epoch": 1.72, "learning_rate": 1.8252337214838068e-05, "loss": 0.1071, "step": 9393 }, { "epoch": 1.72, "learning_rate": 1.8251917874014854e-05, "loss": 0.0911, "step": 9394 }, { "epoch": 1.72, "learning_rate": 1.825149848770687e-05, "loss": 0.0947, "step": 9395 }, { "epoch": 1.72, "learning_rate": 1.8251079055916438e-05, "loss": 0.0962, "step": 9396 }, { "epoch": 1.72, "learning_rate": 1.825065957864586e-05, "loss": 0.0941, "step": 9397 }, { "epoch": 1.72, "learning_rate": 1.8250240055897455e-05, "loss": 0.091, "step": 9398 }, { "epoch": 1.72, "learning_rate": 1.8249820487673535e-05, "loss": 0.0935, "step": 9399 }, { "epoch": 1.72, "learning_rate": 1.824940087397641e-05, "loss": 0.0883, "step": 9400 }, { "epoch": 1.72, "learning_rate": 1.8248981214808395e-05, "loss": 0.0935, "step": 9401 }, { "epoch": 1.72, "learning_rate": 1.8248561510171803e-05, "loss": 0.1053, "step": 9402 }, { "epoch": 1.72, "learning_rate": 1.8248141760068943e-05, "loss": 0.0937, "step": 9403 }, { "epoch": 1.72, "learning_rate": 1.8247721964502135e-05, "loss": 0.0825, "step": 9404 }, { "epoch": 1.72, "learning_rate": 1.824730212347369e-05, "loss": 0.094, "step": 9405 }, { "epoch": 1.72, "learning_rate": 1.8246882236985924e-05, "loss": 0.0864, "step": 9406 }, { "epoch": 1.73, "learning_rate": 1.8246462305041147e-05, "loss": 0.0921, "step": 9407 }, { "epoch": 1.73, "learning_rate": 1.8246042327641678e-05, "loss": 0.0926, "step": 9408 }, { "epoch": 1.73, "learning_rate": 1.824562230478983e-05, "loss": 0.1008, "step": 9409 }, { "epoch": 1.73, "learning_rate": 1.8245202236487918e-05, "loss": 0.0825, "step": 9410 }, { "epoch": 1.73, "learning_rate": 1.824478212273826e-05, "loss": 0.0902, "step": 9411 }, { "epoch": 1.73, "learning_rate": 1.824436196354317e-05, "loss": 0.0952, "step": 9412 }, { "epoch": 1.73, "learning_rate": 1.8243941758904965e-05, "loss": 0.11, "step": 9413 }, { "epoch": 1.73, "learning_rate": 1.8243521508825956e-05, "loss": 0.0968, "step": 9414 }, { "epoch": 1.73, "learning_rate": 1.8243101213308464e-05, "loss": 0.1018, "step": 9415 }, { "epoch": 1.73, "learning_rate": 1.8242680872354807e-05, "loss": 0.0769, "step": 9416 }, { "epoch": 1.73, "learning_rate": 1.82422604859673e-05, "loss": 0.0965, "step": 9417 }, { "epoch": 1.73, "learning_rate": 1.824184005414826e-05, "loss": 0.09, "step": 9418 }, { "epoch": 1.73, "learning_rate": 1.82414195769e-05, "loss": 0.095, "step": 9419 }, { "epoch": 1.73, "learning_rate": 1.8240999054224845e-05, "loss": 0.111, "step": 9420 }, { "epoch": 1.73, "learning_rate": 1.8240578486125113e-05, "loss": 0.0964, "step": 9421 }, { "epoch": 1.73, "learning_rate": 1.8240157872603118e-05, "loss": 0.093, "step": 9422 }, { "epoch": 1.73, "learning_rate": 1.8239737213661176e-05, "loss": 0.093, "step": 9423 }, { "epoch": 1.73, "learning_rate": 1.8239316509301613e-05, "loss": 0.1087, "step": 9424 }, { "epoch": 1.73, "learning_rate": 1.8238895759526745e-05, "loss": 0.0984, "step": 9425 }, { "epoch": 1.73, "learning_rate": 1.823847496433889e-05, "loss": 0.1016, "step": 9426 }, { "epoch": 1.73, "learning_rate": 1.8238054123740365e-05, "loss": 0.094, "step": 9427 }, { "epoch": 1.73, "learning_rate": 1.8237633237733495e-05, "loss": 0.0962, "step": 9428 }, { "epoch": 1.73, "learning_rate": 1.82372123063206e-05, "loss": 0.0919, "step": 9429 }, { "epoch": 1.73, "learning_rate": 1.8236791329503995e-05, "loss": 0.0982, "step": 9430 }, { "epoch": 1.73, "learning_rate": 1.8236370307286004e-05, "loss": 0.0878, "step": 9431 }, { "epoch": 1.73, "learning_rate": 1.8235949239668945e-05, "loss": 0.0966, "step": 9432 }, { "epoch": 1.73, "learning_rate": 1.8235528126655144e-05, "loss": 0.094, "step": 9433 }, { "epoch": 1.73, "learning_rate": 1.8235106968246917e-05, "loss": 0.0921, "step": 9434 }, { "epoch": 1.73, "learning_rate": 1.823468576444659e-05, "loss": 0.0989, "step": 9435 }, { "epoch": 1.73, "learning_rate": 1.823426451525648e-05, "loss": 0.0943, "step": 9436 }, { "epoch": 1.73, "learning_rate": 1.823384322067891e-05, "loss": 0.0902, "step": 9437 }, { "epoch": 1.73, "learning_rate": 1.8233421880716203e-05, "loss": 0.091, "step": 9438 }, { "epoch": 1.73, "learning_rate": 1.823300049537068e-05, "loss": 0.0931, "step": 9439 }, { "epoch": 1.73, "learning_rate": 1.8232579064644672e-05, "loss": 0.094, "step": 9440 }, { "epoch": 1.73, "learning_rate": 1.823215758854049e-05, "loss": 0.0934, "step": 9441 }, { "epoch": 1.73, "learning_rate": 1.8231736067060465e-05, "loss": 0.0944, "step": 9442 }, { "epoch": 1.73, "learning_rate": 1.823131450020692e-05, "loss": 0.1074, "step": 9443 }, { "epoch": 1.73, "learning_rate": 1.8230892887982172e-05, "loss": 0.092, "step": 9444 }, { "epoch": 1.73, "learning_rate": 1.8230471230388552e-05, "loss": 0.0884, "step": 9445 }, { "epoch": 1.73, "learning_rate": 1.8230049527428382e-05, "loss": 0.0985, "step": 9446 }, { "epoch": 1.73, "learning_rate": 1.8229627779103984e-05, "loss": 0.1002, "step": 9447 }, { "epoch": 1.73, "learning_rate": 1.8229205985417685e-05, "loss": 0.0808, "step": 9448 }, { "epoch": 1.73, "learning_rate": 1.8228784146371813e-05, "loss": 0.0981, "step": 9449 }, { "epoch": 1.73, "learning_rate": 1.8228362261968686e-05, "loss": 0.1059, "step": 9450 }, { "epoch": 1.73, "learning_rate": 1.822794033221064e-05, "loss": 0.0904, "step": 9451 }, { "epoch": 1.73, "learning_rate": 1.8227518357099987e-05, "loss": 0.1015, "step": 9452 }, { "epoch": 1.73, "learning_rate": 1.822709633663906e-05, "loss": 0.0918, "step": 9453 }, { "epoch": 1.73, "learning_rate": 1.822667427083019e-05, "loss": 0.0931, "step": 9454 }, { "epoch": 1.73, "learning_rate": 1.8226252159675696e-05, "loss": 0.0933, "step": 9455 }, { "epoch": 1.73, "learning_rate": 1.822583000317791e-05, "loss": 0.0944, "step": 9456 }, { "epoch": 1.73, "learning_rate": 1.8225407801339154e-05, "loss": 0.085, "step": 9457 }, { "epoch": 1.73, "learning_rate": 1.822498555416176e-05, "loss": 0.0939, "step": 9458 }, { "epoch": 1.73, "learning_rate": 1.822456326164805e-05, "loss": 0.0967, "step": 9459 }, { "epoch": 1.73, "learning_rate": 1.8224140923800356e-05, "loss": 0.0883, "step": 9460 }, { "epoch": 1.74, "learning_rate": 1.8223718540621004e-05, "loss": 0.0915, "step": 9461 }, { "epoch": 1.74, "learning_rate": 1.822329611211232e-05, "loss": 0.0985, "step": 9462 }, { "epoch": 1.74, "learning_rate": 1.822287363827664e-05, "loss": 0.0938, "step": 9463 }, { "epoch": 1.74, "learning_rate": 1.8222451119116288e-05, "loss": 0.0914, "step": 9464 }, { "epoch": 1.74, "learning_rate": 1.822202855463359e-05, "loss": 0.0933, "step": 9465 }, { "epoch": 1.74, "learning_rate": 1.8221605944830875e-05, "loss": 0.092, "step": 9466 }, { "epoch": 1.74, "learning_rate": 1.822118328971048e-05, "loss": 0.088, "step": 9467 }, { "epoch": 1.74, "learning_rate": 1.822076058927473e-05, "loss": 0.0919, "step": 9468 }, { "epoch": 1.74, "learning_rate": 1.8220337843525953e-05, "loss": 0.0982, "step": 9469 }, { "epoch": 1.74, "learning_rate": 1.8219915052466484e-05, "loss": 0.0837, "step": 9470 }, { "epoch": 1.74, "learning_rate": 1.821949221609865e-05, "loss": 0.1074, "step": 9471 }, { "epoch": 1.74, "learning_rate": 1.821906933442478e-05, "loss": 0.0934, "step": 9472 }, { "epoch": 1.74, "learning_rate": 1.8218646407447206e-05, "loss": 0.0859, "step": 9473 }, { "epoch": 1.74, "learning_rate": 1.8218223435168266e-05, "loss": 0.092, "step": 9474 }, { "epoch": 1.74, "learning_rate": 1.8217800417590282e-05, "loss": 0.0919, "step": 9475 }, { "epoch": 1.74, "learning_rate": 1.8217377354715592e-05, "loss": 0.0986, "step": 9476 }, { "epoch": 1.74, "learning_rate": 1.8216954246546523e-05, "loss": 0.0982, "step": 9477 }, { "epoch": 1.74, "learning_rate": 1.821653109308541e-05, "loss": 0.0942, "step": 9478 }, { "epoch": 1.74, "learning_rate": 1.8216107894334586e-05, "loss": 0.0864, "step": 9479 }, { "epoch": 1.74, "learning_rate": 1.8215684650296384e-05, "loss": 0.103, "step": 9480 }, { "epoch": 1.74, "learning_rate": 1.8215261360973134e-05, "loss": 0.0889, "step": 9481 }, { "epoch": 1.74, "learning_rate": 1.821483802636717e-05, "loss": 0.0867, "step": 9482 }, { "epoch": 1.74, "learning_rate": 1.821441464648083e-05, "loss": 0.0884, "step": 9483 }, { "epoch": 1.74, "learning_rate": 1.8213991221316442e-05, "loss": 0.0909, "step": 9484 }, { "epoch": 1.74, "learning_rate": 1.8213567750876338e-05, "loss": 0.0955, "step": 9485 }, { "epoch": 1.74, "learning_rate": 1.8213144235162862e-05, "loss": 0.0964, "step": 9486 }, { "epoch": 1.74, "learning_rate": 1.821272067417834e-05, "loss": 0.1058, "step": 9487 }, { "epoch": 1.74, "learning_rate": 1.821229706792511e-05, "loss": 0.0911, "step": 9488 }, { "epoch": 1.74, "learning_rate": 1.8211873416405507e-05, "loss": 0.1007, "step": 9489 }, { "epoch": 1.74, "learning_rate": 1.8211449719621862e-05, "loss": 0.0939, "step": 9490 }, { "epoch": 1.74, "learning_rate": 1.8211025977576517e-05, "loss": 0.0901, "step": 9491 }, { "epoch": 1.74, "learning_rate": 1.8210602190271803e-05, "loss": 0.091, "step": 9492 }, { "epoch": 1.74, "learning_rate": 1.8210178357710057e-05, "loss": 0.0873, "step": 9493 }, { "epoch": 1.74, "learning_rate": 1.8209754479893615e-05, "loss": 0.0913, "step": 9494 }, { "epoch": 1.74, "learning_rate": 1.8209330556824817e-05, "loss": 0.0933, "step": 9495 }, { "epoch": 1.74, "learning_rate": 1.8208906588505994e-05, "loss": 0.1067, "step": 9496 }, { "epoch": 1.74, "learning_rate": 1.8208482574939487e-05, "loss": 0.1004, "step": 9497 }, { "epoch": 1.74, "learning_rate": 1.8208058516127632e-05, "loss": 0.1072, "step": 9498 }, { "epoch": 1.74, "learning_rate": 1.8207634412072765e-05, "loss": 0.0963, "step": 9499 }, { "epoch": 1.74, "learning_rate": 1.8207210262777225e-05, "loss": 0.0978, "step": 9500 }, { "epoch": 1.74, "learning_rate": 1.820678606824335e-05, "loss": 0.1043, "step": 9501 }, { "epoch": 1.74, "learning_rate": 1.820636182847348e-05, "loss": 0.0998, "step": 9502 }, { "epoch": 1.74, "learning_rate": 1.8205937543469948e-05, "loss": 0.0958, "step": 9503 }, { "epoch": 1.74, "learning_rate": 1.82055132132351e-05, "loss": 0.1004, "step": 9504 }, { "epoch": 1.74, "learning_rate": 1.8205088837771265e-05, "loss": 0.1041, "step": 9505 }, { "epoch": 1.74, "learning_rate": 1.8204664417080792e-05, "loss": 0.0819, "step": 9506 }, { "epoch": 1.74, "learning_rate": 1.8204239951166017e-05, "loss": 0.0905, "step": 9507 }, { "epoch": 1.74, "learning_rate": 1.8203815440029277e-05, "loss": 0.0972, "step": 9508 }, { "epoch": 1.74, "learning_rate": 1.8203390883672914e-05, "loss": 0.0901, "step": 9509 }, { "epoch": 1.74, "learning_rate": 1.8202966282099267e-05, "loss": 0.0955, "step": 9510 }, { "epoch": 1.74, "learning_rate": 1.820254163531068e-05, "loss": 0.105, "step": 9511 }, { "epoch": 1.74, "learning_rate": 1.8202116943309494e-05, "loss": 0.0937, "step": 9512 }, { "epoch": 1.74, "learning_rate": 1.820169220609804e-05, "loss": 0.0945, "step": 9513 }, { "epoch": 1.74, "learning_rate": 1.8201267423678675e-05, "loss": 0.0848, "step": 9514 }, { "epoch": 1.74, "learning_rate": 1.8200842596053726e-05, "loss": 0.0743, "step": 9515 }, { "epoch": 1.75, "learning_rate": 1.8200417723225542e-05, "loss": 0.0903, "step": 9516 }, { "epoch": 1.75, "learning_rate": 1.8199992805196465e-05, "loss": 0.1036, "step": 9517 }, { "epoch": 1.75, "learning_rate": 1.8199567841968836e-05, "loss": 0.0931, "step": 9518 }, { "epoch": 1.75, "learning_rate": 1.8199142833544993e-05, "loss": 0.0968, "step": 9519 }, { "epoch": 1.75, "learning_rate": 1.8198717779927286e-05, "loss": 0.0913, "step": 9520 }, { "epoch": 1.75, "learning_rate": 1.8198292681118054e-05, "loss": 0.0994, "step": 9521 }, { "epoch": 1.75, "learning_rate": 1.819786753711964e-05, "loss": 0.094, "step": 9522 }, { "epoch": 1.75, "learning_rate": 1.819744234793439e-05, "loss": 0.0972, "step": 9523 }, { "epoch": 1.75, "learning_rate": 1.819701711356464e-05, "loss": 0.0915, "step": 9524 }, { "epoch": 1.75, "learning_rate": 1.8196591834012743e-05, "loss": 0.101, "step": 9525 }, { "epoch": 1.75, "learning_rate": 1.8196166509281044e-05, "loss": 0.0935, "step": 9526 }, { "epoch": 1.75, "learning_rate": 1.8195741139371877e-05, "loss": 0.0801, "step": 9527 }, { "epoch": 1.75, "learning_rate": 1.8195315724287596e-05, "loss": 0.0981, "step": 9528 }, { "epoch": 1.75, "learning_rate": 1.819489026403054e-05, "loss": 0.0908, "step": 9529 }, { "epoch": 1.75, "learning_rate": 1.819446475860306e-05, "loss": 0.0819, "step": 9530 }, { "epoch": 1.75, "learning_rate": 1.81940392080075e-05, "loss": 0.0967, "step": 9531 }, { "epoch": 1.75, "learning_rate": 1.81936136122462e-05, "loss": 0.0928, "step": 9532 }, { "epoch": 1.75, "learning_rate": 1.8193187971321512e-05, "loss": 0.0966, "step": 9533 }, { "epoch": 1.75, "learning_rate": 1.819276228523578e-05, "loss": 0.1102, "step": 9534 }, { "epoch": 1.75, "learning_rate": 1.819233655399135e-05, "loss": 0.1012, "step": 9535 }, { "epoch": 1.75, "learning_rate": 1.8191910777590566e-05, "loss": 0.0923, "step": 9536 }, { "epoch": 1.75, "learning_rate": 1.8191484956035777e-05, "loss": 0.0962, "step": 9537 }, { "epoch": 1.75, "learning_rate": 1.8191059089329335e-05, "loss": 0.0971, "step": 9538 }, { "epoch": 1.75, "learning_rate": 1.819063317747358e-05, "loss": 0.0869, "step": 9539 }, { "epoch": 1.75, "learning_rate": 1.819020722047087e-05, "loss": 0.0884, "step": 9540 }, { "epoch": 1.75, "learning_rate": 1.818978121832354e-05, "loss": 0.0932, "step": 9541 }, { "epoch": 1.75, "learning_rate": 1.8189355171033946e-05, "loss": 0.0862, "step": 9542 }, { "epoch": 1.75, "learning_rate": 1.8188929078604435e-05, "loss": 0.1035, "step": 9543 }, { "epoch": 1.75, "learning_rate": 1.8188502941037354e-05, "loss": 0.0939, "step": 9544 }, { "epoch": 1.75, "learning_rate": 1.8188076758335053e-05, "loss": 0.1097, "step": 9545 }, { "epoch": 1.75, "learning_rate": 1.818765053049988e-05, "loss": 0.0987, "step": 9546 }, { "epoch": 1.75, "learning_rate": 1.8187224257534188e-05, "loss": 0.0953, "step": 9547 }, { "epoch": 1.75, "learning_rate": 1.818679793944032e-05, "loss": 0.0965, "step": 9548 }, { "epoch": 1.75, "learning_rate": 1.8186371576220634e-05, "loss": 0.0984, "step": 9549 }, { "epoch": 1.75, "learning_rate": 1.8185945167877476e-05, "loss": 0.1026, "step": 9550 }, { "epoch": 1.75, "learning_rate": 1.8185518714413195e-05, "loss": 0.078, "step": 9551 }, { "epoch": 1.75, "learning_rate": 1.8185092215830142e-05, "loss": 0.109, "step": 9552 }, { "epoch": 1.75, "learning_rate": 1.818466567213067e-05, "loss": 0.0829, "step": 9553 }, { "epoch": 1.75, "learning_rate": 1.818423908331713e-05, "loss": 0.101, "step": 9554 }, { "epoch": 1.75, "learning_rate": 1.818381244939187e-05, "loss": 0.1001, "step": 9555 }, { "epoch": 1.75, "learning_rate": 1.8183385770357246e-05, "loss": 0.084, "step": 9556 }, { "epoch": 1.75, "learning_rate": 1.8182959046215608e-05, "loss": 0.0879, "step": 9557 }, { "epoch": 1.75, "learning_rate": 1.8182532276969306e-05, "loss": 0.0974, "step": 9558 }, { "epoch": 1.75, "learning_rate": 1.81821054626207e-05, "loss": 0.095, "step": 9559 }, { "epoch": 1.75, "learning_rate": 1.818167860317213e-05, "loss": 0.0844, "step": 9560 }, { "epoch": 1.75, "learning_rate": 1.818125169862596e-05, "loss": 0.0866, "step": 9561 }, { "epoch": 1.75, "learning_rate": 1.818082474898454e-05, "loss": 0.0867, "step": 9562 }, { "epoch": 1.75, "learning_rate": 1.818039775425022e-05, "loss": 0.0922, "step": 9563 }, { "epoch": 1.75, "learning_rate": 1.8179970714425355e-05, "loss": 0.0893, "step": 9564 }, { "epoch": 1.75, "learning_rate": 1.81795436295123e-05, "loss": 0.0889, "step": 9565 }, { "epoch": 1.75, "learning_rate": 1.817911649951341e-05, "loss": 0.1051, "step": 9566 }, { "epoch": 1.75, "learning_rate": 1.817868932443104e-05, "loss": 0.0926, "step": 9567 }, { "epoch": 1.75, "learning_rate": 1.817826210426754e-05, "loss": 0.1059, "step": 9568 }, { "epoch": 1.75, "learning_rate": 1.8177834839025268e-05, "loss": 0.0939, "step": 9569 }, { "epoch": 1.75, "learning_rate": 1.8177407528706577e-05, "loss": 0.1107, "step": 9570 }, { "epoch": 1.76, "learning_rate": 1.8176980173313828e-05, "loss": 0.0987, "step": 9571 }, { "epoch": 1.76, "learning_rate": 1.817655277284937e-05, "loss": 0.0947, "step": 9572 }, { "epoch": 1.76, "learning_rate": 1.8176125327315563e-05, "loss": 0.094, "step": 9573 }, { "epoch": 1.76, "learning_rate": 1.8175697836714766e-05, "loss": 0.0948, "step": 9574 }, { "epoch": 1.76, "learning_rate": 1.8175270301049326e-05, "loss": 0.0972, "step": 9575 }, { "epoch": 1.76, "learning_rate": 1.8174842720321605e-05, "loss": 0.0855, "step": 9576 }, { "epoch": 1.76, "learning_rate": 1.8174415094533957e-05, "loss": 0.097, "step": 9577 }, { "epoch": 1.76, "learning_rate": 1.8173987423688743e-05, "loss": 0.0951, "step": 9578 }, { "epoch": 1.76, "learning_rate": 1.8173559707788323e-05, "loss": 0.0919, "step": 9579 }, { "epoch": 1.76, "learning_rate": 1.8173131946835047e-05, "loss": 0.0863, "step": 9580 }, { "epoch": 1.76, "learning_rate": 1.8172704140831273e-05, "loss": 0.0958, "step": 9581 }, { "epoch": 1.76, "learning_rate": 1.8172276289779366e-05, "loss": 0.1011, "step": 9582 }, { "epoch": 1.76, "learning_rate": 1.8171848393681682e-05, "loss": 0.0938, "step": 9583 }, { "epoch": 1.76, "learning_rate": 1.8171420452540574e-05, "loss": 0.093, "step": 9584 }, { "epoch": 1.76, "learning_rate": 1.817099246635841e-05, "loss": 0.0973, "step": 9585 }, { "epoch": 1.76, "learning_rate": 1.8170564435137542e-05, "loss": 0.0994, "step": 9586 }, { "epoch": 1.76, "learning_rate": 1.817013635888033e-05, "loss": 0.093, "step": 9587 }, { "epoch": 1.76, "learning_rate": 1.8169708237589135e-05, "loss": 0.0845, "step": 9588 }, { "epoch": 1.76, "learning_rate": 1.8169280071266317e-05, "loss": 0.103, "step": 9589 }, { "epoch": 1.76, "learning_rate": 1.816885185991424e-05, "loss": 0.091, "step": 9590 }, { "epoch": 1.76, "learning_rate": 1.8168423603535258e-05, "loss": 0.108, "step": 9591 }, { "epoch": 1.76, "learning_rate": 1.8167995302131732e-05, "loss": 0.0971, "step": 9592 }, { "epoch": 1.76, "learning_rate": 1.8167566955706023e-05, "loss": 0.1, "step": 9593 }, { "epoch": 1.76, "learning_rate": 1.8167138564260496e-05, "loss": 0.0864, "step": 9594 }, { "epoch": 1.76, "learning_rate": 1.816671012779751e-05, "loss": 0.1013, "step": 9595 }, { "epoch": 1.76, "learning_rate": 1.8166281646319424e-05, "loss": 0.0854, "step": 9596 }, { "epoch": 1.76, "learning_rate": 1.8165853119828606e-05, "loss": 0.0915, "step": 9597 }, { "epoch": 1.76, "learning_rate": 1.816542454832741e-05, "loss": 0.0892, "step": 9598 }, { "epoch": 1.76, "learning_rate": 1.8164995931818207e-05, "loss": 0.0968, "step": 9599 }, { "epoch": 1.76, "learning_rate": 1.8164567270303354e-05, "loss": 0.09, "step": 9600 }, { "epoch": 1.76, "learning_rate": 1.8164138563785213e-05, "loss": 0.0984, "step": 9601 }, { "epoch": 1.76, "learning_rate": 1.8163709812266153e-05, "loss": 0.0928, "step": 9602 }, { "epoch": 1.76, "learning_rate": 1.8163281015748532e-05, "loss": 0.0937, "step": 9603 }, { "epoch": 1.76, "learning_rate": 1.8162852174234712e-05, "loss": 0.0918, "step": 9604 }, { "epoch": 1.76, "learning_rate": 1.816242328772706e-05, "loss": 0.1054, "step": 9605 }, { "epoch": 1.76, "learning_rate": 1.8161994356227943e-05, "loss": 0.0934, "step": 9606 }, { "epoch": 1.76, "learning_rate": 1.8161565379739717e-05, "loss": 0.0991, "step": 9607 }, { "epoch": 1.76, "learning_rate": 1.816113635826475e-05, "loss": 0.0986, "step": 9608 }, { "epoch": 1.76, "learning_rate": 1.8160707291805418e-05, "loss": 0.0904, "step": 9609 }, { "epoch": 1.76, "learning_rate": 1.8160278180364068e-05, "loss": 0.0944, "step": 9610 }, { "epoch": 1.76, "learning_rate": 1.8159849023943076e-05, "loss": 0.1067, "step": 9611 }, { "epoch": 1.76, "learning_rate": 1.8159419822544808e-05, "loss": 0.0973, "step": 9612 }, { "epoch": 1.76, "learning_rate": 1.815899057617162e-05, "loss": 0.0983, "step": 9613 }, { "epoch": 1.76, "learning_rate": 1.815856128482589e-05, "loss": 0.0868, "step": 9614 }, { "epoch": 1.76, "learning_rate": 1.8158131948509977e-05, "loss": 0.0904, "step": 9615 }, { "epoch": 1.76, "learning_rate": 1.8157702567226248e-05, "loss": 0.0826, "step": 9616 }, { "epoch": 1.76, "learning_rate": 1.8157273140977073e-05, "loss": 0.1007, "step": 9617 }, { "epoch": 1.76, "learning_rate": 1.8156843669764818e-05, "loss": 0.0947, "step": 9618 }, { "epoch": 1.76, "learning_rate": 1.8156414153591847e-05, "loss": 0.0869, "step": 9619 }, { "epoch": 1.76, "learning_rate": 1.8155984592460532e-05, "loss": 0.0929, "step": 9620 }, { "epoch": 1.76, "learning_rate": 1.8155554986373242e-05, "loss": 0.0996, "step": 9621 }, { "epoch": 1.76, "learning_rate": 1.8155125335332338e-05, "loss": 0.0953, "step": 9622 }, { "epoch": 1.76, "learning_rate": 1.8154695639340195e-05, "loss": 0.0919, "step": 9623 }, { "epoch": 1.76, "learning_rate": 1.8154265898399178e-05, "loss": 0.0849, "step": 9624 }, { "epoch": 1.77, "learning_rate": 1.8153836112511655e-05, "loss": 0.0941, "step": 9625 }, { "epoch": 1.77, "learning_rate": 1.8153406281679995e-05, "loss": 0.0979, "step": 9626 }, { "epoch": 1.77, "learning_rate": 1.8152976405906572e-05, "loss": 0.0869, "step": 9627 }, { "epoch": 1.77, "learning_rate": 1.815254648519375e-05, "loss": 0.083, "step": 9628 }, { "epoch": 1.77, "learning_rate": 1.81521165195439e-05, "loss": 0.1, "step": 9629 }, { "epoch": 1.77, "learning_rate": 1.8151686508959394e-05, "loss": 0.0966, "step": 9630 }, { "epoch": 1.77, "learning_rate": 1.81512564534426e-05, "loss": 0.1039, "step": 9631 }, { "epoch": 1.77, "learning_rate": 1.815082635299589e-05, "loss": 0.0858, "step": 9632 }, { "epoch": 1.77, "learning_rate": 1.8150396207621633e-05, "loss": 0.103, "step": 9633 }, { "epoch": 1.77, "learning_rate": 1.81499660173222e-05, "loss": 0.0981, "step": 9634 }, { "epoch": 1.77, "learning_rate": 1.814953578209997e-05, "loss": 0.1046, "step": 9635 }, { "epoch": 1.77, "learning_rate": 1.8149105501957304e-05, "loss": 0.0919, "step": 9636 }, { "epoch": 1.77, "learning_rate": 1.8148675176896576e-05, "loss": 0.086, "step": 9637 }, { "epoch": 1.77, "learning_rate": 1.8148244806920158e-05, "loss": 0.0852, "step": 9638 }, { "epoch": 1.77, "learning_rate": 1.814781439203043e-05, "loss": 0.0892, "step": 9639 }, { "epoch": 1.77, "learning_rate": 1.8147383932229752e-05, "loss": 0.0976, "step": 9640 }, { "epoch": 1.77, "learning_rate": 1.8146953427520506e-05, "loss": 0.0945, "step": 9641 }, { "epoch": 1.77, "learning_rate": 1.8146522877905062e-05, "loss": 0.0699, "step": 9642 }, { "epoch": 1.77, "learning_rate": 1.8146092283385793e-05, "loss": 0.1007, "step": 9643 }, { "epoch": 1.77, "learning_rate": 1.814566164396507e-05, "loss": 0.0899, "step": 9644 }, { "epoch": 1.77, "learning_rate": 1.8145230959645272e-05, "loss": 0.099, "step": 9645 }, { "epoch": 1.77, "learning_rate": 1.814480023042877e-05, "loss": 0.1345, "step": 9646 }, { "epoch": 1.77, "learning_rate": 1.8144369456317937e-05, "loss": 0.1134, "step": 9647 }, { "epoch": 1.77, "learning_rate": 1.814393863731515e-05, "loss": 0.0969, "step": 9648 }, { "epoch": 1.77, "learning_rate": 1.8143507773422778e-05, "loss": 0.1113, "step": 9649 }, { "epoch": 1.77, "learning_rate": 1.8143076864643206e-05, "loss": 0.1078, "step": 9650 }, { "epoch": 1.77, "learning_rate": 1.8142645910978803e-05, "loss": 0.0933, "step": 9651 }, { "epoch": 1.77, "learning_rate": 1.814221491243194e-05, "loss": 0.0939, "step": 9652 }, { "epoch": 1.77, "learning_rate": 1.8141783869005002e-05, "loss": 0.1029, "step": 9653 }, { "epoch": 1.77, "learning_rate": 1.814135278070036e-05, "loss": 0.1042, "step": 9654 }, { "epoch": 1.77, "learning_rate": 1.8140921647520393e-05, "loss": 0.096, "step": 9655 }, { "epoch": 1.77, "learning_rate": 1.8140490469467472e-05, "loss": 0.0944, "step": 9656 }, { "epoch": 1.77, "learning_rate": 1.8140059246543976e-05, "loss": 0.0907, "step": 9657 }, { "epoch": 1.77, "learning_rate": 1.8139627978752285e-05, "loss": 0.0993, "step": 9658 }, { "epoch": 1.77, "learning_rate": 1.8139196666094773e-05, "loss": 0.0876, "step": 9659 }, { "epoch": 1.77, "learning_rate": 1.813876530857382e-05, "loss": 0.1069, "step": 9660 }, { "epoch": 1.77, "learning_rate": 1.81383339061918e-05, "loss": 0.0904, "step": 9661 }, { "epoch": 1.77, "learning_rate": 1.8137902458951094e-05, "loss": 0.0917, "step": 9662 }, { "epoch": 1.77, "learning_rate": 1.8137470966854077e-05, "loss": 0.0921, "step": 9663 }, { "epoch": 1.77, "learning_rate": 1.813703942990313e-05, "loss": 0.0908, "step": 9664 }, { "epoch": 1.77, "learning_rate": 1.813660784810063e-05, "loss": 0.0875, "step": 9665 }, { "epoch": 1.77, "learning_rate": 1.813617622144896e-05, "loss": 0.0943, "step": 9666 }, { "epoch": 1.77, "learning_rate": 1.8135744549950497e-05, "loss": 0.1011, "step": 9667 }, { "epoch": 1.77, "learning_rate": 1.8135312833607616e-05, "loss": 0.0956, "step": 9668 }, { "epoch": 1.77, "learning_rate": 1.8134881072422702e-05, "loss": 0.0899, "step": 9669 }, { "epoch": 1.77, "learning_rate": 1.813444926639813e-05, "loss": 0.0894, "step": 9670 }, { "epoch": 1.77, "learning_rate": 1.8134017415536288e-05, "loss": 0.093, "step": 9671 }, { "epoch": 1.77, "learning_rate": 1.8133585519839547e-05, "loss": 0.0949, "step": 9672 }, { "epoch": 1.77, "learning_rate": 1.813315357931029e-05, "loss": 0.0943, "step": 9673 }, { "epoch": 1.77, "learning_rate": 1.8132721593950905e-05, "loss": 0.0903, "step": 9674 }, { "epoch": 1.77, "learning_rate": 1.813228956376377e-05, "loss": 0.1033, "step": 9675 }, { "epoch": 1.77, "learning_rate": 1.813185748875126e-05, "loss": 0.0868, "step": 9676 }, { "epoch": 1.77, "learning_rate": 1.8131425368915763e-05, "loss": 0.0956, "step": 9677 }, { "epoch": 1.77, "learning_rate": 1.8130993204259658e-05, "loss": 0.0954, "step": 9678 }, { "epoch": 1.77, "learning_rate": 1.8130560994785325e-05, "loss": 0.0873, "step": 9679 }, { "epoch": 1.78, "learning_rate": 1.8130128740495153e-05, "loss": 0.1139, "step": 9680 }, { "epoch": 1.78, "learning_rate": 1.812969644139152e-05, "loss": 0.0939, "step": 9681 }, { "epoch": 1.78, "learning_rate": 1.812926409747681e-05, "loss": 0.097, "step": 9682 }, { "epoch": 1.78, "learning_rate": 1.8128831708753406e-05, "loss": 0.0942, "step": 9683 }, { "epoch": 1.78, "learning_rate": 1.8128399275223693e-05, "loss": 0.0842, "step": 9684 }, { "epoch": 1.78, "learning_rate": 1.8127966796890047e-05, "loss": 0.0932, "step": 9685 }, { "epoch": 1.78, "learning_rate": 1.8127534273754864e-05, "loss": 0.0903, "step": 9686 }, { "epoch": 1.78, "learning_rate": 1.8127101705820518e-05, "loss": 0.1002, "step": 9687 }, { "epoch": 1.78, "learning_rate": 1.81266690930894e-05, "loss": 0.0997, "step": 9688 }, { "epoch": 1.78, "learning_rate": 1.812623643556389e-05, "loss": 0.0877, "step": 9689 }, { "epoch": 1.78, "learning_rate": 1.8125803733246373e-05, "loss": 0.0767, "step": 9690 }, { "epoch": 1.78, "learning_rate": 1.8125370986139236e-05, "loss": 0.0947, "step": 9691 }, { "epoch": 1.78, "learning_rate": 1.8124938194244862e-05, "loss": 0.0933, "step": 9692 }, { "epoch": 1.78, "learning_rate": 1.8124505357565644e-05, "loss": 0.0951, "step": 9693 }, { "epoch": 1.78, "learning_rate": 1.8124072476103957e-05, "loss": 0.0938, "step": 9694 }, { "epoch": 1.78, "learning_rate": 1.8123639549862192e-05, "loss": 0.0845, "step": 9695 }, { "epoch": 1.78, "learning_rate": 1.812320657884274e-05, "loss": 0.0891, "step": 9696 }, { "epoch": 1.78, "learning_rate": 1.812277356304798e-05, "loss": 0.0978, "step": 9697 }, { "epoch": 1.78, "learning_rate": 1.81223405024803e-05, "loss": 0.0859, "step": 9698 }, { "epoch": 1.78, "learning_rate": 1.8121907397142093e-05, "loss": 0.0876, "step": 9699 }, { "epoch": 1.78, "learning_rate": 1.812147424703574e-05, "loss": 0.0953, "step": 9700 }, { "epoch": 1.78, "learning_rate": 1.812104105216363e-05, "loss": 0.0948, "step": 9701 }, { "epoch": 1.78, "learning_rate": 1.8120607812528152e-05, "loss": 0.0831, "step": 9702 }, { "epoch": 1.78, "learning_rate": 1.8120174528131692e-05, "loss": 0.0866, "step": 9703 }, { "epoch": 1.78, "learning_rate": 1.8119741198976643e-05, "loss": 0.0889, "step": 9704 }, { "epoch": 1.78, "learning_rate": 1.811930782506539e-05, "loss": 0.1025, "step": 9705 }, { "epoch": 1.78, "learning_rate": 1.811887440640032e-05, "loss": 0.1047, "step": 9706 }, { "epoch": 1.78, "learning_rate": 1.8118440942983825e-05, "loss": 0.0915, "step": 9707 }, { "epoch": 1.78, "learning_rate": 1.8118007434818293e-05, "loss": 0.0958, "step": 9708 }, { "epoch": 1.78, "learning_rate": 1.8117573881906114e-05, "loss": 0.0901, "step": 9709 }, { "epoch": 1.78, "learning_rate": 1.811714028424968e-05, "loss": 0.0977, "step": 9710 }, { "epoch": 1.78, "learning_rate": 1.8116706641851374e-05, "loss": 0.0973, "step": 9711 }, { "epoch": 1.78, "learning_rate": 1.8116272954713594e-05, "loss": 0.092, "step": 9712 }, { "epoch": 1.78, "learning_rate": 1.8115839222838726e-05, "loss": 0.0858, "step": 9713 }, { "epoch": 1.78, "learning_rate": 1.811540544622916e-05, "loss": 0.0933, "step": 9714 }, { "epoch": 1.78, "learning_rate": 1.8114971624887294e-05, "loss": 0.0997, "step": 9715 }, { "epoch": 1.78, "learning_rate": 1.8114537758815515e-05, "loss": 0.0832, "step": 9716 }, { "epoch": 1.78, "learning_rate": 1.8114103848016208e-05, "loss": 0.0972, "step": 9717 }, { "epoch": 1.78, "learning_rate": 1.8113669892491773e-05, "loss": 0.0831, "step": 9718 }, { "epoch": 1.78, "learning_rate": 1.81132358922446e-05, "loss": 0.0931, "step": 9719 }, { "epoch": 1.78, "learning_rate": 1.811280184727708e-05, "loss": 0.1033, "step": 9720 }, { "epoch": 1.78, "learning_rate": 1.8112367757591607e-05, "loss": 0.0805, "step": 9721 }, { "epoch": 1.78, "learning_rate": 1.811193362319057e-05, "loss": 0.0895, "step": 9722 }, { "epoch": 1.78, "learning_rate": 1.811149944407637e-05, "loss": 0.0964, "step": 9723 }, { "epoch": 1.78, "learning_rate": 1.811106522025139e-05, "loss": 0.1013, "step": 9724 }, { "epoch": 1.78, "learning_rate": 1.811063095171803e-05, "loss": 0.1014, "step": 9725 }, { "epoch": 1.78, "learning_rate": 1.8110196638478684e-05, "loss": 0.1004, "step": 9726 }, { "epoch": 1.78, "learning_rate": 1.8109762280535743e-05, "loss": 0.0848, "step": 9727 }, { "epoch": 1.78, "learning_rate": 1.81093278778916e-05, "loss": 0.106, "step": 9728 }, { "epoch": 1.78, "learning_rate": 1.8108893430548656e-05, "loss": 0.0968, "step": 9729 }, { "epoch": 1.78, "learning_rate": 1.81084589385093e-05, "loss": 0.1026, "step": 9730 }, { "epoch": 1.78, "learning_rate": 1.810802440177593e-05, "loss": 0.082, "step": 9731 }, { "epoch": 1.78, "learning_rate": 1.8107589820350938e-05, "loss": 0.0932, "step": 9732 }, { "epoch": 1.78, "learning_rate": 1.810715519423672e-05, "loss": 0.0876, "step": 9733 }, { "epoch": 1.79, "learning_rate": 1.8106720523435674e-05, "loss": 0.0973, "step": 9734 }, { "epoch": 1.79, "learning_rate": 1.8106285807950192e-05, "loss": 0.0987, "step": 9735 }, { "epoch": 1.79, "learning_rate": 1.8105851047782672e-05, "loss": 0.1025, "step": 9736 }, { "epoch": 1.79, "learning_rate": 1.8105416242935515e-05, "loss": 0.0825, "step": 9737 }, { "epoch": 1.79, "learning_rate": 1.8104981393411112e-05, "loss": 0.1042, "step": 9738 }, { "epoch": 1.79, "learning_rate": 1.810454649921186e-05, "loss": 0.0894, "step": 9739 }, { "epoch": 1.79, "learning_rate": 1.810411156034016e-05, "loss": 0.0944, "step": 9740 }, { "epoch": 1.79, "learning_rate": 1.8103676576798407e-05, "loss": 0.0931, "step": 9741 }, { "epoch": 1.79, "learning_rate": 1.8103241548588998e-05, "loss": 0.0869, "step": 9742 }, { "epoch": 1.79, "learning_rate": 1.8102806475714336e-05, "loss": 0.0838, "step": 9743 }, { "epoch": 1.79, "learning_rate": 1.810237135817681e-05, "loss": 0.0831, "step": 9744 }, { "epoch": 1.79, "learning_rate": 1.8101936195978825e-05, "loss": 0.0976, "step": 9745 }, { "epoch": 1.79, "learning_rate": 1.8101500989122778e-05, "loss": 0.0907, "step": 9746 }, { "epoch": 1.79, "learning_rate": 1.8101065737611066e-05, "loss": 0.0858, "step": 9747 }, { "epoch": 1.79, "learning_rate": 1.810063044144609e-05, "loss": 0.0909, "step": 9748 }, { "epoch": 1.79, "learning_rate": 1.810019510063025e-05, "loss": 0.0921, "step": 9749 }, { "epoch": 1.79, "learning_rate": 1.8099759715165947e-05, "loss": 0.092, "step": 9750 }, { "epoch": 1.79, "learning_rate": 1.8099324285055576e-05, "loss": 0.0924, "step": 9751 }, { "epoch": 1.79, "learning_rate": 1.8098888810301544e-05, "loss": 0.0805, "step": 9752 }, { "epoch": 1.79, "learning_rate": 1.809845329090624e-05, "loss": 0.0989, "step": 9753 }, { "epoch": 1.79, "learning_rate": 1.8098017726872076e-05, "loss": 0.0902, "step": 9754 }, { "epoch": 1.79, "learning_rate": 1.8097582118201447e-05, "loss": 0.0994, "step": 9755 }, { "epoch": 1.79, "learning_rate": 1.8097146464896756e-05, "loss": 0.1059, "step": 9756 }, { "epoch": 1.79, "learning_rate": 1.8096710766960405e-05, "loss": 0.0914, "step": 9757 }, { "epoch": 1.79, "learning_rate": 1.8096275024394796e-05, "loss": 0.0882, "step": 9758 }, { "epoch": 1.79, "learning_rate": 1.8095839237202328e-05, "loss": 0.0922, "step": 9759 }, { "epoch": 1.79, "learning_rate": 1.8095403405385402e-05, "loss": 0.0991, "step": 9760 }, { "epoch": 1.79, "learning_rate": 1.809496752894643e-05, "loss": 0.0984, "step": 9761 }, { "epoch": 1.79, "learning_rate": 1.80945316078878e-05, "loss": 0.0942, "step": 9762 }, { "epoch": 1.79, "learning_rate": 1.8094095642211924e-05, "loss": 0.0893, "step": 9763 }, { "epoch": 1.79, "learning_rate": 1.8093659631921206e-05, "loss": 0.1016, "step": 9764 }, { "epoch": 1.79, "learning_rate": 1.8093223577018042e-05, "loss": 0.0858, "step": 9765 }, { "epoch": 1.79, "learning_rate": 1.8092787477504844e-05, "loss": 0.0918, "step": 9766 }, { "epoch": 1.79, "learning_rate": 1.809235133338401e-05, "loss": 0.093, "step": 9767 }, { "epoch": 1.79, "learning_rate": 1.8091915144657943e-05, "loss": 0.086, "step": 9768 }, { "epoch": 1.79, "learning_rate": 1.8091478911329054e-05, "loss": 0.085, "step": 9769 }, { "epoch": 1.79, "learning_rate": 1.8091042633399744e-05, "loss": 0.1024, "step": 9770 }, { "epoch": 1.79, "learning_rate": 1.8090606310872414e-05, "loss": 0.1022, "step": 9771 }, { "epoch": 1.79, "learning_rate": 1.8090169943749477e-05, "loss": 0.0887, "step": 9772 }, { "epoch": 1.79, "learning_rate": 1.8089733532033328e-05, "loss": 0.0767, "step": 9773 }, { "epoch": 1.79, "learning_rate": 1.8089297075726384e-05, "loss": 0.0907, "step": 9774 }, { "epoch": 1.79, "learning_rate": 1.8088860574831045e-05, "loss": 0.1011, "step": 9775 }, { "epoch": 1.79, "learning_rate": 1.8088424029349712e-05, "loss": 0.0939, "step": 9776 }, { "epoch": 1.79, "learning_rate": 1.8087987439284796e-05, "loss": 0.0902, "step": 9777 }, { "epoch": 1.79, "learning_rate": 1.808755080463871e-05, "loss": 0.0818, "step": 9778 }, { "epoch": 1.79, "learning_rate": 1.808711412541385e-05, "loss": 0.0849, "step": 9779 }, { "epoch": 1.79, "learning_rate": 1.808667740161263e-05, "loss": 0.0942, "step": 9780 }, { "epoch": 1.79, "learning_rate": 1.8086240633237453e-05, "loss": 0.0868, "step": 9781 }, { "epoch": 1.79, "learning_rate": 1.808580382029073e-05, "loss": 0.0962, "step": 9782 }, { "epoch": 1.79, "learning_rate": 1.808536696277486e-05, "loss": 0.0917, "step": 9783 }, { "epoch": 1.79, "learning_rate": 1.8084930060692268e-05, "loss": 0.0854, "step": 9784 }, { "epoch": 1.79, "learning_rate": 1.8084493114045346e-05, "loss": 0.0916, "step": 9785 }, { "epoch": 1.79, "learning_rate": 1.808405612283651e-05, "loss": 0.0864, "step": 9786 }, { "epoch": 1.79, "learning_rate": 1.8083619087068166e-05, "loss": 0.085, "step": 9787 }, { "epoch": 1.79, "learning_rate": 1.808318200674273e-05, "loss": 0.0859, "step": 9788 }, { "epoch": 1.8, "learning_rate": 1.80827448818626e-05, "loss": 0.0854, "step": 9789 }, { "epoch": 1.8, "learning_rate": 1.808230771243019e-05, "loss": 0.0881, "step": 9790 }, { "epoch": 1.8, "learning_rate": 1.8081870498447914e-05, "loss": 0.0999, "step": 9791 }, { "epoch": 1.8, "learning_rate": 1.8081433239918172e-05, "loss": 0.09, "step": 9792 }, { "epoch": 1.8, "learning_rate": 1.8080995936843388e-05, "loss": 0.0764, "step": 9793 }, { "epoch": 1.8, "learning_rate": 1.8080558589225964e-05, "loss": 0.0932, "step": 9794 }, { "epoch": 1.8, "learning_rate": 1.808012119706831e-05, "loss": 0.0982, "step": 9795 }, { "epoch": 1.8, "learning_rate": 1.807968376037284e-05, "loss": 0.0843, "step": 9796 }, { "epoch": 1.8, "learning_rate": 1.807924627914196e-05, "loss": 0.0877, "step": 9797 }, { "epoch": 1.8, "learning_rate": 1.807880875337809e-05, "loss": 0.1035, "step": 9798 }, { "epoch": 1.8, "learning_rate": 1.8078371183083634e-05, "loss": 0.0901, "step": 9799 }, { "epoch": 1.8, "learning_rate": 1.8077933568261006e-05, "loss": 0.0959, "step": 9800 }, { "epoch": 1.8, "learning_rate": 1.807749590891262e-05, "loss": 0.1025, "step": 9801 }, { "epoch": 1.8, "learning_rate": 1.8077058205040888e-05, "loss": 0.0973, "step": 9802 }, { "epoch": 1.8, "learning_rate": 1.8076620456648225e-05, "loss": 0.076, "step": 9803 }, { "epoch": 1.8, "learning_rate": 1.8076182663737036e-05, "loss": 0.0958, "step": 9804 }, { "epoch": 1.8, "learning_rate": 1.807574482630974e-05, "loss": 0.0897, "step": 9805 }, { "epoch": 1.8, "learning_rate": 1.8075306944368752e-05, "loss": 0.097, "step": 9806 }, { "epoch": 1.8, "learning_rate": 1.807486901791648e-05, "loss": 0.0915, "step": 9807 }, { "epoch": 1.8, "learning_rate": 1.807443104695534e-05, "loss": 0.0929, "step": 9808 }, { "epoch": 1.8, "learning_rate": 1.807399303148775e-05, "loss": 0.0865, "step": 9809 }, { "epoch": 1.8, "learning_rate": 1.8073554971516118e-05, "loss": 0.0908, "step": 9810 }, { "epoch": 1.8, "learning_rate": 1.8073116867042863e-05, "loss": 0.0975, "step": 9811 }, { "epoch": 1.8, "learning_rate": 1.80726787180704e-05, "loss": 0.0954, "step": 9812 }, { "epoch": 1.8, "learning_rate": 1.8072240524601144e-05, "loss": 0.0974, "step": 9813 }, { "epoch": 1.8, "learning_rate": 1.8071802286637505e-05, "loss": 0.0864, "step": 9814 }, { "epoch": 1.8, "learning_rate": 1.8071364004181906e-05, "loss": 0.1031, "step": 9815 }, { "epoch": 1.8, "learning_rate": 1.807092567723676e-05, "loss": 0.0993, "step": 9816 }, { "epoch": 1.8, "learning_rate": 1.8070487305804478e-05, "loss": 0.0902, "step": 9817 }, { "epoch": 1.8, "learning_rate": 1.8070048889887484e-05, "loss": 0.0929, "step": 9818 }, { "epoch": 1.8, "learning_rate": 1.8069610429488193e-05, "loss": 0.0989, "step": 9819 }, { "epoch": 1.8, "learning_rate": 1.806917192460902e-05, "loss": 0.0927, "step": 9820 }, { "epoch": 1.8, "learning_rate": 1.806873337525238e-05, "loss": 0.0908, "step": 9821 }, { "epoch": 1.8, "learning_rate": 1.8068294781420692e-05, "loss": 0.0929, "step": 9822 }, { "epoch": 1.8, "learning_rate": 1.8067856143116377e-05, "loss": 0.0926, "step": 9823 }, { "epoch": 1.8, "learning_rate": 1.806741746034185e-05, "loss": 0.0889, "step": 9824 }, { "epoch": 1.8, "learning_rate": 1.8066978733099526e-05, "loss": 0.0997, "step": 9825 }, { "epoch": 1.8, "learning_rate": 1.8066539961391828e-05, "loss": 0.0851, "step": 9826 }, { "epoch": 1.8, "learning_rate": 1.8066101145221172e-05, "loss": 0.0956, "step": 9827 }, { "epoch": 1.8, "learning_rate": 1.8065662284589977e-05, "loss": 0.0996, "step": 9828 }, { "epoch": 1.8, "learning_rate": 1.8065223379500665e-05, "loss": 0.093, "step": 9829 }, { "epoch": 1.8, "learning_rate": 1.8064784429955652e-05, "loss": 0.0963, "step": 9830 }, { "epoch": 1.8, "learning_rate": 1.8064345435957357e-05, "loss": 0.1009, "step": 9831 }, { "epoch": 1.8, "learning_rate": 1.8063906397508203e-05, "loss": 0.0956, "step": 9832 }, { "epoch": 1.8, "learning_rate": 1.8063467314610605e-05, "loss": 0.09, "step": 9833 }, { "epoch": 1.8, "learning_rate": 1.8063028187266987e-05, "loss": 0.1048, "step": 9834 }, { "epoch": 1.8, "learning_rate": 1.806258901547977e-05, "loss": 0.0963, "step": 9835 }, { "epoch": 1.8, "learning_rate": 1.8062149799251373e-05, "loss": 0.1027, "step": 9836 }, { "epoch": 1.8, "learning_rate": 1.806171053858422e-05, "loss": 0.0759, "step": 9837 }, { "epoch": 1.8, "learning_rate": 1.8061271233480725e-05, "loss": 0.0995, "step": 9838 }, { "epoch": 1.8, "learning_rate": 1.806083188394332e-05, "loss": 0.0918, "step": 9839 }, { "epoch": 1.8, "learning_rate": 1.8060392489974413e-05, "loss": 0.0889, "step": 9840 }, { "epoch": 1.8, "learning_rate": 1.805995305157644e-05, "loss": 0.1002, "step": 9841 }, { "epoch": 1.8, "learning_rate": 1.8059513568751813e-05, "loss": 0.1131, "step": 9842 }, { "epoch": 1.81, "learning_rate": 1.805907404150296e-05, "loss": 0.0901, "step": 9843 }, { "epoch": 1.81, "learning_rate": 1.8058634469832306e-05, "loss": 0.0938, "step": 9844 }, { "epoch": 1.81, "learning_rate": 1.8058194853742264e-05, "loss": 0.0932, "step": 9845 }, { "epoch": 1.81, "learning_rate": 1.8057755193235268e-05, "loss": 0.0844, "step": 9846 }, { "epoch": 1.81, "learning_rate": 1.805731548831373e-05, "loss": 0.0944, "step": 9847 }, { "epoch": 1.81, "learning_rate": 1.8056875738980086e-05, "loss": 0.1047, "step": 9848 }, { "epoch": 1.81, "learning_rate": 1.8056435945236752e-05, "loss": 0.1021, "step": 9849 }, { "epoch": 1.81, "learning_rate": 1.8055996107086158e-05, "loss": 0.0913, "step": 9850 }, { "epoch": 1.81, "learning_rate": 1.8055556224530718e-05, "loss": 0.088, "step": 9851 }, { "epoch": 1.81, "learning_rate": 1.805511629757287e-05, "loss": 0.1012, "step": 9852 }, { "epoch": 1.81, "learning_rate": 1.8054676326215027e-05, "loss": 0.0939, "step": 9853 }, { "epoch": 1.81, "learning_rate": 1.8054236310459624e-05, "loss": 0.0917, "step": 9854 }, { "epoch": 1.81, "learning_rate": 1.805379625030908e-05, "loss": 0.0845, "step": 9855 }, { "epoch": 1.81, "learning_rate": 1.805335614576582e-05, "loss": 0.088, "step": 9856 }, { "epoch": 1.81, "learning_rate": 1.8052915996832275e-05, "loss": 0.0901, "step": 9857 }, { "epoch": 1.81, "learning_rate": 1.805247580351087e-05, "loss": 0.0961, "step": 9858 }, { "epoch": 1.81, "learning_rate": 1.8052035565804024e-05, "loss": 0.0942, "step": 9859 }, { "epoch": 1.81, "learning_rate": 1.805159528371417e-05, "loss": 0.097, "step": 9860 }, { "epoch": 1.81, "learning_rate": 1.805115495724374e-05, "loss": 0.0832, "step": 9861 }, { "epoch": 1.81, "learning_rate": 1.805071458639515e-05, "loss": 0.0867, "step": 9862 }, { "epoch": 1.81, "learning_rate": 1.8050274171170835e-05, "loss": 0.0948, "step": 9863 }, { "epoch": 1.81, "learning_rate": 1.804983371157322e-05, "loss": 0.107, "step": 9864 }, { "epoch": 1.81, "learning_rate": 1.8049393207604734e-05, "loss": 0.0959, "step": 9865 }, { "epoch": 1.81, "learning_rate": 1.80489526592678e-05, "loss": 0.0995, "step": 9866 }, { "epoch": 1.81, "learning_rate": 1.8048512066564855e-05, "loss": 0.0919, "step": 9867 }, { "epoch": 1.81, "learning_rate": 1.804807142949832e-05, "loss": 0.0894, "step": 9868 }, { "epoch": 1.81, "learning_rate": 1.8047630748070627e-05, "loss": 0.0956, "step": 9869 }, { "epoch": 1.81, "learning_rate": 1.8047190022284205e-05, "loss": 0.0887, "step": 9870 }, { "epoch": 1.81, "learning_rate": 1.8046749252141487e-05, "loss": 0.1046, "step": 9871 }, { "epoch": 1.81, "learning_rate": 1.8046308437644894e-05, "loss": 0.0907, "step": 9872 }, { "epoch": 1.81, "learning_rate": 1.8045867578796863e-05, "loss": 0.1004, "step": 9873 }, { "epoch": 1.81, "learning_rate": 1.8045426675599817e-05, "loss": 0.0975, "step": 9874 }, { "epoch": 1.81, "learning_rate": 1.8044985728056195e-05, "loss": 0.0989, "step": 9875 }, { "epoch": 1.81, "learning_rate": 1.8044544736168423e-05, "loss": 0.0836, "step": 9876 }, { "epoch": 1.81, "learning_rate": 1.804410369993893e-05, "loss": 0.0908, "step": 9877 }, { "epoch": 1.81, "learning_rate": 1.804366261937015e-05, "loss": 0.0812, "step": 9878 }, { "epoch": 1.81, "learning_rate": 1.8043221494464513e-05, "loss": 0.1042, "step": 9879 }, { "epoch": 1.81, "learning_rate": 1.804278032522445e-05, "loss": 0.0983, "step": 9880 }, { "epoch": 1.81, "learning_rate": 1.8042339111652395e-05, "loss": 0.0864, "step": 9881 }, { "epoch": 1.81, "learning_rate": 1.804189785375078e-05, "loss": 0.0975, "step": 9882 }, { "epoch": 1.81, "learning_rate": 1.804145655152203e-05, "loss": 0.0863, "step": 9883 }, { "epoch": 1.81, "learning_rate": 1.8041015204968587e-05, "loss": 0.0956, "step": 9884 }, { "epoch": 1.81, "learning_rate": 1.804057381409288e-05, "loss": 0.1084, "step": 9885 }, { "epoch": 1.81, "learning_rate": 1.804013237889734e-05, "loss": 0.0812, "step": 9886 }, { "epoch": 1.81, "learning_rate": 1.8039690899384403e-05, "loss": 0.1019, "step": 9887 }, { "epoch": 1.81, "learning_rate": 1.80392493755565e-05, "loss": 0.0941, "step": 9888 }, { "epoch": 1.81, "learning_rate": 1.803880780741607e-05, "loss": 0.0929, "step": 9889 }, { "epoch": 1.81, "learning_rate": 1.8038366194965538e-05, "loss": 0.0823, "step": 9890 }, { "epoch": 1.81, "learning_rate": 1.8037924538207345e-05, "loss": 0.0839, "step": 9891 }, { "epoch": 1.81, "learning_rate": 1.8037482837143928e-05, "loss": 0.0928, "step": 9892 }, { "epoch": 1.81, "learning_rate": 1.8037041091777713e-05, "loss": 0.0921, "step": 9893 }, { "epoch": 1.81, "learning_rate": 1.8036599302111136e-05, "loss": 0.0896, "step": 9894 }, { "epoch": 1.81, "learning_rate": 1.8036157468146644e-05, "loss": 0.0797, "step": 9895 }, { "epoch": 1.81, "learning_rate": 1.8035715589886655e-05, "loss": 0.0959, "step": 9896 }, { "epoch": 1.81, "learning_rate": 1.803527366733362e-05, "loss": 0.0884, "step": 9897 }, { "epoch": 1.82, "learning_rate": 1.8034831700489963e-05, "loss": 0.0832, "step": 9898 }, { "epoch": 1.82, "learning_rate": 1.8034389689358128e-05, "loss": 0.086, "step": 9899 }, { "epoch": 1.82, "learning_rate": 1.803394763394055e-05, "loss": 0.0999, "step": 9900 }, { "epoch": 1.82, "learning_rate": 1.8033505534239665e-05, "loss": 0.0928, "step": 9901 }, { "epoch": 1.82, "learning_rate": 1.8033063390257905e-05, "loss": 0.1063, "step": 9902 }, { "epoch": 1.82, "learning_rate": 1.8032621201997712e-05, "loss": 0.089, "step": 9903 }, { "epoch": 1.82, "learning_rate": 1.8032178969461527e-05, "loss": 0.0975, "step": 9904 }, { "epoch": 1.82, "learning_rate": 1.803173669265178e-05, "loss": 0.0984, "step": 9905 }, { "epoch": 1.82, "learning_rate": 1.8031294371570913e-05, "loss": 0.0976, "step": 9906 }, { "epoch": 1.82, "learning_rate": 1.8030852006221365e-05, "loss": 0.0964, "step": 9907 }, { "epoch": 1.82, "learning_rate": 1.803040959660557e-05, "loss": 0.0932, "step": 9908 }, { "epoch": 1.82, "learning_rate": 1.802996714272597e-05, "loss": 0.1064, "step": 9909 }, { "epoch": 1.82, "learning_rate": 1.8029524644585005e-05, "loss": 0.0969, "step": 9910 }, { "epoch": 1.82, "learning_rate": 1.802908210218511e-05, "loss": 0.0814, "step": 9911 }, { "epoch": 1.82, "learning_rate": 1.8028639515528726e-05, "loss": 0.106, "step": 9912 }, { "epoch": 1.82, "learning_rate": 1.8028196884618293e-05, "loss": 0.0896, "step": 9913 }, { "epoch": 1.82, "learning_rate": 1.8027754209456253e-05, "loss": 0.1063, "step": 9914 }, { "epoch": 1.82, "learning_rate": 1.802731149004504e-05, "loss": 0.0911, "step": 9915 }, { "epoch": 1.82, "learning_rate": 1.80268687263871e-05, "loss": 0.0942, "step": 9916 }, { "epoch": 1.82, "learning_rate": 1.8026425918484875e-05, "loss": 0.0964, "step": 9917 }, { "epoch": 1.82, "learning_rate": 1.8025983066340796e-05, "loss": 0.0861, "step": 9918 }, { "epoch": 1.82, "learning_rate": 1.8025540169957315e-05, "loss": 0.0837, "step": 9919 }, { "epoch": 1.82, "learning_rate": 1.8025097229336864e-05, "loss": 0.0893, "step": 9920 }, { "epoch": 1.82, "learning_rate": 1.8024654244481892e-05, "loss": 0.0929, "step": 9921 }, { "epoch": 1.82, "learning_rate": 1.8024211215394837e-05, "loss": 0.0946, "step": 9922 }, { "epoch": 1.82, "learning_rate": 1.8023768142078143e-05, "loss": 0.0866, "step": 9923 }, { "epoch": 1.82, "learning_rate": 1.802332502453425e-05, "loss": 0.1003, "step": 9924 }, { "epoch": 1.82, "learning_rate": 1.8022881862765603e-05, "loss": 0.0949, "step": 9925 }, { "epoch": 1.82, "learning_rate": 1.8022438656774642e-05, "loss": 0.0782, "step": 9926 }, { "epoch": 1.82, "learning_rate": 1.802199540656381e-05, "loss": 0.0823, "step": 9927 }, { "epoch": 1.82, "learning_rate": 1.8021552112135554e-05, "loss": 0.1075, "step": 9928 }, { "epoch": 1.82, "learning_rate": 1.802110877349231e-05, "loss": 0.0852, "step": 9929 }, { "epoch": 1.82, "learning_rate": 1.8020665390636534e-05, "loss": 0.0788, "step": 9930 }, { "epoch": 1.82, "learning_rate": 1.802022196357066e-05, "loss": 0.0943, "step": 9931 }, { "epoch": 1.82, "learning_rate": 1.801977849229713e-05, "loss": 0.0868, "step": 9932 }, { "epoch": 1.82, "learning_rate": 1.8019334976818393e-05, "loss": 0.0929, "step": 9933 }, { "epoch": 1.82, "learning_rate": 1.8018891417136903e-05, "loss": 0.1017, "step": 9934 }, { "epoch": 1.82, "learning_rate": 1.8018447813255088e-05, "loss": 0.0927, "step": 9935 }, { "epoch": 1.82, "learning_rate": 1.8018004165175403e-05, "loss": 0.0871, "step": 9936 }, { "epoch": 1.82, "learning_rate": 1.801756047290029e-05, "loss": 0.0884, "step": 9937 }, { "epoch": 1.82, "learning_rate": 1.8017116736432197e-05, "loss": 0.093, "step": 9938 }, { "epoch": 1.82, "learning_rate": 1.8016672955773566e-05, "loss": 0.0884, "step": 9939 }, { "epoch": 1.82, "learning_rate": 1.8016229130926846e-05, "loss": 0.0924, "step": 9940 }, { "epoch": 1.82, "learning_rate": 1.8015785261894486e-05, "loss": 0.0788, "step": 9941 }, { "epoch": 1.82, "learning_rate": 1.801534134867893e-05, "loss": 0.1006, "step": 9942 }, { "epoch": 1.82, "learning_rate": 1.8014897391282626e-05, "loss": 0.0876, "step": 9943 }, { "epoch": 1.82, "learning_rate": 1.8014453389708014e-05, "loss": 0.1116, "step": 9944 }, { "epoch": 1.82, "learning_rate": 1.8014009343957554e-05, "loss": 0.0891, "step": 9945 }, { "epoch": 1.82, "learning_rate": 1.8013565254033682e-05, "loss": 0.0845, "step": 9946 }, { "epoch": 1.82, "learning_rate": 1.8013121119938852e-05, "loss": 0.0849, "step": 9947 }, { "epoch": 1.82, "learning_rate": 1.8012676941675512e-05, "loss": 0.1057, "step": 9948 }, { "epoch": 1.82, "learning_rate": 1.801223271924611e-05, "loss": 0.084, "step": 9949 }, { "epoch": 1.82, "learning_rate": 1.801178845265309e-05, "loss": 0.099, "step": 9950 }, { "epoch": 1.82, "learning_rate": 1.80113441418989e-05, "loss": 0.0905, "step": 9951 }, { "epoch": 1.83, "learning_rate": 1.8010899786986004e-05, "loss": 0.0827, "step": 9952 }, { "epoch": 1.83, "learning_rate": 1.8010455387916834e-05, "loss": 0.0959, "step": 9953 }, { "epoch": 1.83, "learning_rate": 1.8010010944693846e-05, "loss": 0.1, "step": 9954 }, { "epoch": 1.83, "learning_rate": 1.8009566457319493e-05, "loss": 0.0903, "step": 9955 }, { "epoch": 1.83, "learning_rate": 1.800912192579622e-05, "loss": 0.0884, "step": 9956 }, { "epoch": 1.83, "learning_rate": 1.8008677350126483e-05, "loss": 0.0901, "step": 9957 }, { "epoch": 1.83, "learning_rate": 1.8008232730312724e-05, "loss": 0.099, "step": 9958 }, { "epoch": 1.83, "learning_rate": 1.8007788066357402e-05, "loss": 0.0813, "step": 9959 }, { "epoch": 1.83, "learning_rate": 1.8007343358262963e-05, "loss": 0.1024, "step": 9960 }, { "epoch": 1.83, "learning_rate": 1.800689860603186e-05, "loss": 0.0981, "step": 9961 }, { "epoch": 1.83, "learning_rate": 1.8006453809666545e-05, "loss": 0.0964, "step": 9962 }, { "epoch": 1.83, "learning_rate": 1.800600896916947e-05, "loss": 0.0976, "step": 9963 }, { "epoch": 1.83, "learning_rate": 1.800556408454308e-05, "loss": 0.0953, "step": 9964 }, { "epoch": 1.83, "learning_rate": 1.800511915578984e-05, "loss": 0.096, "step": 9965 }, { "epoch": 1.83, "learning_rate": 1.8004674182912194e-05, "loss": 0.0946, "step": 9966 }, { "epoch": 1.83, "learning_rate": 1.8004229165912598e-05, "loss": 0.1044, "step": 9967 }, { "epoch": 1.83, "learning_rate": 1.8003784104793498e-05, "loss": 0.0847, "step": 9968 }, { "epoch": 1.83, "learning_rate": 1.8003338999557355e-05, "loss": 0.0879, "step": 9969 }, { "epoch": 1.83, "learning_rate": 1.800289385020662e-05, "loss": 0.0957, "step": 9970 }, { "epoch": 1.83, "learning_rate": 1.8002448656743747e-05, "loss": 0.0865, "step": 9971 }, { "epoch": 1.83, "learning_rate": 1.800200341917119e-05, "loss": 0.0786, "step": 9972 }, { "epoch": 1.83, "learning_rate": 1.8001558137491403e-05, "loss": 0.0852, "step": 9973 }, { "epoch": 1.83, "learning_rate": 1.800111281170684e-05, "loss": 0.0947, "step": 9974 }, { "epoch": 1.83, "learning_rate": 1.8000667441819956e-05, "loss": 0.106, "step": 9975 }, { "epoch": 1.83, "learning_rate": 1.8000222027833202e-05, "loss": 0.0992, "step": 9976 }, { "epoch": 1.83, "learning_rate": 1.7999776569749042e-05, "loss": 0.098, "step": 9977 }, { "epoch": 1.83, "learning_rate": 1.7999331067569923e-05, "loss": 0.0987, "step": 9978 }, { "epoch": 1.83, "learning_rate": 1.7998885521298303e-05, "loss": 0.0965, "step": 9979 }, { "epoch": 1.83, "learning_rate": 1.799843993093664e-05, "loss": 0.0933, "step": 9980 }, { "epoch": 1.83, "learning_rate": 1.799799429648739e-05, "loss": 0.0909, "step": 9981 }, { "epoch": 1.83, "learning_rate": 1.799754861795301e-05, "loss": 0.0928, "step": 9982 }, { "epoch": 1.83, "learning_rate": 1.7997102895335946e-05, "loss": 0.0962, "step": 9983 }, { "epoch": 1.83, "learning_rate": 1.799665712863867e-05, "loss": 0.0993, "step": 9984 }, { "epoch": 1.83, "learning_rate": 1.799621131786363e-05, "loss": 0.0876, "step": 9985 }, { "epoch": 1.83, "learning_rate": 1.7995765463013287e-05, "loss": 0.0973, "step": 9986 }, { "epoch": 1.83, "learning_rate": 1.7995319564090098e-05, "loss": 0.0881, "step": 9987 }, { "epoch": 1.83, "learning_rate": 1.7994873621096523e-05, "loss": 0.0808, "step": 9988 }, { "epoch": 1.83, "learning_rate": 1.7994427634035016e-05, "loss": 0.0792, "step": 9989 }, { "epoch": 1.83, "learning_rate": 1.7993981602908032e-05, "loss": 0.1033, "step": 9990 }, { "epoch": 1.83, "learning_rate": 1.7993535527718035e-05, "loss": 0.0928, "step": 9991 }, { "epoch": 1.83, "learning_rate": 1.7993089408467485e-05, "loss": 0.0984, "step": 9992 }, { "epoch": 1.83, "learning_rate": 1.7992643245158842e-05, "loss": 0.0836, "step": 9993 }, { "epoch": 1.83, "learning_rate": 1.7992197037794558e-05, "loss": 0.0851, "step": 9994 }, { "epoch": 1.83, "learning_rate": 1.7991750786377097e-05, "loss": 0.0988, "step": 9995 }, { "epoch": 1.83, "learning_rate": 1.799130449090892e-05, "loss": 0.0877, "step": 9996 }, { "epoch": 1.83, "learning_rate": 1.7990858151392485e-05, "loss": 0.0918, "step": 9997 }, { "epoch": 1.83, "learning_rate": 1.799041176783025e-05, "loss": 0.0968, "step": 9998 }, { "epoch": 1.83, "learning_rate": 1.7989965340224685e-05, "loss": 0.0891, "step": 9999 }, { "epoch": 1.83, "learning_rate": 1.7989518868578234e-05, "loss": 0.0967, "step": 10000 }, { "epoch": 1.83, "learning_rate": 1.7989072352893374e-05, "loss": 0.0959, "step": 10001 }, { "epoch": 1.83, "learning_rate": 1.798862579317256e-05, "loss": 0.0856, "step": 10002 }, { "epoch": 1.83, "learning_rate": 1.7988179189418252e-05, "loss": 0.0891, "step": 10003 }, { "epoch": 1.83, "learning_rate": 1.7987732541632916e-05, "loss": 0.0885, "step": 10004 }, { "epoch": 1.83, "learning_rate": 1.798728584981901e-05, "loss": 0.078, "step": 10005 }, { "epoch": 1.83, "learning_rate": 1.7986839113978995e-05, "loss": 0.0823, "step": 10006 }, { "epoch": 1.84, "learning_rate": 1.7986392334115333e-05, "loss": 0.1004, "step": 10007 }, { "epoch": 1.84, "learning_rate": 1.7985945510230496e-05, "loss": 0.096, "step": 10008 }, { "epoch": 1.84, "learning_rate": 1.7985498642326938e-05, "loss": 0.0893, "step": 10009 }, { "epoch": 1.84, "learning_rate": 1.7985051730407123e-05, "loss": 0.0974, "step": 10010 }, { "epoch": 1.84, "learning_rate": 1.7984604774473518e-05, "loss": 0.0943, "step": 10011 }, { "epoch": 1.84, "learning_rate": 1.7984157774528582e-05, "loss": 0.0907, "step": 10012 }, { "epoch": 1.84, "learning_rate": 1.7983710730574783e-05, "loss": 0.0987, "step": 10013 }, { "epoch": 1.84, "learning_rate": 1.7983263642614583e-05, "loss": 0.0905, "step": 10014 }, { "epoch": 1.84, "learning_rate": 1.7982816510650445e-05, "loss": 0.0972, "step": 10015 }, { "epoch": 1.84, "learning_rate": 1.7982369334684836e-05, "loss": 0.1103, "step": 10016 }, { "epoch": 1.84, "learning_rate": 1.7981922114720223e-05, "loss": 0.0992, "step": 10017 }, { "epoch": 1.84, "learning_rate": 1.7981474850759065e-05, "loss": 0.0932, "step": 10018 }, { "epoch": 1.84, "learning_rate": 1.7981027542803832e-05, "loss": 0.1066, "step": 10019 }, { "epoch": 1.84, "learning_rate": 1.798058019085699e-05, "loss": 0.0953, "step": 10020 }, { "epoch": 1.84, "learning_rate": 1.7980132794921002e-05, "loss": 0.0905, "step": 10021 }, { "epoch": 1.84, "learning_rate": 1.7979685354998334e-05, "loss": 0.0796, "step": 10022 }, { "epoch": 1.84, "learning_rate": 1.7979237871091455e-05, "loss": 0.0955, "step": 10023 }, { "epoch": 1.84, "learning_rate": 1.7978790343202826e-05, "loss": 0.0926, "step": 10024 }, { "epoch": 1.84, "learning_rate": 1.7978342771334922e-05, "loss": 0.0807, "step": 10025 }, { "epoch": 1.84, "learning_rate": 1.7977895155490203e-05, "loss": 0.0903, "step": 10026 }, { "epoch": 1.84, "learning_rate": 1.797744749567114e-05, "loss": 0.0891, "step": 10027 }, { "epoch": 1.84, "learning_rate": 1.7976999791880197e-05, "loss": 0.0815, "step": 10028 }, { "epoch": 1.84, "learning_rate": 1.797655204411985e-05, "loss": 0.0926, "step": 10029 }, { "epoch": 1.84, "learning_rate": 1.797610425239256e-05, "loss": 0.0931, "step": 10030 }, { "epoch": 1.84, "learning_rate": 1.7975656416700794e-05, "loss": 0.093, "step": 10031 }, { "epoch": 1.84, "learning_rate": 1.7975208537047022e-05, "loss": 0.0975, "step": 10032 }, { "epoch": 1.84, "learning_rate": 1.797476061343372e-05, "loss": 0.0894, "step": 10033 }, { "epoch": 1.84, "learning_rate": 1.7974312645863342e-05, "loss": 0.0936, "step": 10034 }, { "epoch": 1.84, "learning_rate": 1.7973864634338372e-05, "loss": 0.0788, "step": 10035 }, { "epoch": 1.84, "learning_rate": 1.797341657886127e-05, "loss": 0.0953, "step": 10036 }, { "epoch": 1.84, "learning_rate": 1.797296847943451e-05, "loss": 0.0994, "step": 10037 }, { "epoch": 1.84, "learning_rate": 1.797252033606056e-05, "loss": 0.0839, "step": 10038 }, { "epoch": 1.84, "learning_rate": 1.7972072148741896e-05, "loss": 0.0852, "step": 10039 }, { "epoch": 1.84, "learning_rate": 1.7971623917480978e-05, "loss": 0.0907, "step": 10040 }, { "epoch": 1.84, "learning_rate": 1.7971175642280286e-05, "loss": 0.095, "step": 10041 }, { "epoch": 1.84, "learning_rate": 1.7970727323142284e-05, "loss": 0.1024, "step": 10042 }, { "epoch": 1.84, "learning_rate": 1.7970278960069446e-05, "loss": 0.1031, "step": 10043 }, { "epoch": 1.84, "learning_rate": 1.7969830553064245e-05, "loss": 0.0788, "step": 10044 }, { "epoch": 1.84, "learning_rate": 1.7969382102129153e-05, "loss": 0.0892, "step": 10045 }, { "epoch": 1.84, "learning_rate": 1.7968933607266638e-05, "loss": 0.0744, "step": 10046 }, { "epoch": 1.84, "learning_rate": 1.7968485068479173e-05, "loss": 0.086, "step": 10047 }, { "epoch": 1.84, "learning_rate": 1.7968036485769235e-05, "loss": 0.0899, "step": 10048 }, { "epoch": 1.84, "learning_rate": 1.7967587859139293e-05, "loss": 0.0838, "step": 10049 }, { "epoch": 1.84, "learning_rate": 1.7967139188591817e-05, "loss": 0.0863, "step": 10050 }, { "epoch": 1.84, "learning_rate": 1.7966690474129285e-05, "loss": 0.0871, "step": 10051 }, { "epoch": 1.84, "learning_rate": 1.7966241715754167e-05, "loss": 0.104, "step": 10052 }, { "epoch": 1.84, "learning_rate": 1.796579291346894e-05, "loss": 0.0888, "step": 10053 }, { "epoch": 1.84, "learning_rate": 1.7965344067276076e-05, "loss": 0.1011, "step": 10054 }, { "epoch": 1.84, "learning_rate": 1.7964895177178047e-05, "loss": 0.0934, "step": 10055 }, { "epoch": 1.84, "learning_rate": 1.796444624317733e-05, "loss": 0.0797, "step": 10056 }, { "epoch": 1.84, "learning_rate": 1.7963997265276402e-05, "loss": 0.0896, "step": 10057 }, { "epoch": 1.84, "learning_rate": 1.796354824347773e-05, "loss": 0.0926, "step": 10058 }, { "epoch": 1.84, "learning_rate": 1.7963099177783798e-05, "loss": 0.0916, "step": 10059 }, { "epoch": 1.84, "learning_rate": 1.7962650068197072e-05, "loss": 0.0893, "step": 10060 }, { "epoch": 1.85, "learning_rate": 1.7962200914720033e-05, "loss": 0.0893, "step": 10061 }, { "epoch": 1.85, "learning_rate": 1.7961751717355155e-05, "loss": 0.1032, "step": 10062 }, { "epoch": 1.85, "learning_rate": 1.796130247610492e-05, "loss": 0.0964, "step": 10063 }, { "epoch": 1.85, "learning_rate": 1.7960853190971795e-05, "loss": 0.0812, "step": 10064 }, { "epoch": 1.85, "learning_rate": 1.796040386195826e-05, "loss": 0.0892, "step": 10065 }, { "epoch": 1.85, "learning_rate": 1.7959954489066797e-05, "loss": 0.0852, "step": 10066 }, { "epoch": 1.85, "learning_rate": 1.7959505072299873e-05, "loss": 0.0877, "step": 10067 }, { "epoch": 1.85, "learning_rate": 1.7959055611659976e-05, "loss": 0.0913, "step": 10068 }, { "epoch": 1.85, "learning_rate": 1.7958606107149577e-05, "loss": 0.0867, "step": 10069 }, { "epoch": 1.85, "learning_rate": 1.7958156558771152e-05, "loss": 0.0834, "step": 10070 }, { "epoch": 1.85, "learning_rate": 1.795770696652718e-05, "loss": 0.0989, "step": 10071 }, { "epoch": 1.85, "learning_rate": 1.7957257330420147e-05, "loss": 0.0772, "step": 10072 }, { "epoch": 1.85, "learning_rate": 1.795680765045252e-05, "loss": 0.0813, "step": 10073 }, { "epoch": 1.85, "learning_rate": 1.7956357926626786e-05, "loss": 0.0827, "step": 10074 }, { "epoch": 1.85, "learning_rate": 1.7955908158945418e-05, "loss": 0.0899, "step": 10075 }, { "epoch": 1.85, "learning_rate": 1.7955458347410898e-05, "loss": 0.0852, "step": 10076 }, { "epoch": 1.85, "learning_rate": 1.7955008492025707e-05, "loss": 0.087, "step": 10077 }, { "epoch": 1.85, "learning_rate": 1.7954558592792323e-05, "loss": 0.0923, "step": 10078 }, { "epoch": 1.85, "learning_rate": 1.7954108649713224e-05, "loss": 0.0837, "step": 10079 }, { "epoch": 1.85, "learning_rate": 1.7953658662790893e-05, "loss": 0.0991, "step": 10080 }, { "epoch": 1.85, "learning_rate": 1.7953208632027808e-05, "loss": 0.0821, "step": 10081 }, { "epoch": 1.85, "learning_rate": 1.795275855742645e-05, "loss": 0.0916, "step": 10082 }, { "epoch": 1.85, "learning_rate": 1.7952308438989302e-05, "loss": 0.0851, "step": 10083 }, { "epoch": 1.85, "learning_rate": 1.7951858276718845e-05, "loss": 0.0867, "step": 10084 }, { "epoch": 1.85, "learning_rate": 1.7951408070617558e-05, "loss": 0.0871, "step": 10085 }, { "epoch": 1.85, "learning_rate": 1.7950957820687923e-05, "loss": 0.093, "step": 10086 }, { "epoch": 1.85, "learning_rate": 1.795050752693242e-05, "loss": 0.0897, "step": 10087 }, { "epoch": 1.85, "learning_rate": 1.795005718935354e-05, "loss": 0.0885, "step": 10088 }, { "epoch": 1.85, "learning_rate": 1.794960680795375e-05, "loss": 0.094, "step": 10089 }, { "epoch": 1.85, "learning_rate": 1.7949156382735548e-05, "loss": 0.0937, "step": 10090 }, { "epoch": 1.85, "learning_rate": 1.7948705913701406e-05, "loss": 0.0941, "step": 10091 }, { "epoch": 1.85, "learning_rate": 1.7948255400853808e-05, "loss": 0.0798, "step": 10092 }, { "epoch": 1.85, "learning_rate": 1.7947804844195243e-05, "loss": 0.0907, "step": 10093 }, { "epoch": 1.85, "learning_rate": 1.7947354243728196e-05, "loss": 0.0886, "step": 10094 }, { "epoch": 1.85, "learning_rate": 1.794690359945514e-05, "loss": 0.1056, "step": 10095 }, { "epoch": 1.85, "learning_rate": 1.7946452911378568e-05, "loss": 0.0931, "step": 10096 }, { "epoch": 1.85, "learning_rate": 1.794600217950096e-05, "loss": 0.1007, "step": 10097 }, { "epoch": 1.85, "learning_rate": 1.7945551403824803e-05, "loss": 0.0831, "step": 10098 }, { "epoch": 1.85, "learning_rate": 1.794510058435258e-05, "loss": 0.0875, "step": 10099 }, { "epoch": 1.85, "learning_rate": 1.7944649721086776e-05, "loss": 0.0749, "step": 10100 }, { "epoch": 1.85, "learning_rate": 1.7944198814029875e-05, "loss": 0.0957, "step": 10101 }, { "epoch": 1.85, "learning_rate": 1.7943747863184366e-05, "loss": 0.0951, "step": 10102 }, { "epoch": 1.85, "learning_rate": 1.7943296868552735e-05, "loss": 0.0983, "step": 10103 }, { "epoch": 1.85, "learning_rate": 1.7942845830137462e-05, "loss": 0.0869, "step": 10104 }, { "epoch": 1.85, "learning_rate": 1.794239474794104e-05, "loss": 0.0951, "step": 10105 }, { "epoch": 1.85, "learning_rate": 1.794194362196595e-05, "loss": 0.0982, "step": 10106 }, { "epoch": 1.85, "learning_rate": 1.794149245221468e-05, "loss": 0.102, "step": 10107 }, { "epoch": 1.85, "learning_rate": 1.794104123868972e-05, "loss": 0.0952, "step": 10108 }, { "epoch": 1.85, "learning_rate": 1.794058998139355e-05, "loss": 0.094, "step": 10109 }, { "epoch": 1.85, "learning_rate": 1.7940138680328666e-05, "loss": 0.0881, "step": 10110 }, { "epoch": 1.85, "learning_rate": 1.793968733549755e-05, "loss": 0.0923, "step": 10111 }, { "epoch": 1.85, "learning_rate": 1.7939235946902694e-05, "loss": 0.0882, "step": 10112 }, { "epoch": 1.85, "learning_rate": 1.793878451454658e-05, "loss": 0.0858, "step": 10113 }, { "epoch": 1.85, "learning_rate": 1.7938333038431702e-05, "loss": 0.0924, "step": 10114 }, { "epoch": 1.85, "learning_rate": 1.7937881518560548e-05, "loss": 0.0886, "step": 10115 }, { "epoch": 1.86, "learning_rate": 1.79374299549356e-05, "loss": 0.0858, "step": 10116 }, { "epoch": 1.86, "learning_rate": 1.7936978347559354e-05, "loss": 0.0904, "step": 10117 }, { "epoch": 1.86, "learning_rate": 1.79365266964343e-05, "loss": 0.087, "step": 10118 }, { "epoch": 1.86, "learning_rate": 1.7936075001562922e-05, "loss": 0.0819, "step": 10119 }, { "epoch": 1.86, "learning_rate": 1.7935623262947715e-05, "loss": 0.0903, "step": 10120 }, { "epoch": 1.86, "learning_rate": 1.7935171480591164e-05, "loss": 0.0954, "step": 10121 }, { "epoch": 1.86, "learning_rate": 1.7934719654495763e-05, "loss": 0.0956, "step": 10122 }, { "epoch": 1.86, "learning_rate": 1.7934267784664002e-05, "loss": 0.0996, "step": 10123 }, { "epoch": 1.86, "learning_rate": 1.793381587109837e-05, "loss": 0.0821, "step": 10124 }, { "epoch": 1.86, "learning_rate": 1.7933363913801362e-05, "loss": 0.0861, "step": 10125 }, { "epoch": 1.86, "learning_rate": 1.7932911912775463e-05, "loss": 0.0903, "step": 10126 }, { "epoch": 1.86, "learning_rate": 1.7932459868023172e-05, "loss": 0.0892, "step": 10127 }, { "epoch": 1.86, "learning_rate": 1.7932007779546972e-05, "loss": 0.0888, "step": 10128 }, { "epoch": 1.86, "learning_rate": 1.7931555647349358e-05, "loss": 0.087, "step": 10129 }, { "epoch": 1.86, "learning_rate": 1.7931103471432826e-05, "loss": 0.0855, "step": 10130 }, { "epoch": 1.86, "learning_rate": 1.7930651251799866e-05, "loss": 0.0949, "step": 10131 }, { "epoch": 1.86, "learning_rate": 1.7930198988452973e-05, "loss": 0.0907, "step": 10132 }, { "epoch": 1.86, "learning_rate": 1.7929746681394633e-05, "loss": 0.0834, "step": 10133 }, { "epoch": 1.86, "learning_rate": 1.7929294330627346e-05, "loss": 0.0888, "step": 10134 }, { "epoch": 1.86, "learning_rate": 1.7928841936153602e-05, "loss": 0.1036, "step": 10135 }, { "epoch": 1.86, "learning_rate": 1.7928389497975897e-05, "loss": 0.0963, "step": 10136 }, { "epoch": 1.86, "learning_rate": 1.792793701609672e-05, "loss": 0.0903, "step": 10137 }, { "epoch": 1.86, "learning_rate": 1.792748449051857e-05, "loss": 0.0879, "step": 10138 }, { "epoch": 1.86, "learning_rate": 1.792703192124394e-05, "loss": 0.1055, "step": 10139 }, { "epoch": 1.86, "learning_rate": 1.7926579308275325e-05, "loss": 0.0882, "step": 10140 }, { "epoch": 1.86, "learning_rate": 1.7926126651615218e-05, "loss": 0.0892, "step": 10141 }, { "epoch": 1.86, "learning_rate": 1.7925673951266118e-05, "loss": 0.1048, "step": 10142 }, { "epoch": 1.86, "learning_rate": 1.7925221207230514e-05, "loss": 0.0991, "step": 10143 }, { "epoch": 1.86, "learning_rate": 1.7924768419510906e-05, "loss": 0.0809, "step": 10144 }, { "epoch": 1.86, "learning_rate": 1.792431558810979e-05, "loss": 0.1016, "step": 10145 }, { "epoch": 1.86, "learning_rate": 1.792386271302966e-05, "loss": 0.0907, "step": 10146 }, { "epoch": 1.86, "learning_rate": 1.7923409794273013e-05, "loss": 0.0903, "step": 10147 }, { "epoch": 1.86, "learning_rate": 1.7922956831842344e-05, "loss": 0.091, "step": 10148 }, { "epoch": 1.86, "learning_rate": 1.7922503825740153e-05, "loss": 0.0904, "step": 10149 }, { "epoch": 1.86, "learning_rate": 1.7922050775968935e-05, "loss": 0.1019, "step": 10150 }, { "epoch": 1.86, "learning_rate": 1.792159768253119e-05, "loss": 0.0843, "step": 10151 }, { "epoch": 1.86, "learning_rate": 1.792114454542941e-05, "loss": 0.0879, "step": 10152 }, { "epoch": 1.86, "learning_rate": 1.7920691364666096e-05, "loss": 0.0954, "step": 10153 }, { "epoch": 1.86, "learning_rate": 1.7920238140243745e-05, "loss": 0.082, "step": 10154 }, { "epoch": 1.86, "learning_rate": 1.791978487216486e-05, "loss": 0.1002, "step": 10155 }, { "epoch": 1.86, "learning_rate": 1.791933156043193e-05, "loss": 0.0974, "step": 10156 }, { "epoch": 1.86, "learning_rate": 1.791887820504746e-05, "loss": 0.095, "step": 10157 }, { "epoch": 1.86, "learning_rate": 1.791842480601395e-05, "loss": 0.0763, "step": 10158 }, { "epoch": 1.86, "learning_rate": 1.7917971363333896e-05, "loss": 0.0949, "step": 10159 }, { "epoch": 1.86, "learning_rate": 1.7917517877009798e-05, "loss": 0.0964, "step": 10160 }, { "epoch": 1.86, "learning_rate": 1.7917064347044157e-05, "loss": 0.0942, "step": 10161 }, { "epoch": 1.86, "learning_rate": 1.7916610773439473e-05, "loss": 0.1075, "step": 10162 }, { "epoch": 1.86, "learning_rate": 1.791615715619824e-05, "loss": 0.1066, "step": 10163 }, { "epoch": 1.86, "learning_rate": 1.7915703495322967e-05, "loss": 0.0977, "step": 10164 }, { "epoch": 1.86, "learning_rate": 1.791524979081615e-05, "loss": 0.0999, "step": 10165 }, { "epoch": 1.86, "learning_rate": 1.7914796042680292e-05, "loss": 0.0903, "step": 10166 }, { "epoch": 1.86, "learning_rate": 1.7914342250917892e-05, "loss": 0.0829, "step": 10167 }, { "epoch": 1.86, "learning_rate": 1.7913888415531452e-05, "loss": 0.0896, "step": 10168 }, { "epoch": 1.86, "learning_rate": 1.7913434536523473e-05, "loss": 0.1024, "step": 10169 }, { "epoch": 1.87, "learning_rate": 1.7912980613896457e-05, "loss": 0.0924, "step": 10170 }, { "epoch": 1.87, "learning_rate": 1.791252664765291e-05, "loss": 0.0822, "step": 10171 }, { "epoch": 1.87, "learning_rate": 1.7912072637795326e-05, "loss": 0.1001, "step": 10172 }, { "epoch": 1.87, "learning_rate": 1.7911618584326213e-05, "loss": 0.1119, "step": 10173 }, { "epoch": 1.87, "learning_rate": 1.7911164487248077e-05, "loss": 0.0997, "step": 10174 }, { "epoch": 1.87, "learning_rate": 1.7910710346563417e-05, "loss": 0.0764, "step": 10175 }, { "epoch": 1.87, "learning_rate": 1.791025616227473e-05, "loss": 0.0971, "step": 10176 }, { "epoch": 1.87, "learning_rate": 1.790980193438453e-05, "loss": 0.0967, "step": 10177 }, { "epoch": 1.87, "learning_rate": 1.7909347662895316e-05, "loss": 0.0861, "step": 10178 }, { "epoch": 1.87, "learning_rate": 1.7908893347809592e-05, "loss": 0.1, "step": 10179 }, { "epoch": 1.87, "learning_rate": 1.7908438989129863e-05, "loss": 0.0794, "step": 10180 }, { "epoch": 1.87, "learning_rate": 1.7907984586858634e-05, "loss": 0.0886, "step": 10181 }, { "epoch": 1.87, "learning_rate": 1.7907530140998405e-05, "loss": 0.1087, "step": 10182 }, { "epoch": 1.87, "learning_rate": 1.790707565155169e-05, "loss": 0.0938, "step": 10183 }, { "epoch": 1.87, "learning_rate": 1.7906621118520987e-05, "loss": 0.0904, "step": 10184 }, { "epoch": 1.87, "learning_rate": 1.7906166541908798e-05, "loss": 0.0818, "step": 10185 }, { "epoch": 1.87, "learning_rate": 1.790571192171764e-05, "loss": 0.0835, "step": 10186 }, { "epoch": 1.87, "learning_rate": 1.7905257257950008e-05, "loss": 0.0882, "step": 10187 }, { "epoch": 1.87, "learning_rate": 1.7904802550608415e-05, "loss": 0.1, "step": 10188 }, { "epoch": 1.87, "learning_rate": 1.790434779969537e-05, "loss": 0.0774, "step": 10189 }, { "epoch": 1.87, "learning_rate": 1.790389300521337e-05, "loss": 0.0798, "step": 10190 }, { "epoch": 1.87, "learning_rate": 1.7903438167164923e-05, "loss": 0.0969, "step": 10191 }, { "epoch": 1.87, "learning_rate": 1.7902983285552548e-05, "loss": 0.086, "step": 10192 }, { "epoch": 1.87, "learning_rate": 1.7902528360378736e-05, "loss": 0.0886, "step": 10193 }, { "epoch": 1.87, "learning_rate": 1.7902073391646007e-05, "loss": 0.1, "step": 10194 }, { "epoch": 1.87, "learning_rate": 1.7901618379356862e-05, "loss": 0.0793, "step": 10195 }, { "epoch": 1.87, "learning_rate": 1.7901163323513815e-05, "loss": 0.0865, "step": 10196 }, { "epoch": 1.87, "learning_rate": 1.7900708224119366e-05, "loss": 0.0827, "step": 10197 }, { "epoch": 1.87, "learning_rate": 1.790025308117603e-05, "loss": 0.095, "step": 10198 }, { "epoch": 1.87, "learning_rate": 1.7899797894686316e-05, "loss": 0.0857, "step": 10199 }, { "epoch": 1.87, "learning_rate": 1.789934266465273e-05, "loss": 0.0791, "step": 10200 }, { "epoch": 1.87, "learning_rate": 1.7898887391077782e-05, "loss": 0.0912, "step": 10201 }, { "epoch": 1.87, "learning_rate": 1.789843207396398e-05, "loss": 0.0932, "step": 10202 }, { "epoch": 1.87, "learning_rate": 1.7897976713313836e-05, "loss": 0.0811, "step": 10203 }, { "epoch": 1.87, "learning_rate": 1.7897521309129858e-05, "loss": 0.0928, "step": 10204 }, { "epoch": 1.87, "learning_rate": 1.7897065861414563e-05, "loss": 0.0881, "step": 10205 }, { "epoch": 1.87, "learning_rate": 1.7896610370170452e-05, "loss": 0.0923, "step": 10206 }, { "epoch": 1.87, "learning_rate": 1.7896154835400038e-05, "loss": 0.1078, "step": 10207 }, { "epoch": 1.87, "learning_rate": 1.7895699257105836e-05, "loss": 0.0839, "step": 10208 }, { "epoch": 1.87, "learning_rate": 1.7895243635290355e-05, "loss": 0.0933, "step": 10209 }, { "epoch": 1.87, "learning_rate": 1.7894787969956105e-05, "loss": 0.1069, "step": 10210 }, { "epoch": 1.87, "learning_rate": 1.78943322611056e-05, "loss": 0.0773, "step": 10211 }, { "epoch": 1.87, "learning_rate": 1.7893876508741347e-05, "loss": 0.0849, "step": 10212 }, { "epoch": 1.87, "learning_rate": 1.7893420712865864e-05, "loss": 0.0942, "step": 10213 }, { "epoch": 1.87, "learning_rate": 1.7892964873481667e-05, "loss": 0.0972, "step": 10214 }, { "epoch": 1.87, "learning_rate": 1.7892508990591256e-05, "loss": 0.0874, "step": 10215 }, { "epoch": 1.87, "learning_rate": 1.789205306419715e-05, "loss": 0.0895, "step": 10216 }, { "epoch": 1.87, "learning_rate": 1.7891597094301864e-05, "loss": 0.0931, "step": 10217 }, { "epoch": 1.87, "learning_rate": 1.7891141080907914e-05, "loss": 0.0923, "step": 10218 }, { "epoch": 1.87, "learning_rate": 1.7890685024017802e-05, "loss": 0.08, "step": 10219 }, { "epoch": 1.87, "learning_rate": 1.7890228923634057e-05, "loss": 0.0854, "step": 10220 }, { "epoch": 1.87, "learning_rate": 1.788977277975918e-05, "loss": 0.0929, "step": 10221 }, { "epoch": 1.87, "learning_rate": 1.7889316592395692e-05, "loss": 0.0954, "step": 10222 }, { "epoch": 1.87, "learning_rate": 1.7888860361546108e-05, "loss": 0.0824, "step": 10223 }, { "epoch": 1.87, "learning_rate": 1.788840408721294e-05, "loss": 0.0925, "step": 10224 }, { "epoch": 1.88, "learning_rate": 1.7887947769398702e-05, "loss": 0.0864, "step": 10225 }, { "epoch": 1.88, "learning_rate": 1.7887491408105912e-05, "loss": 0.0907, "step": 10226 }, { "epoch": 1.88, "learning_rate": 1.7887035003337082e-05, "loss": 0.0977, "step": 10227 }, { "epoch": 1.88, "learning_rate": 1.7886578555094733e-05, "loss": 0.0983, "step": 10228 }, { "epoch": 1.88, "learning_rate": 1.788612206338138e-05, "loss": 0.0888, "step": 10229 }, { "epoch": 1.88, "learning_rate": 1.7885665528199536e-05, "loss": 0.1041, "step": 10230 }, { "epoch": 1.88, "learning_rate": 1.7885208949551715e-05, "loss": 0.0905, "step": 10231 }, { "epoch": 1.88, "learning_rate": 1.7884752327440444e-05, "loss": 0.0932, "step": 10232 }, { "epoch": 1.88, "learning_rate": 1.7884295661868232e-05, "loss": 0.0958, "step": 10233 }, { "epoch": 1.88, "learning_rate": 1.7883838952837595e-05, "loss": 0.0869, "step": 10234 }, { "epoch": 1.88, "learning_rate": 1.7883382200351054e-05, "loss": 0.0833, "step": 10235 }, { "epoch": 1.88, "learning_rate": 1.7882925404411127e-05, "loss": 0.0892, "step": 10236 }, { "epoch": 1.88, "learning_rate": 1.7882468565020327e-05, "loss": 0.0908, "step": 10237 }, { "epoch": 1.88, "learning_rate": 1.788201168218118e-05, "loss": 0.0998, "step": 10238 }, { "epoch": 1.88, "learning_rate": 1.7881554755896196e-05, "loss": 0.0968, "step": 10239 }, { "epoch": 1.88, "learning_rate": 1.78810977861679e-05, "loss": 0.0907, "step": 10240 }, { "epoch": 1.88, "learning_rate": 1.7880640772998807e-05, "loss": 0.0963, "step": 10241 }, { "epoch": 1.88, "learning_rate": 1.7880183716391436e-05, "loss": 0.0938, "step": 10242 }, { "epoch": 1.88, "learning_rate": 1.787972661634831e-05, "loss": 0.0942, "step": 10243 }, { "epoch": 1.88, "learning_rate": 1.7879269472871947e-05, "loss": 0.0987, "step": 10244 }, { "epoch": 1.88, "learning_rate": 1.7878812285964867e-05, "loss": 0.0932, "step": 10245 }, { "epoch": 1.88, "learning_rate": 1.7878355055629585e-05, "loss": 0.0991, "step": 10246 }, { "epoch": 1.88, "learning_rate": 1.7877897781868628e-05, "loss": 0.089, "step": 10247 }, { "epoch": 1.88, "learning_rate": 1.7877440464684513e-05, "loss": 0.0998, "step": 10248 }, { "epoch": 1.88, "learning_rate": 1.7876983104079765e-05, "loss": 0.0879, "step": 10249 }, { "epoch": 1.88, "learning_rate": 1.78765257000569e-05, "loss": 0.089, "step": 10250 }, { "epoch": 1.88, "learning_rate": 1.7876068252618438e-05, "loss": 0.096, "step": 10251 }, { "epoch": 1.88, "learning_rate": 1.7875610761766907e-05, "loss": 0.0974, "step": 10252 }, { "epoch": 1.88, "learning_rate": 1.787515322750482e-05, "loss": 0.099, "step": 10253 }, { "epoch": 1.88, "learning_rate": 1.787469564983471e-05, "loss": 0.0878, "step": 10254 }, { "epoch": 1.88, "learning_rate": 1.7874238028759086e-05, "loss": 0.1045, "step": 10255 }, { "epoch": 1.88, "learning_rate": 1.787378036428048e-05, "loss": 0.0966, "step": 10256 }, { "epoch": 1.88, "learning_rate": 1.7873322656401417e-05, "loss": 0.0823, "step": 10257 }, { "epoch": 1.88, "learning_rate": 1.787286490512441e-05, "loss": 0.0848, "step": 10258 }, { "epoch": 1.88, "learning_rate": 1.787240711045199e-05, "loss": 0.092, "step": 10259 }, { "epoch": 1.88, "learning_rate": 1.7871949272386678e-05, "loss": 0.0921, "step": 10260 }, { "epoch": 1.88, "learning_rate": 1.7871491390930993e-05, "loss": 0.094, "step": 10261 }, { "epoch": 1.88, "learning_rate": 1.7871033466087464e-05, "loss": 0.0942, "step": 10262 }, { "epoch": 1.88, "learning_rate": 1.7870575497858617e-05, "loss": 0.1103, "step": 10263 }, { "epoch": 1.88, "learning_rate": 1.787011748624697e-05, "loss": 0.0899, "step": 10264 }, { "epoch": 1.88, "learning_rate": 1.786965943125506e-05, "loss": 0.1065, "step": 10265 }, { "epoch": 1.88, "learning_rate": 1.7869201332885393e-05, "loss": 0.0958, "step": 10266 }, { "epoch": 1.88, "learning_rate": 1.7868743191140504e-05, "loss": 0.0891, "step": 10267 }, { "epoch": 1.88, "learning_rate": 1.786828500602292e-05, "loss": 0.0842, "step": 10268 }, { "epoch": 1.88, "learning_rate": 1.7867826777535168e-05, "loss": 0.0995, "step": 10269 }, { "epoch": 1.88, "learning_rate": 1.786736850567977e-05, "loss": 0.0837, "step": 10270 }, { "epoch": 1.88, "learning_rate": 1.786691019045925e-05, "loss": 0.0959, "step": 10271 }, { "epoch": 1.88, "learning_rate": 1.7866451831876138e-05, "loss": 0.0942, "step": 10272 }, { "epoch": 1.88, "learning_rate": 1.786599342993296e-05, "loss": 0.089, "step": 10273 }, { "epoch": 1.88, "learning_rate": 1.7865534984632237e-05, "loss": 0.096, "step": 10274 }, { "epoch": 1.88, "learning_rate": 1.7865076495976507e-05, "loss": 0.0822, "step": 10275 }, { "epoch": 1.88, "learning_rate": 1.786461796396829e-05, "loss": 0.0856, "step": 10276 }, { "epoch": 1.88, "learning_rate": 1.7864159388610112e-05, "loss": 0.0953, "step": 10277 }, { "epoch": 1.88, "learning_rate": 1.7863700769904505e-05, "loss": 0.0869, "step": 10278 }, { "epoch": 1.89, "learning_rate": 1.7863242107853996e-05, "loss": 0.0973, "step": 10279 }, { "epoch": 1.89, "learning_rate": 1.786278340246111e-05, "loss": 0.0961, "step": 10280 }, { "epoch": 1.89, "learning_rate": 1.786232465372838e-05, "loss": 0.0867, "step": 10281 }, { "epoch": 1.89, "learning_rate": 1.7861865861658326e-05, "loss": 0.0976, "step": 10282 }, { "epoch": 1.89, "learning_rate": 1.786140702625349e-05, "loss": 0.0918, "step": 10283 }, { "epoch": 1.89, "learning_rate": 1.7860948147516393e-05, "loss": 0.0835, "step": 10284 }, { "epoch": 1.89, "learning_rate": 1.7860489225449565e-05, "loss": 0.0895, "step": 10285 }, { "epoch": 1.89, "learning_rate": 1.7860030260055538e-05, "loss": 0.0875, "step": 10286 }, { "epoch": 1.89, "learning_rate": 1.785957125133684e-05, "loss": 0.0841, "step": 10287 }, { "epoch": 1.89, "learning_rate": 1.7859112199295996e-05, "loss": 0.0934, "step": 10288 }, { "epoch": 1.89, "learning_rate": 1.7858653103935544e-05, "loss": 0.0805, "step": 10289 }, { "epoch": 1.89, "learning_rate": 1.7858193965258013e-05, "loss": 0.0924, "step": 10290 }, { "epoch": 1.89, "learning_rate": 1.7857734783265935e-05, "loss": 0.0891, "step": 10291 }, { "epoch": 1.89, "learning_rate": 1.7857275557961837e-05, "loss": 0.091, "step": 10292 }, { "epoch": 1.89, "learning_rate": 1.785681628934825e-05, "loss": 0.092, "step": 10293 }, { "epoch": 1.89, "learning_rate": 1.785635697742771e-05, "loss": 0.0923, "step": 10294 }, { "epoch": 1.89, "learning_rate": 1.7855897622202744e-05, "loss": 0.0812, "step": 10295 }, { "epoch": 1.89, "learning_rate": 1.7855438223675888e-05, "loss": 0.0984, "step": 10296 }, { "epoch": 1.89, "learning_rate": 1.785497878184967e-05, "loss": 0.0878, "step": 10297 }, { "epoch": 1.89, "learning_rate": 1.7854519296726628e-05, "loss": 0.0832, "step": 10298 }, { "epoch": 1.89, "learning_rate": 1.7854059768309292e-05, "loss": 0.0868, "step": 10299 }, { "epoch": 1.89, "learning_rate": 1.785360019660019e-05, "loss": 0.0984, "step": 10300 }, { "epoch": 1.89, "learning_rate": 1.7853140581601864e-05, "loss": 0.0973, "step": 10301 }, { "epoch": 1.89, "learning_rate": 1.785268092331684e-05, "loss": 0.0959, "step": 10302 }, { "epoch": 1.89, "learning_rate": 1.7852221221747658e-05, "loss": 0.0989, "step": 10303 }, { "epoch": 1.89, "learning_rate": 1.785176147689685e-05, "loss": 0.0937, "step": 10304 }, { "epoch": 1.89, "learning_rate": 1.7851301688766944e-05, "loss": 0.0955, "step": 10305 }, { "epoch": 1.89, "learning_rate": 1.7850841857360482e-05, "loss": 0.0965, "step": 10306 }, { "epoch": 1.89, "learning_rate": 1.7850381982679998e-05, "loss": 0.0786, "step": 10307 }, { "epoch": 1.89, "learning_rate": 1.7849922064728022e-05, "loss": 0.0986, "step": 10308 }, { "epoch": 1.89, "learning_rate": 1.784946210350709e-05, "loss": 0.0901, "step": 10309 }, { "epoch": 1.89, "learning_rate": 1.7849002099019737e-05, "loss": 0.087, "step": 10310 }, { "epoch": 1.89, "learning_rate": 1.7848542051268504e-05, "loss": 0.0919, "step": 10311 }, { "epoch": 1.89, "learning_rate": 1.7848081960255927e-05, "loss": 0.0903, "step": 10312 }, { "epoch": 1.89, "learning_rate": 1.7847621825984533e-05, "loss": 0.0833, "step": 10313 }, { "epoch": 1.89, "learning_rate": 1.7847161648456865e-05, "loss": 0.0849, "step": 10314 }, { "epoch": 1.89, "learning_rate": 1.7846701427675455e-05, "loss": 0.097, "step": 10315 }, { "epoch": 1.89, "learning_rate": 1.784624116364285e-05, "loss": 0.078, "step": 10316 }, { "epoch": 1.89, "learning_rate": 1.7845780856361575e-05, "loss": 0.104, "step": 10317 }, { "epoch": 1.89, "learning_rate": 1.7845320505834176e-05, "loss": 0.0776, "step": 10318 }, { "epoch": 1.89, "learning_rate": 1.7844860112063184e-05, "loss": 0.103, "step": 10319 }, { "epoch": 1.89, "learning_rate": 1.784439967505114e-05, "loss": 0.0905, "step": 10320 }, { "epoch": 1.89, "learning_rate": 1.7843939194800577e-05, "loss": 0.0994, "step": 10321 }, { "epoch": 1.89, "learning_rate": 1.7843478671314044e-05, "loss": 0.0934, "step": 10322 }, { "epoch": 1.89, "learning_rate": 1.784301810459407e-05, "loss": 0.094, "step": 10323 }, { "epoch": 1.89, "learning_rate": 1.7842557494643192e-05, "loss": 0.0927, "step": 10324 }, { "epoch": 1.89, "learning_rate": 1.784209684146396e-05, "loss": 0.1033, "step": 10325 }, { "epoch": 1.89, "learning_rate": 1.7841636145058902e-05, "loss": 0.0824, "step": 10326 }, { "epoch": 1.89, "learning_rate": 1.7841175405430565e-05, "loss": 0.0958, "step": 10327 }, { "epoch": 1.89, "learning_rate": 1.7840714622581486e-05, "loss": 0.0856, "step": 10328 }, { "epoch": 1.89, "learning_rate": 1.78402537965142e-05, "loss": 0.1018, "step": 10329 }, { "epoch": 1.89, "learning_rate": 1.7839792927231253e-05, "loss": 0.0907, "step": 10330 }, { "epoch": 1.89, "learning_rate": 1.7839332014735186e-05, "loss": 0.085, "step": 10331 }, { "epoch": 1.89, "learning_rate": 1.7838871059028537e-05, "loss": 0.1002, "step": 10332 }, { "epoch": 1.89, "learning_rate": 1.7838410060113846e-05, "loss": 0.101, "step": 10333 }, { "epoch": 1.9, "learning_rate": 1.7837949017993655e-05, "loss": 0.0843, "step": 10334 }, { "epoch": 1.9, "learning_rate": 1.7837487932670507e-05, "loss": 0.0965, "step": 10335 }, { "epoch": 1.9, "learning_rate": 1.783702680414694e-05, "loss": 0.0825, "step": 10336 }, { "epoch": 1.9, "learning_rate": 1.7836565632425497e-05, "loss": 0.0958, "step": 10337 }, { "epoch": 1.9, "learning_rate": 1.783610441750872e-05, "loss": 0.0919, "step": 10338 }, { "epoch": 1.9, "learning_rate": 1.7835643159399156e-05, "loss": 0.0854, "step": 10339 }, { "epoch": 1.9, "learning_rate": 1.783518185809934e-05, "loss": 0.0944, "step": 10340 }, { "epoch": 1.9, "learning_rate": 1.783472051361182e-05, "loss": 0.0894, "step": 10341 }, { "epoch": 1.9, "learning_rate": 1.7834259125939136e-05, "loss": 0.0886, "step": 10342 }, { "epoch": 1.9, "learning_rate": 1.7833797695083833e-05, "loss": 0.0988, "step": 10343 }, { "epoch": 1.9, "learning_rate": 1.7833336221048453e-05, "loss": 0.0912, "step": 10344 }, { "epoch": 1.9, "learning_rate": 1.7832874703835538e-05, "loss": 0.087, "step": 10345 }, { "epoch": 1.9, "learning_rate": 1.7832413143447636e-05, "loss": 0.086, "step": 10346 }, { "epoch": 1.9, "learning_rate": 1.7831951539887292e-05, "loss": 0.0843, "step": 10347 }, { "epoch": 1.9, "learning_rate": 1.7831489893157042e-05, "loss": 0.0842, "step": 10348 }, { "epoch": 1.9, "learning_rate": 1.7831028203259443e-05, "loss": 0.0887, "step": 10349 }, { "epoch": 1.9, "learning_rate": 1.7830566470197027e-05, "loss": 0.0818, "step": 10350 }, { "epoch": 1.9, "learning_rate": 1.7830104693972347e-05, "loss": 0.0977, "step": 10351 }, { "epoch": 1.9, "learning_rate": 1.7829642874587947e-05, "loss": 0.0872, "step": 10352 }, { "epoch": 1.9, "learning_rate": 1.7829181012046374e-05, "loss": 0.0881, "step": 10353 }, { "epoch": 1.9, "learning_rate": 1.7828719106350167e-05, "loss": 0.0938, "step": 10354 }, { "epoch": 1.9, "learning_rate": 1.782825715750188e-05, "loss": 0.1049, "step": 10355 }, { "epoch": 1.9, "learning_rate": 1.7827795165504057e-05, "loss": 0.0952, "step": 10356 }, { "epoch": 1.9, "learning_rate": 1.7827333130359243e-05, "loss": 0.0948, "step": 10357 }, { "epoch": 1.9, "learning_rate": 1.7826871052069986e-05, "loss": 0.0925, "step": 10358 }, { "epoch": 1.9, "learning_rate": 1.7826408930638827e-05, "loss": 0.1004, "step": 10359 }, { "epoch": 1.9, "learning_rate": 1.7825946766068325e-05, "loss": 0.0941, "step": 10360 }, { "epoch": 1.9, "learning_rate": 1.782548455836102e-05, "loss": 0.0858, "step": 10361 }, { "epoch": 1.9, "learning_rate": 1.782502230751946e-05, "loss": 0.0964, "step": 10362 }, { "epoch": 1.9, "learning_rate": 1.7824560013546194e-05, "loss": 0.0982, "step": 10363 }, { "epoch": 1.9, "learning_rate": 1.7824097676443767e-05, "loss": 0.0885, "step": 10364 }, { "epoch": 1.9, "learning_rate": 1.7823635296214736e-05, "loss": 0.0878, "step": 10365 }, { "epoch": 1.9, "learning_rate": 1.7823172872861637e-05, "loss": 0.094, "step": 10366 }, { "epoch": 1.9, "learning_rate": 1.782271040638703e-05, "loss": 0.0849, "step": 10367 }, { "epoch": 1.9, "learning_rate": 1.7822247896793462e-05, "loss": 0.0864, "step": 10368 }, { "epoch": 1.9, "learning_rate": 1.7821785344083474e-05, "loss": 0.0979, "step": 10369 }, { "epoch": 1.9, "learning_rate": 1.7821322748259627e-05, "loss": 0.0991, "step": 10370 }, { "epoch": 1.9, "learning_rate": 1.7820860109324467e-05, "loss": 0.0936, "step": 10371 }, { "epoch": 1.9, "learning_rate": 1.782039742728054e-05, "loss": 0.0849, "step": 10372 }, { "epoch": 1.9, "learning_rate": 1.78199347021304e-05, "loss": 0.0924, "step": 10373 }, { "epoch": 1.9, "learning_rate": 1.7819471933876594e-05, "loss": 0.0904, "step": 10374 }, { "epoch": 1.9, "learning_rate": 1.781900912252168e-05, "loss": 0.0916, "step": 10375 }, { "epoch": 1.9, "learning_rate": 1.7818546268068203e-05, "loss": 0.0908, "step": 10376 }, { "epoch": 1.9, "learning_rate": 1.781808337051871e-05, "loss": 0.094, "step": 10377 }, { "epoch": 1.9, "learning_rate": 1.781762042987577e-05, "loss": 0.0857, "step": 10378 }, { "epoch": 1.9, "learning_rate": 1.7817157446141915e-05, "loss": 0.0849, "step": 10379 }, { "epoch": 1.9, "learning_rate": 1.7816694419319705e-05, "loss": 0.0712, "step": 10380 }, { "epoch": 1.9, "learning_rate": 1.7816231349411694e-05, "loss": 0.0873, "step": 10381 }, { "epoch": 1.9, "learning_rate": 1.7815768236420433e-05, "loss": 0.0928, "step": 10382 }, { "epoch": 1.9, "learning_rate": 1.7815305080348474e-05, "loss": 0.1039, "step": 10383 }, { "epoch": 1.9, "learning_rate": 1.781484188119837e-05, "loss": 0.0988, "step": 10384 }, { "epoch": 1.9, "learning_rate": 1.7814378638972677e-05, "loss": 0.0944, "step": 10385 }, { "epoch": 1.9, "learning_rate": 1.781391535367394e-05, "loss": 0.0912, "step": 10386 }, { "epoch": 1.9, "learning_rate": 1.7813452025304723e-05, "loss": 0.0889, "step": 10387 }, { "epoch": 1.91, "learning_rate": 1.7812988653867576e-05, "loss": 0.0866, "step": 10388 }, { "epoch": 1.91, "learning_rate": 1.7812525239365047e-05, "loss": 0.0897, "step": 10389 }, { "epoch": 1.91, "learning_rate": 1.7812061781799703e-05, "loss": 0.099, "step": 10390 }, { "epoch": 1.91, "learning_rate": 1.7811598281174086e-05, "loss": 0.0865, "step": 10391 }, { "epoch": 1.91, "learning_rate": 1.781113473749076e-05, "loss": 0.0845, "step": 10392 }, { "epoch": 1.91, "learning_rate": 1.7810671150752274e-05, "loss": 0.1011, "step": 10393 }, { "epoch": 1.91, "learning_rate": 1.7810207520961187e-05, "loss": 0.088, "step": 10394 }, { "epoch": 1.91, "learning_rate": 1.780974384812005e-05, "loss": 0.095, "step": 10395 }, { "epoch": 1.91, "learning_rate": 1.7809280132231426e-05, "loss": 0.0818, "step": 10396 }, { "epoch": 1.91, "learning_rate": 1.7808816373297867e-05, "loss": 0.087, "step": 10397 }, { "epoch": 1.91, "learning_rate": 1.7808352571321926e-05, "loss": 0.0809, "step": 10398 }, { "epoch": 1.91, "learning_rate": 1.7807888726306164e-05, "loss": 0.0901, "step": 10399 }, { "epoch": 1.91, "learning_rate": 1.7807424838253136e-05, "loss": 0.0892, "step": 10400 }, { "epoch": 1.91, "learning_rate": 1.7806960907165403e-05, "loss": 0.0875, "step": 10401 }, { "epoch": 1.91, "learning_rate": 1.7806496933045516e-05, "loss": 0.0954, "step": 10402 }, { "epoch": 1.91, "learning_rate": 1.7806032915896034e-05, "loss": 0.0952, "step": 10403 }, { "epoch": 1.91, "learning_rate": 1.7805568855719518e-05, "loss": 0.0893, "step": 10404 }, { "epoch": 1.91, "learning_rate": 1.780510475251852e-05, "loss": 0.0974, "step": 10405 }, { "epoch": 1.91, "learning_rate": 1.7804640606295604e-05, "loss": 0.0918, "step": 10406 }, { "epoch": 1.91, "learning_rate": 1.7804176417053327e-05, "loss": 0.0872, "step": 10407 }, { "epoch": 1.91, "learning_rate": 1.7803712184794247e-05, "loss": 0.0895, "step": 10408 }, { "epoch": 1.91, "learning_rate": 1.780324790952092e-05, "loss": 0.0786, "step": 10409 }, { "epoch": 1.91, "learning_rate": 1.7802783591235907e-05, "loss": 0.095, "step": 10410 }, { "epoch": 1.91, "learning_rate": 1.780231922994177e-05, "loss": 0.0916, "step": 10411 }, { "epoch": 1.91, "learning_rate": 1.780185482564107e-05, "loss": 0.0914, "step": 10412 }, { "epoch": 1.91, "learning_rate": 1.7801390378336358e-05, "loss": 0.0856, "step": 10413 }, { "epoch": 1.91, "learning_rate": 1.7800925888030202e-05, "loss": 0.0944, "step": 10414 }, { "epoch": 1.91, "learning_rate": 1.780046135472516e-05, "loss": 0.0994, "step": 10415 }, { "epoch": 1.91, "learning_rate": 1.779999677842379e-05, "loss": 0.0981, "step": 10416 }, { "epoch": 1.91, "learning_rate": 1.779953215912866e-05, "loss": 0.0865, "step": 10417 }, { "epoch": 1.91, "learning_rate": 1.779906749684232e-05, "loss": 0.0902, "step": 10418 }, { "epoch": 1.91, "learning_rate": 1.779860279156734e-05, "loss": 0.0893, "step": 10419 }, { "epoch": 1.91, "learning_rate": 1.779813804330628e-05, "loss": 0.0977, "step": 10420 }, { "epoch": 1.91, "learning_rate": 1.7797673252061698e-05, "loss": 0.0956, "step": 10421 }, { "epoch": 1.91, "learning_rate": 1.779720841783616e-05, "loss": 0.0864, "step": 10422 }, { "epoch": 1.91, "learning_rate": 1.7796743540632226e-05, "loss": 0.0995, "step": 10423 }, { "epoch": 1.91, "learning_rate": 1.7796278620452458e-05, "loss": 0.0935, "step": 10424 }, { "epoch": 1.91, "learning_rate": 1.7795813657299423e-05, "loss": 0.091, "step": 10425 }, { "epoch": 1.91, "learning_rate": 1.7795348651175674e-05, "loss": 0.0853, "step": 10426 }, { "epoch": 1.91, "learning_rate": 1.7794883602083787e-05, "loss": 0.0953, "step": 10427 }, { "epoch": 1.91, "learning_rate": 1.7794418510026313e-05, "loss": 0.0979, "step": 10428 }, { "epoch": 1.91, "learning_rate": 1.7793953375005827e-05, "loss": 0.0895, "step": 10429 }, { "epoch": 1.91, "learning_rate": 1.7793488197024887e-05, "loss": 0.1048, "step": 10430 }, { "epoch": 1.91, "learning_rate": 1.7793022976086054e-05, "loss": 0.0924, "step": 10431 }, { "epoch": 1.91, "learning_rate": 1.77925577121919e-05, "loss": 0.1038, "step": 10432 }, { "epoch": 1.91, "learning_rate": 1.7792092405344984e-05, "loss": 0.0992, "step": 10433 }, { "epoch": 1.91, "learning_rate": 1.779162705554787e-05, "loss": 0.087, "step": 10434 }, { "epoch": 1.91, "learning_rate": 1.7791161662803125e-05, "loss": 0.097, "step": 10435 }, { "epoch": 1.91, "learning_rate": 1.779069622711332e-05, "loss": 0.0807, "step": 10436 }, { "epoch": 1.91, "learning_rate": 1.779023074848101e-05, "loss": 0.0887, "step": 10437 }, { "epoch": 1.91, "learning_rate": 1.7789765226908764e-05, "loss": 0.0963, "step": 10438 }, { "epoch": 1.91, "learning_rate": 1.778929966239915e-05, "loss": 0.0795, "step": 10439 }, { "epoch": 1.91, "learning_rate": 1.7788834054954736e-05, "loss": 0.0956, "step": 10440 }, { "epoch": 1.91, "learning_rate": 1.7788368404578087e-05, "loss": 0.0862, "step": 10441 }, { "epoch": 1.91, "learning_rate": 1.778790271127177e-05, "loss": 0.0822, "step": 10442 }, { "epoch": 1.92, "learning_rate": 1.7787436975038345e-05, "loss": 0.0909, "step": 10443 }, { "epoch": 1.92, "learning_rate": 1.778697119588039e-05, "loss": 0.101, "step": 10444 }, { "epoch": 1.92, "learning_rate": 1.7786505373800465e-05, "loss": 0.0858, "step": 10445 }, { "epoch": 1.92, "learning_rate": 1.778603950880114e-05, "loss": 0.0924, "step": 10446 }, { "epoch": 1.92, "learning_rate": 1.7785573600884984e-05, "loss": 0.1005, "step": 10447 }, { "epoch": 1.92, "learning_rate": 1.7785107650054564e-05, "loss": 0.0785, "step": 10448 }, { "epoch": 1.92, "learning_rate": 1.7784641656312445e-05, "loss": 0.0973, "step": 10449 }, { "epoch": 1.92, "learning_rate": 1.7784175619661203e-05, "loss": 0.092, "step": 10450 }, { "epoch": 1.92, "learning_rate": 1.77837095401034e-05, "loss": 0.0879, "step": 10451 }, { "epoch": 1.92, "learning_rate": 1.7783243417641613e-05, "loss": 0.0922, "step": 10452 }, { "epoch": 1.92, "learning_rate": 1.7782777252278398e-05, "loss": 0.105, "step": 10453 }, { "epoch": 1.92, "learning_rate": 1.778231104401634e-05, "loss": 0.0945, "step": 10454 }, { "epoch": 1.92, "learning_rate": 1.7781844792857997e-05, "loss": 0.0951, "step": 10455 }, { "epoch": 1.92, "learning_rate": 1.778137849880594e-05, "loss": 0.096, "step": 10456 }, { "epoch": 1.92, "learning_rate": 1.778091216186275e-05, "loss": 0.077, "step": 10457 }, { "epoch": 1.92, "learning_rate": 1.7780445782030985e-05, "loss": 0.1055, "step": 10458 }, { "epoch": 1.92, "learning_rate": 1.777997935931322e-05, "loss": 0.102, "step": 10459 }, { "epoch": 1.92, "learning_rate": 1.777951289371203e-05, "loss": 0.0853, "step": 10460 }, { "epoch": 1.92, "learning_rate": 1.7779046385229977e-05, "loss": 0.0786, "step": 10461 }, { "epoch": 1.92, "learning_rate": 1.7778579833869644e-05, "loss": 0.1033, "step": 10462 }, { "epoch": 1.92, "learning_rate": 1.7778113239633594e-05, "loss": 0.0819, "step": 10463 }, { "epoch": 1.92, "learning_rate": 1.77776466025244e-05, "loss": 0.0885, "step": 10464 }, { "epoch": 1.92, "learning_rate": 1.777717992254464e-05, "loss": 0.0786, "step": 10465 }, { "epoch": 1.92, "learning_rate": 1.777671319969688e-05, "loss": 0.0845, "step": 10466 }, { "epoch": 1.92, "learning_rate": 1.7776246433983694e-05, "loss": 0.09, "step": 10467 }, { "epoch": 1.92, "learning_rate": 1.7775779625407657e-05, "loss": 0.0846, "step": 10468 }, { "epoch": 1.92, "learning_rate": 1.777531277397134e-05, "loss": 0.0866, "step": 10469 }, { "epoch": 1.92, "learning_rate": 1.7774845879677314e-05, "loss": 0.1013, "step": 10470 }, { "epoch": 1.92, "learning_rate": 1.777437894252816e-05, "loss": 0.0879, "step": 10471 }, { "epoch": 1.92, "learning_rate": 1.777391196252644e-05, "loss": 0.089, "step": 10472 }, { "epoch": 1.92, "learning_rate": 1.7773444939674745e-05, "loss": 0.0999, "step": 10473 }, { "epoch": 1.92, "learning_rate": 1.7772977873975633e-05, "loss": 0.0969, "step": 10474 }, { "epoch": 1.92, "learning_rate": 1.7772510765431684e-05, "loss": 0.0974, "step": 10475 }, { "epoch": 1.92, "learning_rate": 1.777204361404548e-05, "loss": 0.0889, "step": 10476 }, { "epoch": 1.92, "learning_rate": 1.7771576419819584e-05, "loss": 0.0831, "step": 10477 }, { "epoch": 1.92, "learning_rate": 1.777110918275658e-05, "loss": 0.0911, "step": 10478 }, { "epoch": 1.92, "learning_rate": 1.7770641902859037e-05, "loss": 0.0816, "step": 10479 }, { "epoch": 1.92, "learning_rate": 1.7770174580129532e-05, "loss": 0.0873, "step": 10480 }, { "epoch": 1.92, "learning_rate": 1.776970721457065e-05, "loss": 0.0909, "step": 10481 }, { "epoch": 1.92, "learning_rate": 1.7769239806184952e-05, "loss": 0.0931, "step": 10482 }, { "epoch": 1.92, "learning_rate": 1.7768772354975027e-05, "loss": 0.095, "step": 10483 }, { "epoch": 1.92, "learning_rate": 1.7768304860943442e-05, "loss": 0.0961, "step": 10484 }, { "epoch": 1.92, "learning_rate": 1.7767837324092784e-05, "loss": 0.0988, "step": 10485 }, { "epoch": 1.92, "learning_rate": 1.7767369744425623e-05, "loss": 0.0988, "step": 10486 }, { "epoch": 1.92, "learning_rate": 1.7766902121944535e-05, "loss": 0.0924, "step": 10487 }, { "epoch": 1.92, "learning_rate": 1.7766434456652107e-05, "loss": 0.0968, "step": 10488 }, { "epoch": 1.92, "learning_rate": 1.7765966748550906e-05, "loss": 0.1031, "step": 10489 }, { "epoch": 1.92, "learning_rate": 1.7765498997643512e-05, "loss": 0.0897, "step": 10490 }, { "epoch": 1.92, "learning_rate": 1.776503120393251e-05, "loss": 0.0953, "step": 10491 }, { "epoch": 1.92, "learning_rate": 1.7764563367420475e-05, "loss": 0.0897, "step": 10492 }, { "epoch": 1.92, "learning_rate": 1.776409548810998e-05, "loss": 0.0928, "step": 10493 }, { "epoch": 1.92, "learning_rate": 1.776362756600361e-05, "loss": 0.0821, "step": 10494 }, { "epoch": 1.92, "learning_rate": 1.7763159601103945e-05, "loss": 0.0906, "step": 10495 }, { "epoch": 1.92, "learning_rate": 1.7762691593413564e-05, "loss": 0.0923, "step": 10496 }, { "epoch": 1.92, "learning_rate": 1.776222354293504e-05, "loss": 0.0961, "step": 10497 }, { "epoch": 1.93, "learning_rate": 1.776175544967096e-05, "loss": 0.0812, "step": 10498 }, { "epoch": 1.93, "learning_rate": 1.7761287313623906e-05, "loss": 0.0844, "step": 10499 }, { "epoch": 1.93, "learning_rate": 1.776081913479645e-05, "loss": 0.0861, "step": 10500 }, { "epoch": 1.93, "learning_rate": 1.7760350913191177e-05, "loss": 0.0883, "step": 10501 }, { "epoch": 1.93, "learning_rate": 1.775988264881067e-05, "loss": 0.0964, "step": 10502 }, { "epoch": 1.93, "learning_rate": 1.775941434165751e-05, "loss": 0.0867, "step": 10503 }, { "epoch": 1.93, "learning_rate": 1.7758945991734273e-05, "loss": 0.0882, "step": 10504 }, { "epoch": 1.93, "learning_rate": 1.7758477599043543e-05, "loss": 0.0922, "step": 10505 }, { "epoch": 1.93, "learning_rate": 1.7758009163587906e-05, "loss": 0.0922, "step": 10506 }, { "epoch": 1.93, "learning_rate": 1.7757540685369936e-05, "loss": 0.0826, "step": 10507 }, { "epoch": 1.93, "learning_rate": 1.7757072164392225e-05, "loss": 0.0999, "step": 10508 }, { "epoch": 1.93, "learning_rate": 1.7756603600657348e-05, "loss": 0.0903, "step": 10509 }, { "epoch": 1.93, "learning_rate": 1.775613499416789e-05, "loss": 0.0876, "step": 10510 }, { "epoch": 1.93, "learning_rate": 1.7755666344926433e-05, "loss": 0.0929, "step": 10511 }, { "epoch": 1.93, "learning_rate": 1.775519765293557e-05, "loss": 0.098, "step": 10512 }, { "epoch": 1.93, "learning_rate": 1.7754728918197865e-05, "loss": 0.086, "step": 10513 }, { "epoch": 1.93, "learning_rate": 1.7754260140715918e-05, "loss": 0.087, "step": 10514 }, { "epoch": 1.93, "learning_rate": 1.7753791320492305e-05, "loss": 0.094, "step": 10515 }, { "epoch": 1.93, "learning_rate": 1.7753322457529615e-05, "loss": 0.0918, "step": 10516 }, { "epoch": 1.93, "learning_rate": 1.7752853551830427e-05, "loss": 0.0808, "step": 10517 }, { "epoch": 1.93, "learning_rate": 1.775238460339733e-05, "loss": 0.0837, "step": 10518 }, { "epoch": 1.93, "learning_rate": 1.775191561223291e-05, "loss": 0.0974, "step": 10519 }, { "epoch": 1.93, "learning_rate": 1.7751446578339748e-05, "loss": 0.0937, "step": 10520 }, { "epoch": 1.93, "learning_rate": 1.775097750172043e-05, "loss": 0.1018, "step": 10521 }, { "epoch": 1.93, "learning_rate": 1.775050838237754e-05, "loss": 0.0948, "step": 10522 }, { "epoch": 1.93, "learning_rate": 1.775003922031367e-05, "loss": 0.0937, "step": 10523 }, { "epoch": 1.93, "learning_rate": 1.77495700155314e-05, "loss": 0.102, "step": 10524 }, { "epoch": 1.93, "learning_rate": 1.774910076803332e-05, "loss": 0.0902, "step": 10525 }, { "epoch": 1.93, "learning_rate": 1.7748631477822015e-05, "loss": 0.0729, "step": 10526 }, { "epoch": 1.93, "learning_rate": 1.7748162144900073e-05, "loss": 0.1043, "step": 10527 }, { "epoch": 1.93, "learning_rate": 1.774769276927008e-05, "loss": 0.0874, "step": 10528 }, { "epoch": 1.93, "learning_rate": 1.7747223350934618e-05, "loss": 0.0882, "step": 10529 }, { "epoch": 1.93, "learning_rate": 1.7746753889896283e-05, "loss": 0.0946, "step": 10530 }, { "epoch": 1.93, "learning_rate": 1.7746284386157658e-05, "loss": 0.0808, "step": 10531 }, { "epoch": 1.93, "learning_rate": 1.7745814839721332e-05, "loss": 0.0977, "step": 10532 }, { "epoch": 1.93, "learning_rate": 1.7745345250589897e-05, "loss": 0.0868, "step": 10533 }, { "epoch": 1.93, "learning_rate": 1.774487561876593e-05, "loss": 0.08, "step": 10534 }, { "epoch": 1.93, "learning_rate": 1.7744405944252034e-05, "loss": 0.0921, "step": 10535 }, { "epoch": 1.93, "learning_rate": 1.774393622705079e-05, "loss": 0.1019, "step": 10536 }, { "epoch": 1.93, "learning_rate": 1.7743466467164784e-05, "loss": 0.0777, "step": 10537 }, { "epoch": 1.93, "learning_rate": 1.774299666459661e-05, "loss": 0.1031, "step": 10538 }, { "epoch": 1.93, "learning_rate": 1.774252681934886e-05, "loss": 0.0902, "step": 10539 }, { "epoch": 1.93, "learning_rate": 1.7742056931424124e-05, "loss": 0.0882, "step": 10540 }, { "epoch": 1.93, "learning_rate": 1.7741587000824984e-05, "loss": 0.0932, "step": 10541 }, { "epoch": 1.93, "learning_rate": 1.7741117027554034e-05, "loss": 0.0933, "step": 10542 }, { "epoch": 1.93, "learning_rate": 1.7740647011613868e-05, "loss": 0.0874, "step": 10543 }, { "epoch": 1.93, "learning_rate": 1.774017695300707e-05, "loss": 0.1003, "step": 10544 }, { "epoch": 1.93, "learning_rate": 1.773970685173624e-05, "loss": 0.0888, "step": 10545 }, { "epoch": 1.93, "learning_rate": 1.7739236707803966e-05, "loss": 0.0911, "step": 10546 }, { "epoch": 1.93, "learning_rate": 1.7738766521212834e-05, "loss": 0.0937, "step": 10547 }, { "epoch": 1.93, "learning_rate": 1.773829629196544e-05, "loss": 0.0943, "step": 10548 }, { "epoch": 1.93, "learning_rate": 1.7737826020064377e-05, "loss": 0.0916, "step": 10549 }, { "epoch": 1.93, "learning_rate": 1.7737355705512235e-05, "loss": 0.0784, "step": 10550 }, { "epoch": 1.93, "learning_rate": 1.773688534831161e-05, "loss": 0.0854, "step": 10551 }, { "epoch": 1.94, "learning_rate": 1.773641494846509e-05, "loss": 0.0982, "step": 10552 }, { "epoch": 1.94, "learning_rate": 1.7735944505975268e-05, "loss": 0.0836, "step": 10553 }, { "epoch": 1.94, "learning_rate": 1.7735474020844738e-05, "loss": 0.0772, "step": 10554 }, { "epoch": 1.94, "learning_rate": 1.7735003493076094e-05, "loss": 0.0945, "step": 10555 }, { "epoch": 1.94, "learning_rate": 1.773453292267193e-05, "loss": 0.105, "step": 10556 }, { "epoch": 1.94, "learning_rate": 1.773406230963484e-05, "loss": 0.0902, "step": 10557 }, { "epoch": 1.94, "learning_rate": 1.7733591653967418e-05, "loss": 0.0965, "step": 10558 }, { "epoch": 1.94, "learning_rate": 1.7733120955672257e-05, "loss": 0.0964, "step": 10559 }, { "epoch": 1.94, "learning_rate": 1.773265021475195e-05, "loss": 0.0881, "step": 10560 }, { "epoch": 1.94, "learning_rate": 1.7732179431209097e-05, "loss": 0.0897, "step": 10561 }, { "epoch": 1.94, "learning_rate": 1.773170860504629e-05, "loss": 0.0892, "step": 10562 }, { "epoch": 1.94, "learning_rate": 1.7731237736266122e-05, "loss": 0.085, "step": 10563 }, { "epoch": 1.94, "learning_rate": 1.773076682487119e-05, "loss": 0.087, "step": 10564 }, { "epoch": 1.94, "learning_rate": 1.7730295870864092e-05, "loss": 0.1014, "step": 10565 }, { "epoch": 1.94, "learning_rate": 1.7729824874247425e-05, "loss": 0.0828, "step": 10566 }, { "epoch": 1.94, "learning_rate": 1.7729353835023773e-05, "loss": 0.0932, "step": 10567 }, { "epoch": 1.94, "learning_rate": 1.772888275319575e-05, "loss": 0.0953, "step": 10568 }, { "epoch": 1.94, "learning_rate": 1.772841162876594e-05, "loss": 0.0775, "step": 10569 }, { "epoch": 1.94, "learning_rate": 1.7727940461736948e-05, "loss": 0.082, "step": 10570 }, { "epoch": 1.94, "learning_rate": 1.7727469252111365e-05, "loss": 0.0897, "step": 10571 }, { "epoch": 1.94, "learning_rate": 1.772699799989179e-05, "loss": 0.0751, "step": 10572 }, { "epoch": 1.94, "learning_rate": 1.7726526705080818e-05, "loss": 0.0913, "step": 10573 }, { "epoch": 1.94, "learning_rate": 1.7726055367681058e-05, "loss": 0.0846, "step": 10574 }, { "epoch": 1.94, "learning_rate": 1.7725583987695093e-05, "loss": 0.0962, "step": 10575 }, { "epoch": 1.94, "learning_rate": 1.772511256512553e-05, "loss": 0.0957, "step": 10576 }, { "epoch": 1.94, "learning_rate": 1.7724641099974965e-05, "loss": 0.0906, "step": 10577 }, { "epoch": 1.94, "learning_rate": 1.7724169592245996e-05, "loss": 0.0825, "step": 10578 }, { "epoch": 1.94, "learning_rate": 1.7723698041941225e-05, "loss": 0.0806, "step": 10579 }, { "epoch": 1.94, "learning_rate": 1.772322644906325e-05, "loss": 0.0875, "step": 10580 }, { "epoch": 1.94, "learning_rate": 1.772275481361467e-05, "loss": 0.0938, "step": 10581 }, { "epoch": 1.94, "learning_rate": 1.772228313559808e-05, "loss": 0.0945, "step": 10582 }, { "epoch": 1.94, "learning_rate": 1.7721811415016092e-05, "loss": 0.0937, "step": 10583 }, { "epoch": 1.94, "learning_rate": 1.7721339651871294e-05, "loss": 0.0809, "step": 10584 }, { "epoch": 1.94, "learning_rate": 1.7720867846166292e-05, "loss": 0.1006, "step": 10585 }, { "epoch": 1.94, "learning_rate": 1.7720395997903685e-05, "loss": 0.0874, "step": 10586 }, { "epoch": 1.94, "learning_rate": 1.7719924107086074e-05, "loss": 0.0948, "step": 10587 }, { "epoch": 1.94, "learning_rate": 1.7719452173716063e-05, "loss": 0.0872, "step": 10588 }, { "epoch": 1.94, "learning_rate": 1.7718980197796247e-05, "loss": 0.0825, "step": 10589 }, { "epoch": 1.94, "learning_rate": 1.7718508179329233e-05, "loss": 0.0814, "step": 10590 }, { "epoch": 1.94, "learning_rate": 1.771803611831762e-05, "loss": 0.0924, "step": 10591 }, { "epoch": 1.94, "learning_rate": 1.7717564014764014e-05, "loss": 0.0929, "step": 10592 }, { "epoch": 1.94, "learning_rate": 1.7717091868671016e-05, "loss": 0.0929, "step": 10593 }, { "epoch": 1.94, "learning_rate": 1.771661968004122e-05, "loss": 0.0888, "step": 10594 }, { "epoch": 1.94, "learning_rate": 1.771614744887724e-05, "loss": 0.0807, "step": 10595 }, { "epoch": 1.94, "learning_rate": 1.7715675175181672e-05, "loss": 0.0925, "step": 10596 }, { "epoch": 1.94, "learning_rate": 1.7715202858957123e-05, "loss": 0.0961, "step": 10597 }, { "epoch": 1.94, "learning_rate": 1.7714730500206197e-05, "loss": 0.0971, "step": 10598 }, { "epoch": 1.94, "learning_rate": 1.7714258098931493e-05, "loss": 0.1059, "step": 10599 }, { "epoch": 1.94, "learning_rate": 1.771378565513562e-05, "loss": 0.09, "step": 10600 }, { "epoch": 1.94, "learning_rate": 1.771331316882118e-05, "loss": 0.0928, "step": 10601 }, { "epoch": 1.94, "learning_rate": 1.7712840639990773e-05, "loss": 0.0835, "step": 10602 }, { "epoch": 1.94, "learning_rate": 1.771236806864701e-05, "loss": 0.1073, "step": 10603 }, { "epoch": 1.94, "learning_rate": 1.7711895454792496e-05, "loss": 0.0913, "step": 10604 }, { "epoch": 1.94, "learning_rate": 1.771142279842983e-05, "loss": 0.0889, "step": 10605 }, { "epoch": 1.94, "learning_rate": 1.7710950099561625e-05, "loss": 0.0926, "step": 10606 }, { "epoch": 1.95, "learning_rate": 1.771047735819048e-05, "loss": 0.0937, "step": 10607 }, { "epoch": 1.95, "learning_rate": 1.7710004574319e-05, "loss": 0.0918, "step": 10608 }, { "epoch": 1.95, "learning_rate": 1.7709531747949796e-05, "loss": 0.1031, "step": 10609 }, { "epoch": 1.95, "learning_rate": 1.7709058879085475e-05, "loss": 0.0926, "step": 10610 }, { "epoch": 1.95, "learning_rate": 1.770858596772864e-05, "loss": 0.0934, "step": 10611 }, { "epoch": 1.95, "learning_rate": 1.7708113013881897e-05, "loss": 0.0798, "step": 10612 }, { "epoch": 1.95, "learning_rate": 1.7707640017547855e-05, "loss": 0.095, "step": 10613 }, { "epoch": 1.95, "learning_rate": 1.770716697872912e-05, "loss": 0.0982, "step": 10614 }, { "epoch": 1.95, "learning_rate": 1.7706693897428303e-05, "loss": 0.09, "step": 10615 }, { "epoch": 1.95, "learning_rate": 1.7706220773648006e-05, "loss": 0.0867, "step": 10616 }, { "epoch": 1.95, "learning_rate": 1.770574760739084e-05, "loss": 0.0829, "step": 10617 }, { "epoch": 1.95, "learning_rate": 1.7705274398659415e-05, "loss": 0.0889, "step": 10618 }, { "epoch": 1.95, "learning_rate": 1.7704801147456333e-05, "loss": 0.0876, "step": 10619 }, { "epoch": 1.95, "learning_rate": 1.770432785378421e-05, "loss": 0.0835, "step": 10620 }, { "epoch": 1.95, "learning_rate": 1.7703854517645645e-05, "loss": 0.0902, "step": 10621 }, { "epoch": 1.95, "learning_rate": 1.7703381139043262e-05, "loss": 0.0672, "step": 10622 }, { "epoch": 1.95, "learning_rate": 1.7702907717979657e-05, "loss": 0.0782, "step": 10623 }, { "epoch": 1.95, "learning_rate": 1.7702434254457447e-05, "loss": 0.0936, "step": 10624 }, { "epoch": 1.95, "learning_rate": 1.7701960748479236e-05, "loss": 0.1088, "step": 10625 }, { "epoch": 1.95, "learning_rate": 1.7701487200047638e-05, "loss": 0.0835, "step": 10626 }, { "epoch": 1.95, "learning_rate": 1.770101360916526e-05, "loss": 0.0989, "step": 10627 }, { "epoch": 1.95, "learning_rate": 1.770053997583472e-05, "loss": 0.0958, "step": 10628 }, { "epoch": 1.95, "learning_rate": 1.7700066300058617e-05, "loss": 0.0976, "step": 10629 }, { "epoch": 1.95, "learning_rate": 1.7699592581839573e-05, "loss": 0.089, "step": 10630 }, { "epoch": 1.95, "learning_rate": 1.769911882118019e-05, "loss": 0.0831, "step": 10631 }, { "epoch": 1.95, "learning_rate": 1.769864501808309e-05, "loss": 0.0944, "step": 10632 }, { "epoch": 1.95, "learning_rate": 1.7698171172550874e-05, "loss": 0.0924, "step": 10633 }, { "epoch": 1.95, "learning_rate": 1.7697697284586157e-05, "loss": 0.0934, "step": 10634 }, { "epoch": 1.95, "learning_rate": 1.7697223354191553e-05, "loss": 0.0969, "step": 10635 }, { "epoch": 1.95, "learning_rate": 1.7696749381369676e-05, "loss": 0.0779, "step": 10636 }, { "epoch": 1.95, "learning_rate": 1.769627536612313e-05, "loss": 0.0913, "step": 10637 }, { "epoch": 1.95, "learning_rate": 1.7695801308454543e-05, "loss": 0.091, "step": 10638 }, { "epoch": 1.95, "learning_rate": 1.7695327208366514e-05, "loss": 0.1102, "step": 10639 }, { "epoch": 1.95, "learning_rate": 1.769485306586166e-05, "loss": 0.0907, "step": 10640 }, { "epoch": 1.95, "learning_rate": 1.7694378880942598e-05, "loss": 0.0935, "step": 10641 }, { "epoch": 1.95, "learning_rate": 1.7693904653611938e-05, "loss": 0.0904, "step": 10642 }, { "epoch": 1.95, "learning_rate": 1.7693430383872297e-05, "loss": 0.0901, "step": 10643 }, { "epoch": 1.95, "learning_rate": 1.7692956071726284e-05, "loss": 0.0889, "step": 10644 }, { "epoch": 1.95, "learning_rate": 1.769248171717652e-05, "loss": 0.0802, "step": 10645 }, { "epoch": 1.95, "learning_rate": 1.7692007320225615e-05, "loss": 0.0909, "step": 10646 }, { "epoch": 1.95, "learning_rate": 1.7691532880876185e-05, "loss": 0.0821, "step": 10647 }, { "epoch": 1.95, "learning_rate": 1.7691058399130848e-05, "loss": 0.1003, "step": 10648 }, { "epoch": 1.95, "learning_rate": 1.7690583874992214e-05, "loss": 0.0925, "step": 10649 }, { "epoch": 1.95, "learning_rate": 1.7690109308462907e-05, "loss": 0.075, "step": 10650 }, { "epoch": 1.95, "learning_rate": 1.768963469954553e-05, "loss": 0.1011, "step": 10651 }, { "epoch": 1.95, "learning_rate": 1.768916004824271e-05, "loss": 0.0959, "step": 10652 }, { "epoch": 1.95, "learning_rate": 1.768868535455706e-05, "loss": 0.0852, "step": 10653 }, { "epoch": 1.95, "learning_rate": 1.76882106184912e-05, "loss": 0.0935, "step": 10654 }, { "epoch": 1.95, "learning_rate": 1.7687735840047733e-05, "loss": 0.08, "step": 10655 }, { "epoch": 1.95, "learning_rate": 1.7687261019229294e-05, "loss": 0.0854, "step": 10656 }, { "epoch": 1.95, "learning_rate": 1.7686786156038488e-05, "loss": 0.0821, "step": 10657 }, { "epoch": 1.95, "learning_rate": 1.768631125047794e-05, "loss": 0.0896, "step": 10658 }, { "epoch": 1.95, "learning_rate": 1.768583630255026e-05, "loss": 0.0955, "step": 10659 }, { "epoch": 1.95, "learning_rate": 1.7685361312258074e-05, "loss": 0.0759, "step": 10660 }, { "epoch": 1.96, "learning_rate": 1.7684886279603996e-05, "loss": 0.0885, "step": 10661 }, { "epoch": 1.96, "learning_rate": 1.7684411204590644e-05, "loss": 0.0861, "step": 10662 }, { "epoch": 1.96, "learning_rate": 1.7683936087220636e-05, "loss": 0.0892, "step": 10663 }, { "epoch": 1.96, "learning_rate": 1.7683460927496593e-05, "loss": 0.0935, "step": 10664 }, { "epoch": 1.96, "learning_rate": 1.7682985725421136e-05, "loss": 0.0776, "step": 10665 }, { "epoch": 1.96, "learning_rate": 1.768251048099688e-05, "loss": 0.0867, "step": 10666 }, { "epoch": 1.96, "learning_rate": 1.7682035194226447e-05, "loss": 0.099, "step": 10667 }, { "epoch": 1.96, "learning_rate": 1.7681559865112454e-05, "loss": 0.0967, "step": 10668 }, { "epoch": 1.96, "learning_rate": 1.7681084493657526e-05, "loss": 0.0962, "step": 10669 }, { "epoch": 1.96, "learning_rate": 1.7680609079864276e-05, "loss": 0.0912, "step": 10670 }, { "epoch": 1.96, "learning_rate": 1.7680133623735334e-05, "loss": 0.0894, "step": 10671 }, { "epoch": 1.96, "learning_rate": 1.767965812527331e-05, "loss": 0.0962, "step": 10672 }, { "epoch": 1.96, "learning_rate": 1.7679182584480834e-05, "loss": 0.0968, "step": 10673 }, { "epoch": 1.96, "learning_rate": 1.767870700136052e-05, "loss": 0.0837, "step": 10674 }, { "epoch": 1.96, "learning_rate": 1.7678231375915002e-05, "loss": 0.0933, "step": 10675 }, { "epoch": 1.96, "learning_rate": 1.7677755708146885e-05, "loss": 0.0897, "step": 10676 }, { "epoch": 1.96, "learning_rate": 1.76772799980588e-05, "loss": 0.0881, "step": 10677 }, { "epoch": 1.96, "learning_rate": 1.767680424565337e-05, "loss": 0.0902, "step": 10678 }, { "epoch": 1.96, "learning_rate": 1.767632845093321e-05, "loss": 0.0843, "step": 10679 }, { "epoch": 1.96, "learning_rate": 1.7675852613900955e-05, "loss": 0.0917, "step": 10680 }, { "epoch": 1.96, "learning_rate": 1.7675376734559214e-05, "loss": 0.0886, "step": 10681 }, { "epoch": 1.96, "learning_rate": 1.767490081291062e-05, "loss": 0.0985, "step": 10682 }, { "epoch": 1.96, "learning_rate": 1.7674424848957793e-05, "loss": 0.0771, "step": 10683 }, { "epoch": 1.96, "learning_rate": 1.7673948842703357e-05, "loss": 0.0892, "step": 10684 }, { "epoch": 1.96, "learning_rate": 1.7673472794149932e-05, "loss": 0.0846, "step": 10685 }, { "epoch": 1.96, "learning_rate": 1.7672996703300147e-05, "loss": 0.0777, "step": 10686 }, { "epoch": 1.96, "learning_rate": 1.7672520570156625e-05, "loss": 0.1133, "step": 10687 }, { "epoch": 1.96, "learning_rate": 1.7672044394721988e-05, "loss": 0.0935, "step": 10688 }, { "epoch": 1.96, "learning_rate": 1.7671568176998865e-05, "loss": 0.0852, "step": 10689 }, { "epoch": 1.96, "learning_rate": 1.767109191698988e-05, "loss": 0.0887, "step": 10690 }, { "epoch": 1.96, "learning_rate": 1.7670615614697652e-05, "loss": 0.0982, "step": 10691 }, { "epoch": 1.96, "learning_rate": 1.7670139270124813e-05, "loss": 0.0859, "step": 10692 }, { "epoch": 1.96, "learning_rate": 1.7669662883273988e-05, "loss": 0.0898, "step": 10693 }, { "epoch": 1.96, "learning_rate": 1.76691864541478e-05, "loss": 0.0903, "step": 10694 }, { "epoch": 1.96, "learning_rate": 1.7668709982748876e-05, "loss": 0.0843, "step": 10695 }, { "epoch": 1.96, "learning_rate": 1.7668233469079845e-05, "loss": 0.0885, "step": 10696 }, { "epoch": 1.96, "learning_rate": 1.766775691314333e-05, "loss": 0.0868, "step": 10697 }, { "epoch": 1.96, "learning_rate": 1.766728031494196e-05, "loss": 0.0858, "step": 10698 }, { "epoch": 1.96, "learning_rate": 1.766680367447836e-05, "loss": 0.0912, "step": 10699 }, { "epoch": 1.96, "learning_rate": 1.766632699175516e-05, "loss": 0.0889, "step": 10700 }, { "epoch": 1.96, "learning_rate": 1.7665850266774985e-05, "loss": 0.0989, "step": 10701 }, { "epoch": 1.96, "learning_rate": 1.7665373499540464e-05, "loss": 0.0944, "step": 10702 }, { "epoch": 1.96, "learning_rate": 1.7664896690054225e-05, "loss": 0.0766, "step": 10703 }, { "epoch": 1.96, "learning_rate": 1.7664419838318894e-05, "loss": 0.0887, "step": 10704 }, { "epoch": 1.96, "learning_rate": 1.7663942944337104e-05, "loss": 0.0982, "step": 10705 }, { "epoch": 1.96, "learning_rate": 1.766346600811148e-05, "loss": 0.0984, "step": 10706 }, { "epoch": 1.96, "learning_rate": 1.766298902964465e-05, "loss": 0.0922, "step": 10707 }, { "epoch": 1.96, "learning_rate": 1.766251200893925e-05, "loss": 0.0879, "step": 10708 }, { "epoch": 1.96, "learning_rate": 1.7662034945997903e-05, "loss": 0.0903, "step": 10709 }, { "epoch": 1.96, "learning_rate": 1.7661557840823236e-05, "loss": 0.0833, "step": 10710 }, { "epoch": 1.96, "learning_rate": 1.7661080693417886e-05, "loss": 0.084, "step": 10711 }, { "epoch": 1.96, "learning_rate": 1.7660603503784477e-05, "loss": 0.079, "step": 10712 }, { "epoch": 1.96, "learning_rate": 1.766012627192564e-05, "loss": 0.0888, "step": 10713 }, { "epoch": 1.96, "learning_rate": 1.7659648997844013e-05, "loss": 0.0891, "step": 10714 }, { "epoch": 1.96, "learning_rate": 1.765917168154222e-05, "loss": 0.0924, "step": 10715 }, { "epoch": 1.97, "learning_rate": 1.765869432302289e-05, "loss": 0.087, "step": 10716 }, { "epoch": 1.97, "learning_rate": 1.7658216922288664e-05, "loss": 0.0756, "step": 10717 }, { "epoch": 1.97, "learning_rate": 1.765773947934216e-05, "loss": 0.0899, "step": 10718 }, { "epoch": 1.97, "learning_rate": 1.765726199418602e-05, "loss": 0.0929, "step": 10719 }, { "epoch": 1.97, "learning_rate": 1.7656784466822872e-05, "loss": 0.1048, "step": 10720 }, { "epoch": 1.97, "learning_rate": 1.765630689725535e-05, "loss": 0.0893, "step": 10721 }, { "epoch": 1.97, "learning_rate": 1.7655829285486082e-05, "loss": 0.0882, "step": 10722 }, { "epoch": 1.97, "learning_rate": 1.7655351631517702e-05, "loss": 0.0938, "step": 10723 }, { "epoch": 1.97, "learning_rate": 1.765487393535285e-05, "loss": 0.0912, "step": 10724 }, { "epoch": 1.97, "learning_rate": 1.765439619699415e-05, "loss": 0.0853, "step": 10725 }, { "epoch": 1.97, "learning_rate": 1.7653918416444243e-05, "loss": 0.0815, "step": 10726 }, { "epoch": 1.97, "learning_rate": 1.7653440593705754e-05, "loss": 0.0932, "step": 10727 }, { "epoch": 1.97, "learning_rate": 1.765296272878132e-05, "loss": 0.1037, "step": 10728 }, { "epoch": 1.97, "learning_rate": 1.7652484821673577e-05, "loss": 0.099, "step": 10729 }, { "epoch": 1.97, "learning_rate": 1.7652006872385165e-05, "loss": 0.114, "step": 10730 }, { "epoch": 1.97, "learning_rate": 1.7651528880918706e-05, "loss": 0.0834, "step": 10731 }, { "epoch": 1.97, "learning_rate": 1.765105084727684e-05, "loss": 0.0893, "step": 10732 }, { "epoch": 1.97, "learning_rate": 1.7650572771462206e-05, "loss": 0.085, "step": 10733 }, { "epoch": 1.97, "learning_rate": 1.765009465347743e-05, "loss": 0.0889, "step": 10734 }, { "epoch": 1.97, "learning_rate": 1.764961649332516e-05, "loss": 0.0878, "step": 10735 }, { "epoch": 1.97, "learning_rate": 1.764913829100802e-05, "loss": 0.1046, "step": 10736 }, { "epoch": 1.97, "learning_rate": 1.7648660046528655e-05, "loss": 0.0918, "step": 10737 }, { "epoch": 1.97, "learning_rate": 1.764818175988969e-05, "loss": 0.0808, "step": 10738 }, { "epoch": 1.97, "learning_rate": 1.764770343109377e-05, "loss": 0.0892, "step": 10739 }, { "epoch": 1.97, "learning_rate": 1.764722506014353e-05, "loss": 0.0937, "step": 10740 }, { "epoch": 1.97, "learning_rate": 1.764674664704161e-05, "loss": 0.0828, "step": 10741 }, { "epoch": 1.97, "learning_rate": 1.764626819179064e-05, "loss": 0.0958, "step": 10742 }, { "epoch": 1.97, "learning_rate": 1.764578969439326e-05, "loss": 0.0873, "step": 10743 }, { "epoch": 1.97, "learning_rate": 1.7645311154852112e-05, "loss": 0.0973, "step": 10744 }, { "epoch": 1.97, "learning_rate": 1.7644832573169826e-05, "loss": 0.104, "step": 10745 }, { "epoch": 1.97, "learning_rate": 1.7644353949349047e-05, "loss": 0.0863, "step": 10746 }, { "epoch": 1.97, "learning_rate": 1.7643875283392407e-05, "loss": 0.0984, "step": 10747 }, { "epoch": 1.97, "learning_rate": 1.764339657530255e-05, "loss": 0.0897, "step": 10748 }, { "epoch": 1.97, "learning_rate": 1.764291782508211e-05, "loss": 0.0854, "step": 10749 }, { "epoch": 1.97, "learning_rate": 1.7642439032733726e-05, "loss": 0.0777, "step": 10750 }, { "epoch": 1.97, "learning_rate": 1.764196019826004e-05, "loss": 0.1022, "step": 10751 }, { "epoch": 1.97, "learning_rate": 1.764148132166369e-05, "loss": 0.0789, "step": 10752 }, { "epoch": 1.97, "learning_rate": 1.7641002402947317e-05, "loss": 0.0741, "step": 10753 }, { "epoch": 1.97, "learning_rate": 1.764052344211356e-05, "loss": 0.0866, "step": 10754 }, { "epoch": 1.97, "learning_rate": 1.7640044439165063e-05, "loss": 0.0906, "step": 10755 }, { "epoch": 1.97, "learning_rate": 1.7639565394104458e-05, "loss": 0.089, "step": 10756 }, { "epoch": 1.97, "learning_rate": 1.7639086306934386e-05, "loss": 0.0878, "step": 10757 }, { "epoch": 1.97, "learning_rate": 1.7638607177657497e-05, "loss": 0.0909, "step": 10758 }, { "epoch": 1.97, "learning_rate": 1.7638128006276422e-05, "loss": 0.0877, "step": 10759 }, { "epoch": 1.97, "learning_rate": 1.7637648792793812e-05, "loss": 0.0971, "step": 10760 }, { "epoch": 1.97, "learning_rate": 1.7637169537212296e-05, "loss": 0.0883, "step": 10761 }, { "epoch": 1.97, "learning_rate": 1.763669023953453e-05, "loss": 0.095, "step": 10762 }, { "epoch": 1.97, "learning_rate": 1.7636210899763143e-05, "loss": 0.0865, "step": 10763 }, { "epoch": 1.97, "learning_rate": 1.7635731517900782e-05, "loss": 0.0984, "step": 10764 }, { "epoch": 1.97, "learning_rate": 1.7635252093950095e-05, "loss": 0.0986, "step": 10765 }, { "epoch": 1.97, "learning_rate": 1.7634772627913716e-05, "loss": 0.086, "step": 10766 }, { "epoch": 1.97, "learning_rate": 1.763429311979429e-05, "loss": 0.1031, "step": 10767 }, { "epoch": 1.97, "learning_rate": 1.7633813569594464e-05, "loss": 0.0972, "step": 10768 }, { "epoch": 1.97, "learning_rate": 1.763333397731688e-05, "loss": 0.0982, "step": 10769 }, { "epoch": 1.98, "learning_rate": 1.763285434296418e-05, "loss": 0.0908, "step": 10770 }, { "epoch": 1.98, "learning_rate": 1.7632374666539004e-05, "loss": 0.0786, "step": 10771 }, { "epoch": 1.98, "learning_rate": 1.7631894948044005e-05, "loss": 0.0987, "step": 10772 }, { "epoch": 1.98, "learning_rate": 1.7631415187481818e-05, "loss": 0.0878, "step": 10773 }, { "epoch": 1.98, "learning_rate": 1.7630935384855095e-05, "loss": 0.0865, "step": 10774 }, { "epoch": 1.98, "learning_rate": 1.7630455540166473e-05, "loss": 0.0927, "step": 10775 }, { "epoch": 1.98, "learning_rate": 1.7629975653418607e-05, "loss": 0.0825, "step": 10776 }, { "epoch": 1.98, "learning_rate": 1.7629495724614132e-05, "loss": 0.0819, "step": 10777 }, { "epoch": 1.98, "learning_rate": 1.76290157537557e-05, "loss": 0.0872, "step": 10778 }, { "epoch": 1.98, "learning_rate": 1.7628535740845954e-05, "loss": 0.1005, "step": 10779 }, { "epoch": 1.98, "learning_rate": 1.7628055685887543e-05, "loss": 0.0843, "step": 10780 }, { "epoch": 1.98, "learning_rate": 1.7627575588883104e-05, "loss": 0.1001, "step": 10781 }, { "epoch": 1.98, "learning_rate": 1.7627095449835297e-05, "loss": 0.0905, "step": 10782 }, { "epoch": 1.98, "learning_rate": 1.7626615268746755e-05, "loss": 0.0771, "step": 10783 }, { "epoch": 1.98, "learning_rate": 1.7626135045620137e-05, "loss": 0.0985, "step": 10784 }, { "epoch": 1.98, "learning_rate": 1.762565478045808e-05, "loss": 0.0911, "step": 10785 }, { "epoch": 1.98, "learning_rate": 1.7625174473263234e-05, "loss": 0.0906, "step": 10786 }, { "epoch": 1.98, "learning_rate": 1.762469412403825e-05, "loss": 0.0802, "step": 10787 }, { "epoch": 1.98, "learning_rate": 1.7624213732785772e-05, "loss": 0.1102, "step": 10788 }, { "epoch": 1.98, "learning_rate": 1.7623733299508454e-05, "loss": 0.0989, "step": 10789 }, { "epoch": 1.98, "learning_rate": 1.7623252824208933e-05, "loss": 0.0892, "step": 10790 }, { "epoch": 1.98, "learning_rate": 1.7622772306889866e-05, "loss": 0.0964, "step": 10791 }, { "epoch": 1.98, "learning_rate": 1.76222917475539e-05, "loss": 0.0936, "step": 10792 }, { "epoch": 1.98, "learning_rate": 1.7621811146203682e-05, "loss": 0.0865, "step": 10793 }, { "epoch": 1.98, "learning_rate": 1.7621330502841864e-05, "loss": 0.0913, "step": 10794 }, { "epoch": 1.98, "learning_rate": 1.7620849817471094e-05, "loss": 0.0939, "step": 10795 }, { "epoch": 1.98, "learning_rate": 1.762036909009402e-05, "loss": 0.0916, "step": 10796 }, { "epoch": 1.98, "learning_rate": 1.7619888320713295e-05, "loss": 0.0924, "step": 10797 }, { "epoch": 1.98, "learning_rate": 1.7619407509331562e-05, "loss": 0.0902, "step": 10798 }, { "epoch": 1.98, "learning_rate": 1.7618926655951482e-05, "loss": 0.0926, "step": 10799 }, { "epoch": 1.98, "learning_rate": 1.76184457605757e-05, "loss": 0.0794, "step": 10800 }, { "epoch": 1.98, "learning_rate": 1.761796482320686e-05, "loss": 0.085, "step": 10801 }, { "epoch": 1.98, "learning_rate": 1.7617483843847628e-05, "loss": 0.087, "step": 10802 }, { "epoch": 1.98, "learning_rate": 1.761700282250064e-05, "loss": 0.0712, "step": 10803 }, { "epoch": 1.98, "learning_rate": 1.7616521759168557e-05, "loss": 0.0954, "step": 10804 }, { "epoch": 1.98, "learning_rate": 1.7616040653854027e-05, "loss": 0.1041, "step": 10805 }, { "epoch": 1.98, "learning_rate": 1.76155595065597e-05, "loss": 0.0944, "step": 10806 }, { "epoch": 1.98, "learning_rate": 1.7615078317288235e-05, "loss": 0.0932, "step": 10807 }, { "epoch": 1.98, "learning_rate": 1.761459708604228e-05, "loss": 0.0982, "step": 10808 }, { "epoch": 1.98, "learning_rate": 1.7614115812824484e-05, "loss": 0.0897, "step": 10809 }, { "epoch": 1.98, "learning_rate": 1.76136344976375e-05, "loss": 0.0894, "step": 10810 }, { "epoch": 1.98, "learning_rate": 1.7613153140483993e-05, "loss": 0.0942, "step": 10811 }, { "epoch": 1.98, "learning_rate": 1.76126717413666e-05, "loss": 0.0902, "step": 10812 }, { "epoch": 1.98, "learning_rate": 1.761219030028799e-05, "loss": 0.0907, "step": 10813 }, { "epoch": 1.98, "learning_rate": 1.7611708817250803e-05, "loss": 0.0896, "step": 10814 }, { "epoch": 1.98, "learning_rate": 1.76112272922577e-05, "loss": 0.0784, "step": 10815 }, { "epoch": 1.98, "learning_rate": 1.761074572531133e-05, "loss": 0.0777, "step": 10816 }, { "epoch": 1.98, "learning_rate": 1.7610264116414356e-05, "loss": 0.092, "step": 10817 }, { "epoch": 1.98, "learning_rate": 1.7609782465569426e-05, "loss": 0.0885, "step": 10818 }, { "epoch": 1.98, "learning_rate": 1.7609300772779197e-05, "loss": 0.0867, "step": 10819 }, { "epoch": 1.98, "learning_rate": 1.7608819038046326e-05, "loss": 0.0816, "step": 10820 }, { "epoch": 1.98, "learning_rate": 1.7608337261373466e-05, "loss": 0.0903, "step": 10821 }, { "epoch": 1.98, "learning_rate": 1.760785544276327e-05, "loss": 0.1032, "step": 10822 }, { "epoch": 1.98, "learning_rate": 1.76073735822184e-05, "loss": 0.0895, "step": 10823 }, { "epoch": 1.98, "learning_rate": 1.7606891679741505e-05, "loss": 0.0921, "step": 10824 }, { "epoch": 1.99, "learning_rate": 1.7606409735335247e-05, "loss": 0.0937, "step": 10825 }, { "epoch": 1.99, "learning_rate": 1.760592774900228e-05, "loss": 0.0989, "step": 10826 }, { "epoch": 1.99, "learning_rate": 1.760544572074526e-05, "loss": 0.0912, "step": 10827 }, { "epoch": 1.99, "learning_rate": 1.7604963650566847e-05, "loss": 0.0747, "step": 10828 }, { "epoch": 1.99, "learning_rate": 1.7604481538469692e-05, "loss": 0.0916, "step": 10829 }, { "epoch": 1.99, "learning_rate": 1.7603999384456464e-05, "loss": 0.0921, "step": 10830 }, { "epoch": 1.99, "learning_rate": 1.760351718852981e-05, "loss": 0.0797, "step": 10831 }, { "epoch": 1.99, "learning_rate": 1.7603034950692386e-05, "loss": 0.0936, "step": 10832 }, { "epoch": 1.99, "learning_rate": 1.760255267094686e-05, "loss": 0.0813, "step": 10833 }, { "epoch": 1.99, "learning_rate": 1.7602070349295886e-05, "loss": 0.0859, "step": 10834 }, { "epoch": 1.99, "learning_rate": 1.760158798574212e-05, "loss": 0.0889, "step": 10835 }, { "epoch": 1.99, "learning_rate": 1.7601105580288225e-05, "loss": 0.0936, "step": 10836 }, { "epoch": 1.99, "learning_rate": 1.7600623132936856e-05, "loss": 0.0909, "step": 10837 }, { "epoch": 1.99, "learning_rate": 1.7600140643690674e-05, "loss": 0.0879, "step": 10838 }, { "epoch": 1.99, "learning_rate": 1.759965811255234e-05, "loss": 0.0955, "step": 10839 }, { "epoch": 1.99, "learning_rate": 1.7599175539524514e-05, "loss": 0.0884, "step": 10840 }, { "epoch": 1.99, "learning_rate": 1.7598692924609852e-05, "loss": 0.0949, "step": 10841 }, { "epoch": 1.99, "learning_rate": 1.7598210267811016e-05, "loss": 0.0941, "step": 10842 }, { "epoch": 1.99, "learning_rate": 1.759772756913067e-05, "loss": 0.0954, "step": 10843 }, { "epoch": 1.99, "learning_rate": 1.759724482857147e-05, "loss": 0.0895, "step": 10844 }, { "epoch": 1.99, "learning_rate": 1.759676204613608e-05, "loss": 0.083, "step": 10845 }, { "epoch": 1.99, "learning_rate": 1.7596279221827153e-05, "loss": 0.1001, "step": 10846 }, { "epoch": 1.99, "learning_rate": 1.7595796355647367e-05, "loss": 0.0839, "step": 10847 }, { "epoch": 1.99, "learning_rate": 1.7595313447599368e-05, "loss": 0.0959, "step": 10848 }, { "epoch": 1.99, "learning_rate": 1.759483049768582e-05, "loss": 0.0945, "step": 10849 }, { "epoch": 1.99, "learning_rate": 1.7594347505909395e-05, "loss": 0.0955, "step": 10850 }, { "epoch": 1.99, "learning_rate": 1.7593864472272746e-05, "loss": 0.0918, "step": 10851 }, { "epoch": 1.99, "learning_rate": 1.759338139677854e-05, "loss": 0.0946, "step": 10852 }, { "epoch": 1.99, "learning_rate": 1.7592898279429434e-05, "loss": 0.106, "step": 10853 }, { "epoch": 1.99, "learning_rate": 1.7592415120228096e-05, "loss": 0.0821, "step": 10854 }, { "epoch": 1.99, "learning_rate": 1.7591931919177188e-05, "loss": 0.083, "step": 10855 }, { "epoch": 1.99, "learning_rate": 1.7591448676279376e-05, "loss": 0.0638, "step": 10856 }, { "epoch": 1.99, "learning_rate": 1.7590965391537316e-05, "loss": 0.086, "step": 10857 }, { "epoch": 1.99, "learning_rate": 1.759048206495368e-05, "loss": 0.0852, "step": 10858 }, { "epoch": 1.99, "learning_rate": 1.758999869653113e-05, "loss": 0.0771, "step": 10859 }, { "epoch": 1.99, "learning_rate": 1.758951528627233e-05, "loss": 0.0808, "step": 10860 }, { "epoch": 1.99, "learning_rate": 1.758903183417994e-05, "loss": 0.0864, "step": 10861 }, { "epoch": 1.99, "learning_rate": 1.7588548340256632e-05, "loss": 0.0995, "step": 10862 }, { "epoch": 1.99, "learning_rate": 1.7588064804505066e-05, "loss": 0.0794, "step": 10863 }, { "epoch": 1.99, "learning_rate": 1.758758122692791e-05, "loss": 0.0918, "step": 10864 }, { "epoch": 1.99, "learning_rate": 1.7587097607527827e-05, "loss": 0.0919, "step": 10865 }, { "epoch": 1.99, "learning_rate": 1.7586613946307486e-05, "loss": 0.0909, "step": 10866 }, { "epoch": 1.99, "learning_rate": 1.7586130243269553e-05, "loss": 0.0928, "step": 10867 }, { "epoch": 1.99, "learning_rate": 1.758564649841669e-05, "loss": 0.0922, "step": 10868 }, { "epoch": 1.99, "learning_rate": 1.7585162711751565e-05, "loss": 0.0894, "step": 10869 }, { "epoch": 1.99, "learning_rate": 1.7584678883276848e-05, "loss": 0.0807, "step": 10870 }, { "epoch": 1.99, "learning_rate": 1.75841950129952e-05, "loss": 0.0874, "step": 10871 }, { "epoch": 1.99, "learning_rate": 1.7583711100909294e-05, "loss": 0.0911, "step": 10872 }, { "epoch": 1.99, "learning_rate": 1.7583227147021794e-05, "loss": 0.0938, "step": 10873 }, { "epoch": 1.99, "learning_rate": 1.758274315133537e-05, "loss": 0.0922, "step": 10874 }, { "epoch": 1.99, "learning_rate": 1.7582259113852684e-05, "loss": 0.0901, "step": 10875 }, { "epoch": 1.99, "learning_rate": 1.758177503457641e-05, "loss": 0.0796, "step": 10876 }, { "epoch": 1.99, "learning_rate": 1.7581290913509214e-05, "loss": 0.0942, "step": 10877 }, { "epoch": 1.99, "learning_rate": 1.7580806750653767e-05, "loss": 0.0976, "step": 10878 }, { "epoch": 2.0, "learning_rate": 1.7580322546012734e-05, "loss": 0.0786, "step": 10879 }, { "epoch": 2.0, "learning_rate": 1.7579838299588785e-05, "loss": 0.1038, "step": 10880 }, { "epoch": 2.0, "learning_rate": 1.7579354011384594e-05, "loss": 0.0859, "step": 10881 }, { "epoch": 2.0, "learning_rate": 1.757886968140282e-05, "loss": 0.0976, "step": 10882 }, { "epoch": 2.0, "learning_rate": 1.757838530964614e-05, "loss": 0.0854, "step": 10883 }, { "epoch": 2.0, "learning_rate": 1.7577900896117226e-05, "loss": 0.0989, "step": 10884 }, { "epoch": 2.0, "learning_rate": 1.757741644081874e-05, "loss": 0.0937, "step": 10885 }, { "epoch": 2.0, "learning_rate": 1.7576931943753362e-05, "loss": 0.0878, "step": 10886 }, { "epoch": 2.0, "learning_rate": 1.7576447404923757e-05, "loss": 0.0981, "step": 10887 }, { "epoch": 2.0, "learning_rate": 1.7575962824332595e-05, "loss": 0.0994, "step": 10888 }, { "epoch": 2.0, "learning_rate": 1.7575478201982546e-05, "loss": 0.092, "step": 10889 }, { "epoch": 2.0, "learning_rate": 1.7574993537876287e-05, "loss": 0.1001, "step": 10890 }, { "epoch": 2.0, "learning_rate": 1.7574508832016484e-05, "loss": 0.0754, "step": 10891 }, { "epoch": 2.0, "learning_rate": 1.7574024084405814e-05, "loss": 0.0925, "step": 10892 }, { "epoch": 2.0, "learning_rate": 1.7573539295046943e-05, "loss": 0.096, "step": 10893 }, { "epoch": 2.0, "learning_rate": 1.7573054463942548e-05, "loss": 0.0861, "step": 10894 }, { "epoch": 2.0, "learning_rate": 1.7572569591095295e-05, "loss": 0.1059, "step": 10895 }, { "epoch": 2.0, "learning_rate": 1.7572084676507865e-05, "loss": 0.0862, "step": 10896 }, { "epoch": 2.0, "learning_rate": 1.7571599720182926e-05, "loss": 0.0845, "step": 10897 }, { "epoch": 2.0, "learning_rate": 1.7571114722123154e-05, "loss": 0.0934, "step": 10898 }, { "epoch": 2.0, "learning_rate": 1.7570629682331216e-05, "loss": 0.0904, "step": 10899 }, { "epoch": 2.0, "learning_rate": 1.7570144600809794e-05, "loss": 0.0951, "step": 10900 }, { "epoch": 2.0, "learning_rate": 1.7569659477561557e-05, "loss": 0.0874, "step": 10901 }, { "epoch": 2.0, "learning_rate": 1.7569174312589178e-05, "loss": 0.0945, "step": 10902 }, { "epoch": 2.0, "learning_rate": 1.756868910589533e-05, "loss": 0.0879, "step": 10903 }, { "epoch": 2.0, "learning_rate": 1.7568203857482695e-05, "loss": 0.0985, "step": 10904 }, { "epoch": 2.0, "learning_rate": 1.756771856735394e-05, "loss": 0.0857, "step": 10905 }, { "epoch": 2.0, "learning_rate": 1.7567233235511745e-05, "loss": 0.082, "step": 10906 }, { "epoch": 2.0, "learning_rate": 1.7566747861958783e-05, "loss": 0.0863, "step": 10907 }, { "epoch": 2.0, "learning_rate": 1.7566262446697727e-05, "loss": 0.0804, "step": 10908 }, { "epoch": 2.0, "learning_rate": 1.756577698973126e-05, "loss": 0.0749, "step": 10909 }, { "epoch": 2.0, "learning_rate": 1.756529149106205e-05, "loss": 0.0904, "step": 10910 }, { "epoch": 2.0, "learning_rate": 1.7564805950692777e-05, "loss": 0.094, "step": 10911 }, { "epoch": 2.0, "learning_rate": 1.7564320368626115e-05, "loss": 0.0803, "step": 10912 }, { "epoch": 2.0, "learning_rate": 1.7563834744864744e-05, "loss": 0.0985, "step": 10913 }, { "epoch": 2.0, "learning_rate": 1.7563349079411337e-05, "loss": 0.0846, "step": 10914 }, { "epoch": 2.0, "learning_rate": 1.756286337226857e-05, "loss": 0.0804, "step": 10915 }, { "epoch": 2.0, "learning_rate": 1.756237762343913e-05, "loss": 0.0944, "step": 10916 }, { "epoch": 2.0, "learning_rate": 1.7561891832925682e-05, "loss": 0.0854, "step": 10917 }, { "epoch": 2.0, "learning_rate": 1.756140600073091e-05, "loss": 0.0847, "step": 10918 }, { "epoch": 2.0, "learning_rate": 1.756092012685749e-05, "loss": 0.0771, "step": 10919 }, { "epoch": 2.0, "learning_rate": 1.7560434211308105e-05, "loss": 0.0931, "step": 10920 }, { "epoch": 2.0, "learning_rate": 1.7559948254085424e-05, "loss": 0.0917, "step": 10921 }, { "epoch": 2.0, "learning_rate": 1.7559462255192134e-05, "loss": 0.091, "step": 10922 }, { "epoch": 2.0, "learning_rate": 1.7558976214630912e-05, "loss": 0.0856, "step": 10923 }, { "epoch": 2.0, "learning_rate": 1.7558490132404437e-05, "loss": 0.0775, "step": 10924 }, { "epoch": 2.0, "learning_rate": 1.7558004008515383e-05, "loss": 0.0777, "step": 10925 }, { "epoch": 2.0, "learning_rate": 1.755751784296644e-05, "loss": 0.0934, "step": 10926 }, { "epoch": 2.0, "learning_rate": 1.7557031635760277e-05, "loss": 0.0911, "step": 10927 }, { "epoch": 2.0, "learning_rate": 1.7556545386899578e-05, "loss": 0.0916, "step": 10928 }, { "epoch": 2.0, "learning_rate": 1.7556059096387024e-05, "loss": 0.1024, "step": 10929 }, { "epoch": 2.0, "learning_rate": 1.7555572764225297e-05, "loss": 0.0966, "step": 10930 }, { "epoch": 2.0, "learning_rate": 1.7555086390417075e-05, "loss": 0.0812, "step": 10931 }, { "epoch": 2.0, "learning_rate": 1.755459997496504e-05, "loss": 0.0826, "step": 10932 }, { "epoch": 2.0, "learning_rate": 1.7554113517871872e-05, "loss": 0.0902, "step": 10933 }, { "epoch": 2.01, "learning_rate": 1.7553627019140256e-05, "loss": 0.0792, "step": 10934 }, { "epoch": 2.01, "learning_rate": 1.755314047877287e-05, "loss": 0.0847, "step": 10935 }, { "epoch": 2.01, "learning_rate": 1.7552653896772396e-05, "loss": 0.0866, "step": 10936 }, { "epoch": 2.01, "learning_rate": 1.7552167273141516e-05, "loss": 0.0876, "step": 10937 }, { "epoch": 2.01, "learning_rate": 1.7551680607882914e-05, "loss": 0.086, "step": 10938 }, { "epoch": 2.01, "learning_rate": 1.7551193900999273e-05, "loss": 0.0823, "step": 10939 }, { "epoch": 2.01, "learning_rate": 1.7550707152493273e-05, "loss": 0.0876, "step": 10940 }, { "epoch": 2.01, "learning_rate": 1.7550220362367595e-05, "loss": 0.0861, "step": 10941 }, { "epoch": 2.01, "learning_rate": 1.7549733530624928e-05, "loss": 0.0906, "step": 10942 }, { "epoch": 2.01, "learning_rate": 1.7549246657267955e-05, "loss": 0.083, "step": 10943 }, { "epoch": 2.01, "learning_rate": 1.7548759742299354e-05, "loss": 0.0804, "step": 10944 }, { "epoch": 2.01, "learning_rate": 1.7548272785721815e-05, "loss": 0.0803, "step": 10945 }, { "epoch": 2.01, "learning_rate": 1.7547785787538018e-05, "loss": 0.0761, "step": 10946 }, { "epoch": 2.01, "learning_rate": 1.7547298747750652e-05, "loss": 0.0748, "step": 10947 }, { "epoch": 2.01, "learning_rate": 1.7546811666362398e-05, "loss": 0.0783, "step": 10948 }, { "epoch": 2.01, "learning_rate": 1.754632454337594e-05, "loss": 0.0875, "step": 10949 }, { "epoch": 2.01, "learning_rate": 1.7545837378793963e-05, "loss": 0.0892, "step": 10950 }, { "epoch": 2.01, "learning_rate": 1.7545350172619155e-05, "loss": 0.0861, "step": 10951 }, { "epoch": 2.01, "learning_rate": 1.75448629248542e-05, "loss": 0.07, "step": 10952 }, { "epoch": 2.01, "learning_rate": 1.754437563550178e-05, "loss": 0.0929, "step": 10953 }, { "epoch": 2.01, "learning_rate": 1.7543888304564588e-05, "loss": 0.0811, "step": 10954 }, { "epoch": 2.01, "learning_rate": 1.754340093204531e-05, "loss": 0.0714, "step": 10955 }, { "epoch": 2.01, "learning_rate": 1.7542913517946627e-05, "loss": 0.0876, "step": 10956 }, { "epoch": 2.01, "learning_rate": 1.7542426062271224e-05, "loss": 0.0787, "step": 10957 }, { "epoch": 2.01, "learning_rate": 1.7541938565021797e-05, "loss": 0.067, "step": 10958 }, { "epoch": 2.01, "learning_rate": 1.7541451026201027e-05, "loss": 0.0702, "step": 10959 }, { "epoch": 2.01, "learning_rate": 1.7540963445811603e-05, "loss": 0.0774, "step": 10960 }, { "epoch": 2.01, "learning_rate": 1.754047582385621e-05, "loss": 0.0832, "step": 10961 }, { "epoch": 2.01, "learning_rate": 1.753998816033754e-05, "loss": 0.0837, "step": 10962 }, { "epoch": 2.01, "learning_rate": 1.7539500455258275e-05, "loss": 0.0862, "step": 10963 }, { "epoch": 2.01, "learning_rate": 1.7539012708621105e-05, "loss": 0.0937, "step": 10964 }, { "epoch": 2.01, "learning_rate": 1.7538524920428725e-05, "loss": 0.089, "step": 10965 }, { "epoch": 2.01, "learning_rate": 1.753803709068382e-05, "loss": 0.0771, "step": 10966 }, { "epoch": 2.01, "learning_rate": 1.7537549219389075e-05, "loss": 0.0715, "step": 10967 }, { "epoch": 2.01, "learning_rate": 1.7537061306547182e-05, "loss": 0.0806, "step": 10968 }, { "epoch": 2.01, "learning_rate": 1.753657335216083e-05, "loss": 0.0824, "step": 10969 }, { "epoch": 2.01, "learning_rate": 1.753608535623271e-05, "loss": 0.0948, "step": 10970 }, { "epoch": 2.01, "learning_rate": 1.7535597318765512e-05, "loss": 0.0921, "step": 10971 }, { "epoch": 2.01, "learning_rate": 1.7535109239761924e-05, "loss": 0.0863, "step": 10972 }, { "epoch": 2.01, "learning_rate": 1.7534621119224638e-05, "loss": 0.0778, "step": 10973 }, { "epoch": 2.01, "learning_rate": 1.7534132957156344e-05, "loss": 0.0845, "step": 10974 }, { "epoch": 2.01, "learning_rate": 1.7533644753559732e-05, "loss": 0.0735, "step": 10975 }, { "epoch": 2.01, "learning_rate": 1.7533156508437493e-05, "loss": 0.0735, "step": 10976 }, { "epoch": 2.01, "learning_rate": 1.7532668221792323e-05, "loss": 0.0864, "step": 10977 }, { "epoch": 2.01, "learning_rate": 1.7532179893626905e-05, "loss": 0.084, "step": 10978 }, { "epoch": 2.01, "learning_rate": 1.7531691523943938e-05, "loss": 0.0801, "step": 10979 }, { "epoch": 2.01, "learning_rate": 1.753120311274611e-05, "loss": 0.0932, "step": 10980 }, { "epoch": 2.01, "learning_rate": 1.7530714660036112e-05, "loss": 0.0734, "step": 10981 }, { "epoch": 2.01, "learning_rate": 1.7530226165816638e-05, "loss": 0.0876, "step": 10982 }, { "epoch": 2.01, "learning_rate": 1.752973763009038e-05, "loss": 0.0895, "step": 10983 }, { "epoch": 2.01, "learning_rate": 1.752924905286004e-05, "loss": 0.0887, "step": 10984 }, { "epoch": 2.01, "learning_rate": 1.7528760434128293e-05, "loss": 0.0793, "step": 10985 }, { "epoch": 2.01, "learning_rate": 1.7528271773897846e-05, "loss": 0.087, "step": 10986 }, { "epoch": 2.01, "learning_rate": 1.7527783072171387e-05, "loss": 0.0808, "step": 10987 }, { "epoch": 2.02, "learning_rate": 1.7527294328951616e-05, "loss": 0.0833, "step": 10988 }, { "epoch": 2.02, "learning_rate": 1.7526805544241217e-05, "loss": 0.0926, "step": 10989 }, { "epoch": 2.02, "learning_rate": 1.7526316718042893e-05, "loss": 0.081, "step": 10990 }, { "epoch": 2.02, "learning_rate": 1.7525827850359335e-05, "loss": 0.0774, "step": 10991 }, { "epoch": 2.02, "learning_rate": 1.7525338941193236e-05, "loss": 0.0867, "step": 10992 }, { "epoch": 2.02, "learning_rate": 1.7524849990547294e-05, "loss": 0.0786, "step": 10993 }, { "epoch": 2.02, "learning_rate": 1.7524360998424198e-05, "loss": 0.0783, "step": 10994 }, { "epoch": 2.02, "learning_rate": 1.7523871964826652e-05, "loss": 0.0725, "step": 10995 }, { "epoch": 2.02, "learning_rate": 1.7523382889757346e-05, "loss": 0.092, "step": 10996 }, { "epoch": 2.02, "learning_rate": 1.7522893773218978e-05, "loss": 0.0971, "step": 10997 }, { "epoch": 2.02, "learning_rate": 1.7522404615214244e-05, "loss": 0.0802, "step": 10998 }, { "epoch": 2.02, "learning_rate": 1.7521915415745838e-05, "loss": 0.0934, "step": 10999 }, { "epoch": 2.02, "learning_rate": 1.752142617481646e-05, "loss": 0.0905, "step": 11000 }, { "epoch": 2.02, "learning_rate": 1.7520936892428803e-05, "loss": 0.0795, "step": 11001 }, { "epoch": 2.02, "learning_rate": 1.7520447568585565e-05, "loss": 0.0688, "step": 11002 }, { "epoch": 2.02, "learning_rate": 1.7519958203289446e-05, "loss": 0.0754, "step": 11003 }, { "epoch": 2.02, "learning_rate": 1.7519468796543143e-05, "loss": 0.0718, "step": 11004 }, { "epoch": 2.02, "learning_rate": 1.7518979348349346e-05, "loss": 0.0764, "step": 11005 }, { "epoch": 2.02, "learning_rate": 1.7518489858710762e-05, "loss": 0.0919, "step": 11006 }, { "epoch": 2.02, "learning_rate": 1.7518000327630088e-05, "loss": 0.0728, "step": 11007 }, { "epoch": 2.02, "learning_rate": 1.7517510755110015e-05, "loss": 0.0816, "step": 11008 }, { "epoch": 2.02, "learning_rate": 1.751702114115325e-05, "loss": 0.0871, "step": 11009 }, { "epoch": 2.02, "learning_rate": 1.7516531485762484e-05, "loss": 0.0773, "step": 11010 }, { "epoch": 2.02, "learning_rate": 1.7516041788940428e-05, "loss": 0.0912, "step": 11011 }, { "epoch": 2.02, "learning_rate": 1.751555205068977e-05, "loss": 0.084, "step": 11012 }, { "epoch": 2.02, "learning_rate": 1.751506227101321e-05, "loss": 0.0888, "step": 11013 }, { "epoch": 2.02, "learning_rate": 1.7514572449913453e-05, "loss": 0.0786, "step": 11014 }, { "epoch": 2.02, "learning_rate": 1.7514082587393195e-05, "loss": 0.0747, "step": 11015 }, { "epoch": 2.02, "learning_rate": 1.7513592683455143e-05, "loss": 0.0834, "step": 11016 }, { "epoch": 2.02, "learning_rate": 1.751310273810199e-05, "loss": 0.0856, "step": 11017 }, { "epoch": 2.02, "learning_rate": 1.7512612751336434e-05, "loss": 0.0891, "step": 11018 }, { "epoch": 2.02, "learning_rate": 1.7512122723161184e-05, "loss": 0.0907, "step": 11019 }, { "epoch": 2.02, "learning_rate": 1.7511632653578938e-05, "loss": 0.0836, "step": 11020 }, { "epoch": 2.02, "learning_rate": 1.7511142542592393e-05, "loss": 0.0845, "step": 11021 }, { "epoch": 2.02, "learning_rate": 1.751065239020426e-05, "loss": 0.0704, "step": 11022 }, { "epoch": 2.02, "learning_rate": 1.7510162196417233e-05, "loss": 0.0809, "step": 11023 }, { "epoch": 2.02, "learning_rate": 1.7509671961234015e-05, "loss": 0.0791, "step": 11024 }, { "epoch": 2.02, "learning_rate": 1.750918168465731e-05, "loss": 0.0837, "step": 11025 }, { "epoch": 2.02, "learning_rate": 1.750869136668982e-05, "loss": 0.0801, "step": 11026 }, { "epoch": 2.02, "learning_rate": 1.7508201007334244e-05, "loss": 0.089, "step": 11027 }, { "epoch": 2.02, "learning_rate": 1.750771060659329e-05, "loss": 0.0901, "step": 11028 }, { "epoch": 2.02, "learning_rate": 1.7507220164469663e-05, "loss": 0.0808, "step": 11029 }, { "epoch": 2.02, "learning_rate": 1.750672968096606e-05, "loss": 0.0806, "step": 11030 }, { "epoch": 2.02, "learning_rate": 1.7506239156085186e-05, "loss": 0.0769, "step": 11031 }, { "epoch": 2.02, "learning_rate": 1.750574858982975e-05, "loss": 0.0877, "step": 11032 }, { "epoch": 2.02, "learning_rate": 1.7505257982202447e-05, "loss": 0.0845, "step": 11033 }, { "epoch": 2.02, "learning_rate": 1.7504767333205985e-05, "loss": 0.093, "step": 11034 }, { "epoch": 2.02, "learning_rate": 1.7504276642843074e-05, "loss": 0.0749, "step": 11035 }, { "epoch": 2.02, "learning_rate": 1.7503785911116413e-05, "loss": 0.0818, "step": 11036 }, { "epoch": 2.02, "learning_rate": 1.750329513802871e-05, "loss": 0.0911, "step": 11037 }, { "epoch": 2.02, "learning_rate": 1.7502804323582665e-05, "loss": 0.0757, "step": 11038 }, { "epoch": 2.02, "learning_rate": 1.7502313467780988e-05, "loss": 0.0753, "step": 11039 }, { "epoch": 2.02, "learning_rate": 1.7501822570626384e-05, "loss": 0.0806, "step": 11040 }, { "epoch": 2.02, "learning_rate": 1.750133163212156e-05, "loss": 0.0862, "step": 11041 }, { "epoch": 2.02, "learning_rate": 1.7500840652269217e-05, "loss": 0.075, "step": 11042 }, { "epoch": 2.03, "learning_rate": 1.7500349631072065e-05, "loss": 0.0849, "step": 11043 }, { "epoch": 2.03, "learning_rate": 1.749985856853281e-05, "loss": 0.0839, "step": 11044 }, { "epoch": 2.03, "learning_rate": 1.749936746465416e-05, "loss": 0.0768, "step": 11045 }, { "epoch": 2.03, "learning_rate": 1.749887631943882e-05, "loss": 0.0829, "step": 11046 }, { "epoch": 2.03, "learning_rate": 1.7498385132889502e-05, "loss": 0.0749, "step": 11047 }, { "epoch": 2.03, "learning_rate": 1.74978939050089e-05, "loss": 0.0747, "step": 11048 }, { "epoch": 2.03, "learning_rate": 1.749740263579974e-05, "loss": 0.0932, "step": 11049 }, { "epoch": 2.03, "learning_rate": 1.749691132526472e-05, "loss": 0.0906, "step": 11050 }, { "epoch": 2.03, "learning_rate": 1.7496419973406546e-05, "loss": 0.0807, "step": 11051 }, { "epoch": 2.03, "learning_rate": 1.7495928580227933e-05, "loss": 0.0854, "step": 11052 }, { "epoch": 2.03, "learning_rate": 1.749543714573158e-05, "loss": 0.0806, "step": 11053 }, { "epoch": 2.03, "learning_rate": 1.749494566992021e-05, "loss": 0.0878, "step": 11054 }, { "epoch": 2.03, "learning_rate": 1.7494454152796518e-05, "loss": 0.0937, "step": 11055 }, { "epoch": 2.03, "learning_rate": 1.7493962594363224e-05, "loss": 0.076, "step": 11056 }, { "epoch": 2.03, "learning_rate": 1.7493470994623027e-05, "loss": 0.0839, "step": 11057 }, { "epoch": 2.03, "learning_rate": 1.7492979353578646e-05, "loss": 0.0896, "step": 11058 }, { "epoch": 2.03, "learning_rate": 1.7492487671232784e-05, "loss": 0.0818, "step": 11059 }, { "epoch": 2.03, "learning_rate": 1.749199594758816e-05, "loss": 0.0916, "step": 11060 }, { "epoch": 2.03, "learning_rate": 1.749150418264747e-05, "loss": 0.0859, "step": 11061 }, { "epoch": 2.03, "learning_rate": 1.749101237641344e-05, "loss": 0.0755, "step": 11062 }, { "epoch": 2.03, "learning_rate": 1.7490520528888778e-05, "loss": 0.0848, "step": 11063 }, { "epoch": 2.03, "learning_rate": 1.7490028640076185e-05, "loss": 0.0916, "step": 11064 }, { "epoch": 2.03, "learning_rate": 1.7489536709978382e-05, "loss": 0.0875, "step": 11065 }, { "epoch": 2.03, "learning_rate": 1.7489044738598075e-05, "loss": 0.0873, "step": 11066 }, { "epoch": 2.03, "learning_rate": 1.7488552725937976e-05, "loss": 0.0883, "step": 11067 }, { "epoch": 2.03, "learning_rate": 1.74880606720008e-05, "loss": 0.086, "step": 11068 }, { "epoch": 2.03, "learning_rate": 1.748756857678926e-05, "loss": 0.0765, "step": 11069 }, { "epoch": 2.03, "learning_rate": 1.7487076440306064e-05, "loss": 0.08, "step": 11070 }, { "epoch": 2.03, "learning_rate": 1.748658426255393e-05, "loss": 0.0929, "step": 11071 }, { "epoch": 2.03, "learning_rate": 1.748609204353557e-05, "loss": 0.0797, "step": 11072 }, { "epoch": 2.03, "learning_rate": 1.7485599783253692e-05, "loss": 0.1003, "step": 11073 }, { "epoch": 2.03, "learning_rate": 1.7485107481711014e-05, "loss": 0.0833, "step": 11074 }, { "epoch": 2.03, "learning_rate": 1.748461513891025e-05, "loss": 0.0862, "step": 11075 }, { "epoch": 2.03, "learning_rate": 1.7484122754854107e-05, "loss": 0.0712, "step": 11076 }, { "epoch": 2.03, "learning_rate": 1.7483630329545305e-05, "loss": 0.1, "step": 11077 }, { "epoch": 2.03, "learning_rate": 1.748313786298656e-05, "loss": 0.0721, "step": 11078 }, { "epoch": 2.03, "learning_rate": 1.748264535518058e-05, "loss": 0.0819, "step": 11079 }, { "epoch": 2.03, "learning_rate": 1.7482152806130088e-05, "loss": 0.0789, "step": 11080 }, { "epoch": 2.03, "learning_rate": 1.748166021583779e-05, "loss": 0.0748, "step": 11081 }, { "epoch": 2.03, "learning_rate": 1.748116758430641e-05, "loss": 0.0863, "step": 11082 }, { "epoch": 2.03, "learning_rate": 1.748067491153866e-05, "loss": 0.0868, "step": 11083 }, { "epoch": 2.03, "learning_rate": 1.7480182197537248e-05, "loss": 0.0883, "step": 11084 }, { "epoch": 2.03, "learning_rate": 1.7479689442304902e-05, "loss": 0.0893, "step": 11085 }, { "epoch": 2.03, "learning_rate": 1.747919664584433e-05, "loss": 0.0995, "step": 11086 }, { "epoch": 2.03, "learning_rate": 1.7478703808158252e-05, "loss": 0.0809, "step": 11087 }, { "epoch": 2.03, "learning_rate": 1.7478210929249383e-05, "loss": 0.089, "step": 11088 }, { "epoch": 2.03, "learning_rate": 1.747771800912044e-05, "loss": 0.0745, "step": 11089 }, { "epoch": 2.03, "learning_rate": 1.7477225047774143e-05, "loss": 0.0925, "step": 11090 }, { "epoch": 2.03, "learning_rate": 1.74767320452132e-05, "loss": 0.0934, "step": 11091 }, { "epoch": 2.03, "learning_rate": 1.747623900144034e-05, "loss": 0.0931, "step": 11092 }, { "epoch": 2.03, "learning_rate": 1.7475745916458274e-05, "loss": 0.0812, "step": 11093 }, { "epoch": 2.03, "learning_rate": 1.7475252790269723e-05, "loss": 0.0818, "step": 11094 }, { "epoch": 2.03, "learning_rate": 1.74747596228774e-05, "loss": 0.0878, "step": 11095 }, { "epoch": 2.03, "learning_rate": 1.7474266414284028e-05, "loss": 0.0803, "step": 11096 }, { "epoch": 2.04, "learning_rate": 1.7473773164492326e-05, "loss": 0.0883, "step": 11097 }, { "epoch": 2.04, "learning_rate": 1.747327987350501e-05, "loss": 0.0861, "step": 11098 }, { "epoch": 2.04, "learning_rate": 1.7472786541324802e-05, "loss": 0.0883, "step": 11099 }, { "epoch": 2.04, "learning_rate": 1.747229316795442e-05, "loss": 0.0831, "step": 11100 }, { "epoch": 2.04, "learning_rate": 1.747179975339658e-05, "loss": 0.0841, "step": 11101 }, { "epoch": 2.04, "learning_rate": 1.7471306297654005e-05, "loss": 0.0786, "step": 11102 }, { "epoch": 2.04, "learning_rate": 1.7470812800729415e-05, "loss": 0.0769, "step": 11103 }, { "epoch": 2.04, "learning_rate": 1.747031926262553e-05, "loss": 0.0798, "step": 11104 }, { "epoch": 2.04, "learning_rate": 1.7469825683345072e-05, "loss": 0.0816, "step": 11105 }, { "epoch": 2.04, "learning_rate": 1.746933206289076e-05, "loss": 0.0844, "step": 11106 }, { "epoch": 2.04, "learning_rate": 1.746883840126531e-05, "loss": 0.0914, "step": 11107 }, { "epoch": 2.04, "learning_rate": 1.746834469847145e-05, "loss": 0.0902, "step": 11108 }, { "epoch": 2.04, "learning_rate": 1.7467850954511903e-05, "loss": 0.0832, "step": 11109 }, { "epoch": 2.04, "learning_rate": 1.746735716938938e-05, "loss": 0.0813, "step": 11110 }, { "epoch": 2.04, "learning_rate": 1.7466863343106615e-05, "loss": 0.0899, "step": 11111 }, { "epoch": 2.04, "learning_rate": 1.746636947566632e-05, "loss": 0.0954, "step": 11112 }, { "epoch": 2.04, "learning_rate": 1.7465875567071228e-05, "loss": 0.0955, "step": 11113 }, { "epoch": 2.04, "learning_rate": 1.7465381617324048e-05, "loss": 0.0829, "step": 11114 }, { "epoch": 2.04, "learning_rate": 1.7464887626427513e-05, "loss": 0.0834, "step": 11115 }, { "epoch": 2.04, "learning_rate": 1.746439359438434e-05, "loss": 0.0887, "step": 11116 }, { "epoch": 2.04, "learning_rate": 1.746389952119726e-05, "loss": 0.0843, "step": 11117 }, { "epoch": 2.04, "learning_rate": 1.7463405406868986e-05, "loss": 0.0725, "step": 11118 }, { "epoch": 2.04, "learning_rate": 1.7462911251402247e-05, "loss": 0.0719, "step": 11119 }, { "epoch": 2.04, "learning_rate": 1.7462417054799766e-05, "loss": 0.0861, "step": 11120 }, { "epoch": 2.04, "learning_rate": 1.746192281706427e-05, "loss": 0.0787, "step": 11121 }, { "epoch": 2.04, "learning_rate": 1.746142853819848e-05, "loss": 0.0851, "step": 11122 }, { "epoch": 2.04, "learning_rate": 1.746093421820512e-05, "loss": 0.0813, "step": 11123 }, { "epoch": 2.04, "learning_rate": 1.7460439857086916e-05, "loss": 0.0813, "step": 11124 }, { "epoch": 2.04, "learning_rate": 1.7459945454846594e-05, "loss": 0.0869, "step": 11125 }, { "epoch": 2.04, "learning_rate": 1.7459451011486875e-05, "loss": 0.075, "step": 11126 }, { "epoch": 2.04, "learning_rate": 1.7458956527010493e-05, "loss": 0.0794, "step": 11127 }, { "epoch": 2.04, "learning_rate": 1.745846200142016e-05, "loss": 0.0884, "step": 11128 }, { "epoch": 2.04, "learning_rate": 1.7457967434718614e-05, "loss": 0.0805, "step": 11129 }, { "epoch": 2.04, "learning_rate": 1.7457472826908576e-05, "loss": 0.086, "step": 11130 }, { "epoch": 2.04, "learning_rate": 1.7456978177992775e-05, "loss": 0.0806, "step": 11131 }, { "epoch": 2.04, "learning_rate": 1.7456483487973934e-05, "loss": 0.0835, "step": 11132 }, { "epoch": 2.04, "learning_rate": 1.745598875685478e-05, "loss": 0.0849, "step": 11133 }, { "epoch": 2.04, "learning_rate": 1.7455493984638044e-05, "loss": 0.0802, "step": 11134 }, { "epoch": 2.04, "learning_rate": 1.745499917132645e-05, "loss": 0.0844, "step": 11135 }, { "epoch": 2.04, "learning_rate": 1.7454504316922725e-05, "loss": 0.0913, "step": 11136 }, { "epoch": 2.04, "learning_rate": 1.74540094214296e-05, "loss": 0.0754, "step": 11137 }, { "epoch": 2.04, "learning_rate": 1.7453514484849794e-05, "loss": 0.0787, "step": 11138 }, { "epoch": 2.04, "learning_rate": 1.745301950718605e-05, "loss": 0.0949, "step": 11139 }, { "epoch": 2.04, "learning_rate": 1.745252448844108e-05, "loss": 0.0841, "step": 11140 }, { "epoch": 2.04, "learning_rate": 1.7452029428617625e-05, "loss": 0.0926, "step": 11141 }, { "epoch": 2.04, "learning_rate": 1.745153432771841e-05, "loss": 0.0844, "step": 11142 }, { "epoch": 2.04, "learning_rate": 1.745103918574616e-05, "loss": 0.0786, "step": 11143 }, { "epoch": 2.04, "learning_rate": 1.745054400270361e-05, "loss": 0.077, "step": 11144 }, { "epoch": 2.04, "learning_rate": 1.7450048778593483e-05, "loss": 0.0901, "step": 11145 }, { "epoch": 2.04, "learning_rate": 1.744955351341852e-05, "loss": 0.0844, "step": 11146 }, { "epoch": 2.04, "learning_rate": 1.7449058207181434e-05, "loss": 0.0893, "step": 11147 }, { "epoch": 2.04, "learning_rate": 1.744856285988497e-05, "loss": 0.0774, "step": 11148 }, { "epoch": 2.04, "learning_rate": 1.7448067471531854e-05, "loss": 0.0978, "step": 11149 }, { "epoch": 2.04, "learning_rate": 1.7447572042124814e-05, "loss": 0.0832, "step": 11150 }, { "epoch": 2.04, "learning_rate": 1.744707657166658e-05, "loss": 0.0807, "step": 11151 }, { "epoch": 2.05, "learning_rate": 1.744658106015989e-05, "loss": 0.0789, "step": 11152 }, { "epoch": 2.05, "learning_rate": 1.744608550760747e-05, "loss": 0.0967, "step": 11153 }, { "epoch": 2.05, "learning_rate": 1.7445589914012048e-05, "loss": 0.0771, "step": 11154 }, { "epoch": 2.05, "learning_rate": 1.7445094279376366e-05, "loss": 0.079, "step": 11155 }, { "epoch": 2.05, "learning_rate": 1.7444598603703147e-05, "loss": 0.0843, "step": 11156 }, { "epoch": 2.05, "learning_rate": 1.7444102886995125e-05, "loss": 0.0869, "step": 11157 }, { "epoch": 2.05, "learning_rate": 1.7443607129255035e-05, "loss": 0.0812, "step": 11158 }, { "epoch": 2.05, "learning_rate": 1.7443111330485608e-05, "loss": 0.0764, "step": 11159 }, { "epoch": 2.05, "learning_rate": 1.7442615490689577e-05, "loss": 0.0863, "step": 11160 }, { "epoch": 2.05, "learning_rate": 1.7442119609869674e-05, "loss": 0.0893, "step": 11161 }, { "epoch": 2.05, "learning_rate": 1.7441623688028633e-05, "loss": 0.0856, "step": 11162 }, { "epoch": 2.05, "learning_rate": 1.7441127725169187e-05, "loss": 0.0831, "step": 11163 }, { "epoch": 2.05, "learning_rate": 1.7440631721294076e-05, "loss": 0.0879, "step": 11164 }, { "epoch": 2.05, "learning_rate": 1.7440135676406026e-05, "loss": 0.0737, "step": 11165 }, { "epoch": 2.05, "learning_rate": 1.743963959050777e-05, "loss": 0.0887, "step": 11166 }, { "epoch": 2.05, "learning_rate": 1.7439143463602052e-05, "loss": 0.0876, "step": 11167 }, { "epoch": 2.05, "learning_rate": 1.7438647295691597e-05, "loss": 0.0881, "step": 11168 }, { "epoch": 2.05, "learning_rate": 1.7438151086779143e-05, "loss": 0.0777, "step": 11169 }, { "epoch": 2.05, "learning_rate": 1.743765483686743e-05, "loss": 0.0787, "step": 11170 }, { "epoch": 2.05, "learning_rate": 1.7437158545959184e-05, "loss": 0.0875, "step": 11171 }, { "epoch": 2.05, "learning_rate": 1.7436662214057153e-05, "loss": 0.0814, "step": 11172 }, { "epoch": 2.05, "learning_rate": 1.743616584116406e-05, "loss": 0.0883, "step": 11173 }, { "epoch": 2.05, "learning_rate": 1.7435669427282648e-05, "loss": 0.0823, "step": 11174 }, { "epoch": 2.05, "learning_rate": 1.7435172972415655e-05, "loss": 0.0829, "step": 11175 }, { "epoch": 2.05, "learning_rate": 1.743467647656581e-05, "loss": 0.088, "step": 11176 }, { "epoch": 2.05, "learning_rate": 1.7434179939735857e-05, "loss": 0.0775, "step": 11177 }, { "epoch": 2.05, "learning_rate": 1.7433683361928527e-05, "loss": 0.089, "step": 11178 }, { "epoch": 2.05, "learning_rate": 1.743318674314656e-05, "loss": 0.0871, "step": 11179 }, { "epoch": 2.05, "learning_rate": 1.74326900833927e-05, "loss": 0.0691, "step": 11180 }, { "epoch": 2.05, "learning_rate": 1.743219338266967e-05, "loss": 0.0827, "step": 11181 }, { "epoch": 2.05, "learning_rate": 1.7431696640980223e-05, "loss": 0.0714, "step": 11182 }, { "epoch": 2.05, "learning_rate": 1.7431199858327085e-05, "loss": 0.0797, "step": 11183 }, { "epoch": 2.05, "learning_rate": 1.7430703034713e-05, "loss": 0.0845, "step": 11184 }, { "epoch": 2.05, "learning_rate": 1.7430206170140707e-05, "loss": 0.0861, "step": 11185 }, { "epoch": 2.05, "learning_rate": 1.7429709264612945e-05, "loss": 0.0832, "step": 11186 }, { "epoch": 2.05, "learning_rate": 1.742921231813245e-05, "loss": 0.0813, "step": 11187 }, { "epoch": 2.05, "learning_rate": 1.7428715330701963e-05, "loss": 0.0806, "step": 11188 }, { "epoch": 2.05, "learning_rate": 1.7428218302324222e-05, "loss": 0.0892, "step": 11189 }, { "epoch": 2.05, "learning_rate": 1.742772123300197e-05, "loss": 0.0942, "step": 11190 }, { "epoch": 2.05, "learning_rate": 1.742722412273794e-05, "loss": 0.0764, "step": 11191 }, { "epoch": 2.05, "learning_rate": 1.7426726971534883e-05, "loss": 0.0924, "step": 11192 }, { "epoch": 2.05, "learning_rate": 1.742622977939553e-05, "loss": 0.086, "step": 11193 }, { "epoch": 2.05, "learning_rate": 1.742573254632262e-05, "loss": 0.0761, "step": 11194 }, { "epoch": 2.05, "learning_rate": 1.74252352723189e-05, "loss": 0.0923, "step": 11195 }, { "epoch": 2.05, "learning_rate": 1.7424737957387114e-05, "loss": 0.0822, "step": 11196 }, { "epoch": 2.05, "learning_rate": 1.7424240601529993e-05, "loss": 0.0903, "step": 11197 }, { "epoch": 2.05, "learning_rate": 1.7423743204750288e-05, "loss": 0.0743, "step": 11198 }, { "epoch": 2.05, "learning_rate": 1.7423245767050736e-05, "loss": 0.0875, "step": 11199 }, { "epoch": 2.05, "learning_rate": 1.7422748288434078e-05, "loss": 0.0857, "step": 11200 }, { "epoch": 2.05, "learning_rate": 1.7422250768903056e-05, "loss": 0.1018, "step": 11201 }, { "epoch": 2.05, "learning_rate": 1.742175320846042e-05, "loss": 0.0877, "step": 11202 }, { "epoch": 2.05, "learning_rate": 1.74212556071089e-05, "loss": 0.087, "step": 11203 }, { "epoch": 2.05, "learning_rate": 1.742075796485125e-05, "loss": 0.0891, "step": 11204 }, { "epoch": 2.05, "learning_rate": 1.7420260281690206e-05, "loss": 0.0752, "step": 11205 }, { "epoch": 2.06, "learning_rate": 1.7419762557628513e-05, "loss": 0.0775, "step": 11206 }, { "epoch": 2.06, "learning_rate": 1.7419264792668915e-05, "loss": 0.088, "step": 11207 }, { "epoch": 2.06, "learning_rate": 1.7418766986814157e-05, "loss": 0.0713, "step": 11208 }, { "epoch": 2.06, "learning_rate": 1.741826914006698e-05, "loss": 0.0709, "step": 11209 }, { "epoch": 2.06, "learning_rate": 1.7417771252430128e-05, "loss": 0.0812, "step": 11210 }, { "epoch": 2.06, "learning_rate": 1.7417273323906347e-05, "loss": 0.0914, "step": 11211 }, { "epoch": 2.06, "learning_rate": 1.7416775354498387e-05, "loss": 0.0808, "step": 11212 }, { "epoch": 2.06, "learning_rate": 1.7416277344208984e-05, "loss": 0.0935, "step": 11213 }, { "epoch": 2.06, "learning_rate": 1.7415779293040887e-05, "loss": 0.0939, "step": 11214 }, { "epoch": 2.06, "learning_rate": 1.741528120099684e-05, "loss": 0.0951, "step": 11215 }, { "epoch": 2.06, "learning_rate": 1.741478306807959e-05, "loss": 0.0734, "step": 11216 }, { "epoch": 2.06, "learning_rate": 1.7414284894291883e-05, "loss": 0.0889, "step": 11217 }, { "epoch": 2.06, "learning_rate": 1.7413786679636463e-05, "loss": 0.0827, "step": 11218 }, { "epoch": 2.06, "learning_rate": 1.7413288424116077e-05, "loss": 0.0894, "step": 11219 }, { "epoch": 2.06, "learning_rate": 1.7412790127733473e-05, "loss": 0.0783, "step": 11220 }, { "epoch": 2.06, "learning_rate": 1.7412291790491392e-05, "loss": 0.0795, "step": 11221 }, { "epoch": 2.06, "learning_rate": 1.741179341239259e-05, "loss": 0.0766, "step": 11222 }, { "epoch": 2.06, "learning_rate": 1.7411294993439808e-05, "loss": 0.0839, "step": 11223 }, { "epoch": 2.06, "learning_rate": 1.7410796533635793e-05, "loss": 0.0745, "step": 11224 }, { "epoch": 2.06, "learning_rate": 1.7410298032983296e-05, "loss": 0.072, "step": 11225 }, { "epoch": 2.06, "learning_rate": 1.7409799491485058e-05, "loss": 0.076, "step": 11226 }, { "epoch": 2.06, "learning_rate": 1.7409300909143833e-05, "loss": 0.0762, "step": 11227 }, { "epoch": 2.06, "learning_rate": 1.740880228596237e-05, "loss": 0.0879, "step": 11228 }, { "epoch": 2.06, "learning_rate": 1.7408303621943417e-05, "loss": 0.0844, "step": 11229 }, { "epoch": 2.06, "learning_rate": 1.7407804917089718e-05, "loss": 0.0894, "step": 11230 }, { "epoch": 2.06, "learning_rate": 1.7407306171404024e-05, "loss": 0.0917, "step": 11231 }, { "epoch": 2.06, "learning_rate": 1.7406807384889088e-05, "loss": 0.0921, "step": 11232 }, { "epoch": 2.06, "learning_rate": 1.7406308557547652e-05, "loss": 0.0829, "step": 11233 }, { "epoch": 2.06, "learning_rate": 1.740580968938247e-05, "loss": 0.0914, "step": 11234 }, { "epoch": 2.06, "learning_rate": 1.7405310780396298e-05, "loss": 0.0884, "step": 11235 }, { "epoch": 2.06, "learning_rate": 1.740481183059187e-05, "loss": 0.0917, "step": 11236 }, { "epoch": 2.06, "learning_rate": 1.7404312839971953e-05, "loss": 0.0785, "step": 11237 }, { "epoch": 2.06, "learning_rate": 1.7403813808539287e-05, "loss": 0.0891, "step": 11238 }, { "epoch": 2.06, "learning_rate": 1.7403314736296625e-05, "loss": 0.0899, "step": 11239 }, { "epoch": 2.06, "learning_rate": 1.7402815623246724e-05, "loss": 0.082, "step": 11240 }, { "epoch": 2.06, "learning_rate": 1.7402316469392325e-05, "loss": 0.0908, "step": 11241 }, { "epoch": 2.06, "learning_rate": 1.7401817274736187e-05, "loss": 0.0849, "step": 11242 }, { "epoch": 2.06, "learning_rate": 1.7401318039281053e-05, "loss": 0.0811, "step": 11243 }, { "epoch": 2.06, "learning_rate": 1.7400818763029685e-05, "loss": 0.071, "step": 11244 }, { "epoch": 2.06, "learning_rate": 1.740031944598483e-05, "loss": 0.097, "step": 11245 }, { "epoch": 2.06, "learning_rate": 1.7399820088149245e-05, "loss": 0.0959, "step": 11246 }, { "epoch": 2.06, "learning_rate": 1.7399320689525672e-05, "loss": 0.0868, "step": 11247 }, { "epoch": 2.06, "learning_rate": 1.7398821250116874e-05, "loss": 0.0822, "step": 11248 }, { "epoch": 2.06, "learning_rate": 1.7398321769925597e-05, "loss": 0.0721, "step": 11249 }, { "epoch": 2.06, "learning_rate": 1.73978222489546e-05, "loss": 0.0818, "step": 11250 }, { "epoch": 2.06, "learning_rate": 1.739732268720663e-05, "loss": 0.0759, "step": 11251 }, { "epoch": 2.06, "learning_rate": 1.7396823084684445e-05, "loss": 0.0785, "step": 11252 }, { "epoch": 2.06, "learning_rate": 1.7396323441390798e-05, "loss": 0.086, "step": 11253 }, { "epoch": 2.06, "learning_rate": 1.7395823757328445e-05, "loss": 0.087, "step": 11254 }, { "epoch": 2.06, "learning_rate": 1.7395324032500137e-05, "loss": 0.0842, "step": 11255 }, { "epoch": 2.06, "learning_rate": 1.739482426690863e-05, "loss": 0.0794, "step": 11256 }, { "epoch": 2.06, "learning_rate": 1.739432446055668e-05, "loss": 0.0884, "step": 11257 }, { "epoch": 2.06, "learning_rate": 1.7393824613447036e-05, "loss": 0.0741, "step": 11258 }, { "epoch": 2.06, "learning_rate": 1.739332472558246e-05, "loss": 0.0826, "step": 11259 }, { "epoch": 2.06, "learning_rate": 1.7392824796965703e-05, "loss": 0.0724, "step": 11260 }, { "epoch": 2.07, "learning_rate": 1.7392324827599526e-05, "loss": 0.083, "step": 11261 }, { "epoch": 2.07, "learning_rate": 1.7391824817486684e-05, "loss": 0.0741, "step": 11262 }, { "epoch": 2.07, "learning_rate": 1.7391324766629923e-05, "loss": 0.0747, "step": 11263 }, { "epoch": 2.07, "learning_rate": 1.739082467503201e-05, "loss": 0.0951, "step": 11264 }, { "epoch": 2.07, "learning_rate": 1.7390324542695697e-05, "loss": 0.0941, "step": 11265 }, { "epoch": 2.07, "learning_rate": 1.7389824369623745e-05, "loss": 0.0787, "step": 11266 }, { "epoch": 2.07, "learning_rate": 1.7389324155818906e-05, "loss": 0.0958, "step": 11267 }, { "epoch": 2.07, "learning_rate": 1.738882390128394e-05, "loss": 0.0736, "step": 11268 }, { "epoch": 2.07, "learning_rate": 1.73883236060216e-05, "loss": 0.0911, "step": 11269 }, { "epoch": 2.07, "learning_rate": 1.738782327003465e-05, "loss": 0.0798, "step": 11270 }, { "epoch": 2.07, "learning_rate": 1.7387322893325845e-05, "loss": 0.092, "step": 11271 }, { "epoch": 2.07, "learning_rate": 1.7386822475897944e-05, "loss": 0.0826, "step": 11272 }, { "epoch": 2.07, "learning_rate": 1.7386322017753702e-05, "loss": 0.0787, "step": 11273 }, { "epoch": 2.07, "learning_rate": 1.7385821518895883e-05, "loss": 0.0865, "step": 11274 }, { "epoch": 2.07, "learning_rate": 1.738532097932724e-05, "loss": 0.0865, "step": 11275 }, { "epoch": 2.07, "learning_rate": 1.7384820399050535e-05, "loss": 0.0842, "step": 11276 }, { "epoch": 2.07, "learning_rate": 1.738431977806853e-05, "loss": 0.0832, "step": 11277 }, { "epoch": 2.07, "learning_rate": 1.738381911638398e-05, "loss": 0.0846, "step": 11278 }, { "epoch": 2.07, "learning_rate": 1.7383318413999644e-05, "loss": 0.0774, "step": 11279 }, { "epoch": 2.07, "learning_rate": 1.7382817670918286e-05, "loss": 0.0871, "step": 11280 }, { "epoch": 2.07, "learning_rate": 1.7382316887142666e-05, "loss": 0.0711, "step": 11281 }, { "epoch": 2.07, "learning_rate": 1.7381816062675536e-05, "loss": 0.0835, "step": 11282 }, { "epoch": 2.07, "learning_rate": 1.7381315197519667e-05, "loss": 0.0864, "step": 11283 }, { "epoch": 2.07, "learning_rate": 1.7380814291677818e-05, "loss": 0.088, "step": 11284 }, { "epoch": 2.07, "learning_rate": 1.7380313345152742e-05, "loss": 0.0781, "step": 11285 }, { "epoch": 2.07, "learning_rate": 1.737981235794721e-05, "loss": 0.0756, "step": 11286 }, { "epoch": 2.07, "learning_rate": 1.7379311330063977e-05, "loss": 0.0773, "step": 11287 }, { "epoch": 2.07, "learning_rate": 1.737881026150581e-05, "loss": 0.0906, "step": 11288 }, { "epoch": 2.07, "learning_rate": 1.7378309152275464e-05, "loss": 0.0754, "step": 11289 }, { "epoch": 2.07, "learning_rate": 1.737780800237571e-05, "loss": 0.0844, "step": 11290 }, { "epoch": 2.07, "learning_rate": 1.7377306811809306e-05, "loss": 0.0719, "step": 11291 }, { "epoch": 2.07, "learning_rate": 1.7376805580579008e-05, "loss": 0.0811, "step": 11292 }, { "epoch": 2.07, "learning_rate": 1.7376304308687587e-05, "loss": 0.0905, "step": 11293 }, { "epoch": 2.07, "learning_rate": 1.7375802996137806e-05, "loss": 0.0833, "step": 11294 }, { "epoch": 2.07, "learning_rate": 1.7375301642932422e-05, "loss": 0.0852, "step": 11295 }, { "epoch": 2.07, "learning_rate": 1.7374800249074207e-05, "loss": 0.0701, "step": 11296 }, { "epoch": 2.07, "learning_rate": 1.737429881456592e-05, "loss": 0.0833, "step": 11297 }, { "epoch": 2.07, "learning_rate": 1.737379733941032e-05, "loss": 0.0815, "step": 11298 }, { "epoch": 2.07, "learning_rate": 1.7373295823610184e-05, "loss": 0.0929, "step": 11299 }, { "epoch": 2.07, "learning_rate": 1.7372794267168264e-05, "loss": 0.0806, "step": 11300 }, { "epoch": 2.07, "learning_rate": 1.7372292670087332e-05, "loss": 0.0933, "step": 11301 }, { "epoch": 2.07, "learning_rate": 1.7371791032370147e-05, "loss": 0.0796, "step": 11302 }, { "epoch": 2.07, "learning_rate": 1.7371289354019478e-05, "loss": 0.0674, "step": 11303 }, { "epoch": 2.07, "learning_rate": 1.7370787635038088e-05, "loss": 0.0838, "step": 11304 }, { "epoch": 2.07, "learning_rate": 1.737028587542875e-05, "loss": 0.0828, "step": 11305 }, { "epoch": 2.07, "learning_rate": 1.7369784075194215e-05, "loss": 0.0729, "step": 11306 }, { "epoch": 2.07, "learning_rate": 1.7369282234337263e-05, "loss": 0.0781, "step": 11307 }, { "epoch": 2.07, "learning_rate": 1.7368780352860652e-05, "loss": 0.0978, "step": 11308 }, { "epoch": 2.07, "learning_rate": 1.7368278430767154e-05, "loss": 0.0905, "step": 11309 }, { "epoch": 2.07, "learning_rate": 1.736777646805953e-05, "loss": 0.0831, "step": 11310 }, { "epoch": 2.07, "learning_rate": 1.7367274464740552e-05, "loss": 0.0786, "step": 11311 }, { "epoch": 2.07, "learning_rate": 1.7366772420812983e-05, "loss": 0.0765, "step": 11312 }, { "epoch": 2.07, "learning_rate": 1.7366270336279593e-05, "loss": 0.0792, "step": 11313 }, { "epoch": 2.07, "learning_rate": 1.7365768211143143e-05, "loss": 0.0861, "step": 11314 }, { "epoch": 2.08, "learning_rate": 1.7365266045406412e-05, "loss": 0.0916, "step": 11315 }, { "epoch": 2.08, "learning_rate": 1.736476383907216e-05, "loss": 0.0729, "step": 11316 }, { "epoch": 2.08, "learning_rate": 1.7364261592143158e-05, "loss": 0.0744, "step": 11317 }, { "epoch": 2.08, "learning_rate": 1.7363759304622175e-05, "loss": 0.095, "step": 11318 }, { "epoch": 2.08, "learning_rate": 1.7363256976511972e-05, "loss": 0.0819, "step": 11319 }, { "epoch": 2.08, "learning_rate": 1.7362754607815332e-05, "loss": 0.0827, "step": 11320 }, { "epoch": 2.08, "learning_rate": 1.7362252198535012e-05, "loss": 0.0815, "step": 11321 }, { "epoch": 2.08, "learning_rate": 1.7361749748673785e-05, "loss": 0.0766, "step": 11322 }, { "epoch": 2.08, "learning_rate": 1.736124725823442e-05, "loss": 0.077, "step": 11323 }, { "epoch": 2.08, "learning_rate": 1.736074472721969e-05, "loss": 0.0848, "step": 11324 }, { "epoch": 2.08, "learning_rate": 1.7360242155632362e-05, "loss": 0.0885, "step": 11325 }, { "epoch": 2.08, "learning_rate": 1.7359739543475207e-05, "loss": 0.0843, "step": 11326 }, { "epoch": 2.08, "learning_rate": 1.7359236890750998e-05, "loss": 0.0834, "step": 11327 }, { "epoch": 2.08, "learning_rate": 1.7358734197462496e-05, "loss": 0.0863, "step": 11328 }, { "epoch": 2.08, "learning_rate": 1.7358231463612487e-05, "loss": 0.0839, "step": 11329 }, { "epoch": 2.08, "learning_rate": 1.7357728689203728e-05, "loss": 0.0797, "step": 11330 }, { "epoch": 2.08, "learning_rate": 1.7357225874238997e-05, "loss": 0.0891, "step": 11331 }, { "epoch": 2.08, "learning_rate": 1.7356723018721067e-05, "loss": 0.0924, "step": 11332 }, { "epoch": 2.08, "learning_rate": 1.7356220122652705e-05, "loss": 0.0764, "step": 11333 }, { "epoch": 2.08, "learning_rate": 1.7355717186036686e-05, "loss": 0.0765, "step": 11334 }, { "epoch": 2.08, "learning_rate": 1.7355214208875778e-05, "loss": 0.0771, "step": 11335 }, { "epoch": 2.08, "learning_rate": 1.7354711191172763e-05, "loss": 0.0821, "step": 11336 }, { "epoch": 2.08, "learning_rate": 1.7354208132930402e-05, "loss": 0.0906, "step": 11337 }, { "epoch": 2.08, "learning_rate": 1.7353705034151477e-05, "loss": 0.0802, "step": 11338 }, { "epoch": 2.08, "learning_rate": 1.735320189483876e-05, "loss": 0.0794, "step": 11339 }, { "epoch": 2.08, "learning_rate": 1.7352698714995017e-05, "loss": 0.0836, "step": 11340 }, { "epoch": 2.08, "learning_rate": 1.735219549462303e-05, "loss": 0.0965, "step": 11341 }, { "epoch": 2.08, "learning_rate": 1.7351692233725565e-05, "loss": 0.0928, "step": 11342 }, { "epoch": 2.08, "learning_rate": 1.7351188932305402e-05, "loss": 0.0897, "step": 11343 }, { "epoch": 2.08, "learning_rate": 1.7350685590365314e-05, "loss": 0.0874, "step": 11344 }, { "epoch": 2.08, "learning_rate": 1.735018220790808e-05, "loss": 0.0892, "step": 11345 }, { "epoch": 2.08, "learning_rate": 1.734967878493646e-05, "loss": 0.0832, "step": 11346 }, { "epoch": 2.08, "learning_rate": 1.7349175321453244e-05, "loss": 0.085, "step": 11347 }, { "epoch": 2.08, "learning_rate": 1.73486718174612e-05, "loss": 0.0819, "step": 11348 }, { "epoch": 2.08, "learning_rate": 1.7348168272963106e-05, "loss": 0.0819, "step": 11349 }, { "epoch": 2.08, "learning_rate": 1.7347664687961733e-05, "loss": 0.0966, "step": 11350 }, { "epoch": 2.08, "learning_rate": 1.7347161062459863e-05, "loss": 0.0782, "step": 11351 }, { "epoch": 2.08, "learning_rate": 1.734665739646027e-05, "loss": 0.0865, "step": 11352 }, { "epoch": 2.08, "learning_rate": 1.734615368996573e-05, "loss": 0.082, "step": 11353 }, { "epoch": 2.08, "learning_rate": 1.734564994297901e-05, "loss": 0.0741, "step": 11354 }, { "epoch": 2.08, "learning_rate": 1.7345146155502906e-05, "loss": 0.0801, "step": 11355 }, { "epoch": 2.08, "learning_rate": 1.734464232754018e-05, "loss": 0.0868, "step": 11356 }, { "epoch": 2.08, "learning_rate": 1.7344138459093612e-05, "loss": 0.0827, "step": 11357 }, { "epoch": 2.08, "learning_rate": 1.7343634550165986e-05, "loss": 0.0816, "step": 11358 }, { "epoch": 2.08, "learning_rate": 1.7343130600760068e-05, "loss": 0.0938, "step": 11359 }, { "epoch": 2.08, "learning_rate": 1.7342626610878647e-05, "loss": 0.0844, "step": 11360 }, { "epoch": 2.08, "learning_rate": 1.734212258052449e-05, "loss": 0.0949, "step": 11361 }, { "epoch": 2.08, "learning_rate": 1.7341618509700388e-05, "loss": 0.0784, "step": 11362 }, { "epoch": 2.08, "learning_rate": 1.734111439840911e-05, "loss": 0.0765, "step": 11363 }, { "epoch": 2.08, "learning_rate": 1.7340610246653438e-05, "loss": 0.0808, "step": 11364 }, { "epoch": 2.08, "learning_rate": 1.734010605443615e-05, "loss": 0.0733, "step": 11365 }, { "epoch": 2.08, "learning_rate": 1.7339601821760027e-05, "loss": 0.0863, "step": 11366 }, { "epoch": 2.08, "learning_rate": 1.7339097548627844e-05, "loss": 0.0827, "step": 11367 }, { "epoch": 2.08, "learning_rate": 1.7338593235042384e-05, "loss": 0.0702, "step": 11368 }, { "epoch": 2.08, "learning_rate": 1.733808888100643e-05, "loss": 0.0753, "step": 11369 }, { "epoch": 2.09, "learning_rate": 1.7337584486522753e-05, "loss": 0.0854, "step": 11370 }, { "epoch": 2.09, "learning_rate": 1.7337080051594138e-05, "loss": 0.085, "step": 11371 }, { "epoch": 2.09, "learning_rate": 1.733657557622337e-05, "loss": 0.0839, "step": 11372 }, { "epoch": 2.09, "learning_rate": 1.733607106041322e-05, "loss": 0.0848, "step": 11373 }, { "epoch": 2.09, "learning_rate": 1.7335566504166483e-05, "loss": 0.0849, "step": 11374 }, { "epoch": 2.09, "learning_rate": 1.7335061907485925e-05, "loss": 0.0904, "step": 11375 }, { "epoch": 2.09, "learning_rate": 1.7334557270374337e-05, "loss": 0.0804, "step": 11376 }, { "epoch": 2.09, "learning_rate": 1.7334052592834493e-05, "loss": 0.08, "step": 11377 }, { "epoch": 2.09, "learning_rate": 1.7333547874869182e-05, "loss": 0.0935, "step": 11378 }, { "epoch": 2.09, "learning_rate": 1.7333043116481185e-05, "loss": 0.0919, "step": 11379 }, { "epoch": 2.09, "learning_rate": 1.733253831767328e-05, "loss": 0.094, "step": 11380 }, { "epoch": 2.09, "learning_rate": 1.7332033478448257e-05, "loss": 0.0938, "step": 11381 }, { "epoch": 2.09, "learning_rate": 1.733152859880889e-05, "loss": 0.1031, "step": 11382 }, { "epoch": 2.09, "learning_rate": 1.7331023678757963e-05, "loss": 0.0829, "step": 11383 }, { "epoch": 2.09, "learning_rate": 1.7330518718298263e-05, "loss": 0.0891, "step": 11384 }, { "epoch": 2.09, "learning_rate": 1.7330013717432574e-05, "loss": 0.0815, "step": 11385 }, { "epoch": 2.09, "learning_rate": 1.732950867616368e-05, "loss": 0.0977, "step": 11386 }, { "epoch": 2.09, "learning_rate": 1.7329003594494358e-05, "loss": 0.0868, "step": 11387 }, { "epoch": 2.09, "learning_rate": 1.7328498472427398e-05, "loss": 0.0857, "step": 11388 }, { "epoch": 2.09, "learning_rate": 1.7327993309965583e-05, "loss": 0.082, "step": 11389 }, { "epoch": 2.09, "learning_rate": 1.7327488107111697e-05, "loss": 0.08, "step": 11390 }, { "epoch": 2.09, "learning_rate": 1.7326982863868526e-05, "loss": 0.0912, "step": 11391 }, { "epoch": 2.09, "learning_rate": 1.732647758023885e-05, "loss": 0.1026, "step": 11392 }, { "epoch": 2.09, "learning_rate": 1.7325972256225464e-05, "loss": 0.0803, "step": 11393 }, { "epoch": 2.09, "learning_rate": 1.7325466891831143e-05, "loss": 0.0843, "step": 11394 }, { "epoch": 2.09, "learning_rate": 1.7324961487058677e-05, "loss": 0.0824, "step": 11395 }, { "epoch": 2.09, "learning_rate": 1.732445604191085e-05, "loss": 0.0867, "step": 11396 }, { "epoch": 2.09, "learning_rate": 1.7323950556390453e-05, "loss": 0.0792, "step": 11397 }, { "epoch": 2.09, "learning_rate": 1.732344503050027e-05, "loss": 0.0824, "step": 11398 }, { "epoch": 2.09, "learning_rate": 1.732293946424308e-05, "loss": 0.0897, "step": 11399 }, { "epoch": 2.09, "learning_rate": 1.7322433857621685e-05, "loss": 0.0818, "step": 11400 }, { "epoch": 2.09, "learning_rate": 1.732192821063886e-05, "loss": 0.086, "step": 11401 }, { "epoch": 2.09, "learning_rate": 1.7321422523297392e-05, "loss": 0.0794, "step": 11402 }, { "epoch": 2.09, "learning_rate": 1.7320916795600073e-05, "loss": 0.0827, "step": 11403 }, { "epoch": 2.09, "learning_rate": 1.7320411027549687e-05, "loss": 0.0863, "step": 11404 }, { "epoch": 2.09, "learning_rate": 1.731990521914903e-05, "loss": 0.0752, "step": 11405 }, { "epoch": 2.09, "learning_rate": 1.731939937040088e-05, "loss": 0.0771, "step": 11406 }, { "epoch": 2.09, "learning_rate": 1.731889348130803e-05, "loss": 0.0815, "step": 11407 }, { "epoch": 2.09, "learning_rate": 1.7318387551873266e-05, "loss": 0.0756, "step": 11408 }, { "epoch": 2.09, "learning_rate": 1.731788158209938e-05, "loss": 0.0875, "step": 11409 }, { "epoch": 2.09, "learning_rate": 1.7317375571989158e-05, "loss": 0.0756, "step": 11410 }, { "epoch": 2.09, "learning_rate": 1.731686952154539e-05, "loss": 0.094, "step": 11411 }, { "epoch": 2.09, "learning_rate": 1.7316363430770868e-05, "loss": 0.0719, "step": 11412 }, { "epoch": 2.09, "learning_rate": 1.731585729966838e-05, "loss": 0.0789, "step": 11413 }, { "epoch": 2.09, "learning_rate": 1.7315351128240713e-05, "loss": 0.0819, "step": 11414 }, { "epoch": 2.09, "learning_rate": 1.731484491649066e-05, "loss": 0.0832, "step": 11415 }, { "epoch": 2.09, "learning_rate": 1.731433866442101e-05, "loss": 0.074, "step": 11416 }, { "epoch": 2.09, "learning_rate": 1.7313832372034554e-05, "loss": 0.0748, "step": 11417 }, { "epoch": 2.09, "learning_rate": 1.7313326039334083e-05, "loss": 0.0915, "step": 11418 }, { "epoch": 2.09, "learning_rate": 1.7312819666322388e-05, "loss": 0.0881, "step": 11419 }, { "epoch": 2.09, "learning_rate": 1.731231325300226e-05, "loss": 0.0814, "step": 11420 }, { "epoch": 2.09, "learning_rate": 1.7311806799376487e-05, "loss": 0.08, "step": 11421 }, { "epoch": 2.09, "learning_rate": 1.7311300305447868e-05, "loss": 0.0838, "step": 11422 }, { "epoch": 2.09, "learning_rate": 1.7310793771219187e-05, "loss": 0.0832, "step": 11423 }, { "epoch": 2.09, "learning_rate": 1.731028719669324e-05, "loss": 0.0786, "step": 11424 }, { "epoch": 2.1, "learning_rate": 1.730978058187282e-05, "loss": 0.0802, "step": 11425 }, { "epoch": 2.1, "learning_rate": 1.7309273926760716e-05, "loss": 0.0841, "step": 11426 }, { "epoch": 2.1, "learning_rate": 1.7308767231359723e-05, "loss": 0.076, "step": 11427 }, { "epoch": 2.1, "learning_rate": 1.7308260495672638e-05, "loss": 0.0664, "step": 11428 }, { "epoch": 2.1, "learning_rate": 1.7307753719702244e-05, "loss": 0.0954, "step": 11429 }, { "epoch": 2.1, "learning_rate": 1.7307246903451344e-05, "loss": 0.0685, "step": 11430 }, { "epoch": 2.1, "learning_rate": 1.730674004692273e-05, "loss": 0.0736, "step": 11431 }, { "epoch": 2.1, "learning_rate": 1.730623315011919e-05, "loss": 0.0893, "step": 11432 }, { "epoch": 2.1, "learning_rate": 1.730572621304352e-05, "loss": 0.081, "step": 11433 }, { "epoch": 2.1, "learning_rate": 1.7305219235698522e-05, "loss": 0.0816, "step": 11434 }, { "epoch": 2.1, "learning_rate": 1.730471221808698e-05, "loss": 0.0784, "step": 11435 }, { "epoch": 2.1, "learning_rate": 1.7304205160211694e-05, "loss": 0.0796, "step": 11436 }, { "epoch": 2.1, "learning_rate": 1.730369806207546e-05, "loss": 0.0911, "step": 11437 }, { "epoch": 2.1, "learning_rate": 1.7303190923681067e-05, "loss": 0.083, "step": 11438 }, { "epoch": 2.1, "learning_rate": 1.730268374503132e-05, "loss": 0.0719, "step": 11439 }, { "epoch": 2.1, "learning_rate": 1.7302176526129006e-05, "loss": 0.0942, "step": 11440 }, { "epoch": 2.1, "learning_rate": 1.730166926697692e-05, "loss": 0.0917, "step": 11441 }, { "epoch": 2.1, "learning_rate": 1.7301161967577865e-05, "loss": 0.0765, "step": 11442 }, { "epoch": 2.1, "learning_rate": 1.7300654627934637e-05, "loss": 0.0827, "step": 11443 }, { "epoch": 2.1, "learning_rate": 1.7300147248050027e-05, "loss": 0.073, "step": 11444 }, { "epoch": 2.1, "learning_rate": 1.7299639827926835e-05, "loss": 0.0757, "step": 11445 }, { "epoch": 2.1, "learning_rate": 1.7299132367567856e-05, "loss": 0.0766, "step": 11446 }, { "epoch": 2.1, "learning_rate": 1.7298624866975894e-05, "loss": 0.0746, "step": 11447 }, { "epoch": 2.1, "learning_rate": 1.7298117326153732e-05, "loss": 0.0784, "step": 11448 }, { "epoch": 2.1, "learning_rate": 1.7297609745104184e-05, "loss": 0.0726, "step": 11449 }, { "epoch": 2.1, "learning_rate": 1.7297102123830037e-05, "loss": 0.0867, "step": 11450 }, { "epoch": 2.1, "learning_rate": 1.729659446233409e-05, "loss": 0.0655, "step": 11451 }, { "epoch": 2.1, "learning_rate": 1.7296086760619148e-05, "loss": 0.0835, "step": 11452 }, { "epoch": 2.1, "learning_rate": 1.7295579018688e-05, "loss": 0.0934, "step": 11453 }, { "epoch": 2.1, "learning_rate": 1.729507123654345e-05, "loss": 0.0726, "step": 11454 }, { "epoch": 2.1, "learning_rate": 1.72945634141883e-05, "loss": 0.0928, "step": 11455 }, { "epoch": 2.1, "learning_rate": 1.7294055551625343e-05, "loss": 0.0819, "step": 11456 }, { "epoch": 2.1, "learning_rate": 1.7293547648857383e-05, "loss": 0.0848, "step": 11457 }, { "epoch": 2.1, "learning_rate": 1.7293039705887214e-05, "loss": 0.0754, "step": 11458 }, { "epoch": 2.1, "learning_rate": 1.7292531722717644e-05, "loss": 0.0886, "step": 11459 }, { "epoch": 2.1, "learning_rate": 1.7292023699351465e-05, "loss": 0.0859, "step": 11460 }, { "epoch": 2.1, "learning_rate": 1.729151563579148e-05, "loss": 0.0891, "step": 11461 }, { "epoch": 2.1, "learning_rate": 1.7291007532040493e-05, "loss": 0.0888, "step": 11462 }, { "epoch": 2.1, "learning_rate": 1.72904993881013e-05, "loss": 0.0955, "step": 11463 }, { "epoch": 2.1, "learning_rate": 1.7289991203976702e-05, "loss": 0.0807, "step": 11464 }, { "epoch": 2.1, "learning_rate": 1.7289482979669503e-05, "loss": 0.0745, "step": 11465 }, { "epoch": 2.1, "learning_rate": 1.7288974715182503e-05, "loss": 0.0899, "step": 11466 }, { "epoch": 2.1, "learning_rate": 1.7288466410518505e-05, "loss": 0.0889, "step": 11467 }, { "epoch": 2.1, "learning_rate": 1.7287958065680307e-05, "loss": 0.0798, "step": 11468 }, { "epoch": 2.1, "learning_rate": 1.7287449680670717e-05, "loss": 0.0814, "step": 11469 }, { "epoch": 2.1, "learning_rate": 1.728694125549253e-05, "loss": 0.0857, "step": 11470 }, { "epoch": 2.1, "learning_rate": 1.7286432790148554e-05, "loss": 0.0761, "step": 11471 }, { "epoch": 2.1, "learning_rate": 1.728592428464159e-05, "loss": 0.0836, "step": 11472 }, { "epoch": 2.1, "learning_rate": 1.7285415738974437e-05, "loss": 0.0789, "step": 11473 }, { "epoch": 2.1, "learning_rate": 1.7284907153149907e-05, "loss": 0.0749, "step": 11474 }, { "epoch": 2.1, "learning_rate": 1.7284398527170797e-05, "loss": 0.08, "step": 11475 }, { "epoch": 2.1, "learning_rate": 1.728388986103991e-05, "loss": 0.0903, "step": 11476 }, { "epoch": 2.1, "learning_rate": 1.728338115476005e-05, "loss": 0.079, "step": 11477 }, { "epoch": 2.1, "learning_rate": 1.7282872408334025e-05, "loss": 0.0857, "step": 11478 }, { "epoch": 2.11, "learning_rate": 1.7282363621764634e-05, "loss": 0.0828, "step": 11479 }, { "epoch": 2.11, "learning_rate": 1.7281854795054688e-05, "loss": 0.089, "step": 11480 }, { "epoch": 2.11, "learning_rate": 1.7281345928206988e-05, "loss": 0.0779, "step": 11481 }, { "epoch": 2.11, "learning_rate": 1.7280837021224335e-05, "loss": 0.0919, "step": 11482 }, { "epoch": 2.11, "learning_rate": 1.728032807410954e-05, "loss": 0.0979, "step": 11483 }, { "epoch": 2.11, "learning_rate": 1.7279819086865403e-05, "loss": 0.0827, "step": 11484 }, { "epoch": 2.11, "learning_rate": 1.7279310059494733e-05, "loss": 0.0832, "step": 11485 }, { "epoch": 2.11, "learning_rate": 1.727880099200034e-05, "loss": 0.0828, "step": 11486 }, { "epoch": 2.11, "learning_rate": 1.7278291884385022e-05, "loss": 0.0718, "step": 11487 }, { "epoch": 2.11, "learning_rate": 1.7277782736651588e-05, "loss": 0.09, "step": 11488 }, { "epoch": 2.11, "learning_rate": 1.7277273548802846e-05, "loss": 0.0838, "step": 11489 }, { "epoch": 2.11, "learning_rate": 1.7276764320841605e-05, "loss": 0.0829, "step": 11490 }, { "epoch": 2.11, "learning_rate": 1.7276255052770664e-05, "loss": 0.0875, "step": 11491 }, { "epoch": 2.11, "learning_rate": 1.7275745744592838e-05, "loss": 0.0704, "step": 11492 }, { "epoch": 2.11, "learning_rate": 1.727523639631093e-05, "loss": 0.0749, "step": 11493 }, { "epoch": 2.11, "learning_rate": 1.7274727007927747e-05, "loss": 0.0924, "step": 11494 }, { "epoch": 2.11, "learning_rate": 1.72742175794461e-05, "loss": 0.0777, "step": 11495 }, { "epoch": 2.11, "learning_rate": 1.7273708110868795e-05, "loss": 0.0751, "step": 11496 }, { "epoch": 2.11, "learning_rate": 1.727319860219864e-05, "loss": 0.0732, "step": 11497 }, { "epoch": 2.11, "learning_rate": 1.7272689053438443e-05, "loss": 0.0873, "step": 11498 }, { "epoch": 2.11, "learning_rate": 1.7272179464591015e-05, "loss": 0.0771, "step": 11499 }, { "epoch": 2.11, "learning_rate": 1.7271669835659163e-05, "loss": 0.0792, "step": 11500 }, { "epoch": 2.11, "learning_rate": 1.7271160166645695e-05, "loss": 0.0814, "step": 11501 }, { "epoch": 2.11, "learning_rate": 1.7270650457553422e-05, "loss": 0.0838, "step": 11502 }, { "epoch": 2.11, "learning_rate": 1.727014070838516e-05, "loss": 0.0807, "step": 11503 }, { "epoch": 2.11, "learning_rate": 1.72696309191437e-05, "loss": 0.0997, "step": 11504 }, { "epoch": 2.11, "learning_rate": 1.7269121089831874e-05, "loss": 0.0792, "step": 11505 }, { "epoch": 2.11, "learning_rate": 1.7268611220452477e-05, "loss": 0.0891, "step": 11506 }, { "epoch": 2.11, "learning_rate": 1.726810131100833e-05, "loss": 0.077, "step": 11507 }, { "epoch": 2.11, "learning_rate": 1.7267591361502233e-05, "loss": 0.0822, "step": 11508 }, { "epoch": 2.11, "learning_rate": 1.7267081371937006e-05, "loss": 0.0821, "step": 11509 }, { "epoch": 2.11, "learning_rate": 1.7266571342315452e-05, "loss": 0.0845, "step": 11510 }, { "epoch": 2.11, "learning_rate": 1.7266061272640387e-05, "loss": 0.0886, "step": 11511 }, { "epoch": 2.11, "learning_rate": 1.7265551162914626e-05, "loss": 0.0738, "step": 11512 }, { "epoch": 2.11, "learning_rate": 1.7265041013140975e-05, "loss": 0.074, "step": 11513 }, { "epoch": 2.11, "learning_rate": 1.726453082332225e-05, "loss": 0.0898, "step": 11514 }, { "epoch": 2.11, "learning_rate": 1.726402059346126e-05, "loss": 0.0807, "step": 11515 }, { "epoch": 2.11, "learning_rate": 1.726351032356081e-05, "loss": 0.0939, "step": 11516 }, { "epoch": 2.11, "learning_rate": 1.7263000013623733e-05, "loss": 0.0978, "step": 11517 }, { "epoch": 2.11, "learning_rate": 1.726248966365282e-05, "loss": 0.0891, "step": 11518 }, { "epoch": 2.11, "learning_rate": 1.7261979273650898e-05, "loss": 0.0893, "step": 11519 }, { "epoch": 2.11, "learning_rate": 1.726146884362078e-05, "loss": 0.069, "step": 11520 }, { "epoch": 2.11, "learning_rate": 1.7260958373565272e-05, "loss": 0.0904, "step": 11521 }, { "epoch": 2.11, "learning_rate": 1.726044786348719e-05, "loss": 0.0729, "step": 11522 }, { "epoch": 2.11, "learning_rate": 1.725993731338935e-05, "loss": 0.086, "step": 11523 }, { "epoch": 2.11, "learning_rate": 1.725942672327457e-05, "loss": 0.0788, "step": 11524 }, { "epoch": 2.11, "learning_rate": 1.7258916093145654e-05, "loss": 0.0757, "step": 11525 }, { "epoch": 2.11, "learning_rate": 1.7258405423005422e-05, "loss": 0.073, "step": 11526 }, { "epoch": 2.11, "learning_rate": 1.725789471285669e-05, "loss": 0.0824, "step": 11527 }, { "epoch": 2.11, "learning_rate": 1.7257383962702278e-05, "loss": 0.0881, "step": 11528 }, { "epoch": 2.11, "learning_rate": 1.725687317254499e-05, "loss": 0.0869, "step": 11529 }, { "epoch": 2.11, "learning_rate": 1.7256362342387647e-05, "loss": 0.0754, "step": 11530 }, { "epoch": 2.11, "learning_rate": 1.7255851472233067e-05, "loss": 0.0851, "step": 11531 }, { "epoch": 2.11, "learning_rate": 1.725534056208406e-05, "loss": 0.0729, "step": 11532 }, { "epoch": 2.11, "learning_rate": 1.725482961194345e-05, "loss": 0.0826, "step": 11533 }, { "epoch": 2.12, "learning_rate": 1.725431862181405e-05, "loss": 0.0807, "step": 11534 }, { "epoch": 2.12, "learning_rate": 1.7253807591698673e-05, "loss": 0.0784, "step": 11535 }, { "epoch": 2.12, "learning_rate": 1.725329652160014e-05, "loss": 0.0871, "step": 11536 }, { "epoch": 2.12, "learning_rate": 1.7252785411521268e-05, "loss": 0.0841, "step": 11537 }, { "epoch": 2.12, "learning_rate": 1.725227426146487e-05, "loss": 0.0763, "step": 11538 }, { "epoch": 2.12, "learning_rate": 1.7251763071433767e-05, "loss": 0.084, "step": 11539 }, { "epoch": 2.12, "learning_rate": 1.7251251841430778e-05, "loss": 0.0734, "step": 11540 }, { "epoch": 2.12, "learning_rate": 1.7250740571458716e-05, "loss": 0.0837, "step": 11541 }, { "epoch": 2.12, "learning_rate": 1.72502292615204e-05, "loss": 0.0778, "step": 11542 }, { "epoch": 2.12, "learning_rate": 1.7249717911618656e-05, "loss": 0.083, "step": 11543 }, { "epoch": 2.12, "learning_rate": 1.7249206521756297e-05, "loss": 0.0802, "step": 11544 }, { "epoch": 2.12, "learning_rate": 1.724869509193614e-05, "loss": 0.0872, "step": 11545 }, { "epoch": 2.12, "learning_rate": 1.7248183622161003e-05, "loss": 0.0795, "step": 11546 }, { "epoch": 2.12, "learning_rate": 1.7247672112433714e-05, "loss": 0.0798, "step": 11547 }, { "epoch": 2.12, "learning_rate": 1.7247160562757083e-05, "loss": 0.0877, "step": 11548 }, { "epoch": 2.12, "learning_rate": 1.724664897313393e-05, "loss": 0.0722, "step": 11549 }, { "epoch": 2.12, "learning_rate": 1.7246137343567082e-05, "loss": 0.0987, "step": 11550 }, { "epoch": 2.12, "learning_rate": 1.7245625674059354e-05, "loss": 0.0749, "step": 11551 }, { "epoch": 2.12, "learning_rate": 1.724511396461357e-05, "loss": 0.0902, "step": 11552 }, { "epoch": 2.12, "learning_rate": 1.7244602215232545e-05, "loss": 0.0784, "step": 11553 }, { "epoch": 2.12, "learning_rate": 1.7244090425919105e-05, "loss": 0.0863, "step": 11554 }, { "epoch": 2.12, "learning_rate": 1.7243578596676068e-05, "loss": 0.0902, "step": 11555 }, { "epoch": 2.12, "learning_rate": 1.7243066727506256e-05, "loss": 0.0856, "step": 11556 }, { "epoch": 2.12, "learning_rate": 1.724255481841249e-05, "loss": 0.0788, "step": 11557 }, { "epoch": 2.12, "learning_rate": 1.7242042869397592e-05, "loss": 0.0744, "step": 11558 }, { "epoch": 2.12, "learning_rate": 1.7241530880464385e-05, "loss": 0.0682, "step": 11559 }, { "epoch": 2.12, "learning_rate": 1.7241018851615693e-05, "loss": 0.0873, "step": 11560 }, { "epoch": 2.12, "learning_rate": 1.724050678285433e-05, "loss": 0.085, "step": 11561 }, { "epoch": 2.12, "learning_rate": 1.7239994674183126e-05, "loss": 0.0798, "step": 11562 }, { "epoch": 2.12, "learning_rate": 1.7239482525604903e-05, "loss": 0.0877, "step": 11563 }, { "epoch": 2.12, "learning_rate": 1.7238970337122484e-05, "loss": 0.0988, "step": 11564 }, { "epoch": 2.12, "learning_rate": 1.7238458108738684e-05, "loss": 0.0791, "step": 11565 }, { "epoch": 2.12, "learning_rate": 1.723794584045634e-05, "loss": 0.0791, "step": 11566 }, { "epoch": 2.12, "learning_rate": 1.7237433532278268e-05, "loss": 0.087, "step": 11567 }, { "epoch": 2.12, "learning_rate": 1.7236921184207293e-05, "loss": 0.0868, "step": 11568 }, { "epoch": 2.12, "learning_rate": 1.7236408796246236e-05, "loss": 0.0798, "step": 11569 }, { "epoch": 2.12, "learning_rate": 1.7235896368397927e-05, "loss": 0.0836, "step": 11570 }, { "epoch": 2.12, "learning_rate": 1.7235383900665184e-05, "loss": 0.084, "step": 11571 }, { "epoch": 2.12, "learning_rate": 1.723487139305084e-05, "loss": 0.0899, "step": 11572 }, { "epoch": 2.12, "learning_rate": 1.723435884555771e-05, "loss": 0.0852, "step": 11573 }, { "epoch": 2.12, "learning_rate": 1.7233846258188632e-05, "loss": 0.0753, "step": 11574 }, { "epoch": 2.12, "learning_rate": 1.723333363094642e-05, "loss": 0.0801, "step": 11575 }, { "epoch": 2.12, "learning_rate": 1.7232820963833905e-05, "loss": 0.0901, "step": 11576 }, { "epoch": 2.12, "learning_rate": 1.7232308256853908e-05, "loss": 0.0799, "step": 11577 }, { "epoch": 2.12, "learning_rate": 1.723179551000926e-05, "loss": 0.0732, "step": 11578 }, { "epoch": 2.12, "learning_rate": 1.7231282723302785e-05, "loss": 0.091, "step": 11579 }, { "epoch": 2.12, "learning_rate": 1.7230769896737315e-05, "loss": 0.0846, "step": 11580 }, { "epoch": 2.12, "learning_rate": 1.7230257030315667e-05, "loss": 0.0855, "step": 11581 }, { "epoch": 2.12, "learning_rate": 1.7229744124040673e-05, "loss": 0.0819, "step": 11582 }, { "epoch": 2.12, "learning_rate": 1.722923117791516e-05, "loss": 0.0806, "step": 11583 }, { "epoch": 2.12, "learning_rate": 1.7228718191941958e-05, "loss": 0.0723, "step": 11584 }, { "epoch": 2.12, "learning_rate": 1.722820516612389e-05, "loss": 0.0819, "step": 11585 }, { "epoch": 2.12, "learning_rate": 1.7227692100463785e-05, "loss": 0.0882, "step": 11586 }, { "epoch": 2.12, "learning_rate": 1.7227178994964476e-05, "loss": 0.0757, "step": 11587 }, { "epoch": 2.13, "learning_rate": 1.7226665849628785e-05, "loss": 0.0874, "step": 11588 }, { "epoch": 2.13, "learning_rate": 1.7226152664459542e-05, "loss": 0.0828, "step": 11589 }, { "epoch": 2.13, "learning_rate": 1.7225639439459575e-05, "loss": 0.0778, "step": 11590 }, { "epoch": 2.13, "learning_rate": 1.7225126174631715e-05, "loss": 0.0795, "step": 11591 }, { "epoch": 2.13, "learning_rate": 1.722461286997879e-05, "loss": 0.0919, "step": 11592 }, { "epoch": 2.13, "learning_rate": 1.722409952550363e-05, "loss": 0.0871, "step": 11593 }, { "epoch": 2.13, "learning_rate": 1.7223586141209063e-05, "loss": 0.0842, "step": 11594 }, { "epoch": 2.13, "learning_rate": 1.722307271709792e-05, "loss": 0.0784, "step": 11595 }, { "epoch": 2.13, "learning_rate": 1.7222559253173033e-05, "loss": 0.0796, "step": 11596 }, { "epoch": 2.13, "learning_rate": 1.722204574943723e-05, "loss": 0.0839, "step": 11597 }, { "epoch": 2.13, "learning_rate": 1.7221532205893342e-05, "loss": 0.0748, "step": 11598 }, { "epoch": 2.13, "learning_rate": 1.7221018622544197e-05, "loss": 0.087, "step": 11599 }, { "epoch": 2.13, "learning_rate": 1.722050499939263e-05, "loss": 0.085, "step": 11600 }, { "epoch": 2.13, "learning_rate": 1.721999133644147e-05, "loss": 0.0863, "step": 11601 }, { "epoch": 2.13, "learning_rate": 1.7219477633693548e-05, "loss": 0.0841, "step": 11602 }, { "epoch": 2.13, "learning_rate": 1.7218963891151696e-05, "loss": 0.0714, "step": 11603 }, { "epoch": 2.13, "learning_rate": 1.7218450108818747e-05, "loss": 0.0806, "step": 11604 }, { "epoch": 2.13, "learning_rate": 1.721793628669753e-05, "loss": 0.0807, "step": 11605 }, { "epoch": 2.13, "learning_rate": 1.7217422424790878e-05, "loss": 0.0768, "step": 11606 }, { "epoch": 2.13, "learning_rate": 1.721690852310163e-05, "loss": 0.0855, "step": 11607 }, { "epoch": 2.13, "learning_rate": 1.721639458163261e-05, "loss": 0.0815, "step": 11608 }, { "epoch": 2.13, "learning_rate": 1.7215880600386654e-05, "loss": 0.0794, "step": 11609 }, { "epoch": 2.13, "learning_rate": 1.7215366579366594e-05, "loss": 0.0741, "step": 11610 }, { "epoch": 2.13, "learning_rate": 1.7214852518575263e-05, "loss": 0.0845, "step": 11611 }, { "epoch": 2.13, "learning_rate": 1.7214338418015502e-05, "loss": 0.0843, "step": 11612 }, { "epoch": 2.13, "learning_rate": 1.7213824277690133e-05, "loss": 0.0736, "step": 11613 }, { "epoch": 2.13, "learning_rate": 1.7213310097601998e-05, "loss": 0.084, "step": 11614 }, { "epoch": 2.13, "learning_rate": 1.721279587775393e-05, "loss": 0.0777, "step": 11615 }, { "epoch": 2.13, "learning_rate": 1.7212281618148757e-05, "loss": 0.0868, "step": 11616 }, { "epoch": 2.13, "learning_rate": 1.7211767318789323e-05, "loss": 0.0912, "step": 11617 }, { "epoch": 2.13, "learning_rate": 1.721125297967846e-05, "loss": 0.0822, "step": 11618 }, { "epoch": 2.13, "learning_rate": 1.7210738600818998e-05, "loss": 0.0879, "step": 11619 }, { "epoch": 2.13, "learning_rate": 1.7210224182213777e-05, "loss": 0.0797, "step": 11620 }, { "epoch": 2.13, "learning_rate": 1.7209709723865632e-05, "loss": 0.0835, "step": 11621 }, { "epoch": 2.13, "learning_rate": 1.72091952257774e-05, "loss": 0.0779, "step": 11622 }, { "epoch": 2.13, "learning_rate": 1.720868068795191e-05, "loss": 0.0777, "step": 11623 }, { "epoch": 2.13, "learning_rate": 1.7208166110392008e-05, "loss": 0.0804, "step": 11624 }, { "epoch": 2.13, "learning_rate": 1.7207651493100522e-05, "loss": 0.0872, "step": 11625 }, { "epoch": 2.13, "learning_rate": 1.7207136836080295e-05, "loss": 0.0748, "step": 11626 }, { "epoch": 2.13, "learning_rate": 1.7206622139334158e-05, "loss": 0.0831, "step": 11627 }, { "epoch": 2.13, "learning_rate": 1.720610740286495e-05, "loss": 0.0843, "step": 11628 }, { "epoch": 2.13, "learning_rate": 1.7205592626675515e-05, "loss": 0.0807, "step": 11629 }, { "epoch": 2.13, "learning_rate": 1.720507781076868e-05, "loss": 0.0885, "step": 11630 }, { "epoch": 2.13, "learning_rate": 1.7204562955147288e-05, "loss": 0.0831, "step": 11631 }, { "epoch": 2.13, "learning_rate": 1.7204048059814175e-05, "loss": 0.0879, "step": 11632 }, { "epoch": 2.13, "learning_rate": 1.7203533124772184e-05, "loss": 0.0799, "step": 11633 }, { "epoch": 2.13, "learning_rate": 1.720301815002415e-05, "loss": 0.0893, "step": 11634 }, { "epoch": 2.13, "learning_rate": 1.7202503135572907e-05, "loss": 0.0787, "step": 11635 }, { "epoch": 2.13, "learning_rate": 1.72019880814213e-05, "loss": 0.0832, "step": 11636 }, { "epoch": 2.13, "learning_rate": 1.7201472987572164e-05, "loss": 0.0849, "step": 11637 }, { "epoch": 2.13, "learning_rate": 1.7200957854028344e-05, "loss": 0.0724, "step": 11638 }, { "epoch": 2.13, "learning_rate": 1.7200442680792672e-05, "loss": 0.0823, "step": 11639 }, { "epoch": 2.13, "learning_rate": 1.7199927467867995e-05, "loss": 0.0858, "step": 11640 }, { "epoch": 2.13, "learning_rate": 1.7199412215257146e-05, "loss": 0.0751, "step": 11641 }, { "epoch": 2.13, "learning_rate": 1.7198896922962967e-05, "loss": 0.0824, "step": 11642 }, { "epoch": 2.14, "learning_rate": 1.7198381590988304e-05, "loss": 0.0863, "step": 11643 }, { "epoch": 2.14, "learning_rate": 1.719786621933599e-05, "loss": 0.0791, "step": 11644 }, { "epoch": 2.14, "learning_rate": 1.719735080800887e-05, "loss": 0.0766, "step": 11645 }, { "epoch": 2.14, "learning_rate": 1.719683535700978e-05, "loss": 0.0791, "step": 11646 }, { "epoch": 2.14, "learning_rate": 1.7196319866341574e-05, "loss": 0.0871, "step": 11647 }, { "epoch": 2.14, "learning_rate": 1.7195804336007077e-05, "loss": 0.082, "step": 11648 }, { "epoch": 2.14, "learning_rate": 1.7195288766009138e-05, "loss": 0.0848, "step": 11649 }, { "epoch": 2.14, "learning_rate": 1.71947731563506e-05, "loss": 0.0741, "step": 11650 }, { "epoch": 2.14, "learning_rate": 1.7194257507034303e-05, "loss": 0.0769, "step": 11651 }, { "epoch": 2.14, "learning_rate": 1.7193741818063096e-05, "loss": 0.0784, "step": 11652 }, { "epoch": 2.14, "learning_rate": 1.7193226089439806e-05, "loss": 0.0658, "step": 11653 }, { "epoch": 2.14, "learning_rate": 1.719271032116729e-05, "loss": 0.0775, "step": 11654 }, { "epoch": 2.14, "learning_rate": 1.7192194513248388e-05, "loss": 0.0782, "step": 11655 }, { "epoch": 2.14, "learning_rate": 1.719167866568594e-05, "loss": 0.0797, "step": 11656 }, { "epoch": 2.14, "learning_rate": 1.719116277848279e-05, "loss": 0.0858, "step": 11657 }, { "epoch": 2.14, "learning_rate": 1.7190646851641782e-05, "loss": 0.0736, "step": 11658 }, { "epoch": 2.14, "learning_rate": 1.719013088516576e-05, "loss": 0.0824, "step": 11659 }, { "epoch": 2.14, "learning_rate": 1.718961487905757e-05, "loss": 0.0864, "step": 11660 }, { "epoch": 2.14, "learning_rate": 1.7189098833320052e-05, "loss": 0.088, "step": 11661 }, { "epoch": 2.14, "learning_rate": 1.7188582747956057e-05, "loss": 0.0715, "step": 11662 }, { "epoch": 2.14, "learning_rate": 1.718806662296842e-05, "loss": 0.0778, "step": 11663 }, { "epoch": 2.14, "learning_rate": 1.7187550458359995e-05, "loss": 0.0796, "step": 11664 }, { "epoch": 2.14, "learning_rate": 1.718703425413362e-05, "loss": 0.0818, "step": 11665 }, { "epoch": 2.14, "learning_rate": 1.7186518010292146e-05, "loss": 0.0795, "step": 11666 }, { "epoch": 2.14, "learning_rate": 1.7186001726838414e-05, "loss": 0.0796, "step": 11667 }, { "epoch": 2.14, "learning_rate": 1.7185485403775275e-05, "loss": 0.0764, "step": 11668 }, { "epoch": 2.14, "learning_rate": 1.718496904110557e-05, "loss": 0.0776, "step": 11669 }, { "epoch": 2.14, "learning_rate": 1.718445263883215e-05, "loss": 0.0789, "step": 11670 }, { "epoch": 2.14, "learning_rate": 1.718393619695786e-05, "loss": 0.0833, "step": 11671 }, { "epoch": 2.14, "learning_rate": 1.7183419715485543e-05, "loss": 0.0833, "step": 11672 }, { "epoch": 2.14, "learning_rate": 1.718290319441805e-05, "loss": 0.0826, "step": 11673 }, { "epoch": 2.14, "learning_rate": 1.7182386633758225e-05, "loss": 0.0747, "step": 11674 }, { "epoch": 2.14, "learning_rate": 1.7181870033508915e-05, "loss": 0.0922, "step": 11675 }, { "epoch": 2.14, "learning_rate": 1.7181353393672968e-05, "loss": 0.0925, "step": 11676 }, { "epoch": 2.14, "learning_rate": 1.7180836714253236e-05, "loss": 0.0804, "step": 11677 }, { "epoch": 2.14, "learning_rate": 1.7180319995252563e-05, "loss": 0.0822, "step": 11678 }, { "epoch": 2.14, "learning_rate": 1.71798032366738e-05, "loss": 0.0914, "step": 11679 }, { "epoch": 2.14, "learning_rate": 1.717928643851979e-05, "loss": 0.0822, "step": 11680 }, { "epoch": 2.14, "learning_rate": 1.7178769600793386e-05, "loss": 0.0866, "step": 11681 }, { "epoch": 2.14, "learning_rate": 1.7178252723497434e-05, "loss": 0.0873, "step": 11682 }, { "epoch": 2.14, "learning_rate": 1.717773580663479e-05, "loss": 0.0787, "step": 11683 }, { "epoch": 2.14, "learning_rate": 1.7177218850208292e-05, "loss": 0.0808, "step": 11684 }, { "epoch": 2.14, "learning_rate": 1.71767018542208e-05, "loss": 0.0836, "step": 11685 }, { "epoch": 2.14, "learning_rate": 1.7176184818675158e-05, "loss": 0.0826, "step": 11686 }, { "epoch": 2.14, "learning_rate": 1.7175667743574217e-05, "loss": 0.082, "step": 11687 }, { "epoch": 2.14, "learning_rate": 1.7175150628920826e-05, "loss": 0.0739, "step": 11688 }, { "epoch": 2.14, "learning_rate": 1.717463347471784e-05, "loss": 0.0785, "step": 11689 }, { "epoch": 2.14, "learning_rate": 1.7174116280968102e-05, "loss": 0.0817, "step": 11690 }, { "epoch": 2.14, "learning_rate": 1.717359904767447e-05, "loss": 0.0756, "step": 11691 }, { "epoch": 2.14, "learning_rate": 1.7173081774839792e-05, "loss": 0.0822, "step": 11692 }, { "epoch": 2.14, "learning_rate": 1.7172564462466915e-05, "loss": 0.0862, "step": 11693 }, { "epoch": 2.14, "learning_rate": 1.71720471105587e-05, "loss": 0.0751, "step": 11694 }, { "epoch": 2.14, "learning_rate": 1.717152971911799e-05, "loss": 0.0714, "step": 11695 }, { "epoch": 2.14, "learning_rate": 1.717101228814764e-05, "loss": 0.0701, "step": 11696 }, { "epoch": 2.15, "learning_rate": 1.7170494817650504e-05, "loss": 0.0783, "step": 11697 }, { "epoch": 2.15, "learning_rate": 1.7169977307629434e-05, "loss": 0.0934, "step": 11698 }, { "epoch": 2.15, "learning_rate": 1.7169459758087278e-05, "loss": 0.0873, "step": 11699 }, { "epoch": 2.15, "learning_rate": 1.7168942169026894e-05, "loss": 0.0773, "step": 11700 }, { "epoch": 2.15, "learning_rate": 1.7168424540451133e-05, "loss": 0.0766, "step": 11701 }, { "epoch": 2.15, "learning_rate": 1.7167906872362846e-05, "loss": 0.0947, "step": 11702 }, { "epoch": 2.15, "learning_rate": 1.716738916476489e-05, "loss": 0.0829, "step": 11703 }, { "epoch": 2.15, "learning_rate": 1.7166871417660116e-05, "loss": 0.0717, "step": 11704 }, { "epoch": 2.15, "learning_rate": 1.716635363105138e-05, "loss": 0.0923, "step": 11705 }, { "epoch": 2.15, "learning_rate": 1.7165835804941532e-05, "loss": 0.0803, "step": 11706 }, { "epoch": 2.15, "learning_rate": 1.7165317939333433e-05, "loss": 0.0823, "step": 11707 }, { "epoch": 2.15, "learning_rate": 1.716480003422993e-05, "loss": 0.0756, "step": 11708 }, { "epoch": 2.15, "learning_rate": 1.7164282089633885e-05, "loss": 0.0965, "step": 11709 }, { "epoch": 2.15, "learning_rate": 1.7163764105548147e-05, "loss": 0.0672, "step": 11710 }, { "epoch": 2.15, "learning_rate": 1.7163246081975575e-05, "loss": 0.0818, "step": 11711 }, { "epoch": 2.15, "learning_rate": 1.716272801891902e-05, "loss": 0.0781, "step": 11712 }, { "epoch": 2.15, "learning_rate": 1.716220991638134e-05, "loss": 0.0908, "step": 11713 }, { "epoch": 2.15, "learning_rate": 1.7161691774365395e-05, "loss": 0.0973, "step": 11714 }, { "epoch": 2.15, "learning_rate": 1.7161173592874037e-05, "loss": 0.0746, "step": 11715 }, { "epoch": 2.15, "learning_rate": 1.716065537191012e-05, "loss": 0.084, "step": 11716 }, { "epoch": 2.15, "learning_rate": 1.7160137111476505e-05, "loss": 0.0856, "step": 11717 }, { "epoch": 2.15, "learning_rate": 1.7159618811576043e-05, "loss": 0.0937, "step": 11718 }, { "epoch": 2.15, "learning_rate": 1.7159100472211595e-05, "loss": 0.0865, "step": 11719 }, { "epoch": 2.15, "learning_rate": 1.715858209338602e-05, "loss": 0.087, "step": 11720 }, { "epoch": 2.15, "learning_rate": 1.7158063675102172e-05, "loss": 0.0827, "step": 11721 }, { "epoch": 2.15, "learning_rate": 1.715754521736291e-05, "loss": 0.0804, "step": 11722 }, { "epoch": 2.15, "learning_rate": 1.715702672017109e-05, "loss": 0.0712, "step": 11723 }, { "epoch": 2.15, "learning_rate": 1.7156508183529572e-05, "loss": 0.0841, "step": 11724 }, { "epoch": 2.15, "learning_rate": 1.715598960744121e-05, "loss": 0.0928, "step": 11725 }, { "epoch": 2.15, "learning_rate": 1.7155470991908873e-05, "loss": 0.0805, "step": 11726 }, { "epoch": 2.15, "learning_rate": 1.7154952336935405e-05, "loss": 0.0806, "step": 11727 }, { "epoch": 2.15, "learning_rate": 1.7154433642523676e-05, "loss": 0.0845, "step": 11728 }, { "epoch": 2.15, "learning_rate": 1.7153914908676536e-05, "loss": 0.0826, "step": 11729 }, { "epoch": 2.15, "learning_rate": 1.7153396135396854e-05, "loss": 0.0805, "step": 11730 }, { "epoch": 2.15, "learning_rate": 1.7152877322687484e-05, "loss": 0.0804, "step": 11731 }, { "epoch": 2.15, "learning_rate": 1.7152358470551285e-05, "loss": 0.0815, "step": 11732 }, { "epoch": 2.15, "learning_rate": 1.7151839578991123e-05, "loss": 0.0844, "step": 11733 }, { "epoch": 2.15, "learning_rate": 1.7151320648009852e-05, "loss": 0.0799, "step": 11734 }, { "epoch": 2.15, "learning_rate": 1.715080167761033e-05, "loss": 0.0885, "step": 11735 }, { "epoch": 2.15, "learning_rate": 1.7150282667795427e-05, "loss": 0.0813, "step": 11736 }, { "epoch": 2.15, "learning_rate": 1.7149763618567994e-05, "loss": 0.0701, "step": 11737 }, { "epoch": 2.15, "learning_rate": 1.71492445299309e-05, "loss": 0.0802, "step": 11738 }, { "epoch": 2.15, "learning_rate": 1.7148725401887002e-05, "loss": 0.0877, "step": 11739 }, { "epoch": 2.15, "learning_rate": 1.7148206234439156e-05, "loss": 0.0742, "step": 11740 }, { "epoch": 2.15, "learning_rate": 1.7147687027590238e-05, "loss": 0.0869, "step": 11741 }, { "epoch": 2.15, "learning_rate": 1.71471677813431e-05, "loss": 0.0836, "step": 11742 }, { "epoch": 2.15, "learning_rate": 1.7146648495700602e-05, "loss": 0.0868, "step": 11743 }, { "epoch": 2.15, "learning_rate": 1.7146129170665614e-05, "loss": 0.074, "step": 11744 }, { "epoch": 2.15, "learning_rate": 1.714560980624099e-05, "loss": 0.0928, "step": 11745 }, { "epoch": 2.15, "learning_rate": 1.7145090402429604e-05, "loss": 0.0922, "step": 11746 }, { "epoch": 2.15, "learning_rate": 1.7144570959234306e-05, "loss": 0.0853, "step": 11747 }, { "epoch": 2.15, "learning_rate": 1.714405147665797e-05, "loss": 0.0726, "step": 11748 }, { "epoch": 2.15, "learning_rate": 1.7143531954703453e-05, "loss": 0.0941, "step": 11749 }, { "epoch": 2.15, "learning_rate": 1.7143012393373618e-05, "loss": 0.0902, "step": 11750 }, { "epoch": 2.15, "learning_rate": 1.7142492792671337e-05, "loss": 0.0874, "step": 11751 }, { "epoch": 2.16, "learning_rate": 1.7141973152599463e-05, "loss": 0.07, "step": 11752 }, { "epoch": 2.16, "learning_rate": 1.7141453473160868e-05, "loss": 0.0788, "step": 11753 }, { "epoch": 2.16, "learning_rate": 1.7140933754358413e-05, "loss": 0.0882, "step": 11754 }, { "epoch": 2.16, "learning_rate": 1.7140413996194966e-05, "loss": 0.0783, "step": 11755 }, { "epoch": 2.16, "learning_rate": 1.713989419867339e-05, "loss": 0.079, "step": 11756 }, { "epoch": 2.16, "learning_rate": 1.7139374361796547e-05, "loss": 0.0782, "step": 11757 }, { "epoch": 2.16, "learning_rate": 1.713885448556731e-05, "loss": 0.0799, "step": 11758 }, { "epoch": 2.16, "learning_rate": 1.7138334569988533e-05, "loss": 0.0847, "step": 11759 }, { "epoch": 2.16, "learning_rate": 1.7137814615063094e-05, "loss": 0.0762, "step": 11760 }, { "epoch": 2.16, "learning_rate": 1.713729462079385e-05, "loss": 0.0917, "step": 11761 }, { "epoch": 2.16, "learning_rate": 1.7136774587183677e-05, "loss": 0.0945, "step": 11762 }, { "epoch": 2.16, "learning_rate": 1.7136254514235428e-05, "loss": 0.0896, "step": 11763 }, { "epoch": 2.16, "learning_rate": 1.713573440195198e-05, "loss": 0.1004, "step": 11764 }, { "epoch": 2.16, "learning_rate": 1.7135214250336196e-05, "loss": 0.0845, "step": 11765 }, { "epoch": 2.16, "learning_rate": 1.7134694059390942e-05, "loss": 0.0597, "step": 11766 }, { "epoch": 2.16, "learning_rate": 1.713417382911909e-05, "loss": 0.0897, "step": 11767 }, { "epoch": 2.16, "learning_rate": 1.713365355952351e-05, "loss": 0.1004, "step": 11768 }, { "epoch": 2.16, "learning_rate": 1.7133133250607058e-05, "loss": 0.0914, "step": 11769 }, { "epoch": 2.16, "learning_rate": 1.7132612902372607e-05, "loss": 0.0867, "step": 11770 }, { "epoch": 2.16, "learning_rate": 1.713209251482303e-05, "loss": 0.0769, "step": 11771 }, { "epoch": 2.16, "learning_rate": 1.713157208796119e-05, "loss": 0.0803, "step": 11772 }, { "epoch": 2.16, "learning_rate": 1.713105162178996e-05, "loss": 0.0729, "step": 11773 }, { "epoch": 2.16, "learning_rate": 1.7130531116312202e-05, "loss": 0.0855, "step": 11774 }, { "epoch": 2.16, "learning_rate": 1.7130010571530796e-05, "loss": 0.0807, "step": 11775 }, { "epoch": 2.16, "learning_rate": 1.7129489987448603e-05, "loss": 0.0737, "step": 11776 }, { "epoch": 2.16, "learning_rate": 1.712896936406849e-05, "loss": 0.0872, "step": 11777 }, { "epoch": 2.16, "learning_rate": 1.712844870139333e-05, "loss": 0.0768, "step": 11778 }, { "epoch": 2.16, "learning_rate": 1.7127927999426002e-05, "loss": 0.0865, "step": 11779 }, { "epoch": 2.16, "learning_rate": 1.712740725816936e-05, "loss": 0.0822, "step": 11780 }, { "epoch": 2.16, "learning_rate": 1.7126886477626285e-05, "loss": 0.08, "step": 11781 }, { "epoch": 2.16, "learning_rate": 1.7126365657799648e-05, "loss": 0.0752, "step": 11782 }, { "epoch": 2.16, "learning_rate": 1.712584479869231e-05, "loss": 0.0814, "step": 11783 }, { "epoch": 2.16, "learning_rate": 1.7125323900307152e-05, "loss": 0.0756, "step": 11784 }, { "epoch": 2.16, "learning_rate": 1.7124802962647044e-05, "loss": 0.0872, "step": 11785 }, { "epoch": 2.16, "learning_rate": 1.712428198571485e-05, "loss": 0.0882, "step": 11786 }, { "epoch": 2.16, "learning_rate": 1.712376096951345e-05, "loss": 0.0759, "step": 11787 }, { "epoch": 2.16, "learning_rate": 1.712323991404571e-05, "loss": 0.0718, "step": 11788 }, { "epoch": 2.16, "learning_rate": 1.712271881931451e-05, "loss": 0.0824, "step": 11789 }, { "epoch": 2.16, "learning_rate": 1.712219768532271e-05, "loss": 0.0806, "step": 11790 }, { "epoch": 2.16, "learning_rate": 1.71216765120732e-05, "loss": 0.0619, "step": 11791 }, { "epoch": 2.16, "learning_rate": 1.7121155299568833e-05, "loss": 0.0916, "step": 11792 }, { "epoch": 2.16, "learning_rate": 1.7120634047812492e-05, "loss": 0.0942, "step": 11793 }, { "epoch": 2.16, "learning_rate": 1.712011275680705e-05, "loss": 0.0748, "step": 11794 }, { "epoch": 2.16, "learning_rate": 1.711959142655538e-05, "loss": 0.0875, "step": 11795 }, { "epoch": 2.16, "learning_rate": 1.711907005706036e-05, "loss": 0.0812, "step": 11796 }, { "epoch": 2.16, "learning_rate": 1.7118548648324852e-05, "loss": 0.0791, "step": 11797 }, { "epoch": 2.16, "learning_rate": 1.711802720035174e-05, "loss": 0.0811, "step": 11798 }, { "epoch": 2.16, "learning_rate": 1.7117505713143896e-05, "loss": 0.0759, "step": 11799 }, { "epoch": 2.16, "learning_rate": 1.7116984186704197e-05, "loss": 0.0866, "step": 11800 }, { "epoch": 2.16, "learning_rate": 1.7116462621035504e-05, "loss": 0.0762, "step": 11801 }, { "epoch": 2.16, "learning_rate": 1.7115941016140713e-05, "loss": 0.0793, "step": 11802 }, { "epoch": 2.16, "learning_rate": 1.7115419372022682e-05, "loss": 0.0804, "step": 11803 }, { "epoch": 2.16, "learning_rate": 1.7114897688684295e-05, "loss": 0.0854, "step": 11804 }, { "epoch": 2.16, "learning_rate": 1.7114375966128427e-05, "loss": 0.085, "step": 11805 }, { "epoch": 2.17, "learning_rate": 1.711385420435795e-05, "loss": 0.0742, "step": 11806 }, { "epoch": 2.17, "learning_rate": 1.7113332403375743e-05, "loss": 0.078, "step": 11807 }, { "epoch": 2.17, "learning_rate": 1.711281056318468e-05, "loss": 0.076, "step": 11808 }, { "epoch": 2.17, "learning_rate": 1.7112288683787637e-05, "loss": 0.0887, "step": 11809 }, { "epoch": 2.17, "learning_rate": 1.7111766765187496e-05, "loss": 0.085, "step": 11810 }, { "epoch": 2.17, "learning_rate": 1.711124480738713e-05, "loss": 0.0867, "step": 11811 }, { "epoch": 2.17, "learning_rate": 1.7110722810389413e-05, "loss": 0.0881, "step": 11812 }, { "epoch": 2.17, "learning_rate": 1.7110200774197225e-05, "loss": 0.0719, "step": 11813 }, { "epoch": 2.17, "learning_rate": 1.7109678698813447e-05, "loss": 0.0857, "step": 11814 }, { "epoch": 2.17, "learning_rate": 1.710915658424095e-05, "loss": 0.0757, "step": 11815 }, { "epoch": 2.17, "learning_rate": 1.7108634430482615e-05, "loss": 0.0814, "step": 11816 }, { "epoch": 2.17, "learning_rate": 1.7108112237541323e-05, "loss": 0.085, "step": 11817 }, { "epoch": 2.17, "learning_rate": 1.710759000541995e-05, "loss": 0.0839, "step": 11818 }, { "epoch": 2.17, "learning_rate": 1.7107067734121376e-05, "loss": 0.0846, "step": 11819 }, { "epoch": 2.17, "learning_rate": 1.7106545423648475e-05, "loss": 0.0864, "step": 11820 }, { "epoch": 2.17, "learning_rate": 1.710602307400413e-05, "loss": 0.0841, "step": 11821 }, { "epoch": 2.17, "learning_rate": 1.710550068519122e-05, "loss": 0.0843, "step": 11822 }, { "epoch": 2.17, "learning_rate": 1.710497825721262e-05, "loss": 0.072, "step": 11823 }, { "epoch": 2.17, "learning_rate": 1.710445579007122e-05, "loss": 0.0812, "step": 11824 }, { "epoch": 2.17, "learning_rate": 1.7103933283769886e-05, "loss": 0.0791, "step": 11825 }, { "epoch": 2.17, "learning_rate": 1.710341073831151e-05, "loss": 0.0842, "step": 11826 }, { "epoch": 2.17, "learning_rate": 1.7102888153698966e-05, "loss": 0.0775, "step": 11827 }, { "epoch": 2.17, "learning_rate": 1.710236552993514e-05, "loss": 0.0743, "step": 11828 }, { "epoch": 2.17, "learning_rate": 1.7101842867022906e-05, "loss": 0.0787, "step": 11829 }, { "epoch": 2.17, "learning_rate": 1.710132016496514e-05, "loss": 0.0848, "step": 11830 }, { "epoch": 2.17, "learning_rate": 1.710079742376474e-05, "loss": 0.0818, "step": 11831 }, { "epoch": 2.17, "learning_rate": 1.710027464342458e-05, "loss": 0.0793, "step": 11832 }, { "epoch": 2.17, "learning_rate": 1.7099751823947536e-05, "loss": 0.0812, "step": 11833 }, { "epoch": 2.17, "learning_rate": 1.7099228965336494e-05, "loss": 0.0864, "step": 11834 }, { "epoch": 2.17, "learning_rate": 1.7098706067594336e-05, "loss": 0.086, "step": 11835 }, { "epoch": 2.17, "learning_rate": 1.7098183130723942e-05, "loss": 0.082, "step": 11836 }, { "epoch": 2.17, "learning_rate": 1.7097660154728197e-05, "loss": 0.0839, "step": 11837 }, { "epoch": 2.17, "learning_rate": 1.7097137139609983e-05, "loss": 0.0892, "step": 11838 }, { "epoch": 2.17, "learning_rate": 1.7096614085372186e-05, "loss": 0.0792, "step": 11839 }, { "epoch": 2.17, "learning_rate": 1.7096090992017683e-05, "loss": 0.0674, "step": 11840 }, { "epoch": 2.17, "learning_rate": 1.709556785954936e-05, "loss": 0.0834, "step": 11841 }, { "epoch": 2.17, "learning_rate": 1.70950446879701e-05, "loss": 0.0833, "step": 11842 }, { "epoch": 2.17, "learning_rate": 1.709452147728279e-05, "loss": 0.0867, "step": 11843 }, { "epoch": 2.17, "learning_rate": 1.7093998227490306e-05, "loss": 0.0721, "step": 11844 }, { "epoch": 2.17, "learning_rate": 1.7093474938595546e-05, "loss": 0.0885, "step": 11845 }, { "epoch": 2.17, "learning_rate": 1.709295161060138e-05, "loss": 0.0862, "step": 11846 }, { "epoch": 2.17, "learning_rate": 1.70924282435107e-05, "loss": 0.0688, "step": 11847 }, { "epoch": 2.17, "learning_rate": 1.709190483732639e-05, "loss": 0.0899, "step": 11848 }, { "epoch": 2.17, "learning_rate": 1.7091381392051333e-05, "loss": 0.0965, "step": 11849 }, { "epoch": 2.17, "learning_rate": 1.7090857907688416e-05, "loss": 0.0657, "step": 11850 }, { "epoch": 2.17, "learning_rate": 1.7090334384240526e-05, "loss": 0.0855, "step": 11851 }, { "epoch": 2.17, "learning_rate": 1.7089810821710543e-05, "loss": 0.0878, "step": 11852 }, { "epoch": 2.17, "learning_rate": 1.708928722010136e-05, "loss": 0.0729, "step": 11853 }, { "epoch": 2.17, "learning_rate": 1.708876357941586e-05, "loss": 0.0835, "step": 11854 }, { "epoch": 2.17, "learning_rate": 1.708823989965693e-05, "loss": 0.0725, "step": 11855 }, { "epoch": 2.17, "learning_rate": 1.7087716180827453e-05, "loss": 0.0794, "step": 11856 }, { "epoch": 2.17, "learning_rate": 1.7087192422930316e-05, "loss": 0.0757, "step": 11857 }, { "epoch": 2.17, "learning_rate": 1.7086668625968413e-05, "loss": 0.0785, "step": 11858 }, { "epoch": 2.17, "learning_rate": 1.7086144789944623e-05, "loss": 0.0669, "step": 11859 }, { "epoch": 2.17, "learning_rate": 1.7085620914861837e-05, "loss": 0.0749, "step": 11860 }, { "epoch": 2.18, "learning_rate": 1.7085097000722945e-05, "loss": 0.0804, "step": 11861 }, { "epoch": 2.18, "learning_rate": 1.7084573047530832e-05, "loss": 0.0839, "step": 11862 }, { "epoch": 2.18, "learning_rate": 1.7084049055288387e-05, "loss": 0.0904, "step": 11863 }, { "epoch": 2.18, "learning_rate": 1.7083525023998495e-05, "loss": 0.0873, "step": 11864 }, { "epoch": 2.18, "learning_rate": 1.708300095366405e-05, "loss": 0.0737, "step": 11865 }, { "epoch": 2.18, "learning_rate": 1.7082476844287933e-05, "loss": 0.09, "step": 11866 }, { "epoch": 2.18, "learning_rate": 1.7081952695873045e-05, "loss": 0.0728, "step": 11867 }, { "epoch": 2.18, "learning_rate": 1.7081428508422262e-05, "loss": 0.0836, "step": 11868 }, { "epoch": 2.18, "learning_rate": 1.7080904281938483e-05, "loss": 0.0966, "step": 11869 }, { "epoch": 2.18, "learning_rate": 1.708038001642459e-05, "loss": 0.0758, "step": 11870 }, { "epoch": 2.18, "learning_rate": 1.7079855711883476e-05, "loss": 0.0814, "step": 11871 }, { "epoch": 2.18, "learning_rate": 1.7079331368318036e-05, "loss": 0.0799, "step": 11872 }, { "epoch": 2.18, "learning_rate": 1.7078806985731153e-05, "loss": 0.0847, "step": 11873 }, { "epoch": 2.18, "learning_rate": 1.707828256412572e-05, "loss": 0.0673, "step": 11874 }, { "epoch": 2.18, "learning_rate": 1.707775810350463e-05, "loss": 0.0805, "step": 11875 }, { "epoch": 2.18, "learning_rate": 1.7077233603870767e-05, "loss": 0.0916, "step": 11876 }, { "epoch": 2.18, "learning_rate": 1.707670906522703e-05, "loss": 0.0733, "step": 11877 }, { "epoch": 2.18, "learning_rate": 1.707618448757631e-05, "loss": 0.0824, "step": 11878 }, { "epoch": 2.18, "learning_rate": 1.707565987092149e-05, "loss": 0.0874, "step": 11879 }, { "epoch": 2.18, "learning_rate": 1.7075135215265466e-05, "loss": 0.0705, "step": 11880 }, { "epoch": 2.18, "learning_rate": 1.7074610520611133e-05, "loss": 0.09, "step": 11881 }, { "epoch": 2.18, "learning_rate": 1.7074085786961383e-05, "loss": 0.0805, "step": 11882 }, { "epoch": 2.18, "learning_rate": 1.7073561014319104e-05, "loss": 0.0784, "step": 11883 }, { "epoch": 2.18, "learning_rate": 1.707303620268719e-05, "loss": 0.0819, "step": 11884 }, { "epoch": 2.18, "learning_rate": 1.7072511352068536e-05, "loss": 0.0751, "step": 11885 }, { "epoch": 2.18, "learning_rate": 1.707198646246604e-05, "loss": 0.0802, "step": 11886 }, { "epoch": 2.18, "learning_rate": 1.707146153388258e-05, "loss": 0.0884, "step": 11887 }, { "epoch": 2.18, "learning_rate": 1.7070936566321063e-05, "loss": 0.0846, "step": 11888 }, { "epoch": 2.18, "learning_rate": 1.7070411559784378e-05, "loss": 0.0785, "step": 11889 }, { "epoch": 2.18, "learning_rate": 1.7069886514275413e-05, "loss": 0.0894, "step": 11890 }, { "epoch": 2.18, "learning_rate": 1.7069361429797077e-05, "loss": 0.0867, "step": 11891 }, { "epoch": 2.18, "learning_rate": 1.706883630635225e-05, "loss": 0.0827, "step": 11892 }, { "epoch": 2.18, "learning_rate": 1.706831114394383e-05, "loss": 0.0851, "step": 11893 }, { "epoch": 2.18, "learning_rate": 1.7067785942574718e-05, "loss": 0.0858, "step": 11894 }, { "epoch": 2.18, "learning_rate": 1.7067260702247802e-05, "loss": 0.0803, "step": 11895 }, { "epoch": 2.18, "learning_rate": 1.7066735422965976e-05, "loss": 0.0804, "step": 11896 }, { "epoch": 2.18, "learning_rate": 1.706621010473214e-05, "loss": 0.0874, "step": 11897 }, { "epoch": 2.18, "learning_rate": 1.7065684747549193e-05, "loss": 0.0786, "step": 11898 }, { "epoch": 2.18, "learning_rate": 1.7065159351420023e-05, "loss": 0.0781, "step": 11899 }, { "epoch": 2.18, "learning_rate": 1.706463391634753e-05, "loss": 0.0839, "step": 11900 }, { "epoch": 2.18, "learning_rate": 1.706410844233461e-05, "loss": 0.0707, "step": 11901 }, { "epoch": 2.18, "learning_rate": 1.7063582929384158e-05, "loss": 0.0684, "step": 11902 }, { "epoch": 2.18, "learning_rate": 1.706305737749907e-05, "loss": 0.09, "step": 11903 }, { "epoch": 2.18, "learning_rate": 1.7062531786682246e-05, "loss": 0.0945, "step": 11904 }, { "epoch": 2.18, "learning_rate": 1.7062006156936584e-05, "loss": 0.0782, "step": 11905 }, { "epoch": 2.18, "learning_rate": 1.706148048826497e-05, "loss": 0.0763, "step": 11906 }, { "epoch": 2.18, "learning_rate": 1.7060954780670317e-05, "loss": 0.0922, "step": 11907 }, { "epoch": 2.18, "learning_rate": 1.7060429034155515e-05, "loss": 0.0684, "step": 11908 }, { "epoch": 2.18, "learning_rate": 1.7059903248723463e-05, "loss": 0.0814, "step": 11909 }, { "epoch": 2.18, "learning_rate": 1.7059377424377058e-05, "loss": 0.1057, "step": 11910 }, { "epoch": 2.18, "learning_rate": 1.7058851561119198e-05, "loss": 0.0779, "step": 11911 }, { "epoch": 2.18, "learning_rate": 1.7058325658952785e-05, "loss": 0.0839, "step": 11912 }, { "epoch": 2.18, "learning_rate": 1.7057799717880716e-05, "loss": 0.0896, "step": 11913 }, { "epoch": 2.18, "learning_rate": 1.7057273737905887e-05, "loss": 0.0732, "step": 11914 }, { "epoch": 2.19, "learning_rate": 1.7056747719031203e-05, "loss": 0.0771, "step": 11915 }, { "epoch": 2.19, "learning_rate": 1.705622166125956e-05, "loss": 0.079, "step": 11916 }, { "epoch": 2.19, "learning_rate": 1.7055695564593856e-05, "loss": 0.0764, "step": 11917 }, { "epoch": 2.19, "learning_rate": 1.705516942903699e-05, "loss": 0.0798, "step": 11918 }, { "epoch": 2.19, "learning_rate": 1.7054643254591868e-05, "loss": 0.0816, "step": 11919 }, { "epoch": 2.19, "learning_rate": 1.7054117041261387e-05, "loss": 0.0865, "step": 11920 }, { "epoch": 2.19, "learning_rate": 1.7053590789048448e-05, "loss": 0.0903, "step": 11921 }, { "epoch": 2.19, "learning_rate": 1.705306449795595e-05, "loss": 0.0799, "step": 11922 }, { "epoch": 2.19, "learning_rate": 1.70525381679868e-05, "loss": 0.0829, "step": 11923 }, { "epoch": 2.19, "learning_rate": 1.7052011799143885e-05, "loss": 0.0728, "step": 11924 }, { "epoch": 2.19, "learning_rate": 1.7051485391430124e-05, "loss": 0.078, "step": 11925 }, { "epoch": 2.19, "learning_rate": 1.705095894484841e-05, "loss": 0.0801, "step": 11926 }, { "epoch": 2.19, "learning_rate": 1.7050432459401644e-05, "loss": 0.0744, "step": 11927 }, { "epoch": 2.19, "learning_rate": 1.7049905935092726e-05, "loss": 0.0765, "step": 11928 }, { "epoch": 2.19, "learning_rate": 1.7049379371924563e-05, "loss": 0.0802, "step": 11929 }, { "epoch": 2.19, "learning_rate": 1.7048852769900054e-05, "loss": 0.0792, "step": 11930 }, { "epoch": 2.19, "learning_rate": 1.7048326129022107e-05, "loss": 0.0902, "step": 11931 }, { "epoch": 2.19, "learning_rate": 1.7047799449293618e-05, "loss": 0.072, "step": 11932 }, { "epoch": 2.19, "learning_rate": 1.7047272730717494e-05, "loss": 0.0886, "step": 11933 }, { "epoch": 2.19, "learning_rate": 1.7046745973296642e-05, "loss": 0.0825, "step": 11934 }, { "epoch": 2.19, "learning_rate": 1.7046219177033957e-05, "loss": 0.0798, "step": 11935 }, { "epoch": 2.19, "learning_rate": 1.7045692341932344e-05, "loss": 0.0822, "step": 11936 }, { "epoch": 2.19, "learning_rate": 1.7045165467994714e-05, "loss": 0.0942, "step": 11937 }, { "epoch": 2.19, "learning_rate": 1.7044638555223968e-05, "loss": 0.0866, "step": 11938 }, { "epoch": 2.19, "learning_rate": 1.704411160362301e-05, "loss": 0.0842, "step": 11939 }, { "epoch": 2.19, "learning_rate": 1.704358461319474e-05, "loss": 0.0832, "step": 11940 }, { "epoch": 2.19, "learning_rate": 1.7043057583942063e-05, "loss": 0.089, "step": 11941 }, { "epoch": 2.19, "learning_rate": 1.7042530515867897e-05, "loss": 0.0834, "step": 11942 }, { "epoch": 2.19, "learning_rate": 1.7042003408975132e-05, "loss": 0.0901, "step": 11943 }, { "epoch": 2.19, "learning_rate": 1.704147626326668e-05, "loss": 0.0845, "step": 11944 }, { "epoch": 2.19, "learning_rate": 1.7040949078745447e-05, "loss": 0.073, "step": 11945 }, { "epoch": 2.19, "learning_rate": 1.7040421855414336e-05, "loss": 0.09, "step": 11946 }, { "epoch": 2.19, "learning_rate": 1.7039894593276255e-05, "loss": 0.0893, "step": 11947 }, { "epoch": 2.19, "learning_rate": 1.7039367292334114e-05, "loss": 0.0799, "step": 11948 }, { "epoch": 2.19, "learning_rate": 1.703883995259081e-05, "loss": 0.075, "step": 11949 }, { "epoch": 2.19, "learning_rate": 1.7038312574049257e-05, "loss": 0.0917, "step": 11950 }, { "epoch": 2.19, "learning_rate": 1.7037785156712365e-05, "loss": 0.0776, "step": 11951 }, { "epoch": 2.19, "learning_rate": 1.703725770058303e-05, "loss": 0.081, "step": 11952 }, { "epoch": 2.19, "learning_rate": 1.7036730205664168e-05, "loss": 0.0803, "step": 11953 }, { "epoch": 2.19, "learning_rate": 1.703620267195868e-05, "loss": 0.0826, "step": 11954 }, { "epoch": 2.19, "learning_rate": 1.7035675099469483e-05, "loss": 0.0804, "step": 11955 }, { "epoch": 2.19, "learning_rate": 1.703514748819948e-05, "loss": 0.0877, "step": 11956 }, { "epoch": 2.19, "learning_rate": 1.703461983815158e-05, "loss": 0.0948, "step": 11957 }, { "epoch": 2.19, "learning_rate": 1.7034092149328688e-05, "loss": 0.0736, "step": 11958 }, { "epoch": 2.19, "learning_rate": 1.7033564421733717e-05, "loss": 0.0753, "step": 11959 }, { "epoch": 2.19, "learning_rate": 1.7033036655369574e-05, "loss": 0.0837, "step": 11960 }, { "epoch": 2.19, "learning_rate": 1.7032508850239165e-05, "loss": 0.0688, "step": 11961 }, { "epoch": 2.19, "learning_rate": 1.703198100634541e-05, "loss": 0.0947, "step": 11962 }, { "epoch": 2.19, "learning_rate": 1.7031453123691204e-05, "loss": 0.0794, "step": 11963 }, { "epoch": 2.19, "learning_rate": 1.7030925202279465e-05, "loss": 0.0793, "step": 11964 }, { "epoch": 2.19, "learning_rate": 1.7030397242113103e-05, "loss": 0.0894, "step": 11965 }, { "epoch": 2.19, "learning_rate": 1.7029869243195024e-05, "loss": 0.0859, "step": 11966 }, { "epoch": 2.19, "learning_rate": 1.7029341205528145e-05, "loss": 0.0719, "step": 11967 }, { "epoch": 2.19, "learning_rate": 1.702881312911537e-05, "loss": 0.0885, "step": 11968 }, { "epoch": 2.19, "learning_rate": 1.7028285013959614e-05, "loss": 0.0737, "step": 11969 }, { "epoch": 2.2, "learning_rate": 1.7027756860063783e-05, "loss": 0.0834, "step": 11970 }, { "epoch": 2.2, "learning_rate": 1.7027228667430797e-05, "loss": 0.0816, "step": 11971 }, { "epoch": 2.2, "learning_rate": 1.702670043606356e-05, "loss": 0.0803, "step": 11972 }, { "epoch": 2.2, "learning_rate": 1.7026172165964983e-05, "loss": 0.0833, "step": 11973 }, { "epoch": 2.2, "learning_rate": 1.7025643857137982e-05, "loss": 0.0797, "step": 11974 }, { "epoch": 2.2, "learning_rate": 1.702511550958547e-05, "loss": 0.0813, "step": 11975 }, { "epoch": 2.2, "learning_rate": 1.7024587123310352e-05, "loss": 0.0859, "step": 11976 }, { "epoch": 2.2, "learning_rate": 1.7024058698315548e-05, "loss": 0.083, "step": 11977 }, { "epoch": 2.2, "learning_rate": 1.7023530234603965e-05, "loss": 0.0807, "step": 11978 }, { "epoch": 2.2, "learning_rate": 1.7023001732178524e-05, "loss": 0.0838, "step": 11979 }, { "epoch": 2.2, "learning_rate": 1.702247319104213e-05, "loss": 0.0868, "step": 11980 }, { "epoch": 2.2, "learning_rate": 1.7021944611197695e-05, "loss": 0.0849, "step": 11981 }, { "epoch": 2.2, "learning_rate": 1.7021415992648143e-05, "loss": 0.0742, "step": 11982 }, { "epoch": 2.2, "learning_rate": 1.702088733539638e-05, "loss": 0.0781, "step": 11983 }, { "epoch": 2.2, "learning_rate": 1.702035863944532e-05, "loss": 0.0906, "step": 11984 }, { "epoch": 2.2, "learning_rate": 1.701982990479788e-05, "loss": 0.0758, "step": 11985 }, { "epoch": 2.2, "learning_rate": 1.7019301131456973e-05, "loss": 0.074, "step": 11986 }, { "epoch": 2.2, "learning_rate": 1.7018772319425517e-05, "loss": 0.076, "step": 11987 }, { "epoch": 2.2, "learning_rate": 1.7018243468706417e-05, "loss": 0.0809, "step": 11988 }, { "epoch": 2.2, "learning_rate": 1.7017714579302597e-05, "loss": 0.0831, "step": 11989 }, { "epoch": 2.2, "learning_rate": 1.701718565121697e-05, "loss": 0.0853, "step": 11990 }, { "epoch": 2.2, "learning_rate": 1.7016656684452453e-05, "loss": 0.0906, "step": 11991 }, { "epoch": 2.2, "learning_rate": 1.7016127679011956e-05, "loss": 0.0829, "step": 11992 }, { "epoch": 2.2, "learning_rate": 1.7015598634898406e-05, "loss": 0.0883, "step": 11993 }, { "epoch": 2.2, "learning_rate": 1.7015069552114707e-05, "loss": 0.0789, "step": 11994 }, { "epoch": 2.2, "learning_rate": 1.701454043066378e-05, "loss": 0.083, "step": 11995 }, { "epoch": 2.2, "learning_rate": 1.701401127054854e-05, "loss": 0.0835, "step": 11996 }, { "epoch": 2.2, "learning_rate": 1.701348207177191e-05, "loss": 0.0782, "step": 11997 }, { "epoch": 2.2, "learning_rate": 1.70129528343368e-05, "loss": 0.0728, "step": 11998 }, { "epoch": 2.2, "learning_rate": 1.701242355824613e-05, "loss": 0.0758, "step": 11999 }, { "epoch": 2.2, "learning_rate": 1.7011894243502822e-05, "loss": 0.0838, "step": 12000 }, { "epoch": 2.2, "learning_rate": 1.701136489010978e-05, "loss": 0.0792, "step": 12001 }, { "epoch": 2.2, "learning_rate": 1.7010835498069937e-05, "loss": 0.0875, "step": 12002 }, { "epoch": 2.2, "learning_rate": 1.7010306067386203e-05, "loss": 0.0769, "step": 12003 }, { "epoch": 2.2, "learning_rate": 1.7009776598061496e-05, "loss": 0.0742, "step": 12004 }, { "epoch": 2.2, "learning_rate": 1.7009247090098737e-05, "loss": 0.0823, "step": 12005 }, { "epoch": 2.2, "learning_rate": 1.7008717543500844e-05, "loss": 0.0908, "step": 12006 }, { "epoch": 2.2, "learning_rate": 1.7008187958270737e-05, "loss": 0.0756, "step": 12007 }, { "epoch": 2.2, "learning_rate": 1.7007658334411333e-05, "loss": 0.0755, "step": 12008 }, { "epoch": 2.2, "learning_rate": 1.700712867192555e-05, "loss": 0.0863, "step": 12009 }, { "epoch": 2.2, "learning_rate": 1.700659897081631e-05, "loss": 0.0853, "step": 12010 }, { "epoch": 2.2, "learning_rate": 1.7006069231086534e-05, "loss": 0.0798, "step": 12011 }, { "epoch": 2.2, "learning_rate": 1.700553945273914e-05, "loss": 0.0729, "step": 12012 }, { "epoch": 2.2, "learning_rate": 1.7005009635777047e-05, "loss": 0.0819, "step": 12013 }, { "epoch": 2.2, "learning_rate": 1.7004479780203178e-05, "loss": 0.0757, "step": 12014 }, { "epoch": 2.2, "learning_rate": 1.7003949886020453e-05, "loss": 0.0889, "step": 12015 }, { "epoch": 2.2, "learning_rate": 1.700341995323179e-05, "loss": 0.0881, "step": 12016 }, { "epoch": 2.2, "learning_rate": 1.7002889981840114e-05, "loss": 0.0832, "step": 12017 }, { "epoch": 2.2, "learning_rate": 1.7002359971848344e-05, "loss": 0.0793, "step": 12018 }, { "epoch": 2.2, "learning_rate": 1.70018299232594e-05, "loss": 0.0822, "step": 12019 }, { "epoch": 2.2, "learning_rate": 1.7001299836076207e-05, "loss": 0.07, "step": 12020 }, { "epoch": 2.2, "learning_rate": 1.7000769710301684e-05, "loss": 0.0696, "step": 12021 }, { "epoch": 2.2, "learning_rate": 1.700023954593876e-05, "loss": 0.0853, "step": 12022 }, { "epoch": 2.2, "learning_rate": 1.6999709342990346e-05, "loss": 0.0799, "step": 12023 }, { "epoch": 2.21, "learning_rate": 1.6999179101459372e-05, "loss": 0.0752, "step": 12024 }, { "epoch": 2.21, "learning_rate": 1.699864882134876e-05, "loss": 0.0821, "step": 12025 }, { "epoch": 2.21, "learning_rate": 1.6998118502661428e-05, "loss": 0.0861, "step": 12026 }, { "epoch": 2.21, "learning_rate": 1.6997588145400302e-05, "loss": 0.0883, "step": 12027 }, { "epoch": 2.21, "learning_rate": 1.699705774956831e-05, "loss": 0.0706, "step": 12028 }, { "epoch": 2.21, "learning_rate": 1.6996527315168372e-05, "loss": 0.0872, "step": 12029 }, { "epoch": 2.21, "learning_rate": 1.699599684220341e-05, "loss": 0.0787, "step": 12030 }, { "epoch": 2.21, "learning_rate": 1.699546633067635e-05, "loss": 0.0794, "step": 12031 }, { "epoch": 2.21, "learning_rate": 1.6994935780590115e-05, "loss": 0.0849, "step": 12032 }, { "epoch": 2.21, "learning_rate": 1.6994405191947635e-05, "loss": 0.0841, "step": 12033 }, { "epoch": 2.21, "learning_rate": 1.6993874564751824e-05, "loss": 0.0815, "step": 12034 }, { "epoch": 2.21, "learning_rate": 1.699334389900561e-05, "loss": 0.081, "step": 12035 }, { "epoch": 2.21, "learning_rate": 1.6992813194711926e-05, "loss": 0.0858, "step": 12036 }, { "epoch": 2.21, "learning_rate": 1.699228245187369e-05, "loss": 0.0732, "step": 12037 }, { "epoch": 2.21, "learning_rate": 1.699175167049383e-05, "loss": 0.08, "step": 12038 }, { "epoch": 2.21, "learning_rate": 1.6991220850575272e-05, "loss": 0.0837, "step": 12039 }, { "epoch": 2.21, "learning_rate": 1.6990689992120943e-05, "loss": 0.0812, "step": 12040 }, { "epoch": 2.21, "learning_rate": 1.699015909513376e-05, "loss": 0.0837, "step": 12041 }, { "epoch": 2.21, "learning_rate": 1.698962815961666e-05, "loss": 0.0953, "step": 12042 }, { "epoch": 2.21, "learning_rate": 1.6989097185572564e-05, "loss": 0.0829, "step": 12043 }, { "epoch": 2.21, "learning_rate": 1.6988566173004405e-05, "loss": 0.0753, "step": 12044 }, { "epoch": 2.21, "learning_rate": 1.6988035121915102e-05, "loss": 0.0715, "step": 12045 }, { "epoch": 2.21, "learning_rate": 1.6987504032307585e-05, "loss": 0.0826, "step": 12046 }, { "epoch": 2.21, "learning_rate": 1.6986972904184783e-05, "loss": 0.0857, "step": 12047 }, { "epoch": 2.21, "learning_rate": 1.6986441737549625e-05, "loss": 0.0846, "step": 12048 }, { "epoch": 2.21, "learning_rate": 1.6985910532405032e-05, "loss": 0.0825, "step": 12049 }, { "epoch": 2.21, "learning_rate": 1.698537928875394e-05, "loss": 0.0807, "step": 12050 }, { "epoch": 2.21, "learning_rate": 1.6984848006599276e-05, "loss": 0.0791, "step": 12051 }, { "epoch": 2.21, "learning_rate": 1.6984316685943964e-05, "loss": 0.0827, "step": 12052 }, { "epoch": 2.21, "learning_rate": 1.6983785326790933e-05, "loss": 0.075, "step": 12053 }, { "epoch": 2.21, "learning_rate": 1.698325392914311e-05, "loss": 0.0789, "step": 12054 }, { "epoch": 2.21, "learning_rate": 1.6982722493003434e-05, "loss": 0.0725, "step": 12055 }, { "epoch": 2.21, "learning_rate": 1.6982191018374828e-05, "loss": 0.0774, "step": 12056 }, { "epoch": 2.21, "learning_rate": 1.698165950526022e-05, "loss": 0.0892, "step": 12057 }, { "epoch": 2.21, "learning_rate": 1.698112795366254e-05, "loss": 0.0778, "step": 12058 }, { "epoch": 2.21, "learning_rate": 1.698059636358472e-05, "loss": 0.0793, "step": 12059 }, { "epoch": 2.21, "learning_rate": 1.6980064735029688e-05, "loss": 0.0759, "step": 12060 }, { "epoch": 2.21, "learning_rate": 1.697953306800038e-05, "loss": 0.0906, "step": 12061 }, { "epoch": 2.21, "learning_rate": 1.6979001362499717e-05, "loss": 0.086, "step": 12062 }, { "epoch": 2.21, "learning_rate": 1.6978469618530636e-05, "loss": 0.0823, "step": 12063 }, { "epoch": 2.21, "learning_rate": 1.697793783609607e-05, "loss": 0.0864, "step": 12064 }, { "epoch": 2.21, "learning_rate": 1.6977406015198947e-05, "loss": 0.0849, "step": 12065 }, { "epoch": 2.21, "learning_rate": 1.69768741558422e-05, "loss": 0.0842, "step": 12066 }, { "epoch": 2.21, "learning_rate": 1.6976342258028754e-05, "loss": 0.0752, "step": 12067 }, { "epoch": 2.21, "learning_rate": 1.6975810321761547e-05, "loss": 0.0764, "step": 12068 }, { "epoch": 2.21, "learning_rate": 1.697527834704351e-05, "loss": 0.0807, "step": 12069 }, { "epoch": 2.21, "learning_rate": 1.697474633387758e-05, "loss": 0.0845, "step": 12070 }, { "epoch": 2.21, "learning_rate": 1.6974214282266678e-05, "loss": 0.0899, "step": 12071 }, { "epoch": 2.21, "learning_rate": 1.697368219221375e-05, "loss": 0.0779, "step": 12072 }, { "epoch": 2.21, "learning_rate": 1.6973150063721718e-05, "loss": 0.0946, "step": 12073 }, { "epoch": 2.21, "learning_rate": 1.697261789679352e-05, "loss": 0.0729, "step": 12074 }, { "epoch": 2.21, "learning_rate": 1.6972085691432092e-05, "loss": 0.0769, "step": 12075 }, { "epoch": 2.21, "learning_rate": 1.697155344764036e-05, "loss": 0.0899, "step": 12076 }, { "epoch": 2.21, "learning_rate": 1.6971021165421267e-05, "loss": 0.0787, "step": 12077 }, { "epoch": 2.21, "learning_rate": 1.6970488844777736e-05, "loss": 0.0803, "step": 12078 }, { "epoch": 2.22, "learning_rate": 1.6969956485712712e-05, "loss": 0.0823, "step": 12079 }, { "epoch": 2.22, "learning_rate": 1.6969424088229123e-05, "loss": 0.0829, "step": 12080 }, { "epoch": 2.22, "learning_rate": 1.6968891652329907e-05, "loss": 0.0798, "step": 12081 }, { "epoch": 2.22, "learning_rate": 1.6968359178017993e-05, "loss": 0.0846, "step": 12082 }, { "epoch": 2.22, "learning_rate": 1.6967826665296324e-05, "loss": 0.09, "step": 12083 }, { "epoch": 2.22, "learning_rate": 1.6967294114167827e-05, "loss": 0.0836, "step": 12084 }, { "epoch": 2.22, "learning_rate": 1.696676152463544e-05, "loss": 0.0861, "step": 12085 }, { "epoch": 2.22, "learning_rate": 1.696622889670211e-05, "loss": 0.0837, "step": 12086 }, { "epoch": 2.22, "learning_rate": 1.6965696230370757e-05, "loss": 0.0738, "step": 12087 }, { "epoch": 2.22, "learning_rate": 1.696516352564432e-05, "loss": 0.0789, "step": 12088 }, { "epoch": 2.22, "learning_rate": 1.6964630782525743e-05, "loss": 0.0788, "step": 12089 }, { "epoch": 2.22, "learning_rate": 1.6964098001017958e-05, "loss": 0.0938, "step": 12090 }, { "epoch": 2.22, "learning_rate": 1.6963565181123897e-05, "loss": 0.0712, "step": 12091 }, { "epoch": 2.22, "learning_rate": 1.6963032322846507e-05, "loss": 0.0843, "step": 12092 }, { "epoch": 2.22, "learning_rate": 1.696249942618872e-05, "loss": 0.0842, "step": 12093 }, { "epoch": 2.22, "learning_rate": 1.696196649115347e-05, "loss": 0.0849, "step": 12094 }, { "epoch": 2.22, "learning_rate": 1.69614335177437e-05, "loss": 0.0799, "step": 12095 }, { "epoch": 2.22, "learning_rate": 1.696090050596234e-05, "loss": 0.0775, "step": 12096 }, { "epoch": 2.22, "learning_rate": 1.6960367455812336e-05, "loss": 0.0805, "step": 12097 }, { "epoch": 2.22, "learning_rate": 1.6959834367296626e-05, "loss": 0.0821, "step": 12098 }, { "epoch": 2.22, "learning_rate": 1.6959301240418144e-05, "loss": 0.0788, "step": 12099 }, { "epoch": 2.22, "learning_rate": 1.695876807517983e-05, "loss": 0.0833, "step": 12100 }, { "epoch": 2.22, "learning_rate": 1.6958234871584627e-05, "loss": 0.0776, "step": 12101 }, { "epoch": 2.22, "learning_rate": 1.6957701629635468e-05, "loss": 0.0683, "step": 12102 }, { "epoch": 2.22, "learning_rate": 1.6957168349335293e-05, "loss": 0.0779, "step": 12103 }, { "epoch": 2.22, "learning_rate": 1.6956635030687047e-05, "loss": 0.0826, "step": 12104 }, { "epoch": 2.22, "learning_rate": 1.695610167369366e-05, "loss": 0.0706, "step": 12105 }, { "epoch": 2.22, "learning_rate": 1.6955568278358083e-05, "loss": 0.0711, "step": 12106 }, { "epoch": 2.22, "learning_rate": 1.6955034844683246e-05, "loss": 0.0896, "step": 12107 }, { "epoch": 2.22, "learning_rate": 1.69545013726721e-05, "loss": 0.0822, "step": 12108 }, { "epoch": 2.22, "learning_rate": 1.6953967862327574e-05, "loss": 0.0787, "step": 12109 }, { "epoch": 2.22, "learning_rate": 1.6953434313652616e-05, "loss": 0.0787, "step": 12110 }, { "epoch": 2.22, "learning_rate": 1.6952900726650167e-05, "loss": 0.0825, "step": 12111 }, { "epoch": 2.22, "learning_rate": 1.6952367101323162e-05, "loss": 0.0682, "step": 12112 }, { "epoch": 2.22, "learning_rate": 1.695183343767455e-05, "loss": 0.0683, "step": 12113 }, { "epoch": 2.22, "learning_rate": 1.695129973570727e-05, "loss": 0.08, "step": 12114 }, { "epoch": 2.22, "learning_rate": 1.695076599542426e-05, "loss": 0.073, "step": 12115 }, { "epoch": 2.22, "learning_rate": 1.6950232216828468e-05, "loss": 0.0757, "step": 12116 }, { "epoch": 2.22, "learning_rate": 1.694969839992283e-05, "loss": 0.0744, "step": 12117 }, { "epoch": 2.22, "learning_rate": 1.6949164544710295e-05, "loss": 0.0717, "step": 12118 }, { "epoch": 2.22, "learning_rate": 1.69486306511938e-05, "loss": 0.0832, "step": 12119 }, { "epoch": 2.22, "learning_rate": 1.694809671937629e-05, "loss": 0.0896, "step": 12120 }, { "epoch": 2.22, "learning_rate": 1.694756274926071e-05, "loss": 0.0858, "step": 12121 }, { "epoch": 2.22, "learning_rate": 1.694702874085e-05, "loss": 0.0832, "step": 12122 }, { "epoch": 2.22, "learning_rate": 1.6946494694147103e-05, "loss": 0.0852, "step": 12123 }, { "epoch": 2.22, "learning_rate": 1.6945960609154966e-05, "loss": 0.0735, "step": 12124 }, { "epoch": 2.22, "learning_rate": 1.6945426485876533e-05, "loss": 0.0743, "step": 12125 }, { "epoch": 2.22, "learning_rate": 1.6944892324314745e-05, "loss": 0.0791, "step": 12126 }, { "epoch": 2.22, "learning_rate": 1.6944358124472547e-05, "loss": 0.0818, "step": 12127 }, { "epoch": 2.22, "learning_rate": 1.694382388635289e-05, "loss": 0.0767, "step": 12128 }, { "epoch": 2.22, "learning_rate": 1.6943289609958706e-05, "loss": 0.0795, "step": 12129 }, { "epoch": 2.22, "learning_rate": 1.694275529529295e-05, "loss": 0.0751, "step": 12130 }, { "epoch": 2.22, "learning_rate": 1.6942220942358564e-05, "loss": 0.088, "step": 12131 }, { "epoch": 2.22, "learning_rate": 1.694168655115849e-05, "loss": 0.0852, "step": 12132 }, { "epoch": 2.23, "learning_rate": 1.694115212169568e-05, "loss": 0.0804, "step": 12133 }, { "epoch": 2.23, "learning_rate": 1.694061765397308e-05, "loss": 0.0805, "step": 12134 }, { "epoch": 2.23, "learning_rate": 1.6940083147993628e-05, "loss": 0.0743, "step": 12135 }, { "epoch": 2.23, "learning_rate": 1.6939548603760278e-05, "loss": 0.075, "step": 12136 }, { "epoch": 2.23, "learning_rate": 1.6939014021275973e-05, "loss": 0.0812, "step": 12137 }, { "epoch": 2.23, "learning_rate": 1.693847940054366e-05, "loss": 0.0855, "step": 12138 }, { "epoch": 2.23, "learning_rate": 1.693794474156629e-05, "loss": 0.0738, "step": 12139 }, { "epoch": 2.23, "learning_rate": 1.69374100443468e-05, "loss": 0.1008, "step": 12140 }, { "epoch": 2.23, "learning_rate": 1.6936875308888144e-05, "loss": 0.0871, "step": 12141 }, { "epoch": 2.23, "learning_rate": 1.693634053519327e-05, "loss": 0.0739, "step": 12142 }, { "epoch": 2.23, "learning_rate": 1.6935805723265125e-05, "loss": 0.0911, "step": 12143 }, { "epoch": 2.23, "learning_rate": 1.6935270873106656e-05, "loss": 0.0909, "step": 12144 }, { "epoch": 2.23, "learning_rate": 1.693473598472081e-05, "loss": 0.0706, "step": 12145 }, { "epoch": 2.23, "learning_rate": 1.6934201058110537e-05, "loss": 0.0699, "step": 12146 }, { "epoch": 2.23, "learning_rate": 1.6933666093278786e-05, "loss": 0.0794, "step": 12147 }, { "epoch": 2.23, "learning_rate": 1.6933131090228508e-05, "loss": 0.0965, "step": 12148 }, { "epoch": 2.23, "learning_rate": 1.6932596048962645e-05, "loss": 0.0871, "step": 12149 }, { "epoch": 2.23, "learning_rate": 1.693206096948415e-05, "loss": 0.0759, "step": 12150 }, { "epoch": 2.23, "learning_rate": 1.6931525851795977e-05, "loss": 0.0745, "step": 12151 }, { "epoch": 2.23, "learning_rate": 1.6930990695901067e-05, "loss": 0.0767, "step": 12152 }, { "epoch": 2.23, "learning_rate": 1.6930455501802376e-05, "loss": 0.0839, "step": 12153 }, { "epoch": 2.23, "learning_rate": 1.692992026950285e-05, "loss": 0.0727, "step": 12154 }, { "epoch": 2.23, "learning_rate": 1.6929384999005444e-05, "loss": 0.0789, "step": 12155 }, { "epoch": 2.23, "learning_rate": 1.6928849690313102e-05, "loss": 0.0838, "step": 12156 }, { "epoch": 2.23, "learning_rate": 1.6928314343428784e-05, "loss": 0.0815, "step": 12157 }, { "epoch": 2.23, "learning_rate": 1.692777895835543e-05, "loss": 0.0858, "step": 12158 }, { "epoch": 2.23, "learning_rate": 1.6927243535095995e-05, "loss": 0.08, "step": 12159 }, { "epoch": 2.23, "learning_rate": 1.6926708073653432e-05, "loss": 0.1041, "step": 12160 }, { "epoch": 2.23, "learning_rate": 1.6926172574030695e-05, "loss": 0.0785, "step": 12161 }, { "epoch": 2.23, "learning_rate": 1.692563703623073e-05, "loss": 0.0748, "step": 12162 }, { "epoch": 2.23, "learning_rate": 1.6925101460256495e-05, "loss": 0.0805, "step": 12163 }, { "epoch": 2.23, "learning_rate": 1.6924565846110933e-05, "loss": 0.0752, "step": 12164 }, { "epoch": 2.23, "learning_rate": 1.6924030193797006e-05, "loss": 0.0877, "step": 12165 }, { "epoch": 2.23, "learning_rate": 1.692349450331766e-05, "loss": 0.0835, "step": 12166 }, { "epoch": 2.23, "learning_rate": 1.692295877467585e-05, "loss": 0.0689, "step": 12167 }, { "epoch": 2.23, "learning_rate": 1.6922423007874532e-05, "loss": 0.0758, "step": 12168 }, { "epoch": 2.23, "learning_rate": 1.6921887202916655e-05, "loss": 0.0837, "step": 12169 }, { "epoch": 2.23, "learning_rate": 1.6921351359805172e-05, "loss": 0.0859, "step": 12170 }, { "epoch": 2.23, "learning_rate": 1.6920815478543037e-05, "loss": 0.0746, "step": 12171 }, { "epoch": 2.23, "learning_rate": 1.692027955913321e-05, "loss": 0.0717, "step": 12172 }, { "epoch": 2.23, "learning_rate": 1.6919743601578636e-05, "loss": 0.0732, "step": 12173 }, { "epoch": 2.23, "learning_rate": 1.6919207605882272e-05, "loss": 0.0811, "step": 12174 }, { "epoch": 2.23, "learning_rate": 1.6918671572047077e-05, "loss": 0.0855, "step": 12175 }, { "epoch": 2.23, "learning_rate": 1.6918135500076e-05, "loss": 0.0794, "step": 12176 }, { "epoch": 2.23, "learning_rate": 1.6917599389972e-05, "loss": 0.0807, "step": 12177 }, { "epoch": 2.23, "learning_rate": 1.691706324173803e-05, "loss": 0.0795, "step": 12178 }, { "epoch": 2.23, "learning_rate": 1.6916527055377042e-05, "loss": 0.0854, "step": 12179 }, { "epoch": 2.23, "learning_rate": 1.6915990830891997e-05, "loss": 0.0846, "step": 12180 }, { "epoch": 2.23, "learning_rate": 1.691545456828585e-05, "loss": 0.0771, "step": 12181 }, { "epoch": 2.23, "learning_rate": 1.6914918267561552e-05, "loss": 0.0884, "step": 12182 }, { "epoch": 2.23, "learning_rate": 1.6914381928722065e-05, "loss": 0.084, "step": 12183 }, { "epoch": 2.23, "learning_rate": 1.6913845551770344e-05, "loss": 0.0895, "step": 12184 }, { "epoch": 2.23, "learning_rate": 1.6913309136709343e-05, "loss": 0.0892, "step": 12185 }, { "epoch": 2.23, "learning_rate": 1.6912772683542017e-05, "loss": 0.0675, "step": 12186 }, { "epoch": 2.23, "learning_rate": 1.691223619227133e-05, "loss": 0.081, "step": 12187 }, { "epoch": 2.24, "learning_rate": 1.6911699662900233e-05, "loss": 0.0771, "step": 12188 }, { "epoch": 2.24, "learning_rate": 1.6911163095431686e-05, "loss": 0.0922, "step": 12189 }, { "epoch": 2.24, "learning_rate": 1.691062648986865e-05, "loss": 0.087, "step": 12190 }, { "epoch": 2.24, "learning_rate": 1.6910089846214072e-05, "loss": 0.0861, "step": 12191 }, { "epoch": 2.24, "learning_rate": 1.690955316447092e-05, "loss": 0.086, "step": 12192 }, { "epoch": 2.24, "learning_rate": 1.6909016444642154e-05, "loss": 0.0755, "step": 12193 }, { "epoch": 2.24, "learning_rate": 1.690847968673072e-05, "loss": 0.0835, "step": 12194 }, { "epoch": 2.24, "learning_rate": 1.6907942890739586e-05, "loss": 0.088, "step": 12195 }, { "epoch": 2.24, "learning_rate": 1.690740605667171e-05, "loss": 0.0888, "step": 12196 }, { "epoch": 2.24, "learning_rate": 1.6906869184530052e-05, "loss": 0.0941, "step": 12197 }, { "epoch": 2.24, "learning_rate": 1.6906332274317568e-05, "loss": 0.0723, "step": 12198 }, { "epoch": 2.24, "learning_rate": 1.6905795326037218e-05, "loss": 0.0883, "step": 12199 }, { "epoch": 2.24, "learning_rate": 1.6905258339691963e-05, "loss": 0.0856, "step": 12200 }, { "epoch": 2.24, "learning_rate": 1.690472131528476e-05, "loss": 0.0912, "step": 12201 }, { "epoch": 2.24, "learning_rate": 1.6904184252818572e-05, "loss": 0.0784, "step": 12202 }, { "epoch": 2.24, "learning_rate": 1.6903647152296362e-05, "loss": 0.0731, "step": 12203 }, { "epoch": 2.24, "learning_rate": 1.690311001372108e-05, "loss": 0.0834, "step": 12204 }, { "epoch": 2.24, "learning_rate": 1.6902572837095702e-05, "loss": 0.0793, "step": 12205 }, { "epoch": 2.24, "learning_rate": 1.6902035622423175e-05, "loss": 0.081, "step": 12206 }, { "epoch": 2.24, "learning_rate": 1.690149836970647e-05, "loss": 0.0783, "step": 12207 }, { "epoch": 2.24, "learning_rate": 1.6900961078948542e-05, "loss": 0.0814, "step": 12208 }, { "epoch": 2.24, "learning_rate": 1.690042375015236e-05, "loss": 0.0852, "step": 12209 }, { "epoch": 2.24, "learning_rate": 1.6899886383320873e-05, "loss": 0.0838, "step": 12210 }, { "epoch": 2.24, "learning_rate": 1.6899348978457055e-05, "loss": 0.0767, "step": 12211 }, { "epoch": 2.24, "learning_rate": 1.689881153556386e-05, "loss": 0.0799, "step": 12212 }, { "epoch": 2.24, "learning_rate": 1.689827405464426e-05, "loss": 0.0797, "step": 12213 }, { "epoch": 2.24, "learning_rate": 1.6897736535701208e-05, "loss": 0.0854, "step": 12214 }, { "epoch": 2.24, "learning_rate": 1.6897198978737673e-05, "loss": 0.082, "step": 12215 }, { "epoch": 2.24, "learning_rate": 1.6896661383756612e-05, "loss": 0.0643, "step": 12216 }, { "epoch": 2.24, "learning_rate": 1.6896123750760994e-05, "loss": 0.081, "step": 12217 }, { "epoch": 2.24, "learning_rate": 1.689558607975378e-05, "loss": 0.0776, "step": 12218 }, { "epoch": 2.24, "learning_rate": 1.6895048370737933e-05, "loss": 0.0918, "step": 12219 }, { "epoch": 2.24, "learning_rate": 1.6894510623716417e-05, "loss": 0.0734, "step": 12220 }, { "epoch": 2.24, "learning_rate": 1.68939728386922e-05, "loss": 0.0861, "step": 12221 }, { "epoch": 2.24, "learning_rate": 1.689343501566824e-05, "loss": 0.0722, "step": 12222 }, { "epoch": 2.24, "learning_rate": 1.6892897154647512e-05, "loss": 0.0889, "step": 12223 }, { "epoch": 2.24, "learning_rate": 1.6892359255632967e-05, "loss": 0.0735, "step": 12224 }, { "epoch": 2.24, "learning_rate": 1.6891821318627578e-05, "loss": 0.0743, "step": 12225 }, { "epoch": 2.24, "learning_rate": 1.6891283343634306e-05, "loss": 0.0838, "step": 12226 }, { "epoch": 2.24, "learning_rate": 1.6890745330656123e-05, "loss": 0.0824, "step": 12227 }, { "epoch": 2.24, "learning_rate": 1.6890207279695987e-05, "loss": 0.0749, "step": 12228 }, { "epoch": 2.24, "learning_rate": 1.688966919075687e-05, "loss": 0.08, "step": 12229 }, { "epoch": 2.24, "learning_rate": 1.688913106384173e-05, "loss": 0.0786, "step": 12230 }, { "epoch": 2.24, "learning_rate": 1.688859289895354e-05, "loss": 0.0716, "step": 12231 }, { "epoch": 2.24, "learning_rate": 1.688805469609527e-05, "loss": 0.0771, "step": 12232 }, { "epoch": 2.24, "learning_rate": 1.6887516455269873e-05, "loss": 0.0757, "step": 12233 }, { "epoch": 2.24, "learning_rate": 1.688697817648033e-05, "loss": 0.0875, "step": 12234 }, { "epoch": 2.24, "learning_rate": 1.6886439859729597e-05, "loss": 0.069, "step": 12235 }, { "epoch": 2.24, "learning_rate": 1.6885901505020653e-05, "loss": 0.0776, "step": 12236 }, { "epoch": 2.24, "learning_rate": 1.6885363112356452e-05, "loss": 0.0689, "step": 12237 }, { "epoch": 2.24, "learning_rate": 1.688482468173997e-05, "loss": 0.0772, "step": 12238 }, { "epoch": 2.24, "learning_rate": 1.6884286213174176e-05, "loss": 0.0773, "step": 12239 }, { "epoch": 2.24, "learning_rate": 1.688374770666203e-05, "loss": 0.0948, "step": 12240 }, { "epoch": 2.24, "learning_rate": 1.688320916220651e-05, "loss": 0.0745, "step": 12241 }, { "epoch": 2.25, "learning_rate": 1.6882670579810578e-05, "loss": 0.0808, "step": 12242 }, { "epoch": 2.25, "learning_rate": 1.6882131959477205e-05, "loss": 0.0801, "step": 12243 }, { "epoch": 2.25, "learning_rate": 1.6881593301209356e-05, "loss": 0.0808, "step": 12244 }, { "epoch": 2.25, "learning_rate": 1.688105460501001e-05, "loss": 0.0717, "step": 12245 }, { "epoch": 2.25, "learning_rate": 1.6880515870882125e-05, "loss": 0.0756, "step": 12246 }, { "epoch": 2.25, "learning_rate": 1.687997709882868e-05, "loss": 0.0766, "step": 12247 }, { "epoch": 2.25, "learning_rate": 1.6879438288852635e-05, "loss": 0.0896, "step": 12248 }, { "epoch": 2.25, "learning_rate": 1.6878899440956966e-05, "loss": 0.0823, "step": 12249 }, { "epoch": 2.25, "learning_rate": 1.6878360555144643e-05, "loss": 0.0716, "step": 12250 }, { "epoch": 2.25, "learning_rate": 1.6877821631418637e-05, "loss": 0.0823, "step": 12251 }, { "epoch": 2.25, "learning_rate": 1.6877282669781914e-05, "loss": 0.0824, "step": 12252 }, { "epoch": 2.25, "learning_rate": 1.6876743670237447e-05, "loss": 0.0731, "step": 12253 }, { "epoch": 2.25, "learning_rate": 1.6876204632788212e-05, "loss": 0.0856, "step": 12254 }, { "epoch": 2.25, "learning_rate": 1.6875665557437173e-05, "loss": 0.0753, "step": 12255 }, { "epoch": 2.25, "learning_rate": 1.6875126444187308e-05, "loss": 0.0811, "step": 12256 }, { "epoch": 2.25, "learning_rate": 1.687458729304158e-05, "loss": 0.0719, "step": 12257 }, { "epoch": 2.25, "learning_rate": 1.687404810400297e-05, "loss": 0.0821, "step": 12258 }, { "epoch": 2.25, "learning_rate": 1.6873508877074443e-05, "loss": 0.076, "step": 12259 }, { "epoch": 2.25, "learning_rate": 1.6872969612258976e-05, "loss": 0.0808, "step": 12260 }, { "epoch": 2.25, "learning_rate": 1.6872430309559534e-05, "loss": 0.0865, "step": 12261 }, { "epoch": 2.25, "learning_rate": 1.6871890968979104e-05, "loss": 0.0913, "step": 12262 }, { "epoch": 2.25, "learning_rate": 1.6871351590520642e-05, "loss": 0.0821, "step": 12263 }, { "epoch": 2.25, "learning_rate": 1.6870812174187136e-05, "loss": 0.079, "step": 12264 }, { "epoch": 2.25, "learning_rate": 1.687027271998155e-05, "loss": 0.0852, "step": 12265 }, { "epoch": 2.25, "learning_rate": 1.6869733227906855e-05, "loss": 0.0803, "step": 12266 }, { "epoch": 2.25, "learning_rate": 1.6869193697966037e-05, "loss": 0.0841, "step": 12267 }, { "epoch": 2.25, "learning_rate": 1.6868654130162056e-05, "loss": 0.0702, "step": 12268 }, { "epoch": 2.25, "learning_rate": 1.6868114524497895e-05, "loss": 0.0819, "step": 12269 }, { "epoch": 2.25, "learning_rate": 1.686757488097653e-05, "loss": 0.0774, "step": 12270 }, { "epoch": 2.25, "learning_rate": 1.6867035199600924e-05, "loss": 0.0767, "step": 12271 }, { "epoch": 2.25, "learning_rate": 1.686649548037406e-05, "loss": 0.0848, "step": 12272 }, { "epoch": 2.25, "learning_rate": 1.6865955723298917e-05, "loss": 0.0785, "step": 12273 }, { "epoch": 2.25, "learning_rate": 1.686541592837846e-05, "loss": 0.0791, "step": 12274 }, { "epoch": 2.25, "learning_rate": 1.6864876095615674e-05, "loss": 0.0859, "step": 12275 }, { "epoch": 2.25, "learning_rate": 1.686433622501353e-05, "loss": 0.0775, "step": 12276 }, { "epoch": 2.25, "learning_rate": 1.6863796316575e-05, "loss": 0.0857, "step": 12277 }, { "epoch": 2.25, "learning_rate": 1.686325637030307e-05, "loss": 0.0754, "step": 12278 }, { "epoch": 2.25, "learning_rate": 1.6862716386200702e-05, "loss": 0.0748, "step": 12279 }, { "epoch": 2.25, "learning_rate": 1.6862176364270887e-05, "loss": 0.0702, "step": 12280 }, { "epoch": 2.25, "learning_rate": 1.686163630451659e-05, "loss": 0.0894, "step": 12281 }, { "epoch": 2.25, "learning_rate": 1.68610962069408e-05, "loss": 0.0788, "step": 12282 }, { "epoch": 2.25, "learning_rate": 1.686055607154648e-05, "loss": 0.0715, "step": 12283 }, { "epoch": 2.25, "learning_rate": 1.6860015898336617e-05, "loss": 0.077, "step": 12284 }, { "epoch": 2.25, "learning_rate": 1.6859475687314187e-05, "loss": 0.0815, "step": 12285 }, { "epoch": 2.25, "learning_rate": 1.6858935438482165e-05, "loss": 0.0799, "step": 12286 }, { "epoch": 2.25, "learning_rate": 1.685839515184353e-05, "loss": 0.0753, "step": 12287 }, { "epoch": 2.25, "learning_rate": 1.6857854827401262e-05, "loss": 0.0866, "step": 12288 }, { "epoch": 2.25, "learning_rate": 1.6857314465158335e-05, "loss": 0.0761, "step": 12289 }, { "epoch": 2.25, "learning_rate": 1.6856774065117728e-05, "loss": 0.0816, "step": 12290 }, { "epoch": 2.25, "learning_rate": 1.685623362728243e-05, "loss": 0.0829, "step": 12291 }, { "epoch": 2.25, "learning_rate": 1.6855693151655404e-05, "loss": 0.0707, "step": 12292 }, { "epoch": 2.25, "learning_rate": 1.6855152638239637e-05, "loss": 0.0806, "step": 12293 }, { "epoch": 2.25, "learning_rate": 1.6854612087038113e-05, "loss": 0.0864, "step": 12294 }, { "epoch": 2.25, "learning_rate": 1.6854071498053803e-05, "loss": 0.078, "step": 12295 }, { "epoch": 2.25, "learning_rate": 1.6853530871289693e-05, "loss": 0.074, "step": 12296 }, { "epoch": 2.26, "learning_rate": 1.6852990206748757e-05, "loss": 0.0781, "step": 12297 }, { "epoch": 2.26, "learning_rate": 1.6852449504433977e-05, "loss": 0.0746, "step": 12298 }, { "epoch": 2.26, "learning_rate": 1.6851908764348338e-05, "loss": 0.0755, "step": 12299 }, { "epoch": 2.26, "learning_rate": 1.6851367986494817e-05, "loss": 0.0794, "step": 12300 }, { "epoch": 2.26, "learning_rate": 1.6850827170876397e-05, "loss": 0.0724, "step": 12301 }, { "epoch": 2.26, "learning_rate": 1.6850286317496055e-05, "loss": 0.077, "step": 12302 }, { "epoch": 2.26, "learning_rate": 1.6849745426356774e-05, "loss": 0.0733, "step": 12303 }, { "epoch": 2.26, "learning_rate": 1.6849204497461537e-05, "loss": 0.0746, "step": 12304 }, { "epoch": 2.26, "learning_rate": 1.6848663530813324e-05, "loss": 0.0792, "step": 12305 }, { "epoch": 2.26, "learning_rate": 1.6848122526415116e-05, "loss": 0.069, "step": 12306 }, { "epoch": 2.26, "learning_rate": 1.6847581484269897e-05, "loss": 0.0779, "step": 12307 }, { "epoch": 2.26, "learning_rate": 1.6847040404380647e-05, "loss": 0.07, "step": 12308 }, { "epoch": 2.26, "learning_rate": 1.6846499286750352e-05, "loss": 0.0786, "step": 12309 }, { "epoch": 2.26, "learning_rate": 1.6845958131381992e-05, "loss": 0.0845, "step": 12310 }, { "epoch": 2.26, "learning_rate": 1.6845416938278552e-05, "loss": 0.0746, "step": 12311 }, { "epoch": 2.26, "learning_rate": 1.684487570744301e-05, "loss": 0.09, "step": 12312 }, { "epoch": 2.26, "learning_rate": 1.6844334438878355e-05, "loss": 0.0836, "step": 12313 }, { "epoch": 2.26, "learning_rate": 1.684379313258757e-05, "loss": 0.085, "step": 12314 }, { "epoch": 2.26, "learning_rate": 1.684325178857363e-05, "loss": 0.0807, "step": 12315 }, { "epoch": 2.26, "learning_rate": 1.684271040683953e-05, "loss": 0.0803, "step": 12316 }, { "epoch": 2.26, "learning_rate": 1.6842168987388253e-05, "loss": 0.0774, "step": 12317 }, { "epoch": 2.26, "learning_rate": 1.6841627530222776e-05, "loss": 0.0762, "step": 12318 }, { "epoch": 2.26, "learning_rate": 1.684108603534609e-05, "loss": 0.0885, "step": 12319 }, { "epoch": 2.26, "learning_rate": 1.684054450276118e-05, "loss": 0.0763, "step": 12320 }, { "epoch": 2.26, "learning_rate": 1.6840002932471023e-05, "loss": 0.0769, "step": 12321 }, { "epoch": 2.26, "learning_rate": 1.6839461324478614e-05, "loss": 0.0775, "step": 12322 }, { "epoch": 2.26, "learning_rate": 1.683891967878693e-05, "loss": 0.0796, "step": 12323 }, { "epoch": 2.26, "learning_rate": 1.6838377995398962e-05, "loss": 0.0817, "step": 12324 }, { "epoch": 2.26, "learning_rate": 1.6837836274317693e-05, "loss": 0.0769, "step": 12325 }, { "epoch": 2.26, "learning_rate": 1.6837294515546115e-05, "loss": 0.0877, "step": 12326 }, { "epoch": 2.26, "learning_rate": 1.6836752719087208e-05, "loss": 0.0836, "step": 12327 }, { "epoch": 2.26, "learning_rate": 1.6836210884943954e-05, "loss": 0.0804, "step": 12328 }, { "epoch": 2.26, "learning_rate": 1.6835669013119352e-05, "loss": 0.0788, "step": 12329 }, { "epoch": 2.26, "learning_rate": 1.6835127103616376e-05, "loss": 0.089, "step": 12330 }, { "epoch": 2.26, "learning_rate": 1.6834585156438026e-05, "loss": 0.0779, "step": 12331 }, { "epoch": 2.26, "learning_rate": 1.683404317158728e-05, "loss": 0.0821, "step": 12332 }, { "epoch": 2.26, "learning_rate": 1.6833501149067126e-05, "loss": 0.0839, "step": 12333 }, { "epoch": 2.26, "learning_rate": 1.6832959088880557e-05, "loss": 0.0819, "step": 12334 }, { "epoch": 2.26, "learning_rate": 1.6832416991030556e-05, "loss": 0.0848, "step": 12335 }, { "epoch": 2.26, "learning_rate": 1.683187485552011e-05, "loss": 0.0854, "step": 12336 }, { "epoch": 2.26, "learning_rate": 1.6831332682352212e-05, "loss": 0.0886, "step": 12337 }, { "epoch": 2.26, "learning_rate": 1.683079047152985e-05, "loss": 0.0831, "step": 12338 }, { "epoch": 2.26, "learning_rate": 1.6830248223056008e-05, "loss": 0.0856, "step": 12339 }, { "epoch": 2.26, "learning_rate": 1.682970593693368e-05, "loss": 0.0755, "step": 12340 }, { "epoch": 2.26, "learning_rate": 1.682916361316585e-05, "loss": 0.0819, "step": 12341 }, { "epoch": 2.26, "learning_rate": 1.6828621251755513e-05, "loss": 0.0927, "step": 12342 }, { "epoch": 2.26, "learning_rate": 1.6828078852705654e-05, "loss": 0.077, "step": 12343 }, { "epoch": 2.26, "learning_rate": 1.6827536416019263e-05, "loss": 0.0767, "step": 12344 }, { "epoch": 2.26, "learning_rate": 1.6826993941699336e-05, "loss": 0.0787, "step": 12345 }, { "epoch": 2.26, "learning_rate": 1.6826451429748852e-05, "loss": 0.079, "step": 12346 }, { "epoch": 2.26, "learning_rate": 1.682590888017081e-05, "loss": 0.079, "step": 12347 }, { "epoch": 2.26, "learning_rate": 1.6825366292968205e-05, "loss": 0.0904, "step": 12348 }, { "epoch": 2.26, "learning_rate": 1.6824823668144015e-05, "loss": 0.0853, "step": 12349 }, { "epoch": 2.26, "learning_rate": 1.682428100570124e-05, "loss": 0.0871, "step": 12350 }, { "epoch": 2.26, "learning_rate": 1.6823738305642864e-05, "loss": 0.0825, "step": 12351 }, { "epoch": 2.27, "learning_rate": 1.6823195567971883e-05, "loss": 0.0795, "step": 12352 }, { "epoch": 2.27, "learning_rate": 1.682265279269129e-05, "loss": 0.088, "step": 12353 }, { "epoch": 2.27, "learning_rate": 1.6822109979804075e-05, "loss": 0.0767, "step": 12354 }, { "epoch": 2.27, "learning_rate": 1.6821567129313228e-05, "loss": 0.0807, "step": 12355 }, { "epoch": 2.27, "learning_rate": 1.682102424122175e-05, "loss": 0.0837, "step": 12356 }, { "epoch": 2.27, "learning_rate": 1.682048131553262e-05, "loss": 0.0852, "step": 12357 }, { "epoch": 2.27, "learning_rate": 1.6819938352248838e-05, "loss": 0.0732, "step": 12358 }, { "epoch": 2.27, "learning_rate": 1.6819395351373397e-05, "loss": 0.084, "step": 12359 }, { "epoch": 2.27, "learning_rate": 1.681885231290929e-05, "loss": 0.0844, "step": 12360 }, { "epoch": 2.27, "learning_rate": 1.681830923685951e-05, "loss": 0.0791, "step": 12361 }, { "epoch": 2.27, "learning_rate": 1.6817766123227045e-05, "loss": 0.0783, "step": 12362 }, { "epoch": 2.27, "learning_rate": 1.6817222972014894e-05, "loss": 0.0823, "step": 12363 }, { "epoch": 2.27, "learning_rate": 1.6816679783226054e-05, "loss": 0.0801, "step": 12364 }, { "epoch": 2.27, "learning_rate": 1.6816136556863514e-05, "loss": 0.074, "step": 12365 }, { "epoch": 2.27, "learning_rate": 1.681559329293027e-05, "loss": 0.0758, "step": 12366 }, { "epoch": 2.27, "learning_rate": 1.6815049991429315e-05, "loss": 0.0831, "step": 12367 }, { "epoch": 2.27, "learning_rate": 1.6814506652363647e-05, "loss": 0.0776, "step": 12368 }, { "epoch": 2.27, "learning_rate": 1.6813963275736256e-05, "loss": 0.0818, "step": 12369 }, { "epoch": 2.27, "learning_rate": 1.681341986155014e-05, "loss": 0.0928, "step": 12370 }, { "epoch": 2.27, "learning_rate": 1.6812876409808296e-05, "loss": 0.0938, "step": 12371 }, { "epoch": 2.27, "learning_rate": 1.6812332920513715e-05, "loss": 0.0947, "step": 12372 }, { "epoch": 2.27, "learning_rate": 1.68117893936694e-05, "loss": 0.0833, "step": 12373 }, { "epoch": 2.27, "learning_rate": 1.6811245829278336e-05, "loss": 0.0758, "step": 12374 }, { "epoch": 2.27, "learning_rate": 1.681070222734353e-05, "loss": 0.0876, "step": 12375 }, { "epoch": 2.27, "learning_rate": 1.6810158587867973e-05, "loss": 0.0816, "step": 12376 }, { "epoch": 2.27, "learning_rate": 1.680961491085466e-05, "loss": 0.08, "step": 12377 }, { "epoch": 2.27, "learning_rate": 1.6809071196306593e-05, "loss": 0.076, "step": 12378 }, { "epoch": 2.27, "learning_rate": 1.6808527444226765e-05, "loss": 0.0743, "step": 12379 }, { "epoch": 2.27, "learning_rate": 1.6807983654618174e-05, "loss": 0.0747, "step": 12380 }, { "epoch": 2.27, "learning_rate": 1.6807439827483818e-05, "loss": 0.0834, "step": 12381 }, { "epoch": 2.27, "learning_rate": 1.6806895962826694e-05, "loss": 0.0765, "step": 12382 }, { "epoch": 2.27, "learning_rate": 1.68063520606498e-05, "loss": 0.0731, "step": 12383 }, { "epoch": 2.27, "learning_rate": 1.6805808120956135e-05, "loss": 0.0809, "step": 12384 }, { "epoch": 2.27, "learning_rate": 1.6805264143748696e-05, "loss": 0.082, "step": 12385 }, { "epoch": 2.27, "learning_rate": 1.6804720129030477e-05, "loss": 0.089, "step": 12386 }, { "epoch": 2.27, "learning_rate": 1.6804176076804487e-05, "loss": 0.07, "step": 12387 }, { "epoch": 2.27, "learning_rate": 1.6803631987073715e-05, "loss": 0.0756, "step": 12388 }, { "epoch": 2.27, "learning_rate": 1.6803087859841166e-05, "loss": 0.0809, "step": 12389 }, { "epoch": 2.27, "learning_rate": 1.6802543695109838e-05, "loss": 0.0845, "step": 12390 }, { "epoch": 2.27, "learning_rate": 1.680199949288273e-05, "loss": 0.0793, "step": 12391 }, { "epoch": 2.27, "learning_rate": 1.680145525316284e-05, "loss": 0.0774, "step": 12392 }, { "epoch": 2.27, "learning_rate": 1.6800910975953172e-05, "loss": 0.0835, "step": 12393 }, { "epoch": 2.27, "learning_rate": 1.680036666125672e-05, "loss": 0.0768, "step": 12394 }, { "epoch": 2.27, "learning_rate": 1.679982230907649e-05, "loss": 0.0716, "step": 12395 }, { "epoch": 2.27, "learning_rate": 1.6799277919415478e-05, "loss": 0.076, "step": 12396 }, { "epoch": 2.27, "learning_rate": 1.679873349227669e-05, "loss": 0.0838, "step": 12397 }, { "epoch": 2.27, "learning_rate": 1.6798189027663118e-05, "loss": 0.0697, "step": 12398 }, { "epoch": 2.27, "learning_rate": 1.6797644525577774e-05, "loss": 0.0807, "step": 12399 }, { "epoch": 2.27, "learning_rate": 1.6797099986023654e-05, "loss": 0.081, "step": 12400 }, { "epoch": 2.27, "learning_rate": 1.6796555409003758e-05, "loss": 0.0678, "step": 12401 }, { "epoch": 2.27, "learning_rate": 1.6796010794521086e-05, "loss": 0.069, "step": 12402 }, { "epoch": 2.27, "learning_rate": 1.679546614257865e-05, "loss": 0.0796, "step": 12403 }, { "epoch": 2.27, "learning_rate": 1.6794921453179444e-05, "loss": 0.0782, "step": 12404 }, { "epoch": 2.27, "learning_rate": 1.6794376726326467e-05, "loss": 0.0688, "step": 12405 }, { "epoch": 2.28, "learning_rate": 1.6793831962022728e-05, "loss": 0.0716, "step": 12406 }, { "epoch": 2.28, "learning_rate": 1.679328716027123e-05, "loss": 0.076, "step": 12407 }, { "epoch": 2.28, "learning_rate": 1.6792742321074975e-05, "loss": 0.0875, "step": 12408 }, { "epoch": 2.28, "learning_rate": 1.6792197444436964e-05, "loss": 0.081, "step": 12409 }, { "epoch": 2.28, "learning_rate": 1.67916525303602e-05, "loss": 0.0819, "step": 12410 }, { "epoch": 2.28, "learning_rate": 1.6791107578847688e-05, "loss": 0.0922, "step": 12411 }, { "epoch": 2.28, "learning_rate": 1.6790562589902436e-05, "loss": 0.0735, "step": 12412 }, { "epoch": 2.28, "learning_rate": 1.6790017563527442e-05, "loss": 0.0836, "step": 12413 }, { "epoch": 2.28, "learning_rate": 1.6789472499725708e-05, "loss": 0.0695, "step": 12414 }, { "epoch": 2.28, "learning_rate": 1.6788927398500248e-05, "loss": 0.0933, "step": 12415 }, { "epoch": 2.28, "learning_rate": 1.6788382259854056e-05, "loss": 0.0863, "step": 12416 }, { "epoch": 2.28, "learning_rate": 1.678783708379015e-05, "loss": 0.0756, "step": 12417 }, { "epoch": 2.28, "learning_rate": 1.6787291870311516e-05, "loss": 0.0835, "step": 12418 }, { "epoch": 2.28, "learning_rate": 1.6786746619421177e-05, "loss": 0.0704, "step": 12419 }, { "epoch": 2.28, "learning_rate": 1.678620133112213e-05, "loss": 0.0764, "step": 12420 }, { "epoch": 2.28, "learning_rate": 1.6785656005417386e-05, "loss": 0.0723, "step": 12421 }, { "epoch": 2.28, "learning_rate": 1.6785110642309943e-05, "loss": 0.0828, "step": 12422 }, { "epoch": 2.28, "learning_rate": 1.6784565241802813e-05, "loss": 0.0757, "step": 12423 }, { "epoch": 2.28, "learning_rate": 1.6784019803899e-05, "loss": 0.0759, "step": 12424 }, { "epoch": 2.28, "learning_rate": 1.678347432860151e-05, "loss": 0.0826, "step": 12425 }, { "epoch": 2.28, "learning_rate": 1.678292881591335e-05, "loss": 0.0931, "step": 12426 }, { "epoch": 2.28, "learning_rate": 1.678238326583753e-05, "loss": 0.078, "step": 12427 }, { "epoch": 2.28, "learning_rate": 1.6781837678377053e-05, "loss": 0.0832, "step": 12428 }, { "epoch": 2.28, "learning_rate": 1.6781292053534928e-05, "loss": 0.0752, "step": 12429 }, { "epoch": 2.28, "learning_rate": 1.678074639131416e-05, "loss": 0.0773, "step": 12430 }, { "epoch": 2.28, "learning_rate": 1.6780200691717763e-05, "loss": 0.0789, "step": 12431 }, { "epoch": 2.28, "learning_rate": 1.677965495474874e-05, "loss": 0.0878, "step": 12432 }, { "epoch": 2.28, "learning_rate": 1.6779109180410098e-05, "loss": 0.0736, "step": 12433 }, { "epoch": 2.28, "learning_rate": 1.677856336870485e-05, "loss": 0.0811, "step": 12434 }, { "epoch": 2.28, "learning_rate": 1.6778017519636e-05, "loss": 0.0881, "step": 12435 }, { "epoch": 2.28, "learning_rate": 1.6777471633206558e-05, "loss": 0.0693, "step": 12436 }, { "epoch": 2.28, "learning_rate": 1.6776925709419537e-05, "loss": 0.0822, "step": 12437 }, { "epoch": 2.28, "learning_rate": 1.677637974827794e-05, "loss": 0.0752, "step": 12438 }, { "epoch": 2.28, "learning_rate": 1.677583374978478e-05, "loss": 0.077, "step": 12439 }, { "epoch": 2.28, "learning_rate": 1.6775287713943064e-05, "loss": 0.078, "step": 12440 }, { "epoch": 2.28, "learning_rate": 1.6774741640755806e-05, "loss": 0.0927, "step": 12441 }, { "epoch": 2.28, "learning_rate": 1.6774195530226012e-05, "loss": 0.089, "step": 12442 }, { "epoch": 2.28, "learning_rate": 1.6773649382356695e-05, "loss": 0.0764, "step": 12443 }, { "epoch": 2.28, "learning_rate": 1.677310319715086e-05, "loss": 0.0833, "step": 12444 }, { "epoch": 2.28, "learning_rate": 1.6772556974611527e-05, "loss": 0.0954, "step": 12445 }, { "epoch": 2.28, "learning_rate": 1.6772010714741697e-05, "loss": 0.0737, "step": 12446 }, { "epoch": 2.28, "learning_rate": 1.6771464417544385e-05, "loss": 0.0828, "step": 12447 }, { "epoch": 2.28, "learning_rate": 1.6770918083022607e-05, "loss": 0.0786, "step": 12448 }, { "epoch": 2.28, "learning_rate": 1.6770371711179364e-05, "loss": 0.0806, "step": 12449 }, { "epoch": 2.28, "learning_rate": 1.6769825302017676e-05, "loss": 0.0717, "step": 12450 }, { "epoch": 2.28, "learning_rate": 1.6769278855540555e-05, "loss": 0.0875, "step": 12451 }, { "epoch": 2.28, "learning_rate": 1.676873237175101e-05, "loss": 0.0825, "step": 12452 }, { "epoch": 2.28, "learning_rate": 1.676818585065205e-05, "loss": 0.0804, "step": 12453 }, { "epoch": 2.28, "learning_rate": 1.676763929224669e-05, "loss": 0.0806, "step": 12454 }, { "epoch": 2.28, "learning_rate": 1.6767092696537946e-05, "loss": 0.0806, "step": 12455 }, { "epoch": 2.28, "learning_rate": 1.6766546063528832e-05, "loss": 0.0849, "step": 12456 }, { "epoch": 2.28, "learning_rate": 1.6765999393222356e-05, "loss": 0.0706, "step": 12457 }, { "epoch": 2.28, "learning_rate": 1.676545268562153e-05, "loss": 0.0701, "step": 12458 }, { "epoch": 2.28, "learning_rate": 1.6764905940729373e-05, "loss": 0.0904, "step": 12459 }, { "epoch": 2.28, "learning_rate": 1.6764359158548894e-05, "loss": 0.0843, "step": 12460 }, { "epoch": 2.29, "learning_rate": 1.676381233908311e-05, "loss": 0.078, "step": 12461 }, { "epoch": 2.29, "learning_rate": 1.676326548233503e-05, "loss": 0.0691, "step": 12462 }, { "epoch": 2.29, "learning_rate": 1.676271858830768e-05, "loss": 0.081, "step": 12463 }, { "epoch": 2.29, "learning_rate": 1.676217165700406e-05, "loss": 0.0813, "step": 12464 }, { "epoch": 2.29, "learning_rate": 1.6761624688427195e-05, "loss": 0.0728, "step": 12465 }, { "epoch": 2.29, "learning_rate": 1.6761077682580093e-05, "loss": 0.0794, "step": 12466 }, { "epoch": 2.29, "learning_rate": 1.6760530639465774e-05, "loss": 0.0722, "step": 12467 }, { "epoch": 2.29, "learning_rate": 1.6759983559087257e-05, "loss": 0.072, "step": 12468 }, { "epoch": 2.29, "learning_rate": 1.6759436441447544e-05, "loss": 0.0722, "step": 12469 }, { "epoch": 2.29, "learning_rate": 1.6758889286549663e-05, "loss": 0.0829, "step": 12470 }, { "epoch": 2.29, "learning_rate": 1.6758342094396628e-05, "loss": 0.0921, "step": 12471 }, { "epoch": 2.29, "learning_rate": 1.675779486499145e-05, "loss": 0.0872, "step": 12472 }, { "epoch": 2.29, "learning_rate": 1.6757247598337147e-05, "loss": 0.0675, "step": 12473 }, { "epoch": 2.29, "learning_rate": 1.6756700294436738e-05, "loss": 0.0868, "step": 12474 }, { "epoch": 2.29, "learning_rate": 1.675615295329324e-05, "loss": 0.0798, "step": 12475 }, { "epoch": 2.29, "learning_rate": 1.6755605574909667e-05, "loss": 0.0679, "step": 12476 }, { "epoch": 2.29, "learning_rate": 1.675505815928904e-05, "loss": 0.0649, "step": 12477 }, { "epoch": 2.29, "learning_rate": 1.6754510706434373e-05, "loss": 0.0872, "step": 12478 }, { "epoch": 2.29, "learning_rate": 1.675396321634869e-05, "loss": 0.0774, "step": 12479 }, { "epoch": 2.29, "learning_rate": 1.6753415689034996e-05, "loss": 0.0882, "step": 12480 }, { "epoch": 2.29, "learning_rate": 1.6752868124496316e-05, "loss": 0.0805, "step": 12481 }, { "epoch": 2.29, "learning_rate": 1.6752320522735674e-05, "loss": 0.0706, "step": 12482 }, { "epoch": 2.29, "learning_rate": 1.675177288375608e-05, "loss": 0.0771, "step": 12483 }, { "epoch": 2.29, "learning_rate": 1.6751225207560553e-05, "loss": 0.0841, "step": 12484 }, { "epoch": 2.29, "learning_rate": 1.6750677494152122e-05, "loss": 0.0716, "step": 12485 }, { "epoch": 2.29, "learning_rate": 1.675012974353379e-05, "loss": 0.0891, "step": 12486 }, { "epoch": 2.29, "learning_rate": 1.674958195570859e-05, "loss": 0.0908, "step": 12487 }, { "epoch": 2.29, "learning_rate": 1.6749034130679536e-05, "loss": 0.0849, "step": 12488 }, { "epoch": 2.29, "learning_rate": 1.674848626844965e-05, "loss": 0.0824, "step": 12489 }, { "epoch": 2.29, "learning_rate": 1.6747938369021945e-05, "loss": 0.0704, "step": 12490 }, { "epoch": 2.29, "learning_rate": 1.6747390432399445e-05, "loss": 0.0911, "step": 12491 }, { "epoch": 2.29, "learning_rate": 1.6746842458585174e-05, "loss": 0.0827, "step": 12492 }, { "epoch": 2.29, "learning_rate": 1.6746294447582146e-05, "loss": 0.079, "step": 12493 }, { "epoch": 2.29, "learning_rate": 1.674574639939339e-05, "loss": 0.0696, "step": 12494 }, { "epoch": 2.29, "learning_rate": 1.674519831402192e-05, "loss": 0.0684, "step": 12495 }, { "epoch": 2.29, "learning_rate": 1.6744650191470757e-05, "loss": 0.0803, "step": 12496 }, { "epoch": 2.29, "learning_rate": 1.6744102031742922e-05, "loss": 0.0723, "step": 12497 }, { "epoch": 2.29, "learning_rate": 1.6743553834841444e-05, "loss": 0.0817, "step": 12498 }, { "epoch": 2.29, "learning_rate": 1.674300560076934e-05, "loss": 0.0795, "step": 12499 }, { "epoch": 2.29, "learning_rate": 1.6742457329529623e-05, "loss": 0.0689, "step": 12500 }, { "epoch": 2.29, "learning_rate": 1.674190902112533e-05, "loss": 0.0781, "step": 12501 }, { "epoch": 2.29, "learning_rate": 1.6741360675559475e-05, "loss": 0.0792, "step": 12502 }, { "epoch": 2.29, "learning_rate": 1.6740812292835082e-05, "loss": 0.0766, "step": 12503 }, { "epoch": 2.29, "learning_rate": 1.6740263872955173e-05, "loss": 0.0782, "step": 12504 }, { "epoch": 2.29, "learning_rate": 1.6739715415922772e-05, "loss": 0.0766, "step": 12505 }, { "epoch": 2.29, "learning_rate": 1.6739166921740905e-05, "loss": 0.0796, "step": 12506 }, { "epoch": 2.29, "learning_rate": 1.6738618390412588e-05, "loss": 0.0758, "step": 12507 }, { "epoch": 2.29, "learning_rate": 1.6738069821940846e-05, "loss": 0.0876, "step": 12508 }, { "epoch": 2.29, "learning_rate": 1.6737521216328712e-05, "loss": 0.0734, "step": 12509 }, { "epoch": 2.29, "learning_rate": 1.67369725735792e-05, "loss": 0.083, "step": 12510 }, { "epoch": 2.29, "learning_rate": 1.6736423893695336e-05, "loss": 0.0762, "step": 12511 }, { "epoch": 2.29, "learning_rate": 1.6735875176680148e-05, "loss": 0.0788, "step": 12512 }, { "epoch": 2.29, "learning_rate": 1.6735326422536656e-05, "loss": 0.0719, "step": 12513 }, { "epoch": 2.29, "learning_rate": 1.673477763126789e-05, "loss": 0.0881, "step": 12514 }, { "epoch": 2.3, "learning_rate": 1.673422880287687e-05, "loss": 0.0824, "step": 12515 }, { "epoch": 2.3, "learning_rate": 1.6733679937366624e-05, "loss": 0.0796, "step": 12516 }, { "epoch": 2.3, "learning_rate": 1.6733131034740174e-05, "loss": 0.0793, "step": 12517 }, { "epoch": 2.3, "learning_rate": 1.6732582095000553e-05, "loss": 0.0816, "step": 12518 }, { "epoch": 2.3, "learning_rate": 1.6732033118150775e-05, "loss": 0.0676, "step": 12519 }, { "epoch": 2.3, "learning_rate": 1.673148410419388e-05, "loss": 0.0854, "step": 12520 }, { "epoch": 2.3, "learning_rate": 1.673093505313288e-05, "loss": 0.0791, "step": 12521 }, { "epoch": 2.3, "learning_rate": 1.6730385964970814e-05, "loss": 0.0779, "step": 12522 }, { "epoch": 2.3, "learning_rate": 1.67298368397107e-05, "loss": 0.0796, "step": 12523 }, { "epoch": 2.3, "learning_rate": 1.6729287677355565e-05, "loss": 0.0767, "step": 12524 }, { "epoch": 2.3, "learning_rate": 1.6728738477908446e-05, "loss": 0.0815, "step": 12525 }, { "epoch": 2.3, "learning_rate": 1.6728189241372355e-05, "loss": 0.0871, "step": 12526 }, { "epoch": 2.3, "learning_rate": 1.672763996775033e-05, "loss": 0.0712, "step": 12527 }, { "epoch": 2.3, "learning_rate": 1.6727090657045396e-05, "loss": 0.0776, "step": 12528 }, { "epoch": 2.3, "learning_rate": 1.6726541309260584e-05, "loss": 0.0812, "step": 12529 }, { "epoch": 2.3, "learning_rate": 1.6725991924398918e-05, "loss": 0.0765, "step": 12530 }, { "epoch": 2.3, "learning_rate": 1.6725442502463423e-05, "loss": 0.0876, "step": 12531 }, { "epoch": 2.3, "learning_rate": 1.6724893043457134e-05, "loss": 0.0712, "step": 12532 }, { "epoch": 2.3, "learning_rate": 1.6724343547383075e-05, "loss": 0.0756, "step": 12533 }, { "epoch": 2.3, "learning_rate": 1.672379401424428e-05, "loss": 0.0856, "step": 12534 }, { "epoch": 2.3, "learning_rate": 1.672324444404377e-05, "loss": 0.0712, "step": 12535 }, { "epoch": 2.3, "learning_rate": 1.6722694836784582e-05, "loss": 0.0836, "step": 12536 }, { "epoch": 2.3, "learning_rate": 1.672214519246974e-05, "loss": 0.0828, "step": 12537 }, { "epoch": 2.3, "learning_rate": 1.6721595511102285e-05, "loss": 0.0853, "step": 12538 }, { "epoch": 2.3, "learning_rate": 1.672104579268523e-05, "loss": 0.0832, "step": 12539 }, { "epoch": 2.3, "learning_rate": 1.6720496037221615e-05, "loss": 0.0765, "step": 12540 }, { "epoch": 2.3, "learning_rate": 1.6719946244714467e-05, "loss": 0.0746, "step": 12541 }, { "epoch": 2.3, "learning_rate": 1.6719396415166823e-05, "loss": 0.087, "step": 12542 }, { "epoch": 2.3, "learning_rate": 1.6718846548581704e-05, "loss": 0.0695, "step": 12543 }, { "epoch": 2.3, "learning_rate": 1.6718296644962146e-05, "loss": 0.0682, "step": 12544 }, { "epoch": 2.3, "learning_rate": 1.671774670431118e-05, "loss": 0.0944, "step": 12545 }, { "epoch": 2.3, "learning_rate": 1.6717196726631837e-05, "loss": 0.0719, "step": 12546 }, { "epoch": 2.3, "learning_rate": 1.6716646711927148e-05, "loss": 0.079, "step": 12547 }, { "epoch": 2.3, "learning_rate": 1.6716096660200146e-05, "loss": 0.0783, "step": 12548 }, { "epoch": 2.3, "learning_rate": 1.671554657145386e-05, "loss": 0.0822, "step": 12549 }, { "epoch": 2.3, "learning_rate": 1.6714996445691326e-05, "loss": 0.0706, "step": 12550 }, { "epoch": 2.3, "learning_rate": 1.6714446282915574e-05, "loss": 0.0814, "step": 12551 }, { "epoch": 2.3, "learning_rate": 1.6713896083129635e-05, "loss": 0.0902, "step": 12552 }, { "epoch": 2.3, "learning_rate": 1.6713345846336545e-05, "loss": 0.0919, "step": 12553 }, { "epoch": 2.3, "learning_rate": 1.6712795572539334e-05, "loss": 0.0858, "step": 12554 }, { "epoch": 2.3, "learning_rate": 1.6712245261741037e-05, "loss": 0.0686, "step": 12555 }, { "epoch": 2.3, "learning_rate": 1.6711694913944686e-05, "loss": 0.0807, "step": 12556 }, { "epoch": 2.3, "learning_rate": 1.6711144529153317e-05, "loss": 0.077, "step": 12557 }, { "epoch": 2.3, "learning_rate": 1.6710594107369963e-05, "loss": 0.0862, "step": 12558 }, { "epoch": 2.3, "learning_rate": 1.6710043648597656e-05, "loss": 0.0647, "step": 12559 }, { "epoch": 2.3, "learning_rate": 1.670949315283943e-05, "loss": 0.0666, "step": 12560 }, { "epoch": 2.3, "learning_rate": 1.670894262009832e-05, "loss": 0.0718, "step": 12561 }, { "epoch": 2.3, "learning_rate": 1.6708392050377365e-05, "loss": 0.082, "step": 12562 }, { "epoch": 2.3, "learning_rate": 1.6707841443679593e-05, "loss": 0.0748, "step": 12563 }, { "epoch": 2.3, "learning_rate": 1.670729080000804e-05, "loss": 0.0799, "step": 12564 }, { "epoch": 2.3, "learning_rate": 1.6706740119365746e-05, "loss": 0.0867, "step": 12565 }, { "epoch": 2.3, "learning_rate": 1.6706189401755743e-05, "loss": 0.0741, "step": 12566 }, { "epoch": 2.3, "learning_rate": 1.6705638647181065e-05, "loss": 0.098, "step": 12567 }, { "epoch": 2.3, "learning_rate": 1.6705087855644753e-05, "loss": 0.0857, "step": 12568 }, { "epoch": 2.3, "learning_rate": 1.670453702714984e-05, "loss": 0.066, "step": 12569 }, { "epoch": 2.31, "learning_rate": 1.670398616169936e-05, "loss": 0.0761, "step": 12570 }, { "epoch": 2.31, "learning_rate": 1.670343525929635e-05, "loss": 0.0788, "step": 12571 }, { "epoch": 2.31, "learning_rate": 1.6702884319943845e-05, "loss": 0.0777, "step": 12572 }, { "epoch": 2.31, "learning_rate": 1.6702333343644893e-05, "loss": 0.0747, "step": 12573 }, { "epoch": 2.31, "learning_rate": 1.6701782330402515e-05, "loss": 0.0762, "step": 12574 }, { "epoch": 2.31, "learning_rate": 1.670123128021976e-05, "loss": 0.0727, "step": 12575 }, { "epoch": 2.31, "learning_rate": 1.670068019309966e-05, "loss": 0.0663, "step": 12576 }, { "epoch": 2.31, "learning_rate": 1.6700129069045253e-05, "loss": 0.0702, "step": 12577 }, { "epoch": 2.31, "learning_rate": 1.669957790805958e-05, "loss": 0.0789, "step": 12578 }, { "epoch": 2.31, "learning_rate": 1.6699026710145672e-05, "loss": 0.0732, "step": 12579 }, { "epoch": 2.31, "learning_rate": 1.6698475475306576e-05, "loss": 0.0767, "step": 12580 }, { "epoch": 2.31, "learning_rate": 1.6697924203545325e-05, "loss": 0.0803, "step": 12581 }, { "epoch": 2.31, "learning_rate": 1.669737289486496e-05, "loss": 0.0717, "step": 12582 }, { "epoch": 2.31, "learning_rate": 1.6696821549268516e-05, "loss": 0.0813, "step": 12583 }, { "epoch": 2.31, "learning_rate": 1.6696270166759038e-05, "loss": 0.0854, "step": 12584 }, { "epoch": 2.31, "learning_rate": 1.669571874733956e-05, "loss": 0.0856, "step": 12585 }, { "epoch": 2.31, "learning_rate": 1.6695167291013123e-05, "loss": 0.0849, "step": 12586 }, { "epoch": 2.31, "learning_rate": 1.6694615797782768e-05, "loss": 0.0687, "step": 12587 }, { "epoch": 2.31, "learning_rate": 1.6694064267651534e-05, "loss": 0.0831, "step": 12588 }, { "epoch": 2.31, "learning_rate": 1.669351270062246e-05, "loss": 0.0809, "step": 12589 }, { "epoch": 2.31, "learning_rate": 1.6692961096698587e-05, "loss": 0.0857, "step": 12590 }, { "epoch": 2.31, "learning_rate": 1.6692409455882955e-05, "loss": 0.073, "step": 12591 }, { "epoch": 2.31, "learning_rate": 1.669185777817861e-05, "loss": 0.083, "step": 12592 }, { "epoch": 2.31, "learning_rate": 1.6691306063588583e-05, "loss": 0.0786, "step": 12593 }, { "epoch": 2.31, "learning_rate": 1.669075431211592e-05, "loss": 0.0824, "step": 12594 }, { "epoch": 2.31, "learning_rate": 1.6690202523763667e-05, "loss": 0.0839, "step": 12595 }, { "epoch": 2.31, "learning_rate": 1.6689650698534858e-05, "loss": 0.0773, "step": 12596 }, { "epoch": 2.31, "learning_rate": 1.668909883643254e-05, "loss": 0.0925, "step": 12597 }, { "epoch": 2.31, "learning_rate": 1.6688546937459752e-05, "loss": 0.0862, "step": 12598 }, { "epoch": 2.31, "learning_rate": 1.6687995001619534e-05, "loss": 0.0786, "step": 12599 }, { "epoch": 2.31, "learning_rate": 1.668744302891493e-05, "loss": 0.0718, "step": 12600 }, { "epoch": 2.31, "learning_rate": 1.6686891019348987e-05, "loss": 0.0851, "step": 12601 }, { "epoch": 2.31, "learning_rate": 1.668633897292474e-05, "loss": 0.0761, "step": 12602 }, { "epoch": 2.31, "learning_rate": 1.668578688964524e-05, "loss": 0.0755, "step": 12603 }, { "epoch": 2.31, "learning_rate": 1.6685234769513522e-05, "loss": 0.0781, "step": 12604 }, { "epoch": 2.31, "learning_rate": 1.668468261253264e-05, "loss": 0.0848, "step": 12605 }, { "epoch": 2.31, "learning_rate": 1.6684130418705625e-05, "loss": 0.076, "step": 12606 }, { "epoch": 2.31, "learning_rate": 1.668357818803553e-05, "loss": 0.0686, "step": 12607 }, { "epoch": 2.31, "learning_rate": 1.668302592052539e-05, "loss": 0.0724, "step": 12608 }, { "epoch": 2.31, "learning_rate": 1.6682473616178262e-05, "loss": 0.0663, "step": 12609 }, { "epoch": 2.31, "learning_rate": 1.6681921274997176e-05, "loss": 0.0638, "step": 12610 }, { "epoch": 2.31, "learning_rate": 1.6681368896985186e-05, "loss": 0.07, "step": 12611 }, { "epoch": 2.31, "learning_rate": 1.6680816482145338e-05, "loss": 0.0749, "step": 12612 }, { "epoch": 2.31, "learning_rate": 1.668026403048067e-05, "loss": 0.0816, "step": 12613 }, { "epoch": 2.31, "learning_rate": 1.6679711541994227e-05, "loss": 0.0734, "step": 12614 }, { "epoch": 2.31, "learning_rate": 1.6679159016689062e-05, "loss": 0.0963, "step": 12615 }, { "epoch": 2.31, "learning_rate": 1.667860645456821e-05, "loss": 0.0838, "step": 12616 }, { "epoch": 2.31, "learning_rate": 1.667805385563473e-05, "loss": 0.0829, "step": 12617 }, { "epoch": 2.31, "learning_rate": 1.6677501219891657e-05, "loss": 0.0722, "step": 12618 }, { "epoch": 2.31, "learning_rate": 1.667694854734204e-05, "loss": 0.0793, "step": 12619 }, { "epoch": 2.31, "learning_rate": 1.6676395837988927e-05, "loss": 0.0725, "step": 12620 }, { "epoch": 2.31, "learning_rate": 1.6675843091835368e-05, "loss": 0.0722, "step": 12621 }, { "epoch": 2.31, "learning_rate": 1.66752903088844e-05, "loss": 0.0757, "step": 12622 }, { "epoch": 2.31, "learning_rate": 1.6674737489139077e-05, "loss": 0.0719, "step": 12623 }, { "epoch": 2.32, "learning_rate": 1.6674184632602447e-05, "loss": 0.0769, "step": 12624 }, { "epoch": 2.32, "learning_rate": 1.6673631739277553e-05, "loss": 0.0836, "step": 12625 }, { "epoch": 2.32, "learning_rate": 1.667307880916744e-05, "loss": 0.0722, "step": 12626 }, { "epoch": 2.32, "learning_rate": 1.667252584227517e-05, "loss": 0.0814, "step": 12627 }, { "epoch": 2.32, "learning_rate": 1.6671972838603772e-05, "loss": 0.0746, "step": 12628 }, { "epoch": 2.32, "learning_rate": 1.667141979815631e-05, "loss": 0.0753, "step": 12629 }, { "epoch": 2.32, "learning_rate": 1.667086672093582e-05, "loss": 0.0845, "step": 12630 }, { "epoch": 2.32, "learning_rate": 1.6670313606945362e-05, "loss": 0.0753, "step": 12631 }, { "epoch": 2.32, "learning_rate": 1.6669760456187972e-05, "loss": 0.0811, "step": 12632 }, { "epoch": 2.32, "learning_rate": 1.6669207268666717e-05, "loss": 0.0807, "step": 12633 }, { "epoch": 2.32, "learning_rate": 1.6668654044384624e-05, "loss": 0.072, "step": 12634 }, { "epoch": 2.32, "learning_rate": 1.666810078334476e-05, "loss": 0.0802, "step": 12635 }, { "epoch": 2.32, "learning_rate": 1.666754748555017e-05, "loss": 0.0838, "step": 12636 }, { "epoch": 2.32, "learning_rate": 1.6666994151003896e-05, "loss": 0.0952, "step": 12637 }, { "epoch": 2.32, "learning_rate": 1.6666440779709e-05, "loss": 0.0703, "step": 12638 }, { "epoch": 2.32, "learning_rate": 1.666588737166852e-05, "loss": 0.0749, "step": 12639 }, { "epoch": 2.32, "learning_rate": 1.6665333926885516e-05, "loss": 0.0776, "step": 12640 }, { "epoch": 2.32, "learning_rate": 1.6664780445363034e-05, "loss": 0.0729, "step": 12641 }, { "epoch": 2.32, "learning_rate": 1.666422692710413e-05, "loss": 0.0735, "step": 12642 }, { "epoch": 2.32, "learning_rate": 1.6663673372111847e-05, "loss": 0.0848, "step": 12643 }, { "epoch": 2.32, "learning_rate": 1.666311978038924e-05, "loss": 0.0782, "step": 12644 }, { "epoch": 2.32, "learning_rate": 1.6662566151939363e-05, "loss": 0.0775, "step": 12645 }, { "epoch": 2.32, "learning_rate": 1.6662012486765268e-05, "loss": 0.0723, "step": 12646 }, { "epoch": 2.32, "learning_rate": 1.6661458784869997e-05, "loss": 0.0738, "step": 12647 }, { "epoch": 2.32, "learning_rate": 1.6660905046256614e-05, "loss": 0.0837, "step": 12648 }, { "epoch": 2.32, "learning_rate": 1.6660351270928164e-05, "loss": 0.0857, "step": 12649 }, { "epoch": 2.32, "learning_rate": 1.6659797458887703e-05, "loss": 0.0771, "step": 12650 }, { "epoch": 2.32, "learning_rate": 1.665924361013828e-05, "loss": 0.0722, "step": 12651 }, { "epoch": 2.32, "learning_rate": 1.6658689724682955e-05, "loss": 0.089, "step": 12652 }, { "epoch": 2.32, "learning_rate": 1.6658135802524772e-05, "loss": 0.0729, "step": 12653 }, { "epoch": 2.32, "learning_rate": 1.6657581843666794e-05, "loss": 0.0832, "step": 12654 }, { "epoch": 2.32, "learning_rate": 1.6657027848112064e-05, "loss": 0.0847, "step": 12655 }, { "epoch": 2.32, "learning_rate": 1.6656473815863637e-05, "loss": 0.0815, "step": 12656 }, { "epoch": 2.32, "learning_rate": 1.665591974692458e-05, "loss": 0.0888, "step": 12657 }, { "epoch": 2.32, "learning_rate": 1.665536564129793e-05, "loss": 0.0857, "step": 12658 }, { "epoch": 2.32, "learning_rate": 1.6654811498986752e-05, "loss": 0.074, "step": 12659 }, { "epoch": 2.32, "learning_rate": 1.6654257319994095e-05, "loss": 0.0749, "step": 12660 }, { "epoch": 2.32, "learning_rate": 1.6653703104323015e-05, "loss": 0.0762, "step": 12661 }, { "epoch": 2.32, "learning_rate": 1.665314885197657e-05, "loss": 0.0795, "step": 12662 }, { "epoch": 2.32, "learning_rate": 1.6652594562957813e-05, "loss": 0.0775, "step": 12663 }, { "epoch": 2.32, "learning_rate": 1.6652040237269797e-05, "loss": 0.0836, "step": 12664 }, { "epoch": 2.32, "learning_rate": 1.665148587491558e-05, "loss": 0.0722, "step": 12665 }, { "epoch": 2.32, "learning_rate": 1.665093147589822e-05, "loss": 0.0875, "step": 12666 }, { "epoch": 2.32, "learning_rate": 1.6650377040220766e-05, "loss": 0.0639, "step": 12667 }, { "epoch": 2.32, "learning_rate": 1.664982256788628e-05, "loss": 0.0821, "step": 12668 }, { "epoch": 2.32, "learning_rate": 1.6649268058897815e-05, "loss": 0.0738, "step": 12669 }, { "epoch": 2.32, "learning_rate": 1.664871351325843e-05, "loss": 0.0741, "step": 12670 }, { "epoch": 2.32, "learning_rate": 1.664815893097118e-05, "loss": 0.0744, "step": 12671 }, { "epoch": 2.32, "learning_rate": 1.6647604312039122e-05, "loss": 0.0673, "step": 12672 }, { "epoch": 2.32, "learning_rate": 1.6647049656465312e-05, "loss": 0.0766, "step": 12673 }, { "epoch": 2.32, "learning_rate": 1.6646494964252808e-05, "loss": 0.0679, "step": 12674 }, { "epoch": 2.32, "learning_rate": 1.664594023540467e-05, "loss": 0.0749, "step": 12675 }, { "epoch": 2.32, "learning_rate": 1.6645385469923952e-05, "loss": 0.0744, "step": 12676 }, { "epoch": 2.32, "learning_rate": 1.6644830667813717e-05, "loss": 0.0811, "step": 12677 }, { "epoch": 2.32, "learning_rate": 1.6644275829077015e-05, "loss": 0.0758, "step": 12678 }, { "epoch": 2.33, "learning_rate": 1.6643720953716916e-05, "loss": 0.075, "step": 12679 }, { "epoch": 2.33, "learning_rate": 1.6643166041736468e-05, "loss": 0.0812, "step": 12680 }, { "epoch": 2.33, "learning_rate": 1.664261109313873e-05, "loss": 0.0788, "step": 12681 }, { "epoch": 2.33, "learning_rate": 1.6642056107926767e-05, "loss": 0.0705, "step": 12682 }, { "epoch": 2.33, "learning_rate": 1.6641501086103633e-05, "loss": 0.0849, "step": 12683 }, { "epoch": 2.33, "learning_rate": 1.6640946027672395e-05, "loss": 0.0799, "step": 12684 }, { "epoch": 2.33, "learning_rate": 1.66403909326361e-05, "loss": 0.0692, "step": 12685 }, { "epoch": 2.33, "learning_rate": 1.6639835800997816e-05, "loss": 0.0787, "step": 12686 }, { "epoch": 2.33, "learning_rate": 1.6639280632760605e-05, "loss": 0.0742, "step": 12687 }, { "epoch": 2.33, "learning_rate": 1.663872542792752e-05, "loss": 0.0713, "step": 12688 }, { "epoch": 2.33, "learning_rate": 1.663817018650163e-05, "loss": 0.0789, "step": 12689 }, { "epoch": 2.33, "learning_rate": 1.6637614908485983e-05, "loss": 0.0748, "step": 12690 }, { "epoch": 2.33, "learning_rate": 1.663705959388365e-05, "loss": 0.0818, "step": 12691 }, { "epoch": 2.33, "learning_rate": 1.6636504242697694e-05, "loss": 0.0764, "step": 12692 }, { "epoch": 2.33, "learning_rate": 1.6635948854931166e-05, "loss": 0.0785, "step": 12693 }, { "epoch": 2.33, "learning_rate": 1.6635393430587133e-05, "loss": 0.072, "step": 12694 }, { "epoch": 2.33, "learning_rate": 1.6634837969668654e-05, "loss": 0.084, "step": 12695 }, { "epoch": 2.33, "learning_rate": 1.6634282472178796e-05, "loss": 0.085, "step": 12696 }, { "epoch": 2.33, "learning_rate": 1.6633726938120618e-05, "loss": 0.0823, "step": 12697 }, { "epoch": 2.33, "learning_rate": 1.663317136749718e-05, "loss": 0.0749, "step": 12698 }, { "epoch": 2.33, "learning_rate": 1.6632615760311547e-05, "loss": 0.0705, "step": 12699 }, { "epoch": 2.33, "learning_rate": 1.6632060116566777e-05, "loss": 0.0835, "step": 12700 }, { "epoch": 2.33, "learning_rate": 1.663150443626594e-05, "loss": 0.081, "step": 12701 }, { "epoch": 2.33, "learning_rate": 1.6630948719412097e-05, "loss": 0.0825, "step": 12702 }, { "epoch": 2.33, "learning_rate": 1.6630392966008307e-05, "loss": 0.0735, "step": 12703 }, { "epoch": 2.33, "learning_rate": 1.6629837176057634e-05, "loss": 0.0784, "step": 12704 }, { "epoch": 2.33, "learning_rate": 1.6629281349563143e-05, "loss": 0.0774, "step": 12705 }, { "epoch": 2.33, "learning_rate": 1.6628725486527902e-05, "loss": 0.0778, "step": 12706 }, { "epoch": 2.33, "learning_rate": 1.6628169586954967e-05, "loss": 0.0816, "step": 12707 }, { "epoch": 2.33, "learning_rate": 1.662761365084741e-05, "loss": 0.0863, "step": 12708 }, { "epoch": 2.33, "learning_rate": 1.6627057678208288e-05, "loss": 0.091, "step": 12709 }, { "epoch": 2.33, "learning_rate": 1.662650166904067e-05, "loss": 0.073, "step": 12710 }, { "epoch": 2.33, "learning_rate": 1.662594562334762e-05, "loss": 0.0721, "step": 12711 }, { "epoch": 2.33, "learning_rate": 1.6625389541132197e-05, "loss": 0.0747, "step": 12712 }, { "epoch": 2.33, "learning_rate": 1.6624833422397476e-05, "loss": 0.0838, "step": 12713 }, { "epoch": 2.33, "learning_rate": 1.6624277267146517e-05, "loss": 0.0795, "step": 12714 }, { "epoch": 2.33, "learning_rate": 1.662372107538239e-05, "loss": 0.0812, "step": 12715 }, { "epoch": 2.33, "learning_rate": 1.6623164847108156e-05, "loss": 0.0696, "step": 12716 }, { "epoch": 2.33, "learning_rate": 1.6622608582326882e-05, "loss": 0.075, "step": 12717 }, { "epoch": 2.33, "learning_rate": 1.662205228104163e-05, "loss": 0.0682, "step": 12718 }, { "epoch": 2.33, "learning_rate": 1.6621495943255473e-05, "loss": 0.0731, "step": 12719 }, { "epoch": 2.33, "learning_rate": 1.662093956897148e-05, "loss": 0.0735, "step": 12720 }, { "epoch": 2.33, "learning_rate": 1.6620383158192708e-05, "loss": 0.0847, "step": 12721 }, { "epoch": 2.33, "learning_rate": 1.6619826710922228e-05, "loss": 0.0834, "step": 12722 }, { "epoch": 2.33, "learning_rate": 1.6619270227163113e-05, "loss": 0.0772, "step": 12723 }, { "epoch": 2.33, "learning_rate": 1.6618713706918424e-05, "loss": 0.0764, "step": 12724 }, { "epoch": 2.33, "learning_rate": 1.6618157150191225e-05, "loss": 0.0779, "step": 12725 }, { "epoch": 2.33, "learning_rate": 1.6617600556984595e-05, "loss": 0.0767, "step": 12726 }, { "epoch": 2.33, "learning_rate": 1.6617043927301592e-05, "loss": 0.0809, "step": 12727 }, { "epoch": 2.33, "learning_rate": 1.661648726114529e-05, "loss": 0.0729, "step": 12728 }, { "epoch": 2.33, "learning_rate": 1.6615930558518753e-05, "loss": 0.0729, "step": 12729 }, { "epoch": 2.33, "learning_rate": 1.6615373819425053e-05, "loss": 0.0819, "step": 12730 }, { "epoch": 2.33, "learning_rate": 1.6614817043867256e-05, "loss": 0.0852, "step": 12731 }, { "epoch": 2.33, "learning_rate": 1.6614260231848437e-05, "loss": 0.0833, "step": 12732 }, { "epoch": 2.34, "learning_rate": 1.6613703383371657e-05, "loss": 0.0899, "step": 12733 }, { "epoch": 2.34, "learning_rate": 1.6613146498439988e-05, "loss": 0.0811, "step": 12734 }, { "epoch": 2.34, "learning_rate": 1.6612589577056504e-05, "loss": 0.0812, "step": 12735 }, { "epoch": 2.34, "learning_rate": 1.661203261922427e-05, "loss": 0.0871, "step": 12736 }, { "epoch": 2.34, "learning_rate": 1.6611475624946357e-05, "loss": 0.0925, "step": 12737 }, { "epoch": 2.34, "learning_rate": 1.6610918594225833e-05, "loss": 0.0847, "step": 12738 }, { "epoch": 2.34, "learning_rate": 1.6610361527065773e-05, "loss": 0.0778, "step": 12739 }, { "epoch": 2.34, "learning_rate": 1.6609804423469245e-05, "loss": 0.0798, "step": 12740 }, { "epoch": 2.34, "learning_rate": 1.6609247283439324e-05, "loss": 0.0729, "step": 12741 }, { "epoch": 2.34, "learning_rate": 1.6608690106979074e-05, "loss": 0.0779, "step": 12742 }, { "epoch": 2.34, "learning_rate": 1.660813289409157e-05, "loss": 0.0811, "step": 12743 }, { "epoch": 2.34, "learning_rate": 1.660757564477988e-05, "loss": 0.0836, "step": 12744 }, { "epoch": 2.34, "learning_rate": 1.660701835904708e-05, "loss": 0.0772, "step": 12745 }, { "epoch": 2.34, "learning_rate": 1.660646103689624e-05, "loss": 0.0822, "step": 12746 }, { "epoch": 2.34, "learning_rate": 1.6605903678330433e-05, "loss": 0.0806, "step": 12747 }, { "epoch": 2.34, "learning_rate": 1.660534628335273e-05, "loss": 0.0685, "step": 12748 }, { "epoch": 2.34, "learning_rate": 1.66047888519662e-05, "loss": 0.0814, "step": 12749 }, { "epoch": 2.34, "learning_rate": 1.6604231384173922e-05, "loss": 0.0757, "step": 12750 }, { "epoch": 2.34, "learning_rate": 1.6603673879978966e-05, "loss": 0.0771, "step": 12751 }, { "epoch": 2.34, "learning_rate": 1.6603116339384405e-05, "loss": 0.0775, "step": 12752 }, { "epoch": 2.34, "learning_rate": 1.6602558762393313e-05, "loss": 0.0647, "step": 12753 }, { "epoch": 2.34, "learning_rate": 1.660200114900876e-05, "loss": 0.0764, "step": 12754 }, { "epoch": 2.34, "learning_rate": 1.6601443499233824e-05, "loss": 0.0785, "step": 12755 }, { "epoch": 2.34, "learning_rate": 1.6600885813071576e-05, "loss": 0.0768, "step": 12756 }, { "epoch": 2.34, "learning_rate": 1.660032809052509e-05, "loss": 0.071, "step": 12757 }, { "epoch": 2.34, "learning_rate": 1.659977033159744e-05, "loss": 0.077, "step": 12758 }, { "epoch": 2.34, "learning_rate": 1.6599212536291705e-05, "loss": 0.0781, "step": 12759 }, { "epoch": 2.34, "learning_rate": 1.6598654704610952e-05, "loss": 0.0746, "step": 12760 }, { "epoch": 2.34, "learning_rate": 1.6598096836558265e-05, "loss": 0.0775, "step": 12761 }, { "epoch": 2.34, "learning_rate": 1.6597538932136704e-05, "loss": 0.0908, "step": 12762 }, { "epoch": 2.34, "learning_rate": 1.6596980991349363e-05, "loss": 0.0762, "step": 12763 }, { "epoch": 2.34, "learning_rate": 1.6596423014199303e-05, "loss": 0.0796, "step": 12764 }, { "epoch": 2.34, "learning_rate": 1.659586500068961e-05, "loss": 0.0738, "step": 12765 }, { "epoch": 2.34, "learning_rate": 1.659530695082335e-05, "loss": 0.0873, "step": 12766 }, { "epoch": 2.34, "learning_rate": 1.6594748864603606e-05, "loss": 0.0724, "step": 12767 }, { "epoch": 2.34, "learning_rate": 1.6594190742033453e-05, "loss": 0.0838, "step": 12768 }, { "epoch": 2.34, "learning_rate": 1.659363258311596e-05, "loss": 0.0785, "step": 12769 }, { "epoch": 2.34, "learning_rate": 1.6593074387854216e-05, "loss": 0.0726, "step": 12770 }, { "epoch": 2.34, "learning_rate": 1.659251615625129e-05, "loss": 0.0817, "step": 12771 }, { "epoch": 2.34, "learning_rate": 1.659195788831026e-05, "loss": 0.0735, "step": 12772 }, { "epoch": 2.34, "learning_rate": 1.6591399584034203e-05, "loss": 0.0819, "step": 12773 }, { "epoch": 2.34, "learning_rate": 1.65908412434262e-05, "loss": 0.0872, "step": 12774 }, { "epoch": 2.34, "learning_rate": 1.659028286648932e-05, "loss": 0.0797, "step": 12775 }, { "epoch": 2.34, "learning_rate": 1.658972445322665e-05, "loss": 0.0778, "step": 12776 }, { "epoch": 2.34, "learning_rate": 1.6589166003641267e-05, "loss": 0.0765, "step": 12777 }, { "epoch": 2.34, "learning_rate": 1.6588607517736243e-05, "loss": 0.0859, "step": 12778 }, { "epoch": 2.34, "learning_rate": 1.6588048995514657e-05, "loss": 0.0787, "step": 12779 }, { "epoch": 2.34, "learning_rate": 1.6587490436979597e-05, "loss": 0.0688, "step": 12780 }, { "epoch": 2.34, "learning_rate": 1.6586931842134132e-05, "loss": 0.0882, "step": 12781 }, { "epoch": 2.34, "learning_rate": 1.6586373210981345e-05, "loss": 0.0773, "step": 12782 }, { "epoch": 2.34, "learning_rate": 1.6585814543524313e-05, "loss": 0.0758, "step": 12783 }, { "epoch": 2.34, "learning_rate": 1.658525583976612e-05, "loss": 0.0821, "step": 12784 }, { "epoch": 2.34, "learning_rate": 1.658469709970984e-05, "loss": 0.0629, "step": 12785 }, { "epoch": 2.34, "learning_rate": 1.6584138323358558e-05, "loss": 0.0814, "step": 12786 }, { "epoch": 2.34, "learning_rate": 1.6583579510715352e-05, "loss": 0.0889, "step": 12787 }, { "epoch": 2.35, "learning_rate": 1.6583020661783296e-05, "loss": 0.0857, "step": 12788 }, { "epoch": 2.35, "learning_rate": 1.658246177656548e-05, "loss": 0.0783, "step": 12789 }, { "epoch": 2.35, "learning_rate": 1.658190285506498e-05, "loss": 0.0721, "step": 12790 }, { "epoch": 2.35, "learning_rate": 1.6581343897284875e-05, "loss": 0.0748, "step": 12791 }, { "epoch": 2.35, "learning_rate": 1.6580784903228252e-05, "loss": 0.068, "step": 12792 }, { "epoch": 2.35, "learning_rate": 1.6580225872898185e-05, "loss": 0.0746, "step": 12793 }, { "epoch": 2.35, "learning_rate": 1.6579666806297764e-05, "loss": 0.0873, "step": 12794 }, { "epoch": 2.35, "learning_rate": 1.657910770343006e-05, "loss": 0.0892, "step": 12795 }, { "epoch": 2.35, "learning_rate": 1.657854856429816e-05, "loss": 0.0726, "step": 12796 }, { "epoch": 2.35, "learning_rate": 1.6577989388905148e-05, "loss": 0.0707, "step": 12797 }, { "epoch": 2.35, "learning_rate": 1.6577430177254106e-05, "loss": 0.0791, "step": 12798 }, { "epoch": 2.35, "learning_rate": 1.6576870929348116e-05, "loss": 0.0832, "step": 12799 }, { "epoch": 2.35, "learning_rate": 1.6576311645190254e-05, "loss": 0.0715, "step": 12800 }, { "epoch": 2.35, "learning_rate": 1.6575752324783613e-05, "loss": 0.0735, "step": 12801 }, { "epoch": 2.35, "learning_rate": 1.6575192968131267e-05, "loss": 0.0754, "step": 12802 }, { "epoch": 2.35, "learning_rate": 1.6574633575236306e-05, "loss": 0.0902, "step": 12803 }, { "epoch": 2.35, "learning_rate": 1.657407414610181e-05, "loss": 0.0853, "step": 12804 }, { "epoch": 2.35, "learning_rate": 1.6573514680730864e-05, "loss": 0.0798, "step": 12805 }, { "epoch": 2.35, "learning_rate": 1.6572955179126552e-05, "loss": 0.087, "step": 12806 }, { "epoch": 2.35, "learning_rate": 1.657239564129195e-05, "loss": 0.0725, "step": 12807 }, { "epoch": 2.35, "learning_rate": 1.6571836067230158e-05, "loss": 0.0759, "step": 12808 }, { "epoch": 2.35, "learning_rate": 1.657127645694425e-05, "loss": 0.0773, "step": 12809 }, { "epoch": 2.35, "learning_rate": 1.657071681043731e-05, "loss": 0.0758, "step": 12810 }, { "epoch": 2.35, "learning_rate": 1.6570157127712424e-05, "loss": 0.0816, "step": 12811 }, { "epoch": 2.35, "learning_rate": 1.656959740877268e-05, "loss": 0.0717, "step": 12812 }, { "epoch": 2.35, "learning_rate": 1.656903765362116e-05, "loss": 0.0819, "step": 12813 }, { "epoch": 2.35, "learning_rate": 1.6568477862260952e-05, "loss": 0.073, "step": 12814 }, { "epoch": 2.35, "learning_rate": 1.6567918034695137e-05, "loss": 0.0765, "step": 12815 }, { "epoch": 2.35, "learning_rate": 1.6567358170926812e-05, "loss": 0.0828, "step": 12816 }, { "epoch": 2.35, "learning_rate": 1.6566798270959046e-05, "loss": 0.0749, "step": 12817 }, { "epoch": 2.35, "learning_rate": 1.6566238334794937e-05, "loss": 0.0761, "step": 12818 }, { "epoch": 2.35, "learning_rate": 1.656567836243757e-05, "loss": 0.0775, "step": 12819 }, { "epoch": 2.35, "learning_rate": 1.6565118353890027e-05, "loss": 0.0717, "step": 12820 }, { "epoch": 2.35, "learning_rate": 1.6564558309155403e-05, "loss": 0.0629, "step": 12821 }, { "epoch": 2.35, "learning_rate": 1.6563998228236772e-05, "loss": 0.0834, "step": 12822 }, { "epoch": 2.35, "learning_rate": 1.6563438111137237e-05, "loss": 0.075, "step": 12823 }, { "epoch": 2.35, "learning_rate": 1.6562877957859874e-05, "loss": 0.0808, "step": 12824 }, { "epoch": 2.35, "learning_rate": 1.6562317768407773e-05, "loss": 0.0794, "step": 12825 }, { "epoch": 2.35, "learning_rate": 1.6561757542784022e-05, "loss": 0.0714, "step": 12826 }, { "epoch": 2.35, "learning_rate": 1.6561197280991713e-05, "loss": 0.0883, "step": 12827 }, { "epoch": 2.35, "learning_rate": 1.656063698303393e-05, "loss": 0.0825, "step": 12828 }, { "epoch": 2.35, "learning_rate": 1.6560076648913762e-05, "loss": 0.0668, "step": 12829 }, { "epoch": 2.35, "learning_rate": 1.6559516278634297e-05, "loss": 0.0757, "step": 12830 }, { "epoch": 2.35, "learning_rate": 1.6558955872198624e-05, "loss": 0.0783, "step": 12831 }, { "epoch": 2.35, "learning_rate": 1.6558395429609832e-05, "loss": 0.0787, "step": 12832 }, { "epoch": 2.35, "learning_rate": 1.6557834950871013e-05, "loss": 0.0743, "step": 12833 }, { "epoch": 2.35, "learning_rate": 1.6557274435985253e-05, "loss": 0.0767, "step": 12834 }, { "epoch": 2.35, "learning_rate": 1.655671388495564e-05, "loss": 0.082, "step": 12835 }, { "epoch": 2.35, "learning_rate": 1.6556153297785272e-05, "loss": 0.0747, "step": 12836 }, { "epoch": 2.35, "learning_rate": 1.655559267447723e-05, "loss": 0.0772, "step": 12837 }, { "epoch": 2.35, "learning_rate": 1.6555032015034612e-05, "loss": 0.0791, "step": 12838 }, { "epoch": 2.35, "learning_rate": 1.6554471319460498e-05, "loss": 0.0821, "step": 12839 }, { "epoch": 2.35, "learning_rate": 1.6553910587757986e-05, "loss": 0.0777, "step": 12840 }, { "epoch": 2.35, "learning_rate": 1.6553349819930167e-05, "loss": 0.0819, "step": 12841 }, { "epoch": 2.36, "learning_rate": 1.655278901598013e-05, "loss": 0.0883, "step": 12842 }, { "epoch": 2.36, "learning_rate": 1.6552228175910966e-05, "loss": 0.0712, "step": 12843 }, { "epoch": 2.36, "learning_rate": 1.6551667299725765e-05, "loss": 0.0778, "step": 12844 }, { "epoch": 2.36, "learning_rate": 1.6551106387427623e-05, "loss": 0.0763, "step": 12845 }, { "epoch": 2.36, "learning_rate": 1.655054543901963e-05, "loss": 0.0725, "step": 12846 }, { "epoch": 2.36, "learning_rate": 1.654998445450487e-05, "loss": 0.0845, "step": 12847 }, { "epoch": 2.36, "learning_rate": 1.6549423433886448e-05, "loss": 0.0714, "step": 12848 }, { "epoch": 2.36, "learning_rate": 1.6548862377167454e-05, "loss": 0.0838, "step": 12849 }, { "epoch": 2.36, "learning_rate": 1.654830128435097e-05, "loss": 0.0846, "step": 12850 }, { "epoch": 2.36, "learning_rate": 1.65477401554401e-05, "loss": 0.08, "step": 12851 }, { "epoch": 2.36, "learning_rate": 1.6547178990437932e-05, "loss": 0.0733, "step": 12852 }, { "epoch": 2.36, "learning_rate": 1.6546617789347562e-05, "loss": 0.079, "step": 12853 }, { "epoch": 2.36, "learning_rate": 1.6546056552172078e-05, "loss": 0.0845, "step": 12854 }, { "epoch": 2.36, "learning_rate": 1.654549527891458e-05, "loss": 0.0696, "step": 12855 }, { "epoch": 2.36, "learning_rate": 1.6544933969578156e-05, "loss": 0.0692, "step": 12856 }, { "epoch": 2.36, "learning_rate": 1.6544372624165904e-05, "loss": 0.0738, "step": 12857 }, { "epoch": 2.36, "learning_rate": 1.654381124268092e-05, "loss": 0.0744, "step": 12858 }, { "epoch": 2.36, "learning_rate": 1.6543249825126285e-05, "loss": 0.0773, "step": 12859 }, { "epoch": 2.36, "learning_rate": 1.6542688371505114e-05, "loss": 0.0864, "step": 12860 }, { "epoch": 2.36, "learning_rate": 1.6542126881820485e-05, "loss": 0.0782, "step": 12861 }, { "epoch": 2.36, "learning_rate": 1.6541565356075506e-05, "loss": 0.0788, "step": 12862 }, { "epoch": 2.36, "learning_rate": 1.654100379427326e-05, "loss": 0.0779, "step": 12863 }, { "epoch": 2.36, "learning_rate": 1.6540442196416847e-05, "loss": 0.0779, "step": 12864 }, { "epoch": 2.36, "learning_rate": 1.653988056250937e-05, "loss": 0.0832, "step": 12865 }, { "epoch": 2.36, "learning_rate": 1.653931889255391e-05, "loss": 0.0817, "step": 12866 }, { "epoch": 2.36, "learning_rate": 1.653875718655358e-05, "loss": 0.0751, "step": 12867 }, { "epoch": 2.36, "learning_rate": 1.6538195444511456e-05, "loss": 0.0888, "step": 12868 }, { "epoch": 2.36, "learning_rate": 1.6537633666430655e-05, "loss": 0.0842, "step": 12869 }, { "epoch": 2.36, "learning_rate": 1.6537071852314257e-05, "loss": 0.0803, "step": 12870 }, { "epoch": 2.36, "learning_rate": 1.6536510002165373e-05, "loss": 0.0936, "step": 12871 }, { "epoch": 2.36, "learning_rate": 1.6535948115987088e-05, "loss": 0.0724, "step": 12872 }, { "epoch": 2.36, "learning_rate": 1.6535386193782506e-05, "loss": 0.0711, "step": 12873 }, { "epoch": 2.36, "learning_rate": 1.653482423555472e-05, "loss": 0.0823, "step": 12874 }, { "epoch": 2.36, "learning_rate": 1.653426224130683e-05, "loss": 0.0703, "step": 12875 }, { "epoch": 2.36, "learning_rate": 1.6533700211041936e-05, "loss": 0.0703, "step": 12876 }, { "epoch": 2.36, "learning_rate": 1.653313814476313e-05, "loss": 0.0814, "step": 12877 }, { "epoch": 2.36, "learning_rate": 1.6532576042473518e-05, "loss": 0.0695, "step": 12878 }, { "epoch": 2.36, "learning_rate": 1.653201390417619e-05, "loss": 0.0755, "step": 12879 }, { "epoch": 2.36, "learning_rate": 1.653145172987425e-05, "loss": 0.0729, "step": 12880 }, { "epoch": 2.36, "learning_rate": 1.6530889519570794e-05, "loss": 0.0706, "step": 12881 }, { "epoch": 2.36, "learning_rate": 1.6530327273268923e-05, "loss": 0.0873, "step": 12882 }, { "epoch": 2.36, "learning_rate": 1.6529764990971736e-05, "loss": 0.082, "step": 12883 }, { "epoch": 2.36, "learning_rate": 1.6529202672682328e-05, "loss": 0.0769, "step": 12884 }, { "epoch": 2.36, "learning_rate": 1.6528640318403804e-05, "loss": 0.0685, "step": 12885 }, { "epoch": 2.36, "learning_rate": 1.652807792813926e-05, "loss": 0.0778, "step": 12886 }, { "epoch": 2.36, "learning_rate": 1.65275155018918e-05, "loss": 0.0732, "step": 12887 }, { "epoch": 2.36, "learning_rate": 1.652695303966452e-05, "loss": 0.0776, "step": 12888 }, { "epoch": 2.36, "learning_rate": 1.6526390541460524e-05, "loss": 0.0713, "step": 12889 }, { "epoch": 2.36, "learning_rate": 1.6525828007282908e-05, "loss": 0.0786, "step": 12890 }, { "epoch": 2.36, "learning_rate": 1.6525265437134778e-05, "loss": 0.0755, "step": 12891 }, { "epoch": 2.36, "learning_rate": 1.6524702831019227e-05, "loss": 0.0812, "step": 12892 }, { "epoch": 2.36, "learning_rate": 1.6524140188939365e-05, "loss": 0.0748, "step": 12893 }, { "epoch": 2.36, "learning_rate": 1.652357751089829e-05, "loss": 0.0714, "step": 12894 }, { "epoch": 2.36, "learning_rate": 1.6523014796899103e-05, "loss": 0.0811, "step": 12895 }, { "epoch": 2.36, "learning_rate": 1.6522452046944902e-05, "loss": 0.0765, "step": 12896 }, { "epoch": 2.37, "learning_rate": 1.65218892610388e-05, "loss": 0.0853, "step": 12897 }, { "epoch": 2.37, "learning_rate": 1.652132643918388e-05, "loss": 0.0837, "step": 12898 }, { "epoch": 2.37, "learning_rate": 1.6520763581383264e-05, "loss": 0.0868, "step": 12899 }, { "epoch": 2.37, "learning_rate": 1.6520200687640047e-05, "loss": 0.0699, "step": 12900 }, { "epoch": 2.37, "learning_rate": 1.6519637757957324e-05, "loss": 0.0777, "step": 12901 }, { "epoch": 2.37, "learning_rate": 1.6519074792338212e-05, "loss": 0.0832, "step": 12902 }, { "epoch": 2.37, "learning_rate": 1.6518511790785804e-05, "loss": 0.0873, "step": 12903 }, { "epoch": 2.37, "learning_rate": 1.651794875330321e-05, "loss": 0.0783, "step": 12904 }, { "epoch": 2.37, "learning_rate": 1.6517385679893524e-05, "loss": 0.0797, "step": 12905 }, { "epoch": 2.37, "learning_rate": 1.651682257055986e-05, "loss": 0.07, "step": 12906 }, { "epoch": 2.37, "learning_rate": 1.6516259425305317e-05, "loss": 0.0736, "step": 12907 }, { "epoch": 2.37, "learning_rate": 1.6515696244132996e-05, "loss": 0.0752, "step": 12908 }, { "epoch": 2.37, "learning_rate": 1.6515133027046007e-05, "loss": 0.0732, "step": 12909 }, { "epoch": 2.37, "learning_rate": 1.651456977404745e-05, "loss": 0.0854, "step": 12910 }, { "epoch": 2.37, "learning_rate": 1.6514006485140436e-05, "loss": 0.0826, "step": 12911 }, { "epoch": 2.37, "learning_rate": 1.651344316032806e-05, "loss": 0.0653, "step": 12912 }, { "epoch": 2.37, "learning_rate": 1.6512879799613436e-05, "loss": 0.0872, "step": 12913 }, { "epoch": 2.37, "learning_rate": 1.6512316402999667e-05, "loss": 0.0752, "step": 12914 }, { "epoch": 2.37, "learning_rate": 1.6511752970489855e-05, "loss": 0.0822, "step": 12915 }, { "epoch": 2.37, "learning_rate": 1.6511189502087113e-05, "loss": 0.065, "step": 12916 }, { "epoch": 2.37, "learning_rate": 1.6510625997794535e-05, "loss": 0.081, "step": 12917 }, { "epoch": 2.37, "learning_rate": 1.6510062457615237e-05, "loss": 0.0742, "step": 12918 }, { "epoch": 2.37, "learning_rate": 1.6509498881552324e-05, "loss": 0.0815, "step": 12919 }, { "epoch": 2.37, "learning_rate": 1.65089352696089e-05, "loss": 0.0865, "step": 12920 }, { "epoch": 2.37, "learning_rate": 1.6508371621788072e-05, "loss": 0.0831, "step": 12921 }, { "epoch": 2.37, "learning_rate": 1.6507807938092947e-05, "loss": 0.0699, "step": 12922 }, { "epoch": 2.37, "learning_rate": 1.650724421852663e-05, "loss": 0.0807, "step": 12923 }, { "epoch": 2.37, "learning_rate": 1.6506680463092232e-05, "loss": 0.0916, "step": 12924 }, { "epoch": 2.37, "learning_rate": 1.6506116671792857e-05, "loss": 0.0761, "step": 12925 }, { "epoch": 2.37, "learning_rate": 1.6505552844631618e-05, "loss": 0.0923, "step": 12926 }, { "epoch": 2.37, "learning_rate": 1.6504988981611618e-05, "loss": 0.0821, "step": 12927 }, { "epoch": 2.37, "learning_rate": 1.6504425082735965e-05, "loss": 0.0798, "step": 12928 }, { "epoch": 2.37, "learning_rate": 1.6503861148007772e-05, "loss": 0.0707, "step": 12929 }, { "epoch": 2.37, "learning_rate": 1.650329717743014e-05, "loss": 0.0887, "step": 12930 }, { "epoch": 2.37, "learning_rate": 1.6502733171006185e-05, "loss": 0.0805, "step": 12931 }, { "epoch": 2.37, "learning_rate": 1.650216912873901e-05, "loss": 0.0695, "step": 12932 }, { "epoch": 2.37, "learning_rate": 1.6501605050631725e-05, "loss": 0.084, "step": 12933 }, { "epoch": 2.37, "learning_rate": 1.6501040936687444e-05, "loss": 0.0846, "step": 12934 }, { "epoch": 2.37, "learning_rate": 1.650047678690927e-05, "loss": 0.0828, "step": 12935 }, { "epoch": 2.37, "learning_rate": 1.6499912601300318e-05, "loss": 0.0776, "step": 12936 }, { "epoch": 2.37, "learning_rate": 1.6499348379863697e-05, "loss": 0.079, "step": 12937 }, { "epoch": 2.37, "learning_rate": 1.649878412260251e-05, "loss": 0.0839, "step": 12938 }, { "epoch": 2.37, "learning_rate": 1.6498219829519876e-05, "loss": 0.0804, "step": 12939 }, { "epoch": 2.37, "learning_rate": 1.6497655500618904e-05, "loss": 0.0755, "step": 12940 }, { "epoch": 2.37, "learning_rate": 1.6497091135902703e-05, "loss": 0.0799, "step": 12941 }, { "epoch": 2.37, "learning_rate": 1.6496526735374382e-05, "loss": 0.0789, "step": 12942 }, { "epoch": 2.37, "learning_rate": 1.649596229903705e-05, "loss": 0.0769, "step": 12943 }, { "epoch": 2.37, "learning_rate": 1.6495397826893827e-05, "loss": 0.085, "step": 12944 }, { "epoch": 2.37, "learning_rate": 1.649483331894781e-05, "loss": 0.0691, "step": 12945 }, { "epoch": 2.37, "learning_rate": 1.6494268775202132e-05, "loss": 0.0736, "step": 12946 }, { "epoch": 2.37, "learning_rate": 1.6493704195659886e-05, "loss": 0.0806, "step": 12947 }, { "epoch": 2.37, "learning_rate": 1.649313958032419e-05, "loss": 0.0904, "step": 12948 }, { "epoch": 2.37, "learning_rate": 1.6492574929198156e-05, "loss": 0.0751, "step": 12949 }, { "epoch": 2.37, "learning_rate": 1.6492010242284896e-05, "loss": 0.0789, "step": 12950 }, { "epoch": 2.38, "learning_rate": 1.649144551958753e-05, "loss": 0.081, "step": 12951 }, { "epoch": 2.38, "learning_rate": 1.649088076110916e-05, "loss": 0.0782, "step": 12952 }, { "epoch": 2.38, "learning_rate": 1.64903159668529e-05, "loss": 0.074, "step": 12953 }, { "epoch": 2.38, "learning_rate": 1.648975113682187e-05, "loss": 0.0709, "step": 12954 }, { "epoch": 2.38, "learning_rate": 1.648918627101918e-05, "loss": 0.0854, "step": 12955 }, { "epoch": 2.38, "learning_rate": 1.648862136944794e-05, "loss": 0.0704, "step": 12956 }, { "epoch": 2.38, "learning_rate": 1.648805643211127e-05, "loss": 0.0757, "step": 12957 }, { "epoch": 2.38, "learning_rate": 1.648749145901228e-05, "loss": 0.0704, "step": 12958 }, { "epoch": 2.38, "learning_rate": 1.6486926450154085e-05, "loss": 0.0716, "step": 12959 }, { "epoch": 2.38, "learning_rate": 1.64863614055398e-05, "loss": 0.0824, "step": 12960 }, { "epoch": 2.38, "learning_rate": 1.6485796325172536e-05, "loss": 0.0776, "step": 12961 }, { "epoch": 2.38, "learning_rate": 1.6485231209055416e-05, "loss": 0.0867, "step": 12962 }, { "epoch": 2.38, "learning_rate": 1.6484666057191545e-05, "loss": 0.0793, "step": 12963 }, { "epoch": 2.38, "learning_rate": 1.6484100869584044e-05, "loss": 0.059, "step": 12964 }, { "epoch": 2.38, "learning_rate": 1.648353564623603e-05, "loss": 0.0797, "step": 12965 }, { "epoch": 2.38, "learning_rate": 1.648297038715061e-05, "loss": 0.0718, "step": 12966 }, { "epoch": 2.38, "learning_rate": 1.648240509233091e-05, "loss": 0.0609, "step": 12967 }, { "epoch": 2.38, "learning_rate": 1.6481839761780042e-05, "loss": 0.0832, "step": 12968 }, { "epoch": 2.38, "learning_rate": 1.648127439550112e-05, "loss": 0.0706, "step": 12969 }, { "epoch": 2.38, "learning_rate": 1.648070899349726e-05, "loss": 0.0766, "step": 12970 }, { "epoch": 2.38, "learning_rate": 1.648014355577158e-05, "loss": 0.0876, "step": 12971 }, { "epoch": 2.38, "learning_rate": 1.6479578082327198e-05, "loss": 0.082, "step": 12972 }, { "epoch": 2.38, "learning_rate": 1.647901257316723e-05, "loss": 0.079, "step": 12973 }, { "epoch": 2.38, "learning_rate": 1.6478447028294793e-05, "loss": 0.0779, "step": 12974 }, { "epoch": 2.38, "learning_rate": 1.6477881447713008e-05, "loss": 0.0885, "step": 12975 }, { "epoch": 2.38, "learning_rate": 1.6477315831424983e-05, "loss": 0.0791, "step": 12976 }, { "epoch": 2.38, "learning_rate": 1.647675017943384e-05, "loss": 0.0645, "step": 12977 }, { "epoch": 2.38, "learning_rate": 1.6476184491742704e-05, "loss": 0.0719, "step": 12978 }, { "epoch": 2.38, "learning_rate": 1.6475618768354684e-05, "loss": 0.0771, "step": 12979 }, { "epoch": 2.38, "learning_rate": 1.6475053009272906e-05, "loss": 0.0791, "step": 12980 }, { "epoch": 2.38, "learning_rate": 1.647448721450048e-05, "loss": 0.0835, "step": 12981 }, { "epoch": 2.38, "learning_rate": 1.647392138404053e-05, "loss": 0.0771, "step": 12982 }, { "epoch": 2.38, "learning_rate": 1.647335551789617e-05, "loss": 0.0693, "step": 12983 }, { "epoch": 2.38, "learning_rate": 1.6472789616070528e-05, "loss": 0.0786, "step": 12984 }, { "epoch": 2.38, "learning_rate": 1.647222367856671e-05, "loss": 0.0702, "step": 12985 }, { "epoch": 2.38, "learning_rate": 1.6471657705387855e-05, "loss": 0.0856, "step": 12986 }, { "epoch": 2.38, "learning_rate": 1.6471091696537064e-05, "loss": 0.0893, "step": 12987 }, { "epoch": 2.38, "learning_rate": 1.6470525652017465e-05, "loss": 0.0681, "step": 12988 }, { "epoch": 2.38, "learning_rate": 1.6469959571832178e-05, "loss": 0.09, "step": 12989 }, { "epoch": 2.38, "learning_rate": 1.646939345598432e-05, "loss": 0.0781, "step": 12990 }, { "epoch": 2.38, "learning_rate": 1.646882730447702e-05, "loss": 0.0826, "step": 12991 }, { "epoch": 2.38, "learning_rate": 1.6468261117313385e-05, "loss": 0.0768, "step": 12992 }, { "epoch": 2.38, "learning_rate": 1.6467694894496545e-05, "loss": 0.0856, "step": 12993 }, { "epoch": 2.38, "learning_rate": 1.646712863602962e-05, "loss": 0.0698, "step": 12994 }, { "epoch": 2.38, "learning_rate": 1.646656234191573e-05, "loss": 0.0721, "step": 12995 }, { "epoch": 2.38, "learning_rate": 1.6465996012157996e-05, "loss": 0.0627, "step": 12996 }, { "epoch": 2.38, "learning_rate": 1.6465429646759545e-05, "loss": 0.0781, "step": 12997 }, { "epoch": 2.38, "learning_rate": 1.646486324572349e-05, "loss": 0.0821, "step": 12998 }, { "epoch": 2.38, "learning_rate": 1.646429680905296e-05, "loss": 0.0732, "step": 12999 }, { "epoch": 2.38, "learning_rate": 1.6463730336751072e-05, "loss": 0.0787, "step": 13000 }, { "epoch": 2.38, "learning_rate": 1.646316382882095e-05, "loss": 0.0749, "step": 13001 }, { "epoch": 2.38, "learning_rate": 1.6462597285265723e-05, "loss": 0.0804, "step": 13002 }, { "epoch": 2.38, "learning_rate": 1.6462030706088503e-05, "loss": 0.0768, "step": 13003 }, { "epoch": 2.38, "learning_rate": 1.646146409129242e-05, "loss": 0.0671, "step": 13004 }, { "epoch": 2.38, "learning_rate": 1.6460897440880597e-05, "loss": 0.0743, "step": 13005 }, { "epoch": 2.39, "learning_rate": 1.6460330754856156e-05, "loss": 0.0743, "step": 13006 }, { "epoch": 2.39, "learning_rate": 1.6459764033222217e-05, "loss": 0.0862, "step": 13007 }, { "epoch": 2.39, "learning_rate": 1.645919727598191e-05, "loss": 0.0785, "step": 13008 }, { "epoch": 2.39, "learning_rate": 1.6458630483138354e-05, "loss": 0.0694, "step": 13009 }, { "epoch": 2.39, "learning_rate": 1.6458063654694682e-05, "loss": 0.0724, "step": 13010 }, { "epoch": 2.39, "learning_rate": 1.6457496790654006e-05, "loss": 0.075, "step": 13011 }, { "epoch": 2.39, "learning_rate": 1.6456929891019457e-05, "loss": 0.0703, "step": 13012 }, { "epoch": 2.39, "learning_rate": 1.645636295579416e-05, "loss": 0.0719, "step": 13013 }, { "epoch": 2.39, "learning_rate": 1.6455795984981237e-05, "loss": 0.0757, "step": 13014 }, { "epoch": 2.39, "learning_rate": 1.645522897858382e-05, "loss": 0.0684, "step": 13015 }, { "epoch": 2.39, "learning_rate": 1.6454661936605025e-05, "loss": 0.0738, "step": 13016 }, { "epoch": 2.39, "learning_rate": 1.6454094859047987e-05, "loss": 0.0731, "step": 13017 }, { "epoch": 2.39, "learning_rate": 1.6453527745915822e-05, "loss": 0.0667, "step": 13018 }, { "epoch": 2.39, "learning_rate": 1.6452960597211666e-05, "loss": 0.0779, "step": 13019 }, { "epoch": 2.39, "learning_rate": 1.645239341293863e-05, "loss": 0.071, "step": 13020 }, { "epoch": 2.39, "learning_rate": 1.645182619309986e-05, "loss": 0.0802, "step": 13021 }, { "epoch": 2.39, "learning_rate": 1.6451258937698474e-05, "loss": 0.0719, "step": 13022 }, { "epoch": 2.39, "learning_rate": 1.645069164673759e-05, "loss": 0.0829, "step": 13023 }, { "epoch": 2.39, "learning_rate": 1.6450124320220348e-05, "loss": 0.0798, "step": 13024 }, { "epoch": 2.39, "learning_rate": 1.6449556958149865e-05, "loss": 0.0744, "step": 13025 }, { "epoch": 2.39, "learning_rate": 1.6448989560529282e-05, "loss": 0.0748, "step": 13026 }, { "epoch": 2.39, "learning_rate": 1.6448422127361707e-05, "loss": 0.0741, "step": 13027 }, { "epoch": 2.39, "learning_rate": 1.6447854658650282e-05, "loss": 0.0793, "step": 13028 }, { "epoch": 2.39, "learning_rate": 1.6447287154398133e-05, "loss": 0.079, "step": 13029 }, { "epoch": 2.39, "learning_rate": 1.6446719614608385e-05, "loss": 0.0799, "step": 13030 }, { "epoch": 2.39, "learning_rate": 1.6446152039284166e-05, "loss": 0.0773, "step": 13031 }, { "epoch": 2.39, "learning_rate": 1.6445584428428608e-05, "loss": 0.0687, "step": 13032 }, { "epoch": 2.39, "learning_rate": 1.644501678204484e-05, "loss": 0.0665, "step": 13033 }, { "epoch": 2.39, "learning_rate": 1.644444910013598e-05, "loss": 0.0823, "step": 13034 }, { "epoch": 2.39, "learning_rate": 1.6443881382705174e-05, "loss": 0.0733, "step": 13035 }, { "epoch": 2.39, "learning_rate": 1.644331362975554e-05, "loss": 0.0638, "step": 13036 }, { "epoch": 2.39, "learning_rate": 1.6442745841290207e-05, "loss": 0.0678, "step": 13037 }, { "epoch": 2.39, "learning_rate": 1.6442178017312313e-05, "loss": 0.0753, "step": 13038 }, { "epoch": 2.39, "learning_rate": 1.6441610157824982e-05, "loss": 0.0752, "step": 13039 }, { "epoch": 2.39, "learning_rate": 1.644104226283134e-05, "loss": 0.0826, "step": 13040 }, { "epoch": 2.39, "learning_rate": 1.6440474332334527e-05, "loss": 0.088, "step": 13041 }, { "epoch": 2.39, "learning_rate": 1.643990636633767e-05, "loss": 0.0876, "step": 13042 }, { "epoch": 2.39, "learning_rate": 1.643933836484389e-05, "loss": 0.0818, "step": 13043 }, { "epoch": 2.39, "learning_rate": 1.6438770327856337e-05, "loss": 0.0812, "step": 13044 }, { "epoch": 2.39, "learning_rate": 1.6438202255378123e-05, "loss": 0.0796, "step": 13045 }, { "epoch": 2.39, "learning_rate": 1.6437634147412393e-05, "loss": 0.077, "step": 13046 }, { "epoch": 2.39, "learning_rate": 1.6437066003962267e-05, "loss": 0.0709, "step": 13047 }, { "epoch": 2.39, "learning_rate": 1.6436497825030886e-05, "loss": 0.0953, "step": 13048 }, { "epoch": 2.39, "learning_rate": 1.643592961062138e-05, "loss": 0.0804, "step": 13049 }, { "epoch": 2.39, "learning_rate": 1.6435361360736877e-05, "loss": 0.0832, "step": 13050 }, { "epoch": 2.39, "learning_rate": 1.643479307538051e-05, "loss": 0.0824, "step": 13051 }, { "epoch": 2.39, "learning_rate": 1.6434224754555413e-05, "loss": 0.0781, "step": 13052 }, { "epoch": 2.39, "learning_rate": 1.643365639826472e-05, "loss": 0.0736, "step": 13053 }, { "epoch": 2.39, "learning_rate": 1.6433088006511563e-05, "loss": 0.0675, "step": 13054 }, { "epoch": 2.39, "learning_rate": 1.643251957929907e-05, "loss": 0.0885, "step": 13055 }, { "epoch": 2.39, "learning_rate": 1.6431951116630383e-05, "loss": 0.0707, "step": 13056 }, { "epoch": 2.39, "learning_rate": 1.643138261850863e-05, "loss": 0.0677, "step": 13057 }, { "epoch": 2.39, "learning_rate": 1.643081408493694e-05, "loss": 0.0742, "step": 13058 }, { "epoch": 2.39, "learning_rate": 1.6430245515918458e-05, "loss": 0.0604, "step": 13059 }, { "epoch": 2.4, "learning_rate": 1.6429676911456312e-05, "loss": 0.0849, "step": 13060 }, { "epoch": 2.4, "learning_rate": 1.6429108271553637e-05, "loss": 0.0774, "step": 13061 }, { "epoch": 2.4, "learning_rate": 1.642853959621356e-05, "loss": 0.0841, "step": 13062 }, { "epoch": 2.4, "learning_rate": 1.6427970885439227e-05, "loss": 0.0827, "step": 13063 }, { "epoch": 2.4, "learning_rate": 1.6427402139233768e-05, "loss": 0.0777, "step": 13064 }, { "epoch": 2.4, "learning_rate": 1.642683335760032e-05, "loss": 0.0723, "step": 13065 }, { "epoch": 2.4, "learning_rate": 1.642626454054201e-05, "loss": 0.0803, "step": 13066 }, { "epoch": 2.4, "learning_rate": 1.6425695688061984e-05, "loss": 0.0717, "step": 13067 }, { "epoch": 2.4, "learning_rate": 1.6425126800163373e-05, "loss": 0.076, "step": 13068 }, { "epoch": 2.4, "learning_rate": 1.6424557876849308e-05, "loss": 0.0659, "step": 13069 }, { "epoch": 2.4, "learning_rate": 1.6423988918122933e-05, "loss": 0.0772, "step": 13070 }, { "epoch": 2.4, "learning_rate": 1.642341992398738e-05, "loss": 0.0872, "step": 13071 }, { "epoch": 2.4, "learning_rate": 1.6422850894445784e-05, "loss": 0.0787, "step": 13072 }, { "epoch": 2.4, "learning_rate": 1.6422281829501285e-05, "loss": 0.0783, "step": 13073 }, { "epoch": 2.4, "learning_rate": 1.642171272915702e-05, "loss": 0.077, "step": 13074 }, { "epoch": 2.4, "learning_rate": 1.6421143593416123e-05, "loss": 0.0783, "step": 13075 }, { "epoch": 2.4, "learning_rate": 1.642057442228173e-05, "loss": 0.0598, "step": 13076 }, { "epoch": 2.4, "learning_rate": 1.6420005215756983e-05, "loss": 0.0727, "step": 13077 }, { "epoch": 2.4, "learning_rate": 1.6419435973845014e-05, "loss": 0.0819, "step": 13078 }, { "epoch": 2.4, "learning_rate": 1.6418866696548965e-05, "loss": 0.0798, "step": 13079 }, { "epoch": 2.4, "learning_rate": 1.6418297383871974e-05, "loss": 0.0745, "step": 13080 }, { "epoch": 2.4, "learning_rate": 1.6417728035817175e-05, "loss": 0.0814, "step": 13081 }, { "epoch": 2.4, "learning_rate": 1.641715865238771e-05, "loss": 0.0763, "step": 13082 }, { "epoch": 2.4, "learning_rate": 1.6416589233586714e-05, "loss": 0.0747, "step": 13083 }, { "epoch": 2.4, "learning_rate": 1.6416019779417328e-05, "loss": 0.0735, "step": 13084 }, { "epoch": 2.4, "learning_rate": 1.6415450289882697e-05, "loss": 0.0844, "step": 13085 }, { "epoch": 2.4, "learning_rate": 1.6414880764985947e-05, "loss": 0.0795, "step": 13086 }, { "epoch": 2.4, "learning_rate": 1.6414311204730228e-05, "loss": 0.0823, "step": 13087 }, { "epoch": 2.4, "learning_rate": 1.6413741609118672e-05, "loss": 0.0844, "step": 13088 }, { "epoch": 2.4, "learning_rate": 1.641317197815442e-05, "loss": 0.0736, "step": 13089 }, { "epoch": 2.4, "learning_rate": 1.641260231184062e-05, "loss": 0.0799, "step": 13090 }, { "epoch": 2.4, "learning_rate": 1.6412032610180403e-05, "loss": 0.0777, "step": 13091 }, { "epoch": 2.4, "learning_rate": 1.641146287317691e-05, "loss": 0.0881, "step": 13092 }, { "epoch": 2.4, "learning_rate": 1.6410893100833285e-05, "loss": 0.0803, "step": 13093 }, { "epoch": 2.4, "learning_rate": 1.641032329315267e-05, "loss": 0.0719, "step": 13094 }, { "epoch": 2.4, "learning_rate": 1.64097534501382e-05, "loss": 0.0766, "step": 13095 }, { "epoch": 2.4, "learning_rate": 1.640918357179302e-05, "loss": 0.067, "step": 13096 }, { "epoch": 2.4, "learning_rate": 1.6408613658120266e-05, "loss": 0.0656, "step": 13097 }, { "epoch": 2.4, "learning_rate": 1.640804370912309e-05, "loss": 0.0689, "step": 13098 }, { "epoch": 2.4, "learning_rate": 1.640747372480462e-05, "loss": 0.084, "step": 13099 }, { "epoch": 2.4, "learning_rate": 1.640690370516801e-05, "loss": 0.0728, "step": 13100 }, { "epoch": 2.4, "learning_rate": 1.6406333650216395e-05, "loss": 0.0755, "step": 13101 }, { "epoch": 2.4, "learning_rate": 1.640576355995292e-05, "loss": 0.0656, "step": 13102 }, { "epoch": 2.4, "learning_rate": 1.6405193434380722e-05, "loss": 0.0782, "step": 13103 }, { "epoch": 2.4, "learning_rate": 1.6404623273502953e-05, "loss": 0.0826, "step": 13104 }, { "epoch": 2.4, "learning_rate": 1.6404053077322747e-05, "loss": 0.0861, "step": 13105 }, { "epoch": 2.4, "learning_rate": 1.640348284584325e-05, "loss": 0.0829, "step": 13106 }, { "epoch": 2.4, "learning_rate": 1.6402912579067608e-05, "loss": 0.0799, "step": 13107 }, { "epoch": 2.4, "learning_rate": 1.6402342276998963e-05, "loss": 0.083, "step": 13108 }, { "epoch": 2.4, "learning_rate": 1.6401771939640455e-05, "loss": 0.0719, "step": 13109 }, { "epoch": 2.4, "learning_rate": 1.640120156699523e-05, "loss": 0.0854, "step": 13110 }, { "epoch": 2.4, "learning_rate": 1.6400631159066433e-05, "loss": 0.0837, "step": 13111 }, { "epoch": 2.4, "learning_rate": 1.6400060715857203e-05, "loss": 0.0826, "step": 13112 }, { "epoch": 2.4, "learning_rate": 1.639949023737069e-05, "loss": 0.0853, "step": 13113 }, { "epoch": 2.4, "learning_rate": 1.639891972361004e-05, "loss": 0.0755, "step": 13114 }, { "epoch": 2.41, "learning_rate": 1.6398349174578396e-05, "loss": 0.0853, "step": 13115 }, { "epoch": 2.41, "learning_rate": 1.6397778590278898e-05, "loss": 0.0764, "step": 13116 }, { "epoch": 2.41, "learning_rate": 1.6397207970714694e-05, "loss": 0.0871, "step": 13117 }, { "epoch": 2.41, "learning_rate": 1.6396637315888933e-05, "loss": 0.0733, "step": 13118 }, { "epoch": 2.41, "learning_rate": 1.6396066625804754e-05, "loss": 0.076, "step": 13119 }, { "epoch": 2.41, "learning_rate": 1.6395495900465306e-05, "loss": 0.0675, "step": 13120 }, { "epoch": 2.41, "learning_rate": 1.639492513987374e-05, "loss": 0.0702, "step": 13121 }, { "epoch": 2.41, "learning_rate": 1.639435434403319e-05, "loss": 0.0776, "step": 13122 }, { "epoch": 2.41, "learning_rate": 1.6393783512946814e-05, "loss": 0.0711, "step": 13123 }, { "epoch": 2.41, "learning_rate": 1.6393212646617746e-05, "loss": 0.0759, "step": 13124 }, { "epoch": 2.41, "learning_rate": 1.639264174504915e-05, "loss": 0.0925, "step": 13125 }, { "epoch": 2.41, "learning_rate": 1.6392070808244157e-05, "loss": 0.0812, "step": 13126 }, { "epoch": 2.41, "learning_rate": 1.639149983620592e-05, "loss": 0.0701, "step": 13127 }, { "epoch": 2.41, "learning_rate": 1.639092882893759e-05, "loss": 0.0858, "step": 13128 }, { "epoch": 2.41, "learning_rate": 1.6390357786442306e-05, "loss": 0.0713, "step": 13129 }, { "epoch": 2.41, "learning_rate": 1.6389786708723223e-05, "loss": 0.0779, "step": 13130 }, { "epoch": 2.41, "learning_rate": 1.6389215595783486e-05, "loss": 0.0801, "step": 13131 }, { "epoch": 2.41, "learning_rate": 1.638864444762624e-05, "loss": 0.0742, "step": 13132 }, { "epoch": 2.41, "learning_rate": 1.638807326425464e-05, "loss": 0.0782, "step": 13133 }, { "epoch": 2.41, "learning_rate": 1.6387502045671827e-05, "loss": 0.073, "step": 13134 }, { "epoch": 2.41, "learning_rate": 1.6386930791880958e-05, "loss": 0.0757, "step": 13135 }, { "epoch": 2.41, "learning_rate": 1.6386359502885172e-05, "loss": 0.0671, "step": 13136 }, { "epoch": 2.41, "learning_rate": 1.6385788178687626e-05, "loss": 0.0875, "step": 13137 }, { "epoch": 2.41, "learning_rate": 1.638521681929146e-05, "loss": 0.0673, "step": 13138 }, { "epoch": 2.41, "learning_rate": 1.6384645424699835e-05, "loss": 0.0796, "step": 13139 }, { "epoch": 2.41, "learning_rate": 1.6384073994915898e-05, "loss": 0.0718, "step": 13140 }, { "epoch": 2.41, "learning_rate": 1.638350252994279e-05, "loss": 0.0826, "step": 13141 }, { "epoch": 2.41, "learning_rate": 1.6382931029783664e-05, "loss": 0.0843, "step": 13142 }, { "epoch": 2.41, "learning_rate": 1.638235949444168e-05, "loss": 0.0699, "step": 13143 }, { "epoch": 2.41, "learning_rate": 1.6381787923919975e-05, "loss": 0.0753, "step": 13144 }, { "epoch": 2.41, "learning_rate": 1.638121631822171e-05, "loss": 0.0639, "step": 13145 }, { "epoch": 2.41, "learning_rate": 1.638064467735003e-05, "loss": 0.0778, "step": 13146 }, { "epoch": 2.41, "learning_rate": 1.6380073001308082e-05, "loss": 0.0822, "step": 13147 }, { "epoch": 2.41, "learning_rate": 1.6379501290099024e-05, "loss": 0.0833, "step": 13148 }, { "epoch": 2.41, "learning_rate": 1.637892954372601e-05, "loss": 0.0878, "step": 13149 }, { "epoch": 2.41, "learning_rate": 1.6378357762192184e-05, "loss": 0.0746, "step": 13150 }, { "epoch": 2.41, "learning_rate": 1.63777859455007e-05, "loss": 0.0793, "step": 13151 }, { "epoch": 2.41, "learning_rate": 1.6377214093654713e-05, "loss": 0.0763, "step": 13152 }, { "epoch": 2.41, "learning_rate": 1.637664220665737e-05, "loss": 0.0784, "step": 13153 }, { "epoch": 2.41, "learning_rate": 1.637607028451183e-05, "loss": 0.0743, "step": 13154 }, { "epoch": 2.41, "learning_rate": 1.637549832722124e-05, "loss": 0.0802, "step": 13155 }, { "epoch": 2.41, "learning_rate": 1.637492633478875e-05, "loss": 0.0741, "step": 13156 }, { "epoch": 2.41, "learning_rate": 1.6374354307217516e-05, "loss": 0.0766, "step": 13157 }, { "epoch": 2.41, "learning_rate": 1.6373782244510693e-05, "loss": 0.0794, "step": 13158 }, { "epoch": 2.41, "learning_rate": 1.6373210146671437e-05, "loss": 0.0774, "step": 13159 }, { "epoch": 2.41, "learning_rate": 1.6372638013702893e-05, "loss": 0.0761, "step": 13160 }, { "epoch": 2.41, "learning_rate": 1.637206584560822e-05, "loss": 0.0752, "step": 13161 }, { "epoch": 2.41, "learning_rate": 1.637149364239057e-05, "loss": 0.0732, "step": 13162 }, { "epoch": 2.41, "learning_rate": 1.63709214040531e-05, "loss": 0.0814, "step": 13163 }, { "epoch": 2.41, "learning_rate": 1.637034913059896e-05, "loss": 0.077, "step": 13164 }, { "epoch": 2.41, "learning_rate": 1.636977682203131e-05, "loss": 0.0686, "step": 13165 }, { "epoch": 2.41, "learning_rate": 1.63692044783533e-05, "loss": 0.0745, "step": 13166 }, { "epoch": 2.41, "learning_rate": 1.636863209956808e-05, "loss": 0.0697, "step": 13167 }, { "epoch": 2.41, "learning_rate": 1.6368059685678817e-05, "loss": 0.0625, "step": 13168 }, { "epoch": 2.42, "learning_rate": 1.6367487236688656e-05, "loss": 0.0776, "step": 13169 }, { "epoch": 2.42, "learning_rate": 1.6366914752600757e-05, "loss": 0.0683, "step": 13170 }, { "epoch": 2.42, "learning_rate": 1.6366342233418276e-05, "loss": 0.0693, "step": 13171 }, { "epoch": 2.42, "learning_rate": 1.636576967914437e-05, "loss": 0.0796, "step": 13172 }, { "epoch": 2.42, "learning_rate": 1.6365197089782188e-05, "loss": 0.092, "step": 13173 }, { "epoch": 2.42, "learning_rate": 1.636462446533489e-05, "loss": 0.0689, "step": 13174 }, { "epoch": 2.42, "learning_rate": 1.6364051805805635e-05, "loss": 0.0886, "step": 13175 }, { "epoch": 2.42, "learning_rate": 1.6363479111197577e-05, "loss": 0.0807, "step": 13176 }, { "epoch": 2.42, "learning_rate": 1.6362906381513874e-05, "loss": 0.07, "step": 13177 }, { "epoch": 2.42, "learning_rate": 1.636233361675768e-05, "loss": 0.0694, "step": 13178 }, { "epoch": 2.42, "learning_rate": 1.6361760816932158e-05, "loss": 0.0786, "step": 13179 }, { "epoch": 2.42, "learning_rate": 1.6361187982040458e-05, "loss": 0.0677, "step": 13180 }, { "epoch": 2.42, "learning_rate": 1.636061511208574e-05, "loss": 0.069, "step": 13181 }, { "epoch": 2.42, "learning_rate": 1.6360042207071164e-05, "loss": 0.0726, "step": 13182 }, { "epoch": 2.42, "learning_rate": 1.6359469266999886e-05, "loss": 0.077, "step": 13183 }, { "epoch": 2.42, "learning_rate": 1.6358896291875064e-05, "loss": 0.0887, "step": 13184 }, { "epoch": 2.42, "learning_rate": 1.6358323281699857e-05, "loss": 0.0706, "step": 13185 }, { "epoch": 2.42, "learning_rate": 1.6357750236477424e-05, "loss": 0.071, "step": 13186 }, { "epoch": 2.42, "learning_rate": 1.635717715621092e-05, "loss": 0.0822, "step": 13187 }, { "epoch": 2.42, "learning_rate": 1.635660404090351e-05, "loss": 0.0802, "step": 13188 }, { "epoch": 2.42, "learning_rate": 1.635603089055835e-05, "loss": 0.0887, "step": 13189 }, { "epoch": 2.42, "learning_rate": 1.6355457705178593e-05, "loss": 0.0906, "step": 13190 }, { "epoch": 2.42, "learning_rate": 1.635488448476741e-05, "loss": 0.082, "step": 13191 }, { "epoch": 2.42, "learning_rate": 1.6354311229327954e-05, "loss": 0.0795, "step": 13192 }, { "epoch": 2.42, "learning_rate": 1.6353737938863384e-05, "loss": 0.0889, "step": 13193 }, { "epoch": 2.42, "learning_rate": 1.6353164613376865e-05, "loss": 0.0736, "step": 13194 }, { "epoch": 2.42, "learning_rate": 1.635259125287155e-05, "loss": 0.0863, "step": 13195 }, { "epoch": 2.42, "learning_rate": 1.6352017857350603e-05, "loss": 0.0797, "step": 13196 }, { "epoch": 2.42, "learning_rate": 1.6351444426817186e-05, "loss": 0.0872, "step": 13197 }, { "epoch": 2.42, "learning_rate": 1.635087096127446e-05, "loss": 0.0915, "step": 13198 }, { "epoch": 2.42, "learning_rate": 1.635029746072558e-05, "loss": 0.0748, "step": 13199 }, { "epoch": 2.42, "learning_rate": 1.634972392517372e-05, "loss": 0.0884, "step": 13200 }, { "epoch": 2.42, "learning_rate": 1.6349150354622025e-05, "loss": 0.0817, "step": 13201 }, { "epoch": 2.42, "learning_rate": 1.6348576749073667e-05, "loss": 0.0734, "step": 13202 }, { "epoch": 2.42, "learning_rate": 1.6348003108531804e-05, "loss": 0.0752, "step": 13203 }, { "epoch": 2.42, "learning_rate": 1.6347429432999602e-05, "loss": 0.0716, "step": 13204 }, { "epoch": 2.42, "learning_rate": 1.634685572248022e-05, "loss": 0.08, "step": 13205 }, { "epoch": 2.42, "learning_rate": 1.6346281976976818e-05, "loss": 0.068, "step": 13206 }, { "epoch": 2.42, "learning_rate": 1.6345708196492562e-05, "loss": 0.0664, "step": 13207 }, { "epoch": 2.42, "learning_rate": 1.6345134381030616e-05, "loss": 0.0888, "step": 13208 }, { "epoch": 2.42, "learning_rate": 1.6344560530594137e-05, "loss": 0.0733, "step": 13209 }, { "epoch": 2.42, "learning_rate": 1.6343986645186293e-05, "loss": 0.0676, "step": 13210 }, { "epoch": 2.42, "learning_rate": 1.634341272481025e-05, "loss": 0.0765, "step": 13211 }, { "epoch": 2.42, "learning_rate": 1.634283876946916e-05, "loss": 0.0832, "step": 13212 }, { "epoch": 2.42, "learning_rate": 1.63422647791662e-05, "loss": 0.0852, "step": 13213 }, { "epoch": 2.42, "learning_rate": 1.6341690753904527e-05, "loss": 0.0691, "step": 13214 }, { "epoch": 2.42, "learning_rate": 1.63411166936873e-05, "loss": 0.077, "step": 13215 }, { "epoch": 2.42, "learning_rate": 1.6340542598517697e-05, "loss": 0.0786, "step": 13216 }, { "epoch": 2.42, "learning_rate": 1.6339968468398874e-05, "loss": 0.0921, "step": 13217 }, { "epoch": 2.42, "learning_rate": 1.6339394303333996e-05, "loss": 0.0706, "step": 13218 }, { "epoch": 2.42, "learning_rate": 1.6338820103326224e-05, "loss": 0.0756, "step": 13219 }, { "epoch": 2.42, "learning_rate": 1.6338245868378732e-05, "loss": 0.085, "step": 13220 }, { "epoch": 2.42, "learning_rate": 1.6337671598494675e-05, "loss": 0.0776, "step": 13221 }, { "epoch": 2.42, "learning_rate": 1.6337097293677227e-05, "loss": 0.0765, "step": 13222 }, { "epoch": 2.42, "learning_rate": 1.6336522953929547e-05, "loss": 0.0835, "step": 13223 }, { "epoch": 2.43, "learning_rate": 1.633594857925481e-05, "loss": 0.0689, "step": 13224 }, { "epoch": 2.43, "learning_rate": 1.6335374169656175e-05, "loss": 0.0841, "step": 13225 }, { "epoch": 2.43, "learning_rate": 1.6334799725136804e-05, "loss": 0.0778, "step": 13226 }, { "epoch": 2.43, "learning_rate": 1.6334225245699872e-05, "loss": 0.0762, "step": 13227 }, { "epoch": 2.43, "learning_rate": 1.6333650731348537e-05, "loss": 0.0784, "step": 13228 }, { "epoch": 2.43, "learning_rate": 1.6333076182085977e-05, "loss": 0.076, "step": 13229 }, { "epoch": 2.43, "learning_rate": 1.6332501597915353e-05, "loss": 0.0688, "step": 13230 }, { "epoch": 2.43, "learning_rate": 1.633192697883983e-05, "loss": 0.0731, "step": 13231 }, { "epoch": 2.43, "learning_rate": 1.6331352324862576e-05, "loss": 0.0753, "step": 13232 }, { "epoch": 2.43, "learning_rate": 1.633077763598676e-05, "loss": 0.0727, "step": 13233 }, { "epoch": 2.43, "learning_rate": 1.6330202912215548e-05, "loss": 0.0703, "step": 13234 }, { "epoch": 2.43, "learning_rate": 1.632962815355211e-05, "loss": 0.0713, "step": 13235 }, { "epoch": 2.43, "learning_rate": 1.6329053359999618e-05, "loss": 0.0734, "step": 13236 }, { "epoch": 2.43, "learning_rate": 1.6328478531561233e-05, "loss": 0.0778, "step": 13237 }, { "epoch": 2.43, "learning_rate": 1.6327903668240123e-05, "loss": 0.0617, "step": 13238 }, { "epoch": 2.43, "learning_rate": 1.6327328770039464e-05, "loss": 0.0851, "step": 13239 }, { "epoch": 2.43, "learning_rate": 1.6326753836962416e-05, "loss": 0.0641, "step": 13240 }, { "epoch": 2.43, "learning_rate": 1.6326178869012154e-05, "loss": 0.072, "step": 13241 }, { "epoch": 2.43, "learning_rate": 1.632560386619185e-05, "loss": 0.0767, "step": 13242 }, { "epoch": 2.43, "learning_rate": 1.6325028828504664e-05, "loss": 0.0869, "step": 13243 }, { "epoch": 2.43, "learning_rate": 1.6324453755953772e-05, "loss": 0.0732, "step": 13244 }, { "epoch": 2.43, "learning_rate": 1.6323878648542348e-05, "loss": 0.0799, "step": 13245 }, { "epoch": 2.43, "learning_rate": 1.632330350627355e-05, "loss": 0.0713, "step": 13246 }, { "epoch": 2.43, "learning_rate": 1.6322728329150558e-05, "loss": 0.0788, "step": 13247 }, { "epoch": 2.43, "learning_rate": 1.632215311717654e-05, "loss": 0.0701, "step": 13248 }, { "epoch": 2.43, "learning_rate": 1.6321577870354665e-05, "loss": 0.0751, "step": 13249 }, { "epoch": 2.43, "learning_rate": 1.6321002588688106e-05, "loss": 0.0781, "step": 13250 }, { "epoch": 2.43, "learning_rate": 1.632042727218003e-05, "loss": 0.0865, "step": 13251 }, { "epoch": 2.43, "learning_rate": 1.6319851920833612e-05, "loss": 0.0739, "step": 13252 }, { "epoch": 2.43, "learning_rate": 1.631927653465202e-05, "loss": 0.0803, "step": 13253 }, { "epoch": 2.43, "learning_rate": 1.6318701113638428e-05, "loss": 0.0794, "step": 13254 }, { "epoch": 2.43, "learning_rate": 1.6318125657796013e-05, "loss": 0.0852, "step": 13255 }, { "epoch": 2.43, "learning_rate": 1.6317550167127936e-05, "loss": 0.0899, "step": 13256 }, { "epoch": 2.43, "learning_rate": 1.6316974641637377e-05, "loss": 0.0783, "step": 13257 }, { "epoch": 2.43, "learning_rate": 1.63163990813275e-05, "loss": 0.0757, "step": 13258 }, { "epoch": 2.43, "learning_rate": 1.6315823486201493e-05, "loss": 0.076, "step": 13259 }, { "epoch": 2.43, "learning_rate": 1.6315247856262514e-05, "loss": 0.0798, "step": 13260 }, { "epoch": 2.43, "learning_rate": 1.6314672191513738e-05, "loss": 0.0863, "step": 13261 }, { "epoch": 2.43, "learning_rate": 1.6314096491958348e-05, "loss": 0.0845, "step": 13262 }, { "epoch": 2.43, "learning_rate": 1.6313520757599502e-05, "loss": 0.072, "step": 13263 }, { "epoch": 2.43, "learning_rate": 1.6312944988440386e-05, "loss": 0.0857, "step": 13264 }, { "epoch": 2.43, "learning_rate": 1.631236918448417e-05, "loss": 0.0845, "step": 13265 }, { "epoch": 2.43, "learning_rate": 1.6311793345734027e-05, "loss": 0.0774, "step": 13266 }, { "epoch": 2.43, "learning_rate": 1.631121747219313e-05, "loss": 0.0826, "step": 13267 }, { "epoch": 2.43, "learning_rate": 1.6310641563864654e-05, "loss": 0.0789, "step": 13268 }, { "epoch": 2.43, "learning_rate": 1.6310065620751777e-05, "loss": 0.0836, "step": 13269 }, { "epoch": 2.43, "learning_rate": 1.6309489642857665e-05, "loss": 0.0797, "step": 13270 }, { "epoch": 2.43, "learning_rate": 1.6308913630185504e-05, "loss": 0.0816, "step": 13271 }, { "epoch": 2.43, "learning_rate": 1.630833758273846e-05, "loss": 0.0693, "step": 13272 }, { "epoch": 2.43, "learning_rate": 1.6307761500519714e-05, "loss": 0.0778, "step": 13273 }, { "epoch": 2.43, "learning_rate": 1.6307185383532433e-05, "loss": 0.0874, "step": 13274 }, { "epoch": 2.43, "learning_rate": 1.6306609231779807e-05, "loss": 0.0642, "step": 13275 }, { "epoch": 2.43, "learning_rate": 1.6306033045264997e-05, "loss": 0.0843, "step": 13276 }, { "epoch": 2.43, "learning_rate": 1.6305456823991186e-05, "loss": 0.0727, "step": 13277 }, { "epoch": 2.43, "learning_rate": 1.6304880567961553e-05, "loss": 0.0847, "step": 13278 }, { "epoch": 2.44, "learning_rate": 1.6304304277179267e-05, "loss": 0.063, "step": 13279 }, { "epoch": 2.44, "learning_rate": 1.6303727951647505e-05, "loss": 0.0684, "step": 13280 }, { "epoch": 2.44, "learning_rate": 1.630315159136945e-05, "loss": 0.0722, "step": 13281 }, { "epoch": 2.44, "learning_rate": 1.6302575196348282e-05, "loss": 0.0723, "step": 13282 }, { "epoch": 2.44, "learning_rate": 1.630199876658716e-05, "loss": 0.0815, "step": 13283 }, { "epoch": 2.44, "learning_rate": 1.6301422302089282e-05, "loss": 0.0786, "step": 13284 }, { "epoch": 2.44, "learning_rate": 1.6300845802857815e-05, "loss": 0.0769, "step": 13285 }, { "epoch": 2.44, "learning_rate": 1.6300269268895938e-05, "loss": 0.0825, "step": 13286 }, { "epoch": 2.44, "learning_rate": 1.6299692700206827e-05, "loss": 0.078, "step": 13287 }, { "epoch": 2.44, "learning_rate": 1.6299116096793665e-05, "loss": 0.0776, "step": 13288 }, { "epoch": 2.44, "learning_rate": 1.6298539458659627e-05, "loss": 0.0686, "step": 13289 }, { "epoch": 2.44, "learning_rate": 1.629796278580789e-05, "loss": 0.0728, "step": 13290 }, { "epoch": 2.44, "learning_rate": 1.629738607824164e-05, "loss": 0.0573, "step": 13291 }, { "epoch": 2.44, "learning_rate": 1.6296809335964042e-05, "loss": 0.0757, "step": 13292 }, { "epoch": 2.44, "learning_rate": 1.6296232558978288e-05, "loss": 0.0702, "step": 13293 }, { "epoch": 2.44, "learning_rate": 1.629565574728755e-05, "loss": 0.0882, "step": 13294 }, { "epoch": 2.44, "learning_rate": 1.6295078900895012e-05, "loss": 0.0767, "step": 13295 }, { "epoch": 2.44, "learning_rate": 1.6294502019803854e-05, "loss": 0.0764, "step": 13296 }, { "epoch": 2.44, "learning_rate": 1.629392510401725e-05, "loss": 0.0823, "step": 13297 }, { "epoch": 2.44, "learning_rate": 1.6293348153538383e-05, "loss": 0.0736, "step": 13298 }, { "epoch": 2.44, "learning_rate": 1.629277116837044e-05, "loss": 0.0728, "step": 13299 }, { "epoch": 2.44, "learning_rate": 1.629219414851659e-05, "loss": 0.0752, "step": 13300 }, { "epoch": 2.44, "learning_rate": 1.6291617093980018e-05, "loss": 0.0655, "step": 13301 }, { "epoch": 2.44, "learning_rate": 1.6291040004763906e-05, "loss": 0.0717, "step": 13302 }, { "epoch": 2.44, "learning_rate": 1.6290462880871434e-05, "loss": 0.0748, "step": 13303 }, { "epoch": 2.44, "learning_rate": 1.6289885722305784e-05, "loss": 0.0825, "step": 13304 }, { "epoch": 2.44, "learning_rate": 1.628930852907013e-05, "loss": 0.0826, "step": 13305 }, { "epoch": 2.44, "learning_rate": 1.6288731301167667e-05, "loss": 0.0729, "step": 13306 }, { "epoch": 2.44, "learning_rate": 1.628815403860157e-05, "loss": 0.0791, "step": 13307 }, { "epoch": 2.44, "learning_rate": 1.628757674137502e-05, "loss": 0.081, "step": 13308 }, { "epoch": 2.44, "learning_rate": 1.6286999409491196e-05, "loss": 0.0744, "step": 13309 }, { "epoch": 2.44, "learning_rate": 1.628642204295329e-05, "loss": 0.0774, "step": 13310 }, { "epoch": 2.44, "learning_rate": 1.6285844641764473e-05, "loss": 0.0727, "step": 13311 }, { "epoch": 2.44, "learning_rate": 1.6285267205927934e-05, "loss": 0.0809, "step": 13312 }, { "epoch": 2.44, "learning_rate": 1.6284689735446856e-05, "loss": 0.0768, "step": 13313 }, { "epoch": 2.44, "learning_rate": 1.628411223032442e-05, "loss": 0.0799, "step": 13314 }, { "epoch": 2.44, "learning_rate": 1.6283534690563814e-05, "loss": 0.0892, "step": 13315 }, { "epoch": 2.44, "learning_rate": 1.6282957116168216e-05, "loss": 0.0704, "step": 13316 }, { "epoch": 2.44, "learning_rate": 1.628237950714081e-05, "loss": 0.0746, "step": 13317 }, { "epoch": 2.44, "learning_rate": 1.628180186348478e-05, "loss": 0.0726, "step": 13318 }, { "epoch": 2.44, "learning_rate": 1.628122418520331e-05, "loss": 0.0763, "step": 13319 }, { "epoch": 2.44, "learning_rate": 1.628064647229959e-05, "loss": 0.072, "step": 13320 }, { "epoch": 2.44, "learning_rate": 1.6280068724776795e-05, "loss": 0.078, "step": 13321 }, { "epoch": 2.44, "learning_rate": 1.6279490942638117e-05, "loss": 0.0839, "step": 13322 }, { "epoch": 2.44, "learning_rate": 1.6278913125886734e-05, "loss": 0.0823, "step": 13323 }, { "epoch": 2.44, "learning_rate": 1.6278335274525837e-05, "loss": 0.0877, "step": 13324 }, { "epoch": 2.44, "learning_rate": 1.6277757388558612e-05, "loss": 0.0678, "step": 13325 }, { "epoch": 2.44, "learning_rate": 1.6277179467988238e-05, "loss": 0.0832, "step": 13326 }, { "epoch": 2.44, "learning_rate": 1.6276601512817906e-05, "loss": 0.0794, "step": 13327 }, { "epoch": 2.44, "learning_rate": 1.6276023523050794e-05, "loss": 0.0796, "step": 13328 }, { "epoch": 2.44, "learning_rate": 1.62754454986901e-05, "loss": 0.0856, "step": 13329 }, { "epoch": 2.44, "learning_rate": 1.6274867439738996e-05, "loss": 0.0777, "step": 13330 }, { "epoch": 2.44, "learning_rate": 1.6274289346200683e-05, "loss": 0.0773, "step": 13331 }, { "epoch": 2.44, "learning_rate": 1.6273711218078336e-05, "loss": 0.0832, "step": 13332 }, { "epoch": 2.45, "learning_rate": 1.6273133055375147e-05, "loss": 0.0742, "step": 13333 }, { "epoch": 2.45, "learning_rate": 1.62725548580943e-05, "loss": 0.0767, "step": 13334 }, { "epoch": 2.45, "learning_rate": 1.6271976626238987e-05, "loss": 0.073, "step": 13335 }, { "epoch": 2.45, "learning_rate": 1.6271398359812385e-05, "loss": 0.0796, "step": 13336 }, { "epoch": 2.45, "learning_rate": 1.6270820058817695e-05, "loss": 0.0784, "step": 13337 }, { "epoch": 2.45, "learning_rate": 1.6270241723258097e-05, "loss": 0.0788, "step": 13338 }, { "epoch": 2.45, "learning_rate": 1.626966335313678e-05, "loss": 0.0762, "step": 13339 }, { "epoch": 2.45, "learning_rate": 1.626908494845693e-05, "loss": 0.0748, "step": 13340 }, { "epoch": 2.45, "learning_rate": 1.626850650922174e-05, "loss": 0.0729, "step": 13341 }, { "epoch": 2.45, "learning_rate": 1.626792803543439e-05, "loss": 0.0737, "step": 13342 }, { "epoch": 2.45, "learning_rate": 1.6267349527098076e-05, "loss": 0.0771, "step": 13343 }, { "epoch": 2.45, "learning_rate": 1.6266770984215984e-05, "loss": 0.0781, "step": 13344 }, { "epoch": 2.45, "learning_rate": 1.626619240679131e-05, "loss": 0.0789, "step": 13345 }, { "epoch": 2.45, "learning_rate": 1.6265613794827228e-05, "loss": 0.0652, "step": 13346 }, { "epoch": 2.45, "learning_rate": 1.626503514832694e-05, "loss": 0.0809, "step": 13347 }, { "epoch": 2.45, "learning_rate": 1.626445646729363e-05, "loss": 0.0781, "step": 13348 }, { "epoch": 2.45, "learning_rate": 1.626387775173049e-05, "loss": 0.0642, "step": 13349 }, { "epoch": 2.45, "learning_rate": 1.6263299001640707e-05, "loss": 0.0714, "step": 13350 }, { "epoch": 2.45, "learning_rate": 1.6262720217027477e-05, "loss": 0.0834, "step": 13351 }, { "epoch": 2.45, "learning_rate": 1.6262141397893985e-05, "loss": 0.0704, "step": 13352 }, { "epoch": 2.45, "learning_rate": 1.6261562544243422e-05, "loss": 0.0773, "step": 13353 }, { "epoch": 2.45, "learning_rate": 1.626098365607898e-05, "loss": 0.0898, "step": 13354 }, { "epoch": 2.45, "learning_rate": 1.6260404733403852e-05, "loss": 0.0833, "step": 13355 }, { "epoch": 2.45, "learning_rate": 1.625982577622122e-05, "loss": 0.07, "step": 13356 }, { "epoch": 2.45, "learning_rate": 1.6259246784534287e-05, "loss": 0.0807, "step": 13357 }, { "epoch": 2.45, "learning_rate": 1.6258667758346237e-05, "loss": 0.0833, "step": 13358 }, { "epoch": 2.45, "learning_rate": 1.6258088697660266e-05, "loss": 0.0818, "step": 13359 }, { "epoch": 2.45, "learning_rate": 1.6257509602479563e-05, "loss": 0.0824, "step": 13360 }, { "epoch": 2.45, "learning_rate": 1.625693047280732e-05, "loss": 0.0718, "step": 13361 }, { "epoch": 2.45, "learning_rate": 1.6256351308646726e-05, "loss": 0.0744, "step": 13362 }, { "epoch": 2.45, "learning_rate": 1.6255772110000982e-05, "loss": 0.0777, "step": 13363 }, { "epoch": 2.45, "learning_rate": 1.6255192876873276e-05, "loss": 0.0843, "step": 13364 }, { "epoch": 2.45, "learning_rate": 1.6254613609266795e-05, "loss": 0.0876, "step": 13365 }, { "epoch": 2.45, "learning_rate": 1.6254034307184742e-05, "loss": 0.0707, "step": 13366 }, { "epoch": 2.45, "learning_rate": 1.62534549706303e-05, "loss": 0.0789, "step": 13367 }, { "epoch": 2.45, "learning_rate": 1.6252875599606675e-05, "loss": 0.0747, "step": 13368 }, { "epoch": 2.45, "learning_rate": 1.6252296194117047e-05, "loss": 0.0682, "step": 13369 }, { "epoch": 2.45, "learning_rate": 1.625171675416462e-05, "loss": 0.0745, "step": 13370 }, { "epoch": 2.45, "learning_rate": 1.6251137279752582e-05, "loss": 0.0622, "step": 13371 }, { "epoch": 2.45, "learning_rate": 1.625055777088413e-05, "loss": 0.075, "step": 13372 }, { "epoch": 2.45, "learning_rate": 1.6249978227562454e-05, "loss": 0.0743, "step": 13373 }, { "epoch": 2.45, "learning_rate": 1.6249398649790756e-05, "loss": 0.0715, "step": 13374 }, { "epoch": 2.45, "learning_rate": 1.6248819037572225e-05, "loss": 0.0697, "step": 13375 }, { "epoch": 2.45, "learning_rate": 1.6248239390910054e-05, "loss": 0.0771, "step": 13376 }, { "epoch": 2.45, "learning_rate": 1.6247659709807445e-05, "loss": 0.0694, "step": 13377 }, { "epoch": 2.45, "learning_rate": 1.6247079994267587e-05, "loss": 0.075, "step": 13378 }, { "epoch": 2.45, "learning_rate": 1.6246500244293678e-05, "loss": 0.0797, "step": 13379 }, { "epoch": 2.45, "learning_rate": 1.6245920459888913e-05, "loss": 0.0813, "step": 13380 }, { "epoch": 2.45, "learning_rate": 1.6245340641056487e-05, "loss": 0.075, "step": 13381 }, { "epoch": 2.45, "learning_rate": 1.62447607877996e-05, "loss": 0.077, "step": 13382 }, { "epoch": 2.45, "learning_rate": 1.6244180900121443e-05, "loss": 0.0826, "step": 13383 }, { "epoch": 2.45, "learning_rate": 1.6243600978025215e-05, "loss": 0.0812, "step": 13384 }, { "epoch": 2.45, "learning_rate": 1.6243021021514113e-05, "loss": 0.0683, "step": 13385 }, { "epoch": 2.45, "learning_rate": 1.6242441030591335e-05, "loss": 0.0755, "step": 13386 }, { "epoch": 2.45, "learning_rate": 1.6241861005260072e-05, "loss": 0.0753, "step": 13387 }, { "epoch": 2.46, "learning_rate": 1.6241280945523524e-05, "loss": 0.0721, "step": 13388 }, { "epoch": 2.46, "learning_rate": 1.6240700851384893e-05, "loss": 0.0699, "step": 13389 }, { "epoch": 2.46, "learning_rate": 1.624012072284737e-05, "loss": 0.0907, "step": 13390 }, { "epoch": 2.46, "learning_rate": 1.6239540559914157e-05, "loss": 0.0668, "step": 13391 }, { "epoch": 2.46, "learning_rate": 1.623896036258845e-05, "loss": 0.0731, "step": 13392 }, { "epoch": 2.46, "learning_rate": 1.6238380130873447e-05, "loss": 0.0774, "step": 13393 }, { "epoch": 2.46, "learning_rate": 1.6237799864772348e-05, "loss": 0.0654, "step": 13394 }, { "epoch": 2.46, "learning_rate": 1.6237219564288348e-05, "loss": 0.0755, "step": 13395 }, { "epoch": 2.46, "learning_rate": 1.623663922942465e-05, "loss": 0.0781, "step": 13396 }, { "epoch": 2.46, "learning_rate": 1.623605886018445e-05, "loss": 0.0705, "step": 13397 }, { "epoch": 2.46, "learning_rate": 1.6235478456570942e-05, "loss": 0.0747, "step": 13398 }, { "epoch": 2.46, "learning_rate": 1.6234898018587336e-05, "loss": 0.0797, "step": 13399 }, { "epoch": 2.46, "learning_rate": 1.6234317546236824e-05, "loss": 0.0817, "step": 13400 }, { "epoch": 2.46, "learning_rate": 1.623373703952261e-05, "loss": 0.0749, "step": 13401 }, { "epoch": 2.46, "learning_rate": 1.623315649844789e-05, "loss": 0.0894, "step": 13402 }, { "epoch": 2.46, "learning_rate": 1.6232575923015864e-05, "loss": 0.0827, "step": 13403 }, { "epoch": 2.46, "learning_rate": 1.6231995313229737e-05, "loss": 0.0705, "step": 13404 }, { "epoch": 2.46, "learning_rate": 1.6231414669092704e-05, "loss": 0.0866, "step": 13405 }, { "epoch": 2.46, "learning_rate": 1.6230833990607967e-05, "loss": 0.0813, "step": 13406 }, { "epoch": 2.46, "learning_rate": 1.6230253277778727e-05, "loss": 0.094, "step": 13407 }, { "epoch": 2.46, "learning_rate": 1.6229672530608186e-05, "loss": 0.0746, "step": 13408 }, { "epoch": 2.46, "learning_rate": 1.6229091749099542e-05, "loss": 0.0778, "step": 13409 }, { "epoch": 2.46, "learning_rate": 1.6228510933256e-05, "loss": 0.0879, "step": 13410 }, { "epoch": 2.46, "learning_rate": 1.622793008308076e-05, "loss": 0.0771, "step": 13411 }, { "epoch": 2.46, "learning_rate": 1.622734919857702e-05, "loss": 0.0772, "step": 13412 }, { "epoch": 2.46, "learning_rate": 1.622676827974799e-05, "loss": 0.0838, "step": 13413 }, { "epoch": 2.46, "learning_rate": 1.6226187326596865e-05, "loss": 0.0857, "step": 13414 }, { "epoch": 2.46, "learning_rate": 1.622560633912685e-05, "loss": 0.0804, "step": 13415 }, { "epoch": 2.46, "learning_rate": 1.6225025317341148e-05, "loss": 0.0733, "step": 13416 }, { "epoch": 2.46, "learning_rate": 1.622444426124296e-05, "loss": 0.0748, "step": 13417 }, { "epoch": 2.46, "learning_rate": 1.6223863170835488e-05, "loss": 0.0659, "step": 13418 }, { "epoch": 2.46, "learning_rate": 1.622328204612194e-05, "loss": 0.0904, "step": 13419 }, { "epoch": 2.46, "learning_rate": 1.622270088710551e-05, "loss": 0.0802, "step": 13420 }, { "epoch": 2.46, "learning_rate": 1.622211969378941e-05, "loss": 0.0727, "step": 13421 }, { "epoch": 2.46, "learning_rate": 1.6221538466176842e-05, "loss": 0.0807, "step": 13422 }, { "epoch": 2.46, "learning_rate": 1.6220957204271006e-05, "loss": 0.0705, "step": 13423 }, { "epoch": 2.46, "learning_rate": 1.6220375908075108e-05, "loss": 0.0765, "step": 13424 }, { "epoch": 2.46, "learning_rate": 1.6219794577592353e-05, "loss": 0.0719, "step": 13425 }, { "epoch": 2.46, "learning_rate": 1.6219213212825943e-05, "loss": 0.0775, "step": 13426 }, { "epoch": 2.46, "learning_rate": 1.6218631813779087e-05, "loss": 0.0722, "step": 13427 }, { "epoch": 2.46, "learning_rate": 1.6218050380454985e-05, "loss": 0.0728, "step": 13428 }, { "epoch": 2.46, "learning_rate": 1.6217468912856848e-05, "loss": 0.0603, "step": 13429 }, { "epoch": 2.46, "learning_rate": 1.6216887410987872e-05, "loss": 0.076, "step": 13430 }, { "epoch": 2.46, "learning_rate": 1.6216305874851267e-05, "loss": 0.0758, "step": 13431 }, { "epoch": 2.46, "learning_rate": 1.621572430445024e-05, "loss": 0.073, "step": 13432 }, { "epoch": 2.46, "learning_rate": 1.6215142699787994e-05, "loss": 0.0805, "step": 13433 }, { "epoch": 2.46, "learning_rate": 1.6214561060867736e-05, "loss": 0.0684, "step": 13434 }, { "epoch": 2.46, "learning_rate": 1.621397938769267e-05, "loss": 0.0747, "step": 13435 }, { "epoch": 2.46, "learning_rate": 1.621339768026601e-05, "loss": 0.0771, "step": 13436 }, { "epoch": 2.46, "learning_rate": 1.6212815938590947e-05, "loss": 0.0798, "step": 13437 }, { "epoch": 2.46, "learning_rate": 1.6212234162670702e-05, "loss": 0.0663, "step": 13438 }, { "epoch": 2.46, "learning_rate": 1.621165235250848e-05, "loss": 0.0709, "step": 13439 }, { "epoch": 2.46, "learning_rate": 1.621107050810748e-05, "loss": 0.072, "step": 13440 }, { "epoch": 2.46, "learning_rate": 1.6210488629470916e-05, "loss": 0.0754, "step": 13441 }, { "epoch": 2.47, "learning_rate": 1.6209906716601995e-05, "loss": 0.0817, "step": 13442 }, { "epoch": 2.47, "learning_rate": 1.620932476950392e-05, "loss": 0.07, "step": 13443 }, { "epoch": 2.47, "learning_rate": 1.62087427881799e-05, "loss": 0.0914, "step": 13444 }, { "epoch": 2.47, "learning_rate": 1.6208160772633147e-05, "loss": 0.0726, "step": 13445 }, { "epoch": 2.47, "learning_rate": 1.6207578722866862e-05, "loss": 0.0733, "step": 13446 }, { "epoch": 2.47, "learning_rate": 1.6206996638884265e-05, "loss": 0.0706, "step": 13447 }, { "epoch": 2.47, "learning_rate": 1.6206414520688554e-05, "loss": 0.0858, "step": 13448 }, { "epoch": 2.47, "learning_rate": 1.620583236828294e-05, "loss": 0.0728, "step": 13449 }, { "epoch": 2.47, "learning_rate": 1.6205250181670633e-05, "loss": 0.0797, "step": 13450 }, { "epoch": 2.47, "learning_rate": 1.620466796085484e-05, "loss": 0.077, "step": 13451 }, { "epoch": 2.47, "learning_rate": 1.6204085705838774e-05, "loss": 0.08, "step": 13452 }, { "epoch": 2.47, "learning_rate": 1.620350341662564e-05, "loss": 0.0669, "step": 13453 }, { "epoch": 2.47, "learning_rate": 1.6202921093218654e-05, "loss": 0.0743, "step": 13454 }, { "epoch": 2.47, "learning_rate": 1.6202338735621016e-05, "loss": 0.0836, "step": 13455 }, { "epoch": 2.47, "learning_rate": 1.6201756343835946e-05, "loss": 0.074, "step": 13456 }, { "epoch": 2.47, "learning_rate": 1.6201173917866646e-05, "loss": 0.0862, "step": 13457 }, { "epoch": 2.47, "learning_rate": 1.6200591457716334e-05, "loss": 0.0821, "step": 13458 }, { "epoch": 2.47, "learning_rate": 1.6200008963388216e-05, "loss": 0.0688, "step": 13459 }, { "epoch": 2.47, "learning_rate": 1.6199426434885503e-05, "loss": 0.071, "step": 13460 }, { "epoch": 2.47, "learning_rate": 1.6198843872211404e-05, "loss": 0.0727, "step": 13461 }, { "epoch": 2.47, "learning_rate": 1.6198261275369134e-05, "loss": 0.0816, "step": 13462 }, { "epoch": 2.47, "learning_rate": 1.6197678644361902e-05, "loss": 0.0814, "step": 13463 }, { "epoch": 2.47, "learning_rate": 1.619709597919292e-05, "loss": 0.079, "step": 13464 }, { "epoch": 2.47, "learning_rate": 1.61965132798654e-05, "loss": 0.0678, "step": 13465 }, { "epoch": 2.47, "learning_rate": 1.6195930546382554e-05, "loss": 0.0754, "step": 13466 }, { "epoch": 2.47, "learning_rate": 1.6195347778747593e-05, "loss": 0.0827, "step": 13467 }, { "epoch": 2.47, "learning_rate": 1.6194764976963733e-05, "loss": 0.0739, "step": 13468 }, { "epoch": 2.47, "learning_rate": 1.619418214103418e-05, "loss": 0.0759, "step": 13469 }, { "epoch": 2.47, "learning_rate": 1.619359927096215e-05, "loss": 0.0602, "step": 13470 }, { "epoch": 2.47, "learning_rate": 1.619301636675086e-05, "loss": 0.0687, "step": 13471 }, { "epoch": 2.47, "learning_rate": 1.6192433428403513e-05, "loss": 0.0723, "step": 13472 }, { "epoch": 2.47, "learning_rate": 1.619185045592333e-05, "loss": 0.0872, "step": 13473 }, { "epoch": 2.47, "learning_rate": 1.6191267449313524e-05, "loss": 0.0805, "step": 13474 }, { "epoch": 2.47, "learning_rate": 1.6190684408577303e-05, "loss": 0.0802, "step": 13475 }, { "epoch": 2.47, "learning_rate": 1.6190101333717888e-05, "loss": 0.0778, "step": 13476 }, { "epoch": 2.47, "learning_rate": 1.618951822473849e-05, "loss": 0.0718, "step": 13477 }, { "epoch": 2.47, "learning_rate": 1.6188935081642317e-05, "loss": 0.0729, "step": 13478 }, { "epoch": 2.47, "learning_rate": 1.6188351904432594e-05, "loss": 0.0678, "step": 13479 }, { "epoch": 2.47, "learning_rate": 1.618776869311253e-05, "loss": 0.0746, "step": 13480 }, { "epoch": 2.47, "learning_rate": 1.6187185447685338e-05, "loss": 0.0674, "step": 13481 }, { "epoch": 2.47, "learning_rate": 1.618660216815423e-05, "loss": 0.0794, "step": 13482 }, { "epoch": 2.47, "learning_rate": 1.6186018854522436e-05, "loss": 0.0821, "step": 13483 }, { "epoch": 2.47, "learning_rate": 1.6185435506793153e-05, "loss": 0.067, "step": 13484 }, { "epoch": 2.47, "learning_rate": 1.618485212496961e-05, "loss": 0.0709, "step": 13485 }, { "epoch": 2.47, "learning_rate": 1.6184268709055017e-05, "loss": 0.0757, "step": 13486 }, { "epoch": 2.47, "learning_rate": 1.6183685259052587e-05, "loss": 0.0707, "step": 13487 }, { "epoch": 2.47, "learning_rate": 1.618310177496554e-05, "loss": 0.0816, "step": 13488 }, { "epoch": 2.47, "learning_rate": 1.6182518256797095e-05, "loss": 0.0773, "step": 13489 }, { "epoch": 2.47, "learning_rate": 1.618193470455046e-05, "loss": 0.0734, "step": 13490 }, { "epoch": 2.47, "learning_rate": 1.618135111822886e-05, "loss": 0.0785, "step": 13491 }, { "epoch": 2.47, "learning_rate": 1.6180767497835503e-05, "loss": 0.0852, "step": 13492 }, { "epoch": 2.47, "learning_rate": 1.618018384337361e-05, "loss": 0.0635, "step": 13493 }, { "epoch": 2.47, "learning_rate": 1.6179600154846405e-05, "loss": 0.0738, "step": 13494 }, { "epoch": 2.47, "learning_rate": 1.6179016432257094e-05, "loss": 0.0764, "step": 13495 }, { "epoch": 2.47, "learning_rate": 1.6178432675608903e-05, "loss": 0.0776, "step": 13496 }, { "epoch": 2.48, "learning_rate": 1.617784888490504e-05, "loss": 0.0638, "step": 13497 }, { "epoch": 2.48, "learning_rate": 1.617726506014874e-05, "loss": 0.0889, "step": 13498 }, { "epoch": 2.48, "learning_rate": 1.61766812013432e-05, "loss": 0.0744, "step": 13499 }, { "epoch": 2.48, "learning_rate": 1.6176097308491657e-05, "loss": 0.0757, "step": 13500 }, { "epoch": 2.48, "learning_rate": 1.6175513381597315e-05, "loss": 0.0769, "step": 13501 }, { "epoch": 2.48, "learning_rate": 1.61749294206634e-05, "loss": 0.0652, "step": 13502 }, { "epoch": 2.48, "learning_rate": 1.617434542569313e-05, "loss": 0.0705, "step": 13503 }, { "epoch": 2.48, "learning_rate": 1.617376139668972e-05, "loss": 0.0789, "step": 13504 }, { "epoch": 2.48, "learning_rate": 1.6173177333656398e-05, "loss": 0.0733, "step": 13505 }, { "epoch": 2.48, "learning_rate": 1.6172593236596376e-05, "loss": 0.0664, "step": 13506 }, { "epoch": 2.48, "learning_rate": 1.6172009105512873e-05, "loss": 0.0811, "step": 13507 }, { "epoch": 2.48, "learning_rate": 1.6171424940409112e-05, "loss": 0.0578, "step": 13508 }, { "epoch": 2.48, "learning_rate": 1.6170840741288318e-05, "loss": 0.0739, "step": 13509 }, { "epoch": 2.48, "learning_rate": 1.61702565081537e-05, "loss": 0.0778, "step": 13510 }, { "epoch": 2.48, "learning_rate": 1.6169672241008482e-05, "loss": 0.0673, "step": 13511 }, { "epoch": 2.48, "learning_rate": 1.6169087939855888e-05, "loss": 0.0653, "step": 13512 }, { "epoch": 2.48, "learning_rate": 1.616850360469914e-05, "loss": 0.0672, "step": 13513 }, { "epoch": 2.48, "learning_rate": 1.616791923554145e-05, "loss": 0.0713, "step": 13514 }, { "epoch": 2.48, "learning_rate": 1.6167334832386047e-05, "loss": 0.075, "step": 13515 }, { "epoch": 2.48, "learning_rate": 1.616675039523615e-05, "loss": 0.0684, "step": 13516 }, { "epoch": 2.48, "learning_rate": 1.616616592409498e-05, "loss": 0.0688, "step": 13517 }, { "epoch": 2.48, "learning_rate": 1.6165581418965763e-05, "loss": 0.0625, "step": 13518 }, { "epoch": 2.48, "learning_rate": 1.6164996879851714e-05, "loss": 0.0841, "step": 13519 }, { "epoch": 2.48, "learning_rate": 1.6164412306756055e-05, "loss": 0.0791, "step": 13520 }, { "epoch": 2.48, "learning_rate": 1.616382769968201e-05, "loss": 0.0794, "step": 13521 }, { "epoch": 2.48, "learning_rate": 1.616324305863281e-05, "loss": 0.06, "step": 13522 }, { "epoch": 2.48, "learning_rate": 1.6162658383611666e-05, "loss": 0.0803, "step": 13523 }, { "epoch": 2.48, "learning_rate": 1.61620736746218e-05, "loss": 0.0843, "step": 13524 }, { "epoch": 2.48, "learning_rate": 1.6161488931666445e-05, "loss": 0.0767, "step": 13525 }, { "epoch": 2.48, "learning_rate": 1.6160904154748817e-05, "loss": 0.0748, "step": 13526 }, { "epoch": 2.48, "learning_rate": 1.616031934387214e-05, "loss": 0.0702, "step": 13527 }, { "epoch": 2.48, "learning_rate": 1.6159734499039642e-05, "loss": 0.0674, "step": 13528 }, { "epoch": 2.48, "learning_rate": 1.615914962025454e-05, "loss": 0.0711, "step": 13529 }, { "epoch": 2.48, "learning_rate": 1.6158564707520058e-05, "loss": 0.0774, "step": 13530 }, { "epoch": 2.48, "learning_rate": 1.615797976083943e-05, "loss": 0.0784, "step": 13531 }, { "epoch": 2.48, "learning_rate": 1.615739478021587e-05, "loss": 0.0647, "step": 13532 }, { "epoch": 2.48, "learning_rate": 1.6156809765652606e-05, "loss": 0.0732, "step": 13533 }, { "epoch": 2.48, "learning_rate": 1.615622471715286e-05, "loss": 0.0683, "step": 13534 }, { "epoch": 2.48, "learning_rate": 1.615563963471986e-05, "loss": 0.0732, "step": 13535 }, { "epoch": 2.48, "learning_rate": 1.6155054518356833e-05, "loss": 0.073, "step": 13536 }, { "epoch": 2.48, "learning_rate": 1.6154469368066998e-05, "loss": 0.0761, "step": 13537 }, { "epoch": 2.48, "learning_rate": 1.6153884183853588e-05, "loss": 0.074, "step": 13538 }, { "epoch": 2.48, "learning_rate": 1.615329896571982e-05, "loss": 0.0733, "step": 13539 }, { "epoch": 2.48, "learning_rate": 1.6152713713668925e-05, "loss": 0.0884, "step": 13540 }, { "epoch": 2.48, "learning_rate": 1.6152128427704127e-05, "loss": 0.0799, "step": 13541 }, { "epoch": 2.48, "learning_rate": 1.6151543107828657e-05, "loss": 0.0668, "step": 13542 }, { "epoch": 2.48, "learning_rate": 1.6150957754045732e-05, "loss": 0.0685, "step": 13543 }, { "epoch": 2.48, "learning_rate": 1.6150372366358586e-05, "loss": 0.0692, "step": 13544 }, { "epoch": 2.48, "learning_rate": 1.6149786944770447e-05, "loss": 0.0734, "step": 13545 }, { "epoch": 2.48, "learning_rate": 1.6149201489284534e-05, "loss": 0.0839, "step": 13546 }, { "epoch": 2.48, "learning_rate": 1.614861599990408e-05, "loss": 0.0684, "step": 13547 }, { "epoch": 2.48, "learning_rate": 1.6148030476632308e-05, "loss": 0.0705, "step": 13548 }, { "epoch": 2.48, "learning_rate": 1.6147444919472453e-05, "loss": 0.0682, "step": 13549 }, { "epoch": 2.48, "learning_rate": 1.6146859328427732e-05, "loss": 0.0769, "step": 13550 }, { "epoch": 2.49, "learning_rate": 1.614627370350138e-05, "loss": 0.0761, "step": 13551 }, { "epoch": 2.49, "learning_rate": 1.6145688044696625e-05, "loss": 0.0823, "step": 13552 }, { "epoch": 2.49, "learning_rate": 1.6145102352016692e-05, "loss": 0.0601, "step": 13553 }, { "epoch": 2.49, "learning_rate": 1.6144516625464813e-05, "loss": 0.0841, "step": 13554 }, { "epoch": 2.49, "learning_rate": 1.6143930865044214e-05, "loss": 0.086, "step": 13555 }, { "epoch": 2.49, "learning_rate": 1.6143345070758122e-05, "loss": 0.0759, "step": 13556 }, { "epoch": 2.49, "learning_rate": 1.614275924260977e-05, "loss": 0.0751, "step": 13557 }, { "epoch": 2.49, "learning_rate": 1.614217338060238e-05, "loss": 0.0701, "step": 13558 }, { "epoch": 2.49, "learning_rate": 1.614158748473919e-05, "loss": 0.0651, "step": 13559 }, { "epoch": 2.49, "learning_rate": 1.614100155502343e-05, "loss": 0.0814, "step": 13560 }, { "epoch": 2.49, "learning_rate": 1.6140415591458323e-05, "loss": 0.0749, "step": 13561 }, { "epoch": 2.49, "learning_rate": 1.61398295940471e-05, "loss": 0.0866, "step": 13562 }, { "epoch": 2.49, "learning_rate": 1.6139243562792994e-05, "loss": 0.0753, "step": 13563 }, { "epoch": 2.49, "learning_rate": 1.6138657497699234e-05, "loss": 0.0752, "step": 13564 }, { "epoch": 2.49, "learning_rate": 1.6138071398769045e-05, "loss": 0.0832, "step": 13565 }, { "epoch": 2.49, "learning_rate": 1.6137485266005667e-05, "loss": 0.0836, "step": 13566 }, { "epoch": 2.49, "learning_rate": 1.6136899099412325e-05, "loss": 0.0701, "step": 13567 }, { "epoch": 2.49, "learning_rate": 1.6136312898992255e-05, "loss": 0.0721, "step": 13568 }, { "epoch": 2.49, "learning_rate": 1.613572666474868e-05, "loss": 0.0737, "step": 13569 }, { "epoch": 2.49, "learning_rate": 1.6135140396684835e-05, "loss": 0.0828, "step": 13570 }, { "epoch": 2.49, "learning_rate": 1.6134554094803956e-05, "loss": 0.0661, "step": 13571 }, { "epoch": 2.49, "learning_rate": 1.613396775910927e-05, "loss": 0.0719, "step": 13572 }, { "epoch": 2.49, "learning_rate": 1.6133381389604014e-05, "loss": 0.077, "step": 13573 }, { "epoch": 2.49, "learning_rate": 1.613279498629141e-05, "loss": 0.0745, "step": 13574 }, { "epoch": 2.49, "learning_rate": 1.61322085491747e-05, "loss": 0.067, "step": 13575 }, { "epoch": 2.49, "learning_rate": 1.613162207825711e-05, "loss": 0.0776, "step": 13576 }, { "epoch": 2.49, "learning_rate": 1.6131035573541876e-05, "loss": 0.0818, "step": 13577 }, { "epoch": 2.49, "learning_rate": 1.6130449035032234e-05, "loss": 0.0771, "step": 13578 }, { "epoch": 2.49, "learning_rate": 1.6129862462731408e-05, "loss": 0.0791, "step": 13579 }, { "epoch": 2.49, "learning_rate": 1.6129275856642638e-05, "loss": 0.0815, "step": 13580 }, { "epoch": 2.49, "learning_rate": 1.612868921676916e-05, "loss": 0.068, "step": 13581 }, { "epoch": 2.49, "learning_rate": 1.61281025431142e-05, "loss": 0.0865, "step": 13582 }, { "epoch": 2.49, "learning_rate": 1.6127515835680993e-05, "loss": 0.0762, "step": 13583 }, { "epoch": 2.49, "learning_rate": 1.612692909447278e-05, "loss": 0.0811, "step": 13584 }, { "epoch": 2.49, "learning_rate": 1.6126342319492784e-05, "loss": 0.0836, "step": 13585 }, { "epoch": 2.49, "learning_rate": 1.6125755510744253e-05, "loss": 0.0746, "step": 13586 }, { "epoch": 2.49, "learning_rate": 1.612516866823041e-05, "loss": 0.0751, "step": 13587 }, { "epoch": 2.49, "learning_rate": 1.6124581791954494e-05, "loss": 0.0788, "step": 13588 }, { "epoch": 2.49, "learning_rate": 1.612399488191974e-05, "loss": 0.0771, "step": 13589 }, { "epoch": 2.49, "learning_rate": 1.6123407938129384e-05, "loss": 0.0807, "step": 13590 }, { "epoch": 2.49, "learning_rate": 1.6122820960586657e-05, "loss": 0.0774, "step": 13591 }, { "epoch": 2.49, "learning_rate": 1.61222339492948e-05, "loss": 0.0804, "step": 13592 }, { "epoch": 2.49, "learning_rate": 1.6121646904257047e-05, "loss": 0.0805, "step": 13593 }, { "epoch": 2.49, "learning_rate": 1.612105982547663e-05, "loss": 0.0677, "step": 13594 }, { "epoch": 2.49, "learning_rate": 1.612047271295679e-05, "loss": 0.0738, "step": 13595 }, { "epoch": 2.49, "learning_rate": 1.611988556670076e-05, "loss": 0.0719, "step": 13596 }, { "epoch": 2.49, "learning_rate": 1.611929838671178e-05, "loss": 0.0766, "step": 13597 }, { "epoch": 2.49, "learning_rate": 1.611871117299308e-05, "loss": 0.0737, "step": 13598 }, { "epoch": 2.49, "learning_rate": 1.61181239255479e-05, "loss": 0.0656, "step": 13599 }, { "epoch": 2.49, "learning_rate": 1.6117536644379482e-05, "loss": 0.0884, "step": 13600 }, { "epoch": 2.49, "learning_rate": 1.6116949329491055e-05, "loss": 0.0817, "step": 13601 }, { "epoch": 2.49, "learning_rate": 1.611636198088586e-05, "loss": 0.082, "step": 13602 }, { "epoch": 2.49, "learning_rate": 1.6115774598567136e-05, "loss": 0.0746, "step": 13603 }, { "epoch": 2.49, "learning_rate": 1.611518718253812e-05, "loss": 0.0889, "step": 13604 }, { "epoch": 2.49, "learning_rate": 1.6114599732802044e-05, "loss": 0.0753, "step": 13605 }, { "epoch": 2.5, "learning_rate": 1.6114012249362157e-05, "loss": 0.0684, "step": 13606 }, { "epoch": 2.5, "learning_rate": 1.611342473222169e-05, "loss": 0.0738, "step": 13607 }, { "epoch": 2.5, "learning_rate": 1.6112837181383883e-05, "loss": 0.0763, "step": 13608 }, { "epoch": 2.5, "learning_rate": 1.6112249596851973e-05, "loss": 0.0871, "step": 13609 }, { "epoch": 2.5, "learning_rate": 1.61116619786292e-05, "loss": 0.0792, "step": 13610 }, { "epoch": 2.5, "learning_rate": 1.6111074326718803e-05, "loss": 0.081, "step": 13611 }, { "epoch": 2.5, "learning_rate": 1.6110486641124018e-05, "loss": 0.078, "step": 13612 }, { "epoch": 2.5, "learning_rate": 1.6109898921848092e-05, "loss": 0.0662, "step": 13613 }, { "epoch": 2.5, "learning_rate": 1.610931116889426e-05, "loss": 0.0737, "step": 13614 }, { "epoch": 2.5, "learning_rate": 1.610872338226576e-05, "loss": 0.0665, "step": 13615 }, { "epoch": 2.5, "learning_rate": 1.6108135561965833e-05, "loss": 0.0706, "step": 13616 }, { "epoch": 2.5, "learning_rate": 1.610754770799772e-05, "loss": 0.0692, "step": 13617 }, { "epoch": 2.5, "learning_rate": 1.6106959820364662e-05, "loss": 0.0804, "step": 13618 }, { "epoch": 2.5, "learning_rate": 1.61063718990699e-05, "loss": 0.0684, "step": 13619 }, { "epoch": 2.5, "learning_rate": 1.610578394411667e-05, "loss": 0.0756, "step": 13620 }, { "epoch": 2.5, "learning_rate": 1.6105195955508217e-05, "loss": 0.0682, "step": 13621 }, { "epoch": 2.5, "learning_rate": 1.6104607933247783e-05, "loss": 0.0857, "step": 13622 }, { "epoch": 2.5, "learning_rate": 1.6104019877338603e-05, "loss": 0.0778, "step": 13623 }, { "epoch": 2.5, "learning_rate": 1.6103431787783925e-05, "loss": 0.0772, "step": 13624 }, { "epoch": 2.5, "learning_rate": 1.6102843664586986e-05, "loss": 0.0718, "step": 13625 }, { "epoch": 2.5, "learning_rate": 1.610225550775103e-05, "loss": 0.0677, "step": 13626 }, { "epoch": 2.5, "learning_rate": 1.6101667317279303e-05, "loss": 0.0729, "step": 13627 }, { "epoch": 2.5, "learning_rate": 1.6101079093175037e-05, "loss": 0.0713, "step": 13628 }, { "epoch": 2.5, "learning_rate": 1.6100490835441483e-05, "loss": 0.0718, "step": 13629 }, { "epoch": 2.5, "learning_rate": 1.6099902544081883e-05, "loss": 0.081, "step": 13630 }, { "epoch": 2.5, "learning_rate": 1.6099314219099473e-05, "loss": 0.0747, "step": 13631 }, { "epoch": 2.5, "learning_rate": 1.6098725860497502e-05, "loss": 0.0744, "step": 13632 }, { "epoch": 2.5, "learning_rate": 1.609813746827921e-05, "loss": 0.0739, "step": 13633 }, { "epoch": 2.5, "learning_rate": 1.6097549042447844e-05, "loss": 0.0809, "step": 13634 }, { "epoch": 2.5, "learning_rate": 1.609696058300664e-05, "loss": 0.0754, "step": 13635 }, { "epoch": 2.5, "learning_rate": 1.609637208995885e-05, "loss": 0.0778, "step": 13636 }, { "epoch": 2.5, "learning_rate": 1.6095783563307718e-05, "loss": 0.0811, "step": 13637 }, { "epoch": 2.5, "learning_rate": 1.6095195003056475e-05, "loss": 0.073, "step": 13638 }, { "epoch": 2.5, "learning_rate": 1.6094606409208383e-05, "loss": 0.0809, "step": 13639 }, { "epoch": 2.5, "learning_rate": 1.6094017781766672e-05, "loss": 0.0728, "step": 13640 }, { "epoch": 2.5, "learning_rate": 1.6093429120734592e-05, "loss": 0.0864, "step": 13641 }, { "epoch": 2.5, "learning_rate": 1.6092840426115393e-05, "loss": 0.0599, "step": 13642 }, { "epoch": 2.5, "learning_rate": 1.609225169791231e-05, "loss": 0.0715, "step": 13643 }, { "epoch": 2.5, "learning_rate": 1.6091662936128596e-05, "loss": 0.0824, "step": 13644 }, { "epoch": 2.5, "learning_rate": 1.6091074140767492e-05, "loss": 0.0774, "step": 13645 }, { "epoch": 2.5, "learning_rate": 1.6090485311832246e-05, "loss": 0.0728, "step": 13646 }, { "epoch": 2.5, "learning_rate": 1.60898964493261e-05, "loss": 0.0716, "step": 13647 }, { "epoch": 2.5, "learning_rate": 1.6089307553252303e-05, "loss": 0.0702, "step": 13648 }, { "epoch": 2.5, "learning_rate": 1.60887186236141e-05, "loss": 0.0732, "step": 13649 }, { "epoch": 2.5, "learning_rate": 1.608812966041474e-05, "loss": 0.0768, "step": 13650 }, { "epoch": 2.5, "learning_rate": 1.608754066365746e-05, "loss": 0.0669, "step": 13651 }, { "epoch": 2.5, "learning_rate": 1.6086951633345518e-05, "loss": 0.0794, "step": 13652 }, { "epoch": 2.5, "learning_rate": 1.6086362569482156e-05, "loss": 0.0711, "step": 13653 }, { "epoch": 2.5, "learning_rate": 1.6085773472070622e-05, "loss": 0.0735, "step": 13654 }, { "epoch": 2.5, "learning_rate": 1.608518434111416e-05, "loss": 0.078, "step": 13655 }, { "epoch": 2.5, "learning_rate": 1.6084595176616017e-05, "loss": 0.0671, "step": 13656 }, { "epoch": 2.5, "learning_rate": 1.6084005978579445e-05, "loss": 0.0711, "step": 13657 }, { "epoch": 2.5, "learning_rate": 1.608341674700769e-05, "loss": 0.079, "step": 13658 }, { "epoch": 2.5, "learning_rate": 1.6082827481904e-05, "loss": 0.0723, "step": 13659 }, { "epoch": 2.51, "learning_rate": 1.6082238183271625e-05, "loss": 0.0863, "step": 13660 }, { "epoch": 2.51, "learning_rate": 1.6081648851113806e-05, "loss": 0.0723, "step": 13661 }, { "epoch": 2.51, "learning_rate": 1.6081059485433798e-05, "loss": 0.0646, "step": 13662 }, { "epoch": 2.51, "learning_rate": 1.6080470086234846e-05, "loss": 0.0885, "step": 13663 }, { "epoch": 2.51, "learning_rate": 1.60798806535202e-05, "loss": 0.0724, "step": 13664 }, { "epoch": 2.51, "learning_rate": 1.6079291187293112e-05, "loss": 0.0739, "step": 13665 }, { "epoch": 2.51, "learning_rate": 1.607870168755683e-05, "loss": 0.0761, "step": 13666 }, { "epoch": 2.51, "learning_rate": 1.6078112154314597e-05, "loss": 0.0689, "step": 13667 }, { "epoch": 2.51, "learning_rate": 1.6077522587569673e-05, "loss": 0.0785, "step": 13668 }, { "epoch": 2.51, "learning_rate": 1.60769329873253e-05, "loss": 0.0817, "step": 13669 }, { "epoch": 2.51, "learning_rate": 1.6076343353584727e-05, "loss": 0.0793, "step": 13670 }, { "epoch": 2.51, "learning_rate": 1.6075753686351205e-05, "loss": 0.0834, "step": 13671 }, { "epoch": 2.51, "learning_rate": 1.6075163985627992e-05, "loss": 0.0785, "step": 13672 }, { "epoch": 2.51, "learning_rate": 1.607457425141833e-05, "loss": 0.064, "step": 13673 }, { "epoch": 2.51, "learning_rate": 1.6073984483725475e-05, "loss": 0.0833, "step": 13674 }, { "epoch": 2.51, "learning_rate": 1.6073394682552676e-05, "loss": 0.0844, "step": 13675 }, { "epoch": 2.51, "learning_rate": 1.607280484790318e-05, "loss": 0.065, "step": 13676 }, { "epoch": 2.51, "learning_rate": 1.6072214979780242e-05, "loss": 0.0772, "step": 13677 }, { "epoch": 2.51, "learning_rate": 1.6071625078187113e-05, "loss": 0.0754, "step": 13678 }, { "epoch": 2.51, "learning_rate": 1.6071035143127045e-05, "loss": 0.077, "step": 13679 }, { "epoch": 2.51, "learning_rate": 1.6070445174603287e-05, "loss": 0.074, "step": 13680 }, { "epoch": 2.51, "learning_rate": 1.6069855172619098e-05, "loss": 0.0882, "step": 13681 }, { "epoch": 2.51, "learning_rate": 1.606926513717772e-05, "loss": 0.0767, "step": 13682 }, { "epoch": 2.51, "learning_rate": 1.606867506828241e-05, "loss": 0.0793, "step": 13683 }, { "epoch": 2.51, "learning_rate": 1.6068084965936426e-05, "loss": 0.0761, "step": 13684 }, { "epoch": 2.51, "learning_rate": 1.6067494830143014e-05, "loss": 0.0743, "step": 13685 }, { "epoch": 2.51, "learning_rate": 1.6066904660905424e-05, "loss": 0.0903, "step": 13686 }, { "epoch": 2.51, "learning_rate": 1.6066314458226915e-05, "loss": 0.0848, "step": 13687 }, { "epoch": 2.51, "learning_rate": 1.606572422211074e-05, "loss": 0.0801, "step": 13688 }, { "epoch": 2.51, "learning_rate": 1.6065133952560154e-05, "loss": 0.0932, "step": 13689 }, { "epoch": 2.51, "learning_rate": 1.6064543649578403e-05, "loss": 0.08, "step": 13690 }, { "epoch": 2.51, "learning_rate": 1.6063953313168747e-05, "loss": 0.0728, "step": 13691 }, { "epoch": 2.51, "learning_rate": 1.6063362943334436e-05, "loss": 0.075, "step": 13692 }, { "epoch": 2.51, "learning_rate": 1.606277254007873e-05, "loss": 0.0799, "step": 13693 }, { "epoch": 2.51, "learning_rate": 1.606218210340488e-05, "loss": 0.0724, "step": 13694 }, { "epoch": 2.51, "learning_rate": 1.6061591633316137e-05, "loss": 0.088, "step": 13695 }, { "epoch": 2.51, "learning_rate": 1.6061001129815758e-05, "loss": 0.0668, "step": 13696 }, { "epoch": 2.51, "learning_rate": 1.6060410592907004e-05, "loss": 0.0687, "step": 13697 }, { "epoch": 2.51, "learning_rate": 1.605982002259312e-05, "loss": 0.0735, "step": 13698 }, { "epoch": 2.51, "learning_rate": 1.605922941887737e-05, "loss": 0.0811, "step": 13699 }, { "epoch": 2.51, "learning_rate": 1.6058638781763e-05, "loss": 0.0742, "step": 13700 }, { "epoch": 2.51, "learning_rate": 1.6058048111253274e-05, "loss": 0.0729, "step": 13701 }, { "epoch": 2.51, "learning_rate": 1.6057457407351445e-05, "loss": 0.0761, "step": 13702 }, { "epoch": 2.51, "learning_rate": 1.605686667006077e-05, "loss": 0.0697, "step": 13703 }, { "epoch": 2.51, "learning_rate": 1.6056275899384502e-05, "loss": 0.0659, "step": 13704 }, { "epoch": 2.51, "learning_rate": 1.6055685095325903e-05, "loss": 0.071, "step": 13705 }, { "epoch": 2.51, "learning_rate": 1.6055094257888222e-05, "loss": 0.0844, "step": 13706 }, { "epoch": 2.51, "learning_rate": 1.6054503387074723e-05, "loss": 0.0688, "step": 13707 }, { "epoch": 2.51, "learning_rate": 1.6053912482888657e-05, "loss": 0.0718, "step": 13708 }, { "epoch": 2.51, "learning_rate": 1.605332154533328e-05, "loss": 0.0677, "step": 13709 }, { "epoch": 2.51, "learning_rate": 1.605273057441186e-05, "loss": 0.0713, "step": 13710 }, { "epoch": 2.51, "learning_rate": 1.6052139570127643e-05, "loss": 0.0735, "step": 13711 }, { "epoch": 2.51, "learning_rate": 1.6051548532483893e-05, "loss": 0.0744, "step": 13712 }, { "epoch": 2.51, "learning_rate": 1.6050957461483863e-05, "loss": 0.0762, "step": 13713 }, { "epoch": 2.51, "learning_rate": 1.6050366357130817e-05, "loss": 0.0747, "step": 13714 }, { "epoch": 2.52, "learning_rate": 1.604977521942801e-05, "loss": 0.0669, "step": 13715 }, { "epoch": 2.52, "learning_rate": 1.6049184048378695e-05, "loss": 0.063, "step": 13716 }, { "epoch": 2.52, "learning_rate": 1.6048592843986145e-05, "loss": 0.0822, "step": 13717 }, { "epoch": 2.52, "learning_rate": 1.6048001606253602e-05, "loss": 0.079, "step": 13718 }, { "epoch": 2.52, "learning_rate": 1.6047410335184333e-05, "loss": 0.0694, "step": 13719 }, { "epoch": 2.52, "learning_rate": 1.6046819030781596e-05, "loss": 0.0844, "step": 13720 }, { "epoch": 2.52, "learning_rate": 1.6046227693048655e-05, "loss": 0.078, "step": 13721 }, { "epoch": 2.52, "learning_rate": 1.6045636321988763e-05, "loss": 0.0802, "step": 13722 }, { "epoch": 2.52, "learning_rate": 1.604504491760518e-05, "loss": 0.0718, "step": 13723 }, { "epoch": 2.52, "learning_rate": 1.6044453479901173e-05, "loss": 0.0807, "step": 13724 }, { "epoch": 2.52, "learning_rate": 1.6043862008879995e-05, "loss": 0.08, "step": 13725 }, { "epoch": 2.52, "learning_rate": 1.6043270504544906e-05, "loss": 0.0724, "step": 13726 }, { "epoch": 2.52, "learning_rate": 1.604267896689917e-05, "loss": 0.0792, "step": 13727 }, { "epoch": 2.52, "learning_rate": 1.6042087395946044e-05, "loss": 0.0736, "step": 13728 }, { "epoch": 2.52, "learning_rate": 1.6041495791688792e-05, "loss": 0.0843, "step": 13729 }, { "epoch": 2.52, "learning_rate": 1.6040904154130674e-05, "loss": 0.0733, "step": 13730 }, { "epoch": 2.52, "learning_rate": 1.604031248327495e-05, "loss": 0.0702, "step": 13731 }, { "epoch": 2.52, "learning_rate": 1.6039720779124882e-05, "loss": 0.0804, "step": 13732 }, { "epoch": 2.52, "learning_rate": 1.603912904168373e-05, "loss": 0.0787, "step": 13733 }, { "epoch": 2.52, "learning_rate": 1.603853727095476e-05, "loss": 0.0624, "step": 13734 }, { "epoch": 2.52, "learning_rate": 1.603794546694123e-05, "loss": 0.0788, "step": 13735 }, { "epoch": 2.52, "learning_rate": 1.6037353629646405e-05, "loss": 0.0694, "step": 13736 }, { "epoch": 2.52, "learning_rate": 1.603676175907354e-05, "loss": 0.0771, "step": 13737 }, { "epoch": 2.52, "learning_rate": 1.603616985522591e-05, "loss": 0.0789, "step": 13738 }, { "epoch": 2.52, "learning_rate": 1.6035577918106767e-05, "loss": 0.0746, "step": 13739 }, { "epoch": 2.52, "learning_rate": 1.6034985947719378e-05, "loss": 0.0662, "step": 13740 }, { "epoch": 2.52, "learning_rate": 1.6034393944067003e-05, "loss": 0.083, "step": 13741 }, { "epoch": 2.52, "learning_rate": 1.6033801907152908e-05, "loss": 0.0723, "step": 13742 }, { "epoch": 2.52, "learning_rate": 1.6033209836980356e-05, "loss": 0.0673, "step": 13743 }, { "epoch": 2.52, "learning_rate": 1.603261773355261e-05, "loss": 0.0816, "step": 13744 }, { "epoch": 2.52, "learning_rate": 1.6032025596872934e-05, "loss": 0.0634, "step": 13745 }, { "epoch": 2.52, "learning_rate": 1.603143342694459e-05, "loss": 0.0761, "step": 13746 }, { "epoch": 2.52, "learning_rate": 1.6030841223770846e-05, "loss": 0.0822, "step": 13747 }, { "epoch": 2.52, "learning_rate": 1.6030248987354965e-05, "loss": 0.0678, "step": 13748 }, { "epoch": 2.52, "learning_rate": 1.6029656717700204e-05, "loss": 0.0737, "step": 13749 }, { "epoch": 2.52, "learning_rate": 1.602906441480984e-05, "loss": 0.073, "step": 13750 }, { "epoch": 2.52, "learning_rate": 1.602847207868713e-05, "loss": 0.0729, "step": 13751 }, { "epoch": 2.52, "learning_rate": 1.602787970933534e-05, "loss": 0.0767, "step": 13752 }, { "epoch": 2.52, "learning_rate": 1.6027287306757735e-05, "loss": 0.0748, "step": 13753 }, { "epoch": 2.52, "learning_rate": 1.6026694870957583e-05, "loss": 0.0683, "step": 13754 }, { "epoch": 2.52, "learning_rate": 1.6026102401938148e-05, "loss": 0.075, "step": 13755 }, { "epoch": 2.52, "learning_rate": 1.602550989970269e-05, "loss": 0.0693, "step": 13756 }, { "epoch": 2.52, "learning_rate": 1.6024917364254484e-05, "loss": 0.0779, "step": 13757 }, { "epoch": 2.52, "learning_rate": 1.6024324795596794e-05, "loss": 0.0778, "step": 13758 }, { "epoch": 2.52, "learning_rate": 1.6023732193732886e-05, "loss": 0.0734, "step": 13759 }, { "epoch": 2.52, "learning_rate": 1.602313955866602e-05, "loss": 0.0717, "step": 13760 }, { "epoch": 2.52, "learning_rate": 1.6022546890399473e-05, "loss": 0.0683, "step": 13761 }, { "epoch": 2.52, "learning_rate": 1.60219541889365e-05, "loss": 0.0676, "step": 13762 }, { "epoch": 2.52, "learning_rate": 1.602136145428038e-05, "loss": 0.0696, "step": 13763 }, { "epoch": 2.52, "learning_rate": 1.6020768686434372e-05, "loss": 0.0745, "step": 13764 }, { "epoch": 2.52, "learning_rate": 1.602017588540175e-05, "loss": 0.0828, "step": 13765 }, { "epoch": 2.52, "learning_rate": 1.601958305118577e-05, "loss": 0.0711, "step": 13766 }, { "epoch": 2.52, "learning_rate": 1.6018990183789713e-05, "loss": 0.0779, "step": 13767 }, { "epoch": 2.52, "learning_rate": 1.601839728321684e-05, "loss": 0.0858, "step": 13768 }, { "epoch": 2.53, "learning_rate": 1.6017804349470417e-05, "loss": 0.0679, "step": 13769 }, { "epoch": 2.53, "learning_rate": 1.601721138255372e-05, "loss": 0.0845, "step": 13770 }, { "epoch": 2.53, "learning_rate": 1.6016618382470014e-05, "loss": 0.0781, "step": 13771 }, { "epoch": 2.53, "learning_rate": 1.6016025349222562e-05, "loss": 0.0625, "step": 13772 }, { "epoch": 2.53, "learning_rate": 1.6015432282814635e-05, "loss": 0.0791, "step": 13773 }, { "epoch": 2.53, "learning_rate": 1.601483918324951e-05, "loss": 0.0735, "step": 13774 }, { "epoch": 2.53, "learning_rate": 1.6014246050530447e-05, "loss": 0.0831, "step": 13775 }, { "epoch": 2.53, "learning_rate": 1.6013652884660723e-05, "loss": 0.0848, "step": 13776 }, { "epoch": 2.53, "learning_rate": 1.6013059685643598e-05, "loss": 0.0712, "step": 13777 }, { "epoch": 2.53, "learning_rate": 1.601246645348235e-05, "loss": 0.0771, "step": 13778 }, { "epoch": 2.53, "learning_rate": 1.601187318818025e-05, "loss": 0.0836, "step": 13779 }, { "epoch": 2.53, "learning_rate": 1.6011279889740558e-05, "loss": 0.0761, "step": 13780 }, { "epoch": 2.53, "learning_rate": 1.6010686558166554e-05, "loss": 0.0639, "step": 13781 }, { "epoch": 2.53, "learning_rate": 1.6010093193461504e-05, "loss": 0.0731, "step": 13782 }, { "epoch": 2.53, "learning_rate": 1.6009499795628676e-05, "loss": 0.0854, "step": 13783 }, { "epoch": 2.53, "learning_rate": 1.6008906364671348e-05, "loss": 0.0887, "step": 13784 }, { "epoch": 2.53, "learning_rate": 1.600831290059279e-05, "loss": 0.0754, "step": 13785 }, { "epoch": 2.53, "learning_rate": 1.6007719403396264e-05, "loss": 0.0749, "step": 13786 }, { "epoch": 2.53, "learning_rate": 1.600712587308505e-05, "loss": 0.0703, "step": 13787 }, { "epoch": 2.53, "learning_rate": 1.6006532309662422e-05, "loss": 0.0702, "step": 13788 }, { "epoch": 2.53, "learning_rate": 1.6005938713131644e-05, "loss": 0.0678, "step": 13789 }, { "epoch": 2.53, "learning_rate": 1.6005345083495992e-05, "loss": 0.0641, "step": 13790 }, { "epoch": 2.53, "learning_rate": 1.6004751420758733e-05, "loss": 0.0674, "step": 13791 }, { "epoch": 2.53, "learning_rate": 1.6004157724923147e-05, "loss": 0.0864, "step": 13792 }, { "epoch": 2.53, "learning_rate": 1.6003563995992503e-05, "loss": 0.0805, "step": 13793 }, { "epoch": 2.53, "learning_rate": 1.6002970233970074e-05, "loss": 0.0817, "step": 13794 }, { "epoch": 2.53, "learning_rate": 1.600237643885913e-05, "loss": 0.0753, "step": 13795 }, { "epoch": 2.53, "learning_rate": 1.600178261066295e-05, "loss": 0.0715, "step": 13796 }, { "epoch": 2.53, "learning_rate": 1.60011887493848e-05, "loss": 0.0693, "step": 13797 }, { "epoch": 2.53, "learning_rate": 1.600059485502796e-05, "loss": 0.0763, "step": 13798 }, { "epoch": 2.53, "learning_rate": 1.6000000927595697e-05, "loss": 0.0707, "step": 13799 }, { "epoch": 2.53, "learning_rate": 1.5999406967091293e-05, "loss": 0.0649, "step": 13800 }, { "epoch": 2.53, "learning_rate": 1.5998812973518015e-05, "loss": 0.0734, "step": 13801 }, { "epoch": 2.53, "learning_rate": 1.599821894687914e-05, "loss": 0.0643, "step": 13802 }, { "epoch": 2.53, "learning_rate": 1.599762488717794e-05, "loss": 0.0684, "step": 13803 }, { "epoch": 2.53, "learning_rate": 1.599703079441769e-05, "loss": 0.0641, "step": 13804 }, { "epoch": 2.53, "learning_rate": 1.599643666860167e-05, "loss": 0.0841, "step": 13805 }, { "epoch": 2.53, "learning_rate": 1.5995842509733147e-05, "loss": 0.0689, "step": 13806 }, { "epoch": 2.53, "learning_rate": 1.5995248317815402e-05, "loss": 0.0752, "step": 13807 }, { "epoch": 2.53, "learning_rate": 1.5994654092851706e-05, "loss": 0.0865, "step": 13808 }, { "epoch": 2.53, "learning_rate": 1.5994059834845337e-05, "loss": 0.0771, "step": 13809 }, { "epoch": 2.53, "learning_rate": 1.599346554379957e-05, "loss": 0.0665, "step": 13810 }, { "epoch": 2.53, "learning_rate": 1.599287121971768e-05, "loss": 0.0831, "step": 13811 }, { "epoch": 2.53, "learning_rate": 1.5992276862602945e-05, "loss": 0.0741, "step": 13812 }, { "epoch": 2.53, "learning_rate": 1.5991682472458636e-05, "loss": 0.0677, "step": 13813 }, { "epoch": 2.53, "learning_rate": 1.5991088049288035e-05, "loss": 0.0809, "step": 13814 }, { "epoch": 2.53, "learning_rate": 1.5990493593094417e-05, "loss": 0.0653, "step": 13815 }, { "epoch": 2.53, "learning_rate": 1.598989910388106e-05, "loss": 0.0832, "step": 13816 }, { "epoch": 2.53, "learning_rate": 1.5989304581651234e-05, "loss": 0.0787, "step": 13817 }, { "epoch": 2.53, "learning_rate": 1.5988710026408223e-05, "loss": 0.0762, "step": 13818 }, { "epoch": 2.53, "learning_rate": 1.59881154381553e-05, "loss": 0.0736, "step": 13819 }, { "epoch": 2.53, "learning_rate": 1.598752081689575e-05, "loss": 0.067, "step": 13820 }, { "epoch": 2.53, "learning_rate": 1.598692616263284e-05, "loss": 0.0716, "step": 13821 }, { "epoch": 2.53, "learning_rate": 1.5986331475369853e-05, "loss": 0.0817, "step": 13822 }, { "epoch": 2.53, "learning_rate": 1.598573675511007e-05, "loss": 0.0631, "step": 13823 }, { "epoch": 2.54, "learning_rate": 1.5985142001856763e-05, "loss": 0.0699, "step": 13824 }, { "epoch": 2.54, "learning_rate": 1.5984547215613214e-05, "loss": 0.0709, "step": 13825 }, { "epoch": 2.54, "learning_rate": 1.5983952396382698e-05, "loss": 0.0731, "step": 13826 }, { "epoch": 2.54, "learning_rate": 1.59833575441685e-05, "loss": 0.0718, "step": 13827 }, { "epoch": 2.54, "learning_rate": 1.5982762658973894e-05, "loss": 0.0723, "step": 13828 }, { "epoch": 2.54, "learning_rate": 1.598216774080216e-05, "loss": 0.0686, "step": 13829 }, { "epoch": 2.54, "learning_rate": 1.5981572789656575e-05, "loss": 0.0678, "step": 13830 }, { "epoch": 2.54, "learning_rate": 1.5980977805540424e-05, "loss": 0.0677, "step": 13831 }, { "epoch": 2.54, "learning_rate": 1.598038278845698e-05, "loss": 0.0696, "step": 13832 }, { "epoch": 2.54, "learning_rate": 1.5979787738409527e-05, "loss": 0.0645, "step": 13833 }, { "epoch": 2.54, "learning_rate": 1.5979192655401344e-05, "loss": 0.0689, "step": 13834 }, { "epoch": 2.54, "learning_rate": 1.5978597539435712e-05, "loss": 0.0678, "step": 13835 }, { "epoch": 2.54, "learning_rate": 1.597800239051591e-05, "loss": 0.0673, "step": 13836 }, { "epoch": 2.54, "learning_rate": 1.5977407208645218e-05, "loss": 0.0771, "step": 13837 }, { "epoch": 2.54, "learning_rate": 1.597681199382692e-05, "loss": 0.0749, "step": 13838 }, { "epoch": 2.54, "learning_rate": 1.5976216746064294e-05, "loss": 0.0733, "step": 13839 }, { "epoch": 2.54, "learning_rate": 1.597562146536062e-05, "loss": 0.077, "step": 13840 }, { "epoch": 2.54, "learning_rate": 1.5975026151719176e-05, "loss": 0.0686, "step": 13841 }, { "epoch": 2.54, "learning_rate": 1.5974430805143254e-05, "loss": 0.0836, "step": 13842 }, { "epoch": 2.54, "learning_rate": 1.5973835425636128e-05, "loss": 0.0733, "step": 13843 }, { "epoch": 2.54, "learning_rate": 1.597324001320108e-05, "loss": 0.0692, "step": 13844 }, { "epoch": 2.54, "learning_rate": 1.5972644567841395e-05, "loss": 0.064, "step": 13845 }, { "epoch": 2.54, "learning_rate": 1.597204908956035e-05, "loss": 0.0738, "step": 13846 }, { "epoch": 2.54, "learning_rate": 1.5971453578361234e-05, "loss": 0.0789, "step": 13847 }, { "epoch": 2.54, "learning_rate": 1.5970858034247326e-05, "loss": 0.074, "step": 13848 }, { "epoch": 2.54, "learning_rate": 1.5970262457221905e-05, "loss": 0.0745, "step": 13849 }, { "epoch": 2.54, "learning_rate": 1.5969666847288257e-05, "loss": 0.0575, "step": 13850 }, { "epoch": 2.54, "learning_rate": 1.5969071204449667e-05, "loss": 0.0772, "step": 13851 }, { "epoch": 2.54, "learning_rate": 1.596847552870942e-05, "loss": 0.0766, "step": 13852 }, { "epoch": 2.54, "learning_rate": 1.5967879820070797e-05, "loss": 0.0779, "step": 13853 }, { "epoch": 2.54, "learning_rate": 1.5967284078537076e-05, "loss": 0.0747, "step": 13854 }, { "epoch": 2.54, "learning_rate": 1.5966688304111545e-05, "loss": 0.0806, "step": 13855 }, { "epoch": 2.54, "learning_rate": 1.5966092496797493e-05, "loss": 0.0759, "step": 13856 }, { "epoch": 2.54, "learning_rate": 1.5965496656598193e-05, "loss": 0.0836, "step": 13857 }, { "epoch": 2.54, "learning_rate": 1.596490078351694e-05, "loss": 0.0718, "step": 13858 }, { "epoch": 2.54, "learning_rate": 1.5964304877557013e-05, "loss": 0.0676, "step": 13859 }, { "epoch": 2.54, "learning_rate": 1.5963708938721702e-05, "loss": 0.0852, "step": 13860 }, { "epoch": 2.54, "learning_rate": 1.596311296701428e-05, "loss": 0.0668, "step": 13861 }, { "epoch": 2.54, "learning_rate": 1.596251696243804e-05, "loss": 0.067, "step": 13862 }, { "epoch": 2.54, "learning_rate": 1.5961920924996274e-05, "loss": 0.0706, "step": 13863 }, { "epoch": 2.54, "learning_rate": 1.5961324854692254e-05, "loss": 0.091, "step": 13864 }, { "epoch": 2.54, "learning_rate": 1.5960728751529275e-05, "loss": 0.0771, "step": 13865 }, { "epoch": 2.54, "learning_rate": 1.596013261551062e-05, "loss": 0.0644, "step": 13866 }, { "epoch": 2.54, "learning_rate": 1.5959536446639572e-05, "loss": 0.079, "step": 13867 }, { "epoch": 2.54, "learning_rate": 1.5958940244919418e-05, "loss": 0.0589, "step": 13868 }, { "epoch": 2.54, "learning_rate": 1.595834401035345e-05, "loss": 0.0872, "step": 13869 }, { "epoch": 2.54, "learning_rate": 1.5957747742944948e-05, "loss": 0.0699, "step": 13870 }, { "epoch": 2.54, "learning_rate": 1.5957151442697202e-05, "loss": 0.0728, "step": 13871 }, { "epoch": 2.54, "learning_rate": 1.5956555109613496e-05, "loss": 0.0661, "step": 13872 }, { "epoch": 2.54, "learning_rate": 1.595595874369712e-05, "loss": 0.0809, "step": 13873 }, { "epoch": 2.54, "learning_rate": 1.5955362344951358e-05, "loss": 0.0709, "step": 13874 }, { "epoch": 2.54, "learning_rate": 1.5954765913379503e-05, "loss": 0.0768, "step": 13875 }, { "epoch": 2.54, "learning_rate": 1.5954169448984836e-05, "loss": 0.0786, "step": 13876 }, { "epoch": 2.54, "learning_rate": 1.5953572951770647e-05, "loss": 0.0724, "step": 13877 }, { "epoch": 2.55, "learning_rate": 1.5952976421740224e-05, "loss": 0.0763, "step": 13878 }, { "epoch": 2.55, "learning_rate": 1.5952379858896856e-05, "loss": 0.0804, "step": 13879 }, { "epoch": 2.55, "learning_rate": 1.5951783263243836e-05, "loss": 0.0738, "step": 13880 }, { "epoch": 2.55, "learning_rate": 1.595118663478444e-05, "loss": 0.0862, "step": 13881 }, { "epoch": 2.55, "learning_rate": 1.595058997352197e-05, "loss": 0.0699, "step": 13882 }, { "epoch": 2.55, "learning_rate": 1.5949993279459702e-05, "loss": 0.0885, "step": 13883 }, { "epoch": 2.55, "learning_rate": 1.5949396552600936e-05, "loss": 0.0759, "step": 13884 }, { "epoch": 2.55, "learning_rate": 1.5948799792948954e-05, "loss": 0.0778, "step": 13885 }, { "epoch": 2.55, "learning_rate": 1.5948203000507053e-05, "loss": 0.0757, "step": 13886 }, { "epoch": 2.55, "learning_rate": 1.5947606175278513e-05, "loss": 0.0669, "step": 13887 }, { "epoch": 2.55, "learning_rate": 1.5947009317266632e-05, "loss": 0.0776, "step": 13888 }, { "epoch": 2.55, "learning_rate": 1.5946412426474692e-05, "loss": 0.0784, "step": 13889 }, { "epoch": 2.55, "learning_rate": 1.594581550290599e-05, "loss": 0.0592, "step": 13890 }, { "epoch": 2.55, "learning_rate": 1.5945218546563815e-05, "loss": 0.0733, "step": 13891 }, { "epoch": 2.55, "learning_rate": 1.5944621557451455e-05, "loss": 0.0736, "step": 13892 }, { "epoch": 2.55, "learning_rate": 1.59440245355722e-05, "loss": 0.0674, "step": 13893 }, { "epoch": 2.55, "learning_rate": 1.5943427480929344e-05, "loss": 0.0718, "step": 13894 }, { "epoch": 2.55, "learning_rate": 1.5942830393526176e-05, "loss": 0.0772, "step": 13895 }, { "epoch": 2.55, "learning_rate": 1.5942233273365987e-05, "loss": 0.063, "step": 13896 }, { "epoch": 2.55, "learning_rate": 1.594163612045207e-05, "loss": 0.088, "step": 13897 }, { "epoch": 2.55, "learning_rate": 1.5941038934787715e-05, "loss": 0.0726, "step": 13898 }, { "epoch": 2.55, "learning_rate": 1.5940441716376217e-05, "loss": 0.0817, "step": 13899 }, { "epoch": 2.55, "learning_rate": 1.5939844465220862e-05, "loss": 0.0741, "step": 13900 }, { "epoch": 2.55, "learning_rate": 1.5939247181324943e-05, "loss": 0.063, "step": 13901 }, { "epoch": 2.55, "learning_rate": 1.593864986469176e-05, "loss": 0.0744, "step": 13902 }, { "epoch": 2.55, "learning_rate": 1.5938052515324597e-05, "loss": 0.0696, "step": 13903 }, { "epoch": 2.55, "learning_rate": 1.5937455133226747e-05, "loss": 0.0719, "step": 13904 }, { "epoch": 2.55, "learning_rate": 1.593685771840151e-05, "loss": 0.0609, "step": 13905 }, { "epoch": 2.55, "learning_rate": 1.5936260270852174e-05, "loss": 0.0638, "step": 13906 }, { "epoch": 2.55, "learning_rate": 1.593566279058203e-05, "loss": 0.0741, "step": 13907 }, { "epoch": 2.55, "learning_rate": 1.5935065277594374e-05, "loss": 0.0685, "step": 13908 }, { "epoch": 2.55, "learning_rate": 1.5934467731892497e-05, "loss": 0.0685, "step": 13909 }, { "epoch": 2.55, "learning_rate": 1.5933870153479695e-05, "loss": 0.0799, "step": 13910 }, { "epoch": 2.55, "learning_rate": 1.5933272542359267e-05, "loss": 0.0714, "step": 13911 }, { "epoch": 2.55, "learning_rate": 1.5932674898534495e-05, "loss": 0.0779, "step": 13912 }, { "epoch": 2.55, "learning_rate": 1.5932077222008687e-05, "loss": 0.0779, "step": 13913 }, { "epoch": 2.55, "learning_rate": 1.593147951278512e-05, "loss": 0.0723, "step": 13914 }, { "epoch": 2.55, "learning_rate": 1.593088177086711e-05, "loss": 0.0703, "step": 13915 }, { "epoch": 2.55, "learning_rate": 1.5930283996257936e-05, "loss": 0.0762, "step": 13916 }, { "epoch": 2.55, "learning_rate": 1.5929686188960897e-05, "loss": 0.0704, "step": 13917 }, { "epoch": 2.55, "learning_rate": 1.5929088348979292e-05, "loss": 0.0711, "step": 13918 }, { "epoch": 2.55, "learning_rate": 1.592849047631641e-05, "loss": 0.0685, "step": 13919 }, { "epoch": 2.55, "learning_rate": 1.592789257097555e-05, "loss": 0.076, "step": 13920 }, { "epoch": 2.55, "learning_rate": 1.5927294632960006e-05, "loss": 0.0781, "step": 13921 }, { "epoch": 2.55, "learning_rate": 1.592669666227308e-05, "loss": 0.0717, "step": 13922 }, { "epoch": 2.55, "learning_rate": 1.5926098658918057e-05, "loss": 0.0623, "step": 13923 }, { "epoch": 2.55, "learning_rate": 1.5925500622898243e-05, "loss": 0.0778, "step": 13924 }, { "epoch": 2.55, "learning_rate": 1.592490255421693e-05, "loss": 0.0705, "step": 13925 }, { "epoch": 2.55, "learning_rate": 1.5924304452877417e-05, "loss": 0.0723, "step": 13926 }, { "epoch": 2.55, "learning_rate": 1.5923706318882996e-05, "loss": 0.0819, "step": 13927 }, { "epoch": 2.55, "learning_rate": 1.592310815223697e-05, "loss": 0.076, "step": 13928 }, { "epoch": 2.55, "learning_rate": 1.592250995294263e-05, "loss": 0.0785, "step": 13929 }, { "epoch": 2.55, "learning_rate": 1.5921911721003282e-05, "loss": 0.0909, "step": 13930 }, { "epoch": 2.55, "learning_rate": 1.592131345642221e-05, "loss": 0.0677, "step": 13931 }, { "epoch": 2.55, "learning_rate": 1.5920715159202725e-05, "loss": 0.0789, "step": 13932 }, { "epoch": 2.56, "learning_rate": 1.5920116829348116e-05, "loss": 0.0762, "step": 13933 }, { "epoch": 2.56, "learning_rate": 1.591951846686169e-05, "loss": 0.0683, "step": 13934 }, { "epoch": 2.56, "learning_rate": 1.5918920071746732e-05, "loss": 0.069, "step": 13935 }, { "epoch": 2.56, "learning_rate": 1.591832164400655e-05, "loss": 0.073, "step": 13936 }, { "epoch": 2.56, "learning_rate": 1.5917723183644446e-05, "loss": 0.0639, "step": 13937 }, { "epoch": 2.56, "learning_rate": 1.5917124690663705e-05, "loss": 0.0706, "step": 13938 }, { "epoch": 2.56, "learning_rate": 1.591652616506764e-05, "loss": 0.0629, "step": 13939 }, { "epoch": 2.56, "learning_rate": 1.5915927606859542e-05, "loss": 0.0655, "step": 13940 }, { "epoch": 2.56, "learning_rate": 1.5915329016042712e-05, "loss": 0.0763, "step": 13941 }, { "epoch": 2.56, "learning_rate": 1.5914730392620453e-05, "loss": 0.0844, "step": 13942 }, { "epoch": 2.56, "learning_rate": 1.591413173659606e-05, "loss": 0.0676, "step": 13943 }, { "epoch": 2.56, "learning_rate": 1.5913533047972832e-05, "loss": 0.0799, "step": 13944 }, { "epoch": 2.56, "learning_rate": 1.5912934326754074e-05, "loss": 0.0857, "step": 13945 }, { "epoch": 2.56, "learning_rate": 1.591233557294308e-05, "loss": 0.0647, "step": 13946 }, { "epoch": 2.56, "learning_rate": 1.5911736786543157e-05, "loss": 0.075, "step": 13947 }, { "epoch": 2.56, "learning_rate": 1.5911137967557603e-05, "loss": 0.0716, "step": 13948 }, { "epoch": 2.56, "learning_rate": 1.5910539115989717e-05, "loss": 0.0694, "step": 13949 }, { "epoch": 2.56, "learning_rate": 1.59099402318428e-05, "loss": 0.0823, "step": 13950 }, { "epoch": 2.56, "learning_rate": 1.5909341315120155e-05, "loss": 0.079, "step": 13951 }, { "epoch": 2.56, "learning_rate": 1.5908742365825077e-05, "loss": 0.068, "step": 13952 }, { "epoch": 2.56, "learning_rate": 1.590814338396088e-05, "loss": 0.0661, "step": 13953 }, { "epoch": 2.56, "learning_rate": 1.5907544369530853e-05, "loss": 0.0753, "step": 13954 }, { "epoch": 2.56, "learning_rate": 1.590694532253831e-05, "loss": 0.0636, "step": 13955 }, { "epoch": 2.56, "learning_rate": 1.590634624298654e-05, "loss": 0.08, "step": 13956 }, { "epoch": 2.56, "learning_rate": 1.5905747130878853e-05, "loss": 0.0847, "step": 13957 }, { "epoch": 2.56, "learning_rate": 1.5905147986218546e-05, "loss": 0.0709, "step": 13958 }, { "epoch": 2.56, "learning_rate": 1.590454880900893e-05, "loss": 0.0623, "step": 13959 }, { "epoch": 2.56, "learning_rate": 1.5903949599253302e-05, "loss": 0.068, "step": 13960 }, { "epoch": 2.56, "learning_rate": 1.5903350356954965e-05, "loss": 0.068, "step": 13961 }, { "epoch": 2.56, "learning_rate": 1.590275108211722e-05, "loss": 0.067, "step": 13962 }, { "epoch": 2.56, "learning_rate": 1.5902151774743372e-05, "loss": 0.0801, "step": 13963 }, { "epoch": 2.56, "learning_rate": 1.5901552434836733e-05, "loss": 0.0715, "step": 13964 }, { "epoch": 2.56, "learning_rate": 1.590095306240059e-05, "loss": 0.0738, "step": 13965 }, { "epoch": 2.56, "learning_rate": 1.590035365743826e-05, "loss": 0.073, "step": 13966 }, { "epoch": 2.56, "learning_rate": 1.589975421995304e-05, "loss": 0.0719, "step": 13967 }, { "epoch": 2.56, "learning_rate": 1.589915474994824e-05, "loss": 0.0825, "step": 13968 }, { "epoch": 2.56, "learning_rate": 1.5898555247427158e-05, "loss": 0.0816, "step": 13969 }, { "epoch": 2.56, "learning_rate": 1.5897955712393102e-05, "loss": 0.072, "step": 13970 }, { "epoch": 2.56, "learning_rate": 1.5897356144849374e-05, "loss": 0.0674, "step": 13971 }, { "epoch": 2.56, "learning_rate": 1.5896756544799282e-05, "loss": 0.0754, "step": 13972 }, { "epoch": 2.56, "learning_rate": 1.589615691224613e-05, "loss": 0.0732, "step": 13973 }, { "epoch": 2.56, "learning_rate": 1.5895557247193226e-05, "loss": 0.0664, "step": 13974 }, { "epoch": 2.56, "learning_rate": 1.589495754964387e-05, "loss": 0.076, "step": 13975 }, { "epoch": 2.56, "learning_rate": 1.5894357819601366e-05, "loss": 0.0757, "step": 13976 }, { "epoch": 2.56, "learning_rate": 1.589375805706903e-05, "loss": 0.0692, "step": 13977 }, { "epoch": 2.56, "learning_rate": 1.5893158262050158e-05, "loss": 0.0772, "step": 13978 }, { "epoch": 2.56, "learning_rate": 1.589255843454806e-05, "loss": 0.069, "step": 13979 }, { "epoch": 2.56, "learning_rate": 1.589195857456604e-05, "loss": 0.0712, "step": 13980 }, { "epoch": 2.56, "learning_rate": 1.589135868210741e-05, "loss": 0.0691, "step": 13981 }, { "epoch": 2.56, "learning_rate": 1.5890758757175473e-05, "loss": 0.0725, "step": 13982 }, { "epoch": 2.56, "learning_rate": 1.5890158799773534e-05, "loss": 0.0648, "step": 13983 }, { "epoch": 2.56, "learning_rate": 1.58895588099049e-05, "loss": 0.0768, "step": 13984 }, { "epoch": 2.56, "learning_rate": 1.5888958787572883e-05, "loss": 0.0834, "step": 13985 }, { "epoch": 2.56, "learning_rate": 1.5888358732780783e-05, "loss": 0.0745, "step": 13986 }, { "epoch": 2.57, "learning_rate": 1.5887758645531916e-05, "loss": 0.0747, "step": 13987 }, { "epoch": 2.57, "learning_rate": 1.5887158525829588e-05, "loss": 0.0853, "step": 13988 }, { "epoch": 2.57, "learning_rate": 1.5886558373677095e-05, "loss": 0.0657, "step": 13989 }, { "epoch": 2.57, "learning_rate": 1.5885958189077763e-05, "loss": 0.0684, "step": 13990 }, { "epoch": 2.57, "learning_rate": 1.5885357972034886e-05, "loss": 0.0614, "step": 13991 }, { "epoch": 2.57, "learning_rate": 1.588475772255178e-05, "loss": 0.0806, "step": 13992 }, { "epoch": 2.57, "learning_rate": 1.5884157440631752e-05, "loss": 0.0709, "step": 13993 }, { "epoch": 2.57, "learning_rate": 1.5883557126278112e-05, "loss": 0.0715, "step": 13994 }, { "epoch": 2.57, "learning_rate": 1.5882956779494165e-05, "loss": 0.0781, "step": 13995 }, { "epoch": 2.57, "learning_rate": 1.5882356400283222e-05, "loss": 0.0779, "step": 13996 }, { "epoch": 2.57, "learning_rate": 1.5881755988648594e-05, "loss": 0.072, "step": 13997 }, { "epoch": 2.57, "learning_rate": 1.5881155544593586e-05, "loss": 0.0699, "step": 13998 }, { "epoch": 2.57, "learning_rate": 1.5880555068121515e-05, "loss": 0.0768, "step": 13999 }, { "epoch": 2.57, "learning_rate": 1.5879954559235687e-05, "loss": 0.0634, "step": 14000 }, { "epoch": 2.57, "learning_rate": 1.587935401793941e-05, "loss": 0.0708, "step": 14001 }, { "epoch": 2.57, "learning_rate": 1.5878753444235993e-05, "loss": 0.0765, "step": 14002 }, { "epoch": 2.57, "learning_rate": 1.5878152838128756e-05, "loss": 0.081, "step": 14003 }, { "epoch": 2.57, "learning_rate": 1.5877552199620997e-05, "loss": 0.0784, "step": 14004 }, { "epoch": 2.57, "learning_rate": 1.5876951528716034e-05, "loss": 0.0749, "step": 14005 }, { "epoch": 2.57, "learning_rate": 1.5876350825417177e-05, "loss": 0.0682, "step": 14006 }, { "epoch": 2.57, "learning_rate": 1.5875750089727736e-05, "loss": 0.0657, "step": 14007 }, { "epoch": 2.57, "learning_rate": 1.5875149321651025e-05, "loss": 0.0742, "step": 14008 }, { "epoch": 2.57, "learning_rate": 1.5874548521190353e-05, "loss": 0.0756, "step": 14009 }, { "epoch": 2.57, "learning_rate": 1.587394768834903e-05, "loss": 0.0868, "step": 14010 }, { "epoch": 2.57, "learning_rate": 1.587334682313037e-05, "loss": 0.0701, "step": 14011 }, { "epoch": 2.57, "learning_rate": 1.5872745925537683e-05, "loss": 0.0768, "step": 14012 }, { "epoch": 2.57, "learning_rate": 1.5872144995574287e-05, "loss": 0.0722, "step": 14013 }, { "epoch": 2.57, "learning_rate": 1.5871544033243488e-05, "loss": 0.0882, "step": 14014 }, { "epoch": 2.57, "learning_rate": 1.58709430385486e-05, "loss": 0.0772, "step": 14015 }, { "epoch": 2.57, "learning_rate": 1.5870342011492937e-05, "loss": 0.071, "step": 14016 }, { "epoch": 2.57, "learning_rate": 1.5869740952079813e-05, "loss": 0.0663, "step": 14017 }, { "epoch": 2.57, "learning_rate": 1.5869139860312537e-05, "loss": 0.0754, "step": 14018 }, { "epoch": 2.57, "learning_rate": 1.5868538736194427e-05, "loss": 0.0746, "step": 14019 }, { "epoch": 2.57, "learning_rate": 1.5867937579728792e-05, "loss": 0.0731, "step": 14020 }, { "epoch": 2.57, "learning_rate": 1.5867336390918948e-05, "loss": 0.0851, "step": 14021 }, { "epoch": 2.57, "learning_rate": 1.586673516976821e-05, "loss": 0.0783, "step": 14022 }, { "epoch": 2.57, "learning_rate": 1.5866133916279887e-05, "loss": 0.0754, "step": 14023 }, { "epoch": 2.57, "learning_rate": 1.5865532630457302e-05, "loss": 0.0755, "step": 14024 }, { "epoch": 2.57, "learning_rate": 1.5864931312303757e-05, "loss": 0.0604, "step": 14025 }, { "epoch": 2.57, "learning_rate": 1.5864329961822577e-05, "loss": 0.0584, "step": 14026 }, { "epoch": 2.57, "learning_rate": 1.586372857901707e-05, "loss": 0.0594, "step": 14027 }, { "epoch": 2.57, "learning_rate": 1.5863127163890556e-05, "loss": 0.0828, "step": 14028 }, { "epoch": 2.57, "learning_rate": 1.586252571644635e-05, "loss": 0.0818, "step": 14029 }, { "epoch": 2.57, "learning_rate": 1.5861924236687762e-05, "loss": 0.0715, "step": 14030 }, { "epoch": 2.57, "learning_rate": 1.586132272461811e-05, "loss": 0.0679, "step": 14031 }, { "epoch": 2.57, "learning_rate": 1.5860721180240713e-05, "loss": 0.0849, "step": 14032 }, { "epoch": 2.57, "learning_rate": 1.586011960355888e-05, "loss": 0.0619, "step": 14033 }, { "epoch": 2.57, "learning_rate": 1.5859517994575932e-05, "loss": 0.0704, "step": 14034 }, { "epoch": 2.57, "learning_rate": 1.585891635329518e-05, "loss": 0.0627, "step": 14035 }, { "epoch": 2.57, "learning_rate": 1.5858314679719947e-05, "loss": 0.0675, "step": 14036 }, { "epoch": 2.57, "learning_rate": 1.5857712973853548e-05, "loss": 0.0807, "step": 14037 }, { "epoch": 2.57, "learning_rate": 1.5857111235699296e-05, "loss": 0.0804, "step": 14038 }, { "epoch": 2.57, "learning_rate": 1.5856509465260508e-05, "loss": 0.075, "step": 14039 }, { "epoch": 2.57, "learning_rate": 1.5855907662540505e-05, "loss": 0.0789, "step": 14040 }, { "epoch": 2.57, "learning_rate": 1.58553058275426e-05, "loss": 0.0687, "step": 14041 }, { "epoch": 2.58, "learning_rate": 1.5854703960270113e-05, "loss": 0.0727, "step": 14042 }, { "epoch": 2.58, "learning_rate": 1.585410206072636e-05, "loss": 0.0787, "step": 14043 }, { "epoch": 2.58, "learning_rate": 1.5853500128914664e-05, "loss": 0.0798, "step": 14044 }, { "epoch": 2.58, "learning_rate": 1.585289816483833e-05, "loss": 0.0744, "step": 14045 }, { "epoch": 2.58, "learning_rate": 1.585229616850069e-05, "loss": 0.0795, "step": 14046 }, { "epoch": 2.58, "learning_rate": 1.5851694139905055e-05, "loss": 0.0731, "step": 14047 }, { "epoch": 2.58, "learning_rate": 1.5851092079054743e-05, "loss": 0.0759, "step": 14048 }, { "epoch": 2.58, "learning_rate": 1.5850489985953076e-05, "loss": 0.0853, "step": 14049 }, { "epoch": 2.58, "learning_rate": 1.5849887860603374e-05, "loss": 0.0667, "step": 14050 }, { "epoch": 2.58, "learning_rate": 1.5849285703008947e-05, "loss": 0.0756, "step": 14051 }, { "epoch": 2.58, "learning_rate": 1.5848683513173125e-05, "loss": 0.0748, "step": 14052 }, { "epoch": 2.58, "learning_rate": 1.5848081291099223e-05, "loss": 0.0735, "step": 14053 }, { "epoch": 2.58, "learning_rate": 1.5847479036790557e-05, "loss": 0.0744, "step": 14054 }, { "epoch": 2.58, "learning_rate": 1.5846876750250448e-05, "loss": 0.0691, "step": 14055 }, { "epoch": 2.58, "learning_rate": 1.584627443148222e-05, "loss": 0.0725, "step": 14056 }, { "epoch": 2.58, "learning_rate": 1.584567208048919e-05, "loss": 0.0714, "step": 14057 }, { "epoch": 2.58, "learning_rate": 1.584506969727468e-05, "loss": 0.0708, "step": 14058 }, { "epoch": 2.58, "learning_rate": 1.5844467281842007e-05, "loss": 0.0746, "step": 14059 }, { "epoch": 2.58, "learning_rate": 1.5843864834194493e-05, "loss": 0.0775, "step": 14060 }, { "epoch": 2.58, "learning_rate": 1.584326235433546e-05, "loss": 0.0818, "step": 14061 }, { "epoch": 2.58, "learning_rate": 1.584265984226823e-05, "loss": 0.0768, "step": 14062 }, { "epoch": 2.58, "learning_rate": 1.584205729799612e-05, "loss": 0.0706, "step": 14063 }, { "epoch": 2.58, "learning_rate": 1.5841454721522453e-05, "loss": 0.074, "step": 14064 }, { "epoch": 2.58, "learning_rate": 1.584085211285055e-05, "loss": 0.0681, "step": 14065 }, { "epoch": 2.58, "learning_rate": 1.5840249471983735e-05, "loss": 0.0699, "step": 14066 }, { "epoch": 2.58, "learning_rate": 1.5839646798925327e-05, "loss": 0.072, "step": 14067 }, { "epoch": 2.58, "learning_rate": 1.583904409367865e-05, "loss": 0.0705, "step": 14068 }, { "epoch": 2.58, "learning_rate": 1.5838441356247024e-05, "loss": 0.0768, "step": 14069 }, { "epoch": 2.58, "learning_rate": 1.5837838586633772e-05, "loss": 0.079, "step": 14070 }, { "epoch": 2.58, "learning_rate": 1.583723578484222e-05, "loss": 0.0783, "step": 14071 }, { "epoch": 2.58, "learning_rate": 1.5836632950875688e-05, "loss": 0.0589, "step": 14072 }, { "epoch": 2.58, "learning_rate": 1.5836030084737495e-05, "loss": 0.0782, "step": 14073 }, { "epoch": 2.58, "learning_rate": 1.5835427186430967e-05, "loss": 0.0681, "step": 14074 }, { "epoch": 2.58, "learning_rate": 1.583482425595943e-05, "loss": 0.0739, "step": 14075 }, { "epoch": 2.58, "learning_rate": 1.5834221293326204e-05, "loss": 0.0762, "step": 14076 }, { "epoch": 2.58, "learning_rate": 1.5833618298534614e-05, "loss": 0.0804, "step": 14077 }, { "epoch": 2.58, "learning_rate": 1.5833015271587982e-05, "loss": 0.0748, "step": 14078 }, { "epoch": 2.58, "learning_rate": 1.5832412212489637e-05, "loss": 0.0787, "step": 14079 }, { "epoch": 2.58, "learning_rate": 1.5831809121242894e-05, "loss": 0.0767, "step": 14080 }, { "epoch": 2.58, "learning_rate": 1.5831205997851086e-05, "loss": 0.0856, "step": 14081 }, { "epoch": 2.58, "learning_rate": 1.5830602842317534e-05, "loss": 0.0608, "step": 14082 }, { "epoch": 2.58, "learning_rate": 1.582999965464556e-05, "loss": 0.0706, "step": 14083 }, { "epoch": 2.58, "learning_rate": 1.5829396434838492e-05, "loss": 0.0782, "step": 14084 }, { "epoch": 2.58, "learning_rate": 1.582879318289966e-05, "loss": 0.0768, "step": 14085 }, { "epoch": 2.58, "learning_rate": 1.5828189898832375e-05, "loss": 0.0672, "step": 14086 }, { "epoch": 2.58, "learning_rate": 1.5827586582639974e-05, "loss": 0.0663, "step": 14087 }, { "epoch": 2.58, "learning_rate": 1.582698323432578e-05, "loss": 0.0724, "step": 14088 }, { "epoch": 2.58, "learning_rate": 1.5826379853893115e-05, "loss": 0.0686, "step": 14089 }, { "epoch": 2.58, "learning_rate": 1.5825776441345312e-05, "loss": 0.0663, "step": 14090 }, { "epoch": 2.58, "learning_rate": 1.582517299668569e-05, "loss": 0.0712, "step": 14091 }, { "epoch": 2.58, "learning_rate": 1.582456951991758e-05, "loss": 0.0814, "step": 14092 }, { "epoch": 2.58, "learning_rate": 1.5823966011044303e-05, "loss": 0.0792, "step": 14093 }, { "epoch": 2.58, "learning_rate": 1.5823362470069192e-05, "loss": 0.0681, "step": 14094 }, { "epoch": 2.58, "learning_rate": 1.582275889699557e-05, "loss": 0.0652, "step": 14095 }, { "epoch": 2.58, "learning_rate": 1.5822155291826766e-05, "loss": 0.0652, "step": 14096 }, { "epoch": 2.59, "learning_rate": 1.5821551654566103e-05, "loss": 0.0794, "step": 14097 }, { "epoch": 2.59, "learning_rate": 1.582094798521691e-05, "loss": 0.0804, "step": 14098 }, { "epoch": 2.59, "learning_rate": 1.582034428378252e-05, "loss": 0.0715, "step": 14099 }, { "epoch": 2.59, "learning_rate": 1.5819740550266253e-05, "loss": 0.0705, "step": 14100 }, { "epoch": 2.59, "learning_rate": 1.581913678467144e-05, "loss": 0.0811, "step": 14101 }, { "epoch": 2.59, "learning_rate": 1.581853298700141e-05, "loss": 0.0731, "step": 14102 }, { "epoch": 2.59, "learning_rate": 1.5817929157259486e-05, "loss": 0.0804, "step": 14103 }, { "epoch": 2.59, "learning_rate": 1.5817325295449002e-05, "loss": 0.0698, "step": 14104 }, { "epoch": 2.59, "learning_rate": 1.5816721401573287e-05, "loss": 0.071, "step": 14105 }, { "epoch": 2.59, "learning_rate": 1.5816117475635664e-05, "loss": 0.071, "step": 14106 }, { "epoch": 2.59, "learning_rate": 1.5815513517639466e-05, "loss": 0.0701, "step": 14107 }, { "epoch": 2.59, "learning_rate": 1.581490952758802e-05, "loss": 0.0719, "step": 14108 }, { "epoch": 2.59, "learning_rate": 1.581430550548466e-05, "loss": 0.0834, "step": 14109 }, { "epoch": 2.59, "learning_rate": 1.5813701451332714e-05, "loss": 0.0733, "step": 14110 }, { "epoch": 2.59, "learning_rate": 1.5813097365135505e-05, "loss": 0.0705, "step": 14111 }, { "epoch": 2.59, "learning_rate": 1.5812493246896368e-05, "loss": 0.0728, "step": 14112 }, { "epoch": 2.59, "learning_rate": 1.5811889096618633e-05, "loss": 0.0754, "step": 14113 }, { "epoch": 2.59, "learning_rate": 1.5811284914305627e-05, "loss": 0.0782, "step": 14114 }, { "epoch": 2.59, "learning_rate": 1.581068069996068e-05, "loss": 0.0692, "step": 14115 }, { "epoch": 2.59, "learning_rate": 1.5810076453587133e-05, "loss": 0.0659, "step": 14116 }, { "epoch": 2.59, "learning_rate": 1.5809472175188303e-05, "loss": 0.081, "step": 14117 }, { "epoch": 2.59, "learning_rate": 1.5808867864767527e-05, "loss": 0.0728, "step": 14118 }, { "epoch": 2.59, "learning_rate": 1.5808263522328137e-05, "loss": 0.0706, "step": 14119 }, { "epoch": 2.59, "learning_rate": 1.580765914787346e-05, "loss": 0.0739, "step": 14120 }, { "epoch": 2.59, "learning_rate": 1.5807054741406828e-05, "loss": 0.0666, "step": 14121 }, { "epoch": 2.59, "learning_rate": 1.580645030293158e-05, "loss": 0.0549, "step": 14122 }, { "epoch": 2.59, "learning_rate": 1.5805845832451038e-05, "loss": 0.0731, "step": 14123 }, { "epoch": 2.59, "learning_rate": 1.5805241329968537e-05, "loss": 0.0729, "step": 14124 }, { "epoch": 2.59, "learning_rate": 1.5804636795487414e-05, "loss": 0.0748, "step": 14125 }, { "epoch": 2.59, "learning_rate": 1.580403222901099e-05, "loss": 0.0646, "step": 14126 }, { "epoch": 2.59, "learning_rate": 1.580342763054261e-05, "loss": 0.0793, "step": 14127 }, { "epoch": 2.59, "learning_rate": 1.5802823000085598e-05, "loss": 0.0784, "step": 14128 }, { "epoch": 2.59, "learning_rate": 1.5802218337643293e-05, "loss": 0.0837, "step": 14129 }, { "epoch": 2.59, "learning_rate": 1.5801613643219022e-05, "loss": 0.0716, "step": 14130 }, { "epoch": 2.59, "learning_rate": 1.580100891681612e-05, "loss": 0.0762, "step": 14131 }, { "epoch": 2.59, "learning_rate": 1.5800404158437926e-05, "loss": 0.0845, "step": 14132 }, { "epoch": 2.59, "learning_rate": 1.5799799368087764e-05, "loss": 0.0697, "step": 14133 }, { "epoch": 2.59, "learning_rate": 1.5799194545768976e-05, "loss": 0.0823, "step": 14134 }, { "epoch": 2.59, "learning_rate": 1.5798589691484885e-05, "loss": 0.0832, "step": 14135 }, { "epoch": 2.59, "learning_rate": 1.5797984805238837e-05, "loss": 0.0716, "step": 14136 }, { "epoch": 2.59, "learning_rate": 1.5797379887034158e-05, "loss": 0.0733, "step": 14137 }, { "epoch": 2.59, "learning_rate": 1.5796774936874188e-05, "loss": 0.0753, "step": 14138 }, { "epoch": 2.59, "learning_rate": 1.5796169954762255e-05, "loss": 0.065, "step": 14139 }, { "epoch": 2.59, "learning_rate": 1.57955649407017e-05, "loss": 0.0782, "step": 14140 }, { "epoch": 2.59, "learning_rate": 1.5794959894695855e-05, "loss": 0.0731, "step": 14141 }, { "epoch": 2.59, "learning_rate": 1.5794354816748057e-05, "loss": 0.0769, "step": 14142 }, { "epoch": 2.59, "learning_rate": 1.5793749706861637e-05, "loss": 0.0752, "step": 14143 }, { "epoch": 2.59, "learning_rate": 1.5793144565039934e-05, "loss": 0.0699, "step": 14144 }, { "epoch": 2.59, "learning_rate": 1.5792539391286282e-05, "loss": 0.0702, "step": 14145 }, { "epoch": 2.59, "learning_rate": 1.5791934185604018e-05, "loss": 0.0768, "step": 14146 }, { "epoch": 2.59, "learning_rate": 1.5791328947996475e-05, "loss": 0.0776, "step": 14147 }, { "epoch": 2.59, "learning_rate": 1.579072367846699e-05, "loss": 0.0774, "step": 14148 }, { "epoch": 2.59, "learning_rate": 1.5790118377018907e-05, "loss": 0.0835, "step": 14149 }, { "epoch": 2.59, "learning_rate": 1.578951304365555e-05, "loss": 0.0735, "step": 14150 }, { "epoch": 2.6, "learning_rate": 1.5788907678380264e-05, "loss": 0.0758, "step": 14151 }, { "epoch": 2.6, "learning_rate": 1.5788302281196383e-05, "loss": 0.0693, "step": 14152 }, { "epoch": 2.6, "learning_rate": 1.578769685210724e-05, "loss": 0.0604, "step": 14153 }, { "epoch": 2.6, "learning_rate": 1.578709139111618e-05, "loss": 0.0709, "step": 14154 }, { "epoch": 2.6, "learning_rate": 1.578648589822654e-05, "loss": 0.0809, "step": 14155 }, { "epoch": 2.6, "learning_rate": 1.578588037344165e-05, "loss": 0.0755, "step": 14156 }, { "epoch": 2.6, "learning_rate": 1.5785274816764853e-05, "loss": 0.0747, "step": 14157 }, { "epoch": 2.6, "learning_rate": 1.578466922819949e-05, "loss": 0.0702, "step": 14158 }, { "epoch": 2.6, "learning_rate": 1.5784063607748886e-05, "loss": 0.0793, "step": 14159 }, { "epoch": 2.6, "learning_rate": 1.578345795541639e-05, "loss": 0.0827, "step": 14160 }, { "epoch": 2.6, "learning_rate": 1.5782852271205343e-05, "loss": 0.0691, "step": 14161 }, { "epoch": 2.6, "learning_rate": 1.5782246555119077e-05, "loss": 0.0872, "step": 14162 }, { "epoch": 2.6, "learning_rate": 1.5781640807160932e-05, "loss": 0.0806, "step": 14163 }, { "epoch": 2.6, "learning_rate": 1.5781035027334244e-05, "loss": 0.0786, "step": 14164 }, { "epoch": 2.6, "learning_rate": 1.5780429215642362e-05, "loss": 0.0659, "step": 14165 }, { "epoch": 2.6, "learning_rate": 1.577982337208861e-05, "loss": 0.0736, "step": 14166 }, { "epoch": 2.6, "learning_rate": 1.5779217496676345e-05, "loss": 0.0694, "step": 14167 }, { "epoch": 2.6, "learning_rate": 1.577861158940889e-05, "loss": 0.0636, "step": 14168 }, { "epoch": 2.6, "learning_rate": 1.57780056502896e-05, "loss": 0.0903, "step": 14169 }, { "epoch": 2.6, "learning_rate": 1.5777399679321802e-05, "loss": 0.0807, "step": 14170 }, { "epoch": 2.6, "learning_rate": 1.5776793676508846e-05, "loss": 0.0736, "step": 14171 }, { "epoch": 2.6, "learning_rate": 1.5776187641854064e-05, "loss": 0.081, "step": 14172 }, { "epoch": 2.6, "learning_rate": 1.5775581575360804e-05, "loss": 0.0832, "step": 14173 }, { "epoch": 2.6, "learning_rate": 1.57749754770324e-05, "loss": 0.0717, "step": 14174 }, { "epoch": 2.6, "learning_rate": 1.5774369346872197e-05, "loss": 0.0749, "step": 14175 }, { "epoch": 2.6, "learning_rate": 1.5773763184883538e-05, "loss": 0.0782, "step": 14176 }, { "epoch": 2.6, "learning_rate": 1.5773156991069754e-05, "loss": 0.0754, "step": 14177 }, { "epoch": 2.6, "learning_rate": 1.57725507654342e-05, "loss": 0.0675, "step": 14178 }, { "epoch": 2.6, "learning_rate": 1.5771944507980208e-05, "loss": 0.0773, "step": 14179 }, { "epoch": 2.6, "learning_rate": 1.5771338218711124e-05, "loss": 0.0649, "step": 14180 }, { "epoch": 2.6, "learning_rate": 1.5770731897630284e-05, "loss": 0.0824, "step": 14181 }, { "epoch": 2.6, "learning_rate": 1.577012554474104e-05, "loss": 0.0808, "step": 14182 }, { "epoch": 2.6, "learning_rate": 1.576951916004673e-05, "loss": 0.0777, "step": 14183 }, { "epoch": 2.6, "learning_rate": 1.576891274355069e-05, "loss": 0.0656, "step": 14184 }, { "epoch": 2.6, "learning_rate": 1.576830629525627e-05, "loss": 0.0654, "step": 14185 }, { "epoch": 2.6, "learning_rate": 1.576769981516681e-05, "loss": 0.0723, "step": 14186 }, { "epoch": 2.6, "learning_rate": 1.5767093303285657e-05, "loss": 0.0728, "step": 14187 }, { "epoch": 2.6, "learning_rate": 1.5766486759616146e-05, "loss": 0.0716, "step": 14188 }, { "epoch": 2.6, "learning_rate": 1.5765880184161625e-05, "loss": 0.0819, "step": 14189 }, { "epoch": 2.6, "learning_rate": 1.576527357692544e-05, "loss": 0.0809, "step": 14190 }, { "epoch": 2.6, "learning_rate": 1.576466693791093e-05, "loss": 0.071, "step": 14191 }, { "epoch": 2.6, "learning_rate": 1.5764060267121443e-05, "loss": 0.0786, "step": 14192 }, { "epoch": 2.6, "learning_rate": 1.5763453564560318e-05, "loss": 0.0717, "step": 14193 }, { "epoch": 2.6, "learning_rate": 1.5762846830230904e-05, "loss": 0.0673, "step": 14194 }, { "epoch": 2.6, "learning_rate": 1.576224006413654e-05, "loss": 0.0655, "step": 14195 }, { "epoch": 2.6, "learning_rate": 1.5761633266280577e-05, "loss": 0.0809, "step": 14196 }, { "epoch": 2.6, "learning_rate": 1.5761026436666357e-05, "loss": 0.0806, "step": 14197 }, { "epoch": 2.6, "learning_rate": 1.5760419575297223e-05, "loss": 0.0662, "step": 14198 }, { "epoch": 2.6, "learning_rate": 1.5759812682176525e-05, "loss": 0.0778, "step": 14199 }, { "epoch": 2.6, "learning_rate": 1.57592057573076e-05, "loss": 0.0858, "step": 14200 }, { "epoch": 2.6, "learning_rate": 1.5758598800693797e-05, "loss": 0.085, "step": 14201 }, { "epoch": 2.6, "learning_rate": 1.5757991812338467e-05, "loss": 0.0767, "step": 14202 }, { "epoch": 2.6, "learning_rate": 1.5757384792244948e-05, "loss": 0.0788, "step": 14203 }, { "epoch": 2.6, "learning_rate": 1.575677774041659e-05, "loss": 0.0676, "step": 14204 }, { "epoch": 2.6, "learning_rate": 1.575617065685674e-05, "loss": 0.0789, "step": 14205 }, { "epoch": 2.61, "learning_rate": 1.575556354156874e-05, "loss": 0.0667, "step": 14206 }, { "epoch": 2.61, "learning_rate": 1.5754956394555936e-05, "loss": 0.0709, "step": 14207 }, { "epoch": 2.61, "learning_rate": 1.575434921582168e-05, "loss": 0.0667, "step": 14208 }, { "epoch": 2.61, "learning_rate": 1.575374200536932e-05, "loss": 0.0691, "step": 14209 }, { "epoch": 2.61, "learning_rate": 1.5753134763202194e-05, "loss": 0.0588, "step": 14210 }, { "epoch": 2.61, "learning_rate": 1.575252748932366e-05, "loss": 0.0851, "step": 14211 }, { "epoch": 2.61, "learning_rate": 1.5751920183737054e-05, "loss": 0.0697, "step": 14212 }, { "epoch": 2.61, "learning_rate": 1.5751312846445736e-05, "loss": 0.0752, "step": 14213 }, { "epoch": 2.61, "learning_rate": 1.575070547745304e-05, "loss": 0.0799, "step": 14214 }, { "epoch": 2.61, "learning_rate": 1.5750098076762325e-05, "loss": 0.0772, "step": 14215 }, { "epoch": 2.61, "learning_rate": 1.574949064437693e-05, "loss": 0.0597, "step": 14216 }, { "epoch": 2.61, "learning_rate": 1.574888318030021e-05, "loss": 0.0801, "step": 14217 }, { "epoch": 2.61, "learning_rate": 1.5748275684535515e-05, "loss": 0.071, "step": 14218 }, { "epoch": 2.61, "learning_rate": 1.5747668157086188e-05, "loss": 0.0774, "step": 14219 }, { "epoch": 2.61, "learning_rate": 1.5747060597955578e-05, "loss": 0.0843, "step": 14220 }, { "epoch": 2.61, "learning_rate": 1.5746453007147034e-05, "loss": 0.0693, "step": 14221 }, { "epoch": 2.61, "learning_rate": 1.574584538466391e-05, "loss": 0.0717, "step": 14222 }, { "epoch": 2.61, "learning_rate": 1.574523773050955e-05, "loss": 0.0788, "step": 14223 }, { "epoch": 2.61, "learning_rate": 1.5744630044687307e-05, "loss": 0.0793, "step": 14224 }, { "epoch": 2.61, "learning_rate": 1.5744022327200524e-05, "loss": 0.0774, "step": 14225 }, { "epoch": 2.61, "learning_rate": 1.574341457805256e-05, "loss": 0.0779, "step": 14226 }, { "epoch": 2.61, "learning_rate": 1.5742806797246757e-05, "loss": 0.078, "step": 14227 }, { "epoch": 2.61, "learning_rate": 1.574219898478647e-05, "loss": 0.075, "step": 14228 }, { "epoch": 2.61, "learning_rate": 1.5741591140675046e-05, "loss": 0.0787, "step": 14229 }, { "epoch": 2.61, "learning_rate": 1.574098326491584e-05, "loss": 0.0657, "step": 14230 }, { "epoch": 2.61, "learning_rate": 1.5740375357512198e-05, "loss": 0.0703, "step": 14231 }, { "epoch": 2.61, "learning_rate": 1.5739767418467474e-05, "loss": 0.0787, "step": 14232 }, { "epoch": 2.61, "learning_rate": 1.5739159447785012e-05, "loss": 0.0718, "step": 14233 }, { "epoch": 2.61, "learning_rate": 1.5738551445468174e-05, "loss": 0.0817, "step": 14234 }, { "epoch": 2.61, "learning_rate": 1.5737943411520307e-05, "loss": 0.0682, "step": 14235 }, { "epoch": 2.61, "learning_rate": 1.5737335345944758e-05, "loss": 0.0815, "step": 14236 }, { "epoch": 2.61, "learning_rate": 1.5736727248744884e-05, "loss": 0.0632, "step": 14237 }, { "epoch": 2.61, "learning_rate": 1.5736119119924035e-05, "loss": 0.0757, "step": 14238 }, { "epoch": 2.61, "learning_rate": 1.573551095948556e-05, "loss": 0.0676, "step": 14239 }, { "epoch": 2.61, "learning_rate": 1.573490276743282e-05, "loss": 0.0851, "step": 14240 }, { "epoch": 2.61, "learning_rate": 1.5734294543769158e-05, "loss": 0.0654, "step": 14241 }, { "epoch": 2.61, "learning_rate": 1.5733686288497928e-05, "loss": 0.0687, "step": 14242 }, { "epoch": 2.61, "learning_rate": 1.5733078001622488e-05, "loss": 0.0629, "step": 14243 }, { "epoch": 2.61, "learning_rate": 1.5732469683146188e-05, "loss": 0.0714, "step": 14244 }, { "epoch": 2.61, "learning_rate": 1.5731861333072377e-05, "loss": 0.0717, "step": 14245 }, { "epoch": 2.61, "learning_rate": 1.5731252951404416e-05, "loss": 0.0681, "step": 14246 }, { "epoch": 2.61, "learning_rate": 1.5730644538145654e-05, "loss": 0.0774, "step": 14247 }, { "epoch": 2.61, "learning_rate": 1.5730036093299444e-05, "loss": 0.0676, "step": 14248 }, { "epoch": 2.61, "learning_rate": 1.5729427616869145e-05, "loss": 0.0805, "step": 14249 }, { "epoch": 2.61, "learning_rate": 1.57288191088581e-05, "loss": 0.0716, "step": 14250 }, { "epoch": 2.61, "learning_rate": 1.5728210569269672e-05, "loss": 0.0699, "step": 14251 }, { "epoch": 2.61, "learning_rate": 1.5727601998107217e-05, "loss": 0.0789, "step": 14252 }, { "epoch": 2.61, "learning_rate": 1.5726993395374083e-05, "loss": 0.0921, "step": 14253 }, { "epoch": 2.61, "learning_rate": 1.5726384761073627e-05, "loss": 0.0645, "step": 14254 }, { "epoch": 2.61, "learning_rate": 1.57257760952092e-05, "loss": 0.0713, "step": 14255 }, { "epoch": 2.61, "learning_rate": 1.5725167397784168e-05, "loss": 0.0726, "step": 14256 }, { "epoch": 2.61, "learning_rate": 1.5724558668801874e-05, "loss": 0.0777, "step": 14257 }, { "epoch": 2.61, "learning_rate": 1.5723949908265682e-05, "loss": 0.0724, "step": 14258 }, { "epoch": 2.61, "learning_rate": 1.5723341116178942e-05, "loss": 0.0667, "step": 14259 }, { "epoch": 2.62, "learning_rate": 1.5722732292545012e-05, "loss": 0.0689, "step": 14260 }, { "epoch": 2.62, "learning_rate": 1.5722123437367247e-05, "loss": 0.0623, "step": 14261 }, { "epoch": 2.62, "learning_rate": 1.5721514550649004e-05, "loss": 0.0668, "step": 14262 }, { "epoch": 2.62, "learning_rate": 1.572090563239364e-05, "loss": 0.0796, "step": 14263 }, { "epoch": 2.62, "learning_rate": 1.572029668260451e-05, "loss": 0.0752, "step": 14264 }, { "epoch": 2.62, "learning_rate": 1.571968770128497e-05, "loss": 0.0746, "step": 14265 }, { "epoch": 2.62, "learning_rate": 1.5719078688438377e-05, "loss": 0.0709, "step": 14266 }, { "epoch": 2.62, "learning_rate": 1.5718469644068085e-05, "loss": 0.0766, "step": 14267 }, { "epoch": 2.62, "learning_rate": 1.571786056817746e-05, "loss": 0.0844, "step": 14268 }, { "epoch": 2.62, "learning_rate": 1.5717251460769852e-05, "loss": 0.0659, "step": 14269 }, { "epoch": 2.62, "learning_rate": 1.571664232184862e-05, "loss": 0.079, "step": 14270 }, { "epoch": 2.62, "learning_rate": 1.5716033151417124e-05, "loss": 0.0801, "step": 14271 }, { "epoch": 2.62, "learning_rate": 1.5715423949478714e-05, "loss": 0.0666, "step": 14272 }, { "epoch": 2.62, "learning_rate": 1.5714814716036756e-05, "loss": 0.077, "step": 14273 }, { "epoch": 2.62, "learning_rate": 1.5714205451094602e-05, "loss": 0.0754, "step": 14274 }, { "epoch": 2.62, "learning_rate": 1.571359615465562e-05, "loss": 0.07, "step": 14275 }, { "epoch": 2.62, "learning_rate": 1.571298682672316e-05, "loss": 0.0711, "step": 14276 }, { "epoch": 2.62, "learning_rate": 1.5712377467300582e-05, "loss": 0.0707, "step": 14277 }, { "epoch": 2.62, "learning_rate": 1.5711768076391245e-05, "loss": 0.0723, "step": 14278 }, { "epoch": 2.62, "learning_rate": 1.5711158653998507e-05, "loss": 0.0713, "step": 14279 }, { "epoch": 2.62, "learning_rate": 1.5710549200125732e-05, "loss": 0.0779, "step": 14280 }, { "epoch": 2.62, "learning_rate": 1.5709939714776274e-05, "loss": 0.0765, "step": 14281 }, { "epoch": 2.62, "learning_rate": 1.5709330197953495e-05, "loss": 0.0793, "step": 14282 }, { "epoch": 2.62, "learning_rate": 1.5708720649660754e-05, "loss": 0.0728, "step": 14283 }, { "epoch": 2.62, "learning_rate": 1.5708111069901414e-05, "loss": 0.077, "step": 14284 }, { "epoch": 2.62, "learning_rate": 1.5707501458678828e-05, "loss": 0.0733, "step": 14285 }, { "epoch": 2.62, "learning_rate": 1.570689181599636e-05, "loss": 0.0723, "step": 14286 }, { "epoch": 2.62, "learning_rate": 1.5706282141857375e-05, "loss": 0.0706, "step": 14287 }, { "epoch": 2.62, "learning_rate": 1.5705672436265225e-05, "loss": 0.0725, "step": 14288 }, { "epoch": 2.62, "learning_rate": 1.5705062699223278e-05, "loss": 0.0869, "step": 14289 }, { "epoch": 2.62, "learning_rate": 1.570445293073489e-05, "loss": 0.0797, "step": 14290 }, { "epoch": 2.62, "learning_rate": 1.5703843130803423e-05, "loss": 0.0749, "step": 14291 }, { "epoch": 2.62, "learning_rate": 1.5703233299432238e-05, "loss": 0.0727, "step": 14292 }, { "epoch": 2.62, "learning_rate": 1.5702623436624702e-05, "loss": 0.0733, "step": 14293 }, { "epoch": 2.62, "learning_rate": 1.5702013542384167e-05, "loss": 0.0767, "step": 14294 }, { "epoch": 2.62, "learning_rate": 1.5701403616714003e-05, "loss": 0.0646, "step": 14295 }, { "epoch": 2.62, "learning_rate": 1.5700793659617568e-05, "loss": 0.076, "step": 14296 }, { "epoch": 2.62, "learning_rate": 1.5700183671098223e-05, "loss": 0.0746, "step": 14297 }, { "epoch": 2.62, "learning_rate": 1.5699573651159333e-05, "loss": 0.069, "step": 14298 }, { "epoch": 2.62, "learning_rate": 1.5698963599804266e-05, "loss": 0.0797, "step": 14299 }, { "epoch": 2.62, "learning_rate": 1.569835351703637e-05, "loss": 0.0864, "step": 14300 }, { "epoch": 2.62, "learning_rate": 1.5697743402859015e-05, "loss": 0.0773, "step": 14301 }, { "epoch": 2.62, "learning_rate": 1.569713325727557e-05, "loss": 0.0768, "step": 14302 }, { "epoch": 2.62, "learning_rate": 1.569652308028939e-05, "loss": 0.0758, "step": 14303 }, { "epoch": 2.62, "learning_rate": 1.5695912871903843e-05, "loss": 0.0812, "step": 14304 }, { "epoch": 2.62, "learning_rate": 1.5695302632122292e-05, "loss": 0.0719, "step": 14305 }, { "epoch": 2.62, "learning_rate": 1.5694692360948097e-05, "loss": 0.0658, "step": 14306 }, { "epoch": 2.62, "learning_rate": 1.5694082058384626e-05, "loss": 0.0786, "step": 14307 }, { "epoch": 2.62, "learning_rate": 1.569347172443524e-05, "loss": 0.0817, "step": 14308 }, { "epoch": 2.62, "learning_rate": 1.5692861359103304e-05, "loss": 0.0714, "step": 14309 }, { "epoch": 2.62, "learning_rate": 1.5692250962392182e-05, "loss": 0.0643, "step": 14310 }, { "epoch": 2.62, "learning_rate": 1.5691640534305242e-05, "loss": 0.07, "step": 14311 }, { "epoch": 2.62, "learning_rate": 1.5691030074845845e-05, "loss": 0.0741, "step": 14312 }, { "epoch": 2.62, "learning_rate": 1.569041958401736e-05, "loss": 0.0637, "step": 14313 }, { "epoch": 2.62, "learning_rate": 1.5689809061823144e-05, "loss": 0.0633, "step": 14314 }, { "epoch": 2.63, "learning_rate": 1.568919850826657e-05, "loss": 0.0728, "step": 14315 }, { "epoch": 2.63, "learning_rate": 1.5688587923350997e-05, "loss": 0.0849, "step": 14316 }, { "epoch": 2.63, "learning_rate": 1.5687977307079798e-05, "loss": 0.0823, "step": 14317 }, { "epoch": 2.63, "learning_rate": 1.5687366659456334e-05, "loss": 0.0675, "step": 14318 }, { "epoch": 2.63, "learning_rate": 1.568675598048397e-05, "loss": 0.0713, "step": 14319 }, { "epoch": 2.63, "learning_rate": 1.5686145270166077e-05, "loss": 0.0722, "step": 14320 }, { "epoch": 2.63, "learning_rate": 1.5685534528506015e-05, "loss": 0.0701, "step": 14321 }, { "epoch": 2.63, "learning_rate": 1.568492375550715e-05, "loss": 0.0672, "step": 14322 }, { "epoch": 2.63, "learning_rate": 1.568431295117286e-05, "loss": 0.0753, "step": 14323 }, { "epoch": 2.63, "learning_rate": 1.56837021155065e-05, "loss": 0.0754, "step": 14324 }, { "epoch": 2.63, "learning_rate": 1.568309124851144e-05, "loss": 0.0621, "step": 14325 }, { "epoch": 2.63, "learning_rate": 1.568248035019105e-05, "loss": 0.0736, "step": 14326 }, { "epoch": 2.63, "learning_rate": 1.5681869420548695e-05, "loss": 0.0584, "step": 14327 }, { "epoch": 2.63, "learning_rate": 1.568125845958774e-05, "loss": 0.062, "step": 14328 }, { "epoch": 2.63, "learning_rate": 1.568064746731156e-05, "loss": 0.0807, "step": 14329 }, { "epoch": 2.63, "learning_rate": 1.5680036443723516e-05, "loss": 0.0587, "step": 14330 }, { "epoch": 2.63, "learning_rate": 1.5679425388826977e-05, "loss": 0.0785, "step": 14331 }, { "epoch": 2.63, "learning_rate": 1.567881430262531e-05, "loss": 0.0615, "step": 14332 }, { "epoch": 2.63, "learning_rate": 1.567820318512189e-05, "loss": 0.0809, "step": 14333 }, { "epoch": 2.63, "learning_rate": 1.567759203632008e-05, "loss": 0.071, "step": 14334 }, { "epoch": 2.63, "learning_rate": 1.567698085622325e-05, "loss": 0.0698, "step": 14335 }, { "epoch": 2.63, "learning_rate": 1.5676369644834764e-05, "loss": 0.0724, "step": 14336 }, { "epoch": 2.63, "learning_rate": 1.5675758402157997e-05, "loss": 0.0688, "step": 14337 }, { "epoch": 2.63, "learning_rate": 1.567514712819632e-05, "loss": 0.0722, "step": 14338 }, { "epoch": 2.63, "learning_rate": 1.5674535822953098e-05, "loss": 0.067, "step": 14339 }, { "epoch": 2.63, "learning_rate": 1.56739244864317e-05, "loss": 0.0695, "step": 14340 }, { "epoch": 2.63, "learning_rate": 1.56733131186355e-05, "loss": 0.0768, "step": 14341 }, { "epoch": 2.63, "learning_rate": 1.5672701719567864e-05, "loss": 0.0693, "step": 14342 }, { "epoch": 2.63, "learning_rate": 1.5672090289232164e-05, "loss": 0.0755, "step": 14343 }, { "epoch": 2.63, "learning_rate": 1.5671478827631766e-05, "loss": 0.0767, "step": 14344 }, { "epoch": 2.63, "learning_rate": 1.567086733477005e-05, "loss": 0.0617, "step": 14345 }, { "epoch": 2.63, "learning_rate": 1.5670255810650377e-05, "loss": 0.0784, "step": 14346 }, { "epoch": 2.63, "learning_rate": 1.566964425527612e-05, "loss": 0.0709, "step": 14347 }, { "epoch": 2.63, "learning_rate": 1.566903266865065e-05, "loss": 0.0775, "step": 14348 }, { "epoch": 2.63, "learning_rate": 1.566842105077734e-05, "loss": 0.0707, "step": 14349 }, { "epoch": 2.63, "learning_rate": 1.5667809401659565e-05, "loss": 0.0727, "step": 14350 }, { "epoch": 2.63, "learning_rate": 1.5667197721300684e-05, "loss": 0.0808, "step": 14351 }, { "epoch": 2.63, "learning_rate": 1.5666586009704082e-05, "loss": 0.072, "step": 14352 }, { "epoch": 2.63, "learning_rate": 1.5665974266873123e-05, "loss": 0.0846, "step": 14353 }, { "epoch": 2.63, "learning_rate": 1.5665362492811182e-05, "loss": 0.0701, "step": 14354 }, { "epoch": 2.63, "learning_rate": 1.566475068752163e-05, "loss": 0.0685, "step": 14355 }, { "epoch": 2.63, "learning_rate": 1.566413885100784e-05, "loss": 0.0726, "step": 14356 }, { "epoch": 2.63, "learning_rate": 1.5663526983273184e-05, "loss": 0.0662, "step": 14357 }, { "epoch": 2.63, "learning_rate": 1.5662915084321032e-05, "loss": 0.0737, "step": 14358 }, { "epoch": 2.63, "learning_rate": 1.5662303154154764e-05, "loss": 0.0724, "step": 14359 }, { "epoch": 2.63, "learning_rate": 1.5661691192777745e-05, "loss": 0.0775, "step": 14360 }, { "epoch": 2.63, "learning_rate": 1.566107920019335e-05, "loss": 0.0827, "step": 14361 }, { "epoch": 2.63, "learning_rate": 1.5660467176404955e-05, "loss": 0.0698, "step": 14362 }, { "epoch": 2.63, "learning_rate": 1.5659855121415938e-05, "loss": 0.0694, "step": 14363 }, { "epoch": 2.63, "learning_rate": 1.5659243035229657e-05, "loss": 0.0774, "step": 14364 }, { "epoch": 2.63, "learning_rate": 1.5658630917849504e-05, "loss": 0.0646, "step": 14365 }, { "epoch": 2.63, "learning_rate": 1.5658018769278844e-05, "loss": 0.0726, "step": 14366 }, { "epoch": 2.63, "learning_rate": 1.565740658952105e-05, "loss": 0.076, "step": 14367 }, { "epoch": 2.63, "learning_rate": 1.5656794378579498e-05, "loss": 0.074, "step": 14368 }, { "epoch": 2.64, "learning_rate": 1.5656182136457562e-05, "loss": 0.073, "step": 14369 }, { "epoch": 2.64, "learning_rate": 1.565556986315862e-05, "loss": 0.0741, "step": 14370 }, { "epoch": 2.64, "learning_rate": 1.5654957558686043e-05, "loss": 0.0802, "step": 14371 }, { "epoch": 2.64, "learning_rate": 1.5654345223043206e-05, "loss": 0.069, "step": 14372 }, { "epoch": 2.64, "learning_rate": 1.5653732856233487e-05, "loss": 0.0779, "step": 14373 }, { "epoch": 2.64, "learning_rate": 1.5653120458260263e-05, "loss": 0.0756, "step": 14374 }, { "epoch": 2.64, "learning_rate": 1.56525080291269e-05, "loss": 0.0731, "step": 14375 }, { "epoch": 2.64, "learning_rate": 1.5651895568836785e-05, "loss": 0.0673, "step": 14376 }, { "epoch": 2.64, "learning_rate": 1.565128307739329e-05, "loss": 0.0737, "step": 14377 }, { "epoch": 2.64, "learning_rate": 1.565067055479979e-05, "loss": 0.092, "step": 14378 }, { "epoch": 2.64, "learning_rate": 1.5650058001059658e-05, "loss": 0.0758, "step": 14379 }, { "epoch": 2.64, "learning_rate": 1.5649445416176275e-05, "loss": 0.083, "step": 14380 }, { "epoch": 2.64, "learning_rate": 1.564883280015302e-05, "loss": 0.0622, "step": 14381 }, { "epoch": 2.64, "learning_rate": 1.5648220152993264e-05, "loss": 0.0752, "step": 14382 }, { "epoch": 2.64, "learning_rate": 1.5647607474700386e-05, "loss": 0.0631, "step": 14383 }, { "epoch": 2.64, "learning_rate": 1.5646994765277762e-05, "loss": 0.0791, "step": 14384 }, { "epoch": 2.64, "learning_rate": 1.5646382024728772e-05, "loss": 0.0727, "step": 14385 }, { "epoch": 2.64, "learning_rate": 1.5645769253056792e-05, "loss": 0.0643, "step": 14386 }, { "epoch": 2.64, "learning_rate": 1.56451564502652e-05, "loss": 0.0671, "step": 14387 }, { "epoch": 2.64, "learning_rate": 1.5644543616357375e-05, "loss": 0.0647, "step": 14388 }, { "epoch": 2.64, "learning_rate": 1.5643930751336693e-05, "loss": 0.0647, "step": 14389 }, { "epoch": 2.64, "learning_rate": 1.5643317855206527e-05, "loss": 0.0724, "step": 14390 }, { "epoch": 2.64, "learning_rate": 1.5642704927970265e-05, "loss": 0.0904, "step": 14391 }, { "epoch": 2.64, "learning_rate": 1.5642091969631282e-05, "loss": 0.077, "step": 14392 }, { "epoch": 2.64, "learning_rate": 1.5641478980192953e-05, "loss": 0.0819, "step": 14393 }, { "epoch": 2.64, "learning_rate": 1.564086595965866e-05, "loss": 0.0754, "step": 14394 }, { "epoch": 2.64, "learning_rate": 1.5640252908031786e-05, "loss": 0.0782, "step": 14395 }, { "epoch": 2.64, "learning_rate": 1.5639639825315702e-05, "loss": 0.0675, "step": 14396 }, { "epoch": 2.64, "learning_rate": 1.5639026711513793e-05, "loss": 0.0749, "step": 14397 }, { "epoch": 2.64, "learning_rate": 1.5638413566629437e-05, "loss": 0.0658, "step": 14398 }, { "epoch": 2.64, "learning_rate": 1.5637800390666013e-05, "loss": 0.0661, "step": 14399 }, { "epoch": 2.64, "learning_rate": 1.5637187183626895e-05, "loss": 0.0776, "step": 14400 }, { "epoch": 2.64, "learning_rate": 1.5636573945515477e-05, "loss": 0.0649, "step": 14401 }, { "epoch": 2.64, "learning_rate": 1.5635960676335125e-05, "loss": 0.0847, "step": 14402 }, { "epoch": 2.64, "learning_rate": 1.5635347376089233e-05, "loss": 0.0767, "step": 14403 }, { "epoch": 2.64, "learning_rate": 1.563473404478117e-05, "loss": 0.0688, "step": 14404 }, { "epoch": 2.64, "learning_rate": 1.563412068241432e-05, "loss": 0.0673, "step": 14405 }, { "epoch": 2.64, "learning_rate": 1.5633507288992068e-05, "loss": 0.078, "step": 14406 }, { "epoch": 2.64, "learning_rate": 1.5632893864517793e-05, "loss": 0.0724, "step": 14407 }, { "epoch": 2.64, "learning_rate": 1.563228040899487e-05, "loss": 0.0773, "step": 14408 }, { "epoch": 2.64, "learning_rate": 1.5631666922426688e-05, "loss": 0.0643, "step": 14409 }, { "epoch": 2.64, "learning_rate": 1.5631053404816627e-05, "loss": 0.0787, "step": 14410 }, { "epoch": 2.64, "learning_rate": 1.5630439856168064e-05, "loss": 0.0751, "step": 14411 }, { "epoch": 2.64, "learning_rate": 1.562982627648439e-05, "loss": 0.0637, "step": 14412 }, { "epoch": 2.64, "learning_rate": 1.562921266576898e-05, "loss": 0.0699, "step": 14413 }, { "epoch": 2.64, "learning_rate": 1.5628599024025214e-05, "loss": 0.0734, "step": 14414 }, { "epoch": 2.64, "learning_rate": 1.562798535125648e-05, "loss": 0.0704, "step": 14415 }, { "epoch": 2.64, "learning_rate": 1.5627371647466164e-05, "loss": 0.0802, "step": 14416 }, { "epoch": 2.64, "learning_rate": 1.5626757912657638e-05, "loss": 0.075, "step": 14417 }, { "epoch": 2.64, "learning_rate": 1.562614414683429e-05, "loss": 0.0675, "step": 14418 }, { "epoch": 2.64, "learning_rate": 1.562553034999951e-05, "loss": 0.0831, "step": 14419 }, { "epoch": 2.64, "learning_rate": 1.5624916522156666e-05, "loss": 0.0741, "step": 14420 }, { "epoch": 2.64, "learning_rate": 1.5624302663309157e-05, "loss": 0.0774, "step": 14421 }, { "epoch": 2.64, "learning_rate": 1.5623688773460358e-05, "loss": 0.0688, "step": 14422 }, { "epoch": 2.64, "learning_rate": 1.5623074852613656e-05, "loss": 0.0786, "step": 14423 }, { "epoch": 2.65, "learning_rate": 1.5622460900772433e-05, "loss": 0.0737, "step": 14424 }, { "epoch": 2.65, "learning_rate": 1.562184691794007e-05, "loss": 0.0764, "step": 14425 }, { "epoch": 2.65, "learning_rate": 1.562123290411996e-05, "loss": 0.0783, "step": 14426 }, { "epoch": 2.65, "learning_rate": 1.562061885931548e-05, "loss": 0.0753, "step": 14427 }, { "epoch": 2.65, "learning_rate": 1.5620004783530017e-05, "loss": 0.073, "step": 14428 }, { "epoch": 2.65, "learning_rate": 1.5619390676766956e-05, "loss": 0.0842, "step": 14429 }, { "epoch": 2.65, "learning_rate": 1.5618776539029685e-05, "loss": 0.0677, "step": 14430 }, { "epoch": 2.65, "learning_rate": 1.5618162370321583e-05, "loss": 0.0805, "step": 14431 }, { "epoch": 2.65, "learning_rate": 1.561754817064604e-05, "loss": 0.0667, "step": 14432 }, { "epoch": 2.65, "learning_rate": 1.5616933940006438e-05, "loss": 0.0728, "step": 14433 }, { "epoch": 2.65, "learning_rate": 1.561631967840617e-05, "loss": 0.0669, "step": 14434 }, { "epoch": 2.65, "learning_rate": 1.561570538584861e-05, "loss": 0.0687, "step": 14435 }, { "epoch": 2.65, "learning_rate": 1.5615091062337154e-05, "loss": 0.071, "step": 14436 }, { "epoch": 2.65, "learning_rate": 1.561447670787518e-05, "loss": 0.0765, "step": 14437 }, { "epoch": 2.65, "learning_rate": 1.5613862322466086e-05, "loss": 0.0829, "step": 14438 }, { "epoch": 2.65, "learning_rate": 1.561324790611324e-05, "loss": 0.081, "step": 14439 }, { "epoch": 2.65, "learning_rate": 1.5612633458820052e-05, "loss": 0.073, "step": 14440 }, { "epoch": 2.65, "learning_rate": 1.5612018980589892e-05, "loss": 0.0836, "step": 14441 }, { "epoch": 2.65, "learning_rate": 1.561140447142615e-05, "loss": 0.0723, "step": 14442 }, { "epoch": 2.65, "learning_rate": 1.5610789931332217e-05, "loss": 0.0717, "step": 14443 }, { "epoch": 2.65, "learning_rate": 1.561017536031148e-05, "loss": 0.0699, "step": 14444 }, { "epoch": 2.65, "learning_rate": 1.5609560758367325e-05, "loss": 0.0814, "step": 14445 }, { "epoch": 2.65, "learning_rate": 1.5608946125503135e-05, "loss": 0.0723, "step": 14446 }, { "epoch": 2.65, "learning_rate": 1.5608331461722307e-05, "loss": 0.0761, "step": 14447 }, { "epoch": 2.65, "learning_rate": 1.5607716767028222e-05, "loss": 0.0779, "step": 14448 }, { "epoch": 2.65, "learning_rate": 1.5607102041424272e-05, "loss": 0.062, "step": 14449 }, { "epoch": 2.65, "learning_rate": 1.5606487284913848e-05, "loss": 0.0668, "step": 14450 }, { "epoch": 2.65, "learning_rate": 1.560587249750033e-05, "loss": 0.0758, "step": 14451 }, { "epoch": 2.65, "learning_rate": 1.5605257679187114e-05, "loss": 0.0662, "step": 14452 }, { "epoch": 2.65, "learning_rate": 1.5604642829977584e-05, "loss": 0.0708, "step": 14453 }, { "epoch": 2.65, "learning_rate": 1.5604027949875133e-05, "loss": 0.0842, "step": 14454 }, { "epoch": 2.65, "learning_rate": 1.5603413038883145e-05, "loss": 0.0756, "step": 14455 }, { "epoch": 2.65, "learning_rate": 1.560279809700502e-05, "loss": 0.0623, "step": 14456 }, { "epoch": 2.65, "learning_rate": 1.5602183124244136e-05, "loss": 0.0746, "step": 14457 }, { "epoch": 2.65, "learning_rate": 1.5601568120603887e-05, "loss": 0.0721, "step": 14458 }, { "epoch": 2.65, "learning_rate": 1.560095308608766e-05, "loss": 0.0656, "step": 14459 }, { "epoch": 2.65, "learning_rate": 1.5600338020698854e-05, "loss": 0.0752, "step": 14460 }, { "epoch": 2.65, "learning_rate": 1.5599722924440855e-05, "loss": 0.0758, "step": 14461 }, { "epoch": 2.65, "learning_rate": 1.559910779731705e-05, "loss": 0.0821, "step": 14462 }, { "epoch": 2.65, "learning_rate": 1.559849263933083e-05, "loss": 0.0786, "step": 14463 }, { "epoch": 2.65, "learning_rate": 1.559787745048559e-05, "loss": 0.0658, "step": 14464 }, { "epoch": 2.65, "learning_rate": 1.5597262230784718e-05, "loss": 0.0656, "step": 14465 }, { "epoch": 2.65, "learning_rate": 1.5596646980231604e-05, "loss": 0.0649, "step": 14466 }, { "epoch": 2.65, "learning_rate": 1.5596031698829638e-05, "loss": 0.0707, "step": 14467 }, { "epoch": 2.65, "learning_rate": 1.559541638658222e-05, "loss": 0.0711, "step": 14468 }, { "epoch": 2.65, "learning_rate": 1.559480104349273e-05, "loss": 0.0695, "step": 14469 }, { "epoch": 2.65, "learning_rate": 1.559418566956457e-05, "loss": 0.0741, "step": 14470 }, { "epoch": 2.65, "learning_rate": 1.5593570264801128e-05, "loss": 0.0893, "step": 14471 }, { "epoch": 2.65, "learning_rate": 1.5592954829205793e-05, "loss": 0.0676, "step": 14472 }, { "epoch": 2.65, "learning_rate": 1.5592339362781958e-05, "loss": 0.0748, "step": 14473 }, { "epoch": 2.65, "learning_rate": 1.5591723865533022e-05, "loss": 0.0735, "step": 14474 }, { "epoch": 2.65, "learning_rate": 1.559110833746237e-05, "loss": 0.0701, "step": 14475 }, { "epoch": 2.65, "learning_rate": 1.55904927785734e-05, "loss": 0.0852, "step": 14476 }, { "epoch": 2.65, "learning_rate": 1.55898771888695e-05, "loss": 0.0688, "step": 14477 }, { "epoch": 2.66, "learning_rate": 1.558926156835407e-05, "loss": 0.0789, "step": 14478 }, { "epoch": 2.66, "learning_rate": 1.5588645917030495e-05, "loss": 0.063, "step": 14479 }, { "epoch": 2.66, "learning_rate": 1.5588030234902175e-05, "loss": 0.0752, "step": 14480 }, { "epoch": 2.66, "learning_rate": 1.5587414521972504e-05, "loss": 0.0704, "step": 14481 }, { "epoch": 2.66, "learning_rate": 1.558679877824487e-05, "loss": 0.0674, "step": 14482 }, { "epoch": 2.66, "learning_rate": 1.5586183003722673e-05, "loss": 0.0726, "step": 14483 }, { "epoch": 2.66, "learning_rate": 1.55855671984093e-05, "loss": 0.0639, "step": 14484 }, { "epoch": 2.66, "learning_rate": 1.558495136230815e-05, "loss": 0.0681, "step": 14485 }, { "epoch": 2.66, "learning_rate": 1.558433549542262e-05, "loss": 0.068, "step": 14486 }, { "epoch": 2.66, "learning_rate": 1.55837195977561e-05, "loss": 0.0849, "step": 14487 }, { "epoch": 2.66, "learning_rate": 1.558310366931199e-05, "loss": 0.0695, "step": 14488 }, { "epoch": 2.66, "learning_rate": 1.558248771009368e-05, "loss": 0.0794, "step": 14489 }, { "epoch": 2.66, "learning_rate": 1.5581871720104567e-05, "loss": 0.0551, "step": 14490 }, { "epoch": 2.66, "learning_rate": 1.5581255699348046e-05, "loss": 0.0659, "step": 14491 }, { "epoch": 2.66, "learning_rate": 1.558063964782751e-05, "loss": 0.0668, "step": 14492 }, { "epoch": 2.66, "learning_rate": 1.558002356554636e-05, "loss": 0.0744, "step": 14493 }, { "epoch": 2.66, "learning_rate": 1.5579407452507993e-05, "loss": 0.0721, "step": 14494 }, { "epoch": 2.66, "learning_rate": 1.5578791308715797e-05, "loss": 0.0694, "step": 14495 }, { "epoch": 2.66, "learning_rate": 1.5578175134173174e-05, "loss": 0.0788, "step": 14496 }, { "epoch": 2.66, "learning_rate": 1.557755892888352e-05, "loss": 0.0846, "step": 14497 }, { "epoch": 2.66, "learning_rate": 1.557694269285023e-05, "loss": 0.0586, "step": 14498 }, { "epoch": 2.66, "learning_rate": 1.5576326426076698e-05, "loss": 0.068, "step": 14499 }, { "epoch": 2.66, "learning_rate": 1.5575710128566325e-05, "loss": 0.0849, "step": 14500 }, { "epoch": 2.66, "learning_rate": 1.5575093800322513e-05, "loss": 0.0669, "step": 14501 }, { "epoch": 2.66, "learning_rate": 1.5574477441348647e-05, "loss": 0.0813, "step": 14502 }, { "epoch": 2.66, "learning_rate": 1.5573861051648132e-05, "loss": 0.0707, "step": 14503 }, { "epoch": 2.66, "learning_rate": 1.5573244631224364e-05, "loss": 0.0759, "step": 14504 }, { "epoch": 2.66, "learning_rate": 1.5572628180080743e-05, "loss": 0.0708, "step": 14505 }, { "epoch": 2.66, "learning_rate": 1.5572011698220662e-05, "loss": 0.0782, "step": 14506 }, { "epoch": 2.66, "learning_rate": 1.557139518564753e-05, "loss": 0.0865, "step": 14507 }, { "epoch": 2.66, "learning_rate": 1.557077864236473e-05, "loss": 0.0745, "step": 14508 }, { "epoch": 2.66, "learning_rate": 1.5570162068375666e-05, "loss": 0.0678, "step": 14509 }, { "epoch": 2.66, "learning_rate": 1.5569545463683742e-05, "loss": 0.0698, "step": 14510 }, { "epoch": 2.66, "learning_rate": 1.5568928828292353e-05, "loss": 0.0741, "step": 14511 }, { "epoch": 2.66, "learning_rate": 1.5568312162204896e-05, "loss": 0.0642, "step": 14512 }, { "epoch": 2.66, "learning_rate": 1.5567695465424773e-05, "loss": 0.0731, "step": 14513 }, { "epoch": 2.66, "learning_rate": 1.556707873795538e-05, "loss": 0.0662, "step": 14514 }, { "epoch": 2.66, "learning_rate": 1.556646197980012e-05, "loss": 0.0824, "step": 14515 }, { "epoch": 2.66, "learning_rate": 1.5565845190962395e-05, "loss": 0.0661, "step": 14516 }, { "epoch": 2.66, "learning_rate": 1.5565228371445595e-05, "loss": 0.0774, "step": 14517 }, { "epoch": 2.66, "learning_rate": 1.556461152125313e-05, "loss": 0.0663, "step": 14518 }, { "epoch": 2.66, "learning_rate": 1.5563994640388393e-05, "loss": 0.071, "step": 14519 }, { "epoch": 2.66, "learning_rate": 1.556337772885479e-05, "loss": 0.072, "step": 14520 }, { "epoch": 2.66, "learning_rate": 1.5562760786655714e-05, "loss": 0.0694, "step": 14521 }, { "epoch": 2.66, "learning_rate": 1.5562143813794577e-05, "loss": 0.0777, "step": 14522 }, { "epoch": 2.66, "learning_rate": 1.5561526810274768e-05, "loss": 0.085, "step": 14523 }, { "epoch": 2.66, "learning_rate": 1.5560909776099694e-05, "loss": 0.0847, "step": 14524 }, { "epoch": 2.66, "learning_rate": 1.5560292711272755e-05, "loss": 0.0713, "step": 14525 }, { "epoch": 2.66, "learning_rate": 1.555967561579735e-05, "loss": 0.0695, "step": 14526 }, { "epoch": 2.66, "learning_rate": 1.555905848967689e-05, "loss": 0.0782, "step": 14527 }, { "epoch": 2.66, "learning_rate": 1.5558441332914762e-05, "loss": 0.0701, "step": 14528 }, { "epoch": 2.66, "learning_rate": 1.555782414551438e-05, "loss": 0.0686, "step": 14529 }, { "epoch": 2.66, "learning_rate": 1.555720692747914e-05, "loss": 0.067, "step": 14530 }, { "epoch": 2.66, "learning_rate": 1.5556589678812444e-05, "loss": 0.0814, "step": 14531 }, { "epoch": 2.66, "learning_rate": 1.5555972399517694e-05, "loss": 0.0687, "step": 14532 }, { "epoch": 2.67, "learning_rate": 1.5555355089598295e-05, "loss": 0.0779, "step": 14533 }, { "epoch": 2.67, "learning_rate": 1.555473774905765e-05, "loss": 0.0692, "step": 14534 }, { "epoch": 2.67, "learning_rate": 1.5554120377899157e-05, "loss": 0.0828, "step": 14535 }, { "epoch": 2.67, "learning_rate": 1.5553502976126228e-05, "loss": 0.0635, "step": 14536 }, { "epoch": 2.67, "learning_rate": 1.5552885543742257e-05, "loss": 0.0701, "step": 14537 }, { "epoch": 2.67, "learning_rate": 1.5552268080750652e-05, "loss": 0.0782, "step": 14538 }, { "epoch": 2.67, "learning_rate": 1.5551650587154815e-05, "loss": 0.076, "step": 14539 }, { "epoch": 2.67, "learning_rate": 1.555103306295815e-05, "loss": 0.071, "step": 14540 }, { "epoch": 2.67, "learning_rate": 1.5550415508164058e-05, "loss": 0.0736, "step": 14541 }, { "epoch": 2.67, "learning_rate": 1.5549797922775948e-05, "loss": 0.0789, "step": 14542 }, { "epoch": 2.67, "learning_rate": 1.5549180306797223e-05, "loss": 0.0779, "step": 14543 }, { "epoch": 2.67, "learning_rate": 1.5548562660231284e-05, "loss": 0.0594, "step": 14544 }, { "epoch": 2.67, "learning_rate": 1.5547944983081537e-05, "loss": 0.074, "step": 14545 }, { "epoch": 2.67, "learning_rate": 1.554732727535139e-05, "loss": 0.0738, "step": 14546 }, { "epoch": 2.67, "learning_rate": 1.5546709537044245e-05, "loss": 0.0766, "step": 14547 }, { "epoch": 2.67, "learning_rate": 1.5546091768163506e-05, "loss": 0.0639, "step": 14548 }, { "epoch": 2.67, "learning_rate": 1.554547396871258e-05, "loss": 0.0605, "step": 14549 }, { "epoch": 2.67, "learning_rate": 1.554485613869487e-05, "loss": 0.0691, "step": 14550 }, { "epoch": 2.67, "learning_rate": 1.5544238278113785e-05, "loss": 0.0714, "step": 14551 }, { "epoch": 2.67, "learning_rate": 1.5543620386972724e-05, "loss": 0.0665, "step": 14552 }, { "epoch": 2.67, "learning_rate": 1.5543002465275102e-05, "loss": 0.0605, "step": 14553 }, { "epoch": 2.67, "learning_rate": 1.5542384513024317e-05, "loss": 0.0774, "step": 14554 }, { "epoch": 2.67, "learning_rate": 1.5541766530223785e-05, "loss": 0.0663, "step": 14555 }, { "epoch": 2.67, "learning_rate": 1.55411485168769e-05, "loss": 0.0689, "step": 14556 }, { "epoch": 2.67, "learning_rate": 1.5540530472987077e-05, "loss": 0.0667, "step": 14557 }, { "epoch": 2.67, "learning_rate": 1.5539912398557722e-05, "loss": 0.0671, "step": 14558 }, { "epoch": 2.67, "learning_rate": 1.5539294293592237e-05, "loss": 0.078, "step": 14559 }, { "epoch": 2.67, "learning_rate": 1.5538676158094036e-05, "loss": 0.0743, "step": 14560 }, { "epoch": 2.67, "learning_rate": 1.5538057992066518e-05, "loss": 0.0692, "step": 14561 }, { "epoch": 2.67, "learning_rate": 1.5537439795513095e-05, "loss": 0.0811, "step": 14562 }, { "epoch": 2.67, "learning_rate": 1.5536821568437174e-05, "loss": 0.069, "step": 14563 }, { "epoch": 2.67, "learning_rate": 1.5536203310842166e-05, "loss": 0.0774, "step": 14564 }, { "epoch": 2.67, "learning_rate": 1.5535585022731474e-05, "loss": 0.066, "step": 14565 }, { "epoch": 2.67, "learning_rate": 1.553496670410851e-05, "loss": 0.0703, "step": 14566 }, { "epoch": 2.67, "learning_rate": 1.5534348354976674e-05, "loss": 0.0697, "step": 14567 }, { "epoch": 2.67, "learning_rate": 1.5533729975339383e-05, "loss": 0.0765, "step": 14568 }, { "epoch": 2.67, "learning_rate": 1.5533111565200046e-05, "loss": 0.0715, "step": 14569 }, { "epoch": 2.67, "learning_rate": 1.5532493124562065e-05, "loss": 0.0592, "step": 14570 }, { "epoch": 2.67, "learning_rate": 1.5531874653428854e-05, "loss": 0.0846, "step": 14571 }, { "epoch": 2.67, "learning_rate": 1.5531256151803816e-05, "loss": 0.072, "step": 14572 }, { "epoch": 2.67, "learning_rate": 1.5530637619690373e-05, "loss": 0.0648, "step": 14573 }, { "epoch": 2.67, "learning_rate": 1.553001905709192e-05, "loss": 0.0772, "step": 14574 }, { "epoch": 2.67, "learning_rate": 1.552940046401187e-05, "loss": 0.0727, "step": 14575 }, { "epoch": 2.67, "learning_rate": 1.5528781840453637e-05, "loss": 0.0774, "step": 14576 }, { "epoch": 2.67, "learning_rate": 1.5528163186420632e-05, "loss": 0.0648, "step": 14577 }, { "epoch": 2.67, "learning_rate": 1.552754450191626e-05, "loss": 0.0699, "step": 14578 }, { "epoch": 2.67, "learning_rate": 1.5526925786943932e-05, "loss": 0.0662, "step": 14579 }, { "epoch": 2.67, "learning_rate": 1.5526307041507064e-05, "loss": 0.0789, "step": 14580 }, { "epoch": 2.67, "learning_rate": 1.5525688265609054e-05, "loss": 0.0774, "step": 14581 }, { "epoch": 2.67, "learning_rate": 1.552506945925333e-05, "loss": 0.0732, "step": 14582 }, { "epoch": 2.67, "learning_rate": 1.5524450622443286e-05, "loss": 0.0706, "step": 14583 }, { "epoch": 2.67, "learning_rate": 1.5523831755182348e-05, "loss": 0.0804, "step": 14584 }, { "epoch": 2.67, "learning_rate": 1.5523212857473916e-05, "loss": 0.065, "step": 14585 }, { "epoch": 2.67, "learning_rate": 1.5522593929321406e-05, "loss": 0.0761, "step": 14586 }, { "epoch": 2.68, "learning_rate": 1.5521974970728225e-05, "loss": 0.0727, "step": 14587 }, { "epoch": 2.68, "learning_rate": 1.5521355981697793e-05, "loss": 0.0794, "step": 14588 }, { "epoch": 2.68, "learning_rate": 1.5520736962233518e-05, "loss": 0.0677, "step": 14589 }, { "epoch": 2.68, "learning_rate": 1.552011791233881e-05, "loss": 0.0715, "step": 14590 }, { "epoch": 2.68, "learning_rate": 1.5519498832017083e-05, "loss": 0.0664, "step": 14591 }, { "epoch": 2.68, "learning_rate": 1.551887972127175e-05, "loss": 0.0658, "step": 14592 }, { "epoch": 2.68, "learning_rate": 1.5518260580106222e-05, "loss": 0.0739, "step": 14593 }, { "epoch": 2.68, "learning_rate": 1.551764140852391e-05, "loss": 0.0677, "step": 14594 }, { "epoch": 2.68, "learning_rate": 1.5517022206528233e-05, "loss": 0.0743, "step": 14595 }, { "epoch": 2.68, "learning_rate": 1.55164029741226e-05, "loss": 0.07, "step": 14596 }, { "epoch": 2.68, "learning_rate": 1.551578371131042e-05, "loss": 0.0715, "step": 14597 }, { "epoch": 2.68, "learning_rate": 1.5515164418095115e-05, "loss": 0.0583, "step": 14598 }, { "epoch": 2.68, "learning_rate": 1.5514545094480093e-05, "loss": 0.0641, "step": 14599 }, { "epoch": 2.68, "learning_rate": 1.551392574046877e-05, "loss": 0.0723, "step": 14600 }, { "epoch": 2.68, "learning_rate": 1.5513306356064563e-05, "loss": 0.0662, "step": 14601 }, { "epoch": 2.68, "learning_rate": 1.5512686941270878e-05, "loss": 0.0657, "step": 14602 }, { "epoch": 2.68, "learning_rate": 1.551206749609113e-05, "loss": 0.0901, "step": 14603 }, { "epoch": 2.68, "learning_rate": 1.5511448020528745e-05, "loss": 0.0665, "step": 14604 }, { "epoch": 2.68, "learning_rate": 1.5510828514587122e-05, "loss": 0.0785, "step": 14605 }, { "epoch": 2.68, "learning_rate": 1.551020897826969e-05, "loss": 0.0642, "step": 14606 }, { "epoch": 2.68, "learning_rate": 1.550958941157985e-05, "loss": 0.067, "step": 14607 }, { "epoch": 2.68, "learning_rate": 1.5508969814521026e-05, "loss": 0.0657, "step": 14608 }, { "epoch": 2.68, "learning_rate": 1.5508350187096634e-05, "loss": 0.0665, "step": 14609 }, { "epoch": 2.68, "learning_rate": 1.5507730529310085e-05, "loss": 0.0817, "step": 14610 }, { "epoch": 2.68, "learning_rate": 1.5507110841164798e-05, "loss": 0.071, "step": 14611 }, { "epoch": 2.68, "learning_rate": 1.5506491122664184e-05, "loss": 0.0631, "step": 14612 }, { "epoch": 2.68, "learning_rate": 1.5505871373811663e-05, "loss": 0.0659, "step": 14613 }, { "epoch": 2.68, "learning_rate": 1.5505251594610652e-05, "loss": 0.073, "step": 14614 }, { "epoch": 2.68, "learning_rate": 1.5504631785064564e-05, "loss": 0.0801, "step": 14615 }, { "epoch": 2.68, "learning_rate": 1.5504011945176814e-05, "loss": 0.0651, "step": 14616 }, { "epoch": 2.68, "learning_rate": 1.5503392074950824e-05, "loss": 0.0764, "step": 14617 }, { "epoch": 2.68, "learning_rate": 1.5502772174390006e-05, "loss": 0.0749, "step": 14618 }, { "epoch": 2.68, "learning_rate": 1.5502152243497782e-05, "loss": 0.0762, "step": 14619 }, { "epoch": 2.68, "learning_rate": 1.5501532282277566e-05, "loss": 0.0747, "step": 14620 }, { "epoch": 2.68, "learning_rate": 1.5500912290732772e-05, "loss": 0.0656, "step": 14621 }, { "epoch": 2.68, "learning_rate": 1.5500292268866824e-05, "loss": 0.0681, "step": 14622 }, { "epoch": 2.68, "learning_rate": 1.5499672216683134e-05, "loss": 0.082, "step": 14623 }, { "epoch": 2.68, "learning_rate": 1.549905213418512e-05, "loss": 0.0798, "step": 14624 }, { "epoch": 2.68, "learning_rate": 1.5498432021376206e-05, "loss": 0.0689, "step": 14625 }, { "epoch": 2.68, "learning_rate": 1.5497811878259802e-05, "loss": 0.0761, "step": 14626 }, { "epoch": 2.68, "learning_rate": 1.549719170483933e-05, "loss": 0.0765, "step": 14627 }, { "epoch": 2.68, "learning_rate": 1.5496571501118214e-05, "loss": 0.0684, "step": 14628 }, { "epoch": 2.68, "learning_rate": 1.549595126709986e-05, "loss": 0.0778, "step": 14629 }, { "epoch": 2.68, "learning_rate": 1.5495331002787703e-05, "loss": 0.0718, "step": 14630 }, { "epoch": 2.68, "learning_rate": 1.5494710708185145e-05, "loss": 0.061, "step": 14631 }, { "epoch": 2.68, "learning_rate": 1.5494090383295616e-05, "loss": 0.0752, "step": 14632 }, { "epoch": 2.68, "learning_rate": 1.5493470028122532e-05, "loss": 0.0731, "step": 14633 }, { "epoch": 2.68, "learning_rate": 1.549284964266931e-05, "loss": 0.073, "step": 14634 }, { "epoch": 2.68, "learning_rate": 1.549222922693937e-05, "loss": 0.0736, "step": 14635 }, { "epoch": 2.68, "learning_rate": 1.5491608780936142e-05, "loss": 0.0657, "step": 14636 }, { "epoch": 2.68, "learning_rate": 1.549098830466303e-05, "loss": 0.0897, "step": 14637 }, { "epoch": 2.68, "learning_rate": 1.549036779812347e-05, "loss": 0.0806, "step": 14638 }, { "epoch": 2.68, "learning_rate": 1.5489747261320866e-05, "loss": 0.0784, "step": 14639 }, { "epoch": 2.68, "learning_rate": 1.548912669425865e-05, "loss": 0.0786, "step": 14640 }, { "epoch": 2.68, "learning_rate": 1.548850609694024e-05, "loss": 0.0646, "step": 14641 }, { "epoch": 2.69, "learning_rate": 1.5487885469369054e-05, "loss": 0.0776, "step": 14642 }, { "epoch": 2.69, "learning_rate": 1.5487264811548516e-05, "loss": 0.0735, "step": 14643 }, { "epoch": 2.69, "learning_rate": 1.5486644123482047e-05, "loss": 0.0651, "step": 14644 }, { "epoch": 2.69, "learning_rate": 1.5486023405173064e-05, "loss": 0.0785, "step": 14645 }, { "epoch": 2.69, "learning_rate": 1.5485402656624994e-05, "loss": 0.0719, "step": 14646 }, { "epoch": 2.69, "learning_rate": 1.548478187784125e-05, "loss": 0.0792, "step": 14647 }, { "epoch": 2.69, "learning_rate": 1.5484161068825266e-05, "loss": 0.0663, "step": 14648 }, { "epoch": 2.69, "learning_rate": 1.5483540229580455e-05, "loss": 0.0718, "step": 14649 }, { "epoch": 2.69, "learning_rate": 1.5482919360110246e-05, "loss": 0.0721, "step": 14650 }, { "epoch": 2.69, "learning_rate": 1.5482298460418054e-05, "loss": 0.0713, "step": 14651 }, { "epoch": 2.69, "learning_rate": 1.54816775305073e-05, "loss": 0.0668, "step": 14652 }, { "epoch": 2.69, "learning_rate": 1.548105657038142e-05, "loss": 0.0721, "step": 14653 }, { "epoch": 2.69, "learning_rate": 1.5480435580043823e-05, "loss": 0.075, "step": 14654 }, { "epoch": 2.69, "learning_rate": 1.5479814559497936e-05, "loss": 0.0716, "step": 14655 }, { "epoch": 2.69, "learning_rate": 1.5479193508747185e-05, "loss": 0.0817, "step": 14656 }, { "epoch": 2.69, "learning_rate": 1.5478572427794987e-05, "loss": 0.0667, "step": 14657 }, { "epoch": 2.69, "learning_rate": 1.5477951316644774e-05, "loss": 0.0677, "step": 14658 }, { "epoch": 2.69, "learning_rate": 1.5477330175299964e-05, "loss": 0.0799, "step": 14659 }, { "epoch": 2.69, "learning_rate": 1.547670900376398e-05, "loss": 0.0739, "step": 14660 }, { "epoch": 2.69, "learning_rate": 1.5476087802040248e-05, "loss": 0.0733, "step": 14661 }, { "epoch": 2.69, "learning_rate": 1.5475466570132193e-05, "loss": 0.0774, "step": 14662 }, { "epoch": 2.69, "learning_rate": 1.5474845308043236e-05, "loss": 0.0748, "step": 14663 }, { "epoch": 2.69, "learning_rate": 1.5474224015776806e-05, "loss": 0.0722, "step": 14664 }, { "epoch": 2.69, "learning_rate": 1.547360269333632e-05, "loss": 0.0735, "step": 14665 }, { "epoch": 2.69, "learning_rate": 1.5472981340725217e-05, "loss": 0.0664, "step": 14666 }, { "epoch": 2.69, "learning_rate": 1.5472359957946906e-05, "loss": 0.0795, "step": 14667 }, { "epoch": 2.69, "learning_rate": 1.547173854500482e-05, "loss": 0.0699, "step": 14668 }, { "epoch": 2.69, "learning_rate": 1.5471117101902378e-05, "loss": 0.067, "step": 14669 }, { "epoch": 2.69, "learning_rate": 1.547049562864302e-05, "loss": 0.073, "step": 14670 }, { "epoch": 2.69, "learning_rate": 1.5469874125230157e-05, "loss": 0.0761, "step": 14671 }, { "epoch": 2.69, "learning_rate": 1.5469252591667218e-05, "loss": 0.0599, "step": 14672 }, { "epoch": 2.69, "learning_rate": 1.546863102795763e-05, "loss": 0.0847, "step": 14673 }, { "epoch": 2.69, "learning_rate": 1.5468009434104826e-05, "loss": 0.0638, "step": 14674 }, { "epoch": 2.69, "learning_rate": 1.5467387810112223e-05, "loss": 0.0789, "step": 14675 }, { "epoch": 2.69, "learning_rate": 1.5466766155983248e-05, "loss": 0.071, "step": 14676 }, { "epoch": 2.69, "learning_rate": 1.546614447172133e-05, "loss": 0.0771, "step": 14677 }, { "epoch": 2.69, "learning_rate": 1.5465522757329895e-05, "loss": 0.0716, "step": 14678 }, { "epoch": 2.69, "learning_rate": 1.5464901012812374e-05, "loss": 0.0625, "step": 14679 }, { "epoch": 2.69, "learning_rate": 1.546427923817219e-05, "loss": 0.0678, "step": 14680 }, { "epoch": 2.69, "learning_rate": 1.5463657433412768e-05, "loss": 0.0658, "step": 14681 }, { "epoch": 2.69, "learning_rate": 1.546303559853754e-05, "loss": 0.0644, "step": 14682 }, { "epoch": 2.69, "learning_rate": 1.546241373354993e-05, "loss": 0.0751, "step": 14683 }, { "epoch": 2.69, "learning_rate": 1.546179183845337e-05, "loss": 0.0585, "step": 14684 }, { "epoch": 2.69, "learning_rate": 1.5461169913251285e-05, "loss": 0.0729, "step": 14685 }, { "epoch": 2.69, "learning_rate": 1.5460547957947105e-05, "loss": 0.0816, "step": 14686 }, { "epoch": 2.69, "learning_rate": 1.5459925972544256e-05, "loss": 0.0777, "step": 14687 }, { "epoch": 2.69, "learning_rate": 1.5459303957046167e-05, "loss": 0.0754, "step": 14688 }, { "epoch": 2.69, "learning_rate": 1.545868191145626e-05, "loss": 0.0714, "step": 14689 }, { "epoch": 2.69, "learning_rate": 1.5458059835777978e-05, "loss": 0.0757, "step": 14690 }, { "epoch": 2.69, "learning_rate": 1.545743773001474e-05, "loss": 0.0623, "step": 14691 }, { "epoch": 2.69, "learning_rate": 1.545681559416998e-05, "loss": 0.0712, "step": 14692 }, { "epoch": 2.69, "learning_rate": 1.5456193428247123e-05, "loss": 0.0712, "step": 14693 }, { "epoch": 2.69, "learning_rate": 1.54555712322496e-05, "loss": 0.0778, "step": 14694 }, { "epoch": 2.69, "learning_rate": 1.5454949006180834e-05, "loss": 0.0734, "step": 14695 }, { "epoch": 2.7, "learning_rate": 1.545432675004427e-05, "loss": 0.0692, "step": 14696 }, { "epoch": 2.7, "learning_rate": 1.5453704463843328e-05, "loss": 0.0752, "step": 14697 }, { "epoch": 2.7, "learning_rate": 1.5453082147581434e-05, "loss": 0.0772, "step": 14698 }, { "epoch": 2.7, "learning_rate": 1.545245980126203e-05, "loss": 0.0667, "step": 14699 }, { "epoch": 2.7, "learning_rate": 1.5451837424888533e-05, "loss": 0.0744, "step": 14700 }, { "epoch": 2.7, "learning_rate": 1.5451215018464386e-05, "loss": 0.0678, "step": 14701 }, { "epoch": 2.7, "learning_rate": 1.5450592581993013e-05, "loss": 0.069, "step": 14702 }, { "epoch": 2.7, "learning_rate": 1.5449970115477846e-05, "loss": 0.075, "step": 14703 }, { "epoch": 2.7, "learning_rate": 1.5449347618922316e-05, "loss": 0.0653, "step": 14704 }, { "epoch": 2.7, "learning_rate": 1.5448725092329856e-05, "loss": 0.0706, "step": 14705 }, { "epoch": 2.7, "learning_rate": 1.5448102535703888e-05, "loss": 0.0729, "step": 14706 }, { "epoch": 2.7, "learning_rate": 1.5447479949047857e-05, "loss": 0.0709, "step": 14707 }, { "epoch": 2.7, "learning_rate": 1.544685733236519e-05, "loss": 0.0725, "step": 14708 }, { "epoch": 2.7, "learning_rate": 1.5446234685659314e-05, "loss": 0.0837, "step": 14709 }, { "epoch": 2.7, "learning_rate": 1.544561200893367e-05, "loss": 0.0773, "step": 14710 }, { "epoch": 2.7, "learning_rate": 1.544498930219168e-05, "loss": 0.0741, "step": 14711 }, { "epoch": 2.7, "learning_rate": 1.5444366565436784e-05, "loss": 0.073, "step": 14712 }, { "epoch": 2.7, "learning_rate": 1.544374379867241e-05, "loss": 0.0702, "step": 14713 }, { "epoch": 2.7, "learning_rate": 1.5443121001901994e-05, "loss": 0.0692, "step": 14714 }, { "epoch": 2.7, "learning_rate": 1.544249817512897e-05, "loss": 0.0643, "step": 14715 }, { "epoch": 2.7, "learning_rate": 1.5441875318356764e-05, "loss": 0.0596, "step": 14716 }, { "epoch": 2.7, "learning_rate": 1.544125243158882e-05, "loss": 0.0846, "step": 14717 }, { "epoch": 2.7, "learning_rate": 1.544062951482856e-05, "loss": 0.0651, "step": 14718 }, { "epoch": 2.7, "learning_rate": 1.5440006568079425e-05, "loss": 0.0688, "step": 14719 }, { "epoch": 2.7, "learning_rate": 1.5439383591344844e-05, "loss": 0.0668, "step": 14720 }, { "epoch": 2.7, "learning_rate": 1.543876058462825e-05, "loss": 0.0634, "step": 14721 }, { "epoch": 2.7, "learning_rate": 1.5438137547933086e-05, "loss": 0.062, "step": 14722 }, { "epoch": 2.7, "learning_rate": 1.543751448126278e-05, "loss": 0.0682, "step": 14723 }, { "epoch": 2.7, "learning_rate": 1.5436891384620765e-05, "loss": 0.0627, "step": 14724 }, { "epoch": 2.7, "learning_rate": 1.5436268258010477e-05, "loss": 0.0707, "step": 14725 }, { "epoch": 2.7, "learning_rate": 1.543564510143535e-05, "loss": 0.0762, "step": 14726 }, { "epoch": 2.7, "learning_rate": 1.5435021914898824e-05, "loss": 0.068, "step": 14727 }, { "epoch": 2.7, "learning_rate": 1.543439869840433e-05, "loss": 0.066, "step": 14728 }, { "epoch": 2.7, "learning_rate": 1.54337754519553e-05, "loss": 0.0734, "step": 14729 }, { "epoch": 2.7, "learning_rate": 1.5433152175555173e-05, "loss": 0.0686, "step": 14730 }, { "epoch": 2.7, "learning_rate": 1.543252886920738e-05, "loss": 0.065, "step": 14731 }, { "epoch": 2.7, "learning_rate": 1.5431905532915367e-05, "loss": 0.0699, "step": 14732 }, { "epoch": 2.7, "learning_rate": 1.5431282166682562e-05, "loss": 0.0615, "step": 14733 }, { "epoch": 2.7, "learning_rate": 1.54306587705124e-05, "loss": 0.059, "step": 14734 }, { "epoch": 2.7, "learning_rate": 1.543003534440832e-05, "loss": 0.0743, "step": 14735 }, { "epoch": 2.7, "learning_rate": 1.5429411888373764e-05, "loss": 0.0756, "step": 14736 }, { "epoch": 2.7, "learning_rate": 1.5428788402412154e-05, "loss": 0.0697, "step": 14737 }, { "epoch": 2.7, "learning_rate": 1.5428164886526943e-05, "loss": 0.0759, "step": 14738 }, { "epoch": 2.7, "learning_rate": 1.5427541340721554e-05, "loss": 0.0708, "step": 14739 }, { "epoch": 2.7, "learning_rate": 1.5426917764999432e-05, "loss": 0.068, "step": 14740 }, { "epoch": 2.7, "learning_rate": 1.5426294159364013e-05, "loss": 0.0649, "step": 14741 }, { "epoch": 2.7, "learning_rate": 1.5425670523818732e-05, "loss": 0.0597, "step": 14742 }, { "epoch": 2.7, "learning_rate": 1.542504685836703e-05, "loss": 0.0734, "step": 14743 }, { "epoch": 2.7, "learning_rate": 1.5424423163012335e-05, "loss": 0.077, "step": 14744 }, { "epoch": 2.7, "learning_rate": 1.5423799437758103e-05, "loss": 0.0857, "step": 14745 }, { "epoch": 2.7, "learning_rate": 1.5423175682607755e-05, "loss": 0.0654, "step": 14746 }, { "epoch": 2.7, "learning_rate": 1.5422551897564737e-05, "loss": 0.0676, "step": 14747 }, { "epoch": 2.7, "learning_rate": 1.5421928082632488e-05, "loss": 0.0646, "step": 14748 }, { "epoch": 2.7, "learning_rate": 1.542130423781444e-05, "loss": 0.0719, "step": 14749 }, { "epoch": 2.7, "learning_rate": 1.542068036311404e-05, "loss": 0.0794, "step": 14750 }, { "epoch": 2.71, "learning_rate": 1.542005645853472e-05, "loss": 0.0722, "step": 14751 }, { "epoch": 2.71, "learning_rate": 1.5419432524079926e-05, "loss": 0.066, "step": 14752 }, { "epoch": 2.71, "learning_rate": 1.5418808559753088e-05, "loss": 0.0772, "step": 14753 }, { "epoch": 2.71, "learning_rate": 1.5418184565557655e-05, "loss": 0.0689, "step": 14754 }, { "epoch": 2.71, "learning_rate": 1.5417560541497057e-05, "loss": 0.0766, "step": 14755 }, { "epoch": 2.71, "learning_rate": 1.5416936487574743e-05, "loss": 0.0708, "step": 14756 }, { "epoch": 2.71, "learning_rate": 1.5416312403794145e-05, "loss": 0.0659, "step": 14757 }, { "epoch": 2.71, "learning_rate": 1.5415688290158708e-05, "loss": 0.0705, "step": 14758 }, { "epoch": 2.71, "learning_rate": 1.541506414667187e-05, "loss": 0.0748, "step": 14759 }, { "epoch": 2.71, "learning_rate": 1.541443997333707e-05, "loss": 0.0661, "step": 14760 }, { "epoch": 2.71, "learning_rate": 1.541381577015775e-05, "loss": 0.0803, "step": 14761 }, { "epoch": 2.71, "learning_rate": 1.5413191537137353e-05, "loss": 0.0652, "step": 14762 }, { "epoch": 2.71, "learning_rate": 1.5412567274279316e-05, "loss": 0.0595, "step": 14763 }, { "epoch": 2.71, "learning_rate": 1.541194298158708e-05, "loss": 0.0625, "step": 14764 }, { "epoch": 2.71, "learning_rate": 1.541131865906409e-05, "loss": 0.0808, "step": 14765 }, { "epoch": 2.71, "learning_rate": 1.5410694306713783e-05, "loss": 0.07, "step": 14766 }, { "epoch": 2.71, "learning_rate": 1.5410069924539605e-05, "loss": 0.0719, "step": 14767 }, { "epoch": 2.71, "learning_rate": 1.540944551254499e-05, "loss": 0.0712, "step": 14768 }, { "epoch": 2.71, "learning_rate": 1.540882107073339e-05, "loss": 0.0663, "step": 14769 }, { "epoch": 2.71, "learning_rate": 1.5408196599108238e-05, "loss": 0.0704, "step": 14770 }, { "epoch": 2.71, "learning_rate": 1.540757209767298e-05, "loss": 0.0659, "step": 14771 }, { "epoch": 2.71, "learning_rate": 1.540694756643106e-05, "loss": 0.0683, "step": 14772 }, { "epoch": 2.71, "learning_rate": 1.5406323005385915e-05, "loss": 0.0604, "step": 14773 }, { "epoch": 2.71, "learning_rate": 1.5405698414540992e-05, "loss": 0.0759, "step": 14774 }, { "epoch": 2.71, "learning_rate": 1.540507379389973e-05, "loss": 0.0863, "step": 14775 }, { "epoch": 2.71, "learning_rate": 1.540444914346558e-05, "loss": 0.0792, "step": 14776 }, { "epoch": 2.71, "learning_rate": 1.540382446324198e-05, "loss": 0.0716, "step": 14777 }, { "epoch": 2.71, "learning_rate": 1.5403199753232372e-05, "loss": 0.082, "step": 14778 }, { "epoch": 2.71, "learning_rate": 1.5402575013440196e-05, "loss": 0.0669, "step": 14779 }, { "epoch": 2.71, "learning_rate": 1.5401950243868903e-05, "loss": 0.0631, "step": 14780 }, { "epoch": 2.71, "learning_rate": 1.5401325444521935e-05, "loss": 0.0673, "step": 14781 }, { "epoch": 2.71, "learning_rate": 1.5400700615402735e-05, "loss": 0.0736, "step": 14782 }, { "epoch": 2.71, "learning_rate": 1.5400075756514744e-05, "loss": 0.0746, "step": 14783 }, { "epoch": 2.71, "learning_rate": 1.539945086786141e-05, "loss": 0.0743, "step": 14784 }, { "epoch": 2.71, "learning_rate": 1.5398825949446177e-05, "loss": 0.0724, "step": 14785 }, { "epoch": 2.71, "learning_rate": 1.5398201001272485e-05, "loss": 0.0727, "step": 14786 }, { "epoch": 2.71, "learning_rate": 1.539757602334379e-05, "loss": 0.0773, "step": 14787 }, { "epoch": 2.71, "learning_rate": 1.5396951015663524e-05, "loss": 0.0745, "step": 14788 }, { "epoch": 2.71, "learning_rate": 1.5396325978235137e-05, "loss": 0.0786, "step": 14789 }, { "epoch": 2.71, "learning_rate": 1.5395700911062078e-05, "loss": 0.074, "step": 14790 }, { "epoch": 2.71, "learning_rate": 1.5395075814147787e-05, "loss": 0.0686, "step": 14791 }, { "epoch": 2.71, "learning_rate": 1.539445068749571e-05, "loss": 0.0638, "step": 14792 }, { "epoch": 2.71, "learning_rate": 1.53938255311093e-05, "loss": 0.0644, "step": 14793 }, { "epoch": 2.71, "learning_rate": 1.5393200344991993e-05, "loss": 0.0738, "step": 14794 }, { "epoch": 2.71, "learning_rate": 1.539257512914724e-05, "loss": 0.0654, "step": 14795 }, { "epoch": 2.71, "learning_rate": 1.5391949883578487e-05, "loss": 0.0696, "step": 14796 }, { "epoch": 2.71, "learning_rate": 1.539132460828918e-05, "loss": 0.0734, "step": 14797 }, { "epoch": 2.71, "learning_rate": 1.5390699303282768e-05, "loss": 0.0715, "step": 14798 }, { "epoch": 2.71, "learning_rate": 1.539007396856269e-05, "loss": 0.0656, "step": 14799 }, { "epoch": 2.71, "learning_rate": 1.5389448604132402e-05, "loss": 0.075, "step": 14800 }, { "epoch": 2.71, "learning_rate": 1.5388823209995345e-05, "loss": 0.0713, "step": 14801 }, { "epoch": 2.71, "learning_rate": 1.538819778615497e-05, "loss": 0.0732, "step": 14802 }, { "epoch": 2.71, "learning_rate": 1.5387572332614722e-05, "loss": 0.0707, "step": 14803 }, { "epoch": 2.71, "learning_rate": 1.538694684937805e-05, "loss": 0.0718, "step": 14804 }, { "epoch": 2.72, "learning_rate": 1.5386321336448395e-05, "loss": 0.0671, "step": 14805 }, { "epoch": 2.72, "learning_rate": 1.5385695793829216e-05, "loss": 0.0719, "step": 14806 }, { "epoch": 2.72, "learning_rate": 1.5385070221523958e-05, "loss": 0.08, "step": 14807 }, { "epoch": 2.72, "learning_rate": 1.5384444619536064e-05, "loss": 0.0674, "step": 14808 }, { "epoch": 2.72, "learning_rate": 1.5383818987868982e-05, "loss": 0.0775, "step": 14809 }, { "epoch": 2.72, "learning_rate": 1.5383193326526166e-05, "loss": 0.0665, "step": 14810 }, { "epoch": 2.72, "learning_rate": 1.5382567635511064e-05, "loss": 0.0724, "step": 14811 }, { "epoch": 2.72, "learning_rate": 1.5381941914827118e-05, "loss": 0.0723, "step": 14812 }, { "epoch": 2.72, "learning_rate": 1.5381316164477788e-05, "loss": 0.0648, "step": 14813 }, { "epoch": 2.72, "learning_rate": 1.5380690384466516e-05, "loss": 0.0581, "step": 14814 }, { "epoch": 2.72, "learning_rate": 1.5380064574796748e-05, "loss": 0.0613, "step": 14815 }, { "epoch": 2.72, "learning_rate": 1.537943873547194e-05, "loss": 0.0677, "step": 14816 }, { "epoch": 2.72, "learning_rate": 1.5378812866495544e-05, "loss": 0.0731, "step": 14817 }, { "epoch": 2.72, "learning_rate": 1.5378186967871003e-05, "loss": 0.0656, "step": 14818 }, { "epoch": 2.72, "learning_rate": 1.5377561039601765e-05, "loss": 0.0766, "step": 14819 }, { "epoch": 2.72, "learning_rate": 1.537693508169129e-05, "loss": 0.0669, "step": 14820 }, { "epoch": 2.72, "learning_rate": 1.537630909414302e-05, "loss": 0.0667, "step": 14821 }, { "epoch": 2.72, "learning_rate": 1.537568307696041e-05, "loss": 0.0712, "step": 14822 }, { "epoch": 2.72, "learning_rate": 1.537505703014691e-05, "loss": 0.0625, "step": 14823 }, { "epoch": 2.72, "learning_rate": 1.537443095370597e-05, "loss": 0.0763, "step": 14824 }, { "epoch": 2.72, "learning_rate": 1.537380484764104e-05, "loss": 0.0711, "step": 14825 }, { "epoch": 2.72, "learning_rate": 1.537317871195557e-05, "loss": 0.0663, "step": 14826 }, { "epoch": 2.72, "learning_rate": 1.5372552546653012e-05, "loss": 0.0726, "step": 14827 }, { "epoch": 2.72, "learning_rate": 1.537192635173682e-05, "loss": 0.0722, "step": 14828 }, { "epoch": 2.72, "learning_rate": 1.5371300127210447e-05, "loss": 0.0738, "step": 14829 }, { "epoch": 2.72, "learning_rate": 1.5370673873077335e-05, "loss": 0.0615, "step": 14830 }, { "epoch": 2.72, "learning_rate": 1.537004758934095e-05, "loss": 0.0716, "step": 14831 }, { "epoch": 2.72, "learning_rate": 1.5369421276004733e-05, "loss": 0.0685, "step": 14832 }, { "epoch": 2.72, "learning_rate": 1.5368794933072143e-05, "loss": 0.0636, "step": 14833 }, { "epoch": 2.72, "learning_rate": 1.5368168560546623e-05, "loss": 0.074, "step": 14834 }, { "epoch": 2.72, "learning_rate": 1.536754215843164e-05, "loss": 0.0671, "step": 14835 }, { "epoch": 2.72, "learning_rate": 1.5366915726730634e-05, "loss": 0.0683, "step": 14836 }, { "epoch": 2.72, "learning_rate": 1.5366289265447063e-05, "loss": 0.074, "step": 14837 }, { "epoch": 2.72, "learning_rate": 1.5365662774584383e-05, "loss": 0.0644, "step": 14838 }, { "epoch": 2.72, "learning_rate": 1.5365036254146044e-05, "loss": 0.0757, "step": 14839 }, { "epoch": 2.72, "learning_rate": 1.5364409704135497e-05, "loss": 0.0668, "step": 14840 }, { "epoch": 2.72, "learning_rate": 1.53637831245562e-05, "loss": 0.0745, "step": 14841 }, { "epoch": 2.72, "learning_rate": 1.5363156515411605e-05, "loss": 0.0729, "step": 14842 }, { "epoch": 2.72, "learning_rate": 1.536252987670516e-05, "loss": 0.0688, "step": 14843 }, { "epoch": 2.72, "learning_rate": 1.5361903208440332e-05, "loss": 0.0642, "step": 14844 }, { "epoch": 2.72, "learning_rate": 1.536127651062056e-05, "loss": 0.059, "step": 14845 }, { "epoch": 2.72, "learning_rate": 1.5360649783249314e-05, "loss": 0.0659, "step": 14846 }, { "epoch": 2.72, "learning_rate": 1.536002302633004e-05, "loss": 0.0761, "step": 14847 }, { "epoch": 2.72, "learning_rate": 1.5359396239866188e-05, "loss": 0.0721, "step": 14848 }, { "epoch": 2.72, "learning_rate": 1.5358769423861224e-05, "loss": 0.0617, "step": 14849 }, { "epoch": 2.72, "learning_rate": 1.5358142578318595e-05, "loss": 0.0612, "step": 14850 }, { "epoch": 2.72, "learning_rate": 1.535751570324176e-05, "loss": 0.0803, "step": 14851 }, { "epoch": 2.72, "learning_rate": 1.5356888798634168e-05, "loss": 0.0614, "step": 14852 }, { "epoch": 2.72, "learning_rate": 1.5356261864499286e-05, "loss": 0.0698, "step": 14853 }, { "epoch": 2.72, "learning_rate": 1.5355634900840558e-05, "loss": 0.0646, "step": 14854 }, { "epoch": 2.72, "learning_rate": 1.535500790766145e-05, "loss": 0.0647, "step": 14855 }, { "epoch": 2.72, "learning_rate": 1.535438088496541e-05, "loss": 0.06, "step": 14856 }, { "epoch": 2.72, "learning_rate": 1.53537538327559e-05, "loss": 0.0678, "step": 14857 }, { "epoch": 2.72, "learning_rate": 1.5353126751036367e-05, "loss": 0.0716, "step": 14858 }, { "epoch": 2.72, "learning_rate": 1.535249963981028e-05, "loss": 0.07, "step": 14859 }, { "epoch": 2.73, "learning_rate": 1.5351872499081087e-05, "loss": 0.0712, "step": 14860 }, { "epoch": 2.73, "learning_rate": 1.5351245328852247e-05, "loss": 0.0702, "step": 14861 }, { "epoch": 2.73, "learning_rate": 1.5350618129127216e-05, "loss": 0.073, "step": 14862 }, { "epoch": 2.73, "learning_rate": 1.5349990899909456e-05, "loss": 0.0594, "step": 14863 }, { "epoch": 2.73, "learning_rate": 1.5349363641202418e-05, "loss": 0.0737, "step": 14864 }, { "epoch": 2.73, "learning_rate": 1.5348736353009566e-05, "loss": 0.0704, "step": 14865 }, { "epoch": 2.73, "learning_rate": 1.534810903533435e-05, "loss": 0.0719, "step": 14866 }, { "epoch": 2.73, "learning_rate": 1.5347481688180232e-05, "loss": 0.0665, "step": 14867 }, { "epoch": 2.73, "learning_rate": 1.5346854311550673e-05, "loss": 0.0673, "step": 14868 }, { "epoch": 2.73, "learning_rate": 1.5346226905449127e-05, "loss": 0.0736, "step": 14869 }, { "epoch": 2.73, "learning_rate": 1.534559946987905e-05, "loss": 0.0728, "step": 14870 }, { "epoch": 2.73, "learning_rate": 1.5344972004843907e-05, "loss": 0.0767, "step": 14871 }, { "epoch": 2.73, "learning_rate": 1.5344344510347148e-05, "loss": 0.0723, "step": 14872 }, { "epoch": 2.73, "learning_rate": 1.534371698639224e-05, "loss": 0.0697, "step": 14873 }, { "epoch": 2.73, "learning_rate": 1.534308943298264e-05, "loss": 0.0708, "step": 14874 }, { "epoch": 2.73, "learning_rate": 1.5342461850121806e-05, "loss": 0.0695, "step": 14875 }, { "epoch": 2.73, "learning_rate": 1.5341834237813196e-05, "loss": 0.0706, "step": 14876 }, { "epoch": 2.73, "learning_rate": 1.534120659606027e-05, "loss": 0.0711, "step": 14877 }, { "epoch": 2.73, "learning_rate": 1.534057892486649e-05, "loss": 0.0724, "step": 14878 }, { "epoch": 2.73, "learning_rate": 1.533995122423531e-05, "loss": 0.0642, "step": 14879 }, { "epoch": 2.73, "learning_rate": 1.53393234941702e-05, "loss": 0.0768, "step": 14880 }, { "epoch": 2.73, "learning_rate": 1.5338695734674605e-05, "loss": 0.075, "step": 14881 }, { "epoch": 2.73, "learning_rate": 1.5338067945752002e-05, "loss": 0.06, "step": 14882 }, { "epoch": 2.73, "learning_rate": 1.5337440127405837e-05, "loss": 0.0809, "step": 14883 }, { "epoch": 2.73, "learning_rate": 1.5336812279639584e-05, "loss": 0.0883, "step": 14884 }, { "epoch": 2.73, "learning_rate": 1.5336184402456694e-05, "loss": 0.0723, "step": 14885 }, { "epoch": 2.73, "learning_rate": 1.5335556495860628e-05, "loss": 0.0668, "step": 14886 }, { "epoch": 2.73, "learning_rate": 1.533492855985485e-05, "loss": 0.0725, "step": 14887 }, { "epoch": 2.73, "learning_rate": 1.5334300594442826e-05, "loss": 0.0679, "step": 14888 }, { "epoch": 2.73, "learning_rate": 1.5333672599628005e-05, "loss": 0.0747, "step": 14889 }, { "epoch": 2.73, "learning_rate": 1.533304457541386e-05, "loss": 0.069, "step": 14890 }, { "epoch": 2.73, "learning_rate": 1.533241652180385e-05, "loss": 0.0624, "step": 14891 }, { "epoch": 2.73, "learning_rate": 1.533178843880143e-05, "loss": 0.0697, "step": 14892 }, { "epoch": 2.73, "learning_rate": 1.533116032641007e-05, "loss": 0.0684, "step": 14893 }, { "epoch": 2.73, "learning_rate": 1.533053218463323e-05, "loss": 0.072, "step": 14894 }, { "epoch": 2.73, "learning_rate": 1.532990401347437e-05, "loss": 0.0722, "step": 14895 }, { "epoch": 2.73, "learning_rate": 1.5329275812936955e-05, "loss": 0.069, "step": 14896 }, { "epoch": 2.73, "learning_rate": 1.5328647583024446e-05, "loss": 0.062, "step": 14897 }, { "epoch": 2.73, "learning_rate": 1.5328019323740307e-05, "loss": 0.0717, "step": 14898 }, { "epoch": 2.73, "learning_rate": 1.5327391035088e-05, "loss": 0.073, "step": 14899 }, { "epoch": 2.73, "learning_rate": 1.5326762717070995e-05, "loss": 0.0791, "step": 14900 }, { "epoch": 2.73, "learning_rate": 1.5326134369692743e-05, "loss": 0.0802, "step": 14901 }, { "epoch": 2.73, "learning_rate": 1.5325505992956715e-05, "loss": 0.0709, "step": 14902 }, { "epoch": 2.73, "learning_rate": 1.532487758686637e-05, "loss": 0.068, "step": 14903 }, { "epoch": 2.73, "learning_rate": 1.532424915142518e-05, "loss": 0.0664, "step": 14904 }, { "epoch": 2.73, "learning_rate": 1.53236206866366e-05, "loss": 0.0761, "step": 14905 }, { "epoch": 2.73, "learning_rate": 1.5322992192504104e-05, "loss": 0.0559, "step": 14906 }, { "epoch": 2.73, "learning_rate": 1.5322363669031145e-05, "loss": 0.0521, "step": 14907 }, { "epoch": 2.73, "learning_rate": 1.5321735116221193e-05, "loss": 0.0807, "step": 14908 }, { "epoch": 2.73, "learning_rate": 1.5321106534077714e-05, "loss": 0.0767, "step": 14909 }, { "epoch": 2.73, "learning_rate": 1.532047792260417e-05, "loss": 0.0657, "step": 14910 }, { "epoch": 2.73, "learning_rate": 1.5319849281804025e-05, "loss": 0.0697, "step": 14911 }, { "epoch": 2.73, "learning_rate": 1.531922061168075e-05, "loss": 0.0699, "step": 14912 }, { "epoch": 2.73, "learning_rate": 1.531859191223781e-05, "loss": 0.0593, "step": 14913 }, { "epoch": 2.74, "learning_rate": 1.5317963183478656e-05, "loss": 0.0689, "step": 14914 }, { "epoch": 2.74, "learning_rate": 1.5317334425406773e-05, "loss": 0.0766, "step": 14915 }, { "epoch": 2.74, "learning_rate": 1.531670563802561e-05, "loss": 0.0625, "step": 14916 }, { "epoch": 2.74, "learning_rate": 1.531607682133865e-05, "loss": 0.0626, "step": 14917 }, { "epoch": 2.74, "learning_rate": 1.5315447975349346e-05, "loss": 0.073, "step": 14918 }, { "epoch": 2.74, "learning_rate": 1.5314819100061167e-05, "loss": 0.0641, "step": 14919 }, { "epoch": 2.74, "learning_rate": 1.5314190195477585e-05, "loss": 0.0703, "step": 14920 }, { "epoch": 2.74, "learning_rate": 1.531356126160206e-05, "loss": 0.0705, "step": 14921 }, { "epoch": 2.74, "learning_rate": 1.531293229843806e-05, "loss": 0.06, "step": 14922 }, { "epoch": 2.74, "learning_rate": 1.531230330598905e-05, "loss": 0.0659, "step": 14923 }, { "epoch": 2.74, "learning_rate": 1.5311674284258504e-05, "loss": 0.072, "step": 14924 }, { "epoch": 2.74, "learning_rate": 1.5311045233249885e-05, "loss": 0.0635, "step": 14925 }, { "epoch": 2.74, "learning_rate": 1.5310416152966658e-05, "loss": 0.0582, "step": 14926 }, { "epoch": 2.74, "learning_rate": 1.5309787043412293e-05, "loss": 0.0661, "step": 14927 }, { "epoch": 2.74, "learning_rate": 1.5309157904590257e-05, "loss": 0.0747, "step": 14928 }, { "epoch": 2.74, "learning_rate": 1.530852873650402e-05, "loss": 0.0744, "step": 14929 }, { "epoch": 2.74, "learning_rate": 1.5307899539157047e-05, "loss": 0.0673, "step": 14930 }, { "epoch": 2.74, "learning_rate": 1.530727031255281e-05, "loss": 0.0721, "step": 14931 }, { "epoch": 2.74, "learning_rate": 1.530664105669477e-05, "loss": 0.072, "step": 14932 }, { "epoch": 2.74, "learning_rate": 1.5306011771586404e-05, "loss": 0.0673, "step": 14933 }, { "epoch": 2.74, "learning_rate": 1.5305382457231175e-05, "loss": 0.0713, "step": 14934 }, { "epoch": 2.74, "learning_rate": 1.5304753113632554e-05, "loss": 0.0704, "step": 14935 }, { "epoch": 2.74, "learning_rate": 1.530412374079401e-05, "loss": 0.0754, "step": 14936 }, { "epoch": 2.74, "learning_rate": 1.530349433871901e-05, "loss": 0.0722, "step": 14937 }, { "epoch": 2.74, "learning_rate": 1.5302864907411028e-05, "loss": 0.0601, "step": 14938 }, { "epoch": 2.74, "learning_rate": 1.5302235446873525e-05, "loss": 0.0782, "step": 14939 }, { "epoch": 2.74, "learning_rate": 1.530160595710998e-05, "loss": 0.0732, "step": 14940 }, { "epoch": 2.74, "learning_rate": 1.5300976438123857e-05, "loss": 0.0746, "step": 14941 }, { "epoch": 2.74, "learning_rate": 1.5300346889918634e-05, "loss": 0.073, "step": 14942 }, { "epoch": 2.74, "learning_rate": 1.5299717312497764e-05, "loss": 0.0698, "step": 14943 }, { "epoch": 2.74, "learning_rate": 1.5299087705864737e-05, "loss": 0.0699, "step": 14944 }, { "epoch": 2.74, "learning_rate": 1.529845807002301e-05, "loss": 0.0675, "step": 14945 }, { "epoch": 2.74, "learning_rate": 1.5297828404976062e-05, "loss": 0.077, "step": 14946 }, { "epoch": 2.74, "learning_rate": 1.5297198710727356e-05, "loss": 0.0676, "step": 14947 }, { "epoch": 2.74, "learning_rate": 1.5296568987280368e-05, "loss": 0.0812, "step": 14948 }, { "epoch": 2.74, "learning_rate": 1.5295939234638566e-05, "loss": 0.062, "step": 14949 }, { "epoch": 2.74, "learning_rate": 1.5295309452805423e-05, "loss": 0.0711, "step": 14950 }, { "epoch": 2.74, "learning_rate": 1.529467964178441e-05, "loss": 0.0693, "step": 14951 }, { "epoch": 2.74, "learning_rate": 1.5294049801579e-05, "loss": 0.0581, "step": 14952 }, { "epoch": 2.74, "learning_rate": 1.5293419932192664e-05, "loss": 0.0698, "step": 14953 }, { "epoch": 2.74, "learning_rate": 1.5292790033628878e-05, "loss": 0.0716, "step": 14954 }, { "epoch": 2.74, "learning_rate": 1.5292160105891102e-05, "loss": 0.0647, "step": 14955 }, { "epoch": 2.74, "learning_rate": 1.529153014898282e-05, "loss": 0.0657, "step": 14956 }, { "epoch": 2.74, "learning_rate": 1.5290900162907495e-05, "loss": 0.0713, "step": 14957 }, { "epoch": 2.74, "learning_rate": 1.5290270147668606e-05, "loss": 0.0666, "step": 14958 }, { "epoch": 2.74, "learning_rate": 1.5289640103269626e-05, "loss": 0.0663, "step": 14959 }, { "epoch": 2.74, "learning_rate": 1.5289010029714024e-05, "loss": 0.0585, "step": 14960 }, { "epoch": 2.74, "learning_rate": 1.528837992700528e-05, "loss": 0.0646, "step": 14961 }, { "epoch": 2.74, "learning_rate": 1.5287749795146853e-05, "loss": 0.0626, "step": 14962 }, { "epoch": 2.74, "learning_rate": 1.528711963414223e-05, "loss": 0.0686, "step": 14963 }, { "epoch": 2.74, "learning_rate": 1.528648944399488e-05, "loss": 0.0682, "step": 14964 }, { "epoch": 2.74, "learning_rate": 1.5285859224708276e-05, "loss": 0.0648, "step": 14965 }, { "epoch": 2.74, "learning_rate": 1.5285228976285893e-05, "loss": 0.0683, "step": 14966 }, { "epoch": 2.74, "learning_rate": 1.52845986987312e-05, "loss": 0.0777, "step": 14967 }, { "epoch": 2.74, "learning_rate": 1.5283968392047682e-05, "loss": 0.057, "step": 14968 }, { "epoch": 2.75, "learning_rate": 1.52833380562388e-05, "loss": 0.073, "step": 14969 }, { "epoch": 2.75, "learning_rate": 1.5282707691308038e-05, "loss": 0.0758, "step": 14970 }, { "epoch": 2.75, "learning_rate": 1.5282077297258865e-05, "loss": 0.0741, "step": 14971 }, { "epoch": 2.75, "learning_rate": 1.5281446874094764e-05, "loss": 0.0642, "step": 14972 }, { "epoch": 2.75, "learning_rate": 1.5280816421819197e-05, "loss": 0.0652, "step": 14973 }, { "epoch": 2.75, "learning_rate": 1.5280185940435652e-05, "loss": 0.0668, "step": 14974 }, { "epoch": 2.75, "learning_rate": 1.5279555429947594e-05, "loss": 0.0671, "step": 14975 }, { "epoch": 2.75, "learning_rate": 1.5278924890358506e-05, "loss": 0.0757, "step": 14976 }, { "epoch": 2.75, "learning_rate": 1.5278294321671857e-05, "loss": 0.0812, "step": 14977 }, { "epoch": 2.75, "learning_rate": 1.5277663723891128e-05, "loss": 0.0685, "step": 14978 }, { "epoch": 2.75, "learning_rate": 1.5277033097019794e-05, "loss": 0.0633, "step": 14979 }, { "epoch": 2.75, "learning_rate": 1.527640244106133e-05, "loss": 0.0643, "step": 14980 }, { "epoch": 2.75, "learning_rate": 1.527577175601921e-05, "loss": 0.0713, "step": 14981 }, { "epoch": 2.75, "learning_rate": 1.5275141041896914e-05, "loss": 0.0722, "step": 14982 }, { "epoch": 2.75, "learning_rate": 1.5274510298697917e-05, "loss": 0.054, "step": 14983 }, { "epoch": 2.75, "learning_rate": 1.5273879526425693e-05, "loss": 0.0681, "step": 14984 }, { "epoch": 2.75, "learning_rate": 1.5273248725083726e-05, "loss": 0.0715, "step": 14985 }, { "epoch": 2.75, "learning_rate": 1.5272617894675484e-05, "loss": 0.0745, "step": 14986 }, { "epoch": 2.75, "learning_rate": 1.527198703520445e-05, "loss": 0.0622, "step": 14987 }, { "epoch": 2.75, "learning_rate": 1.5271356146674103e-05, "loss": 0.0651, "step": 14988 }, { "epoch": 2.75, "learning_rate": 1.5270725229087915e-05, "loss": 0.0702, "step": 14989 }, { "epoch": 2.75, "learning_rate": 1.5270094282449365e-05, "loss": 0.0705, "step": 14990 }, { "epoch": 2.75, "learning_rate": 1.5269463306761932e-05, "loss": 0.0691, "step": 14991 }, { "epoch": 2.75, "learning_rate": 1.5268832302029095e-05, "loss": 0.0682, "step": 14992 }, { "epoch": 2.75, "learning_rate": 1.526820126825433e-05, "loss": 0.0641, "step": 14993 }, { "epoch": 2.75, "learning_rate": 1.5267570205441117e-05, "loss": 0.0734, "step": 14994 }, { "epoch": 2.75, "learning_rate": 1.5266939113592937e-05, "loss": 0.0727, "step": 14995 }, { "epoch": 2.75, "learning_rate": 1.5266307992713262e-05, "loss": 0.0674, "step": 14996 }, { "epoch": 2.75, "learning_rate": 1.526567684280557e-05, "loss": 0.073, "step": 14997 }, { "epoch": 2.75, "learning_rate": 1.526504566387335e-05, "loss": 0.0634, "step": 14998 }, { "epoch": 2.75, "learning_rate": 1.5264414455920072e-05, "loss": 0.072, "step": 14999 }, { "epoch": 2.75, "learning_rate": 1.5263783218949216e-05, "loss": 0.0657, "step": 15000 }, { "epoch": 2.75, "learning_rate": 1.526315195296427e-05, "loss": 0.0624, "step": 15001 }, { "epoch": 2.75, "learning_rate": 1.52625206579687e-05, "loss": 0.0704, "step": 15002 }, { "epoch": 2.75, "learning_rate": 1.5261889333965996e-05, "loss": 0.0706, "step": 15003 }, { "epoch": 2.75, "learning_rate": 1.5261257980959635e-05, "loss": 0.0595, "step": 15004 }, { "epoch": 2.75, "learning_rate": 1.52606265989531e-05, "loss": 0.0695, "step": 15005 }, { "epoch": 2.75, "learning_rate": 1.5259995187949864e-05, "loss": 0.0766, "step": 15006 }, { "epoch": 2.75, "learning_rate": 1.5259363747953413e-05, "loss": 0.0688, "step": 15007 }, { "epoch": 2.75, "learning_rate": 1.5258732278967222e-05, "loss": 0.0656, "step": 15008 }, { "epoch": 2.75, "learning_rate": 1.5258100780994778e-05, "loss": 0.0706, "step": 15009 }, { "epoch": 2.75, "learning_rate": 1.5257469254039561e-05, "loss": 0.0734, "step": 15010 }, { "epoch": 2.75, "learning_rate": 1.5256837698105047e-05, "loss": 0.0605, "step": 15011 }, { "epoch": 2.75, "learning_rate": 1.5256206113194725e-05, "loss": 0.064, "step": 15012 }, { "epoch": 2.75, "learning_rate": 1.5255574499312067e-05, "loss": 0.0621, "step": 15013 }, { "epoch": 2.75, "learning_rate": 1.5254942856460561e-05, "loss": 0.0646, "step": 15014 }, { "epoch": 2.75, "learning_rate": 1.5254311184643688e-05, "loss": 0.0714, "step": 15015 }, { "epoch": 2.75, "learning_rate": 1.5253679483864927e-05, "loss": 0.0649, "step": 15016 }, { "epoch": 2.75, "learning_rate": 1.5253047754127766e-05, "loss": 0.0655, "step": 15017 }, { "epoch": 2.75, "learning_rate": 1.5252415995435677e-05, "loss": 0.072, "step": 15018 }, { "epoch": 2.75, "learning_rate": 1.525178420779215e-05, "loss": 0.0701, "step": 15019 }, { "epoch": 2.75, "learning_rate": 1.5251152391200665e-05, "loss": 0.0601, "step": 15020 }, { "epoch": 2.75, "learning_rate": 1.5250520545664707e-05, "loss": 0.0635, "step": 15021 }, { "epoch": 2.75, "learning_rate": 1.5249888671187755e-05, "loss": 0.058, "step": 15022 }, { "epoch": 2.75, "learning_rate": 1.5249256767773296e-05, "loss": 0.0685, "step": 15023 }, { "epoch": 2.76, "learning_rate": 1.5248624835424805e-05, "loss": 0.0702, "step": 15024 }, { "epoch": 2.76, "learning_rate": 1.5247992874145776e-05, "loss": 0.069, "step": 15025 }, { "epoch": 2.76, "learning_rate": 1.5247360883939685e-05, "loss": 0.0717, "step": 15026 }, { "epoch": 2.76, "learning_rate": 1.5246728864810018e-05, "loss": 0.0637, "step": 15027 }, { "epoch": 2.76, "learning_rate": 1.5246096816760263e-05, "loss": 0.0696, "step": 15028 }, { "epoch": 2.76, "learning_rate": 1.5245464739793892e-05, "loss": 0.0689, "step": 15029 }, { "epoch": 2.76, "learning_rate": 1.5244832633914402e-05, "loss": 0.0743, "step": 15030 }, { "epoch": 2.76, "learning_rate": 1.524420049912527e-05, "loss": 0.0764, "step": 15031 }, { "epoch": 2.76, "learning_rate": 1.5243568335429984e-05, "loss": 0.0783, "step": 15032 }, { "epoch": 2.76, "learning_rate": 1.5242936142832023e-05, "loss": 0.0632, "step": 15033 }, { "epoch": 2.76, "learning_rate": 1.5242303921334879e-05, "loss": 0.0636, "step": 15034 }, { "epoch": 2.76, "learning_rate": 1.5241671670942027e-05, "loss": 0.0654, "step": 15035 }, { "epoch": 2.76, "learning_rate": 1.5241039391656963e-05, "loss": 0.0782, "step": 15036 }, { "epoch": 2.76, "learning_rate": 1.5240407083483164e-05, "loss": 0.0614, "step": 15037 }, { "epoch": 2.76, "learning_rate": 1.523977474642412e-05, "loss": 0.0742, "step": 15038 }, { "epoch": 2.76, "learning_rate": 1.5239142380483313e-05, "loss": 0.0679, "step": 15039 }, { "epoch": 2.76, "learning_rate": 1.5238509985664234e-05, "loss": 0.0754, "step": 15040 }, { "epoch": 2.76, "learning_rate": 1.5237877561970364e-05, "loss": 0.0722, "step": 15041 }, { "epoch": 2.76, "learning_rate": 1.523724510940519e-05, "loss": 0.0718, "step": 15042 }, { "epoch": 2.76, "learning_rate": 1.5236612627972197e-05, "loss": 0.0643, "step": 15043 }, { "epoch": 2.76, "learning_rate": 1.5235980117674874e-05, "loss": 0.0618, "step": 15044 }, { "epoch": 2.76, "learning_rate": 1.5235347578516706e-05, "loss": 0.0757, "step": 15045 }, { "epoch": 2.76, "learning_rate": 1.5234715010501178e-05, "loss": 0.0701, "step": 15046 }, { "epoch": 2.76, "learning_rate": 1.5234082413631778e-05, "loss": 0.06, "step": 15047 }, { "epoch": 2.76, "learning_rate": 1.5233449787911995e-05, "loss": 0.0644, "step": 15048 }, { "epoch": 2.76, "learning_rate": 1.5232817133345314e-05, "loss": 0.0786, "step": 15049 }, { "epoch": 2.76, "learning_rate": 1.523218444993522e-05, "loss": 0.0764, "step": 15050 }, { "epoch": 2.76, "learning_rate": 1.5231551737685206e-05, "loss": 0.0733, "step": 15051 }, { "epoch": 2.76, "learning_rate": 1.5230918996598757e-05, "loss": 0.0579, "step": 15052 }, { "epoch": 2.76, "learning_rate": 1.5230286226679356e-05, "loss": 0.0861, "step": 15053 }, { "epoch": 2.76, "learning_rate": 1.5229653427930497e-05, "loss": 0.0753, "step": 15054 }, { "epoch": 2.76, "learning_rate": 1.5229020600355664e-05, "loss": 0.0789, "step": 15055 }, { "epoch": 2.76, "learning_rate": 1.5228387743958349e-05, "loss": 0.0694, "step": 15056 }, { "epoch": 2.76, "learning_rate": 1.5227754858742039e-05, "loss": 0.0803, "step": 15057 }, { "epoch": 2.76, "learning_rate": 1.522712194471022e-05, "loss": 0.0722, "step": 15058 }, { "epoch": 2.76, "learning_rate": 1.522648900186638e-05, "loss": 0.0772, "step": 15059 }, { "epoch": 2.76, "learning_rate": 1.5225856030214015e-05, "loss": 0.0627, "step": 15060 }, { "epoch": 2.76, "learning_rate": 1.5225223029756605e-05, "loss": 0.0606, "step": 15061 }, { "epoch": 2.76, "learning_rate": 1.5224590000497646e-05, "loss": 0.0682, "step": 15062 }, { "epoch": 2.76, "learning_rate": 1.5223956942440626e-05, "loss": 0.0683, "step": 15063 }, { "epoch": 2.76, "learning_rate": 1.5223323855589027e-05, "loss": 0.071, "step": 15064 }, { "epoch": 2.76, "learning_rate": 1.5222690739946351e-05, "loss": 0.069, "step": 15065 }, { "epoch": 2.76, "learning_rate": 1.5222057595516075e-05, "loss": 0.0653, "step": 15066 }, { "epoch": 2.76, "learning_rate": 1.52214244223017e-05, "loss": 0.0648, "step": 15067 }, { "epoch": 2.76, "learning_rate": 1.5220791220306707e-05, "loss": 0.0658, "step": 15068 }, { "epoch": 2.76, "learning_rate": 1.5220157989534597e-05, "loss": 0.0689, "step": 15069 }, { "epoch": 2.76, "learning_rate": 1.521952472998885e-05, "loss": 0.0698, "step": 15070 }, { "epoch": 2.76, "learning_rate": 1.521889144167296e-05, "loss": 0.0689, "step": 15071 }, { "epoch": 2.76, "learning_rate": 1.5218258124590422e-05, "loss": 0.0782, "step": 15072 }, { "epoch": 2.76, "learning_rate": 1.5217624778744718e-05, "loss": 0.0632, "step": 15073 }, { "epoch": 2.76, "learning_rate": 1.5216991404139345e-05, "loss": 0.0617, "step": 15074 }, { "epoch": 2.76, "learning_rate": 1.5216358000777795e-05, "loss": 0.0652, "step": 15075 }, { "epoch": 2.76, "learning_rate": 1.5215724568663555e-05, "loss": 0.0736, "step": 15076 }, { "epoch": 2.76, "learning_rate": 1.521509110780012e-05, "loss": 0.0622, "step": 15077 }, { "epoch": 2.77, "learning_rate": 1.5214457618190981e-05, "loss": 0.0719, "step": 15078 }, { "epoch": 2.77, "learning_rate": 1.5213824099839628e-05, "loss": 0.0648, "step": 15079 }, { "epoch": 2.77, "learning_rate": 1.5213190552749557e-05, "loss": 0.0727, "step": 15080 }, { "epoch": 2.77, "learning_rate": 1.5212556976924257e-05, "loss": 0.0649, "step": 15081 }, { "epoch": 2.77, "learning_rate": 1.5211923372367219e-05, "loss": 0.0791, "step": 15082 }, { "epoch": 2.77, "learning_rate": 1.521128973908194e-05, "loss": 0.0723, "step": 15083 }, { "epoch": 2.77, "learning_rate": 1.5210656077071909e-05, "loss": 0.0778, "step": 15084 }, { "epoch": 2.77, "learning_rate": 1.521002238634062e-05, "loss": 0.064, "step": 15085 }, { "epoch": 2.77, "learning_rate": 1.5209388666891563e-05, "loss": 0.0672, "step": 15086 }, { "epoch": 2.77, "learning_rate": 1.5208754918728237e-05, "loss": 0.0605, "step": 15087 }, { "epoch": 2.77, "learning_rate": 1.5208121141854126e-05, "loss": 0.0761, "step": 15088 }, { "epoch": 2.77, "learning_rate": 1.5207487336272735e-05, "loss": 0.0632, "step": 15089 }, { "epoch": 2.77, "learning_rate": 1.5206853501987549e-05, "loss": 0.0717, "step": 15090 }, { "epoch": 2.77, "learning_rate": 1.5206219639002066e-05, "loss": 0.0628, "step": 15091 }, { "epoch": 2.77, "learning_rate": 1.520558574731978e-05, "loss": 0.0668, "step": 15092 }, { "epoch": 2.77, "learning_rate": 1.520495182694418e-05, "loss": 0.0756, "step": 15093 }, { "epoch": 2.77, "learning_rate": 1.5204317877878768e-05, "loss": 0.0695, "step": 15094 }, { "epoch": 2.77, "learning_rate": 1.5203683900127027e-05, "loss": 0.0746, "step": 15095 }, { "epoch": 2.77, "learning_rate": 1.5203049893692466e-05, "loss": 0.0659, "step": 15096 }, { "epoch": 2.77, "learning_rate": 1.5202415858578565e-05, "loss": 0.0786, "step": 15097 }, { "epoch": 2.77, "learning_rate": 1.520178179478883e-05, "loss": 0.0676, "step": 15098 }, { "epoch": 2.77, "learning_rate": 1.5201147702326751e-05, "loss": 0.077, "step": 15099 }, { "epoch": 2.77, "learning_rate": 1.5200513581195826e-05, "loss": 0.0605, "step": 15100 }, { "epoch": 2.77, "learning_rate": 1.5199879431399548e-05, "loss": 0.0712, "step": 15101 }, { "epoch": 2.77, "learning_rate": 1.5199245252941409e-05, "loss": 0.0703, "step": 15102 }, { "epoch": 2.77, "learning_rate": 1.5198611045824912e-05, "loss": 0.064, "step": 15103 }, { "epoch": 2.77, "learning_rate": 1.5197976810053546e-05, "loss": 0.0709, "step": 15104 }, { "epoch": 2.77, "learning_rate": 1.5197342545630814e-05, "loss": 0.055, "step": 15105 }, { "epoch": 2.77, "learning_rate": 1.5196708252560203e-05, "loss": 0.0715, "step": 15106 }, { "epoch": 2.77, "learning_rate": 1.5196073930845217e-05, "loss": 0.0719, "step": 15107 }, { "epoch": 2.77, "learning_rate": 1.519543958048935e-05, "loss": 0.0741, "step": 15108 }, { "epoch": 2.77, "learning_rate": 1.5194805201496098e-05, "loss": 0.0667, "step": 15109 }, { "epoch": 2.77, "learning_rate": 1.519417079386896e-05, "loss": 0.0669, "step": 15110 }, { "epoch": 2.77, "learning_rate": 1.5193536357611432e-05, "loss": 0.0672, "step": 15111 }, { "epoch": 2.77, "learning_rate": 1.5192901892727006e-05, "loss": 0.0671, "step": 15112 }, { "epoch": 2.77, "learning_rate": 1.5192267399219183e-05, "loss": 0.0661, "step": 15113 }, { "epoch": 2.77, "learning_rate": 1.5191632877091463e-05, "loss": 0.0663, "step": 15114 }, { "epoch": 2.77, "learning_rate": 1.519099832634734e-05, "loss": 0.0652, "step": 15115 }, { "epoch": 2.77, "learning_rate": 1.5190363746990311e-05, "loss": 0.0725, "step": 15116 }, { "epoch": 2.77, "learning_rate": 1.5189729139023877e-05, "loss": 0.0623, "step": 15117 }, { "epoch": 2.77, "learning_rate": 1.5189094502451535e-05, "loss": 0.0718, "step": 15118 }, { "epoch": 2.77, "learning_rate": 1.5188459837276781e-05, "loss": 0.0693, "step": 15119 }, { "epoch": 2.77, "learning_rate": 1.5187825143503116e-05, "loss": 0.0765, "step": 15120 }, { "epoch": 2.77, "learning_rate": 1.5187190421134034e-05, "loss": 0.0694, "step": 15121 }, { "epoch": 2.77, "learning_rate": 1.5186555670173042e-05, "loss": 0.0694, "step": 15122 }, { "epoch": 2.77, "learning_rate": 1.5185920890623634e-05, "loss": 0.0653, "step": 15123 }, { "epoch": 2.77, "learning_rate": 1.5185286082489303e-05, "loss": 0.071, "step": 15124 }, { "epoch": 2.77, "learning_rate": 1.5184651245773557e-05, "loss": 0.0731, "step": 15125 }, { "epoch": 2.77, "learning_rate": 1.5184016380479891e-05, "loss": 0.0636, "step": 15126 }, { "epoch": 2.77, "learning_rate": 1.5183381486611808e-05, "loss": 0.0669, "step": 15127 }, { "epoch": 2.77, "learning_rate": 1.5182746564172802e-05, "loss": 0.0612, "step": 15128 }, { "epoch": 2.77, "learning_rate": 1.5182111613166375e-05, "loss": 0.076, "step": 15129 }, { "epoch": 2.77, "learning_rate": 1.5181476633596028e-05, "loss": 0.0754, "step": 15130 }, { "epoch": 2.77, "learning_rate": 1.5180841625465262e-05, "loss": 0.0657, "step": 15131 }, { "epoch": 2.77, "learning_rate": 1.5180206588777573e-05, "loss": 0.0716, "step": 15132 }, { "epoch": 2.78, "learning_rate": 1.5179571523536465e-05, "loss": 0.0735, "step": 15133 }, { "epoch": 2.78, "learning_rate": 1.5178936429745439e-05, "loss": 0.0665, "step": 15134 }, { "epoch": 2.78, "learning_rate": 1.5178301307407992e-05, "loss": 0.0658, "step": 15135 }, { "epoch": 2.78, "learning_rate": 1.5177666156527629e-05, "loss": 0.0602, "step": 15136 }, { "epoch": 2.78, "learning_rate": 1.5177030977107847e-05, "loss": 0.0616, "step": 15137 }, { "epoch": 2.78, "learning_rate": 1.5176395769152148e-05, "loss": 0.0753, "step": 15138 }, { "epoch": 2.78, "learning_rate": 1.5175760532664035e-05, "loss": 0.0729, "step": 15139 }, { "epoch": 2.78, "learning_rate": 1.5175125267647012e-05, "loss": 0.0704, "step": 15140 }, { "epoch": 2.78, "learning_rate": 1.5174489974104574e-05, "loss": 0.0655, "step": 15141 }, { "epoch": 2.78, "learning_rate": 1.5173854652040225e-05, "loss": 0.0655, "step": 15142 }, { "epoch": 2.78, "learning_rate": 1.5173219301457472e-05, "loss": 0.0721, "step": 15143 }, { "epoch": 2.78, "learning_rate": 1.5172583922359812e-05, "loss": 0.0735, "step": 15144 }, { "epoch": 2.78, "learning_rate": 1.5171948514750746e-05, "loss": 0.0697, "step": 15145 }, { "epoch": 2.78, "learning_rate": 1.517131307863378e-05, "loss": 0.0673, "step": 15146 }, { "epoch": 2.78, "learning_rate": 1.5170677614012414e-05, "loss": 0.0724, "step": 15147 }, { "epoch": 2.78, "learning_rate": 1.5170042120890149e-05, "loss": 0.0649, "step": 15148 }, { "epoch": 2.78, "learning_rate": 1.51694065992705e-05, "loss": 0.0677, "step": 15149 }, { "epoch": 2.78, "learning_rate": 1.5168771049156953e-05, "loss": 0.0724, "step": 15150 }, { "epoch": 2.78, "learning_rate": 1.516813547055302e-05, "loss": 0.0683, "step": 15151 }, { "epoch": 2.78, "learning_rate": 1.5167499863462202e-05, "loss": 0.0648, "step": 15152 }, { "epoch": 2.78, "learning_rate": 1.516686422788801e-05, "loss": 0.0702, "step": 15153 }, { "epoch": 2.78, "learning_rate": 1.5166228563833933e-05, "loss": 0.0742, "step": 15154 }, { "epoch": 2.78, "learning_rate": 1.516559287130349e-05, "loss": 0.0798, "step": 15155 }, { "epoch": 2.78, "learning_rate": 1.5164957150300174e-05, "loss": 0.0645, "step": 15156 }, { "epoch": 2.78, "learning_rate": 1.5164321400827496e-05, "loss": 0.0671, "step": 15157 }, { "epoch": 2.78, "learning_rate": 1.5163685622888955e-05, "loss": 0.0577, "step": 15158 }, { "epoch": 2.78, "learning_rate": 1.5163049816488058e-05, "loss": 0.0567, "step": 15159 }, { "epoch": 2.78, "learning_rate": 1.5162413981628312e-05, "loss": 0.0685, "step": 15160 }, { "epoch": 2.78, "learning_rate": 1.5161778118313217e-05, "loss": 0.0723, "step": 15161 }, { "epoch": 2.78, "learning_rate": 1.5161142226546281e-05, "loss": 0.0657, "step": 15162 }, { "epoch": 2.78, "learning_rate": 1.5160506306331005e-05, "loss": 0.0686, "step": 15163 }, { "epoch": 2.78, "learning_rate": 1.5159870357670902e-05, "loss": 0.0631, "step": 15164 }, { "epoch": 2.78, "learning_rate": 1.5159234380569469e-05, "loss": 0.0662, "step": 15165 }, { "epoch": 2.78, "learning_rate": 1.5158598375030218e-05, "loss": 0.0611, "step": 15166 }, { "epoch": 2.78, "learning_rate": 1.5157962341056653e-05, "loss": 0.0667, "step": 15167 }, { "epoch": 2.78, "learning_rate": 1.5157326278652275e-05, "loss": 0.0585, "step": 15168 }, { "epoch": 2.78, "learning_rate": 1.5156690187820596e-05, "loss": 0.0578, "step": 15169 }, { "epoch": 2.78, "learning_rate": 1.515605406856512e-05, "loss": 0.0704, "step": 15170 }, { "epoch": 2.78, "learning_rate": 1.5155417920889352e-05, "loss": 0.0684, "step": 15171 }, { "epoch": 2.78, "learning_rate": 1.5154781744796798e-05, "loss": 0.061, "step": 15172 }, { "epoch": 2.78, "learning_rate": 1.5154145540290973e-05, "loss": 0.0668, "step": 15173 }, { "epoch": 2.78, "learning_rate": 1.5153509307375368e-05, "loss": 0.0671, "step": 15174 }, { "epoch": 2.78, "learning_rate": 1.5152873046053505e-05, "loss": 0.0685, "step": 15175 }, { "epoch": 2.78, "learning_rate": 1.5152236756328884e-05, "loss": 0.0721, "step": 15176 }, { "epoch": 2.78, "learning_rate": 1.5151600438205012e-05, "loss": 0.0734, "step": 15177 }, { "epoch": 2.78, "learning_rate": 1.5150964091685398e-05, "loss": 0.0686, "step": 15178 }, { "epoch": 2.78, "learning_rate": 1.515032771677355e-05, "loss": 0.0685, "step": 15179 }, { "epoch": 2.78, "learning_rate": 1.5149691313472974e-05, "loss": 0.0573, "step": 15180 }, { "epoch": 2.78, "learning_rate": 1.5149054881787178e-05, "loss": 0.0708, "step": 15181 }, { "epoch": 2.78, "learning_rate": 1.5148418421719674e-05, "loss": 0.0671, "step": 15182 }, { "epoch": 2.78, "learning_rate": 1.5147781933273963e-05, "loss": 0.0639, "step": 15183 }, { "epoch": 2.78, "learning_rate": 1.5147145416453561e-05, "loss": 0.0699, "step": 15184 }, { "epoch": 2.78, "learning_rate": 1.514650887126197e-05, "loss": 0.0655, "step": 15185 }, { "epoch": 2.78, "learning_rate": 1.5145872297702706e-05, "loss": 0.0762, "step": 15186 }, { "epoch": 2.79, "learning_rate": 1.514523569577927e-05, "loss": 0.0735, "step": 15187 }, { "epoch": 2.79, "learning_rate": 1.5144599065495174e-05, "loss": 0.0658, "step": 15188 }, { "epoch": 2.79, "learning_rate": 1.514396240685393e-05, "loss": 0.0679, "step": 15189 }, { "epoch": 2.79, "learning_rate": 1.514332571985904e-05, "loss": 0.0605, "step": 15190 }, { "epoch": 2.79, "learning_rate": 1.5142689004514025e-05, "loss": 0.0697, "step": 15191 }, { "epoch": 2.79, "learning_rate": 1.514205226082238e-05, "loss": 0.0722, "step": 15192 }, { "epoch": 2.79, "learning_rate": 1.5141415488787628e-05, "loss": 0.0738, "step": 15193 }, { "epoch": 2.79, "learning_rate": 1.514077868841327e-05, "loss": 0.0747, "step": 15194 }, { "epoch": 2.79, "learning_rate": 1.5140141859702823e-05, "loss": 0.0652, "step": 15195 }, { "epoch": 2.79, "learning_rate": 1.5139505002659792e-05, "loss": 0.0732, "step": 15196 }, { "epoch": 2.79, "learning_rate": 1.5138868117287689e-05, "loss": 0.0502, "step": 15197 }, { "epoch": 2.79, "learning_rate": 1.5138231203590024e-05, "loss": 0.0611, "step": 15198 }, { "epoch": 2.79, "learning_rate": 1.5137594261570306e-05, "loss": 0.0702, "step": 15199 }, { "epoch": 2.79, "learning_rate": 1.5136957291232053e-05, "loss": 0.0774, "step": 15200 }, { "epoch": 2.79, "learning_rate": 1.5136320292578765e-05, "loss": 0.0686, "step": 15201 }, { "epoch": 2.79, "learning_rate": 1.5135683265613965e-05, "loss": 0.0674, "step": 15202 }, { "epoch": 2.79, "learning_rate": 1.5135046210341155e-05, "loss": 0.0727, "step": 15203 }, { "epoch": 2.79, "learning_rate": 1.5134409126763854e-05, "loss": 0.0704, "step": 15204 }, { "epoch": 2.79, "learning_rate": 1.5133772014885566e-05, "loss": 0.0704, "step": 15205 }, { "epoch": 2.79, "learning_rate": 1.5133134874709807e-05, "loss": 0.0724, "step": 15206 }, { "epoch": 2.79, "learning_rate": 1.5132497706240088e-05, "loss": 0.0637, "step": 15207 }, { "epoch": 2.79, "learning_rate": 1.5131860509479923e-05, "loss": 0.0636, "step": 15208 }, { "epoch": 2.79, "learning_rate": 1.5131223284432823e-05, "loss": 0.0713, "step": 15209 }, { "epoch": 2.79, "learning_rate": 1.5130586031102296e-05, "loss": 0.0701, "step": 15210 }, { "epoch": 2.79, "learning_rate": 1.5129948749491862e-05, "loss": 0.0605, "step": 15211 }, { "epoch": 2.79, "learning_rate": 1.512931143960503e-05, "loss": 0.0706, "step": 15212 }, { "epoch": 2.79, "learning_rate": 1.5128674101445313e-05, "loss": 0.0743, "step": 15213 }, { "epoch": 2.79, "learning_rate": 1.5128036735016225e-05, "loss": 0.0845, "step": 15214 }, { "epoch": 2.79, "learning_rate": 1.5127399340321276e-05, "loss": 0.0736, "step": 15215 }, { "epoch": 2.79, "learning_rate": 1.5126761917363984e-05, "loss": 0.0689, "step": 15216 }, { "epoch": 2.79, "learning_rate": 1.512612446614786e-05, "loss": 0.0592, "step": 15217 }, { "epoch": 2.79, "learning_rate": 1.5125486986676415e-05, "loss": 0.0675, "step": 15218 }, { "epoch": 2.79, "learning_rate": 1.5124849478953168e-05, "loss": 0.0627, "step": 15219 }, { "epoch": 2.79, "learning_rate": 1.5124211942981633e-05, "loss": 0.066, "step": 15220 }, { "epoch": 2.79, "learning_rate": 1.5123574378765318e-05, "loss": 0.0643, "step": 15221 }, { "epoch": 2.79, "learning_rate": 1.5122936786307743e-05, "loss": 0.0784, "step": 15222 }, { "epoch": 2.79, "learning_rate": 1.5122299165612417e-05, "loss": 0.0756, "step": 15223 }, { "epoch": 2.79, "learning_rate": 1.5121661516682862e-05, "loss": 0.0768, "step": 15224 }, { "epoch": 2.79, "learning_rate": 1.5121023839522587e-05, "loss": 0.0782, "step": 15225 }, { "epoch": 2.79, "learning_rate": 1.512038613413511e-05, "loss": 0.0639, "step": 15226 }, { "epoch": 2.79, "learning_rate": 1.5119748400523944e-05, "loss": 0.0658, "step": 15227 }, { "epoch": 2.79, "learning_rate": 1.5119110638692602e-05, "loss": 0.0725, "step": 15228 }, { "epoch": 2.79, "learning_rate": 1.5118472848644605e-05, "loss": 0.0822, "step": 15229 }, { "epoch": 2.79, "learning_rate": 1.5117835030383465e-05, "loss": 0.0768, "step": 15230 }, { "epoch": 2.79, "learning_rate": 1.51171971839127e-05, "loss": 0.0683, "step": 15231 }, { "epoch": 2.79, "learning_rate": 1.5116559309235825e-05, "loss": 0.0747, "step": 15232 }, { "epoch": 2.79, "learning_rate": 1.5115921406356351e-05, "loss": 0.0673, "step": 15233 }, { "epoch": 2.79, "learning_rate": 1.5115283475277799e-05, "loss": 0.0738, "step": 15234 }, { "epoch": 2.79, "learning_rate": 1.511464551600369e-05, "loss": 0.0766, "step": 15235 }, { "epoch": 2.79, "learning_rate": 1.511400752853753e-05, "loss": 0.0564, "step": 15236 }, { "epoch": 2.79, "learning_rate": 1.5113369512882842e-05, "loss": 0.0706, "step": 15237 }, { "epoch": 2.79, "learning_rate": 1.5112731469043143e-05, "loss": 0.0648, "step": 15238 }, { "epoch": 2.79, "learning_rate": 1.5112093397021945e-05, "loss": 0.0663, "step": 15239 }, { "epoch": 2.79, "learning_rate": 1.5111455296822772e-05, "loss": 0.0644, "step": 15240 }, { "epoch": 2.79, "learning_rate": 1.5110817168449137e-05, "loss": 0.0742, "step": 15241 }, { "epoch": 2.8, "learning_rate": 1.5110179011904556e-05, "loss": 0.057, "step": 15242 }, { "epoch": 2.8, "learning_rate": 1.5109540827192548e-05, "loss": 0.0655, "step": 15243 }, { "epoch": 2.8, "learning_rate": 1.5108902614316634e-05, "loss": 0.0704, "step": 15244 }, { "epoch": 2.8, "learning_rate": 1.5108264373280327e-05, "loss": 0.0736, "step": 15245 }, { "epoch": 2.8, "learning_rate": 1.510762610408715e-05, "loss": 0.0768, "step": 15246 }, { "epoch": 2.8, "learning_rate": 1.5106987806740616e-05, "loss": 0.0712, "step": 15247 }, { "epoch": 2.8, "learning_rate": 1.5106349481244244e-05, "loss": 0.0726, "step": 15248 }, { "epoch": 2.8, "learning_rate": 1.5105711127601555e-05, "loss": 0.0672, "step": 15249 }, { "epoch": 2.8, "learning_rate": 1.510507274581607e-05, "loss": 0.0641, "step": 15250 }, { "epoch": 2.8, "learning_rate": 1.51044343358913e-05, "loss": 0.0656, "step": 15251 }, { "epoch": 2.8, "learning_rate": 1.510379589783077e-05, "loss": 0.0818, "step": 15252 }, { "epoch": 2.8, "learning_rate": 1.5103157431638e-05, "loss": 0.0677, "step": 15253 }, { "epoch": 2.8, "learning_rate": 1.5102518937316504e-05, "loss": 0.0732, "step": 15254 }, { "epoch": 2.8, "learning_rate": 1.5101880414869804e-05, "loss": 0.073, "step": 15255 }, { "epoch": 2.8, "learning_rate": 1.5101241864301419e-05, "loss": 0.0666, "step": 15256 }, { "epoch": 2.8, "learning_rate": 1.5100603285614872e-05, "loss": 0.0598, "step": 15257 }, { "epoch": 2.8, "learning_rate": 1.5099964678813677e-05, "loss": 0.0661, "step": 15258 }, { "epoch": 2.8, "learning_rate": 1.5099326043901361e-05, "loss": 0.064, "step": 15259 }, { "epoch": 2.8, "learning_rate": 1.5098687380881436e-05, "loss": 0.0745, "step": 15260 }, { "epoch": 2.8, "learning_rate": 1.509804868975743e-05, "loss": 0.0639, "step": 15261 }, { "epoch": 2.8, "learning_rate": 1.5097409970532862e-05, "loss": 0.0601, "step": 15262 }, { "epoch": 2.8, "learning_rate": 1.5096771223211244e-05, "loss": 0.0624, "step": 15263 }, { "epoch": 2.8, "learning_rate": 1.5096132447796107e-05, "loss": 0.0652, "step": 15264 }, { "epoch": 2.8, "learning_rate": 1.509549364429097e-05, "loss": 0.0715, "step": 15265 }, { "epoch": 2.8, "learning_rate": 1.5094854812699354e-05, "loss": 0.0694, "step": 15266 }, { "epoch": 2.8, "learning_rate": 1.5094215953024773e-05, "loss": 0.0636, "step": 15267 }, { "epoch": 2.8, "learning_rate": 1.509357706527076e-05, "loss": 0.0672, "step": 15268 }, { "epoch": 2.8, "learning_rate": 1.509293814944083e-05, "loss": 0.0671, "step": 15269 }, { "epoch": 2.8, "learning_rate": 1.5092299205538506e-05, "loss": 0.074, "step": 15270 }, { "epoch": 2.8, "learning_rate": 1.5091660233567306e-05, "loss": 0.0732, "step": 15271 }, { "epoch": 2.8, "learning_rate": 1.5091021233530759e-05, "loss": 0.0811, "step": 15272 }, { "epoch": 2.8, "learning_rate": 1.509038220543238e-05, "loss": 0.0681, "step": 15273 }, { "epoch": 2.8, "learning_rate": 1.50897431492757e-05, "loss": 0.066, "step": 15274 }, { "epoch": 2.8, "learning_rate": 1.5089104065064233e-05, "loss": 0.0667, "step": 15275 }, { "epoch": 2.8, "learning_rate": 1.5088464952801505e-05, "loss": 0.0763, "step": 15276 }, { "epoch": 2.8, "learning_rate": 1.5087825812491042e-05, "loss": 0.0653, "step": 15277 }, { "epoch": 2.8, "learning_rate": 1.5087186644136361e-05, "loss": 0.0744, "step": 15278 }, { "epoch": 2.8, "learning_rate": 1.508654744774099e-05, "loss": 0.071, "step": 15279 }, { "epoch": 2.8, "learning_rate": 1.5085908223308447e-05, "loss": 0.0697, "step": 15280 }, { "epoch": 2.8, "learning_rate": 1.5085268970842261e-05, "loss": 0.0813, "step": 15281 }, { "epoch": 2.8, "learning_rate": 1.5084629690345956e-05, "loss": 0.0671, "step": 15282 }, { "epoch": 2.8, "learning_rate": 1.5083990381823048e-05, "loss": 0.0652, "step": 15283 }, { "epoch": 2.8, "learning_rate": 1.5083351045277068e-05, "loss": 0.0722, "step": 15284 }, { "epoch": 2.8, "learning_rate": 1.5082711680711539e-05, "loss": 0.0696, "step": 15285 }, { "epoch": 2.8, "learning_rate": 1.5082072288129983e-05, "loss": 0.0739, "step": 15286 }, { "epoch": 2.8, "learning_rate": 1.5081432867535925e-05, "loss": 0.0579, "step": 15287 }, { "epoch": 2.8, "learning_rate": 1.5080793418932891e-05, "loss": 0.0625, "step": 15288 }, { "epoch": 2.8, "learning_rate": 1.5080153942324404e-05, "loss": 0.0701, "step": 15289 }, { "epoch": 2.8, "learning_rate": 1.5079514437713991e-05, "loss": 0.0618, "step": 15290 }, { "epoch": 2.8, "learning_rate": 1.5078874905105175e-05, "loss": 0.0722, "step": 15291 }, { "epoch": 2.8, "learning_rate": 1.507823534450148e-05, "loss": 0.0726, "step": 15292 }, { "epoch": 2.8, "learning_rate": 1.5077595755906431e-05, "loss": 0.0765, "step": 15293 }, { "epoch": 2.8, "learning_rate": 1.5076956139323562e-05, "loss": 0.0723, "step": 15294 }, { "epoch": 2.8, "learning_rate": 1.5076316494756387e-05, "loss": 0.075, "step": 15295 }, { "epoch": 2.81, "learning_rate": 1.5075676822208434e-05, "loss": 0.0582, "step": 15296 }, { "epoch": 2.81, "learning_rate": 1.5075037121683235e-05, "loss": 0.064, "step": 15297 }, { "epoch": 2.81, "learning_rate": 1.5074397393184313e-05, "loss": 0.0688, "step": 15298 }, { "epoch": 2.81, "learning_rate": 1.5073757636715192e-05, "loss": 0.0633, "step": 15299 }, { "epoch": 2.81, "learning_rate": 1.5073117852279401e-05, "loss": 0.0577, "step": 15300 }, { "epoch": 2.81, "learning_rate": 1.5072478039880468e-05, "loss": 0.0818, "step": 15301 }, { "epoch": 2.81, "learning_rate": 1.5071838199521912e-05, "loss": 0.0696, "step": 15302 }, { "epoch": 2.81, "learning_rate": 1.507119833120727e-05, "loss": 0.0793, "step": 15303 }, { "epoch": 2.81, "learning_rate": 1.507055843494006e-05, "loss": 0.067, "step": 15304 }, { "epoch": 2.81, "learning_rate": 1.5069918510723813e-05, "loss": 0.0756, "step": 15305 }, { "epoch": 2.81, "learning_rate": 1.506927855856206e-05, "loss": 0.0594, "step": 15306 }, { "epoch": 2.81, "learning_rate": 1.5068638578458323e-05, "loss": 0.0762, "step": 15307 }, { "epoch": 2.81, "learning_rate": 1.5067998570416132e-05, "loss": 0.0771, "step": 15308 }, { "epoch": 2.81, "learning_rate": 1.5067358534439013e-05, "loss": 0.0613, "step": 15309 }, { "epoch": 2.81, "learning_rate": 1.5066718470530495e-05, "loss": 0.0787, "step": 15310 }, { "epoch": 2.81, "learning_rate": 1.5066078378694107e-05, "loss": 0.0654, "step": 15311 }, { "epoch": 2.81, "learning_rate": 1.5065438258933376e-05, "loss": 0.0673, "step": 15312 }, { "epoch": 2.81, "learning_rate": 1.5064798111251831e-05, "loss": 0.0792, "step": 15313 }, { "epoch": 2.81, "learning_rate": 1.5064157935652998e-05, "loss": 0.0716, "step": 15314 }, { "epoch": 2.81, "learning_rate": 1.5063517732140413e-05, "loss": 0.0718, "step": 15315 }, { "epoch": 2.81, "learning_rate": 1.5062877500717594e-05, "loss": 0.0686, "step": 15316 }, { "epoch": 2.81, "learning_rate": 1.5062237241388079e-05, "loss": 0.0716, "step": 15317 }, { "epoch": 2.81, "learning_rate": 1.5061596954155394e-05, "loss": 0.067, "step": 15318 }, { "epoch": 2.81, "learning_rate": 1.5060956639023063e-05, "loss": 0.0619, "step": 15319 }, { "epoch": 2.81, "learning_rate": 1.5060316295994624e-05, "loss": 0.0723, "step": 15320 }, { "epoch": 2.81, "learning_rate": 1.5059675925073605e-05, "loss": 0.0622, "step": 15321 }, { "epoch": 2.81, "learning_rate": 1.505903552626353e-05, "loss": 0.0698, "step": 15322 }, { "epoch": 2.81, "learning_rate": 1.5058395099567935e-05, "loss": 0.0647, "step": 15323 }, { "epoch": 2.81, "learning_rate": 1.5057754644990347e-05, "loss": 0.0701, "step": 15324 }, { "epoch": 2.81, "learning_rate": 1.5057114162534298e-05, "loss": 0.0647, "step": 15325 }, { "epoch": 2.81, "learning_rate": 1.5056473652203316e-05, "loss": 0.0743, "step": 15326 }, { "epoch": 2.81, "learning_rate": 1.5055833114000932e-05, "loss": 0.0789, "step": 15327 }, { "epoch": 2.81, "learning_rate": 1.5055192547930681e-05, "loss": 0.0705, "step": 15328 }, { "epoch": 2.81, "learning_rate": 1.5054551953996086e-05, "loss": 0.0681, "step": 15329 }, { "epoch": 2.81, "learning_rate": 1.5053911332200686e-05, "loss": 0.0666, "step": 15330 }, { "epoch": 2.81, "learning_rate": 1.5053270682548005e-05, "loss": 0.0591, "step": 15331 }, { "epoch": 2.81, "learning_rate": 1.505263000504158e-05, "loss": 0.0752, "step": 15332 }, { "epoch": 2.81, "learning_rate": 1.5051989299684937e-05, "loss": 0.0607, "step": 15333 }, { "epoch": 2.81, "learning_rate": 1.5051348566481614e-05, "loss": 0.0673, "step": 15334 }, { "epoch": 2.81, "learning_rate": 1.5050707805435134e-05, "loss": 0.0752, "step": 15335 }, { "epoch": 2.81, "learning_rate": 1.5050067016549042e-05, "loss": 0.0733, "step": 15336 }, { "epoch": 2.81, "learning_rate": 1.504942619982686e-05, "loss": 0.0725, "step": 15337 }, { "epoch": 2.81, "learning_rate": 1.5048785355272118e-05, "loss": 0.065, "step": 15338 }, { "epoch": 2.81, "learning_rate": 1.5048144482888356e-05, "loss": 0.072, "step": 15339 }, { "epoch": 2.81, "learning_rate": 1.5047503582679102e-05, "loss": 0.0619, "step": 15340 }, { "epoch": 2.81, "learning_rate": 1.5046862654647893e-05, "loss": 0.0753, "step": 15341 }, { "epoch": 2.81, "learning_rate": 1.5046221698798257e-05, "loss": 0.0751, "step": 15342 }, { "epoch": 2.81, "learning_rate": 1.504558071513373e-05, "loss": 0.0704, "step": 15343 }, { "epoch": 2.81, "learning_rate": 1.5044939703657839e-05, "loss": 0.0653, "step": 15344 }, { "epoch": 2.81, "learning_rate": 1.5044298664374128e-05, "loss": 0.075, "step": 15345 }, { "epoch": 2.81, "learning_rate": 1.5043657597286122e-05, "loss": 0.0729, "step": 15346 }, { "epoch": 2.81, "learning_rate": 1.5043016502397359e-05, "loss": 0.0746, "step": 15347 }, { "epoch": 2.81, "learning_rate": 1.5042375379711369e-05, "loss": 0.075, "step": 15348 }, { "epoch": 2.81, "learning_rate": 1.5041734229231688e-05, "loss": 0.0703, "step": 15349 }, { "epoch": 2.81, "learning_rate": 1.504109305096185e-05, "loss": 0.0692, "step": 15350 }, { "epoch": 2.82, "learning_rate": 1.5040451844905389e-05, "loss": 0.0719, "step": 15351 }, { "epoch": 2.82, "learning_rate": 1.503981061106584e-05, "loss": 0.0724, "step": 15352 }, { "epoch": 2.82, "learning_rate": 1.5039169349446734e-05, "loss": 0.0649, "step": 15353 }, { "epoch": 2.82, "learning_rate": 1.5038528060051612e-05, "loss": 0.0837, "step": 15354 }, { "epoch": 2.82, "learning_rate": 1.5037886742884003e-05, "loss": 0.0692, "step": 15355 }, { "epoch": 2.82, "learning_rate": 1.5037245397947446e-05, "loss": 0.0718, "step": 15356 }, { "epoch": 2.82, "learning_rate": 1.5036604025245473e-05, "loss": 0.0727, "step": 15357 }, { "epoch": 2.82, "learning_rate": 1.5035962624781621e-05, "loss": 0.0715, "step": 15358 }, { "epoch": 2.82, "learning_rate": 1.5035321196559426e-05, "loss": 0.0602, "step": 15359 }, { "epoch": 2.82, "learning_rate": 1.5034679740582423e-05, "loss": 0.0704, "step": 15360 }, { "epoch": 2.82, "learning_rate": 1.5034038256854145e-05, "loss": 0.0747, "step": 15361 }, { "epoch": 2.82, "learning_rate": 1.5033396745378133e-05, "loss": 0.0665, "step": 15362 }, { "epoch": 2.82, "learning_rate": 1.5032755206157919e-05, "loss": 0.0686, "step": 15363 }, { "epoch": 2.82, "learning_rate": 1.5032113639197037e-05, "loss": 0.0643, "step": 15364 }, { "epoch": 2.82, "learning_rate": 1.5031472044499032e-05, "loss": 0.0593, "step": 15365 }, { "epoch": 2.82, "learning_rate": 1.5030830422067433e-05, "loss": 0.0668, "step": 15366 }, { "epoch": 2.82, "learning_rate": 1.5030188771905778e-05, "loss": 0.0742, "step": 15367 }, { "epoch": 2.82, "learning_rate": 1.5029547094017605e-05, "loss": 0.0595, "step": 15368 }, { "epoch": 2.82, "learning_rate": 1.5028905388406451e-05, "loss": 0.0726, "step": 15369 }, { "epoch": 2.82, "learning_rate": 1.5028263655075854e-05, "loss": 0.0734, "step": 15370 }, { "epoch": 2.82, "learning_rate": 1.502762189402935e-05, "loss": 0.069, "step": 15371 }, { "epoch": 2.82, "learning_rate": 1.5026980105270475e-05, "loss": 0.068, "step": 15372 }, { "epoch": 2.82, "learning_rate": 1.5026338288802764e-05, "loss": 0.0774, "step": 15373 }, { "epoch": 2.82, "learning_rate": 1.5025696444629762e-05, "loss": 0.0656, "step": 15374 }, { "epoch": 2.82, "learning_rate": 1.5025054572755004e-05, "loss": 0.0653, "step": 15375 }, { "epoch": 2.82, "learning_rate": 1.502441267318203e-05, "loss": 0.0666, "step": 15376 }, { "epoch": 2.82, "learning_rate": 1.5023770745914372e-05, "loss": 0.0699, "step": 15377 }, { "epoch": 2.82, "learning_rate": 1.5023128790955572e-05, "loss": 0.0662, "step": 15378 }, { "epoch": 2.82, "learning_rate": 1.5022486808309171e-05, "loss": 0.0544, "step": 15379 }, { "epoch": 2.82, "learning_rate": 1.5021844797978704e-05, "loss": 0.0782, "step": 15380 }, { "epoch": 2.82, "learning_rate": 1.502120275996771e-05, "loss": 0.0625, "step": 15381 }, { "epoch": 2.82, "learning_rate": 1.5020560694279729e-05, "loss": 0.0693, "step": 15382 }, { "epoch": 2.82, "learning_rate": 1.50199186009183e-05, "loss": 0.0677, "step": 15383 }, { "epoch": 2.82, "learning_rate": 1.5019276479886961e-05, "loss": 0.0601, "step": 15384 }, { "epoch": 2.82, "learning_rate": 1.5018634331189254e-05, "loss": 0.0769, "step": 15385 }, { "epoch": 2.82, "learning_rate": 1.5017992154828718e-05, "loss": 0.0753, "step": 15386 }, { "epoch": 2.82, "learning_rate": 1.5017349950808888e-05, "loss": 0.0665, "step": 15387 }, { "epoch": 2.82, "learning_rate": 1.501670771913331e-05, "loss": 0.0636, "step": 15388 }, { "epoch": 2.82, "learning_rate": 1.5016065459805523e-05, "loss": 0.0642, "step": 15389 }, { "epoch": 2.82, "learning_rate": 1.5015423172829065e-05, "loss": 0.0633, "step": 15390 }, { "epoch": 2.82, "learning_rate": 1.5014780858207471e-05, "loss": 0.0605, "step": 15391 }, { "epoch": 2.82, "learning_rate": 1.5014138515944297e-05, "loss": 0.0585, "step": 15392 }, { "epoch": 2.82, "learning_rate": 1.5013496146043065e-05, "loss": 0.0785, "step": 15393 }, { "epoch": 2.82, "learning_rate": 1.501285374850733e-05, "loss": 0.0684, "step": 15394 }, { "epoch": 2.82, "learning_rate": 1.5012211323340626e-05, "loss": 0.062, "step": 15395 }, { "epoch": 2.82, "learning_rate": 1.5011568870546498e-05, "loss": 0.0736, "step": 15396 }, { "epoch": 2.82, "learning_rate": 1.5010926390128481e-05, "loss": 0.0727, "step": 15397 }, { "epoch": 2.82, "learning_rate": 1.5010283882090124e-05, "loss": 0.0657, "step": 15398 }, { "epoch": 2.82, "learning_rate": 1.5009641346434963e-05, "loss": 0.0679, "step": 15399 }, { "epoch": 2.82, "learning_rate": 1.5008998783166542e-05, "loss": 0.0647, "step": 15400 }, { "epoch": 2.82, "learning_rate": 1.5008356192288407e-05, "loss": 0.0721, "step": 15401 }, { "epoch": 2.82, "learning_rate": 1.5007713573804089e-05, "loss": 0.0749, "step": 15402 }, { "epoch": 2.82, "learning_rate": 1.5007070927717137e-05, "loss": 0.0747, "step": 15403 }, { "epoch": 2.82, "learning_rate": 1.5006428254031096e-05, "loss": 0.0714, "step": 15404 }, { "epoch": 2.83, "learning_rate": 1.5005785552749502e-05, "loss": 0.0777, "step": 15405 }, { "epoch": 2.83, "learning_rate": 1.50051428238759e-05, "loss": 0.0676, "step": 15406 }, { "epoch": 2.83, "learning_rate": 1.5004500067413838e-05, "loss": 0.0584, "step": 15407 }, { "epoch": 2.83, "learning_rate": 1.5003857283366849e-05, "loss": 0.0669, "step": 15408 }, { "epoch": 2.83, "learning_rate": 1.5003214471738487e-05, "loss": 0.0777, "step": 15409 }, { "epoch": 2.83, "learning_rate": 1.5002571632532285e-05, "loss": 0.072, "step": 15410 }, { "epoch": 2.83, "learning_rate": 1.5001928765751794e-05, "loss": 0.0659, "step": 15411 }, { "epoch": 2.83, "learning_rate": 1.5001285871400554e-05, "loss": 0.0728, "step": 15412 }, { "epoch": 2.83, "learning_rate": 1.5000642949482109e-05, "loss": 0.0707, "step": 15413 }, { "epoch": 2.83, "learning_rate": 1.5000000000000002e-05, "loss": 0.0636, "step": 15414 }, { "epoch": 2.83, "learning_rate": 1.4999357022957777e-05, "loss": 0.064, "step": 15415 }, { "epoch": 2.83, "learning_rate": 1.4998714018358983e-05, "loss": 0.0782, "step": 15416 }, { "epoch": 2.83, "learning_rate": 1.4998070986207155e-05, "loss": 0.0628, "step": 15417 }, { "epoch": 2.83, "learning_rate": 1.4997427926505846e-05, "loss": 0.0657, "step": 15418 }, { "epoch": 2.83, "learning_rate": 1.4996784839258596e-05, "loss": 0.0755, "step": 15419 }, { "epoch": 2.83, "learning_rate": 1.4996141724468951e-05, "loss": 0.0611, "step": 15420 }, { "epoch": 2.83, "learning_rate": 1.4995498582140452e-05, "loss": 0.0641, "step": 15421 }, { "epoch": 2.83, "learning_rate": 1.4994855412276656e-05, "loss": 0.0633, "step": 15422 }, { "epoch": 2.83, "learning_rate": 1.4994212214881097e-05, "loss": 0.0667, "step": 15423 }, { "epoch": 2.83, "learning_rate": 1.4993568989957318e-05, "loss": 0.069, "step": 15424 }, { "epoch": 2.83, "learning_rate": 1.4992925737508877e-05, "loss": 0.0646, "step": 15425 }, { "epoch": 2.83, "learning_rate": 1.4992282457539307e-05, "loss": 0.0621, "step": 15426 }, { "epoch": 2.83, "learning_rate": 1.4991639150052162e-05, "loss": 0.0744, "step": 15427 }, { "epoch": 2.83, "learning_rate": 1.4990995815050984e-05, "loss": 0.0619, "step": 15428 }, { "epoch": 2.83, "learning_rate": 1.499035245253932e-05, "loss": 0.0643, "step": 15429 }, { "epoch": 2.83, "learning_rate": 1.4989709062520716e-05, "loss": 0.0687, "step": 15430 }, { "epoch": 2.83, "learning_rate": 1.4989065644998721e-05, "loss": 0.083, "step": 15431 }, { "epoch": 2.83, "learning_rate": 1.498842219997688e-05, "loss": 0.0677, "step": 15432 }, { "epoch": 2.83, "learning_rate": 1.4987778727458735e-05, "loss": 0.0671, "step": 15433 }, { "epoch": 2.83, "learning_rate": 1.498713522744784e-05, "loss": 0.0599, "step": 15434 }, { "epoch": 2.83, "learning_rate": 1.4986491699947736e-05, "loss": 0.0768, "step": 15435 }, { "epoch": 2.83, "learning_rate": 1.4985848144961974e-05, "loss": 0.0722, "step": 15436 }, { "epoch": 2.83, "learning_rate": 1.49852045624941e-05, "loss": 0.0691, "step": 15437 }, { "epoch": 2.83, "learning_rate": 1.4984560952547665e-05, "loss": 0.0666, "step": 15438 }, { "epoch": 2.83, "learning_rate": 1.4983917315126208e-05, "loss": 0.0658, "step": 15439 }, { "epoch": 2.83, "learning_rate": 1.4983273650233286e-05, "loss": 0.0699, "step": 15440 }, { "epoch": 2.83, "learning_rate": 1.4982629957872438e-05, "loss": 0.0651, "step": 15441 }, { "epoch": 2.83, "learning_rate": 1.4981986238047222e-05, "loss": 0.0668, "step": 15442 }, { "epoch": 2.83, "learning_rate": 1.4981342490761181e-05, "loss": 0.056, "step": 15443 }, { "epoch": 2.83, "learning_rate": 1.498069871601786e-05, "loss": 0.0707, "step": 15444 }, { "epoch": 2.83, "learning_rate": 1.4980054913820814e-05, "loss": 0.0675, "step": 15445 }, { "epoch": 2.83, "learning_rate": 1.4979411084173587e-05, "loss": 0.0697, "step": 15446 }, { "epoch": 2.83, "learning_rate": 1.4978767227079729e-05, "loss": 0.0683, "step": 15447 }, { "epoch": 2.83, "learning_rate": 1.4978123342542791e-05, "loss": 0.0757, "step": 15448 }, { "epoch": 2.83, "learning_rate": 1.497747943056632e-05, "loss": 0.0706, "step": 15449 }, { "epoch": 2.83, "learning_rate": 1.4976835491153865e-05, "loss": 0.0625, "step": 15450 }, { "epoch": 2.83, "learning_rate": 1.4976191524308976e-05, "loss": 0.0624, "step": 15451 }, { "epoch": 2.83, "learning_rate": 1.4975547530035206e-05, "loss": 0.0679, "step": 15452 }, { "epoch": 2.83, "learning_rate": 1.4974903508336098e-05, "loss": 0.0812, "step": 15453 }, { "epoch": 2.83, "learning_rate": 1.4974259459215205e-05, "loss": 0.0705, "step": 15454 }, { "epoch": 2.83, "learning_rate": 1.4973615382676082e-05, "loss": 0.0722, "step": 15455 }, { "epoch": 2.83, "learning_rate": 1.497297127872227e-05, "loss": 0.0715, "step": 15456 }, { "epoch": 2.83, "learning_rate": 1.4972327147357323e-05, "loss": 0.0723, "step": 15457 }, { "epoch": 2.83, "learning_rate": 1.4971682988584797e-05, "loss": 0.0676, "step": 15458 }, { "epoch": 2.83, "learning_rate": 1.4971038802408233e-05, "loss": 0.0546, "step": 15459 }, { "epoch": 2.84, "learning_rate": 1.4970394588831188e-05, "loss": 0.0702, "step": 15460 }, { "epoch": 2.84, "learning_rate": 1.4969750347857214e-05, "loss": 0.0781, "step": 15461 }, { "epoch": 2.84, "learning_rate": 1.4969106079489856e-05, "loss": 0.0701, "step": 15462 }, { "epoch": 2.84, "learning_rate": 1.4968461783732673e-05, "loss": 0.0643, "step": 15463 }, { "epoch": 2.84, "learning_rate": 1.4967817460589209e-05, "loss": 0.068, "step": 15464 }, { "epoch": 2.84, "learning_rate": 1.496717311006302e-05, "loss": 0.0635, "step": 15465 }, { "epoch": 2.84, "learning_rate": 1.4966528732157658e-05, "loss": 0.0576, "step": 15466 }, { "epoch": 2.84, "learning_rate": 1.4965884326876672e-05, "loss": 0.0659, "step": 15467 }, { "epoch": 2.84, "learning_rate": 1.4965239894223613e-05, "loss": 0.072, "step": 15468 }, { "epoch": 2.84, "learning_rate": 1.4964595434202039e-05, "loss": 0.0767, "step": 15469 }, { "epoch": 2.84, "learning_rate": 1.4963950946815496e-05, "loss": 0.0682, "step": 15470 }, { "epoch": 2.84, "learning_rate": 1.496330643206754e-05, "loss": 0.0618, "step": 15471 }, { "epoch": 2.84, "learning_rate": 1.4962661889961722e-05, "loss": 0.0658, "step": 15472 }, { "epoch": 2.84, "learning_rate": 1.4962017320501599e-05, "loss": 0.0759, "step": 15473 }, { "epoch": 2.84, "learning_rate": 1.4961372723690715e-05, "loss": 0.0589, "step": 15474 }, { "epoch": 2.84, "learning_rate": 1.4960728099532631e-05, "loss": 0.0495, "step": 15475 }, { "epoch": 2.84, "learning_rate": 1.4960083448030897e-05, "loss": 0.0747, "step": 15476 }, { "epoch": 2.84, "learning_rate": 1.4959438769189066e-05, "loss": 0.0683, "step": 15477 }, { "epoch": 2.84, "learning_rate": 1.4958794063010695e-05, "loss": 0.0701, "step": 15478 }, { "epoch": 2.84, "learning_rate": 1.4958149329499332e-05, "loss": 0.0683, "step": 15479 }, { "epoch": 2.84, "learning_rate": 1.4957504568658537e-05, "loss": 0.0709, "step": 15480 }, { "epoch": 2.84, "learning_rate": 1.4956859780491857e-05, "loss": 0.0688, "step": 15481 }, { "epoch": 2.84, "learning_rate": 1.4956214965002854e-05, "loss": 0.08, "step": 15482 }, { "epoch": 2.84, "learning_rate": 1.4955570122195071e-05, "loss": 0.0685, "step": 15483 }, { "epoch": 2.84, "learning_rate": 1.4954925252072077e-05, "loss": 0.0737, "step": 15484 }, { "epoch": 2.84, "learning_rate": 1.4954280354637413e-05, "loss": 0.0685, "step": 15485 }, { "epoch": 2.84, "learning_rate": 1.4953635429894645e-05, "loss": 0.0723, "step": 15486 }, { "epoch": 2.84, "learning_rate": 1.495299047784732e-05, "loss": 0.0614, "step": 15487 }, { "epoch": 2.84, "learning_rate": 1.4952345498498995e-05, "loss": 0.0688, "step": 15488 }, { "epoch": 2.84, "learning_rate": 1.4951700491853224e-05, "loss": 0.07, "step": 15489 }, { "epoch": 2.84, "learning_rate": 1.4951055457913569e-05, "loss": 0.0715, "step": 15490 }, { "epoch": 2.84, "learning_rate": 1.4950410396683577e-05, "loss": 0.0641, "step": 15491 }, { "epoch": 2.84, "learning_rate": 1.4949765308166803e-05, "loss": 0.0843, "step": 15492 }, { "epoch": 2.84, "learning_rate": 1.4949120192366815e-05, "loss": 0.0636, "step": 15493 }, { "epoch": 2.84, "learning_rate": 1.4948475049287154e-05, "loss": 0.0595, "step": 15494 }, { "epoch": 2.84, "learning_rate": 1.4947829878931383e-05, "loss": 0.068, "step": 15495 }, { "epoch": 2.84, "learning_rate": 1.4947184681303065e-05, "loss": 0.0794, "step": 15496 }, { "epoch": 2.84, "learning_rate": 1.494653945640574e-05, "loss": 0.0644, "step": 15497 }, { "epoch": 2.84, "learning_rate": 1.4945894204242978e-05, "loss": 0.0755, "step": 15498 }, { "epoch": 2.84, "learning_rate": 1.4945248924818333e-05, "loss": 0.0769, "step": 15499 }, { "epoch": 2.84, "learning_rate": 1.4944603618135355e-05, "loss": 0.0624, "step": 15500 }, { "epoch": 2.84, "learning_rate": 1.4943958284197609e-05, "loss": 0.0621, "step": 15501 }, { "epoch": 2.84, "learning_rate": 1.4943312923008651e-05, "loss": 0.0624, "step": 15502 }, { "epoch": 2.84, "learning_rate": 1.494266753457203e-05, "loss": 0.0606, "step": 15503 }, { "epoch": 2.84, "learning_rate": 1.4942022118891317e-05, "loss": 0.0761, "step": 15504 }, { "epoch": 2.84, "learning_rate": 1.4941376675970058e-05, "loss": 0.0738, "step": 15505 }, { "epoch": 2.84, "learning_rate": 1.4940731205811819e-05, "loss": 0.0718, "step": 15506 }, { "epoch": 2.84, "learning_rate": 1.494008570842015e-05, "loss": 0.0684, "step": 15507 }, { "epoch": 2.84, "learning_rate": 1.4939440183798612e-05, "loss": 0.067, "step": 15508 }, { "epoch": 2.84, "learning_rate": 1.4938794631950769e-05, "loss": 0.0677, "step": 15509 }, { "epoch": 2.84, "learning_rate": 1.493814905288017e-05, "loss": 0.0678, "step": 15510 }, { "epoch": 2.84, "learning_rate": 1.493750344659038e-05, "loss": 0.0652, "step": 15511 }, { "epoch": 2.84, "learning_rate": 1.4936857813084953e-05, "loss": 0.0705, "step": 15512 }, { "epoch": 2.84, "learning_rate": 1.4936212152367452e-05, "loss": 0.0734, "step": 15513 }, { "epoch": 2.85, "learning_rate": 1.4935566464441431e-05, "loss": 0.0718, "step": 15514 }, { "epoch": 2.85, "learning_rate": 1.4934920749310456e-05, "loss": 0.0692, "step": 15515 }, { "epoch": 2.85, "learning_rate": 1.4934275006978081e-05, "loss": 0.0738, "step": 15516 }, { "epoch": 2.85, "learning_rate": 1.4933629237447867e-05, "loss": 0.0742, "step": 15517 }, { "epoch": 2.85, "learning_rate": 1.4932983440723371e-05, "loss": 0.0579, "step": 15518 }, { "epoch": 2.85, "learning_rate": 1.4932337616808154e-05, "loss": 0.0682, "step": 15519 }, { "epoch": 2.85, "learning_rate": 1.4931691765705782e-05, "loss": 0.0619, "step": 15520 }, { "epoch": 2.85, "learning_rate": 1.4931045887419803e-05, "loss": 0.0653, "step": 15521 }, { "epoch": 2.85, "learning_rate": 1.4930399981953787e-05, "loss": 0.0651, "step": 15522 }, { "epoch": 2.85, "learning_rate": 1.4929754049311292e-05, "loss": 0.0564, "step": 15523 }, { "epoch": 2.85, "learning_rate": 1.4929108089495874e-05, "loss": 0.0633, "step": 15524 }, { "epoch": 2.85, "learning_rate": 1.4928462102511099e-05, "loss": 0.0597, "step": 15525 }, { "epoch": 2.85, "learning_rate": 1.4927816088360526e-05, "loss": 0.0636, "step": 15526 }, { "epoch": 2.85, "learning_rate": 1.492717004704771e-05, "loss": 0.0643, "step": 15527 }, { "epoch": 2.85, "learning_rate": 1.4926523978576222e-05, "loss": 0.0645, "step": 15528 }, { "epoch": 2.85, "learning_rate": 1.492587788294962e-05, "loss": 0.067, "step": 15529 }, { "epoch": 2.85, "learning_rate": 1.4925231760171459e-05, "loss": 0.0604, "step": 15530 }, { "epoch": 2.85, "learning_rate": 1.4924585610245309e-05, "loss": 0.0644, "step": 15531 }, { "epoch": 2.85, "learning_rate": 1.4923939433174726e-05, "loss": 0.0723, "step": 15532 }, { "epoch": 2.85, "learning_rate": 1.4923293228963274e-05, "loss": 0.0791, "step": 15533 }, { "epoch": 2.85, "learning_rate": 1.4922646997614513e-05, "loss": 0.071, "step": 15534 }, { "epoch": 2.85, "learning_rate": 1.4922000739132007e-05, "loss": 0.0712, "step": 15535 }, { "epoch": 2.85, "learning_rate": 1.4921354453519317e-05, "loss": 0.0754, "step": 15536 }, { "epoch": 2.85, "learning_rate": 1.4920708140780008e-05, "loss": 0.063, "step": 15537 }, { "epoch": 2.85, "learning_rate": 1.4920061800917637e-05, "loss": 0.0618, "step": 15538 }, { "epoch": 2.85, "learning_rate": 1.4919415433935776e-05, "loss": 0.0776, "step": 15539 }, { "epoch": 2.85, "learning_rate": 1.4918769039837978e-05, "loss": 0.065, "step": 15540 }, { "epoch": 2.85, "learning_rate": 1.491812261862781e-05, "loss": 0.0596, "step": 15541 }, { "epoch": 2.85, "learning_rate": 1.4917476170308835e-05, "loss": 0.0635, "step": 15542 }, { "epoch": 2.85, "learning_rate": 1.4916829694884615e-05, "loss": 0.0677, "step": 15543 }, { "epoch": 2.85, "learning_rate": 1.4916183192358717e-05, "loss": 0.0655, "step": 15544 }, { "epoch": 2.85, "learning_rate": 1.49155366627347e-05, "loss": 0.0655, "step": 15545 }, { "epoch": 2.85, "learning_rate": 1.491489010601613e-05, "loss": 0.0777, "step": 15546 }, { "epoch": 2.85, "learning_rate": 1.4914243522206572e-05, "loss": 0.0631, "step": 15547 }, { "epoch": 2.85, "learning_rate": 1.491359691130959e-05, "loss": 0.0658, "step": 15548 }, { "epoch": 2.85, "learning_rate": 1.4912950273328744e-05, "loss": 0.0563, "step": 15549 }, { "epoch": 2.85, "learning_rate": 1.49123036082676e-05, "loss": 0.0617, "step": 15550 }, { "epoch": 2.85, "learning_rate": 1.4911656916129727e-05, "loss": 0.0672, "step": 15551 }, { "epoch": 2.85, "learning_rate": 1.4911010196918683e-05, "loss": 0.0578, "step": 15552 }, { "epoch": 2.85, "learning_rate": 1.4910363450638037e-05, "loss": 0.0766, "step": 15553 }, { "epoch": 2.85, "learning_rate": 1.490971667729135e-05, "loss": 0.0615, "step": 15554 }, { "epoch": 2.85, "learning_rate": 1.4909069876882196e-05, "loss": 0.0724, "step": 15555 }, { "epoch": 2.85, "learning_rate": 1.4908423049414127e-05, "loss": 0.0663, "step": 15556 }, { "epoch": 2.85, "learning_rate": 1.4907776194890717e-05, "loss": 0.068, "step": 15557 }, { "epoch": 2.85, "learning_rate": 1.4907129313315532e-05, "loss": 0.0641, "step": 15558 }, { "epoch": 2.85, "learning_rate": 1.4906482404692133e-05, "loss": 0.0603, "step": 15559 }, { "epoch": 2.85, "learning_rate": 1.490583546902409e-05, "loss": 0.0618, "step": 15560 }, { "epoch": 2.85, "learning_rate": 1.4905188506314967e-05, "loss": 0.06, "step": 15561 }, { "epoch": 2.85, "learning_rate": 1.490454151656833e-05, "loss": 0.0756, "step": 15562 }, { "epoch": 2.85, "learning_rate": 1.4903894499787742e-05, "loss": 0.0687, "step": 15563 }, { "epoch": 2.85, "learning_rate": 1.4903247455976777e-05, "loss": 0.0586, "step": 15564 }, { "epoch": 2.85, "learning_rate": 1.4902600385138992e-05, "loss": 0.0775, "step": 15565 }, { "epoch": 2.85, "learning_rate": 1.4901953287277962e-05, "loss": 0.0631, "step": 15566 }, { "epoch": 2.85, "learning_rate": 1.4901306162397252e-05, "loss": 0.0595, "step": 15567 }, { "epoch": 2.85, "learning_rate": 1.4900659010500424e-05, "loss": 0.0694, "step": 15568 }, { "epoch": 2.86, "learning_rate": 1.490001183159105e-05, "loss": 0.066, "step": 15569 }, { "epoch": 2.86, "learning_rate": 1.4899364625672699e-05, "loss": 0.0679, "step": 15570 }, { "epoch": 2.86, "learning_rate": 1.4898717392748934e-05, "loss": 0.0651, "step": 15571 }, { "epoch": 2.86, "learning_rate": 1.4898070132823321e-05, "loss": 0.0536, "step": 15572 }, { "epoch": 2.86, "learning_rate": 1.4897422845899433e-05, "loss": 0.0713, "step": 15573 }, { "epoch": 2.86, "learning_rate": 1.4896775531980836e-05, "loss": 0.0553, "step": 15574 }, { "epoch": 2.86, "learning_rate": 1.4896128191071095e-05, "loss": 0.0731, "step": 15575 }, { "epoch": 2.86, "learning_rate": 1.4895480823173784e-05, "loss": 0.0751, "step": 15576 }, { "epoch": 2.86, "learning_rate": 1.4894833428292468e-05, "loss": 0.0528, "step": 15577 }, { "epoch": 2.86, "learning_rate": 1.489418600643071e-05, "loss": 0.0817, "step": 15578 }, { "epoch": 2.86, "learning_rate": 1.489353855759209e-05, "loss": 0.0684, "step": 15579 }, { "epoch": 2.86, "learning_rate": 1.489289108178017e-05, "loss": 0.0723, "step": 15580 }, { "epoch": 2.86, "learning_rate": 1.4892243578998518e-05, "loss": 0.0718, "step": 15581 }, { "epoch": 2.86, "learning_rate": 1.4891596049250708e-05, "loss": 0.0742, "step": 15582 }, { "epoch": 2.86, "learning_rate": 1.48909484925403e-05, "loss": 0.0621, "step": 15583 }, { "epoch": 2.86, "learning_rate": 1.4890300908870874e-05, "loss": 0.0826, "step": 15584 }, { "epoch": 2.86, "learning_rate": 1.4889653298245993e-05, "loss": 0.0683, "step": 15585 }, { "epoch": 2.86, "learning_rate": 1.4889005660669229e-05, "loss": 0.0641, "step": 15586 }, { "epoch": 2.86, "learning_rate": 1.4888357996144152e-05, "loss": 0.0677, "step": 15587 }, { "epoch": 2.86, "learning_rate": 1.4887710304674331e-05, "loss": 0.069, "step": 15588 }, { "epoch": 2.86, "learning_rate": 1.4887062586263334e-05, "loss": 0.0682, "step": 15589 }, { "epoch": 2.86, "learning_rate": 1.4886414840914735e-05, "loss": 0.0755, "step": 15590 }, { "epoch": 2.86, "learning_rate": 1.4885767068632104e-05, "loss": 0.0749, "step": 15591 }, { "epoch": 2.86, "learning_rate": 1.488511926941901e-05, "loss": 0.0685, "step": 15592 }, { "epoch": 2.86, "learning_rate": 1.4884471443279025e-05, "loss": 0.0687, "step": 15593 }, { "epoch": 2.86, "learning_rate": 1.4883823590215718e-05, "loss": 0.0726, "step": 15594 }, { "epoch": 2.86, "learning_rate": 1.488317571023266e-05, "loss": 0.0708, "step": 15595 }, { "epoch": 2.86, "learning_rate": 1.4882527803333422e-05, "loss": 0.0773, "step": 15596 }, { "epoch": 2.86, "learning_rate": 1.4881879869521582e-05, "loss": 0.0662, "step": 15597 }, { "epoch": 2.86, "learning_rate": 1.48812319088007e-05, "loss": 0.0723, "step": 15598 }, { "epoch": 2.86, "learning_rate": 1.4880583921174356e-05, "loss": 0.0596, "step": 15599 }, { "epoch": 2.86, "learning_rate": 1.4879935906646118e-05, "loss": 0.0697, "step": 15600 }, { "epoch": 2.86, "learning_rate": 1.487928786521956e-05, "loss": 0.067, "step": 15601 }, { "epoch": 2.86, "learning_rate": 1.4878639796898253e-05, "loss": 0.0671, "step": 15602 }, { "epoch": 2.86, "learning_rate": 1.4877991701685767e-05, "loss": 0.0653, "step": 15603 }, { "epoch": 2.86, "learning_rate": 1.4877343579585679e-05, "loss": 0.065, "step": 15604 }, { "epoch": 2.86, "learning_rate": 1.4876695430601558e-05, "loss": 0.0546, "step": 15605 }, { "epoch": 2.86, "learning_rate": 1.4876047254736978e-05, "loss": 0.0731, "step": 15606 }, { "epoch": 2.86, "learning_rate": 1.487539905199551e-05, "loss": 0.0684, "step": 15607 }, { "epoch": 2.86, "learning_rate": 1.4874750822380728e-05, "loss": 0.069, "step": 15608 }, { "epoch": 2.86, "learning_rate": 1.4874102565896206e-05, "loss": 0.0662, "step": 15609 }, { "epoch": 2.86, "learning_rate": 1.4873454282545519e-05, "loss": 0.0645, "step": 15610 }, { "epoch": 2.86, "learning_rate": 1.4872805972332235e-05, "loss": 0.0614, "step": 15611 }, { "epoch": 2.86, "learning_rate": 1.4872157635259932e-05, "loss": 0.0725, "step": 15612 }, { "epoch": 2.86, "learning_rate": 1.4871509271332178e-05, "loss": 0.0552, "step": 15613 }, { "epoch": 2.86, "learning_rate": 1.4870860880552554e-05, "loss": 0.0633, "step": 15614 }, { "epoch": 2.86, "learning_rate": 1.4870212462924634e-05, "loss": 0.0711, "step": 15615 }, { "epoch": 2.86, "learning_rate": 1.4869564018451983e-05, "loss": 0.0665, "step": 15616 }, { "epoch": 2.86, "learning_rate": 1.4868915547138183e-05, "loss": 0.0609, "step": 15617 }, { "epoch": 2.86, "learning_rate": 1.4868267048986806e-05, "loss": 0.0611, "step": 15618 }, { "epoch": 2.86, "learning_rate": 1.4867618524001428e-05, "loss": 0.0653, "step": 15619 }, { "epoch": 2.86, "learning_rate": 1.4866969972185623e-05, "loss": 0.0655, "step": 15620 }, { "epoch": 2.86, "learning_rate": 1.4866321393542964e-05, "loss": 0.0642, "step": 15621 }, { "epoch": 2.86, "learning_rate": 1.4865672788077027e-05, "loss": 0.071, "step": 15622 }, { "epoch": 2.87, "learning_rate": 1.486502415579139e-05, "loss": 0.0597, "step": 15623 }, { "epoch": 2.87, "learning_rate": 1.4864375496689625e-05, "loss": 0.0636, "step": 15624 }, { "epoch": 2.87, "learning_rate": 1.4863726810775308e-05, "loss": 0.0648, "step": 15625 }, { "epoch": 2.87, "learning_rate": 1.4863078098052015e-05, "loss": 0.0658, "step": 15626 }, { "epoch": 2.87, "learning_rate": 1.486242935852332e-05, "loss": 0.071, "step": 15627 }, { "epoch": 2.87, "learning_rate": 1.4861780592192803e-05, "loss": 0.063, "step": 15628 }, { "epoch": 2.87, "learning_rate": 1.4861131799064037e-05, "loss": 0.0671, "step": 15629 }, { "epoch": 2.87, "learning_rate": 1.4860482979140599e-05, "loss": 0.0632, "step": 15630 }, { "epoch": 2.87, "learning_rate": 1.485983413242606e-05, "loss": 0.0586, "step": 15631 }, { "epoch": 2.87, "learning_rate": 1.485918525892401e-05, "loss": 0.0624, "step": 15632 }, { "epoch": 2.87, "learning_rate": 1.4858536358638009e-05, "loss": 0.0622, "step": 15633 }, { "epoch": 2.87, "learning_rate": 1.4857887431571647e-05, "loss": 0.0678, "step": 15634 }, { "epoch": 2.87, "learning_rate": 1.4857238477728493e-05, "loss": 0.0696, "step": 15635 }, { "epoch": 2.87, "learning_rate": 1.4856589497112128e-05, "loss": 0.0616, "step": 15636 }, { "epoch": 2.87, "learning_rate": 1.4855940489726128e-05, "loss": 0.0622, "step": 15637 }, { "epoch": 2.87, "learning_rate": 1.485529145557407e-05, "loss": 0.072, "step": 15638 }, { "epoch": 2.87, "learning_rate": 1.4854642394659531e-05, "loss": 0.0608, "step": 15639 }, { "epoch": 2.87, "learning_rate": 1.4853993306986088e-05, "loss": 0.0668, "step": 15640 }, { "epoch": 2.87, "learning_rate": 1.4853344192557324e-05, "loss": 0.0729, "step": 15641 }, { "epoch": 2.87, "learning_rate": 1.4852695051376811e-05, "loss": 0.0604, "step": 15642 }, { "epoch": 2.87, "learning_rate": 1.485204588344813e-05, "loss": 0.0583, "step": 15643 }, { "epoch": 2.87, "learning_rate": 1.4851396688774857e-05, "loss": 0.071, "step": 15644 }, { "epoch": 2.87, "learning_rate": 1.4850747467360571e-05, "loss": 0.0633, "step": 15645 }, { "epoch": 2.87, "learning_rate": 1.4850098219208855e-05, "loss": 0.0665, "step": 15646 }, { "epoch": 2.87, "learning_rate": 1.4849448944323282e-05, "loss": 0.0536, "step": 15647 }, { "epoch": 2.87, "learning_rate": 1.4848799642707433e-05, "loss": 0.0749, "step": 15648 }, { "epoch": 2.87, "learning_rate": 1.4848150314364884e-05, "loss": 0.0661, "step": 15649 }, { "epoch": 2.87, "learning_rate": 1.484750095929922e-05, "loss": 0.0756, "step": 15650 }, { "epoch": 2.87, "learning_rate": 1.4846851577514017e-05, "loss": 0.0623, "step": 15651 }, { "epoch": 2.87, "learning_rate": 1.4846202169012854e-05, "loss": 0.0709, "step": 15652 }, { "epoch": 2.87, "learning_rate": 1.484555273379931e-05, "loss": 0.0653, "step": 15653 }, { "epoch": 2.87, "learning_rate": 1.4844903271876968e-05, "loss": 0.0623, "step": 15654 }, { "epoch": 2.87, "learning_rate": 1.4844253783249402e-05, "loss": 0.0684, "step": 15655 }, { "epoch": 2.87, "learning_rate": 1.4843604267920201e-05, "loss": 0.0609, "step": 15656 }, { "epoch": 2.87, "learning_rate": 1.484295472589294e-05, "loss": 0.0641, "step": 15657 }, { "epoch": 2.87, "learning_rate": 1.4842305157171192e-05, "loss": 0.0745, "step": 15658 }, { "epoch": 2.87, "learning_rate": 1.4841655561758552e-05, "loss": 0.0726, "step": 15659 }, { "epoch": 2.87, "learning_rate": 1.4841005939658588e-05, "loss": 0.0694, "step": 15660 }, { "epoch": 2.87, "learning_rate": 1.484035629087489e-05, "loss": 0.0761, "step": 15661 }, { "epoch": 2.87, "learning_rate": 1.4839706615411032e-05, "loss": 0.0681, "step": 15662 }, { "epoch": 2.87, "learning_rate": 1.4839056913270598e-05, "loss": 0.0663, "step": 15663 }, { "epoch": 2.87, "learning_rate": 1.4838407184457166e-05, "loss": 0.0762, "step": 15664 }, { "epoch": 2.87, "learning_rate": 1.4837757428974327e-05, "loss": 0.0718, "step": 15665 }, { "epoch": 2.87, "learning_rate": 1.4837107646825648e-05, "loss": 0.0822, "step": 15666 }, { "epoch": 2.87, "learning_rate": 1.4836457838014725e-05, "loss": 0.0647, "step": 15667 }, { "epoch": 2.87, "learning_rate": 1.483580800254513e-05, "loss": 0.0675, "step": 15668 }, { "epoch": 2.87, "learning_rate": 1.4835158140420448e-05, "loss": 0.0626, "step": 15669 }, { "epoch": 2.87, "learning_rate": 1.4834508251644261e-05, "loss": 0.0685, "step": 15670 }, { "epoch": 2.87, "learning_rate": 1.4833858336220152e-05, "loss": 0.0571, "step": 15671 }, { "epoch": 2.87, "learning_rate": 1.4833208394151704e-05, "loss": 0.088, "step": 15672 }, { "epoch": 2.87, "learning_rate": 1.4832558425442494e-05, "loss": 0.0781, "step": 15673 }, { "epoch": 2.87, "learning_rate": 1.4831908430096113e-05, "loss": 0.0676, "step": 15674 }, { "epoch": 2.87, "learning_rate": 1.4831258408116136e-05, "loss": 0.0618, "step": 15675 }, { "epoch": 2.87, "learning_rate": 1.4830608359506152e-05, "loss": 0.06, "step": 15676 }, { "epoch": 2.87, "learning_rate": 1.4829958284269741e-05, "loss": 0.0746, "step": 15677 }, { "epoch": 2.88, "learning_rate": 1.4829308182410487e-05, "loss": 0.0633, "step": 15678 }, { "epoch": 2.88, "learning_rate": 1.4828658053931973e-05, "loss": 0.0618, "step": 15679 }, { "epoch": 2.88, "learning_rate": 1.4828007898837782e-05, "loss": 0.0597, "step": 15680 }, { "epoch": 2.88, "learning_rate": 1.4827357717131499e-05, "loss": 0.0766, "step": 15681 }, { "epoch": 2.88, "learning_rate": 1.4826707508816705e-05, "loss": 0.0559, "step": 15682 }, { "epoch": 2.88, "learning_rate": 1.4826057273896992e-05, "loss": 0.0564, "step": 15683 }, { "epoch": 2.88, "learning_rate": 1.4825407012375932e-05, "loss": 0.0705, "step": 15684 }, { "epoch": 2.88, "learning_rate": 1.4824756724257116e-05, "loss": 0.0665, "step": 15685 }, { "epoch": 2.88, "learning_rate": 1.482410640954413e-05, "loss": 0.0578, "step": 15686 }, { "epoch": 2.88, "learning_rate": 1.4823456068240558e-05, "loss": 0.0697, "step": 15687 }, { "epoch": 2.88, "learning_rate": 1.4822805700349982e-05, "loss": 0.0702, "step": 15688 }, { "epoch": 2.88, "learning_rate": 1.4822155305875986e-05, "loss": 0.076, "step": 15689 }, { "epoch": 2.88, "learning_rate": 1.482150488482216e-05, "loss": 0.0636, "step": 15690 }, { "epoch": 2.88, "learning_rate": 1.4820854437192083e-05, "loss": 0.0657, "step": 15691 }, { "epoch": 2.88, "learning_rate": 1.4820203962989346e-05, "loss": 0.0633, "step": 15692 }, { "epoch": 2.88, "learning_rate": 1.4819553462217529e-05, "loss": 0.0752, "step": 15693 }, { "epoch": 2.88, "learning_rate": 1.4818902934880222e-05, "loss": 0.0674, "step": 15694 }, { "epoch": 2.88, "learning_rate": 1.481825238098101e-05, "loss": 0.0646, "step": 15695 }, { "epoch": 2.88, "learning_rate": 1.4817601800523477e-05, "loss": 0.0648, "step": 15696 }, { "epoch": 2.88, "learning_rate": 1.481695119351121e-05, "loss": 0.0697, "step": 15697 }, { "epoch": 2.88, "learning_rate": 1.4816300559947797e-05, "loss": 0.0632, "step": 15698 }, { "epoch": 2.88, "learning_rate": 1.4815649899836819e-05, "loss": 0.0591, "step": 15699 }, { "epoch": 2.88, "learning_rate": 1.4814999213181867e-05, "loss": 0.0729, "step": 15700 }, { "epoch": 2.88, "learning_rate": 1.4814348499986531e-05, "loss": 0.0626, "step": 15701 }, { "epoch": 2.88, "learning_rate": 1.4813697760254387e-05, "loss": 0.0722, "step": 15702 }, { "epoch": 2.88, "learning_rate": 1.4813046993989031e-05, "loss": 0.0666, "step": 15703 }, { "epoch": 2.88, "learning_rate": 1.4812396201194048e-05, "loss": 0.0718, "step": 15704 }, { "epoch": 2.88, "learning_rate": 1.4811745381873021e-05, "loss": 0.0726, "step": 15705 }, { "epoch": 2.88, "learning_rate": 1.4811094536029544e-05, "loss": 0.0694, "step": 15706 }, { "epoch": 2.88, "learning_rate": 1.48104436636672e-05, "loss": 0.0699, "step": 15707 }, { "epoch": 2.88, "learning_rate": 1.4809792764789577e-05, "loss": 0.0639, "step": 15708 }, { "epoch": 2.88, "learning_rate": 1.4809141839400268e-05, "loss": 0.0797, "step": 15709 }, { "epoch": 2.88, "learning_rate": 1.4808490887502853e-05, "loss": 0.077, "step": 15710 }, { "epoch": 2.88, "learning_rate": 1.4807839909100922e-05, "loss": 0.0653, "step": 15711 }, { "epoch": 2.88, "learning_rate": 1.4807188904198066e-05, "loss": 0.0668, "step": 15712 }, { "epoch": 2.88, "learning_rate": 1.4806537872797874e-05, "loss": 0.0619, "step": 15713 }, { "epoch": 2.88, "learning_rate": 1.4805886814903932e-05, "loss": 0.0656, "step": 15714 }, { "epoch": 2.88, "learning_rate": 1.4805235730519829e-05, "loss": 0.0752, "step": 15715 }, { "epoch": 2.88, "learning_rate": 1.4804584619649153e-05, "loss": 0.0599, "step": 15716 }, { "epoch": 2.88, "learning_rate": 1.4803933482295494e-05, "loss": 0.076, "step": 15717 }, { "epoch": 2.88, "learning_rate": 1.4803282318462443e-05, "loss": 0.0716, "step": 15718 }, { "epoch": 2.88, "learning_rate": 1.4802631128153584e-05, "loss": 0.0666, "step": 15719 }, { "epoch": 2.88, "learning_rate": 1.4801979911372512e-05, "loss": 0.0595, "step": 15720 }, { "epoch": 2.88, "learning_rate": 1.4801328668122816e-05, "loss": 0.0658, "step": 15721 }, { "epoch": 2.88, "learning_rate": 1.4800677398408084e-05, "loss": 0.0705, "step": 15722 }, { "epoch": 2.88, "learning_rate": 1.4800026102231901e-05, "loss": 0.0731, "step": 15723 }, { "epoch": 2.88, "learning_rate": 1.4799374779597866e-05, "loss": 0.0651, "step": 15724 }, { "epoch": 2.88, "learning_rate": 1.4798723430509563e-05, "loss": 0.0654, "step": 15725 }, { "epoch": 2.88, "learning_rate": 1.4798072054970583e-05, "loss": 0.0648, "step": 15726 }, { "epoch": 2.88, "learning_rate": 1.479742065298452e-05, "loss": 0.0588, "step": 15727 }, { "epoch": 2.88, "learning_rate": 1.4796769224554959e-05, "loss": 0.0777, "step": 15728 }, { "epoch": 2.88, "learning_rate": 1.4796117769685497e-05, "loss": 0.0629, "step": 15729 }, { "epoch": 2.88, "learning_rate": 1.4795466288379719e-05, "loss": 0.0639, "step": 15730 }, { "epoch": 2.88, "learning_rate": 1.479481478064122e-05, "loss": 0.061, "step": 15731 }, { "epoch": 2.89, "learning_rate": 1.4794163246473588e-05, "loss": 0.0691, "step": 15732 }, { "epoch": 2.89, "learning_rate": 1.4793511685880418e-05, "loss": 0.0579, "step": 15733 }, { "epoch": 2.89, "learning_rate": 1.4792860098865299e-05, "loss": 0.0666, "step": 15734 }, { "epoch": 2.89, "learning_rate": 1.4792208485431818e-05, "loss": 0.0614, "step": 15735 }, { "epoch": 2.89, "learning_rate": 1.4791556845583576e-05, "loss": 0.0719, "step": 15736 }, { "epoch": 2.89, "learning_rate": 1.4790905179324159e-05, "loss": 0.0608, "step": 15737 }, { "epoch": 2.89, "learning_rate": 1.4790253486657162e-05, "loss": 0.0665, "step": 15738 }, { "epoch": 2.89, "learning_rate": 1.4789601767586174e-05, "loss": 0.0641, "step": 15739 }, { "epoch": 2.89, "learning_rate": 1.4788950022114792e-05, "loss": 0.069, "step": 15740 }, { "epoch": 2.89, "learning_rate": 1.47882982502466e-05, "loss": 0.0676, "step": 15741 }, { "epoch": 2.89, "learning_rate": 1.4787646451985203e-05, "loss": 0.069, "step": 15742 }, { "epoch": 2.89, "learning_rate": 1.4786994627334181e-05, "loss": 0.0719, "step": 15743 }, { "epoch": 2.89, "learning_rate": 1.4786342776297132e-05, "loss": 0.0659, "step": 15744 }, { "epoch": 2.89, "learning_rate": 1.4785690898877654e-05, "loss": 0.0582, "step": 15745 }, { "epoch": 2.89, "learning_rate": 1.478503899507933e-05, "loss": 0.0676, "step": 15746 }, { "epoch": 2.89, "learning_rate": 1.4784387064905763e-05, "loss": 0.0639, "step": 15747 }, { "epoch": 2.89, "learning_rate": 1.4783735108360542e-05, "loss": 0.0742, "step": 15748 }, { "epoch": 2.89, "learning_rate": 1.478308312544726e-05, "loss": 0.0727, "step": 15749 }, { "epoch": 2.89, "learning_rate": 1.478243111616951e-05, "loss": 0.0705, "step": 15750 }, { "epoch": 2.89, "learning_rate": 1.4781779080530894e-05, "loss": 0.0593, "step": 15751 }, { "epoch": 2.89, "learning_rate": 1.4781127018534996e-05, "loss": 0.0585, "step": 15752 }, { "epoch": 2.89, "learning_rate": 1.4780474930185408e-05, "loss": 0.0679, "step": 15753 }, { "epoch": 2.89, "learning_rate": 1.4779822815485741e-05, "loss": 0.0564, "step": 15754 }, { "epoch": 2.89, "learning_rate": 1.477917067443957e-05, "loss": 0.0615, "step": 15755 }, { "epoch": 2.89, "learning_rate": 1.4778518507050504e-05, "loss": 0.0609, "step": 15756 }, { "epoch": 2.89, "learning_rate": 1.4777866313322129e-05, "loss": 0.0708, "step": 15757 }, { "epoch": 2.89, "learning_rate": 1.4777214093258043e-05, "loss": 0.0691, "step": 15758 }, { "epoch": 2.89, "learning_rate": 1.4776561846861838e-05, "loss": 0.0676, "step": 15759 }, { "epoch": 2.89, "learning_rate": 1.4775909574137119e-05, "loss": 0.0748, "step": 15760 }, { "epoch": 2.89, "learning_rate": 1.477525727508747e-05, "loss": 0.0701, "step": 15761 }, { "epoch": 2.89, "learning_rate": 1.4774604949716492e-05, "loss": 0.0663, "step": 15762 }, { "epoch": 2.89, "learning_rate": 1.4773952598027782e-05, "loss": 0.0773, "step": 15763 }, { "epoch": 2.89, "learning_rate": 1.477330022002493e-05, "loss": 0.0666, "step": 15764 }, { "epoch": 2.89, "learning_rate": 1.4772647815711536e-05, "loss": 0.0585, "step": 15765 }, { "epoch": 2.89, "learning_rate": 1.4771995385091198e-05, "loss": 0.06, "step": 15766 }, { "epoch": 2.89, "learning_rate": 1.4771342928167508e-05, "loss": 0.0673, "step": 15767 }, { "epoch": 2.89, "learning_rate": 1.4770690444944061e-05, "loss": 0.0758, "step": 15768 }, { "epoch": 2.89, "learning_rate": 1.4770037935424461e-05, "loss": 0.0722, "step": 15769 }, { "epoch": 2.89, "learning_rate": 1.4769385399612298e-05, "loss": 0.0682, "step": 15770 }, { "epoch": 2.89, "learning_rate": 1.4768732837511171e-05, "loss": 0.0681, "step": 15771 }, { "epoch": 2.89, "learning_rate": 1.4768080249124677e-05, "loss": 0.0662, "step": 15772 }, { "epoch": 2.89, "learning_rate": 1.4767427634456413e-05, "loss": 0.0624, "step": 15773 }, { "epoch": 2.89, "learning_rate": 1.4766774993509977e-05, "loss": 0.0692, "step": 15774 }, { "epoch": 2.89, "learning_rate": 1.4766122326288966e-05, "loss": 0.064, "step": 15775 }, { "epoch": 2.89, "learning_rate": 1.4765469632796974e-05, "loss": 0.0543, "step": 15776 }, { "epoch": 2.89, "learning_rate": 1.4764816913037604e-05, "loss": 0.0623, "step": 15777 }, { "epoch": 2.89, "learning_rate": 1.4764164167014451e-05, "loss": 0.0703, "step": 15778 }, { "epoch": 2.89, "learning_rate": 1.4763511394731113e-05, "loss": 0.0663, "step": 15779 }, { "epoch": 2.89, "learning_rate": 1.476285859619119e-05, "loss": 0.0674, "step": 15780 }, { "epoch": 2.89, "learning_rate": 1.476220577139828e-05, "loss": 0.0627, "step": 15781 }, { "epoch": 2.89, "learning_rate": 1.4761552920355976e-05, "loss": 0.0669, "step": 15782 }, { "epoch": 2.89, "learning_rate": 1.4760900043067884e-05, "loss": 0.0631, "step": 15783 }, { "epoch": 2.89, "learning_rate": 1.47602471395376e-05, "loss": 0.0676, "step": 15784 }, { "epoch": 2.89, "learning_rate": 1.4759594209768721e-05, "loss": 0.0659, "step": 15785 }, { "epoch": 2.89, "learning_rate": 1.4758941253764846e-05, "loss": 0.066, "step": 15786 }, { "epoch": 2.9, "learning_rate": 1.4758288271529577e-05, "loss": 0.063, "step": 15787 }, { "epoch": 2.9, "learning_rate": 1.475763526306651e-05, "loss": 0.0646, "step": 15788 }, { "epoch": 2.9, "learning_rate": 1.4756982228379246e-05, "loss": 0.0678, "step": 15789 }, { "epoch": 2.9, "learning_rate": 1.4756329167471386e-05, "loss": 0.0564, "step": 15790 }, { "epoch": 2.9, "learning_rate": 1.4755676080346528e-05, "loss": 0.0704, "step": 15791 }, { "epoch": 2.9, "learning_rate": 1.4755022967008273e-05, "loss": 0.0749, "step": 15792 }, { "epoch": 2.9, "learning_rate": 1.4754369827460217e-05, "loss": 0.0573, "step": 15793 }, { "epoch": 2.9, "learning_rate": 1.4753716661705966e-05, "loss": 0.0776, "step": 15794 }, { "epoch": 2.9, "learning_rate": 1.4753063469749117e-05, "loss": 0.0638, "step": 15795 }, { "epoch": 2.9, "learning_rate": 1.475241025159327e-05, "loss": 0.0683, "step": 15796 }, { "epoch": 2.9, "learning_rate": 1.4751757007242024e-05, "loss": 0.0667, "step": 15797 }, { "epoch": 2.9, "learning_rate": 1.4751103736698986e-05, "loss": 0.054, "step": 15798 }, { "epoch": 2.9, "learning_rate": 1.4750450439967751e-05, "loss": 0.0564, "step": 15799 }, { "epoch": 2.9, "learning_rate": 1.4749797117051923e-05, "loss": 0.0631, "step": 15800 }, { "epoch": 2.9, "learning_rate": 1.4749143767955102e-05, "loss": 0.0666, "step": 15801 }, { "epoch": 2.9, "learning_rate": 1.4748490392680887e-05, "loss": 0.0574, "step": 15802 }, { "epoch": 2.9, "learning_rate": 1.4747836991232883e-05, "loss": 0.0721, "step": 15803 }, { "epoch": 2.9, "learning_rate": 1.4747183563614692e-05, "loss": 0.0683, "step": 15804 }, { "epoch": 2.9, "learning_rate": 1.4746530109829911e-05, "loss": 0.0729, "step": 15805 }, { "epoch": 2.9, "learning_rate": 1.4745876629882146e-05, "loss": 0.0666, "step": 15806 }, { "epoch": 2.9, "learning_rate": 1.4745223123774999e-05, "loss": 0.0651, "step": 15807 }, { "epoch": 2.9, "learning_rate": 1.474456959151207e-05, "loss": 0.0562, "step": 15808 }, { "epoch": 2.9, "learning_rate": 1.4743916033096964e-05, "loss": 0.0737, "step": 15809 }, { "epoch": 2.9, "learning_rate": 1.474326244853328e-05, "loss": 0.0716, "step": 15810 }, { "epoch": 2.9, "learning_rate": 1.4742608837824622e-05, "loss": 0.071, "step": 15811 }, { "epoch": 2.9, "learning_rate": 1.4741955200974592e-05, "loss": 0.0652, "step": 15812 }, { "epoch": 2.9, "learning_rate": 1.4741301537986797e-05, "loss": 0.0688, "step": 15813 }, { "epoch": 2.9, "learning_rate": 1.4740647848864833e-05, "loss": 0.0608, "step": 15814 }, { "epoch": 2.9, "learning_rate": 1.4739994133612308e-05, "loss": 0.0611, "step": 15815 }, { "epoch": 2.9, "learning_rate": 1.4739340392232829e-05, "loss": 0.0617, "step": 15816 }, { "epoch": 2.9, "learning_rate": 1.4738686624729987e-05, "loss": 0.0696, "step": 15817 }, { "epoch": 2.9, "learning_rate": 1.4738032831107399e-05, "loss": 0.0626, "step": 15818 }, { "epoch": 2.9, "learning_rate": 1.4737379011368663e-05, "loss": 0.0574, "step": 15819 }, { "epoch": 2.9, "learning_rate": 1.473672516551738e-05, "loss": 0.0677, "step": 15820 }, { "epoch": 2.9, "learning_rate": 1.4736071293557157e-05, "loss": 0.0785, "step": 15821 }, { "epoch": 2.9, "learning_rate": 1.47354173954916e-05, "loss": 0.0607, "step": 15822 }, { "epoch": 2.9, "learning_rate": 1.473476347132431e-05, "loss": 0.0628, "step": 15823 }, { "epoch": 2.9, "learning_rate": 1.4734109521058892e-05, "loss": 0.0704, "step": 15824 }, { "epoch": 2.9, "learning_rate": 1.4733455544698954e-05, "loss": 0.0602, "step": 15825 }, { "epoch": 2.9, "learning_rate": 1.4732801542248099e-05, "loss": 0.0639, "step": 15826 }, { "epoch": 2.9, "learning_rate": 1.4732147513709927e-05, "loss": 0.0608, "step": 15827 }, { "epoch": 2.9, "learning_rate": 1.4731493459088051e-05, "loss": 0.0565, "step": 15828 }, { "epoch": 2.9, "learning_rate": 1.473083937838607e-05, "loss": 0.0703, "step": 15829 }, { "epoch": 2.9, "learning_rate": 1.4730185271607588e-05, "loss": 0.0607, "step": 15830 }, { "epoch": 2.9, "learning_rate": 1.4729531138756221e-05, "loss": 0.0696, "step": 15831 }, { "epoch": 2.9, "learning_rate": 1.4728876979835562e-05, "loss": 0.0619, "step": 15832 }, { "epoch": 2.9, "learning_rate": 1.4728222794849227e-05, "loss": 0.067, "step": 15833 }, { "epoch": 2.9, "learning_rate": 1.4727568583800813e-05, "loss": 0.0603, "step": 15834 }, { "epoch": 2.9, "learning_rate": 1.4726914346693932e-05, "loss": 0.0578, "step": 15835 }, { "epoch": 2.9, "learning_rate": 1.4726260083532187e-05, "loss": 0.0615, "step": 15836 }, { "epoch": 2.9, "learning_rate": 1.472560579431919e-05, "loss": 0.0719, "step": 15837 }, { "epoch": 2.9, "learning_rate": 1.472495147905854e-05, "loss": 0.0667, "step": 15838 }, { "epoch": 2.9, "learning_rate": 1.4724297137753846e-05, "loss": 0.0625, "step": 15839 }, { "epoch": 2.9, "learning_rate": 1.472364277040872e-05, "loss": 0.0599, "step": 15840 }, { "epoch": 2.91, "learning_rate": 1.4722988377026757e-05, "loss": 0.0512, "step": 15841 }, { "epoch": 2.91, "learning_rate": 1.4722333957611577e-05, "loss": 0.0673, "step": 15842 }, { "epoch": 2.91, "learning_rate": 1.472167951216678e-05, "loss": 0.0652, "step": 15843 }, { "epoch": 2.91, "learning_rate": 1.4721025040695973e-05, "loss": 0.0631, "step": 15844 }, { "epoch": 2.91, "learning_rate": 1.4720370543202766e-05, "loss": 0.0638, "step": 15845 }, { "epoch": 2.91, "learning_rate": 1.4719716019690771e-05, "loss": 0.0615, "step": 15846 }, { "epoch": 2.91, "learning_rate": 1.4719061470163585e-05, "loss": 0.0489, "step": 15847 }, { "epoch": 2.91, "learning_rate": 1.471840689462482e-05, "loss": 0.0581, "step": 15848 }, { "epoch": 2.91, "learning_rate": 1.4717752293078094e-05, "loss": 0.0706, "step": 15849 }, { "epoch": 2.91, "learning_rate": 1.4717097665527e-05, "loss": 0.0636, "step": 15850 }, { "epoch": 2.91, "learning_rate": 1.4716443011975157e-05, "loss": 0.0531, "step": 15851 }, { "epoch": 2.91, "learning_rate": 1.4715788332426167e-05, "loss": 0.0627, "step": 15852 }, { "epoch": 2.91, "learning_rate": 1.4715133626883642e-05, "loss": 0.0598, "step": 15853 }, { "epoch": 2.91, "learning_rate": 1.4714478895351189e-05, "loss": 0.0727, "step": 15854 }, { "epoch": 2.91, "learning_rate": 1.471382413783242e-05, "loss": 0.0553, "step": 15855 }, { "epoch": 2.91, "learning_rate": 1.4713169354330941e-05, "loss": 0.0755, "step": 15856 }, { "epoch": 2.91, "learning_rate": 1.471251454485036e-05, "loss": 0.0676, "step": 15857 }, { "epoch": 2.91, "learning_rate": 1.4711859709394291e-05, "loss": 0.0594, "step": 15858 }, { "epoch": 2.91, "learning_rate": 1.471120484796634e-05, "loss": 0.0653, "step": 15859 }, { "epoch": 2.91, "learning_rate": 1.4710549960570118e-05, "loss": 0.0635, "step": 15860 }, { "epoch": 2.91, "learning_rate": 1.4709895047209235e-05, "loss": 0.0644, "step": 15861 }, { "epoch": 2.91, "learning_rate": 1.4709240107887299e-05, "loss": 0.0656, "step": 15862 }, { "epoch": 2.91, "learning_rate": 1.4708585142607921e-05, "loss": 0.0768, "step": 15863 }, { "epoch": 2.91, "learning_rate": 1.4707930151374712e-05, "loss": 0.0685, "step": 15864 }, { "epoch": 2.91, "learning_rate": 1.4707275134191278e-05, "loss": 0.0692, "step": 15865 }, { "epoch": 2.91, "learning_rate": 1.470662009106124e-05, "loss": 0.0625, "step": 15866 }, { "epoch": 2.91, "learning_rate": 1.4705965021988198e-05, "loss": 0.0738, "step": 15867 }, { "epoch": 2.91, "learning_rate": 1.4705309926975769e-05, "loss": 0.0584, "step": 15868 }, { "epoch": 2.91, "learning_rate": 1.4704654806027558e-05, "loss": 0.0705, "step": 15869 }, { "epoch": 2.91, "learning_rate": 1.4703999659147181e-05, "loss": 0.0682, "step": 15870 }, { "epoch": 2.91, "learning_rate": 1.4703344486338247e-05, "loss": 0.0714, "step": 15871 }, { "epoch": 2.91, "learning_rate": 1.470268928760437e-05, "loss": 0.0588, "step": 15872 }, { "epoch": 2.91, "learning_rate": 1.4702034062949158e-05, "loss": 0.0578, "step": 15873 }, { "epoch": 2.91, "learning_rate": 1.4701378812376221e-05, "loss": 0.0629, "step": 15874 }, { "epoch": 2.91, "learning_rate": 1.4700723535889178e-05, "loss": 0.0615, "step": 15875 }, { "epoch": 2.91, "learning_rate": 1.4700068233491636e-05, "loss": 0.0662, "step": 15876 }, { "epoch": 2.91, "learning_rate": 1.4699412905187207e-05, "loss": 0.065, "step": 15877 }, { "epoch": 2.91, "learning_rate": 1.4698757550979506e-05, "loss": 0.0671, "step": 15878 }, { "epoch": 2.91, "learning_rate": 1.4698102170872142e-05, "loss": 0.0618, "step": 15879 }, { "epoch": 2.91, "learning_rate": 1.4697446764868729e-05, "loss": 0.0631, "step": 15880 }, { "epoch": 2.91, "learning_rate": 1.4696791332972877e-05, "loss": 0.0693, "step": 15881 }, { "epoch": 2.91, "learning_rate": 1.4696135875188204e-05, "loss": 0.0711, "step": 15882 }, { "epoch": 2.91, "learning_rate": 1.4695480391518317e-05, "loss": 0.0645, "step": 15883 }, { "epoch": 2.91, "learning_rate": 1.4694824881966837e-05, "loss": 0.0552, "step": 15884 }, { "epoch": 2.91, "learning_rate": 1.4694169346537368e-05, "loss": 0.0581, "step": 15885 }, { "epoch": 2.91, "learning_rate": 1.4693513785233528e-05, "loss": 0.0583, "step": 15886 }, { "epoch": 2.91, "learning_rate": 1.4692858198058932e-05, "loss": 0.0616, "step": 15887 }, { "epoch": 2.91, "learning_rate": 1.4692202585017193e-05, "loss": 0.0549, "step": 15888 }, { "epoch": 2.91, "learning_rate": 1.4691546946111918e-05, "loss": 0.0617, "step": 15889 }, { "epoch": 2.91, "learning_rate": 1.469089128134673e-05, "loss": 0.0692, "step": 15890 }, { "epoch": 2.91, "learning_rate": 1.469023559072524e-05, "loss": 0.0721, "step": 15891 }, { "epoch": 2.91, "learning_rate": 1.4689579874251058e-05, "loss": 0.061, "step": 15892 }, { "epoch": 2.91, "learning_rate": 1.4688924131927806e-05, "loss": 0.0603, "step": 15893 }, { "epoch": 2.91, "learning_rate": 1.4688268363759092e-05, "loss": 0.0641, "step": 15894 }, { "epoch": 2.91, "learning_rate": 1.4687612569748536e-05, "loss": 0.0587, "step": 15895 }, { "epoch": 2.92, "learning_rate": 1.4686956749899745e-05, "loss": 0.0693, "step": 15896 }, { "epoch": 2.92, "learning_rate": 1.4686300904216342e-05, "loss": 0.0554, "step": 15897 }, { "epoch": 2.92, "learning_rate": 1.4685645032701937e-05, "loss": 0.0647, "step": 15898 }, { "epoch": 2.92, "learning_rate": 1.4684989135360148e-05, "loss": 0.0692, "step": 15899 }, { "epoch": 2.92, "learning_rate": 1.4684333212194589e-05, "loss": 0.0712, "step": 15900 }, { "epoch": 2.92, "learning_rate": 1.4683677263208873e-05, "loss": 0.0692, "step": 15901 }, { "epoch": 2.92, "learning_rate": 1.4683021288406623e-05, "loss": 0.0673, "step": 15902 }, { "epoch": 2.92, "learning_rate": 1.4682365287791444e-05, "loss": 0.0722, "step": 15903 }, { "epoch": 2.92, "learning_rate": 1.4681709261366963e-05, "loss": 0.0618, "step": 15904 }, { "epoch": 2.92, "learning_rate": 1.4681053209136789e-05, "loss": 0.0504, "step": 15905 }, { "epoch": 2.92, "learning_rate": 1.468039713110454e-05, "loss": 0.0688, "step": 15906 }, { "epoch": 2.92, "learning_rate": 1.467974102727383e-05, "loss": 0.0651, "step": 15907 }, { "epoch": 2.92, "learning_rate": 1.4679084897648283e-05, "loss": 0.0712, "step": 15908 }, { "epoch": 2.92, "learning_rate": 1.4678428742231505e-05, "loss": 0.0683, "step": 15909 }, { "epoch": 2.92, "learning_rate": 1.4677772561027121e-05, "loss": 0.0675, "step": 15910 }, { "epoch": 2.92, "learning_rate": 1.4677116354038744e-05, "loss": 0.0622, "step": 15911 }, { "epoch": 2.92, "learning_rate": 1.4676460121269993e-05, "loss": 0.065, "step": 15912 }, { "epoch": 2.92, "learning_rate": 1.4675803862724484e-05, "loss": 0.0681, "step": 15913 }, { "epoch": 2.92, "learning_rate": 1.4675147578405838e-05, "loss": 0.0736, "step": 15914 }, { "epoch": 2.92, "learning_rate": 1.4674491268317663e-05, "loss": 0.0571, "step": 15915 }, { "epoch": 2.92, "learning_rate": 1.4673834932463584e-05, "loss": 0.0642, "step": 15916 }, { "epoch": 2.92, "learning_rate": 1.4673178570847221e-05, "loss": 0.0631, "step": 15917 }, { "epoch": 2.92, "learning_rate": 1.4672522183472184e-05, "loss": 0.0624, "step": 15918 }, { "epoch": 2.92, "learning_rate": 1.4671865770342098e-05, "loss": 0.0527, "step": 15919 }, { "epoch": 2.92, "learning_rate": 1.4671209331460578e-05, "loss": 0.0569, "step": 15920 }, { "epoch": 2.92, "learning_rate": 1.4670552866831243e-05, "loss": 0.0713, "step": 15921 }, { "epoch": 2.92, "learning_rate": 1.4669896376457709e-05, "loss": 0.0647, "step": 15922 }, { "epoch": 2.92, "learning_rate": 1.4669239860343601e-05, "loss": 0.0693, "step": 15923 }, { "epoch": 2.92, "learning_rate": 1.4668583318492529e-05, "loss": 0.0646, "step": 15924 }, { "epoch": 2.92, "learning_rate": 1.4667926750908116e-05, "loss": 0.0656, "step": 15925 }, { "epoch": 2.92, "learning_rate": 1.4667270157593987e-05, "loss": 0.0512, "step": 15926 }, { "epoch": 2.92, "learning_rate": 1.4666613538553751e-05, "loss": 0.0702, "step": 15927 }, { "epoch": 2.92, "learning_rate": 1.466595689379103e-05, "loss": 0.0687, "step": 15928 }, { "epoch": 2.92, "learning_rate": 1.466530022330945e-05, "loss": 0.0671, "step": 15929 }, { "epoch": 2.92, "learning_rate": 1.4664643527112626e-05, "loss": 0.0635, "step": 15930 }, { "epoch": 2.92, "learning_rate": 1.4663986805204174e-05, "loss": 0.0657, "step": 15931 }, { "epoch": 2.92, "learning_rate": 1.466333005758772e-05, "loss": 0.0676, "step": 15932 }, { "epoch": 2.92, "learning_rate": 1.466267328426688e-05, "loss": 0.0605, "step": 15933 }, { "epoch": 2.92, "learning_rate": 1.4662016485245274e-05, "loss": 0.0653, "step": 15934 }, { "epoch": 2.92, "learning_rate": 1.466135966052653e-05, "loss": 0.0656, "step": 15935 }, { "epoch": 2.92, "learning_rate": 1.4660702810114256e-05, "loss": 0.0745, "step": 15936 }, { "epoch": 2.92, "learning_rate": 1.4660045934012082e-05, "loss": 0.059, "step": 15937 }, { "epoch": 2.92, "learning_rate": 1.4659389032223626e-05, "loss": 0.0632, "step": 15938 }, { "epoch": 2.92, "learning_rate": 1.4658732104752507e-05, "loss": 0.0587, "step": 15939 }, { "epoch": 2.92, "learning_rate": 1.465807515160235e-05, "loss": 0.0741, "step": 15940 }, { "epoch": 2.92, "learning_rate": 1.465741817277677e-05, "loss": 0.0651, "step": 15941 }, { "epoch": 2.92, "learning_rate": 1.4656761168279392e-05, "loss": 0.0614, "step": 15942 }, { "epoch": 2.92, "learning_rate": 1.4656104138113838e-05, "loss": 0.064, "step": 15943 }, { "epoch": 2.92, "learning_rate": 1.4655447082283734e-05, "loss": 0.071, "step": 15944 }, { "epoch": 2.92, "learning_rate": 1.4654790000792689e-05, "loss": 0.0708, "step": 15945 }, { "epoch": 2.92, "learning_rate": 1.4654132893644337e-05, "loss": 0.0651, "step": 15946 }, { "epoch": 2.92, "learning_rate": 1.4653475760842294e-05, "loss": 0.0811, "step": 15947 }, { "epoch": 2.92, "learning_rate": 1.4652818602390185e-05, "loss": 0.054, "step": 15948 }, { "epoch": 2.92, "learning_rate": 1.465216141829163e-05, "loss": 0.0644, "step": 15949 }, { "epoch": 2.92, "learning_rate": 1.4651504208550251e-05, "loss": 0.0665, "step": 15950 }, { "epoch": 2.93, "learning_rate": 1.4650846973169673e-05, "loss": 0.0622, "step": 15951 }, { "epoch": 2.93, "learning_rate": 1.4650189712153517e-05, "loss": 0.0737, "step": 15952 }, { "epoch": 2.93, "learning_rate": 1.4649532425505405e-05, "loss": 0.0693, "step": 15953 }, { "epoch": 2.93, "learning_rate": 1.4648875113228967e-05, "loss": 0.0652, "step": 15954 }, { "epoch": 2.93, "learning_rate": 1.4648217775327815e-05, "loss": 0.0754, "step": 15955 }, { "epoch": 2.93, "learning_rate": 1.464756041180558e-05, "loss": 0.0635, "step": 15956 }, { "epoch": 2.93, "learning_rate": 1.4646903022665884e-05, "loss": 0.0685, "step": 15957 }, { "epoch": 2.93, "learning_rate": 1.4646245607912346e-05, "loss": 0.068, "step": 15958 }, { "epoch": 2.93, "learning_rate": 1.4645588167548597e-05, "loss": 0.0684, "step": 15959 }, { "epoch": 2.93, "learning_rate": 1.4644930701578252e-05, "loss": 0.0703, "step": 15960 }, { "epoch": 2.93, "learning_rate": 1.4644273210004944e-05, "loss": 0.0615, "step": 15961 }, { "epoch": 2.93, "learning_rate": 1.4643615692832294e-05, "loss": 0.0699, "step": 15962 }, { "epoch": 2.93, "learning_rate": 1.4642958150063927e-05, "loss": 0.069, "step": 15963 }, { "epoch": 2.93, "learning_rate": 1.4642300581703462e-05, "loss": 0.0727, "step": 15964 }, { "epoch": 2.93, "learning_rate": 1.464164298775453e-05, "loss": 0.0653, "step": 15965 }, { "epoch": 2.93, "learning_rate": 1.4640985368220753e-05, "loss": 0.0715, "step": 15966 }, { "epoch": 2.93, "learning_rate": 1.4640327723105754e-05, "loss": 0.0781, "step": 15967 }, { "epoch": 2.93, "learning_rate": 1.4639670052413161e-05, "loss": 0.0758, "step": 15968 }, { "epoch": 2.93, "learning_rate": 1.4639012356146596e-05, "loss": 0.0691, "step": 15969 }, { "epoch": 2.93, "learning_rate": 1.4638354634309689e-05, "loss": 0.0635, "step": 15970 }, { "epoch": 2.93, "learning_rate": 1.463769688690606e-05, "loss": 0.0719, "step": 15971 }, { "epoch": 2.93, "learning_rate": 1.4637039113939339e-05, "loss": 0.0648, "step": 15972 }, { "epoch": 2.93, "learning_rate": 1.463638131541315e-05, "loss": 0.0752, "step": 15973 }, { "epoch": 2.93, "learning_rate": 1.463572349133112e-05, "loss": 0.0695, "step": 15974 }, { "epoch": 2.93, "learning_rate": 1.4635065641696868e-05, "loss": 0.0605, "step": 15975 }, { "epoch": 2.93, "learning_rate": 1.4634407766514033e-05, "loss": 0.0586, "step": 15976 }, { "epoch": 2.93, "learning_rate": 1.463374986578623e-05, "loss": 0.0621, "step": 15977 }, { "epoch": 2.93, "learning_rate": 1.4633091939517088e-05, "loss": 0.0764, "step": 15978 }, { "epoch": 2.93, "learning_rate": 1.4632433987710237e-05, "loss": 0.0629, "step": 15979 }, { "epoch": 2.93, "learning_rate": 1.4631776010369302e-05, "loss": 0.0692, "step": 15980 }, { "epoch": 2.93, "learning_rate": 1.4631118007497904e-05, "loss": 0.0671, "step": 15981 }, { "epoch": 2.93, "learning_rate": 1.4630459979099678e-05, "loss": 0.0683, "step": 15982 }, { "epoch": 2.93, "learning_rate": 1.4629801925178252e-05, "loss": 0.0748, "step": 15983 }, { "epoch": 2.93, "learning_rate": 1.4629143845737242e-05, "loss": 0.0572, "step": 15984 }, { "epoch": 2.93, "learning_rate": 1.4628485740780289e-05, "loss": 0.0619, "step": 15985 }, { "epoch": 2.93, "learning_rate": 1.462782761031101e-05, "loss": 0.0678, "step": 15986 }, { "epoch": 2.93, "learning_rate": 1.462716945433304e-05, "loss": 0.0674, "step": 15987 }, { "epoch": 2.93, "learning_rate": 1.4626511272850003e-05, "loss": 0.0647, "step": 15988 }, { "epoch": 2.93, "learning_rate": 1.4625853065865526e-05, "loss": 0.0574, "step": 15989 }, { "epoch": 2.93, "learning_rate": 1.4625194833383239e-05, "loss": 0.0581, "step": 15990 }, { "epoch": 2.93, "learning_rate": 1.462453657540677e-05, "loss": 0.0642, "step": 15991 }, { "epoch": 2.93, "learning_rate": 1.4623878291939747e-05, "loss": 0.0629, "step": 15992 }, { "epoch": 2.93, "learning_rate": 1.4623219982985797e-05, "loss": 0.0623, "step": 15993 }, { "epoch": 2.93, "learning_rate": 1.4622561648548551e-05, "loss": 0.0689, "step": 15994 }, { "epoch": 2.93, "learning_rate": 1.4621903288631636e-05, "loss": 0.0659, "step": 15995 }, { "epoch": 2.93, "learning_rate": 1.4621244903238684e-05, "loss": 0.0523, "step": 15996 }, { "epoch": 2.93, "learning_rate": 1.462058649237332e-05, "loss": 0.0737, "step": 15997 }, { "epoch": 2.93, "learning_rate": 1.4619928056039178e-05, "loss": 0.0451, "step": 15998 }, { "epoch": 2.93, "learning_rate": 1.4619269594239882e-05, "loss": 0.0671, "step": 15999 }, { "epoch": 2.93, "learning_rate": 1.4618611106979063e-05, "loss": 0.0768, "step": 16000 }, { "epoch": 2.93, "learning_rate": 1.4617952594260354e-05, "loss": 0.0642, "step": 16001 }, { "epoch": 2.93, "learning_rate": 1.461729405608738e-05, "loss": 0.0635, "step": 16002 }, { "epoch": 2.93, "learning_rate": 1.4616635492463775e-05, "loss": 0.0539, "step": 16003 }, { "epoch": 2.93, "learning_rate": 1.4615976903393162e-05, "loss": 0.0699, "step": 16004 }, { "epoch": 2.94, "learning_rate": 1.4615318288879184e-05, "loss": 0.0619, "step": 16005 }, { "epoch": 2.94, "learning_rate": 1.4614659648925458e-05, "loss": 0.0657, "step": 16006 }, { "epoch": 2.94, "learning_rate": 1.4614000983535621e-05, "loss": 0.0758, "step": 16007 }, { "epoch": 2.94, "learning_rate": 1.4613342292713303e-05, "loss": 0.0615, "step": 16008 }, { "epoch": 2.94, "learning_rate": 1.4612683576462135e-05, "loss": 0.0613, "step": 16009 }, { "epoch": 2.94, "learning_rate": 1.4612024834785746e-05, "loss": 0.0613, "step": 16010 }, { "epoch": 2.94, "learning_rate": 1.4611366067687764e-05, "loss": 0.0631, "step": 16011 }, { "epoch": 2.94, "learning_rate": 1.461070727517183e-05, "loss": 0.0716, "step": 16012 }, { "epoch": 2.94, "learning_rate": 1.4610048457241569e-05, "loss": 0.0628, "step": 16013 }, { "epoch": 2.94, "learning_rate": 1.4609389613900611e-05, "loss": 0.0655, "step": 16014 }, { "epoch": 2.94, "learning_rate": 1.4608730745152588e-05, "loss": 0.0681, "step": 16015 }, { "epoch": 2.94, "learning_rate": 1.4608071851001135e-05, "loss": 0.0708, "step": 16016 }, { "epoch": 2.94, "learning_rate": 1.4607412931449885e-05, "loss": 0.0593, "step": 16017 }, { "epoch": 2.94, "learning_rate": 1.4606753986502464e-05, "loss": 0.0539, "step": 16018 }, { "epoch": 2.94, "learning_rate": 1.4606095016162507e-05, "loss": 0.0599, "step": 16019 }, { "epoch": 2.94, "learning_rate": 1.4605436020433646e-05, "loss": 0.0638, "step": 16020 }, { "epoch": 2.94, "learning_rate": 1.4604776999319517e-05, "loss": 0.0606, "step": 16021 }, { "epoch": 2.94, "learning_rate": 1.4604117952823745e-05, "loss": 0.0666, "step": 16022 }, { "epoch": 2.94, "learning_rate": 1.4603458880949968e-05, "loss": 0.0672, "step": 16023 }, { "epoch": 2.94, "learning_rate": 1.4602799783701818e-05, "loss": 0.0732, "step": 16024 }, { "epoch": 2.94, "learning_rate": 1.4602140661082928e-05, "loss": 0.0757, "step": 16025 }, { "epoch": 2.94, "learning_rate": 1.4601481513096932e-05, "loss": 0.0604, "step": 16026 }, { "epoch": 2.94, "learning_rate": 1.460082233974746e-05, "loss": 0.0608, "step": 16027 }, { "epoch": 2.94, "learning_rate": 1.4600163141038148e-05, "loss": 0.0717, "step": 16028 }, { "epoch": 2.94, "learning_rate": 1.459950391697263e-05, "loss": 0.064, "step": 16029 }, { "epoch": 2.94, "learning_rate": 1.4598844667554542e-05, "loss": 0.0591, "step": 16030 }, { "epoch": 2.94, "learning_rate": 1.4598185392787507e-05, "loss": 0.068, "step": 16031 }, { "epoch": 2.94, "learning_rate": 1.4597526092675169e-05, "loss": 0.0612, "step": 16032 }, { "epoch": 2.94, "learning_rate": 1.4596866767221162e-05, "loss": 0.0666, "step": 16033 }, { "epoch": 2.94, "learning_rate": 1.4596207416429119e-05, "loss": 0.0663, "step": 16034 }, { "epoch": 2.94, "learning_rate": 1.4595548040302669e-05, "loss": 0.067, "step": 16035 }, { "epoch": 2.94, "learning_rate": 1.4594888638845453e-05, "loss": 0.064, "step": 16036 }, { "epoch": 2.94, "learning_rate": 1.45942292120611e-05, "loss": 0.0666, "step": 16037 }, { "epoch": 2.94, "learning_rate": 1.459356975995325e-05, "loss": 0.0669, "step": 16038 }, { "epoch": 2.94, "learning_rate": 1.4592910282525537e-05, "loss": 0.0691, "step": 16039 }, { "epoch": 2.94, "learning_rate": 1.4592250779781591e-05, "loss": 0.066, "step": 16040 }, { "epoch": 2.94, "learning_rate": 1.4591591251725054e-05, "loss": 0.0727, "step": 16041 }, { "epoch": 2.94, "learning_rate": 1.459093169835956e-05, "loss": 0.0722, "step": 16042 }, { "epoch": 2.94, "learning_rate": 1.4590272119688742e-05, "loss": 0.0611, "step": 16043 }, { "epoch": 2.94, "learning_rate": 1.4589612515716236e-05, "loss": 0.0666, "step": 16044 }, { "epoch": 2.94, "learning_rate": 1.4588952886445677e-05, "loss": 0.0703, "step": 16045 }, { "epoch": 2.94, "learning_rate": 1.4588293231880704e-05, "loss": 0.0694, "step": 16046 }, { "epoch": 2.94, "learning_rate": 1.4587633552024949e-05, "loss": 0.0645, "step": 16047 }, { "epoch": 2.94, "learning_rate": 1.4586973846882054e-05, "loss": 0.0676, "step": 16048 }, { "epoch": 2.94, "learning_rate": 1.4586314116455649e-05, "loss": 0.0655, "step": 16049 }, { "epoch": 2.94, "learning_rate": 1.4585654360749373e-05, "loss": 0.0522, "step": 16050 }, { "epoch": 2.94, "learning_rate": 1.4584994579766865e-05, "loss": 0.0564, "step": 16051 }, { "epoch": 2.94, "learning_rate": 1.4584334773511756e-05, "loss": 0.0556, "step": 16052 }, { "epoch": 2.94, "learning_rate": 1.4583674941987688e-05, "loss": 0.0657, "step": 16053 }, { "epoch": 2.94, "learning_rate": 1.4583015085198298e-05, "loss": 0.0621, "step": 16054 }, { "epoch": 2.94, "learning_rate": 1.4582355203147218e-05, "loss": 0.0634, "step": 16055 }, { "epoch": 2.94, "learning_rate": 1.4581695295838094e-05, "loss": 0.0559, "step": 16056 }, { "epoch": 2.94, "learning_rate": 1.4581035363274554e-05, "loss": 0.0693, "step": 16057 }, { "epoch": 2.94, "learning_rate": 1.458037540546024e-05, "loss": 0.0604, "step": 16058 }, { "epoch": 2.94, "learning_rate": 1.457971542239879e-05, "loss": 0.0691, "step": 16059 }, { "epoch": 2.95, "learning_rate": 1.4579055414093843e-05, "loss": 0.0607, "step": 16060 }, { "epoch": 2.95, "learning_rate": 1.457839538054903e-05, "loss": 0.0599, "step": 16061 }, { "epoch": 2.95, "learning_rate": 1.4577735321768001e-05, "loss": 0.0628, "step": 16062 }, { "epoch": 2.95, "learning_rate": 1.4577075237754386e-05, "loss": 0.0716, "step": 16063 }, { "epoch": 2.95, "learning_rate": 1.4576415128511823e-05, "loss": 0.071, "step": 16064 }, { "epoch": 2.95, "learning_rate": 1.4575754994043956e-05, "loss": 0.0753, "step": 16065 }, { "epoch": 2.95, "learning_rate": 1.4575094834354415e-05, "loss": 0.0683, "step": 16066 }, { "epoch": 2.95, "learning_rate": 1.4574434649446848e-05, "loss": 0.0695, "step": 16067 }, { "epoch": 2.95, "learning_rate": 1.4573774439324888e-05, "loss": 0.0781, "step": 16068 }, { "epoch": 2.95, "learning_rate": 1.4573114203992177e-05, "loss": 0.0553, "step": 16069 }, { "epoch": 2.95, "learning_rate": 1.4572453943452352e-05, "loss": 0.0714, "step": 16070 }, { "epoch": 2.95, "learning_rate": 1.4571793657709057e-05, "loss": 0.0708, "step": 16071 }, { "epoch": 2.95, "learning_rate": 1.457113334676593e-05, "loss": 0.0694, "step": 16072 }, { "epoch": 2.95, "learning_rate": 1.4570473010626601e-05, "loss": 0.067, "step": 16073 }, { "epoch": 2.95, "learning_rate": 1.4569812649294725e-05, "loss": 0.0603, "step": 16074 }, { "epoch": 2.95, "learning_rate": 1.456915226277393e-05, "loss": 0.0615, "step": 16075 }, { "epoch": 2.95, "learning_rate": 1.4568491851067861e-05, "loss": 0.0686, "step": 16076 }, { "epoch": 2.95, "learning_rate": 1.4567831414180161e-05, "loss": 0.067, "step": 16077 }, { "epoch": 2.95, "learning_rate": 1.4567170952114465e-05, "loss": 0.0671, "step": 16078 }, { "epoch": 2.95, "learning_rate": 1.4566510464874415e-05, "loss": 0.0525, "step": 16079 }, { "epoch": 2.95, "learning_rate": 1.4565849952463656e-05, "loss": 0.0677, "step": 16080 }, { "epoch": 2.95, "learning_rate": 1.456518941488582e-05, "loss": 0.0669, "step": 16081 }, { "epoch": 2.95, "learning_rate": 1.4564528852144556e-05, "loss": 0.0638, "step": 16082 }, { "epoch": 2.95, "learning_rate": 1.4563868264243504e-05, "loss": 0.0643, "step": 16083 }, { "epoch": 2.95, "learning_rate": 1.45632076511863e-05, "loss": 0.07, "step": 16084 }, { "epoch": 2.95, "learning_rate": 1.4562547012976589e-05, "loss": 0.0636, "step": 16085 }, { "epoch": 2.95, "learning_rate": 1.4561886349618013e-05, "loss": 0.0683, "step": 16086 }, { "epoch": 2.95, "learning_rate": 1.4561225661114212e-05, "loss": 0.0646, "step": 16087 }, { "epoch": 2.95, "learning_rate": 1.4560564947468824e-05, "loss": 0.0519, "step": 16088 }, { "epoch": 2.95, "learning_rate": 1.4559904208685503e-05, "loss": 0.0579, "step": 16089 }, { "epoch": 2.95, "learning_rate": 1.4559243444767878e-05, "loss": 0.0592, "step": 16090 }, { "epoch": 2.95, "learning_rate": 1.4558582655719598e-05, "loss": 0.0566, "step": 16091 }, { "epoch": 2.95, "learning_rate": 1.4557921841544305e-05, "loss": 0.0665, "step": 16092 }, { "epoch": 2.95, "learning_rate": 1.4557261002245637e-05, "loss": 0.075, "step": 16093 }, { "epoch": 2.95, "learning_rate": 1.4556600137827242e-05, "loss": 0.0578, "step": 16094 }, { "epoch": 2.95, "learning_rate": 1.4555939248292759e-05, "loss": 0.0649, "step": 16095 }, { "epoch": 2.95, "learning_rate": 1.4555278333645833e-05, "loss": 0.0606, "step": 16096 }, { "epoch": 2.95, "learning_rate": 1.4554617393890105e-05, "loss": 0.0761, "step": 16097 }, { "epoch": 2.95, "learning_rate": 1.4553956429029221e-05, "loss": 0.0659, "step": 16098 }, { "epoch": 2.95, "learning_rate": 1.455329543906682e-05, "loss": 0.0683, "step": 16099 }, { "epoch": 2.95, "learning_rate": 1.4552634424006548e-05, "loss": 0.0721, "step": 16100 }, { "epoch": 2.95, "learning_rate": 1.4551973383852052e-05, "loss": 0.0711, "step": 16101 }, { "epoch": 2.95, "learning_rate": 1.455131231860697e-05, "loss": 0.0576, "step": 16102 }, { "epoch": 2.95, "learning_rate": 1.4550651228274948e-05, "loss": 0.0625, "step": 16103 }, { "epoch": 2.95, "learning_rate": 1.454999011285963e-05, "loss": 0.056, "step": 16104 }, { "epoch": 2.95, "learning_rate": 1.4549328972364658e-05, "loss": 0.0611, "step": 16105 }, { "epoch": 2.95, "learning_rate": 1.4548667806793676e-05, "loss": 0.0648, "step": 16106 }, { "epoch": 2.95, "learning_rate": 1.4548006616150338e-05, "loss": 0.0659, "step": 16107 }, { "epoch": 2.95, "learning_rate": 1.4547345400438275e-05, "loss": 0.0599, "step": 16108 }, { "epoch": 2.95, "learning_rate": 1.454668415966114e-05, "loss": 0.0654, "step": 16109 }, { "epoch": 2.95, "learning_rate": 1.4546022893822574e-05, "loss": 0.0678, "step": 16110 }, { "epoch": 2.95, "learning_rate": 1.4545361602926225e-05, "loss": 0.0577, "step": 16111 }, { "epoch": 2.95, "learning_rate": 1.4544700286975735e-05, "loss": 0.0674, "step": 16112 }, { "epoch": 2.95, "learning_rate": 1.4544038945974753e-05, "loss": 0.0702, "step": 16113 }, { "epoch": 2.96, "learning_rate": 1.4543377579926915e-05, "loss": 0.072, "step": 16114 }, { "epoch": 2.96, "learning_rate": 1.4542716188835881e-05, "loss": 0.0825, "step": 16115 }, { "epoch": 2.96, "learning_rate": 1.454205477270529e-05, "loss": 0.06, "step": 16116 }, { "epoch": 2.96, "learning_rate": 1.454139333153878e-05, "loss": 0.064, "step": 16117 }, { "epoch": 2.96, "learning_rate": 1.4540731865340005e-05, "loss": 0.0633, "step": 16118 }, { "epoch": 2.96, "learning_rate": 1.454007037411261e-05, "loss": 0.0654, "step": 16119 }, { "epoch": 2.96, "learning_rate": 1.4539408857860244e-05, "loss": 0.0636, "step": 16120 }, { "epoch": 2.96, "learning_rate": 1.453874731658655e-05, "loss": 0.0616, "step": 16121 }, { "epoch": 2.96, "learning_rate": 1.453808575029517e-05, "loss": 0.0636, "step": 16122 }, { "epoch": 2.96, "learning_rate": 1.4537424158989758e-05, "loss": 0.0721, "step": 16123 }, { "epoch": 2.96, "learning_rate": 1.4536762542673957e-05, "loss": 0.0686, "step": 16124 }, { "epoch": 2.96, "learning_rate": 1.4536100901351413e-05, "loss": 0.0554, "step": 16125 }, { "epoch": 2.96, "learning_rate": 1.4535439235025776e-05, "loss": 0.0623, "step": 16126 }, { "epoch": 2.96, "learning_rate": 1.4534777543700693e-05, "loss": 0.0556, "step": 16127 }, { "epoch": 2.96, "learning_rate": 1.4534115827379809e-05, "loss": 0.065, "step": 16128 }, { "epoch": 2.96, "learning_rate": 1.4533454086066772e-05, "loss": 0.0613, "step": 16129 }, { "epoch": 2.96, "learning_rate": 1.4532792319765231e-05, "loss": 0.0769, "step": 16130 }, { "epoch": 2.96, "learning_rate": 1.4532130528478831e-05, "loss": 0.0569, "step": 16131 }, { "epoch": 2.96, "learning_rate": 1.453146871221122e-05, "loss": 0.0627, "step": 16132 }, { "epoch": 2.96, "learning_rate": 1.4530806870966053e-05, "loss": 0.0794, "step": 16133 }, { "epoch": 2.96, "learning_rate": 1.4530145004746966e-05, "loss": 0.0603, "step": 16134 }, { "epoch": 2.96, "learning_rate": 1.4529483113557617e-05, "loss": 0.0635, "step": 16135 }, { "epoch": 2.96, "learning_rate": 1.452882119740165e-05, "loss": 0.0667, "step": 16136 }, { "epoch": 2.96, "learning_rate": 1.4528159256282714e-05, "loss": 0.0637, "step": 16137 }, { "epoch": 2.96, "learning_rate": 1.452749729020446e-05, "loss": 0.0644, "step": 16138 }, { "epoch": 2.96, "learning_rate": 1.4526835299170535e-05, "loss": 0.0742, "step": 16139 }, { "epoch": 2.96, "learning_rate": 1.4526173283184586e-05, "loss": 0.0618, "step": 16140 }, { "epoch": 2.96, "learning_rate": 1.4525511242250261e-05, "loss": 0.0628, "step": 16141 }, { "epoch": 2.96, "learning_rate": 1.4524849176371219e-05, "loss": 0.0755, "step": 16142 }, { "epoch": 2.96, "learning_rate": 1.4524187085551095e-05, "loss": 0.0606, "step": 16143 }, { "epoch": 2.96, "learning_rate": 1.452352496979355e-05, "loss": 0.0617, "step": 16144 }, { "epoch": 2.96, "learning_rate": 1.452286282910223e-05, "loss": 0.0685, "step": 16145 }, { "epoch": 2.96, "learning_rate": 1.4522200663480784e-05, "loss": 0.0556, "step": 16146 }, { "epoch": 2.96, "learning_rate": 1.4521538472932859e-05, "loss": 0.066, "step": 16147 }, { "epoch": 2.96, "learning_rate": 1.4520876257462111e-05, "loss": 0.0544, "step": 16148 }, { "epoch": 2.96, "learning_rate": 1.4520214017072185e-05, "loss": 0.0668, "step": 16149 }, { "epoch": 2.96, "learning_rate": 1.4519551751766733e-05, "loss": 0.0703, "step": 16150 }, { "epoch": 2.96, "learning_rate": 1.451888946154941e-05, "loss": 0.0612, "step": 16151 }, { "epoch": 2.96, "learning_rate": 1.4518227146423858e-05, "loss": 0.07, "step": 16152 }, { "epoch": 2.96, "learning_rate": 1.4517564806393734e-05, "loss": 0.0637, "step": 16153 }, { "epoch": 2.96, "learning_rate": 1.4516902441462688e-05, "loss": 0.0606, "step": 16154 }, { "epoch": 2.96, "learning_rate": 1.4516240051634367e-05, "loss": 0.0702, "step": 16155 }, { "epoch": 2.96, "learning_rate": 1.4515577636912426e-05, "loss": 0.0676, "step": 16156 }, { "epoch": 2.96, "learning_rate": 1.451491519730052e-05, "loss": 0.0683, "step": 16157 }, { "epoch": 2.96, "learning_rate": 1.451425273280229e-05, "loss": 0.0665, "step": 16158 }, { "epoch": 2.96, "learning_rate": 1.4513590243421394e-05, "loss": 0.0665, "step": 16159 }, { "epoch": 2.96, "learning_rate": 1.4512927729161486e-05, "loss": 0.0579, "step": 16160 }, { "epoch": 2.96, "learning_rate": 1.4512265190026212e-05, "loss": 0.0677, "step": 16161 }, { "epoch": 2.96, "learning_rate": 1.4511602626019226e-05, "loss": 0.07, "step": 16162 }, { "epoch": 2.96, "learning_rate": 1.4510940037144182e-05, "loss": 0.0615, "step": 16163 }, { "epoch": 2.96, "learning_rate": 1.451027742340473e-05, "loss": 0.0629, "step": 16164 }, { "epoch": 2.96, "learning_rate": 1.4509614784804522e-05, "loss": 0.0586, "step": 16165 }, { "epoch": 2.96, "learning_rate": 1.4508952121347216e-05, "loss": 0.065, "step": 16166 }, { "epoch": 2.96, "learning_rate": 1.4508289433036456e-05, "loss": 0.056, "step": 16167 }, { "epoch": 2.96, "learning_rate": 1.4507626719875898e-05, "loss": 0.0713, "step": 16168 }, { "epoch": 2.97, "learning_rate": 1.4506963981869202e-05, "loss": 0.062, "step": 16169 }, { "epoch": 2.97, "learning_rate": 1.450630121902001e-05, "loss": 0.0595, "step": 16170 }, { "epoch": 2.97, "learning_rate": 1.4505638431331982e-05, "loss": 0.0684, "step": 16171 }, { "epoch": 2.97, "learning_rate": 1.4504975618808767e-05, "loss": 0.0548, "step": 16172 }, { "epoch": 2.97, "learning_rate": 1.4504312781454024e-05, "loss": 0.068, "step": 16173 }, { "epoch": 2.97, "learning_rate": 1.45036499192714e-05, "loss": 0.0587, "step": 16174 }, { "epoch": 2.97, "learning_rate": 1.4502987032264557e-05, "loss": 0.0602, "step": 16175 }, { "epoch": 2.97, "learning_rate": 1.4502324120437139e-05, "loss": 0.0648, "step": 16176 }, { "epoch": 2.97, "learning_rate": 1.4501661183792807e-05, "loss": 0.0655, "step": 16177 }, { "epoch": 2.97, "learning_rate": 1.4500998222335214e-05, "loss": 0.0539, "step": 16178 }, { "epoch": 2.97, "learning_rate": 1.4500335236068013e-05, "loss": 0.0669, "step": 16179 }, { "epoch": 2.97, "learning_rate": 1.4499672224994859e-05, "loss": 0.0559, "step": 16180 }, { "epoch": 2.97, "learning_rate": 1.4499009189119407e-05, "loss": 0.0627, "step": 16181 }, { "epoch": 2.97, "learning_rate": 1.4498346128445308e-05, "loss": 0.0667, "step": 16182 }, { "epoch": 2.97, "learning_rate": 1.4497683042976221e-05, "loss": 0.0636, "step": 16183 }, { "epoch": 2.97, "learning_rate": 1.44970199327158e-05, "loss": 0.0645, "step": 16184 }, { "epoch": 2.97, "learning_rate": 1.44963567976677e-05, "loss": 0.0669, "step": 16185 }, { "epoch": 2.97, "learning_rate": 1.4495693637835576e-05, "loss": 0.059, "step": 16186 }, { "epoch": 2.97, "learning_rate": 1.4495030453223083e-05, "loss": 0.0743, "step": 16187 }, { "epoch": 2.97, "learning_rate": 1.4494367243833878e-05, "loss": 0.0688, "step": 16188 }, { "epoch": 2.97, "learning_rate": 1.4493704009671614e-05, "loss": 0.056, "step": 16189 }, { "epoch": 2.97, "learning_rate": 1.4493040750739949e-05, "loss": 0.0756, "step": 16190 }, { "epoch": 2.97, "learning_rate": 1.4492377467042538e-05, "loss": 0.0632, "step": 16191 }, { "epoch": 2.97, "learning_rate": 1.4491714158583037e-05, "loss": 0.0597, "step": 16192 }, { "epoch": 2.97, "learning_rate": 1.4491050825365105e-05, "loss": 0.0619, "step": 16193 }, { "epoch": 2.97, "learning_rate": 1.4490387467392391e-05, "loss": 0.0571, "step": 16194 }, { "epoch": 2.97, "learning_rate": 1.4489724084668556e-05, "loss": 0.0703, "step": 16195 }, { "epoch": 2.97, "learning_rate": 1.448906067719726e-05, "loss": 0.0664, "step": 16196 }, { "epoch": 2.97, "learning_rate": 1.4488397244982156e-05, "loss": 0.0694, "step": 16197 }, { "epoch": 2.97, "learning_rate": 1.4487733788026898e-05, "loss": 0.0697, "step": 16198 }, { "epoch": 2.97, "learning_rate": 1.4487070306335151e-05, "loss": 0.0607, "step": 16199 }, { "epoch": 2.97, "learning_rate": 1.4486406799910567e-05, "loss": 0.0641, "step": 16200 }, { "epoch": 2.97, "learning_rate": 1.4485743268756798e-05, "loss": 0.0597, "step": 16201 }, { "epoch": 2.97, "learning_rate": 1.4485079712877511e-05, "loss": 0.0645, "step": 16202 }, { "epoch": 2.97, "learning_rate": 1.4484416132276358e-05, "loss": 0.0737, "step": 16203 }, { "epoch": 2.97, "learning_rate": 1.4483752526957e-05, "loss": 0.0576, "step": 16204 }, { "epoch": 2.97, "learning_rate": 1.4483088896923092e-05, "loss": 0.0674, "step": 16205 }, { "epoch": 2.97, "learning_rate": 1.4482425242178296e-05, "loss": 0.0658, "step": 16206 }, { "epoch": 2.97, "learning_rate": 1.4481761562726261e-05, "loss": 0.0714, "step": 16207 }, { "epoch": 2.97, "learning_rate": 1.4481097858570657e-05, "loss": 0.0653, "step": 16208 }, { "epoch": 2.97, "learning_rate": 1.4480434129715133e-05, "loss": 0.0691, "step": 16209 }, { "epoch": 2.97, "learning_rate": 1.4479770376163354e-05, "loss": 0.0586, "step": 16210 }, { "epoch": 2.97, "learning_rate": 1.4479106597918975e-05, "loss": 0.0689, "step": 16211 }, { "epoch": 2.97, "learning_rate": 1.4478442794985653e-05, "loss": 0.063, "step": 16212 }, { "epoch": 2.97, "learning_rate": 1.4477778967367051e-05, "loss": 0.0667, "step": 16213 }, { "epoch": 2.97, "learning_rate": 1.4477115115066827e-05, "loss": 0.0567, "step": 16214 }, { "epoch": 2.97, "learning_rate": 1.4476451238088637e-05, "loss": 0.0749, "step": 16215 }, { "epoch": 2.97, "learning_rate": 1.4475787336436146e-05, "loss": 0.0663, "step": 16216 }, { "epoch": 2.97, "learning_rate": 1.447512341011301e-05, "loss": 0.0703, "step": 16217 }, { "epoch": 2.97, "learning_rate": 1.4474459459122887e-05, "loss": 0.0647, "step": 16218 }, { "epoch": 2.97, "learning_rate": 1.4473795483469442e-05, "loss": 0.0637, "step": 16219 }, { "epoch": 2.97, "learning_rate": 1.4473131483156326e-05, "loss": 0.0552, "step": 16220 }, { "epoch": 2.97, "learning_rate": 1.447246745818721e-05, "loss": 0.0693, "step": 16221 }, { "epoch": 2.97, "learning_rate": 1.4471803408565745e-05, "loss": 0.0731, "step": 16222 }, { "epoch": 2.98, "learning_rate": 1.44711393342956e-05, "loss": 0.0627, "step": 16223 }, { "epoch": 2.98, "learning_rate": 1.4470475235380424e-05, "loss": 0.057, "step": 16224 }, { "epoch": 2.98, "learning_rate": 1.4469811111823885e-05, "loss": 0.0573, "step": 16225 }, { "epoch": 2.98, "learning_rate": 1.4469146963629646e-05, "loss": 0.0653, "step": 16226 }, { "epoch": 2.98, "learning_rate": 1.4468482790801358e-05, "loss": 0.0657, "step": 16227 }, { "epoch": 2.98, "learning_rate": 1.4467818593342695e-05, "loss": 0.0643, "step": 16228 }, { "epoch": 2.98, "learning_rate": 1.4467154371257307e-05, "loss": 0.0559, "step": 16229 }, { "epoch": 2.98, "learning_rate": 1.4466490124548862e-05, "loss": 0.0749, "step": 16230 }, { "epoch": 2.98, "learning_rate": 1.4465825853221019e-05, "loss": 0.0565, "step": 16231 }, { "epoch": 2.98, "learning_rate": 1.4465161557277437e-05, "loss": 0.0695, "step": 16232 }, { "epoch": 2.98, "learning_rate": 1.4464497236721779e-05, "loss": 0.0712, "step": 16233 }, { "epoch": 2.98, "learning_rate": 1.446383289155771e-05, "loss": 0.0646, "step": 16234 }, { "epoch": 2.98, "learning_rate": 1.446316852178889e-05, "loss": 0.0669, "step": 16235 }, { "epoch": 2.98, "learning_rate": 1.4462504127418977e-05, "loss": 0.0706, "step": 16236 }, { "epoch": 2.98, "learning_rate": 1.4461839708451641e-05, "loss": 0.0562, "step": 16237 }, { "epoch": 2.98, "learning_rate": 1.4461175264890536e-05, "loss": 0.0662, "step": 16238 }, { "epoch": 2.98, "learning_rate": 1.4460510796739332e-05, "loss": 0.0763, "step": 16239 }, { "epoch": 2.98, "learning_rate": 1.4459846304001688e-05, "loss": 0.072, "step": 16240 }, { "epoch": 2.98, "learning_rate": 1.4459181786681264e-05, "loss": 0.0656, "step": 16241 }, { "epoch": 2.98, "learning_rate": 1.4458517244781724e-05, "loss": 0.0651, "step": 16242 }, { "epoch": 2.98, "learning_rate": 1.4457852678306738e-05, "loss": 0.0669, "step": 16243 }, { "epoch": 2.98, "learning_rate": 1.4457188087259964e-05, "loss": 0.0737, "step": 16244 }, { "epoch": 2.98, "learning_rate": 1.445652347164506e-05, "loss": 0.0637, "step": 16245 }, { "epoch": 2.98, "learning_rate": 1.4455858831465696e-05, "loss": 0.0618, "step": 16246 }, { "epoch": 2.98, "learning_rate": 1.4455194166725532e-05, "loss": 0.0629, "step": 16247 }, { "epoch": 2.98, "learning_rate": 1.4454529477428237e-05, "loss": 0.0718, "step": 16248 }, { "epoch": 2.98, "learning_rate": 1.4453864763577468e-05, "loss": 0.0592, "step": 16249 }, { "epoch": 2.98, "learning_rate": 1.4453200025176895e-05, "loss": 0.0626, "step": 16250 }, { "epoch": 2.98, "learning_rate": 1.4452535262230174e-05, "loss": 0.0661, "step": 16251 }, { "epoch": 2.98, "learning_rate": 1.4451870474740978e-05, "loss": 0.0655, "step": 16252 }, { "epoch": 2.98, "learning_rate": 1.4451205662712969e-05, "loss": 0.0551, "step": 16253 }, { "epoch": 2.98, "learning_rate": 1.4450540826149808e-05, "loss": 0.0679, "step": 16254 }, { "epoch": 2.98, "learning_rate": 1.4449875965055164e-05, "loss": 0.0687, "step": 16255 }, { "epoch": 2.98, "learning_rate": 1.4449211079432694e-05, "loss": 0.0614, "step": 16256 }, { "epoch": 2.98, "learning_rate": 1.4448546169286073e-05, "loss": 0.0632, "step": 16257 }, { "epoch": 2.98, "learning_rate": 1.444788123461896e-05, "loss": 0.0549, "step": 16258 }, { "epoch": 2.98, "learning_rate": 1.4447216275435023e-05, "loss": 0.0595, "step": 16259 }, { "epoch": 2.98, "learning_rate": 1.4446551291737923e-05, "loss": 0.0582, "step": 16260 }, { "epoch": 2.98, "learning_rate": 1.4445886283531331e-05, "loss": 0.0649, "step": 16261 }, { "epoch": 2.98, "learning_rate": 1.4445221250818908e-05, "loss": 0.0596, "step": 16262 }, { "epoch": 2.98, "learning_rate": 1.444455619360432e-05, "loss": 0.0681, "step": 16263 }, { "epoch": 2.98, "learning_rate": 1.4443891111891239e-05, "loss": 0.0604, "step": 16264 }, { "epoch": 2.98, "learning_rate": 1.444322600568332e-05, "loss": 0.0581, "step": 16265 }, { "epoch": 2.98, "learning_rate": 1.444256087498424e-05, "loss": 0.064, "step": 16266 }, { "epoch": 2.98, "learning_rate": 1.4441895719797656e-05, "loss": 0.0742, "step": 16267 }, { "epoch": 2.98, "learning_rate": 1.4441230540127242e-05, "loss": 0.0755, "step": 16268 }, { "epoch": 2.98, "learning_rate": 1.444056533597666e-05, "loss": 0.065, "step": 16269 }, { "epoch": 2.98, "learning_rate": 1.4439900107349578e-05, "loss": 0.0661, "step": 16270 }, { "epoch": 2.98, "learning_rate": 1.4439234854249661e-05, "loss": 0.0618, "step": 16271 }, { "epoch": 2.98, "learning_rate": 1.4438569576680579e-05, "loss": 0.0571, "step": 16272 }, { "epoch": 2.98, "learning_rate": 1.4437904274645997e-05, "loss": 0.0744, "step": 16273 }, { "epoch": 2.98, "learning_rate": 1.4437238948149586e-05, "loss": 0.0538, "step": 16274 }, { "epoch": 2.98, "learning_rate": 1.4436573597195007e-05, "loss": 0.0692, "step": 16275 }, { "epoch": 2.98, "learning_rate": 1.443590822178593e-05, "loss": 0.0579, "step": 16276 }, { "epoch": 2.98, "learning_rate": 1.4435242821926024e-05, "loss": 0.0552, "step": 16277 }, { "epoch": 2.99, "learning_rate": 1.4434577397618954e-05, "loss": 0.0717, "step": 16278 }, { "epoch": 2.99, "learning_rate": 1.4433911948868394e-05, "loss": 0.0698, "step": 16279 }, { "epoch": 2.99, "learning_rate": 1.4433246475678e-05, "loss": 0.0745, "step": 16280 }, { "epoch": 2.99, "learning_rate": 1.4432580978051451e-05, "loss": 0.063, "step": 16281 }, { "epoch": 2.99, "learning_rate": 1.4431915455992416e-05, "loss": 0.0717, "step": 16282 }, { "epoch": 2.99, "learning_rate": 1.4431249909504555e-05, "loss": 0.0789, "step": 16283 }, { "epoch": 2.99, "learning_rate": 1.4430584338591543e-05, "loss": 0.0628, "step": 16284 }, { "epoch": 2.99, "learning_rate": 1.4429918743257044e-05, "loss": 0.0693, "step": 16285 }, { "epoch": 2.99, "learning_rate": 1.442925312350473e-05, "loss": 0.0621, "step": 16286 }, { "epoch": 2.99, "learning_rate": 1.4428587479338268e-05, "loss": 0.0538, "step": 16287 }, { "epoch": 2.99, "learning_rate": 1.442792181076133e-05, "loss": 0.0566, "step": 16288 }, { "epoch": 2.99, "learning_rate": 1.442725611777758e-05, "loss": 0.0644, "step": 16289 }, { "epoch": 2.99, "learning_rate": 1.4426590400390691e-05, "loss": 0.0723, "step": 16290 }, { "epoch": 2.99, "learning_rate": 1.4425924658604335e-05, "loss": 0.0684, "step": 16291 }, { "epoch": 2.99, "learning_rate": 1.4425258892422177e-05, "loss": 0.0695, "step": 16292 }, { "epoch": 2.99, "learning_rate": 1.442459310184789e-05, "loss": 0.0669, "step": 16293 }, { "epoch": 2.99, "learning_rate": 1.442392728688514e-05, "loss": 0.0646, "step": 16294 }, { "epoch": 2.99, "learning_rate": 1.44232614475376e-05, "loss": 0.0689, "step": 16295 }, { "epoch": 2.99, "learning_rate": 1.442259558380894e-05, "loss": 0.0797, "step": 16296 }, { "epoch": 2.99, "learning_rate": 1.442192969570283e-05, "loss": 0.0608, "step": 16297 }, { "epoch": 2.99, "learning_rate": 1.4421263783222936e-05, "loss": 0.0671, "step": 16298 }, { "epoch": 2.99, "learning_rate": 1.4420597846372936e-05, "loss": 0.0627, "step": 16299 }, { "epoch": 2.99, "learning_rate": 1.4419931885156495e-05, "loss": 0.0615, "step": 16300 }, { "epoch": 2.99, "learning_rate": 1.4419265899577288e-05, "loss": 0.0664, "step": 16301 }, { "epoch": 2.99, "learning_rate": 1.4418599889638984e-05, "loss": 0.072, "step": 16302 }, { "epoch": 2.99, "learning_rate": 1.4417933855345254e-05, "loss": 0.0601, "step": 16303 }, { "epoch": 2.99, "learning_rate": 1.4417267796699766e-05, "loss": 0.0532, "step": 16304 }, { "epoch": 2.99, "learning_rate": 1.4416601713706197e-05, "loss": 0.0677, "step": 16305 }, { "epoch": 2.99, "learning_rate": 1.4415935606368215e-05, "loss": 0.0612, "step": 16306 }, { "epoch": 2.99, "learning_rate": 1.4415269474689492e-05, "loss": 0.0664, "step": 16307 }, { "epoch": 2.99, "learning_rate": 1.44146033186737e-05, "loss": 0.0673, "step": 16308 }, { "epoch": 2.99, "learning_rate": 1.4413937138324514e-05, "loss": 0.0753, "step": 16309 }, { "epoch": 2.99, "learning_rate": 1.44132709336456e-05, "loss": 0.0611, "step": 16310 }, { "epoch": 2.99, "learning_rate": 1.4412604704640636e-05, "loss": 0.0682, "step": 16311 }, { "epoch": 2.99, "learning_rate": 1.441193845131329e-05, "loss": 0.0615, "step": 16312 }, { "epoch": 2.99, "learning_rate": 1.4411272173667235e-05, "loss": 0.0683, "step": 16313 }, { "epoch": 2.99, "learning_rate": 1.4410605871706147e-05, "loss": 0.0659, "step": 16314 }, { "epoch": 2.99, "learning_rate": 1.440993954543369e-05, "loss": 0.0639, "step": 16315 }, { "epoch": 2.99, "learning_rate": 1.440927319485355e-05, "loss": 0.0684, "step": 16316 }, { "epoch": 2.99, "learning_rate": 1.4408606819969388e-05, "loss": 0.08, "step": 16317 }, { "epoch": 2.99, "learning_rate": 1.4407940420784885e-05, "loss": 0.0658, "step": 16318 }, { "epoch": 2.99, "learning_rate": 1.440727399730371e-05, "loss": 0.0691, "step": 16319 }, { "epoch": 2.99, "learning_rate": 1.440660754952954e-05, "loss": 0.0636, "step": 16320 }, { "epoch": 2.99, "learning_rate": 1.4405941077466041e-05, "loss": 0.0562, "step": 16321 }, { "epoch": 2.99, "learning_rate": 1.4405274581116896e-05, "loss": 0.0699, "step": 16322 }, { "epoch": 2.99, "learning_rate": 1.4404608060485772e-05, "loss": 0.072, "step": 16323 }, { "epoch": 2.99, "learning_rate": 1.4403941515576344e-05, "loss": 0.0603, "step": 16324 }, { "epoch": 2.99, "learning_rate": 1.4403274946392289e-05, "loss": 0.0609, "step": 16325 }, { "epoch": 2.99, "learning_rate": 1.4402608352937279e-05, "loss": 0.0626, "step": 16326 }, { "epoch": 2.99, "learning_rate": 1.440194173521499e-05, "loss": 0.0609, "step": 16327 }, { "epoch": 2.99, "learning_rate": 1.4401275093229092e-05, "loss": 0.0678, "step": 16328 }, { "epoch": 2.99, "learning_rate": 1.4400608426983267e-05, "loss": 0.0574, "step": 16329 }, { "epoch": 2.99, "learning_rate": 1.4399941736481181e-05, "loss": 0.0595, "step": 16330 }, { "epoch": 2.99, "learning_rate": 1.4399275021726513e-05, "loss": 0.0598, "step": 16331 }, { "epoch": 3.0, "learning_rate": 1.4398608282722943e-05, "loss": 0.0618, "step": 16332 }, { "epoch": 3.0, "learning_rate": 1.4397941519474135e-05, "loss": 0.068, "step": 16333 }, { "epoch": 3.0, "learning_rate": 1.4397274731983775e-05, "loss": 0.0682, "step": 16334 }, { "epoch": 3.0, "learning_rate": 1.4396607920255531e-05, "loss": 0.0605, "step": 16335 }, { "epoch": 3.0, "learning_rate": 1.4395941084293081e-05, "loss": 0.061, "step": 16336 }, { "epoch": 3.0, "learning_rate": 1.43952742241001e-05, "loss": 0.0718, "step": 16337 }, { "epoch": 3.0, "learning_rate": 1.4394607339680268e-05, "loss": 0.0639, "step": 16338 }, { "epoch": 3.0, "learning_rate": 1.4393940431037256e-05, "loss": 0.0667, "step": 16339 }, { "epoch": 3.0, "learning_rate": 1.439327349817474e-05, "loss": 0.0644, "step": 16340 }, { "epoch": 3.0, "learning_rate": 1.4392606541096402e-05, "loss": 0.0676, "step": 16341 }, { "epoch": 3.0, "learning_rate": 1.4391939559805908e-05, "loss": 0.0727, "step": 16342 }, { "epoch": 3.0, "learning_rate": 1.439127255430694e-05, "loss": 0.0647, "step": 16343 }, { "epoch": 3.0, "learning_rate": 1.439060552460318e-05, "loss": 0.0732, "step": 16344 }, { "epoch": 3.0, "learning_rate": 1.4389938470698297e-05, "loss": 0.0586, "step": 16345 }, { "epoch": 3.0, "learning_rate": 1.438927139259597e-05, "loss": 0.0713, "step": 16346 }, { "epoch": 3.0, "learning_rate": 1.4388604290299877e-05, "loss": 0.0637, "step": 16347 }, { "epoch": 3.0, "learning_rate": 1.4387937163813693e-05, "loss": 0.0771, "step": 16348 }, { "epoch": 3.0, "learning_rate": 1.4387270013141098e-05, "loss": 0.0577, "step": 16349 }, { "epoch": 3.0, "learning_rate": 1.4386602838285768e-05, "loss": 0.0679, "step": 16350 }, { "epoch": 3.0, "learning_rate": 1.4385935639251379e-05, "loss": 0.0576, "step": 16351 }, { "epoch": 3.0, "learning_rate": 1.438526841604161e-05, "loss": 0.0632, "step": 16352 }, { "epoch": 3.0, "learning_rate": 1.4384601168660139e-05, "loss": 0.0628, "step": 16353 }, { "epoch": 3.0, "learning_rate": 1.4383933897110645e-05, "loss": 0.0726, "step": 16354 }, { "epoch": 3.0, "learning_rate": 1.4383266601396801e-05, "loss": 0.0551, "step": 16355 }, { "epoch": 3.0, "learning_rate": 1.4382599281522294e-05, "loss": 0.0625, "step": 16356 }, { "epoch": 3.0, "learning_rate": 1.4381931937490793e-05, "loss": 0.0558, "step": 16357 }, { "epoch": 3.0, "learning_rate": 1.4381264569305981e-05, "loss": 0.0621, "step": 16358 }, { "epoch": 3.0, "learning_rate": 1.4380597176971538e-05, "loss": 0.0581, "step": 16359 }, { "epoch": 3.0, "learning_rate": 1.437992976049114e-05, "loss": 0.0444, "step": 16360 }, { "epoch": 3.0, "learning_rate": 1.4379262319868468e-05, "loss": 0.0452, "step": 16361 }, { "epoch": 3.0, "learning_rate": 1.4378594855107197e-05, "loss": 0.043, "step": 16362 }, { "epoch": 3.0, "learning_rate": 1.437792736621101e-05, "loss": 0.0531, "step": 16363 }, { "epoch": 3.0, "learning_rate": 1.4377259853183585e-05, "loss": 0.0522, "step": 16364 }, { "epoch": 3.0, "learning_rate": 1.43765923160286e-05, "loss": 0.0526, "step": 16365 }, { "epoch": 3.0, "learning_rate": 1.4375924754749732e-05, "loss": 0.0507, "step": 16366 }, { "epoch": 3.0, "learning_rate": 1.4375257169350668e-05, "loss": 0.0501, "step": 16367 }, { "epoch": 3.0, "learning_rate": 1.4374589559835088e-05, "loss": 0.0474, "step": 16368 }, { "epoch": 3.0, "learning_rate": 1.4373921926206662e-05, "loss": 0.0602, "step": 16369 }, { "epoch": 3.0, "learning_rate": 1.4373254268469079e-05, "loss": 0.0545, "step": 16370 }, { "epoch": 3.0, "learning_rate": 1.4372586586626015e-05, "loss": 0.0494, "step": 16371 }, { "epoch": 3.0, "learning_rate": 1.4371918880681153e-05, "loss": 0.0547, "step": 16372 }, { "epoch": 3.0, "learning_rate": 1.437125115063817e-05, "loss": 0.0454, "step": 16373 }, { "epoch": 3.0, "learning_rate": 1.4370583396500749e-05, "loss": 0.0434, "step": 16374 }, { "epoch": 3.0, "learning_rate": 1.4369915618272568e-05, "loss": 0.0471, "step": 16375 }, { "epoch": 3.0, "learning_rate": 1.436924781595731e-05, "loss": 0.0538, "step": 16376 }, { "epoch": 3.0, "learning_rate": 1.4368579989558656e-05, "loss": 0.0524, "step": 16377 }, { "epoch": 3.0, "learning_rate": 1.4367912139080287e-05, "loss": 0.0514, "step": 16378 }, { "epoch": 3.0, "learning_rate": 1.4367244264525886e-05, "loss": 0.0478, "step": 16379 }, { "epoch": 3.0, "learning_rate": 1.436657636589913e-05, "loss": 0.0553, "step": 16380 }, { "epoch": 3.0, "learning_rate": 1.4365908443203702e-05, "loss": 0.0559, "step": 16381 }, { "epoch": 3.0, "learning_rate": 1.4365240496443287e-05, "loss": 0.0471, "step": 16382 }, { "epoch": 3.0, "learning_rate": 1.4364572525621565e-05, "loss": 0.0458, "step": 16383 }, { "epoch": 3.0, "learning_rate": 1.4363904530742213e-05, "loss": 0.0564, "step": 16384 }, { "epoch": 3.0, "learning_rate": 1.4363236511808918e-05, "loss": 0.0524, "step": 16385 }, { "epoch": 3.0, "learning_rate": 1.4362568468825362e-05, "loss": 0.0556, "step": 16386 }, { "epoch": 3.01, "learning_rate": 1.4361900401795227e-05, "loss": 0.0522, "step": 16387 }, { "epoch": 3.01, "learning_rate": 1.4361232310722195e-05, "loss": 0.0501, "step": 16388 }, { "epoch": 3.01, "learning_rate": 1.4360564195609948e-05, "loss": 0.055, "step": 16389 }, { "epoch": 3.01, "learning_rate": 1.4359896056462166e-05, "loss": 0.0548, "step": 16390 }, { "epoch": 3.01, "learning_rate": 1.4359227893282539e-05, "loss": 0.0474, "step": 16391 }, { "epoch": 3.01, "learning_rate": 1.4358559706074743e-05, "loss": 0.0569, "step": 16392 }, { "epoch": 3.01, "learning_rate": 1.4357891494842464e-05, "loss": 0.0526, "step": 16393 }, { "epoch": 3.01, "learning_rate": 1.4357223259589386e-05, "loss": 0.0503, "step": 16394 }, { "epoch": 3.01, "learning_rate": 1.4356555000319192e-05, "loss": 0.0522, "step": 16395 }, { "epoch": 3.01, "learning_rate": 1.4355886717035563e-05, "loss": 0.0498, "step": 16396 }, { "epoch": 3.01, "learning_rate": 1.4355218409742187e-05, "loss": 0.0492, "step": 16397 }, { "epoch": 3.01, "learning_rate": 1.4354550078442743e-05, "loss": 0.0517, "step": 16398 }, { "epoch": 3.01, "learning_rate": 1.4353881723140917e-05, "loss": 0.0502, "step": 16399 }, { "epoch": 3.01, "learning_rate": 1.4353213343840393e-05, "loss": 0.0446, "step": 16400 }, { "epoch": 3.01, "learning_rate": 1.4352544940544855e-05, "loss": 0.0545, "step": 16401 }, { "epoch": 3.01, "learning_rate": 1.4351876513257988e-05, "loss": 0.0519, "step": 16402 }, { "epoch": 3.01, "learning_rate": 1.4351208061983477e-05, "loss": 0.0529, "step": 16403 }, { "epoch": 3.01, "learning_rate": 1.4350539586725e-05, "loss": 0.0472, "step": 16404 }, { "epoch": 3.01, "learning_rate": 1.4349871087486254e-05, "loss": 0.0522, "step": 16405 }, { "epoch": 3.01, "learning_rate": 1.4349202564270912e-05, "loss": 0.0491, "step": 16406 }, { "epoch": 3.01, "learning_rate": 1.4348534017082665e-05, "loss": 0.0475, "step": 16407 }, { "epoch": 3.01, "learning_rate": 1.4347865445925194e-05, "loss": 0.0513, "step": 16408 }, { "epoch": 3.01, "learning_rate": 1.4347196850802193e-05, "loss": 0.0434, "step": 16409 }, { "epoch": 3.01, "learning_rate": 1.4346528231717334e-05, "loss": 0.0451, "step": 16410 }, { "epoch": 3.01, "learning_rate": 1.434585958867431e-05, "loss": 0.0478, "step": 16411 }, { "epoch": 3.01, "learning_rate": 1.434519092167681e-05, "loss": 0.0517, "step": 16412 }, { "epoch": 3.01, "learning_rate": 1.4344522230728517e-05, "loss": 0.0511, "step": 16413 }, { "epoch": 3.01, "learning_rate": 1.4343853515833112e-05, "loss": 0.0494, "step": 16414 }, { "epoch": 3.01, "learning_rate": 1.4343184776994288e-05, "loss": 0.0529, "step": 16415 }, { "epoch": 3.01, "learning_rate": 1.4342516014215726e-05, "loss": 0.0575, "step": 16416 }, { "epoch": 3.01, "learning_rate": 1.434184722750111e-05, "loss": 0.0607, "step": 16417 }, { "epoch": 3.01, "learning_rate": 1.4341178416854138e-05, "loss": 0.047, "step": 16418 }, { "epoch": 3.01, "learning_rate": 1.4340509582278484e-05, "loss": 0.044, "step": 16419 }, { "epoch": 3.01, "learning_rate": 1.433984072377784e-05, "loss": 0.0478, "step": 16420 }, { "epoch": 3.01, "learning_rate": 1.4339171841355892e-05, "loss": 0.0388, "step": 16421 }, { "epoch": 3.01, "learning_rate": 1.433850293501633e-05, "loss": 0.0508, "step": 16422 }, { "epoch": 3.01, "learning_rate": 1.4337834004762834e-05, "loss": 0.0514, "step": 16423 }, { "epoch": 3.01, "learning_rate": 1.43371650505991e-05, "loss": 0.0362, "step": 16424 }, { "epoch": 3.01, "learning_rate": 1.4336496072528807e-05, "loss": 0.0577, "step": 16425 }, { "epoch": 3.01, "learning_rate": 1.4335827070555645e-05, "loss": 0.0427, "step": 16426 }, { "epoch": 3.01, "learning_rate": 1.4335158044683309e-05, "loss": 0.0544, "step": 16427 }, { "epoch": 3.01, "learning_rate": 1.4334488994915474e-05, "loss": 0.0481, "step": 16428 }, { "epoch": 3.01, "learning_rate": 1.4333819921255836e-05, "loss": 0.0519, "step": 16429 }, { "epoch": 3.01, "learning_rate": 1.4333150823708083e-05, "loss": 0.0434, "step": 16430 }, { "epoch": 3.01, "learning_rate": 1.4332481702275898e-05, "loss": 0.0535, "step": 16431 }, { "epoch": 3.01, "learning_rate": 1.4331812556962972e-05, "loss": 0.0468, "step": 16432 }, { "epoch": 3.01, "learning_rate": 1.4331143387772996e-05, "loss": 0.0547, "step": 16433 }, { "epoch": 3.01, "learning_rate": 1.4330474194709656e-05, "loss": 0.0603, "step": 16434 }, { "epoch": 3.01, "learning_rate": 1.432980497777664e-05, "loss": 0.0442, "step": 16435 }, { "epoch": 3.01, "learning_rate": 1.432913573697764e-05, "loss": 0.0604, "step": 16436 }, { "epoch": 3.01, "learning_rate": 1.4328466472316338e-05, "loss": 0.0547, "step": 16437 }, { "epoch": 3.01, "learning_rate": 1.432779718379643e-05, "loss": 0.0574, "step": 16438 }, { "epoch": 3.01, "learning_rate": 1.4327127871421605e-05, "loss": 0.0544, "step": 16439 }, { "epoch": 3.01, "learning_rate": 1.4326458535195546e-05, "loss": 0.0604, "step": 16440 }, { "epoch": 3.02, "learning_rate": 1.4325789175121947e-05, "loss": 0.0459, "step": 16441 }, { "epoch": 3.02, "learning_rate": 1.4325119791204496e-05, "loss": 0.0581, "step": 16442 }, { "epoch": 3.02, "learning_rate": 1.4324450383446884e-05, "loss": 0.0482, "step": 16443 }, { "epoch": 3.02, "learning_rate": 1.4323780951852802e-05, "loss": 0.0526, "step": 16444 }, { "epoch": 3.02, "learning_rate": 1.4323111496425934e-05, "loss": 0.0513, "step": 16445 }, { "epoch": 3.02, "learning_rate": 1.4322442017169978e-05, "loss": 0.054, "step": 16446 }, { "epoch": 3.02, "learning_rate": 1.4321772514088622e-05, "loss": 0.0445, "step": 16447 }, { "epoch": 3.02, "learning_rate": 1.4321102987185551e-05, "loss": 0.0582, "step": 16448 }, { "epoch": 3.02, "learning_rate": 1.4320433436464458e-05, "loss": 0.0428, "step": 16449 }, { "epoch": 3.02, "learning_rate": 1.4319763861929037e-05, "loss": 0.0542, "step": 16450 }, { "epoch": 3.02, "learning_rate": 1.4319094263582977e-05, "loss": 0.0443, "step": 16451 }, { "epoch": 3.02, "learning_rate": 1.4318424641429967e-05, "loss": 0.0575, "step": 16452 }, { "epoch": 3.02, "learning_rate": 1.4317754995473699e-05, "loss": 0.0517, "step": 16453 }, { "epoch": 3.02, "learning_rate": 1.4317085325717863e-05, "loss": 0.0489, "step": 16454 }, { "epoch": 3.02, "learning_rate": 1.4316415632166154e-05, "loss": 0.0503, "step": 16455 }, { "epoch": 3.02, "learning_rate": 1.4315745914822262e-05, "loss": 0.046, "step": 16456 }, { "epoch": 3.02, "learning_rate": 1.4315076173689875e-05, "loss": 0.0448, "step": 16457 }, { "epoch": 3.02, "learning_rate": 1.4314406408772687e-05, "loss": 0.0551, "step": 16458 }, { "epoch": 3.02, "learning_rate": 1.4313736620074389e-05, "loss": 0.0476, "step": 16459 }, { "epoch": 3.02, "learning_rate": 1.4313066807598677e-05, "loss": 0.0463, "step": 16460 }, { "epoch": 3.02, "learning_rate": 1.4312396971349236e-05, "loss": 0.0577, "step": 16461 }, { "epoch": 3.02, "learning_rate": 1.4311727111329765e-05, "loss": 0.0398, "step": 16462 }, { "epoch": 3.02, "learning_rate": 1.4311057227543951e-05, "loss": 0.0399, "step": 16463 }, { "epoch": 3.02, "learning_rate": 1.4310387319995492e-05, "loss": 0.0473, "step": 16464 }, { "epoch": 3.02, "learning_rate": 1.4309717388688074e-05, "loss": 0.0581, "step": 16465 }, { "epoch": 3.02, "learning_rate": 1.4309047433625392e-05, "loss": 0.0509, "step": 16466 }, { "epoch": 3.02, "learning_rate": 1.4308377454811141e-05, "loss": 0.0521, "step": 16467 }, { "epoch": 3.02, "learning_rate": 1.4307707452249013e-05, "loss": 0.0534, "step": 16468 }, { "epoch": 3.02, "learning_rate": 1.43070374259427e-05, "loss": 0.0532, "step": 16469 }, { "epoch": 3.02, "learning_rate": 1.4306367375895894e-05, "loss": 0.0531, "step": 16470 }, { "epoch": 3.02, "learning_rate": 1.4305697302112291e-05, "loss": 0.0554, "step": 16471 }, { "epoch": 3.02, "learning_rate": 1.4305027204595586e-05, "loss": 0.056, "step": 16472 }, { "epoch": 3.02, "learning_rate": 1.4304357083349469e-05, "loss": 0.0504, "step": 16473 }, { "epoch": 3.02, "learning_rate": 1.4303686938377634e-05, "loss": 0.0489, "step": 16474 }, { "epoch": 3.02, "learning_rate": 1.4303016769683778e-05, "loss": 0.052, "step": 16475 }, { "epoch": 3.02, "learning_rate": 1.4302346577271591e-05, "loss": 0.0517, "step": 16476 }, { "epoch": 3.02, "learning_rate": 1.4301676361144769e-05, "loss": 0.0447, "step": 16477 }, { "epoch": 3.02, "learning_rate": 1.4301006121307008e-05, "loss": 0.0538, "step": 16478 }, { "epoch": 3.02, "learning_rate": 1.4300335857761995e-05, "loss": 0.052, "step": 16479 }, { "epoch": 3.02, "learning_rate": 1.4299665570513437e-05, "loss": 0.0467, "step": 16480 }, { "epoch": 3.02, "learning_rate": 1.4298995259565016e-05, "loss": 0.0561, "step": 16481 }, { "epoch": 3.02, "learning_rate": 1.4298324924920434e-05, "loss": 0.0585, "step": 16482 }, { "epoch": 3.02, "learning_rate": 1.4297654566583387e-05, "loss": 0.0497, "step": 16483 }, { "epoch": 3.02, "learning_rate": 1.4296984184557564e-05, "loss": 0.0556, "step": 16484 }, { "epoch": 3.02, "learning_rate": 1.4296313778846664e-05, "loss": 0.0499, "step": 16485 }, { "epoch": 3.02, "learning_rate": 1.4295643349454386e-05, "loss": 0.0495, "step": 16486 }, { "epoch": 3.02, "learning_rate": 1.4294972896384415e-05, "loss": 0.0523, "step": 16487 }, { "epoch": 3.02, "learning_rate": 1.4294302419640458e-05, "loss": 0.0521, "step": 16488 }, { "epoch": 3.02, "learning_rate": 1.4293631919226205e-05, "loss": 0.0477, "step": 16489 }, { "epoch": 3.02, "learning_rate": 1.4292961395145347e-05, "loss": 0.0464, "step": 16490 }, { "epoch": 3.02, "learning_rate": 1.429229084740159e-05, "loss": 0.0525, "step": 16491 }, { "epoch": 3.02, "learning_rate": 1.4291620275998624e-05, "loss": 0.0501, "step": 16492 }, { "epoch": 3.02, "learning_rate": 1.4290949680940147e-05, "loss": 0.052, "step": 16493 }, { "epoch": 3.02, "learning_rate": 1.4290279062229852e-05, "loss": 0.0429, "step": 16494 }, { "epoch": 3.02, "learning_rate": 1.4289608419871443e-05, "loss": 0.0568, "step": 16495 }, { "epoch": 3.03, "learning_rate": 1.4288937753868607e-05, "loss": 0.0512, "step": 16496 }, { "epoch": 3.03, "learning_rate": 1.428826706422505e-05, "loss": 0.0579, "step": 16497 }, { "epoch": 3.03, "learning_rate": 1.4287596350944462e-05, "loss": 0.0492, "step": 16498 }, { "epoch": 3.03, "learning_rate": 1.4286925614030542e-05, "loss": 0.0498, "step": 16499 }, { "epoch": 3.03, "learning_rate": 1.428625485348699e-05, "loss": 0.0449, "step": 16500 }, { "epoch": 3.03, "learning_rate": 1.4285584069317499e-05, "loss": 0.0573, "step": 16501 }, { "epoch": 3.03, "learning_rate": 1.428491326152577e-05, "loss": 0.0503, "step": 16502 }, { "epoch": 3.03, "learning_rate": 1.4284242430115495e-05, "loss": 0.046, "step": 16503 }, { "epoch": 3.03, "learning_rate": 1.428357157509038e-05, "loss": 0.0516, "step": 16504 }, { "epoch": 3.03, "learning_rate": 1.4282900696454115e-05, "loss": 0.054, "step": 16505 }, { "epoch": 3.03, "learning_rate": 1.4282229794210404e-05, "loss": 0.0465, "step": 16506 }, { "epoch": 3.03, "learning_rate": 1.4281558868362937e-05, "loss": 0.0555, "step": 16507 }, { "epoch": 3.03, "learning_rate": 1.4280887918915421e-05, "loss": 0.0638, "step": 16508 }, { "epoch": 3.03, "learning_rate": 1.4280216945871548e-05, "loss": 0.0485, "step": 16509 }, { "epoch": 3.03, "learning_rate": 1.4279545949235024e-05, "loss": 0.045, "step": 16510 }, { "epoch": 3.03, "learning_rate": 1.4278874929009538e-05, "loss": 0.0493, "step": 16511 }, { "epoch": 3.03, "learning_rate": 1.4278203885198791e-05, "loss": 0.0541, "step": 16512 }, { "epoch": 3.03, "learning_rate": 1.4277532817806491e-05, "loss": 0.047, "step": 16513 }, { "epoch": 3.03, "learning_rate": 1.4276861726836325e-05, "loss": 0.0527, "step": 16514 }, { "epoch": 3.03, "learning_rate": 1.4276190612291998e-05, "loss": 0.0463, "step": 16515 }, { "epoch": 3.03, "learning_rate": 1.4275519474177209e-05, "loss": 0.0525, "step": 16516 }, { "epoch": 3.03, "learning_rate": 1.4274848312495656e-05, "loss": 0.0541, "step": 16517 }, { "epoch": 3.03, "learning_rate": 1.4274177127251037e-05, "loss": 0.0507, "step": 16518 }, { "epoch": 3.03, "learning_rate": 1.4273505918447054e-05, "loss": 0.0484, "step": 16519 }, { "epoch": 3.03, "learning_rate": 1.4272834686087409e-05, "loss": 0.061, "step": 16520 }, { "epoch": 3.03, "learning_rate": 1.4272163430175797e-05, "loss": 0.05, "step": 16521 }, { "epoch": 3.03, "learning_rate": 1.4271492150715922e-05, "loss": 0.0498, "step": 16522 }, { "epoch": 3.03, "learning_rate": 1.427082084771148e-05, "loss": 0.0462, "step": 16523 }, { "epoch": 3.03, "learning_rate": 1.4270149521166174e-05, "loss": 0.0603, "step": 16524 }, { "epoch": 3.03, "learning_rate": 1.4269478171083704e-05, "loss": 0.0525, "step": 16525 }, { "epoch": 3.03, "learning_rate": 1.4268806797467772e-05, "loss": 0.0575, "step": 16526 }, { "epoch": 3.03, "learning_rate": 1.4268135400322074e-05, "loss": 0.0495, "step": 16527 }, { "epoch": 3.03, "learning_rate": 1.4267463979650316e-05, "loss": 0.0448, "step": 16528 }, { "epoch": 3.03, "learning_rate": 1.4266792535456193e-05, "loss": 0.0562, "step": 16529 }, { "epoch": 3.03, "learning_rate": 1.4266121067743413e-05, "loss": 0.0515, "step": 16530 }, { "epoch": 3.03, "learning_rate": 1.4265449576515675e-05, "loss": 0.0564, "step": 16531 }, { "epoch": 3.03, "learning_rate": 1.4264778061776673e-05, "loss": 0.0485, "step": 16532 }, { "epoch": 3.03, "learning_rate": 1.4264106523530119e-05, "loss": 0.04, "step": 16533 }, { "epoch": 3.03, "learning_rate": 1.4263434961779709e-05, "loss": 0.049, "step": 16534 }, { "epoch": 3.03, "learning_rate": 1.4262763376529148e-05, "loss": 0.0495, "step": 16535 }, { "epoch": 3.03, "learning_rate": 1.4262091767782133e-05, "loss": 0.0436, "step": 16536 }, { "epoch": 3.03, "learning_rate": 1.426142013554237e-05, "loss": 0.0566, "step": 16537 }, { "epoch": 3.03, "learning_rate": 1.4260748479813555e-05, "loss": 0.0521, "step": 16538 }, { "epoch": 3.03, "learning_rate": 1.4260076800599398e-05, "loss": 0.0445, "step": 16539 }, { "epoch": 3.03, "learning_rate": 1.4259405097903598e-05, "loss": 0.0568, "step": 16540 }, { "epoch": 3.03, "learning_rate": 1.4258733371729857e-05, "loss": 0.0533, "step": 16541 }, { "epoch": 3.03, "learning_rate": 1.4258061622081876e-05, "loss": 0.0539, "step": 16542 }, { "epoch": 3.03, "learning_rate": 1.425738984896336e-05, "loss": 0.0537, "step": 16543 }, { "epoch": 3.03, "learning_rate": 1.4256718052378016e-05, "loss": 0.0465, "step": 16544 }, { "epoch": 3.03, "learning_rate": 1.425604623232954e-05, "loss": 0.0546, "step": 16545 }, { "epoch": 3.03, "learning_rate": 1.4255374388821634e-05, "loss": 0.0528, "step": 16546 }, { "epoch": 3.03, "learning_rate": 1.4254702521858006e-05, "loss": 0.0496, "step": 16547 }, { "epoch": 3.03, "learning_rate": 1.4254030631442359e-05, "loss": 0.045, "step": 16548 }, { "epoch": 3.03, "learning_rate": 1.4253358717578396e-05, "loss": 0.0489, "step": 16549 }, { "epoch": 3.04, "learning_rate": 1.425268678026982e-05, "loss": 0.0485, "step": 16550 }, { "epoch": 3.04, "learning_rate": 1.4252014819520335e-05, "loss": 0.0498, "step": 16551 }, { "epoch": 3.04, "learning_rate": 1.4251342835333644e-05, "loss": 0.0514, "step": 16552 }, { "epoch": 3.04, "learning_rate": 1.4250670827713452e-05, "loss": 0.0504, "step": 16553 }, { "epoch": 3.04, "learning_rate": 1.4249998796663463e-05, "loss": 0.0538, "step": 16554 }, { "epoch": 3.04, "learning_rate": 1.4249326742187382e-05, "loss": 0.055, "step": 16555 }, { "epoch": 3.04, "learning_rate": 1.4248654664288908e-05, "loss": 0.0501, "step": 16556 }, { "epoch": 3.04, "learning_rate": 1.4247982562971757e-05, "loss": 0.0496, "step": 16557 }, { "epoch": 3.04, "learning_rate": 1.4247310438239621e-05, "loss": 0.05, "step": 16558 }, { "epoch": 3.04, "learning_rate": 1.4246638290096214e-05, "loss": 0.0591, "step": 16559 }, { "epoch": 3.04, "learning_rate": 1.4245966118545233e-05, "loss": 0.0438, "step": 16560 }, { "epoch": 3.04, "learning_rate": 1.424529392359039e-05, "loss": 0.055, "step": 16561 }, { "epoch": 3.04, "learning_rate": 1.4244621705235386e-05, "loss": 0.0554, "step": 16562 }, { "epoch": 3.04, "learning_rate": 1.424394946348393e-05, "loss": 0.0497, "step": 16563 }, { "epoch": 3.04, "learning_rate": 1.4243277198339724e-05, "loss": 0.0586, "step": 16564 }, { "epoch": 3.04, "learning_rate": 1.4242604909806475e-05, "loss": 0.047, "step": 16565 }, { "epoch": 3.04, "learning_rate": 1.4241932597887888e-05, "loss": 0.0484, "step": 16566 }, { "epoch": 3.04, "learning_rate": 1.4241260262587668e-05, "loss": 0.0513, "step": 16567 }, { "epoch": 3.04, "learning_rate": 1.4240587903909524e-05, "loss": 0.0407, "step": 16568 }, { "epoch": 3.04, "learning_rate": 1.423991552185716e-05, "loss": 0.0506, "step": 16569 }, { "epoch": 3.04, "learning_rate": 1.423924311643428e-05, "loss": 0.0429, "step": 16570 }, { "epoch": 3.04, "learning_rate": 1.4238570687644593e-05, "loss": 0.0482, "step": 16571 }, { "epoch": 3.04, "learning_rate": 1.423789823549181e-05, "loss": 0.0435, "step": 16572 }, { "epoch": 3.04, "learning_rate": 1.4237225759979627e-05, "loss": 0.0511, "step": 16573 }, { "epoch": 3.04, "learning_rate": 1.4236553261111758e-05, "loss": 0.0629, "step": 16574 }, { "epoch": 3.04, "learning_rate": 1.423588073889191e-05, "loss": 0.046, "step": 16575 }, { "epoch": 3.04, "learning_rate": 1.4235208193323787e-05, "loss": 0.0478, "step": 16576 }, { "epoch": 3.04, "learning_rate": 1.4234535624411096e-05, "loss": 0.0438, "step": 16577 }, { "epoch": 3.04, "learning_rate": 1.4233863032157546e-05, "loss": 0.0498, "step": 16578 }, { "epoch": 3.04, "learning_rate": 1.4233190416566845e-05, "loss": 0.0569, "step": 16579 }, { "epoch": 3.04, "learning_rate": 1.4232517777642697e-05, "loss": 0.0488, "step": 16580 }, { "epoch": 3.04, "learning_rate": 1.4231845115388815e-05, "loss": 0.0614, "step": 16581 }, { "epoch": 3.04, "learning_rate": 1.4231172429808899e-05, "loss": 0.0505, "step": 16582 }, { "epoch": 3.04, "learning_rate": 1.4230499720906666e-05, "loss": 0.0536, "step": 16583 }, { "epoch": 3.04, "learning_rate": 1.422982698868582e-05, "loss": 0.0474, "step": 16584 }, { "epoch": 3.04, "learning_rate": 1.4229154233150062e-05, "loss": 0.0529, "step": 16585 }, { "epoch": 3.04, "learning_rate": 1.4228481454303112e-05, "loss": 0.0521, "step": 16586 }, { "epoch": 3.04, "learning_rate": 1.4227808652148674e-05, "loss": 0.0494, "step": 16587 }, { "epoch": 3.04, "learning_rate": 1.4227135826690452e-05, "loss": 0.0583, "step": 16588 }, { "epoch": 3.04, "learning_rate": 1.4226462977932158e-05, "loss": 0.0559, "step": 16589 }, { "epoch": 3.04, "learning_rate": 1.4225790105877507e-05, "loss": 0.0536, "step": 16590 }, { "epoch": 3.04, "learning_rate": 1.4225117210530195e-05, "loss": 0.0492, "step": 16591 }, { "epoch": 3.04, "learning_rate": 1.4224444291893941e-05, "loss": 0.0447, "step": 16592 }, { "epoch": 3.04, "learning_rate": 1.422377134997245e-05, "loss": 0.0488, "step": 16593 }, { "epoch": 3.04, "learning_rate": 1.4223098384769437e-05, "loss": 0.0472, "step": 16594 }, { "epoch": 3.04, "learning_rate": 1.42224253962886e-05, "loss": 0.0415, "step": 16595 }, { "epoch": 3.04, "learning_rate": 1.422175238453366e-05, "loss": 0.043, "step": 16596 }, { "epoch": 3.04, "learning_rate": 1.422107934950832e-05, "loss": 0.0515, "step": 16597 }, { "epoch": 3.04, "learning_rate": 1.422040629121629e-05, "loss": 0.045, "step": 16598 }, { "epoch": 3.04, "learning_rate": 1.4219733209661289e-05, "loss": 0.0509, "step": 16599 }, { "epoch": 3.04, "learning_rate": 1.4219060104847012e-05, "loss": 0.0508, "step": 16600 }, { "epoch": 3.04, "learning_rate": 1.421838697677718e-05, "loss": 0.0533, "step": 16601 }, { "epoch": 3.04, "learning_rate": 1.42177138254555e-05, "loss": 0.0543, "step": 16602 }, { "epoch": 3.04, "learning_rate": 1.421704065088568e-05, "loss": 0.0534, "step": 16603 }, { "epoch": 3.04, "learning_rate": 1.4216367453071435e-05, "loss": 0.0465, "step": 16604 }, { "epoch": 3.05, "learning_rate": 1.4215694232016475e-05, "loss": 0.0427, "step": 16605 }, { "epoch": 3.05, "learning_rate": 1.421502098772451e-05, "loss": 0.0618, "step": 16606 }, { "epoch": 3.05, "learning_rate": 1.4214347720199246e-05, "loss": 0.0554, "step": 16607 }, { "epoch": 3.05, "learning_rate": 1.4213674429444406e-05, "loss": 0.0577, "step": 16608 }, { "epoch": 3.05, "learning_rate": 1.4213001115463688e-05, "loss": 0.0444, "step": 16609 }, { "epoch": 3.05, "learning_rate": 1.421232777826081e-05, "loss": 0.0422, "step": 16610 }, { "epoch": 3.05, "learning_rate": 1.4211654417839485e-05, "loss": 0.042, "step": 16611 }, { "epoch": 3.05, "learning_rate": 1.4210981034203419e-05, "loss": 0.053, "step": 16612 }, { "epoch": 3.05, "learning_rate": 1.421030762735633e-05, "loss": 0.0445, "step": 16613 }, { "epoch": 3.05, "learning_rate": 1.4209634197301925e-05, "loss": 0.0556, "step": 16614 }, { "epoch": 3.05, "learning_rate": 1.4208960744043917e-05, "loss": 0.0484, "step": 16615 }, { "epoch": 3.05, "learning_rate": 1.4208287267586021e-05, "loss": 0.0561, "step": 16616 }, { "epoch": 3.05, "learning_rate": 1.4207613767931947e-05, "loss": 0.0575, "step": 16617 }, { "epoch": 3.05, "learning_rate": 1.4206940245085403e-05, "loss": 0.0559, "step": 16618 }, { "epoch": 3.05, "learning_rate": 1.4206266699050112e-05, "loss": 0.0494, "step": 16619 }, { "epoch": 3.05, "learning_rate": 1.4205593129829775e-05, "loss": 0.0496, "step": 16620 }, { "epoch": 3.05, "learning_rate": 1.4204919537428114e-05, "loss": 0.0543, "step": 16621 }, { "epoch": 3.05, "learning_rate": 1.4204245921848837e-05, "loss": 0.0551, "step": 16622 }, { "epoch": 3.05, "learning_rate": 1.4203572283095657e-05, "loss": 0.0523, "step": 16623 }, { "epoch": 3.05, "learning_rate": 1.4202898621172288e-05, "loss": 0.0469, "step": 16624 }, { "epoch": 3.05, "learning_rate": 1.4202224936082443e-05, "loss": 0.0581, "step": 16625 }, { "epoch": 3.05, "learning_rate": 1.4201551227829837e-05, "loss": 0.0476, "step": 16626 }, { "epoch": 3.05, "learning_rate": 1.4200877496418182e-05, "loss": 0.0553, "step": 16627 }, { "epoch": 3.05, "learning_rate": 1.4200203741851191e-05, "loss": 0.0532, "step": 16628 }, { "epoch": 3.05, "learning_rate": 1.4199529964132582e-05, "loss": 0.0554, "step": 16629 }, { "epoch": 3.05, "learning_rate": 1.4198856163266062e-05, "loss": 0.048, "step": 16630 }, { "epoch": 3.05, "learning_rate": 1.4198182339255349e-05, "loss": 0.058, "step": 16631 }, { "epoch": 3.05, "learning_rate": 1.4197508492104154e-05, "loss": 0.0429, "step": 16632 }, { "epoch": 3.05, "learning_rate": 1.4196834621816198e-05, "loss": 0.0544, "step": 16633 }, { "epoch": 3.05, "learning_rate": 1.419616072839519e-05, "loss": 0.0492, "step": 16634 }, { "epoch": 3.05, "learning_rate": 1.4195486811844845e-05, "loss": 0.0523, "step": 16635 }, { "epoch": 3.05, "learning_rate": 1.419481287216888e-05, "loss": 0.0504, "step": 16636 }, { "epoch": 3.05, "learning_rate": 1.4194138909371007e-05, "loss": 0.0473, "step": 16637 }, { "epoch": 3.05, "learning_rate": 1.4193464923454942e-05, "loss": 0.0477, "step": 16638 }, { "epoch": 3.05, "learning_rate": 1.41927909144244e-05, "loss": 0.0556, "step": 16639 }, { "epoch": 3.05, "learning_rate": 1.4192116882283097e-05, "loss": 0.054, "step": 16640 }, { "epoch": 3.05, "learning_rate": 1.4191442827034748e-05, "loss": 0.0564, "step": 16641 }, { "epoch": 3.05, "learning_rate": 1.4190768748683063e-05, "loss": 0.0561, "step": 16642 }, { "epoch": 3.05, "learning_rate": 1.419009464723177e-05, "loss": 0.0521, "step": 16643 }, { "epoch": 3.05, "learning_rate": 1.4189420522684572e-05, "loss": 0.0486, "step": 16644 }, { "epoch": 3.05, "learning_rate": 1.4188746375045192e-05, "loss": 0.0609, "step": 16645 }, { "epoch": 3.05, "learning_rate": 1.4188072204317344e-05, "loss": 0.0533, "step": 16646 }, { "epoch": 3.05, "learning_rate": 1.4187398010504742e-05, "loss": 0.0571, "step": 16647 }, { "epoch": 3.05, "learning_rate": 1.4186723793611105e-05, "loss": 0.0462, "step": 16648 }, { "epoch": 3.05, "learning_rate": 1.4186049553640152e-05, "loss": 0.0554, "step": 16649 }, { "epoch": 3.05, "learning_rate": 1.4185375290595591e-05, "loss": 0.0561, "step": 16650 }, { "epoch": 3.05, "learning_rate": 1.4184701004481142e-05, "loss": 0.0475, "step": 16651 }, { "epoch": 3.05, "learning_rate": 1.418402669530053e-05, "loss": 0.0417, "step": 16652 }, { "epoch": 3.05, "learning_rate": 1.4183352363057458e-05, "loss": 0.0465, "step": 16653 }, { "epoch": 3.05, "learning_rate": 1.4182678007755653e-05, "loss": 0.0545, "step": 16654 }, { "epoch": 3.05, "learning_rate": 1.4182003629398828e-05, "loss": 0.0508, "step": 16655 }, { "epoch": 3.05, "learning_rate": 1.4181329227990702e-05, "loss": 0.054, "step": 16656 }, { "epoch": 3.05, "learning_rate": 1.418065480353499e-05, "loss": 0.056, "step": 16657 }, { "epoch": 3.05, "learning_rate": 1.4179980356035414e-05, "loss": 0.0498, "step": 16658 }, { "epoch": 3.06, "learning_rate": 1.4179305885495684e-05, "loss": 0.05, "step": 16659 }, { "epoch": 3.06, "learning_rate": 1.4178631391919524e-05, "loss": 0.0569, "step": 16660 }, { "epoch": 3.06, "learning_rate": 1.417795687531065e-05, "loss": 0.0522, "step": 16661 }, { "epoch": 3.06, "learning_rate": 1.417728233567278e-05, "loss": 0.0585, "step": 16662 }, { "epoch": 3.06, "learning_rate": 1.417660777300963e-05, "loss": 0.0478, "step": 16663 }, { "epoch": 3.06, "learning_rate": 1.4175933187324922e-05, "loss": 0.062, "step": 16664 }, { "epoch": 3.06, "learning_rate": 1.4175258578622373e-05, "loss": 0.0539, "step": 16665 }, { "epoch": 3.06, "learning_rate": 1.4174583946905697e-05, "loss": 0.0518, "step": 16666 }, { "epoch": 3.06, "learning_rate": 1.4173909292178622e-05, "loss": 0.0594, "step": 16667 }, { "epoch": 3.06, "learning_rate": 1.4173234614444857e-05, "loss": 0.0454, "step": 16668 }, { "epoch": 3.06, "learning_rate": 1.4172559913708127e-05, "loss": 0.0528, "step": 16669 }, { "epoch": 3.06, "learning_rate": 1.417188518997215e-05, "loss": 0.0626, "step": 16670 }, { "epoch": 3.06, "learning_rate": 1.4171210443240642e-05, "loss": 0.0579, "step": 16671 }, { "epoch": 3.06, "learning_rate": 1.4170535673517325e-05, "loss": 0.0472, "step": 16672 }, { "epoch": 3.06, "learning_rate": 1.416986088080592e-05, "loss": 0.0448, "step": 16673 }, { "epoch": 3.06, "learning_rate": 1.4169186065110142e-05, "loss": 0.0517, "step": 16674 }, { "epoch": 3.06, "learning_rate": 1.416851122643371e-05, "loss": 0.051, "step": 16675 }, { "epoch": 3.06, "learning_rate": 1.416783636478035e-05, "loss": 0.0562, "step": 16676 }, { "epoch": 3.06, "learning_rate": 1.4167161480153774e-05, "loss": 0.0504, "step": 16677 }, { "epoch": 3.06, "learning_rate": 1.4166486572557712e-05, "loss": 0.0491, "step": 16678 }, { "epoch": 3.06, "learning_rate": 1.4165811641995877e-05, "loss": 0.0565, "step": 16679 }, { "epoch": 3.06, "learning_rate": 1.416513668847199e-05, "loss": 0.0516, "step": 16680 }, { "epoch": 3.06, "learning_rate": 1.416446171198977e-05, "loss": 0.0532, "step": 16681 }, { "epoch": 3.06, "learning_rate": 1.4163786712552943e-05, "loss": 0.0523, "step": 16682 }, { "epoch": 3.06, "learning_rate": 1.4163111690165224e-05, "loss": 0.0496, "step": 16683 }, { "epoch": 3.06, "learning_rate": 1.4162436644830333e-05, "loss": 0.0508, "step": 16684 }, { "epoch": 3.06, "learning_rate": 1.4161761576552e-05, "loss": 0.0551, "step": 16685 }, { "epoch": 3.06, "learning_rate": 1.4161086485333935e-05, "loss": 0.0587, "step": 16686 }, { "epoch": 3.06, "learning_rate": 1.4160411371179866e-05, "loss": 0.0506, "step": 16687 }, { "epoch": 3.06, "learning_rate": 1.415973623409351e-05, "loss": 0.0487, "step": 16688 }, { "epoch": 3.06, "learning_rate": 1.4159061074078593e-05, "loss": 0.0505, "step": 16689 }, { "epoch": 3.06, "learning_rate": 1.4158385891138833e-05, "loss": 0.0488, "step": 16690 }, { "epoch": 3.06, "learning_rate": 1.4157710685277953e-05, "loss": 0.0524, "step": 16691 }, { "epoch": 3.06, "learning_rate": 1.4157035456499673e-05, "loss": 0.0468, "step": 16692 }, { "epoch": 3.06, "learning_rate": 1.4156360204807716e-05, "loss": 0.0473, "step": 16693 }, { "epoch": 3.06, "learning_rate": 1.4155684930205806e-05, "loss": 0.045, "step": 16694 }, { "epoch": 3.06, "learning_rate": 1.415500963269766e-05, "loss": 0.0511, "step": 16695 }, { "epoch": 3.06, "learning_rate": 1.4154334312287006e-05, "loss": 0.0455, "step": 16696 }, { "epoch": 3.06, "learning_rate": 1.4153658968977563e-05, "loss": 0.0553, "step": 16697 }, { "epoch": 3.06, "learning_rate": 1.4152983602773058e-05, "loss": 0.0492, "step": 16698 }, { "epoch": 3.06, "learning_rate": 1.4152308213677205e-05, "loss": 0.0466, "step": 16699 }, { "epoch": 3.06, "learning_rate": 1.4151632801693736e-05, "loss": 0.0429, "step": 16700 }, { "epoch": 3.06, "learning_rate": 1.4150957366826366e-05, "loss": 0.0537, "step": 16701 }, { "epoch": 3.06, "learning_rate": 1.4150281909078824e-05, "loss": 0.0497, "step": 16702 }, { "epoch": 3.06, "learning_rate": 1.4149606428454832e-05, "loss": 0.0506, "step": 16703 }, { "epoch": 3.06, "learning_rate": 1.4148930924958108e-05, "loss": 0.0578, "step": 16704 }, { "epoch": 3.06, "learning_rate": 1.4148255398592383e-05, "loss": 0.0516, "step": 16705 }, { "epoch": 3.06, "learning_rate": 1.4147579849361377e-05, "loss": 0.0554, "step": 16706 }, { "epoch": 3.06, "learning_rate": 1.4146904277268813e-05, "loss": 0.0499, "step": 16707 }, { "epoch": 3.06, "learning_rate": 1.4146228682318417e-05, "loss": 0.0433, "step": 16708 }, { "epoch": 3.06, "learning_rate": 1.4145553064513909e-05, "loss": 0.057, "step": 16709 }, { "epoch": 3.06, "learning_rate": 1.4144877423859013e-05, "loss": 0.0466, "step": 16710 }, { "epoch": 3.06, "learning_rate": 1.4144201760357464e-05, "loss": 0.0586, "step": 16711 }, { "epoch": 3.06, "learning_rate": 1.414352607401297e-05, "loss": 0.0522, "step": 16712 }, { "epoch": 3.06, "learning_rate": 1.4142850364829265e-05, "loss": 0.0446, "step": 16713 }, { "epoch": 3.07, "learning_rate": 1.4142174632810074e-05, "loss": 0.0437, "step": 16714 }, { "epoch": 3.07, "learning_rate": 1.4141498877959118e-05, "loss": 0.0457, "step": 16715 }, { "epoch": 3.07, "learning_rate": 1.414082310028012e-05, "loss": 0.0511, "step": 16716 }, { "epoch": 3.07, "learning_rate": 1.4140147299776814e-05, "loss": 0.0587, "step": 16717 }, { "epoch": 3.07, "learning_rate": 1.4139471476452915e-05, "loss": 0.0559, "step": 16718 }, { "epoch": 3.07, "learning_rate": 1.4138795630312152e-05, "loss": 0.0551, "step": 16719 }, { "epoch": 3.07, "learning_rate": 1.4138119761358253e-05, "loss": 0.0494, "step": 16720 }, { "epoch": 3.07, "learning_rate": 1.4137443869594937e-05, "loss": 0.0567, "step": 16721 }, { "epoch": 3.07, "learning_rate": 1.4136767955025938e-05, "loss": 0.0495, "step": 16722 }, { "epoch": 3.07, "learning_rate": 1.4136092017654977e-05, "loss": 0.059, "step": 16723 }, { "epoch": 3.07, "learning_rate": 1.4135416057485777e-05, "loss": 0.0449, "step": 16724 }, { "epoch": 3.07, "learning_rate": 1.4134740074522069e-05, "loss": 0.0488, "step": 16725 }, { "epoch": 3.07, "learning_rate": 1.4134064068767577e-05, "loss": 0.0528, "step": 16726 }, { "epoch": 3.07, "learning_rate": 1.4133388040226026e-05, "loss": 0.0545, "step": 16727 }, { "epoch": 3.07, "learning_rate": 1.4132711988901142e-05, "loss": 0.0466, "step": 16728 }, { "epoch": 3.07, "learning_rate": 1.4132035914796656e-05, "loss": 0.052, "step": 16729 }, { "epoch": 3.07, "learning_rate": 1.4131359817916286e-05, "loss": 0.054, "step": 16730 }, { "epoch": 3.07, "learning_rate": 1.4130683698263767e-05, "loss": 0.0536, "step": 16731 }, { "epoch": 3.07, "learning_rate": 1.4130007555842825e-05, "loss": 0.0473, "step": 16732 }, { "epoch": 3.07, "learning_rate": 1.412933139065718e-05, "loss": 0.0494, "step": 16733 }, { "epoch": 3.07, "learning_rate": 1.4128655202710569e-05, "loss": 0.0538, "step": 16734 }, { "epoch": 3.07, "learning_rate": 1.412797899200671e-05, "loss": 0.048, "step": 16735 }, { "epoch": 3.07, "learning_rate": 1.4127302758549334e-05, "loss": 0.0531, "step": 16736 }, { "epoch": 3.07, "learning_rate": 1.4126626502342168e-05, "loss": 0.0484, "step": 16737 }, { "epoch": 3.07, "learning_rate": 1.4125950223388944e-05, "loss": 0.0569, "step": 16738 }, { "epoch": 3.07, "learning_rate": 1.412527392169338e-05, "loss": 0.0597, "step": 16739 }, { "epoch": 3.07, "learning_rate": 1.4124597597259215e-05, "loss": 0.0535, "step": 16740 }, { "epoch": 3.07, "learning_rate": 1.4123921250090169e-05, "loss": 0.0444, "step": 16741 }, { "epoch": 3.07, "learning_rate": 1.4123244880189972e-05, "loss": 0.0553, "step": 16742 }, { "epoch": 3.07, "learning_rate": 1.4122568487562351e-05, "loss": 0.0596, "step": 16743 }, { "epoch": 3.07, "learning_rate": 1.412189207221104e-05, "loss": 0.0494, "step": 16744 }, { "epoch": 3.07, "learning_rate": 1.4121215634139761e-05, "loss": 0.049, "step": 16745 }, { "epoch": 3.07, "learning_rate": 1.4120539173352245e-05, "loss": 0.0552, "step": 16746 }, { "epoch": 3.07, "learning_rate": 1.4119862689852224e-05, "loss": 0.0561, "step": 16747 }, { "epoch": 3.07, "learning_rate": 1.4119186183643416e-05, "loss": 0.0527, "step": 16748 }, { "epoch": 3.07, "learning_rate": 1.4118509654729561e-05, "loss": 0.0535, "step": 16749 }, { "epoch": 3.07, "learning_rate": 1.4117833103114387e-05, "loss": 0.0485, "step": 16750 }, { "epoch": 3.07, "learning_rate": 1.411715652880162e-05, "loss": 0.0543, "step": 16751 }, { "epoch": 3.07, "learning_rate": 1.4116479931794984e-05, "loss": 0.054, "step": 16752 }, { "epoch": 3.07, "learning_rate": 1.411580331209822e-05, "loss": 0.05, "step": 16753 }, { "epoch": 3.07, "learning_rate": 1.4115126669715049e-05, "loss": 0.0472, "step": 16754 }, { "epoch": 3.07, "learning_rate": 1.4114450004649203e-05, "loss": 0.053, "step": 16755 }, { "epoch": 3.07, "learning_rate": 1.4113773316904417e-05, "loss": 0.0523, "step": 16756 }, { "epoch": 3.07, "learning_rate": 1.411309660648441e-05, "loss": 0.0461, "step": 16757 }, { "epoch": 3.07, "learning_rate": 1.4112419873392922e-05, "loss": 0.058, "step": 16758 }, { "epoch": 3.07, "learning_rate": 1.4111743117633676e-05, "loss": 0.049, "step": 16759 }, { "epoch": 3.07, "learning_rate": 1.4111066339210408e-05, "loss": 0.0541, "step": 16760 }, { "epoch": 3.07, "learning_rate": 1.4110389538126844e-05, "loss": 0.0578, "step": 16761 }, { "epoch": 3.07, "learning_rate": 1.4109712714386721e-05, "loss": 0.046, "step": 16762 }, { "epoch": 3.07, "learning_rate": 1.410903586799376e-05, "loss": 0.0534, "step": 16763 }, { "epoch": 3.07, "learning_rate": 1.4108358998951702e-05, "loss": 0.0512, "step": 16764 }, { "epoch": 3.07, "learning_rate": 1.410768210726427e-05, "loss": 0.0438, "step": 16765 }, { "epoch": 3.07, "learning_rate": 1.4107005192935197e-05, "loss": 0.053, "step": 16766 }, { "epoch": 3.07, "learning_rate": 1.4106328255968218e-05, "loss": 0.0526, "step": 16767 }, { "epoch": 3.08, "learning_rate": 1.410565129636706e-05, "loss": 0.0525, "step": 16768 }, { "epoch": 3.08, "learning_rate": 1.4104974314135455e-05, "loss": 0.0498, "step": 16769 }, { "epoch": 3.08, "learning_rate": 1.4104297309277135e-05, "loss": 0.0627, "step": 16770 }, { "epoch": 3.08, "learning_rate": 1.4103620281795835e-05, "loss": 0.0535, "step": 16771 }, { "epoch": 3.08, "learning_rate": 1.4102943231695281e-05, "loss": 0.051, "step": 16772 }, { "epoch": 3.08, "learning_rate": 1.410226615897921e-05, "loss": 0.0471, "step": 16773 }, { "epoch": 3.08, "learning_rate": 1.410158906365135e-05, "loss": 0.0506, "step": 16774 }, { "epoch": 3.08, "learning_rate": 1.4100911945715436e-05, "loss": 0.048, "step": 16775 }, { "epoch": 3.08, "learning_rate": 1.4100234805175199e-05, "loss": 0.0458, "step": 16776 }, { "epoch": 3.08, "learning_rate": 1.4099557642034371e-05, "loss": 0.0487, "step": 16777 }, { "epoch": 3.08, "learning_rate": 1.4098880456296686e-05, "loss": 0.0543, "step": 16778 }, { "epoch": 3.08, "learning_rate": 1.4098203247965876e-05, "loss": 0.0455, "step": 16779 }, { "epoch": 3.08, "learning_rate": 1.4097526017045675e-05, "loss": 0.0512, "step": 16780 }, { "epoch": 3.08, "learning_rate": 1.4096848763539812e-05, "loss": 0.0534, "step": 16781 }, { "epoch": 3.08, "learning_rate": 1.4096171487452024e-05, "loss": 0.0553, "step": 16782 }, { "epoch": 3.08, "learning_rate": 1.4095494188786044e-05, "loss": 0.0488, "step": 16783 }, { "epoch": 3.08, "learning_rate": 1.4094816867545601e-05, "loss": 0.0502, "step": 16784 }, { "epoch": 3.08, "learning_rate": 1.4094139523734433e-05, "loss": 0.0492, "step": 16785 }, { "epoch": 3.08, "learning_rate": 1.4093462157356274e-05, "loss": 0.0519, "step": 16786 }, { "epoch": 3.08, "learning_rate": 1.4092784768414854e-05, "loss": 0.0423, "step": 16787 }, { "epoch": 3.08, "learning_rate": 1.4092107356913908e-05, "loss": 0.0452, "step": 16788 }, { "epoch": 3.08, "learning_rate": 1.4091429922857171e-05, "loss": 0.0549, "step": 16789 }, { "epoch": 3.08, "learning_rate": 1.4090752466248374e-05, "loss": 0.0542, "step": 16790 }, { "epoch": 3.08, "learning_rate": 1.4090074987091256e-05, "loss": 0.0483, "step": 16791 }, { "epoch": 3.08, "learning_rate": 1.4089397485389548e-05, "loss": 0.0547, "step": 16792 }, { "epoch": 3.08, "learning_rate": 1.4088719961146988e-05, "loss": 0.0543, "step": 16793 }, { "epoch": 3.08, "learning_rate": 1.4088042414367304e-05, "loss": 0.0435, "step": 16794 }, { "epoch": 3.08, "learning_rate": 1.4087364845054236e-05, "loss": 0.0488, "step": 16795 }, { "epoch": 3.08, "learning_rate": 1.4086687253211517e-05, "loss": 0.0472, "step": 16796 }, { "epoch": 3.08, "learning_rate": 1.4086009638842884e-05, "loss": 0.0459, "step": 16797 }, { "epoch": 3.08, "learning_rate": 1.408533200195207e-05, "loss": 0.0491, "step": 16798 }, { "epoch": 3.08, "learning_rate": 1.4084654342542807e-05, "loss": 0.0492, "step": 16799 }, { "epoch": 3.08, "learning_rate": 1.4083976660618834e-05, "loss": 0.053, "step": 16800 }, { "epoch": 3.08, "learning_rate": 1.4083298956183889e-05, "loss": 0.0555, "step": 16801 }, { "epoch": 3.08, "learning_rate": 1.4082621229241702e-05, "loss": 0.0444, "step": 16802 }, { "epoch": 3.08, "learning_rate": 1.4081943479796012e-05, "loss": 0.046, "step": 16803 }, { "epoch": 3.08, "learning_rate": 1.4081265707850556e-05, "loss": 0.0598, "step": 16804 }, { "epoch": 3.08, "learning_rate": 1.4080587913409063e-05, "loss": 0.0514, "step": 16805 }, { "epoch": 3.08, "learning_rate": 1.407991009647528e-05, "loss": 0.0477, "step": 16806 }, { "epoch": 3.08, "learning_rate": 1.407923225705293e-05, "loss": 0.0577, "step": 16807 }, { "epoch": 3.08, "learning_rate": 1.4078554395145761e-05, "loss": 0.0502, "step": 16808 }, { "epoch": 3.08, "learning_rate": 1.4077876510757502e-05, "loss": 0.0503, "step": 16809 }, { "epoch": 3.08, "learning_rate": 1.4077198603891895e-05, "loss": 0.0538, "step": 16810 }, { "epoch": 3.08, "learning_rate": 1.4076520674552675e-05, "loss": 0.0434, "step": 16811 }, { "epoch": 3.08, "learning_rate": 1.4075842722743573e-05, "loss": 0.0506, "step": 16812 }, { "epoch": 3.08, "learning_rate": 1.4075164748468333e-05, "loss": 0.0546, "step": 16813 }, { "epoch": 3.08, "learning_rate": 1.4074486751730687e-05, "loss": 0.0491, "step": 16814 }, { "epoch": 3.08, "learning_rate": 1.4073808732534377e-05, "loss": 0.0499, "step": 16815 }, { "epoch": 3.08, "learning_rate": 1.4073130690883136e-05, "loss": 0.0574, "step": 16816 }, { "epoch": 3.08, "learning_rate": 1.4072452626780705e-05, "loss": 0.0544, "step": 16817 }, { "epoch": 3.08, "learning_rate": 1.4071774540230817e-05, "loss": 0.0473, "step": 16818 }, { "epoch": 3.08, "learning_rate": 1.4071096431237217e-05, "loss": 0.0489, "step": 16819 }, { "epoch": 3.08, "learning_rate": 1.4070418299803636e-05, "loss": 0.0557, "step": 16820 }, { "epoch": 3.08, "learning_rate": 1.4069740145933812e-05, "loss": 0.0547, "step": 16821 }, { "epoch": 3.08, "learning_rate": 1.4069061969631489e-05, "loss": 0.0509, "step": 16822 }, { "epoch": 3.09, "learning_rate": 1.4068383770900395e-05, "loss": 0.0526, "step": 16823 }, { "epoch": 3.09, "learning_rate": 1.406770554974428e-05, "loss": 0.0431, "step": 16824 }, { "epoch": 3.09, "learning_rate": 1.4067027306166874e-05, "loss": 0.0373, "step": 16825 }, { "epoch": 3.09, "learning_rate": 1.4066349040171918e-05, "loss": 0.0456, "step": 16826 }, { "epoch": 3.09, "learning_rate": 1.4065670751763151e-05, "loss": 0.051, "step": 16827 }, { "epoch": 3.09, "learning_rate": 1.4064992440944314e-05, "loss": 0.0438, "step": 16828 }, { "epoch": 3.09, "learning_rate": 1.4064314107719138e-05, "loss": 0.052, "step": 16829 }, { "epoch": 3.09, "learning_rate": 1.4063635752091373e-05, "loss": 0.0504, "step": 16830 }, { "epoch": 3.09, "learning_rate": 1.4062957374064752e-05, "loss": 0.0431, "step": 16831 }, { "epoch": 3.09, "learning_rate": 1.4062278973643011e-05, "loss": 0.0492, "step": 16832 }, { "epoch": 3.09, "learning_rate": 1.4061600550829897e-05, "loss": 0.055, "step": 16833 }, { "epoch": 3.09, "learning_rate": 1.4060922105629145e-05, "loss": 0.0504, "step": 16834 }, { "epoch": 3.09, "learning_rate": 1.4060243638044494e-05, "loss": 0.0559, "step": 16835 }, { "epoch": 3.09, "learning_rate": 1.4059565148079687e-05, "loss": 0.0441, "step": 16836 }, { "epoch": 3.09, "learning_rate": 1.4058886635738458e-05, "loss": 0.0527, "step": 16837 }, { "epoch": 3.09, "learning_rate": 1.4058208101024551e-05, "loss": 0.0422, "step": 16838 }, { "epoch": 3.09, "learning_rate": 1.405752954394171e-05, "loss": 0.0558, "step": 16839 }, { "epoch": 3.09, "learning_rate": 1.4056850964493668e-05, "loss": 0.0455, "step": 16840 }, { "epoch": 3.09, "learning_rate": 1.4056172362684168e-05, "loss": 0.0503, "step": 16841 }, { "epoch": 3.09, "learning_rate": 1.4055493738516955e-05, "loss": 0.0566, "step": 16842 }, { "epoch": 3.09, "learning_rate": 1.405481509199576e-05, "loss": 0.0495, "step": 16843 }, { "epoch": 3.09, "learning_rate": 1.4054136423124331e-05, "loss": 0.0611, "step": 16844 }, { "epoch": 3.09, "learning_rate": 1.405345773190641e-05, "loss": 0.0493, "step": 16845 }, { "epoch": 3.09, "learning_rate": 1.4052779018345732e-05, "loss": 0.0492, "step": 16846 }, { "epoch": 3.09, "learning_rate": 1.4052100282446039e-05, "loss": 0.0531, "step": 16847 }, { "epoch": 3.09, "learning_rate": 1.4051421524211081e-05, "loss": 0.0526, "step": 16848 }, { "epoch": 3.09, "learning_rate": 1.4050742743644588e-05, "loss": 0.057, "step": 16849 }, { "epoch": 3.09, "learning_rate": 1.4050063940750307e-05, "loss": 0.0488, "step": 16850 }, { "epoch": 3.09, "learning_rate": 1.404938511553198e-05, "loss": 0.0574, "step": 16851 }, { "epoch": 3.09, "learning_rate": 1.4048706267993347e-05, "loss": 0.0543, "step": 16852 }, { "epoch": 3.09, "learning_rate": 1.4048027398138149e-05, "loss": 0.0443, "step": 16853 }, { "epoch": 3.09, "learning_rate": 1.4047348505970128e-05, "loss": 0.0558, "step": 16854 }, { "epoch": 3.09, "learning_rate": 1.404666959149303e-05, "loss": 0.0467, "step": 16855 }, { "epoch": 3.09, "learning_rate": 1.4045990654710591e-05, "loss": 0.0536, "step": 16856 }, { "epoch": 3.09, "learning_rate": 1.404531169562656e-05, "loss": 0.0523, "step": 16857 }, { "epoch": 3.09, "learning_rate": 1.4044632714244673e-05, "loss": 0.056, "step": 16858 }, { "epoch": 3.09, "learning_rate": 1.4043953710568678e-05, "loss": 0.0432, "step": 16859 }, { "epoch": 3.09, "learning_rate": 1.4043274684602315e-05, "loss": 0.0533, "step": 16860 }, { "epoch": 3.09, "learning_rate": 1.4042595636349327e-05, "loss": 0.0511, "step": 16861 }, { "epoch": 3.09, "learning_rate": 1.4041916565813457e-05, "loss": 0.0518, "step": 16862 }, { "epoch": 3.09, "learning_rate": 1.4041237472998451e-05, "loss": 0.0502, "step": 16863 }, { "epoch": 3.09, "learning_rate": 1.4040558357908047e-05, "loss": 0.053, "step": 16864 }, { "epoch": 3.09, "learning_rate": 1.403987922054599e-05, "loss": 0.0589, "step": 16865 }, { "epoch": 3.09, "learning_rate": 1.4039200060916027e-05, "loss": 0.0572, "step": 16866 }, { "epoch": 3.09, "learning_rate": 1.4038520879021893e-05, "loss": 0.0527, "step": 16867 }, { "epoch": 3.09, "learning_rate": 1.4037841674867343e-05, "loss": 0.0545, "step": 16868 }, { "epoch": 3.09, "learning_rate": 1.4037162448456111e-05, "loss": 0.0518, "step": 16869 }, { "epoch": 3.09, "learning_rate": 1.4036483199791949e-05, "loss": 0.052, "step": 16870 }, { "epoch": 3.09, "learning_rate": 1.4035803928878595e-05, "loss": 0.0521, "step": 16871 }, { "epoch": 3.09, "learning_rate": 1.4035124635719795e-05, "loss": 0.049, "step": 16872 }, { "epoch": 3.09, "learning_rate": 1.4034445320319292e-05, "loss": 0.046, "step": 16873 }, { "epoch": 3.09, "learning_rate": 1.4033765982680834e-05, "loss": 0.0488, "step": 16874 }, { "epoch": 3.09, "learning_rate": 1.4033086622808164e-05, "loss": 0.0479, "step": 16875 }, { "epoch": 3.09, "learning_rate": 1.403240724070502e-05, "loss": 0.0489, "step": 16876 }, { "epoch": 3.09, "learning_rate": 1.4031727836375159e-05, "loss": 0.0454, "step": 16877 }, { "epoch": 3.1, "learning_rate": 1.403104840982232e-05, "loss": 0.0485, "step": 16878 }, { "epoch": 3.1, "learning_rate": 1.4030368961050248e-05, "loss": 0.0524, "step": 16879 }, { "epoch": 3.1, "learning_rate": 1.4029689490062686e-05, "loss": 0.0473, "step": 16880 }, { "epoch": 3.1, "learning_rate": 1.402900999686338e-05, "loss": 0.0458, "step": 16881 }, { "epoch": 3.1, "learning_rate": 1.4028330481456076e-05, "loss": 0.0523, "step": 16882 }, { "epoch": 3.1, "learning_rate": 1.4027650943844525e-05, "loss": 0.0531, "step": 16883 }, { "epoch": 3.1, "learning_rate": 1.4026971384032463e-05, "loss": 0.0499, "step": 16884 }, { "epoch": 3.1, "learning_rate": 1.402629180202364e-05, "loss": 0.0478, "step": 16885 }, { "epoch": 3.1, "learning_rate": 1.4025612197821805e-05, "loss": 0.0441, "step": 16886 }, { "epoch": 3.1, "learning_rate": 1.4024932571430701e-05, "loss": 0.0514, "step": 16887 }, { "epoch": 3.1, "learning_rate": 1.402425292285407e-05, "loss": 0.0465, "step": 16888 }, { "epoch": 3.1, "learning_rate": 1.4023573252095669e-05, "loss": 0.0442, "step": 16889 }, { "epoch": 3.1, "learning_rate": 1.4022893559159237e-05, "loss": 0.0535, "step": 16890 }, { "epoch": 3.1, "learning_rate": 1.4022213844048516e-05, "loss": 0.0467, "step": 16891 }, { "epoch": 3.1, "learning_rate": 1.4021534106767262e-05, "loss": 0.0452, "step": 16892 }, { "epoch": 3.1, "learning_rate": 1.4020854347319217e-05, "loss": 0.0566, "step": 16893 }, { "epoch": 3.1, "learning_rate": 1.402017456570813e-05, "loss": 0.0472, "step": 16894 }, { "epoch": 3.1, "learning_rate": 1.4019494761937746e-05, "loss": 0.0477, "step": 16895 }, { "epoch": 3.1, "learning_rate": 1.4018814936011814e-05, "loss": 0.0447, "step": 16896 }, { "epoch": 3.1, "learning_rate": 1.4018135087934078e-05, "loss": 0.0518, "step": 16897 }, { "epoch": 3.1, "learning_rate": 1.4017455217708287e-05, "loss": 0.0558, "step": 16898 }, { "epoch": 3.1, "learning_rate": 1.401677532533819e-05, "loss": 0.0575, "step": 16899 }, { "epoch": 3.1, "learning_rate": 1.401609541082753e-05, "loss": 0.0515, "step": 16900 }, { "epoch": 3.1, "learning_rate": 1.4015415474180063e-05, "loss": 0.0511, "step": 16901 }, { "epoch": 3.1, "learning_rate": 1.4014735515399527e-05, "loss": 0.0457, "step": 16902 }, { "epoch": 3.1, "learning_rate": 1.4014055534489679e-05, "loss": 0.0563, "step": 16903 }, { "epoch": 3.1, "learning_rate": 1.4013375531454264e-05, "loss": 0.0473, "step": 16904 }, { "epoch": 3.1, "learning_rate": 1.4012695506297022e-05, "loss": 0.0527, "step": 16905 }, { "epoch": 3.1, "learning_rate": 1.4012015459021715e-05, "loss": 0.05, "step": 16906 }, { "epoch": 3.1, "learning_rate": 1.4011335389632081e-05, "loss": 0.0458, "step": 16907 }, { "epoch": 3.1, "learning_rate": 1.4010655298131875e-05, "loss": 0.0637, "step": 16908 }, { "epoch": 3.1, "learning_rate": 1.400997518452484e-05, "loss": 0.0446, "step": 16909 }, { "epoch": 3.1, "learning_rate": 1.400929504881473e-05, "loss": 0.0477, "step": 16910 }, { "epoch": 3.1, "learning_rate": 1.400861489100529e-05, "loss": 0.053, "step": 16911 }, { "epoch": 3.1, "learning_rate": 1.4007934711100273e-05, "loss": 0.0501, "step": 16912 }, { "epoch": 3.1, "learning_rate": 1.4007254509103425e-05, "loss": 0.0446, "step": 16913 }, { "epoch": 3.1, "learning_rate": 1.4006574285018497e-05, "loss": 0.0484, "step": 16914 }, { "epoch": 3.1, "learning_rate": 1.4005894038849237e-05, "loss": 0.0497, "step": 16915 }, { "epoch": 3.1, "learning_rate": 1.4005213770599394e-05, "loss": 0.0473, "step": 16916 }, { "epoch": 3.1, "learning_rate": 1.4004533480272722e-05, "loss": 0.0591, "step": 16917 }, { "epoch": 3.1, "learning_rate": 1.4003853167872963e-05, "loss": 0.0521, "step": 16918 }, { "epoch": 3.1, "learning_rate": 1.4003172833403877e-05, "loss": 0.0434, "step": 16919 }, { "epoch": 3.1, "learning_rate": 1.4002492476869205e-05, "loss": 0.0432, "step": 16920 }, { "epoch": 3.1, "learning_rate": 1.40018120982727e-05, "loss": 0.0525, "step": 16921 }, { "epoch": 3.1, "learning_rate": 1.4001131697618116e-05, "loss": 0.0463, "step": 16922 }, { "epoch": 3.1, "learning_rate": 1.4000451274909198e-05, "loss": 0.047, "step": 16923 }, { "epoch": 3.1, "learning_rate": 1.3999770830149699e-05, "loss": 0.0479, "step": 16924 }, { "epoch": 3.1, "learning_rate": 1.399909036334337e-05, "loss": 0.059, "step": 16925 }, { "epoch": 3.1, "learning_rate": 1.3998409874493963e-05, "loss": 0.0499, "step": 16926 }, { "epoch": 3.1, "learning_rate": 1.3997729363605224e-05, "loss": 0.0453, "step": 16927 }, { "epoch": 3.1, "learning_rate": 1.399704883068091e-05, "loss": 0.0503, "step": 16928 }, { "epoch": 3.1, "learning_rate": 1.3996368275724767e-05, "loss": 0.05, "step": 16929 }, { "epoch": 3.1, "learning_rate": 1.399568769874055e-05, "loss": 0.0514, "step": 16930 }, { "epoch": 3.1, "learning_rate": 1.3995007099732007e-05, "loss": 0.0421, "step": 16931 }, { "epoch": 3.11, "learning_rate": 1.3994326478702893e-05, "loss": 0.045, "step": 16932 }, { "epoch": 3.11, "learning_rate": 1.3993645835656955e-05, "loss": 0.0545, "step": 16933 }, { "epoch": 3.11, "learning_rate": 1.3992965170597952e-05, "loss": 0.0544, "step": 16934 }, { "epoch": 3.11, "learning_rate": 1.3992284483529628e-05, "loss": 0.0526, "step": 16935 }, { "epoch": 3.11, "learning_rate": 1.399160377445574e-05, "loss": 0.0565, "step": 16936 }, { "epoch": 3.11, "learning_rate": 1.399092304338004e-05, "loss": 0.0494, "step": 16937 }, { "epoch": 3.11, "learning_rate": 1.3990242290306275e-05, "loss": 0.0499, "step": 16938 }, { "epoch": 3.11, "learning_rate": 1.3989561515238204e-05, "loss": 0.0455, "step": 16939 }, { "epoch": 3.11, "learning_rate": 1.3988880718179574e-05, "loss": 0.0506, "step": 16940 }, { "epoch": 3.11, "learning_rate": 1.3988199899134142e-05, "loss": 0.0433, "step": 16941 }, { "epoch": 3.11, "learning_rate": 1.3987519058105657e-05, "loss": 0.0453, "step": 16942 }, { "epoch": 3.11, "learning_rate": 1.3986838195097877e-05, "loss": 0.0545, "step": 16943 }, { "epoch": 3.11, "learning_rate": 1.3986157310114545e-05, "loss": 0.0532, "step": 16944 }, { "epoch": 3.11, "learning_rate": 1.3985476403159425e-05, "loss": 0.0458, "step": 16945 }, { "epoch": 3.11, "learning_rate": 1.3984795474236266e-05, "loss": 0.056, "step": 16946 }, { "epoch": 3.11, "learning_rate": 1.3984114523348819e-05, "loss": 0.0504, "step": 16947 }, { "epoch": 3.11, "learning_rate": 1.3983433550500842e-05, "loss": 0.0549, "step": 16948 }, { "epoch": 3.11, "learning_rate": 1.3982752555696087e-05, "loss": 0.0478, "step": 16949 }, { "epoch": 3.11, "learning_rate": 1.3982071538938303e-05, "loss": 0.0457, "step": 16950 }, { "epoch": 3.11, "learning_rate": 1.3981390500231249e-05, "loss": 0.0482, "step": 16951 }, { "epoch": 3.11, "learning_rate": 1.3980709439578678e-05, "loss": 0.0455, "step": 16952 }, { "epoch": 3.11, "learning_rate": 1.398002835698434e-05, "loss": 0.0471, "step": 16953 }, { "epoch": 3.11, "learning_rate": 1.3979347252451994e-05, "loss": 0.0479, "step": 16954 }, { "epoch": 3.11, "learning_rate": 1.3978666125985394e-05, "loss": 0.0519, "step": 16955 }, { "epoch": 3.11, "learning_rate": 1.3977984977588292e-05, "loss": 0.0463, "step": 16956 }, { "epoch": 3.11, "learning_rate": 1.3977303807264444e-05, "loss": 0.048, "step": 16957 }, { "epoch": 3.11, "learning_rate": 1.3976622615017605e-05, "loss": 0.0553, "step": 16958 }, { "epoch": 3.11, "learning_rate": 1.3975941400851529e-05, "loss": 0.0496, "step": 16959 }, { "epoch": 3.11, "learning_rate": 1.397526016476997e-05, "loss": 0.0573, "step": 16960 }, { "epoch": 3.11, "learning_rate": 1.3974578906776683e-05, "loss": 0.0519, "step": 16961 }, { "epoch": 3.11, "learning_rate": 1.3973897626875424e-05, "loss": 0.0494, "step": 16962 }, { "epoch": 3.11, "learning_rate": 1.397321632506995e-05, "loss": 0.0395, "step": 16963 }, { "epoch": 3.11, "learning_rate": 1.3972535001364014e-05, "loss": 0.0585, "step": 16964 }, { "epoch": 3.11, "learning_rate": 1.3971853655761372e-05, "loss": 0.058, "step": 16965 }, { "epoch": 3.11, "learning_rate": 1.3971172288265779e-05, "loss": 0.0556, "step": 16966 }, { "epoch": 3.11, "learning_rate": 1.3970490898880991e-05, "loss": 0.0487, "step": 16967 }, { "epoch": 3.11, "learning_rate": 1.3969809487610767e-05, "loss": 0.0503, "step": 16968 }, { "epoch": 3.11, "learning_rate": 1.3969128054458857e-05, "loss": 0.055, "step": 16969 }, { "epoch": 3.11, "learning_rate": 1.3968446599429019e-05, "loss": 0.0503, "step": 16970 }, { "epoch": 3.11, "learning_rate": 1.3967765122525013e-05, "loss": 0.0473, "step": 16971 }, { "epoch": 3.11, "learning_rate": 1.3967083623750592e-05, "loss": 0.0426, "step": 16972 }, { "epoch": 3.11, "learning_rate": 1.3966402103109513e-05, "loss": 0.0488, "step": 16973 }, { "epoch": 3.11, "learning_rate": 1.3965720560605533e-05, "loss": 0.0557, "step": 16974 }, { "epoch": 3.11, "learning_rate": 1.3965038996242409e-05, "loss": 0.052, "step": 16975 }, { "epoch": 3.11, "learning_rate": 1.3964357410023896e-05, "loss": 0.0444, "step": 16976 }, { "epoch": 3.11, "learning_rate": 1.396367580195375e-05, "loss": 0.0453, "step": 16977 }, { "epoch": 3.11, "learning_rate": 1.3962994172035736e-05, "loss": 0.0487, "step": 16978 }, { "epoch": 3.11, "learning_rate": 1.3962312520273602e-05, "loss": 0.0616, "step": 16979 }, { "epoch": 3.11, "learning_rate": 1.3961630846671106e-05, "loss": 0.0459, "step": 16980 }, { "epoch": 3.11, "learning_rate": 1.3960949151232013e-05, "loss": 0.0557, "step": 16981 }, { "epoch": 3.11, "learning_rate": 1.396026743396007e-05, "loss": 0.0506, "step": 16982 }, { "epoch": 3.11, "learning_rate": 1.3959585694859043e-05, "loss": 0.0464, "step": 16983 }, { "epoch": 3.11, "learning_rate": 1.3958903933932688e-05, "loss": 0.0471, "step": 16984 }, { "epoch": 3.11, "learning_rate": 1.3958222151184759e-05, "loss": 0.0502, "step": 16985 }, { "epoch": 3.11, "learning_rate": 1.3957540346619019e-05, "loss": 0.0514, "step": 16986 }, { "epoch": 3.12, "learning_rate": 1.3956858520239225e-05, "loss": 0.0509, "step": 16987 }, { "epoch": 3.12, "learning_rate": 1.3956176672049132e-05, "loss": 0.0481, "step": 16988 }, { "epoch": 3.12, "learning_rate": 1.3955494802052498e-05, "loss": 0.0512, "step": 16989 }, { "epoch": 3.12, "learning_rate": 1.3954812910253091e-05, "loss": 0.0525, "step": 16990 }, { "epoch": 3.12, "learning_rate": 1.3954130996654658e-05, "loss": 0.0535, "step": 16991 }, { "epoch": 3.12, "learning_rate": 1.3953449061260962e-05, "loss": 0.0578, "step": 16992 }, { "epoch": 3.12, "learning_rate": 1.3952767104075764e-05, "loss": 0.0597, "step": 16993 }, { "epoch": 3.12, "learning_rate": 1.395208512510282e-05, "loss": 0.0545, "step": 16994 }, { "epoch": 3.12, "learning_rate": 1.3951403124345888e-05, "loss": 0.0483, "step": 16995 }, { "epoch": 3.12, "learning_rate": 1.3950721101808733e-05, "loss": 0.0497, "step": 16996 }, { "epoch": 3.12, "learning_rate": 1.3950039057495108e-05, "loss": 0.0493, "step": 16997 }, { "epoch": 3.12, "learning_rate": 1.3949356991408775e-05, "loss": 0.0495, "step": 16998 }, { "epoch": 3.12, "learning_rate": 1.3948674903553499e-05, "loss": 0.0551, "step": 16999 }, { "epoch": 3.12, "learning_rate": 1.394799279393303e-05, "loss": 0.0531, "step": 17000 }, { "epoch": 3.12, "learning_rate": 1.3947310662551132e-05, "loss": 0.0468, "step": 17001 }, { "epoch": 3.12, "learning_rate": 1.3946628509411567e-05, "loss": 0.0461, "step": 17002 }, { "epoch": 3.12, "learning_rate": 1.3945946334518092e-05, "loss": 0.0582, "step": 17003 }, { "epoch": 3.12, "learning_rate": 1.3945264137874468e-05, "loss": 0.0498, "step": 17004 }, { "epoch": 3.12, "learning_rate": 1.394458191948446e-05, "loss": 0.0615, "step": 17005 }, { "epoch": 3.12, "learning_rate": 1.3943899679351817e-05, "loss": 0.0574, "step": 17006 }, { "epoch": 3.12, "learning_rate": 1.394321741748031e-05, "loss": 0.0415, "step": 17007 }, { "epoch": 3.12, "learning_rate": 1.3942535133873696e-05, "loss": 0.0564, "step": 17008 }, { "epoch": 3.12, "learning_rate": 1.3941852828535738e-05, "loss": 0.052, "step": 17009 }, { "epoch": 3.12, "learning_rate": 1.3941170501470189e-05, "loss": 0.0456, "step": 17010 }, { "epoch": 3.12, "learning_rate": 1.3940488152680823e-05, "loss": 0.0498, "step": 17011 }, { "epoch": 3.12, "learning_rate": 1.3939805782171391e-05, "loss": 0.049, "step": 17012 }, { "epoch": 3.12, "learning_rate": 1.3939123389945657e-05, "loss": 0.0445, "step": 17013 }, { "epoch": 3.12, "learning_rate": 1.3938440976007384e-05, "loss": 0.0517, "step": 17014 }, { "epoch": 3.12, "learning_rate": 1.393775854036033e-05, "loss": 0.0482, "step": 17015 }, { "epoch": 3.12, "learning_rate": 1.3937076083008258e-05, "loss": 0.0523, "step": 17016 }, { "epoch": 3.12, "learning_rate": 1.3936393603954932e-05, "loss": 0.0468, "step": 17017 }, { "epoch": 3.12, "learning_rate": 1.3935711103204113e-05, "loss": 0.0427, "step": 17018 }, { "epoch": 3.12, "learning_rate": 1.393502858075956e-05, "loss": 0.0493, "step": 17019 }, { "epoch": 3.12, "learning_rate": 1.393434603662504e-05, "loss": 0.0606, "step": 17020 }, { "epoch": 3.12, "learning_rate": 1.3933663470804308e-05, "loss": 0.0452, "step": 17021 }, { "epoch": 3.12, "learning_rate": 1.3932980883301135e-05, "loss": 0.0469, "step": 17022 }, { "epoch": 3.12, "learning_rate": 1.393229827411928e-05, "loss": 0.0472, "step": 17023 }, { "epoch": 3.12, "learning_rate": 1.3931615643262502e-05, "loss": 0.0531, "step": 17024 }, { "epoch": 3.12, "learning_rate": 1.3930932990734567e-05, "loss": 0.0524, "step": 17025 }, { "epoch": 3.12, "learning_rate": 1.3930250316539237e-05, "loss": 0.0449, "step": 17026 }, { "epoch": 3.12, "learning_rate": 1.3929567620680278e-05, "loss": 0.0443, "step": 17027 }, { "epoch": 3.12, "learning_rate": 1.3928884903161448e-05, "loss": 0.045, "step": 17028 }, { "epoch": 3.12, "learning_rate": 1.392820216398651e-05, "loss": 0.0484, "step": 17029 }, { "epoch": 3.12, "learning_rate": 1.3927519403159231e-05, "loss": 0.0547, "step": 17030 }, { "epoch": 3.12, "learning_rate": 1.3926836620683376e-05, "loss": 0.0566, "step": 17031 }, { "epoch": 3.12, "learning_rate": 1.3926153816562704e-05, "loss": 0.0509, "step": 17032 }, { "epoch": 3.12, "learning_rate": 1.3925470990800978e-05, "loss": 0.0557, "step": 17033 }, { "epoch": 3.12, "learning_rate": 1.3924788143401966e-05, "loss": 0.0506, "step": 17034 }, { "epoch": 3.12, "learning_rate": 1.392410527436943e-05, "loss": 0.0489, "step": 17035 }, { "epoch": 3.12, "learning_rate": 1.3923422383707133e-05, "loss": 0.0462, "step": 17036 }, { "epoch": 3.12, "learning_rate": 1.392273947141884e-05, "loss": 0.0557, "step": 17037 }, { "epoch": 3.12, "learning_rate": 1.3922056537508317e-05, "loss": 0.0541, "step": 17038 }, { "epoch": 3.12, "learning_rate": 1.3921373581979323e-05, "loss": 0.0476, "step": 17039 }, { "epoch": 3.12, "learning_rate": 1.3920690604835627e-05, "loss": 0.0471, "step": 17040 }, { "epoch": 3.13, "learning_rate": 1.3920007606080995e-05, "loss": 0.0586, "step": 17041 }, { "epoch": 3.13, "learning_rate": 1.3919324585719186e-05, "loss": 0.0506, "step": 17042 }, { "epoch": 3.13, "learning_rate": 1.391864154375397e-05, "loss": 0.0465, "step": 17043 }, { "epoch": 3.13, "learning_rate": 1.3917958480189112e-05, "loss": 0.0449, "step": 17044 }, { "epoch": 3.13, "learning_rate": 1.3917275395028372e-05, "loss": 0.0538, "step": 17045 }, { "epoch": 3.13, "learning_rate": 1.3916592288275518e-05, "loss": 0.0474, "step": 17046 }, { "epoch": 3.13, "learning_rate": 1.3915909159934318e-05, "loss": 0.0519, "step": 17047 }, { "epoch": 3.13, "learning_rate": 1.391522601000853e-05, "loss": 0.0498, "step": 17048 }, { "epoch": 3.13, "learning_rate": 1.3914542838501928e-05, "loss": 0.0525, "step": 17049 }, { "epoch": 3.13, "learning_rate": 1.3913859645418276e-05, "loss": 0.0387, "step": 17050 }, { "epoch": 3.13, "learning_rate": 1.3913176430761335e-05, "loss": 0.0499, "step": 17051 }, { "epoch": 3.13, "learning_rate": 1.3912493194534876e-05, "loss": 0.054, "step": 17052 }, { "epoch": 3.13, "learning_rate": 1.3911809936742662e-05, "loss": 0.0502, "step": 17053 }, { "epoch": 3.13, "learning_rate": 1.3911126657388461e-05, "loss": 0.0556, "step": 17054 }, { "epoch": 3.13, "learning_rate": 1.3910443356476036e-05, "loss": 0.0491, "step": 17055 }, { "epoch": 3.13, "learning_rate": 1.3909760034009156e-05, "loss": 0.05, "step": 17056 }, { "epoch": 3.13, "learning_rate": 1.3909076689991587e-05, "loss": 0.0522, "step": 17057 }, { "epoch": 3.13, "learning_rate": 1.3908393324427098e-05, "loss": 0.0519, "step": 17058 }, { "epoch": 3.13, "learning_rate": 1.3907709937319451e-05, "loss": 0.0569, "step": 17059 }, { "epoch": 3.13, "learning_rate": 1.3907026528672416e-05, "loss": 0.0449, "step": 17060 }, { "epoch": 3.13, "learning_rate": 1.3906343098489759e-05, "loss": 0.0597, "step": 17061 }, { "epoch": 3.13, "learning_rate": 1.3905659646775247e-05, "loss": 0.0558, "step": 17062 }, { "epoch": 3.13, "learning_rate": 1.3904976173532647e-05, "loss": 0.0483, "step": 17063 }, { "epoch": 3.13, "learning_rate": 1.3904292678765732e-05, "loss": 0.0425, "step": 17064 }, { "epoch": 3.13, "learning_rate": 1.390360916247826e-05, "loss": 0.0451, "step": 17065 }, { "epoch": 3.13, "learning_rate": 1.3902925624674e-05, "loss": 0.0392, "step": 17066 }, { "epoch": 3.13, "learning_rate": 1.3902242065356729e-05, "loss": 0.0525, "step": 17067 }, { "epoch": 3.13, "learning_rate": 1.3901558484530203e-05, "loss": 0.0515, "step": 17068 }, { "epoch": 3.13, "learning_rate": 1.3900874882198196e-05, "loss": 0.0516, "step": 17069 }, { "epoch": 3.13, "learning_rate": 1.3900191258364476e-05, "loss": 0.0493, "step": 17070 }, { "epoch": 3.13, "learning_rate": 1.3899507613032812e-05, "loss": 0.0507, "step": 17071 }, { "epoch": 3.13, "learning_rate": 1.3898823946206967e-05, "loss": 0.0483, "step": 17072 }, { "epoch": 3.13, "learning_rate": 1.389814025789072e-05, "loss": 0.0548, "step": 17073 }, { "epoch": 3.13, "learning_rate": 1.3897456548087824e-05, "loss": 0.0524, "step": 17074 }, { "epoch": 3.13, "learning_rate": 1.3896772816802062e-05, "loss": 0.0555, "step": 17075 }, { "epoch": 3.13, "learning_rate": 1.3896089064037197e-05, "loss": 0.0492, "step": 17076 }, { "epoch": 3.13, "learning_rate": 1.3895405289796994e-05, "loss": 0.0445, "step": 17077 }, { "epoch": 3.13, "learning_rate": 1.389472149408523e-05, "loss": 0.0473, "step": 17078 }, { "epoch": 3.13, "learning_rate": 1.3894037676905667e-05, "loss": 0.0465, "step": 17079 }, { "epoch": 3.13, "learning_rate": 1.3893353838262078e-05, "loss": 0.0558, "step": 17080 }, { "epoch": 3.13, "learning_rate": 1.389266997815823e-05, "loss": 0.0549, "step": 17081 }, { "epoch": 3.13, "learning_rate": 1.3891986096597897e-05, "loss": 0.0471, "step": 17082 }, { "epoch": 3.13, "learning_rate": 1.389130219358484e-05, "loss": 0.0468, "step": 17083 }, { "epoch": 3.13, "learning_rate": 1.389061826912284e-05, "loss": 0.0519, "step": 17084 }, { "epoch": 3.13, "learning_rate": 1.3889934323215659e-05, "loss": 0.045, "step": 17085 }, { "epoch": 3.13, "learning_rate": 1.388925035586707e-05, "loss": 0.0472, "step": 17086 }, { "epoch": 3.13, "learning_rate": 1.388856636708084e-05, "loss": 0.0458, "step": 17087 }, { "epoch": 3.13, "learning_rate": 1.3887882356860743e-05, "loss": 0.0494, "step": 17088 }, { "epoch": 3.13, "learning_rate": 1.3887198325210544e-05, "loss": 0.0521, "step": 17089 }, { "epoch": 3.13, "learning_rate": 1.388651427213402e-05, "loss": 0.0494, "step": 17090 }, { "epoch": 3.13, "learning_rate": 1.3885830197634937e-05, "loss": 0.0462, "step": 17091 }, { "epoch": 3.13, "learning_rate": 1.3885146101717066e-05, "loss": 0.0648, "step": 17092 }, { "epoch": 3.13, "learning_rate": 1.3884461984384179e-05, "loss": 0.0526, "step": 17093 }, { "epoch": 3.13, "learning_rate": 1.3883777845640047e-05, "loss": 0.054, "step": 17094 }, { "epoch": 3.13, "learning_rate": 1.3883093685488443e-05, "loss": 0.0523, "step": 17095 }, { "epoch": 3.14, "learning_rate": 1.3882409503933132e-05, "loss": 0.0402, "step": 17096 }, { "epoch": 3.14, "learning_rate": 1.388172530097789e-05, "loss": 0.0509, "step": 17097 }, { "epoch": 3.14, "learning_rate": 1.3881041076626489e-05, "loss": 0.0424, "step": 17098 }, { "epoch": 3.14, "learning_rate": 1.3880356830882695e-05, "loss": 0.0503, "step": 17099 }, { "epoch": 3.14, "learning_rate": 1.3879672563750288e-05, "loss": 0.0481, "step": 17100 }, { "epoch": 3.14, "learning_rate": 1.387898827523303e-05, "loss": 0.0401, "step": 17101 }, { "epoch": 3.14, "learning_rate": 1.38783039653347e-05, "loss": 0.0528, "step": 17102 }, { "epoch": 3.14, "learning_rate": 1.387761963405907e-05, "loss": 0.0443, "step": 17103 }, { "epoch": 3.14, "learning_rate": 1.3876935281409907e-05, "loss": 0.0544, "step": 17104 }, { "epoch": 3.14, "learning_rate": 1.3876250907390985e-05, "loss": 0.0461, "step": 17105 }, { "epoch": 3.14, "learning_rate": 1.387556651200608e-05, "loss": 0.0588, "step": 17106 }, { "epoch": 3.14, "learning_rate": 1.3874882095258959e-05, "loss": 0.0468, "step": 17107 }, { "epoch": 3.14, "learning_rate": 1.3874197657153397e-05, "loss": 0.0488, "step": 17108 }, { "epoch": 3.14, "learning_rate": 1.387351319769317e-05, "loss": 0.058, "step": 17109 }, { "epoch": 3.14, "learning_rate": 1.3872828716882042e-05, "loss": 0.0521, "step": 17110 }, { "epoch": 3.14, "learning_rate": 1.3872144214723797e-05, "loss": 0.0543, "step": 17111 }, { "epoch": 3.14, "learning_rate": 1.3871459691222203e-05, "loss": 0.0527, "step": 17112 }, { "epoch": 3.14, "learning_rate": 1.3870775146381028e-05, "loss": 0.0484, "step": 17113 }, { "epoch": 3.14, "learning_rate": 1.3870090580204052e-05, "loss": 0.0512, "step": 17114 }, { "epoch": 3.14, "learning_rate": 1.3869405992695043e-05, "loss": 0.0549, "step": 17115 }, { "epoch": 3.14, "learning_rate": 1.386872138385778e-05, "loss": 0.0493, "step": 17116 }, { "epoch": 3.14, "learning_rate": 1.3868036753696034e-05, "loss": 0.0427, "step": 17117 }, { "epoch": 3.14, "learning_rate": 1.3867352102213582e-05, "loss": 0.047, "step": 17118 }, { "epoch": 3.14, "learning_rate": 1.3866667429414188e-05, "loss": 0.0462, "step": 17119 }, { "epoch": 3.14, "learning_rate": 1.3865982735301637e-05, "loss": 0.05, "step": 17120 }, { "epoch": 3.14, "learning_rate": 1.3865298019879698e-05, "loss": 0.053, "step": 17121 }, { "epoch": 3.14, "learning_rate": 1.3864613283152147e-05, "loss": 0.055, "step": 17122 }, { "epoch": 3.14, "learning_rate": 1.3863928525122754e-05, "loss": 0.0468, "step": 17123 }, { "epoch": 3.14, "learning_rate": 1.3863243745795299e-05, "loss": 0.0512, "step": 17124 }, { "epoch": 3.14, "learning_rate": 1.3862558945173552e-05, "loss": 0.0527, "step": 17125 }, { "epoch": 3.14, "learning_rate": 1.3861874123261289e-05, "loss": 0.0546, "step": 17126 }, { "epoch": 3.14, "learning_rate": 1.3861189280062288e-05, "loss": 0.0546, "step": 17127 }, { "epoch": 3.14, "learning_rate": 1.386050441558032e-05, "loss": 0.0484, "step": 17128 }, { "epoch": 3.14, "learning_rate": 1.3859819529819164e-05, "loss": 0.0529, "step": 17129 }, { "epoch": 3.14, "learning_rate": 1.3859134622782588e-05, "loss": 0.0536, "step": 17130 }, { "epoch": 3.14, "learning_rate": 1.3858449694474372e-05, "loss": 0.0496, "step": 17131 }, { "epoch": 3.14, "learning_rate": 1.3857764744898292e-05, "loss": 0.0497, "step": 17132 }, { "epoch": 3.14, "learning_rate": 1.3857079774058124e-05, "loss": 0.0484, "step": 17133 }, { "epoch": 3.14, "learning_rate": 1.3856394781957639e-05, "loss": 0.0538, "step": 17134 }, { "epoch": 3.14, "learning_rate": 1.3855709768600618e-05, "loss": 0.0543, "step": 17135 }, { "epoch": 3.14, "learning_rate": 1.3855024733990831e-05, "loss": 0.0486, "step": 17136 }, { "epoch": 3.14, "learning_rate": 1.385433967813206e-05, "loss": 0.052, "step": 17137 }, { "epoch": 3.14, "learning_rate": 1.3853654601028079e-05, "loss": 0.0539, "step": 17138 }, { "epoch": 3.14, "learning_rate": 1.3852969502682663e-05, "loss": 0.0543, "step": 17139 }, { "epoch": 3.14, "learning_rate": 1.3852284383099587e-05, "loss": 0.0484, "step": 17140 }, { "epoch": 3.14, "learning_rate": 1.3851599242282635e-05, "loss": 0.0524, "step": 17141 }, { "epoch": 3.14, "learning_rate": 1.3850914080235572e-05, "loss": 0.0486, "step": 17142 }, { "epoch": 3.14, "learning_rate": 1.3850228896962178e-05, "loss": 0.0447, "step": 17143 }, { "epoch": 3.14, "learning_rate": 1.384954369246624e-05, "loss": 0.0434, "step": 17144 }, { "epoch": 3.14, "learning_rate": 1.3848858466751521e-05, "loss": 0.0448, "step": 17145 }, { "epoch": 3.14, "learning_rate": 1.3848173219821806e-05, "loss": 0.0486, "step": 17146 }, { "epoch": 3.14, "learning_rate": 1.3847487951680871e-05, "loss": 0.0656, "step": 17147 }, { "epoch": 3.14, "learning_rate": 1.3846802662332493e-05, "loss": 0.0608, "step": 17148 }, { "epoch": 3.14, "learning_rate": 1.3846117351780445e-05, "loss": 0.0535, "step": 17149 }, { "epoch": 3.15, "learning_rate": 1.3845432020028511e-05, "loss": 0.0472, "step": 17150 }, { "epoch": 3.15, "learning_rate": 1.3844746667080464e-05, "loss": 0.0519, "step": 17151 }, { "epoch": 3.15, "learning_rate": 1.3844061292940084e-05, "loss": 0.0504, "step": 17152 }, { "epoch": 3.15, "learning_rate": 1.3843375897611148e-05, "loss": 0.0523, "step": 17153 }, { "epoch": 3.15, "learning_rate": 1.3842690481097431e-05, "loss": 0.0579, "step": 17154 }, { "epoch": 3.15, "learning_rate": 1.3842005043402718e-05, "loss": 0.0454, "step": 17155 }, { "epoch": 3.15, "learning_rate": 1.3841319584530784e-05, "loss": 0.0533, "step": 17156 }, { "epoch": 3.15, "learning_rate": 1.3840634104485404e-05, "loss": 0.051, "step": 17157 }, { "epoch": 3.15, "learning_rate": 1.3839948603270357e-05, "loss": 0.0461, "step": 17158 }, { "epoch": 3.15, "learning_rate": 1.383926308088943e-05, "loss": 0.0506, "step": 17159 }, { "epoch": 3.15, "learning_rate": 1.3838577537346389e-05, "loss": 0.0557, "step": 17160 }, { "epoch": 3.15, "learning_rate": 1.3837891972645018e-05, "loss": 0.0488, "step": 17161 }, { "epoch": 3.15, "learning_rate": 1.3837206386789101e-05, "loss": 0.0467, "step": 17162 }, { "epoch": 3.15, "learning_rate": 1.3836520779782409e-05, "loss": 0.0605, "step": 17163 }, { "epoch": 3.15, "learning_rate": 1.3835835151628728e-05, "loss": 0.0513, "step": 17164 }, { "epoch": 3.15, "learning_rate": 1.3835149502331832e-05, "loss": 0.0503, "step": 17165 }, { "epoch": 3.15, "learning_rate": 1.3834463831895501e-05, "loss": 0.0532, "step": 17166 }, { "epoch": 3.15, "learning_rate": 1.3833778140323514e-05, "loss": 0.0531, "step": 17167 }, { "epoch": 3.15, "learning_rate": 1.3833092427619658e-05, "loss": 0.0545, "step": 17168 }, { "epoch": 3.15, "learning_rate": 1.38324066937877e-05, "loss": 0.0543, "step": 17169 }, { "epoch": 3.15, "learning_rate": 1.3831720938831432e-05, "loss": 0.0447, "step": 17170 }, { "epoch": 3.15, "learning_rate": 1.3831035162754627e-05, "loss": 0.0484, "step": 17171 }, { "epoch": 3.15, "learning_rate": 1.3830349365561064e-05, "loss": 0.0459, "step": 17172 }, { "epoch": 3.15, "learning_rate": 1.3829663547254527e-05, "loss": 0.0527, "step": 17173 }, { "epoch": 3.15, "learning_rate": 1.3828977707838798e-05, "loss": 0.0473, "step": 17174 }, { "epoch": 3.15, "learning_rate": 1.382829184731765e-05, "loss": 0.0466, "step": 17175 }, { "epoch": 3.15, "learning_rate": 1.3827605965694865e-05, "loss": 0.0443, "step": 17176 }, { "epoch": 3.15, "learning_rate": 1.3826920062974234e-05, "loss": 0.0469, "step": 17177 }, { "epoch": 3.15, "learning_rate": 1.3826234139159525e-05, "loss": 0.0442, "step": 17178 }, { "epoch": 3.15, "learning_rate": 1.3825548194254523e-05, "loss": 0.0464, "step": 17179 }, { "epoch": 3.15, "learning_rate": 1.3824862228263012e-05, "loss": 0.0527, "step": 17180 }, { "epoch": 3.15, "learning_rate": 1.3824176241188771e-05, "loss": 0.0543, "step": 17181 }, { "epoch": 3.15, "learning_rate": 1.382349023303558e-05, "loss": 0.0597, "step": 17182 }, { "epoch": 3.15, "learning_rate": 1.3822804203807223e-05, "loss": 0.055, "step": 17183 }, { "epoch": 3.15, "learning_rate": 1.3822118153507478e-05, "loss": 0.0451, "step": 17184 }, { "epoch": 3.15, "learning_rate": 1.3821432082140128e-05, "loss": 0.0536, "step": 17185 }, { "epoch": 3.15, "learning_rate": 1.3820745989708957e-05, "loss": 0.0636, "step": 17186 }, { "epoch": 3.15, "learning_rate": 1.3820059876217741e-05, "loss": 0.0584, "step": 17187 }, { "epoch": 3.15, "learning_rate": 1.3819373741670268e-05, "loss": 0.0525, "step": 17188 }, { "epoch": 3.15, "learning_rate": 1.3818687586070316e-05, "loss": 0.0544, "step": 17189 }, { "epoch": 3.15, "learning_rate": 1.381800140942167e-05, "loss": 0.0546, "step": 17190 }, { "epoch": 3.15, "learning_rate": 1.3817315211728111e-05, "loss": 0.0461, "step": 17191 }, { "epoch": 3.15, "learning_rate": 1.381662899299342e-05, "loss": 0.0405, "step": 17192 }, { "epoch": 3.15, "learning_rate": 1.3815942753221381e-05, "loss": 0.058, "step": 17193 }, { "epoch": 3.15, "learning_rate": 1.3815256492415774e-05, "loss": 0.0509, "step": 17194 }, { "epoch": 3.15, "learning_rate": 1.3814570210580386e-05, "loss": 0.0637, "step": 17195 }, { "epoch": 3.15, "learning_rate": 1.3813883907718997e-05, "loss": 0.058, "step": 17196 }, { "epoch": 3.15, "learning_rate": 1.3813197583835391e-05, "loss": 0.0485, "step": 17197 }, { "epoch": 3.15, "learning_rate": 1.381251123893335e-05, "loss": 0.05, "step": 17198 }, { "epoch": 3.15, "learning_rate": 1.3811824873016662e-05, "loss": 0.0557, "step": 17199 }, { "epoch": 3.15, "learning_rate": 1.38111384860891e-05, "loss": 0.0472, "step": 17200 }, { "epoch": 3.15, "learning_rate": 1.3810452078154457e-05, "loss": 0.0465, "step": 17201 }, { "epoch": 3.15, "learning_rate": 1.3809765649216509e-05, "loss": 0.0454, "step": 17202 }, { "epoch": 3.15, "learning_rate": 1.3809079199279047e-05, "loss": 0.052, "step": 17203 }, { "epoch": 3.15, "learning_rate": 1.3808392728345852e-05, "loss": 0.0502, "step": 17204 }, { "epoch": 3.16, "learning_rate": 1.3807706236420704e-05, "loss": 0.0553, "step": 17205 }, { "epoch": 3.16, "learning_rate": 1.380701972350739e-05, "loss": 0.0469, "step": 17206 }, { "epoch": 3.16, "learning_rate": 1.3806333189609698e-05, "loss": 0.0498, "step": 17207 }, { "epoch": 3.16, "learning_rate": 1.3805646634731404e-05, "loss": 0.0546, "step": 17208 }, { "epoch": 3.16, "learning_rate": 1.38049600588763e-05, "loss": 0.0523, "step": 17209 }, { "epoch": 3.16, "learning_rate": 1.3804273462048165e-05, "loss": 0.0448, "step": 17210 }, { "epoch": 3.16, "learning_rate": 1.3803586844250784e-05, "loss": 0.0482, "step": 17211 }, { "epoch": 3.16, "learning_rate": 1.3802900205487948e-05, "loss": 0.0582, "step": 17212 }, { "epoch": 3.16, "learning_rate": 1.3802213545763432e-05, "loss": 0.0414, "step": 17213 }, { "epoch": 3.16, "learning_rate": 1.3801526865081028e-05, "loss": 0.0495, "step": 17214 }, { "epoch": 3.16, "learning_rate": 1.3800840163444517e-05, "loss": 0.0511, "step": 17215 }, { "epoch": 3.16, "learning_rate": 1.3800153440857691e-05, "loss": 0.0614, "step": 17216 }, { "epoch": 3.16, "learning_rate": 1.3799466697324327e-05, "loss": 0.0454, "step": 17217 }, { "epoch": 3.16, "learning_rate": 1.3798779932848212e-05, "loss": 0.0491, "step": 17218 }, { "epoch": 3.16, "learning_rate": 1.3798093147433135e-05, "loss": 0.0536, "step": 17219 }, { "epoch": 3.16, "learning_rate": 1.3797406341082877e-05, "loss": 0.0449, "step": 17220 }, { "epoch": 3.16, "learning_rate": 1.379671951380123e-05, "loss": 0.0562, "step": 17221 }, { "epoch": 3.16, "learning_rate": 1.3796032665591973e-05, "loss": 0.0497, "step": 17222 }, { "epoch": 3.16, "learning_rate": 1.3795345796458898e-05, "loss": 0.0496, "step": 17223 }, { "epoch": 3.16, "learning_rate": 1.3794658906405786e-05, "loss": 0.0485, "step": 17224 }, { "epoch": 3.16, "learning_rate": 1.3793971995436426e-05, "loss": 0.0515, "step": 17225 }, { "epoch": 3.16, "learning_rate": 1.3793285063554604e-05, "loss": 0.0503, "step": 17226 }, { "epoch": 3.16, "learning_rate": 1.3792598110764107e-05, "loss": 0.0526, "step": 17227 }, { "epoch": 3.16, "learning_rate": 1.3791911137068716e-05, "loss": 0.0516, "step": 17228 }, { "epoch": 3.16, "learning_rate": 1.3791224142472224e-05, "loss": 0.0487, "step": 17229 }, { "epoch": 3.16, "learning_rate": 1.3790537126978418e-05, "loss": 0.0534, "step": 17230 }, { "epoch": 3.16, "learning_rate": 1.378985009059108e-05, "loss": 0.0517, "step": 17231 }, { "epoch": 3.16, "learning_rate": 1.3789163033314e-05, "loss": 0.0543, "step": 17232 }, { "epoch": 3.16, "learning_rate": 1.3788475955150964e-05, "loss": 0.0559, "step": 17233 }, { "epoch": 3.16, "learning_rate": 1.3787788856105762e-05, "loss": 0.0577, "step": 17234 }, { "epoch": 3.16, "learning_rate": 1.3787101736182177e-05, "loss": 0.0559, "step": 17235 }, { "epoch": 3.16, "learning_rate": 1.3786414595384001e-05, "loss": 0.0495, "step": 17236 }, { "epoch": 3.16, "learning_rate": 1.3785727433715018e-05, "loss": 0.0515, "step": 17237 }, { "epoch": 3.16, "learning_rate": 1.3785040251179013e-05, "loss": 0.0437, "step": 17238 }, { "epoch": 3.16, "learning_rate": 1.3784353047779781e-05, "loss": 0.0495, "step": 17239 }, { "epoch": 3.16, "learning_rate": 1.3783665823521106e-05, "loss": 0.0598, "step": 17240 }, { "epoch": 3.16, "learning_rate": 1.3782978578406777e-05, "loss": 0.0462, "step": 17241 }, { "epoch": 3.16, "learning_rate": 1.3782291312440582e-05, "loss": 0.0489, "step": 17242 }, { "epoch": 3.16, "learning_rate": 1.3781604025626305e-05, "loss": 0.0574, "step": 17243 }, { "epoch": 3.16, "learning_rate": 1.3780916717967738e-05, "loss": 0.047, "step": 17244 }, { "epoch": 3.16, "learning_rate": 1.3780229389468676e-05, "loss": 0.0508, "step": 17245 }, { "epoch": 3.16, "learning_rate": 1.3779542040132896e-05, "loss": 0.0493, "step": 17246 }, { "epoch": 3.16, "learning_rate": 1.377885466996419e-05, "loss": 0.0552, "step": 17247 }, { "epoch": 3.16, "learning_rate": 1.3778167278966352e-05, "loss": 0.0529, "step": 17248 }, { "epoch": 3.16, "learning_rate": 1.3777479867143167e-05, "loss": 0.0483, "step": 17249 }, { "epoch": 3.16, "learning_rate": 1.3776792434498423e-05, "loss": 0.0468, "step": 17250 }, { "epoch": 3.16, "learning_rate": 1.3776104981035912e-05, "loss": 0.0581, "step": 17251 }, { "epoch": 3.16, "learning_rate": 1.3775417506759421e-05, "loss": 0.0504, "step": 17252 }, { "epoch": 3.16, "learning_rate": 1.3774730011672738e-05, "loss": 0.0533, "step": 17253 }, { "epoch": 3.16, "learning_rate": 1.3774042495779662e-05, "loss": 0.0531, "step": 17254 }, { "epoch": 3.16, "learning_rate": 1.3773354959083968e-05, "loss": 0.0453, "step": 17255 }, { "epoch": 3.16, "learning_rate": 1.3772667401589456e-05, "loss": 0.0552, "step": 17256 }, { "epoch": 3.16, "learning_rate": 1.3771979823299915e-05, "loss": 0.0496, "step": 17257 }, { "epoch": 3.16, "learning_rate": 1.3771292224219128e-05, "loss": 0.0556, "step": 17258 }, { "epoch": 3.17, "learning_rate": 1.3770604604350895e-05, "loss": 0.0611, "step": 17259 }, { "epoch": 3.17, "learning_rate": 1.3769916963698997e-05, "loss": 0.0496, "step": 17260 }, { "epoch": 3.17, "learning_rate": 1.376922930226723e-05, "loss": 0.0513, "step": 17261 }, { "epoch": 3.17, "learning_rate": 1.376854162005938e-05, "loss": 0.0494, "step": 17262 }, { "epoch": 3.17, "learning_rate": 1.3767853917079249e-05, "loss": 0.0503, "step": 17263 }, { "epoch": 3.17, "learning_rate": 1.376716619333061e-05, "loss": 0.0443, "step": 17264 }, { "epoch": 3.17, "learning_rate": 1.3766478448817264e-05, "loss": 0.047, "step": 17265 }, { "epoch": 3.17, "learning_rate": 1.3765790683543004e-05, "loss": 0.0606, "step": 17266 }, { "epoch": 3.17, "learning_rate": 1.3765102897511615e-05, "loss": 0.0573, "step": 17267 }, { "epoch": 3.17, "learning_rate": 1.3764415090726893e-05, "loss": 0.051, "step": 17268 }, { "epoch": 3.17, "learning_rate": 1.3763727263192626e-05, "loss": 0.0459, "step": 17269 }, { "epoch": 3.17, "learning_rate": 1.3763039414912605e-05, "loss": 0.0538, "step": 17270 }, { "epoch": 3.17, "learning_rate": 1.376235154589062e-05, "loss": 0.0548, "step": 17271 }, { "epoch": 3.17, "learning_rate": 1.376166365613047e-05, "loss": 0.0442, "step": 17272 }, { "epoch": 3.17, "learning_rate": 1.3760975745635938e-05, "loss": 0.0521, "step": 17273 }, { "epoch": 3.17, "learning_rate": 1.3760287814410822e-05, "loss": 0.0464, "step": 17274 }, { "epoch": 3.17, "learning_rate": 1.3759599862458912e-05, "loss": 0.0458, "step": 17275 }, { "epoch": 3.17, "learning_rate": 1.3758911889784e-05, "loss": 0.0572, "step": 17276 }, { "epoch": 3.17, "learning_rate": 1.3758223896389873e-05, "loss": 0.0556, "step": 17277 }, { "epoch": 3.17, "learning_rate": 1.3757535882280334e-05, "loss": 0.0559, "step": 17278 }, { "epoch": 3.17, "learning_rate": 1.3756847847459164e-05, "loss": 0.0416, "step": 17279 }, { "epoch": 3.17, "learning_rate": 1.375615979193016e-05, "loss": 0.0507, "step": 17280 }, { "epoch": 3.17, "learning_rate": 1.3755471715697122e-05, "loss": 0.0403, "step": 17281 }, { "epoch": 3.17, "learning_rate": 1.375478361876383e-05, "loss": 0.0522, "step": 17282 }, { "epoch": 3.17, "learning_rate": 1.3754095501134084e-05, "loss": 0.0441, "step": 17283 }, { "epoch": 3.17, "learning_rate": 1.3753407362811678e-05, "loss": 0.0496, "step": 17284 }, { "epoch": 3.17, "learning_rate": 1.37527192038004e-05, "loss": 0.0511, "step": 17285 }, { "epoch": 3.17, "learning_rate": 1.3752031024104047e-05, "loss": 0.0512, "step": 17286 }, { "epoch": 3.17, "learning_rate": 1.375134282372641e-05, "loss": 0.0521, "step": 17287 }, { "epoch": 3.17, "learning_rate": 1.3750654602671285e-05, "loss": 0.0441, "step": 17288 }, { "epoch": 3.17, "learning_rate": 1.3749966360942464e-05, "loss": 0.0511, "step": 17289 }, { "epoch": 3.17, "learning_rate": 1.3749278098543739e-05, "loss": 0.0479, "step": 17290 }, { "epoch": 3.17, "learning_rate": 1.3748589815478906e-05, "loss": 0.0565, "step": 17291 }, { "epoch": 3.17, "learning_rate": 1.3747901511751757e-05, "loss": 0.0518, "step": 17292 }, { "epoch": 3.17, "learning_rate": 1.3747213187366088e-05, "loss": 0.0445, "step": 17293 }, { "epoch": 3.17, "learning_rate": 1.3746524842325694e-05, "loss": 0.0513, "step": 17294 }, { "epoch": 3.17, "learning_rate": 1.3745836476634367e-05, "loss": 0.0504, "step": 17295 }, { "epoch": 3.17, "learning_rate": 1.3745148090295899e-05, "loss": 0.05, "step": 17296 }, { "epoch": 3.17, "learning_rate": 1.3744459683314086e-05, "loss": 0.0542, "step": 17297 }, { "epoch": 3.17, "learning_rate": 1.3743771255692729e-05, "loss": 0.0457, "step": 17298 }, { "epoch": 3.17, "learning_rate": 1.3743082807435615e-05, "loss": 0.0559, "step": 17299 }, { "epoch": 3.17, "learning_rate": 1.3742394338546536e-05, "loss": 0.0483, "step": 17300 }, { "epoch": 3.17, "learning_rate": 1.3741705849029297e-05, "loss": 0.0483, "step": 17301 }, { "epoch": 3.17, "learning_rate": 1.3741017338887685e-05, "loss": 0.0532, "step": 17302 }, { "epoch": 3.17, "learning_rate": 1.37403288081255e-05, "loss": 0.0471, "step": 17303 }, { "epoch": 3.17, "learning_rate": 1.3739640256746534e-05, "loss": 0.0507, "step": 17304 }, { "epoch": 3.17, "learning_rate": 1.3738951684754585e-05, "loss": 0.0516, "step": 17305 }, { "epoch": 3.17, "learning_rate": 1.3738263092153443e-05, "loss": 0.045, "step": 17306 }, { "epoch": 3.17, "learning_rate": 1.3737574478946912e-05, "loss": 0.0434, "step": 17307 }, { "epoch": 3.17, "learning_rate": 1.3736885845138776e-05, "loss": 0.0504, "step": 17308 }, { "epoch": 3.17, "learning_rate": 1.3736197190732844e-05, "loss": 0.0454, "step": 17309 }, { "epoch": 3.17, "learning_rate": 1.3735508515732903e-05, "loss": 0.051, "step": 17310 }, { "epoch": 3.17, "learning_rate": 1.3734819820142753e-05, "loss": 0.0493, "step": 17311 }, { "epoch": 3.17, "learning_rate": 1.3734131103966187e-05, "loss": 0.0491, "step": 17312 }, { "epoch": 3.17, "learning_rate": 1.3733442367207004e-05, "loss": 0.0549, "step": 17313 }, { "epoch": 3.18, "learning_rate": 1.3732753609868999e-05, "loss": 0.051, "step": 17314 }, { "epoch": 3.18, "learning_rate": 1.3732064831955968e-05, "loss": 0.0514, "step": 17315 }, { "epoch": 3.18, "learning_rate": 1.3731376033471708e-05, "loss": 0.0481, "step": 17316 }, { "epoch": 3.18, "learning_rate": 1.3730687214420016e-05, "loss": 0.0586, "step": 17317 }, { "epoch": 3.18, "learning_rate": 1.3729998374804686e-05, "loss": 0.0417, "step": 17318 }, { "epoch": 3.18, "learning_rate": 1.3729309514629523e-05, "loss": 0.0467, "step": 17319 }, { "epoch": 3.18, "learning_rate": 1.3728620633898318e-05, "loss": 0.0478, "step": 17320 }, { "epoch": 3.18, "learning_rate": 1.3727931732614865e-05, "loss": 0.0483, "step": 17321 }, { "epoch": 3.18, "learning_rate": 1.3727242810782967e-05, "loss": 0.0489, "step": 17322 }, { "epoch": 3.18, "learning_rate": 1.372655386840642e-05, "loss": 0.0594, "step": 17323 }, { "epoch": 3.18, "learning_rate": 1.3725864905489019e-05, "loss": 0.0515, "step": 17324 }, { "epoch": 3.18, "learning_rate": 1.3725175922034566e-05, "loss": 0.0524, "step": 17325 }, { "epoch": 3.18, "learning_rate": 1.3724486918046853e-05, "loss": 0.0483, "step": 17326 }, { "epoch": 3.18, "learning_rate": 1.3723797893529682e-05, "loss": 0.0396, "step": 17327 }, { "epoch": 3.18, "learning_rate": 1.372310884848685e-05, "loss": 0.0511, "step": 17328 }, { "epoch": 3.18, "learning_rate": 1.3722419782922156e-05, "loss": 0.0547, "step": 17329 }, { "epoch": 3.18, "learning_rate": 1.3721730696839393e-05, "loss": 0.057, "step": 17330 }, { "epoch": 3.18, "learning_rate": 1.3721041590242368e-05, "loss": 0.0459, "step": 17331 }, { "epoch": 3.18, "learning_rate": 1.3720352463134875e-05, "loss": 0.0559, "step": 17332 }, { "epoch": 3.18, "learning_rate": 1.3719663315520705e-05, "loss": 0.0507, "step": 17333 }, { "epoch": 3.18, "learning_rate": 1.3718974147403671e-05, "loss": 0.0479, "step": 17334 }, { "epoch": 3.18, "learning_rate": 1.3718284958787562e-05, "loss": 0.0463, "step": 17335 }, { "epoch": 3.18, "learning_rate": 1.3717595749676178e-05, "loss": 0.0457, "step": 17336 }, { "epoch": 3.18, "learning_rate": 1.371690652007332e-05, "loss": 0.0504, "step": 17337 }, { "epoch": 3.18, "learning_rate": 1.3716217269982788e-05, "loss": 0.0541, "step": 17338 }, { "epoch": 3.18, "learning_rate": 1.3715527999408376e-05, "loss": 0.0564, "step": 17339 }, { "epoch": 3.18, "learning_rate": 1.3714838708353888e-05, "loss": 0.0478, "step": 17340 }, { "epoch": 3.18, "learning_rate": 1.3714149396823122e-05, "loss": 0.0475, "step": 17341 }, { "epoch": 3.18, "learning_rate": 1.3713460064819876e-05, "loss": 0.0502, "step": 17342 }, { "epoch": 3.18, "learning_rate": 1.3712770712347956e-05, "loss": 0.0506, "step": 17343 }, { "epoch": 3.18, "learning_rate": 1.3712081339411152e-05, "loss": 0.046, "step": 17344 }, { "epoch": 3.18, "learning_rate": 1.3711391946013271e-05, "loss": 0.0496, "step": 17345 }, { "epoch": 3.18, "learning_rate": 1.3710702532158112e-05, "loss": 0.0453, "step": 17346 }, { "epoch": 3.18, "learning_rate": 1.3710013097849468e-05, "loss": 0.0543, "step": 17347 }, { "epoch": 3.18, "learning_rate": 1.3709323643091148e-05, "loss": 0.0502, "step": 17348 }, { "epoch": 3.18, "learning_rate": 1.3708634167886952e-05, "loss": 0.0545, "step": 17349 }, { "epoch": 3.18, "learning_rate": 1.3707944672240673e-05, "loss": 0.0449, "step": 17350 }, { "epoch": 3.18, "learning_rate": 1.3707255156156118e-05, "loss": 0.0539, "step": 17351 }, { "epoch": 3.18, "learning_rate": 1.3706565619637089e-05, "loss": 0.0534, "step": 17352 }, { "epoch": 3.18, "learning_rate": 1.370587606268738e-05, "loss": 0.0496, "step": 17353 }, { "epoch": 3.18, "learning_rate": 1.3705186485310797e-05, "loss": 0.0571, "step": 17354 }, { "epoch": 3.18, "learning_rate": 1.3704496887511139e-05, "loss": 0.0494, "step": 17355 }, { "epoch": 3.18, "learning_rate": 1.3703807269292204e-05, "loss": 0.0564, "step": 17356 }, { "epoch": 3.18, "learning_rate": 1.3703117630657799e-05, "loss": 0.0534, "step": 17357 }, { "epoch": 3.18, "learning_rate": 1.3702427971611724e-05, "loss": 0.0495, "step": 17358 }, { "epoch": 3.18, "learning_rate": 1.3701738292157778e-05, "loss": 0.06, "step": 17359 }, { "epoch": 3.18, "learning_rate": 1.3701048592299763e-05, "loss": 0.0438, "step": 17360 }, { "epoch": 3.18, "learning_rate": 1.3700358872041483e-05, "loss": 0.0503, "step": 17361 }, { "epoch": 3.18, "learning_rate": 1.3699669131386739e-05, "loss": 0.0451, "step": 17362 }, { "epoch": 3.18, "learning_rate": 1.3698979370339332e-05, "loss": 0.0482, "step": 17363 }, { "epoch": 3.18, "learning_rate": 1.3698289588903062e-05, "loss": 0.0547, "step": 17364 }, { "epoch": 3.18, "learning_rate": 1.3697599787081734e-05, "loss": 0.055, "step": 17365 }, { "epoch": 3.18, "learning_rate": 1.3696909964879148e-05, "loss": 0.0483, "step": 17366 }, { "epoch": 3.18, "learning_rate": 1.369622012229911e-05, "loss": 0.0529, "step": 17367 }, { "epoch": 3.19, "learning_rate": 1.369553025934542e-05, "loss": 0.0494, "step": 17368 }, { "epoch": 3.19, "learning_rate": 1.3694840376021879e-05, "loss": 0.0428, "step": 17369 }, { "epoch": 3.19, "learning_rate": 1.3694150472332292e-05, "loss": 0.0463, "step": 17370 }, { "epoch": 3.19, "learning_rate": 1.3693460548280462e-05, "loss": 0.0414, "step": 17371 }, { "epoch": 3.19, "learning_rate": 1.369277060387019e-05, "loss": 0.0508, "step": 17372 }, { "epoch": 3.19, "learning_rate": 1.369208063910528e-05, "loss": 0.0578, "step": 17373 }, { "epoch": 3.19, "learning_rate": 1.3691390653989536e-05, "loss": 0.0452, "step": 17374 }, { "epoch": 3.19, "learning_rate": 1.369070064852676e-05, "loss": 0.0516, "step": 17375 }, { "epoch": 3.19, "learning_rate": 1.3690010622720756e-05, "loss": 0.0581, "step": 17376 }, { "epoch": 3.19, "learning_rate": 1.3689320576575323e-05, "loss": 0.0411, "step": 17377 }, { "epoch": 3.19, "learning_rate": 1.3688630510094271e-05, "loss": 0.0503, "step": 17378 }, { "epoch": 3.19, "learning_rate": 1.3687940423281401e-05, "loss": 0.046, "step": 17379 }, { "epoch": 3.19, "learning_rate": 1.368725031614052e-05, "loss": 0.0474, "step": 17380 }, { "epoch": 3.19, "learning_rate": 1.3686560188675424e-05, "loss": 0.0527, "step": 17381 }, { "epoch": 3.19, "learning_rate": 1.3685870040889926e-05, "loss": 0.0557, "step": 17382 }, { "epoch": 3.19, "learning_rate": 1.368517987278782e-05, "loss": 0.0465, "step": 17383 }, { "epoch": 3.19, "learning_rate": 1.3684489684372925e-05, "loss": 0.0568, "step": 17384 }, { "epoch": 3.19, "learning_rate": 1.368379947564903e-05, "loss": 0.048, "step": 17385 }, { "epoch": 3.19, "learning_rate": 1.3683109246619947e-05, "loss": 0.0475, "step": 17386 }, { "epoch": 3.19, "learning_rate": 1.3682418997289481e-05, "loss": 0.0498, "step": 17387 }, { "epoch": 3.19, "learning_rate": 1.3681728727661436e-05, "loss": 0.0426, "step": 17388 }, { "epoch": 3.19, "learning_rate": 1.3681038437739616e-05, "loss": 0.0565, "step": 17389 }, { "epoch": 3.19, "learning_rate": 1.3680348127527823e-05, "loss": 0.0514, "step": 17390 }, { "epoch": 3.19, "learning_rate": 1.3679657797029868e-05, "loss": 0.0415, "step": 17391 }, { "epoch": 3.19, "learning_rate": 1.3678967446249549e-05, "loss": 0.0569, "step": 17392 }, { "epoch": 3.19, "learning_rate": 1.367827707519068e-05, "loss": 0.0523, "step": 17393 }, { "epoch": 3.19, "learning_rate": 1.3677586683857059e-05, "loss": 0.0507, "step": 17394 }, { "epoch": 3.19, "learning_rate": 1.3676896272252493e-05, "loss": 0.0532, "step": 17395 }, { "epoch": 3.19, "learning_rate": 1.3676205840380793e-05, "loss": 0.0486, "step": 17396 }, { "epoch": 3.19, "learning_rate": 1.3675515388245755e-05, "loss": 0.0468, "step": 17397 }, { "epoch": 3.19, "learning_rate": 1.3674824915851193e-05, "loss": 0.0464, "step": 17398 }, { "epoch": 3.19, "learning_rate": 1.367413442320091e-05, "loss": 0.0478, "step": 17399 }, { "epoch": 3.19, "learning_rate": 1.3673443910298712e-05, "loss": 0.0525, "step": 17400 }, { "epoch": 3.19, "learning_rate": 1.3672753377148401e-05, "loss": 0.0572, "step": 17401 }, { "epoch": 3.19, "learning_rate": 1.3672062823753795e-05, "loss": 0.0514, "step": 17402 }, { "epoch": 3.19, "learning_rate": 1.3671372250118686e-05, "loss": 0.0503, "step": 17403 }, { "epoch": 3.19, "learning_rate": 1.367068165624689e-05, "loss": 0.0565, "step": 17404 }, { "epoch": 3.19, "learning_rate": 1.3669991042142212e-05, "loss": 0.0511, "step": 17405 }, { "epoch": 3.19, "learning_rate": 1.3669300407808454e-05, "loss": 0.0514, "step": 17406 }, { "epoch": 3.19, "learning_rate": 1.3668609753249428e-05, "loss": 0.0537, "step": 17407 }, { "epoch": 3.19, "learning_rate": 1.3667919078468941e-05, "loss": 0.0523, "step": 17408 }, { "epoch": 3.19, "learning_rate": 1.3667228383470796e-05, "loss": 0.044, "step": 17409 }, { "epoch": 3.19, "learning_rate": 1.3666537668258802e-05, "loss": 0.0558, "step": 17410 }, { "epoch": 3.19, "learning_rate": 1.3665846932836768e-05, "loss": 0.043, "step": 17411 }, { "epoch": 3.19, "learning_rate": 1.3665156177208497e-05, "loss": 0.0513, "step": 17412 }, { "epoch": 3.19, "learning_rate": 1.36644654013778e-05, "loss": 0.0403, "step": 17413 }, { "epoch": 3.19, "learning_rate": 1.3663774605348486e-05, "loss": 0.0432, "step": 17414 }, { "epoch": 3.19, "learning_rate": 1.3663083789124359e-05, "loss": 0.0563, "step": 17415 }, { "epoch": 3.19, "learning_rate": 1.366239295270923e-05, "loss": 0.0466, "step": 17416 }, { "epoch": 3.19, "learning_rate": 1.3661702096106903e-05, "loss": 0.059, "step": 17417 }, { "epoch": 3.19, "learning_rate": 1.366101121932119e-05, "loss": 0.0466, "step": 17418 }, { "epoch": 3.19, "learning_rate": 1.3660320322355896e-05, "loss": 0.0489, "step": 17419 }, { "epoch": 3.19, "learning_rate": 1.365962940521483e-05, "loss": 0.0525, "step": 17420 }, { "epoch": 3.19, "learning_rate": 1.3658938467901803e-05, "loss": 0.0435, "step": 17421 }, { "epoch": 3.19, "learning_rate": 1.3658247510420622e-05, "loss": 0.0458, "step": 17422 }, { "epoch": 3.2, "learning_rate": 1.3657556532775093e-05, "loss": 0.0549, "step": 17423 }, { "epoch": 3.2, "learning_rate": 1.3656865534969029e-05, "loss": 0.0496, "step": 17424 }, { "epoch": 3.2, "learning_rate": 1.3656174517006231e-05, "loss": 0.0509, "step": 17425 }, { "epoch": 3.2, "learning_rate": 1.3655483478890519e-05, "loss": 0.0456, "step": 17426 }, { "epoch": 3.2, "learning_rate": 1.3654792420625696e-05, "loss": 0.0541, "step": 17427 }, { "epoch": 3.2, "learning_rate": 1.365410134221557e-05, "loss": 0.0484, "step": 17428 }, { "epoch": 3.2, "learning_rate": 1.3653410243663953e-05, "loss": 0.0552, "step": 17429 }, { "epoch": 3.2, "learning_rate": 1.3652719124974652e-05, "loss": 0.0506, "step": 17430 }, { "epoch": 3.2, "learning_rate": 1.3652027986151478e-05, "loss": 0.0456, "step": 17431 }, { "epoch": 3.2, "learning_rate": 1.3651336827198239e-05, "loss": 0.0441, "step": 17432 }, { "epoch": 3.2, "learning_rate": 1.3650645648118748e-05, "loss": 0.0478, "step": 17433 }, { "epoch": 3.2, "learning_rate": 1.3649954448916809e-05, "loss": 0.0552, "step": 17434 }, { "epoch": 3.2, "learning_rate": 1.3649263229596241e-05, "loss": 0.0514, "step": 17435 }, { "epoch": 3.2, "learning_rate": 1.3648571990160843e-05, "loss": 0.0488, "step": 17436 }, { "epoch": 3.2, "learning_rate": 1.3647880730614434e-05, "loss": 0.0551, "step": 17437 }, { "epoch": 3.2, "learning_rate": 1.3647189450960823e-05, "loss": 0.0564, "step": 17438 }, { "epoch": 3.2, "learning_rate": 1.3646498151203813e-05, "loss": 0.0447, "step": 17439 }, { "epoch": 3.2, "learning_rate": 1.3645806831347224e-05, "loss": 0.0537, "step": 17440 }, { "epoch": 3.2, "learning_rate": 1.364511549139486e-05, "loss": 0.0445, "step": 17441 }, { "epoch": 3.2, "learning_rate": 1.3644424131350536e-05, "loss": 0.0493, "step": 17442 }, { "epoch": 3.2, "learning_rate": 1.3643732751218058e-05, "loss": 0.0505, "step": 17443 }, { "epoch": 3.2, "learning_rate": 1.3643041351001243e-05, "loss": 0.0518, "step": 17444 }, { "epoch": 3.2, "learning_rate": 1.3642349930703896e-05, "loss": 0.0477, "step": 17445 }, { "epoch": 3.2, "learning_rate": 1.364165849032983e-05, "loss": 0.0446, "step": 17446 }, { "epoch": 3.2, "learning_rate": 1.3640967029882858e-05, "loss": 0.0501, "step": 17447 }, { "epoch": 3.2, "learning_rate": 1.3640275549366795e-05, "loss": 0.0549, "step": 17448 }, { "epoch": 3.2, "learning_rate": 1.3639584048785444e-05, "loss": 0.0457, "step": 17449 }, { "epoch": 3.2, "learning_rate": 1.3638892528142619e-05, "loss": 0.0512, "step": 17450 }, { "epoch": 3.2, "learning_rate": 1.3638200987442136e-05, "loss": 0.0484, "step": 17451 }, { "epoch": 3.2, "learning_rate": 1.3637509426687803e-05, "loss": 0.0564, "step": 17452 }, { "epoch": 3.2, "learning_rate": 1.3636817845883435e-05, "loss": 0.0445, "step": 17453 }, { "epoch": 3.2, "learning_rate": 1.3636126245032836e-05, "loss": 0.0476, "step": 17454 }, { "epoch": 3.2, "learning_rate": 1.3635434624139827e-05, "loss": 0.0549, "step": 17455 }, { "epoch": 3.2, "learning_rate": 1.3634742983208218e-05, "loss": 0.0504, "step": 17456 }, { "epoch": 3.2, "learning_rate": 1.3634051322241821e-05, "loss": 0.0498, "step": 17457 }, { "epoch": 3.2, "learning_rate": 1.3633359641244446e-05, "loss": 0.05, "step": 17458 }, { "epoch": 3.2, "learning_rate": 1.3632667940219911e-05, "loss": 0.0541, "step": 17459 }, { "epoch": 3.2, "learning_rate": 1.3631976219172023e-05, "loss": 0.0477, "step": 17460 }, { "epoch": 3.2, "learning_rate": 1.3631284478104597e-05, "loss": 0.0457, "step": 17461 }, { "epoch": 3.2, "learning_rate": 1.3630592717021444e-05, "loss": 0.0521, "step": 17462 }, { "epoch": 3.2, "learning_rate": 1.3629900935926382e-05, "loss": 0.0471, "step": 17463 }, { "epoch": 3.2, "learning_rate": 1.3629209134823218e-05, "loss": 0.0519, "step": 17464 }, { "epoch": 3.2, "learning_rate": 1.362851731371577e-05, "loss": 0.0469, "step": 17465 }, { "epoch": 3.2, "learning_rate": 1.3627825472607849e-05, "loss": 0.0419, "step": 17466 }, { "epoch": 3.2, "learning_rate": 1.362713361150327e-05, "loss": 0.0601, "step": 17467 }, { "epoch": 3.2, "learning_rate": 1.3626441730405846e-05, "loss": 0.0493, "step": 17468 }, { "epoch": 3.2, "learning_rate": 1.3625749829319386e-05, "loss": 0.0504, "step": 17469 }, { "epoch": 3.2, "learning_rate": 1.3625057908247714e-05, "loss": 0.048, "step": 17470 }, { "epoch": 3.2, "learning_rate": 1.3624365967194636e-05, "loss": 0.0527, "step": 17471 }, { "epoch": 3.2, "learning_rate": 1.3623674006163966e-05, "loss": 0.0471, "step": 17472 }, { "epoch": 3.2, "learning_rate": 1.3622982025159519e-05, "loss": 0.049, "step": 17473 }, { "epoch": 3.2, "learning_rate": 1.3622290024185114e-05, "loss": 0.0532, "step": 17474 }, { "epoch": 3.2, "learning_rate": 1.3621598003244558e-05, "loss": 0.0431, "step": 17475 }, { "epoch": 3.2, "learning_rate": 1.3620905962341672e-05, "loss": 0.0513, "step": 17476 }, { "epoch": 3.21, "learning_rate": 1.3620213901480265e-05, "loss": 0.0477, "step": 17477 }, { "epoch": 3.21, "learning_rate": 1.3619521820664154e-05, "loss": 0.0496, "step": 17478 }, { "epoch": 3.21, "learning_rate": 1.3618829719897158e-05, "loss": 0.047, "step": 17479 }, { "epoch": 3.21, "learning_rate": 1.3618137599183086e-05, "loss": 0.0549, "step": 17480 }, { "epoch": 3.21, "learning_rate": 1.361744545852575e-05, "loss": 0.0472, "step": 17481 }, { "epoch": 3.21, "learning_rate": 1.3616753297928977e-05, "loss": 0.049, "step": 17482 }, { "epoch": 3.21, "learning_rate": 1.3616061117396571e-05, "loss": 0.0576, "step": 17483 }, { "epoch": 3.21, "learning_rate": 1.3615368916932352e-05, "loss": 0.0497, "step": 17484 }, { "epoch": 3.21, "learning_rate": 1.3614676696540134e-05, "loss": 0.0556, "step": 17485 }, { "epoch": 3.21, "learning_rate": 1.3613984456223736e-05, "loss": 0.0647, "step": 17486 }, { "epoch": 3.21, "learning_rate": 1.3613292195986967e-05, "loss": 0.0501, "step": 17487 }, { "epoch": 3.21, "learning_rate": 1.3612599915833654e-05, "loss": 0.0592, "step": 17488 }, { "epoch": 3.21, "learning_rate": 1.3611907615767596e-05, "loss": 0.0457, "step": 17489 }, { "epoch": 3.21, "learning_rate": 1.3611215295792625e-05, "loss": 0.0428, "step": 17490 }, { "epoch": 3.21, "learning_rate": 1.3610522955912551e-05, "loss": 0.0465, "step": 17491 }, { "epoch": 3.21, "learning_rate": 1.3609830596131185e-05, "loss": 0.0453, "step": 17492 }, { "epoch": 3.21, "learning_rate": 1.3609138216452352e-05, "loss": 0.054, "step": 17493 }, { "epoch": 3.21, "learning_rate": 1.3608445816879865e-05, "loss": 0.0528, "step": 17494 }, { "epoch": 3.21, "learning_rate": 1.3607753397417539e-05, "loss": 0.0447, "step": 17495 }, { "epoch": 3.21, "learning_rate": 1.360706095806919e-05, "loss": 0.0478, "step": 17496 }, { "epoch": 3.21, "learning_rate": 1.360636849883864e-05, "loss": 0.053, "step": 17497 }, { "epoch": 3.21, "learning_rate": 1.3605676019729698e-05, "loss": 0.0563, "step": 17498 }, { "epoch": 3.21, "learning_rate": 1.3604983520746189e-05, "loss": 0.0392, "step": 17499 }, { "epoch": 3.21, "learning_rate": 1.3604291001891924e-05, "loss": 0.049, "step": 17500 }, { "epoch": 3.21, "learning_rate": 1.3603598463170723e-05, "loss": 0.0515, "step": 17501 }, { "epoch": 3.21, "learning_rate": 1.3602905904586405e-05, "loss": 0.0435, "step": 17502 }, { "epoch": 3.21, "learning_rate": 1.3602213326142786e-05, "loss": 0.0512, "step": 17503 }, { "epoch": 3.21, "learning_rate": 1.360152072784368e-05, "loss": 0.0477, "step": 17504 }, { "epoch": 3.21, "learning_rate": 1.3600828109692905e-05, "loss": 0.0495, "step": 17505 }, { "epoch": 3.21, "learning_rate": 1.3600135471694288e-05, "loss": 0.046, "step": 17506 }, { "epoch": 3.21, "learning_rate": 1.3599442813851633e-05, "loss": 0.0582, "step": 17507 }, { "epoch": 3.21, "learning_rate": 1.3598750136168768e-05, "loss": 0.0546, "step": 17508 }, { "epoch": 3.21, "learning_rate": 1.3598057438649509e-05, "loss": 0.0515, "step": 17509 }, { "epoch": 3.21, "learning_rate": 1.3597364721297672e-05, "loss": 0.0449, "step": 17510 }, { "epoch": 3.21, "learning_rate": 1.3596671984117073e-05, "loss": 0.0508, "step": 17511 }, { "epoch": 3.21, "learning_rate": 1.359597922711154e-05, "loss": 0.0542, "step": 17512 }, { "epoch": 3.21, "learning_rate": 1.3595286450284882e-05, "loss": 0.0586, "step": 17513 }, { "epoch": 3.21, "learning_rate": 1.359459365364092e-05, "loss": 0.0514, "step": 17514 }, { "epoch": 3.21, "learning_rate": 1.3593900837183477e-05, "loss": 0.0527, "step": 17515 }, { "epoch": 3.21, "learning_rate": 1.3593208000916365e-05, "loss": 0.0532, "step": 17516 }, { "epoch": 3.21, "learning_rate": 1.3592515144843405e-05, "loss": 0.056, "step": 17517 }, { "epoch": 3.21, "learning_rate": 1.3591822268968422e-05, "loss": 0.0463, "step": 17518 }, { "epoch": 3.21, "learning_rate": 1.3591129373295228e-05, "loss": 0.0446, "step": 17519 }, { "epoch": 3.21, "learning_rate": 1.3590436457827645e-05, "loss": 0.0451, "step": 17520 }, { "epoch": 3.21, "learning_rate": 1.3589743522569495e-05, "loss": 0.0503, "step": 17521 }, { "epoch": 3.21, "learning_rate": 1.3589050567524589e-05, "loss": 0.0554, "step": 17522 }, { "epoch": 3.21, "learning_rate": 1.3588357592696755e-05, "loss": 0.0527, "step": 17523 }, { "epoch": 3.21, "learning_rate": 1.3587664598089812e-05, "loss": 0.0554, "step": 17524 }, { "epoch": 3.21, "learning_rate": 1.3586971583707573e-05, "loss": 0.0581, "step": 17525 }, { "epoch": 3.21, "learning_rate": 1.3586278549553867e-05, "loss": 0.0574, "step": 17526 }, { "epoch": 3.21, "learning_rate": 1.3585585495632508e-05, "loss": 0.0471, "step": 17527 }, { "epoch": 3.21, "learning_rate": 1.3584892421947315e-05, "loss": 0.0508, "step": 17528 }, { "epoch": 3.21, "learning_rate": 1.3584199328502115e-05, "loss": 0.0599, "step": 17529 }, { "epoch": 3.21, "learning_rate": 1.3583506215300723e-05, "loss": 0.0484, "step": 17530 }, { "epoch": 3.21, "learning_rate": 1.358281308234696e-05, "loss": 0.0366, "step": 17531 }, { "epoch": 3.22, "learning_rate": 1.3582119929644649e-05, "loss": 0.0565, "step": 17532 }, { "epoch": 3.22, "learning_rate": 1.3581426757197607e-05, "loss": 0.0517, "step": 17533 }, { "epoch": 3.22, "learning_rate": 1.3580733565009659e-05, "loss": 0.0466, "step": 17534 }, { "epoch": 3.22, "learning_rate": 1.358004035308462e-05, "loss": 0.048, "step": 17535 }, { "epoch": 3.22, "learning_rate": 1.357934712142632e-05, "loss": 0.0492, "step": 17536 }, { "epoch": 3.22, "learning_rate": 1.357865387003857e-05, "loss": 0.0454, "step": 17537 }, { "epoch": 3.22, "learning_rate": 1.35779605989252e-05, "loss": 0.0596, "step": 17538 }, { "epoch": 3.22, "learning_rate": 1.3577267308090026e-05, "loss": 0.0454, "step": 17539 }, { "epoch": 3.22, "learning_rate": 1.3576573997536867e-05, "loss": 0.0479, "step": 17540 }, { "epoch": 3.22, "learning_rate": 1.3575880667269552e-05, "loss": 0.0485, "step": 17541 }, { "epoch": 3.22, "learning_rate": 1.3575187317291897e-05, "loss": 0.0602, "step": 17542 }, { "epoch": 3.22, "learning_rate": 1.357449394760773e-05, "loss": 0.0474, "step": 17543 }, { "epoch": 3.22, "learning_rate": 1.3573800558220867e-05, "loss": 0.0544, "step": 17544 }, { "epoch": 3.22, "learning_rate": 1.3573107149135129e-05, "loss": 0.0483, "step": 17545 }, { "epoch": 3.22, "learning_rate": 1.3572413720354343e-05, "loss": 0.0467, "step": 17546 }, { "epoch": 3.22, "learning_rate": 1.3571720271882327e-05, "loss": 0.0451, "step": 17547 }, { "epoch": 3.22, "learning_rate": 1.3571026803722906e-05, "loss": 0.0484, "step": 17548 }, { "epoch": 3.22, "learning_rate": 1.35703333158799e-05, "loss": 0.0545, "step": 17549 }, { "epoch": 3.22, "learning_rate": 1.3569639808357136e-05, "loss": 0.0577, "step": 17550 }, { "epoch": 3.22, "learning_rate": 1.3568946281158432e-05, "loss": 0.0517, "step": 17551 }, { "epoch": 3.22, "learning_rate": 1.3568252734287614e-05, "loss": 0.0493, "step": 17552 }, { "epoch": 3.22, "learning_rate": 1.3567559167748504e-05, "loss": 0.0478, "step": 17553 }, { "epoch": 3.22, "learning_rate": 1.3566865581544923e-05, "loss": 0.0523, "step": 17554 }, { "epoch": 3.22, "learning_rate": 1.3566171975680695e-05, "loss": 0.0625, "step": 17555 }, { "epoch": 3.22, "learning_rate": 1.3565478350159645e-05, "loss": 0.0442, "step": 17556 }, { "epoch": 3.22, "learning_rate": 1.3564784704985592e-05, "loss": 0.0472, "step": 17557 }, { "epoch": 3.22, "learning_rate": 1.3564091040162363e-05, "loss": 0.0568, "step": 17558 }, { "epoch": 3.22, "learning_rate": 1.3563397355693783e-05, "loss": 0.0506, "step": 17559 }, { "epoch": 3.22, "learning_rate": 1.356270365158367e-05, "loss": 0.0572, "step": 17560 }, { "epoch": 3.22, "learning_rate": 1.3562009927835853e-05, "loss": 0.061, "step": 17561 }, { "epoch": 3.22, "learning_rate": 1.3561316184454154e-05, "loss": 0.0454, "step": 17562 }, { "epoch": 3.22, "learning_rate": 1.3560622421442395e-05, "loss": 0.049, "step": 17563 }, { "epoch": 3.22, "learning_rate": 1.3559928638804402e-05, "loss": 0.046, "step": 17564 }, { "epoch": 3.22, "learning_rate": 1.3559234836544001e-05, "loss": 0.0432, "step": 17565 }, { "epoch": 3.22, "learning_rate": 1.3558541014665015e-05, "loss": 0.0563, "step": 17566 }, { "epoch": 3.22, "learning_rate": 1.3557847173171264e-05, "loss": 0.048, "step": 17567 }, { "epoch": 3.22, "learning_rate": 1.355715331206658e-05, "loss": 0.0487, "step": 17568 }, { "epoch": 3.22, "learning_rate": 1.355645943135478e-05, "loss": 0.0576, "step": 17569 }, { "epoch": 3.22, "learning_rate": 1.3555765531039692e-05, "loss": 0.0549, "step": 17570 }, { "epoch": 3.22, "learning_rate": 1.3555071611125143e-05, "loss": 0.0503, "step": 17571 }, { "epoch": 3.22, "learning_rate": 1.3554377671614957e-05, "loss": 0.053, "step": 17572 }, { "epoch": 3.22, "learning_rate": 1.3553683712512955e-05, "loss": 0.0466, "step": 17573 }, { "epoch": 3.22, "learning_rate": 1.355298973382297e-05, "loss": 0.0463, "step": 17574 }, { "epoch": 3.22, "learning_rate": 1.3552295735548817e-05, "loss": 0.0497, "step": 17575 }, { "epoch": 3.22, "learning_rate": 1.355160171769433e-05, "loss": 0.0588, "step": 17576 }, { "epoch": 3.22, "learning_rate": 1.3550907680263332e-05, "loss": 0.045, "step": 17577 }, { "epoch": 3.22, "learning_rate": 1.3550213623259643e-05, "loss": 0.0427, "step": 17578 }, { "epoch": 3.22, "learning_rate": 1.3549519546687097e-05, "loss": 0.0527, "step": 17579 }, { "epoch": 3.22, "learning_rate": 1.3548825450549515e-05, "loss": 0.0539, "step": 17580 }, { "epoch": 3.22, "learning_rate": 1.3548131334850725e-05, "loss": 0.0569, "step": 17581 }, { "epoch": 3.22, "learning_rate": 1.3547437199594549e-05, "loss": 0.0528, "step": 17582 }, { "epoch": 3.22, "learning_rate": 1.354674304478482e-05, "loss": 0.0539, "step": 17583 }, { "epoch": 3.22, "learning_rate": 1.3546048870425356e-05, "loss": 0.0549, "step": 17584 }, { "epoch": 3.22, "learning_rate": 1.354535467651999e-05, "loss": 0.0455, "step": 17585 }, { "epoch": 3.23, "learning_rate": 1.3544660463072548e-05, "loss": 0.0526, "step": 17586 }, { "epoch": 3.23, "learning_rate": 1.354396623008685e-05, "loss": 0.0542, "step": 17587 }, { "epoch": 3.23, "learning_rate": 1.3543271977566732e-05, "loss": 0.0545, "step": 17588 }, { "epoch": 3.23, "learning_rate": 1.3542577705516014e-05, "loss": 0.0459, "step": 17589 }, { "epoch": 3.23, "learning_rate": 1.3541883413938523e-05, "loss": 0.0552, "step": 17590 }, { "epoch": 3.23, "learning_rate": 1.3541189102838089e-05, "loss": 0.046, "step": 17591 }, { "epoch": 3.23, "learning_rate": 1.3540494772218537e-05, "loss": 0.053, "step": 17592 }, { "epoch": 3.23, "learning_rate": 1.3539800422083696e-05, "loss": 0.0499, "step": 17593 }, { "epoch": 3.23, "learning_rate": 1.353910605243739e-05, "loss": 0.0451, "step": 17594 }, { "epoch": 3.23, "learning_rate": 1.353841166328345e-05, "loss": 0.0482, "step": 17595 }, { "epoch": 3.23, "learning_rate": 1.3537717254625703e-05, "loss": 0.0473, "step": 17596 }, { "epoch": 3.23, "learning_rate": 1.3537022826467974e-05, "loss": 0.0415, "step": 17597 }, { "epoch": 3.23, "learning_rate": 1.3536328378814094e-05, "loss": 0.0612, "step": 17598 }, { "epoch": 3.23, "learning_rate": 1.3535633911667887e-05, "loss": 0.0519, "step": 17599 }, { "epoch": 3.23, "learning_rate": 1.3534939425033184e-05, "loss": 0.0617, "step": 17600 }, { "epoch": 3.23, "learning_rate": 1.3534244918913815e-05, "loss": 0.046, "step": 17601 }, { "epoch": 3.23, "learning_rate": 1.35335503933136e-05, "loss": 0.0496, "step": 17602 }, { "epoch": 3.23, "learning_rate": 1.3532855848236375e-05, "loss": 0.0492, "step": 17603 }, { "epoch": 3.23, "learning_rate": 1.3532161283685965e-05, "loss": 0.0517, "step": 17604 }, { "epoch": 3.23, "learning_rate": 1.35314666996662e-05, "loss": 0.0494, "step": 17605 }, { "epoch": 3.23, "learning_rate": 1.3530772096180909e-05, "loss": 0.0515, "step": 17606 }, { "epoch": 3.23, "learning_rate": 1.3530077473233918e-05, "loss": 0.0537, "step": 17607 }, { "epoch": 3.23, "learning_rate": 1.3529382830829057e-05, "loss": 0.045, "step": 17608 }, { "epoch": 3.23, "learning_rate": 1.3528688168970154e-05, "loss": 0.048, "step": 17609 }, { "epoch": 3.23, "learning_rate": 1.3527993487661041e-05, "loss": 0.0453, "step": 17610 }, { "epoch": 3.23, "learning_rate": 1.3527298786905546e-05, "loss": 0.0547, "step": 17611 }, { "epoch": 3.23, "learning_rate": 1.3526604066707496e-05, "loss": 0.0454, "step": 17612 }, { "epoch": 3.23, "learning_rate": 1.3525909327070724e-05, "loss": 0.0508, "step": 17613 }, { "epoch": 3.23, "learning_rate": 1.3525214567999055e-05, "loss": 0.0528, "step": 17614 }, { "epoch": 3.23, "learning_rate": 1.352451978949632e-05, "loss": 0.0429, "step": 17615 }, { "epoch": 3.23, "learning_rate": 1.352382499156635e-05, "loss": 0.0464, "step": 17616 }, { "epoch": 3.23, "learning_rate": 1.3523130174212973e-05, "loss": 0.0563, "step": 17617 }, { "epoch": 3.23, "learning_rate": 1.352243533744002e-05, "loss": 0.0457, "step": 17618 }, { "epoch": 3.23, "learning_rate": 1.3521740481251326e-05, "loss": 0.047, "step": 17619 }, { "epoch": 3.23, "learning_rate": 1.3521045605650712e-05, "loss": 0.048, "step": 17620 }, { "epoch": 3.23, "learning_rate": 1.3520350710642011e-05, "loss": 0.0408, "step": 17621 }, { "epoch": 3.23, "learning_rate": 1.3519655796229057e-05, "loss": 0.049, "step": 17622 }, { "epoch": 3.23, "learning_rate": 1.3518960862415678e-05, "loss": 0.0418, "step": 17623 }, { "epoch": 3.23, "learning_rate": 1.3518265909205704e-05, "loss": 0.058, "step": 17624 }, { "epoch": 3.23, "learning_rate": 1.3517570936602963e-05, "loss": 0.0492, "step": 17625 }, { "epoch": 3.23, "learning_rate": 1.351687594461129e-05, "loss": 0.0537, "step": 17626 }, { "epoch": 3.23, "learning_rate": 1.3516180933234514e-05, "loss": 0.0516, "step": 17627 }, { "epoch": 3.23, "learning_rate": 1.3515485902476467e-05, "loss": 0.0558, "step": 17628 }, { "epoch": 3.23, "learning_rate": 1.351479085234098e-05, "loss": 0.0514, "step": 17629 }, { "epoch": 3.23, "learning_rate": 1.3514095782831885e-05, "loss": 0.0484, "step": 17630 }, { "epoch": 3.23, "learning_rate": 1.3513400693953008e-05, "loss": 0.0438, "step": 17631 }, { "epoch": 3.23, "learning_rate": 1.3512705585708186e-05, "loss": 0.056, "step": 17632 }, { "epoch": 3.23, "learning_rate": 1.3512010458101248e-05, "loss": 0.0509, "step": 17633 }, { "epoch": 3.23, "learning_rate": 1.3511315311136027e-05, "loss": 0.0488, "step": 17634 }, { "epoch": 3.23, "learning_rate": 1.3510620144816351e-05, "loss": 0.0569, "step": 17635 }, { "epoch": 3.23, "learning_rate": 1.3509924959146059e-05, "loss": 0.0537, "step": 17636 }, { "epoch": 3.23, "learning_rate": 1.3509229754128975e-05, "loss": 0.0499, "step": 17637 }, { "epoch": 3.23, "learning_rate": 1.3508534529768935e-05, "loss": 0.0492, "step": 17638 }, { "epoch": 3.23, "learning_rate": 1.350783928606977e-05, "loss": 0.0405, "step": 17639 }, { "epoch": 3.23, "learning_rate": 1.3507144023035315e-05, "loss": 0.0504, "step": 17640 }, { "epoch": 3.24, "learning_rate": 1.3506448740669398e-05, "loss": 0.0508, "step": 17641 }, { "epoch": 3.24, "learning_rate": 1.3505753438975855e-05, "loss": 0.0462, "step": 17642 }, { "epoch": 3.24, "learning_rate": 1.3505058117958515e-05, "loss": 0.044, "step": 17643 }, { "epoch": 3.24, "learning_rate": 1.3504362777621211e-05, "loss": 0.0474, "step": 17644 }, { "epoch": 3.24, "learning_rate": 1.3503667417967783e-05, "loss": 0.0549, "step": 17645 }, { "epoch": 3.24, "learning_rate": 1.3502972039002053e-05, "loss": 0.0579, "step": 17646 }, { "epoch": 3.24, "learning_rate": 1.350227664072786e-05, "loss": 0.0506, "step": 17647 }, { "epoch": 3.24, "learning_rate": 1.3501581223149039e-05, "loss": 0.0439, "step": 17648 }, { "epoch": 3.24, "learning_rate": 1.3500885786269418e-05, "loss": 0.0548, "step": 17649 }, { "epoch": 3.24, "learning_rate": 1.350019033009283e-05, "loss": 0.0523, "step": 17650 }, { "epoch": 3.24, "learning_rate": 1.3499494854623117e-05, "loss": 0.0458, "step": 17651 }, { "epoch": 3.24, "learning_rate": 1.3498799359864103e-05, "loss": 0.0422, "step": 17652 }, { "epoch": 3.24, "learning_rate": 1.3498103845819624e-05, "loss": 0.0525, "step": 17653 }, { "epoch": 3.24, "learning_rate": 1.3497408312493518e-05, "loss": 0.0499, "step": 17654 }, { "epoch": 3.24, "learning_rate": 1.349671275988961e-05, "loss": 0.0508, "step": 17655 }, { "epoch": 3.24, "learning_rate": 1.3496017188011743e-05, "loss": 0.0452, "step": 17656 }, { "epoch": 3.24, "learning_rate": 1.3495321596863747e-05, "loss": 0.0516, "step": 17657 }, { "epoch": 3.24, "learning_rate": 1.3494625986449457e-05, "loss": 0.0447, "step": 17658 }, { "epoch": 3.24, "learning_rate": 1.3493930356772705e-05, "loss": 0.0522, "step": 17659 }, { "epoch": 3.24, "learning_rate": 1.3493234707837329e-05, "loss": 0.0562, "step": 17660 }, { "epoch": 3.24, "learning_rate": 1.349253903964716e-05, "loss": 0.0541, "step": 17661 }, { "epoch": 3.24, "learning_rate": 1.3491843352206033e-05, "loss": 0.0565, "step": 17662 }, { "epoch": 3.24, "learning_rate": 1.3491147645517786e-05, "loss": 0.0536, "step": 17663 }, { "epoch": 3.24, "learning_rate": 1.3490451919586249e-05, "loss": 0.0503, "step": 17664 }, { "epoch": 3.24, "learning_rate": 1.3489756174415263e-05, "loss": 0.0481, "step": 17665 }, { "epoch": 3.24, "learning_rate": 1.3489060410008656e-05, "loss": 0.0389, "step": 17666 }, { "epoch": 3.24, "learning_rate": 1.3488364626370267e-05, "loss": 0.0465, "step": 17667 }, { "epoch": 3.24, "learning_rate": 1.3487668823503928e-05, "loss": 0.0487, "step": 17668 }, { "epoch": 3.24, "learning_rate": 1.3486973001413484e-05, "loss": 0.0636, "step": 17669 }, { "epoch": 3.24, "learning_rate": 1.3486277160102756e-05, "loss": 0.049, "step": 17670 }, { "epoch": 3.24, "learning_rate": 1.3485581299575587e-05, "loss": 0.0383, "step": 17671 }, { "epoch": 3.24, "learning_rate": 1.3484885419835818e-05, "loss": 0.0604, "step": 17672 }, { "epoch": 3.24, "learning_rate": 1.3484189520887273e-05, "loss": 0.0464, "step": 17673 }, { "epoch": 3.24, "learning_rate": 1.3483493602733797e-05, "loss": 0.0479, "step": 17674 }, { "epoch": 3.24, "learning_rate": 1.3482797665379221e-05, "loss": 0.0512, "step": 17675 }, { "epoch": 3.24, "learning_rate": 1.3482101708827384e-05, "loss": 0.0476, "step": 17676 }, { "epoch": 3.24, "learning_rate": 1.3481405733082118e-05, "loss": 0.0539, "step": 17677 }, { "epoch": 3.24, "learning_rate": 1.3480709738147266e-05, "loss": 0.045, "step": 17678 }, { "epoch": 3.24, "learning_rate": 1.3480013724026657e-05, "loss": 0.054, "step": 17679 }, { "epoch": 3.24, "learning_rate": 1.3479317690724135e-05, "loss": 0.0588, "step": 17680 }, { "epoch": 3.24, "learning_rate": 1.3478621638243528e-05, "loss": 0.0474, "step": 17681 }, { "epoch": 3.24, "learning_rate": 1.3477925566588679e-05, "loss": 0.0472, "step": 17682 }, { "epoch": 3.24, "learning_rate": 1.3477229475763422e-05, "loss": 0.0522, "step": 17683 }, { "epoch": 3.24, "learning_rate": 1.3476533365771596e-05, "loss": 0.0542, "step": 17684 }, { "epoch": 3.24, "learning_rate": 1.3475837236617036e-05, "loss": 0.0474, "step": 17685 }, { "epoch": 3.24, "learning_rate": 1.3475141088303579e-05, "loss": 0.0486, "step": 17686 }, { "epoch": 3.24, "learning_rate": 1.3474444920835063e-05, "loss": 0.0449, "step": 17687 }, { "epoch": 3.24, "learning_rate": 1.3473748734215325e-05, "loss": 0.0483, "step": 17688 }, { "epoch": 3.24, "learning_rate": 1.3473052528448203e-05, "loss": 0.0433, "step": 17689 }, { "epoch": 3.24, "learning_rate": 1.3472356303537534e-05, "loss": 0.0439, "step": 17690 }, { "epoch": 3.24, "learning_rate": 1.3471660059487157e-05, "loss": 0.0458, "step": 17691 }, { "epoch": 3.24, "learning_rate": 1.3470963796300909e-05, "loss": 0.0509, "step": 17692 }, { "epoch": 3.24, "learning_rate": 1.3470267513982625e-05, "loss": 0.0532, "step": 17693 }, { "epoch": 3.24, "learning_rate": 1.3469571212536145e-05, "loss": 0.0516, "step": 17694 }, { "epoch": 3.25, "learning_rate": 1.3468874891965307e-05, "loss": 0.0433, "step": 17695 }, { "epoch": 3.25, "learning_rate": 1.3468178552273954e-05, "loss": 0.0466, "step": 17696 }, { "epoch": 3.25, "learning_rate": 1.3467482193465915e-05, "loss": 0.0516, "step": 17697 }, { "epoch": 3.25, "learning_rate": 1.3466785815545033e-05, "loss": 0.0556, "step": 17698 }, { "epoch": 3.25, "learning_rate": 1.3466089418515147e-05, "loss": 0.0637, "step": 17699 }, { "epoch": 3.25, "learning_rate": 1.3465393002380096e-05, "loss": 0.0501, "step": 17700 }, { "epoch": 3.25, "learning_rate": 1.3464696567143718e-05, "loss": 0.0433, "step": 17701 }, { "epoch": 3.25, "learning_rate": 1.346400011280985e-05, "loss": 0.0464, "step": 17702 }, { "epoch": 3.25, "learning_rate": 1.3463303639382331e-05, "loss": 0.0528, "step": 17703 }, { "epoch": 3.25, "learning_rate": 1.3462607146865001e-05, "loss": 0.0535, "step": 17704 }, { "epoch": 3.25, "learning_rate": 1.3461910635261703e-05, "loss": 0.0448, "step": 17705 }, { "epoch": 3.25, "learning_rate": 1.3461214104576267e-05, "loss": 0.046, "step": 17706 }, { "epoch": 3.25, "learning_rate": 1.3460517554812543e-05, "loss": 0.048, "step": 17707 }, { "epoch": 3.25, "learning_rate": 1.3459820985974363e-05, "loss": 0.0491, "step": 17708 }, { "epoch": 3.25, "learning_rate": 1.345912439806557e-05, "loss": 0.0494, "step": 17709 }, { "epoch": 3.25, "learning_rate": 1.3458427791090001e-05, "loss": 0.0443, "step": 17710 }, { "epoch": 3.25, "learning_rate": 1.3457731165051498e-05, "loss": 0.0602, "step": 17711 }, { "epoch": 3.25, "learning_rate": 1.3457034519953898e-05, "loss": 0.0515, "step": 17712 }, { "epoch": 3.25, "learning_rate": 1.3456337855801044e-05, "loss": 0.0473, "step": 17713 }, { "epoch": 3.25, "learning_rate": 1.3455641172596774e-05, "loss": 0.0495, "step": 17714 }, { "epoch": 3.25, "learning_rate": 1.345494447034493e-05, "loss": 0.0421, "step": 17715 }, { "epoch": 3.25, "learning_rate": 1.345424774904935e-05, "loss": 0.0424, "step": 17716 }, { "epoch": 3.25, "learning_rate": 1.3453551008713877e-05, "loss": 0.0492, "step": 17717 }, { "epoch": 3.25, "learning_rate": 1.345285424934235e-05, "loss": 0.0399, "step": 17718 }, { "epoch": 3.25, "learning_rate": 1.3452157470938606e-05, "loss": 0.0532, "step": 17719 }, { "epoch": 3.25, "learning_rate": 1.3451460673506493e-05, "loss": 0.0463, "step": 17720 }, { "epoch": 3.25, "learning_rate": 1.3450763857049844e-05, "loss": 0.0462, "step": 17721 }, { "epoch": 3.25, "learning_rate": 1.3450067021572508e-05, "loss": 0.0465, "step": 17722 }, { "epoch": 3.25, "learning_rate": 1.344937016707832e-05, "loss": 0.0465, "step": 17723 }, { "epoch": 3.25, "learning_rate": 1.3448673293571121e-05, "loss": 0.0437, "step": 17724 }, { "epoch": 3.25, "learning_rate": 1.3447976401054756e-05, "loss": 0.041, "step": 17725 }, { "epoch": 3.25, "learning_rate": 1.3447279489533065e-05, "loss": 0.0553, "step": 17726 }, { "epoch": 3.25, "learning_rate": 1.3446582559009888e-05, "loss": 0.044, "step": 17727 }, { "epoch": 3.25, "learning_rate": 1.344588560948907e-05, "loss": 0.0536, "step": 17728 }, { "epoch": 3.25, "learning_rate": 1.3445188640974448e-05, "loss": 0.0487, "step": 17729 }, { "epoch": 3.25, "learning_rate": 1.344449165346986e-05, "loss": 0.0496, "step": 17730 }, { "epoch": 3.25, "learning_rate": 1.3443794646979161e-05, "loss": 0.0497, "step": 17731 }, { "epoch": 3.25, "learning_rate": 1.3443097621506178e-05, "loss": 0.0456, "step": 17732 }, { "epoch": 3.25, "learning_rate": 1.3442400577054767e-05, "loss": 0.0465, "step": 17733 }, { "epoch": 3.25, "learning_rate": 1.3441703513628758e-05, "loss": 0.0515, "step": 17734 }, { "epoch": 3.25, "learning_rate": 1.3441006431232002e-05, "loss": 0.054, "step": 17735 }, { "epoch": 3.25, "learning_rate": 1.3440309329868337e-05, "loss": 0.0462, "step": 17736 }, { "epoch": 3.25, "learning_rate": 1.3439612209541608e-05, "loss": 0.0409, "step": 17737 }, { "epoch": 3.25, "learning_rate": 1.3438915070255654e-05, "loss": 0.0522, "step": 17738 }, { "epoch": 3.25, "learning_rate": 1.3438217912014318e-05, "loss": 0.0484, "step": 17739 }, { "epoch": 3.25, "learning_rate": 1.3437520734821447e-05, "loss": 0.0504, "step": 17740 }, { "epoch": 3.25, "learning_rate": 1.3436823538680882e-05, "loss": 0.056, "step": 17741 }, { "epoch": 3.25, "learning_rate": 1.3436126323596462e-05, "loss": 0.0495, "step": 17742 }, { "epoch": 3.25, "learning_rate": 1.3435429089572036e-05, "loss": 0.0532, "step": 17743 }, { "epoch": 3.25, "learning_rate": 1.343473183661144e-05, "loss": 0.0403, "step": 17744 }, { "epoch": 3.25, "learning_rate": 1.3434034564718526e-05, "loss": 0.055, "step": 17745 }, { "epoch": 3.25, "learning_rate": 1.3433337273897133e-05, "loss": 0.0482, "step": 17746 }, { "epoch": 3.25, "learning_rate": 1.3432639964151104e-05, "loss": 0.0606, "step": 17747 }, { "epoch": 3.25, "learning_rate": 1.343194263548428e-05, "loss": 0.0446, "step": 17748 }, { "epoch": 3.25, "learning_rate": 1.3431245287900514e-05, "loss": 0.0497, "step": 17749 }, { "epoch": 3.26, "learning_rate": 1.3430547921403636e-05, "loss": 0.0512, "step": 17750 }, { "epoch": 3.26, "learning_rate": 1.3429850535997502e-05, "loss": 0.0505, "step": 17751 }, { "epoch": 3.26, "learning_rate": 1.3429153131685952e-05, "loss": 0.0429, "step": 17752 }, { "epoch": 3.26, "learning_rate": 1.3428455708472827e-05, "loss": 0.0466, "step": 17753 }, { "epoch": 3.26, "learning_rate": 1.3427758266361974e-05, "loss": 0.0556, "step": 17754 }, { "epoch": 3.26, "learning_rate": 1.3427060805357242e-05, "loss": 0.0435, "step": 17755 }, { "epoch": 3.26, "learning_rate": 1.3426363325462464e-05, "loss": 0.0474, "step": 17756 }, { "epoch": 3.26, "learning_rate": 1.3425665826681495e-05, "loss": 0.0501, "step": 17757 }, { "epoch": 3.26, "learning_rate": 1.3424968309018175e-05, "loss": 0.0449, "step": 17758 }, { "epoch": 3.26, "learning_rate": 1.3424270772476349e-05, "loss": 0.0493, "step": 17759 }, { "epoch": 3.26, "learning_rate": 1.3423573217059863e-05, "loss": 0.0514, "step": 17760 }, { "epoch": 3.26, "learning_rate": 1.342287564277256e-05, "loss": 0.0563, "step": 17761 }, { "epoch": 3.26, "learning_rate": 1.3422178049618287e-05, "loss": 0.0542, "step": 17762 }, { "epoch": 3.26, "learning_rate": 1.3421480437600887e-05, "loss": 0.0485, "step": 17763 }, { "epoch": 3.26, "learning_rate": 1.3420782806724209e-05, "loss": 0.0469, "step": 17764 }, { "epoch": 3.26, "learning_rate": 1.3420085156992093e-05, "loss": 0.0525, "step": 17765 }, { "epoch": 3.26, "learning_rate": 1.3419387488408391e-05, "loss": 0.0503, "step": 17766 }, { "epoch": 3.26, "learning_rate": 1.3418689800976943e-05, "loss": 0.0595, "step": 17767 }, { "epoch": 3.26, "learning_rate": 1.3417992094701598e-05, "loss": 0.0469, "step": 17768 }, { "epoch": 3.26, "learning_rate": 1.34172943695862e-05, "loss": 0.0476, "step": 17769 }, { "epoch": 3.26, "learning_rate": 1.3416596625634595e-05, "loss": 0.0514, "step": 17770 }, { "epoch": 3.26, "learning_rate": 1.341589886285063e-05, "loss": 0.0466, "step": 17771 }, { "epoch": 3.26, "learning_rate": 1.3415201081238147e-05, "loss": 0.0587, "step": 17772 }, { "epoch": 3.26, "learning_rate": 1.3414503280801002e-05, "loss": 0.0474, "step": 17773 }, { "epoch": 3.26, "learning_rate": 1.3413805461543027e-05, "loss": 0.0495, "step": 17774 }, { "epoch": 3.26, "learning_rate": 1.3413107623468085e-05, "loss": 0.0502, "step": 17775 }, { "epoch": 3.26, "learning_rate": 1.3412409766580008e-05, "loss": 0.0484, "step": 17776 }, { "epoch": 3.26, "learning_rate": 1.3411711890882651e-05, "loss": 0.0462, "step": 17777 }, { "epoch": 3.26, "learning_rate": 1.3411013996379857e-05, "loss": 0.0532, "step": 17778 }, { "epoch": 3.26, "learning_rate": 1.3410316083075473e-05, "loss": 0.0522, "step": 17779 }, { "epoch": 3.26, "learning_rate": 1.3409618150973349e-05, "loss": 0.0496, "step": 17780 }, { "epoch": 3.26, "learning_rate": 1.3408920200077327e-05, "loss": 0.0478, "step": 17781 }, { "epoch": 3.26, "learning_rate": 1.340822223039126e-05, "loss": 0.0471, "step": 17782 }, { "epoch": 3.26, "learning_rate": 1.340752424191899e-05, "loss": 0.0474, "step": 17783 }, { "epoch": 3.26, "learning_rate": 1.3406826234664369e-05, "loss": 0.0475, "step": 17784 }, { "epoch": 3.26, "learning_rate": 1.3406128208631239e-05, "loss": 0.0471, "step": 17785 }, { "epoch": 3.26, "learning_rate": 1.3405430163823454e-05, "loss": 0.0476, "step": 17786 }, { "epoch": 3.26, "learning_rate": 1.3404732100244857e-05, "loss": 0.0503, "step": 17787 }, { "epoch": 3.26, "learning_rate": 1.3404034017899297e-05, "loss": 0.0472, "step": 17788 }, { "epoch": 3.26, "learning_rate": 1.3403335916790617e-05, "loss": 0.0495, "step": 17789 }, { "epoch": 3.26, "learning_rate": 1.3402637796922675e-05, "loss": 0.0518, "step": 17790 }, { "epoch": 3.26, "learning_rate": 1.3401939658299315e-05, "loss": 0.0531, "step": 17791 }, { "epoch": 3.26, "learning_rate": 1.3401241500924377e-05, "loss": 0.0489, "step": 17792 }, { "epoch": 3.26, "learning_rate": 1.3400543324801722e-05, "loss": 0.0586, "step": 17793 }, { "epoch": 3.26, "learning_rate": 1.3399845129935191e-05, "loss": 0.0434, "step": 17794 }, { "epoch": 3.26, "learning_rate": 1.3399146916328634e-05, "loss": 0.0485, "step": 17795 }, { "epoch": 3.26, "learning_rate": 1.3398448683985901e-05, "loss": 0.0548, "step": 17796 }, { "epoch": 3.26, "learning_rate": 1.3397750432910837e-05, "loss": 0.0443, "step": 17797 }, { "epoch": 3.26, "learning_rate": 1.3397052163107291e-05, "loss": 0.0545, "step": 17798 }, { "epoch": 3.26, "learning_rate": 1.339635387457912e-05, "loss": 0.0564, "step": 17799 }, { "epoch": 3.26, "learning_rate": 1.3395655567330164e-05, "loss": 0.0452, "step": 17800 }, { "epoch": 3.26, "learning_rate": 1.3394957241364273e-05, "loss": 0.0454, "step": 17801 }, { "epoch": 3.26, "learning_rate": 1.3394258896685299e-05, "loss": 0.0519, "step": 17802 }, { "epoch": 3.26, "learning_rate": 1.3393560533297092e-05, "loss": 0.0534, "step": 17803 }, { "epoch": 3.26, "learning_rate": 1.3392862151203499e-05, "loss": 0.0468, "step": 17804 }, { "epoch": 3.27, "learning_rate": 1.3392163750408371e-05, "loss": 0.0369, "step": 17805 }, { "epoch": 3.27, "learning_rate": 1.3391465330915557e-05, "loss": 0.046, "step": 17806 }, { "epoch": 3.27, "learning_rate": 1.3390766892728904e-05, "loss": 0.0498, "step": 17807 }, { "epoch": 3.27, "learning_rate": 1.3390068435852267e-05, "loss": 0.0531, "step": 17808 }, { "epoch": 3.27, "learning_rate": 1.3389369960289492e-05, "loss": 0.0554, "step": 17809 }, { "epoch": 3.27, "learning_rate": 1.338867146604443e-05, "loss": 0.0497, "step": 17810 }, { "epoch": 3.27, "learning_rate": 1.3387972953120933e-05, "loss": 0.0436, "step": 17811 }, { "epoch": 3.27, "learning_rate": 1.338727442152285e-05, "loss": 0.0458, "step": 17812 }, { "epoch": 3.27, "learning_rate": 1.338657587125403e-05, "loss": 0.0459, "step": 17813 }, { "epoch": 3.27, "learning_rate": 1.3385877302318324e-05, "loss": 0.0474, "step": 17814 }, { "epoch": 3.27, "learning_rate": 1.3385178714719584e-05, "loss": 0.047, "step": 17815 }, { "epoch": 3.27, "learning_rate": 1.338448010846166e-05, "loss": 0.0531, "step": 17816 }, { "epoch": 3.27, "learning_rate": 1.33837814835484e-05, "loss": 0.0461, "step": 17817 }, { "epoch": 3.27, "learning_rate": 1.338308283998366e-05, "loss": 0.0463, "step": 17818 }, { "epoch": 3.27, "learning_rate": 1.3382384177771286e-05, "loss": 0.0488, "step": 17819 }, { "epoch": 3.27, "learning_rate": 1.3381685496915133e-05, "loss": 0.0429, "step": 17820 }, { "epoch": 3.27, "learning_rate": 1.3380986797419049e-05, "loss": 0.0413, "step": 17821 }, { "epoch": 3.27, "learning_rate": 1.3380288079286887e-05, "loss": 0.047, "step": 17822 }, { "epoch": 3.27, "learning_rate": 1.3379589342522498e-05, "loss": 0.0451, "step": 17823 }, { "epoch": 3.27, "learning_rate": 1.3378890587129733e-05, "loss": 0.052, "step": 17824 }, { "epoch": 3.27, "learning_rate": 1.3378191813112444e-05, "loss": 0.0541, "step": 17825 }, { "epoch": 3.27, "learning_rate": 1.3377493020474485e-05, "loss": 0.0536, "step": 17826 }, { "epoch": 3.27, "learning_rate": 1.3376794209219702e-05, "loss": 0.0513, "step": 17827 }, { "epoch": 3.27, "learning_rate": 1.3376095379351952e-05, "loss": 0.0513, "step": 17828 }, { "epoch": 3.27, "learning_rate": 1.3375396530875084e-05, "loss": 0.0433, "step": 17829 }, { "epoch": 3.27, "learning_rate": 1.3374697663792952e-05, "loss": 0.0555, "step": 17830 }, { "epoch": 3.27, "learning_rate": 1.3373998778109404e-05, "loss": 0.0505, "step": 17831 }, { "epoch": 3.27, "learning_rate": 1.3373299873828303e-05, "loss": 0.0467, "step": 17832 }, { "epoch": 3.27, "learning_rate": 1.337260095095349e-05, "loss": 0.0469, "step": 17833 }, { "epoch": 3.27, "learning_rate": 1.3371902009488819e-05, "loss": 0.051, "step": 17834 }, { "epoch": 3.27, "learning_rate": 1.337120304943815e-05, "loss": 0.0487, "step": 17835 }, { "epoch": 3.27, "learning_rate": 1.3370504070805325e-05, "loss": 0.0593, "step": 17836 }, { "epoch": 3.27, "learning_rate": 1.336980507359421e-05, "loss": 0.045, "step": 17837 }, { "epoch": 3.27, "learning_rate": 1.3369106057808645e-05, "loss": 0.051, "step": 17838 }, { "epoch": 3.27, "learning_rate": 1.336840702345249e-05, "loss": 0.045, "step": 17839 }, { "epoch": 3.27, "learning_rate": 1.3367707970529595e-05, "loss": 0.0394, "step": 17840 }, { "epoch": 3.27, "learning_rate": 1.3367008899043819e-05, "loss": 0.0475, "step": 17841 }, { "epoch": 3.27, "learning_rate": 1.3366309808999005e-05, "loss": 0.0482, "step": 17842 }, { "epoch": 3.27, "learning_rate": 1.3365610700399017e-05, "loss": 0.048, "step": 17843 }, { "epoch": 3.27, "learning_rate": 1.3364911573247706e-05, "loss": 0.0446, "step": 17844 }, { "epoch": 3.27, "learning_rate": 1.3364212427548917e-05, "loss": 0.0568, "step": 17845 }, { "epoch": 3.27, "learning_rate": 1.3363513263306516e-05, "loss": 0.0551, "step": 17846 }, { "epoch": 3.27, "learning_rate": 1.3362814080524346e-05, "loss": 0.0462, "step": 17847 }, { "epoch": 3.27, "learning_rate": 1.3362114879206271e-05, "loss": 0.0506, "step": 17848 }, { "epoch": 3.27, "learning_rate": 1.3361415659356137e-05, "loss": 0.0505, "step": 17849 }, { "epoch": 3.27, "learning_rate": 1.3360716420977803e-05, "loss": 0.0426, "step": 17850 }, { "epoch": 3.27, "learning_rate": 1.3360017164075119e-05, "loss": 0.0414, "step": 17851 }, { "epoch": 3.27, "learning_rate": 1.3359317888651944e-05, "loss": 0.0459, "step": 17852 }, { "epoch": 3.27, "learning_rate": 1.335861859471213e-05, "loss": 0.0501, "step": 17853 }, { "epoch": 3.27, "learning_rate": 1.335791928225953e-05, "loss": 0.054, "step": 17854 }, { "epoch": 3.27, "learning_rate": 1.3357219951297999e-05, "loss": 0.0422, "step": 17855 }, { "epoch": 3.27, "learning_rate": 1.3356520601831398e-05, "loss": 0.0448, "step": 17856 }, { "epoch": 3.27, "learning_rate": 1.3355821233863573e-05, "loss": 0.0442, "step": 17857 }, { "epoch": 3.27, "learning_rate": 1.3355121847398382e-05, "loss": 0.0521, "step": 17858 }, { "epoch": 3.28, "learning_rate": 1.3354422442439683e-05, "loss": 0.0434, "step": 17859 }, { "epoch": 3.28, "learning_rate": 1.3353723018991328e-05, "loss": 0.0522, "step": 17860 }, { "epoch": 3.28, "learning_rate": 1.3353023577057174e-05, "loss": 0.0433, "step": 17861 }, { "epoch": 3.28, "learning_rate": 1.3352324116641073e-05, "loss": 0.0457, "step": 17862 }, { "epoch": 3.28, "learning_rate": 1.3351624637746885e-05, "loss": 0.0523, "step": 17863 }, { "epoch": 3.28, "learning_rate": 1.3350925140378465e-05, "loss": 0.0473, "step": 17864 }, { "epoch": 3.28, "learning_rate": 1.3350225624539665e-05, "loss": 0.0545, "step": 17865 }, { "epoch": 3.28, "learning_rate": 1.3349526090234343e-05, "loss": 0.05, "step": 17866 }, { "epoch": 3.28, "learning_rate": 1.3348826537466355e-05, "loss": 0.0468, "step": 17867 }, { "epoch": 3.28, "learning_rate": 1.3348126966239558e-05, "loss": 0.0508, "step": 17868 }, { "epoch": 3.28, "learning_rate": 1.3347427376557805e-05, "loss": 0.0549, "step": 17869 }, { "epoch": 3.28, "learning_rate": 1.3346727768424952e-05, "loss": 0.0454, "step": 17870 }, { "epoch": 3.28, "learning_rate": 1.334602814184486e-05, "loss": 0.0534, "step": 17871 }, { "epoch": 3.28, "learning_rate": 1.3345328496821384e-05, "loss": 0.0429, "step": 17872 }, { "epoch": 3.28, "learning_rate": 1.3344628833358375e-05, "loss": 0.0499, "step": 17873 }, { "epoch": 3.28, "learning_rate": 1.3343929151459696e-05, "loss": 0.0519, "step": 17874 }, { "epoch": 3.28, "learning_rate": 1.3343229451129201e-05, "loss": 0.0543, "step": 17875 }, { "epoch": 3.28, "learning_rate": 1.3342529732370745e-05, "loss": 0.0486, "step": 17876 }, { "epoch": 3.28, "learning_rate": 1.3341829995188189e-05, "loss": 0.0505, "step": 17877 }, { "epoch": 3.28, "learning_rate": 1.3341130239585387e-05, "loss": 0.0441, "step": 17878 }, { "epoch": 3.28, "learning_rate": 1.3340430465566197e-05, "loss": 0.0456, "step": 17879 }, { "epoch": 3.28, "learning_rate": 1.3339730673134475e-05, "loss": 0.0472, "step": 17880 }, { "epoch": 3.28, "learning_rate": 1.333903086229408e-05, "loss": 0.0488, "step": 17881 }, { "epoch": 3.28, "learning_rate": 1.333833103304887e-05, "loss": 0.044, "step": 17882 }, { "epoch": 3.28, "learning_rate": 1.3337631185402699e-05, "loss": 0.044, "step": 17883 }, { "epoch": 3.28, "learning_rate": 1.3336931319359428e-05, "loss": 0.0574, "step": 17884 }, { "epoch": 3.28, "learning_rate": 1.3336231434922915e-05, "loss": 0.0527, "step": 17885 }, { "epoch": 3.28, "learning_rate": 1.3335531532097012e-05, "loss": 0.0606, "step": 17886 }, { "epoch": 3.28, "learning_rate": 1.3334831610885584e-05, "loss": 0.0371, "step": 17887 }, { "epoch": 3.28, "learning_rate": 1.3334131671292486e-05, "loss": 0.0479, "step": 17888 }, { "epoch": 3.28, "learning_rate": 1.3333431713321575e-05, "loss": 0.0585, "step": 17889 }, { "epoch": 3.28, "learning_rate": 1.333273173697671e-05, "loss": 0.0456, "step": 17890 }, { "epoch": 3.28, "learning_rate": 1.333203174226175e-05, "loss": 0.0493, "step": 17891 }, { "epoch": 3.28, "learning_rate": 1.3331331729180554e-05, "loss": 0.039, "step": 17892 }, { "epoch": 3.28, "learning_rate": 1.3330631697736977e-05, "loss": 0.0451, "step": 17893 }, { "epoch": 3.28, "learning_rate": 1.3329931647934884e-05, "loss": 0.0464, "step": 17894 }, { "epoch": 3.28, "learning_rate": 1.3329231579778126e-05, "loss": 0.0485, "step": 17895 }, { "epoch": 3.28, "learning_rate": 1.3328531493270566e-05, "loss": 0.0438, "step": 17896 }, { "epoch": 3.28, "learning_rate": 1.3327831388416066e-05, "loss": 0.0541, "step": 17897 }, { "epoch": 3.28, "learning_rate": 1.3327131265218479e-05, "loss": 0.0408, "step": 17898 }, { "epoch": 3.28, "learning_rate": 1.3326431123681667e-05, "loss": 0.0544, "step": 17899 }, { "epoch": 3.28, "learning_rate": 1.3325730963809487e-05, "loss": 0.0405, "step": 17900 }, { "epoch": 3.28, "learning_rate": 1.3325030785605804e-05, "loss": 0.0492, "step": 17901 }, { "epoch": 3.28, "learning_rate": 1.3324330589074468e-05, "loss": 0.0482, "step": 17902 }, { "epoch": 3.28, "learning_rate": 1.3323630374219347e-05, "loss": 0.0528, "step": 17903 }, { "epoch": 3.28, "learning_rate": 1.3322930141044297e-05, "loss": 0.0433, "step": 17904 }, { "epoch": 3.28, "learning_rate": 1.3322229889553178e-05, "loss": 0.0492, "step": 17905 }, { "epoch": 3.28, "learning_rate": 1.3321529619749849e-05, "loss": 0.0459, "step": 17906 }, { "epoch": 3.28, "learning_rate": 1.3320829331638173e-05, "loss": 0.0606, "step": 17907 }, { "epoch": 3.28, "learning_rate": 1.3320129025222008e-05, "loss": 0.0495, "step": 17908 }, { "epoch": 3.28, "learning_rate": 1.3319428700505212e-05, "loss": 0.0426, "step": 17909 }, { "epoch": 3.28, "learning_rate": 1.331872835749165e-05, "loss": 0.0448, "step": 17910 }, { "epoch": 3.28, "learning_rate": 1.3318027996185177e-05, "loss": 0.0494, "step": 17911 }, { "epoch": 3.28, "learning_rate": 1.331732761658966e-05, "loss": 0.0572, "step": 17912 }, { "epoch": 3.28, "learning_rate": 1.3316627218708948e-05, "loss": 0.0454, "step": 17913 }, { "epoch": 3.29, "learning_rate": 1.3315926802546914e-05, "loss": 0.0508, "step": 17914 }, { "epoch": 3.29, "learning_rate": 1.3315226368107414e-05, "loss": 0.0409, "step": 17915 }, { "epoch": 3.29, "learning_rate": 1.3314525915394308e-05, "loss": 0.051, "step": 17916 }, { "epoch": 3.29, "learning_rate": 1.3313825444411456e-05, "loss": 0.0436, "step": 17917 }, { "epoch": 3.29, "learning_rate": 1.3313124955162724e-05, "loss": 0.0439, "step": 17918 }, { "epoch": 3.29, "learning_rate": 1.3312424447651968e-05, "loss": 0.0506, "step": 17919 }, { "epoch": 3.29, "learning_rate": 1.3311723921883048e-05, "loss": 0.0429, "step": 17920 }, { "epoch": 3.29, "learning_rate": 1.3311023377859834e-05, "loss": 0.0436, "step": 17921 }, { "epoch": 3.29, "learning_rate": 1.3310322815586174e-05, "loss": 0.0464, "step": 17922 }, { "epoch": 3.29, "learning_rate": 1.3309622235065941e-05, "loss": 0.0499, "step": 17923 }, { "epoch": 3.29, "learning_rate": 1.3308921636302995e-05, "loss": 0.0444, "step": 17924 }, { "epoch": 3.29, "learning_rate": 1.3308221019301192e-05, "loss": 0.0533, "step": 17925 }, { "epoch": 3.29, "learning_rate": 1.3307520384064399e-05, "loss": 0.0497, "step": 17926 }, { "epoch": 3.29, "learning_rate": 1.3306819730596476e-05, "loss": 0.0544, "step": 17927 }, { "epoch": 3.29, "learning_rate": 1.3306119058901285e-05, "loss": 0.049, "step": 17928 }, { "epoch": 3.29, "learning_rate": 1.3305418368982687e-05, "loss": 0.0567, "step": 17929 }, { "epoch": 3.29, "learning_rate": 1.3304717660844548e-05, "loss": 0.0418, "step": 17930 }, { "epoch": 3.29, "learning_rate": 1.3304016934490724e-05, "loss": 0.0416, "step": 17931 }, { "epoch": 3.29, "learning_rate": 1.3303316189925086e-05, "loss": 0.0524, "step": 17932 }, { "epoch": 3.29, "learning_rate": 1.3302615427151488e-05, "loss": 0.0452, "step": 17933 }, { "epoch": 3.29, "learning_rate": 1.3301914646173799e-05, "loss": 0.0478, "step": 17934 }, { "epoch": 3.29, "learning_rate": 1.3301213846995876e-05, "loss": 0.0414, "step": 17935 }, { "epoch": 3.29, "learning_rate": 1.3300513029621589e-05, "loss": 0.054, "step": 17936 }, { "epoch": 3.29, "learning_rate": 1.3299812194054793e-05, "loss": 0.0537, "step": 17937 }, { "epoch": 3.29, "learning_rate": 1.3299111340299357e-05, "loss": 0.0565, "step": 17938 }, { "epoch": 3.29, "learning_rate": 1.3298410468359144e-05, "loss": 0.0547, "step": 17939 }, { "epoch": 3.29, "learning_rate": 1.3297709578238012e-05, "loss": 0.0536, "step": 17940 }, { "epoch": 3.29, "learning_rate": 1.3297008669939829e-05, "loss": 0.0439, "step": 17941 }, { "epoch": 3.29, "learning_rate": 1.3296307743468456e-05, "loss": 0.0471, "step": 17942 }, { "epoch": 3.29, "learning_rate": 1.3295606798827758e-05, "loss": 0.0484, "step": 17943 }, { "epoch": 3.29, "learning_rate": 1.3294905836021594e-05, "loss": 0.0536, "step": 17944 }, { "epoch": 3.29, "learning_rate": 1.3294204855053837e-05, "loss": 0.0417, "step": 17945 }, { "epoch": 3.29, "learning_rate": 1.3293503855928344e-05, "loss": 0.0446, "step": 17946 }, { "epoch": 3.29, "learning_rate": 1.3292802838648981e-05, "loss": 0.0477, "step": 17947 }, { "epoch": 3.29, "learning_rate": 1.3292101803219613e-05, "loss": 0.0425, "step": 17948 }, { "epoch": 3.29, "learning_rate": 1.32914007496441e-05, "loss": 0.0557, "step": 17949 }, { "epoch": 3.29, "learning_rate": 1.3290699677926313e-05, "loss": 0.0499, "step": 17950 }, { "epoch": 3.29, "learning_rate": 1.3289998588070109e-05, "loss": 0.0411, "step": 17951 }, { "epoch": 3.29, "learning_rate": 1.3289297480079357e-05, "loss": 0.0405, "step": 17952 }, { "epoch": 3.29, "learning_rate": 1.328859635395792e-05, "loss": 0.0546, "step": 17953 }, { "epoch": 3.29, "learning_rate": 1.3287895209709662e-05, "loss": 0.0481, "step": 17954 }, { "epoch": 3.29, "learning_rate": 1.3287194047338448e-05, "loss": 0.0541, "step": 17955 }, { "epoch": 3.29, "learning_rate": 1.3286492866848143e-05, "loss": 0.0382, "step": 17956 }, { "epoch": 3.29, "learning_rate": 1.3285791668242616e-05, "loss": 0.0511, "step": 17957 }, { "epoch": 3.29, "learning_rate": 1.3285090451525726e-05, "loss": 0.0426, "step": 17958 }, { "epoch": 3.29, "learning_rate": 1.328438921670134e-05, "loss": 0.047, "step": 17959 }, { "epoch": 3.29, "learning_rate": 1.3283687963773326e-05, "loss": 0.0438, "step": 17960 }, { "epoch": 3.29, "learning_rate": 1.3282986692745545e-05, "loss": 0.0531, "step": 17961 }, { "epoch": 3.29, "learning_rate": 1.3282285403621864e-05, "loss": 0.0513, "step": 17962 }, { "epoch": 3.29, "learning_rate": 1.328158409640615e-05, "loss": 0.0484, "step": 17963 }, { "epoch": 3.29, "learning_rate": 1.3280882771102265e-05, "loss": 0.0494, "step": 17964 }, { "epoch": 3.29, "learning_rate": 1.3280181427714081e-05, "loss": 0.0439, "step": 17965 }, { "epoch": 3.29, "learning_rate": 1.327948006624546e-05, "loss": 0.0484, "step": 17966 }, { "epoch": 3.29, "learning_rate": 1.3278778686700266e-05, "loss": 0.0497, "step": 17967 }, { "epoch": 3.3, "learning_rate": 1.327807728908237e-05, "loss": 0.0462, "step": 17968 }, { "epoch": 3.3, "learning_rate": 1.3277375873395632e-05, "loss": 0.056, "step": 17969 }, { "epoch": 3.3, "learning_rate": 1.327667443964392e-05, "loss": 0.0466, "step": 17970 }, { "epoch": 3.3, "learning_rate": 1.3275972987831107e-05, "loss": 0.0518, "step": 17971 }, { "epoch": 3.3, "learning_rate": 1.3275271517961051e-05, "loss": 0.0483, "step": 17972 }, { "epoch": 3.3, "learning_rate": 1.327457003003762e-05, "loss": 0.0505, "step": 17973 }, { "epoch": 3.3, "learning_rate": 1.3273868524064684e-05, "loss": 0.0506, "step": 17974 }, { "epoch": 3.3, "learning_rate": 1.327316700004611e-05, "loss": 0.0509, "step": 17975 }, { "epoch": 3.3, "learning_rate": 1.3272465457985759e-05, "loss": 0.0413, "step": 17976 }, { "epoch": 3.3, "learning_rate": 1.3271763897887504e-05, "loss": 0.0465, "step": 17977 }, { "epoch": 3.3, "learning_rate": 1.327106231975521e-05, "loss": 0.042, "step": 17978 }, { "epoch": 3.3, "learning_rate": 1.3270360723592743e-05, "loss": 0.0513, "step": 17979 }, { "epoch": 3.3, "learning_rate": 1.3269659109403974e-05, "loss": 0.0446, "step": 17980 }, { "epoch": 3.3, "learning_rate": 1.3268957477192764e-05, "loss": 0.0491, "step": 17981 }, { "epoch": 3.3, "learning_rate": 1.3268255826962984e-05, "loss": 0.0553, "step": 17982 }, { "epoch": 3.3, "learning_rate": 1.3267554158718503e-05, "loss": 0.0433, "step": 17983 }, { "epoch": 3.3, "learning_rate": 1.3266852472463186e-05, "loss": 0.0475, "step": 17984 }, { "epoch": 3.3, "learning_rate": 1.32661507682009e-05, "loss": 0.0403, "step": 17985 }, { "epoch": 3.3, "learning_rate": 1.3265449045935517e-05, "loss": 0.0474, "step": 17986 }, { "epoch": 3.3, "learning_rate": 1.3264747305670901e-05, "loss": 0.0464, "step": 17987 }, { "epoch": 3.3, "learning_rate": 1.3264045547410919e-05, "loss": 0.0465, "step": 17988 }, { "epoch": 3.3, "learning_rate": 1.3263343771159447e-05, "loss": 0.0556, "step": 17989 }, { "epoch": 3.3, "learning_rate": 1.3262641976920343e-05, "loss": 0.0483, "step": 17990 }, { "epoch": 3.3, "learning_rate": 1.3261940164697483e-05, "loss": 0.0446, "step": 17991 }, { "epoch": 3.3, "learning_rate": 1.3261238334494732e-05, "loss": 0.0473, "step": 17992 }, { "epoch": 3.3, "learning_rate": 1.3260536486315955e-05, "loss": 0.0446, "step": 17993 }, { "epoch": 3.3, "learning_rate": 1.3259834620165028e-05, "loss": 0.043, "step": 17994 }, { "epoch": 3.3, "learning_rate": 1.3259132736045816e-05, "loss": 0.0623, "step": 17995 }, { "epoch": 3.3, "learning_rate": 1.325843083396219e-05, "loss": 0.0454, "step": 17996 }, { "epoch": 3.3, "learning_rate": 1.3257728913918013e-05, "loss": 0.049, "step": 17997 }, { "epoch": 3.3, "learning_rate": 1.3257026975917162e-05, "loss": 0.0549, "step": 17998 }, { "epoch": 3.3, "learning_rate": 1.3256325019963496e-05, "loss": 0.0466, "step": 17999 }, { "epoch": 3.3, "learning_rate": 1.3255623046060895e-05, "loss": 0.0591, "step": 18000 }, { "epoch": 3.3, "learning_rate": 1.3254921054213224e-05, "loss": 0.0434, "step": 18001 }, { "epoch": 3.3, "learning_rate": 1.325421904442435e-05, "loss": 0.0475, "step": 18002 }, { "epoch": 3.3, "learning_rate": 1.3253517016698146e-05, "loss": 0.0511, "step": 18003 }, { "epoch": 3.3, "learning_rate": 1.3252814971038477e-05, "loss": 0.0457, "step": 18004 }, { "epoch": 3.3, "learning_rate": 1.3252112907449216e-05, "loss": 0.058, "step": 18005 }, { "epoch": 3.3, "learning_rate": 1.3251410825934235e-05, "loss": 0.0546, "step": 18006 }, { "epoch": 3.3, "learning_rate": 1.3250708726497401e-05, "loss": 0.0417, "step": 18007 }, { "epoch": 3.3, "learning_rate": 1.3250006609142584e-05, "loss": 0.0503, "step": 18008 }, { "epoch": 3.3, "learning_rate": 1.3249304473873653e-05, "loss": 0.0491, "step": 18009 }, { "epoch": 3.3, "learning_rate": 1.324860232069448e-05, "loss": 0.0387, "step": 18010 }, { "epoch": 3.3, "learning_rate": 1.3247900149608936e-05, "loss": 0.0557, "step": 18011 }, { "epoch": 3.3, "learning_rate": 1.3247197960620888e-05, "loss": 0.0553, "step": 18012 }, { "epoch": 3.3, "learning_rate": 1.3246495753734212e-05, "loss": 0.0512, "step": 18013 }, { "epoch": 3.3, "learning_rate": 1.3245793528952775e-05, "loss": 0.0556, "step": 18014 }, { "epoch": 3.3, "learning_rate": 1.3245091286280443e-05, "loss": 0.0563, "step": 18015 }, { "epoch": 3.3, "learning_rate": 1.3244389025721099e-05, "loss": 0.0472, "step": 18016 }, { "epoch": 3.3, "learning_rate": 1.3243686747278602e-05, "loss": 0.0488, "step": 18017 }, { "epoch": 3.3, "learning_rate": 1.3242984450956829e-05, "loss": 0.0507, "step": 18018 }, { "epoch": 3.3, "learning_rate": 1.324228213675965e-05, "loss": 0.0507, "step": 18019 }, { "epoch": 3.3, "learning_rate": 1.3241579804690935e-05, "loss": 0.0555, "step": 18020 }, { "epoch": 3.3, "learning_rate": 1.3240877454754555e-05, "loss": 0.054, "step": 18021 }, { "epoch": 3.3, "learning_rate": 1.3240175086954387e-05, "loss": 0.0594, "step": 18022 }, { "epoch": 3.31, "learning_rate": 1.3239472701294294e-05, "loss": 0.047, "step": 18023 }, { "epoch": 3.31, "learning_rate": 1.3238770297778152e-05, "loss": 0.0551, "step": 18024 }, { "epoch": 3.31, "learning_rate": 1.3238067876409835e-05, "loss": 0.0466, "step": 18025 }, { "epoch": 3.31, "learning_rate": 1.323736543719321e-05, "loss": 0.0467, "step": 18026 }, { "epoch": 3.31, "learning_rate": 1.3236662980132151e-05, "loss": 0.0487, "step": 18027 }, { "epoch": 3.31, "learning_rate": 1.323596050523053e-05, "loss": 0.0427, "step": 18028 }, { "epoch": 3.31, "learning_rate": 1.3235258012492218e-05, "loss": 0.0463, "step": 18029 }, { "epoch": 3.31, "learning_rate": 1.323455550192109e-05, "loss": 0.0532, "step": 18030 }, { "epoch": 3.31, "learning_rate": 1.3233852973521015e-05, "loss": 0.059, "step": 18031 }, { "epoch": 3.31, "learning_rate": 1.3233150427295864e-05, "loss": 0.0539, "step": 18032 }, { "epoch": 3.31, "learning_rate": 1.3232447863249516e-05, "loss": 0.0444, "step": 18033 }, { "epoch": 3.31, "learning_rate": 1.323174528138584e-05, "loss": 0.0463, "step": 18034 }, { "epoch": 3.31, "learning_rate": 1.3231042681708705e-05, "loss": 0.0534, "step": 18035 }, { "epoch": 3.31, "learning_rate": 1.323034006422199e-05, "loss": 0.0535, "step": 18036 }, { "epoch": 3.31, "learning_rate": 1.3229637428929562e-05, "loss": 0.0523, "step": 18037 }, { "epoch": 3.31, "learning_rate": 1.32289347758353e-05, "loss": 0.0533, "step": 18038 }, { "epoch": 3.31, "learning_rate": 1.3228232104943073e-05, "loss": 0.0459, "step": 18039 }, { "epoch": 3.31, "learning_rate": 1.3227529416256755e-05, "loss": 0.0516, "step": 18040 }, { "epoch": 3.31, "learning_rate": 1.3226826709780216e-05, "loss": 0.0513, "step": 18041 }, { "epoch": 3.31, "learning_rate": 1.3226123985517337e-05, "loss": 0.0488, "step": 18042 }, { "epoch": 3.31, "learning_rate": 1.3225421243471983e-05, "loss": 0.0498, "step": 18043 }, { "epoch": 3.31, "learning_rate": 1.3224718483648035e-05, "loss": 0.0502, "step": 18044 }, { "epoch": 3.31, "learning_rate": 1.3224015706049362e-05, "loss": 0.0437, "step": 18045 }, { "epoch": 3.31, "learning_rate": 1.322331291067984e-05, "loss": 0.0536, "step": 18046 }, { "epoch": 3.31, "learning_rate": 1.322261009754334e-05, "loss": 0.0466, "step": 18047 }, { "epoch": 3.31, "learning_rate": 1.322190726664374e-05, "loss": 0.0528, "step": 18048 }, { "epoch": 3.31, "learning_rate": 1.3221204417984907e-05, "loss": 0.047, "step": 18049 }, { "epoch": 3.31, "learning_rate": 1.322050155157072e-05, "loss": 0.0402, "step": 18050 }, { "epoch": 3.31, "learning_rate": 1.321979866740506e-05, "loss": 0.0472, "step": 18051 }, { "epoch": 3.31, "learning_rate": 1.3219095765491789e-05, "loss": 0.0418, "step": 18052 }, { "epoch": 3.31, "learning_rate": 1.3218392845834789e-05, "loss": 0.0559, "step": 18053 }, { "epoch": 3.31, "learning_rate": 1.321768990843793e-05, "loss": 0.0512, "step": 18054 }, { "epoch": 3.31, "learning_rate": 1.3216986953305088e-05, "loss": 0.0518, "step": 18055 }, { "epoch": 3.31, "learning_rate": 1.3216283980440143e-05, "loss": 0.0521, "step": 18056 }, { "epoch": 3.31, "learning_rate": 1.321558098984696e-05, "loss": 0.0489, "step": 18057 }, { "epoch": 3.31, "learning_rate": 1.3214877981529424e-05, "loss": 0.0551, "step": 18058 }, { "epoch": 3.31, "learning_rate": 1.3214174955491401e-05, "loss": 0.0562, "step": 18059 }, { "epoch": 3.31, "learning_rate": 1.3213471911736775e-05, "loss": 0.0485, "step": 18060 }, { "epoch": 3.31, "learning_rate": 1.3212768850269413e-05, "loss": 0.0425, "step": 18061 }, { "epoch": 3.31, "learning_rate": 1.3212065771093194e-05, "loss": 0.0425, "step": 18062 }, { "epoch": 3.31, "learning_rate": 1.3211362674211996e-05, "loss": 0.0463, "step": 18063 }, { "epoch": 3.31, "learning_rate": 1.3210659559629688e-05, "loss": 0.0504, "step": 18064 }, { "epoch": 3.31, "learning_rate": 1.320995642735015e-05, "loss": 0.0507, "step": 18065 }, { "epoch": 3.31, "learning_rate": 1.3209253277377261e-05, "loss": 0.0458, "step": 18066 }, { "epoch": 3.31, "learning_rate": 1.3208550109714891e-05, "loss": 0.0438, "step": 18067 }, { "epoch": 3.31, "learning_rate": 1.3207846924366913e-05, "loss": 0.046, "step": 18068 }, { "epoch": 3.31, "learning_rate": 1.3207143721337214e-05, "loss": 0.0506, "step": 18069 }, { "epoch": 3.31, "learning_rate": 1.3206440500629657e-05, "loss": 0.0417, "step": 18070 }, { "epoch": 3.31, "learning_rate": 1.320573726224813e-05, "loss": 0.053, "step": 18071 }, { "epoch": 3.31, "learning_rate": 1.3205034006196505e-05, "loss": 0.0442, "step": 18072 }, { "epoch": 3.31, "learning_rate": 1.3204330732478653e-05, "loss": 0.0534, "step": 18073 }, { "epoch": 3.31, "learning_rate": 1.3203627441098457e-05, "loss": 0.0442, "step": 18074 }, { "epoch": 3.31, "learning_rate": 1.3202924132059793e-05, "loss": 0.0407, "step": 18075 }, { "epoch": 3.31, "learning_rate": 1.3202220805366533e-05, "loss": 0.0479, "step": 18076 }, { "epoch": 3.32, "learning_rate": 1.320151746102256e-05, "loss": 0.0477, "step": 18077 }, { "epoch": 3.32, "learning_rate": 1.3200814099031747e-05, "loss": 0.0589, "step": 18078 }, { "epoch": 3.32, "learning_rate": 1.3200110719397969e-05, "loss": 0.0522, "step": 18079 }, { "epoch": 3.32, "learning_rate": 1.3199407322125109e-05, "loss": 0.0458, "step": 18080 }, { "epoch": 3.32, "learning_rate": 1.3198703907217038e-05, "loss": 0.0438, "step": 18081 }, { "epoch": 3.32, "learning_rate": 1.319800047467764e-05, "loss": 0.0517, "step": 18082 }, { "epoch": 3.32, "learning_rate": 1.3197297024510784e-05, "loss": 0.0587, "step": 18083 }, { "epoch": 3.32, "learning_rate": 1.3196593556720356e-05, "loss": 0.0504, "step": 18084 }, { "epoch": 3.32, "learning_rate": 1.3195890071310226e-05, "loss": 0.0473, "step": 18085 }, { "epoch": 3.32, "learning_rate": 1.3195186568284278e-05, "loss": 0.0433, "step": 18086 }, { "epoch": 3.32, "learning_rate": 1.3194483047646385e-05, "loss": 0.0525, "step": 18087 }, { "epoch": 3.32, "learning_rate": 1.3193779509400425e-05, "loss": 0.0455, "step": 18088 }, { "epoch": 3.32, "learning_rate": 1.3193075953550282e-05, "loss": 0.0484, "step": 18089 }, { "epoch": 3.32, "learning_rate": 1.3192372380099827e-05, "loss": 0.0525, "step": 18090 }, { "epoch": 3.32, "learning_rate": 1.319166878905294e-05, "loss": 0.0511, "step": 18091 }, { "epoch": 3.32, "learning_rate": 1.31909651804135e-05, "loss": 0.0466, "step": 18092 }, { "epoch": 3.32, "learning_rate": 1.3190261554185388e-05, "loss": 0.045, "step": 18093 }, { "epoch": 3.32, "learning_rate": 1.3189557910372478e-05, "loss": 0.047, "step": 18094 }, { "epoch": 3.32, "learning_rate": 1.318885424897865e-05, "loss": 0.053, "step": 18095 }, { "epoch": 3.32, "learning_rate": 1.3188150570007781e-05, "loss": 0.0546, "step": 18096 }, { "epoch": 3.32, "learning_rate": 1.3187446873463752e-05, "loss": 0.0516, "step": 18097 }, { "epoch": 3.32, "learning_rate": 1.318674315935044e-05, "loss": 0.0482, "step": 18098 }, { "epoch": 3.32, "learning_rate": 1.3186039427671728e-05, "loss": 0.0427, "step": 18099 }, { "epoch": 3.32, "learning_rate": 1.318533567843149e-05, "loss": 0.0473, "step": 18100 }, { "epoch": 3.32, "learning_rate": 1.3184631911633608e-05, "loss": 0.0581, "step": 18101 }, { "epoch": 3.32, "learning_rate": 1.3183928127281961e-05, "loss": 0.0519, "step": 18102 }, { "epoch": 3.32, "learning_rate": 1.3183224325380426e-05, "loss": 0.0482, "step": 18103 }, { "epoch": 3.32, "learning_rate": 1.3182520505932885e-05, "loss": 0.0495, "step": 18104 }, { "epoch": 3.32, "learning_rate": 1.3181816668943217e-05, "loss": 0.0395, "step": 18105 }, { "epoch": 3.32, "learning_rate": 1.3181112814415299e-05, "loss": 0.0541, "step": 18106 }, { "epoch": 3.32, "learning_rate": 1.3180408942353014e-05, "loss": 0.0487, "step": 18107 }, { "epoch": 3.32, "learning_rate": 1.3179705052760241e-05, "loss": 0.0457, "step": 18108 }, { "epoch": 3.32, "learning_rate": 1.3179001145640856e-05, "loss": 0.0566, "step": 18109 }, { "epoch": 3.32, "learning_rate": 1.3178297220998744e-05, "loss": 0.0434, "step": 18110 }, { "epoch": 3.32, "learning_rate": 1.3177593278837786e-05, "loss": 0.0536, "step": 18111 }, { "epoch": 3.32, "learning_rate": 1.3176889319161856e-05, "loss": 0.0459, "step": 18112 }, { "epoch": 3.32, "learning_rate": 1.317618534197484e-05, "loss": 0.0492, "step": 18113 }, { "epoch": 3.32, "learning_rate": 1.3175481347280614e-05, "loss": 0.0516, "step": 18114 }, { "epoch": 3.32, "learning_rate": 1.3174777335083063e-05, "loss": 0.0546, "step": 18115 }, { "epoch": 3.32, "learning_rate": 1.3174073305386061e-05, "loss": 0.051, "step": 18116 }, { "epoch": 3.32, "learning_rate": 1.3173369258193496e-05, "loss": 0.0448, "step": 18117 }, { "epoch": 3.32, "learning_rate": 1.3172665193509242e-05, "loss": 0.0503, "step": 18118 }, { "epoch": 3.32, "learning_rate": 1.3171961111337186e-05, "loss": 0.0464, "step": 18119 }, { "epoch": 3.32, "learning_rate": 1.3171257011681209e-05, "loss": 0.0511, "step": 18120 }, { "epoch": 3.32, "learning_rate": 1.3170552894545181e-05, "loss": 0.0501, "step": 18121 }, { "epoch": 3.32, "learning_rate": 1.3169848759932997e-05, "loss": 0.0483, "step": 18122 }, { "epoch": 3.32, "learning_rate": 1.3169144607848532e-05, "loss": 0.041, "step": 18123 }, { "epoch": 3.32, "learning_rate": 1.3168440438295664e-05, "loss": 0.0443, "step": 18124 }, { "epoch": 3.32, "learning_rate": 1.3167736251278283e-05, "loss": 0.0475, "step": 18125 }, { "epoch": 3.32, "learning_rate": 1.3167032046800263e-05, "loss": 0.05, "step": 18126 }, { "epoch": 3.32, "learning_rate": 1.3166327824865485e-05, "loss": 0.0459, "step": 18127 }, { "epoch": 3.32, "learning_rate": 1.3165623585477838e-05, "loss": 0.0489, "step": 18128 }, { "epoch": 3.32, "learning_rate": 1.31649193286412e-05, "loss": 0.0541, "step": 18129 }, { "epoch": 3.32, "learning_rate": 1.3164215054359452e-05, "loss": 0.0509, "step": 18130 }, { "epoch": 3.32, "learning_rate": 1.3163510762636476e-05, "loss": 0.0537, "step": 18131 }, { "epoch": 3.33, "learning_rate": 1.3162806453476153e-05, "loss": 0.0463, "step": 18132 }, { "epoch": 3.33, "learning_rate": 1.316210212688237e-05, "loss": 0.0454, "step": 18133 }, { "epoch": 3.33, "learning_rate": 1.3161397782859005e-05, "loss": 0.0463, "step": 18134 }, { "epoch": 3.33, "learning_rate": 1.316069342140994e-05, "loss": 0.0417, "step": 18135 }, { "epoch": 3.33, "learning_rate": 1.3159989042539056e-05, "loss": 0.0455, "step": 18136 }, { "epoch": 3.33, "learning_rate": 1.3159284646250246e-05, "loss": 0.0467, "step": 18137 }, { "epoch": 3.33, "learning_rate": 1.3158580232547378e-05, "loss": 0.0528, "step": 18138 }, { "epoch": 3.33, "learning_rate": 1.3157875801434346e-05, "loss": 0.0474, "step": 18139 }, { "epoch": 3.33, "learning_rate": 1.3157171352915026e-05, "loss": 0.0472, "step": 18140 }, { "epoch": 3.33, "learning_rate": 1.3156466886993304e-05, "loss": 0.0524, "step": 18141 }, { "epoch": 3.33, "learning_rate": 1.3155762403673065e-05, "loss": 0.0556, "step": 18142 }, { "epoch": 3.33, "learning_rate": 1.3155057902958185e-05, "loss": 0.0546, "step": 18143 }, { "epoch": 3.33, "learning_rate": 1.3154353384852559e-05, "loss": 0.0401, "step": 18144 }, { "epoch": 3.33, "learning_rate": 1.3153648849360057e-05, "loss": 0.038, "step": 18145 }, { "epoch": 3.33, "learning_rate": 1.315294429648457e-05, "loss": 0.0548, "step": 18146 }, { "epoch": 3.33, "learning_rate": 1.315223972622998e-05, "loss": 0.0504, "step": 18147 }, { "epoch": 3.33, "learning_rate": 1.3151535138600174e-05, "loss": 0.0471, "step": 18148 }, { "epoch": 3.33, "learning_rate": 1.3150830533599029e-05, "loss": 0.0466, "step": 18149 }, { "epoch": 3.33, "learning_rate": 1.3150125911230433e-05, "loss": 0.0455, "step": 18150 }, { "epoch": 3.33, "learning_rate": 1.3149421271498268e-05, "loss": 0.0529, "step": 18151 }, { "epoch": 3.33, "learning_rate": 1.3148716614406423e-05, "loss": 0.063, "step": 18152 }, { "epoch": 3.33, "learning_rate": 1.3148011939958776e-05, "loss": 0.0453, "step": 18153 }, { "epoch": 3.33, "learning_rate": 1.3147307248159211e-05, "loss": 0.0475, "step": 18154 }, { "epoch": 3.33, "learning_rate": 1.314660253901162e-05, "loss": 0.0495, "step": 18155 }, { "epoch": 3.33, "learning_rate": 1.3145897812519879e-05, "loss": 0.0559, "step": 18156 }, { "epoch": 3.33, "learning_rate": 1.3145193068687875e-05, "loss": 0.0456, "step": 18157 }, { "epoch": 3.33, "learning_rate": 1.3144488307519494e-05, "loss": 0.0541, "step": 18158 }, { "epoch": 3.33, "learning_rate": 1.3143783529018621e-05, "loss": 0.0447, "step": 18159 }, { "epoch": 3.33, "learning_rate": 1.3143078733189137e-05, "loss": 0.0476, "step": 18160 }, { "epoch": 3.33, "learning_rate": 1.3142373920034932e-05, "loss": 0.0463, "step": 18161 }, { "epoch": 3.33, "learning_rate": 1.3141669089559885e-05, "loss": 0.0438, "step": 18162 }, { "epoch": 3.33, "learning_rate": 1.3140964241767886e-05, "loss": 0.0503, "step": 18163 }, { "epoch": 3.33, "learning_rate": 1.314025937666282e-05, "loss": 0.049, "step": 18164 }, { "epoch": 3.33, "learning_rate": 1.313955449424857e-05, "loss": 0.0487, "step": 18165 }, { "epoch": 3.33, "learning_rate": 1.313884959452902e-05, "loss": 0.0453, "step": 18166 }, { "epoch": 3.33, "learning_rate": 1.313814467750806e-05, "loss": 0.0524, "step": 18167 }, { "epoch": 3.33, "learning_rate": 1.3137439743189573e-05, "loss": 0.0527, "step": 18168 }, { "epoch": 3.33, "learning_rate": 1.3136734791577441e-05, "loss": 0.0522, "step": 18169 }, { "epoch": 3.33, "learning_rate": 1.3136029822675558e-05, "loss": 0.047, "step": 18170 }, { "epoch": 3.33, "learning_rate": 1.3135324836487802e-05, "loss": 0.0501, "step": 18171 }, { "epoch": 3.33, "learning_rate": 1.3134619833018065e-05, "loss": 0.0446, "step": 18172 }, { "epoch": 3.33, "learning_rate": 1.3133914812270227e-05, "loss": 0.0527, "step": 18173 }, { "epoch": 3.33, "learning_rate": 1.313320977424818e-05, "loss": 0.0544, "step": 18174 }, { "epoch": 3.33, "learning_rate": 1.3132504718955805e-05, "loss": 0.0504, "step": 18175 }, { "epoch": 3.33, "learning_rate": 1.3131799646396992e-05, "loss": 0.0436, "step": 18176 }, { "epoch": 3.33, "learning_rate": 1.3131094556575625e-05, "loss": 0.043, "step": 18177 }, { "epoch": 3.33, "learning_rate": 1.3130389449495591e-05, "loss": 0.0533, "step": 18178 }, { "epoch": 3.33, "learning_rate": 1.312968432516078e-05, "loss": 0.0455, "step": 18179 }, { "epoch": 3.33, "learning_rate": 1.3128979183575072e-05, "loss": 0.0411, "step": 18180 }, { "epoch": 3.33, "learning_rate": 1.3128274024742357e-05, "loss": 0.0471, "step": 18181 }, { "epoch": 3.33, "learning_rate": 1.3127568848666527e-05, "loss": 0.0515, "step": 18182 }, { "epoch": 3.33, "learning_rate": 1.312686365535146e-05, "loss": 0.0513, "step": 18183 }, { "epoch": 3.33, "learning_rate": 1.3126158444801051e-05, "loss": 0.0474, "step": 18184 }, { "epoch": 3.33, "learning_rate": 1.3125453217019182e-05, "loss": 0.0573, "step": 18185 }, { "epoch": 3.34, "learning_rate": 1.3124747972009742e-05, "loss": 0.0576, "step": 18186 }, { "epoch": 3.34, "learning_rate": 1.3124042709776611e-05, "loss": 0.0493, "step": 18187 }, { "epoch": 3.34, "learning_rate": 1.3123337430323693e-05, "loss": 0.038, "step": 18188 }, { "epoch": 3.34, "learning_rate": 1.312263213365486e-05, "loss": 0.0615, "step": 18189 }, { "epoch": 3.34, "learning_rate": 1.3121926819774009e-05, "loss": 0.0486, "step": 18190 }, { "epoch": 3.34, "learning_rate": 1.3121221488685023e-05, "loss": 0.0487, "step": 18191 }, { "epoch": 3.34, "learning_rate": 1.312051614039179e-05, "loss": 0.0513, "step": 18192 }, { "epoch": 3.34, "learning_rate": 1.31198107748982e-05, "loss": 0.0521, "step": 18193 }, { "epoch": 3.34, "learning_rate": 1.311910539220814e-05, "loss": 0.0537, "step": 18194 }, { "epoch": 3.34, "learning_rate": 1.3118399992325498e-05, "loss": 0.0492, "step": 18195 }, { "epoch": 3.34, "learning_rate": 1.3117694575254163e-05, "loss": 0.0539, "step": 18196 }, { "epoch": 3.34, "learning_rate": 1.3116989140998022e-05, "loss": 0.0461, "step": 18197 }, { "epoch": 3.34, "learning_rate": 1.3116283689560963e-05, "loss": 0.046, "step": 18198 }, { "epoch": 3.34, "learning_rate": 1.3115578220946874e-05, "loss": 0.0462, "step": 18199 }, { "epoch": 3.34, "learning_rate": 1.3114872735159647e-05, "loss": 0.0534, "step": 18200 }, { "epoch": 3.34, "learning_rate": 1.311416723220317e-05, "loss": 0.047, "step": 18201 }, { "epoch": 3.34, "learning_rate": 1.3113461712081326e-05, "loss": 0.0551, "step": 18202 }, { "epoch": 3.34, "learning_rate": 1.311275617479801e-05, "loss": 0.0546, "step": 18203 }, { "epoch": 3.34, "learning_rate": 1.311205062035711e-05, "loss": 0.0514, "step": 18204 }, { "epoch": 3.34, "learning_rate": 1.3111345048762512e-05, "loss": 0.0404, "step": 18205 }, { "epoch": 3.34, "learning_rate": 1.3110639460018111e-05, "loss": 0.0486, "step": 18206 }, { "epoch": 3.34, "learning_rate": 1.310993385412779e-05, "loss": 0.0451, "step": 18207 }, { "epoch": 3.34, "learning_rate": 1.310922823109544e-05, "loss": 0.0494, "step": 18208 }, { "epoch": 3.34, "learning_rate": 1.3108522590924953e-05, "loss": 0.0472, "step": 18209 }, { "epoch": 3.34, "learning_rate": 1.3107816933620215e-05, "loss": 0.0499, "step": 18210 }, { "epoch": 3.34, "learning_rate": 1.3107111259185116e-05, "loss": 0.0522, "step": 18211 }, { "epoch": 3.34, "learning_rate": 1.310640556762355e-05, "loss": 0.0339, "step": 18212 }, { "epoch": 3.34, "learning_rate": 1.3105699858939401e-05, "loss": 0.0403, "step": 18213 }, { "epoch": 3.34, "learning_rate": 1.3104994133136563e-05, "loss": 0.0565, "step": 18214 }, { "epoch": 3.34, "learning_rate": 1.3104288390218927e-05, "loss": 0.0569, "step": 18215 }, { "epoch": 3.34, "learning_rate": 1.3103582630190377e-05, "loss": 0.0542, "step": 18216 }, { "epoch": 3.34, "learning_rate": 1.3102876853054808e-05, "loss": 0.0512, "step": 18217 }, { "epoch": 3.34, "learning_rate": 1.3102171058816112e-05, "loss": 0.0502, "step": 18218 }, { "epoch": 3.34, "learning_rate": 1.3101465247478173e-05, "loss": 0.0533, "step": 18219 }, { "epoch": 3.34, "learning_rate": 1.3100759419044886e-05, "loss": 0.0391, "step": 18220 }, { "epoch": 3.34, "learning_rate": 1.310005357352014e-05, "loss": 0.0424, "step": 18221 }, { "epoch": 3.34, "learning_rate": 1.3099347710907824e-05, "loss": 0.045, "step": 18222 }, { "epoch": 3.34, "learning_rate": 1.3098641831211836e-05, "loss": 0.0504, "step": 18223 }, { "epoch": 3.34, "learning_rate": 1.3097935934436057e-05, "loss": 0.0449, "step": 18224 }, { "epoch": 3.34, "learning_rate": 1.3097230020584384e-05, "loss": 0.0551, "step": 18225 }, { "epoch": 3.34, "learning_rate": 1.3096524089660707e-05, "loss": 0.0473, "step": 18226 }, { "epoch": 3.34, "learning_rate": 1.3095818141668917e-05, "loss": 0.042, "step": 18227 }, { "epoch": 3.34, "learning_rate": 1.3095112176612906e-05, "loss": 0.0508, "step": 18228 }, { "epoch": 3.34, "learning_rate": 1.309440619449656e-05, "loss": 0.0466, "step": 18229 }, { "epoch": 3.34, "learning_rate": 1.3093700195323778e-05, "loss": 0.0482, "step": 18230 }, { "epoch": 3.34, "learning_rate": 1.3092994179098444e-05, "loss": 0.0451, "step": 18231 }, { "epoch": 3.34, "learning_rate": 1.3092288145824461e-05, "loss": 0.0513, "step": 18232 }, { "epoch": 3.34, "learning_rate": 1.3091582095505707e-05, "loss": 0.0466, "step": 18233 }, { "epoch": 3.34, "learning_rate": 1.3090876028146083e-05, "loss": 0.0528, "step": 18234 }, { "epoch": 3.34, "learning_rate": 1.3090169943749475e-05, "loss": 0.0413, "step": 18235 }, { "epoch": 3.34, "learning_rate": 1.3089463842319781e-05, "loss": 0.0472, "step": 18236 }, { "epoch": 3.34, "learning_rate": 1.3088757723860885e-05, "loss": 0.0556, "step": 18237 }, { "epoch": 3.34, "learning_rate": 1.3088051588376692e-05, "loss": 0.0525, "step": 18238 }, { "epoch": 3.34, "learning_rate": 1.3087345435871082e-05, "loss": 0.0455, "step": 18239 }, { "epoch": 3.34, "learning_rate": 1.308663926634795e-05, "loss": 0.049, "step": 18240 }, { "epoch": 3.35, "learning_rate": 1.3085933079811195e-05, "loss": 0.0492, "step": 18241 }, { "epoch": 3.35, "learning_rate": 1.3085226876264699e-05, "loss": 0.0415, "step": 18242 }, { "epoch": 3.35, "learning_rate": 1.3084520655712363e-05, "loss": 0.0433, "step": 18243 }, { "epoch": 3.35, "learning_rate": 1.3083814418158075e-05, "loss": 0.0437, "step": 18244 }, { "epoch": 3.35, "learning_rate": 1.3083108163605734e-05, "loss": 0.0489, "step": 18245 }, { "epoch": 3.35, "learning_rate": 1.3082401892059223e-05, "loss": 0.0492, "step": 18246 }, { "epoch": 3.35, "learning_rate": 1.3081695603522446e-05, "loss": 0.0483, "step": 18247 }, { "epoch": 3.35, "learning_rate": 1.3080989297999287e-05, "loss": 0.0573, "step": 18248 }, { "epoch": 3.35, "learning_rate": 1.308028297549364e-05, "loss": 0.0512, "step": 18249 }, { "epoch": 3.35, "learning_rate": 1.307957663600941e-05, "loss": 0.0547, "step": 18250 }, { "epoch": 3.35, "learning_rate": 1.3078870279550474e-05, "loss": 0.0496, "step": 18251 }, { "epoch": 3.35, "learning_rate": 1.3078163906120736e-05, "loss": 0.0415, "step": 18252 }, { "epoch": 3.35, "learning_rate": 1.3077457515724086e-05, "loss": 0.0496, "step": 18253 }, { "epoch": 3.35, "learning_rate": 1.3076751108364416e-05, "loss": 0.0442, "step": 18254 }, { "epoch": 3.35, "learning_rate": 1.3076044684045623e-05, "loss": 0.0533, "step": 18255 }, { "epoch": 3.35, "learning_rate": 1.3075338242771602e-05, "loss": 0.0525, "step": 18256 }, { "epoch": 3.35, "learning_rate": 1.307463178454624e-05, "loss": 0.057, "step": 18257 }, { "epoch": 3.35, "learning_rate": 1.3073925309373438e-05, "loss": 0.0549, "step": 18258 }, { "epoch": 3.35, "learning_rate": 1.3073218817257091e-05, "loss": 0.0554, "step": 18259 }, { "epoch": 3.35, "learning_rate": 1.3072512308201085e-05, "loss": 0.0544, "step": 18260 }, { "epoch": 3.35, "learning_rate": 1.3071805782209322e-05, "loss": 0.0452, "step": 18261 }, { "epoch": 3.35, "learning_rate": 1.3071099239285691e-05, "loss": 0.0434, "step": 18262 }, { "epoch": 3.35, "learning_rate": 1.3070392679434091e-05, "loss": 0.0553, "step": 18263 }, { "epoch": 3.35, "learning_rate": 1.3069686102658414e-05, "loss": 0.0519, "step": 18264 }, { "epoch": 3.35, "learning_rate": 1.3068979508962556e-05, "loss": 0.043, "step": 18265 }, { "epoch": 3.35, "learning_rate": 1.306827289835041e-05, "loss": 0.0457, "step": 18266 }, { "epoch": 3.35, "learning_rate": 1.3067566270825873e-05, "loss": 0.0475, "step": 18267 }, { "epoch": 3.35, "learning_rate": 1.3066859626392838e-05, "loss": 0.0615, "step": 18268 }, { "epoch": 3.35, "learning_rate": 1.3066152965055202e-05, "loss": 0.0401, "step": 18269 }, { "epoch": 3.35, "learning_rate": 1.3065446286816856e-05, "loss": 0.0499, "step": 18270 }, { "epoch": 3.35, "learning_rate": 1.3064739591681703e-05, "loss": 0.0478, "step": 18271 }, { "epoch": 3.35, "learning_rate": 1.306403287965363e-05, "loss": 0.0452, "step": 18272 }, { "epoch": 3.35, "learning_rate": 1.3063326150736535e-05, "loss": 0.0481, "step": 18273 }, { "epoch": 3.35, "learning_rate": 1.3062619404934319e-05, "loss": 0.0538, "step": 18274 }, { "epoch": 3.35, "learning_rate": 1.306191264225087e-05, "loss": 0.061, "step": 18275 }, { "epoch": 3.35, "learning_rate": 1.3061205862690085e-05, "loss": 0.0494, "step": 18276 }, { "epoch": 3.35, "learning_rate": 1.3060499066255864e-05, "loss": 0.0539, "step": 18277 }, { "epoch": 3.35, "learning_rate": 1.3059792252952102e-05, "loss": 0.0453, "step": 18278 }, { "epoch": 3.35, "learning_rate": 1.3059085422782691e-05, "loss": 0.0508, "step": 18279 }, { "epoch": 3.35, "learning_rate": 1.3058378575751532e-05, "loss": 0.0455, "step": 18280 }, { "epoch": 3.35, "learning_rate": 1.3057671711862516e-05, "loss": 0.0531, "step": 18281 }, { "epoch": 3.35, "learning_rate": 1.3056964831119543e-05, "loss": 0.041, "step": 18282 }, { "epoch": 3.35, "learning_rate": 1.3056257933526508e-05, "loss": 0.0604, "step": 18283 }, { "epoch": 3.35, "learning_rate": 1.3055551019087307e-05, "loss": 0.06, "step": 18284 }, { "epoch": 3.35, "learning_rate": 1.3054844087805839e-05, "loss": 0.0556, "step": 18285 }, { "epoch": 3.35, "learning_rate": 1.3054137139685996e-05, "loss": 0.0499, "step": 18286 }, { "epoch": 3.35, "learning_rate": 1.305343017473168e-05, "loss": 0.0522, "step": 18287 }, { "epoch": 3.35, "learning_rate": 1.3052723192946786e-05, "loss": 0.0533, "step": 18288 }, { "epoch": 3.35, "learning_rate": 1.3052016194335208e-05, "loss": 0.0427, "step": 18289 }, { "epoch": 3.35, "learning_rate": 1.3051309178900844e-05, "loss": 0.0422, "step": 18290 }, { "epoch": 3.35, "learning_rate": 1.3050602146647596e-05, "loss": 0.0463, "step": 18291 }, { "epoch": 3.35, "learning_rate": 1.3049895097579357e-05, "loss": 0.049, "step": 18292 }, { "epoch": 3.35, "learning_rate": 1.3049188031700023e-05, "loss": 0.0434, "step": 18293 }, { "epoch": 3.35, "learning_rate": 1.3048480949013494e-05, "loss": 0.0416, "step": 18294 }, { "epoch": 3.36, "learning_rate": 1.3047773849523666e-05, "loss": 0.0487, "step": 18295 }, { "epoch": 3.36, "learning_rate": 1.3047066733234436e-05, "loss": 0.0428, "step": 18296 }, { "epoch": 3.36, "learning_rate": 1.3046359600149705e-05, "loss": 0.0553, "step": 18297 }, { "epoch": 3.36, "learning_rate": 1.304565245027337e-05, "loss": 0.0481, "step": 18298 }, { "epoch": 3.36, "learning_rate": 1.3044945283609322e-05, "loss": 0.0554, "step": 18299 }, { "epoch": 3.36, "learning_rate": 1.304423810016147e-05, "loss": 0.0441, "step": 18300 }, { "epoch": 3.36, "learning_rate": 1.30435308999337e-05, "loss": 0.0401, "step": 18301 }, { "epoch": 3.36, "learning_rate": 1.3042823682929922e-05, "loss": 0.0549, "step": 18302 }, { "epoch": 3.36, "learning_rate": 1.3042116449154027e-05, "loss": 0.0414, "step": 18303 }, { "epoch": 3.36, "learning_rate": 1.3041409198609915e-05, "loss": 0.0449, "step": 18304 }, { "epoch": 3.36, "learning_rate": 1.3040701931301483e-05, "loss": 0.0559, "step": 18305 }, { "epoch": 3.36, "learning_rate": 1.3039994647232633e-05, "loss": 0.0417, "step": 18306 }, { "epoch": 3.36, "learning_rate": 1.303928734640726e-05, "loss": 0.0477, "step": 18307 }, { "epoch": 3.36, "learning_rate": 1.3038580028829262e-05, "loss": 0.0525, "step": 18308 }, { "epoch": 3.36, "learning_rate": 1.3037872694502542e-05, "loss": 0.0458, "step": 18309 }, { "epoch": 3.36, "learning_rate": 1.3037165343430995e-05, "loss": 0.0454, "step": 18310 }, { "epoch": 3.36, "learning_rate": 1.3036457975618522e-05, "loss": 0.0452, "step": 18311 }, { "epoch": 3.36, "learning_rate": 1.3035750591069023e-05, "loss": 0.0485, "step": 18312 }, { "epoch": 3.36, "learning_rate": 1.3035043189786394e-05, "loss": 0.0473, "step": 18313 }, { "epoch": 3.36, "learning_rate": 1.3034335771774538e-05, "loss": 0.0498, "step": 18314 }, { "epoch": 3.36, "learning_rate": 1.3033628337037349e-05, "loss": 0.0433, "step": 18315 }, { "epoch": 3.36, "learning_rate": 1.3032920885578733e-05, "loss": 0.0475, "step": 18316 }, { "epoch": 3.36, "learning_rate": 1.3032213417402582e-05, "loss": 0.0472, "step": 18317 }, { "epoch": 3.36, "learning_rate": 1.3031505932512803e-05, "loss": 0.0539, "step": 18318 }, { "epoch": 3.36, "learning_rate": 1.3030798430913289e-05, "loss": 0.0418, "step": 18319 }, { "epoch": 3.36, "learning_rate": 1.3030090912607946e-05, "loss": 0.04, "step": 18320 }, { "epoch": 3.36, "learning_rate": 1.302938337760067e-05, "loss": 0.0461, "step": 18321 }, { "epoch": 3.36, "learning_rate": 1.3028675825895363e-05, "loss": 0.0538, "step": 18322 }, { "epoch": 3.36, "learning_rate": 1.3027968257495923e-05, "loss": 0.0492, "step": 18323 }, { "epoch": 3.36, "learning_rate": 1.302726067240625e-05, "loss": 0.0437, "step": 18324 }, { "epoch": 3.36, "learning_rate": 1.3026553070630245e-05, "loss": 0.0478, "step": 18325 }, { "epoch": 3.36, "learning_rate": 1.3025845452171808e-05, "loss": 0.0541, "step": 18326 }, { "epoch": 3.36, "learning_rate": 1.3025137817034845e-05, "loss": 0.0605, "step": 18327 }, { "epoch": 3.36, "learning_rate": 1.3024430165223245e-05, "loss": 0.0483, "step": 18328 }, { "epoch": 3.36, "learning_rate": 1.3023722496740919e-05, "loss": 0.0426, "step": 18329 }, { "epoch": 3.36, "learning_rate": 1.3023014811591763e-05, "loss": 0.0466, "step": 18330 }, { "epoch": 3.36, "learning_rate": 1.302230710977968e-05, "loss": 0.0489, "step": 18331 }, { "epoch": 3.36, "learning_rate": 1.3021599391308564e-05, "loss": 0.0491, "step": 18332 }, { "epoch": 3.36, "learning_rate": 1.3020891656182328e-05, "loss": 0.0513, "step": 18333 }, { "epoch": 3.36, "learning_rate": 1.3020183904404863e-05, "loss": 0.0518, "step": 18334 }, { "epoch": 3.36, "learning_rate": 1.3019476135980073e-05, "loss": 0.0481, "step": 18335 }, { "epoch": 3.36, "learning_rate": 1.3018768350911863e-05, "loss": 0.052, "step": 18336 }, { "epoch": 3.36, "learning_rate": 1.3018060549204127e-05, "loss": 0.0515, "step": 18337 }, { "epoch": 3.36, "learning_rate": 1.3017352730860773e-05, "loss": 0.0502, "step": 18338 }, { "epoch": 3.36, "learning_rate": 1.30166448958857e-05, "loss": 0.0501, "step": 18339 }, { "epoch": 3.36, "learning_rate": 1.301593704428281e-05, "loss": 0.0457, "step": 18340 }, { "epoch": 3.36, "learning_rate": 1.3015229176056001e-05, "loss": 0.0432, "step": 18341 }, { "epoch": 3.36, "learning_rate": 1.3014521291209183e-05, "loss": 0.0383, "step": 18342 }, { "epoch": 3.36, "learning_rate": 1.3013813389746249e-05, "loss": 0.0481, "step": 18343 }, { "epoch": 3.36, "learning_rate": 1.3013105471671108e-05, "loss": 0.0484, "step": 18344 }, { "epoch": 3.36, "learning_rate": 1.301239753698766e-05, "loss": 0.0504, "step": 18345 }, { "epoch": 3.36, "learning_rate": 1.3011689585699801e-05, "loss": 0.0491, "step": 18346 }, { "epoch": 3.36, "learning_rate": 1.3010981617811442e-05, "loss": 0.0604, "step": 18347 }, { "epoch": 3.36, "learning_rate": 1.301027363332648e-05, "loss": 0.0522, "step": 18348 }, { "epoch": 3.36, "learning_rate": 1.3009565632248822e-05, "loss": 0.0486, "step": 18349 }, { "epoch": 3.37, "learning_rate": 1.3008857614582364e-05, "loss": 0.0471, "step": 18350 }, { "epoch": 3.37, "learning_rate": 1.3008149580331018e-05, "loss": 0.0485, "step": 18351 }, { "epoch": 3.37, "learning_rate": 1.3007441529498675e-05, "loss": 0.0553, "step": 18352 }, { "epoch": 3.37, "learning_rate": 1.3006733462089247e-05, "loss": 0.0442, "step": 18353 }, { "epoch": 3.37, "learning_rate": 1.3006025378106633e-05, "loss": 0.0465, "step": 18354 }, { "epoch": 3.37, "learning_rate": 1.3005317277554737e-05, "loss": 0.061, "step": 18355 }, { "epoch": 3.37, "learning_rate": 1.3004609160437463e-05, "loss": 0.0497, "step": 18356 }, { "epoch": 3.37, "learning_rate": 1.3003901026758711e-05, "loss": 0.0505, "step": 18357 }, { "epoch": 3.37, "learning_rate": 1.3003192876522385e-05, "loss": 0.0481, "step": 18358 }, { "epoch": 3.37, "learning_rate": 1.300248470973239e-05, "loss": 0.0453, "step": 18359 }, { "epoch": 3.37, "learning_rate": 1.3001776526392634e-05, "loss": 0.0402, "step": 18360 }, { "epoch": 3.37, "learning_rate": 1.3001068326507009e-05, "loss": 0.0532, "step": 18361 }, { "epoch": 3.37, "learning_rate": 1.3000360110079428e-05, "loss": 0.0507, "step": 18362 }, { "epoch": 3.37, "learning_rate": 1.2999651877113792e-05, "loss": 0.0403, "step": 18363 }, { "epoch": 3.37, "learning_rate": 1.2998943627614005e-05, "loss": 0.0471, "step": 18364 }, { "epoch": 3.37, "learning_rate": 1.2998235361583968e-05, "loss": 0.0463, "step": 18365 }, { "epoch": 3.37, "learning_rate": 1.299752707902759e-05, "loss": 0.0527, "step": 18366 }, { "epoch": 3.37, "learning_rate": 1.2996818779948773e-05, "loss": 0.0407, "step": 18367 }, { "epoch": 3.37, "learning_rate": 1.2996110464351417e-05, "loss": 0.0505, "step": 18368 }, { "epoch": 3.37, "learning_rate": 1.2995402132239434e-05, "loss": 0.0446, "step": 18369 }, { "epoch": 3.37, "learning_rate": 1.299469378361672e-05, "loss": 0.0436, "step": 18370 }, { "epoch": 3.37, "learning_rate": 1.2993985418487185e-05, "loss": 0.0521, "step": 18371 }, { "epoch": 3.37, "learning_rate": 1.2993277036854734e-05, "loss": 0.0501, "step": 18372 }, { "epoch": 3.37, "learning_rate": 1.2992568638723269e-05, "loss": 0.0463, "step": 18373 }, { "epoch": 3.37, "learning_rate": 1.2991860224096695e-05, "loss": 0.0494, "step": 18374 }, { "epoch": 3.37, "learning_rate": 1.299115179297892e-05, "loss": 0.0522, "step": 18375 }, { "epoch": 3.37, "learning_rate": 1.2990443345373845e-05, "loss": 0.0426, "step": 18376 }, { "epoch": 3.37, "learning_rate": 1.2989734881285374e-05, "loss": 0.0453, "step": 18377 }, { "epoch": 3.37, "learning_rate": 1.2989026400717418e-05, "loss": 0.0539, "step": 18378 }, { "epoch": 3.37, "learning_rate": 1.2988317903673873e-05, "loss": 0.0562, "step": 18379 }, { "epoch": 3.37, "learning_rate": 1.2987609390158653e-05, "loss": 0.0466, "step": 18380 }, { "epoch": 3.37, "learning_rate": 1.2986900860175662e-05, "loss": 0.0568, "step": 18381 }, { "epoch": 3.37, "learning_rate": 1.29861923137288e-05, "loss": 0.0495, "step": 18382 }, { "epoch": 3.37, "learning_rate": 1.2985483750821976e-05, "loss": 0.0511, "step": 18383 }, { "epoch": 3.37, "learning_rate": 1.2984775171459099e-05, "loss": 0.0503, "step": 18384 }, { "epoch": 3.37, "learning_rate": 1.2984066575644067e-05, "loss": 0.0466, "step": 18385 }, { "epoch": 3.37, "learning_rate": 1.2983357963380796e-05, "loss": 0.0542, "step": 18386 }, { "epoch": 3.37, "learning_rate": 1.2982649334673182e-05, "loss": 0.0482, "step": 18387 }, { "epoch": 3.37, "learning_rate": 1.2981940689525134e-05, "loss": 0.0558, "step": 18388 }, { "epoch": 3.37, "learning_rate": 1.2981232027940562e-05, "loss": 0.0457, "step": 18389 }, { "epoch": 3.37, "learning_rate": 1.2980523349923368e-05, "loss": 0.054, "step": 18390 }, { "epoch": 3.37, "learning_rate": 1.2979814655477458e-05, "loss": 0.0418, "step": 18391 }, { "epoch": 3.37, "learning_rate": 1.2979105944606743e-05, "loss": 0.0467, "step": 18392 }, { "epoch": 3.37, "learning_rate": 1.2978397217315125e-05, "loss": 0.0483, "step": 18393 }, { "epoch": 3.37, "learning_rate": 1.2977688473606507e-05, "loss": 0.0578, "step": 18394 }, { "epoch": 3.37, "learning_rate": 1.2976979713484807e-05, "loss": 0.0453, "step": 18395 }, { "epoch": 3.37, "learning_rate": 1.297627093695392e-05, "loss": 0.051, "step": 18396 }, { "epoch": 3.37, "learning_rate": 1.2975562144017762e-05, "loss": 0.0515, "step": 18397 }, { "epoch": 3.37, "learning_rate": 1.2974853334680236e-05, "loss": 0.0516, "step": 18398 }, { "epoch": 3.37, "learning_rate": 1.2974144508945246e-05, "loss": 0.051, "step": 18399 }, { "epoch": 3.37, "learning_rate": 1.2973435666816702e-05, "loss": 0.0468, "step": 18400 }, { "epoch": 3.37, "learning_rate": 1.2972726808298513e-05, "loss": 0.0503, "step": 18401 }, { "epoch": 3.37, "learning_rate": 1.2972017933394582e-05, "loss": 0.0558, "step": 18402 }, { "epoch": 3.37, "learning_rate": 1.2971309042108817e-05, "loss": 0.0516, "step": 18403 }, { "epoch": 3.38, "learning_rate": 1.2970600134445133e-05, "loss": 0.0491, "step": 18404 }, { "epoch": 3.38, "learning_rate": 1.2969891210407426e-05, "loss": 0.0503, "step": 18405 }, { "epoch": 3.38, "learning_rate": 1.296918226999961e-05, "loss": 0.0449, "step": 18406 }, { "epoch": 3.38, "learning_rate": 1.2968473313225593e-05, "loss": 0.0447, "step": 18407 }, { "epoch": 3.38, "learning_rate": 1.2967764340089281e-05, "loss": 0.0482, "step": 18408 }, { "epoch": 3.38, "learning_rate": 1.2967055350594584e-05, "loss": 0.0529, "step": 18409 }, { "epoch": 3.38, "learning_rate": 1.2966346344745407e-05, "loss": 0.0418, "step": 18410 }, { "epoch": 3.38, "learning_rate": 1.2965637322545658e-05, "loss": 0.0453, "step": 18411 }, { "epoch": 3.38, "learning_rate": 1.2964928283999248e-05, "loss": 0.0425, "step": 18412 }, { "epoch": 3.38, "learning_rate": 1.2964219229110087e-05, "loss": 0.0413, "step": 18413 }, { "epoch": 3.38, "learning_rate": 1.2963510157882074e-05, "loss": 0.0528, "step": 18414 }, { "epoch": 3.38, "learning_rate": 1.2962801070319127e-05, "loss": 0.0431, "step": 18415 }, { "epoch": 3.38, "learning_rate": 1.296209196642515e-05, "loss": 0.0516, "step": 18416 }, { "epoch": 3.38, "learning_rate": 1.2961382846204056e-05, "loss": 0.0451, "step": 18417 }, { "epoch": 3.38, "learning_rate": 1.2960673709659743e-05, "loss": 0.0474, "step": 18418 }, { "epoch": 3.38, "learning_rate": 1.2959964556796137e-05, "loss": 0.052, "step": 18419 }, { "epoch": 3.38, "learning_rate": 1.2959255387617132e-05, "loss": 0.0487, "step": 18420 }, { "epoch": 3.38, "learning_rate": 1.2958546202126638e-05, "loss": 0.0532, "step": 18421 }, { "epoch": 3.38, "learning_rate": 1.2957837000328575e-05, "loss": 0.0352, "step": 18422 }, { "epoch": 3.38, "learning_rate": 1.2957127782226842e-05, "loss": 0.0534, "step": 18423 }, { "epoch": 3.38, "learning_rate": 1.295641854782535e-05, "loss": 0.054, "step": 18424 }, { "epoch": 3.38, "learning_rate": 1.2955709297128013e-05, "loss": 0.044, "step": 18425 }, { "epoch": 3.38, "learning_rate": 1.2955000030138734e-05, "loss": 0.0529, "step": 18426 }, { "epoch": 3.38, "learning_rate": 1.2954290746861426e-05, "loss": 0.0503, "step": 18427 }, { "epoch": 3.38, "learning_rate": 1.29535814473e-05, "loss": 0.051, "step": 18428 }, { "epoch": 3.38, "learning_rate": 1.2952872131458362e-05, "loss": 0.0439, "step": 18429 }, { "epoch": 3.38, "learning_rate": 1.2952162799340424e-05, "loss": 0.0468, "step": 18430 }, { "epoch": 3.38, "learning_rate": 1.2951453450950098e-05, "loss": 0.0462, "step": 18431 }, { "epoch": 3.38, "learning_rate": 1.295074408629129e-05, "loss": 0.0539, "step": 18432 }, { "epoch": 3.38, "learning_rate": 1.295003470536791e-05, "loss": 0.0483, "step": 18433 }, { "epoch": 3.38, "learning_rate": 1.2949325308183871e-05, "loss": 0.0528, "step": 18434 }, { "epoch": 3.38, "learning_rate": 1.2948615894743082e-05, "loss": 0.0474, "step": 18435 }, { "epoch": 3.38, "learning_rate": 1.2947906465049452e-05, "loss": 0.0503, "step": 18436 }, { "epoch": 3.38, "learning_rate": 1.2947197019106894e-05, "loss": 0.0447, "step": 18437 }, { "epoch": 3.38, "learning_rate": 1.2946487556919314e-05, "loss": 0.0463, "step": 18438 }, { "epoch": 3.38, "learning_rate": 1.2945778078490627e-05, "loss": 0.0421, "step": 18439 }, { "epoch": 3.38, "learning_rate": 1.2945068583824745e-05, "loss": 0.0591, "step": 18440 }, { "epoch": 3.38, "learning_rate": 1.2944359072925571e-05, "loss": 0.0521, "step": 18441 }, { "epoch": 3.38, "learning_rate": 1.2943649545797023e-05, "loss": 0.0462, "step": 18442 }, { "epoch": 3.38, "learning_rate": 1.2942940002443011e-05, "loss": 0.0556, "step": 18443 }, { "epoch": 3.38, "learning_rate": 1.2942230442867444e-05, "loss": 0.0466, "step": 18444 }, { "epoch": 3.38, "learning_rate": 1.2941520867074232e-05, "loss": 0.0413, "step": 18445 }, { "epoch": 3.38, "learning_rate": 1.2940811275067295e-05, "loss": 0.0439, "step": 18446 }, { "epoch": 3.38, "learning_rate": 1.2940101666850531e-05, "loss": 0.0435, "step": 18447 }, { "epoch": 3.38, "learning_rate": 1.2939392042427858e-05, "loss": 0.0414, "step": 18448 }, { "epoch": 3.38, "learning_rate": 1.2938682401803188e-05, "loss": 0.051, "step": 18449 }, { "epoch": 3.38, "learning_rate": 1.2937972744980431e-05, "loss": 0.0426, "step": 18450 }, { "epoch": 3.38, "learning_rate": 1.2937263071963502e-05, "loss": 0.0413, "step": 18451 }, { "epoch": 3.38, "learning_rate": 1.293655338275631e-05, "loss": 0.0431, "step": 18452 }, { "epoch": 3.38, "learning_rate": 1.2935843677362765e-05, "loss": 0.0475, "step": 18453 }, { "epoch": 3.38, "learning_rate": 1.293513395578678e-05, "loss": 0.051, "step": 18454 }, { "epoch": 3.38, "learning_rate": 1.2934424218032271e-05, "loss": 0.0453, "step": 18455 }, { "epoch": 3.38, "learning_rate": 1.2933714464103143e-05, "loss": 0.0454, "step": 18456 }, { "epoch": 3.38, "learning_rate": 1.2933004694003315e-05, "loss": 0.0471, "step": 18457 }, { "epoch": 3.38, "learning_rate": 1.2932294907736695e-05, "loss": 0.0464, "step": 18458 }, { "epoch": 3.39, "learning_rate": 1.2931585105307199e-05, "loss": 0.0528, "step": 18459 }, { "epoch": 3.39, "learning_rate": 1.2930875286718737e-05, "loss": 0.0581, "step": 18460 }, { "epoch": 3.39, "learning_rate": 1.293016545197522e-05, "loss": 0.0512, "step": 18461 }, { "epoch": 3.39, "learning_rate": 1.2929455601080562e-05, "loss": 0.048, "step": 18462 }, { "epoch": 3.39, "learning_rate": 1.292874573403868e-05, "loss": 0.0509, "step": 18463 }, { "epoch": 3.39, "learning_rate": 1.2928035850853477e-05, "loss": 0.0487, "step": 18464 }, { "epoch": 3.39, "learning_rate": 1.2927325951528873e-05, "loss": 0.0493, "step": 18465 }, { "epoch": 3.39, "learning_rate": 1.2926616036068782e-05, "loss": 0.0451, "step": 18466 }, { "epoch": 3.39, "learning_rate": 1.2925906104477115e-05, "loss": 0.0485, "step": 18467 }, { "epoch": 3.39, "learning_rate": 1.2925196156757784e-05, "loss": 0.0493, "step": 18468 }, { "epoch": 3.39, "learning_rate": 1.2924486192914705e-05, "loss": 0.046, "step": 18469 }, { "epoch": 3.39, "learning_rate": 1.2923776212951788e-05, "loss": 0.0512, "step": 18470 }, { "epoch": 3.39, "learning_rate": 1.2923066216872946e-05, "loss": 0.0427, "step": 18471 }, { "epoch": 3.39, "learning_rate": 1.2922356204682101e-05, "loss": 0.0489, "step": 18472 }, { "epoch": 3.39, "learning_rate": 1.2921646176383154e-05, "loss": 0.0582, "step": 18473 }, { "epoch": 3.39, "learning_rate": 1.2920936131980026e-05, "loss": 0.045, "step": 18474 }, { "epoch": 3.39, "learning_rate": 1.292022607147663e-05, "loss": 0.0489, "step": 18475 }, { "epoch": 3.39, "learning_rate": 1.2919515994876878e-05, "loss": 0.0581, "step": 18476 }, { "epoch": 3.39, "learning_rate": 1.2918805902184688e-05, "loss": 0.0489, "step": 18477 }, { "epoch": 3.39, "learning_rate": 1.2918095793403973e-05, "loss": 0.0402, "step": 18478 }, { "epoch": 3.39, "learning_rate": 1.2917385668538643e-05, "loss": 0.0412, "step": 18479 }, { "epoch": 3.39, "learning_rate": 1.2916675527592612e-05, "loss": 0.0506, "step": 18480 }, { "epoch": 3.39, "learning_rate": 1.2915965370569803e-05, "loss": 0.0484, "step": 18481 }, { "epoch": 3.39, "learning_rate": 1.291525519747412e-05, "loss": 0.0537, "step": 18482 }, { "epoch": 3.39, "learning_rate": 1.2914545008309484e-05, "loss": 0.0471, "step": 18483 }, { "epoch": 3.39, "learning_rate": 1.2913834803079813e-05, "loss": 0.0525, "step": 18484 }, { "epoch": 3.39, "learning_rate": 1.2913124581789007e-05, "loss": 0.0493, "step": 18485 }, { "epoch": 3.39, "learning_rate": 1.2912414344440993e-05, "loss": 0.0527, "step": 18486 }, { "epoch": 3.39, "learning_rate": 1.2911704091039685e-05, "loss": 0.0477, "step": 18487 }, { "epoch": 3.39, "learning_rate": 1.2910993821588995e-05, "loss": 0.0587, "step": 18488 }, { "epoch": 3.39, "learning_rate": 1.2910283536092835e-05, "loss": 0.053, "step": 18489 }, { "epoch": 3.39, "learning_rate": 1.2909573234555133e-05, "loss": 0.0456, "step": 18490 }, { "epoch": 3.39, "learning_rate": 1.2908862916979787e-05, "loss": 0.0558, "step": 18491 }, { "epoch": 3.39, "learning_rate": 1.2908152583370723e-05, "loss": 0.0506, "step": 18492 }, { "epoch": 3.39, "learning_rate": 1.2907442233731855e-05, "loss": 0.0441, "step": 18493 }, { "epoch": 3.39, "learning_rate": 1.2906731868067096e-05, "loss": 0.0434, "step": 18494 }, { "epoch": 3.39, "learning_rate": 1.2906021486380363e-05, "loss": 0.0444, "step": 18495 }, { "epoch": 3.39, "learning_rate": 1.2905311088675571e-05, "loss": 0.053, "step": 18496 }, { "epoch": 3.39, "learning_rate": 1.2904600674956637e-05, "loss": 0.0445, "step": 18497 }, { "epoch": 3.39, "learning_rate": 1.2903890245227474e-05, "loss": 0.0494, "step": 18498 }, { "epoch": 3.39, "learning_rate": 1.2903179799492005e-05, "loss": 0.0392, "step": 18499 }, { "epoch": 3.39, "learning_rate": 1.2902469337754136e-05, "loss": 0.0535, "step": 18500 }, { "epoch": 3.39, "learning_rate": 1.290175886001779e-05, "loss": 0.0486, "step": 18501 }, { "epoch": 3.39, "learning_rate": 1.2901048366286881e-05, "loss": 0.0482, "step": 18502 }, { "epoch": 3.39, "learning_rate": 1.2900337856565326e-05, "loss": 0.0467, "step": 18503 }, { "epoch": 3.39, "learning_rate": 1.2899627330857039e-05, "loss": 0.0395, "step": 18504 }, { "epoch": 3.39, "learning_rate": 1.289891678916594e-05, "loss": 0.0545, "step": 18505 }, { "epoch": 3.39, "learning_rate": 1.2898206231495944e-05, "loss": 0.0453, "step": 18506 }, { "epoch": 3.39, "learning_rate": 1.2897495657850965e-05, "loss": 0.0444, "step": 18507 }, { "epoch": 3.39, "learning_rate": 1.2896785068234925e-05, "loss": 0.0375, "step": 18508 }, { "epoch": 3.39, "learning_rate": 1.2896074462651736e-05, "loss": 0.042, "step": 18509 }, { "epoch": 3.39, "learning_rate": 1.2895363841105315e-05, "loss": 0.0439, "step": 18510 }, { "epoch": 3.39, "learning_rate": 1.2894653203599583e-05, "loss": 0.0438, "step": 18511 }, { "epoch": 3.39, "learning_rate": 1.2893942550138453e-05, "loss": 0.0458, "step": 18512 }, { "epoch": 3.4, "learning_rate": 1.2893231880725845e-05, "loss": 0.0477, "step": 18513 }, { "epoch": 3.4, "learning_rate": 1.2892521195365679e-05, "loss": 0.0533, "step": 18514 }, { "epoch": 3.4, "learning_rate": 1.2891810494061865e-05, "loss": 0.0528, "step": 18515 }, { "epoch": 3.4, "learning_rate": 1.289109977681832e-05, "loss": 0.0489, "step": 18516 }, { "epoch": 3.4, "learning_rate": 1.2890389043638974e-05, "loss": 0.0524, "step": 18517 }, { "epoch": 3.4, "learning_rate": 1.2889678294527728e-05, "loss": 0.0443, "step": 18518 }, { "epoch": 3.4, "learning_rate": 1.288896752948851e-05, "loss": 0.0489, "step": 18519 }, { "epoch": 3.4, "learning_rate": 1.2888256748525236e-05, "loss": 0.0549, "step": 18520 }, { "epoch": 3.4, "learning_rate": 1.2887545951641824e-05, "loss": 0.0397, "step": 18521 }, { "epoch": 3.4, "learning_rate": 1.2886835138842186e-05, "loss": 0.0417, "step": 18522 }, { "epoch": 3.4, "learning_rate": 1.2886124310130252e-05, "loss": 0.0389, "step": 18523 }, { "epoch": 3.4, "learning_rate": 1.2885413465509929e-05, "loss": 0.0495, "step": 18524 }, { "epoch": 3.4, "learning_rate": 1.288470260498514e-05, "loss": 0.0545, "step": 18525 }, { "epoch": 3.4, "learning_rate": 1.2883991728559805e-05, "loss": 0.0408, "step": 18526 }, { "epoch": 3.4, "learning_rate": 1.2883280836237837e-05, "loss": 0.0512, "step": 18527 }, { "epoch": 3.4, "learning_rate": 1.2882569928023159e-05, "loss": 0.0534, "step": 18528 }, { "epoch": 3.4, "learning_rate": 1.2881859003919688e-05, "loss": 0.0464, "step": 18529 }, { "epoch": 3.4, "learning_rate": 1.2881148063931344e-05, "loss": 0.0527, "step": 18530 }, { "epoch": 3.4, "learning_rate": 1.2880437108062043e-05, "loss": 0.0493, "step": 18531 }, { "epoch": 3.4, "learning_rate": 1.2879726136315705e-05, "loss": 0.0508, "step": 18532 }, { "epoch": 3.4, "learning_rate": 1.2879015148696245e-05, "loss": 0.0465, "step": 18533 }, { "epoch": 3.4, "learning_rate": 1.2878304145207593e-05, "loss": 0.057, "step": 18534 }, { "epoch": 3.4, "learning_rate": 1.2877593125853656e-05, "loss": 0.0423, "step": 18535 }, { "epoch": 3.4, "learning_rate": 1.287688209063836e-05, "loss": 0.0467, "step": 18536 }, { "epoch": 3.4, "learning_rate": 1.2876171039565623e-05, "loss": 0.05, "step": 18537 }, { "epoch": 3.4, "learning_rate": 1.2875459972639367e-05, "loss": 0.0475, "step": 18538 }, { "epoch": 3.4, "learning_rate": 1.2874748889863504e-05, "loss": 0.0461, "step": 18539 }, { "epoch": 3.4, "learning_rate": 1.2874037791241959e-05, "loss": 0.0467, "step": 18540 }, { "epoch": 3.4, "learning_rate": 1.2873326676778651e-05, "loss": 0.0474, "step": 18541 }, { "epoch": 3.4, "learning_rate": 1.2872615546477496e-05, "loss": 0.0482, "step": 18542 }, { "epoch": 3.4, "learning_rate": 1.2871904400342418e-05, "loss": 0.0414, "step": 18543 }, { "epoch": 3.4, "learning_rate": 1.2871193238377337e-05, "loss": 0.0482, "step": 18544 }, { "epoch": 3.4, "learning_rate": 1.2870482060586173e-05, "loss": 0.0436, "step": 18545 }, { "epoch": 3.4, "learning_rate": 1.2869770866972841e-05, "loss": 0.0469, "step": 18546 }, { "epoch": 3.4, "learning_rate": 1.286905965754127e-05, "loss": 0.0496, "step": 18547 }, { "epoch": 3.4, "learning_rate": 1.286834843229537e-05, "loss": 0.0489, "step": 18548 }, { "epoch": 3.4, "learning_rate": 1.2867637191239068e-05, "loss": 0.0445, "step": 18549 }, { "epoch": 3.4, "learning_rate": 1.2866925934376285e-05, "loss": 0.0332, "step": 18550 }, { "epoch": 3.4, "learning_rate": 1.2866214661710931e-05, "loss": 0.0488, "step": 18551 }, { "epoch": 3.4, "learning_rate": 1.286550337324694e-05, "loss": 0.0502, "step": 18552 }, { "epoch": 3.4, "learning_rate": 1.286479206898823e-05, "loss": 0.0425, "step": 18553 }, { "epoch": 3.4, "learning_rate": 1.2864080748938717e-05, "loss": 0.0393, "step": 18554 }, { "epoch": 3.4, "learning_rate": 1.2863369413102325e-05, "loss": 0.0429, "step": 18555 }, { "epoch": 3.4, "learning_rate": 1.286265806148297e-05, "loss": 0.0485, "step": 18556 }, { "epoch": 3.4, "learning_rate": 1.2861946694084578e-05, "loss": 0.0446, "step": 18557 }, { "epoch": 3.4, "learning_rate": 1.286123531091107e-05, "loss": 0.045, "step": 18558 }, { "epoch": 3.4, "learning_rate": 1.2860523911966367e-05, "loss": 0.0481, "step": 18559 }, { "epoch": 3.4, "learning_rate": 1.2859812497254388e-05, "loss": 0.0513, "step": 18560 }, { "epoch": 3.4, "learning_rate": 1.2859101066779056e-05, "loss": 0.0477, "step": 18561 }, { "epoch": 3.4, "learning_rate": 1.2858389620544288e-05, "loss": 0.0571, "step": 18562 }, { "epoch": 3.4, "learning_rate": 1.2857678158554014e-05, "loss": 0.0519, "step": 18563 }, { "epoch": 3.4, "learning_rate": 1.2856966680812148e-05, "loss": 0.0526, "step": 18564 }, { "epoch": 3.4, "learning_rate": 1.2856255187322616e-05, "loss": 0.0481, "step": 18565 }, { "epoch": 3.4, "learning_rate": 1.285554367808934e-05, "loss": 0.0416, "step": 18566 }, { "epoch": 3.4, "learning_rate": 1.2854832153116239e-05, "loss": 0.0353, "step": 18567 }, { "epoch": 3.41, "learning_rate": 1.2854120612407237e-05, "loss": 0.05, "step": 18568 }, { "epoch": 3.41, "learning_rate": 1.2853409055966253e-05, "loss": 0.0419, "step": 18569 }, { "epoch": 3.41, "learning_rate": 1.2852697483797216e-05, "loss": 0.0433, "step": 18570 }, { "epoch": 3.41, "learning_rate": 1.285198589590404e-05, "loss": 0.0472, "step": 18571 }, { "epoch": 3.41, "learning_rate": 1.285127429229065e-05, "loss": 0.0508, "step": 18572 }, { "epoch": 3.41, "learning_rate": 1.2850562672960973e-05, "loss": 0.0458, "step": 18573 }, { "epoch": 3.41, "learning_rate": 1.2849851037918925e-05, "loss": 0.0428, "step": 18574 }, { "epoch": 3.41, "learning_rate": 1.284913938716843e-05, "loss": 0.0551, "step": 18575 }, { "epoch": 3.41, "learning_rate": 1.2848427720713417e-05, "loss": 0.0502, "step": 18576 }, { "epoch": 3.41, "learning_rate": 1.2847716038557798e-05, "loss": 0.0458, "step": 18577 }, { "epoch": 3.41, "learning_rate": 1.2847004340705505e-05, "loss": 0.0485, "step": 18578 }, { "epoch": 3.41, "learning_rate": 1.2846292627160459e-05, "loss": 0.0418, "step": 18579 }, { "epoch": 3.41, "learning_rate": 1.2845580897926578e-05, "loss": 0.0496, "step": 18580 }, { "epoch": 3.41, "learning_rate": 1.284486915300779e-05, "loss": 0.049, "step": 18581 }, { "epoch": 3.41, "learning_rate": 1.2844157392408013e-05, "loss": 0.0432, "step": 18582 }, { "epoch": 3.41, "learning_rate": 1.2843445616131177e-05, "loss": 0.0458, "step": 18583 }, { "epoch": 3.41, "learning_rate": 1.2842733824181201e-05, "loss": 0.0467, "step": 18584 }, { "epoch": 3.41, "learning_rate": 1.2842022016562015e-05, "loss": 0.0417, "step": 18585 }, { "epoch": 3.41, "learning_rate": 1.2841310193277529e-05, "loss": 0.0389, "step": 18586 }, { "epoch": 3.41, "learning_rate": 1.284059835433168e-05, "loss": 0.0475, "step": 18587 }, { "epoch": 3.41, "learning_rate": 1.2839886499728384e-05, "loss": 0.0434, "step": 18588 }, { "epoch": 3.41, "learning_rate": 1.2839174629471569e-05, "loss": 0.0512, "step": 18589 }, { "epoch": 3.41, "learning_rate": 1.2838462743565154e-05, "loss": 0.0526, "step": 18590 }, { "epoch": 3.41, "learning_rate": 1.2837750842013068e-05, "loss": 0.0471, "step": 18591 }, { "epoch": 3.41, "learning_rate": 1.2837038924819232e-05, "loss": 0.0423, "step": 18592 }, { "epoch": 3.41, "learning_rate": 1.2836326991987571e-05, "loss": 0.0427, "step": 18593 }, { "epoch": 3.41, "learning_rate": 1.2835615043522012e-05, "loss": 0.0397, "step": 18594 }, { "epoch": 3.41, "learning_rate": 1.283490307942647e-05, "loss": 0.0457, "step": 18595 }, { "epoch": 3.41, "learning_rate": 1.283419109970488e-05, "loss": 0.0458, "step": 18596 }, { "epoch": 3.41, "learning_rate": 1.2833479104361163e-05, "loss": 0.0547, "step": 18597 }, { "epoch": 3.41, "learning_rate": 1.2832767093399242e-05, "loss": 0.0422, "step": 18598 }, { "epoch": 3.41, "learning_rate": 1.283205506682304e-05, "loss": 0.0435, "step": 18599 }, { "epoch": 3.41, "learning_rate": 1.2831343024636488e-05, "loss": 0.0526, "step": 18600 }, { "epoch": 3.41, "learning_rate": 1.2830630966843507e-05, "loss": 0.0478, "step": 18601 }, { "epoch": 3.41, "learning_rate": 1.2829918893448016e-05, "loss": 0.0442, "step": 18602 }, { "epoch": 3.41, "learning_rate": 1.2829206804453953e-05, "loss": 0.0357, "step": 18603 }, { "epoch": 3.41, "learning_rate": 1.2828494699865233e-05, "loss": 0.0471, "step": 18604 }, { "epoch": 3.41, "learning_rate": 1.2827782579685781e-05, "loss": 0.05, "step": 18605 }, { "epoch": 3.41, "learning_rate": 1.282707044391953e-05, "loss": 0.0433, "step": 18606 }, { "epoch": 3.41, "learning_rate": 1.2826358292570398e-05, "loss": 0.0492, "step": 18607 }, { "epoch": 3.41, "learning_rate": 1.2825646125642312e-05, "loss": 0.0543, "step": 18608 }, { "epoch": 3.41, "learning_rate": 1.28249339431392e-05, "loss": 0.0454, "step": 18609 }, { "epoch": 3.41, "learning_rate": 1.2824221745064984e-05, "loss": 0.054, "step": 18610 }, { "epoch": 3.41, "learning_rate": 1.2823509531423592e-05, "loss": 0.0491, "step": 18611 }, { "epoch": 3.41, "learning_rate": 1.2822797302218954e-05, "loss": 0.0511, "step": 18612 }, { "epoch": 3.41, "learning_rate": 1.2822085057454985e-05, "loss": 0.0564, "step": 18613 }, { "epoch": 3.41, "learning_rate": 1.2821372797135621e-05, "loss": 0.0456, "step": 18614 }, { "epoch": 3.41, "learning_rate": 1.2820660521264784e-05, "loss": 0.0537, "step": 18615 }, { "epoch": 3.41, "learning_rate": 1.2819948229846397e-05, "loss": 0.0443, "step": 18616 }, { "epoch": 3.41, "learning_rate": 1.281923592288439e-05, "loss": 0.0392, "step": 18617 }, { "epoch": 3.41, "learning_rate": 1.2818523600382691e-05, "loss": 0.0422, "step": 18618 }, { "epoch": 3.41, "learning_rate": 1.281781126234522e-05, "loss": 0.0513, "step": 18619 }, { "epoch": 3.41, "learning_rate": 1.2817098908775909e-05, "loss": 0.0498, "step": 18620 }, { "epoch": 3.41, "learning_rate": 1.2816386539678685e-05, "loss": 0.0498, "step": 18621 }, { "epoch": 3.42, "learning_rate": 1.281567415505747e-05, "loss": 0.0373, "step": 18622 }, { "epoch": 3.42, "learning_rate": 1.2814961754916193e-05, "loss": 0.0561, "step": 18623 }, { "epoch": 3.42, "learning_rate": 1.2814249339258783e-05, "loss": 0.0504, "step": 18624 }, { "epoch": 3.42, "learning_rate": 1.2813536908089163e-05, "loss": 0.0463, "step": 18625 }, { "epoch": 3.42, "learning_rate": 1.2812824461411263e-05, "loss": 0.037, "step": 18626 }, { "epoch": 3.42, "learning_rate": 1.2812111999229009e-05, "loss": 0.0525, "step": 18627 }, { "epoch": 3.42, "learning_rate": 1.2811399521546324e-05, "loss": 0.039, "step": 18628 }, { "epoch": 3.42, "learning_rate": 1.2810687028367142e-05, "loss": 0.0478, "step": 18629 }, { "epoch": 3.42, "learning_rate": 1.2809974519695388e-05, "loss": 0.0448, "step": 18630 }, { "epoch": 3.42, "learning_rate": 1.2809261995534987e-05, "loss": 0.044, "step": 18631 }, { "epoch": 3.42, "learning_rate": 1.280854945588987e-05, "loss": 0.0464, "step": 18632 }, { "epoch": 3.42, "learning_rate": 1.2807836900763962e-05, "loss": 0.0455, "step": 18633 }, { "epoch": 3.42, "learning_rate": 1.2807124330161188e-05, "loss": 0.0449, "step": 18634 }, { "epoch": 3.42, "learning_rate": 1.2806411744085483e-05, "loss": 0.0502, "step": 18635 }, { "epoch": 3.42, "learning_rate": 1.2805699142540768e-05, "loss": 0.043, "step": 18636 }, { "epoch": 3.42, "learning_rate": 1.2804986525530974e-05, "loss": 0.0455, "step": 18637 }, { "epoch": 3.42, "learning_rate": 1.2804273893060028e-05, "loss": 0.0411, "step": 18638 }, { "epoch": 3.42, "learning_rate": 1.2803561245131859e-05, "loss": 0.0536, "step": 18639 }, { "epoch": 3.42, "learning_rate": 1.2802848581750397e-05, "loss": 0.0485, "step": 18640 }, { "epoch": 3.42, "learning_rate": 1.2802135902919566e-05, "loss": 0.0454, "step": 18641 }, { "epoch": 3.42, "learning_rate": 1.2801423208643297e-05, "loss": 0.0387, "step": 18642 }, { "epoch": 3.42, "learning_rate": 1.280071049892552e-05, "loss": 0.0489, "step": 18643 }, { "epoch": 3.42, "learning_rate": 1.2799997773770157e-05, "loss": 0.0466, "step": 18644 }, { "epoch": 3.42, "learning_rate": 1.2799285033181144e-05, "loss": 0.0432, "step": 18645 }, { "epoch": 3.42, "learning_rate": 1.27985722771624e-05, "loss": 0.0419, "step": 18646 }, { "epoch": 3.42, "learning_rate": 1.2797859505717866e-05, "loss": 0.0527, "step": 18647 }, { "epoch": 3.42, "learning_rate": 1.2797146718851464e-05, "loss": 0.0519, "step": 18648 }, { "epoch": 3.42, "learning_rate": 1.2796433916567122e-05, "loss": 0.0516, "step": 18649 }, { "epoch": 3.42, "learning_rate": 1.2795721098868771e-05, "loss": 0.0436, "step": 18650 }, { "epoch": 3.42, "learning_rate": 1.2795008265760342e-05, "loss": 0.0515, "step": 18651 }, { "epoch": 3.42, "learning_rate": 1.2794295417245758e-05, "loss": 0.0426, "step": 18652 }, { "epoch": 3.42, "learning_rate": 1.2793582553328958e-05, "loss": 0.048, "step": 18653 }, { "epoch": 3.42, "learning_rate": 1.2792869674013863e-05, "loss": 0.0581, "step": 18654 }, { "epoch": 3.42, "learning_rate": 1.2792156779304402e-05, "loss": 0.0442, "step": 18655 }, { "epoch": 3.42, "learning_rate": 1.2791443869204513e-05, "loss": 0.0476, "step": 18656 }, { "epoch": 3.42, "learning_rate": 1.2790730943718114e-05, "loss": 0.0454, "step": 18657 }, { "epoch": 3.42, "learning_rate": 1.2790018002849145e-05, "loss": 0.0475, "step": 18658 }, { "epoch": 3.42, "learning_rate": 1.2789305046601528e-05, "loss": 0.0476, "step": 18659 }, { "epoch": 3.42, "learning_rate": 1.27885920749792e-05, "loss": 0.0501, "step": 18660 }, { "epoch": 3.42, "learning_rate": 1.2787879087986082e-05, "loss": 0.0445, "step": 18661 }, { "epoch": 3.42, "learning_rate": 1.2787166085626116e-05, "loss": 0.0461, "step": 18662 }, { "epoch": 3.42, "learning_rate": 1.2786453067903218e-05, "loss": 0.0441, "step": 18663 }, { "epoch": 3.42, "learning_rate": 1.278574003482133e-05, "loss": 0.055, "step": 18664 }, { "epoch": 3.42, "learning_rate": 1.2785026986384378e-05, "loss": 0.0498, "step": 18665 }, { "epoch": 3.42, "learning_rate": 1.278431392259629e-05, "loss": 0.0434, "step": 18666 }, { "epoch": 3.42, "learning_rate": 1.2783600843461e-05, "loss": 0.0446, "step": 18667 }, { "epoch": 3.42, "learning_rate": 1.2782887748982435e-05, "loss": 0.0434, "step": 18668 }, { "epoch": 3.42, "learning_rate": 1.2782174639164528e-05, "loss": 0.0419, "step": 18669 }, { "epoch": 3.42, "learning_rate": 1.2781461514011208e-05, "loss": 0.0454, "step": 18670 }, { "epoch": 3.42, "learning_rate": 1.2780748373526413e-05, "loss": 0.0518, "step": 18671 }, { "epoch": 3.42, "learning_rate": 1.278003521771406e-05, "loss": 0.0424, "step": 18672 }, { "epoch": 3.42, "learning_rate": 1.2779322046578091e-05, "loss": 0.0458, "step": 18673 }, { "epoch": 3.42, "learning_rate": 1.2778608860122434e-05, "loss": 0.0458, "step": 18674 }, { "epoch": 3.42, "learning_rate": 1.277789565835102e-05, "loss": 0.0448, "step": 18675 }, { "epoch": 3.42, "learning_rate": 1.2777182441267781e-05, "loss": 0.0457, "step": 18676 }, { "epoch": 3.43, "learning_rate": 1.2776469208876647e-05, "loss": 0.0439, "step": 18677 }, { "epoch": 3.43, "learning_rate": 1.277575596118155e-05, "loss": 0.0468, "step": 18678 }, { "epoch": 3.43, "learning_rate": 1.2775042698186418e-05, "loss": 0.047, "step": 18679 }, { "epoch": 3.43, "learning_rate": 1.2774329419895188e-05, "loss": 0.0392, "step": 18680 }, { "epoch": 3.43, "learning_rate": 1.2773616126311787e-05, "loss": 0.0429, "step": 18681 }, { "epoch": 3.43, "learning_rate": 1.2772902817440152e-05, "loss": 0.0525, "step": 18682 }, { "epoch": 3.43, "learning_rate": 1.277218949328421e-05, "loss": 0.0391, "step": 18683 }, { "epoch": 3.43, "learning_rate": 1.2771476153847898e-05, "loss": 0.0486, "step": 18684 }, { "epoch": 3.43, "learning_rate": 1.2770762799135139e-05, "loss": 0.0478, "step": 18685 }, { "epoch": 3.43, "learning_rate": 1.2770049429149873e-05, "loss": 0.0476, "step": 18686 }, { "epoch": 3.43, "learning_rate": 1.276933604389603e-05, "loss": 0.0459, "step": 18687 }, { "epoch": 3.43, "learning_rate": 1.2768622643377541e-05, "loss": 0.0444, "step": 18688 }, { "epoch": 3.43, "learning_rate": 1.2767909227598342e-05, "loss": 0.0513, "step": 18689 }, { "epoch": 3.43, "learning_rate": 1.2767195796562359e-05, "loss": 0.0481, "step": 18690 }, { "epoch": 3.43, "learning_rate": 1.2766482350273531e-05, "loss": 0.0503, "step": 18691 }, { "epoch": 3.43, "learning_rate": 1.2765768888735784e-05, "loss": 0.052, "step": 18692 }, { "epoch": 3.43, "learning_rate": 1.2765055411953055e-05, "loss": 0.0486, "step": 18693 }, { "epoch": 3.43, "learning_rate": 1.2764341919929276e-05, "loss": 0.0469, "step": 18694 }, { "epoch": 3.43, "learning_rate": 1.2763628412668383e-05, "loss": 0.054, "step": 18695 }, { "epoch": 3.43, "learning_rate": 1.2762914890174303e-05, "loss": 0.0534, "step": 18696 }, { "epoch": 3.43, "learning_rate": 1.2762201352450969e-05, "loss": 0.0454, "step": 18697 }, { "epoch": 3.43, "learning_rate": 1.2761487799502321e-05, "loss": 0.0425, "step": 18698 }, { "epoch": 3.43, "learning_rate": 1.2760774231332283e-05, "loss": 0.0388, "step": 18699 }, { "epoch": 3.43, "learning_rate": 1.2760060647944794e-05, "loss": 0.041, "step": 18700 }, { "epoch": 3.43, "learning_rate": 1.2759347049343786e-05, "loss": 0.0487, "step": 18701 }, { "epoch": 3.43, "learning_rate": 1.2758633435533195e-05, "loss": 0.0522, "step": 18702 }, { "epoch": 3.43, "learning_rate": 1.275791980651695e-05, "loss": 0.0467, "step": 18703 }, { "epoch": 3.43, "learning_rate": 1.2757206162298987e-05, "loss": 0.0444, "step": 18704 }, { "epoch": 3.43, "learning_rate": 1.2756492502883236e-05, "loss": 0.0534, "step": 18705 }, { "epoch": 3.43, "learning_rate": 1.2755778828273638e-05, "loss": 0.0478, "step": 18706 }, { "epoch": 3.43, "learning_rate": 1.2755065138474124e-05, "loss": 0.0426, "step": 18707 }, { "epoch": 3.43, "learning_rate": 1.2754351433488622e-05, "loss": 0.0457, "step": 18708 }, { "epoch": 3.43, "learning_rate": 1.275363771332107e-05, "loss": 0.0507, "step": 18709 }, { "epoch": 3.43, "learning_rate": 1.2752923977975404e-05, "loss": 0.0449, "step": 18710 }, { "epoch": 3.43, "learning_rate": 1.2752210227455557e-05, "loss": 0.0463, "step": 18711 }, { "epoch": 3.43, "learning_rate": 1.2751496461765463e-05, "loss": 0.0424, "step": 18712 }, { "epoch": 3.43, "learning_rate": 1.2750782680909055e-05, "loss": 0.0451, "step": 18713 }, { "epoch": 3.43, "learning_rate": 1.2750068884890268e-05, "loss": 0.0373, "step": 18714 }, { "epoch": 3.43, "learning_rate": 1.2749355073713036e-05, "loss": 0.0473, "step": 18715 }, { "epoch": 3.43, "learning_rate": 1.2748641247381297e-05, "loss": 0.0528, "step": 18716 }, { "epoch": 3.43, "learning_rate": 1.274792740589898e-05, "loss": 0.0434, "step": 18717 }, { "epoch": 3.43, "learning_rate": 1.2747213549270027e-05, "loss": 0.0459, "step": 18718 }, { "epoch": 3.43, "learning_rate": 1.2746499677498367e-05, "loss": 0.0476, "step": 18719 }, { "epoch": 3.43, "learning_rate": 1.2745785790587937e-05, "loss": 0.0476, "step": 18720 }, { "epoch": 3.43, "learning_rate": 1.2745071888542668e-05, "loss": 0.0556, "step": 18721 }, { "epoch": 3.43, "learning_rate": 1.2744357971366503e-05, "loss": 0.0492, "step": 18722 }, { "epoch": 3.43, "learning_rate": 1.2743644039063367e-05, "loss": 0.0414, "step": 18723 }, { "epoch": 3.43, "learning_rate": 1.2742930091637206e-05, "loss": 0.0405, "step": 18724 }, { "epoch": 3.43, "learning_rate": 1.2742216129091948e-05, "loss": 0.052, "step": 18725 }, { "epoch": 3.43, "learning_rate": 1.2741502151431531e-05, "loss": 0.0464, "step": 18726 }, { "epoch": 3.43, "learning_rate": 1.274078815865989e-05, "loss": 0.0436, "step": 18727 }, { "epoch": 3.43, "learning_rate": 1.2740074150780959e-05, "loss": 0.0517, "step": 18728 }, { "epoch": 3.43, "learning_rate": 1.2739360127798677e-05, "loss": 0.0527, "step": 18729 }, { "epoch": 3.43, "learning_rate": 1.2738646089716977e-05, "loss": 0.0465, "step": 18730 }, { "epoch": 3.43, "learning_rate": 1.2737932036539795e-05, "loss": 0.0415, "step": 18731 }, { "epoch": 3.44, "learning_rate": 1.2737217968271067e-05, "loss": 0.0522, "step": 18732 }, { "epoch": 3.44, "learning_rate": 1.2736503884914733e-05, "loss": 0.0419, "step": 18733 }, { "epoch": 3.44, "learning_rate": 1.2735789786474719e-05, "loss": 0.0468, "step": 18734 }, { "epoch": 3.44, "learning_rate": 1.2735075672954972e-05, "loss": 0.0485, "step": 18735 }, { "epoch": 3.44, "learning_rate": 1.2734361544359423e-05, "loss": 0.0525, "step": 18736 }, { "epoch": 3.44, "learning_rate": 1.273364740069201e-05, "loss": 0.0461, "step": 18737 }, { "epoch": 3.44, "learning_rate": 1.2732933241956667e-05, "loss": 0.0538, "step": 18738 }, { "epoch": 3.44, "learning_rate": 1.2732219068157335e-05, "loss": 0.0521, "step": 18739 }, { "epoch": 3.44, "learning_rate": 1.2731504879297946e-05, "loss": 0.0486, "step": 18740 }, { "epoch": 3.44, "learning_rate": 1.2730790675382437e-05, "loss": 0.048, "step": 18741 }, { "epoch": 3.44, "learning_rate": 1.2730076456414747e-05, "loss": 0.0423, "step": 18742 }, { "epoch": 3.44, "learning_rate": 1.2729362222398809e-05, "loss": 0.0516, "step": 18743 }, { "epoch": 3.44, "learning_rate": 1.2728647973338564e-05, "loss": 0.0486, "step": 18744 }, { "epoch": 3.44, "learning_rate": 1.2727933709237947e-05, "loss": 0.0453, "step": 18745 }, { "epoch": 3.44, "learning_rate": 1.2727219430100899e-05, "loss": 0.05, "step": 18746 }, { "epoch": 3.44, "learning_rate": 1.2726505135931349e-05, "loss": 0.0528, "step": 18747 }, { "epoch": 3.44, "learning_rate": 1.2725790826733242e-05, "loss": 0.042, "step": 18748 }, { "epoch": 3.44, "learning_rate": 1.272507650251051e-05, "loss": 0.0548, "step": 18749 }, { "epoch": 3.44, "learning_rate": 1.2724362163267092e-05, "loss": 0.0541, "step": 18750 }, { "epoch": 3.44, "learning_rate": 1.272364780900693e-05, "loss": 0.0458, "step": 18751 }, { "epoch": 3.44, "learning_rate": 1.2722933439733951e-05, "loss": 0.0439, "step": 18752 }, { "epoch": 3.44, "learning_rate": 1.2722219055452105e-05, "loss": 0.0511, "step": 18753 }, { "epoch": 3.44, "learning_rate": 1.272150465616532e-05, "loss": 0.0446, "step": 18754 }, { "epoch": 3.44, "learning_rate": 1.272079024187754e-05, "loss": 0.042, "step": 18755 }, { "epoch": 3.44, "learning_rate": 1.2720075812592698e-05, "loss": 0.0442, "step": 18756 }, { "epoch": 3.44, "learning_rate": 1.2719361368314737e-05, "loss": 0.0476, "step": 18757 }, { "epoch": 3.44, "learning_rate": 1.2718646909047592e-05, "loss": 0.0533, "step": 18758 }, { "epoch": 3.44, "learning_rate": 1.27179324347952e-05, "loss": 0.0444, "step": 18759 }, { "epoch": 3.44, "learning_rate": 1.2717217945561504e-05, "loss": 0.0457, "step": 18760 }, { "epoch": 3.44, "learning_rate": 1.2716503441350435e-05, "loss": 0.0449, "step": 18761 }, { "epoch": 3.44, "learning_rate": 1.2715788922165937e-05, "loss": 0.0499, "step": 18762 }, { "epoch": 3.44, "learning_rate": 1.2715074388011948e-05, "loss": 0.0384, "step": 18763 }, { "epoch": 3.44, "learning_rate": 1.2714359838892403e-05, "loss": 0.0506, "step": 18764 }, { "epoch": 3.44, "learning_rate": 1.2713645274811244e-05, "loss": 0.047, "step": 18765 }, { "epoch": 3.44, "learning_rate": 1.271293069577241e-05, "loss": 0.0507, "step": 18766 }, { "epoch": 3.44, "learning_rate": 1.2712216101779835e-05, "loss": 0.0436, "step": 18767 }, { "epoch": 3.44, "learning_rate": 1.2711501492837463e-05, "loss": 0.0407, "step": 18768 }, { "epoch": 3.44, "learning_rate": 1.2710786868949234e-05, "loss": 0.0405, "step": 18769 }, { "epoch": 3.44, "learning_rate": 1.2710072230119081e-05, "loss": 0.0534, "step": 18770 }, { "epoch": 3.44, "learning_rate": 1.270935757635095e-05, "loss": 0.0446, "step": 18771 }, { "epoch": 3.44, "learning_rate": 1.2708642907648774e-05, "loss": 0.0413, "step": 18772 }, { "epoch": 3.44, "learning_rate": 1.2707928224016495e-05, "loss": 0.0476, "step": 18773 }, { "epoch": 3.44, "learning_rate": 1.270721352545805e-05, "loss": 0.0481, "step": 18774 }, { "epoch": 3.44, "learning_rate": 1.2706498811977386e-05, "loss": 0.0503, "step": 18775 }, { "epoch": 3.44, "learning_rate": 1.2705784083578432e-05, "loss": 0.0433, "step": 18776 }, { "epoch": 3.44, "learning_rate": 1.2705069340265136e-05, "loss": 0.0486, "step": 18777 }, { "epoch": 3.44, "learning_rate": 1.2704354582041431e-05, "loss": 0.0421, "step": 18778 }, { "epoch": 3.44, "learning_rate": 1.2703639808911264e-05, "loss": 0.0483, "step": 18779 }, { "epoch": 3.44, "learning_rate": 1.270292502087857e-05, "loss": 0.0499, "step": 18780 }, { "epoch": 3.44, "learning_rate": 1.2702210217947289e-05, "loss": 0.0546, "step": 18781 }, { "epoch": 3.44, "learning_rate": 1.2701495400121363e-05, "loss": 0.0453, "step": 18782 }, { "epoch": 3.44, "learning_rate": 1.2700780567404728e-05, "loss": 0.0485, "step": 18783 }, { "epoch": 3.44, "learning_rate": 1.2700065719801331e-05, "loss": 0.0422, "step": 18784 }, { "epoch": 3.44, "learning_rate": 1.2699350857315105e-05, "loss": 0.0468, "step": 18785 }, { "epoch": 3.45, "learning_rate": 1.2698635979949996e-05, "loss": 0.0459, "step": 18786 }, { "epoch": 3.45, "learning_rate": 1.2697921087709943e-05, "loss": 0.0489, "step": 18787 }, { "epoch": 3.45, "learning_rate": 1.2697206180598884e-05, "loss": 0.0502, "step": 18788 }, { "epoch": 3.45, "learning_rate": 1.2696491258620763e-05, "loss": 0.0421, "step": 18789 }, { "epoch": 3.45, "learning_rate": 1.2695776321779518e-05, "loss": 0.0475, "step": 18790 }, { "epoch": 3.45, "learning_rate": 1.269506137007909e-05, "loss": 0.052, "step": 18791 }, { "epoch": 3.45, "learning_rate": 1.2694346403523424e-05, "loss": 0.0489, "step": 18792 }, { "epoch": 3.45, "learning_rate": 1.2693631422116455e-05, "loss": 0.0496, "step": 18793 }, { "epoch": 3.45, "learning_rate": 1.2692916425862125e-05, "loss": 0.0486, "step": 18794 }, { "epoch": 3.45, "learning_rate": 1.2692201414764378e-05, "loss": 0.046, "step": 18795 }, { "epoch": 3.45, "learning_rate": 1.2691486388827154e-05, "loss": 0.045, "step": 18796 }, { "epoch": 3.45, "learning_rate": 1.2690771348054396e-05, "loss": 0.0511, "step": 18797 }, { "epoch": 3.45, "learning_rate": 1.2690056292450041e-05, "loss": 0.0507, "step": 18798 }, { "epoch": 3.45, "learning_rate": 1.2689341222018033e-05, "loss": 0.048, "step": 18799 }, { "epoch": 3.45, "learning_rate": 1.268862613676231e-05, "loss": 0.0431, "step": 18800 }, { "epoch": 3.45, "learning_rate": 1.2687911036686824e-05, "loss": 0.0459, "step": 18801 }, { "epoch": 3.45, "learning_rate": 1.2687195921795503e-05, "loss": 0.0471, "step": 18802 }, { "epoch": 3.45, "learning_rate": 1.2686480792092298e-05, "loss": 0.0514, "step": 18803 }, { "epoch": 3.45, "learning_rate": 1.2685765647581146e-05, "loss": 0.0548, "step": 18804 }, { "epoch": 3.45, "learning_rate": 1.2685050488265993e-05, "loss": 0.0406, "step": 18805 }, { "epoch": 3.45, "learning_rate": 1.2684335314150777e-05, "loss": 0.0434, "step": 18806 }, { "epoch": 3.45, "learning_rate": 1.2683620125239444e-05, "loss": 0.0516, "step": 18807 }, { "epoch": 3.45, "learning_rate": 1.2682904921535932e-05, "loss": 0.0466, "step": 18808 }, { "epoch": 3.45, "learning_rate": 1.2682189703044185e-05, "loss": 0.0503, "step": 18809 }, { "epoch": 3.45, "learning_rate": 1.268147446976815e-05, "loss": 0.0503, "step": 18810 }, { "epoch": 3.45, "learning_rate": 1.2680759221711758e-05, "loss": 0.0461, "step": 18811 }, { "epoch": 3.45, "learning_rate": 1.2680043958878962e-05, "loss": 0.052, "step": 18812 }, { "epoch": 3.45, "learning_rate": 1.2679328681273701e-05, "loss": 0.0499, "step": 18813 }, { "epoch": 3.45, "learning_rate": 1.2678613388899916e-05, "loss": 0.0543, "step": 18814 }, { "epoch": 3.45, "learning_rate": 1.2677898081761555e-05, "loss": 0.0482, "step": 18815 }, { "epoch": 3.45, "learning_rate": 1.2677182759862554e-05, "loss": 0.0469, "step": 18816 }, { "epoch": 3.45, "learning_rate": 1.267646742320686e-05, "loss": 0.0432, "step": 18817 }, { "epoch": 3.45, "learning_rate": 1.2675752071798413e-05, "loss": 0.0448, "step": 18818 }, { "epoch": 3.45, "learning_rate": 1.267503670564116e-05, "loss": 0.0442, "step": 18819 }, { "epoch": 3.45, "learning_rate": 1.2674321324739038e-05, "loss": 0.0511, "step": 18820 }, { "epoch": 3.45, "learning_rate": 1.2673605929095998e-05, "loss": 0.0506, "step": 18821 }, { "epoch": 3.45, "learning_rate": 1.2672890518715982e-05, "loss": 0.0485, "step": 18822 }, { "epoch": 3.45, "learning_rate": 1.2672175093602928e-05, "loss": 0.0436, "step": 18823 }, { "epoch": 3.45, "learning_rate": 1.2671459653760781e-05, "loss": 0.0475, "step": 18824 }, { "epoch": 3.45, "learning_rate": 1.2670744199193488e-05, "loss": 0.047, "step": 18825 }, { "epoch": 3.45, "learning_rate": 1.2670028729904988e-05, "loss": 0.054, "step": 18826 }, { "epoch": 3.45, "learning_rate": 1.2669313245899226e-05, "loss": 0.052, "step": 18827 }, { "epoch": 3.45, "learning_rate": 1.2668597747180151e-05, "loss": 0.0459, "step": 18828 }, { "epoch": 3.45, "learning_rate": 1.26678822337517e-05, "loss": 0.0427, "step": 18829 }, { "epoch": 3.45, "learning_rate": 1.266716670561782e-05, "loss": 0.0488, "step": 18830 }, { "epoch": 3.45, "learning_rate": 1.2666451162782455e-05, "loss": 0.0419, "step": 18831 }, { "epoch": 3.45, "learning_rate": 1.266573560524955e-05, "loss": 0.0477, "step": 18832 }, { "epoch": 3.45, "learning_rate": 1.2665020033023043e-05, "loss": 0.0448, "step": 18833 }, { "epoch": 3.45, "learning_rate": 1.2664304446106889e-05, "loss": 0.0484, "step": 18834 }, { "epoch": 3.45, "learning_rate": 1.2663588844505026e-05, "loss": 0.0504, "step": 18835 }, { "epoch": 3.45, "learning_rate": 1.2662873228221394e-05, "loss": 0.0521, "step": 18836 }, { "epoch": 3.45, "learning_rate": 1.266215759725995e-05, "loss": 0.0468, "step": 18837 }, { "epoch": 3.45, "learning_rate": 1.2661441951624624e-05, "loss": 0.0414, "step": 18838 }, { "epoch": 3.45, "learning_rate": 1.2660726291319372e-05, "loss": 0.0467, "step": 18839 }, { "epoch": 3.45, "learning_rate": 1.266001061634813e-05, "loss": 0.047, "step": 18840 }, { "epoch": 3.46, "learning_rate": 1.265929492671485e-05, "loss": 0.0551, "step": 18841 }, { "epoch": 3.46, "learning_rate": 1.2658579222423473e-05, "loss": 0.0381, "step": 18842 }, { "epoch": 3.46, "learning_rate": 1.2657863503477947e-05, "loss": 0.0447, "step": 18843 }, { "epoch": 3.46, "learning_rate": 1.2657147769882215e-05, "loss": 0.0465, "step": 18844 }, { "epoch": 3.46, "learning_rate": 1.2656432021640218e-05, "loss": 0.044, "step": 18845 }, { "epoch": 3.46, "learning_rate": 1.2655716258755913e-05, "loss": 0.0488, "step": 18846 }, { "epoch": 3.46, "learning_rate": 1.2655000481233231e-05, "loss": 0.0398, "step": 18847 }, { "epoch": 3.46, "learning_rate": 1.2654284689076125e-05, "loss": 0.0515, "step": 18848 }, { "epoch": 3.46, "learning_rate": 1.2653568882288543e-05, "loss": 0.0456, "step": 18849 }, { "epoch": 3.46, "learning_rate": 1.2652853060874427e-05, "loss": 0.0502, "step": 18850 }, { "epoch": 3.46, "learning_rate": 1.2652137224837718e-05, "loss": 0.0585, "step": 18851 }, { "epoch": 3.46, "learning_rate": 1.265142137418237e-05, "loss": 0.035, "step": 18852 }, { "epoch": 3.46, "learning_rate": 1.2650705508912324e-05, "loss": 0.0477, "step": 18853 }, { "epoch": 3.46, "learning_rate": 1.264998962903153e-05, "loss": 0.0494, "step": 18854 }, { "epoch": 3.46, "learning_rate": 1.2649273734543928e-05, "loss": 0.0462, "step": 18855 }, { "epoch": 3.46, "learning_rate": 1.2648557825453467e-05, "loss": 0.0448, "step": 18856 }, { "epoch": 3.46, "learning_rate": 1.2647841901764095e-05, "loss": 0.0477, "step": 18857 }, { "epoch": 3.46, "learning_rate": 1.2647125963479757e-05, "loss": 0.0479, "step": 18858 }, { "epoch": 3.46, "learning_rate": 1.2646410010604397e-05, "loss": 0.0443, "step": 18859 }, { "epoch": 3.46, "learning_rate": 1.2645694043141963e-05, "loss": 0.0409, "step": 18860 }, { "epoch": 3.46, "learning_rate": 1.2644978061096402e-05, "loss": 0.0377, "step": 18861 }, { "epoch": 3.46, "learning_rate": 1.264426206447166e-05, "loss": 0.0506, "step": 18862 }, { "epoch": 3.46, "learning_rate": 1.2643546053271682e-05, "loss": 0.0406, "step": 18863 }, { "epoch": 3.46, "learning_rate": 1.264283002750042e-05, "loss": 0.0441, "step": 18864 }, { "epoch": 3.46, "learning_rate": 1.2642113987161815e-05, "loss": 0.0436, "step": 18865 }, { "epoch": 3.46, "learning_rate": 1.2641397932259814e-05, "loss": 0.0434, "step": 18866 }, { "epoch": 3.46, "learning_rate": 1.2640681862798367e-05, "loss": 0.0498, "step": 18867 }, { "epoch": 3.46, "learning_rate": 1.263996577878142e-05, "loss": 0.0477, "step": 18868 }, { "epoch": 3.46, "learning_rate": 1.263924968021292e-05, "loss": 0.0429, "step": 18869 }, { "epoch": 3.46, "learning_rate": 1.2638533567096813e-05, "loss": 0.0513, "step": 18870 }, { "epoch": 3.46, "learning_rate": 1.2637817439437048e-05, "loss": 0.0399, "step": 18871 }, { "epoch": 3.46, "learning_rate": 1.263710129723757e-05, "loss": 0.05, "step": 18872 }, { "epoch": 3.46, "learning_rate": 1.263638514050233e-05, "loss": 0.0423, "step": 18873 }, { "epoch": 3.46, "learning_rate": 1.2635668969235272e-05, "loss": 0.0433, "step": 18874 }, { "epoch": 3.46, "learning_rate": 1.2634952783440347e-05, "loss": 0.0529, "step": 18875 }, { "epoch": 3.46, "learning_rate": 1.2634236583121498e-05, "loss": 0.0542, "step": 18876 }, { "epoch": 3.46, "learning_rate": 1.2633520368282675e-05, "loss": 0.0432, "step": 18877 }, { "epoch": 3.46, "learning_rate": 1.2632804138927828e-05, "loss": 0.0511, "step": 18878 }, { "epoch": 3.46, "learning_rate": 1.2632087895060903e-05, "loss": 0.0479, "step": 18879 }, { "epoch": 3.46, "learning_rate": 1.2631371636685845e-05, "loss": 0.0519, "step": 18880 }, { "epoch": 3.46, "learning_rate": 1.2630655363806605e-05, "loss": 0.0386, "step": 18881 }, { "epoch": 3.46, "learning_rate": 1.2629939076427133e-05, "loss": 0.0404, "step": 18882 }, { "epoch": 3.46, "learning_rate": 1.2629222774551377e-05, "loss": 0.0528, "step": 18883 }, { "epoch": 3.46, "learning_rate": 1.262850645818328e-05, "loss": 0.0508, "step": 18884 }, { "epoch": 3.46, "learning_rate": 1.2627790127326796e-05, "loss": 0.0463, "step": 18885 }, { "epoch": 3.46, "learning_rate": 1.262707378198587e-05, "loss": 0.0493, "step": 18886 }, { "epoch": 3.46, "learning_rate": 1.2626357422164456e-05, "loss": 0.0423, "step": 18887 }, { "epoch": 3.46, "learning_rate": 1.2625641047866493e-05, "loss": 0.0445, "step": 18888 }, { "epoch": 3.46, "learning_rate": 1.2624924659095934e-05, "loss": 0.0472, "step": 18889 }, { "epoch": 3.46, "learning_rate": 1.2624208255856734e-05, "loss": 0.0481, "step": 18890 }, { "epoch": 3.46, "learning_rate": 1.2623491838152832e-05, "loss": 0.0484, "step": 18891 }, { "epoch": 3.46, "learning_rate": 1.2622775405988184e-05, "loss": 0.0541, "step": 18892 }, { "epoch": 3.46, "learning_rate": 1.2622058959366737e-05, "loss": 0.0428, "step": 18893 }, { "epoch": 3.46, "learning_rate": 1.262134249829244e-05, "loss": 0.0401, "step": 18894 }, { "epoch": 3.47, "learning_rate": 1.2620626022769239e-05, "loss": 0.0493, "step": 18895 }, { "epoch": 3.47, "learning_rate": 1.2619909532801087e-05, "loss": 0.0469, "step": 18896 }, { "epoch": 3.47, "learning_rate": 1.2619193028391931e-05, "loss": 0.0398, "step": 18897 }, { "epoch": 3.47, "learning_rate": 1.2618476509545724e-05, "loss": 0.0558, "step": 18898 }, { "epoch": 3.47, "learning_rate": 1.2617759976266413e-05, "loss": 0.0566, "step": 18899 }, { "epoch": 3.47, "learning_rate": 1.2617043428557947e-05, "loss": 0.0467, "step": 18900 }, { "epoch": 3.47, "learning_rate": 1.2616326866424278e-05, "loss": 0.0483, "step": 18901 }, { "epoch": 3.47, "learning_rate": 1.2615610289869351e-05, "loss": 0.0509, "step": 18902 }, { "epoch": 3.47, "learning_rate": 1.2614893698897121e-05, "loss": 0.0454, "step": 18903 }, { "epoch": 3.47, "learning_rate": 1.2614177093511533e-05, "loss": 0.0389, "step": 18904 }, { "epoch": 3.47, "learning_rate": 1.2613460473716543e-05, "loss": 0.0464, "step": 18905 }, { "epoch": 3.47, "learning_rate": 1.2612743839516095e-05, "loss": 0.0457, "step": 18906 }, { "epoch": 3.47, "learning_rate": 1.2612027190914144e-05, "loss": 0.0527, "step": 18907 }, { "epoch": 3.47, "learning_rate": 1.2611310527914638e-05, "loss": 0.0473, "step": 18908 }, { "epoch": 3.47, "learning_rate": 1.2610593850521524e-05, "loss": 0.0462, "step": 18909 }, { "epoch": 3.47, "learning_rate": 1.260987715873876e-05, "loss": 0.0452, "step": 18910 }, { "epoch": 3.47, "learning_rate": 1.2609160452570288e-05, "loss": 0.0463, "step": 18911 }, { "epoch": 3.47, "learning_rate": 1.2608443732020066e-05, "loss": 0.0408, "step": 18912 }, { "epoch": 3.47, "learning_rate": 1.2607726997092036e-05, "loss": 0.0522, "step": 18913 }, { "epoch": 3.47, "learning_rate": 1.2607010247790158e-05, "loss": 0.0465, "step": 18914 }, { "epoch": 3.47, "learning_rate": 1.2606293484118377e-05, "loss": 0.0404, "step": 18915 }, { "epoch": 3.47, "learning_rate": 1.2605576706080647e-05, "loss": 0.051, "step": 18916 }, { "epoch": 3.47, "learning_rate": 1.2604859913680913e-05, "loss": 0.0534, "step": 18917 }, { "epoch": 3.47, "learning_rate": 1.2604143106923135e-05, "loss": 0.0448, "step": 18918 }, { "epoch": 3.47, "learning_rate": 1.2603426285811256e-05, "loss": 0.0468, "step": 18919 }, { "epoch": 3.47, "learning_rate": 1.260270945034923e-05, "loss": 0.0475, "step": 18920 }, { "epoch": 3.47, "learning_rate": 1.260199260054101e-05, "loss": 0.042, "step": 18921 }, { "epoch": 3.47, "learning_rate": 1.2601275736390547e-05, "loss": 0.0492, "step": 18922 }, { "epoch": 3.47, "learning_rate": 1.260055885790179e-05, "loss": 0.0474, "step": 18923 }, { "epoch": 3.47, "learning_rate": 1.2599841965078689e-05, "loss": 0.0489, "step": 18924 }, { "epoch": 3.47, "learning_rate": 1.2599125057925203e-05, "loss": 0.0457, "step": 18925 }, { "epoch": 3.47, "learning_rate": 1.2598408136445278e-05, "loss": 0.0444, "step": 18926 }, { "epoch": 3.47, "learning_rate": 1.2597691200642862e-05, "loss": 0.0452, "step": 18927 }, { "epoch": 3.47, "learning_rate": 1.2596974250521914e-05, "loss": 0.0431, "step": 18928 }, { "epoch": 3.47, "learning_rate": 1.2596257286086387e-05, "loss": 0.0517, "step": 18929 }, { "epoch": 3.47, "learning_rate": 1.2595540307340222e-05, "loss": 0.0402, "step": 18930 }, { "epoch": 3.47, "learning_rate": 1.259482331428738e-05, "loss": 0.0406, "step": 18931 }, { "epoch": 3.47, "learning_rate": 1.2594106306931818e-05, "loss": 0.0461, "step": 18932 }, { "epoch": 3.47, "learning_rate": 1.2593389285277473e-05, "loss": 0.0521, "step": 18933 }, { "epoch": 3.47, "learning_rate": 1.2592672249328306e-05, "loss": 0.0415, "step": 18934 }, { "epoch": 3.47, "learning_rate": 1.2591955199088273e-05, "loss": 0.0485, "step": 18935 }, { "epoch": 3.47, "learning_rate": 1.259123813456132e-05, "loss": 0.0454, "step": 18936 }, { "epoch": 3.47, "learning_rate": 1.2590521055751399e-05, "loss": 0.0478, "step": 18937 }, { "epoch": 3.47, "learning_rate": 1.2589803962662473e-05, "loss": 0.0542, "step": 18938 }, { "epoch": 3.47, "learning_rate": 1.2589086855298477e-05, "loss": 0.0488, "step": 18939 }, { "epoch": 3.47, "learning_rate": 1.2588369733663379e-05, "loss": 0.0481, "step": 18940 }, { "epoch": 3.47, "learning_rate": 1.2587652597761126e-05, "loss": 0.0432, "step": 18941 }, { "epoch": 3.47, "learning_rate": 1.2586935447595674e-05, "loss": 0.0519, "step": 18942 }, { "epoch": 3.47, "learning_rate": 1.258621828317097e-05, "loss": 0.0487, "step": 18943 }, { "epoch": 3.47, "learning_rate": 1.2585501104490971e-05, "loss": 0.0452, "step": 18944 }, { "epoch": 3.47, "learning_rate": 1.2584783911559631e-05, "loss": 0.0435, "step": 18945 }, { "epoch": 3.47, "learning_rate": 1.2584066704380895e-05, "loss": 0.0494, "step": 18946 }, { "epoch": 3.47, "learning_rate": 1.2583349482958731e-05, "loss": 0.0477, "step": 18947 }, { "epoch": 3.47, "learning_rate": 1.2582632247297078e-05, "loss": 0.0508, "step": 18948 }, { "epoch": 3.47, "learning_rate": 1.2581914997399899e-05, "loss": 0.0401, "step": 18949 }, { "epoch": 3.48, "learning_rate": 1.2581197733271143e-05, "loss": 0.0499, "step": 18950 }, { "epoch": 3.48, "learning_rate": 1.2580480454914765e-05, "loss": 0.0475, "step": 18951 }, { "epoch": 3.48, "learning_rate": 1.2579763162334718e-05, "loss": 0.0486, "step": 18952 }, { "epoch": 3.48, "learning_rate": 1.2579045855534955e-05, "loss": 0.0459, "step": 18953 }, { "epoch": 3.48, "learning_rate": 1.2578328534519433e-05, "loss": 0.0473, "step": 18954 }, { "epoch": 3.48, "learning_rate": 1.2577611199292104e-05, "loss": 0.0505, "step": 18955 }, { "epoch": 3.48, "learning_rate": 1.2576893849856919e-05, "loss": 0.0469, "step": 18956 }, { "epoch": 3.48, "learning_rate": 1.2576176486217834e-05, "loss": 0.0435, "step": 18957 }, { "epoch": 3.48, "learning_rate": 1.2575459108378807e-05, "loss": 0.0492, "step": 18958 }, { "epoch": 3.48, "learning_rate": 1.2574741716343787e-05, "loss": 0.0455, "step": 18959 }, { "epoch": 3.48, "learning_rate": 1.2574024310116732e-05, "loss": 0.0454, "step": 18960 }, { "epoch": 3.48, "learning_rate": 1.2573306889701593e-05, "loss": 0.0489, "step": 18961 }, { "epoch": 3.48, "learning_rate": 1.257258945510233e-05, "loss": 0.0505, "step": 18962 }, { "epoch": 3.48, "learning_rate": 1.257187200632289e-05, "loss": 0.0464, "step": 18963 }, { "epoch": 3.48, "learning_rate": 1.2571154543367233e-05, "loss": 0.0489, "step": 18964 }, { "epoch": 3.48, "learning_rate": 1.257043706623931e-05, "loss": 0.0453, "step": 18965 }, { "epoch": 3.48, "learning_rate": 1.2569719574943076e-05, "loss": 0.048, "step": 18966 }, { "epoch": 3.48, "learning_rate": 1.256900206948249e-05, "loss": 0.0487, "step": 18967 }, { "epoch": 3.48, "learning_rate": 1.2568284549861505e-05, "loss": 0.0381, "step": 18968 }, { "epoch": 3.48, "learning_rate": 1.2567567016084076e-05, "loss": 0.0379, "step": 18969 }, { "epoch": 3.48, "learning_rate": 1.2566849468154157e-05, "loss": 0.0494, "step": 18970 }, { "epoch": 3.48, "learning_rate": 1.2566131906075702e-05, "loss": 0.052, "step": 18971 }, { "epoch": 3.48, "learning_rate": 1.2565414329852668e-05, "loss": 0.0495, "step": 18972 }, { "epoch": 3.48, "learning_rate": 1.2564696739489014e-05, "loss": 0.0582, "step": 18973 }, { "epoch": 3.48, "learning_rate": 1.2563979134988686e-05, "loss": 0.0449, "step": 18974 }, { "epoch": 3.48, "learning_rate": 1.2563261516355647e-05, "loss": 0.0511, "step": 18975 }, { "epoch": 3.48, "learning_rate": 1.256254388359385e-05, "loss": 0.0537, "step": 18976 }, { "epoch": 3.48, "learning_rate": 1.2561826236707253e-05, "loss": 0.0449, "step": 18977 }, { "epoch": 3.48, "learning_rate": 1.2561108575699808e-05, "loss": 0.0387, "step": 18978 }, { "epoch": 3.48, "learning_rate": 1.2560390900575472e-05, "loss": 0.0457, "step": 18979 }, { "epoch": 3.48, "learning_rate": 1.2559673211338204e-05, "loss": 0.0419, "step": 18980 }, { "epoch": 3.48, "learning_rate": 1.2558955507991955e-05, "loss": 0.0463, "step": 18981 }, { "epoch": 3.48, "learning_rate": 1.2558237790540684e-05, "loss": 0.0431, "step": 18982 }, { "epoch": 3.48, "learning_rate": 1.2557520058988346e-05, "loss": 0.0416, "step": 18983 }, { "epoch": 3.48, "learning_rate": 1.2556802313338899e-05, "loss": 0.046, "step": 18984 }, { "epoch": 3.48, "learning_rate": 1.2556084553596297e-05, "loss": 0.0504, "step": 18985 }, { "epoch": 3.48, "learning_rate": 1.2555366779764495e-05, "loss": 0.0479, "step": 18986 }, { "epoch": 3.48, "learning_rate": 1.2554648991847453e-05, "loss": 0.0502, "step": 18987 }, { "epoch": 3.48, "learning_rate": 1.2553931189849127e-05, "loss": 0.042, "step": 18988 }, { "epoch": 3.48, "learning_rate": 1.2553213373773472e-05, "loss": 0.0448, "step": 18989 }, { "epoch": 3.48, "learning_rate": 1.2552495543624443e-05, "loss": 0.0447, "step": 18990 }, { "epoch": 3.48, "learning_rate": 1.2551777699406002e-05, "loss": 0.0535, "step": 18991 }, { "epoch": 3.48, "learning_rate": 1.25510598411221e-05, "loss": 0.04, "step": 18992 }, { "epoch": 3.48, "learning_rate": 1.2550341968776698e-05, "loss": 0.0524, "step": 18993 }, { "epoch": 3.48, "learning_rate": 1.254962408237375e-05, "loss": 0.0522, "step": 18994 }, { "epoch": 3.48, "learning_rate": 1.2548906181917216e-05, "loss": 0.0409, "step": 18995 }, { "epoch": 3.48, "learning_rate": 1.254818826741105e-05, "loss": 0.0453, "step": 18996 }, { "epoch": 3.48, "learning_rate": 1.2547470338859211e-05, "loss": 0.0397, "step": 18997 }, { "epoch": 3.48, "learning_rate": 1.2546752396265655e-05, "loss": 0.0553, "step": 18998 }, { "epoch": 3.48, "learning_rate": 1.254603443963434e-05, "loss": 0.0429, "step": 18999 }, { "epoch": 3.48, "learning_rate": 1.2545316468969226e-05, "loss": 0.0458, "step": 19000 }, { "epoch": 3.48, "learning_rate": 1.2544598484274263e-05, "loss": 0.0508, "step": 19001 }, { "epoch": 3.48, "learning_rate": 1.254388048555342e-05, "loss": 0.043, "step": 19002 }, { "epoch": 3.48, "learning_rate": 1.2543162472810643e-05, "loss": 0.0466, "step": 19003 }, { "epoch": 3.49, "learning_rate": 1.2542444446049899e-05, "loss": 0.0407, "step": 19004 }, { "epoch": 3.49, "learning_rate": 1.2541726405275138e-05, "loss": 0.0415, "step": 19005 }, { "epoch": 3.49, "learning_rate": 1.2541008350490325e-05, "loss": 0.0413, "step": 19006 }, { "epoch": 3.49, "learning_rate": 1.2540290281699413e-05, "loss": 0.0415, "step": 19007 }, { "epoch": 3.49, "learning_rate": 1.2539572198906358e-05, "loss": 0.037, "step": 19008 }, { "epoch": 3.49, "learning_rate": 1.2538854102115128e-05, "loss": 0.042, "step": 19009 }, { "epoch": 3.49, "learning_rate": 1.253813599132967e-05, "loss": 0.0477, "step": 19010 }, { "epoch": 3.49, "learning_rate": 1.253741786655395e-05, "loss": 0.0422, "step": 19011 }, { "epoch": 3.49, "learning_rate": 1.2536699727791918e-05, "loss": 0.0473, "step": 19012 }, { "epoch": 3.49, "learning_rate": 1.2535981575047544e-05, "loss": 0.0506, "step": 19013 }, { "epoch": 3.49, "learning_rate": 1.2535263408324774e-05, "loss": 0.0439, "step": 19014 }, { "epoch": 3.49, "learning_rate": 1.2534545227627578e-05, "loss": 0.0538, "step": 19015 }, { "epoch": 3.49, "learning_rate": 1.253382703295991e-05, "loss": 0.045, "step": 19016 }, { "epoch": 3.49, "learning_rate": 1.253310882432572e-05, "loss": 0.0563, "step": 19017 }, { "epoch": 3.49, "learning_rate": 1.2532390601728985e-05, "loss": 0.0529, "step": 19018 }, { "epoch": 3.49, "learning_rate": 1.2531672365173646e-05, "loss": 0.041, "step": 19019 }, { "epoch": 3.49, "learning_rate": 1.2530954114663673e-05, "loss": 0.05, "step": 19020 }, { "epoch": 3.49, "learning_rate": 1.2530235850203024e-05, "loss": 0.0447, "step": 19021 }, { "epoch": 3.49, "learning_rate": 1.2529517571795651e-05, "loss": 0.0462, "step": 19022 }, { "epoch": 3.49, "learning_rate": 1.2528799279445519e-05, "loss": 0.0556, "step": 19023 }, { "epoch": 3.49, "learning_rate": 1.2528080973156587e-05, "loss": 0.0412, "step": 19024 }, { "epoch": 3.49, "learning_rate": 1.2527362652932812e-05, "loss": 0.0456, "step": 19025 }, { "epoch": 3.49, "learning_rate": 1.2526644318778156e-05, "loss": 0.0585, "step": 19026 }, { "epoch": 3.49, "learning_rate": 1.252592597069658e-05, "loss": 0.0489, "step": 19027 }, { "epoch": 3.49, "learning_rate": 1.2525207608692039e-05, "loss": 0.0461, "step": 19028 }, { "epoch": 3.49, "learning_rate": 1.2524489232768491e-05, "loss": 0.0413, "step": 19029 }, { "epoch": 3.49, "learning_rate": 1.2523770842929903e-05, "loss": 0.0421, "step": 19030 }, { "epoch": 3.49, "learning_rate": 1.252305243918023e-05, "loss": 0.0522, "step": 19031 }, { "epoch": 3.49, "learning_rate": 1.2522334021523434e-05, "loss": 0.0445, "step": 19032 }, { "epoch": 3.49, "learning_rate": 1.2521615589963472e-05, "loss": 0.0424, "step": 19033 }, { "epoch": 3.49, "learning_rate": 1.2520897144504304e-05, "loss": 0.0473, "step": 19034 }, { "epoch": 3.49, "learning_rate": 1.2520178685149895e-05, "loss": 0.0497, "step": 19035 }, { "epoch": 3.49, "learning_rate": 1.2519460211904203e-05, "loss": 0.05, "step": 19036 }, { "epoch": 3.49, "learning_rate": 1.2518741724771185e-05, "loss": 0.0421, "step": 19037 }, { "epoch": 3.49, "learning_rate": 1.2518023223754806e-05, "loss": 0.0483, "step": 19038 }, { "epoch": 3.49, "learning_rate": 1.2517304708859021e-05, "loss": 0.0504, "step": 19039 }, { "epoch": 3.49, "learning_rate": 1.2516586180087795e-05, "loss": 0.0426, "step": 19040 }, { "epoch": 3.49, "learning_rate": 1.2515867637445088e-05, "loss": 0.0376, "step": 19041 }, { "epoch": 3.49, "learning_rate": 1.251514908093486e-05, "loss": 0.0368, "step": 19042 }, { "epoch": 3.49, "learning_rate": 1.2514430510561066e-05, "loss": 0.0499, "step": 19043 }, { "epoch": 3.49, "learning_rate": 1.2513711926327676e-05, "loss": 0.0424, "step": 19044 }, { "epoch": 3.49, "learning_rate": 1.2512993328238648e-05, "loss": 0.0445, "step": 19045 }, { "epoch": 3.49, "learning_rate": 1.2512274716297941e-05, "loss": 0.055, "step": 19046 }, { "epoch": 3.49, "learning_rate": 1.2511556090509516e-05, "loss": 0.0494, "step": 19047 }, { "epoch": 3.49, "learning_rate": 1.2510837450877335e-05, "loss": 0.0462, "step": 19048 }, { "epoch": 3.49, "learning_rate": 1.2510118797405361e-05, "loss": 0.0479, "step": 19049 }, { "epoch": 3.49, "learning_rate": 1.2509400130097551e-05, "loss": 0.0456, "step": 19050 }, { "epoch": 3.49, "learning_rate": 1.2508681448957871e-05, "loss": 0.0432, "step": 19051 }, { "epoch": 3.49, "learning_rate": 1.2507962753990277e-05, "loss": 0.0443, "step": 19052 }, { "epoch": 3.49, "learning_rate": 1.2507244045198737e-05, "loss": 0.0469, "step": 19053 }, { "epoch": 3.49, "learning_rate": 1.2506525322587207e-05, "loss": 0.0458, "step": 19054 }, { "epoch": 3.49, "learning_rate": 1.2505806586159652e-05, "loss": 0.0508, "step": 19055 }, { "epoch": 3.49, "learning_rate": 1.2505087835920032e-05, "loss": 0.0458, "step": 19056 }, { "epoch": 3.49, "learning_rate": 1.250436907187231e-05, "loss": 0.0449, "step": 19057 }, { "epoch": 3.49, "learning_rate": 1.2503650294020442e-05, "loss": 0.0449, "step": 19058 }, { "epoch": 3.5, "learning_rate": 1.25029315023684e-05, "loss": 0.0474, "step": 19059 }, { "epoch": 3.5, "learning_rate": 1.250221269692014e-05, "loss": 0.0475, "step": 19060 }, { "epoch": 3.5, "learning_rate": 1.2501493877679622e-05, "loss": 0.0449, "step": 19061 }, { "epoch": 3.5, "learning_rate": 1.2500775044650814e-05, "loss": 0.0417, "step": 19062 }, { "epoch": 3.5, "learning_rate": 1.2500056197837678e-05, "loss": 0.049, "step": 19063 }, { "epoch": 3.5, "learning_rate": 1.2499337337244171e-05, "loss": 0.0435, "step": 19064 }, { "epoch": 3.5, "learning_rate": 1.2498618462874256e-05, "loss": 0.0452, "step": 19065 }, { "epoch": 3.5, "learning_rate": 1.24978995747319e-05, "loss": 0.0386, "step": 19066 }, { "epoch": 3.5, "learning_rate": 1.249718067282106e-05, "loss": 0.0446, "step": 19067 }, { "epoch": 3.5, "learning_rate": 1.2496461757145707e-05, "loss": 0.0451, "step": 19068 }, { "epoch": 3.5, "learning_rate": 1.2495742827709794e-05, "loss": 0.0447, "step": 19069 }, { "epoch": 3.5, "learning_rate": 1.249502388451729e-05, "loss": 0.0483, "step": 19070 }, { "epoch": 3.5, "learning_rate": 1.2494304927572158e-05, "loss": 0.0553, "step": 19071 }, { "epoch": 3.5, "learning_rate": 1.2493585956878354e-05, "loss": 0.0504, "step": 19072 }, { "epoch": 3.5, "learning_rate": 1.2492866972439848e-05, "loss": 0.043, "step": 19073 }, { "epoch": 3.5, "learning_rate": 1.2492147974260602e-05, "loss": 0.0385, "step": 19074 }, { "epoch": 3.5, "learning_rate": 1.2491428962344576e-05, "loss": 0.0473, "step": 19075 }, { "epoch": 3.5, "learning_rate": 1.2490709936695735e-05, "loss": 0.047, "step": 19076 }, { "epoch": 3.5, "learning_rate": 1.2489990897318046e-05, "loss": 0.0438, "step": 19077 }, { "epoch": 3.5, "learning_rate": 1.2489271844215463e-05, "loss": 0.0456, "step": 19078 }, { "epoch": 3.5, "learning_rate": 1.2488552777391958e-05, "loss": 0.0498, "step": 19079 }, { "epoch": 3.5, "learning_rate": 1.2487833696851495e-05, "loss": 0.0531, "step": 19080 }, { "epoch": 3.5, "learning_rate": 1.2487114602598029e-05, "loss": 0.054, "step": 19081 }, { "epoch": 3.5, "learning_rate": 1.2486395494635533e-05, "loss": 0.0437, "step": 19082 }, { "epoch": 3.5, "learning_rate": 1.2485676372967966e-05, "loss": 0.0496, "step": 19083 }, { "epoch": 3.5, "learning_rate": 1.2484957237599292e-05, "loss": 0.0474, "step": 19084 }, { "epoch": 3.5, "learning_rate": 1.2484238088533473e-05, "loss": 0.0499, "step": 19085 }, { "epoch": 3.5, "learning_rate": 1.2483518925774481e-05, "loss": 0.0453, "step": 19086 }, { "epoch": 3.5, "learning_rate": 1.2482799749326271e-05, "loss": 0.0412, "step": 19087 }, { "epoch": 3.5, "learning_rate": 1.248208055919281e-05, "loss": 0.0515, "step": 19088 }, { "epoch": 3.5, "learning_rate": 1.2481361355378066e-05, "loss": 0.0388, "step": 19089 }, { "epoch": 3.5, "learning_rate": 1.2480642137885997e-05, "loss": 0.046, "step": 19090 }, { "epoch": 3.5, "learning_rate": 1.2479922906720571e-05, "loss": 0.0459, "step": 19091 }, { "epoch": 3.5, "learning_rate": 1.2479203661885752e-05, "loss": 0.0577, "step": 19092 }, { "epoch": 3.5, "learning_rate": 1.2478484403385506e-05, "loss": 0.0414, "step": 19093 }, { "epoch": 3.5, "learning_rate": 1.2477765131223793e-05, "loss": 0.0454, "step": 19094 }, { "epoch": 3.5, "learning_rate": 1.2477045845404585e-05, "loss": 0.0468, "step": 19095 }, { "epoch": 3.5, "learning_rate": 1.2476326545931836e-05, "loss": 0.0526, "step": 19096 }, { "epoch": 3.5, "learning_rate": 1.2475607232809522e-05, "loss": 0.0538, "step": 19097 }, { "epoch": 3.5, "learning_rate": 1.2474887906041602e-05, "loss": 0.0521, "step": 19098 }, { "epoch": 3.5, "learning_rate": 1.2474168565632042e-05, "loss": 0.0478, "step": 19099 }, { "epoch": 3.5, "learning_rate": 1.2473449211584802e-05, "loss": 0.0482, "step": 19100 }, { "epoch": 3.5, "learning_rate": 1.247272984390386e-05, "loss": 0.04, "step": 19101 }, { "epoch": 3.5, "learning_rate": 1.2472010462593169e-05, "loss": 0.0448, "step": 19102 }, { "epoch": 3.5, "learning_rate": 1.2471291067656696e-05, "loss": 0.0538, "step": 19103 }, { "epoch": 3.5, "learning_rate": 1.2470571659098415e-05, "loss": 0.0511, "step": 19104 }, { "epoch": 3.5, "learning_rate": 1.2469852236922279e-05, "loss": 0.0467, "step": 19105 }, { "epoch": 3.5, "learning_rate": 1.2469132801132262e-05, "loss": 0.0429, "step": 19106 }, { "epoch": 3.5, "learning_rate": 1.2468413351732328e-05, "loss": 0.0494, "step": 19107 }, { "epoch": 3.5, "learning_rate": 1.2467693888726442e-05, "loss": 0.0469, "step": 19108 }, { "epoch": 3.5, "learning_rate": 1.246697441211857e-05, "loss": 0.045, "step": 19109 }, { "epoch": 3.5, "learning_rate": 1.2466254921912674e-05, "loss": 0.0507, "step": 19110 }, { "epoch": 3.5, "learning_rate": 1.2465535418112724e-05, "loss": 0.0504, "step": 19111 }, { "epoch": 3.5, "learning_rate": 1.2464815900722687e-05, "loss": 0.0511, "step": 19112 }, { "epoch": 3.51, "learning_rate": 1.2464096369746528e-05, "loss": 0.0442, "step": 19113 }, { "epoch": 3.51, "learning_rate": 1.2463376825188207e-05, "loss": 0.04, "step": 19114 }, { "epoch": 3.51, "learning_rate": 1.2462657267051698e-05, "loss": 0.045, "step": 19115 }, { "epoch": 3.51, "learning_rate": 1.2461937695340967e-05, "loss": 0.0537, "step": 19116 }, { "epoch": 3.51, "learning_rate": 1.2461218110059974e-05, "loss": 0.0511, "step": 19117 }, { "epoch": 3.51, "learning_rate": 1.2460498511212691e-05, "loss": 0.0469, "step": 19118 }, { "epoch": 3.51, "learning_rate": 1.2459778898803083e-05, "loss": 0.051, "step": 19119 }, { "epoch": 3.51, "learning_rate": 1.2459059272835115e-05, "loss": 0.0547, "step": 19120 }, { "epoch": 3.51, "learning_rate": 1.2458339633312752e-05, "loss": 0.0419, "step": 19121 }, { "epoch": 3.51, "learning_rate": 1.2457619980239967e-05, "loss": 0.0481, "step": 19122 }, { "epoch": 3.51, "learning_rate": 1.2456900313620723e-05, "loss": 0.0465, "step": 19123 }, { "epoch": 3.51, "learning_rate": 1.2456180633458987e-05, "loss": 0.0503, "step": 19124 }, { "epoch": 3.51, "learning_rate": 1.2455460939758725e-05, "loss": 0.0504, "step": 19125 }, { "epoch": 3.51, "learning_rate": 1.2454741232523905e-05, "loss": 0.0526, "step": 19126 }, { "epoch": 3.51, "learning_rate": 1.2454021511758495e-05, "loss": 0.0457, "step": 19127 }, { "epoch": 3.51, "learning_rate": 1.245330177746646e-05, "loss": 0.0416, "step": 19128 }, { "epoch": 3.51, "learning_rate": 1.2452582029651765e-05, "loss": 0.044, "step": 19129 }, { "epoch": 3.51, "learning_rate": 1.2451862268318384e-05, "loss": 0.0407, "step": 19130 }, { "epoch": 3.51, "learning_rate": 1.245114249347028e-05, "loss": 0.052, "step": 19131 }, { "epoch": 3.51, "learning_rate": 1.245042270511142e-05, "loss": 0.0452, "step": 19132 }, { "epoch": 3.51, "learning_rate": 1.2449702903245775e-05, "loss": 0.0462, "step": 19133 }, { "epoch": 3.51, "learning_rate": 1.244898308787731e-05, "loss": 0.0547, "step": 19134 }, { "epoch": 3.51, "learning_rate": 1.2448263259009991e-05, "loss": 0.0438, "step": 19135 }, { "epoch": 3.51, "learning_rate": 1.2447543416647789e-05, "loss": 0.055, "step": 19136 }, { "epoch": 3.51, "learning_rate": 1.244682356079467e-05, "loss": 0.0416, "step": 19137 }, { "epoch": 3.51, "learning_rate": 1.2446103691454598e-05, "loss": 0.0535, "step": 19138 }, { "epoch": 3.51, "learning_rate": 1.2445383808631551e-05, "loss": 0.0429, "step": 19139 }, { "epoch": 3.51, "learning_rate": 1.2444663912329489e-05, "loss": 0.0448, "step": 19140 }, { "epoch": 3.51, "learning_rate": 1.2443944002552381e-05, "loss": 0.0535, "step": 19141 }, { "epoch": 3.51, "learning_rate": 1.2443224079304199e-05, "loss": 0.0469, "step": 19142 }, { "epoch": 3.51, "learning_rate": 1.2442504142588909e-05, "loss": 0.0426, "step": 19143 }, { "epoch": 3.51, "learning_rate": 1.2441784192410478e-05, "loss": 0.0475, "step": 19144 }, { "epoch": 3.51, "learning_rate": 1.2441064228772874e-05, "loss": 0.0422, "step": 19145 }, { "epoch": 3.51, "learning_rate": 1.2440344251680067e-05, "loss": 0.0444, "step": 19146 }, { "epoch": 3.51, "learning_rate": 1.2439624261136025e-05, "loss": 0.0425, "step": 19147 }, { "epoch": 3.51, "learning_rate": 1.243890425714472e-05, "loss": 0.051, "step": 19148 }, { "epoch": 3.51, "learning_rate": 1.2438184239710114e-05, "loss": 0.0465, "step": 19149 }, { "epoch": 3.51, "learning_rate": 1.243746420883618e-05, "loss": 0.0552, "step": 19150 }, { "epoch": 3.51, "learning_rate": 1.2436744164526886e-05, "loss": 0.0346, "step": 19151 }, { "epoch": 3.51, "learning_rate": 1.2436024106786202e-05, "loss": 0.0473, "step": 19152 }, { "epoch": 3.51, "learning_rate": 1.2435304035618095e-05, "loss": 0.0484, "step": 19153 }, { "epoch": 3.51, "learning_rate": 1.2434583951026538e-05, "loss": 0.0468, "step": 19154 }, { "epoch": 3.51, "learning_rate": 1.2433863853015495e-05, "loss": 0.0451, "step": 19155 }, { "epoch": 3.51, "learning_rate": 1.2433143741588938e-05, "loss": 0.0478, "step": 19156 }, { "epoch": 3.51, "learning_rate": 1.2432423616750838e-05, "loss": 0.0431, "step": 19157 }, { "epoch": 3.51, "learning_rate": 1.2431703478505158e-05, "loss": 0.0451, "step": 19158 }, { "epoch": 3.51, "learning_rate": 1.2430983326855873e-05, "loss": 0.0518, "step": 19159 }, { "epoch": 3.51, "learning_rate": 1.243026316180695e-05, "loss": 0.055, "step": 19160 }, { "epoch": 3.51, "learning_rate": 1.242954298336236e-05, "loss": 0.0422, "step": 19161 }, { "epoch": 3.51, "learning_rate": 1.2428822791526072e-05, "loss": 0.046, "step": 19162 }, { "epoch": 3.51, "learning_rate": 1.2428102586302058e-05, "loss": 0.0448, "step": 19163 }, { "epoch": 3.51, "learning_rate": 1.2427382367694283e-05, "loss": 0.0461, "step": 19164 }, { "epoch": 3.51, "learning_rate": 1.242666213570672e-05, "loss": 0.0405, "step": 19165 }, { "epoch": 3.51, "learning_rate": 1.242594189034334e-05, "loss": 0.0439, "step": 19166 }, { "epoch": 3.51, "learning_rate": 1.2425221631608108e-05, "loss": 0.0526, "step": 19167 }, { "epoch": 3.52, "learning_rate": 1.2424501359505002e-05, "loss": 0.0436, "step": 19168 }, { "epoch": 3.52, "learning_rate": 1.2423781074037985e-05, "loss": 0.0412, "step": 19169 }, { "epoch": 3.52, "learning_rate": 1.2423060775211032e-05, "loss": 0.0489, "step": 19170 }, { "epoch": 3.52, "learning_rate": 1.2422340463028108e-05, "loss": 0.0477, "step": 19171 }, { "epoch": 3.52, "learning_rate": 1.2421620137493191e-05, "loss": 0.0531, "step": 19172 }, { "epoch": 3.52, "learning_rate": 1.2420899798610243e-05, "loss": 0.047, "step": 19173 }, { "epoch": 3.52, "learning_rate": 1.242017944638324e-05, "loss": 0.0442, "step": 19174 }, { "epoch": 3.52, "learning_rate": 1.2419459080816152e-05, "loss": 0.0473, "step": 19175 }, { "epoch": 3.52, "learning_rate": 1.241873870191295e-05, "loss": 0.0426, "step": 19176 }, { "epoch": 3.52, "learning_rate": 1.24180183096776e-05, "loss": 0.0556, "step": 19177 }, { "epoch": 3.52, "learning_rate": 1.2417297904114078e-05, "loss": 0.0468, "step": 19178 }, { "epoch": 3.52, "learning_rate": 1.2416577485226355e-05, "loss": 0.0419, "step": 19179 }, { "epoch": 3.52, "learning_rate": 1.2415857053018396e-05, "loss": 0.0404, "step": 19180 }, { "epoch": 3.52, "learning_rate": 1.2415136607494182e-05, "loss": 0.0442, "step": 19181 }, { "epoch": 3.52, "learning_rate": 1.2414416148657675e-05, "loss": 0.0556, "step": 19182 }, { "epoch": 3.52, "learning_rate": 1.2413695676512849e-05, "loss": 0.0462, "step": 19183 }, { "epoch": 3.52, "learning_rate": 1.2412975191063678e-05, "loss": 0.0482, "step": 19184 }, { "epoch": 3.52, "learning_rate": 1.2412254692314132e-05, "loss": 0.0488, "step": 19185 }, { "epoch": 3.52, "learning_rate": 1.2411534180268179e-05, "loss": 0.0425, "step": 19186 }, { "epoch": 3.52, "learning_rate": 1.2410813654929794e-05, "loss": 0.0414, "step": 19187 }, { "epoch": 3.52, "learning_rate": 1.2410093116302947e-05, "loss": 0.0449, "step": 19188 }, { "epoch": 3.52, "learning_rate": 1.240937256439161e-05, "loss": 0.0469, "step": 19189 }, { "epoch": 3.52, "learning_rate": 1.240865199919976e-05, "loss": 0.0455, "step": 19190 }, { "epoch": 3.52, "learning_rate": 1.2407931420731356e-05, "loss": 0.0483, "step": 19191 }, { "epoch": 3.52, "learning_rate": 1.2407210828990384e-05, "loss": 0.0536, "step": 19192 }, { "epoch": 3.52, "learning_rate": 1.2406490223980807e-05, "loss": 0.0495, "step": 19193 }, { "epoch": 3.52, "learning_rate": 1.2405769605706599e-05, "loss": 0.0474, "step": 19194 }, { "epoch": 3.52, "learning_rate": 1.2405048974171734e-05, "loss": 0.0392, "step": 19195 }, { "epoch": 3.52, "learning_rate": 1.2404328329380182e-05, "loss": 0.0433, "step": 19196 }, { "epoch": 3.52, "learning_rate": 1.2403607671335913e-05, "loss": 0.0448, "step": 19197 }, { "epoch": 3.52, "learning_rate": 1.2402887000042904e-05, "loss": 0.0414, "step": 19198 }, { "epoch": 3.52, "learning_rate": 1.2402166315505131e-05, "loss": 0.0597, "step": 19199 }, { "epoch": 3.52, "learning_rate": 1.2401445617726554e-05, "loss": 0.0426, "step": 19200 }, { "epoch": 3.52, "learning_rate": 1.2400724906711155e-05, "loss": 0.0472, "step": 19201 }, { "epoch": 3.52, "learning_rate": 1.2400004182462904e-05, "loss": 0.047, "step": 19202 }, { "epoch": 3.52, "learning_rate": 1.2399283444985772e-05, "loss": 0.0536, "step": 19203 }, { "epoch": 3.52, "learning_rate": 1.2398562694283738e-05, "loss": 0.0463, "step": 19204 }, { "epoch": 3.52, "learning_rate": 1.2397841930360766e-05, "loss": 0.0472, "step": 19205 }, { "epoch": 3.52, "learning_rate": 1.2397121153220833e-05, "loss": 0.041, "step": 19206 }, { "epoch": 3.52, "learning_rate": 1.2396400362867914e-05, "loss": 0.0422, "step": 19207 }, { "epoch": 3.52, "learning_rate": 1.2395679559305982e-05, "loss": 0.0428, "step": 19208 }, { "epoch": 3.52, "learning_rate": 1.2394958742539003e-05, "loss": 0.0485, "step": 19209 }, { "epoch": 3.52, "learning_rate": 1.239423791257096e-05, "loss": 0.0415, "step": 19210 }, { "epoch": 3.52, "learning_rate": 1.2393517069405818e-05, "loss": 0.0534, "step": 19211 }, { "epoch": 3.52, "learning_rate": 1.2392796213047555e-05, "loss": 0.0477, "step": 19212 }, { "epoch": 3.52, "learning_rate": 1.2392075343500145e-05, "loss": 0.0382, "step": 19213 }, { "epoch": 3.52, "learning_rate": 1.239135446076756e-05, "loss": 0.0535, "step": 19214 }, { "epoch": 3.52, "learning_rate": 1.2390633564853768e-05, "loss": 0.0475, "step": 19215 }, { "epoch": 3.52, "learning_rate": 1.2389912655762753e-05, "loss": 0.0453, "step": 19216 }, { "epoch": 3.52, "learning_rate": 1.2389191733498481e-05, "loss": 0.0429, "step": 19217 }, { "epoch": 3.52, "learning_rate": 1.2388470798064929e-05, "loss": 0.0504, "step": 19218 }, { "epoch": 3.52, "learning_rate": 1.2387749849466069e-05, "loss": 0.0459, "step": 19219 }, { "epoch": 3.52, "learning_rate": 1.2387028887705879e-05, "loss": 0.0403, "step": 19220 }, { "epoch": 3.52, "learning_rate": 1.2386307912788326e-05, "loss": 0.05, "step": 19221 }, { "epoch": 3.53, "learning_rate": 1.238558692471739e-05, "loss": 0.0472, "step": 19222 }, { "epoch": 3.53, "learning_rate": 1.2384865923497042e-05, "loss": 0.0524, "step": 19223 }, { "epoch": 3.53, "learning_rate": 1.2384144909131258e-05, "loss": 0.0463, "step": 19224 }, { "epoch": 3.53, "learning_rate": 1.2383423881624013e-05, "loss": 0.0451, "step": 19225 }, { "epoch": 3.53, "learning_rate": 1.2382702840979275e-05, "loss": 0.0481, "step": 19226 }, { "epoch": 3.53, "learning_rate": 1.2381981787201027e-05, "loss": 0.0437, "step": 19227 }, { "epoch": 3.53, "learning_rate": 1.2381260720293237e-05, "loss": 0.042, "step": 19228 }, { "epoch": 3.53, "learning_rate": 1.2380539640259885e-05, "loss": 0.044, "step": 19229 }, { "epoch": 3.53, "learning_rate": 1.2379818547104942e-05, "loss": 0.0449, "step": 19230 }, { "epoch": 3.53, "learning_rate": 1.2379097440832383e-05, "loss": 0.0492, "step": 19231 }, { "epoch": 3.53, "learning_rate": 1.2378376321446184e-05, "loss": 0.0493, "step": 19232 }, { "epoch": 3.53, "learning_rate": 1.2377655188950315e-05, "loss": 0.045, "step": 19233 }, { "epoch": 3.53, "learning_rate": 1.237693404334876e-05, "loss": 0.0431, "step": 19234 }, { "epoch": 3.53, "learning_rate": 1.2376212884645484e-05, "loss": 0.051, "step": 19235 }, { "epoch": 3.53, "learning_rate": 1.2375491712844472e-05, "loss": 0.0429, "step": 19236 }, { "epoch": 3.53, "learning_rate": 1.2374770527949692e-05, "loss": 0.0431, "step": 19237 }, { "epoch": 3.53, "learning_rate": 1.2374049329965118e-05, "loss": 0.0472, "step": 19238 }, { "epoch": 3.53, "learning_rate": 1.2373328118894732e-05, "loss": 0.047, "step": 19239 }, { "epoch": 3.53, "learning_rate": 1.2372606894742507e-05, "loss": 0.0512, "step": 19240 }, { "epoch": 3.53, "learning_rate": 1.2371885657512415e-05, "loss": 0.0423, "step": 19241 }, { "epoch": 3.53, "learning_rate": 1.2371164407208434e-05, "loss": 0.0444, "step": 19242 }, { "epoch": 3.53, "learning_rate": 1.2370443143834541e-05, "loss": 0.0451, "step": 19243 }, { "epoch": 3.53, "learning_rate": 1.2369721867394706e-05, "loss": 0.045, "step": 19244 }, { "epoch": 3.53, "learning_rate": 1.2369000577892912e-05, "loss": 0.0393, "step": 19245 }, { "epoch": 3.53, "learning_rate": 1.2368279275333129e-05, "loss": 0.0371, "step": 19246 }, { "epoch": 3.53, "learning_rate": 1.2367557959719336e-05, "loss": 0.0417, "step": 19247 }, { "epoch": 3.53, "learning_rate": 1.2366836631055506e-05, "loss": 0.0449, "step": 19248 }, { "epoch": 3.53, "learning_rate": 1.236611528934562e-05, "loss": 0.0465, "step": 19249 }, { "epoch": 3.53, "learning_rate": 1.236539393459365e-05, "loss": 0.0482, "step": 19250 }, { "epoch": 3.53, "learning_rate": 1.2364672566803573e-05, "loss": 0.0397, "step": 19251 }, { "epoch": 3.53, "learning_rate": 1.2363951185979368e-05, "loss": 0.0461, "step": 19252 }, { "epoch": 3.53, "learning_rate": 1.2363229792125007e-05, "loss": 0.0519, "step": 19253 }, { "epoch": 3.53, "learning_rate": 1.2362508385244466e-05, "loss": 0.0446, "step": 19254 }, { "epoch": 3.53, "learning_rate": 1.2361786965341726e-05, "loss": 0.0478, "step": 19255 }, { "epoch": 3.53, "learning_rate": 1.2361065532420762e-05, "loss": 0.0463, "step": 19256 }, { "epoch": 3.53, "learning_rate": 1.2360344086485545e-05, "loss": 0.041, "step": 19257 }, { "epoch": 3.53, "learning_rate": 1.2359622627540059e-05, "loss": 0.0377, "step": 19258 }, { "epoch": 3.53, "learning_rate": 1.2358901155588276e-05, "loss": 0.0525, "step": 19259 }, { "epoch": 3.53, "learning_rate": 1.2358179670634176e-05, "loss": 0.0516, "step": 19260 }, { "epoch": 3.53, "learning_rate": 1.2357458172681735e-05, "loss": 0.0446, "step": 19261 }, { "epoch": 3.53, "learning_rate": 1.2356736661734929e-05, "loss": 0.0412, "step": 19262 }, { "epoch": 3.53, "learning_rate": 1.2356015137797735e-05, "loss": 0.0502, "step": 19263 }, { "epoch": 3.53, "learning_rate": 1.2355293600874132e-05, "loss": 0.0413, "step": 19264 }, { "epoch": 3.53, "learning_rate": 1.2354572050968092e-05, "loss": 0.0442, "step": 19265 }, { "epoch": 3.53, "learning_rate": 1.2353850488083597e-05, "loss": 0.0469, "step": 19266 }, { "epoch": 3.53, "learning_rate": 1.2353128912224625e-05, "loss": 0.049, "step": 19267 }, { "epoch": 3.53, "learning_rate": 1.2352407323395148e-05, "loss": 0.0424, "step": 19268 }, { "epoch": 3.53, "learning_rate": 1.235168572159915e-05, "loss": 0.0468, "step": 19269 }, { "epoch": 3.53, "learning_rate": 1.2350964106840603e-05, "loss": 0.0386, "step": 19270 }, { "epoch": 3.53, "learning_rate": 1.235024247912349e-05, "loss": 0.0427, "step": 19271 }, { "epoch": 3.53, "learning_rate": 1.234952083845178e-05, "loss": 0.0484, "step": 19272 }, { "epoch": 3.53, "learning_rate": 1.234879918482946e-05, "loss": 0.0433, "step": 19273 }, { "epoch": 3.53, "learning_rate": 1.2348077518260504e-05, "loss": 0.0427, "step": 19274 }, { "epoch": 3.53, "learning_rate": 1.2347355838748887e-05, "loss": 0.0501, "step": 19275 }, { "epoch": 3.53, "learning_rate": 1.2346634146298593e-05, "loss": 0.0441, "step": 19276 }, { "epoch": 3.54, "learning_rate": 1.2345912440913593e-05, "loss": 0.0535, "step": 19277 }, { "epoch": 3.54, "learning_rate": 1.234519072259787e-05, "loss": 0.047, "step": 19278 }, { "epoch": 3.54, "learning_rate": 1.2344468991355402e-05, "loss": 0.0484, "step": 19279 }, { "epoch": 3.54, "learning_rate": 1.2343747247190166e-05, "loss": 0.0491, "step": 19280 }, { "epoch": 3.54, "learning_rate": 1.234302549010614e-05, "loss": 0.0426, "step": 19281 }, { "epoch": 3.54, "learning_rate": 1.2342303720107303e-05, "loss": 0.0442, "step": 19282 }, { "epoch": 3.54, "learning_rate": 1.2341581937197631e-05, "loss": 0.0452, "step": 19283 }, { "epoch": 3.54, "learning_rate": 1.2340860141381103e-05, "loss": 0.0531, "step": 19284 }, { "epoch": 3.54, "learning_rate": 1.2340138332661703e-05, "loss": 0.0521, "step": 19285 }, { "epoch": 3.54, "learning_rate": 1.2339416511043405e-05, "loss": 0.0484, "step": 19286 }, { "epoch": 3.54, "learning_rate": 1.2338694676530187e-05, "loss": 0.049, "step": 19287 }, { "epoch": 3.54, "learning_rate": 1.233797282912603e-05, "loss": 0.0535, "step": 19288 }, { "epoch": 3.54, "learning_rate": 1.2337250968834912e-05, "loss": 0.0466, "step": 19289 }, { "epoch": 3.54, "learning_rate": 1.2336529095660813e-05, "loss": 0.0471, "step": 19290 }, { "epoch": 3.54, "learning_rate": 1.233580720960771e-05, "loss": 0.0489, "step": 19291 }, { "epoch": 3.54, "learning_rate": 1.2335085310679581e-05, "loss": 0.0475, "step": 19292 }, { "epoch": 3.54, "learning_rate": 1.2334363398880409e-05, "loss": 0.0442, "step": 19293 }, { "epoch": 3.54, "learning_rate": 1.2333641474214173e-05, "loss": 0.0408, "step": 19294 }, { "epoch": 3.54, "learning_rate": 1.2332919536684846e-05, "loss": 0.0508, "step": 19295 }, { "epoch": 3.54, "learning_rate": 1.2332197586296415e-05, "loss": 0.0437, "step": 19296 }, { "epoch": 3.54, "learning_rate": 1.2331475623052855e-05, "loss": 0.0459, "step": 19297 }, { "epoch": 3.54, "learning_rate": 1.233075364695815e-05, "loss": 0.0449, "step": 19298 }, { "epoch": 3.54, "learning_rate": 1.2330031658016273e-05, "loss": 0.0401, "step": 19299 }, { "epoch": 3.54, "learning_rate": 1.2329309656231207e-05, "loss": 0.0351, "step": 19300 }, { "epoch": 3.54, "learning_rate": 1.232858764160693e-05, "loss": 0.0494, "step": 19301 }, { "epoch": 3.54, "learning_rate": 1.2327865614147429e-05, "loss": 0.0444, "step": 19302 }, { "epoch": 3.54, "learning_rate": 1.2327143573856672e-05, "loss": 0.0516, "step": 19303 }, { "epoch": 3.54, "learning_rate": 1.232642152073865e-05, "loss": 0.0477, "step": 19304 }, { "epoch": 3.54, "learning_rate": 1.2325699454797334e-05, "loss": 0.0487, "step": 19305 }, { "epoch": 3.54, "learning_rate": 1.2324977376036711e-05, "loss": 0.0393, "step": 19306 }, { "epoch": 3.54, "learning_rate": 1.2324255284460757e-05, "loss": 0.0459, "step": 19307 }, { "epoch": 3.54, "learning_rate": 1.2323533180073455e-05, "loss": 0.048, "step": 19308 }, { "epoch": 3.54, "learning_rate": 1.2322811062878783e-05, "loss": 0.0438, "step": 19309 }, { "epoch": 3.54, "learning_rate": 1.2322088932880718e-05, "loss": 0.0518, "step": 19310 }, { "epoch": 3.54, "learning_rate": 1.2321366790083252e-05, "loss": 0.0471, "step": 19311 }, { "epoch": 3.54, "learning_rate": 1.232064463449035e-05, "loss": 0.0511, "step": 19312 }, { "epoch": 3.54, "learning_rate": 1.2319922466106004e-05, "loss": 0.0388, "step": 19313 }, { "epoch": 3.54, "learning_rate": 1.2319200284934191e-05, "loss": 0.049, "step": 19314 }, { "epoch": 3.54, "learning_rate": 1.2318478090978893e-05, "loss": 0.0498, "step": 19315 }, { "epoch": 3.54, "learning_rate": 1.231775588424409e-05, "loss": 0.0421, "step": 19316 }, { "epoch": 3.54, "learning_rate": 1.231703366473376e-05, "loss": 0.0488, "step": 19317 }, { "epoch": 3.54, "learning_rate": 1.2316311432451886e-05, "loss": 0.0435, "step": 19318 }, { "epoch": 3.54, "learning_rate": 1.2315589187402448e-05, "loss": 0.0424, "step": 19319 }, { "epoch": 3.54, "learning_rate": 1.2314866929589434e-05, "loss": 0.0482, "step": 19320 }, { "epoch": 3.54, "learning_rate": 1.2314144659016812e-05, "loss": 0.0524, "step": 19321 }, { "epoch": 3.54, "learning_rate": 1.2313422375688573e-05, "loss": 0.05, "step": 19322 }, { "epoch": 3.54, "learning_rate": 1.2312700079608696e-05, "loss": 0.0424, "step": 19323 }, { "epoch": 3.54, "learning_rate": 1.2311977770781165e-05, "loss": 0.0407, "step": 19324 }, { "epoch": 3.54, "learning_rate": 1.2311255449209953e-05, "loss": 0.0431, "step": 19325 }, { "epoch": 3.54, "learning_rate": 1.2310533114899052e-05, "loss": 0.0468, "step": 19326 }, { "epoch": 3.54, "learning_rate": 1.2309810767852435e-05, "loss": 0.051, "step": 19327 }, { "epoch": 3.54, "learning_rate": 1.2309088408074086e-05, "loss": 0.0406, "step": 19328 }, { "epoch": 3.54, "learning_rate": 1.2308366035567992e-05, "loss": 0.0481, "step": 19329 }, { "epoch": 3.54, "learning_rate": 1.2307643650338126e-05, "loss": 0.0408, "step": 19330 }, { "epoch": 3.55, "learning_rate": 1.2306921252388476e-05, "loss": 0.0432, "step": 19331 }, { "epoch": 3.55, "learning_rate": 1.2306198841723023e-05, "loss": 0.0475, "step": 19332 }, { "epoch": 3.55, "learning_rate": 1.2305476418345746e-05, "loss": 0.0427, "step": 19333 }, { "epoch": 3.55, "learning_rate": 1.2304753982260627e-05, "loss": 0.0437, "step": 19334 }, { "epoch": 3.55, "learning_rate": 1.2304031533471656e-05, "loss": 0.0509, "step": 19335 }, { "epoch": 3.55, "learning_rate": 1.2303309071982806e-05, "loss": 0.0467, "step": 19336 }, { "epoch": 3.55, "learning_rate": 1.2302586597798061e-05, "loss": 0.0482, "step": 19337 }, { "epoch": 3.55, "learning_rate": 1.2301864110921409e-05, "loss": 0.0421, "step": 19338 }, { "epoch": 3.55, "learning_rate": 1.2301141611356821e-05, "loss": 0.045, "step": 19339 }, { "epoch": 3.55, "learning_rate": 1.2300419099108291e-05, "loss": 0.0433, "step": 19340 }, { "epoch": 3.55, "learning_rate": 1.2299696574179799e-05, "loss": 0.0541, "step": 19341 }, { "epoch": 3.55, "learning_rate": 1.2298974036575324e-05, "loss": 0.0491, "step": 19342 }, { "epoch": 3.55, "learning_rate": 1.2298251486298847e-05, "loss": 0.0442, "step": 19343 }, { "epoch": 3.55, "learning_rate": 1.2297528923354358e-05, "loss": 0.0443, "step": 19344 }, { "epoch": 3.55, "learning_rate": 1.2296806347745833e-05, "loss": 0.0442, "step": 19345 }, { "epoch": 3.55, "learning_rate": 1.229608375947726e-05, "loss": 0.0503, "step": 19346 }, { "epoch": 3.55, "learning_rate": 1.229536115855262e-05, "loss": 0.044, "step": 19347 }, { "epoch": 3.55, "learning_rate": 1.2294638544975892e-05, "loss": 0.0424, "step": 19348 }, { "epoch": 3.55, "learning_rate": 1.2293915918751065e-05, "loss": 0.0493, "step": 19349 }, { "epoch": 3.55, "learning_rate": 1.229319327988212e-05, "loss": 0.0525, "step": 19350 }, { "epoch": 3.55, "learning_rate": 1.229247062837304e-05, "loss": 0.0467, "step": 19351 }, { "epoch": 3.55, "learning_rate": 1.2291747964227806e-05, "loss": 0.0498, "step": 19352 }, { "epoch": 3.55, "learning_rate": 1.2291025287450407e-05, "loss": 0.0444, "step": 19353 }, { "epoch": 3.55, "learning_rate": 1.229030259804482e-05, "loss": 0.0446, "step": 19354 }, { "epoch": 3.55, "learning_rate": 1.2289579896015032e-05, "loss": 0.0435, "step": 19355 }, { "epoch": 3.55, "learning_rate": 1.2288857181365025e-05, "loss": 0.0497, "step": 19356 }, { "epoch": 3.55, "learning_rate": 1.2288134454098786e-05, "loss": 0.0424, "step": 19357 }, { "epoch": 3.55, "learning_rate": 1.2287411714220294e-05, "loss": 0.0452, "step": 19358 }, { "epoch": 3.55, "learning_rate": 1.2286688961733537e-05, "loss": 0.0341, "step": 19359 }, { "epoch": 3.55, "learning_rate": 1.2285966196642496e-05, "loss": 0.045, "step": 19360 }, { "epoch": 3.55, "learning_rate": 1.2285243418951154e-05, "loss": 0.0445, "step": 19361 }, { "epoch": 3.55, "learning_rate": 1.2284520628663502e-05, "loss": 0.045, "step": 19362 }, { "epoch": 3.55, "learning_rate": 1.2283797825783512e-05, "loss": 0.042, "step": 19363 }, { "epoch": 3.55, "learning_rate": 1.228307501031518e-05, "loss": 0.055, "step": 19364 }, { "epoch": 3.55, "learning_rate": 1.2282352182262483e-05, "loss": 0.0461, "step": 19365 }, { "epoch": 3.55, "learning_rate": 1.2281629341629408e-05, "loss": 0.0374, "step": 19366 }, { "epoch": 3.55, "learning_rate": 1.2280906488419939e-05, "loss": 0.0493, "step": 19367 }, { "epoch": 3.55, "learning_rate": 1.2280183622638058e-05, "loss": 0.0377, "step": 19368 }, { "epoch": 3.55, "learning_rate": 1.2279460744287755e-05, "loss": 0.0481, "step": 19369 }, { "epoch": 3.55, "learning_rate": 1.2278737853373007e-05, "loss": 0.0405, "step": 19370 }, { "epoch": 3.55, "learning_rate": 1.2278014949897806e-05, "loss": 0.0432, "step": 19371 }, { "epoch": 3.55, "learning_rate": 1.2277292033866129e-05, "loss": 0.0444, "step": 19372 }, { "epoch": 3.55, "learning_rate": 1.2276569105281967e-05, "loss": 0.0461, "step": 19373 }, { "epoch": 3.55, "learning_rate": 1.2275846164149303e-05, "loss": 0.046, "step": 19374 }, { "epoch": 3.55, "learning_rate": 1.2275123210472122e-05, "loss": 0.0476, "step": 19375 }, { "epoch": 3.55, "learning_rate": 1.2274400244254409e-05, "loss": 0.0464, "step": 19376 }, { "epoch": 3.55, "learning_rate": 1.2273677265500147e-05, "loss": 0.047, "step": 19377 }, { "epoch": 3.55, "learning_rate": 1.2272954274213322e-05, "loss": 0.0535, "step": 19378 }, { "epoch": 3.55, "learning_rate": 1.2272231270397925e-05, "loss": 0.0439, "step": 19379 }, { "epoch": 3.55, "learning_rate": 1.2271508254057933e-05, "loss": 0.0405, "step": 19380 }, { "epoch": 3.55, "learning_rate": 1.2270785225197331e-05, "loss": 0.0457, "step": 19381 }, { "epoch": 3.55, "learning_rate": 1.2270062183820111e-05, "loss": 0.0468, "step": 19382 }, { "epoch": 3.55, "learning_rate": 1.2269339129930256e-05, "loss": 0.0457, "step": 19383 }, { "epoch": 3.55, "learning_rate": 1.2268616063531749e-05, "loss": 0.0441, "step": 19384 }, { "epoch": 3.55, "learning_rate": 1.2267892984628575e-05, "loss": 0.0457, "step": 19385 }, { "epoch": 3.56, "learning_rate": 1.2267169893224724e-05, "loss": 0.0512, "step": 19386 }, { "epoch": 3.56, "learning_rate": 1.2266446789324178e-05, "loss": 0.044, "step": 19387 }, { "epoch": 3.56, "learning_rate": 1.2265723672930927e-05, "loss": 0.0489, "step": 19388 }, { "epoch": 3.56, "learning_rate": 1.226500054404895e-05, "loss": 0.0458, "step": 19389 }, { "epoch": 3.56, "learning_rate": 1.226427740268224e-05, "loss": 0.0425, "step": 19390 }, { "epoch": 3.56, "learning_rate": 1.2263554248834781e-05, "loss": 0.0426, "step": 19391 }, { "epoch": 3.56, "learning_rate": 1.2262831082510555e-05, "loss": 0.0411, "step": 19392 }, { "epoch": 3.56, "learning_rate": 1.2262107903713554e-05, "loss": 0.0442, "step": 19393 }, { "epoch": 3.56, "learning_rate": 1.2261384712447759e-05, "loss": 0.0525, "step": 19394 }, { "epoch": 3.56, "learning_rate": 1.226066150871716e-05, "loss": 0.0481, "step": 19395 }, { "epoch": 3.56, "learning_rate": 1.2259938292525737e-05, "loss": 0.0533, "step": 19396 }, { "epoch": 3.56, "learning_rate": 1.2259215063877488e-05, "loss": 0.0441, "step": 19397 }, { "epoch": 3.56, "learning_rate": 1.2258491822776389e-05, "loss": 0.0486, "step": 19398 }, { "epoch": 3.56, "learning_rate": 1.225776856922643e-05, "loss": 0.0385, "step": 19399 }, { "epoch": 3.56, "learning_rate": 1.2257045303231598e-05, "loss": 0.0502, "step": 19400 }, { "epoch": 3.56, "learning_rate": 1.2256322024795882e-05, "loss": 0.0432, "step": 19401 }, { "epoch": 3.56, "learning_rate": 1.2255598733923264e-05, "loss": 0.0494, "step": 19402 }, { "epoch": 3.56, "learning_rate": 1.2254875430617732e-05, "loss": 0.0402, "step": 19403 }, { "epoch": 3.56, "learning_rate": 1.2254152114883276e-05, "loss": 0.0499, "step": 19404 }, { "epoch": 3.56, "learning_rate": 1.2253428786723878e-05, "loss": 0.046, "step": 19405 }, { "epoch": 3.56, "learning_rate": 1.2252705446143532e-05, "loss": 0.0448, "step": 19406 }, { "epoch": 3.56, "learning_rate": 1.2251982093146216e-05, "loss": 0.0448, "step": 19407 }, { "epoch": 3.56, "learning_rate": 1.2251258727735925e-05, "loss": 0.0463, "step": 19408 }, { "epoch": 3.56, "learning_rate": 1.225053534991664e-05, "loss": 0.0474, "step": 19409 }, { "epoch": 3.56, "learning_rate": 1.2249811959692355e-05, "loss": 0.0428, "step": 19410 }, { "epoch": 3.56, "learning_rate": 1.2249088557067052e-05, "loss": 0.0422, "step": 19411 }, { "epoch": 3.56, "learning_rate": 1.2248365142044722e-05, "loss": 0.041, "step": 19412 }, { "epoch": 3.56, "learning_rate": 1.224764171462935e-05, "loss": 0.0483, "step": 19413 }, { "epoch": 3.56, "learning_rate": 1.224691827482492e-05, "loss": 0.0442, "step": 19414 }, { "epoch": 3.56, "learning_rate": 1.224619482263543e-05, "loss": 0.0477, "step": 19415 }, { "epoch": 3.56, "learning_rate": 1.2245471358064858e-05, "loss": 0.0524, "step": 19416 }, { "epoch": 3.56, "learning_rate": 1.2244747881117196e-05, "loss": 0.0494, "step": 19417 }, { "epoch": 3.56, "learning_rate": 1.2244024391796432e-05, "loss": 0.0388, "step": 19418 }, { "epoch": 3.56, "learning_rate": 1.2243300890106551e-05, "loss": 0.0523, "step": 19419 }, { "epoch": 3.56, "learning_rate": 1.2242577376051543e-05, "loss": 0.0411, "step": 19420 }, { "epoch": 3.56, "learning_rate": 1.2241853849635399e-05, "loss": 0.0474, "step": 19421 }, { "epoch": 3.56, "learning_rate": 1.2241130310862102e-05, "loss": 0.0487, "step": 19422 }, { "epoch": 3.56, "learning_rate": 1.2240406759735642e-05, "loss": 0.0508, "step": 19423 }, { "epoch": 3.56, "learning_rate": 1.2239683196260008e-05, "loss": 0.0478, "step": 19424 }, { "epoch": 3.56, "learning_rate": 1.2238959620439186e-05, "loss": 0.0403, "step": 19425 }, { "epoch": 3.56, "learning_rate": 1.2238236032277169e-05, "loss": 0.047, "step": 19426 }, { "epoch": 3.56, "learning_rate": 1.223751243177794e-05, "loss": 0.0479, "step": 19427 }, { "epoch": 3.56, "learning_rate": 1.2236788818945492e-05, "loss": 0.047, "step": 19428 }, { "epoch": 3.56, "learning_rate": 1.223606519378381e-05, "loss": 0.042, "step": 19429 }, { "epoch": 3.56, "learning_rate": 1.2235341556296888e-05, "loss": 0.0439, "step": 19430 }, { "epoch": 3.56, "learning_rate": 1.2234617906488707e-05, "loss": 0.0524, "step": 19431 }, { "epoch": 3.56, "learning_rate": 1.223389424436326e-05, "loss": 0.0466, "step": 19432 }, { "epoch": 3.56, "learning_rate": 1.2233170569924538e-05, "loss": 0.0423, "step": 19433 }, { "epoch": 3.56, "learning_rate": 1.2232446883176525e-05, "loss": 0.0379, "step": 19434 }, { "epoch": 3.56, "learning_rate": 1.2231723184123212e-05, "loss": 0.0477, "step": 19435 }, { "epoch": 3.56, "learning_rate": 1.2230999472768593e-05, "loss": 0.0451, "step": 19436 }, { "epoch": 3.56, "learning_rate": 1.2230275749116647e-05, "loss": 0.0476, "step": 19437 }, { "epoch": 3.56, "learning_rate": 1.222955201317137e-05, "loss": 0.0501, "step": 19438 }, { "epoch": 3.56, "learning_rate": 1.2228828264936755e-05, "loss": 0.0538, "step": 19439 }, { "epoch": 3.57, "learning_rate": 1.222810450441678e-05, "loss": 0.041, "step": 19440 }, { "epoch": 3.57, "learning_rate": 1.2227380731615443e-05, "loss": 0.0496, "step": 19441 }, { "epoch": 3.57, "learning_rate": 1.2226656946536734e-05, "loss": 0.0487, "step": 19442 }, { "epoch": 3.57, "learning_rate": 1.2225933149184636e-05, "loss": 0.0448, "step": 19443 }, { "epoch": 3.57, "learning_rate": 1.2225209339563144e-05, "loss": 0.0453, "step": 19444 }, { "epoch": 3.57, "learning_rate": 1.2224485517676248e-05, "loss": 0.0429, "step": 19445 }, { "epoch": 3.57, "learning_rate": 1.2223761683527933e-05, "loss": 0.0398, "step": 19446 }, { "epoch": 3.57, "learning_rate": 1.2223037837122192e-05, "loss": 0.0462, "step": 19447 }, { "epoch": 3.57, "learning_rate": 1.2222313978463017e-05, "loss": 0.042, "step": 19448 }, { "epoch": 3.57, "learning_rate": 1.2221590107554388e-05, "loss": 0.0479, "step": 19449 }, { "epoch": 3.57, "learning_rate": 1.2220866224400309e-05, "loss": 0.047, "step": 19450 }, { "epoch": 3.57, "learning_rate": 1.2220142329004763e-05, "loss": 0.0493, "step": 19451 }, { "epoch": 3.57, "learning_rate": 1.2219418421371738e-05, "loss": 0.0523, "step": 19452 }, { "epoch": 3.57, "learning_rate": 1.2218694501505229e-05, "loss": 0.0478, "step": 19453 }, { "epoch": 3.57, "learning_rate": 1.221797056940922e-05, "loss": 0.0428, "step": 19454 }, { "epoch": 3.57, "learning_rate": 1.2217246625087709e-05, "loss": 0.0406, "step": 19455 }, { "epoch": 3.57, "learning_rate": 1.2216522668544681e-05, "loss": 0.0494, "step": 19456 }, { "epoch": 3.57, "learning_rate": 1.2215798699784129e-05, "loss": 0.0475, "step": 19457 }, { "epoch": 3.57, "learning_rate": 1.221507471881004e-05, "loss": 0.0455, "step": 19458 }, { "epoch": 3.57, "learning_rate": 1.2214350725626412e-05, "loss": 0.0583, "step": 19459 }, { "epoch": 3.57, "learning_rate": 1.2213626720237228e-05, "loss": 0.0455, "step": 19460 }, { "epoch": 3.57, "learning_rate": 1.2212902702646483e-05, "loss": 0.0529, "step": 19461 }, { "epoch": 3.57, "learning_rate": 1.2212178672858163e-05, "loss": 0.0432, "step": 19462 }, { "epoch": 3.57, "learning_rate": 1.2211454630876267e-05, "loss": 0.0412, "step": 19463 }, { "epoch": 3.57, "learning_rate": 1.221073057670478e-05, "loss": 0.0474, "step": 19464 }, { "epoch": 3.57, "learning_rate": 1.2210006510347694e-05, "loss": 0.0556, "step": 19465 }, { "epoch": 3.57, "learning_rate": 1.2209282431808998e-05, "loss": 0.0449, "step": 19466 }, { "epoch": 3.57, "learning_rate": 1.2208558341092687e-05, "loss": 0.0446, "step": 19467 }, { "epoch": 3.57, "learning_rate": 1.220783423820275e-05, "loss": 0.0434, "step": 19468 }, { "epoch": 3.57, "learning_rate": 1.220711012314318e-05, "loss": 0.0527, "step": 19469 }, { "epoch": 3.57, "learning_rate": 1.220638599591797e-05, "loss": 0.0474, "step": 19470 }, { "epoch": 3.57, "learning_rate": 1.2205661856531103e-05, "loss": 0.0506, "step": 19471 }, { "epoch": 3.57, "learning_rate": 1.220493770498658e-05, "loss": 0.0502, "step": 19472 }, { "epoch": 3.57, "learning_rate": 1.2204213541288385e-05, "loss": 0.0423, "step": 19473 }, { "epoch": 3.57, "learning_rate": 1.220348936544052e-05, "loss": 0.0399, "step": 19474 }, { "epoch": 3.57, "learning_rate": 1.2202765177446965e-05, "loss": 0.0504, "step": 19475 }, { "epoch": 3.57, "learning_rate": 1.2202040977311716e-05, "loss": 0.0513, "step": 19476 }, { "epoch": 3.57, "learning_rate": 1.2201316765038769e-05, "loss": 0.0459, "step": 19477 }, { "epoch": 3.57, "learning_rate": 1.220059254063211e-05, "loss": 0.0472, "step": 19478 }, { "epoch": 3.57, "learning_rate": 1.2199868304095735e-05, "loss": 0.0504, "step": 19479 }, { "epoch": 3.57, "learning_rate": 1.2199144055433631e-05, "loss": 0.0468, "step": 19480 }, { "epoch": 3.57, "learning_rate": 1.21984197946498e-05, "loss": 0.045, "step": 19481 }, { "epoch": 3.57, "learning_rate": 1.219769552174822e-05, "loss": 0.0457, "step": 19482 }, { "epoch": 3.57, "learning_rate": 1.2196971236732896e-05, "loss": 0.0425, "step": 19483 }, { "epoch": 3.57, "learning_rate": 1.2196246939607812e-05, "loss": 0.0479, "step": 19484 }, { "epoch": 3.57, "learning_rate": 1.2195522630376963e-05, "loss": 0.0448, "step": 19485 }, { "epoch": 3.57, "learning_rate": 1.2194798309044345e-05, "loss": 0.0439, "step": 19486 }, { "epoch": 3.57, "learning_rate": 1.2194073975613946e-05, "loss": 0.0453, "step": 19487 }, { "epoch": 3.57, "learning_rate": 1.219334963008976e-05, "loss": 0.0447, "step": 19488 }, { "epoch": 3.57, "learning_rate": 1.219262527247578e-05, "loss": 0.0377, "step": 19489 }, { "epoch": 3.57, "learning_rate": 1.2191900902775997e-05, "loss": 0.0455, "step": 19490 }, { "epoch": 3.57, "learning_rate": 1.2191176520994403e-05, "loss": 0.0438, "step": 19491 }, { "epoch": 3.57, "learning_rate": 1.2190452127134997e-05, "loss": 0.0482, "step": 19492 }, { "epoch": 3.57, "learning_rate": 1.2189727721201765e-05, "loss": 0.0482, "step": 19493 }, { "epoch": 3.57, "learning_rate": 1.2189003303198702e-05, "loss": 0.0522, "step": 19494 }, { "epoch": 3.58, "learning_rate": 1.2188278873129803e-05, "loss": 0.0446, "step": 19495 }, { "epoch": 3.58, "learning_rate": 1.218755443099906e-05, "loss": 0.0487, "step": 19496 }, { "epoch": 3.58, "learning_rate": 1.2186829976810466e-05, "loss": 0.0494, "step": 19497 }, { "epoch": 3.58, "learning_rate": 1.2186105510568012e-05, "loss": 0.053, "step": 19498 }, { "epoch": 3.58, "learning_rate": 1.2185381032275694e-05, "loss": 0.0503, "step": 19499 }, { "epoch": 3.58, "learning_rate": 1.2184656541937502e-05, "loss": 0.0434, "step": 19500 }, { "epoch": 3.58, "learning_rate": 1.2183932039557438e-05, "loss": 0.0552, "step": 19501 }, { "epoch": 3.58, "learning_rate": 1.2183207525139485e-05, "loss": 0.048, "step": 19502 }, { "epoch": 3.58, "learning_rate": 1.2182482998687642e-05, "loss": 0.0411, "step": 19503 }, { "epoch": 3.58, "learning_rate": 1.2181758460205901e-05, "loss": 0.0443, "step": 19504 }, { "epoch": 3.58, "learning_rate": 1.2181033909698257e-05, "loss": 0.0499, "step": 19505 }, { "epoch": 3.58, "learning_rate": 1.2180309347168699e-05, "loss": 0.0448, "step": 19506 }, { "epoch": 3.58, "learning_rate": 1.217958477262123e-05, "loss": 0.0486, "step": 19507 }, { "epoch": 3.58, "learning_rate": 1.217886018605984e-05, "loss": 0.0429, "step": 19508 }, { "epoch": 3.58, "learning_rate": 1.2178135587488515e-05, "loss": 0.0472, "step": 19509 }, { "epoch": 3.58, "learning_rate": 1.2177410976911264e-05, "loss": 0.0503, "step": 19510 }, { "epoch": 3.58, "learning_rate": 1.2176686354332066e-05, "loss": 0.0462, "step": 19511 }, { "epoch": 3.58, "learning_rate": 1.2175961719754924e-05, "loss": 0.0402, "step": 19512 }, { "epoch": 3.58, "learning_rate": 1.2175237073183829e-05, "loss": 0.0465, "step": 19513 }, { "epoch": 3.58, "learning_rate": 1.2174512414622776e-05, "loss": 0.0436, "step": 19514 }, { "epoch": 3.58, "learning_rate": 1.217378774407576e-05, "loss": 0.0403, "step": 19515 }, { "epoch": 3.58, "learning_rate": 1.2173063061546779e-05, "loss": 0.0404, "step": 19516 }, { "epoch": 3.58, "learning_rate": 1.2172338367039818e-05, "loss": 0.0404, "step": 19517 }, { "epoch": 3.58, "learning_rate": 1.2171613660558879e-05, "loss": 0.0448, "step": 19518 }, { "epoch": 3.58, "learning_rate": 1.2170888942107957e-05, "loss": 0.0438, "step": 19519 }, { "epoch": 3.58, "learning_rate": 1.2170164211691042e-05, "loss": 0.0433, "step": 19520 }, { "epoch": 3.58, "learning_rate": 1.2169439469312132e-05, "loss": 0.0539, "step": 19521 }, { "epoch": 3.58, "learning_rate": 1.2168714714975218e-05, "loss": 0.0493, "step": 19522 }, { "epoch": 3.58, "learning_rate": 1.2167989948684302e-05, "loss": 0.0453, "step": 19523 }, { "epoch": 3.58, "learning_rate": 1.216726517044337e-05, "loss": 0.0463, "step": 19524 }, { "epoch": 3.58, "learning_rate": 1.2166540380256426e-05, "loss": 0.0467, "step": 19525 }, { "epoch": 3.58, "learning_rate": 1.2165815578127458e-05, "loss": 0.0429, "step": 19526 }, { "epoch": 3.58, "learning_rate": 1.2165090764060463e-05, "loss": 0.0482, "step": 19527 }, { "epoch": 3.58, "learning_rate": 1.2164365938059442e-05, "loss": 0.0344, "step": 19528 }, { "epoch": 3.58, "learning_rate": 1.216364110012838e-05, "loss": 0.0424, "step": 19529 }, { "epoch": 3.58, "learning_rate": 1.2162916250271279e-05, "loss": 0.044, "step": 19530 }, { "epoch": 3.58, "learning_rate": 1.2162191388492133e-05, "loss": 0.0484, "step": 19531 }, { "epoch": 3.58, "learning_rate": 1.2161466514794937e-05, "loss": 0.0419, "step": 19532 }, { "epoch": 3.58, "learning_rate": 1.216074162918369e-05, "loss": 0.0474, "step": 19533 }, { "epoch": 3.58, "learning_rate": 1.216001673166238e-05, "loss": 0.0558, "step": 19534 }, { "epoch": 3.58, "learning_rate": 1.2159291822235008e-05, "loss": 0.0457, "step": 19535 }, { "epoch": 3.58, "learning_rate": 1.215856690090557e-05, "loss": 0.0447, "step": 19536 }, { "epoch": 3.58, "learning_rate": 1.2157841967678064e-05, "loss": 0.0551, "step": 19537 }, { "epoch": 3.58, "learning_rate": 1.2157117022556478e-05, "loss": 0.0501, "step": 19538 }, { "epoch": 3.58, "learning_rate": 1.2156392065544813e-05, "loss": 0.0415, "step": 19539 }, { "epoch": 3.58, "learning_rate": 1.2155667096647065e-05, "loss": 0.0415, "step": 19540 }, { "epoch": 3.58, "learning_rate": 1.2154942115867232e-05, "loss": 0.0509, "step": 19541 }, { "epoch": 3.58, "learning_rate": 1.2154217123209304e-05, "loss": 0.0451, "step": 19542 }, { "epoch": 3.58, "learning_rate": 1.2153492118677284e-05, "loss": 0.0508, "step": 19543 }, { "epoch": 3.58, "learning_rate": 1.215276710227516e-05, "loss": 0.0417, "step": 19544 }, { "epoch": 3.58, "learning_rate": 1.2152042074006935e-05, "loss": 0.0409, "step": 19545 }, { "epoch": 3.58, "learning_rate": 1.2151317033876607e-05, "loss": 0.0424, "step": 19546 }, { "epoch": 3.58, "learning_rate": 1.2150591981888167e-05, "loss": 0.0388, "step": 19547 }, { "epoch": 3.58, "learning_rate": 1.2149866918045614e-05, "loss": 0.0436, "step": 19548 }, { "epoch": 3.58, "learning_rate": 1.2149141842352945e-05, "loss": 0.0469, "step": 19549 }, { "epoch": 3.59, "learning_rate": 1.2148416754814154e-05, "loss": 0.044, "step": 19550 }, { "epoch": 3.59, "learning_rate": 1.214769165543324e-05, "loss": 0.0493, "step": 19551 }, { "epoch": 3.59, "learning_rate": 1.21469665442142e-05, "loss": 0.0423, "step": 19552 }, { "epoch": 3.59, "learning_rate": 1.2146241421161027e-05, "loss": 0.0498, "step": 19553 }, { "epoch": 3.59, "learning_rate": 1.2145516286277725e-05, "loss": 0.0434, "step": 19554 }, { "epoch": 3.59, "learning_rate": 1.2144791139568287e-05, "loss": 0.0456, "step": 19555 }, { "epoch": 3.59, "learning_rate": 1.2144065981036708e-05, "loss": 0.0469, "step": 19556 }, { "epoch": 3.59, "learning_rate": 1.2143340810686988e-05, "loss": 0.0423, "step": 19557 }, { "epoch": 3.59, "learning_rate": 1.2142615628523122e-05, "loss": 0.042, "step": 19558 }, { "epoch": 3.59, "learning_rate": 1.2141890434549109e-05, "loss": 0.0471, "step": 19559 }, { "epoch": 3.59, "learning_rate": 1.2141165228768948e-05, "loss": 0.047, "step": 19560 }, { "epoch": 3.59, "learning_rate": 1.2140440011186631e-05, "loss": 0.0454, "step": 19561 }, { "epoch": 3.59, "learning_rate": 1.2139714781806157e-05, "loss": 0.0446, "step": 19562 }, { "epoch": 3.59, "learning_rate": 1.213898954063153e-05, "loss": 0.0449, "step": 19563 }, { "epoch": 3.59, "learning_rate": 1.213826428766674e-05, "loss": 0.046, "step": 19564 }, { "epoch": 3.59, "learning_rate": 1.2137539022915787e-05, "loss": 0.0446, "step": 19565 }, { "epoch": 3.59, "learning_rate": 1.213681374638267e-05, "loss": 0.0369, "step": 19566 }, { "epoch": 3.59, "learning_rate": 1.2136088458071384e-05, "loss": 0.0487, "step": 19567 }, { "epoch": 3.59, "learning_rate": 1.2135363157985929e-05, "loss": 0.0436, "step": 19568 }, { "epoch": 3.59, "learning_rate": 1.2134637846130303e-05, "loss": 0.0515, "step": 19569 }, { "epoch": 3.59, "learning_rate": 1.2133912522508501e-05, "loss": 0.0351, "step": 19570 }, { "epoch": 3.59, "learning_rate": 1.2133187187124525e-05, "loss": 0.045, "step": 19571 }, { "epoch": 3.59, "learning_rate": 1.2132461839982375e-05, "loss": 0.0508, "step": 19572 }, { "epoch": 3.59, "learning_rate": 1.2131736481086038e-05, "loss": 0.0503, "step": 19573 }, { "epoch": 3.59, "learning_rate": 1.2131011110439523e-05, "loss": 0.0452, "step": 19574 }, { "epoch": 3.59, "learning_rate": 1.2130285728046824e-05, "loss": 0.0388, "step": 19575 }, { "epoch": 3.59, "learning_rate": 1.2129560333911943e-05, "loss": 0.047, "step": 19576 }, { "epoch": 3.59, "learning_rate": 1.212883492803887e-05, "loss": 0.0512, "step": 19577 }, { "epoch": 3.59, "learning_rate": 1.2128109510431617e-05, "loss": 0.0432, "step": 19578 }, { "epoch": 3.59, "learning_rate": 1.2127384081094167e-05, "loss": 0.0484, "step": 19579 }, { "epoch": 3.59, "learning_rate": 1.2126658640030531e-05, "loss": 0.0435, "step": 19580 }, { "epoch": 3.59, "learning_rate": 1.21259331872447e-05, "loss": 0.0424, "step": 19581 }, { "epoch": 3.59, "learning_rate": 1.2125207722740677e-05, "loss": 0.0484, "step": 19582 }, { "epoch": 3.59, "learning_rate": 1.2124482246522457e-05, "loss": 0.0369, "step": 19583 }, { "epoch": 3.59, "learning_rate": 1.2123756758594044e-05, "loss": 0.0444, "step": 19584 }, { "epoch": 3.59, "learning_rate": 1.2123031258959433e-05, "loss": 0.0482, "step": 19585 }, { "epoch": 3.59, "learning_rate": 1.2122305747622622e-05, "loss": 0.0484, "step": 19586 }, { "epoch": 3.59, "learning_rate": 1.2121580224587618e-05, "loss": 0.0445, "step": 19587 }, { "epoch": 3.59, "learning_rate": 1.2120854689858405e-05, "loss": 0.0435, "step": 19588 }, { "epoch": 3.59, "learning_rate": 1.2120129143438997e-05, "loss": 0.0484, "step": 19589 }, { "epoch": 3.59, "learning_rate": 1.2119403585333386e-05, "loss": 0.041, "step": 19590 }, { "epoch": 3.59, "learning_rate": 1.2118678015545575e-05, "loss": 0.0478, "step": 19591 }, { "epoch": 3.59, "learning_rate": 1.211795243407956e-05, "loss": 0.0405, "step": 19592 }, { "epoch": 3.59, "learning_rate": 1.211722684093934e-05, "loss": 0.042, "step": 19593 }, { "epoch": 3.59, "learning_rate": 1.2116501236128917e-05, "loss": 0.0513, "step": 19594 }, { "epoch": 3.59, "learning_rate": 1.2115775619652287e-05, "loss": 0.04, "step": 19595 }, { "epoch": 3.59, "learning_rate": 1.2115049991513458e-05, "loss": 0.0392, "step": 19596 }, { "epoch": 3.59, "learning_rate": 1.211432435171642e-05, "loss": 0.0417, "step": 19597 }, { "epoch": 3.59, "learning_rate": 1.2113598700265177e-05, "loss": 0.0437, "step": 19598 }, { "epoch": 3.59, "learning_rate": 1.2112873037163728e-05, "loss": 0.048, "step": 19599 }, { "epoch": 3.59, "learning_rate": 1.2112147362416076e-05, "loss": 0.0385, "step": 19600 }, { "epoch": 3.59, "learning_rate": 1.2111421676026214e-05, "loss": 0.0472, "step": 19601 }, { "epoch": 3.59, "learning_rate": 1.211069597799815e-05, "loss": 0.0496, "step": 19602 }, { "epoch": 3.59, "learning_rate": 1.210997026833588e-05, "loss": 0.0441, "step": 19603 }, { "epoch": 3.6, "learning_rate": 1.2109244547043401e-05, "loss": 0.0422, "step": 19604 }, { "epoch": 3.6, "learning_rate": 1.2108518814124723e-05, "loss": 0.0404, "step": 19605 }, { "epoch": 3.6, "learning_rate": 1.2107793069583833e-05, "loss": 0.0365, "step": 19606 }, { "epoch": 3.6, "learning_rate": 1.2107067313424743e-05, "loss": 0.0403, "step": 19607 }, { "epoch": 3.6, "learning_rate": 1.2106341545651445e-05, "loss": 0.0495, "step": 19608 }, { "epoch": 3.6, "learning_rate": 1.2105615766267947e-05, "loss": 0.0481, "step": 19609 }, { "epoch": 3.6, "learning_rate": 1.2104889975278242e-05, "loss": 0.0398, "step": 19610 }, { "epoch": 3.6, "learning_rate": 1.2104164172686337e-05, "loss": 0.0448, "step": 19611 }, { "epoch": 3.6, "learning_rate": 1.2103438358496226e-05, "loss": 0.0457, "step": 19612 }, { "epoch": 3.6, "learning_rate": 1.2102712532711915e-05, "loss": 0.043, "step": 19613 }, { "epoch": 3.6, "learning_rate": 1.2101986695337407e-05, "loss": 0.042, "step": 19614 }, { "epoch": 3.6, "learning_rate": 1.2101260846376694e-05, "loss": 0.0489, "step": 19615 }, { "epoch": 3.6, "learning_rate": 1.2100534985833785e-05, "loss": 0.0526, "step": 19616 }, { "epoch": 3.6, "learning_rate": 1.2099809113712678e-05, "loss": 0.0499, "step": 19617 }, { "epoch": 3.6, "learning_rate": 1.2099083230017373e-05, "loss": 0.0502, "step": 19618 }, { "epoch": 3.6, "learning_rate": 1.2098357334751873e-05, "loss": 0.0472, "step": 19619 }, { "epoch": 3.6, "learning_rate": 1.2097631427920178e-05, "loss": 0.042, "step": 19620 }, { "epoch": 3.6, "learning_rate": 1.2096905509526286e-05, "loss": 0.0459, "step": 19621 }, { "epoch": 3.6, "learning_rate": 1.2096179579574205e-05, "loss": 0.0511, "step": 19622 }, { "epoch": 3.6, "learning_rate": 1.2095453638067933e-05, "loss": 0.0417, "step": 19623 }, { "epoch": 3.6, "learning_rate": 1.2094727685011472e-05, "loss": 0.0494, "step": 19624 }, { "epoch": 3.6, "learning_rate": 1.209400172040882e-05, "loss": 0.042, "step": 19625 }, { "epoch": 3.6, "learning_rate": 1.2093275744263983e-05, "loss": 0.0452, "step": 19626 }, { "epoch": 3.6, "learning_rate": 1.2092549756580962e-05, "loss": 0.0371, "step": 19627 }, { "epoch": 3.6, "learning_rate": 1.2091823757363759e-05, "loss": 0.0446, "step": 19628 }, { "epoch": 3.6, "learning_rate": 1.2091097746616372e-05, "loss": 0.0561, "step": 19629 }, { "epoch": 3.6, "learning_rate": 1.2090371724342804e-05, "loss": 0.0485, "step": 19630 }, { "epoch": 3.6, "learning_rate": 1.2089645690547061e-05, "loss": 0.0446, "step": 19631 }, { "epoch": 3.6, "learning_rate": 1.2088919645233142e-05, "loss": 0.0362, "step": 19632 }, { "epoch": 3.6, "learning_rate": 1.2088193588405047e-05, "loss": 0.0547, "step": 19633 }, { "epoch": 3.6, "learning_rate": 1.208746752006678e-05, "loss": 0.0476, "step": 19634 }, { "epoch": 3.6, "learning_rate": 1.2086741440222345e-05, "loss": 0.0488, "step": 19635 }, { "epoch": 3.6, "learning_rate": 1.208601534887574e-05, "loss": 0.043, "step": 19636 }, { "epoch": 3.6, "learning_rate": 1.2085289246030973e-05, "loss": 0.0491, "step": 19637 }, { "epoch": 3.6, "learning_rate": 1.2084563131692039e-05, "loss": 0.0455, "step": 19638 }, { "epoch": 3.6, "learning_rate": 1.2083837005862946e-05, "loss": 0.0415, "step": 19639 }, { "epoch": 3.6, "learning_rate": 1.2083110868547694e-05, "loss": 0.0452, "step": 19640 }, { "epoch": 3.6, "learning_rate": 1.2082384719750286e-05, "loss": 0.039, "step": 19641 }, { "epoch": 3.6, "learning_rate": 1.2081658559474726e-05, "loss": 0.0376, "step": 19642 }, { "epoch": 3.6, "learning_rate": 1.2080932387725016e-05, "loss": 0.0477, "step": 19643 }, { "epoch": 3.6, "learning_rate": 1.2080206204505156e-05, "loss": 0.0567, "step": 19644 }, { "epoch": 3.6, "learning_rate": 1.2079480009819149e-05, "loss": 0.0428, "step": 19645 }, { "epoch": 3.6, "learning_rate": 1.2078753803671003e-05, "loss": 0.0409, "step": 19646 }, { "epoch": 3.6, "learning_rate": 1.2078027586064718e-05, "loss": 0.0445, "step": 19647 }, { "epoch": 3.6, "learning_rate": 1.2077301357004292e-05, "loss": 0.0454, "step": 19648 }, { "epoch": 3.6, "learning_rate": 1.2076575116493736e-05, "loss": 0.0487, "step": 19649 }, { "epoch": 3.6, "learning_rate": 1.2075848864537046e-05, "loss": 0.0444, "step": 19650 }, { "epoch": 3.6, "learning_rate": 1.207512260113823e-05, "loss": 0.0444, "step": 19651 }, { "epoch": 3.6, "learning_rate": 1.207439632630129e-05, "loss": 0.0502, "step": 19652 }, { "epoch": 3.6, "learning_rate": 1.207367004003023e-05, "loss": 0.0491, "step": 19653 }, { "epoch": 3.6, "learning_rate": 1.2072943742329048e-05, "loss": 0.0393, "step": 19654 }, { "epoch": 3.6, "learning_rate": 1.2072217433201756e-05, "loss": 0.0449, "step": 19655 }, { "epoch": 3.6, "learning_rate": 1.2071491112652351e-05, "loss": 0.0447, "step": 19656 }, { "epoch": 3.6, "learning_rate": 1.2070764780684839e-05, "loss": 0.0443, "step": 19657 }, { "epoch": 3.6, "learning_rate": 1.2070038437303224e-05, "loss": 0.0438, "step": 19658 }, { "epoch": 3.61, "learning_rate": 1.2069312082511506e-05, "loss": 0.0392, "step": 19659 }, { "epoch": 3.61, "learning_rate": 1.2068585716313692e-05, "loss": 0.0445, "step": 19660 }, { "epoch": 3.61, "learning_rate": 1.2067859338713786e-05, "loss": 0.0436, "step": 19661 }, { "epoch": 3.61, "learning_rate": 1.2067132949715791e-05, "loss": 0.0437, "step": 19662 }, { "epoch": 3.61, "learning_rate": 1.2066406549323707e-05, "loss": 0.0488, "step": 19663 }, { "epoch": 3.61, "learning_rate": 1.206568013754155e-05, "loss": 0.0426, "step": 19664 }, { "epoch": 3.61, "learning_rate": 1.2064953714373309e-05, "loss": 0.0384, "step": 19665 }, { "epoch": 3.61, "learning_rate": 1.2064227279822995e-05, "loss": 0.0418, "step": 19666 }, { "epoch": 3.61, "learning_rate": 1.2063500833894616e-05, "loss": 0.0417, "step": 19667 }, { "epoch": 3.61, "learning_rate": 1.2062774376592168e-05, "loss": 0.0537, "step": 19668 }, { "epoch": 3.61, "learning_rate": 1.206204790791966e-05, "loss": 0.0418, "step": 19669 }, { "epoch": 3.61, "learning_rate": 1.2061321427881097e-05, "loss": 0.0426, "step": 19670 }, { "epoch": 3.61, "learning_rate": 1.2060594936480483e-05, "loss": 0.0408, "step": 19671 }, { "epoch": 3.61, "learning_rate": 1.2059868433721818e-05, "loss": 0.0361, "step": 19672 }, { "epoch": 3.61, "learning_rate": 1.2059141919609112e-05, "loss": 0.0471, "step": 19673 }, { "epoch": 3.61, "learning_rate": 1.2058415394146366e-05, "loss": 0.046, "step": 19674 }, { "epoch": 3.61, "learning_rate": 1.2057688857337588e-05, "loss": 0.0376, "step": 19675 }, { "epoch": 3.61, "learning_rate": 1.2056962309186781e-05, "loss": 0.0496, "step": 19676 }, { "epoch": 3.61, "learning_rate": 1.2056235749697949e-05, "loss": 0.0321, "step": 19677 }, { "epoch": 3.61, "learning_rate": 1.2055509178875098e-05, "loss": 0.0411, "step": 19678 }, { "epoch": 3.61, "learning_rate": 1.205478259672223e-05, "loss": 0.0467, "step": 19679 }, { "epoch": 3.61, "learning_rate": 1.2054056003243355e-05, "loss": 0.0432, "step": 19680 }, { "epoch": 3.61, "learning_rate": 1.205332939844247e-05, "loss": 0.0423, "step": 19681 }, { "epoch": 3.61, "learning_rate": 1.2052602782323592e-05, "loss": 0.0511, "step": 19682 }, { "epoch": 3.61, "learning_rate": 1.2051876154890713e-05, "loss": 0.037, "step": 19683 }, { "epoch": 3.61, "learning_rate": 1.205114951614785e-05, "loss": 0.0504, "step": 19684 }, { "epoch": 3.61, "learning_rate": 1.2050422866099e-05, "loss": 0.0476, "step": 19685 }, { "epoch": 3.61, "learning_rate": 1.2049696204748173e-05, "loss": 0.0478, "step": 19686 }, { "epoch": 3.61, "learning_rate": 1.204896953209937e-05, "loss": 0.0426, "step": 19687 }, { "epoch": 3.61, "learning_rate": 1.2048242848156602e-05, "loss": 0.0458, "step": 19688 }, { "epoch": 3.61, "learning_rate": 1.204751615292387e-05, "loss": 0.0489, "step": 19689 }, { "epoch": 3.61, "learning_rate": 1.2046789446405179e-05, "loss": 0.0488, "step": 19690 }, { "epoch": 3.61, "learning_rate": 1.204606272860454e-05, "loss": 0.0346, "step": 19691 }, { "epoch": 3.61, "learning_rate": 1.2045335999525953e-05, "loss": 0.0462, "step": 19692 }, { "epoch": 3.61, "learning_rate": 1.2044609259173425e-05, "loss": 0.0479, "step": 19693 }, { "epoch": 3.61, "learning_rate": 1.2043882507550965e-05, "loss": 0.0393, "step": 19694 }, { "epoch": 3.61, "learning_rate": 1.2043155744662579e-05, "loss": 0.0439, "step": 19695 }, { "epoch": 3.61, "learning_rate": 1.2042428970512267e-05, "loss": 0.0453, "step": 19696 }, { "epoch": 3.61, "learning_rate": 1.204170218510404e-05, "loss": 0.0492, "step": 19697 }, { "epoch": 3.61, "learning_rate": 1.20409753884419e-05, "loss": 0.0424, "step": 19698 }, { "epoch": 3.61, "learning_rate": 1.2040248580529858e-05, "loss": 0.0457, "step": 19699 }, { "epoch": 3.61, "learning_rate": 1.2039521761371922e-05, "loss": 0.0413, "step": 19700 }, { "epoch": 3.61, "learning_rate": 1.2038794930972089e-05, "loss": 0.0436, "step": 19701 }, { "epoch": 3.61, "learning_rate": 1.2038068089334371e-05, "loss": 0.0419, "step": 19702 }, { "epoch": 3.61, "learning_rate": 1.2037341236462776e-05, "loss": 0.0507, "step": 19703 }, { "epoch": 3.61, "learning_rate": 1.2036614372361309e-05, "loss": 0.0459, "step": 19704 }, { "epoch": 3.61, "learning_rate": 1.2035887497033974e-05, "loss": 0.0456, "step": 19705 }, { "epoch": 3.61, "learning_rate": 1.203516061048478e-05, "loss": 0.0444, "step": 19706 }, { "epoch": 3.61, "learning_rate": 1.203443371271773e-05, "loss": 0.053, "step": 19707 }, { "epoch": 3.61, "learning_rate": 1.2033706803736836e-05, "loss": 0.0521, "step": 19708 }, { "epoch": 3.61, "learning_rate": 1.2032979883546104e-05, "loss": 0.0426, "step": 19709 }, { "epoch": 3.61, "learning_rate": 1.2032252952149536e-05, "loss": 0.0489, "step": 19710 }, { "epoch": 3.61, "learning_rate": 1.2031526009551146e-05, "loss": 0.0442, "step": 19711 }, { "epoch": 3.61, "learning_rate": 1.2030799055754936e-05, "loss": 0.0432, "step": 19712 }, { "epoch": 3.62, "learning_rate": 1.2030072090764912e-05, "loss": 0.0528, "step": 19713 }, { "epoch": 3.62, "learning_rate": 1.2029345114585086e-05, "loss": 0.0416, "step": 19714 }, { "epoch": 3.62, "learning_rate": 1.2028618127219461e-05, "loss": 0.0379, "step": 19715 }, { "epoch": 3.62, "learning_rate": 1.2027891128672042e-05, "loss": 0.0469, "step": 19716 }, { "epoch": 3.62, "learning_rate": 1.2027164118946843e-05, "loss": 0.0447, "step": 19717 }, { "epoch": 3.62, "learning_rate": 1.2026437098047867e-05, "loss": 0.0486, "step": 19718 }, { "epoch": 3.62, "learning_rate": 1.2025710065979123e-05, "loss": 0.0418, "step": 19719 }, { "epoch": 3.62, "learning_rate": 1.2024983022744618e-05, "loss": 0.0457, "step": 19720 }, { "epoch": 3.62, "learning_rate": 1.202425596834836e-05, "loss": 0.0427, "step": 19721 }, { "epoch": 3.62, "learning_rate": 1.2023528902794355e-05, "loss": 0.0449, "step": 19722 }, { "epoch": 3.62, "learning_rate": 1.2022801826086609e-05, "loss": 0.0481, "step": 19723 }, { "epoch": 3.62, "learning_rate": 1.2022074738229134e-05, "loss": 0.0466, "step": 19724 }, { "epoch": 3.62, "learning_rate": 1.2021347639225933e-05, "loss": 0.0398, "step": 19725 }, { "epoch": 3.62, "learning_rate": 1.2020620529081022e-05, "loss": 0.042, "step": 19726 }, { "epoch": 3.62, "learning_rate": 1.2019893407798397e-05, "loss": 0.0452, "step": 19727 }, { "epoch": 3.62, "learning_rate": 1.2019166275382075e-05, "loss": 0.0472, "step": 19728 }, { "epoch": 3.62, "learning_rate": 1.2018439131836062e-05, "loss": 0.0384, "step": 19729 }, { "epoch": 3.62, "learning_rate": 1.2017711977164364e-05, "loss": 0.049, "step": 19730 }, { "epoch": 3.62, "learning_rate": 1.2016984811370991e-05, "loss": 0.0459, "step": 19731 }, { "epoch": 3.62, "learning_rate": 1.201625763445995e-05, "loss": 0.0484, "step": 19732 }, { "epoch": 3.62, "learning_rate": 1.2015530446435251e-05, "loss": 0.0449, "step": 19733 }, { "epoch": 3.62, "learning_rate": 1.2014803247300897e-05, "loss": 0.0544, "step": 19734 }, { "epoch": 3.62, "learning_rate": 1.2014076037060907e-05, "loss": 0.0386, "step": 19735 }, { "epoch": 3.62, "learning_rate": 1.2013348815719276e-05, "loss": 0.0523, "step": 19736 }, { "epoch": 3.62, "learning_rate": 1.2012621583280021e-05, "loss": 0.0403, "step": 19737 }, { "epoch": 3.62, "learning_rate": 1.2011894339747153e-05, "loss": 0.0444, "step": 19738 }, { "epoch": 3.62, "learning_rate": 1.201116708512467e-05, "loss": 0.0402, "step": 19739 }, { "epoch": 3.62, "learning_rate": 1.2010439819416591e-05, "loss": 0.0496, "step": 19740 }, { "epoch": 3.62, "learning_rate": 1.2009712542626922e-05, "loss": 0.0461, "step": 19741 }, { "epoch": 3.62, "learning_rate": 1.2008985254759666e-05, "loss": 0.0449, "step": 19742 }, { "epoch": 3.62, "learning_rate": 1.2008257955818836e-05, "loss": 0.0489, "step": 19743 }, { "epoch": 3.62, "learning_rate": 1.2007530645808445e-05, "loss": 0.0492, "step": 19744 }, { "epoch": 3.62, "learning_rate": 1.2006803324732496e-05, "loss": 0.0417, "step": 19745 }, { "epoch": 3.62, "learning_rate": 1.2006075992595e-05, "loss": 0.0398, "step": 19746 }, { "epoch": 3.62, "learning_rate": 1.2005348649399965e-05, "loss": 0.0484, "step": 19747 }, { "epoch": 3.62, "learning_rate": 1.2004621295151405e-05, "loss": 0.0453, "step": 19748 }, { "epoch": 3.62, "learning_rate": 1.2003893929853321e-05, "loss": 0.041, "step": 19749 }, { "epoch": 3.62, "learning_rate": 1.2003166553509732e-05, "loss": 0.0507, "step": 19750 }, { "epoch": 3.62, "learning_rate": 1.2002439166124635e-05, "loss": 0.0543, "step": 19751 }, { "epoch": 3.62, "learning_rate": 1.2001711767702051e-05, "loss": 0.0465, "step": 19752 }, { "epoch": 3.62, "learning_rate": 1.2000984358245987e-05, "loss": 0.0505, "step": 19753 }, { "epoch": 3.62, "learning_rate": 1.2000256937760446e-05, "loss": 0.0473, "step": 19754 }, { "epoch": 3.62, "learning_rate": 1.1999529506249443e-05, "loss": 0.0462, "step": 19755 }, { "epoch": 3.62, "learning_rate": 1.1998802063716988e-05, "loss": 0.0445, "step": 19756 }, { "epoch": 3.62, "learning_rate": 1.1998074610167087e-05, "loss": 0.042, "step": 19757 }, { "epoch": 3.62, "learning_rate": 1.1997347145603752e-05, "loss": 0.0349, "step": 19758 }, { "epoch": 3.62, "learning_rate": 1.1996619670030996e-05, "loss": 0.0462, "step": 19759 }, { "epoch": 3.62, "learning_rate": 1.1995892183452821e-05, "loss": 0.0484, "step": 19760 }, { "epoch": 3.62, "learning_rate": 1.1995164685873243e-05, "loss": 0.0456, "step": 19761 }, { "epoch": 3.62, "learning_rate": 1.1994437177296274e-05, "loss": 0.0472, "step": 19762 }, { "epoch": 3.62, "learning_rate": 1.1993709657725918e-05, "loss": 0.0335, "step": 19763 }, { "epoch": 3.62, "learning_rate": 1.1992982127166186e-05, "loss": 0.0449, "step": 19764 }, { "epoch": 3.62, "learning_rate": 1.1992254585621093e-05, "loss": 0.0448, "step": 19765 }, { "epoch": 3.62, "learning_rate": 1.1991527033094644e-05, "loss": 0.0499, "step": 19766 }, { "epoch": 3.62, "learning_rate": 1.199079946959085e-05, "loss": 0.0529, "step": 19767 }, { "epoch": 3.63, "learning_rate": 1.1990071895113727e-05, "loss": 0.0416, "step": 19768 }, { "epoch": 3.63, "learning_rate": 1.1989344309667278e-05, "loss": 0.0452, "step": 19769 }, { "epoch": 3.63, "learning_rate": 1.1988616713255516e-05, "loss": 0.0489, "step": 19770 }, { "epoch": 3.63, "learning_rate": 1.1987889105882453e-05, "loss": 0.0464, "step": 19771 }, { "epoch": 3.63, "learning_rate": 1.19871614875521e-05, "loss": 0.0381, "step": 19772 }, { "epoch": 3.63, "learning_rate": 1.1986433858268464e-05, "loss": 0.0386, "step": 19773 }, { "epoch": 3.63, "learning_rate": 1.198570621803556e-05, "loss": 0.0474, "step": 19774 }, { "epoch": 3.63, "learning_rate": 1.1984978566857397e-05, "loss": 0.0437, "step": 19775 }, { "epoch": 3.63, "learning_rate": 1.1984250904737981e-05, "loss": 0.043, "step": 19776 }, { "epoch": 3.63, "learning_rate": 1.1983523231681335e-05, "loss": 0.0464, "step": 19777 }, { "epoch": 3.63, "learning_rate": 1.1982795547691457e-05, "loss": 0.0444, "step": 19778 }, { "epoch": 3.63, "learning_rate": 1.1982067852772366e-05, "loss": 0.0488, "step": 19779 }, { "epoch": 3.63, "learning_rate": 1.198134014692807e-05, "loss": 0.0399, "step": 19780 }, { "epoch": 3.63, "learning_rate": 1.1980612430162582e-05, "loss": 0.0381, "step": 19781 }, { "epoch": 3.63, "learning_rate": 1.1979884702479909e-05, "loss": 0.05, "step": 19782 }, { "epoch": 3.63, "learning_rate": 1.1979156963884066e-05, "loss": 0.0465, "step": 19783 }, { "epoch": 3.63, "learning_rate": 1.1978429214379067e-05, "loss": 0.0465, "step": 19784 }, { "epoch": 3.63, "learning_rate": 1.1977701453968912e-05, "loss": 0.0458, "step": 19785 }, { "epoch": 3.63, "learning_rate": 1.1976973682657629e-05, "loss": 0.0493, "step": 19786 }, { "epoch": 3.63, "learning_rate": 1.1976245900449215e-05, "loss": 0.0433, "step": 19787 }, { "epoch": 3.63, "learning_rate": 1.1975518107347687e-05, "loss": 0.0388, "step": 19788 }, { "epoch": 3.63, "learning_rate": 1.197479030335706e-05, "loss": 0.0395, "step": 19789 }, { "epoch": 3.63, "learning_rate": 1.1974062488481341e-05, "loss": 0.0455, "step": 19790 }, { "epoch": 3.63, "learning_rate": 1.1973334662724545e-05, "loss": 0.0401, "step": 19791 }, { "epoch": 3.63, "learning_rate": 1.1972606826090682e-05, "loss": 0.0458, "step": 19792 }, { "epoch": 3.63, "learning_rate": 1.1971878978583762e-05, "loss": 0.0451, "step": 19793 }, { "epoch": 3.63, "learning_rate": 1.1971151120207798e-05, "loss": 0.0427, "step": 19794 }, { "epoch": 3.63, "learning_rate": 1.1970423250966807e-05, "loss": 0.0487, "step": 19795 }, { "epoch": 3.63, "learning_rate": 1.1969695370864793e-05, "loss": 0.0384, "step": 19796 }, { "epoch": 3.63, "learning_rate": 1.1968967479905773e-05, "loss": 0.0442, "step": 19797 }, { "epoch": 3.63, "learning_rate": 1.1968239578093761e-05, "loss": 0.0489, "step": 19798 }, { "epoch": 3.63, "learning_rate": 1.1967511665432763e-05, "loss": 0.0441, "step": 19799 }, { "epoch": 3.63, "learning_rate": 1.1966783741926794e-05, "loss": 0.0423, "step": 19800 }, { "epoch": 3.63, "learning_rate": 1.1966055807579871e-05, "loss": 0.0464, "step": 19801 }, { "epoch": 3.63, "learning_rate": 1.1965327862395996e-05, "loss": 0.0474, "step": 19802 }, { "epoch": 3.63, "learning_rate": 1.1964599906379196e-05, "loss": 0.0545, "step": 19803 }, { "epoch": 3.63, "learning_rate": 1.196387193953347e-05, "loss": 0.0475, "step": 19804 }, { "epoch": 3.63, "learning_rate": 1.1963143961862836e-05, "loss": 0.0492, "step": 19805 }, { "epoch": 3.63, "learning_rate": 1.1962415973371306e-05, "loss": 0.0433, "step": 19806 }, { "epoch": 3.63, "learning_rate": 1.1961687974062897e-05, "loss": 0.0513, "step": 19807 }, { "epoch": 3.63, "learning_rate": 1.1960959963941614e-05, "loss": 0.0492, "step": 19808 }, { "epoch": 3.63, "learning_rate": 1.1960231943011476e-05, "loss": 0.0433, "step": 19809 }, { "epoch": 3.63, "learning_rate": 1.1959503911276493e-05, "loss": 0.043, "step": 19810 }, { "epoch": 3.63, "learning_rate": 1.1958775868740675e-05, "loss": 0.041, "step": 19811 }, { "epoch": 3.63, "learning_rate": 1.1958047815408045e-05, "loss": 0.0441, "step": 19812 }, { "epoch": 3.63, "learning_rate": 1.1957319751282605e-05, "loss": 0.0445, "step": 19813 }, { "epoch": 3.63, "learning_rate": 1.1956591676368374e-05, "loss": 0.044, "step": 19814 }, { "epoch": 3.63, "learning_rate": 1.1955863590669363e-05, "loss": 0.0463, "step": 19815 }, { "epoch": 3.63, "learning_rate": 1.195513549418959e-05, "loss": 0.0453, "step": 19816 }, { "epoch": 3.63, "learning_rate": 1.195440738693306e-05, "loss": 0.0364, "step": 19817 }, { "epoch": 3.63, "learning_rate": 1.1953679268903794e-05, "loss": 0.0514, "step": 19818 }, { "epoch": 3.63, "learning_rate": 1.1952951140105797e-05, "loss": 0.0412, "step": 19819 }, { "epoch": 3.63, "learning_rate": 1.195222300054309e-05, "loss": 0.0409, "step": 19820 }, { "epoch": 3.63, "learning_rate": 1.1951494850219688e-05, "loss": 0.0464, "step": 19821 }, { "epoch": 3.64, "learning_rate": 1.1950766689139595e-05, "loss": 0.0395, "step": 19822 }, { "epoch": 3.64, "learning_rate": 1.1950038517306832e-05, "loss": 0.0448, "step": 19823 }, { "epoch": 3.64, "learning_rate": 1.1949310334725413e-05, "loss": 0.036, "step": 19824 }, { "epoch": 3.64, "learning_rate": 1.1948582141399347e-05, "loss": 0.0392, "step": 19825 }, { "epoch": 3.64, "learning_rate": 1.1947853937332652e-05, "loss": 0.0439, "step": 19826 }, { "epoch": 3.64, "learning_rate": 1.1947125722529343e-05, "loss": 0.048, "step": 19827 }, { "epoch": 3.64, "learning_rate": 1.1946397496993427e-05, "loss": 0.0537, "step": 19828 }, { "epoch": 3.64, "learning_rate": 1.1945669260728923e-05, "loss": 0.0405, "step": 19829 }, { "epoch": 3.64, "learning_rate": 1.1944941013739849e-05, "loss": 0.0408, "step": 19830 }, { "epoch": 3.64, "learning_rate": 1.1944212756030209e-05, "loss": 0.0481, "step": 19831 }, { "epoch": 3.64, "learning_rate": 1.1943484487604025e-05, "loss": 0.0437, "step": 19832 }, { "epoch": 3.64, "learning_rate": 1.194275620846531e-05, "loss": 0.0458, "step": 19833 }, { "epoch": 3.64, "learning_rate": 1.1942027918618076e-05, "loss": 0.0428, "step": 19834 }, { "epoch": 3.64, "learning_rate": 1.1941299618066338e-05, "loss": 0.037, "step": 19835 }, { "epoch": 3.64, "learning_rate": 1.1940571306814115e-05, "loss": 0.0503, "step": 19836 }, { "epoch": 3.64, "learning_rate": 1.1939842984865412e-05, "loss": 0.0431, "step": 19837 }, { "epoch": 3.64, "learning_rate": 1.193911465222425e-05, "loss": 0.0557, "step": 19838 }, { "epoch": 3.64, "learning_rate": 1.1938386308894648e-05, "loss": 0.0388, "step": 19839 }, { "epoch": 3.64, "learning_rate": 1.1937657954880607e-05, "loss": 0.0421, "step": 19840 }, { "epoch": 3.64, "learning_rate": 1.1936929590186157e-05, "loss": 0.0494, "step": 19841 }, { "epoch": 3.64, "learning_rate": 1.1936201214815304e-05, "loss": 0.0459, "step": 19842 }, { "epoch": 3.64, "learning_rate": 1.1935472828772063e-05, "loss": 0.041, "step": 19843 }, { "epoch": 3.64, "learning_rate": 1.1934744432060447e-05, "loss": 0.0448, "step": 19844 }, { "epoch": 3.64, "learning_rate": 1.1934016024684483e-05, "loss": 0.0401, "step": 19845 }, { "epoch": 3.64, "learning_rate": 1.1933287606648169e-05, "loss": 0.0402, "step": 19846 }, { "epoch": 3.64, "learning_rate": 1.1932559177955533e-05, "loss": 0.0461, "step": 19847 }, { "epoch": 3.64, "learning_rate": 1.1931830738610587e-05, "loss": 0.0439, "step": 19848 }, { "epoch": 3.64, "learning_rate": 1.193110228861734e-05, "loss": 0.0459, "step": 19849 }, { "epoch": 3.64, "learning_rate": 1.1930373827979815e-05, "loss": 0.0427, "step": 19850 }, { "epoch": 3.64, "learning_rate": 1.1929645356702025e-05, "loss": 0.0449, "step": 19851 }, { "epoch": 3.64, "learning_rate": 1.1928916874787983e-05, "loss": 0.0426, "step": 19852 }, { "epoch": 3.64, "learning_rate": 1.1928188382241703e-05, "loss": 0.0408, "step": 19853 }, { "epoch": 3.64, "learning_rate": 1.192745987906721e-05, "loss": 0.0406, "step": 19854 }, { "epoch": 3.64, "learning_rate": 1.1926731365268507e-05, "loss": 0.0429, "step": 19855 }, { "epoch": 3.64, "learning_rate": 1.192600284084962e-05, "loss": 0.0384, "step": 19856 }, { "epoch": 3.64, "learning_rate": 1.1925274305814559e-05, "loss": 0.0337, "step": 19857 }, { "epoch": 3.64, "learning_rate": 1.1924545760167339e-05, "loss": 0.039, "step": 19858 }, { "epoch": 3.64, "learning_rate": 1.192381720391198e-05, "loss": 0.0444, "step": 19859 }, { "epoch": 3.64, "learning_rate": 1.1923088637052494e-05, "loss": 0.0396, "step": 19860 }, { "epoch": 3.64, "learning_rate": 1.19223600595929e-05, "loss": 0.0404, "step": 19861 }, { "epoch": 3.64, "learning_rate": 1.192163147153721e-05, "loss": 0.0444, "step": 19862 }, { "epoch": 3.64, "learning_rate": 1.1920902872889445e-05, "loss": 0.0483, "step": 19863 }, { "epoch": 3.64, "learning_rate": 1.1920174263653615e-05, "loss": 0.0488, "step": 19864 }, { "epoch": 3.64, "learning_rate": 1.191944564383374e-05, "loss": 0.04, "step": 19865 }, { "epoch": 3.64, "learning_rate": 1.1918717013433839e-05, "loss": 0.048, "step": 19866 }, { "epoch": 3.64, "learning_rate": 1.1917988372457923e-05, "loss": 0.0439, "step": 19867 }, { "epoch": 3.64, "learning_rate": 1.191725972091001e-05, "loss": 0.048, "step": 19868 }, { "epoch": 3.64, "learning_rate": 1.1916531058794117e-05, "loss": 0.0488, "step": 19869 }, { "epoch": 3.64, "learning_rate": 1.1915802386114259e-05, "loss": 0.0396, "step": 19870 }, { "epoch": 3.64, "learning_rate": 1.1915073702874452e-05, "loss": 0.0488, "step": 19871 }, { "epoch": 3.64, "learning_rate": 1.1914345009078718e-05, "loss": 0.0483, "step": 19872 }, { "epoch": 3.64, "learning_rate": 1.1913616304731064e-05, "loss": 0.0396, "step": 19873 }, { "epoch": 3.64, "learning_rate": 1.1912887589835516e-05, "loss": 0.0462, "step": 19874 }, { "epoch": 3.64, "learning_rate": 1.1912158864396087e-05, "loss": 0.0404, "step": 19875 }, { "epoch": 3.64, "learning_rate": 1.1911430128416793e-05, "loss": 0.0476, "step": 19876 }, { "epoch": 3.65, "learning_rate": 1.191070138190165e-05, "loss": 0.0463, "step": 19877 }, { "epoch": 3.65, "learning_rate": 1.1909972624854678e-05, "loss": 0.0464, "step": 19878 }, { "epoch": 3.65, "learning_rate": 1.1909243857279888e-05, "loss": 0.0427, "step": 19879 }, { "epoch": 3.65, "learning_rate": 1.1908515079181308e-05, "loss": 0.0506, "step": 19880 }, { "epoch": 3.65, "learning_rate": 1.1907786290562946e-05, "loss": 0.0413, "step": 19881 }, { "epoch": 3.65, "learning_rate": 1.1907057491428818e-05, "loss": 0.0418, "step": 19882 }, { "epoch": 3.65, "learning_rate": 1.1906328681782946e-05, "loss": 0.0472, "step": 19883 }, { "epoch": 3.65, "learning_rate": 1.1905599861629345e-05, "loss": 0.0391, "step": 19884 }, { "epoch": 3.65, "learning_rate": 1.1904871030972033e-05, "loss": 0.0577, "step": 19885 }, { "epoch": 3.65, "learning_rate": 1.1904142189815028e-05, "loss": 0.0421, "step": 19886 }, { "epoch": 3.65, "learning_rate": 1.1903413338162348e-05, "loss": 0.0355, "step": 19887 }, { "epoch": 3.65, "learning_rate": 1.1902684476018007e-05, "loss": 0.0472, "step": 19888 }, { "epoch": 3.65, "learning_rate": 1.1901955603386028e-05, "loss": 0.0435, "step": 19889 }, { "epoch": 3.65, "learning_rate": 1.190122672027042e-05, "loss": 0.0499, "step": 19890 }, { "epoch": 3.65, "learning_rate": 1.1900497826675207e-05, "loss": 0.0492, "step": 19891 }, { "epoch": 3.65, "learning_rate": 1.1899768922604409e-05, "loss": 0.0367, "step": 19892 }, { "epoch": 3.65, "learning_rate": 1.189904000806204e-05, "loss": 0.0451, "step": 19893 }, { "epoch": 3.65, "learning_rate": 1.1898311083052113e-05, "loss": 0.04, "step": 19894 }, { "epoch": 3.65, "learning_rate": 1.1897582147578654e-05, "loss": 0.0436, "step": 19895 }, { "epoch": 3.65, "learning_rate": 1.189685320164568e-05, "loss": 0.0433, "step": 19896 }, { "epoch": 3.65, "learning_rate": 1.18961242452572e-05, "loss": 0.0433, "step": 19897 }, { "epoch": 3.65, "learning_rate": 1.1895395278417247e-05, "loss": 0.0446, "step": 19898 }, { "epoch": 3.65, "learning_rate": 1.1894666301129825e-05, "loss": 0.0365, "step": 19899 }, { "epoch": 3.65, "learning_rate": 1.189393731339896e-05, "loss": 0.0447, "step": 19900 }, { "epoch": 3.65, "learning_rate": 1.1893208315228668e-05, "loss": 0.0514, "step": 19901 }, { "epoch": 3.65, "learning_rate": 1.1892479306622968e-05, "loss": 0.0408, "step": 19902 }, { "epoch": 3.65, "learning_rate": 1.1891750287585877e-05, "loss": 0.0456, "step": 19903 }, { "epoch": 3.65, "learning_rate": 1.1891021258121413e-05, "loss": 0.0423, "step": 19904 }, { "epoch": 3.65, "learning_rate": 1.1890292218233599e-05, "loss": 0.0415, "step": 19905 }, { "epoch": 3.65, "learning_rate": 1.1889563167926446e-05, "loss": 0.0493, "step": 19906 }, { "epoch": 3.65, "learning_rate": 1.188883410720398e-05, "loss": 0.0463, "step": 19907 }, { "epoch": 3.65, "learning_rate": 1.1888105036070213e-05, "loss": 0.0529, "step": 19908 }, { "epoch": 3.65, "learning_rate": 1.1887375954529167e-05, "loss": 0.0431, "step": 19909 }, { "epoch": 3.65, "learning_rate": 1.1886646862584863e-05, "loss": 0.0455, "step": 19910 }, { "epoch": 3.65, "learning_rate": 1.1885917760241317e-05, "loss": 0.0427, "step": 19911 }, { "epoch": 3.65, "learning_rate": 1.1885188647502546e-05, "loss": 0.0489, "step": 19912 }, { "epoch": 3.65, "learning_rate": 1.1884459524372574e-05, "loss": 0.0489, "step": 19913 }, { "epoch": 3.65, "learning_rate": 1.1883730390855415e-05, "loss": 0.0545, "step": 19914 }, { "epoch": 3.65, "learning_rate": 1.1883001246955089e-05, "loss": 0.042, "step": 19915 }, { "epoch": 3.65, "learning_rate": 1.1882272092675618e-05, "loss": 0.0506, "step": 19916 }, { "epoch": 3.65, "learning_rate": 1.1881542928021018e-05, "loss": 0.0538, "step": 19917 }, { "epoch": 3.65, "learning_rate": 1.1880813752995307e-05, "loss": 0.0369, "step": 19918 }, { "epoch": 3.65, "learning_rate": 1.188008456760251e-05, "loss": 0.0565, "step": 19919 }, { "epoch": 3.65, "learning_rate": 1.1879355371846644e-05, "loss": 0.0405, "step": 19920 }, { "epoch": 3.65, "learning_rate": 1.1878626165731721e-05, "loss": 0.0434, "step": 19921 }, { "epoch": 3.65, "learning_rate": 1.1877896949261773e-05, "loss": 0.0456, "step": 19922 }, { "epoch": 3.65, "learning_rate": 1.1877167722440813e-05, "loss": 0.0469, "step": 19923 }, { "epoch": 3.65, "learning_rate": 1.1876438485272854e-05, "loss": 0.0448, "step": 19924 }, { "epoch": 3.65, "learning_rate": 1.187570923776193e-05, "loss": 0.0394, "step": 19925 }, { "epoch": 3.65, "learning_rate": 1.1874979979912047e-05, "loss": 0.0412, "step": 19926 }, { "epoch": 3.65, "learning_rate": 1.1874250711727232e-05, "loss": 0.0409, "step": 19927 }, { "epoch": 3.65, "learning_rate": 1.1873521433211504e-05, "loss": 0.0489, "step": 19928 }, { "epoch": 3.65, "learning_rate": 1.187279214436888e-05, "loss": 0.05, "step": 19929 }, { "epoch": 3.65, "learning_rate": 1.1872062845203382e-05, "loss": 0.0481, "step": 19930 }, { "epoch": 3.66, "learning_rate": 1.1871333535719035e-05, "loss": 0.0515, "step": 19931 }, { "epoch": 3.66, "learning_rate": 1.1870604215919847e-05, "loss": 0.0413, "step": 19932 }, { "epoch": 3.66, "learning_rate": 1.1869874885809848e-05, "loss": 0.0423, "step": 19933 }, { "epoch": 3.66, "learning_rate": 1.1869145545393057e-05, "loss": 0.0428, "step": 19934 }, { "epoch": 3.66, "learning_rate": 1.186841619467349e-05, "loss": 0.0431, "step": 19935 }, { "epoch": 3.66, "learning_rate": 1.1867686833655168e-05, "loss": 0.0438, "step": 19936 }, { "epoch": 3.66, "learning_rate": 1.1866957462342112e-05, "loss": 0.0424, "step": 19937 }, { "epoch": 3.66, "learning_rate": 1.1866228080738345e-05, "loss": 0.0468, "step": 19938 }, { "epoch": 3.66, "learning_rate": 1.186549868884788e-05, "loss": 0.0437, "step": 19939 }, { "epoch": 3.66, "learning_rate": 1.1864769286674749e-05, "loss": 0.0454, "step": 19940 }, { "epoch": 3.66, "learning_rate": 1.1864039874222961e-05, "loss": 0.0441, "step": 19941 }, { "epoch": 3.66, "learning_rate": 1.1863310451496546e-05, "loss": 0.0427, "step": 19942 }, { "epoch": 3.66, "learning_rate": 1.1862581018499517e-05, "loss": 0.0452, "step": 19943 }, { "epoch": 3.66, "learning_rate": 1.1861851575235899e-05, "loss": 0.0472, "step": 19944 }, { "epoch": 3.66, "learning_rate": 1.1861122121709711e-05, "loss": 0.0476, "step": 19945 }, { "epoch": 3.66, "learning_rate": 1.1860392657924975e-05, "loss": 0.0502, "step": 19946 }, { "epoch": 3.66, "learning_rate": 1.1859663183885713e-05, "loss": 0.0392, "step": 19947 }, { "epoch": 3.66, "learning_rate": 1.1858933699595937e-05, "loss": 0.0498, "step": 19948 }, { "epoch": 3.66, "learning_rate": 1.1858204205059682e-05, "loss": 0.0413, "step": 19949 }, { "epoch": 3.66, "learning_rate": 1.1857474700280957e-05, "loss": 0.0459, "step": 19950 }, { "epoch": 3.66, "learning_rate": 1.1856745185263791e-05, "loss": 0.0468, "step": 19951 }, { "epoch": 3.66, "learning_rate": 1.1856015660012203e-05, "loss": 0.0486, "step": 19952 }, { "epoch": 3.66, "learning_rate": 1.1855286124530211e-05, "loss": 0.0542, "step": 19953 }, { "epoch": 3.66, "learning_rate": 1.185455657882184e-05, "loss": 0.0389, "step": 19954 }, { "epoch": 3.66, "learning_rate": 1.1853827022891108e-05, "loss": 0.0438, "step": 19955 }, { "epoch": 3.66, "learning_rate": 1.1853097456742039e-05, "loss": 0.0428, "step": 19956 }, { "epoch": 3.66, "learning_rate": 1.1852367880378654e-05, "loss": 0.0408, "step": 19957 }, { "epoch": 3.66, "learning_rate": 1.1851638293804973e-05, "loss": 0.0429, "step": 19958 }, { "epoch": 3.66, "learning_rate": 1.1850908697025017e-05, "loss": 0.0411, "step": 19959 }, { "epoch": 3.66, "learning_rate": 1.185017909004281e-05, "loss": 0.0394, "step": 19960 }, { "epoch": 3.66, "learning_rate": 1.1849449472862373e-05, "loss": 0.0467, "step": 19961 }, { "epoch": 3.66, "learning_rate": 1.1848719845487726e-05, "loss": 0.0486, "step": 19962 }, { "epoch": 3.66, "learning_rate": 1.1847990207922892e-05, "loss": 0.0503, "step": 19963 }, { "epoch": 3.66, "learning_rate": 1.1847260560171895e-05, "loss": 0.0485, "step": 19964 }, { "epoch": 3.66, "learning_rate": 1.1846530902238751e-05, "loss": 0.0504, "step": 19965 }, { "epoch": 3.66, "learning_rate": 1.184580123412749e-05, "loss": 0.0395, "step": 19966 }, { "epoch": 3.66, "learning_rate": 1.1845071555842128e-05, "loss": 0.0412, "step": 19967 }, { "epoch": 3.66, "learning_rate": 1.1844341867386685e-05, "loss": 0.0383, "step": 19968 }, { "epoch": 3.66, "learning_rate": 1.1843612168765188e-05, "loss": 0.0405, "step": 19969 }, { "epoch": 3.66, "learning_rate": 1.1842882459981658e-05, "loss": 0.0416, "step": 19970 }, { "epoch": 3.66, "learning_rate": 1.1842152741040117e-05, "loss": 0.0431, "step": 19971 }, { "epoch": 3.66, "learning_rate": 1.1841423011944585e-05, "loss": 0.0431, "step": 19972 }, { "epoch": 3.66, "learning_rate": 1.184069327269909e-05, "loss": 0.0322, "step": 19973 }, { "epoch": 3.66, "learning_rate": 1.1839963523307647e-05, "loss": 0.0461, "step": 19974 }, { "epoch": 3.66, "learning_rate": 1.1839233763774285e-05, "loss": 0.0417, "step": 19975 }, { "epoch": 3.66, "learning_rate": 1.1838503994103024e-05, "loss": 0.0444, "step": 19976 }, { "epoch": 3.66, "learning_rate": 1.1837774214297881e-05, "loss": 0.0398, "step": 19977 }, { "epoch": 3.66, "learning_rate": 1.1837044424362887e-05, "loss": 0.046, "step": 19978 }, { "epoch": 3.66, "learning_rate": 1.1836314624302058e-05, "loss": 0.045, "step": 19979 }, { "epoch": 3.66, "learning_rate": 1.1835584814119422e-05, "loss": 0.0401, "step": 19980 }, { "epoch": 3.66, "learning_rate": 1.1834854993819e-05, "loss": 0.0403, "step": 19981 }, { "epoch": 3.66, "learning_rate": 1.1834125163404815e-05, "loss": 0.0446, "step": 19982 }, { "epoch": 3.66, "learning_rate": 1.1833395322880888e-05, "loss": 0.0465, "step": 19983 }, { "epoch": 3.66, "learning_rate": 1.1832665472251243e-05, "loss": 0.0522, "step": 19984 }, { "epoch": 3.66, "learning_rate": 1.1831935611519902e-05, "loss": 0.0385, "step": 19985 }, { "epoch": 3.67, "learning_rate": 1.183120574069089e-05, "loss": 0.0459, "step": 19986 }, { "epoch": 3.67, "learning_rate": 1.183047585976823e-05, "loss": 0.0445, "step": 19987 }, { "epoch": 3.67, "learning_rate": 1.1829745968755943e-05, "loss": 0.0424, "step": 19988 }, { "epoch": 3.67, "learning_rate": 1.1829016067658054e-05, "loss": 0.0477, "step": 19989 }, { "epoch": 3.67, "learning_rate": 1.1828286156478586e-05, "loss": 0.0476, "step": 19990 }, { "epoch": 3.67, "learning_rate": 1.182755623522156e-05, "loss": 0.0469, "step": 19991 }, { "epoch": 3.67, "learning_rate": 1.1826826303891001e-05, "loss": 0.0426, "step": 19992 }, { "epoch": 3.67, "learning_rate": 1.1826096362490936e-05, "loss": 0.0458, "step": 19993 }, { "epoch": 3.67, "learning_rate": 1.1825366411025381e-05, "loss": 0.0476, "step": 19994 }, { "epoch": 3.67, "learning_rate": 1.1824636449498366e-05, "loss": 0.044, "step": 19995 }, { "epoch": 3.67, "learning_rate": 1.182390647791391e-05, "loss": 0.0437, "step": 19996 }, { "epoch": 3.67, "learning_rate": 1.1823176496276043e-05, "loss": 0.0482, "step": 19997 }, { "epoch": 3.67, "learning_rate": 1.1822446504588783e-05, "loss": 0.0465, "step": 19998 }, { "epoch": 3.67, "learning_rate": 1.1821716502856154e-05, "loss": 0.0442, "step": 19999 }, { "epoch": 3.67, "learning_rate": 1.182098649108218e-05, "loss": 0.0464, "step": 20000 }, { "epoch": 3.67, "learning_rate": 1.1820256469270885e-05, "loss": 0.0341, "step": 20001 }, { "epoch": 3.67, "learning_rate": 1.1819526437426298e-05, "loss": 0.0456, "step": 20002 }, { "epoch": 3.67, "learning_rate": 1.1818796395552435e-05, "loss": 0.0417, "step": 20003 }, { "epoch": 3.67, "learning_rate": 1.1818066343653323e-05, "loss": 0.0458, "step": 20004 }, { "epoch": 3.67, "learning_rate": 1.181733628173299e-05, "loss": 0.0492, "step": 20005 }, { "epoch": 3.67, "learning_rate": 1.1816606209795456e-05, "loss": 0.0491, "step": 20006 }, { "epoch": 3.67, "learning_rate": 1.1815876127844744e-05, "loss": 0.0484, "step": 20007 }, { "epoch": 3.67, "learning_rate": 1.1815146035884883e-05, "loss": 0.0585, "step": 20008 }, { "epoch": 3.67, "learning_rate": 1.1814415933919894e-05, "loss": 0.0398, "step": 20009 }, { "epoch": 3.67, "learning_rate": 1.1813685821953798e-05, "loss": 0.0367, "step": 20010 }, { "epoch": 3.67, "learning_rate": 1.1812955699990628e-05, "loss": 0.0455, "step": 20011 }, { "epoch": 3.67, "learning_rate": 1.1812225568034401e-05, "loss": 0.0405, "step": 20012 }, { "epoch": 3.67, "learning_rate": 1.1811495426089146e-05, "loss": 0.0461, "step": 20013 }, { "epoch": 3.67, "learning_rate": 1.1810765274158884e-05, "loss": 0.0521, "step": 20014 }, { "epoch": 3.67, "learning_rate": 1.1810035112247644e-05, "loss": 0.0453, "step": 20015 }, { "epoch": 3.67, "learning_rate": 1.1809304940359442e-05, "loss": 0.0499, "step": 20016 }, { "epoch": 3.67, "learning_rate": 1.1808574758498317e-05, "loss": 0.0469, "step": 20017 }, { "epoch": 3.67, "learning_rate": 1.180784456666828e-05, "loss": 0.0468, "step": 20018 }, { "epoch": 3.67, "learning_rate": 1.1807114364873363e-05, "loss": 0.0413, "step": 20019 }, { "epoch": 3.67, "learning_rate": 1.180638415311759e-05, "loss": 0.0467, "step": 20020 }, { "epoch": 3.67, "learning_rate": 1.1805653931404985e-05, "loss": 0.0479, "step": 20021 }, { "epoch": 3.67, "learning_rate": 1.1804923699739572e-05, "loss": 0.0398, "step": 20022 }, { "epoch": 3.67, "learning_rate": 1.180419345812538e-05, "loss": 0.0417, "step": 20023 }, { "epoch": 3.67, "learning_rate": 1.180346320656643e-05, "loss": 0.0425, "step": 20024 }, { "epoch": 3.67, "learning_rate": 1.1802732945066745e-05, "loss": 0.0373, "step": 20025 }, { "epoch": 3.67, "learning_rate": 1.180200267363036e-05, "loss": 0.0495, "step": 20026 }, { "epoch": 3.67, "learning_rate": 1.180127239226129e-05, "loss": 0.0444, "step": 20027 }, { "epoch": 3.67, "learning_rate": 1.1800542100963564e-05, "loss": 0.0469, "step": 20028 }, { "epoch": 3.67, "learning_rate": 1.1799811799741209e-05, "loss": 0.0437, "step": 20029 }, { "epoch": 3.67, "learning_rate": 1.179908148859825e-05, "loss": 0.0409, "step": 20030 }, { "epoch": 3.67, "learning_rate": 1.179835116753871e-05, "loss": 0.0403, "step": 20031 }, { "epoch": 3.67, "learning_rate": 1.1797620836566619e-05, "loss": 0.0426, "step": 20032 }, { "epoch": 3.67, "learning_rate": 1.1796890495685998e-05, "loss": 0.0472, "step": 20033 }, { "epoch": 3.67, "learning_rate": 1.1796160144900874e-05, "loss": 0.0442, "step": 20034 }, { "epoch": 3.67, "learning_rate": 1.1795429784215276e-05, "loss": 0.0501, "step": 20035 }, { "epoch": 3.67, "learning_rate": 1.1794699413633222e-05, "loss": 0.0522, "step": 20036 }, { "epoch": 3.67, "learning_rate": 1.1793969033158746e-05, "loss": 0.0412, "step": 20037 }, { "epoch": 3.67, "learning_rate": 1.179323864279587e-05, "loss": 0.0406, "step": 20038 }, { "epoch": 3.67, "learning_rate": 1.179250824254862e-05, "loss": 0.0503, "step": 20039 }, { "epoch": 3.68, "learning_rate": 1.1791777832421022e-05, "loss": 0.0388, "step": 20040 }, { "epoch": 3.68, "learning_rate": 1.1791047412417104e-05, "loss": 0.0516, "step": 20041 }, { "epoch": 3.68, "learning_rate": 1.1790316982540892e-05, "loss": 0.0373, "step": 20042 }, { "epoch": 3.68, "learning_rate": 1.178958654279641e-05, "loss": 0.0451, "step": 20043 }, { "epoch": 3.68, "learning_rate": 1.1788856093187686e-05, "loss": 0.046, "step": 20044 }, { "epoch": 3.68, "learning_rate": 1.1788125633718741e-05, "loss": 0.0342, "step": 20045 }, { "epoch": 3.68, "learning_rate": 1.178739516439361e-05, "loss": 0.044, "step": 20046 }, { "epoch": 3.68, "learning_rate": 1.1786664685216312e-05, "loss": 0.0446, "step": 20047 }, { "epoch": 3.68, "learning_rate": 1.1785934196190877e-05, "loss": 0.0456, "step": 20048 }, { "epoch": 3.68, "learning_rate": 1.1785203697321333e-05, "loss": 0.0402, "step": 20049 }, { "epoch": 3.68, "learning_rate": 1.1784473188611703e-05, "loss": 0.0449, "step": 20050 }, { "epoch": 3.68, "learning_rate": 1.1783742670066016e-05, "loss": 0.0406, "step": 20051 }, { "epoch": 3.68, "learning_rate": 1.1783012141688297e-05, "loss": 0.0418, "step": 20052 }, { "epoch": 3.68, "learning_rate": 1.1782281603482573e-05, "loss": 0.0495, "step": 20053 }, { "epoch": 3.68, "learning_rate": 1.1781551055452869e-05, "loss": 0.0445, "step": 20054 }, { "epoch": 3.68, "learning_rate": 1.1780820497603215e-05, "loss": 0.0407, "step": 20055 }, { "epoch": 3.68, "learning_rate": 1.178008992993764e-05, "loss": 0.0479, "step": 20056 }, { "epoch": 3.68, "learning_rate": 1.1779359352460164e-05, "loss": 0.0563, "step": 20057 }, { "epoch": 3.68, "learning_rate": 1.177862876517482e-05, "loss": 0.0393, "step": 20058 }, { "epoch": 3.68, "learning_rate": 1.177789816808563e-05, "loss": 0.0488, "step": 20059 }, { "epoch": 3.68, "learning_rate": 1.1777167561196623e-05, "loss": 0.0469, "step": 20060 }, { "epoch": 3.68, "learning_rate": 1.1776436944511831e-05, "loss": 0.044, "step": 20061 }, { "epoch": 3.68, "learning_rate": 1.1775706318035276e-05, "loss": 0.0464, "step": 20062 }, { "epoch": 3.68, "learning_rate": 1.1774975681770981e-05, "loss": 0.0531, "step": 20063 }, { "epoch": 3.68, "learning_rate": 1.1774245035722984e-05, "loss": 0.0443, "step": 20064 }, { "epoch": 3.68, "learning_rate": 1.1773514379895305e-05, "loss": 0.0403, "step": 20065 }, { "epoch": 3.68, "learning_rate": 1.1772783714291973e-05, "loss": 0.0488, "step": 20066 }, { "epoch": 3.68, "learning_rate": 1.1772053038917017e-05, "loss": 0.0474, "step": 20067 }, { "epoch": 3.68, "learning_rate": 1.1771322353774461e-05, "loss": 0.0411, "step": 20068 }, { "epoch": 3.68, "learning_rate": 1.1770591658868336e-05, "loss": 0.0436, "step": 20069 }, { "epoch": 3.68, "learning_rate": 1.176986095420267e-05, "loss": 0.0517, "step": 20070 }, { "epoch": 3.68, "learning_rate": 1.1769130239781486e-05, "loss": 0.0431, "step": 20071 }, { "epoch": 3.68, "learning_rate": 1.1768399515608814e-05, "loss": 0.048, "step": 20072 }, { "epoch": 3.68, "learning_rate": 1.1767668781688685e-05, "loss": 0.0484, "step": 20073 }, { "epoch": 3.68, "learning_rate": 1.1766938038025125e-05, "loss": 0.0448, "step": 20074 }, { "epoch": 3.68, "learning_rate": 1.1766207284622157e-05, "loss": 0.0422, "step": 20075 }, { "epoch": 3.68, "learning_rate": 1.1765476521483817e-05, "loss": 0.0491, "step": 20076 }, { "epoch": 3.68, "learning_rate": 1.1764745748614128e-05, "loss": 0.0475, "step": 20077 }, { "epoch": 3.68, "learning_rate": 1.1764014966017117e-05, "loss": 0.0449, "step": 20078 }, { "epoch": 3.68, "learning_rate": 1.1763284173696819e-05, "loss": 0.0459, "step": 20079 }, { "epoch": 3.68, "learning_rate": 1.176255337165725e-05, "loss": 0.0391, "step": 20080 }, { "epoch": 3.68, "learning_rate": 1.1761822559902451e-05, "loss": 0.0384, "step": 20081 }, { "epoch": 3.68, "learning_rate": 1.1761091738436442e-05, "loss": 0.0428, "step": 20082 }, { "epoch": 3.68, "learning_rate": 1.1760360907263257e-05, "loss": 0.0552, "step": 20083 }, { "epoch": 3.68, "learning_rate": 1.1759630066386921e-05, "loss": 0.0387, "step": 20084 }, { "epoch": 3.68, "learning_rate": 1.1758899215811461e-05, "loss": 0.0482, "step": 20085 }, { "epoch": 3.68, "learning_rate": 1.1758168355540908e-05, "loss": 0.045, "step": 20086 }, { "epoch": 3.68, "learning_rate": 1.1757437485579289e-05, "loss": 0.043, "step": 20087 }, { "epoch": 3.68, "learning_rate": 1.1756706605930637e-05, "loss": 0.0477, "step": 20088 }, { "epoch": 3.68, "learning_rate": 1.1755975716598973e-05, "loss": 0.0485, "step": 20089 }, { "epoch": 3.68, "learning_rate": 1.1755244817588331e-05, "loss": 0.0401, "step": 20090 }, { "epoch": 3.68, "learning_rate": 1.1754513908902739e-05, "loss": 0.0478, "step": 20091 }, { "epoch": 3.68, "learning_rate": 1.1753782990546226e-05, "loss": 0.0462, "step": 20092 }, { "epoch": 3.68, "learning_rate": 1.1753052062522817e-05, "loss": 0.0432, "step": 20093 }, { "epoch": 3.68, "learning_rate": 1.1752321124836548e-05, "loss": 0.0521, "step": 20094 }, { "epoch": 3.69, "learning_rate": 1.1751590177491441e-05, "loss": 0.0399, "step": 20095 }, { "epoch": 3.69, "learning_rate": 1.1750859220491528e-05, "loss": 0.0353, "step": 20096 }, { "epoch": 3.69, "learning_rate": 1.175012825384084e-05, "loss": 0.0378, "step": 20097 }, { "epoch": 3.69, "learning_rate": 1.1749397277543401e-05, "loss": 0.0478, "step": 20098 }, { "epoch": 3.69, "learning_rate": 1.1748666291603246e-05, "loss": 0.0392, "step": 20099 }, { "epoch": 3.69, "learning_rate": 1.17479352960244e-05, "loss": 0.0414, "step": 20100 }, { "epoch": 3.69, "learning_rate": 1.1747204290810894e-05, "loss": 0.0478, "step": 20101 }, { "epoch": 3.69, "learning_rate": 1.1746473275966754e-05, "loss": 0.0531, "step": 20102 }, { "epoch": 3.69, "learning_rate": 1.174574225149602e-05, "loss": 0.0426, "step": 20103 }, { "epoch": 3.69, "learning_rate": 1.1745011217402709e-05, "loss": 0.0435, "step": 20104 }, { "epoch": 3.69, "learning_rate": 1.1744280173690854e-05, "loss": 0.0355, "step": 20105 }, { "epoch": 3.69, "learning_rate": 1.174354912036449e-05, "loss": 0.0495, "step": 20106 }, { "epoch": 3.69, "learning_rate": 1.1742818057427636e-05, "loss": 0.0491, "step": 20107 }, { "epoch": 3.69, "learning_rate": 1.1742086984884333e-05, "loss": 0.0432, "step": 20108 }, { "epoch": 3.69, "learning_rate": 1.1741355902738605e-05, "loss": 0.0386, "step": 20109 }, { "epoch": 3.69, "learning_rate": 1.1740624810994483e-05, "loss": 0.0511, "step": 20110 }, { "epoch": 3.69, "learning_rate": 1.1739893709655992e-05, "loss": 0.0442, "step": 20111 }, { "epoch": 3.69, "learning_rate": 1.1739162598727172e-05, "loss": 0.0436, "step": 20112 }, { "epoch": 3.69, "learning_rate": 1.1738431478212043e-05, "loss": 0.0443, "step": 20113 }, { "epoch": 3.69, "learning_rate": 1.173770034811464e-05, "loss": 0.0494, "step": 20114 }, { "epoch": 3.69, "learning_rate": 1.1736969208438996e-05, "loss": 0.0458, "step": 20115 }, { "epoch": 3.69, "learning_rate": 1.1736238059189131e-05, "loss": 0.0463, "step": 20116 }, { "epoch": 3.69, "learning_rate": 1.1735506900369082e-05, "loss": 0.042, "step": 20117 }, { "epoch": 3.69, "learning_rate": 1.1734775731982881e-05, "loss": 0.0475, "step": 20118 }, { "epoch": 3.69, "learning_rate": 1.1734044554034556e-05, "loss": 0.0438, "step": 20119 }, { "epoch": 3.69, "learning_rate": 1.1733313366528135e-05, "loss": 0.0439, "step": 20120 }, { "epoch": 3.69, "learning_rate": 1.173258216946765e-05, "loss": 0.0376, "step": 20121 }, { "epoch": 3.69, "learning_rate": 1.173185096285713e-05, "loss": 0.0497, "step": 20122 }, { "epoch": 3.69, "learning_rate": 1.1731119746700608e-05, "loss": 0.0416, "step": 20123 }, { "epoch": 3.69, "learning_rate": 1.1730388521002115e-05, "loss": 0.0476, "step": 20124 }, { "epoch": 3.69, "learning_rate": 1.1729657285765678e-05, "loss": 0.0429, "step": 20125 }, { "epoch": 3.69, "learning_rate": 1.1728926040995332e-05, "loss": 0.0474, "step": 20126 }, { "epoch": 3.69, "learning_rate": 1.1728194786695103e-05, "loss": 0.0423, "step": 20127 }, { "epoch": 3.69, "learning_rate": 1.1727463522869025e-05, "loss": 0.0454, "step": 20128 }, { "epoch": 3.69, "learning_rate": 1.1726732249521129e-05, "loss": 0.0392, "step": 20129 }, { "epoch": 3.69, "learning_rate": 1.1726000966655442e-05, "loss": 0.0479, "step": 20130 }, { "epoch": 3.69, "learning_rate": 1.1725269674275996e-05, "loss": 0.04, "step": 20131 }, { "epoch": 3.69, "learning_rate": 1.1724538372386824e-05, "loss": 0.0462, "step": 20132 }, { "epoch": 3.69, "learning_rate": 1.1723807060991958e-05, "loss": 0.04, "step": 20133 }, { "epoch": 3.69, "learning_rate": 1.1723075740095427e-05, "loss": 0.043, "step": 20134 }, { "epoch": 3.69, "learning_rate": 1.1722344409701261e-05, "loss": 0.0405, "step": 20135 }, { "epoch": 3.69, "learning_rate": 1.1721613069813493e-05, "loss": 0.0458, "step": 20136 }, { "epoch": 3.69, "learning_rate": 1.1720881720436153e-05, "loss": 0.0486, "step": 20137 }, { "epoch": 3.69, "learning_rate": 1.1720150361573273e-05, "loss": 0.043, "step": 20138 }, { "epoch": 3.69, "learning_rate": 1.1719418993228883e-05, "loss": 0.0442, "step": 20139 }, { "epoch": 3.69, "learning_rate": 1.1718687615407014e-05, "loss": 0.0463, "step": 20140 }, { "epoch": 3.69, "learning_rate": 1.17179562281117e-05, "loss": 0.0402, "step": 20141 }, { "epoch": 3.69, "learning_rate": 1.1717224831346972e-05, "loss": 0.0469, "step": 20142 }, { "epoch": 3.69, "learning_rate": 1.1716493425116859e-05, "loss": 0.0447, "step": 20143 }, { "epoch": 3.69, "learning_rate": 1.1715762009425395e-05, "loss": 0.0413, "step": 20144 }, { "epoch": 3.69, "learning_rate": 1.1715030584276613e-05, "loss": 0.0514, "step": 20145 }, { "epoch": 3.69, "learning_rate": 1.1714299149674538e-05, "loss": 0.0405, "step": 20146 }, { "epoch": 3.69, "learning_rate": 1.171356770562321e-05, "loss": 0.0372, "step": 20147 }, { "epoch": 3.69, "learning_rate": 1.1712836252126654e-05, "loss": 0.0418, "step": 20148 }, { "epoch": 3.7, "learning_rate": 1.1712104789188903e-05, "loss": 0.0521, "step": 20149 }, { "epoch": 3.7, "learning_rate": 1.1711373316813994e-05, "loss": 0.0441, "step": 20150 }, { "epoch": 3.7, "learning_rate": 1.1710641835005952e-05, "loss": 0.0456, "step": 20151 }, { "epoch": 3.7, "learning_rate": 1.1709910343768813e-05, "loss": 0.0385, "step": 20152 }, { "epoch": 3.7, "learning_rate": 1.1709178843106608e-05, "loss": 0.0404, "step": 20153 }, { "epoch": 3.7, "learning_rate": 1.170844733302337e-05, "loss": 0.0435, "step": 20154 }, { "epoch": 3.7, "learning_rate": 1.1707715813523129e-05, "loss": 0.0397, "step": 20155 }, { "epoch": 3.7, "learning_rate": 1.170698428460992e-05, "loss": 0.0433, "step": 20156 }, { "epoch": 3.7, "learning_rate": 1.1706252746287772e-05, "loss": 0.0365, "step": 20157 }, { "epoch": 3.7, "learning_rate": 1.170552119856072e-05, "loss": 0.0421, "step": 20158 }, { "epoch": 3.7, "learning_rate": 1.1704789641432796e-05, "loss": 0.047, "step": 20159 }, { "epoch": 3.7, "learning_rate": 1.1704058074908028e-05, "loss": 0.0364, "step": 20160 }, { "epoch": 3.7, "learning_rate": 1.1703326498990456e-05, "loss": 0.0464, "step": 20161 }, { "epoch": 3.7, "learning_rate": 1.1702594913684109e-05, "loss": 0.0373, "step": 20162 }, { "epoch": 3.7, "learning_rate": 1.1701863318993017e-05, "loss": 0.041, "step": 20163 }, { "epoch": 3.7, "learning_rate": 1.1701131714921212e-05, "loss": 0.0515, "step": 20164 }, { "epoch": 3.7, "learning_rate": 1.1700400101472733e-05, "loss": 0.0433, "step": 20165 }, { "epoch": 3.7, "learning_rate": 1.1699668478651607e-05, "loss": 0.0475, "step": 20166 }, { "epoch": 3.7, "learning_rate": 1.1698936846461868e-05, "loss": 0.0465, "step": 20167 }, { "epoch": 3.7, "learning_rate": 1.1698205204907554e-05, "loss": 0.0435, "step": 20168 }, { "epoch": 3.7, "learning_rate": 1.1697473553992686e-05, "loss": 0.0494, "step": 20169 }, { "epoch": 3.7, "learning_rate": 1.1696741893721308e-05, "loss": 0.0473, "step": 20170 }, { "epoch": 3.7, "learning_rate": 1.1696010224097449e-05, "loss": 0.0423, "step": 20171 }, { "epoch": 3.7, "learning_rate": 1.1695278545125142e-05, "loss": 0.0381, "step": 20172 }, { "epoch": 3.7, "learning_rate": 1.1694546856808416e-05, "loss": 0.043, "step": 20173 }, { "epoch": 3.7, "learning_rate": 1.1693815159151313e-05, "loss": 0.034, "step": 20174 }, { "epoch": 3.7, "learning_rate": 1.1693083452157858e-05, "loss": 0.0373, "step": 20175 }, { "epoch": 3.7, "learning_rate": 1.1692351735832088e-05, "loss": 0.0453, "step": 20176 }, { "epoch": 3.7, "learning_rate": 1.1691620010178037e-05, "loss": 0.0481, "step": 20177 }, { "epoch": 3.7, "learning_rate": 1.1690888275199736e-05, "loss": 0.0485, "step": 20178 }, { "epoch": 3.7, "learning_rate": 1.169015653090122e-05, "loss": 0.0483, "step": 20179 }, { "epoch": 3.7, "learning_rate": 1.1689424777286522e-05, "loss": 0.0425, "step": 20180 }, { "epoch": 3.7, "learning_rate": 1.1688693014359673e-05, "loss": 0.0385, "step": 20181 }, { "epoch": 3.7, "learning_rate": 1.168796124212471e-05, "loss": 0.0363, "step": 20182 }, { "epoch": 3.7, "learning_rate": 1.1687229460585666e-05, "loss": 0.0391, "step": 20183 }, { "epoch": 3.7, "learning_rate": 1.168649766974657e-05, "loss": 0.0436, "step": 20184 }, { "epoch": 3.7, "learning_rate": 1.1685765869611463e-05, "loss": 0.0423, "step": 20185 }, { "epoch": 3.7, "learning_rate": 1.1685034060184373e-05, "loss": 0.0414, "step": 20186 }, { "epoch": 3.7, "learning_rate": 1.1684302241469337e-05, "loss": 0.0431, "step": 20187 }, { "epoch": 3.7, "learning_rate": 1.1683570413470384e-05, "loss": 0.0429, "step": 20188 }, { "epoch": 3.7, "learning_rate": 1.1682838576191557e-05, "loss": 0.039, "step": 20189 }, { "epoch": 3.7, "learning_rate": 1.1682106729636884e-05, "loss": 0.0537, "step": 20190 }, { "epoch": 3.7, "learning_rate": 1.1681374873810393e-05, "loss": 0.04, "step": 20191 }, { "epoch": 3.7, "learning_rate": 1.1680643008716132e-05, "loss": 0.0439, "step": 20192 }, { "epoch": 3.7, "learning_rate": 1.1679911134358121e-05, "loss": 0.0466, "step": 20193 }, { "epoch": 3.7, "learning_rate": 1.1679179250740405e-05, "loss": 0.0387, "step": 20194 }, { "epoch": 3.7, "learning_rate": 1.167844735786701e-05, "loss": 0.0425, "step": 20195 }, { "epoch": 3.7, "learning_rate": 1.1677715455741979e-05, "loss": 0.041, "step": 20196 }, { "epoch": 3.7, "learning_rate": 1.1676983544369337e-05, "loss": 0.0481, "step": 20197 }, { "epoch": 3.7, "learning_rate": 1.1676251623753124e-05, "loss": 0.0469, "step": 20198 }, { "epoch": 3.7, "learning_rate": 1.167551969389737e-05, "loss": 0.0438, "step": 20199 }, { "epoch": 3.7, "learning_rate": 1.1674787754806114e-05, "loss": 0.0458, "step": 20200 }, { "epoch": 3.7, "learning_rate": 1.167405580648339e-05, "loss": 0.0458, "step": 20201 }, { "epoch": 3.7, "learning_rate": 1.1673323848933229e-05, "loss": 0.0462, "step": 20202 }, { "epoch": 3.7, "learning_rate": 1.1672591882159669e-05, "loss": 0.0472, "step": 20203 }, { "epoch": 3.71, "learning_rate": 1.1671859906166745e-05, "loss": 0.0447, "step": 20204 }, { "epoch": 3.71, "learning_rate": 1.1671127920958488e-05, "loss": 0.0451, "step": 20205 }, { "epoch": 3.71, "learning_rate": 1.1670395926538935e-05, "loss": 0.0388, "step": 20206 }, { "epoch": 3.71, "learning_rate": 1.166966392291212e-05, "loss": 0.0396, "step": 20207 }, { "epoch": 3.71, "learning_rate": 1.1668931910082078e-05, "loss": 0.051, "step": 20208 }, { "epoch": 3.71, "learning_rate": 1.1668199888052844e-05, "loss": 0.0416, "step": 20209 }, { "epoch": 3.71, "learning_rate": 1.1667467856828456e-05, "loss": 0.0444, "step": 20210 }, { "epoch": 3.71, "learning_rate": 1.1666735816412944e-05, "loss": 0.0427, "step": 20211 }, { "epoch": 3.71, "learning_rate": 1.1666003766810345e-05, "loss": 0.0598, "step": 20212 }, { "epoch": 3.71, "learning_rate": 1.1665271708024695e-05, "loss": 0.0477, "step": 20213 }, { "epoch": 3.71, "learning_rate": 1.1664539640060028e-05, "loss": 0.0457, "step": 20214 }, { "epoch": 3.71, "learning_rate": 1.166380756292038e-05, "loss": 0.0451, "step": 20215 }, { "epoch": 3.71, "learning_rate": 1.1663075476609786e-05, "loss": 0.0414, "step": 20216 }, { "epoch": 3.71, "learning_rate": 1.1662343381132278e-05, "loss": 0.0475, "step": 20217 }, { "epoch": 3.71, "learning_rate": 1.1661611276491898e-05, "loss": 0.0414, "step": 20218 }, { "epoch": 3.71, "learning_rate": 1.1660879162692676e-05, "loss": 0.0461, "step": 20219 }, { "epoch": 3.71, "learning_rate": 1.166014703973865e-05, "loss": 0.0431, "step": 20220 }, { "epoch": 3.71, "learning_rate": 1.1659414907633852e-05, "loss": 0.0437, "step": 20221 }, { "epoch": 3.71, "learning_rate": 1.1658682766382324e-05, "loss": 0.0437, "step": 20222 }, { "epoch": 3.71, "learning_rate": 1.1657950615988096e-05, "loss": 0.0448, "step": 20223 }, { "epoch": 3.71, "learning_rate": 1.1657218456455207e-05, "loss": 0.0328, "step": 20224 }, { "epoch": 3.71, "learning_rate": 1.1656486287787686e-05, "loss": 0.0445, "step": 20225 }, { "epoch": 3.71, "learning_rate": 1.1655754109989576e-05, "loss": 0.0412, "step": 20226 }, { "epoch": 3.71, "learning_rate": 1.1655021923064915e-05, "loss": 0.0548, "step": 20227 }, { "epoch": 3.71, "learning_rate": 1.165428972701773e-05, "loss": 0.0434, "step": 20228 }, { "epoch": 3.71, "learning_rate": 1.1653557521852061e-05, "loss": 0.0396, "step": 20229 }, { "epoch": 3.71, "learning_rate": 1.1652825307571947e-05, "loss": 0.0479, "step": 20230 }, { "epoch": 3.71, "learning_rate": 1.165209308418142e-05, "loss": 0.0346, "step": 20231 }, { "epoch": 3.71, "learning_rate": 1.1651360851684517e-05, "loss": 0.0517, "step": 20232 }, { "epoch": 3.71, "learning_rate": 1.1650628610085275e-05, "loss": 0.0445, "step": 20233 }, { "epoch": 3.71, "learning_rate": 1.1649896359387729e-05, "loss": 0.0343, "step": 20234 }, { "epoch": 3.71, "learning_rate": 1.1649164099595913e-05, "loss": 0.0445, "step": 20235 }, { "epoch": 3.71, "learning_rate": 1.1648431830713868e-05, "loss": 0.0413, "step": 20236 }, { "epoch": 3.71, "learning_rate": 1.1647699552745628e-05, "loss": 0.0429, "step": 20237 }, { "epoch": 3.71, "learning_rate": 1.164696726569523e-05, "loss": 0.0485, "step": 20238 }, { "epoch": 3.71, "learning_rate": 1.164623496956671e-05, "loss": 0.0433, "step": 20239 }, { "epoch": 3.71, "learning_rate": 1.1645502664364103e-05, "loss": 0.0386, "step": 20240 }, { "epoch": 3.71, "learning_rate": 1.1644770350091448e-05, "loss": 0.0522, "step": 20241 }, { "epoch": 3.71, "learning_rate": 1.1644038026752781e-05, "loss": 0.0378, "step": 20242 }, { "epoch": 3.71, "learning_rate": 1.1643305694352136e-05, "loss": 0.0504, "step": 20243 }, { "epoch": 3.71, "learning_rate": 1.1642573352893552e-05, "loss": 0.0373, "step": 20244 }, { "epoch": 3.71, "learning_rate": 1.1641841002381067e-05, "loss": 0.0559, "step": 20245 }, { "epoch": 3.71, "learning_rate": 1.1641108642818712e-05, "loss": 0.0467, "step": 20246 }, { "epoch": 3.71, "learning_rate": 1.1640376274210532e-05, "loss": 0.04, "step": 20247 }, { "epoch": 3.71, "learning_rate": 1.1639643896560557e-05, "loss": 0.0359, "step": 20248 }, { "epoch": 3.71, "learning_rate": 1.1638911509872827e-05, "loss": 0.0374, "step": 20249 }, { "epoch": 3.71, "learning_rate": 1.1638179114151378e-05, "loss": 0.0459, "step": 20250 }, { "epoch": 3.71, "learning_rate": 1.1637446709400248e-05, "loss": 0.0461, "step": 20251 }, { "epoch": 3.71, "learning_rate": 1.1636714295623472e-05, "loss": 0.0436, "step": 20252 }, { "epoch": 3.71, "learning_rate": 1.163598187282509e-05, "loss": 0.0482, "step": 20253 }, { "epoch": 3.71, "learning_rate": 1.163524944100914e-05, "loss": 0.0398, "step": 20254 }, { "epoch": 3.71, "learning_rate": 1.1634517000179652e-05, "loss": 0.0371, "step": 20255 }, { "epoch": 3.71, "learning_rate": 1.163378455034067e-05, "loss": 0.0405, "step": 20256 }, { "epoch": 3.71, "learning_rate": 1.163305209149623e-05, "loss": 0.0395, "step": 20257 }, { "epoch": 3.72, "learning_rate": 1.1632319623650368e-05, "loss": 0.0442, "step": 20258 }, { "epoch": 3.72, "learning_rate": 1.163158714680712e-05, "loss": 0.0411, "step": 20259 }, { "epoch": 3.72, "learning_rate": 1.163085466097053e-05, "loss": 0.0438, "step": 20260 }, { "epoch": 3.72, "learning_rate": 1.1630122166144626e-05, "loss": 0.0439, "step": 20261 }, { "epoch": 3.72, "learning_rate": 1.1629389662333453e-05, "loss": 0.0468, "step": 20262 }, { "epoch": 3.72, "learning_rate": 1.1628657149541044e-05, "loss": 0.0498, "step": 20263 }, { "epoch": 3.72, "learning_rate": 1.1627924627771443e-05, "loss": 0.0446, "step": 20264 }, { "epoch": 3.72, "learning_rate": 1.162719209702868e-05, "loss": 0.0424, "step": 20265 }, { "epoch": 3.72, "learning_rate": 1.1626459557316796e-05, "loss": 0.048, "step": 20266 }, { "epoch": 3.72, "learning_rate": 1.1625727008639831e-05, "loss": 0.0296, "step": 20267 }, { "epoch": 3.72, "learning_rate": 1.1624994451001817e-05, "loss": 0.052, "step": 20268 }, { "epoch": 3.72, "learning_rate": 1.16242618844068e-05, "loss": 0.0347, "step": 20269 }, { "epoch": 3.72, "learning_rate": 1.162352930885881e-05, "loss": 0.0491, "step": 20270 }, { "epoch": 3.72, "learning_rate": 1.162279672436189e-05, "loss": 0.056, "step": 20271 }, { "epoch": 3.72, "learning_rate": 1.1622064130920074e-05, "loss": 0.0358, "step": 20272 }, { "epoch": 3.72, "learning_rate": 1.1621331528537405e-05, "loss": 0.0451, "step": 20273 }, { "epoch": 3.72, "learning_rate": 1.1620598917217917e-05, "loss": 0.0378, "step": 20274 }, { "epoch": 3.72, "learning_rate": 1.1619866296965654e-05, "loss": 0.047, "step": 20275 }, { "epoch": 3.72, "learning_rate": 1.1619133667784645e-05, "loss": 0.0423, "step": 20276 }, { "epoch": 3.72, "learning_rate": 1.1618401029678935e-05, "loss": 0.041, "step": 20277 }, { "epoch": 3.72, "learning_rate": 1.1617668382652563e-05, "loss": 0.0486, "step": 20278 }, { "epoch": 3.72, "learning_rate": 1.1616935726709562e-05, "loss": 0.0385, "step": 20279 }, { "epoch": 3.72, "learning_rate": 1.1616203061853972e-05, "loss": 0.048, "step": 20280 }, { "epoch": 3.72, "learning_rate": 1.1615470388089836e-05, "loss": 0.042, "step": 20281 }, { "epoch": 3.72, "learning_rate": 1.1614737705421189e-05, "loss": 0.0374, "step": 20282 }, { "epoch": 3.72, "learning_rate": 1.161400501385207e-05, "loss": 0.0489, "step": 20283 }, { "epoch": 3.72, "learning_rate": 1.1613272313386516e-05, "loss": 0.0442, "step": 20284 }, { "epoch": 3.72, "learning_rate": 1.1612539604028567e-05, "loss": 0.0427, "step": 20285 }, { "epoch": 3.72, "learning_rate": 1.1611806885782264e-05, "loss": 0.0437, "step": 20286 }, { "epoch": 3.72, "learning_rate": 1.1611074158651645e-05, "loss": 0.0494, "step": 20287 }, { "epoch": 3.72, "learning_rate": 1.1610341422640743e-05, "loss": 0.0433, "step": 20288 }, { "epoch": 3.72, "learning_rate": 1.1609608677753602e-05, "loss": 0.0366, "step": 20289 }, { "epoch": 3.72, "learning_rate": 1.1608875923994262e-05, "loss": 0.0413, "step": 20290 }, { "epoch": 3.72, "learning_rate": 1.160814316136676e-05, "loss": 0.0444, "step": 20291 }, { "epoch": 3.72, "learning_rate": 1.1607410389875138e-05, "loss": 0.0431, "step": 20292 }, { "epoch": 3.72, "learning_rate": 1.1606677609523429e-05, "loss": 0.0497, "step": 20293 }, { "epoch": 3.72, "learning_rate": 1.1605944820315674e-05, "loss": 0.0443, "step": 20294 }, { "epoch": 3.72, "learning_rate": 1.1605212022255916e-05, "loss": 0.0419, "step": 20295 }, { "epoch": 3.72, "learning_rate": 1.1604479215348192e-05, "loss": 0.0479, "step": 20296 }, { "epoch": 3.72, "learning_rate": 1.1603746399596539e-05, "loss": 0.0391, "step": 20297 }, { "epoch": 3.72, "learning_rate": 1.1603013575005e-05, "loss": 0.0411, "step": 20298 }, { "epoch": 3.72, "learning_rate": 1.1602280741577612e-05, "loss": 0.0402, "step": 20299 }, { "epoch": 3.72, "learning_rate": 1.1601547899318415e-05, "loss": 0.038, "step": 20300 }, { "epoch": 3.72, "learning_rate": 1.1600815048231449e-05, "loss": 0.0448, "step": 20301 }, { "epoch": 3.72, "learning_rate": 1.1600082188320752e-05, "loss": 0.0438, "step": 20302 }, { "epoch": 3.72, "learning_rate": 1.1599349319590362e-05, "loss": 0.0475, "step": 20303 }, { "epoch": 3.72, "learning_rate": 1.1598616442044328e-05, "loss": 0.0455, "step": 20304 }, { "epoch": 3.72, "learning_rate": 1.1597883555686678e-05, "loss": 0.0404, "step": 20305 }, { "epoch": 3.72, "learning_rate": 1.1597150660521457e-05, "loss": 0.0441, "step": 20306 }, { "epoch": 3.72, "learning_rate": 1.1596417756552704e-05, "loss": 0.0449, "step": 20307 }, { "epoch": 3.72, "learning_rate": 1.1595684843784458e-05, "loss": 0.0454, "step": 20308 }, { "epoch": 3.72, "learning_rate": 1.1594951922220763e-05, "loss": 0.0383, "step": 20309 }, { "epoch": 3.72, "learning_rate": 1.1594218991865653e-05, "loss": 0.0405, "step": 20310 }, { "epoch": 3.72, "learning_rate": 1.1593486052723171e-05, "loss": 0.0383, "step": 20311 }, { "epoch": 3.72, "learning_rate": 1.1592753104797356e-05, "loss": 0.0475, "step": 20312 }, { "epoch": 3.73, "learning_rate": 1.1592020148092252e-05, "loss": 0.0393, "step": 20313 }, { "epoch": 3.73, "learning_rate": 1.159128718261189e-05, "loss": 0.0398, "step": 20314 }, { "epoch": 3.73, "learning_rate": 1.1590554208360318e-05, "loss": 0.0536, "step": 20315 }, { "epoch": 3.73, "learning_rate": 1.1589821225341575e-05, "loss": 0.0472, "step": 20316 }, { "epoch": 3.73, "learning_rate": 1.15890882335597e-05, "loss": 0.0492, "step": 20317 }, { "epoch": 3.73, "learning_rate": 1.1588355233018736e-05, "loss": 0.0457, "step": 20318 }, { "epoch": 3.73, "learning_rate": 1.1587622223722717e-05, "loss": 0.0489, "step": 20319 }, { "epoch": 3.73, "learning_rate": 1.1586889205675689e-05, "loss": 0.0364, "step": 20320 }, { "epoch": 3.73, "learning_rate": 1.1586156178881687e-05, "loss": 0.0505, "step": 20321 }, { "epoch": 3.73, "learning_rate": 1.158542314334476e-05, "loss": 0.0431, "step": 20322 }, { "epoch": 3.73, "learning_rate": 1.1584690099068938e-05, "loss": 0.0406, "step": 20323 }, { "epoch": 3.73, "learning_rate": 1.1583957046058272e-05, "loss": 0.0447, "step": 20324 }, { "epoch": 3.73, "learning_rate": 1.1583223984316795e-05, "loss": 0.0384, "step": 20325 }, { "epoch": 3.73, "learning_rate": 1.1582490913848552e-05, "loss": 0.0478, "step": 20326 }, { "epoch": 3.73, "learning_rate": 1.158175783465758e-05, "loss": 0.0514, "step": 20327 }, { "epoch": 3.73, "learning_rate": 1.1581024746747925e-05, "loss": 0.041, "step": 20328 }, { "epoch": 3.73, "learning_rate": 1.158029165012362e-05, "loss": 0.0479, "step": 20329 }, { "epoch": 3.73, "learning_rate": 1.1579558544788712e-05, "loss": 0.0385, "step": 20330 }, { "epoch": 3.73, "learning_rate": 1.1578825430747245e-05, "loss": 0.0422, "step": 20331 }, { "epoch": 3.73, "learning_rate": 1.1578092308003248e-05, "loss": 0.0381, "step": 20332 }, { "epoch": 3.73, "learning_rate": 1.1577359176560774e-05, "loss": 0.0457, "step": 20333 }, { "epoch": 3.73, "learning_rate": 1.1576626036423856e-05, "loss": 0.0414, "step": 20334 }, { "epoch": 3.73, "learning_rate": 1.1575892887596542e-05, "loss": 0.0437, "step": 20335 }, { "epoch": 3.73, "learning_rate": 1.1575159730082866e-05, "loss": 0.0475, "step": 20336 }, { "epoch": 3.73, "learning_rate": 1.1574426563886875e-05, "loss": 0.0494, "step": 20337 }, { "epoch": 3.73, "learning_rate": 1.1573693389012605e-05, "loss": 0.0451, "step": 20338 }, { "epoch": 3.73, "learning_rate": 1.1572960205464106e-05, "loss": 0.0385, "step": 20339 }, { "epoch": 3.73, "learning_rate": 1.1572227013245411e-05, "loss": 0.0434, "step": 20340 }, { "epoch": 3.73, "learning_rate": 1.1571493812360562e-05, "loss": 0.0381, "step": 20341 }, { "epoch": 3.73, "learning_rate": 1.1570760602813603e-05, "loss": 0.0382, "step": 20342 }, { "epoch": 3.73, "learning_rate": 1.1570027384608572e-05, "loss": 0.0469, "step": 20343 }, { "epoch": 3.73, "learning_rate": 1.1569294157749518e-05, "loss": 0.0408, "step": 20344 }, { "epoch": 3.73, "learning_rate": 1.1568560922240474e-05, "loss": 0.0461, "step": 20345 }, { "epoch": 3.73, "learning_rate": 1.156782767808549e-05, "loss": 0.0363, "step": 20346 }, { "epoch": 3.73, "learning_rate": 1.1567094425288599e-05, "loss": 0.0403, "step": 20347 }, { "epoch": 3.73, "learning_rate": 1.1566361163853848e-05, "loss": 0.038, "step": 20348 }, { "epoch": 3.73, "learning_rate": 1.1565627893785278e-05, "loss": 0.0469, "step": 20349 }, { "epoch": 3.73, "learning_rate": 1.156489461508693e-05, "loss": 0.0353, "step": 20350 }, { "epoch": 3.73, "learning_rate": 1.1564161327762848e-05, "loss": 0.0467, "step": 20351 }, { "epoch": 3.73, "learning_rate": 1.156342803181707e-05, "loss": 0.0487, "step": 20352 }, { "epoch": 3.73, "learning_rate": 1.1562694727253641e-05, "loss": 0.0421, "step": 20353 }, { "epoch": 3.73, "learning_rate": 1.1561961414076598e-05, "loss": 0.0469, "step": 20354 }, { "epoch": 3.73, "learning_rate": 1.1561228092289996e-05, "loss": 0.0345, "step": 20355 }, { "epoch": 3.73, "learning_rate": 1.156049476189786e-05, "loss": 0.0468, "step": 20356 }, { "epoch": 3.73, "learning_rate": 1.1559761422904244e-05, "loss": 0.05, "step": 20357 }, { "epoch": 3.73, "learning_rate": 1.1559028075313186e-05, "loss": 0.0404, "step": 20358 }, { "epoch": 3.73, "learning_rate": 1.1558294719128729e-05, "loss": 0.0438, "step": 20359 }, { "epoch": 3.73, "learning_rate": 1.1557561354354913e-05, "loss": 0.0416, "step": 20360 }, { "epoch": 3.73, "learning_rate": 1.1556827980995787e-05, "loss": 0.037, "step": 20361 }, { "epoch": 3.73, "learning_rate": 1.1556094599055385e-05, "loss": 0.0394, "step": 20362 }, { "epoch": 3.73, "learning_rate": 1.1555361208537753e-05, "loss": 0.0374, "step": 20363 }, { "epoch": 3.73, "learning_rate": 1.1554627809446934e-05, "loss": 0.0439, "step": 20364 }, { "epoch": 3.73, "learning_rate": 1.155389440178697e-05, "loss": 0.0432, "step": 20365 }, { "epoch": 3.73, "learning_rate": 1.1553160985561905e-05, "loss": 0.0389, "step": 20366 }, { "epoch": 3.74, "learning_rate": 1.1552427560775778e-05, "loss": 0.0379, "step": 20367 }, { "epoch": 3.74, "learning_rate": 1.1551694127432636e-05, "loss": 0.0447, "step": 20368 }, { "epoch": 3.74, "learning_rate": 1.1550960685536518e-05, "loss": 0.0469, "step": 20369 }, { "epoch": 3.74, "learning_rate": 1.155022723509147e-05, "loss": 0.0431, "step": 20370 }, { "epoch": 3.74, "learning_rate": 1.1549493776101532e-05, "loss": 0.0426, "step": 20371 }, { "epoch": 3.74, "learning_rate": 1.1548760308570746e-05, "loss": 0.0413, "step": 20372 }, { "epoch": 3.74, "learning_rate": 1.1548026832503162e-05, "loss": 0.0455, "step": 20373 }, { "epoch": 3.74, "learning_rate": 1.1547293347902813e-05, "loss": 0.0328, "step": 20374 }, { "epoch": 3.74, "learning_rate": 1.154655985477375e-05, "loss": 0.0494, "step": 20375 }, { "epoch": 3.74, "learning_rate": 1.1545826353120011e-05, "loss": 0.0415, "step": 20376 }, { "epoch": 3.74, "learning_rate": 1.1545092842945643e-05, "loss": 0.0467, "step": 20377 }, { "epoch": 3.74, "learning_rate": 1.1544359324254685e-05, "loss": 0.0447, "step": 20378 }, { "epoch": 3.74, "learning_rate": 1.1543625797051184e-05, "loss": 0.0441, "step": 20379 }, { "epoch": 3.74, "learning_rate": 1.1542892261339177e-05, "loss": 0.0505, "step": 20380 }, { "epoch": 3.74, "learning_rate": 1.1542158717122718e-05, "loss": 0.0412, "step": 20381 }, { "epoch": 3.74, "learning_rate": 1.1541425164405842e-05, "loss": 0.0461, "step": 20382 }, { "epoch": 3.74, "learning_rate": 1.154069160319259e-05, "loss": 0.0373, "step": 20383 }, { "epoch": 3.74, "learning_rate": 1.1539958033487015e-05, "loss": 0.0399, "step": 20384 }, { "epoch": 3.74, "learning_rate": 1.1539224455293154e-05, "loss": 0.0441, "step": 20385 }, { "epoch": 3.74, "learning_rate": 1.1538490868615051e-05, "loss": 0.0525, "step": 20386 }, { "epoch": 3.74, "learning_rate": 1.1537757273456751e-05, "loss": 0.0428, "step": 20387 }, { "epoch": 3.74, "learning_rate": 1.1537023669822295e-05, "loss": 0.0454, "step": 20388 }, { "epoch": 3.74, "learning_rate": 1.1536290057715728e-05, "loss": 0.0492, "step": 20389 }, { "epoch": 3.74, "learning_rate": 1.15355564371411e-05, "loss": 0.0412, "step": 20390 }, { "epoch": 3.74, "learning_rate": 1.1534822808102444e-05, "loss": 0.0377, "step": 20391 }, { "epoch": 3.74, "learning_rate": 1.1534089170603808e-05, "loss": 0.0424, "step": 20392 }, { "epoch": 3.74, "learning_rate": 1.1533355524649238e-05, "loss": 0.0388, "step": 20393 }, { "epoch": 3.74, "learning_rate": 1.1532621870242778e-05, "loss": 0.0451, "step": 20394 }, { "epoch": 3.74, "learning_rate": 1.153188820738847e-05, "loss": 0.0516, "step": 20395 }, { "epoch": 3.74, "learning_rate": 1.1531154536090357e-05, "loss": 0.0482, "step": 20396 }, { "epoch": 3.74, "learning_rate": 1.1530420856352485e-05, "loss": 0.0417, "step": 20397 }, { "epoch": 3.74, "learning_rate": 1.1529687168178895e-05, "loss": 0.0421, "step": 20398 }, { "epoch": 3.74, "learning_rate": 1.1528953471573639e-05, "loss": 0.0436, "step": 20399 }, { "epoch": 3.74, "learning_rate": 1.1528219766540749e-05, "loss": 0.0426, "step": 20400 }, { "epoch": 3.74, "learning_rate": 1.1527486053084281e-05, "loss": 0.048, "step": 20401 }, { "epoch": 3.74, "learning_rate": 1.1526752331208271e-05, "loss": 0.0466, "step": 20402 }, { "epoch": 3.74, "learning_rate": 1.1526018600916769e-05, "loss": 0.0434, "step": 20403 }, { "epoch": 3.74, "learning_rate": 1.1525284862213815e-05, "loss": 0.0479, "step": 20404 }, { "epoch": 3.74, "learning_rate": 1.1524551115103455e-05, "loss": 0.0382, "step": 20405 }, { "epoch": 3.74, "learning_rate": 1.1523817359589734e-05, "loss": 0.0482, "step": 20406 }, { "epoch": 3.74, "learning_rate": 1.1523083595676696e-05, "loss": 0.0362, "step": 20407 }, { "epoch": 3.74, "learning_rate": 1.1522349823368385e-05, "loss": 0.0408, "step": 20408 }, { "epoch": 3.74, "learning_rate": 1.1521616042668846e-05, "loss": 0.0433, "step": 20409 }, { "epoch": 3.74, "learning_rate": 1.1520882253582126e-05, "loss": 0.0423, "step": 20410 }, { "epoch": 3.74, "learning_rate": 1.1520148456112266e-05, "loss": 0.043, "step": 20411 }, { "epoch": 3.74, "learning_rate": 1.1519414650263313e-05, "loss": 0.0448, "step": 20412 }, { "epoch": 3.74, "learning_rate": 1.1518680836039307e-05, "loss": 0.0468, "step": 20413 }, { "epoch": 3.74, "learning_rate": 1.1517947013444302e-05, "loss": 0.0383, "step": 20414 }, { "epoch": 3.74, "learning_rate": 1.1517213182482335e-05, "loss": 0.0553, "step": 20415 }, { "epoch": 3.74, "learning_rate": 1.151647934315745e-05, "loss": 0.0409, "step": 20416 }, { "epoch": 3.74, "learning_rate": 1.15157454954737e-05, "loss": 0.0448, "step": 20417 }, { "epoch": 3.74, "learning_rate": 1.1515011639435124e-05, "loss": 0.0457, "step": 20418 }, { "epoch": 3.74, "learning_rate": 1.1514277775045768e-05, "loss": 0.0452, "step": 20419 }, { "epoch": 3.74, "learning_rate": 1.1513543902309679e-05, "loss": 0.0393, "step": 20420 }, { "epoch": 3.74, "learning_rate": 1.1512810021230898e-05, "loss": 0.0433, "step": 20421 }, { "epoch": 3.75, "learning_rate": 1.1512076131813472e-05, "loss": 0.0418, "step": 20422 }, { "epoch": 3.75, "learning_rate": 1.151134223406145e-05, "loss": 0.0503, "step": 20423 }, { "epoch": 3.75, "learning_rate": 1.1510608327978873e-05, "loss": 0.0467, "step": 20424 }, { "epoch": 3.75, "learning_rate": 1.1509874413569788e-05, "loss": 0.0415, "step": 20425 }, { "epoch": 3.75, "learning_rate": 1.1509140490838241e-05, "loss": 0.0425, "step": 20426 }, { "epoch": 3.75, "learning_rate": 1.1508406559788274e-05, "loss": 0.0458, "step": 20427 }, { "epoch": 3.75, "learning_rate": 1.1507672620423936e-05, "loss": 0.0545, "step": 20428 }, { "epoch": 3.75, "learning_rate": 1.150693867274927e-05, "loss": 0.0453, "step": 20429 }, { "epoch": 3.75, "learning_rate": 1.1506204716768324e-05, "loss": 0.0506, "step": 20430 }, { "epoch": 3.75, "learning_rate": 1.1505470752485139e-05, "loss": 0.0322, "step": 20431 }, { "epoch": 3.75, "learning_rate": 1.1504736779903769e-05, "loss": 0.0433, "step": 20432 }, { "epoch": 3.75, "learning_rate": 1.150400279902825e-05, "loss": 0.0475, "step": 20433 }, { "epoch": 3.75, "learning_rate": 1.1503268809862636e-05, "loss": 0.043, "step": 20434 }, { "epoch": 3.75, "learning_rate": 1.150253481241097e-05, "loss": 0.0351, "step": 20435 }, { "epoch": 3.75, "learning_rate": 1.1501800806677293e-05, "loss": 0.0482, "step": 20436 }, { "epoch": 3.75, "learning_rate": 1.1501066792665656e-05, "loss": 0.047, "step": 20437 }, { "epoch": 3.75, "learning_rate": 1.1500332770380103e-05, "loss": 0.0445, "step": 20438 }, { "epoch": 3.75, "learning_rate": 1.1499598739824682e-05, "loss": 0.0458, "step": 20439 }, { "epoch": 3.75, "learning_rate": 1.1498864701003436e-05, "loss": 0.0403, "step": 20440 }, { "epoch": 3.75, "learning_rate": 1.1498130653920416e-05, "loss": 0.0402, "step": 20441 }, { "epoch": 3.75, "learning_rate": 1.149739659857966e-05, "loss": 0.0386, "step": 20442 }, { "epoch": 3.75, "learning_rate": 1.1496662534985224e-05, "loss": 0.0432, "step": 20443 }, { "epoch": 3.75, "learning_rate": 1.1495928463141145e-05, "loss": 0.0456, "step": 20444 }, { "epoch": 3.75, "learning_rate": 1.1495194383051475e-05, "loss": 0.0403, "step": 20445 }, { "epoch": 3.75, "learning_rate": 1.149446029472026e-05, "loss": 0.0445, "step": 20446 }, { "epoch": 3.75, "learning_rate": 1.1493726198151543e-05, "loss": 0.0547, "step": 20447 }, { "epoch": 3.75, "learning_rate": 1.1492992093349371e-05, "loss": 0.041, "step": 20448 }, { "epoch": 3.75, "learning_rate": 1.1492257980317792e-05, "loss": 0.0481, "step": 20449 }, { "epoch": 3.75, "learning_rate": 1.1491523859060855e-05, "loss": 0.0428, "step": 20450 }, { "epoch": 3.75, "learning_rate": 1.1490789729582602e-05, "loss": 0.0493, "step": 20451 }, { "epoch": 3.75, "learning_rate": 1.149005559188708e-05, "loss": 0.0411, "step": 20452 }, { "epoch": 3.75, "learning_rate": 1.1489321445978337e-05, "loss": 0.0357, "step": 20453 }, { "epoch": 3.75, "learning_rate": 1.148858729186042e-05, "loss": 0.0399, "step": 20454 }, { "epoch": 3.75, "learning_rate": 1.1487853129537374e-05, "loss": 0.0447, "step": 20455 }, { "epoch": 3.75, "learning_rate": 1.1487118959013249e-05, "loss": 0.0356, "step": 20456 }, { "epoch": 3.75, "learning_rate": 1.1486384780292088e-05, "loss": 0.0399, "step": 20457 }, { "epoch": 3.75, "learning_rate": 1.1485650593377939e-05, "loss": 0.0434, "step": 20458 }, { "epoch": 3.75, "learning_rate": 1.1484916398274847e-05, "loss": 0.045, "step": 20459 }, { "epoch": 3.75, "learning_rate": 1.1484182194986864e-05, "loss": 0.0511, "step": 20460 }, { "epoch": 3.75, "learning_rate": 1.1483447983518032e-05, "loss": 0.0425, "step": 20461 }, { "epoch": 3.75, "learning_rate": 1.1482713763872401e-05, "loss": 0.0415, "step": 20462 }, { "epoch": 3.75, "learning_rate": 1.1481979536054017e-05, "loss": 0.0402, "step": 20463 }, { "epoch": 3.75, "learning_rate": 1.1481245300066928e-05, "loss": 0.0505, "step": 20464 }, { "epoch": 3.75, "learning_rate": 1.148051105591518e-05, "loss": 0.0392, "step": 20465 }, { "epoch": 3.75, "learning_rate": 1.1479776803602818e-05, "loss": 0.0457, "step": 20466 }, { "epoch": 3.75, "learning_rate": 1.1479042543133895e-05, "loss": 0.0355, "step": 20467 }, { "epoch": 3.75, "learning_rate": 1.1478308274512453e-05, "loss": 0.0433, "step": 20468 }, { "epoch": 3.75, "learning_rate": 1.1477573997742541e-05, "loss": 0.0476, "step": 20469 }, { "epoch": 3.75, "learning_rate": 1.1476839712828207e-05, "loss": 0.0448, "step": 20470 }, { "epoch": 3.75, "learning_rate": 1.1476105419773498e-05, "loss": 0.0477, "step": 20471 }, { "epoch": 3.75, "learning_rate": 1.147537111858246e-05, "loss": 0.045, "step": 20472 }, { "epoch": 3.75, "learning_rate": 1.1474636809259146e-05, "loss": 0.0395, "step": 20473 }, { "epoch": 3.75, "learning_rate": 1.1473902491807596e-05, "loss": 0.0346, "step": 20474 }, { "epoch": 3.75, "learning_rate": 1.1473168166231859e-05, "loss": 0.0441, "step": 20475 }, { "epoch": 3.75, "learning_rate": 1.1472433832535988e-05, "loss": 0.0376, "step": 20476 }, { "epoch": 3.76, "learning_rate": 1.1471699490724025e-05, "loss": 0.0374, "step": 20477 }, { "epoch": 3.76, "learning_rate": 1.1470965140800022e-05, "loss": 0.0461, "step": 20478 }, { "epoch": 3.76, "learning_rate": 1.1470230782768023e-05, "loss": 0.0365, "step": 20479 }, { "epoch": 3.76, "learning_rate": 1.1469496416632078e-05, "loss": 0.0443, "step": 20480 }, { "epoch": 3.76, "learning_rate": 1.1468762042396236e-05, "loss": 0.0443, "step": 20481 }, { "epoch": 3.76, "learning_rate": 1.146802766006454e-05, "loss": 0.0459, "step": 20482 }, { "epoch": 3.76, "learning_rate": 1.1467293269641045e-05, "loss": 0.0385, "step": 20483 }, { "epoch": 3.76, "learning_rate": 1.146655887112979e-05, "loss": 0.047, "step": 20484 }, { "epoch": 3.76, "learning_rate": 1.1465824464534834e-05, "loss": 0.0452, "step": 20485 }, { "epoch": 3.76, "learning_rate": 1.1465090049860214e-05, "loss": 0.0438, "step": 20486 }, { "epoch": 3.76, "learning_rate": 1.1464355627109987e-05, "loss": 0.0378, "step": 20487 }, { "epoch": 3.76, "learning_rate": 1.1463621196288195e-05, "loss": 0.0454, "step": 20488 }, { "epoch": 3.76, "learning_rate": 1.146288675739889e-05, "loss": 0.0404, "step": 20489 }, { "epoch": 3.76, "learning_rate": 1.1462152310446117e-05, "loss": 0.0484, "step": 20490 }, { "epoch": 3.76, "learning_rate": 1.146141785543393e-05, "loss": 0.0425, "step": 20491 }, { "epoch": 3.76, "learning_rate": 1.1460683392366372e-05, "loss": 0.0432, "step": 20492 }, { "epoch": 3.76, "learning_rate": 1.145994892124749e-05, "loss": 0.0422, "step": 20493 }, { "epoch": 3.76, "learning_rate": 1.1459214442081339e-05, "loss": 0.037, "step": 20494 }, { "epoch": 3.76, "learning_rate": 1.1458479954871962e-05, "loss": 0.0469, "step": 20495 }, { "epoch": 3.76, "learning_rate": 1.145774545962341e-05, "loss": 0.0433, "step": 20496 }, { "epoch": 3.76, "learning_rate": 1.1457010956339731e-05, "loss": 0.047, "step": 20497 }, { "epoch": 3.76, "learning_rate": 1.1456276445024975e-05, "loss": 0.0514, "step": 20498 }, { "epoch": 3.76, "learning_rate": 1.1455541925683185e-05, "loss": 0.0427, "step": 20499 }, { "epoch": 3.76, "learning_rate": 1.1454807398318417e-05, "loss": 0.0417, "step": 20500 }, { "epoch": 3.76, "learning_rate": 1.1454072862934715e-05, "loss": 0.05, "step": 20501 }, { "epoch": 3.76, "learning_rate": 1.1453338319536129e-05, "loss": 0.0457, "step": 20502 }, { "epoch": 3.76, "learning_rate": 1.1452603768126711e-05, "loss": 0.0478, "step": 20503 }, { "epoch": 3.76, "learning_rate": 1.1451869208710504e-05, "loss": 0.0418, "step": 20504 }, { "epoch": 3.76, "learning_rate": 1.145113464129156e-05, "loss": 0.0425, "step": 20505 }, { "epoch": 3.76, "learning_rate": 1.145040006587393e-05, "loss": 0.0434, "step": 20506 }, { "epoch": 3.76, "learning_rate": 1.144966548246166e-05, "loss": 0.041, "step": 20507 }, { "epoch": 3.76, "learning_rate": 1.1448930891058798e-05, "loss": 0.044, "step": 20508 }, { "epoch": 3.76, "learning_rate": 1.1448196291669398e-05, "loss": 0.0483, "step": 20509 }, { "epoch": 3.76, "learning_rate": 1.1447461684297505e-05, "loss": 0.0467, "step": 20510 }, { "epoch": 3.76, "learning_rate": 1.1446727068947168e-05, "loss": 0.0462, "step": 20511 }, { "epoch": 3.76, "learning_rate": 1.1445992445622441e-05, "loss": 0.0515, "step": 20512 }, { "epoch": 3.76, "learning_rate": 1.1445257814327365e-05, "loss": 0.0479, "step": 20513 }, { "epoch": 3.76, "learning_rate": 1.1444523175065996e-05, "loss": 0.0442, "step": 20514 }, { "epoch": 3.76, "learning_rate": 1.1443788527842384e-05, "loss": 0.0427, "step": 20515 }, { "epoch": 3.76, "learning_rate": 1.1443053872660571e-05, "loss": 0.0439, "step": 20516 }, { "epoch": 3.76, "learning_rate": 1.1442319209524613e-05, "loss": 0.0441, "step": 20517 }, { "epoch": 3.76, "learning_rate": 1.1441584538438562e-05, "loss": 0.0427, "step": 20518 }, { "epoch": 3.76, "learning_rate": 1.1440849859406458e-05, "loss": 0.0451, "step": 20519 }, { "epoch": 3.76, "learning_rate": 1.1440115172432357e-05, "loss": 0.0526, "step": 20520 }, { "epoch": 3.76, "learning_rate": 1.143938047752031e-05, "loss": 0.0445, "step": 20521 }, { "epoch": 3.76, "learning_rate": 1.1438645774674359e-05, "loss": 0.0392, "step": 20522 }, { "epoch": 3.76, "learning_rate": 1.1437911063898564e-05, "loss": 0.0504, "step": 20523 }, { "epoch": 3.76, "learning_rate": 1.1437176345196967e-05, "loss": 0.0424, "step": 20524 }, { "epoch": 3.76, "learning_rate": 1.1436441618573621e-05, "loss": 0.0332, "step": 20525 }, { "epoch": 3.76, "learning_rate": 1.1435706884032574e-05, "loss": 0.0424, "step": 20526 }, { "epoch": 3.76, "learning_rate": 1.1434972141577881e-05, "loss": 0.0471, "step": 20527 }, { "epoch": 3.76, "learning_rate": 1.143423739121358e-05, "loss": 0.0407, "step": 20528 }, { "epoch": 3.76, "learning_rate": 1.1433502632943735e-05, "loss": 0.046, "step": 20529 }, { "epoch": 3.76, "learning_rate": 1.143276786677239e-05, "loss": 0.0392, "step": 20530 }, { "epoch": 3.77, "learning_rate": 1.1432033092703594e-05, "loss": 0.0449, "step": 20531 }, { "epoch": 3.77, "learning_rate": 1.1431298310741397e-05, "loss": 0.0512, "step": 20532 }, { "epoch": 3.77, "learning_rate": 1.143056352088985e-05, "loss": 0.0479, "step": 20533 }, { "epoch": 3.77, "learning_rate": 1.1429828723153005e-05, "loss": 0.0419, "step": 20534 }, { "epoch": 3.77, "learning_rate": 1.1429093917534907e-05, "loss": 0.0487, "step": 20535 }, { "epoch": 3.77, "learning_rate": 1.1428359104039617e-05, "loss": 0.0474, "step": 20536 }, { "epoch": 3.77, "learning_rate": 1.142762428267117e-05, "loss": 0.0387, "step": 20537 }, { "epoch": 3.77, "learning_rate": 1.1426889453433629e-05, "loss": 0.0377, "step": 20538 }, { "epoch": 3.77, "learning_rate": 1.142615461633104e-05, "loss": 0.0435, "step": 20539 }, { "epoch": 3.77, "learning_rate": 1.1425419771367454e-05, "loss": 0.0381, "step": 20540 }, { "epoch": 3.77, "learning_rate": 1.1424684918546918e-05, "loss": 0.0426, "step": 20541 }, { "epoch": 3.77, "learning_rate": 1.1423950057873488e-05, "loss": 0.0389, "step": 20542 }, { "epoch": 3.77, "learning_rate": 1.142321518935121e-05, "loss": 0.0358, "step": 20543 }, { "epoch": 3.77, "learning_rate": 1.1422480312984137e-05, "loss": 0.0402, "step": 20544 }, { "epoch": 3.77, "learning_rate": 1.142174542877632e-05, "loss": 0.0397, "step": 20545 }, { "epoch": 3.77, "learning_rate": 1.1421010536731807e-05, "loss": 0.0348, "step": 20546 }, { "epoch": 3.77, "learning_rate": 1.1420275636854652e-05, "loss": 0.0432, "step": 20547 }, { "epoch": 3.77, "learning_rate": 1.1419540729148903e-05, "loss": 0.0418, "step": 20548 }, { "epoch": 3.77, "learning_rate": 1.1418805813618614e-05, "loss": 0.0458, "step": 20549 }, { "epoch": 3.77, "learning_rate": 1.1418070890267834e-05, "loss": 0.0504, "step": 20550 }, { "epoch": 3.77, "learning_rate": 1.1417335959100614e-05, "loss": 0.0359, "step": 20551 }, { "epoch": 3.77, "learning_rate": 1.1416601020121005e-05, "loss": 0.043, "step": 20552 }, { "epoch": 3.77, "learning_rate": 1.1415866073333059e-05, "loss": 0.0439, "step": 20553 }, { "epoch": 3.77, "learning_rate": 1.1415131118740826e-05, "loss": 0.049, "step": 20554 }, { "epoch": 3.77, "learning_rate": 1.1414396156348353e-05, "loss": 0.0403, "step": 20555 }, { "epoch": 3.77, "learning_rate": 1.1413661186159697e-05, "loss": 0.0413, "step": 20556 }, { "epoch": 3.77, "learning_rate": 1.141292620817891e-05, "loss": 0.0428, "step": 20557 }, { "epoch": 3.77, "learning_rate": 1.1412191222410038e-05, "loss": 0.0488, "step": 20558 }, { "epoch": 3.77, "learning_rate": 1.1411456228857135e-05, "loss": 0.0497, "step": 20559 }, { "epoch": 3.77, "learning_rate": 1.1410721227524256e-05, "loss": 0.0413, "step": 20560 }, { "epoch": 3.77, "learning_rate": 1.1409986218415442e-05, "loss": 0.0395, "step": 20561 }, { "epoch": 3.77, "learning_rate": 1.140925120153476e-05, "loss": 0.0504, "step": 20562 }, { "epoch": 3.77, "learning_rate": 1.1408516176886246e-05, "loss": 0.0505, "step": 20563 }, { "epoch": 3.77, "learning_rate": 1.1407781144473957e-05, "loss": 0.0418, "step": 20564 }, { "epoch": 3.77, "learning_rate": 1.1407046104301948e-05, "loss": 0.0407, "step": 20565 }, { "epoch": 3.77, "learning_rate": 1.1406311056374266e-05, "loss": 0.0406, "step": 20566 }, { "epoch": 3.77, "learning_rate": 1.1405576000694968e-05, "loss": 0.0418, "step": 20567 }, { "epoch": 3.77, "learning_rate": 1.1404840937268098e-05, "loss": 0.0465, "step": 20568 }, { "epoch": 3.77, "learning_rate": 1.1404105866097714e-05, "loss": 0.0463, "step": 20569 }, { "epoch": 3.77, "learning_rate": 1.1403370787187863e-05, "loss": 0.0444, "step": 20570 }, { "epoch": 3.77, "learning_rate": 1.1402635700542606e-05, "loss": 0.0455, "step": 20571 }, { "epoch": 3.77, "learning_rate": 1.1401900606165981e-05, "loss": 0.0488, "step": 20572 }, { "epoch": 3.77, "learning_rate": 1.140116550406205e-05, "loss": 0.0432, "step": 20573 }, { "epoch": 3.77, "learning_rate": 1.1400430394234865e-05, "loss": 0.0409, "step": 20574 }, { "epoch": 3.77, "learning_rate": 1.1399695276688469e-05, "loss": 0.047, "step": 20575 }, { "epoch": 3.77, "learning_rate": 1.1398960151426924e-05, "loss": 0.044, "step": 20576 }, { "epoch": 3.77, "learning_rate": 1.1398225018454273e-05, "loss": 0.046, "step": 20577 }, { "epoch": 3.77, "learning_rate": 1.1397489877774579e-05, "loss": 0.0493, "step": 20578 }, { "epoch": 3.77, "learning_rate": 1.1396754729391883e-05, "loss": 0.0477, "step": 20579 }, { "epoch": 3.77, "learning_rate": 1.1396019573310247e-05, "loss": 0.0413, "step": 20580 }, { "epoch": 3.77, "learning_rate": 1.1395284409533713e-05, "loss": 0.0516, "step": 20581 }, { "epoch": 3.77, "learning_rate": 1.1394549238066341e-05, "loss": 0.037, "step": 20582 }, { "epoch": 3.77, "learning_rate": 1.1393814058912183e-05, "loss": 0.0504, "step": 20583 }, { "epoch": 3.77, "learning_rate": 1.1393078872075288e-05, "loss": 0.0442, "step": 20584 }, { "epoch": 3.77, "learning_rate": 1.139234367755971e-05, "loss": 0.0426, "step": 20585 }, { "epoch": 3.78, "learning_rate": 1.13916084753695e-05, "loss": 0.0498, "step": 20586 }, { "epoch": 3.78, "learning_rate": 1.1390873265508712e-05, "loss": 0.0434, "step": 20587 }, { "epoch": 3.78, "learning_rate": 1.1390138047981396e-05, "loss": 0.044, "step": 20588 }, { "epoch": 3.78, "learning_rate": 1.1389402822791612e-05, "loss": 0.0428, "step": 20589 }, { "epoch": 3.78, "learning_rate": 1.13886675899434e-05, "loss": 0.0438, "step": 20590 }, { "epoch": 3.78, "learning_rate": 1.1387932349440822e-05, "loss": 0.0487, "step": 20591 }, { "epoch": 3.78, "learning_rate": 1.138719710128793e-05, "loss": 0.0428, "step": 20592 }, { "epoch": 3.78, "learning_rate": 1.1386461845488776e-05, "loss": 0.0446, "step": 20593 }, { "epoch": 3.78, "learning_rate": 1.1385726582047407e-05, "loss": 0.0428, "step": 20594 }, { "epoch": 3.78, "learning_rate": 1.1384991310967888e-05, "loss": 0.0401, "step": 20595 }, { "epoch": 3.78, "learning_rate": 1.1384256032254259e-05, "loss": 0.0423, "step": 20596 }, { "epoch": 3.78, "learning_rate": 1.138352074591058e-05, "loss": 0.0468, "step": 20597 }, { "epoch": 3.78, "learning_rate": 1.1382785451940903e-05, "loss": 0.0425, "step": 20598 }, { "epoch": 3.78, "learning_rate": 1.1382050150349277e-05, "loss": 0.0459, "step": 20599 }, { "epoch": 3.78, "learning_rate": 1.1381314841139762e-05, "loss": 0.0362, "step": 20600 }, { "epoch": 3.78, "learning_rate": 1.1380579524316406e-05, "loss": 0.043, "step": 20601 }, { "epoch": 3.78, "learning_rate": 1.1379844199883263e-05, "loss": 0.0409, "step": 20602 }, { "epoch": 3.78, "learning_rate": 1.1379108867844384e-05, "loss": 0.045, "step": 20603 }, { "epoch": 3.78, "learning_rate": 1.137837352820383e-05, "loss": 0.0463, "step": 20604 }, { "epoch": 3.78, "learning_rate": 1.1377638180965645e-05, "loss": 0.0447, "step": 20605 }, { "epoch": 3.78, "learning_rate": 1.1376902826133885e-05, "loss": 0.0436, "step": 20606 }, { "epoch": 3.78, "learning_rate": 1.1376167463712608e-05, "loss": 0.0429, "step": 20607 }, { "epoch": 3.78, "learning_rate": 1.1375432093705861e-05, "loss": 0.0458, "step": 20608 }, { "epoch": 3.78, "learning_rate": 1.1374696716117702e-05, "loss": 0.0434, "step": 20609 }, { "epoch": 3.78, "learning_rate": 1.1373961330952181e-05, "loss": 0.0511, "step": 20610 }, { "epoch": 3.78, "learning_rate": 1.1373225938213355e-05, "loss": 0.0483, "step": 20611 }, { "epoch": 3.78, "learning_rate": 1.137249053790527e-05, "loss": 0.0519, "step": 20612 }, { "epoch": 3.78, "learning_rate": 1.1371755130031992e-05, "loss": 0.0441, "step": 20613 }, { "epoch": 3.78, "learning_rate": 1.1371019714597562e-05, "loss": 0.0431, "step": 20614 }, { "epoch": 3.78, "learning_rate": 1.1370284291606042e-05, "loss": 0.0408, "step": 20615 }, { "epoch": 3.78, "learning_rate": 1.1369548861061486e-05, "loss": 0.0454, "step": 20616 }, { "epoch": 3.78, "learning_rate": 1.136881342296794e-05, "loss": 0.0427, "step": 20617 }, { "epoch": 3.78, "learning_rate": 1.1368077977329463e-05, "loss": 0.0432, "step": 20618 }, { "epoch": 3.78, "learning_rate": 1.1367342524150109e-05, "loss": 0.0431, "step": 20619 }, { "epoch": 3.78, "learning_rate": 1.136660706343393e-05, "loss": 0.0401, "step": 20620 }, { "epoch": 3.78, "learning_rate": 1.1365871595184983e-05, "loss": 0.053, "step": 20621 }, { "epoch": 3.78, "learning_rate": 1.1365136119407318e-05, "loss": 0.0446, "step": 20622 }, { "epoch": 3.78, "learning_rate": 1.1364400636104988e-05, "loss": 0.0426, "step": 20623 }, { "epoch": 3.78, "learning_rate": 1.1363665145282055e-05, "loss": 0.0377, "step": 20624 }, { "epoch": 3.78, "learning_rate": 1.1362929646942564e-05, "loss": 0.0454, "step": 20625 }, { "epoch": 3.78, "learning_rate": 1.1362194141090574e-05, "loss": 0.0477, "step": 20626 }, { "epoch": 3.78, "learning_rate": 1.136145862773014e-05, "loss": 0.0509, "step": 20627 }, { "epoch": 3.78, "learning_rate": 1.1360723106865313e-05, "loss": 0.0474, "step": 20628 }, { "epoch": 3.78, "learning_rate": 1.135998757850015e-05, "loss": 0.0434, "step": 20629 }, { "epoch": 3.78, "learning_rate": 1.13592520426387e-05, "loss": 0.0385, "step": 20630 }, { "epoch": 3.78, "learning_rate": 1.1358516499285021e-05, "loss": 0.0382, "step": 20631 }, { "epoch": 3.78, "learning_rate": 1.135778094844317e-05, "loss": 0.0346, "step": 20632 }, { "epoch": 3.78, "learning_rate": 1.1357045390117197e-05, "loss": 0.0504, "step": 20633 }, { "epoch": 3.78, "learning_rate": 1.1356309824311156e-05, "loss": 0.0316, "step": 20634 }, { "epoch": 3.78, "learning_rate": 1.1355574251029107e-05, "loss": 0.045, "step": 20635 }, { "epoch": 3.78, "learning_rate": 1.1354838670275099e-05, "loss": 0.0379, "step": 20636 }, { "epoch": 3.78, "learning_rate": 1.135410308205319e-05, "loss": 0.0399, "step": 20637 }, { "epoch": 3.78, "learning_rate": 1.1353367486367433e-05, "loss": 0.039, "step": 20638 }, { "epoch": 3.78, "learning_rate": 1.1352631883221883e-05, "loss": 0.0416, "step": 20639 }, { "epoch": 3.79, "learning_rate": 1.1351896272620593e-05, "loss": 0.0468, "step": 20640 }, { "epoch": 3.79, "learning_rate": 1.1351160654567617e-05, "loss": 0.0325, "step": 20641 }, { "epoch": 3.79, "learning_rate": 1.1350425029067014e-05, "loss": 0.0465, "step": 20642 }, { "epoch": 3.79, "learning_rate": 1.1349689396122838e-05, "loss": 0.0512, "step": 20643 }, { "epoch": 3.79, "learning_rate": 1.134895375573914e-05, "loss": 0.0423, "step": 20644 }, { "epoch": 3.79, "learning_rate": 1.134821810791998e-05, "loss": 0.0469, "step": 20645 }, { "epoch": 3.79, "learning_rate": 1.134748245266941e-05, "loss": 0.0406, "step": 20646 }, { "epoch": 3.79, "learning_rate": 1.134674678999148e-05, "loss": 0.0476, "step": 20647 }, { "epoch": 3.79, "learning_rate": 1.1346011119890258e-05, "loss": 0.0445, "step": 20648 }, { "epoch": 3.79, "learning_rate": 1.1345275442369787e-05, "loss": 0.0436, "step": 20649 }, { "epoch": 3.79, "learning_rate": 1.1344539757434125e-05, "loss": 0.0494, "step": 20650 }, { "epoch": 3.79, "learning_rate": 1.1343804065087332e-05, "loss": 0.0387, "step": 20651 }, { "epoch": 3.79, "learning_rate": 1.1343068365333454e-05, "loss": 0.04, "step": 20652 }, { "epoch": 3.79, "learning_rate": 1.1342332658176556e-05, "loss": 0.0417, "step": 20653 }, { "epoch": 3.79, "learning_rate": 1.1341596943620688e-05, "loss": 0.0408, "step": 20654 }, { "epoch": 3.79, "learning_rate": 1.1340861221669908e-05, "loss": 0.048, "step": 20655 }, { "epoch": 3.79, "learning_rate": 1.1340125492328265e-05, "loss": 0.0401, "step": 20656 }, { "epoch": 3.79, "learning_rate": 1.1339389755599823e-05, "loss": 0.0375, "step": 20657 }, { "epoch": 3.79, "learning_rate": 1.133865401148863e-05, "loss": 0.0449, "step": 20658 }, { "epoch": 3.79, "learning_rate": 1.1337918259998748e-05, "loss": 0.0427, "step": 20659 }, { "epoch": 3.79, "learning_rate": 1.1337182501134229e-05, "loss": 0.0415, "step": 20660 }, { "epoch": 3.79, "learning_rate": 1.1336446734899127e-05, "loss": 0.0452, "step": 20661 }, { "epoch": 3.79, "learning_rate": 1.1335710961297498e-05, "loss": 0.0424, "step": 20662 }, { "epoch": 3.79, "learning_rate": 1.1334975180333403e-05, "loss": 0.038, "step": 20663 }, { "epoch": 3.79, "learning_rate": 1.133423939201089e-05, "loss": 0.0435, "step": 20664 }, { "epoch": 3.79, "learning_rate": 1.133350359633402e-05, "loss": 0.0407, "step": 20665 }, { "epoch": 3.79, "learning_rate": 1.1332767793306848e-05, "loss": 0.0388, "step": 20666 }, { "epoch": 3.79, "learning_rate": 1.1332031982933425e-05, "loss": 0.0383, "step": 20667 }, { "epoch": 3.79, "learning_rate": 1.1331296165217813e-05, "loss": 0.0474, "step": 20668 }, { "epoch": 3.79, "learning_rate": 1.1330560340164064e-05, "loss": 0.0479, "step": 20669 }, { "epoch": 3.79, "learning_rate": 1.1329824507776237e-05, "loss": 0.0441, "step": 20670 }, { "epoch": 3.79, "learning_rate": 1.1329088668058386e-05, "loss": 0.0452, "step": 20671 }, { "epoch": 3.79, "learning_rate": 1.1328352821014567e-05, "loss": 0.0422, "step": 20672 }, { "epoch": 3.79, "learning_rate": 1.1327616966648836e-05, "loss": 0.038, "step": 20673 }, { "epoch": 3.79, "learning_rate": 1.1326881104965246e-05, "loss": 0.0407, "step": 20674 }, { "epoch": 3.79, "learning_rate": 1.1326145235967863e-05, "loss": 0.0501, "step": 20675 }, { "epoch": 3.79, "learning_rate": 1.1325409359660728e-05, "loss": 0.044, "step": 20676 }, { "epoch": 3.79, "learning_rate": 1.1324673476047912e-05, "loss": 0.0421, "step": 20677 }, { "epoch": 3.79, "learning_rate": 1.1323937585133461e-05, "loss": 0.0425, "step": 20678 }, { "epoch": 3.79, "learning_rate": 1.132320168692144e-05, "loss": 0.0414, "step": 20679 }, { "epoch": 3.79, "learning_rate": 1.1322465781415896e-05, "loss": 0.0448, "step": 20680 }, { "epoch": 3.79, "learning_rate": 1.1321729868620893e-05, "loss": 0.0343, "step": 20681 }, { "epoch": 3.79, "learning_rate": 1.132099394854048e-05, "loss": 0.0385, "step": 20682 }, { "epoch": 3.79, "learning_rate": 1.1320258021178718e-05, "loss": 0.0427, "step": 20683 }, { "epoch": 3.79, "learning_rate": 1.1319522086539666e-05, "loss": 0.036, "step": 20684 }, { "epoch": 3.79, "learning_rate": 1.1318786144627373e-05, "loss": 0.0502, "step": 20685 }, { "epoch": 3.79, "learning_rate": 1.1318050195445903e-05, "loss": 0.0366, "step": 20686 }, { "epoch": 3.79, "learning_rate": 1.1317314238999309e-05, "loss": 0.0425, "step": 20687 }, { "epoch": 3.79, "learning_rate": 1.1316578275291648e-05, "loss": 0.0415, "step": 20688 }, { "epoch": 3.79, "learning_rate": 1.1315842304326974e-05, "loss": 0.0478, "step": 20689 }, { "epoch": 3.79, "learning_rate": 1.1315106326109351e-05, "loss": 0.0387, "step": 20690 }, { "epoch": 3.79, "learning_rate": 1.131437034064283e-05, "loss": 0.0462, "step": 20691 }, { "epoch": 3.79, "learning_rate": 1.1313634347931466e-05, "loss": 0.0548, "step": 20692 }, { "epoch": 3.79, "learning_rate": 1.1312898347979321e-05, "loss": 0.0425, "step": 20693 }, { "epoch": 3.79, "learning_rate": 1.131216234079045e-05, "loss": 0.0452, "step": 20694 }, { "epoch": 3.8, "learning_rate": 1.1311426326368906e-05, "loss": 0.0427, "step": 20695 }, { "epoch": 3.8, "learning_rate": 1.1310690304718754e-05, "loss": 0.0489, "step": 20696 }, { "epoch": 3.8, "learning_rate": 1.1309954275844044e-05, "loss": 0.0461, "step": 20697 }, { "epoch": 3.8, "learning_rate": 1.1309218239748836e-05, "loss": 0.041, "step": 20698 }, { "epoch": 3.8, "learning_rate": 1.1308482196437187e-05, "loss": 0.0522, "step": 20699 }, { "epoch": 3.8, "learning_rate": 1.1307746145913149e-05, "loss": 0.0443, "step": 20700 }, { "epoch": 3.8, "learning_rate": 1.1307010088180788e-05, "loss": 0.0475, "step": 20701 }, { "epoch": 3.8, "learning_rate": 1.1306274023244157e-05, "loss": 0.049, "step": 20702 }, { "epoch": 3.8, "learning_rate": 1.1305537951107311e-05, "loss": 0.0411, "step": 20703 }, { "epoch": 3.8, "learning_rate": 1.1304801871774309e-05, "loss": 0.0369, "step": 20704 }, { "epoch": 3.8, "learning_rate": 1.1304065785249212e-05, "loss": 0.043, "step": 20705 }, { "epoch": 3.8, "learning_rate": 1.1303329691536072e-05, "loss": 0.0463, "step": 20706 }, { "epoch": 3.8, "learning_rate": 1.1302593590638946e-05, "loss": 0.0466, "step": 20707 }, { "epoch": 3.8, "learning_rate": 1.1301857482561895e-05, "loss": 0.0389, "step": 20708 }, { "epoch": 3.8, "learning_rate": 1.1301121367308973e-05, "loss": 0.0448, "step": 20709 }, { "epoch": 3.8, "learning_rate": 1.1300385244884242e-05, "loss": 0.0416, "step": 20710 }, { "epoch": 3.8, "learning_rate": 1.1299649115291757e-05, "loss": 0.0458, "step": 20711 }, { "epoch": 3.8, "learning_rate": 1.1298912978535577e-05, "loss": 0.0412, "step": 20712 }, { "epoch": 3.8, "learning_rate": 1.1298176834619756e-05, "loss": 0.0437, "step": 20713 }, { "epoch": 3.8, "learning_rate": 1.1297440683548355e-05, "loss": 0.0473, "step": 20714 }, { "epoch": 3.8, "learning_rate": 1.129670452532543e-05, "loss": 0.0408, "step": 20715 }, { "epoch": 3.8, "learning_rate": 1.129596835995504e-05, "loss": 0.0452, "step": 20716 }, { "epoch": 3.8, "learning_rate": 1.1295232187441243e-05, "loss": 0.0451, "step": 20717 }, { "epoch": 3.8, "learning_rate": 1.1294496007788091e-05, "loss": 0.0405, "step": 20718 }, { "epoch": 3.8, "learning_rate": 1.129375982099965e-05, "loss": 0.0467, "step": 20719 }, { "epoch": 3.8, "learning_rate": 1.1293023627079975e-05, "loss": 0.0363, "step": 20720 }, { "epoch": 3.8, "learning_rate": 1.1292287426033123e-05, "loss": 0.0378, "step": 20721 }, { "epoch": 3.8, "learning_rate": 1.1291551217863152e-05, "loss": 0.049, "step": 20722 }, { "epoch": 3.8, "learning_rate": 1.1290815002574123e-05, "loss": 0.0347, "step": 20723 }, { "epoch": 3.8, "learning_rate": 1.1290078780170088e-05, "loss": 0.0409, "step": 20724 }, { "epoch": 3.8, "learning_rate": 1.1289342550655111e-05, "loss": 0.0455, "step": 20725 }, { "epoch": 3.8, "learning_rate": 1.1288606314033247e-05, "loss": 0.0453, "step": 20726 }, { "epoch": 3.8, "learning_rate": 1.1287870070308553e-05, "loss": 0.0457, "step": 20727 }, { "epoch": 3.8, "learning_rate": 1.1287133819485092e-05, "loss": 0.0452, "step": 20728 }, { "epoch": 3.8, "learning_rate": 1.1286397561566916e-05, "loss": 0.0428, "step": 20729 }, { "epoch": 3.8, "learning_rate": 1.1285661296558088e-05, "loss": 0.041, "step": 20730 }, { "epoch": 3.8, "learning_rate": 1.1284925024462665e-05, "loss": 0.0423, "step": 20731 }, { "epoch": 3.8, "learning_rate": 1.1284188745284705e-05, "loss": 0.0397, "step": 20732 }, { "epoch": 3.8, "learning_rate": 1.1283452459028266e-05, "loss": 0.0445, "step": 20733 }, { "epoch": 3.8, "learning_rate": 1.128271616569741e-05, "loss": 0.0518, "step": 20734 }, { "epoch": 3.8, "learning_rate": 1.128197986529619e-05, "loss": 0.046, "step": 20735 }, { "epoch": 3.8, "learning_rate": 1.1281243557828664e-05, "loss": 0.0403, "step": 20736 }, { "epoch": 3.8, "learning_rate": 1.12805072432989e-05, "loss": 0.041, "step": 20737 }, { "epoch": 3.8, "learning_rate": 1.1279770921710944e-05, "loss": 0.0399, "step": 20738 }, { "epoch": 3.8, "learning_rate": 1.1279034593068862e-05, "loss": 0.0449, "step": 20739 }, { "epoch": 3.8, "learning_rate": 1.1278298257376714e-05, "loss": 0.0438, "step": 20740 }, { "epoch": 3.8, "learning_rate": 1.1277561914638555e-05, "loss": 0.0372, "step": 20741 }, { "epoch": 3.8, "learning_rate": 1.1276825564858443e-05, "loss": 0.0357, "step": 20742 }, { "epoch": 3.8, "learning_rate": 1.1276089208040442e-05, "loss": 0.05, "step": 20743 }, { "epoch": 3.8, "learning_rate": 1.1275352844188605e-05, "loss": 0.0386, "step": 20744 }, { "epoch": 3.8, "learning_rate": 1.1274616473306992e-05, "loss": 0.0438, "step": 20745 }, { "epoch": 3.8, "learning_rate": 1.1273880095399667e-05, "loss": 0.0489, "step": 20746 }, { "epoch": 3.8, "learning_rate": 1.1273143710470682e-05, "loss": 0.0391, "step": 20747 }, { "epoch": 3.8, "learning_rate": 1.1272407318524099e-05, "loss": 0.0488, "step": 20748 }, { "epoch": 3.81, "learning_rate": 1.127167091956398e-05, "loss": 0.0509, "step": 20749 }, { "epoch": 3.81, "learning_rate": 1.1270934513594378e-05, "loss": 0.0471, "step": 20750 }, { "epoch": 3.81, "learning_rate": 1.1270198100619354e-05, "loss": 0.0377, "step": 20751 }, { "epoch": 3.81, "learning_rate": 1.1269461680642975e-05, "loss": 0.0451, "step": 20752 }, { "epoch": 3.81, "learning_rate": 1.1268725253669286e-05, "loss": 0.0457, "step": 20753 }, { "epoch": 3.81, "learning_rate": 1.1267988819702355e-05, "loss": 0.0391, "step": 20754 }, { "epoch": 3.81, "learning_rate": 1.1267252378746246e-05, "loss": 0.0399, "step": 20755 }, { "epoch": 3.81, "learning_rate": 1.1266515930805006e-05, "loss": 0.0466, "step": 20756 }, { "epoch": 3.81, "learning_rate": 1.1265779475882704e-05, "loss": 0.051, "step": 20757 }, { "epoch": 3.81, "learning_rate": 1.1265043013983391e-05, "loss": 0.0414, "step": 20758 }, { "epoch": 3.81, "learning_rate": 1.1264306545111136e-05, "loss": 0.0416, "step": 20759 }, { "epoch": 3.81, "learning_rate": 1.1263570069269992e-05, "loss": 0.0389, "step": 20760 }, { "epoch": 3.81, "learning_rate": 1.1262833586464022e-05, "loss": 0.0469, "step": 20761 }, { "epoch": 3.81, "learning_rate": 1.126209709669728e-05, "loss": 0.0344, "step": 20762 }, { "epoch": 3.81, "learning_rate": 1.1261360599973832e-05, "loss": 0.0437, "step": 20763 }, { "epoch": 3.81, "learning_rate": 1.1260624096297735e-05, "loss": 0.0404, "step": 20764 }, { "epoch": 3.81, "learning_rate": 1.1259887585673045e-05, "loss": 0.0475, "step": 20765 }, { "epoch": 3.81, "learning_rate": 1.1259151068103829e-05, "loss": 0.0428, "step": 20766 }, { "epoch": 3.81, "learning_rate": 1.1258414543594143e-05, "loss": 0.0436, "step": 20767 }, { "epoch": 3.81, "learning_rate": 1.1257678012148045e-05, "loss": 0.0451, "step": 20768 }, { "epoch": 3.81, "learning_rate": 1.1256941473769594e-05, "loss": 0.0408, "step": 20769 }, { "epoch": 3.81, "learning_rate": 1.1256204928462858e-05, "loss": 0.0483, "step": 20770 }, { "epoch": 3.81, "learning_rate": 1.1255468376231884e-05, "loss": 0.0485, "step": 20771 }, { "epoch": 3.81, "learning_rate": 1.1254731817080745e-05, "loss": 0.0467, "step": 20772 }, { "epoch": 3.81, "learning_rate": 1.125399525101349e-05, "loss": 0.0347, "step": 20773 }, { "epoch": 3.81, "learning_rate": 1.1253258678034187e-05, "loss": 0.0516, "step": 20774 }, { "epoch": 3.81, "learning_rate": 1.1252522098146891e-05, "loss": 0.049, "step": 20775 }, { "epoch": 3.81, "learning_rate": 1.1251785511355666e-05, "loss": 0.0389, "step": 20776 }, { "epoch": 3.81, "learning_rate": 1.125104891766457e-05, "loss": 0.0368, "step": 20777 }, { "epoch": 3.81, "learning_rate": 1.1250312317077659e-05, "loss": 0.0417, "step": 20778 }, { "epoch": 3.81, "learning_rate": 1.1249575709599e-05, "loss": 0.0394, "step": 20779 }, { "epoch": 3.81, "learning_rate": 1.1248839095232651e-05, "loss": 0.0353, "step": 20780 }, { "epoch": 3.81, "learning_rate": 1.1248102473982672e-05, "loss": 0.0405, "step": 20781 }, { "epoch": 3.81, "learning_rate": 1.124736584585312e-05, "loss": 0.0385, "step": 20782 }, { "epoch": 3.81, "learning_rate": 1.1246629210848062e-05, "loss": 0.0366, "step": 20783 }, { "epoch": 3.81, "learning_rate": 1.1245892568971553e-05, "loss": 0.0455, "step": 20784 }, { "epoch": 3.81, "learning_rate": 1.1245155920227655e-05, "loss": 0.0404, "step": 20785 }, { "epoch": 3.81, "learning_rate": 1.1244419264620425e-05, "loss": 0.0444, "step": 20786 }, { "epoch": 3.81, "learning_rate": 1.1243682602153931e-05, "loss": 0.0415, "step": 20787 }, { "epoch": 3.81, "learning_rate": 1.124294593283223e-05, "loss": 0.0419, "step": 20788 }, { "epoch": 3.81, "learning_rate": 1.1242209256659377e-05, "loss": 0.0382, "step": 20789 }, { "epoch": 3.81, "learning_rate": 1.1241472573639442e-05, "loss": 0.0397, "step": 20790 }, { "epoch": 3.81, "learning_rate": 1.124073588377648e-05, "loss": 0.039, "step": 20791 }, { "epoch": 3.81, "learning_rate": 1.1239999187074553e-05, "loss": 0.0486, "step": 20792 }, { "epoch": 3.81, "learning_rate": 1.123926248353772e-05, "loss": 0.0356, "step": 20793 }, { "epoch": 3.81, "learning_rate": 1.1238525773170046e-05, "loss": 0.049, "step": 20794 }, { "epoch": 3.81, "learning_rate": 1.1237789055975585e-05, "loss": 0.0429, "step": 20795 }, { "epoch": 3.81, "learning_rate": 1.1237052331958405e-05, "loss": 0.0479, "step": 20796 }, { "epoch": 3.81, "learning_rate": 1.1236315601122561e-05, "loss": 0.0501, "step": 20797 }, { "epoch": 3.81, "learning_rate": 1.1235578863472118e-05, "loss": 0.0431, "step": 20798 }, { "epoch": 3.81, "learning_rate": 1.1234842119011136e-05, "loss": 0.0378, "step": 20799 }, { "epoch": 3.81, "learning_rate": 1.1234105367743676e-05, "loss": 0.0377, "step": 20800 }, { "epoch": 3.81, "learning_rate": 1.1233368609673796e-05, "loss": 0.0378, "step": 20801 }, { "epoch": 3.81, "learning_rate": 1.1232631844805559e-05, "loss": 0.0398, "step": 20802 }, { "epoch": 3.81, "learning_rate": 1.123189507314303e-05, "loss": 0.0476, "step": 20803 }, { "epoch": 3.82, "learning_rate": 1.123115829469026e-05, "loss": 0.0422, "step": 20804 }, { "epoch": 3.82, "learning_rate": 1.1230421509451323e-05, "loss": 0.0428, "step": 20805 }, { "epoch": 3.82, "learning_rate": 1.1229684717430269e-05, "loss": 0.0394, "step": 20806 }, { "epoch": 3.82, "learning_rate": 1.1228947918631167e-05, "loss": 0.046, "step": 20807 }, { "epoch": 3.82, "learning_rate": 1.1228211113058075e-05, "loss": 0.0401, "step": 20808 }, { "epoch": 3.82, "learning_rate": 1.1227474300715056e-05, "loss": 0.0375, "step": 20809 }, { "epoch": 3.82, "learning_rate": 1.1226737481606168e-05, "loss": 0.0407, "step": 20810 }, { "epoch": 3.82, "learning_rate": 1.1226000655735475e-05, "loss": 0.0452, "step": 20811 }, { "epoch": 3.82, "learning_rate": 1.1225263823107038e-05, "loss": 0.0378, "step": 20812 }, { "epoch": 3.82, "learning_rate": 1.1224526983724915e-05, "loss": 0.0427, "step": 20813 }, { "epoch": 3.82, "learning_rate": 1.1223790137593174e-05, "loss": 0.0437, "step": 20814 }, { "epoch": 3.82, "learning_rate": 1.122305328471587e-05, "loss": 0.0409, "step": 20815 }, { "epoch": 3.82, "learning_rate": 1.122231642509707e-05, "loss": 0.0452, "step": 20816 }, { "epoch": 3.82, "learning_rate": 1.1221579558740834e-05, "loss": 0.0487, "step": 20817 }, { "epoch": 3.82, "learning_rate": 1.1220842685651222e-05, "loss": 0.0417, "step": 20818 }, { "epoch": 3.82, "learning_rate": 1.1220105805832296e-05, "loss": 0.038, "step": 20819 }, { "epoch": 3.82, "learning_rate": 1.1219368919288118e-05, "loss": 0.0466, "step": 20820 }, { "epoch": 3.82, "learning_rate": 1.1218632026022752e-05, "loss": 0.0397, "step": 20821 }, { "epoch": 3.82, "learning_rate": 1.1217895126040252e-05, "loss": 0.0444, "step": 20822 }, { "epoch": 3.82, "learning_rate": 1.1217158219344693e-05, "loss": 0.0462, "step": 20823 }, { "epoch": 3.82, "learning_rate": 1.1216421305940124e-05, "loss": 0.0522, "step": 20824 }, { "epoch": 3.82, "learning_rate": 1.1215684385830614e-05, "loss": 0.0432, "step": 20825 }, { "epoch": 3.82, "learning_rate": 1.1214947459020222e-05, "loss": 0.0483, "step": 20826 }, { "epoch": 3.82, "learning_rate": 1.1214210525513014e-05, "loss": 0.0367, "step": 20827 }, { "epoch": 3.82, "learning_rate": 1.1213473585313046e-05, "loss": 0.0447, "step": 20828 }, { "epoch": 3.82, "learning_rate": 1.1212736638424386e-05, "loss": 0.0444, "step": 20829 }, { "epoch": 3.82, "learning_rate": 1.1211999684851094e-05, "loss": 0.0441, "step": 20830 }, { "epoch": 3.82, "learning_rate": 1.1211262724597226e-05, "loss": 0.0412, "step": 20831 }, { "epoch": 3.82, "learning_rate": 1.1210525757666857e-05, "loss": 0.0429, "step": 20832 }, { "epoch": 3.82, "learning_rate": 1.1209788784064035e-05, "loss": 0.0474, "step": 20833 }, { "epoch": 3.82, "learning_rate": 1.1209051803792832e-05, "loss": 0.0471, "step": 20834 }, { "epoch": 3.82, "learning_rate": 1.1208314816857306e-05, "loss": 0.0555, "step": 20835 }, { "epoch": 3.82, "learning_rate": 1.1207577823261521e-05, "loss": 0.0403, "step": 20836 }, { "epoch": 3.82, "learning_rate": 1.1206840823009536e-05, "loss": 0.0418, "step": 20837 }, { "epoch": 3.82, "learning_rate": 1.1206103816105422e-05, "loss": 0.0421, "step": 20838 }, { "epoch": 3.82, "learning_rate": 1.1205366802553231e-05, "loss": 0.0511, "step": 20839 }, { "epoch": 3.82, "learning_rate": 1.120462978235703e-05, "loss": 0.0453, "step": 20840 }, { "epoch": 3.82, "learning_rate": 1.1203892755520886e-05, "loss": 0.0399, "step": 20841 }, { "epoch": 3.82, "learning_rate": 1.1203155722048853e-05, "loss": 0.0435, "step": 20842 }, { "epoch": 3.82, "learning_rate": 1.1202418681944998e-05, "loss": 0.0442, "step": 20843 }, { "epoch": 3.82, "learning_rate": 1.1201681635213383e-05, "loss": 0.0385, "step": 20844 }, { "epoch": 3.82, "learning_rate": 1.120094458185807e-05, "loss": 0.0414, "step": 20845 }, { "epoch": 3.82, "learning_rate": 1.1200207521883122e-05, "loss": 0.043, "step": 20846 }, { "epoch": 3.82, "learning_rate": 1.1199470455292608e-05, "loss": 0.0449, "step": 20847 }, { "epoch": 3.82, "learning_rate": 1.1198733382090576e-05, "loss": 0.0478, "step": 20848 }, { "epoch": 3.82, "learning_rate": 1.1197996302281103e-05, "loss": 0.0454, "step": 20849 }, { "epoch": 3.82, "learning_rate": 1.1197259215868244e-05, "loss": 0.0388, "step": 20850 }, { "epoch": 3.82, "learning_rate": 1.1196522122856065e-05, "loss": 0.0508, "step": 20851 }, { "epoch": 3.82, "learning_rate": 1.1195785023248628e-05, "loss": 0.04, "step": 20852 }, { "epoch": 3.82, "learning_rate": 1.1195047917049997e-05, "loss": 0.0386, "step": 20853 }, { "epoch": 3.82, "learning_rate": 1.1194310804264232e-05, "loss": 0.0433, "step": 20854 }, { "epoch": 3.82, "learning_rate": 1.1193573684895398e-05, "loss": 0.0385, "step": 20855 }, { "epoch": 3.82, "learning_rate": 1.119283655894756e-05, "loss": 0.0382, "step": 20856 }, { "epoch": 3.82, "learning_rate": 1.1192099426424776e-05, "loss": 0.0433, "step": 20857 }, { "epoch": 3.83, "learning_rate": 1.1191362287331113e-05, "loss": 0.0392, "step": 20858 }, { "epoch": 3.83, "learning_rate": 1.1190625141670635e-05, "loss": 0.0442, "step": 20859 }, { "epoch": 3.83, "learning_rate": 1.1189887989447401e-05, "loss": 0.0408, "step": 20860 }, { "epoch": 3.83, "learning_rate": 1.1189150830665478e-05, "loss": 0.0374, "step": 20861 }, { "epoch": 3.83, "learning_rate": 1.1188413665328929e-05, "loss": 0.0386, "step": 20862 }, { "epoch": 3.83, "learning_rate": 1.1187676493441812e-05, "loss": 0.0392, "step": 20863 }, { "epoch": 3.83, "learning_rate": 1.1186939315008197e-05, "loss": 0.0365, "step": 20864 }, { "epoch": 3.83, "learning_rate": 1.1186202130032146e-05, "loss": 0.0403, "step": 20865 }, { "epoch": 3.83, "learning_rate": 1.1185464938517718e-05, "loss": 0.0485, "step": 20866 }, { "epoch": 3.83, "learning_rate": 1.1184727740468981e-05, "loss": 0.0388, "step": 20867 }, { "epoch": 3.83, "learning_rate": 1.1183990535889996e-05, "loss": 0.0392, "step": 20868 }, { "epoch": 3.83, "learning_rate": 1.118325332478483e-05, "loss": 0.0451, "step": 20869 }, { "epoch": 3.83, "learning_rate": 1.1182516107157542e-05, "loss": 0.0431, "step": 20870 }, { "epoch": 3.83, "learning_rate": 1.1181778883012198e-05, "loss": 0.0483, "step": 20871 }, { "epoch": 3.83, "learning_rate": 1.118104165235286e-05, "loss": 0.0397, "step": 20872 }, { "epoch": 3.83, "learning_rate": 1.1180304415183592e-05, "loss": 0.0409, "step": 20873 }, { "epoch": 3.83, "learning_rate": 1.1179567171508463e-05, "loss": 0.0448, "step": 20874 }, { "epoch": 3.83, "learning_rate": 1.1178829921331526e-05, "loss": 0.0502, "step": 20875 }, { "epoch": 3.83, "learning_rate": 1.1178092664656854e-05, "loss": 0.041, "step": 20876 }, { "epoch": 3.83, "learning_rate": 1.117735540148851e-05, "loss": 0.0459, "step": 20877 }, { "epoch": 3.83, "learning_rate": 1.1176618131830553e-05, "loss": 0.0409, "step": 20878 }, { "epoch": 3.83, "learning_rate": 1.1175880855687048e-05, "loss": 0.0443, "step": 20879 }, { "epoch": 3.83, "learning_rate": 1.1175143573062063e-05, "loss": 0.0429, "step": 20880 }, { "epoch": 3.83, "learning_rate": 1.1174406283959656e-05, "loss": 0.0405, "step": 20881 }, { "epoch": 3.83, "learning_rate": 1.1173668988383897e-05, "loss": 0.0377, "step": 20882 }, { "epoch": 3.83, "learning_rate": 1.1172931686338846e-05, "loss": 0.0446, "step": 20883 }, { "epoch": 3.83, "learning_rate": 1.1172194377828567e-05, "loss": 0.0438, "step": 20884 }, { "epoch": 3.83, "learning_rate": 1.1171457062857124e-05, "loss": 0.0403, "step": 20885 }, { "epoch": 3.83, "learning_rate": 1.1170719741428586e-05, "loss": 0.0373, "step": 20886 }, { "epoch": 3.83, "learning_rate": 1.1169982413547012e-05, "loss": 0.0371, "step": 20887 }, { "epoch": 3.83, "learning_rate": 1.1169245079216467e-05, "loss": 0.0463, "step": 20888 }, { "epoch": 3.83, "learning_rate": 1.1168507738441017e-05, "loss": 0.0387, "step": 20889 }, { "epoch": 3.83, "learning_rate": 1.1167770391224721e-05, "loss": 0.0401, "step": 20890 }, { "epoch": 3.83, "learning_rate": 1.1167033037571653e-05, "loss": 0.0384, "step": 20891 }, { "epoch": 3.83, "learning_rate": 1.1166295677485865e-05, "loss": 0.0377, "step": 20892 }, { "epoch": 3.83, "learning_rate": 1.1165558310971432e-05, "loss": 0.0411, "step": 20893 }, { "epoch": 3.83, "learning_rate": 1.1164820938032414e-05, "loss": 0.0402, "step": 20894 }, { "epoch": 3.83, "learning_rate": 1.1164083558672876e-05, "loss": 0.0418, "step": 20895 }, { "epoch": 3.83, "learning_rate": 1.1163346172896881e-05, "loss": 0.0425, "step": 20896 }, { "epoch": 3.83, "learning_rate": 1.1162608780708498e-05, "loss": 0.0407, "step": 20897 }, { "epoch": 3.83, "learning_rate": 1.1161871382111783e-05, "loss": 0.049, "step": 20898 }, { "epoch": 3.83, "learning_rate": 1.1161133977110807e-05, "loss": 0.0494, "step": 20899 }, { "epoch": 3.83, "learning_rate": 1.1160396565709637e-05, "loss": 0.0468, "step": 20900 }, { "epoch": 3.83, "learning_rate": 1.115965914791233e-05, "loss": 0.0439, "step": 20901 }, { "epoch": 3.83, "learning_rate": 1.1158921723722955e-05, "loss": 0.0433, "step": 20902 }, { "epoch": 3.83, "learning_rate": 1.1158184293145578e-05, "loss": 0.0518, "step": 20903 }, { "epoch": 3.83, "learning_rate": 1.1157446856184263e-05, "loss": 0.0486, "step": 20904 }, { "epoch": 3.83, "learning_rate": 1.115670941284307e-05, "loss": 0.0445, "step": 20905 }, { "epoch": 3.83, "learning_rate": 1.115597196312607e-05, "loss": 0.0423, "step": 20906 }, { "epoch": 3.83, "learning_rate": 1.1155234507037326e-05, "loss": 0.0377, "step": 20907 }, { "epoch": 3.83, "learning_rate": 1.11544970445809e-05, "loss": 0.045, "step": 20908 }, { "epoch": 3.83, "learning_rate": 1.1153759575760864e-05, "loss": 0.0391, "step": 20909 }, { "epoch": 3.83, "learning_rate": 1.1153022100581273e-05, "loss": 0.0356, "step": 20910 }, { "epoch": 3.83, "learning_rate": 1.1152284619046199e-05, "loss": 0.0419, "step": 20911 }, { "epoch": 3.83, "learning_rate": 1.1151547131159706e-05, "loss": 0.0413, "step": 20912 }, { "epoch": 3.84, "learning_rate": 1.1150809636925857e-05, "loss": 0.0468, "step": 20913 }, { "epoch": 3.84, "learning_rate": 1.1150072136348717e-05, "loss": 0.0396, "step": 20914 }, { "epoch": 3.84, "learning_rate": 1.1149334629432356e-05, "loss": 0.0422, "step": 20915 }, { "epoch": 3.84, "learning_rate": 1.1148597116180835e-05, "loss": 0.0426, "step": 20916 }, { "epoch": 3.84, "learning_rate": 1.1147859596598215e-05, "loss": 0.042, "step": 20917 }, { "epoch": 3.84, "learning_rate": 1.1147122070688574e-05, "loss": 0.0426, "step": 20918 }, { "epoch": 3.84, "learning_rate": 1.1146384538455963e-05, "loss": 0.0411, "step": 20919 }, { "epoch": 3.84, "learning_rate": 1.1145646999904455e-05, "loss": 0.0404, "step": 20920 }, { "epoch": 3.84, "learning_rate": 1.1144909455038116e-05, "loss": 0.0435, "step": 20921 }, { "epoch": 3.84, "learning_rate": 1.1144171903861008e-05, "loss": 0.0431, "step": 20922 }, { "epoch": 3.84, "learning_rate": 1.1143434346377194e-05, "loss": 0.042, "step": 20923 }, { "epoch": 3.84, "learning_rate": 1.114269678259075e-05, "loss": 0.0489, "step": 20924 }, { "epoch": 3.84, "learning_rate": 1.1141959212505729e-05, "loss": 0.043, "step": 20925 }, { "epoch": 3.84, "learning_rate": 1.1141221636126202e-05, "loss": 0.0367, "step": 20926 }, { "epoch": 3.84, "learning_rate": 1.114048405345624e-05, "loss": 0.0413, "step": 20927 }, { "epoch": 3.84, "learning_rate": 1.1139746464499899e-05, "loss": 0.0452, "step": 20928 }, { "epoch": 3.84, "learning_rate": 1.113900886926125e-05, "loss": 0.045, "step": 20929 }, { "epoch": 3.84, "learning_rate": 1.1138271267744355e-05, "loss": 0.0437, "step": 20930 }, { "epoch": 3.84, "learning_rate": 1.1137533659953286e-05, "loss": 0.044, "step": 20931 }, { "epoch": 3.84, "learning_rate": 1.1136796045892102e-05, "loss": 0.0392, "step": 20932 }, { "epoch": 3.84, "learning_rate": 1.1136058425564876e-05, "loss": 0.0425, "step": 20933 }, { "epoch": 3.84, "learning_rate": 1.1135320798975662e-05, "loss": 0.0418, "step": 20934 }, { "epoch": 3.84, "learning_rate": 1.1134583166128538e-05, "loss": 0.0424, "step": 20935 }, { "epoch": 3.84, "learning_rate": 1.1133845527027567e-05, "loss": 0.044, "step": 20936 }, { "epoch": 3.84, "learning_rate": 1.1133107881676808e-05, "loss": 0.0364, "step": 20937 }, { "epoch": 3.84, "learning_rate": 1.1132370230080334e-05, "loss": 0.0506, "step": 20938 }, { "epoch": 3.84, "learning_rate": 1.1131632572242208e-05, "loss": 0.0404, "step": 20939 }, { "epoch": 3.84, "learning_rate": 1.1130894908166499e-05, "loss": 0.0429, "step": 20940 }, { "epoch": 3.84, "learning_rate": 1.1130157237857266e-05, "loss": 0.0331, "step": 20941 }, { "epoch": 3.84, "learning_rate": 1.1129419561318584e-05, "loss": 0.0528, "step": 20942 }, { "epoch": 3.84, "learning_rate": 1.1128681878554513e-05, "loss": 0.0318, "step": 20943 }, { "epoch": 3.84, "learning_rate": 1.1127944189569122e-05, "loss": 0.0461, "step": 20944 }, { "epoch": 3.84, "learning_rate": 1.1127206494366476e-05, "loss": 0.0405, "step": 20945 }, { "epoch": 3.84, "learning_rate": 1.1126468792950642e-05, "loss": 0.0386, "step": 20946 }, { "epoch": 3.84, "learning_rate": 1.1125731085325687e-05, "loss": 0.0471, "step": 20947 }, { "epoch": 3.84, "learning_rate": 1.1124993371495671e-05, "loss": 0.0389, "step": 20948 }, { "epoch": 3.84, "learning_rate": 1.1124255651464671e-05, "loss": 0.0457, "step": 20949 }, { "epoch": 3.84, "learning_rate": 1.1123517925236742e-05, "loss": 0.0432, "step": 20950 }, { "epoch": 3.84, "learning_rate": 1.1122780192815961e-05, "loss": 0.0356, "step": 20951 }, { "epoch": 3.84, "learning_rate": 1.1122042454206386e-05, "loss": 0.0437, "step": 20952 }, { "epoch": 3.84, "learning_rate": 1.1121304709412086e-05, "loss": 0.0384, "step": 20953 }, { "epoch": 3.84, "learning_rate": 1.1120566958437131e-05, "loss": 0.0424, "step": 20954 }, { "epoch": 3.84, "learning_rate": 1.1119829201285583e-05, "loss": 0.0346, "step": 20955 }, { "epoch": 3.84, "learning_rate": 1.1119091437961511e-05, "loss": 0.0409, "step": 20956 }, { "epoch": 3.84, "learning_rate": 1.111835366846898e-05, "loss": 0.0463, "step": 20957 }, { "epoch": 3.84, "learning_rate": 1.1117615892812057e-05, "loss": 0.0445, "step": 20958 }, { "epoch": 3.84, "learning_rate": 1.111687811099481e-05, "loss": 0.0387, "step": 20959 }, { "epoch": 3.84, "learning_rate": 1.1116140323021305e-05, "loss": 0.0344, "step": 20960 }, { "epoch": 3.84, "learning_rate": 1.1115402528895606e-05, "loss": 0.0493, "step": 20961 }, { "epoch": 3.84, "learning_rate": 1.1114664728621785e-05, "loss": 0.0506, "step": 20962 }, { "epoch": 3.84, "learning_rate": 1.1113926922203906e-05, "loss": 0.0331, "step": 20963 }, { "epoch": 3.84, "learning_rate": 1.1113189109646034e-05, "loss": 0.0442, "step": 20964 }, { "epoch": 3.84, "learning_rate": 1.1112451290952238e-05, "loss": 0.0479, "step": 20965 }, { "epoch": 3.84, "learning_rate": 1.1111713466126586e-05, "loss": 0.0449, "step": 20966 }, { "epoch": 3.85, "learning_rate": 1.1110975635173138e-05, "loss": 0.0382, "step": 20967 }, { "epoch": 3.85, "learning_rate": 1.1110237798095973e-05, "loss": 0.0566, "step": 20968 }, { "epoch": 3.85, "learning_rate": 1.1109499954899147e-05, "loss": 0.0424, "step": 20969 }, { "epoch": 3.85, "learning_rate": 1.1108762105586732e-05, "loss": 0.0467, "step": 20970 }, { "epoch": 3.85, "learning_rate": 1.1108024250162797e-05, "loss": 0.0427, "step": 20971 }, { "epoch": 3.85, "learning_rate": 1.1107286388631404e-05, "loss": 0.0408, "step": 20972 }, { "epoch": 3.85, "learning_rate": 1.1106548520996622e-05, "loss": 0.0359, "step": 20973 }, { "epoch": 3.85, "learning_rate": 1.110581064726252e-05, "loss": 0.0527, "step": 20974 }, { "epoch": 3.85, "learning_rate": 1.1105072767433163e-05, "loss": 0.0416, "step": 20975 }, { "epoch": 3.85, "learning_rate": 1.1104334881512617e-05, "loss": 0.0495, "step": 20976 }, { "epoch": 3.85, "learning_rate": 1.1103596989504957e-05, "loss": 0.0364, "step": 20977 }, { "epoch": 3.85, "learning_rate": 1.110285909141424e-05, "loss": 0.0477, "step": 20978 }, { "epoch": 3.85, "learning_rate": 1.1102121187244538e-05, "loss": 0.04, "step": 20979 }, { "epoch": 3.85, "learning_rate": 1.110138327699992e-05, "loss": 0.0478, "step": 20980 }, { "epoch": 3.85, "learning_rate": 1.110064536068445e-05, "loss": 0.0426, "step": 20981 }, { "epoch": 3.85, "learning_rate": 1.1099907438302197e-05, "loss": 0.0432, "step": 20982 }, { "epoch": 3.85, "learning_rate": 1.1099169509857231e-05, "loss": 0.0368, "step": 20983 }, { "epoch": 3.85, "learning_rate": 1.1098431575353614e-05, "loss": 0.0408, "step": 20984 }, { "epoch": 3.85, "learning_rate": 1.1097693634795417e-05, "loss": 0.0319, "step": 20985 }, { "epoch": 3.85, "learning_rate": 1.1096955688186708e-05, "loss": 0.043, "step": 20986 }, { "epoch": 3.85, "learning_rate": 1.1096217735531552e-05, "loss": 0.0443, "step": 20987 }, { "epoch": 3.85, "learning_rate": 1.109547977683402e-05, "loss": 0.0448, "step": 20988 }, { "epoch": 3.85, "learning_rate": 1.1094741812098176e-05, "loss": 0.041, "step": 20989 }, { "epoch": 3.85, "learning_rate": 1.1094003841328091e-05, "loss": 0.0447, "step": 20990 }, { "epoch": 3.85, "learning_rate": 1.1093265864527832e-05, "loss": 0.0425, "step": 20991 }, { "epoch": 3.85, "learning_rate": 1.1092527881701462e-05, "loss": 0.0384, "step": 20992 }, { "epoch": 3.85, "learning_rate": 1.1091789892853058e-05, "loss": 0.0334, "step": 20993 }, { "epoch": 3.85, "learning_rate": 1.1091051897986677e-05, "loss": 0.0399, "step": 20994 }, { "epoch": 3.85, "learning_rate": 1.1090313897106398e-05, "loss": 0.0357, "step": 20995 }, { "epoch": 3.85, "learning_rate": 1.1089575890216277e-05, "loss": 0.0449, "step": 20996 }, { "epoch": 3.85, "learning_rate": 1.1088837877320393e-05, "loss": 0.0435, "step": 20997 }, { "epoch": 3.85, "learning_rate": 1.1088099858422807e-05, "loss": 0.0457, "step": 20998 }, { "epoch": 3.85, "learning_rate": 1.108736183352759e-05, "loss": 0.0462, "step": 20999 }, { "epoch": 3.85, "learning_rate": 1.108662380263881e-05, "loss": 0.0499, "step": 21000 }, { "epoch": 3.85, "learning_rate": 1.1085885765760533e-05, "loss": 0.0407, "step": 21001 }, { "epoch": 3.85, "learning_rate": 1.108514772289683e-05, "loss": 0.0391, "step": 21002 }, { "epoch": 3.85, "learning_rate": 1.1084409674051762e-05, "loss": 0.0384, "step": 21003 }, { "epoch": 3.85, "learning_rate": 1.1083671619229409e-05, "loss": 0.0475, "step": 21004 }, { "epoch": 3.85, "learning_rate": 1.1082933558433828e-05, "loss": 0.0377, "step": 21005 }, { "epoch": 3.85, "learning_rate": 1.1082195491669092e-05, "loss": 0.043, "step": 21006 }, { "epoch": 3.85, "learning_rate": 1.1081457418939271e-05, "loss": 0.0411, "step": 21007 }, { "epoch": 3.85, "learning_rate": 1.1080719340248431e-05, "loss": 0.0413, "step": 21008 }, { "epoch": 3.85, "learning_rate": 1.107998125560064e-05, "loss": 0.0453, "step": 21009 }, { "epoch": 3.85, "learning_rate": 1.1079243164999972e-05, "loss": 0.0446, "step": 21010 }, { "epoch": 3.85, "learning_rate": 1.1078505068450485e-05, "loss": 0.0472, "step": 21011 }, { "epoch": 3.85, "learning_rate": 1.107776696595625e-05, "loss": 0.0412, "step": 21012 }, { "epoch": 3.85, "learning_rate": 1.1077028857521345e-05, "loss": 0.0452, "step": 21013 }, { "epoch": 3.85, "learning_rate": 1.1076290743149827e-05, "loss": 0.0389, "step": 21014 }, { "epoch": 3.85, "learning_rate": 1.107555262284577e-05, "loss": 0.0324, "step": 21015 }, { "epoch": 3.85, "learning_rate": 1.1074814496613245e-05, "loss": 0.0399, "step": 21016 }, { "epoch": 3.85, "learning_rate": 1.1074076364456316e-05, "loss": 0.0489, "step": 21017 }, { "epoch": 3.85, "learning_rate": 1.1073338226379049e-05, "loss": 0.0489, "step": 21018 }, { "epoch": 3.85, "learning_rate": 1.1072600082385523e-05, "loss": 0.0475, "step": 21019 }, { "epoch": 3.85, "learning_rate": 1.1071861932479794e-05, "loss": 0.0434, "step": 21020 }, { "epoch": 3.85, "learning_rate": 1.107112377666594e-05, "loss": 0.0371, "step": 21021 }, { "epoch": 3.86, "learning_rate": 1.107038561494803e-05, "loss": 0.0534, "step": 21022 }, { "epoch": 3.86, "learning_rate": 1.1069647447330122e-05, "loss": 0.0389, "step": 21023 }, { "epoch": 3.86, "learning_rate": 1.1068909273816297e-05, "loss": 0.0381, "step": 21024 }, { "epoch": 3.86, "learning_rate": 1.1068171094410618e-05, "loss": 0.0489, "step": 21025 }, { "epoch": 3.86, "learning_rate": 1.1067432909117155e-05, "loss": 0.0493, "step": 21026 }, { "epoch": 3.86, "learning_rate": 1.1066694717939975e-05, "loss": 0.0443, "step": 21027 }, { "epoch": 3.86, "learning_rate": 1.1065956520883152e-05, "loss": 0.0417, "step": 21028 }, { "epoch": 3.86, "learning_rate": 1.106521831795075e-05, "loss": 0.0497, "step": 21029 }, { "epoch": 3.86, "learning_rate": 1.106448010914684e-05, "loss": 0.0367, "step": 21030 }, { "epoch": 3.86, "learning_rate": 1.106374189447549e-05, "loss": 0.0438, "step": 21031 }, { "epoch": 3.86, "learning_rate": 1.106300367394077e-05, "loss": 0.0393, "step": 21032 }, { "epoch": 3.86, "learning_rate": 1.106226544754675e-05, "loss": 0.0451, "step": 21033 }, { "epoch": 3.86, "learning_rate": 1.1061527215297496e-05, "loss": 0.0363, "step": 21034 }, { "epoch": 3.86, "learning_rate": 1.1060788977197081e-05, "loss": 0.0382, "step": 21035 }, { "epoch": 3.86, "learning_rate": 1.106005073324957e-05, "loss": 0.0374, "step": 21036 }, { "epoch": 3.86, "learning_rate": 1.1059312483459039e-05, "loss": 0.0462, "step": 21037 }, { "epoch": 3.86, "learning_rate": 1.1058574227829549e-05, "loss": 0.0439, "step": 21038 }, { "epoch": 3.86, "learning_rate": 1.1057835966365173e-05, "loss": 0.0387, "step": 21039 }, { "epoch": 3.86, "learning_rate": 1.105709769906998e-05, "loss": 0.0455, "step": 21040 }, { "epoch": 3.86, "learning_rate": 1.1056359425948042e-05, "loss": 0.0366, "step": 21041 }, { "epoch": 3.86, "learning_rate": 1.1055621147003426e-05, "loss": 0.0398, "step": 21042 }, { "epoch": 3.86, "learning_rate": 1.10548828622402e-05, "loss": 0.0384, "step": 21043 }, { "epoch": 3.86, "learning_rate": 1.1054144571662435e-05, "loss": 0.0473, "step": 21044 }, { "epoch": 3.86, "learning_rate": 1.1053406275274203e-05, "loss": 0.0383, "step": 21045 }, { "epoch": 3.86, "learning_rate": 1.105266797307957e-05, "loss": 0.0378, "step": 21046 }, { "epoch": 3.86, "learning_rate": 1.1051929665082602e-05, "loss": 0.0462, "step": 21047 }, { "epoch": 3.86, "learning_rate": 1.1051191351287376e-05, "loss": 0.0389, "step": 21048 }, { "epoch": 3.86, "learning_rate": 1.1050453031697958e-05, "loss": 0.0367, "step": 21049 }, { "epoch": 3.86, "learning_rate": 1.1049714706318422e-05, "loss": 0.0389, "step": 21050 }, { "epoch": 3.86, "learning_rate": 1.1048976375152832e-05, "loss": 0.0429, "step": 21051 }, { "epoch": 3.86, "learning_rate": 1.1048238038205256e-05, "loss": 0.0429, "step": 21052 }, { "epoch": 3.86, "learning_rate": 1.104749969547977e-05, "loss": 0.0459, "step": 21053 }, { "epoch": 3.86, "learning_rate": 1.1046761346980444e-05, "loss": 0.0412, "step": 21054 }, { "epoch": 3.86, "learning_rate": 1.1046022992711343e-05, "loss": 0.0454, "step": 21055 }, { "epoch": 3.86, "learning_rate": 1.1045284632676535e-05, "loss": 0.0505, "step": 21056 }, { "epoch": 3.86, "learning_rate": 1.1044546266880097e-05, "loss": 0.047, "step": 21057 }, { "epoch": 3.86, "learning_rate": 1.1043807895326095e-05, "loss": 0.0396, "step": 21058 }, { "epoch": 3.86, "learning_rate": 1.10430695180186e-05, "loss": 0.048, "step": 21059 }, { "epoch": 3.86, "learning_rate": 1.1042331134961682e-05, "loss": 0.0451, "step": 21060 }, { "epoch": 3.86, "learning_rate": 1.1041592746159408e-05, "loss": 0.0397, "step": 21061 }, { "epoch": 3.86, "learning_rate": 1.104085435161585e-05, "loss": 0.0524, "step": 21062 }, { "epoch": 3.86, "learning_rate": 1.1040115951335084e-05, "loss": 0.0394, "step": 21063 }, { "epoch": 3.86, "learning_rate": 1.103937754532117e-05, "loss": 0.0462, "step": 21064 }, { "epoch": 3.86, "learning_rate": 1.1038639133578181e-05, "loss": 0.0422, "step": 21065 }, { "epoch": 3.86, "learning_rate": 1.1037900716110193e-05, "loss": 0.0409, "step": 21066 }, { "epoch": 3.86, "learning_rate": 1.103716229292127e-05, "loss": 0.0412, "step": 21067 }, { "epoch": 3.86, "learning_rate": 1.1036423864015485e-05, "loss": 0.044, "step": 21068 }, { "epoch": 3.86, "learning_rate": 1.1035685429396906e-05, "loss": 0.0442, "step": 21069 }, { "epoch": 3.86, "learning_rate": 1.1034946989069606e-05, "loss": 0.0393, "step": 21070 }, { "epoch": 3.86, "learning_rate": 1.103420854303765e-05, "loss": 0.046, "step": 21071 }, { "epoch": 3.86, "learning_rate": 1.1033470091305119e-05, "loss": 0.0393, "step": 21072 }, { "epoch": 3.86, "learning_rate": 1.1032731633876071e-05, "loss": 0.0452, "step": 21073 }, { "epoch": 3.86, "learning_rate": 1.1031993170754584e-05, "loss": 0.0396, "step": 21074 }, { "epoch": 3.86, "learning_rate": 1.103125470194473e-05, "loss": 0.0375, "step": 21075 }, { "epoch": 3.87, "learning_rate": 1.103051622745057e-05, "loss": 0.0428, "step": 21076 }, { "epoch": 3.87, "learning_rate": 1.1029777747276182e-05, "loss": 0.0378, "step": 21077 }, { "epoch": 3.87, "learning_rate": 1.1029039261425636e-05, "loss": 0.0467, "step": 21078 }, { "epoch": 3.87, "learning_rate": 1.1028300769903002e-05, "loss": 0.0464, "step": 21079 }, { "epoch": 3.87, "learning_rate": 1.102756227271235e-05, "loss": 0.0436, "step": 21080 }, { "epoch": 3.87, "learning_rate": 1.1026823769857751e-05, "loss": 0.0467, "step": 21081 }, { "epoch": 3.87, "learning_rate": 1.1026085261343271e-05, "loss": 0.0422, "step": 21082 }, { "epoch": 3.87, "learning_rate": 1.1025346747172989e-05, "loss": 0.0494, "step": 21083 }, { "epoch": 3.87, "learning_rate": 1.1024608227350973e-05, "loss": 0.0423, "step": 21084 }, { "epoch": 3.87, "learning_rate": 1.1023869701881288e-05, "loss": 0.0432, "step": 21085 }, { "epoch": 3.87, "learning_rate": 1.1023131170768013e-05, "loss": 0.0493, "step": 21086 }, { "epoch": 3.87, "learning_rate": 1.1022392634015216e-05, "loss": 0.0378, "step": 21087 }, { "epoch": 3.87, "learning_rate": 1.1021654091626963e-05, "loss": 0.0386, "step": 21088 }, { "epoch": 3.87, "learning_rate": 1.1020915543607329e-05, "loss": 0.0402, "step": 21089 }, { "epoch": 3.87, "learning_rate": 1.102017698996039e-05, "loss": 0.0427, "step": 21090 }, { "epoch": 3.87, "learning_rate": 1.1019438430690203e-05, "loss": 0.0469, "step": 21091 }, { "epoch": 3.87, "learning_rate": 1.1018699865800853e-05, "loss": 0.0447, "step": 21092 }, { "epoch": 3.87, "learning_rate": 1.1017961295296404e-05, "loss": 0.0466, "step": 21093 }, { "epoch": 3.87, "learning_rate": 1.1017222719180929e-05, "loss": 0.0504, "step": 21094 }, { "epoch": 3.87, "learning_rate": 1.1016484137458497e-05, "loss": 0.0467, "step": 21095 }, { "epoch": 3.87, "learning_rate": 1.1015745550133186e-05, "loss": 0.0436, "step": 21096 }, { "epoch": 3.87, "learning_rate": 1.1015006957209057e-05, "loss": 0.037, "step": 21097 }, { "epoch": 3.87, "learning_rate": 1.1014268358690186e-05, "loss": 0.0404, "step": 21098 }, { "epoch": 3.87, "learning_rate": 1.1013529754580646e-05, "loss": 0.0377, "step": 21099 }, { "epoch": 3.87, "learning_rate": 1.1012791144884502e-05, "loss": 0.0326, "step": 21100 }, { "epoch": 3.87, "learning_rate": 1.1012052529605835e-05, "loss": 0.0362, "step": 21101 }, { "epoch": 3.87, "learning_rate": 1.1011313908748706e-05, "loss": 0.0437, "step": 21102 }, { "epoch": 3.87, "learning_rate": 1.1010575282317195e-05, "loss": 0.045, "step": 21103 }, { "epoch": 3.87, "learning_rate": 1.1009836650315364e-05, "loss": 0.04, "step": 21104 }, { "epoch": 3.87, "learning_rate": 1.1009098012747298e-05, "loss": 0.0415, "step": 21105 }, { "epoch": 3.87, "learning_rate": 1.1008359369617052e-05, "loss": 0.044, "step": 21106 }, { "epoch": 3.87, "learning_rate": 1.1007620720928709e-05, "loss": 0.0414, "step": 21107 }, { "epoch": 3.87, "learning_rate": 1.100688206668634e-05, "loss": 0.0392, "step": 21108 }, { "epoch": 3.87, "learning_rate": 1.100614340689401e-05, "loss": 0.0428, "step": 21109 }, { "epoch": 3.87, "learning_rate": 1.1005404741555794e-05, "loss": 0.0479, "step": 21110 }, { "epoch": 3.87, "learning_rate": 1.1004666070675765e-05, "loss": 0.0333, "step": 21111 }, { "epoch": 3.87, "learning_rate": 1.100392739425799e-05, "loss": 0.0444, "step": 21112 }, { "epoch": 3.87, "learning_rate": 1.1003188712306545e-05, "loss": 0.0431, "step": 21113 }, { "epoch": 3.87, "learning_rate": 1.1002450024825504e-05, "loss": 0.0345, "step": 21114 }, { "epoch": 3.87, "learning_rate": 1.100171133181893e-05, "loss": 0.0471, "step": 21115 }, { "epoch": 3.87, "learning_rate": 1.10009726332909e-05, "loss": 0.037, "step": 21116 }, { "epoch": 3.87, "learning_rate": 1.1000233929245488e-05, "loss": 0.0445, "step": 21117 }, { "epoch": 3.87, "learning_rate": 1.0999495219686762e-05, "loss": 0.043, "step": 21118 }, { "epoch": 3.87, "learning_rate": 1.0998756504618794e-05, "loss": 0.0388, "step": 21119 }, { "epoch": 3.87, "learning_rate": 1.099801778404566e-05, "loss": 0.0439, "step": 21120 }, { "epoch": 3.87, "learning_rate": 1.0997279057971426e-05, "loss": 0.0416, "step": 21121 }, { "epoch": 3.87, "learning_rate": 1.0996540326400168e-05, "loss": 0.042, "step": 21122 }, { "epoch": 3.87, "learning_rate": 1.0995801589335955e-05, "loss": 0.0376, "step": 21123 }, { "epoch": 3.87, "learning_rate": 1.0995062846782861e-05, "loss": 0.0393, "step": 21124 }, { "epoch": 3.87, "learning_rate": 1.0994324098744954e-05, "loss": 0.0464, "step": 21125 }, { "epoch": 3.87, "learning_rate": 1.0993585345226314e-05, "loss": 0.0434, "step": 21126 }, { "epoch": 3.87, "learning_rate": 1.0992846586231008e-05, "loss": 0.0398, "step": 21127 }, { "epoch": 3.87, "learning_rate": 1.0992107821763105e-05, "loss": 0.0434, "step": 21128 }, { "epoch": 3.87, "learning_rate": 1.0991369051826683e-05, "loss": 0.0406, "step": 21129 }, { "epoch": 3.87, "learning_rate": 1.099063027642581e-05, "loss": 0.0414, "step": 21130 }, { "epoch": 3.88, "learning_rate": 1.0989891495564562e-05, "loss": 0.0376, "step": 21131 }, { "epoch": 3.88, "learning_rate": 1.098915270924701e-05, "loss": 0.039, "step": 21132 }, { "epoch": 3.88, "learning_rate": 1.0988413917477218e-05, "loss": 0.0419, "step": 21133 }, { "epoch": 3.88, "learning_rate": 1.0987675120259271e-05, "loss": 0.0453, "step": 21134 }, { "epoch": 3.88, "learning_rate": 1.0986936317597235e-05, "loss": 0.0363, "step": 21135 }, { "epoch": 3.88, "learning_rate": 1.0986197509495182e-05, "loss": 0.0373, "step": 21136 }, { "epoch": 3.88, "learning_rate": 1.0985458695957184e-05, "loss": 0.0428, "step": 21137 }, { "epoch": 3.88, "learning_rate": 1.0984719876987317e-05, "loss": 0.0352, "step": 21138 }, { "epoch": 3.88, "learning_rate": 1.0983981052589651e-05, "loss": 0.0268, "step": 21139 }, { "epoch": 3.88, "learning_rate": 1.0983242222768257e-05, "loss": 0.042, "step": 21140 }, { "epoch": 3.88, "learning_rate": 1.098250338752721e-05, "loss": 0.0376, "step": 21141 }, { "epoch": 3.88, "learning_rate": 1.0981764546870578e-05, "loss": 0.0563, "step": 21142 }, { "epoch": 3.88, "learning_rate": 1.0981025700802442e-05, "loss": 0.0475, "step": 21143 }, { "epoch": 3.88, "learning_rate": 1.0980286849326867e-05, "loss": 0.0357, "step": 21144 }, { "epoch": 3.88, "learning_rate": 1.0979547992447927e-05, "loss": 0.0404, "step": 21145 }, { "epoch": 3.88, "learning_rate": 1.0978809130169696e-05, "loss": 0.0408, "step": 21146 }, { "epoch": 3.88, "learning_rate": 1.0978070262496248e-05, "loss": 0.0433, "step": 21147 }, { "epoch": 3.88, "learning_rate": 1.097733138943165e-05, "loss": 0.0334, "step": 21148 }, { "epoch": 3.88, "learning_rate": 1.0976592510979982e-05, "loss": 0.0455, "step": 21149 }, { "epoch": 3.88, "learning_rate": 1.0975853627145313e-05, "loss": 0.0382, "step": 21150 }, { "epoch": 3.88, "learning_rate": 1.0975114737931713e-05, "loss": 0.0393, "step": 21151 }, { "epoch": 3.88, "learning_rate": 1.0974375843343261e-05, "loss": 0.0413, "step": 21152 }, { "epoch": 3.88, "learning_rate": 1.0973636943384024e-05, "loss": 0.0471, "step": 21153 }, { "epoch": 3.88, "learning_rate": 1.0972898038058077e-05, "loss": 0.0347, "step": 21154 }, { "epoch": 3.88, "learning_rate": 1.0972159127369497e-05, "loss": 0.0491, "step": 21155 }, { "epoch": 3.88, "learning_rate": 1.097142021132235e-05, "loss": 0.0427, "step": 21156 }, { "epoch": 3.88, "learning_rate": 1.097068128992071e-05, "loss": 0.0408, "step": 21157 }, { "epoch": 3.88, "learning_rate": 1.0969942363168657e-05, "loss": 0.0417, "step": 21158 }, { "epoch": 3.88, "learning_rate": 1.0969203431070257e-05, "loss": 0.0377, "step": 21159 }, { "epoch": 3.88, "learning_rate": 1.0968464493629585e-05, "loss": 0.04, "step": 21160 }, { "epoch": 3.88, "learning_rate": 1.0967725550850716e-05, "loss": 0.0422, "step": 21161 }, { "epoch": 3.88, "learning_rate": 1.0966986602737717e-05, "loss": 0.0346, "step": 21162 }, { "epoch": 3.88, "learning_rate": 1.0966247649294668e-05, "loss": 0.0446, "step": 21163 }, { "epoch": 3.88, "learning_rate": 1.0965508690525637e-05, "loss": 0.0497, "step": 21164 }, { "epoch": 3.88, "learning_rate": 1.09647697264347e-05, "loss": 0.0401, "step": 21165 }, { "epoch": 3.88, "learning_rate": 1.096403075702593e-05, "loss": 0.0451, "step": 21166 }, { "epoch": 3.88, "learning_rate": 1.0963291782303402e-05, "loss": 0.041, "step": 21167 }, { "epoch": 3.88, "learning_rate": 1.0962552802271184e-05, "loss": 0.0435, "step": 21168 }, { "epoch": 3.88, "learning_rate": 1.0961813816933355e-05, "loss": 0.0484, "step": 21169 }, { "epoch": 3.88, "learning_rate": 1.0961074826293986e-05, "loss": 0.0416, "step": 21170 }, { "epoch": 3.88, "learning_rate": 1.0960335830357147e-05, "loss": 0.0331, "step": 21171 }, { "epoch": 3.88, "learning_rate": 1.0959596829126918e-05, "loss": 0.0387, "step": 21172 }, { "epoch": 3.88, "learning_rate": 1.0958857822607365e-05, "loss": 0.045, "step": 21173 }, { "epoch": 3.88, "learning_rate": 1.0958118810802567e-05, "loss": 0.0409, "step": 21174 }, { "epoch": 3.88, "learning_rate": 1.0957379793716596e-05, "loss": 0.0505, "step": 21175 }, { "epoch": 3.88, "learning_rate": 1.0956640771353525e-05, "loss": 0.0366, "step": 21176 }, { "epoch": 3.88, "learning_rate": 1.0955901743717427e-05, "loss": 0.0397, "step": 21177 }, { "epoch": 3.88, "learning_rate": 1.0955162710812378e-05, "loss": 0.0372, "step": 21178 }, { "epoch": 3.88, "learning_rate": 1.0954423672642447e-05, "loss": 0.0393, "step": 21179 }, { "epoch": 3.88, "learning_rate": 1.0953684629211714e-05, "loss": 0.0346, "step": 21180 }, { "epoch": 3.88, "learning_rate": 1.0952945580524243e-05, "loss": 0.0392, "step": 21181 }, { "epoch": 3.88, "learning_rate": 1.0952206526584122e-05, "loss": 0.0504, "step": 21182 }, { "epoch": 3.88, "learning_rate": 1.0951467467395411e-05, "loss": 0.046, "step": 21183 }, { "epoch": 3.88, "learning_rate": 1.0950728402962187e-05, "loss": 0.0422, "step": 21184 }, { "epoch": 3.89, "learning_rate": 1.094998933328853e-05, "loss": 0.0378, "step": 21185 }, { "epoch": 3.89, "learning_rate": 1.0949250258378507e-05, "loss": 0.0486, "step": 21186 }, { "epoch": 3.89, "learning_rate": 1.0948511178236197e-05, "loss": 0.0465, "step": 21187 }, { "epoch": 3.89, "learning_rate": 1.0947772092865669e-05, "loss": 0.0378, "step": 21188 }, { "epoch": 3.89, "learning_rate": 1.0947033002271001e-05, "loss": 0.0375, "step": 21189 }, { "epoch": 3.89, "learning_rate": 1.0946293906456262e-05, "loss": 0.0478, "step": 21190 }, { "epoch": 3.89, "learning_rate": 1.0945554805425532e-05, "loss": 0.0348, "step": 21191 }, { "epoch": 3.89, "learning_rate": 1.094481569918288e-05, "loss": 0.05, "step": 21192 }, { "epoch": 3.89, "learning_rate": 1.094407658773238e-05, "loss": 0.0402, "step": 21193 }, { "epoch": 3.89, "learning_rate": 1.0943337471078113e-05, "loss": 0.0419, "step": 21194 }, { "epoch": 3.89, "learning_rate": 1.0942598349224143e-05, "loss": 0.0388, "step": 21195 }, { "epoch": 3.89, "learning_rate": 1.0941859222174549e-05, "loss": 0.0359, "step": 21196 }, { "epoch": 3.89, "learning_rate": 1.0941120089933409e-05, "loss": 0.0453, "step": 21197 }, { "epoch": 3.89, "learning_rate": 1.0940380952504788e-05, "loss": 0.0427, "step": 21198 }, { "epoch": 3.89, "learning_rate": 1.0939641809892767e-05, "loss": 0.0414, "step": 21199 }, { "epoch": 3.89, "learning_rate": 1.0938902662101421e-05, "loss": 0.042, "step": 21200 }, { "epoch": 3.89, "learning_rate": 1.0938163509134817e-05, "loss": 0.0514, "step": 21201 }, { "epoch": 3.89, "learning_rate": 1.0937424350997035e-05, "loss": 0.042, "step": 21202 }, { "epoch": 3.89, "learning_rate": 1.093668518769215e-05, "loss": 0.0429, "step": 21203 }, { "epoch": 3.89, "learning_rate": 1.0935946019224233e-05, "loss": 0.0387, "step": 21204 }, { "epoch": 3.89, "learning_rate": 1.093520684559736e-05, "loss": 0.0343, "step": 21205 }, { "epoch": 3.89, "learning_rate": 1.0934467666815605e-05, "loss": 0.0398, "step": 21206 }, { "epoch": 3.89, "learning_rate": 1.0933728482883041e-05, "loss": 0.0391, "step": 21207 }, { "epoch": 3.89, "learning_rate": 1.0932989293803746e-05, "loss": 0.0398, "step": 21208 }, { "epoch": 3.89, "learning_rate": 1.0932250099581791e-05, "loss": 0.0411, "step": 21209 }, { "epoch": 3.89, "learning_rate": 1.0931510900221249e-05, "loss": 0.0323, "step": 21210 }, { "epoch": 3.89, "learning_rate": 1.0930771695726201e-05, "loss": 0.0408, "step": 21211 }, { "epoch": 3.89, "learning_rate": 1.0930032486100717e-05, "loss": 0.0431, "step": 21212 }, { "epoch": 3.89, "learning_rate": 1.0929293271348871e-05, "loss": 0.0413, "step": 21213 }, { "epoch": 3.89, "learning_rate": 1.0928554051474741e-05, "loss": 0.0392, "step": 21214 }, { "epoch": 3.89, "learning_rate": 1.0927814826482397e-05, "loss": 0.0362, "step": 21215 }, { "epoch": 3.89, "learning_rate": 1.0927075596375916e-05, "loss": 0.0389, "step": 21216 }, { "epoch": 3.89, "learning_rate": 1.0926336361159373e-05, "loss": 0.0418, "step": 21217 }, { "epoch": 3.89, "learning_rate": 1.0925597120836844e-05, "loss": 0.0446, "step": 21218 }, { "epoch": 3.89, "learning_rate": 1.0924857875412398e-05, "loss": 0.0438, "step": 21219 }, { "epoch": 3.89, "learning_rate": 1.0924118624890117e-05, "loss": 0.0426, "step": 21220 }, { "epoch": 3.89, "learning_rate": 1.092337936927407e-05, "loss": 0.0479, "step": 21221 }, { "epoch": 3.89, "learning_rate": 1.0922640108568337e-05, "loss": 0.0426, "step": 21222 }, { "epoch": 3.89, "learning_rate": 1.0921900842776989e-05, "loss": 0.0465, "step": 21223 }, { "epoch": 3.89, "learning_rate": 1.0921161571904101e-05, "loss": 0.0433, "step": 21224 }, { "epoch": 3.89, "learning_rate": 1.092042229595375e-05, "loss": 0.0401, "step": 21225 }, { "epoch": 3.89, "learning_rate": 1.091968301493001e-05, "loss": 0.0454, "step": 21226 }, { "epoch": 3.89, "learning_rate": 1.0918943728836954e-05, "loss": 0.0368, "step": 21227 }, { "epoch": 3.89, "learning_rate": 1.0918204437678659e-05, "loss": 0.0416, "step": 21228 }, { "epoch": 3.89, "learning_rate": 1.0917465141459203e-05, "loss": 0.0437, "step": 21229 }, { "epoch": 3.89, "learning_rate": 1.0916725840182653e-05, "loss": 0.0435, "step": 21230 }, { "epoch": 3.89, "learning_rate": 1.0915986533853091e-05, "loss": 0.0417, "step": 21231 }, { "epoch": 3.89, "learning_rate": 1.091524722247459e-05, "loss": 0.0345, "step": 21232 }, { "epoch": 3.89, "learning_rate": 1.0914507906051224e-05, "loss": 0.049, "step": 21233 }, { "epoch": 3.89, "learning_rate": 1.0913768584587068e-05, "loss": 0.0419, "step": 21234 }, { "epoch": 3.89, "learning_rate": 1.0913029258086202e-05, "loss": 0.0444, "step": 21235 }, { "epoch": 3.89, "learning_rate": 1.0912289926552696e-05, "loss": 0.0439, "step": 21236 }, { "epoch": 3.89, "learning_rate": 1.0911550589990622e-05, "loss": 0.0339, "step": 21237 }, { "epoch": 3.89, "learning_rate": 1.0910811248404064e-05, "loss": 0.0451, "step": 21238 }, { "epoch": 3.89, "learning_rate": 1.0910071901797093e-05, "loss": 0.0383, "step": 21239 }, { "epoch": 3.9, "learning_rate": 1.0909332550173783e-05, "loss": 0.0366, "step": 21240 }, { "epoch": 3.9, "learning_rate": 1.0908593193538213e-05, "loss": 0.0435, "step": 21241 }, { "epoch": 3.9, "learning_rate": 1.0907853831894453e-05, "loss": 0.0411, "step": 21242 }, { "epoch": 3.9, "learning_rate": 1.090711446524658e-05, "loss": 0.0386, "step": 21243 }, { "epoch": 3.9, "learning_rate": 1.0906375093598678e-05, "loss": 0.0418, "step": 21244 }, { "epoch": 3.9, "learning_rate": 1.0905635716954808e-05, "loss": 0.0391, "step": 21245 }, { "epoch": 3.9, "learning_rate": 1.0904896335319055e-05, "loss": 0.0353, "step": 21246 }, { "epoch": 3.9, "learning_rate": 1.0904156948695496e-05, "loss": 0.0459, "step": 21247 }, { "epoch": 3.9, "learning_rate": 1.0903417557088195e-05, "loss": 0.0384, "step": 21248 }, { "epoch": 3.9, "learning_rate": 1.090267816050124e-05, "loss": 0.0424, "step": 21249 }, { "epoch": 3.9, "learning_rate": 1.0901938758938704e-05, "loss": 0.0357, "step": 21250 }, { "epoch": 3.9, "learning_rate": 1.0901199352404658e-05, "loss": 0.0395, "step": 21251 }, { "epoch": 3.9, "learning_rate": 1.090045994090318e-05, "loss": 0.0371, "step": 21252 }, { "epoch": 3.9, "learning_rate": 1.0899720524438347e-05, "loss": 0.0356, "step": 21253 }, { "epoch": 3.9, "learning_rate": 1.089898110301423e-05, "loss": 0.0389, "step": 21254 }, { "epoch": 3.9, "learning_rate": 1.0898241676634908e-05, "loss": 0.038, "step": 21255 }, { "epoch": 3.9, "learning_rate": 1.0897502245304462e-05, "loss": 0.0383, "step": 21256 }, { "epoch": 3.9, "learning_rate": 1.089676280902696e-05, "loss": 0.0385, "step": 21257 }, { "epoch": 3.9, "learning_rate": 1.0896023367806478e-05, "loss": 0.0424, "step": 21258 }, { "epoch": 3.9, "learning_rate": 1.0895283921647098e-05, "loss": 0.0498, "step": 21259 }, { "epoch": 3.9, "learning_rate": 1.089454447055289e-05, "loss": 0.0461, "step": 21260 }, { "epoch": 3.9, "learning_rate": 1.0893805014527929e-05, "loss": 0.043, "step": 21261 }, { "epoch": 3.9, "learning_rate": 1.08930655535763e-05, "loss": 0.0387, "step": 21262 }, { "epoch": 3.9, "learning_rate": 1.0892326087702068e-05, "loss": 0.0451, "step": 21263 }, { "epoch": 3.9, "learning_rate": 1.0891586616909316e-05, "loss": 0.0409, "step": 21264 }, { "epoch": 3.9, "learning_rate": 1.0890847141202116e-05, "loss": 0.0373, "step": 21265 }, { "epoch": 3.9, "learning_rate": 1.0890107660584548e-05, "loss": 0.0407, "step": 21266 }, { "epoch": 3.9, "learning_rate": 1.0889368175060685e-05, "loss": 0.0467, "step": 21267 }, { "epoch": 3.9, "learning_rate": 1.0888628684634603e-05, "loss": 0.042, "step": 21268 }, { "epoch": 3.9, "learning_rate": 1.0887889189310381e-05, "loss": 0.0349, "step": 21269 }, { "epoch": 3.9, "learning_rate": 1.0887149689092089e-05, "loss": 0.0392, "step": 21270 }, { "epoch": 3.9, "learning_rate": 1.0886410183983812e-05, "loss": 0.0461, "step": 21271 }, { "epoch": 3.9, "learning_rate": 1.0885670673989616e-05, "loss": 0.0431, "step": 21272 }, { "epoch": 3.9, "learning_rate": 1.0884931159113585e-05, "loss": 0.0418, "step": 21273 }, { "epoch": 3.9, "learning_rate": 1.0884191639359793e-05, "loss": 0.0464, "step": 21274 }, { "epoch": 3.9, "learning_rate": 1.0883452114732316e-05, "loss": 0.0439, "step": 21275 }, { "epoch": 3.9, "learning_rate": 1.088271258523523e-05, "loss": 0.0463, "step": 21276 }, { "epoch": 3.9, "learning_rate": 1.0881973050872612e-05, "loss": 0.0364, "step": 21277 }, { "epoch": 3.9, "learning_rate": 1.088123351164854e-05, "loss": 0.0468, "step": 21278 }, { "epoch": 3.9, "learning_rate": 1.0880493967567084e-05, "loss": 0.0388, "step": 21279 }, { "epoch": 3.9, "learning_rate": 1.0879754418632327e-05, "loss": 0.0491, "step": 21280 }, { "epoch": 3.9, "learning_rate": 1.087901486484834e-05, "loss": 0.0424, "step": 21281 }, { "epoch": 3.9, "learning_rate": 1.0878275306219206e-05, "loss": 0.0402, "step": 21282 }, { "epoch": 3.9, "learning_rate": 1.0877535742748996e-05, "loss": 0.0488, "step": 21283 }, { "epoch": 3.9, "learning_rate": 1.087679617444179e-05, "loss": 0.0373, "step": 21284 }, { "epoch": 3.9, "learning_rate": 1.0876056601301663e-05, "loss": 0.0374, "step": 21285 }, { "epoch": 3.9, "learning_rate": 1.087531702333269e-05, "loss": 0.0412, "step": 21286 }, { "epoch": 3.9, "learning_rate": 1.0874577440538947e-05, "loss": 0.05, "step": 21287 }, { "epoch": 3.9, "learning_rate": 1.0873837852924516e-05, "loss": 0.0427, "step": 21288 }, { "epoch": 3.9, "learning_rate": 1.0873098260493474e-05, "loss": 0.0443, "step": 21289 }, { "epoch": 3.9, "learning_rate": 1.0872358663249887e-05, "loss": 0.0332, "step": 21290 }, { "epoch": 3.9, "learning_rate": 1.087161906119784e-05, "loss": 0.0424, "step": 21291 }, { "epoch": 3.9, "learning_rate": 1.0870879454341409e-05, "loss": 0.041, "step": 21292 }, { "epoch": 3.9, "learning_rate": 1.0870139842684673e-05, "loss": 0.0431, "step": 21293 }, { "epoch": 3.91, "learning_rate": 1.0869400226231704e-05, "loss": 0.0443, "step": 21294 }, { "epoch": 3.91, "learning_rate": 1.0868660604986578e-05, "loss": 0.0424, "step": 21295 }, { "epoch": 3.91, "learning_rate": 1.0867920978953375e-05, "loss": 0.0362, "step": 21296 }, { "epoch": 3.91, "learning_rate": 1.0867181348136174e-05, "loss": 0.0414, "step": 21297 }, { "epoch": 3.91, "learning_rate": 1.0866441712539047e-05, "loss": 0.0441, "step": 21298 }, { "epoch": 3.91, "learning_rate": 1.0865702072166073e-05, "loss": 0.0374, "step": 21299 }, { "epoch": 3.91, "learning_rate": 1.0864962427021331e-05, "loss": 0.0372, "step": 21300 }, { "epoch": 3.91, "learning_rate": 1.0864222777108895e-05, "loss": 0.0406, "step": 21301 }, { "epoch": 3.91, "learning_rate": 1.0863483122432843e-05, "loss": 0.0416, "step": 21302 }, { "epoch": 3.91, "learning_rate": 1.086274346299725e-05, "loss": 0.0415, "step": 21303 }, { "epoch": 3.91, "learning_rate": 1.0862003798806195e-05, "loss": 0.0418, "step": 21304 }, { "epoch": 3.91, "learning_rate": 1.0861264129863756e-05, "loss": 0.0359, "step": 21305 }, { "epoch": 3.91, "learning_rate": 1.0860524456174011e-05, "loss": 0.0332, "step": 21306 }, { "epoch": 3.91, "learning_rate": 1.085978477774103e-05, "loss": 0.0399, "step": 21307 }, { "epoch": 3.91, "learning_rate": 1.0859045094568898e-05, "loss": 0.0479, "step": 21308 }, { "epoch": 3.91, "learning_rate": 1.085830540666169e-05, "loss": 0.0345, "step": 21309 }, { "epoch": 3.91, "learning_rate": 1.085756571402348e-05, "loss": 0.043, "step": 21310 }, { "epoch": 3.91, "learning_rate": 1.085682601665835e-05, "loss": 0.0468, "step": 21311 }, { "epoch": 3.91, "learning_rate": 1.0856086314570375e-05, "loss": 0.0386, "step": 21312 }, { "epoch": 3.91, "learning_rate": 1.0855346607763633e-05, "loss": 0.0389, "step": 21313 }, { "epoch": 3.91, "learning_rate": 1.0854606896242195e-05, "loss": 0.0455, "step": 21314 }, { "epoch": 3.91, "learning_rate": 1.0853867180010152e-05, "loss": 0.0412, "step": 21315 }, { "epoch": 3.91, "learning_rate": 1.0853127459071566e-05, "loss": 0.0466, "step": 21316 }, { "epoch": 3.91, "learning_rate": 1.0852387733430522e-05, "loss": 0.0418, "step": 21317 }, { "epoch": 3.91, "learning_rate": 1.0851648003091102e-05, "loss": 0.0422, "step": 21318 }, { "epoch": 3.91, "learning_rate": 1.0850908268057374e-05, "loss": 0.0385, "step": 21319 }, { "epoch": 3.91, "learning_rate": 1.0850168528333422e-05, "loss": 0.0424, "step": 21320 }, { "epoch": 3.91, "learning_rate": 1.084942878392332e-05, "loss": 0.039, "step": 21321 }, { "epoch": 3.91, "learning_rate": 1.0848689034831148e-05, "loss": 0.0415, "step": 21322 }, { "epoch": 3.91, "learning_rate": 1.0847949281060979e-05, "loss": 0.033, "step": 21323 }, { "epoch": 3.91, "learning_rate": 1.0847209522616897e-05, "loss": 0.0406, "step": 21324 }, { "epoch": 3.91, "learning_rate": 1.0846469759502974e-05, "loss": 0.0435, "step": 21325 }, { "epoch": 3.91, "learning_rate": 1.0845729991723291e-05, "loss": 0.0528, "step": 21326 }, { "epoch": 3.91, "learning_rate": 1.0844990219281924e-05, "loss": 0.0419, "step": 21327 }, { "epoch": 3.91, "learning_rate": 1.084425044218295e-05, "loss": 0.0354, "step": 21328 }, { "epoch": 3.91, "learning_rate": 1.0843510660430447e-05, "loss": 0.0403, "step": 21329 }, { "epoch": 3.91, "learning_rate": 1.0842770874028499e-05, "loss": 0.0415, "step": 21330 }, { "epoch": 3.91, "learning_rate": 1.0842031082981175e-05, "loss": 0.0417, "step": 21331 }, { "epoch": 3.91, "learning_rate": 1.0841291287292555e-05, "loss": 0.0316, "step": 21332 }, { "epoch": 3.91, "learning_rate": 1.0840551486966719e-05, "loss": 0.0411, "step": 21333 }, { "epoch": 3.91, "learning_rate": 1.0839811682007742e-05, "loss": 0.0408, "step": 21334 }, { "epoch": 3.91, "learning_rate": 1.0839071872419704e-05, "loss": 0.0423, "step": 21335 }, { "epoch": 3.91, "learning_rate": 1.0838332058206685e-05, "loss": 0.0377, "step": 21336 }, { "epoch": 3.91, "learning_rate": 1.0837592239372757e-05, "loss": 0.0405, "step": 21337 }, { "epoch": 3.91, "learning_rate": 1.0836852415921998e-05, "loss": 0.0415, "step": 21338 }, { "epoch": 3.91, "learning_rate": 1.0836112587858496e-05, "loss": 0.0399, "step": 21339 }, { "epoch": 3.91, "learning_rate": 1.0835372755186315e-05, "loss": 0.0416, "step": 21340 }, { "epoch": 3.91, "learning_rate": 1.0834632917909544e-05, "loss": 0.0283, "step": 21341 }, { "epoch": 3.91, "learning_rate": 1.0833893076032258e-05, "loss": 0.0336, "step": 21342 }, { "epoch": 3.91, "learning_rate": 1.083315322955853e-05, "loss": 0.0354, "step": 21343 }, { "epoch": 3.91, "learning_rate": 1.0832413378492443e-05, "loss": 0.0478, "step": 21344 }, { "epoch": 3.91, "learning_rate": 1.0831673522838076e-05, "loss": 0.0449, "step": 21345 }, { "epoch": 3.91, "learning_rate": 1.0830933662599504e-05, "loss": 0.0426, "step": 21346 }, { "epoch": 3.91, "learning_rate": 1.0830193797780803e-05, "loss": 0.0313, "step": 21347 }, { "epoch": 3.91, "learning_rate": 1.082945392838606e-05, "loss": 0.0393, "step": 21348 }, { "epoch": 3.92, "learning_rate": 1.0828714054419343e-05, "loss": 0.0451, "step": 21349 }, { "epoch": 3.92, "learning_rate": 1.0827974175884737e-05, "loss": 0.0442, "step": 21350 }, { "epoch": 3.92, "learning_rate": 1.0827234292786317e-05, "loss": 0.0383, "step": 21351 }, { "epoch": 3.92, "learning_rate": 1.0826494405128163e-05, "loss": 0.0427, "step": 21352 }, { "epoch": 3.92, "learning_rate": 1.082575451291435e-05, "loss": 0.0364, "step": 21353 }, { "epoch": 3.92, "learning_rate": 1.0825014616148961e-05, "loss": 0.0419, "step": 21354 }, { "epoch": 3.92, "learning_rate": 1.0824274714836074e-05, "loss": 0.0403, "step": 21355 }, { "epoch": 3.92, "learning_rate": 1.082353480897976e-05, "loss": 0.0431, "step": 21356 }, { "epoch": 3.92, "learning_rate": 1.0822794898584109e-05, "loss": 0.0389, "step": 21357 }, { "epoch": 3.92, "learning_rate": 1.0822054983653187e-05, "loss": 0.0449, "step": 21358 }, { "epoch": 3.92, "learning_rate": 1.0821315064191084e-05, "loss": 0.0438, "step": 21359 }, { "epoch": 3.92, "learning_rate": 1.082057514020187e-05, "loss": 0.0358, "step": 21360 }, { "epoch": 3.92, "learning_rate": 1.0819835211689629e-05, "loss": 0.0427, "step": 21361 }, { "epoch": 3.92, "learning_rate": 1.0819095278658436e-05, "loss": 0.0373, "step": 21362 }, { "epoch": 3.92, "learning_rate": 1.081835534111237e-05, "loss": 0.0382, "step": 21363 }, { "epoch": 3.92, "learning_rate": 1.0817615399055513e-05, "loss": 0.0336, "step": 21364 }, { "epoch": 3.92, "learning_rate": 1.0816875452491936e-05, "loss": 0.0346, "step": 21365 }, { "epoch": 3.92, "learning_rate": 1.0816135501425726e-05, "loss": 0.0395, "step": 21366 }, { "epoch": 3.92, "learning_rate": 1.0815395545860957e-05, "loss": 0.0373, "step": 21367 }, { "epoch": 3.92, "learning_rate": 1.0814655585801708e-05, "loss": 0.0432, "step": 21368 }, { "epoch": 3.92, "learning_rate": 1.0813915621252057e-05, "loss": 0.0356, "step": 21369 }, { "epoch": 3.92, "learning_rate": 1.0813175652216088e-05, "loss": 0.0403, "step": 21370 }, { "epoch": 3.92, "learning_rate": 1.0812435678697872e-05, "loss": 0.0414, "step": 21371 }, { "epoch": 3.92, "learning_rate": 1.0811695700701495e-05, "loss": 0.0485, "step": 21372 }, { "epoch": 3.92, "learning_rate": 1.0810955718231026e-05, "loss": 0.0456, "step": 21373 }, { "epoch": 3.92, "learning_rate": 1.0810215731290557e-05, "loss": 0.0461, "step": 21374 }, { "epoch": 3.92, "learning_rate": 1.0809475739884159e-05, "loss": 0.0381, "step": 21375 }, { "epoch": 3.92, "learning_rate": 1.0808735744015909e-05, "loss": 0.0397, "step": 21376 }, { "epoch": 3.92, "learning_rate": 1.080799574368989e-05, "loss": 0.0367, "step": 21377 }, { "epoch": 3.92, "learning_rate": 1.0807255738910178e-05, "loss": 0.0353, "step": 21378 }, { "epoch": 3.92, "learning_rate": 1.0806515729680856e-05, "loss": 0.0383, "step": 21379 }, { "epoch": 3.92, "learning_rate": 1.0805775716005999e-05, "loss": 0.0372, "step": 21380 }, { "epoch": 3.92, "learning_rate": 1.0805035697889688e-05, "loss": 0.0399, "step": 21381 }, { "epoch": 3.92, "learning_rate": 1.0804295675335998e-05, "loss": 0.0414, "step": 21382 }, { "epoch": 3.92, "learning_rate": 1.0803555648349016e-05, "loss": 0.039, "step": 21383 }, { "epoch": 3.92, "learning_rate": 1.0802815616932815e-05, "loss": 0.0369, "step": 21384 }, { "epoch": 3.92, "learning_rate": 1.0802075581091474e-05, "loss": 0.0395, "step": 21385 }, { "epoch": 3.92, "learning_rate": 1.0801335540829075e-05, "loss": 0.0369, "step": 21386 }, { "epoch": 3.92, "learning_rate": 1.0800595496149695e-05, "loss": 0.0441, "step": 21387 }, { "epoch": 3.92, "learning_rate": 1.0799855447057417e-05, "loss": 0.0389, "step": 21388 }, { "epoch": 3.92, "learning_rate": 1.0799115393556314e-05, "loss": 0.0436, "step": 21389 }, { "epoch": 3.92, "learning_rate": 1.0798375335650468e-05, "loss": 0.0471, "step": 21390 }, { "epoch": 3.92, "learning_rate": 1.0797635273343958e-05, "loss": 0.04, "step": 21391 }, { "epoch": 3.92, "learning_rate": 1.0796895206640866e-05, "loss": 0.0409, "step": 21392 }, { "epoch": 3.92, "learning_rate": 1.0796155135545265e-05, "loss": 0.0444, "step": 21393 }, { "epoch": 3.92, "learning_rate": 1.0795415060061242e-05, "loss": 0.0416, "step": 21394 }, { "epoch": 3.92, "learning_rate": 1.0794674980192872e-05, "loss": 0.0384, "step": 21395 }, { "epoch": 3.92, "learning_rate": 1.0793934895944236e-05, "loss": 0.0489, "step": 21396 }, { "epoch": 3.92, "learning_rate": 1.079319480731941e-05, "loss": 0.0387, "step": 21397 }, { "epoch": 3.92, "learning_rate": 1.0792454714322478e-05, "loss": 0.0439, "step": 21398 }, { "epoch": 3.92, "learning_rate": 1.0791714616957513e-05, "loss": 0.0395, "step": 21399 }, { "epoch": 3.92, "learning_rate": 1.0790974515228601e-05, "loss": 0.0412, "step": 21400 }, { "epoch": 3.92, "learning_rate": 1.079023440913982e-05, "loss": 0.036, "step": 21401 }, { "epoch": 3.92, "learning_rate": 1.0789494298695246e-05, "loss": 0.0427, "step": 21402 }, { "epoch": 3.92, "learning_rate": 1.0788754183898962e-05, "loss": 0.0433, "step": 21403 }, { "epoch": 3.93, "learning_rate": 1.0788014064755048e-05, "loss": 0.0369, "step": 21404 }, { "epoch": 3.93, "learning_rate": 1.078727394126758e-05, "loss": 0.0395, "step": 21405 }, { "epoch": 3.93, "learning_rate": 1.078653381344064e-05, "loss": 0.0445, "step": 21406 }, { "epoch": 3.93, "learning_rate": 1.078579368127831e-05, "loss": 0.0386, "step": 21407 }, { "epoch": 3.93, "learning_rate": 1.0785053544784662e-05, "loss": 0.0393, "step": 21408 }, { "epoch": 3.93, "learning_rate": 1.0784313403963782e-05, "loss": 0.0422, "step": 21409 }, { "epoch": 3.93, "learning_rate": 1.078357325881975e-05, "loss": 0.043, "step": 21410 }, { "epoch": 3.93, "learning_rate": 1.078283310935664e-05, "loss": 0.0426, "step": 21411 }, { "epoch": 3.93, "learning_rate": 1.078209295557854e-05, "loss": 0.0416, "step": 21412 }, { "epoch": 3.93, "learning_rate": 1.0781352797489524e-05, "loss": 0.0448, "step": 21413 }, { "epoch": 3.93, "learning_rate": 1.0780612635093672e-05, "loss": 0.0352, "step": 21414 }, { "epoch": 3.93, "learning_rate": 1.0779872468395066e-05, "loss": 0.0347, "step": 21415 }, { "epoch": 3.93, "learning_rate": 1.0779132297397785e-05, "loss": 0.0393, "step": 21416 }, { "epoch": 3.93, "learning_rate": 1.0778392122105908e-05, "loss": 0.0463, "step": 21417 }, { "epoch": 3.93, "learning_rate": 1.0777651942523513e-05, "loss": 0.0435, "step": 21418 }, { "epoch": 3.93, "learning_rate": 1.0776911758654687e-05, "loss": 0.0463, "step": 21419 }, { "epoch": 3.93, "learning_rate": 1.07761715705035e-05, "loss": 0.0459, "step": 21420 }, { "epoch": 3.93, "learning_rate": 1.0775431378074039e-05, "loss": 0.0452, "step": 21421 }, { "epoch": 3.93, "learning_rate": 1.0774691181370385e-05, "loss": 0.0417, "step": 21422 }, { "epoch": 3.93, "learning_rate": 1.0773950980396613e-05, "loss": 0.0393, "step": 21423 }, { "epoch": 3.93, "learning_rate": 1.0773210775156802e-05, "loss": 0.0422, "step": 21424 }, { "epoch": 3.93, "learning_rate": 1.077247056565504e-05, "loss": 0.0483, "step": 21425 }, { "epoch": 3.93, "learning_rate": 1.07717303518954e-05, "loss": 0.0483, "step": 21426 }, { "epoch": 3.93, "learning_rate": 1.0770990133881965e-05, "loss": 0.0324, "step": 21427 }, { "epoch": 3.93, "learning_rate": 1.0770249911618814e-05, "loss": 0.0478, "step": 21428 }, { "epoch": 3.93, "learning_rate": 1.0769509685110026e-05, "loss": 0.0433, "step": 21429 }, { "epoch": 3.93, "learning_rate": 1.0768769454359684e-05, "loss": 0.0329, "step": 21430 }, { "epoch": 3.93, "learning_rate": 1.0768029219371867e-05, "loss": 0.0418, "step": 21431 }, { "epoch": 3.93, "learning_rate": 1.0767288980150655e-05, "loss": 0.0323, "step": 21432 }, { "epoch": 3.93, "learning_rate": 1.0766548736700123e-05, "loss": 0.0433, "step": 21433 }, { "epoch": 3.93, "learning_rate": 1.0765808489024365e-05, "loss": 0.0431, "step": 21434 }, { "epoch": 3.93, "learning_rate": 1.0765068237127447e-05, "loss": 0.0456, "step": 21435 }, { "epoch": 3.93, "learning_rate": 1.0764327981013456e-05, "loss": 0.0423, "step": 21436 }, { "epoch": 3.93, "learning_rate": 1.0763587720686473e-05, "loss": 0.0407, "step": 21437 }, { "epoch": 3.93, "learning_rate": 1.0762847456150576e-05, "loss": 0.0393, "step": 21438 }, { "epoch": 3.93, "learning_rate": 1.0762107187409845e-05, "loss": 0.0426, "step": 21439 }, { "epoch": 3.93, "learning_rate": 1.0761366914468363e-05, "loss": 0.0375, "step": 21440 }, { "epoch": 3.93, "learning_rate": 1.0760626637330206e-05, "loss": 0.0443, "step": 21441 }, { "epoch": 3.93, "learning_rate": 1.0759886355999459e-05, "loss": 0.045, "step": 21442 }, { "epoch": 3.93, "learning_rate": 1.07591460704802e-05, "loss": 0.0331, "step": 21443 }, { "epoch": 3.93, "learning_rate": 1.0758405780776508e-05, "loss": 0.0435, "step": 21444 }, { "epoch": 3.93, "learning_rate": 1.075766548689247e-05, "loss": 0.0374, "step": 21445 }, { "epoch": 3.93, "learning_rate": 1.0756925188832159e-05, "loss": 0.0491, "step": 21446 }, { "epoch": 3.93, "learning_rate": 1.075618488659966e-05, "loss": 0.0411, "step": 21447 }, { "epoch": 3.93, "learning_rate": 1.0755444580199052e-05, "loss": 0.0388, "step": 21448 }, { "epoch": 3.93, "learning_rate": 1.0754704269634417e-05, "loss": 0.0365, "step": 21449 }, { "epoch": 3.93, "learning_rate": 1.0753963954909831e-05, "loss": 0.0415, "step": 21450 }, { "epoch": 3.93, "learning_rate": 1.075322363602938e-05, "loss": 0.0366, "step": 21451 }, { "epoch": 3.93, "learning_rate": 1.0752483312997144e-05, "loss": 0.0394, "step": 21452 }, { "epoch": 3.93, "learning_rate": 1.0751742985817199e-05, "loss": 0.0427, "step": 21453 }, { "epoch": 3.93, "learning_rate": 1.0751002654493632e-05, "loss": 0.0434, "step": 21454 }, { "epoch": 3.93, "learning_rate": 1.075026231903052e-05, "loss": 0.0422, "step": 21455 }, { "epoch": 3.93, "learning_rate": 1.0749521979431946e-05, "loss": 0.0398, "step": 21456 }, { "epoch": 3.93, "learning_rate": 1.0748781635701985e-05, "loss": 0.0445, "step": 21457 }, { "epoch": 3.94, "learning_rate": 1.0748041287844726e-05, "loss": 0.0331, "step": 21458 }, { "epoch": 3.94, "learning_rate": 1.0747300935864245e-05, "loss": 0.0314, "step": 21459 }, { "epoch": 3.94, "learning_rate": 1.074656057976462e-05, "loss": 0.0379, "step": 21460 }, { "epoch": 3.94, "learning_rate": 1.0745820219549943e-05, "loss": 0.0402, "step": 21461 }, { "epoch": 3.94, "learning_rate": 1.074507985522428e-05, "loss": 0.0373, "step": 21462 }, { "epoch": 3.94, "learning_rate": 1.0744339486791723e-05, "loss": 0.0424, "step": 21463 }, { "epoch": 3.94, "learning_rate": 1.0743599114256351e-05, "loss": 0.0455, "step": 21464 }, { "epoch": 3.94, "learning_rate": 1.0742858737622242e-05, "loss": 0.0371, "step": 21465 }, { "epoch": 3.94, "learning_rate": 1.0742118356893476e-05, "loss": 0.0409, "step": 21466 }, { "epoch": 3.94, "learning_rate": 1.0741377972074139e-05, "loss": 0.0369, "step": 21467 }, { "epoch": 3.94, "learning_rate": 1.0740637583168304e-05, "loss": 0.0414, "step": 21468 }, { "epoch": 3.94, "learning_rate": 1.0739897190180066e-05, "loss": 0.0415, "step": 21469 }, { "epoch": 3.94, "learning_rate": 1.0739156793113491e-05, "loss": 0.0326, "step": 21470 }, { "epoch": 3.94, "learning_rate": 1.0738416391972668e-05, "loss": 0.0362, "step": 21471 }, { "epoch": 3.94, "learning_rate": 1.0737675986761678e-05, "loss": 0.0405, "step": 21472 }, { "epoch": 3.94, "learning_rate": 1.0736935577484598e-05, "loss": 0.041, "step": 21473 }, { "epoch": 3.94, "learning_rate": 1.0736195164145512e-05, "loss": 0.0334, "step": 21474 }, { "epoch": 3.94, "learning_rate": 1.0735454746748503e-05, "loss": 0.0416, "step": 21475 }, { "epoch": 3.94, "learning_rate": 1.0734714325297651e-05, "loss": 0.0367, "step": 21476 }, { "epoch": 3.94, "learning_rate": 1.0733973899797032e-05, "loss": 0.034, "step": 21477 }, { "epoch": 3.94, "learning_rate": 1.0733233470250737e-05, "loss": 0.0374, "step": 21478 }, { "epoch": 3.94, "learning_rate": 1.0732493036662839e-05, "loss": 0.0338, "step": 21479 }, { "epoch": 3.94, "learning_rate": 1.0731752599037421e-05, "loss": 0.0382, "step": 21480 }, { "epoch": 3.94, "learning_rate": 1.0731012157378568e-05, "loss": 0.0415, "step": 21481 }, { "epoch": 3.94, "learning_rate": 1.0730271711690359e-05, "loss": 0.0404, "step": 21482 }, { "epoch": 3.94, "learning_rate": 1.0729531261976873e-05, "loss": 0.0366, "step": 21483 }, { "epoch": 3.94, "learning_rate": 1.0728790808242195e-05, "loss": 0.0433, "step": 21484 }, { "epoch": 3.94, "learning_rate": 1.0728050350490407e-05, "loss": 0.0363, "step": 21485 }, { "epoch": 3.94, "learning_rate": 1.0727309888725584e-05, "loss": 0.0416, "step": 21486 }, { "epoch": 3.94, "learning_rate": 1.0726569422951816e-05, "loss": 0.0437, "step": 21487 }, { "epoch": 3.94, "learning_rate": 1.0725828953173177e-05, "loss": 0.0387, "step": 21488 }, { "epoch": 3.94, "learning_rate": 1.0725088479393752e-05, "loss": 0.0438, "step": 21489 }, { "epoch": 3.94, "learning_rate": 1.0724348001617626e-05, "loss": 0.0434, "step": 21490 }, { "epoch": 3.94, "learning_rate": 1.0723607519848873e-05, "loss": 0.0412, "step": 21491 }, { "epoch": 3.94, "learning_rate": 1.072286703409158e-05, "loss": 0.0408, "step": 21492 }, { "epoch": 3.94, "learning_rate": 1.0722126544349826e-05, "loss": 0.0383, "step": 21493 }, { "epoch": 3.94, "learning_rate": 1.0721386050627695e-05, "loss": 0.0434, "step": 21494 }, { "epoch": 3.94, "learning_rate": 1.0720645552929266e-05, "loss": 0.0355, "step": 21495 }, { "epoch": 3.94, "learning_rate": 1.0719905051258625e-05, "loss": 0.0405, "step": 21496 }, { "epoch": 3.94, "learning_rate": 1.0719164545619845e-05, "loss": 0.037, "step": 21497 }, { "epoch": 3.94, "learning_rate": 1.0718424036017016e-05, "loss": 0.0334, "step": 21498 }, { "epoch": 3.94, "learning_rate": 1.0717683522454219e-05, "loss": 0.0443, "step": 21499 }, { "epoch": 3.94, "learning_rate": 1.0716943004935532e-05, "loss": 0.0421, "step": 21500 }, { "epoch": 3.94, "learning_rate": 1.0716202483465035e-05, "loss": 0.0335, "step": 21501 }, { "epoch": 3.94, "learning_rate": 1.0715461958046818e-05, "loss": 0.0432, "step": 21502 }, { "epoch": 3.94, "learning_rate": 1.0714721428684955e-05, "loss": 0.0415, "step": 21503 }, { "epoch": 3.94, "learning_rate": 1.071398089538353e-05, "loss": 0.0417, "step": 21504 }, { "epoch": 3.94, "learning_rate": 1.071324035814663e-05, "loss": 0.03, "step": 21505 }, { "epoch": 3.94, "learning_rate": 1.0712499816978328e-05, "loss": 0.0439, "step": 21506 }, { "epoch": 3.94, "learning_rate": 1.0711759271882713e-05, "loss": 0.044, "step": 21507 }, { "epoch": 3.94, "learning_rate": 1.0711018722863864e-05, "loss": 0.0457, "step": 21508 }, { "epoch": 3.94, "learning_rate": 1.0710278169925861e-05, "loss": 0.0424, "step": 21509 }, { "epoch": 3.94, "learning_rate": 1.0709537613072789e-05, "loss": 0.0342, "step": 21510 }, { "epoch": 3.94, "learning_rate": 1.070879705230873e-05, "loss": 0.035, "step": 21511 }, { "epoch": 3.94, "learning_rate": 1.0708056487637765e-05, "loss": 0.0408, "step": 21512 }, { "epoch": 3.95, "learning_rate": 1.0707315919063974e-05, "loss": 0.0341, "step": 21513 }, { "epoch": 3.95, "learning_rate": 1.0706575346591445e-05, "loss": 0.0376, "step": 21514 }, { "epoch": 3.95, "learning_rate": 1.0705834770224253e-05, "loss": 0.0451, "step": 21515 }, { "epoch": 3.95, "learning_rate": 1.0705094189966485e-05, "loss": 0.0391, "step": 21516 }, { "epoch": 3.95, "learning_rate": 1.070435360582222e-05, "loss": 0.0398, "step": 21517 }, { "epoch": 3.95, "learning_rate": 1.0703613017795541e-05, "loss": 0.0348, "step": 21518 }, { "epoch": 3.95, "learning_rate": 1.070287242589053e-05, "loss": 0.0402, "step": 21519 }, { "epoch": 3.95, "learning_rate": 1.0702131830111274e-05, "loss": 0.0388, "step": 21520 }, { "epoch": 3.95, "learning_rate": 1.0701391230461847e-05, "loss": 0.0478, "step": 21521 }, { "epoch": 3.95, "learning_rate": 1.0700650626946334e-05, "loss": 0.0482, "step": 21522 }, { "epoch": 3.95, "learning_rate": 1.0699910019568824e-05, "loss": 0.0374, "step": 21523 }, { "epoch": 3.95, "learning_rate": 1.0699169408333387e-05, "loss": 0.0433, "step": 21524 }, { "epoch": 3.95, "learning_rate": 1.0698428793244116e-05, "loss": 0.0431, "step": 21525 }, { "epoch": 3.95, "learning_rate": 1.0697688174305087e-05, "loss": 0.0382, "step": 21526 }, { "epoch": 3.95, "learning_rate": 1.0696947551520387e-05, "loss": 0.0427, "step": 21527 }, { "epoch": 3.95, "learning_rate": 1.0696206924894092e-05, "loss": 0.0385, "step": 21528 }, { "epoch": 3.95, "learning_rate": 1.0695466294430292e-05, "loss": 0.0422, "step": 21529 }, { "epoch": 3.95, "learning_rate": 1.0694725660133064e-05, "loss": 0.035, "step": 21530 }, { "epoch": 3.95, "learning_rate": 1.069398502200649e-05, "loss": 0.0429, "step": 21531 }, { "epoch": 3.95, "learning_rate": 1.0693244380054657e-05, "loss": 0.044, "step": 21532 }, { "epoch": 3.95, "learning_rate": 1.0692503734281643e-05, "loss": 0.0397, "step": 21533 }, { "epoch": 3.95, "learning_rate": 1.0691763084691533e-05, "loss": 0.0375, "step": 21534 }, { "epoch": 3.95, "learning_rate": 1.069102243128841e-05, "loss": 0.0362, "step": 21535 }, { "epoch": 3.95, "learning_rate": 1.0690281774076353e-05, "loss": 0.0363, "step": 21536 }, { "epoch": 3.95, "learning_rate": 1.0689541113059446e-05, "loss": 0.0384, "step": 21537 }, { "epoch": 3.95, "learning_rate": 1.0688800448241775e-05, "loss": 0.0371, "step": 21538 }, { "epoch": 3.95, "learning_rate": 1.0688059779627417e-05, "loss": 0.0408, "step": 21539 }, { "epoch": 3.95, "learning_rate": 1.068731910722046e-05, "loss": 0.0396, "step": 21540 }, { "epoch": 3.95, "learning_rate": 1.068657843102498e-05, "loss": 0.0383, "step": 21541 }, { "epoch": 3.95, "learning_rate": 1.068583775104507e-05, "loss": 0.0393, "step": 21542 }, { "epoch": 3.95, "learning_rate": 1.0685097067284801e-05, "loss": 0.0481, "step": 21543 }, { "epoch": 3.95, "learning_rate": 1.0684356379748264e-05, "loss": 0.039, "step": 21544 }, { "epoch": 3.95, "learning_rate": 1.0683615688439536e-05, "loss": 0.0445, "step": 21545 }, { "epoch": 3.95, "learning_rate": 1.0682874993362705e-05, "loss": 0.0434, "step": 21546 }, { "epoch": 3.95, "learning_rate": 1.0682134294521848e-05, "loss": 0.0396, "step": 21547 }, { "epoch": 3.95, "learning_rate": 1.0681393591921052e-05, "loss": 0.0373, "step": 21548 }, { "epoch": 3.95, "learning_rate": 1.0680652885564398e-05, "loss": 0.0402, "step": 21549 }, { "epoch": 3.95, "learning_rate": 1.067991217545597e-05, "loss": 0.0435, "step": 21550 }, { "epoch": 3.95, "learning_rate": 1.067917146159985e-05, "loss": 0.0436, "step": 21551 }, { "epoch": 3.95, "learning_rate": 1.0678430744000123e-05, "loss": 0.039, "step": 21552 }, { "epoch": 3.95, "learning_rate": 1.067769002266087e-05, "loss": 0.0413, "step": 21553 }, { "epoch": 3.95, "learning_rate": 1.0676949297586168e-05, "loss": 0.0422, "step": 21554 }, { "epoch": 3.95, "learning_rate": 1.0676208568780113e-05, "loss": 0.0327, "step": 21555 }, { "epoch": 3.95, "learning_rate": 1.0675467836246778e-05, "loss": 0.0491, "step": 21556 }, { "epoch": 3.95, "learning_rate": 1.0674727099990246e-05, "loss": 0.0429, "step": 21557 }, { "epoch": 3.95, "learning_rate": 1.0673986360014603e-05, "loss": 0.0418, "step": 21558 }, { "epoch": 3.95, "learning_rate": 1.0673245616323933e-05, "loss": 0.04, "step": 21559 }, { "epoch": 3.95, "learning_rate": 1.0672504868922316e-05, "loss": 0.0471, "step": 21560 }, { "epoch": 3.95, "learning_rate": 1.0671764117813836e-05, "loss": 0.0321, "step": 21561 }, { "epoch": 3.95, "learning_rate": 1.067102336300258e-05, "loss": 0.0395, "step": 21562 }, { "epoch": 3.95, "learning_rate": 1.0670282604492623e-05, "loss": 0.0413, "step": 21563 }, { "epoch": 3.95, "learning_rate": 1.0669541842288056e-05, "loss": 0.0431, "step": 21564 }, { "epoch": 3.95, "learning_rate": 1.0668801076392956e-05, "loss": 0.0333, "step": 21565 }, { "epoch": 3.95, "learning_rate": 1.066806030681141e-05, "loss": 0.0432, "step": 21566 }, { "epoch": 3.96, "learning_rate": 1.06673195335475e-05, "loss": 0.0356, "step": 21567 }, { "epoch": 3.96, "learning_rate": 1.0666578756605311e-05, "loss": 0.0338, "step": 21568 }, { "epoch": 3.96, "learning_rate": 1.0665837975988923e-05, "loss": 0.0384, "step": 21569 }, { "epoch": 3.96, "learning_rate": 1.0665097191702418e-05, "loss": 0.0387, "step": 21570 }, { "epoch": 3.96, "learning_rate": 1.0664356403749884e-05, "loss": 0.0407, "step": 21571 }, { "epoch": 3.96, "learning_rate": 1.06636156121354e-05, "loss": 0.0364, "step": 21572 }, { "epoch": 3.96, "learning_rate": 1.0662874816863056e-05, "loss": 0.0386, "step": 21573 }, { "epoch": 3.96, "learning_rate": 1.0662134017936924e-05, "loss": 0.0386, "step": 21574 }, { "epoch": 3.96, "learning_rate": 1.0661393215361098e-05, "loss": 0.0396, "step": 21575 }, { "epoch": 3.96, "learning_rate": 1.0660652409139658e-05, "loss": 0.0391, "step": 21576 }, { "epoch": 3.96, "learning_rate": 1.0659911599276682e-05, "loss": 0.0422, "step": 21577 }, { "epoch": 3.96, "learning_rate": 1.0659170785776262e-05, "loss": 0.0378, "step": 21578 }, { "epoch": 3.96, "learning_rate": 1.0658429968642476e-05, "loss": 0.0398, "step": 21579 }, { "epoch": 3.96, "learning_rate": 1.065768914787941e-05, "loss": 0.0311, "step": 21580 }, { "epoch": 3.96, "learning_rate": 1.0656948323491142e-05, "loss": 0.0337, "step": 21581 }, { "epoch": 3.96, "learning_rate": 1.0656207495481762e-05, "loss": 0.0436, "step": 21582 }, { "epoch": 3.96, "learning_rate": 1.0655466663855349e-05, "loss": 0.0481, "step": 21583 }, { "epoch": 3.96, "learning_rate": 1.0654725828615991e-05, "loss": 0.0323, "step": 21584 }, { "epoch": 3.96, "learning_rate": 1.0653984989767766e-05, "loss": 0.042, "step": 21585 }, { "epoch": 3.96, "learning_rate": 1.0653244147314763e-05, "loss": 0.0346, "step": 21586 }, { "epoch": 3.96, "learning_rate": 1.0652503301261064e-05, "loss": 0.043, "step": 21587 }, { "epoch": 3.96, "learning_rate": 1.0651762451610747e-05, "loss": 0.0319, "step": 21588 }, { "epoch": 3.96, "learning_rate": 1.0651021598367905e-05, "loss": 0.0474, "step": 21589 }, { "epoch": 3.96, "learning_rate": 1.0650280741536612e-05, "loss": 0.0433, "step": 21590 }, { "epoch": 3.96, "learning_rate": 1.064953988112096e-05, "loss": 0.0447, "step": 21591 }, { "epoch": 3.96, "learning_rate": 1.0648799017125027e-05, "loss": 0.039, "step": 21592 }, { "epoch": 3.96, "learning_rate": 1.06480581495529e-05, "loss": 0.0362, "step": 21593 }, { "epoch": 3.96, "learning_rate": 1.064731727840866e-05, "loss": 0.0424, "step": 21594 }, { "epoch": 3.96, "learning_rate": 1.0646576403696394e-05, "loss": 0.0403, "step": 21595 }, { "epoch": 3.96, "learning_rate": 1.064583552542018e-05, "loss": 0.0372, "step": 21596 }, { "epoch": 3.96, "learning_rate": 1.0645094643584111e-05, "loss": 0.0389, "step": 21597 }, { "epoch": 3.96, "learning_rate": 1.0644353758192262e-05, "loss": 0.0377, "step": 21598 }, { "epoch": 3.96, "learning_rate": 1.0643612869248718e-05, "loss": 0.0371, "step": 21599 }, { "epoch": 3.96, "learning_rate": 1.064287197675757e-05, "loss": 0.0403, "step": 21600 }, { "epoch": 3.96, "learning_rate": 1.0642131080722891e-05, "loss": 0.0355, "step": 21601 }, { "epoch": 3.96, "learning_rate": 1.0641390181148772e-05, "loss": 0.0404, "step": 21602 }, { "epoch": 3.96, "learning_rate": 1.0640649278039297e-05, "loss": 0.0364, "step": 21603 }, { "epoch": 3.96, "learning_rate": 1.0639908371398548e-05, "loss": 0.0396, "step": 21604 }, { "epoch": 3.96, "learning_rate": 1.0639167461230607e-05, "loss": 0.0363, "step": 21605 }, { "epoch": 3.96, "learning_rate": 1.0638426547539562e-05, "loss": 0.0335, "step": 21606 }, { "epoch": 3.96, "learning_rate": 1.0637685630329494e-05, "loss": 0.0461, "step": 21607 }, { "epoch": 3.96, "learning_rate": 1.0636944709604486e-05, "loss": 0.043, "step": 21608 }, { "epoch": 3.96, "learning_rate": 1.063620378536863e-05, "loss": 0.04, "step": 21609 }, { "epoch": 3.96, "learning_rate": 1.0635462857625998e-05, "loss": 0.0445, "step": 21610 }, { "epoch": 3.96, "learning_rate": 1.0634721926380682e-05, "loss": 0.0386, "step": 21611 }, { "epoch": 3.96, "learning_rate": 1.0633980991636762e-05, "loss": 0.0461, "step": 21612 }, { "epoch": 3.96, "learning_rate": 1.0633240053398329e-05, "loss": 0.0347, "step": 21613 }, { "epoch": 3.96, "learning_rate": 1.0632499111669453e-05, "loss": 0.042, "step": 21614 }, { "epoch": 3.96, "learning_rate": 1.0631758166454236e-05, "loss": 0.0389, "step": 21615 }, { "epoch": 3.96, "learning_rate": 1.063101721775675e-05, "loss": 0.0412, "step": 21616 }, { "epoch": 3.96, "learning_rate": 1.063027626558108e-05, "loss": 0.0446, "step": 21617 }, { "epoch": 3.96, "learning_rate": 1.0629535309931317e-05, "loss": 0.0383, "step": 21618 }, { "epoch": 3.96, "learning_rate": 1.0628794350811537e-05, "loss": 0.0404, "step": 21619 }, { "epoch": 3.96, "learning_rate": 1.0628053388225829e-05, "loss": 0.0391, "step": 21620 }, { "epoch": 3.96, "learning_rate": 1.0627312422178275e-05, "loss": 0.0432, "step": 21621 }, { "epoch": 3.97, "learning_rate": 1.0626571452672962e-05, "loss": 0.0452, "step": 21622 }, { "epoch": 3.97, "learning_rate": 1.0625830479713972e-05, "loss": 0.0402, "step": 21623 }, { "epoch": 3.97, "learning_rate": 1.062508950330539e-05, "loss": 0.0427, "step": 21624 }, { "epoch": 3.97, "learning_rate": 1.0624348523451298e-05, "loss": 0.0455, "step": 21625 }, { "epoch": 3.97, "learning_rate": 1.0623607540155782e-05, "loss": 0.0347, "step": 21626 }, { "epoch": 3.97, "learning_rate": 1.062286655342293e-05, "loss": 0.0474, "step": 21627 }, { "epoch": 3.97, "learning_rate": 1.0622125563256823e-05, "loss": 0.0371, "step": 21628 }, { "epoch": 3.97, "learning_rate": 1.0621384569661542e-05, "loss": 0.036, "step": 21629 }, { "epoch": 3.97, "learning_rate": 1.0620643572641177e-05, "loss": 0.0413, "step": 21630 }, { "epoch": 3.97, "learning_rate": 1.061990257219981e-05, "loss": 0.0429, "step": 21631 }, { "epoch": 3.97, "learning_rate": 1.0619161568341525e-05, "loss": 0.032, "step": 21632 }, { "epoch": 3.97, "learning_rate": 1.0618420561070406e-05, "loss": 0.037, "step": 21633 }, { "epoch": 3.97, "learning_rate": 1.0617679550390538e-05, "loss": 0.0392, "step": 21634 }, { "epoch": 3.97, "learning_rate": 1.0616938536306008e-05, "loss": 0.0399, "step": 21635 }, { "epoch": 3.97, "learning_rate": 1.0616197518820896e-05, "loss": 0.0369, "step": 21636 }, { "epoch": 3.97, "learning_rate": 1.0615456497939292e-05, "loss": 0.035, "step": 21637 }, { "epoch": 3.97, "learning_rate": 1.0614715473665276e-05, "loss": 0.0426, "step": 21638 }, { "epoch": 3.97, "learning_rate": 1.0613974446002933e-05, "loss": 0.0413, "step": 21639 }, { "epoch": 3.97, "learning_rate": 1.0613233414956349e-05, "loss": 0.0373, "step": 21640 }, { "epoch": 3.97, "learning_rate": 1.061249238052961e-05, "loss": 0.044, "step": 21641 }, { "epoch": 3.97, "learning_rate": 1.0611751342726795e-05, "loss": 0.0405, "step": 21642 }, { "epoch": 3.97, "learning_rate": 1.0611010301551992e-05, "loss": 0.0469, "step": 21643 }, { "epoch": 3.97, "learning_rate": 1.0610269257009289e-05, "loss": 0.0442, "step": 21644 }, { "epoch": 3.97, "learning_rate": 1.0609528209102767e-05, "loss": 0.035, "step": 21645 }, { "epoch": 3.97, "learning_rate": 1.060878715783651e-05, "loss": 0.0407, "step": 21646 }, { "epoch": 3.97, "learning_rate": 1.0608046103214605e-05, "loss": 0.0382, "step": 21647 }, { "epoch": 3.97, "learning_rate": 1.0607305045241134e-05, "loss": 0.0418, "step": 21648 }, { "epoch": 3.97, "learning_rate": 1.0606563983920183e-05, "loss": 0.0477, "step": 21649 }, { "epoch": 3.97, "learning_rate": 1.0605822919255841e-05, "loss": 0.0422, "step": 21650 }, { "epoch": 3.97, "learning_rate": 1.0605081851252189e-05, "loss": 0.036, "step": 21651 }, { "epoch": 3.97, "learning_rate": 1.0604340779913305e-05, "loss": 0.0421, "step": 21652 }, { "epoch": 3.97, "learning_rate": 1.0603599705243288e-05, "loss": 0.0516, "step": 21653 }, { "epoch": 3.97, "learning_rate": 1.0602858627246208e-05, "loss": 0.0437, "step": 21654 }, { "epoch": 3.97, "learning_rate": 1.0602117545926163e-05, "loss": 0.0437, "step": 21655 }, { "epoch": 3.97, "learning_rate": 1.0601376461287229e-05, "loss": 0.0359, "step": 21656 }, { "epoch": 3.97, "learning_rate": 1.0600635373333497e-05, "loss": 0.0375, "step": 21657 }, { "epoch": 3.97, "learning_rate": 1.0599894282069043e-05, "loss": 0.0394, "step": 21658 }, { "epoch": 3.97, "learning_rate": 1.0599153187497966e-05, "loss": 0.0476, "step": 21659 }, { "epoch": 3.97, "learning_rate": 1.0598412089624335e-05, "loss": 0.0382, "step": 21660 }, { "epoch": 3.97, "learning_rate": 1.0597670988452245e-05, "loss": 0.0388, "step": 21661 }, { "epoch": 3.97, "learning_rate": 1.0596929883985782e-05, "loss": 0.0416, "step": 21662 }, { "epoch": 3.97, "learning_rate": 1.0596188776229022e-05, "loss": 0.0391, "step": 21663 }, { "epoch": 3.97, "learning_rate": 1.0595447665186058e-05, "loss": 0.0423, "step": 21664 }, { "epoch": 3.97, "learning_rate": 1.0594706550860974e-05, "loss": 0.0374, "step": 21665 }, { "epoch": 3.97, "learning_rate": 1.059396543325785e-05, "loss": 0.0425, "step": 21666 }, { "epoch": 3.97, "learning_rate": 1.0593224312380777e-05, "loss": 0.0417, "step": 21667 }, { "epoch": 3.97, "learning_rate": 1.0592483188233839e-05, "loss": 0.0316, "step": 21668 }, { "epoch": 3.97, "learning_rate": 1.0591742060821115e-05, "loss": 0.0397, "step": 21669 }, { "epoch": 3.97, "learning_rate": 1.0591000930146699e-05, "loss": 0.0349, "step": 21670 }, { "epoch": 3.97, "learning_rate": 1.0590259796214672e-05, "loss": 0.0362, "step": 21671 }, { "epoch": 3.97, "learning_rate": 1.058951865902912e-05, "loss": 0.0417, "step": 21672 }, { "epoch": 3.97, "learning_rate": 1.0588777518594124e-05, "loss": 0.0397, "step": 21673 }, { "epoch": 3.97, "learning_rate": 1.0588036374913775e-05, "loss": 0.0439, "step": 21674 }, { "epoch": 3.97, "learning_rate": 1.0587295227992154e-05, "loss": 0.0384, "step": 21675 }, { "epoch": 3.98, "learning_rate": 1.0586554077833346e-05, "loss": 0.0391, "step": 21676 }, { "epoch": 3.98, "learning_rate": 1.0585812924441446e-05, "loss": 0.0399, "step": 21677 }, { "epoch": 3.98, "learning_rate": 1.0585071767820524e-05, "loss": 0.0394, "step": 21678 }, { "epoch": 3.98, "learning_rate": 1.0584330607974673e-05, "loss": 0.0408, "step": 21679 }, { "epoch": 3.98, "learning_rate": 1.0583589444907982e-05, "loss": 0.0405, "step": 21680 }, { "epoch": 3.98, "learning_rate": 1.0582848278624532e-05, "loss": 0.0341, "step": 21681 }, { "epoch": 3.98, "learning_rate": 1.0582107109128404e-05, "loss": 0.0402, "step": 21682 }, { "epoch": 3.98, "learning_rate": 1.0581365936423693e-05, "loss": 0.0342, "step": 21683 }, { "epoch": 3.98, "learning_rate": 1.0580624760514477e-05, "loss": 0.0414, "step": 21684 }, { "epoch": 3.98, "learning_rate": 1.0579883581404842e-05, "loss": 0.0392, "step": 21685 }, { "epoch": 3.98, "learning_rate": 1.057914239909888e-05, "loss": 0.0399, "step": 21686 }, { "epoch": 3.98, "learning_rate": 1.0578401213600667e-05, "loss": 0.0366, "step": 21687 }, { "epoch": 3.98, "learning_rate": 1.0577660024914294e-05, "loss": 0.0459, "step": 21688 }, { "epoch": 3.98, "learning_rate": 1.0576918833043847e-05, "loss": 0.0433, "step": 21689 }, { "epoch": 3.98, "learning_rate": 1.057617763799341e-05, "loss": 0.0388, "step": 21690 }, { "epoch": 3.98, "learning_rate": 1.0575436439767064e-05, "loss": 0.0364, "step": 21691 }, { "epoch": 3.98, "learning_rate": 1.0574695238368905e-05, "loss": 0.0316, "step": 21692 }, { "epoch": 3.98, "learning_rate": 1.0573954033803006e-05, "loss": 0.0371, "step": 21693 }, { "epoch": 3.98, "learning_rate": 1.0573212826073464e-05, "loss": 0.0446, "step": 21694 }, { "epoch": 3.98, "learning_rate": 1.0572471615184361e-05, "loss": 0.0415, "step": 21695 }, { "epoch": 3.98, "learning_rate": 1.0571730401139775e-05, "loss": 0.048, "step": 21696 }, { "epoch": 3.98, "learning_rate": 1.05709891839438e-05, "loss": 0.0401, "step": 21697 }, { "epoch": 3.98, "learning_rate": 1.0570247963600521e-05, "loss": 0.0375, "step": 21698 }, { "epoch": 3.98, "learning_rate": 1.0569506740114022e-05, "loss": 0.0421, "step": 21699 }, { "epoch": 3.98, "learning_rate": 1.0568765513488388e-05, "loss": 0.0417, "step": 21700 }, { "epoch": 3.98, "learning_rate": 1.0568024283727704e-05, "loss": 0.0516, "step": 21701 }, { "epoch": 3.98, "learning_rate": 1.0567283050836054e-05, "loss": 0.0309, "step": 21702 }, { "epoch": 3.98, "learning_rate": 1.056654181481753e-05, "loss": 0.0381, "step": 21703 }, { "epoch": 3.98, "learning_rate": 1.0565800575676217e-05, "loss": 0.0405, "step": 21704 }, { "epoch": 3.98, "learning_rate": 1.0565059333416194e-05, "loss": 0.0391, "step": 21705 }, { "epoch": 3.98, "learning_rate": 1.0564318088041552e-05, "loss": 0.04, "step": 21706 }, { "epoch": 3.98, "learning_rate": 1.0563576839556375e-05, "loss": 0.0373, "step": 21707 }, { "epoch": 3.98, "learning_rate": 1.056283558796475e-05, "loss": 0.0427, "step": 21708 }, { "epoch": 3.98, "learning_rate": 1.056209433327076e-05, "loss": 0.0386, "step": 21709 }, { "epoch": 3.98, "learning_rate": 1.0561353075478496e-05, "loss": 0.0426, "step": 21710 }, { "epoch": 3.98, "learning_rate": 1.0560611814592036e-05, "loss": 0.0396, "step": 21711 }, { "epoch": 3.98, "learning_rate": 1.0559870550615475e-05, "loss": 0.0481, "step": 21712 }, { "epoch": 3.98, "learning_rate": 1.0559129283552893e-05, "loss": 0.0368, "step": 21713 }, { "epoch": 3.98, "learning_rate": 1.0558388013408378e-05, "loss": 0.0433, "step": 21714 }, { "epoch": 3.98, "learning_rate": 1.0557646740186014e-05, "loss": 0.0389, "step": 21715 }, { "epoch": 3.98, "learning_rate": 1.055690546388989e-05, "loss": 0.0352, "step": 21716 }, { "epoch": 3.98, "learning_rate": 1.0556164184524088e-05, "loss": 0.0358, "step": 21717 }, { "epoch": 3.98, "learning_rate": 1.0555422902092698e-05, "loss": 0.0358, "step": 21718 }, { "epoch": 3.98, "learning_rate": 1.0554681616599803e-05, "loss": 0.0325, "step": 21719 }, { "epoch": 3.98, "learning_rate": 1.0553940328049486e-05, "loss": 0.0347, "step": 21720 }, { "epoch": 3.98, "learning_rate": 1.055319903644584e-05, "loss": 0.0367, "step": 21721 }, { "epoch": 3.98, "learning_rate": 1.0552457741792949e-05, "loss": 0.0424, "step": 21722 }, { "epoch": 3.98, "learning_rate": 1.05517164440949e-05, "loss": 0.0385, "step": 21723 }, { "epoch": 3.98, "learning_rate": 1.0550975143355772e-05, "loss": 0.0467, "step": 21724 }, { "epoch": 3.98, "learning_rate": 1.055023383957966e-05, "loss": 0.037, "step": 21725 }, { "epoch": 3.98, "learning_rate": 1.0549492532770645e-05, "loss": 0.0339, "step": 21726 }, { "epoch": 3.98, "learning_rate": 1.0548751222932815e-05, "loss": 0.0411, "step": 21727 }, { "epoch": 3.98, "learning_rate": 1.0548009910070253e-05, "loss": 0.044, "step": 21728 }, { "epoch": 3.98, "learning_rate": 1.0547268594187049e-05, "loss": 0.0399, "step": 21729 }, { "epoch": 3.98, "learning_rate": 1.0546527275287285e-05, "loss": 0.0419, "step": 21730 }, { "epoch": 3.99, "learning_rate": 1.0545785953375054e-05, "loss": 0.0408, "step": 21731 }, { "epoch": 3.99, "learning_rate": 1.0545044628454437e-05, "loss": 0.0395, "step": 21732 }, { "epoch": 3.99, "learning_rate": 1.054430330052952e-05, "loss": 0.0404, "step": 21733 }, { "epoch": 3.99, "learning_rate": 1.0543561969604392e-05, "loss": 0.0337, "step": 21734 }, { "epoch": 3.99, "learning_rate": 1.0542820635683136e-05, "loss": 0.0421, "step": 21735 }, { "epoch": 3.99, "learning_rate": 1.0542079298769843e-05, "loss": 0.0454, "step": 21736 }, { "epoch": 3.99, "learning_rate": 1.0541337958868594e-05, "loss": 0.0396, "step": 21737 }, { "epoch": 3.99, "learning_rate": 1.0540596615983477e-05, "loss": 0.0337, "step": 21738 }, { "epoch": 3.99, "learning_rate": 1.0539855270118579e-05, "loss": 0.0416, "step": 21739 }, { "epoch": 3.99, "learning_rate": 1.0539113921277986e-05, "loss": 0.0463, "step": 21740 }, { "epoch": 3.99, "learning_rate": 1.0538372569465784e-05, "loss": 0.0312, "step": 21741 }, { "epoch": 3.99, "learning_rate": 1.053763121468606e-05, "loss": 0.0433, "step": 21742 }, { "epoch": 3.99, "learning_rate": 1.0536889856942903e-05, "loss": 0.0452, "step": 21743 }, { "epoch": 3.99, "learning_rate": 1.0536148496240393e-05, "loss": 0.0423, "step": 21744 }, { "epoch": 3.99, "learning_rate": 1.0535407132582623e-05, "loss": 0.0331, "step": 21745 }, { "epoch": 3.99, "learning_rate": 1.0534665765973673e-05, "loss": 0.0335, "step": 21746 }, { "epoch": 3.99, "learning_rate": 1.0533924396417635e-05, "loss": 0.042, "step": 21747 }, { "epoch": 3.99, "learning_rate": 1.0533183023918594e-05, "loss": 0.0404, "step": 21748 }, { "epoch": 3.99, "learning_rate": 1.0532441648480632e-05, "loss": 0.0447, "step": 21749 }, { "epoch": 3.99, "learning_rate": 1.053170027010784e-05, "loss": 0.0442, "step": 21750 }, { "epoch": 3.99, "learning_rate": 1.0530958888804307e-05, "loss": 0.0389, "step": 21751 }, { "epoch": 3.99, "learning_rate": 1.0530217504574114e-05, "loss": 0.0376, "step": 21752 }, { "epoch": 3.99, "learning_rate": 1.0529476117421349e-05, "loss": 0.0391, "step": 21753 }, { "epoch": 3.99, "learning_rate": 1.0528734727350103e-05, "loss": 0.0397, "step": 21754 }, { "epoch": 3.99, "learning_rate": 1.0527993334364452e-05, "loss": 0.0388, "step": 21755 }, { "epoch": 3.99, "learning_rate": 1.0527251938468495e-05, "loss": 0.042, "step": 21756 }, { "epoch": 3.99, "learning_rate": 1.0526510539666311e-05, "loss": 0.0318, "step": 21757 }, { "epoch": 3.99, "learning_rate": 1.0525769137961989e-05, "loss": 0.0374, "step": 21758 }, { "epoch": 3.99, "learning_rate": 1.0525027733359617e-05, "loss": 0.0375, "step": 21759 }, { "epoch": 3.99, "learning_rate": 1.0524286325863277e-05, "loss": 0.042, "step": 21760 }, { "epoch": 3.99, "learning_rate": 1.052354491547706e-05, "loss": 0.0419, "step": 21761 }, { "epoch": 3.99, "learning_rate": 1.0522803502205049e-05, "loss": 0.0439, "step": 21762 }, { "epoch": 3.99, "learning_rate": 1.0522062086051337e-05, "loss": 0.0407, "step": 21763 }, { "epoch": 3.99, "learning_rate": 1.0521320667020001e-05, "loss": 0.0446, "step": 21764 }, { "epoch": 3.99, "learning_rate": 1.0520579245115136e-05, "loss": 0.0384, "step": 21765 }, { "epoch": 3.99, "learning_rate": 1.0519837820340827e-05, "loss": 0.0397, "step": 21766 }, { "epoch": 3.99, "learning_rate": 1.0519096392701157e-05, "loss": 0.0449, "step": 21767 }, { "epoch": 3.99, "learning_rate": 1.0518354962200219e-05, "loss": 0.035, "step": 21768 }, { "epoch": 3.99, "learning_rate": 1.0517613528842096e-05, "loss": 0.045, "step": 21769 }, { "epoch": 3.99, "learning_rate": 1.0516872092630874e-05, "loss": 0.0453, "step": 21770 }, { "epoch": 3.99, "learning_rate": 1.051613065357064e-05, "loss": 0.0371, "step": 21771 }, { "epoch": 3.99, "learning_rate": 1.0515389211665483e-05, "loss": 0.0395, "step": 21772 }, { "epoch": 3.99, "learning_rate": 1.0514647766919488e-05, "loss": 0.0431, "step": 21773 }, { "epoch": 3.99, "learning_rate": 1.0513906319336741e-05, "loss": 0.0456, "step": 21774 }, { "epoch": 3.99, "learning_rate": 1.0513164868921333e-05, "loss": 0.031, "step": 21775 }, { "epoch": 3.99, "learning_rate": 1.0512423415677347e-05, "loss": 0.0394, "step": 21776 }, { "epoch": 3.99, "learning_rate": 1.0511681959608872e-05, "loss": 0.045, "step": 21777 }, { "epoch": 3.99, "learning_rate": 1.0510940500719992e-05, "loss": 0.0506, "step": 21778 }, { "epoch": 3.99, "learning_rate": 1.0510199039014799e-05, "loss": 0.0403, "step": 21779 }, { "epoch": 3.99, "learning_rate": 1.0509457574497373e-05, "loss": 0.044, "step": 21780 }, { "epoch": 3.99, "learning_rate": 1.050871610717181e-05, "loss": 0.0405, "step": 21781 }, { "epoch": 3.99, "learning_rate": 1.0507974637042186e-05, "loss": 0.0415, "step": 21782 }, { "epoch": 3.99, "learning_rate": 1.0507233164112598e-05, "loss": 0.0406, "step": 21783 }, { "epoch": 3.99, "learning_rate": 1.0506491688387128e-05, "loss": 0.0473, "step": 21784 }, { "epoch": 4.0, "learning_rate": 1.0505750209869865e-05, "loss": 0.0407, "step": 21785 }, { "epoch": 4.0, "learning_rate": 1.0505008728564894e-05, "loss": 0.0506, "step": 21786 }, { "epoch": 4.0, "learning_rate": 1.0504267244476302e-05, "loss": 0.0405, "step": 21787 }, { "epoch": 4.0, "learning_rate": 1.0503525757608178e-05, "loss": 0.0458, "step": 21788 }, { "epoch": 4.0, "learning_rate": 1.0502784267964607e-05, "loss": 0.0483, "step": 21789 }, { "epoch": 4.0, "learning_rate": 1.050204277554968e-05, "loss": 0.0394, "step": 21790 }, { "epoch": 4.0, "learning_rate": 1.050130128036748e-05, "loss": 0.0475, "step": 21791 }, { "epoch": 4.0, "learning_rate": 1.0500559782422096e-05, "loss": 0.0438, "step": 21792 }, { "epoch": 4.0, "learning_rate": 1.0499818281717613e-05, "loss": 0.0439, "step": 21793 }, { "epoch": 4.0, "learning_rate": 1.049907677825812e-05, "loss": 0.0412, "step": 21794 }, { "epoch": 4.0, "learning_rate": 1.0498335272047709e-05, "loss": 0.0459, "step": 21795 }, { "epoch": 4.0, "learning_rate": 1.049759376309046e-05, "loss": 0.043, "step": 21796 }, { "epoch": 4.0, "learning_rate": 1.0496852251390458e-05, "loss": 0.0439, "step": 21797 }, { "epoch": 4.0, "learning_rate": 1.0496110736951798e-05, "loss": 0.0403, "step": 21798 }, { "epoch": 4.0, "learning_rate": 1.0495369219778564e-05, "loss": 0.0376, "step": 21799 }, { "epoch": 4.0, "learning_rate": 1.0494627699874844e-05, "loss": 0.0362, "step": 21800 }, { "epoch": 4.0, "learning_rate": 1.0493886177244724e-05, "loss": 0.0452, "step": 21801 }, { "epoch": 4.0, "learning_rate": 1.049314465189229e-05, "loss": 0.0459, "step": 21802 }, { "epoch": 4.0, "learning_rate": 1.0492403123821636e-05, "loss": 0.0441, "step": 21803 }, { "epoch": 4.0, "learning_rate": 1.049166159303684e-05, "loss": 0.0452, "step": 21804 }, { "epoch": 4.0, "learning_rate": 1.0490920059541996e-05, "loss": 0.0371, "step": 21805 }, { "epoch": 4.0, "learning_rate": 1.0490178523341185e-05, "loss": 0.0376, "step": 21806 }, { "epoch": 4.0, "learning_rate": 1.0489436984438502e-05, "loss": 0.0432, "step": 21807 }, { "epoch": 4.0, "learning_rate": 1.048869544283803e-05, "loss": 0.0372, "step": 21808 }, { "epoch": 4.0, "learning_rate": 1.0487953898543858e-05, "loss": 0.0446, "step": 21809 }, { "epoch": 4.0, "learning_rate": 1.0487212351560074e-05, "loss": 0.0413, "step": 21810 }, { "epoch": 4.0, "learning_rate": 1.0486470801890762e-05, "loss": 0.043, "step": 21811 }, { "epoch": 4.0, "learning_rate": 1.0485729249540014e-05, "loss": 0.0395, "step": 21812 }, { "epoch": 4.0, "learning_rate": 1.0484987694511912e-05, "loss": 0.0284, "step": 21813 }, { "epoch": 4.0, "learning_rate": 1.0484246136810547e-05, "loss": 0.0315, "step": 21814 }, { "epoch": 4.0, "learning_rate": 1.0483504576440005e-05, "loss": 0.0266, "step": 21815 }, { "epoch": 4.0, "learning_rate": 1.0482763013404377e-05, "loss": 0.0276, "step": 21816 }, { "epoch": 4.0, "learning_rate": 1.0482021447707746e-05, "loss": 0.0293, "step": 21817 }, { "epoch": 4.0, "learning_rate": 1.0481279879354203e-05, "loss": 0.032, "step": 21818 }, { "epoch": 4.0, "learning_rate": 1.0480538308347834e-05, "loss": 0.0317, "step": 21819 }, { "epoch": 4.0, "learning_rate": 1.0479796734692726e-05, "loss": 0.0313, "step": 21820 }, { "epoch": 4.0, "learning_rate": 1.0479055158392965e-05, "loss": 0.0305, "step": 21821 }, { "epoch": 4.0, "learning_rate": 1.0478313579452645e-05, "loss": 0.0245, "step": 21822 }, { "epoch": 4.0, "learning_rate": 1.0477571997875848e-05, "loss": 0.03, "step": 21823 }, { "epoch": 4.0, "learning_rate": 1.0476830413666661e-05, "loss": 0.0255, "step": 21824 }, { "epoch": 4.0, "learning_rate": 1.0476088826829175e-05, "loss": 0.0291, "step": 21825 }, { "epoch": 4.0, "learning_rate": 1.0475347237367475e-05, "loss": 0.0302, "step": 21826 }, { "epoch": 4.0, "learning_rate": 1.0474605645285652e-05, "loss": 0.0315, "step": 21827 }, { "epoch": 4.0, "learning_rate": 1.047386405058779e-05, "loss": 0.0285, "step": 21828 }, { "epoch": 4.0, "learning_rate": 1.0473122453277979e-05, "loss": 0.0244, "step": 21829 }, { "epoch": 4.0, "learning_rate": 1.0472380853360304e-05, "loss": 0.0296, "step": 21830 }, { "epoch": 4.0, "learning_rate": 1.0471639250838858e-05, "loss": 0.0307, "step": 21831 }, { "epoch": 4.0, "learning_rate": 1.0470897645717724e-05, "loss": 0.0229, "step": 21832 }, { "epoch": 4.0, "learning_rate": 1.047015603800099e-05, "loss": 0.0337, "step": 21833 }, { "epoch": 4.0, "learning_rate": 1.0469414427692748e-05, "loss": 0.0226, "step": 21834 }, { "epoch": 4.0, "learning_rate": 1.0468672814797078e-05, "loss": 0.0265, "step": 21835 }, { "epoch": 4.0, "learning_rate": 1.0467931199318076e-05, "loss": 0.0286, "step": 21836 }, { "epoch": 4.0, "learning_rate": 1.0467189581259825e-05, "loss": 0.0233, "step": 21837 }, { "epoch": 4.0, "learning_rate": 1.0466447960626414e-05, "loss": 0.0311, "step": 21838 }, { "epoch": 4.0, "learning_rate": 1.0465706337421931e-05, "loss": 0.0277, "step": 21839 }, { "epoch": 4.01, "learning_rate": 1.0464964711650466e-05, "loss": 0.0252, "step": 21840 }, { "epoch": 4.01, "learning_rate": 1.04642230833161e-05, "loss": 0.0293, "step": 21841 }, { "epoch": 4.01, "learning_rate": 1.0463481452422929e-05, "loss": 0.0308, "step": 21842 }, { "epoch": 4.01, "learning_rate": 1.0462739818975039e-05, "loss": 0.0312, "step": 21843 }, { "epoch": 4.01, "learning_rate": 1.0461998182976512e-05, "loss": 0.028, "step": 21844 }, { "epoch": 4.01, "learning_rate": 1.0461256544431442e-05, "loss": 0.026, "step": 21845 }, { "epoch": 4.01, "learning_rate": 1.0460514903343915e-05, "loss": 0.0263, "step": 21846 }, { "epoch": 4.01, "learning_rate": 1.0459773259718019e-05, "loss": 0.0246, "step": 21847 }, { "epoch": 4.01, "learning_rate": 1.045903161355784e-05, "loss": 0.0289, "step": 21848 }, { "epoch": 4.01, "learning_rate": 1.0458289964867473e-05, "loss": 0.0261, "step": 21849 }, { "epoch": 4.01, "learning_rate": 1.0457548313650995e-05, "loss": 0.0279, "step": 21850 }, { "epoch": 4.01, "learning_rate": 1.0456806659912504e-05, "loss": 0.0288, "step": 21851 }, { "epoch": 4.01, "learning_rate": 1.0456065003656085e-05, "loss": 0.0265, "step": 21852 }, { "epoch": 4.01, "learning_rate": 1.0455323344885822e-05, "loss": 0.029, "step": 21853 }, { "epoch": 4.01, "learning_rate": 1.0454581683605807e-05, "loss": 0.0275, "step": 21854 }, { "epoch": 4.01, "learning_rate": 1.0453840019820128e-05, "loss": 0.0239, "step": 21855 }, { "epoch": 4.01, "learning_rate": 1.0453098353532871e-05, "loss": 0.0272, "step": 21856 }, { "epoch": 4.01, "learning_rate": 1.0452356684748124e-05, "loss": 0.0231, "step": 21857 }, { "epoch": 4.01, "learning_rate": 1.045161501346998e-05, "loss": 0.0256, "step": 21858 }, { "epoch": 4.01, "learning_rate": 1.0450873339702519e-05, "loss": 0.0304, "step": 21859 }, { "epoch": 4.01, "learning_rate": 1.0450131663449837e-05, "loss": 0.0262, "step": 21860 }, { "epoch": 4.01, "learning_rate": 1.0449389984716018e-05, "loss": 0.0266, "step": 21861 }, { "epoch": 4.01, "learning_rate": 1.044864830350515e-05, "loss": 0.0278, "step": 21862 }, { "epoch": 4.01, "learning_rate": 1.0447906619821325e-05, "loss": 0.0203, "step": 21863 }, { "epoch": 4.01, "learning_rate": 1.0447164933668626e-05, "loss": 0.0218, "step": 21864 }, { "epoch": 4.01, "learning_rate": 1.0446423245051143e-05, "loss": 0.0276, "step": 21865 }, { "epoch": 4.01, "learning_rate": 1.0445681553972961e-05, "loss": 0.0269, "step": 21866 }, { "epoch": 4.01, "learning_rate": 1.044493986043818e-05, "loss": 0.0268, "step": 21867 }, { "epoch": 4.01, "learning_rate": 1.0444198164450874e-05, "loss": 0.0248, "step": 21868 }, { "epoch": 4.01, "learning_rate": 1.0443456466015138e-05, "loss": 0.0248, "step": 21869 }, { "epoch": 4.01, "learning_rate": 1.0442714765135062e-05, "loss": 0.0266, "step": 21870 }, { "epoch": 4.01, "learning_rate": 1.0441973061814732e-05, "loss": 0.0271, "step": 21871 }, { "epoch": 4.01, "learning_rate": 1.0441231356058235e-05, "loss": 0.0228, "step": 21872 }, { "epoch": 4.01, "learning_rate": 1.0440489647869658e-05, "loss": 0.0246, "step": 21873 }, { "epoch": 4.01, "learning_rate": 1.0439747937253093e-05, "loss": 0.0291, "step": 21874 }, { "epoch": 4.01, "learning_rate": 1.0439006224212629e-05, "loss": 0.0247, "step": 21875 }, { "epoch": 4.01, "learning_rate": 1.0438264508752353e-05, "loss": 0.0289, "step": 21876 }, { "epoch": 4.01, "learning_rate": 1.0437522790876348e-05, "loss": 0.029, "step": 21877 }, { "epoch": 4.01, "learning_rate": 1.0436781070588712e-05, "loss": 0.0294, "step": 21878 }, { "epoch": 4.01, "learning_rate": 1.0436039347893528e-05, "loss": 0.0251, "step": 21879 }, { "epoch": 4.01, "learning_rate": 1.0435297622794883e-05, "loss": 0.0248, "step": 21880 }, { "epoch": 4.01, "learning_rate": 1.0434555895296868e-05, "loss": 0.0294, "step": 21881 }, { "epoch": 4.01, "learning_rate": 1.0433814165403571e-05, "loss": 0.0261, "step": 21882 }, { "epoch": 4.01, "learning_rate": 1.0433072433119078e-05, "loss": 0.0255, "step": 21883 }, { "epoch": 4.01, "learning_rate": 1.0432330698447482e-05, "loss": 0.0271, "step": 21884 }, { "epoch": 4.01, "learning_rate": 1.0431588961392869e-05, "loss": 0.0299, "step": 21885 }, { "epoch": 4.01, "learning_rate": 1.0430847221959326e-05, "loss": 0.0234, "step": 21886 }, { "epoch": 4.01, "learning_rate": 1.0430105480150945e-05, "loss": 0.0298, "step": 21887 }, { "epoch": 4.01, "learning_rate": 1.0429363735971811e-05, "loss": 0.0321, "step": 21888 }, { "epoch": 4.01, "learning_rate": 1.0428621989426016e-05, "loss": 0.0279, "step": 21889 }, { "epoch": 4.01, "learning_rate": 1.0427880240517646e-05, "loss": 0.0288, "step": 21890 }, { "epoch": 4.01, "learning_rate": 1.0427138489250787e-05, "loss": 0.0325, "step": 21891 }, { "epoch": 4.01, "learning_rate": 1.0426396735629532e-05, "loss": 0.0192, "step": 21892 }, { "epoch": 4.01, "learning_rate": 1.0425654979657971e-05, "loss": 0.0286, "step": 21893 }, { "epoch": 4.02, "learning_rate": 1.0424913221340187e-05, "loss": 0.0228, "step": 21894 }, { "epoch": 4.02, "learning_rate": 1.0424171460680273e-05, "loss": 0.0289, "step": 21895 }, { "epoch": 4.02, "learning_rate": 1.0423429697682314e-05, "loss": 0.0257, "step": 21896 }, { "epoch": 4.02, "learning_rate": 1.0422687932350402e-05, "loss": 0.0253, "step": 21897 }, { "epoch": 4.02, "learning_rate": 1.0421946164688626e-05, "loss": 0.0261, "step": 21898 }, { "epoch": 4.02, "learning_rate": 1.042120439470107e-05, "loss": 0.0219, "step": 21899 }, { "epoch": 4.02, "learning_rate": 1.0420462622391828e-05, "loss": 0.0277, "step": 21900 }, { "epoch": 4.02, "learning_rate": 1.041972084776498e-05, "loss": 0.0327, "step": 21901 }, { "epoch": 4.02, "learning_rate": 1.0418979070824628e-05, "loss": 0.0313, "step": 21902 }, { "epoch": 4.02, "learning_rate": 1.041823729157485e-05, "loss": 0.0227, "step": 21903 }, { "epoch": 4.02, "learning_rate": 1.041749551001974e-05, "loss": 0.0248, "step": 21904 }, { "epoch": 4.02, "learning_rate": 1.0416753726163385e-05, "loss": 0.0316, "step": 21905 }, { "epoch": 4.02, "learning_rate": 1.0416011940009872e-05, "loss": 0.0304, "step": 21906 }, { "epoch": 4.02, "learning_rate": 1.0415270151563293e-05, "loss": 0.022, "step": 21907 }, { "epoch": 4.02, "learning_rate": 1.0414528360827733e-05, "loss": 0.0354, "step": 21908 }, { "epoch": 4.02, "learning_rate": 1.0413786567807286e-05, "loss": 0.0322, "step": 21909 }, { "epoch": 4.02, "learning_rate": 1.0413044772506034e-05, "loss": 0.0243, "step": 21910 }, { "epoch": 4.02, "learning_rate": 1.0412302974928074e-05, "loss": 0.029, "step": 21911 }, { "epoch": 4.02, "learning_rate": 1.0411561175077487e-05, "loss": 0.0302, "step": 21912 }, { "epoch": 4.02, "learning_rate": 1.0410819372958365e-05, "loss": 0.0302, "step": 21913 }, { "epoch": 4.02, "learning_rate": 1.0410077568574799e-05, "loss": 0.0242, "step": 21914 }, { "epoch": 4.02, "learning_rate": 1.0409335761930877e-05, "loss": 0.0303, "step": 21915 }, { "epoch": 4.02, "learning_rate": 1.040859395303068e-05, "loss": 0.0296, "step": 21916 }, { "epoch": 4.02, "learning_rate": 1.0407852141878311e-05, "loss": 0.0254, "step": 21917 }, { "epoch": 4.02, "learning_rate": 1.040711032847785e-05, "loss": 0.0309, "step": 21918 }, { "epoch": 4.02, "learning_rate": 1.0406368512833384e-05, "loss": 0.0314, "step": 21919 }, { "epoch": 4.02, "learning_rate": 1.0405626694949009e-05, "loss": 0.0303, "step": 21920 }, { "epoch": 4.02, "learning_rate": 1.0404884874828807e-05, "loss": 0.0266, "step": 21921 }, { "epoch": 4.02, "learning_rate": 1.0404143052476871e-05, "loss": 0.0281, "step": 21922 }, { "epoch": 4.02, "learning_rate": 1.0403401227897291e-05, "loss": 0.0304, "step": 21923 }, { "epoch": 4.02, "learning_rate": 1.0402659401094154e-05, "loss": 0.0277, "step": 21924 }, { "epoch": 4.02, "learning_rate": 1.0401917572071546e-05, "loss": 0.0301, "step": 21925 }, { "epoch": 4.02, "learning_rate": 1.0401175740833562e-05, "loss": 0.0257, "step": 21926 }, { "epoch": 4.02, "learning_rate": 1.0400433907384283e-05, "loss": 0.0259, "step": 21927 }, { "epoch": 4.02, "learning_rate": 1.0399692071727807e-05, "loss": 0.0245, "step": 21928 }, { "epoch": 4.02, "learning_rate": 1.039895023386822e-05, "loss": 0.0309, "step": 21929 }, { "epoch": 4.02, "learning_rate": 1.0398208393809609e-05, "loss": 0.0297, "step": 21930 }, { "epoch": 4.02, "learning_rate": 1.0397466551556061e-05, "loss": 0.0222, "step": 21931 }, { "epoch": 4.02, "learning_rate": 1.0396724707111672e-05, "loss": 0.0315, "step": 21932 }, { "epoch": 4.02, "learning_rate": 1.0395982860480524e-05, "loss": 0.029, "step": 21933 }, { "epoch": 4.02, "learning_rate": 1.039524101166671e-05, "loss": 0.0221, "step": 21934 }, { "epoch": 4.02, "learning_rate": 1.0394499160674321e-05, "loss": 0.0253, "step": 21935 }, { "epoch": 4.02, "learning_rate": 1.039375730750744e-05, "loss": 0.0286, "step": 21936 }, { "epoch": 4.02, "learning_rate": 1.0393015452170161e-05, "loss": 0.032, "step": 21937 }, { "epoch": 4.02, "learning_rate": 1.039227359466657e-05, "loss": 0.031, "step": 21938 }, { "epoch": 4.02, "learning_rate": 1.039153173500076e-05, "loss": 0.0217, "step": 21939 }, { "epoch": 4.02, "learning_rate": 1.0390789873176818e-05, "loss": 0.0344, "step": 21940 }, { "epoch": 4.02, "learning_rate": 1.0390048009198832e-05, "loss": 0.0288, "step": 21941 }, { "epoch": 4.02, "learning_rate": 1.0389306143070892e-05, "loss": 0.0317, "step": 21942 }, { "epoch": 4.02, "learning_rate": 1.0388564274797086e-05, "loss": 0.0247, "step": 21943 }, { "epoch": 4.02, "learning_rate": 1.0387822404381509e-05, "loss": 0.0249, "step": 21944 }, { "epoch": 4.02, "learning_rate": 1.0387080531828241e-05, "loss": 0.0288, "step": 21945 }, { "epoch": 4.02, "learning_rate": 1.0386338657141379e-05, "loss": 0.0236, "step": 21946 }, { "epoch": 4.02, "learning_rate": 1.038559678032501e-05, "loss": 0.0216, "step": 21947 }, { "epoch": 4.02, "learning_rate": 1.0384854901383218e-05, "loss": 0.026, "step": 21948 }, { "epoch": 4.03, "learning_rate": 1.0384113020320102e-05, "loss": 0.0285, "step": 21949 }, { "epoch": 4.03, "learning_rate": 1.0383371137139744e-05, "loss": 0.0223, "step": 21950 }, { "epoch": 4.03, "learning_rate": 1.0382629251846235e-05, "loss": 0.0293, "step": 21951 }, { "epoch": 4.03, "learning_rate": 1.0381887364443663e-05, "loss": 0.0269, "step": 21952 }, { "epoch": 4.03, "learning_rate": 1.0381145474936123e-05, "loss": 0.0344, "step": 21953 }, { "epoch": 4.03, "learning_rate": 1.0380403583327696e-05, "loss": 0.0295, "step": 21954 }, { "epoch": 4.03, "learning_rate": 1.0379661689622477e-05, "loss": 0.0213, "step": 21955 }, { "epoch": 4.03, "learning_rate": 1.0378919793824555e-05, "loss": 0.0229, "step": 21956 }, { "epoch": 4.03, "learning_rate": 1.0378177895938018e-05, "loss": 0.0284, "step": 21957 }, { "epoch": 4.03, "learning_rate": 1.0377435995966956e-05, "loss": 0.0264, "step": 21958 }, { "epoch": 4.03, "learning_rate": 1.0376694093915456e-05, "loss": 0.0272, "step": 21959 }, { "epoch": 4.03, "learning_rate": 1.037595218978761e-05, "loss": 0.0229, "step": 21960 }, { "epoch": 4.03, "learning_rate": 1.0375210283587504e-05, "loss": 0.0254, "step": 21961 }, { "epoch": 4.03, "learning_rate": 1.0374468375319238e-05, "loss": 0.03, "step": 21962 }, { "epoch": 4.03, "learning_rate": 1.0373726464986886e-05, "loss": 0.0286, "step": 21963 }, { "epoch": 4.03, "learning_rate": 1.0372984552594547e-05, "loss": 0.0239, "step": 21964 }, { "epoch": 4.03, "learning_rate": 1.037224263814631e-05, "loss": 0.0292, "step": 21965 }, { "epoch": 4.03, "learning_rate": 1.037150072164626e-05, "loss": 0.0232, "step": 21966 }, { "epoch": 4.03, "learning_rate": 1.0370758803098493e-05, "loss": 0.0235, "step": 21967 }, { "epoch": 4.03, "learning_rate": 1.0370016882507095e-05, "loss": 0.0213, "step": 21968 }, { "epoch": 4.03, "learning_rate": 1.036927495987615e-05, "loss": 0.0256, "step": 21969 }, { "epoch": 4.03, "learning_rate": 1.0368533035209758e-05, "loss": 0.0282, "step": 21970 }, { "epoch": 4.03, "learning_rate": 1.0367791108512001e-05, "loss": 0.0299, "step": 21971 }, { "epoch": 4.03, "learning_rate": 1.036704917978697e-05, "loss": 0.0251, "step": 21972 }, { "epoch": 4.03, "learning_rate": 1.0366307249038759e-05, "loss": 0.0302, "step": 21973 }, { "epoch": 4.03, "learning_rate": 1.036556531627145e-05, "loss": 0.0278, "step": 21974 }, { "epoch": 4.03, "learning_rate": 1.0364823381489137e-05, "loss": 0.0277, "step": 21975 }, { "epoch": 4.03, "learning_rate": 1.036408144469591e-05, "loss": 0.0222, "step": 21976 }, { "epoch": 4.03, "learning_rate": 1.0363339505895857e-05, "loss": 0.0287, "step": 21977 }, { "epoch": 4.03, "learning_rate": 1.0362597565093069e-05, "loss": 0.0276, "step": 21978 }, { "epoch": 4.03, "learning_rate": 1.0361855622291637e-05, "loss": 0.0316, "step": 21979 }, { "epoch": 4.03, "learning_rate": 1.0361113677495642e-05, "loss": 0.0263, "step": 21980 }, { "epoch": 4.03, "learning_rate": 1.0360371730709185e-05, "loss": 0.024, "step": 21981 }, { "epoch": 4.03, "learning_rate": 1.0359629781936349e-05, "loss": 0.0316, "step": 21982 }, { "epoch": 4.03, "learning_rate": 1.0358887831181223e-05, "loss": 0.0285, "step": 21983 }, { "epoch": 4.03, "learning_rate": 1.0358145878447903e-05, "loss": 0.0283, "step": 21984 }, { "epoch": 4.03, "learning_rate": 1.0357403923740473e-05, "loss": 0.0273, "step": 21985 }, { "epoch": 4.03, "learning_rate": 1.0356661967063024e-05, "loss": 0.0284, "step": 21986 }, { "epoch": 4.03, "learning_rate": 1.0355920008419645e-05, "loss": 0.0286, "step": 21987 }, { "epoch": 4.03, "learning_rate": 1.0355178047814427e-05, "loss": 0.0266, "step": 21988 }, { "epoch": 4.03, "learning_rate": 1.0354436085251458e-05, "loss": 0.0248, "step": 21989 }, { "epoch": 4.03, "learning_rate": 1.0353694120734832e-05, "loss": 0.0224, "step": 21990 }, { "epoch": 4.03, "learning_rate": 1.0352952154268634e-05, "loss": 0.0287, "step": 21991 }, { "epoch": 4.03, "learning_rate": 1.0352210185856957e-05, "loss": 0.0197, "step": 21992 }, { "epoch": 4.03, "learning_rate": 1.0351468215503888e-05, "loss": 0.0269, "step": 21993 }, { "epoch": 4.03, "learning_rate": 1.0350726243213519e-05, "loss": 0.0328, "step": 21994 }, { "epoch": 4.03, "learning_rate": 1.0349984268989936e-05, "loss": 0.0292, "step": 21995 }, { "epoch": 4.03, "learning_rate": 1.0349242292837232e-05, "loss": 0.0282, "step": 21996 }, { "epoch": 4.03, "learning_rate": 1.0348500314759499e-05, "loss": 0.0295, "step": 21997 }, { "epoch": 4.03, "learning_rate": 1.034775833476082e-05, "loss": 0.0269, "step": 21998 }, { "epoch": 4.03, "learning_rate": 1.034701635284529e-05, "loss": 0.0258, "step": 21999 }, { "epoch": 4.03, "learning_rate": 1.0346274369017002e-05, "loss": 0.0313, "step": 22000 }, { "epoch": 4.03, "learning_rate": 1.0345532383280039e-05, "loss": 0.0229, "step": 22001 }, { "epoch": 4.03, "learning_rate": 1.034479039563849e-05, "loss": 0.0254, "step": 22002 }, { "epoch": 4.04, "learning_rate": 1.0344048406096454e-05, "loss": 0.0284, "step": 22003 }, { "epoch": 4.04, "learning_rate": 1.0343306414658014e-05, "loss": 0.0235, "step": 22004 }, { "epoch": 4.04, "learning_rate": 1.0342564421327257e-05, "loss": 0.0237, "step": 22005 }, { "epoch": 4.04, "learning_rate": 1.0341822426108281e-05, "loss": 0.0256, "step": 22006 }, { "epoch": 4.04, "learning_rate": 1.0341080429005168e-05, "loss": 0.0214, "step": 22007 }, { "epoch": 4.04, "learning_rate": 1.0340338430022014e-05, "loss": 0.0285, "step": 22008 }, { "epoch": 4.04, "learning_rate": 1.0339596429162908e-05, "loss": 0.0275, "step": 22009 }, { "epoch": 4.04, "learning_rate": 1.0338854426431937e-05, "loss": 0.0298, "step": 22010 }, { "epoch": 4.04, "learning_rate": 1.033811242183319e-05, "loss": 0.0307, "step": 22011 }, { "epoch": 4.04, "learning_rate": 1.0337370415370764e-05, "loss": 0.0292, "step": 22012 }, { "epoch": 4.04, "learning_rate": 1.033662840704874e-05, "loss": 0.0263, "step": 22013 }, { "epoch": 4.04, "learning_rate": 1.0335886396871215e-05, "loss": 0.0276, "step": 22014 }, { "epoch": 4.04, "learning_rate": 1.033514438484228e-05, "loss": 0.0302, "step": 22015 }, { "epoch": 4.04, "learning_rate": 1.0334402370966015e-05, "loss": 0.0254, "step": 22016 }, { "epoch": 4.04, "learning_rate": 1.0333660355246517e-05, "loss": 0.0329, "step": 22017 }, { "epoch": 4.04, "learning_rate": 1.0332918337687879e-05, "loss": 0.025, "step": 22018 }, { "epoch": 4.04, "learning_rate": 1.0332176318294185e-05, "loss": 0.021, "step": 22019 }, { "epoch": 4.04, "learning_rate": 1.0331434297069525e-05, "loss": 0.0259, "step": 22020 }, { "epoch": 4.04, "learning_rate": 1.0330692274017997e-05, "loss": 0.03, "step": 22021 }, { "epoch": 4.04, "learning_rate": 1.032995024914368e-05, "loss": 0.0314, "step": 22022 }, { "epoch": 4.04, "learning_rate": 1.0329208222450674e-05, "loss": 0.0253, "step": 22023 }, { "epoch": 4.04, "learning_rate": 1.0328466193943064e-05, "loss": 0.0253, "step": 22024 }, { "epoch": 4.04, "learning_rate": 1.0327724163624934e-05, "loss": 0.0269, "step": 22025 }, { "epoch": 4.04, "learning_rate": 1.0326982131500386e-05, "loss": 0.0272, "step": 22026 }, { "epoch": 4.04, "learning_rate": 1.0326240097573505e-05, "loss": 0.0227, "step": 22027 }, { "epoch": 4.04, "learning_rate": 1.0325498061848383e-05, "loss": 0.0255, "step": 22028 }, { "epoch": 4.04, "learning_rate": 1.0324756024329103e-05, "loss": 0.0229, "step": 22029 }, { "epoch": 4.04, "learning_rate": 1.0324013985019765e-05, "loss": 0.0212, "step": 22030 }, { "epoch": 4.04, "learning_rate": 1.032327194392445e-05, "loss": 0.0269, "step": 22031 }, { "epoch": 4.04, "learning_rate": 1.0322529901047255e-05, "loss": 0.0231, "step": 22032 }, { "epoch": 4.04, "learning_rate": 1.0321787856392266e-05, "loss": 0.0277, "step": 22033 }, { "epoch": 4.04, "learning_rate": 1.0321045809963577e-05, "loss": 0.0276, "step": 22034 }, { "epoch": 4.04, "learning_rate": 1.0320303761765276e-05, "loss": 0.0343, "step": 22035 }, { "epoch": 4.04, "learning_rate": 1.031956171180145e-05, "loss": 0.0355, "step": 22036 }, { "epoch": 4.04, "learning_rate": 1.0318819660076196e-05, "loss": 0.0283, "step": 22037 }, { "epoch": 4.04, "learning_rate": 1.0318077606593598e-05, "loss": 0.0271, "step": 22038 }, { "epoch": 4.04, "learning_rate": 1.0317335551357753e-05, "loss": 0.0259, "step": 22039 }, { "epoch": 4.04, "learning_rate": 1.0316593494372743e-05, "loss": 0.0322, "step": 22040 }, { "epoch": 4.04, "learning_rate": 1.0315851435642662e-05, "loss": 0.0298, "step": 22041 }, { "epoch": 4.04, "learning_rate": 1.03151093751716e-05, "loss": 0.0285, "step": 22042 }, { "epoch": 4.04, "learning_rate": 1.0314367312963651e-05, "loss": 0.031, "step": 22043 }, { "epoch": 4.04, "learning_rate": 1.03136252490229e-05, "loss": 0.0292, "step": 22044 }, { "epoch": 4.04, "learning_rate": 1.0312883183353439e-05, "loss": 0.0251, "step": 22045 }, { "epoch": 4.04, "learning_rate": 1.031214111595936e-05, "loss": 0.0221, "step": 22046 }, { "epoch": 4.04, "learning_rate": 1.0311399046844752e-05, "loss": 0.0316, "step": 22047 }, { "epoch": 4.04, "learning_rate": 1.0310656976013704e-05, "loss": 0.0259, "step": 22048 }, { "epoch": 4.04, "learning_rate": 1.0309914903470306e-05, "loss": 0.0251, "step": 22049 }, { "epoch": 4.04, "learning_rate": 1.0309172829218651e-05, "loss": 0.0259, "step": 22050 }, { "epoch": 4.04, "learning_rate": 1.0308430753262829e-05, "loss": 0.0283, "step": 22051 }, { "epoch": 4.04, "learning_rate": 1.0307688675606928e-05, "loss": 0.0244, "step": 22052 }, { "epoch": 4.04, "learning_rate": 1.0306946596255044e-05, "loss": 0.0216, "step": 22053 }, { "epoch": 4.04, "learning_rate": 1.0306204515211259e-05, "loss": 0.0269, "step": 22054 }, { "epoch": 4.04, "learning_rate": 1.0305462432479665e-05, "loss": 0.0264, "step": 22055 }, { "epoch": 4.04, "learning_rate": 1.030472034806436e-05, "loss": 0.0322, "step": 22056 }, { "epoch": 4.04, "learning_rate": 1.030397826196943e-05, "loss": 0.0256, "step": 22057 }, { "epoch": 4.05, "learning_rate": 1.0303236174198961e-05, "loss": 0.0312, "step": 22058 }, { "epoch": 4.05, "learning_rate": 1.030249408475705e-05, "loss": 0.0289, "step": 22059 }, { "epoch": 4.05, "learning_rate": 1.0301751993647782e-05, "loss": 0.0279, "step": 22060 }, { "epoch": 4.05, "learning_rate": 1.030100990087525e-05, "loss": 0.025, "step": 22061 }, { "epoch": 4.05, "learning_rate": 1.0300267806443547e-05, "loss": 0.0256, "step": 22062 }, { "epoch": 4.05, "learning_rate": 1.029952571035676e-05, "loss": 0.0289, "step": 22063 }, { "epoch": 4.05, "learning_rate": 1.0298783612618977e-05, "loss": 0.0201, "step": 22064 }, { "epoch": 4.05, "learning_rate": 1.0298041513234296e-05, "loss": 0.0291, "step": 22065 }, { "epoch": 4.05, "learning_rate": 1.0297299412206798e-05, "loss": 0.0227, "step": 22066 }, { "epoch": 4.05, "learning_rate": 1.0296557309540582e-05, "loss": 0.0286, "step": 22067 }, { "epoch": 4.05, "learning_rate": 1.0295815205239737e-05, "loss": 0.0204, "step": 22068 }, { "epoch": 4.05, "learning_rate": 1.029507309930835e-05, "loss": 0.0328, "step": 22069 }, { "epoch": 4.05, "learning_rate": 1.0294330991750515e-05, "loss": 0.0237, "step": 22070 }, { "epoch": 4.05, "learning_rate": 1.0293588882570318e-05, "loss": 0.0221, "step": 22071 }, { "epoch": 4.05, "learning_rate": 1.0292846771771852e-05, "loss": 0.0276, "step": 22072 }, { "epoch": 4.05, "learning_rate": 1.0292104659359208e-05, "loss": 0.0224, "step": 22073 }, { "epoch": 4.05, "learning_rate": 1.0291362545336478e-05, "loss": 0.027, "step": 22074 }, { "epoch": 4.05, "learning_rate": 1.0290620429707747e-05, "loss": 0.0307, "step": 22075 }, { "epoch": 4.05, "learning_rate": 1.0289878312477113e-05, "loss": 0.033, "step": 22076 }, { "epoch": 4.05, "learning_rate": 1.0289136193648664e-05, "loss": 0.0326, "step": 22077 }, { "epoch": 4.05, "learning_rate": 1.0288394073226486e-05, "loss": 0.0294, "step": 22078 }, { "epoch": 4.05, "learning_rate": 1.0287651951214675e-05, "loss": 0.0249, "step": 22079 }, { "epoch": 4.05, "learning_rate": 1.028690982761732e-05, "loss": 0.0321, "step": 22080 }, { "epoch": 4.05, "learning_rate": 1.0286167702438512e-05, "loss": 0.0266, "step": 22081 }, { "epoch": 4.05, "learning_rate": 1.0285425575682337e-05, "loss": 0.0298, "step": 22082 }, { "epoch": 4.05, "learning_rate": 1.0284683447352894e-05, "loss": 0.0269, "step": 22083 }, { "epoch": 4.05, "learning_rate": 1.0283941317454266e-05, "loss": 0.0309, "step": 22084 }, { "epoch": 4.05, "learning_rate": 1.0283199185990548e-05, "loss": 0.0231, "step": 22085 }, { "epoch": 4.05, "learning_rate": 1.028245705296583e-05, "loss": 0.0295, "step": 22086 }, { "epoch": 4.05, "learning_rate": 1.0281714918384203e-05, "loss": 0.0348, "step": 22087 }, { "epoch": 4.05, "learning_rate": 1.0280972782249754e-05, "loss": 0.0246, "step": 22088 }, { "epoch": 4.05, "learning_rate": 1.028023064456658e-05, "loss": 0.0323, "step": 22089 }, { "epoch": 4.05, "learning_rate": 1.0279488505338766e-05, "loss": 0.0317, "step": 22090 }, { "epoch": 4.05, "learning_rate": 1.0278746364570402e-05, "loss": 0.0305, "step": 22091 }, { "epoch": 4.05, "learning_rate": 1.0278004222265587e-05, "loss": 0.0306, "step": 22092 }, { "epoch": 4.05, "learning_rate": 1.0277262078428404e-05, "loss": 0.0268, "step": 22093 }, { "epoch": 4.05, "learning_rate": 1.0276519933062942e-05, "loss": 0.033, "step": 22094 }, { "epoch": 4.05, "learning_rate": 1.02757777861733e-05, "loss": 0.0264, "step": 22095 }, { "epoch": 4.05, "learning_rate": 1.0275035637763564e-05, "loss": 0.0273, "step": 22096 }, { "epoch": 4.05, "learning_rate": 1.0274293487837823e-05, "loss": 0.0245, "step": 22097 }, { "epoch": 4.05, "learning_rate": 1.0273551336400174e-05, "loss": 0.0332, "step": 22098 }, { "epoch": 4.05, "learning_rate": 1.0272809183454701e-05, "loss": 0.0252, "step": 22099 }, { "epoch": 4.05, "learning_rate": 1.0272067029005497e-05, "loss": 0.0303, "step": 22100 }, { "epoch": 4.05, "learning_rate": 1.0271324873056656e-05, "loss": 0.0277, "step": 22101 }, { "epoch": 4.05, "learning_rate": 1.027058271561226e-05, "loss": 0.0287, "step": 22102 }, { "epoch": 4.05, "learning_rate": 1.0269840556676411e-05, "loss": 0.0247, "step": 22103 }, { "epoch": 4.05, "learning_rate": 1.0269098396253193e-05, "loss": 0.0268, "step": 22104 }, { "epoch": 4.05, "learning_rate": 1.0268356234346699e-05, "loss": 0.0324, "step": 22105 }, { "epoch": 4.05, "learning_rate": 1.0267614070961017e-05, "loss": 0.0258, "step": 22106 }, { "epoch": 4.05, "learning_rate": 1.0266871906100242e-05, "loss": 0.0269, "step": 22107 }, { "epoch": 4.05, "learning_rate": 1.026612973976846e-05, "loss": 0.0273, "step": 22108 }, { "epoch": 4.05, "learning_rate": 1.0265387571969768e-05, "loss": 0.0289, "step": 22109 }, { "epoch": 4.05, "learning_rate": 1.0264645402708255e-05, "loss": 0.0313, "step": 22110 }, { "epoch": 4.05, "learning_rate": 1.0263903231988006e-05, "loss": 0.0282, "step": 22111 }, { "epoch": 4.06, "learning_rate": 1.0263161059813118e-05, "loss": 0.0271, "step": 22112 }, { "epoch": 4.06, "learning_rate": 1.026241888618768e-05, "loss": 0.0257, "step": 22113 }, { "epoch": 4.06, "learning_rate": 1.0261676711115782e-05, "loss": 0.0279, "step": 22114 }, { "epoch": 4.06, "learning_rate": 1.0260934534601516e-05, "loss": 0.0263, "step": 22115 }, { "epoch": 4.06, "learning_rate": 1.0260192356648974e-05, "loss": 0.0246, "step": 22116 }, { "epoch": 4.06, "learning_rate": 1.0259450177262245e-05, "loss": 0.0262, "step": 22117 }, { "epoch": 4.06, "learning_rate": 1.025870799644542e-05, "loss": 0.0252, "step": 22118 }, { "epoch": 4.06, "learning_rate": 1.0257965814202592e-05, "loss": 0.0302, "step": 22119 }, { "epoch": 4.06, "learning_rate": 1.025722363053785e-05, "loss": 0.0289, "step": 22120 }, { "epoch": 4.06, "learning_rate": 1.0256481445455285e-05, "loss": 0.0282, "step": 22121 }, { "epoch": 4.06, "learning_rate": 1.0255739258958988e-05, "loss": 0.0276, "step": 22122 }, { "epoch": 4.06, "learning_rate": 1.025499707105305e-05, "loss": 0.024, "step": 22123 }, { "epoch": 4.06, "learning_rate": 1.0254254881741562e-05, "loss": 0.0301, "step": 22124 }, { "epoch": 4.06, "learning_rate": 1.0253512691028617e-05, "loss": 0.0315, "step": 22125 }, { "epoch": 4.06, "learning_rate": 1.02527704989183e-05, "loss": 0.0257, "step": 22126 }, { "epoch": 4.06, "learning_rate": 1.025202830541471e-05, "loss": 0.0259, "step": 22127 }, { "epoch": 4.06, "learning_rate": 1.0251286110521933e-05, "loss": 0.021, "step": 22128 }, { "epoch": 4.06, "learning_rate": 1.025054391424406e-05, "loss": 0.0336, "step": 22129 }, { "epoch": 4.06, "learning_rate": 1.0249801716585187e-05, "loss": 0.0266, "step": 22130 }, { "epoch": 4.06, "learning_rate": 1.0249059517549395e-05, "loss": 0.0286, "step": 22131 }, { "epoch": 4.06, "learning_rate": 1.0248317317140785e-05, "loss": 0.0286, "step": 22132 }, { "epoch": 4.06, "learning_rate": 1.0247575115363444e-05, "loss": 0.0231, "step": 22133 }, { "epoch": 4.06, "learning_rate": 1.0246832912221463e-05, "loss": 0.026, "step": 22134 }, { "epoch": 4.06, "learning_rate": 1.024609070771893e-05, "loss": 0.0287, "step": 22135 }, { "epoch": 4.06, "learning_rate": 1.0245348501859945e-05, "loss": 0.0199, "step": 22136 }, { "epoch": 4.06, "learning_rate": 1.024460629464859e-05, "loss": 0.0272, "step": 22137 }, { "epoch": 4.06, "learning_rate": 1.024386408608896e-05, "loss": 0.0334, "step": 22138 }, { "epoch": 4.06, "learning_rate": 1.0243121876185144e-05, "loss": 0.0262, "step": 22139 }, { "epoch": 4.06, "learning_rate": 1.0242379664941236e-05, "loss": 0.0246, "step": 22140 }, { "epoch": 4.06, "learning_rate": 1.0241637452361323e-05, "loss": 0.0254, "step": 22141 }, { "epoch": 4.06, "learning_rate": 1.0240895238449504e-05, "loss": 0.0281, "step": 22142 }, { "epoch": 4.06, "learning_rate": 1.024015302320986e-05, "loss": 0.0309, "step": 22143 }, { "epoch": 4.06, "learning_rate": 1.0239410806646489e-05, "loss": 0.0306, "step": 22144 }, { "epoch": 4.06, "learning_rate": 1.0238668588763479e-05, "loss": 0.0296, "step": 22145 }, { "epoch": 4.06, "learning_rate": 1.0237926369564924e-05, "loss": 0.0274, "step": 22146 }, { "epoch": 4.06, "learning_rate": 1.023718414905491e-05, "loss": 0.0275, "step": 22147 }, { "epoch": 4.06, "learning_rate": 1.0236441927237534e-05, "loss": 0.0312, "step": 22148 }, { "epoch": 4.06, "learning_rate": 1.0235699704116886e-05, "loss": 0.028, "step": 22149 }, { "epoch": 4.06, "learning_rate": 1.023495747969705e-05, "loss": 0.0293, "step": 22150 }, { "epoch": 4.06, "learning_rate": 1.023421525398213e-05, "loss": 0.0266, "step": 22151 }, { "epoch": 4.06, "learning_rate": 1.0233473026976206e-05, "loss": 0.0217, "step": 22152 }, { "epoch": 4.06, "learning_rate": 1.0232730798683373e-05, "loss": 0.0257, "step": 22153 }, { "epoch": 4.06, "learning_rate": 1.0231988569107722e-05, "loss": 0.0239, "step": 22154 }, { "epoch": 4.06, "learning_rate": 1.0231246338253346e-05, "loss": 0.0285, "step": 22155 }, { "epoch": 4.06, "learning_rate": 1.0230504106124335e-05, "loss": 0.0267, "step": 22156 }, { "epoch": 4.06, "learning_rate": 1.0229761872724781e-05, "loss": 0.034, "step": 22157 }, { "epoch": 4.06, "learning_rate": 1.0229019638058772e-05, "loss": 0.0316, "step": 22158 }, { "epoch": 4.06, "learning_rate": 1.02282774021304e-05, "loss": 0.0261, "step": 22159 }, { "epoch": 4.06, "learning_rate": 1.0227535164943763e-05, "loss": 0.0259, "step": 22160 }, { "epoch": 4.06, "learning_rate": 1.0226792926502942e-05, "loss": 0.0193, "step": 22161 }, { "epoch": 4.06, "learning_rate": 1.0226050686812034e-05, "loss": 0.0246, "step": 22162 }, { "epoch": 4.06, "learning_rate": 1.0225308445875133e-05, "loss": 0.0285, "step": 22163 }, { "epoch": 4.06, "learning_rate": 1.0224566203696322e-05, "loss": 0.0272, "step": 22164 }, { "epoch": 4.06, "learning_rate": 1.0223823960279697e-05, "loss": 0.0273, "step": 22165 }, { "epoch": 4.06, "learning_rate": 1.0223081715629353e-05, "loss": 0.0252, "step": 22166 }, { "epoch": 4.07, "learning_rate": 1.0222339469749374e-05, "loss": 0.0286, "step": 22167 }, { "epoch": 4.07, "learning_rate": 1.0221597222643853e-05, "loss": 0.0269, "step": 22168 }, { "epoch": 4.07, "learning_rate": 1.0220854974316888e-05, "loss": 0.0242, "step": 22169 }, { "epoch": 4.07, "learning_rate": 1.0220112724772559e-05, "loss": 0.0273, "step": 22170 }, { "epoch": 4.07, "learning_rate": 1.0219370474014969e-05, "loss": 0.0294, "step": 22171 }, { "epoch": 4.07, "learning_rate": 1.0218628222048202e-05, "loss": 0.023, "step": 22172 }, { "epoch": 4.07, "learning_rate": 1.0217885968876351e-05, "loss": 0.0273, "step": 22173 }, { "epoch": 4.07, "learning_rate": 1.0217143714503508e-05, "loss": 0.0256, "step": 22174 }, { "epoch": 4.07, "learning_rate": 1.0216401458933762e-05, "loss": 0.0335, "step": 22175 }, { "epoch": 4.07, "learning_rate": 1.0215659202171209e-05, "loss": 0.0294, "step": 22176 }, { "epoch": 4.07, "learning_rate": 1.0214916944219932e-05, "loss": 0.0284, "step": 22177 }, { "epoch": 4.07, "learning_rate": 1.0214174685084036e-05, "loss": 0.0261, "step": 22178 }, { "epoch": 4.07, "learning_rate": 1.0213432424767598e-05, "loss": 0.0238, "step": 22179 }, { "epoch": 4.07, "learning_rate": 1.0212690163274717e-05, "loss": 0.0286, "step": 22180 }, { "epoch": 4.07, "learning_rate": 1.0211947900609482e-05, "loss": 0.0286, "step": 22181 }, { "epoch": 4.07, "learning_rate": 1.0211205636775987e-05, "loss": 0.0275, "step": 22182 }, { "epoch": 4.07, "learning_rate": 1.021046337177832e-05, "loss": 0.0278, "step": 22183 }, { "epoch": 4.07, "learning_rate": 1.0209721105620576e-05, "loss": 0.0243, "step": 22184 }, { "epoch": 4.07, "learning_rate": 1.0208978838306842e-05, "loss": 0.0281, "step": 22185 }, { "epoch": 4.07, "learning_rate": 1.0208236569841212e-05, "loss": 0.0271, "step": 22186 }, { "epoch": 4.07, "learning_rate": 1.0207494300227781e-05, "loss": 0.0251, "step": 22187 }, { "epoch": 4.07, "learning_rate": 1.020675202947063e-05, "loss": 0.0247, "step": 22188 }, { "epoch": 4.07, "learning_rate": 1.020600975757386e-05, "loss": 0.0257, "step": 22189 }, { "epoch": 4.07, "learning_rate": 1.020526748454156e-05, "loss": 0.0235, "step": 22190 }, { "epoch": 4.07, "learning_rate": 1.0204525210377822e-05, "loss": 0.0218, "step": 22191 }, { "epoch": 4.07, "learning_rate": 1.0203782935086733e-05, "loss": 0.0293, "step": 22192 }, { "epoch": 4.07, "learning_rate": 1.0203040658672391e-05, "loss": 0.026, "step": 22193 }, { "epoch": 4.07, "learning_rate": 1.0202298381138882e-05, "loss": 0.0355, "step": 22194 }, { "epoch": 4.07, "learning_rate": 1.02015561024903e-05, "loss": 0.0226, "step": 22195 }, { "epoch": 4.07, "learning_rate": 1.020081382273074e-05, "loss": 0.0265, "step": 22196 }, { "epoch": 4.07, "learning_rate": 1.0200071541864283e-05, "loss": 0.0328, "step": 22197 }, { "epoch": 4.07, "learning_rate": 1.019932925989503e-05, "loss": 0.0283, "step": 22198 }, { "epoch": 4.07, "learning_rate": 1.019858697682707e-05, "loss": 0.0247, "step": 22199 }, { "epoch": 4.07, "learning_rate": 1.0197844692664494e-05, "loss": 0.0269, "step": 22200 }, { "epoch": 4.07, "learning_rate": 1.0197102407411392e-05, "loss": 0.0317, "step": 22201 }, { "epoch": 4.07, "learning_rate": 1.0196360121071857e-05, "loss": 0.0231, "step": 22202 }, { "epoch": 4.07, "learning_rate": 1.019561783364998e-05, "loss": 0.0282, "step": 22203 }, { "epoch": 4.07, "learning_rate": 1.0194875545149854e-05, "loss": 0.0233, "step": 22204 }, { "epoch": 4.07, "learning_rate": 1.019413325557557e-05, "loss": 0.0317, "step": 22205 }, { "epoch": 4.07, "learning_rate": 1.0193390964931218e-05, "loss": 0.0223, "step": 22206 }, { "epoch": 4.07, "learning_rate": 1.0192648673220891e-05, "loss": 0.0282, "step": 22207 }, { "epoch": 4.07, "learning_rate": 1.019190638044868e-05, "loss": 0.0231, "step": 22208 }, { "epoch": 4.07, "learning_rate": 1.0191164086618675e-05, "loss": 0.0266, "step": 22209 }, { "epoch": 4.07, "learning_rate": 1.019042179173497e-05, "loss": 0.0198, "step": 22210 }, { "epoch": 4.07, "learning_rate": 1.0189679495801656e-05, "loss": 0.0287, "step": 22211 }, { "epoch": 4.07, "learning_rate": 1.0188937198822823e-05, "loss": 0.0233, "step": 22212 }, { "epoch": 4.07, "learning_rate": 1.0188194900802565e-05, "loss": 0.0264, "step": 22213 }, { "epoch": 4.07, "learning_rate": 1.0187452601744971e-05, "loss": 0.0186, "step": 22214 }, { "epoch": 4.07, "learning_rate": 1.0186710301654138e-05, "loss": 0.0216, "step": 22215 }, { "epoch": 4.07, "learning_rate": 1.018596800053415e-05, "loss": 0.0231, "step": 22216 }, { "epoch": 4.07, "learning_rate": 1.0185225698389103e-05, "loss": 0.0258, "step": 22217 }, { "epoch": 4.07, "learning_rate": 1.0184483395223085e-05, "loss": 0.027, "step": 22218 }, { "epoch": 4.07, "learning_rate": 1.0183741091040193e-05, "loss": 0.0268, "step": 22219 }, { "epoch": 4.07, "learning_rate": 1.0182998785844515e-05, "loss": 0.0253, "step": 22220 }, { "epoch": 4.08, "learning_rate": 1.0182256479640142e-05, "loss": 0.0283, "step": 22221 }, { "epoch": 4.08, "learning_rate": 1.018151417243117e-05, "loss": 0.0259, "step": 22222 }, { "epoch": 4.08, "learning_rate": 1.0180771864221685e-05, "loss": 0.0229, "step": 22223 }, { "epoch": 4.08, "learning_rate": 1.0180029555015784e-05, "loss": 0.024, "step": 22224 }, { "epoch": 4.08, "learning_rate": 1.0179287244817555e-05, "loss": 0.0282, "step": 22225 }, { "epoch": 4.08, "learning_rate": 1.017854493363109e-05, "loss": 0.03, "step": 22226 }, { "epoch": 4.08, "learning_rate": 1.017780262146048e-05, "loss": 0.0257, "step": 22227 }, { "epoch": 4.08, "learning_rate": 1.0177060308309818e-05, "loss": 0.031, "step": 22228 }, { "epoch": 4.08, "learning_rate": 1.0176317994183198e-05, "loss": 0.0225, "step": 22229 }, { "epoch": 4.08, "learning_rate": 1.0175575679084704e-05, "loss": 0.029, "step": 22230 }, { "epoch": 4.08, "learning_rate": 1.0174833363018438e-05, "loss": 0.0224, "step": 22231 }, { "epoch": 4.08, "learning_rate": 1.0174091045988483e-05, "loss": 0.0255, "step": 22232 }, { "epoch": 4.08, "learning_rate": 1.0173348727998936e-05, "loss": 0.03, "step": 22233 }, { "epoch": 4.08, "learning_rate": 1.0172606409053887e-05, "loss": 0.0262, "step": 22234 }, { "epoch": 4.08, "learning_rate": 1.0171864089157426e-05, "loss": 0.0256, "step": 22235 }, { "epoch": 4.08, "learning_rate": 1.0171121768313646e-05, "loss": 0.0289, "step": 22236 }, { "epoch": 4.08, "learning_rate": 1.0170379446526641e-05, "loss": 0.035, "step": 22237 }, { "epoch": 4.08, "learning_rate": 1.01696371238005e-05, "loss": 0.0293, "step": 22238 }, { "epoch": 4.08, "learning_rate": 1.0168894800139311e-05, "loss": 0.0276, "step": 22239 }, { "epoch": 4.08, "learning_rate": 1.0168152475547177e-05, "loss": 0.0316, "step": 22240 }, { "epoch": 4.08, "learning_rate": 1.0167410150028176e-05, "loss": 0.0259, "step": 22241 }, { "epoch": 4.08, "learning_rate": 1.016666782358641e-05, "loss": 0.027, "step": 22242 }, { "epoch": 4.08, "learning_rate": 1.0165925496225965e-05, "loss": 0.0295, "step": 22243 }, { "epoch": 4.08, "learning_rate": 1.0165183167950937e-05, "loss": 0.0227, "step": 22244 }, { "epoch": 4.08, "learning_rate": 1.0164440838765413e-05, "loss": 0.0268, "step": 22245 }, { "epoch": 4.08, "learning_rate": 1.016369850867349e-05, "loss": 0.025, "step": 22246 }, { "epoch": 4.08, "learning_rate": 1.0162956177679256e-05, "loss": 0.0241, "step": 22247 }, { "epoch": 4.08, "learning_rate": 1.0162213845786802e-05, "loss": 0.0266, "step": 22248 }, { "epoch": 4.08, "learning_rate": 1.0161471513000226e-05, "loss": 0.0238, "step": 22249 }, { "epoch": 4.08, "learning_rate": 1.0160729179323609e-05, "loss": 0.0254, "step": 22250 }, { "epoch": 4.08, "learning_rate": 1.0159986844761052e-05, "loss": 0.028, "step": 22251 }, { "epoch": 4.08, "learning_rate": 1.0159244509316643e-05, "loss": 0.0236, "step": 22252 }, { "epoch": 4.08, "learning_rate": 1.0158502172994477e-05, "loss": 0.0225, "step": 22253 }, { "epoch": 4.08, "learning_rate": 1.015775983579864e-05, "loss": 0.0337, "step": 22254 }, { "epoch": 4.08, "learning_rate": 1.015701749773323e-05, "loss": 0.0226, "step": 22255 }, { "epoch": 4.08, "learning_rate": 1.0156275158802333e-05, "loss": 0.0237, "step": 22256 }, { "epoch": 4.08, "learning_rate": 1.0155532819010045e-05, "loss": 0.0264, "step": 22257 }, { "epoch": 4.08, "learning_rate": 1.0154790478360458e-05, "loss": 0.0226, "step": 22258 }, { "epoch": 4.08, "learning_rate": 1.0154048136857662e-05, "loss": 0.0282, "step": 22259 }, { "epoch": 4.08, "learning_rate": 1.0153305794505748e-05, "loss": 0.0294, "step": 22260 }, { "epoch": 4.08, "learning_rate": 1.015256345130881e-05, "loss": 0.0256, "step": 22261 }, { "epoch": 4.08, "learning_rate": 1.0151821107270937e-05, "loss": 0.0303, "step": 22262 }, { "epoch": 4.08, "learning_rate": 1.015107876239622e-05, "loss": 0.0275, "step": 22263 }, { "epoch": 4.08, "learning_rate": 1.015033641668876e-05, "loss": 0.0287, "step": 22264 }, { "epoch": 4.08, "learning_rate": 1.0149594070152638e-05, "loss": 0.0278, "step": 22265 }, { "epoch": 4.08, "learning_rate": 1.014885172279195e-05, "loss": 0.0295, "step": 22266 }, { "epoch": 4.08, "learning_rate": 1.0148109374610788e-05, "loss": 0.0241, "step": 22267 }, { "epoch": 4.08, "learning_rate": 1.0147367025613243e-05, "loss": 0.024, "step": 22268 }, { "epoch": 4.08, "learning_rate": 1.0146624675803408e-05, "loss": 0.0242, "step": 22269 }, { "epoch": 4.08, "learning_rate": 1.0145882325185377e-05, "loss": 0.0284, "step": 22270 }, { "epoch": 4.08, "learning_rate": 1.0145139973763239e-05, "loss": 0.0259, "step": 22271 }, { "epoch": 4.08, "learning_rate": 1.0144397621541082e-05, "loss": 0.0328, "step": 22272 }, { "epoch": 4.08, "learning_rate": 1.0143655268523005e-05, "loss": 0.0242, "step": 22273 }, { "epoch": 4.08, "learning_rate": 1.0142912914713094e-05, "loss": 0.021, "step": 22274 }, { "epoch": 4.08, "learning_rate": 1.0142170560115448e-05, "loss": 0.0287, "step": 22275 }, { "epoch": 4.09, "learning_rate": 1.0141428204734153e-05, "loss": 0.0252, "step": 22276 }, { "epoch": 4.09, "learning_rate": 1.01406858485733e-05, "loss": 0.0247, "step": 22277 }, { "epoch": 4.09, "learning_rate": 1.0139943491636986e-05, "loss": 0.0259, "step": 22278 }, { "epoch": 4.09, "learning_rate": 1.0139201133929298e-05, "loss": 0.0276, "step": 22279 }, { "epoch": 4.09, "learning_rate": 1.0138458775454331e-05, "loss": 0.0269, "step": 22280 }, { "epoch": 4.09, "learning_rate": 1.0137716416216174e-05, "loss": 0.0393, "step": 22281 }, { "epoch": 4.09, "learning_rate": 1.0136974056218927e-05, "loss": 0.0225, "step": 22282 }, { "epoch": 4.09, "learning_rate": 1.013623169546667e-05, "loss": 0.0264, "step": 22283 }, { "epoch": 4.09, "learning_rate": 1.0135489333963502e-05, "loss": 0.0252, "step": 22284 }, { "epoch": 4.09, "learning_rate": 1.0134746971713514e-05, "loss": 0.0306, "step": 22285 }, { "epoch": 4.09, "learning_rate": 1.0134004608720798e-05, "loss": 0.0306, "step": 22286 }, { "epoch": 4.09, "learning_rate": 1.0133262244989446e-05, "loss": 0.0248, "step": 22287 }, { "epoch": 4.09, "learning_rate": 1.0132519880523548e-05, "loss": 0.0246, "step": 22288 }, { "epoch": 4.09, "learning_rate": 1.0131777515327197e-05, "loss": 0.0306, "step": 22289 }, { "epoch": 4.09, "learning_rate": 1.0131035149404485e-05, "loss": 0.0261, "step": 22290 }, { "epoch": 4.09, "learning_rate": 1.0130292782759508e-05, "loss": 0.0214, "step": 22291 }, { "epoch": 4.09, "learning_rate": 1.012955041539635e-05, "loss": 0.0271, "step": 22292 }, { "epoch": 4.09, "learning_rate": 1.0128808047319108e-05, "loss": 0.0253, "step": 22293 }, { "epoch": 4.09, "learning_rate": 1.0128065678531873e-05, "loss": 0.0266, "step": 22294 }, { "epoch": 4.09, "learning_rate": 1.012732330903874e-05, "loss": 0.0291, "step": 22295 }, { "epoch": 4.09, "learning_rate": 1.0126580938843795e-05, "loss": 0.0273, "step": 22296 }, { "epoch": 4.09, "learning_rate": 1.0125838567951132e-05, "loss": 0.0242, "step": 22297 }, { "epoch": 4.09, "learning_rate": 1.0125096196364843e-05, "loss": 0.026, "step": 22298 }, { "epoch": 4.09, "learning_rate": 1.0124353824089024e-05, "loss": 0.0305, "step": 22299 }, { "epoch": 4.09, "learning_rate": 1.0123611451127763e-05, "loss": 0.0265, "step": 22300 }, { "epoch": 4.09, "learning_rate": 1.0122869077485154e-05, "loss": 0.0237, "step": 22301 }, { "epoch": 4.09, "learning_rate": 1.0122126703165284e-05, "loss": 0.0256, "step": 22302 }, { "epoch": 4.09, "learning_rate": 1.0121384328172252e-05, "loss": 0.0282, "step": 22303 }, { "epoch": 4.09, "learning_rate": 1.0120641952510147e-05, "loss": 0.0272, "step": 22304 }, { "epoch": 4.09, "learning_rate": 1.0119899576183061e-05, "loss": 0.0239, "step": 22305 }, { "epoch": 4.09, "learning_rate": 1.0119157199195085e-05, "loss": 0.0213, "step": 22306 }, { "epoch": 4.09, "learning_rate": 1.0118414821550308e-05, "loss": 0.0281, "step": 22307 }, { "epoch": 4.09, "learning_rate": 1.0117672443252829e-05, "loss": 0.0258, "step": 22308 }, { "epoch": 4.09, "learning_rate": 1.0116930064306736e-05, "loss": 0.0236, "step": 22309 }, { "epoch": 4.09, "learning_rate": 1.0116187684716123e-05, "loss": 0.0248, "step": 22310 }, { "epoch": 4.09, "learning_rate": 1.0115445304485081e-05, "loss": 0.0259, "step": 22311 }, { "epoch": 4.09, "learning_rate": 1.0114702923617701e-05, "loss": 0.031, "step": 22312 }, { "epoch": 4.09, "learning_rate": 1.0113960542118075e-05, "loss": 0.0228, "step": 22313 }, { "epoch": 4.09, "learning_rate": 1.0113218159990296e-05, "loss": 0.0256, "step": 22314 }, { "epoch": 4.09, "learning_rate": 1.0112475777238457e-05, "loss": 0.0323, "step": 22315 }, { "epoch": 4.09, "learning_rate": 1.0111733393866648e-05, "loss": 0.0281, "step": 22316 }, { "epoch": 4.09, "learning_rate": 1.0110991009878963e-05, "loss": 0.0258, "step": 22317 }, { "epoch": 4.09, "learning_rate": 1.0110248625279491e-05, "loss": 0.0234, "step": 22318 }, { "epoch": 4.09, "learning_rate": 1.0109506240072326e-05, "loss": 0.027, "step": 22319 }, { "epoch": 4.09, "learning_rate": 1.0108763854261561e-05, "loss": 0.0301, "step": 22320 }, { "epoch": 4.09, "learning_rate": 1.0108021467851287e-05, "loss": 0.0298, "step": 22321 }, { "epoch": 4.09, "learning_rate": 1.0107279080845595e-05, "loss": 0.0274, "step": 22322 }, { "epoch": 4.09, "learning_rate": 1.010653669324858e-05, "loss": 0.0274, "step": 22323 }, { "epoch": 4.09, "learning_rate": 1.0105794305064333e-05, "loss": 0.026, "step": 22324 }, { "epoch": 4.09, "learning_rate": 1.0105051916296942e-05, "loss": 0.0296, "step": 22325 }, { "epoch": 4.09, "learning_rate": 1.0104309526950508e-05, "loss": 0.0299, "step": 22326 }, { "epoch": 4.09, "learning_rate": 1.0103567137029111e-05, "loss": 0.0258, "step": 22327 }, { "epoch": 4.09, "learning_rate": 1.0102824746536852e-05, "loss": 0.0251, "step": 22328 }, { "epoch": 4.09, "learning_rate": 1.0102082355477822e-05, "loss": 0.0213, "step": 22329 }, { "epoch": 4.09, "learning_rate": 1.0101339963856112e-05, "loss": 0.0311, "step": 22330 }, { "epoch": 4.1, "learning_rate": 1.010059757167581e-05, "loss": 0.0269, "step": 22331 }, { "epoch": 4.1, "learning_rate": 1.0099855178941016e-05, "loss": 0.0204, "step": 22332 }, { "epoch": 4.1, "learning_rate": 1.0099112785655814e-05, "loss": 0.0223, "step": 22333 }, { "epoch": 4.1, "learning_rate": 1.0098370391824303e-05, "loss": 0.0262, "step": 22334 }, { "epoch": 4.1, "learning_rate": 1.0097627997450574e-05, "loss": 0.0253, "step": 22335 }, { "epoch": 4.1, "learning_rate": 1.009688560253871e-05, "loss": 0.0279, "step": 22336 }, { "epoch": 4.1, "learning_rate": 1.0096143207092817e-05, "loss": 0.0266, "step": 22337 }, { "epoch": 4.1, "learning_rate": 1.0095400811116978e-05, "loss": 0.0314, "step": 22338 }, { "epoch": 4.1, "learning_rate": 1.0094658414615288e-05, "loss": 0.0294, "step": 22339 }, { "epoch": 4.1, "learning_rate": 1.0093916017591836e-05, "loss": 0.0275, "step": 22340 }, { "epoch": 4.1, "learning_rate": 1.0093173620050723e-05, "loss": 0.0252, "step": 22341 }, { "epoch": 4.1, "learning_rate": 1.0092431221996026e-05, "loss": 0.0249, "step": 22342 }, { "epoch": 4.1, "learning_rate": 1.009168882343185e-05, "loss": 0.0204, "step": 22343 }, { "epoch": 4.1, "learning_rate": 1.009094642436229e-05, "loss": 0.0254, "step": 22344 }, { "epoch": 4.1, "learning_rate": 1.009020402479142e-05, "loss": 0.0266, "step": 22345 }, { "epoch": 4.1, "learning_rate": 1.0089461624723349e-05, "loss": 0.0215, "step": 22346 }, { "epoch": 4.1, "learning_rate": 1.0088719224162163e-05, "loss": 0.0314, "step": 22347 }, { "epoch": 4.1, "learning_rate": 1.0087976823111956e-05, "loss": 0.0224, "step": 22348 }, { "epoch": 4.1, "learning_rate": 1.0087234421576812e-05, "loss": 0.0315, "step": 22349 }, { "epoch": 4.1, "learning_rate": 1.0086492019560839e-05, "loss": 0.0241, "step": 22350 }, { "epoch": 4.1, "learning_rate": 1.0085749617068113e-05, "loss": 0.0294, "step": 22351 }, { "epoch": 4.1, "learning_rate": 1.0085007214102736e-05, "loss": 0.026, "step": 22352 }, { "epoch": 4.1, "learning_rate": 1.0084264810668797e-05, "loss": 0.0251, "step": 22353 }, { "epoch": 4.1, "learning_rate": 1.008352240677039e-05, "loss": 0.0245, "step": 22354 }, { "epoch": 4.1, "learning_rate": 1.0082780002411602e-05, "loss": 0.0289, "step": 22355 }, { "epoch": 4.1, "learning_rate": 1.0082037597596532e-05, "loss": 0.0248, "step": 22356 }, { "epoch": 4.1, "learning_rate": 1.0081295192329269e-05, "loss": 0.0276, "step": 22357 }, { "epoch": 4.1, "learning_rate": 1.0080552786613899e-05, "loss": 0.0261, "step": 22358 }, { "epoch": 4.1, "learning_rate": 1.0079810380454528e-05, "loss": 0.0281, "step": 22359 }, { "epoch": 4.1, "learning_rate": 1.0079067973855234e-05, "loss": 0.0284, "step": 22360 }, { "epoch": 4.1, "learning_rate": 1.007832556682012e-05, "loss": 0.029, "step": 22361 }, { "epoch": 4.1, "learning_rate": 1.0077583159353272e-05, "loss": 0.0184, "step": 22362 }, { "epoch": 4.1, "learning_rate": 1.0076840751458785e-05, "loss": 0.0276, "step": 22363 }, { "epoch": 4.1, "learning_rate": 1.0076098343140748e-05, "loss": 0.0305, "step": 22364 }, { "epoch": 4.1, "learning_rate": 1.0075355934403258e-05, "loss": 0.0316, "step": 22365 }, { "epoch": 4.1, "learning_rate": 1.0074613525250401e-05, "loss": 0.026, "step": 22366 }, { "epoch": 4.1, "learning_rate": 1.0073871115686272e-05, "loss": 0.0286, "step": 22367 }, { "epoch": 4.1, "learning_rate": 1.0073128705714967e-05, "loss": 0.0253, "step": 22368 }, { "epoch": 4.1, "learning_rate": 1.0072386295340572e-05, "loss": 0.0295, "step": 22369 }, { "epoch": 4.1, "learning_rate": 1.0071643884567184e-05, "loss": 0.0262, "step": 22370 }, { "epoch": 4.1, "learning_rate": 1.0070901473398893e-05, "loss": 0.0261, "step": 22371 }, { "epoch": 4.1, "learning_rate": 1.007015906183979e-05, "loss": 0.0274, "step": 22372 }, { "epoch": 4.1, "learning_rate": 1.006941664989397e-05, "loss": 0.0248, "step": 22373 }, { "epoch": 4.1, "learning_rate": 1.0068674237565523e-05, "loss": 0.0234, "step": 22374 }, { "epoch": 4.1, "learning_rate": 1.0067931824858542e-05, "loss": 0.0335, "step": 22375 }, { "epoch": 4.1, "learning_rate": 1.0067189411777118e-05, "loss": 0.0324, "step": 22376 }, { "epoch": 4.1, "learning_rate": 1.0066446998325349e-05, "loss": 0.0282, "step": 22377 }, { "epoch": 4.1, "learning_rate": 1.0065704584507318e-05, "loss": 0.0262, "step": 22378 }, { "epoch": 4.1, "learning_rate": 1.0064962170327123e-05, "loss": 0.0329, "step": 22379 }, { "epoch": 4.1, "learning_rate": 1.0064219755788857e-05, "loss": 0.0241, "step": 22380 }, { "epoch": 4.1, "learning_rate": 1.0063477340896609e-05, "loss": 0.0253, "step": 22381 }, { "epoch": 4.1, "learning_rate": 1.0062734925654472e-05, "loss": 0.0235, "step": 22382 }, { "epoch": 4.1, "learning_rate": 1.0061992510066538e-05, "loss": 0.0239, "step": 22383 }, { "epoch": 4.1, "learning_rate": 1.0061250094136898e-05, "loss": 0.0369, "step": 22384 }, { "epoch": 4.11, "learning_rate": 1.0060507677869648e-05, "loss": 0.0252, "step": 22385 }, { "epoch": 4.11, "learning_rate": 1.0059765261268882e-05, "loss": 0.0261, "step": 22386 }, { "epoch": 4.11, "learning_rate": 1.0059022844338684e-05, "loss": 0.0315, "step": 22387 }, { "epoch": 4.11, "learning_rate": 1.0058280427083152e-05, "loss": 0.0301, "step": 22388 }, { "epoch": 4.11, "learning_rate": 1.0057538009506378e-05, "loss": 0.0332, "step": 22389 }, { "epoch": 4.11, "learning_rate": 1.0056795591612452e-05, "loss": 0.0298, "step": 22390 }, { "epoch": 4.11, "learning_rate": 1.0056053173405467e-05, "loss": 0.0275, "step": 22391 }, { "epoch": 4.11, "learning_rate": 1.0055310754889516e-05, "loss": 0.0243, "step": 22392 }, { "epoch": 4.11, "learning_rate": 1.0054568336068688e-05, "loss": 0.0303, "step": 22393 }, { "epoch": 4.11, "learning_rate": 1.0053825916947084e-05, "loss": 0.0215, "step": 22394 }, { "epoch": 4.11, "learning_rate": 1.0053083497528787e-05, "loss": 0.027, "step": 22395 }, { "epoch": 4.11, "learning_rate": 1.0052341077817892e-05, "loss": 0.0279, "step": 22396 }, { "epoch": 4.11, "learning_rate": 1.005159865781849e-05, "loss": 0.025, "step": 22397 }, { "epoch": 4.11, "learning_rate": 1.0050856237534678e-05, "loss": 0.0264, "step": 22398 }, { "epoch": 4.11, "learning_rate": 1.0050113816970546e-05, "loss": 0.0246, "step": 22399 }, { "epoch": 4.11, "learning_rate": 1.0049371396130185e-05, "loss": 0.0266, "step": 22400 }, { "epoch": 4.11, "learning_rate": 1.0048628975017686e-05, "loss": 0.0208, "step": 22401 }, { "epoch": 4.11, "learning_rate": 1.0047886553637142e-05, "loss": 0.026, "step": 22402 }, { "epoch": 4.11, "learning_rate": 1.004714413199265e-05, "loss": 0.0311, "step": 22403 }, { "epoch": 4.11, "learning_rate": 1.0046401710088296e-05, "loss": 0.0266, "step": 22404 }, { "epoch": 4.11, "learning_rate": 1.0045659287928173e-05, "loss": 0.0245, "step": 22405 }, { "epoch": 4.11, "learning_rate": 1.0044916865516378e-05, "loss": 0.0261, "step": 22406 }, { "epoch": 4.11, "learning_rate": 1.0044174442857e-05, "loss": 0.0294, "step": 22407 }, { "epoch": 4.11, "learning_rate": 1.0043432019954131e-05, "loss": 0.0305, "step": 22408 }, { "epoch": 4.11, "learning_rate": 1.0042689596811864e-05, "loss": 0.0273, "step": 22409 }, { "epoch": 4.11, "learning_rate": 1.004194717343429e-05, "loss": 0.0279, "step": 22410 }, { "epoch": 4.11, "learning_rate": 1.0041204749825499e-05, "loss": 0.0286, "step": 22411 }, { "epoch": 4.11, "learning_rate": 1.0040462325989592e-05, "loss": 0.0373, "step": 22412 }, { "epoch": 4.11, "learning_rate": 1.0039719901930654e-05, "loss": 0.0272, "step": 22413 }, { "epoch": 4.11, "learning_rate": 1.0038977477652779e-05, "loss": 0.0311, "step": 22414 }, { "epoch": 4.11, "learning_rate": 1.0038235053160058e-05, "loss": 0.0257, "step": 22415 }, { "epoch": 4.11, "learning_rate": 1.0037492628456586e-05, "loss": 0.0231, "step": 22416 }, { "epoch": 4.11, "learning_rate": 1.003675020354645e-05, "loss": 0.0248, "step": 22417 }, { "epoch": 4.11, "learning_rate": 1.0036007778433752e-05, "loss": 0.0209, "step": 22418 }, { "epoch": 4.11, "learning_rate": 1.0035265353122577e-05, "loss": 0.0284, "step": 22419 }, { "epoch": 4.11, "learning_rate": 1.0034522927617014e-05, "loss": 0.0305, "step": 22420 }, { "epoch": 4.11, "learning_rate": 1.0033780501921164e-05, "loss": 0.0222, "step": 22421 }, { "epoch": 4.11, "learning_rate": 1.0033038076039113e-05, "loss": 0.025, "step": 22422 }, { "epoch": 4.11, "learning_rate": 1.0032295649974957e-05, "loss": 0.03, "step": 22423 }, { "epoch": 4.11, "learning_rate": 1.0031553223732784e-05, "loss": 0.0225, "step": 22424 }, { "epoch": 4.11, "learning_rate": 1.0030810797316692e-05, "loss": 0.025, "step": 22425 }, { "epoch": 4.11, "learning_rate": 1.0030068370730766e-05, "loss": 0.026, "step": 22426 }, { "epoch": 4.11, "learning_rate": 1.002932594397911e-05, "loss": 0.0279, "step": 22427 }, { "epoch": 4.11, "learning_rate": 1.00285835170658e-05, "loss": 0.0281, "step": 22428 }, { "epoch": 4.11, "learning_rate": 1.0027841089994941e-05, "loss": 0.0227, "step": 22429 }, { "epoch": 4.11, "learning_rate": 1.0027098662770625e-05, "loss": 0.0222, "step": 22430 }, { "epoch": 4.11, "learning_rate": 1.0026356235396935e-05, "loss": 0.0299, "step": 22431 }, { "epoch": 4.11, "learning_rate": 1.002561380787797e-05, "loss": 0.0309, "step": 22432 }, { "epoch": 4.11, "learning_rate": 1.0024871380217822e-05, "loss": 0.0262, "step": 22433 }, { "epoch": 4.11, "learning_rate": 1.0024128952420583e-05, "loss": 0.0222, "step": 22434 }, { "epoch": 4.11, "learning_rate": 1.002338652449034e-05, "loss": 0.0293, "step": 22435 }, { "epoch": 4.11, "learning_rate": 1.0022644096431196e-05, "loss": 0.0207, "step": 22436 }, { "epoch": 4.11, "learning_rate": 1.0021901668247234e-05, "loss": 0.0232, "step": 22437 }, { "epoch": 4.11, "learning_rate": 1.0021159239942551e-05, "loss": 0.0301, "step": 22438 }, { "epoch": 4.11, "learning_rate": 1.0020416811521237e-05, "loss": 0.0263, "step": 22439 }, { "epoch": 4.12, "learning_rate": 1.0019674382987382e-05, "loss": 0.0339, "step": 22440 }, { "epoch": 4.12, "learning_rate": 1.0018931954345088e-05, "loss": 0.0245, "step": 22441 }, { "epoch": 4.12, "learning_rate": 1.0018189525598436e-05, "loss": 0.0236, "step": 22442 }, { "epoch": 4.12, "learning_rate": 1.0017447096751525e-05, "loss": 0.0248, "step": 22443 }, { "epoch": 4.12, "learning_rate": 1.001670466780844e-05, "loss": 0.0282, "step": 22444 }, { "epoch": 4.12, "learning_rate": 1.0015962238773287e-05, "loss": 0.0193, "step": 22445 }, { "epoch": 4.12, "learning_rate": 1.0015219809650143e-05, "loss": 0.0262, "step": 22446 }, { "epoch": 4.12, "learning_rate": 1.0014477380443111e-05, "loss": 0.0247, "step": 22447 }, { "epoch": 4.12, "learning_rate": 1.0013734951156278e-05, "loss": 0.0272, "step": 22448 }, { "epoch": 4.12, "learning_rate": 1.0012992521793736e-05, "loss": 0.0298, "step": 22449 }, { "epoch": 4.12, "learning_rate": 1.001225009235958e-05, "loss": 0.0266, "step": 22450 }, { "epoch": 4.12, "learning_rate": 1.0011507662857903e-05, "loss": 0.0263, "step": 22451 }, { "epoch": 4.12, "learning_rate": 1.0010765233292794e-05, "loss": 0.028, "step": 22452 }, { "epoch": 4.12, "learning_rate": 1.0010022803668344e-05, "loss": 0.0275, "step": 22453 }, { "epoch": 4.12, "learning_rate": 1.0009280373988654e-05, "loss": 0.0272, "step": 22454 }, { "epoch": 4.12, "learning_rate": 1.0008537944257805e-05, "loss": 0.024, "step": 22455 }, { "epoch": 4.12, "learning_rate": 1.0007795514479898e-05, "loss": 0.0273, "step": 22456 }, { "epoch": 4.12, "learning_rate": 1.0007053084659021e-05, "loss": 0.024, "step": 22457 }, { "epoch": 4.12, "learning_rate": 1.0006310654799267e-05, "loss": 0.0255, "step": 22458 }, { "epoch": 4.12, "learning_rate": 1.0005568224904729e-05, "loss": 0.0257, "step": 22459 }, { "epoch": 4.12, "learning_rate": 1.0004825794979498e-05, "loss": 0.0278, "step": 22460 }, { "epoch": 4.12, "learning_rate": 1.0004083365027666e-05, "loss": 0.0287, "step": 22461 }, { "epoch": 4.12, "learning_rate": 1.0003340935053329e-05, "loss": 0.0233, "step": 22462 }, { "epoch": 4.12, "learning_rate": 1.0002598505060577e-05, "loss": 0.026, "step": 22463 }, { "epoch": 4.12, "learning_rate": 1.00018560750535e-05, "loss": 0.0257, "step": 22464 }, { "epoch": 4.12, "learning_rate": 1.0001113645036193e-05, "loss": 0.0268, "step": 22465 }, { "epoch": 4.12, "learning_rate": 1.0000371215012748e-05, "loss": 0.0297, "step": 22466 }, { "epoch": 4.12, "learning_rate": 9.999628784987253e-06, "loss": 0.0217, "step": 22467 }, { "epoch": 4.12, "learning_rate": 9.99888635496381e-06, "loss": 0.0288, "step": 22468 }, { "epoch": 4.12, "learning_rate": 9.998143924946505e-06, "loss": 0.0231, "step": 22469 }, { "epoch": 4.12, "learning_rate": 9.997401494939426e-06, "loss": 0.0255, "step": 22470 }, { "epoch": 4.12, "learning_rate": 9.996659064946673e-06, "loss": 0.0265, "step": 22471 }, { "epoch": 4.12, "learning_rate": 9.995916634972334e-06, "loss": 0.0265, "step": 22472 }, { "epoch": 4.12, "learning_rate": 9.995174205020504e-06, "loss": 0.0244, "step": 22473 }, { "epoch": 4.12, "learning_rate": 9.994431775095273e-06, "loss": 0.0308, "step": 22474 }, { "epoch": 4.12, "learning_rate": 9.993689345200738e-06, "loss": 0.0259, "step": 22475 }, { "epoch": 4.12, "learning_rate": 9.99294691534098e-06, "loss": 0.0247, "step": 22476 }, { "epoch": 4.12, "learning_rate": 9.992204485520105e-06, "loss": 0.0293, "step": 22477 }, { "epoch": 4.12, "learning_rate": 9.991462055742197e-06, "loss": 0.0265, "step": 22478 }, { "epoch": 4.12, "learning_rate": 9.99071962601135e-06, "loss": 0.0227, "step": 22479 }, { "epoch": 4.12, "learning_rate": 9.989977196331656e-06, "loss": 0.0202, "step": 22480 }, { "epoch": 4.12, "learning_rate": 9.989234766707209e-06, "loss": 0.0274, "step": 22481 }, { "epoch": 4.12, "learning_rate": 9.988492337142099e-06, "loss": 0.0291, "step": 22482 }, { "epoch": 4.12, "learning_rate": 9.987749907640421e-06, "loss": 0.0362, "step": 22483 }, { "epoch": 4.12, "learning_rate": 9.987007478206269e-06, "loss": 0.0275, "step": 22484 }, { "epoch": 4.12, "learning_rate": 9.986265048843723e-06, "loss": 0.0216, "step": 22485 }, { "epoch": 4.12, "learning_rate": 9.985522619556894e-06, "loss": 0.0309, "step": 22486 }, { "epoch": 4.12, "learning_rate": 9.98478019034986e-06, "loss": 0.0297, "step": 22487 }, { "epoch": 4.12, "learning_rate": 9.984037761226718e-06, "loss": 0.0271, "step": 22488 }, { "epoch": 4.12, "learning_rate": 9.98329533219156e-06, "loss": 0.0274, "step": 22489 }, { "epoch": 4.12, "learning_rate": 9.982552903248478e-06, "loss": 0.0253, "step": 22490 }, { "epoch": 4.12, "learning_rate": 9.981810474401567e-06, "loss": 0.023, "step": 22491 }, { "epoch": 4.12, "learning_rate": 9.981068045654916e-06, "loss": 0.0286, "step": 22492 }, { "epoch": 4.12, "learning_rate": 9.980325617012616e-06, "loss": 0.032, "step": 22493 }, { "epoch": 4.13, "learning_rate": 9.979583188478765e-06, "loss": 0.0229, "step": 22494 }, { "epoch": 4.13, "learning_rate": 9.978840760057454e-06, "loss": 0.0292, "step": 22495 }, { "epoch": 4.13, "learning_rate": 9.97809833175277e-06, "loss": 0.0227, "step": 22496 }, { "epoch": 4.13, "learning_rate": 9.977355903568809e-06, "loss": 0.0213, "step": 22497 }, { "epoch": 4.13, "learning_rate": 9.97661347550966e-06, "loss": 0.0235, "step": 22498 }, { "epoch": 4.13, "learning_rate": 9.97587104757942e-06, "loss": 0.0321, "step": 22499 }, { "epoch": 4.13, "learning_rate": 9.975128619782182e-06, "loss": 0.0202, "step": 22500 }, { "epoch": 4.13, "learning_rate": 9.974386192122035e-06, "loss": 0.027, "step": 22501 }, { "epoch": 4.13, "learning_rate": 9.973643764603067e-06, "loss": 0.0251, "step": 22502 }, { "epoch": 4.13, "learning_rate": 9.97290133722938e-06, "loss": 0.0248, "step": 22503 }, { "epoch": 4.13, "learning_rate": 9.972158910005062e-06, "loss": 0.0271, "step": 22504 }, { "epoch": 4.13, "learning_rate": 9.971416482934203e-06, "loss": 0.0261, "step": 22505 }, { "epoch": 4.13, "learning_rate": 9.970674056020897e-06, "loss": 0.0257, "step": 22506 }, { "epoch": 4.13, "learning_rate": 9.969931629269234e-06, "loss": 0.0223, "step": 22507 }, { "epoch": 4.13, "learning_rate": 9.969189202683311e-06, "loss": 0.0236, "step": 22508 }, { "epoch": 4.13, "learning_rate": 9.968446776267217e-06, "loss": 0.0301, "step": 22509 }, { "epoch": 4.13, "learning_rate": 9.96770435002505e-06, "loss": 0.0247, "step": 22510 }, { "epoch": 4.13, "learning_rate": 9.966961923960889e-06, "loss": 0.0292, "step": 22511 }, { "epoch": 4.13, "learning_rate": 9.966219498078839e-06, "loss": 0.0212, "step": 22512 }, { "epoch": 4.13, "learning_rate": 9.965477072382989e-06, "loss": 0.035, "step": 22513 }, { "epoch": 4.13, "learning_rate": 9.96473464687743e-06, "loss": 0.0232, "step": 22514 }, { "epoch": 4.13, "learning_rate": 9.963992221566253e-06, "loss": 0.0299, "step": 22515 }, { "epoch": 4.13, "learning_rate": 9.963249796453551e-06, "loss": 0.0308, "step": 22516 }, { "epoch": 4.13, "learning_rate": 9.962507371543417e-06, "loss": 0.0196, "step": 22517 }, { "epoch": 4.13, "learning_rate": 9.961764946839944e-06, "loss": 0.0248, "step": 22518 }, { "epoch": 4.13, "learning_rate": 9.961022522347226e-06, "loss": 0.0248, "step": 22519 }, { "epoch": 4.13, "learning_rate": 9.960280098069346e-06, "loss": 0.0248, "step": 22520 }, { "epoch": 4.13, "learning_rate": 9.95953767401041e-06, "loss": 0.023, "step": 22521 }, { "epoch": 4.13, "learning_rate": 9.958795250174503e-06, "loss": 0.0309, "step": 22522 }, { "epoch": 4.13, "learning_rate": 9.958052826565715e-06, "loss": 0.0245, "step": 22523 }, { "epoch": 4.13, "learning_rate": 9.95731040318814e-06, "loss": 0.0294, "step": 22524 }, { "epoch": 4.13, "learning_rate": 9.956567980045872e-06, "loss": 0.032, "step": 22525 }, { "epoch": 4.13, "learning_rate": 9.955825557143003e-06, "loss": 0.0311, "step": 22526 }, { "epoch": 4.13, "learning_rate": 9.955083134483623e-06, "loss": 0.0322, "step": 22527 }, { "epoch": 4.13, "learning_rate": 9.95434071207183e-06, "loss": 0.0287, "step": 22528 }, { "epoch": 4.13, "learning_rate": 9.953598289911706e-06, "loss": 0.0297, "step": 22529 }, { "epoch": 4.13, "learning_rate": 9.952855868007351e-06, "loss": 0.025, "step": 22530 }, { "epoch": 4.13, "learning_rate": 9.952113446362861e-06, "loss": 0.0278, "step": 22531 }, { "epoch": 4.13, "learning_rate": 9.951371024982319e-06, "loss": 0.0261, "step": 22532 }, { "epoch": 4.13, "learning_rate": 9.950628603869818e-06, "loss": 0.0278, "step": 22533 }, { "epoch": 4.13, "learning_rate": 9.949886183029457e-06, "loss": 0.0272, "step": 22534 }, { "epoch": 4.13, "learning_rate": 9.949143762465324e-06, "loss": 0.0257, "step": 22535 }, { "epoch": 4.13, "learning_rate": 9.948401342181511e-06, "loss": 0.025, "step": 22536 }, { "epoch": 4.13, "learning_rate": 9.947658922182113e-06, "loss": 0.0274, "step": 22537 }, { "epoch": 4.13, "learning_rate": 9.946916502471215e-06, "loss": 0.0251, "step": 22538 }, { "epoch": 4.13, "learning_rate": 9.946174083052919e-06, "loss": 0.0311, "step": 22539 }, { "epoch": 4.13, "learning_rate": 9.945431663931313e-06, "loss": 0.0315, "step": 22540 }, { "epoch": 4.13, "learning_rate": 9.94468924511049e-06, "loss": 0.0233, "step": 22541 }, { "epoch": 4.13, "learning_rate": 9.943946826594536e-06, "loss": 0.0293, "step": 22542 }, { "epoch": 4.13, "learning_rate": 9.943204408387551e-06, "loss": 0.0261, "step": 22543 }, { "epoch": 4.13, "learning_rate": 9.942461990493627e-06, "loss": 0.0293, "step": 22544 }, { "epoch": 4.13, "learning_rate": 9.94171957291685e-06, "loss": 0.0228, "step": 22545 }, { "epoch": 4.13, "learning_rate": 9.940977155661317e-06, "loss": 0.0285, "step": 22546 }, { "epoch": 4.13, "learning_rate": 9.94023473873112e-06, "loss": 0.0232, "step": 22547 }, { "epoch": 4.13, "learning_rate": 9.939492322130353e-06, "loss": 0.0221, "step": 22548 }, { "epoch": 4.14, "learning_rate": 9.938749905863105e-06, "loss": 0.0249, "step": 22549 }, { "epoch": 4.14, "learning_rate": 9.938007489933467e-06, "loss": 0.029, "step": 22550 }, { "epoch": 4.14, "learning_rate": 9.937265074345532e-06, "loss": 0.028, "step": 22551 }, { "epoch": 4.14, "learning_rate": 9.936522659103395e-06, "loss": 0.0284, "step": 22552 }, { "epoch": 4.14, "learning_rate": 9.935780244211146e-06, "loss": 0.0248, "step": 22553 }, { "epoch": 4.14, "learning_rate": 9.935037829672878e-06, "loss": 0.0232, "step": 22554 }, { "epoch": 4.14, "learning_rate": 9.934295415492682e-06, "loss": 0.0248, "step": 22555 }, { "epoch": 4.14, "learning_rate": 9.933553001674653e-06, "loss": 0.0268, "step": 22556 }, { "epoch": 4.14, "learning_rate": 9.932810588222884e-06, "loss": 0.0232, "step": 22557 }, { "epoch": 4.14, "learning_rate": 9.932068175141462e-06, "loss": 0.0307, "step": 22558 }, { "epoch": 4.14, "learning_rate": 9.93132576243448e-06, "loss": 0.0311, "step": 22559 }, { "epoch": 4.14, "learning_rate": 9.930583350106032e-06, "loss": 0.0256, "step": 22560 }, { "epoch": 4.14, "learning_rate": 9.929840938160213e-06, "loss": 0.03, "step": 22561 }, { "epoch": 4.14, "learning_rate": 9.92909852660111e-06, "loss": 0.0269, "step": 22562 }, { "epoch": 4.14, "learning_rate": 9.92835611543282e-06, "loss": 0.0222, "step": 22563 }, { "epoch": 4.14, "learning_rate": 9.92761370465943e-06, "loss": 0.0271, "step": 22564 }, { "epoch": 4.14, "learning_rate": 9.926871294285035e-06, "loss": 0.025, "step": 22565 }, { "epoch": 4.14, "learning_rate": 9.926128884313732e-06, "loss": 0.0254, "step": 22566 }, { "epoch": 4.14, "learning_rate": 9.925386474749604e-06, "loss": 0.0287, "step": 22567 }, { "epoch": 4.14, "learning_rate": 9.924644065596749e-06, "loss": 0.0261, "step": 22568 }, { "epoch": 4.14, "learning_rate": 9.923901656859256e-06, "loss": 0.0243, "step": 22569 }, { "epoch": 4.14, "learning_rate": 9.923159248541218e-06, "loss": 0.026, "step": 22570 }, { "epoch": 4.14, "learning_rate": 9.922416840646731e-06, "loss": 0.0232, "step": 22571 }, { "epoch": 4.14, "learning_rate": 9.921674433179884e-06, "loss": 0.0237, "step": 22572 }, { "epoch": 4.14, "learning_rate": 9.920932026144766e-06, "loss": 0.0197, "step": 22573 }, { "epoch": 4.14, "learning_rate": 9.920189619545475e-06, "loss": 0.0244, "step": 22574 }, { "epoch": 4.14, "learning_rate": 9.919447213386103e-06, "loss": 0.0211, "step": 22575 }, { "epoch": 4.14, "learning_rate": 9.918704807670736e-06, "loss": 0.0303, "step": 22576 }, { "epoch": 4.14, "learning_rate": 9.917962402403468e-06, "loss": 0.0292, "step": 22577 }, { "epoch": 4.14, "learning_rate": 9.9172199975884e-06, "loss": 0.0268, "step": 22578 }, { "epoch": 4.14, "learning_rate": 9.916477593229614e-06, "loss": 0.0368, "step": 22579 }, { "epoch": 4.14, "learning_rate": 9.915735189331205e-06, "loss": 0.0283, "step": 22580 }, { "epoch": 4.14, "learning_rate": 9.914992785897266e-06, "loss": 0.0247, "step": 22581 }, { "epoch": 4.14, "learning_rate": 9.914250382931887e-06, "loss": 0.0275, "step": 22582 }, { "epoch": 4.14, "learning_rate": 9.913507980439164e-06, "loss": 0.0259, "step": 22583 }, { "epoch": 4.14, "learning_rate": 9.912765578423189e-06, "loss": 0.0247, "step": 22584 }, { "epoch": 4.14, "learning_rate": 9.912023176888051e-06, "loss": 0.0312, "step": 22585 }, { "epoch": 4.14, "learning_rate": 9.911280775837837e-06, "loss": 0.027, "step": 22586 }, { "epoch": 4.14, "learning_rate": 9.910538375276653e-06, "loss": 0.0233, "step": 22587 }, { "epoch": 4.14, "learning_rate": 9.909795975208581e-06, "loss": 0.0299, "step": 22588 }, { "epoch": 4.14, "learning_rate": 9.909053575637717e-06, "loss": 0.0246, "step": 22589 }, { "epoch": 4.14, "learning_rate": 9.908311176568151e-06, "loss": 0.0252, "step": 22590 }, { "epoch": 4.14, "learning_rate": 9.907568778003972e-06, "loss": 0.0211, "step": 22591 }, { "epoch": 4.14, "learning_rate": 9.90682637994928e-06, "loss": 0.0258, "step": 22592 }, { "epoch": 4.14, "learning_rate": 9.906083982408167e-06, "loss": 0.0223, "step": 22593 }, { "epoch": 4.14, "learning_rate": 9.905341585384718e-06, "loss": 0.0275, "step": 22594 }, { "epoch": 4.14, "learning_rate": 9.904599188883025e-06, "loss": 0.0277, "step": 22595 }, { "epoch": 4.14, "learning_rate": 9.903856792907186e-06, "loss": 0.0267, "step": 22596 }, { "epoch": 4.14, "learning_rate": 9.903114397461291e-06, "loss": 0.0244, "step": 22597 }, { "epoch": 4.14, "learning_rate": 9.902372002549432e-06, "loss": 0.0248, "step": 22598 }, { "epoch": 4.14, "learning_rate": 9.901629608175698e-06, "loss": 0.0244, "step": 22599 }, { "epoch": 4.14, "learning_rate": 9.900887214344186e-06, "loss": 0.0225, "step": 22600 }, { "epoch": 4.14, "learning_rate": 9.900144821058988e-06, "loss": 0.0261, "step": 22601 }, { "epoch": 4.14, "learning_rate": 9.899402428324193e-06, "loss": 0.0237, "step": 22602 }, { "epoch": 4.15, "learning_rate": 9.898660036143893e-06, "loss": 0.0271, "step": 22603 }, { "epoch": 4.15, "learning_rate": 9.89791764452218e-06, "loss": 0.0255, "step": 22604 }, { "epoch": 4.15, "learning_rate": 9.89717525346315e-06, "loss": 0.025, "step": 22605 }, { "epoch": 4.15, "learning_rate": 9.896432862970892e-06, "loss": 0.0263, "step": 22606 }, { "epoch": 4.15, "learning_rate": 9.895690473049497e-06, "loss": 0.023, "step": 22607 }, { "epoch": 4.15, "learning_rate": 9.894948083703058e-06, "loss": 0.0275, "step": 22608 }, { "epoch": 4.15, "learning_rate": 9.894205694935669e-06, "loss": 0.0244, "step": 22609 }, { "epoch": 4.15, "learning_rate": 9.893463306751421e-06, "loss": 0.0243, "step": 22610 }, { "epoch": 4.15, "learning_rate": 9.892720919154409e-06, "loss": 0.0269, "step": 22611 }, { "epoch": 4.15, "learning_rate": 9.891978532148718e-06, "loss": 0.024, "step": 22612 }, { "epoch": 4.15, "learning_rate": 9.891236145738442e-06, "loss": 0.0289, "step": 22613 }, { "epoch": 4.15, "learning_rate": 9.890493759927676e-06, "loss": 0.0341, "step": 22614 }, { "epoch": 4.15, "learning_rate": 9.889751374720514e-06, "loss": 0.0254, "step": 22615 }, { "epoch": 4.15, "learning_rate": 9.889008990121042e-06, "loss": 0.0198, "step": 22616 }, { "epoch": 4.15, "learning_rate": 9.888266606133354e-06, "loss": 0.0281, "step": 22617 }, { "epoch": 4.15, "learning_rate": 9.887524222761546e-06, "loss": 0.0253, "step": 22618 }, { "epoch": 4.15, "learning_rate": 9.886781840009705e-06, "loss": 0.0297, "step": 22619 }, { "epoch": 4.15, "learning_rate": 9.88603945788193e-06, "loss": 0.0253, "step": 22620 }, { "epoch": 4.15, "learning_rate": 9.8852970763823e-06, "loss": 0.0271, "step": 22621 }, { "epoch": 4.15, "learning_rate": 9.884554695514922e-06, "loss": 0.0239, "step": 22622 }, { "epoch": 4.15, "learning_rate": 9.88381231528388e-06, "loss": 0.0219, "step": 22623 }, { "epoch": 4.15, "learning_rate": 9.883069935693267e-06, "loss": 0.0266, "step": 22624 }, { "epoch": 4.15, "learning_rate": 9.882327556747174e-06, "loss": 0.0263, "step": 22625 }, { "epoch": 4.15, "learning_rate": 9.881585178449694e-06, "loss": 0.0325, "step": 22626 }, { "epoch": 4.15, "learning_rate": 9.88084280080492e-06, "loss": 0.0272, "step": 22627 }, { "epoch": 4.15, "learning_rate": 9.880100423816942e-06, "loss": 0.0239, "step": 22628 }, { "epoch": 4.15, "learning_rate": 9.879358047489858e-06, "loss": 0.0207, "step": 22629 }, { "epoch": 4.15, "learning_rate": 9.878615671827746e-06, "loss": 0.0258, "step": 22630 }, { "epoch": 4.15, "learning_rate": 9.877873296834718e-06, "loss": 0.0334, "step": 22631 }, { "epoch": 4.15, "learning_rate": 9.87713092251485e-06, "loss": 0.0223, "step": 22632 }, { "epoch": 4.15, "learning_rate": 9.87638854887224e-06, "loss": 0.0269, "step": 22633 }, { "epoch": 4.15, "learning_rate": 9.875646175910979e-06, "loss": 0.027, "step": 22634 }, { "epoch": 4.15, "learning_rate": 9.874903803635156e-06, "loss": 0.0304, "step": 22635 }, { "epoch": 4.15, "learning_rate": 9.87416143204887e-06, "loss": 0.0265, "step": 22636 }, { "epoch": 4.15, "learning_rate": 9.873419061156209e-06, "loss": 0.027, "step": 22637 }, { "epoch": 4.15, "learning_rate": 9.872676690961266e-06, "loss": 0.022, "step": 22638 }, { "epoch": 4.15, "learning_rate": 9.871934321468127e-06, "loss": 0.0287, "step": 22639 }, { "epoch": 4.15, "learning_rate": 9.871191952680893e-06, "loss": 0.0252, "step": 22640 }, { "epoch": 4.15, "learning_rate": 9.870449584603653e-06, "loss": 0.025, "step": 22641 }, { "epoch": 4.15, "learning_rate": 9.869707217240497e-06, "loss": 0.0214, "step": 22642 }, { "epoch": 4.15, "learning_rate": 9.868964850595517e-06, "loss": 0.0357, "step": 22643 }, { "epoch": 4.15, "learning_rate": 9.868222484672805e-06, "loss": 0.0251, "step": 22644 }, { "epoch": 4.15, "learning_rate": 9.867480119476454e-06, "loss": 0.0269, "step": 22645 }, { "epoch": 4.15, "learning_rate": 9.866737755010557e-06, "loss": 0.0223, "step": 22646 }, { "epoch": 4.15, "learning_rate": 9.865995391279207e-06, "loss": 0.0274, "step": 22647 }, { "epoch": 4.15, "learning_rate": 9.865253028286487e-06, "loss": 0.0283, "step": 22648 }, { "epoch": 4.15, "learning_rate": 9.8645106660365e-06, "loss": 0.0239, "step": 22649 }, { "epoch": 4.15, "learning_rate": 9.863768304533335e-06, "loss": 0.029, "step": 22650 }, { "epoch": 4.15, "learning_rate": 9.86302594378108e-06, "loss": 0.0262, "step": 22651 }, { "epoch": 4.15, "learning_rate": 9.862283583783826e-06, "loss": 0.0269, "step": 22652 }, { "epoch": 4.15, "learning_rate": 9.86154122454567e-06, "loss": 0.0237, "step": 22653 }, { "epoch": 4.15, "learning_rate": 9.860798866070704e-06, "loss": 0.0286, "step": 22654 }, { "epoch": 4.15, "learning_rate": 9.860056508363019e-06, "loss": 0.0282, "step": 22655 }, { "epoch": 4.15, "learning_rate": 9.859314151426705e-06, "loss": 0.0254, "step": 22656 }, { "epoch": 4.15, "learning_rate": 9.858571795265849e-06, "loss": 0.0237, "step": 22657 }, { "epoch": 4.16, "learning_rate": 9.857829439884557e-06, "loss": 0.0283, "step": 22658 }, { "epoch": 4.16, "learning_rate": 9.857087085286908e-06, "loss": 0.0275, "step": 22659 }, { "epoch": 4.16, "learning_rate": 9.856344731476999e-06, "loss": 0.0245, "step": 22660 }, { "epoch": 4.16, "learning_rate": 9.85560237845892e-06, "loss": 0.0244, "step": 22661 }, { "epoch": 4.16, "learning_rate": 9.854860026236764e-06, "loss": 0.0245, "step": 22662 }, { "epoch": 4.16, "learning_rate": 9.854117674814624e-06, "loss": 0.0266, "step": 22663 }, { "epoch": 4.16, "learning_rate": 9.853375324196595e-06, "loss": 0.0237, "step": 22664 }, { "epoch": 4.16, "learning_rate": 9.85263297438676e-06, "loss": 0.0305, "step": 22665 }, { "epoch": 4.16, "learning_rate": 9.851890625389211e-06, "loss": 0.0245, "step": 22666 }, { "epoch": 4.16, "learning_rate": 9.851148277208053e-06, "loss": 0.0291, "step": 22667 }, { "epoch": 4.16, "learning_rate": 9.850405929847367e-06, "loss": 0.03, "step": 22668 }, { "epoch": 4.16, "learning_rate": 9.849663583311245e-06, "loss": 0.0197, "step": 22669 }, { "epoch": 4.16, "learning_rate": 9.84892123760378e-06, "loss": 0.0326, "step": 22670 }, { "epoch": 4.16, "learning_rate": 9.848178892729065e-06, "loss": 0.0247, "step": 22671 }, { "epoch": 4.16, "learning_rate": 9.847436548691194e-06, "loss": 0.0287, "step": 22672 }, { "epoch": 4.16, "learning_rate": 9.846694205494257e-06, "loss": 0.0307, "step": 22673 }, { "epoch": 4.16, "learning_rate": 9.84595186314234e-06, "loss": 0.0277, "step": 22674 }, { "epoch": 4.16, "learning_rate": 9.845209521639543e-06, "loss": 0.0246, "step": 22675 }, { "epoch": 4.16, "learning_rate": 9.844467180989957e-06, "loss": 0.0256, "step": 22676 }, { "epoch": 4.16, "learning_rate": 9.843724841197669e-06, "loss": 0.0242, "step": 22677 }, { "epoch": 4.16, "learning_rate": 9.842982502266773e-06, "loss": 0.0264, "step": 22678 }, { "epoch": 4.16, "learning_rate": 9.842240164201362e-06, "loss": 0.0276, "step": 22679 }, { "epoch": 4.16, "learning_rate": 9.841497827005526e-06, "loss": 0.0266, "step": 22680 }, { "epoch": 4.16, "learning_rate": 9.840755490683358e-06, "loss": 0.0275, "step": 22681 }, { "epoch": 4.16, "learning_rate": 9.840013155238952e-06, "loss": 0.0202, "step": 22682 }, { "epoch": 4.16, "learning_rate": 9.839270820676393e-06, "loss": 0.0261, "step": 22683 }, { "epoch": 4.16, "learning_rate": 9.838528486999778e-06, "loss": 0.0302, "step": 22684 }, { "epoch": 4.16, "learning_rate": 9.837786154213201e-06, "loss": 0.0226, "step": 22685 }, { "epoch": 4.16, "learning_rate": 9.83704382232075e-06, "loss": 0.0318, "step": 22686 }, { "epoch": 4.16, "learning_rate": 9.836301491326515e-06, "loss": 0.0305, "step": 22687 }, { "epoch": 4.16, "learning_rate": 9.835559161234587e-06, "loss": 0.0215, "step": 22688 }, { "epoch": 4.16, "learning_rate": 9.834816832049065e-06, "loss": 0.0242, "step": 22689 }, { "epoch": 4.16, "learning_rate": 9.834074503774036e-06, "loss": 0.0296, "step": 22690 }, { "epoch": 4.16, "learning_rate": 9.833332176413595e-06, "loss": 0.0262, "step": 22691 }, { "epoch": 4.16, "learning_rate": 9.832589849971824e-06, "loss": 0.0325, "step": 22692 }, { "epoch": 4.16, "learning_rate": 9.831847524452826e-06, "loss": 0.0242, "step": 22693 }, { "epoch": 4.16, "learning_rate": 9.83110519986069e-06, "loss": 0.0259, "step": 22694 }, { "epoch": 4.16, "learning_rate": 9.830362876199505e-06, "loss": 0.0245, "step": 22695 }, { "epoch": 4.16, "learning_rate": 9.829620553473364e-06, "loss": 0.0225, "step": 22696 }, { "epoch": 4.16, "learning_rate": 9.828878231686355e-06, "loss": 0.0262, "step": 22697 }, { "epoch": 4.16, "learning_rate": 9.828135910842576e-06, "loss": 0.023, "step": 22698 }, { "epoch": 4.16, "learning_rate": 9.827393590946116e-06, "loss": 0.0258, "step": 22699 }, { "epoch": 4.16, "learning_rate": 9.82665127200107e-06, "loss": 0.0234, "step": 22700 }, { "epoch": 4.16, "learning_rate": 9.825908954011517e-06, "loss": 0.0225, "step": 22701 }, { "epoch": 4.16, "learning_rate": 9.825166636981566e-06, "loss": 0.0271, "step": 22702 }, { "epoch": 4.16, "learning_rate": 9.824424320915297e-06, "loss": 0.0249, "step": 22703 }, { "epoch": 4.16, "learning_rate": 9.823682005816809e-06, "loss": 0.0239, "step": 22704 }, { "epoch": 4.16, "learning_rate": 9.822939691690185e-06, "loss": 0.0245, "step": 22705 }, { "epoch": 4.16, "learning_rate": 9.822197378539523e-06, "loss": 0.0277, "step": 22706 }, { "epoch": 4.16, "learning_rate": 9.821455066368914e-06, "loss": 0.0299, "step": 22707 }, { "epoch": 4.16, "learning_rate": 9.820712755182448e-06, "loss": 0.0253, "step": 22708 }, { "epoch": 4.16, "learning_rate": 9.819970444984221e-06, "loss": 0.0218, "step": 22709 }, { "epoch": 4.16, "learning_rate": 9.819228135778315e-06, "loss": 0.0279, "step": 22710 }, { "epoch": 4.16, "learning_rate": 9.818485827568833e-06, "loss": 0.0211, "step": 22711 }, { "epoch": 4.17, "learning_rate": 9.81774352035986e-06, "loss": 0.0242, "step": 22712 }, { "epoch": 4.17, "learning_rate": 9.817001214155488e-06, "loss": 0.0253, "step": 22713 }, { "epoch": 4.17, "learning_rate": 9.816258908959809e-06, "loss": 0.0244, "step": 22714 }, { "epoch": 4.17, "learning_rate": 9.815516604776916e-06, "loss": 0.0227, "step": 22715 }, { "epoch": 4.17, "learning_rate": 9.8147743016109e-06, "loss": 0.0224, "step": 22716 }, { "epoch": 4.17, "learning_rate": 9.814031999465853e-06, "loss": 0.0329, "step": 22717 }, { "epoch": 4.17, "learning_rate": 9.813289698345868e-06, "loss": 0.025, "step": 22718 }, { "epoch": 4.17, "learning_rate": 9.812547398255027e-06, "loss": 0.0261, "step": 22719 }, { "epoch": 4.17, "learning_rate": 9.811805099197438e-06, "loss": 0.0281, "step": 22720 }, { "epoch": 4.17, "learning_rate": 9.81106280117718e-06, "loss": 0.0267, "step": 22721 }, { "epoch": 4.17, "learning_rate": 9.810320504198349e-06, "loss": 0.0264, "step": 22722 }, { "epoch": 4.17, "learning_rate": 9.809578208265032e-06, "loss": 0.0243, "step": 22723 }, { "epoch": 4.17, "learning_rate": 9.808835913381327e-06, "loss": 0.0248, "step": 22724 }, { "epoch": 4.17, "learning_rate": 9.808093619551324e-06, "loss": 0.0233, "step": 22725 }, { "epoch": 4.17, "learning_rate": 9.807351326779114e-06, "loss": 0.0266, "step": 22726 }, { "epoch": 4.17, "learning_rate": 9.806609035068783e-06, "loss": 0.0249, "step": 22727 }, { "epoch": 4.17, "learning_rate": 9.805866744424432e-06, "loss": 0.0264, "step": 22728 }, { "epoch": 4.17, "learning_rate": 9.80512445485015e-06, "loss": 0.0245, "step": 22729 }, { "epoch": 4.17, "learning_rate": 9.804382166350023e-06, "loss": 0.0247, "step": 22730 }, { "epoch": 4.17, "learning_rate": 9.803639878928148e-06, "loss": 0.0297, "step": 22731 }, { "epoch": 4.17, "learning_rate": 9.802897592588611e-06, "loss": 0.0217, "step": 22732 }, { "epoch": 4.17, "learning_rate": 9.80215530733551e-06, "loss": 0.0278, "step": 22733 }, { "epoch": 4.17, "learning_rate": 9.801413023172934e-06, "loss": 0.0297, "step": 22734 }, { "epoch": 4.17, "learning_rate": 9.800670740104972e-06, "loss": 0.0215, "step": 22735 }, { "epoch": 4.17, "learning_rate": 9.799928458135717e-06, "loss": 0.0297, "step": 22736 }, { "epoch": 4.17, "learning_rate": 9.799186177269264e-06, "loss": 0.0279, "step": 22737 }, { "epoch": 4.17, "learning_rate": 9.798443897509702e-06, "loss": 0.0261, "step": 22738 }, { "epoch": 4.17, "learning_rate": 9.797701618861122e-06, "loss": 0.0228, "step": 22739 }, { "epoch": 4.17, "learning_rate": 9.796959341327614e-06, "loss": 0.0282, "step": 22740 }, { "epoch": 4.17, "learning_rate": 9.796217064913269e-06, "loss": 0.0245, "step": 22741 }, { "epoch": 4.17, "learning_rate": 9.79547478962218e-06, "loss": 0.0356, "step": 22742 }, { "epoch": 4.17, "learning_rate": 9.794732515458441e-06, "loss": 0.0255, "step": 22743 }, { "epoch": 4.17, "learning_rate": 9.793990242426143e-06, "loss": 0.0225, "step": 22744 }, { "epoch": 4.17, "learning_rate": 9.79324797052937e-06, "loss": 0.0266, "step": 22745 }, { "epoch": 4.17, "learning_rate": 9.792505699772224e-06, "loss": 0.0285, "step": 22746 }, { "epoch": 4.17, "learning_rate": 9.791763430158791e-06, "loss": 0.0187, "step": 22747 }, { "epoch": 4.17, "learning_rate": 9.791021161693163e-06, "loss": 0.0237, "step": 22748 }, { "epoch": 4.17, "learning_rate": 9.790278894379428e-06, "loss": 0.0303, "step": 22749 }, { "epoch": 4.17, "learning_rate": 9.789536628221682e-06, "loss": 0.0333, "step": 22750 }, { "epoch": 4.17, "learning_rate": 9.788794363224015e-06, "loss": 0.0263, "step": 22751 }, { "epoch": 4.17, "learning_rate": 9.78805209939052e-06, "loss": 0.018, "step": 22752 }, { "epoch": 4.17, "learning_rate": 9.787309836725286e-06, "loss": 0.0253, "step": 22753 }, { "epoch": 4.17, "learning_rate": 9.786567575232404e-06, "loss": 0.0228, "step": 22754 }, { "epoch": 4.17, "learning_rate": 9.785825314915968e-06, "loss": 0.0247, "step": 22755 }, { "epoch": 4.17, "learning_rate": 9.78508305578007e-06, "loss": 0.0255, "step": 22756 }, { "epoch": 4.17, "learning_rate": 9.784340797828796e-06, "loss": 0.028, "step": 22757 }, { "epoch": 4.17, "learning_rate": 9.78359854106624e-06, "loss": 0.0221, "step": 22758 }, { "epoch": 4.17, "learning_rate": 9.782856285496494e-06, "loss": 0.0282, "step": 22759 }, { "epoch": 4.17, "learning_rate": 9.782114031123652e-06, "loss": 0.0224, "step": 22760 }, { "epoch": 4.17, "learning_rate": 9.781371777951801e-06, "loss": 0.0269, "step": 22761 }, { "epoch": 4.17, "learning_rate": 9.780629525985034e-06, "loss": 0.0289, "step": 22762 }, { "epoch": 4.17, "learning_rate": 9.77988727522744e-06, "loss": 0.0319, "step": 22763 }, { "epoch": 4.17, "learning_rate": 9.779145025683114e-06, "loss": 0.0251, "step": 22764 }, { "epoch": 4.17, "learning_rate": 9.778402777356149e-06, "loss": 0.0289, "step": 22765 }, { "epoch": 4.17, "learning_rate": 9.777660530250631e-06, "loss": 0.0274, "step": 22766 }, { "epoch": 4.18, "learning_rate": 9.776918284370652e-06, "loss": 0.0284, "step": 22767 }, { "epoch": 4.18, "learning_rate": 9.776176039720305e-06, "loss": 0.0336, "step": 22768 }, { "epoch": 4.18, "learning_rate": 9.775433796303681e-06, "loss": 0.0323, "step": 22769 }, { "epoch": 4.18, "learning_rate": 9.774691554124872e-06, "loss": 0.0226, "step": 22770 }, { "epoch": 4.18, "learning_rate": 9.773949313187971e-06, "loss": 0.0226, "step": 22771 }, { "epoch": 4.18, "learning_rate": 9.773207073497058e-06, "loss": 0.0244, "step": 22772 }, { "epoch": 4.18, "learning_rate": 9.77246483505624e-06, "loss": 0.024, "step": 22773 }, { "epoch": 4.18, "learning_rate": 9.771722597869602e-06, "loss": 0.0257, "step": 22774 }, { "epoch": 4.18, "learning_rate": 9.770980361941233e-06, "loss": 0.0322, "step": 22775 }, { "epoch": 4.18, "learning_rate": 9.770238127275222e-06, "loss": 0.0262, "step": 22776 }, { "epoch": 4.18, "learning_rate": 9.769495893875667e-06, "loss": 0.0297, "step": 22777 }, { "epoch": 4.18, "learning_rate": 9.768753661746657e-06, "loss": 0.0244, "step": 22778 }, { "epoch": 4.18, "learning_rate": 9.76801143089228e-06, "loss": 0.0249, "step": 22779 }, { "epoch": 4.18, "learning_rate": 9.767269201316629e-06, "loss": 0.027, "step": 22780 }, { "epoch": 4.18, "learning_rate": 9.766526973023797e-06, "loss": 0.0239, "step": 22781 }, { "epoch": 4.18, "learning_rate": 9.765784746017873e-06, "loss": 0.0288, "step": 22782 }, { "epoch": 4.18, "learning_rate": 9.765042520302951e-06, "loss": 0.0266, "step": 22783 }, { "epoch": 4.18, "learning_rate": 9.76430029588312e-06, "loss": 0.0218, "step": 22784 }, { "epoch": 4.18, "learning_rate": 9.763558072762467e-06, "loss": 0.0266, "step": 22785 }, { "epoch": 4.18, "learning_rate": 9.762815850945093e-06, "loss": 0.027, "step": 22786 }, { "epoch": 4.18, "learning_rate": 9.76207363043508e-06, "loss": 0.026, "step": 22787 }, { "epoch": 4.18, "learning_rate": 9.761331411236524e-06, "loss": 0.0293, "step": 22788 }, { "epoch": 4.18, "learning_rate": 9.760589193353513e-06, "loss": 0.0279, "step": 22789 }, { "epoch": 4.18, "learning_rate": 9.759846976790141e-06, "loss": 0.0238, "step": 22790 }, { "epoch": 4.18, "learning_rate": 9.7591047615505e-06, "loss": 0.0225, "step": 22791 }, { "epoch": 4.18, "learning_rate": 9.75836254763868e-06, "loss": 0.031, "step": 22792 }, { "epoch": 4.18, "learning_rate": 9.757620335058769e-06, "loss": 0.0231, "step": 22793 }, { "epoch": 4.18, "learning_rate": 9.75687812381486e-06, "loss": 0.0285, "step": 22794 }, { "epoch": 4.18, "learning_rate": 9.756135913911044e-06, "loss": 0.0259, "step": 22795 }, { "epoch": 4.18, "learning_rate": 9.755393705351415e-06, "loss": 0.0281, "step": 22796 }, { "epoch": 4.18, "learning_rate": 9.75465149814006e-06, "loss": 0.0275, "step": 22797 }, { "epoch": 4.18, "learning_rate": 9.75390929228107e-06, "loss": 0.0257, "step": 22798 }, { "epoch": 4.18, "learning_rate": 9.75316708777854e-06, "loss": 0.0243, "step": 22799 }, { "epoch": 4.18, "learning_rate": 9.752424884636559e-06, "loss": 0.024, "step": 22800 }, { "epoch": 4.18, "learning_rate": 9.75168268285922e-06, "loss": 0.0279, "step": 22801 }, { "epoch": 4.18, "learning_rate": 9.750940482450603e-06, "loss": 0.0251, "step": 22802 }, { "epoch": 4.18, "learning_rate": 9.750198283414818e-06, "loss": 0.0287, "step": 22803 }, { "epoch": 4.18, "learning_rate": 9.749456085755942e-06, "loss": 0.0265, "step": 22804 }, { "epoch": 4.18, "learning_rate": 9.748713889478069e-06, "loss": 0.0273, "step": 22805 }, { "epoch": 4.18, "learning_rate": 9.747971694585294e-06, "loss": 0.0248, "step": 22806 }, { "epoch": 4.18, "learning_rate": 9.7472295010817e-06, "loss": 0.0217, "step": 22807 }, { "epoch": 4.18, "learning_rate": 9.746487308971387e-06, "loss": 0.0237, "step": 22808 }, { "epoch": 4.18, "learning_rate": 9.74574511825844e-06, "loss": 0.0251, "step": 22809 }, { "epoch": 4.18, "learning_rate": 9.745002928946955e-06, "loss": 0.0233, "step": 22810 }, { "epoch": 4.18, "learning_rate": 9.744260741041012e-06, "loss": 0.0306, "step": 22811 }, { "epoch": 4.18, "learning_rate": 9.743518554544719e-06, "loss": 0.0249, "step": 22812 }, { "epoch": 4.18, "learning_rate": 9.742776369462153e-06, "loss": 0.026, "step": 22813 }, { "epoch": 4.18, "learning_rate": 9.742034185797411e-06, "loss": 0.0202, "step": 22814 }, { "epoch": 4.18, "learning_rate": 9.741292003554581e-06, "loss": 0.0243, "step": 22815 }, { "epoch": 4.18, "learning_rate": 9.740549822737755e-06, "loss": 0.0279, "step": 22816 }, { "epoch": 4.18, "learning_rate": 9.739807643351028e-06, "loss": 0.0251, "step": 22817 }, { "epoch": 4.18, "learning_rate": 9.739065465398488e-06, "loss": 0.0256, "step": 22818 }, { "epoch": 4.18, "learning_rate": 9.738323288884223e-06, "loss": 0.024, "step": 22819 }, { "epoch": 4.18, "learning_rate": 9.737581113812322e-06, "loss": 0.0278, "step": 22820 }, { "epoch": 4.19, "learning_rate": 9.736838940186885e-06, "loss": 0.0307, "step": 22821 }, { "epoch": 4.19, "learning_rate": 9.736096768011997e-06, "loss": 0.0243, "step": 22822 }, { "epoch": 4.19, "learning_rate": 9.73535459729175e-06, "loss": 0.0236, "step": 22823 }, { "epoch": 4.19, "learning_rate": 9.734612428030234e-06, "loss": 0.0243, "step": 22824 }, { "epoch": 4.19, "learning_rate": 9.73387026023154e-06, "loss": 0.0301, "step": 22825 }, { "epoch": 4.19, "learning_rate": 9.73312809389976e-06, "loss": 0.0223, "step": 22826 }, { "epoch": 4.19, "learning_rate": 9.732385929038987e-06, "loss": 0.0275, "step": 22827 }, { "epoch": 4.19, "learning_rate": 9.731643765653308e-06, "loss": 0.019, "step": 22828 }, { "epoch": 4.19, "learning_rate": 9.730901603746807e-06, "loss": 0.0256, "step": 22829 }, { "epoch": 4.19, "learning_rate": 9.730159443323592e-06, "loss": 0.0253, "step": 22830 }, { "epoch": 4.19, "learning_rate": 9.729417284387743e-06, "loss": 0.0214, "step": 22831 }, { "epoch": 4.19, "learning_rate": 9.72867512694335e-06, "loss": 0.0319, "step": 22832 }, { "epoch": 4.19, "learning_rate": 9.727932970994504e-06, "loss": 0.0282, "step": 22833 }, { "epoch": 4.19, "learning_rate": 9.7271908165453e-06, "loss": 0.0242, "step": 22834 }, { "epoch": 4.19, "learning_rate": 9.72644866359983e-06, "loss": 0.0192, "step": 22835 }, { "epoch": 4.19, "learning_rate": 9.72570651216218e-06, "loss": 0.0253, "step": 22836 }, { "epoch": 4.19, "learning_rate": 9.724964362236441e-06, "loss": 0.022, "step": 22837 }, { "epoch": 4.19, "learning_rate": 9.724222213826699e-06, "loss": 0.0258, "step": 22838 }, { "epoch": 4.19, "learning_rate": 9.72348006693706e-06, "loss": 0.0274, "step": 22839 }, { "epoch": 4.19, "learning_rate": 9.722737921571601e-06, "loss": 0.0224, "step": 22840 }, { "epoch": 4.19, "learning_rate": 9.721995777734418e-06, "loss": 0.0303, "step": 22841 }, { "epoch": 4.19, "learning_rate": 9.721253635429598e-06, "loss": 0.0319, "step": 22842 }, { "epoch": 4.19, "learning_rate": 9.720511494661237e-06, "loss": 0.0256, "step": 22843 }, { "epoch": 4.19, "learning_rate": 9.719769355433424e-06, "loss": 0.0221, "step": 22844 }, { "epoch": 4.19, "learning_rate": 9.71902721775025e-06, "loss": 0.0288, "step": 22845 }, { "epoch": 4.19, "learning_rate": 9.718285081615802e-06, "loss": 0.0272, "step": 22846 }, { "epoch": 4.19, "learning_rate": 9.717542947034172e-06, "loss": 0.0325, "step": 22847 }, { "epoch": 4.19, "learning_rate": 9.716800814009455e-06, "loss": 0.0202, "step": 22848 }, { "epoch": 4.19, "learning_rate": 9.716058682545737e-06, "loss": 0.0289, "step": 22849 }, { "epoch": 4.19, "learning_rate": 9.715316552647111e-06, "loss": 0.0263, "step": 22850 }, { "epoch": 4.19, "learning_rate": 9.714574424317664e-06, "loss": 0.0289, "step": 22851 }, { "epoch": 4.19, "learning_rate": 9.713832297561491e-06, "loss": 0.0253, "step": 22852 }, { "epoch": 4.19, "learning_rate": 9.713090172382683e-06, "loss": 0.0225, "step": 22853 }, { "epoch": 4.19, "learning_rate": 9.71234804878533e-06, "loss": 0.0308, "step": 22854 }, { "epoch": 4.19, "learning_rate": 9.711605926773514e-06, "loss": 0.0272, "step": 22855 }, { "epoch": 4.19, "learning_rate": 9.710863806351338e-06, "loss": 0.0271, "step": 22856 }, { "epoch": 4.19, "learning_rate": 9.71012168752289e-06, "loss": 0.0236, "step": 22857 }, { "epoch": 4.19, "learning_rate": 9.709379570292255e-06, "loss": 0.021, "step": 22858 }, { "epoch": 4.19, "learning_rate": 9.708637454663527e-06, "loss": 0.0323, "step": 22859 }, { "epoch": 4.19, "learning_rate": 9.707895340640793e-06, "loss": 0.0293, "step": 22860 }, { "epoch": 4.19, "learning_rate": 9.707153228228151e-06, "loss": 0.0231, "step": 22861 }, { "epoch": 4.19, "learning_rate": 9.706411117429686e-06, "loss": 0.0187, "step": 22862 }, { "epoch": 4.19, "learning_rate": 9.705669008249492e-06, "loss": 0.0262, "step": 22863 }, { "epoch": 4.19, "learning_rate": 9.704926900691651e-06, "loss": 0.0304, "step": 22864 }, { "epoch": 4.19, "learning_rate": 9.704184794760266e-06, "loss": 0.0271, "step": 22865 }, { "epoch": 4.19, "learning_rate": 9.70344269045942e-06, "loss": 0.0259, "step": 22866 }, { "epoch": 4.19, "learning_rate": 9.702700587793204e-06, "loss": 0.0229, "step": 22867 }, { "epoch": 4.19, "learning_rate": 9.70195848676571e-06, "loss": 0.022, "step": 22868 }, { "epoch": 4.19, "learning_rate": 9.701216387381025e-06, "loss": 0.026, "step": 22869 }, { "epoch": 4.19, "learning_rate": 9.700474289643245e-06, "loss": 0.028, "step": 22870 }, { "epoch": 4.19, "learning_rate": 9.699732193556456e-06, "loss": 0.0256, "step": 22871 }, { "epoch": 4.19, "learning_rate": 9.698990099124755e-06, "loss": 0.0294, "step": 22872 }, { "epoch": 4.19, "learning_rate": 9.69824800635222e-06, "loss": 0.0214, "step": 22873 }, { "epoch": 4.19, "learning_rate": 9.697505915242956e-06, "loss": 0.0221, "step": 22874 }, { "epoch": 4.19, "learning_rate": 9.696763825801044e-06, "loss": 0.0257, "step": 22875 }, { "epoch": 4.2, "learning_rate": 9.696021738030575e-06, "loss": 0.0233, "step": 22876 }, { "epoch": 4.2, "learning_rate": 9.695279651935641e-06, "loss": 0.03, "step": 22877 }, { "epoch": 4.2, "learning_rate": 9.694537567520335e-06, "loss": 0.026, "step": 22878 }, { "epoch": 4.2, "learning_rate": 9.693795484788745e-06, "loss": 0.0253, "step": 22879 }, { "epoch": 4.2, "learning_rate": 9.693053403744961e-06, "loss": 0.0287, "step": 22880 }, { "epoch": 4.2, "learning_rate": 9.692311324393075e-06, "loss": 0.0237, "step": 22881 }, { "epoch": 4.2, "learning_rate": 9.691569246737171e-06, "loss": 0.0299, "step": 22882 }, { "epoch": 4.2, "learning_rate": 9.690827170781352e-06, "loss": 0.0251, "step": 22883 }, { "epoch": 4.2, "learning_rate": 9.690085096529697e-06, "loss": 0.0225, "step": 22884 }, { "epoch": 4.2, "learning_rate": 9.689343023986303e-06, "loss": 0.0255, "step": 22885 }, { "epoch": 4.2, "learning_rate": 9.688600953155251e-06, "loss": 0.0244, "step": 22886 }, { "epoch": 4.2, "learning_rate": 9.687858884040643e-06, "loss": 0.0302, "step": 22887 }, { "epoch": 4.2, "learning_rate": 9.687116816646563e-06, "loss": 0.0262, "step": 22888 }, { "epoch": 4.2, "learning_rate": 9.686374750977104e-06, "loss": 0.0286, "step": 22889 }, { "epoch": 4.2, "learning_rate": 9.685632687036355e-06, "loss": 0.0241, "step": 22890 }, { "epoch": 4.2, "learning_rate": 9.684890624828401e-06, "loss": 0.0303, "step": 22891 }, { "epoch": 4.2, "learning_rate": 9.684148564357342e-06, "loss": 0.0212, "step": 22892 }, { "epoch": 4.2, "learning_rate": 9.683406505627262e-06, "loss": 0.021, "step": 22893 }, { "epoch": 4.2, "learning_rate": 9.682664448642252e-06, "loss": 0.0259, "step": 22894 }, { "epoch": 4.2, "learning_rate": 9.681922393406404e-06, "loss": 0.0279, "step": 22895 }, { "epoch": 4.2, "learning_rate": 9.681180339923805e-06, "loss": 0.0224, "step": 22896 }, { "epoch": 4.2, "learning_rate": 9.680438288198551e-06, "loss": 0.0297, "step": 22897 }, { "epoch": 4.2, "learning_rate": 9.679696238234727e-06, "loss": 0.028, "step": 22898 }, { "epoch": 4.2, "learning_rate": 9.678954190036428e-06, "loss": 0.0299, "step": 22899 }, { "epoch": 4.2, "learning_rate": 9.678212143607734e-06, "loss": 0.0247, "step": 22900 }, { "epoch": 4.2, "learning_rate": 9.677470098952748e-06, "loss": 0.0259, "step": 22901 }, { "epoch": 4.2, "learning_rate": 9.676728056075553e-06, "loss": 0.023, "step": 22902 }, { "epoch": 4.2, "learning_rate": 9.67598601498024e-06, "loss": 0.0303, "step": 22903 }, { "epoch": 4.2, "learning_rate": 9.675243975670898e-06, "loss": 0.0213, "step": 22904 }, { "epoch": 4.2, "learning_rate": 9.67450193815162e-06, "loss": 0.0238, "step": 22905 }, { "epoch": 4.2, "learning_rate": 9.673759902426496e-06, "loss": 0.0227, "step": 22906 }, { "epoch": 4.2, "learning_rate": 9.673017868499615e-06, "loss": 0.0272, "step": 22907 }, { "epoch": 4.2, "learning_rate": 9.672275836375064e-06, "loss": 0.0268, "step": 22908 }, { "epoch": 4.2, "learning_rate": 9.67153380605694e-06, "loss": 0.0233, "step": 22909 }, { "epoch": 4.2, "learning_rate": 9.670791777549331e-06, "loss": 0.0254, "step": 22910 }, { "epoch": 4.2, "learning_rate": 9.670049750856323e-06, "loss": 0.0264, "step": 22911 }, { "epoch": 4.2, "learning_rate": 9.669307725982008e-06, "loss": 0.0257, "step": 22912 }, { "epoch": 4.2, "learning_rate": 9.668565702930476e-06, "loss": 0.0264, "step": 22913 }, { "epoch": 4.2, "learning_rate": 9.667823681705818e-06, "loss": 0.0233, "step": 22914 }, { "epoch": 4.2, "learning_rate": 9.667081662312125e-06, "loss": 0.0245, "step": 22915 }, { "epoch": 4.2, "learning_rate": 9.666339644753484e-06, "loss": 0.0245, "step": 22916 }, { "epoch": 4.2, "learning_rate": 9.665597629033986e-06, "loss": 0.0246, "step": 22917 }, { "epoch": 4.2, "learning_rate": 9.664855615157723e-06, "loss": 0.0285, "step": 22918 }, { "epoch": 4.2, "learning_rate": 9.664113603128787e-06, "loss": 0.0319, "step": 22919 }, { "epoch": 4.2, "learning_rate": 9.663371592951261e-06, "loss": 0.0318, "step": 22920 }, { "epoch": 4.2, "learning_rate": 9.66262958462924e-06, "loss": 0.0298, "step": 22921 }, { "epoch": 4.2, "learning_rate": 9.66188757816681e-06, "loss": 0.0264, "step": 22922 }, { "epoch": 4.2, "learning_rate": 9.661145573568067e-06, "loss": 0.0198, "step": 22923 }, { "epoch": 4.2, "learning_rate": 9.660403570837095e-06, "loss": 0.0308, "step": 22924 }, { "epoch": 4.2, "learning_rate": 9.65966156997799e-06, "loss": 0.0295, "step": 22925 }, { "epoch": 4.2, "learning_rate": 9.658919570994832e-06, "loss": 0.0281, "step": 22926 }, { "epoch": 4.2, "learning_rate": 9.65817757389172e-06, "loss": 0.0287, "step": 22927 }, { "epoch": 4.2, "learning_rate": 9.657435578672746e-06, "loss": 0.0212, "step": 22928 }, { "epoch": 4.2, "learning_rate": 9.656693585341992e-06, "loss": 0.0275, "step": 22929 }, { "epoch": 4.21, "learning_rate": 9.655951593903549e-06, "loss": 0.0271, "step": 22930 }, { "epoch": 4.21, "learning_rate": 9.65520960436151e-06, "loss": 0.0243, "step": 22931 }, { "epoch": 4.21, "learning_rate": 9.654467616719964e-06, "loss": 0.0265, "step": 22932 }, { "epoch": 4.21, "learning_rate": 9.653725630983e-06, "loss": 0.0239, "step": 22933 }, { "epoch": 4.21, "learning_rate": 9.652983647154712e-06, "loss": 0.0221, "step": 22934 }, { "epoch": 4.21, "learning_rate": 9.65224166523918e-06, "loss": 0.0262, "step": 22935 }, { "epoch": 4.21, "learning_rate": 9.651499685240504e-06, "loss": 0.0266, "step": 22936 }, { "epoch": 4.21, "learning_rate": 9.650757707162773e-06, "loss": 0.0292, "step": 22937 }, { "epoch": 4.21, "learning_rate": 9.650015731010069e-06, "loss": 0.0255, "step": 22938 }, { "epoch": 4.21, "learning_rate": 9.649273756786486e-06, "loss": 0.0238, "step": 22939 }, { "epoch": 4.21, "learning_rate": 9.648531784496116e-06, "loss": 0.0218, "step": 22940 }, { "epoch": 4.21, "learning_rate": 9.647789814143047e-06, "loss": 0.0258, "step": 22941 }, { "epoch": 4.21, "learning_rate": 9.647047845731369e-06, "loss": 0.024, "step": 22942 }, { "epoch": 4.21, "learning_rate": 9.646305879265173e-06, "loss": 0.0321, "step": 22943 }, { "epoch": 4.21, "learning_rate": 9.645563914748542e-06, "loss": 0.0226, "step": 22944 }, { "epoch": 4.21, "learning_rate": 9.644821952185574e-06, "loss": 0.0256, "step": 22945 }, { "epoch": 4.21, "learning_rate": 9.64407999158036e-06, "loss": 0.0247, "step": 22946 }, { "epoch": 4.21, "learning_rate": 9.643338032936981e-06, "loss": 0.0208, "step": 22947 }, { "epoch": 4.21, "learning_rate": 9.64259607625953e-06, "loss": 0.0193, "step": 22948 }, { "epoch": 4.21, "learning_rate": 9.6418541215521e-06, "loss": 0.0269, "step": 22949 }, { "epoch": 4.21, "learning_rate": 9.641112168818778e-06, "loss": 0.0249, "step": 22950 }, { "epoch": 4.21, "learning_rate": 9.640370218063655e-06, "loss": 0.0274, "step": 22951 }, { "epoch": 4.21, "learning_rate": 9.639628269290822e-06, "loss": 0.0259, "step": 22952 }, { "epoch": 4.21, "learning_rate": 9.638886322504358e-06, "loss": 0.0204, "step": 22953 }, { "epoch": 4.21, "learning_rate": 9.638144377708366e-06, "loss": 0.0242, "step": 22954 }, { "epoch": 4.21, "learning_rate": 9.637402434906935e-06, "loss": 0.0252, "step": 22955 }, { "epoch": 4.21, "learning_rate": 9.636660494104147e-06, "loss": 0.0257, "step": 22956 }, { "epoch": 4.21, "learning_rate": 9.635918555304092e-06, "loss": 0.0272, "step": 22957 }, { "epoch": 4.21, "learning_rate": 9.635176618510864e-06, "loss": 0.0273, "step": 22958 }, { "epoch": 4.21, "learning_rate": 9.634434683728553e-06, "loss": 0.0266, "step": 22959 }, { "epoch": 4.21, "learning_rate": 9.633692750961246e-06, "loss": 0.0267, "step": 22960 }, { "epoch": 4.21, "learning_rate": 9.63295082021303e-06, "loss": 0.0261, "step": 22961 }, { "epoch": 4.21, "learning_rate": 9.632208891488e-06, "loss": 0.0333, "step": 22962 }, { "epoch": 4.21, "learning_rate": 9.631466964790244e-06, "loss": 0.0199, "step": 22963 }, { "epoch": 4.21, "learning_rate": 9.630725040123853e-06, "loss": 0.0287, "step": 22964 }, { "epoch": 4.21, "learning_rate": 9.629983117492912e-06, "loss": 0.0292, "step": 22965 }, { "epoch": 4.21, "learning_rate": 9.629241196901509e-06, "loss": 0.0249, "step": 22966 }, { "epoch": 4.21, "learning_rate": 9.62849927835374e-06, "loss": 0.0274, "step": 22967 }, { "epoch": 4.21, "learning_rate": 9.627757361853692e-06, "loss": 0.0249, "step": 22968 }, { "epoch": 4.21, "learning_rate": 9.627015447405454e-06, "loss": 0.0287, "step": 22969 }, { "epoch": 4.21, "learning_rate": 9.626273535013114e-06, "loss": 0.0292, "step": 22970 }, { "epoch": 4.21, "learning_rate": 9.625531624680765e-06, "loss": 0.0251, "step": 22971 }, { "epoch": 4.21, "learning_rate": 9.624789716412498e-06, "loss": 0.024, "step": 22972 }, { "epoch": 4.21, "learning_rate": 9.624047810212394e-06, "loss": 0.0243, "step": 22973 }, { "epoch": 4.21, "learning_rate": 9.623305906084549e-06, "loss": 0.0195, "step": 22974 }, { "epoch": 4.21, "learning_rate": 9.622564004033049e-06, "loss": 0.0262, "step": 22975 }, { "epoch": 4.21, "learning_rate": 9.621822104061986e-06, "loss": 0.0218, "step": 22976 }, { "epoch": 4.21, "learning_rate": 9.621080206175448e-06, "loss": 0.0229, "step": 22977 }, { "epoch": 4.21, "learning_rate": 9.620338310377526e-06, "loss": 0.0265, "step": 22978 }, { "epoch": 4.21, "learning_rate": 9.619596416672306e-06, "loss": 0.0199, "step": 22979 }, { "epoch": 4.21, "learning_rate": 9.61885452506388e-06, "loss": 0.0309, "step": 22980 }, { "epoch": 4.21, "learning_rate": 9.61811263555634e-06, "loss": 0.0233, "step": 22981 }, { "epoch": 4.21, "learning_rate": 9.61737074815377e-06, "loss": 0.03, "step": 22982 }, { "epoch": 4.21, "learning_rate": 9.616628862860256e-06, "loss": 0.0303, "step": 22983 }, { "epoch": 4.21, "learning_rate": 9.615886979679901e-06, "loss": 0.0249, "step": 22984 }, { "epoch": 4.22, "learning_rate": 9.615145098616784e-06, "loss": 0.0245, "step": 22985 }, { "epoch": 4.22, "learning_rate": 9.614403219674995e-06, "loss": 0.0222, "step": 22986 }, { "epoch": 4.22, "learning_rate": 9.613661342858624e-06, "loss": 0.0231, "step": 22987 }, { "epoch": 4.22, "learning_rate": 9.61291946817176e-06, "loss": 0.0277, "step": 22988 }, { "epoch": 4.22, "learning_rate": 9.612177595618494e-06, "loss": 0.0274, "step": 22989 }, { "epoch": 4.22, "learning_rate": 9.611435725202916e-06, "loss": 0.027, "step": 22990 }, { "epoch": 4.22, "learning_rate": 9.610693856929113e-06, "loss": 0.0263, "step": 22991 }, { "epoch": 4.22, "learning_rate": 9.609951990801168e-06, "loss": 0.0302, "step": 22992 }, { "epoch": 4.22, "learning_rate": 9.609210126823186e-06, "loss": 0.0303, "step": 22993 }, { "epoch": 4.22, "learning_rate": 9.608468264999242e-06, "loss": 0.0268, "step": 22994 }, { "epoch": 4.22, "learning_rate": 9.607726405333431e-06, "loss": 0.0303, "step": 22995 }, { "epoch": 4.22, "learning_rate": 9.606984547829844e-06, "loss": 0.0272, "step": 22996 }, { "epoch": 4.22, "learning_rate": 9.606242692492561e-06, "loss": 0.0223, "step": 22997 }, { "epoch": 4.22, "learning_rate": 9.605500839325682e-06, "loss": 0.0202, "step": 22998 }, { "epoch": 4.22, "learning_rate": 9.604758988333293e-06, "loss": 0.0239, "step": 22999 }, { "epoch": 4.22, "learning_rate": 9.60401713951948e-06, "loss": 0.0252, "step": 23000 }, { "epoch": 4.22, "learning_rate": 9.603275292888328e-06, "loss": 0.0206, "step": 23001 }, { "epoch": 4.22, "learning_rate": 9.60253344844394e-06, "loss": 0.0259, "step": 23002 }, { "epoch": 4.22, "learning_rate": 9.601791606190395e-06, "loss": 0.0234, "step": 23003 }, { "epoch": 4.22, "learning_rate": 9.601049766131784e-06, "loss": 0.0233, "step": 23004 }, { "epoch": 4.22, "learning_rate": 9.600307928272193e-06, "loss": 0.0296, "step": 23005 }, { "epoch": 4.22, "learning_rate": 9.599566092615716e-06, "loss": 0.0218, "step": 23006 }, { "epoch": 4.22, "learning_rate": 9.598824259166441e-06, "loss": 0.0299, "step": 23007 }, { "epoch": 4.22, "learning_rate": 9.598082427928458e-06, "loss": 0.0273, "step": 23008 }, { "epoch": 4.22, "learning_rate": 9.597340598905851e-06, "loss": 0.0246, "step": 23009 }, { "epoch": 4.22, "learning_rate": 9.596598772102709e-06, "loss": 0.0272, "step": 23010 }, { "epoch": 4.22, "learning_rate": 9.59585694752313e-06, "loss": 0.028, "step": 23011 }, { "epoch": 4.22, "learning_rate": 9.595115125171195e-06, "loss": 0.0235, "step": 23012 }, { "epoch": 4.22, "learning_rate": 9.594373305050995e-06, "loss": 0.0286, "step": 23013 }, { "epoch": 4.22, "learning_rate": 9.593631487166616e-06, "loss": 0.0285, "step": 23014 }, { "epoch": 4.22, "learning_rate": 9.592889671522151e-06, "loss": 0.0224, "step": 23015 }, { "epoch": 4.22, "learning_rate": 9.59214785812169e-06, "loss": 0.0269, "step": 23016 }, { "epoch": 4.22, "learning_rate": 9.591406046969321e-06, "loss": 0.0285, "step": 23017 }, { "epoch": 4.22, "learning_rate": 9.59066423806913e-06, "loss": 0.0285, "step": 23018 }, { "epoch": 4.22, "learning_rate": 9.589922431425203e-06, "loss": 0.0276, "step": 23019 }, { "epoch": 4.22, "learning_rate": 9.589180627041636e-06, "loss": 0.0318, "step": 23020 }, { "epoch": 4.22, "learning_rate": 9.588438824922516e-06, "loss": 0.0217, "step": 23021 }, { "epoch": 4.22, "learning_rate": 9.587697025071931e-06, "loss": 0.0249, "step": 23022 }, { "epoch": 4.22, "learning_rate": 9.586955227493968e-06, "loss": 0.0286, "step": 23023 }, { "epoch": 4.22, "learning_rate": 9.586213432192717e-06, "loss": 0.0228, "step": 23024 }, { "epoch": 4.22, "learning_rate": 9.585471639172269e-06, "loss": 0.0279, "step": 23025 }, { "epoch": 4.22, "learning_rate": 9.584729848436712e-06, "loss": 0.0311, "step": 23026 }, { "epoch": 4.22, "learning_rate": 9.583988059990133e-06, "loss": 0.0278, "step": 23027 }, { "epoch": 4.22, "learning_rate": 9.58324627383662e-06, "loss": 0.0303, "step": 23028 }, { "epoch": 4.22, "learning_rate": 9.582504489980263e-06, "loss": 0.027, "step": 23029 }, { "epoch": 4.22, "learning_rate": 9.581762708425153e-06, "loss": 0.0332, "step": 23030 }, { "epoch": 4.22, "learning_rate": 9.581020929175377e-06, "loss": 0.0249, "step": 23031 }, { "epoch": 4.22, "learning_rate": 9.580279152235019e-06, "loss": 0.0281, "step": 23032 }, { "epoch": 4.22, "learning_rate": 9.579537377608177e-06, "loss": 0.0195, "step": 23033 }, { "epoch": 4.22, "learning_rate": 9.578795605298933e-06, "loss": 0.0249, "step": 23034 }, { "epoch": 4.22, "learning_rate": 9.57805383531138e-06, "loss": 0.0297, "step": 23035 }, { "epoch": 4.22, "learning_rate": 9.577312067649598e-06, "loss": 0.0239, "step": 23036 }, { "epoch": 4.22, "learning_rate": 9.576570302317687e-06, "loss": 0.0273, "step": 23037 }, { "epoch": 4.22, "learning_rate": 9.57582853931973e-06, "loss": 0.0284, "step": 23038 }, { "epoch": 4.23, "learning_rate": 9.575086778659817e-06, "loss": 0.0289, "step": 23039 }, { "epoch": 4.23, "learning_rate": 9.574345020342032e-06, "loss": 0.0256, "step": 23040 }, { "epoch": 4.23, "learning_rate": 9.57360326437047e-06, "loss": 0.0283, "step": 23041 }, { "epoch": 4.23, "learning_rate": 9.572861510749215e-06, "loss": 0.0241, "step": 23042 }, { "epoch": 4.23, "learning_rate": 9.572119759482358e-06, "loss": 0.0259, "step": 23043 }, { "epoch": 4.23, "learning_rate": 9.57137801057399e-06, "loss": 0.026, "step": 23044 }, { "epoch": 4.23, "learning_rate": 9.570636264028189e-06, "loss": 0.0237, "step": 23045 }, { "epoch": 4.23, "learning_rate": 9.569894519849058e-06, "loss": 0.0243, "step": 23046 }, { "epoch": 4.23, "learning_rate": 9.569152778040677e-06, "loss": 0.0249, "step": 23047 }, { "epoch": 4.23, "learning_rate": 9.568411038607135e-06, "loss": 0.0249, "step": 23048 }, { "epoch": 4.23, "learning_rate": 9.567669301552522e-06, "loss": 0.0207, "step": 23049 }, { "epoch": 4.23, "learning_rate": 9.566927566880924e-06, "loss": 0.0271, "step": 23050 }, { "epoch": 4.23, "learning_rate": 9.566185834596432e-06, "loss": 0.0277, "step": 23051 }, { "epoch": 4.23, "learning_rate": 9.565444104703135e-06, "loss": 0.0193, "step": 23052 }, { "epoch": 4.23, "learning_rate": 9.564702377205122e-06, "loss": 0.0238, "step": 23053 }, { "epoch": 4.23, "learning_rate": 9.563960652106474e-06, "loss": 0.0237, "step": 23054 }, { "epoch": 4.23, "learning_rate": 9.56321892941129e-06, "loss": 0.0297, "step": 23055 }, { "epoch": 4.23, "learning_rate": 9.562477209123653e-06, "loss": 0.0253, "step": 23056 }, { "epoch": 4.23, "learning_rate": 9.561735491247652e-06, "loss": 0.0362, "step": 23057 }, { "epoch": 4.23, "learning_rate": 9.560993775787373e-06, "loss": 0.0215, "step": 23058 }, { "epoch": 4.23, "learning_rate": 9.560252062746907e-06, "loss": 0.0261, "step": 23059 }, { "epoch": 4.23, "learning_rate": 9.559510352130344e-06, "loss": 0.0251, "step": 23060 }, { "epoch": 4.23, "learning_rate": 9.55876864394177e-06, "loss": 0.0248, "step": 23061 }, { "epoch": 4.23, "learning_rate": 9.558026938185275e-06, "loss": 0.0243, "step": 23062 }, { "epoch": 4.23, "learning_rate": 9.55728523486494e-06, "loss": 0.0251, "step": 23063 }, { "epoch": 4.23, "learning_rate": 9.556543533984864e-06, "loss": 0.0301, "step": 23064 }, { "epoch": 4.23, "learning_rate": 9.555801835549129e-06, "loss": 0.0214, "step": 23065 }, { "epoch": 4.23, "learning_rate": 9.555060139561826e-06, "loss": 0.0238, "step": 23066 }, { "epoch": 4.23, "learning_rate": 9.554318446027039e-06, "loss": 0.0227, "step": 23067 }, { "epoch": 4.23, "learning_rate": 9.55357675494886e-06, "loss": 0.0257, "step": 23068 }, { "epoch": 4.23, "learning_rate": 9.552835066331379e-06, "loss": 0.0273, "step": 23069 }, { "epoch": 4.23, "learning_rate": 9.55209338017868e-06, "loss": 0.0262, "step": 23070 }, { "epoch": 4.23, "learning_rate": 9.551351696494854e-06, "loss": 0.0283, "step": 23071 }, { "epoch": 4.23, "learning_rate": 9.550610015283983e-06, "loss": 0.022, "step": 23072 }, { "epoch": 4.23, "learning_rate": 9.549868336550167e-06, "loss": 0.0241, "step": 23073 }, { "epoch": 4.23, "learning_rate": 9.549126660297483e-06, "loss": 0.021, "step": 23074 }, { "epoch": 4.23, "learning_rate": 9.548384986530025e-06, "loss": 0.0287, "step": 23075 }, { "epoch": 4.23, "learning_rate": 9.547643315251877e-06, "loss": 0.0298, "step": 23076 }, { "epoch": 4.23, "learning_rate": 9.546901646467132e-06, "loss": 0.0258, "step": 23077 }, { "epoch": 4.23, "learning_rate": 9.546159980179875e-06, "loss": 0.025, "step": 23078 }, { "epoch": 4.23, "learning_rate": 9.545418316394196e-06, "loss": 0.022, "step": 23079 }, { "epoch": 4.23, "learning_rate": 9.544676655114183e-06, "loss": 0.0305, "step": 23080 }, { "epoch": 4.23, "learning_rate": 9.543934996343917e-06, "loss": 0.0248, "step": 23081 }, { "epoch": 4.23, "learning_rate": 9.543193340087499e-06, "loss": 0.0236, "step": 23082 }, { "epoch": 4.23, "learning_rate": 9.542451686349008e-06, "loss": 0.029, "step": 23083 }, { "epoch": 4.23, "learning_rate": 9.541710035132532e-06, "loss": 0.0258, "step": 23084 }, { "epoch": 4.23, "learning_rate": 9.540968386442161e-06, "loss": 0.0226, "step": 23085 }, { "epoch": 4.23, "learning_rate": 9.540226740281985e-06, "loss": 0.0242, "step": 23086 }, { "epoch": 4.23, "learning_rate": 9.53948509665609e-06, "loss": 0.0197, "step": 23087 }, { "epoch": 4.23, "learning_rate": 9.538743455568565e-06, "loss": 0.0264, "step": 23088 }, { "epoch": 4.23, "learning_rate": 9.53800181702349e-06, "loss": 0.0272, "step": 23089 }, { "epoch": 4.23, "learning_rate": 9.537260181024966e-06, "loss": 0.0278, "step": 23090 }, { "epoch": 4.23, "learning_rate": 9.536518547577075e-06, "loss": 0.0234, "step": 23091 }, { "epoch": 4.23, "learning_rate": 9.535776916683903e-06, "loss": 0.022, "step": 23092 }, { "epoch": 4.23, "learning_rate": 9.53503528834954e-06, "loss": 0.0317, "step": 23093 }, { "epoch": 4.24, "learning_rate": 9.53429366257807e-06, "loss": 0.0301, "step": 23094 }, { "epoch": 4.24, "learning_rate": 9.533552039373588e-06, "loss": 0.0291, "step": 23095 }, { "epoch": 4.24, "learning_rate": 9.532810418740179e-06, "loss": 0.0209, "step": 23096 }, { "epoch": 4.24, "learning_rate": 9.53206880068193e-06, "loss": 0.0302, "step": 23097 }, { "epoch": 4.24, "learning_rate": 9.531327185202922e-06, "loss": 0.0249, "step": 23098 }, { "epoch": 4.24, "learning_rate": 9.530585572307256e-06, "loss": 0.0278, "step": 23099 }, { "epoch": 4.24, "learning_rate": 9.529843961999014e-06, "loss": 0.0242, "step": 23100 }, { "epoch": 4.24, "learning_rate": 9.52910235428228e-06, "loss": 0.0264, "step": 23101 }, { "epoch": 4.24, "learning_rate": 9.528360749161147e-06, "loss": 0.0214, "step": 23102 }, { "epoch": 4.24, "learning_rate": 9.527619146639698e-06, "loss": 0.0198, "step": 23103 }, { "epoch": 4.24, "learning_rate": 9.526877546722023e-06, "loss": 0.0268, "step": 23104 }, { "epoch": 4.24, "learning_rate": 9.526135949412213e-06, "loss": 0.0244, "step": 23105 }, { "epoch": 4.24, "learning_rate": 9.525394354714353e-06, "loss": 0.0243, "step": 23106 }, { "epoch": 4.24, "learning_rate": 9.524652762632525e-06, "loss": 0.0272, "step": 23107 }, { "epoch": 4.24, "learning_rate": 9.523911173170826e-06, "loss": 0.0305, "step": 23108 }, { "epoch": 4.24, "learning_rate": 9.523169586333342e-06, "loss": 0.0225, "step": 23109 }, { "epoch": 4.24, "learning_rate": 9.522428002124157e-06, "loss": 0.0261, "step": 23110 }, { "epoch": 4.24, "learning_rate": 9.521686420547358e-06, "loss": 0.026, "step": 23111 }, { "epoch": 4.24, "learning_rate": 9.520944841607036e-06, "loss": 0.0216, "step": 23112 }, { "epoch": 4.24, "learning_rate": 9.520203265307278e-06, "loss": 0.0229, "step": 23113 }, { "epoch": 4.24, "learning_rate": 9.519461691652169e-06, "loss": 0.0241, "step": 23114 }, { "epoch": 4.24, "learning_rate": 9.518720120645802e-06, "loss": 0.0228, "step": 23115 }, { "epoch": 4.24, "learning_rate": 9.517978552292254e-06, "loss": 0.0229, "step": 23116 }, { "epoch": 4.24, "learning_rate": 9.517236986595624e-06, "loss": 0.0241, "step": 23117 }, { "epoch": 4.24, "learning_rate": 9.516495423559998e-06, "loss": 0.0249, "step": 23118 }, { "epoch": 4.24, "learning_rate": 9.515753863189458e-06, "loss": 0.021, "step": 23119 }, { "epoch": 4.24, "learning_rate": 9.515012305488091e-06, "loss": 0.0195, "step": 23120 }, { "epoch": 4.24, "learning_rate": 9.514270750459991e-06, "loss": 0.0248, "step": 23121 }, { "epoch": 4.24, "learning_rate": 9.51352919810924e-06, "loss": 0.0281, "step": 23122 }, { "epoch": 4.24, "learning_rate": 9.51278764843993e-06, "loss": 0.0256, "step": 23123 }, { "epoch": 4.24, "learning_rate": 9.512046101456147e-06, "loss": 0.0247, "step": 23124 }, { "epoch": 4.24, "learning_rate": 9.51130455716197e-06, "loss": 0.0265, "step": 23125 }, { "epoch": 4.24, "learning_rate": 9.5105630155615e-06, "loss": 0.0277, "step": 23126 }, { "epoch": 4.24, "learning_rate": 9.509821476658818e-06, "loss": 0.0226, "step": 23127 }, { "epoch": 4.24, "learning_rate": 9.50907994045801e-06, "loss": 0.029, "step": 23128 }, { "epoch": 4.24, "learning_rate": 9.508338406963164e-06, "loss": 0.0229, "step": 23129 }, { "epoch": 4.24, "learning_rate": 9.50759687617837e-06, "loss": 0.0222, "step": 23130 }, { "epoch": 4.24, "learning_rate": 9.506855348107711e-06, "loss": 0.0259, "step": 23131 }, { "epoch": 4.24, "learning_rate": 9.50611382275528e-06, "loss": 0.0228, "step": 23132 }, { "epoch": 4.24, "learning_rate": 9.505372300125158e-06, "loss": 0.025, "step": 23133 }, { "epoch": 4.24, "learning_rate": 9.504630780221437e-06, "loss": 0.0236, "step": 23134 }, { "epoch": 4.24, "learning_rate": 9.503889263048205e-06, "loss": 0.0215, "step": 23135 }, { "epoch": 4.24, "learning_rate": 9.503147748609544e-06, "loss": 0.0267, "step": 23136 }, { "epoch": 4.24, "learning_rate": 9.502406236909547e-06, "loss": 0.0241, "step": 23137 }, { "epoch": 4.24, "learning_rate": 9.501664727952294e-06, "loss": 0.0244, "step": 23138 }, { "epoch": 4.24, "learning_rate": 9.50092322174188e-06, "loss": 0.0271, "step": 23139 }, { "epoch": 4.24, "learning_rate": 9.500181718282388e-06, "loss": 0.0267, "step": 23140 }, { "epoch": 4.24, "learning_rate": 9.499440217577907e-06, "loss": 0.0232, "step": 23141 }, { "epoch": 4.24, "learning_rate": 9.498698719632521e-06, "loss": 0.0243, "step": 23142 }, { "epoch": 4.24, "learning_rate": 9.497957224450321e-06, "loss": 0.028, "step": 23143 }, { "epoch": 4.24, "learning_rate": 9.497215732035396e-06, "loss": 0.0281, "step": 23144 }, { "epoch": 4.24, "learning_rate": 9.496474242391826e-06, "loss": 0.0238, "step": 23145 }, { "epoch": 4.24, "learning_rate": 9.495732755523703e-06, "loss": 0.0244, "step": 23146 }, { "epoch": 4.24, "learning_rate": 9.49499127143511e-06, "loss": 0.0216, "step": 23147 }, { "epoch": 4.25, "learning_rate": 9.494249790130138e-06, "loss": 0.0293, "step": 23148 }, { "epoch": 4.25, "learning_rate": 9.493508311612874e-06, "loss": 0.0219, "step": 23149 }, { "epoch": 4.25, "learning_rate": 9.492766835887405e-06, "loss": 0.0275, "step": 23150 }, { "epoch": 4.25, "learning_rate": 9.492025362957812e-06, "loss": 0.0209, "step": 23151 }, { "epoch": 4.25, "learning_rate": 9.491283892828192e-06, "loss": 0.0238, "step": 23152 }, { "epoch": 4.25, "learning_rate": 9.490542425502629e-06, "loss": 0.0285, "step": 23153 }, { "epoch": 4.25, "learning_rate": 9.489800960985206e-06, "loss": 0.0262, "step": 23154 }, { "epoch": 4.25, "learning_rate": 9.489059499280011e-06, "loss": 0.0237, "step": 23155 }, { "epoch": 4.25, "learning_rate": 9.488318040391132e-06, "loss": 0.024, "step": 23156 }, { "epoch": 4.25, "learning_rate": 9.487576584322655e-06, "loss": 0.0243, "step": 23157 }, { "epoch": 4.25, "learning_rate": 9.48683513107867e-06, "loss": 0.0225, "step": 23158 }, { "epoch": 4.25, "learning_rate": 9.48609368066326e-06, "loss": 0.0267, "step": 23159 }, { "epoch": 4.25, "learning_rate": 9.485352233080514e-06, "loss": 0.0297, "step": 23160 }, { "epoch": 4.25, "learning_rate": 9.484610788334519e-06, "loss": 0.0262, "step": 23161 }, { "epoch": 4.25, "learning_rate": 9.483869346429365e-06, "loss": 0.0236, "step": 23162 }, { "epoch": 4.25, "learning_rate": 9.483127907369131e-06, "loss": 0.0224, "step": 23163 }, { "epoch": 4.25, "learning_rate": 9.482386471157905e-06, "loss": 0.0237, "step": 23164 }, { "epoch": 4.25, "learning_rate": 9.481645037799783e-06, "loss": 0.0268, "step": 23165 }, { "epoch": 4.25, "learning_rate": 9.480903607298844e-06, "loss": 0.0222, "step": 23166 }, { "epoch": 4.25, "learning_rate": 9.480162179659175e-06, "loss": 0.0258, "step": 23167 }, { "epoch": 4.25, "learning_rate": 9.479420754884867e-06, "loss": 0.0281, "step": 23168 }, { "epoch": 4.25, "learning_rate": 9.47867933298e-06, "loss": 0.0227, "step": 23169 }, { "epoch": 4.25, "learning_rate": 9.477937913948668e-06, "loss": 0.0238, "step": 23170 }, { "epoch": 4.25, "learning_rate": 9.477196497794955e-06, "loss": 0.028, "step": 23171 }, { "epoch": 4.25, "learning_rate": 9.476455084522945e-06, "loss": 0.0283, "step": 23172 }, { "epoch": 4.25, "learning_rate": 9.475713674136726e-06, "loss": 0.0284, "step": 23173 }, { "epoch": 4.25, "learning_rate": 9.474972266640388e-06, "loss": 0.0358, "step": 23174 }, { "epoch": 4.25, "learning_rate": 9.474230862038013e-06, "loss": 0.0311, "step": 23175 }, { "epoch": 4.25, "learning_rate": 9.47348946033369e-06, "loss": 0.0243, "step": 23176 }, { "epoch": 4.25, "learning_rate": 9.472748061531509e-06, "loss": 0.0291, "step": 23177 }, { "epoch": 4.25, "learning_rate": 9.472006665635547e-06, "loss": 0.0256, "step": 23178 }, { "epoch": 4.25, "learning_rate": 9.471265272649902e-06, "loss": 0.0232, "step": 23179 }, { "epoch": 4.25, "learning_rate": 9.470523882578654e-06, "loss": 0.0231, "step": 23180 }, { "epoch": 4.25, "learning_rate": 9.469782495425889e-06, "loss": 0.0277, "step": 23181 }, { "epoch": 4.25, "learning_rate": 9.469041111195696e-06, "loss": 0.024, "step": 23182 }, { "epoch": 4.25, "learning_rate": 9.468299729892161e-06, "loss": 0.0261, "step": 23183 }, { "epoch": 4.25, "learning_rate": 9.46755835151937e-06, "loss": 0.0268, "step": 23184 }, { "epoch": 4.25, "learning_rate": 9.466816976081411e-06, "loss": 0.0242, "step": 23185 }, { "epoch": 4.25, "learning_rate": 9.466075603582367e-06, "loss": 0.0254, "step": 23186 }, { "epoch": 4.25, "learning_rate": 9.465334234026328e-06, "loss": 0.0301, "step": 23187 }, { "epoch": 4.25, "learning_rate": 9.46459286741738e-06, "loss": 0.0348, "step": 23188 }, { "epoch": 4.25, "learning_rate": 9.463851503759612e-06, "loss": 0.0276, "step": 23189 }, { "epoch": 4.25, "learning_rate": 9.463110143057102e-06, "loss": 0.0246, "step": 23190 }, { "epoch": 4.25, "learning_rate": 9.462368785313941e-06, "loss": 0.023, "step": 23191 }, { "epoch": 4.25, "learning_rate": 9.461627430534218e-06, "loss": 0.0238, "step": 23192 }, { "epoch": 4.25, "learning_rate": 9.460886078722017e-06, "loss": 0.0217, "step": 23193 }, { "epoch": 4.25, "learning_rate": 9.460144729881426e-06, "loss": 0.0252, "step": 23194 }, { "epoch": 4.25, "learning_rate": 9.459403384016525e-06, "loss": 0.0264, "step": 23195 }, { "epoch": 4.25, "learning_rate": 9.458662041131409e-06, "loss": 0.0273, "step": 23196 }, { "epoch": 4.25, "learning_rate": 9.45792070123016e-06, "loss": 0.027, "step": 23197 }, { "epoch": 4.25, "learning_rate": 9.457179364316867e-06, "loss": 0.0241, "step": 23198 }, { "epoch": 4.25, "learning_rate": 9.456438030395613e-06, "loss": 0.0257, "step": 23199 }, { "epoch": 4.25, "learning_rate": 9.455696699470483e-06, "loss": 0.0249, "step": 23200 }, { "epoch": 4.25, "learning_rate": 9.454955371545565e-06, "loss": 0.0277, "step": 23201 }, { "epoch": 4.25, "learning_rate": 9.454214046624949e-06, "loss": 0.0237, "step": 23202 }, { "epoch": 4.26, "learning_rate": 9.453472724712716e-06, "loss": 0.0304, "step": 23203 }, { "epoch": 4.26, "learning_rate": 9.452731405812953e-06, "loss": 0.0228, "step": 23204 }, { "epoch": 4.26, "learning_rate": 9.451990089929749e-06, "loss": 0.0248, "step": 23205 }, { "epoch": 4.26, "learning_rate": 9.451248777067189e-06, "loss": 0.027, "step": 23206 }, { "epoch": 4.26, "learning_rate": 9.45050746722936e-06, "loss": 0.0238, "step": 23207 }, { "epoch": 4.26, "learning_rate": 9.449766160420345e-06, "loss": 0.0267, "step": 23208 }, { "epoch": 4.26, "learning_rate": 9.44902485664423e-06, "loss": 0.0317, "step": 23209 }, { "epoch": 4.26, "learning_rate": 9.448283555905105e-06, "loss": 0.023, "step": 23210 }, { "epoch": 4.26, "learning_rate": 9.447542258207052e-06, "loss": 0.027, "step": 23211 }, { "epoch": 4.26, "learning_rate": 9.446800963554161e-06, "loss": 0.0283, "step": 23212 }, { "epoch": 4.26, "learning_rate": 9.446059671950514e-06, "loss": 0.0232, "step": 23213 }, { "epoch": 4.26, "learning_rate": 9.4453183834002e-06, "loss": 0.0188, "step": 23214 }, { "epoch": 4.26, "learning_rate": 9.444577097907307e-06, "loss": 0.0268, "step": 23215 }, { "epoch": 4.26, "learning_rate": 9.443835815475917e-06, "loss": 0.0236, "step": 23216 }, { "epoch": 4.26, "learning_rate": 9.443094536110112e-06, "loss": 0.0182, "step": 23217 }, { "epoch": 4.26, "learning_rate": 9.442353259813988e-06, "loss": 0.0304, "step": 23218 }, { "epoch": 4.26, "learning_rate": 9.441611986591625e-06, "loss": 0.0262, "step": 23219 }, { "epoch": 4.26, "learning_rate": 9.440870716447109e-06, "loss": 0.0294, "step": 23220 }, { "epoch": 4.26, "learning_rate": 9.440129449384527e-06, "loss": 0.0282, "step": 23221 }, { "epoch": 4.26, "learning_rate": 9.439388185407964e-06, "loss": 0.02, "step": 23222 }, { "epoch": 4.26, "learning_rate": 9.438646924521506e-06, "loss": 0.0283, "step": 23223 }, { "epoch": 4.26, "learning_rate": 9.437905666729242e-06, "loss": 0.0289, "step": 23224 }, { "epoch": 4.26, "learning_rate": 9.437164412035255e-06, "loss": 0.0272, "step": 23225 }, { "epoch": 4.26, "learning_rate": 9.436423160443625e-06, "loss": 0.0199, "step": 23226 }, { "epoch": 4.26, "learning_rate": 9.435681911958451e-06, "loss": 0.0288, "step": 23227 }, { "epoch": 4.26, "learning_rate": 9.434940666583809e-06, "loss": 0.0303, "step": 23228 }, { "epoch": 4.26, "learning_rate": 9.434199424323788e-06, "loss": 0.0251, "step": 23229 }, { "epoch": 4.26, "learning_rate": 9.433458185182472e-06, "loss": 0.0247, "step": 23230 }, { "epoch": 4.26, "learning_rate": 9.432716949163946e-06, "loss": 0.0247, "step": 23231 }, { "epoch": 4.26, "learning_rate": 9.431975716272299e-06, "loss": 0.0216, "step": 23232 }, { "epoch": 4.26, "learning_rate": 9.431234486511615e-06, "loss": 0.0261, "step": 23233 }, { "epoch": 4.26, "learning_rate": 9.430493259885983e-06, "loss": 0.0273, "step": 23234 }, { "epoch": 4.26, "learning_rate": 9.429752036399479e-06, "loss": 0.0299, "step": 23235 }, { "epoch": 4.26, "learning_rate": 9.429010816056202e-06, "loss": 0.0294, "step": 23236 }, { "epoch": 4.26, "learning_rate": 9.428269598860226e-06, "loss": 0.0258, "step": 23237 }, { "epoch": 4.26, "learning_rate": 9.427528384815644e-06, "loss": 0.0229, "step": 23238 }, { "epoch": 4.26, "learning_rate": 9.426787173926536e-06, "loss": 0.0272, "step": 23239 }, { "epoch": 4.26, "learning_rate": 9.426045966196992e-06, "loss": 0.0244, "step": 23240 }, { "epoch": 4.26, "learning_rate": 9.425304761631098e-06, "loss": 0.0261, "step": 23241 }, { "epoch": 4.26, "learning_rate": 9.42456356023294e-06, "loss": 0.0288, "step": 23242 }, { "epoch": 4.26, "learning_rate": 9.423822362006596e-06, "loss": 0.0224, "step": 23243 }, { "epoch": 4.26, "learning_rate": 9.423081166956155e-06, "loss": 0.0246, "step": 23244 }, { "epoch": 4.26, "learning_rate": 9.422339975085707e-06, "loss": 0.0303, "step": 23245 }, { "epoch": 4.26, "learning_rate": 9.421598786399336e-06, "loss": 0.021, "step": 23246 }, { "epoch": 4.26, "learning_rate": 9.420857600901125e-06, "loss": 0.0237, "step": 23247 }, { "epoch": 4.26, "learning_rate": 9.420116418595158e-06, "loss": 0.0222, "step": 23248 }, { "epoch": 4.26, "learning_rate": 9.419375239485525e-06, "loss": 0.0234, "step": 23249 }, { "epoch": 4.26, "learning_rate": 9.41863406357631e-06, "loss": 0.0343, "step": 23250 }, { "epoch": 4.26, "learning_rate": 9.4178928908716e-06, "loss": 0.0341, "step": 23251 }, { "epoch": 4.26, "learning_rate": 9.417151721375475e-06, "loss": 0.0243, "step": 23252 }, { "epoch": 4.26, "learning_rate": 9.41641055509202e-06, "loss": 0.0266, "step": 23253 }, { "epoch": 4.26, "learning_rate": 9.415669392025329e-06, "loss": 0.0298, "step": 23254 }, { "epoch": 4.26, "learning_rate": 9.41492823217948e-06, "loss": 0.0284, "step": 23255 }, { "epoch": 4.26, "learning_rate": 9.414187075558561e-06, "loss": 0.0228, "step": 23256 }, { "epoch": 4.26, "learning_rate": 9.413445922166654e-06, "loss": 0.0281, "step": 23257 }, { "epoch": 4.27, "learning_rate": 9.412704772007849e-06, "loss": 0.0252, "step": 23258 }, { "epoch": 4.27, "learning_rate": 9.411963625086228e-06, "loss": 0.0248, "step": 23259 }, { "epoch": 4.27, "learning_rate": 9.411222481405882e-06, "loss": 0.0226, "step": 23260 }, { "epoch": 4.27, "learning_rate": 9.410481340970882e-06, "loss": 0.0245, "step": 23261 }, { "epoch": 4.27, "learning_rate": 9.40974020378533e-06, "loss": 0.0251, "step": 23262 }, { "epoch": 4.27, "learning_rate": 9.408999069853304e-06, "loss": 0.0243, "step": 23263 }, { "epoch": 4.27, "learning_rate": 9.408257939178887e-06, "loss": 0.0278, "step": 23264 }, { "epoch": 4.27, "learning_rate": 9.407516811766166e-06, "loss": 0.0256, "step": 23265 }, { "epoch": 4.27, "learning_rate": 9.406775687619223e-06, "loss": 0.0256, "step": 23266 }, { "epoch": 4.27, "learning_rate": 9.406034566742151e-06, "loss": 0.0278, "step": 23267 }, { "epoch": 4.27, "learning_rate": 9.405293449139031e-06, "loss": 0.0274, "step": 23268 }, { "epoch": 4.27, "learning_rate": 9.404552334813947e-06, "loss": 0.0307, "step": 23269 }, { "epoch": 4.27, "learning_rate": 9.403811223770978e-06, "loss": 0.0228, "step": 23270 }, { "epoch": 4.27, "learning_rate": 9.403070116014221e-06, "loss": 0.0269, "step": 23271 }, { "epoch": 4.27, "learning_rate": 9.402329011547757e-06, "loss": 0.0253, "step": 23272 }, { "epoch": 4.27, "learning_rate": 9.401587910375668e-06, "loss": 0.0237, "step": 23273 }, { "epoch": 4.27, "learning_rate": 9.40084681250204e-06, "loss": 0.0285, "step": 23274 }, { "epoch": 4.27, "learning_rate": 9.400105717930956e-06, "loss": 0.0243, "step": 23275 }, { "epoch": 4.27, "learning_rate": 9.399364626666506e-06, "loss": 0.0267, "step": 23276 }, { "epoch": 4.27, "learning_rate": 9.398623538712773e-06, "loss": 0.0214, "step": 23277 }, { "epoch": 4.27, "learning_rate": 9.397882454073842e-06, "loss": 0.0257, "step": 23278 }, { "epoch": 4.27, "learning_rate": 9.397141372753791e-06, "loss": 0.0215, "step": 23279 }, { "epoch": 4.27, "learning_rate": 9.396400294756715e-06, "loss": 0.0314, "step": 23280 }, { "epoch": 4.27, "learning_rate": 9.395659220086697e-06, "loss": 0.0201, "step": 23281 }, { "epoch": 4.27, "learning_rate": 9.394918148747818e-06, "loss": 0.0204, "step": 23282 }, { "epoch": 4.27, "learning_rate": 9.394177080744164e-06, "loss": 0.0245, "step": 23283 }, { "epoch": 4.27, "learning_rate": 9.393436016079818e-06, "loss": 0.0327, "step": 23284 }, { "epoch": 4.27, "learning_rate": 9.392694954758868e-06, "loss": 0.0209, "step": 23285 }, { "epoch": 4.27, "learning_rate": 9.391953896785398e-06, "loss": 0.0287, "step": 23286 }, { "epoch": 4.27, "learning_rate": 9.391212842163496e-06, "loss": 0.0217, "step": 23287 }, { "epoch": 4.27, "learning_rate": 9.390471790897235e-06, "loss": 0.0235, "step": 23288 }, { "epoch": 4.27, "learning_rate": 9.389730742990714e-06, "loss": 0.0225, "step": 23289 }, { "epoch": 4.27, "learning_rate": 9.388989698448011e-06, "loss": 0.0293, "step": 23290 }, { "epoch": 4.27, "learning_rate": 9.38824865727321e-06, "loss": 0.0198, "step": 23291 }, { "epoch": 4.27, "learning_rate": 9.387507619470396e-06, "loss": 0.0218, "step": 23292 }, { "epoch": 4.27, "learning_rate": 9.386766585043654e-06, "loss": 0.0239, "step": 23293 }, { "epoch": 4.27, "learning_rate": 9.38602555399707e-06, "loss": 0.0249, "step": 23294 }, { "epoch": 4.27, "learning_rate": 9.385284526334728e-06, "loss": 0.0276, "step": 23295 }, { "epoch": 4.27, "learning_rate": 9.384543502060713e-06, "loss": 0.0284, "step": 23296 }, { "epoch": 4.27, "learning_rate": 9.383802481179104e-06, "loss": 0.0238, "step": 23297 }, { "epoch": 4.27, "learning_rate": 9.383061463693996e-06, "loss": 0.0195, "step": 23298 }, { "epoch": 4.27, "learning_rate": 9.382320449609466e-06, "loss": 0.021, "step": 23299 }, { "epoch": 4.27, "learning_rate": 9.3815794389296e-06, "loss": 0.0191, "step": 23300 }, { "epoch": 4.27, "learning_rate": 9.380838431658479e-06, "loss": 0.0298, "step": 23301 }, { "epoch": 4.27, "learning_rate": 9.380097427800194e-06, "loss": 0.0232, "step": 23302 }, { "epoch": 4.27, "learning_rate": 9.379356427358825e-06, "loss": 0.0261, "step": 23303 }, { "epoch": 4.27, "learning_rate": 9.37861543033846e-06, "loss": 0.0288, "step": 23304 }, { "epoch": 4.27, "learning_rate": 9.377874436743184e-06, "loss": 0.0285, "step": 23305 }, { "epoch": 4.27, "learning_rate": 9.377133446577072e-06, "loss": 0.0282, "step": 23306 }, { "epoch": 4.27, "learning_rate": 9.37639245984422e-06, "loss": 0.0299, "step": 23307 }, { "epoch": 4.27, "learning_rate": 9.375651476548705e-06, "loss": 0.0252, "step": 23308 }, { "epoch": 4.27, "learning_rate": 9.374910496694614e-06, "loss": 0.0262, "step": 23309 }, { "epoch": 4.27, "learning_rate": 9.37416952028603e-06, "loss": 0.0246, "step": 23310 }, { "epoch": 4.27, "learning_rate": 9.37342854732704e-06, "loss": 0.0278, "step": 23311 }, { "epoch": 4.28, "learning_rate": 9.372687577821726e-06, "loss": 0.0228, "step": 23312 }, { "epoch": 4.28, "learning_rate": 9.371946611774174e-06, "loss": 0.0246, "step": 23313 }, { "epoch": 4.28, "learning_rate": 9.371205649188465e-06, "loss": 0.0233, "step": 23314 }, { "epoch": 4.28, "learning_rate": 9.370464690068685e-06, "loss": 0.0279, "step": 23315 }, { "epoch": 4.28, "learning_rate": 9.369723734418921e-06, "loss": 0.0248, "step": 23316 }, { "epoch": 4.28, "learning_rate": 9.368982782243255e-06, "loss": 0.0232, "step": 23317 }, { "epoch": 4.28, "learning_rate": 9.368241833545769e-06, "loss": 0.0257, "step": 23318 }, { "epoch": 4.28, "learning_rate": 9.367500888330546e-06, "loss": 0.0231, "step": 23319 }, { "epoch": 4.28, "learning_rate": 9.366759946601676e-06, "loss": 0.0224, "step": 23320 }, { "epoch": 4.28, "learning_rate": 9.36601900836324e-06, "loss": 0.0242, "step": 23321 }, { "epoch": 4.28, "learning_rate": 9.365278073619321e-06, "loss": 0.0273, "step": 23322 }, { "epoch": 4.28, "learning_rate": 9.364537142374003e-06, "loss": 0.0263, "step": 23323 }, { "epoch": 4.28, "learning_rate": 9.363796214631373e-06, "loss": 0.0283, "step": 23324 }, { "epoch": 4.28, "learning_rate": 9.363055290395516e-06, "loss": 0.0242, "step": 23325 }, { "epoch": 4.28, "learning_rate": 9.36231436967051e-06, "loss": 0.0266, "step": 23326 }, { "epoch": 4.28, "learning_rate": 9.361573452460441e-06, "loss": 0.0219, "step": 23327 }, { "epoch": 4.28, "learning_rate": 9.360832538769395e-06, "loss": 0.0219, "step": 23328 }, { "epoch": 4.28, "learning_rate": 9.360091628601455e-06, "loss": 0.0316, "step": 23329 }, { "epoch": 4.28, "learning_rate": 9.359350721960706e-06, "loss": 0.0255, "step": 23330 }, { "epoch": 4.28, "learning_rate": 9.35860981885123e-06, "loss": 0.0208, "step": 23331 }, { "epoch": 4.28, "learning_rate": 9.35786891927711e-06, "loss": 0.0304, "step": 23332 }, { "epoch": 4.28, "learning_rate": 9.357128023242434e-06, "loss": 0.0271, "step": 23333 }, { "epoch": 4.28, "learning_rate": 9.356387130751286e-06, "loss": 0.0266, "step": 23334 }, { "epoch": 4.28, "learning_rate": 9.355646241807744e-06, "loss": 0.0273, "step": 23335 }, { "epoch": 4.28, "learning_rate": 9.354905356415895e-06, "loss": 0.0275, "step": 23336 }, { "epoch": 4.28, "learning_rate": 9.354164474579822e-06, "loss": 0.0287, "step": 23337 }, { "epoch": 4.28, "learning_rate": 9.35342359630361e-06, "loss": 0.0234, "step": 23338 }, { "epoch": 4.28, "learning_rate": 9.352682721591342e-06, "loss": 0.0257, "step": 23339 }, { "epoch": 4.28, "learning_rate": 9.351941850447103e-06, "loss": 0.0258, "step": 23340 }, { "epoch": 4.28, "learning_rate": 9.351200982874973e-06, "loss": 0.0303, "step": 23341 }, { "epoch": 4.28, "learning_rate": 9.350460118879042e-06, "loss": 0.0233, "step": 23342 }, { "epoch": 4.28, "learning_rate": 9.349719258463391e-06, "loss": 0.0174, "step": 23343 }, { "epoch": 4.28, "learning_rate": 9.348978401632101e-06, "loss": 0.0274, "step": 23344 }, { "epoch": 4.28, "learning_rate": 9.348237548389255e-06, "loss": 0.0232, "step": 23345 }, { "epoch": 4.28, "learning_rate": 9.347496698738941e-06, "loss": 0.0238, "step": 23346 }, { "epoch": 4.28, "learning_rate": 9.346755852685239e-06, "loss": 0.0279, "step": 23347 }, { "epoch": 4.28, "learning_rate": 9.346015010232235e-06, "loss": 0.0265, "step": 23348 }, { "epoch": 4.28, "learning_rate": 9.345274171384016e-06, "loss": 0.0229, "step": 23349 }, { "epoch": 4.28, "learning_rate": 9.344533336144653e-06, "loss": 0.0276, "step": 23350 }, { "epoch": 4.28, "learning_rate": 9.343792504518241e-06, "loss": 0.0218, "step": 23351 }, { "epoch": 4.28, "learning_rate": 9.343051676508864e-06, "loss": 0.0306, "step": 23352 }, { "epoch": 4.28, "learning_rate": 9.342310852120597e-06, "loss": 0.0235, "step": 23353 }, { "epoch": 4.28, "learning_rate": 9.341570031357527e-06, "loss": 0.0268, "step": 23354 }, { "epoch": 4.28, "learning_rate": 9.340829214223741e-06, "loss": 0.0283, "step": 23355 }, { "epoch": 4.28, "learning_rate": 9.34008840072332e-06, "loss": 0.0242, "step": 23356 }, { "epoch": 4.28, "learning_rate": 9.339347590860347e-06, "loss": 0.0237, "step": 23357 }, { "epoch": 4.28, "learning_rate": 9.338606784638907e-06, "loss": 0.027, "step": 23358 }, { "epoch": 4.28, "learning_rate": 9.337865982063076e-06, "loss": 0.0218, "step": 23359 }, { "epoch": 4.28, "learning_rate": 9.337125183136949e-06, "loss": 0.0237, "step": 23360 }, { "epoch": 4.28, "learning_rate": 9.336384387864603e-06, "loss": 0.027, "step": 23361 }, { "epoch": 4.28, "learning_rate": 9.33564359625012e-06, "loss": 0.0241, "step": 23362 }, { "epoch": 4.28, "learning_rate": 9.334902808297585e-06, "loss": 0.0201, "step": 23363 }, { "epoch": 4.28, "learning_rate": 9.33416202401108e-06, "loss": 0.0249, "step": 23364 }, { "epoch": 4.28, "learning_rate": 9.333421243394692e-06, "loss": 0.0284, "step": 23365 }, { "epoch": 4.28, "learning_rate": 9.332680466452502e-06, "loss": 0.021, "step": 23366 }, { "epoch": 4.29, "learning_rate": 9.33193969318859e-06, "loss": 0.0306, "step": 23367 }, { "epoch": 4.29, "learning_rate": 9.331198923607044e-06, "loss": 0.0291, "step": 23368 }, { "epoch": 4.29, "learning_rate": 9.330458157711946e-06, "loss": 0.0241, "step": 23369 }, { "epoch": 4.29, "learning_rate": 9.32971739550738e-06, "loss": 0.0227, "step": 23370 }, { "epoch": 4.29, "learning_rate": 9.328976636997426e-06, "loss": 0.0296, "step": 23371 }, { "epoch": 4.29, "learning_rate": 9.328235882186165e-06, "loss": 0.0228, "step": 23372 }, { "epoch": 4.29, "learning_rate": 9.327495131077686e-06, "loss": 0.0275, "step": 23373 }, { "epoch": 4.29, "learning_rate": 9.326754383676072e-06, "loss": 0.0195, "step": 23374 }, { "epoch": 4.29, "learning_rate": 9.3260136399854e-06, "loss": 0.0282, "step": 23375 }, { "epoch": 4.29, "learning_rate": 9.325272900009756e-06, "loss": 0.0265, "step": 23376 }, { "epoch": 4.29, "learning_rate": 9.324532163753227e-06, "loss": 0.0212, "step": 23377 }, { "epoch": 4.29, "learning_rate": 9.32379143121989e-06, "loss": 0.0267, "step": 23378 }, { "epoch": 4.29, "learning_rate": 9.323050702413834e-06, "loss": 0.0217, "step": 23379 }, { "epoch": 4.29, "learning_rate": 9.322309977339137e-06, "loss": 0.0253, "step": 23380 }, { "epoch": 4.29, "learning_rate": 9.32156925599988e-06, "loss": 0.0336, "step": 23381 }, { "epoch": 4.29, "learning_rate": 9.32082853840015e-06, "loss": 0.0242, "step": 23382 }, { "epoch": 4.29, "learning_rate": 9.320087824544031e-06, "loss": 0.0234, "step": 23383 }, { "epoch": 4.29, "learning_rate": 9.319347114435604e-06, "loss": 0.0208, "step": 23384 }, { "epoch": 4.29, "learning_rate": 9.31860640807895e-06, "loss": 0.0314, "step": 23385 }, { "epoch": 4.29, "learning_rate": 9.317865705478153e-06, "loss": 0.0259, "step": 23386 }, { "epoch": 4.29, "learning_rate": 9.317125006637299e-06, "loss": 0.0255, "step": 23387 }, { "epoch": 4.29, "learning_rate": 9.316384311560469e-06, "loss": 0.0235, "step": 23388 }, { "epoch": 4.29, "learning_rate": 9.315643620251738e-06, "loss": 0.0258, "step": 23389 }, { "epoch": 4.29, "learning_rate": 9.314902932715202e-06, "loss": 0.0275, "step": 23390 }, { "epoch": 4.29, "learning_rate": 9.314162248954934e-06, "loss": 0.0238, "step": 23391 }, { "epoch": 4.29, "learning_rate": 9.313421568975021e-06, "loss": 0.0275, "step": 23392 }, { "epoch": 4.29, "learning_rate": 9.312680892779543e-06, "loss": 0.0191, "step": 23393 }, { "epoch": 4.29, "learning_rate": 9.311940220372585e-06, "loss": 0.0265, "step": 23394 }, { "epoch": 4.29, "learning_rate": 9.311199551758228e-06, "loss": 0.0234, "step": 23395 }, { "epoch": 4.29, "learning_rate": 9.310458886940557e-06, "loss": 0.0203, "step": 23396 }, { "epoch": 4.29, "learning_rate": 9.309718225923653e-06, "loss": 0.0258, "step": 23397 }, { "epoch": 4.29, "learning_rate": 9.30897756871159e-06, "loss": 0.0245, "step": 23398 }, { "epoch": 4.29, "learning_rate": 9.30823691530847e-06, "loss": 0.0239, "step": 23399 }, { "epoch": 4.29, "learning_rate": 9.30749626571836e-06, "loss": 0.0263, "step": 23400 }, { "epoch": 4.29, "learning_rate": 9.306755619945346e-06, "loss": 0.0261, "step": 23401 }, { "epoch": 4.29, "learning_rate": 9.306014977993513e-06, "loss": 0.0268, "step": 23402 }, { "epoch": 4.29, "learning_rate": 9.305274339866938e-06, "loss": 0.0276, "step": 23403 }, { "epoch": 4.29, "learning_rate": 9.304533705569711e-06, "loss": 0.0271, "step": 23404 }, { "epoch": 4.29, "learning_rate": 9.303793075105911e-06, "loss": 0.0232, "step": 23405 }, { "epoch": 4.29, "learning_rate": 9.30305244847962e-06, "loss": 0.0169, "step": 23406 }, { "epoch": 4.29, "learning_rate": 9.302311825694912e-06, "loss": 0.0265, "step": 23407 }, { "epoch": 4.29, "learning_rate": 9.301571206755887e-06, "loss": 0.0215, "step": 23408 }, { "epoch": 4.29, "learning_rate": 9.300830591666615e-06, "loss": 0.0264, "step": 23409 }, { "epoch": 4.29, "learning_rate": 9.300089980431181e-06, "loss": 0.0283, "step": 23410 }, { "epoch": 4.29, "learning_rate": 9.299349373053667e-06, "loss": 0.0268, "step": 23411 }, { "epoch": 4.29, "learning_rate": 9.298608769538155e-06, "loss": 0.0282, "step": 23412 }, { "epoch": 4.29, "learning_rate": 9.29786816988873e-06, "loss": 0.0218, "step": 23413 }, { "epoch": 4.29, "learning_rate": 9.297127574109472e-06, "loss": 0.026, "step": 23414 }, { "epoch": 4.29, "learning_rate": 9.296386982204464e-06, "loss": 0.0219, "step": 23415 }, { "epoch": 4.29, "learning_rate": 9.29564639417778e-06, "loss": 0.0204, "step": 23416 }, { "epoch": 4.29, "learning_rate": 9.294905810033518e-06, "loss": 0.0294, "step": 23417 }, { "epoch": 4.29, "learning_rate": 9.29416522977575e-06, "loss": 0.0238, "step": 23418 }, { "epoch": 4.29, "learning_rate": 9.29342465340856e-06, "loss": 0.0342, "step": 23419 }, { "epoch": 4.29, "learning_rate": 9.292684080936027e-06, "loss": 0.0232, "step": 23420 }, { "epoch": 4.3, "learning_rate": 9.291943512362236e-06, "loss": 0.023, "step": 23421 }, { "epoch": 4.3, "learning_rate": 9.291202947691272e-06, "loss": 0.0244, "step": 23422 }, { "epoch": 4.3, "learning_rate": 9.290462386927215e-06, "loss": 0.0292, "step": 23423 }, { "epoch": 4.3, "learning_rate": 9.289721830074142e-06, "loss": 0.0256, "step": 23424 }, { "epoch": 4.3, "learning_rate": 9.288981277136138e-06, "loss": 0.0278, "step": 23425 }, { "epoch": 4.3, "learning_rate": 9.28824072811729e-06, "loss": 0.0242, "step": 23426 }, { "epoch": 4.3, "learning_rate": 9.287500183021676e-06, "loss": 0.024, "step": 23427 }, { "epoch": 4.3, "learning_rate": 9.286759641853375e-06, "loss": 0.0266, "step": 23428 }, { "epoch": 4.3, "learning_rate": 9.28601910461647e-06, "loss": 0.0251, "step": 23429 }, { "epoch": 4.3, "learning_rate": 9.285278571315046e-06, "loss": 0.0252, "step": 23430 }, { "epoch": 4.3, "learning_rate": 9.284538041953186e-06, "loss": 0.025, "step": 23431 }, { "epoch": 4.3, "learning_rate": 9.283797516534968e-06, "loss": 0.0258, "step": 23432 }, { "epoch": 4.3, "learning_rate": 9.283056995064475e-06, "loss": 0.0259, "step": 23433 }, { "epoch": 4.3, "learning_rate": 9.282316477545783e-06, "loss": 0.0304, "step": 23434 }, { "epoch": 4.3, "learning_rate": 9.281575963982985e-06, "loss": 0.0243, "step": 23435 }, { "epoch": 4.3, "learning_rate": 9.280835454380157e-06, "loss": 0.0246, "step": 23436 }, { "epoch": 4.3, "learning_rate": 9.28009494874138e-06, "loss": 0.0265, "step": 23437 }, { "epoch": 4.3, "learning_rate": 9.279354447070736e-06, "loss": 0.0207, "step": 23438 }, { "epoch": 4.3, "learning_rate": 9.278613949372307e-06, "loss": 0.0215, "step": 23439 }, { "epoch": 4.3, "learning_rate": 9.277873455650175e-06, "loss": 0.022, "step": 23440 }, { "epoch": 4.3, "learning_rate": 9.277132965908425e-06, "loss": 0.0287, "step": 23441 }, { "epoch": 4.3, "learning_rate": 9.276392480151127e-06, "loss": 0.023, "step": 23442 }, { "epoch": 4.3, "learning_rate": 9.275651998382377e-06, "loss": 0.0253, "step": 23443 }, { "epoch": 4.3, "learning_rate": 9.27491152060625e-06, "loss": 0.0229, "step": 23444 }, { "epoch": 4.3, "learning_rate": 9.274171046826826e-06, "loss": 0.0244, "step": 23445 }, { "epoch": 4.3, "learning_rate": 9.273430577048189e-06, "loss": 0.0237, "step": 23446 }, { "epoch": 4.3, "learning_rate": 9.272690111274418e-06, "loss": 0.0239, "step": 23447 }, { "epoch": 4.3, "learning_rate": 9.271949649509596e-06, "loss": 0.0214, "step": 23448 }, { "epoch": 4.3, "learning_rate": 9.271209191757807e-06, "loss": 0.0252, "step": 23449 }, { "epoch": 4.3, "learning_rate": 9.27046873802313e-06, "loss": 0.024, "step": 23450 }, { "epoch": 4.3, "learning_rate": 9.269728288309641e-06, "loss": 0.0273, "step": 23451 }, { "epoch": 4.3, "learning_rate": 9.268987842621434e-06, "loss": 0.0225, "step": 23452 }, { "epoch": 4.3, "learning_rate": 9.26824740096258e-06, "loss": 0.027, "step": 23453 }, { "epoch": 4.3, "learning_rate": 9.267506963337165e-06, "loss": 0.0286, "step": 23454 }, { "epoch": 4.3, "learning_rate": 9.266766529749267e-06, "loss": 0.0234, "step": 23455 }, { "epoch": 4.3, "learning_rate": 9.266026100202968e-06, "loss": 0.0261, "step": 23456 }, { "epoch": 4.3, "learning_rate": 9.265285674702352e-06, "loss": 0.0229, "step": 23457 }, { "epoch": 4.3, "learning_rate": 9.264545253251498e-06, "loss": 0.0345, "step": 23458 }, { "epoch": 4.3, "learning_rate": 9.263804835854491e-06, "loss": 0.0249, "step": 23459 }, { "epoch": 4.3, "learning_rate": 9.263064422515402e-06, "loss": 0.0296, "step": 23460 }, { "epoch": 4.3, "learning_rate": 9.262324013238327e-06, "loss": 0.026, "step": 23461 }, { "epoch": 4.3, "learning_rate": 9.261583608027336e-06, "loss": 0.0243, "step": 23462 }, { "epoch": 4.3, "learning_rate": 9.260843206886512e-06, "loss": 0.0219, "step": 23463 }, { "epoch": 4.3, "learning_rate": 9.260102809819939e-06, "loss": 0.0316, "step": 23464 }, { "epoch": 4.3, "learning_rate": 9.259362416831696e-06, "loss": 0.0238, "step": 23465 }, { "epoch": 4.3, "learning_rate": 9.258622027925865e-06, "loss": 0.0242, "step": 23466 }, { "epoch": 4.3, "learning_rate": 9.257881643106527e-06, "loss": 0.0231, "step": 23467 }, { "epoch": 4.3, "learning_rate": 9.257141262377765e-06, "loss": 0.0226, "step": 23468 }, { "epoch": 4.3, "learning_rate": 9.25640088574365e-06, "loss": 0.033, "step": 23469 }, { "epoch": 4.3, "learning_rate": 9.255660513208278e-06, "loss": 0.0239, "step": 23470 }, { "epoch": 4.3, "learning_rate": 9.254920144775722e-06, "loss": 0.0271, "step": 23471 }, { "epoch": 4.3, "learning_rate": 9.254179780450062e-06, "loss": 0.0272, "step": 23472 }, { "epoch": 4.3, "learning_rate": 9.25343942023538e-06, "loss": 0.0246, "step": 23473 }, { "epoch": 4.3, "learning_rate": 9.252699064135759e-06, "loss": 0.0256, "step": 23474 }, { "epoch": 4.3, "learning_rate": 9.251958712155277e-06, "loss": 0.0243, "step": 23475 }, { "epoch": 4.31, "learning_rate": 9.251218364298016e-06, "loss": 0.0222, "step": 23476 }, { "epoch": 4.31, "learning_rate": 9.250478020568061e-06, "loss": 0.03, "step": 23477 }, { "epoch": 4.31, "learning_rate": 9.24973768096948e-06, "loss": 0.0259, "step": 23478 }, { "epoch": 4.31, "learning_rate": 9.248997345506371e-06, "loss": 0.0304, "step": 23479 }, { "epoch": 4.31, "learning_rate": 9.248257014182803e-06, "loss": 0.0238, "step": 23480 }, { "epoch": 4.31, "learning_rate": 9.247516687002861e-06, "loss": 0.0254, "step": 23481 }, { "epoch": 4.31, "learning_rate": 9.24677636397062e-06, "loss": 0.0265, "step": 23482 }, { "epoch": 4.31, "learning_rate": 9.24603604509017e-06, "loss": 0.0297, "step": 23483 }, { "epoch": 4.31, "learning_rate": 9.245295730365586e-06, "loss": 0.023, "step": 23484 }, { "epoch": 4.31, "learning_rate": 9.24455541980095e-06, "loss": 0.0212, "step": 23485 }, { "epoch": 4.31, "learning_rate": 9.243815113400345e-06, "loss": 0.0246, "step": 23486 }, { "epoch": 4.31, "learning_rate": 9.243074811167841e-06, "loss": 0.0218, "step": 23487 }, { "epoch": 4.31, "learning_rate": 9.242334513107534e-06, "loss": 0.027, "step": 23488 }, { "epoch": 4.31, "learning_rate": 9.241594219223494e-06, "loss": 0.022, "step": 23489 }, { "epoch": 4.31, "learning_rate": 9.240853929519804e-06, "loss": 0.0304, "step": 23490 }, { "epoch": 4.31, "learning_rate": 9.240113644000543e-06, "loss": 0.0363, "step": 23491 }, { "epoch": 4.31, "learning_rate": 9.239373362669797e-06, "loss": 0.0269, "step": 23492 }, { "epoch": 4.31, "learning_rate": 9.23863308553164e-06, "loss": 0.0243, "step": 23493 }, { "epoch": 4.31, "learning_rate": 9.237892812590158e-06, "loss": 0.0213, "step": 23494 }, { "epoch": 4.31, "learning_rate": 9.237152543849426e-06, "loss": 0.0257, "step": 23495 }, { "epoch": 4.31, "learning_rate": 9.236412279313529e-06, "loss": 0.0249, "step": 23496 }, { "epoch": 4.31, "learning_rate": 9.235672018986547e-06, "loss": 0.0263, "step": 23497 }, { "epoch": 4.31, "learning_rate": 9.234931762872557e-06, "loss": 0.0229, "step": 23498 }, { "epoch": 4.31, "learning_rate": 9.23419151097564e-06, "loss": 0.0212, "step": 23499 }, { "epoch": 4.31, "learning_rate": 9.233451263299876e-06, "loss": 0.0239, "step": 23500 }, { "epoch": 4.31, "learning_rate": 9.232711019849349e-06, "loss": 0.023, "step": 23501 }, { "epoch": 4.31, "learning_rate": 9.231970780628136e-06, "loss": 0.0233, "step": 23502 }, { "epoch": 4.31, "learning_rate": 9.23123054564032e-06, "loss": 0.0223, "step": 23503 }, { "epoch": 4.31, "learning_rate": 9.230490314889975e-06, "loss": 0.0303, "step": 23504 }, { "epoch": 4.31, "learning_rate": 9.229750088381189e-06, "loss": 0.0196, "step": 23505 }, { "epoch": 4.31, "learning_rate": 9.22900986611804e-06, "loss": 0.0249, "step": 23506 }, { "epoch": 4.31, "learning_rate": 9.228269648104605e-06, "loss": 0.0267, "step": 23507 }, { "epoch": 4.31, "learning_rate": 9.227529434344963e-06, "loss": 0.0227, "step": 23508 }, { "epoch": 4.31, "learning_rate": 9.226789224843198e-06, "loss": 0.0279, "step": 23509 }, { "epoch": 4.31, "learning_rate": 9.22604901960339e-06, "loss": 0.0287, "step": 23510 }, { "epoch": 4.31, "learning_rate": 9.22530881862962e-06, "loss": 0.0234, "step": 23511 }, { "epoch": 4.31, "learning_rate": 9.224568621925965e-06, "loss": 0.0214, "step": 23512 }, { "epoch": 4.31, "learning_rate": 9.2238284294965e-06, "loss": 0.0246, "step": 23513 }, { "epoch": 4.31, "learning_rate": 9.223088241345317e-06, "loss": 0.0238, "step": 23514 }, { "epoch": 4.31, "learning_rate": 9.22234805747649e-06, "loss": 0.0264, "step": 23515 }, { "epoch": 4.31, "learning_rate": 9.221607877894097e-06, "loss": 0.0264, "step": 23516 }, { "epoch": 4.31, "learning_rate": 9.220867702602217e-06, "loss": 0.0233, "step": 23517 }, { "epoch": 4.31, "learning_rate": 9.220127531604936e-06, "loss": 0.0226, "step": 23518 }, { "epoch": 4.31, "learning_rate": 9.21938736490633e-06, "loss": 0.0273, "step": 23519 }, { "epoch": 4.31, "learning_rate": 9.218647202510479e-06, "loss": 0.0317, "step": 23520 }, { "epoch": 4.31, "learning_rate": 9.217907044421465e-06, "loss": 0.023, "step": 23521 }, { "epoch": 4.31, "learning_rate": 9.21716689064336e-06, "loss": 0.0258, "step": 23522 }, { "epoch": 4.31, "learning_rate": 9.216426741180253e-06, "loss": 0.0249, "step": 23523 }, { "epoch": 4.31, "learning_rate": 9.215686596036221e-06, "loss": 0.0244, "step": 23524 }, { "epoch": 4.31, "learning_rate": 9.214946455215343e-06, "loss": 0.0195, "step": 23525 }, { "epoch": 4.31, "learning_rate": 9.214206318721695e-06, "loss": 0.0264, "step": 23526 }, { "epoch": 4.31, "learning_rate": 9.213466186559362e-06, "loss": 0.0264, "step": 23527 }, { "epoch": 4.31, "learning_rate": 9.212726058732423e-06, "loss": 0.0214, "step": 23528 }, { "epoch": 4.31, "learning_rate": 9.211985935244956e-06, "loss": 0.0258, "step": 23529 }, { "epoch": 4.32, "learning_rate": 9.211245816101043e-06, "loss": 0.025, "step": 23530 }, { "epoch": 4.32, "learning_rate": 9.210505701304755e-06, "loss": 0.0282, "step": 23531 }, { "epoch": 4.32, "learning_rate": 9.209765590860183e-06, "loss": 0.0262, "step": 23532 }, { "epoch": 4.32, "learning_rate": 9.209025484771402e-06, "loss": 0.0223, "step": 23533 }, { "epoch": 4.32, "learning_rate": 9.20828538304249e-06, "loss": 0.0244, "step": 23534 }, { "epoch": 4.32, "learning_rate": 9.207545285677526e-06, "loss": 0.0215, "step": 23535 }, { "epoch": 4.32, "learning_rate": 9.206805192680592e-06, "loss": 0.0273, "step": 23536 }, { "epoch": 4.32, "learning_rate": 9.206065104055767e-06, "loss": 0.0289, "step": 23537 }, { "epoch": 4.32, "learning_rate": 9.20532501980713e-06, "loss": 0.0225, "step": 23538 }, { "epoch": 4.32, "learning_rate": 9.204584939938763e-06, "loss": 0.0255, "step": 23539 }, { "epoch": 4.32, "learning_rate": 9.203844864454735e-06, "loss": 0.0233, "step": 23540 }, { "epoch": 4.32, "learning_rate": 9.203104793359137e-06, "loss": 0.0263, "step": 23541 }, { "epoch": 4.32, "learning_rate": 9.202364726656045e-06, "loss": 0.0267, "step": 23542 }, { "epoch": 4.32, "learning_rate": 9.201624664349537e-06, "loss": 0.0232, "step": 23543 }, { "epoch": 4.32, "learning_rate": 9.200884606443689e-06, "loss": 0.0244, "step": 23544 }, { "epoch": 4.32, "learning_rate": 9.200144552942587e-06, "loss": 0.0257, "step": 23545 }, { "epoch": 4.32, "learning_rate": 9.199404503850306e-06, "loss": 0.031, "step": 23546 }, { "epoch": 4.32, "learning_rate": 9.198664459170927e-06, "loss": 0.0258, "step": 23547 }, { "epoch": 4.32, "learning_rate": 9.197924418908526e-06, "loss": 0.0234, "step": 23548 }, { "epoch": 4.32, "learning_rate": 9.197184383067186e-06, "loss": 0.0195, "step": 23549 }, { "epoch": 4.32, "learning_rate": 9.196444351650989e-06, "loss": 0.0306, "step": 23550 }, { "epoch": 4.32, "learning_rate": 9.195704324664005e-06, "loss": 0.0257, "step": 23551 }, { "epoch": 4.32, "learning_rate": 9.194964302110317e-06, "loss": 0.0238, "step": 23552 }, { "epoch": 4.32, "learning_rate": 9.194224283994004e-06, "loss": 0.0235, "step": 23553 }, { "epoch": 4.32, "learning_rate": 9.193484270319148e-06, "loss": 0.0277, "step": 23554 }, { "epoch": 4.32, "learning_rate": 9.192744261089825e-06, "loss": 0.0195, "step": 23555 }, { "epoch": 4.32, "learning_rate": 9.192004256310114e-06, "loss": 0.0226, "step": 23556 }, { "epoch": 4.32, "learning_rate": 9.191264255984093e-06, "loss": 0.0281, "step": 23557 }, { "epoch": 4.32, "learning_rate": 9.190524260115844e-06, "loss": 0.0222, "step": 23558 }, { "epoch": 4.32, "learning_rate": 9.189784268709447e-06, "loss": 0.0236, "step": 23559 }, { "epoch": 4.32, "learning_rate": 9.189044281768976e-06, "loss": 0.0188, "step": 23560 }, { "epoch": 4.32, "learning_rate": 9.18830429929851e-06, "loss": 0.0212, "step": 23561 }, { "epoch": 4.32, "learning_rate": 9.187564321302131e-06, "loss": 0.0251, "step": 23562 }, { "epoch": 4.32, "learning_rate": 9.186824347783916e-06, "loss": 0.0196, "step": 23563 }, { "epoch": 4.32, "learning_rate": 9.186084378747944e-06, "loss": 0.0226, "step": 23564 }, { "epoch": 4.32, "learning_rate": 9.185344414198297e-06, "loss": 0.0279, "step": 23565 }, { "epoch": 4.32, "learning_rate": 9.184604454139045e-06, "loss": 0.0222, "step": 23566 }, { "epoch": 4.32, "learning_rate": 9.183864498574276e-06, "loss": 0.0247, "step": 23567 }, { "epoch": 4.32, "learning_rate": 9.183124547508068e-06, "loss": 0.0272, "step": 23568 }, { "epoch": 4.32, "learning_rate": 9.182384600944494e-06, "loss": 0.0236, "step": 23569 }, { "epoch": 4.32, "learning_rate": 9.18164465888763e-06, "loss": 0.0214, "step": 23570 }, { "epoch": 4.32, "learning_rate": 9.180904721341567e-06, "loss": 0.0243, "step": 23571 }, { "epoch": 4.32, "learning_rate": 9.180164788310374e-06, "loss": 0.0203, "step": 23572 }, { "epoch": 4.32, "learning_rate": 9.179424859798131e-06, "loss": 0.0267, "step": 23573 }, { "epoch": 4.32, "learning_rate": 9.178684935808919e-06, "loss": 0.0274, "step": 23574 }, { "epoch": 4.32, "learning_rate": 9.177945016346811e-06, "loss": 0.0277, "step": 23575 }, { "epoch": 4.32, "learning_rate": 9.177205101415893e-06, "loss": 0.035, "step": 23576 }, { "epoch": 4.32, "learning_rate": 9.176465191020243e-06, "loss": 0.0288, "step": 23577 }, { "epoch": 4.32, "learning_rate": 9.175725285163932e-06, "loss": 0.0261, "step": 23578 }, { "epoch": 4.32, "learning_rate": 9.174985383851039e-06, "loss": 0.0259, "step": 23579 }, { "epoch": 4.32, "learning_rate": 9.174245487085652e-06, "loss": 0.0248, "step": 23580 }, { "epoch": 4.32, "learning_rate": 9.173505594871842e-06, "loss": 0.0191, "step": 23581 }, { "epoch": 4.32, "learning_rate": 9.172765707213687e-06, "loss": 0.0239, "step": 23582 }, { "epoch": 4.32, "learning_rate": 9.172025824115268e-06, "loss": 0.0233, "step": 23583 }, { "epoch": 4.32, "learning_rate": 9.171285945580659e-06, "loss": 0.0236, "step": 23584 }, { "epoch": 4.33, "learning_rate": 9.170546071613943e-06, "loss": 0.0169, "step": 23585 }, { "epoch": 4.33, "learning_rate": 9.169806202219199e-06, "loss": 0.0252, "step": 23586 }, { "epoch": 4.33, "learning_rate": 9.169066337400501e-06, "loss": 0.0237, "step": 23587 }, { "epoch": 4.33, "learning_rate": 9.168326477161926e-06, "loss": 0.0249, "step": 23588 }, { "epoch": 4.33, "learning_rate": 9.167586621507559e-06, "loss": 0.0248, "step": 23589 }, { "epoch": 4.33, "learning_rate": 9.166846770441473e-06, "loss": 0.0183, "step": 23590 }, { "epoch": 4.33, "learning_rate": 9.166106923967747e-06, "loss": 0.0264, "step": 23591 }, { "epoch": 4.33, "learning_rate": 9.165367082090459e-06, "loss": 0.0246, "step": 23592 }, { "epoch": 4.33, "learning_rate": 9.164627244813685e-06, "loss": 0.0251, "step": 23593 }, { "epoch": 4.33, "learning_rate": 9.163887412141507e-06, "loss": 0.0282, "step": 23594 }, { "epoch": 4.33, "learning_rate": 9.163147584078003e-06, "loss": 0.0276, "step": 23595 }, { "epoch": 4.33, "learning_rate": 9.162407760627248e-06, "loss": 0.0292, "step": 23596 }, { "epoch": 4.33, "learning_rate": 9.16166794179332e-06, "loss": 0.0253, "step": 23597 }, { "epoch": 4.33, "learning_rate": 9.160928127580298e-06, "loss": 0.025, "step": 23598 }, { "epoch": 4.33, "learning_rate": 9.16018831799226e-06, "loss": 0.024, "step": 23599 }, { "epoch": 4.33, "learning_rate": 9.159448513033284e-06, "loss": 0.0273, "step": 23600 }, { "epoch": 4.33, "learning_rate": 9.158708712707446e-06, "loss": 0.0177, "step": 23601 }, { "epoch": 4.33, "learning_rate": 9.157968917018826e-06, "loss": 0.0213, "step": 23602 }, { "epoch": 4.33, "learning_rate": 9.157229125971503e-06, "loss": 0.0238, "step": 23603 }, { "epoch": 4.33, "learning_rate": 9.156489339569555e-06, "loss": 0.0238, "step": 23604 }, { "epoch": 4.33, "learning_rate": 9.155749557817054e-06, "loss": 0.0235, "step": 23605 }, { "epoch": 4.33, "learning_rate": 9.15500978071808e-06, "loss": 0.025, "step": 23606 }, { "epoch": 4.33, "learning_rate": 9.154270008276712e-06, "loss": 0.0241, "step": 23607 }, { "epoch": 4.33, "learning_rate": 9.15353024049703e-06, "loss": 0.0282, "step": 23608 }, { "epoch": 4.33, "learning_rate": 9.152790477383108e-06, "loss": 0.0222, "step": 23609 }, { "epoch": 4.33, "learning_rate": 9.152050718939023e-06, "loss": 0.0307, "step": 23610 }, { "epoch": 4.33, "learning_rate": 9.151310965168857e-06, "loss": 0.0256, "step": 23611 }, { "epoch": 4.33, "learning_rate": 9.150571216076683e-06, "loss": 0.0252, "step": 23612 }, { "epoch": 4.33, "learning_rate": 9.149831471666583e-06, "loss": 0.0215, "step": 23613 }, { "epoch": 4.33, "learning_rate": 9.149091731942629e-06, "loss": 0.0245, "step": 23614 }, { "epoch": 4.33, "learning_rate": 9.148351996908901e-06, "loss": 0.0247, "step": 23615 }, { "epoch": 4.33, "learning_rate": 9.14761226656948e-06, "loss": 0.0289, "step": 23616 }, { "epoch": 4.33, "learning_rate": 9.146872540928437e-06, "loss": 0.0238, "step": 23617 }, { "epoch": 4.33, "learning_rate": 9.146132819989855e-06, "loss": 0.022, "step": 23618 }, { "epoch": 4.33, "learning_rate": 9.145393103757805e-06, "loss": 0.0258, "step": 23619 }, { "epoch": 4.33, "learning_rate": 9.14465339223637e-06, "loss": 0.0266, "step": 23620 }, { "epoch": 4.33, "learning_rate": 9.143913685429628e-06, "loss": 0.021, "step": 23621 }, { "epoch": 4.33, "learning_rate": 9.143173983341655e-06, "loss": 0.0269, "step": 23622 }, { "epoch": 4.33, "learning_rate": 9.14243428597652e-06, "loss": 0.0267, "step": 23623 }, { "epoch": 4.33, "learning_rate": 9.141694593338314e-06, "loss": 0.0198, "step": 23624 }, { "epoch": 4.33, "learning_rate": 9.140954905431105e-06, "loss": 0.03, "step": 23625 }, { "epoch": 4.33, "learning_rate": 9.140215222258974e-06, "loss": 0.0302, "step": 23626 }, { "epoch": 4.33, "learning_rate": 9.139475543825995e-06, "loss": 0.0258, "step": 23627 }, { "epoch": 4.33, "learning_rate": 9.138735870136245e-06, "loss": 0.0252, "step": 23628 }, { "epoch": 4.33, "learning_rate": 9.137996201193807e-06, "loss": 0.0222, "step": 23629 }, { "epoch": 4.33, "learning_rate": 9.137256537002752e-06, "loss": 0.0226, "step": 23630 }, { "epoch": 4.33, "learning_rate": 9.136516877567164e-06, "loss": 0.0292, "step": 23631 }, { "epoch": 4.33, "learning_rate": 9.135777222891106e-06, "loss": 0.0222, "step": 23632 }, { "epoch": 4.33, "learning_rate": 9.135037572978672e-06, "loss": 0.0266, "step": 23633 }, { "epoch": 4.33, "learning_rate": 9.134297927833929e-06, "loss": 0.0209, "step": 23634 }, { "epoch": 4.33, "learning_rate": 9.133558287460955e-06, "loss": 0.0233, "step": 23635 }, { "epoch": 4.33, "learning_rate": 9.13281865186383e-06, "loss": 0.0255, "step": 23636 }, { "epoch": 4.33, "learning_rate": 9.132079021046625e-06, "loss": 0.0255, "step": 23637 }, { "epoch": 4.33, "learning_rate": 9.131339395013424e-06, "loss": 0.0233, "step": 23638 }, { "epoch": 4.34, "learning_rate": 9.1305997737683e-06, "loss": 0.0307, "step": 23639 }, { "epoch": 4.34, "learning_rate": 9.129860157315332e-06, "loss": 0.0234, "step": 23640 }, { "epoch": 4.34, "learning_rate": 9.129120545658589e-06, "loss": 0.02, "step": 23641 }, { "epoch": 4.34, "learning_rate": 9.128380938802161e-06, "loss": 0.0207, "step": 23642 }, { "epoch": 4.34, "learning_rate": 9.127641336750116e-06, "loss": 0.0241, "step": 23643 }, { "epoch": 4.34, "learning_rate": 9.126901739506531e-06, "loss": 0.0238, "step": 23644 }, { "epoch": 4.34, "learning_rate": 9.126162147075484e-06, "loss": 0.0293, "step": 23645 }, { "epoch": 4.34, "learning_rate": 9.125422559461053e-06, "loss": 0.0175, "step": 23646 }, { "epoch": 4.34, "learning_rate": 9.124682976667313e-06, "loss": 0.0288, "step": 23647 }, { "epoch": 4.34, "learning_rate": 9.12394339869834e-06, "loss": 0.032, "step": 23648 }, { "epoch": 4.34, "learning_rate": 9.123203825558215e-06, "loss": 0.0245, "step": 23649 }, { "epoch": 4.34, "learning_rate": 9.122464257251004e-06, "loss": 0.0233, "step": 23650 }, { "epoch": 4.34, "learning_rate": 9.121724693780797e-06, "loss": 0.0256, "step": 23651 }, { "epoch": 4.34, "learning_rate": 9.120985135151662e-06, "loss": 0.0245, "step": 23652 }, { "epoch": 4.34, "learning_rate": 9.120245581367678e-06, "loss": 0.0267, "step": 23653 }, { "epoch": 4.34, "learning_rate": 9.119506032432918e-06, "loss": 0.0246, "step": 23654 }, { "epoch": 4.34, "learning_rate": 9.118766488351466e-06, "loss": 0.0225, "step": 23655 }, { "epoch": 4.34, "learning_rate": 9.11802694912739e-06, "loss": 0.0309, "step": 23656 }, { "epoch": 4.34, "learning_rate": 9.117287414764771e-06, "loss": 0.024, "step": 23657 }, { "epoch": 4.34, "learning_rate": 9.116547885267688e-06, "loss": 0.0278, "step": 23658 }, { "epoch": 4.34, "learning_rate": 9.115808360640207e-06, "loss": 0.0249, "step": 23659 }, { "epoch": 4.34, "learning_rate": 9.115068840886418e-06, "loss": 0.0192, "step": 23660 }, { "epoch": 4.34, "learning_rate": 9.114329326010387e-06, "loss": 0.0247, "step": 23661 }, { "epoch": 4.34, "learning_rate": 9.113589816016193e-06, "loss": 0.025, "step": 23662 }, { "epoch": 4.34, "learning_rate": 9.112850310907911e-06, "loss": 0.0268, "step": 23663 }, { "epoch": 4.34, "learning_rate": 9.112110810689622e-06, "loss": 0.024, "step": 23664 }, { "epoch": 4.34, "learning_rate": 9.111371315365399e-06, "loss": 0.025, "step": 23665 }, { "epoch": 4.34, "learning_rate": 9.11063182493932e-06, "loss": 0.0311, "step": 23666 }, { "epoch": 4.34, "learning_rate": 9.109892339415455e-06, "loss": 0.0259, "step": 23667 }, { "epoch": 4.34, "learning_rate": 9.109152858797882e-06, "loss": 0.0249, "step": 23668 }, { "epoch": 4.34, "learning_rate": 9.108413383090686e-06, "loss": 0.0256, "step": 23669 }, { "epoch": 4.34, "learning_rate": 9.107673912297934e-06, "loss": 0.0257, "step": 23670 }, { "epoch": 4.34, "learning_rate": 9.106934446423704e-06, "loss": 0.0244, "step": 23671 }, { "epoch": 4.34, "learning_rate": 9.10619498547207e-06, "loss": 0.0214, "step": 23672 }, { "epoch": 4.34, "learning_rate": 9.105455529447113e-06, "loss": 0.0265, "step": 23673 }, { "epoch": 4.34, "learning_rate": 9.104716078352906e-06, "loss": 0.0269, "step": 23674 }, { "epoch": 4.34, "learning_rate": 9.103976632193527e-06, "loss": 0.0232, "step": 23675 }, { "epoch": 4.34, "learning_rate": 9.103237190973042e-06, "loss": 0.0254, "step": 23676 }, { "epoch": 4.34, "learning_rate": 9.10249775469554e-06, "loss": 0.0338, "step": 23677 }, { "epoch": 4.34, "learning_rate": 9.101758323365094e-06, "loss": 0.0215, "step": 23678 }, { "epoch": 4.34, "learning_rate": 9.101018896985774e-06, "loss": 0.0276, "step": 23679 }, { "epoch": 4.34, "learning_rate": 9.100279475561658e-06, "loss": 0.0281, "step": 23680 }, { "epoch": 4.34, "learning_rate": 9.099540059096822e-06, "loss": 0.0279, "step": 23681 }, { "epoch": 4.34, "learning_rate": 9.098800647595345e-06, "loss": 0.0272, "step": 23682 }, { "epoch": 4.34, "learning_rate": 9.0980612410613e-06, "loss": 0.029, "step": 23683 }, { "epoch": 4.34, "learning_rate": 9.097321839498763e-06, "loss": 0.0232, "step": 23684 }, { "epoch": 4.34, "learning_rate": 9.096582442911803e-06, "loss": 0.0255, "step": 23685 }, { "epoch": 4.34, "learning_rate": 9.095843051304507e-06, "loss": 0.0249, "step": 23686 }, { "epoch": 4.34, "learning_rate": 9.095103664680947e-06, "loss": 0.025, "step": 23687 }, { "epoch": 4.34, "learning_rate": 9.094364283045196e-06, "loss": 0.031, "step": 23688 }, { "epoch": 4.34, "learning_rate": 9.093624906401328e-06, "loss": 0.0256, "step": 23689 }, { "epoch": 4.34, "learning_rate": 9.092885534753419e-06, "loss": 0.0214, "step": 23690 }, { "epoch": 4.34, "learning_rate": 9.09214616810555e-06, "loss": 0.0244, "step": 23691 }, { "epoch": 4.34, "learning_rate": 9.09140680646179e-06, "loss": 0.0228, "step": 23692 }, { "epoch": 4.34, "learning_rate": 9.090667449826222e-06, "loss": 0.0246, "step": 23693 }, { "epoch": 4.35, "learning_rate": 9.089928098202909e-06, "loss": 0.0272, "step": 23694 }, { "epoch": 4.35, "learning_rate": 9.089188751595937e-06, "loss": 0.0232, "step": 23695 }, { "epoch": 4.35, "learning_rate": 9.08844941000938e-06, "loss": 0.0275, "step": 23696 }, { "epoch": 4.35, "learning_rate": 9.08771007344731e-06, "loss": 0.0218, "step": 23697 }, { "epoch": 4.35, "learning_rate": 9.086970741913801e-06, "loss": 0.0279, "step": 23698 }, { "epoch": 4.35, "learning_rate": 9.086231415412932e-06, "loss": 0.0262, "step": 23699 }, { "epoch": 4.35, "learning_rate": 9.085492093948777e-06, "loss": 0.024, "step": 23700 }, { "epoch": 4.35, "learning_rate": 9.084752777525414e-06, "loss": 0.027, "step": 23701 }, { "epoch": 4.35, "learning_rate": 9.084013466146914e-06, "loss": 0.0249, "step": 23702 }, { "epoch": 4.35, "learning_rate": 9.083274159817349e-06, "loss": 0.0289, "step": 23703 }, { "epoch": 4.35, "learning_rate": 9.082534858540799e-06, "loss": 0.0243, "step": 23704 }, { "epoch": 4.35, "learning_rate": 9.081795562321345e-06, "loss": 0.0237, "step": 23705 }, { "epoch": 4.35, "learning_rate": 9.08105627116305e-06, "loss": 0.0302, "step": 23706 }, { "epoch": 4.35, "learning_rate": 9.080316985069992e-06, "loss": 0.0258, "step": 23707 }, { "epoch": 4.35, "learning_rate": 9.079577704046251e-06, "loss": 0.024, "step": 23708 }, { "epoch": 4.35, "learning_rate": 9.0788384280959e-06, "loss": 0.0301, "step": 23709 }, { "epoch": 4.35, "learning_rate": 9.078099157223014e-06, "loss": 0.0263, "step": 23710 }, { "epoch": 4.35, "learning_rate": 9.077359891431668e-06, "loss": 0.029, "step": 23711 }, { "epoch": 4.35, "learning_rate": 9.07662063072593e-06, "loss": 0.0228, "step": 23712 }, { "epoch": 4.35, "learning_rate": 9.075881375109888e-06, "loss": 0.0287, "step": 23713 }, { "epoch": 4.35, "learning_rate": 9.075142124587607e-06, "loss": 0.0264, "step": 23714 }, { "epoch": 4.35, "learning_rate": 9.074402879163163e-06, "loss": 0.0232, "step": 23715 }, { "epoch": 4.35, "learning_rate": 9.073663638840629e-06, "loss": 0.0252, "step": 23716 }, { "epoch": 4.35, "learning_rate": 9.072924403624088e-06, "loss": 0.0228, "step": 23717 }, { "epoch": 4.35, "learning_rate": 9.072185173517606e-06, "loss": 0.0292, "step": 23718 }, { "epoch": 4.35, "learning_rate": 9.071445948525262e-06, "loss": 0.0216, "step": 23719 }, { "epoch": 4.35, "learning_rate": 9.070706728651134e-06, "loss": 0.0306, "step": 23720 }, { "epoch": 4.35, "learning_rate": 9.069967513899284e-06, "loss": 0.0233, "step": 23721 }, { "epoch": 4.35, "learning_rate": 9.069228304273802e-06, "loss": 0.0239, "step": 23722 }, { "epoch": 4.35, "learning_rate": 9.068489099778753e-06, "loss": 0.0246, "step": 23723 }, { "epoch": 4.35, "learning_rate": 9.067749900418214e-06, "loss": 0.0308, "step": 23724 }, { "epoch": 4.35, "learning_rate": 9.067010706196256e-06, "loss": 0.0232, "step": 23725 }, { "epoch": 4.35, "learning_rate": 9.06627151711696e-06, "loss": 0.0249, "step": 23726 }, { "epoch": 4.35, "learning_rate": 9.065532333184398e-06, "loss": 0.0221, "step": 23727 }, { "epoch": 4.35, "learning_rate": 9.064793154402643e-06, "loss": 0.0187, "step": 23728 }, { "epoch": 4.35, "learning_rate": 9.064053980775768e-06, "loss": 0.0262, "step": 23729 }, { "epoch": 4.35, "learning_rate": 9.063314812307852e-06, "loss": 0.0294, "step": 23730 }, { "epoch": 4.35, "learning_rate": 9.062575649002968e-06, "loss": 0.0243, "step": 23731 }, { "epoch": 4.35, "learning_rate": 9.061836490865187e-06, "loss": 0.0233, "step": 23732 }, { "epoch": 4.35, "learning_rate": 9.061097337898585e-06, "loss": 0.0322, "step": 23733 }, { "epoch": 4.35, "learning_rate": 9.060358190107235e-06, "loss": 0.0228, "step": 23734 }, { "epoch": 4.35, "learning_rate": 9.059619047495215e-06, "loss": 0.029, "step": 23735 }, { "epoch": 4.35, "learning_rate": 9.058879910066596e-06, "loss": 0.02, "step": 23736 }, { "epoch": 4.35, "learning_rate": 9.058140777825455e-06, "loss": 0.0251, "step": 23737 }, { "epoch": 4.35, "learning_rate": 9.057401650775859e-06, "loss": 0.0221, "step": 23738 }, { "epoch": 4.35, "learning_rate": 9.05666252892189e-06, "loss": 0.0267, "step": 23739 }, { "epoch": 4.35, "learning_rate": 9.055923412267622e-06, "loss": 0.0257, "step": 23740 }, { "epoch": 4.35, "learning_rate": 9.055184300817125e-06, "loss": 0.0276, "step": 23741 }, { "epoch": 4.35, "learning_rate": 9.054445194574473e-06, "loss": 0.0264, "step": 23742 }, { "epoch": 4.35, "learning_rate": 9.05370609354374e-06, "loss": 0.0259, "step": 23743 }, { "epoch": 4.35, "learning_rate": 9.052966997729002e-06, "loss": 0.0244, "step": 23744 }, { "epoch": 4.35, "learning_rate": 9.052227907134333e-06, "loss": 0.0259, "step": 23745 }, { "epoch": 4.35, "learning_rate": 9.051488821763807e-06, "loss": 0.0214, "step": 23746 }, { "epoch": 4.35, "learning_rate": 9.050749741621493e-06, "loss": 0.0281, "step": 23747 }, { "epoch": 4.36, "learning_rate": 9.050010666711471e-06, "loss": 0.0292, "step": 23748 }, { "epoch": 4.36, "learning_rate": 9.049271597037816e-06, "loss": 0.0324, "step": 23749 }, { "epoch": 4.36, "learning_rate": 9.048532532604594e-06, "loss": 0.0224, "step": 23750 }, { "epoch": 4.36, "learning_rate": 9.047793473415882e-06, "loss": 0.0219, "step": 23751 }, { "epoch": 4.36, "learning_rate": 9.047054419475756e-06, "loss": 0.0214, "step": 23752 }, { "epoch": 4.36, "learning_rate": 9.04631537078829e-06, "loss": 0.0245, "step": 23753 }, { "epoch": 4.36, "learning_rate": 9.045576327357554e-06, "loss": 0.0261, "step": 23754 }, { "epoch": 4.36, "learning_rate": 9.044837289187626e-06, "loss": 0.0246, "step": 23755 }, { "epoch": 4.36, "learning_rate": 9.044098256282574e-06, "loss": 0.0206, "step": 23756 }, { "epoch": 4.36, "learning_rate": 9.043359228646476e-06, "loss": 0.0266, "step": 23757 }, { "epoch": 4.36, "learning_rate": 9.042620206283407e-06, "loss": 0.0262, "step": 23758 }, { "epoch": 4.36, "learning_rate": 9.041881189197437e-06, "loss": 0.0195, "step": 23759 }, { "epoch": 4.36, "learning_rate": 9.041142177392637e-06, "loss": 0.0309, "step": 23760 }, { "epoch": 4.36, "learning_rate": 9.040403170873087e-06, "loss": 0.0297, "step": 23761 }, { "epoch": 4.36, "learning_rate": 9.039664169642855e-06, "loss": 0.0265, "step": 23762 }, { "epoch": 4.36, "learning_rate": 9.038925173706018e-06, "loss": 0.0288, "step": 23763 }, { "epoch": 4.36, "learning_rate": 9.038186183066648e-06, "loss": 0.0238, "step": 23764 }, { "epoch": 4.36, "learning_rate": 9.037447197728816e-06, "loss": 0.0273, "step": 23765 }, { "epoch": 4.36, "learning_rate": 9.0367082176966e-06, "loss": 0.0256, "step": 23766 }, { "epoch": 4.36, "learning_rate": 9.035969242974074e-06, "loss": 0.0273, "step": 23767 }, { "epoch": 4.36, "learning_rate": 9.035230273565303e-06, "loss": 0.0275, "step": 23768 }, { "epoch": 4.36, "learning_rate": 9.034491309474367e-06, "loss": 0.0283, "step": 23769 }, { "epoch": 4.36, "learning_rate": 9.033752350705335e-06, "loss": 0.0234, "step": 23770 }, { "epoch": 4.36, "learning_rate": 9.033013397262286e-06, "loss": 0.0267, "step": 23771 }, { "epoch": 4.36, "learning_rate": 9.03227444914929e-06, "loss": 0.0301, "step": 23772 }, { "epoch": 4.36, "learning_rate": 9.031535506370417e-06, "loss": 0.0223, "step": 23773 }, { "epoch": 4.36, "learning_rate": 9.030796568929744e-06, "loss": 0.025, "step": 23774 }, { "epoch": 4.36, "learning_rate": 9.030057636831344e-06, "loss": 0.0286, "step": 23775 }, { "epoch": 4.36, "learning_rate": 9.029318710079291e-06, "loss": 0.0237, "step": 23776 }, { "epoch": 4.36, "learning_rate": 9.028579788677655e-06, "loss": 0.0207, "step": 23777 }, { "epoch": 4.36, "learning_rate": 9.027840872630506e-06, "loss": 0.0315, "step": 23778 }, { "epoch": 4.36, "learning_rate": 9.027101961941925e-06, "loss": 0.018, "step": 23779 }, { "epoch": 4.36, "learning_rate": 9.026363056615978e-06, "loss": 0.0266, "step": 23780 }, { "epoch": 4.36, "learning_rate": 9.025624156656744e-06, "loss": 0.027, "step": 23781 }, { "epoch": 4.36, "learning_rate": 9.024885262068289e-06, "loss": 0.0238, "step": 23782 }, { "epoch": 4.36, "learning_rate": 9.024146372854689e-06, "loss": 0.0258, "step": 23783 }, { "epoch": 4.36, "learning_rate": 9.02340748902002e-06, "loss": 0.0278, "step": 23784 }, { "epoch": 4.36, "learning_rate": 9.022668610568353e-06, "loss": 0.019, "step": 23785 }, { "epoch": 4.36, "learning_rate": 9.021929737503757e-06, "loss": 0.0232, "step": 23786 }, { "epoch": 4.36, "learning_rate": 9.021190869830306e-06, "loss": 0.0202, "step": 23787 }, { "epoch": 4.36, "learning_rate": 9.020452007552076e-06, "loss": 0.0231, "step": 23788 }, { "epoch": 4.36, "learning_rate": 9.019713150673138e-06, "loss": 0.029, "step": 23789 }, { "epoch": 4.36, "learning_rate": 9.018974299197562e-06, "loss": 0.0225, "step": 23790 }, { "epoch": 4.36, "learning_rate": 9.018235453129422e-06, "loss": 0.0275, "step": 23791 }, { "epoch": 4.36, "learning_rate": 9.017496612472792e-06, "loss": 0.0248, "step": 23792 }, { "epoch": 4.36, "learning_rate": 9.016757777231744e-06, "loss": 0.0231, "step": 23793 }, { "epoch": 4.36, "learning_rate": 9.016018947410354e-06, "loss": 0.0229, "step": 23794 }, { "epoch": 4.36, "learning_rate": 9.015280123012688e-06, "loss": 0.0245, "step": 23795 }, { "epoch": 4.36, "learning_rate": 9.014541304042817e-06, "loss": 0.0213, "step": 23796 }, { "epoch": 4.36, "learning_rate": 9.013802490504822e-06, "loss": 0.0252, "step": 23797 }, { "epoch": 4.36, "learning_rate": 9.013063682402769e-06, "loss": 0.0257, "step": 23798 }, { "epoch": 4.36, "learning_rate": 9.012324879740732e-06, "loss": 0.0261, "step": 23799 }, { "epoch": 4.36, "learning_rate": 9.011586082522781e-06, "loss": 0.026, "step": 23800 }, { "epoch": 4.36, "learning_rate": 9.010847290752996e-06, "loss": 0.0249, "step": 23801 }, { "epoch": 4.36, "learning_rate": 9.010108504435441e-06, "loss": 0.0233, "step": 23802 }, { "epoch": 4.37, "learning_rate": 9.009369723574194e-06, "loss": 0.0249, "step": 23803 }, { "epoch": 4.37, "learning_rate": 9.008630948173317e-06, "loss": 0.0224, "step": 23804 }, { "epoch": 4.37, "learning_rate": 9.007892178236896e-06, "loss": 0.0198, "step": 23805 }, { "epoch": 4.37, "learning_rate": 9.007153413768997e-06, "loss": 0.0249, "step": 23806 }, { "epoch": 4.37, "learning_rate": 9.006414654773689e-06, "loss": 0.0232, "step": 23807 }, { "epoch": 4.37, "learning_rate": 9.005675901255047e-06, "loss": 0.0252, "step": 23808 }, { "epoch": 4.37, "learning_rate": 9.00493715321714e-06, "loss": 0.0297, "step": 23809 }, { "epoch": 4.37, "learning_rate": 9.004198410664047e-06, "loss": 0.021, "step": 23810 }, { "epoch": 4.37, "learning_rate": 9.003459673599835e-06, "loss": 0.0263, "step": 23811 }, { "epoch": 4.37, "learning_rate": 9.002720942028577e-06, "loss": 0.0214, "step": 23812 }, { "epoch": 4.37, "learning_rate": 9.00198221595434e-06, "loss": 0.0267, "step": 23813 }, { "epoch": 4.37, "learning_rate": 9.001243495381207e-06, "loss": 0.0274, "step": 23814 }, { "epoch": 4.37, "learning_rate": 9.00050478031324e-06, "loss": 0.0315, "step": 23815 }, { "epoch": 4.37, "learning_rate": 8.999766070754514e-06, "loss": 0.0254, "step": 23816 }, { "epoch": 4.37, "learning_rate": 8.999027366709101e-06, "loss": 0.0239, "step": 23817 }, { "epoch": 4.37, "learning_rate": 8.998288668181072e-06, "loss": 0.025, "step": 23818 }, { "epoch": 4.37, "learning_rate": 8.9975499751745e-06, "loss": 0.0226, "step": 23819 }, { "epoch": 4.37, "learning_rate": 8.996811287693458e-06, "loss": 0.027, "step": 23820 }, { "epoch": 4.37, "learning_rate": 8.996072605742014e-06, "loss": 0.0188, "step": 23821 }, { "epoch": 4.37, "learning_rate": 8.995333929324236e-06, "loss": 0.024, "step": 23822 }, { "epoch": 4.37, "learning_rate": 8.99459525844421e-06, "loss": 0.0202, "step": 23823 }, { "epoch": 4.37, "learning_rate": 8.993856593105993e-06, "loss": 0.0286, "step": 23824 }, { "epoch": 4.37, "learning_rate": 8.993117933313665e-06, "loss": 0.0289, "step": 23825 }, { "epoch": 4.37, "learning_rate": 8.992379279071291e-06, "loss": 0.0313, "step": 23826 }, { "epoch": 4.37, "learning_rate": 8.991640630382948e-06, "loss": 0.0277, "step": 23827 }, { "epoch": 4.37, "learning_rate": 8.990901987252706e-06, "loss": 0.0238, "step": 23828 }, { "epoch": 4.37, "learning_rate": 8.990163349684638e-06, "loss": 0.0279, "step": 23829 }, { "epoch": 4.37, "learning_rate": 8.989424717682812e-06, "loss": 0.0324, "step": 23830 }, { "epoch": 4.37, "learning_rate": 8.988686091251292e-06, "loss": 0.025, "step": 23831 }, { "epoch": 4.37, "learning_rate": 8.98794747039417e-06, "loss": 0.0218, "step": 23832 }, { "epoch": 4.37, "learning_rate": 8.9872088551155e-06, "loss": 0.0215, "step": 23833 }, { "epoch": 4.37, "learning_rate": 8.986470245419359e-06, "loss": 0.0214, "step": 23834 }, { "epoch": 4.37, "learning_rate": 8.985731641309816e-06, "loss": 0.0247, "step": 23835 }, { "epoch": 4.37, "learning_rate": 8.984993042790946e-06, "loss": 0.0281, "step": 23836 }, { "epoch": 4.37, "learning_rate": 8.984254449866819e-06, "loss": 0.0237, "step": 23837 }, { "epoch": 4.37, "learning_rate": 8.983515862541505e-06, "loss": 0.0283, "step": 23838 }, { "epoch": 4.37, "learning_rate": 8.982777280819074e-06, "loss": 0.0224, "step": 23839 }, { "epoch": 4.37, "learning_rate": 8.982038704703596e-06, "loss": 0.0275, "step": 23840 }, { "epoch": 4.37, "learning_rate": 8.98130013419915e-06, "loss": 0.0261, "step": 23841 }, { "epoch": 4.37, "learning_rate": 8.980561569309799e-06, "loss": 0.0243, "step": 23842 }, { "epoch": 4.37, "learning_rate": 8.979823010039617e-06, "loss": 0.0224, "step": 23843 }, { "epoch": 4.37, "learning_rate": 8.979084456392671e-06, "loss": 0.0242, "step": 23844 }, { "epoch": 4.37, "learning_rate": 8.97834590837304e-06, "loss": 0.0226, "step": 23845 }, { "epoch": 4.37, "learning_rate": 8.977607365984789e-06, "loss": 0.0227, "step": 23846 }, { "epoch": 4.37, "learning_rate": 8.976868829231992e-06, "loss": 0.0271, "step": 23847 }, { "epoch": 4.37, "learning_rate": 8.976130298118717e-06, "loss": 0.0256, "step": 23848 }, { "epoch": 4.37, "learning_rate": 8.975391772649029e-06, "loss": 0.0285, "step": 23849 }, { "epoch": 4.37, "learning_rate": 8.974653252827014e-06, "loss": 0.0211, "step": 23850 }, { "epoch": 4.37, "learning_rate": 8.97391473865673e-06, "loss": 0.025, "step": 23851 }, { "epoch": 4.37, "learning_rate": 8.973176230142254e-06, "loss": 0.0252, "step": 23852 }, { "epoch": 4.37, "learning_rate": 8.972437727287652e-06, "loss": 0.0274, "step": 23853 }, { "epoch": 4.37, "learning_rate": 8.971699230097001e-06, "loss": 0.0266, "step": 23854 }, { "epoch": 4.37, "learning_rate": 8.970960738574366e-06, "loss": 0.0224, "step": 23855 }, { "epoch": 4.37, "learning_rate": 8.970222252723823e-06, "loss": 0.0187, "step": 23856 }, { "epoch": 4.38, "learning_rate": 8.969483772549432e-06, "loss": 0.0267, "step": 23857 }, { "epoch": 4.38, "learning_rate": 8.968745298055274e-06, "loss": 0.0299, "step": 23858 }, { "epoch": 4.38, "learning_rate": 8.968006829245419e-06, "loss": 0.0265, "step": 23859 }, { "epoch": 4.38, "learning_rate": 8.967268366123932e-06, "loss": 0.0263, "step": 23860 }, { "epoch": 4.38, "learning_rate": 8.966529908694886e-06, "loss": 0.0258, "step": 23861 }, { "epoch": 4.38, "learning_rate": 8.965791456962349e-06, "loss": 0.0234, "step": 23862 }, { "epoch": 4.38, "learning_rate": 8.965053010930397e-06, "loss": 0.0218, "step": 23863 }, { "epoch": 4.38, "learning_rate": 8.964314570603096e-06, "loss": 0.0274, "step": 23864 }, { "epoch": 4.38, "learning_rate": 8.96357613598452e-06, "loss": 0.0257, "step": 23865 }, { "epoch": 4.38, "learning_rate": 8.962837707078731e-06, "loss": 0.0242, "step": 23866 }, { "epoch": 4.38, "learning_rate": 8.96209928388981e-06, "loss": 0.0242, "step": 23867 }, { "epoch": 4.38, "learning_rate": 8.96136086642182e-06, "loss": 0.0263, "step": 23868 }, { "epoch": 4.38, "learning_rate": 8.960622454678835e-06, "loss": 0.0222, "step": 23869 }, { "epoch": 4.38, "learning_rate": 8.959884048664921e-06, "loss": 0.0217, "step": 23870 }, { "epoch": 4.38, "learning_rate": 8.959145648384148e-06, "loss": 0.0238, "step": 23871 }, { "epoch": 4.38, "learning_rate": 8.958407253840593e-06, "loss": 0.0252, "step": 23872 }, { "epoch": 4.38, "learning_rate": 8.95766886503832e-06, "loss": 0.0223, "step": 23873 }, { "epoch": 4.38, "learning_rate": 8.956930481981404e-06, "loss": 0.0292, "step": 23874 }, { "epoch": 4.38, "learning_rate": 8.956192104673905e-06, "loss": 0.0255, "step": 23875 }, { "epoch": 4.38, "learning_rate": 8.955453733119905e-06, "loss": 0.028, "step": 23876 }, { "epoch": 4.38, "learning_rate": 8.954715367323468e-06, "loss": 0.0287, "step": 23877 }, { "epoch": 4.38, "learning_rate": 8.953977007288662e-06, "loss": 0.0256, "step": 23878 }, { "epoch": 4.38, "learning_rate": 8.95323865301956e-06, "loss": 0.025, "step": 23879 }, { "epoch": 4.38, "learning_rate": 8.952500304520231e-06, "loss": 0.0258, "step": 23880 }, { "epoch": 4.38, "learning_rate": 8.951761961794746e-06, "loss": 0.0241, "step": 23881 }, { "epoch": 4.38, "learning_rate": 8.951023624847173e-06, "loss": 0.0253, "step": 23882 }, { "epoch": 4.38, "learning_rate": 8.950285293681585e-06, "loss": 0.0217, "step": 23883 }, { "epoch": 4.38, "learning_rate": 8.949546968302042e-06, "loss": 0.0237, "step": 23884 }, { "epoch": 4.38, "learning_rate": 8.948808648712627e-06, "loss": 0.0228, "step": 23885 }, { "epoch": 4.38, "learning_rate": 8.948070334917401e-06, "loss": 0.0277, "step": 23886 }, { "epoch": 4.38, "learning_rate": 8.947332026920437e-06, "loss": 0.0234, "step": 23887 }, { "epoch": 4.38, "learning_rate": 8.9465937247258e-06, "loss": 0.0275, "step": 23888 }, { "epoch": 4.38, "learning_rate": 8.945855428337567e-06, "loss": 0.0219, "step": 23889 }, { "epoch": 4.38, "learning_rate": 8.945117137759802e-06, "loss": 0.0231, "step": 23890 }, { "epoch": 4.38, "learning_rate": 8.944378852996578e-06, "loss": 0.0222, "step": 23891 }, { "epoch": 4.38, "learning_rate": 8.943640574051963e-06, "loss": 0.0219, "step": 23892 }, { "epoch": 4.38, "learning_rate": 8.942902300930019e-06, "loss": 0.0226, "step": 23893 }, { "epoch": 4.38, "learning_rate": 8.94216403363483e-06, "loss": 0.0236, "step": 23894 }, { "epoch": 4.38, "learning_rate": 8.941425772170456e-06, "loss": 0.0238, "step": 23895 }, { "epoch": 4.38, "learning_rate": 8.940687516540968e-06, "loss": 0.0286, "step": 23896 }, { "epoch": 4.38, "learning_rate": 8.939949266750431e-06, "loss": 0.0206, "step": 23897 }, { "epoch": 4.38, "learning_rate": 8.939211022802922e-06, "loss": 0.0209, "step": 23898 }, { "epoch": 4.38, "learning_rate": 8.938472784702506e-06, "loss": 0.0213, "step": 23899 }, { "epoch": 4.38, "learning_rate": 8.937734552453254e-06, "loss": 0.0285, "step": 23900 }, { "epoch": 4.38, "learning_rate": 8.93699632605923e-06, "loss": 0.0267, "step": 23901 }, { "epoch": 4.38, "learning_rate": 8.93625810552451e-06, "loss": 0.0269, "step": 23902 }, { "epoch": 4.38, "learning_rate": 8.935519890853163e-06, "loss": 0.0262, "step": 23903 }, { "epoch": 4.38, "learning_rate": 8.934781682049253e-06, "loss": 0.0232, "step": 23904 }, { "epoch": 4.38, "learning_rate": 8.934043479116852e-06, "loss": 0.0226, "step": 23905 }, { "epoch": 4.38, "learning_rate": 8.933305282060025e-06, "loss": 0.019, "step": 23906 }, { "epoch": 4.38, "learning_rate": 8.932567090882847e-06, "loss": 0.0261, "step": 23907 }, { "epoch": 4.38, "learning_rate": 8.931828905589385e-06, "loss": 0.0332, "step": 23908 }, { "epoch": 4.38, "learning_rate": 8.931090726183705e-06, "loss": 0.0231, "step": 23909 }, { "epoch": 4.38, "learning_rate": 8.930352552669878e-06, "loss": 0.0205, "step": 23910 }, { "epoch": 4.38, "learning_rate": 8.929614385051974e-06, "loss": 0.0291, "step": 23911 }, { "epoch": 4.39, "learning_rate": 8.928876223334061e-06, "loss": 0.0283, "step": 23912 }, { "epoch": 4.39, "learning_rate": 8.928138067520209e-06, "loss": 0.0256, "step": 23913 }, { "epoch": 4.39, "learning_rate": 8.927399917614482e-06, "loss": 0.0265, "step": 23914 }, { "epoch": 4.39, "learning_rate": 8.926661773620951e-06, "loss": 0.0251, "step": 23915 }, { "epoch": 4.39, "learning_rate": 8.925923635543687e-06, "loss": 0.0257, "step": 23916 }, { "epoch": 4.39, "learning_rate": 8.925185503386757e-06, "loss": 0.025, "step": 23917 }, { "epoch": 4.39, "learning_rate": 8.924447377154231e-06, "loss": 0.0245, "step": 23918 }, { "epoch": 4.39, "learning_rate": 8.923709256850173e-06, "loss": 0.0199, "step": 23919 }, { "epoch": 4.39, "learning_rate": 8.922971142478656e-06, "loss": 0.0265, "step": 23920 }, { "epoch": 4.39, "learning_rate": 8.922233034043751e-06, "loss": 0.02, "step": 23921 }, { "epoch": 4.39, "learning_rate": 8.92149493154952e-06, "loss": 0.0226, "step": 23922 }, { "epoch": 4.39, "learning_rate": 8.920756835000035e-06, "loss": 0.0235, "step": 23923 }, { "epoch": 4.39, "learning_rate": 8.920018744399361e-06, "loss": 0.0258, "step": 23924 }, { "epoch": 4.39, "learning_rate": 8.91928065975157e-06, "loss": 0.0245, "step": 23925 }, { "epoch": 4.39, "learning_rate": 8.91854258106073e-06, "loss": 0.0275, "step": 23926 }, { "epoch": 4.39, "learning_rate": 8.91780450833091e-06, "loss": 0.0222, "step": 23927 }, { "epoch": 4.39, "learning_rate": 8.917066441566173e-06, "loss": 0.0248, "step": 23928 }, { "epoch": 4.39, "learning_rate": 8.916328380770595e-06, "loss": 0.033, "step": 23929 }, { "epoch": 4.39, "learning_rate": 8.915590325948242e-06, "loss": 0.025, "step": 23930 }, { "epoch": 4.39, "learning_rate": 8.914852277103177e-06, "loss": 0.0288, "step": 23931 }, { "epoch": 4.39, "learning_rate": 8.91411423423947e-06, "loss": 0.0318, "step": 23932 }, { "epoch": 4.39, "learning_rate": 8.913376197361194e-06, "loss": 0.0194, "step": 23933 }, { "epoch": 4.39, "learning_rate": 8.912638166472412e-06, "loss": 0.0245, "step": 23934 }, { "epoch": 4.39, "learning_rate": 8.911900141577195e-06, "loss": 0.0238, "step": 23935 }, { "epoch": 4.39, "learning_rate": 8.911162122679613e-06, "loss": 0.0242, "step": 23936 }, { "epoch": 4.39, "learning_rate": 8.910424109783723e-06, "loss": 0.0215, "step": 23937 }, { "epoch": 4.39, "learning_rate": 8.909686102893606e-06, "loss": 0.0278, "step": 23938 }, { "epoch": 4.39, "learning_rate": 8.908948102013326e-06, "loss": 0.0271, "step": 23939 }, { "epoch": 4.39, "learning_rate": 8.908210107146948e-06, "loss": 0.0259, "step": 23940 }, { "epoch": 4.39, "learning_rate": 8.90747211829854e-06, "loss": 0.0221, "step": 23941 }, { "epoch": 4.39, "learning_rate": 8.906734135472173e-06, "loss": 0.0302, "step": 23942 }, { "epoch": 4.39, "learning_rate": 8.905996158671912e-06, "loss": 0.0292, "step": 23943 }, { "epoch": 4.39, "learning_rate": 8.905258187901826e-06, "loss": 0.0273, "step": 23944 }, { "epoch": 4.39, "learning_rate": 8.904520223165986e-06, "loss": 0.0254, "step": 23945 }, { "epoch": 4.39, "learning_rate": 8.903782264468448e-06, "loss": 0.0287, "step": 23946 }, { "epoch": 4.39, "learning_rate": 8.903044311813294e-06, "loss": 0.0241, "step": 23947 }, { "epoch": 4.39, "learning_rate": 8.902306365204587e-06, "loss": 0.0258, "step": 23948 }, { "epoch": 4.39, "learning_rate": 8.90156842464639e-06, "loss": 0.0312, "step": 23949 }, { "epoch": 4.39, "learning_rate": 8.900830490142772e-06, "loss": 0.0213, "step": 23950 }, { "epoch": 4.39, "learning_rate": 8.900092561697805e-06, "loss": 0.0256, "step": 23951 }, { "epoch": 4.39, "learning_rate": 8.899354639315553e-06, "loss": 0.0284, "step": 23952 }, { "epoch": 4.39, "learning_rate": 8.898616723000083e-06, "loss": 0.0245, "step": 23953 }, { "epoch": 4.39, "learning_rate": 8.897878812755463e-06, "loss": 0.0226, "step": 23954 }, { "epoch": 4.39, "learning_rate": 8.897140908585762e-06, "loss": 0.031, "step": 23955 }, { "epoch": 4.39, "learning_rate": 8.896403010495046e-06, "loss": 0.0228, "step": 23956 }, { "epoch": 4.39, "learning_rate": 8.895665118487386e-06, "loss": 0.0271, "step": 23957 }, { "epoch": 4.39, "learning_rate": 8.894927232566842e-06, "loss": 0.0317, "step": 23958 }, { "epoch": 4.39, "learning_rate": 8.894189352737483e-06, "loss": 0.0295, "step": 23959 }, { "epoch": 4.39, "learning_rate": 8.893451479003381e-06, "loss": 0.0305, "step": 23960 }, { "epoch": 4.39, "learning_rate": 8.892713611368599e-06, "loss": 0.0294, "step": 23961 }, { "epoch": 4.39, "learning_rate": 8.891975749837208e-06, "loss": 0.0245, "step": 23962 }, { "epoch": 4.39, "learning_rate": 8.891237894413268e-06, "loss": 0.0217, "step": 23963 }, { "epoch": 4.39, "learning_rate": 8.890500045100854e-06, "loss": 0.0264, "step": 23964 }, { "epoch": 4.39, "learning_rate": 8.88976220190403e-06, "loss": 0.0264, "step": 23965 }, { "epoch": 4.4, "learning_rate": 8.889024364826863e-06, "loss": 0.0309, "step": 23966 }, { "epoch": 4.4, "learning_rate": 8.88828653387342e-06, "loss": 0.0282, "step": 23967 }, { "epoch": 4.4, "learning_rate": 8.887548709047765e-06, "loss": 0.0268, "step": 23968 }, { "epoch": 4.4, "learning_rate": 8.88681089035397e-06, "loss": 0.0234, "step": 23969 }, { "epoch": 4.4, "learning_rate": 8.886073077796097e-06, "loss": 0.0329, "step": 23970 }, { "epoch": 4.4, "learning_rate": 8.885335271378218e-06, "loss": 0.0301, "step": 23971 }, { "epoch": 4.4, "learning_rate": 8.884597471104394e-06, "loss": 0.0242, "step": 23972 }, { "epoch": 4.4, "learning_rate": 8.883859676978697e-06, "loss": 0.0301, "step": 23973 }, { "epoch": 4.4, "learning_rate": 8.883121889005194e-06, "loss": 0.0229, "step": 23974 }, { "epoch": 4.4, "learning_rate": 8.882384107187946e-06, "loss": 0.0272, "step": 23975 }, { "epoch": 4.4, "learning_rate": 8.881646331531026e-06, "loss": 0.0252, "step": 23976 }, { "epoch": 4.4, "learning_rate": 8.880908562038492e-06, "loss": 0.0238, "step": 23977 }, { "epoch": 4.4, "learning_rate": 8.88017079871442e-06, "loss": 0.026, "step": 23978 }, { "epoch": 4.4, "learning_rate": 8.879433041562872e-06, "loss": 0.0247, "step": 23979 }, { "epoch": 4.4, "learning_rate": 8.878695290587917e-06, "loss": 0.0255, "step": 23980 }, { "epoch": 4.4, "learning_rate": 8.877957545793616e-06, "loss": 0.0257, "step": 23981 }, { "epoch": 4.4, "learning_rate": 8.877219807184042e-06, "loss": 0.0247, "step": 23982 }, { "epoch": 4.4, "learning_rate": 8.876482074763261e-06, "loss": 0.0284, "step": 23983 }, { "epoch": 4.4, "learning_rate": 8.875744348535336e-06, "loss": 0.0243, "step": 23984 }, { "epoch": 4.4, "learning_rate": 8.875006628504327e-06, "loss": 0.0206, "step": 23985 }, { "epoch": 4.4, "learning_rate": 8.874268914674318e-06, "loss": 0.0319, "step": 23986 }, { "epoch": 4.4, "learning_rate": 8.873531207049361e-06, "loss": 0.0305, "step": 23987 }, { "epoch": 4.4, "learning_rate": 8.872793505633526e-06, "loss": 0.0273, "step": 23988 }, { "epoch": 4.4, "learning_rate": 8.872055810430881e-06, "loss": 0.025, "step": 23989 }, { "epoch": 4.4, "learning_rate": 8.871318121445489e-06, "loss": 0.0261, "step": 23990 }, { "epoch": 4.4, "learning_rate": 8.870580438681417e-06, "loss": 0.0206, "step": 23991 }, { "epoch": 4.4, "learning_rate": 8.869842762142737e-06, "loss": 0.0283, "step": 23992 }, { "epoch": 4.4, "learning_rate": 8.869105091833508e-06, "loss": 0.0317, "step": 23993 }, { "epoch": 4.4, "learning_rate": 8.868367427757792e-06, "loss": 0.0219, "step": 23994 }, { "epoch": 4.4, "learning_rate": 8.86762976991967e-06, "loss": 0.0269, "step": 23995 }, { "epoch": 4.4, "learning_rate": 8.866892118323196e-06, "loss": 0.0275, "step": 23996 }, { "epoch": 4.4, "learning_rate": 8.86615447297244e-06, "loss": 0.0246, "step": 23997 }, { "epoch": 4.4, "learning_rate": 8.865416833871465e-06, "loss": 0.0267, "step": 23998 }, { "epoch": 4.4, "learning_rate": 8.864679201024338e-06, "loss": 0.0267, "step": 23999 }, { "epoch": 4.4, "learning_rate": 8.86394157443513e-06, "loss": 0.0286, "step": 24000 }, { "epoch": 4.4, "learning_rate": 8.863203954107902e-06, "loss": 0.023, "step": 24001 }, { "epoch": 4.4, "learning_rate": 8.862466340046719e-06, "loss": 0.0198, "step": 24002 }, { "epoch": 4.4, "learning_rate": 8.861728732255643e-06, "loss": 0.0251, "step": 24003 }, { "epoch": 4.4, "learning_rate": 8.860991130738754e-06, "loss": 0.0267, "step": 24004 }, { "epoch": 4.4, "learning_rate": 8.860253535500105e-06, "loss": 0.0237, "step": 24005 }, { "epoch": 4.4, "learning_rate": 8.859515946543763e-06, "loss": 0.0238, "step": 24006 }, { "epoch": 4.4, "learning_rate": 8.858778363873796e-06, "loss": 0.0209, "step": 24007 }, { "epoch": 4.4, "learning_rate": 8.858040787494271e-06, "loss": 0.0205, "step": 24008 }, { "epoch": 4.4, "learning_rate": 8.857303217409253e-06, "loss": 0.0238, "step": 24009 }, { "epoch": 4.4, "learning_rate": 8.856565653622808e-06, "loss": 0.0254, "step": 24010 }, { "epoch": 4.4, "learning_rate": 8.855828096139e-06, "loss": 0.0243, "step": 24011 }, { "epoch": 4.4, "learning_rate": 8.855090544961886e-06, "loss": 0.0175, "step": 24012 }, { "epoch": 4.4, "learning_rate": 8.854353000095548e-06, "loss": 0.0208, "step": 24013 }, { "epoch": 4.4, "learning_rate": 8.85361546154404e-06, "loss": 0.0244, "step": 24014 }, { "epoch": 4.4, "learning_rate": 8.852877929311432e-06, "loss": 0.0249, "step": 24015 }, { "epoch": 4.4, "learning_rate": 8.852140403401785e-06, "loss": 0.0247, "step": 24016 }, { "epoch": 4.4, "learning_rate": 8.851402883819168e-06, "loss": 0.0225, "step": 24017 }, { "epoch": 4.4, "learning_rate": 8.850665370567645e-06, "loss": 0.0254, "step": 24018 }, { "epoch": 4.4, "learning_rate": 8.849927863651287e-06, "loss": 0.0322, "step": 24019 }, { "epoch": 4.4, "learning_rate": 8.849190363074148e-06, "loss": 0.0271, "step": 24020 }, { "epoch": 4.41, "learning_rate": 8.848452868840299e-06, "loss": 0.0225, "step": 24021 }, { "epoch": 4.41, "learning_rate": 8.847715380953804e-06, "loss": 0.0226, "step": 24022 }, { "epoch": 4.41, "learning_rate": 8.84697789941873e-06, "loss": 0.0283, "step": 24023 }, { "epoch": 4.41, "learning_rate": 8.846240424239141e-06, "loss": 0.0223, "step": 24024 }, { "epoch": 4.41, "learning_rate": 8.845502955419101e-06, "loss": 0.0253, "step": 24025 }, { "epoch": 4.41, "learning_rate": 8.844765492962677e-06, "loss": 0.026, "step": 24026 }, { "epoch": 4.41, "learning_rate": 8.844028036873931e-06, "loss": 0.025, "step": 24027 }, { "epoch": 4.41, "learning_rate": 8.843290587156934e-06, "loss": 0.0287, "step": 24028 }, { "epoch": 4.41, "learning_rate": 8.842553143815739e-06, "loss": 0.0236, "step": 24029 }, { "epoch": 4.41, "learning_rate": 8.841815706854426e-06, "loss": 0.03, "step": 24030 }, { "epoch": 4.41, "learning_rate": 8.841078276277047e-06, "loss": 0.0257, "step": 24031 }, { "epoch": 4.41, "learning_rate": 8.840340852087674e-06, "loss": 0.0239, "step": 24032 }, { "epoch": 4.41, "learning_rate": 8.839603434290368e-06, "loss": 0.0271, "step": 24033 }, { "epoch": 4.41, "learning_rate": 8.838866022889195e-06, "loss": 0.0252, "step": 24034 }, { "epoch": 4.41, "learning_rate": 8.838128617888218e-06, "loss": 0.0225, "step": 24035 }, { "epoch": 4.41, "learning_rate": 8.837391219291507e-06, "loss": 0.0221, "step": 24036 }, { "epoch": 4.41, "learning_rate": 8.836653827103124e-06, "loss": 0.0234, "step": 24037 }, { "epoch": 4.41, "learning_rate": 8.835916441327125e-06, "loss": 0.024, "step": 24038 }, { "epoch": 4.41, "learning_rate": 8.83517906196759e-06, "loss": 0.0226, "step": 24039 }, { "epoch": 4.41, "learning_rate": 8.83444168902857e-06, "loss": 0.0279, "step": 24040 }, { "epoch": 4.41, "learning_rate": 8.833704322514138e-06, "loss": 0.0264, "step": 24041 }, { "epoch": 4.41, "learning_rate": 8.832966962428352e-06, "loss": 0.0208, "step": 24042 }, { "epoch": 4.41, "learning_rate": 8.83222960877528e-06, "loss": 0.0228, "step": 24043 }, { "epoch": 4.41, "learning_rate": 8.831492261558987e-06, "loss": 0.0235, "step": 24044 }, { "epoch": 4.41, "learning_rate": 8.830754920783536e-06, "loss": 0.0275, "step": 24045 }, { "epoch": 4.41, "learning_rate": 8.830017586452993e-06, "loss": 0.0274, "step": 24046 }, { "epoch": 4.41, "learning_rate": 8.829280258571414e-06, "loss": 0.0255, "step": 24047 }, { "epoch": 4.41, "learning_rate": 8.828542937142877e-06, "loss": 0.0262, "step": 24048 }, { "epoch": 4.41, "learning_rate": 8.827805622171437e-06, "loss": 0.0264, "step": 24049 }, { "epoch": 4.41, "learning_rate": 8.827068313661159e-06, "loss": 0.0241, "step": 24050 }, { "epoch": 4.41, "learning_rate": 8.826331011616107e-06, "loss": 0.0272, "step": 24051 }, { "epoch": 4.41, "learning_rate": 8.825593716040345e-06, "loss": 0.0252, "step": 24052 }, { "epoch": 4.41, "learning_rate": 8.82485642693794e-06, "loss": 0.0237, "step": 24053 }, { "epoch": 4.41, "learning_rate": 8.824119144312953e-06, "loss": 0.0242, "step": 24054 }, { "epoch": 4.41, "learning_rate": 8.823381868169452e-06, "loss": 0.0189, "step": 24055 }, { "epoch": 4.41, "learning_rate": 8.822644598511491e-06, "loss": 0.0277, "step": 24056 }, { "epoch": 4.41, "learning_rate": 8.821907335343148e-06, "loss": 0.0275, "step": 24057 }, { "epoch": 4.41, "learning_rate": 8.821170078668475e-06, "loss": 0.0267, "step": 24058 }, { "epoch": 4.41, "learning_rate": 8.820432828491542e-06, "loss": 0.0301, "step": 24059 }, { "epoch": 4.41, "learning_rate": 8.819695584816408e-06, "loss": 0.0261, "step": 24060 }, { "epoch": 4.41, "learning_rate": 8.818958347647141e-06, "loss": 0.0297, "step": 24061 }, { "epoch": 4.41, "learning_rate": 8.818221116987805e-06, "loss": 0.0245, "step": 24062 }, { "epoch": 4.41, "learning_rate": 8.817483892842461e-06, "loss": 0.0224, "step": 24063 }, { "epoch": 4.41, "learning_rate": 8.816746675215176e-06, "loss": 0.0295, "step": 24064 }, { "epoch": 4.41, "learning_rate": 8.816009464110004e-06, "loss": 0.0271, "step": 24065 }, { "epoch": 4.41, "learning_rate": 8.815272259531022e-06, "loss": 0.024, "step": 24066 }, { "epoch": 4.41, "learning_rate": 8.814535061482285e-06, "loss": 0.0262, "step": 24067 }, { "epoch": 4.41, "learning_rate": 8.813797869967859e-06, "loss": 0.0245, "step": 24068 }, { "epoch": 4.41, "learning_rate": 8.813060684991805e-06, "loss": 0.0236, "step": 24069 }, { "epoch": 4.41, "learning_rate": 8.81232350655819e-06, "loss": 0.0278, "step": 24070 }, { "epoch": 4.41, "learning_rate": 8.811586334671075e-06, "loss": 0.0246, "step": 24071 }, { "epoch": 4.41, "learning_rate": 8.810849169334525e-06, "loss": 0.0225, "step": 24072 }, { "epoch": 4.41, "learning_rate": 8.810112010552604e-06, "loss": 0.0216, "step": 24073 }, { "epoch": 4.41, "learning_rate": 8.809374858329366e-06, "loss": 0.0273, "step": 24074 }, { "epoch": 4.42, "learning_rate": 8.808637712668889e-06, "loss": 0.019, "step": 24075 }, { "epoch": 4.42, "learning_rate": 8.807900573575227e-06, "loss": 0.0232, "step": 24076 }, { "epoch": 4.42, "learning_rate": 8.807163441052446e-06, "loss": 0.0232, "step": 24077 }, { "epoch": 4.42, "learning_rate": 8.806426315104603e-06, "loss": 0.0235, "step": 24078 }, { "epoch": 4.42, "learning_rate": 8.805689195735771e-06, "loss": 0.0281, "step": 24079 }, { "epoch": 4.42, "learning_rate": 8.804952082950008e-06, "loss": 0.0249, "step": 24080 }, { "epoch": 4.42, "learning_rate": 8.804214976751375e-06, "loss": 0.0292, "step": 24081 }, { "epoch": 4.42, "learning_rate": 8.803477877143935e-06, "loss": 0.0237, "step": 24082 }, { "epoch": 4.42, "learning_rate": 8.802740784131756e-06, "loss": 0.0322, "step": 24083 }, { "epoch": 4.42, "learning_rate": 8.802003697718902e-06, "loss": 0.0229, "step": 24084 }, { "epoch": 4.42, "learning_rate": 8.801266617909426e-06, "loss": 0.028, "step": 24085 }, { "epoch": 4.42, "learning_rate": 8.800529544707399e-06, "loss": 0.0221, "step": 24086 }, { "epoch": 4.42, "learning_rate": 8.799792478116878e-06, "loss": 0.0256, "step": 24087 }, { "epoch": 4.42, "learning_rate": 8.799055418141931e-06, "loss": 0.024, "step": 24088 }, { "epoch": 4.42, "learning_rate": 8.79831836478662e-06, "loss": 0.0268, "step": 24089 }, { "epoch": 4.42, "learning_rate": 8.797581318055007e-06, "loss": 0.028, "step": 24090 }, { "epoch": 4.42, "learning_rate": 8.796844277951148e-06, "loss": 0.028, "step": 24091 }, { "epoch": 4.42, "learning_rate": 8.796107244479116e-06, "loss": 0.0279, "step": 24092 }, { "epoch": 4.42, "learning_rate": 8.795370217642971e-06, "loss": 0.0245, "step": 24093 }, { "epoch": 4.42, "learning_rate": 8.79463319744677e-06, "loss": 0.0241, "step": 24094 }, { "epoch": 4.42, "learning_rate": 8.793896183894583e-06, "loss": 0.0204, "step": 24095 }, { "epoch": 4.42, "learning_rate": 8.793159176990464e-06, "loss": 0.0249, "step": 24096 }, { "epoch": 4.42, "learning_rate": 8.79242217673848e-06, "loss": 0.0218, "step": 24097 }, { "epoch": 4.42, "learning_rate": 8.791685183142696e-06, "loss": 0.0216, "step": 24098 }, { "epoch": 4.42, "learning_rate": 8.790948196207175e-06, "loss": 0.0242, "step": 24099 }, { "epoch": 4.42, "learning_rate": 8.790211215935965e-06, "loss": 0.0215, "step": 24100 }, { "epoch": 4.42, "learning_rate": 8.789474242333146e-06, "loss": 0.0273, "step": 24101 }, { "epoch": 4.42, "learning_rate": 8.788737275402775e-06, "loss": 0.0261, "step": 24102 }, { "epoch": 4.42, "learning_rate": 8.788000315148911e-06, "loss": 0.025, "step": 24103 }, { "epoch": 4.42, "learning_rate": 8.787263361575617e-06, "loss": 0.0226, "step": 24104 }, { "epoch": 4.42, "learning_rate": 8.786526414686954e-06, "loss": 0.0249, "step": 24105 }, { "epoch": 4.42, "learning_rate": 8.78578947448699e-06, "loss": 0.0202, "step": 24106 }, { "epoch": 4.42, "learning_rate": 8.78505254097978e-06, "loss": 0.0289, "step": 24107 }, { "epoch": 4.42, "learning_rate": 8.784315614169391e-06, "loss": 0.0221, "step": 24108 }, { "epoch": 4.42, "learning_rate": 8.783578694059878e-06, "loss": 0.0207, "step": 24109 }, { "epoch": 4.42, "learning_rate": 8.78284178065531e-06, "loss": 0.0231, "step": 24110 }, { "epoch": 4.42, "learning_rate": 8.78210487395975e-06, "loss": 0.0258, "step": 24111 }, { "epoch": 4.42, "learning_rate": 8.781367973977255e-06, "loss": 0.0261, "step": 24112 }, { "epoch": 4.42, "learning_rate": 8.780631080711886e-06, "loss": 0.023, "step": 24113 }, { "epoch": 4.42, "learning_rate": 8.779894194167708e-06, "loss": 0.0204, "step": 24114 }, { "epoch": 4.42, "learning_rate": 8.779157314348782e-06, "loss": 0.0301, "step": 24115 }, { "epoch": 4.42, "learning_rate": 8.77842044125917e-06, "loss": 0.0222, "step": 24116 }, { "epoch": 4.42, "learning_rate": 8.777683574902936e-06, "loss": 0.0282, "step": 24117 }, { "epoch": 4.42, "learning_rate": 8.77694671528413e-06, "loss": 0.0195, "step": 24118 }, { "epoch": 4.42, "learning_rate": 8.776209862406828e-06, "loss": 0.0217, "step": 24119 }, { "epoch": 4.42, "learning_rate": 8.77547301627509e-06, "loss": 0.0282, "step": 24120 }, { "epoch": 4.42, "learning_rate": 8.77473617689297e-06, "loss": 0.0249, "step": 24121 }, { "epoch": 4.42, "learning_rate": 8.773999344264529e-06, "loss": 0.0258, "step": 24122 }, { "epoch": 4.42, "learning_rate": 8.773262518393834e-06, "loss": 0.0253, "step": 24123 }, { "epoch": 4.42, "learning_rate": 8.772525699284947e-06, "loss": 0.0261, "step": 24124 }, { "epoch": 4.42, "learning_rate": 8.771788886941928e-06, "loss": 0.0237, "step": 24125 }, { "epoch": 4.42, "learning_rate": 8.771052081368838e-06, "loss": 0.0271, "step": 24126 }, { "epoch": 4.42, "learning_rate": 8.77031528256973e-06, "loss": 0.0181, "step": 24127 }, { "epoch": 4.42, "learning_rate": 8.769578490548679e-06, "loss": 0.0276, "step": 24128 }, { "epoch": 4.42, "learning_rate": 8.768841705309741e-06, "loss": 0.0271, "step": 24129 }, { "epoch": 4.43, "learning_rate": 8.768104926856976e-06, "loss": 0.0243, "step": 24130 }, { "epoch": 4.43, "learning_rate": 8.767368155194443e-06, "loss": 0.0246, "step": 24131 }, { "epoch": 4.43, "learning_rate": 8.766631390326207e-06, "loss": 0.0242, "step": 24132 }, { "epoch": 4.43, "learning_rate": 8.76589463225633e-06, "loss": 0.0235, "step": 24133 }, { "epoch": 4.43, "learning_rate": 8.765157880988868e-06, "loss": 0.0238, "step": 24134 }, { "epoch": 4.43, "learning_rate": 8.764421136527882e-06, "loss": 0.0221, "step": 24135 }, { "epoch": 4.43, "learning_rate": 8.763684398877439e-06, "loss": 0.0285, "step": 24136 }, { "epoch": 4.43, "learning_rate": 8.7629476680416e-06, "loss": 0.0307, "step": 24137 }, { "epoch": 4.43, "learning_rate": 8.762210944024418e-06, "loss": 0.02, "step": 24138 }, { "epoch": 4.43, "learning_rate": 8.76147422682996e-06, "loss": 0.0221, "step": 24139 }, { "epoch": 4.43, "learning_rate": 8.760737516462282e-06, "loss": 0.0263, "step": 24140 }, { "epoch": 4.43, "learning_rate": 8.76000081292545e-06, "loss": 0.0235, "step": 24141 }, { "epoch": 4.43, "learning_rate": 8.759264116223523e-06, "loss": 0.028, "step": 24142 }, { "epoch": 4.43, "learning_rate": 8.758527426360561e-06, "loss": 0.0241, "step": 24143 }, { "epoch": 4.43, "learning_rate": 8.757790743340621e-06, "loss": 0.0251, "step": 24144 }, { "epoch": 4.43, "learning_rate": 8.757054067167773e-06, "loss": 0.0242, "step": 24145 }, { "epoch": 4.43, "learning_rate": 8.756317397846072e-06, "loss": 0.0274, "step": 24146 }, { "epoch": 4.43, "learning_rate": 8.755580735379578e-06, "loss": 0.0251, "step": 24147 }, { "epoch": 4.43, "learning_rate": 8.754844079772352e-06, "loss": 0.0255, "step": 24148 }, { "epoch": 4.43, "learning_rate": 8.75410743102845e-06, "loss": 0.0218, "step": 24149 }, { "epoch": 4.43, "learning_rate": 8.753370789151941e-06, "loss": 0.0203, "step": 24150 }, { "epoch": 4.43, "learning_rate": 8.752634154146882e-06, "loss": 0.0248, "step": 24151 }, { "epoch": 4.43, "learning_rate": 8.751897526017332e-06, "loss": 0.0247, "step": 24152 }, { "epoch": 4.43, "learning_rate": 8.75116090476735e-06, "loss": 0.024, "step": 24153 }, { "epoch": 4.43, "learning_rate": 8.750424290401001e-06, "loss": 0.0242, "step": 24154 }, { "epoch": 4.43, "learning_rate": 8.749687682922345e-06, "loss": 0.0219, "step": 24155 }, { "epoch": 4.43, "learning_rate": 8.748951082335436e-06, "loss": 0.0297, "step": 24156 }, { "epoch": 4.43, "learning_rate": 8.748214488644334e-06, "loss": 0.029, "step": 24157 }, { "epoch": 4.43, "learning_rate": 8.74747790185311e-06, "loss": 0.0279, "step": 24158 }, { "epoch": 4.43, "learning_rate": 8.746741321965817e-06, "loss": 0.0252, "step": 24159 }, { "epoch": 4.43, "learning_rate": 8.746004748986512e-06, "loss": 0.0283, "step": 24160 }, { "epoch": 4.43, "learning_rate": 8.74526818291926e-06, "loss": 0.0321, "step": 24161 }, { "epoch": 4.43, "learning_rate": 8.744531623768116e-06, "loss": 0.0216, "step": 24162 }, { "epoch": 4.43, "learning_rate": 8.743795071537146e-06, "loss": 0.0252, "step": 24163 }, { "epoch": 4.43, "learning_rate": 8.743058526230409e-06, "loss": 0.0207, "step": 24164 }, { "epoch": 4.43, "learning_rate": 8.74232198785196e-06, "loss": 0.025, "step": 24165 }, { "epoch": 4.43, "learning_rate": 8.741585456405858e-06, "loss": 0.0249, "step": 24166 }, { "epoch": 4.43, "learning_rate": 8.740848931896173e-06, "loss": 0.0222, "step": 24167 }, { "epoch": 4.43, "learning_rate": 8.740112414326956e-06, "loss": 0.0231, "step": 24168 }, { "epoch": 4.43, "learning_rate": 8.73937590370227e-06, "loss": 0.0246, "step": 24169 }, { "epoch": 4.43, "learning_rate": 8.738639400026173e-06, "loss": 0.0262, "step": 24170 }, { "epoch": 4.43, "learning_rate": 8.73790290330272e-06, "loss": 0.0238, "step": 24171 }, { "epoch": 4.43, "learning_rate": 8.737166413535981e-06, "loss": 0.0256, "step": 24172 }, { "epoch": 4.43, "learning_rate": 8.736429930730012e-06, "loss": 0.0277, "step": 24173 }, { "epoch": 4.43, "learning_rate": 8.735693454888869e-06, "loss": 0.0294, "step": 24174 }, { "epoch": 4.43, "learning_rate": 8.73495698601661e-06, "loss": 0.0235, "step": 24175 }, { "epoch": 4.43, "learning_rate": 8.734220524117301e-06, "loss": 0.0252, "step": 24176 }, { "epoch": 4.43, "learning_rate": 8.733484069194997e-06, "loss": 0.0228, "step": 24177 }, { "epoch": 4.43, "learning_rate": 8.73274762125376e-06, "loss": 0.0282, "step": 24178 }, { "epoch": 4.43, "learning_rate": 8.732011180297646e-06, "loss": 0.0258, "step": 24179 }, { "epoch": 4.43, "learning_rate": 8.731274746330714e-06, "loss": 0.0268, "step": 24180 }, { "epoch": 4.43, "learning_rate": 8.73053831935703e-06, "loss": 0.0253, "step": 24181 }, { "epoch": 4.43, "learning_rate": 8.729801899380647e-06, "loss": 0.0242, "step": 24182 }, { "epoch": 4.43, "learning_rate": 8.729065486405628e-06, "loss": 0.0218, "step": 24183 }, { "epoch": 4.43, "learning_rate": 8.728329080436022e-06, "loss": 0.0185, "step": 24184 }, { "epoch": 4.44, "learning_rate": 8.727592681475903e-06, "loss": 0.0257, "step": 24185 }, { "epoch": 4.44, "learning_rate": 8.726856289529321e-06, "loss": 0.0322, "step": 24186 }, { "epoch": 4.44, "learning_rate": 8.726119904600337e-06, "loss": 0.0238, "step": 24187 }, { "epoch": 4.44, "learning_rate": 8.725383526693008e-06, "loss": 0.0242, "step": 24188 }, { "epoch": 4.44, "learning_rate": 8.724647155811397e-06, "loss": 0.0268, "step": 24189 }, { "epoch": 4.44, "learning_rate": 8.72391079195956e-06, "loss": 0.0298, "step": 24190 }, { "epoch": 4.44, "learning_rate": 8.723174435141559e-06, "loss": 0.025, "step": 24191 }, { "epoch": 4.44, "learning_rate": 8.72243808536145e-06, "loss": 0.0225, "step": 24192 }, { "epoch": 4.44, "learning_rate": 8.721701742623288e-06, "loss": 0.0196, "step": 24193 }, { "epoch": 4.44, "learning_rate": 8.72096540693114e-06, "loss": 0.0302, "step": 24194 }, { "epoch": 4.44, "learning_rate": 8.720229078289059e-06, "loss": 0.0244, "step": 24195 }, { "epoch": 4.44, "learning_rate": 8.719492756701105e-06, "loss": 0.0226, "step": 24196 }, { "epoch": 4.44, "learning_rate": 8.718756442171337e-06, "loss": 0.0237, "step": 24197 }, { "epoch": 4.44, "learning_rate": 8.718020134703812e-06, "loss": 0.0258, "step": 24198 }, { "epoch": 4.44, "learning_rate": 8.717283834302593e-06, "loss": 0.0253, "step": 24199 }, { "epoch": 4.44, "learning_rate": 8.716547540971738e-06, "loss": 0.03, "step": 24200 }, { "epoch": 4.44, "learning_rate": 8.7158112547153e-06, "loss": 0.026, "step": 24201 }, { "epoch": 4.44, "learning_rate": 8.715074975537338e-06, "loss": 0.0199, "step": 24202 }, { "epoch": 4.44, "learning_rate": 8.714338703441914e-06, "loss": 0.0227, "step": 24203 }, { "epoch": 4.44, "learning_rate": 8.713602438433086e-06, "loss": 0.0251, "step": 24204 }, { "epoch": 4.44, "learning_rate": 8.712866180514913e-06, "loss": 0.0279, "step": 24205 }, { "epoch": 4.44, "learning_rate": 8.712129929691449e-06, "loss": 0.0202, "step": 24206 }, { "epoch": 4.44, "learning_rate": 8.711393685966756e-06, "loss": 0.022, "step": 24207 }, { "epoch": 4.44, "learning_rate": 8.710657449344892e-06, "loss": 0.0216, "step": 24208 }, { "epoch": 4.44, "learning_rate": 8.709921219829916e-06, "loss": 0.0216, "step": 24209 }, { "epoch": 4.44, "learning_rate": 8.709184997425877e-06, "loss": 0.0214, "step": 24210 }, { "epoch": 4.44, "learning_rate": 8.70844878213685e-06, "loss": 0.025, "step": 24211 }, { "epoch": 4.44, "learning_rate": 8.707712573966878e-06, "loss": 0.0211, "step": 24212 }, { "epoch": 4.44, "learning_rate": 8.706976372920027e-06, "loss": 0.0206, "step": 24213 }, { "epoch": 4.44, "learning_rate": 8.706240179000351e-06, "loss": 0.0263, "step": 24214 }, { "epoch": 4.44, "learning_rate": 8.705503992211909e-06, "loss": 0.0314, "step": 24215 }, { "epoch": 4.44, "learning_rate": 8.70476781255876e-06, "loss": 0.0263, "step": 24216 }, { "epoch": 4.44, "learning_rate": 8.704031640044963e-06, "loss": 0.0185, "step": 24217 }, { "epoch": 4.44, "learning_rate": 8.703295474674573e-06, "loss": 0.0211, "step": 24218 }, { "epoch": 4.44, "learning_rate": 8.702559316451646e-06, "loss": 0.0249, "step": 24219 }, { "epoch": 4.44, "learning_rate": 8.701823165380247e-06, "loss": 0.0226, "step": 24220 }, { "epoch": 4.44, "learning_rate": 8.701087021464426e-06, "loss": 0.0247, "step": 24221 }, { "epoch": 4.44, "learning_rate": 8.700350884708245e-06, "loss": 0.0267, "step": 24222 }, { "epoch": 4.44, "learning_rate": 8.69961475511576e-06, "loss": 0.025, "step": 24223 }, { "epoch": 4.44, "learning_rate": 8.698878632691027e-06, "loss": 0.0294, "step": 24224 }, { "epoch": 4.44, "learning_rate": 8.698142517438108e-06, "loss": 0.0234, "step": 24225 }, { "epoch": 4.44, "learning_rate": 8.697406409361056e-06, "loss": 0.0255, "step": 24226 }, { "epoch": 4.44, "learning_rate": 8.696670308463935e-06, "loss": 0.0241, "step": 24227 }, { "epoch": 4.44, "learning_rate": 8.69593421475079e-06, "loss": 0.0198, "step": 24228 }, { "epoch": 4.44, "learning_rate": 8.695198128225693e-06, "loss": 0.0232, "step": 24229 }, { "epoch": 4.44, "learning_rate": 8.694462048892692e-06, "loss": 0.0217, "step": 24230 }, { "epoch": 4.44, "learning_rate": 8.693725976755846e-06, "loss": 0.025, "step": 24231 }, { "epoch": 4.44, "learning_rate": 8.692989911819216e-06, "loss": 0.0261, "step": 24232 }, { "epoch": 4.44, "learning_rate": 8.692253854086853e-06, "loss": 0.0229, "step": 24233 }, { "epoch": 4.44, "learning_rate": 8.691517803562816e-06, "loss": 0.0283, "step": 24234 }, { "epoch": 4.44, "learning_rate": 8.690781760251167e-06, "loss": 0.0238, "step": 24235 }, { "epoch": 4.44, "learning_rate": 8.690045724155961e-06, "loss": 0.021, "step": 24236 }, { "epoch": 4.44, "learning_rate": 8.689309695281248e-06, "loss": 0.0261, "step": 24237 }, { "epoch": 4.44, "learning_rate": 8.688573673631096e-06, "loss": 0.0218, "step": 24238 }, { "epoch": 4.45, "learning_rate": 8.687837659209556e-06, "loss": 0.0281, "step": 24239 }, { "epoch": 4.45, "learning_rate": 8.687101652020684e-06, "loss": 0.0279, "step": 24240 }, { "epoch": 4.45, "learning_rate": 8.686365652068536e-06, "loss": 0.021, "step": 24241 }, { "epoch": 4.45, "learning_rate": 8.685629659357173e-06, "loss": 0.032, "step": 24242 }, { "epoch": 4.45, "learning_rate": 8.68489367389065e-06, "loss": 0.021, "step": 24243 }, { "epoch": 4.45, "learning_rate": 8.68415769567303e-06, "loss": 0.0289, "step": 24244 }, { "epoch": 4.45, "learning_rate": 8.683421724708357e-06, "loss": 0.0285, "step": 24245 }, { "epoch": 4.45, "learning_rate": 8.682685761000691e-06, "loss": 0.0243, "step": 24246 }, { "epoch": 4.45, "learning_rate": 8.681949804554099e-06, "loss": 0.0256, "step": 24247 }, { "epoch": 4.45, "learning_rate": 8.68121385537263e-06, "loss": 0.0252, "step": 24248 }, { "epoch": 4.45, "learning_rate": 8.680477913460339e-06, "loss": 0.0259, "step": 24249 }, { "epoch": 4.45, "learning_rate": 8.679741978821284e-06, "loss": 0.0236, "step": 24250 }, { "epoch": 4.45, "learning_rate": 8.679006051459522e-06, "loss": 0.0272, "step": 24251 }, { "epoch": 4.45, "learning_rate": 8.678270131379112e-06, "loss": 0.0312, "step": 24252 }, { "epoch": 4.45, "learning_rate": 8.677534218584109e-06, "loss": 0.0271, "step": 24253 }, { "epoch": 4.45, "learning_rate": 8.676798313078565e-06, "loss": 0.0266, "step": 24254 }, { "epoch": 4.45, "learning_rate": 8.676062414866537e-06, "loss": 0.026, "step": 24255 }, { "epoch": 4.45, "learning_rate": 8.675326523952092e-06, "loss": 0.0278, "step": 24256 }, { "epoch": 4.45, "learning_rate": 8.674590640339274e-06, "loss": 0.0267, "step": 24257 }, { "epoch": 4.45, "learning_rate": 8.673854764032144e-06, "loss": 0.0269, "step": 24258 }, { "epoch": 4.45, "learning_rate": 8.673118895034754e-06, "loss": 0.0224, "step": 24259 }, { "epoch": 4.45, "learning_rate": 8.672383033351167e-06, "loss": 0.0247, "step": 24260 }, { "epoch": 4.45, "learning_rate": 8.671647178985436e-06, "loss": 0.0266, "step": 24261 }, { "epoch": 4.45, "learning_rate": 8.67091133194162e-06, "loss": 0.0269, "step": 24262 }, { "epoch": 4.45, "learning_rate": 8.670175492223763e-06, "loss": 0.0297, "step": 24263 }, { "epoch": 4.45, "learning_rate": 8.669439659835935e-06, "loss": 0.0253, "step": 24264 }, { "epoch": 4.45, "learning_rate": 8.66870383478219e-06, "loss": 0.0201, "step": 24265 }, { "epoch": 4.45, "learning_rate": 8.667968017066579e-06, "loss": 0.0256, "step": 24266 }, { "epoch": 4.45, "learning_rate": 8.667232206693157e-06, "loss": 0.0214, "step": 24267 }, { "epoch": 4.45, "learning_rate": 8.666496403665982e-06, "loss": 0.027, "step": 24268 }, { "epoch": 4.45, "learning_rate": 8.665760607989111e-06, "loss": 0.0274, "step": 24269 }, { "epoch": 4.45, "learning_rate": 8.6650248196666e-06, "loss": 0.028, "step": 24270 }, { "epoch": 4.45, "learning_rate": 8.664289038702505e-06, "loss": 0.0275, "step": 24271 }, { "epoch": 4.45, "learning_rate": 8.663553265100875e-06, "loss": 0.0245, "step": 24272 }, { "epoch": 4.45, "learning_rate": 8.662817498865773e-06, "loss": 0.0255, "step": 24273 }, { "epoch": 4.45, "learning_rate": 8.662081740001255e-06, "loss": 0.0251, "step": 24274 }, { "epoch": 4.45, "learning_rate": 8.661345988511372e-06, "loss": 0.0323, "step": 24275 }, { "epoch": 4.45, "learning_rate": 8.660610244400182e-06, "loss": 0.0212, "step": 24276 }, { "epoch": 4.45, "learning_rate": 8.659874507671737e-06, "loss": 0.0215, "step": 24277 }, { "epoch": 4.45, "learning_rate": 8.659138778330097e-06, "loss": 0.0234, "step": 24278 }, { "epoch": 4.45, "learning_rate": 8.658403056379314e-06, "loss": 0.0257, "step": 24279 }, { "epoch": 4.45, "learning_rate": 8.657667341823449e-06, "loss": 0.0219, "step": 24280 }, { "epoch": 4.45, "learning_rate": 8.656931634666546e-06, "loss": 0.0312, "step": 24281 }, { "epoch": 4.45, "learning_rate": 8.656195934912671e-06, "loss": 0.0202, "step": 24282 }, { "epoch": 4.45, "learning_rate": 8.655460242565878e-06, "loss": 0.0215, "step": 24283 }, { "epoch": 4.45, "learning_rate": 8.654724557630218e-06, "loss": 0.0223, "step": 24284 }, { "epoch": 4.45, "learning_rate": 8.653988880109744e-06, "loss": 0.0242, "step": 24285 }, { "epoch": 4.45, "learning_rate": 8.65325321000852e-06, "loss": 0.0274, "step": 24286 }, { "epoch": 4.45, "learning_rate": 8.652517547330595e-06, "loss": 0.0233, "step": 24287 }, { "epoch": 4.45, "learning_rate": 8.651781892080023e-06, "loss": 0.0252, "step": 24288 }, { "epoch": 4.45, "learning_rate": 8.651046244260863e-06, "loss": 0.0261, "step": 24289 }, { "epoch": 4.45, "learning_rate": 8.650310603877164e-06, "loss": 0.0262, "step": 24290 }, { "epoch": 4.45, "learning_rate": 8.649574970932988e-06, "loss": 0.0224, "step": 24291 }, { "epoch": 4.45, "learning_rate": 8.648839345432387e-06, "loss": 0.0233, "step": 24292 }, { "epoch": 4.45, "learning_rate": 8.648103727379412e-06, "loss": 0.0241, "step": 24293 }, { "epoch": 4.46, "learning_rate": 8.64736811677812e-06, "loss": 0.0216, "step": 24294 }, { "epoch": 4.46, "learning_rate": 8.64663251363257e-06, "loss": 0.0266, "step": 24295 }, { "epoch": 4.46, "learning_rate": 8.645896917946812e-06, "loss": 0.0287, "step": 24296 }, { "epoch": 4.46, "learning_rate": 8.645161329724903e-06, "loss": 0.0246, "step": 24297 }, { "epoch": 4.46, "learning_rate": 8.644425748970898e-06, "loss": 0.0253, "step": 24298 }, { "epoch": 4.46, "learning_rate": 8.643690175688844e-06, "loss": 0.0255, "step": 24299 }, { "epoch": 4.46, "learning_rate": 8.642954609882808e-06, "loss": 0.0227, "step": 24300 }, { "epoch": 4.46, "learning_rate": 8.642219051556834e-06, "loss": 0.0237, "step": 24301 }, { "epoch": 4.46, "learning_rate": 8.641483500714982e-06, "loss": 0.0269, "step": 24302 }, { "epoch": 4.46, "learning_rate": 8.640747957361303e-06, "loss": 0.0271, "step": 24303 }, { "epoch": 4.46, "learning_rate": 8.640012421499856e-06, "loss": 0.0242, "step": 24304 }, { "epoch": 4.46, "learning_rate": 8.639276893134689e-06, "loss": 0.0278, "step": 24305 }, { "epoch": 4.46, "learning_rate": 8.638541372269863e-06, "loss": 0.02, "step": 24306 }, { "epoch": 4.46, "learning_rate": 8.63780585890943e-06, "loss": 0.0226, "step": 24307 }, { "epoch": 4.46, "learning_rate": 8.637070353057436e-06, "loss": 0.0274, "step": 24308 }, { "epoch": 4.46, "learning_rate": 8.636334854717949e-06, "loss": 0.0287, "step": 24309 }, { "epoch": 4.46, "learning_rate": 8.635599363895013e-06, "loss": 0.0245, "step": 24310 }, { "epoch": 4.46, "learning_rate": 8.634863880592687e-06, "loss": 0.0206, "step": 24311 }, { "epoch": 4.46, "learning_rate": 8.63412840481502e-06, "loss": 0.0209, "step": 24312 }, { "epoch": 4.46, "learning_rate": 8.633392936566072e-06, "loss": 0.0253, "step": 24313 }, { "epoch": 4.46, "learning_rate": 8.632657475849895e-06, "loss": 0.0267, "step": 24314 }, { "epoch": 4.46, "learning_rate": 8.63192202267054e-06, "loss": 0.0274, "step": 24315 }, { "epoch": 4.46, "learning_rate": 8.63118657703206e-06, "loss": 0.0272, "step": 24316 }, { "epoch": 4.46, "learning_rate": 8.630451138938517e-06, "loss": 0.025, "step": 24317 }, { "epoch": 4.46, "learning_rate": 8.62971570839396e-06, "loss": 0.0249, "step": 24318 }, { "epoch": 4.46, "learning_rate": 8.62898028540244e-06, "loss": 0.0229, "step": 24319 }, { "epoch": 4.46, "learning_rate": 8.628244869968013e-06, "loss": 0.024, "step": 24320 }, { "epoch": 4.46, "learning_rate": 8.62750946209473e-06, "loss": 0.026, "step": 24321 }, { "epoch": 4.46, "learning_rate": 8.62677406178665e-06, "loss": 0.0233, "step": 24322 }, { "epoch": 4.46, "learning_rate": 8.626038669047822e-06, "loss": 0.0237, "step": 24323 }, { "epoch": 4.46, "learning_rate": 8.625303283882301e-06, "loss": 0.0277, "step": 24324 }, { "epoch": 4.46, "learning_rate": 8.624567906294139e-06, "loss": 0.0203, "step": 24325 }, { "epoch": 4.46, "learning_rate": 8.623832536287394e-06, "loss": 0.0242, "step": 24326 }, { "epoch": 4.46, "learning_rate": 8.623097173866117e-06, "loss": 0.0241, "step": 24327 }, { "epoch": 4.46, "learning_rate": 8.62236181903436e-06, "loss": 0.0234, "step": 24328 }, { "epoch": 4.46, "learning_rate": 8.621626471796176e-06, "loss": 0.0202, "step": 24329 }, { "epoch": 4.46, "learning_rate": 8.620891132155616e-06, "loss": 0.0252, "step": 24330 }, { "epoch": 4.46, "learning_rate": 8.62015580011674e-06, "loss": 0.0236, "step": 24331 }, { "epoch": 4.46, "learning_rate": 8.619420475683597e-06, "loss": 0.0251, "step": 24332 }, { "epoch": 4.46, "learning_rate": 8.618685158860241e-06, "loss": 0.0228, "step": 24333 }, { "epoch": 4.46, "learning_rate": 8.617949849650723e-06, "loss": 0.023, "step": 24334 }, { "epoch": 4.46, "learning_rate": 8.6172145480591e-06, "loss": 0.0202, "step": 24335 }, { "epoch": 4.46, "learning_rate": 8.616479254089424e-06, "loss": 0.0257, "step": 24336 }, { "epoch": 4.46, "learning_rate": 8.615743967745744e-06, "loss": 0.024, "step": 24337 }, { "epoch": 4.46, "learning_rate": 8.615008689032115e-06, "loss": 0.0248, "step": 24338 }, { "epoch": 4.46, "learning_rate": 8.614273417952593e-06, "loss": 0.0289, "step": 24339 }, { "epoch": 4.46, "learning_rate": 8.613538154511227e-06, "loss": 0.0235, "step": 24340 }, { "epoch": 4.46, "learning_rate": 8.612802898712071e-06, "loss": 0.0278, "step": 24341 }, { "epoch": 4.46, "learning_rate": 8.61206765055918e-06, "loss": 0.0238, "step": 24342 }, { "epoch": 4.46, "learning_rate": 8.611332410056601e-06, "loss": 0.0222, "step": 24343 }, { "epoch": 4.46, "learning_rate": 8.610597177208393e-06, "loss": 0.0249, "step": 24344 }, { "epoch": 4.46, "learning_rate": 8.609861952018607e-06, "loss": 0.0243, "step": 24345 }, { "epoch": 4.46, "learning_rate": 8.609126734491293e-06, "loss": 0.024, "step": 24346 }, { "epoch": 4.46, "learning_rate": 8.608391524630504e-06, "loss": 0.0219, "step": 24347 }, { "epoch": 4.47, "learning_rate": 8.607656322440295e-06, "loss": 0.0237, "step": 24348 }, { "epoch": 4.47, "learning_rate": 8.606921127924716e-06, "loss": 0.0239, "step": 24349 }, { "epoch": 4.47, "learning_rate": 8.60618594108782e-06, "loss": 0.0239, "step": 24350 }, { "epoch": 4.47, "learning_rate": 8.60545076193366e-06, "loss": 0.0207, "step": 24351 }, { "epoch": 4.47, "learning_rate": 8.604715590466287e-06, "loss": 0.0274, "step": 24352 }, { "epoch": 4.47, "learning_rate": 8.603980426689757e-06, "loss": 0.0178, "step": 24353 }, { "epoch": 4.47, "learning_rate": 8.60324527060812e-06, "loss": 0.0249, "step": 24354 }, { "epoch": 4.47, "learning_rate": 8.602510122225426e-06, "loss": 0.018, "step": 24355 }, { "epoch": 4.47, "learning_rate": 8.601774981545728e-06, "loss": 0.0269, "step": 24356 }, { "epoch": 4.47, "learning_rate": 8.601039848573081e-06, "loss": 0.0278, "step": 24357 }, { "epoch": 4.47, "learning_rate": 8.600304723311534e-06, "loss": 0.0246, "step": 24358 }, { "epoch": 4.47, "learning_rate": 8.599569605765141e-06, "loss": 0.025, "step": 24359 }, { "epoch": 4.47, "learning_rate": 8.598834495937956e-06, "loss": 0.022, "step": 24360 }, { "epoch": 4.47, "learning_rate": 8.598099393834019e-06, "loss": 0.0246, "step": 24361 }, { "epoch": 4.47, "learning_rate": 8.597364299457397e-06, "loss": 0.0238, "step": 24362 }, { "epoch": 4.47, "learning_rate": 8.596629212812138e-06, "loss": 0.0221, "step": 24363 }, { "epoch": 4.47, "learning_rate": 8.59589413390229e-06, "loss": 0.028, "step": 24364 }, { "epoch": 4.47, "learning_rate": 8.595159062731904e-06, "loss": 0.0195, "step": 24365 }, { "epoch": 4.47, "learning_rate": 8.594423999305036e-06, "loss": 0.0194, "step": 24366 }, { "epoch": 4.47, "learning_rate": 8.593688943625736e-06, "loss": 0.0268, "step": 24367 }, { "epoch": 4.47, "learning_rate": 8.592953895698055e-06, "loss": 0.0215, "step": 24368 }, { "epoch": 4.47, "learning_rate": 8.592218855526043e-06, "loss": 0.0228, "step": 24369 }, { "epoch": 4.47, "learning_rate": 8.591483823113756e-06, "loss": 0.0266, "step": 24370 }, { "epoch": 4.47, "learning_rate": 8.590748798465244e-06, "loss": 0.0275, "step": 24371 }, { "epoch": 4.47, "learning_rate": 8.59001378158456e-06, "loss": 0.0223, "step": 24372 }, { "epoch": 4.47, "learning_rate": 8.58927877247575e-06, "loss": 0.0213, "step": 24373 }, { "epoch": 4.47, "learning_rate": 8.588543771142866e-06, "loss": 0.0283, "step": 24374 }, { "epoch": 4.47, "learning_rate": 8.587808777589964e-06, "loss": 0.0212, "step": 24375 }, { "epoch": 4.47, "learning_rate": 8.587073791821094e-06, "loss": 0.0286, "step": 24376 }, { "epoch": 4.47, "learning_rate": 8.586338813840304e-06, "loss": 0.0246, "step": 24377 }, { "epoch": 4.47, "learning_rate": 8.585603843651649e-06, "loss": 0.022, "step": 24378 }, { "epoch": 4.47, "learning_rate": 8.58486888125918e-06, "loss": 0.0273, "step": 24379 }, { "epoch": 4.47, "learning_rate": 8.584133926666945e-06, "loss": 0.0245, "step": 24380 }, { "epoch": 4.47, "learning_rate": 8.583398979878999e-06, "loss": 0.0243, "step": 24381 }, { "epoch": 4.47, "learning_rate": 8.58266404089939e-06, "loss": 0.0279, "step": 24382 }, { "epoch": 4.47, "learning_rate": 8.581929109732167e-06, "loss": 0.0201, "step": 24383 }, { "epoch": 4.47, "learning_rate": 8.581194186381388e-06, "loss": 0.0266, "step": 24384 }, { "epoch": 4.47, "learning_rate": 8.580459270851099e-06, "loss": 0.0232, "step": 24385 }, { "epoch": 4.47, "learning_rate": 8.579724363145351e-06, "loss": 0.0249, "step": 24386 }, { "epoch": 4.47, "learning_rate": 8.578989463268194e-06, "loss": 0.0244, "step": 24387 }, { "epoch": 4.47, "learning_rate": 8.578254571223683e-06, "loss": 0.0243, "step": 24388 }, { "epoch": 4.47, "learning_rate": 8.577519687015864e-06, "loss": 0.0248, "step": 24389 }, { "epoch": 4.47, "learning_rate": 8.576784810648795e-06, "loss": 0.0232, "step": 24390 }, { "epoch": 4.47, "learning_rate": 8.576049942126514e-06, "loss": 0.0235, "step": 24391 }, { "epoch": 4.47, "learning_rate": 8.575315081453084e-06, "loss": 0.0244, "step": 24392 }, { "epoch": 4.47, "learning_rate": 8.574580228632549e-06, "loss": 0.0262, "step": 24393 }, { "epoch": 4.47, "learning_rate": 8.573845383668963e-06, "loss": 0.0237, "step": 24394 }, { "epoch": 4.47, "learning_rate": 8.573110546566373e-06, "loss": 0.025, "step": 24395 }, { "epoch": 4.47, "learning_rate": 8.57237571732883e-06, "loss": 0.0247, "step": 24396 }, { "epoch": 4.47, "learning_rate": 8.571640895960386e-06, "loss": 0.0242, "step": 24397 }, { "epoch": 4.47, "learning_rate": 8.570906082465094e-06, "loss": 0.025, "step": 24398 }, { "epoch": 4.47, "learning_rate": 8.570171276846998e-06, "loss": 0.0278, "step": 24399 }, { "epoch": 4.47, "learning_rate": 8.56943647911015e-06, "loss": 0.0265, "step": 24400 }, { "epoch": 4.47, "learning_rate": 8.568701689258606e-06, "loss": 0.025, "step": 24401 }, { "epoch": 4.47, "learning_rate": 8.56796690729641e-06, "loss": 0.0258, "step": 24402 }, { "epoch": 4.48, "learning_rate": 8.567232133227614e-06, "loss": 0.022, "step": 24403 }, { "epoch": 4.48, "learning_rate": 8.566497367056268e-06, "loss": 0.0271, "step": 24404 }, { "epoch": 4.48, "learning_rate": 8.56576260878642e-06, "loss": 0.0234, "step": 24405 }, { "epoch": 4.48, "learning_rate": 8.565027858422124e-06, "loss": 0.0212, "step": 24406 }, { "epoch": 4.48, "learning_rate": 8.564293115967431e-06, "loss": 0.0301, "step": 24407 }, { "epoch": 4.48, "learning_rate": 8.563558381426384e-06, "loss": 0.0247, "step": 24408 }, { "epoch": 4.48, "learning_rate": 8.562823654803035e-06, "loss": 0.0235, "step": 24409 }, { "epoch": 4.48, "learning_rate": 8.56208893610144e-06, "loss": 0.0222, "step": 24410 }, { "epoch": 4.48, "learning_rate": 8.561354225325643e-06, "loss": 0.0207, "step": 24411 }, { "epoch": 4.48, "learning_rate": 8.560619522479694e-06, "loss": 0.0193, "step": 24412 }, { "epoch": 4.48, "learning_rate": 8.559884827567644e-06, "loss": 0.0255, "step": 24413 }, { "epoch": 4.48, "learning_rate": 8.559150140593543e-06, "loss": 0.0225, "step": 24414 }, { "epoch": 4.48, "learning_rate": 8.558415461561442e-06, "loss": 0.0205, "step": 24415 }, { "epoch": 4.48, "learning_rate": 8.55768079047539e-06, "loss": 0.0246, "step": 24416 }, { "epoch": 4.48, "learning_rate": 8.556946127339434e-06, "loss": 0.0222, "step": 24417 }, { "epoch": 4.48, "learning_rate": 8.556211472157618e-06, "loss": 0.0232, "step": 24418 }, { "epoch": 4.48, "learning_rate": 8.555476824934006e-06, "loss": 0.0215, "step": 24419 }, { "epoch": 4.48, "learning_rate": 8.554742185672637e-06, "loss": 0.0279, "step": 24420 }, { "epoch": 4.48, "learning_rate": 8.554007554377564e-06, "loss": 0.026, "step": 24421 }, { "epoch": 4.48, "learning_rate": 8.553272931052832e-06, "loss": 0.0219, "step": 24422 }, { "epoch": 4.48, "learning_rate": 8.552538315702497e-06, "loss": 0.0216, "step": 24423 }, { "epoch": 4.48, "learning_rate": 8.551803708330604e-06, "loss": 0.0281, "step": 24424 }, { "epoch": 4.48, "learning_rate": 8.551069108941206e-06, "loss": 0.026, "step": 24425 }, { "epoch": 4.48, "learning_rate": 8.550334517538345e-06, "loss": 0.0229, "step": 24426 }, { "epoch": 4.48, "learning_rate": 8.54959993412607e-06, "loss": 0.025, "step": 24427 }, { "epoch": 4.48, "learning_rate": 8.548865358708443e-06, "loss": 0.0191, "step": 24428 }, { "epoch": 4.48, "learning_rate": 8.5481307912895e-06, "loss": 0.0266, "step": 24429 }, { "epoch": 4.48, "learning_rate": 8.547396231873294e-06, "loss": 0.0271, "step": 24430 }, { "epoch": 4.48, "learning_rate": 8.546661680463873e-06, "loss": 0.0224, "step": 24431 }, { "epoch": 4.48, "learning_rate": 8.545927137065287e-06, "loss": 0.0231, "step": 24432 }, { "epoch": 4.48, "learning_rate": 8.545192601681586e-06, "loss": 0.0253, "step": 24433 }, { "epoch": 4.48, "learning_rate": 8.544458074316819e-06, "loss": 0.0266, "step": 24434 }, { "epoch": 4.48, "learning_rate": 8.543723554975032e-06, "loss": 0.0271, "step": 24435 }, { "epoch": 4.48, "learning_rate": 8.54298904366027e-06, "loss": 0.0233, "step": 24436 }, { "epoch": 4.48, "learning_rate": 8.542254540376594e-06, "loss": 0.0289, "step": 24437 }, { "epoch": 4.48, "learning_rate": 8.541520045128041e-06, "loss": 0.0248, "step": 24438 }, { "epoch": 4.48, "learning_rate": 8.540785557918664e-06, "loss": 0.0232, "step": 24439 }, { "epoch": 4.48, "learning_rate": 8.54005107875251e-06, "loss": 0.0268, "step": 24440 }, { "epoch": 4.48, "learning_rate": 8.539316607633632e-06, "loss": 0.0267, "step": 24441 }, { "epoch": 4.48, "learning_rate": 8.538582144566073e-06, "loss": 0.0212, "step": 24442 }, { "epoch": 4.48, "learning_rate": 8.537847689553886e-06, "loss": 0.0295, "step": 24443 }, { "epoch": 4.48, "learning_rate": 8.53711324260111e-06, "loss": 0.0232, "step": 24444 }, { "epoch": 4.48, "learning_rate": 8.536378803711808e-06, "loss": 0.026, "step": 24445 }, { "epoch": 4.48, "learning_rate": 8.535644372890018e-06, "loss": 0.0261, "step": 24446 }, { "epoch": 4.48, "learning_rate": 8.53490995013979e-06, "loss": 0.0248, "step": 24447 }, { "epoch": 4.48, "learning_rate": 8.534175535465173e-06, "loss": 0.0225, "step": 24448 }, { "epoch": 4.48, "learning_rate": 8.53344112887021e-06, "loss": 0.0226, "step": 24449 }, { "epoch": 4.48, "learning_rate": 8.53270673035896e-06, "loss": 0.0288, "step": 24450 }, { "epoch": 4.48, "learning_rate": 8.531972339935463e-06, "loss": 0.0246, "step": 24451 }, { "epoch": 4.48, "learning_rate": 8.531237957603771e-06, "loss": 0.0278, "step": 24452 }, { "epoch": 4.48, "learning_rate": 8.530503583367922e-06, "loss": 0.0204, "step": 24453 }, { "epoch": 4.48, "learning_rate": 8.52976921723198e-06, "loss": 0.0253, "step": 24454 }, { "epoch": 4.48, "learning_rate": 8.529034859199981e-06, "loss": 0.0224, "step": 24455 }, { "epoch": 4.48, "learning_rate": 8.528300509275979e-06, "loss": 0.0225, "step": 24456 }, { "epoch": 4.49, "learning_rate": 8.527566167464016e-06, "loss": 0.0252, "step": 24457 }, { "epoch": 4.49, "learning_rate": 8.526831833768143e-06, "loss": 0.0215, "step": 24458 }, { "epoch": 4.49, "learning_rate": 8.526097508192407e-06, "loss": 0.0218, "step": 24459 }, { "epoch": 4.49, "learning_rate": 8.52536319074086e-06, "loss": 0.0248, "step": 24460 }, { "epoch": 4.49, "learning_rate": 8.524628881417543e-06, "loss": 0.0229, "step": 24461 }, { "epoch": 4.49, "learning_rate": 8.523894580226503e-06, "loss": 0.0249, "step": 24462 }, { "epoch": 4.49, "learning_rate": 8.523160287171797e-06, "loss": 0.0249, "step": 24463 }, { "epoch": 4.49, "learning_rate": 8.522426002257462e-06, "loss": 0.0263, "step": 24464 }, { "epoch": 4.49, "learning_rate": 8.52169172548755e-06, "loss": 0.0239, "step": 24465 }, { "epoch": 4.49, "learning_rate": 8.520957456866107e-06, "loss": 0.023, "step": 24466 }, { "epoch": 4.49, "learning_rate": 8.520223196397182e-06, "loss": 0.0179, "step": 24467 }, { "epoch": 4.49, "learning_rate": 8.519488944084823e-06, "loss": 0.0219, "step": 24468 }, { "epoch": 4.49, "learning_rate": 8.518754699933073e-06, "loss": 0.0256, "step": 24469 }, { "epoch": 4.49, "learning_rate": 8.518020463945987e-06, "loss": 0.0252, "step": 24470 }, { "epoch": 4.49, "learning_rate": 8.517286236127599e-06, "loss": 0.0264, "step": 24471 }, { "epoch": 4.49, "learning_rate": 8.51655201648197e-06, "loss": 0.0206, "step": 24472 }, { "epoch": 4.49, "learning_rate": 8.51581780501314e-06, "loss": 0.0301, "step": 24473 }, { "epoch": 4.49, "learning_rate": 8.515083601725156e-06, "loss": 0.028, "step": 24474 }, { "epoch": 4.49, "learning_rate": 8.514349406622064e-06, "loss": 0.0255, "step": 24475 }, { "epoch": 4.49, "learning_rate": 8.513615219707916e-06, "loss": 0.0211, "step": 24476 }, { "epoch": 4.49, "learning_rate": 8.512881040986755e-06, "loss": 0.0211, "step": 24477 }, { "epoch": 4.49, "learning_rate": 8.51214687046263e-06, "loss": 0.0273, "step": 24478 }, { "epoch": 4.49, "learning_rate": 8.511412708139585e-06, "loss": 0.0243, "step": 24479 }, { "epoch": 4.49, "learning_rate": 8.510678554021663e-06, "loss": 0.0255, "step": 24480 }, { "epoch": 4.49, "learning_rate": 8.509944408112924e-06, "loss": 0.0288, "step": 24481 }, { "epoch": 4.49, "learning_rate": 8.509210270417403e-06, "loss": 0.023, "step": 24482 }, { "epoch": 4.49, "learning_rate": 8.50847614093915e-06, "loss": 0.0289, "step": 24483 }, { "epoch": 4.49, "learning_rate": 8.507742019682208e-06, "loss": 0.021, "step": 24484 }, { "epoch": 4.49, "learning_rate": 8.50700790665063e-06, "loss": 0.0183, "step": 24485 }, { "epoch": 4.49, "learning_rate": 8.50627380184846e-06, "loss": 0.0205, "step": 24486 }, { "epoch": 4.49, "learning_rate": 8.505539705279744e-06, "loss": 0.0251, "step": 24487 }, { "epoch": 4.49, "learning_rate": 8.50480561694853e-06, "loss": 0.0172, "step": 24488 }, { "epoch": 4.49, "learning_rate": 8.504071536858856e-06, "loss": 0.0233, "step": 24489 }, { "epoch": 4.49, "learning_rate": 8.503337465014781e-06, "loss": 0.024, "step": 24490 }, { "epoch": 4.49, "learning_rate": 8.502603401420342e-06, "loss": 0.0169, "step": 24491 }, { "epoch": 4.49, "learning_rate": 8.50186934607959e-06, "loss": 0.0262, "step": 24492 }, { "epoch": 4.49, "learning_rate": 8.501135298996565e-06, "loss": 0.0225, "step": 24493 }, { "epoch": 4.49, "learning_rate": 8.500401260175321e-06, "loss": 0.025, "step": 24494 }, { "epoch": 4.49, "learning_rate": 8.499667229619899e-06, "loss": 0.0243, "step": 24495 }, { "epoch": 4.49, "learning_rate": 8.498933207334348e-06, "loss": 0.0215, "step": 24496 }, { "epoch": 4.49, "learning_rate": 8.498199193322709e-06, "loss": 0.0246, "step": 24497 }, { "epoch": 4.49, "learning_rate": 8.497465187589033e-06, "loss": 0.0227, "step": 24498 }, { "epoch": 4.49, "learning_rate": 8.496731190137368e-06, "loss": 0.0225, "step": 24499 }, { "epoch": 4.49, "learning_rate": 8.495997200971753e-06, "loss": 0.0233, "step": 24500 }, { "epoch": 4.49, "learning_rate": 8.495263220096236e-06, "loss": 0.0259, "step": 24501 }, { "epoch": 4.49, "learning_rate": 8.494529247514863e-06, "loss": 0.0251, "step": 24502 }, { "epoch": 4.49, "learning_rate": 8.49379528323168e-06, "loss": 0.0242, "step": 24503 }, { "epoch": 4.49, "learning_rate": 8.493061327250734e-06, "loss": 0.0241, "step": 24504 }, { "epoch": 4.49, "learning_rate": 8.492327379576069e-06, "loss": 0.0285, "step": 24505 }, { "epoch": 4.49, "learning_rate": 8.491593440211728e-06, "loss": 0.0246, "step": 24506 }, { "epoch": 4.49, "learning_rate": 8.490859509161762e-06, "loss": 0.0189, "step": 24507 }, { "epoch": 4.49, "learning_rate": 8.490125586430215e-06, "loss": 0.0181, "step": 24508 }, { "epoch": 4.49, "learning_rate": 8.48939167202113e-06, "loss": 0.0284, "step": 24509 }, { "epoch": 4.49, "learning_rate": 8.488657765938555e-06, "loss": 0.0222, "step": 24510 }, { "epoch": 4.49, "learning_rate": 8.487923868186528e-06, "loss": 0.0232, "step": 24511 }, { "epoch": 4.5, "learning_rate": 8.487189978769106e-06, "loss": 0.0223, "step": 24512 }, { "epoch": 4.5, "learning_rate": 8.486456097690326e-06, "loss": 0.0241, "step": 24513 }, { "epoch": 4.5, "learning_rate": 8.485722224954237e-06, "loss": 0.0249, "step": 24514 }, { "epoch": 4.5, "learning_rate": 8.484988360564878e-06, "loss": 0.0238, "step": 24515 }, { "epoch": 4.5, "learning_rate": 8.484254504526302e-06, "loss": 0.0235, "step": 24516 }, { "epoch": 4.5, "learning_rate": 8.483520656842552e-06, "loss": 0.0225, "step": 24517 }, { "epoch": 4.5, "learning_rate": 8.48278681751767e-06, "loss": 0.0268, "step": 24518 }, { "epoch": 4.5, "learning_rate": 8.482052986555703e-06, "loss": 0.0234, "step": 24519 }, { "epoch": 4.5, "learning_rate": 8.481319163960694e-06, "loss": 0.0263, "step": 24520 }, { "epoch": 4.5, "learning_rate": 8.480585349736692e-06, "loss": 0.0253, "step": 24521 }, { "epoch": 4.5, "learning_rate": 8.479851543887737e-06, "loss": 0.0259, "step": 24522 }, { "epoch": 4.5, "learning_rate": 8.479117746417879e-06, "loss": 0.0229, "step": 24523 }, { "epoch": 4.5, "learning_rate": 8.478383957331154e-06, "loss": 0.021, "step": 24524 }, { "epoch": 4.5, "learning_rate": 8.477650176631616e-06, "loss": 0.0261, "step": 24525 }, { "epoch": 4.5, "learning_rate": 8.476916404323307e-06, "loss": 0.0278, "step": 24526 }, { "epoch": 4.5, "learning_rate": 8.47618264041027e-06, "loss": 0.0257, "step": 24527 }, { "epoch": 4.5, "learning_rate": 8.475448884896546e-06, "loss": 0.0264, "step": 24528 }, { "epoch": 4.5, "learning_rate": 8.474715137786187e-06, "loss": 0.0254, "step": 24529 }, { "epoch": 4.5, "learning_rate": 8.473981399083234e-06, "loss": 0.0277, "step": 24530 }, { "epoch": 4.5, "learning_rate": 8.47324766879173e-06, "loss": 0.0221, "step": 24531 }, { "epoch": 4.5, "learning_rate": 8.472513946915724e-06, "loss": 0.0275, "step": 24532 }, { "epoch": 4.5, "learning_rate": 8.47178023345925e-06, "loss": 0.0249, "step": 24533 }, { "epoch": 4.5, "learning_rate": 8.471046528426364e-06, "loss": 0.0234, "step": 24534 }, { "epoch": 4.5, "learning_rate": 8.470312831821107e-06, "loss": 0.0301, "step": 24535 }, { "epoch": 4.5, "learning_rate": 8.46957914364752e-06, "loss": 0.024, "step": 24536 }, { "epoch": 4.5, "learning_rate": 8.468845463909644e-06, "loss": 0.029, "step": 24537 }, { "epoch": 4.5, "learning_rate": 8.468111792611532e-06, "loss": 0.023, "step": 24538 }, { "epoch": 4.5, "learning_rate": 8.467378129757224e-06, "loss": 0.0232, "step": 24539 }, { "epoch": 4.5, "learning_rate": 8.466644475350763e-06, "loss": 0.0182, "step": 24540 }, { "epoch": 4.5, "learning_rate": 8.465910829396192e-06, "loss": 0.0259, "step": 24541 }, { "epoch": 4.5, "learning_rate": 8.465177191897557e-06, "loss": 0.0263, "step": 24542 }, { "epoch": 4.5, "learning_rate": 8.464443562858903e-06, "loss": 0.0246, "step": 24543 }, { "epoch": 4.5, "learning_rate": 8.463709942284273e-06, "loss": 0.0312, "step": 24544 }, { "epoch": 4.5, "learning_rate": 8.46297633017771e-06, "loss": 0.0212, "step": 24545 }, { "epoch": 4.5, "learning_rate": 8.462242726543252e-06, "loss": 0.0291, "step": 24546 }, { "epoch": 4.5, "learning_rate": 8.461509131384952e-06, "loss": 0.0269, "step": 24547 }, { "epoch": 4.5, "learning_rate": 8.46077554470685e-06, "loss": 0.024, "step": 24548 }, { "epoch": 4.5, "learning_rate": 8.460041966512988e-06, "loss": 0.0229, "step": 24549 }, { "epoch": 4.5, "learning_rate": 8.45930839680741e-06, "loss": 0.0225, "step": 24550 }, { "epoch": 4.5, "learning_rate": 8.458574835594162e-06, "loss": 0.0213, "step": 24551 }, { "epoch": 4.5, "learning_rate": 8.457841282877284e-06, "loss": 0.0254, "step": 24552 }, { "epoch": 4.5, "learning_rate": 8.457107738660826e-06, "loss": 0.0249, "step": 24553 }, { "epoch": 4.5, "learning_rate": 8.456374202948821e-06, "loss": 0.0267, "step": 24554 }, { "epoch": 4.5, "learning_rate": 8.455640675745318e-06, "loss": 0.0251, "step": 24555 }, { "epoch": 4.5, "learning_rate": 8.45490715705436e-06, "loss": 0.0252, "step": 24556 }, { "epoch": 4.5, "learning_rate": 8.454173646879992e-06, "loss": 0.0221, "step": 24557 }, { "epoch": 4.5, "learning_rate": 8.453440145226254e-06, "loss": 0.0203, "step": 24558 }, { "epoch": 4.5, "learning_rate": 8.452706652097187e-06, "loss": 0.0222, "step": 24559 }, { "epoch": 4.5, "learning_rate": 8.451973167496841e-06, "loss": 0.0267, "step": 24560 }, { "epoch": 4.5, "learning_rate": 8.451239691429256e-06, "loss": 0.0292, "step": 24561 }, { "epoch": 4.5, "learning_rate": 8.450506223898473e-06, "loss": 0.0203, "step": 24562 }, { "epoch": 4.5, "learning_rate": 8.449772764908534e-06, "loss": 0.0258, "step": 24563 }, { "epoch": 4.5, "learning_rate": 8.449039314463485e-06, "loss": 0.0269, "step": 24564 }, { "epoch": 4.5, "learning_rate": 8.448305872567367e-06, "loss": 0.0335, "step": 24565 }, { "epoch": 4.51, "learning_rate": 8.447572439224226e-06, "loss": 0.0268, "step": 24566 }, { "epoch": 4.51, "learning_rate": 8.4468390144381e-06, "loss": 0.0259, "step": 24567 }, { "epoch": 4.51, "learning_rate": 8.446105598213032e-06, "loss": 0.0232, "step": 24568 }, { "epoch": 4.51, "learning_rate": 8.445372190553069e-06, "loss": 0.022, "step": 24569 }, { "epoch": 4.51, "learning_rate": 8.444638791462252e-06, "loss": 0.023, "step": 24570 }, { "epoch": 4.51, "learning_rate": 8.44390540094462e-06, "loss": 0.0274, "step": 24571 }, { "epoch": 4.51, "learning_rate": 8.443172019004215e-06, "loss": 0.0238, "step": 24572 }, { "epoch": 4.51, "learning_rate": 8.442438645645089e-06, "loss": 0.0251, "step": 24573 }, { "epoch": 4.51, "learning_rate": 8.441705280871275e-06, "loss": 0.0225, "step": 24574 }, { "epoch": 4.51, "learning_rate": 8.440971924686817e-06, "loss": 0.023, "step": 24575 }, { "epoch": 4.51, "learning_rate": 8.440238577095758e-06, "loss": 0.0249, "step": 24576 }, { "epoch": 4.51, "learning_rate": 8.43950523810214e-06, "loss": 0.0212, "step": 24577 }, { "epoch": 4.51, "learning_rate": 8.438771907710007e-06, "loss": 0.0175, "step": 24578 }, { "epoch": 4.51, "learning_rate": 8.438038585923403e-06, "loss": 0.0243, "step": 24579 }, { "epoch": 4.51, "learning_rate": 8.437305272746364e-06, "loss": 0.0216, "step": 24580 }, { "epoch": 4.51, "learning_rate": 8.43657196818293e-06, "loss": 0.0284, "step": 24581 }, { "epoch": 4.51, "learning_rate": 8.435838672237155e-06, "loss": 0.0252, "step": 24582 }, { "epoch": 4.51, "learning_rate": 8.435105384913071e-06, "loss": 0.0296, "step": 24583 }, { "epoch": 4.51, "learning_rate": 8.434372106214725e-06, "loss": 0.0223, "step": 24584 }, { "epoch": 4.51, "learning_rate": 8.433638836146154e-06, "loss": 0.0208, "step": 24585 }, { "epoch": 4.51, "learning_rate": 8.432905574711401e-06, "loss": 0.0259, "step": 24586 }, { "epoch": 4.51, "learning_rate": 8.432172321914513e-06, "loss": 0.0241, "step": 24587 }, { "epoch": 4.51, "learning_rate": 8.431439077759529e-06, "loss": 0.0245, "step": 24588 }, { "epoch": 4.51, "learning_rate": 8.430705842250487e-06, "loss": 0.0239, "step": 24589 }, { "epoch": 4.51, "learning_rate": 8.429972615391426e-06, "loss": 0.0221, "step": 24590 }, { "epoch": 4.51, "learning_rate": 8.4292393971864e-06, "loss": 0.0217, "step": 24591 }, { "epoch": 4.51, "learning_rate": 8.428506187639443e-06, "loss": 0.0213, "step": 24592 }, { "epoch": 4.51, "learning_rate": 8.427772986754594e-06, "loss": 0.0298, "step": 24593 }, { "epoch": 4.51, "learning_rate": 8.427039794535896e-06, "loss": 0.0218, "step": 24594 }, { "epoch": 4.51, "learning_rate": 8.426306610987393e-06, "loss": 0.0216, "step": 24595 }, { "epoch": 4.51, "learning_rate": 8.425573436113126e-06, "loss": 0.0199, "step": 24596 }, { "epoch": 4.51, "learning_rate": 8.424840269917137e-06, "loss": 0.0247, "step": 24597 }, { "epoch": 4.51, "learning_rate": 8.424107112403463e-06, "loss": 0.0255, "step": 24598 }, { "epoch": 4.51, "learning_rate": 8.423373963576145e-06, "loss": 0.0224, "step": 24599 }, { "epoch": 4.51, "learning_rate": 8.42264082343923e-06, "loss": 0.0295, "step": 24600 }, { "epoch": 4.51, "learning_rate": 8.421907691996755e-06, "loss": 0.029, "step": 24601 }, { "epoch": 4.51, "learning_rate": 8.421174569252762e-06, "loss": 0.0315, "step": 24602 }, { "epoch": 4.51, "learning_rate": 8.420441455211288e-06, "loss": 0.0236, "step": 24603 }, { "epoch": 4.51, "learning_rate": 8.419708349876382e-06, "loss": 0.0283, "step": 24604 }, { "epoch": 4.51, "learning_rate": 8.418975253252079e-06, "loss": 0.0249, "step": 24605 }, { "epoch": 4.51, "learning_rate": 8.418242165342424e-06, "loss": 0.0253, "step": 24606 }, { "epoch": 4.51, "learning_rate": 8.417509086151453e-06, "loss": 0.0276, "step": 24607 }, { "epoch": 4.51, "learning_rate": 8.416776015683208e-06, "loss": 0.0201, "step": 24608 }, { "epoch": 4.51, "learning_rate": 8.416042953941733e-06, "loss": 0.0266, "step": 24609 }, { "epoch": 4.51, "learning_rate": 8.415309900931063e-06, "loss": 0.0205, "step": 24610 }, { "epoch": 4.51, "learning_rate": 8.414576856655245e-06, "loss": 0.0239, "step": 24611 }, { "epoch": 4.51, "learning_rate": 8.413843821118313e-06, "loss": 0.0279, "step": 24612 }, { "epoch": 4.51, "learning_rate": 8.413110794324314e-06, "loss": 0.0256, "step": 24613 }, { "epoch": 4.51, "learning_rate": 8.412377776277286e-06, "loss": 0.0225, "step": 24614 }, { "epoch": 4.51, "learning_rate": 8.41164476698127e-06, "loss": 0.0237, "step": 24615 }, { "epoch": 4.51, "learning_rate": 8.410911766440305e-06, "loss": 0.0249, "step": 24616 }, { "epoch": 4.51, "learning_rate": 8.410178774658426e-06, "loss": 0.0249, "step": 24617 }, { "epoch": 4.51, "learning_rate": 8.409445791639684e-06, "loss": 0.0287, "step": 24618 }, { "epoch": 4.51, "learning_rate": 8.408712817388113e-06, "loss": 0.029, "step": 24619 }, { "epoch": 4.51, "learning_rate": 8.407979851907753e-06, "loss": 0.0241, "step": 24620 }, { "epoch": 4.52, "learning_rate": 8.407246895202646e-06, "loss": 0.0245, "step": 24621 }, { "epoch": 4.52, "learning_rate": 8.40651394727683e-06, "loss": 0.0276, "step": 24622 }, { "epoch": 4.52, "learning_rate": 8.40578100813435e-06, "loss": 0.0229, "step": 24623 }, { "epoch": 4.52, "learning_rate": 8.405048077779244e-06, "loss": 0.0186, "step": 24624 }, { "epoch": 4.52, "learning_rate": 8.404315156215542e-06, "loss": 0.025, "step": 24625 }, { "epoch": 4.52, "learning_rate": 8.403582243447299e-06, "loss": 0.0285, "step": 24626 }, { "epoch": 4.52, "learning_rate": 8.402849339478546e-06, "loss": 0.026, "step": 24627 }, { "epoch": 4.52, "learning_rate": 8.402116444313325e-06, "loss": 0.0259, "step": 24628 }, { "epoch": 4.52, "learning_rate": 8.401383557955677e-06, "loss": 0.0247, "step": 24629 }, { "epoch": 4.52, "learning_rate": 8.400650680409636e-06, "loss": 0.0282, "step": 24630 }, { "epoch": 4.52, "learning_rate": 8.39991781167925e-06, "loss": 0.0237, "step": 24631 }, { "epoch": 4.52, "learning_rate": 8.399184951768555e-06, "loss": 0.0205, "step": 24632 }, { "epoch": 4.52, "learning_rate": 8.39845210068159e-06, "loss": 0.0238, "step": 24633 }, { "epoch": 4.52, "learning_rate": 8.39771925842239e-06, "loss": 0.0248, "step": 24634 }, { "epoch": 4.52, "learning_rate": 8.396986424995003e-06, "loss": 0.0204, "step": 24635 }, { "epoch": 4.52, "learning_rate": 8.396253600403463e-06, "loss": 0.0244, "step": 24636 }, { "epoch": 4.52, "learning_rate": 8.39552078465181e-06, "loss": 0.0231, "step": 24637 }, { "epoch": 4.52, "learning_rate": 8.394787977744088e-06, "loss": 0.0228, "step": 24638 }, { "epoch": 4.52, "learning_rate": 8.394055179684328e-06, "loss": 0.0218, "step": 24639 }, { "epoch": 4.52, "learning_rate": 8.393322390476574e-06, "loss": 0.021, "step": 24640 }, { "epoch": 4.52, "learning_rate": 8.392589610124865e-06, "loss": 0.0227, "step": 24641 }, { "epoch": 4.52, "learning_rate": 8.391856838633243e-06, "loss": 0.0244, "step": 24642 }, { "epoch": 4.52, "learning_rate": 8.391124076005737e-06, "loss": 0.0204, "step": 24643 }, { "epoch": 4.52, "learning_rate": 8.3903913222464e-06, "loss": 0.0254, "step": 24644 }, { "epoch": 4.52, "learning_rate": 8.389658577359262e-06, "loss": 0.0256, "step": 24645 }, { "epoch": 4.52, "learning_rate": 8.38892584134836e-06, "loss": 0.0224, "step": 24646 }, { "epoch": 4.52, "learning_rate": 8.388193114217737e-06, "loss": 0.0238, "step": 24647 }, { "epoch": 4.52, "learning_rate": 8.387460395971434e-06, "loss": 0.025, "step": 24648 }, { "epoch": 4.52, "learning_rate": 8.386727686613487e-06, "loss": 0.0285, "step": 24649 }, { "epoch": 4.52, "learning_rate": 8.385994986147934e-06, "loss": 0.0265, "step": 24650 }, { "epoch": 4.52, "learning_rate": 8.385262294578816e-06, "loss": 0.0259, "step": 24651 }, { "epoch": 4.52, "learning_rate": 8.384529611910164e-06, "loss": 0.0191, "step": 24652 }, { "epoch": 4.52, "learning_rate": 8.38379693814603e-06, "loss": 0.0263, "step": 24653 }, { "epoch": 4.52, "learning_rate": 8.383064273290441e-06, "loss": 0.0214, "step": 24654 }, { "epoch": 4.52, "learning_rate": 8.382331617347442e-06, "loss": 0.0264, "step": 24655 }, { "epoch": 4.52, "learning_rate": 8.381598970321067e-06, "loss": 0.0261, "step": 24656 }, { "epoch": 4.52, "learning_rate": 8.380866332215356e-06, "loss": 0.0235, "step": 24657 }, { "epoch": 4.52, "learning_rate": 8.38013370303435e-06, "loss": 0.0304, "step": 24658 }, { "epoch": 4.52, "learning_rate": 8.379401082782084e-06, "loss": 0.0258, "step": 24659 }, { "epoch": 4.52, "learning_rate": 8.378668471462598e-06, "loss": 0.0188, "step": 24660 }, { "epoch": 4.52, "learning_rate": 8.377935869079925e-06, "loss": 0.03, "step": 24661 }, { "epoch": 4.52, "learning_rate": 8.377203275638114e-06, "loss": 0.025, "step": 24662 }, { "epoch": 4.52, "learning_rate": 8.376470691141193e-06, "loss": 0.0233, "step": 24663 }, { "epoch": 4.52, "learning_rate": 8.375738115593205e-06, "loss": 0.0202, "step": 24664 }, { "epoch": 4.52, "learning_rate": 8.375005548998185e-06, "loss": 0.0216, "step": 24665 }, { "epoch": 4.52, "learning_rate": 8.374272991360172e-06, "loss": 0.0216, "step": 24666 }, { "epoch": 4.52, "learning_rate": 8.373540442683206e-06, "loss": 0.0228, "step": 24667 }, { "epoch": 4.52, "learning_rate": 8.372807902971325e-06, "loss": 0.0251, "step": 24668 }, { "epoch": 4.52, "learning_rate": 8.372075372228558e-06, "loss": 0.025, "step": 24669 }, { "epoch": 4.52, "learning_rate": 8.371342850458954e-06, "loss": 0.0224, "step": 24670 }, { "epoch": 4.52, "learning_rate": 8.370610337666549e-06, "loss": 0.0229, "step": 24671 }, { "epoch": 4.52, "learning_rate": 8.369877833855376e-06, "loss": 0.0239, "step": 24672 }, { "epoch": 4.52, "learning_rate": 8.369145339029475e-06, "loss": 0.028, "step": 24673 }, { "epoch": 4.52, "learning_rate": 8.368412853192881e-06, "loss": 0.0215, "step": 24674 }, { "epoch": 4.53, "learning_rate": 8.367680376349635e-06, "loss": 0.0262, "step": 24675 }, { "epoch": 4.53, "learning_rate": 8.366947908503773e-06, "loss": 0.0248, "step": 24676 }, { "epoch": 4.53, "learning_rate": 8.366215449659335e-06, "loss": 0.0238, "step": 24677 }, { "epoch": 4.53, "learning_rate": 8.36548299982035e-06, "loss": 0.0238, "step": 24678 }, { "epoch": 4.53, "learning_rate": 8.364750558990862e-06, "loss": 0.0306, "step": 24679 }, { "epoch": 4.53, "learning_rate": 8.364018127174912e-06, "loss": 0.0248, "step": 24680 }, { "epoch": 4.53, "learning_rate": 8.36328570437653e-06, "loss": 0.0273, "step": 24681 }, { "epoch": 4.53, "learning_rate": 8.362553290599755e-06, "loss": 0.027, "step": 24682 }, { "epoch": 4.53, "learning_rate": 8.361820885848623e-06, "loss": 0.0256, "step": 24683 }, { "epoch": 4.53, "learning_rate": 8.361088490127176e-06, "loss": 0.0212, "step": 24684 }, { "epoch": 4.53, "learning_rate": 8.360356103439445e-06, "loss": 0.0191, "step": 24685 }, { "epoch": 4.53, "learning_rate": 8.359623725789474e-06, "loss": 0.0263, "step": 24686 }, { "epoch": 4.53, "learning_rate": 8.358891357181288e-06, "loss": 0.0269, "step": 24687 }, { "epoch": 4.53, "learning_rate": 8.358158997618936e-06, "loss": 0.021, "step": 24688 }, { "epoch": 4.53, "learning_rate": 8.357426647106451e-06, "loss": 0.0266, "step": 24689 }, { "epoch": 4.53, "learning_rate": 8.356694305647868e-06, "loss": 0.025, "step": 24690 }, { "epoch": 4.53, "learning_rate": 8.355961973247224e-06, "loss": 0.021, "step": 24691 }, { "epoch": 4.53, "learning_rate": 8.355229649908553e-06, "loss": 0.021, "step": 24692 }, { "epoch": 4.53, "learning_rate": 8.354497335635899e-06, "loss": 0.022, "step": 24693 }, { "epoch": 4.53, "learning_rate": 8.353765030433293e-06, "loss": 0.0205, "step": 24694 }, { "epoch": 4.53, "learning_rate": 8.353032734304775e-06, "loss": 0.0247, "step": 24695 }, { "epoch": 4.53, "learning_rate": 8.352300447254372e-06, "loss": 0.0217, "step": 24696 }, { "epoch": 4.53, "learning_rate": 8.351568169286134e-06, "loss": 0.0253, "step": 24697 }, { "epoch": 4.53, "learning_rate": 8.35083590040409e-06, "loss": 0.0231, "step": 24698 }, { "epoch": 4.53, "learning_rate": 8.350103640612276e-06, "loss": 0.0258, "step": 24699 }, { "epoch": 4.53, "learning_rate": 8.349371389914728e-06, "loss": 0.0247, "step": 24700 }, { "epoch": 4.53, "learning_rate": 8.348639148315486e-06, "loss": 0.0194, "step": 24701 }, { "epoch": 4.53, "learning_rate": 8.347906915818584e-06, "loss": 0.0224, "step": 24702 }, { "epoch": 4.53, "learning_rate": 8.347174692428056e-06, "loss": 0.0286, "step": 24703 }, { "epoch": 4.53, "learning_rate": 8.346442478147942e-06, "loss": 0.0266, "step": 24704 }, { "epoch": 4.53, "learning_rate": 8.34571027298227e-06, "loss": 0.0217, "step": 24705 }, { "epoch": 4.53, "learning_rate": 8.344978076935088e-06, "loss": 0.0248, "step": 24706 }, { "epoch": 4.53, "learning_rate": 8.344245890010425e-06, "loss": 0.0262, "step": 24707 }, { "epoch": 4.53, "learning_rate": 8.343513712212317e-06, "loss": 0.0271, "step": 24708 }, { "epoch": 4.53, "learning_rate": 8.342781543544799e-06, "loss": 0.0203, "step": 24709 }, { "epoch": 4.53, "learning_rate": 8.342049384011907e-06, "loss": 0.0235, "step": 24710 }, { "epoch": 4.53, "learning_rate": 8.34131723361768e-06, "loss": 0.0255, "step": 24711 }, { "epoch": 4.53, "learning_rate": 8.34058509236615e-06, "loss": 0.0241, "step": 24712 }, { "epoch": 4.53, "learning_rate": 8.339852960261357e-06, "loss": 0.0248, "step": 24713 }, { "epoch": 4.53, "learning_rate": 8.339120837307324e-06, "loss": 0.0246, "step": 24714 }, { "epoch": 4.53, "learning_rate": 8.338388723508107e-06, "loss": 0.023, "step": 24715 }, { "epoch": 4.53, "learning_rate": 8.337656618867724e-06, "loss": 0.0188, "step": 24716 }, { "epoch": 4.53, "learning_rate": 8.33692452339022e-06, "loss": 0.0253, "step": 24717 }, { "epoch": 4.53, "learning_rate": 8.336192437079622e-06, "loss": 0.0218, "step": 24718 }, { "epoch": 4.53, "learning_rate": 8.335460359939974e-06, "loss": 0.0264, "step": 24719 }, { "epoch": 4.53, "learning_rate": 8.334728291975306e-06, "loss": 0.0257, "step": 24720 }, { "epoch": 4.53, "learning_rate": 8.333996233189657e-06, "loss": 0.0223, "step": 24721 }, { "epoch": 4.53, "learning_rate": 8.333264183587058e-06, "loss": 0.0248, "step": 24722 }, { "epoch": 4.53, "learning_rate": 8.332532143171546e-06, "loss": 0.0258, "step": 24723 }, { "epoch": 4.53, "learning_rate": 8.331800111947158e-06, "loss": 0.0217, "step": 24724 }, { "epoch": 4.53, "learning_rate": 8.331068089917926e-06, "loss": 0.0248, "step": 24725 }, { "epoch": 4.53, "learning_rate": 8.330336077087883e-06, "loss": 0.0235, "step": 24726 }, { "epoch": 4.53, "learning_rate": 8.329604073461069e-06, "loss": 0.0232, "step": 24727 }, { "epoch": 4.53, "learning_rate": 8.328872079041514e-06, "loss": 0.0254, "step": 24728 }, { "epoch": 4.53, "learning_rate": 8.32814009383326e-06, "loss": 0.0251, "step": 24729 }, { "epoch": 4.54, "learning_rate": 8.327408117840333e-06, "loss": 0.0237, "step": 24730 }, { "epoch": 4.54, "learning_rate": 8.32667615106677e-06, "loss": 0.0259, "step": 24731 }, { "epoch": 4.54, "learning_rate": 8.325944193516611e-06, "loss": 0.0227, "step": 24732 }, { "epoch": 4.54, "learning_rate": 8.325212245193889e-06, "loss": 0.0253, "step": 24733 }, { "epoch": 4.54, "learning_rate": 8.324480306102633e-06, "loss": 0.0225, "step": 24734 }, { "epoch": 4.54, "learning_rate": 8.323748376246881e-06, "loss": 0.0177, "step": 24735 }, { "epoch": 4.54, "learning_rate": 8.323016455630666e-06, "loss": 0.0237, "step": 24736 }, { "epoch": 4.54, "learning_rate": 8.322284544258025e-06, "loss": 0.0228, "step": 24737 }, { "epoch": 4.54, "learning_rate": 8.321552642132991e-06, "loss": 0.0262, "step": 24738 }, { "epoch": 4.54, "learning_rate": 8.320820749259598e-06, "loss": 0.0257, "step": 24739 }, { "epoch": 4.54, "learning_rate": 8.320088865641879e-06, "loss": 0.0204, "step": 24740 }, { "epoch": 4.54, "learning_rate": 8.319356991283871e-06, "loss": 0.0248, "step": 24741 }, { "epoch": 4.54, "learning_rate": 8.318625126189608e-06, "loss": 0.0217, "step": 24742 }, { "epoch": 4.54, "learning_rate": 8.317893270363121e-06, "loss": 0.0245, "step": 24743 }, { "epoch": 4.54, "learning_rate": 8.317161423808446e-06, "loss": 0.0266, "step": 24744 }, { "epoch": 4.54, "learning_rate": 8.316429586529616e-06, "loss": 0.0236, "step": 24745 }, { "epoch": 4.54, "learning_rate": 8.315697758530665e-06, "loss": 0.0212, "step": 24746 }, { "epoch": 4.54, "learning_rate": 8.314965939815629e-06, "loss": 0.0225, "step": 24747 }, { "epoch": 4.54, "learning_rate": 8.31423413038854e-06, "loss": 0.024, "step": 24748 }, { "epoch": 4.54, "learning_rate": 8.31350233025343e-06, "loss": 0.0305, "step": 24749 }, { "epoch": 4.54, "learning_rate": 8.312770539414336e-06, "loss": 0.0219, "step": 24750 }, { "epoch": 4.54, "learning_rate": 8.312038757875294e-06, "loss": 0.0267, "step": 24751 }, { "epoch": 4.54, "learning_rate": 8.31130698564033e-06, "loss": 0.0223, "step": 24752 }, { "epoch": 4.54, "learning_rate": 8.310575222713478e-06, "loss": 0.026, "step": 24753 }, { "epoch": 4.54, "learning_rate": 8.309843469098781e-06, "loss": 0.0264, "step": 24754 }, { "epoch": 4.54, "learning_rate": 8.309111724800265e-06, "loss": 0.0198, "step": 24755 }, { "epoch": 4.54, "learning_rate": 8.308379989821966e-06, "loss": 0.0282, "step": 24756 }, { "epoch": 4.54, "learning_rate": 8.307648264167914e-06, "loss": 0.0239, "step": 24757 }, { "epoch": 4.54, "learning_rate": 8.306916547842142e-06, "loss": 0.0345, "step": 24758 }, { "epoch": 4.54, "learning_rate": 8.306184840848688e-06, "loss": 0.0278, "step": 24759 }, { "epoch": 4.54, "learning_rate": 8.305453143191586e-06, "loss": 0.0264, "step": 24760 }, { "epoch": 4.54, "learning_rate": 8.304721454874864e-06, "loss": 0.0257, "step": 24761 }, { "epoch": 4.54, "learning_rate": 8.303989775902555e-06, "loss": 0.0242, "step": 24762 }, { "epoch": 4.54, "learning_rate": 8.303258106278695e-06, "loss": 0.0208, "step": 24763 }, { "epoch": 4.54, "learning_rate": 8.302526446007316e-06, "loss": 0.021, "step": 24764 }, { "epoch": 4.54, "learning_rate": 8.30179479509245e-06, "loss": 0.0252, "step": 24765 }, { "epoch": 4.54, "learning_rate": 8.301063153538135e-06, "loss": 0.026, "step": 24766 }, { "epoch": 4.54, "learning_rate": 8.300331521348395e-06, "loss": 0.0219, "step": 24767 }, { "epoch": 4.54, "learning_rate": 8.299599898527268e-06, "loss": 0.0267, "step": 24768 }, { "epoch": 4.54, "learning_rate": 8.298868285078791e-06, "loss": 0.029, "step": 24769 }, { "epoch": 4.54, "learning_rate": 8.298136681006988e-06, "loss": 0.0256, "step": 24770 }, { "epoch": 4.54, "learning_rate": 8.297405086315895e-06, "loss": 0.0242, "step": 24771 }, { "epoch": 4.54, "learning_rate": 8.296673501009547e-06, "loss": 0.026, "step": 24772 }, { "epoch": 4.54, "learning_rate": 8.295941925091973e-06, "loss": 0.0263, "step": 24773 }, { "epoch": 4.54, "learning_rate": 8.295210358567207e-06, "loss": 0.0224, "step": 24774 }, { "epoch": 4.54, "learning_rate": 8.294478801439281e-06, "loss": 0.0196, "step": 24775 }, { "epoch": 4.54, "learning_rate": 8.29374725371223e-06, "loss": 0.0244, "step": 24776 }, { "epoch": 4.54, "learning_rate": 8.293015715390083e-06, "loss": 0.0252, "step": 24777 }, { "epoch": 4.54, "learning_rate": 8.292284186476875e-06, "loss": 0.0253, "step": 24778 }, { "epoch": 4.54, "learning_rate": 8.291552666976634e-06, "loss": 0.0212, "step": 24779 }, { "epoch": 4.54, "learning_rate": 8.290821156893395e-06, "loss": 0.0287, "step": 24780 }, { "epoch": 4.54, "learning_rate": 8.29008965623119e-06, "loss": 0.0215, "step": 24781 }, { "epoch": 4.54, "learning_rate": 8.289358164994052e-06, "loss": 0.0257, "step": 24782 }, { "epoch": 4.54, "learning_rate": 8.288626683186011e-06, "loss": 0.0283, "step": 24783 }, { "epoch": 4.55, "learning_rate": 8.287895210811099e-06, "loss": 0.0213, "step": 24784 }, { "epoch": 4.55, "learning_rate": 8.287163747873351e-06, "loss": 0.0253, "step": 24785 }, { "epoch": 4.55, "learning_rate": 8.286432294376794e-06, "loss": 0.029, "step": 24786 }, { "epoch": 4.55, "learning_rate": 8.285700850325467e-06, "loss": 0.0223, "step": 24787 }, { "epoch": 4.55, "learning_rate": 8.284969415723392e-06, "loss": 0.0231, "step": 24788 }, { "epoch": 4.55, "learning_rate": 8.284237990574607e-06, "loss": 0.0225, "step": 24789 }, { "epoch": 4.55, "learning_rate": 8.283506574883143e-06, "loss": 0.0264, "step": 24790 }, { "epoch": 4.55, "learning_rate": 8.28277516865303e-06, "loss": 0.0252, "step": 24791 }, { "epoch": 4.55, "learning_rate": 8.282043771888301e-06, "loss": 0.0255, "step": 24792 }, { "epoch": 4.55, "learning_rate": 8.281312384592986e-06, "loss": 0.0211, "step": 24793 }, { "epoch": 4.55, "learning_rate": 8.28058100677112e-06, "loss": 0.0254, "step": 24794 }, { "epoch": 4.55, "learning_rate": 8.27984963842673e-06, "loss": 0.0262, "step": 24795 }, { "epoch": 4.55, "learning_rate": 8.279118279563852e-06, "loss": 0.0264, "step": 24796 }, { "epoch": 4.55, "learning_rate": 8.278386930186509e-06, "loss": 0.0216, "step": 24797 }, { "epoch": 4.55, "learning_rate": 8.277655590298742e-06, "loss": 0.0202, "step": 24798 }, { "epoch": 4.55, "learning_rate": 8.276924259904576e-06, "loss": 0.0237, "step": 24799 }, { "epoch": 4.55, "learning_rate": 8.276192939008046e-06, "loss": 0.024, "step": 24800 }, { "epoch": 4.55, "learning_rate": 8.275461627613178e-06, "loss": 0.0247, "step": 24801 }, { "epoch": 4.55, "learning_rate": 8.274730325724005e-06, "loss": 0.0231, "step": 24802 }, { "epoch": 4.55, "learning_rate": 8.273999033344561e-06, "loss": 0.0233, "step": 24803 }, { "epoch": 4.55, "learning_rate": 8.273267750478874e-06, "loss": 0.0297, "step": 24804 }, { "epoch": 4.55, "learning_rate": 8.272536477130978e-06, "loss": 0.026, "step": 24805 }, { "epoch": 4.55, "learning_rate": 8.271805213304897e-06, "loss": 0.0281, "step": 24806 }, { "epoch": 4.55, "learning_rate": 8.271073959004671e-06, "loss": 0.0296, "step": 24807 }, { "epoch": 4.55, "learning_rate": 8.270342714234323e-06, "loss": 0.0294, "step": 24808 }, { "epoch": 4.55, "learning_rate": 8.269611478997887e-06, "loss": 0.0235, "step": 24809 }, { "epoch": 4.55, "learning_rate": 8.268880253299394e-06, "loss": 0.02, "step": 24810 }, { "epoch": 4.55, "learning_rate": 8.268149037142871e-06, "loss": 0.0266, "step": 24811 }, { "epoch": 4.55, "learning_rate": 8.267417830532352e-06, "loss": 0.0239, "step": 24812 }, { "epoch": 4.55, "learning_rate": 8.266686633471867e-06, "loss": 0.0216, "step": 24813 }, { "epoch": 4.55, "learning_rate": 8.265955445965451e-06, "loss": 0.0234, "step": 24814 }, { "epoch": 4.55, "learning_rate": 8.265224268017119e-06, "loss": 0.0235, "step": 24815 }, { "epoch": 4.55, "learning_rate": 8.26449309963092e-06, "loss": 0.0276, "step": 24816 }, { "epoch": 4.55, "learning_rate": 8.263761940810872e-06, "loss": 0.0209, "step": 24817 }, { "epoch": 4.55, "learning_rate": 8.26303079156101e-06, "loss": 0.0239, "step": 24818 }, { "epoch": 4.55, "learning_rate": 8.262299651885363e-06, "loss": 0.0288, "step": 24819 }, { "epoch": 4.55, "learning_rate": 8.261568521787957e-06, "loss": 0.0235, "step": 24820 }, { "epoch": 4.55, "learning_rate": 8.26083740127283e-06, "loss": 0.0242, "step": 24821 }, { "epoch": 4.55, "learning_rate": 8.26010629034401e-06, "loss": 0.0235, "step": 24822 }, { "epoch": 4.55, "learning_rate": 8.259375189005523e-06, "loss": 0.0277, "step": 24823 }, { "epoch": 4.55, "learning_rate": 8.258644097261396e-06, "loss": 0.0277, "step": 24824 }, { "epoch": 4.55, "learning_rate": 8.257913015115669e-06, "loss": 0.0223, "step": 24825 }, { "epoch": 4.55, "learning_rate": 8.257181942572366e-06, "loss": 0.0254, "step": 24826 }, { "epoch": 4.55, "learning_rate": 8.256450879635516e-06, "loss": 0.0251, "step": 24827 }, { "epoch": 4.55, "learning_rate": 8.255719826309148e-06, "loss": 0.0252, "step": 24828 }, { "epoch": 4.55, "learning_rate": 8.254988782597295e-06, "loss": 0.0281, "step": 24829 }, { "epoch": 4.55, "learning_rate": 8.254257748503983e-06, "loss": 0.0205, "step": 24830 }, { "epoch": 4.55, "learning_rate": 8.253526724033248e-06, "loss": 0.0224, "step": 24831 }, { "epoch": 4.55, "learning_rate": 8.25279570918911e-06, "loss": 0.0184, "step": 24832 }, { "epoch": 4.55, "learning_rate": 8.252064703975601e-06, "loss": 0.0261, "step": 24833 }, { "epoch": 4.55, "learning_rate": 8.251333708396758e-06, "loss": 0.0236, "step": 24834 }, { "epoch": 4.55, "learning_rate": 8.250602722456602e-06, "loss": 0.0242, "step": 24835 }, { "epoch": 4.55, "learning_rate": 8.249871746159164e-06, "loss": 0.0225, "step": 24836 }, { "epoch": 4.55, "learning_rate": 8.249140779508473e-06, "loss": 0.0263, "step": 24837 }, { "epoch": 4.55, "learning_rate": 8.24840982250856e-06, "loss": 0.0243, "step": 24838 }, { "epoch": 4.56, "learning_rate": 8.247678875163456e-06, "loss": 0.0224, "step": 24839 }, { "epoch": 4.56, "learning_rate": 8.246947937477187e-06, "loss": 0.019, "step": 24840 }, { "epoch": 4.56, "learning_rate": 8.24621700945378e-06, "loss": 0.0198, "step": 24841 }, { "epoch": 4.56, "learning_rate": 8.245486091097261e-06, "loss": 0.0191, "step": 24842 }, { "epoch": 4.56, "learning_rate": 8.24475518241167e-06, "loss": 0.025, "step": 24843 }, { "epoch": 4.56, "learning_rate": 8.24402428340103e-06, "loss": 0.0248, "step": 24844 }, { "epoch": 4.56, "learning_rate": 8.243293394069367e-06, "loss": 0.0234, "step": 24845 }, { "epoch": 4.56, "learning_rate": 8.242562514420711e-06, "loss": 0.0262, "step": 24846 }, { "epoch": 4.56, "learning_rate": 8.241831644459093e-06, "loss": 0.0238, "step": 24847 }, { "epoch": 4.56, "learning_rate": 8.24110078418854e-06, "loss": 0.0251, "step": 24848 }, { "epoch": 4.56, "learning_rate": 8.240369933613084e-06, "loss": 0.0259, "step": 24849 }, { "epoch": 4.56, "learning_rate": 8.239639092736743e-06, "loss": 0.0189, "step": 24850 }, { "epoch": 4.56, "learning_rate": 8.238908261563556e-06, "loss": 0.0205, "step": 24851 }, { "epoch": 4.56, "learning_rate": 8.238177440097552e-06, "loss": 0.0224, "step": 24852 }, { "epoch": 4.56, "learning_rate": 8.237446628342751e-06, "loss": 0.0218, "step": 24853 }, { "epoch": 4.56, "learning_rate": 8.236715826303188e-06, "loss": 0.0236, "step": 24854 }, { "epoch": 4.56, "learning_rate": 8.235985033982883e-06, "loss": 0.0227, "step": 24855 }, { "epoch": 4.56, "learning_rate": 8.235254251385875e-06, "loss": 0.0227, "step": 24856 }, { "epoch": 4.56, "learning_rate": 8.234523478516185e-06, "loss": 0.0215, "step": 24857 }, { "epoch": 4.56, "learning_rate": 8.233792715377846e-06, "loss": 0.0201, "step": 24858 }, { "epoch": 4.56, "learning_rate": 8.233061961974877e-06, "loss": 0.0263, "step": 24859 }, { "epoch": 4.56, "learning_rate": 8.232331218311316e-06, "loss": 0.0237, "step": 24860 }, { "epoch": 4.56, "learning_rate": 8.23160048439119e-06, "loss": 0.0278, "step": 24861 }, { "epoch": 4.56, "learning_rate": 8.230869760218519e-06, "loss": 0.0256, "step": 24862 }, { "epoch": 4.56, "learning_rate": 8.230139045797334e-06, "loss": 0.022, "step": 24863 }, { "epoch": 4.56, "learning_rate": 8.229408341131665e-06, "loss": 0.0225, "step": 24864 }, { "epoch": 4.56, "learning_rate": 8.22867764622554e-06, "loss": 0.0206, "step": 24865 }, { "epoch": 4.56, "learning_rate": 8.227946961082986e-06, "loss": 0.0231, "step": 24866 }, { "epoch": 4.56, "learning_rate": 8.22721628570803e-06, "loss": 0.0273, "step": 24867 }, { "epoch": 4.56, "learning_rate": 8.226485620104695e-06, "loss": 0.0248, "step": 24868 }, { "epoch": 4.56, "learning_rate": 8.225754964277018e-06, "loss": 0.0224, "step": 24869 }, { "epoch": 4.56, "learning_rate": 8.22502431822902e-06, "loss": 0.0215, "step": 24870 }, { "epoch": 4.56, "learning_rate": 8.22429368196473e-06, "loss": 0.0261, "step": 24871 }, { "epoch": 4.56, "learning_rate": 8.223563055488174e-06, "loss": 0.0197, "step": 24872 }, { "epoch": 4.56, "learning_rate": 8.222832438803376e-06, "loss": 0.0235, "step": 24873 }, { "epoch": 4.56, "learning_rate": 8.222101831914371e-06, "loss": 0.0157, "step": 24874 }, { "epoch": 4.56, "learning_rate": 8.221371234825183e-06, "loss": 0.0243, "step": 24875 }, { "epoch": 4.56, "learning_rate": 8.22064064753984e-06, "loss": 0.0215, "step": 24876 }, { "epoch": 4.56, "learning_rate": 8.219910070062362e-06, "loss": 0.0298, "step": 24877 }, { "epoch": 4.56, "learning_rate": 8.219179502396786e-06, "loss": 0.0219, "step": 24878 }, { "epoch": 4.56, "learning_rate": 8.218448944547135e-06, "loss": 0.0208, "step": 24879 }, { "epoch": 4.56, "learning_rate": 8.217718396517432e-06, "loss": 0.0276, "step": 24880 }, { "epoch": 4.56, "learning_rate": 8.216987858311706e-06, "loss": 0.0218, "step": 24881 }, { "epoch": 4.56, "learning_rate": 8.216257329933987e-06, "loss": 0.019, "step": 24882 }, { "epoch": 4.56, "learning_rate": 8.215526811388299e-06, "loss": 0.0243, "step": 24883 }, { "epoch": 4.56, "learning_rate": 8.214796302678669e-06, "loss": 0.02, "step": 24884 }, { "epoch": 4.56, "learning_rate": 8.214065803809128e-06, "loss": 0.02, "step": 24885 }, { "epoch": 4.56, "learning_rate": 8.213335314783688e-06, "loss": 0.0244, "step": 24886 }, { "epoch": 4.56, "learning_rate": 8.212604835606394e-06, "loss": 0.0216, "step": 24887 }, { "epoch": 4.56, "learning_rate": 8.211874366281262e-06, "loss": 0.0256, "step": 24888 }, { "epoch": 4.56, "learning_rate": 8.21114390681232e-06, "loss": 0.0229, "step": 24889 }, { "epoch": 4.56, "learning_rate": 8.210413457203591e-06, "loss": 0.0221, "step": 24890 }, { "epoch": 4.56, "learning_rate": 8.20968301745911e-06, "loss": 0.0249, "step": 24891 }, { "epoch": 4.56, "learning_rate": 8.208952587582897e-06, "loss": 0.0219, "step": 24892 }, { "epoch": 4.57, "learning_rate": 8.20822216757898e-06, "loss": 0.0197, "step": 24893 }, { "epoch": 4.57, "learning_rate": 8.207491757451384e-06, "loss": 0.0258, "step": 24894 }, { "epoch": 4.57, "learning_rate": 8.206761357204132e-06, "loss": 0.0299, "step": 24895 }, { "epoch": 4.57, "learning_rate": 8.20603096684126e-06, "loss": 0.0191, "step": 24896 }, { "epoch": 4.57, "learning_rate": 8.205300586366782e-06, "loss": 0.0218, "step": 24897 }, { "epoch": 4.57, "learning_rate": 8.20457021578473e-06, "loss": 0.0227, "step": 24898 }, { "epoch": 4.57, "learning_rate": 8.203839855099128e-06, "loss": 0.0283, "step": 24899 }, { "epoch": 4.57, "learning_rate": 8.203109504314005e-06, "loss": 0.0235, "step": 24900 }, { "epoch": 4.57, "learning_rate": 8.202379163433385e-06, "loss": 0.023, "step": 24901 }, { "epoch": 4.57, "learning_rate": 8.201648832461291e-06, "loss": 0.0211, "step": 24902 }, { "epoch": 4.57, "learning_rate": 8.200918511401752e-06, "loss": 0.026, "step": 24903 }, { "epoch": 4.57, "learning_rate": 8.200188200258793e-06, "loss": 0.0232, "step": 24904 }, { "epoch": 4.57, "learning_rate": 8.199457899036438e-06, "loss": 0.0242, "step": 24905 }, { "epoch": 4.57, "learning_rate": 8.198727607738715e-06, "loss": 0.0209, "step": 24906 }, { "epoch": 4.57, "learning_rate": 8.197997326369645e-06, "loss": 0.022, "step": 24907 }, { "epoch": 4.57, "learning_rate": 8.197267054933256e-06, "loss": 0.0191, "step": 24908 }, { "epoch": 4.57, "learning_rate": 8.196536793433573e-06, "loss": 0.0263, "step": 24909 }, { "epoch": 4.57, "learning_rate": 8.195806541874624e-06, "loss": 0.0249, "step": 24910 }, { "epoch": 4.57, "learning_rate": 8.19507630026043e-06, "loss": 0.0269, "step": 24911 }, { "epoch": 4.57, "learning_rate": 8.194346068595016e-06, "loss": 0.0209, "step": 24912 }, { "epoch": 4.57, "learning_rate": 8.193615846882411e-06, "loss": 0.0248, "step": 24913 }, { "epoch": 4.57, "learning_rate": 8.192885635126639e-06, "loss": 0.022, "step": 24914 }, { "epoch": 4.57, "learning_rate": 8.192155433331723e-06, "loss": 0.0224, "step": 24915 }, { "epoch": 4.57, "learning_rate": 8.191425241501688e-06, "loss": 0.0217, "step": 24916 }, { "epoch": 4.57, "learning_rate": 8.190695059640556e-06, "loss": 0.0252, "step": 24917 }, { "epoch": 4.57, "learning_rate": 8.18996488775236e-06, "loss": 0.0272, "step": 24918 }, { "epoch": 4.57, "learning_rate": 8.189234725841119e-06, "loss": 0.0283, "step": 24919 }, { "epoch": 4.57, "learning_rate": 8.188504573910859e-06, "loss": 0.0232, "step": 24920 }, { "epoch": 4.57, "learning_rate": 8.187774431965599e-06, "loss": 0.0224, "step": 24921 }, { "epoch": 4.57, "learning_rate": 8.187044300009373e-06, "loss": 0.0203, "step": 24922 }, { "epoch": 4.57, "learning_rate": 8.186314178046203e-06, "loss": 0.0279, "step": 24923 }, { "epoch": 4.57, "learning_rate": 8.185584066080113e-06, "loss": 0.0223, "step": 24924 }, { "epoch": 4.57, "learning_rate": 8.18485396411512e-06, "loss": 0.0291, "step": 24925 }, { "epoch": 4.57, "learning_rate": 8.184123872155258e-06, "loss": 0.0211, "step": 24926 }, { "epoch": 4.57, "learning_rate": 8.183393790204547e-06, "loss": 0.0273, "step": 24927 }, { "epoch": 4.57, "learning_rate": 8.182663718267013e-06, "loss": 0.0265, "step": 24928 }, { "epoch": 4.57, "learning_rate": 8.181933656346678e-06, "loss": 0.0196, "step": 24929 }, { "epoch": 4.57, "learning_rate": 8.181203604447567e-06, "loss": 0.0226, "step": 24930 }, { "epoch": 4.57, "learning_rate": 8.180473562573705e-06, "loss": 0.0205, "step": 24931 }, { "epoch": 4.57, "learning_rate": 8.179743530729118e-06, "loss": 0.0234, "step": 24932 }, { "epoch": 4.57, "learning_rate": 8.179013508917825e-06, "loss": 0.0285, "step": 24933 }, { "epoch": 4.57, "learning_rate": 8.178283497143851e-06, "loss": 0.0239, "step": 24934 }, { "epoch": 4.57, "learning_rate": 8.177553495411222e-06, "loss": 0.023, "step": 24935 }, { "epoch": 4.57, "learning_rate": 8.17682350372396e-06, "loss": 0.0268, "step": 24936 }, { "epoch": 4.57, "learning_rate": 8.176093522086091e-06, "loss": 0.0296, "step": 24937 }, { "epoch": 4.57, "learning_rate": 8.175363550501639e-06, "loss": 0.0226, "step": 24938 }, { "epoch": 4.57, "learning_rate": 8.17463358897462e-06, "loss": 0.0227, "step": 24939 }, { "epoch": 4.57, "learning_rate": 8.173903637509066e-06, "loss": 0.0229, "step": 24940 }, { "epoch": 4.57, "learning_rate": 8.173173696109002e-06, "loss": 0.0278, "step": 24941 }, { "epoch": 4.57, "learning_rate": 8.172443764778444e-06, "loss": 0.0259, "step": 24942 }, { "epoch": 4.57, "learning_rate": 8.171713843521419e-06, "loss": 0.0219, "step": 24943 }, { "epoch": 4.57, "learning_rate": 8.170983932341949e-06, "loss": 0.0235, "step": 24944 }, { "epoch": 4.57, "learning_rate": 8.17025403124406e-06, "loss": 0.0231, "step": 24945 }, { "epoch": 4.57, "learning_rate": 8.169524140231772e-06, "loss": 0.0251, "step": 24946 }, { "epoch": 4.57, "learning_rate": 8.168794259309115e-06, "loss": 0.0235, "step": 24947 }, { "epoch": 4.58, "learning_rate": 8.1680643884801e-06, "loss": 0.0247, "step": 24948 }, { "epoch": 4.58, "learning_rate": 8.167334527748759e-06, "loss": 0.0254, "step": 24949 }, { "epoch": 4.58, "learning_rate": 8.166604677119117e-06, "loss": 0.0276, "step": 24950 }, { "epoch": 4.58, "learning_rate": 8.165874836595189e-06, "loss": 0.0245, "step": 24951 }, { "epoch": 4.58, "learning_rate": 8.165145006181002e-06, "loss": 0.0247, "step": 24952 }, { "epoch": 4.58, "learning_rate": 8.16441518588058e-06, "loss": 0.0262, "step": 24953 }, { "epoch": 4.58, "learning_rate": 8.163685375697944e-06, "loss": 0.023, "step": 24954 }, { "epoch": 4.58, "learning_rate": 8.162955575637118e-06, "loss": 0.0232, "step": 24955 }, { "epoch": 4.58, "learning_rate": 8.16222578570212e-06, "loss": 0.0235, "step": 24956 }, { "epoch": 4.58, "learning_rate": 8.161496005896981e-06, "loss": 0.0247, "step": 24957 }, { "epoch": 4.58, "learning_rate": 8.160766236225717e-06, "loss": 0.0235, "step": 24958 }, { "epoch": 4.58, "learning_rate": 8.160036476692357e-06, "loss": 0.0242, "step": 24959 }, { "epoch": 4.58, "learning_rate": 8.159306727300915e-06, "loss": 0.0211, "step": 24960 }, { "epoch": 4.58, "learning_rate": 8.158576988055416e-06, "loss": 0.0258, "step": 24961 }, { "epoch": 4.58, "learning_rate": 8.157847258959885e-06, "loss": 0.0239, "step": 24962 }, { "epoch": 4.58, "learning_rate": 8.157117540018344e-06, "loss": 0.0249, "step": 24963 }, { "epoch": 4.58, "learning_rate": 8.156387831234815e-06, "loss": 0.0216, "step": 24964 }, { "epoch": 4.58, "learning_rate": 8.155658132613317e-06, "loss": 0.0251, "step": 24965 }, { "epoch": 4.58, "learning_rate": 8.154928444157876e-06, "loss": 0.0246, "step": 24966 }, { "epoch": 4.58, "learning_rate": 8.154198765872514e-06, "loss": 0.0185, "step": 24967 }, { "epoch": 4.58, "learning_rate": 8.153469097761252e-06, "loss": 0.019, "step": 24968 }, { "epoch": 4.58, "learning_rate": 8.15273943982811e-06, "loss": 0.0249, "step": 24969 }, { "epoch": 4.58, "learning_rate": 8.15200979207711e-06, "loss": 0.0206, "step": 24970 }, { "epoch": 4.58, "learning_rate": 8.151280154512277e-06, "loss": 0.0207, "step": 24971 }, { "epoch": 4.58, "learning_rate": 8.15055052713763e-06, "loss": 0.0225, "step": 24972 }, { "epoch": 4.58, "learning_rate": 8.149820909957193e-06, "loss": 0.0255, "step": 24973 }, { "epoch": 4.58, "learning_rate": 8.149091302974985e-06, "loss": 0.0265, "step": 24974 }, { "epoch": 4.58, "learning_rate": 8.14836170619503e-06, "loss": 0.0201, "step": 24975 }, { "epoch": 4.58, "learning_rate": 8.14763211962135e-06, "loss": 0.0212, "step": 24976 }, { "epoch": 4.58, "learning_rate": 8.146902543257966e-06, "loss": 0.0258, "step": 24977 }, { "epoch": 4.58, "learning_rate": 8.146172977108893e-06, "loss": 0.0251, "step": 24978 }, { "epoch": 4.58, "learning_rate": 8.145443421178162e-06, "loss": 0.0277, "step": 24979 }, { "epoch": 4.58, "learning_rate": 8.144713875469792e-06, "loss": 0.0258, "step": 24980 }, { "epoch": 4.58, "learning_rate": 8.1439843399878e-06, "loss": 0.0239, "step": 24981 }, { "epoch": 4.58, "learning_rate": 8.14325481473621e-06, "loss": 0.0224, "step": 24982 }, { "epoch": 4.58, "learning_rate": 8.142525299719041e-06, "loss": 0.0221, "step": 24983 }, { "epoch": 4.58, "learning_rate": 8.14179579494032e-06, "loss": 0.0271, "step": 24984 }, { "epoch": 4.58, "learning_rate": 8.141066300404064e-06, "loss": 0.0237, "step": 24985 }, { "epoch": 4.58, "learning_rate": 8.140336816114294e-06, "loss": 0.0275, "step": 24986 }, { "epoch": 4.58, "learning_rate": 8.139607342075025e-06, "loss": 0.023, "step": 24987 }, { "epoch": 4.58, "learning_rate": 8.13887787829029e-06, "loss": 0.0264, "step": 24988 }, { "epoch": 4.58, "learning_rate": 8.138148424764105e-06, "loss": 0.0226, "step": 24989 }, { "epoch": 4.58, "learning_rate": 8.137418981500485e-06, "loss": 0.0224, "step": 24990 }, { "epoch": 4.58, "learning_rate": 8.136689548503459e-06, "loss": 0.0215, "step": 24991 }, { "epoch": 4.58, "learning_rate": 8.135960125777039e-06, "loss": 0.0212, "step": 24992 }, { "epoch": 4.58, "learning_rate": 8.135230713325253e-06, "loss": 0.0243, "step": 24993 }, { "epoch": 4.58, "learning_rate": 8.134501311152123e-06, "loss": 0.0237, "step": 24994 }, { "epoch": 4.58, "learning_rate": 8.133771919261661e-06, "loss": 0.0258, "step": 24995 }, { "epoch": 4.58, "learning_rate": 8.13304253765789e-06, "loss": 0.0216, "step": 24996 }, { "epoch": 4.58, "learning_rate": 8.132313166344837e-06, "loss": 0.021, "step": 24997 }, { "epoch": 4.58, "learning_rate": 8.131583805326516e-06, "loss": 0.0254, "step": 24998 }, { "epoch": 4.58, "learning_rate": 8.130854454606948e-06, "loss": 0.0247, "step": 24999 }, { "epoch": 4.58, "learning_rate": 8.130125114190154e-06, "loss": 0.0175, "step": 25000 }, { "epoch": 4.58, "learning_rate": 8.129395784080153e-06, "loss": 0.0282, "step": 25001 }, { "epoch": 4.58, "learning_rate": 8.128666464280967e-06, "loss": 0.0229, "step": 25002 }, { "epoch": 4.59, "learning_rate": 8.12793715479662e-06, "loss": 0.026, "step": 25003 }, { "epoch": 4.59, "learning_rate": 8.127207855631123e-06, "loss": 0.0223, "step": 25004 }, { "epoch": 4.59, "learning_rate": 8.126478566788496e-06, "loss": 0.0265, "step": 25005 }, { "epoch": 4.59, "learning_rate": 8.125749288272771e-06, "loss": 0.0256, "step": 25006 }, { "epoch": 4.59, "learning_rate": 8.125020020087956e-06, "loss": 0.0286, "step": 25007 }, { "epoch": 4.59, "learning_rate": 8.124290762238075e-06, "loss": 0.0214, "step": 25008 }, { "epoch": 4.59, "learning_rate": 8.123561514727146e-06, "loss": 0.0248, "step": 25009 }, { "epoch": 4.59, "learning_rate": 8.122832277559192e-06, "loss": 0.0258, "step": 25010 }, { "epoch": 4.59, "learning_rate": 8.122103050738229e-06, "loss": 0.0227, "step": 25011 }, { "epoch": 4.59, "learning_rate": 8.121373834268282e-06, "loss": 0.028, "step": 25012 }, { "epoch": 4.59, "learning_rate": 8.120644628153362e-06, "loss": 0.0243, "step": 25013 }, { "epoch": 4.59, "learning_rate": 8.11991543239749e-06, "loss": 0.0235, "step": 25014 }, { "epoch": 4.59, "learning_rate": 8.119186247004694e-06, "loss": 0.0227, "step": 25015 }, { "epoch": 4.59, "learning_rate": 8.118457071978987e-06, "loss": 0.0299, "step": 25016 }, { "epoch": 4.59, "learning_rate": 8.117727907324386e-06, "loss": 0.0235, "step": 25017 }, { "epoch": 4.59, "learning_rate": 8.116998753044913e-06, "loss": 0.0232, "step": 25018 }, { "epoch": 4.59, "learning_rate": 8.116269609144588e-06, "loss": 0.0252, "step": 25019 }, { "epoch": 4.59, "learning_rate": 8.11554047562743e-06, "loss": 0.0252, "step": 25020 }, { "epoch": 4.59, "learning_rate": 8.114811352497458e-06, "loss": 0.0203, "step": 25021 }, { "epoch": 4.59, "learning_rate": 8.11408223975869e-06, "loss": 0.0245, "step": 25022 }, { "epoch": 4.59, "learning_rate": 8.11335313741514e-06, "loss": 0.0212, "step": 25023 }, { "epoch": 4.59, "learning_rate": 8.112624045470834e-06, "loss": 0.0233, "step": 25024 }, { "epoch": 4.59, "learning_rate": 8.11189496392979e-06, "loss": 0.0253, "step": 25025 }, { "epoch": 4.59, "learning_rate": 8.111165892796025e-06, "loss": 0.0203, "step": 25026 }, { "epoch": 4.59, "learning_rate": 8.110436832073556e-06, "loss": 0.0259, "step": 25027 }, { "epoch": 4.59, "learning_rate": 8.109707781766405e-06, "loss": 0.0224, "step": 25028 }, { "epoch": 4.59, "learning_rate": 8.108978741878589e-06, "loss": 0.0309, "step": 25029 }, { "epoch": 4.59, "learning_rate": 8.108249712414128e-06, "loss": 0.0203, "step": 25030 }, { "epoch": 4.59, "learning_rate": 8.107520693377033e-06, "loss": 0.0229, "step": 25031 }, { "epoch": 4.59, "learning_rate": 8.106791684771335e-06, "loss": 0.021, "step": 25032 }, { "epoch": 4.59, "learning_rate": 8.106062686601042e-06, "loss": 0.0208, "step": 25033 }, { "epoch": 4.59, "learning_rate": 8.105333698870178e-06, "loss": 0.0234, "step": 25034 }, { "epoch": 4.59, "learning_rate": 8.104604721582758e-06, "loss": 0.0233, "step": 25035 }, { "epoch": 4.59, "learning_rate": 8.103875754742799e-06, "loss": 0.0291, "step": 25036 }, { "epoch": 4.59, "learning_rate": 8.103146798354322e-06, "loss": 0.0166, "step": 25037 }, { "epoch": 4.59, "learning_rate": 8.102417852421348e-06, "loss": 0.023, "step": 25038 }, { "epoch": 4.59, "learning_rate": 8.10168891694789e-06, "loss": 0.0179, "step": 25039 }, { "epoch": 4.59, "learning_rate": 8.100959991937962e-06, "loss": 0.0185, "step": 25040 }, { "epoch": 4.59, "learning_rate": 8.100231077395593e-06, "loss": 0.0231, "step": 25041 }, { "epoch": 4.59, "learning_rate": 8.099502173324794e-06, "loss": 0.0242, "step": 25042 }, { "epoch": 4.59, "learning_rate": 8.098773279729582e-06, "loss": 0.0238, "step": 25043 }, { "epoch": 4.59, "learning_rate": 8.098044396613979e-06, "loss": 0.02, "step": 25044 }, { "epoch": 4.59, "learning_rate": 8.097315523981993e-06, "loss": 0.0224, "step": 25045 }, { "epoch": 4.59, "learning_rate": 8.096586661837653e-06, "loss": 0.0257, "step": 25046 }, { "epoch": 4.59, "learning_rate": 8.095857810184973e-06, "loss": 0.0285, "step": 25047 }, { "epoch": 4.59, "learning_rate": 8.09512896902797e-06, "loss": 0.0219, "step": 25048 }, { "epoch": 4.59, "learning_rate": 8.094400138370655e-06, "loss": 0.0256, "step": 25049 }, { "epoch": 4.59, "learning_rate": 8.093671318217059e-06, "loss": 0.031, "step": 25050 }, { "epoch": 4.59, "learning_rate": 8.092942508571187e-06, "loss": 0.021, "step": 25051 }, { "epoch": 4.59, "learning_rate": 8.09221370943706e-06, "loss": 0.0236, "step": 25052 }, { "epoch": 4.59, "learning_rate": 8.091484920818695e-06, "loss": 0.0261, "step": 25053 }, { "epoch": 4.59, "learning_rate": 8.09075614272011e-06, "loss": 0.0278, "step": 25054 }, { "epoch": 4.59, "learning_rate": 8.090027375145325e-06, "loss": 0.022, "step": 25055 }, { "epoch": 4.59, "learning_rate": 8.089298618098353e-06, "loss": 0.0272, "step": 25056 }, { "epoch": 4.6, "learning_rate": 8.088569871583212e-06, "loss": 0.0219, "step": 25057 }, { "epoch": 4.6, "learning_rate": 8.087841135603913e-06, "loss": 0.0238, "step": 25058 }, { "epoch": 4.6, "learning_rate": 8.087112410164485e-06, "loss": 0.0255, "step": 25059 }, { "epoch": 4.6, "learning_rate": 8.086383695268937e-06, "loss": 0.0237, "step": 25060 }, { "epoch": 4.6, "learning_rate": 8.085654990921287e-06, "loss": 0.0197, "step": 25061 }, { "epoch": 4.6, "learning_rate": 8.084926297125548e-06, "loss": 0.0215, "step": 25062 }, { "epoch": 4.6, "learning_rate": 8.084197613885744e-06, "loss": 0.0223, "step": 25063 }, { "epoch": 4.6, "learning_rate": 8.083468941205886e-06, "loss": 0.0233, "step": 25064 }, { "epoch": 4.6, "learning_rate": 8.082740279089993e-06, "loss": 0.0207, "step": 25065 }, { "epoch": 4.6, "learning_rate": 8.082011627542084e-06, "loss": 0.0236, "step": 25066 }, { "epoch": 4.6, "learning_rate": 8.081282986566163e-06, "loss": 0.0202, "step": 25067 }, { "epoch": 4.6, "learning_rate": 8.080554356166261e-06, "loss": 0.0235, "step": 25068 }, { "epoch": 4.6, "learning_rate": 8.079825736346389e-06, "loss": 0.0237, "step": 25069 }, { "epoch": 4.6, "learning_rate": 8.07909712711056e-06, "loss": 0.0254, "step": 25070 }, { "epoch": 4.6, "learning_rate": 8.078368528462791e-06, "loss": 0.0223, "step": 25071 }, { "epoch": 4.6, "learning_rate": 8.077639940407104e-06, "loss": 0.0234, "step": 25072 }, { "epoch": 4.6, "learning_rate": 8.076911362947508e-06, "loss": 0.0246, "step": 25073 }, { "epoch": 4.6, "learning_rate": 8.076182796088023e-06, "loss": 0.0224, "step": 25074 }, { "epoch": 4.6, "learning_rate": 8.075454239832664e-06, "loss": 0.0227, "step": 25075 }, { "epoch": 4.6, "learning_rate": 8.074725694185443e-06, "loss": 0.0279, "step": 25076 }, { "epoch": 4.6, "learning_rate": 8.073997159150384e-06, "loss": 0.0229, "step": 25077 }, { "epoch": 4.6, "learning_rate": 8.073268634731495e-06, "loss": 0.0212, "step": 25078 }, { "epoch": 4.6, "learning_rate": 8.072540120932795e-06, "loss": 0.0254, "step": 25079 }, { "epoch": 4.6, "learning_rate": 8.071811617758296e-06, "loss": 0.0246, "step": 25080 }, { "epoch": 4.6, "learning_rate": 8.07108312521202e-06, "loss": 0.022, "step": 25081 }, { "epoch": 4.6, "learning_rate": 8.070354643297978e-06, "loss": 0.0233, "step": 25082 }, { "epoch": 4.6, "learning_rate": 8.069626172020187e-06, "loss": 0.0209, "step": 25083 }, { "epoch": 4.6, "learning_rate": 8.068897711382659e-06, "loss": 0.024, "step": 25084 }, { "epoch": 4.6, "learning_rate": 8.068169261389416e-06, "loss": 0.0216, "step": 25085 }, { "epoch": 4.6, "learning_rate": 8.06744082204447e-06, "loss": 0.0227, "step": 25086 }, { "epoch": 4.6, "learning_rate": 8.066712393351833e-06, "loss": 0.0259, "step": 25087 }, { "epoch": 4.6, "learning_rate": 8.065983975315524e-06, "loss": 0.0249, "step": 25088 }, { "epoch": 4.6, "learning_rate": 8.065255567939551e-06, "loss": 0.0248, "step": 25089 }, { "epoch": 4.6, "learning_rate": 8.06452717122794e-06, "loss": 0.0229, "step": 25090 }, { "epoch": 4.6, "learning_rate": 8.0637987851847e-06, "loss": 0.0179, "step": 25091 }, { "epoch": 4.6, "learning_rate": 8.06307040981385e-06, "loss": 0.0269, "step": 25092 }, { "epoch": 4.6, "learning_rate": 8.06234204511939e-06, "loss": 0.021, "step": 25093 }, { "epoch": 4.6, "learning_rate": 8.061613691105355e-06, "loss": 0.0191, "step": 25094 }, { "epoch": 4.6, "learning_rate": 8.060885347775751e-06, "loss": 0.0287, "step": 25095 }, { "epoch": 4.6, "learning_rate": 8.06015701513459e-06, "loss": 0.0189, "step": 25096 }, { "epoch": 4.6, "learning_rate": 8.059428693185892e-06, "loss": 0.0271, "step": 25097 }, { "epoch": 4.6, "learning_rate": 8.058700381933662e-06, "loss": 0.0211, "step": 25098 }, { "epoch": 4.6, "learning_rate": 8.057972081381926e-06, "loss": 0.0303, "step": 25099 }, { "epoch": 4.6, "learning_rate": 8.057243791534692e-06, "loss": 0.0228, "step": 25100 }, { "epoch": 4.6, "learning_rate": 8.056515512395979e-06, "loss": 0.0239, "step": 25101 }, { "epoch": 4.6, "learning_rate": 8.055787243969791e-06, "loss": 0.0227, "step": 25102 }, { "epoch": 4.6, "learning_rate": 8.055058986260153e-06, "loss": 0.0208, "step": 25103 }, { "epoch": 4.6, "learning_rate": 8.054330739271078e-06, "loss": 0.0239, "step": 25104 }, { "epoch": 4.6, "learning_rate": 8.053602503006576e-06, "loss": 0.0237, "step": 25105 }, { "epoch": 4.6, "learning_rate": 8.05287427747066e-06, "loss": 0.0226, "step": 25106 }, { "epoch": 4.6, "learning_rate": 8.05214606266735e-06, "loss": 0.0168, "step": 25107 }, { "epoch": 4.6, "learning_rate": 8.051417858600655e-06, "loss": 0.0248, "step": 25108 }, { "epoch": 4.6, "learning_rate": 8.05068966527459e-06, "loss": 0.0228, "step": 25109 }, { "epoch": 4.6, "learning_rate": 8.049961482693173e-06, "loss": 0.0253, "step": 25110 }, { "epoch": 4.6, "learning_rate": 8.049233310860407e-06, "loss": 0.0181, "step": 25111 }, { "epoch": 4.61, "learning_rate": 8.048505149780316e-06, "loss": 0.0261, "step": 25112 }, { "epoch": 4.61, "learning_rate": 8.047776999456913e-06, "loss": 0.0241, "step": 25113 }, { "epoch": 4.61, "learning_rate": 8.047048859894206e-06, "loss": 0.0284, "step": 25114 }, { "epoch": 4.61, "learning_rate": 8.046320731096211e-06, "loss": 0.0203, "step": 25115 }, { "epoch": 4.61, "learning_rate": 8.045592613066944e-06, "loss": 0.0223, "step": 25116 }, { "epoch": 4.61, "learning_rate": 8.044864505810415e-06, "loss": 0.0233, "step": 25117 }, { "epoch": 4.61, "learning_rate": 8.044136409330638e-06, "loss": 0.0258, "step": 25118 }, { "epoch": 4.61, "learning_rate": 8.043408323631631e-06, "loss": 0.0205, "step": 25119 }, { "epoch": 4.61, "learning_rate": 8.042680248717397e-06, "loss": 0.0233, "step": 25120 }, { "epoch": 4.61, "learning_rate": 8.041952184591958e-06, "loss": 0.0359, "step": 25121 }, { "epoch": 4.61, "learning_rate": 8.041224131259327e-06, "loss": 0.0237, "step": 25122 }, { "epoch": 4.61, "learning_rate": 8.040496088723512e-06, "loss": 0.0242, "step": 25123 }, { "epoch": 4.61, "learning_rate": 8.039768056988527e-06, "loss": 0.0251, "step": 25124 }, { "epoch": 4.61, "learning_rate": 8.039040036058389e-06, "loss": 0.0299, "step": 25125 }, { "epoch": 4.61, "learning_rate": 8.038312025937108e-06, "loss": 0.0216, "step": 25126 }, { "epoch": 4.61, "learning_rate": 8.037584026628695e-06, "loss": 0.0212, "step": 25127 }, { "epoch": 4.61, "learning_rate": 8.03685603813717e-06, "loss": 0.0255, "step": 25128 }, { "epoch": 4.61, "learning_rate": 8.036128060466532e-06, "loss": 0.0277, "step": 25129 }, { "epoch": 4.61, "learning_rate": 8.035400093620806e-06, "loss": 0.0274, "step": 25130 }, { "epoch": 4.61, "learning_rate": 8.034672137604005e-06, "loss": 0.0328, "step": 25131 }, { "epoch": 4.61, "learning_rate": 8.033944192420134e-06, "loss": 0.0228, "step": 25132 }, { "epoch": 4.61, "learning_rate": 8.033216258073207e-06, "loss": 0.023, "step": 25133 }, { "epoch": 4.61, "learning_rate": 8.032488334567238e-06, "loss": 0.0243, "step": 25134 }, { "epoch": 4.61, "learning_rate": 8.031760421906242e-06, "loss": 0.0248, "step": 25135 }, { "epoch": 4.61, "learning_rate": 8.031032520094229e-06, "loss": 0.029, "step": 25136 }, { "epoch": 4.61, "learning_rate": 8.030304629135207e-06, "loss": 0.0214, "step": 25137 }, { "epoch": 4.61, "learning_rate": 8.029576749033195e-06, "loss": 0.0245, "step": 25138 }, { "epoch": 4.61, "learning_rate": 8.028848879792204e-06, "loss": 0.026, "step": 25139 }, { "epoch": 4.61, "learning_rate": 8.028121021416241e-06, "loss": 0.021, "step": 25140 }, { "epoch": 4.61, "learning_rate": 8.027393173909323e-06, "loss": 0.0242, "step": 25141 }, { "epoch": 4.61, "learning_rate": 8.026665337275456e-06, "loss": 0.0225, "step": 25142 }, { "epoch": 4.61, "learning_rate": 8.02593751151866e-06, "loss": 0.0257, "step": 25143 }, { "epoch": 4.61, "learning_rate": 8.025209696642942e-06, "loss": 0.0265, "step": 25144 }, { "epoch": 4.61, "learning_rate": 8.024481892652314e-06, "loss": 0.0274, "step": 25145 }, { "epoch": 4.61, "learning_rate": 8.023754099550786e-06, "loss": 0.0246, "step": 25146 }, { "epoch": 4.61, "learning_rate": 8.023026317342375e-06, "loss": 0.0246, "step": 25147 }, { "epoch": 4.61, "learning_rate": 8.02229854603109e-06, "loss": 0.0249, "step": 25148 }, { "epoch": 4.61, "learning_rate": 8.02157078562094e-06, "loss": 0.0212, "step": 25149 }, { "epoch": 4.61, "learning_rate": 8.020843036115939e-06, "loss": 0.025, "step": 25150 }, { "epoch": 4.61, "learning_rate": 8.020115297520093e-06, "loss": 0.0232, "step": 25151 }, { "epoch": 4.61, "learning_rate": 8.019387569837422e-06, "loss": 0.0225, "step": 25152 }, { "epoch": 4.61, "learning_rate": 8.018659853071933e-06, "loss": 0.0229, "step": 25153 }, { "epoch": 4.61, "learning_rate": 8.017932147227637e-06, "loss": 0.0227, "step": 25154 }, { "epoch": 4.61, "learning_rate": 8.017204452308544e-06, "loss": 0.0203, "step": 25155 }, { "epoch": 4.61, "learning_rate": 8.016476768318668e-06, "loss": 0.0222, "step": 25156 }, { "epoch": 4.61, "learning_rate": 8.01574909526202e-06, "loss": 0.0262, "step": 25157 }, { "epoch": 4.61, "learning_rate": 8.015021433142608e-06, "loss": 0.0263, "step": 25158 }, { "epoch": 4.61, "learning_rate": 8.01429378196444e-06, "loss": 0.0208, "step": 25159 }, { "epoch": 4.61, "learning_rate": 8.013566141731537e-06, "loss": 0.0234, "step": 25160 }, { "epoch": 4.61, "learning_rate": 8.012838512447902e-06, "loss": 0.0244, "step": 25161 }, { "epoch": 4.61, "learning_rate": 8.012110894117549e-06, "loss": 0.0235, "step": 25162 }, { "epoch": 4.61, "learning_rate": 8.011383286744486e-06, "loss": 0.0242, "step": 25163 }, { "epoch": 4.61, "learning_rate": 8.010655690332724e-06, "loss": 0.0195, "step": 25164 }, { "epoch": 4.61, "learning_rate": 8.009928104886275e-06, "loss": 0.0354, "step": 25165 }, { "epoch": 4.62, "learning_rate": 8.009200530409153e-06, "loss": 0.0199, "step": 25166 }, { "epoch": 4.62, "learning_rate": 8.008472966905361e-06, "loss": 0.0248, "step": 25167 }, { "epoch": 4.62, "learning_rate": 8.007745414378909e-06, "loss": 0.0192, "step": 25168 }, { "epoch": 4.62, "learning_rate": 8.007017872833816e-06, "loss": 0.0214, "step": 25169 }, { "epoch": 4.62, "learning_rate": 8.006290342274085e-06, "loss": 0.0281, "step": 25170 }, { "epoch": 4.62, "learning_rate": 8.00556282270373e-06, "loss": 0.0267, "step": 25171 }, { "epoch": 4.62, "learning_rate": 8.004835314126758e-06, "loss": 0.0221, "step": 25172 }, { "epoch": 4.62, "learning_rate": 8.00410781654718e-06, "loss": 0.0214, "step": 25173 }, { "epoch": 4.62, "learning_rate": 8.003380329969007e-06, "loss": 0.0196, "step": 25174 }, { "epoch": 4.62, "learning_rate": 8.002652854396251e-06, "loss": 0.0234, "step": 25175 }, { "epoch": 4.62, "learning_rate": 8.001925389832918e-06, "loss": 0.0221, "step": 25176 }, { "epoch": 4.62, "learning_rate": 8.001197936283014e-06, "loss": 0.0239, "step": 25177 }, { "epoch": 4.62, "learning_rate": 8.000470493750559e-06, "loss": 0.0232, "step": 25178 }, { "epoch": 4.62, "learning_rate": 7.999743062239557e-06, "loss": 0.0235, "step": 25179 }, { "epoch": 4.62, "learning_rate": 7.999015641754018e-06, "loss": 0.0279, "step": 25180 }, { "epoch": 4.62, "learning_rate": 7.998288232297949e-06, "loss": 0.0222, "step": 25181 }, { "epoch": 4.62, "learning_rate": 7.997560833875365e-06, "loss": 0.021, "step": 25182 }, { "epoch": 4.62, "learning_rate": 7.996833446490272e-06, "loss": 0.02, "step": 25183 }, { "epoch": 4.62, "learning_rate": 7.996106070146682e-06, "loss": 0.0238, "step": 25184 }, { "epoch": 4.62, "learning_rate": 7.9953787048486e-06, "loss": 0.0253, "step": 25185 }, { "epoch": 4.62, "learning_rate": 7.994651350600036e-06, "loss": 0.0203, "step": 25186 }, { "epoch": 4.62, "learning_rate": 7.993924007405003e-06, "loss": 0.0192, "step": 25187 }, { "epoch": 4.62, "learning_rate": 7.993196675267507e-06, "loss": 0.0236, "step": 25188 }, { "epoch": 4.62, "learning_rate": 7.992469354191559e-06, "loss": 0.02, "step": 25189 }, { "epoch": 4.62, "learning_rate": 7.991742044181164e-06, "loss": 0.0253, "step": 25190 }, { "epoch": 4.62, "learning_rate": 7.991014745240336e-06, "loss": 0.0221, "step": 25191 }, { "epoch": 4.62, "learning_rate": 7.990287457373083e-06, "loss": 0.0204, "step": 25192 }, { "epoch": 4.62, "learning_rate": 7.989560180583414e-06, "loss": 0.0251, "step": 25193 }, { "epoch": 4.62, "learning_rate": 7.988832914875333e-06, "loss": 0.0239, "step": 25194 }, { "epoch": 4.62, "learning_rate": 7.988105660252852e-06, "loss": 0.0256, "step": 25195 }, { "epoch": 4.62, "learning_rate": 7.98737841671998e-06, "loss": 0.0217, "step": 25196 }, { "epoch": 4.62, "learning_rate": 7.986651184280727e-06, "loss": 0.0235, "step": 25197 }, { "epoch": 4.62, "learning_rate": 7.985923962939098e-06, "loss": 0.0234, "step": 25198 }, { "epoch": 4.62, "learning_rate": 7.985196752699103e-06, "loss": 0.0242, "step": 25199 }, { "epoch": 4.62, "learning_rate": 7.984469553564752e-06, "loss": 0.0251, "step": 25200 }, { "epoch": 4.62, "learning_rate": 7.983742365540053e-06, "loss": 0.0216, "step": 25201 }, { "epoch": 4.62, "learning_rate": 7.983015188629014e-06, "loss": 0.0252, "step": 25202 }, { "epoch": 4.62, "learning_rate": 7.982288022835641e-06, "loss": 0.0169, "step": 25203 }, { "epoch": 4.62, "learning_rate": 7.981560868163941e-06, "loss": 0.0231, "step": 25204 }, { "epoch": 4.62, "learning_rate": 7.980833724617928e-06, "loss": 0.0216, "step": 25205 }, { "epoch": 4.62, "learning_rate": 7.980106592201606e-06, "loss": 0.0196, "step": 25206 }, { "epoch": 4.62, "learning_rate": 7.979379470918985e-06, "loss": 0.0211, "step": 25207 }, { "epoch": 4.62, "learning_rate": 7.978652360774067e-06, "loss": 0.0249, "step": 25208 }, { "epoch": 4.62, "learning_rate": 7.977925261770869e-06, "loss": 0.0234, "step": 25209 }, { "epoch": 4.62, "learning_rate": 7.977198173913394e-06, "loss": 0.026, "step": 25210 }, { "epoch": 4.62, "learning_rate": 7.976471097205651e-06, "loss": 0.0197, "step": 25211 }, { "epoch": 4.62, "learning_rate": 7.975744031651642e-06, "loss": 0.0233, "step": 25212 }, { "epoch": 4.62, "learning_rate": 7.975016977255384e-06, "loss": 0.0217, "step": 25213 }, { "epoch": 4.62, "learning_rate": 7.974289934020879e-06, "loss": 0.0217, "step": 25214 }, { "epoch": 4.62, "learning_rate": 7.973562901952135e-06, "loss": 0.0251, "step": 25215 }, { "epoch": 4.62, "learning_rate": 7.97283588105316e-06, "loss": 0.0243, "step": 25216 }, { "epoch": 4.62, "learning_rate": 7.972108871327958e-06, "loss": 0.0225, "step": 25217 }, { "epoch": 4.62, "learning_rate": 7.971381872780542e-06, "loss": 0.0212, "step": 25218 }, { "epoch": 4.62, "learning_rate": 7.970654885414916e-06, "loss": 0.0217, "step": 25219 }, { "epoch": 4.62, "learning_rate": 7.969927909235091e-06, "loss": 0.0184, "step": 25220 }, { "epoch": 4.63, "learning_rate": 7.969200944245065e-06, "loss": 0.025, "step": 25221 }, { "epoch": 4.63, "learning_rate": 7.968473990448857e-06, "loss": 0.0251, "step": 25222 }, { "epoch": 4.63, "learning_rate": 7.967747047850465e-06, "loss": 0.0211, "step": 25223 }, { "epoch": 4.63, "learning_rate": 7.9670201164539e-06, "loss": 0.021, "step": 25224 }, { "epoch": 4.63, "learning_rate": 7.966293196263166e-06, "loss": 0.0229, "step": 25225 }, { "epoch": 4.63, "learning_rate": 7.965566287282272e-06, "loss": 0.0212, "step": 25226 }, { "epoch": 4.63, "learning_rate": 7.964839389515225e-06, "loss": 0.0211, "step": 25227 }, { "epoch": 4.63, "learning_rate": 7.964112502966031e-06, "loss": 0.0252, "step": 25228 }, { "epoch": 4.63, "learning_rate": 7.963385627638698e-06, "loss": 0.0227, "step": 25229 }, { "epoch": 4.63, "learning_rate": 7.962658763537226e-06, "loss": 0.0232, "step": 25230 }, { "epoch": 4.63, "learning_rate": 7.96193191066563e-06, "loss": 0.0209, "step": 25231 }, { "epoch": 4.63, "learning_rate": 7.961205069027915e-06, "loss": 0.0287, "step": 25232 }, { "epoch": 4.63, "learning_rate": 7.960478238628085e-06, "loss": 0.0216, "step": 25233 }, { "epoch": 4.63, "learning_rate": 7.959751419470142e-06, "loss": 0.0226, "step": 25234 }, { "epoch": 4.63, "learning_rate": 7.9590246115581e-06, "loss": 0.0249, "step": 25235 }, { "epoch": 4.63, "learning_rate": 7.958297814895963e-06, "loss": 0.0241, "step": 25236 }, { "epoch": 4.63, "learning_rate": 7.957571029487736e-06, "loss": 0.025, "step": 25237 }, { "epoch": 4.63, "learning_rate": 7.956844255337426e-06, "loss": 0.0229, "step": 25238 }, { "epoch": 4.63, "learning_rate": 7.956117492449034e-06, "loss": 0.0215, "step": 25239 }, { "epoch": 4.63, "learning_rate": 7.955390740826577e-06, "loss": 0.0265, "step": 25240 }, { "epoch": 4.63, "learning_rate": 7.95466400047405e-06, "loss": 0.0184, "step": 25241 }, { "epoch": 4.63, "learning_rate": 7.953937271395465e-06, "loss": 0.0258, "step": 25242 }, { "epoch": 4.63, "learning_rate": 7.953210553594821e-06, "loss": 0.0289, "step": 25243 }, { "epoch": 4.63, "learning_rate": 7.952483847076133e-06, "loss": 0.0169, "step": 25244 }, { "epoch": 4.63, "learning_rate": 7.9517571518434e-06, "loss": 0.031, "step": 25245 }, { "epoch": 4.63, "learning_rate": 7.951030467900633e-06, "loss": 0.0201, "step": 25246 }, { "epoch": 4.63, "learning_rate": 7.950303795251833e-06, "loss": 0.0205, "step": 25247 }, { "epoch": 4.63, "learning_rate": 7.949577133900999e-06, "loss": 0.0258, "step": 25248 }, { "epoch": 4.63, "learning_rate": 7.948850483852153e-06, "loss": 0.0285, "step": 25249 }, { "epoch": 4.63, "learning_rate": 7.948123845109289e-06, "loss": 0.0218, "step": 25250 }, { "epoch": 4.63, "learning_rate": 7.947397217676413e-06, "loss": 0.025, "step": 25251 }, { "epoch": 4.63, "learning_rate": 7.94667060155753e-06, "loss": 0.0177, "step": 25252 }, { "epoch": 4.63, "learning_rate": 7.945943996756648e-06, "loss": 0.026, "step": 25253 }, { "epoch": 4.63, "learning_rate": 7.945217403277771e-06, "loss": 0.0268, "step": 25254 }, { "epoch": 4.63, "learning_rate": 7.944490821124909e-06, "loss": 0.0206, "step": 25255 }, { "epoch": 4.63, "learning_rate": 7.943764250302056e-06, "loss": 0.0205, "step": 25256 }, { "epoch": 4.63, "learning_rate": 7.94303769081322e-06, "loss": 0.024, "step": 25257 }, { "epoch": 4.63, "learning_rate": 7.942311142662414e-06, "loss": 0.0289, "step": 25258 }, { "epoch": 4.63, "learning_rate": 7.941584605853636e-06, "loss": 0.0226, "step": 25259 }, { "epoch": 4.63, "learning_rate": 7.940858080390891e-06, "loss": 0.0239, "step": 25260 }, { "epoch": 4.63, "learning_rate": 7.940131566278183e-06, "loss": 0.0228, "step": 25261 }, { "epoch": 4.63, "learning_rate": 7.93940506351952e-06, "loss": 0.0286, "step": 25262 }, { "epoch": 4.63, "learning_rate": 7.938678572118904e-06, "loss": 0.0209, "step": 25263 }, { "epoch": 4.63, "learning_rate": 7.937952092080345e-06, "loss": 0.0247, "step": 25264 }, { "epoch": 4.63, "learning_rate": 7.937225623407832e-06, "loss": 0.0235, "step": 25265 }, { "epoch": 4.63, "learning_rate": 7.936499166105387e-06, "loss": 0.02, "step": 25266 }, { "epoch": 4.63, "learning_rate": 7.935772720177007e-06, "loss": 0.0213, "step": 25267 }, { "epoch": 4.63, "learning_rate": 7.935046285626695e-06, "loss": 0.0186, "step": 25268 }, { "epoch": 4.63, "learning_rate": 7.934319862458456e-06, "loss": 0.0182, "step": 25269 }, { "epoch": 4.63, "learning_rate": 7.933593450676291e-06, "loss": 0.0251, "step": 25270 }, { "epoch": 4.63, "learning_rate": 7.932867050284212e-06, "loss": 0.0197, "step": 25271 }, { "epoch": 4.63, "learning_rate": 7.932140661286216e-06, "loss": 0.0263, "step": 25272 }, { "epoch": 4.63, "learning_rate": 7.931414283686313e-06, "loss": 0.0246, "step": 25273 }, { "epoch": 4.63, "learning_rate": 7.930687917488495e-06, "loss": 0.0225, "step": 25274 }, { "epoch": 4.64, "learning_rate": 7.929961562696779e-06, "loss": 0.0206, "step": 25275 }, { "epoch": 4.64, "learning_rate": 7.929235219315164e-06, "loss": 0.0265, "step": 25276 }, { "epoch": 4.64, "learning_rate": 7.928508887347652e-06, "loss": 0.0249, "step": 25277 }, { "epoch": 4.64, "learning_rate": 7.927782566798247e-06, "loss": 0.0215, "step": 25278 }, { "epoch": 4.64, "learning_rate": 7.927056257670953e-06, "loss": 0.0203, "step": 25279 }, { "epoch": 4.64, "learning_rate": 7.926329959969775e-06, "loss": 0.0195, "step": 25280 }, { "epoch": 4.64, "learning_rate": 7.925603673698712e-06, "loss": 0.0204, "step": 25281 }, { "epoch": 4.64, "learning_rate": 7.924877398861775e-06, "loss": 0.0196, "step": 25282 }, { "epoch": 4.64, "learning_rate": 7.924151135462954e-06, "loss": 0.0236, "step": 25283 }, { "epoch": 4.64, "learning_rate": 7.923424883506267e-06, "loss": 0.0218, "step": 25284 }, { "epoch": 4.64, "learning_rate": 7.922698642995711e-06, "loss": 0.0253, "step": 25285 }, { "epoch": 4.64, "learning_rate": 7.921972413935287e-06, "loss": 0.0221, "step": 25286 }, { "epoch": 4.64, "learning_rate": 7.921246196328999e-06, "loss": 0.0249, "step": 25287 }, { "epoch": 4.64, "learning_rate": 7.920519990180851e-06, "loss": 0.0246, "step": 25288 }, { "epoch": 4.64, "learning_rate": 7.919793795494847e-06, "loss": 0.0197, "step": 25289 }, { "epoch": 4.64, "learning_rate": 7.919067612274986e-06, "loss": 0.0241, "step": 25290 }, { "epoch": 4.64, "learning_rate": 7.918341440525279e-06, "loss": 0.0203, "step": 25291 }, { "epoch": 4.64, "learning_rate": 7.917615280249714e-06, "loss": 0.0229, "step": 25292 }, { "epoch": 4.64, "learning_rate": 7.916889131452307e-06, "loss": 0.0234, "step": 25293 }, { "epoch": 4.64, "learning_rate": 7.916162994137056e-06, "loss": 0.0219, "step": 25294 }, { "epoch": 4.64, "learning_rate": 7.915436868307964e-06, "loss": 0.0263, "step": 25295 }, { "epoch": 4.64, "learning_rate": 7.91471075396903e-06, "loss": 0.0181, "step": 25296 }, { "epoch": 4.64, "learning_rate": 7.913984651124261e-06, "loss": 0.0251, "step": 25297 }, { "epoch": 4.64, "learning_rate": 7.913258559777659e-06, "loss": 0.0228, "step": 25298 }, { "epoch": 4.64, "learning_rate": 7.912532479933222e-06, "loss": 0.0229, "step": 25299 }, { "epoch": 4.64, "learning_rate": 7.911806411594958e-06, "loss": 0.0219, "step": 25300 }, { "epoch": 4.64, "learning_rate": 7.91108035476686e-06, "loss": 0.0222, "step": 25301 }, { "epoch": 4.64, "learning_rate": 7.910354309452942e-06, "loss": 0.0238, "step": 25302 }, { "epoch": 4.64, "learning_rate": 7.909628275657199e-06, "loss": 0.0221, "step": 25303 }, { "epoch": 4.64, "learning_rate": 7.908902253383632e-06, "loss": 0.0219, "step": 25304 }, { "epoch": 4.64, "learning_rate": 7.908176242636245e-06, "loss": 0.0246, "step": 25305 }, { "epoch": 4.64, "learning_rate": 7.90745024341904e-06, "loss": 0.0188, "step": 25306 }, { "epoch": 4.64, "learning_rate": 7.906724255736019e-06, "loss": 0.0268, "step": 25307 }, { "epoch": 4.64, "learning_rate": 7.905998279591183e-06, "loss": 0.0211, "step": 25308 }, { "epoch": 4.64, "learning_rate": 7.90527231498853e-06, "loss": 0.0197, "step": 25309 }, { "epoch": 4.64, "learning_rate": 7.904546361932069e-06, "loss": 0.0263, "step": 25310 }, { "epoch": 4.64, "learning_rate": 7.903820420425798e-06, "loss": 0.0272, "step": 25311 }, { "epoch": 4.64, "learning_rate": 7.903094490473717e-06, "loss": 0.0222, "step": 25312 }, { "epoch": 4.64, "learning_rate": 7.902368572079828e-06, "loss": 0.0251, "step": 25313 }, { "epoch": 4.64, "learning_rate": 7.90164266524813e-06, "loss": 0.0192, "step": 25314 }, { "epoch": 4.64, "learning_rate": 7.900916769982629e-06, "loss": 0.0239, "step": 25315 }, { "epoch": 4.64, "learning_rate": 7.900190886287326e-06, "loss": 0.0237, "step": 25316 }, { "epoch": 4.64, "learning_rate": 7.899465014166218e-06, "loss": 0.0236, "step": 25317 }, { "epoch": 4.64, "learning_rate": 7.898739153623306e-06, "loss": 0.0219, "step": 25318 }, { "epoch": 4.64, "learning_rate": 7.898013304662594e-06, "loss": 0.0234, "step": 25319 }, { "epoch": 4.64, "learning_rate": 7.897287467288088e-06, "loss": 0.0233, "step": 25320 }, { "epoch": 4.64, "learning_rate": 7.896561641503777e-06, "loss": 0.0225, "step": 25321 }, { "epoch": 4.64, "learning_rate": 7.895835827313668e-06, "loss": 0.0222, "step": 25322 }, { "epoch": 4.64, "learning_rate": 7.89511002472176e-06, "loss": 0.025, "step": 25323 }, { "epoch": 4.64, "learning_rate": 7.894384233732057e-06, "loss": 0.0281, "step": 25324 }, { "epoch": 4.64, "learning_rate": 7.893658454348558e-06, "loss": 0.0193, "step": 25325 }, { "epoch": 4.64, "learning_rate": 7.89293268657526e-06, "loss": 0.0232, "step": 25326 }, { "epoch": 4.64, "learning_rate": 7.892206930416167e-06, "loss": 0.0214, "step": 25327 }, { "epoch": 4.64, "learning_rate": 7.89148118587528e-06, "loss": 0.0266, "step": 25328 }, { "epoch": 4.64, "learning_rate": 7.8907554529566e-06, "loss": 0.0219, "step": 25329 }, { "epoch": 4.65, "learning_rate": 7.890029731664126e-06, "loss": 0.0187, "step": 25330 }, { "epoch": 4.65, "learning_rate": 7.889304022001854e-06, "loss": 0.0223, "step": 25331 }, { "epoch": 4.65, "learning_rate": 7.888578323973788e-06, "loss": 0.0213, "step": 25332 }, { "epoch": 4.65, "learning_rate": 7.887852637583927e-06, "loss": 0.0227, "step": 25333 }, { "epoch": 4.65, "learning_rate": 7.887126962836274e-06, "loss": 0.0244, "step": 25334 }, { "epoch": 4.65, "learning_rate": 7.886401299734827e-06, "loss": 0.0204, "step": 25335 }, { "epoch": 4.65, "learning_rate": 7.885675648283581e-06, "loss": 0.0268, "step": 25336 }, { "epoch": 4.65, "learning_rate": 7.884950008486545e-06, "loss": 0.0208, "step": 25337 }, { "epoch": 4.65, "learning_rate": 7.884224380347714e-06, "loss": 0.0211, "step": 25338 }, { "epoch": 4.65, "learning_rate": 7.883498763871088e-06, "loss": 0.0239, "step": 25339 }, { "epoch": 4.65, "learning_rate": 7.882773159060663e-06, "loss": 0.0161, "step": 25340 }, { "epoch": 4.65, "learning_rate": 7.882047565920442e-06, "loss": 0.0193, "step": 25341 }, { "epoch": 4.65, "learning_rate": 7.881321984454428e-06, "loss": 0.0258, "step": 25342 }, { "epoch": 4.65, "learning_rate": 7.880596414666615e-06, "loss": 0.0172, "step": 25343 }, { "epoch": 4.65, "learning_rate": 7.879870856561006e-06, "loss": 0.0243, "step": 25344 }, { "epoch": 4.65, "learning_rate": 7.879145310141595e-06, "loss": 0.0209, "step": 25345 }, { "epoch": 4.65, "learning_rate": 7.878419775412387e-06, "loss": 0.0247, "step": 25346 }, { "epoch": 4.65, "learning_rate": 7.877694252377381e-06, "loss": 0.0198, "step": 25347 }, { "epoch": 4.65, "learning_rate": 7.876968741040572e-06, "loss": 0.0246, "step": 25348 }, { "epoch": 4.65, "learning_rate": 7.876243241405959e-06, "loss": 0.0206, "step": 25349 }, { "epoch": 4.65, "learning_rate": 7.875517753477544e-06, "loss": 0.0229, "step": 25350 }, { "epoch": 4.65, "learning_rate": 7.874792277259326e-06, "loss": 0.0201, "step": 25351 }, { "epoch": 4.65, "learning_rate": 7.874066812755303e-06, "loss": 0.0251, "step": 25352 }, { "epoch": 4.65, "learning_rate": 7.873341359969474e-06, "loss": 0.0221, "step": 25353 }, { "epoch": 4.65, "learning_rate": 7.872615918905833e-06, "loss": 0.0215, "step": 25354 }, { "epoch": 4.65, "learning_rate": 7.871890489568387e-06, "loss": 0.0219, "step": 25355 }, { "epoch": 4.65, "learning_rate": 7.871165071961132e-06, "loss": 0.0233, "step": 25356 }, { "epoch": 4.65, "learning_rate": 7.870439666088062e-06, "loss": 0.0224, "step": 25357 }, { "epoch": 4.65, "learning_rate": 7.869714271953177e-06, "loss": 0.0225, "step": 25358 }, { "epoch": 4.65, "learning_rate": 7.868988889560479e-06, "loss": 0.022, "step": 25359 }, { "epoch": 4.65, "learning_rate": 7.868263518913965e-06, "loss": 0.0215, "step": 25360 }, { "epoch": 4.65, "learning_rate": 7.867538160017632e-06, "loss": 0.0191, "step": 25361 }, { "epoch": 4.65, "learning_rate": 7.866812812875475e-06, "loss": 0.0211, "step": 25362 }, { "epoch": 4.65, "learning_rate": 7.8660874774915e-06, "loss": 0.0248, "step": 25363 }, { "epoch": 4.65, "learning_rate": 7.865362153869698e-06, "loss": 0.0213, "step": 25364 }, { "epoch": 4.65, "learning_rate": 7.864636842014075e-06, "loss": 0.02, "step": 25365 }, { "epoch": 4.65, "learning_rate": 7.863911541928621e-06, "loss": 0.023, "step": 25366 }, { "epoch": 4.65, "learning_rate": 7.863186253617332e-06, "loss": 0.0234, "step": 25367 }, { "epoch": 4.65, "learning_rate": 7.862460977084214e-06, "loss": 0.0228, "step": 25368 }, { "epoch": 4.65, "learning_rate": 7.861735712333262e-06, "loss": 0.0232, "step": 25369 }, { "epoch": 4.65, "learning_rate": 7.861010459368473e-06, "loss": 0.0226, "step": 25370 }, { "epoch": 4.65, "learning_rate": 7.860285218193841e-06, "loss": 0.0214, "step": 25371 }, { "epoch": 4.65, "learning_rate": 7.85955998881337e-06, "loss": 0.024, "step": 25372 }, { "epoch": 4.65, "learning_rate": 7.858834771231055e-06, "loss": 0.0229, "step": 25373 }, { "epoch": 4.65, "learning_rate": 7.858109565450896e-06, "loss": 0.0229, "step": 25374 }, { "epoch": 4.65, "learning_rate": 7.857384371476883e-06, "loss": 0.018, "step": 25375 }, { "epoch": 4.65, "learning_rate": 7.856659189313016e-06, "loss": 0.0221, "step": 25376 }, { "epoch": 4.65, "learning_rate": 7.855934018963296e-06, "loss": 0.0254, "step": 25377 }, { "epoch": 4.65, "learning_rate": 7.855208860431718e-06, "loss": 0.0204, "step": 25378 }, { "epoch": 4.65, "learning_rate": 7.854483713722278e-06, "loss": 0.0282, "step": 25379 }, { "epoch": 4.65, "learning_rate": 7.853758578838973e-06, "loss": 0.0273, "step": 25380 }, { "epoch": 4.65, "learning_rate": 7.853033455785803e-06, "loss": 0.0245, "step": 25381 }, { "epoch": 4.65, "learning_rate": 7.852308344566763e-06, "loss": 0.0257, "step": 25382 }, { "epoch": 4.65, "learning_rate": 7.851583245185851e-06, "loss": 0.0245, "step": 25383 }, { "epoch": 4.66, "learning_rate": 7.850858157647061e-06, "loss": 0.023, "step": 25384 }, { "epoch": 4.66, "learning_rate": 7.850133081954388e-06, "loss": 0.024, "step": 25385 }, { "epoch": 4.66, "learning_rate": 7.849408018111836e-06, "loss": 0.0203, "step": 25386 }, { "epoch": 4.66, "learning_rate": 7.848682966123395e-06, "loss": 0.0271, "step": 25387 }, { "epoch": 4.66, "learning_rate": 7.847957925993066e-06, "loss": 0.0251, "step": 25388 }, { "epoch": 4.66, "learning_rate": 7.84723289772484e-06, "loss": 0.024, "step": 25389 }, { "epoch": 4.66, "learning_rate": 7.84650788132272e-06, "loss": 0.0211, "step": 25390 }, { "epoch": 4.66, "learning_rate": 7.845782876790698e-06, "loss": 0.0235, "step": 25391 }, { "epoch": 4.66, "learning_rate": 7.845057884132775e-06, "loss": 0.0257, "step": 25392 }, { "epoch": 4.66, "learning_rate": 7.844332903352933e-06, "loss": 0.0227, "step": 25393 }, { "epoch": 4.66, "learning_rate": 7.843607934455189e-06, "loss": 0.0188, "step": 25394 }, { "epoch": 4.66, "learning_rate": 7.842882977443524e-06, "loss": 0.0235, "step": 25395 }, { "epoch": 4.66, "learning_rate": 7.84215803232194e-06, "loss": 0.0225, "step": 25396 }, { "epoch": 4.66, "learning_rate": 7.841433099094431e-06, "loss": 0.0205, "step": 25397 }, { "epoch": 4.66, "learning_rate": 7.840708177764992e-06, "loss": 0.0175, "step": 25398 }, { "epoch": 4.66, "learning_rate": 7.839983268337622e-06, "loss": 0.0227, "step": 25399 }, { "epoch": 4.66, "learning_rate": 7.839258370816314e-06, "loss": 0.0229, "step": 25400 }, { "epoch": 4.66, "learning_rate": 7.838533485205066e-06, "loss": 0.0304, "step": 25401 }, { "epoch": 4.66, "learning_rate": 7.837808611507867e-06, "loss": 0.0296, "step": 25402 }, { "epoch": 4.66, "learning_rate": 7.837083749728725e-06, "loss": 0.0232, "step": 25403 }, { "epoch": 4.66, "learning_rate": 7.836358899871623e-06, "loss": 0.0215, "step": 25404 }, { "epoch": 4.66, "learning_rate": 7.835634061940563e-06, "loss": 0.0212, "step": 25405 }, { "epoch": 4.66, "learning_rate": 7.834909235939538e-06, "loss": 0.0249, "step": 25406 }, { "epoch": 4.66, "learning_rate": 7.834184421872543e-06, "loss": 0.0214, "step": 25407 }, { "epoch": 4.66, "learning_rate": 7.833459619743577e-06, "loss": 0.0231, "step": 25408 }, { "epoch": 4.66, "learning_rate": 7.832734829556633e-06, "loss": 0.0216, "step": 25409 }, { "epoch": 4.66, "learning_rate": 7.832010051315703e-06, "loss": 0.0219, "step": 25410 }, { "epoch": 4.66, "learning_rate": 7.83128528502478e-06, "loss": 0.0248, "step": 25411 }, { "epoch": 4.66, "learning_rate": 7.830560530687871e-06, "loss": 0.0266, "step": 25412 }, { "epoch": 4.66, "learning_rate": 7.829835788308961e-06, "loss": 0.019, "step": 25413 }, { "epoch": 4.66, "learning_rate": 7.829111057892047e-06, "loss": 0.0225, "step": 25414 }, { "epoch": 4.66, "learning_rate": 7.828386339441121e-06, "loss": 0.0251, "step": 25415 }, { "epoch": 4.66, "learning_rate": 7.827661632960182e-06, "loss": 0.02, "step": 25416 }, { "epoch": 4.66, "learning_rate": 7.826936938453225e-06, "loss": 0.0182, "step": 25417 }, { "epoch": 4.66, "learning_rate": 7.826212255924243e-06, "loss": 0.0208, "step": 25418 }, { "epoch": 4.66, "learning_rate": 7.825487585377227e-06, "loss": 0.0271, "step": 25419 }, { "epoch": 4.66, "learning_rate": 7.824762926816173e-06, "loss": 0.0239, "step": 25420 }, { "epoch": 4.66, "learning_rate": 7.82403828024508e-06, "loss": 0.0183, "step": 25421 }, { "epoch": 4.66, "learning_rate": 7.823313645667938e-06, "loss": 0.0188, "step": 25422 }, { "epoch": 4.66, "learning_rate": 7.822589023088743e-06, "loss": 0.0216, "step": 25423 }, { "epoch": 4.66, "learning_rate": 7.821864412511485e-06, "loss": 0.0206, "step": 25424 }, { "epoch": 4.66, "learning_rate": 7.821139813940163e-06, "loss": 0.0235, "step": 25425 }, { "epoch": 4.66, "learning_rate": 7.820415227378771e-06, "loss": 0.021, "step": 25426 }, { "epoch": 4.66, "learning_rate": 7.819690652831303e-06, "loss": 0.0203, "step": 25427 }, { "epoch": 4.66, "learning_rate": 7.81896609030175e-06, "loss": 0.0225, "step": 25428 }, { "epoch": 4.66, "learning_rate": 7.8182415397941e-06, "loss": 0.0246, "step": 25429 }, { "epoch": 4.66, "learning_rate": 7.817517001312363e-06, "loss": 0.0189, "step": 25430 }, { "epoch": 4.66, "learning_rate": 7.81679247486052e-06, "loss": 0.0215, "step": 25431 }, { "epoch": 4.66, "learning_rate": 7.816067960442569e-06, "loss": 0.0253, "step": 25432 }, { "epoch": 4.66, "learning_rate": 7.8153434580625e-06, "loss": 0.0234, "step": 25433 }, { "epoch": 4.66, "learning_rate": 7.81461896772431e-06, "loss": 0.021, "step": 25434 }, { "epoch": 4.66, "learning_rate": 7.813894489431992e-06, "loss": 0.0192, "step": 25435 }, { "epoch": 4.66, "learning_rate": 7.81317002318954e-06, "loss": 0.0213, "step": 25436 }, { "epoch": 4.66, "learning_rate": 7.81244556900094e-06, "loss": 0.0227, "step": 25437 }, { "epoch": 4.66, "learning_rate": 7.811721126870198e-06, "loss": 0.0253, "step": 25438 }, { "epoch": 4.67, "learning_rate": 7.8109966968013e-06, "loss": 0.0223, "step": 25439 }, { "epoch": 4.67, "learning_rate": 7.810272278798237e-06, "loss": 0.0227, "step": 25440 }, { "epoch": 4.67, "learning_rate": 7.809547872865007e-06, "loss": 0.0208, "step": 25441 }, { "epoch": 4.67, "learning_rate": 7.808823479005595e-06, "loss": 0.0188, "step": 25442 }, { "epoch": 4.67, "learning_rate": 7.808099097224005e-06, "loss": 0.0195, "step": 25443 }, { "epoch": 4.67, "learning_rate": 7.807374727524223e-06, "loss": 0.0243, "step": 25444 }, { "epoch": 4.67, "learning_rate": 7.806650369910244e-06, "loss": 0.0212, "step": 25445 }, { "epoch": 4.67, "learning_rate": 7.805926024386055e-06, "loss": 0.0306, "step": 25446 }, { "epoch": 4.67, "learning_rate": 7.805201690955657e-06, "loss": 0.0251, "step": 25447 }, { "epoch": 4.67, "learning_rate": 7.804477369623038e-06, "loss": 0.0238, "step": 25448 }, { "epoch": 4.67, "learning_rate": 7.803753060392192e-06, "loss": 0.0206, "step": 25449 }, { "epoch": 4.67, "learning_rate": 7.80302876326711e-06, "loss": 0.0235, "step": 25450 }, { "epoch": 4.67, "learning_rate": 7.80230447825178e-06, "loss": 0.0196, "step": 25451 }, { "epoch": 4.67, "learning_rate": 7.801580205350206e-06, "loss": 0.0223, "step": 25452 }, { "epoch": 4.67, "learning_rate": 7.80085594456637e-06, "loss": 0.022, "step": 25453 }, { "epoch": 4.67, "learning_rate": 7.800131695904272e-06, "loss": 0.022, "step": 25454 }, { "epoch": 4.67, "learning_rate": 7.799407459367891e-06, "loss": 0.0193, "step": 25455 }, { "epoch": 4.67, "learning_rate": 7.798683234961235e-06, "loss": 0.022, "step": 25456 }, { "epoch": 4.67, "learning_rate": 7.797959022688286e-06, "loss": 0.0237, "step": 25457 }, { "epoch": 4.67, "learning_rate": 7.79723482255304e-06, "loss": 0.0222, "step": 25458 }, { "epoch": 4.67, "learning_rate": 7.796510634559487e-06, "loss": 0.0189, "step": 25459 }, { "epoch": 4.67, "learning_rate": 7.795786458711615e-06, "loss": 0.0241, "step": 25460 }, { "epoch": 4.67, "learning_rate": 7.795062295013422e-06, "loss": 0.024, "step": 25461 }, { "epoch": 4.67, "learning_rate": 7.794338143468899e-06, "loss": 0.025, "step": 25462 }, { "epoch": 4.67, "learning_rate": 7.793614004082037e-06, "loss": 0.0225, "step": 25463 }, { "epoch": 4.67, "learning_rate": 7.792889876856821e-06, "loss": 0.0209, "step": 25464 }, { "epoch": 4.67, "learning_rate": 7.792165761797252e-06, "loss": 0.0249, "step": 25465 }, { "epoch": 4.67, "learning_rate": 7.791441658907318e-06, "loss": 0.0253, "step": 25466 }, { "epoch": 4.67, "learning_rate": 7.790717568191007e-06, "loss": 0.0244, "step": 25467 }, { "epoch": 4.67, "learning_rate": 7.78999348965231e-06, "loss": 0.0204, "step": 25468 }, { "epoch": 4.67, "learning_rate": 7.789269423295224e-06, "loss": 0.0204, "step": 25469 }, { "epoch": 4.67, "learning_rate": 7.788545369123736e-06, "loss": 0.0204, "step": 25470 }, { "epoch": 4.67, "learning_rate": 7.787821327141838e-06, "loss": 0.0235, "step": 25471 }, { "epoch": 4.67, "learning_rate": 7.787097297353524e-06, "loss": 0.019, "step": 25472 }, { "epoch": 4.67, "learning_rate": 7.786373279762773e-06, "loss": 0.0181, "step": 25473 }, { "epoch": 4.67, "learning_rate": 7.785649274373592e-06, "loss": 0.0225, "step": 25474 }, { "epoch": 4.67, "learning_rate": 7.784925281189962e-06, "loss": 0.0238, "step": 25475 }, { "epoch": 4.67, "learning_rate": 7.784201300215875e-06, "loss": 0.0171, "step": 25476 }, { "epoch": 4.67, "learning_rate": 7.78347733145532e-06, "loss": 0.0184, "step": 25477 }, { "epoch": 4.67, "learning_rate": 7.782753374912292e-06, "loss": 0.0178, "step": 25478 }, { "epoch": 4.67, "learning_rate": 7.782029430590781e-06, "loss": 0.0226, "step": 25479 }, { "epoch": 4.67, "learning_rate": 7.781305498494776e-06, "loss": 0.0273, "step": 25480 }, { "epoch": 4.67, "learning_rate": 7.780581578628267e-06, "loss": 0.0267, "step": 25481 }, { "epoch": 4.67, "learning_rate": 7.779857670995239e-06, "loss": 0.0207, "step": 25482 }, { "epoch": 4.67, "learning_rate": 7.779133775599694e-06, "loss": 0.0288, "step": 25483 }, { "epoch": 4.67, "learning_rate": 7.778409892445614e-06, "loss": 0.0266, "step": 25484 }, { "epoch": 4.67, "learning_rate": 7.77768602153699e-06, "loss": 0.0231, "step": 25485 }, { "epoch": 4.67, "learning_rate": 7.77696216287781e-06, "loss": 0.0223, "step": 25486 }, { "epoch": 4.67, "learning_rate": 7.776238316472069e-06, "loss": 0.0192, "step": 25487 }, { "epoch": 4.67, "learning_rate": 7.775514482323756e-06, "loss": 0.0218, "step": 25488 }, { "epoch": 4.67, "learning_rate": 7.774790660436857e-06, "loss": 0.0238, "step": 25489 }, { "epoch": 4.67, "learning_rate": 7.774066850815362e-06, "loss": 0.0211, "step": 25490 }, { "epoch": 4.67, "learning_rate": 7.773343053463268e-06, "loss": 0.025, "step": 25491 }, { "epoch": 4.67, "learning_rate": 7.772619268384559e-06, "loss": 0.0256, "step": 25492 }, { "epoch": 4.68, "learning_rate": 7.771895495583223e-06, "loss": 0.0196, "step": 25493 }, { "epoch": 4.68, "learning_rate": 7.77117173506325e-06, "loss": 0.0242, "step": 25494 }, { "epoch": 4.68, "learning_rate": 7.77044798682863e-06, "loss": 0.0197, "step": 25495 }, { "epoch": 4.68, "learning_rate": 7.769724250883355e-06, "loss": 0.0248, "step": 25496 }, { "epoch": 4.68, "learning_rate": 7.769000527231412e-06, "loss": 0.0247, "step": 25497 }, { "epoch": 4.68, "learning_rate": 7.76827681587679e-06, "loss": 0.0339, "step": 25498 }, { "epoch": 4.68, "learning_rate": 7.767553116823477e-06, "loss": 0.0236, "step": 25499 }, { "epoch": 4.68, "learning_rate": 7.766829430075464e-06, "loss": 0.0251, "step": 25500 }, { "epoch": 4.68, "learning_rate": 7.766105755636744e-06, "loss": 0.0158, "step": 25501 }, { "epoch": 4.68, "learning_rate": 7.765382093511298e-06, "loss": 0.0249, "step": 25502 }, { "epoch": 4.68, "learning_rate": 7.764658443703117e-06, "loss": 0.0243, "step": 25503 }, { "epoch": 4.68, "learning_rate": 7.763934806216191e-06, "loss": 0.0219, "step": 25504 }, { "epoch": 4.68, "learning_rate": 7.763211181054511e-06, "loss": 0.0248, "step": 25505 }, { "epoch": 4.68, "learning_rate": 7.762487568222061e-06, "loss": 0.0253, "step": 25506 }, { "epoch": 4.68, "learning_rate": 7.761763967722834e-06, "loss": 0.0171, "step": 25507 }, { "epoch": 4.68, "learning_rate": 7.761040379560814e-06, "loss": 0.0232, "step": 25508 }, { "epoch": 4.68, "learning_rate": 7.760316803739994e-06, "loss": 0.0217, "step": 25509 }, { "epoch": 4.68, "learning_rate": 7.759593240264362e-06, "loss": 0.0258, "step": 25510 }, { "epoch": 4.68, "learning_rate": 7.758869689137903e-06, "loss": 0.025, "step": 25511 }, { "epoch": 4.68, "learning_rate": 7.758146150364606e-06, "loss": 0.0204, "step": 25512 }, { "epoch": 4.68, "learning_rate": 7.757422623948458e-06, "loss": 0.0223, "step": 25513 }, { "epoch": 4.68, "learning_rate": 7.75669910989345e-06, "loss": 0.0221, "step": 25514 }, { "epoch": 4.68, "learning_rate": 7.755975608203571e-06, "loss": 0.0201, "step": 25515 }, { "epoch": 4.68, "learning_rate": 7.75525211888281e-06, "loss": 0.0275, "step": 25516 }, { "epoch": 4.68, "learning_rate": 7.754528641935144e-06, "loss": 0.0198, "step": 25517 }, { "epoch": 4.68, "learning_rate": 7.753805177364573e-06, "loss": 0.0222, "step": 25518 }, { "epoch": 4.68, "learning_rate": 7.753081725175083e-06, "loss": 0.0216, "step": 25519 }, { "epoch": 4.68, "learning_rate": 7.752358285370656e-06, "loss": 0.0212, "step": 25520 }, { "epoch": 4.68, "learning_rate": 7.751634857955281e-06, "loss": 0.0254, "step": 25521 }, { "epoch": 4.68, "learning_rate": 7.75091144293295e-06, "loss": 0.0279, "step": 25522 }, { "epoch": 4.68, "learning_rate": 7.750188040307648e-06, "loss": 0.0259, "step": 25523 }, { "epoch": 4.68, "learning_rate": 7.749464650083361e-06, "loss": 0.0214, "step": 25524 }, { "epoch": 4.68, "learning_rate": 7.74874127226408e-06, "loss": 0.0217, "step": 25525 }, { "epoch": 4.68, "learning_rate": 7.748017906853786e-06, "loss": 0.0212, "step": 25526 }, { "epoch": 4.68, "learning_rate": 7.747294553856472e-06, "loss": 0.0227, "step": 25527 }, { "epoch": 4.68, "learning_rate": 7.746571213276125e-06, "loss": 0.0264, "step": 25528 }, { "epoch": 4.68, "learning_rate": 7.74584788511673e-06, "loss": 0.0208, "step": 25529 }, { "epoch": 4.68, "learning_rate": 7.74512456938227e-06, "loss": 0.0246, "step": 25530 }, { "epoch": 4.68, "learning_rate": 7.744401266076738e-06, "loss": 0.0259, "step": 25531 }, { "epoch": 4.68, "learning_rate": 7.743677975204122e-06, "loss": 0.0243, "step": 25532 }, { "epoch": 4.68, "learning_rate": 7.742954696768404e-06, "loss": 0.0281, "step": 25533 }, { "epoch": 4.68, "learning_rate": 7.742231430773574e-06, "loss": 0.0224, "step": 25534 }, { "epoch": 4.68, "learning_rate": 7.741508177223611e-06, "loss": 0.0204, "step": 25535 }, { "epoch": 4.68, "learning_rate": 7.740784936122514e-06, "loss": 0.0186, "step": 25536 }, { "epoch": 4.68, "learning_rate": 7.740061707474265e-06, "loss": 0.0236, "step": 25537 }, { "epoch": 4.68, "learning_rate": 7.739338491282846e-06, "loss": 0.0212, "step": 25538 }, { "epoch": 4.68, "learning_rate": 7.738615287552244e-06, "loss": 0.0231, "step": 25539 }, { "epoch": 4.68, "learning_rate": 7.73789209628645e-06, "loss": 0.024, "step": 25540 }, { "epoch": 4.68, "learning_rate": 7.737168917489448e-06, "loss": 0.0226, "step": 25541 }, { "epoch": 4.68, "learning_rate": 7.736445751165222e-06, "loss": 0.0219, "step": 25542 }, { "epoch": 4.68, "learning_rate": 7.73572259731776e-06, "loss": 0.0251, "step": 25543 }, { "epoch": 4.68, "learning_rate": 7.73499945595105e-06, "loss": 0.0183, "step": 25544 }, { "epoch": 4.68, "learning_rate": 7.734276327069075e-06, "loss": 0.0194, "step": 25545 }, { "epoch": 4.68, "learning_rate": 7.733553210675825e-06, "loss": 0.0193, "step": 25546 }, { "epoch": 4.68, "learning_rate": 7.73283010677528e-06, "loss": 0.0212, "step": 25547 }, { "epoch": 4.69, "learning_rate": 7.732107015371428e-06, "loss": 0.0236, "step": 25548 }, { "epoch": 4.69, "learning_rate": 7.731383936468255e-06, "loss": 0.0246, "step": 25549 }, { "epoch": 4.69, "learning_rate": 7.730660870069749e-06, "loss": 0.0185, "step": 25550 }, { "epoch": 4.69, "learning_rate": 7.729937816179892e-06, "loss": 0.0189, "step": 25551 }, { "epoch": 4.69, "learning_rate": 7.729214774802669e-06, "loss": 0.026, "step": 25552 }, { "epoch": 4.69, "learning_rate": 7.72849174594207e-06, "loss": 0.0192, "step": 25553 }, { "epoch": 4.69, "learning_rate": 7.727768729602078e-06, "loss": 0.0179, "step": 25554 }, { "epoch": 4.69, "learning_rate": 7.72704572578668e-06, "loss": 0.0229, "step": 25555 }, { "epoch": 4.69, "learning_rate": 7.726322734499856e-06, "loss": 0.0204, "step": 25556 }, { "epoch": 4.69, "learning_rate": 7.725599755745593e-06, "loss": 0.021, "step": 25557 }, { "epoch": 4.69, "learning_rate": 7.72487678952788e-06, "loss": 0.0211, "step": 25558 }, { "epoch": 4.69, "learning_rate": 7.724153835850699e-06, "loss": 0.0252, "step": 25559 }, { "epoch": 4.69, "learning_rate": 7.723430894718036e-06, "loss": 0.0226, "step": 25560 }, { "epoch": 4.69, "learning_rate": 7.722707966133873e-06, "loss": 0.0215, "step": 25561 }, { "epoch": 4.69, "learning_rate": 7.721985050102198e-06, "loss": 0.0201, "step": 25562 }, { "epoch": 4.69, "learning_rate": 7.721262146626998e-06, "loss": 0.0214, "step": 25563 }, { "epoch": 4.69, "learning_rate": 7.720539255712252e-06, "loss": 0.0168, "step": 25564 }, { "epoch": 4.69, "learning_rate": 7.719816377361942e-06, "loss": 0.0205, "step": 25565 }, { "epoch": 4.69, "learning_rate": 7.719093511580064e-06, "loss": 0.0275, "step": 25566 }, { "epoch": 4.69, "learning_rate": 7.718370658370595e-06, "loss": 0.0275, "step": 25567 }, { "epoch": 4.69, "learning_rate": 7.717647817737519e-06, "loss": 0.0243, "step": 25568 }, { "epoch": 4.69, "learning_rate": 7.716924989684823e-06, "loss": 0.021, "step": 25569 }, { "epoch": 4.69, "learning_rate": 7.716202174216488e-06, "loss": 0.0261, "step": 25570 }, { "epoch": 4.69, "learning_rate": 7.7154793713365e-06, "loss": 0.0301, "step": 25571 }, { "epoch": 4.69, "learning_rate": 7.714756581048848e-06, "loss": 0.0263, "step": 25572 }, { "epoch": 4.69, "learning_rate": 7.714033803357507e-06, "loss": 0.0222, "step": 25573 }, { "epoch": 4.69, "learning_rate": 7.713311038266463e-06, "loss": 0.0214, "step": 25574 }, { "epoch": 4.69, "learning_rate": 7.712588285779708e-06, "loss": 0.02, "step": 25575 }, { "epoch": 4.69, "learning_rate": 7.711865545901217e-06, "loss": 0.0178, "step": 25576 }, { "epoch": 4.69, "learning_rate": 7.711142818634978e-06, "loss": 0.0279, "step": 25577 }, { "epoch": 4.69, "learning_rate": 7.710420103984973e-06, "loss": 0.0216, "step": 25578 }, { "epoch": 4.69, "learning_rate": 7.709697401955182e-06, "loss": 0.024, "step": 25579 }, { "epoch": 4.69, "learning_rate": 7.708974712549596e-06, "loss": 0.0247, "step": 25580 }, { "epoch": 4.69, "learning_rate": 7.708252035772198e-06, "loss": 0.0239, "step": 25581 }, { "epoch": 4.69, "learning_rate": 7.707529371626966e-06, "loss": 0.0206, "step": 25582 }, { "epoch": 4.69, "learning_rate": 7.706806720117881e-06, "loss": 0.0226, "step": 25583 }, { "epoch": 4.69, "learning_rate": 7.706084081248937e-06, "loss": 0.0228, "step": 25584 }, { "epoch": 4.69, "learning_rate": 7.705361455024111e-06, "loss": 0.0234, "step": 25585 }, { "epoch": 4.69, "learning_rate": 7.704638841447385e-06, "loss": 0.0234, "step": 25586 }, { "epoch": 4.69, "learning_rate": 7.703916240522744e-06, "loss": 0.0165, "step": 25587 }, { "epoch": 4.69, "learning_rate": 7.703193652254167e-06, "loss": 0.0222, "step": 25588 }, { "epoch": 4.69, "learning_rate": 7.702471076645643e-06, "loss": 0.0253, "step": 25589 }, { "epoch": 4.69, "learning_rate": 7.701748513701157e-06, "loss": 0.0242, "step": 25590 }, { "epoch": 4.69, "learning_rate": 7.701025963424683e-06, "loss": 0.0247, "step": 25591 }, { "epoch": 4.69, "learning_rate": 7.700303425820203e-06, "loss": 0.0196, "step": 25592 }, { "epoch": 4.69, "learning_rate": 7.69958090089171e-06, "loss": 0.0239, "step": 25593 }, { "epoch": 4.69, "learning_rate": 7.69885838864318e-06, "loss": 0.021, "step": 25594 }, { "epoch": 4.69, "learning_rate": 7.698135889078598e-06, "loss": 0.0216, "step": 25595 }, { "epoch": 4.69, "learning_rate": 7.69741340220194e-06, "loss": 0.0229, "step": 25596 }, { "epoch": 4.69, "learning_rate": 7.696690928017196e-06, "loss": 0.0214, "step": 25597 }, { "epoch": 4.69, "learning_rate": 7.695968466528347e-06, "loss": 0.0247, "step": 25598 }, { "epoch": 4.69, "learning_rate": 7.695246017739375e-06, "loss": 0.0236, "step": 25599 }, { "epoch": 4.69, "learning_rate": 7.694523581654259e-06, "loss": 0.0205, "step": 25600 }, { "epoch": 4.69, "learning_rate": 7.693801158276978e-06, "loss": 0.0258, "step": 25601 }, { "epoch": 4.7, "learning_rate": 7.693078747611526e-06, "loss": 0.0233, "step": 25602 }, { "epoch": 4.7, "learning_rate": 7.692356349661878e-06, "loss": 0.0207, "step": 25603 }, { "epoch": 4.7, "learning_rate": 7.691633964432013e-06, "loss": 0.0223, "step": 25604 }, { "epoch": 4.7, "learning_rate": 7.690911591925913e-06, "loss": 0.0261, "step": 25605 }, { "epoch": 4.7, "learning_rate": 7.690189232147566e-06, "loss": 0.0187, "step": 25606 }, { "epoch": 4.7, "learning_rate": 7.689466885100951e-06, "loss": 0.0174, "step": 25607 }, { "epoch": 4.7, "learning_rate": 7.68874455079005e-06, "loss": 0.0195, "step": 25608 }, { "epoch": 4.7, "learning_rate": 7.688022229218842e-06, "loss": 0.0255, "step": 25609 }, { "epoch": 4.7, "learning_rate": 7.687299920391305e-06, "loss": 0.0243, "step": 25610 }, { "epoch": 4.7, "learning_rate": 7.686577624311428e-06, "loss": 0.0206, "step": 25611 }, { "epoch": 4.7, "learning_rate": 7.68585534098319e-06, "loss": 0.0244, "step": 25612 }, { "epoch": 4.7, "learning_rate": 7.685133070410571e-06, "loss": 0.0215, "step": 25613 }, { "epoch": 4.7, "learning_rate": 7.68441081259755e-06, "loss": 0.0253, "step": 25614 }, { "epoch": 4.7, "learning_rate": 7.683688567548115e-06, "loss": 0.0237, "step": 25615 }, { "epoch": 4.7, "learning_rate": 7.682966335266243e-06, "loss": 0.0221, "step": 25616 }, { "epoch": 4.7, "learning_rate": 7.682244115755915e-06, "loss": 0.0219, "step": 25617 }, { "epoch": 4.7, "learning_rate": 7.681521909021107e-06, "loss": 0.0219, "step": 25618 }, { "epoch": 4.7, "learning_rate": 7.68079971506581e-06, "loss": 0.0223, "step": 25619 }, { "epoch": 4.7, "learning_rate": 7.680077533893998e-06, "loss": 0.0252, "step": 25620 }, { "epoch": 4.7, "learning_rate": 7.679355365509652e-06, "loss": 0.0184, "step": 25621 }, { "epoch": 4.7, "learning_rate": 7.678633209916753e-06, "loss": 0.0216, "step": 25622 }, { "epoch": 4.7, "learning_rate": 7.677911067119282e-06, "loss": 0.0185, "step": 25623 }, { "epoch": 4.7, "learning_rate": 7.67718893712122e-06, "loss": 0.0172, "step": 25624 }, { "epoch": 4.7, "learning_rate": 7.676466819926548e-06, "loss": 0.0249, "step": 25625 }, { "epoch": 4.7, "learning_rate": 7.675744715539248e-06, "loss": 0.0243, "step": 25626 }, { "epoch": 4.7, "learning_rate": 7.675022623963289e-06, "loss": 0.0239, "step": 25627 }, { "epoch": 4.7, "learning_rate": 7.674300545202667e-06, "loss": 0.0233, "step": 25628 }, { "epoch": 4.7, "learning_rate": 7.673578479261354e-06, "loss": 0.0216, "step": 25629 }, { "epoch": 4.7, "learning_rate": 7.67285642614333e-06, "loss": 0.0202, "step": 25630 }, { "epoch": 4.7, "learning_rate": 7.672134385852576e-06, "loss": 0.0262, "step": 25631 }, { "epoch": 4.7, "learning_rate": 7.67141235839307e-06, "loss": 0.019, "step": 25632 }, { "epoch": 4.7, "learning_rate": 7.670690343768795e-06, "loss": 0.0235, "step": 25633 }, { "epoch": 4.7, "learning_rate": 7.66996834198373e-06, "loss": 0.0178, "step": 25634 }, { "epoch": 4.7, "learning_rate": 7.669246353041855e-06, "loss": 0.024, "step": 25635 }, { "epoch": 4.7, "learning_rate": 7.668524376947144e-06, "loss": 0.0221, "step": 25636 }, { "epoch": 4.7, "learning_rate": 7.667802413703589e-06, "loss": 0.0181, "step": 25637 }, { "epoch": 4.7, "learning_rate": 7.667080463315157e-06, "loss": 0.0207, "step": 25638 }, { "epoch": 4.7, "learning_rate": 7.666358525785832e-06, "loss": 0.0244, "step": 25639 }, { "epoch": 4.7, "learning_rate": 7.665636601119594e-06, "loss": 0.0216, "step": 25640 }, { "epoch": 4.7, "learning_rate": 7.66491468932042e-06, "loss": 0.0211, "step": 25641 }, { "epoch": 4.7, "learning_rate": 7.664192790392293e-06, "loss": 0.0221, "step": 25642 }, { "epoch": 4.7, "learning_rate": 7.66347090433919e-06, "loss": 0.02, "step": 25643 }, { "epoch": 4.7, "learning_rate": 7.662749031165093e-06, "loss": 0.0207, "step": 25644 }, { "epoch": 4.7, "learning_rate": 7.66202717087397e-06, "loss": 0.018, "step": 25645 }, { "epoch": 4.7, "learning_rate": 7.661305323469815e-06, "loss": 0.0215, "step": 25646 }, { "epoch": 4.7, "learning_rate": 7.660583488956599e-06, "loss": 0.0244, "step": 25647 }, { "epoch": 4.7, "learning_rate": 7.6598616673383e-06, "loss": 0.0268, "step": 25648 }, { "epoch": 4.7, "learning_rate": 7.659139858618897e-06, "loss": 0.0233, "step": 25649 }, { "epoch": 4.7, "learning_rate": 7.658418062802372e-06, "loss": 0.0195, "step": 25650 }, { "epoch": 4.7, "learning_rate": 7.657696279892702e-06, "loss": 0.0236, "step": 25651 }, { "epoch": 4.7, "learning_rate": 7.656974509893863e-06, "loss": 0.0178, "step": 25652 }, { "epoch": 4.7, "learning_rate": 7.65625275280984e-06, "loss": 0.0234, "step": 25653 }, { "epoch": 4.7, "learning_rate": 7.6555310086446e-06, "loss": 0.0218, "step": 25654 }, { "epoch": 4.7, "learning_rate": 7.654809277402133e-06, "loss": 0.0225, "step": 25655 }, { "epoch": 4.7, "learning_rate": 7.654087559086409e-06, "loss": 0.0236, "step": 25656 }, { "epoch": 4.71, "learning_rate": 7.653365853701412e-06, "loss": 0.0209, "step": 25657 }, { "epoch": 4.71, "learning_rate": 7.652644161251115e-06, "loss": 0.0172, "step": 25658 }, { "epoch": 4.71, "learning_rate": 7.651922481739498e-06, "loss": 0.0236, "step": 25659 }, { "epoch": 4.71, "learning_rate": 7.651200815170541e-06, "loss": 0.0195, "step": 25660 }, { "epoch": 4.71, "learning_rate": 7.650479161548221e-06, "loss": 0.0247, "step": 25661 }, { "epoch": 4.71, "learning_rate": 7.649757520876517e-06, "loss": 0.0172, "step": 25662 }, { "epoch": 4.71, "learning_rate": 7.649035893159398e-06, "loss": 0.016, "step": 25663 }, { "epoch": 4.71, "learning_rate": 7.648314278400854e-06, "loss": 0.0238, "step": 25664 }, { "epoch": 4.71, "learning_rate": 7.647592676604854e-06, "loss": 0.0228, "step": 25665 }, { "epoch": 4.71, "learning_rate": 7.64687108777538e-06, "loss": 0.0208, "step": 25666 }, { "epoch": 4.71, "learning_rate": 7.646149511916405e-06, "loss": 0.0231, "step": 25667 }, { "epoch": 4.71, "learning_rate": 7.64542794903191e-06, "loss": 0.0231, "step": 25668 }, { "epoch": 4.71, "learning_rate": 7.644706399125871e-06, "loss": 0.0214, "step": 25669 }, { "epoch": 4.71, "learning_rate": 7.64398486220227e-06, "loss": 0.0199, "step": 25670 }, { "epoch": 4.71, "learning_rate": 7.643263338265071e-06, "loss": 0.0292, "step": 25671 }, { "epoch": 4.71, "learning_rate": 7.642541827318264e-06, "loss": 0.0198, "step": 25672 }, { "epoch": 4.71, "learning_rate": 7.641820329365826e-06, "loss": 0.0217, "step": 25673 }, { "epoch": 4.71, "learning_rate": 7.641098844411726e-06, "loss": 0.0235, "step": 25674 }, { "epoch": 4.71, "learning_rate": 7.640377372459944e-06, "loss": 0.0187, "step": 25675 }, { "epoch": 4.71, "learning_rate": 7.639655913514457e-06, "loss": 0.02, "step": 25676 }, { "epoch": 4.71, "learning_rate": 7.638934467579241e-06, "loss": 0.0248, "step": 25677 }, { "epoch": 4.71, "learning_rate": 7.638213034658276e-06, "loss": 0.0242, "step": 25678 }, { "epoch": 4.71, "learning_rate": 7.637491614755539e-06, "loss": 0.0183, "step": 25679 }, { "epoch": 4.71, "learning_rate": 7.636770207874995e-06, "loss": 0.0239, "step": 25680 }, { "epoch": 4.71, "learning_rate": 7.636048814020633e-06, "loss": 0.0239, "step": 25681 }, { "epoch": 4.71, "learning_rate": 7.635327433196429e-06, "loss": 0.0216, "step": 25682 }, { "epoch": 4.71, "learning_rate": 7.634606065406352e-06, "loss": 0.0233, "step": 25683 }, { "epoch": 4.71, "learning_rate": 7.633884710654383e-06, "loss": 0.0221, "step": 25684 }, { "epoch": 4.71, "learning_rate": 7.633163368944494e-06, "loss": 0.0221, "step": 25685 }, { "epoch": 4.71, "learning_rate": 7.632442040280667e-06, "loss": 0.0246, "step": 25686 }, { "epoch": 4.71, "learning_rate": 7.631720724666874e-06, "loss": 0.0224, "step": 25687 }, { "epoch": 4.71, "learning_rate": 7.630999422107095e-06, "loss": 0.0186, "step": 25688 }, { "epoch": 4.71, "learning_rate": 7.630278132605295e-06, "loss": 0.0262, "step": 25689 }, { "epoch": 4.71, "learning_rate": 7.629556856165464e-06, "loss": 0.024, "step": 25690 }, { "epoch": 4.71, "learning_rate": 7.62883559279157e-06, "loss": 0.0219, "step": 25691 }, { "epoch": 4.71, "learning_rate": 7.628114342487589e-06, "loss": 0.0204, "step": 25692 }, { "epoch": 4.71, "learning_rate": 7.627393105257497e-06, "loss": 0.0224, "step": 25693 }, { "epoch": 4.71, "learning_rate": 7.626671881105269e-06, "loss": 0.0252, "step": 25694 }, { "epoch": 4.71, "learning_rate": 7.625950670034883e-06, "loss": 0.0197, "step": 25695 }, { "epoch": 4.71, "learning_rate": 7.6252294720503125e-06, "loss": 0.0213, "step": 25696 }, { "epoch": 4.71, "learning_rate": 7.624508287155534e-06, "loss": 0.0228, "step": 25697 }, { "epoch": 4.71, "learning_rate": 7.623787115354515e-06, "loss": 0.0184, "step": 25698 }, { "epoch": 4.71, "learning_rate": 7.6230659566512435e-06, "loss": 0.02, "step": 25699 }, { "epoch": 4.71, "learning_rate": 7.622344811049688e-06, "loss": 0.0224, "step": 25700 }, { "epoch": 4.71, "learning_rate": 7.621623678553822e-06, "loss": 0.0254, "step": 25701 }, { "epoch": 4.71, "learning_rate": 7.62090255916762e-06, "loss": 0.0214, "step": 25702 }, { "epoch": 4.71, "learning_rate": 7.620181452895061e-06, "loss": 0.0226, "step": 25703 }, { "epoch": 4.71, "learning_rate": 7.6194603597401174e-06, "loss": 0.0188, "step": 25704 }, { "epoch": 4.71, "learning_rate": 7.618739279706765e-06, "loss": 0.0213, "step": 25705 }, { "epoch": 4.71, "learning_rate": 7.618018212798978e-06, "loss": 0.02, "step": 25706 }, { "epoch": 4.71, "learning_rate": 7.6172971590207255e-06, "loss": 0.0189, "step": 25707 }, { "epoch": 4.71, "learning_rate": 7.61657611837599e-06, "loss": 0.0269, "step": 25708 }, { "epoch": 4.71, "learning_rate": 7.6158550908687455e-06, "loss": 0.0253, "step": 25709 }, { "epoch": 4.71, "learning_rate": 7.615134076502961e-06, "loss": 0.0218, "step": 25710 }, { "epoch": 4.72, "learning_rate": 7.614413075282612e-06, "loss": 0.0237, "step": 25711 }, { "epoch": 4.72, "learning_rate": 7.613692087211676e-06, "loss": 0.022, "step": 25712 }, { "epoch": 4.72, "learning_rate": 7.6129711122941254e-06, "loss": 0.0208, "step": 25713 }, { "epoch": 4.72, "learning_rate": 7.612250150533933e-06, "loss": 0.0231, "step": 25714 }, { "epoch": 4.72, "learning_rate": 7.611529201935076e-06, "loss": 0.0199, "step": 25715 }, { "epoch": 4.72, "learning_rate": 7.610808266501521e-06, "loss": 0.021, "step": 25716 }, { "epoch": 4.72, "learning_rate": 7.610087344237251e-06, "loss": 0.0239, "step": 25717 }, { "epoch": 4.72, "learning_rate": 7.609366435146235e-06, "loss": 0.0242, "step": 25718 }, { "epoch": 4.72, "learning_rate": 7.608645539232447e-06, "loss": 0.02, "step": 25719 }, { "epoch": 4.72, "learning_rate": 7.607924656499858e-06, "loss": 0.0271, "step": 25720 }, { "epoch": 4.72, "learning_rate": 7.607203786952446e-06, "loss": 0.0199, "step": 25721 }, { "epoch": 4.72, "learning_rate": 7.606482930594185e-06, "loss": 0.0149, "step": 25722 }, { "epoch": 4.72, "learning_rate": 7.605762087429044e-06, "loss": 0.0202, "step": 25723 }, { "epoch": 4.72, "learning_rate": 7.605041257460997e-06, "loss": 0.0203, "step": 25724 }, { "epoch": 4.72, "learning_rate": 7.6043204406940205e-06, "loss": 0.0239, "step": 25725 }, { "epoch": 4.72, "learning_rate": 7.603599637132089e-06, "loss": 0.0294, "step": 25726 }, { "epoch": 4.72, "learning_rate": 7.602878846779169e-06, "loss": 0.0271, "step": 25727 }, { "epoch": 4.72, "learning_rate": 7.602158069639238e-06, "loss": 0.0249, "step": 25728 }, { "epoch": 4.72, "learning_rate": 7.601437305716266e-06, "loss": 0.0207, "step": 25729 }, { "epoch": 4.72, "learning_rate": 7.600716555014229e-06, "loss": 0.0211, "step": 25730 }, { "epoch": 4.72, "learning_rate": 7.599995817537099e-06, "loss": 0.0204, "step": 25731 }, { "epoch": 4.72, "learning_rate": 7.599275093288848e-06, "loss": 0.022, "step": 25732 }, { "epoch": 4.72, "learning_rate": 7.598554382273448e-06, "loss": 0.0277, "step": 25733 }, { "epoch": 4.72, "learning_rate": 7.597833684494873e-06, "loss": 0.0227, "step": 25734 }, { "epoch": 4.72, "learning_rate": 7.597112999957098e-06, "loss": 0.0267, "step": 25735 }, { "epoch": 4.72, "learning_rate": 7.596392328664091e-06, "loss": 0.0209, "step": 25736 }, { "epoch": 4.72, "learning_rate": 7.595671670619824e-06, "loss": 0.028, "step": 25737 }, { "epoch": 4.72, "learning_rate": 7.59495102582827e-06, "loss": 0.0229, "step": 25738 }, { "epoch": 4.72, "learning_rate": 7.594230394293404e-06, "loss": 0.0245, "step": 25739 }, { "epoch": 4.72, "learning_rate": 7.593509776019196e-06, "loss": 0.0274, "step": 25740 }, { "epoch": 4.72, "learning_rate": 7.59278917100962e-06, "loss": 0.0219, "step": 25741 }, { "epoch": 4.72, "learning_rate": 7.5920685792686434e-06, "loss": 0.0176, "step": 25742 }, { "epoch": 4.72, "learning_rate": 7.591348000800243e-06, "loss": 0.0243, "step": 25743 }, { "epoch": 4.72, "learning_rate": 7.590627435608392e-06, "loss": 0.0232, "step": 25744 }, { "epoch": 4.72, "learning_rate": 7.589906883697056e-06, "loss": 0.0274, "step": 25745 }, { "epoch": 4.72, "learning_rate": 7.5891863450702055e-06, "loss": 0.0243, "step": 25746 }, { "epoch": 4.72, "learning_rate": 7.588465819731823e-06, "loss": 0.018, "step": 25747 }, { "epoch": 4.72, "learning_rate": 7.587745307685872e-06, "loss": 0.0205, "step": 25748 }, { "epoch": 4.72, "learning_rate": 7.587024808936324e-06, "loss": 0.0248, "step": 25749 }, { "epoch": 4.72, "learning_rate": 7.586304323487153e-06, "loss": 0.0234, "step": 25750 }, { "epoch": 4.72, "learning_rate": 7.5855838513423265e-06, "loss": 0.0272, "step": 25751 }, { "epoch": 4.72, "learning_rate": 7.58486339250582e-06, "loss": 0.0238, "step": 25752 }, { "epoch": 4.72, "learning_rate": 7.584142946981605e-06, "loss": 0.0259, "step": 25753 }, { "epoch": 4.72, "learning_rate": 7.5834225147736505e-06, "loss": 0.022, "step": 25754 }, { "epoch": 4.72, "learning_rate": 7.582702095885921e-06, "loss": 0.0201, "step": 25755 }, { "epoch": 4.72, "learning_rate": 7.581981690322402e-06, "loss": 0.0203, "step": 25756 }, { "epoch": 4.72, "learning_rate": 7.581261298087054e-06, "loss": 0.0248, "step": 25757 }, { "epoch": 4.72, "learning_rate": 7.580540919183851e-06, "loss": 0.0231, "step": 25758 }, { "epoch": 4.72, "learning_rate": 7.579820553616762e-06, "loss": 0.0215, "step": 25759 }, { "epoch": 4.72, "learning_rate": 7.579100201389757e-06, "loss": 0.0234, "step": 25760 }, { "epoch": 4.72, "learning_rate": 7.578379862506811e-06, "loss": 0.024, "step": 25761 }, { "epoch": 4.72, "learning_rate": 7.577659536971895e-06, "loss": 0.0255, "step": 25762 }, { "epoch": 4.72, "learning_rate": 7.5769392247889725e-06, "loss": 0.0198, "step": 25763 }, { "epoch": 4.72, "learning_rate": 7.576218925962016e-06, "loss": 0.0255, "step": 25764 }, { "epoch": 4.72, "learning_rate": 7.575498640495e-06, "loss": 0.0256, "step": 25765 }, { "epoch": 4.73, "learning_rate": 7.574778368391893e-06, "loss": 0.027, "step": 25766 }, { "epoch": 4.73, "learning_rate": 7.574058109656664e-06, "loss": 0.0259, "step": 25767 }, { "epoch": 4.73, "learning_rate": 7.573337864293283e-06, "loss": 0.0228, "step": 25768 }, { "epoch": 4.73, "learning_rate": 7.5726176323057185e-06, "loss": 0.0226, "step": 25769 }, { "epoch": 4.73, "learning_rate": 7.571897413697945e-06, "loss": 0.0245, "step": 25770 }, { "epoch": 4.73, "learning_rate": 7.571177208473932e-06, "loss": 0.0186, "step": 25771 }, { "epoch": 4.73, "learning_rate": 7.570457016637644e-06, "loss": 0.022, "step": 25772 }, { "epoch": 4.73, "learning_rate": 7.569736838193054e-06, "loss": 0.0196, "step": 25773 }, { "epoch": 4.73, "learning_rate": 7.569016673144132e-06, "loss": 0.0273, "step": 25774 }, { "epoch": 4.73, "learning_rate": 7.568296521494846e-06, "loss": 0.0222, "step": 25775 }, { "epoch": 4.73, "learning_rate": 7.567576383249168e-06, "loss": 0.0206, "step": 25776 }, { "epoch": 4.73, "learning_rate": 7.566856258411064e-06, "loss": 0.0219, "step": 25777 }, { "epoch": 4.73, "learning_rate": 7.566136146984506e-06, "loss": 0.0196, "step": 25778 }, { "epoch": 4.73, "learning_rate": 7.565416048973464e-06, "loss": 0.0232, "step": 25779 }, { "epoch": 4.73, "learning_rate": 7.564695964381908e-06, "loss": 0.0198, "step": 25780 }, { "epoch": 4.73, "learning_rate": 7.563975893213802e-06, "loss": 0.0255, "step": 25781 }, { "epoch": 4.73, "learning_rate": 7.563255835473115e-06, "loss": 0.0235, "step": 25782 }, { "epoch": 4.73, "learning_rate": 7.562535791163823e-06, "loss": 0.0194, "step": 25783 }, { "epoch": 4.73, "learning_rate": 7.56181576028989e-06, "loss": 0.0183, "step": 25784 }, { "epoch": 4.73, "learning_rate": 7.561095742855286e-06, "loss": 0.0186, "step": 25785 }, { "epoch": 4.73, "learning_rate": 7.5603757388639765e-06, "loss": 0.0226, "step": 25786 }, { "epoch": 4.73, "learning_rate": 7.559655748319935e-06, "loss": 0.0216, "step": 25787 }, { "epoch": 4.73, "learning_rate": 7.558935771227129e-06, "loss": 0.0247, "step": 25788 }, { "epoch": 4.73, "learning_rate": 7.558215807589528e-06, "loss": 0.023, "step": 25789 }, { "epoch": 4.73, "learning_rate": 7.557495857411097e-06, "loss": 0.0228, "step": 25790 }, { "epoch": 4.73, "learning_rate": 7.556775920695804e-06, "loss": 0.024, "step": 25791 }, { "epoch": 4.73, "learning_rate": 7.55605599744762e-06, "loss": 0.023, "step": 25792 }, { "epoch": 4.73, "learning_rate": 7.555336087670514e-06, "loss": 0.0264, "step": 25793 }, { "epoch": 4.73, "learning_rate": 7.554616191368452e-06, "loss": 0.0215, "step": 25794 }, { "epoch": 4.73, "learning_rate": 7.553896308545401e-06, "loss": 0.0195, "step": 25795 }, { "epoch": 4.73, "learning_rate": 7.553176439205333e-06, "loss": 0.0221, "step": 25796 }, { "epoch": 4.73, "learning_rate": 7.552456583352215e-06, "loss": 0.0204, "step": 25797 }, { "epoch": 4.73, "learning_rate": 7.5517367409900146e-06, "loss": 0.0228, "step": 25798 }, { "epoch": 4.73, "learning_rate": 7.551016912122692e-06, "loss": 0.0254, "step": 25799 }, { "epoch": 4.73, "learning_rate": 7.550297096754228e-06, "loss": 0.0234, "step": 25800 }, { "epoch": 4.73, "learning_rate": 7.549577294888581e-06, "loss": 0.0245, "step": 25801 }, { "epoch": 4.73, "learning_rate": 7.548857506529723e-06, "loss": 0.0234, "step": 25802 }, { "epoch": 4.73, "learning_rate": 7.54813773168162e-06, "loss": 0.0246, "step": 25803 }, { "epoch": 4.73, "learning_rate": 7.547417970348235e-06, "loss": 0.0237, "step": 25804 }, { "epoch": 4.73, "learning_rate": 7.546698222533544e-06, "loss": 0.0222, "step": 25805 }, { "epoch": 4.73, "learning_rate": 7.545978488241508e-06, "loss": 0.0224, "step": 25806 }, { "epoch": 4.73, "learning_rate": 7.545258767476099e-06, "loss": 0.0254, "step": 25807 }, { "epoch": 4.73, "learning_rate": 7.544539060241275e-06, "loss": 0.0239, "step": 25808 }, { "epoch": 4.73, "learning_rate": 7.543819366541015e-06, "loss": 0.0245, "step": 25809 }, { "epoch": 4.73, "learning_rate": 7.543099686379279e-06, "loss": 0.0159, "step": 25810 }, { "epoch": 4.73, "learning_rate": 7.542380019760035e-06, "loss": 0.0234, "step": 25811 }, { "epoch": 4.73, "learning_rate": 7.541660366687249e-06, "loss": 0.0214, "step": 25812 }, { "epoch": 4.73, "learning_rate": 7.540940727164887e-06, "loss": 0.016, "step": 25813 }, { "epoch": 4.73, "learning_rate": 7.5402211011969195e-06, "loss": 0.0201, "step": 25814 }, { "epoch": 4.73, "learning_rate": 7.539501488787311e-06, "loss": 0.0216, "step": 25815 }, { "epoch": 4.73, "learning_rate": 7.53878188994003e-06, "loss": 0.0216, "step": 25816 }, { "epoch": 4.73, "learning_rate": 7.538062304659035e-06, "loss": 0.0234, "step": 25817 }, { "epoch": 4.73, "learning_rate": 7.537342732948304e-06, "loss": 0.0231, "step": 25818 }, { "epoch": 4.73, "learning_rate": 7.536623174811796e-06, "loss": 0.0198, "step": 25819 }, { "epoch": 4.74, "learning_rate": 7.535903630253478e-06, "loss": 0.0211, "step": 25820 }, { "epoch": 4.74, "learning_rate": 7.535184099277315e-06, "loss": 0.0222, "step": 25821 }, { "epoch": 4.74, "learning_rate": 7.534464581887278e-06, "loss": 0.0237, "step": 25822 }, { "epoch": 4.74, "learning_rate": 7.533745078087328e-06, "loss": 0.0201, "step": 25823 }, { "epoch": 4.74, "learning_rate": 7.533025587881434e-06, "loss": 0.0226, "step": 25824 }, { "epoch": 4.74, "learning_rate": 7.5323061112735625e-06, "loss": 0.0214, "step": 25825 }, { "epoch": 4.74, "learning_rate": 7.5315866482676724e-06, "loss": 0.0225, "step": 25826 }, { "epoch": 4.74, "learning_rate": 7.5308671988677395e-06, "loss": 0.0214, "step": 25827 }, { "epoch": 4.74, "learning_rate": 7.5301477630777244e-06, "loss": 0.0249, "step": 25828 }, { "epoch": 4.74, "learning_rate": 7.52942834090159e-06, "loss": 0.0247, "step": 25829 }, { "epoch": 4.74, "learning_rate": 7.5287089323433035e-06, "loss": 0.0225, "step": 25830 }, { "epoch": 4.74, "learning_rate": 7.527989537406834e-06, "loss": 0.0215, "step": 25831 }, { "epoch": 4.74, "learning_rate": 7.527270156096144e-06, "loss": 0.0204, "step": 25832 }, { "epoch": 4.74, "learning_rate": 7.5265507884152e-06, "loss": 0.0201, "step": 25833 }, { "epoch": 4.74, "learning_rate": 7.525831434367965e-06, "loss": 0.0244, "step": 25834 }, { "epoch": 4.74, "learning_rate": 7.5251120939584e-06, "loss": 0.0203, "step": 25835 }, { "epoch": 4.74, "learning_rate": 7.524392767190482e-06, "loss": 0.0215, "step": 25836 }, { "epoch": 4.74, "learning_rate": 7.523673454068165e-06, "loss": 0.0211, "step": 25837 }, { "epoch": 4.74, "learning_rate": 7.522954154595421e-06, "loss": 0.0233, "step": 25838 }, { "epoch": 4.74, "learning_rate": 7.522234868776208e-06, "loss": 0.0191, "step": 25839 }, { "epoch": 4.74, "learning_rate": 7.521515596614497e-06, "loss": 0.021, "step": 25840 }, { "epoch": 4.74, "learning_rate": 7.520796338114249e-06, "loss": 0.0228, "step": 25841 }, { "epoch": 4.74, "learning_rate": 7.520077093279433e-06, "loss": 0.0206, "step": 25842 }, { "epoch": 4.74, "learning_rate": 7.5193578621140074e-06, "loss": 0.0229, "step": 25843 }, { "epoch": 4.74, "learning_rate": 7.518638644621936e-06, "loss": 0.0211, "step": 25844 }, { "epoch": 4.74, "learning_rate": 7.517919440807192e-06, "loss": 0.0237, "step": 25845 }, { "epoch": 4.74, "learning_rate": 7.517200250673733e-06, "loss": 0.0212, "step": 25846 }, { "epoch": 4.74, "learning_rate": 7.516481074225523e-06, "loss": 0.0215, "step": 25847 }, { "epoch": 4.74, "learning_rate": 7.515761911466527e-06, "loss": 0.0221, "step": 25848 }, { "epoch": 4.74, "learning_rate": 7.515042762400711e-06, "loss": 0.0203, "step": 25849 }, { "epoch": 4.74, "learning_rate": 7.514323627032036e-06, "loss": 0.0209, "step": 25850 }, { "epoch": 4.74, "learning_rate": 7.513604505364472e-06, "loss": 0.022, "step": 25851 }, { "epoch": 4.74, "learning_rate": 7.51288539740197e-06, "loss": 0.0238, "step": 25852 }, { "epoch": 4.74, "learning_rate": 7.512166303148507e-06, "loss": 0.0186, "step": 25853 }, { "epoch": 4.74, "learning_rate": 7.511447222608044e-06, "loss": 0.0238, "step": 25854 }, { "epoch": 4.74, "learning_rate": 7.510728155784539e-06, "loss": 0.0165, "step": 25855 }, { "epoch": 4.74, "learning_rate": 7.51000910268196e-06, "loss": 0.0239, "step": 25856 }, { "epoch": 4.74, "learning_rate": 7.509290063304266e-06, "loss": 0.0192, "step": 25857 }, { "epoch": 4.74, "learning_rate": 7.508571037655427e-06, "loss": 0.0206, "step": 25858 }, { "epoch": 4.74, "learning_rate": 7.507852025739402e-06, "loss": 0.0214, "step": 25859 }, { "epoch": 4.74, "learning_rate": 7.5071330275601574e-06, "loss": 0.0208, "step": 25860 }, { "epoch": 4.74, "learning_rate": 7.506414043121647e-06, "loss": 0.022, "step": 25861 }, { "epoch": 4.74, "learning_rate": 7.5056950724278456e-06, "loss": 0.0237, "step": 25862 }, { "epoch": 4.74, "learning_rate": 7.504976115482713e-06, "loss": 0.0263, "step": 25863 }, { "epoch": 4.74, "learning_rate": 7.504257172290208e-06, "loss": 0.0208, "step": 25864 }, { "epoch": 4.74, "learning_rate": 7.503538242854297e-06, "loss": 0.0239, "step": 25865 }, { "epoch": 4.74, "learning_rate": 7.502819327178938e-06, "loss": 0.0222, "step": 25866 }, { "epoch": 4.74, "learning_rate": 7.502100425268102e-06, "loss": 0.0238, "step": 25867 }, { "epoch": 4.74, "learning_rate": 7.501381537125745e-06, "loss": 0.0249, "step": 25868 }, { "epoch": 4.74, "learning_rate": 7.500662662755834e-06, "loss": 0.0234, "step": 25869 }, { "epoch": 4.74, "learning_rate": 7.499943802162324e-06, "loss": 0.0183, "step": 25870 }, { "epoch": 4.74, "learning_rate": 7.499224955349187e-06, "loss": 0.0252, "step": 25871 }, { "epoch": 4.74, "learning_rate": 7.4985061223203795e-06, "loss": 0.0242, "step": 25872 }, { "epoch": 4.74, "learning_rate": 7.497787303079864e-06, "loss": 0.0283, "step": 25873 }, { "epoch": 4.74, "learning_rate": 7.497068497631601e-06, "loss": 0.0236, "step": 25874 }, { "epoch": 4.75, "learning_rate": 7.496349705979559e-06, "loss": 0.0244, "step": 25875 }, { "epoch": 4.75, "learning_rate": 7.495630928127695e-06, "loss": 0.0197, "step": 25876 }, { "epoch": 4.75, "learning_rate": 7.4949121640799714e-06, "loss": 0.0238, "step": 25877 }, { "epoch": 4.75, "learning_rate": 7.494193413840354e-06, "loss": 0.0274, "step": 25878 }, { "epoch": 4.75, "learning_rate": 7.493474677412795e-06, "loss": 0.0208, "step": 25879 }, { "epoch": 4.75, "learning_rate": 7.4927559548012675e-06, "loss": 0.0233, "step": 25880 }, { "epoch": 4.75, "learning_rate": 7.492037246009726e-06, "loss": 0.0223, "step": 25881 }, { "epoch": 4.75, "learning_rate": 7.491318551042134e-06, "loss": 0.0214, "step": 25882 }, { "epoch": 4.75, "learning_rate": 7.490599869902451e-06, "loss": 0.0207, "step": 25883 }, { "epoch": 4.75, "learning_rate": 7.489881202594642e-06, "loss": 0.0191, "step": 25884 }, { "epoch": 4.75, "learning_rate": 7.489162549122667e-06, "loss": 0.0217, "step": 25885 }, { "epoch": 4.75, "learning_rate": 7.488443909490487e-06, "loss": 0.0252, "step": 25886 }, { "epoch": 4.75, "learning_rate": 7.487725283702066e-06, "loss": 0.021, "step": 25887 }, { "epoch": 4.75, "learning_rate": 7.4870066717613545e-06, "loss": 0.0224, "step": 25888 }, { "epoch": 4.75, "learning_rate": 7.486288073672327e-06, "loss": 0.0191, "step": 25889 }, { "epoch": 4.75, "learning_rate": 7.485569489438937e-06, "loss": 0.0214, "step": 25890 }, { "epoch": 4.75, "learning_rate": 7.4848509190651475e-06, "loss": 0.0206, "step": 25891 }, { "epoch": 4.75, "learning_rate": 7.484132362554915e-06, "loss": 0.0253, "step": 25892 }, { "epoch": 4.75, "learning_rate": 7.4834138199122075e-06, "loss": 0.0216, "step": 25893 }, { "epoch": 4.75, "learning_rate": 7.482695291140982e-06, "loss": 0.0266, "step": 25894 }, { "epoch": 4.75, "learning_rate": 7.481976776245199e-06, "loss": 0.022, "step": 25895 }, { "epoch": 4.75, "learning_rate": 7.48125827522882e-06, "loss": 0.022, "step": 25896 }, { "epoch": 4.75, "learning_rate": 7.480539788095798e-06, "loss": 0.0223, "step": 25897 }, { "epoch": 4.75, "learning_rate": 7.479821314850107e-06, "loss": 0.0182, "step": 25898 }, { "epoch": 4.75, "learning_rate": 7.479102855495698e-06, "loss": 0.0244, "step": 25899 }, { "epoch": 4.75, "learning_rate": 7.478384410036532e-06, "loss": 0.0208, "step": 25900 }, { "epoch": 4.75, "learning_rate": 7.477665978476569e-06, "loss": 0.024, "step": 25901 }, { "epoch": 4.75, "learning_rate": 7.476947560819771e-06, "loss": 0.0177, "step": 25902 }, { "epoch": 4.75, "learning_rate": 7.4762291570701e-06, "loss": 0.0221, "step": 25903 }, { "epoch": 4.75, "learning_rate": 7.475510767231511e-06, "loss": 0.0202, "step": 25904 }, { "epoch": 4.75, "learning_rate": 7.474792391307963e-06, "loss": 0.0216, "step": 25905 }, { "epoch": 4.75, "learning_rate": 7.474074029303422e-06, "loss": 0.0195, "step": 25906 }, { "epoch": 4.75, "learning_rate": 7.4733556812218455e-06, "loss": 0.0211, "step": 25907 }, { "epoch": 4.75, "learning_rate": 7.472637347067191e-06, "loss": 0.0179, "step": 25908 }, { "epoch": 4.75, "learning_rate": 7.4719190268434176e-06, "loss": 0.0232, "step": 25909 }, { "epoch": 4.75, "learning_rate": 7.471200720554484e-06, "loss": 0.0188, "step": 25910 }, { "epoch": 4.75, "learning_rate": 7.470482428204352e-06, "loss": 0.0212, "step": 25911 }, { "epoch": 4.75, "learning_rate": 7.469764149796981e-06, "loss": 0.0183, "step": 25912 }, { "epoch": 4.75, "learning_rate": 7.46904588533633e-06, "loss": 0.0206, "step": 25913 }, { "epoch": 4.75, "learning_rate": 7.468327634826354e-06, "loss": 0.0215, "step": 25914 }, { "epoch": 4.75, "learning_rate": 7.467609398271018e-06, "loss": 0.021, "step": 25915 }, { "epoch": 4.75, "learning_rate": 7.46689117567428e-06, "loss": 0.0259, "step": 25916 }, { "epoch": 4.75, "learning_rate": 7.4661729670400965e-06, "loss": 0.0189, "step": 25917 }, { "epoch": 4.75, "learning_rate": 7.465454772372426e-06, "loss": 0.0247, "step": 25918 }, { "epoch": 4.75, "learning_rate": 7.464736591675226e-06, "loss": 0.0242, "step": 25919 }, { "epoch": 4.75, "learning_rate": 7.46401842495246e-06, "loss": 0.0204, "step": 25920 }, { "epoch": 4.75, "learning_rate": 7.4633002722080824e-06, "loss": 0.0222, "step": 25921 }, { "epoch": 4.75, "learning_rate": 7.462582133446055e-06, "loss": 0.0257, "step": 25922 }, { "epoch": 4.75, "learning_rate": 7.461864008670331e-06, "loss": 0.0193, "step": 25923 }, { "epoch": 4.75, "learning_rate": 7.461145897884874e-06, "loss": 0.0226, "step": 25924 }, { "epoch": 4.75, "learning_rate": 7.460427801093644e-06, "loss": 0.0207, "step": 25925 }, { "epoch": 4.75, "learning_rate": 7.4597097183005915e-06, "loss": 0.0177, "step": 25926 }, { "epoch": 4.75, "learning_rate": 7.458991649509677e-06, "loss": 0.0256, "step": 25927 }, { "epoch": 4.75, "learning_rate": 7.458273594724863e-06, "loss": 0.0227, "step": 25928 }, { "epoch": 4.75, "learning_rate": 7.4575555539501046e-06, "loss": 0.0268, "step": 25929 }, { "epoch": 4.76, "learning_rate": 7.456837527189358e-06, "loss": 0.019, "step": 25930 }, { "epoch": 4.76, "learning_rate": 7.456119514446583e-06, "loss": 0.0185, "step": 25931 }, { "epoch": 4.76, "learning_rate": 7.455401515725736e-06, "loss": 0.0186, "step": 25932 }, { "epoch": 4.76, "learning_rate": 7.454683531030775e-06, "loss": 0.0209, "step": 25933 }, { "epoch": 4.76, "learning_rate": 7.453965560365662e-06, "loss": 0.0274, "step": 25934 }, { "epoch": 4.76, "learning_rate": 7.453247603734348e-06, "loss": 0.0246, "step": 25935 }, { "epoch": 4.76, "learning_rate": 7.452529661140791e-06, "loss": 0.024, "step": 25936 }, { "epoch": 4.76, "learning_rate": 7.451811732588952e-06, "loss": 0.0264, "step": 25937 }, { "epoch": 4.76, "learning_rate": 7.451093818082787e-06, "loss": 0.02, "step": 25938 }, { "epoch": 4.76, "learning_rate": 7.450375917626253e-06, "loss": 0.0192, "step": 25939 }, { "epoch": 4.76, "learning_rate": 7.449658031223308e-06, "loss": 0.0275, "step": 25940 }, { "epoch": 4.76, "learning_rate": 7.448940158877901e-06, "loss": 0.0206, "step": 25941 }, { "epoch": 4.76, "learning_rate": 7.448222300594001e-06, "loss": 0.0187, "step": 25942 }, { "epoch": 4.76, "learning_rate": 7.44750445637556e-06, "loss": 0.0228, "step": 25943 }, { "epoch": 4.76, "learning_rate": 7.446786626226532e-06, "loss": 0.0268, "step": 25944 }, { "epoch": 4.76, "learning_rate": 7.446068810150875e-06, "loss": 0.0256, "step": 25945 }, { "epoch": 4.76, "learning_rate": 7.445351008152548e-06, "loss": 0.0206, "step": 25946 }, { "epoch": 4.76, "learning_rate": 7.444633220235506e-06, "loss": 0.0218, "step": 25947 }, { "epoch": 4.76, "learning_rate": 7.443915446403706e-06, "loss": 0.0205, "step": 25948 }, { "epoch": 4.76, "learning_rate": 7.443197686661104e-06, "loss": 0.0263, "step": 25949 }, { "epoch": 4.76, "learning_rate": 7.4424799410116555e-06, "loss": 0.0203, "step": 25950 }, { "epoch": 4.76, "learning_rate": 7.441762209459318e-06, "loss": 0.0183, "step": 25951 }, { "epoch": 4.76, "learning_rate": 7.441044492008049e-06, "loss": 0.0234, "step": 25952 }, { "epoch": 4.76, "learning_rate": 7.440326788661801e-06, "loss": 0.0161, "step": 25953 }, { "epoch": 4.76, "learning_rate": 7.4396090994245295e-06, "loss": 0.0186, "step": 25954 }, { "epoch": 4.76, "learning_rate": 7.438891424300194e-06, "loss": 0.0235, "step": 25955 }, { "epoch": 4.76, "learning_rate": 7.438173763292751e-06, "loss": 0.0234, "step": 25956 }, { "epoch": 4.76, "learning_rate": 7.437456116406152e-06, "loss": 0.0228, "step": 25957 }, { "epoch": 4.76, "learning_rate": 7.436738483644354e-06, "loss": 0.0217, "step": 25958 }, { "epoch": 4.76, "learning_rate": 7.436020865011317e-06, "loss": 0.0255, "step": 25959 }, { "epoch": 4.76, "learning_rate": 7.435303260510991e-06, "loss": 0.0229, "step": 25960 }, { "epoch": 4.76, "learning_rate": 7.434585670147336e-06, "loss": 0.0183, "step": 25961 }, { "epoch": 4.76, "learning_rate": 7.433868093924304e-06, "loss": 0.0204, "step": 25962 }, { "epoch": 4.76, "learning_rate": 7.433150531845848e-06, "loss": 0.0235, "step": 25963 }, { "epoch": 4.76, "learning_rate": 7.432432983915927e-06, "loss": 0.0206, "step": 25964 }, { "epoch": 4.76, "learning_rate": 7.431715450138497e-06, "loss": 0.0201, "step": 25965 }, { "epoch": 4.76, "learning_rate": 7.430997930517512e-06, "loss": 0.029, "step": 25966 }, { "epoch": 4.76, "learning_rate": 7.4302804250569234e-06, "loss": 0.0293, "step": 25967 }, { "epoch": 4.76, "learning_rate": 7.4295629337606924e-06, "loss": 0.0193, "step": 25968 }, { "epoch": 4.76, "learning_rate": 7.428845456632771e-06, "loss": 0.0184, "step": 25969 }, { "epoch": 4.76, "learning_rate": 7.428127993677116e-06, "loss": 0.0223, "step": 25970 }, { "epoch": 4.76, "learning_rate": 7.4274105448976766e-06, "loss": 0.029, "step": 25971 }, { "epoch": 4.76, "learning_rate": 7.426693110298408e-06, "loss": 0.0218, "step": 25972 }, { "epoch": 4.76, "learning_rate": 7.4259756898832715e-06, "loss": 0.0249, "step": 25973 }, { "epoch": 4.76, "learning_rate": 7.425258283656216e-06, "loss": 0.0185, "step": 25974 }, { "epoch": 4.76, "learning_rate": 7.424540891621197e-06, "loss": 0.0232, "step": 25975 }, { "epoch": 4.76, "learning_rate": 7.423823513782167e-06, "loss": 0.0198, "step": 25976 }, { "epoch": 4.76, "learning_rate": 7.423106150143083e-06, "loss": 0.0238, "step": 25977 }, { "epoch": 4.76, "learning_rate": 7.4223888007079005e-06, "loss": 0.0249, "step": 25978 }, { "epoch": 4.76, "learning_rate": 7.421671465480571e-06, "loss": 0.0173, "step": 25979 }, { "epoch": 4.76, "learning_rate": 7.420954144465044e-06, "loss": 0.0245, "step": 25980 }, { "epoch": 4.76, "learning_rate": 7.4202368376652845e-06, "loss": 0.0214, "step": 25981 }, { "epoch": 4.76, "learning_rate": 7.4195195450852385e-06, "loss": 0.0185, "step": 25982 }, { "epoch": 4.76, "learning_rate": 7.4188022667288595e-06, "loss": 0.0241, "step": 25983 }, { "epoch": 4.77, "learning_rate": 7.418085002600104e-06, "loss": 0.0235, "step": 25984 }, { "epoch": 4.77, "learning_rate": 7.417367752702922e-06, "loss": 0.0247, "step": 25985 }, { "epoch": 4.77, "learning_rate": 7.416650517041273e-06, "loss": 0.0242, "step": 25986 }, { "epoch": 4.77, "learning_rate": 7.415933295619106e-06, "loss": 0.0201, "step": 25987 }, { "epoch": 4.77, "learning_rate": 7.4152160884403755e-06, "loss": 0.0175, "step": 25988 }, { "epoch": 4.77, "learning_rate": 7.41449889550903e-06, "loss": 0.0233, "step": 25989 }, { "epoch": 4.77, "learning_rate": 7.413781716829034e-06, "loss": 0.0197, "step": 25990 }, { "epoch": 4.77, "learning_rate": 7.41306455240433e-06, "loss": 0.0175, "step": 25991 }, { "epoch": 4.77, "learning_rate": 7.412347402238875e-06, "loss": 0.0207, "step": 25992 }, { "epoch": 4.77, "learning_rate": 7.4116302663366225e-06, "loss": 0.0235, "step": 25993 }, { "epoch": 4.77, "learning_rate": 7.410913144701521e-06, "loss": 0.022, "step": 25994 }, { "epoch": 4.77, "learning_rate": 7.410196037337531e-06, "loss": 0.0229, "step": 25995 }, { "epoch": 4.77, "learning_rate": 7.409478944248602e-06, "loss": 0.0202, "step": 25996 }, { "epoch": 4.77, "learning_rate": 7.408761865438685e-06, "loss": 0.0251, "step": 25997 }, { "epoch": 4.77, "learning_rate": 7.408044800911729e-06, "loss": 0.0263, "step": 25998 }, { "epoch": 4.77, "learning_rate": 7.407327750671696e-06, "loss": 0.0179, "step": 25999 }, { "epoch": 4.77, "learning_rate": 7.406610714722531e-06, "loss": 0.0179, "step": 26000 }, { "epoch": 4.77, "learning_rate": 7.405893693068189e-06, "loss": 0.018, "step": 26001 }, { "epoch": 4.77, "learning_rate": 7.40517668571262e-06, "loss": 0.0203, "step": 26002 }, { "epoch": 4.77, "learning_rate": 7.404459692659779e-06, "loss": 0.0247, "step": 26003 }, { "epoch": 4.77, "learning_rate": 7.4037427139136176e-06, "loss": 0.0251, "step": 26004 }, { "epoch": 4.77, "learning_rate": 7.40302574947809e-06, "loss": 0.0214, "step": 26005 }, { "epoch": 4.77, "learning_rate": 7.402308799357141e-06, "loss": 0.0189, "step": 26006 }, { "epoch": 4.77, "learning_rate": 7.401591863554725e-06, "loss": 0.0217, "step": 26007 }, { "epoch": 4.77, "learning_rate": 7.4008749420748e-06, "loss": 0.0186, "step": 26008 }, { "epoch": 4.77, "learning_rate": 7.400158034921312e-06, "loss": 0.0259, "step": 26009 }, { "epoch": 4.77, "learning_rate": 7.399441142098214e-06, "loss": 0.0222, "step": 26010 }, { "epoch": 4.77, "learning_rate": 7.398724263609455e-06, "loss": 0.0231, "step": 26011 }, { "epoch": 4.77, "learning_rate": 7.398007399458991e-06, "loss": 0.0245, "step": 26012 }, { "epoch": 4.77, "learning_rate": 7.397290549650771e-06, "loss": 0.0215, "step": 26013 }, { "epoch": 4.77, "learning_rate": 7.396573714188748e-06, "loss": 0.0192, "step": 26014 }, { "epoch": 4.77, "learning_rate": 7.395856893076871e-06, "loss": 0.0224, "step": 26015 }, { "epoch": 4.77, "learning_rate": 7.395140086319086e-06, "loss": 0.0195, "step": 26016 }, { "epoch": 4.77, "learning_rate": 7.394423293919357e-06, "loss": 0.0261, "step": 26017 }, { "epoch": 4.77, "learning_rate": 7.393706515881626e-06, "loss": 0.0232, "step": 26018 }, { "epoch": 4.77, "learning_rate": 7.392989752209845e-06, "loss": 0.0198, "step": 26019 }, { "epoch": 4.77, "learning_rate": 7.392273002907964e-06, "loss": 0.0226, "step": 26020 }, { "epoch": 4.77, "learning_rate": 7.391556267979938e-06, "loss": 0.0207, "step": 26021 }, { "epoch": 4.77, "learning_rate": 7.390839547429713e-06, "loss": 0.0218, "step": 26022 }, { "epoch": 4.77, "learning_rate": 7.390122841261246e-06, "loss": 0.0241, "step": 26023 }, { "epoch": 4.77, "learning_rate": 7.38940614947848e-06, "loss": 0.0228, "step": 26024 }, { "epoch": 4.77, "learning_rate": 7.388689472085364e-06, "loss": 0.025, "step": 26025 }, { "epoch": 4.77, "learning_rate": 7.38797280908586e-06, "loss": 0.0248, "step": 26026 }, { "epoch": 4.77, "learning_rate": 7.387256160483906e-06, "loss": 0.0271, "step": 26027 }, { "epoch": 4.77, "learning_rate": 7.38653952628346e-06, "loss": 0.0273, "step": 26028 }, { "epoch": 4.77, "learning_rate": 7.3858229064884665e-06, "loss": 0.0214, "step": 26029 }, { "epoch": 4.77, "learning_rate": 7.385106301102882e-06, "loss": 0.0231, "step": 26030 }, { "epoch": 4.77, "learning_rate": 7.384389710130651e-06, "loss": 0.0272, "step": 26031 }, { "epoch": 4.77, "learning_rate": 7.383673133575728e-06, "loss": 0.0197, "step": 26032 }, { "epoch": 4.77, "learning_rate": 7.3829565714420525e-06, "loss": 0.0187, "step": 26033 }, { "epoch": 4.77, "learning_rate": 7.38224002373359e-06, "loss": 0.022, "step": 26034 }, { "epoch": 4.77, "learning_rate": 7.381523490454279e-06, "loss": 0.0201, "step": 26035 }, { "epoch": 4.77, "learning_rate": 7.380806971608071e-06, "loss": 0.023, "step": 26036 }, { "epoch": 4.77, "learning_rate": 7.380090467198917e-06, "loss": 0.0241, "step": 26037 }, { "epoch": 4.77, "learning_rate": 7.379373977230764e-06, "loss": 0.0219, "step": 26038 }, { "epoch": 4.78, "learning_rate": 7.378657501707565e-06, "loss": 0.0213, "step": 26039 }, { "epoch": 4.78, "learning_rate": 7.3779410406332655e-06, "loss": 0.0222, "step": 26040 }, { "epoch": 4.78, "learning_rate": 7.377224594011821e-06, "loss": 0.0221, "step": 26041 }, { "epoch": 4.78, "learning_rate": 7.3765081618471675e-06, "loss": 0.0201, "step": 26042 }, { "epoch": 4.78, "learning_rate": 7.37579174414327e-06, "loss": 0.0243, "step": 26043 }, { "epoch": 4.78, "learning_rate": 7.375075340904068e-06, "loss": 0.0221, "step": 26044 }, { "epoch": 4.78, "learning_rate": 7.374358952133511e-06, "loss": 0.02, "step": 26045 }, { "epoch": 4.78, "learning_rate": 7.373642577835551e-06, "loss": 0.0182, "step": 26046 }, { "epoch": 4.78, "learning_rate": 7.372926218014131e-06, "loss": 0.0192, "step": 26047 }, { "epoch": 4.78, "learning_rate": 7.372209872673206e-06, "loss": 0.0193, "step": 26048 }, { "epoch": 4.78, "learning_rate": 7.371493541816721e-06, "loss": 0.0213, "step": 26049 }, { "epoch": 4.78, "learning_rate": 7.370777225448629e-06, "loss": 0.0221, "step": 26050 }, { "epoch": 4.78, "learning_rate": 7.370060923572866e-06, "loss": 0.0231, "step": 26051 }, { "epoch": 4.78, "learning_rate": 7.369344636193396e-06, "loss": 0.0219, "step": 26052 }, { "epoch": 4.78, "learning_rate": 7.368628363314158e-06, "loss": 0.0245, "step": 26053 }, { "epoch": 4.78, "learning_rate": 7.367912104939102e-06, "loss": 0.0194, "step": 26054 }, { "epoch": 4.78, "learning_rate": 7.367195861072174e-06, "loss": 0.0217, "step": 26055 }, { "epoch": 4.78, "learning_rate": 7.366479631717327e-06, "loss": 0.0194, "step": 26056 }, { "epoch": 4.78, "learning_rate": 7.3657634168785054e-06, "loss": 0.0179, "step": 26057 }, { "epoch": 4.78, "learning_rate": 7.3650472165596575e-06, "loss": 0.0194, "step": 26058 }, { "epoch": 4.78, "learning_rate": 7.364331030764733e-06, "loss": 0.0209, "step": 26059 }, { "epoch": 4.78, "learning_rate": 7.363614859497671e-06, "loss": 0.0212, "step": 26060 }, { "epoch": 4.78, "learning_rate": 7.362898702762433e-06, "loss": 0.0205, "step": 26061 }, { "epoch": 4.78, "learning_rate": 7.362182560562956e-06, "loss": 0.0221, "step": 26062 }, { "epoch": 4.78, "learning_rate": 7.36146643290319e-06, "loss": 0.0232, "step": 26063 }, { "epoch": 4.78, "learning_rate": 7.360750319787082e-06, "loss": 0.0207, "step": 26064 }, { "epoch": 4.78, "learning_rate": 7.360034221218581e-06, "loss": 0.0227, "step": 26065 }, { "epoch": 4.78, "learning_rate": 7.359318137201635e-06, "loss": 0.0224, "step": 26066 }, { "epoch": 4.78, "learning_rate": 7.358602067740188e-06, "loss": 0.0239, "step": 26067 }, { "epoch": 4.78, "learning_rate": 7.357886012838192e-06, "loss": 0.0225, "step": 26068 }, { "epoch": 4.78, "learning_rate": 7.357169972499583e-06, "loss": 0.0222, "step": 26069 }, { "epoch": 4.78, "learning_rate": 7.356453946728321e-06, "loss": 0.0196, "step": 26070 }, { "epoch": 4.78, "learning_rate": 7.355737935528344e-06, "loss": 0.0234, "step": 26071 }, { "epoch": 4.78, "learning_rate": 7.355021938903602e-06, "loss": 0.0249, "step": 26072 }, { "epoch": 4.78, "learning_rate": 7.3543059568580385e-06, "loss": 0.0215, "step": 26073 }, { "epoch": 4.78, "learning_rate": 7.353589989395605e-06, "loss": 0.0244, "step": 26074 }, { "epoch": 4.78, "learning_rate": 7.352874036520246e-06, "loss": 0.0175, "step": 26075 }, { "epoch": 4.78, "learning_rate": 7.352158098235906e-06, "loss": 0.0234, "step": 26076 }, { "epoch": 4.78, "learning_rate": 7.351442174546532e-06, "loss": 0.0226, "step": 26077 }, { "epoch": 4.78, "learning_rate": 7.350726265456073e-06, "loss": 0.0219, "step": 26078 }, { "epoch": 4.78, "learning_rate": 7.350010370968475e-06, "loss": 0.0224, "step": 26079 }, { "epoch": 4.78, "learning_rate": 7.349294491087679e-06, "loss": 0.0241, "step": 26080 }, { "epoch": 4.78, "learning_rate": 7.3485786258176325e-06, "loss": 0.0233, "step": 26081 }, { "epoch": 4.78, "learning_rate": 7.347862775162284e-06, "loss": 0.0237, "step": 26082 }, { "epoch": 4.78, "learning_rate": 7.347146939125578e-06, "loss": 0.0216, "step": 26083 }, { "epoch": 4.78, "learning_rate": 7.34643111771146e-06, "loss": 0.0252, "step": 26084 }, { "epoch": 4.78, "learning_rate": 7.3457153109238754e-06, "loss": 0.024, "step": 26085 }, { "epoch": 4.78, "learning_rate": 7.34499951876677e-06, "loss": 0.0207, "step": 26086 }, { "epoch": 4.78, "learning_rate": 7.344283741244091e-06, "loss": 0.0262, "step": 26087 }, { "epoch": 4.78, "learning_rate": 7.343567978359783e-06, "loss": 0.0205, "step": 26088 }, { "epoch": 4.78, "learning_rate": 7.3428522301177894e-06, "loss": 0.0211, "step": 26089 }, { "epoch": 4.78, "learning_rate": 7.342136496522057e-06, "loss": 0.0216, "step": 26090 }, { "epoch": 4.78, "learning_rate": 7.341420777576528e-06, "loss": 0.0247, "step": 26091 }, { "epoch": 4.78, "learning_rate": 7.340705073285151e-06, "loss": 0.0246, "step": 26092 }, { "epoch": 4.79, "learning_rate": 7.339989383651871e-06, "loss": 0.0197, "step": 26093 }, { "epoch": 4.79, "learning_rate": 7.339273708680634e-06, "loss": 0.0217, "step": 26094 }, { "epoch": 4.79, "learning_rate": 7.338558048375377e-06, "loss": 0.023, "step": 26095 }, { "epoch": 4.79, "learning_rate": 7.337842402740054e-06, "loss": 0.0196, "step": 26096 }, { "epoch": 4.79, "learning_rate": 7.3371267717786065e-06, "loss": 0.0235, "step": 26097 }, { "epoch": 4.79, "learning_rate": 7.336411155494978e-06, "loss": 0.0197, "step": 26098 }, { "epoch": 4.79, "learning_rate": 7.335695553893114e-06, "loss": 0.0217, "step": 26099 }, { "epoch": 4.79, "learning_rate": 7.334979966976956e-06, "loss": 0.0225, "step": 26100 }, { "epoch": 4.79, "learning_rate": 7.334264394750453e-06, "loss": 0.0185, "step": 26101 }, { "epoch": 4.79, "learning_rate": 7.333548837217548e-06, "loss": 0.0224, "step": 26102 }, { "epoch": 4.79, "learning_rate": 7.332833294382185e-06, "loss": 0.0209, "step": 26103 }, { "epoch": 4.79, "learning_rate": 7.332117766248301e-06, "loss": 0.0255, "step": 26104 }, { "epoch": 4.79, "learning_rate": 7.33140225281985e-06, "loss": 0.0194, "step": 26105 }, { "epoch": 4.79, "learning_rate": 7.3306867541007766e-06, "loss": 0.0152, "step": 26106 }, { "epoch": 4.79, "learning_rate": 7.329971270095018e-06, "loss": 0.0228, "step": 26107 }, { "epoch": 4.79, "learning_rate": 7.329255800806516e-06, "loss": 0.0185, "step": 26108 }, { "epoch": 4.79, "learning_rate": 7.328540346239222e-06, "loss": 0.0183, "step": 26109 }, { "epoch": 4.79, "learning_rate": 7.327824906397076e-06, "loss": 0.0182, "step": 26110 }, { "epoch": 4.79, "learning_rate": 7.3271094812840225e-06, "loss": 0.0212, "step": 26111 }, { "epoch": 4.79, "learning_rate": 7.326394070904006e-06, "loss": 0.0235, "step": 26112 }, { "epoch": 4.79, "learning_rate": 7.325678675260961e-06, "loss": 0.022, "step": 26113 }, { "epoch": 4.79, "learning_rate": 7.324963294358842e-06, "loss": 0.0173, "step": 26114 }, { "epoch": 4.79, "learning_rate": 7.324247928201591e-06, "loss": 0.0212, "step": 26115 }, { "epoch": 4.79, "learning_rate": 7.323532576793145e-06, "loss": 0.0204, "step": 26116 }, { "epoch": 4.79, "learning_rate": 7.32281724013745e-06, "loss": 0.0254, "step": 26117 }, { "epoch": 4.79, "learning_rate": 7.322101918238449e-06, "loss": 0.0249, "step": 26118 }, { "epoch": 4.79, "learning_rate": 7.321386611100085e-06, "loss": 0.0191, "step": 26119 }, { "epoch": 4.79, "learning_rate": 7.320671318726302e-06, "loss": 0.0206, "step": 26120 }, { "epoch": 4.79, "learning_rate": 7.319956041121044e-06, "loss": 0.0238, "step": 26121 }, { "epoch": 4.79, "learning_rate": 7.319240778288243e-06, "loss": 0.0263, "step": 26122 }, { "epoch": 4.79, "learning_rate": 7.3185255302318545e-06, "loss": 0.0215, "step": 26123 }, { "epoch": 4.79, "learning_rate": 7.317810296955819e-06, "loss": 0.0221, "step": 26124 }, { "epoch": 4.79, "learning_rate": 7.317095078464072e-06, "loss": 0.0219, "step": 26125 }, { "epoch": 4.79, "learning_rate": 7.31637987476056e-06, "loss": 0.0198, "step": 26126 }, { "epoch": 4.79, "learning_rate": 7.315664685849225e-06, "loss": 0.0215, "step": 26127 }, { "epoch": 4.79, "learning_rate": 7.3149495117340106e-06, "loss": 0.0225, "step": 26128 }, { "epoch": 4.79, "learning_rate": 7.314234352418856e-06, "loss": 0.0197, "step": 26129 }, { "epoch": 4.79, "learning_rate": 7.3135192079077045e-06, "loss": 0.0227, "step": 26130 }, { "epoch": 4.79, "learning_rate": 7.312804078204499e-06, "loss": 0.0198, "step": 26131 }, { "epoch": 4.79, "learning_rate": 7.31208896331318e-06, "loss": 0.0201, "step": 26132 }, { "epoch": 4.79, "learning_rate": 7.311373863237692e-06, "loss": 0.0185, "step": 26133 }, { "epoch": 4.79, "learning_rate": 7.310658777981973e-06, "loss": 0.024, "step": 26134 }, { "epoch": 4.79, "learning_rate": 7.309943707549963e-06, "loss": 0.0218, "step": 26135 }, { "epoch": 4.79, "learning_rate": 7.309228651945608e-06, "loss": 0.0171, "step": 26136 }, { "epoch": 4.79, "learning_rate": 7.308513611172848e-06, "loss": 0.0246, "step": 26137 }, { "epoch": 4.79, "learning_rate": 7.307798585235623e-06, "loss": 0.0208, "step": 26138 }, { "epoch": 4.79, "learning_rate": 7.307083574137876e-06, "loss": 0.0183, "step": 26139 }, { "epoch": 4.79, "learning_rate": 7.306368577883547e-06, "loss": 0.0218, "step": 26140 }, { "epoch": 4.79, "learning_rate": 7.3056535964765806e-06, "loss": 0.0208, "step": 26141 }, { "epoch": 4.79, "learning_rate": 7.304938629920912e-06, "loss": 0.023, "step": 26142 }, { "epoch": 4.79, "learning_rate": 7.304223678220485e-06, "loss": 0.0224, "step": 26143 }, { "epoch": 4.79, "learning_rate": 7.303508741379239e-06, "loss": 0.0206, "step": 26144 }, { "epoch": 4.79, "learning_rate": 7.302793819401117e-06, "loss": 0.0229, "step": 26145 }, { "epoch": 4.79, "learning_rate": 7.30207891229006e-06, "loss": 0.0237, "step": 26146 }, { "epoch": 4.79, "learning_rate": 7.301364020050006e-06, "loss": 0.0164, "step": 26147 }, { "epoch": 4.8, "learning_rate": 7.3006491426848945e-06, "loss": 0.0153, "step": 26148 }, { "epoch": 4.8, "learning_rate": 7.299934280198672e-06, "loss": 0.0203, "step": 26149 }, { "epoch": 4.8, "learning_rate": 7.299219432595276e-06, "loss": 0.0159, "step": 26150 }, { "epoch": 4.8, "learning_rate": 7.298504599878643e-06, "loss": 0.0182, "step": 26151 }, { "epoch": 4.8, "learning_rate": 7.297789782052716e-06, "loss": 0.0274, "step": 26152 }, { "epoch": 4.8, "learning_rate": 7.297074979121435e-06, "loss": 0.0174, "step": 26153 }, { "epoch": 4.8, "learning_rate": 7.2963601910887405e-06, "loss": 0.0172, "step": 26154 }, { "epoch": 4.8, "learning_rate": 7.295645417958571e-06, "loss": 0.0276, "step": 26155 }, { "epoch": 4.8, "learning_rate": 7.2949306597348694e-06, "loss": 0.0234, "step": 26156 }, { "epoch": 4.8, "learning_rate": 7.29421591642157e-06, "loss": 0.0261, "step": 26157 }, { "epoch": 4.8, "learning_rate": 7.293501188022618e-06, "loss": 0.0229, "step": 26158 }, { "epoch": 4.8, "learning_rate": 7.2927864745419526e-06, "loss": 0.0225, "step": 26159 }, { "epoch": 4.8, "learning_rate": 7.29207177598351e-06, "loss": 0.0209, "step": 26160 }, { "epoch": 4.8, "learning_rate": 7.291357092351228e-06, "loss": 0.0231, "step": 26161 }, { "epoch": 4.8, "learning_rate": 7.290642423649054e-06, "loss": 0.0212, "step": 26162 }, { "epoch": 4.8, "learning_rate": 7.28992776988092e-06, "loss": 0.0234, "step": 26163 }, { "epoch": 4.8, "learning_rate": 7.289213131050769e-06, "loss": 0.0154, "step": 26164 }, { "epoch": 4.8, "learning_rate": 7.288498507162538e-06, "loss": 0.02, "step": 26165 }, { "epoch": 4.8, "learning_rate": 7.287783898220164e-06, "loss": 0.0229, "step": 26166 }, { "epoch": 4.8, "learning_rate": 7.287069304227592e-06, "loss": 0.0188, "step": 26167 }, { "epoch": 4.8, "learning_rate": 7.28635472518876e-06, "loss": 0.0211, "step": 26168 }, { "epoch": 4.8, "learning_rate": 7.285640161107601e-06, "loss": 0.0304, "step": 26169 }, { "epoch": 4.8, "learning_rate": 7.284925611988054e-06, "loss": 0.0279, "step": 26170 }, { "epoch": 4.8, "learning_rate": 7.284211077834066e-06, "loss": 0.0298, "step": 26171 }, { "epoch": 4.8, "learning_rate": 7.283496558649568e-06, "loss": 0.0247, "step": 26172 }, { "epoch": 4.8, "learning_rate": 7.282782054438501e-06, "loss": 0.0194, "step": 26173 }, { "epoch": 4.8, "learning_rate": 7.2820675652048025e-06, "loss": 0.024, "step": 26174 }, { "epoch": 4.8, "learning_rate": 7.28135309095241e-06, "loss": 0.0245, "step": 26175 }, { "epoch": 4.8, "learning_rate": 7.280638631685264e-06, "loss": 0.0261, "step": 26176 }, { "epoch": 4.8, "learning_rate": 7.279924187407305e-06, "loss": 0.0163, "step": 26177 }, { "epoch": 4.8, "learning_rate": 7.2792097581224654e-06, "loss": 0.0184, "step": 26178 }, { "epoch": 4.8, "learning_rate": 7.278495343834679e-06, "loss": 0.0228, "step": 26179 }, { "epoch": 4.8, "learning_rate": 7.2777809445478985e-06, "loss": 0.022, "step": 26180 }, { "epoch": 4.8, "learning_rate": 7.2770665602660505e-06, "loss": 0.0259, "step": 26181 }, { "epoch": 4.8, "learning_rate": 7.276352190993076e-06, "loss": 0.0225, "step": 26182 }, { "epoch": 4.8, "learning_rate": 7.2756378367329095e-06, "loss": 0.0201, "step": 26183 }, { "epoch": 4.8, "learning_rate": 7.274923497489492e-06, "loss": 0.0228, "step": 26184 }, { "epoch": 4.8, "learning_rate": 7.274209173266762e-06, "loss": 0.0197, "step": 26185 }, { "epoch": 4.8, "learning_rate": 7.273494864068655e-06, "loss": 0.0218, "step": 26186 }, { "epoch": 4.8, "learning_rate": 7.272780569899107e-06, "loss": 0.0185, "step": 26187 }, { "epoch": 4.8, "learning_rate": 7.272066290762054e-06, "loss": 0.0141, "step": 26188 }, { "epoch": 4.8, "learning_rate": 7.271352026661439e-06, "loss": 0.0238, "step": 26189 }, { "epoch": 4.8, "learning_rate": 7.270637777601195e-06, "loss": 0.022, "step": 26190 }, { "epoch": 4.8, "learning_rate": 7.2699235435852574e-06, "loss": 0.0243, "step": 26191 }, { "epoch": 4.8, "learning_rate": 7.269209324617565e-06, "loss": 0.0231, "step": 26192 }, { "epoch": 4.8, "learning_rate": 7.268495120702057e-06, "loss": 0.0218, "step": 26193 }, { "epoch": 4.8, "learning_rate": 7.267780931842667e-06, "loss": 0.0206, "step": 26194 }, { "epoch": 4.8, "learning_rate": 7.267066758043335e-06, "loss": 0.0249, "step": 26195 }, { "epoch": 4.8, "learning_rate": 7.266352599307994e-06, "loss": 0.0201, "step": 26196 }, { "epoch": 4.8, "learning_rate": 7.265638455640578e-06, "loss": 0.0245, "step": 26197 }, { "epoch": 4.8, "learning_rate": 7.2649243270450295e-06, "loss": 0.0219, "step": 26198 }, { "epoch": 4.8, "learning_rate": 7.264210213525282e-06, "loss": 0.0228, "step": 26199 }, { "epoch": 4.8, "learning_rate": 7.263496115085274e-06, "loss": 0.0208, "step": 26200 }, { "epoch": 4.8, "learning_rate": 7.262782031728934e-06, "loss": 0.0165, "step": 26201 }, { "epoch": 4.81, "learning_rate": 7.262067963460208e-06, "loss": 0.0244, "step": 26202 }, { "epoch": 4.81, "learning_rate": 7.261353910283026e-06, "loss": 0.0246, "step": 26203 }, { "epoch": 4.81, "learning_rate": 7.260639872201328e-06, "loss": 0.0198, "step": 26204 }, { "epoch": 4.81, "learning_rate": 7.259925849219041e-06, "loss": 0.0202, "step": 26205 }, { "epoch": 4.81, "learning_rate": 7.259211841340113e-06, "loss": 0.0277, "step": 26206 }, { "epoch": 4.81, "learning_rate": 7.258497848568473e-06, "loss": 0.0194, "step": 26207 }, { "epoch": 4.81, "learning_rate": 7.257783870908055e-06, "loss": 0.0227, "step": 26208 }, { "epoch": 4.81, "learning_rate": 7.257069908362797e-06, "loss": 0.0231, "step": 26209 }, { "epoch": 4.81, "learning_rate": 7.256355960936633e-06, "loss": 0.0235, "step": 26210 }, { "epoch": 4.81, "learning_rate": 7.2556420286335e-06, "loss": 0.0228, "step": 26211 }, { "epoch": 4.81, "learning_rate": 7.254928111457333e-06, "loss": 0.0229, "step": 26212 }, { "epoch": 4.81, "learning_rate": 7.254214209412069e-06, "loss": 0.0222, "step": 26213 }, { "epoch": 4.81, "learning_rate": 7.253500322501634e-06, "loss": 0.0238, "step": 26214 }, { "epoch": 4.81, "learning_rate": 7.252786450729976e-06, "loss": 0.0274, "step": 26215 }, { "epoch": 4.81, "learning_rate": 7.2520725941010204e-06, "loss": 0.0243, "step": 26216 }, { "epoch": 4.81, "learning_rate": 7.251358752618707e-06, "loss": 0.0192, "step": 26217 }, { "epoch": 4.81, "learning_rate": 7.250644926286967e-06, "loss": 0.0242, "step": 26218 }, { "epoch": 4.81, "learning_rate": 7.249931115109734e-06, "loss": 0.0296, "step": 26219 }, { "epoch": 4.81, "learning_rate": 7.249217319090949e-06, "loss": 0.0227, "step": 26220 }, { "epoch": 4.81, "learning_rate": 7.248503538234542e-06, "loss": 0.0228, "step": 26221 }, { "epoch": 4.81, "learning_rate": 7.247789772544449e-06, "loss": 0.0221, "step": 26222 }, { "epoch": 4.81, "learning_rate": 7.247076022024597e-06, "loss": 0.0255, "step": 26223 }, { "epoch": 4.81, "learning_rate": 7.246362286678934e-06, "loss": 0.0236, "step": 26224 }, { "epoch": 4.81, "learning_rate": 7.245648566511383e-06, "loss": 0.0226, "step": 26225 }, { "epoch": 4.81, "learning_rate": 7.244934861525883e-06, "loss": 0.0228, "step": 26226 }, { "epoch": 4.81, "learning_rate": 7.244221171726367e-06, "loss": 0.0213, "step": 26227 }, { "epoch": 4.81, "learning_rate": 7.243507497116764e-06, "loss": 0.0238, "step": 26228 }, { "epoch": 4.81, "learning_rate": 7.242793837701016e-06, "loss": 0.0191, "step": 26229 }, { "epoch": 4.81, "learning_rate": 7.242080193483052e-06, "loss": 0.0196, "step": 26230 }, { "epoch": 4.81, "learning_rate": 7.241366564466809e-06, "loss": 0.0206, "step": 26231 }, { "epoch": 4.81, "learning_rate": 7.240652950656213e-06, "loss": 0.0204, "step": 26232 }, { "epoch": 4.81, "learning_rate": 7.239939352055208e-06, "loss": 0.0236, "step": 26233 }, { "epoch": 4.81, "learning_rate": 7.239225768667721e-06, "loss": 0.0207, "step": 26234 }, { "epoch": 4.81, "learning_rate": 7.238512200497685e-06, "loss": 0.0211, "step": 26235 }, { "epoch": 4.81, "learning_rate": 7.237798647549033e-06, "loss": 0.0207, "step": 26236 }, { "epoch": 4.81, "learning_rate": 7.237085109825701e-06, "loss": 0.0139, "step": 26237 }, { "epoch": 4.81, "learning_rate": 7.236371587331621e-06, "loss": 0.0198, "step": 26238 }, { "epoch": 4.81, "learning_rate": 7.235658080070724e-06, "loss": 0.0231, "step": 26239 }, { "epoch": 4.81, "learning_rate": 7.234944588046948e-06, "loss": 0.0238, "step": 26240 }, { "epoch": 4.81, "learning_rate": 7.234231111264216e-06, "loss": 0.02, "step": 26241 }, { "epoch": 4.81, "learning_rate": 7.233517649726473e-06, "loss": 0.0198, "step": 26242 }, { "epoch": 4.81, "learning_rate": 7.232804203437645e-06, "loss": 0.0201, "step": 26243 }, { "epoch": 4.81, "learning_rate": 7.232090772401663e-06, "loss": 0.0189, "step": 26244 }, { "epoch": 4.81, "learning_rate": 7.2313773566224595e-06, "loss": 0.0196, "step": 26245 }, { "epoch": 4.81, "learning_rate": 7.230663956103972e-06, "loss": 0.0227, "step": 26246 }, { "epoch": 4.81, "learning_rate": 7.229950570850128e-06, "loss": 0.021, "step": 26247 }, { "epoch": 4.81, "learning_rate": 7.229237200864863e-06, "loss": 0.025, "step": 26248 }, { "epoch": 4.81, "learning_rate": 7.228523846152109e-06, "loss": 0.0192, "step": 26249 }, { "epoch": 4.81, "learning_rate": 7.22781050671579e-06, "loss": 0.019, "step": 26250 }, { "epoch": 4.81, "learning_rate": 7.22709718255985e-06, "loss": 0.0226, "step": 26251 }, { "epoch": 4.81, "learning_rate": 7.226383873688215e-06, "loss": 0.024, "step": 26252 }, { "epoch": 4.81, "learning_rate": 7.225670580104816e-06, "loss": 0.0179, "step": 26253 }, { "epoch": 4.81, "learning_rate": 7.224957301813585e-06, "loss": 0.0176, "step": 26254 }, { "epoch": 4.81, "learning_rate": 7.224244038818454e-06, "loss": 0.0199, "step": 26255 }, { "epoch": 4.81, "learning_rate": 7.223530791123357e-06, "loss": 0.0204, "step": 26256 }, { "epoch": 4.82, "learning_rate": 7.2228175587322245e-06, "loss": 0.0228, "step": 26257 }, { "epoch": 4.82, "learning_rate": 7.22210434164898e-06, "loss": 0.0176, "step": 26258 }, { "epoch": 4.82, "learning_rate": 7.221391139877566e-06, "loss": 0.0184, "step": 26259 }, { "epoch": 4.82, "learning_rate": 7.220677953421911e-06, "loss": 0.0182, "step": 26260 }, { "epoch": 4.82, "learning_rate": 7.219964782285943e-06, "loss": 0.0228, "step": 26261 }, { "epoch": 4.82, "learning_rate": 7.219251626473593e-06, "loss": 0.0204, "step": 26262 }, { "epoch": 4.82, "learning_rate": 7.218538485988793e-06, "loss": 0.0215, "step": 26263 }, { "epoch": 4.82, "learning_rate": 7.217825360835475e-06, "loss": 0.0219, "step": 26264 }, { "epoch": 4.82, "learning_rate": 7.217112251017569e-06, "loss": 0.0205, "step": 26265 }, { "epoch": 4.82, "learning_rate": 7.216399156539007e-06, "loss": 0.0221, "step": 26266 }, { "epoch": 4.82, "learning_rate": 7.215686077403712e-06, "loss": 0.0202, "step": 26267 }, { "epoch": 4.82, "learning_rate": 7.214973013615625e-06, "loss": 0.0211, "step": 26268 }, { "epoch": 4.82, "learning_rate": 7.214259965178674e-06, "loss": 0.0204, "step": 26269 }, { "epoch": 4.82, "learning_rate": 7.2135469320967846e-06, "loss": 0.0173, "step": 26270 }, { "epoch": 4.82, "learning_rate": 7.21283391437389e-06, "loss": 0.0179, "step": 26271 }, { "epoch": 4.82, "learning_rate": 7.212120912013918e-06, "loss": 0.0219, "step": 26272 }, { "epoch": 4.82, "learning_rate": 7.211407925020804e-06, "loss": 0.0224, "step": 26273 }, { "epoch": 4.82, "learning_rate": 7.210694953398474e-06, "loss": 0.0219, "step": 26274 }, { "epoch": 4.82, "learning_rate": 7.209981997150861e-06, "loss": 0.0179, "step": 26275 }, { "epoch": 4.82, "learning_rate": 7.209269056281887e-06, "loss": 0.0244, "step": 26276 }, { "epoch": 4.82, "learning_rate": 7.20855613079549e-06, "loss": 0.0156, "step": 26277 }, { "epoch": 4.82, "learning_rate": 7.207843220695601e-06, "loss": 0.0225, "step": 26278 }, { "epoch": 4.82, "learning_rate": 7.207130325986143e-06, "loss": 0.0231, "step": 26279 }, { "epoch": 4.82, "learning_rate": 7.206417446671046e-06, "loss": 0.0226, "step": 26280 }, { "epoch": 4.82, "learning_rate": 7.2057045827542415e-06, "loss": 0.0189, "step": 26281 }, { "epoch": 4.82, "learning_rate": 7.20499173423966e-06, "loss": 0.022, "step": 26282 }, { "epoch": 4.82, "learning_rate": 7.204278901131229e-06, "loss": 0.0234, "step": 26283 }, { "epoch": 4.82, "learning_rate": 7.203566083432883e-06, "loss": 0.026, "step": 26284 }, { "epoch": 4.82, "learning_rate": 7.202853281148538e-06, "loss": 0.0185, "step": 26285 }, { "epoch": 4.82, "learning_rate": 7.202140494282135e-06, "loss": 0.0249, "step": 26286 }, { "epoch": 4.82, "learning_rate": 7.201427722837602e-06, "loss": 0.0226, "step": 26287 }, { "epoch": 4.82, "learning_rate": 7.200714966818862e-06, "loss": 0.0159, "step": 26288 }, { "epoch": 4.82, "learning_rate": 7.200002226229845e-06, "loss": 0.0224, "step": 26289 }, { "epoch": 4.82, "learning_rate": 7.1992895010744844e-06, "loss": 0.0253, "step": 26290 }, { "epoch": 4.82, "learning_rate": 7.198576791356705e-06, "loss": 0.0231, "step": 26291 }, { "epoch": 4.82, "learning_rate": 7.197864097080436e-06, "loss": 0.0204, "step": 26292 }, { "epoch": 4.82, "learning_rate": 7.197151418249608e-06, "loss": 0.0221, "step": 26293 }, { "epoch": 4.82, "learning_rate": 7.19643875486814e-06, "loss": 0.018, "step": 26294 }, { "epoch": 4.82, "learning_rate": 7.1957261069399745e-06, "loss": 0.0185, "step": 26295 }, { "epoch": 4.82, "learning_rate": 7.19501347446903e-06, "loss": 0.0228, "step": 26296 }, { "epoch": 4.82, "learning_rate": 7.194300857459237e-06, "loss": 0.0201, "step": 26297 }, { "epoch": 4.82, "learning_rate": 7.193588255914521e-06, "loss": 0.0223, "step": 26298 }, { "epoch": 4.82, "learning_rate": 7.192875669838815e-06, "loss": 0.0229, "step": 26299 }, { "epoch": 4.82, "learning_rate": 7.192163099236042e-06, "loss": 0.0197, "step": 26300 }, { "epoch": 4.82, "learning_rate": 7.1914505441101345e-06, "loss": 0.0255, "step": 26301 }, { "epoch": 4.82, "learning_rate": 7.190738004465018e-06, "loss": 0.0214, "step": 26302 }, { "epoch": 4.82, "learning_rate": 7.190025480304614e-06, "loss": 0.0188, "step": 26303 }, { "epoch": 4.82, "learning_rate": 7.18931297163286e-06, "loss": 0.0187, "step": 26304 }, { "epoch": 4.82, "learning_rate": 7.188600478453678e-06, "loss": 0.0204, "step": 26305 }, { "epoch": 4.82, "learning_rate": 7.187888000770996e-06, "loss": 0.0245, "step": 26306 }, { "epoch": 4.82, "learning_rate": 7.18717553858874e-06, "loss": 0.0296, "step": 26307 }, { "epoch": 4.82, "learning_rate": 7.186463091910838e-06, "loss": 0.022, "step": 26308 }, { "epoch": 4.82, "learning_rate": 7.18575066074122e-06, "loss": 0.0182, "step": 26309 }, { "epoch": 4.82, "learning_rate": 7.185038245083808e-06, "loss": 0.016, "step": 26310 }, { "epoch": 4.83, "learning_rate": 7.18432584494253e-06, "loss": 0.0201, "step": 26311 }, { "epoch": 4.83, "learning_rate": 7.183613460321317e-06, "loss": 0.0223, "step": 26312 }, { "epoch": 4.83, "learning_rate": 7.182901091224094e-06, "loss": 0.0234, "step": 26313 }, { "epoch": 4.83, "learning_rate": 7.1821887376547825e-06, "loss": 0.0179, "step": 26314 }, { "epoch": 4.83, "learning_rate": 7.181476399617315e-06, "loss": 0.0206, "step": 26315 }, { "epoch": 4.83, "learning_rate": 7.1807640771156115e-06, "loss": 0.0194, "step": 26316 }, { "epoch": 4.83, "learning_rate": 7.180051770153606e-06, "loss": 0.024, "step": 26317 }, { "epoch": 4.83, "learning_rate": 7.179339478735221e-06, "loss": 0.0194, "step": 26318 }, { "epoch": 4.83, "learning_rate": 7.178627202864383e-06, "loss": 0.0208, "step": 26319 }, { "epoch": 4.83, "learning_rate": 7.1779149425450154e-06, "loss": 0.0226, "step": 26320 }, { "epoch": 4.83, "learning_rate": 7.1772026977810495e-06, "loss": 0.0183, "step": 26321 }, { "epoch": 4.83, "learning_rate": 7.176490468576409e-06, "loss": 0.0173, "step": 26322 }, { "epoch": 4.83, "learning_rate": 7.175778254935019e-06, "loss": 0.0147, "step": 26323 }, { "epoch": 4.83, "learning_rate": 7.175066056860805e-06, "loss": 0.0217, "step": 26324 }, { "epoch": 4.83, "learning_rate": 7.17435387435769e-06, "loss": 0.0293, "step": 26325 }, { "epoch": 4.83, "learning_rate": 7.173641707429606e-06, "loss": 0.0212, "step": 26326 }, { "epoch": 4.83, "learning_rate": 7.1729295560804745e-06, "loss": 0.0196, "step": 26327 }, { "epoch": 4.83, "learning_rate": 7.17221742031422e-06, "loss": 0.0265, "step": 26328 }, { "epoch": 4.83, "learning_rate": 7.171505300134769e-06, "loss": 0.0209, "step": 26329 }, { "epoch": 4.83, "learning_rate": 7.170793195546048e-06, "loss": 0.0187, "step": 26330 }, { "epoch": 4.83, "learning_rate": 7.170081106551985e-06, "loss": 0.018, "step": 26331 }, { "epoch": 4.83, "learning_rate": 7.169369033156499e-06, "loss": 0.0224, "step": 26332 }, { "epoch": 4.83, "learning_rate": 7.168656975363512e-06, "loss": 0.0213, "step": 26333 }, { "epoch": 4.83, "learning_rate": 7.16794493317696e-06, "loss": 0.0224, "step": 26334 }, { "epoch": 4.83, "learning_rate": 7.167232906600761e-06, "loss": 0.0195, "step": 26335 }, { "epoch": 4.83, "learning_rate": 7.16652089563884e-06, "loss": 0.0268, "step": 26336 }, { "epoch": 4.83, "learning_rate": 7.165808900295121e-06, "loss": 0.0234, "step": 26337 }, { "epoch": 4.83, "learning_rate": 7.16509692057353e-06, "loss": 0.0196, "step": 26338 }, { "epoch": 4.83, "learning_rate": 7.164384956477992e-06, "loss": 0.0225, "step": 26339 }, { "epoch": 4.83, "learning_rate": 7.163673008012433e-06, "loss": 0.0221, "step": 26340 }, { "epoch": 4.83, "learning_rate": 7.162961075180772e-06, "loss": 0.0212, "step": 26341 }, { "epoch": 4.83, "learning_rate": 7.162249157986935e-06, "loss": 0.0188, "step": 26342 }, { "epoch": 4.83, "learning_rate": 7.1615372564348485e-06, "loss": 0.0214, "step": 26343 }, { "epoch": 4.83, "learning_rate": 7.160825370528435e-06, "loss": 0.0235, "step": 26344 }, { "epoch": 4.83, "learning_rate": 7.1601135002716194e-06, "loss": 0.0248, "step": 26345 }, { "epoch": 4.83, "learning_rate": 7.159401645668324e-06, "loss": 0.0175, "step": 26346 }, { "epoch": 4.83, "learning_rate": 7.158689806722471e-06, "loss": 0.0204, "step": 26347 }, { "epoch": 4.83, "learning_rate": 7.157977983437989e-06, "loss": 0.0224, "step": 26348 }, { "epoch": 4.83, "learning_rate": 7.1572661758188e-06, "loss": 0.0219, "step": 26349 }, { "epoch": 4.83, "learning_rate": 7.156554383868826e-06, "loss": 0.0198, "step": 26350 }, { "epoch": 4.83, "learning_rate": 7.155842607591987e-06, "loss": 0.0198, "step": 26351 }, { "epoch": 4.83, "learning_rate": 7.155130846992213e-06, "loss": 0.0213, "step": 26352 }, { "epoch": 4.83, "learning_rate": 7.154419102073426e-06, "loss": 0.0246, "step": 26353 }, { "epoch": 4.83, "learning_rate": 7.153707372839546e-06, "loss": 0.0235, "step": 26354 }, { "epoch": 4.83, "learning_rate": 7.152995659294497e-06, "loss": 0.0243, "step": 26355 }, { "epoch": 4.83, "learning_rate": 7.152283961442201e-06, "loss": 0.0212, "step": 26356 }, { "epoch": 4.83, "learning_rate": 7.151572279286586e-06, "loss": 0.0201, "step": 26357 }, { "epoch": 4.83, "learning_rate": 7.1508606128315715e-06, "loss": 0.0206, "step": 26358 }, { "epoch": 4.83, "learning_rate": 7.1501489620810804e-06, "loss": 0.0206, "step": 26359 }, { "epoch": 4.83, "learning_rate": 7.1494373270390305e-06, "loss": 0.0225, "step": 26360 }, { "epoch": 4.83, "learning_rate": 7.148725707709352e-06, "loss": 0.0233, "step": 26361 }, { "epoch": 4.83, "learning_rate": 7.148014104095964e-06, "loss": 0.0186, "step": 26362 }, { "epoch": 4.83, "learning_rate": 7.1473025162027886e-06, "loss": 0.0212, "step": 26363 }, { "epoch": 4.83, "learning_rate": 7.146590944033746e-06, "loss": 0.0242, "step": 26364 }, { "epoch": 4.83, "learning_rate": 7.145879387592764e-06, "loss": 0.022, "step": 26365 }, { "epoch": 4.84, "learning_rate": 7.145167846883762e-06, "loss": 0.022, "step": 26366 }, { "epoch": 4.84, "learning_rate": 7.144456321910665e-06, "loss": 0.0223, "step": 26367 }, { "epoch": 4.84, "learning_rate": 7.143744812677387e-06, "loss": 0.0215, "step": 26368 }, { "epoch": 4.84, "learning_rate": 7.143033319187853e-06, "loss": 0.0197, "step": 26369 }, { "epoch": 4.84, "learning_rate": 7.142321841445989e-06, "loss": 0.0245, "step": 26370 }, { "epoch": 4.84, "learning_rate": 7.141610379455715e-06, "loss": 0.0223, "step": 26371 }, { "epoch": 4.84, "learning_rate": 7.14089893322095e-06, "loss": 0.0217, "step": 26372 }, { "epoch": 4.84, "learning_rate": 7.140187502745615e-06, "loss": 0.0214, "step": 26373 }, { "epoch": 4.84, "learning_rate": 7.139476088033635e-06, "loss": 0.0226, "step": 26374 }, { "epoch": 4.84, "learning_rate": 7.138764689088932e-06, "loss": 0.0194, "step": 26375 }, { "epoch": 4.84, "learning_rate": 7.138053305915426e-06, "loss": 0.0232, "step": 26376 }, { "epoch": 4.84, "learning_rate": 7.137341938517035e-06, "loss": 0.0225, "step": 26377 }, { "epoch": 4.84, "learning_rate": 7.136630586897679e-06, "loss": 0.0206, "step": 26378 }, { "epoch": 4.84, "learning_rate": 7.1359192510612864e-06, "loss": 0.026, "step": 26379 }, { "epoch": 4.84, "learning_rate": 7.135207931011773e-06, "loss": 0.0217, "step": 26380 }, { "epoch": 4.84, "learning_rate": 7.13449662675306e-06, "loss": 0.0208, "step": 26381 }, { "epoch": 4.84, "learning_rate": 7.133785338289068e-06, "loss": 0.0241, "step": 26382 }, { "epoch": 4.84, "learning_rate": 7.133074065623719e-06, "loss": 0.0222, "step": 26383 }, { "epoch": 4.84, "learning_rate": 7.132362808760934e-06, "loss": 0.0281, "step": 26384 }, { "epoch": 4.84, "learning_rate": 7.131651567704635e-06, "loss": 0.0139, "step": 26385 }, { "epoch": 4.84, "learning_rate": 7.130940342458732e-06, "loss": 0.0193, "step": 26386 }, { "epoch": 4.84, "learning_rate": 7.13022913302716e-06, "loss": 0.0186, "step": 26387 }, { "epoch": 4.84, "learning_rate": 7.129517939413831e-06, "loss": 0.0213, "step": 26388 }, { "epoch": 4.84, "learning_rate": 7.1288067616226655e-06, "loss": 0.0238, "step": 26389 }, { "epoch": 4.84, "learning_rate": 7.128095599657584e-06, "loss": 0.0254, "step": 26390 }, { "epoch": 4.84, "learning_rate": 7.127384453522506e-06, "loss": 0.0207, "step": 26391 }, { "epoch": 4.84, "learning_rate": 7.126673323221352e-06, "loss": 0.0233, "step": 26392 }, { "epoch": 4.84, "learning_rate": 7.125962208758044e-06, "loss": 0.0168, "step": 26393 }, { "epoch": 4.84, "learning_rate": 7.125251110136502e-06, "loss": 0.0195, "step": 26394 }, { "epoch": 4.84, "learning_rate": 7.124540027360636e-06, "loss": 0.0276, "step": 26395 }, { "epoch": 4.84, "learning_rate": 7.123828960434378e-06, "loss": 0.0221, "step": 26396 }, { "epoch": 4.84, "learning_rate": 7.123117909361641e-06, "loss": 0.0196, "step": 26397 }, { "epoch": 4.84, "learning_rate": 7.122406874146346e-06, "loss": 0.0167, "step": 26398 }, { "epoch": 4.84, "learning_rate": 7.121695854792412e-06, "loss": 0.0233, "step": 26399 }, { "epoch": 4.84, "learning_rate": 7.120984851303755e-06, "loss": 0.0214, "step": 26400 }, { "epoch": 4.84, "learning_rate": 7.120273863684299e-06, "loss": 0.0249, "step": 26401 }, { "epoch": 4.84, "learning_rate": 7.119562891937961e-06, "loss": 0.0208, "step": 26402 }, { "epoch": 4.84, "learning_rate": 7.118851936068663e-06, "loss": 0.018, "step": 26403 }, { "epoch": 4.84, "learning_rate": 7.118140996080313e-06, "loss": 0.0244, "step": 26404 }, { "epoch": 4.84, "learning_rate": 7.117430071976843e-06, "loss": 0.0249, "step": 26405 }, { "epoch": 4.84, "learning_rate": 7.1167191637621655e-06, "loss": 0.0213, "step": 26406 }, { "epoch": 4.84, "learning_rate": 7.116008271440199e-06, "loss": 0.0237, "step": 26407 }, { "epoch": 4.84, "learning_rate": 7.1152973950148615e-06, "loss": 0.0216, "step": 26408 }, { "epoch": 4.84, "learning_rate": 7.114586534490072e-06, "loss": 0.0254, "step": 26409 }, { "epoch": 4.84, "learning_rate": 7.11387568986975e-06, "loss": 0.0199, "step": 26410 }, { "epoch": 4.84, "learning_rate": 7.113164861157815e-06, "loss": 0.0216, "step": 26411 }, { "epoch": 4.84, "learning_rate": 7.112454048358182e-06, "loss": 0.0219, "step": 26412 }, { "epoch": 4.84, "learning_rate": 7.111743251474766e-06, "loss": 0.0198, "step": 26413 }, { "epoch": 4.84, "learning_rate": 7.111032470511492e-06, "loss": 0.0199, "step": 26414 }, { "epoch": 4.84, "learning_rate": 7.110321705472275e-06, "loss": 0.0255, "step": 26415 }, { "epoch": 4.84, "learning_rate": 7.1096109563610326e-06, "loss": 0.02, "step": 26416 }, { "epoch": 4.84, "learning_rate": 7.1089002231816796e-06, "loss": 0.0224, "step": 26417 }, { "epoch": 4.84, "learning_rate": 7.108189505938139e-06, "loss": 0.0254, "step": 26418 }, { "epoch": 4.84, "learning_rate": 7.107478804634324e-06, "loss": 0.0209, "step": 26419 }, { "epoch": 4.85, "learning_rate": 7.106768119274157e-06, "loss": 0.0185, "step": 26420 }, { "epoch": 4.85, "learning_rate": 7.10605744986155e-06, "loss": 0.0205, "step": 26421 }, { "epoch": 4.85, "learning_rate": 7.1053467964004165e-06, "loss": 0.0201, "step": 26422 }, { "epoch": 4.85, "learning_rate": 7.104636158894687e-06, "loss": 0.024, "step": 26423 }, { "epoch": 4.85, "learning_rate": 7.103925537348268e-06, "loss": 0.0153, "step": 26424 }, { "epoch": 4.85, "learning_rate": 7.10321493176508e-06, "loss": 0.024, "step": 26425 }, { "epoch": 4.85, "learning_rate": 7.102504342149037e-06, "loss": 0.0209, "step": 26426 }, { "epoch": 4.85, "learning_rate": 7.10179376850406e-06, "loss": 0.0258, "step": 26427 }, { "epoch": 4.85, "learning_rate": 7.1010832108340636e-06, "loss": 0.0169, "step": 26428 }, { "epoch": 4.85, "learning_rate": 7.100372669142965e-06, "loss": 0.0258, "step": 26429 }, { "epoch": 4.85, "learning_rate": 7.09966214343468e-06, "loss": 0.0256, "step": 26430 }, { "epoch": 4.85, "learning_rate": 7.098951633713121e-06, "loss": 0.0223, "step": 26431 }, { "epoch": 4.85, "learning_rate": 7.098241139982214e-06, "loss": 0.0192, "step": 26432 }, { "epoch": 4.85, "learning_rate": 7.097530662245867e-06, "loss": 0.0248, "step": 26433 }, { "epoch": 4.85, "learning_rate": 7.096820200508e-06, "loss": 0.017, "step": 26434 }, { "epoch": 4.85, "learning_rate": 7.096109754772526e-06, "loss": 0.0186, "step": 26435 }, { "epoch": 4.85, "learning_rate": 7.095399325043365e-06, "loss": 0.0203, "step": 26436 }, { "epoch": 4.85, "learning_rate": 7.094688911324431e-06, "loss": 0.0221, "step": 26437 }, { "epoch": 4.85, "learning_rate": 7.0939785136196415e-06, "loss": 0.0237, "step": 26438 }, { "epoch": 4.85, "learning_rate": 7.093268131932905e-06, "loss": 0.0205, "step": 26439 }, { "epoch": 4.85, "learning_rate": 7.092557766268147e-06, "loss": 0.0195, "step": 26440 }, { "epoch": 4.85, "learning_rate": 7.091847416629281e-06, "loss": 0.0242, "step": 26441 }, { "epoch": 4.85, "learning_rate": 7.091137083020217e-06, "loss": 0.0222, "step": 26442 }, { "epoch": 4.85, "learning_rate": 7.090426765444873e-06, "loss": 0.0193, "step": 26443 }, { "epoch": 4.85, "learning_rate": 7.089716463907164e-06, "loss": 0.0159, "step": 26444 }, { "epoch": 4.85, "learning_rate": 7.089006178411008e-06, "loss": 0.0258, "step": 26445 }, { "epoch": 4.85, "learning_rate": 7.088295908960317e-06, "loss": 0.0201, "step": 26446 }, { "epoch": 4.85, "learning_rate": 7.08758565555901e-06, "loss": 0.0196, "step": 26447 }, { "epoch": 4.85, "learning_rate": 7.0868754182109945e-06, "loss": 0.0221, "step": 26448 }, { "epoch": 4.85, "learning_rate": 7.086165196920192e-06, "loss": 0.0208, "step": 26449 }, { "epoch": 4.85, "learning_rate": 7.085454991690518e-06, "loss": 0.0243, "step": 26450 }, { "epoch": 4.85, "learning_rate": 7.084744802525883e-06, "loss": 0.0165, "step": 26451 }, { "epoch": 4.85, "learning_rate": 7.084034629430201e-06, "loss": 0.0233, "step": 26452 }, { "epoch": 4.85, "learning_rate": 7.0833244724073875e-06, "loss": 0.0196, "step": 26453 }, { "epoch": 4.85, "learning_rate": 7.0826143314613595e-06, "loss": 0.0202, "step": 26454 }, { "epoch": 4.85, "learning_rate": 7.081904206596031e-06, "loss": 0.0233, "step": 26455 }, { "epoch": 4.85, "learning_rate": 7.081194097815316e-06, "loss": 0.0212, "step": 26456 }, { "epoch": 4.85, "learning_rate": 7.08048400512312e-06, "loss": 0.0228, "step": 26457 }, { "epoch": 4.85, "learning_rate": 7.079773928523372e-06, "loss": 0.0185, "step": 26458 }, { "epoch": 4.85, "learning_rate": 7.079063868019978e-06, "loss": 0.022, "step": 26459 }, { "epoch": 4.85, "learning_rate": 7.078353823616851e-06, "loss": 0.0222, "step": 26460 }, { "epoch": 4.85, "learning_rate": 7.077643795317904e-06, "loss": 0.021, "step": 26461 }, { "epoch": 4.85, "learning_rate": 7.076933783127054e-06, "loss": 0.0231, "step": 26462 }, { "epoch": 4.85, "learning_rate": 7.076223787048215e-06, "loss": 0.0237, "step": 26463 }, { "epoch": 4.85, "learning_rate": 7.075513807085299e-06, "loss": 0.0199, "step": 26464 }, { "epoch": 4.85, "learning_rate": 7.07480384324222e-06, "loss": 0.0182, "step": 26465 }, { "epoch": 4.85, "learning_rate": 7.0740938955228854e-06, "loss": 0.0188, "step": 26466 }, { "epoch": 4.85, "learning_rate": 7.07338396393122e-06, "loss": 0.0195, "step": 26467 }, { "epoch": 4.85, "learning_rate": 7.072674048471129e-06, "loss": 0.0188, "step": 26468 }, { "epoch": 4.85, "learning_rate": 7.071964149146526e-06, "loss": 0.0186, "step": 26469 }, { "epoch": 4.85, "learning_rate": 7.071254265961325e-06, "loss": 0.0246, "step": 26470 }, { "epoch": 4.85, "learning_rate": 7.07054439891944e-06, "loss": 0.0254, "step": 26471 }, { "epoch": 4.85, "learning_rate": 7.069834548024783e-06, "loss": 0.0201, "step": 26472 }, { "epoch": 4.85, "learning_rate": 7.069124713281268e-06, "loss": 0.0217, "step": 26473 }, { "epoch": 4.85, "learning_rate": 7.068414894692805e-06, "loss": 0.0204, "step": 26474 }, { "epoch": 4.86, "learning_rate": 7.067705092263305e-06, "loss": 0.0232, "step": 26475 }, { "epoch": 4.86, "learning_rate": 7.066995305996688e-06, "loss": 0.0191, "step": 26476 }, { "epoch": 4.86, "learning_rate": 7.06628553589686e-06, "loss": 0.0198, "step": 26477 }, { "epoch": 4.86, "learning_rate": 7.065575781967734e-06, "loss": 0.0175, "step": 26478 }, { "epoch": 4.86, "learning_rate": 7.064866044213221e-06, "loss": 0.0271, "step": 26479 }, { "epoch": 4.86, "learning_rate": 7.064156322637238e-06, "loss": 0.0191, "step": 26480 }, { "epoch": 4.86, "learning_rate": 7.063446617243695e-06, "loss": 0.0172, "step": 26481 }, { "epoch": 4.86, "learning_rate": 7.0627369280365e-06, "loss": 0.0191, "step": 26482 }, { "epoch": 4.86, "learning_rate": 7.062027255019572e-06, "loss": 0.0225, "step": 26483 }, { "epoch": 4.86, "learning_rate": 7.061317598196813e-06, "loss": 0.0153, "step": 26484 }, { "epoch": 4.86, "learning_rate": 7.060607957572145e-06, "loss": 0.0255, "step": 26485 }, { "epoch": 4.86, "learning_rate": 7.059898333149473e-06, "loss": 0.0185, "step": 26486 }, { "epoch": 4.86, "learning_rate": 7.059188724932711e-06, "loss": 0.0224, "step": 26487 }, { "epoch": 4.86, "learning_rate": 7.058479132925766e-06, "loss": 0.0212, "step": 26488 }, { "epoch": 4.86, "learning_rate": 7.057769557132556e-06, "loss": 0.0222, "step": 26489 }, { "epoch": 4.86, "learning_rate": 7.05705999755699e-06, "loss": 0.0226, "step": 26490 }, { "epoch": 4.86, "learning_rate": 7.056350454202978e-06, "loss": 0.0223, "step": 26491 }, { "epoch": 4.86, "learning_rate": 7.0556409270744285e-06, "loss": 0.0195, "step": 26492 }, { "epoch": 4.86, "learning_rate": 7.054931416175257e-06, "loss": 0.0226, "step": 26493 }, { "epoch": 4.86, "learning_rate": 7.0542219215093755e-06, "loss": 0.0218, "step": 26494 }, { "epoch": 4.86, "learning_rate": 7.05351244308069e-06, "loss": 0.0208, "step": 26495 }, { "epoch": 4.86, "learning_rate": 7.052802980893112e-06, "loss": 0.02, "step": 26496 }, { "epoch": 4.86, "learning_rate": 7.052093534950551e-06, "loss": 0.018, "step": 26497 }, { "epoch": 4.86, "learning_rate": 7.051384105256922e-06, "loss": 0.0237, "step": 26498 }, { "epoch": 4.86, "learning_rate": 7.050674691816133e-06, "loss": 0.0204, "step": 26499 }, { "epoch": 4.86, "learning_rate": 7.049965294632093e-06, "loss": 0.021, "step": 26500 }, { "epoch": 4.86, "learning_rate": 7.0492559137087125e-06, "loss": 0.0165, "step": 26501 }, { "epoch": 4.86, "learning_rate": 7.0485465490499035e-06, "loss": 0.0195, "step": 26502 }, { "epoch": 4.86, "learning_rate": 7.0478372006595786e-06, "loss": 0.022, "step": 26503 }, { "epoch": 4.86, "learning_rate": 7.047127868541641e-06, "loss": 0.0255, "step": 26504 }, { "epoch": 4.86, "learning_rate": 7.046418552700005e-06, "loss": 0.0236, "step": 26505 }, { "epoch": 4.86, "learning_rate": 7.045709253138575e-06, "loss": 0.0196, "step": 26506 }, { "epoch": 4.86, "learning_rate": 7.044999969861268e-06, "loss": 0.0182, "step": 26507 }, { "epoch": 4.86, "learning_rate": 7.044290702871992e-06, "loss": 0.0235, "step": 26508 }, { "epoch": 4.86, "learning_rate": 7.043581452174653e-06, "loss": 0.0186, "step": 26509 }, { "epoch": 4.86, "learning_rate": 7.04287221777316e-06, "loss": 0.0174, "step": 26510 }, { "epoch": 4.86, "learning_rate": 7.042162999671427e-06, "loss": 0.0203, "step": 26511 }, { "epoch": 4.86, "learning_rate": 7.041453797873363e-06, "loss": 0.0203, "step": 26512 }, { "epoch": 4.86, "learning_rate": 7.040744612382874e-06, "loss": 0.0219, "step": 26513 }, { "epoch": 4.86, "learning_rate": 7.0400354432038676e-06, "loss": 0.0207, "step": 26514 }, { "epoch": 4.86, "learning_rate": 7.039326290340256e-06, "loss": 0.0235, "step": 26515 }, { "epoch": 4.86, "learning_rate": 7.038617153795948e-06, "loss": 0.0172, "step": 26516 }, { "epoch": 4.86, "learning_rate": 7.037908033574851e-06, "loss": 0.0232, "step": 26517 }, { "epoch": 4.86, "learning_rate": 7.037198929680877e-06, "loss": 0.0184, "step": 26518 }, { "epoch": 4.86, "learning_rate": 7.036489842117926e-06, "loss": 0.0184, "step": 26519 }, { "epoch": 4.86, "learning_rate": 7.035780770889917e-06, "loss": 0.023, "step": 26520 }, { "epoch": 4.86, "learning_rate": 7.035071716000756e-06, "loss": 0.0218, "step": 26521 }, { "epoch": 4.86, "learning_rate": 7.034362677454345e-06, "loss": 0.019, "step": 26522 }, { "epoch": 4.86, "learning_rate": 7.0336536552545955e-06, "loss": 0.0215, "step": 26523 }, { "epoch": 4.86, "learning_rate": 7.03294464940542e-06, "loss": 0.0216, "step": 26524 }, { "epoch": 4.86, "learning_rate": 7.032235659910721e-06, "loss": 0.0246, "step": 26525 }, { "epoch": 4.86, "learning_rate": 7.031526686774409e-06, "loss": 0.0226, "step": 26526 }, { "epoch": 4.86, "learning_rate": 7.030817730000394e-06, "loss": 0.0235, "step": 26527 }, { "epoch": 4.86, "learning_rate": 7.030108789592576e-06, "loss": 0.0194, "step": 26528 }, { "epoch": 4.87, "learning_rate": 7.02939986555487e-06, "loss": 0.0194, "step": 26529 }, { "epoch": 4.87, "learning_rate": 7.028690957891184e-06, "loss": 0.0207, "step": 26530 }, { "epoch": 4.87, "learning_rate": 7.027982066605423e-06, "loss": 0.018, "step": 26531 }, { "epoch": 4.87, "learning_rate": 7.02727319170149e-06, "loss": 0.0258, "step": 26532 }, { "epoch": 4.87, "learning_rate": 7.0265643331833e-06, "loss": 0.0162, "step": 26533 }, { "epoch": 4.87, "learning_rate": 7.025855491054757e-06, "loss": 0.014, "step": 26534 }, { "epoch": 4.87, "learning_rate": 7.0251466653197686e-06, "loss": 0.0211, "step": 26535 }, { "epoch": 4.87, "learning_rate": 7.024437855982242e-06, "loss": 0.017, "step": 26536 }, { "epoch": 4.87, "learning_rate": 7.023729063046079e-06, "loss": 0.0223, "step": 26537 }, { "epoch": 4.87, "learning_rate": 7.0230202865151955e-06, "loss": 0.0202, "step": 26538 }, { "epoch": 4.87, "learning_rate": 7.022311526393495e-06, "loss": 0.0191, "step": 26539 }, { "epoch": 4.87, "learning_rate": 7.021602782684882e-06, "loss": 0.0203, "step": 26540 }, { "epoch": 4.87, "learning_rate": 7.0208940553932614e-06, "loss": 0.0225, "step": 26541 }, { "epoch": 4.87, "learning_rate": 7.020185344522544e-06, "loss": 0.0206, "step": 26542 }, { "epoch": 4.87, "learning_rate": 7.019476650076635e-06, "loss": 0.018, "step": 26543 }, { "epoch": 4.87, "learning_rate": 7.018767972059441e-06, "loss": 0.0199, "step": 26544 }, { "epoch": 4.87, "learning_rate": 7.018059310474866e-06, "loss": 0.0209, "step": 26545 }, { "epoch": 4.87, "learning_rate": 7.017350665326821e-06, "loss": 0.0221, "step": 26546 }, { "epoch": 4.87, "learning_rate": 7.016642036619207e-06, "loss": 0.0248, "step": 26547 }, { "epoch": 4.87, "learning_rate": 7.015933424355935e-06, "loss": 0.0193, "step": 26548 }, { "epoch": 4.87, "learning_rate": 7.015224828540905e-06, "loss": 0.0238, "step": 26549 }, { "epoch": 4.87, "learning_rate": 7.014516249178025e-06, "loss": 0.0256, "step": 26550 }, { "epoch": 4.87, "learning_rate": 7.013807686271202e-06, "loss": 0.0196, "step": 26551 }, { "epoch": 4.87, "learning_rate": 7.013099139824342e-06, "loss": 0.0203, "step": 26552 }, { "epoch": 4.87, "learning_rate": 7.012390609841349e-06, "loss": 0.026, "step": 26553 }, { "epoch": 4.87, "learning_rate": 7.011682096326127e-06, "loss": 0.0179, "step": 26554 }, { "epoch": 4.87, "learning_rate": 7.010973599282587e-06, "loss": 0.0206, "step": 26555 }, { "epoch": 4.87, "learning_rate": 7.010265118714628e-06, "loss": 0.0169, "step": 26556 }, { "epoch": 4.87, "learning_rate": 7.009556654626161e-06, "loss": 0.0218, "step": 26557 }, { "epoch": 4.87, "learning_rate": 7.008848207021085e-06, "loss": 0.0237, "step": 26558 }, { "epoch": 4.87, "learning_rate": 7.008139775903306e-06, "loss": 0.0194, "step": 26559 }, { "epoch": 4.87, "learning_rate": 7.007431361276733e-06, "loss": 0.019, "step": 26560 }, { "epoch": 4.87, "learning_rate": 7.006722963145269e-06, "loss": 0.0243, "step": 26561 }, { "epoch": 4.87, "learning_rate": 7.006014581512816e-06, "loss": 0.0214, "step": 26562 }, { "epoch": 4.87, "learning_rate": 7.005306216383281e-06, "loss": 0.0212, "step": 26563 }, { "epoch": 4.87, "learning_rate": 7.004597867760569e-06, "loss": 0.0195, "step": 26564 }, { "epoch": 4.87, "learning_rate": 7.003889535648587e-06, "loss": 0.0195, "step": 26565 }, { "epoch": 4.87, "learning_rate": 7.003181220051234e-06, "loss": 0.0162, "step": 26566 }, { "epoch": 4.87, "learning_rate": 7.002472920972411e-06, "loss": 0.0222, "step": 26567 }, { "epoch": 4.87, "learning_rate": 7.001764638416034e-06, "loss": 0.0202, "step": 26568 }, { "epoch": 4.87, "learning_rate": 7.001056372385999e-06, "loss": 0.022, "step": 26569 }, { "epoch": 4.87, "learning_rate": 7.000348122886211e-06, "loss": 0.0201, "step": 26570 }, { "epoch": 4.87, "learning_rate": 6.999639889920575e-06, "loss": 0.0221, "step": 26571 }, { "epoch": 4.87, "learning_rate": 6.998931673492993e-06, "loss": 0.0225, "step": 26572 }, { "epoch": 4.87, "learning_rate": 6.998223473607369e-06, "loss": 0.0187, "step": 26573 }, { "epoch": 4.87, "learning_rate": 6.997515290267611e-06, "loss": 0.0214, "step": 26574 }, { "epoch": 4.87, "learning_rate": 6.996807123477618e-06, "loss": 0.0237, "step": 26575 }, { "epoch": 4.87, "learning_rate": 6.99609897324129e-06, "loss": 0.0209, "step": 26576 }, { "epoch": 4.87, "learning_rate": 6.995390839562541e-06, "loss": 0.0197, "step": 26577 }, { "epoch": 4.87, "learning_rate": 6.994682722445265e-06, "loss": 0.0205, "step": 26578 }, { "epoch": 4.87, "learning_rate": 6.9939746218933695e-06, "loss": 0.023, "step": 26579 }, { "epoch": 4.87, "learning_rate": 6.993266537910757e-06, "loss": 0.0202, "step": 26580 }, { "epoch": 4.87, "learning_rate": 6.992558470501325e-06, "loss": 0.0224, "step": 26581 }, { "epoch": 4.87, "learning_rate": 6.991850419668986e-06, "loss": 0.0215, "step": 26582 }, { "epoch": 4.87, "learning_rate": 6.991142385417639e-06, "loss": 0.0215, "step": 26583 }, { "epoch": 4.88, "learning_rate": 6.990434367751182e-06, "loss": 0.0193, "step": 26584 }, { "epoch": 4.88, "learning_rate": 6.989726366673519e-06, "loss": 0.0222, "step": 26585 }, { "epoch": 4.88, "learning_rate": 6.989018382188562e-06, "loss": 0.0162, "step": 26586 }, { "epoch": 4.88, "learning_rate": 6.9883104143002016e-06, "loss": 0.0241, "step": 26587 }, { "epoch": 4.88, "learning_rate": 6.987602463012346e-06, "loss": 0.0234, "step": 26588 }, { "epoch": 4.88, "learning_rate": 6.986894528328893e-06, "loss": 0.0206, "step": 26589 }, { "epoch": 4.88, "learning_rate": 6.986186610253751e-06, "loss": 0.0173, "step": 26590 }, { "epoch": 4.88, "learning_rate": 6.98547870879082e-06, "loss": 0.0197, "step": 26591 }, { "epoch": 4.88, "learning_rate": 6.984770823944002e-06, "loss": 0.0215, "step": 26592 }, { "epoch": 4.88, "learning_rate": 6.984062955717196e-06, "loss": 0.0218, "step": 26593 }, { "epoch": 4.88, "learning_rate": 6.983355104114301e-06, "loss": 0.0202, "step": 26594 }, { "epoch": 4.88, "learning_rate": 6.98264726913923e-06, "loss": 0.0233, "step": 26595 }, { "epoch": 4.88, "learning_rate": 6.981939450795876e-06, "loss": 0.02, "step": 26596 }, { "epoch": 4.88, "learning_rate": 6.981231649088142e-06, "loss": 0.019, "step": 26597 }, { "epoch": 4.88, "learning_rate": 6.980523864019928e-06, "loss": 0.0245, "step": 26598 }, { "epoch": 4.88, "learning_rate": 6.97981609559514e-06, "loss": 0.0258, "step": 26599 }, { "epoch": 4.88, "learning_rate": 6.979108343817676e-06, "loss": 0.0179, "step": 26600 }, { "epoch": 4.88, "learning_rate": 6.978400608691438e-06, "loss": 0.0221, "step": 26601 }, { "epoch": 4.88, "learning_rate": 6.977692890220326e-06, "loss": 0.02, "step": 26602 }, { "epoch": 4.88, "learning_rate": 6.976985188408239e-06, "loss": 0.0258, "step": 26603 }, { "epoch": 4.88, "learning_rate": 6.9762775032590835e-06, "loss": 0.0227, "step": 26604 }, { "epoch": 4.88, "learning_rate": 6.975569834776757e-06, "loss": 0.0223, "step": 26605 }, { "epoch": 4.88, "learning_rate": 6.974862182965161e-06, "loss": 0.0191, "step": 26606 }, { "epoch": 4.88, "learning_rate": 6.974154547828191e-06, "loss": 0.022, "step": 26607 }, { "epoch": 4.88, "learning_rate": 6.973446929369757e-06, "loss": 0.022, "step": 26608 }, { "epoch": 4.88, "learning_rate": 6.972739327593754e-06, "loss": 0.0188, "step": 26609 }, { "epoch": 4.88, "learning_rate": 6.9720317425040825e-06, "loss": 0.0198, "step": 26610 }, { "epoch": 4.88, "learning_rate": 6.971324174104644e-06, "loss": 0.02, "step": 26611 }, { "epoch": 4.88, "learning_rate": 6.970616622399333e-06, "loss": 0.0189, "step": 26612 }, { "epoch": 4.88, "learning_rate": 6.969909087392057e-06, "loss": 0.0187, "step": 26613 }, { "epoch": 4.88, "learning_rate": 6.9692015690867135e-06, "loss": 0.018, "step": 26614 }, { "epoch": 4.88, "learning_rate": 6.9684940674872015e-06, "loss": 0.0196, "step": 26615 }, { "epoch": 4.88, "learning_rate": 6.9677865825974185e-06, "loss": 0.0217, "step": 26616 }, { "epoch": 4.88, "learning_rate": 6.96707911442127e-06, "loss": 0.023, "step": 26617 }, { "epoch": 4.88, "learning_rate": 6.966371662962652e-06, "loss": 0.0172, "step": 26618 }, { "epoch": 4.88, "learning_rate": 6.965664228225469e-06, "loss": 0.0178, "step": 26619 }, { "epoch": 4.88, "learning_rate": 6.964956810213606e-06, "loss": 0.0173, "step": 26620 }, { "epoch": 4.88, "learning_rate": 6.96424940893098e-06, "loss": 0.0177, "step": 26621 }, { "epoch": 4.88, "learning_rate": 6.96354202438148e-06, "loss": 0.0263, "step": 26622 }, { "epoch": 4.88, "learning_rate": 6.962834656569007e-06, "loss": 0.0239, "step": 26623 }, { "epoch": 4.88, "learning_rate": 6.962127305497462e-06, "loss": 0.0189, "step": 26624 }, { "epoch": 4.88, "learning_rate": 6.9614199711707395e-06, "loss": 0.0236, "step": 26625 }, { "epoch": 4.88, "learning_rate": 6.960712653592743e-06, "loss": 0.0216, "step": 26626 }, { "epoch": 4.88, "learning_rate": 6.96000535276737e-06, "loss": 0.0176, "step": 26627 }, { "epoch": 4.88, "learning_rate": 6.959298068698522e-06, "loss": 0.0183, "step": 26628 }, { "epoch": 4.88, "learning_rate": 6.958590801390086e-06, "loss": 0.026, "step": 26629 }, { "epoch": 4.88, "learning_rate": 6.9578835508459764e-06, "loss": 0.0228, "step": 26630 }, { "epoch": 4.88, "learning_rate": 6.95717631707008e-06, "loss": 0.0215, "step": 26631 }, { "epoch": 4.88, "learning_rate": 6.956469100066301e-06, "loss": 0.016, "step": 26632 }, { "epoch": 4.88, "learning_rate": 6.955761899838534e-06, "loss": 0.0213, "step": 26633 }, { "epoch": 4.88, "learning_rate": 6.955054716390679e-06, "loss": 0.0179, "step": 26634 }, { "epoch": 4.88, "learning_rate": 6.954347549726634e-06, "loss": 0.0166, "step": 26635 }, { "epoch": 4.88, "learning_rate": 6.953640399850297e-06, "loss": 0.0205, "step": 26636 }, { "epoch": 4.88, "learning_rate": 6.952933266765567e-06, "loss": 0.017, "step": 26637 }, { "epoch": 4.89, "learning_rate": 6.952226150476335e-06, "loss": 0.0201, "step": 26638 }, { "epoch": 4.89, "learning_rate": 6.95151905098651e-06, "loss": 0.021, "step": 26639 }, { "epoch": 4.89, "learning_rate": 6.950811968299981e-06, "loss": 0.0219, "step": 26640 }, { "epoch": 4.89, "learning_rate": 6.950104902420647e-06, "loss": 0.0197, "step": 26641 }, { "epoch": 4.89, "learning_rate": 6.949397853352407e-06, "loss": 0.0225, "step": 26642 }, { "epoch": 4.89, "learning_rate": 6.948690821099156e-06, "loss": 0.0207, "step": 26643 }, { "epoch": 4.89, "learning_rate": 6.947983805664795e-06, "loss": 0.0206, "step": 26644 }, { "epoch": 4.89, "learning_rate": 6.947276807053218e-06, "loss": 0.0211, "step": 26645 }, { "epoch": 4.89, "learning_rate": 6.946569825268325e-06, "loss": 0.023, "step": 26646 }, { "epoch": 4.89, "learning_rate": 6.945862860314003e-06, "loss": 0.0233, "step": 26647 }, { "epoch": 4.89, "learning_rate": 6.945155912194165e-06, "loss": 0.0217, "step": 26648 }, { "epoch": 4.89, "learning_rate": 6.944448980912696e-06, "loss": 0.0203, "step": 26649 }, { "epoch": 4.89, "learning_rate": 6.943742066473496e-06, "loss": 0.018, "step": 26650 }, { "epoch": 4.89, "learning_rate": 6.9430351688804585e-06, "loss": 0.022, "step": 26651 }, { "epoch": 4.89, "learning_rate": 6.942328288137485e-06, "loss": 0.0217, "step": 26652 }, { "epoch": 4.89, "learning_rate": 6.9416214242484716e-06, "loss": 0.0237, "step": 26653 }, { "epoch": 4.89, "learning_rate": 6.94091457721731e-06, "loss": 0.019, "step": 26654 }, { "epoch": 4.89, "learning_rate": 6.9402077470479026e-06, "loss": 0.0225, "step": 26655 }, { "epoch": 4.89, "learning_rate": 6.9395009337441355e-06, "loss": 0.0192, "step": 26656 }, { "epoch": 4.89, "learning_rate": 6.938794137309916e-06, "loss": 0.023, "step": 26657 }, { "epoch": 4.89, "learning_rate": 6.938087357749135e-06, "loss": 0.0181, "step": 26658 }, { "epoch": 4.89, "learning_rate": 6.937380595065686e-06, "loss": 0.0266, "step": 26659 }, { "epoch": 4.89, "learning_rate": 6.936673849263466e-06, "loss": 0.0197, "step": 26660 }, { "epoch": 4.89, "learning_rate": 6.935967120346372e-06, "loss": 0.0217, "step": 26661 }, { "epoch": 4.89, "learning_rate": 6.9352604083183004e-06, "loss": 0.0239, "step": 26662 }, { "epoch": 4.89, "learning_rate": 6.934553713183145e-06, "loss": 0.0214, "step": 26663 }, { "epoch": 4.89, "learning_rate": 6.933847034944804e-06, "loss": 0.0224, "step": 26664 }, { "epoch": 4.89, "learning_rate": 6.933140373607164e-06, "loss": 0.025, "step": 26665 }, { "epoch": 4.89, "learning_rate": 6.93243372917413e-06, "loss": 0.02, "step": 26666 }, { "epoch": 4.89, "learning_rate": 6.931727101649594e-06, "loss": 0.0191, "step": 26667 }, { "epoch": 4.89, "learning_rate": 6.931020491037448e-06, "loss": 0.0189, "step": 26668 }, { "epoch": 4.89, "learning_rate": 6.9303138973415875e-06, "loss": 0.0214, "step": 26669 }, { "epoch": 4.89, "learning_rate": 6.929607320565911e-06, "loss": 0.0181, "step": 26670 }, { "epoch": 4.89, "learning_rate": 6.928900760714311e-06, "loss": 0.0171, "step": 26671 }, { "epoch": 4.89, "learning_rate": 6.928194217790682e-06, "loss": 0.0222, "step": 26672 }, { "epoch": 4.89, "learning_rate": 6.927487691798915e-06, "loss": 0.0185, "step": 26673 }, { "epoch": 4.89, "learning_rate": 6.926781182742912e-06, "loss": 0.0208, "step": 26674 }, { "epoch": 4.89, "learning_rate": 6.926074690626564e-06, "loss": 0.0235, "step": 26675 }, { "epoch": 4.89, "learning_rate": 6.9253682154537625e-06, "loss": 0.0184, "step": 26676 }, { "epoch": 4.89, "learning_rate": 6.924661757228403e-06, "loss": 0.0185, "step": 26677 }, { "epoch": 4.89, "learning_rate": 6.923955315954378e-06, "loss": 0.0239, "step": 26678 }, { "epoch": 4.89, "learning_rate": 6.923248891635585e-06, "loss": 0.0163, "step": 26679 }, { "epoch": 4.89, "learning_rate": 6.922542484275918e-06, "loss": 0.0174, "step": 26680 }, { "epoch": 4.89, "learning_rate": 6.921836093879271e-06, "loss": 0.0174, "step": 26681 }, { "epoch": 4.89, "learning_rate": 6.921129720449526e-06, "loss": 0.0193, "step": 26682 }, { "epoch": 4.89, "learning_rate": 6.920423363990594e-06, "loss": 0.0173, "step": 26683 }, { "epoch": 4.89, "learning_rate": 6.9197170245063605e-06, "loss": 0.023, "step": 26684 }, { "epoch": 4.89, "learning_rate": 6.919010702000717e-06, "loss": 0.024, "step": 26685 }, { "epoch": 4.89, "learning_rate": 6.9183043964775595e-06, "loss": 0.0169, "step": 26686 }, { "epoch": 4.89, "learning_rate": 6.917598107940778e-06, "loss": 0.0222, "step": 26687 }, { "epoch": 4.89, "learning_rate": 6.91689183639427e-06, "loss": 0.0242, "step": 26688 }, { "epoch": 4.89, "learning_rate": 6.916185581841926e-06, "loss": 0.0218, "step": 26689 }, { "epoch": 4.89, "learning_rate": 6.915479344287642e-06, "loss": 0.0253, "step": 26690 }, { "epoch": 4.89, "learning_rate": 6.914773123735301e-06, "loss": 0.0191, "step": 26691 }, { "epoch": 4.89, "learning_rate": 6.914066920188809e-06, "loss": 0.0207, "step": 26692 }, { "epoch": 4.9, "learning_rate": 6.913360733652052e-06, "loss": 0.0191, "step": 26693 }, { "epoch": 4.9, "learning_rate": 6.9126545641289225e-06, "loss": 0.0189, "step": 26694 }, { "epoch": 4.9, "learning_rate": 6.911948411623311e-06, "loss": 0.0175, "step": 26695 }, { "epoch": 4.9, "learning_rate": 6.911242276139114e-06, "loss": 0.0263, "step": 26696 }, { "epoch": 4.9, "learning_rate": 6.9105361576802224e-06, "loss": 0.0238, "step": 26697 }, { "epoch": 4.9, "learning_rate": 6.909830056250527e-06, "loss": 0.0185, "step": 26698 }, { "epoch": 4.9, "learning_rate": 6.909123971853924e-06, "loss": 0.0205, "step": 26699 }, { "epoch": 4.9, "learning_rate": 6.908417904494295e-06, "loss": 0.0213, "step": 26700 }, { "epoch": 4.9, "learning_rate": 6.9077118541755425e-06, "loss": 0.0199, "step": 26701 }, { "epoch": 4.9, "learning_rate": 6.907005820901558e-06, "loss": 0.0218, "step": 26702 }, { "epoch": 4.9, "learning_rate": 6.906299804676226e-06, "loss": 0.0165, "step": 26703 }, { "epoch": 4.9, "learning_rate": 6.905593805503442e-06, "loss": 0.0222, "step": 26704 }, { "epoch": 4.9, "learning_rate": 6.9048878233870984e-06, "loss": 0.021, "step": 26705 }, { "epoch": 4.9, "learning_rate": 6.904181858331087e-06, "loss": 0.0207, "step": 26706 }, { "epoch": 4.9, "learning_rate": 6.903475910339297e-06, "loss": 0.0168, "step": 26707 }, { "epoch": 4.9, "learning_rate": 6.902769979415621e-06, "loss": 0.0191, "step": 26708 }, { "epoch": 4.9, "learning_rate": 6.902064065563945e-06, "loss": 0.023, "step": 26709 }, { "epoch": 4.9, "learning_rate": 6.901358168788167e-06, "loss": 0.0228, "step": 26710 }, { "epoch": 4.9, "learning_rate": 6.900652289092178e-06, "loss": 0.02, "step": 26711 }, { "epoch": 4.9, "learning_rate": 6.899946426479865e-06, "loss": 0.0225, "step": 26712 }, { "epoch": 4.9, "learning_rate": 6.899240580955118e-06, "loss": 0.0261, "step": 26713 }, { "epoch": 4.9, "learning_rate": 6.898534752521831e-06, "loss": 0.0206, "step": 26714 }, { "epoch": 4.9, "learning_rate": 6.897828941183893e-06, "loss": 0.0182, "step": 26715 }, { "epoch": 4.9, "learning_rate": 6.897123146945194e-06, "loss": 0.0205, "step": 26716 }, { "epoch": 4.9, "learning_rate": 6.896417369809622e-06, "loss": 0.0219, "step": 26717 }, { "epoch": 4.9, "learning_rate": 6.895711609781075e-06, "loss": 0.0214, "step": 26718 }, { "epoch": 4.9, "learning_rate": 6.895005866863439e-06, "loss": 0.0179, "step": 26719 }, { "epoch": 4.9, "learning_rate": 6.894300141060602e-06, "loss": 0.0268, "step": 26720 }, { "epoch": 4.9, "learning_rate": 6.893594432376453e-06, "loss": 0.0225, "step": 26721 }, { "epoch": 4.9, "learning_rate": 6.892888740814885e-06, "loss": 0.0155, "step": 26722 }, { "epoch": 4.9, "learning_rate": 6.892183066379788e-06, "loss": 0.0208, "step": 26723 }, { "epoch": 4.9, "learning_rate": 6.891477409075051e-06, "loss": 0.023, "step": 26724 }, { "epoch": 4.9, "learning_rate": 6.890771768904563e-06, "loss": 0.0213, "step": 26725 }, { "epoch": 4.9, "learning_rate": 6.890066145872212e-06, "loss": 0.0196, "step": 26726 }, { "epoch": 4.9, "learning_rate": 6.8893605399818905e-06, "loss": 0.0222, "step": 26727 }, { "epoch": 4.9, "learning_rate": 6.88865495123749e-06, "loss": 0.0184, "step": 26728 }, { "epoch": 4.9, "learning_rate": 6.887949379642893e-06, "loss": 0.0179, "step": 26729 }, { "epoch": 4.9, "learning_rate": 6.887243825201994e-06, "loss": 0.017, "step": 26730 }, { "epoch": 4.9, "learning_rate": 6.8865382879186755e-06, "loss": 0.021, "step": 26731 }, { "epoch": 4.9, "learning_rate": 6.885832767796834e-06, "loss": 0.0221, "step": 26732 }, { "epoch": 4.9, "learning_rate": 6.885127264840356e-06, "loss": 0.014, "step": 26733 }, { "epoch": 4.9, "learning_rate": 6.884421779053128e-06, "loss": 0.0252, "step": 26734 }, { "epoch": 4.9, "learning_rate": 6.883716310439039e-06, "loss": 0.0259, "step": 26735 }, { "epoch": 4.9, "learning_rate": 6.883010859001979e-06, "loss": 0.0206, "step": 26736 }, { "epoch": 4.9, "learning_rate": 6.882305424745841e-06, "loss": 0.0206, "step": 26737 }, { "epoch": 4.9, "learning_rate": 6.881600007674505e-06, "loss": 0.0196, "step": 26738 }, { "epoch": 4.9, "learning_rate": 6.880894607791864e-06, "loss": 0.0177, "step": 26739 }, { "epoch": 4.9, "learning_rate": 6.880189225101801e-06, "loss": 0.0198, "step": 26740 }, { "epoch": 4.9, "learning_rate": 6.879483859608211e-06, "loss": 0.0179, "step": 26741 }, { "epoch": 4.9, "learning_rate": 6.87877851131498e-06, "loss": 0.0195, "step": 26742 }, { "epoch": 4.9, "learning_rate": 6.8780731802259934e-06, "loss": 0.017, "step": 26743 }, { "epoch": 4.9, "learning_rate": 6.877367866345139e-06, "loss": 0.0197, "step": 26744 }, { "epoch": 4.9, "learning_rate": 6.876662569676309e-06, "loss": 0.0187, "step": 26745 }, { "epoch": 4.9, "learning_rate": 6.8759572902233895e-06, "loss": 0.0143, "step": 26746 }, { "epoch": 4.91, "learning_rate": 6.875252027990264e-06, "loss": 0.0216, "step": 26747 }, { "epoch": 4.91, "learning_rate": 6.87454678298082e-06, "loss": 0.0262, "step": 26748 }, { "epoch": 4.91, "learning_rate": 6.873841555198952e-06, "loss": 0.0202, "step": 26749 }, { "epoch": 4.91, "learning_rate": 6.873136344648541e-06, "loss": 0.0176, "step": 26750 }, { "epoch": 4.91, "learning_rate": 6.8724311513334765e-06, "loss": 0.0179, "step": 26751 }, { "epoch": 4.91, "learning_rate": 6.871725975257643e-06, "loss": 0.0213, "step": 26752 }, { "epoch": 4.91, "learning_rate": 6.871020816424929e-06, "loss": 0.019, "step": 26753 }, { "epoch": 4.91, "learning_rate": 6.870315674839223e-06, "loss": 0.0166, "step": 26754 }, { "epoch": 4.91, "learning_rate": 6.869610550504412e-06, "loss": 0.0175, "step": 26755 }, { "epoch": 4.91, "learning_rate": 6.868905443424379e-06, "loss": 0.0181, "step": 26756 }, { "epoch": 4.91, "learning_rate": 6.868200353603008e-06, "loss": 0.022, "step": 26757 }, { "epoch": 4.91, "learning_rate": 6.867495281044197e-06, "loss": 0.0208, "step": 26758 }, { "epoch": 4.91, "learning_rate": 6.866790225751823e-06, "loss": 0.0155, "step": 26759 }, { "epoch": 4.91, "learning_rate": 6.866085187729776e-06, "loss": 0.019, "step": 26760 }, { "epoch": 4.91, "learning_rate": 6.865380166981939e-06, "loss": 0.02, "step": 26761 }, { "epoch": 4.91, "learning_rate": 6.864675163512198e-06, "loss": 0.0221, "step": 26762 }, { "epoch": 4.91, "learning_rate": 6.863970177324444e-06, "loss": 0.0209, "step": 26763 }, { "epoch": 4.91, "learning_rate": 6.863265208422562e-06, "loss": 0.0241, "step": 26764 }, { "epoch": 4.91, "learning_rate": 6.8625602568104335e-06, "loss": 0.0197, "step": 26765 }, { "epoch": 4.91, "learning_rate": 6.861855322491943e-06, "loss": 0.0161, "step": 26766 }, { "epoch": 4.91, "learning_rate": 6.861150405470982e-06, "loss": 0.0203, "step": 26767 }, { "epoch": 4.91, "learning_rate": 6.860445505751434e-06, "loss": 0.0205, "step": 26768 }, { "epoch": 4.91, "learning_rate": 6.859740623337183e-06, "loss": 0.0183, "step": 26769 }, { "epoch": 4.91, "learning_rate": 6.859035758232115e-06, "loss": 0.0207, "step": 26770 }, { "epoch": 4.91, "learning_rate": 6.858330910440116e-06, "loss": 0.0212, "step": 26771 }, { "epoch": 4.91, "learning_rate": 6.857626079965072e-06, "loss": 0.0191, "step": 26772 }, { "epoch": 4.91, "learning_rate": 6.856921266810867e-06, "loss": 0.0224, "step": 26773 }, { "epoch": 4.91, "learning_rate": 6.8562164709813846e-06, "loss": 0.0216, "step": 26774 }, { "epoch": 4.91, "learning_rate": 6.8555116924805086e-06, "loss": 0.0202, "step": 26775 }, { "epoch": 4.91, "learning_rate": 6.854806931312128e-06, "loss": 0.022, "step": 26776 }, { "epoch": 4.91, "learning_rate": 6.854102187480125e-06, "loss": 0.0234, "step": 26777 }, { "epoch": 4.91, "learning_rate": 6.853397460988385e-06, "loss": 0.0206, "step": 26778 }, { "epoch": 4.91, "learning_rate": 6.852692751840788e-06, "loss": 0.0179, "step": 26779 }, { "epoch": 4.91, "learning_rate": 6.851988060041226e-06, "loss": 0.0221, "step": 26780 }, { "epoch": 4.91, "learning_rate": 6.851283385593579e-06, "loss": 0.022, "step": 26781 }, { "epoch": 4.91, "learning_rate": 6.850578728501736e-06, "loss": 0.0204, "step": 26782 }, { "epoch": 4.91, "learning_rate": 6.849874088769572e-06, "loss": 0.0207, "step": 26783 }, { "epoch": 4.91, "learning_rate": 6.849169466400975e-06, "loss": 0.0264, "step": 26784 }, { "epoch": 4.91, "learning_rate": 6.848464861399831e-06, "loss": 0.021, "step": 26785 }, { "epoch": 4.91, "learning_rate": 6.847760273770023e-06, "loss": 0.0208, "step": 26786 }, { "epoch": 4.91, "learning_rate": 6.847055703515433e-06, "loss": 0.0228, "step": 26787 }, { "epoch": 4.91, "learning_rate": 6.846351150639945e-06, "loss": 0.0201, "step": 26788 }, { "epoch": 4.91, "learning_rate": 6.845646615147445e-06, "loss": 0.0178, "step": 26789 }, { "epoch": 4.91, "learning_rate": 6.8449420970418155e-06, "loss": 0.0217, "step": 26790 }, { "epoch": 4.91, "learning_rate": 6.844237596326941e-06, "loss": 0.0153, "step": 26791 }, { "epoch": 4.91, "learning_rate": 6.843533113006699e-06, "loss": 0.02, "step": 26792 }, { "epoch": 4.91, "learning_rate": 6.842828647084977e-06, "loss": 0.0195, "step": 26793 }, { "epoch": 4.91, "learning_rate": 6.842124198565659e-06, "loss": 0.0229, "step": 26794 }, { "epoch": 4.91, "learning_rate": 6.841419767452625e-06, "loss": 0.0217, "step": 26795 }, { "epoch": 4.91, "learning_rate": 6.84071535374976e-06, "loss": 0.0182, "step": 26796 }, { "epoch": 4.91, "learning_rate": 6.840010957460943e-06, "loss": 0.0211, "step": 26797 }, { "epoch": 4.91, "learning_rate": 6.839306578590063e-06, "loss": 0.0244, "step": 26798 }, { "epoch": 4.91, "learning_rate": 6.838602217141e-06, "loss": 0.0201, "step": 26799 }, { "epoch": 4.91, "learning_rate": 6.837897873117636e-06, "loss": 0.016, "step": 26800 }, { "epoch": 4.91, "learning_rate": 6.837193546523846e-06, "loss": 0.025, "step": 26801 }, { "epoch": 4.92, "learning_rate": 6.836489237363527e-06, "loss": 0.0187, "step": 26802 }, { "epoch": 4.92, "learning_rate": 6.8357849456405515e-06, "loss": 0.0179, "step": 26803 }, { "epoch": 4.92, "learning_rate": 6.835080671358802e-06, "loss": 0.0211, "step": 26804 }, { "epoch": 4.92, "learning_rate": 6.834376414522164e-06, "loss": 0.0189, "step": 26805 }, { "epoch": 4.92, "learning_rate": 6.833672175134514e-06, "loss": 0.0201, "step": 26806 }, { "epoch": 4.92, "learning_rate": 6.83296795319974e-06, "loss": 0.0206, "step": 26807 }, { "epoch": 4.92, "learning_rate": 6.83226374872172e-06, "loss": 0.0217, "step": 26808 }, { "epoch": 4.92, "learning_rate": 6.831559561704339e-06, "loss": 0.0205, "step": 26809 }, { "epoch": 4.92, "learning_rate": 6.83085539215147e-06, "loss": 0.0216, "step": 26810 }, { "epoch": 4.92, "learning_rate": 6.830151240067007e-06, "loss": 0.0141, "step": 26811 }, { "epoch": 4.92, "learning_rate": 6.8294471054548206e-06, "loss": 0.0199, "step": 26812 }, { "epoch": 4.92, "learning_rate": 6.828742988318797e-06, "loss": 0.0164, "step": 26813 }, { "epoch": 4.92, "learning_rate": 6.828038888662817e-06, "loss": 0.0219, "step": 26814 }, { "epoch": 4.92, "learning_rate": 6.8273348064907574e-06, "loss": 0.024, "step": 26815 }, { "epoch": 4.92, "learning_rate": 6.826630741806506e-06, "loss": 0.0206, "step": 26816 }, { "epoch": 4.92, "learning_rate": 6.82592669461394e-06, "loss": 0.0193, "step": 26817 }, { "epoch": 4.92, "learning_rate": 6.825222664916943e-06, "loss": 0.0151, "step": 26818 }, { "epoch": 4.92, "learning_rate": 6.824518652719386e-06, "loss": 0.0184, "step": 26819 }, { "epoch": 4.92, "learning_rate": 6.823814658025163e-06, "loss": 0.0182, "step": 26820 }, { "epoch": 4.92, "learning_rate": 6.8231106808381464e-06, "loss": 0.0187, "step": 26821 }, { "epoch": 4.92, "learning_rate": 6.822406721162218e-06, "loss": 0.0206, "step": 26822 }, { "epoch": 4.92, "learning_rate": 6.821702779001256e-06, "loss": 0.0187, "step": 26823 }, { "epoch": 4.92, "learning_rate": 6.820998854359144e-06, "loss": 0.0212, "step": 26824 }, { "epoch": 4.92, "learning_rate": 6.820294947239763e-06, "loss": 0.0196, "step": 26825 }, { "epoch": 4.92, "learning_rate": 6.819591057646988e-06, "loss": 0.0197, "step": 26826 }, { "epoch": 4.92, "learning_rate": 6.8188871855847065e-06, "loss": 0.0201, "step": 26827 }, { "epoch": 4.92, "learning_rate": 6.818183331056784e-06, "loss": 0.0163, "step": 26828 }, { "epoch": 4.92, "learning_rate": 6.817479494067119e-06, "loss": 0.0194, "step": 26829 }, { "epoch": 4.92, "learning_rate": 6.816775674619578e-06, "loss": 0.0173, "step": 26830 }, { "epoch": 4.92, "learning_rate": 6.816071872718043e-06, "loss": 0.0214, "step": 26831 }, { "epoch": 4.92, "learning_rate": 6.815368088366393e-06, "loss": 0.023, "step": 26832 }, { "epoch": 4.92, "learning_rate": 6.814664321568511e-06, "loss": 0.0216, "step": 26833 }, { "epoch": 4.92, "learning_rate": 6.813960572328274e-06, "loss": 0.0203, "step": 26834 }, { "epoch": 4.92, "learning_rate": 6.813256840649564e-06, "loss": 0.0178, "step": 26835 }, { "epoch": 4.92, "learning_rate": 6.812553126536253e-06, "loss": 0.0276, "step": 26836 }, { "epoch": 4.92, "learning_rate": 6.811849429992221e-06, "loss": 0.0222, "step": 26837 }, { "epoch": 4.92, "learning_rate": 6.811145751021355e-06, "loss": 0.0187, "step": 26838 }, { "epoch": 4.92, "learning_rate": 6.810442089627527e-06, "loss": 0.0202, "step": 26839 }, { "epoch": 4.92, "learning_rate": 6.809738445814617e-06, "loss": 0.0207, "step": 26840 }, { "epoch": 4.92, "learning_rate": 6.8090348195865e-06, "loss": 0.0192, "step": 26841 }, { "epoch": 4.92, "learning_rate": 6.808331210947061e-06, "loss": 0.0201, "step": 26842 }, { "epoch": 4.92, "learning_rate": 6.807627619900174e-06, "loss": 0.0212, "step": 26843 }, { "epoch": 4.92, "learning_rate": 6.806924046449723e-06, "loss": 0.0194, "step": 26844 }, { "epoch": 4.92, "learning_rate": 6.806220490599573e-06, "loss": 0.0227, "step": 26845 }, { "epoch": 4.92, "learning_rate": 6.805516952353616e-06, "loss": 0.0211, "step": 26846 }, { "epoch": 4.92, "learning_rate": 6.804813431715726e-06, "loss": 0.0183, "step": 26847 }, { "epoch": 4.92, "learning_rate": 6.804109928689777e-06, "loss": 0.0203, "step": 26848 }, { "epoch": 4.92, "learning_rate": 6.803406443279648e-06, "loss": 0.0225, "step": 26849 }, { "epoch": 4.92, "learning_rate": 6.802702975489217e-06, "loss": 0.0217, "step": 26850 }, { "epoch": 4.92, "learning_rate": 6.801999525322362e-06, "loss": 0.0159, "step": 26851 }, { "epoch": 4.92, "learning_rate": 6.801296092782963e-06, "loss": 0.0159, "step": 26852 }, { "epoch": 4.92, "learning_rate": 6.800592677874895e-06, "loss": 0.0217, "step": 26853 }, { "epoch": 4.92, "learning_rate": 6.799889280602031e-06, "loss": 0.0225, "step": 26854 }, { "epoch": 4.92, "learning_rate": 6.7991859009682545e-06, "loss": 0.0171, "step": 26855 }, { "epoch": 4.92, "learning_rate": 6.798482538977445e-06, "loss": 0.0165, "step": 26856 }, { "epoch": 4.93, "learning_rate": 6.797779194633469e-06, "loss": 0.0179, "step": 26857 }, { "epoch": 4.93, "learning_rate": 6.797075867940211e-06, "loss": 0.0131, "step": 26858 }, { "epoch": 4.93, "learning_rate": 6.796372558901545e-06, "loss": 0.0233, "step": 26859 }, { "epoch": 4.93, "learning_rate": 6.795669267521348e-06, "loss": 0.0192, "step": 26860 }, { "epoch": 4.93, "learning_rate": 6.7949659938035e-06, "loss": 0.0212, "step": 26861 }, { "epoch": 4.93, "learning_rate": 6.794262737751874e-06, "loss": 0.0226, "step": 26862 }, { "epoch": 4.93, "learning_rate": 6.793559499370342e-06, "loss": 0.0178, "step": 26863 }, { "epoch": 4.93, "learning_rate": 6.7928562786627895e-06, "loss": 0.025, "step": 26864 }, { "epoch": 4.93, "learning_rate": 6.79215307563309e-06, "loss": 0.0202, "step": 26865 }, { "epoch": 4.93, "learning_rate": 6.791449890285116e-06, "loss": 0.0194, "step": 26866 }, { "epoch": 4.93, "learning_rate": 6.7907467226227445e-06, "loss": 0.02, "step": 26867 }, { "epoch": 4.93, "learning_rate": 6.790043572649851e-06, "loss": 0.0189, "step": 26868 }, { "epoch": 4.93, "learning_rate": 6.789340440370313e-06, "loss": 0.0163, "step": 26869 }, { "epoch": 4.93, "learning_rate": 6.788637325788008e-06, "loss": 0.0197, "step": 26870 }, { "epoch": 4.93, "learning_rate": 6.78793422890681e-06, "loss": 0.0174, "step": 26871 }, { "epoch": 4.93, "learning_rate": 6.787231149730589e-06, "loss": 0.0214, "step": 26872 }, { "epoch": 4.93, "learning_rate": 6.786528088263227e-06, "loss": 0.0239, "step": 26873 }, { "epoch": 4.93, "learning_rate": 6.7858250445086e-06, "loss": 0.0237, "step": 26874 }, { "epoch": 4.93, "learning_rate": 6.78512201847058e-06, "loss": 0.021, "step": 26875 }, { "epoch": 4.93, "learning_rate": 6.78441901015304e-06, "loss": 0.022, "step": 26876 }, { "epoch": 4.93, "learning_rate": 6.7837160195598605e-06, "loss": 0.0199, "step": 26877 }, { "epoch": 4.93, "learning_rate": 6.783013046694913e-06, "loss": 0.02, "step": 26878 }, { "epoch": 4.93, "learning_rate": 6.782310091562073e-06, "loss": 0.0182, "step": 26879 }, { "epoch": 4.93, "learning_rate": 6.781607154165217e-06, "loss": 0.0216, "step": 26880 }, { "epoch": 4.93, "learning_rate": 6.780904234508212e-06, "loss": 0.0211, "step": 26881 }, { "epoch": 4.93, "learning_rate": 6.780201332594944e-06, "loss": 0.0211, "step": 26882 }, { "epoch": 4.93, "learning_rate": 6.77949844842928e-06, "loss": 0.019, "step": 26883 }, { "epoch": 4.93, "learning_rate": 6.778795582015096e-06, "loss": 0.019, "step": 26884 }, { "epoch": 4.93, "learning_rate": 6.778092733356264e-06, "loss": 0.0207, "step": 26885 }, { "epoch": 4.93, "learning_rate": 6.777389902456662e-06, "loss": 0.0224, "step": 26886 }, { "epoch": 4.93, "learning_rate": 6.776687089320164e-06, "loss": 0.0194, "step": 26887 }, { "epoch": 4.93, "learning_rate": 6.77598429395064e-06, "loss": 0.0238, "step": 26888 }, { "epoch": 4.93, "learning_rate": 6.775281516351971e-06, "loss": 0.0196, "step": 26889 }, { "epoch": 4.93, "learning_rate": 6.7745787565280165e-06, "loss": 0.0174, "step": 26890 }, { "epoch": 4.93, "learning_rate": 6.773876014482667e-06, "loss": 0.0214, "step": 26891 }, { "epoch": 4.93, "learning_rate": 6.7731732902197865e-06, "loss": 0.0198, "step": 26892 }, { "epoch": 4.93, "learning_rate": 6.772470583743252e-06, "loss": 0.0184, "step": 26893 }, { "epoch": 4.93, "learning_rate": 6.77176789505693e-06, "loss": 0.023, "step": 26894 }, { "epoch": 4.93, "learning_rate": 6.771065224164703e-06, "loss": 0.0209, "step": 26895 }, { "epoch": 4.93, "learning_rate": 6.770362571070439e-06, "loss": 0.0238, "step": 26896 }, { "epoch": 4.93, "learning_rate": 6.769659935778013e-06, "loss": 0.0228, "step": 26897 }, { "epoch": 4.93, "learning_rate": 6.768957318291296e-06, "loss": 0.0227, "step": 26898 }, { "epoch": 4.93, "learning_rate": 6.768254718614162e-06, "loss": 0.024, "step": 26899 }, { "epoch": 4.93, "learning_rate": 6.767552136750487e-06, "loss": 0.0209, "step": 26900 }, { "epoch": 4.93, "learning_rate": 6.7668495727041385e-06, "loss": 0.0219, "step": 26901 }, { "epoch": 4.93, "learning_rate": 6.766147026478991e-06, "loss": 0.0177, "step": 26902 }, { "epoch": 4.93, "learning_rate": 6.765444498078914e-06, "loss": 0.0161, "step": 26903 }, { "epoch": 4.93, "learning_rate": 6.764741987507785e-06, "loss": 0.021, "step": 26904 }, { "epoch": 4.93, "learning_rate": 6.764039494769473e-06, "loss": 0.0174, "step": 26905 }, { "epoch": 4.93, "learning_rate": 6.763337019867853e-06, "loss": 0.0254, "step": 26906 }, { "epoch": 4.93, "learning_rate": 6.762634562806791e-06, "loss": 0.0233, "step": 26907 }, { "epoch": 4.93, "learning_rate": 6.761932123590167e-06, "loss": 0.0238, "step": 26908 }, { "epoch": 4.93, "learning_rate": 6.76122970222185e-06, "loss": 0.0221, "step": 26909 }, { "epoch": 4.93, "learning_rate": 6.76052729870571e-06, "loss": 0.0292, "step": 26910 }, { "epoch": 4.94, "learning_rate": 6.759824913045619e-06, "loss": 0.0226, "step": 26911 }, { "epoch": 4.94, "learning_rate": 6.759122545245446e-06, "loss": 0.0214, "step": 26912 }, { "epoch": 4.94, "learning_rate": 6.758420195309068e-06, "loss": 0.0171, "step": 26913 }, { "epoch": 4.94, "learning_rate": 6.757717863240354e-06, "loss": 0.0213, "step": 26914 }, { "epoch": 4.94, "learning_rate": 6.757015549043174e-06, "loss": 0.0187, "step": 26915 }, { "epoch": 4.94, "learning_rate": 6.7563132527213995e-06, "loss": 0.0198, "step": 26916 }, { "epoch": 4.94, "learning_rate": 6.755610974278903e-06, "loss": 0.0196, "step": 26917 }, { "epoch": 4.94, "learning_rate": 6.754908713719559e-06, "loss": 0.0248, "step": 26918 }, { "epoch": 4.94, "learning_rate": 6.754206471047231e-06, "loss": 0.0192, "step": 26919 }, { "epoch": 4.94, "learning_rate": 6.753504246265793e-06, "loss": 0.0183, "step": 26920 }, { "epoch": 4.94, "learning_rate": 6.7528020393791136e-06, "loss": 0.0184, "step": 26921 }, { "epoch": 4.94, "learning_rate": 6.752099850391068e-06, "loss": 0.0215, "step": 26922 }, { "epoch": 4.94, "learning_rate": 6.751397679305522e-06, "loss": 0.0158, "step": 26923 }, { "epoch": 4.94, "learning_rate": 6.750695526126351e-06, "loss": 0.0201, "step": 26924 }, { "epoch": 4.94, "learning_rate": 6.749993390857419e-06, "loss": 0.0223, "step": 26925 }, { "epoch": 4.94, "learning_rate": 6.749291273502601e-06, "loss": 0.0203, "step": 26926 }, { "epoch": 4.94, "learning_rate": 6.7485891740657695e-06, "loss": 0.0178, "step": 26927 }, { "epoch": 4.94, "learning_rate": 6.747887092550787e-06, "loss": 0.02, "step": 26928 }, { "epoch": 4.94, "learning_rate": 6.7471850289615246e-06, "loss": 0.0206, "step": 26929 }, { "epoch": 4.94, "learning_rate": 6.746482983301858e-06, "loss": 0.0193, "step": 26930 }, { "epoch": 4.94, "learning_rate": 6.745780955575654e-06, "loss": 0.0211, "step": 26931 }, { "epoch": 4.94, "learning_rate": 6.74507894578678e-06, "loss": 0.0207, "step": 26932 }, { "epoch": 4.94, "learning_rate": 6.744376953939107e-06, "loss": 0.0262, "step": 26933 }, { "epoch": 4.94, "learning_rate": 6.7436749800365035e-06, "loss": 0.023, "step": 26934 }, { "epoch": 4.94, "learning_rate": 6.7429730240828416e-06, "loss": 0.0206, "step": 26935 }, { "epoch": 4.94, "learning_rate": 6.74227108608199e-06, "loss": 0.0247, "step": 26936 }, { "epoch": 4.94, "learning_rate": 6.741569166037816e-06, "loss": 0.0179, "step": 26937 }, { "epoch": 4.94, "learning_rate": 6.740867263954185e-06, "loss": 0.0183, "step": 26938 }, { "epoch": 4.94, "learning_rate": 6.740165379834973e-06, "loss": 0.0255, "step": 26939 }, { "epoch": 4.94, "learning_rate": 6.739463513684046e-06, "loss": 0.0212, "step": 26940 }, { "epoch": 4.94, "learning_rate": 6.738761665505272e-06, "loss": 0.0244, "step": 26941 }, { "epoch": 4.94, "learning_rate": 6.738059835302523e-06, "loss": 0.0155, "step": 26942 }, { "epoch": 4.94, "learning_rate": 6.7373580230796575e-06, "loss": 0.0174, "step": 26943 }, { "epoch": 4.94, "learning_rate": 6.736656228840556e-06, "loss": 0.0142, "step": 26944 }, { "epoch": 4.94, "learning_rate": 6.735954452589083e-06, "loss": 0.0269, "step": 26945 }, { "epoch": 4.94, "learning_rate": 6.735252694329104e-06, "loss": 0.019, "step": 26946 }, { "epoch": 4.94, "learning_rate": 6.734550954064487e-06, "loss": 0.0187, "step": 26947 }, { "epoch": 4.94, "learning_rate": 6.733849231799103e-06, "loss": 0.0179, "step": 26948 }, { "epoch": 4.94, "learning_rate": 6.733147527536818e-06, "loss": 0.022, "step": 26949 }, { "epoch": 4.94, "learning_rate": 6.732445841281501e-06, "loss": 0.0196, "step": 26950 }, { "epoch": 4.94, "learning_rate": 6.731744173037018e-06, "loss": 0.0224, "step": 26951 }, { "epoch": 4.94, "learning_rate": 6.731042522807239e-06, "loss": 0.0219, "step": 26952 }, { "epoch": 4.94, "learning_rate": 6.73034089059603e-06, "loss": 0.0208, "step": 26953 }, { "epoch": 4.94, "learning_rate": 6.729639276407261e-06, "loss": 0.0164, "step": 26954 }, { "epoch": 4.94, "learning_rate": 6.728937680244794e-06, "loss": 0.0212, "step": 26955 }, { "epoch": 4.94, "learning_rate": 6.728236102112498e-06, "loss": 0.022, "step": 26956 }, { "epoch": 4.94, "learning_rate": 6.7275345420142426e-06, "loss": 0.0201, "step": 26957 }, { "epoch": 4.94, "learning_rate": 6.726832999953894e-06, "loss": 0.019, "step": 26958 }, { "epoch": 4.94, "learning_rate": 6.726131475935318e-06, "loss": 0.025, "step": 26959 }, { "epoch": 4.94, "learning_rate": 6.72542996996238e-06, "loss": 0.0211, "step": 26960 }, { "epoch": 4.94, "learning_rate": 6.724728482038952e-06, "loss": 0.0151, "step": 26961 }, { "epoch": 4.94, "learning_rate": 6.724027012168896e-06, "loss": 0.021, "step": 26962 }, { "epoch": 4.94, "learning_rate": 6.723325560356082e-06, "loss": 0.0214, "step": 26963 }, { "epoch": 4.94, "learning_rate": 6.7226241266043735e-06, "loss": 0.0218, "step": 26964 }, { "epoch": 4.94, "learning_rate": 6.721922710917634e-06, "loss": 0.0169, "step": 26965 }, { "epoch": 4.95, "learning_rate": 6.721221313299738e-06, "loss": 0.0203, "step": 26966 }, { "epoch": 4.95, "learning_rate": 6.720519933754543e-06, "loss": 0.0236, "step": 26967 }, { "epoch": 4.95, "learning_rate": 6.719818572285921e-06, "loss": 0.021, "step": 26968 }, { "epoch": 4.95, "learning_rate": 6.719117228897734e-06, "loss": 0.0209, "step": 26969 }, { "epoch": 4.95, "learning_rate": 6.718415903593851e-06, "loss": 0.0196, "step": 26970 }, { "epoch": 4.95, "learning_rate": 6.717714596378138e-06, "loss": 0.0202, "step": 26971 }, { "epoch": 4.95, "learning_rate": 6.71701330725446e-06, "loss": 0.0202, "step": 26972 }, { "epoch": 4.95, "learning_rate": 6.716312036226676e-06, "loss": 0.0178, "step": 26973 }, { "epoch": 4.95, "learning_rate": 6.715610783298662e-06, "loss": 0.0199, "step": 26974 }, { "epoch": 4.95, "learning_rate": 6.714909548474277e-06, "loss": 0.0186, "step": 26975 }, { "epoch": 4.95, "learning_rate": 6.714208331757387e-06, "loss": 0.0187, "step": 26976 }, { "epoch": 4.95, "learning_rate": 6.7135071331518575e-06, "loss": 0.0185, "step": 26977 }, { "epoch": 4.95, "learning_rate": 6.712805952661552e-06, "loss": 0.0199, "step": 26978 }, { "epoch": 4.95, "learning_rate": 6.7121047902903395e-06, "loss": 0.0187, "step": 26979 }, { "epoch": 4.95, "learning_rate": 6.711403646042083e-06, "loss": 0.0225, "step": 26980 }, { "epoch": 4.95, "learning_rate": 6.710702519920648e-06, "loss": 0.0223, "step": 26981 }, { "epoch": 4.95, "learning_rate": 6.710001411929891e-06, "loss": 0.0235, "step": 26982 }, { "epoch": 4.95, "learning_rate": 6.70930032207369e-06, "loss": 0.0194, "step": 26983 }, { "epoch": 4.95, "learning_rate": 6.7085992503559004e-06, "loss": 0.0226, "step": 26984 }, { "epoch": 4.95, "learning_rate": 6.707898196780389e-06, "loss": 0.0231, "step": 26985 }, { "epoch": 4.95, "learning_rate": 6.70719716135102e-06, "loss": 0.0186, "step": 26986 }, { "epoch": 4.95, "learning_rate": 6.706496144071654e-06, "loss": 0.0211, "step": 26987 }, { "epoch": 4.95, "learning_rate": 6.705795144946163e-06, "loss": 0.0196, "step": 26988 }, { "epoch": 4.95, "learning_rate": 6.705094163978407e-06, "loss": 0.018, "step": 26989 }, { "epoch": 4.95, "learning_rate": 6.704393201172248e-06, "loss": 0.0182, "step": 26990 }, { "epoch": 4.95, "learning_rate": 6.703692256531545e-06, "loss": 0.0187, "step": 26991 }, { "epoch": 4.95, "learning_rate": 6.7029913300601755e-06, "loss": 0.0191, "step": 26992 }, { "epoch": 4.95, "learning_rate": 6.702290421761993e-06, "loss": 0.0216, "step": 26993 }, { "epoch": 4.95, "learning_rate": 6.701589531640861e-06, "loss": 0.0229, "step": 26994 }, { "epoch": 4.95, "learning_rate": 6.700888659700646e-06, "loss": 0.0195, "step": 26995 }, { "epoch": 4.95, "learning_rate": 6.700187805945208e-06, "loss": 0.0177, "step": 26996 }, { "epoch": 4.95, "learning_rate": 6.699486970378414e-06, "loss": 0.018, "step": 26997 }, { "epoch": 4.95, "learning_rate": 6.698786153004126e-06, "loss": 0.0192, "step": 26998 }, { "epoch": 4.95, "learning_rate": 6.698085353826207e-06, "loss": 0.0208, "step": 26999 }, { "epoch": 4.95, "learning_rate": 6.697384572848512e-06, "loss": 0.0223, "step": 27000 }, { "epoch": 4.95, "learning_rate": 6.696683810074918e-06, "loss": 0.0187, "step": 27001 }, { "epoch": 4.95, "learning_rate": 6.6959830655092775e-06, "loss": 0.0173, "step": 27002 }, { "epoch": 4.95, "learning_rate": 6.6952823391554565e-06, "loss": 0.0176, "step": 27003 }, { "epoch": 4.95, "learning_rate": 6.6945816310173136e-06, "loss": 0.0153, "step": 27004 }, { "epoch": 4.95, "learning_rate": 6.693880941098718e-06, "loss": 0.0238, "step": 27005 }, { "epoch": 4.95, "learning_rate": 6.6931802694035265e-06, "loss": 0.0204, "step": 27006 }, { "epoch": 4.95, "learning_rate": 6.692479615935606e-06, "loss": 0.0204, "step": 27007 }, { "epoch": 4.95, "learning_rate": 6.691778980698812e-06, "loss": 0.0196, "step": 27008 }, { "epoch": 4.95, "learning_rate": 6.691078363697007e-06, "loss": 0.0201, "step": 27009 }, { "epoch": 4.95, "learning_rate": 6.69037776493406e-06, "loss": 0.0197, "step": 27010 }, { "epoch": 4.95, "learning_rate": 6.689677184413827e-06, "loss": 0.0202, "step": 27011 }, { "epoch": 4.95, "learning_rate": 6.688976622140173e-06, "loss": 0.0176, "step": 27012 }, { "epoch": 4.95, "learning_rate": 6.688276078116953e-06, "loss": 0.0174, "step": 27013 }, { "epoch": 4.95, "learning_rate": 6.687575552348036e-06, "loss": 0.0166, "step": 27014 }, { "epoch": 4.95, "learning_rate": 6.68687504483728e-06, "loss": 0.0236, "step": 27015 }, { "epoch": 4.95, "learning_rate": 6.686174555588547e-06, "loss": 0.0243, "step": 27016 }, { "epoch": 4.95, "learning_rate": 6.6854740846056965e-06, "loss": 0.0192, "step": 27017 }, { "epoch": 4.95, "learning_rate": 6.684773631892587e-06, "loss": 0.0172, "step": 27018 }, { "epoch": 4.95, "learning_rate": 6.684073197453088e-06, "loss": 0.0182, "step": 27019 }, { "epoch": 4.96, "learning_rate": 6.683372781291053e-06, "loss": 0.0235, "step": 27020 }, { "epoch": 4.96, "learning_rate": 6.682672383410347e-06, "loss": 0.022, "step": 27021 }, { "epoch": 4.96, "learning_rate": 6.681972003814825e-06, "loss": 0.0197, "step": 27022 }, { "epoch": 4.96, "learning_rate": 6.681271642508354e-06, "loss": 0.0199, "step": 27023 }, { "epoch": 4.96, "learning_rate": 6.6805712994947894e-06, "loss": 0.0171, "step": 27024 }, { "epoch": 4.96, "learning_rate": 6.6798709747779965e-06, "loss": 0.0183, "step": 27025 }, { "epoch": 4.96, "learning_rate": 6.679170668361827e-06, "loss": 0.0181, "step": 27026 }, { "epoch": 4.96, "learning_rate": 6.67847038025015e-06, "loss": 0.0213, "step": 27027 }, { "epoch": 4.96, "learning_rate": 6.677770110446825e-06, "loss": 0.0161, "step": 27028 }, { "epoch": 4.96, "learning_rate": 6.677069858955707e-06, "loss": 0.0202, "step": 27029 }, { "epoch": 4.96, "learning_rate": 6.6763696257806565e-06, "loss": 0.0173, "step": 27030 }, { "epoch": 4.96, "learning_rate": 6.6756694109255335e-06, "loss": 0.022, "step": 27031 }, { "epoch": 4.96, "learning_rate": 6.674969214394201e-06, "loss": 0.0161, "step": 27032 }, { "epoch": 4.96, "learning_rate": 6.674269036190515e-06, "loss": 0.0158, "step": 27033 }, { "epoch": 4.96, "learning_rate": 6.67356887631834e-06, "loss": 0.0178, "step": 27034 }, { "epoch": 4.96, "learning_rate": 6.672868734781521e-06, "loss": 0.0202, "step": 27035 }, { "epoch": 4.96, "learning_rate": 6.672168611583938e-06, "loss": 0.019, "step": 27036 }, { "epoch": 4.96, "learning_rate": 6.671468506729435e-06, "loss": 0.0166, "step": 27037 }, { "epoch": 4.96, "learning_rate": 6.670768420221876e-06, "loss": 0.0159, "step": 27038 }, { "epoch": 4.96, "learning_rate": 6.67006835206512e-06, "loss": 0.0169, "step": 27039 }, { "epoch": 4.96, "learning_rate": 6.669368302263024e-06, "loss": 0.025, "step": 27040 }, { "epoch": 4.96, "learning_rate": 6.668668270819448e-06, "loss": 0.0142, "step": 27041 }, { "epoch": 4.96, "learning_rate": 6.667968257738252e-06, "loss": 0.014, "step": 27042 }, { "epoch": 4.96, "learning_rate": 6.6672682630232945e-06, "loss": 0.0162, "step": 27043 }, { "epoch": 4.96, "learning_rate": 6.666568286678426e-06, "loss": 0.0171, "step": 27044 }, { "epoch": 4.96, "learning_rate": 6.6658683287075185e-06, "loss": 0.0177, "step": 27045 }, { "epoch": 4.96, "learning_rate": 6.66516838911442e-06, "loss": 0.0205, "step": 27046 }, { "epoch": 4.96, "learning_rate": 6.664468467902992e-06, "loss": 0.0233, "step": 27047 }, { "epoch": 4.96, "learning_rate": 6.663768565077091e-06, "loss": 0.0175, "step": 27048 }, { "epoch": 4.96, "learning_rate": 6.663068680640574e-06, "loss": 0.0152, "step": 27049 }, { "epoch": 4.96, "learning_rate": 6.662368814597303e-06, "loss": 0.0175, "step": 27050 }, { "epoch": 4.96, "learning_rate": 6.661668966951134e-06, "loss": 0.0169, "step": 27051 }, { "epoch": 4.96, "learning_rate": 6.660969137705924e-06, "loss": 0.0202, "step": 27052 }, { "epoch": 4.96, "learning_rate": 6.660269326865525e-06, "loss": 0.0194, "step": 27053 }, { "epoch": 4.96, "learning_rate": 6.6595695344338075e-06, "loss": 0.0197, "step": 27054 }, { "epoch": 4.96, "learning_rate": 6.658869760414617e-06, "loss": 0.0171, "step": 27055 }, { "epoch": 4.96, "learning_rate": 6.658170004811815e-06, "loss": 0.0206, "step": 27056 }, { "epoch": 4.96, "learning_rate": 6.657470267629256e-06, "loss": 0.0204, "step": 27057 }, { "epoch": 4.96, "learning_rate": 6.656770548870802e-06, "loss": 0.0179, "step": 27058 }, { "epoch": 4.96, "learning_rate": 6.656070848540307e-06, "loss": 0.0229, "step": 27059 }, { "epoch": 4.96, "learning_rate": 6.655371166641628e-06, "loss": 0.0222, "step": 27060 }, { "epoch": 4.96, "learning_rate": 6.654671503178623e-06, "loss": 0.0207, "step": 27061 }, { "epoch": 4.96, "learning_rate": 6.65397185815514e-06, "loss": 0.0194, "step": 27062 }, { "epoch": 4.96, "learning_rate": 6.65327223157505e-06, "loss": 0.0202, "step": 27063 }, { "epoch": 4.96, "learning_rate": 6.6525726234421994e-06, "loss": 0.0186, "step": 27064 }, { "epoch": 4.96, "learning_rate": 6.651873033760447e-06, "loss": 0.0238, "step": 27065 }, { "epoch": 4.96, "learning_rate": 6.6511734625336465e-06, "loss": 0.0219, "step": 27066 }, { "epoch": 4.96, "learning_rate": 6.650473909765659e-06, "loss": 0.0206, "step": 27067 }, { "epoch": 4.96, "learning_rate": 6.649774375460338e-06, "loss": 0.0146, "step": 27068 }, { "epoch": 4.96, "learning_rate": 6.649074859621538e-06, "loss": 0.0184, "step": 27069 }, { "epoch": 4.96, "learning_rate": 6.648375362253119e-06, "loss": 0.0176, "step": 27070 }, { "epoch": 4.96, "learning_rate": 6.647675883358926e-06, "loss": 0.0174, "step": 27071 }, { "epoch": 4.96, "learning_rate": 6.64697642294283e-06, "loss": 0.0187, "step": 27072 }, { "epoch": 4.96, "learning_rate": 6.646276981008675e-06, "loss": 0.0163, "step": 27073 }, { "epoch": 4.96, "learning_rate": 6.645577557560321e-06, "loss": 0.0219, "step": 27074 }, { "epoch": 4.97, "learning_rate": 6.64487815260162e-06, "loss": 0.0179, "step": 27075 }, { "epoch": 4.97, "learning_rate": 6.64417876613643e-06, "loss": 0.0163, "step": 27076 }, { "epoch": 4.97, "learning_rate": 6.643479398168606e-06, "loss": 0.0224, "step": 27077 }, { "epoch": 4.97, "learning_rate": 6.642780048702002e-06, "loss": 0.0181, "step": 27078 }, { "epoch": 4.97, "learning_rate": 6.642080717740471e-06, "loss": 0.0173, "step": 27079 }, { "epoch": 4.97, "learning_rate": 6.6413814052878725e-06, "loss": 0.0194, "step": 27080 }, { "epoch": 4.97, "learning_rate": 6.64068211134806e-06, "loss": 0.0157, "step": 27081 }, { "epoch": 4.97, "learning_rate": 6.639982835924884e-06, "loss": 0.0203, "step": 27082 }, { "epoch": 4.97, "learning_rate": 6.639283579022202e-06, "loss": 0.0188, "step": 27083 }, { "epoch": 4.97, "learning_rate": 6.638584340643865e-06, "loss": 0.0153, "step": 27084 }, { "epoch": 4.97, "learning_rate": 6.637885120793732e-06, "loss": 0.027, "step": 27085 }, { "epoch": 4.97, "learning_rate": 6.637185919475655e-06, "loss": 0.0236, "step": 27086 }, { "epoch": 4.97, "learning_rate": 6.636486736693488e-06, "loss": 0.0214, "step": 27087 }, { "epoch": 4.97, "learning_rate": 6.635787572451083e-06, "loss": 0.0218, "step": 27088 }, { "epoch": 4.97, "learning_rate": 6.635088426752298e-06, "loss": 0.0221, "step": 27089 }, { "epoch": 4.97, "learning_rate": 6.634389299600985e-06, "loss": 0.0202, "step": 27090 }, { "epoch": 4.97, "learning_rate": 6.633690191000996e-06, "loss": 0.0189, "step": 27091 }, { "epoch": 4.97, "learning_rate": 6.632991100956187e-06, "loss": 0.0223, "step": 27092 }, { "epoch": 4.97, "learning_rate": 6.632292029470406e-06, "loss": 0.0175, "step": 27093 }, { "epoch": 4.97, "learning_rate": 6.631592976547513e-06, "loss": 0.0198, "step": 27094 }, { "epoch": 4.97, "learning_rate": 6.630893942191359e-06, "loss": 0.0198, "step": 27095 }, { "epoch": 4.97, "learning_rate": 6.630194926405795e-06, "loss": 0.0229, "step": 27096 }, { "epoch": 4.97, "learning_rate": 6.629495929194673e-06, "loss": 0.0173, "step": 27097 }, { "epoch": 4.97, "learning_rate": 6.628796950561853e-06, "loss": 0.0214, "step": 27098 }, { "epoch": 4.97, "learning_rate": 6.628097990511184e-06, "loss": 0.0221, "step": 27099 }, { "epoch": 4.97, "learning_rate": 6.627399049046515e-06, "loss": 0.0189, "step": 27100 }, { "epoch": 4.97, "learning_rate": 6.6267001261717015e-06, "loss": 0.0219, "step": 27101 }, { "epoch": 4.97, "learning_rate": 6.626001221890596e-06, "loss": 0.0206, "step": 27102 }, { "epoch": 4.97, "learning_rate": 6.625302336207051e-06, "loss": 0.0208, "step": 27103 }, { "epoch": 4.97, "learning_rate": 6.624603469124919e-06, "loss": 0.0188, "step": 27104 }, { "epoch": 4.97, "learning_rate": 6.623904620648054e-06, "loss": 0.02, "step": 27105 }, { "epoch": 4.97, "learning_rate": 6.623205790780299e-06, "loss": 0.0182, "step": 27106 }, { "epoch": 4.97, "learning_rate": 6.6225069795255184e-06, "loss": 0.0204, "step": 27107 }, { "epoch": 4.97, "learning_rate": 6.62180818688756e-06, "loss": 0.0173, "step": 27108 }, { "epoch": 4.97, "learning_rate": 6.62110941287027e-06, "loss": 0.0173, "step": 27109 }, { "epoch": 4.97, "learning_rate": 6.620410657477504e-06, "loss": 0.0212, "step": 27110 }, { "epoch": 4.97, "learning_rate": 6.619711920713114e-06, "loss": 0.0217, "step": 27111 }, { "epoch": 4.97, "learning_rate": 6.619013202580953e-06, "loss": 0.0247, "step": 27112 }, { "epoch": 4.97, "learning_rate": 6.618314503084871e-06, "loss": 0.0211, "step": 27113 }, { "epoch": 4.97, "learning_rate": 6.6176158222287175e-06, "loss": 0.0191, "step": 27114 }, { "epoch": 4.97, "learning_rate": 6.6169171600163415e-06, "loss": 0.0208, "step": 27115 }, { "epoch": 4.97, "learning_rate": 6.6162185164516e-06, "loss": 0.0218, "step": 27116 }, { "epoch": 4.97, "learning_rate": 6.615519891538344e-06, "loss": 0.0181, "step": 27117 }, { "epoch": 4.97, "learning_rate": 6.614821285280419e-06, "loss": 0.0249, "step": 27118 }, { "epoch": 4.97, "learning_rate": 6.614122697681677e-06, "loss": 0.0175, "step": 27119 }, { "epoch": 4.97, "learning_rate": 6.613424128745973e-06, "loss": 0.0191, "step": 27120 }, { "epoch": 4.97, "learning_rate": 6.612725578477153e-06, "loss": 0.0183, "step": 27121 }, { "epoch": 4.97, "learning_rate": 6.612027046879071e-06, "loss": 0.0212, "step": 27122 }, { "epoch": 4.97, "learning_rate": 6.611328533955574e-06, "loss": 0.0212, "step": 27123 }, { "epoch": 4.97, "learning_rate": 6.610630039710511e-06, "loss": 0.0195, "step": 27124 }, { "epoch": 4.97, "learning_rate": 6.609931564147736e-06, "loss": 0.0263, "step": 27125 }, { "epoch": 4.97, "learning_rate": 6.609233107271101e-06, "loss": 0.0178, "step": 27126 }, { "epoch": 4.97, "learning_rate": 6.60853466908445e-06, "loss": 0.0214, "step": 27127 }, { "epoch": 4.97, "learning_rate": 6.607836249591632e-06, "loss": 0.0209, "step": 27128 }, { "epoch": 4.98, "learning_rate": 6.607137848796504e-06, "loss": 0.0215, "step": 27129 }, { "epoch": 4.98, "learning_rate": 6.606439466702912e-06, "loss": 0.0209, "step": 27130 }, { "epoch": 4.98, "learning_rate": 6.605741103314703e-06, "loss": 0.0148, "step": 27131 }, { "epoch": 4.98, "learning_rate": 6.605042758635729e-06, "loss": 0.019, "step": 27132 }, { "epoch": 4.98, "learning_rate": 6.6043444326698394e-06, "loss": 0.0199, "step": 27133 }, { "epoch": 4.98, "learning_rate": 6.603646125420883e-06, "loss": 0.0184, "step": 27134 }, { "epoch": 4.98, "learning_rate": 6.60294783689271e-06, "loss": 0.0179, "step": 27135 }, { "epoch": 4.98, "learning_rate": 6.602249567089167e-06, "loss": 0.0222, "step": 27136 }, { "epoch": 4.98, "learning_rate": 6.6015513160141036e-06, "loss": 0.0191, "step": 27137 }, { "epoch": 4.98, "learning_rate": 6.600853083671368e-06, "loss": 0.0239, "step": 27138 }, { "epoch": 4.98, "learning_rate": 6.600154870064812e-06, "loss": 0.0216, "step": 27139 }, { "epoch": 4.98, "learning_rate": 6.599456675198282e-06, "loss": 0.0197, "step": 27140 }, { "epoch": 4.98, "learning_rate": 6.598758499075621e-06, "loss": 0.022, "step": 27141 }, { "epoch": 4.98, "learning_rate": 6.59806034170069e-06, "loss": 0.0177, "step": 27142 }, { "epoch": 4.98, "learning_rate": 6.597362203077329e-06, "loss": 0.0138, "step": 27143 }, { "epoch": 4.98, "learning_rate": 6.596664083209385e-06, "loss": 0.0169, "step": 27144 }, { "epoch": 4.98, "learning_rate": 6.595965982100709e-06, "loss": 0.0165, "step": 27145 }, { "epoch": 4.98, "learning_rate": 6.595267899755147e-06, "loss": 0.0209, "step": 27146 }, { "epoch": 4.98, "learning_rate": 6.594569836176549e-06, "loss": 0.0182, "step": 27147 }, { "epoch": 4.98, "learning_rate": 6.593871791368762e-06, "loss": 0.0194, "step": 27148 }, { "epoch": 4.98, "learning_rate": 6.5931737653356345e-06, "loss": 0.0185, "step": 27149 }, { "epoch": 4.98, "learning_rate": 6.59247575808101e-06, "loss": 0.017, "step": 27150 }, { "epoch": 4.98, "learning_rate": 6.591777769608742e-06, "loss": 0.0177, "step": 27151 }, { "epoch": 4.98, "learning_rate": 6.591079799922675e-06, "loss": 0.0166, "step": 27152 }, { "epoch": 4.98, "learning_rate": 6.5903818490266554e-06, "loss": 0.0169, "step": 27153 }, { "epoch": 4.98, "learning_rate": 6.589683916924527e-06, "loss": 0.0183, "step": 27154 }, { "epoch": 4.98, "learning_rate": 6.588986003620146e-06, "loss": 0.0184, "step": 27155 }, { "epoch": 4.98, "learning_rate": 6.588288109117352e-06, "loss": 0.0236, "step": 27156 }, { "epoch": 4.98, "learning_rate": 6.587590233419995e-06, "loss": 0.0168, "step": 27157 }, { "epoch": 4.98, "learning_rate": 6.58689237653192e-06, "loss": 0.0179, "step": 27158 }, { "epoch": 4.98, "learning_rate": 6.586194538456972e-06, "loss": 0.0185, "step": 27159 }, { "epoch": 4.98, "learning_rate": 6.585496719199001e-06, "loss": 0.0161, "step": 27160 }, { "epoch": 4.98, "learning_rate": 6.584798918761854e-06, "loss": 0.0223, "step": 27161 }, { "epoch": 4.98, "learning_rate": 6.584101137149375e-06, "loss": 0.0206, "step": 27162 }, { "epoch": 4.98, "learning_rate": 6.583403374365406e-06, "loss": 0.0144, "step": 27163 }, { "epoch": 4.98, "learning_rate": 6.5827056304138035e-06, "loss": 0.0243, "step": 27164 }, { "epoch": 4.98, "learning_rate": 6.5820079052984055e-06, "loss": 0.0183, "step": 27165 }, { "epoch": 4.98, "learning_rate": 6.58131019902306e-06, "loss": 0.0221, "step": 27166 }, { "epoch": 4.98, "learning_rate": 6.580612511591613e-06, "loss": 0.0192, "step": 27167 }, { "epoch": 4.98, "learning_rate": 6.579914843007906e-06, "loss": 0.0191, "step": 27168 }, { "epoch": 4.98, "learning_rate": 6.579217193275793e-06, "loss": 0.0181, "step": 27169 }, { "epoch": 4.98, "learning_rate": 6.578519562399117e-06, "loss": 0.0178, "step": 27170 }, { "epoch": 4.98, "learning_rate": 6.577821950381717e-06, "loss": 0.0199, "step": 27171 }, { "epoch": 4.98, "learning_rate": 6.577124357227441e-06, "loss": 0.0189, "step": 27172 }, { "epoch": 4.98, "learning_rate": 6.576426782940141e-06, "loss": 0.0214, "step": 27173 }, { "epoch": 4.98, "learning_rate": 6.575729227523655e-06, "loss": 0.0166, "step": 27174 }, { "epoch": 4.98, "learning_rate": 6.575031690981829e-06, "loss": 0.0206, "step": 27175 }, { "epoch": 4.98, "learning_rate": 6.574334173318508e-06, "loss": 0.0219, "step": 27176 }, { "epoch": 4.98, "learning_rate": 6.573636674537537e-06, "loss": 0.0188, "step": 27177 }, { "epoch": 4.98, "learning_rate": 6.572939194642762e-06, "loss": 0.023, "step": 27178 }, { "epoch": 4.98, "learning_rate": 6.5722417336380284e-06, "loss": 0.0172, "step": 27179 }, { "epoch": 4.98, "learning_rate": 6.571544291527177e-06, "loss": 0.019, "step": 27180 }, { "epoch": 4.98, "learning_rate": 6.570846868314049e-06, "loss": 0.0156, "step": 27181 }, { "epoch": 4.98, "learning_rate": 6.5701494640024994e-06, "loss": 0.0198, "step": 27182 }, { "epoch": 4.98, "learning_rate": 6.569452078596366e-06, "loss": 0.0233, "step": 27183 }, { "epoch": 4.99, "learning_rate": 6.568754712099492e-06, "loss": 0.0217, "step": 27184 }, { "epoch": 4.99, "learning_rate": 6.568057364515721e-06, "loss": 0.022, "step": 27185 }, { "epoch": 4.99, "learning_rate": 6.567360035848899e-06, "loss": 0.0154, "step": 27186 }, { "epoch": 4.99, "learning_rate": 6.5666627261028695e-06, "loss": 0.0217, "step": 27187 }, { "epoch": 4.99, "learning_rate": 6.565965435281478e-06, "loss": 0.0207, "step": 27188 }, { "epoch": 4.99, "learning_rate": 6.565268163388563e-06, "loss": 0.0203, "step": 27189 }, { "epoch": 4.99, "learning_rate": 6.564570910427967e-06, "loss": 0.0181, "step": 27190 }, { "epoch": 4.99, "learning_rate": 6.563873676403541e-06, "loss": 0.021, "step": 27191 }, { "epoch": 4.99, "learning_rate": 6.563176461319123e-06, "loss": 0.0208, "step": 27192 }, { "epoch": 4.99, "learning_rate": 6.562479265178556e-06, "loss": 0.0207, "step": 27193 }, { "epoch": 4.99, "learning_rate": 6.561782087985682e-06, "loss": 0.0212, "step": 27194 }, { "epoch": 4.99, "learning_rate": 6.561084929744348e-06, "loss": 0.0195, "step": 27195 }, { "epoch": 4.99, "learning_rate": 6.5603877904583955e-06, "loss": 0.0256, "step": 27196 }, { "epoch": 4.99, "learning_rate": 6.5596906701316665e-06, "loss": 0.0202, "step": 27197 }, { "epoch": 4.99, "learning_rate": 6.558993568768002e-06, "loss": 0.0215, "step": 27198 }, { "epoch": 4.99, "learning_rate": 6.558296486371243e-06, "loss": 0.0212, "step": 27199 }, { "epoch": 4.99, "learning_rate": 6.5575994229452374e-06, "loss": 0.0174, "step": 27200 }, { "epoch": 4.99, "learning_rate": 6.556902378493823e-06, "loss": 0.0243, "step": 27201 }, { "epoch": 4.99, "learning_rate": 6.556205353020845e-06, "loss": 0.0219, "step": 27202 }, { "epoch": 4.99, "learning_rate": 6.555508346530141e-06, "loss": 0.019, "step": 27203 }, { "epoch": 4.99, "learning_rate": 6.554811359025557e-06, "loss": 0.0182, "step": 27204 }, { "epoch": 4.99, "learning_rate": 6.554114390510935e-06, "loss": 0.022, "step": 27205 }, { "epoch": 4.99, "learning_rate": 6.553417440990115e-06, "loss": 0.0149, "step": 27206 }, { "epoch": 4.99, "learning_rate": 6.5527205104669346e-06, "loss": 0.0192, "step": 27207 }, { "epoch": 4.99, "learning_rate": 6.552023598945245e-06, "loss": 0.0196, "step": 27208 }, { "epoch": 4.99, "learning_rate": 6.55132670642888e-06, "loss": 0.0208, "step": 27209 }, { "epoch": 4.99, "learning_rate": 6.550629832921683e-06, "loss": 0.0206, "step": 27210 }, { "epoch": 4.99, "learning_rate": 6.5499329784274956e-06, "loss": 0.0165, "step": 27211 }, { "epoch": 4.99, "learning_rate": 6.549236142950156e-06, "loss": 0.023, "step": 27212 }, { "epoch": 4.99, "learning_rate": 6.548539326493511e-06, "loss": 0.0225, "step": 27213 }, { "epoch": 4.99, "learning_rate": 6.547842529061395e-06, "loss": 0.0194, "step": 27214 }, { "epoch": 4.99, "learning_rate": 6.547145750657656e-06, "loss": 0.0153, "step": 27215 }, { "epoch": 4.99, "learning_rate": 6.546448991286124e-06, "loss": 0.0204, "step": 27216 }, { "epoch": 4.99, "learning_rate": 6.5457522509506525e-06, "loss": 0.0211, "step": 27217 }, { "epoch": 4.99, "learning_rate": 6.5450555296550734e-06, "loss": 0.0157, "step": 27218 }, { "epoch": 4.99, "learning_rate": 6.5443588274032286e-06, "loss": 0.0153, "step": 27219 }, { "epoch": 4.99, "learning_rate": 6.543662144198959e-06, "loss": 0.0205, "step": 27220 }, { "epoch": 4.99, "learning_rate": 6.542965480046104e-06, "loss": 0.0183, "step": 27221 }, { "epoch": 4.99, "learning_rate": 6.542268834948504e-06, "loss": 0.0202, "step": 27222 }, { "epoch": 4.99, "learning_rate": 6.541572208910001e-06, "loss": 0.0212, "step": 27223 }, { "epoch": 4.99, "learning_rate": 6.540875601934435e-06, "loss": 0.0214, "step": 27224 }, { "epoch": 4.99, "learning_rate": 6.540179014025636e-06, "loss": 0.0223, "step": 27225 }, { "epoch": 4.99, "learning_rate": 6.53948244518746e-06, "loss": 0.0187, "step": 27226 }, { "epoch": 4.99, "learning_rate": 6.5387858954237335e-06, "loss": 0.0181, "step": 27227 }, { "epoch": 4.99, "learning_rate": 6.538089364738301e-06, "loss": 0.0189, "step": 27228 }, { "epoch": 4.99, "learning_rate": 6.537392853135e-06, "loss": 0.0198, "step": 27229 }, { "epoch": 4.99, "learning_rate": 6.536696360617672e-06, "loss": 0.0199, "step": 27230 }, { "epoch": 4.99, "learning_rate": 6.5359998871901545e-06, "loss": 0.0198, "step": 27231 }, { "epoch": 4.99, "learning_rate": 6.535303432856286e-06, "loss": 0.0225, "step": 27232 }, { "epoch": 4.99, "learning_rate": 6.534606997619909e-06, "loss": 0.0208, "step": 27233 }, { "epoch": 4.99, "learning_rate": 6.533910581484854e-06, "loss": 0.0169, "step": 27234 }, { "epoch": 4.99, "learning_rate": 6.53321418445497e-06, "loss": 0.0216, "step": 27235 }, { "epoch": 4.99, "learning_rate": 6.532517806534089e-06, "loss": 0.0182, "step": 27236 }, { "epoch": 4.99, "learning_rate": 6.5318214477260515e-06, "loss": 0.015, "step": 27237 }, { "epoch": 5.0, "learning_rate": 6.531125108034693e-06, "loss": 0.0193, "step": 27238 }, { "epoch": 5.0, "learning_rate": 6.530428787463856e-06, "loss": 0.0238, "step": 27239 }, { "epoch": 5.0, "learning_rate": 6.5297324860173775e-06, "loss": 0.0191, "step": 27240 }, { "epoch": 5.0, "learning_rate": 6.529036203699094e-06, "loss": 0.0212, "step": 27241 }, { "epoch": 5.0, "learning_rate": 6.5283399405128465e-06, "loss": 0.0193, "step": 27242 }, { "epoch": 5.0, "learning_rate": 6.527643696462465e-06, "loss": 0.0201, "step": 27243 }, { "epoch": 5.0, "learning_rate": 6.526947471551799e-06, "loss": 0.0184, "step": 27244 }, { "epoch": 5.0, "learning_rate": 6.526251265784678e-06, "loss": 0.0157, "step": 27245 }, { "epoch": 5.0, "learning_rate": 6.52555507916494e-06, "loss": 0.0163, "step": 27246 }, { "epoch": 5.0, "learning_rate": 6.524858911696423e-06, "loss": 0.0168, "step": 27247 }, { "epoch": 5.0, "learning_rate": 6.524162763382966e-06, "loss": 0.0247, "step": 27248 }, { "epoch": 5.0, "learning_rate": 6.523466634228407e-06, "loss": 0.0208, "step": 27249 }, { "epoch": 5.0, "learning_rate": 6.52277052423658e-06, "loss": 0.0252, "step": 27250 }, { "epoch": 5.0, "learning_rate": 6.5220744334113265e-06, "loss": 0.0274, "step": 27251 }, { "epoch": 5.0, "learning_rate": 6.521378361756473e-06, "loss": 0.0206, "step": 27252 }, { "epoch": 5.0, "learning_rate": 6.5206823092758695e-06, "loss": 0.0208, "step": 27253 }, { "epoch": 5.0, "learning_rate": 6.519986275973346e-06, "loss": 0.0234, "step": 27254 }, { "epoch": 5.0, "learning_rate": 6.519290261852737e-06, "loss": 0.0207, "step": 27255 }, { "epoch": 5.0, "learning_rate": 6.518594266917883e-06, "loss": 0.0152, "step": 27256 }, { "epoch": 5.0, "learning_rate": 6.517898291172618e-06, "loss": 0.0213, "step": 27257 }, { "epoch": 5.0, "learning_rate": 6.517202334620781e-06, "loss": 0.0182, "step": 27258 }, { "epoch": 5.0, "learning_rate": 6.516506397266208e-06, "loss": 0.0191, "step": 27259 }, { "epoch": 5.0, "learning_rate": 6.515810479112727e-06, "loss": 0.0142, "step": 27260 }, { "epoch": 5.0, "learning_rate": 6.515114580164184e-06, "loss": 0.0191, "step": 27261 }, { "epoch": 5.0, "learning_rate": 6.514418700424414e-06, "loss": 0.0192, "step": 27262 }, { "epoch": 5.0, "learning_rate": 6.513722839897247e-06, "loss": 0.016, "step": 27263 }, { "epoch": 5.0, "learning_rate": 6.513026998586522e-06, "loss": 0.0214, "step": 27264 }, { "epoch": 5.0, "learning_rate": 6.5123311764960715e-06, "loss": 0.018, "step": 27265 }, { "epoch": 5.0, "learning_rate": 6.511635373629736e-06, "loss": 0.0081, "step": 27266 }, { "epoch": 5.0, "learning_rate": 6.510939589991347e-06, "loss": 0.0102, "step": 27267 }, { "epoch": 5.0, "learning_rate": 6.510243825584743e-06, "loss": 0.0111, "step": 27268 }, { "epoch": 5.0, "learning_rate": 6.50954808041375e-06, "loss": 0.0127, "step": 27269 }, { "epoch": 5.0, "learning_rate": 6.508852354482216e-06, "loss": 0.011, "step": 27270 }, { "epoch": 5.0, "learning_rate": 6.50815664779397e-06, "loss": 0.0101, "step": 27271 }, { "epoch": 5.0, "learning_rate": 6.507460960352845e-06, "loss": 0.0092, "step": 27272 }, { "epoch": 5.0, "learning_rate": 6.506765292162675e-06, "loss": 0.0081, "step": 27273 }, { "epoch": 5.0, "learning_rate": 6.5060696432272974e-06, "loss": 0.0103, "step": 27274 }, { "epoch": 5.0, "learning_rate": 6.505374013550546e-06, "loss": 0.0091, "step": 27275 }, { "epoch": 5.0, "learning_rate": 6.504678403136256e-06, "loss": 0.0082, "step": 27276 }, { "epoch": 5.0, "learning_rate": 6.503982811988262e-06, "loss": 0.012, "step": 27277 }, { "epoch": 5.0, "learning_rate": 6.50328724011039e-06, "loss": 0.01, "step": 27278 }, { "epoch": 5.0, "learning_rate": 6.502591687506485e-06, "loss": 0.01, "step": 27279 }, { "epoch": 5.0, "learning_rate": 6.50189615418038e-06, "loss": 0.0107, "step": 27280 }, { "epoch": 5.0, "learning_rate": 6.501200640135902e-06, "loss": 0.0075, "step": 27281 }, { "epoch": 5.0, "learning_rate": 6.500505145376885e-06, "loss": 0.0089, "step": 27282 }, { "epoch": 5.0, "learning_rate": 6.499809669907169e-06, "loss": 0.013, "step": 27283 }, { "epoch": 5.0, "learning_rate": 6.499114213730584e-06, "loss": 0.0114, "step": 27284 }, { "epoch": 5.0, "learning_rate": 6.498418776850963e-06, "loss": 0.0128, "step": 27285 }, { "epoch": 5.0, "learning_rate": 6.497723359272143e-06, "loss": 0.0083, "step": 27286 }, { "epoch": 5.0, "learning_rate": 6.497027960997947e-06, "loss": 0.0089, "step": 27287 }, { "epoch": 5.0, "learning_rate": 6.496332582032218e-06, "loss": 0.0107, "step": 27288 }, { "epoch": 5.0, "learning_rate": 6.49563722237879e-06, "loss": 0.0114, "step": 27289 }, { "epoch": 5.0, "learning_rate": 6.494941882041489e-06, "loss": 0.0085, "step": 27290 }, { "epoch": 5.0, "learning_rate": 6.4942465610241475e-06, "loss": 0.0093, "step": 27291 }, { "epoch": 5.0, "learning_rate": 6.493551259330604e-06, "loss": 0.0101, "step": 27292 }, { "epoch": 5.01, "learning_rate": 6.492855976964687e-06, "loss": 0.0092, "step": 27293 }, { "epoch": 5.01, "learning_rate": 6.4921607139302315e-06, "loss": 0.0094, "step": 27294 }, { "epoch": 5.01, "learning_rate": 6.49146547023107e-06, "loss": 0.0138, "step": 27295 }, { "epoch": 5.01, "learning_rate": 6.490770245871025e-06, "loss": 0.0106, "step": 27296 }, { "epoch": 5.01, "learning_rate": 6.490075040853942e-06, "loss": 0.0092, "step": 27297 }, { "epoch": 5.01, "learning_rate": 6.489379855183651e-06, "loss": 0.007, "step": 27298 }, { "epoch": 5.01, "learning_rate": 6.488684688863978e-06, "loss": 0.0118, "step": 27299 }, { "epoch": 5.01, "learning_rate": 6.487989541898754e-06, "loss": 0.0128, "step": 27300 }, { "epoch": 5.01, "learning_rate": 6.4872944142918155e-06, "loss": 0.009, "step": 27301 }, { "epoch": 5.01, "learning_rate": 6.486599306046995e-06, "loss": 0.0109, "step": 27302 }, { "epoch": 5.01, "learning_rate": 6.4859042171681195e-06, "loss": 0.0074, "step": 27303 }, { "epoch": 5.01, "learning_rate": 6.485209147659024e-06, "loss": 0.0093, "step": 27304 }, { "epoch": 5.01, "learning_rate": 6.484514097523533e-06, "loss": 0.0096, "step": 27305 }, { "epoch": 5.01, "learning_rate": 6.483819066765489e-06, "loss": 0.0134, "step": 27306 }, { "epoch": 5.01, "learning_rate": 6.483124055388714e-06, "loss": 0.0088, "step": 27307 }, { "epoch": 5.01, "learning_rate": 6.4824290633970425e-06, "loss": 0.0065, "step": 27308 }, { "epoch": 5.01, "learning_rate": 6.481734090794301e-06, "loss": 0.0116, "step": 27309 }, { "epoch": 5.01, "learning_rate": 6.481039137584326e-06, "loss": 0.0074, "step": 27310 }, { "epoch": 5.01, "learning_rate": 6.480344203770945e-06, "loss": 0.0102, "step": 27311 }, { "epoch": 5.01, "learning_rate": 6.479649289357991e-06, "loss": 0.0094, "step": 27312 }, { "epoch": 5.01, "learning_rate": 6.47895439434929e-06, "loss": 0.0091, "step": 27313 }, { "epoch": 5.01, "learning_rate": 6.478259518748675e-06, "loss": 0.0122, "step": 27314 }, { "epoch": 5.01, "learning_rate": 6.477564662559981e-06, "loss": 0.0084, "step": 27315 }, { "epoch": 5.01, "learning_rate": 6.476869825787029e-06, "loss": 0.0094, "step": 27316 }, { "epoch": 5.01, "learning_rate": 6.476175008433654e-06, "loss": 0.01, "step": 27317 }, { "epoch": 5.01, "learning_rate": 6.475480210503683e-06, "loss": 0.0099, "step": 27318 }, { "epoch": 5.01, "learning_rate": 6.4747854320009494e-06, "loss": 0.0081, "step": 27319 }, { "epoch": 5.01, "learning_rate": 6.474090672929281e-06, "loss": 0.0116, "step": 27320 }, { "epoch": 5.01, "learning_rate": 6.473395933292508e-06, "loss": 0.0115, "step": 27321 }, { "epoch": 5.01, "learning_rate": 6.472701213094455e-06, "loss": 0.0069, "step": 27322 }, { "epoch": 5.01, "learning_rate": 6.47200651233896e-06, "loss": 0.008, "step": 27323 }, { "epoch": 5.01, "learning_rate": 6.47131183102985e-06, "loss": 0.0092, "step": 27324 }, { "epoch": 5.01, "learning_rate": 6.470617169170948e-06, "loss": 0.0112, "step": 27325 }, { "epoch": 5.01, "learning_rate": 6.469922526766088e-06, "loss": 0.0084, "step": 27326 }, { "epoch": 5.01, "learning_rate": 6.469227903819095e-06, "loss": 0.0083, "step": 27327 }, { "epoch": 5.01, "learning_rate": 6.468533300333802e-06, "loss": 0.0096, "step": 27328 }, { "epoch": 5.01, "learning_rate": 6.467838716314038e-06, "loss": 0.0104, "step": 27329 }, { "epoch": 5.01, "learning_rate": 6.467144151763629e-06, "loss": 0.0077, "step": 27330 }, { "epoch": 5.01, "learning_rate": 6.466449606686401e-06, "loss": 0.0104, "step": 27331 }, { "epoch": 5.01, "learning_rate": 6.465755081086189e-06, "loss": 0.0076, "step": 27332 }, { "epoch": 5.01, "learning_rate": 6.465060574966819e-06, "loss": 0.0099, "step": 27333 }, { "epoch": 5.01, "learning_rate": 6.464366088332117e-06, "loss": 0.0114, "step": 27334 }, { "epoch": 5.01, "learning_rate": 6.4636716211859076e-06, "loss": 0.0116, "step": 27335 }, { "epoch": 5.01, "learning_rate": 6.462977173532028e-06, "loss": 0.0088, "step": 27336 }, { "epoch": 5.01, "learning_rate": 6.4622827453743e-06, "loss": 0.0108, "step": 27337 }, { "epoch": 5.01, "learning_rate": 6.461588336716552e-06, "loss": 0.01, "step": 27338 }, { "epoch": 5.01, "learning_rate": 6.460893947562612e-06, "loss": 0.0079, "step": 27339 }, { "epoch": 5.01, "learning_rate": 6.460199577916306e-06, "loss": 0.0105, "step": 27340 }, { "epoch": 5.01, "learning_rate": 6.459505227781464e-06, "loss": 0.0075, "step": 27341 }, { "epoch": 5.01, "learning_rate": 6.458810897161915e-06, "loss": 0.008, "step": 27342 }, { "epoch": 5.01, "learning_rate": 6.458116586061482e-06, "loss": 0.0063, "step": 27343 }, { "epoch": 5.01, "learning_rate": 6.4574222944839895e-06, "loss": 0.0106, "step": 27344 }, { "epoch": 5.01, "learning_rate": 6.4567280224332715e-06, "loss": 0.0081, "step": 27345 }, { "epoch": 5.01, "learning_rate": 6.45603376991315e-06, "loss": 0.0101, "step": 27346 }, { "epoch": 5.02, "learning_rate": 6.4553395369274554e-06, "loss": 0.0078, "step": 27347 }, { "epoch": 5.02, "learning_rate": 6.454645323480011e-06, "loss": 0.01, "step": 27348 }, { "epoch": 5.02, "learning_rate": 6.453951129574644e-06, "loss": 0.0119, "step": 27349 }, { "epoch": 5.02, "learning_rate": 6.453256955215182e-06, "loss": 0.0103, "step": 27350 }, { "epoch": 5.02, "learning_rate": 6.452562800405453e-06, "loss": 0.0079, "step": 27351 }, { "epoch": 5.02, "learning_rate": 6.45186866514928e-06, "loss": 0.0104, "step": 27352 }, { "epoch": 5.02, "learning_rate": 6.451174549450488e-06, "loss": 0.0102, "step": 27353 }, { "epoch": 5.02, "learning_rate": 6.450480453312906e-06, "loss": 0.0101, "step": 27354 }, { "epoch": 5.02, "learning_rate": 6.449786376740359e-06, "loss": 0.0087, "step": 27355 }, { "epoch": 5.02, "learning_rate": 6.449092319736674e-06, "loss": 0.0115, "step": 27356 }, { "epoch": 5.02, "learning_rate": 6.448398282305672e-06, "loss": 0.0095, "step": 27357 }, { "epoch": 5.02, "learning_rate": 6.4477042644511834e-06, "loss": 0.0086, "step": 27358 }, { "epoch": 5.02, "learning_rate": 6.4470102661770325e-06, "loss": 0.0126, "step": 27359 }, { "epoch": 5.02, "learning_rate": 6.446316287487046e-06, "loss": 0.012, "step": 27360 }, { "epoch": 5.02, "learning_rate": 6.445622328385047e-06, "loss": 0.01, "step": 27361 }, { "epoch": 5.02, "learning_rate": 6.444928388874858e-06, "loss": 0.0089, "step": 27362 }, { "epoch": 5.02, "learning_rate": 6.4442344689603085e-06, "loss": 0.0092, "step": 27363 }, { "epoch": 5.02, "learning_rate": 6.443540568645223e-06, "loss": 0.0099, "step": 27364 }, { "epoch": 5.02, "learning_rate": 6.442846687933425e-06, "loss": 0.0087, "step": 27365 }, { "epoch": 5.02, "learning_rate": 6.442152826828737e-06, "loss": 0.0117, "step": 27366 }, { "epoch": 5.02, "learning_rate": 6.441458985334987e-06, "loss": 0.0122, "step": 27367 }, { "epoch": 5.02, "learning_rate": 6.4407651634559995e-06, "loss": 0.0077, "step": 27368 }, { "epoch": 5.02, "learning_rate": 6.440071361195601e-06, "loss": 0.0099, "step": 27369 }, { "epoch": 5.02, "learning_rate": 6.439377578557609e-06, "loss": 0.0076, "step": 27370 }, { "epoch": 5.02, "learning_rate": 6.438683815545849e-06, "loss": 0.0098, "step": 27371 }, { "epoch": 5.02, "learning_rate": 6.4379900721641495e-06, "loss": 0.0114, "step": 27372 }, { "epoch": 5.02, "learning_rate": 6.437296348416333e-06, "loss": 0.0113, "step": 27373 }, { "epoch": 5.02, "learning_rate": 6.436602644306222e-06, "loss": 0.0091, "step": 27374 }, { "epoch": 5.02, "learning_rate": 6.435908959837638e-06, "loss": 0.0061, "step": 27375 }, { "epoch": 5.02, "learning_rate": 6.43521529501441e-06, "loss": 0.0074, "step": 27376 }, { "epoch": 5.02, "learning_rate": 6.434521649840359e-06, "loss": 0.0099, "step": 27377 }, { "epoch": 5.02, "learning_rate": 6.43382802431931e-06, "loss": 0.0094, "step": 27378 }, { "epoch": 5.02, "learning_rate": 6.433134418455082e-06, "loss": 0.0093, "step": 27379 }, { "epoch": 5.02, "learning_rate": 6.432440832251499e-06, "loss": 0.0085, "step": 27380 }, { "epoch": 5.02, "learning_rate": 6.431747265712388e-06, "loss": 0.0108, "step": 27381 }, { "epoch": 5.02, "learning_rate": 6.431053718841569e-06, "loss": 0.0096, "step": 27382 }, { "epoch": 5.02, "learning_rate": 6.430360191642866e-06, "loss": 0.0079, "step": 27383 }, { "epoch": 5.02, "learning_rate": 6.429666684120099e-06, "loss": 0.0118, "step": 27384 }, { "epoch": 5.02, "learning_rate": 6.428973196277095e-06, "loss": 0.0117, "step": 27385 }, { "epoch": 5.02, "learning_rate": 6.428279728117674e-06, "loss": 0.0095, "step": 27386 }, { "epoch": 5.02, "learning_rate": 6.427586279645663e-06, "loss": 0.01, "step": 27387 }, { "epoch": 5.02, "learning_rate": 6.426892850864872e-06, "loss": 0.0067, "step": 27388 }, { "epoch": 5.02, "learning_rate": 6.426199441779138e-06, "loss": 0.0077, "step": 27389 }, { "epoch": 5.02, "learning_rate": 6.425506052392274e-06, "loss": 0.0095, "step": 27390 }, { "epoch": 5.02, "learning_rate": 6.424812682708103e-06, "loss": 0.0076, "step": 27391 }, { "epoch": 5.02, "learning_rate": 6.424119332730451e-06, "loss": 0.0083, "step": 27392 }, { "epoch": 5.02, "learning_rate": 6.423426002463133e-06, "loss": 0.0123, "step": 27393 }, { "epoch": 5.02, "learning_rate": 6.422732691909979e-06, "loss": 0.009, "step": 27394 }, { "epoch": 5.02, "learning_rate": 6.422039401074805e-06, "loss": 0.0102, "step": 27395 }, { "epoch": 5.02, "learning_rate": 6.421346129961435e-06, "loss": 0.007, "step": 27396 }, { "epoch": 5.02, "learning_rate": 6.4206528785736824e-06, "loss": 0.008, "step": 27397 }, { "epoch": 5.02, "learning_rate": 6.419959646915382e-06, "loss": 0.0085, "step": 27398 }, { "epoch": 5.02, "learning_rate": 6.4192664349903455e-06, "loss": 0.0077, "step": 27399 }, { "epoch": 5.02, "learning_rate": 6.418573242802397e-06, "loss": 0.0104, "step": 27400 }, { "epoch": 5.02, "learning_rate": 6.417880070355356e-06, "loss": 0.0072, "step": 27401 }, { "epoch": 5.03, "learning_rate": 6.417186917653042e-06, "loss": 0.008, "step": 27402 }, { "epoch": 5.03, "learning_rate": 6.4164937846992805e-06, "loss": 0.0093, "step": 27403 }, { "epoch": 5.03, "learning_rate": 6.415800671497887e-06, "loss": 0.0151, "step": 27404 }, { "epoch": 5.03, "learning_rate": 6.415107578052688e-06, "loss": 0.0061, "step": 27405 }, { "epoch": 5.03, "learning_rate": 6.414414504367495e-06, "loss": 0.0091, "step": 27406 }, { "epoch": 5.03, "learning_rate": 6.413721450446138e-06, "loss": 0.0109, "step": 27407 }, { "epoch": 5.03, "learning_rate": 6.41302841629243e-06, "loss": 0.0094, "step": 27408 }, { "epoch": 5.03, "learning_rate": 6.412335401910193e-06, "loss": 0.0077, "step": 27409 }, { "epoch": 5.03, "learning_rate": 6.411642407303247e-06, "loss": 0.0095, "step": 27410 }, { "epoch": 5.03, "learning_rate": 6.410949432475413e-06, "loss": 0.0103, "step": 27411 }, { "epoch": 5.03, "learning_rate": 6.410256477430509e-06, "loss": 0.0096, "step": 27412 }, { "epoch": 5.03, "learning_rate": 6.409563542172359e-06, "loss": 0.0082, "step": 27413 }, { "epoch": 5.03, "learning_rate": 6.408870626704778e-06, "loss": 0.01, "step": 27414 }, { "epoch": 5.03, "learning_rate": 6.40817773103158e-06, "loss": 0.0083, "step": 27415 }, { "epoch": 5.03, "learning_rate": 6.407484855156596e-06, "loss": 0.0119, "step": 27416 }, { "epoch": 5.03, "learning_rate": 6.406791999083639e-06, "loss": 0.0099, "step": 27417 }, { "epoch": 5.03, "learning_rate": 6.406099162816528e-06, "loss": 0.0118, "step": 27418 }, { "epoch": 5.03, "learning_rate": 6.405406346359081e-06, "loss": 0.0096, "step": 27419 }, { "epoch": 5.03, "learning_rate": 6.40471354971512e-06, "loss": 0.0078, "step": 27420 }, { "epoch": 5.03, "learning_rate": 6.4040207728884626e-06, "loss": 0.0112, "step": 27421 }, { "epoch": 5.03, "learning_rate": 6.403328015882929e-06, "loss": 0.01, "step": 27422 }, { "epoch": 5.03, "learning_rate": 6.402635278702334e-06, "loss": 0.0077, "step": 27423 }, { "epoch": 5.03, "learning_rate": 6.401942561350492e-06, "loss": 0.0091, "step": 27424 }, { "epoch": 5.03, "learning_rate": 6.401249863831235e-06, "loss": 0.0075, "step": 27425 }, { "epoch": 5.03, "learning_rate": 6.400557186148371e-06, "loss": 0.0135, "step": 27426 }, { "epoch": 5.03, "learning_rate": 6.399864528305718e-06, "loss": 0.0085, "step": 27427 }, { "epoch": 5.03, "learning_rate": 6.399171890307095e-06, "loss": 0.0094, "step": 27428 }, { "epoch": 5.03, "learning_rate": 6.398479272156324e-06, "loss": 0.01, "step": 27429 }, { "epoch": 5.03, "learning_rate": 6.397786673857218e-06, "loss": 0.0093, "step": 27430 }, { "epoch": 5.03, "learning_rate": 6.3970940954136e-06, "loss": 0.0096, "step": 27431 }, { "epoch": 5.03, "learning_rate": 6.39640153682928e-06, "loss": 0.0105, "step": 27432 }, { "epoch": 5.03, "learning_rate": 6.395708998108077e-06, "loss": 0.0099, "step": 27433 }, { "epoch": 5.03, "learning_rate": 6.395016479253815e-06, "loss": 0.0116, "step": 27434 }, { "epoch": 5.03, "learning_rate": 6.394323980270305e-06, "loss": 0.0068, "step": 27435 }, { "epoch": 5.03, "learning_rate": 6.393631501161366e-06, "loss": 0.01, "step": 27436 }, { "epoch": 5.03, "learning_rate": 6.392939041930811e-06, "loss": 0.0121, "step": 27437 }, { "epoch": 5.03, "learning_rate": 6.392246602582464e-06, "loss": 0.0116, "step": 27438 }, { "epoch": 5.03, "learning_rate": 6.391554183120139e-06, "loss": 0.0074, "step": 27439 }, { "epoch": 5.03, "learning_rate": 6.390861783547653e-06, "loss": 0.0086, "step": 27440 }, { "epoch": 5.03, "learning_rate": 6.390169403868815e-06, "loss": 0.0112, "step": 27441 }, { "epoch": 5.03, "learning_rate": 6.3894770440874525e-06, "loss": 0.0097, "step": 27442 }, { "epoch": 5.03, "learning_rate": 6.388784704207378e-06, "loss": 0.0075, "step": 27443 }, { "epoch": 5.03, "learning_rate": 6.388092384232406e-06, "loss": 0.0106, "step": 27444 }, { "epoch": 5.03, "learning_rate": 6.387400084166353e-06, "loss": 0.0108, "step": 27445 }, { "epoch": 5.03, "learning_rate": 6.3867078040130326e-06, "loss": 0.011, "step": 27446 }, { "epoch": 5.03, "learning_rate": 6.386015543776267e-06, "loss": 0.0082, "step": 27447 }, { "epoch": 5.03, "learning_rate": 6.385323303459867e-06, "loss": 0.0095, "step": 27448 }, { "epoch": 5.03, "learning_rate": 6.3846310830676525e-06, "loss": 0.0114, "step": 27449 }, { "epoch": 5.03, "learning_rate": 6.383938882603431e-06, "loss": 0.0076, "step": 27450 }, { "epoch": 5.03, "learning_rate": 6.383246702071026e-06, "loss": 0.0088, "step": 27451 }, { "epoch": 5.03, "learning_rate": 6.382554541474251e-06, "loss": 0.0118, "step": 27452 }, { "epoch": 5.03, "learning_rate": 6.381862400816919e-06, "loss": 0.0107, "step": 27453 }, { "epoch": 5.03, "learning_rate": 6.3811702801028465e-06, "loss": 0.0092, "step": 27454 }, { "epoch": 5.03, "learning_rate": 6.380478179335846e-06, "loss": 0.0115, "step": 27455 }, { "epoch": 5.04, "learning_rate": 6.379786098519736e-06, "loss": 0.0115, "step": 27456 }, { "epoch": 5.04, "learning_rate": 6.379094037658331e-06, "loss": 0.0107, "step": 27457 }, { "epoch": 5.04, "learning_rate": 6.378401996755446e-06, "loss": 0.0102, "step": 27458 }, { "epoch": 5.04, "learning_rate": 6.3777099758148874e-06, "loss": 0.0094, "step": 27459 }, { "epoch": 5.04, "learning_rate": 6.377017974840482e-06, "loss": 0.0097, "step": 27460 }, { "epoch": 5.04, "learning_rate": 6.376325993836039e-06, "loss": 0.0092, "step": 27461 }, { "epoch": 5.04, "learning_rate": 6.375634032805369e-06, "loss": 0.0095, "step": 27462 }, { "epoch": 5.04, "learning_rate": 6.374942091752289e-06, "loss": 0.0117, "step": 27463 }, { "epoch": 5.04, "learning_rate": 6.374250170680613e-06, "loss": 0.0109, "step": 27464 }, { "epoch": 5.04, "learning_rate": 6.373558269594158e-06, "loss": 0.0091, "step": 27465 }, { "epoch": 5.04, "learning_rate": 6.372866388496732e-06, "loss": 0.0114, "step": 27466 }, { "epoch": 5.04, "learning_rate": 6.372174527392155e-06, "loss": 0.0078, "step": 27467 }, { "epoch": 5.04, "learning_rate": 6.371482686284231e-06, "loss": 0.0098, "step": 27468 }, { "epoch": 5.04, "learning_rate": 6.3707908651767845e-06, "loss": 0.0111, "step": 27469 }, { "epoch": 5.04, "learning_rate": 6.370099064073623e-06, "loss": 0.0112, "step": 27470 }, { "epoch": 5.04, "learning_rate": 6.369407282978559e-06, "loss": 0.0122, "step": 27471 }, { "epoch": 5.04, "learning_rate": 6.3687155218954066e-06, "loss": 0.0126, "step": 27472 }, { "epoch": 5.04, "learning_rate": 6.368023780827981e-06, "loss": 0.0084, "step": 27473 }, { "epoch": 5.04, "learning_rate": 6.367332059780092e-06, "loss": 0.0094, "step": 27474 }, { "epoch": 5.04, "learning_rate": 6.366640358755556e-06, "loss": 0.0085, "step": 27475 }, { "epoch": 5.04, "learning_rate": 6.365948677758185e-06, "loss": 0.0083, "step": 27476 }, { "epoch": 5.04, "learning_rate": 6.365257016791782e-06, "loss": 0.0087, "step": 27477 }, { "epoch": 5.04, "learning_rate": 6.364565375860174e-06, "loss": 0.0082, "step": 27478 }, { "epoch": 5.04, "learning_rate": 6.363873754967167e-06, "loss": 0.0079, "step": 27479 }, { "epoch": 5.04, "learning_rate": 6.363182154116572e-06, "loss": 0.0097, "step": 27480 }, { "epoch": 5.04, "learning_rate": 6.362490573312199e-06, "loss": 0.011, "step": 27481 }, { "epoch": 5.04, "learning_rate": 6.361799012557867e-06, "loss": 0.0088, "step": 27482 }, { "epoch": 5.04, "learning_rate": 6.361107471857382e-06, "loss": 0.009, "step": 27483 }, { "epoch": 5.04, "learning_rate": 6.36041595121456e-06, "loss": 0.0083, "step": 27484 }, { "epoch": 5.04, "learning_rate": 6.359724450633207e-06, "loss": 0.0125, "step": 27485 }, { "epoch": 5.04, "learning_rate": 6.359032970117141e-06, "loss": 0.0084, "step": 27486 }, { "epoch": 5.04, "learning_rate": 6.358341509670172e-06, "loss": 0.0105, "step": 27487 }, { "epoch": 5.04, "learning_rate": 6.3576500692961085e-06, "loss": 0.0098, "step": 27488 }, { "epoch": 5.04, "learning_rate": 6.356958648998762e-06, "loss": 0.0099, "step": 27489 }, { "epoch": 5.04, "learning_rate": 6.356267248781945e-06, "loss": 0.009, "step": 27490 }, { "epoch": 5.04, "learning_rate": 6.355575868649468e-06, "loss": 0.0096, "step": 27491 }, { "epoch": 5.04, "learning_rate": 6.354884508605142e-06, "loss": 0.0083, "step": 27492 }, { "epoch": 5.04, "learning_rate": 6.35419316865278e-06, "loss": 0.0104, "step": 27493 }, { "epoch": 5.04, "learning_rate": 6.353501848796187e-06, "loss": 0.0099, "step": 27494 }, { "epoch": 5.04, "learning_rate": 6.352810549039181e-06, "loss": 0.0108, "step": 27495 }, { "epoch": 5.04, "learning_rate": 6.352119269385569e-06, "loss": 0.0106, "step": 27496 }, { "epoch": 5.04, "learning_rate": 6.35142800983916e-06, "loss": 0.0091, "step": 27497 }, { "epoch": 5.04, "learning_rate": 6.350736770403765e-06, "loss": 0.0121, "step": 27498 }, { "epoch": 5.04, "learning_rate": 6.350045551083193e-06, "loss": 0.0104, "step": 27499 }, { "epoch": 5.04, "learning_rate": 6.349354351881256e-06, "loss": 0.0094, "step": 27500 }, { "epoch": 5.04, "learning_rate": 6.348663172801763e-06, "loss": 0.0086, "step": 27501 }, { "epoch": 5.04, "learning_rate": 6.347972013848527e-06, "loss": 0.0094, "step": 27502 }, { "epoch": 5.04, "learning_rate": 6.34728087502535e-06, "loss": 0.01, "step": 27503 }, { "epoch": 5.04, "learning_rate": 6.34658975633605e-06, "loss": 0.013, "step": 27504 }, { "epoch": 5.04, "learning_rate": 6.345898657784434e-06, "loss": 0.0081, "step": 27505 }, { "epoch": 5.04, "learning_rate": 6.345207579374308e-06, "loss": 0.0126, "step": 27506 }, { "epoch": 5.04, "learning_rate": 6.344516521109485e-06, "loss": 0.0105, "step": 27507 }, { "epoch": 5.04, "learning_rate": 6.343825482993769e-06, "loss": 0.0081, "step": 27508 }, { "epoch": 5.04, "learning_rate": 6.3431344650309754e-06, "loss": 0.0106, "step": 27509 }, { "epoch": 5.04, "learning_rate": 6.34244346722491e-06, "loss": 0.0092, "step": 27510 }, { "epoch": 5.05, "learning_rate": 6.341752489579382e-06, "loss": 0.0103, "step": 27511 }, { "epoch": 5.05, "learning_rate": 6.341061532098198e-06, "loss": 0.0096, "step": 27512 }, { "epoch": 5.05, "learning_rate": 6.34037059478517e-06, "loss": 0.0094, "step": 27513 }, { "epoch": 5.05, "learning_rate": 6.339679677644108e-06, "loss": 0.0099, "step": 27514 }, { "epoch": 5.05, "learning_rate": 6.338988780678814e-06, "loss": 0.009, "step": 27515 }, { "epoch": 5.05, "learning_rate": 6.338297903893099e-06, "loss": 0.0088, "step": 27516 }, { "epoch": 5.05, "learning_rate": 6.337607047290774e-06, "loss": 0.0103, "step": 27517 }, { "epoch": 5.05, "learning_rate": 6.336916210875644e-06, "loss": 0.0105, "step": 27518 }, { "epoch": 5.05, "learning_rate": 6.3362253946515166e-06, "loss": 0.0095, "step": 27519 }, { "epoch": 5.05, "learning_rate": 6.335534598622202e-06, "loss": 0.0109, "step": 27520 }, { "epoch": 5.05, "learning_rate": 6.334843822791505e-06, "loss": 0.008, "step": 27521 }, { "epoch": 5.05, "learning_rate": 6.3341530671632355e-06, "loss": 0.0093, "step": 27522 }, { "epoch": 5.05, "learning_rate": 6.333462331741202e-06, "loss": 0.0091, "step": 27523 }, { "epoch": 5.05, "learning_rate": 6.332771616529209e-06, "loss": 0.0098, "step": 27524 }, { "epoch": 5.05, "learning_rate": 6.332080921531063e-06, "loss": 0.0134, "step": 27525 }, { "epoch": 5.05, "learning_rate": 6.331390246750574e-06, "loss": 0.0086, "step": 27526 }, { "epoch": 5.05, "learning_rate": 6.3306995921915485e-06, "loss": 0.0104, "step": 27527 }, { "epoch": 5.05, "learning_rate": 6.330008957857792e-06, "loss": 0.0136, "step": 27528 }, { "epoch": 5.05, "learning_rate": 6.329318343753115e-06, "loss": 0.0087, "step": 27529 }, { "epoch": 5.05, "learning_rate": 6.3286277498813145e-06, "loss": 0.0087, "step": 27530 }, { "epoch": 5.05, "learning_rate": 6.327937176246208e-06, "loss": 0.009, "step": 27531 }, { "epoch": 5.05, "learning_rate": 6.3272466228516e-06, "loss": 0.0095, "step": 27532 }, { "epoch": 5.05, "learning_rate": 6.3265560897012924e-06, "loss": 0.0103, "step": 27533 }, { "epoch": 5.05, "learning_rate": 6.325865576799093e-06, "loss": 0.0085, "step": 27534 }, { "epoch": 5.05, "learning_rate": 6.325175084148809e-06, "loss": 0.0081, "step": 27535 }, { "epoch": 5.05, "learning_rate": 6.324484611754247e-06, "loss": 0.0105, "step": 27536 }, { "epoch": 5.05, "learning_rate": 6.3237941596192124e-06, "loss": 0.0092, "step": 27537 }, { "epoch": 5.05, "learning_rate": 6.323103727747507e-06, "loss": 0.009, "step": 27538 }, { "epoch": 5.05, "learning_rate": 6.322413316142942e-06, "loss": 0.0108, "step": 27539 }, { "epoch": 5.05, "learning_rate": 6.321722924809322e-06, "loss": 0.0097, "step": 27540 }, { "epoch": 5.05, "learning_rate": 6.321032553750453e-06, "loss": 0.0097, "step": 27541 }, { "epoch": 5.05, "learning_rate": 6.320342202970138e-06, "loss": 0.0122, "step": 27542 }, { "epoch": 5.05, "learning_rate": 6.319651872472179e-06, "loss": 0.0088, "step": 27543 }, { "epoch": 5.05, "learning_rate": 6.318961562260388e-06, "loss": 0.0093, "step": 27544 }, { "epoch": 5.05, "learning_rate": 6.318271272338568e-06, "loss": 0.0079, "step": 27545 }, { "epoch": 5.05, "learning_rate": 6.3175810027105215e-06, "loss": 0.0078, "step": 27546 }, { "epoch": 5.05, "learning_rate": 6.316890753380053e-06, "loss": 0.0089, "step": 27547 }, { "epoch": 5.05, "learning_rate": 6.31620052435097e-06, "loss": 0.0079, "step": 27548 }, { "epoch": 5.05, "learning_rate": 6.315510315627078e-06, "loss": 0.0089, "step": 27549 }, { "epoch": 5.05, "learning_rate": 6.314820127212181e-06, "loss": 0.0112, "step": 27550 }, { "epoch": 5.05, "learning_rate": 6.314129959110081e-06, "loss": 0.0083, "step": 27551 }, { "epoch": 5.05, "learning_rate": 6.313439811324578e-06, "loss": 0.0074, "step": 27552 }, { "epoch": 5.05, "learning_rate": 6.312749683859485e-06, "loss": 0.0125, "step": 27553 }, { "epoch": 5.05, "learning_rate": 6.312059576718601e-06, "loss": 0.0136, "step": 27554 }, { "epoch": 5.05, "learning_rate": 6.311369489905732e-06, "loss": 0.0087, "step": 27555 }, { "epoch": 5.05, "learning_rate": 6.3106794234246794e-06, "loss": 0.0094, "step": 27556 }, { "epoch": 5.05, "learning_rate": 6.309989377279249e-06, "loss": 0.0103, "step": 27557 }, { "epoch": 5.05, "learning_rate": 6.309299351473244e-06, "loss": 0.0116, "step": 27558 }, { "epoch": 5.05, "learning_rate": 6.30860934601047e-06, "loss": 0.01, "step": 27559 }, { "epoch": 5.05, "learning_rate": 6.307919360894725e-06, "loss": 0.0091, "step": 27560 }, { "epoch": 5.05, "learning_rate": 6.307229396129813e-06, "loss": 0.0082, "step": 27561 }, { "epoch": 5.05, "learning_rate": 6.306539451719541e-06, "loss": 0.0089, "step": 27562 }, { "epoch": 5.05, "learning_rate": 6.3058495276677114e-06, "loss": 0.0103, "step": 27563 }, { "epoch": 5.05, "learning_rate": 6.305159623978123e-06, "loss": 0.0094, "step": 27564 }, { "epoch": 5.06, "learning_rate": 6.3044697406545814e-06, "loss": 0.0098, "step": 27565 }, { "epoch": 5.06, "learning_rate": 6.30377987770089e-06, "loss": 0.0095, "step": 27566 }, { "epoch": 5.06, "learning_rate": 6.303090035120855e-06, "loss": 0.0083, "step": 27567 }, { "epoch": 5.06, "learning_rate": 6.3024002129182705e-06, "loss": 0.0091, "step": 27568 }, { "epoch": 5.06, "learning_rate": 6.301710411096939e-06, "loss": 0.0149, "step": 27569 }, { "epoch": 5.06, "learning_rate": 6.301020629660673e-06, "loss": 0.0097, "step": 27570 }, { "epoch": 5.06, "learning_rate": 6.300330868613264e-06, "loss": 0.0079, "step": 27571 }, { "epoch": 5.06, "learning_rate": 6.299641127958519e-06, "loss": 0.0069, "step": 27572 }, { "epoch": 5.06, "learning_rate": 6.298951407700239e-06, "loss": 0.0108, "step": 27573 }, { "epoch": 5.06, "learning_rate": 6.298261707842224e-06, "loss": 0.0076, "step": 27574 }, { "epoch": 5.06, "learning_rate": 6.297572028388277e-06, "loss": 0.009, "step": 27575 }, { "epoch": 5.06, "learning_rate": 6.2968823693422035e-06, "loss": 0.0098, "step": 27576 }, { "epoch": 5.06, "learning_rate": 6.2961927307078e-06, "loss": 0.0083, "step": 27577 }, { "epoch": 5.06, "learning_rate": 6.295503112488863e-06, "loss": 0.0119, "step": 27578 }, { "epoch": 5.06, "learning_rate": 6.294813514689206e-06, "loss": 0.0101, "step": 27579 }, { "epoch": 5.06, "learning_rate": 6.294123937312622e-06, "loss": 0.0095, "step": 27580 }, { "epoch": 5.06, "learning_rate": 6.293434380362915e-06, "loss": 0.009, "step": 27581 }, { "epoch": 5.06, "learning_rate": 6.292744843843883e-06, "loss": 0.0094, "step": 27582 }, { "epoch": 5.06, "learning_rate": 6.292055327759327e-06, "loss": 0.0117, "step": 27583 }, { "epoch": 5.06, "learning_rate": 6.29136583211305e-06, "loss": 0.0094, "step": 27584 }, { "epoch": 5.06, "learning_rate": 6.290676356908854e-06, "loss": 0.0098, "step": 27585 }, { "epoch": 5.06, "learning_rate": 6.289986902150536e-06, "loss": 0.0095, "step": 27586 }, { "epoch": 5.06, "learning_rate": 6.289297467841891e-06, "loss": 0.0121, "step": 27587 }, { "epoch": 5.06, "learning_rate": 6.288608053986732e-06, "loss": 0.0067, "step": 27588 }, { "epoch": 5.06, "learning_rate": 6.287918660588851e-06, "loss": 0.0127, "step": 27589 }, { "epoch": 5.06, "learning_rate": 6.287229287652048e-06, "loss": 0.0127, "step": 27590 }, { "epoch": 5.06, "learning_rate": 6.286539935180124e-06, "loss": 0.0072, "step": 27591 }, { "epoch": 5.06, "learning_rate": 6.28585060317688e-06, "loss": 0.0115, "step": 27592 }, { "epoch": 5.06, "learning_rate": 6.285161291646113e-06, "loss": 0.011, "step": 27593 }, { "epoch": 5.06, "learning_rate": 6.284472000591629e-06, "loss": 0.0077, "step": 27594 }, { "epoch": 5.06, "learning_rate": 6.283782730017218e-06, "loss": 0.0097, "step": 27595 }, { "epoch": 5.06, "learning_rate": 6.283093479926681e-06, "loss": 0.0114, "step": 27596 }, { "epoch": 5.06, "learning_rate": 6.282404250323825e-06, "loss": 0.0117, "step": 27597 }, { "epoch": 5.06, "learning_rate": 6.281715041212442e-06, "loss": 0.0107, "step": 27598 }, { "epoch": 5.06, "learning_rate": 6.281025852596332e-06, "loss": 0.0095, "step": 27599 }, { "epoch": 5.06, "learning_rate": 6.280336684479293e-06, "loss": 0.0091, "step": 27600 }, { "epoch": 5.06, "learning_rate": 6.279647536865129e-06, "loss": 0.0092, "step": 27601 }, { "epoch": 5.06, "learning_rate": 6.278958409757634e-06, "loss": 0.0115, "step": 27602 }, { "epoch": 5.06, "learning_rate": 6.278269303160609e-06, "loss": 0.0085, "step": 27603 }, { "epoch": 5.06, "learning_rate": 6.27758021707785e-06, "loss": 0.0105, "step": 27604 }, { "epoch": 5.06, "learning_rate": 6.27689115151315e-06, "loss": 0.0077, "step": 27605 }, { "epoch": 5.06, "learning_rate": 6.27620210647032e-06, "loss": 0.0125, "step": 27606 }, { "epoch": 5.06, "learning_rate": 6.275513081953151e-06, "loss": 0.0079, "step": 27607 }, { "epoch": 5.06, "learning_rate": 6.274824077965438e-06, "loss": 0.0114, "step": 27608 }, { "epoch": 5.06, "learning_rate": 6.274135094510982e-06, "loss": 0.0116, "step": 27609 }, { "epoch": 5.06, "learning_rate": 6.2734461315935826e-06, "loss": 0.0094, "step": 27610 }, { "epoch": 5.06, "learning_rate": 6.272757189217034e-06, "loss": 0.0073, "step": 27611 }, { "epoch": 5.06, "learning_rate": 6.272068267385139e-06, "loss": 0.012, "step": 27612 }, { "epoch": 5.06, "learning_rate": 6.271379366101684e-06, "loss": 0.0117, "step": 27613 }, { "epoch": 5.06, "learning_rate": 6.2706904853704796e-06, "loss": 0.0104, "step": 27614 }, { "epoch": 5.06, "learning_rate": 6.270001625195314e-06, "loss": 0.0081, "step": 27615 }, { "epoch": 5.06, "learning_rate": 6.269312785579988e-06, "loss": 0.0105, "step": 27616 }, { "epoch": 5.06, "learning_rate": 6.2686239665282965e-06, "loss": 0.0096, "step": 27617 }, { "epoch": 5.06, "learning_rate": 6.267935168044035e-06, "loss": 0.012, "step": 27618 }, { "epoch": 5.06, "learning_rate": 6.267246390131005e-06, "loss": 0.0091, "step": 27619 }, { "epoch": 5.07, "learning_rate": 6.266557632793e-06, "loss": 0.0112, "step": 27620 }, { "epoch": 5.07, "learning_rate": 6.265868896033818e-06, "loss": 0.0097, "step": 27621 }, { "epoch": 5.07, "learning_rate": 6.265180179857249e-06, "loss": 0.0091, "step": 27622 }, { "epoch": 5.07, "learning_rate": 6.2644914842671e-06, "loss": 0.0097, "step": 27623 }, { "epoch": 5.07, "learning_rate": 6.26380280926716e-06, "loss": 0.0107, "step": 27624 }, { "epoch": 5.07, "learning_rate": 6.263114154861226e-06, "loss": 0.0094, "step": 27625 }, { "epoch": 5.07, "learning_rate": 6.262425521053095e-06, "loss": 0.0109, "step": 27626 }, { "epoch": 5.07, "learning_rate": 6.2617369078465585e-06, "loss": 0.0074, "step": 27627 }, { "epoch": 5.07, "learning_rate": 6.261048315245419e-06, "loss": 0.0084, "step": 27628 }, { "epoch": 5.07, "learning_rate": 6.260359743253468e-06, "loss": 0.0074, "step": 27629 }, { "epoch": 5.07, "learning_rate": 6.2596711918745045e-06, "loss": 0.0103, "step": 27630 }, { "epoch": 5.07, "learning_rate": 6.258982661112315e-06, "loss": 0.0083, "step": 27631 }, { "epoch": 5.07, "learning_rate": 6.2582941509707066e-06, "loss": 0.0077, "step": 27632 }, { "epoch": 5.07, "learning_rate": 6.257605661453466e-06, "loss": 0.0108, "step": 27633 }, { "epoch": 5.07, "learning_rate": 6.2569171925643915e-06, "loss": 0.0114, "step": 27634 }, { "epoch": 5.07, "learning_rate": 6.256228744307275e-06, "loss": 0.0105, "step": 27635 }, { "epoch": 5.07, "learning_rate": 6.255540316685914e-06, "loss": 0.0084, "step": 27636 }, { "epoch": 5.07, "learning_rate": 6.254851909704104e-06, "loss": 0.0078, "step": 27637 }, { "epoch": 5.07, "learning_rate": 6.254163523365637e-06, "loss": 0.0097, "step": 27638 }, { "epoch": 5.07, "learning_rate": 6.253475157674311e-06, "loss": 0.014, "step": 27639 }, { "epoch": 5.07, "learning_rate": 6.252786812633912e-06, "loss": 0.0111, "step": 27640 }, { "epoch": 5.07, "learning_rate": 6.252098488248246e-06, "loss": 0.0088, "step": 27641 }, { "epoch": 5.07, "learning_rate": 6.2514101845210986e-06, "loss": 0.0113, "step": 27642 }, { "epoch": 5.07, "learning_rate": 6.250721901456265e-06, "loss": 0.0118, "step": 27643 }, { "epoch": 5.07, "learning_rate": 6.250033639057539e-06, "loss": 0.0117, "step": 27644 }, { "epoch": 5.07, "learning_rate": 6.249345397328718e-06, "loss": 0.0123, "step": 27645 }, { "epoch": 5.07, "learning_rate": 6.248657176273591e-06, "loss": 0.0124, "step": 27646 }, { "epoch": 5.07, "learning_rate": 6.247968975895956e-06, "loss": 0.0086, "step": 27647 }, { "epoch": 5.07, "learning_rate": 6.2472807961996055e-06, "loss": 0.0102, "step": 27648 }, { "epoch": 5.07, "learning_rate": 6.246592637188324e-06, "loss": 0.0111, "step": 27649 }, { "epoch": 5.07, "learning_rate": 6.245904498865918e-06, "loss": 0.009, "step": 27650 }, { "epoch": 5.07, "learning_rate": 6.245216381236173e-06, "loss": 0.0087, "step": 27651 }, { "epoch": 5.07, "learning_rate": 6.2445282843028835e-06, "loss": 0.0071, "step": 27652 }, { "epoch": 5.07, "learning_rate": 6.24384020806984e-06, "loss": 0.0082, "step": 27653 }, { "epoch": 5.07, "learning_rate": 6.243152152540839e-06, "loss": 0.0095, "step": 27654 }, { "epoch": 5.07, "learning_rate": 6.242464117719671e-06, "loss": 0.0074, "step": 27655 }, { "epoch": 5.07, "learning_rate": 6.241776103610128e-06, "loss": 0.0087, "step": 27656 }, { "epoch": 5.07, "learning_rate": 6.241088110216007e-06, "loss": 0.0107, "step": 27657 }, { "epoch": 5.07, "learning_rate": 6.240400137541089e-06, "loss": 0.0082, "step": 27658 }, { "epoch": 5.07, "learning_rate": 6.239712185589181e-06, "loss": 0.007, "step": 27659 }, { "epoch": 5.07, "learning_rate": 6.239024254364064e-06, "loss": 0.0116, "step": 27660 }, { "epoch": 5.07, "learning_rate": 6.238336343869535e-06, "loss": 0.009, "step": 27661 }, { "epoch": 5.07, "learning_rate": 6.237648454109381e-06, "loss": 0.0089, "step": 27662 }, { "epoch": 5.07, "learning_rate": 6.2369605850874e-06, "loss": 0.0087, "step": 27663 }, { "epoch": 5.07, "learning_rate": 6.236272736807378e-06, "loss": 0.0083, "step": 27664 }, { "epoch": 5.07, "learning_rate": 6.235584909273112e-06, "loss": 0.0071, "step": 27665 }, { "epoch": 5.07, "learning_rate": 6.234897102488386e-06, "loss": 0.0137, "step": 27666 }, { "epoch": 5.07, "learning_rate": 6.234209316456997e-06, "loss": 0.0103, "step": 27667 }, { "epoch": 5.07, "learning_rate": 6.233521551182738e-06, "loss": 0.0111, "step": 27668 }, { "epoch": 5.07, "learning_rate": 6.232833806669394e-06, "loss": 0.011, "step": 27669 }, { "epoch": 5.07, "learning_rate": 6.232146082920758e-06, "loss": 0.0106, "step": 27670 }, { "epoch": 5.07, "learning_rate": 6.231458379940619e-06, "loss": 0.0081, "step": 27671 }, { "epoch": 5.07, "learning_rate": 6.230770697732772e-06, "loss": 0.0076, "step": 27672 }, { "epoch": 5.07, "learning_rate": 6.230083036301005e-06, "loss": 0.0096, "step": 27673 }, { "epoch": 5.08, "learning_rate": 6.22939539564911e-06, "loss": 0.0082, "step": 27674 }, { "epoch": 5.08, "learning_rate": 6.228707775780871e-06, "loss": 0.0128, "step": 27675 }, { "epoch": 5.08, "learning_rate": 6.228020176700087e-06, "loss": 0.0115, "step": 27676 }, { "epoch": 5.08, "learning_rate": 6.227332598410547e-06, "loss": 0.0074, "step": 27677 }, { "epoch": 5.08, "learning_rate": 6.226645040916035e-06, "loss": 0.0081, "step": 27678 }, { "epoch": 5.08, "learning_rate": 6.225957504220343e-06, "loss": 0.0104, "step": 27679 }, { "epoch": 5.08, "learning_rate": 6.225269988327261e-06, "loss": 0.0065, "step": 27680 }, { "epoch": 5.08, "learning_rate": 6.224582493240581e-06, "loss": 0.0086, "step": 27681 }, { "epoch": 5.08, "learning_rate": 6.22389501896409e-06, "loss": 0.0082, "step": 27682 }, { "epoch": 5.08, "learning_rate": 6.223207565501581e-06, "loss": 0.0093, "step": 27683 }, { "epoch": 5.08, "learning_rate": 6.2225201328568356e-06, "loss": 0.0088, "step": 27684 }, { "epoch": 5.08, "learning_rate": 6.221832721033649e-06, "loss": 0.0105, "step": 27685 }, { "epoch": 5.08, "learning_rate": 6.221145330035813e-06, "loss": 0.0105, "step": 27686 }, { "epoch": 5.08, "learning_rate": 6.22045795986711e-06, "loss": 0.0053, "step": 27687 }, { "epoch": 5.08, "learning_rate": 6.21977061053133e-06, "loss": 0.0077, "step": 27688 }, { "epoch": 5.08, "learning_rate": 6.219083282032262e-06, "loss": 0.0078, "step": 27689 }, { "epoch": 5.08, "learning_rate": 6.218395974373698e-06, "loss": 0.0112, "step": 27690 }, { "epoch": 5.08, "learning_rate": 6.2177086875594225e-06, "loss": 0.0092, "step": 27691 }, { "epoch": 5.08, "learning_rate": 6.217021421593229e-06, "loss": 0.0097, "step": 27692 }, { "epoch": 5.08, "learning_rate": 6.216334176478895e-06, "loss": 0.011, "step": 27693 }, { "epoch": 5.08, "learning_rate": 6.21564695222022e-06, "loss": 0.0093, "step": 27694 }, { "epoch": 5.08, "learning_rate": 6.21495974882099e-06, "loss": 0.009, "step": 27695 }, { "epoch": 5.08, "learning_rate": 6.214272566284988e-06, "loss": 0.0089, "step": 27696 }, { "epoch": 5.08, "learning_rate": 6.213585404616002e-06, "loss": 0.0083, "step": 27697 }, { "epoch": 5.08, "learning_rate": 6.212898263817825e-06, "loss": 0.0091, "step": 27698 }, { "epoch": 5.08, "learning_rate": 6.21221114389424e-06, "loss": 0.0125, "step": 27699 }, { "epoch": 5.08, "learning_rate": 6.211524044849038e-06, "loss": 0.0125, "step": 27700 }, { "epoch": 5.08, "learning_rate": 6.210836966686005e-06, "loss": 0.009, "step": 27701 }, { "epoch": 5.08, "learning_rate": 6.210149909408921e-06, "loss": 0.0108, "step": 27702 }, { "epoch": 5.08, "learning_rate": 6.209462873021583e-06, "loss": 0.0082, "step": 27703 }, { "epoch": 5.08, "learning_rate": 6.2087758575277775e-06, "loss": 0.0115, "step": 27704 }, { "epoch": 5.08, "learning_rate": 6.208088862931288e-06, "loss": 0.0111, "step": 27705 }, { "epoch": 5.08, "learning_rate": 6.207401889235897e-06, "loss": 0.0097, "step": 27706 }, { "epoch": 5.08, "learning_rate": 6.206714936445398e-06, "loss": 0.0087, "step": 27707 }, { "epoch": 5.08, "learning_rate": 6.206028004563576e-06, "loss": 0.0084, "step": 27708 }, { "epoch": 5.08, "learning_rate": 6.205341093594216e-06, "loss": 0.0089, "step": 27709 }, { "epoch": 5.08, "learning_rate": 6.204654203541108e-06, "loss": 0.0125, "step": 27710 }, { "epoch": 5.08, "learning_rate": 6.2039673344080265e-06, "loss": 0.0105, "step": 27711 }, { "epoch": 5.08, "learning_rate": 6.203280486198771e-06, "loss": 0.0085, "step": 27712 }, { "epoch": 5.08, "learning_rate": 6.202593658917124e-06, "loss": 0.0112, "step": 27713 }, { "epoch": 5.08, "learning_rate": 6.201906852566869e-06, "loss": 0.0075, "step": 27714 }, { "epoch": 5.08, "learning_rate": 6.20122006715179e-06, "loss": 0.0092, "step": 27715 }, { "epoch": 5.08, "learning_rate": 6.2005333026756766e-06, "loss": 0.0095, "step": 27716 }, { "epoch": 5.08, "learning_rate": 6.199846559142313e-06, "loss": 0.0126, "step": 27717 }, { "epoch": 5.08, "learning_rate": 6.199159836555484e-06, "loss": 0.0091, "step": 27718 }, { "epoch": 5.08, "learning_rate": 6.198473134918972e-06, "loss": 0.0145, "step": 27719 }, { "epoch": 5.08, "learning_rate": 6.19778645423657e-06, "loss": 0.0086, "step": 27720 }, { "epoch": 5.08, "learning_rate": 6.197099794512056e-06, "loss": 0.0083, "step": 27721 }, { "epoch": 5.08, "learning_rate": 6.196413155749219e-06, "loss": 0.0083, "step": 27722 }, { "epoch": 5.08, "learning_rate": 6.1957265379518396e-06, "loss": 0.0085, "step": 27723 }, { "epoch": 5.08, "learning_rate": 6.195039941123704e-06, "loss": 0.0125, "step": 27724 }, { "epoch": 5.08, "learning_rate": 6.194353365268597e-06, "loss": 0.0101, "step": 27725 }, { "epoch": 5.08, "learning_rate": 6.193666810390306e-06, "loss": 0.0117, "step": 27726 }, { "epoch": 5.08, "learning_rate": 6.1929802764926105e-06, "loss": 0.0111, "step": 27727 }, { "epoch": 5.08, "learning_rate": 6.192293763579297e-06, "loss": 0.0106, "step": 27728 }, { "epoch": 5.09, "learning_rate": 6.19160727165415e-06, "loss": 0.0087, "step": 27729 }, { "epoch": 5.09, "learning_rate": 6.190920800720957e-06, "loss": 0.0083, "step": 27730 }, { "epoch": 5.09, "learning_rate": 6.190234350783493e-06, "loss": 0.0089, "step": 27731 }, { "epoch": 5.09, "learning_rate": 6.1895479218455475e-06, "loss": 0.008, "step": 27732 }, { "epoch": 5.09, "learning_rate": 6.188861513910902e-06, "loss": 0.008, "step": 27733 }, { "epoch": 5.09, "learning_rate": 6.188175126983343e-06, "loss": 0.0118, "step": 27734 }, { "epoch": 5.09, "learning_rate": 6.187488761066651e-06, "loss": 0.0074, "step": 27735 }, { "epoch": 5.09, "learning_rate": 6.18680241616461e-06, "loss": 0.011, "step": 27736 }, { "epoch": 5.09, "learning_rate": 6.186116092281003e-06, "loss": 0.0119, "step": 27737 }, { "epoch": 5.09, "learning_rate": 6.185429789419615e-06, "loss": 0.0095, "step": 27738 }, { "epoch": 5.09, "learning_rate": 6.184743507584229e-06, "loss": 0.0089, "step": 27739 }, { "epoch": 5.09, "learning_rate": 6.184057246778624e-06, "loss": 0.012, "step": 27740 }, { "epoch": 5.09, "learning_rate": 6.183371007006581e-06, "loss": 0.0106, "step": 27741 }, { "epoch": 5.09, "learning_rate": 6.182684788271893e-06, "loss": 0.0085, "step": 27742 }, { "epoch": 5.09, "learning_rate": 6.181998590578333e-06, "loss": 0.0124, "step": 27743 }, { "epoch": 5.09, "learning_rate": 6.181312413929687e-06, "loss": 0.0103, "step": 27744 }, { "epoch": 5.09, "learning_rate": 6.180626258329736e-06, "loss": 0.0087, "step": 27745 }, { "epoch": 5.09, "learning_rate": 6.17994012378226e-06, "loss": 0.0086, "step": 27746 }, { "epoch": 5.09, "learning_rate": 6.179254010291046e-06, "loss": 0.0103, "step": 27747 }, { "epoch": 5.09, "learning_rate": 6.178567917859876e-06, "loss": 0.0099, "step": 27748 }, { "epoch": 5.09, "learning_rate": 6.177881846492526e-06, "loss": 0.012, "step": 27749 }, { "epoch": 5.09, "learning_rate": 6.177195796192777e-06, "loss": 0.0119, "step": 27750 }, { "epoch": 5.09, "learning_rate": 6.176509766964422e-06, "loss": 0.01, "step": 27751 }, { "epoch": 5.09, "learning_rate": 6.17582375881123e-06, "loss": 0.0097, "step": 27752 }, { "epoch": 5.09, "learning_rate": 6.17513777173699e-06, "loss": 0.0082, "step": 27753 }, { "epoch": 5.09, "learning_rate": 6.174451805745478e-06, "loss": 0.0119, "step": 27754 }, { "epoch": 5.09, "learning_rate": 6.173765860840477e-06, "loss": 0.0094, "step": 27755 }, { "epoch": 5.09, "learning_rate": 6.173079937025769e-06, "loss": 0.0099, "step": 27756 }, { "epoch": 5.09, "learning_rate": 6.172394034305136e-06, "loss": 0.0091, "step": 27757 }, { "epoch": 5.09, "learning_rate": 6.171708152682355e-06, "loss": 0.0084, "step": 27758 }, { "epoch": 5.09, "learning_rate": 6.171022292161204e-06, "loss": 0.0092, "step": 27759 }, { "epoch": 5.09, "learning_rate": 6.170336452745474e-06, "loss": 0.011, "step": 27760 }, { "epoch": 5.09, "learning_rate": 6.169650634438937e-06, "loss": 0.011, "step": 27761 }, { "epoch": 5.09, "learning_rate": 6.168964837245377e-06, "loss": 0.0073, "step": 27762 }, { "epoch": 5.09, "learning_rate": 6.1682790611685715e-06, "loss": 0.0098, "step": 27763 }, { "epoch": 5.09, "learning_rate": 6.1675933062123e-06, "loss": 0.0121, "step": 27764 }, { "epoch": 5.09, "learning_rate": 6.166907572380345e-06, "loss": 0.0109, "step": 27765 }, { "epoch": 5.09, "learning_rate": 6.166221859676488e-06, "loss": 0.008, "step": 27766 }, { "epoch": 5.09, "learning_rate": 6.165536168104503e-06, "loss": 0.0091, "step": 27767 }, { "epoch": 5.09, "learning_rate": 6.164850497668171e-06, "loss": 0.0087, "step": 27768 }, { "epoch": 5.09, "learning_rate": 6.1641648483712755e-06, "loss": 0.0088, "step": 27769 }, { "epoch": 5.09, "learning_rate": 6.163479220217592e-06, "loss": 0.0099, "step": 27770 }, { "epoch": 5.09, "learning_rate": 6.162793613210902e-06, "loss": 0.0108, "step": 27771 }, { "epoch": 5.09, "learning_rate": 6.162108027354982e-06, "loss": 0.0076, "step": 27772 }, { "epoch": 5.09, "learning_rate": 6.161422462653613e-06, "loss": 0.0094, "step": 27773 }, { "epoch": 5.09, "learning_rate": 6.160736919110575e-06, "loss": 0.0096, "step": 27774 }, { "epoch": 5.09, "learning_rate": 6.160051396729645e-06, "loss": 0.009, "step": 27775 }, { "epoch": 5.09, "learning_rate": 6.159365895514601e-06, "loss": 0.0089, "step": 27776 }, { "epoch": 5.09, "learning_rate": 6.1586804154692205e-06, "loss": 0.0094, "step": 27777 }, { "epoch": 5.09, "learning_rate": 6.157994956597284e-06, "loss": 0.0082, "step": 27778 }, { "epoch": 5.09, "learning_rate": 6.15730951890257e-06, "loss": 0.008, "step": 27779 }, { "epoch": 5.09, "learning_rate": 6.156624102388858e-06, "loss": 0.0092, "step": 27780 }, { "epoch": 5.09, "learning_rate": 6.155938707059919e-06, "loss": 0.0097, "step": 27781 }, { "epoch": 5.09, "learning_rate": 6.155253332919539e-06, "loss": 0.0108, "step": 27782 }, { "epoch": 5.09, "learning_rate": 6.154567979971493e-06, "loss": 0.0091, "step": 27783 }, { "epoch": 5.1, "learning_rate": 6.1538826482195605e-06, "loss": 0.0114, "step": 27784 }, { "epoch": 5.1, "learning_rate": 6.153197337667513e-06, "loss": 0.0106, "step": 27785 }, { "epoch": 5.1, "learning_rate": 6.152512048319133e-06, "loss": 0.0081, "step": 27786 }, { "epoch": 5.1, "learning_rate": 6.1518267801781965e-06, "loss": 0.0117, "step": 27787 }, { "epoch": 5.1, "learning_rate": 6.151141533248481e-06, "loss": 0.0103, "step": 27788 }, { "epoch": 5.1, "learning_rate": 6.150456307533765e-06, "loss": 0.0068, "step": 27789 }, { "epoch": 5.1, "learning_rate": 6.149771103037821e-06, "loss": 0.0087, "step": 27790 }, { "epoch": 5.1, "learning_rate": 6.149085919764431e-06, "loss": 0.0079, "step": 27791 }, { "epoch": 5.1, "learning_rate": 6.1484007577173695e-06, "loss": 0.0095, "step": 27792 }, { "epoch": 5.1, "learning_rate": 6.147715616900416e-06, "loss": 0.007, "step": 27793 }, { "epoch": 5.1, "learning_rate": 6.147030497317336e-06, "loss": 0.0108, "step": 27794 }, { "epoch": 5.1, "learning_rate": 6.1463453989719225e-06, "loss": 0.0089, "step": 27795 }, { "epoch": 5.1, "learning_rate": 6.145660321867941e-06, "loss": 0.0085, "step": 27796 }, { "epoch": 5.1, "learning_rate": 6.14497526600917e-06, "loss": 0.0078, "step": 27797 }, { "epoch": 5.1, "learning_rate": 6.144290231399386e-06, "loss": 0.0111, "step": 27798 }, { "epoch": 5.1, "learning_rate": 6.143605218042362e-06, "loss": 0.0111, "step": 27799 }, { "epoch": 5.1, "learning_rate": 6.142920225941879e-06, "loss": 0.008, "step": 27800 }, { "epoch": 5.1, "learning_rate": 6.142235255101709e-06, "loss": 0.006, "step": 27801 }, { "epoch": 5.1, "learning_rate": 6.141550305525632e-06, "loss": 0.0103, "step": 27802 }, { "epoch": 5.1, "learning_rate": 6.140865377217413e-06, "loss": 0.0093, "step": 27803 }, { "epoch": 5.1, "learning_rate": 6.140180470180841e-06, "loss": 0.0082, "step": 27804 }, { "epoch": 5.1, "learning_rate": 6.139495584419682e-06, "loss": 0.0081, "step": 27805 }, { "epoch": 5.1, "learning_rate": 6.138810719937715e-06, "loss": 0.0096, "step": 27806 }, { "epoch": 5.1, "learning_rate": 6.138125876738712e-06, "loss": 0.0106, "step": 27807 }, { "epoch": 5.1, "learning_rate": 6.13744105482645e-06, "loss": 0.0086, "step": 27808 }, { "epoch": 5.1, "learning_rate": 6.1367562542047035e-06, "loss": 0.0072, "step": 27809 }, { "epoch": 5.1, "learning_rate": 6.1360714748772475e-06, "loss": 0.0123, "step": 27810 }, { "epoch": 5.1, "learning_rate": 6.135386716847857e-06, "loss": 0.0086, "step": 27811 }, { "epoch": 5.1, "learning_rate": 6.1347019801203014e-06, "loss": 0.0087, "step": 27812 }, { "epoch": 5.1, "learning_rate": 6.134017264698364e-06, "loss": 0.0107, "step": 27813 }, { "epoch": 5.1, "learning_rate": 6.133332570585813e-06, "loss": 0.0085, "step": 27814 }, { "epoch": 5.1, "learning_rate": 6.1326478977864235e-06, "loss": 0.0093, "step": 27815 }, { "epoch": 5.1, "learning_rate": 6.131963246303968e-06, "loss": 0.008, "step": 27816 }, { "epoch": 5.1, "learning_rate": 6.13127861614222e-06, "loss": 0.008, "step": 27817 }, { "epoch": 5.1, "learning_rate": 6.1305940073049575e-06, "loss": 0.0094, "step": 27818 }, { "epoch": 5.1, "learning_rate": 6.129909419795952e-06, "loss": 0.0091, "step": 27819 }, { "epoch": 5.1, "learning_rate": 6.129224853618977e-06, "loss": 0.0076, "step": 27820 }, { "epoch": 5.1, "learning_rate": 6.1285403087778e-06, "loss": 0.0103, "step": 27821 }, { "epoch": 5.1, "learning_rate": 6.127855785276206e-06, "loss": 0.0092, "step": 27822 }, { "epoch": 5.1, "learning_rate": 6.127171283117958e-06, "loss": 0.0096, "step": 27823 }, { "epoch": 5.1, "learning_rate": 6.126486802306835e-06, "loss": 0.0093, "step": 27824 }, { "epoch": 5.1, "learning_rate": 6.125802342846603e-06, "loss": 0.0097, "step": 27825 }, { "epoch": 5.1, "learning_rate": 6.125117904741043e-06, "loss": 0.0122, "step": 27826 }, { "epoch": 5.1, "learning_rate": 6.124433487993924e-06, "loss": 0.0096, "step": 27827 }, { "epoch": 5.1, "learning_rate": 6.123749092609017e-06, "loss": 0.0101, "step": 27828 }, { "epoch": 5.1, "learning_rate": 6.123064718590099e-06, "loss": 0.011, "step": 27829 }, { "epoch": 5.1, "learning_rate": 6.122380365940932e-06, "loss": 0.0112, "step": 27830 }, { "epoch": 5.1, "learning_rate": 6.121696034665302e-06, "loss": 0.0094, "step": 27831 }, { "epoch": 5.1, "learning_rate": 6.121011724766972e-06, "loss": 0.0096, "step": 27832 }, { "epoch": 5.1, "learning_rate": 6.120327436249717e-06, "loss": 0.0087, "step": 27833 }, { "epoch": 5.1, "learning_rate": 6.119643169117306e-06, "loss": 0.0087, "step": 27834 }, { "epoch": 5.1, "learning_rate": 6.118958923373514e-06, "loss": 0.0137, "step": 27835 }, { "epoch": 5.1, "learning_rate": 6.118274699022111e-06, "loss": 0.0094, "step": 27836 }, { "epoch": 5.1, "learning_rate": 6.117590496066872e-06, "loss": 0.0107, "step": 27837 }, { "epoch": 5.11, "learning_rate": 6.116906314511563e-06, "loss": 0.0075, "step": 27838 }, { "epoch": 5.11, "learning_rate": 6.116222154359952e-06, "loss": 0.0087, "step": 27839 }, { "epoch": 5.11, "learning_rate": 6.115538015615823e-06, "loss": 0.0074, "step": 27840 }, { "epoch": 5.11, "learning_rate": 6.114853898282938e-06, "loss": 0.0114, "step": 27841 }, { "epoch": 5.11, "learning_rate": 6.114169802365066e-06, "loss": 0.0088, "step": 27842 }, { "epoch": 5.11, "learning_rate": 6.113485727865982e-06, "loss": 0.0072, "step": 27843 }, { "epoch": 5.11, "learning_rate": 6.112801674789457e-06, "loss": 0.0124, "step": 27844 }, { "epoch": 5.11, "learning_rate": 6.112117643139261e-06, "loss": 0.0076, "step": 27845 }, { "epoch": 5.11, "learning_rate": 6.111433632919163e-06, "loss": 0.0089, "step": 27846 }, { "epoch": 5.11, "learning_rate": 6.1107496441329315e-06, "loss": 0.0102, "step": 27847 }, { "epoch": 5.11, "learning_rate": 6.110065676784341e-06, "loss": 0.0089, "step": 27848 }, { "epoch": 5.11, "learning_rate": 6.109381730877162e-06, "loss": 0.0109, "step": 27849 }, { "epoch": 5.11, "learning_rate": 6.10869780641516e-06, "loss": 0.0097, "step": 27850 }, { "epoch": 5.11, "learning_rate": 6.108013903402107e-06, "loss": 0.0085, "step": 27851 }, { "epoch": 5.11, "learning_rate": 6.107330021841771e-06, "loss": 0.0094, "step": 27852 }, { "epoch": 5.11, "learning_rate": 6.106646161737923e-06, "loss": 0.0078, "step": 27853 }, { "epoch": 5.11, "learning_rate": 6.105962323094335e-06, "loss": 0.0085, "step": 27854 }, { "epoch": 5.11, "learning_rate": 6.105278505914775e-06, "loss": 0.0106, "step": 27855 }, { "epoch": 5.11, "learning_rate": 6.1045947102030055e-06, "loss": 0.0103, "step": 27856 }, { "epoch": 5.11, "learning_rate": 6.103910935962804e-06, "loss": 0.0108, "step": 27857 }, { "epoch": 5.11, "learning_rate": 6.103227183197941e-06, "loss": 0.0083, "step": 27858 }, { "epoch": 5.11, "learning_rate": 6.102543451912177e-06, "loss": 0.0089, "step": 27859 }, { "epoch": 5.11, "learning_rate": 6.1018597421092866e-06, "loss": 0.0126, "step": 27860 }, { "epoch": 5.11, "learning_rate": 6.101176053793032e-06, "loss": 0.0108, "step": 27861 }, { "epoch": 5.11, "learning_rate": 6.100492386967189e-06, "loss": 0.0105, "step": 27862 }, { "epoch": 5.11, "learning_rate": 6.0998087416355245e-06, "loss": 0.0076, "step": 27863 }, { "epoch": 5.11, "learning_rate": 6.0991251178018075e-06, "loss": 0.0076, "step": 27864 }, { "epoch": 5.11, "learning_rate": 6.098441515469798e-06, "loss": 0.0068, "step": 27865 }, { "epoch": 5.11, "learning_rate": 6.0977579346432745e-06, "loss": 0.0122, "step": 27866 }, { "epoch": 5.11, "learning_rate": 6.097074375326002e-06, "loss": 0.0117, "step": 27867 }, { "epoch": 5.11, "learning_rate": 6.096390837521746e-06, "loss": 0.0112, "step": 27868 }, { "epoch": 5.11, "learning_rate": 6.095707321234272e-06, "loss": 0.0099, "step": 27869 }, { "epoch": 5.11, "learning_rate": 6.0950238264673546e-06, "loss": 0.0113, "step": 27870 }, { "epoch": 5.11, "learning_rate": 6.094340353224755e-06, "loss": 0.0088, "step": 27871 }, { "epoch": 5.11, "learning_rate": 6.093656901510243e-06, "loss": 0.0078, "step": 27872 }, { "epoch": 5.11, "learning_rate": 6.09297347132759e-06, "loss": 0.0089, "step": 27873 }, { "epoch": 5.11, "learning_rate": 6.09229006268055e-06, "loss": 0.0097, "step": 27874 }, { "epoch": 5.11, "learning_rate": 6.091606675572905e-06, "loss": 0.0085, "step": 27875 }, { "epoch": 5.11, "learning_rate": 6.090923310008417e-06, "loss": 0.0098, "step": 27876 }, { "epoch": 5.11, "learning_rate": 6.090239965990849e-06, "loss": 0.0122, "step": 27877 }, { "epoch": 5.11, "learning_rate": 6.0895566435239674e-06, "loss": 0.0103, "step": 27878 }, { "epoch": 5.11, "learning_rate": 6.088873342611543e-06, "loss": 0.0101, "step": 27879 }, { "epoch": 5.11, "learning_rate": 6.088190063257341e-06, "loss": 0.0108, "step": 27880 }, { "epoch": 5.11, "learning_rate": 6.087506805465127e-06, "loss": 0.0103, "step": 27881 }, { "epoch": 5.11, "learning_rate": 6.086823569238669e-06, "loss": 0.0103, "step": 27882 }, { "epoch": 5.11, "learning_rate": 6.086140354581726e-06, "loss": 0.0074, "step": 27883 }, { "epoch": 5.11, "learning_rate": 6.085457161498074e-06, "loss": 0.012, "step": 27884 }, { "epoch": 5.11, "learning_rate": 6.0847739899914725e-06, "loss": 0.0109, "step": 27885 }, { "epoch": 5.11, "learning_rate": 6.084090840065689e-06, "loss": 0.0101, "step": 27886 }, { "epoch": 5.11, "learning_rate": 6.083407711724485e-06, "loss": 0.0073, "step": 27887 }, { "epoch": 5.11, "learning_rate": 6.082724604971632e-06, "loss": 0.0151, "step": 27888 }, { "epoch": 5.11, "learning_rate": 6.0820415198108925e-06, "loss": 0.0124, "step": 27889 }, { "epoch": 5.11, "learning_rate": 6.081358456246032e-06, "loss": 0.0098, "step": 27890 }, { "epoch": 5.11, "learning_rate": 6.080675414280818e-06, "loss": 0.0101, "step": 27891 }, { "epoch": 5.11, "learning_rate": 6.079992393919006e-06, "loss": 0.0108, "step": 27892 }, { "epoch": 5.12, "learning_rate": 6.079309395164374e-06, "loss": 0.0107, "step": 27893 }, { "epoch": 5.12, "learning_rate": 6.07862641802068e-06, "loss": 0.0067, "step": 27894 }, { "epoch": 5.12, "learning_rate": 6.077943462491689e-06, "loss": 0.0134, "step": 27895 }, { "epoch": 5.12, "learning_rate": 6.077260528581161e-06, "loss": 0.0099, "step": 27896 }, { "epoch": 5.12, "learning_rate": 6.076577616292869e-06, "loss": 0.0101, "step": 27897 }, { "epoch": 5.12, "learning_rate": 6.075894725630574e-06, "loss": 0.0106, "step": 27898 }, { "epoch": 5.12, "learning_rate": 6.0752118565980354e-06, "loss": 0.0098, "step": 27899 }, { "epoch": 5.12, "learning_rate": 6.074529009199022e-06, "loss": 0.0113, "step": 27900 }, { "epoch": 5.12, "learning_rate": 6.073846183437297e-06, "loss": 0.0094, "step": 27901 }, { "epoch": 5.12, "learning_rate": 6.073163379316628e-06, "loss": 0.0097, "step": 27902 }, { "epoch": 5.12, "learning_rate": 6.072480596840771e-06, "loss": 0.0075, "step": 27903 }, { "epoch": 5.12, "learning_rate": 6.0717978360134935e-06, "loss": 0.0088, "step": 27904 }, { "epoch": 5.12, "learning_rate": 6.071115096838556e-06, "loss": 0.0099, "step": 27905 }, { "epoch": 5.12, "learning_rate": 6.070432379319726e-06, "loss": 0.0099, "step": 27906 }, { "epoch": 5.12, "learning_rate": 6.069749683460765e-06, "loss": 0.0098, "step": 27907 }, { "epoch": 5.12, "learning_rate": 6.069067009265436e-06, "loss": 0.0097, "step": 27908 }, { "epoch": 5.12, "learning_rate": 6.0683843567375e-06, "loss": 0.0096, "step": 27909 }, { "epoch": 5.12, "learning_rate": 6.067701725880723e-06, "loss": 0.0097, "step": 27910 }, { "epoch": 5.12, "learning_rate": 6.067019116698868e-06, "loss": 0.0105, "step": 27911 }, { "epoch": 5.12, "learning_rate": 6.066336529195693e-06, "loss": 0.0118, "step": 27912 }, { "epoch": 5.12, "learning_rate": 6.065653963374965e-06, "loss": 0.0117, "step": 27913 }, { "epoch": 5.12, "learning_rate": 6.0649714192404415e-06, "loss": 0.0085, "step": 27914 }, { "epoch": 5.12, "learning_rate": 6.06428889679589e-06, "loss": 0.0081, "step": 27915 }, { "epoch": 5.12, "learning_rate": 6.0636063960450695e-06, "loss": 0.0118, "step": 27916 }, { "epoch": 5.12, "learning_rate": 6.062923916991745e-06, "loss": 0.0085, "step": 27917 }, { "epoch": 5.12, "learning_rate": 6.062241459639672e-06, "loss": 0.0119, "step": 27918 }, { "epoch": 5.12, "learning_rate": 6.061559023992619e-06, "loss": 0.0108, "step": 27919 }, { "epoch": 5.12, "learning_rate": 6.060876610054347e-06, "loss": 0.0104, "step": 27920 }, { "epoch": 5.12, "learning_rate": 6.060194217828613e-06, "loss": 0.0106, "step": 27921 }, { "epoch": 5.12, "learning_rate": 6.059511847319177e-06, "loss": 0.0121, "step": 27922 }, { "epoch": 5.12, "learning_rate": 6.05882949852981e-06, "loss": 0.0118, "step": 27923 }, { "epoch": 5.12, "learning_rate": 6.058147171464266e-06, "loss": 0.0097, "step": 27924 }, { "epoch": 5.12, "learning_rate": 6.057464866126306e-06, "loss": 0.0112, "step": 27925 }, { "epoch": 5.12, "learning_rate": 6.0567825825196936e-06, "loss": 0.0098, "step": 27926 }, { "epoch": 5.12, "learning_rate": 6.056100320648184e-06, "loss": 0.0101, "step": 27927 }, { "epoch": 5.12, "learning_rate": 6.055418080515545e-06, "loss": 0.0094, "step": 27928 }, { "epoch": 5.12, "learning_rate": 6.054735862125536e-06, "loss": 0.0112, "step": 27929 }, { "epoch": 5.12, "learning_rate": 6.054053665481912e-06, "loss": 0.0078, "step": 27930 }, { "epoch": 5.12, "learning_rate": 6.0533714905884355e-06, "loss": 0.0073, "step": 27931 }, { "epoch": 5.12, "learning_rate": 6.052689337448869e-06, "loss": 0.0093, "step": 27932 }, { "epoch": 5.12, "learning_rate": 6.052007206066973e-06, "loss": 0.0072, "step": 27933 }, { "epoch": 5.12, "learning_rate": 6.051325096446505e-06, "loss": 0.0076, "step": 27934 }, { "epoch": 5.12, "learning_rate": 6.050643008591226e-06, "loss": 0.0075, "step": 27935 }, { "epoch": 5.12, "learning_rate": 6.049960942504893e-06, "loss": 0.0121, "step": 27936 }, { "epoch": 5.12, "learning_rate": 6.049278898191269e-06, "loss": 0.0086, "step": 27937 }, { "epoch": 5.12, "learning_rate": 6.048596875654116e-06, "loss": 0.0081, "step": 27938 }, { "epoch": 5.12, "learning_rate": 6.047914874897185e-06, "loss": 0.0075, "step": 27939 }, { "epoch": 5.12, "learning_rate": 6.04723289592424e-06, "loss": 0.01, "step": 27940 }, { "epoch": 5.12, "learning_rate": 6.046550938739042e-06, "loss": 0.0102, "step": 27941 }, { "epoch": 5.12, "learning_rate": 6.045869003345346e-06, "loss": 0.0103, "step": 27942 }, { "epoch": 5.12, "learning_rate": 6.045187089746913e-06, "loss": 0.0116, "step": 27943 }, { "epoch": 5.12, "learning_rate": 6.044505197947503e-06, "loss": 0.0083, "step": 27944 }, { "epoch": 5.12, "learning_rate": 6.043823327950871e-06, "loss": 0.0092, "step": 27945 }, { "epoch": 5.12, "learning_rate": 6.043141479760778e-06, "loss": 0.0088, "step": 27946 }, { "epoch": 5.13, "learning_rate": 6.042459653380985e-06, "loss": 0.008, "step": 27947 }, { "epoch": 5.13, "learning_rate": 6.041777848815244e-06, "loss": 0.0114, "step": 27948 }, { "epoch": 5.13, "learning_rate": 6.041096066067316e-06, "loss": 0.0104, "step": 27949 }, { "epoch": 5.13, "learning_rate": 6.040414305140959e-06, "loss": 0.007, "step": 27950 }, { "epoch": 5.13, "learning_rate": 6.039732566039932e-06, "loss": 0.0087, "step": 27951 }, { "epoch": 5.13, "learning_rate": 6.039050848767992e-06, "loss": 0.0116, "step": 27952 }, { "epoch": 5.13, "learning_rate": 6.038369153328894e-06, "loss": 0.0092, "step": 27953 }, { "epoch": 5.13, "learning_rate": 6.037687479726402e-06, "loss": 0.0094, "step": 27954 }, { "epoch": 5.13, "learning_rate": 6.0370058279642674e-06, "loss": 0.0079, "step": 27955 }, { "epoch": 5.13, "learning_rate": 6.036324198046253e-06, "loss": 0.0087, "step": 27956 }, { "epoch": 5.13, "learning_rate": 6.035642589976109e-06, "loss": 0.0106, "step": 27957 }, { "epoch": 5.13, "learning_rate": 6.034961003757595e-06, "loss": 0.0097, "step": 27958 }, { "epoch": 5.13, "learning_rate": 6.0342794393944705e-06, "loss": 0.0076, "step": 27959 }, { "epoch": 5.13, "learning_rate": 6.033597896890491e-06, "loss": 0.0095, "step": 27960 }, { "epoch": 5.13, "learning_rate": 6.032916376249412e-06, "loss": 0.0087, "step": 27961 }, { "epoch": 5.13, "learning_rate": 6.03223487747499e-06, "loss": 0.0141, "step": 27962 }, { "epoch": 5.13, "learning_rate": 6.031553400570982e-06, "loss": 0.0066, "step": 27963 }, { "epoch": 5.13, "learning_rate": 6.030871945541146e-06, "loss": 0.0097, "step": 27964 }, { "epoch": 5.13, "learning_rate": 6.03019051238924e-06, "loss": 0.011, "step": 27965 }, { "epoch": 5.13, "learning_rate": 6.0295091011190135e-06, "loss": 0.0103, "step": 27966 }, { "epoch": 5.13, "learning_rate": 6.028827711734224e-06, "loss": 0.0087, "step": 27967 }, { "epoch": 5.13, "learning_rate": 6.028146344238632e-06, "loss": 0.0145, "step": 27968 }, { "epoch": 5.13, "learning_rate": 6.02746499863599e-06, "loss": 0.0098, "step": 27969 }, { "epoch": 5.13, "learning_rate": 6.0267836749300535e-06, "loss": 0.0089, "step": 27970 }, { "epoch": 5.13, "learning_rate": 6.026102373124576e-06, "loss": 0.0075, "step": 27971 }, { "epoch": 5.13, "learning_rate": 6.025421093223318e-06, "loss": 0.009, "step": 27972 }, { "epoch": 5.13, "learning_rate": 6.024739835230034e-06, "loss": 0.0086, "step": 27973 }, { "epoch": 5.13, "learning_rate": 6.024058599148477e-06, "loss": 0.013, "step": 27974 }, { "epoch": 5.13, "learning_rate": 6.023377384982396e-06, "loss": 0.0084, "step": 27975 }, { "epoch": 5.13, "learning_rate": 6.022696192735558e-06, "loss": 0.0112, "step": 27976 }, { "epoch": 5.13, "learning_rate": 6.022015022411711e-06, "loss": 0.0116, "step": 27977 }, { "epoch": 5.13, "learning_rate": 6.02133387401461e-06, "loss": 0.0111, "step": 27978 }, { "epoch": 5.13, "learning_rate": 6.020652747548008e-06, "loss": 0.016, "step": 27979 }, { "epoch": 5.13, "learning_rate": 6.019971643015662e-06, "loss": 0.0094, "step": 27980 }, { "epoch": 5.13, "learning_rate": 6.019290560421326e-06, "loss": 0.009, "step": 27981 }, { "epoch": 5.13, "learning_rate": 6.018609499768754e-06, "loss": 0.0087, "step": 27982 }, { "epoch": 5.13, "learning_rate": 6.0179284610617015e-06, "loss": 0.008, "step": 27983 }, { "epoch": 5.13, "learning_rate": 6.017247444303915e-06, "loss": 0.0089, "step": 27984 }, { "epoch": 5.13, "learning_rate": 6.01656644949916e-06, "loss": 0.0105, "step": 27985 }, { "epoch": 5.13, "learning_rate": 6.015885476651182e-06, "loss": 0.0082, "step": 27986 }, { "epoch": 5.13, "learning_rate": 6.0152045257637355e-06, "loss": 0.0072, "step": 27987 }, { "epoch": 5.13, "learning_rate": 6.014523596840577e-06, "loss": 0.0078, "step": 27988 }, { "epoch": 5.13, "learning_rate": 6.0138426898854544e-06, "loss": 0.011, "step": 27989 }, { "epoch": 5.13, "learning_rate": 6.013161804902127e-06, "loss": 0.0073, "step": 27990 }, { "epoch": 5.13, "learning_rate": 6.012480941894346e-06, "loss": 0.0103, "step": 27991 }, { "epoch": 5.13, "learning_rate": 6.011800100865863e-06, "loss": 0.0093, "step": 27992 }, { "epoch": 5.13, "learning_rate": 6.011119281820426e-06, "loss": 0.0097, "step": 27993 }, { "epoch": 5.13, "learning_rate": 6.010438484761801e-06, "loss": 0.0126, "step": 27994 }, { "epoch": 5.13, "learning_rate": 6.009757709693728e-06, "loss": 0.0095, "step": 27995 }, { "epoch": 5.13, "learning_rate": 6.009076956619965e-06, "loss": 0.0101, "step": 27996 }, { "epoch": 5.13, "learning_rate": 6.008396225544261e-06, "loss": 0.0084, "step": 27997 }, { "epoch": 5.13, "learning_rate": 6.007715516470373e-06, "loss": 0.0127, "step": 27998 }, { "epoch": 5.13, "learning_rate": 6.0070348294020496e-06, "loss": 0.0082, "step": 27999 }, { "epoch": 5.13, "learning_rate": 6.006354164343047e-06, "loss": 0.0108, "step": 28000 }, { "epoch": 5.13, "learning_rate": 6.005673521297112e-06, "loss": 0.0105, "step": 28001 }, { "epoch": 5.14, "learning_rate": 6.004992900267993e-06, "loss": 0.007, "step": 28002 }, { "epoch": 5.14, "learning_rate": 6.004312301259454e-06, "loss": 0.0105, "step": 28003 }, { "epoch": 5.14, "learning_rate": 6.0036317242752365e-06, "loss": 0.0116, "step": 28004 }, { "epoch": 5.14, "learning_rate": 6.0029511693190936e-06, "loss": 0.0073, "step": 28005 }, { "epoch": 5.14, "learning_rate": 6.002270636394776e-06, "loss": 0.0094, "step": 28006 }, { "epoch": 5.14, "learning_rate": 6.0015901255060394e-06, "loss": 0.0107, "step": 28007 }, { "epoch": 5.14, "learning_rate": 6.000909636656631e-06, "loss": 0.0084, "step": 28008 }, { "epoch": 5.14, "learning_rate": 6.000229169850304e-06, "loss": 0.014, "step": 28009 }, { "epoch": 5.14, "learning_rate": 5.9995487250908065e-06, "loss": 0.0104, "step": 28010 }, { "epoch": 5.14, "learning_rate": 5.998868302381885e-06, "loss": 0.0101, "step": 28011 }, { "epoch": 5.14, "learning_rate": 5.998187901727301e-06, "loss": 0.0083, "step": 28012 }, { "epoch": 5.14, "learning_rate": 5.997507523130799e-06, "loss": 0.0101, "step": 28013 }, { "epoch": 5.14, "learning_rate": 5.996827166596129e-06, "loss": 0.0117, "step": 28014 }, { "epoch": 5.14, "learning_rate": 5.996146832127037e-06, "loss": 0.0116, "step": 28015 }, { "epoch": 5.14, "learning_rate": 5.995466519727281e-06, "loss": 0.0086, "step": 28016 }, { "epoch": 5.14, "learning_rate": 5.994786229400607e-06, "loss": 0.0076, "step": 28017 }, { "epoch": 5.14, "learning_rate": 5.994105961150768e-06, "loss": 0.0106, "step": 28018 }, { "epoch": 5.14, "learning_rate": 5.993425714981509e-06, "loss": 0.0109, "step": 28019 }, { "epoch": 5.14, "learning_rate": 5.992745490896577e-06, "loss": 0.009, "step": 28020 }, { "epoch": 5.14, "learning_rate": 5.99206528889973e-06, "loss": 0.0099, "step": 28021 }, { "epoch": 5.14, "learning_rate": 5.991385108994712e-06, "loss": 0.0073, "step": 28022 }, { "epoch": 5.14, "learning_rate": 5.990704951185273e-06, "loss": 0.0127, "step": 28023 }, { "epoch": 5.14, "learning_rate": 5.990024815475162e-06, "loss": 0.0118, "step": 28024 }, { "epoch": 5.14, "learning_rate": 5.98934470186813e-06, "loss": 0.0083, "step": 28025 }, { "epoch": 5.14, "learning_rate": 5.9886646103679216e-06, "loss": 0.0109, "step": 28026 }, { "epoch": 5.14, "learning_rate": 5.987984540978291e-06, "loss": 0.0114, "step": 28027 }, { "epoch": 5.14, "learning_rate": 5.9873044937029776e-06, "loss": 0.0091, "step": 28028 }, { "epoch": 5.14, "learning_rate": 5.986624468545741e-06, "loss": 0.0106, "step": 28029 }, { "epoch": 5.14, "learning_rate": 5.985944465510325e-06, "loss": 0.0096, "step": 28030 }, { "epoch": 5.14, "learning_rate": 5.985264484600475e-06, "loss": 0.0093, "step": 28031 }, { "epoch": 5.14, "learning_rate": 5.9845845258199405e-06, "loss": 0.008, "step": 28032 }, { "epoch": 5.14, "learning_rate": 5.983904589172471e-06, "loss": 0.0126, "step": 28033 }, { "epoch": 5.14, "learning_rate": 5.983224674661814e-06, "loss": 0.0063, "step": 28034 }, { "epoch": 5.14, "learning_rate": 5.982544782291716e-06, "loss": 0.0077, "step": 28035 }, { "epoch": 5.14, "learning_rate": 5.981864912065928e-06, "loss": 0.0113, "step": 28036 }, { "epoch": 5.14, "learning_rate": 5.981185063988189e-06, "loss": 0.0111, "step": 28037 }, { "epoch": 5.14, "learning_rate": 5.980505238062257e-06, "loss": 0.0094, "step": 28038 }, { "epoch": 5.14, "learning_rate": 5.979825434291873e-06, "loss": 0.0101, "step": 28039 }, { "epoch": 5.14, "learning_rate": 5.979145652680784e-06, "loss": 0.0114, "step": 28040 }, { "epoch": 5.14, "learning_rate": 5.978465893232741e-06, "loss": 0.0107, "step": 28041 }, { "epoch": 5.14, "learning_rate": 5.977786155951484e-06, "loss": 0.008, "step": 28042 }, { "epoch": 5.14, "learning_rate": 5.977106440840768e-06, "loss": 0.012, "step": 28043 }, { "epoch": 5.14, "learning_rate": 5.976426747904334e-06, "loss": 0.0072, "step": 28044 }, { "epoch": 5.14, "learning_rate": 5.9757470771459326e-06, "loss": 0.0096, "step": 28045 }, { "epoch": 5.14, "learning_rate": 5.975067428569301e-06, "loss": 0.0083, "step": 28046 }, { "epoch": 5.14, "learning_rate": 5.974387802178198e-06, "loss": 0.0127, "step": 28047 }, { "epoch": 5.14, "learning_rate": 5.973708197976362e-06, "loss": 0.0088, "step": 28048 }, { "epoch": 5.14, "learning_rate": 5.973028615967542e-06, "loss": 0.0086, "step": 28049 }, { "epoch": 5.14, "learning_rate": 5.9723490561554785e-06, "loss": 0.0154, "step": 28050 }, { "epoch": 5.14, "learning_rate": 5.9716695185439244e-06, "loss": 0.0089, "step": 28051 }, { "epoch": 5.14, "learning_rate": 5.970990003136623e-06, "loss": 0.0081, "step": 28052 }, { "epoch": 5.14, "learning_rate": 5.970310509937318e-06, "loss": 0.0074, "step": 28053 }, { "epoch": 5.14, "learning_rate": 5.969631038949758e-06, "loss": 0.0096, "step": 28054 }, { "epoch": 5.14, "learning_rate": 5.96895159017768e-06, "loss": 0.0108, "step": 28055 }, { "epoch": 5.15, "learning_rate": 5.968272163624842e-06, "loss": 0.0074, "step": 28056 }, { "epoch": 5.15, "learning_rate": 5.96759275929498e-06, "loss": 0.0085, "step": 28057 }, { "epoch": 5.15, "learning_rate": 5.966913377191841e-06, "loss": 0.0088, "step": 28058 }, { "epoch": 5.15, "learning_rate": 5.966234017319168e-06, "loss": 0.0093, "step": 28059 }, { "epoch": 5.15, "learning_rate": 5.965554679680709e-06, "loss": 0.0093, "step": 28060 }, { "epoch": 5.15, "learning_rate": 5.964875364280209e-06, "loss": 0.0101, "step": 28061 }, { "epoch": 5.15, "learning_rate": 5.964196071121409e-06, "loss": 0.0117, "step": 28062 }, { "epoch": 5.15, "learning_rate": 5.963516800208056e-06, "loss": 0.0107, "step": 28063 }, { "epoch": 5.15, "learning_rate": 5.9628375515438894e-06, "loss": 0.0091, "step": 28064 }, { "epoch": 5.15, "learning_rate": 5.962158325132661e-06, "loss": 0.0105, "step": 28065 }, { "epoch": 5.15, "learning_rate": 5.961479120978109e-06, "loss": 0.0067, "step": 28066 }, { "epoch": 5.15, "learning_rate": 5.960799939083979e-06, "loss": 0.0078, "step": 28067 }, { "epoch": 5.15, "learning_rate": 5.960120779454011e-06, "loss": 0.0081, "step": 28068 }, { "epoch": 5.15, "learning_rate": 5.959441642091956e-06, "loss": 0.0085, "step": 28069 }, { "epoch": 5.15, "learning_rate": 5.9587625270015515e-06, "loss": 0.0117, "step": 28070 }, { "epoch": 5.15, "learning_rate": 5.958083434186543e-06, "loss": 0.0108, "step": 28071 }, { "epoch": 5.15, "learning_rate": 5.957404363650676e-06, "loss": 0.0085, "step": 28072 }, { "epoch": 5.15, "learning_rate": 5.956725315397684e-06, "loss": 0.0089, "step": 28073 }, { "epoch": 5.15, "learning_rate": 5.956046289431324e-06, "loss": 0.0079, "step": 28074 }, { "epoch": 5.15, "learning_rate": 5.955367285755329e-06, "loss": 0.0081, "step": 28075 }, { "epoch": 5.15, "learning_rate": 5.954688304373444e-06, "loss": 0.0078, "step": 28076 }, { "epoch": 5.15, "learning_rate": 5.954009345289411e-06, "loss": 0.0103, "step": 28077 }, { "epoch": 5.15, "learning_rate": 5.953330408506974e-06, "loss": 0.0081, "step": 28078 }, { "epoch": 5.15, "learning_rate": 5.952651494029875e-06, "loss": 0.0111, "step": 28079 }, { "epoch": 5.15, "learning_rate": 5.951972601861855e-06, "loss": 0.0082, "step": 28080 }, { "epoch": 5.15, "learning_rate": 5.951293732006656e-06, "loss": 0.0111, "step": 28081 }, { "epoch": 5.15, "learning_rate": 5.950614884468023e-06, "loss": 0.0066, "step": 28082 }, { "epoch": 5.15, "learning_rate": 5.949936059249697e-06, "loss": 0.0103, "step": 28083 }, { "epoch": 5.15, "learning_rate": 5.949257256355415e-06, "loss": 0.0111, "step": 28084 }, { "epoch": 5.15, "learning_rate": 5.948578475788924e-06, "loss": 0.0093, "step": 28085 }, { "epoch": 5.15, "learning_rate": 5.9478997175539595e-06, "loss": 0.0126, "step": 28086 }, { "epoch": 5.15, "learning_rate": 5.9472209816542705e-06, "loss": 0.0099, "step": 28087 }, { "epoch": 5.15, "learning_rate": 5.946542268093594e-06, "loss": 0.0089, "step": 28088 }, { "epoch": 5.15, "learning_rate": 5.94586357687567e-06, "loss": 0.0065, "step": 28089 }, { "epoch": 5.15, "learning_rate": 5.94518490800424e-06, "loss": 0.0107, "step": 28090 }, { "epoch": 5.15, "learning_rate": 5.944506261483049e-06, "loss": 0.0088, "step": 28091 }, { "epoch": 5.15, "learning_rate": 5.943827637315836e-06, "loss": 0.0112, "step": 28092 }, { "epoch": 5.15, "learning_rate": 5.943149035506337e-06, "loss": 0.0108, "step": 28093 }, { "epoch": 5.15, "learning_rate": 5.942470456058296e-06, "loss": 0.009, "step": 28094 }, { "epoch": 5.15, "learning_rate": 5.941791898975451e-06, "loss": 0.0128, "step": 28095 }, { "epoch": 5.15, "learning_rate": 5.941113364261546e-06, "loss": 0.0089, "step": 28096 }, { "epoch": 5.15, "learning_rate": 5.9404348519203185e-06, "loss": 0.0091, "step": 28097 }, { "epoch": 5.15, "learning_rate": 5.93975636195551e-06, "loss": 0.0095, "step": 28098 }, { "epoch": 5.15, "learning_rate": 5.939077894370857e-06, "loss": 0.0055, "step": 28099 }, { "epoch": 5.15, "learning_rate": 5.938399449170104e-06, "loss": 0.0095, "step": 28100 }, { "epoch": 5.15, "learning_rate": 5.93772102635699e-06, "loss": 0.0126, "step": 28101 }, { "epoch": 5.15, "learning_rate": 5.937042625935252e-06, "loss": 0.0105, "step": 28102 }, { "epoch": 5.15, "learning_rate": 5.936364247908628e-06, "loss": 0.009, "step": 28103 }, { "epoch": 5.15, "learning_rate": 5.935685892280862e-06, "loss": 0.008, "step": 28104 }, { "epoch": 5.15, "learning_rate": 5.93500755905569e-06, "loss": 0.0101, "step": 28105 }, { "epoch": 5.15, "learning_rate": 5.934329248236851e-06, "loss": 0.0133, "step": 28106 }, { "epoch": 5.15, "learning_rate": 5.933650959828087e-06, "loss": 0.0134, "step": 28107 }, { "epoch": 5.15, "learning_rate": 5.932972693833128e-06, "loss": 0.0088, "step": 28108 }, { "epoch": 5.15, "learning_rate": 5.9322944502557226e-06, "loss": 0.0087, "step": 28109 }, { "epoch": 5.15, "learning_rate": 5.931616229099608e-06, "loss": 0.0077, "step": 28110 }, { "epoch": 5.16, "learning_rate": 5.930938030368517e-06, "loss": 0.0107, "step": 28111 }, { "epoch": 5.16, "learning_rate": 5.930259854066189e-06, "loss": 0.0089, "step": 28112 }, { "epoch": 5.16, "learning_rate": 5.929581700196368e-06, "loss": 0.0124, "step": 28113 }, { "epoch": 5.16, "learning_rate": 5.928903568762786e-06, "loss": 0.0095, "step": 28114 }, { "epoch": 5.16, "learning_rate": 5.928225459769183e-06, "loss": 0.0116, "step": 28115 }, { "epoch": 5.16, "learning_rate": 5.9275473732193004e-06, "loss": 0.008, "step": 28116 }, { "epoch": 5.16, "learning_rate": 5.926869309116865e-06, "loss": 0.009, "step": 28117 }, { "epoch": 5.16, "learning_rate": 5.926191267465625e-06, "loss": 0.0082, "step": 28118 }, { "epoch": 5.16, "learning_rate": 5.925513248269317e-06, "loss": 0.0102, "step": 28119 }, { "epoch": 5.16, "learning_rate": 5.924835251531673e-06, "loss": 0.0072, "step": 28120 }, { "epoch": 5.16, "learning_rate": 5.924157277256431e-06, "loss": 0.0098, "step": 28121 }, { "epoch": 5.16, "learning_rate": 5.92347932544733e-06, "loss": 0.0095, "step": 28122 }, { "epoch": 5.16, "learning_rate": 5.9228013961081075e-06, "loss": 0.0099, "step": 28123 }, { "epoch": 5.16, "learning_rate": 5.922123489242499e-06, "loss": 0.0115, "step": 28124 }, { "epoch": 5.16, "learning_rate": 5.921445604854239e-06, "loss": 0.0103, "step": 28125 }, { "epoch": 5.16, "learning_rate": 5.92076774294707e-06, "loss": 0.0087, "step": 28126 }, { "epoch": 5.16, "learning_rate": 5.920089903524724e-06, "loss": 0.0078, "step": 28127 }, { "epoch": 5.16, "learning_rate": 5.919412086590939e-06, "loss": 0.0096, "step": 28128 }, { "epoch": 5.16, "learning_rate": 5.91873429214945e-06, "loss": 0.0077, "step": 28129 }, { "epoch": 5.16, "learning_rate": 5.91805652020399e-06, "loss": 0.0079, "step": 28130 }, { "epoch": 5.16, "learning_rate": 5.9173787707583e-06, "loss": 0.0131, "step": 28131 }, { "epoch": 5.16, "learning_rate": 5.916701043816113e-06, "loss": 0.0081, "step": 28132 }, { "epoch": 5.16, "learning_rate": 5.916023339381167e-06, "loss": 0.009, "step": 28133 }, { "epoch": 5.16, "learning_rate": 5.915345657457194e-06, "loss": 0.0104, "step": 28134 }, { "epoch": 5.16, "learning_rate": 5.914667998047934e-06, "loss": 0.0077, "step": 28135 }, { "epoch": 5.16, "learning_rate": 5.913990361157118e-06, "loss": 0.0101, "step": 28136 }, { "epoch": 5.16, "learning_rate": 5.913312746788485e-06, "loss": 0.0077, "step": 28137 }, { "epoch": 5.16, "learning_rate": 5.9126351549457675e-06, "loss": 0.0085, "step": 28138 }, { "epoch": 5.16, "learning_rate": 5.911957585632697e-06, "loss": 0.0111, "step": 28139 }, { "epoch": 5.16, "learning_rate": 5.911280038853016e-06, "loss": 0.0108, "step": 28140 }, { "epoch": 5.16, "learning_rate": 5.910602514610453e-06, "loss": 0.0095, "step": 28141 }, { "epoch": 5.16, "learning_rate": 5.909925012908746e-06, "loss": 0.0092, "step": 28142 }, { "epoch": 5.16, "learning_rate": 5.909247533751626e-06, "loss": 0.0092, "step": 28143 }, { "epoch": 5.16, "learning_rate": 5.908570077142832e-06, "loss": 0.0101, "step": 28144 }, { "epoch": 5.16, "learning_rate": 5.907892643086095e-06, "loss": 0.0093, "step": 28145 }, { "epoch": 5.16, "learning_rate": 5.9072152315851505e-06, "loss": 0.012, "step": 28146 }, { "epoch": 5.16, "learning_rate": 5.9065378426437314e-06, "loss": 0.0084, "step": 28147 }, { "epoch": 5.16, "learning_rate": 5.905860476265568e-06, "loss": 0.0119, "step": 28148 }, { "epoch": 5.16, "learning_rate": 5.905183132454402e-06, "loss": 0.0092, "step": 28149 }, { "epoch": 5.16, "learning_rate": 5.9045058112139605e-06, "loss": 0.0087, "step": 28150 }, { "epoch": 5.16, "learning_rate": 5.903828512547979e-06, "loss": 0.01, "step": 28151 }, { "epoch": 5.16, "learning_rate": 5.90315123646019e-06, "loss": 0.0119, "step": 28152 }, { "epoch": 5.16, "learning_rate": 5.902473982954328e-06, "loss": 0.0079, "step": 28153 }, { "epoch": 5.16, "learning_rate": 5.901796752034128e-06, "loss": 0.0087, "step": 28154 }, { "epoch": 5.16, "learning_rate": 5.901119543703317e-06, "loss": 0.0116, "step": 28155 }, { "epoch": 5.16, "learning_rate": 5.900442357965627e-06, "loss": 0.0095, "step": 28156 }, { "epoch": 5.16, "learning_rate": 5.899765194824804e-06, "loss": 0.0108, "step": 28157 }, { "epoch": 5.16, "learning_rate": 5.8990880542845676e-06, "loss": 0.0103, "step": 28158 }, { "epoch": 5.16, "learning_rate": 5.898410936348653e-06, "loss": 0.0087, "step": 28159 }, { "epoch": 5.16, "learning_rate": 5.897733841020794e-06, "loss": 0.0073, "step": 28160 }, { "epoch": 5.16, "learning_rate": 5.89705676830472e-06, "loss": 0.0136, "step": 28161 }, { "epoch": 5.16, "learning_rate": 5.896379718204168e-06, "loss": 0.0082, "step": 28162 }, { "epoch": 5.16, "learning_rate": 5.8957026907228685e-06, "loss": 0.0106, "step": 28163 }, { "epoch": 5.16, "learning_rate": 5.89502568586455e-06, "loss": 0.0102, "step": 28164 }, { "epoch": 5.17, "learning_rate": 5.894348703632942e-06, "loss": 0.0125, "step": 28165 }, { "epoch": 5.17, "learning_rate": 5.893671744031786e-06, "loss": 0.0135, "step": 28166 }, { "epoch": 5.17, "learning_rate": 5.8929948070648044e-06, "loss": 0.0113, "step": 28167 }, { "epoch": 5.17, "learning_rate": 5.892317892735734e-06, "loss": 0.0074, "step": 28168 }, { "epoch": 5.17, "learning_rate": 5.8916410010483025e-06, "loss": 0.0091, "step": 28169 }, { "epoch": 5.17, "learning_rate": 5.890964132006241e-06, "loss": 0.0104, "step": 28170 }, { "epoch": 5.17, "learning_rate": 5.890287285613282e-06, "loss": 0.0124, "step": 28171 }, { "epoch": 5.17, "learning_rate": 5.889610461873159e-06, "loss": 0.0089, "step": 28172 }, { "epoch": 5.17, "learning_rate": 5.888933660789595e-06, "loss": 0.0122, "step": 28173 }, { "epoch": 5.17, "learning_rate": 5.888256882366323e-06, "loss": 0.0128, "step": 28174 }, { "epoch": 5.17, "learning_rate": 5.887580126607082e-06, "loss": 0.0103, "step": 28175 }, { "epoch": 5.17, "learning_rate": 5.886903393515593e-06, "loss": 0.0104, "step": 28176 }, { "epoch": 5.17, "learning_rate": 5.8862266830955885e-06, "loss": 0.0078, "step": 28177 }, { "epoch": 5.17, "learning_rate": 5.885549995350797e-06, "loss": 0.0093, "step": 28178 }, { "epoch": 5.17, "learning_rate": 5.884873330284953e-06, "loss": 0.0111, "step": 28179 }, { "epoch": 5.17, "learning_rate": 5.884196687901782e-06, "loss": 0.0095, "step": 28180 }, { "epoch": 5.17, "learning_rate": 5.8835200682050184e-06, "loss": 0.0112, "step": 28181 }, { "epoch": 5.17, "learning_rate": 5.8828434711983875e-06, "loss": 0.0111, "step": 28182 }, { "epoch": 5.17, "learning_rate": 5.882166896885615e-06, "loss": 0.0094, "step": 28183 }, { "epoch": 5.17, "learning_rate": 5.881490345270439e-06, "loss": 0.0098, "step": 28184 }, { "epoch": 5.17, "learning_rate": 5.8808138163565854e-06, "loss": 0.0076, "step": 28185 }, { "epoch": 5.17, "learning_rate": 5.880137310147782e-06, "loss": 0.0094, "step": 28186 }, { "epoch": 5.17, "learning_rate": 5.879460826647756e-06, "loss": 0.006, "step": 28187 }, { "epoch": 5.17, "learning_rate": 5.87878436586024e-06, "loss": 0.009, "step": 28188 }, { "epoch": 5.17, "learning_rate": 5.878107927788962e-06, "loss": 0.009, "step": 28189 }, { "epoch": 5.17, "learning_rate": 5.8774315124376516e-06, "loss": 0.008, "step": 28190 }, { "epoch": 5.17, "learning_rate": 5.8767551198100326e-06, "loss": 0.0089, "step": 28191 }, { "epoch": 5.17, "learning_rate": 5.876078749909835e-06, "loss": 0.0098, "step": 28192 }, { "epoch": 5.17, "learning_rate": 5.8754024027407885e-06, "loss": 0.0083, "step": 28193 }, { "epoch": 5.17, "learning_rate": 5.8747260783066215e-06, "loss": 0.0097, "step": 28194 }, { "epoch": 5.17, "learning_rate": 5.8740497766110616e-06, "loss": 0.0088, "step": 28195 }, { "epoch": 5.17, "learning_rate": 5.873373497657832e-06, "loss": 0.0099, "step": 28196 }, { "epoch": 5.17, "learning_rate": 5.872697241450669e-06, "loss": 0.01, "step": 28197 }, { "epoch": 5.17, "learning_rate": 5.872021007993294e-06, "loss": 0.0085, "step": 28198 }, { "epoch": 5.17, "learning_rate": 5.871344797289438e-06, "loss": 0.0118, "step": 28199 }, { "epoch": 5.17, "learning_rate": 5.870668609342823e-06, "loss": 0.0109, "step": 28200 }, { "epoch": 5.17, "learning_rate": 5.8699924441571795e-06, "loss": 0.0087, "step": 28201 }, { "epoch": 5.17, "learning_rate": 5.869316301736234e-06, "loss": 0.0095, "step": 28202 }, { "epoch": 5.17, "learning_rate": 5.868640182083716e-06, "loss": 0.0093, "step": 28203 }, { "epoch": 5.17, "learning_rate": 5.86796408520335e-06, "loss": 0.0103, "step": 28204 }, { "epoch": 5.17, "learning_rate": 5.867288011098859e-06, "loss": 0.0104, "step": 28205 }, { "epoch": 5.17, "learning_rate": 5.866611959773978e-06, "loss": 0.0095, "step": 28206 }, { "epoch": 5.17, "learning_rate": 5.865935931232427e-06, "loss": 0.0071, "step": 28207 }, { "epoch": 5.17, "learning_rate": 5.8652599254779356e-06, "loss": 0.0108, "step": 28208 }, { "epoch": 5.17, "learning_rate": 5.864583942514224e-06, "loss": 0.0104, "step": 28209 }, { "epoch": 5.17, "learning_rate": 5.863907982345027e-06, "loss": 0.0102, "step": 28210 }, { "epoch": 5.17, "learning_rate": 5.863232044974064e-06, "loss": 0.0113, "step": 28211 }, { "epoch": 5.17, "learning_rate": 5.862556130405064e-06, "loss": 0.0076, "step": 28212 }, { "epoch": 5.17, "learning_rate": 5.8618802386417505e-06, "loss": 0.0094, "step": 28213 }, { "epoch": 5.17, "learning_rate": 5.861204369687849e-06, "loss": 0.0076, "step": 28214 }, { "epoch": 5.17, "learning_rate": 5.860528523547088e-06, "loss": 0.0073, "step": 28215 }, { "epoch": 5.17, "learning_rate": 5.85985270022319e-06, "loss": 0.0094, "step": 28216 }, { "epoch": 5.17, "learning_rate": 5.859176899719883e-06, "loss": 0.0097, "step": 28217 }, { "epoch": 5.17, "learning_rate": 5.858501122040885e-06, "loss": 0.0125, "step": 28218 }, { "epoch": 5.17, "learning_rate": 5.857825367189931e-06, "loss": 0.0067, "step": 28219 }, { "epoch": 5.18, "learning_rate": 5.857149635170738e-06, "loss": 0.0103, "step": 28220 }, { "epoch": 5.18, "learning_rate": 5.856473925987034e-06, "loss": 0.0093, "step": 28221 }, { "epoch": 5.18, "learning_rate": 5.855798239642543e-06, "loss": 0.0109, "step": 28222 }, { "epoch": 5.18, "learning_rate": 5.855122576140987e-06, "loss": 0.0096, "step": 28223 }, { "epoch": 5.18, "learning_rate": 5.8544469354860935e-06, "loss": 0.0093, "step": 28224 }, { "epoch": 5.18, "learning_rate": 5.853771317681587e-06, "loss": 0.0067, "step": 28225 }, { "epoch": 5.18, "learning_rate": 5.853095722731192e-06, "loss": 0.0099, "step": 28226 }, { "epoch": 5.18, "learning_rate": 5.8524201506386236e-06, "loss": 0.0067, "step": 28227 }, { "epoch": 5.18, "learning_rate": 5.851744601407618e-06, "loss": 0.0115, "step": 28228 }, { "epoch": 5.18, "learning_rate": 5.851069075041893e-06, "loss": 0.0094, "step": 28229 }, { "epoch": 5.18, "learning_rate": 5.850393571545172e-06, "loss": 0.0119, "step": 28230 }, { "epoch": 5.18, "learning_rate": 5.849718090921177e-06, "loss": 0.0087, "step": 28231 }, { "epoch": 5.18, "learning_rate": 5.849042633173635e-06, "loss": 0.0096, "step": 28232 }, { "epoch": 5.18, "learning_rate": 5.848367198306266e-06, "loss": 0.0078, "step": 28233 }, { "epoch": 5.18, "learning_rate": 5.847691786322795e-06, "loss": 0.0104, "step": 28234 }, { "epoch": 5.18, "learning_rate": 5.847016397226947e-06, "loss": 0.0112, "step": 28235 }, { "epoch": 5.18, "learning_rate": 5.846341031022436e-06, "loss": 0.0106, "step": 28236 }, { "epoch": 5.18, "learning_rate": 5.845665687712995e-06, "loss": 0.0108, "step": 28237 }, { "epoch": 5.18, "learning_rate": 5.844990367302342e-06, "loss": 0.011, "step": 28238 }, { "epoch": 5.18, "learning_rate": 5.844315069794199e-06, "loss": 0.0082, "step": 28239 }, { "epoch": 5.18, "learning_rate": 5.8436397951922865e-06, "loss": 0.0124, "step": 28240 }, { "epoch": 5.18, "learning_rate": 5.842964543500329e-06, "loss": 0.0112, "step": 28241 }, { "epoch": 5.18, "learning_rate": 5.842289314722051e-06, "loss": 0.009, "step": 28242 }, { "epoch": 5.18, "learning_rate": 5.84161410886117e-06, "loss": 0.0095, "step": 28243 }, { "epoch": 5.18, "learning_rate": 5.840938925921411e-06, "loss": 0.0111, "step": 28244 }, { "epoch": 5.18, "learning_rate": 5.84026376590649e-06, "loss": 0.0104, "step": 28245 }, { "epoch": 5.18, "learning_rate": 5.839588628820138e-06, "loss": 0.0109, "step": 28246 }, { "epoch": 5.18, "learning_rate": 5.838913514666068e-06, "loss": 0.0081, "step": 28247 }, { "epoch": 5.18, "learning_rate": 5.838238423448006e-06, "loss": 0.0091, "step": 28248 }, { "epoch": 5.18, "learning_rate": 5.837563355169667e-06, "loss": 0.0097, "step": 28249 }, { "epoch": 5.18, "learning_rate": 5.836888309834779e-06, "loss": 0.012, "step": 28250 }, { "epoch": 5.18, "learning_rate": 5.836213287447061e-06, "loss": 0.0098, "step": 28251 }, { "epoch": 5.18, "learning_rate": 5.8355382880102316e-06, "loss": 0.0097, "step": 28252 }, { "epoch": 5.18, "learning_rate": 5.834863311528014e-06, "loss": 0.0111, "step": 28253 }, { "epoch": 5.18, "learning_rate": 5.834188358004127e-06, "loss": 0.007, "step": 28254 }, { "epoch": 5.18, "learning_rate": 5.833513427442291e-06, "loss": 0.0081, "step": 28255 }, { "epoch": 5.18, "learning_rate": 5.8328385198462265e-06, "loss": 0.0126, "step": 28256 }, { "epoch": 5.18, "learning_rate": 5.8321636352196544e-06, "loss": 0.012, "step": 28257 }, { "epoch": 5.18, "learning_rate": 5.83148877356629e-06, "loss": 0.0112, "step": 28258 }, { "epoch": 5.18, "learning_rate": 5.830813934889865e-06, "loss": 0.0109, "step": 28259 }, { "epoch": 5.18, "learning_rate": 5.830139119194087e-06, "loss": 0.0099, "step": 28260 }, { "epoch": 5.18, "learning_rate": 5.8294643264826786e-06, "loss": 0.0118, "step": 28261 }, { "epoch": 5.18, "learning_rate": 5.828789556759359e-06, "loss": 0.0091, "step": 28262 }, { "epoch": 5.18, "learning_rate": 5.828114810027852e-06, "loss": 0.009, "step": 28263 }, { "epoch": 5.18, "learning_rate": 5.827440086291873e-06, "loss": 0.0087, "step": 28264 }, { "epoch": 5.18, "learning_rate": 5.8267653855551445e-06, "loss": 0.0093, "step": 28265 }, { "epoch": 5.18, "learning_rate": 5.826090707821383e-06, "loss": 0.0089, "step": 28266 }, { "epoch": 5.18, "learning_rate": 5.8254160530943026e-06, "loss": 0.0086, "step": 28267 }, { "epoch": 5.18, "learning_rate": 5.82474142137763e-06, "loss": 0.0073, "step": 28268 }, { "epoch": 5.18, "learning_rate": 5.8240668126750795e-06, "loss": 0.0089, "step": 28269 }, { "epoch": 5.18, "learning_rate": 5.823392226990372e-06, "loss": 0.0079, "step": 28270 }, { "epoch": 5.18, "learning_rate": 5.822717664327224e-06, "loss": 0.0089, "step": 28271 }, { "epoch": 5.18, "learning_rate": 5.822043124689354e-06, "loss": 0.0104, "step": 28272 }, { "epoch": 5.18, "learning_rate": 5.8213686080804795e-06, "loss": 0.0088, "step": 28273 }, { "epoch": 5.19, "learning_rate": 5.82069411450432e-06, "loss": 0.0075, "step": 28274 }, { "epoch": 5.19, "learning_rate": 5.820019643964591e-06, "loss": 0.0076, "step": 28275 }, { "epoch": 5.19, "learning_rate": 5.81934519646501e-06, "loss": 0.0089, "step": 28276 }, { "epoch": 5.19, "learning_rate": 5.818670772009303e-06, "loss": 0.0111, "step": 28277 }, { "epoch": 5.19, "learning_rate": 5.817996370601176e-06, "loss": 0.0078, "step": 28278 }, { "epoch": 5.19, "learning_rate": 5.8173219922443516e-06, "loss": 0.0074, "step": 28279 }, { "epoch": 5.19, "learning_rate": 5.816647636942543e-06, "loss": 0.0104, "step": 28280 }, { "epoch": 5.19, "learning_rate": 5.8159733046994735e-06, "loss": 0.0089, "step": 28281 }, { "epoch": 5.19, "learning_rate": 5.815298995518857e-06, "loss": 0.0103, "step": 28282 }, { "epoch": 5.19, "learning_rate": 5.814624709404412e-06, "loss": 0.0076, "step": 28283 }, { "epoch": 5.19, "learning_rate": 5.813950446359853e-06, "loss": 0.01, "step": 28284 }, { "epoch": 5.19, "learning_rate": 5.813276206388896e-06, "loss": 0.0108, "step": 28285 }, { "epoch": 5.19, "learning_rate": 5.812601989495259e-06, "loss": 0.0078, "step": 28286 }, { "epoch": 5.19, "learning_rate": 5.811927795682659e-06, "loss": 0.0105, "step": 28287 }, { "epoch": 5.19, "learning_rate": 5.811253624954811e-06, "loss": 0.0123, "step": 28288 }, { "epoch": 5.19, "learning_rate": 5.81057947731543e-06, "loss": 0.0075, "step": 28289 }, { "epoch": 5.19, "learning_rate": 5.809905352768235e-06, "loss": 0.0107, "step": 28290 }, { "epoch": 5.19, "learning_rate": 5.809231251316938e-06, "loss": 0.0088, "step": 28291 }, { "epoch": 5.19, "learning_rate": 5.808557172965258e-06, "loss": 0.0089, "step": 28292 }, { "epoch": 5.19, "learning_rate": 5.807883117716904e-06, "loss": 0.0097, "step": 28293 }, { "epoch": 5.19, "learning_rate": 5.8072090855756e-06, "loss": 0.0091, "step": 28294 }, { "epoch": 5.19, "learning_rate": 5.806535076545063e-06, "loss": 0.0096, "step": 28295 }, { "epoch": 5.19, "learning_rate": 5.805861090628999e-06, "loss": 0.0081, "step": 28296 }, { "epoch": 5.19, "learning_rate": 5.805187127831127e-06, "loss": 0.0096, "step": 28297 }, { "epoch": 5.19, "learning_rate": 5.804513188155156e-06, "loss": 0.0093, "step": 28298 }, { "epoch": 5.19, "learning_rate": 5.803839271604812e-06, "loss": 0.0072, "step": 28299 }, { "epoch": 5.19, "learning_rate": 5.803165378183804e-06, "loss": 0.01, "step": 28300 }, { "epoch": 5.19, "learning_rate": 5.802491507895846e-06, "loss": 0.0115, "step": 28301 }, { "epoch": 5.19, "learning_rate": 5.801817660744656e-06, "loss": 0.0087, "step": 28302 }, { "epoch": 5.19, "learning_rate": 5.801143836733943e-06, "loss": 0.0104, "step": 28303 }, { "epoch": 5.19, "learning_rate": 5.800470035867422e-06, "loss": 0.0091, "step": 28304 }, { "epoch": 5.19, "learning_rate": 5.799796258148811e-06, "loss": 0.0109, "step": 28305 }, { "epoch": 5.19, "learning_rate": 5.799122503581821e-06, "loss": 0.0108, "step": 28306 }, { "epoch": 5.19, "learning_rate": 5.798448772170167e-06, "loss": 0.0106, "step": 28307 }, { "epoch": 5.19, "learning_rate": 5.797775063917559e-06, "loss": 0.0113, "step": 28308 }, { "epoch": 5.19, "learning_rate": 5.797101378827716e-06, "loss": 0.0078, "step": 28309 }, { "epoch": 5.19, "learning_rate": 5.796427716904347e-06, "loss": 0.0081, "step": 28310 }, { "epoch": 5.19, "learning_rate": 5.795754078151163e-06, "loss": 0.0097, "step": 28311 }, { "epoch": 5.19, "learning_rate": 5.795080462571887e-06, "loss": 0.0089, "step": 28312 }, { "epoch": 5.19, "learning_rate": 5.794406870170228e-06, "loss": 0.0096, "step": 28313 }, { "epoch": 5.19, "learning_rate": 5.793733300949894e-06, "loss": 0.0105, "step": 28314 }, { "epoch": 5.19, "learning_rate": 5.7930597549145965e-06, "loss": 0.0074, "step": 28315 }, { "epoch": 5.19, "learning_rate": 5.7923862320680555e-06, "loss": 0.0077, "step": 28316 }, { "epoch": 5.19, "learning_rate": 5.791712732413981e-06, "loss": 0.0081, "step": 28317 }, { "epoch": 5.19, "learning_rate": 5.791039255956083e-06, "loss": 0.0109, "step": 28318 }, { "epoch": 5.19, "learning_rate": 5.790365802698081e-06, "loss": 0.0115, "step": 28319 }, { "epoch": 5.19, "learning_rate": 5.789692372643672e-06, "loss": 0.0102, "step": 28320 }, { "epoch": 5.19, "learning_rate": 5.789018965796582e-06, "loss": 0.0105, "step": 28321 }, { "epoch": 5.19, "learning_rate": 5.7883455821605175e-06, "loss": 0.0081, "step": 28322 }, { "epoch": 5.19, "learning_rate": 5.787672221739191e-06, "loss": 0.0086, "step": 28323 }, { "epoch": 5.19, "learning_rate": 5.786998884536315e-06, "loss": 0.008, "step": 28324 }, { "epoch": 5.19, "learning_rate": 5.786325570555599e-06, "loss": 0.0073, "step": 28325 }, { "epoch": 5.19, "learning_rate": 5.785652279800755e-06, "loss": 0.0122, "step": 28326 }, { "epoch": 5.19, "learning_rate": 5.784979012275494e-06, "loss": 0.0106, "step": 28327 }, { "epoch": 5.19, "learning_rate": 5.784305767983529e-06, "loss": 0.0115, "step": 28328 }, { "epoch": 5.2, "learning_rate": 5.783632546928563e-06, "loss": 0.0075, "step": 28329 }, { "epoch": 5.2, "learning_rate": 5.782959349114319e-06, "loss": 0.0128, "step": 28330 }, { "epoch": 5.2, "learning_rate": 5.7822861745445055e-06, "loss": 0.0092, "step": 28331 }, { "epoch": 5.2, "learning_rate": 5.781613023222825e-06, "loss": 0.0102, "step": 28332 }, { "epoch": 5.2, "learning_rate": 5.780939895152989e-06, "loss": 0.011, "step": 28333 }, { "epoch": 5.2, "learning_rate": 5.780266790338715e-06, "loss": 0.0102, "step": 28334 }, { "epoch": 5.2, "learning_rate": 5.779593708783708e-06, "loss": 0.0101, "step": 28335 }, { "epoch": 5.2, "learning_rate": 5.7789206504916815e-06, "loss": 0.0088, "step": 28336 }, { "epoch": 5.2, "learning_rate": 5.778247615466341e-06, "loss": 0.0124, "step": 28337 }, { "epoch": 5.2, "learning_rate": 5.777574603711401e-06, "loss": 0.01, "step": 28338 }, { "epoch": 5.2, "learning_rate": 5.776901615230568e-06, "loss": 0.0093, "step": 28339 }, { "epoch": 5.2, "learning_rate": 5.776228650027551e-06, "loss": 0.0105, "step": 28340 }, { "epoch": 5.2, "learning_rate": 5.775555708106062e-06, "loss": 0.0119, "step": 28341 }, { "epoch": 5.2, "learning_rate": 5.774882789469808e-06, "loss": 0.0085, "step": 28342 }, { "epoch": 5.2, "learning_rate": 5.774209894122499e-06, "loss": 0.0122, "step": 28343 }, { "epoch": 5.2, "learning_rate": 5.7735370220678445e-06, "loss": 0.0092, "step": 28344 }, { "epoch": 5.2, "learning_rate": 5.772864173309552e-06, "loss": 0.0087, "step": 28345 }, { "epoch": 5.2, "learning_rate": 5.772191347851328e-06, "loss": 0.0081, "step": 28346 }, { "epoch": 5.2, "learning_rate": 5.771518545696888e-06, "loss": 0.0081, "step": 28347 }, { "epoch": 5.2, "learning_rate": 5.7708457668499375e-06, "loss": 0.0086, "step": 28348 }, { "epoch": 5.2, "learning_rate": 5.770173011314188e-06, "loss": 0.009, "step": 28349 }, { "epoch": 5.2, "learning_rate": 5.769500279093341e-06, "loss": 0.0076, "step": 28350 }, { "epoch": 5.2, "learning_rate": 5.768827570191102e-06, "loss": 0.0106, "step": 28351 }, { "epoch": 5.2, "learning_rate": 5.768154884611188e-06, "loss": 0.0077, "step": 28352 }, { "epoch": 5.2, "learning_rate": 5.767482222357306e-06, "loss": 0.0071, "step": 28353 }, { "epoch": 5.2, "learning_rate": 5.766809583433158e-06, "loss": 0.0092, "step": 28354 }, { "epoch": 5.2, "learning_rate": 5.766136967842457e-06, "loss": 0.0067, "step": 28355 }, { "epoch": 5.2, "learning_rate": 5.7654643755889074e-06, "loss": 0.0095, "step": 28356 }, { "epoch": 5.2, "learning_rate": 5.764791806676217e-06, "loss": 0.0073, "step": 28357 }, { "epoch": 5.2, "learning_rate": 5.764119261108094e-06, "loss": 0.0092, "step": 28358 }, { "epoch": 5.2, "learning_rate": 5.763446738888241e-06, "loss": 0.0084, "step": 28359 }, { "epoch": 5.2, "learning_rate": 5.762774240020377e-06, "loss": 0.0096, "step": 28360 }, { "epoch": 5.2, "learning_rate": 5.7621017645081955e-06, "loss": 0.0118, "step": 28361 }, { "epoch": 5.2, "learning_rate": 5.761429312355409e-06, "loss": 0.0093, "step": 28362 }, { "epoch": 5.2, "learning_rate": 5.7607568835657235e-06, "loss": 0.0089, "step": 28363 }, { "epoch": 5.2, "learning_rate": 5.760084478142842e-06, "loss": 0.0095, "step": 28364 }, { "epoch": 5.2, "learning_rate": 5.759412096090478e-06, "loss": 0.0074, "step": 28365 }, { "epoch": 5.2, "learning_rate": 5.7587397374123366e-06, "loss": 0.0104, "step": 28366 }, { "epoch": 5.2, "learning_rate": 5.758067402112118e-06, "loss": 0.0086, "step": 28367 }, { "epoch": 5.2, "learning_rate": 5.7573950901935274e-06, "loss": 0.0069, "step": 28368 }, { "epoch": 5.2, "learning_rate": 5.756722801660277e-06, "loss": 0.0095, "step": 28369 }, { "epoch": 5.2, "learning_rate": 5.756050536516073e-06, "loss": 0.0098, "step": 28370 }, { "epoch": 5.2, "learning_rate": 5.755378294764615e-06, "loss": 0.0103, "step": 28371 }, { "epoch": 5.2, "learning_rate": 5.754706076409613e-06, "loss": 0.0103, "step": 28372 }, { "epoch": 5.2, "learning_rate": 5.75403388145477e-06, "loss": 0.0072, "step": 28373 }, { "epoch": 5.2, "learning_rate": 5.753361709903791e-06, "loss": 0.0104, "step": 28374 }, { "epoch": 5.2, "learning_rate": 5.752689561760383e-06, "loss": 0.0087, "step": 28375 }, { "epoch": 5.2, "learning_rate": 5.7520174370282475e-06, "loss": 0.0078, "step": 28376 }, { "epoch": 5.2, "learning_rate": 5.751345335711089e-06, "loss": 0.0101, "step": 28377 }, { "epoch": 5.2, "learning_rate": 5.750673257812624e-06, "loss": 0.0116, "step": 28378 }, { "epoch": 5.2, "learning_rate": 5.750001203336542e-06, "loss": 0.0103, "step": 28379 }, { "epoch": 5.2, "learning_rate": 5.749329172286553e-06, "loss": 0.0122, "step": 28380 }, { "epoch": 5.2, "learning_rate": 5.7486571646663555e-06, "loss": 0.0082, "step": 28381 }, { "epoch": 5.2, "learning_rate": 5.747985180479666e-06, "loss": 0.0099, "step": 28382 }, { "epoch": 5.21, "learning_rate": 5.74731321973018e-06, "loss": 0.0109, "step": 28383 }, { "epoch": 5.21, "learning_rate": 5.746641282421608e-06, "loss": 0.0088, "step": 28384 }, { "epoch": 5.21, "learning_rate": 5.7459693685576445e-06, "loss": 0.0094, "step": 28385 }, { "epoch": 5.21, "learning_rate": 5.745297478141994e-06, "loss": 0.0091, "step": 28386 }, { "epoch": 5.21, "learning_rate": 5.744625611178367e-06, "loss": 0.0086, "step": 28387 }, { "epoch": 5.21, "learning_rate": 5.743953767670465e-06, "loss": 0.0089, "step": 28388 }, { "epoch": 5.21, "learning_rate": 5.743281947621988e-06, "loss": 0.0103, "step": 28389 }, { "epoch": 5.21, "learning_rate": 5.742610151036641e-06, "loss": 0.0109, "step": 28390 }, { "epoch": 5.21, "learning_rate": 5.741938377918126e-06, "loss": 0.0085, "step": 28391 }, { "epoch": 5.21, "learning_rate": 5.741266628270146e-06, "loss": 0.0104, "step": 28392 }, { "epoch": 5.21, "learning_rate": 5.740594902096406e-06, "loss": 0.0082, "step": 28393 }, { "epoch": 5.21, "learning_rate": 5.739923199400606e-06, "loss": 0.0088, "step": 28394 }, { "epoch": 5.21, "learning_rate": 5.739251520186444e-06, "loss": 0.0095, "step": 28395 }, { "epoch": 5.21, "learning_rate": 5.738579864457636e-06, "loss": 0.0118, "step": 28396 }, { "epoch": 5.21, "learning_rate": 5.737908232217871e-06, "loss": 0.0108, "step": 28397 }, { "epoch": 5.21, "learning_rate": 5.737236623470857e-06, "loss": 0.0074, "step": 28398 }, { "epoch": 5.21, "learning_rate": 5.736565038220289e-06, "loss": 0.0107, "step": 28399 }, { "epoch": 5.21, "learning_rate": 5.7358934764698806e-06, "loss": 0.01, "step": 28400 }, { "epoch": 5.21, "learning_rate": 5.735221938223326e-06, "loss": 0.0113, "step": 28401 }, { "epoch": 5.21, "learning_rate": 5.734550423484332e-06, "loss": 0.009, "step": 28402 }, { "epoch": 5.21, "learning_rate": 5.733878932256591e-06, "loss": 0.0072, "step": 28403 }, { "epoch": 5.21, "learning_rate": 5.733207464543807e-06, "loss": 0.0093, "step": 28404 }, { "epoch": 5.21, "learning_rate": 5.732536020349687e-06, "loss": 0.0086, "step": 28405 }, { "epoch": 5.21, "learning_rate": 5.731864599677928e-06, "loss": 0.0134, "step": 28406 }, { "epoch": 5.21, "learning_rate": 5.731193202532232e-06, "loss": 0.0072, "step": 28407 }, { "epoch": 5.21, "learning_rate": 5.730521828916298e-06, "loss": 0.0067, "step": 28408 }, { "epoch": 5.21, "learning_rate": 5.7298504788338295e-06, "loss": 0.0096, "step": 28409 }, { "epoch": 5.21, "learning_rate": 5.729179152288523e-06, "loss": 0.0098, "step": 28410 }, { "epoch": 5.21, "learning_rate": 5.728507849284082e-06, "loss": 0.0102, "step": 28411 }, { "epoch": 5.21, "learning_rate": 5.7278365698242025e-06, "loss": 0.0095, "step": 28412 }, { "epoch": 5.21, "learning_rate": 5.727165313912595e-06, "loss": 0.0099, "step": 28413 }, { "epoch": 5.21, "learning_rate": 5.726494081552949e-06, "loss": 0.0109, "step": 28414 }, { "epoch": 5.21, "learning_rate": 5.725822872748967e-06, "loss": 0.0088, "step": 28415 }, { "epoch": 5.21, "learning_rate": 5.72515168750435e-06, "loss": 0.0078, "step": 28416 }, { "epoch": 5.21, "learning_rate": 5.724480525822793e-06, "loss": 0.0092, "step": 28417 }, { "epoch": 5.21, "learning_rate": 5.723809387708004e-06, "loss": 0.0087, "step": 28418 }, { "epoch": 5.21, "learning_rate": 5.723138273163676e-06, "loss": 0.0104, "step": 28419 }, { "epoch": 5.21, "learning_rate": 5.722467182193516e-06, "loss": 0.0088, "step": 28420 }, { "epoch": 5.21, "learning_rate": 5.721796114801208e-06, "loss": 0.0079, "step": 28421 }, { "epoch": 5.21, "learning_rate": 5.721125070990465e-06, "loss": 0.0087, "step": 28422 }, { "epoch": 5.21, "learning_rate": 5.72045405076498e-06, "loss": 0.0074, "step": 28423 }, { "epoch": 5.21, "learning_rate": 5.719783054128453e-06, "loss": 0.011, "step": 28424 }, { "epoch": 5.21, "learning_rate": 5.7191120810845814e-06, "loss": 0.0103, "step": 28425 }, { "epoch": 5.21, "learning_rate": 5.718441131637065e-06, "loss": 0.0094, "step": 28426 }, { "epoch": 5.21, "learning_rate": 5.7177702057896015e-06, "loss": 0.0077, "step": 28427 }, { "epoch": 5.21, "learning_rate": 5.717099303545888e-06, "loss": 0.0092, "step": 28428 }, { "epoch": 5.21, "learning_rate": 5.716428424909623e-06, "loss": 0.0104, "step": 28429 }, { "epoch": 5.21, "learning_rate": 5.715757569884503e-06, "loss": 0.01, "step": 28430 }, { "epoch": 5.21, "learning_rate": 5.715086738474235e-06, "loss": 0.0069, "step": 28431 }, { "epoch": 5.21, "learning_rate": 5.714415930682504e-06, "loss": 0.0123, "step": 28432 }, { "epoch": 5.21, "learning_rate": 5.713745146513014e-06, "loss": 0.0089, "step": 28433 }, { "epoch": 5.21, "learning_rate": 5.713074385969457e-06, "loss": 0.0094, "step": 28434 }, { "epoch": 5.21, "learning_rate": 5.712403649055538e-06, "loss": 0.0103, "step": 28435 }, { "epoch": 5.21, "learning_rate": 5.711732935774952e-06, "loss": 0.0093, "step": 28436 }, { "epoch": 5.21, "learning_rate": 5.711062246131393e-06, "loss": 0.0088, "step": 28437 }, { "epoch": 5.22, "learning_rate": 5.7103915801285635e-06, "loss": 0.0093, "step": 28438 }, { "epoch": 5.22, "learning_rate": 5.709720937770149e-06, "loss": 0.008, "step": 28439 }, { "epoch": 5.22, "learning_rate": 5.7090503190598556e-06, "loss": 0.0082, "step": 28440 }, { "epoch": 5.22, "learning_rate": 5.708379724001378e-06, "loss": 0.0091, "step": 28441 }, { "epoch": 5.22, "learning_rate": 5.707709152598413e-06, "loss": 0.0076, "step": 28442 }, { "epoch": 5.22, "learning_rate": 5.707038604854654e-06, "loss": 0.0098, "step": 28443 }, { "epoch": 5.22, "learning_rate": 5.7063680807738005e-06, "loss": 0.0091, "step": 28444 }, { "epoch": 5.22, "learning_rate": 5.705697580359546e-06, "loss": 0.0074, "step": 28445 }, { "epoch": 5.22, "learning_rate": 5.7050271036155865e-06, "loss": 0.0108, "step": 28446 }, { "epoch": 5.22, "learning_rate": 5.70435665054562e-06, "loss": 0.007, "step": 28447 }, { "epoch": 5.22, "learning_rate": 5.703686221153334e-06, "loss": 0.0115, "step": 28448 }, { "epoch": 5.22, "learning_rate": 5.703015815442439e-06, "loss": 0.01, "step": 28449 }, { "epoch": 5.22, "learning_rate": 5.702345433416619e-06, "loss": 0.0098, "step": 28450 }, { "epoch": 5.22, "learning_rate": 5.701675075079569e-06, "loss": 0.0092, "step": 28451 }, { "epoch": 5.22, "learning_rate": 5.7010047404349836e-06, "loss": 0.0115, "step": 28452 }, { "epoch": 5.22, "learning_rate": 5.700334429486565e-06, "loss": 0.0066, "step": 28453 }, { "epoch": 5.22, "learning_rate": 5.699664142238004e-06, "loss": 0.0098, "step": 28454 }, { "epoch": 5.22, "learning_rate": 5.698993878692996e-06, "loss": 0.0078, "step": 28455 }, { "epoch": 5.22, "learning_rate": 5.698323638855236e-06, "loss": 0.0088, "step": 28456 }, { "epoch": 5.22, "learning_rate": 5.6976534227284106e-06, "loss": 0.0079, "step": 28457 }, { "epoch": 5.22, "learning_rate": 5.696983230316224e-06, "loss": 0.0099, "step": 28458 }, { "epoch": 5.22, "learning_rate": 5.6963130616223675e-06, "loss": 0.0092, "step": 28459 }, { "epoch": 5.22, "learning_rate": 5.6956429166505336e-06, "loss": 0.0088, "step": 28460 }, { "epoch": 5.22, "learning_rate": 5.694972795404416e-06, "loss": 0.0096, "step": 28461 }, { "epoch": 5.22, "learning_rate": 5.694302697887711e-06, "loss": 0.0091, "step": 28462 }, { "epoch": 5.22, "learning_rate": 5.693632624104109e-06, "loss": 0.0109, "step": 28463 }, { "epoch": 5.22, "learning_rate": 5.692962574057304e-06, "loss": 0.0105, "step": 28464 }, { "epoch": 5.22, "learning_rate": 5.692292547750989e-06, "loss": 0.0097, "step": 28465 }, { "epoch": 5.22, "learning_rate": 5.691622545188859e-06, "loss": 0.0087, "step": 28466 }, { "epoch": 5.22, "learning_rate": 5.6909525663746125e-06, "loss": 0.0113, "step": 28467 }, { "epoch": 5.22, "learning_rate": 5.690282611311933e-06, "loss": 0.0116, "step": 28468 }, { "epoch": 5.22, "learning_rate": 5.689612680004515e-06, "loss": 0.0117, "step": 28469 }, { "epoch": 5.22, "learning_rate": 5.68894277245605e-06, "loss": 0.0126, "step": 28470 }, { "epoch": 5.22, "learning_rate": 5.688272888670236e-06, "loss": 0.0107, "step": 28471 }, { "epoch": 5.22, "learning_rate": 5.687603028650764e-06, "loss": 0.0107, "step": 28472 }, { "epoch": 5.22, "learning_rate": 5.686933192401329e-06, "loss": 0.0105, "step": 28473 }, { "epoch": 5.22, "learning_rate": 5.686263379925612e-06, "loss": 0.0072, "step": 28474 }, { "epoch": 5.22, "learning_rate": 5.685593591227315e-06, "loss": 0.0074, "step": 28475 }, { "epoch": 5.22, "learning_rate": 5.684923826310128e-06, "loss": 0.0103, "step": 28476 }, { "epoch": 5.22, "learning_rate": 5.6842540851777415e-06, "loss": 0.0111, "step": 28477 }, { "epoch": 5.22, "learning_rate": 5.683584367833848e-06, "loss": 0.0126, "step": 28478 }, { "epoch": 5.22, "learning_rate": 5.682914674282138e-06, "loss": 0.0083, "step": 28479 }, { "epoch": 5.22, "learning_rate": 5.682245004526305e-06, "loss": 0.0119, "step": 28480 }, { "epoch": 5.22, "learning_rate": 5.681575358570038e-06, "loss": 0.0123, "step": 28481 }, { "epoch": 5.22, "learning_rate": 5.680905736417028e-06, "loss": 0.0079, "step": 28482 }, { "epoch": 5.22, "learning_rate": 5.680236138070963e-06, "loss": 0.0099, "step": 28483 }, { "epoch": 5.22, "learning_rate": 5.679566563535541e-06, "loss": 0.0094, "step": 28484 }, { "epoch": 5.22, "learning_rate": 5.678897012814455e-06, "loss": 0.0101, "step": 28485 }, { "epoch": 5.22, "learning_rate": 5.678227485911385e-06, "loss": 0.0075, "step": 28486 }, { "epoch": 5.22, "learning_rate": 5.677557982830022e-06, "loss": 0.0091, "step": 28487 }, { "epoch": 5.22, "learning_rate": 5.676888503574065e-06, "loss": 0.0103, "step": 28488 }, { "epoch": 5.22, "learning_rate": 5.6762190481472e-06, "loss": 0.0122, "step": 28489 }, { "epoch": 5.22, "learning_rate": 5.675549616553117e-06, "loss": 0.0081, "step": 28490 }, { "epoch": 5.22, "learning_rate": 5.67488020879551e-06, "loss": 0.0092, "step": 28491 }, { "epoch": 5.23, "learning_rate": 5.674210824878056e-06, "loss": 0.0125, "step": 28492 }, { "epoch": 5.23, "learning_rate": 5.673541464804457e-06, "loss": 0.0062, "step": 28493 }, { "epoch": 5.23, "learning_rate": 5.672872128578399e-06, "loss": 0.0067, "step": 28494 }, { "epoch": 5.23, "learning_rate": 5.672202816203571e-06, "loss": 0.012, "step": 28495 }, { "epoch": 5.23, "learning_rate": 5.671533527683664e-06, "loss": 0.0109, "step": 28496 }, { "epoch": 5.23, "learning_rate": 5.670864263022364e-06, "loss": 0.0071, "step": 28497 }, { "epoch": 5.23, "learning_rate": 5.670195022223363e-06, "loss": 0.0081, "step": 28498 }, { "epoch": 5.23, "learning_rate": 5.669525805290348e-06, "loss": 0.0122, "step": 28499 }, { "epoch": 5.23, "learning_rate": 5.668856612227007e-06, "loss": 0.0078, "step": 28500 }, { "epoch": 5.23, "learning_rate": 5.668187443037028e-06, "loss": 0.0099, "step": 28501 }, { "epoch": 5.23, "learning_rate": 5.667518297724103e-06, "loss": 0.011, "step": 28502 }, { "epoch": 5.23, "learning_rate": 5.666849176291924e-06, "loss": 0.0084, "step": 28503 }, { "epoch": 5.23, "learning_rate": 5.666180078744169e-06, "loss": 0.008, "step": 28504 }, { "epoch": 5.23, "learning_rate": 5.665511005084527e-06, "loss": 0.0101, "step": 28505 }, { "epoch": 5.23, "learning_rate": 5.664841955316696e-06, "loss": 0.0087, "step": 28506 }, { "epoch": 5.23, "learning_rate": 5.664172929444355e-06, "loss": 0.0089, "step": 28507 }, { "epoch": 5.23, "learning_rate": 5.663503927471195e-06, "loss": 0.0104, "step": 28508 }, { "epoch": 5.23, "learning_rate": 5.662834949400904e-06, "loss": 0.0089, "step": 28509 }, { "epoch": 5.23, "learning_rate": 5.662165995237168e-06, "loss": 0.0137, "step": 28510 }, { "epoch": 5.23, "learning_rate": 5.661497064983674e-06, "loss": 0.0091, "step": 28511 }, { "epoch": 5.23, "learning_rate": 5.660828158644109e-06, "loss": 0.0109, "step": 28512 }, { "epoch": 5.23, "learning_rate": 5.660159276222164e-06, "loss": 0.0056, "step": 28513 }, { "epoch": 5.23, "learning_rate": 5.65949041772152e-06, "loss": 0.0129, "step": 28514 }, { "epoch": 5.23, "learning_rate": 5.658821583145867e-06, "loss": 0.0075, "step": 28515 }, { "epoch": 5.23, "learning_rate": 5.65815277249889e-06, "loss": 0.0098, "step": 28516 }, { "epoch": 5.23, "learning_rate": 5.65748398578428e-06, "loss": 0.0078, "step": 28517 }, { "epoch": 5.23, "learning_rate": 5.656815223005714e-06, "loss": 0.0109, "step": 28518 }, { "epoch": 5.23, "learning_rate": 5.656146484166888e-06, "loss": 0.0088, "step": 28519 }, { "epoch": 5.23, "learning_rate": 5.655477769271488e-06, "loss": 0.0094, "step": 28520 }, { "epoch": 5.23, "learning_rate": 5.654809078323194e-06, "loss": 0.0077, "step": 28521 }, { "epoch": 5.23, "learning_rate": 5.654140411325693e-06, "loss": 0.0072, "step": 28522 }, { "epoch": 5.23, "learning_rate": 5.653471768282668e-06, "loss": 0.0095, "step": 28523 }, { "epoch": 5.23, "learning_rate": 5.652803149197812e-06, "loss": 0.0086, "step": 28524 }, { "epoch": 5.23, "learning_rate": 5.652134554074806e-06, "loss": 0.0083, "step": 28525 }, { "epoch": 5.23, "learning_rate": 5.651465982917339e-06, "loss": 0.0107, "step": 28526 }, { "epoch": 5.23, "learning_rate": 5.6507974357290915e-06, "loss": 0.0102, "step": 28527 }, { "epoch": 5.23, "learning_rate": 5.65012891251375e-06, "loss": 0.01, "step": 28528 }, { "epoch": 5.23, "learning_rate": 5.649460413275e-06, "loss": 0.0082, "step": 28529 }, { "epoch": 5.23, "learning_rate": 5.6487919380165275e-06, "loss": 0.0071, "step": 28530 }, { "epoch": 5.23, "learning_rate": 5.648123486742015e-06, "loss": 0.0078, "step": 28531 }, { "epoch": 5.23, "learning_rate": 5.647455059455149e-06, "loss": 0.0093, "step": 28532 }, { "epoch": 5.23, "learning_rate": 5.64678665615961e-06, "loss": 0.0102, "step": 28533 }, { "epoch": 5.23, "learning_rate": 5.646118276859087e-06, "loss": 0.0076, "step": 28534 }, { "epoch": 5.23, "learning_rate": 5.6454499215572625e-06, "loss": 0.0107, "step": 28535 }, { "epoch": 5.23, "learning_rate": 5.644781590257815e-06, "loss": 0.01, "step": 28536 }, { "epoch": 5.23, "learning_rate": 5.644113282964436e-06, "loss": 0.0098, "step": 28537 }, { "epoch": 5.23, "learning_rate": 5.643444999680814e-06, "loss": 0.0086, "step": 28538 }, { "epoch": 5.23, "learning_rate": 5.642776740410618e-06, "loss": 0.0086, "step": 28539 }, { "epoch": 5.23, "learning_rate": 5.642108505157537e-06, "loss": 0.0082, "step": 28540 }, { "epoch": 5.23, "learning_rate": 5.6414402939252575e-06, "loss": 0.0089, "step": 28541 }, { "epoch": 5.23, "learning_rate": 5.640772106717464e-06, "loss": 0.01, "step": 28542 }, { "epoch": 5.23, "learning_rate": 5.6401039435378354e-06, "loss": 0.0102, "step": 28543 }, { "epoch": 5.23, "learning_rate": 5.639435804390057e-06, "loss": 0.0092, "step": 28544 }, { "epoch": 5.23, "learning_rate": 5.638767689277808e-06, "loss": 0.0067, "step": 28545 }, { "epoch": 5.23, "learning_rate": 5.638099598204776e-06, "loss": 0.0102, "step": 28546 }, { "epoch": 5.24, "learning_rate": 5.637431531174641e-06, "loss": 0.0081, "step": 28547 }, { "epoch": 5.24, "learning_rate": 5.636763488191085e-06, "loss": 0.011, "step": 28548 }, { "epoch": 5.24, "learning_rate": 5.636095469257787e-06, "loss": 0.0095, "step": 28549 }, { "epoch": 5.24, "learning_rate": 5.6354274743784406e-06, "loss": 0.0099, "step": 28550 }, { "epoch": 5.24, "learning_rate": 5.6347595035567175e-06, "loss": 0.0133, "step": 28551 }, { "epoch": 5.24, "learning_rate": 5.634091556796302e-06, "loss": 0.0098, "step": 28552 }, { "epoch": 5.24, "learning_rate": 5.633423634100874e-06, "loss": 0.0091, "step": 28553 }, { "epoch": 5.24, "learning_rate": 5.632755735474115e-06, "loss": 0.0089, "step": 28554 }, { "epoch": 5.24, "learning_rate": 5.632087860919713e-06, "loss": 0.0092, "step": 28555 }, { "epoch": 5.24, "learning_rate": 5.6314200104413486e-06, "loss": 0.0082, "step": 28556 }, { "epoch": 5.24, "learning_rate": 5.630752184042696e-06, "loss": 0.009, "step": 28557 }, { "epoch": 5.24, "learning_rate": 5.630084381727434e-06, "loss": 0.009, "step": 28558 }, { "epoch": 5.24, "learning_rate": 5.629416603499255e-06, "loss": 0.0126, "step": 28559 }, { "epoch": 5.24, "learning_rate": 5.6287488493618335e-06, "loss": 0.0082, "step": 28560 }, { "epoch": 5.24, "learning_rate": 5.628081119318851e-06, "loss": 0.0076, "step": 28561 }, { "epoch": 5.24, "learning_rate": 5.627413413373988e-06, "loss": 0.0076, "step": 28562 }, { "epoch": 5.24, "learning_rate": 5.626745731530923e-06, "loss": 0.0085, "step": 28563 }, { "epoch": 5.24, "learning_rate": 5.626078073793339e-06, "loss": 0.0083, "step": 28564 }, { "epoch": 5.24, "learning_rate": 5.625410440164916e-06, "loss": 0.0094, "step": 28565 }, { "epoch": 5.24, "learning_rate": 5.624742830649332e-06, "loss": 0.0116, "step": 28566 }, { "epoch": 5.24, "learning_rate": 5.6240752452502685e-06, "loss": 0.0115, "step": 28567 }, { "epoch": 5.24, "learning_rate": 5.623407683971405e-06, "loss": 0.0113, "step": 28568 }, { "epoch": 5.24, "learning_rate": 5.622740146816421e-06, "loss": 0.0094, "step": 28569 }, { "epoch": 5.24, "learning_rate": 5.622072633788995e-06, "loss": 0.0077, "step": 28570 }, { "epoch": 5.24, "learning_rate": 5.621405144892804e-06, "loss": 0.0146, "step": 28571 }, { "epoch": 5.24, "learning_rate": 5.620737680131534e-06, "loss": 0.0107, "step": 28572 }, { "epoch": 5.24, "learning_rate": 5.620070239508862e-06, "loss": 0.008, "step": 28573 }, { "epoch": 5.24, "learning_rate": 5.619402823028466e-06, "loss": 0.0093, "step": 28574 }, { "epoch": 5.24, "learning_rate": 5.6187354306940225e-06, "loss": 0.0105, "step": 28575 }, { "epoch": 5.24, "learning_rate": 5.6180680625092074e-06, "loss": 0.0083, "step": 28576 }, { "epoch": 5.24, "learning_rate": 5.617400718477709e-06, "loss": 0.009, "step": 28577 }, { "epoch": 5.24, "learning_rate": 5.6167333986032e-06, "loss": 0.0083, "step": 28578 }, { "epoch": 5.24, "learning_rate": 5.6160661028893596e-06, "loss": 0.0118, "step": 28579 }, { "epoch": 5.24, "learning_rate": 5.6153988313398645e-06, "loss": 0.0097, "step": 28580 }, { "epoch": 5.24, "learning_rate": 5.614731583958393e-06, "loss": 0.0065, "step": 28581 }, { "epoch": 5.24, "learning_rate": 5.614064360748626e-06, "loss": 0.0076, "step": 28582 }, { "epoch": 5.24, "learning_rate": 5.6133971617142365e-06, "loss": 0.0096, "step": 28583 }, { "epoch": 5.24, "learning_rate": 5.612729986858907e-06, "loss": 0.0096, "step": 28584 }, { "epoch": 5.24, "learning_rate": 5.612062836186312e-06, "loss": 0.0106, "step": 28585 }, { "epoch": 5.24, "learning_rate": 5.611395709700128e-06, "loss": 0.0072, "step": 28586 }, { "epoch": 5.24, "learning_rate": 5.610728607404034e-06, "loss": 0.0071, "step": 28587 }, { "epoch": 5.24, "learning_rate": 5.610061529301708e-06, "loss": 0.0087, "step": 28588 }, { "epoch": 5.24, "learning_rate": 5.609394475396821e-06, "loss": 0.0114, "step": 28589 }, { "epoch": 5.24, "learning_rate": 5.608727445693058e-06, "loss": 0.0093, "step": 28590 }, { "epoch": 5.24, "learning_rate": 5.6080604401940944e-06, "loss": 0.0117, "step": 28591 }, { "epoch": 5.24, "learning_rate": 5.607393458903606e-06, "loss": 0.0123, "step": 28592 }, { "epoch": 5.24, "learning_rate": 5.606726501825261e-06, "loss": 0.0076, "step": 28593 }, { "epoch": 5.24, "learning_rate": 5.606059568962746e-06, "loss": 0.0096, "step": 28594 }, { "epoch": 5.24, "learning_rate": 5.6053926603197335e-06, "loss": 0.0098, "step": 28595 }, { "epoch": 5.24, "learning_rate": 5.6047257758998996e-06, "loss": 0.0104, "step": 28596 }, { "epoch": 5.24, "learning_rate": 5.6040589157069206e-06, "loss": 0.0097, "step": 28597 }, { "epoch": 5.24, "learning_rate": 5.603392079744472e-06, "loss": 0.0092, "step": 28598 }, { "epoch": 5.24, "learning_rate": 5.6027252680162295e-06, "loss": 0.009, "step": 28599 }, { "epoch": 5.24, "learning_rate": 5.602058480525867e-06, "loss": 0.0089, "step": 28600 }, { "epoch": 5.25, "learning_rate": 5.6013917172770626e-06, "loss": 0.0098, "step": 28601 }, { "epoch": 5.25, "learning_rate": 5.600724978273485e-06, "loss": 0.0074, "step": 28602 }, { "epoch": 5.25, "learning_rate": 5.600058263518824e-06, "loss": 0.0089, "step": 28603 }, { "epoch": 5.25, "learning_rate": 5.599391573016739e-06, "loss": 0.0124, "step": 28604 }, { "epoch": 5.25, "learning_rate": 5.598724906770912e-06, "loss": 0.0117, "step": 28605 }, { "epoch": 5.25, "learning_rate": 5.598058264785016e-06, "loss": 0.0074, "step": 28606 }, { "epoch": 5.25, "learning_rate": 5.597391647062722e-06, "loss": 0.0101, "step": 28607 }, { "epoch": 5.25, "learning_rate": 5.5967250536077125e-06, "loss": 0.0103, "step": 28608 }, { "epoch": 5.25, "learning_rate": 5.5960584844236565e-06, "loss": 0.0095, "step": 28609 }, { "epoch": 5.25, "learning_rate": 5.595391939514234e-06, "loss": 0.0107, "step": 28610 }, { "epoch": 5.25, "learning_rate": 5.594725418883107e-06, "loss": 0.0115, "step": 28611 }, { "epoch": 5.25, "learning_rate": 5.5940589225339605e-06, "loss": 0.0083, "step": 28612 }, { "epoch": 5.25, "learning_rate": 5.593392450470464e-06, "loss": 0.0084, "step": 28613 }, { "epoch": 5.25, "learning_rate": 5.5927260026962915e-06, "loss": 0.0096, "step": 28614 }, { "epoch": 5.25, "learning_rate": 5.592059579215117e-06, "loss": 0.014, "step": 28615 }, { "epoch": 5.25, "learning_rate": 5.591393180030613e-06, "loss": 0.0099, "step": 28616 }, { "epoch": 5.25, "learning_rate": 5.590726805146454e-06, "loss": 0.0089, "step": 28617 }, { "epoch": 5.25, "learning_rate": 5.590060454566311e-06, "loss": 0.009, "step": 28618 }, { "epoch": 5.25, "learning_rate": 5.589394128293859e-06, "loss": 0.0096, "step": 28619 }, { "epoch": 5.25, "learning_rate": 5.588727826332765e-06, "loss": 0.008, "step": 28620 }, { "epoch": 5.25, "learning_rate": 5.588061548686715e-06, "loss": 0.0098, "step": 28621 }, { "epoch": 5.25, "learning_rate": 5.5873952953593694e-06, "loss": 0.0095, "step": 28622 }, { "epoch": 5.25, "learning_rate": 5.586729066354404e-06, "loss": 0.0069, "step": 28623 }, { "epoch": 5.25, "learning_rate": 5.586062861675487e-06, "loss": 0.0064, "step": 28624 }, { "epoch": 5.25, "learning_rate": 5.585396681326299e-06, "loss": 0.0072, "step": 28625 }, { "epoch": 5.25, "learning_rate": 5.5847305253105085e-06, "loss": 0.0116, "step": 28626 }, { "epoch": 5.25, "learning_rate": 5.5840643936317895e-06, "loss": 0.0078, "step": 28627 }, { "epoch": 5.25, "learning_rate": 5.583398286293808e-06, "loss": 0.0091, "step": 28628 }, { "epoch": 5.25, "learning_rate": 5.582732203300236e-06, "loss": 0.0091, "step": 28629 }, { "epoch": 5.25, "learning_rate": 5.58206614465475e-06, "loss": 0.0094, "step": 28630 }, { "epoch": 5.25, "learning_rate": 5.581400110361019e-06, "loss": 0.0085, "step": 28631 }, { "epoch": 5.25, "learning_rate": 5.580734100422714e-06, "loss": 0.0082, "step": 28632 }, { "epoch": 5.25, "learning_rate": 5.580068114843506e-06, "loss": 0.0093, "step": 28633 }, { "epoch": 5.25, "learning_rate": 5.579402153627067e-06, "loss": 0.0103, "step": 28634 }, { "epoch": 5.25, "learning_rate": 5.578736216777065e-06, "loss": 0.0085, "step": 28635 }, { "epoch": 5.25, "learning_rate": 5.5780703042971754e-06, "loss": 0.0093, "step": 28636 }, { "epoch": 5.25, "learning_rate": 5.57740441619106e-06, "loss": 0.0106, "step": 28637 }, { "epoch": 5.25, "learning_rate": 5.5767385524624e-06, "loss": 0.0088, "step": 28638 }, { "epoch": 5.25, "learning_rate": 5.576072713114864e-06, "loss": 0.0089, "step": 28639 }, { "epoch": 5.25, "learning_rate": 5.575406898152115e-06, "loss": 0.0066, "step": 28640 }, { "epoch": 5.25, "learning_rate": 5.574741107577826e-06, "loss": 0.0114, "step": 28641 }, { "epoch": 5.25, "learning_rate": 5.574075341395665e-06, "loss": 0.0093, "step": 28642 }, { "epoch": 5.25, "learning_rate": 5.573409599609308e-06, "loss": 0.0091, "step": 28643 }, { "epoch": 5.25, "learning_rate": 5.572743882222421e-06, "loss": 0.0106, "step": 28644 }, { "epoch": 5.25, "learning_rate": 5.572078189238677e-06, "loss": 0.0084, "step": 28645 }, { "epoch": 5.25, "learning_rate": 5.5714125206617346e-06, "loss": 0.0117, "step": 28646 }, { "epoch": 5.25, "learning_rate": 5.570746876495275e-06, "loss": 0.0098, "step": 28647 }, { "epoch": 5.25, "learning_rate": 5.57008125674296e-06, "loss": 0.0069, "step": 28648 }, { "epoch": 5.25, "learning_rate": 5.569415661408462e-06, "loss": 0.0096, "step": 28649 }, { "epoch": 5.25, "learning_rate": 5.568750090495448e-06, "loss": 0.0088, "step": 28650 }, { "epoch": 5.25, "learning_rate": 5.5680845440075885e-06, "loss": 0.0118, "step": 28651 }, { "epoch": 5.25, "learning_rate": 5.56741902194855e-06, "loss": 0.0116, "step": 28652 }, { "epoch": 5.25, "learning_rate": 5.5667535243220015e-06, "loss": 0.0097, "step": 28653 }, { "epoch": 5.25, "learning_rate": 5.566088051131613e-06, "loss": 0.0089, "step": 28654 }, { "epoch": 5.25, "learning_rate": 5.565422602381045e-06, "loss": 0.0079, "step": 28655 }, { "epoch": 5.26, "learning_rate": 5.564757178073976e-06, "loss": 0.0077, "step": 28656 }, { "epoch": 5.26, "learning_rate": 5.564091778214074e-06, "loss": 0.0092, "step": 28657 }, { "epoch": 5.26, "learning_rate": 5.563426402804999e-06, "loss": 0.0093, "step": 28658 }, { "epoch": 5.26, "learning_rate": 5.562761051850421e-06, "loss": 0.0093, "step": 28659 }, { "epoch": 5.26, "learning_rate": 5.562095725354002e-06, "loss": 0.0079, "step": 28660 }, { "epoch": 5.26, "learning_rate": 5.56143042331942e-06, "loss": 0.0072, "step": 28661 }, { "epoch": 5.26, "learning_rate": 5.56076514575034e-06, "loss": 0.0096, "step": 28662 }, { "epoch": 5.26, "learning_rate": 5.560099892650428e-06, "loss": 0.0082, "step": 28663 }, { "epoch": 5.26, "learning_rate": 5.559434664023342e-06, "loss": 0.0097, "step": 28664 }, { "epoch": 5.26, "learning_rate": 5.558769459872762e-06, "loss": 0.012, "step": 28665 }, { "epoch": 5.26, "learning_rate": 5.558104280202346e-06, "loss": 0.008, "step": 28666 }, { "epoch": 5.26, "learning_rate": 5.557439125015765e-06, "loss": 0.0072, "step": 28667 }, { "epoch": 5.26, "learning_rate": 5.556773994316682e-06, "loss": 0.0104, "step": 28668 }, { "epoch": 5.26, "learning_rate": 5.556108888108767e-06, "loss": 0.0085, "step": 28669 }, { "epoch": 5.26, "learning_rate": 5.555443806395681e-06, "loss": 0.0097, "step": 28670 }, { "epoch": 5.26, "learning_rate": 5.554778749181095e-06, "loss": 0.0103, "step": 28671 }, { "epoch": 5.26, "learning_rate": 5.554113716468673e-06, "loss": 0.0094, "step": 28672 }, { "epoch": 5.26, "learning_rate": 5.5534487082620755e-06, "loss": 0.0121, "step": 28673 }, { "epoch": 5.26, "learning_rate": 5.5527837245649806e-06, "loss": 0.0105, "step": 28674 }, { "epoch": 5.26, "learning_rate": 5.552118765381043e-06, "loss": 0.0098, "step": 28675 }, { "epoch": 5.26, "learning_rate": 5.551453830713931e-06, "loss": 0.0111, "step": 28676 }, { "epoch": 5.26, "learning_rate": 5.550788920567305e-06, "loss": 0.0084, "step": 28677 }, { "epoch": 5.26, "learning_rate": 5.550124034944839e-06, "loss": 0.011, "step": 28678 }, { "epoch": 5.26, "learning_rate": 5.549459173850193e-06, "loss": 0.0076, "step": 28679 }, { "epoch": 5.26, "learning_rate": 5.5487943372870334e-06, "loss": 0.0088, "step": 28680 }, { "epoch": 5.26, "learning_rate": 5.5481295252590275e-06, "loss": 0.0099, "step": 28681 }, { "epoch": 5.26, "learning_rate": 5.547464737769827e-06, "loss": 0.0105, "step": 28682 }, { "epoch": 5.26, "learning_rate": 5.5467999748231095e-06, "loss": 0.0092, "step": 28683 }, { "epoch": 5.26, "learning_rate": 5.546135236422534e-06, "loss": 0.0099, "step": 28684 }, { "epoch": 5.26, "learning_rate": 5.545470522571768e-06, "loss": 0.0086, "step": 28685 }, { "epoch": 5.26, "learning_rate": 5.54480583327447e-06, "loss": 0.0085, "step": 28686 }, { "epoch": 5.26, "learning_rate": 5.544141168534309e-06, "loss": 0.0067, "step": 28687 }, { "epoch": 5.26, "learning_rate": 5.543476528354945e-06, "loss": 0.0079, "step": 28688 }, { "epoch": 5.26, "learning_rate": 5.542811912740042e-06, "loss": 0.0085, "step": 28689 }, { "epoch": 5.26, "learning_rate": 5.542147321693261e-06, "loss": 0.0098, "step": 28690 }, { "epoch": 5.26, "learning_rate": 5.541482755218272e-06, "loss": 0.0082, "step": 28691 }, { "epoch": 5.26, "learning_rate": 5.54081821331874e-06, "loss": 0.0097, "step": 28692 }, { "epoch": 5.26, "learning_rate": 5.540153695998318e-06, "loss": 0.008, "step": 28693 }, { "epoch": 5.26, "learning_rate": 5.539489203260673e-06, "loss": 0.0097, "step": 28694 }, { "epoch": 5.26, "learning_rate": 5.538824735109463e-06, "loss": 0.01, "step": 28695 }, { "epoch": 5.26, "learning_rate": 5.5381602915483605e-06, "loss": 0.0112, "step": 28696 }, { "epoch": 5.26, "learning_rate": 5.5374958725810224e-06, "loss": 0.0085, "step": 28697 }, { "epoch": 5.26, "learning_rate": 5.536831478211113e-06, "loss": 0.0109, "step": 28698 }, { "epoch": 5.26, "learning_rate": 5.536167108442292e-06, "loss": 0.0105, "step": 28699 }, { "epoch": 5.26, "learning_rate": 5.535502763278222e-06, "loss": 0.0092, "step": 28700 }, { "epoch": 5.26, "learning_rate": 5.534838442722567e-06, "loss": 0.009, "step": 28701 }, { "epoch": 5.26, "learning_rate": 5.534174146778985e-06, "loss": 0.0092, "step": 28702 }, { "epoch": 5.26, "learning_rate": 5.533509875451142e-06, "loss": 0.0079, "step": 28703 }, { "epoch": 5.26, "learning_rate": 5.532845628742696e-06, "loss": 0.009, "step": 28704 }, { "epoch": 5.26, "learning_rate": 5.532181406657308e-06, "loss": 0.008, "step": 28705 }, { "epoch": 5.26, "learning_rate": 5.531517209198642e-06, "loss": 0.0065, "step": 28706 }, { "epoch": 5.26, "learning_rate": 5.530853036370359e-06, "loss": 0.0092, "step": 28707 }, { "epoch": 5.26, "learning_rate": 5.530188888176114e-06, "loss": 0.011, "step": 28708 }, { "epoch": 5.26, "learning_rate": 5.529524764619576e-06, "loss": 0.0081, "step": 28709 }, { "epoch": 5.26, "learning_rate": 5.5288606657044075e-06, "loss": 0.0095, "step": 28710 }, { "epoch": 5.27, "learning_rate": 5.5281965914342584e-06, "loss": 0.007, "step": 28711 }, { "epoch": 5.27, "learning_rate": 5.527532541812795e-06, "loss": 0.0096, "step": 28712 }, { "epoch": 5.27, "learning_rate": 5.526868516843673e-06, "loss": 0.0094, "step": 28713 }, { "epoch": 5.27, "learning_rate": 5.5262045165305615e-06, "loss": 0.0078, "step": 28714 }, { "epoch": 5.27, "learning_rate": 5.5255405408771144e-06, "loss": 0.0085, "step": 28715 }, { "epoch": 5.27, "learning_rate": 5.5248765898869925e-06, "loss": 0.009, "step": 28716 }, { "epoch": 5.27, "learning_rate": 5.524212663563856e-06, "loss": 0.0096, "step": 28717 }, { "epoch": 5.27, "learning_rate": 5.523548761911363e-06, "loss": 0.0083, "step": 28718 }, { "epoch": 5.27, "learning_rate": 5.522884884933177e-06, "loss": 0.0079, "step": 28719 }, { "epoch": 5.27, "learning_rate": 5.5222210326329525e-06, "loss": 0.009, "step": 28720 }, { "epoch": 5.27, "learning_rate": 5.521557205014351e-06, "loss": 0.0091, "step": 28721 }, { "epoch": 5.27, "learning_rate": 5.52089340208103e-06, "loss": 0.0081, "step": 28722 }, { "epoch": 5.27, "learning_rate": 5.5202296238366505e-06, "loss": 0.0114, "step": 28723 }, { "epoch": 5.27, "learning_rate": 5.519565870284871e-06, "loss": 0.0089, "step": 28724 }, { "epoch": 5.27, "learning_rate": 5.518902141429347e-06, "loss": 0.0095, "step": 28725 }, { "epoch": 5.27, "learning_rate": 5.518238437273738e-06, "loss": 0.0093, "step": 28726 }, { "epoch": 5.27, "learning_rate": 5.517574757821707e-06, "loss": 0.0095, "step": 28727 }, { "epoch": 5.27, "learning_rate": 5.516911103076912e-06, "loss": 0.0081, "step": 28728 }, { "epoch": 5.27, "learning_rate": 5.516247473043003e-06, "loss": 0.0093, "step": 28729 }, { "epoch": 5.27, "learning_rate": 5.515583867723642e-06, "loss": 0.0112, "step": 28730 }, { "epoch": 5.27, "learning_rate": 5.514920287122489e-06, "loss": 0.0117, "step": 28731 }, { "epoch": 5.27, "learning_rate": 5.514256731243202e-06, "loss": 0.0096, "step": 28732 }, { "epoch": 5.27, "learning_rate": 5.513593200089437e-06, "loss": 0.0104, "step": 28733 }, { "epoch": 5.27, "learning_rate": 5.5129296936648525e-06, "loss": 0.0087, "step": 28734 }, { "epoch": 5.27, "learning_rate": 5.512266211973103e-06, "loss": 0.0096, "step": 28735 }, { "epoch": 5.27, "learning_rate": 5.511602755017848e-06, "loss": 0.0115, "step": 28736 }, { "epoch": 5.27, "learning_rate": 5.510939322802742e-06, "loss": 0.0096, "step": 28737 }, { "epoch": 5.27, "learning_rate": 5.510275915331445e-06, "loss": 0.0067, "step": 28738 }, { "epoch": 5.27, "learning_rate": 5.509612532607613e-06, "loss": 0.0095, "step": 28739 }, { "epoch": 5.27, "learning_rate": 5.508949174634901e-06, "loss": 0.0071, "step": 28740 }, { "epoch": 5.27, "learning_rate": 5.508285841416967e-06, "loss": 0.0096, "step": 28741 }, { "epoch": 5.27, "learning_rate": 5.507622532957466e-06, "loss": 0.0081, "step": 28742 }, { "epoch": 5.27, "learning_rate": 5.506959249260051e-06, "loss": 0.0081, "step": 28743 }, { "epoch": 5.27, "learning_rate": 5.5062959903283855e-06, "loss": 0.009, "step": 28744 }, { "epoch": 5.27, "learning_rate": 5.505632756166123e-06, "loss": 0.0139, "step": 28745 }, { "epoch": 5.27, "learning_rate": 5.5049695467769216e-06, "loss": 0.0079, "step": 28746 }, { "epoch": 5.27, "learning_rate": 5.5043063621644285e-06, "loss": 0.008, "step": 28747 }, { "epoch": 5.27, "learning_rate": 5.5036432023323e-06, "loss": 0.0103, "step": 28748 }, { "epoch": 5.27, "learning_rate": 5.5029800672842e-06, "loss": 0.0114, "step": 28749 }, { "epoch": 5.27, "learning_rate": 5.50231695702378e-06, "loss": 0.0097, "step": 28750 }, { "epoch": 5.27, "learning_rate": 5.501653871554693e-06, "loss": 0.0115, "step": 28751 }, { "epoch": 5.27, "learning_rate": 5.5009908108805965e-06, "loss": 0.0096, "step": 28752 }, { "epoch": 5.27, "learning_rate": 5.500327775005143e-06, "loss": 0.0081, "step": 28753 }, { "epoch": 5.27, "learning_rate": 5.499664763931989e-06, "loss": 0.0075, "step": 28754 }, { "epoch": 5.27, "learning_rate": 5.499001777664787e-06, "loss": 0.0076, "step": 28755 }, { "epoch": 5.27, "learning_rate": 5.498338816207195e-06, "loss": 0.0077, "step": 28756 }, { "epoch": 5.27, "learning_rate": 5.497675879562863e-06, "loss": 0.012, "step": 28757 }, { "epoch": 5.27, "learning_rate": 5.4970129677354476e-06, "loss": 0.0109, "step": 28758 }, { "epoch": 5.27, "learning_rate": 5.496350080728603e-06, "loss": 0.0063, "step": 28759 }, { "epoch": 5.27, "learning_rate": 5.495687218545981e-06, "loss": 0.0099, "step": 28760 }, { "epoch": 5.27, "learning_rate": 5.495024381191232e-06, "loss": 0.0088, "step": 28761 }, { "epoch": 5.27, "learning_rate": 5.49436156866802e-06, "loss": 0.0129, "step": 28762 }, { "epoch": 5.27, "learning_rate": 5.4936987809799915e-06, "loss": 0.0083, "step": 28763 }, { "epoch": 5.27, "learning_rate": 5.493036018130805e-06, "loss": 0.0095, "step": 28764 }, { "epoch": 5.28, "learning_rate": 5.492373280124101e-06, "loss": 0.0102, "step": 28765 }, { "epoch": 5.28, "learning_rate": 5.4917105669635464e-06, "loss": 0.0088, "step": 28766 }, { "epoch": 5.28, "learning_rate": 5.491047878652789e-06, "loss": 0.0094, "step": 28767 }, { "epoch": 5.28, "learning_rate": 5.490385215195479e-06, "loss": 0.0109, "step": 28768 }, { "epoch": 5.28, "learning_rate": 5.4897225765952735e-06, "loss": 0.009, "step": 28769 }, { "epoch": 5.28, "learning_rate": 5.489059962855822e-06, "loss": 0.0101, "step": 28770 }, { "epoch": 5.28, "learning_rate": 5.488397373980778e-06, "loss": 0.0068, "step": 28771 }, { "epoch": 5.28, "learning_rate": 5.487734809973793e-06, "loss": 0.009, "step": 28772 }, { "epoch": 5.28, "learning_rate": 5.487072270838519e-06, "loss": 0.0098, "step": 28773 }, { "epoch": 5.28, "learning_rate": 5.486409756578606e-06, "loss": 0.0115, "step": 28774 }, { "epoch": 5.28, "learning_rate": 5.485747267197714e-06, "loss": 0.0084, "step": 28775 }, { "epoch": 5.28, "learning_rate": 5.4850848026994865e-06, "loss": 0.0105, "step": 28776 }, { "epoch": 5.28, "learning_rate": 5.484422363087576e-06, "loss": 0.0102, "step": 28777 }, { "epoch": 5.28, "learning_rate": 5.483759948365637e-06, "loss": 0.0086, "step": 28778 }, { "epoch": 5.28, "learning_rate": 5.4830975585373135e-06, "loss": 0.0098, "step": 28779 }, { "epoch": 5.28, "learning_rate": 5.482435193606266e-06, "loss": 0.0091, "step": 28780 }, { "epoch": 5.28, "learning_rate": 5.481772853576143e-06, "loss": 0.0107, "step": 28781 }, { "epoch": 5.28, "learning_rate": 5.481110538450596e-06, "loss": 0.0093, "step": 28782 }, { "epoch": 5.28, "learning_rate": 5.4804482482332676e-06, "loss": 0.0073, "step": 28783 }, { "epoch": 5.28, "learning_rate": 5.479785982927817e-06, "loss": 0.013, "step": 28784 }, { "epoch": 5.28, "learning_rate": 5.479123742537893e-06, "loss": 0.0109, "step": 28785 }, { "epoch": 5.28, "learning_rate": 5.478461527067143e-06, "loss": 0.0118, "step": 28786 }, { "epoch": 5.28, "learning_rate": 5.477799336519221e-06, "loss": 0.0052, "step": 28787 }, { "epoch": 5.28, "learning_rate": 5.477137170897773e-06, "loss": 0.0107, "step": 28788 }, { "epoch": 5.28, "learning_rate": 5.476475030206452e-06, "loss": 0.0091, "step": 28789 }, { "epoch": 5.28, "learning_rate": 5.475812914448907e-06, "loss": 0.0097, "step": 28790 }, { "epoch": 5.28, "learning_rate": 5.475150823628786e-06, "loss": 0.0085, "step": 28791 }, { "epoch": 5.28, "learning_rate": 5.4744887577497364e-06, "loss": 0.0085, "step": 28792 }, { "epoch": 5.28, "learning_rate": 5.47382671681542e-06, "loss": 0.0092, "step": 28793 }, { "epoch": 5.28, "learning_rate": 5.473164700829471e-06, "loss": 0.0081, "step": 28794 }, { "epoch": 5.28, "learning_rate": 5.4725027097955445e-06, "loss": 0.0096, "step": 28795 }, { "epoch": 5.28, "learning_rate": 5.471840743717286e-06, "loss": 0.0116, "step": 28796 }, { "epoch": 5.28, "learning_rate": 5.471178802598351e-06, "loss": 0.01, "step": 28797 }, { "epoch": 5.28, "learning_rate": 5.470516886442384e-06, "loss": 0.0091, "step": 28798 }, { "epoch": 5.28, "learning_rate": 5.469854995253039e-06, "loss": 0.0056, "step": 28799 }, { "epoch": 5.28, "learning_rate": 5.469193129033954e-06, "loss": 0.009, "step": 28800 }, { "epoch": 5.28, "learning_rate": 5.46853128778878e-06, "loss": 0.0091, "step": 28801 }, { "epoch": 5.28, "learning_rate": 5.4678694715211714e-06, "loss": 0.009, "step": 28802 }, { "epoch": 5.28, "learning_rate": 5.467207680234772e-06, "loss": 0.007, "step": 28803 }, { "epoch": 5.28, "learning_rate": 5.46654591393323e-06, "loss": 0.0108, "step": 28804 }, { "epoch": 5.28, "learning_rate": 5.465884172620193e-06, "loss": 0.0086, "step": 28805 }, { "epoch": 5.28, "learning_rate": 5.465222456299309e-06, "loss": 0.0079, "step": 28806 }, { "epoch": 5.28, "learning_rate": 5.464560764974225e-06, "loss": 0.0086, "step": 28807 }, { "epoch": 5.28, "learning_rate": 5.463899098648589e-06, "loss": 0.0095, "step": 28808 }, { "epoch": 5.28, "learning_rate": 5.463237457326047e-06, "loss": 0.0101, "step": 28809 }, { "epoch": 5.28, "learning_rate": 5.462575841010242e-06, "loss": 0.0126, "step": 28810 }, { "epoch": 5.28, "learning_rate": 5.4619142497048336e-06, "loss": 0.0103, "step": 28811 }, { "epoch": 5.28, "learning_rate": 5.461252683413456e-06, "loss": 0.0093, "step": 28812 }, { "epoch": 5.28, "learning_rate": 5.46059114213976e-06, "loss": 0.0106, "step": 28813 }, { "epoch": 5.28, "learning_rate": 5.459929625887388e-06, "loss": 0.0089, "step": 28814 }, { "epoch": 5.28, "learning_rate": 5.459268134659994e-06, "loss": 0.0071, "step": 28815 }, { "epoch": 5.28, "learning_rate": 5.458606668461222e-06, "loss": 0.0101, "step": 28816 }, { "epoch": 5.28, "learning_rate": 5.457945227294719e-06, "loss": 0.0112, "step": 28817 }, { "epoch": 5.28, "learning_rate": 5.45728381116412e-06, "loss": 0.0083, "step": 28818 }, { "epoch": 5.28, "learning_rate": 5.456622420073084e-06, "loss": 0.0096, "step": 28819 }, { "epoch": 5.29, "learning_rate": 5.45596105402525e-06, "loss": 0.0066, "step": 28820 }, { "epoch": 5.29, "learning_rate": 5.455299713024267e-06, "loss": 0.0111, "step": 28821 }, { "epoch": 5.29, "learning_rate": 5.454638397073777e-06, "loss": 0.0087, "step": 28822 }, { "epoch": 5.29, "learning_rate": 5.453977106177428e-06, "loss": 0.0091, "step": 28823 }, { "epoch": 5.29, "learning_rate": 5.453315840338863e-06, "loss": 0.0087, "step": 28824 }, { "epoch": 5.29, "learning_rate": 5.452654599561727e-06, "loss": 0.0114, "step": 28825 }, { "epoch": 5.29, "learning_rate": 5.451993383849667e-06, "loss": 0.0084, "step": 28826 }, { "epoch": 5.29, "learning_rate": 5.451332193206321e-06, "loss": 0.0084, "step": 28827 }, { "epoch": 5.29, "learning_rate": 5.450671027635346e-06, "loss": 0.0094, "step": 28828 }, { "epoch": 5.29, "learning_rate": 5.4500098871403765e-06, "loss": 0.01, "step": 28829 }, { "epoch": 5.29, "learning_rate": 5.449348771725058e-06, "loss": 0.0119, "step": 28830 }, { "epoch": 5.29, "learning_rate": 5.448687681393036e-06, "loss": 0.0108, "step": 28831 }, { "epoch": 5.29, "learning_rate": 5.44802661614795e-06, "loss": 0.0086, "step": 28832 }, { "epoch": 5.29, "learning_rate": 5.447365575993451e-06, "loss": 0.0106, "step": 28833 }, { "epoch": 5.29, "learning_rate": 5.446704560933181e-06, "loss": 0.0088, "step": 28834 }, { "epoch": 5.29, "learning_rate": 5.446043570970785e-06, "loss": 0.0101, "step": 28835 }, { "epoch": 5.29, "learning_rate": 5.445382606109897e-06, "loss": 0.0081, "step": 28836 }, { "epoch": 5.29, "learning_rate": 5.444721666354169e-06, "loss": 0.0092, "step": 28837 }, { "epoch": 5.29, "learning_rate": 5.4440607517072434e-06, "loss": 0.0123, "step": 28838 }, { "epoch": 5.29, "learning_rate": 5.44339986217276e-06, "loss": 0.0095, "step": 28839 }, { "epoch": 5.29, "learning_rate": 5.4427389977543645e-06, "loss": 0.0115, "step": 28840 }, { "epoch": 5.29, "learning_rate": 5.442078158455699e-06, "loss": 0.0078, "step": 28841 }, { "epoch": 5.29, "learning_rate": 5.441417344280404e-06, "loss": 0.0105, "step": 28842 }, { "epoch": 5.29, "learning_rate": 5.440756555232125e-06, "loss": 0.0092, "step": 28843 }, { "epoch": 5.29, "learning_rate": 5.440095791314501e-06, "loss": 0.0091, "step": 28844 }, { "epoch": 5.29, "learning_rate": 5.439435052531172e-06, "loss": 0.0096, "step": 28845 }, { "epoch": 5.29, "learning_rate": 5.4387743388857925e-06, "loss": 0.0097, "step": 28846 }, { "epoch": 5.29, "learning_rate": 5.438113650381993e-06, "loss": 0.0098, "step": 28847 }, { "epoch": 5.29, "learning_rate": 5.437452987023416e-06, "loss": 0.0077, "step": 28848 }, { "epoch": 5.29, "learning_rate": 5.436792348813702e-06, "loss": 0.0085, "step": 28849 }, { "epoch": 5.29, "learning_rate": 5.436131735756498e-06, "loss": 0.0068, "step": 28850 }, { "epoch": 5.29, "learning_rate": 5.435471147855444e-06, "loss": 0.0075, "step": 28851 }, { "epoch": 5.29, "learning_rate": 5.434810585114181e-06, "loss": 0.0092, "step": 28852 }, { "epoch": 5.29, "learning_rate": 5.434150047536351e-06, "loss": 0.009, "step": 28853 }, { "epoch": 5.29, "learning_rate": 5.433489535125587e-06, "loss": 0.0108, "step": 28854 }, { "epoch": 5.29, "learning_rate": 5.432829047885537e-06, "loss": 0.0102, "step": 28855 }, { "epoch": 5.29, "learning_rate": 5.432168585819842e-06, "loss": 0.0097, "step": 28856 }, { "epoch": 5.29, "learning_rate": 5.431508148932141e-06, "loss": 0.0109, "step": 28857 }, { "epoch": 5.29, "learning_rate": 5.430847737226073e-06, "loss": 0.0068, "step": 28858 }, { "epoch": 5.29, "learning_rate": 5.4301873507052805e-06, "loss": 0.0084, "step": 28859 }, { "epoch": 5.29, "learning_rate": 5.429526989373402e-06, "loss": 0.0099, "step": 28860 }, { "epoch": 5.29, "learning_rate": 5.428866653234077e-06, "loss": 0.0099, "step": 28861 }, { "epoch": 5.29, "learning_rate": 5.428206342290947e-06, "loss": 0.0119, "step": 28862 }, { "epoch": 5.29, "learning_rate": 5.427546056547645e-06, "loss": 0.0089, "step": 28863 }, { "epoch": 5.29, "learning_rate": 5.426885796007827e-06, "loss": 0.0123, "step": 28864 }, { "epoch": 5.29, "learning_rate": 5.426225560675116e-06, "loss": 0.0085, "step": 28865 }, { "epoch": 5.29, "learning_rate": 5.425565350553157e-06, "loss": 0.0074, "step": 28866 }, { "epoch": 5.29, "learning_rate": 5.424905165645587e-06, "loss": 0.0083, "step": 28867 }, { "epoch": 5.29, "learning_rate": 5.424245005956048e-06, "loss": 0.0091, "step": 28868 }, { "epoch": 5.29, "learning_rate": 5.423584871488179e-06, "loss": 0.0059, "step": 28869 }, { "epoch": 5.29, "learning_rate": 5.422924762245616e-06, "loss": 0.0091, "step": 28870 }, { "epoch": 5.29, "learning_rate": 5.422264678232001e-06, "loss": 0.0074, "step": 28871 }, { "epoch": 5.29, "learning_rate": 5.421604619450969e-06, "loss": 0.0106, "step": 28872 }, { "epoch": 5.29, "learning_rate": 5.42094458590616e-06, "loss": 0.0092, "step": 28873 }, { "epoch": 5.3, "learning_rate": 5.420284577601211e-06, "loss": 0.0096, "step": 28874 }, { "epoch": 5.3, "learning_rate": 5.419624594539763e-06, "loss": 0.01, "step": 28875 }, { "epoch": 5.3, "learning_rate": 5.41896463672545e-06, "loss": 0.0094, "step": 28876 }, { "epoch": 5.3, "learning_rate": 5.418304704161911e-06, "loss": 0.0078, "step": 28877 }, { "epoch": 5.3, "learning_rate": 5.417644796852783e-06, "loss": 0.0083, "step": 28878 }, { "epoch": 5.3, "learning_rate": 5.416984914801706e-06, "loss": 0.0105, "step": 28879 }, { "epoch": 5.3, "learning_rate": 5.416325058012311e-06, "loss": 0.0118, "step": 28880 }, { "epoch": 5.3, "learning_rate": 5.415665226488244e-06, "loss": 0.0095, "step": 28881 }, { "epoch": 5.3, "learning_rate": 5.415005420233141e-06, "loss": 0.0091, "step": 28882 }, { "epoch": 5.3, "learning_rate": 5.414345639250631e-06, "loss": 0.006, "step": 28883 }, { "epoch": 5.3, "learning_rate": 5.413685883544356e-06, "loss": 0.0116, "step": 28884 }, { "epoch": 5.3, "learning_rate": 5.4130261531179485e-06, "loss": 0.0082, "step": 28885 }, { "epoch": 5.3, "learning_rate": 5.412366447975052e-06, "loss": 0.0123, "step": 28886 }, { "epoch": 5.3, "learning_rate": 5.411706768119298e-06, "loss": 0.012, "step": 28887 }, { "epoch": 5.3, "learning_rate": 5.411047113554324e-06, "loss": 0.0108, "step": 28888 }, { "epoch": 5.3, "learning_rate": 5.410387484283767e-06, "loss": 0.0082, "step": 28889 }, { "epoch": 5.3, "learning_rate": 5.409727880311262e-06, "loss": 0.0108, "step": 28890 }, { "epoch": 5.3, "learning_rate": 5.409068301640443e-06, "loss": 0.0107, "step": 28891 }, { "epoch": 5.3, "learning_rate": 5.408408748274947e-06, "loss": 0.0089, "step": 28892 }, { "epoch": 5.3, "learning_rate": 5.40774922021841e-06, "loss": 0.0076, "step": 28893 }, { "epoch": 5.3, "learning_rate": 5.407089717474467e-06, "loss": 0.0096, "step": 28894 }, { "epoch": 5.3, "learning_rate": 5.406430240046754e-06, "loss": 0.0085, "step": 28895 }, { "epoch": 5.3, "learning_rate": 5.405770787938905e-06, "loss": 0.0094, "step": 28896 }, { "epoch": 5.3, "learning_rate": 5.405111361154553e-06, "loss": 0.0092, "step": 28897 }, { "epoch": 5.3, "learning_rate": 5.404451959697332e-06, "loss": 0.0086, "step": 28898 }, { "epoch": 5.3, "learning_rate": 5.403792583570884e-06, "loss": 0.0098, "step": 28899 }, { "epoch": 5.3, "learning_rate": 5.403133232778842e-06, "loss": 0.0087, "step": 28900 }, { "epoch": 5.3, "learning_rate": 5.4024739073248344e-06, "loss": 0.0099, "step": 28901 }, { "epoch": 5.3, "learning_rate": 5.401814607212493e-06, "loss": 0.0095, "step": 28902 }, { "epoch": 5.3, "learning_rate": 5.4011553324454625e-06, "loss": 0.0087, "step": 28903 }, { "epoch": 5.3, "learning_rate": 5.4004960830273715e-06, "loss": 0.0091, "step": 28904 }, { "epoch": 5.3, "learning_rate": 5.399836858961853e-06, "loss": 0.0105, "step": 28905 }, { "epoch": 5.3, "learning_rate": 5.399177660252546e-06, "loss": 0.011, "step": 28906 }, { "epoch": 5.3, "learning_rate": 5.3985184869030705e-06, "loss": 0.0101, "step": 28907 }, { "epoch": 5.3, "learning_rate": 5.397859338917074e-06, "loss": 0.0072, "step": 28908 }, { "epoch": 5.3, "learning_rate": 5.397200216298184e-06, "loss": 0.0095, "step": 28909 }, { "epoch": 5.3, "learning_rate": 5.3965411190500365e-06, "loss": 0.0094, "step": 28910 }, { "epoch": 5.3, "learning_rate": 5.39588204717626e-06, "loss": 0.01, "step": 28911 }, { "epoch": 5.3, "learning_rate": 5.395223000680489e-06, "loss": 0.0102, "step": 28912 }, { "epoch": 5.3, "learning_rate": 5.394563979566358e-06, "loss": 0.0111, "step": 28913 }, { "epoch": 5.3, "learning_rate": 5.393904983837498e-06, "loss": 0.0088, "step": 28914 }, { "epoch": 5.3, "learning_rate": 5.3932460134975415e-06, "loss": 0.0083, "step": 28915 }, { "epoch": 5.3, "learning_rate": 5.392587068550117e-06, "loss": 0.0073, "step": 28916 }, { "epoch": 5.3, "learning_rate": 5.391928148998864e-06, "loss": 0.0115, "step": 28917 }, { "epoch": 5.3, "learning_rate": 5.391269254847416e-06, "loss": 0.0076, "step": 28918 }, { "epoch": 5.3, "learning_rate": 5.390610386099395e-06, "loss": 0.0086, "step": 28919 }, { "epoch": 5.3, "learning_rate": 5.389951542758434e-06, "loss": 0.0119, "step": 28920 }, { "epoch": 5.3, "learning_rate": 5.389292724828171e-06, "loss": 0.0075, "step": 28921 }, { "epoch": 5.3, "learning_rate": 5.388633932312235e-06, "loss": 0.0091, "step": 28922 }, { "epoch": 5.3, "learning_rate": 5.387975165214259e-06, "loss": 0.0097, "step": 28923 }, { "epoch": 5.3, "learning_rate": 5.387316423537869e-06, "loss": 0.0092, "step": 28924 }, { "epoch": 5.3, "learning_rate": 5.3866577072867e-06, "loss": 0.0116, "step": 28925 }, { "epoch": 5.3, "learning_rate": 5.385999016464382e-06, "loss": 0.0082, "step": 28926 }, { "epoch": 5.3, "learning_rate": 5.385340351074545e-06, "loss": 0.0081, "step": 28927 }, { "epoch": 5.3, "learning_rate": 5.3846817111208205e-06, "loss": 0.0094, "step": 28928 }, { "epoch": 5.31, "learning_rate": 5.3840230966068385e-06, "loss": 0.0116, "step": 28929 }, { "epoch": 5.31, "learning_rate": 5.3833645075362295e-06, "loss": 0.0092, "step": 28930 }, { "epoch": 5.31, "learning_rate": 5.382705943912623e-06, "loss": 0.0078, "step": 28931 }, { "epoch": 5.31, "learning_rate": 5.38204740573965e-06, "loss": 0.0094, "step": 28932 }, { "epoch": 5.31, "learning_rate": 5.381388893020935e-06, "loss": 0.0059, "step": 28933 }, { "epoch": 5.31, "learning_rate": 5.380730405760118e-06, "loss": 0.0101, "step": 28934 }, { "epoch": 5.31, "learning_rate": 5.380071943960823e-06, "loss": 0.0098, "step": 28935 }, { "epoch": 5.31, "learning_rate": 5.379413507626683e-06, "loss": 0.0081, "step": 28936 }, { "epoch": 5.31, "learning_rate": 5.3787550967613215e-06, "loss": 0.0097, "step": 28937 }, { "epoch": 5.31, "learning_rate": 5.3780967113683635e-06, "loss": 0.0087, "step": 28938 }, { "epoch": 5.31, "learning_rate": 5.37743835145145e-06, "loss": 0.0103, "step": 28939 }, { "epoch": 5.31, "learning_rate": 5.3767800170142045e-06, "loss": 0.0063, "step": 28940 }, { "epoch": 5.31, "learning_rate": 5.376121708060257e-06, "loss": 0.0081, "step": 28941 }, { "epoch": 5.31, "learning_rate": 5.375463424593234e-06, "loss": 0.0087, "step": 28942 }, { "epoch": 5.31, "learning_rate": 5.374805166616765e-06, "loss": 0.0093, "step": 28943 }, { "epoch": 5.31, "learning_rate": 5.3741469341344785e-06, "loss": 0.0085, "step": 28944 }, { "epoch": 5.31, "learning_rate": 5.3734887271500025e-06, "loss": 0.0115, "step": 28945 }, { "epoch": 5.31, "learning_rate": 5.37283054566696e-06, "loss": 0.011, "step": 28946 }, { "epoch": 5.31, "learning_rate": 5.372172389688992e-06, "loss": 0.0068, "step": 28947 }, { "epoch": 5.31, "learning_rate": 5.371514259219716e-06, "loss": 0.0074, "step": 28948 }, { "epoch": 5.31, "learning_rate": 5.37085615426276e-06, "loss": 0.0074, "step": 28949 }, { "epoch": 5.31, "learning_rate": 5.370198074821754e-06, "loss": 0.0089, "step": 28950 }, { "epoch": 5.31, "learning_rate": 5.369540020900321e-06, "loss": 0.0113, "step": 28951 }, { "epoch": 5.31, "learning_rate": 5.368881992502095e-06, "loss": 0.0108, "step": 28952 }, { "epoch": 5.31, "learning_rate": 5.368223989630705e-06, "loss": 0.0087, "step": 28953 }, { "epoch": 5.31, "learning_rate": 5.367566012289769e-06, "loss": 0.0081, "step": 28954 }, { "epoch": 5.31, "learning_rate": 5.366908060482914e-06, "loss": 0.0081, "step": 28955 }, { "epoch": 5.31, "learning_rate": 5.366250134213773e-06, "loss": 0.0093, "step": 28956 }, { "epoch": 5.31, "learning_rate": 5.36559223348597e-06, "loss": 0.0085, "step": 28957 }, { "epoch": 5.31, "learning_rate": 5.364934358303131e-06, "loss": 0.009, "step": 28958 }, { "epoch": 5.31, "learning_rate": 5.364276508668883e-06, "loss": 0.0101, "step": 28959 }, { "epoch": 5.31, "learning_rate": 5.363618684586851e-06, "loss": 0.0087, "step": 28960 }, { "epoch": 5.31, "learning_rate": 5.362960886060662e-06, "loss": 0.0093, "step": 28961 }, { "epoch": 5.31, "learning_rate": 5.362303113093941e-06, "loss": 0.0101, "step": 28962 }, { "epoch": 5.31, "learning_rate": 5.361645365690313e-06, "loss": 0.0092, "step": 28963 }, { "epoch": 5.31, "learning_rate": 5.360987643853403e-06, "loss": 0.0131, "step": 28964 }, { "epoch": 5.31, "learning_rate": 5.360329947586843e-06, "loss": 0.0088, "step": 28965 }, { "epoch": 5.31, "learning_rate": 5.35967227689425e-06, "loss": 0.0107, "step": 28966 }, { "epoch": 5.31, "learning_rate": 5.359014631779253e-06, "loss": 0.0082, "step": 28967 }, { "epoch": 5.31, "learning_rate": 5.358357012245474e-06, "loss": 0.0079, "step": 28968 }, { "epoch": 5.31, "learning_rate": 5.357699418296538e-06, "loss": 0.0065, "step": 28969 }, { "epoch": 5.31, "learning_rate": 5.357041849936076e-06, "loss": 0.0097, "step": 28970 }, { "epoch": 5.31, "learning_rate": 5.35638430716771e-06, "loss": 0.0081, "step": 28971 }, { "epoch": 5.31, "learning_rate": 5.3557267899950595e-06, "loss": 0.0073, "step": 28972 }, { "epoch": 5.31, "learning_rate": 5.355069298421747e-06, "loss": 0.0096, "step": 28973 }, { "epoch": 5.31, "learning_rate": 5.354411832451406e-06, "loss": 0.0087, "step": 28974 }, { "epoch": 5.31, "learning_rate": 5.353754392087655e-06, "loss": 0.0096, "step": 28975 }, { "epoch": 5.31, "learning_rate": 5.35309697733412e-06, "loss": 0.0077, "step": 28976 }, { "epoch": 5.31, "learning_rate": 5.352439588194423e-06, "loss": 0.0101, "step": 28977 }, { "epoch": 5.31, "learning_rate": 5.351782224672188e-06, "loss": 0.0101, "step": 28978 }, { "epoch": 5.31, "learning_rate": 5.351124886771037e-06, "loss": 0.006, "step": 28979 }, { "epoch": 5.31, "learning_rate": 5.350467574494595e-06, "loss": 0.0097, "step": 28980 }, { "epoch": 5.31, "learning_rate": 5.349810287846486e-06, "loss": 0.0119, "step": 28981 }, { "epoch": 5.31, "learning_rate": 5.349153026830327e-06, "loss": 0.0077, "step": 28982 }, { "epoch": 5.32, "learning_rate": 5.348495791449753e-06, "loss": 0.01, "step": 28983 }, { "epoch": 5.32, "learning_rate": 5.347838581708375e-06, "loss": 0.0083, "step": 28984 }, { "epoch": 5.32, "learning_rate": 5.34718139760982e-06, "loss": 0.0104, "step": 28985 }, { "epoch": 5.32, "learning_rate": 5.3465242391577064e-06, "loss": 0.012, "step": 28986 }, { "epoch": 5.32, "learning_rate": 5.345867106355665e-06, "loss": 0.0066, "step": 28987 }, { "epoch": 5.32, "learning_rate": 5.34520999920731e-06, "loss": 0.0078, "step": 28988 }, { "epoch": 5.32, "learning_rate": 5.344552917716274e-06, "loss": 0.0111, "step": 28989 }, { "epoch": 5.32, "learning_rate": 5.343895861886166e-06, "loss": 0.0073, "step": 28990 }, { "epoch": 5.32, "learning_rate": 5.34323883172061e-06, "loss": 0.0102, "step": 28991 }, { "epoch": 5.32, "learning_rate": 5.342581827223232e-06, "loss": 0.0095, "step": 28992 }, { "epoch": 5.32, "learning_rate": 5.341924848397655e-06, "loss": 0.0079, "step": 28993 }, { "epoch": 5.32, "learning_rate": 5.341267895247496e-06, "loss": 0.0083, "step": 28994 }, { "epoch": 5.32, "learning_rate": 5.340610967776378e-06, "loss": 0.0089, "step": 28995 }, { "epoch": 5.32, "learning_rate": 5.339954065987921e-06, "loss": 0.0101, "step": 28996 }, { "epoch": 5.32, "learning_rate": 5.339297189885746e-06, "loss": 0.0077, "step": 28997 }, { "epoch": 5.32, "learning_rate": 5.338640339473475e-06, "loss": 0.0079, "step": 28998 }, { "epoch": 5.32, "learning_rate": 5.337983514754723e-06, "loss": 0.0088, "step": 28999 }, { "epoch": 5.32, "learning_rate": 5.337326715733124e-06, "loss": 0.0071, "step": 29000 }, { "epoch": 5.32, "learning_rate": 5.3366699424122845e-06, "loss": 0.009, "step": 29001 }, { "epoch": 5.32, "learning_rate": 5.33601319479583e-06, "loss": 0.0127, "step": 29002 }, { "epoch": 5.32, "learning_rate": 5.335356472887381e-06, "loss": 0.0093, "step": 29003 }, { "epoch": 5.32, "learning_rate": 5.33469977669055e-06, "loss": 0.0085, "step": 29004 }, { "epoch": 5.32, "learning_rate": 5.3340431062089685e-06, "loss": 0.0061, "step": 29005 }, { "epoch": 5.32, "learning_rate": 5.33338646144625e-06, "loss": 0.0097, "step": 29006 }, { "epoch": 5.32, "learning_rate": 5.33272984240602e-06, "loss": 0.0112, "step": 29007 }, { "epoch": 5.32, "learning_rate": 5.3320732490918844e-06, "loss": 0.012, "step": 29008 }, { "epoch": 5.32, "learning_rate": 5.331416681507473e-06, "loss": 0.0114, "step": 29009 }, { "epoch": 5.32, "learning_rate": 5.330760139656403e-06, "loss": 0.0093, "step": 29010 }, { "epoch": 5.32, "learning_rate": 5.3301036235422935e-06, "loss": 0.007, "step": 29011 }, { "epoch": 5.32, "learning_rate": 5.32944713316876e-06, "loss": 0.0091, "step": 29012 }, { "epoch": 5.32, "learning_rate": 5.3287906685394255e-06, "loss": 0.0092, "step": 29013 }, { "epoch": 5.32, "learning_rate": 5.328134229657904e-06, "loss": 0.0083, "step": 29014 }, { "epoch": 5.32, "learning_rate": 5.327477816527817e-06, "loss": 0.0102, "step": 29015 }, { "epoch": 5.32, "learning_rate": 5.326821429152783e-06, "loss": 0.0081, "step": 29016 }, { "epoch": 5.32, "learning_rate": 5.326165067536414e-06, "loss": 0.0102, "step": 29017 }, { "epoch": 5.32, "learning_rate": 5.325508731682341e-06, "loss": 0.0073, "step": 29018 }, { "epoch": 5.32, "learning_rate": 5.324852421594169e-06, "loss": 0.01, "step": 29019 }, { "epoch": 5.32, "learning_rate": 5.324196137275519e-06, "loss": 0.0072, "step": 29020 }, { "epoch": 5.32, "learning_rate": 5.323539878730006e-06, "loss": 0.0082, "step": 29021 }, { "epoch": 5.32, "learning_rate": 5.322883645961256e-06, "loss": 0.0131, "step": 29022 }, { "epoch": 5.32, "learning_rate": 5.32222743897288e-06, "loss": 0.0085, "step": 29023 }, { "epoch": 5.32, "learning_rate": 5.321571257768496e-06, "loss": 0.0063, "step": 29024 }, { "epoch": 5.32, "learning_rate": 5.320915102351725e-06, "loss": 0.0095, "step": 29025 }, { "epoch": 5.32, "learning_rate": 5.320258972726171e-06, "loss": 0.0104, "step": 29026 }, { "epoch": 5.32, "learning_rate": 5.319602868895463e-06, "loss": 0.0076, "step": 29027 }, { "epoch": 5.32, "learning_rate": 5.318946790863214e-06, "loss": 0.0086, "step": 29028 }, { "epoch": 5.32, "learning_rate": 5.318290738633041e-06, "loss": 0.0124, "step": 29029 }, { "epoch": 5.32, "learning_rate": 5.317634712208557e-06, "loss": 0.0081, "step": 29030 }, { "epoch": 5.32, "learning_rate": 5.316978711593381e-06, "loss": 0.0112, "step": 29031 }, { "epoch": 5.32, "learning_rate": 5.3163227367911285e-06, "loss": 0.0075, "step": 29032 }, { "epoch": 5.32, "learning_rate": 5.315666787805416e-06, "loss": 0.0083, "step": 29033 }, { "epoch": 5.32, "learning_rate": 5.315010864639856e-06, "loss": 0.0075, "step": 29034 }, { "epoch": 5.32, "learning_rate": 5.314354967298063e-06, "loss": 0.0078, "step": 29035 }, { "epoch": 5.32, "learning_rate": 5.3136990957836624e-06, "loss": 0.0103, "step": 29036 }, { "epoch": 5.32, "learning_rate": 5.313043250100258e-06, "loss": 0.01, "step": 29037 }, { "epoch": 5.33, "learning_rate": 5.31238743025147e-06, "loss": 0.0089, "step": 29038 }, { "epoch": 5.33, "learning_rate": 5.3117316362409075e-06, "loss": 0.0091, "step": 29039 }, { "epoch": 5.33, "learning_rate": 5.311075868072195e-06, "loss": 0.0073, "step": 29040 }, { "epoch": 5.33, "learning_rate": 5.310420125748941e-06, "loss": 0.0087, "step": 29041 }, { "epoch": 5.33, "learning_rate": 5.309764409274763e-06, "loss": 0.0069, "step": 29042 }, { "epoch": 5.33, "learning_rate": 5.309108718653275e-06, "loss": 0.0093, "step": 29043 }, { "epoch": 5.33, "learning_rate": 5.308453053888083e-06, "loss": 0.0126, "step": 29044 }, { "epoch": 5.33, "learning_rate": 5.307797414982812e-06, "loss": 0.0097, "step": 29045 }, { "epoch": 5.33, "learning_rate": 5.3071418019410704e-06, "loss": 0.0118, "step": 29046 }, { "epoch": 5.33, "learning_rate": 5.3064862147664734e-06, "loss": 0.0086, "step": 29047 }, { "epoch": 5.33, "learning_rate": 5.305830653462634e-06, "loss": 0.0102, "step": 29048 }, { "epoch": 5.33, "learning_rate": 5.3051751180331675e-06, "loss": 0.0092, "step": 29049 }, { "epoch": 5.33, "learning_rate": 5.3045196084816845e-06, "loss": 0.01, "step": 29050 }, { "epoch": 5.33, "learning_rate": 5.3038641248118e-06, "loss": 0.0095, "step": 29051 }, { "epoch": 5.33, "learning_rate": 5.3032086670271225e-06, "loss": 0.0078, "step": 29052 }, { "epoch": 5.33, "learning_rate": 5.302553235131273e-06, "loss": 0.0088, "step": 29053 }, { "epoch": 5.33, "learning_rate": 5.3018978291278635e-06, "loss": 0.0079, "step": 29054 }, { "epoch": 5.33, "learning_rate": 5.301242449020499e-06, "loss": 0.007, "step": 29055 }, { "epoch": 5.33, "learning_rate": 5.300587094812797e-06, "loss": 0.0081, "step": 29056 }, { "epoch": 5.33, "learning_rate": 5.299931766508365e-06, "loss": 0.0076, "step": 29057 }, { "epoch": 5.33, "learning_rate": 5.299276464110823e-06, "loss": 0.0106, "step": 29058 }, { "epoch": 5.33, "learning_rate": 5.298621187623778e-06, "loss": 0.0091, "step": 29059 }, { "epoch": 5.33, "learning_rate": 5.297965937050849e-06, "loss": 0.0079, "step": 29060 }, { "epoch": 5.33, "learning_rate": 5.297310712395634e-06, "loss": 0.0069, "step": 29061 }, { "epoch": 5.33, "learning_rate": 5.296655513661755e-06, "loss": 0.0089, "step": 29062 }, { "epoch": 5.33, "learning_rate": 5.296000340852823e-06, "loss": 0.0103, "step": 29063 }, { "epoch": 5.33, "learning_rate": 5.295345193972445e-06, "loss": 0.0092, "step": 29064 }, { "epoch": 5.33, "learning_rate": 5.2946900730242355e-06, "loss": 0.0074, "step": 29065 }, { "epoch": 5.33, "learning_rate": 5.294034978011805e-06, "loss": 0.0094, "step": 29066 }, { "epoch": 5.33, "learning_rate": 5.293379908938764e-06, "loss": 0.0063, "step": 29067 }, { "epoch": 5.33, "learning_rate": 5.2927248658087225e-06, "loss": 0.0121, "step": 29068 }, { "epoch": 5.33, "learning_rate": 5.292069848625292e-06, "loss": 0.011, "step": 29069 }, { "epoch": 5.33, "learning_rate": 5.291414857392079e-06, "loss": 0.0098, "step": 29070 }, { "epoch": 5.33, "learning_rate": 5.290759892112702e-06, "loss": 0.0082, "step": 29071 }, { "epoch": 5.33, "learning_rate": 5.29010495279077e-06, "loss": 0.0094, "step": 29072 }, { "epoch": 5.33, "learning_rate": 5.289450039429887e-06, "loss": 0.0117, "step": 29073 }, { "epoch": 5.33, "learning_rate": 5.28879515203366e-06, "loss": 0.0096, "step": 29074 }, { "epoch": 5.33, "learning_rate": 5.28814029060571e-06, "loss": 0.0084, "step": 29075 }, { "epoch": 5.33, "learning_rate": 5.28748545514964e-06, "loss": 0.0105, "step": 29076 }, { "epoch": 5.33, "learning_rate": 5.286830645669061e-06, "loss": 0.0097, "step": 29077 }, { "epoch": 5.33, "learning_rate": 5.2861758621675865e-06, "loss": 0.0081, "step": 29078 }, { "epoch": 5.33, "learning_rate": 5.285521104648812e-06, "loss": 0.0133, "step": 29079 }, { "epoch": 5.33, "learning_rate": 5.284866373116361e-06, "loss": 0.0109, "step": 29080 }, { "epoch": 5.33, "learning_rate": 5.284211667573835e-06, "loss": 0.0085, "step": 29081 }, { "epoch": 5.33, "learning_rate": 5.283556988024847e-06, "loss": 0.0053, "step": 29082 }, { "epoch": 5.33, "learning_rate": 5.282902334473003e-06, "loss": 0.0108, "step": 29083 }, { "epoch": 5.33, "learning_rate": 5.28224770692191e-06, "loss": 0.0105, "step": 29084 }, { "epoch": 5.33, "learning_rate": 5.28159310537518e-06, "loss": 0.0105, "step": 29085 }, { "epoch": 5.33, "learning_rate": 5.280938529836419e-06, "loss": 0.0081, "step": 29086 }, { "epoch": 5.33, "learning_rate": 5.280283980309236e-06, "loss": 0.0066, "step": 29087 }, { "epoch": 5.33, "learning_rate": 5.279629456797233e-06, "loss": 0.0103, "step": 29088 }, { "epoch": 5.33, "learning_rate": 5.2789749593040265e-06, "loss": 0.009, "step": 29089 }, { "epoch": 5.33, "learning_rate": 5.2783204878332265e-06, "loss": 0.0086, "step": 29090 }, { "epoch": 5.33, "learning_rate": 5.277666042388429e-06, "loss": 0.0072, "step": 29091 }, { "epoch": 5.34, "learning_rate": 5.277011622973243e-06, "loss": 0.0099, "step": 29092 }, { "epoch": 5.34, "learning_rate": 5.276357229591285e-06, "loss": 0.009, "step": 29093 }, { "epoch": 5.34, "learning_rate": 5.275702862246155e-06, "loss": 0.0102, "step": 29094 }, { "epoch": 5.34, "learning_rate": 5.275048520941461e-06, "loss": 0.0113, "step": 29095 }, { "epoch": 5.34, "learning_rate": 5.274394205680817e-06, "loss": 0.0086, "step": 29096 }, { "epoch": 5.34, "learning_rate": 5.2737399164678124e-06, "loss": 0.0083, "step": 29097 }, { "epoch": 5.34, "learning_rate": 5.273085653306069e-06, "loss": 0.0075, "step": 29098 }, { "epoch": 5.34, "learning_rate": 5.2724314161991884e-06, "loss": 0.0056, "step": 29099 }, { "epoch": 5.34, "learning_rate": 5.271777205150776e-06, "loss": 0.0074, "step": 29100 }, { "epoch": 5.34, "learning_rate": 5.271123020164439e-06, "loss": 0.0109, "step": 29101 }, { "epoch": 5.34, "learning_rate": 5.270468861243783e-06, "loss": 0.0089, "step": 29102 }, { "epoch": 5.34, "learning_rate": 5.2698147283924126e-06, "loss": 0.0086, "step": 29103 }, { "epoch": 5.34, "learning_rate": 5.269160621613936e-06, "loss": 0.0128, "step": 29104 }, { "epoch": 5.34, "learning_rate": 5.2685065409119504e-06, "loss": 0.0069, "step": 29105 }, { "epoch": 5.34, "learning_rate": 5.267852486290073e-06, "loss": 0.0121, "step": 29106 }, { "epoch": 5.34, "learning_rate": 5.267198457751908e-06, "loss": 0.0127, "step": 29107 }, { "epoch": 5.34, "learning_rate": 5.26654445530105e-06, "loss": 0.0105, "step": 29108 }, { "epoch": 5.34, "learning_rate": 5.2658904789411114e-06, "loss": 0.006, "step": 29109 }, { "epoch": 5.34, "learning_rate": 5.265236528675691e-06, "loss": 0.0081, "step": 29110 }, { "epoch": 5.34, "learning_rate": 5.264582604508401e-06, "loss": 0.0059, "step": 29111 }, { "epoch": 5.34, "learning_rate": 5.263928706442844e-06, "loss": 0.0082, "step": 29112 }, { "epoch": 5.34, "learning_rate": 5.263274834482623e-06, "loss": 0.0094, "step": 29113 }, { "epoch": 5.34, "learning_rate": 5.262620988631342e-06, "loss": 0.0098, "step": 29114 }, { "epoch": 5.34, "learning_rate": 5.261967168892603e-06, "loss": 0.0086, "step": 29115 }, { "epoch": 5.34, "learning_rate": 5.2613133752700145e-06, "loss": 0.0076, "step": 29116 }, { "epoch": 5.34, "learning_rate": 5.260659607767176e-06, "loss": 0.0085, "step": 29117 }, { "epoch": 5.34, "learning_rate": 5.260005866387694e-06, "loss": 0.0114, "step": 29118 }, { "epoch": 5.34, "learning_rate": 5.259352151135171e-06, "loss": 0.01, "step": 29119 }, { "epoch": 5.34, "learning_rate": 5.2586984620132085e-06, "loss": 0.0133, "step": 29120 }, { "epoch": 5.34, "learning_rate": 5.2580447990254126e-06, "loss": 0.0104, "step": 29121 }, { "epoch": 5.34, "learning_rate": 5.257391162175384e-06, "loss": 0.0082, "step": 29122 }, { "epoch": 5.34, "learning_rate": 5.2567375514667215e-06, "loss": 0.0082, "step": 29123 }, { "epoch": 5.34, "learning_rate": 5.256083966903038e-06, "loss": 0.0068, "step": 29124 }, { "epoch": 5.34, "learning_rate": 5.255430408487935e-06, "loss": 0.007, "step": 29125 }, { "epoch": 5.34, "learning_rate": 5.254776876225006e-06, "loss": 0.0083, "step": 29126 }, { "epoch": 5.34, "learning_rate": 5.254123370117854e-06, "loss": 0.0075, "step": 29127 }, { "epoch": 5.34, "learning_rate": 5.253469890170091e-06, "loss": 0.0091, "step": 29128 }, { "epoch": 5.34, "learning_rate": 5.252816436385311e-06, "loss": 0.0125, "step": 29129 }, { "epoch": 5.34, "learning_rate": 5.2521630087671196e-06, "loss": 0.0116, "step": 29130 }, { "epoch": 5.34, "learning_rate": 5.251509607319115e-06, "loss": 0.0088, "step": 29131 }, { "epoch": 5.34, "learning_rate": 5.250856232044902e-06, "loss": 0.0135, "step": 29132 }, { "epoch": 5.34, "learning_rate": 5.25020288294808e-06, "loss": 0.0097, "step": 29133 }, { "epoch": 5.34, "learning_rate": 5.249549560032252e-06, "loss": 0.0073, "step": 29134 }, { "epoch": 5.34, "learning_rate": 5.248896263301017e-06, "loss": 0.0088, "step": 29135 }, { "epoch": 5.34, "learning_rate": 5.248242992757975e-06, "loss": 0.0085, "step": 29136 }, { "epoch": 5.34, "learning_rate": 5.247589748406734e-06, "loss": 0.0072, "step": 29137 }, { "epoch": 5.34, "learning_rate": 5.246936530250888e-06, "loss": 0.0102, "step": 29138 }, { "epoch": 5.34, "learning_rate": 5.246283338294038e-06, "loss": 0.0094, "step": 29139 }, { "epoch": 5.34, "learning_rate": 5.245630172539786e-06, "loss": 0.0088, "step": 29140 }, { "epoch": 5.34, "learning_rate": 5.2449770329917285e-06, "loss": 0.0082, "step": 29141 }, { "epoch": 5.34, "learning_rate": 5.244323919653472e-06, "loss": 0.0108, "step": 29142 }, { "epoch": 5.34, "learning_rate": 5.2436708325286175e-06, "loss": 0.0081, "step": 29143 }, { "epoch": 5.34, "learning_rate": 5.243017771620757e-06, "loss": 0.0066, "step": 29144 }, { "epoch": 5.34, "learning_rate": 5.24236473693349e-06, "loss": 0.0097, "step": 29145 }, { "epoch": 5.34, "learning_rate": 5.241711728470425e-06, "loss": 0.0088, "step": 29146 }, { "epoch": 5.35, "learning_rate": 5.2410587462351556e-06, "loss": 0.0108, "step": 29147 }, { "epoch": 5.35, "learning_rate": 5.240405790231282e-06, "loss": 0.0083, "step": 29148 }, { "epoch": 5.35, "learning_rate": 5.239752860462404e-06, "loss": 0.0105, "step": 29149 }, { "epoch": 5.35, "learning_rate": 5.239099956932118e-06, "loss": 0.0097, "step": 29150 }, { "epoch": 5.35, "learning_rate": 5.2384470796440245e-06, "loss": 0.0091, "step": 29151 }, { "epoch": 5.35, "learning_rate": 5.2377942286017245e-06, "loss": 0.0085, "step": 29152 }, { "epoch": 5.35, "learning_rate": 5.237141403808812e-06, "loss": 0.0057, "step": 29153 }, { "epoch": 5.35, "learning_rate": 5.236488605268889e-06, "loss": 0.0095, "step": 29154 }, { "epoch": 5.35, "learning_rate": 5.235835832985552e-06, "loss": 0.0084, "step": 29155 }, { "epoch": 5.35, "learning_rate": 5.2351830869624e-06, "loss": 0.01, "step": 29156 }, { "epoch": 5.35, "learning_rate": 5.23453036720303e-06, "loss": 0.0126, "step": 29157 }, { "epoch": 5.35, "learning_rate": 5.2338776737110365e-06, "loss": 0.0083, "step": 29158 }, { "epoch": 5.35, "learning_rate": 5.233225006490023e-06, "loss": 0.011, "step": 29159 }, { "epoch": 5.35, "learning_rate": 5.232572365543588e-06, "loss": 0.0072, "step": 29160 }, { "epoch": 5.35, "learning_rate": 5.231919750875327e-06, "loss": 0.0093, "step": 29161 }, { "epoch": 5.35, "learning_rate": 5.231267162488835e-06, "loss": 0.0099, "step": 29162 }, { "epoch": 5.35, "learning_rate": 5.230614600387705e-06, "loss": 0.008, "step": 29163 }, { "epoch": 5.35, "learning_rate": 5.229962064575542e-06, "loss": 0.0099, "step": 29164 }, { "epoch": 5.35, "learning_rate": 5.229309555055941e-06, "loss": 0.0077, "step": 29165 }, { "epoch": 5.35, "learning_rate": 5.2286570718324956e-06, "loss": 0.0127, "step": 29166 }, { "epoch": 5.35, "learning_rate": 5.228004614908806e-06, "loss": 0.0099, "step": 29167 }, { "epoch": 5.35, "learning_rate": 5.227352184288466e-06, "loss": 0.0071, "step": 29168 }, { "epoch": 5.35, "learning_rate": 5.226699779975073e-06, "loss": 0.0074, "step": 29169 }, { "epoch": 5.35, "learning_rate": 5.226047401972223e-06, "loss": 0.0095, "step": 29170 }, { "epoch": 5.35, "learning_rate": 5.225395050283511e-06, "loss": 0.0078, "step": 29171 }, { "epoch": 5.35, "learning_rate": 5.224742724912533e-06, "loss": 0.0098, "step": 29172 }, { "epoch": 5.35, "learning_rate": 5.224090425862885e-06, "loss": 0.0086, "step": 29173 }, { "epoch": 5.35, "learning_rate": 5.2234381531381626e-06, "loss": 0.0085, "step": 29174 }, { "epoch": 5.35, "learning_rate": 5.2227859067419624e-06, "loss": 0.0076, "step": 29175 }, { "epoch": 5.35, "learning_rate": 5.222133686677873e-06, "loss": 0.0079, "step": 29176 }, { "epoch": 5.35, "learning_rate": 5.221481492949499e-06, "loss": 0.0084, "step": 29177 }, { "epoch": 5.35, "learning_rate": 5.2208293255604305e-06, "loss": 0.0085, "step": 29178 }, { "epoch": 5.35, "learning_rate": 5.2201771845142666e-06, "loss": 0.0088, "step": 29179 }, { "epoch": 5.35, "learning_rate": 5.219525069814591e-06, "loss": 0.012, "step": 29180 }, { "epoch": 5.35, "learning_rate": 5.218872981465009e-06, "loss": 0.009, "step": 29181 }, { "epoch": 5.35, "learning_rate": 5.21822091946911e-06, "loss": 0.0072, "step": 29182 }, { "epoch": 5.35, "learning_rate": 5.217568883830491e-06, "loss": 0.0097, "step": 29183 }, { "epoch": 5.35, "learning_rate": 5.216916874552742e-06, "loss": 0.0078, "step": 29184 }, { "epoch": 5.35, "learning_rate": 5.216264891639461e-06, "loss": 0.0092, "step": 29185 }, { "epoch": 5.35, "learning_rate": 5.215612935094239e-06, "loss": 0.0083, "step": 29186 }, { "epoch": 5.35, "learning_rate": 5.214961004920671e-06, "loss": 0.009, "step": 29187 }, { "epoch": 5.35, "learning_rate": 5.21430910112235e-06, "loss": 0.0106, "step": 29188 }, { "epoch": 5.35, "learning_rate": 5.213657223702866e-06, "loss": 0.0101, "step": 29189 }, { "epoch": 5.35, "learning_rate": 5.2130053726658224e-06, "loss": 0.0112, "step": 29190 }, { "epoch": 5.35, "learning_rate": 5.2123535480148035e-06, "loss": 0.0082, "step": 29191 }, { "epoch": 5.35, "learning_rate": 5.211701749753402e-06, "loss": 0.0102, "step": 29192 }, { "epoch": 5.35, "learning_rate": 5.211049977885213e-06, "loss": 0.0094, "step": 29193 }, { "epoch": 5.35, "learning_rate": 5.210398232413824e-06, "loss": 0.0106, "step": 29194 }, { "epoch": 5.35, "learning_rate": 5.209746513342838e-06, "loss": 0.0113, "step": 29195 }, { "epoch": 5.35, "learning_rate": 5.209094820675841e-06, "loss": 0.0105, "step": 29196 }, { "epoch": 5.35, "learning_rate": 5.208443154416429e-06, "loss": 0.0098, "step": 29197 }, { "epoch": 5.35, "learning_rate": 5.207791514568182e-06, "loss": 0.0071, "step": 29198 }, { "epoch": 5.35, "learning_rate": 5.207139901134706e-06, "loss": 0.0131, "step": 29199 }, { "epoch": 5.35, "learning_rate": 5.206488314119586e-06, "loss": 0.0091, "step": 29200 }, { "epoch": 5.36, "learning_rate": 5.205836753526414e-06, "loss": 0.0077, "step": 29201 }, { "epoch": 5.36, "learning_rate": 5.205185219358784e-06, "loss": 0.0091, "step": 29202 }, { "epoch": 5.36, "learning_rate": 5.204533711620285e-06, "loss": 0.0081, "step": 29203 }, { "epoch": 5.36, "learning_rate": 5.203882230314508e-06, "loss": 0.011, "step": 29204 }, { "epoch": 5.36, "learning_rate": 5.203230775445043e-06, "loss": 0.0062, "step": 29205 }, { "epoch": 5.36, "learning_rate": 5.202579347015484e-06, "loss": 0.0115, "step": 29206 }, { "epoch": 5.36, "learning_rate": 5.201927945029417e-06, "loss": 0.0068, "step": 29207 }, { "epoch": 5.36, "learning_rate": 5.201276569490441e-06, "loss": 0.0098, "step": 29208 }, { "epoch": 5.36, "learning_rate": 5.200625220402139e-06, "loss": 0.009, "step": 29209 }, { "epoch": 5.36, "learning_rate": 5.199973897768103e-06, "loss": 0.011, "step": 29210 }, { "epoch": 5.36, "learning_rate": 5.199322601591918e-06, "loss": 0.0086, "step": 29211 }, { "epoch": 5.36, "learning_rate": 5.198671331877185e-06, "loss": 0.0099, "step": 29212 }, { "epoch": 5.36, "learning_rate": 5.198020088627487e-06, "loss": 0.0098, "step": 29213 }, { "epoch": 5.36, "learning_rate": 5.1973688718464195e-06, "loss": 0.0104, "step": 29214 }, { "epoch": 5.36, "learning_rate": 5.196717681537563e-06, "loss": 0.0079, "step": 29215 }, { "epoch": 5.36, "learning_rate": 5.1960665177045075e-06, "loss": 0.0091, "step": 29216 }, { "epoch": 5.36, "learning_rate": 5.19541538035085e-06, "loss": 0.0102, "step": 29217 }, { "epoch": 5.36, "learning_rate": 5.194764269480175e-06, "loss": 0.0104, "step": 29218 }, { "epoch": 5.36, "learning_rate": 5.194113185096071e-06, "loss": 0.008, "step": 29219 }, { "epoch": 5.36, "learning_rate": 5.193462127202129e-06, "loss": 0.0122, "step": 29220 }, { "epoch": 5.36, "learning_rate": 5.192811095801935e-06, "loss": 0.0091, "step": 29221 }, { "epoch": 5.36, "learning_rate": 5.192160090899081e-06, "loss": 0.0088, "step": 29222 }, { "epoch": 5.36, "learning_rate": 5.191509112497152e-06, "loss": 0.0084, "step": 29223 }, { "epoch": 5.36, "learning_rate": 5.190858160599736e-06, "loss": 0.0074, "step": 29224 }, { "epoch": 5.36, "learning_rate": 5.190207235210421e-06, "loss": 0.0076, "step": 29225 }, { "epoch": 5.36, "learning_rate": 5.189556336332804e-06, "loss": 0.0078, "step": 29226 }, { "epoch": 5.36, "learning_rate": 5.18890546397046e-06, "loss": 0.0087, "step": 29227 }, { "epoch": 5.36, "learning_rate": 5.188254618126982e-06, "loss": 0.0101, "step": 29228 }, { "epoch": 5.36, "learning_rate": 5.187603798805953e-06, "loss": 0.0108, "step": 29229 }, { "epoch": 5.36, "learning_rate": 5.18695300601097e-06, "loss": 0.0099, "step": 29230 }, { "epoch": 5.36, "learning_rate": 5.186302239745614e-06, "loss": 0.0104, "step": 29231 }, { "epoch": 5.36, "learning_rate": 5.185651500013476e-06, "loss": 0.0116, "step": 29232 }, { "epoch": 5.36, "learning_rate": 5.185000786818133e-06, "loss": 0.0107, "step": 29233 }, { "epoch": 5.36, "learning_rate": 5.184350100163182e-06, "loss": 0.0107, "step": 29234 }, { "epoch": 5.36, "learning_rate": 5.1836994400522066e-06, "loss": 0.0089, "step": 29235 }, { "epoch": 5.36, "learning_rate": 5.183048806488793e-06, "loss": 0.0082, "step": 29236 }, { "epoch": 5.36, "learning_rate": 5.182398199476526e-06, "loss": 0.0076, "step": 29237 }, { "epoch": 5.36, "learning_rate": 5.181747619018993e-06, "loss": 0.0096, "step": 29238 }, { "epoch": 5.36, "learning_rate": 5.18109706511978e-06, "loss": 0.0095, "step": 29239 }, { "epoch": 5.36, "learning_rate": 5.180446537782473e-06, "loss": 0.0119, "step": 29240 }, { "epoch": 5.36, "learning_rate": 5.179796037010658e-06, "loss": 0.0117, "step": 29241 }, { "epoch": 5.36, "learning_rate": 5.179145562807917e-06, "loss": 0.0084, "step": 29242 }, { "epoch": 5.36, "learning_rate": 5.178495115177842e-06, "loss": 0.0103, "step": 29243 }, { "epoch": 5.36, "learning_rate": 5.177844694124018e-06, "loss": 0.0088, "step": 29244 }, { "epoch": 5.36, "learning_rate": 5.177194299650023e-06, "loss": 0.0079, "step": 29245 }, { "epoch": 5.36, "learning_rate": 5.176543931759447e-06, "loss": 0.0082, "step": 29246 }, { "epoch": 5.36, "learning_rate": 5.1758935904558694e-06, "loss": 0.0071, "step": 29247 }, { "epoch": 5.36, "learning_rate": 5.175243275742883e-06, "loss": 0.0079, "step": 29248 }, { "epoch": 5.36, "learning_rate": 5.17459298762407e-06, "loss": 0.01, "step": 29249 }, { "epoch": 5.36, "learning_rate": 5.173942726103016e-06, "loss": 0.0062, "step": 29250 }, { "epoch": 5.36, "learning_rate": 5.173292491183295e-06, "loss": 0.0074, "step": 29251 }, { "epoch": 5.36, "learning_rate": 5.172642282868503e-06, "loss": 0.0085, "step": 29252 }, { "epoch": 5.36, "learning_rate": 5.17199210116222e-06, "loss": 0.0092, "step": 29253 }, { "epoch": 5.36, "learning_rate": 5.171341946068029e-06, "loss": 0.0094, "step": 29254 }, { "epoch": 5.36, "learning_rate": 5.170691817589516e-06, "loss": 0.0097, "step": 29255 }, { "epoch": 5.37, "learning_rate": 5.170041715730262e-06, "loss": 0.0133, "step": 29256 }, { "epoch": 5.37, "learning_rate": 5.16939164049385e-06, "loss": 0.0082, "step": 29257 }, { "epoch": 5.37, "learning_rate": 5.168741591883866e-06, "loss": 0.0085, "step": 29258 }, { "epoch": 5.37, "learning_rate": 5.168091569903892e-06, "loss": 0.0084, "step": 29259 }, { "epoch": 5.37, "learning_rate": 5.1674415745575045e-06, "loss": 0.0075, "step": 29260 }, { "epoch": 5.37, "learning_rate": 5.166791605848301e-06, "loss": 0.0065, "step": 29261 }, { "epoch": 5.37, "learning_rate": 5.166141663779853e-06, "loss": 0.0074, "step": 29262 }, { "epoch": 5.37, "learning_rate": 5.165491748355743e-06, "loss": 0.0088, "step": 29263 }, { "epoch": 5.37, "learning_rate": 5.164841859579552e-06, "loss": 0.0106, "step": 29264 }, { "epoch": 5.37, "learning_rate": 5.164191997454871e-06, "loss": 0.008, "step": 29265 }, { "epoch": 5.37, "learning_rate": 5.163542161985278e-06, "loss": 0.0132, "step": 29266 }, { "epoch": 5.37, "learning_rate": 5.162892353174351e-06, "loss": 0.0095, "step": 29267 }, { "epoch": 5.37, "learning_rate": 5.162242571025682e-06, "loss": 0.0114, "step": 29268 }, { "epoch": 5.37, "learning_rate": 5.161592815542835e-06, "loss": 0.0084, "step": 29269 }, { "epoch": 5.37, "learning_rate": 5.160943086729406e-06, "loss": 0.0095, "step": 29270 }, { "epoch": 5.37, "learning_rate": 5.160293384588973e-06, "loss": 0.0079, "step": 29271 }, { "epoch": 5.37, "learning_rate": 5.159643709125115e-06, "loss": 0.0111, "step": 29272 }, { "epoch": 5.37, "learning_rate": 5.158994060341415e-06, "loss": 0.0094, "step": 29273 }, { "epoch": 5.37, "learning_rate": 5.158344438241454e-06, "loss": 0.0102, "step": 29274 }, { "epoch": 5.37, "learning_rate": 5.15769484282881e-06, "loss": 0.0083, "step": 29275 }, { "epoch": 5.37, "learning_rate": 5.157045274107066e-06, "loss": 0.0071, "step": 29276 }, { "epoch": 5.37, "learning_rate": 5.156395732079799e-06, "loss": 0.008, "step": 29277 }, { "epoch": 5.37, "learning_rate": 5.155746216750594e-06, "loss": 0.007, "step": 29278 }, { "epoch": 5.37, "learning_rate": 5.155096728123037e-06, "loss": 0.0082, "step": 29279 }, { "epoch": 5.37, "learning_rate": 5.154447266200693e-06, "loss": 0.0079, "step": 29280 }, { "epoch": 5.37, "learning_rate": 5.1537978309871505e-06, "loss": 0.0084, "step": 29281 }, { "epoch": 5.37, "learning_rate": 5.153148422485984e-06, "loss": 0.0089, "step": 29282 }, { "epoch": 5.37, "learning_rate": 5.15249904070078e-06, "loss": 0.0081, "step": 29283 }, { "epoch": 5.37, "learning_rate": 5.151849685635116e-06, "loss": 0.0105, "step": 29284 }, { "epoch": 5.37, "learning_rate": 5.15120035729257e-06, "loss": 0.0066, "step": 29285 }, { "epoch": 5.37, "learning_rate": 5.15055105567672e-06, "loss": 0.0087, "step": 29286 }, { "epoch": 5.37, "learning_rate": 5.149901780791148e-06, "loss": 0.0126, "step": 29287 }, { "epoch": 5.37, "learning_rate": 5.14925253263943e-06, "loss": 0.0091, "step": 29288 }, { "epoch": 5.37, "learning_rate": 5.148603311225146e-06, "loss": 0.0088, "step": 29289 }, { "epoch": 5.37, "learning_rate": 5.147954116551875e-06, "loss": 0.0078, "step": 29290 }, { "epoch": 5.37, "learning_rate": 5.147304948623193e-06, "loss": 0.0096, "step": 29291 }, { "epoch": 5.37, "learning_rate": 5.146655807442679e-06, "loss": 0.0077, "step": 29292 }, { "epoch": 5.37, "learning_rate": 5.146006693013914e-06, "loss": 0.0072, "step": 29293 }, { "epoch": 5.37, "learning_rate": 5.145357605340473e-06, "loss": 0.0083, "step": 29294 }, { "epoch": 5.37, "learning_rate": 5.144708544425931e-06, "loss": 0.008, "step": 29295 }, { "epoch": 5.37, "learning_rate": 5.144059510273873e-06, "loss": 0.009, "step": 29296 }, { "epoch": 5.37, "learning_rate": 5.143410502887877e-06, "loss": 0.0102, "step": 29297 }, { "epoch": 5.37, "learning_rate": 5.1427615222715114e-06, "loss": 0.0089, "step": 29298 }, { "epoch": 5.37, "learning_rate": 5.142112568428359e-06, "loss": 0.0077, "step": 29299 }, { "epoch": 5.37, "learning_rate": 5.141463641361991e-06, "loss": 0.0093, "step": 29300 }, { "epoch": 5.37, "learning_rate": 5.140814741075995e-06, "loss": 0.0112, "step": 29301 }, { "epoch": 5.37, "learning_rate": 5.14016586757394e-06, "loss": 0.0075, "step": 29302 }, { "epoch": 5.37, "learning_rate": 5.139517020859405e-06, "loss": 0.0054, "step": 29303 }, { "epoch": 5.37, "learning_rate": 5.138868200935966e-06, "loss": 0.0093, "step": 29304 }, { "epoch": 5.37, "learning_rate": 5.138219407807199e-06, "loss": 0.007, "step": 29305 }, { "epoch": 5.37, "learning_rate": 5.137570641476682e-06, "loss": 0.0104, "step": 29306 }, { "epoch": 5.37, "learning_rate": 5.136921901947989e-06, "loss": 0.0108, "step": 29307 }, { "epoch": 5.37, "learning_rate": 5.136273189224695e-06, "loss": 0.0091, "step": 29308 }, { "epoch": 5.37, "learning_rate": 5.135624503310379e-06, "loss": 0.0097, "step": 29309 }, { "epoch": 5.38, "learning_rate": 5.134975844208614e-06, "loss": 0.0102, "step": 29310 }, { "epoch": 5.38, "learning_rate": 5.134327211922976e-06, "loss": 0.0075, "step": 29311 }, { "epoch": 5.38, "learning_rate": 5.13367860645704e-06, "loss": 0.0069, "step": 29312 }, { "epoch": 5.38, "learning_rate": 5.1330300278143786e-06, "loss": 0.0089, "step": 29313 }, { "epoch": 5.38, "learning_rate": 5.132381475998573e-06, "loss": 0.0084, "step": 29314 }, { "epoch": 5.38, "learning_rate": 5.131732951013199e-06, "loss": 0.0082, "step": 29315 }, { "epoch": 5.38, "learning_rate": 5.131084452861822e-06, "loss": 0.0081, "step": 29316 }, { "epoch": 5.38, "learning_rate": 5.1304359815480185e-06, "loss": 0.0106, "step": 29317 }, { "epoch": 5.38, "learning_rate": 5.12978753707537e-06, "loss": 0.0102, "step": 29318 }, { "epoch": 5.38, "learning_rate": 5.129139119447446e-06, "loss": 0.0107, "step": 29319 }, { "epoch": 5.38, "learning_rate": 5.128490728667823e-06, "loss": 0.0091, "step": 29320 }, { "epoch": 5.38, "learning_rate": 5.127842364740076e-06, "loss": 0.0085, "step": 29321 }, { "epoch": 5.38, "learning_rate": 5.127194027667768e-06, "loss": 0.0095, "step": 29322 }, { "epoch": 5.38, "learning_rate": 5.126545717454485e-06, "loss": 0.01, "step": 29323 }, { "epoch": 5.38, "learning_rate": 5.125897434103795e-06, "loss": 0.0075, "step": 29324 }, { "epoch": 5.38, "learning_rate": 5.125249177619274e-06, "loss": 0.0082, "step": 29325 }, { "epoch": 5.38, "learning_rate": 5.124600948004493e-06, "loss": 0.0107, "step": 29326 }, { "epoch": 5.38, "learning_rate": 5.123952745263025e-06, "loss": 0.0094, "step": 29327 }, { "epoch": 5.38, "learning_rate": 5.1233045693984465e-06, "loss": 0.0082, "step": 29328 }, { "epoch": 5.38, "learning_rate": 5.1226564204143245e-06, "loss": 0.01, "step": 29329 }, { "epoch": 5.38, "learning_rate": 5.122008298314232e-06, "loss": 0.009, "step": 29330 }, { "epoch": 5.38, "learning_rate": 5.1213602031017485e-06, "loss": 0.0105, "step": 29331 }, { "epoch": 5.38, "learning_rate": 5.120712134780441e-06, "loss": 0.008, "step": 29332 }, { "epoch": 5.38, "learning_rate": 5.120064093353887e-06, "loss": 0.0073, "step": 29333 }, { "epoch": 5.38, "learning_rate": 5.119416078825649e-06, "loss": 0.0073, "step": 29334 }, { "epoch": 5.38, "learning_rate": 5.118768091199301e-06, "loss": 0.0072, "step": 29335 }, { "epoch": 5.38, "learning_rate": 5.118120130478421e-06, "loss": 0.0084, "step": 29336 }, { "epoch": 5.38, "learning_rate": 5.117472196666578e-06, "loss": 0.0097, "step": 29337 }, { "epoch": 5.38, "learning_rate": 5.116824289767342e-06, "loss": 0.0094, "step": 29338 }, { "epoch": 5.38, "learning_rate": 5.1161764097842844e-06, "loss": 0.0089, "step": 29339 }, { "epoch": 5.38, "learning_rate": 5.115528556720978e-06, "loss": 0.0101, "step": 29340 }, { "epoch": 5.38, "learning_rate": 5.114880730580991e-06, "loss": 0.0107, "step": 29341 }, { "epoch": 5.38, "learning_rate": 5.114232931367897e-06, "loss": 0.0107, "step": 29342 }, { "epoch": 5.38, "learning_rate": 5.113585159085266e-06, "loss": 0.0108, "step": 29343 }, { "epoch": 5.38, "learning_rate": 5.112937413736667e-06, "loss": 0.0102, "step": 29344 }, { "epoch": 5.38, "learning_rate": 5.112289695325673e-06, "loss": 0.0104, "step": 29345 }, { "epoch": 5.38, "learning_rate": 5.111642003855853e-06, "loss": 0.0105, "step": 29346 }, { "epoch": 5.38, "learning_rate": 5.110994339330775e-06, "loss": 0.0052, "step": 29347 }, { "epoch": 5.38, "learning_rate": 5.110346701754007e-06, "loss": 0.0071, "step": 29348 }, { "epoch": 5.38, "learning_rate": 5.1096990911291265e-06, "loss": 0.0093, "step": 29349 }, { "epoch": 5.38, "learning_rate": 5.1090515074596995e-06, "loss": 0.0108, "step": 29350 }, { "epoch": 5.38, "learning_rate": 5.108403950749299e-06, "loss": 0.0088, "step": 29351 }, { "epoch": 5.38, "learning_rate": 5.107756421001486e-06, "loss": 0.008, "step": 29352 }, { "epoch": 5.38, "learning_rate": 5.107108918219832e-06, "loss": 0.008, "step": 29353 }, { "epoch": 5.38, "learning_rate": 5.1064614424079106e-06, "loss": 0.01, "step": 29354 }, { "epoch": 5.38, "learning_rate": 5.105813993569289e-06, "loss": 0.0064, "step": 29355 }, { "epoch": 5.38, "learning_rate": 5.105166571707536e-06, "loss": 0.0074, "step": 29356 }, { "epoch": 5.38, "learning_rate": 5.104519176826219e-06, "loss": 0.0114, "step": 29357 }, { "epoch": 5.38, "learning_rate": 5.103871808928906e-06, "loss": 0.0087, "step": 29358 }, { "epoch": 5.38, "learning_rate": 5.1032244680191674e-06, "loss": 0.0114, "step": 29359 }, { "epoch": 5.38, "learning_rate": 5.10257715410057e-06, "loss": 0.0084, "step": 29360 }, { "epoch": 5.38, "learning_rate": 5.101929867176678e-06, "loss": 0.0112, "step": 29361 }, { "epoch": 5.38, "learning_rate": 5.1012826072510715e-06, "loss": 0.0096, "step": 29362 }, { "epoch": 5.38, "learning_rate": 5.100635374327305e-06, "loss": 0.0095, "step": 29363 }, { "epoch": 5.38, "learning_rate": 5.099988168408952e-06, "loss": 0.0084, "step": 29364 }, { "epoch": 5.39, "learning_rate": 5.099340989499579e-06, "loss": 0.0071, "step": 29365 }, { "epoch": 5.39, "learning_rate": 5.098693837602751e-06, "loss": 0.0081, "step": 29366 }, { "epoch": 5.39, "learning_rate": 5.098046712722038e-06, "loss": 0.0094, "step": 29367 }, { "epoch": 5.39, "learning_rate": 5.097399614861012e-06, "loss": 0.0124, "step": 29368 }, { "epoch": 5.39, "learning_rate": 5.0967525440232314e-06, "loss": 0.0098, "step": 29369 }, { "epoch": 5.39, "learning_rate": 5.096105500212259e-06, "loss": 0.0093, "step": 29370 }, { "epoch": 5.39, "learning_rate": 5.095458483431674e-06, "loss": 0.0077, "step": 29371 }, { "epoch": 5.39, "learning_rate": 5.094811493685037e-06, "loss": 0.0121, "step": 29372 }, { "epoch": 5.39, "learning_rate": 5.0941645309759126e-06, "loss": 0.0101, "step": 29373 }, { "epoch": 5.39, "learning_rate": 5.0935175953078685e-06, "loss": 0.0069, "step": 29374 }, { "epoch": 5.39, "learning_rate": 5.09287068668447e-06, "loss": 0.009, "step": 29375 }, { "epoch": 5.39, "learning_rate": 5.092223805109284e-06, "loss": 0.0085, "step": 29376 }, { "epoch": 5.39, "learning_rate": 5.091576950585875e-06, "loss": 0.0076, "step": 29377 }, { "epoch": 5.39, "learning_rate": 5.0909301231178095e-06, "loss": 0.0097, "step": 29378 }, { "epoch": 5.39, "learning_rate": 5.090283322708648e-06, "loss": 0.0088, "step": 29379 }, { "epoch": 5.39, "learning_rate": 5.089636549361967e-06, "loss": 0.011, "step": 29380 }, { "epoch": 5.39, "learning_rate": 5.088989803081321e-06, "loss": 0.0093, "step": 29381 }, { "epoch": 5.39, "learning_rate": 5.088343083870279e-06, "loss": 0.0113, "step": 29382 }, { "epoch": 5.39, "learning_rate": 5.0876963917323995e-06, "loss": 0.0095, "step": 29383 }, { "epoch": 5.39, "learning_rate": 5.087049726671258e-06, "loss": 0.0082, "step": 29384 }, { "epoch": 5.39, "learning_rate": 5.086403088690413e-06, "loss": 0.0079, "step": 29385 }, { "epoch": 5.39, "learning_rate": 5.085756477793432e-06, "loss": 0.0093, "step": 29386 }, { "epoch": 5.39, "learning_rate": 5.085109893983874e-06, "loss": 0.008, "step": 29387 }, { "epoch": 5.39, "learning_rate": 5.0844633372653005e-06, "loss": 0.0076, "step": 29388 }, { "epoch": 5.39, "learning_rate": 5.0838168076412844e-06, "loss": 0.0107, "step": 29389 }, { "epoch": 5.39, "learning_rate": 5.0831703051153854e-06, "loss": 0.0083, "step": 29390 }, { "epoch": 5.39, "learning_rate": 5.082523829691167e-06, "loss": 0.0098, "step": 29391 }, { "epoch": 5.39, "learning_rate": 5.0818773813721935e-06, "loss": 0.0077, "step": 29392 }, { "epoch": 5.39, "learning_rate": 5.081230960162025e-06, "loss": 0.0105, "step": 29393 }, { "epoch": 5.39, "learning_rate": 5.080584566064228e-06, "loss": 0.0095, "step": 29394 }, { "epoch": 5.39, "learning_rate": 5.079938199082363e-06, "loss": 0.009, "step": 29395 }, { "epoch": 5.39, "learning_rate": 5.079291859219995e-06, "loss": 0.0103, "step": 29396 }, { "epoch": 5.39, "learning_rate": 5.078645546480682e-06, "loss": 0.0067, "step": 29397 }, { "epoch": 5.39, "learning_rate": 5.077999260867996e-06, "loss": 0.0099, "step": 29398 }, { "epoch": 5.39, "learning_rate": 5.077353002385493e-06, "loss": 0.0065, "step": 29399 }, { "epoch": 5.39, "learning_rate": 5.0767067710367326e-06, "loss": 0.0085, "step": 29400 }, { "epoch": 5.39, "learning_rate": 5.076060566825276e-06, "loss": 0.007, "step": 29401 }, { "epoch": 5.39, "learning_rate": 5.075414389754694e-06, "loss": 0.0097, "step": 29402 }, { "epoch": 5.39, "learning_rate": 5.074768239828541e-06, "loss": 0.0079, "step": 29403 }, { "epoch": 5.39, "learning_rate": 5.0741221170503865e-06, "loss": 0.0082, "step": 29404 }, { "epoch": 5.39, "learning_rate": 5.073476021423779e-06, "loss": 0.0081, "step": 29405 }, { "epoch": 5.39, "learning_rate": 5.07282995295229e-06, "loss": 0.0087, "step": 29406 }, { "epoch": 5.39, "learning_rate": 5.072183911639479e-06, "loss": 0.0095, "step": 29407 }, { "epoch": 5.39, "learning_rate": 5.071537897488904e-06, "loss": 0.0101, "step": 29408 }, { "epoch": 5.39, "learning_rate": 5.0708919105041274e-06, "loss": 0.0084, "step": 29409 }, { "epoch": 5.39, "learning_rate": 5.0702459506887125e-06, "loss": 0.0082, "step": 29410 }, { "epoch": 5.39, "learning_rate": 5.069600018046215e-06, "loss": 0.0096, "step": 29411 }, { "epoch": 5.39, "learning_rate": 5.0689541125802e-06, "loss": 0.0083, "step": 29412 }, { "epoch": 5.39, "learning_rate": 5.068308234294224e-06, "loss": 0.0101, "step": 29413 }, { "epoch": 5.39, "learning_rate": 5.067662383191845e-06, "loss": 0.0073, "step": 29414 }, { "epoch": 5.39, "learning_rate": 5.067016559276633e-06, "loss": 0.0088, "step": 29415 }, { "epoch": 5.39, "learning_rate": 5.066370762552138e-06, "loss": 0.0085, "step": 29416 }, { "epoch": 5.39, "learning_rate": 5.065724993021923e-06, "loss": 0.0053, "step": 29417 }, { "epoch": 5.39, "learning_rate": 5.065079250689548e-06, "loss": 0.0077, "step": 29418 }, { "epoch": 5.4, "learning_rate": 5.064433535558569e-06, "loss": 0.0056, "step": 29419 }, { "epoch": 5.4, "learning_rate": 5.063787847632551e-06, "loss": 0.0103, "step": 29420 }, { "epoch": 5.4, "learning_rate": 5.063142186915049e-06, "loss": 0.0087, "step": 29421 }, { "epoch": 5.4, "learning_rate": 5.062496553409627e-06, "loss": 0.0091, "step": 29422 }, { "epoch": 5.4, "learning_rate": 5.061850947119833e-06, "loss": 0.007, "step": 29423 }, { "epoch": 5.4, "learning_rate": 5.061205368049235e-06, "loss": 0.0125, "step": 29424 }, { "epoch": 5.4, "learning_rate": 5.060559816201389e-06, "loss": 0.0082, "step": 29425 }, { "epoch": 5.4, "learning_rate": 5.059914291579852e-06, "loss": 0.01, "step": 29426 }, { "epoch": 5.4, "learning_rate": 5.059268794188185e-06, "loss": 0.0102, "step": 29427 }, { "epoch": 5.4, "learning_rate": 5.058623324029944e-06, "loss": 0.0099, "step": 29428 }, { "epoch": 5.4, "learning_rate": 5.057977881108686e-06, "loss": 0.0089, "step": 29429 }, { "epoch": 5.4, "learning_rate": 5.057332465427971e-06, "loss": 0.0081, "step": 29430 }, { "epoch": 5.4, "learning_rate": 5.056687076991354e-06, "loss": 0.011, "step": 29431 }, { "epoch": 5.4, "learning_rate": 5.056041715802391e-06, "loss": 0.0098, "step": 29432 }, { "epoch": 5.4, "learning_rate": 5.055396381864648e-06, "loss": 0.0084, "step": 29433 }, { "epoch": 5.4, "learning_rate": 5.054751075181674e-06, "loss": 0.0095, "step": 29434 }, { "epoch": 5.4, "learning_rate": 5.054105795757027e-06, "loss": 0.0082, "step": 29435 }, { "epoch": 5.4, "learning_rate": 5.053460543594259e-06, "loss": 0.0078, "step": 29436 }, { "epoch": 5.4, "learning_rate": 5.0528153186969396e-06, "loss": 0.008, "step": 29437 }, { "epoch": 5.4, "learning_rate": 5.052170121068616e-06, "loss": 0.0117, "step": 29438 }, { "epoch": 5.4, "learning_rate": 5.051524950712848e-06, "loss": 0.0131, "step": 29439 }, { "epoch": 5.4, "learning_rate": 5.050879807633192e-06, "loss": 0.01, "step": 29440 }, { "epoch": 5.4, "learning_rate": 5.050234691833196e-06, "loss": 0.0085, "step": 29441 }, { "epoch": 5.4, "learning_rate": 5.049589603316427e-06, "loss": 0.0083, "step": 29442 }, { "epoch": 5.4, "learning_rate": 5.0489445420864355e-06, "loss": 0.0077, "step": 29443 }, { "epoch": 5.4, "learning_rate": 5.048299508146776e-06, "loss": 0.0106, "step": 29444 }, { "epoch": 5.4, "learning_rate": 5.047654501501008e-06, "loss": 0.0088, "step": 29445 }, { "epoch": 5.4, "learning_rate": 5.047009522152684e-06, "loss": 0.0092, "step": 29446 }, { "epoch": 5.4, "learning_rate": 5.046364570105359e-06, "loss": 0.0085, "step": 29447 }, { "epoch": 5.4, "learning_rate": 5.0457196453625875e-06, "loss": 0.0071, "step": 29448 }, { "epoch": 5.4, "learning_rate": 5.045074747927927e-06, "loss": 0.0075, "step": 29449 }, { "epoch": 5.4, "learning_rate": 5.044429877804926e-06, "loss": 0.0074, "step": 29450 }, { "epoch": 5.4, "learning_rate": 5.043785034997153e-06, "loss": 0.0085, "step": 29451 }, { "epoch": 5.4, "learning_rate": 5.043140219508147e-06, "loss": 0.0126, "step": 29452 }, { "epoch": 5.4, "learning_rate": 5.042495431341469e-06, "loss": 0.0076, "step": 29453 }, { "epoch": 5.4, "learning_rate": 5.0418506705006676e-06, "loss": 0.0097, "step": 29454 }, { "epoch": 5.4, "learning_rate": 5.041205936989306e-06, "loss": 0.0084, "step": 29455 }, { "epoch": 5.4, "learning_rate": 5.040561230810934e-06, "loss": 0.0096, "step": 29456 }, { "epoch": 5.4, "learning_rate": 5.039916551969104e-06, "loss": 0.0094, "step": 29457 }, { "epoch": 5.4, "learning_rate": 5.039271900467371e-06, "loss": 0.007, "step": 29458 }, { "epoch": 5.4, "learning_rate": 5.038627276309287e-06, "loss": 0.0097, "step": 29459 }, { "epoch": 5.4, "learning_rate": 5.0379826794984055e-06, "loss": 0.0092, "step": 29460 }, { "epoch": 5.4, "learning_rate": 5.03733811003828e-06, "loss": 0.0104, "step": 29461 }, { "epoch": 5.4, "learning_rate": 5.036693567932463e-06, "loss": 0.0084, "step": 29462 }, { "epoch": 5.4, "learning_rate": 5.036049053184506e-06, "loss": 0.0105, "step": 29463 }, { "epoch": 5.4, "learning_rate": 5.035404565797966e-06, "loss": 0.0082, "step": 29464 }, { "epoch": 5.4, "learning_rate": 5.03476010577639e-06, "loss": 0.0078, "step": 29465 }, { "epoch": 5.4, "learning_rate": 5.0341156731233324e-06, "loss": 0.0061, "step": 29466 }, { "epoch": 5.4, "learning_rate": 5.033471267842342e-06, "loss": 0.0077, "step": 29467 }, { "epoch": 5.4, "learning_rate": 5.032826889936979e-06, "loss": 0.0115, "step": 29468 }, { "epoch": 5.4, "learning_rate": 5.032182539410795e-06, "loss": 0.0082, "step": 29469 }, { "epoch": 5.4, "learning_rate": 5.031538216267332e-06, "loss": 0.0094, "step": 29470 }, { "epoch": 5.4, "learning_rate": 5.030893920510147e-06, "loss": 0.0081, "step": 29471 }, { "epoch": 5.4, "learning_rate": 5.030249652142787e-06, "loss": 0.0087, "step": 29472 }, { "epoch": 5.4, "learning_rate": 5.029605411168812e-06, "loss": 0.0081, "step": 29473 }, { "epoch": 5.41, "learning_rate": 5.028961197591769e-06, "loss": 0.0091, "step": 29474 }, { "epoch": 5.41, "learning_rate": 5.02831701141521e-06, "loss": 0.0085, "step": 29475 }, { "epoch": 5.41, "learning_rate": 5.027672852642677e-06, "loss": 0.0093, "step": 29476 }, { "epoch": 5.41, "learning_rate": 5.0270287212777334e-06, "loss": 0.0083, "step": 29477 }, { "epoch": 5.41, "learning_rate": 5.026384617323923e-06, "loss": 0.0114, "step": 29478 }, { "epoch": 5.41, "learning_rate": 5.025740540784796e-06, "loss": 0.0092, "step": 29479 }, { "epoch": 5.41, "learning_rate": 5.025096491663906e-06, "loss": 0.0101, "step": 29480 }, { "epoch": 5.41, "learning_rate": 5.024452469964798e-06, "loss": 0.0092, "step": 29481 }, { "epoch": 5.41, "learning_rate": 5.023808475691025e-06, "loss": 0.0111, "step": 29482 }, { "epoch": 5.41, "learning_rate": 5.023164508846138e-06, "loss": 0.009, "step": 29483 }, { "epoch": 5.41, "learning_rate": 5.022520569433684e-06, "loss": 0.0084, "step": 29484 }, { "epoch": 5.41, "learning_rate": 5.021876657457209e-06, "loss": 0.0069, "step": 29485 }, { "epoch": 5.41, "learning_rate": 5.02123277292027e-06, "loss": 0.0071, "step": 29486 }, { "epoch": 5.41, "learning_rate": 5.020588915826417e-06, "loss": 0.009, "step": 29487 }, { "epoch": 5.41, "learning_rate": 5.019945086179192e-06, "loss": 0.0094, "step": 29488 }, { "epoch": 5.41, "learning_rate": 5.019301283982141e-06, "loss": 0.0061, "step": 29489 }, { "epoch": 5.41, "learning_rate": 5.0186575092388225e-06, "loss": 0.0099, "step": 29490 }, { "epoch": 5.41, "learning_rate": 5.01801376195278e-06, "loss": 0.0077, "step": 29491 }, { "epoch": 5.41, "learning_rate": 5.0173700421275614e-06, "loss": 0.01, "step": 29492 }, { "epoch": 5.41, "learning_rate": 5.016726349766721e-06, "loss": 0.0076, "step": 29493 }, { "epoch": 5.41, "learning_rate": 5.0160826848737945e-06, "loss": 0.0059, "step": 29494 }, { "epoch": 5.41, "learning_rate": 5.0154390474523395e-06, "loss": 0.0093, "step": 29495 }, { "epoch": 5.41, "learning_rate": 5.0147954375059015e-06, "loss": 0.0106, "step": 29496 }, { "epoch": 5.41, "learning_rate": 5.014151855038029e-06, "loss": 0.0084, "step": 29497 }, { "epoch": 5.41, "learning_rate": 5.013508300052267e-06, "loss": 0.0067, "step": 29498 }, { "epoch": 5.41, "learning_rate": 5.012864772552165e-06, "loss": 0.0104, "step": 29499 }, { "epoch": 5.41, "learning_rate": 5.012221272541269e-06, "loss": 0.0117, "step": 29500 }, { "epoch": 5.41, "learning_rate": 5.011577800023126e-06, "loss": 0.011, "step": 29501 }, { "epoch": 5.41, "learning_rate": 5.010934355001282e-06, "loss": 0.0075, "step": 29502 }, { "epoch": 5.41, "learning_rate": 5.010290937479283e-06, "loss": 0.0089, "step": 29503 }, { "epoch": 5.41, "learning_rate": 5.00964754746068e-06, "loss": 0.0095, "step": 29504 }, { "epoch": 5.41, "learning_rate": 5.00900418494902e-06, "loss": 0.0088, "step": 29505 }, { "epoch": 5.41, "learning_rate": 5.008360849947842e-06, "loss": 0.0068, "step": 29506 }, { "epoch": 5.41, "learning_rate": 5.0077175424606926e-06, "loss": 0.0125, "step": 29507 }, { "epoch": 5.41, "learning_rate": 5.007074262491125e-06, "loss": 0.0066, "step": 29508 }, { "epoch": 5.41, "learning_rate": 5.006431010042682e-06, "loss": 0.0085, "step": 29509 }, { "epoch": 5.41, "learning_rate": 5.005787785118906e-06, "loss": 0.0067, "step": 29510 }, { "epoch": 5.41, "learning_rate": 5.0051445877233465e-06, "loss": 0.0086, "step": 29511 }, { "epoch": 5.41, "learning_rate": 5.004501417859547e-06, "loss": 0.0077, "step": 29512 }, { "epoch": 5.41, "learning_rate": 5.0038582755310516e-06, "loss": 0.0076, "step": 29513 }, { "epoch": 5.41, "learning_rate": 5.003215160741407e-06, "loss": 0.0091, "step": 29514 }, { "epoch": 5.41, "learning_rate": 5.002572073494157e-06, "loss": 0.0078, "step": 29515 }, { "epoch": 5.41, "learning_rate": 5.001929013792848e-06, "loss": 0.0093, "step": 29516 }, { "epoch": 5.41, "learning_rate": 5.0012859816410216e-06, "loss": 0.0083, "step": 29517 }, { "epoch": 5.41, "learning_rate": 5.000642977042226e-06, "loss": 0.008, "step": 29518 }, { "epoch": 5.41, "learning_rate": 5.000000000000003e-06, "loss": 0.0084, "step": 29519 }, { "epoch": 5.41, "learning_rate": 4.999357050517893e-06, "loss": 0.0096, "step": 29520 }, { "epoch": 5.41, "learning_rate": 4.998714128599447e-06, "loss": 0.0082, "step": 29521 }, { "epoch": 5.41, "learning_rate": 4.99807123424821e-06, "loss": 0.0073, "step": 29522 }, { "epoch": 5.41, "learning_rate": 4.997428367467718e-06, "loss": 0.01, "step": 29523 }, { "epoch": 5.41, "learning_rate": 4.996785528261518e-06, "loss": 0.0088, "step": 29524 }, { "epoch": 5.41, "learning_rate": 4.99614271663315e-06, "loss": 0.0097, "step": 29525 }, { "epoch": 5.41, "learning_rate": 4.995499932586164e-06, "loss": 0.0067, "step": 29526 }, { "epoch": 5.41, "learning_rate": 4.9948571761241e-06, "loss": 0.0102, "step": 29527 }, { "epoch": 5.42, "learning_rate": 4.994214447250501e-06, "loss": 0.0097, "step": 29528 }, { "epoch": 5.42, "learning_rate": 4.9935717459689085e-06, "loss": 0.0089, "step": 29529 }, { "epoch": 5.42, "learning_rate": 4.992929072282865e-06, "loss": 0.0117, "step": 29530 }, { "epoch": 5.42, "learning_rate": 4.992286426195915e-06, "loss": 0.0129, "step": 29531 }, { "epoch": 5.42, "learning_rate": 4.991643807711599e-06, "loss": 0.0077, "step": 29532 }, { "epoch": 5.42, "learning_rate": 4.991001216833456e-06, "loss": 0.0085, "step": 29533 }, { "epoch": 5.42, "learning_rate": 4.990358653565039e-06, "loss": 0.013, "step": 29534 }, { "epoch": 5.42, "learning_rate": 4.98971611790988e-06, "loss": 0.0092, "step": 29535 }, { "epoch": 5.42, "learning_rate": 4.989073609871521e-06, "loss": 0.0092, "step": 29536 }, { "epoch": 5.42, "learning_rate": 4.988431129453507e-06, "loss": 0.008, "step": 29537 }, { "epoch": 5.42, "learning_rate": 4.987788676659374e-06, "loss": 0.0108, "step": 29538 }, { "epoch": 5.42, "learning_rate": 4.98714625149267e-06, "loss": 0.0075, "step": 29539 }, { "epoch": 5.42, "learning_rate": 4.986503853956939e-06, "loss": 0.0077, "step": 29540 }, { "epoch": 5.42, "learning_rate": 4.985861484055711e-06, "loss": 0.007, "step": 29541 }, { "epoch": 5.42, "learning_rate": 4.985219141792529e-06, "loss": 0.0085, "step": 29542 }, { "epoch": 5.42, "learning_rate": 4.984576827170939e-06, "loss": 0.0116, "step": 29543 }, { "epoch": 5.42, "learning_rate": 4.983934540194481e-06, "loss": 0.0125, "step": 29544 }, { "epoch": 5.42, "learning_rate": 4.983292280866692e-06, "loss": 0.0084, "step": 29545 }, { "epoch": 5.42, "learning_rate": 4.982650049191114e-06, "loss": 0.0088, "step": 29546 }, { "epoch": 5.42, "learning_rate": 4.9820078451712865e-06, "loss": 0.008, "step": 29547 }, { "epoch": 5.42, "learning_rate": 4.981365668810749e-06, "loss": 0.0115, "step": 29548 }, { "epoch": 5.42, "learning_rate": 4.980723520113041e-06, "loss": 0.0072, "step": 29549 }, { "epoch": 5.42, "learning_rate": 4.980081399081703e-06, "loss": 0.0096, "step": 29550 }, { "epoch": 5.42, "learning_rate": 4.979439305720271e-06, "loss": 0.0088, "step": 29551 }, { "epoch": 5.42, "learning_rate": 4.978797240032294e-06, "loss": 0.0091, "step": 29552 }, { "epoch": 5.42, "learning_rate": 4.978155202021301e-06, "loss": 0.0081, "step": 29553 }, { "epoch": 5.42, "learning_rate": 4.977513191690834e-06, "loss": 0.0108, "step": 29554 }, { "epoch": 5.42, "learning_rate": 4.97687120904443e-06, "loss": 0.0117, "step": 29555 }, { "epoch": 5.42, "learning_rate": 4.976229254085629e-06, "loss": 0.0049, "step": 29556 }, { "epoch": 5.42, "learning_rate": 4.9755873268179715e-06, "loss": 0.0094, "step": 29557 }, { "epoch": 5.42, "learning_rate": 4.974945427244999e-06, "loss": 0.009, "step": 29558 }, { "epoch": 5.42, "learning_rate": 4.974303555370241e-06, "loss": 0.0065, "step": 29559 }, { "epoch": 5.42, "learning_rate": 4.973661711197236e-06, "loss": 0.0102, "step": 29560 }, { "epoch": 5.42, "learning_rate": 4.973019894729529e-06, "loss": 0.0109, "step": 29561 }, { "epoch": 5.42, "learning_rate": 4.9723781059706546e-06, "loss": 0.0085, "step": 29562 }, { "epoch": 5.42, "learning_rate": 4.971736344924149e-06, "loss": 0.0078, "step": 29563 }, { "epoch": 5.42, "learning_rate": 4.971094611593551e-06, "loss": 0.008, "step": 29564 }, { "epoch": 5.42, "learning_rate": 4.970452905982397e-06, "loss": 0.0066, "step": 29565 }, { "epoch": 5.42, "learning_rate": 4.969811228094225e-06, "loss": 0.0093, "step": 29566 }, { "epoch": 5.42, "learning_rate": 4.969169577932571e-06, "loss": 0.0107, "step": 29567 }, { "epoch": 5.42, "learning_rate": 4.968527955500972e-06, "loss": 0.0097, "step": 29568 }, { "epoch": 5.42, "learning_rate": 4.967886360802965e-06, "loss": 0.0077, "step": 29569 }, { "epoch": 5.42, "learning_rate": 4.967244793842086e-06, "loss": 0.0089, "step": 29570 }, { "epoch": 5.42, "learning_rate": 4.966603254621872e-06, "loss": 0.007, "step": 29571 }, { "epoch": 5.42, "learning_rate": 4.9659617431458584e-06, "loss": 0.0085, "step": 29572 }, { "epoch": 5.42, "learning_rate": 4.965320259417578e-06, "loss": 0.009, "step": 29573 }, { "epoch": 5.42, "learning_rate": 4.964678803440574e-06, "loss": 0.0086, "step": 29574 }, { "epoch": 5.42, "learning_rate": 4.964037375218379e-06, "loss": 0.0102, "step": 29575 }, { "epoch": 5.42, "learning_rate": 4.963395974754532e-06, "loss": 0.0108, "step": 29576 }, { "epoch": 5.42, "learning_rate": 4.96275460205256e-06, "loss": 0.0082, "step": 29577 }, { "epoch": 5.42, "learning_rate": 4.962113257115998e-06, "loss": 0.0084, "step": 29578 }, { "epoch": 5.42, "learning_rate": 4.96147193994839e-06, "loss": 0.01, "step": 29579 }, { "epoch": 5.42, "learning_rate": 4.960830650553267e-06, "loss": 0.0115, "step": 29580 }, { "epoch": 5.42, "learning_rate": 4.960189388934163e-06, "loss": 0.0095, "step": 29581 }, { "epoch": 5.42, "learning_rate": 4.959548155094614e-06, "loss": 0.0097, "step": 29582 }, { "epoch": 5.43, "learning_rate": 4.9589069490381535e-06, "loss": 0.0086, "step": 29583 }, { "epoch": 5.43, "learning_rate": 4.958265770768315e-06, "loss": 0.0082, "step": 29584 }, { "epoch": 5.43, "learning_rate": 4.957624620288635e-06, "loss": 0.0105, "step": 29585 }, { "epoch": 5.43, "learning_rate": 4.956983497602642e-06, "loss": 0.0094, "step": 29586 }, { "epoch": 5.43, "learning_rate": 4.956342402713882e-06, "loss": 0.0075, "step": 29587 }, { "epoch": 5.43, "learning_rate": 4.955701335625876e-06, "loss": 0.0074, "step": 29588 }, { "epoch": 5.43, "learning_rate": 4.955060296342163e-06, "loss": 0.0097, "step": 29589 }, { "epoch": 5.43, "learning_rate": 4.954419284866276e-06, "loss": 0.009, "step": 29590 }, { "epoch": 5.43, "learning_rate": 4.953778301201744e-06, "loss": 0.0082, "step": 29591 }, { "epoch": 5.43, "learning_rate": 4.953137345352108e-06, "loss": 0.0078, "step": 29592 }, { "epoch": 5.43, "learning_rate": 4.952496417320898e-06, "loss": 0.0085, "step": 29593 }, { "epoch": 5.43, "learning_rate": 4.9518555171116485e-06, "loss": 0.0107, "step": 29594 }, { "epoch": 5.43, "learning_rate": 4.951214644727883e-06, "loss": 0.0103, "step": 29595 }, { "epoch": 5.43, "learning_rate": 4.950573800173144e-06, "loss": 0.0084, "step": 29596 }, { "epoch": 5.43, "learning_rate": 4.949932983450961e-06, "loss": 0.0111, "step": 29597 }, { "epoch": 5.43, "learning_rate": 4.9492921945648656e-06, "loss": 0.0103, "step": 29598 }, { "epoch": 5.43, "learning_rate": 4.94865143351839e-06, "loss": 0.0071, "step": 29599 }, { "epoch": 5.43, "learning_rate": 4.948010700315066e-06, "loss": 0.0084, "step": 29600 }, { "epoch": 5.43, "learning_rate": 4.947369994958424e-06, "loss": 0.0072, "step": 29601 }, { "epoch": 5.43, "learning_rate": 4.946729317452e-06, "loss": 0.0095, "step": 29602 }, { "epoch": 5.43, "learning_rate": 4.946088667799319e-06, "loss": 0.0091, "step": 29603 }, { "epoch": 5.43, "learning_rate": 4.945448046003913e-06, "loss": 0.0083, "step": 29604 }, { "epoch": 5.43, "learning_rate": 4.9448074520693246e-06, "loss": 0.0099, "step": 29605 }, { "epoch": 5.43, "learning_rate": 4.944166885999071e-06, "loss": 0.0091, "step": 29606 }, { "epoch": 5.43, "learning_rate": 4.943526347796689e-06, "loss": 0.0064, "step": 29607 }, { "epoch": 5.43, "learning_rate": 4.942885837465707e-06, "loss": 0.0073, "step": 29608 }, { "epoch": 5.43, "learning_rate": 4.942245355009654e-06, "loss": 0.0101, "step": 29609 }, { "epoch": 5.43, "learning_rate": 4.941604900432065e-06, "loss": 0.0102, "step": 29610 }, { "epoch": 5.43, "learning_rate": 4.9409644737364706e-06, "loss": 0.0075, "step": 29611 }, { "epoch": 5.43, "learning_rate": 4.9403240749264014e-06, "loss": 0.0086, "step": 29612 }, { "epoch": 5.43, "learning_rate": 4.939683704005377e-06, "loss": 0.0103, "step": 29613 }, { "epoch": 5.43, "learning_rate": 4.939043360976938e-06, "loss": 0.0115, "step": 29614 }, { "epoch": 5.43, "learning_rate": 4.938403045844611e-06, "loss": 0.0113, "step": 29615 }, { "epoch": 5.43, "learning_rate": 4.937762758611925e-06, "loss": 0.0109, "step": 29616 }, { "epoch": 5.43, "learning_rate": 4.9371224992824085e-06, "loss": 0.0068, "step": 29617 }, { "epoch": 5.43, "learning_rate": 4.936482267859592e-06, "loss": 0.0067, "step": 29618 }, { "epoch": 5.43, "learning_rate": 4.935842064347004e-06, "loss": 0.0074, "step": 29619 }, { "epoch": 5.43, "learning_rate": 4.935201888748172e-06, "loss": 0.0097, "step": 29620 }, { "epoch": 5.43, "learning_rate": 4.934561741066628e-06, "loss": 0.0094, "step": 29621 }, { "epoch": 5.43, "learning_rate": 4.933921621305893e-06, "loss": 0.0096, "step": 29622 }, { "epoch": 5.43, "learning_rate": 4.933281529469509e-06, "loss": 0.0085, "step": 29623 }, { "epoch": 5.43, "learning_rate": 4.932641465560991e-06, "loss": 0.0084, "step": 29624 }, { "epoch": 5.43, "learning_rate": 4.932001429583874e-06, "loss": 0.0079, "step": 29625 }, { "epoch": 5.43, "learning_rate": 4.931361421541678e-06, "loss": 0.0102, "step": 29626 }, { "epoch": 5.43, "learning_rate": 4.930721441437941e-06, "loss": 0.0114, "step": 29627 }, { "epoch": 5.43, "learning_rate": 4.930081489276186e-06, "loss": 0.0069, "step": 29628 }, { "epoch": 5.43, "learning_rate": 4.929441565059942e-06, "loss": 0.0087, "step": 29629 }, { "epoch": 5.43, "learning_rate": 4.928801668792738e-06, "loss": 0.0099, "step": 29630 }, { "epoch": 5.43, "learning_rate": 4.92816180047809e-06, "loss": 0.0084, "step": 29631 }, { "epoch": 5.43, "learning_rate": 4.927521960119536e-06, "loss": 0.0059, "step": 29632 }, { "epoch": 5.43, "learning_rate": 4.9268821477206005e-06, "loss": 0.01, "step": 29633 }, { "epoch": 5.43, "learning_rate": 4.92624236328481e-06, "loss": 0.0117, "step": 29634 }, { "epoch": 5.43, "learning_rate": 4.92560260681569e-06, "loss": 0.0103, "step": 29635 }, { "epoch": 5.43, "learning_rate": 4.924962878316767e-06, "loss": 0.007, "step": 29636 }, { "epoch": 5.43, "learning_rate": 4.924323177791568e-06, "loss": 0.0092, "step": 29637 }, { "epoch": 5.44, "learning_rate": 4.923683505243618e-06, "loss": 0.0099, "step": 29638 }, { "epoch": 5.44, "learning_rate": 4.923043860676441e-06, "loss": 0.0099, "step": 29639 }, { "epoch": 5.44, "learning_rate": 4.922404244093567e-06, "loss": 0.0084, "step": 29640 }, { "epoch": 5.44, "learning_rate": 4.921764655498526e-06, "loss": 0.0077, "step": 29641 }, { "epoch": 5.44, "learning_rate": 4.921125094894831e-06, "loss": 0.0066, "step": 29642 }, { "epoch": 5.44, "learning_rate": 4.920485562286015e-06, "loss": 0.0077, "step": 29643 }, { "epoch": 5.44, "learning_rate": 4.919846057675597e-06, "loss": 0.0072, "step": 29644 }, { "epoch": 5.44, "learning_rate": 4.919206581067111e-06, "loss": 0.0081, "step": 29645 }, { "epoch": 5.44, "learning_rate": 4.918567132464076e-06, "loss": 0.0071, "step": 29646 }, { "epoch": 5.44, "learning_rate": 4.917927711870023e-06, "loss": 0.0093, "step": 29647 }, { "epoch": 5.44, "learning_rate": 4.917288319288464e-06, "loss": 0.0096, "step": 29648 }, { "epoch": 5.44, "learning_rate": 4.916648954722935e-06, "loss": 0.0091, "step": 29649 }, { "epoch": 5.44, "learning_rate": 4.9160096181769545e-06, "loss": 0.0093, "step": 29650 }, { "epoch": 5.44, "learning_rate": 4.9153703096540485e-06, "loss": 0.0078, "step": 29651 }, { "epoch": 5.44, "learning_rate": 4.914731029157741e-06, "loss": 0.007, "step": 29652 }, { "epoch": 5.44, "learning_rate": 4.914091776691555e-06, "loss": 0.0077, "step": 29653 }, { "epoch": 5.44, "learning_rate": 4.9134525522590145e-06, "loss": 0.0086, "step": 29654 }, { "epoch": 5.44, "learning_rate": 4.9128133558636425e-06, "loss": 0.0121, "step": 29655 }, { "epoch": 5.44, "learning_rate": 4.912174187508963e-06, "loss": 0.0099, "step": 29656 }, { "epoch": 5.44, "learning_rate": 4.911535047198495e-06, "loss": 0.0106, "step": 29657 }, { "epoch": 5.44, "learning_rate": 4.910895934935768e-06, "loss": 0.0104, "step": 29658 }, { "epoch": 5.44, "learning_rate": 4.910256850724306e-06, "loss": 0.0073, "step": 29659 }, { "epoch": 5.44, "learning_rate": 4.9096177945676225e-06, "loss": 0.0078, "step": 29660 }, { "epoch": 5.44, "learning_rate": 4.908978766469242e-06, "loss": 0.009, "step": 29661 }, { "epoch": 5.44, "learning_rate": 4.908339766432695e-06, "loss": 0.0062, "step": 29662 }, { "epoch": 5.44, "learning_rate": 4.907700794461496e-06, "loss": 0.0117, "step": 29663 }, { "epoch": 5.44, "learning_rate": 4.907061850559172e-06, "loss": 0.0098, "step": 29664 }, { "epoch": 5.44, "learning_rate": 4.906422934729245e-06, "loss": 0.009, "step": 29665 }, { "epoch": 5.44, "learning_rate": 4.905784046975226e-06, "loss": 0.008, "step": 29666 }, { "epoch": 5.44, "learning_rate": 4.905145187300648e-06, "loss": 0.0108, "step": 29667 }, { "epoch": 5.44, "learning_rate": 4.9045063557090315e-06, "loss": 0.0096, "step": 29668 }, { "epoch": 5.44, "learning_rate": 4.903867552203893e-06, "loss": 0.0077, "step": 29669 }, { "epoch": 5.44, "learning_rate": 4.903228776788758e-06, "loss": 0.0068, "step": 29670 }, { "epoch": 5.44, "learning_rate": 4.902590029467143e-06, "loss": 0.0095, "step": 29671 }, { "epoch": 5.44, "learning_rate": 4.901951310242573e-06, "loss": 0.0067, "step": 29672 }, { "epoch": 5.44, "learning_rate": 4.901312619118565e-06, "loss": 0.0068, "step": 29673 }, { "epoch": 5.44, "learning_rate": 4.900673956098644e-06, "loss": 0.007, "step": 29674 }, { "epoch": 5.44, "learning_rate": 4.900035321186322e-06, "loss": 0.0069, "step": 29675 }, { "epoch": 5.44, "learning_rate": 4.899396714385133e-06, "loss": 0.0123, "step": 29676 }, { "epoch": 5.44, "learning_rate": 4.898758135698584e-06, "loss": 0.0094, "step": 29677 }, { "epoch": 5.44, "learning_rate": 4.898119585130201e-06, "loss": 0.0089, "step": 29678 }, { "epoch": 5.44, "learning_rate": 4.897481062683498e-06, "loss": 0.0076, "step": 29679 }, { "epoch": 5.44, "learning_rate": 4.896842568362002e-06, "loss": 0.0097, "step": 29680 }, { "epoch": 5.44, "learning_rate": 4.896204102169231e-06, "loss": 0.0132, "step": 29681 }, { "epoch": 5.44, "learning_rate": 4.895565664108702e-06, "loss": 0.0091, "step": 29682 }, { "epoch": 5.44, "learning_rate": 4.894927254183937e-06, "loss": 0.0087, "step": 29683 }, { "epoch": 5.44, "learning_rate": 4.894288872398446e-06, "loss": 0.0089, "step": 29684 }, { "epoch": 5.44, "learning_rate": 4.893650518755758e-06, "loss": 0.0081, "step": 29685 }, { "epoch": 5.44, "learning_rate": 4.893012193259388e-06, "loss": 0.0088, "step": 29686 }, { "epoch": 5.44, "learning_rate": 4.892373895912854e-06, "loss": 0.006, "step": 29687 }, { "epoch": 5.44, "learning_rate": 4.891735626719676e-06, "loss": 0.0079, "step": 29688 }, { "epoch": 5.44, "learning_rate": 4.89109738568337e-06, "loss": 0.0072, "step": 29689 }, { "epoch": 5.44, "learning_rate": 4.890459172807455e-06, "loss": 0.0065, "step": 29690 }, { "epoch": 5.44, "learning_rate": 4.889820988095448e-06, "loss": 0.0077, "step": 29691 }, { "epoch": 5.45, "learning_rate": 4.889182831550865e-06, "loss": 0.0107, "step": 29692 }, { "epoch": 5.45, "learning_rate": 4.88854470317723e-06, "loss": 0.0097, "step": 29693 }, { "epoch": 5.45, "learning_rate": 4.887906602978059e-06, "loss": 0.0093, "step": 29694 }, { "epoch": 5.45, "learning_rate": 4.887268530956863e-06, "loss": 0.0066, "step": 29695 }, { "epoch": 5.45, "learning_rate": 4.886630487117163e-06, "loss": 0.0117, "step": 29696 }, { "epoch": 5.45, "learning_rate": 4.885992471462472e-06, "loss": 0.0089, "step": 29697 }, { "epoch": 5.45, "learning_rate": 4.885354483996313e-06, "loss": 0.0082, "step": 29698 }, { "epoch": 5.45, "learning_rate": 4.884716524722201e-06, "loss": 0.0081, "step": 29699 }, { "epoch": 5.45, "learning_rate": 4.884078593643651e-06, "loss": 0.0075, "step": 29700 }, { "epoch": 5.45, "learning_rate": 4.8834406907641784e-06, "loss": 0.0084, "step": 29701 }, { "epoch": 5.45, "learning_rate": 4.882802816087302e-06, "loss": 0.0073, "step": 29702 }, { "epoch": 5.45, "learning_rate": 4.8821649696165366e-06, "loss": 0.0094, "step": 29703 }, { "epoch": 5.45, "learning_rate": 4.881527151355396e-06, "loss": 0.0103, "step": 29704 }, { "epoch": 5.45, "learning_rate": 4.880889361307399e-06, "loss": 0.008, "step": 29705 }, { "epoch": 5.45, "learning_rate": 4.8802515994760604e-06, "loss": 0.0085, "step": 29706 }, { "epoch": 5.45, "learning_rate": 4.879613865864894e-06, "loss": 0.0068, "step": 29707 }, { "epoch": 5.45, "learning_rate": 4.878976160477417e-06, "loss": 0.0083, "step": 29708 }, { "epoch": 5.45, "learning_rate": 4.878338483317142e-06, "loss": 0.0076, "step": 29709 }, { "epoch": 5.45, "learning_rate": 4.877700834387582e-06, "loss": 0.009, "step": 29710 }, { "epoch": 5.45, "learning_rate": 4.877063213692259e-06, "loss": 0.0084, "step": 29711 }, { "epoch": 5.45, "learning_rate": 4.8764256212346875e-06, "loss": 0.0074, "step": 29712 }, { "epoch": 5.45, "learning_rate": 4.875788057018373e-06, "loss": 0.0067, "step": 29713 }, { "epoch": 5.45, "learning_rate": 4.875150521046832e-06, "loss": 0.0108, "step": 29714 }, { "epoch": 5.45, "learning_rate": 4.874513013323585e-06, "loss": 0.0108, "step": 29715 }, { "epoch": 5.45, "learning_rate": 4.873875533852142e-06, "loss": 0.0087, "step": 29716 }, { "epoch": 5.45, "learning_rate": 4.8732380826360185e-06, "loss": 0.0077, "step": 29717 }, { "epoch": 5.45, "learning_rate": 4.8726006596787255e-06, "loss": 0.0083, "step": 29718 }, { "epoch": 5.45, "learning_rate": 4.871963264983778e-06, "loss": 0.0084, "step": 29719 }, { "epoch": 5.45, "learning_rate": 4.87132589855469e-06, "loss": 0.0094, "step": 29720 }, { "epoch": 5.45, "learning_rate": 4.870688560394973e-06, "loss": 0.0093, "step": 29721 }, { "epoch": 5.45, "learning_rate": 4.8700512505081395e-06, "loss": 0.0117, "step": 29722 }, { "epoch": 5.45, "learning_rate": 4.869413968897706e-06, "loss": 0.0072, "step": 29723 }, { "epoch": 5.45, "learning_rate": 4.868776715567182e-06, "loss": 0.0098, "step": 29724 }, { "epoch": 5.45, "learning_rate": 4.868139490520081e-06, "loss": 0.0089, "step": 29725 }, { "epoch": 5.45, "learning_rate": 4.867502293759915e-06, "loss": 0.0087, "step": 29726 }, { "epoch": 5.45, "learning_rate": 4.866865125290196e-06, "loss": 0.0115, "step": 29727 }, { "epoch": 5.45, "learning_rate": 4.8662279851144344e-06, "loss": 0.0079, "step": 29728 }, { "epoch": 5.45, "learning_rate": 4.8655908732361476e-06, "loss": 0.0077, "step": 29729 }, { "epoch": 5.45, "learning_rate": 4.8649537896588485e-06, "loss": 0.0077, "step": 29730 }, { "epoch": 5.45, "learning_rate": 4.86431673438604e-06, "loss": 0.0084, "step": 29731 }, { "epoch": 5.45, "learning_rate": 4.863679707421234e-06, "loss": 0.006, "step": 29732 }, { "epoch": 5.45, "learning_rate": 4.86304270876795e-06, "loss": 0.0121, "step": 29733 }, { "epoch": 5.45, "learning_rate": 4.862405738429694e-06, "loss": 0.0102, "step": 29734 }, { "epoch": 5.45, "learning_rate": 4.861768796409979e-06, "loss": 0.009, "step": 29735 }, { "epoch": 5.45, "learning_rate": 4.861131882712314e-06, "loss": 0.0077, "step": 29736 }, { "epoch": 5.45, "learning_rate": 4.8604949973402115e-06, "loss": 0.0075, "step": 29737 }, { "epoch": 5.45, "learning_rate": 4.85985814029718e-06, "loss": 0.0089, "step": 29738 }, { "epoch": 5.45, "learning_rate": 4.859221311586731e-06, "loss": 0.0098, "step": 29739 }, { "epoch": 5.45, "learning_rate": 4.8585845112123755e-06, "loss": 0.0092, "step": 29740 }, { "epoch": 5.45, "learning_rate": 4.857947739177622e-06, "loss": 0.0105, "step": 29741 }, { "epoch": 5.45, "learning_rate": 4.857310995485981e-06, "loss": 0.0087, "step": 29742 }, { "epoch": 5.45, "learning_rate": 4.856674280140962e-06, "loss": 0.0085, "step": 29743 }, { "epoch": 5.45, "learning_rate": 4.856037593146074e-06, "loss": 0.0079, "step": 29744 }, { "epoch": 5.45, "learning_rate": 4.855400934504826e-06, "loss": 0.0064, "step": 29745 }, { "epoch": 5.45, "learning_rate": 4.854764304220731e-06, "loss": 0.0103, "step": 29746 }, { "epoch": 5.46, "learning_rate": 4.854127702297296e-06, "loss": 0.0078, "step": 29747 }, { "epoch": 5.46, "learning_rate": 4.853491128738033e-06, "loss": 0.009, "step": 29748 }, { "epoch": 5.46, "learning_rate": 4.852854583546444e-06, "loss": 0.0078, "step": 29749 }, { "epoch": 5.46, "learning_rate": 4.852218066726037e-06, "loss": 0.0117, "step": 29750 }, { "epoch": 5.46, "learning_rate": 4.851581578280327e-06, "loss": 0.0093, "step": 29751 }, { "epoch": 5.46, "learning_rate": 4.850945118212822e-06, "loss": 0.0117, "step": 29752 }, { "epoch": 5.46, "learning_rate": 4.850308686527028e-06, "loss": 0.008, "step": 29753 }, { "epoch": 5.46, "learning_rate": 4.849672283226453e-06, "loss": 0.0096, "step": 29754 }, { "epoch": 5.46, "learning_rate": 4.849035908314604e-06, "loss": 0.0112, "step": 29755 }, { "epoch": 5.46, "learning_rate": 4.848399561794991e-06, "loss": 0.0127, "step": 29756 }, { "epoch": 5.46, "learning_rate": 4.84776324367112e-06, "loss": 0.0064, "step": 29757 }, { "epoch": 5.46, "learning_rate": 4.847126953946498e-06, "loss": 0.01, "step": 29758 }, { "epoch": 5.46, "learning_rate": 4.846490692624635e-06, "loss": 0.0095, "step": 29759 }, { "epoch": 5.46, "learning_rate": 4.845854459709034e-06, "loss": 0.0082, "step": 29760 }, { "epoch": 5.46, "learning_rate": 4.8452182552032045e-06, "loss": 0.0084, "step": 29761 }, { "epoch": 5.46, "learning_rate": 4.844582079110652e-06, "loss": 0.0075, "step": 29762 }, { "epoch": 5.46, "learning_rate": 4.843945931434881e-06, "loss": 0.0077, "step": 29763 }, { "epoch": 5.46, "learning_rate": 4.843309812179405e-06, "loss": 0.0092, "step": 29764 }, { "epoch": 5.46, "learning_rate": 4.842673721347726e-06, "loss": 0.0123, "step": 29765 }, { "epoch": 5.46, "learning_rate": 4.842037658943353e-06, "loss": 0.0084, "step": 29766 }, { "epoch": 5.46, "learning_rate": 4.841401624969782e-06, "loss": 0.0071, "step": 29767 }, { "epoch": 5.46, "learning_rate": 4.840765619430531e-06, "loss": 0.0085, "step": 29768 }, { "epoch": 5.46, "learning_rate": 4.8401296423291e-06, "loss": 0.0124, "step": 29769 }, { "epoch": 5.46, "learning_rate": 4.839493693668995e-06, "loss": 0.0089, "step": 29770 }, { "epoch": 5.46, "learning_rate": 4.838857773453722e-06, "loss": 0.007, "step": 29771 }, { "epoch": 5.46, "learning_rate": 4.838221881686785e-06, "loss": 0.0063, "step": 29772 }, { "epoch": 5.46, "learning_rate": 4.8375860183716916e-06, "loss": 0.009, "step": 29773 }, { "epoch": 5.46, "learning_rate": 4.836950183511944e-06, "loss": 0.0131, "step": 29774 }, { "epoch": 5.46, "learning_rate": 4.8363143771110474e-06, "loss": 0.0077, "step": 29775 }, { "epoch": 5.46, "learning_rate": 4.835678599172504e-06, "loss": 0.0107, "step": 29776 }, { "epoch": 5.46, "learning_rate": 4.835042849699829e-06, "loss": 0.0089, "step": 29777 }, { "epoch": 5.46, "learning_rate": 4.834407128696515e-06, "loss": 0.0105, "step": 29778 }, { "epoch": 5.46, "learning_rate": 4.83377143616607e-06, "loss": 0.0095, "step": 29779 }, { "epoch": 5.46, "learning_rate": 4.8331357721119965e-06, "loss": 0.0089, "step": 29780 }, { "epoch": 5.46, "learning_rate": 4.832500136537797e-06, "loss": 0.0094, "step": 29781 }, { "epoch": 5.46, "learning_rate": 4.831864529446982e-06, "loss": 0.0083, "step": 29782 }, { "epoch": 5.46, "learning_rate": 4.831228950843049e-06, "loss": 0.0097, "step": 29783 }, { "epoch": 5.46, "learning_rate": 4.8305934007295086e-06, "loss": 0.0077, "step": 29784 }, { "epoch": 5.46, "learning_rate": 4.82995787910985e-06, "loss": 0.0094, "step": 29785 }, { "epoch": 5.46, "learning_rate": 4.829322385987588e-06, "loss": 0.0092, "step": 29786 }, { "epoch": 5.46, "learning_rate": 4.8286869213662235e-06, "loss": 0.0091, "step": 29787 }, { "epoch": 5.46, "learning_rate": 4.828051485249258e-06, "loss": 0.0086, "step": 29788 }, { "epoch": 5.46, "learning_rate": 4.827416077640192e-06, "loss": 0.0082, "step": 29789 }, { "epoch": 5.46, "learning_rate": 4.8267806985425316e-06, "loss": 0.0074, "step": 29790 }, { "epoch": 5.46, "learning_rate": 4.826145347959776e-06, "loss": 0.0084, "step": 29791 }, { "epoch": 5.46, "learning_rate": 4.825510025895429e-06, "loss": 0.0113, "step": 29792 }, { "epoch": 5.46, "learning_rate": 4.824874732352992e-06, "loss": 0.009, "step": 29793 }, { "epoch": 5.46, "learning_rate": 4.8242394673359625e-06, "loss": 0.0078, "step": 29794 }, { "epoch": 5.46, "learning_rate": 4.823604230847855e-06, "loss": 0.0071, "step": 29795 }, { "epoch": 5.46, "learning_rate": 4.822969022892158e-06, "loss": 0.0079, "step": 29796 }, { "epoch": 5.46, "learning_rate": 4.822333843472376e-06, "loss": 0.0082, "step": 29797 }, { "epoch": 5.46, "learning_rate": 4.821698692592008e-06, "loss": 0.0093, "step": 29798 }, { "epoch": 5.46, "learning_rate": 4.821063570254563e-06, "loss": 0.0103, "step": 29799 }, { "epoch": 5.46, "learning_rate": 4.820428476463535e-06, "loss": 0.0101, "step": 29800 }, { "epoch": 5.47, "learning_rate": 4.8197934112224325e-06, "loss": 0.0082, "step": 29801 }, { "epoch": 5.47, "learning_rate": 4.819158374534745e-06, "loss": 0.0054, "step": 29802 }, { "epoch": 5.47, "learning_rate": 4.8185233664039745e-06, "loss": 0.0082, "step": 29803 }, { "epoch": 5.47, "learning_rate": 4.817888386833628e-06, "loss": 0.0094, "step": 29804 }, { "epoch": 5.47, "learning_rate": 4.8172534358272016e-06, "loss": 0.0082, "step": 29805 }, { "epoch": 5.47, "learning_rate": 4.816618513388196e-06, "loss": 0.0062, "step": 29806 }, { "epoch": 5.47, "learning_rate": 4.815983619520111e-06, "loss": 0.0063, "step": 29807 }, { "epoch": 5.47, "learning_rate": 4.815348754226446e-06, "loss": 0.0062, "step": 29808 }, { "epoch": 5.47, "learning_rate": 4.814713917510699e-06, "loss": 0.0083, "step": 29809 }, { "epoch": 5.47, "learning_rate": 4.814079109376372e-06, "loss": 0.008, "step": 29810 }, { "epoch": 5.47, "learning_rate": 4.813444329826961e-06, "loss": 0.0067, "step": 29811 }, { "epoch": 5.47, "learning_rate": 4.812809578865964e-06, "loss": 0.009, "step": 29812 }, { "epoch": 5.47, "learning_rate": 4.812174856496889e-06, "loss": 0.0094, "step": 29813 }, { "epoch": 5.47, "learning_rate": 4.811540162723222e-06, "loss": 0.0078, "step": 29814 }, { "epoch": 5.47, "learning_rate": 4.810905497548471e-06, "loss": 0.0092, "step": 29815 }, { "epoch": 5.47, "learning_rate": 4.810270860976124e-06, "loss": 0.0091, "step": 29816 }, { "epoch": 5.47, "learning_rate": 4.80963625300969e-06, "loss": 0.0078, "step": 29817 }, { "epoch": 5.47, "learning_rate": 4.809001673652662e-06, "loss": 0.0091, "step": 29818 }, { "epoch": 5.47, "learning_rate": 4.808367122908544e-06, "loss": 0.0083, "step": 29819 }, { "epoch": 5.47, "learning_rate": 4.807732600780819e-06, "loss": 0.0065, "step": 29820 }, { "epoch": 5.47, "learning_rate": 4.807098107272997e-06, "loss": 0.0088, "step": 29821 }, { "epoch": 5.47, "learning_rate": 4.806463642388573e-06, "loss": 0.0104, "step": 29822 }, { "epoch": 5.47, "learning_rate": 4.805829206131042e-06, "loss": 0.0087, "step": 29823 }, { "epoch": 5.47, "learning_rate": 4.805194798503903e-06, "loss": 0.0084, "step": 29824 }, { "epoch": 5.47, "learning_rate": 4.804560419510651e-06, "loss": 0.0087, "step": 29825 }, { "epoch": 5.47, "learning_rate": 4.803926069154785e-06, "loss": 0.0078, "step": 29826 }, { "epoch": 5.47, "learning_rate": 4.803291747439799e-06, "loss": 0.0073, "step": 29827 }, { "epoch": 5.47, "learning_rate": 4.802657454369192e-06, "loss": 0.0078, "step": 29828 }, { "epoch": 5.47, "learning_rate": 4.802023189946454e-06, "loss": 0.0079, "step": 29829 }, { "epoch": 5.47, "learning_rate": 4.80138895417509e-06, "loss": 0.007, "step": 29830 }, { "epoch": 5.47, "learning_rate": 4.800754747058596e-06, "loss": 0.0087, "step": 29831 }, { "epoch": 5.47, "learning_rate": 4.800120568600459e-06, "loss": 0.0078, "step": 29832 }, { "epoch": 5.47, "learning_rate": 4.79948641880418e-06, "loss": 0.0086, "step": 29833 }, { "epoch": 5.47, "learning_rate": 4.79885229767325e-06, "loss": 0.0086, "step": 29834 }, { "epoch": 5.47, "learning_rate": 4.79821820521117e-06, "loss": 0.0096, "step": 29835 }, { "epoch": 5.47, "learning_rate": 4.797584141421435e-06, "loss": 0.0112, "step": 29836 }, { "epoch": 5.47, "learning_rate": 4.796950106307542e-06, "loss": 0.0094, "step": 29837 }, { "epoch": 5.47, "learning_rate": 4.796316099872974e-06, "loss": 0.008, "step": 29838 }, { "epoch": 5.47, "learning_rate": 4.7956821221212365e-06, "loss": 0.0072, "step": 29839 }, { "epoch": 5.47, "learning_rate": 4.795048173055822e-06, "loss": 0.0112, "step": 29840 }, { "epoch": 5.47, "learning_rate": 4.794414252680224e-06, "loss": 0.0057, "step": 29841 }, { "epoch": 5.47, "learning_rate": 4.793780360997936e-06, "loss": 0.0091, "step": 29842 }, { "epoch": 5.47, "learning_rate": 4.793146498012452e-06, "loss": 0.0067, "step": 29843 }, { "epoch": 5.47, "learning_rate": 4.792512663727268e-06, "loss": 0.0082, "step": 29844 }, { "epoch": 5.47, "learning_rate": 4.791878858145875e-06, "loss": 0.009, "step": 29845 }, { "epoch": 5.47, "learning_rate": 4.791245081271768e-06, "loss": 0.008, "step": 29846 }, { "epoch": 5.47, "learning_rate": 4.790611333108437e-06, "loss": 0.0102, "step": 29847 }, { "epoch": 5.47, "learning_rate": 4.789977613659385e-06, "loss": 0.0105, "step": 29848 }, { "epoch": 5.47, "learning_rate": 4.789343922928096e-06, "loss": 0.0063, "step": 29849 }, { "epoch": 5.47, "learning_rate": 4.788710260918065e-06, "loss": 0.0081, "step": 29850 }, { "epoch": 5.47, "learning_rate": 4.78807662763278e-06, "loss": 0.0078, "step": 29851 }, { "epoch": 5.47, "learning_rate": 4.787443023075744e-06, "loss": 0.0079, "step": 29852 }, { "epoch": 5.47, "learning_rate": 4.786809447250444e-06, "loss": 0.0099, "step": 29853 }, { "epoch": 5.47, "learning_rate": 4.7861759001603716e-06, "loss": 0.0073, "step": 29854 }, { "epoch": 5.47, "learning_rate": 4.785542381809024e-06, "loss": 0.0081, "step": 29855 }, { "epoch": 5.48, "learning_rate": 4.784908892199882e-06, "loss": 0.0092, "step": 29856 }, { "epoch": 5.48, "learning_rate": 4.784275431336447e-06, "loss": 0.0061, "step": 29857 }, { "epoch": 5.48, "learning_rate": 4.783641999222208e-06, "loss": 0.007, "step": 29858 }, { "epoch": 5.48, "learning_rate": 4.783008595860657e-06, "loss": 0.006, "step": 29859 }, { "epoch": 5.48, "learning_rate": 4.7823752212552855e-06, "loss": 0.0092, "step": 29860 }, { "epoch": 5.48, "learning_rate": 4.7817418754095835e-06, "loss": 0.0077, "step": 29861 }, { "epoch": 5.48, "learning_rate": 4.7811085583270426e-06, "loss": 0.0066, "step": 29862 }, { "epoch": 5.48, "learning_rate": 4.780475270011153e-06, "loss": 0.0102, "step": 29863 }, { "epoch": 5.48, "learning_rate": 4.7798420104654074e-06, "loss": 0.0092, "step": 29864 }, { "epoch": 5.48, "learning_rate": 4.779208779693289e-06, "loss": 0.0105, "step": 29865 }, { "epoch": 5.48, "learning_rate": 4.778575577698303e-06, "loss": 0.0062, "step": 29866 }, { "epoch": 5.48, "learning_rate": 4.777942404483927e-06, "loss": 0.0102, "step": 29867 }, { "epoch": 5.48, "learning_rate": 4.777309260053655e-06, "loss": 0.0109, "step": 29868 }, { "epoch": 5.48, "learning_rate": 4.776676144410973e-06, "loss": 0.0091, "step": 29869 }, { "epoch": 5.48, "learning_rate": 4.7760430575593775e-06, "loss": 0.0084, "step": 29870 }, { "epoch": 5.48, "learning_rate": 4.775409999502356e-06, "loss": 0.009, "step": 29871 }, { "epoch": 5.48, "learning_rate": 4.774776970243396e-06, "loss": 0.0085, "step": 29872 }, { "epoch": 5.48, "learning_rate": 4.7741439697859884e-06, "loss": 0.0062, "step": 29873 }, { "epoch": 5.48, "learning_rate": 4.7735109981336216e-06, "loss": 0.0071, "step": 29874 }, { "epoch": 5.48, "learning_rate": 4.772878055289784e-06, "loss": 0.0089, "step": 29875 }, { "epoch": 5.48, "learning_rate": 4.772245141257965e-06, "loss": 0.0081, "step": 29876 }, { "epoch": 5.48, "learning_rate": 4.771612256041654e-06, "loss": 0.0058, "step": 29877 }, { "epoch": 5.48, "learning_rate": 4.770979399644339e-06, "loss": 0.006, "step": 29878 }, { "epoch": 5.48, "learning_rate": 4.770346572069507e-06, "loss": 0.0106, "step": 29879 }, { "epoch": 5.48, "learning_rate": 4.769713773320648e-06, "loss": 0.0083, "step": 29880 }, { "epoch": 5.48, "learning_rate": 4.769081003401249e-06, "loss": 0.0109, "step": 29881 }, { "epoch": 5.48, "learning_rate": 4.768448262314795e-06, "loss": 0.0097, "step": 29882 }, { "epoch": 5.48, "learning_rate": 4.767815550064778e-06, "loss": 0.0087, "step": 29883 }, { "epoch": 5.48, "learning_rate": 4.7671828666546906e-06, "loss": 0.0099, "step": 29884 }, { "epoch": 5.48, "learning_rate": 4.766550212088009e-06, "loss": 0.0085, "step": 29885 }, { "epoch": 5.48, "learning_rate": 4.765917586368225e-06, "loss": 0.0104, "step": 29886 }, { "epoch": 5.48, "learning_rate": 4.765284989498823e-06, "loss": 0.0085, "step": 29887 }, { "epoch": 5.48, "learning_rate": 4.764652421483295e-06, "loss": 0.0094, "step": 29888 }, { "epoch": 5.48, "learning_rate": 4.764019882325127e-06, "loss": 0.008, "step": 29889 }, { "epoch": 5.48, "learning_rate": 4.763387372027804e-06, "loss": 0.0092, "step": 29890 }, { "epoch": 5.48, "learning_rate": 4.762754890594813e-06, "loss": 0.0071, "step": 29891 }, { "epoch": 5.48, "learning_rate": 4.7621224380296385e-06, "loss": 0.0087, "step": 29892 }, { "epoch": 5.48, "learning_rate": 4.761490014335768e-06, "loss": 0.0106, "step": 29893 }, { "epoch": 5.48, "learning_rate": 4.760857619516688e-06, "loss": 0.0093, "step": 29894 }, { "epoch": 5.48, "learning_rate": 4.760225253575883e-06, "loss": 0.0087, "step": 29895 }, { "epoch": 5.48, "learning_rate": 4.759592916516839e-06, "loss": 0.0069, "step": 29896 }, { "epoch": 5.48, "learning_rate": 4.7589606083430415e-06, "loss": 0.0092, "step": 29897 }, { "epoch": 5.48, "learning_rate": 4.7583283290579764e-06, "loss": 0.011, "step": 29898 }, { "epoch": 5.48, "learning_rate": 4.757696078665128e-06, "loss": 0.0088, "step": 29899 }, { "epoch": 5.48, "learning_rate": 4.757063857167979e-06, "loss": 0.0083, "step": 29900 }, { "epoch": 5.48, "learning_rate": 4.756431664570019e-06, "loss": 0.0056, "step": 29901 }, { "epoch": 5.48, "learning_rate": 4.755799500874735e-06, "loss": 0.0083, "step": 29902 }, { "epoch": 5.48, "learning_rate": 4.755167366085603e-06, "loss": 0.0099, "step": 29903 }, { "epoch": 5.48, "learning_rate": 4.754535260206108e-06, "loss": 0.0097, "step": 29904 }, { "epoch": 5.48, "learning_rate": 4.7539031832397405e-06, "loss": 0.0082, "step": 29905 }, { "epoch": 5.48, "learning_rate": 4.753271135189982e-06, "loss": 0.0069, "step": 29906 }, { "epoch": 5.48, "learning_rate": 4.752639116060317e-06, "loss": 0.0081, "step": 29907 }, { "epoch": 5.48, "learning_rate": 4.75200712585423e-06, "loss": 0.0081, "step": 29908 }, { "epoch": 5.48, "learning_rate": 4.7513751645751956e-06, "loss": 0.0069, "step": 29909 }, { "epoch": 5.49, "learning_rate": 4.750743232226709e-06, "loss": 0.0085, "step": 29910 }, { "epoch": 5.49, "learning_rate": 4.750111328812248e-06, "loss": 0.0105, "step": 29911 }, { "epoch": 5.49, "learning_rate": 4.749479454335297e-06, "loss": 0.0099, "step": 29912 }, { "epoch": 5.49, "learning_rate": 4.7488476087993374e-06, "loss": 0.0069, "step": 29913 }, { "epoch": 5.49, "learning_rate": 4.748215792207853e-06, "loss": 0.0074, "step": 29914 }, { "epoch": 5.49, "learning_rate": 4.747584004564327e-06, "loss": 0.007, "step": 29915 }, { "epoch": 5.49, "learning_rate": 4.7469522458722406e-06, "loss": 0.0092, "step": 29916 }, { "epoch": 5.49, "learning_rate": 4.746320516135072e-06, "loss": 0.0096, "step": 29917 }, { "epoch": 5.49, "learning_rate": 4.745688815356312e-06, "loss": 0.0092, "step": 29918 }, { "epoch": 5.49, "learning_rate": 4.745057143539439e-06, "loss": 0.0083, "step": 29919 }, { "epoch": 5.49, "learning_rate": 4.744425500687937e-06, "loss": 0.0082, "step": 29920 }, { "epoch": 5.49, "learning_rate": 4.743793886805281e-06, "loss": 0.0071, "step": 29921 }, { "epoch": 5.49, "learning_rate": 4.743162301894952e-06, "loss": 0.0049, "step": 29922 }, { "epoch": 5.49, "learning_rate": 4.7425307459604404e-06, "loss": 0.0084, "step": 29923 }, { "epoch": 5.49, "learning_rate": 4.741899219005223e-06, "loss": 0.0078, "step": 29924 }, { "epoch": 5.49, "learning_rate": 4.741267721032779e-06, "loss": 0.0065, "step": 29925 }, { "epoch": 5.49, "learning_rate": 4.740636252046591e-06, "loss": 0.0061, "step": 29926 }, { "epoch": 5.49, "learning_rate": 4.74000481205014e-06, "loss": 0.0065, "step": 29927 }, { "epoch": 5.49, "learning_rate": 4.739373401046904e-06, "loss": 0.0091, "step": 29928 }, { "epoch": 5.49, "learning_rate": 4.738742019040366e-06, "loss": 0.0096, "step": 29929 }, { "epoch": 5.49, "learning_rate": 4.738110666034006e-06, "loss": 0.0067, "step": 29930 }, { "epoch": 5.49, "learning_rate": 4.737479342031301e-06, "loss": 0.0068, "step": 29931 }, { "epoch": 5.49, "learning_rate": 4.736848047035736e-06, "loss": 0.0093, "step": 29932 }, { "epoch": 5.49, "learning_rate": 4.736216781050787e-06, "loss": 0.01, "step": 29933 }, { "epoch": 5.49, "learning_rate": 4.735585544079933e-06, "loss": 0.0091, "step": 29934 }, { "epoch": 5.49, "learning_rate": 4.734954336126652e-06, "loss": 0.0089, "step": 29935 }, { "epoch": 5.49, "learning_rate": 4.734323157194429e-06, "loss": 0.009, "step": 29936 }, { "epoch": 5.49, "learning_rate": 4.733692007286742e-06, "loss": 0.0097, "step": 29937 }, { "epoch": 5.49, "learning_rate": 4.733060886407071e-06, "loss": 0.0066, "step": 29938 }, { "epoch": 5.49, "learning_rate": 4.732429794558887e-06, "loss": 0.0091, "step": 29939 }, { "epoch": 5.49, "learning_rate": 4.73179873174567e-06, "loss": 0.008, "step": 29940 }, { "epoch": 5.49, "learning_rate": 4.731167697970906e-06, "loss": 0.0094, "step": 29941 }, { "epoch": 5.49, "learning_rate": 4.7305366932380695e-06, "loss": 0.008, "step": 29942 }, { "epoch": 5.49, "learning_rate": 4.729905717550638e-06, "loss": 0.0098, "step": 29943 }, { "epoch": 5.49, "learning_rate": 4.729274770912089e-06, "loss": 0.0084, "step": 29944 }, { "epoch": 5.49, "learning_rate": 4.7286438533259015e-06, "loss": 0.0071, "step": 29945 }, { "epoch": 5.49, "learning_rate": 4.728012964795552e-06, "loss": 0.0105, "step": 29946 }, { "epoch": 5.49, "learning_rate": 4.727382105324518e-06, "loss": 0.0086, "step": 29947 }, { "epoch": 5.49, "learning_rate": 4.726751274916276e-06, "loss": 0.0087, "step": 29948 }, { "epoch": 5.49, "learning_rate": 4.726120473574309e-06, "loss": 0.0082, "step": 29949 }, { "epoch": 5.49, "learning_rate": 4.725489701302088e-06, "loss": 0.0105, "step": 29950 }, { "epoch": 5.49, "learning_rate": 4.724858958103091e-06, "loss": 0.0068, "step": 29951 }, { "epoch": 5.49, "learning_rate": 4.724228243980794e-06, "loss": 0.0076, "step": 29952 }, { "epoch": 5.49, "learning_rate": 4.7235975589386715e-06, "loss": 0.0082, "step": 29953 }, { "epoch": 5.49, "learning_rate": 4.722966902980208e-06, "loss": 0.0087, "step": 29954 }, { "epoch": 5.49, "learning_rate": 4.7223362761088755e-06, "loss": 0.0086, "step": 29955 }, { "epoch": 5.49, "learning_rate": 4.721705678328147e-06, "loss": 0.007, "step": 29956 }, { "epoch": 5.49, "learning_rate": 4.721075109641497e-06, "loss": 0.0096, "step": 29957 }, { "epoch": 5.49, "learning_rate": 4.720444570052407e-06, "loss": 0.0084, "step": 29958 }, { "epoch": 5.49, "learning_rate": 4.7198140595643505e-06, "loss": 0.0072, "step": 29959 }, { "epoch": 5.49, "learning_rate": 4.719183578180803e-06, "loss": 0.0102, "step": 29960 }, { "epoch": 5.49, "learning_rate": 4.7185531259052395e-06, "loss": 0.009, "step": 29961 }, { "epoch": 5.49, "learning_rate": 4.717922702741136e-06, "loss": 0.0088, "step": 29962 }, { "epoch": 5.49, "learning_rate": 4.717292308691964e-06, "loss": 0.0095, "step": 29963 }, { "epoch": 5.49, "learning_rate": 4.716661943761201e-06, "loss": 0.0066, "step": 29964 }, { "epoch": 5.5, "learning_rate": 4.716031607952323e-06, "loss": 0.009, "step": 29965 }, { "epoch": 5.5, "learning_rate": 4.715401301268797e-06, "loss": 0.0084, "step": 29966 }, { "epoch": 5.5, "learning_rate": 4.714771023714112e-06, "loss": 0.0073, "step": 29967 }, { "epoch": 5.5, "learning_rate": 4.714140775291728e-06, "loss": 0.0073, "step": 29968 }, { "epoch": 5.5, "learning_rate": 4.713510556005124e-06, "loss": 0.0103, "step": 29969 }, { "epoch": 5.5, "learning_rate": 4.712880365857769e-06, "loss": 0.0096, "step": 29970 }, { "epoch": 5.5, "learning_rate": 4.712250204853147e-06, "loss": 0.0091, "step": 29971 }, { "epoch": 5.5, "learning_rate": 4.711620072994725e-06, "loss": 0.0085, "step": 29972 }, { "epoch": 5.5, "learning_rate": 4.71098997028598e-06, "loss": 0.0087, "step": 29973 }, { "epoch": 5.5, "learning_rate": 4.710359896730379e-06, "loss": 0.0105, "step": 29974 }, { "epoch": 5.5, "learning_rate": 4.709729852331395e-06, "loss": 0.0074, "step": 29975 }, { "epoch": 5.5, "learning_rate": 4.709099837092507e-06, "loss": 0.0075, "step": 29976 }, { "epoch": 5.5, "learning_rate": 4.7084698510171845e-06, "loss": 0.0093, "step": 29977 }, { "epoch": 5.5, "learning_rate": 4.707839894108901e-06, "loss": 0.0082, "step": 29978 }, { "epoch": 5.5, "learning_rate": 4.707209966371127e-06, "loss": 0.0086, "step": 29979 }, { "epoch": 5.5, "learning_rate": 4.706580067807336e-06, "loss": 0.0063, "step": 29980 }, { "epoch": 5.5, "learning_rate": 4.705950198421e-06, "loss": 0.0065, "step": 29981 }, { "epoch": 5.5, "learning_rate": 4.70532035821559e-06, "loss": 0.0068, "step": 29982 }, { "epoch": 5.5, "learning_rate": 4.704690547194579e-06, "loss": 0.0074, "step": 29983 }, { "epoch": 5.5, "learning_rate": 4.7040607653614335e-06, "loss": 0.009, "step": 29984 }, { "epoch": 5.5, "learning_rate": 4.703431012719636e-06, "loss": 0.0074, "step": 29985 }, { "epoch": 5.5, "learning_rate": 4.702801289272648e-06, "loss": 0.0082, "step": 29986 }, { "epoch": 5.5, "learning_rate": 4.702171595023943e-06, "loss": 0.007, "step": 29987 }, { "epoch": 5.5, "learning_rate": 4.7015419299769896e-06, "loss": 0.0085, "step": 29988 }, { "epoch": 5.5, "learning_rate": 4.700912294135264e-06, "loss": 0.0064, "step": 29989 }, { "epoch": 5.5, "learning_rate": 4.700282687502233e-06, "loss": 0.0084, "step": 29990 }, { "epoch": 5.5, "learning_rate": 4.699653110081373e-06, "loss": 0.0119, "step": 29991 }, { "epoch": 5.5, "learning_rate": 4.699023561876146e-06, "loss": 0.0067, "step": 29992 }, { "epoch": 5.5, "learning_rate": 4.698394042890021e-06, "loss": 0.0104, "step": 29993 }, { "epoch": 5.5, "learning_rate": 4.6977645531264745e-06, "loss": 0.0087, "step": 29994 }, { "epoch": 5.5, "learning_rate": 4.697135092588976e-06, "loss": 0.0073, "step": 29995 }, { "epoch": 5.5, "learning_rate": 4.696505661280992e-06, "loss": 0.0079, "step": 29996 }, { "epoch": 5.5, "learning_rate": 4.695876259205994e-06, "loss": 0.0064, "step": 29997 }, { "epoch": 5.5, "learning_rate": 4.695246886367449e-06, "loss": 0.0074, "step": 29998 }, { "epoch": 5.5, "learning_rate": 4.694617542768829e-06, "loss": 0.0109, "step": 29999 }, { "epoch": 5.5, "learning_rate": 4.693988228413601e-06, "loss": 0.0063, "step": 30000 }, { "epoch": 5.5, "learning_rate": 4.69335894330523e-06, "loss": 0.0097, "step": 30001 }, { "epoch": 5.5, "learning_rate": 4.692729687447196e-06, "loss": 0.0088, "step": 30002 }, { "epoch": 5.5, "learning_rate": 4.692100460842958e-06, "loss": 0.008, "step": 30003 }, { "epoch": 5.5, "learning_rate": 4.691471263495985e-06, "loss": 0.0094, "step": 30004 }, { "epoch": 5.5, "learning_rate": 4.690842095409747e-06, "loss": 0.0084, "step": 30005 }, { "epoch": 5.5, "learning_rate": 4.690212956587708e-06, "loss": 0.008, "step": 30006 }, { "epoch": 5.5, "learning_rate": 4.689583847033345e-06, "loss": 0.0084, "step": 30007 }, { "epoch": 5.5, "learning_rate": 4.688954766750118e-06, "loss": 0.0073, "step": 30008 }, { "epoch": 5.5, "learning_rate": 4.688325715741501e-06, "loss": 0.0084, "step": 30009 }, { "epoch": 5.5, "learning_rate": 4.6876966940109505e-06, "loss": 0.0105, "step": 30010 }, { "epoch": 5.5, "learning_rate": 4.687067701561944e-06, "loss": 0.0099, "step": 30011 }, { "epoch": 5.5, "learning_rate": 4.686438738397944e-06, "loss": 0.0076, "step": 30012 }, { "epoch": 5.5, "learning_rate": 4.685809804522417e-06, "loss": 0.0079, "step": 30013 }, { "epoch": 5.5, "learning_rate": 4.6851808999388334e-06, "loss": 0.0107, "step": 30014 }, { "epoch": 5.5, "learning_rate": 4.684552024650656e-06, "loss": 0.0066, "step": 30015 }, { "epoch": 5.5, "learning_rate": 4.6839231786613525e-06, "loss": 0.0058, "step": 30016 }, { "epoch": 5.5, "learning_rate": 4.6832943619743885e-06, "loss": 0.0093, "step": 30017 }, { "epoch": 5.5, "learning_rate": 4.682665574593231e-06, "loss": 0.0084, "step": 30018 }, { "epoch": 5.51, "learning_rate": 4.682036816521343e-06, "loss": 0.0095, "step": 30019 }, { "epoch": 5.51, "learning_rate": 4.681408087762198e-06, "loss": 0.0098, "step": 30020 }, { "epoch": 5.51, "learning_rate": 4.680779388319253e-06, "loss": 0.0104, "step": 30021 }, { "epoch": 5.51, "learning_rate": 4.680150718195977e-06, "loss": 0.0084, "step": 30022 }, { "epoch": 5.51, "learning_rate": 4.679522077395831e-06, "loss": 0.0077, "step": 30023 }, { "epoch": 5.51, "learning_rate": 4.678893465922288e-06, "loss": 0.008, "step": 30024 }, { "epoch": 5.51, "learning_rate": 4.678264883778809e-06, "loss": 0.0089, "step": 30025 }, { "epoch": 5.51, "learning_rate": 4.677636330968858e-06, "loss": 0.0088, "step": 30026 }, { "epoch": 5.51, "learning_rate": 4.677007807495904e-06, "loss": 0.0104, "step": 30027 }, { "epoch": 5.51, "learning_rate": 4.6763793133634005e-06, "loss": 0.0072, "step": 30028 }, { "epoch": 5.51, "learning_rate": 4.675750848574823e-06, "loss": 0.0091, "step": 30029 }, { "epoch": 5.51, "learning_rate": 4.6751224131336315e-06, "loss": 0.0073, "step": 30030 }, { "epoch": 5.51, "learning_rate": 4.674494007043288e-06, "loss": 0.0071, "step": 30031 }, { "epoch": 5.51, "learning_rate": 4.6738656303072615e-06, "loss": 0.0052, "step": 30032 }, { "epoch": 5.51, "learning_rate": 4.67323728292901e-06, "loss": 0.0106, "step": 30033 }, { "epoch": 5.51, "learning_rate": 4.672608964912001e-06, "loss": 0.0084, "step": 30034 }, { "epoch": 5.51, "learning_rate": 4.671980676259695e-06, "loss": 0.0098, "step": 30035 }, { "epoch": 5.51, "learning_rate": 4.6713524169755575e-06, "loss": 0.007, "step": 30036 }, { "epoch": 5.51, "learning_rate": 4.670724187063046e-06, "loss": 0.006, "step": 30037 }, { "epoch": 5.51, "learning_rate": 4.670095986525635e-06, "loss": 0.0088, "step": 30038 }, { "epoch": 5.51, "learning_rate": 4.669467815366776e-06, "loss": 0.0081, "step": 30039 }, { "epoch": 5.51, "learning_rate": 4.668839673589934e-06, "loss": 0.0086, "step": 30040 }, { "epoch": 5.51, "learning_rate": 4.668211561198571e-06, "loss": 0.0093, "step": 30041 }, { "epoch": 5.51, "learning_rate": 4.667583478196154e-06, "loss": 0.0061, "step": 30042 }, { "epoch": 5.51, "learning_rate": 4.666955424586141e-06, "loss": 0.0053, "step": 30043 }, { "epoch": 5.51, "learning_rate": 4.666327400371995e-06, "loss": 0.0085, "step": 30044 }, { "epoch": 5.51, "learning_rate": 4.665699405557178e-06, "loss": 0.0085, "step": 30045 }, { "epoch": 5.51, "learning_rate": 4.66507144014515e-06, "loss": 0.0098, "step": 30046 }, { "epoch": 5.51, "learning_rate": 4.664443504139374e-06, "loss": 0.0078, "step": 30047 }, { "epoch": 5.51, "learning_rate": 4.66381559754331e-06, "loss": 0.0075, "step": 30048 }, { "epoch": 5.51, "learning_rate": 4.663187720360419e-06, "loss": 0.01, "step": 30049 }, { "epoch": 5.51, "learning_rate": 4.662559872594163e-06, "loss": 0.0102, "step": 30050 }, { "epoch": 5.51, "learning_rate": 4.661932054248002e-06, "loss": 0.0101, "step": 30051 }, { "epoch": 5.51, "learning_rate": 4.661304265325396e-06, "loss": 0.0063, "step": 30052 }, { "epoch": 5.51, "learning_rate": 4.660676505829807e-06, "loss": 0.0115, "step": 30053 }, { "epoch": 5.51, "learning_rate": 4.6600487757646885e-06, "loss": 0.0109, "step": 30054 }, { "epoch": 5.51, "learning_rate": 4.659421075133511e-06, "loss": 0.0059, "step": 30055 }, { "epoch": 5.51, "learning_rate": 4.658793403939734e-06, "loss": 0.008, "step": 30056 }, { "epoch": 5.51, "learning_rate": 4.6581657621868095e-06, "loss": 0.0086, "step": 30057 }, { "epoch": 5.51, "learning_rate": 4.657538149878199e-06, "loss": 0.0084, "step": 30058 }, { "epoch": 5.51, "learning_rate": 4.656910567017361e-06, "loss": 0.0084, "step": 30059 }, { "epoch": 5.51, "learning_rate": 4.65628301360776e-06, "loss": 0.0061, "step": 30060 }, { "epoch": 5.51, "learning_rate": 4.6556554896528514e-06, "loss": 0.0107, "step": 30061 }, { "epoch": 5.51, "learning_rate": 4.655027995156101e-06, "loss": 0.0086, "step": 30062 }, { "epoch": 5.51, "learning_rate": 4.654400530120953e-06, "loss": 0.0112, "step": 30063 }, { "epoch": 5.51, "learning_rate": 4.6537730945508774e-06, "loss": 0.0057, "step": 30064 }, { "epoch": 5.51, "learning_rate": 4.65314568844933e-06, "loss": 0.0081, "step": 30065 }, { "epoch": 5.51, "learning_rate": 4.6525183118197695e-06, "loss": 0.0079, "step": 30066 }, { "epoch": 5.51, "learning_rate": 4.651890964665653e-06, "loss": 0.0093, "step": 30067 }, { "epoch": 5.51, "learning_rate": 4.6512636469904385e-06, "loss": 0.0101, "step": 30068 }, { "epoch": 5.51, "learning_rate": 4.650636358797585e-06, "loss": 0.0078, "step": 30069 }, { "epoch": 5.51, "learning_rate": 4.650009100090548e-06, "loss": 0.0104, "step": 30070 }, { "epoch": 5.51, "learning_rate": 4.6493818708727866e-06, "loss": 0.006, "step": 30071 }, { "epoch": 5.51, "learning_rate": 4.648754671147755e-06, "loss": 0.0084, "step": 30072 }, { "epoch": 5.51, "learning_rate": 4.648127500918915e-06, "loss": 0.0089, "step": 30073 }, { "epoch": 5.52, "learning_rate": 4.6475003601897265e-06, "loss": 0.0084, "step": 30074 }, { "epoch": 5.52, "learning_rate": 4.646873248963637e-06, "loss": 0.0069, "step": 30075 }, { "epoch": 5.52, "learning_rate": 4.646246167244105e-06, "loss": 0.0083, "step": 30076 }, { "epoch": 5.52, "learning_rate": 4.645619115034592e-06, "loss": 0.0091, "step": 30077 }, { "epoch": 5.52, "learning_rate": 4.644992092338552e-06, "loss": 0.0098, "step": 30078 }, { "epoch": 5.52, "learning_rate": 4.644365099159443e-06, "loss": 0.0087, "step": 30079 }, { "epoch": 5.52, "learning_rate": 4.64373813550072e-06, "loss": 0.0067, "step": 30080 }, { "epoch": 5.52, "learning_rate": 4.643111201365832e-06, "loss": 0.0078, "step": 30081 }, { "epoch": 5.52, "learning_rate": 4.642484296758244e-06, "loss": 0.0085, "step": 30082 }, { "epoch": 5.52, "learning_rate": 4.641857421681408e-06, "loss": 0.0085, "step": 30083 }, { "epoch": 5.52, "learning_rate": 4.641230576138778e-06, "loss": 0.0078, "step": 30084 }, { "epoch": 5.52, "learning_rate": 4.640603760133813e-06, "loss": 0.0073, "step": 30085 }, { "epoch": 5.52, "learning_rate": 4.639976973669965e-06, "loss": 0.0073, "step": 30086 }, { "epoch": 5.52, "learning_rate": 4.639350216750689e-06, "loss": 0.009, "step": 30087 }, { "epoch": 5.52, "learning_rate": 4.638723489379441e-06, "loss": 0.0087, "step": 30088 }, { "epoch": 5.52, "learning_rate": 4.638096791559673e-06, "loss": 0.0124, "step": 30089 }, { "epoch": 5.52, "learning_rate": 4.637470123294838e-06, "loss": 0.0082, "step": 30090 }, { "epoch": 5.52, "learning_rate": 4.636843484588398e-06, "loss": 0.0064, "step": 30091 }, { "epoch": 5.52, "learning_rate": 4.6362168754438055e-06, "loss": 0.0063, "step": 30092 }, { "epoch": 5.52, "learning_rate": 4.635590295864508e-06, "loss": 0.0056, "step": 30093 }, { "epoch": 5.52, "learning_rate": 4.634963745853957e-06, "loss": 0.0087, "step": 30094 }, { "epoch": 5.52, "learning_rate": 4.6343372254156175e-06, "loss": 0.01, "step": 30095 }, { "epoch": 5.52, "learning_rate": 4.633710734552937e-06, "loss": 0.0099, "step": 30096 }, { "epoch": 5.52, "learning_rate": 4.633084273269367e-06, "loss": 0.0095, "step": 30097 }, { "epoch": 5.52, "learning_rate": 4.632457841568363e-06, "loss": 0.0076, "step": 30098 }, { "epoch": 5.52, "learning_rate": 4.631831439453377e-06, "loss": 0.0099, "step": 30099 }, { "epoch": 5.52, "learning_rate": 4.631205066927862e-06, "loss": 0.0072, "step": 30100 }, { "epoch": 5.52, "learning_rate": 4.630578723995269e-06, "loss": 0.0092, "step": 30101 }, { "epoch": 5.52, "learning_rate": 4.629952410659054e-06, "loss": 0.0079, "step": 30102 }, { "epoch": 5.52, "learning_rate": 4.629326126922666e-06, "loss": 0.0076, "step": 30103 }, { "epoch": 5.52, "learning_rate": 4.628699872789558e-06, "loss": 0.0117, "step": 30104 }, { "epoch": 5.52, "learning_rate": 4.628073648263182e-06, "loss": 0.0092, "step": 30105 }, { "epoch": 5.52, "learning_rate": 4.627447453346991e-06, "loss": 0.0065, "step": 30106 }, { "epoch": 5.52, "learning_rate": 4.626821288044432e-06, "loss": 0.0073, "step": 30107 }, { "epoch": 5.52, "learning_rate": 4.626195152358963e-06, "loss": 0.0083, "step": 30108 }, { "epoch": 5.52, "learning_rate": 4.6255690462940364e-06, "loss": 0.011, "step": 30109 }, { "epoch": 5.52, "learning_rate": 4.624942969853094e-06, "loss": 0.009, "step": 30110 }, { "epoch": 5.52, "learning_rate": 4.624316923039593e-06, "loss": 0.0109, "step": 30111 }, { "epoch": 5.52, "learning_rate": 4.623690905856979e-06, "loss": 0.0092, "step": 30112 }, { "epoch": 5.52, "learning_rate": 4.62306491830871e-06, "loss": 0.0095, "step": 30113 }, { "epoch": 5.52, "learning_rate": 4.622438960398234e-06, "loss": 0.0087, "step": 30114 }, { "epoch": 5.52, "learning_rate": 4.621813032129e-06, "loss": 0.0104, "step": 30115 }, { "epoch": 5.52, "learning_rate": 4.621187133504459e-06, "loss": 0.0076, "step": 30116 }, { "epoch": 5.52, "learning_rate": 4.620561264528061e-06, "loss": 0.0105, "step": 30117 }, { "epoch": 5.52, "learning_rate": 4.619935425203254e-06, "loss": 0.0082, "step": 30118 }, { "epoch": 5.52, "learning_rate": 4.619309615533489e-06, "loss": 0.0095, "step": 30119 }, { "epoch": 5.52, "learning_rate": 4.618683835522216e-06, "loss": 0.0082, "step": 30120 }, { "epoch": 5.52, "learning_rate": 4.618058085172884e-06, "loss": 0.0079, "step": 30121 }, { "epoch": 5.52, "learning_rate": 4.6174323644889414e-06, "loss": 0.0055, "step": 30122 }, { "epoch": 5.52, "learning_rate": 4.616806673473838e-06, "loss": 0.0105, "step": 30123 }, { "epoch": 5.52, "learning_rate": 4.6161810121310225e-06, "loss": 0.0091, "step": 30124 }, { "epoch": 5.52, "learning_rate": 4.61555538046394e-06, "loss": 0.0079, "step": 30125 }, { "epoch": 5.52, "learning_rate": 4.614929778476045e-06, "loss": 0.0087, "step": 30126 }, { "epoch": 5.52, "learning_rate": 4.6143042061707875e-06, "loss": 0.0065, "step": 30127 }, { "epoch": 5.53, "learning_rate": 4.613678663551607e-06, "loss": 0.0106, "step": 30128 }, { "epoch": 5.53, "learning_rate": 4.613053150621953e-06, "loss": 0.0066, "step": 30129 }, { "epoch": 5.53, "learning_rate": 4.612427667385281e-06, "loss": 0.0074, "step": 30130 }, { "epoch": 5.53, "learning_rate": 4.611802213845032e-06, "loss": 0.0064, "step": 30131 }, { "epoch": 5.53, "learning_rate": 4.611176790004657e-06, "loss": 0.0117, "step": 30132 }, { "epoch": 5.53, "learning_rate": 4.610551395867599e-06, "loss": 0.0067, "step": 30133 }, { "epoch": 5.53, "learning_rate": 4.609926031437311e-06, "loss": 0.0097, "step": 30134 }, { "epoch": 5.53, "learning_rate": 4.609300696717236e-06, "loss": 0.0101, "step": 30135 }, { "epoch": 5.53, "learning_rate": 4.608675391710821e-06, "loss": 0.0063, "step": 30136 }, { "epoch": 5.53, "learning_rate": 4.608050116421516e-06, "loss": 0.0064, "step": 30137 }, { "epoch": 5.53, "learning_rate": 4.607424870852758e-06, "loss": 0.007, "step": 30138 }, { "epoch": 5.53, "learning_rate": 4.606799655008009e-06, "loss": 0.0087, "step": 30139 }, { "epoch": 5.53, "learning_rate": 4.606174468890704e-06, "loss": 0.0071, "step": 30140 }, { "epoch": 5.53, "learning_rate": 4.605549312504292e-06, "loss": 0.0058, "step": 30141 }, { "epoch": 5.53, "learning_rate": 4.604924185852217e-06, "loss": 0.0075, "step": 30142 }, { "epoch": 5.53, "learning_rate": 4.604299088937923e-06, "loss": 0.0089, "step": 30143 }, { "epoch": 5.53, "learning_rate": 4.603674021764863e-06, "loss": 0.0091, "step": 30144 }, { "epoch": 5.53, "learning_rate": 4.603048984336483e-06, "loss": 0.0082, "step": 30145 }, { "epoch": 5.53, "learning_rate": 4.6024239766562175e-06, "loss": 0.0069, "step": 30146 }, { "epoch": 5.53, "learning_rate": 4.601798998727515e-06, "loss": 0.01, "step": 30147 }, { "epoch": 5.53, "learning_rate": 4.601174050553826e-06, "loss": 0.0103, "step": 30148 }, { "epoch": 5.53, "learning_rate": 4.600549132138593e-06, "loss": 0.0116, "step": 30149 }, { "epoch": 5.53, "learning_rate": 4.599924243485259e-06, "loss": 0.0086, "step": 30150 }, { "epoch": 5.53, "learning_rate": 4.599299384597269e-06, "loss": 0.0107, "step": 30151 }, { "epoch": 5.53, "learning_rate": 4.598674555478068e-06, "loss": 0.0089, "step": 30152 }, { "epoch": 5.53, "learning_rate": 4.598049756131099e-06, "loss": 0.0074, "step": 30153 }, { "epoch": 5.53, "learning_rate": 4.597424986559805e-06, "loss": 0.0078, "step": 30154 }, { "epoch": 5.53, "learning_rate": 4.596800246767633e-06, "loss": 0.0101, "step": 30155 }, { "epoch": 5.53, "learning_rate": 4.596175536758024e-06, "loss": 0.0085, "step": 30156 }, { "epoch": 5.53, "learning_rate": 4.595550856534422e-06, "loss": 0.0083, "step": 30157 }, { "epoch": 5.53, "learning_rate": 4.5949262061002695e-06, "loss": 0.0086, "step": 30158 }, { "epoch": 5.53, "learning_rate": 4.594301585459011e-06, "loss": 0.0057, "step": 30159 }, { "epoch": 5.53, "learning_rate": 4.593676994614086e-06, "loss": 0.0069, "step": 30160 }, { "epoch": 5.53, "learning_rate": 4.593052433568943e-06, "loss": 0.0091, "step": 30161 }, { "epoch": 5.53, "learning_rate": 4.592427902327022e-06, "loss": 0.0087, "step": 30162 }, { "epoch": 5.53, "learning_rate": 4.591803400891768e-06, "loss": 0.0097, "step": 30163 }, { "epoch": 5.53, "learning_rate": 4.591178929266617e-06, "loss": 0.0082, "step": 30164 }, { "epoch": 5.53, "learning_rate": 4.590554487455011e-06, "loss": 0.0094, "step": 30165 }, { "epoch": 5.53, "learning_rate": 4.589930075460398e-06, "loss": 0.0078, "step": 30166 }, { "epoch": 5.53, "learning_rate": 4.589305693286219e-06, "loss": 0.0088, "step": 30167 }, { "epoch": 5.53, "learning_rate": 4.588681340935912e-06, "loss": 0.0055, "step": 30168 }, { "epoch": 5.53, "learning_rate": 4.588057018412922e-06, "loss": 0.0081, "step": 30169 }, { "epoch": 5.53, "learning_rate": 4.587432725720687e-06, "loss": 0.0097, "step": 30170 }, { "epoch": 5.53, "learning_rate": 4.586808462862651e-06, "loss": 0.0055, "step": 30171 }, { "epoch": 5.53, "learning_rate": 4.586184229842252e-06, "loss": 0.0078, "step": 30172 }, { "epoch": 5.53, "learning_rate": 4.58556002666293e-06, "loss": 0.0095, "step": 30173 }, { "epoch": 5.53, "learning_rate": 4.584935853328134e-06, "loss": 0.01, "step": 30174 }, { "epoch": 5.53, "learning_rate": 4.584311709841296e-06, "loss": 0.0072, "step": 30175 }, { "epoch": 5.53, "learning_rate": 4.583687596205858e-06, "loss": 0.0082, "step": 30176 }, { "epoch": 5.53, "learning_rate": 4.583063512425262e-06, "loss": 0.0086, "step": 30177 }, { "epoch": 5.53, "learning_rate": 4.582439458502942e-06, "loss": 0.0057, "step": 30178 }, { "epoch": 5.53, "learning_rate": 4.581815434442346e-06, "loss": 0.0079, "step": 30179 }, { "epoch": 5.53, "learning_rate": 4.5811914402469115e-06, "loss": 0.0059, "step": 30180 }, { "epoch": 5.53, "learning_rate": 4.580567475920079e-06, "loss": 0.0079, "step": 30181 }, { "epoch": 5.53, "learning_rate": 4.57994354146528e-06, "loss": 0.0063, "step": 30182 }, { "epoch": 5.54, "learning_rate": 4.579319636885962e-06, "loss": 0.0094, "step": 30183 }, { "epoch": 5.54, "learning_rate": 4.57869576218556e-06, "loss": 0.0089, "step": 30184 }, { "epoch": 5.54, "learning_rate": 4.578071917367515e-06, "loss": 0.0073, "step": 30185 }, { "epoch": 5.54, "learning_rate": 4.577448102435265e-06, "loss": 0.0107, "step": 30186 }, { "epoch": 5.54, "learning_rate": 4.576824317392247e-06, "loss": 0.0083, "step": 30187 }, { "epoch": 5.54, "learning_rate": 4.576200562241901e-06, "loss": 0.0085, "step": 30188 }, { "epoch": 5.54, "learning_rate": 4.575576836987665e-06, "loss": 0.0083, "step": 30189 }, { "epoch": 5.54, "learning_rate": 4.574953141632976e-06, "loss": 0.0083, "step": 30190 }, { "epoch": 5.54, "learning_rate": 4.5743294761812695e-06, "loss": 0.0069, "step": 30191 }, { "epoch": 5.54, "learning_rate": 4.573705840635992e-06, "loss": 0.0095, "step": 30192 }, { "epoch": 5.54, "learning_rate": 4.573082235000572e-06, "loss": 0.009, "step": 30193 }, { "epoch": 5.54, "learning_rate": 4.57245865927845e-06, "loss": 0.0063, "step": 30194 }, { "epoch": 5.54, "learning_rate": 4.5718351134730635e-06, "loss": 0.0069, "step": 30195 }, { "epoch": 5.54, "learning_rate": 4.571211597587844e-06, "loss": 0.0096, "step": 30196 }, { "epoch": 5.54, "learning_rate": 4.570588111626238e-06, "loss": 0.0072, "step": 30197 }, { "epoch": 5.54, "learning_rate": 4.569964655591678e-06, "loss": 0.0084, "step": 30198 }, { "epoch": 5.54, "learning_rate": 4.569341229487604e-06, "loss": 0.0095, "step": 30199 }, { "epoch": 5.54, "learning_rate": 4.56871783331744e-06, "loss": 0.0093, "step": 30200 }, { "epoch": 5.54, "learning_rate": 4.568094467084634e-06, "loss": 0.0084, "step": 30201 }, { "epoch": 5.54, "learning_rate": 4.567471130792619e-06, "loss": 0.0079, "step": 30202 }, { "epoch": 5.54, "learning_rate": 4.56684782444483e-06, "loss": 0.0111, "step": 30203 }, { "epoch": 5.54, "learning_rate": 4.566224548044703e-06, "loss": 0.0071, "step": 30204 }, { "epoch": 5.54, "learning_rate": 4.5656013015956746e-06, "loss": 0.0097, "step": 30205 }, { "epoch": 5.54, "learning_rate": 4.5649780851011784e-06, "loss": 0.0061, "step": 30206 }, { "epoch": 5.54, "learning_rate": 4.56435489856465e-06, "loss": 0.0087, "step": 30207 }, { "epoch": 5.54, "learning_rate": 4.563731741989526e-06, "loss": 0.0106, "step": 30208 }, { "epoch": 5.54, "learning_rate": 4.5631086153792344e-06, "loss": 0.0089, "step": 30209 }, { "epoch": 5.54, "learning_rate": 4.562485518737224e-06, "loss": 0.008, "step": 30210 }, { "epoch": 5.54, "learning_rate": 4.561862452066917e-06, "loss": 0.0075, "step": 30211 }, { "epoch": 5.54, "learning_rate": 4.561239415371752e-06, "loss": 0.0071, "step": 30212 }, { "epoch": 5.54, "learning_rate": 4.560616408655158e-06, "loss": 0.0078, "step": 30213 }, { "epoch": 5.54, "learning_rate": 4.559993431920578e-06, "loss": 0.0085, "step": 30214 }, { "epoch": 5.54, "learning_rate": 4.559370485171442e-06, "loss": 0.0057, "step": 30215 }, { "epoch": 5.54, "learning_rate": 4.558747568411188e-06, "loss": 0.008, "step": 30216 }, { "epoch": 5.54, "learning_rate": 4.55812468164324e-06, "loss": 0.0081, "step": 30217 }, { "epoch": 5.54, "learning_rate": 4.557501824871033e-06, "loss": 0.0071, "step": 30218 }, { "epoch": 5.54, "learning_rate": 4.5568789980980066e-06, "loss": 0.008, "step": 30219 }, { "epoch": 5.54, "learning_rate": 4.5562562013275906e-06, "loss": 0.0096, "step": 30220 }, { "epoch": 5.54, "learning_rate": 4.5556334345632185e-06, "loss": 0.0084, "step": 30221 }, { "epoch": 5.54, "learning_rate": 4.555010697808322e-06, "loss": 0.0102, "step": 30222 }, { "epoch": 5.54, "learning_rate": 4.554387991066334e-06, "loss": 0.0105, "step": 30223 }, { "epoch": 5.54, "learning_rate": 4.5537653143406865e-06, "loss": 0.0075, "step": 30224 }, { "epoch": 5.54, "learning_rate": 4.5531426676348135e-06, "loss": 0.0069, "step": 30225 }, { "epoch": 5.54, "learning_rate": 4.552520050952142e-06, "loss": 0.0084, "step": 30226 }, { "epoch": 5.54, "learning_rate": 4.55189746429611e-06, "loss": 0.0075, "step": 30227 }, { "epoch": 5.54, "learning_rate": 4.551274907670151e-06, "loss": 0.0088, "step": 30228 }, { "epoch": 5.54, "learning_rate": 4.550652381077689e-06, "loss": 0.0066, "step": 30229 }, { "epoch": 5.54, "learning_rate": 4.5500298845221594e-06, "loss": 0.01, "step": 30230 }, { "epoch": 5.54, "learning_rate": 4.549407418006988e-06, "loss": 0.0098, "step": 30231 }, { "epoch": 5.54, "learning_rate": 4.5487849815356145e-06, "loss": 0.0083, "step": 30232 }, { "epoch": 5.54, "learning_rate": 4.5481625751114655e-06, "loss": 0.0093, "step": 30233 }, { "epoch": 5.54, "learning_rate": 4.547540198737976e-06, "loss": 0.0103, "step": 30234 }, { "epoch": 5.54, "learning_rate": 4.546917852418567e-06, "loss": 0.006, "step": 30235 }, { "epoch": 5.54, "learning_rate": 4.5462955361566764e-06, "loss": 0.008, "step": 30236 }, { "epoch": 5.55, "learning_rate": 4.545673249955732e-06, "loss": 0.0074, "step": 30237 }, { "epoch": 5.55, "learning_rate": 4.545050993819166e-06, "loss": 0.0059, "step": 30238 }, { "epoch": 5.55, "learning_rate": 4.544428767750406e-06, "loss": 0.008, "step": 30239 }, { "epoch": 5.55, "learning_rate": 4.5438065717528825e-06, "loss": 0.0098, "step": 30240 }, { "epoch": 5.55, "learning_rate": 4.5431844058300246e-06, "loss": 0.0072, "step": 30241 }, { "epoch": 5.55, "learning_rate": 4.542562269985263e-06, "loss": 0.0103, "step": 30242 }, { "epoch": 5.55, "learning_rate": 4.541940164222025e-06, "loss": 0.0062, "step": 30243 }, { "epoch": 5.55, "learning_rate": 4.541318088543738e-06, "loss": 0.009, "step": 30244 }, { "epoch": 5.55, "learning_rate": 4.540696042953837e-06, "loss": 0.0076, "step": 30245 }, { "epoch": 5.55, "learning_rate": 4.540074027455751e-06, "loss": 0.0103, "step": 30246 }, { "epoch": 5.55, "learning_rate": 4.539452042052901e-06, "loss": 0.011, "step": 30247 }, { "epoch": 5.55, "learning_rate": 4.538830086748719e-06, "loss": 0.0088, "step": 30248 }, { "epoch": 5.55, "learning_rate": 4.53820816154663e-06, "loss": 0.0078, "step": 30249 }, { "epoch": 5.55, "learning_rate": 4.537586266450069e-06, "loss": 0.0114, "step": 30250 }, { "epoch": 5.55, "learning_rate": 4.536964401462461e-06, "loss": 0.0092, "step": 30251 }, { "epoch": 5.55, "learning_rate": 4.5363425665872375e-06, "loss": 0.0067, "step": 30252 }, { "epoch": 5.55, "learning_rate": 4.535720761827813e-06, "loss": 0.0082, "step": 30253 }, { "epoch": 5.55, "learning_rate": 4.535098987187629e-06, "loss": 0.0093, "step": 30254 }, { "epoch": 5.55, "learning_rate": 4.534477242670105e-06, "loss": 0.0071, "step": 30255 }, { "epoch": 5.55, "learning_rate": 4.5338555282786725e-06, "loss": 0.0066, "step": 30256 }, { "epoch": 5.55, "learning_rate": 4.533233844016756e-06, "loss": 0.0063, "step": 30257 }, { "epoch": 5.55, "learning_rate": 4.532612189887782e-06, "loss": 0.0071, "step": 30258 }, { "epoch": 5.55, "learning_rate": 4.531990565895178e-06, "loss": 0.0094, "step": 30259 }, { "epoch": 5.55, "learning_rate": 4.53136897204237e-06, "loss": 0.0076, "step": 30260 }, { "epoch": 5.55, "learning_rate": 4.530747408332785e-06, "loss": 0.0083, "step": 30261 }, { "epoch": 5.55, "learning_rate": 4.530125874769844e-06, "loss": 0.0062, "step": 30262 }, { "epoch": 5.55, "learning_rate": 4.529504371356985e-06, "loss": 0.0081, "step": 30263 }, { "epoch": 5.55, "learning_rate": 4.528882898097622e-06, "loss": 0.0073, "step": 30264 }, { "epoch": 5.55, "learning_rate": 4.528261454995185e-06, "loss": 0.0079, "step": 30265 }, { "epoch": 5.55, "learning_rate": 4.527640042053095e-06, "loss": 0.0086, "step": 30266 }, { "epoch": 5.55, "learning_rate": 4.527018659274785e-06, "loss": 0.0087, "step": 30267 }, { "epoch": 5.55, "learning_rate": 4.526397306663678e-06, "loss": 0.0086, "step": 30268 }, { "epoch": 5.55, "learning_rate": 4.525775984223195e-06, "loss": 0.0111, "step": 30269 }, { "epoch": 5.55, "learning_rate": 4.525154691956768e-06, "loss": 0.007, "step": 30270 }, { "epoch": 5.55, "learning_rate": 4.524533429867809e-06, "loss": 0.0074, "step": 30271 }, { "epoch": 5.55, "learning_rate": 4.523912197959754e-06, "loss": 0.0066, "step": 30272 }, { "epoch": 5.55, "learning_rate": 4.523290996236023e-06, "loss": 0.0079, "step": 30273 }, { "epoch": 5.55, "learning_rate": 4.52266982470004e-06, "loss": 0.0081, "step": 30274 }, { "epoch": 5.55, "learning_rate": 4.522048683355229e-06, "loss": 0.0064, "step": 30275 }, { "epoch": 5.55, "learning_rate": 4.521427572205015e-06, "loss": 0.0088, "step": 30276 }, { "epoch": 5.55, "learning_rate": 4.520806491252819e-06, "loss": 0.0093, "step": 30277 }, { "epoch": 5.55, "learning_rate": 4.520185440502067e-06, "loss": 0.0083, "step": 30278 }, { "epoch": 5.55, "learning_rate": 4.519564419956179e-06, "loss": 0.0103, "step": 30279 }, { "epoch": 5.55, "learning_rate": 4.518943429618582e-06, "loss": 0.0097, "step": 30280 }, { "epoch": 5.55, "learning_rate": 4.518322469492701e-06, "loss": 0.0104, "step": 30281 }, { "epoch": 5.55, "learning_rate": 4.517701539581951e-06, "loss": 0.0089, "step": 30282 }, { "epoch": 5.55, "learning_rate": 4.517080639889761e-06, "loss": 0.0109, "step": 30283 }, { "epoch": 5.55, "learning_rate": 4.516459770419544e-06, "loss": 0.006, "step": 30284 }, { "epoch": 5.55, "learning_rate": 4.515838931174735e-06, "loss": 0.0086, "step": 30285 }, { "epoch": 5.55, "learning_rate": 4.51521812215875e-06, "loss": 0.0106, "step": 30286 }, { "epoch": 5.55, "learning_rate": 4.5145973433750104e-06, "loss": 0.0086, "step": 30287 }, { "epoch": 5.55, "learning_rate": 4.5139765948269395e-06, "loss": 0.008, "step": 30288 }, { "epoch": 5.55, "learning_rate": 4.5133558765179576e-06, "loss": 0.01, "step": 30289 }, { "epoch": 5.55, "learning_rate": 4.512735188451487e-06, "loss": 0.0089, "step": 30290 }, { "epoch": 5.55, "learning_rate": 4.5121145306309484e-06, "loss": 0.0083, "step": 30291 }, { "epoch": 5.56, "learning_rate": 4.511493903059764e-06, "loss": 0.0088, "step": 30292 }, { "epoch": 5.56, "learning_rate": 4.510873305741352e-06, "loss": 0.0104, "step": 30293 }, { "epoch": 5.56, "learning_rate": 4.510252738679136e-06, "loss": 0.01, "step": 30294 }, { "epoch": 5.56, "learning_rate": 4.509632201876536e-06, "loss": 0.0072, "step": 30295 }, { "epoch": 5.56, "learning_rate": 4.509011695336971e-06, "loss": 0.0116, "step": 30296 }, { "epoch": 5.56, "learning_rate": 4.508391219063859e-06, "loss": 0.0103, "step": 30297 }, { "epoch": 5.56, "learning_rate": 4.507770773060626e-06, "loss": 0.0079, "step": 30298 }, { "epoch": 5.56, "learning_rate": 4.507150357330695e-06, "loss": 0.0079, "step": 30299 }, { "epoch": 5.56, "learning_rate": 4.506529971877475e-06, "loss": 0.0069, "step": 30300 }, { "epoch": 5.56, "learning_rate": 4.505909616704387e-06, "loss": 0.0075, "step": 30301 }, { "epoch": 5.56, "learning_rate": 4.505289291814857e-06, "loss": 0.0078, "step": 30302 }, { "epoch": 5.56, "learning_rate": 4.5046689972123015e-06, "loss": 0.0079, "step": 30303 }, { "epoch": 5.56, "learning_rate": 4.504048732900139e-06, "loss": 0.0123, "step": 30304 }, { "epoch": 5.56, "learning_rate": 4.503428498881789e-06, "loss": 0.0078, "step": 30305 }, { "epoch": 5.56, "learning_rate": 4.50280829516067e-06, "loss": 0.0087, "step": 30306 }, { "epoch": 5.56, "learning_rate": 4.5021881217401995e-06, "loss": 0.0058, "step": 30307 }, { "epoch": 5.56, "learning_rate": 4.501567978623797e-06, "loss": 0.0074, "step": 30308 }, { "epoch": 5.56, "learning_rate": 4.500947865814882e-06, "loss": 0.0072, "step": 30309 }, { "epoch": 5.56, "learning_rate": 4.50032778331687e-06, "loss": 0.0095, "step": 30310 }, { "epoch": 5.56, "learning_rate": 4.49970773113318e-06, "loss": 0.0079, "step": 30311 }, { "epoch": 5.56, "learning_rate": 4.499087709267231e-06, "loss": 0.008, "step": 30312 }, { "epoch": 5.56, "learning_rate": 4.498467717722439e-06, "loss": 0.0106, "step": 30313 }, { "epoch": 5.56, "learning_rate": 4.4978477565022215e-06, "loss": 0.0083, "step": 30314 }, { "epoch": 5.56, "learning_rate": 4.497227825609992e-06, "loss": 0.0105, "step": 30315 }, { "epoch": 5.56, "learning_rate": 4.4966079250491755e-06, "loss": 0.0075, "step": 30316 }, { "epoch": 5.56, "learning_rate": 4.49598805482319e-06, "loss": 0.0087, "step": 30317 }, { "epoch": 5.56, "learning_rate": 4.495368214935443e-06, "loss": 0.0069, "step": 30318 }, { "epoch": 5.56, "learning_rate": 4.494748405389351e-06, "loss": 0.0082, "step": 30319 }, { "epoch": 5.56, "learning_rate": 4.494128626188339e-06, "loss": 0.0093, "step": 30320 }, { "epoch": 5.56, "learning_rate": 4.493508877335818e-06, "loss": 0.0088, "step": 30321 }, { "epoch": 5.56, "learning_rate": 4.4928891588352055e-06, "loss": 0.0081, "step": 30322 }, { "epoch": 5.56, "learning_rate": 4.492269470689922e-06, "loss": 0.0072, "step": 30323 }, { "epoch": 5.56, "learning_rate": 4.491649812903369e-06, "loss": 0.0094, "step": 30324 }, { "epoch": 5.56, "learning_rate": 4.491030185478976e-06, "loss": 0.0084, "step": 30325 }, { "epoch": 5.56, "learning_rate": 4.490410588420153e-06, "loss": 0.008, "step": 30326 }, { "epoch": 5.56, "learning_rate": 4.489791021730316e-06, "loss": 0.0084, "step": 30327 }, { "epoch": 5.56, "learning_rate": 4.489171485412881e-06, "loss": 0.0085, "step": 30328 }, { "epoch": 5.56, "learning_rate": 4.48855197947126e-06, "loss": 0.0063, "step": 30329 }, { "epoch": 5.56, "learning_rate": 4.487932503908872e-06, "loss": 0.0078, "step": 30330 }, { "epoch": 5.56, "learning_rate": 4.4873130587291284e-06, "loss": 0.0108, "step": 30331 }, { "epoch": 5.56, "learning_rate": 4.48669364393544e-06, "loss": 0.0105, "step": 30332 }, { "epoch": 5.56, "learning_rate": 4.486074259531229e-06, "loss": 0.0096, "step": 30333 }, { "epoch": 5.56, "learning_rate": 4.485454905519906e-06, "loss": 0.0102, "step": 30334 }, { "epoch": 5.56, "learning_rate": 4.4848355819048896e-06, "loss": 0.0061, "step": 30335 }, { "epoch": 5.56, "learning_rate": 4.484216288689583e-06, "loss": 0.0087, "step": 30336 }, { "epoch": 5.56, "learning_rate": 4.483597025877403e-06, "loss": 0.0072, "step": 30337 }, { "epoch": 5.56, "learning_rate": 4.482977793471769e-06, "loss": 0.0115, "step": 30338 }, { "epoch": 5.56, "learning_rate": 4.48235859147609e-06, "loss": 0.0067, "step": 30339 }, { "epoch": 5.56, "learning_rate": 4.481739419893781e-06, "loss": 0.0074, "step": 30340 }, { "epoch": 5.56, "learning_rate": 4.481120278728253e-06, "loss": 0.0068, "step": 30341 }, { "epoch": 5.56, "learning_rate": 4.480501167982919e-06, "loss": 0.0073, "step": 30342 }, { "epoch": 5.56, "learning_rate": 4.479882087661192e-06, "loss": 0.0066, "step": 30343 }, { "epoch": 5.56, "learning_rate": 4.4792630377664845e-06, "loss": 0.0081, "step": 30344 }, { "epoch": 5.56, "learning_rate": 4.478644018302209e-06, "loss": 0.0095, "step": 30345 }, { "epoch": 5.57, "learning_rate": 4.478025029271775e-06, "loss": 0.0123, "step": 30346 }, { "epoch": 5.57, "learning_rate": 4.477406070678599e-06, "loss": 0.0065, "step": 30347 }, { "epoch": 5.57, "learning_rate": 4.476787142526088e-06, "loss": 0.0094, "step": 30348 }, { "epoch": 5.57, "learning_rate": 4.476168244817657e-06, "loss": 0.0069, "step": 30349 }, { "epoch": 5.57, "learning_rate": 4.475549377556711e-06, "loss": 0.0086, "step": 30350 }, { "epoch": 5.57, "learning_rate": 4.474930540746672e-06, "loss": 0.0057, "step": 30351 }, { "epoch": 5.57, "learning_rate": 4.474311734390944e-06, "loss": 0.0074, "step": 30352 }, { "epoch": 5.57, "learning_rate": 4.473692958492942e-06, "loss": 0.0065, "step": 30353 }, { "epoch": 5.57, "learning_rate": 4.473074213056068e-06, "loss": 0.0115, "step": 30354 }, { "epoch": 5.57, "learning_rate": 4.472455498083742e-06, "loss": 0.0082, "step": 30355 }, { "epoch": 5.57, "learning_rate": 4.47183681357937e-06, "loss": 0.0089, "step": 30356 }, { "epoch": 5.57, "learning_rate": 4.471218159546363e-06, "loss": 0.0073, "step": 30357 }, { "epoch": 5.57, "learning_rate": 4.470599535988131e-06, "loss": 0.0082, "step": 30358 }, { "epoch": 5.57, "learning_rate": 4.4699809429080855e-06, "loss": 0.0078, "step": 30359 }, { "epoch": 5.57, "learning_rate": 4.469362380309633e-06, "loss": 0.0087, "step": 30360 }, { "epoch": 5.57, "learning_rate": 4.468743848196185e-06, "loss": 0.0085, "step": 30361 }, { "epoch": 5.57, "learning_rate": 4.468125346571151e-06, "loss": 0.0086, "step": 30362 }, { "epoch": 5.57, "learning_rate": 4.467506875437934e-06, "loss": 0.0091, "step": 30363 }, { "epoch": 5.57, "learning_rate": 4.466888434799959e-06, "loss": 0.0073, "step": 30364 }, { "epoch": 5.57, "learning_rate": 4.466270024660618e-06, "loss": 0.0069, "step": 30365 }, { "epoch": 5.57, "learning_rate": 4.465651645023329e-06, "loss": 0.0063, "step": 30366 }, { "epoch": 5.57, "learning_rate": 4.465033295891497e-06, "loss": 0.0067, "step": 30367 }, { "epoch": 5.57, "learning_rate": 4.464414977268526e-06, "loss": 0.0085, "step": 30368 }, { "epoch": 5.57, "learning_rate": 4.463796689157834e-06, "loss": 0.0085, "step": 30369 }, { "epoch": 5.57, "learning_rate": 4.463178431562828e-06, "loss": 0.0078, "step": 30370 }, { "epoch": 5.57, "learning_rate": 4.462560204486909e-06, "loss": 0.0081, "step": 30371 }, { "epoch": 5.57, "learning_rate": 4.461942007933483e-06, "loss": 0.0065, "step": 30372 }, { "epoch": 5.57, "learning_rate": 4.4613238419059665e-06, "loss": 0.0075, "step": 30373 }, { "epoch": 5.57, "learning_rate": 4.460705706407764e-06, "loss": 0.0084, "step": 30374 }, { "epoch": 5.57, "learning_rate": 4.46008760144228e-06, "loss": 0.0084, "step": 30375 }, { "epoch": 5.57, "learning_rate": 4.459469527012924e-06, "loss": 0.0074, "step": 30376 }, { "epoch": 5.57, "learning_rate": 4.458851483123101e-06, "loss": 0.0084, "step": 30377 }, { "epoch": 5.57, "learning_rate": 4.4582334697762186e-06, "loss": 0.0084, "step": 30378 }, { "epoch": 5.57, "learning_rate": 4.457615486975683e-06, "loss": 0.0081, "step": 30379 }, { "epoch": 5.57, "learning_rate": 4.4569975347249005e-06, "loss": 0.0091, "step": 30380 }, { "epoch": 5.57, "learning_rate": 4.456379613027275e-06, "loss": 0.0086, "step": 30381 }, { "epoch": 5.57, "learning_rate": 4.455761721886221e-06, "loss": 0.0102, "step": 30382 }, { "epoch": 5.57, "learning_rate": 4.455143861305136e-06, "loss": 0.0047, "step": 30383 }, { "epoch": 5.57, "learning_rate": 4.454526031287426e-06, "loss": 0.0077, "step": 30384 }, { "epoch": 5.57, "learning_rate": 4.453908231836496e-06, "loss": 0.0092, "step": 30385 }, { "epoch": 5.57, "learning_rate": 4.453290462955757e-06, "loss": 0.0075, "step": 30386 }, { "epoch": 5.57, "learning_rate": 4.4526727246486116e-06, "loss": 0.0071, "step": 30387 }, { "epoch": 5.57, "learning_rate": 4.452055016918466e-06, "loss": 0.0095, "step": 30388 }, { "epoch": 5.57, "learning_rate": 4.451437339768722e-06, "loss": 0.0072, "step": 30389 }, { "epoch": 5.57, "learning_rate": 4.4508196932027805e-06, "loss": 0.01, "step": 30390 }, { "epoch": 5.57, "learning_rate": 4.450202077224054e-06, "loss": 0.0109, "step": 30391 }, { "epoch": 5.57, "learning_rate": 4.449584491835943e-06, "loss": 0.0078, "step": 30392 }, { "epoch": 5.57, "learning_rate": 4.448966937041854e-06, "loss": 0.0072, "step": 30393 }, { "epoch": 5.57, "learning_rate": 4.4483494128451885e-06, "loss": 0.0079, "step": 30394 }, { "epoch": 5.57, "learning_rate": 4.447731919249352e-06, "loss": 0.0091, "step": 30395 }, { "epoch": 5.57, "learning_rate": 4.447114456257746e-06, "loss": 0.009, "step": 30396 }, { "epoch": 5.57, "learning_rate": 4.446497023873776e-06, "loss": 0.0092, "step": 30397 }, { "epoch": 5.57, "learning_rate": 4.4458796221008436e-06, "loss": 0.0085, "step": 30398 }, { "epoch": 5.57, "learning_rate": 4.44526225094235e-06, "loss": 0.0081, "step": 30399 }, { "epoch": 5.57, "learning_rate": 4.444644910401709e-06, "loss": 0.0061, "step": 30400 }, { "epoch": 5.58, "learning_rate": 4.44402760048231e-06, "loss": 0.0067, "step": 30401 }, { "epoch": 5.58, "learning_rate": 4.443410321187562e-06, "loss": 0.0085, "step": 30402 }, { "epoch": 5.58, "learning_rate": 4.442793072520864e-06, "loss": 0.0061, "step": 30403 }, { "epoch": 5.58, "learning_rate": 4.442175854485622e-06, "loss": 0.0098, "step": 30404 }, { "epoch": 5.58, "learning_rate": 4.441558667085239e-06, "loss": 0.0088, "step": 30405 }, { "epoch": 5.58, "learning_rate": 4.440941510323117e-06, "loss": 0.0132, "step": 30406 }, { "epoch": 5.58, "learning_rate": 4.440324384202649e-06, "loss": 0.0074, "step": 30407 }, { "epoch": 5.58, "learning_rate": 4.439707288727247e-06, "loss": 0.0113, "step": 30408 }, { "epoch": 5.58, "learning_rate": 4.439090223900307e-06, "loss": 0.0095, "step": 30409 }, { "epoch": 5.58, "learning_rate": 4.438473189725234e-06, "loss": 0.0078, "step": 30410 }, { "epoch": 5.58, "learning_rate": 4.437856186205427e-06, "loss": 0.0064, "step": 30411 }, { "epoch": 5.58, "learning_rate": 4.437239213344286e-06, "loss": 0.0068, "step": 30412 }, { "epoch": 5.58, "learning_rate": 4.436622271145214e-06, "loss": 0.0059, "step": 30413 }, { "epoch": 5.58, "learning_rate": 4.4360053596116095e-06, "loss": 0.0126, "step": 30414 }, { "epoch": 5.58, "learning_rate": 4.435388478746874e-06, "loss": 0.0069, "step": 30415 }, { "epoch": 5.58, "learning_rate": 4.434771628554404e-06, "loss": 0.0074, "step": 30416 }, { "epoch": 5.58, "learning_rate": 4.4341548090376105e-06, "loss": 0.0067, "step": 30417 }, { "epoch": 5.58, "learning_rate": 4.433538020199882e-06, "loss": 0.0085, "step": 30418 }, { "epoch": 5.58, "learning_rate": 4.432921262044622e-06, "loss": 0.0101, "step": 30419 }, { "epoch": 5.58, "learning_rate": 4.4323045345752315e-06, "loss": 0.0087, "step": 30420 }, { "epoch": 5.58, "learning_rate": 4.431687837795106e-06, "loss": 0.0074, "step": 30421 }, { "epoch": 5.58, "learning_rate": 4.431071171707649e-06, "loss": 0.0091, "step": 30422 }, { "epoch": 5.58, "learning_rate": 4.43045453631626e-06, "loss": 0.0081, "step": 30423 }, { "epoch": 5.58, "learning_rate": 4.429837931624338e-06, "loss": 0.0081, "step": 30424 }, { "epoch": 5.58, "learning_rate": 4.429221357635274e-06, "loss": 0.0074, "step": 30425 }, { "epoch": 5.58, "learning_rate": 4.428604814352475e-06, "loss": 0.0077, "step": 30426 }, { "epoch": 5.58, "learning_rate": 4.427988301779338e-06, "loss": 0.0082, "step": 30427 }, { "epoch": 5.58, "learning_rate": 4.427371819919258e-06, "loss": 0.0076, "step": 30428 }, { "epoch": 5.58, "learning_rate": 4.426755368775637e-06, "loss": 0.0093, "step": 30429 }, { "epoch": 5.58, "learning_rate": 4.42613894835187e-06, "loss": 0.0096, "step": 30430 }, { "epoch": 5.58, "learning_rate": 4.425522558651356e-06, "loss": 0.0076, "step": 30431 }, { "epoch": 5.58, "learning_rate": 4.4249061996774924e-06, "loss": 0.0063, "step": 30432 }, { "epoch": 5.58, "learning_rate": 4.424289871433677e-06, "loss": 0.0098, "step": 30433 }, { "epoch": 5.58, "learning_rate": 4.423673573923302e-06, "loss": 0.0076, "step": 30434 }, { "epoch": 5.58, "learning_rate": 4.423057307149776e-06, "loss": 0.0092, "step": 30435 }, { "epoch": 5.58, "learning_rate": 4.4224410711164855e-06, "loss": 0.0072, "step": 30436 }, { "epoch": 5.58, "learning_rate": 4.42182486582683e-06, "loss": 0.0064, "step": 30437 }, { "epoch": 5.58, "learning_rate": 4.421208691284204e-06, "loss": 0.0096, "step": 30438 }, { "epoch": 5.58, "learning_rate": 4.420592547492008e-06, "loss": 0.0071, "step": 30439 }, { "epoch": 5.58, "learning_rate": 4.4199764344536385e-06, "loss": 0.0071, "step": 30440 }, { "epoch": 5.58, "learning_rate": 4.4193603521724895e-06, "loss": 0.0118, "step": 30441 }, { "epoch": 5.58, "learning_rate": 4.41874430065196e-06, "loss": 0.0081, "step": 30442 }, { "epoch": 5.58, "learning_rate": 4.4181282798954365e-06, "loss": 0.0068, "step": 30443 }, { "epoch": 5.58, "learning_rate": 4.417512289906322e-06, "loss": 0.0092, "step": 30444 }, { "epoch": 5.58, "learning_rate": 4.416896330688012e-06, "loss": 0.0065, "step": 30445 }, { "epoch": 5.58, "learning_rate": 4.416280402243901e-06, "loss": 0.0093, "step": 30446 }, { "epoch": 5.58, "learning_rate": 4.415664504577382e-06, "loss": 0.01, "step": 30447 }, { "epoch": 5.58, "learning_rate": 4.415048637691851e-06, "loss": 0.0084, "step": 30448 }, { "epoch": 5.58, "learning_rate": 4.414432801590703e-06, "loss": 0.0091, "step": 30449 }, { "epoch": 5.58, "learning_rate": 4.413816996277333e-06, "loss": 0.0081, "step": 30450 }, { "epoch": 5.58, "learning_rate": 4.4132012217551346e-06, "loss": 0.0086, "step": 30451 }, { "epoch": 5.58, "learning_rate": 4.412585478027497e-06, "loss": 0.0092, "step": 30452 }, { "epoch": 5.58, "learning_rate": 4.411969765097828e-06, "loss": 0.0064, "step": 30453 }, { "epoch": 5.58, "learning_rate": 4.411354082969508e-06, "loss": 0.0081, "step": 30454 }, { "epoch": 5.58, "learning_rate": 4.410738431645936e-06, "loss": 0.0081, "step": 30455 }, { "epoch": 5.59, "learning_rate": 4.410122811130499e-06, "loss": 0.0109, "step": 30456 }, { "epoch": 5.59, "learning_rate": 4.409507221426601e-06, "loss": 0.0071, "step": 30457 }, { "epoch": 5.59, "learning_rate": 4.408891662537631e-06, "loss": 0.0093, "step": 30458 }, { "epoch": 5.59, "learning_rate": 4.40827613446698e-06, "loss": 0.0089, "step": 30459 }, { "epoch": 5.59, "learning_rate": 4.407660637218042e-06, "loss": 0.0074, "step": 30460 }, { "epoch": 5.59, "learning_rate": 4.407045170794211e-06, "loss": 0.0085, "step": 30461 }, { "epoch": 5.59, "learning_rate": 4.406429735198876e-06, "loss": 0.0059, "step": 30462 }, { "epoch": 5.59, "learning_rate": 4.405814330435433e-06, "loss": 0.0074, "step": 30463 }, { "epoch": 5.59, "learning_rate": 4.405198956507272e-06, "loss": 0.0119, "step": 30464 }, { "epoch": 5.59, "learning_rate": 4.4045836134177845e-06, "loss": 0.0071, "step": 30465 }, { "epoch": 5.59, "learning_rate": 4.4039683011703635e-06, "loss": 0.0085, "step": 30466 }, { "epoch": 5.59, "learning_rate": 4.403353019768401e-06, "loss": 0.0094, "step": 30467 }, { "epoch": 5.59, "learning_rate": 4.402737769215287e-06, "loss": 0.0088, "step": 30468 }, { "epoch": 5.59, "learning_rate": 4.402122549514412e-06, "loss": 0.0092, "step": 30469 }, { "epoch": 5.59, "learning_rate": 4.401507360669169e-06, "loss": 0.0055, "step": 30470 }, { "epoch": 5.59, "learning_rate": 4.400892202682955e-06, "loss": 0.0079, "step": 30471 }, { "epoch": 5.59, "learning_rate": 4.400277075559149e-06, "loss": 0.0094, "step": 30472 }, { "epoch": 5.59, "learning_rate": 4.399661979301147e-06, "loss": 0.0063, "step": 30473 }, { "epoch": 5.59, "learning_rate": 4.3990469139123375e-06, "loss": 0.008, "step": 30474 }, { "epoch": 5.59, "learning_rate": 4.398431879396114e-06, "loss": 0.0074, "step": 30475 }, { "epoch": 5.59, "learning_rate": 4.397816875755867e-06, "loss": 0.0073, "step": 30476 }, { "epoch": 5.59, "learning_rate": 4.3972019029949845e-06, "loss": 0.0091, "step": 30477 }, { "epoch": 5.59, "learning_rate": 4.396586961116856e-06, "loss": 0.0104, "step": 30478 }, { "epoch": 5.59, "learning_rate": 4.39597205012487e-06, "loss": 0.0069, "step": 30479 }, { "epoch": 5.59, "learning_rate": 4.39535717002242e-06, "loss": 0.0058, "step": 30480 }, { "epoch": 5.59, "learning_rate": 4.3947423208128906e-06, "loss": 0.0085, "step": 30481 }, { "epoch": 5.59, "learning_rate": 4.394127502499673e-06, "loss": 0.0093, "step": 30482 }, { "epoch": 5.59, "learning_rate": 4.393512715086157e-06, "loss": 0.0072, "step": 30483 }, { "epoch": 5.59, "learning_rate": 4.3928979585757295e-06, "loss": 0.0051, "step": 30484 }, { "epoch": 5.59, "learning_rate": 4.392283232971779e-06, "loss": 0.0098, "step": 30485 }, { "epoch": 5.59, "learning_rate": 4.391668538277696e-06, "loss": 0.0057, "step": 30486 }, { "epoch": 5.59, "learning_rate": 4.391053874496863e-06, "loss": 0.0091, "step": 30487 }, { "epoch": 5.59, "learning_rate": 4.390439241632677e-06, "loss": 0.0087, "step": 30488 }, { "epoch": 5.59, "learning_rate": 4.389824639688524e-06, "loss": 0.0095, "step": 30489 }, { "epoch": 5.59, "learning_rate": 4.389210068667786e-06, "loss": 0.0067, "step": 30490 }, { "epoch": 5.59, "learning_rate": 4.38859552857385e-06, "loss": 0.0065, "step": 30491 }, { "epoch": 5.59, "learning_rate": 4.3879810194101105e-06, "loss": 0.0076, "step": 30492 }, { "epoch": 5.59, "learning_rate": 4.387366541179951e-06, "loss": 0.0076, "step": 30493 }, { "epoch": 5.59, "learning_rate": 4.386752093886758e-06, "loss": 0.0085, "step": 30494 }, { "epoch": 5.59, "learning_rate": 4.3861376775339226e-06, "loss": 0.0068, "step": 30495 }, { "epoch": 5.59, "learning_rate": 4.385523292124821e-06, "loss": 0.0064, "step": 30496 }, { "epoch": 5.59, "learning_rate": 4.3849089376628506e-06, "loss": 0.0075, "step": 30497 }, { "epoch": 5.59, "learning_rate": 4.384294614151393e-06, "loss": 0.0077, "step": 30498 }, { "epoch": 5.59, "learning_rate": 4.383680321593836e-06, "loss": 0.0048, "step": 30499 }, { "epoch": 5.59, "learning_rate": 4.383066059993564e-06, "loss": 0.0093, "step": 30500 }, { "epoch": 5.59, "learning_rate": 4.382451829353963e-06, "loss": 0.0105, "step": 30501 }, { "epoch": 5.59, "learning_rate": 4.38183762967842e-06, "loss": 0.0059, "step": 30502 }, { "epoch": 5.59, "learning_rate": 4.381223460970318e-06, "loss": 0.0108, "step": 30503 }, { "epoch": 5.59, "learning_rate": 4.380609323233046e-06, "loss": 0.0077, "step": 30504 }, { "epoch": 5.59, "learning_rate": 4.379995216469982e-06, "loss": 0.0056, "step": 30505 }, { "epoch": 5.59, "learning_rate": 4.379381140684521e-06, "loss": 0.0075, "step": 30506 }, { "epoch": 5.59, "learning_rate": 4.378767095880045e-06, "loss": 0.0078, "step": 30507 }, { "epoch": 5.59, "learning_rate": 4.378153082059934e-06, "loss": 0.009, "step": 30508 }, { "epoch": 5.59, "learning_rate": 4.377539099227571e-06, "loss": 0.007, "step": 30509 }, { "epoch": 5.6, "learning_rate": 4.376925147386347e-06, "loss": 0.0081, "step": 30510 }, { "epoch": 5.6, "learning_rate": 4.3763112265396445e-06, "loss": 0.0052, "step": 30511 }, { "epoch": 5.6, "learning_rate": 4.375697336690845e-06, "loss": 0.0079, "step": 30512 }, { "epoch": 5.6, "learning_rate": 4.375083477843335e-06, "loss": 0.0078, "step": 30513 }, { "epoch": 5.6, "learning_rate": 4.3744696500004955e-06, "loss": 0.0095, "step": 30514 }, { "epoch": 5.6, "learning_rate": 4.373855853165711e-06, "loss": 0.0076, "step": 30515 }, { "epoch": 5.6, "learning_rate": 4.373242087342366e-06, "loss": 0.0072, "step": 30516 }, { "epoch": 5.6, "learning_rate": 4.372628352533842e-06, "loss": 0.0088, "step": 30517 }, { "epoch": 5.6, "learning_rate": 4.3720146487435226e-06, "loss": 0.0104, "step": 30518 }, { "epoch": 5.6, "learning_rate": 4.371400975974789e-06, "loss": 0.0089, "step": 30519 }, { "epoch": 5.6, "learning_rate": 4.370787334231026e-06, "loss": 0.0082, "step": 30520 }, { "epoch": 5.6, "learning_rate": 4.370173723515615e-06, "loss": 0.0099, "step": 30521 }, { "epoch": 5.6, "learning_rate": 4.369560143831935e-06, "loss": 0.0087, "step": 30522 }, { "epoch": 5.6, "learning_rate": 4.3689465951833745e-06, "loss": 0.0069, "step": 30523 }, { "epoch": 5.6, "learning_rate": 4.368333077573316e-06, "loss": 0.0082, "step": 30524 }, { "epoch": 5.6, "learning_rate": 4.367719591005134e-06, "loss": 0.0086, "step": 30525 }, { "epoch": 5.6, "learning_rate": 4.367106135482214e-06, "loss": 0.0115, "step": 30526 }, { "epoch": 5.6, "learning_rate": 4.3664927110079324e-06, "loss": 0.0074, "step": 30527 }, { "epoch": 5.6, "learning_rate": 4.3658793175856795e-06, "loss": 0.008, "step": 30528 }, { "epoch": 5.6, "learning_rate": 4.365265955218832e-06, "loss": 0.008, "step": 30529 }, { "epoch": 5.6, "learning_rate": 4.364652623910769e-06, "loss": 0.007, "step": 30530 }, { "epoch": 5.6, "learning_rate": 4.364039323664874e-06, "loss": 0.0087, "step": 30531 }, { "epoch": 5.6, "learning_rate": 4.363426054484525e-06, "loss": 0.0077, "step": 30532 }, { "epoch": 5.6, "learning_rate": 4.362812816373105e-06, "loss": 0.0082, "step": 30533 }, { "epoch": 5.6, "learning_rate": 4.362199609333993e-06, "loss": 0.0084, "step": 30534 }, { "epoch": 5.6, "learning_rate": 4.361586433370565e-06, "loss": 0.0085, "step": 30535 }, { "epoch": 5.6, "learning_rate": 4.360973288486211e-06, "loss": 0.0093, "step": 30536 }, { "epoch": 5.6, "learning_rate": 4.360360174684301e-06, "loss": 0.0069, "step": 30537 }, { "epoch": 5.6, "learning_rate": 4.3597470919682186e-06, "loss": 0.0097, "step": 30538 }, { "epoch": 5.6, "learning_rate": 4.359134040341342e-06, "loss": 0.0076, "step": 30539 }, { "epoch": 5.6, "learning_rate": 4.358521019807047e-06, "loss": 0.0061, "step": 30540 }, { "epoch": 5.6, "learning_rate": 4.357908030368719e-06, "loss": 0.0078, "step": 30541 }, { "epoch": 5.6, "learning_rate": 4.357295072029739e-06, "loss": 0.0056, "step": 30542 }, { "epoch": 5.6, "learning_rate": 4.356682144793477e-06, "loss": 0.0102, "step": 30543 }, { "epoch": 5.6, "learning_rate": 4.3560692486633114e-06, "loss": 0.0092, "step": 30544 }, { "epoch": 5.6, "learning_rate": 4.355456383642628e-06, "loss": 0.0068, "step": 30545 }, { "epoch": 5.6, "learning_rate": 4.354843549734801e-06, "loss": 0.0076, "step": 30546 }, { "epoch": 5.6, "learning_rate": 4.3542307469432086e-06, "loss": 0.0074, "step": 30547 }, { "epoch": 5.6, "learning_rate": 4.353617975271229e-06, "loss": 0.0078, "step": 30548 }, { "epoch": 5.6, "learning_rate": 4.353005234722239e-06, "loss": 0.0065, "step": 30549 }, { "epoch": 5.6, "learning_rate": 4.352392525299617e-06, "loss": 0.0096, "step": 30550 }, { "epoch": 5.6, "learning_rate": 4.35177984700674e-06, "loss": 0.0072, "step": 30551 }, { "epoch": 5.6, "learning_rate": 4.351167199846984e-06, "loss": 0.0087, "step": 30552 }, { "epoch": 5.6, "learning_rate": 4.350554583823723e-06, "loss": 0.0064, "step": 30553 }, { "epoch": 5.6, "learning_rate": 4.349941998940344e-06, "loss": 0.0071, "step": 30554 }, { "epoch": 5.6, "learning_rate": 4.349329445200216e-06, "loss": 0.0094, "step": 30555 }, { "epoch": 5.6, "learning_rate": 4.348716922606714e-06, "loss": 0.0071, "step": 30556 }, { "epoch": 5.6, "learning_rate": 4.348104431163215e-06, "loss": 0.0103, "step": 30557 }, { "epoch": 5.6, "learning_rate": 4.347491970873099e-06, "loss": 0.0104, "step": 30558 }, { "epoch": 5.6, "learning_rate": 4.34687954173974e-06, "loss": 0.0102, "step": 30559 }, { "epoch": 5.6, "learning_rate": 4.3462671437665175e-06, "loss": 0.011, "step": 30560 }, { "epoch": 5.6, "learning_rate": 4.345654776956798e-06, "loss": 0.0098, "step": 30561 }, { "epoch": 5.6, "learning_rate": 4.3450424413139596e-06, "loss": 0.0093, "step": 30562 }, { "epoch": 5.6, "learning_rate": 4.3444301368413836e-06, "loss": 0.006, "step": 30563 }, { "epoch": 5.6, "learning_rate": 4.34381786354244e-06, "loss": 0.0065, "step": 30564 }, { "epoch": 5.61, "learning_rate": 4.343205621420505e-06, "loss": 0.0087, "step": 30565 }, { "epoch": 5.61, "learning_rate": 4.342593410478953e-06, "loss": 0.0087, "step": 30566 }, { "epoch": 5.61, "learning_rate": 4.34198123072116e-06, "loss": 0.0101, "step": 30567 }, { "epoch": 5.61, "learning_rate": 4.341369082150498e-06, "loss": 0.0082, "step": 30568 }, { "epoch": 5.61, "learning_rate": 4.340756964770343e-06, "loss": 0.0094, "step": 30569 }, { "epoch": 5.61, "learning_rate": 4.340144878584067e-06, "loss": 0.0058, "step": 30570 }, { "epoch": 5.61, "learning_rate": 4.339532823595045e-06, "loss": 0.0072, "step": 30571 }, { "epoch": 5.61, "learning_rate": 4.338920799806652e-06, "loss": 0.007, "step": 30572 }, { "epoch": 5.61, "learning_rate": 4.338308807222259e-06, "loss": 0.007, "step": 30573 }, { "epoch": 5.61, "learning_rate": 4.337696845845242e-06, "loss": 0.0055, "step": 30574 }, { "epoch": 5.61, "learning_rate": 4.337084915678967e-06, "loss": 0.0097, "step": 30575 }, { "epoch": 5.61, "learning_rate": 4.336473016726817e-06, "loss": 0.0095, "step": 30576 }, { "epoch": 5.61, "learning_rate": 4.3358611489921606e-06, "loss": 0.0075, "step": 30577 }, { "epoch": 5.61, "learning_rate": 4.335249312478374e-06, "loss": 0.0097, "step": 30578 }, { "epoch": 5.61, "learning_rate": 4.334637507188823e-06, "loss": 0.0069, "step": 30579 }, { "epoch": 5.61, "learning_rate": 4.3340257331268775e-06, "loss": 0.0073, "step": 30580 }, { "epoch": 5.61, "learning_rate": 4.3334139902959194e-06, "loss": 0.0097, "step": 30581 }, { "epoch": 5.61, "learning_rate": 4.332802278699317e-06, "loss": 0.0093, "step": 30582 }, { "epoch": 5.61, "learning_rate": 4.33219059834044e-06, "loss": 0.0105, "step": 30583 }, { "epoch": 5.61, "learning_rate": 4.331578949222661e-06, "loss": 0.0083, "step": 30584 }, { "epoch": 5.61, "learning_rate": 4.330967331349353e-06, "loss": 0.0088, "step": 30585 }, { "epoch": 5.61, "learning_rate": 4.330355744723884e-06, "loss": 0.0087, "step": 30586 }, { "epoch": 5.61, "learning_rate": 4.3297441893496284e-06, "loss": 0.01, "step": 30587 }, { "epoch": 5.61, "learning_rate": 4.329132665229952e-06, "loss": 0.006, "step": 30588 }, { "epoch": 5.61, "learning_rate": 4.328521172368235e-06, "loss": 0.008, "step": 30589 }, { "epoch": 5.61, "learning_rate": 4.32790971076784e-06, "loss": 0.009, "step": 30590 }, { "epoch": 5.61, "learning_rate": 4.32729828043214e-06, "loss": 0.0097, "step": 30591 }, { "epoch": 5.61, "learning_rate": 4.3266868813645034e-06, "loss": 0.0072, "step": 30592 }, { "epoch": 5.61, "learning_rate": 4.3260755135682985e-06, "loss": 0.008, "step": 30593 }, { "epoch": 5.61, "learning_rate": 4.325464177046902e-06, "loss": 0.0111, "step": 30594 }, { "epoch": 5.61, "learning_rate": 4.32485287180368e-06, "loss": 0.007, "step": 30595 }, { "epoch": 5.61, "learning_rate": 4.324241597842005e-06, "loss": 0.0093, "step": 30596 }, { "epoch": 5.61, "learning_rate": 4.323630355165237e-06, "loss": 0.007, "step": 30597 }, { "epoch": 5.61, "learning_rate": 4.323019143776754e-06, "loss": 0.0065, "step": 30598 }, { "epoch": 5.61, "learning_rate": 4.322407963679923e-06, "loss": 0.0075, "step": 30599 }, { "epoch": 5.61, "learning_rate": 4.321796814878112e-06, "loss": 0.0076, "step": 30600 }, { "epoch": 5.61, "learning_rate": 4.3211856973746904e-06, "loss": 0.0076, "step": 30601 }, { "epoch": 5.61, "learning_rate": 4.320574611173026e-06, "loss": 0.0073, "step": 30602 }, { "epoch": 5.61, "learning_rate": 4.319963556276488e-06, "loss": 0.0058, "step": 30603 }, { "epoch": 5.61, "learning_rate": 4.319352532688444e-06, "loss": 0.0058, "step": 30604 }, { "epoch": 5.61, "learning_rate": 4.318741540412261e-06, "loss": 0.0072, "step": 30605 }, { "epoch": 5.61, "learning_rate": 4.318130579451305e-06, "loss": 0.0069, "step": 30606 }, { "epoch": 5.61, "learning_rate": 4.317519649808952e-06, "loss": 0.0081, "step": 30607 }, { "epoch": 5.61, "learning_rate": 4.316908751488562e-06, "loss": 0.0092, "step": 30608 }, { "epoch": 5.61, "learning_rate": 4.3162978844935034e-06, "loss": 0.0081, "step": 30609 }, { "epoch": 5.61, "learning_rate": 4.315687048827141e-06, "loss": 0.0052, "step": 30610 }, { "epoch": 5.61, "learning_rate": 4.315076244492847e-06, "loss": 0.0097, "step": 30611 }, { "epoch": 5.61, "learning_rate": 4.314465471493987e-06, "loss": 0.0088, "step": 30612 }, { "epoch": 5.61, "learning_rate": 4.313854729833926e-06, "loss": 0.0068, "step": 30613 }, { "epoch": 5.61, "learning_rate": 4.3132440195160355e-06, "loss": 0.0078, "step": 30614 }, { "epoch": 5.61, "learning_rate": 4.312633340543669e-06, "loss": 0.0064, "step": 30615 }, { "epoch": 5.61, "learning_rate": 4.312022692920205e-06, "loss": 0.0089, "step": 30616 }, { "epoch": 5.61, "learning_rate": 4.311412076649004e-06, "loss": 0.0085, "step": 30617 }, { "epoch": 5.61, "learning_rate": 4.310801491733434e-06, "loss": 0.0089, "step": 30618 }, { "epoch": 5.62, "learning_rate": 4.310190938176859e-06, "loss": 0.0059, "step": 30619 }, { "epoch": 5.62, "learning_rate": 4.309580415982646e-06, "loss": 0.0081, "step": 30620 }, { "epoch": 5.62, "learning_rate": 4.308969925154157e-06, "loss": 0.0066, "step": 30621 }, { "epoch": 5.62, "learning_rate": 4.30835946569476e-06, "loss": 0.0059, "step": 30622 }, { "epoch": 5.62, "learning_rate": 4.30774903760782e-06, "loss": 0.0082, "step": 30623 }, { "epoch": 5.62, "learning_rate": 4.307138640896696e-06, "loss": 0.0082, "step": 30624 }, { "epoch": 5.62, "learning_rate": 4.3065282755647655e-06, "loss": 0.0065, "step": 30625 }, { "epoch": 5.62, "learning_rate": 4.30591794161538e-06, "loss": 0.0066, "step": 30626 }, { "epoch": 5.62, "learning_rate": 4.305307639051907e-06, "loss": 0.0061, "step": 30627 }, { "epoch": 5.62, "learning_rate": 4.304697367877711e-06, "loss": 0.007, "step": 30628 }, { "epoch": 5.62, "learning_rate": 4.304087128096157e-06, "loss": 0.0072, "step": 30629 }, { "epoch": 5.62, "learning_rate": 4.30347691971061e-06, "loss": 0.0093, "step": 30630 }, { "epoch": 5.62, "learning_rate": 4.302866742724432e-06, "loss": 0.0069, "step": 30631 }, { "epoch": 5.62, "learning_rate": 4.302256597140988e-06, "loss": 0.009, "step": 30632 }, { "epoch": 5.62, "learning_rate": 4.301646482963633e-06, "loss": 0.0092, "step": 30633 }, { "epoch": 5.62, "learning_rate": 4.301036400195739e-06, "loss": 0.007, "step": 30634 }, { "epoch": 5.62, "learning_rate": 4.300426348840667e-06, "loss": 0.0075, "step": 30635 }, { "epoch": 5.62, "learning_rate": 4.299816328901778e-06, "loss": 0.0087, "step": 30636 }, { "epoch": 5.62, "learning_rate": 4.2992063403824345e-06, "loss": 0.0071, "step": 30637 }, { "epoch": 5.62, "learning_rate": 4.298596383286e-06, "loss": 0.008, "step": 30638 }, { "epoch": 5.62, "learning_rate": 4.297986457615836e-06, "loss": 0.0082, "step": 30639 }, { "epoch": 5.62, "learning_rate": 4.297376563375303e-06, "loss": 0.008, "step": 30640 }, { "epoch": 5.62, "learning_rate": 4.29676670056776e-06, "loss": 0.0104, "step": 30641 }, { "epoch": 5.62, "learning_rate": 4.296156869196578e-06, "loss": 0.0099, "step": 30642 }, { "epoch": 5.62, "learning_rate": 4.2955470692651156e-06, "loss": 0.0069, "step": 30643 }, { "epoch": 5.62, "learning_rate": 4.294937300776728e-06, "loss": 0.0085, "step": 30644 }, { "epoch": 5.62, "learning_rate": 4.294327563734779e-06, "loss": 0.006, "step": 30645 }, { "epoch": 5.62, "learning_rate": 4.293717858142628e-06, "loss": 0.0067, "step": 30646 }, { "epoch": 5.62, "learning_rate": 4.29310818400364e-06, "loss": 0.0058, "step": 30647 }, { "epoch": 5.62, "learning_rate": 4.292498541321174e-06, "loss": 0.0081, "step": 30648 }, { "epoch": 5.62, "learning_rate": 4.291888930098594e-06, "loss": 0.0071, "step": 30649 }, { "epoch": 5.62, "learning_rate": 4.291279350339247e-06, "loss": 0.0076, "step": 30650 }, { "epoch": 5.62, "learning_rate": 4.290669802046506e-06, "loss": 0.0076, "step": 30651 }, { "epoch": 5.62, "learning_rate": 4.290060285223728e-06, "loss": 0.0078, "step": 30652 }, { "epoch": 5.62, "learning_rate": 4.289450799874271e-06, "loss": 0.0078, "step": 30653 }, { "epoch": 5.62, "learning_rate": 4.288841346001495e-06, "loss": 0.008, "step": 30654 }, { "epoch": 5.62, "learning_rate": 4.288231923608759e-06, "loss": 0.0094, "step": 30655 }, { "epoch": 5.62, "learning_rate": 4.2876225326994225e-06, "loss": 0.0067, "step": 30656 }, { "epoch": 5.62, "learning_rate": 4.287013173276845e-06, "loss": 0.0087, "step": 30657 }, { "epoch": 5.62, "learning_rate": 4.286403845344384e-06, "loss": 0.0071, "step": 30658 }, { "epoch": 5.62, "learning_rate": 4.285794548905395e-06, "loss": 0.0102, "step": 30659 }, { "epoch": 5.62, "learning_rate": 4.285185283963244e-06, "loss": 0.0075, "step": 30660 }, { "epoch": 5.62, "learning_rate": 4.284576050521291e-06, "loss": 0.0068, "step": 30661 }, { "epoch": 5.62, "learning_rate": 4.283966848582883e-06, "loss": 0.0071, "step": 30662 }, { "epoch": 5.62, "learning_rate": 4.28335767815138e-06, "loss": 0.0086, "step": 30663 }, { "epoch": 5.62, "learning_rate": 4.282748539230148e-06, "loss": 0.0059, "step": 30664 }, { "epoch": 5.62, "learning_rate": 4.282139431822541e-06, "loss": 0.0063, "step": 30665 }, { "epoch": 5.62, "learning_rate": 4.281530355931913e-06, "loss": 0.0083, "step": 30666 }, { "epoch": 5.62, "learning_rate": 4.2809213115616285e-06, "loss": 0.0074, "step": 30667 }, { "epoch": 5.62, "learning_rate": 4.280312298715033e-06, "loss": 0.0063, "step": 30668 }, { "epoch": 5.62, "learning_rate": 4.279703317395492e-06, "loss": 0.012, "step": 30669 }, { "epoch": 5.62, "learning_rate": 4.279094367606362e-06, "loss": 0.0069, "step": 30670 }, { "epoch": 5.62, "learning_rate": 4.2784854493509976e-06, "loss": 0.0074, "step": 30671 }, { "epoch": 5.62, "learning_rate": 4.277876562632755e-06, "loss": 0.0076, "step": 30672 }, { "epoch": 5.62, "learning_rate": 4.277267707454992e-06, "loss": 0.008, "step": 30673 }, { "epoch": 5.63, "learning_rate": 4.276658883821062e-06, "loss": 0.0075, "step": 30674 }, { "epoch": 5.63, "learning_rate": 4.2760500917343224e-06, "loss": 0.0093, "step": 30675 }, { "epoch": 5.63, "learning_rate": 4.275441331198129e-06, "loss": 0.0084, "step": 30676 }, { "epoch": 5.63, "learning_rate": 4.274832602215834e-06, "loss": 0.0049, "step": 30677 }, { "epoch": 5.63, "learning_rate": 4.274223904790799e-06, "loss": 0.0098, "step": 30678 }, { "epoch": 5.63, "learning_rate": 4.273615238926379e-06, "loss": 0.0071, "step": 30679 }, { "epoch": 5.63, "learning_rate": 4.2730066046259246e-06, "loss": 0.0076, "step": 30680 }, { "epoch": 5.63, "learning_rate": 4.272398001892787e-06, "loss": 0.0066, "step": 30681 }, { "epoch": 5.63, "learning_rate": 4.271789430730329e-06, "loss": 0.0098, "step": 30682 }, { "epoch": 5.63, "learning_rate": 4.271180891141902e-06, "loss": 0.0078, "step": 30683 }, { "epoch": 5.63, "learning_rate": 4.2705723831308595e-06, "loss": 0.0083, "step": 30684 }, { "epoch": 5.63, "learning_rate": 4.269963906700557e-06, "loss": 0.0081, "step": 30685 }, { "epoch": 5.63, "learning_rate": 4.269355461854347e-06, "loss": 0.0072, "step": 30686 }, { "epoch": 5.63, "learning_rate": 4.268747048595585e-06, "loss": 0.0083, "step": 30687 }, { "epoch": 5.63, "learning_rate": 4.268138666927624e-06, "loss": 0.0073, "step": 30688 }, { "epoch": 5.63, "learning_rate": 4.2675303168538165e-06, "loss": 0.0092, "step": 30689 }, { "epoch": 5.63, "learning_rate": 4.266921998377514e-06, "loss": 0.0097, "step": 30690 }, { "epoch": 5.63, "learning_rate": 4.266313711502075e-06, "loss": 0.0071, "step": 30691 }, { "epoch": 5.63, "learning_rate": 4.265705456230848e-06, "loss": 0.0065, "step": 30692 }, { "epoch": 5.63, "learning_rate": 4.265097232567186e-06, "loss": 0.0084, "step": 30693 }, { "epoch": 5.63, "learning_rate": 4.264489040514439e-06, "loss": 0.0063, "step": 30694 }, { "epoch": 5.63, "learning_rate": 4.263880880075967e-06, "loss": 0.0073, "step": 30695 }, { "epoch": 5.63, "learning_rate": 4.263272751255121e-06, "loss": 0.0074, "step": 30696 }, { "epoch": 5.63, "learning_rate": 4.262664654055247e-06, "loss": 0.0073, "step": 30697 }, { "epoch": 5.63, "learning_rate": 4.262056588479699e-06, "loss": 0.0068, "step": 30698 }, { "epoch": 5.63, "learning_rate": 4.261448554531826e-06, "loss": 0.0086, "step": 30699 }, { "epoch": 5.63, "learning_rate": 4.260840552214988e-06, "loss": 0.01, "step": 30700 }, { "epoch": 5.63, "learning_rate": 4.260232581532531e-06, "loss": 0.008, "step": 30701 }, { "epoch": 5.63, "learning_rate": 4.2596246424878055e-06, "loss": 0.007, "step": 30702 }, { "epoch": 5.63, "learning_rate": 4.259016735084164e-06, "loss": 0.0078, "step": 30703 }, { "epoch": 5.63, "learning_rate": 4.258408859324956e-06, "loss": 0.0085, "step": 30704 }, { "epoch": 5.63, "learning_rate": 4.257801015213533e-06, "loss": 0.0046, "step": 30705 }, { "epoch": 5.63, "learning_rate": 4.257193202753246e-06, "loss": 0.0083, "step": 30706 }, { "epoch": 5.63, "learning_rate": 4.256585421947445e-06, "loss": 0.0099, "step": 30707 }, { "epoch": 5.63, "learning_rate": 4.255977672799479e-06, "loss": 0.0098, "step": 30708 }, { "epoch": 5.63, "learning_rate": 4.255369955312698e-06, "loss": 0.0066, "step": 30709 }, { "epoch": 5.63, "learning_rate": 4.254762269490453e-06, "loss": 0.0085, "step": 30710 }, { "epoch": 5.63, "learning_rate": 4.2541546153360934e-06, "loss": 0.0107, "step": 30711 }, { "epoch": 5.63, "learning_rate": 4.253546992852965e-06, "loss": 0.0083, "step": 30712 }, { "epoch": 5.63, "learning_rate": 4.252939402044423e-06, "loss": 0.0074, "step": 30713 }, { "epoch": 5.63, "learning_rate": 4.252331842913817e-06, "loss": 0.0072, "step": 30714 }, { "epoch": 5.63, "learning_rate": 4.25172431546449e-06, "loss": 0.0075, "step": 30715 }, { "epoch": 5.63, "learning_rate": 4.251116819699787e-06, "loss": 0.0084, "step": 30716 }, { "epoch": 5.63, "learning_rate": 4.250509355623069e-06, "loss": 0.0065, "step": 30717 }, { "epoch": 5.63, "learning_rate": 4.249901923237677e-06, "loss": 0.0072, "step": 30718 }, { "epoch": 5.63, "learning_rate": 4.249294522546962e-06, "loss": 0.009, "step": 30719 }, { "epoch": 5.63, "learning_rate": 4.248687153554269e-06, "loss": 0.0064, "step": 30720 }, { "epoch": 5.63, "learning_rate": 4.248079816262947e-06, "loss": 0.0099, "step": 30721 }, { "epoch": 5.63, "learning_rate": 4.247472510676344e-06, "loss": 0.0092, "step": 30722 }, { "epoch": 5.63, "learning_rate": 4.246865236797807e-06, "loss": 0.0065, "step": 30723 }, { "epoch": 5.63, "learning_rate": 4.246257994630684e-06, "loss": 0.0079, "step": 30724 }, { "epoch": 5.63, "learning_rate": 4.245650784178321e-06, "loss": 0.0072, "step": 30725 }, { "epoch": 5.63, "learning_rate": 4.245043605444066e-06, "loss": 0.0068, "step": 30726 }, { "epoch": 5.63, "learning_rate": 4.2444364584312654e-06, "loss": 0.0081, "step": 30727 }, { "epoch": 5.64, "learning_rate": 4.2438293431432665e-06, "loss": 0.0089, "step": 30728 }, { "epoch": 5.64, "learning_rate": 4.243222259583415e-06, "loss": 0.0081, "step": 30729 }, { "epoch": 5.64, "learning_rate": 4.242615207755054e-06, "loss": 0.0071, "step": 30730 }, { "epoch": 5.64, "learning_rate": 4.242008187661536e-06, "loss": 0.0063, "step": 30731 }, { "epoch": 5.64, "learning_rate": 4.241401199306206e-06, "loss": 0.0087, "step": 30732 }, { "epoch": 5.64, "learning_rate": 4.2407942426924064e-06, "loss": 0.0084, "step": 30733 }, { "epoch": 5.64, "learning_rate": 4.240187317823479e-06, "loss": 0.0072, "step": 30734 }, { "epoch": 5.64, "learning_rate": 4.2395804247027775e-06, "loss": 0.0074, "step": 30735 }, { "epoch": 5.64, "learning_rate": 4.238973563333644e-06, "loss": 0.0066, "step": 30736 }, { "epoch": 5.64, "learning_rate": 4.2383667337194235e-06, "loss": 0.0074, "step": 30737 }, { "epoch": 5.64, "learning_rate": 4.237759935863459e-06, "loss": 0.0102, "step": 30738 }, { "epoch": 5.64, "learning_rate": 4.237153169769098e-06, "loss": 0.0085, "step": 30739 }, { "epoch": 5.64, "learning_rate": 4.236546435439685e-06, "loss": 0.0079, "step": 30740 }, { "epoch": 5.64, "learning_rate": 4.235939732878561e-06, "loss": 0.0098, "step": 30741 }, { "epoch": 5.64, "learning_rate": 4.235333062089072e-06, "loss": 0.0059, "step": 30742 }, { "epoch": 5.64, "learning_rate": 4.234726423074564e-06, "loss": 0.0066, "step": 30743 }, { "epoch": 5.64, "learning_rate": 4.234119815838378e-06, "loss": 0.0067, "step": 30744 }, { "epoch": 5.64, "learning_rate": 4.233513240383858e-06, "loss": 0.0093, "step": 30745 }, { "epoch": 5.64, "learning_rate": 4.23290669671435e-06, "loss": 0.0066, "step": 30746 }, { "epoch": 5.64, "learning_rate": 4.23230018483319e-06, "loss": 0.0094, "step": 30747 }, { "epoch": 5.64, "learning_rate": 4.231693704743731e-06, "loss": 0.0081, "step": 30748 }, { "epoch": 5.64, "learning_rate": 4.231087256449311e-06, "loss": 0.0077, "step": 30749 }, { "epoch": 5.64, "learning_rate": 4.230480839953277e-06, "loss": 0.0072, "step": 30750 }, { "epoch": 5.64, "learning_rate": 4.229874455258964e-06, "loss": 0.0083, "step": 30751 }, { "epoch": 5.64, "learning_rate": 4.229268102369715e-06, "loss": 0.0074, "step": 30752 }, { "epoch": 5.64, "learning_rate": 4.228661781288879e-06, "loss": 0.0083, "step": 30753 }, { "epoch": 5.64, "learning_rate": 4.228055492019794e-06, "loss": 0.0069, "step": 30754 }, { "epoch": 5.64, "learning_rate": 4.227449234565803e-06, "loss": 0.0086, "step": 30755 }, { "epoch": 5.64, "learning_rate": 4.2268430089302455e-06, "loss": 0.0074, "step": 30756 }, { "epoch": 5.64, "learning_rate": 4.226236815116467e-06, "loss": 0.0063, "step": 30757 }, { "epoch": 5.64, "learning_rate": 4.2256306531278045e-06, "loss": 0.0086, "step": 30758 }, { "epoch": 5.64, "learning_rate": 4.225024522967602e-06, "loss": 0.0053, "step": 30759 }, { "epoch": 5.64, "learning_rate": 4.2244184246391995e-06, "loss": 0.01, "step": 30760 }, { "epoch": 5.64, "learning_rate": 4.223812358145938e-06, "loss": 0.0099, "step": 30761 }, { "epoch": 5.64, "learning_rate": 4.223206323491158e-06, "loss": 0.0085, "step": 30762 }, { "epoch": 5.64, "learning_rate": 4.222600320678199e-06, "loss": 0.0082, "step": 30763 }, { "epoch": 5.64, "learning_rate": 4.221994349710404e-06, "loss": 0.0093, "step": 30764 }, { "epoch": 5.64, "learning_rate": 4.221388410591107e-06, "loss": 0.0067, "step": 30765 }, { "epoch": 5.64, "learning_rate": 4.220782503323657e-06, "loss": 0.0084, "step": 30766 }, { "epoch": 5.64, "learning_rate": 4.220176627911389e-06, "loss": 0.0079, "step": 30767 }, { "epoch": 5.64, "learning_rate": 4.219570784357646e-06, "loss": 0.0058, "step": 30768 }, { "epoch": 5.64, "learning_rate": 4.218964972665757e-06, "loss": 0.0092, "step": 30769 }, { "epoch": 5.64, "learning_rate": 4.218359192839072e-06, "loss": 0.008, "step": 30770 }, { "epoch": 5.64, "learning_rate": 4.217753444880927e-06, "loss": 0.0107, "step": 30771 }, { "epoch": 5.64, "learning_rate": 4.21714772879466e-06, "loss": 0.0076, "step": 30772 }, { "epoch": 5.64, "learning_rate": 4.216542044583611e-06, "loss": 0.0086, "step": 30773 }, { "epoch": 5.64, "learning_rate": 4.215936392251117e-06, "loss": 0.0106, "step": 30774 }, { "epoch": 5.64, "learning_rate": 4.215330771800517e-06, "loss": 0.0069, "step": 30775 }, { "epoch": 5.64, "learning_rate": 4.21472518323515e-06, "loss": 0.0063, "step": 30776 }, { "epoch": 5.64, "learning_rate": 4.214119626558353e-06, "loss": 0.0087, "step": 30777 }, { "epoch": 5.64, "learning_rate": 4.21351410177346e-06, "loss": 0.0069, "step": 30778 }, { "epoch": 5.64, "learning_rate": 4.212908608883821e-06, "loss": 0.0061, "step": 30779 }, { "epoch": 5.64, "learning_rate": 4.212303147892761e-06, "loss": 0.0071, "step": 30780 }, { "epoch": 5.64, "learning_rate": 4.211697718803622e-06, "loss": 0.0063, "step": 30781 }, { "epoch": 5.64, "learning_rate": 4.211092321619741e-06, "loss": 0.0095, "step": 30782 }, { "epoch": 5.65, "learning_rate": 4.2104869563444505e-06, "loss": 0.008, "step": 30783 }, { "epoch": 5.65, "learning_rate": 4.209881622981096e-06, "loss": 0.0069, "step": 30784 }, { "epoch": 5.65, "learning_rate": 4.209276321533009e-06, "loss": 0.0054, "step": 30785 }, { "epoch": 5.65, "learning_rate": 4.20867105200353e-06, "loss": 0.0096, "step": 30786 }, { "epoch": 5.65, "learning_rate": 4.208065814395985e-06, "loss": 0.0089, "step": 30787 }, { "epoch": 5.65, "learning_rate": 4.20746060871372e-06, "loss": 0.0061, "step": 30788 }, { "epoch": 5.65, "learning_rate": 4.206855434960069e-06, "loss": 0.0103, "step": 30789 }, { "epoch": 5.65, "learning_rate": 4.206250293138366e-06, "loss": 0.008, "step": 30790 }, { "epoch": 5.65, "learning_rate": 4.2056451832519475e-06, "loss": 0.0064, "step": 30791 }, { "epoch": 5.65, "learning_rate": 4.205040105304148e-06, "loss": 0.0061, "step": 30792 }, { "epoch": 5.65, "learning_rate": 4.204435059298303e-06, "loss": 0.0068, "step": 30793 }, { "epoch": 5.65, "learning_rate": 4.203830045237748e-06, "loss": 0.0075, "step": 30794 }, { "epoch": 5.65, "learning_rate": 4.203225063125816e-06, "loss": 0.0059, "step": 30795 }, { "epoch": 5.65, "learning_rate": 4.2026201129658415e-06, "loss": 0.0089, "step": 30796 }, { "epoch": 5.65, "learning_rate": 4.202015194761169e-06, "loss": 0.0075, "step": 30797 }, { "epoch": 5.65, "learning_rate": 4.2014103085151195e-06, "loss": 0.0055, "step": 30798 }, { "epoch": 5.65, "learning_rate": 4.200805454231032e-06, "loss": 0.0076, "step": 30799 }, { "epoch": 5.65, "learning_rate": 4.200200631912237e-06, "loss": 0.0091, "step": 30800 }, { "epoch": 5.65, "learning_rate": 4.199595841562077e-06, "loss": 0.0086, "step": 30801 }, { "epoch": 5.65, "learning_rate": 4.198991083183879e-06, "loss": 0.0089, "step": 30802 }, { "epoch": 5.65, "learning_rate": 4.1983863567809825e-06, "loss": 0.009, "step": 30803 }, { "epoch": 5.65, "learning_rate": 4.1977816623567114e-06, "loss": 0.009, "step": 30804 }, { "epoch": 5.65, "learning_rate": 4.197176999914402e-06, "loss": 0.0054, "step": 30805 }, { "epoch": 5.65, "learning_rate": 4.196572369457391e-06, "loss": 0.0068, "step": 30806 }, { "epoch": 5.65, "learning_rate": 4.19596777098901e-06, "loss": 0.005, "step": 30807 }, { "epoch": 5.65, "learning_rate": 4.19536320451259e-06, "loss": 0.0085, "step": 30808 }, { "epoch": 5.65, "learning_rate": 4.194758670031465e-06, "loss": 0.0077, "step": 30809 }, { "epoch": 5.65, "learning_rate": 4.194154167548965e-06, "loss": 0.009, "step": 30810 }, { "epoch": 5.65, "learning_rate": 4.193549697068424e-06, "loss": 0.01, "step": 30811 }, { "epoch": 5.65, "learning_rate": 4.192945258593173e-06, "loss": 0.0083, "step": 30812 }, { "epoch": 5.65, "learning_rate": 4.192340852126541e-06, "loss": 0.0078, "step": 30813 }, { "epoch": 5.65, "learning_rate": 4.191736477671864e-06, "loss": 0.0058, "step": 30814 }, { "epoch": 5.65, "learning_rate": 4.1911321352324764e-06, "loss": 0.0079, "step": 30815 }, { "epoch": 5.65, "learning_rate": 4.190527824811701e-06, "loss": 0.0095, "step": 30816 }, { "epoch": 5.65, "learning_rate": 4.189923546412872e-06, "loss": 0.0069, "step": 30817 }, { "epoch": 5.65, "learning_rate": 4.1893193000393176e-06, "loss": 0.0066, "step": 30818 }, { "epoch": 5.65, "learning_rate": 4.188715085694374e-06, "loss": 0.0081, "step": 30819 }, { "epoch": 5.65, "learning_rate": 4.188110903381369e-06, "loss": 0.0079, "step": 30820 }, { "epoch": 5.65, "learning_rate": 4.187506753103637e-06, "loss": 0.0058, "step": 30821 }, { "epoch": 5.65, "learning_rate": 4.186902634864497e-06, "loss": 0.007, "step": 30822 }, { "epoch": 5.65, "learning_rate": 4.1862985486672895e-06, "loss": 0.0066, "step": 30823 }, { "epoch": 5.65, "learning_rate": 4.18569449451534e-06, "loss": 0.0093, "step": 30824 }, { "epoch": 5.65, "learning_rate": 4.1850904724119775e-06, "loss": 0.0087, "step": 30825 }, { "epoch": 5.65, "learning_rate": 4.184486482360535e-06, "loss": 0.0099, "step": 30826 }, { "epoch": 5.65, "learning_rate": 4.183882524364338e-06, "loss": 0.0066, "step": 30827 }, { "epoch": 5.65, "learning_rate": 4.1832785984267165e-06, "loss": 0.0065, "step": 30828 }, { "epoch": 5.65, "learning_rate": 4.1826747045509995e-06, "loss": 0.0063, "step": 30829 }, { "epoch": 5.65, "learning_rate": 4.182070842740517e-06, "loss": 0.0095, "step": 30830 }, { "epoch": 5.65, "learning_rate": 4.1814670129985924e-06, "loss": 0.0081, "step": 30831 }, { "epoch": 5.65, "learning_rate": 4.180863215328561e-06, "loss": 0.0071, "step": 30832 }, { "epoch": 5.65, "learning_rate": 4.180259449733751e-06, "loss": 0.0073, "step": 30833 }, { "epoch": 5.65, "learning_rate": 4.179655716217484e-06, "loss": 0.0086, "step": 30834 }, { "epoch": 5.65, "learning_rate": 4.179052014783092e-06, "loss": 0.0076, "step": 30835 }, { "epoch": 5.65, "learning_rate": 4.178448345433898e-06, "loss": 0.0067, "step": 30836 }, { "epoch": 5.66, "learning_rate": 4.177844708173235e-06, "loss": 0.0073, "step": 30837 }, { "epoch": 5.66, "learning_rate": 4.17724110300443e-06, "loss": 0.0073, "step": 30838 }, { "epoch": 5.66, "learning_rate": 4.176637529930811e-06, "loss": 0.0091, "step": 30839 }, { "epoch": 5.66, "learning_rate": 4.176033988955697e-06, "loss": 0.0074, "step": 30840 }, { "epoch": 5.66, "learning_rate": 4.175430480082422e-06, "loss": 0.0081, "step": 30841 }, { "epoch": 5.66, "learning_rate": 4.174827003314311e-06, "loss": 0.0087, "step": 30842 }, { "epoch": 5.66, "learning_rate": 4.1742235586546894e-06, "loss": 0.0111, "step": 30843 }, { "epoch": 5.66, "learning_rate": 4.1736201461068856e-06, "loss": 0.0081, "step": 30844 }, { "epoch": 5.66, "learning_rate": 4.173016765674222e-06, "loss": 0.0072, "step": 30845 }, { "epoch": 5.66, "learning_rate": 4.172413417360029e-06, "loss": 0.007, "step": 30846 }, { "epoch": 5.66, "learning_rate": 4.171810101167628e-06, "loss": 0.01, "step": 30847 }, { "epoch": 5.66, "learning_rate": 4.171206817100348e-06, "loss": 0.0103, "step": 30848 }, { "epoch": 5.66, "learning_rate": 4.170603565161506e-06, "loss": 0.0059, "step": 30849 }, { "epoch": 5.66, "learning_rate": 4.170000345354443e-06, "loss": 0.0071, "step": 30850 }, { "epoch": 5.66, "learning_rate": 4.169397157682471e-06, "loss": 0.0082, "step": 30851 }, { "epoch": 5.66, "learning_rate": 4.168794002148918e-06, "loss": 0.0045, "step": 30852 }, { "epoch": 5.66, "learning_rate": 4.1681908787571055e-06, "loss": 0.008, "step": 30853 }, { "epoch": 5.66, "learning_rate": 4.167587787510366e-06, "loss": 0.0072, "step": 30854 }, { "epoch": 5.66, "learning_rate": 4.166984728412018e-06, "loss": 0.0091, "step": 30855 }, { "epoch": 5.66, "learning_rate": 4.166381701465387e-06, "loss": 0.0082, "step": 30856 }, { "epoch": 5.66, "learning_rate": 4.165778706673801e-06, "loss": 0.006, "step": 30857 }, { "epoch": 5.66, "learning_rate": 4.165175744040572e-06, "loss": 0.01, "step": 30858 }, { "epoch": 5.66, "learning_rate": 4.1645728135690345e-06, "loss": 0.0092, "step": 30859 }, { "epoch": 5.66, "learning_rate": 4.163969915262508e-06, "loss": 0.0087, "step": 30860 }, { "epoch": 5.66, "learning_rate": 4.1633670491243165e-06, "loss": 0.0065, "step": 30861 }, { "epoch": 5.66, "learning_rate": 4.162764215157783e-06, "loss": 0.0087, "step": 30862 }, { "epoch": 5.66, "learning_rate": 4.162161413366229e-06, "loss": 0.0079, "step": 30863 }, { "epoch": 5.66, "learning_rate": 4.1615586437529785e-06, "loss": 0.0068, "step": 30864 }, { "epoch": 5.66, "learning_rate": 4.160955906321354e-06, "loss": 0.008, "step": 30865 }, { "epoch": 5.66, "learning_rate": 4.160353201074673e-06, "loss": 0.0083, "step": 30866 }, { "epoch": 5.66, "learning_rate": 4.159750528016265e-06, "loss": 0.0077, "step": 30867 }, { "epoch": 5.66, "learning_rate": 4.159147887149454e-06, "loss": 0.007, "step": 30868 }, { "epoch": 5.66, "learning_rate": 4.1585452784775515e-06, "loss": 0.0068, "step": 30869 }, { "epoch": 5.66, "learning_rate": 4.157942702003885e-06, "loss": 0.0072, "step": 30870 }, { "epoch": 5.66, "learning_rate": 4.157340157731772e-06, "loss": 0.0073, "step": 30871 }, { "epoch": 5.66, "learning_rate": 4.15673764566454e-06, "loss": 0.008, "step": 30872 }, { "epoch": 5.66, "learning_rate": 4.1561351658055075e-06, "loss": 0.0084, "step": 30873 }, { "epoch": 5.66, "learning_rate": 4.155532718157995e-06, "loss": 0.0071, "step": 30874 }, { "epoch": 5.66, "learning_rate": 4.154930302725322e-06, "loss": 0.0063, "step": 30875 }, { "epoch": 5.66, "learning_rate": 4.154327919510811e-06, "loss": 0.0089, "step": 30876 }, { "epoch": 5.66, "learning_rate": 4.153725568517781e-06, "loss": 0.0107, "step": 30877 }, { "epoch": 5.66, "learning_rate": 4.153123249749553e-06, "loss": 0.01, "step": 30878 }, { "epoch": 5.66, "learning_rate": 4.152520963209446e-06, "loss": 0.0078, "step": 30879 }, { "epoch": 5.66, "learning_rate": 4.151918708900782e-06, "loss": 0.0096, "step": 30880 }, { "epoch": 5.66, "learning_rate": 4.151316486826877e-06, "loss": 0.0104, "step": 30881 }, { "epoch": 5.66, "learning_rate": 4.150714296991054e-06, "loss": 0.0065, "step": 30882 }, { "epoch": 5.66, "learning_rate": 4.15011213939663e-06, "loss": 0.0095, "step": 30883 }, { "epoch": 5.66, "learning_rate": 4.149510014046922e-06, "loss": 0.0081, "step": 30884 }, { "epoch": 5.66, "learning_rate": 4.148907920945255e-06, "loss": 0.0084, "step": 30885 }, { "epoch": 5.66, "learning_rate": 4.148305860094949e-06, "loss": 0.0076, "step": 30886 }, { "epoch": 5.66, "learning_rate": 4.147703831499314e-06, "loss": 0.0108, "step": 30887 }, { "epoch": 5.66, "learning_rate": 4.147101835161673e-06, "loss": 0.0066, "step": 30888 }, { "epoch": 5.66, "learning_rate": 4.1464998710853395e-06, "loss": 0.0064, "step": 30889 }, { "epoch": 5.66, "learning_rate": 4.14589793927364e-06, "loss": 0.0075, "step": 30890 }, { "epoch": 5.66, "learning_rate": 4.145296039729888e-06, "loss": 0.0083, "step": 30891 }, { "epoch": 5.67, "learning_rate": 4.144694172457401e-06, "loss": 0.0069, "step": 30892 }, { "epoch": 5.67, "learning_rate": 4.144092337459498e-06, "loss": 0.0072, "step": 30893 }, { "epoch": 5.67, "learning_rate": 4.143490534739494e-06, "loss": 0.0055, "step": 30894 }, { "epoch": 5.67, "learning_rate": 4.142888764300708e-06, "loss": 0.0065, "step": 30895 }, { "epoch": 5.67, "learning_rate": 4.142287026146456e-06, "loss": 0.0066, "step": 30896 }, { "epoch": 5.67, "learning_rate": 4.141685320280055e-06, "loss": 0.01, "step": 30897 }, { "epoch": 5.67, "learning_rate": 4.141083646704822e-06, "loss": 0.0054, "step": 30898 }, { "epoch": 5.67, "learning_rate": 4.140482005424073e-06, "loss": 0.008, "step": 30899 }, { "epoch": 5.67, "learning_rate": 4.139880396441125e-06, "loss": 0.0059, "step": 30900 }, { "epoch": 5.67, "learning_rate": 4.139278819759294e-06, "loss": 0.0075, "step": 30901 }, { "epoch": 5.67, "learning_rate": 4.1386772753818904e-06, "loss": 0.0081, "step": 30902 }, { "epoch": 5.67, "learning_rate": 4.138075763312239e-06, "loss": 0.0083, "step": 30903 }, { "epoch": 5.67, "learning_rate": 4.137474283553656e-06, "loss": 0.0079, "step": 30904 }, { "epoch": 5.67, "learning_rate": 4.136872836109448e-06, "loss": 0.0075, "step": 30905 }, { "epoch": 5.67, "learning_rate": 4.13627142098293e-06, "loss": 0.0082, "step": 30906 }, { "epoch": 5.67, "learning_rate": 4.1356700381774255e-06, "loss": 0.0087, "step": 30907 }, { "epoch": 5.67, "learning_rate": 4.135068687696244e-06, "loss": 0.0071, "step": 30908 }, { "epoch": 5.67, "learning_rate": 4.134467369542703e-06, "loss": 0.0098, "step": 30909 }, { "epoch": 5.67, "learning_rate": 4.133866083720118e-06, "loss": 0.0074, "step": 30910 }, { "epoch": 5.67, "learning_rate": 4.1332648302317935e-06, "loss": 0.0097, "step": 30911 }, { "epoch": 5.67, "learning_rate": 4.132663609081055e-06, "loss": 0.0083, "step": 30912 }, { "epoch": 5.67, "learning_rate": 4.132062420271211e-06, "loss": 0.0068, "step": 30913 }, { "epoch": 5.67, "learning_rate": 4.1314612638055766e-06, "loss": 0.0096, "step": 30914 }, { "epoch": 5.67, "learning_rate": 4.1308601396874655e-06, "loss": 0.0087, "step": 30915 }, { "epoch": 5.67, "learning_rate": 4.130259047920191e-06, "loss": 0.0075, "step": 30916 }, { "epoch": 5.67, "learning_rate": 4.129657988507065e-06, "loss": 0.008, "step": 30917 }, { "epoch": 5.67, "learning_rate": 4.129056961451403e-06, "loss": 0.0086, "step": 30918 }, { "epoch": 5.67, "learning_rate": 4.128455966756512e-06, "loss": 0.0093, "step": 30919 }, { "epoch": 5.67, "learning_rate": 4.127855004425713e-06, "loss": 0.0083, "step": 30920 }, { "epoch": 5.67, "learning_rate": 4.127254074462317e-06, "loss": 0.0063, "step": 30921 }, { "epoch": 5.67, "learning_rate": 4.1266531768696345e-06, "loss": 0.0091, "step": 30922 }, { "epoch": 5.67, "learning_rate": 4.126052311650977e-06, "loss": 0.0111, "step": 30923 }, { "epoch": 5.67, "learning_rate": 4.12545147880965e-06, "loss": 0.008, "step": 30924 }, { "epoch": 5.67, "learning_rate": 4.124850678348977e-06, "loss": 0.0099, "step": 30925 }, { "epoch": 5.67, "learning_rate": 4.124249910272265e-06, "loss": 0.0086, "step": 30926 }, { "epoch": 5.67, "learning_rate": 4.123649174582824e-06, "loss": 0.0052, "step": 30927 }, { "epoch": 5.67, "learning_rate": 4.123048471283968e-06, "loss": 0.0071, "step": 30928 }, { "epoch": 5.67, "learning_rate": 4.122447800379006e-06, "loss": 0.006, "step": 30929 }, { "epoch": 5.67, "learning_rate": 4.121847161871248e-06, "loss": 0.0078, "step": 30930 }, { "epoch": 5.67, "learning_rate": 4.121246555764008e-06, "loss": 0.0067, "step": 30931 }, { "epoch": 5.67, "learning_rate": 4.1206459820605935e-06, "loss": 0.0063, "step": 30932 }, { "epoch": 5.67, "learning_rate": 4.120045440764318e-06, "loss": 0.0081, "step": 30933 }, { "epoch": 5.67, "learning_rate": 4.119444931878488e-06, "loss": 0.0061, "step": 30934 }, { "epoch": 5.67, "learning_rate": 4.118844455406415e-06, "loss": 0.0081, "step": 30935 }, { "epoch": 5.67, "learning_rate": 4.11824401135141e-06, "loss": 0.0066, "step": 30936 }, { "epoch": 5.67, "learning_rate": 4.117643599716778e-06, "loss": 0.0099, "step": 30937 }, { "epoch": 5.67, "learning_rate": 4.1170432205058364e-06, "loss": 0.0067, "step": 30938 }, { "epoch": 5.67, "learning_rate": 4.11644287372189e-06, "loss": 0.0065, "step": 30939 }, { "epoch": 5.67, "learning_rate": 4.115842559368251e-06, "loss": 0.0076, "step": 30940 }, { "epoch": 5.67, "learning_rate": 4.11524227744822e-06, "loss": 0.0062, "step": 30941 }, { "epoch": 5.67, "learning_rate": 4.114642027965114e-06, "loss": 0.0083, "step": 30942 }, { "epoch": 5.67, "learning_rate": 4.11404181092224e-06, "loss": 0.0078, "step": 30943 }, { "epoch": 5.67, "learning_rate": 4.1134416263229045e-06, "loss": 0.0076, "step": 30944 }, { "epoch": 5.67, "learning_rate": 4.112841474170417e-06, "loss": 0.0075, "step": 30945 }, { "epoch": 5.68, "learning_rate": 4.112241354468085e-06, "loss": 0.0077, "step": 30946 }, { "epoch": 5.68, "learning_rate": 4.111641267219218e-06, "loss": 0.0092, "step": 30947 }, { "epoch": 5.68, "learning_rate": 4.111041212427121e-06, "loss": 0.0078, "step": 30948 }, { "epoch": 5.68, "learning_rate": 4.110441190095102e-06, "loss": 0.0078, "step": 30949 }, { "epoch": 5.68, "learning_rate": 4.1098412002264665e-06, "loss": 0.0077, "step": 30950 }, { "epoch": 5.68, "learning_rate": 4.109241242824531e-06, "loss": 0.0081, "step": 30951 }, { "epoch": 5.68, "learning_rate": 4.108641317892592e-06, "loss": 0.0078, "step": 30952 }, { "epoch": 5.68, "learning_rate": 4.108041425433961e-06, "loss": 0.0074, "step": 30953 }, { "epoch": 5.68, "learning_rate": 4.107441565451944e-06, "loss": 0.0091, "step": 30954 }, { "epoch": 5.68, "learning_rate": 4.106841737949843e-06, "loss": 0.0069, "step": 30955 }, { "epoch": 5.68, "learning_rate": 4.106241942930972e-06, "loss": 0.0064, "step": 30956 }, { "epoch": 5.68, "learning_rate": 4.105642180398637e-06, "loss": 0.009, "step": 30957 }, { "epoch": 5.68, "learning_rate": 4.105042450356136e-06, "loss": 0.0075, "step": 30958 }, { "epoch": 5.68, "learning_rate": 4.104442752806776e-06, "loss": 0.0091, "step": 30959 }, { "epoch": 5.68, "learning_rate": 4.10384308775387e-06, "loss": 0.007, "step": 30960 }, { "epoch": 5.68, "learning_rate": 4.1032434552007186e-06, "loss": 0.0088, "step": 30961 }, { "epoch": 5.68, "learning_rate": 4.102643855150628e-06, "loss": 0.0077, "step": 30962 }, { "epoch": 5.68, "learning_rate": 4.102044287606901e-06, "loss": 0.0071, "step": 30963 }, { "epoch": 5.68, "learning_rate": 4.101444752572845e-06, "loss": 0.0072, "step": 30964 }, { "epoch": 5.68, "learning_rate": 4.100845250051764e-06, "loss": 0.0086, "step": 30965 }, { "epoch": 5.68, "learning_rate": 4.100245780046962e-06, "loss": 0.011, "step": 30966 }, { "epoch": 5.68, "learning_rate": 4.099646342561744e-06, "loss": 0.0099, "step": 30967 }, { "epoch": 5.68, "learning_rate": 4.099046937599409e-06, "loss": 0.0102, "step": 30968 }, { "epoch": 5.68, "learning_rate": 4.0984475651632725e-06, "loss": 0.0091, "step": 30969 }, { "epoch": 5.68, "learning_rate": 4.097848225256629e-06, "loss": 0.0074, "step": 30970 }, { "epoch": 5.68, "learning_rate": 4.0972489178827835e-06, "loss": 0.0087, "step": 30971 }, { "epoch": 5.68, "learning_rate": 4.096649643045037e-06, "loss": 0.0065, "step": 30972 }, { "epoch": 5.68, "learning_rate": 4.0960504007466995e-06, "loss": 0.0054, "step": 30973 }, { "epoch": 5.68, "learning_rate": 4.095451190991071e-06, "loss": 0.0075, "step": 30974 }, { "epoch": 5.68, "learning_rate": 4.094852013781456e-06, "loss": 0.007, "step": 30975 }, { "epoch": 5.68, "learning_rate": 4.094252869121153e-06, "loss": 0.0058, "step": 30976 }, { "epoch": 5.68, "learning_rate": 4.093653757013461e-06, "loss": 0.0078, "step": 30977 }, { "epoch": 5.68, "learning_rate": 4.0930546774616944e-06, "loss": 0.0064, "step": 30978 }, { "epoch": 5.68, "learning_rate": 4.0924556304691465e-06, "loss": 0.006, "step": 30979 }, { "epoch": 5.68, "learning_rate": 4.091856616039123e-06, "loss": 0.0055, "step": 30980 }, { "epoch": 5.68, "learning_rate": 4.091257634174922e-06, "loss": 0.0094, "step": 30981 }, { "epoch": 5.68, "learning_rate": 4.090658684879849e-06, "loss": 0.0105, "step": 30982 }, { "epoch": 5.68, "learning_rate": 4.090059768157204e-06, "loss": 0.0064, "step": 30983 }, { "epoch": 5.68, "learning_rate": 4.089460884010288e-06, "loss": 0.0074, "step": 30984 }, { "epoch": 5.68, "learning_rate": 4.088862032442401e-06, "loss": 0.0081, "step": 30985 }, { "epoch": 5.68, "learning_rate": 4.088263213456842e-06, "loss": 0.0094, "step": 30986 }, { "epoch": 5.68, "learning_rate": 4.087664427056922e-06, "loss": 0.0108, "step": 30987 }, { "epoch": 5.68, "learning_rate": 4.08706567324593e-06, "loss": 0.0068, "step": 30988 }, { "epoch": 5.68, "learning_rate": 4.086466952027171e-06, "loss": 0.0066, "step": 30989 }, { "epoch": 5.68, "learning_rate": 4.0858682634039414e-06, "loss": 0.0084, "step": 30990 }, { "epoch": 5.68, "learning_rate": 4.085269607379549e-06, "loss": 0.0069, "step": 30991 }, { "epoch": 5.68, "learning_rate": 4.084670983957287e-06, "loss": 0.0057, "step": 30992 }, { "epoch": 5.68, "learning_rate": 4.0840723931404615e-06, "loss": 0.009, "step": 30993 }, { "epoch": 5.68, "learning_rate": 4.083473834932361e-06, "loss": 0.0069, "step": 30994 }, { "epoch": 5.68, "learning_rate": 4.082875309336294e-06, "loss": 0.0089, "step": 30995 }, { "epoch": 5.68, "learning_rate": 4.082276816355557e-06, "loss": 0.0052, "step": 30996 }, { "epoch": 5.68, "learning_rate": 4.081678355993449e-06, "loss": 0.0102, "step": 30997 }, { "epoch": 5.68, "learning_rate": 4.081079928253269e-06, "loss": 0.0078, "step": 30998 }, { "epoch": 5.68, "learning_rate": 4.080481533138315e-06, "loss": 0.0069, "step": 30999 }, { "epoch": 5.68, "learning_rate": 4.079883170651885e-06, "loss": 0.0079, "step": 31000 }, { "epoch": 5.69, "learning_rate": 4.079284840797278e-06, "loss": 0.0062, "step": 31001 }, { "epoch": 5.69, "learning_rate": 4.078686543577792e-06, "loss": 0.011, "step": 31002 }, { "epoch": 5.69, "learning_rate": 4.078088278996721e-06, "loss": 0.0121, "step": 31003 }, { "epoch": 5.69, "learning_rate": 4.077490047057372e-06, "loss": 0.0081, "step": 31004 }, { "epoch": 5.69, "learning_rate": 4.076891847763034e-06, "loss": 0.0087, "step": 31005 }, { "epoch": 5.69, "learning_rate": 4.076293681117006e-06, "loss": 0.009, "step": 31006 }, { "epoch": 5.69, "learning_rate": 4.0756955471225875e-06, "loss": 0.0111, "step": 31007 }, { "epoch": 5.69, "learning_rate": 4.075097445783069e-06, "loss": 0.0085, "step": 31008 }, { "epoch": 5.69, "learning_rate": 4.074499377101757e-06, "loss": 0.0087, "step": 31009 }, { "epoch": 5.69, "learning_rate": 4.0739013410819426e-06, "loss": 0.0083, "step": 31010 }, { "epoch": 5.69, "learning_rate": 4.073303337726926e-06, "loss": 0.0078, "step": 31011 }, { "epoch": 5.69, "learning_rate": 4.072705367039994e-06, "loss": 0.0091, "step": 31012 }, { "epoch": 5.69, "learning_rate": 4.072107429024452e-06, "loss": 0.009, "step": 31013 }, { "epoch": 5.69, "learning_rate": 4.071509523683592e-06, "loss": 0.0066, "step": 31014 }, { "epoch": 5.69, "learning_rate": 4.070911651020711e-06, "loss": 0.0077, "step": 31015 }, { "epoch": 5.69, "learning_rate": 4.070313811039105e-06, "loss": 0.011, "step": 31016 }, { "epoch": 5.69, "learning_rate": 4.069716003742067e-06, "loss": 0.0064, "step": 31017 }, { "epoch": 5.69, "learning_rate": 4.069118229132893e-06, "loss": 0.0076, "step": 31018 }, { "epoch": 5.69, "learning_rate": 4.068520487214879e-06, "loss": 0.0081, "step": 31019 }, { "epoch": 5.69, "learning_rate": 4.067922777991319e-06, "loss": 0.0075, "step": 31020 }, { "epoch": 5.69, "learning_rate": 4.067325101465503e-06, "loss": 0.0079, "step": 31021 }, { "epoch": 5.69, "learning_rate": 4.066727457640739e-06, "loss": 0.0068, "step": 31022 }, { "epoch": 5.69, "learning_rate": 4.066129846520308e-06, "loss": 0.0066, "step": 31023 }, { "epoch": 5.69, "learning_rate": 4.065532268107507e-06, "loss": 0.0079, "step": 31024 }, { "epoch": 5.69, "learning_rate": 4.064934722405629e-06, "loss": 0.009, "step": 31025 }, { "epoch": 5.69, "learning_rate": 4.0643372094179735e-06, "loss": 0.0079, "step": 31026 }, { "epoch": 5.69, "learning_rate": 4.0637397291478294e-06, "loss": 0.0059, "step": 31027 }, { "epoch": 5.69, "learning_rate": 4.063142281598492e-06, "loss": 0.0067, "step": 31028 }, { "epoch": 5.69, "learning_rate": 4.062544866773256e-06, "loss": 0.0083, "step": 31029 }, { "epoch": 5.69, "learning_rate": 4.0619474846754045e-06, "loss": 0.0086, "step": 31030 }, { "epoch": 5.69, "learning_rate": 4.061350135308243e-06, "loss": 0.0108, "step": 31031 }, { "epoch": 5.69, "learning_rate": 4.060752818675056e-06, "loss": 0.0092, "step": 31032 }, { "epoch": 5.69, "learning_rate": 4.060155534779141e-06, "loss": 0.0058, "step": 31033 }, { "epoch": 5.69, "learning_rate": 4.059558283623787e-06, "loss": 0.0065, "step": 31034 }, { "epoch": 5.69, "learning_rate": 4.058961065212287e-06, "loss": 0.0086, "step": 31035 }, { "epoch": 5.69, "learning_rate": 4.058363879547933e-06, "loss": 0.0084, "step": 31036 }, { "epoch": 5.69, "learning_rate": 4.057766726634016e-06, "loss": 0.0095, "step": 31037 }, { "epoch": 5.69, "learning_rate": 4.057169606473828e-06, "loss": 0.0073, "step": 31038 }, { "epoch": 5.69, "learning_rate": 4.056572519070656e-06, "loss": 0.0062, "step": 31039 }, { "epoch": 5.69, "learning_rate": 4.055975464427804e-06, "loss": 0.006, "step": 31040 }, { "epoch": 5.69, "learning_rate": 4.055378442548551e-06, "loss": 0.0071, "step": 31041 }, { "epoch": 5.69, "learning_rate": 4.054781453436191e-06, "loss": 0.007, "step": 31042 }, { "epoch": 5.69, "learning_rate": 4.0541844970940106e-06, "loss": 0.007, "step": 31043 }, { "epoch": 5.69, "learning_rate": 4.053587573525308e-06, "loss": 0.0078, "step": 31044 }, { "epoch": 5.69, "learning_rate": 4.052990682733371e-06, "loss": 0.0084, "step": 31045 }, { "epoch": 5.69, "learning_rate": 4.052393824721489e-06, "loss": 0.0067, "step": 31046 }, { "epoch": 5.69, "learning_rate": 4.051796999492951e-06, "loss": 0.0065, "step": 31047 }, { "epoch": 5.69, "learning_rate": 4.051200207051047e-06, "loss": 0.0075, "step": 31048 }, { "epoch": 5.69, "learning_rate": 4.050603447399066e-06, "loss": 0.0077, "step": 31049 }, { "epoch": 5.69, "learning_rate": 4.0500067205403e-06, "loss": 0.0054, "step": 31050 }, { "epoch": 5.69, "learning_rate": 4.049410026478035e-06, "loss": 0.0065, "step": 31051 }, { "epoch": 5.69, "learning_rate": 4.0488133652155625e-06, "loss": 0.005, "step": 31052 }, { "epoch": 5.69, "learning_rate": 4.048216736756169e-06, "loss": 0.0069, "step": 31053 }, { "epoch": 5.69, "learning_rate": 4.047620141103146e-06, "loss": 0.0066, "step": 31054 }, { "epoch": 5.7, "learning_rate": 4.0470235782597775e-06, "loss": 0.0098, "step": 31055 }, { "epoch": 5.7, "learning_rate": 4.0464270482293534e-06, "loss": 0.0094, "step": 31056 }, { "epoch": 5.7, "learning_rate": 4.045830551015165e-06, "loss": 0.0085, "step": 31057 }, { "epoch": 5.7, "learning_rate": 4.045234086620502e-06, "loss": 0.0105, "step": 31058 }, { "epoch": 5.7, "learning_rate": 4.044637655048645e-06, "loss": 0.0091, "step": 31059 }, { "epoch": 5.7, "learning_rate": 4.044041256302885e-06, "loss": 0.0072, "step": 31060 }, { "epoch": 5.7, "learning_rate": 4.0434448903865055e-06, "loss": 0.009, "step": 31061 }, { "epoch": 5.7, "learning_rate": 4.042848557302799e-06, "loss": 0.0058, "step": 31062 }, { "epoch": 5.7, "learning_rate": 4.042252257055054e-06, "loss": 0.0111, "step": 31063 }, { "epoch": 5.7, "learning_rate": 4.041655989646556e-06, "loss": 0.0058, "step": 31064 }, { "epoch": 5.7, "learning_rate": 4.041059755080582e-06, "loss": 0.0079, "step": 31065 }, { "epoch": 5.7, "learning_rate": 4.040463553360431e-06, "loss": 0.0078, "step": 31066 }, { "epoch": 5.7, "learning_rate": 4.039867384489385e-06, "loss": 0.0084, "step": 31067 }, { "epoch": 5.7, "learning_rate": 4.039271248470728e-06, "loss": 0.007, "step": 31068 }, { "epoch": 5.7, "learning_rate": 4.038675145307747e-06, "loss": 0.0071, "step": 31069 }, { "epoch": 5.7, "learning_rate": 4.03807907500373e-06, "loss": 0.0072, "step": 31070 }, { "epoch": 5.7, "learning_rate": 4.03748303756196e-06, "loss": 0.008, "step": 31071 }, { "epoch": 5.7, "learning_rate": 4.036887032985723e-06, "loss": 0.0068, "step": 31072 }, { "epoch": 5.7, "learning_rate": 4.036291061278306e-06, "loss": 0.0059, "step": 31073 }, { "epoch": 5.7, "learning_rate": 4.035695122442987e-06, "loss": 0.0075, "step": 31074 }, { "epoch": 5.7, "learning_rate": 4.035099216483061e-06, "loss": 0.0091, "step": 31075 }, { "epoch": 5.7, "learning_rate": 4.03450334340181e-06, "loss": 0.0081, "step": 31076 }, { "epoch": 5.7, "learning_rate": 4.033907503202514e-06, "loss": 0.0094, "step": 31077 }, { "epoch": 5.7, "learning_rate": 4.033311695888455e-06, "loss": 0.0089, "step": 31078 }, { "epoch": 5.7, "learning_rate": 4.032715921462926e-06, "loss": 0.008, "step": 31079 }, { "epoch": 5.7, "learning_rate": 4.0321201799292065e-06, "loss": 0.0096, "step": 31080 }, { "epoch": 5.7, "learning_rate": 4.031524471290581e-06, "loss": 0.0071, "step": 31081 }, { "epoch": 5.7, "learning_rate": 4.030928795550335e-06, "loss": 0.0094, "step": 31082 }, { "epoch": 5.7, "learning_rate": 4.0303331527117425e-06, "loss": 0.009, "step": 31083 }, { "epoch": 5.7, "learning_rate": 4.029737542778097e-06, "loss": 0.0063, "step": 31084 }, { "epoch": 5.7, "learning_rate": 4.029141965752678e-06, "loss": 0.0074, "step": 31085 }, { "epoch": 5.7, "learning_rate": 4.028546421638769e-06, "loss": 0.0072, "step": 31086 }, { "epoch": 5.7, "learning_rate": 4.027950910439652e-06, "loss": 0.0065, "step": 31087 }, { "epoch": 5.7, "learning_rate": 4.027355432158609e-06, "loss": 0.0103, "step": 31088 }, { "epoch": 5.7, "learning_rate": 4.026759986798923e-06, "loss": 0.0085, "step": 31089 }, { "epoch": 5.7, "learning_rate": 4.026164574363875e-06, "loss": 0.0063, "step": 31090 }, { "epoch": 5.7, "learning_rate": 4.02556919485675e-06, "loss": 0.0071, "step": 31091 }, { "epoch": 5.7, "learning_rate": 4.024973848280822e-06, "loss": 0.0087, "step": 31092 }, { "epoch": 5.7, "learning_rate": 4.024378534639383e-06, "loss": 0.0071, "step": 31093 }, { "epoch": 5.7, "learning_rate": 4.023783253935711e-06, "loss": 0.0084, "step": 31094 }, { "epoch": 5.7, "learning_rate": 4.023188006173084e-06, "loss": 0.0078, "step": 31095 }, { "epoch": 5.7, "learning_rate": 4.0225927913547815e-06, "loss": 0.0062, "step": 31096 }, { "epoch": 5.7, "learning_rate": 4.021997609484089e-06, "loss": 0.0076, "step": 31097 }, { "epoch": 5.7, "learning_rate": 4.021402460564288e-06, "loss": 0.0093, "step": 31098 }, { "epoch": 5.7, "learning_rate": 4.020807344598655e-06, "loss": 0.0066, "step": 31099 }, { "epoch": 5.7, "learning_rate": 4.020212261590474e-06, "loss": 0.0064, "step": 31100 }, { "epoch": 5.7, "learning_rate": 4.019617211543021e-06, "loss": 0.005, "step": 31101 }, { "epoch": 5.7, "learning_rate": 4.019022194459579e-06, "loss": 0.0052, "step": 31102 }, { "epoch": 5.7, "learning_rate": 4.018427210343427e-06, "loss": 0.0072, "step": 31103 }, { "epoch": 5.7, "learning_rate": 4.017832259197843e-06, "loss": 0.0066, "step": 31104 }, { "epoch": 5.7, "learning_rate": 4.017237341026109e-06, "loss": 0.0092, "step": 31105 }, { "epoch": 5.7, "learning_rate": 4.016642455831503e-06, "loss": 0.0069, "step": 31106 }, { "epoch": 5.7, "learning_rate": 4.016047603617303e-06, "loss": 0.0087, "step": 31107 }, { "epoch": 5.7, "learning_rate": 4.0154527843867906e-06, "loss": 0.0098, "step": 31108 }, { "epoch": 5.7, "learning_rate": 4.014857998143238e-06, "loss": 0.0072, "step": 31109 }, { "epoch": 5.71, "learning_rate": 4.014263244889931e-06, "loss": 0.0071, "step": 31110 }, { "epoch": 5.71, "learning_rate": 4.01366852463015e-06, "loss": 0.0059, "step": 31111 }, { "epoch": 5.71, "learning_rate": 4.013073837367164e-06, "loss": 0.0065, "step": 31112 }, { "epoch": 5.71, "learning_rate": 4.012479183104257e-06, "loss": 0.009, "step": 31113 }, { "epoch": 5.71, "learning_rate": 4.011884561844699e-06, "loss": 0.0068, "step": 31114 }, { "epoch": 5.71, "learning_rate": 4.011289973591779e-06, "loss": 0.0072, "step": 31115 }, { "epoch": 5.71, "learning_rate": 4.010695418348768e-06, "loss": 0.0068, "step": 31116 }, { "epoch": 5.71, "learning_rate": 4.0101008961189436e-06, "loss": 0.0086, "step": 31117 }, { "epoch": 5.71, "learning_rate": 4.009506406905584e-06, "loss": 0.0071, "step": 31118 }, { "epoch": 5.71, "learning_rate": 4.008911950711967e-06, "loss": 0.0051, "step": 31119 }, { "epoch": 5.71, "learning_rate": 4.008317527541366e-06, "loss": 0.0074, "step": 31120 }, { "epoch": 5.71, "learning_rate": 4.0077231373970596e-06, "loss": 0.0083, "step": 31121 }, { "epoch": 5.71, "learning_rate": 4.0071287802823186e-06, "loss": 0.0066, "step": 31122 }, { "epoch": 5.71, "learning_rate": 4.0065344562004336e-06, "loss": 0.0076, "step": 31123 }, { "epoch": 5.71, "learning_rate": 4.005940165154666e-06, "loss": 0.0086, "step": 31124 }, { "epoch": 5.71, "learning_rate": 4.005345907148298e-06, "loss": 0.008, "step": 31125 }, { "epoch": 5.71, "learning_rate": 4.004751682184602e-06, "loss": 0.0059, "step": 31126 }, { "epoch": 5.71, "learning_rate": 4.004157490266854e-06, "loss": 0.0064, "step": 31127 }, { "epoch": 5.71, "learning_rate": 4.003563331398331e-06, "loss": 0.0065, "step": 31128 }, { "epoch": 5.71, "learning_rate": 4.002969205582314e-06, "loss": 0.007, "step": 31129 }, { "epoch": 5.71, "learning_rate": 4.002375112822065e-06, "loss": 0.0078, "step": 31130 }, { "epoch": 5.71, "learning_rate": 4.001781053120863e-06, "loss": 0.0074, "step": 31131 }, { "epoch": 5.71, "learning_rate": 4.001187026481987e-06, "loss": 0.0068, "step": 31132 }, { "epoch": 5.71, "learning_rate": 4.000593032908709e-06, "loss": 0.0086, "step": 31133 }, { "epoch": 5.71, "learning_rate": 3.999999072404304e-06, "loss": 0.0085, "step": 31134 }, { "epoch": 5.71, "learning_rate": 3.9994051449720425e-06, "loss": 0.0082, "step": 31135 }, { "epoch": 5.71, "learning_rate": 3.998811250615202e-06, "loss": 0.0085, "step": 31136 }, { "epoch": 5.71, "learning_rate": 3.9982173893370534e-06, "loss": 0.0091, "step": 31137 }, { "epoch": 5.71, "learning_rate": 3.997623561140872e-06, "loss": 0.0091, "step": 31138 }, { "epoch": 5.71, "learning_rate": 3.99702976602993e-06, "loss": 0.0085, "step": 31139 }, { "epoch": 5.71, "learning_rate": 3.996436004007496e-06, "loss": 0.0092, "step": 31140 }, { "epoch": 5.71, "learning_rate": 3.995842275076856e-06, "loss": 0.0082, "step": 31141 }, { "epoch": 5.71, "learning_rate": 3.99524857924127e-06, "loss": 0.0081, "step": 31142 }, { "epoch": 5.71, "learning_rate": 3.994654916504014e-06, "loss": 0.0092, "step": 31143 }, { "epoch": 5.71, "learning_rate": 3.994061286868361e-06, "loss": 0.0081, "step": 31144 }, { "epoch": 5.71, "learning_rate": 3.993467690337579e-06, "loss": 0.0071, "step": 31145 }, { "epoch": 5.71, "learning_rate": 3.992874126914949e-06, "loss": 0.0063, "step": 31146 }, { "epoch": 5.71, "learning_rate": 3.992280596603739e-06, "loss": 0.006, "step": 31147 }, { "epoch": 5.71, "learning_rate": 3.991687099407216e-06, "loss": 0.0095, "step": 31148 }, { "epoch": 5.71, "learning_rate": 3.991093635328652e-06, "loss": 0.0064, "step": 31149 }, { "epoch": 5.71, "learning_rate": 3.990500204371324e-06, "loss": 0.0075, "step": 31150 }, { "epoch": 5.71, "learning_rate": 3.9899068065384995e-06, "loss": 0.006, "step": 31151 }, { "epoch": 5.71, "learning_rate": 3.989313441833449e-06, "loss": 0.008, "step": 31152 }, { "epoch": 5.71, "learning_rate": 3.988720110259443e-06, "loss": 0.0091, "step": 31153 }, { "epoch": 5.71, "learning_rate": 3.988126811819753e-06, "loss": 0.0101, "step": 31154 }, { "epoch": 5.71, "learning_rate": 3.9875335465176505e-06, "loss": 0.0064, "step": 31155 }, { "epoch": 5.71, "learning_rate": 3.986940314356402e-06, "loss": 0.0071, "step": 31156 }, { "epoch": 5.71, "learning_rate": 3.986347115339281e-06, "loss": 0.0086, "step": 31157 }, { "epoch": 5.71, "learning_rate": 3.985753949469554e-06, "loss": 0.0065, "step": 31158 }, { "epoch": 5.71, "learning_rate": 3.985160816750493e-06, "loss": 0.0082, "step": 31159 }, { "epoch": 5.71, "learning_rate": 3.984567717185367e-06, "loss": 0.0062, "step": 31160 }, { "epoch": 5.71, "learning_rate": 3.983974650777444e-06, "loss": 0.0064, "step": 31161 }, { "epoch": 5.71, "learning_rate": 3.98338161752999e-06, "loss": 0.0055, "step": 31162 }, { "epoch": 5.71, "learning_rate": 3.98278861744628e-06, "loss": 0.007, "step": 31163 }, { "epoch": 5.72, "learning_rate": 3.982195650529583e-06, "loss": 0.0081, "step": 31164 }, { "epoch": 5.72, "learning_rate": 3.981602716783165e-06, "loss": 0.0109, "step": 31165 }, { "epoch": 5.72, "learning_rate": 3.981009816210292e-06, "loss": 0.0054, "step": 31166 }, { "epoch": 5.72, "learning_rate": 3.98041694881423e-06, "loss": 0.008, "step": 31167 }, { "epoch": 5.72, "learning_rate": 3.979824114598254e-06, "loss": 0.0075, "step": 31168 }, { "epoch": 5.72, "learning_rate": 3.9792313135656295e-06, "loss": 0.0064, "step": 31169 }, { "epoch": 5.72, "learning_rate": 3.978638545719622e-06, "loss": 0.0089, "step": 31170 }, { "epoch": 5.72, "learning_rate": 3.9780458110635005e-06, "loss": 0.007, "step": 31171 }, { "epoch": 5.72, "learning_rate": 3.977453109600531e-06, "loss": 0.0067, "step": 31172 }, { "epoch": 5.72, "learning_rate": 3.976860441333982e-06, "loss": 0.01, "step": 31173 }, { "epoch": 5.72, "learning_rate": 3.976267806267118e-06, "loss": 0.0063, "step": 31174 }, { "epoch": 5.72, "learning_rate": 3.975675204403204e-06, "loss": 0.0067, "step": 31175 }, { "epoch": 5.72, "learning_rate": 3.975082635745516e-06, "loss": 0.0068, "step": 31176 }, { "epoch": 5.72, "learning_rate": 3.974490100297311e-06, "loss": 0.0065, "step": 31177 }, { "epoch": 5.72, "learning_rate": 3.973897598061858e-06, "loss": 0.0073, "step": 31178 }, { "epoch": 5.72, "learning_rate": 3.973305129042422e-06, "loss": 0.0072, "step": 31179 }, { "epoch": 5.72, "learning_rate": 3.972712693242267e-06, "loss": 0.0079, "step": 31180 }, { "epoch": 5.72, "learning_rate": 3.972120290664663e-06, "loss": 0.0084, "step": 31181 }, { "epoch": 5.72, "learning_rate": 3.971527921312874e-06, "loss": 0.0083, "step": 31182 }, { "epoch": 5.72, "learning_rate": 3.970935585190167e-06, "loss": 0.0094, "step": 31183 }, { "epoch": 5.72, "learning_rate": 3.970343282299796e-06, "loss": 0.0092, "step": 31184 }, { "epoch": 5.72, "learning_rate": 3.96975101264504e-06, "loss": 0.0079, "step": 31185 }, { "epoch": 5.72, "learning_rate": 3.9691587762291576e-06, "loss": 0.0095, "step": 31186 }, { "epoch": 5.72, "learning_rate": 3.968566573055412e-06, "loss": 0.0079, "step": 31187 }, { "epoch": 5.72, "learning_rate": 3.96797440312707e-06, "loss": 0.0057, "step": 31188 }, { "epoch": 5.72, "learning_rate": 3.967382266447393e-06, "loss": 0.0056, "step": 31189 }, { "epoch": 5.72, "learning_rate": 3.966790163019647e-06, "loss": 0.0098, "step": 31190 }, { "epoch": 5.72, "learning_rate": 3.966198092847095e-06, "loss": 0.0062, "step": 31191 }, { "epoch": 5.72, "learning_rate": 3.965606055933001e-06, "loss": 0.01, "step": 31192 }, { "epoch": 5.72, "learning_rate": 3.965014052280624e-06, "loss": 0.0065, "step": 31193 }, { "epoch": 5.72, "learning_rate": 3.9644220818932365e-06, "loss": 0.0107, "step": 31194 }, { "epoch": 5.72, "learning_rate": 3.963830144774094e-06, "loss": 0.0074, "step": 31195 }, { "epoch": 5.72, "learning_rate": 3.963238240926461e-06, "loss": 0.0073, "step": 31196 }, { "epoch": 5.72, "learning_rate": 3.962646370353597e-06, "loss": 0.0098, "step": 31197 }, { "epoch": 5.72, "learning_rate": 3.96205453305877e-06, "loss": 0.0057, "step": 31198 }, { "epoch": 5.72, "learning_rate": 3.96146272904524e-06, "loss": 0.0093, "step": 31199 }, { "epoch": 5.72, "learning_rate": 3.96087095831627e-06, "loss": 0.0081, "step": 31200 }, { "epoch": 5.72, "learning_rate": 3.9602792208751235e-06, "loss": 0.0062, "step": 31201 }, { "epoch": 5.72, "learning_rate": 3.959687516725053e-06, "loss": 0.0081, "step": 31202 }, { "epoch": 5.72, "learning_rate": 3.9590958458693285e-06, "loss": 0.0067, "step": 31203 }, { "epoch": 5.72, "learning_rate": 3.958504208311211e-06, "loss": 0.0055, "step": 31204 }, { "epoch": 5.72, "learning_rate": 3.957912604053959e-06, "loss": 0.011, "step": 31205 }, { "epoch": 5.72, "learning_rate": 3.957321033100834e-06, "loss": 0.0084, "step": 31206 }, { "epoch": 5.72, "learning_rate": 3.956729495455097e-06, "loss": 0.0069, "step": 31207 }, { "epoch": 5.72, "learning_rate": 3.956137991120009e-06, "loss": 0.0089, "step": 31208 }, { "epoch": 5.72, "learning_rate": 3.955546520098831e-06, "loss": 0.0071, "step": 31209 }, { "epoch": 5.72, "learning_rate": 3.95495508239482e-06, "loss": 0.0072, "step": 31210 }, { "epoch": 5.72, "learning_rate": 3.954363678011236e-06, "loss": 0.008, "step": 31211 }, { "epoch": 5.72, "learning_rate": 3.953772306951348e-06, "loss": 0.0091, "step": 31212 }, { "epoch": 5.72, "learning_rate": 3.953180969218405e-06, "loss": 0.0068, "step": 31213 }, { "epoch": 5.72, "learning_rate": 3.9525896648156706e-06, "loss": 0.0083, "step": 31214 }, { "epoch": 5.72, "learning_rate": 3.9519983937464e-06, "loss": 0.011, "step": 31215 }, { "epoch": 5.72, "learning_rate": 3.951407156013859e-06, "loss": 0.0077, "step": 31216 }, { "epoch": 5.72, "learning_rate": 3.950815951621304e-06, "loss": 0.0056, "step": 31217 }, { "epoch": 5.72, "learning_rate": 3.950224780571996e-06, "loss": 0.0064, "step": 31218 }, { "epoch": 5.73, "learning_rate": 3.949633642869187e-06, "loss": 0.0088, "step": 31219 }, { "epoch": 5.73, "learning_rate": 3.949042538516137e-06, "loss": 0.0075, "step": 31220 }, { "epoch": 5.73, "learning_rate": 3.948451467516109e-06, "loss": 0.0078, "step": 31221 }, { "epoch": 5.73, "learning_rate": 3.947860429872358e-06, "loss": 0.0085, "step": 31222 }, { "epoch": 5.73, "learning_rate": 3.947269425588143e-06, "loss": 0.0092, "step": 31223 }, { "epoch": 5.73, "learning_rate": 3.946678454666719e-06, "loss": 0.0061, "step": 31224 }, { "epoch": 5.73, "learning_rate": 3.9460875171113464e-06, "loss": 0.0085, "step": 31225 }, { "epoch": 5.73, "learning_rate": 3.945496612925282e-06, "loss": 0.0075, "step": 31226 }, { "epoch": 5.73, "learning_rate": 3.944905742111781e-06, "loss": 0.0047, "step": 31227 }, { "epoch": 5.73, "learning_rate": 3.944314904674098e-06, "loss": 0.0071, "step": 31228 }, { "epoch": 5.73, "learning_rate": 3.943724100615497e-06, "loss": 0.0071, "step": 31229 }, { "epoch": 5.73, "learning_rate": 3.943133329939234e-06, "loss": 0.0087, "step": 31230 }, { "epoch": 5.73, "learning_rate": 3.942542592648558e-06, "loss": 0.0054, "step": 31231 }, { "epoch": 5.73, "learning_rate": 3.94195188874673e-06, "loss": 0.0055, "step": 31232 }, { "epoch": 5.73, "learning_rate": 3.941361218237001e-06, "loss": 0.0114, "step": 31233 }, { "epoch": 5.73, "learning_rate": 3.940770581122634e-06, "loss": 0.0063, "step": 31234 }, { "epoch": 5.73, "learning_rate": 3.940179977406883e-06, "loss": 0.0099, "step": 31235 }, { "epoch": 5.73, "learning_rate": 3.939589407093004e-06, "loss": 0.0075, "step": 31236 }, { "epoch": 5.73, "learning_rate": 3.938998870184243e-06, "loss": 0.0099, "step": 31237 }, { "epoch": 5.73, "learning_rate": 3.938408366683867e-06, "loss": 0.0067, "step": 31238 }, { "epoch": 5.73, "learning_rate": 3.937817896595125e-06, "loss": 0.0089, "step": 31239 }, { "epoch": 5.73, "learning_rate": 3.937227459921274e-06, "loss": 0.0086, "step": 31240 }, { "epoch": 5.73, "learning_rate": 3.936637056665566e-06, "loss": 0.0077, "step": 31241 }, { "epoch": 5.73, "learning_rate": 3.936046686831257e-06, "loss": 0.0094, "step": 31242 }, { "epoch": 5.73, "learning_rate": 3.935456350421601e-06, "loss": 0.0099, "step": 31243 }, { "epoch": 5.73, "learning_rate": 3.934866047439851e-06, "loss": 0.0082, "step": 31244 }, { "epoch": 5.73, "learning_rate": 3.934275777889263e-06, "loss": 0.0066, "step": 31245 }, { "epoch": 5.73, "learning_rate": 3.933685541773084e-06, "loss": 0.0083, "step": 31246 }, { "epoch": 5.73, "learning_rate": 3.933095339094576e-06, "loss": 0.0099, "step": 31247 }, { "epoch": 5.73, "learning_rate": 3.932505169856993e-06, "loss": 0.0099, "step": 31248 }, { "epoch": 5.73, "learning_rate": 3.931915034063579e-06, "loss": 0.008, "step": 31249 }, { "epoch": 5.73, "learning_rate": 3.931324931717588e-06, "loss": 0.007, "step": 31250 }, { "epoch": 5.73, "learning_rate": 3.930734862822281e-06, "loss": 0.0085, "step": 31251 }, { "epoch": 5.73, "learning_rate": 3.930144827380905e-06, "loss": 0.0065, "step": 31252 }, { "epoch": 5.73, "learning_rate": 3.929554825396713e-06, "loss": 0.0059, "step": 31253 }, { "epoch": 5.73, "learning_rate": 3.92896485687296e-06, "loss": 0.0094, "step": 31254 }, { "epoch": 5.73, "learning_rate": 3.9283749218128885e-06, "loss": 0.0093, "step": 31255 }, { "epoch": 5.73, "learning_rate": 3.9277850202197595e-06, "loss": 0.0081, "step": 31256 }, { "epoch": 5.73, "learning_rate": 3.927195152096822e-06, "loss": 0.0066, "step": 31257 }, { "epoch": 5.73, "learning_rate": 3.926605317447329e-06, "loss": 0.0053, "step": 31258 }, { "epoch": 5.73, "learning_rate": 3.926015516274527e-06, "loss": 0.0073, "step": 31259 }, { "epoch": 5.73, "learning_rate": 3.925425748581671e-06, "loss": 0.0082, "step": 31260 }, { "epoch": 5.73, "learning_rate": 3.92483601437201e-06, "loss": 0.0071, "step": 31261 }, { "epoch": 5.73, "learning_rate": 3.924246313648796e-06, "loss": 0.0062, "step": 31262 }, { "epoch": 5.73, "learning_rate": 3.923656646415278e-06, "loss": 0.0087, "step": 31263 }, { "epoch": 5.73, "learning_rate": 3.923067012674704e-06, "loss": 0.0055, "step": 31264 }, { "epoch": 5.73, "learning_rate": 3.922477412430333e-06, "loss": 0.0069, "step": 31265 }, { "epoch": 5.73, "learning_rate": 3.921887845685407e-06, "loss": 0.0094, "step": 31266 }, { "epoch": 5.73, "learning_rate": 3.921298312443176e-06, "loss": 0.0088, "step": 31267 }, { "epoch": 5.73, "learning_rate": 3.920708812706889e-06, "loss": 0.0066, "step": 31268 }, { "epoch": 5.73, "learning_rate": 3.920119346479799e-06, "loss": 0.0076, "step": 31269 }, { "epoch": 5.73, "learning_rate": 3.919529913765155e-06, "loss": 0.0078, "step": 31270 }, { "epoch": 5.73, "learning_rate": 3.918940514566205e-06, "loss": 0.0093, "step": 31271 }, { "epoch": 5.73, "learning_rate": 3.9183511488862005e-06, "loss": 0.0092, "step": 31272 }, { "epoch": 5.74, "learning_rate": 3.917761816728379e-06, "loss": 0.0079, "step": 31273 }, { "epoch": 5.74, "learning_rate": 3.917172518096001e-06, "loss": 0.0074, "step": 31274 }, { "epoch": 5.74, "learning_rate": 3.916583252992311e-06, "loss": 0.0077, "step": 31275 }, { "epoch": 5.74, "learning_rate": 3.915994021420556e-06, "loss": 0.0056, "step": 31276 }, { "epoch": 5.74, "learning_rate": 3.915404823383985e-06, "loss": 0.0077, "step": 31277 }, { "epoch": 5.74, "learning_rate": 3.914815658885844e-06, "loss": 0.0069, "step": 31278 }, { "epoch": 5.74, "learning_rate": 3.914226527929383e-06, "loss": 0.007, "step": 31279 }, { "epoch": 5.74, "learning_rate": 3.9136374305178486e-06, "loss": 0.0078, "step": 31280 }, { "epoch": 5.74, "learning_rate": 3.913048366654482e-06, "loss": 0.0067, "step": 31281 }, { "epoch": 5.74, "learning_rate": 3.9124593363425385e-06, "loss": 0.0079, "step": 31282 }, { "epoch": 5.74, "learning_rate": 3.9118703395852674e-06, "loss": 0.0079, "step": 31283 }, { "epoch": 5.74, "learning_rate": 3.911281376385905e-06, "loss": 0.0074, "step": 31284 }, { "epoch": 5.74, "learning_rate": 3.910692446747703e-06, "loss": 0.0049, "step": 31285 }, { "epoch": 5.74, "learning_rate": 3.9101035506739015e-06, "loss": 0.0055, "step": 31286 }, { "epoch": 5.74, "learning_rate": 3.909514688167757e-06, "loss": 0.008, "step": 31287 }, { "epoch": 5.74, "learning_rate": 3.9089258592325095e-06, "loss": 0.0058, "step": 31288 }, { "epoch": 5.74, "learning_rate": 3.908337063871406e-06, "loss": 0.0073, "step": 31289 }, { "epoch": 5.74, "learning_rate": 3.907748302087692e-06, "loss": 0.0068, "step": 31290 }, { "epoch": 5.74, "learning_rate": 3.90715957388461e-06, "loss": 0.0061, "step": 31291 }, { "epoch": 5.74, "learning_rate": 3.906570879265409e-06, "loss": 0.0075, "step": 31292 }, { "epoch": 5.74, "learning_rate": 3.9059822182333305e-06, "loss": 0.0064, "step": 31293 }, { "epoch": 5.74, "learning_rate": 3.9053935907916215e-06, "loss": 0.0081, "step": 31294 }, { "epoch": 5.74, "learning_rate": 3.904804996943526e-06, "loss": 0.0092, "step": 31295 }, { "epoch": 5.74, "learning_rate": 3.904216436692288e-06, "loss": 0.0081, "step": 31296 }, { "epoch": 5.74, "learning_rate": 3.903627910041152e-06, "loss": 0.0065, "step": 31297 }, { "epoch": 5.74, "learning_rate": 3.9030394169933625e-06, "loss": 0.0089, "step": 31298 }, { "epoch": 5.74, "learning_rate": 3.902450957552158e-06, "loss": 0.0066, "step": 31299 }, { "epoch": 5.74, "learning_rate": 3.901862531720791e-06, "loss": 0.0085, "step": 31300 }, { "epoch": 5.74, "learning_rate": 3.901274139502504e-06, "loss": 0.0087, "step": 31301 }, { "epoch": 5.74, "learning_rate": 3.900685780900532e-06, "loss": 0.0083, "step": 31302 }, { "epoch": 5.74, "learning_rate": 3.90009745591812e-06, "loss": 0.0081, "step": 31303 }, { "epoch": 5.74, "learning_rate": 3.899509164558517e-06, "loss": 0.0092, "step": 31304 }, { "epoch": 5.74, "learning_rate": 3.898920906824964e-06, "loss": 0.008, "step": 31305 }, { "epoch": 5.74, "learning_rate": 3.8983326827207e-06, "loss": 0.009, "step": 31306 }, { "epoch": 5.74, "learning_rate": 3.89774449224897e-06, "loss": 0.0078, "step": 31307 }, { "epoch": 5.74, "learning_rate": 3.897156335413015e-06, "loss": 0.0061, "step": 31308 }, { "epoch": 5.74, "learning_rate": 3.896568212216078e-06, "loss": 0.0071, "step": 31309 }, { "epoch": 5.74, "learning_rate": 3.895980122661399e-06, "loss": 0.0072, "step": 31310 }, { "epoch": 5.74, "learning_rate": 3.895392066752222e-06, "loss": 0.0077, "step": 31311 }, { "epoch": 5.74, "learning_rate": 3.894804044491786e-06, "loss": 0.0089, "step": 31312 }, { "epoch": 5.74, "learning_rate": 3.894216055883333e-06, "loss": 0.0081, "step": 31313 }, { "epoch": 5.74, "learning_rate": 3.893628100930103e-06, "loss": 0.0057, "step": 31314 }, { "epoch": 5.74, "learning_rate": 3.89304017963534e-06, "loss": 0.0078, "step": 31315 }, { "epoch": 5.74, "learning_rate": 3.8924522920022825e-06, "loss": 0.0073, "step": 31316 }, { "epoch": 5.74, "learning_rate": 3.891864438034167e-06, "loss": 0.0085, "step": 31317 }, { "epoch": 5.74, "learning_rate": 3.891276617734241e-06, "loss": 0.0075, "step": 31318 }, { "epoch": 5.74, "learning_rate": 3.890688831105744e-06, "loss": 0.005, "step": 31319 }, { "epoch": 5.74, "learning_rate": 3.890101078151912e-06, "loss": 0.007, "step": 31320 }, { "epoch": 5.74, "learning_rate": 3.889513358875981e-06, "loss": 0.008, "step": 31321 }, { "epoch": 5.74, "learning_rate": 3.8889256732812e-06, "loss": 0.0095, "step": 31322 }, { "epoch": 5.74, "learning_rate": 3.888338021370803e-06, "loss": 0.0085, "step": 31323 }, { "epoch": 5.74, "learning_rate": 3.887750403148031e-06, "loss": 0.0086, "step": 31324 }, { "epoch": 5.74, "learning_rate": 3.88716281861612e-06, "loss": 0.0092, "step": 31325 }, { "epoch": 5.74, "learning_rate": 3.886575267778313e-06, "loss": 0.0084, "step": 31326 }, { "epoch": 5.74, "learning_rate": 3.885987750637845e-06, "loss": 0.0079, "step": 31327 }, { "epoch": 5.75, "learning_rate": 3.885400267197956e-06, "loss": 0.0056, "step": 31328 }, { "epoch": 5.75, "learning_rate": 3.8848128174618835e-06, "loss": 0.0056, "step": 31329 }, { "epoch": 5.75, "learning_rate": 3.884225401432867e-06, "loss": 0.006, "step": 31330 }, { "epoch": 5.75, "learning_rate": 3.883638019114143e-06, "loss": 0.0074, "step": 31331 }, { "epoch": 5.75, "learning_rate": 3.88305067050895e-06, "loss": 0.0061, "step": 31332 }, { "epoch": 5.75, "learning_rate": 3.8824633556205235e-06, "loss": 0.0085, "step": 31333 }, { "epoch": 5.75, "learning_rate": 3.881876074452101e-06, "loss": 0.0074, "step": 31334 }, { "epoch": 5.75, "learning_rate": 3.881288827006923e-06, "loss": 0.0065, "step": 31335 }, { "epoch": 5.75, "learning_rate": 3.880701613288225e-06, "loss": 0.008, "step": 31336 }, { "epoch": 5.75, "learning_rate": 3.880114433299245e-06, "loss": 0.0057, "step": 31337 }, { "epoch": 5.75, "learning_rate": 3.879527287043215e-06, "loss": 0.0054, "step": 31338 }, { "epoch": 5.75, "learning_rate": 3.878940174523371e-06, "loss": 0.0065, "step": 31339 }, { "epoch": 5.75, "learning_rate": 3.878353095742956e-06, "loss": 0.0071, "step": 31340 }, { "epoch": 5.75, "learning_rate": 3.877766050705201e-06, "loss": 0.0064, "step": 31341 }, { "epoch": 5.75, "learning_rate": 3.877179039413343e-06, "loss": 0.0073, "step": 31342 }, { "epoch": 5.75, "learning_rate": 3.876592061870619e-06, "loss": 0.0073, "step": 31343 }, { "epoch": 5.75, "learning_rate": 3.876005118080263e-06, "loss": 0.0074, "step": 31344 }, { "epoch": 5.75, "learning_rate": 3.875418208045508e-06, "loss": 0.0076, "step": 31345 }, { "epoch": 5.75, "learning_rate": 3.874831331769593e-06, "loss": 0.0089, "step": 31346 }, { "epoch": 5.75, "learning_rate": 3.874244489255751e-06, "loss": 0.0073, "step": 31347 }, { "epoch": 5.75, "learning_rate": 3.8736576805072165e-06, "loss": 0.0071, "step": 31348 }, { "epoch": 5.75, "learning_rate": 3.8730709055272255e-06, "loss": 0.0086, "step": 31349 }, { "epoch": 5.75, "learning_rate": 3.87248416431901e-06, "loss": 0.0077, "step": 31350 }, { "epoch": 5.75, "learning_rate": 3.871897456885806e-06, "loss": 0.0081, "step": 31351 }, { "epoch": 5.75, "learning_rate": 3.871310783230843e-06, "loss": 0.0075, "step": 31352 }, { "epoch": 5.75, "learning_rate": 3.870724143357362e-06, "loss": 0.0076, "step": 31353 }, { "epoch": 5.75, "learning_rate": 3.870137537268592e-06, "loss": 0.0093, "step": 31354 }, { "epoch": 5.75, "learning_rate": 3.869550964967773e-06, "loss": 0.0054, "step": 31355 }, { "epoch": 5.75, "learning_rate": 3.868964426458125e-06, "loss": 0.0089, "step": 31356 }, { "epoch": 5.75, "learning_rate": 3.868377921742892e-06, "loss": 0.0062, "step": 31357 }, { "epoch": 5.75, "learning_rate": 3.867791450825304e-06, "loss": 0.0084, "step": 31358 }, { "epoch": 5.75, "learning_rate": 3.867205013708592e-06, "loss": 0.0079, "step": 31359 }, { "epoch": 5.75, "learning_rate": 3.86661861039599e-06, "loss": 0.0063, "step": 31360 }, { "epoch": 5.75, "learning_rate": 3.8660322408907305e-06, "loss": 0.0086, "step": 31361 }, { "epoch": 5.75, "learning_rate": 3.865445905196044e-06, "loss": 0.0072, "step": 31362 }, { "epoch": 5.75, "learning_rate": 3.864859603315165e-06, "loss": 0.0072, "step": 31363 }, { "epoch": 5.75, "learning_rate": 3.864273335251323e-06, "loss": 0.0092, "step": 31364 }, { "epoch": 5.75, "learning_rate": 3.863687101007747e-06, "loss": 0.0072, "step": 31365 }, { "epoch": 5.75, "learning_rate": 3.863100900587677e-06, "loss": 0.0059, "step": 31366 }, { "epoch": 5.75, "learning_rate": 3.862514733994337e-06, "loss": 0.0073, "step": 31367 }, { "epoch": 5.75, "learning_rate": 3.861928601230957e-06, "loss": 0.0069, "step": 31368 }, { "epoch": 5.75, "learning_rate": 3.861342502300773e-06, "loss": 0.0059, "step": 31369 }, { "epoch": 5.75, "learning_rate": 3.860756437207009e-06, "loss": 0.0071, "step": 31370 }, { "epoch": 5.75, "learning_rate": 3.860170405952901e-06, "loss": 0.0065, "step": 31371 }, { "epoch": 5.75, "learning_rate": 3.859584408541682e-06, "loss": 0.0069, "step": 31372 }, { "epoch": 5.75, "learning_rate": 3.858998444976575e-06, "loss": 0.0073, "step": 31373 }, { "epoch": 5.75, "learning_rate": 3.858412515260809e-06, "loss": 0.0071, "step": 31374 }, { "epoch": 5.75, "learning_rate": 3.857826619397619e-06, "loss": 0.006, "step": 31375 }, { "epoch": 5.75, "learning_rate": 3.857240757390234e-06, "loss": 0.0076, "step": 31376 }, { "epoch": 5.75, "learning_rate": 3.8566549292418806e-06, "loss": 0.0072, "step": 31377 }, { "epoch": 5.75, "learning_rate": 3.85606913495579e-06, "loss": 0.007, "step": 31378 }, { "epoch": 5.75, "learning_rate": 3.855483374535191e-06, "loss": 0.0076, "step": 31379 }, { "epoch": 5.75, "learning_rate": 3.85489764798331e-06, "loss": 0.0088, "step": 31380 }, { "epoch": 5.75, "learning_rate": 3.854311955303378e-06, "loss": 0.0053, "step": 31381 }, { "epoch": 5.75, "learning_rate": 3.853726296498622e-06, "loss": 0.0067, "step": 31382 }, { "epoch": 5.76, "learning_rate": 3.853140671572268e-06, "loss": 0.0081, "step": 31383 }, { "epoch": 5.76, "learning_rate": 3.852555080527553e-06, "loss": 0.0071, "step": 31384 }, { "epoch": 5.76, "learning_rate": 3.851969523367695e-06, "loss": 0.0055, "step": 31385 }, { "epoch": 5.76, "learning_rate": 3.8513840000959244e-06, "loss": 0.0096, "step": 31386 }, { "epoch": 5.76, "learning_rate": 3.850798510715466e-06, "loss": 0.0064, "step": 31387 }, { "epoch": 5.76, "learning_rate": 3.850213055229556e-06, "loss": 0.0077, "step": 31388 }, { "epoch": 5.76, "learning_rate": 3.849627633641413e-06, "loss": 0.0087, "step": 31389 }, { "epoch": 5.76, "learning_rate": 3.849042245954271e-06, "loss": 0.0072, "step": 31390 }, { "epoch": 5.76, "learning_rate": 3.848456892171348e-06, "loss": 0.0058, "step": 31391 }, { "epoch": 5.76, "learning_rate": 3.847871572295873e-06, "loss": 0.0081, "step": 31392 }, { "epoch": 5.76, "learning_rate": 3.8472862863310755e-06, "loss": 0.0087, "step": 31393 }, { "epoch": 5.76, "learning_rate": 3.846701034280182e-06, "loss": 0.0056, "step": 31394 }, { "epoch": 5.76, "learning_rate": 3.846115816146415e-06, "loss": 0.0092, "step": 31395 }, { "epoch": 5.76, "learning_rate": 3.845530631933003e-06, "loss": 0.0069, "step": 31396 }, { "epoch": 5.76, "learning_rate": 3.844945481643169e-06, "loss": 0.0087, "step": 31397 }, { "epoch": 5.76, "learning_rate": 3.844360365280141e-06, "loss": 0.0073, "step": 31398 }, { "epoch": 5.76, "learning_rate": 3.843775282847142e-06, "loss": 0.0063, "step": 31399 }, { "epoch": 5.76, "learning_rate": 3.8431902343473985e-06, "loss": 0.0066, "step": 31400 }, { "epoch": 5.76, "learning_rate": 3.842605219784131e-06, "loss": 0.0085, "step": 31401 }, { "epoch": 5.76, "learning_rate": 3.842020239160575e-06, "loss": 0.0086, "step": 31402 }, { "epoch": 5.76, "learning_rate": 3.841435292479944e-06, "loss": 0.0081, "step": 31403 }, { "epoch": 5.76, "learning_rate": 3.840850379745466e-06, "loss": 0.0091, "step": 31404 }, { "epoch": 5.76, "learning_rate": 3.840265500960361e-06, "loss": 0.0067, "step": 31405 }, { "epoch": 5.76, "learning_rate": 3.83968065612786e-06, "loss": 0.0068, "step": 31406 }, { "epoch": 5.76, "learning_rate": 3.839095845251184e-06, "loss": 0.0082, "step": 31407 }, { "epoch": 5.76, "learning_rate": 3.83851106833356e-06, "loss": 0.0066, "step": 31408 }, { "epoch": 5.76, "learning_rate": 3.837926325378201e-06, "loss": 0.0089, "step": 31409 }, { "epoch": 5.76, "learning_rate": 3.837341616388338e-06, "loss": 0.0072, "step": 31410 }, { "epoch": 5.76, "learning_rate": 3.836756941367194e-06, "loss": 0.0066, "step": 31411 }, { "epoch": 5.76, "learning_rate": 3.836172300317989e-06, "loss": 0.01, "step": 31412 }, { "epoch": 5.76, "learning_rate": 3.8355876932439475e-06, "loss": 0.0057, "step": 31413 }, { "epoch": 5.76, "learning_rate": 3.835003120148291e-06, "loss": 0.0084, "step": 31414 }, { "epoch": 5.76, "learning_rate": 3.834418581034242e-06, "loss": 0.0059, "step": 31415 }, { "epoch": 5.76, "learning_rate": 3.8338340759050205e-06, "loss": 0.0071, "step": 31416 }, { "epoch": 5.76, "learning_rate": 3.833249604763851e-06, "loss": 0.0091, "step": 31417 }, { "epoch": 5.76, "learning_rate": 3.832665167613952e-06, "loss": 0.0068, "step": 31418 }, { "epoch": 5.76, "learning_rate": 3.8320807644585525e-06, "loss": 0.0081, "step": 31419 }, { "epoch": 5.76, "learning_rate": 3.8314963953008655e-06, "loss": 0.007, "step": 31420 }, { "epoch": 5.76, "learning_rate": 3.830912060144115e-06, "loss": 0.007, "step": 31421 }, { "epoch": 5.76, "learning_rate": 3.830327758991521e-06, "loss": 0.0069, "step": 31422 }, { "epoch": 5.76, "learning_rate": 3.829743491846301e-06, "loss": 0.0082, "step": 31423 }, { "epoch": 5.76, "learning_rate": 3.8291592587116855e-06, "loss": 0.0072, "step": 31424 }, { "epoch": 5.76, "learning_rate": 3.828575059590886e-06, "loss": 0.0062, "step": 31425 }, { "epoch": 5.76, "learning_rate": 3.82799089448713e-06, "loss": 0.0084, "step": 31426 }, { "epoch": 5.76, "learning_rate": 3.827406763403626e-06, "loss": 0.0064, "step": 31427 }, { "epoch": 5.76, "learning_rate": 3.826822666343603e-06, "loss": 0.0056, "step": 31428 }, { "epoch": 5.76, "learning_rate": 3.8262386033102794e-06, "loss": 0.006, "step": 31429 }, { "epoch": 5.76, "learning_rate": 3.825654574306873e-06, "loss": 0.0094, "step": 31430 }, { "epoch": 5.76, "learning_rate": 3.8250705793366026e-06, "loss": 0.0076, "step": 31431 }, { "epoch": 5.76, "learning_rate": 3.824486618402689e-06, "loss": 0.0055, "step": 31432 }, { "epoch": 5.76, "learning_rate": 3.823902691508349e-06, "loss": 0.007, "step": 31433 }, { "epoch": 5.76, "learning_rate": 3.823318798656801e-06, "loss": 0.0091, "step": 31434 }, { "epoch": 5.76, "learning_rate": 3.822734939851266e-06, "loss": 0.0088, "step": 31435 }, { "epoch": 5.76, "learning_rate": 3.822151115094956e-06, "loss": 0.0064, "step": 31436 }, { "epoch": 5.77, "learning_rate": 3.821567324391102e-06, "loss": 0.0083, "step": 31437 }, { "epoch": 5.77, "learning_rate": 3.8209835677429094e-06, "loss": 0.0083, "step": 31438 }, { "epoch": 5.77, "learning_rate": 3.820399845153601e-06, "loss": 0.0054, "step": 31439 }, { "epoch": 5.77, "learning_rate": 3.819816156626389e-06, "loss": 0.0059, "step": 31440 }, { "epoch": 5.77, "learning_rate": 3.819232502164499e-06, "loss": 0.0088, "step": 31441 }, { "epoch": 5.77, "learning_rate": 3.818648881771145e-06, "loss": 0.006, "step": 31442 }, { "epoch": 5.77, "learning_rate": 3.818065295449542e-06, "loss": 0.0095, "step": 31443 }, { "epoch": 5.77, "learning_rate": 3.8174817432029125e-06, "loss": 0.0081, "step": 31444 }, { "epoch": 5.77, "learning_rate": 3.81689822503446e-06, "loss": 0.0083, "step": 31445 }, { "epoch": 5.77, "learning_rate": 3.816314740947413e-06, "loss": 0.0079, "step": 31446 }, { "epoch": 5.77, "learning_rate": 3.815731290944985e-06, "loss": 0.0067, "step": 31447 }, { "epoch": 5.77, "learning_rate": 3.815147875030391e-06, "loss": 0.0075, "step": 31448 }, { "epoch": 5.77, "learning_rate": 3.814564493206846e-06, "loss": 0.0072, "step": 31449 }, { "epoch": 5.77, "learning_rate": 3.8139811454775677e-06, "loss": 0.0089, "step": 31450 }, { "epoch": 5.77, "learning_rate": 3.813397831845769e-06, "loss": 0.0061, "step": 31451 }, { "epoch": 5.77, "learning_rate": 3.8128145523146664e-06, "loss": 0.0052, "step": 31452 }, { "epoch": 5.77, "learning_rate": 3.812231306887473e-06, "loss": 0.0064, "step": 31453 }, { "epoch": 5.77, "learning_rate": 3.811648095567407e-06, "loss": 0.0071, "step": 31454 }, { "epoch": 5.77, "learning_rate": 3.8110649183576866e-06, "loss": 0.0078, "step": 31455 }, { "epoch": 5.77, "learning_rate": 3.8104817752615174e-06, "loss": 0.0059, "step": 31456 }, { "epoch": 5.77, "learning_rate": 3.8098986662821168e-06, "loss": 0.0081, "step": 31457 }, { "epoch": 5.77, "learning_rate": 3.8093155914226974e-06, "loss": 0.0069, "step": 31458 }, { "epoch": 5.77, "learning_rate": 3.8087325506864793e-06, "loss": 0.0085, "step": 31459 }, { "epoch": 5.77, "learning_rate": 3.8081495440766715e-06, "loss": 0.0078, "step": 31460 }, { "epoch": 5.77, "learning_rate": 3.8075665715964892e-06, "loss": 0.0053, "step": 31461 }, { "epoch": 5.77, "learning_rate": 3.8069836332491437e-06, "loss": 0.0059, "step": 31462 }, { "epoch": 5.77, "learning_rate": 3.8064007290378515e-06, "loss": 0.0072, "step": 31463 }, { "epoch": 5.77, "learning_rate": 3.8058178589658224e-06, "loss": 0.0105, "step": 31464 }, { "epoch": 5.77, "learning_rate": 3.80523502303627e-06, "loss": 0.0061, "step": 31465 }, { "epoch": 5.77, "learning_rate": 3.8046522212524086e-06, "loss": 0.0076, "step": 31466 }, { "epoch": 5.77, "learning_rate": 3.8040694536174482e-06, "loss": 0.0058, "step": 31467 }, { "epoch": 5.77, "learning_rate": 3.8034867201346026e-06, "loss": 0.0067, "step": 31468 }, { "epoch": 5.77, "learning_rate": 3.8029040208070833e-06, "loss": 0.0056, "step": 31469 }, { "epoch": 5.77, "learning_rate": 3.8023213556381013e-06, "loss": 0.0069, "step": 31470 }, { "epoch": 5.77, "learning_rate": 3.801738724630867e-06, "loss": 0.0057, "step": 31471 }, { "epoch": 5.77, "learning_rate": 3.8011561277885965e-06, "loss": 0.0086, "step": 31472 }, { "epoch": 5.77, "learning_rate": 3.8005735651145026e-06, "loss": 0.0078, "step": 31473 }, { "epoch": 5.77, "learning_rate": 3.79999103661179e-06, "loss": 0.0079, "step": 31474 }, { "epoch": 5.77, "learning_rate": 3.7994085422836703e-06, "loss": 0.0059, "step": 31475 }, { "epoch": 5.77, "learning_rate": 3.7988260821333534e-06, "loss": 0.0102, "step": 31476 }, { "epoch": 5.77, "learning_rate": 3.7982436561640558e-06, "loss": 0.0064, "step": 31477 }, { "epoch": 5.77, "learning_rate": 3.7976612643789847e-06, "loss": 0.0075, "step": 31478 }, { "epoch": 5.77, "learning_rate": 3.7970789067813497e-06, "loss": 0.0072, "step": 31479 }, { "epoch": 5.77, "learning_rate": 3.7964965833743606e-06, "loss": 0.0059, "step": 31480 }, { "epoch": 5.77, "learning_rate": 3.795914294161228e-06, "loss": 0.0069, "step": 31481 }, { "epoch": 5.77, "learning_rate": 3.795332039145162e-06, "loss": 0.0085, "step": 31482 }, { "epoch": 5.77, "learning_rate": 3.794749818329371e-06, "loss": 0.0082, "step": 31483 }, { "epoch": 5.77, "learning_rate": 3.794167631717064e-06, "loss": 0.0062, "step": 31484 }, { "epoch": 5.77, "learning_rate": 3.79358547931145e-06, "loss": 0.0105, "step": 31485 }, { "epoch": 5.77, "learning_rate": 3.793003361115738e-06, "loss": 0.0085, "step": 31486 }, { "epoch": 5.77, "learning_rate": 3.7924212771331382e-06, "loss": 0.007, "step": 31487 }, { "epoch": 5.77, "learning_rate": 3.791839227366857e-06, "loss": 0.0058, "step": 31488 }, { "epoch": 5.77, "learning_rate": 3.7912572118200998e-06, "loss": 0.0079, "step": 31489 }, { "epoch": 5.77, "learning_rate": 3.7906752304960815e-06, "loss": 0.0093, "step": 31490 }, { "epoch": 5.77, "learning_rate": 3.790093283398011e-06, "loss": 0.0077, "step": 31491 }, { "epoch": 5.78, "learning_rate": 3.7895113705290875e-06, "loss": 0.0061, "step": 31492 }, { "epoch": 5.78, "learning_rate": 3.78892949189252e-06, "loss": 0.0103, "step": 31493 }, { "epoch": 5.78, "learning_rate": 3.7883476474915226e-06, "loss": 0.0074, "step": 31494 }, { "epoch": 5.78, "learning_rate": 3.787765837329297e-06, "loss": 0.0077, "step": 31495 }, { "epoch": 5.78, "learning_rate": 3.7871840614090526e-06, "loss": 0.005, "step": 31496 }, { "epoch": 5.78, "learning_rate": 3.7866023197339984e-06, "loss": 0.0093, "step": 31497 }, { "epoch": 5.78, "learning_rate": 3.7860206123073316e-06, "loss": 0.0058, "step": 31498 }, { "epoch": 5.78, "learning_rate": 3.7854389391322677e-06, "loss": 0.0096, "step": 31499 }, { "epoch": 5.78, "learning_rate": 3.7848573002120095e-06, "loss": 0.0075, "step": 31500 }, { "epoch": 5.78, "learning_rate": 3.7842756955497637e-06, "loss": 0.0064, "step": 31501 }, { "epoch": 5.78, "learning_rate": 3.783694125148737e-06, "loss": 0.007, "step": 31502 }, { "epoch": 5.78, "learning_rate": 3.783112589012132e-06, "loss": 0.008, "step": 31503 }, { "epoch": 5.78, "learning_rate": 3.7825310871431575e-06, "loss": 0.0081, "step": 31504 }, { "epoch": 5.78, "learning_rate": 3.781949619545017e-06, "loss": 0.0071, "step": 31505 }, { "epoch": 5.78, "learning_rate": 3.7813681862209116e-06, "loss": 0.007, "step": 31506 }, { "epoch": 5.78, "learning_rate": 3.7807867871740554e-06, "loss": 0.0049, "step": 31507 }, { "epoch": 5.78, "learning_rate": 3.7802054224076467e-06, "loss": 0.0065, "step": 31508 }, { "epoch": 5.78, "learning_rate": 3.779624091924896e-06, "loss": 0.0064, "step": 31509 }, { "epoch": 5.78, "learning_rate": 3.779042795728999e-06, "loss": 0.0066, "step": 31510 }, { "epoch": 5.78, "learning_rate": 3.77846153382316e-06, "loss": 0.0085, "step": 31511 }, { "epoch": 5.78, "learning_rate": 3.777880306210591e-06, "loss": 0.008, "step": 31512 }, { "epoch": 5.78, "learning_rate": 3.7772991128944913e-06, "loss": 0.0089, "step": 31513 }, { "epoch": 5.78, "learning_rate": 3.776717953878064e-06, "loss": 0.009, "step": 31514 }, { "epoch": 5.78, "learning_rate": 3.776136829164514e-06, "loss": 0.0078, "step": 31515 }, { "epoch": 5.78, "learning_rate": 3.775555738757043e-06, "loss": 0.0072, "step": 31516 }, { "epoch": 5.78, "learning_rate": 3.774974682658855e-06, "loss": 0.0085, "step": 31517 }, { "epoch": 5.78, "learning_rate": 3.7743936608731526e-06, "loss": 0.0089, "step": 31518 }, { "epoch": 5.78, "learning_rate": 3.7738126734031376e-06, "loss": 0.0098, "step": 31519 }, { "epoch": 5.78, "learning_rate": 3.773231720252014e-06, "loss": 0.0064, "step": 31520 }, { "epoch": 5.78, "learning_rate": 3.7726508014229825e-06, "loss": 0.0067, "step": 31521 }, { "epoch": 5.78, "learning_rate": 3.7720699169192455e-06, "loss": 0.0091, "step": 31522 }, { "epoch": 5.78, "learning_rate": 3.7714890667440043e-06, "loss": 0.0084, "step": 31523 }, { "epoch": 5.78, "learning_rate": 3.770908250900459e-06, "loss": 0.0082, "step": 31524 }, { "epoch": 5.78, "learning_rate": 3.7703274693918156e-06, "loss": 0.0086, "step": 31525 }, { "epoch": 5.78, "learning_rate": 3.7697467222212746e-06, "loss": 0.0099, "step": 31526 }, { "epoch": 5.78, "learning_rate": 3.769166009392038e-06, "loss": 0.0081, "step": 31527 }, { "epoch": 5.78, "learning_rate": 3.7685853309073016e-06, "loss": 0.0076, "step": 31528 }, { "epoch": 5.78, "learning_rate": 3.7680046867702647e-06, "loss": 0.0062, "step": 31529 }, { "epoch": 5.78, "learning_rate": 3.7674240769841365e-06, "loss": 0.0054, "step": 31530 }, { "epoch": 5.78, "learning_rate": 3.766843501552111e-06, "loss": 0.0086, "step": 31531 }, { "epoch": 5.78, "learning_rate": 3.7662629604773913e-06, "loss": 0.0076, "step": 31532 }, { "epoch": 5.78, "learning_rate": 3.7656824537631762e-06, "loss": 0.0052, "step": 31533 }, { "epoch": 5.78, "learning_rate": 3.7651019814126656e-06, "loss": 0.01, "step": 31534 }, { "epoch": 5.78, "learning_rate": 3.7645215434290582e-06, "loss": 0.0058, "step": 31535 }, { "epoch": 5.78, "learning_rate": 3.763941139815556e-06, "loss": 0.0081, "step": 31536 }, { "epoch": 5.78, "learning_rate": 3.7633607705753506e-06, "loss": 0.006, "step": 31537 }, { "epoch": 5.78, "learning_rate": 3.7627804357116547e-06, "loss": 0.0084, "step": 31538 }, { "epoch": 5.78, "learning_rate": 3.762200135227656e-06, "loss": 0.0072, "step": 31539 }, { "epoch": 5.78, "learning_rate": 3.761619869126556e-06, "loss": 0.0076, "step": 31540 }, { "epoch": 5.78, "learning_rate": 3.761039637411553e-06, "loss": 0.0118, "step": 31541 }, { "epoch": 5.78, "learning_rate": 3.760459440085843e-06, "loss": 0.0088, "step": 31542 }, { "epoch": 5.78, "learning_rate": 3.7598792771526293e-06, "loss": 0.007, "step": 31543 }, { "epoch": 5.78, "learning_rate": 3.759299148615111e-06, "loss": 0.0088, "step": 31544 }, { "epoch": 5.78, "learning_rate": 3.7587190544764785e-06, "loss": 0.0077, "step": 31545 }, { "epoch": 5.79, "learning_rate": 3.7581389947399293e-06, "loss": 0.0053, "step": 31546 }, { "epoch": 5.79, "learning_rate": 3.757558969408668e-06, "loss": 0.0093, "step": 31547 }, { "epoch": 5.79, "learning_rate": 3.7569789784858877e-06, "loss": 0.0066, "step": 31548 }, { "epoch": 5.79, "learning_rate": 3.7563990219747857e-06, "loss": 0.0055, "step": 31549 }, { "epoch": 5.79, "learning_rate": 3.7558190998785583e-06, "loss": 0.0092, "step": 31550 }, { "epoch": 5.79, "learning_rate": 3.755239212200402e-06, "loss": 0.0102, "step": 31551 }, { "epoch": 5.79, "learning_rate": 3.7546593589435142e-06, "loss": 0.0066, "step": 31552 }, { "epoch": 5.79, "learning_rate": 3.7540795401110906e-06, "loss": 0.0078, "step": 31553 }, { "epoch": 5.79, "learning_rate": 3.7534997557063257e-06, "loss": 0.0054, "step": 31554 }, { "epoch": 5.79, "learning_rate": 3.7529200057324143e-06, "loss": 0.0068, "step": 31555 }, { "epoch": 5.79, "learning_rate": 3.7523402901925597e-06, "loss": 0.0069, "step": 31556 }, { "epoch": 5.79, "learning_rate": 3.7517606090899494e-06, "loss": 0.0086, "step": 31557 }, { "epoch": 5.79, "learning_rate": 3.751180962427781e-06, "loss": 0.0071, "step": 31558 }, { "epoch": 5.79, "learning_rate": 3.7506013502092453e-06, "loss": 0.0073, "step": 31559 }, { "epoch": 5.79, "learning_rate": 3.7500217724375453e-06, "loss": 0.0067, "step": 31560 }, { "epoch": 5.79, "learning_rate": 3.749442229115872e-06, "loss": 0.0054, "step": 31561 }, { "epoch": 5.79, "learning_rate": 3.7488627202474227e-06, "loss": 0.0057, "step": 31562 }, { "epoch": 5.79, "learning_rate": 3.7482832458353857e-06, "loss": 0.0103, "step": 31563 }, { "epoch": 5.79, "learning_rate": 3.747703805882953e-06, "loss": 0.0091, "step": 31564 }, { "epoch": 5.79, "learning_rate": 3.7471244003933283e-06, "loss": 0.0056, "step": 31565 }, { "epoch": 5.79, "learning_rate": 3.746545029369699e-06, "loss": 0.0077, "step": 31566 }, { "epoch": 5.79, "learning_rate": 3.745965692815261e-06, "loss": 0.0069, "step": 31567 }, { "epoch": 5.79, "learning_rate": 3.7453863907332067e-06, "loss": 0.0072, "step": 31568 }, { "epoch": 5.79, "learning_rate": 3.744807123126729e-06, "loss": 0.0069, "step": 31569 }, { "epoch": 5.79, "learning_rate": 3.7442278899990215e-06, "loss": 0.0062, "step": 31570 }, { "epoch": 5.79, "learning_rate": 3.7436486913532756e-06, "loss": 0.0065, "step": 31571 }, { "epoch": 5.79, "learning_rate": 3.7430695271926842e-06, "loss": 0.0064, "step": 31572 }, { "epoch": 5.79, "learning_rate": 3.7424903975204407e-06, "loss": 0.0081, "step": 31573 }, { "epoch": 5.79, "learning_rate": 3.7419113023397367e-06, "loss": 0.0085, "step": 31574 }, { "epoch": 5.79, "learning_rate": 3.7413322416537646e-06, "loss": 0.0062, "step": 31575 }, { "epoch": 5.79, "learning_rate": 3.7407532154657157e-06, "loss": 0.0077, "step": 31576 }, { "epoch": 5.79, "learning_rate": 3.7401742237787786e-06, "loss": 0.007, "step": 31577 }, { "epoch": 5.79, "learning_rate": 3.73959526659615e-06, "loss": 0.0088, "step": 31578 }, { "epoch": 5.79, "learning_rate": 3.7390163439210202e-06, "loss": 0.0069, "step": 31579 }, { "epoch": 5.79, "learning_rate": 3.7384374557565818e-06, "loss": 0.0059, "step": 31580 }, { "epoch": 5.79, "learning_rate": 3.737858602106016e-06, "loss": 0.0074, "step": 31581 }, { "epoch": 5.79, "learning_rate": 3.7372797829725238e-06, "loss": 0.0068, "step": 31582 }, { "epoch": 5.79, "learning_rate": 3.736700998359293e-06, "loss": 0.0099, "step": 31583 }, { "epoch": 5.79, "learning_rate": 3.7361222482695114e-06, "loss": 0.0069, "step": 31584 }, { "epoch": 5.79, "learning_rate": 3.7355435327063717e-06, "loss": 0.0073, "step": 31585 }, { "epoch": 5.79, "learning_rate": 3.734964851673063e-06, "loss": 0.0077, "step": 31586 }, { "epoch": 5.79, "learning_rate": 3.734386205172774e-06, "loss": 0.0067, "step": 31587 }, { "epoch": 5.79, "learning_rate": 3.733807593208696e-06, "loss": 0.0077, "step": 31588 }, { "epoch": 5.79, "learning_rate": 3.7332290157840167e-06, "loss": 0.0076, "step": 31589 }, { "epoch": 5.79, "learning_rate": 3.732650472901923e-06, "loss": 0.0092, "step": 31590 }, { "epoch": 5.79, "learning_rate": 3.732071964565613e-06, "loss": 0.0087, "step": 31591 }, { "epoch": 5.79, "learning_rate": 3.7314934907782662e-06, "loss": 0.0065, "step": 31592 }, { "epoch": 5.79, "learning_rate": 3.7309150515430737e-06, "loss": 0.0088, "step": 31593 }, { "epoch": 5.79, "learning_rate": 3.7303366468632242e-06, "loss": 0.0071, "step": 31594 }, { "epoch": 5.79, "learning_rate": 3.7297582767419027e-06, "loss": 0.009, "step": 31595 }, { "epoch": 5.79, "learning_rate": 3.729179941182305e-06, "loss": 0.0074, "step": 31596 }, { "epoch": 5.79, "learning_rate": 3.7286016401876133e-06, "loss": 0.0081, "step": 31597 }, { "epoch": 5.79, "learning_rate": 3.7280233737610194e-06, "loss": 0.0085, "step": 31598 }, { "epoch": 5.79, "learning_rate": 3.7274451419057012e-06, "loss": 0.0064, "step": 31599 }, { "epoch": 5.79, "learning_rate": 3.7268669446248564e-06, "loss": 0.0076, "step": 31600 }, { "epoch": 5.8, "learning_rate": 3.7262887819216664e-06, "loss": 0.0085, "step": 31601 }, { "epoch": 5.8, "learning_rate": 3.7257106537993206e-06, "loss": 0.0103, "step": 31602 }, { "epoch": 5.8, "learning_rate": 3.7251325602610044e-06, "loss": 0.0076, "step": 31603 }, { "epoch": 5.8, "learning_rate": 3.724554501309905e-06, "loss": 0.0078, "step": 31604 }, { "epoch": 5.8, "learning_rate": 3.723976476949208e-06, "loss": 0.0059, "step": 31605 }, { "epoch": 5.8, "learning_rate": 3.7233984871820993e-06, "loss": 0.0079, "step": 31606 }, { "epoch": 5.8, "learning_rate": 3.7228205320117662e-06, "loss": 0.0073, "step": 31607 }, { "epoch": 5.8, "learning_rate": 3.722242611441389e-06, "loss": 0.0065, "step": 31608 }, { "epoch": 5.8, "learning_rate": 3.7216647254741645e-06, "loss": 0.0068, "step": 31609 }, { "epoch": 5.8, "learning_rate": 3.7210868741132676e-06, "loss": 0.0064, "step": 31610 }, { "epoch": 5.8, "learning_rate": 3.7205090573618875e-06, "loss": 0.0064, "step": 31611 }, { "epoch": 5.8, "learning_rate": 3.7199312752232053e-06, "loss": 0.0069, "step": 31612 }, { "epoch": 5.8, "learning_rate": 3.7193535277004124e-06, "loss": 0.0089, "step": 31613 }, { "epoch": 5.8, "learning_rate": 3.718775814796689e-06, "loss": 0.0062, "step": 31614 }, { "epoch": 5.8, "learning_rate": 3.7181981365152208e-06, "loss": 0.0065, "step": 31615 }, { "epoch": 5.8, "learning_rate": 3.7176204928591963e-06, "loss": 0.0093, "step": 31616 }, { "epoch": 5.8, "learning_rate": 3.7170428838317863e-06, "loss": 0.0066, "step": 31617 }, { "epoch": 5.8, "learning_rate": 3.716465309436188e-06, "loss": 0.0083, "step": 31618 }, { "epoch": 5.8, "learning_rate": 3.715887769675579e-06, "loss": 0.0078, "step": 31619 }, { "epoch": 5.8, "learning_rate": 3.715310264553145e-06, "loss": 0.0065, "step": 31620 }, { "epoch": 5.8, "learning_rate": 3.7147327940720668e-06, "loss": 0.008, "step": 31621 }, { "epoch": 5.8, "learning_rate": 3.7141553582355294e-06, "loss": 0.0094, "step": 31622 }, { "epoch": 5.8, "learning_rate": 3.7135779570467146e-06, "loss": 0.006, "step": 31623 }, { "epoch": 5.8, "learning_rate": 3.713000590508805e-06, "loss": 0.006, "step": 31624 }, { "epoch": 5.8, "learning_rate": 3.7124232586249842e-06, "loss": 0.0083, "step": 31625 }, { "epoch": 5.8, "learning_rate": 3.7118459613984303e-06, "loss": 0.006, "step": 31626 }, { "epoch": 5.8, "learning_rate": 3.7112686988323353e-06, "loss": 0.0051, "step": 31627 }, { "epoch": 5.8, "learning_rate": 3.710691470929871e-06, "loss": 0.0066, "step": 31628 }, { "epoch": 5.8, "learning_rate": 3.710114277694222e-06, "loss": 0.0046, "step": 31629 }, { "epoch": 5.8, "learning_rate": 3.709537119128569e-06, "loss": 0.0071, "step": 31630 }, { "epoch": 5.8, "learning_rate": 3.7089599952360967e-06, "loss": 0.0057, "step": 31631 }, { "epoch": 5.8, "learning_rate": 3.708382906019985e-06, "loss": 0.0086, "step": 31632 }, { "epoch": 5.8, "learning_rate": 3.7078058514834136e-06, "loss": 0.0067, "step": 31633 }, { "epoch": 5.8, "learning_rate": 3.7072288316295635e-06, "loss": 0.0064, "step": 31634 }, { "epoch": 5.8, "learning_rate": 3.7066518464616165e-06, "loss": 0.0057, "step": 31635 }, { "epoch": 5.8, "learning_rate": 3.706074895982751e-06, "loss": 0.0068, "step": 31636 }, { "epoch": 5.8, "learning_rate": 3.705497980196149e-06, "loss": 0.0088, "step": 31637 }, { "epoch": 5.8, "learning_rate": 3.7049210991049888e-06, "loss": 0.0073, "step": 31638 }, { "epoch": 5.8, "learning_rate": 3.704344252712452e-06, "loss": 0.0066, "step": 31639 }, { "epoch": 5.8, "learning_rate": 3.703767441021716e-06, "loss": 0.0069, "step": 31640 }, { "epoch": 5.8, "learning_rate": 3.7031906640359616e-06, "loss": 0.007, "step": 31641 }, { "epoch": 5.8, "learning_rate": 3.702613921758368e-06, "loss": 0.0083, "step": 31642 }, { "epoch": 5.8, "learning_rate": 3.70203721419211e-06, "loss": 0.0076, "step": 31643 }, { "epoch": 5.8, "learning_rate": 3.7014605413403746e-06, "loss": 0.0074, "step": 31644 }, { "epoch": 5.8, "learning_rate": 3.7008839032063404e-06, "loss": 0.0053, "step": 31645 }, { "epoch": 5.8, "learning_rate": 3.7003072997931768e-06, "loss": 0.0059, "step": 31646 }, { "epoch": 5.8, "learning_rate": 3.6997307311040674e-06, "loss": 0.0076, "step": 31647 }, { "epoch": 5.8, "learning_rate": 3.699154197142186e-06, "loss": 0.0076, "step": 31648 }, { "epoch": 5.8, "learning_rate": 3.698577697910719e-06, "loss": 0.0071, "step": 31649 }, { "epoch": 5.8, "learning_rate": 3.698001233412839e-06, "loss": 0.0091, "step": 31650 }, { "epoch": 5.8, "learning_rate": 3.697424803651727e-06, "loss": 0.0063, "step": 31651 }, { "epoch": 5.8, "learning_rate": 3.69684840863055e-06, "loss": 0.007, "step": 31652 }, { "epoch": 5.8, "learning_rate": 3.6962720483524948e-06, "loss": 0.0064, "step": 31653 }, { "epoch": 5.8, "learning_rate": 3.695695722820737e-06, "loss": 0.007, "step": 31654 }, { "epoch": 5.81, "learning_rate": 3.695119432038452e-06, "loss": 0.0054, "step": 31655 }, { "epoch": 5.81, "learning_rate": 3.6945431760088166e-06, "loss": 0.0078, "step": 31656 }, { "epoch": 5.81, "learning_rate": 3.6939669547350064e-06, "loss": 0.0063, "step": 31657 }, { "epoch": 5.81, "learning_rate": 3.6933907682201985e-06, "loss": 0.0083, "step": 31658 }, { "epoch": 5.81, "learning_rate": 3.6928146164675682e-06, "loss": 0.0064, "step": 31659 }, { "epoch": 5.81, "learning_rate": 3.6922384994802905e-06, "loss": 0.0058, "step": 31660 }, { "epoch": 5.81, "learning_rate": 3.69166241726154e-06, "loss": 0.005, "step": 31661 }, { "epoch": 5.81, "learning_rate": 3.6910863698144973e-06, "loss": 0.0062, "step": 31662 }, { "epoch": 5.81, "learning_rate": 3.6905103571423374e-06, "loss": 0.0067, "step": 31663 }, { "epoch": 5.81, "learning_rate": 3.689934379248229e-06, "loss": 0.0057, "step": 31664 }, { "epoch": 5.81, "learning_rate": 3.6893584361353463e-06, "loss": 0.0082, "step": 31665 }, { "epoch": 5.81, "learning_rate": 3.6887825278068713e-06, "loss": 0.0071, "step": 31666 }, { "epoch": 5.81, "learning_rate": 3.6882066542659746e-06, "loss": 0.0072, "step": 31667 }, { "epoch": 5.81, "learning_rate": 3.6876308155158314e-06, "loss": 0.0051, "step": 31668 }, { "epoch": 5.81, "learning_rate": 3.6870550115596184e-06, "loss": 0.0079, "step": 31669 }, { "epoch": 5.81, "learning_rate": 3.6864792424004993e-06, "loss": 0.0096, "step": 31670 }, { "epoch": 5.81, "learning_rate": 3.6859035080416572e-06, "loss": 0.0052, "step": 31671 }, { "epoch": 5.81, "learning_rate": 3.685327808486263e-06, "loss": 0.0047, "step": 31672 }, { "epoch": 5.81, "learning_rate": 3.68475214373749e-06, "loss": 0.0068, "step": 31673 }, { "epoch": 5.81, "learning_rate": 3.684176513798512e-06, "loss": 0.0105, "step": 31674 }, { "epoch": 5.81, "learning_rate": 3.6836009186724996e-06, "loss": 0.0091, "step": 31675 }, { "epoch": 5.81, "learning_rate": 3.6830253583626275e-06, "loss": 0.0061, "step": 31676 }, { "epoch": 5.81, "learning_rate": 3.6824498328720683e-06, "loss": 0.0088, "step": 31677 }, { "epoch": 5.81, "learning_rate": 3.681874342203993e-06, "loss": 0.0061, "step": 31678 }, { "epoch": 5.81, "learning_rate": 3.6812988863615706e-06, "loss": 0.0075, "step": 31679 }, { "epoch": 5.81, "learning_rate": 3.6807234653479807e-06, "loss": 0.0056, "step": 31680 }, { "epoch": 5.81, "learning_rate": 3.6801480791663947e-06, "loss": 0.0051, "step": 31681 }, { "epoch": 5.81, "learning_rate": 3.6795727278199768e-06, "loss": 0.0066, "step": 31682 }, { "epoch": 5.81, "learning_rate": 3.6789974113118986e-06, "loss": 0.0071, "step": 31683 }, { "epoch": 5.81, "learning_rate": 3.6784221296453382e-06, "loss": 0.009, "step": 31684 }, { "epoch": 5.81, "learning_rate": 3.6778468828234628e-06, "loss": 0.0082, "step": 31685 }, { "epoch": 5.81, "learning_rate": 3.677271670849444e-06, "loss": 0.0058, "step": 31686 }, { "epoch": 5.81, "learning_rate": 3.6766964937264516e-06, "loss": 0.0069, "step": 31687 }, { "epoch": 5.81, "learning_rate": 3.6761213514576564e-06, "loss": 0.0067, "step": 31688 }, { "epoch": 5.81, "learning_rate": 3.6755462440462288e-06, "loss": 0.0072, "step": 31689 }, { "epoch": 5.81, "learning_rate": 3.6749711714953373e-06, "loss": 0.0076, "step": 31690 }, { "epoch": 5.81, "learning_rate": 3.6743961338081547e-06, "loss": 0.007, "step": 31691 }, { "epoch": 5.81, "learning_rate": 3.6738211309878467e-06, "loss": 0.0077, "step": 31692 }, { "epoch": 5.81, "learning_rate": 3.673246163037587e-06, "loss": 0.0064, "step": 31693 }, { "epoch": 5.81, "learning_rate": 3.6726712299605414e-06, "loss": 0.006, "step": 31694 }, { "epoch": 5.81, "learning_rate": 3.67209633175988e-06, "loss": 0.0067, "step": 31695 }, { "epoch": 5.81, "learning_rate": 3.671521468438769e-06, "loss": 0.0078, "step": 31696 }, { "epoch": 5.81, "learning_rate": 3.6709466400003834e-06, "loss": 0.006, "step": 31697 }, { "epoch": 5.81, "learning_rate": 3.670371846447891e-06, "loss": 0.0079, "step": 31698 }, { "epoch": 5.81, "learning_rate": 3.6697970877844547e-06, "loss": 0.0074, "step": 31699 }, { "epoch": 5.81, "learning_rate": 3.6692223640132452e-06, "loss": 0.0053, "step": 31700 }, { "epoch": 5.81, "learning_rate": 3.6686476751374266e-06, "loss": 0.007, "step": 31701 }, { "epoch": 5.81, "learning_rate": 3.6680730211601735e-06, "loss": 0.0065, "step": 31702 }, { "epoch": 5.81, "learning_rate": 3.6674984020846503e-06, "loss": 0.0061, "step": 31703 }, { "epoch": 5.81, "learning_rate": 3.6669238179140242e-06, "loss": 0.0069, "step": 31704 }, { "epoch": 5.81, "learning_rate": 3.6663492686514623e-06, "loss": 0.0067, "step": 31705 }, { "epoch": 5.81, "learning_rate": 3.665774754300132e-06, "loss": 0.0048, "step": 31706 }, { "epoch": 5.81, "learning_rate": 3.6652002748632e-06, "loss": 0.008, "step": 31707 }, { "epoch": 5.81, "learning_rate": 3.664625830343832e-06, "loss": 0.0067, "step": 31708 }, { "epoch": 5.81, "learning_rate": 3.6640514207451905e-06, "loss": 0.009, "step": 31709 }, { "epoch": 5.82, "learning_rate": 3.6634770460704538e-06, "loss": 0.0064, "step": 31710 }, { "epoch": 5.82, "learning_rate": 3.662902706322777e-06, "loss": 0.006, "step": 31711 }, { "epoch": 5.82, "learning_rate": 3.6623284015053285e-06, "loss": 0.0071, "step": 31712 }, { "epoch": 5.82, "learning_rate": 3.6617541316212745e-06, "loss": 0.0064, "step": 31713 }, { "epoch": 5.82, "learning_rate": 3.6611798966737767e-06, "loss": 0.0073, "step": 31714 }, { "epoch": 5.82, "learning_rate": 3.6606056966660077e-06, "loss": 0.0067, "step": 31715 }, { "epoch": 5.82, "learning_rate": 3.6600315316011315e-06, "loss": 0.0075, "step": 31716 }, { "epoch": 5.82, "learning_rate": 3.659457401482306e-06, "loss": 0.0081, "step": 31717 }, { "epoch": 5.82, "learning_rate": 3.658883306312697e-06, "loss": 0.0061, "step": 31718 }, { "epoch": 5.82, "learning_rate": 3.658309246095475e-06, "loss": 0.0066, "step": 31719 }, { "epoch": 5.82, "learning_rate": 3.6577352208338015e-06, "loss": 0.0056, "step": 31720 }, { "epoch": 5.82, "learning_rate": 3.6571612305308403e-06, "loss": 0.0062, "step": 31721 }, { "epoch": 5.82, "learning_rate": 3.656587275189755e-06, "loss": 0.0053, "step": 31722 }, { "epoch": 5.82, "learning_rate": 3.656013354813708e-06, "loss": 0.0063, "step": 31723 }, { "epoch": 5.82, "learning_rate": 3.655439469405865e-06, "loss": 0.0091, "step": 31724 }, { "epoch": 5.82, "learning_rate": 3.6548656189693876e-06, "loss": 0.0069, "step": 31725 }, { "epoch": 5.82, "learning_rate": 3.6542918035074404e-06, "loss": 0.005, "step": 31726 }, { "epoch": 5.82, "learning_rate": 3.6537180230231817e-06, "loss": 0.0089, "step": 31727 }, { "epoch": 5.82, "learning_rate": 3.6531442775197845e-06, "loss": 0.0072, "step": 31728 }, { "epoch": 5.82, "learning_rate": 3.652570567000402e-06, "loss": 0.0059, "step": 31729 }, { "epoch": 5.82, "learning_rate": 3.651996891468198e-06, "loss": 0.0054, "step": 31730 }, { "epoch": 5.82, "learning_rate": 3.651423250926337e-06, "loss": 0.0078, "step": 31731 }, { "epoch": 5.82, "learning_rate": 3.6508496453779764e-06, "loss": 0.0082, "step": 31732 }, { "epoch": 5.82, "learning_rate": 3.6502760748262843e-06, "loss": 0.0069, "step": 31733 }, { "epoch": 5.82, "learning_rate": 3.649702539274422e-06, "loss": 0.0096, "step": 31734 }, { "epoch": 5.82, "learning_rate": 3.649129038725545e-06, "loss": 0.0065, "step": 31735 }, { "epoch": 5.82, "learning_rate": 3.648555573182815e-06, "loss": 0.0062, "step": 31736 }, { "epoch": 5.82, "learning_rate": 3.647982142649399e-06, "loss": 0.0054, "step": 31737 }, { "epoch": 5.82, "learning_rate": 3.6474087471284526e-06, "loss": 0.0074, "step": 31738 }, { "epoch": 5.82, "learning_rate": 3.6468353866231386e-06, "loss": 0.0066, "step": 31739 }, { "epoch": 5.82, "learning_rate": 3.646262061136617e-06, "loss": 0.0064, "step": 31740 }, { "epoch": 5.82, "learning_rate": 3.645688770672049e-06, "loss": 0.0074, "step": 31741 }, { "epoch": 5.82, "learning_rate": 3.6451155152325924e-06, "loss": 0.008, "step": 31742 }, { "epoch": 5.82, "learning_rate": 3.644542294821407e-06, "loss": 0.0081, "step": 31743 }, { "epoch": 5.82, "learning_rate": 3.643969109441655e-06, "loss": 0.0059, "step": 31744 }, { "epoch": 5.82, "learning_rate": 3.6433959590964927e-06, "loss": 0.0069, "step": 31745 }, { "epoch": 5.82, "learning_rate": 3.6428228437890824e-06, "loss": 0.0068, "step": 31746 }, { "epoch": 5.82, "learning_rate": 3.6422497635225793e-06, "loss": 0.0059, "step": 31747 }, { "epoch": 5.82, "learning_rate": 3.641676718300147e-06, "loss": 0.0065, "step": 31748 }, { "epoch": 5.82, "learning_rate": 3.6411037081249366e-06, "loss": 0.0089, "step": 31749 }, { "epoch": 5.82, "learning_rate": 3.640530733000115e-06, "loss": 0.0073, "step": 31750 }, { "epoch": 5.82, "learning_rate": 3.6399577929288378e-06, "loss": 0.0083, "step": 31751 }, { "epoch": 5.82, "learning_rate": 3.6393848879142647e-06, "loss": 0.0107, "step": 31752 }, { "epoch": 5.82, "learning_rate": 3.638812017959548e-06, "loss": 0.0056, "step": 31753 }, { "epoch": 5.82, "learning_rate": 3.6382391830678453e-06, "loss": 0.0072, "step": 31754 }, { "epoch": 5.82, "learning_rate": 3.637666383242321e-06, "loss": 0.01, "step": 31755 }, { "epoch": 5.82, "learning_rate": 3.6370936184861273e-06, "loss": 0.0067, "step": 31756 }, { "epoch": 5.82, "learning_rate": 3.6365208888024237e-06, "loss": 0.0083, "step": 31757 }, { "epoch": 5.82, "learning_rate": 3.6359481941943664e-06, "loss": 0.0064, "step": 31758 }, { "epoch": 5.82, "learning_rate": 3.635375534665111e-06, "loss": 0.0066, "step": 31759 }, { "epoch": 5.82, "learning_rate": 3.634802910217815e-06, "loss": 0.0092, "step": 31760 }, { "epoch": 5.82, "learning_rate": 3.634230320855634e-06, "loss": 0.0052, "step": 31761 }, { "epoch": 5.82, "learning_rate": 3.633657766581722e-06, "loss": 0.0073, "step": 31762 }, { "epoch": 5.82, "learning_rate": 3.633085247399244e-06, "loss": 0.0079, "step": 31763 }, { "epoch": 5.83, "learning_rate": 3.6325127633113454e-06, "loss": 0.0091, "step": 31764 }, { "epoch": 5.83, "learning_rate": 3.6319403143211862e-06, "loss": 0.0054, "step": 31765 }, { "epoch": 5.83, "learning_rate": 3.6313679004319213e-06, "loss": 0.0087, "step": 31766 }, { "epoch": 5.83, "learning_rate": 3.6307955216467027e-06, "loss": 0.0059, "step": 31767 }, { "epoch": 5.83, "learning_rate": 3.6302231779686915e-06, "loss": 0.0073, "step": 31768 }, { "epoch": 5.83, "learning_rate": 3.629650869401039e-06, "loss": 0.0076, "step": 31769 }, { "epoch": 5.83, "learning_rate": 3.6290785959469033e-06, "loss": 0.0042, "step": 31770 }, { "epoch": 5.83, "learning_rate": 3.6285063576094294e-06, "loss": 0.0086, "step": 31771 }, { "epoch": 5.83, "learning_rate": 3.62793415439178e-06, "loss": 0.0069, "step": 31772 }, { "epoch": 5.83, "learning_rate": 3.627361986297109e-06, "loss": 0.0057, "step": 31773 }, { "epoch": 5.83, "learning_rate": 3.6267898533285663e-06, "loss": 0.0081, "step": 31774 }, { "epoch": 5.83, "learning_rate": 3.6262177554893075e-06, "loss": 0.0082, "step": 31775 }, { "epoch": 5.83, "learning_rate": 3.625645692782486e-06, "loss": 0.0063, "step": 31776 }, { "epoch": 5.83, "learning_rate": 3.625073665211255e-06, "loss": 0.0061, "step": 31777 }, { "epoch": 5.83, "learning_rate": 3.6245016727787662e-06, "loss": 0.0082, "step": 31778 }, { "epoch": 5.83, "learning_rate": 3.6239297154881748e-06, "loss": 0.0071, "step": 31779 }, { "epoch": 5.83, "learning_rate": 3.623357793342628e-06, "loss": 0.0061, "step": 31780 }, { "epoch": 5.83, "learning_rate": 3.6227859063452898e-06, "loss": 0.0069, "step": 31781 }, { "epoch": 5.83, "learning_rate": 3.622214054499301e-06, "loss": 0.0072, "step": 31782 }, { "epoch": 5.83, "learning_rate": 3.6216422378078187e-06, "loss": 0.007, "step": 31783 }, { "epoch": 5.83, "learning_rate": 3.621070456273994e-06, "loss": 0.0066, "step": 31784 }, { "epoch": 5.83, "learning_rate": 3.6204987099009735e-06, "loss": 0.007, "step": 31785 }, { "epoch": 5.83, "learning_rate": 3.6199269986919185e-06, "loss": 0.0068, "step": 31786 }, { "epoch": 5.83, "learning_rate": 3.6193553226499745e-06, "loss": 0.0061, "step": 31787 }, { "epoch": 5.83, "learning_rate": 3.618783681778296e-06, "loss": 0.0052, "step": 31788 }, { "epoch": 5.83, "learning_rate": 3.6182120760800265e-06, "loss": 0.0087, "step": 31789 }, { "epoch": 5.83, "learning_rate": 3.617640505558324e-06, "loss": 0.008, "step": 31790 }, { "epoch": 5.83, "learning_rate": 3.6170689702163364e-06, "loss": 0.0085, "step": 31791 }, { "epoch": 5.83, "learning_rate": 3.616497470057214e-06, "loss": 0.0059, "step": 31792 }, { "epoch": 5.83, "learning_rate": 3.615926005084107e-06, "loss": 0.0082, "step": 31793 }, { "epoch": 5.83, "learning_rate": 3.6153545753001663e-06, "loss": 0.0055, "step": 31794 }, { "epoch": 5.83, "learning_rate": 3.6147831807085397e-06, "loss": 0.0062, "step": 31795 }, { "epoch": 5.83, "learning_rate": 3.614211821312379e-06, "loss": 0.0073, "step": 31796 }, { "epoch": 5.83, "learning_rate": 3.613640497114832e-06, "loss": 0.009, "step": 31797 }, { "epoch": 5.83, "learning_rate": 3.6130692081190444e-06, "loss": 0.0058, "step": 31798 }, { "epoch": 5.83, "learning_rate": 3.612497954328176e-06, "loss": 0.0095, "step": 31799 }, { "epoch": 5.83, "learning_rate": 3.6119267357453646e-06, "loss": 0.009, "step": 31800 }, { "epoch": 5.83, "learning_rate": 3.611355552373763e-06, "loss": 0.0064, "step": 31801 }, { "epoch": 5.83, "learning_rate": 3.6107844042165164e-06, "loss": 0.0075, "step": 31802 }, { "epoch": 5.83, "learning_rate": 3.6102132912767784e-06, "loss": 0.0052, "step": 31803 }, { "epoch": 5.83, "learning_rate": 3.609642213557695e-06, "loss": 0.0081, "step": 31804 }, { "epoch": 5.83, "learning_rate": 3.609071171062417e-06, "loss": 0.009, "step": 31805 }, { "epoch": 5.83, "learning_rate": 3.608500163794084e-06, "loss": 0.0074, "step": 31806 }, { "epoch": 5.83, "learning_rate": 3.6079291917558455e-06, "loss": 0.0056, "step": 31807 }, { "epoch": 5.83, "learning_rate": 3.6073582549508533e-06, "loss": 0.0068, "step": 31808 }, { "epoch": 5.83, "learning_rate": 3.606787353382254e-06, "loss": 0.0082, "step": 31809 }, { "epoch": 5.83, "learning_rate": 3.606216487053191e-06, "loss": 0.0067, "step": 31810 }, { "epoch": 5.83, "learning_rate": 3.605645655966813e-06, "loss": 0.0064, "step": 31811 }, { "epoch": 5.83, "learning_rate": 3.6050748601262662e-06, "loss": 0.0062, "step": 31812 }, { "epoch": 5.83, "learning_rate": 3.604504099534696e-06, "loss": 0.005, "step": 31813 }, { "epoch": 5.83, "learning_rate": 3.6039333741952496e-06, "loss": 0.0077, "step": 31814 }, { "epoch": 5.83, "learning_rate": 3.603362684111069e-06, "loss": 0.0074, "step": 31815 }, { "epoch": 5.83, "learning_rate": 3.602792029285306e-06, "loss": 0.0072, "step": 31816 }, { "epoch": 5.83, "learning_rate": 3.602221409721107e-06, "loss": 0.0065, "step": 31817 }, { "epoch": 5.83, "learning_rate": 3.60165082542161e-06, "loss": 0.0078, "step": 31818 }, { "epoch": 5.84, "learning_rate": 3.601080276389963e-06, "loss": 0.0058, "step": 31819 }, { "epoch": 5.84, "learning_rate": 3.6005097626293083e-06, "loss": 0.0058, "step": 31820 }, { "epoch": 5.84, "learning_rate": 3.5999392841427973e-06, "loss": 0.0047, "step": 31821 }, { "epoch": 5.84, "learning_rate": 3.5993688409335716e-06, "loss": 0.0073, "step": 31822 }, { "epoch": 5.84, "learning_rate": 3.5987984330047776e-06, "loss": 0.0086, "step": 31823 }, { "epoch": 5.84, "learning_rate": 3.598228060359549e-06, "loss": 0.007, "step": 31824 }, { "epoch": 5.84, "learning_rate": 3.5976577230010414e-06, "loss": 0.0073, "step": 31825 }, { "epoch": 5.84, "learning_rate": 3.5970874209323947e-06, "loss": 0.0078, "step": 31826 }, { "epoch": 5.84, "learning_rate": 3.5965171541567523e-06, "loss": 0.0076, "step": 31827 }, { "epoch": 5.84, "learning_rate": 3.5959469226772557e-06, "loss": 0.0056, "step": 31828 }, { "epoch": 5.84, "learning_rate": 3.5953767264970516e-06, "loss": 0.0068, "step": 31829 }, { "epoch": 5.84, "learning_rate": 3.59480656561928e-06, "loss": 0.0075, "step": 31830 }, { "epoch": 5.84, "learning_rate": 3.594236440047085e-06, "loss": 0.0063, "step": 31831 }, { "epoch": 5.84, "learning_rate": 3.593666349783609e-06, "loss": 0.0057, "step": 31832 }, { "epoch": 5.84, "learning_rate": 3.5930962948319903e-06, "loss": 0.0063, "step": 31833 }, { "epoch": 5.84, "learning_rate": 3.592526275195379e-06, "loss": 0.0056, "step": 31834 }, { "epoch": 5.84, "learning_rate": 3.5919562908769156e-06, "loss": 0.0066, "step": 31835 }, { "epoch": 5.84, "learning_rate": 3.591386341879737e-06, "loss": 0.0073, "step": 31836 }, { "epoch": 5.84, "learning_rate": 3.5908164282069825e-06, "loss": 0.0076, "step": 31837 }, { "epoch": 5.84, "learning_rate": 3.5902465498618024e-06, "loss": 0.0055, "step": 31838 }, { "epoch": 5.84, "learning_rate": 3.5896767068473325e-06, "loss": 0.0045, "step": 31839 }, { "epoch": 5.84, "learning_rate": 3.5891068991667154e-06, "loss": 0.0071, "step": 31840 }, { "epoch": 5.84, "learning_rate": 3.588537126823094e-06, "loss": 0.0075, "step": 31841 }, { "epoch": 5.84, "learning_rate": 3.5879673898196e-06, "loss": 0.0057, "step": 31842 }, { "epoch": 5.84, "learning_rate": 3.587397688159383e-06, "loss": 0.0072, "step": 31843 }, { "epoch": 5.84, "learning_rate": 3.58682802184558e-06, "loss": 0.0059, "step": 31844 }, { "epoch": 5.84, "learning_rate": 3.586258390881332e-06, "loss": 0.0072, "step": 31845 }, { "epoch": 5.84, "learning_rate": 3.5856887952697774e-06, "loss": 0.0054, "step": 31846 }, { "epoch": 5.84, "learning_rate": 3.5851192350140563e-06, "loss": 0.0076, "step": 31847 }, { "epoch": 5.84, "learning_rate": 3.584549710117309e-06, "loss": 0.0049, "step": 31848 }, { "epoch": 5.84, "learning_rate": 3.5839802205826724e-06, "loss": 0.007, "step": 31849 }, { "epoch": 5.84, "learning_rate": 3.5834107664132887e-06, "loss": 0.0089, "step": 31850 }, { "epoch": 5.84, "learning_rate": 3.5828413476122913e-06, "loss": 0.0081, "step": 31851 }, { "epoch": 5.84, "learning_rate": 3.5822719641828295e-06, "loss": 0.0061, "step": 31852 }, { "epoch": 5.84, "learning_rate": 3.581702616128031e-06, "loss": 0.0088, "step": 31853 }, { "epoch": 5.84, "learning_rate": 3.581133303451039e-06, "loss": 0.0069, "step": 31854 }, { "epoch": 5.84, "learning_rate": 3.580564026154987e-06, "loss": 0.0073, "step": 31855 }, { "epoch": 5.84, "learning_rate": 3.5799947842430194e-06, "loss": 0.0073, "step": 31856 }, { "epoch": 5.84, "learning_rate": 3.579425577718272e-06, "loss": 0.0078, "step": 31857 }, { "epoch": 5.84, "learning_rate": 3.5788564065838803e-06, "loss": 0.0067, "step": 31858 }, { "epoch": 5.84, "learning_rate": 3.578287270842986e-06, "loss": 0.0081, "step": 31859 }, { "epoch": 5.84, "learning_rate": 3.577718170498716e-06, "loss": 0.0068, "step": 31860 }, { "epoch": 5.84, "learning_rate": 3.5771491055542172e-06, "loss": 0.0041, "step": 31861 }, { "epoch": 5.84, "learning_rate": 3.5765800760126237e-06, "loss": 0.007, "step": 31862 }, { "epoch": 5.84, "learning_rate": 3.57601108187707e-06, "loss": 0.0083, "step": 31863 }, { "epoch": 5.84, "learning_rate": 3.5754421231506953e-06, "loss": 0.0066, "step": 31864 }, { "epoch": 5.84, "learning_rate": 3.574873199836633e-06, "loss": 0.0052, "step": 31865 }, { "epoch": 5.84, "learning_rate": 3.5743043119380207e-06, "loss": 0.0056, "step": 31866 }, { "epoch": 5.84, "learning_rate": 3.573735459457993e-06, "loss": 0.0072, "step": 31867 }, { "epoch": 5.84, "learning_rate": 3.5731666423996825e-06, "loss": 0.0059, "step": 31868 }, { "epoch": 5.84, "learning_rate": 3.5725978607662325e-06, "loss": 0.0053, "step": 31869 }, { "epoch": 5.84, "learning_rate": 3.5720291145607767e-06, "loss": 0.0069, "step": 31870 }, { "epoch": 5.84, "learning_rate": 3.571460403786443e-06, "loss": 0.0093, "step": 31871 }, { "epoch": 5.84, "learning_rate": 3.5708917284463707e-06, "loss": 0.0077, "step": 31872 }, { "epoch": 5.85, "learning_rate": 3.57032308854369e-06, "loss": 0.0053, "step": 31873 }, { "epoch": 5.85, "learning_rate": 3.5697544840815423e-06, "loss": 0.0086, "step": 31874 }, { "epoch": 5.85, "learning_rate": 3.5691859150630594e-06, "loss": 0.0047, "step": 31875 }, { "epoch": 5.85, "learning_rate": 3.568617381491374e-06, "loss": 0.0073, "step": 31876 }, { "epoch": 5.85, "learning_rate": 3.5680488833696193e-06, "loss": 0.0074, "step": 31877 }, { "epoch": 5.85, "learning_rate": 3.5674804207009306e-06, "loss": 0.0068, "step": 31878 }, { "epoch": 5.85, "learning_rate": 3.566911993488441e-06, "loss": 0.0054, "step": 31879 }, { "epoch": 5.85, "learning_rate": 3.5663436017352828e-06, "loss": 0.0062, "step": 31880 }, { "epoch": 5.85, "learning_rate": 3.5657752454445895e-06, "loss": 0.0072, "step": 31881 }, { "epoch": 5.85, "learning_rate": 3.5652069246194944e-06, "loss": 0.0081, "step": 31882 }, { "epoch": 5.85, "learning_rate": 3.564638639263128e-06, "loss": 0.0072, "step": 31883 }, { "epoch": 5.85, "learning_rate": 3.5640703893786254e-06, "loss": 0.0071, "step": 31884 }, { "epoch": 5.85, "learning_rate": 3.563502174969117e-06, "loss": 0.0068, "step": 31885 }, { "epoch": 5.85, "learning_rate": 3.562933996037732e-06, "loss": 0.0085, "step": 31886 }, { "epoch": 5.85, "learning_rate": 3.5623658525876094e-06, "loss": 0.0098, "step": 31887 }, { "epoch": 5.85, "learning_rate": 3.56179774462188e-06, "loss": 0.0083, "step": 31888 }, { "epoch": 5.85, "learning_rate": 3.561229672143669e-06, "loss": 0.0074, "step": 31889 }, { "epoch": 5.85, "learning_rate": 3.5606616351561072e-06, "loss": 0.0056, "step": 31890 }, { "epoch": 5.85, "learning_rate": 3.5600936336623337e-06, "loss": 0.0072, "step": 31891 }, { "epoch": 5.85, "learning_rate": 3.5595256676654744e-06, "loss": 0.0057, "step": 31892 }, { "epoch": 5.85, "learning_rate": 3.55895773716866e-06, "loss": 0.006, "step": 31893 }, { "epoch": 5.85, "learning_rate": 3.558389842175022e-06, "loss": 0.0077, "step": 31894 }, { "epoch": 5.85, "learning_rate": 3.5578219826876903e-06, "loss": 0.0095, "step": 31895 }, { "epoch": 5.85, "learning_rate": 3.5572541587097943e-06, "loss": 0.0072, "step": 31896 }, { "epoch": 5.85, "learning_rate": 3.5566863702444644e-06, "loss": 0.0077, "step": 31897 }, { "epoch": 5.85, "learning_rate": 3.5561186172948304e-06, "loss": 0.0072, "step": 31898 }, { "epoch": 5.85, "learning_rate": 3.555550899864021e-06, "loss": 0.0066, "step": 31899 }, { "epoch": 5.85, "learning_rate": 3.5549832179551667e-06, "loss": 0.0054, "step": 31900 }, { "epoch": 5.85, "learning_rate": 3.5544155715713958e-06, "loss": 0.0051, "step": 31901 }, { "epoch": 5.85, "learning_rate": 3.553847960715837e-06, "loss": 0.0086, "step": 31902 }, { "epoch": 5.85, "learning_rate": 3.55328038539162e-06, "loss": 0.0059, "step": 31903 }, { "epoch": 5.85, "learning_rate": 3.552712845601868e-06, "loss": 0.005, "step": 31904 }, { "epoch": 5.85, "learning_rate": 3.5521453413497178e-06, "loss": 0.005, "step": 31905 }, { "epoch": 5.85, "learning_rate": 3.5515778726382967e-06, "loss": 0.0067, "step": 31906 }, { "epoch": 5.85, "learning_rate": 3.5510104394707257e-06, "loss": 0.0061, "step": 31907 }, { "epoch": 5.85, "learning_rate": 3.5504430418501336e-06, "loss": 0.0074, "step": 31908 }, { "epoch": 5.85, "learning_rate": 3.549875679779654e-06, "loss": 0.0081, "step": 31909 }, { "epoch": 5.85, "learning_rate": 3.5493083532624106e-06, "loss": 0.0091, "step": 31910 }, { "epoch": 5.85, "learning_rate": 3.5487410623015307e-06, "loss": 0.0045, "step": 31911 }, { "epoch": 5.85, "learning_rate": 3.548173806900144e-06, "loss": 0.0064, "step": 31912 }, { "epoch": 5.85, "learning_rate": 3.5476065870613684e-06, "loss": 0.0065, "step": 31913 }, { "epoch": 5.85, "learning_rate": 3.547039402788339e-06, "loss": 0.0094, "step": 31914 }, { "epoch": 5.85, "learning_rate": 3.54647225408418e-06, "loss": 0.0065, "step": 31915 }, { "epoch": 5.85, "learning_rate": 3.5459051409520174e-06, "loss": 0.0099, "step": 31916 }, { "epoch": 5.85, "learning_rate": 3.5453380633949764e-06, "loss": 0.0044, "step": 31917 }, { "epoch": 5.85, "learning_rate": 3.5447710214161834e-06, "loss": 0.0067, "step": 31918 }, { "epoch": 5.85, "learning_rate": 3.544204015018764e-06, "loss": 0.0057, "step": 31919 }, { "epoch": 5.85, "learning_rate": 3.5436370442058434e-06, "loss": 0.0069, "step": 31920 }, { "epoch": 5.85, "learning_rate": 3.543070108980543e-06, "loss": 0.0115, "step": 31921 }, { "epoch": 5.85, "learning_rate": 3.5425032093459957e-06, "loss": 0.0072, "step": 31922 }, { "epoch": 5.85, "learning_rate": 3.54193634530532e-06, "loss": 0.0063, "step": 31923 }, { "epoch": 5.85, "learning_rate": 3.541369516861648e-06, "loss": 0.0075, "step": 31924 }, { "epoch": 5.85, "learning_rate": 3.540802724018094e-06, "loss": 0.0072, "step": 31925 }, { "epoch": 5.85, "learning_rate": 3.540235966777783e-06, "loss": 0.0084, "step": 31926 }, { "epoch": 5.85, "learning_rate": 3.5396692451438464e-06, "loss": 0.007, "step": 31927 }, { "epoch": 5.86, "learning_rate": 3.539102559119405e-06, "loss": 0.0086, "step": 31928 }, { "epoch": 5.86, "learning_rate": 3.538535908707581e-06, "loss": 0.0078, "step": 31929 }, { "epoch": 5.86, "learning_rate": 3.5379692939114987e-06, "loss": 0.0061, "step": 31930 }, { "epoch": 5.86, "learning_rate": 3.5374027147342803e-06, "loss": 0.0085, "step": 31931 }, { "epoch": 5.86, "learning_rate": 3.536836171179051e-06, "loss": 0.0069, "step": 31932 }, { "epoch": 5.86, "learning_rate": 3.5362696632489312e-06, "loss": 0.007, "step": 31933 }, { "epoch": 5.86, "learning_rate": 3.535703190947044e-06, "loss": 0.0065, "step": 31934 }, { "epoch": 5.86, "learning_rate": 3.5351367542765126e-06, "loss": 0.0094, "step": 31935 }, { "epoch": 5.86, "learning_rate": 3.5345703532404595e-06, "loss": 0.0086, "step": 31936 }, { "epoch": 5.86, "learning_rate": 3.534003987842005e-06, "loss": 0.0071, "step": 31937 }, { "epoch": 5.86, "learning_rate": 3.5334376580842734e-06, "loss": 0.0056, "step": 31938 }, { "epoch": 5.86, "learning_rate": 3.5328713639703805e-06, "loss": 0.006, "step": 31939 }, { "epoch": 5.86, "learning_rate": 3.532305105503455e-06, "loss": 0.0093, "step": 31940 }, { "epoch": 5.86, "learning_rate": 3.5317388826866174e-06, "loss": 0.0092, "step": 31941 }, { "epoch": 5.86, "learning_rate": 3.5311726955229885e-06, "loss": 0.0064, "step": 31942 }, { "epoch": 5.86, "learning_rate": 3.53060654401568e-06, "loss": 0.0069, "step": 31943 }, { "epoch": 5.86, "learning_rate": 3.5300404281678237e-06, "loss": 0.0077, "step": 31944 }, { "epoch": 5.86, "learning_rate": 3.5294743479825367e-06, "loss": 0.0084, "step": 31945 }, { "epoch": 5.86, "learning_rate": 3.5289083034629378e-06, "loss": 0.0078, "step": 31946 }, { "epoch": 5.86, "learning_rate": 3.5283422946121482e-06, "loss": 0.0082, "step": 31947 }, { "epoch": 5.86, "learning_rate": 3.527776321433288e-06, "loss": 0.0038, "step": 31948 }, { "epoch": 5.86, "learning_rate": 3.527210383929476e-06, "loss": 0.0071, "step": 31949 }, { "epoch": 5.86, "learning_rate": 3.5266444821038327e-06, "loss": 0.0062, "step": 31950 }, { "epoch": 5.86, "learning_rate": 3.526078615959475e-06, "loss": 0.0068, "step": 31951 }, { "epoch": 5.86, "learning_rate": 3.525512785499522e-06, "loss": 0.008, "step": 31952 }, { "epoch": 5.86, "learning_rate": 3.5249469907270993e-06, "loss": 0.0073, "step": 31953 }, { "epoch": 5.86, "learning_rate": 3.5243812316453187e-06, "loss": 0.007, "step": 31954 }, { "epoch": 5.86, "learning_rate": 3.5238155082573e-06, "loss": 0.005, "step": 31955 }, { "epoch": 5.86, "learning_rate": 3.5232498205661614e-06, "loss": 0.0055, "step": 31956 }, { "epoch": 5.86, "learning_rate": 3.522684168575018e-06, "loss": 0.0076, "step": 31957 }, { "epoch": 5.86, "learning_rate": 3.5221185522869948e-06, "loss": 0.0089, "step": 31958 }, { "epoch": 5.86, "learning_rate": 3.52155297170521e-06, "loss": 0.0057, "step": 31959 }, { "epoch": 5.86, "learning_rate": 3.5209874268327737e-06, "loss": 0.009, "step": 31960 }, { "epoch": 5.86, "learning_rate": 3.5204219176728016e-06, "loss": 0.0069, "step": 31961 }, { "epoch": 5.86, "learning_rate": 3.51985644422842e-06, "loss": 0.0059, "step": 31962 }, { "epoch": 5.86, "learning_rate": 3.519291006502742e-06, "loss": 0.0063, "step": 31963 }, { "epoch": 5.86, "learning_rate": 3.5187256044988838e-06, "loss": 0.0063, "step": 31964 }, { "epoch": 5.86, "learning_rate": 3.518160238219961e-06, "loss": 0.0088, "step": 31965 }, { "epoch": 5.86, "learning_rate": 3.5175949076690906e-06, "loss": 0.0092, "step": 31966 }, { "epoch": 5.86, "learning_rate": 3.5170296128493896e-06, "loss": 0.0052, "step": 31967 }, { "epoch": 5.86, "learning_rate": 3.516464353763974e-06, "loss": 0.0075, "step": 31968 }, { "epoch": 5.86, "learning_rate": 3.5158991304159572e-06, "loss": 0.009, "step": 31969 }, { "epoch": 5.86, "learning_rate": 3.515333942808454e-06, "loss": 0.0076, "step": 31970 }, { "epoch": 5.86, "learning_rate": 3.514768790944588e-06, "loss": 0.0053, "step": 31971 }, { "epoch": 5.86, "learning_rate": 3.5142036748274666e-06, "loss": 0.0079, "step": 31972 }, { "epoch": 5.86, "learning_rate": 3.5136385944602046e-06, "loss": 0.0071, "step": 31973 }, { "epoch": 5.86, "learning_rate": 3.513073549845917e-06, "loss": 0.0064, "step": 31974 }, { "epoch": 5.86, "learning_rate": 3.512508540987721e-06, "loss": 0.0056, "step": 31975 }, { "epoch": 5.86, "learning_rate": 3.5119435678887328e-06, "loss": 0.0077, "step": 31976 }, { "epoch": 5.86, "learning_rate": 3.511378630552065e-06, "loss": 0.0063, "step": 31977 }, { "epoch": 5.86, "learning_rate": 3.5108137289808275e-06, "loss": 0.0055, "step": 31978 }, { "epoch": 5.86, "learning_rate": 3.510248863178133e-06, "loss": 0.0075, "step": 31979 }, { "epoch": 5.86, "learning_rate": 3.5096840331471017e-06, "loss": 0.0059, "step": 31980 }, { "epoch": 5.86, "learning_rate": 3.509119238890846e-06, "loss": 0.0069, "step": 31981 }, { "epoch": 5.87, "learning_rate": 3.5085544804124748e-06, "loss": 0.0086, "step": 31982 }, { "epoch": 5.87, "learning_rate": 3.5079897577151047e-06, "loss": 0.0119, "step": 31983 }, { "epoch": 5.87, "learning_rate": 3.5074250708018465e-06, "loss": 0.0091, "step": 31984 }, { "epoch": 5.87, "learning_rate": 3.5068604196758136e-06, "loss": 0.0072, "step": 31985 }, { "epoch": 5.87, "learning_rate": 3.5062958043401187e-06, "loss": 0.0083, "step": 31986 }, { "epoch": 5.87, "learning_rate": 3.5057312247978736e-06, "loss": 0.0062, "step": 31987 }, { "epoch": 5.87, "learning_rate": 3.5051666810521858e-06, "loss": 0.0046, "step": 31988 }, { "epoch": 5.87, "learning_rate": 3.5046021731061786e-06, "loss": 0.0095, "step": 31989 }, { "epoch": 5.87, "learning_rate": 3.5040377009629534e-06, "loss": 0.0055, "step": 31990 }, { "epoch": 5.87, "learning_rate": 3.503473264625624e-06, "loss": 0.0065, "step": 31991 }, { "epoch": 5.87, "learning_rate": 3.5029088640973e-06, "loss": 0.0077, "step": 31992 }, { "epoch": 5.87, "learning_rate": 3.5023444993810963e-06, "loss": 0.0072, "step": 31993 }, { "epoch": 5.87, "learning_rate": 3.5017801704801224e-06, "loss": 0.0074, "step": 31994 }, { "epoch": 5.87, "learning_rate": 3.5012158773974926e-06, "loss": 0.0075, "step": 31995 }, { "epoch": 5.87, "learning_rate": 3.5006516201363063e-06, "loss": 0.0083, "step": 31996 }, { "epoch": 5.87, "learning_rate": 3.500087398699683e-06, "loss": 0.0073, "step": 31997 }, { "epoch": 5.87, "learning_rate": 3.4995232130907307e-06, "loss": 0.0072, "step": 31998 }, { "epoch": 5.87, "learning_rate": 3.4989590633125583e-06, "loss": 0.0063, "step": 31999 }, { "epoch": 5.87, "learning_rate": 3.498394949368277e-06, "loss": 0.0088, "step": 32000 }, { "epoch": 5.87, "learning_rate": 3.4978308712609934e-06, "loss": 0.0084, "step": 32001 }, { "epoch": 5.87, "learning_rate": 3.4972668289938194e-06, "loss": 0.007, "step": 32002 }, { "epoch": 5.87, "learning_rate": 3.4967028225698628e-06, "loss": 0.0079, "step": 32003 }, { "epoch": 5.87, "learning_rate": 3.4961388519922325e-06, "loss": 0.008, "step": 32004 }, { "epoch": 5.87, "learning_rate": 3.495574917264034e-06, "loss": 0.0068, "step": 32005 }, { "epoch": 5.87, "learning_rate": 3.4950110183883857e-06, "loss": 0.0075, "step": 32006 }, { "epoch": 5.87, "learning_rate": 3.4944471553683857e-06, "loss": 0.0078, "step": 32007 }, { "epoch": 5.87, "learning_rate": 3.493883328207146e-06, "loss": 0.006, "step": 32008 }, { "epoch": 5.87, "learning_rate": 3.493319536907773e-06, "loss": 0.0069, "step": 32009 }, { "epoch": 5.87, "learning_rate": 3.4927557814733714e-06, "loss": 0.0071, "step": 32010 }, { "epoch": 5.87, "learning_rate": 3.492192061907057e-06, "loss": 0.0079, "step": 32011 }, { "epoch": 5.87, "learning_rate": 3.491628378211931e-06, "loss": 0.006, "step": 32012 }, { "epoch": 5.87, "learning_rate": 3.491064730391106e-06, "loss": 0.0096, "step": 32013 }, { "epoch": 5.87, "learning_rate": 3.490501118447678e-06, "loss": 0.0075, "step": 32014 }, { "epoch": 5.87, "learning_rate": 3.4899375423847637e-06, "loss": 0.0074, "step": 32015 }, { "epoch": 5.87, "learning_rate": 3.4893740022054656e-06, "loss": 0.0088, "step": 32016 }, { "epoch": 5.87, "learning_rate": 3.4888104979128913e-06, "loss": 0.0063, "step": 32017 }, { "epoch": 5.87, "learning_rate": 3.4882470295101456e-06, "loss": 0.0071, "step": 32018 }, { "epoch": 5.87, "learning_rate": 3.487683597000335e-06, "loss": 0.0088, "step": 32019 }, { "epoch": 5.87, "learning_rate": 3.4871202003865656e-06, "loss": 0.0078, "step": 32020 }, { "epoch": 5.87, "learning_rate": 3.486556839671942e-06, "loss": 0.007, "step": 32021 }, { "epoch": 5.87, "learning_rate": 3.485993514859569e-06, "loss": 0.0055, "step": 32022 }, { "epoch": 5.87, "learning_rate": 3.4854302259525495e-06, "loss": 0.0049, "step": 32023 }, { "epoch": 5.87, "learning_rate": 3.4848669729539975e-06, "loss": 0.0094, "step": 32024 }, { "epoch": 5.87, "learning_rate": 3.4843037558670077e-06, "loss": 0.0075, "step": 32025 }, { "epoch": 5.87, "learning_rate": 3.4837405746946896e-06, "loss": 0.008, "step": 32026 }, { "epoch": 5.87, "learning_rate": 3.483177429440142e-06, "loss": 0.0059, "step": 32027 }, { "epoch": 5.87, "learning_rate": 3.4826143201064756e-06, "loss": 0.0084, "step": 32028 }, { "epoch": 5.87, "learning_rate": 3.4820512466967926e-06, "loss": 0.0096, "step": 32029 }, { "epoch": 5.87, "learning_rate": 3.4814882092141968e-06, "loss": 0.0062, "step": 32030 }, { "epoch": 5.87, "learning_rate": 3.480925207661793e-06, "loss": 0.0068, "step": 32031 }, { "epoch": 5.87, "learning_rate": 3.4803622420426754e-06, "loss": 0.0058, "step": 32032 }, { "epoch": 5.87, "learning_rate": 3.479799312359957e-06, "loss": 0.0059, "step": 32033 }, { "epoch": 5.87, "learning_rate": 3.479236418616738e-06, "loss": 0.0077, "step": 32034 }, { "epoch": 5.87, "learning_rate": 3.4786735608161205e-06, "loss": 0.0094, "step": 32035 }, { "epoch": 5.87, "learning_rate": 3.4781107389612066e-06, "loss": 0.0063, "step": 32036 }, { "epoch": 5.88, "learning_rate": 3.4775479530551003e-06, "loss": 0.0069, "step": 32037 }, { "epoch": 5.88, "learning_rate": 3.4769852031009012e-06, "loss": 0.0066, "step": 32038 }, { "epoch": 5.88, "learning_rate": 3.476422489101713e-06, "loss": 0.0066, "step": 32039 }, { "epoch": 5.88, "learning_rate": 3.4758598110606366e-06, "loss": 0.0069, "step": 32040 }, { "epoch": 5.88, "learning_rate": 3.4752971689807713e-06, "loss": 0.0049, "step": 32041 }, { "epoch": 5.88, "learning_rate": 3.4747345628652274e-06, "loss": 0.0086, "step": 32042 }, { "epoch": 5.88, "learning_rate": 3.4741719927170957e-06, "loss": 0.0074, "step": 32043 }, { "epoch": 5.88, "learning_rate": 3.4736094585394807e-06, "loss": 0.0064, "step": 32044 }, { "epoch": 5.88, "learning_rate": 3.47304696033548e-06, "loss": 0.0079, "step": 32045 }, { "epoch": 5.88, "learning_rate": 3.4724844981082005e-06, "loss": 0.0076, "step": 32046 }, { "epoch": 5.88, "learning_rate": 3.4719220718607403e-06, "loss": 0.0066, "step": 32047 }, { "epoch": 5.88, "learning_rate": 3.471359681596198e-06, "loss": 0.0064, "step": 32048 }, { "epoch": 5.88, "learning_rate": 3.4707973273176742e-06, "loss": 0.0086, "step": 32049 }, { "epoch": 5.88, "learning_rate": 3.4702350090282677e-06, "loss": 0.0063, "step": 32050 }, { "epoch": 5.88, "learning_rate": 3.469672726731079e-06, "loss": 0.0061, "step": 32051 }, { "epoch": 5.88, "learning_rate": 3.469110480429209e-06, "loss": 0.0086, "step": 32052 }, { "epoch": 5.88, "learning_rate": 3.4685482701257534e-06, "loss": 0.0056, "step": 32053 }, { "epoch": 5.88, "learning_rate": 3.4679860958238134e-06, "loss": 0.0061, "step": 32054 }, { "epoch": 5.88, "learning_rate": 3.4674239575264856e-06, "loss": 0.0071, "step": 32055 }, { "epoch": 5.88, "learning_rate": 3.4668618552368715e-06, "loss": 0.0058, "step": 32056 }, { "epoch": 5.88, "learning_rate": 3.4662997889580674e-06, "loss": 0.0071, "step": 32057 }, { "epoch": 5.88, "learning_rate": 3.4657377586931695e-06, "loss": 0.0057, "step": 32058 }, { "epoch": 5.88, "learning_rate": 3.4651757644452798e-06, "loss": 0.0073, "step": 32059 }, { "epoch": 5.88, "learning_rate": 3.4646138062174996e-06, "loss": 0.0078, "step": 32060 }, { "epoch": 5.88, "learning_rate": 3.4640518840129168e-06, "loss": 0.0075, "step": 32061 }, { "epoch": 5.88, "learning_rate": 3.4634899978346324e-06, "loss": 0.0059, "step": 32062 }, { "epoch": 5.88, "learning_rate": 3.4629281476857423e-06, "loss": 0.006, "step": 32063 }, { "epoch": 5.88, "learning_rate": 3.462366333569348e-06, "loss": 0.0084, "step": 32064 }, { "epoch": 5.88, "learning_rate": 3.4618045554885438e-06, "loss": 0.0056, "step": 32065 }, { "epoch": 5.88, "learning_rate": 3.4612428134464293e-06, "loss": 0.0042, "step": 32066 }, { "epoch": 5.88, "learning_rate": 3.460681107446091e-06, "loss": 0.0081, "step": 32067 }, { "epoch": 5.88, "learning_rate": 3.4601194374906356e-06, "loss": 0.0096, "step": 32068 }, { "epoch": 5.88, "learning_rate": 3.4595578035831545e-06, "loss": 0.0073, "step": 32069 }, { "epoch": 5.88, "learning_rate": 3.4589962057267436e-06, "loss": 0.0082, "step": 32070 }, { "epoch": 5.88, "learning_rate": 3.4584346439245e-06, "loss": 0.0059, "step": 32071 }, { "epoch": 5.88, "learning_rate": 3.457873118179517e-06, "loss": 0.0064, "step": 32072 }, { "epoch": 5.88, "learning_rate": 3.4573116284948906e-06, "loss": 0.0056, "step": 32073 }, { "epoch": 5.88, "learning_rate": 3.4567501748737153e-06, "loss": 0.0065, "step": 32074 }, { "epoch": 5.88, "learning_rate": 3.4561887573190877e-06, "loss": 0.0074, "step": 32075 }, { "epoch": 5.88, "learning_rate": 3.4556273758340965e-06, "loss": 0.0066, "step": 32076 }, { "epoch": 5.88, "learning_rate": 3.455066030421844e-06, "loss": 0.0102, "step": 32077 }, { "epoch": 5.88, "learning_rate": 3.4545047210854256e-06, "loss": 0.0059, "step": 32078 }, { "epoch": 5.88, "learning_rate": 3.4539434478279254e-06, "loss": 0.0061, "step": 32079 }, { "epoch": 5.88, "learning_rate": 3.4533822106524393e-06, "loss": 0.006, "step": 32080 }, { "epoch": 5.88, "learning_rate": 3.4528210095620683e-06, "loss": 0.007, "step": 32081 }, { "epoch": 5.88, "learning_rate": 3.4522598445598998e-06, "loss": 0.0067, "step": 32082 }, { "epoch": 5.88, "learning_rate": 3.45169871564903e-06, "loss": 0.0068, "step": 32083 }, { "epoch": 5.88, "learning_rate": 3.4511376228325523e-06, "loss": 0.0085, "step": 32084 }, { "epoch": 5.88, "learning_rate": 3.4505765661135515e-06, "loss": 0.0068, "step": 32085 }, { "epoch": 5.88, "learning_rate": 3.450015545495129e-06, "loss": 0.0067, "step": 32086 }, { "epoch": 5.88, "learning_rate": 3.4494545609803742e-06, "loss": 0.0083, "step": 32087 }, { "epoch": 5.88, "learning_rate": 3.4488936125723794e-06, "loss": 0.0055, "step": 32088 }, { "epoch": 5.88, "learning_rate": 3.448332700274237e-06, "loss": 0.0087, "step": 32089 }, { "epoch": 5.88, "learning_rate": 3.4477718240890377e-06, "loss": 0.0077, "step": 32090 }, { "epoch": 5.89, "learning_rate": 3.447210984019874e-06, "loss": 0.0081, "step": 32091 }, { "epoch": 5.89, "learning_rate": 3.446650180069837e-06, "loss": 0.0049, "step": 32092 }, { "epoch": 5.89, "learning_rate": 3.4460894122420143e-06, "loss": 0.0075, "step": 32093 }, { "epoch": 5.89, "learning_rate": 3.4455286805395025e-06, "loss": 0.0076, "step": 32094 }, { "epoch": 5.89, "learning_rate": 3.4449679849653918e-06, "loss": 0.007, "step": 32095 }, { "epoch": 5.89, "learning_rate": 3.444407325522773e-06, "loss": 0.0073, "step": 32096 }, { "epoch": 5.89, "learning_rate": 3.4438467022147328e-06, "loss": 0.0055, "step": 32097 }, { "epoch": 5.89, "learning_rate": 3.4432861150443585e-06, "loss": 0.0065, "step": 32098 }, { "epoch": 5.89, "learning_rate": 3.442725564014748e-06, "loss": 0.0065, "step": 32099 }, { "epoch": 5.89, "learning_rate": 3.442165049128989e-06, "loss": 0.0092, "step": 32100 }, { "epoch": 5.89, "learning_rate": 3.441604570390169e-06, "loss": 0.0093, "step": 32101 }, { "epoch": 5.89, "learning_rate": 3.4410441278013786e-06, "loss": 0.0063, "step": 32102 }, { "epoch": 5.89, "learning_rate": 3.4404837213657074e-06, "loss": 0.0072, "step": 32103 }, { "epoch": 5.89, "learning_rate": 3.439923351086242e-06, "loss": 0.0057, "step": 32104 }, { "epoch": 5.89, "learning_rate": 3.4393630169660743e-06, "loss": 0.0042, "step": 32105 }, { "epoch": 5.89, "learning_rate": 3.4388027190082906e-06, "loss": 0.0108, "step": 32106 }, { "epoch": 5.89, "learning_rate": 3.43824245721598e-06, "loss": 0.0083, "step": 32107 }, { "epoch": 5.89, "learning_rate": 3.4376822315922308e-06, "loss": 0.005, "step": 32108 }, { "epoch": 5.89, "learning_rate": 3.437122042140131e-06, "loss": 0.0066, "step": 32109 }, { "epoch": 5.89, "learning_rate": 3.436561888862767e-06, "loss": 0.0057, "step": 32110 }, { "epoch": 5.89, "learning_rate": 3.4360017717632254e-06, "loss": 0.009, "step": 32111 }, { "epoch": 5.89, "learning_rate": 3.4354416908446e-06, "loss": 0.0073, "step": 32112 }, { "epoch": 5.89, "learning_rate": 3.4348816461099754e-06, "loss": 0.0063, "step": 32113 }, { "epoch": 5.89, "learning_rate": 3.4343216375624344e-06, "loss": 0.0064, "step": 32114 }, { "epoch": 5.89, "learning_rate": 3.4337616652050665e-06, "loss": 0.0064, "step": 32115 }, { "epoch": 5.89, "learning_rate": 3.4332017290409547e-06, "loss": 0.0078, "step": 32116 }, { "epoch": 5.89, "learning_rate": 3.432641829073192e-06, "loss": 0.0055, "step": 32117 }, { "epoch": 5.89, "learning_rate": 3.4320819653048622e-06, "loss": 0.007, "step": 32118 }, { "epoch": 5.89, "learning_rate": 3.4315221377390496e-06, "loss": 0.007, "step": 32119 }, { "epoch": 5.89, "learning_rate": 3.4309623463788413e-06, "loss": 0.0057, "step": 32120 }, { "epoch": 5.89, "learning_rate": 3.430402591227322e-06, "loss": 0.0058, "step": 32121 }, { "epoch": 5.89, "learning_rate": 3.4298428722875777e-06, "loss": 0.0065, "step": 32122 }, { "epoch": 5.89, "learning_rate": 3.4292831895626944e-06, "loss": 0.0087, "step": 32123 }, { "epoch": 5.89, "learning_rate": 3.428723543055752e-06, "loss": 0.0077, "step": 32124 }, { "epoch": 5.89, "learning_rate": 3.428163932769846e-06, "loss": 0.0076, "step": 32125 }, { "epoch": 5.89, "learning_rate": 3.42760435870805e-06, "loss": 0.0046, "step": 32126 }, { "epoch": 5.89, "learning_rate": 3.427044820873454e-06, "loss": 0.0077, "step": 32127 }, { "epoch": 5.89, "learning_rate": 3.4264853192691404e-06, "loss": 0.0063, "step": 32128 }, { "epoch": 5.89, "learning_rate": 3.425925853898191e-06, "loss": 0.0074, "step": 32129 }, { "epoch": 5.89, "learning_rate": 3.425366424763695e-06, "loss": 0.0089, "step": 32130 }, { "epoch": 5.89, "learning_rate": 3.4248070318687365e-06, "loss": 0.006, "step": 32131 }, { "epoch": 5.89, "learning_rate": 3.424247675216393e-06, "loss": 0.0084, "step": 32132 }, { "epoch": 5.89, "learning_rate": 3.4236883548097464e-06, "loss": 0.0058, "step": 32133 }, { "epoch": 5.89, "learning_rate": 3.423129070651888e-06, "loss": 0.0063, "step": 32134 }, { "epoch": 5.89, "learning_rate": 3.422569822745896e-06, "loss": 0.0082, "step": 32135 }, { "epoch": 5.89, "learning_rate": 3.4220106110948525e-06, "loss": 0.0093, "step": 32136 }, { "epoch": 5.89, "learning_rate": 3.4214514357018404e-06, "loss": 0.0091, "step": 32137 }, { "epoch": 5.89, "learning_rate": 3.420892296569943e-06, "loss": 0.0061, "step": 32138 }, { "epoch": 5.89, "learning_rate": 3.420333193702241e-06, "loss": 0.0046, "step": 32139 }, { "epoch": 5.89, "learning_rate": 3.419774127101817e-06, "loss": 0.0057, "step": 32140 }, { "epoch": 5.89, "learning_rate": 3.4192150967717507e-06, "loss": 0.0061, "step": 32141 }, { "epoch": 5.89, "learning_rate": 3.4186561027151234e-06, "loss": 0.0069, "step": 32142 }, { "epoch": 5.89, "learning_rate": 3.418097144935024e-06, "loss": 0.0039, "step": 32143 }, { "epoch": 5.89, "learning_rate": 3.4175382234345234e-06, "loss": 0.0053, "step": 32144 }, { "epoch": 5.89, "learning_rate": 3.4169793382167073e-06, "loss": 0.0067, "step": 32145 }, { "epoch": 5.9, "learning_rate": 3.416420489284651e-06, "loss": 0.0066, "step": 32146 }, { "epoch": 5.9, "learning_rate": 3.4158616766414433e-06, "loss": 0.0076, "step": 32147 }, { "epoch": 5.9, "learning_rate": 3.4153029002901593e-06, "loss": 0.0086, "step": 32148 }, { "epoch": 5.9, "learning_rate": 3.414744160233885e-06, "loss": 0.007, "step": 32149 }, { "epoch": 5.9, "learning_rate": 3.4141854564756904e-06, "loss": 0.0044, "step": 32150 }, { "epoch": 5.9, "learning_rate": 3.413626789018657e-06, "loss": 0.0076, "step": 32151 }, { "epoch": 5.9, "learning_rate": 3.4130681578658697e-06, "loss": 0.0058, "step": 32152 }, { "epoch": 5.9, "learning_rate": 3.4125095630204063e-06, "loss": 0.0057, "step": 32153 }, { "epoch": 5.9, "learning_rate": 3.4119510044853434e-06, "loss": 0.0048, "step": 32154 }, { "epoch": 5.9, "learning_rate": 3.411392482263761e-06, "loss": 0.007, "step": 32155 }, { "epoch": 5.9, "learning_rate": 3.4108339963587376e-06, "loss": 0.0072, "step": 32156 }, { "epoch": 5.9, "learning_rate": 3.4102755467733527e-06, "loss": 0.0072, "step": 32157 }, { "epoch": 5.9, "learning_rate": 3.409717133510683e-06, "loss": 0.0067, "step": 32158 }, { "epoch": 5.9, "learning_rate": 3.409158756573806e-06, "loss": 0.0077, "step": 32159 }, { "epoch": 5.9, "learning_rate": 3.4086004159658003e-06, "loss": 0.0077, "step": 32160 }, { "epoch": 5.9, "learning_rate": 3.4080421116897442e-06, "loss": 0.0073, "step": 32161 }, { "epoch": 5.9, "learning_rate": 3.4074838437487144e-06, "loss": 0.0092, "step": 32162 }, { "epoch": 5.9, "learning_rate": 3.4069256121457884e-06, "loss": 0.0068, "step": 32163 }, { "epoch": 5.9, "learning_rate": 3.4063674168840386e-06, "loss": 0.0086, "step": 32164 }, { "epoch": 5.9, "learning_rate": 3.4058092579665502e-06, "loss": 0.0085, "step": 32165 }, { "epoch": 5.9, "learning_rate": 3.4052511353963946e-06, "loss": 0.0077, "step": 32166 }, { "epoch": 5.9, "learning_rate": 3.404693049176654e-06, "loss": 0.0058, "step": 32167 }, { "epoch": 5.9, "learning_rate": 3.4041349993103954e-06, "loss": 0.0062, "step": 32168 }, { "epoch": 5.9, "learning_rate": 3.403576985800696e-06, "loss": 0.0069, "step": 32169 }, { "epoch": 5.9, "learning_rate": 3.403019008650639e-06, "loss": 0.0057, "step": 32170 }, { "epoch": 5.9, "learning_rate": 3.402461067863295e-06, "loss": 0.007, "step": 32171 }, { "epoch": 5.9, "learning_rate": 3.4019031634417397e-06, "loss": 0.0076, "step": 32172 }, { "epoch": 5.9, "learning_rate": 3.4013452953890503e-06, "loss": 0.0065, "step": 32173 }, { "epoch": 5.9, "learning_rate": 3.4007874637082983e-06, "loss": 0.0061, "step": 32174 }, { "epoch": 5.9, "learning_rate": 3.4002296684025614e-06, "loss": 0.0065, "step": 32175 }, { "epoch": 5.9, "learning_rate": 3.399671909474913e-06, "loss": 0.0088, "step": 32176 }, { "epoch": 5.9, "learning_rate": 3.399114186928425e-06, "loss": 0.0075, "step": 32177 }, { "epoch": 5.9, "learning_rate": 3.3985565007661804e-06, "loss": 0.0062, "step": 32178 }, { "epoch": 5.9, "learning_rate": 3.3979988509912443e-06, "loss": 0.0071, "step": 32179 }, { "epoch": 5.9, "learning_rate": 3.3974412376066923e-06, "loss": 0.0059, "step": 32180 }, { "epoch": 5.9, "learning_rate": 3.3968836606155986e-06, "loss": 0.0087, "step": 32181 }, { "epoch": 5.9, "learning_rate": 3.396326120021034e-06, "loss": 0.0064, "step": 32182 }, { "epoch": 5.9, "learning_rate": 3.395768615826078e-06, "loss": 0.0067, "step": 32183 }, { "epoch": 5.9, "learning_rate": 3.3952111480338003e-06, "loss": 0.0054, "step": 32184 }, { "epoch": 5.9, "learning_rate": 3.394653716647277e-06, "loss": 0.0061, "step": 32185 }, { "epoch": 5.9, "learning_rate": 3.3940963216695697e-06, "loss": 0.0066, "step": 32186 }, { "epoch": 5.9, "learning_rate": 3.3935389631037617e-06, "loss": 0.0049, "step": 32187 }, { "epoch": 5.9, "learning_rate": 3.3929816409529225e-06, "loss": 0.0059, "step": 32188 }, { "epoch": 5.9, "learning_rate": 3.392424355220123e-06, "loss": 0.0055, "step": 32189 }, { "epoch": 5.9, "learning_rate": 3.3918671059084353e-06, "loss": 0.0058, "step": 32190 }, { "epoch": 5.9, "learning_rate": 3.39130989302093e-06, "loss": 0.0044, "step": 32191 }, { "epoch": 5.9, "learning_rate": 3.39075271656068e-06, "loss": 0.0077, "step": 32192 }, { "epoch": 5.9, "learning_rate": 3.3901955765307557e-06, "loss": 0.0049, "step": 32193 }, { "epoch": 5.9, "learning_rate": 3.389638472934229e-06, "loss": 0.0074, "step": 32194 }, { "epoch": 5.9, "learning_rate": 3.389081405774166e-06, "loss": 0.0057, "step": 32195 }, { "epoch": 5.9, "learning_rate": 3.388524375053648e-06, "loss": 0.0068, "step": 32196 }, { "epoch": 5.9, "learning_rate": 3.387967380775734e-06, "loss": 0.0052, "step": 32197 }, { "epoch": 5.9, "learning_rate": 3.3874104229435e-06, "loss": 0.0076, "step": 32198 }, { "epoch": 5.9, "learning_rate": 3.386853501560011e-06, "loss": 0.0056, "step": 32199 }, { "epoch": 5.91, "learning_rate": 3.386296616628344e-06, "loss": 0.0047, "step": 32200 }, { "epoch": 5.91, "learning_rate": 3.3857397681515646e-06, "loss": 0.0102, "step": 32201 }, { "epoch": 5.91, "learning_rate": 3.385182956132743e-06, "loss": 0.0052, "step": 32202 }, { "epoch": 5.91, "learning_rate": 3.3846261805749514e-06, "loss": 0.0057, "step": 32203 }, { "epoch": 5.91, "learning_rate": 3.384069441481248e-06, "loss": 0.0055, "step": 32204 }, { "epoch": 5.91, "learning_rate": 3.3835127388547127e-06, "loss": 0.0057, "step": 32205 }, { "epoch": 5.91, "learning_rate": 3.38295607269841e-06, "loss": 0.0065, "step": 32206 }, { "epoch": 5.91, "learning_rate": 3.382399443015408e-06, "loss": 0.0085, "step": 32207 }, { "epoch": 5.91, "learning_rate": 3.3818428498087753e-06, "loss": 0.0066, "step": 32208 }, { "epoch": 5.91, "learning_rate": 3.3812862930815806e-06, "loss": 0.0065, "step": 32209 }, { "epoch": 5.91, "learning_rate": 3.3807297728368902e-06, "loss": 0.0068, "step": 32210 }, { "epoch": 5.91, "learning_rate": 3.3801732890777727e-06, "loss": 0.0082, "step": 32211 }, { "epoch": 5.91, "learning_rate": 3.379616841807295e-06, "loss": 0.0078, "step": 32212 }, { "epoch": 5.91, "learning_rate": 3.3790604310285214e-06, "loss": 0.0059, "step": 32213 }, { "epoch": 5.91, "learning_rate": 3.3785040567445282e-06, "loss": 0.0047, "step": 32214 }, { "epoch": 5.91, "learning_rate": 3.377947718958372e-06, "loss": 0.0085, "step": 32215 }, { "epoch": 5.91, "learning_rate": 3.3773914176731237e-06, "loss": 0.0077, "step": 32216 }, { "epoch": 5.91, "learning_rate": 3.3768351528918465e-06, "loss": 0.0058, "step": 32217 }, { "epoch": 5.91, "learning_rate": 3.376278924617611e-06, "loss": 0.0083, "step": 32218 }, { "epoch": 5.91, "learning_rate": 3.3757227328534814e-06, "loss": 0.0061, "step": 32219 }, { "epoch": 5.91, "learning_rate": 3.3751665776025277e-06, "loss": 0.0075, "step": 32220 }, { "epoch": 5.91, "learning_rate": 3.374610458867803e-06, "loss": 0.0072, "step": 32221 }, { "epoch": 5.91, "learning_rate": 3.374054376652385e-06, "loss": 0.0067, "step": 32222 }, { "epoch": 5.91, "learning_rate": 3.373498330959334e-06, "loss": 0.0088, "step": 32223 }, { "epoch": 5.91, "learning_rate": 3.372942321791716e-06, "loss": 0.0091, "step": 32224 }, { "epoch": 5.91, "learning_rate": 3.3723863491525943e-06, "loss": 0.0069, "step": 32225 }, { "epoch": 5.91, "learning_rate": 3.371830413045035e-06, "loss": 0.0069, "step": 32226 }, { "epoch": 5.91, "learning_rate": 3.3712745134721014e-06, "loss": 0.0072, "step": 32227 }, { "epoch": 5.91, "learning_rate": 3.3707186504368584e-06, "loss": 0.0055, "step": 32228 }, { "epoch": 5.91, "learning_rate": 3.370162823942369e-06, "loss": 0.0065, "step": 32229 }, { "epoch": 5.91, "learning_rate": 3.369607033991694e-06, "loss": 0.0061, "step": 32230 }, { "epoch": 5.91, "learning_rate": 3.369051280587904e-06, "loss": 0.0049, "step": 32231 }, { "epoch": 5.91, "learning_rate": 3.368495563734062e-06, "loss": 0.0091, "step": 32232 }, { "epoch": 5.91, "learning_rate": 3.3679398834332242e-06, "loss": 0.0071, "step": 32233 }, { "epoch": 5.91, "learning_rate": 3.367384239688458e-06, "loss": 0.0072, "step": 32234 }, { "epoch": 5.91, "learning_rate": 3.366828632502821e-06, "loss": 0.0086, "step": 32235 }, { "epoch": 5.91, "learning_rate": 3.3662730618793835e-06, "loss": 0.0072, "step": 32236 }, { "epoch": 5.91, "learning_rate": 3.365717527821205e-06, "loss": 0.0054, "step": 32237 }, { "epoch": 5.91, "learning_rate": 3.3651620303313492e-06, "loss": 0.0075, "step": 32238 }, { "epoch": 5.91, "learning_rate": 3.3646065694128692e-06, "loss": 0.0075, "step": 32239 }, { "epoch": 5.91, "learning_rate": 3.3640511450688373e-06, "loss": 0.0064, "step": 32240 }, { "epoch": 5.91, "learning_rate": 3.363495757302311e-06, "loss": 0.0099, "step": 32241 }, { "epoch": 5.91, "learning_rate": 3.362940406116351e-06, "loss": 0.0079, "step": 32242 }, { "epoch": 5.91, "learning_rate": 3.362385091514019e-06, "loss": 0.006, "step": 32243 }, { "epoch": 5.91, "learning_rate": 3.361829813498376e-06, "loss": 0.0063, "step": 32244 }, { "epoch": 5.91, "learning_rate": 3.3612745720724827e-06, "loss": 0.0047, "step": 32245 }, { "epoch": 5.91, "learning_rate": 3.360719367239399e-06, "loss": 0.0076, "step": 32246 }, { "epoch": 5.91, "learning_rate": 3.360164199002186e-06, "loss": 0.0097, "step": 32247 }, { "epoch": 5.91, "learning_rate": 3.3596090673639006e-06, "loss": 0.0067, "step": 32248 }, { "epoch": 5.91, "learning_rate": 3.3590539723276083e-06, "loss": 0.0069, "step": 32249 }, { "epoch": 5.91, "learning_rate": 3.3584989138963695e-06, "loss": 0.0054, "step": 32250 }, { "epoch": 5.91, "learning_rate": 3.3579438920732375e-06, "loss": 0.0069, "step": 32251 }, { "epoch": 5.91, "learning_rate": 3.3573889068612707e-06, "loss": 0.0068, "step": 32252 }, { "epoch": 5.91, "learning_rate": 3.3568339582635356e-06, "loss": 0.0055, "step": 32253 }, { "epoch": 5.91, "learning_rate": 3.3562790462830864e-06, "loss": 0.0079, "step": 32254 }, { "epoch": 5.92, "learning_rate": 3.3557241709229836e-06, "loss": 0.0087, "step": 32255 }, { "epoch": 5.92, "learning_rate": 3.3551693321862877e-06, "loss": 0.0058, "step": 32256 }, { "epoch": 5.92, "learning_rate": 3.354614530076048e-06, "loss": 0.006, "step": 32257 }, { "epoch": 5.92, "learning_rate": 3.354059764595331e-06, "loss": 0.0068, "step": 32258 }, { "epoch": 5.92, "learning_rate": 3.353505035747193e-06, "loss": 0.0075, "step": 32259 }, { "epoch": 5.92, "learning_rate": 3.352950343534691e-06, "loss": 0.0066, "step": 32260 }, { "epoch": 5.92, "learning_rate": 3.3523956879608822e-06, "loss": 0.0078, "step": 32261 }, { "epoch": 5.92, "learning_rate": 3.3518410690288237e-06, "loss": 0.0074, "step": 32262 }, { "epoch": 5.92, "learning_rate": 3.351286486741574e-06, "loss": 0.0085, "step": 32263 }, { "epoch": 5.92, "learning_rate": 3.350731941102188e-06, "loss": 0.0059, "step": 32264 }, { "epoch": 5.92, "learning_rate": 3.350177432113724e-06, "loss": 0.0048, "step": 32265 }, { "epoch": 5.92, "learning_rate": 3.349622959779234e-06, "loss": 0.0056, "step": 32266 }, { "epoch": 5.92, "learning_rate": 3.349068524101785e-06, "loss": 0.0093, "step": 32267 }, { "epoch": 5.92, "learning_rate": 3.348514125084422e-06, "loss": 0.0058, "step": 32268 }, { "epoch": 5.92, "learning_rate": 3.3479597627302063e-06, "loss": 0.0074, "step": 32269 }, { "epoch": 5.92, "learning_rate": 3.3474054370421882e-06, "loss": 0.0067, "step": 32270 }, { "epoch": 5.92, "learning_rate": 3.34685114802343e-06, "loss": 0.0066, "step": 32271 }, { "epoch": 5.92, "learning_rate": 3.3462968956769846e-06, "loss": 0.0079, "step": 32272 }, { "epoch": 5.92, "learning_rate": 3.3457426800059066e-06, "loss": 0.0063, "step": 32273 }, { "epoch": 5.92, "learning_rate": 3.3451885010132513e-06, "loss": 0.0057, "step": 32274 }, { "epoch": 5.92, "learning_rate": 3.3446343587020726e-06, "loss": 0.0075, "step": 32275 }, { "epoch": 5.92, "learning_rate": 3.3440802530754246e-06, "loss": 0.0066, "step": 32276 }, { "epoch": 5.92, "learning_rate": 3.343526184136363e-06, "loss": 0.0066, "step": 32277 }, { "epoch": 5.92, "learning_rate": 3.342972151887941e-06, "loss": 0.0081, "step": 32278 }, { "epoch": 5.92, "learning_rate": 3.3424181563332115e-06, "loss": 0.0051, "step": 32279 }, { "epoch": 5.92, "learning_rate": 3.3418641974752297e-06, "loss": 0.0056, "step": 32280 }, { "epoch": 5.92, "learning_rate": 3.341310275317049e-06, "loss": 0.0065, "step": 32281 }, { "epoch": 5.92, "learning_rate": 3.3407563898617213e-06, "loss": 0.0065, "step": 32282 }, { "epoch": 5.92, "learning_rate": 3.3402025411122975e-06, "loss": 0.0071, "step": 32283 }, { "epoch": 5.92, "learning_rate": 3.339648729071836e-06, "loss": 0.0068, "step": 32284 }, { "epoch": 5.92, "learning_rate": 3.33909495374339e-06, "loss": 0.006, "step": 32285 }, { "epoch": 5.92, "learning_rate": 3.338541215130007e-06, "loss": 0.0077, "step": 32286 }, { "epoch": 5.92, "learning_rate": 3.3379875132347395e-06, "loss": 0.006, "step": 32287 }, { "epoch": 5.92, "learning_rate": 3.3374338480606374e-06, "loss": 0.0069, "step": 32288 }, { "epoch": 5.92, "learning_rate": 3.3368802196107596e-06, "loss": 0.0067, "step": 32289 }, { "epoch": 5.92, "learning_rate": 3.3363266278881546e-06, "loss": 0.0073, "step": 32290 }, { "epoch": 5.92, "learning_rate": 3.3357730728958725e-06, "loss": 0.0069, "step": 32291 }, { "epoch": 5.92, "learning_rate": 3.335219554636966e-06, "loss": 0.0064, "step": 32292 }, { "epoch": 5.92, "learning_rate": 3.3346660731144855e-06, "loss": 0.0068, "step": 32293 }, { "epoch": 5.92, "learning_rate": 3.3341126283314817e-06, "loss": 0.0075, "step": 32294 }, { "epoch": 5.92, "learning_rate": 3.333559220291005e-06, "loss": 0.0049, "step": 32295 }, { "epoch": 5.92, "learning_rate": 3.333005848996106e-06, "loss": 0.0063, "step": 32296 }, { "epoch": 5.92, "learning_rate": 3.3324525144498352e-06, "loss": 0.0056, "step": 32297 }, { "epoch": 5.92, "learning_rate": 3.331899216655242e-06, "loss": 0.0054, "step": 32298 }, { "epoch": 5.92, "learning_rate": 3.331345955615377e-06, "loss": 0.0069, "step": 32299 }, { "epoch": 5.92, "learning_rate": 3.33079273133329e-06, "loss": 0.0051, "step": 32300 }, { "epoch": 5.92, "learning_rate": 3.330239543812025e-06, "loss": 0.0075, "step": 32301 }, { "epoch": 5.92, "learning_rate": 3.3296863930546397e-06, "loss": 0.0078, "step": 32302 }, { "epoch": 5.92, "learning_rate": 3.329133279064183e-06, "loss": 0.0077, "step": 32303 }, { "epoch": 5.92, "learning_rate": 3.3285802018436963e-06, "loss": 0.0054, "step": 32304 }, { "epoch": 5.92, "learning_rate": 3.328027161396228e-06, "loss": 0.0059, "step": 32305 }, { "epoch": 5.92, "learning_rate": 3.3274741577248337e-06, "loss": 0.0077, "step": 32306 }, { "epoch": 5.92, "learning_rate": 3.3269211908325596e-06, "loss": 0.0057, "step": 32307 }, { "epoch": 5.92, "learning_rate": 3.32636826072245e-06, "loss": 0.0072, "step": 32308 }, { "epoch": 5.92, "learning_rate": 3.325815367397557e-06, "loss": 0.0059, "step": 32309 }, { "epoch": 5.93, "learning_rate": 3.325262510860925e-06, "loss": 0.0059, "step": 32310 }, { "epoch": 5.93, "learning_rate": 3.324709691115602e-06, "loss": 0.0064, "step": 32311 }, { "epoch": 5.93, "learning_rate": 3.3241569081646364e-06, "loss": 0.0063, "step": 32312 }, { "epoch": 5.93, "learning_rate": 3.3236041620110736e-06, "loss": 0.0068, "step": 32313 }, { "epoch": 5.93, "learning_rate": 3.3230514526579617e-06, "loss": 0.0067, "step": 32314 }, { "epoch": 5.93, "learning_rate": 3.3224987801083463e-06, "loss": 0.005, "step": 32315 }, { "epoch": 5.93, "learning_rate": 3.3219461443652745e-06, "loss": 0.0078, "step": 32316 }, { "epoch": 5.93, "learning_rate": 3.3213935454317914e-06, "loss": 0.0059, "step": 32317 }, { "epoch": 5.93, "learning_rate": 3.3208409833109447e-06, "loss": 0.0045, "step": 32318 }, { "epoch": 5.93, "learning_rate": 3.320288458005774e-06, "loss": 0.0068, "step": 32319 }, { "epoch": 5.93, "learning_rate": 3.319735969519334e-06, "loss": 0.0065, "step": 32320 }, { "epoch": 5.93, "learning_rate": 3.3191835178546695e-06, "loss": 0.0072, "step": 32321 }, { "epoch": 5.93, "learning_rate": 3.3186311030148177e-06, "loss": 0.0043, "step": 32322 }, { "epoch": 5.93, "learning_rate": 3.3180787250028247e-06, "loss": 0.0059, "step": 32323 }, { "epoch": 5.93, "learning_rate": 3.317526383821742e-06, "loss": 0.0068, "step": 32324 }, { "epoch": 5.93, "learning_rate": 3.3169740794746106e-06, "loss": 0.0073, "step": 32325 }, { "epoch": 5.93, "learning_rate": 3.3164218119644742e-06, "loss": 0.0078, "step": 32326 }, { "epoch": 5.93, "learning_rate": 3.3158695812943774e-06, "loss": 0.0071, "step": 32327 }, { "epoch": 5.93, "learning_rate": 3.315317387467365e-06, "loss": 0.0059, "step": 32328 }, { "epoch": 5.93, "learning_rate": 3.3147652304864784e-06, "loss": 0.0053, "step": 32329 }, { "epoch": 5.93, "learning_rate": 3.3142131103547627e-06, "loss": 0.0077, "step": 32330 }, { "epoch": 5.93, "learning_rate": 3.3136610270752613e-06, "loss": 0.0049, "step": 32331 }, { "epoch": 5.93, "learning_rate": 3.3131089806510173e-06, "loss": 0.0077, "step": 32332 }, { "epoch": 5.93, "learning_rate": 3.3125569710850723e-06, "loss": 0.0064, "step": 32333 }, { "epoch": 5.93, "learning_rate": 3.3120049983804715e-06, "loss": 0.0103, "step": 32334 }, { "epoch": 5.93, "learning_rate": 3.3114530625402543e-06, "loss": 0.007, "step": 32335 }, { "epoch": 5.93, "learning_rate": 3.3109011635674617e-06, "loss": 0.0048, "step": 32336 }, { "epoch": 5.93, "learning_rate": 3.310349301465142e-06, "loss": 0.0059, "step": 32337 }, { "epoch": 5.93, "learning_rate": 3.309797476236334e-06, "loss": 0.0053, "step": 32338 }, { "epoch": 5.93, "learning_rate": 3.3092456878840816e-06, "loss": 0.0065, "step": 32339 }, { "epoch": 5.93, "learning_rate": 3.308693936411421e-06, "loss": 0.0069, "step": 32340 }, { "epoch": 5.93, "learning_rate": 3.308142221821393e-06, "loss": 0.0047, "step": 32341 }, { "epoch": 5.93, "learning_rate": 3.307590544117044e-06, "loss": 0.0081, "step": 32342 }, { "epoch": 5.93, "learning_rate": 3.3070389033014138e-06, "loss": 0.0048, "step": 32343 }, { "epoch": 5.93, "learning_rate": 3.3064872993775423e-06, "loss": 0.0049, "step": 32344 }, { "epoch": 5.93, "learning_rate": 3.3059357323484685e-06, "loss": 0.0084, "step": 32345 }, { "epoch": 5.93, "learning_rate": 3.3053842022172345e-06, "loss": 0.006, "step": 32346 }, { "epoch": 5.93, "learning_rate": 3.304832708986879e-06, "loss": 0.0044, "step": 32347 }, { "epoch": 5.93, "learning_rate": 3.304281252660443e-06, "loss": 0.0072, "step": 32348 }, { "epoch": 5.93, "learning_rate": 3.303729833240963e-06, "loss": 0.0067, "step": 32349 }, { "epoch": 5.93, "learning_rate": 3.3031784507314866e-06, "loss": 0.0043, "step": 32350 }, { "epoch": 5.93, "learning_rate": 3.3026271051350445e-06, "loss": 0.0039, "step": 32351 }, { "epoch": 5.93, "learning_rate": 3.302075796454678e-06, "loss": 0.0063, "step": 32352 }, { "epoch": 5.93, "learning_rate": 3.301524524693427e-06, "loss": 0.007, "step": 32353 }, { "epoch": 5.93, "learning_rate": 3.3009732898543266e-06, "loss": 0.006, "step": 32354 }, { "epoch": 5.93, "learning_rate": 3.3004220919404217e-06, "loss": 0.0047, "step": 32355 }, { "epoch": 5.93, "learning_rate": 3.2998709309547473e-06, "loss": 0.0072, "step": 32356 }, { "epoch": 5.93, "learning_rate": 3.299319806900345e-06, "loss": 0.0062, "step": 32357 }, { "epoch": 5.93, "learning_rate": 3.298768719780242e-06, "loss": 0.0066, "step": 32358 }, { "epoch": 5.93, "learning_rate": 3.2982176695974853e-06, "loss": 0.0061, "step": 32359 }, { "epoch": 5.93, "learning_rate": 3.297666656355111e-06, "loss": 0.005, "step": 32360 }, { "epoch": 5.93, "learning_rate": 3.2971156800561543e-06, "loss": 0.0065, "step": 32361 }, { "epoch": 5.93, "learning_rate": 3.2965647407036526e-06, "loss": 0.0076, "step": 32362 }, { "epoch": 5.93, "learning_rate": 3.2960138383006445e-06, "loss": 0.0062, "step": 32363 }, { "epoch": 5.94, "learning_rate": 3.295462972850164e-06, "loss": 0.0062, "step": 32364 }, { "epoch": 5.94, "learning_rate": 3.294912144355249e-06, "loss": 0.0087, "step": 32365 }, { "epoch": 5.94, "learning_rate": 3.2943613528189358e-06, "loss": 0.0048, "step": 32366 }, { "epoch": 5.94, "learning_rate": 3.2938105982442557e-06, "loss": 0.0079, "step": 32367 }, { "epoch": 5.94, "learning_rate": 3.293259880634256e-06, "loss": 0.0059, "step": 32368 }, { "epoch": 5.94, "learning_rate": 3.2927091999919614e-06, "loss": 0.0061, "step": 32369 }, { "epoch": 5.94, "learning_rate": 3.292158556320412e-06, "loss": 0.0063, "step": 32370 }, { "epoch": 5.94, "learning_rate": 3.2916079496226407e-06, "loss": 0.0062, "step": 32371 }, { "epoch": 5.94, "learning_rate": 3.29105737990168e-06, "loss": 0.0081, "step": 32372 }, { "epoch": 5.94, "learning_rate": 3.2905068471605717e-06, "loss": 0.0074, "step": 32373 }, { "epoch": 5.94, "learning_rate": 3.289956351402347e-06, "loss": 0.0051, "step": 32374 }, { "epoch": 5.94, "learning_rate": 3.289405892630043e-06, "loss": 0.0045, "step": 32375 }, { "epoch": 5.94, "learning_rate": 3.288855470846685e-06, "loss": 0.0057, "step": 32376 }, { "epoch": 5.94, "learning_rate": 3.2883050860553154e-06, "loss": 0.0081, "step": 32377 }, { "epoch": 5.94, "learning_rate": 3.287754738258966e-06, "loss": 0.0055, "step": 32378 }, { "epoch": 5.94, "learning_rate": 3.2872044274606696e-06, "loss": 0.0046, "step": 32379 }, { "epoch": 5.94, "learning_rate": 3.2866541536634587e-06, "loss": 0.0048, "step": 32380 }, { "epoch": 5.94, "learning_rate": 3.286103916870369e-06, "loss": 0.0059, "step": 32381 }, { "epoch": 5.94, "learning_rate": 3.285553717084431e-06, "loss": 0.0069, "step": 32382 }, { "epoch": 5.94, "learning_rate": 3.2850035543086777e-06, "loss": 0.0048, "step": 32383 }, { "epoch": 5.94, "learning_rate": 3.284453428546143e-06, "loss": 0.0064, "step": 32384 }, { "epoch": 5.94, "learning_rate": 3.2839033397998544e-06, "loss": 0.0047, "step": 32385 }, { "epoch": 5.94, "learning_rate": 3.2833532880728547e-06, "loss": 0.0052, "step": 32386 }, { "epoch": 5.94, "learning_rate": 3.282803273368167e-06, "loss": 0.0056, "step": 32387 }, { "epoch": 5.94, "learning_rate": 3.2822532956888233e-06, "loss": 0.0065, "step": 32388 }, { "epoch": 5.94, "learning_rate": 3.281703355037854e-06, "loss": 0.0066, "step": 32389 }, { "epoch": 5.94, "learning_rate": 3.2811534514182965e-06, "loss": 0.0075, "step": 32390 }, { "epoch": 5.94, "learning_rate": 3.2806035848331787e-06, "loss": 0.0073, "step": 32391 }, { "epoch": 5.94, "learning_rate": 3.280053755285535e-06, "loss": 0.0047, "step": 32392 }, { "epoch": 5.94, "learning_rate": 3.279503962778389e-06, "loss": 0.0068, "step": 32393 }, { "epoch": 5.94, "learning_rate": 3.2789542073147704e-06, "loss": 0.0065, "step": 32394 }, { "epoch": 5.94, "learning_rate": 3.2784044888977186e-06, "loss": 0.0078, "step": 32395 }, { "epoch": 5.94, "learning_rate": 3.2778548075302575e-06, "loss": 0.0066, "step": 32396 }, { "epoch": 5.94, "learning_rate": 3.2773051632154184e-06, "loss": 0.0069, "step": 32397 }, { "epoch": 5.94, "learning_rate": 3.2767555559562303e-06, "loss": 0.007, "step": 32398 }, { "epoch": 5.94, "learning_rate": 3.2762059857557237e-06, "loss": 0.0064, "step": 32399 }, { "epoch": 5.94, "learning_rate": 3.275656452616928e-06, "loss": 0.007, "step": 32400 }, { "epoch": 5.94, "learning_rate": 3.2751069565428706e-06, "loss": 0.0078, "step": 32401 }, { "epoch": 5.94, "learning_rate": 3.274557497536578e-06, "loss": 0.0061, "step": 32402 }, { "epoch": 5.94, "learning_rate": 3.274008075601085e-06, "loss": 0.0049, "step": 32403 }, { "epoch": 5.94, "learning_rate": 3.2734586907394207e-06, "loss": 0.0051, "step": 32404 }, { "epoch": 5.94, "learning_rate": 3.2729093429546055e-06, "loss": 0.0064, "step": 32405 }, { "epoch": 5.94, "learning_rate": 3.2723600322496728e-06, "loss": 0.0051, "step": 32406 }, { "epoch": 5.94, "learning_rate": 3.271810758627646e-06, "loss": 0.0038, "step": 32407 }, { "epoch": 5.94, "learning_rate": 3.271261522091558e-06, "loss": 0.0058, "step": 32408 }, { "epoch": 5.94, "learning_rate": 3.270712322644435e-06, "loss": 0.0062, "step": 32409 }, { "epoch": 5.94, "learning_rate": 3.270163160289307e-06, "loss": 0.0077, "step": 32410 }, { "epoch": 5.94, "learning_rate": 3.2696140350291894e-06, "loss": 0.0064, "step": 32411 }, { "epoch": 5.94, "learning_rate": 3.269064946867121e-06, "loss": 0.0096, "step": 32412 }, { "epoch": 5.94, "learning_rate": 3.2685158958061245e-06, "loss": 0.0087, "step": 32413 }, { "epoch": 5.94, "learning_rate": 3.2679668818492263e-06, "loss": 0.0045, "step": 32414 }, { "epoch": 5.94, "learning_rate": 3.2674179049994514e-06, "loss": 0.0049, "step": 32415 }, { "epoch": 5.94, "learning_rate": 3.2668689652598273e-06, "loss": 0.0055, "step": 32416 }, { "epoch": 5.94, "learning_rate": 3.266320062633379e-06, "loss": 0.0048, "step": 32417 }, { "epoch": 5.94, "learning_rate": 3.2657711971231333e-06, "loss": 0.0068, "step": 32418 }, { "epoch": 5.95, "learning_rate": 3.265222368732114e-06, "loss": 0.0049, "step": 32419 }, { "epoch": 5.95, "learning_rate": 3.264673577463343e-06, "loss": 0.0069, "step": 32420 }, { "epoch": 5.95, "learning_rate": 3.264124823319855e-06, "loss": 0.0058, "step": 32421 }, { "epoch": 5.95, "learning_rate": 3.2635761063046666e-06, "loss": 0.0051, "step": 32422 }, { "epoch": 5.95, "learning_rate": 3.263027426420804e-06, "loss": 0.0078, "step": 32423 }, { "epoch": 5.95, "learning_rate": 3.262478783671292e-06, "loss": 0.0061, "step": 32424 }, { "epoch": 5.95, "learning_rate": 3.2619301780591526e-06, "loss": 0.0078, "step": 32425 }, { "epoch": 5.95, "learning_rate": 3.261381609587414e-06, "loss": 0.0077, "step": 32426 }, { "epoch": 5.95, "learning_rate": 3.260833078259098e-06, "loss": 0.0077, "step": 32427 }, { "epoch": 5.95, "learning_rate": 3.2602845840772324e-06, "loss": 0.0059, "step": 32428 }, { "epoch": 5.95, "learning_rate": 3.259736127044828e-06, "loss": 0.0049, "step": 32429 }, { "epoch": 5.95, "learning_rate": 3.2591877071649203e-06, "loss": 0.0067, "step": 32430 }, { "epoch": 5.95, "learning_rate": 3.258639324440527e-06, "loss": 0.006, "step": 32431 }, { "epoch": 5.95, "learning_rate": 3.258090978874672e-06, "loss": 0.0061, "step": 32432 }, { "epoch": 5.95, "learning_rate": 3.257542670470377e-06, "loss": 0.0051, "step": 32433 }, { "epoch": 5.95, "learning_rate": 3.256994399230666e-06, "loss": 0.0077, "step": 32434 }, { "epoch": 5.95, "learning_rate": 3.256446165158559e-06, "loss": 0.009, "step": 32435 }, { "epoch": 5.95, "learning_rate": 3.255897968257079e-06, "loss": 0.0058, "step": 32436 }, { "epoch": 5.95, "learning_rate": 3.2553498085292468e-06, "loss": 0.0051, "step": 32437 }, { "epoch": 5.95, "learning_rate": 3.254801685978082e-06, "loss": 0.0059, "step": 32438 }, { "epoch": 5.95, "learning_rate": 3.254253600606614e-06, "loss": 0.0068, "step": 32439 }, { "epoch": 5.95, "learning_rate": 3.2537055524178563e-06, "loss": 0.0065, "step": 32440 }, { "epoch": 5.95, "learning_rate": 3.253157541414831e-06, "loss": 0.0048, "step": 32441 }, { "epoch": 5.95, "learning_rate": 3.2526095676005553e-06, "loss": 0.0072, "step": 32442 }, { "epoch": 5.95, "learning_rate": 3.252061630978057e-06, "loss": 0.0085, "step": 32443 }, { "epoch": 5.95, "learning_rate": 3.2515137315503542e-06, "loss": 0.0082, "step": 32444 }, { "epoch": 5.95, "learning_rate": 3.250965869320465e-06, "loss": 0.0066, "step": 32445 }, { "epoch": 5.95, "learning_rate": 3.2504180442914137e-06, "loss": 0.0043, "step": 32446 }, { "epoch": 5.95, "learning_rate": 3.24987025646621e-06, "loss": 0.0055, "step": 32447 }, { "epoch": 5.95, "learning_rate": 3.2493225058478827e-06, "loss": 0.009, "step": 32448 }, { "epoch": 5.95, "learning_rate": 3.2487747924394476e-06, "loss": 0.0072, "step": 32449 }, { "epoch": 5.95, "learning_rate": 3.2482271162439237e-06, "loss": 0.0067, "step": 32450 }, { "epoch": 5.95, "learning_rate": 3.2476794772643304e-06, "loss": 0.0076, "step": 32451 }, { "epoch": 5.95, "learning_rate": 3.2471318755036864e-06, "loss": 0.007, "step": 32452 }, { "epoch": 5.95, "learning_rate": 3.24658431096501e-06, "loss": 0.0062, "step": 32453 }, { "epoch": 5.95, "learning_rate": 3.2460367836513184e-06, "loss": 0.0066, "step": 32454 }, { "epoch": 5.95, "learning_rate": 3.2454892935656267e-06, "loss": 0.0059, "step": 32455 }, { "epoch": 5.95, "learning_rate": 3.2449418407109603e-06, "loss": 0.0051, "step": 32456 }, { "epoch": 5.95, "learning_rate": 3.244394425090336e-06, "loss": 0.0064, "step": 32457 }, { "epoch": 5.95, "learning_rate": 3.243847046706764e-06, "loss": 0.0075, "step": 32458 }, { "epoch": 5.95, "learning_rate": 3.2432997055632655e-06, "loss": 0.0071, "step": 32459 }, { "epoch": 5.95, "learning_rate": 3.2427524016628542e-06, "loss": 0.0064, "step": 32460 }, { "epoch": 5.95, "learning_rate": 3.2422051350085527e-06, "loss": 0.0052, "step": 32461 }, { "epoch": 5.95, "learning_rate": 3.2416579056033758e-06, "loss": 0.0059, "step": 32462 }, { "epoch": 5.95, "learning_rate": 3.241110713450338e-06, "loss": 0.0051, "step": 32463 }, { "epoch": 5.95, "learning_rate": 3.2405635585524566e-06, "loss": 0.0067, "step": 32464 }, { "epoch": 5.95, "learning_rate": 3.2400164409127477e-06, "loss": 0.0053, "step": 32465 }, { "epoch": 5.95, "learning_rate": 3.2394693605342265e-06, "loss": 0.0069, "step": 32466 }, { "epoch": 5.95, "learning_rate": 3.238922317419908e-06, "loss": 0.0078, "step": 32467 }, { "epoch": 5.95, "learning_rate": 3.238375311572808e-06, "loss": 0.0061, "step": 32468 }, { "epoch": 5.95, "learning_rate": 3.2378283429959424e-06, "loss": 0.0095, "step": 32469 }, { "epoch": 5.95, "learning_rate": 3.2372814116923244e-06, "loss": 0.0095, "step": 32470 }, { "epoch": 5.95, "learning_rate": 3.2367345176649713e-06, "loss": 0.0071, "step": 32471 }, { "epoch": 5.95, "learning_rate": 3.2361876609168943e-06, "loss": 0.0059, "step": 32472 }, { "epoch": 5.96, "learning_rate": 3.2356408414511063e-06, "loss": 0.0084, "step": 32473 }, { "epoch": 5.96, "learning_rate": 3.235094059270628e-06, "loss": 0.0059, "step": 32474 }, { "epoch": 5.96, "learning_rate": 3.2345473143784734e-06, "loss": 0.0051, "step": 32475 }, { "epoch": 5.96, "learning_rate": 3.2340006067776497e-06, "loss": 0.0074, "step": 32476 }, { "epoch": 5.96, "learning_rate": 3.233453936471169e-06, "loss": 0.006, "step": 32477 }, { "epoch": 5.96, "learning_rate": 3.2329073034620517e-06, "loss": 0.0059, "step": 32478 }, { "epoch": 5.96, "learning_rate": 3.2323607077533092e-06, "loss": 0.0039, "step": 32479 }, { "epoch": 5.96, "learning_rate": 3.2318141493479515e-06, "loss": 0.0071, "step": 32480 }, { "epoch": 5.96, "learning_rate": 3.231267628248994e-06, "loss": 0.0061, "step": 32481 }, { "epoch": 5.96, "learning_rate": 3.230721144459448e-06, "loss": 0.0067, "step": 32482 }, { "epoch": 5.96, "learning_rate": 3.2301746979823245e-06, "loss": 0.0061, "step": 32483 }, { "epoch": 5.96, "learning_rate": 3.229628288820638e-06, "loss": 0.0051, "step": 32484 }, { "epoch": 5.96, "learning_rate": 3.229081916977397e-06, "loss": 0.0083, "step": 32485 }, { "epoch": 5.96, "learning_rate": 3.2285355824556163e-06, "loss": 0.0078, "step": 32486 }, { "epoch": 5.96, "learning_rate": 3.227989285258306e-06, "loss": 0.0062, "step": 32487 }, { "epoch": 5.96, "learning_rate": 3.227443025388478e-06, "loss": 0.007, "step": 32488 }, { "epoch": 5.96, "learning_rate": 3.226896802849142e-06, "loss": 0.0079, "step": 32489 }, { "epoch": 5.96, "learning_rate": 3.22635061764331e-06, "loss": 0.005, "step": 32490 }, { "epoch": 5.96, "learning_rate": 3.2258044697739887e-06, "loss": 0.0063, "step": 32491 }, { "epoch": 5.96, "learning_rate": 3.225258359244194e-06, "loss": 0.005, "step": 32492 }, { "epoch": 5.96, "learning_rate": 3.224712286056939e-06, "loss": 0.0067, "step": 32493 }, { "epoch": 5.96, "learning_rate": 3.2241662502152236e-06, "loss": 0.007, "step": 32494 }, { "epoch": 5.96, "learning_rate": 3.2236202517220607e-06, "loss": 0.0063, "step": 32495 }, { "epoch": 5.96, "learning_rate": 3.2230742905804647e-06, "loss": 0.0071, "step": 32496 }, { "epoch": 5.96, "learning_rate": 3.2225283667934414e-06, "loss": 0.0061, "step": 32497 }, { "epoch": 5.96, "learning_rate": 3.2219824803640008e-06, "loss": 0.0064, "step": 32498 }, { "epoch": 5.96, "learning_rate": 3.221436631295155e-06, "loss": 0.0062, "step": 32499 }, { "epoch": 5.96, "learning_rate": 3.2208908195899036e-06, "loss": 0.0075, "step": 32500 }, { "epoch": 5.96, "learning_rate": 3.220345045251263e-06, "loss": 0.0049, "step": 32501 }, { "epoch": 5.96, "learning_rate": 3.2197993082822387e-06, "loss": 0.0062, "step": 32502 }, { "epoch": 5.96, "learning_rate": 3.2192536086858406e-06, "loss": 0.0065, "step": 32503 }, { "epoch": 5.96, "learning_rate": 3.218707946465075e-06, "loss": 0.0077, "step": 32504 }, { "epoch": 5.96, "learning_rate": 3.2181623216229508e-06, "loss": 0.0068, "step": 32505 }, { "epoch": 5.96, "learning_rate": 3.2176167341624743e-06, "loss": 0.0054, "step": 32506 }, { "epoch": 5.96, "learning_rate": 3.217071184086653e-06, "loss": 0.0063, "step": 32507 }, { "epoch": 5.96, "learning_rate": 3.2165256713984906e-06, "loss": 0.005, "step": 32508 }, { "epoch": 5.96, "learning_rate": 3.215980196101002e-06, "loss": 0.0049, "step": 32509 }, { "epoch": 5.96, "learning_rate": 3.215434758197189e-06, "loss": 0.0046, "step": 32510 }, { "epoch": 5.96, "learning_rate": 3.214889357690061e-06, "loss": 0.0075, "step": 32511 }, { "epoch": 5.96, "learning_rate": 3.2143439945826193e-06, "loss": 0.0083, "step": 32512 }, { "epoch": 5.96, "learning_rate": 3.21379866887787e-06, "loss": 0.0064, "step": 32513 }, { "epoch": 5.96, "learning_rate": 3.2132533805788234e-06, "loss": 0.0067, "step": 32514 }, { "epoch": 5.96, "learning_rate": 3.212708129688483e-06, "loss": 0.0101, "step": 32515 }, { "epoch": 5.96, "learning_rate": 3.2121629162098555e-06, "loss": 0.0063, "step": 32516 }, { "epoch": 5.96, "learning_rate": 3.211617740145945e-06, "loss": 0.0069, "step": 32517 }, { "epoch": 5.96, "learning_rate": 3.2110726014997563e-06, "loss": 0.0039, "step": 32518 }, { "epoch": 5.96, "learning_rate": 3.210527500274294e-06, "loss": 0.0044, "step": 32519 }, { "epoch": 5.96, "learning_rate": 3.209982436472563e-06, "loss": 0.0057, "step": 32520 }, { "epoch": 5.96, "learning_rate": 3.209437410097569e-06, "loss": 0.0067, "step": 32521 }, { "epoch": 5.96, "learning_rate": 3.208892421152314e-06, "loss": 0.0069, "step": 32522 }, { "epoch": 5.96, "learning_rate": 3.2083474696398034e-06, "loss": 0.005, "step": 32523 }, { "epoch": 5.96, "learning_rate": 3.207802555563041e-06, "loss": 0.0035, "step": 32524 }, { "epoch": 5.96, "learning_rate": 3.2072576789250296e-06, "loss": 0.0053, "step": 32525 }, { "epoch": 5.96, "learning_rate": 3.2067128397287696e-06, "loss": 0.0062, "step": 32526 }, { "epoch": 5.96, "learning_rate": 3.206168037977271e-06, "loss": 0.0071, "step": 32527 }, { "epoch": 5.97, "learning_rate": 3.205623273673534e-06, "loss": 0.0065, "step": 32528 }, { "epoch": 5.97, "learning_rate": 3.2050785468205635e-06, "loss": 0.0054, "step": 32529 }, { "epoch": 5.97, "learning_rate": 3.2045338574213526e-06, "loss": 0.0094, "step": 32530 }, { "epoch": 5.97, "learning_rate": 3.2039892054789136e-06, "loss": 0.0065, "step": 32531 }, { "epoch": 5.97, "learning_rate": 3.2034445909962453e-06, "loss": 0.0074, "step": 32532 }, { "epoch": 5.97, "learning_rate": 3.20290001397635e-06, "loss": 0.0078, "step": 32533 }, { "epoch": 5.97, "learning_rate": 3.2023554744222286e-06, "loss": 0.0064, "step": 32534 }, { "epoch": 5.97, "learning_rate": 3.2018109723368828e-06, "loss": 0.0068, "step": 32535 }, { "epoch": 5.97, "learning_rate": 3.2012665077233152e-06, "loss": 0.0047, "step": 32536 }, { "epoch": 5.97, "learning_rate": 3.2007220805845253e-06, "loss": 0.007, "step": 32537 }, { "epoch": 5.97, "learning_rate": 3.2001776909235137e-06, "loss": 0.0093, "step": 32538 }, { "epoch": 5.97, "learning_rate": 3.1996333387432797e-06, "loss": 0.0069, "step": 32539 }, { "epoch": 5.97, "learning_rate": 3.199089024046833e-06, "loss": 0.0048, "step": 32540 }, { "epoch": 5.97, "learning_rate": 3.198544746837163e-06, "loss": 0.0066, "step": 32541 }, { "epoch": 5.97, "learning_rate": 3.1980005071172738e-06, "loss": 0.0075, "step": 32542 }, { "epoch": 5.97, "learning_rate": 3.1974563048901653e-06, "loss": 0.0093, "step": 32543 }, { "epoch": 5.97, "learning_rate": 3.1969121401588333e-06, "loss": 0.0064, "step": 32544 }, { "epoch": 5.97, "learning_rate": 3.1963680129262843e-06, "loss": 0.0087, "step": 32545 }, { "epoch": 5.97, "learning_rate": 3.195823923195518e-06, "loss": 0.0065, "step": 32546 }, { "epoch": 5.97, "learning_rate": 3.1952798709695264e-06, "loss": 0.0054, "step": 32547 }, { "epoch": 5.97, "learning_rate": 3.1947358562513076e-06, "loss": 0.0056, "step": 32548 }, { "epoch": 5.97, "learning_rate": 3.1941918790438675e-06, "loss": 0.0076, "step": 32549 }, { "epoch": 5.97, "learning_rate": 3.1936479393502017e-06, "loss": 0.0042, "step": 32550 }, { "epoch": 5.97, "learning_rate": 3.193104037173308e-06, "loss": 0.0073, "step": 32551 }, { "epoch": 5.97, "learning_rate": 3.1925601725161837e-06, "loss": 0.0058, "step": 32552 }, { "epoch": 5.97, "learning_rate": 3.1920163453818276e-06, "loss": 0.005, "step": 32553 }, { "epoch": 5.97, "learning_rate": 3.1914725557732372e-06, "loss": 0.0045, "step": 32554 }, { "epoch": 5.97, "learning_rate": 3.1909288036934093e-06, "loss": 0.0071, "step": 32555 }, { "epoch": 5.97, "learning_rate": 3.1903850891453415e-06, "loss": 0.0076, "step": 32556 }, { "epoch": 5.97, "learning_rate": 3.1898414121320277e-06, "loss": 0.0058, "step": 32557 }, { "epoch": 5.97, "learning_rate": 3.1892977726564735e-06, "loss": 0.0051, "step": 32558 }, { "epoch": 5.97, "learning_rate": 3.1887541707216664e-06, "loss": 0.0069, "step": 32559 }, { "epoch": 5.97, "learning_rate": 3.188210606330605e-06, "loss": 0.0073, "step": 32560 }, { "epoch": 5.97, "learning_rate": 3.1876670794862843e-06, "loss": 0.0045, "step": 32561 }, { "epoch": 5.97, "learning_rate": 3.1871235901917053e-06, "loss": 0.0068, "step": 32562 }, { "epoch": 5.97, "learning_rate": 3.1865801384498607e-06, "loss": 0.0059, "step": 32563 }, { "epoch": 5.97, "learning_rate": 3.186036724263748e-06, "loss": 0.0057, "step": 32564 }, { "epoch": 5.97, "learning_rate": 3.1854933476363593e-06, "loss": 0.0065, "step": 32565 }, { "epoch": 5.97, "learning_rate": 3.1849500085706862e-06, "loss": 0.0046, "step": 32566 }, { "epoch": 5.97, "learning_rate": 3.1844067070697314e-06, "loss": 0.0051, "step": 32567 }, { "epoch": 5.97, "learning_rate": 3.1838634431364877e-06, "loss": 0.0086, "step": 32568 }, { "epoch": 5.97, "learning_rate": 3.1833202167739473e-06, "loss": 0.0083, "step": 32569 }, { "epoch": 5.97, "learning_rate": 3.1827770279851066e-06, "loss": 0.0062, "step": 32570 }, { "epoch": 5.97, "learning_rate": 3.1822338767729577e-06, "loss": 0.0062, "step": 32571 }, { "epoch": 5.97, "learning_rate": 3.1816907631404957e-06, "loss": 0.0066, "step": 32572 }, { "epoch": 5.97, "learning_rate": 3.1811476870907145e-06, "loss": 0.0081, "step": 32573 }, { "epoch": 5.97, "learning_rate": 3.180604648626606e-06, "loss": 0.0049, "step": 32574 }, { "epoch": 5.97, "learning_rate": 3.1800616477511625e-06, "loss": 0.0075, "step": 32575 }, { "epoch": 5.97, "learning_rate": 3.179518684467384e-06, "loss": 0.0081, "step": 32576 }, { "epoch": 5.97, "learning_rate": 3.1789757587782568e-06, "loss": 0.0067, "step": 32577 }, { "epoch": 5.97, "learning_rate": 3.178432870686774e-06, "loss": 0.006, "step": 32578 }, { "epoch": 5.97, "learning_rate": 3.177890020195925e-06, "loss": 0.005, "step": 32579 }, { "epoch": 5.97, "learning_rate": 3.1773472073087108e-06, "loss": 0.0072, "step": 32580 }, { "epoch": 5.97, "learning_rate": 3.176804432028118e-06, "loss": 0.0072, "step": 32581 }, { "epoch": 5.98, "learning_rate": 3.1762616943571423e-06, "loss": 0.0076, "step": 32582 }, { "epoch": 5.98, "learning_rate": 3.175718994298764e-06, "loss": 0.0043, "step": 32583 }, { "epoch": 5.98, "learning_rate": 3.175176331855988e-06, "loss": 0.0048, "step": 32584 }, { "epoch": 5.98, "learning_rate": 3.174633707031799e-06, "loss": 0.0074, "step": 32585 }, { "epoch": 5.98, "learning_rate": 3.1740911198291892e-06, "loss": 0.0077, "step": 32586 }, { "epoch": 5.98, "learning_rate": 3.1735485702511482e-06, "loss": 0.0076, "step": 32587 }, { "epoch": 5.98, "learning_rate": 3.1730060583006684e-06, "loss": 0.0077, "step": 32588 }, { "epoch": 5.98, "learning_rate": 3.172463583980738e-06, "loss": 0.0084, "step": 32589 }, { "epoch": 5.98, "learning_rate": 3.17192114729435e-06, "loss": 0.0074, "step": 32590 }, { "epoch": 5.98, "learning_rate": 3.1713787482444912e-06, "loss": 0.0063, "step": 32591 }, { "epoch": 5.98, "learning_rate": 3.1708363868341494e-06, "loss": 0.0043, "step": 32592 }, { "epoch": 5.98, "learning_rate": 3.170294063066325e-06, "loss": 0.0076, "step": 32593 }, { "epoch": 5.98, "learning_rate": 3.1697517769439955e-06, "loss": 0.0058, "step": 32594 }, { "epoch": 5.98, "learning_rate": 3.1692095284701553e-06, "loss": 0.008, "step": 32595 }, { "epoch": 5.98, "learning_rate": 3.168667317647791e-06, "loss": 0.0054, "step": 32596 }, { "epoch": 5.98, "learning_rate": 3.16812514447989e-06, "loss": 0.0075, "step": 32597 }, { "epoch": 5.98, "learning_rate": 3.167583008969446e-06, "loss": 0.0062, "step": 32598 }, { "epoch": 5.98, "learning_rate": 3.1670409111194454e-06, "loss": 0.0062, "step": 32599 }, { "epoch": 5.98, "learning_rate": 3.1664988509328776e-06, "loss": 0.006, "step": 32600 }, { "epoch": 5.98, "learning_rate": 3.165956828412722e-06, "loss": 0.0063, "step": 32601 }, { "epoch": 5.98, "learning_rate": 3.165414843561976e-06, "loss": 0.0065, "step": 32602 }, { "epoch": 5.98, "learning_rate": 3.164872896383624e-06, "loss": 0.0047, "step": 32603 }, { "epoch": 5.98, "learning_rate": 3.1643309868806516e-06, "loss": 0.0057, "step": 32604 }, { "epoch": 5.98, "learning_rate": 3.1637891150560474e-06, "loss": 0.0073, "step": 32605 }, { "epoch": 5.98, "learning_rate": 3.163247280912798e-06, "loss": 0.007, "step": 32606 }, { "epoch": 5.98, "learning_rate": 3.1627054844538896e-06, "loss": 0.0069, "step": 32607 }, { "epoch": 5.98, "learning_rate": 3.162163725682309e-06, "loss": 0.0068, "step": 32608 }, { "epoch": 5.98, "learning_rate": 3.1616220046010416e-06, "loss": 0.0081, "step": 32609 }, { "epoch": 5.98, "learning_rate": 3.1610803212130705e-06, "loss": 0.008, "step": 32610 }, { "epoch": 5.98, "learning_rate": 3.160538675521392e-06, "loss": 0.0046, "step": 32611 }, { "epoch": 5.98, "learning_rate": 3.1599970675289817e-06, "loss": 0.0051, "step": 32612 }, { "epoch": 5.98, "learning_rate": 3.159455497238827e-06, "loss": 0.0077, "step": 32613 }, { "epoch": 5.98, "learning_rate": 3.15891396465391e-06, "loss": 0.0058, "step": 32614 }, { "epoch": 5.98, "learning_rate": 3.1583724697772244e-06, "loss": 0.0071, "step": 32615 }, { "epoch": 5.98, "learning_rate": 3.1578310126117483e-06, "loss": 0.0069, "step": 32616 }, { "epoch": 5.98, "learning_rate": 3.1572895931604696e-06, "loss": 0.0059, "step": 32617 }, { "epoch": 5.98, "learning_rate": 3.1567482114263725e-06, "loss": 0.0052, "step": 32618 }, { "epoch": 5.98, "learning_rate": 3.1562068674124347e-06, "loss": 0.0054, "step": 32619 }, { "epoch": 5.98, "learning_rate": 3.155665561121648e-06, "loss": 0.0079, "step": 32620 }, { "epoch": 5.98, "learning_rate": 3.1551242925569925e-06, "loss": 0.0062, "step": 32621 }, { "epoch": 5.98, "learning_rate": 3.1545830617214524e-06, "loss": 0.0062, "step": 32622 }, { "epoch": 5.98, "learning_rate": 3.154041868618011e-06, "loss": 0.0052, "step": 32623 }, { "epoch": 5.98, "learning_rate": 3.153500713249651e-06, "loss": 0.0044, "step": 32624 }, { "epoch": 5.98, "learning_rate": 3.152959595619355e-06, "loss": 0.0054, "step": 32625 }, { "epoch": 5.98, "learning_rate": 3.152418515730107e-06, "loss": 0.0056, "step": 32626 }, { "epoch": 5.98, "learning_rate": 3.151877473584888e-06, "loss": 0.007, "step": 32627 }, { "epoch": 5.98, "learning_rate": 3.151336469186678e-06, "loss": 0.0046, "step": 32628 }, { "epoch": 5.98, "learning_rate": 3.1507955025384673e-06, "loss": 0.0057, "step": 32629 }, { "epoch": 5.98, "learning_rate": 3.15025457364323e-06, "loss": 0.0062, "step": 32630 }, { "epoch": 5.98, "learning_rate": 3.14971368250395e-06, "loss": 0.0066, "step": 32631 }, { "epoch": 5.98, "learning_rate": 3.1491728291236044e-06, "loss": 0.0069, "step": 32632 }, { "epoch": 5.98, "learning_rate": 3.148632013505183e-06, "loss": 0.0062, "step": 32633 }, { "epoch": 5.98, "learning_rate": 3.1480912356516623e-06, "loss": 0.0055, "step": 32634 }, { "epoch": 5.98, "learning_rate": 3.147550495566023e-06, "loss": 0.0079, "step": 32635 }, { "epoch": 5.98, "learning_rate": 3.1470097932512456e-06, "loss": 0.0059, "step": 32636 }, { "epoch": 5.99, "learning_rate": 3.146469128710311e-06, "loss": 0.0049, "step": 32637 }, { "epoch": 5.99, "learning_rate": 3.1459285019461995e-06, "loss": 0.0078, "step": 32638 }, { "epoch": 5.99, "learning_rate": 3.1453879129618904e-06, "loss": 0.0074, "step": 32639 }, { "epoch": 5.99, "learning_rate": 3.144847361760364e-06, "loss": 0.0065, "step": 32640 }, { "epoch": 5.99, "learning_rate": 3.1443068483445992e-06, "loss": 0.0041, "step": 32641 }, { "epoch": 5.99, "learning_rate": 3.143766372717576e-06, "loss": 0.0053, "step": 32642 }, { "epoch": 5.99, "learning_rate": 3.1432259348822726e-06, "loss": 0.0055, "step": 32643 }, { "epoch": 5.99, "learning_rate": 3.1426855348416685e-06, "loss": 0.0073, "step": 32644 }, { "epoch": 5.99, "learning_rate": 3.1421451725987396e-06, "loss": 0.0055, "step": 32645 }, { "epoch": 5.99, "learning_rate": 3.1416048481564698e-06, "loss": 0.006, "step": 32646 }, { "epoch": 5.99, "learning_rate": 3.141064561517839e-06, "loss": 0.0059, "step": 32647 }, { "epoch": 5.99, "learning_rate": 3.1405243126858175e-06, "loss": 0.0067, "step": 32648 }, { "epoch": 5.99, "learning_rate": 3.139984101663386e-06, "loss": 0.0053, "step": 32649 }, { "epoch": 5.99, "learning_rate": 3.1394439284535206e-06, "loss": 0.0056, "step": 32650 }, { "epoch": 5.99, "learning_rate": 3.1389037930592028e-06, "loss": 0.0091, "step": 32651 }, { "epoch": 5.99, "learning_rate": 3.1383636954834098e-06, "loss": 0.0052, "step": 32652 }, { "epoch": 5.99, "learning_rate": 3.137823635729118e-06, "loss": 0.0071, "step": 32653 }, { "epoch": 5.99, "learning_rate": 3.1372836137992977e-06, "loss": 0.0051, "step": 32654 }, { "epoch": 5.99, "learning_rate": 3.1367436296969346e-06, "loss": 0.0077, "step": 32655 }, { "epoch": 5.99, "learning_rate": 3.136203683425001e-06, "loss": 0.0082, "step": 32656 }, { "epoch": 5.99, "learning_rate": 3.1356637749864736e-06, "loss": 0.0054, "step": 32657 }, { "epoch": 5.99, "learning_rate": 3.1351239043843287e-06, "loss": 0.0043, "step": 32658 }, { "epoch": 5.99, "learning_rate": 3.1345840716215404e-06, "loss": 0.0056, "step": 32659 }, { "epoch": 5.99, "learning_rate": 3.134044276701087e-06, "loss": 0.006, "step": 32660 }, { "epoch": 5.99, "learning_rate": 3.133504519625942e-06, "loss": 0.0057, "step": 32661 }, { "epoch": 5.99, "learning_rate": 3.1329648003990797e-06, "loss": 0.0082, "step": 32662 }, { "epoch": 5.99, "learning_rate": 3.1324251190234743e-06, "loss": 0.0075, "step": 32663 }, { "epoch": 5.99, "learning_rate": 3.1318854755021054e-06, "loss": 0.0051, "step": 32664 }, { "epoch": 5.99, "learning_rate": 3.1313458698379483e-06, "loss": 0.0059, "step": 32665 }, { "epoch": 5.99, "learning_rate": 3.1308063020339698e-06, "loss": 0.0054, "step": 32666 }, { "epoch": 5.99, "learning_rate": 3.1302667720931445e-06, "loss": 0.0062, "step": 32667 }, { "epoch": 5.99, "learning_rate": 3.129727280018454e-06, "loss": 0.006, "step": 32668 }, { "epoch": 5.99, "learning_rate": 3.1291878258128673e-06, "loss": 0.0058, "step": 32669 }, { "epoch": 5.99, "learning_rate": 3.1286484094793577e-06, "loss": 0.0056, "step": 32670 }, { "epoch": 5.99, "learning_rate": 3.1281090310209027e-06, "loss": 0.0065, "step": 32671 }, { "epoch": 5.99, "learning_rate": 3.127569690440465e-06, "loss": 0.0061, "step": 32672 }, { "epoch": 5.99, "learning_rate": 3.127030387741028e-06, "loss": 0.005, "step": 32673 }, { "epoch": 5.99, "learning_rate": 3.12649112292556e-06, "loss": 0.0063, "step": 32674 }, { "epoch": 5.99, "learning_rate": 3.1259518959970335e-06, "loss": 0.005, "step": 32675 }, { "epoch": 5.99, "learning_rate": 3.1254127069584228e-06, "loss": 0.0057, "step": 32676 }, { "epoch": 5.99, "learning_rate": 3.124873555812696e-06, "loss": 0.0055, "step": 32677 }, { "epoch": 5.99, "learning_rate": 3.1243344425628298e-06, "loss": 0.0064, "step": 32678 }, { "epoch": 5.99, "learning_rate": 3.1237953672117916e-06, "loss": 0.0075, "step": 32679 }, { "epoch": 5.99, "learning_rate": 3.123256329762555e-06, "loss": 0.0059, "step": 32680 }, { "epoch": 5.99, "learning_rate": 3.122717330218087e-06, "loss": 0.008, "step": 32681 }, { "epoch": 5.99, "learning_rate": 3.1221783685813655e-06, "loss": 0.0068, "step": 32682 }, { "epoch": 5.99, "learning_rate": 3.1216394448553612e-06, "loss": 0.0055, "step": 32683 }, { "epoch": 5.99, "learning_rate": 3.121100559043038e-06, "loss": 0.0056, "step": 32684 }, { "epoch": 5.99, "learning_rate": 3.1205617111473676e-06, "loss": 0.0064, "step": 32685 }, { "epoch": 5.99, "learning_rate": 3.1200229011713236e-06, "loss": 0.0052, "step": 32686 }, { "epoch": 5.99, "learning_rate": 3.1194841291178755e-06, "loss": 0.0061, "step": 32687 }, { "epoch": 5.99, "learning_rate": 3.1189453949899927e-06, "loss": 0.0063, "step": 32688 }, { "epoch": 5.99, "learning_rate": 3.1184066987906434e-06, "loss": 0.0069, "step": 32689 }, { "epoch": 5.99, "learning_rate": 3.117868040522798e-06, "loss": 0.0057, "step": 32690 }, { "epoch": 6.0, "learning_rate": 3.1173294201894243e-06, "loss": 0.0057, "step": 32691 }, { "epoch": 6.0, "learning_rate": 3.116790837793493e-06, "loss": 0.0059, "step": 32692 }, { "epoch": 6.0, "learning_rate": 3.1162522933379714e-06, "loss": 0.0066, "step": 32693 }, { "epoch": 6.0, "learning_rate": 3.1157137868258282e-06, "loss": 0.0061, "step": 32694 }, { "epoch": 6.0, "learning_rate": 3.115175318260032e-06, "loss": 0.0092, "step": 32695 }, { "epoch": 6.0, "learning_rate": 3.1146368876435505e-06, "loss": 0.0048, "step": 32696 }, { "epoch": 6.0, "learning_rate": 3.114098494979353e-06, "loss": 0.0059, "step": 32697 }, { "epoch": 6.0, "learning_rate": 3.1135601402704017e-06, "loss": 0.0078, "step": 32698 }, { "epoch": 6.0, "learning_rate": 3.113021823519672e-06, "loss": 0.0049, "step": 32699 }, { "epoch": 6.0, "learning_rate": 3.1124835447301304e-06, "loss": 0.007, "step": 32700 }, { "epoch": 6.0, "learning_rate": 3.1119453039047374e-06, "loss": 0.005, "step": 32701 }, { "epoch": 6.0, "learning_rate": 3.111407101046463e-06, "loss": 0.0061, "step": 32702 }, { "epoch": 6.0, "learning_rate": 3.110868936158271e-06, "loss": 0.0055, "step": 32703 }, { "epoch": 6.0, "learning_rate": 3.110330809243134e-06, "loss": 0.0076, "step": 32704 }, { "epoch": 6.0, "learning_rate": 3.1097927203040155e-06, "loss": 0.0072, "step": 32705 }, { "epoch": 6.0, "learning_rate": 3.1092546693438808e-06, "loss": 0.006, "step": 32706 }, { "epoch": 6.0, "learning_rate": 3.1087166563656957e-06, "loss": 0.0062, "step": 32707 }, { "epoch": 6.0, "learning_rate": 3.1081786813724258e-06, "loss": 0.0077, "step": 32708 }, { "epoch": 6.0, "learning_rate": 3.1076407443670364e-06, "loss": 0.0064, "step": 32709 }, { "epoch": 6.0, "learning_rate": 3.107102845352493e-06, "loss": 0.0068, "step": 32710 }, { "epoch": 6.0, "learning_rate": 3.1065649843317558e-06, "loss": 0.0089, "step": 32711 }, { "epoch": 6.0, "learning_rate": 3.1060271613078018e-06, "loss": 0.0065, "step": 32712 }, { "epoch": 6.0, "learning_rate": 3.1054893762835834e-06, "loss": 0.0054, "step": 32713 }, { "epoch": 6.0, "learning_rate": 3.1049516292620696e-06, "loss": 0.0076, "step": 32714 }, { "epoch": 6.0, "learning_rate": 3.1044139202462244e-06, "loss": 0.0042, "step": 32715 }, { "epoch": 6.0, "learning_rate": 3.1038762492390075e-06, "loss": 0.0049, "step": 32716 }, { "epoch": 6.0, "learning_rate": 3.103338616243389e-06, "loss": 0.0053, "step": 32717 }, { "epoch": 6.0, "learning_rate": 3.1028010212623337e-06, "loss": 0.0062, "step": 32718 }, { "epoch": 6.0, "learning_rate": 3.102263464298797e-06, "loss": 0.0025, "step": 32719 }, { "epoch": 6.0, "learning_rate": 3.1017259453557423e-06, "loss": 0.0017, "step": 32720 }, { "epoch": 6.0, "learning_rate": 3.101188464436139e-06, "loss": 0.0021, "step": 32721 }, { "epoch": 6.0, "learning_rate": 3.1006510215429476e-06, "loss": 0.003, "step": 32722 }, { "epoch": 6.0, "learning_rate": 3.1001136166791277e-06, "loss": 0.0039, "step": 32723 }, { "epoch": 6.0, "learning_rate": 3.0995762498476446e-06, "loss": 0.0029, "step": 32724 }, { "epoch": 6.0, "learning_rate": 3.099038921051458e-06, "loss": 0.0023, "step": 32725 }, { "epoch": 6.0, "learning_rate": 3.098501630293531e-06, "loss": 0.0023, "step": 32726 }, { "epoch": 6.0, "learning_rate": 3.0979643775768253e-06, "loss": 0.0025, "step": 32727 }, { "epoch": 6.0, "learning_rate": 3.0974271629043007e-06, "loss": 0.0027, "step": 32728 }, { "epoch": 6.0, "learning_rate": 3.096889986278916e-06, "loss": 0.0031, "step": 32729 }, { "epoch": 6.0, "learning_rate": 3.0963528477036433e-06, "loss": 0.0025, "step": 32730 }, { "epoch": 6.0, "learning_rate": 3.09581574718143e-06, "loss": 0.0024, "step": 32731 }, { "epoch": 6.0, "learning_rate": 3.095278684715244e-06, "loss": 0.0027, "step": 32732 }, { "epoch": 6.0, "learning_rate": 3.0947416603080393e-06, "loss": 0.0024, "step": 32733 }, { "epoch": 6.0, "learning_rate": 3.094204673962784e-06, "loss": 0.0031, "step": 32734 }, { "epoch": 6.0, "learning_rate": 3.0936677256824344e-06, "loss": 0.0032, "step": 32735 }, { "epoch": 6.0, "learning_rate": 3.0931308154699523e-06, "loss": 0.002, "step": 32736 }, { "epoch": 6.0, "learning_rate": 3.0925939433282926e-06, "loss": 0.0014, "step": 32737 }, { "epoch": 6.0, "learning_rate": 3.0920571092604137e-06, "loss": 0.0026, "step": 32738 }, { "epoch": 6.0, "learning_rate": 3.0915203132692805e-06, "loss": 0.0016, "step": 32739 }, { "epoch": 6.0, "learning_rate": 3.0909835553578505e-06, "loss": 0.0023, "step": 32740 }, { "epoch": 6.0, "learning_rate": 3.0904468355290795e-06, "loss": 0.0031, "step": 32741 }, { "epoch": 6.0, "learning_rate": 3.089910153785929e-06, "loss": 0.0018, "step": 32742 }, { "epoch": 6.0, "learning_rate": 3.089373510131354e-06, "loss": 0.0032, "step": 32743 }, { "epoch": 6.0, "learning_rate": 3.0888369045683155e-06, "loss": 0.0024, "step": 32744 }, { "epoch": 6.0, "learning_rate": 3.088300337099769e-06, "loss": 0.0023, "step": 32745 }, { "epoch": 6.01, "learning_rate": 3.0877638077286732e-06, "loss": 0.0023, "step": 32746 }, { "epoch": 6.01, "learning_rate": 3.0872273164579846e-06, "loss": 0.0038, "step": 32747 }, { "epoch": 6.01, "learning_rate": 3.0866908632906623e-06, "loss": 0.0028, "step": 32748 }, { "epoch": 6.01, "learning_rate": 3.0861544482296613e-06, "loss": 0.0021, "step": 32749 }, { "epoch": 6.01, "learning_rate": 3.0856180712779384e-06, "loss": 0.002, "step": 32750 }, { "epoch": 6.01, "learning_rate": 3.0850817324384473e-06, "loss": 0.0028, "step": 32751 }, { "epoch": 6.01, "learning_rate": 3.084545431714152e-06, "loss": 0.0017, "step": 32752 }, { "epoch": 6.01, "learning_rate": 3.0840091691080045e-06, "loss": 0.0053, "step": 32753 }, { "epoch": 6.01, "learning_rate": 3.0834729446229626e-06, "loss": 0.0038, "step": 32754 }, { "epoch": 6.01, "learning_rate": 3.0829367582619774e-06, "loss": 0.0014, "step": 32755 }, { "epoch": 6.01, "learning_rate": 3.0824006100280023e-06, "loss": 0.0017, "step": 32756 }, { "epoch": 6.01, "learning_rate": 3.081864499924002e-06, "loss": 0.0015, "step": 32757 }, { "epoch": 6.01, "learning_rate": 3.0813284279529255e-06, "loss": 0.0029, "step": 32758 }, { "epoch": 6.01, "learning_rate": 3.08079239411773e-06, "loss": 0.0023, "step": 32759 }, { "epoch": 6.01, "learning_rate": 3.0802563984213675e-06, "loss": 0.0026, "step": 32760 }, { "epoch": 6.01, "learning_rate": 3.0797204408667946e-06, "loss": 0.0024, "step": 32761 }, { "epoch": 6.01, "learning_rate": 3.0791845214569647e-06, "loss": 0.0019, "step": 32762 }, { "epoch": 6.01, "learning_rate": 3.0786486401948323e-06, "loss": 0.0026, "step": 32763 }, { "epoch": 6.01, "learning_rate": 3.0781127970833467e-06, "loss": 0.003, "step": 32764 }, { "epoch": 6.01, "learning_rate": 3.0775769921254726e-06, "loss": 0.0029, "step": 32765 }, { "epoch": 6.01, "learning_rate": 3.077041225324152e-06, "loss": 0.003, "step": 32766 }, { "epoch": 6.01, "learning_rate": 3.0765054966823436e-06, "loss": 0.0022, "step": 32767 }, { "epoch": 6.01, "learning_rate": 3.0759698062029985e-06, "loss": 0.0037, "step": 32768 }, { "epoch": 6.01, "learning_rate": 3.0754341538890673e-06, "loss": 0.0026, "step": 32769 }, { "epoch": 6.01, "learning_rate": 3.0748985397435085e-06, "loss": 0.0026, "step": 32770 }, { "epoch": 6.01, "learning_rate": 3.074362963769271e-06, "loss": 0.0025, "step": 32771 }, { "epoch": 6.01, "learning_rate": 3.073827425969309e-06, "loss": 0.0019, "step": 32772 }, { "epoch": 6.01, "learning_rate": 3.0732919263465678e-06, "loss": 0.0019, "step": 32773 }, { "epoch": 6.01, "learning_rate": 3.0727564649040066e-06, "loss": 0.0025, "step": 32774 }, { "epoch": 6.01, "learning_rate": 3.0722210416445743e-06, "loss": 0.0015, "step": 32775 }, { "epoch": 6.01, "learning_rate": 3.0716856565712206e-06, "loss": 0.0026, "step": 32776 }, { "epoch": 6.01, "learning_rate": 3.071150309686899e-06, "loss": 0.0023, "step": 32777 }, { "epoch": 6.01, "learning_rate": 3.070615000994559e-06, "loss": 0.0031, "step": 32778 }, { "epoch": 6.01, "learning_rate": 3.0700797304971507e-06, "loss": 0.0032, "step": 32779 }, { "epoch": 6.01, "learning_rate": 3.069544498197626e-06, "loss": 0.002, "step": 32780 }, { "epoch": 6.01, "learning_rate": 3.0690093040989356e-06, "loss": 0.0019, "step": 32781 }, { "epoch": 6.01, "learning_rate": 3.0684741482040237e-06, "loss": 0.0024, "step": 32782 }, { "epoch": 6.01, "learning_rate": 3.067939030515851e-06, "loss": 0.0022, "step": 32783 }, { "epoch": 6.01, "learning_rate": 3.0674039510373576e-06, "loss": 0.0022, "step": 32784 }, { "epoch": 6.01, "learning_rate": 3.0668689097714967e-06, "loss": 0.0022, "step": 32785 }, { "epoch": 6.01, "learning_rate": 3.0663339067212127e-06, "loss": 0.002, "step": 32786 }, { "epoch": 6.01, "learning_rate": 3.0657989418894627e-06, "loss": 0.0029, "step": 32787 }, { "epoch": 6.01, "learning_rate": 3.065264015279191e-06, "loss": 0.0022, "step": 32788 }, { "epoch": 6.01, "learning_rate": 3.0647291268933466e-06, "loss": 0.0013, "step": 32789 }, { "epoch": 6.01, "learning_rate": 3.064194276734881e-06, "loss": 0.0026, "step": 32790 }, { "epoch": 6.01, "learning_rate": 3.063659464806732e-06, "loss": 0.0016, "step": 32791 }, { "epoch": 6.01, "learning_rate": 3.0631246911118586e-06, "loss": 0.0023, "step": 32792 }, { "epoch": 6.01, "learning_rate": 3.0625899556532036e-06, "loss": 0.0031, "step": 32793 }, { "epoch": 6.01, "learning_rate": 3.062055258433716e-06, "loss": 0.0024, "step": 32794 }, { "epoch": 6.01, "learning_rate": 3.0615205994563412e-06, "loss": 0.0024, "step": 32795 }, { "epoch": 6.01, "learning_rate": 3.060985978724029e-06, "loss": 0.0021, "step": 32796 }, { "epoch": 6.01, "learning_rate": 3.0604513962397243e-06, "loss": 0.0029, "step": 32797 }, { "epoch": 6.01, "learning_rate": 3.0599168520063725e-06, "loss": 0.003, "step": 32798 }, { "epoch": 6.01, "learning_rate": 3.0593823460269234e-06, "loss": 0.0022, "step": 32799 }, { "epoch": 6.02, "learning_rate": 3.0588478783043174e-06, "loss": 0.0023, "step": 32800 }, { "epoch": 6.02, "learning_rate": 3.0583134488415112e-06, "loss": 0.0026, "step": 32801 }, { "epoch": 6.02, "learning_rate": 3.057779057641441e-06, "loss": 0.0024, "step": 32802 }, { "epoch": 6.02, "learning_rate": 3.057244704707054e-06, "loss": 0.0016, "step": 32803 }, { "epoch": 6.02, "learning_rate": 3.0567103900412955e-06, "loss": 0.0032, "step": 32804 }, { "epoch": 6.02, "learning_rate": 3.0561761136471134e-06, "loss": 0.0024, "step": 32805 }, { "epoch": 6.02, "learning_rate": 3.055641875527453e-06, "loss": 0.0032, "step": 32806 }, { "epoch": 6.02, "learning_rate": 3.055107675685259e-06, "loss": 0.002, "step": 32807 }, { "epoch": 6.02, "learning_rate": 3.054573514123471e-06, "loss": 0.0024, "step": 32808 }, { "epoch": 6.02, "learning_rate": 3.054039390845035e-06, "loss": 0.0027, "step": 32809 }, { "epoch": 6.02, "learning_rate": 3.0535053058528984e-06, "loss": 0.0018, "step": 32810 }, { "epoch": 6.02, "learning_rate": 3.0529712591500026e-06, "loss": 0.0022, "step": 32811 }, { "epoch": 6.02, "learning_rate": 3.052437250739293e-06, "loss": 0.0019, "step": 32812 }, { "epoch": 6.02, "learning_rate": 3.0519032806237127e-06, "loss": 0.0029, "step": 32813 }, { "epoch": 6.02, "learning_rate": 3.0513693488062033e-06, "loss": 0.0013, "step": 32814 }, { "epoch": 6.02, "learning_rate": 3.050835455289709e-06, "loss": 0.0027, "step": 32815 }, { "epoch": 6.02, "learning_rate": 3.0503016000771725e-06, "loss": 0.0027, "step": 32816 }, { "epoch": 6.02, "learning_rate": 3.049767783171533e-06, "loss": 0.002, "step": 32817 }, { "epoch": 6.02, "learning_rate": 3.04923400457574e-06, "loss": 0.0025, "step": 32818 }, { "epoch": 6.02, "learning_rate": 3.0487002642927357e-06, "loss": 0.0022, "step": 32819 }, { "epoch": 6.02, "learning_rate": 3.048166562325454e-06, "loss": 0.0019, "step": 32820 }, { "epoch": 6.02, "learning_rate": 3.047632898676841e-06, "loss": 0.0031, "step": 32821 }, { "epoch": 6.02, "learning_rate": 3.0470992733498363e-06, "loss": 0.0021, "step": 32822 }, { "epoch": 6.02, "learning_rate": 3.046565686347386e-06, "loss": 0.0021, "step": 32823 }, { "epoch": 6.02, "learning_rate": 3.046032137672428e-06, "loss": 0.0022, "step": 32824 }, { "epoch": 6.02, "learning_rate": 3.0454986273279073e-06, "loss": 0.002, "step": 32825 }, { "epoch": 6.02, "learning_rate": 3.0449651553167547e-06, "loss": 0.0023, "step": 32826 }, { "epoch": 6.02, "learning_rate": 3.0444317216419207e-06, "loss": 0.0019, "step": 32827 }, { "epoch": 6.02, "learning_rate": 3.0438983263063416e-06, "loss": 0.0016, "step": 32828 }, { "epoch": 6.02, "learning_rate": 3.0433649693129573e-06, "loss": 0.0024, "step": 32829 }, { "epoch": 6.02, "learning_rate": 3.0428316506647093e-06, "loss": 0.0019, "step": 32830 }, { "epoch": 6.02, "learning_rate": 3.0422983703645358e-06, "loss": 0.0028, "step": 32831 }, { "epoch": 6.02, "learning_rate": 3.041765128415377e-06, "loss": 0.0025, "step": 32832 }, { "epoch": 6.02, "learning_rate": 3.041231924820172e-06, "loss": 0.0019, "step": 32833 }, { "epoch": 6.02, "learning_rate": 3.040698759581858e-06, "loss": 0.0021, "step": 32834 }, { "epoch": 6.02, "learning_rate": 3.0401656327033736e-06, "loss": 0.0027, "step": 32835 }, { "epoch": 6.02, "learning_rate": 3.0396325441876627e-06, "loss": 0.0021, "step": 32836 }, { "epoch": 6.02, "learning_rate": 3.039099494037663e-06, "loss": 0.0022, "step": 32837 }, { "epoch": 6.02, "learning_rate": 3.0385664822563064e-06, "loss": 0.0022, "step": 32838 }, { "epoch": 6.02, "learning_rate": 3.0380335088465327e-06, "loss": 0.0026, "step": 32839 }, { "epoch": 6.02, "learning_rate": 3.0375005738112826e-06, "loss": 0.0027, "step": 32840 }, { "epoch": 6.02, "learning_rate": 3.036967677153494e-06, "loss": 0.0027, "step": 32841 }, { "epoch": 6.02, "learning_rate": 3.036434818876103e-06, "loss": 0.0019, "step": 32842 }, { "epoch": 6.02, "learning_rate": 3.035901998982048e-06, "loss": 0.0025, "step": 32843 }, { "epoch": 6.02, "learning_rate": 3.0353692174742586e-06, "loss": 0.0029, "step": 32844 }, { "epoch": 6.02, "learning_rate": 3.034836474355681e-06, "loss": 0.0023, "step": 32845 }, { "epoch": 6.02, "learning_rate": 3.0343037696292476e-06, "loss": 0.003, "step": 32846 }, { "epoch": 6.02, "learning_rate": 3.0337711032978947e-06, "loss": 0.0032, "step": 32847 }, { "epoch": 6.02, "learning_rate": 3.0332384753645595e-06, "loss": 0.0024, "step": 32848 }, { "epoch": 6.02, "learning_rate": 3.0327058858321754e-06, "loss": 0.0018, "step": 32849 }, { "epoch": 6.02, "learning_rate": 3.0321733347036807e-06, "loss": 0.0021, "step": 32850 }, { "epoch": 6.02, "learning_rate": 3.0316408219820103e-06, "loss": 0.002, "step": 32851 }, { "epoch": 6.02, "learning_rate": 3.0311083476700988e-06, "loss": 0.0034, "step": 32852 }, { "epoch": 6.02, "learning_rate": 3.0305759117708777e-06, "loss": 0.0025, "step": 32853 }, { "epoch": 6.02, "learning_rate": 3.030043514287292e-06, "loss": 0.0038, "step": 32854 }, { "epoch": 6.03, "learning_rate": 3.029511155222267e-06, "loss": 0.003, "step": 32855 }, { "epoch": 6.03, "learning_rate": 3.0289788345787385e-06, "loss": 0.0018, "step": 32856 }, { "epoch": 6.03, "learning_rate": 3.0284465523596396e-06, "loss": 0.0023, "step": 32857 }, { "epoch": 6.03, "learning_rate": 3.02791430856791e-06, "loss": 0.0027, "step": 32858 }, { "epoch": 6.03, "learning_rate": 3.02738210320648e-06, "loss": 0.0019, "step": 32859 }, { "epoch": 6.03, "learning_rate": 3.026849936278283e-06, "loss": 0.0015, "step": 32860 }, { "epoch": 6.03, "learning_rate": 3.0263178077862533e-06, "loss": 0.0025, "step": 32861 }, { "epoch": 6.03, "learning_rate": 3.025785717733322e-06, "loss": 0.002, "step": 32862 }, { "epoch": 6.03, "learning_rate": 3.025253666122424e-06, "loss": 0.0028, "step": 32863 }, { "epoch": 6.03, "learning_rate": 3.0247216529564906e-06, "loss": 0.0027, "step": 32864 }, { "epoch": 6.03, "learning_rate": 3.024189678238455e-06, "loss": 0.0024, "step": 32865 }, { "epoch": 6.03, "learning_rate": 3.0236577419712498e-06, "loss": 0.0016, "step": 32866 }, { "epoch": 6.03, "learning_rate": 3.023125844157806e-06, "loss": 0.0025, "step": 32867 }, { "epoch": 6.03, "learning_rate": 3.0225939848010567e-06, "loss": 0.002, "step": 32868 }, { "epoch": 6.03, "learning_rate": 3.0220621639039327e-06, "loss": 0.0025, "step": 32869 }, { "epoch": 6.03, "learning_rate": 3.021530381469362e-06, "loss": 0.0027, "step": 32870 }, { "epoch": 6.03, "learning_rate": 3.020998637500282e-06, "loss": 0.0023, "step": 32871 }, { "epoch": 6.03, "learning_rate": 3.020466931999625e-06, "loss": 0.0028, "step": 32872 }, { "epoch": 6.03, "learning_rate": 3.019935264970314e-06, "loss": 0.0013, "step": 32873 }, { "epoch": 6.03, "learning_rate": 3.0194036364152845e-06, "loss": 0.0016, "step": 32874 }, { "epoch": 6.03, "learning_rate": 3.018872046337461e-06, "loss": 0.0033, "step": 32875 }, { "epoch": 6.03, "learning_rate": 3.018340494739783e-06, "loss": 0.0016, "step": 32876 }, { "epoch": 6.03, "learning_rate": 3.0178089816251743e-06, "loss": 0.0021, "step": 32877 }, { "epoch": 6.03, "learning_rate": 3.0172775069965675e-06, "loss": 0.0018, "step": 32878 }, { "epoch": 6.03, "learning_rate": 3.0167460708568898e-06, "loss": 0.002, "step": 32879 }, { "epoch": 6.03, "learning_rate": 3.0162146732090712e-06, "loss": 0.0017, "step": 32880 }, { "epoch": 6.03, "learning_rate": 3.0156833140560414e-06, "loss": 0.002, "step": 32881 }, { "epoch": 6.03, "learning_rate": 3.015151993400729e-06, "loss": 0.0023, "step": 32882 }, { "epoch": 6.03, "learning_rate": 3.014620711246061e-06, "loss": 0.0024, "step": 32883 }, { "epoch": 6.03, "learning_rate": 3.0140894675949696e-06, "loss": 0.0014, "step": 32884 }, { "epoch": 6.03, "learning_rate": 3.013558262450379e-06, "loss": 0.0024, "step": 32885 }, { "epoch": 6.03, "learning_rate": 3.0130270958152196e-06, "loss": 0.0027, "step": 32886 }, { "epoch": 6.03, "learning_rate": 3.0124959676924183e-06, "loss": 0.0021, "step": 32887 }, { "epoch": 6.03, "learning_rate": 3.0119648780849e-06, "loss": 0.0027, "step": 32888 }, { "epoch": 6.03, "learning_rate": 3.0114338269955976e-06, "loss": 0.0027, "step": 32889 }, { "epoch": 6.03, "learning_rate": 3.0109028144274388e-06, "loss": 0.0035, "step": 32890 }, { "epoch": 6.03, "learning_rate": 3.010371840383345e-06, "loss": 0.0014, "step": 32891 }, { "epoch": 6.03, "learning_rate": 3.0098409048662412e-06, "loss": 0.0032, "step": 32892 }, { "epoch": 6.03, "learning_rate": 3.0093100078790628e-06, "loss": 0.0019, "step": 32893 }, { "epoch": 6.03, "learning_rate": 3.00877914942473e-06, "loss": 0.0025, "step": 32894 }, { "epoch": 6.03, "learning_rate": 3.0082483295061714e-06, "loss": 0.0018, "step": 32895 }, { "epoch": 6.03, "learning_rate": 3.007717548126311e-06, "loss": 0.0029, "step": 32896 }, { "epoch": 6.03, "learning_rate": 3.007186805288076e-06, "loss": 0.0015, "step": 32897 }, { "epoch": 6.03, "learning_rate": 3.0066561009943908e-06, "loss": 0.0015, "step": 32898 }, { "epoch": 6.03, "learning_rate": 3.0061254352481806e-06, "loss": 0.0014, "step": 32899 }, { "epoch": 6.03, "learning_rate": 3.005594808052371e-06, "loss": 0.0027, "step": 32900 }, { "epoch": 6.03, "learning_rate": 3.0050642194098866e-06, "loss": 0.0032, "step": 32901 }, { "epoch": 6.03, "learning_rate": 3.004533669323653e-06, "loss": 0.0014, "step": 32902 }, { "epoch": 6.03, "learning_rate": 3.0040031577965924e-06, "loss": 0.0011, "step": 32903 }, { "epoch": 6.03, "learning_rate": 3.003472684831631e-06, "loss": 0.0025, "step": 32904 }, { "epoch": 6.03, "learning_rate": 3.002942250431692e-06, "loss": 0.0024, "step": 32905 }, { "epoch": 6.03, "learning_rate": 3.002411854599696e-06, "loss": 0.0026, "step": 32906 }, { "epoch": 6.03, "learning_rate": 3.0018814973385736e-06, "loss": 0.0012, "step": 32907 }, { "epoch": 6.03, "learning_rate": 3.0013511786512463e-06, "loss": 0.0032, "step": 32908 }, { "epoch": 6.04, "learning_rate": 3.0008208985406327e-06, "loss": 0.0018, "step": 32909 }, { "epoch": 6.04, "learning_rate": 3.0002906570096555e-06, "loss": 0.0039, "step": 32910 }, { "epoch": 6.04, "learning_rate": 2.9997604540612426e-06, "loss": 0.0028, "step": 32911 }, { "epoch": 6.04, "learning_rate": 2.999230289698315e-06, "loss": 0.0023, "step": 32912 }, { "epoch": 6.04, "learning_rate": 2.9987001639237932e-06, "loss": 0.0028, "step": 32913 }, { "epoch": 6.04, "learning_rate": 2.998170076740601e-06, "loss": 0.0031, "step": 32914 }, { "epoch": 6.04, "learning_rate": 2.997640028151658e-06, "loss": 0.0023, "step": 32915 }, { "epoch": 6.04, "learning_rate": 2.997110018159888e-06, "loss": 0.0035, "step": 32916 }, { "epoch": 6.04, "learning_rate": 2.9965800467682115e-06, "loss": 0.0021, "step": 32917 }, { "epoch": 6.04, "learning_rate": 2.9960501139795506e-06, "loss": 0.0019, "step": 32918 }, { "epoch": 6.04, "learning_rate": 2.9955202197968248e-06, "loss": 0.0021, "step": 32919 }, { "epoch": 6.04, "learning_rate": 2.994990364222956e-06, "loss": 0.0021, "step": 32920 }, { "epoch": 6.04, "learning_rate": 2.9944605472608646e-06, "loss": 0.0023, "step": 32921 }, { "epoch": 6.04, "learning_rate": 2.99393076891347e-06, "loss": 0.0018, "step": 32922 }, { "epoch": 6.04, "learning_rate": 2.99340102918369e-06, "loss": 0.0025, "step": 32923 }, { "epoch": 6.04, "learning_rate": 2.9928713280744516e-06, "loss": 0.0023, "step": 32924 }, { "epoch": 6.04, "learning_rate": 2.9923416655886705e-06, "loss": 0.0025, "step": 32925 }, { "epoch": 6.04, "learning_rate": 2.9918120417292694e-06, "loss": 0.0019, "step": 32926 }, { "epoch": 6.04, "learning_rate": 2.9912824564991607e-06, "loss": 0.0017, "step": 32927 }, { "epoch": 6.04, "learning_rate": 2.9907529099012646e-06, "loss": 0.004, "step": 32928 }, { "epoch": 6.04, "learning_rate": 2.9902234019385056e-06, "loss": 0.0016, "step": 32929 }, { "epoch": 6.04, "learning_rate": 2.9896939326138e-06, "loss": 0.002, "step": 32930 }, { "epoch": 6.04, "learning_rate": 2.9891645019300655e-06, "loss": 0.0039, "step": 32931 }, { "epoch": 6.04, "learning_rate": 2.98863510989022e-06, "loss": 0.0019, "step": 32932 }, { "epoch": 6.04, "learning_rate": 2.988105756497183e-06, "loss": 0.0023, "step": 32933 }, { "epoch": 6.04, "learning_rate": 2.98757644175387e-06, "loss": 0.0025, "step": 32934 }, { "epoch": 6.04, "learning_rate": 2.9870471656632004e-06, "loss": 0.0015, "step": 32935 }, { "epoch": 6.04, "learning_rate": 2.986517928228092e-06, "loss": 0.0022, "step": 32936 }, { "epoch": 6.04, "learning_rate": 2.98598872945146e-06, "loss": 0.0032, "step": 32937 }, { "epoch": 6.04, "learning_rate": 2.985459569336223e-06, "loss": 0.0034, "step": 32938 }, { "epoch": 6.04, "learning_rate": 2.984930447885297e-06, "loss": 0.0026, "step": 32939 }, { "epoch": 6.04, "learning_rate": 2.984401365101599e-06, "loss": 0.0024, "step": 32940 }, { "epoch": 6.04, "learning_rate": 2.9838723209880414e-06, "loss": 0.0025, "step": 32941 }, { "epoch": 6.04, "learning_rate": 2.983343315547548e-06, "loss": 0.0019, "step": 32942 }, { "epoch": 6.04, "learning_rate": 2.98281434878303e-06, "loss": 0.002, "step": 32943 }, { "epoch": 6.04, "learning_rate": 2.9822854206974074e-06, "loss": 0.0032, "step": 32944 }, { "epoch": 6.04, "learning_rate": 2.981756531293585e-06, "loss": 0.0021, "step": 32945 }, { "epoch": 6.04, "learning_rate": 2.9812276805744878e-06, "loss": 0.0013, "step": 32946 }, { "epoch": 6.04, "learning_rate": 2.9806988685430284e-06, "loss": 0.0022, "step": 32947 }, { "epoch": 6.04, "learning_rate": 2.980170095202122e-06, "loss": 0.0017, "step": 32948 }, { "epoch": 6.04, "learning_rate": 2.9796413605546816e-06, "loss": 0.0014, "step": 32949 }, { "epoch": 6.04, "learning_rate": 2.979112664603623e-06, "loss": 0.0028, "step": 32950 }, { "epoch": 6.04, "learning_rate": 2.9785840073518603e-06, "loss": 0.0024, "step": 32951 }, { "epoch": 6.04, "learning_rate": 2.9780553888023057e-06, "loss": 0.0034, "step": 32952 }, { "epoch": 6.04, "learning_rate": 2.977526808957876e-06, "loss": 0.0026, "step": 32953 }, { "epoch": 6.04, "learning_rate": 2.976998267821478e-06, "loss": 0.0032, "step": 32954 }, { "epoch": 6.04, "learning_rate": 2.976469765396037e-06, "loss": 0.0029, "step": 32955 }, { "epoch": 6.04, "learning_rate": 2.9759413016844572e-06, "loss": 0.002, "step": 32956 }, { "epoch": 6.04, "learning_rate": 2.975412876689653e-06, "loss": 0.0023, "step": 32957 }, { "epoch": 6.04, "learning_rate": 2.974884490414537e-06, "loss": 0.0017, "step": 32958 }, { "epoch": 6.04, "learning_rate": 2.9743561428620193e-06, "loss": 0.0012, "step": 32959 }, { "epoch": 6.04, "learning_rate": 2.973827834035018e-06, "loss": 0.0021, "step": 32960 }, { "epoch": 6.04, "learning_rate": 2.973299563936446e-06, "loss": 0.0024, "step": 32961 }, { "epoch": 6.04, "learning_rate": 2.9727713325692078e-06, "loss": 0.0028, "step": 32962 }, { "epoch": 6.04, "learning_rate": 2.972243139936216e-06, "loss": 0.0015, "step": 32963 }, { "epoch": 6.05, "learning_rate": 2.971714986040387e-06, "loss": 0.0029, "step": 32964 }, { "epoch": 6.05, "learning_rate": 2.9711868708846302e-06, "loss": 0.0027, "step": 32965 }, { "epoch": 6.05, "learning_rate": 2.9706587944718567e-06, "loss": 0.002, "step": 32966 }, { "epoch": 6.05, "learning_rate": 2.9701307568049764e-06, "loss": 0.002, "step": 32967 }, { "epoch": 6.05, "learning_rate": 2.969602757886899e-06, "loss": 0.0027, "step": 32968 }, { "epoch": 6.05, "learning_rate": 2.969074797720537e-06, "loss": 0.0022, "step": 32969 }, { "epoch": 6.05, "learning_rate": 2.968546876308799e-06, "loss": 0.002, "step": 32970 }, { "epoch": 6.05, "learning_rate": 2.9680189936545955e-06, "loss": 0.0016, "step": 32971 }, { "epoch": 6.05, "learning_rate": 2.967491149760833e-06, "loss": 0.0014, "step": 32972 }, { "epoch": 6.05, "learning_rate": 2.966963344630429e-06, "loss": 0.0025, "step": 32973 }, { "epoch": 6.05, "learning_rate": 2.9664355782662858e-06, "loss": 0.0014, "step": 32974 }, { "epoch": 6.05, "learning_rate": 2.9659078506713144e-06, "loss": 0.0022, "step": 32975 }, { "epoch": 6.05, "learning_rate": 2.965380161848421e-06, "loss": 0.0029, "step": 32976 }, { "epoch": 6.05, "learning_rate": 2.964852511800519e-06, "loss": 0.0013, "step": 32977 }, { "epoch": 6.05, "learning_rate": 2.9643249005305165e-06, "loss": 0.003, "step": 32978 }, { "epoch": 6.05, "learning_rate": 2.963797328041321e-06, "loss": 0.0015, "step": 32979 }, { "epoch": 6.05, "learning_rate": 2.963269794335838e-06, "loss": 0.002, "step": 32980 }, { "epoch": 6.05, "learning_rate": 2.962742299416973e-06, "loss": 0.0027, "step": 32981 }, { "epoch": 6.05, "learning_rate": 2.962214843287641e-06, "loss": 0.0015, "step": 32982 }, { "epoch": 6.05, "learning_rate": 2.9616874259507445e-06, "loss": 0.0019, "step": 32983 }, { "epoch": 6.05, "learning_rate": 2.9611600474091916e-06, "loss": 0.0016, "step": 32984 }, { "epoch": 6.05, "learning_rate": 2.9606327076658914e-06, "loss": 0.0028, "step": 32985 }, { "epoch": 6.05, "learning_rate": 2.9601054067237466e-06, "loss": 0.0029, "step": 32986 }, { "epoch": 6.05, "learning_rate": 2.9595781445856665e-06, "loss": 0.0036, "step": 32987 }, { "epoch": 6.05, "learning_rate": 2.959050921254557e-06, "loss": 0.0014, "step": 32988 }, { "epoch": 6.05, "learning_rate": 2.9585237367333207e-06, "loss": 0.0017, "step": 32989 }, { "epoch": 6.05, "learning_rate": 2.957996591024869e-06, "loss": 0.003, "step": 32990 }, { "epoch": 6.05, "learning_rate": 2.9574694841321082e-06, "loss": 0.0026, "step": 32991 }, { "epoch": 6.05, "learning_rate": 2.9569424160579376e-06, "loss": 0.0029, "step": 32992 }, { "epoch": 6.05, "learning_rate": 2.956415386805266e-06, "loss": 0.0034, "step": 32993 }, { "epoch": 6.05, "learning_rate": 2.9558883963769934e-06, "loss": 0.0033, "step": 32994 }, { "epoch": 6.05, "learning_rate": 2.955361444776034e-06, "loss": 0.0012, "step": 32995 }, { "epoch": 6.05, "learning_rate": 2.9548345320052863e-06, "loss": 0.0016, "step": 32996 }, { "epoch": 6.05, "learning_rate": 2.9543076580676587e-06, "loss": 0.002, "step": 32997 }, { "epoch": 6.05, "learning_rate": 2.953780822966046e-06, "loss": 0.003, "step": 32998 }, { "epoch": 6.05, "learning_rate": 2.9532540267033614e-06, "loss": 0.0018, "step": 32999 }, { "epoch": 6.05, "learning_rate": 2.952727269282506e-06, "loss": 0.0017, "step": 33000 }, { "epoch": 6.05, "learning_rate": 2.9522005507063835e-06, "loss": 0.0022, "step": 33001 }, { "epoch": 6.05, "learning_rate": 2.951673870977897e-06, "loss": 0.004, "step": 33002 }, { "epoch": 6.05, "learning_rate": 2.9511472300999478e-06, "loss": 0.003, "step": 33003 }, { "epoch": 6.05, "learning_rate": 2.9506206280754415e-06, "loss": 0.0028, "step": 33004 }, { "epoch": 6.05, "learning_rate": 2.9500940649072786e-06, "loss": 0.0018, "step": 33005 }, { "epoch": 6.05, "learning_rate": 2.949567540598363e-06, "loss": 0.0019, "step": 33006 }, { "epoch": 6.05, "learning_rate": 2.9490410551515927e-06, "loss": 0.0022, "step": 33007 }, { "epoch": 6.05, "learning_rate": 2.948514608569879e-06, "loss": 0.0022, "step": 33008 }, { "epoch": 6.05, "learning_rate": 2.9479882008561156e-06, "loss": 0.0016, "step": 33009 }, { "epoch": 6.05, "learning_rate": 2.9474618320132063e-06, "loss": 0.0017, "step": 33010 }, { "epoch": 6.05, "learning_rate": 2.946935502044053e-06, "loss": 0.0027, "step": 33011 }, { "epoch": 6.05, "learning_rate": 2.946409210951553e-06, "loss": 0.002, "step": 33012 }, { "epoch": 6.05, "learning_rate": 2.945882958738614e-06, "loss": 0.0021, "step": 33013 }, { "epoch": 6.05, "learning_rate": 2.9453567454081334e-06, "loss": 0.0012, "step": 33014 }, { "epoch": 6.05, "learning_rate": 2.944830570963014e-06, "loss": 0.0012, "step": 33015 }, { "epoch": 6.05, "learning_rate": 2.9443044354061477e-06, "loss": 0.0019, "step": 33016 }, { "epoch": 6.05, "learning_rate": 2.943778338740445e-06, "loss": 0.002, "step": 33017 }, { "epoch": 6.06, "learning_rate": 2.9432522809688002e-06, "loss": 0.0031, "step": 33018 }, { "epoch": 6.06, "learning_rate": 2.9427262620941142e-06, "loss": 0.0026, "step": 33019 }, { "epoch": 6.06, "learning_rate": 2.9422002821192876e-06, "loss": 0.0014, "step": 33020 }, { "epoch": 6.06, "learning_rate": 2.941674341047217e-06, "loss": 0.0025, "step": 33021 }, { "epoch": 6.06, "learning_rate": 2.941148438880803e-06, "loss": 0.003, "step": 33022 }, { "epoch": 6.06, "learning_rate": 2.940622575622946e-06, "loss": 0.0037, "step": 33023 }, { "epoch": 6.06, "learning_rate": 2.94009675127654e-06, "loss": 0.0026, "step": 33024 }, { "epoch": 6.06, "learning_rate": 2.939570965844485e-06, "loss": 0.0038, "step": 33025 }, { "epoch": 6.06, "learning_rate": 2.9390452193296847e-06, "loss": 0.0025, "step": 33026 }, { "epoch": 6.06, "learning_rate": 2.9385195117350307e-06, "loss": 0.0031, "step": 33027 }, { "epoch": 6.06, "learning_rate": 2.9379938430634223e-06, "loss": 0.0027, "step": 33028 }, { "epoch": 6.06, "learning_rate": 2.9374682133177547e-06, "loss": 0.0035, "step": 33029 }, { "epoch": 6.06, "learning_rate": 2.9369426225009302e-06, "loss": 0.0017, "step": 33030 }, { "epoch": 6.06, "learning_rate": 2.9364170706158436e-06, "loss": 0.0015, "step": 33031 }, { "epoch": 6.06, "learning_rate": 2.9358915576653926e-06, "loss": 0.0017, "step": 33032 }, { "epoch": 6.06, "learning_rate": 2.9353660836524743e-06, "loss": 0.0014, "step": 33033 }, { "epoch": 6.06, "learning_rate": 2.9348406485799786e-06, "loss": 0.0024, "step": 33034 }, { "epoch": 6.06, "learning_rate": 2.934315252450809e-06, "loss": 0.0024, "step": 33035 }, { "epoch": 6.06, "learning_rate": 2.933789895267859e-06, "loss": 0.0016, "step": 33036 }, { "epoch": 6.06, "learning_rate": 2.933264577034025e-06, "loss": 0.0015, "step": 33037 }, { "epoch": 6.06, "learning_rate": 2.932739297752203e-06, "loss": 0.0016, "step": 33038 }, { "epoch": 6.06, "learning_rate": 2.9322140574252865e-06, "loss": 0.0026, "step": 33039 }, { "epoch": 6.06, "learning_rate": 2.9316888560561727e-06, "loss": 0.0015, "step": 33040 }, { "epoch": 6.06, "learning_rate": 2.9311636936477548e-06, "loss": 0.0026, "step": 33041 }, { "epoch": 6.06, "learning_rate": 2.930638570202926e-06, "loss": 0.002, "step": 33042 }, { "epoch": 6.06, "learning_rate": 2.930113485724585e-06, "loss": 0.0021, "step": 33043 }, { "epoch": 6.06, "learning_rate": 2.929588440215628e-06, "loss": 0.002, "step": 33044 }, { "epoch": 6.06, "learning_rate": 2.9290634336789412e-06, "loss": 0.0022, "step": 33045 }, { "epoch": 6.06, "learning_rate": 2.9285384661174233e-06, "loss": 0.002, "step": 33046 }, { "epoch": 6.06, "learning_rate": 2.9280135375339637e-06, "loss": 0.0018, "step": 33047 }, { "epoch": 6.06, "learning_rate": 2.9274886479314626e-06, "loss": 0.0023, "step": 33048 }, { "epoch": 6.06, "learning_rate": 2.9269637973128096e-06, "loss": 0.0024, "step": 33049 }, { "epoch": 6.06, "learning_rate": 2.926438985680897e-06, "loss": 0.002, "step": 33050 }, { "epoch": 6.06, "learning_rate": 2.9259142130386188e-06, "loss": 0.0035, "step": 33051 }, { "epoch": 6.06, "learning_rate": 2.925389479388867e-06, "loss": 0.0024, "step": 33052 }, { "epoch": 6.06, "learning_rate": 2.9248647847345346e-06, "loss": 0.0015, "step": 33053 }, { "epoch": 6.06, "learning_rate": 2.9243401290785135e-06, "loss": 0.0018, "step": 33054 }, { "epoch": 6.06, "learning_rate": 2.9238155124236954e-06, "loss": 0.003, "step": 33055 }, { "epoch": 6.06, "learning_rate": 2.923290934772971e-06, "loss": 0.0022, "step": 33056 }, { "epoch": 6.06, "learning_rate": 2.9227663961292342e-06, "loss": 0.0028, "step": 33057 }, { "epoch": 6.06, "learning_rate": 2.9222418964953738e-06, "loss": 0.0017, "step": 33058 }, { "epoch": 6.06, "learning_rate": 2.9217174358742817e-06, "loss": 0.0026, "step": 33059 }, { "epoch": 6.06, "learning_rate": 2.921193014268846e-06, "loss": 0.0032, "step": 33060 }, { "epoch": 6.06, "learning_rate": 2.9206686316819645e-06, "loss": 0.0031, "step": 33061 }, { "epoch": 6.06, "learning_rate": 2.9201442881165254e-06, "loss": 0.002, "step": 33062 }, { "epoch": 6.06, "learning_rate": 2.9196199835754133e-06, "loss": 0.0014, "step": 33063 }, { "epoch": 6.06, "learning_rate": 2.919095718061522e-06, "loss": 0.0022, "step": 33064 }, { "epoch": 6.06, "learning_rate": 2.9185714915777386e-06, "loss": 0.0021, "step": 33065 }, { "epoch": 6.06, "learning_rate": 2.9180473041269574e-06, "loss": 0.0041, "step": 33066 }, { "epoch": 6.06, "learning_rate": 2.9175231557120666e-06, "loss": 0.002, "step": 33067 }, { "epoch": 6.06, "learning_rate": 2.9169990463359556e-06, "loss": 0.0017, "step": 33068 }, { "epoch": 6.06, "learning_rate": 2.9164749760015067e-06, "loss": 0.0015, "step": 33069 }, { "epoch": 6.06, "learning_rate": 2.915950944711616e-06, "loss": 0.0014, "step": 33070 }, { "epoch": 6.06, "learning_rate": 2.9154269524691703e-06, "loss": 0.0018, "step": 33071 }, { "epoch": 6.06, "learning_rate": 2.914902999277057e-06, "loss": 0.003, "step": 33072 }, { "epoch": 6.07, "learning_rate": 2.9143790851381636e-06, "loss": 0.0016, "step": 33073 }, { "epoch": 6.07, "learning_rate": 2.9138552100553796e-06, "loss": 0.0028, "step": 33074 }, { "epoch": 6.07, "learning_rate": 2.9133313740315916e-06, "loss": 0.0024, "step": 33075 }, { "epoch": 6.07, "learning_rate": 2.9128075770696872e-06, "loss": 0.0032, "step": 33076 }, { "epoch": 6.07, "learning_rate": 2.9122838191725532e-06, "loss": 0.0036, "step": 33077 }, { "epoch": 6.07, "learning_rate": 2.911760100343074e-06, "loss": 0.0019, "step": 33078 }, { "epoch": 6.07, "learning_rate": 2.9112364205841416e-06, "loss": 0.0021, "step": 33079 }, { "epoch": 6.07, "learning_rate": 2.9107127798986433e-06, "loss": 0.0024, "step": 33080 }, { "epoch": 6.07, "learning_rate": 2.91018917828946e-06, "loss": 0.0018, "step": 33081 }, { "epoch": 6.07, "learning_rate": 2.909665615759476e-06, "loss": 0.003, "step": 33082 }, { "epoch": 6.07, "learning_rate": 2.9091420923115853e-06, "loss": 0.0015, "step": 33083 }, { "epoch": 6.07, "learning_rate": 2.908618607948669e-06, "loss": 0.0019, "step": 33084 }, { "epoch": 6.07, "learning_rate": 2.9080951626736133e-06, "loss": 0.0021, "step": 33085 }, { "epoch": 6.07, "learning_rate": 2.907571756489306e-06, "loss": 0.0025, "step": 33086 }, { "epoch": 6.07, "learning_rate": 2.907048389398622e-06, "loss": 0.002, "step": 33087 }, { "epoch": 6.07, "learning_rate": 2.906525061404458e-06, "loss": 0.0019, "step": 33088 }, { "epoch": 6.07, "learning_rate": 2.9060017725096943e-06, "loss": 0.0015, "step": 33089 }, { "epoch": 6.07, "learning_rate": 2.9054785227172143e-06, "loss": 0.0023, "step": 33090 }, { "epoch": 6.07, "learning_rate": 2.9049553120299035e-06, "loss": 0.0027, "step": 33091 }, { "epoch": 6.07, "learning_rate": 2.9044321404506446e-06, "loss": 0.002, "step": 33092 }, { "epoch": 6.07, "learning_rate": 2.903909007982322e-06, "loss": 0.0039, "step": 33093 }, { "epoch": 6.07, "learning_rate": 2.90338591462782e-06, "loss": 0.0023, "step": 33094 }, { "epoch": 6.07, "learning_rate": 2.9028628603900165e-06, "loss": 0.0019, "step": 33095 }, { "epoch": 6.07, "learning_rate": 2.9023398452718033e-06, "loss": 0.0012, "step": 33096 }, { "epoch": 6.07, "learning_rate": 2.9018168692760595e-06, "loss": 0.0016, "step": 33097 }, { "epoch": 6.07, "learning_rate": 2.90129393240567e-06, "loss": 0.0012, "step": 33098 }, { "epoch": 6.07, "learning_rate": 2.9007710346635122e-06, "loss": 0.0019, "step": 33099 }, { "epoch": 6.07, "learning_rate": 2.9002481760524657e-06, "loss": 0.0025, "step": 33100 }, { "epoch": 6.07, "learning_rate": 2.8997253565754225e-06, "loss": 0.0023, "step": 33101 }, { "epoch": 6.07, "learning_rate": 2.8992025762352593e-06, "loss": 0.0022, "step": 33102 }, { "epoch": 6.07, "learning_rate": 2.8986798350348576e-06, "loss": 0.0028, "step": 33103 }, { "epoch": 6.07, "learning_rate": 2.8981571329770986e-06, "loss": 0.0017, "step": 33104 }, { "epoch": 6.07, "learning_rate": 2.8976344700648653e-06, "loss": 0.0014, "step": 33105 }, { "epoch": 6.07, "learning_rate": 2.897111846301035e-06, "loss": 0.0027, "step": 33106 }, { "epoch": 6.07, "learning_rate": 2.8965892616884928e-06, "loss": 0.0019, "step": 33107 }, { "epoch": 6.07, "learning_rate": 2.896066716230116e-06, "loss": 0.0018, "step": 33108 }, { "epoch": 6.07, "learning_rate": 2.8955442099287854e-06, "loss": 0.003, "step": 33109 }, { "epoch": 6.07, "learning_rate": 2.8950217427873816e-06, "loss": 0.0019, "step": 33110 }, { "epoch": 6.07, "learning_rate": 2.894499314808785e-06, "loss": 0.0047, "step": 33111 }, { "epoch": 6.07, "learning_rate": 2.893976925995875e-06, "loss": 0.0018, "step": 33112 }, { "epoch": 6.07, "learning_rate": 2.8934545763515266e-06, "loss": 0.002, "step": 33113 }, { "epoch": 6.07, "learning_rate": 2.8929322658786263e-06, "loss": 0.0016, "step": 33114 }, { "epoch": 6.07, "learning_rate": 2.8924099945800533e-06, "loss": 0.0025, "step": 33115 }, { "epoch": 6.07, "learning_rate": 2.8918877624586794e-06, "loss": 0.0024, "step": 33116 }, { "epoch": 6.07, "learning_rate": 2.891365569517385e-06, "loss": 0.0037, "step": 33117 }, { "epoch": 6.07, "learning_rate": 2.8908434157590516e-06, "loss": 0.0015, "step": 33118 }, { "epoch": 6.07, "learning_rate": 2.8903213011865562e-06, "loss": 0.0019, "step": 33119 }, { "epoch": 6.07, "learning_rate": 2.8897992258027762e-06, "loss": 0.0022, "step": 33120 }, { "epoch": 6.07, "learning_rate": 2.88927718961059e-06, "loss": 0.0032, "step": 33121 }, { "epoch": 6.07, "learning_rate": 2.8887551926128744e-06, "loss": 0.0024, "step": 33122 }, { "epoch": 6.07, "learning_rate": 2.8882332348125064e-06, "loss": 0.002, "step": 33123 }, { "epoch": 6.07, "learning_rate": 2.8877113162123637e-06, "loss": 0.0015, "step": 33124 }, { "epoch": 6.07, "learning_rate": 2.887189436815323e-06, "loss": 0.0014, "step": 33125 }, { "epoch": 6.07, "learning_rate": 2.8866675966242574e-06, "loss": 0.0031, "step": 33126 }, { "epoch": 6.08, "learning_rate": 2.8861457956420535e-06, "loss": 0.002, "step": 33127 }, { "epoch": 6.08, "learning_rate": 2.8856240338715768e-06, "loss": 0.0014, "step": 33128 }, { "epoch": 6.08, "learning_rate": 2.885102311315707e-06, "loss": 0.0012, "step": 33129 }, { "epoch": 6.08, "learning_rate": 2.88458062797732e-06, "loss": 0.0016, "step": 33130 }, { "epoch": 6.08, "learning_rate": 2.884058983859289e-06, "loss": 0.0024, "step": 33131 }, { "epoch": 6.08, "learning_rate": 2.883537378964494e-06, "loss": 0.0034, "step": 33132 }, { "epoch": 6.08, "learning_rate": 2.883015813295811e-06, "loss": 0.0022, "step": 33133 }, { "epoch": 6.08, "learning_rate": 2.882494286856108e-06, "loss": 0.0018, "step": 33134 }, { "epoch": 6.08, "learning_rate": 2.8819727996482604e-06, "loss": 0.0037, "step": 33135 }, { "epoch": 6.08, "learning_rate": 2.8814513516751485e-06, "loss": 0.002, "step": 33136 }, { "epoch": 6.08, "learning_rate": 2.8809299429396444e-06, "loss": 0.0019, "step": 33137 }, { "epoch": 6.08, "learning_rate": 2.8804085734446208e-06, "loss": 0.0017, "step": 33138 }, { "epoch": 6.08, "learning_rate": 2.8798872431929513e-06, "loss": 0.002, "step": 33139 }, { "epoch": 6.08, "learning_rate": 2.8793659521875107e-06, "loss": 0.002, "step": 33140 }, { "epoch": 6.08, "learning_rate": 2.8788447004311713e-06, "loss": 0.0015, "step": 33141 }, { "epoch": 6.08, "learning_rate": 2.8783234879268076e-06, "loss": 0.0017, "step": 33142 }, { "epoch": 6.08, "learning_rate": 2.8778023146772904e-06, "loss": 0.0012, "step": 33143 }, { "epoch": 6.08, "learning_rate": 2.8772811806854918e-06, "loss": 0.0033, "step": 33144 }, { "epoch": 6.08, "learning_rate": 2.876760085954292e-06, "loss": 0.0019, "step": 33145 }, { "epoch": 6.08, "learning_rate": 2.876239030486554e-06, "loss": 0.0027, "step": 33146 }, { "epoch": 6.08, "learning_rate": 2.875718014285154e-06, "loss": 0.0014, "step": 33147 }, { "epoch": 6.08, "learning_rate": 2.8751970373529592e-06, "loss": 0.0029, "step": 33148 }, { "epoch": 6.08, "learning_rate": 2.8746760996928492e-06, "loss": 0.0016, "step": 33149 }, { "epoch": 6.08, "learning_rate": 2.874155201307691e-06, "loss": 0.0025, "step": 33150 }, { "epoch": 6.08, "learning_rate": 2.87363434220036e-06, "loss": 0.0018, "step": 33151 }, { "epoch": 6.08, "learning_rate": 2.8731135223737196e-06, "loss": 0.0018, "step": 33152 }, { "epoch": 6.08, "learning_rate": 2.872592741830641e-06, "loss": 0.0016, "step": 33153 }, { "epoch": 6.08, "learning_rate": 2.872072000574002e-06, "loss": 0.0024, "step": 33154 }, { "epoch": 6.08, "learning_rate": 2.8715512986066695e-06, "loss": 0.0022, "step": 33155 }, { "epoch": 6.08, "learning_rate": 2.871030635931513e-06, "loss": 0.0015, "step": 33156 }, { "epoch": 6.08, "learning_rate": 2.8705100125514017e-06, "loss": 0.0015, "step": 33157 }, { "epoch": 6.08, "learning_rate": 2.869989428469208e-06, "loss": 0.0024, "step": 33158 }, { "epoch": 6.08, "learning_rate": 2.869468883687798e-06, "loss": 0.002, "step": 33159 }, { "epoch": 6.08, "learning_rate": 2.8689483782100435e-06, "loss": 0.002, "step": 33160 }, { "epoch": 6.08, "learning_rate": 2.8684279120388114e-06, "loss": 0.0015, "step": 33161 }, { "epoch": 6.08, "learning_rate": 2.8679074851769728e-06, "loss": 0.0022, "step": 33162 }, { "epoch": 6.08, "learning_rate": 2.8673870976273956e-06, "loss": 0.0022, "step": 33163 }, { "epoch": 6.08, "learning_rate": 2.866866749392947e-06, "loss": 0.0015, "step": 33164 }, { "epoch": 6.08, "learning_rate": 2.866346440476496e-06, "loss": 0.0012, "step": 33165 }, { "epoch": 6.08, "learning_rate": 2.8658261708809075e-06, "loss": 0.002, "step": 33166 }, { "epoch": 6.08, "learning_rate": 2.8653059406090555e-06, "loss": 0.002, "step": 33167 }, { "epoch": 6.08, "learning_rate": 2.8647857496638044e-06, "loss": 0.0017, "step": 33168 }, { "epoch": 6.08, "learning_rate": 2.864265598048025e-06, "loss": 0.0021, "step": 33169 }, { "epoch": 6.08, "learning_rate": 2.863745485764573e-06, "loss": 0.0017, "step": 33170 }, { "epoch": 6.08, "learning_rate": 2.8632254128163274e-06, "loss": 0.0021, "step": 33171 }, { "epoch": 6.08, "learning_rate": 2.862705379206151e-06, "loss": 0.0019, "step": 33172 }, { "epoch": 6.08, "learning_rate": 2.8621853849369085e-06, "loss": 0.003, "step": 33173 }, { "epoch": 6.08, "learning_rate": 2.8616654300114678e-06, "loss": 0.0016, "step": 33174 }, { "epoch": 6.08, "learning_rate": 2.861145514432695e-06, "loss": 0.0025, "step": 33175 }, { "epoch": 6.08, "learning_rate": 2.8606256382034546e-06, "loss": 0.0017, "step": 33176 }, { "epoch": 6.08, "learning_rate": 2.8601058013266135e-06, "loss": 0.0022, "step": 33177 }, { "epoch": 6.08, "learning_rate": 2.8595860038050373e-06, "loss": 0.0015, "step": 33178 }, { "epoch": 6.08, "learning_rate": 2.8590662456415852e-06, "loss": 0.0026, "step": 33179 }, { "epoch": 6.08, "learning_rate": 2.8585465268391355e-06, "loss": 0.0021, "step": 33180 }, { "epoch": 6.08, "learning_rate": 2.8580268474005404e-06, "loss": 0.0021, "step": 33181 }, { "epoch": 6.09, "learning_rate": 2.857507207328668e-06, "loss": 0.002, "step": 33182 }, { "epoch": 6.09, "learning_rate": 2.8569876066263847e-06, "loss": 0.0028, "step": 33183 }, { "epoch": 6.09, "learning_rate": 2.856468045296549e-06, "loss": 0.0024, "step": 33184 }, { "epoch": 6.09, "learning_rate": 2.8559485233420327e-06, "loss": 0.0019, "step": 33185 }, { "epoch": 6.09, "learning_rate": 2.855429040765695e-06, "loss": 0.0019, "step": 33186 }, { "epoch": 6.09, "learning_rate": 2.854909597570402e-06, "loss": 0.0026, "step": 33187 }, { "epoch": 6.09, "learning_rate": 2.85439019375901e-06, "loss": 0.0032, "step": 33188 }, { "epoch": 6.09, "learning_rate": 2.8538708293343886e-06, "loss": 0.0028, "step": 33189 }, { "epoch": 6.09, "learning_rate": 2.8533515042994e-06, "loss": 0.002, "step": 33190 }, { "epoch": 6.09, "learning_rate": 2.8528322186569035e-06, "loss": 0.0017, "step": 33191 }, { "epoch": 6.09, "learning_rate": 2.8523129724097643e-06, "loss": 0.0021, "step": 33192 }, { "epoch": 6.09, "learning_rate": 2.851793765560844e-06, "loss": 0.0016, "step": 33193 }, { "epoch": 6.09, "learning_rate": 2.851274598113003e-06, "loss": 0.0019, "step": 33194 }, { "epoch": 6.09, "learning_rate": 2.8507554700691043e-06, "loss": 0.0014, "step": 33195 }, { "epoch": 6.09, "learning_rate": 2.850236381432009e-06, "loss": 0.0027, "step": 33196 }, { "epoch": 6.09, "learning_rate": 2.8497173322045747e-06, "loss": 0.0021, "step": 33197 }, { "epoch": 6.09, "learning_rate": 2.8491983223896722e-06, "loss": 0.001, "step": 33198 }, { "epoch": 6.09, "learning_rate": 2.8486793519901532e-06, "loss": 0.0021, "step": 33199 }, { "epoch": 6.09, "learning_rate": 2.8481604210088807e-06, "loss": 0.0022, "step": 33200 }, { "epoch": 6.09, "learning_rate": 2.847641529448714e-06, "loss": 0.0015, "step": 33201 }, { "epoch": 6.09, "learning_rate": 2.847122677312516e-06, "loss": 0.0015, "step": 33202 }, { "epoch": 6.09, "learning_rate": 2.8466038646031468e-06, "loss": 0.0015, "step": 33203 }, { "epoch": 6.09, "learning_rate": 2.846085091323464e-06, "loss": 0.0016, "step": 33204 }, { "epoch": 6.09, "learning_rate": 2.8455663574763304e-06, "loss": 0.0018, "step": 33205 }, { "epoch": 6.09, "learning_rate": 2.8450476630645973e-06, "loss": 0.003, "step": 33206 }, { "epoch": 6.09, "learning_rate": 2.8445290080911312e-06, "loss": 0.0014, "step": 33207 }, { "epoch": 6.09, "learning_rate": 2.8440103925587904e-06, "loss": 0.002, "step": 33208 }, { "epoch": 6.09, "learning_rate": 2.8434918164704306e-06, "loss": 0.0031, "step": 33209 }, { "epoch": 6.09, "learning_rate": 2.8429732798289122e-06, "loss": 0.0027, "step": 33210 }, { "epoch": 6.09, "learning_rate": 2.842454782637092e-06, "loss": 0.0021, "step": 33211 }, { "epoch": 6.09, "learning_rate": 2.8419363248978294e-06, "loss": 0.0029, "step": 33212 }, { "epoch": 6.09, "learning_rate": 2.841417906613981e-06, "loss": 0.0022, "step": 33213 }, { "epoch": 6.09, "learning_rate": 2.8408995277884054e-06, "loss": 0.003, "step": 33214 }, { "epoch": 6.09, "learning_rate": 2.8403811884239564e-06, "loss": 0.0024, "step": 33215 }, { "epoch": 6.09, "learning_rate": 2.839862888523499e-06, "loss": 0.0039, "step": 33216 }, { "epoch": 6.09, "learning_rate": 2.839344628089884e-06, "loss": 0.0019, "step": 33217 }, { "epoch": 6.09, "learning_rate": 2.8388264071259674e-06, "loss": 0.0009, "step": 33218 }, { "epoch": 6.09, "learning_rate": 2.8383082256346052e-06, "loss": 0.0021, "step": 33219 }, { "epoch": 6.09, "learning_rate": 2.837790083618658e-06, "loss": 0.0025, "step": 33220 }, { "epoch": 6.09, "learning_rate": 2.83727198108098e-06, "loss": 0.0018, "step": 33221 }, { "epoch": 6.09, "learning_rate": 2.8367539180244276e-06, "loss": 0.0028, "step": 33222 }, { "epoch": 6.09, "learning_rate": 2.8362358944518544e-06, "loss": 0.0022, "step": 33223 }, { "epoch": 6.09, "learning_rate": 2.835717910366117e-06, "loss": 0.0026, "step": 33224 }, { "epoch": 6.09, "learning_rate": 2.8351999657700713e-06, "loss": 0.0018, "step": 33225 }, { "epoch": 6.09, "learning_rate": 2.8346820606665703e-06, "loss": 0.0022, "step": 33226 }, { "epoch": 6.09, "learning_rate": 2.8341641950584696e-06, "loss": 0.0043, "step": 33227 }, { "epoch": 6.09, "learning_rate": 2.833646368948624e-06, "loss": 0.0024, "step": 33228 }, { "epoch": 6.09, "learning_rate": 2.833128582339887e-06, "loss": 0.0024, "step": 33229 }, { "epoch": 6.09, "learning_rate": 2.8326108352351133e-06, "loss": 0.0025, "step": 33230 }, { "epoch": 6.09, "learning_rate": 2.8320931276371565e-06, "loss": 0.002, "step": 33231 }, { "epoch": 6.09, "learning_rate": 2.8315754595488678e-06, "loss": 0.0024, "step": 33232 }, { "epoch": 6.09, "learning_rate": 2.831057830973106e-06, "loss": 0.0029, "step": 33233 }, { "epoch": 6.09, "learning_rate": 2.8305402419127246e-06, "loss": 0.0036, "step": 33234 }, { "epoch": 6.09, "learning_rate": 2.83002269237057e-06, "loss": 0.0029, "step": 33235 }, { "epoch": 6.09, "learning_rate": 2.829505182349499e-06, "loss": 0.0025, "step": 33236 }, { "epoch": 6.1, "learning_rate": 2.828987711852359e-06, "loss": 0.0018, "step": 33237 }, { "epoch": 6.1, "learning_rate": 2.8284702808820106e-06, "loss": 0.0015, "step": 33238 }, { "epoch": 6.1, "learning_rate": 2.827952889441302e-06, "loss": 0.0024, "step": 33239 }, { "epoch": 6.1, "learning_rate": 2.8274355375330885e-06, "loss": 0.0032, "step": 33240 }, { "epoch": 6.1, "learning_rate": 2.826918225160211e-06, "loss": 0.0023, "step": 33241 }, { "epoch": 6.1, "learning_rate": 2.826400952325533e-06, "loss": 0.0029, "step": 33242 }, { "epoch": 6.1, "learning_rate": 2.8258837190319e-06, "loss": 0.0021, "step": 33243 }, { "epoch": 6.1, "learning_rate": 2.8253665252821637e-06, "loss": 0.0018, "step": 33244 }, { "epoch": 6.1, "learning_rate": 2.8248493710791757e-06, "loss": 0.0035, "step": 33245 }, { "epoch": 6.1, "learning_rate": 2.8243322564257856e-06, "loss": 0.0039, "step": 33246 }, { "epoch": 6.1, "learning_rate": 2.823815181324845e-06, "loss": 0.0041, "step": 33247 }, { "epoch": 6.1, "learning_rate": 2.823298145779203e-06, "loss": 0.002, "step": 33248 }, { "epoch": 6.1, "learning_rate": 2.8227811497917103e-06, "loss": 0.0023, "step": 33249 }, { "epoch": 6.1, "learning_rate": 2.822264193365212e-06, "loss": 0.0016, "step": 33250 }, { "epoch": 6.1, "learning_rate": 2.8217472765025654e-06, "loss": 0.0032, "step": 33251 }, { "epoch": 6.1, "learning_rate": 2.821230399206618e-06, "loss": 0.0024, "step": 33252 }, { "epoch": 6.1, "learning_rate": 2.820713561480215e-06, "loss": 0.0023, "step": 33253 }, { "epoch": 6.1, "learning_rate": 2.820196763326203e-06, "loss": 0.0021, "step": 33254 }, { "epoch": 6.1, "learning_rate": 2.819680004747438e-06, "loss": 0.0019, "step": 33255 }, { "epoch": 6.1, "learning_rate": 2.8191632857467656e-06, "loss": 0.0022, "step": 33256 }, { "epoch": 6.1, "learning_rate": 2.8186466063270325e-06, "loss": 0.0026, "step": 33257 }, { "epoch": 6.1, "learning_rate": 2.818129966491091e-06, "loss": 0.0016, "step": 33258 }, { "epoch": 6.1, "learning_rate": 2.817613366241779e-06, "loss": 0.0029, "step": 33259 }, { "epoch": 6.1, "learning_rate": 2.8170968055819547e-06, "loss": 0.0017, "step": 33260 }, { "epoch": 6.1, "learning_rate": 2.8165802845144598e-06, "loss": 0.0033, "step": 33261 }, { "epoch": 6.1, "learning_rate": 2.816063803042143e-06, "loss": 0.0021, "step": 33262 }, { "epoch": 6.1, "learning_rate": 2.8155473611678507e-06, "loss": 0.0023, "step": 33263 }, { "epoch": 6.1, "learning_rate": 2.8150309588944304e-06, "loss": 0.0021, "step": 33264 }, { "epoch": 6.1, "learning_rate": 2.8145145962247266e-06, "loss": 0.0016, "step": 33265 }, { "epoch": 6.1, "learning_rate": 2.8139982731615877e-06, "loss": 0.0031, "step": 33266 }, { "epoch": 6.1, "learning_rate": 2.813481989707858e-06, "loss": 0.0027, "step": 33267 }, { "epoch": 6.1, "learning_rate": 2.8129657458663806e-06, "loss": 0.0022, "step": 33268 }, { "epoch": 6.1, "learning_rate": 2.812449541640011e-06, "loss": 0.0024, "step": 33269 }, { "epoch": 6.1, "learning_rate": 2.811933377031585e-06, "loss": 0.0028, "step": 33270 }, { "epoch": 6.1, "learning_rate": 2.811417252043951e-06, "loss": 0.0017, "step": 33271 }, { "epoch": 6.1, "learning_rate": 2.81090116667995e-06, "loss": 0.0019, "step": 33272 }, { "epoch": 6.1, "learning_rate": 2.8103851209424326e-06, "loss": 0.0023, "step": 33273 }, { "epoch": 6.1, "learning_rate": 2.8098691148342414e-06, "loss": 0.0038, "step": 33274 }, { "epoch": 6.1, "learning_rate": 2.8093531483582204e-06, "loss": 0.0018, "step": 33275 }, { "epoch": 6.1, "learning_rate": 2.8088372215172135e-06, "loss": 0.0022, "step": 33276 }, { "epoch": 6.1, "learning_rate": 2.8083213343140637e-06, "loss": 0.0015, "step": 33277 }, { "epoch": 6.1, "learning_rate": 2.8078054867516145e-06, "loss": 0.0028, "step": 33278 }, { "epoch": 6.1, "learning_rate": 2.8072896788327107e-06, "loss": 0.0021, "step": 33279 }, { "epoch": 6.1, "learning_rate": 2.8067739105601954e-06, "loss": 0.0023, "step": 33280 }, { "epoch": 6.1, "learning_rate": 2.8062581819369107e-06, "loss": 0.0031, "step": 33281 }, { "epoch": 6.1, "learning_rate": 2.8057424929656985e-06, "loss": 0.0024, "step": 33282 }, { "epoch": 6.1, "learning_rate": 2.8052268436494023e-06, "loss": 0.0019, "step": 33283 }, { "epoch": 6.1, "learning_rate": 2.8047112339908642e-06, "loss": 0.0025, "step": 33284 }, { "epoch": 6.1, "learning_rate": 2.8041956639929244e-06, "loss": 0.0021, "step": 33285 }, { "epoch": 6.1, "learning_rate": 2.8036801336584285e-06, "loss": 0.0034, "step": 33286 }, { "epoch": 6.1, "learning_rate": 2.8031646429902204e-06, "loss": 0.0021, "step": 33287 }, { "epoch": 6.1, "learning_rate": 2.8026491919911335e-06, "loss": 0.0034, "step": 33288 }, { "epoch": 6.1, "learning_rate": 2.8021337806640137e-06, "loss": 0.0021, "step": 33289 }, { "epoch": 6.1, "learning_rate": 2.8016184090116973e-06, "loss": 0.0025, "step": 33290 }, { "epoch": 6.11, "learning_rate": 2.801103077037033e-06, "loss": 0.0031, "step": 33291 }, { "epoch": 6.11, "learning_rate": 2.8005877847428564e-06, "loss": 0.002, "step": 33292 }, { "epoch": 6.11, "learning_rate": 2.8000725321320075e-06, "loss": 0.004, "step": 33293 }, { "epoch": 6.11, "learning_rate": 2.799557319207329e-06, "loss": 0.0019, "step": 33294 }, { "epoch": 6.11, "learning_rate": 2.7990421459716587e-06, "loss": 0.0016, "step": 33295 }, { "epoch": 6.11, "learning_rate": 2.7985270124278373e-06, "loss": 0.0014, "step": 33296 }, { "epoch": 6.11, "learning_rate": 2.7980119185787035e-06, "loss": 0.0014, "step": 33297 }, { "epoch": 6.11, "learning_rate": 2.7974968644270926e-06, "loss": 0.0034, "step": 33298 }, { "epoch": 6.11, "learning_rate": 2.796981849975855e-06, "loss": 0.0012, "step": 33299 }, { "epoch": 6.11, "learning_rate": 2.796466875227819e-06, "loss": 0.0013, "step": 33300 }, { "epoch": 6.11, "learning_rate": 2.795951940185827e-06, "loss": 0.0024, "step": 33301 }, { "epoch": 6.11, "learning_rate": 2.7954370448527158e-06, "loss": 0.0023, "step": 33302 }, { "epoch": 6.11, "learning_rate": 2.7949221892313218e-06, "loss": 0.0016, "step": 33303 }, { "epoch": 6.11, "learning_rate": 2.794407373324488e-06, "loss": 0.0025, "step": 33304 }, { "epoch": 6.11, "learning_rate": 2.793892597135053e-06, "loss": 0.0026, "step": 33305 }, { "epoch": 6.11, "learning_rate": 2.7933778606658466e-06, "loss": 0.0014, "step": 33306 }, { "epoch": 6.11, "learning_rate": 2.792863163919709e-06, "loss": 0.002, "step": 33307 }, { "epoch": 6.11, "learning_rate": 2.7923485068994803e-06, "loss": 0.0025, "step": 33308 }, { "epoch": 6.11, "learning_rate": 2.7918338896079957e-06, "loss": 0.0016, "step": 33309 }, { "epoch": 6.11, "learning_rate": 2.791319312048092e-06, "loss": 0.0018, "step": 33310 }, { "epoch": 6.11, "learning_rate": 2.790804774222605e-06, "loss": 0.0016, "step": 33311 }, { "epoch": 6.11, "learning_rate": 2.790290276134371e-06, "loss": 0.0031, "step": 33312 }, { "epoch": 6.11, "learning_rate": 2.7897758177862255e-06, "loss": 0.0022, "step": 33313 }, { "epoch": 6.11, "learning_rate": 2.7892613991810046e-06, "loss": 0.0018, "step": 33314 }, { "epoch": 6.11, "learning_rate": 2.788747020321545e-06, "loss": 0.0013, "step": 33315 }, { "epoch": 6.11, "learning_rate": 2.7882326812106797e-06, "loss": 0.0016, "step": 33316 }, { "epoch": 6.11, "learning_rate": 2.7877183818512445e-06, "loss": 0.0022, "step": 33317 }, { "epoch": 6.11, "learning_rate": 2.7872041222460755e-06, "loss": 0.0025, "step": 33318 }, { "epoch": 6.11, "learning_rate": 2.7866899023980055e-06, "loss": 0.0011, "step": 33319 }, { "epoch": 6.11, "learning_rate": 2.786175722309871e-06, "loss": 0.0017, "step": 33320 }, { "epoch": 6.11, "learning_rate": 2.7856615819845e-06, "loss": 0.0021, "step": 33321 }, { "epoch": 6.11, "learning_rate": 2.7851474814247358e-06, "loss": 0.0028, "step": 33322 }, { "epoch": 6.11, "learning_rate": 2.78463342063341e-06, "loss": 0.0024, "step": 33323 }, { "epoch": 6.11, "learning_rate": 2.784119399613352e-06, "loss": 0.0014, "step": 33324 }, { "epoch": 6.11, "learning_rate": 2.7836054183673924e-06, "loss": 0.0022, "step": 33325 }, { "epoch": 6.11, "learning_rate": 2.783091476898373e-06, "loss": 0.0021, "step": 33326 }, { "epoch": 6.11, "learning_rate": 2.782577575209122e-06, "loss": 0.0022, "step": 33327 }, { "epoch": 6.11, "learning_rate": 2.782063713302472e-06, "loss": 0.0014, "step": 33328 }, { "epoch": 6.11, "learning_rate": 2.781549891181257e-06, "loss": 0.0019, "step": 33329 }, { "epoch": 6.11, "learning_rate": 2.781036108848307e-06, "loss": 0.0017, "step": 33330 }, { "epoch": 6.11, "learning_rate": 2.780522366306455e-06, "loss": 0.002, "step": 33331 }, { "epoch": 6.11, "learning_rate": 2.780008663558534e-06, "loss": 0.003, "step": 33332 }, { "epoch": 6.11, "learning_rate": 2.779495000607374e-06, "loss": 0.0012, "step": 33333 }, { "epoch": 6.11, "learning_rate": 2.778981377455806e-06, "loss": 0.0034, "step": 33334 }, { "epoch": 6.11, "learning_rate": 2.778467794106663e-06, "loss": 0.0023, "step": 33335 }, { "epoch": 6.11, "learning_rate": 2.777954250562773e-06, "loss": 0.0017, "step": 33336 }, { "epoch": 6.11, "learning_rate": 2.77744074682697e-06, "loss": 0.0018, "step": 33337 }, { "epoch": 6.11, "learning_rate": 2.7769272829020788e-06, "loss": 0.0019, "step": 33338 }, { "epoch": 6.11, "learning_rate": 2.7764138587909372e-06, "loss": 0.002, "step": 33339 }, { "epoch": 6.11, "learning_rate": 2.7759004744963713e-06, "loss": 0.0036, "step": 33340 }, { "epoch": 6.11, "learning_rate": 2.775387130021214e-06, "loss": 0.0025, "step": 33341 }, { "epoch": 6.11, "learning_rate": 2.7748738253682905e-06, "loss": 0.0021, "step": 33342 }, { "epoch": 6.11, "learning_rate": 2.774360560540428e-06, "loss": 0.0016, "step": 33343 }, { "epoch": 6.11, "learning_rate": 2.7738473355404615e-06, "loss": 0.0018, "step": 33344 }, { "epoch": 6.11, "learning_rate": 2.7733341503712187e-06, "loss": 0.0024, "step": 33345 }, { "epoch": 6.12, "learning_rate": 2.7728210050355265e-06, "loss": 0.0024, "step": 33346 }, { "epoch": 6.12, "learning_rate": 2.772307899536215e-06, "loss": 0.0021, "step": 33347 }, { "epoch": 6.12, "learning_rate": 2.7717948338761112e-06, "loss": 0.0024, "step": 33348 }, { "epoch": 6.12, "learning_rate": 2.771281808058044e-06, "loss": 0.0016, "step": 33349 }, { "epoch": 6.12, "learning_rate": 2.7707688220848405e-06, "loss": 0.0025, "step": 33350 }, { "epoch": 6.12, "learning_rate": 2.7702558759593267e-06, "loss": 0.0025, "step": 33351 }, { "epoch": 6.12, "learning_rate": 2.769742969684337e-06, "loss": 0.002, "step": 33352 }, { "epoch": 6.12, "learning_rate": 2.76923010326269e-06, "loss": 0.0017, "step": 33353 }, { "epoch": 6.12, "learning_rate": 2.7687172766972172e-06, "loss": 0.0014, "step": 33354 }, { "epoch": 6.12, "learning_rate": 2.7682044899907436e-06, "loss": 0.0026, "step": 33355 }, { "epoch": 6.12, "learning_rate": 2.7676917431460924e-06, "loss": 0.0025, "step": 33356 }, { "epoch": 6.12, "learning_rate": 2.767179036166098e-06, "loss": 0.0019, "step": 33357 }, { "epoch": 6.12, "learning_rate": 2.766666369053581e-06, "loss": 0.0038, "step": 33358 }, { "epoch": 6.12, "learning_rate": 2.7661537418113725e-06, "loss": 0.0021, "step": 33359 }, { "epoch": 6.12, "learning_rate": 2.765641154442288e-06, "loss": 0.0033, "step": 33360 }, { "epoch": 6.12, "learning_rate": 2.765128606949161e-06, "loss": 0.0023, "step": 33361 }, { "epoch": 6.12, "learning_rate": 2.764616099334816e-06, "loss": 0.0025, "step": 33362 }, { "epoch": 6.12, "learning_rate": 2.7641036316020753e-06, "loss": 0.0017, "step": 33363 }, { "epoch": 6.12, "learning_rate": 2.7635912037537658e-06, "loss": 0.0018, "step": 33364 }, { "epoch": 6.12, "learning_rate": 2.7630788157927104e-06, "loss": 0.0024, "step": 33365 }, { "epoch": 6.12, "learning_rate": 2.7625664677217344e-06, "loss": 0.0026, "step": 33366 }, { "epoch": 6.12, "learning_rate": 2.762054159543662e-06, "loss": 0.0017, "step": 33367 }, { "epoch": 6.12, "learning_rate": 2.7615418912613158e-06, "loss": 0.0037, "step": 33368 }, { "epoch": 6.12, "learning_rate": 2.761029662877518e-06, "loss": 0.0018, "step": 33369 }, { "epoch": 6.12, "learning_rate": 2.7605174743951004e-06, "loss": 0.0025, "step": 33370 }, { "epoch": 6.12, "learning_rate": 2.760005325816877e-06, "loss": 0.0029, "step": 33371 }, { "epoch": 6.12, "learning_rate": 2.7594932171456734e-06, "loss": 0.0023, "step": 33372 }, { "epoch": 6.12, "learning_rate": 2.75898114838431e-06, "loss": 0.0021, "step": 33373 }, { "epoch": 6.12, "learning_rate": 2.758469119535615e-06, "loss": 0.003, "step": 33374 }, { "epoch": 6.12, "learning_rate": 2.757957130602408e-06, "loss": 0.0029, "step": 33375 }, { "epoch": 6.12, "learning_rate": 2.7574451815875105e-06, "loss": 0.0011, "step": 33376 }, { "epoch": 6.12, "learning_rate": 2.756933272493748e-06, "loss": 0.0018, "step": 33377 }, { "epoch": 6.12, "learning_rate": 2.7564214033239336e-06, "loss": 0.0013, "step": 33378 }, { "epoch": 6.12, "learning_rate": 2.755909574080897e-06, "loss": 0.0021, "step": 33379 }, { "epoch": 6.12, "learning_rate": 2.7553977847674562e-06, "loss": 0.0016, "step": 33380 }, { "epoch": 6.12, "learning_rate": 2.7548860353864325e-06, "loss": 0.0028, "step": 33381 }, { "epoch": 6.12, "learning_rate": 2.7543743259406476e-06, "loss": 0.0019, "step": 33382 }, { "epoch": 6.12, "learning_rate": 2.7538626564329195e-06, "loss": 0.003, "step": 33383 }, { "epoch": 6.12, "learning_rate": 2.753351026866071e-06, "loss": 0.002, "step": 33384 }, { "epoch": 6.12, "learning_rate": 2.7528394372429222e-06, "loss": 0.002, "step": 33385 }, { "epoch": 6.12, "learning_rate": 2.752327887566292e-06, "loss": 0.0026, "step": 33386 }, { "epoch": 6.12, "learning_rate": 2.751816377838996e-06, "loss": 0.0022, "step": 33387 }, { "epoch": 6.12, "learning_rate": 2.7513049080638653e-06, "loss": 0.0018, "step": 33388 }, { "epoch": 6.12, "learning_rate": 2.7507934782437086e-06, "loss": 0.0016, "step": 33389 }, { "epoch": 6.12, "learning_rate": 2.7502820883813473e-06, "loss": 0.0027, "step": 33390 }, { "epoch": 6.12, "learning_rate": 2.749770738479598e-06, "loss": 0.0029, "step": 33391 }, { "epoch": 6.12, "learning_rate": 2.7492594285412865e-06, "loss": 0.0012, "step": 33392 }, { "epoch": 6.12, "learning_rate": 2.7487481585692257e-06, "loss": 0.0011, "step": 33393 }, { "epoch": 6.12, "learning_rate": 2.748236928566238e-06, "loss": 0.0022, "step": 33394 }, { "epoch": 6.12, "learning_rate": 2.747725738535135e-06, "loss": 0.0028, "step": 33395 }, { "epoch": 6.12, "learning_rate": 2.747214588478736e-06, "loss": 0.0025, "step": 33396 }, { "epoch": 6.12, "learning_rate": 2.7467034783998627e-06, "loss": 0.0027, "step": 33397 }, { "epoch": 6.12, "learning_rate": 2.746192408301329e-06, "loss": 0.0032, "step": 33398 }, { "epoch": 6.12, "learning_rate": 2.7456813781859528e-06, "loss": 0.002, "step": 33399 }, { "epoch": 6.13, "learning_rate": 2.7451703880565506e-06, "loss": 0.0014, "step": 33400 }, { "epoch": 6.13, "learning_rate": 2.74465943791594e-06, "loss": 0.0021, "step": 33401 }, { "epoch": 6.13, "learning_rate": 2.7441485277669355e-06, "loss": 0.0015, "step": 33402 }, { "epoch": 6.13, "learning_rate": 2.7436376576123557e-06, "loss": 0.0016, "step": 33403 }, { "epoch": 6.13, "learning_rate": 2.7431268274550105e-06, "loss": 0.0026, "step": 33404 }, { "epoch": 6.13, "learning_rate": 2.742616037297725e-06, "loss": 0.0019, "step": 33405 }, { "epoch": 6.13, "learning_rate": 2.742105287143312e-06, "loss": 0.0023, "step": 33406 }, { "epoch": 6.13, "learning_rate": 2.741594576994582e-06, "loss": 0.0018, "step": 33407 }, { "epoch": 6.13, "learning_rate": 2.7410839068543515e-06, "loss": 0.0033, "step": 33408 }, { "epoch": 6.13, "learning_rate": 2.7405732767254347e-06, "loss": 0.002, "step": 33409 }, { "epoch": 6.13, "learning_rate": 2.7400626866106504e-06, "loss": 0.0016, "step": 33410 }, { "epoch": 6.13, "learning_rate": 2.739552136512812e-06, "loss": 0.0016, "step": 33411 }, { "epoch": 6.13, "learning_rate": 2.739041626434734e-06, "loss": 0.0021, "step": 33412 }, { "epoch": 6.13, "learning_rate": 2.7385311563792226e-06, "loss": 0.0018, "step": 33413 }, { "epoch": 6.13, "learning_rate": 2.738020726349102e-06, "loss": 0.002, "step": 33414 }, { "epoch": 6.13, "learning_rate": 2.73751033634718e-06, "loss": 0.0027, "step": 33415 }, { "epoch": 6.13, "learning_rate": 2.736999986376272e-06, "loss": 0.0036, "step": 33416 }, { "epoch": 6.13, "learning_rate": 2.736489676439189e-06, "loss": 0.0026, "step": 33417 }, { "epoch": 6.13, "learning_rate": 2.7359794065387458e-06, "loss": 0.0028, "step": 33418 }, { "epoch": 6.13, "learning_rate": 2.7354691766777554e-06, "loss": 0.0027, "step": 33419 }, { "epoch": 6.13, "learning_rate": 2.7349589868590275e-06, "loss": 0.002, "step": 33420 }, { "epoch": 6.13, "learning_rate": 2.7344488370853763e-06, "loss": 0.0026, "step": 33421 }, { "epoch": 6.13, "learning_rate": 2.7339387273596106e-06, "loss": 0.0016, "step": 33422 }, { "epoch": 6.13, "learning_rate": 2.733428657684547e-06, "loss": 0.0015, "step": 33423 }, { "epoch": 6.13, "learning_rate": 2.7329186280629993e-06, "loss": 0.0024, "step": 33424 }, { "epoch": 6.13, "learning_rate": 2.73240863849777e-06, "loss": 0.0024, "step": 33425 }, { "epoch": 6.13, "learning_rate": 2.7318986889916722e-06, "loss": 0.0021, "step": 33426 }, { "epoch": 6.13, "learning_rate": 2.7313887795475224e-06, "loss": 0.0015, "step": 33427 }, { "epoch": 6.13, "learning_rate": 2.7308789101681275e-06, "loss": 0.0025, "step": 33428 }, { "epoch": 6.13, "learning_rate": 2.730369080856299e-06, "loss": 0.0021, "step": 33429 }, { "epoch": 6.13, "learning_rate": 2.7298592916148482e-06, "loss": 0.0016, "step": 33430 }, { "epoch": 6.13, "learning_rate": 2.7293495424465776e-06, "loss": 0.0013, "step": 33431 }, { "epoch": 6.13, "learning_rate": 2.7288398333543063e-06, "loss": 0.0026, "step": 33432 }, { "epoch": 6.13, "learning_rate": 2.72833016434084e-06, "loss": 0.0023, "step": 33433 }, { "epoch": 6.13, "learning_rate": 2.727820535408988e-06, "loss": 0.0025, "step": 33434 }, { "epoch": 6.13, "learning_rate": 2.7273109465615598e-06, "loss": 0.0023, "step": 33435 }, { "epoch": 6.13, "learning_rate": 2.726801397801364e-06, "loss": 0.0013, "step": 33436 }, { "epoch": 6.13, "learning_rate": 2.7262918891312094e-06, "loss": 0.0013, "step": 33437 }, { "epoch": 6.13, "learning_rate": 2.725782420553904e-06, "loss": 0.0021, "step": 33438 }, { "epoch": 6.13, "learning_rate": 2.7252729920722564e-06, "loss": 0.0044, "step": 33439 }, { "epoch": 6.13, "learning_rate": 2.7247636036890714e-06, "loss": 0.0015, "step": 33440 }, { "epoch": 6.13, "learning_rate": 2.724254255407166e-06, "loss": 0.0018, "step": 33441 }, { "epoch": 6.13, "learning_rate": 2.7237449472293388e-06, "loss": 0.0027, "step": 33442 }, { "epoch": 6.13, "learning_rate": 2.7232356791584e-06, "loss": 0.0018, "step": 33443 }, { "epoch": 6.13, "learning_rate": 2.722726451197153e-06, "loss": 0.0015, "step": 33444 }, { "epoch": 6.13, "learning_rate": 2.7222172633484113e-06, "loss": 0.0025, "step": 33445 }, { "epoch": 6.13, "learning_rate": 2.7217081156149794e-06, "loss": 0.0026, "step": 33446 }, { "epoch": 6.13, "learning_rate": 2.721199007999662e-06, "loss": 0.0025, "step": 33447 }, { "epoch": 6.13, "learning_rate": 2.72068994050527e-06, "loss": 0.0016, "step": 33448 }, { "epoch": 6.13, "learning_rate": 2.720180913134599e-06, "loss": 0.0016, "step": 33449 }, { "epoch": 6.13, "learning_rate": 2.7196719258904647e-06, "loss": 0.0015, "step": 33450 }, { "epoch": 6.13, "learning_rate": 2.719162978775669e-06, "loss": 0.0016, "step": 33451 }, { "epoch": 6.13, "learning_rate": 2.7186540717930175e-06, "loss": 0.0014, "step": 33452 }, { "epoch": 6.13, "learning_rate": 2.718145204945315e-06, "loss": 0.0025, "step": 33453 }, { "epoch": 6.13, "learning_rate": 2.7176363782353677e-06, "loss": 0.0026, "step": 33454 }, { "epoch": 6.14, "learning_rate": 2.717127591665979e-06, "loss": 0.002, "step": 33455 }, { "epoch": 6.14, "learning_rate": 2.7166188452399535e-06, "loss": 0.0024, "step": 33456 }, { "epoch": 6.14, "learning_rate": 2.716110138960092e-06, "loss": 0.0036, "step": 33457 }, { "epoch": 6.14, "learning_rate": 2.7156014728292056e-06, "loss": 0.0028, "step": 33458 }, { "epoch": 6.14, "learning_rate": 2.715092846850097e-06, "loss": 0.0026, "step": 33459 }, { "epoch": 6.14, "learning_rate": 2.7145842610255645e-06, "loss": 0.002, "step": 33460 }, { "epoch": 6.14, "learning_rate": 2.7140757153584153e-06, "loss": 0.0022, "step": 33461 }, { "epoch": 6.14, "learning_rate": 2.713567209851448e-06, "loss": 0.0027, "step": 33462 }, { "epoch": 6.14, "learning_rate": 2.7130587445074707e-06, "loss": 0.0017, "step": 33463 }, { "epoch": 6.14, "learning_rate": 2.7125503193292857e-06, "loss": 0.0018, "step": 33464 }, { "epoch": 6.14, "learning_rate": 2.7120419343196934e-06, "loss": 0.0019, "step": 33465 }, { "epoch": 6.14, "learning_rate": 2.7115335894814963e-06, "loss": 0.0023, "step": 33466 }, { "epoch": 6.14, "learning_rate": 2.711025284817498e-06, "loss": 0.0022, "step": 33467 }, { "epoch": 6.14, "learning_rate": 2.7105170203304975e-06, "loss": 0.0024, "step": 33468 }, { "epoch": 6.14, "learning_rate": 2.710008796023299e-06, "loss": 0.0011, "step": 33469 }, { "epoch": 6.14, "learning_rate": 2.709500611898703e-06, "loss": 0.0018, "step": 33470 }, { "epoch": 6.14, "learning_rate": 2.70899246795951e-06, "loss": 0.0013, "step": 33471 }, { "epoch": 6.14, "learning_rate": 2.7084843642085213e-06, "loss": 0.0025, "step": 33472 }, { "epoch": 6.14, "learning_rate": 2.707976300648538e-06, "loss": 0.0022, "step": 33473 }, { "epoch": 6.14, "learning_rate": 2.70746827728236e-06, "loss": 0.0017, "step": 33474 }, { "epoch": 6.14, "learning_rate": 2.706960294112785e-06, "loss": 0.0013, "step": 33475 }, { "epoch": 6.14, "learning_rate": 2.706452351142618e-06, "loss": 0.0022, "step": 33476 }, { "epoch": 6.14, "learning_rate": 2.70594444837466e-06, "loss": 0.0021, "step": 33477 }, { "epoch": 6.14, "learning_rate": 2.7054365858117036e-06, "loss": 0.0035, "step": 33478 }, { "epoch": 6.14, "learning_rate": 2.704928763456549e-06, "loss": 0.0027, "step": 33479 }, { "epoch": 6.14, "learning_rate": 2.7044209813120013e-06, "loss": 0.0018, "step": 33480 }, { "epoch": 6.14, "learning_rate": 2.7039132393808556e-06, "loss": 0.0013, "step": 33481 }, { "epoch": 6.14, "learning_rate": 2.703405537665911e-06, "loss": 0.003, "step": 33482 }, { "epoch": 6.14, "learning_rate": 2.7028978761699664e-06, "loss": 0.0019, "step": 33483 }, { "epoch": 6.14, "learning_rate": 2.7023902548958193e-06, "loss": 0.0014, "step": 33484 }, { "epoch": 6.14, "learning_rate": 2.7018826738462677e-06, "loss": 0.0016, "step": 33485 }, { "epoch": 6.14, "learning_rate": 2.701375133024111e-06, "loss": 0.0027, "step": 33486 }, { "epoch": 6.14, "learning_rate": 2.700867632432145e-06, "loss": 0.0023, "step": 33487 }, { "epoch": 6.14, "learning_rate": 2.7003601720731675e-06, "loss": 0.0022, "step": 33488 }, { "epoch": 6.14, "learning_rate": 2.699852751949975e-06, "loss": 0.002, "step": 33489 }, { "epoch": 6.14, "learning_rate": 2.6993453720653664e-06, "loss": 0.0016, "step": 33490 }, { "epoch": 6.14, "learning_rate": 2.6988380324221364e-06, "loss": 0.0023, "step": 33491 }, { "epoch": 6.14, "learning_rate": 2.6983307330230826e-06, "loss": 0.0017, "step": 33492 }, { "epoch": 6.14, "learning_rate": 2.6978234738709974e-06, "loss": 0.0023, "step": 33493 }, { "epoch": 6.14, "learning_rate": 2.6973162549686837e-06, "loss": 0.0016, "step": 33494 }, { "epoch": 6.14, "learning_rate": 2.6968090763189368e-06, "loss": 0.0035, "step": 33495 }, { "epoch": 6.14, "learning_rate": 2.696301937924546e-06, "loss": 0.0031, "step": 33496 }, { "epoch": 6.14, "learning_rate": 2.695794839788307e-06, "loss": 0.0026, "step": 33497 }, { "epoch": 6.14, "learning_rate": 2.6952877819130218e-06, "loss": 0.0014, "step": 33498 }, { "epoch": 6.14, "learning_rate": 2.694780764301481e-06, "loss": 0.0025, "step": 33499 }, { "epoch": 6.14, "learning_rate": 2.6942737869564795e-06, "loss": 0.0012, "step": 33500 }, { "epoch": 6.14, "learning_rate": 2.6937668498808125e-06, "loss": 0.002, "step": 33501 }, { "epoch": 6.14, "learning_rate": 2.693259953077274e-06, "loss": 0.0021, "step": 33502 }, { "epoch": 6.14, "learning_rate": 2.692753096548657e-06, "loss": 0.0018, "step": 33503 }, { "epoch": 6.14, "learning_rate": 2.6922462802977566e-06, "loss": 0.0028, "step": 33504 }, { "epoch": 6.14, "learning_rate": 2.6917395043273665e-06, "loss": 0.0019, "step": 33505 }, { "epoch": 6.14, "learning_rate": 2.6912327686402786e-06, "loss": 0.0026, "step": 33506 }, { "epoch": 6.14, "learning_rate": 2.690726073239287e-06, "loss": 0.0024, "step": 33507 }, { "epoch": 6.14, "learning_rate": 2.6902194181271846e-06, "loss": 0.0017, "step": 33508 }, { "epoch": 6.15, "learning_rate": 2.6897128033067642e-06, "loss": 0.0029, "step": 33509 }, { "epoch": 6.15, "learning_rate": 2.689206228780814e-06, "loss": 0.002, "step": 33510 }, { "epoch": 6.15, "learning_rate": 2.6886996945521348e-06, "loss": 0.0022, "step": 33511 }, { "epoch": 6.15, "learning_rate": 2.6881932006235135e-06, "loss": 0.003, "step": 33512 }, { "epoch": 6.15, "learning_rate": 2.6876867469977465e-06, "loss": 0.0018, "step": 33513 }, { "epoch": 6.15, "learning_rate": 2.687180333677617e-06, "loss": 0.0016, "step": 33514 }, { "epoch": 6.15, "learning_rate": 2.6866739606659187e-06, "loss": 0.0035, "step": 33515 }, { "epoch": 6.15, "learning_rate": 2.686167627965447e-06, "loss": 0.0026, "step": 33516 }, { "epoch": 6.15, "learning_rate": 2.685661335578992e-06, "loss": 0.0012, "step": 33517 }, { "epoch": 6.15, "learning_rate": 2.685155083509342e-06, "loss": 0.002, "step": 33518 }, { "epoch": 6.15, "learning_rate": 2.684648871759289e-06, "loss": 0.0032, "step": 33519 }, { "epoch": 6.15, "learning_rate": 2.6841427003316233e-06, "loss": 0.0033, "step": 33520 }, { "epoch": 6.15, "learning_rate": 2.6836365692291334e-06, "loss": 0.0032, "step": 33521 }, { "epoch": 6.15, "learning_rate": 2.6831304784546106e-06, "loss": 0.0022, "step": 33522 }, { "epoch": 6.15, "learning_rate": 2.6826244280108438e-06, "loss": 0.0021, "step": 33523 }, { "epoch": 6.15, "learning_rate": 2.6821184179006234e-06, "loss": 0.0019, "step": 33524 }, { "epoch": 6.15, "learning_rate": 2.681612448126737e-06, "loss": 0.002, "step": 33525 }, { "epoch": 6.15, "learning_rate": 2.681106518691974e-06, "loss": 0.0028, "step": 33526 }, { "epoch": 6.15, "learning_rate": 2.6806006295991236e-06, "loss": 0.0017, "step": 33527 }, { "epoch": 6.15, "learning_rate": 2.6800947808509713e-06, "loss": 0.0011, "step": 33528 }, { "epoch": 6.15, "learning_rate": 2.679588972450311e-06, "loss": 0.0026, "step": 33529 }, { "epoch": 6.15, "learning_rate": 2.6790832043999283e-06, "loss": 0.0011, "step": 33530 }, { "epoch": 6.15, "learning_rate": 2.678577476702612e-06, "loss": 0.0026, "step": 33531 }, { "epoch": 6.15, "learning_rate": 2.678071789361143e-06, "loss": 0.0021, "step": 33532 }, { "epoch": 6.15, "learning_rate": 2.6775661423783173e-06, "loss": 0.0014, "step": 33533 }, { "epoch": 6.15, "learning_rate": 2.677060535756918e-06, "loss": 0.0022, "step": 33534 }, { "epoch": 6.15, "learning_rate": 2.676554969499733e-06, "loss": 0.0027, "step": 33535 }, { "epoch": 6.15, "learning_rate": 2.6760494436095475e-06, "loss": 0.0016, "step": 33536 }, { "epoch": 6.15, "learning_rate": 2.6755439580891507e-06, "loss": 0.0023, "step": 33537 }, { "epoch": 6.15, "learning_rate": 2.675038512941326e-06, "loss": 0.0016, "step": 33538 }, { "epoch": 6.15, "learning_rate": 2.674533108168861e-06, "loss": 0.0022, "step": 33539 }, { "epoch": 6.15, "learning_rate": 2.674027743774541e-06, "loss": 0.0036, "step": 33540 }, { "epoch": 6.15, "learning_rate": 2.6735224197611485e-06, "loss": 0.0027, "step": 33541 }, { "epoch": 6.15, "learning_rate": 2.673017136131478e-06, "loss": 0.0026, "step": 33542 }, { "epoch": 6.15, "learning_rate": 2.6725118928883065e-06, "loss": 0.0025, "step": 33543 }, { "epoch": 6.15, "learning_rate": 2.6720066900344212e-06, "loss": 0.0037, "step": 33544 }, { "epoch": 6.15, "learning_rate": 2.671501527572605e-06, "loss": 0.0014, "step": 33545 }, { "epoch": 6.15, "learning_rate": 2.6709964055056426e-06, "loss": 0.0036, "step": 33546 }, { "epoch": 6.15, "learning_rate": 2.6704913238363227e-06, "loss": 0.0029, "step": 33547 }, { "epoch": 6.15, "learning_rate": 2.6699862825674293e-06, "loss": 0.0014, "step": 33548 }, { "epoch": 6.15, "learning_rate": 2.669481281701739e-06, "loss": 0.0039, "step": 33549 }, { "epoch": 6.15, "learning_rate": 2.668976321242038e-06, "loss": 0.0022, "step": 33550 }, { "epoch": 6.15, "learning_rate": 2.6684714011911138e-06, "loss": 0.0024, "step": 33551 }, { "epoch": 6.15, "learning_rate": 2.667966521551747e-06, "loss": 0.0015, "step": 33552 }, { "epoch": 6.15, "learning_rate": 2.667461682326721e-06, "loss": 0.0028, "step": 33553 }, { "epoch": 6.15, "learning_rate": 2.6669568835188174e-06, "loss": 0.0028, "step": 33554 }, { "epoch": 6.15, "learning_rate": 2.6664521251308184e-06, "loss": 0.0031, "step": 33555 }, { "epoch": 6.15, "learning_rate": 2.665947407165509e-06, "loss": 0.003, "step": 33556 }, { "epoch": 6.15, "learning_rate": 2.665442729625668e-06, "loss": 0.0014, "step": 33557 }, { "epoch": 6.15, "learning_rate": 2.664938092514079e-06, "loss": 0.0018, "step": 33558 }, { "epoch": 6.15, "learning_rate": 2.6644334958335193e-06, "loss": 0.0035, "step": 33559 }, { "epoch": 6.15, "learning_rate": 2.663928939586781e-06, "loss": 0.0019, "step": 33560 }, { "epoch": 6.15, "learning_rate": 2.6634244237766347e-06, "loss": 0.0019, "step": 33561 }, { "epoch": 6.15, "learning_rate": 2.6629199484058655e-06, "loss": 0.0025, "step": 33562 }, { "epoch": 6.15, "learning_rate": 2.66241551347725e-06, "loss": 0.0028, "step": 33563 }, { "epoch": 6.16, "learning_rate": 2.661911118993574e-06, "loss": 0.0028, "step": 33564 }, { "epoch": 6.16, "learning_rate": 2.661406764957618e-06, "loss": 0.0019, "step": 33565 }, { "epoch": 6.16, "learning_rate": 2.6609024513721616e-06, "loss": 0.002, "step": 33566 }, { "epoch": 6.16, "learning_rate": 2.66039817823998e-06, "loss": 0.0017, "step": 33567 }, { "epoch": 6.16, "learning_rate": 2.6598939455638517e-06, "loss": 0.0031, "step": 33568 }, { "epoch": 6.16, "learning_rate": 2.659389753346564e-06, "loss": 0.002, "step": 33569 }, { "epoch": 6.16, "learning_rate": 2.6588856015908913e-06, "loss": 0.0046, "step": 33570 }, { "epoch": 6.16, "learning_rate": 2.658381490299614e-06, "loss": 0.0036, "step": 33571 }, { "epoch": 6.16, "learning_rate": 2.65787741947551e-06, "loss": 0.0022, "step": 33572 }, { "epoch": 6.16, "learning_rate": 2.657373389121357e-06, "loss": 0.0019, "step": 33573 }, { "epoch": 6.16, "learning_rate": 2.6568693992399343e-06, "loss": 0.0018, "step": 33574 }, { "epoch": 6.16, "learning_rate": 2.6563654498340187e-06, "loss": 0.0015, "step": 33575 }, { "epoch": 6.16, "learning_rate": 2.6558615409063893e-06, "loss": 0.0029, "step": 33576 }, { "epoch": 6.16, "learning_rate": 2.6553576724598205e-06, "loss": 0.0032, "step": 33577 }, { "epoch": 6.16, "learning_rate": 2.6548538444970983e-06, "loss": 0.0025, "step": 33578 }, { "epoch": 6.16, "learning_rate": 2.6543500570209903e-06, "loss": 0.0033, "step": 33579 }, { "epoch": 6.16, "learning_rate": 2.6538463100342773e-06, "loss": 0.002, "step": 33580 }, { "epoch": 6.16, "learning_rate": 2.6533426035397312e-06, "loss": 0.0039, "step": 33581 }, { "epoch": 6.16, "learning_rate": 2.6528389375401376e-06, "loss": 0.002, "step": 33582 }, { "epoch": 6.16, "learning_rate": 2.6523353120382666e-06, "loss": 0.0021, "step": 33583 }, { "epoch": 6.16, "learning_rate": 2.6518317270369e-06, "loss": 0.0038, "step": 33584 }, { "epoch": 6.16, "learning_rate": 2.6513281825388013e-06, "loss": 0.0026, "step": 33585 }, { "epoch": 6.16, "learning_rate": 2.6508246785467585e-06, "loss": 0.0022, "step": 33586 }, { "epoch": 6.16, "learning_rate": 2.6503212150635404e-06, "loss": 0.0019, "step": 33587 }, { "epoch": 6.16, "learning_rate": 2.6498177920919253e-06, "loss": 0.0018, "step": 33588 }, { "epoch": 6.16, "learning_rate": 2.6493144096346867e-06, "loss": 0.0024, "step": 33589 }, { "epoch": 6.16, "learning_rate": 2.6488110676945987e-06, "loss": 0.0012, "step": 33590 }, { "epoch": 6.16, "learning_rate": 2.6483077662744363e-06, "loss": 0.0023, "step": 33591 }, { "epoch": 6.16, "learning_rate": 2.647804505376974e-06, "loss": 0.0023, "step": 33592 }, { "epoch": 6.16, "learning_rate": 2.647301285004986e-06, "loss": 0.0036, "step": 33593 }, { "epoch": 6.16, "learning_rate": 2.6467981051612425e-06, "loss": 0.0015, "step": 33594 }, { "epoch": 6.16, "learning_rate": 2.646294965848525e-06, "loss": 0.003, "step": 33595 }, { "epoch": 6.16, "learning_rate": 2.6457918670695994e-06, "loss": 0.0017, "step": 33596 }, { "epoch": 6.16, "learning_rate": 2.645288808827242e-06, "loss": 0.0022, "step": 33597 }, { "epoch": 6.16, "learning_rate": 2.644785791124225e-06, "loss": 0.0034, "step": 33598 }, { "epoch": 6.16, "learning_rate": 2.644282813963317e-06, "loss": 0.0022, "step": 33599 }, { "epoch": 6.16, "learning_rate": 2.6437798773472976e-06, "loss": 0.0021, "step": 33600 }, { "epoch": 6.16, "learning_rate": 2.643276981278936e-06, "loss": 0.0017, "step": 33601 }, { "epoch": 6.16, "learning_rate": 2.6427741257610074e-06, "loss": 0.0023, "step": 33602 }, { "epoch": 6.16, "learning_rate": 2.642271310796275e-06, "loss": 0.0014, "step": 33603 }, { "epoch": 6.16, "learning_rate": 2.6417685363875168e-06, "loss": 0.0011, "step": 33604 }, { "epoch": 6.16, "learning_rate": 2.6412658025375037e-06, "loss": 0.0034, "step": 33605 }, { "epoch": 6.16, "learning_rate": 2.640763109249006e-06, "loss": 0.0029, "step": 33606 }, { "epoch": 6.16, "learning_rate": 2.6402604565247946e-06, "loss": 0.0018, "step": 33607 }, { "epoch": 6.16, "learning_rate": 2.63975784436764e-06, "loss": 0.0034, "step": 33608 }, { "epoch": 6.16, "learning_rate": 2.639255272780311e-06, "loss": 0.0024, "step": 33609 }, { "epoch": 6.16, "learning_rate": 2.638752741765581e-06, "loss": 0.0025, "step": 33610 }, { "epoch": 6.16, "learning_rate": 2.6382502513262177e-06, "loss": 0.0033, "step": 33611 }, { "epoch": 6.16, "learning_rate": 2.637747801464989e-06, "loss": 0.0027, "step": 33612 }, { "epoch": 6.16, "learning_rate": 2.637245392184673e-06, "loss": 0.0036, "step": 33613 }, { "epoch": 6.16, "learning_rate": 2.6367430234880286e-06, "loss": 0.0023, "step": 33614 }, { "epoch": 6.16, "learning_rate": 2.6362406953778307e-06, "loss": 0.0012, "step": 33615 }, { "epoch": 6.16, "learning_rate": 2.6357384078568425e-06, "loss": 0.0044, "step": 33616 }, { "epoch": 6.16, "learning_rate": 2.635236160927841e-06, "loss": 0.0026, "step": 33617 }, { "epoch": 6.17, "learning_rate": 2.634733954593589e-06, "loss": 0.0022, "step": 33618 }, { "epoch": 6.17, "learning_rate": 2.634231788856857e-06, "loss": 0.0012, "step": 33619 }, { "epoch": 6.17, "learning_rate": 2.6337296637204134e-06, "loss": 0.0021, "step": 33620 }, { "epoch": 6.17, "learning_rate": 2.63322757918702e-06, "loss": 0.0021, "step": 33621 }, { "epoch": 6.17, "learning_rate": 2.6327255352594515e-06, "loss": 0.0016, "step": 33622 }, { "epoch": 6.17, "learning_rate": 2.632223531940472e-06, "loss": 0.0032, "step": 33623 }, { "epoch": 6.17, "learning_rate": 2.631721569232849e-06, "loss": 0.0028, "step": 33624 }, { "epoch": 6.17, "learning_rate": 2.63121964713935e-06, "loss": 0.0034, "step": 33625 }, { "epoch": 6.17, "learning_rate": 2.63071776566274e-06, "loss": 0.0037, "step": 33626 }, { "epoch": 6.17, "learning_rate": 2.6302159248057867e-06, "loss": 0.0019, "step": 33627 }, { "epoch": 6.17, "learning_rate": 2.6297141245712564e-06, "loss": 0.0016, "step": 33628 }, { "epoch": 6.17, "learning_rate": 2.6292123649619106e-06, "loss": 0.002, "step": 33629 }, { "epoch": 6.17, "learning_rate": 2.6287106459805235e-06, "loss": 0.0027, "step": 33630 }, { "epoch": 6.17, "learning_rate": 2.628208967629857e-06, "loss": 0.0022, "step": 33631 }, { "epoch": 6.17, "learning_rate": 2.627707329912673e-06, "loss": 0.0034, "step": 33632 }, { "epoch": 6.17, "learning_rate": 2.6272057328317403e-06, "loss": 0.0021, "step": 33633 }, { "epoch": 6.17, "learning_rate": 2.626704176389818e-06, "loss": 0.0022, "step": 33634 }, { "epoch": 6.17, "learning_rate": 2.6262026605896784e-06, "loss": 0.0022, "step": 33635 }, { "epoch": 6.17, "learning_rate": 2.625701185434082e-06, "loss": 0.0029, "step": 33636 }, { "epoch": 6.17, "learning_rate": 2.6251997509257943e-06, "loss": 0.0027, "step": 33637 }, { "epoch": 6.17, "learning_rate": 2.624698357067578e-06, "loss": 0.0022, "step": 33638 }, { "epoch": 6.17, "learning_rate": 2.6241970038621966e-06, "loss": 0.0023, "step": 33639 }, { "epoch": 6.17, "learning_rate": 2.6236956913124147e-06, "loss": 0.0012, "step": 33640 }, { "epoch": 6.17, "learning_rate": 2.6231944194209948e-06, "loss": 0.0033, "step": 33641 }, { "epoch": 6.17, "learning_rate": 2.622693188190699e-06, "loss": 0.0019, "step": 33642 }, { "epoch": 6.17, "learning_rate": 2.6221919976242927e-06, "loss": 0.0023, "step": 33643 }, { "epoch": 6.17, "learning_rate": 2.621690847724536e-06, "loss": 0.0021, "step": 33644 }, { "epoch": 6.17, "learning_rate": 2.621189738494193e-06, "loss": 0.0028, "step": 33645 }, { "epoch": 6.17, "learning_rate": 2.6206886699360244e-06, "loss": 0.0026, "step": 33646 }, { "epoch": 6.17, "learning_rate": 2.6201876420527907e-06, "loss": 0.0018, "step": 33647 }, { "epoch": 6.17, "learning_rate": 2.6196866548472576e-06, "loss": 0.0037, "step": 33648 }, { "epoch": 6.17, "learning_rate": 2.6191857083221873e-06, "loss": 0.0027, "step": 33649 }, { "epoch": 6.17, "learning_rate": 2.6186848024803356e-06, "loss": 0.0018, "step": 33650 }, { "epoch": 6.17, "learning_rate": 2.618183937324467e-06, "loss": 0.0015, "step": 33651 }, { "epoch": 6.17, "learning_rate": 2.6176831128573376e-06, "loss": 0.0028, "step": 33652 }, { "epoch": 6.17, "learning_rate": 2.6171823290817157e-06, "loss": 0.0026, "step": 33653 }, { "epoch": 6.17, "learning_rate": 2.616681586000357e-06, "loss": 0.0025, "step": 33654 }, { "epoch": 6.17, "learning_rate": 2.616180883616026e-06, "loss": 0.0014, "step": 33655 }, { "epoch": 6.17, "learning_rate": 2.6156802219314723e-06, "loss": 0.0028, "step": 33656 }, { "epoch": 6.17, "learning_rate": 2.6151796009494656e-06, "loss": 0.0028, "step": 33657 }, { "epoch": 6.17, "learning_rate": 2.6146790206727613e-06, "loss": 0.0006, "step": 33658 }, { "epoch": 6.17, "learning_rate": 2.6141784811041203e-06, "loss": 0.002, "step": 33659 }, { "epoch": 6.17, "learning_rate": 2.6136779822463e-06, "loss": 0.0029, "step": 33660 }, { "epoch": 6.17, "learning_rate": 2.6131775241020594e-06, "loss": 0.0014, "step": 33661 }, { "epoch": 6.17, "learning_rate": 2.6126771066741575e-06, "loss": 0.0045, "step": 33662 }, { "epoch": 6.17, "learning_rate": 2.612176729965353e-06, "loss": 0.0016, "step": 33663 }, { "epoch": 6.17, "learning_rate": 2.611676393978403e-06, "loss": 0.0021, "step": 33664 }, { "epoch": 6.17, "learning_rate": 2.611176098716063e-06, "loss": 0.0011, "step": 33665 }, { "epoch": 6.17, "learning_rate": 2.6106758441810964e-06, "loss": 0.0017, "step": 33666 }, { "epoch": 6.17, "learning_rate": 2.6101756303762604e-06, "loss": 0.0014, "step": 33667 }, { "epoch": 6.17, "learning_rate": 2.609675457304307e-06, "loss": 0.0017, "step": 33668 }, { "epoch": 6.17, "learning_rate": 2.609175324967992e-06, "loss": 0.0036, "step": 33669 }, { "epoch": 6.17, "learning_rate": 2.6086752333700792e-06, "loss": 0.0028, "step": 33670 }, { "epoch": 6.17, "learning_rate": 2.6081751825133204e-06, "loss": 0.0023, "step": 33671 }, { "epoch": 6.17, "learning_rate": 2.6076751724004744e-06, "loss": 0.0012, "step": 33672 }, { "epoch": 6.18, "learning_rate": 2.607175203034299e-06, "loss": 0.0018, "step": 33673 }, { "epoch": 6.18, "learning_rate": 2.6066752744175418e-06, "loss": 0.0022, "step": 33674 }, { "epoch": 6.18, "learning_rate": 2.6061753865529647e-06, "loss": 0.0031, "step": 33675 }, { "epoch": 6.18, "learning_rate": 2.605675539443324e-06, "loss": 0.0021, "step": 33676 }, { "epoch": 6.18, "learning_rate": 2.605175733091372e-06, "loss": 0.0026, "step": 33677 }, { "epoch": 6.18, "learning_rate": 2.6046759674998646e-06, "loss": 0.0015, "step": 33678 }, { "epoch": 6.18, "learning_rate": 2.6041762426715565e-06, "loss": 0.0021, "step": 33679 }, { "epoch": 6.18, "learning_rate": 2.6036765586092027e-06, "loss": 0.0016, "step": 33680 }, { "epoch": 6.18, "learning_rate": 2.603176915315557e-06, "loss": 0.0021, "step": 33681 }, { "epoch": 6.18, "learning_rate": 2.6026773127933704e-06, "loss": 0.0026, "step": 33682 }, { "epoch": 6.18, "learning_rate": 2.6021777510454028e-06, "loss": 0.0021, "step": 33683 }, { "epoch": 6.18, "learning_rate": 2.601678230074404e-06, "loss": 0.003, "step": 33684 }, { "epoch": 6.18, "learning_rate": 2.601178749883131e-06, "loss": 0.0042, "step": 33685 }, { "epoch": 6.18, "learning_rate": 2.600679310474332e-06, "loss": 0.0023, "step": 33686 }, { "epoch": 6.18, "learning_rate": 2.600179911850759e-06, "loss": 0.0016, "step": 33687 }, { "epoch": 6.18, "learning_rate": 2.59968055401517e-06, "loss": 0.0012, "step": 33688 }, { "epoch": 6.18, "learning_rate": 2.599181236970315e-06, "loss": 0.0024, "step": 33689 }, { "epoch": 6.18, "learning_rate": 2.598681960718947e-06, "loss": 0.0034, "step": 33690 }, { "epoch": 6.18, "learning_rate": 2.598182725263818e-06, "loss": 0.0032, "step": 33691 }, { "epoch": 6.18, "learning_rate": 2.5976835306076787e-06, "loss": 0.0033, "step": 33692 }, { "epoch": 6.18, "learning_rate": 2.59718437675328e-06, "loss": 0.002, "step": 33693 }, { "epoch": 6.18, "learning_rate": 2.596685263703376e-06, "loss": 0.0018, "step": 33694 }, { "epoch": 6.18, "learning_rate": 2.5961861914607154e-06, "loss": 0.0027, "step": 33695 }, { "epoch": 6.18, "learning_rate": 2.5956871600280508e-06, "loss": 0.0024, "step": 33696 }, { "epoch": 6.18, "learning_rate": 2.5951881694081317e-06, "loss": 0.0022, "step": 33697 }, { "epoch": 6.18, "learning_rate": 2.5946892196037086e-06, "loss": 0.0025, "step": 33698 }, { "epoch": 6.18, "learning_rate": 2.5941903106175314e-06, "loss": 0.0035, "step": 33699 }, { "epoch": 6.18, "learning_rate": 2.5936914424523485e-06, "loss": 0.002, "step": 33700 }, { "epoch": 6.18, "learning_rate": 2.5931926151109156e-06, "loss": 0.0019, "step": 33701 }, { "epoch": 6.18, "learning_rate": 2.5926938285959802e-06, "loss": 0.0017, "step": 33702 }, { "epoch": 6.18, "learning_rate": 2.592195082910287e-06, "loss": 0.0021, "step": 33703 }, { "epoch": 6.18, "learning_rate": 2.5916963780565897e-06, "loss": 0.0027, "step": 33704 }, { "epoch": 6.18, "learning_rate": 2.591197714037631e-06, "loss": 0.0015, "step": 33705 }, { "epoch": 6.18, "learning_rate": 2.590699090856167e-06, "loss": 0.002, "step": 33706 }, { "epoch": 6.18, "learning_rate": 2.590200508514944e-06, "loss": 0.0012, "step": 33707 }, { "epoch": 6.18, "learning_rate": 2.589701967016709e-06, "loss": 0.0025, "step": 33708 }, { "epoch": 6.18, "learning_rate": 2.589203466364211e-06, "loss": 0.0013, "step": 33709 }, { "epoch": 6.18, "learning_rate": 2.588705006560196e-06, "loss": 0.0033, "step": 33710 }, { "epoch": 6.18, "learning_rate": 2.5882065876074123e-06, "loss": 0.0027, "step": 33711 }, { "epoch": 6.18, "learning_rate": 2.5877082095086092e-06, "loss": 0.0025, "step": 33712 }, { "epoch": 6.18, "learning_rate": 2.5872098722665284e-06, "loss": 0.0019, "step": 33713 }, { "epoch": 6.18, "learning_rate": 2.5867115758839258e-06, "loss": 0.0028, "step": 33714 }, { "epoch": 6.18, "learning_rate": 2.5862133203635397e-06, "loss": 0.0028, "step": 33715 }, { "epoch": 6.18, "learning_rate": 2.585715105708121e-06, "loss": 0.0019, "step": 33716 }, { "epoch": 6.18, "learning_rate": 2.585216931920412e-06, "loss": 0.002, "step": 33717 }, { "epoch": 6.18, "learning_rate": 2.5847187990031595e-06, "loss": 0.0032, "step": 33718 }, { "epoch": 6.18, "learning_rate": 2.584220706959114e-06, "loss": 0.0015, "step": 33719 }, { "epoch": 6.18, "learning_rate": 2.58372265579102e-06, "loss": 0.0021, "step": 33720 }, { "epoch": 6.18, "learning_rate": 2.5832246455016176e-06, "loss": 0.0022, "step": 33721 }, { "epoch": 6.18, "learning_rate": 2.582726676093652e-06, "loss": 0.0012, "step": 33722 }, { "epoch": 6.18, "learning_rate": 2.582228747569874e-06, "loss": 0.0026, "step": 33723 }, { "epoch": 6.18, "learning_rate": 2.5817308599330237e-06, "loss": 0.0017, "step": 33724 }, { "epoch": 6.18, "learning_rate": 2.5812330131858477e-06, "loss": 0.0018, "step": 33725 }, { "epoch": 6.18, "learning_rate": 2.5807352073310887e-06, "loss": 0.0027, "step": 33726 }, { "epoch": 6.19, "learning_rate": 2.5802374423714917e-06, "loss": 0.0024, "step": 33727 }, { "epoch": 6.19, "learning_rate": 2.5797397183097984e-06, "loss": 0.0027, "step": 33728 }, { "epoch": 6.19, "learning_rate": 2.5792420351487546e-06, "loss": 0.0021, "step": 33729 }, { "epoch": 6.19, "learning_rate": 2.578744392891103e-06, "loss": 0.0025, "step": 33730 }, { "epoch": 6.19, "learning_rate": 2.5782467915395813e-06, "loss": 0.0033, "step": 33731 }, { "epoch": 6.19, "learning_rate": 2.5777492310969453e-06, "loss": 0.0022, "step": 33732 }, { "epoch": 6.19, "learning_rate": 2.5772517115659245e-06, "loss": 0.0027, "step": 33733 }, { "epoch": 6.19, "learning_rate": 2.5767542329492677e-06, "loss": 0.0023, "step": 33734 }, { "epoch": 6.19, "learning_rate": 2.576256795249712e-06, "loss": 0.0034, "step": 33735 }, { "epoch": 6.19, "learning_rate": 2.5757593984700057e-06, "loss": 0.0022, "step": 33736 }, { "epoch": 6.19, "learning_rate": 2.575262042612887e-06, "loss": 0.0027, "step": 33737 }, { "epoch": 6.19, "learning_rate": 2.5747647276811017e-06, "loss": 0.0023, "step": 33738 }, { "epoch": 6.19, "learning_rate": 2.5742674536773837e-06, "loss": 0.0021, "step": 33739 }, { "epoch": 6.19, "learning_rate": 2.5737702206044755e-06, "loss": 0.0028, "step": 33740 }, { "epoch": 6.19, "learning_rate": 2.573273028465122e-06, "loss": 0.0015, "step": 33741 }, { "epoch": 6.19, "learning_rate": 2.5727758772620616e-06, "loss": 0.0029, "step": 33742 }, { "epoch": 6.19, "learning_rate": 2.572278766998034e-06, "loss": 0.0024, "step": 33743 }, { "epoch": 6.19, "learning_rate": 2.5717816976757803e-06, "loss": 0.0019, "step": 33744 }, { "epoch": 6.19, "learning_rate": 2.5712846692980396e-06, "loss": 0.0023, "step": 33745 }, { "epoch": 6.19, "learning_rate": 2.5707876818675524e-06, "loss": 0.0019, "step": 33746 }, { "epoch": 6.19, "learning_rate": 2.5702907353870578e-06, "loss": 0.0022, "step": 33747 }, { "epoch": 6.19, "learning_rate": 2.569793829859294e-06, "loss": 0.0027, "step": 33748 }, { "epoch": 6.19, "learning_rate": 2.5692969652870015e-06, "loss": 0.0025, "step": 33749 }, { "epoch": 6.19, "learning_rate": 2.5688001416729167e-06, "loss": 0.0028, "step": 33750 }, { "epoch": 6.19, "learning_rate": 2.5683033590197817e-06, "loss": 0.0019, "step": 33751 }, { "epoch": 6.19, "learning_rate": 2.5678066173303305e-06, "loss": 0.0018, "step": 33752 }, { "epoch": 6.19, "learning_rate": 2.5673099166073022e-06, "loss": 0.0018, "step": 33753 }, { "epoch": 6.19, "learning_rate": 2.5668132568534377e-06, "loss": 0.0021, "step": 33754 }, { "epoch": 6.19, "learning_rate": 2.5663166380714734e-06, "loss": 0.0027, "step": 33755 }, { "epoch": 6.19, "learning_rate": 2.565820060264148e-06, "loss": 0.0027, "step": 33756 }, { "epoch": 6.19, "learning_rate": 2.565323523434191e-06, "loss": 0.0019, "step": 33757 }, { "epoch": 6.19, "learning_rate": 2.564827027584348e-06, "loss": 0.0021, "step": 33758 }, { "epoch": 6.19, "learning_rate": 2.5643305727173527e-06, "loss": 0.0021, "step": 33759 }, { "epoch": 6.19, "learning_rate": 2.563834158835942e-06, "loss": 0.0021, "step": 33760 }, { "epoch": 6.19, "learning_rate": 2.56333778594285e-06, "loss": 0.0016, "step": 33761 }, { "epoch": 6.19, "learning_rate": 2.562841454040815e-06, "loss": 0.0034, "step": 33762 }, { "epoch": 6.19, "learning_rate": 2.5623451631325734e-06, "loss": 0.0024, "step": 33763 }, { "epoch": 6.19, "learning_rate": 2.561848913220858e-06, "loss": 0.0022, "step": 33764 }, { "epoch": 6.19, "learning_rate": 2.5613527043084065e-06, "loss": 0.003, "step": 33765 }, { "epoch": 6.19, "learning_rate": 2.56085653639795e-06, "loss": 0.002, "step": 33766 }, { "epoch": 6.19, "learning_rate": 2.560360409492233e-06, "loss": 0.0007, "step": 33767 }, { "epoch": 6.19, "learning_rate": 2.559864323593979e-06, "loss": 0.0014, "step": 33768 }, { "epoch": 6.19, "learning_rate": 2.5593682787059295e-06, "loss": 0.0024, "step": 33769 }, { "epoch": 6.19, "learning_rate": 2.558872274830815e-06, "loss": 0.0023, "step": 33770 }, { "epoch": 6.19, "learning_rate": 2.5583763119713677e-06, "loss": 0.0032, "step": 33771 }, { "epoch": 6.19, "learning_rate": 2.5578803901303282e-06, "loss": 0.0026, "step": 33772 }, { "epoch": 6.19, "learning_rate": 2.5573845093104256e-06, "loss": 0.0027, "step": 33773 }, { "epoch": 6.19, "learning_rate": 2.5568886695143967e-06, "loss": 0.0027, "step": 33774 }, { "epoch": 6.19, "learning_rate": 2.556392870744967e-06, "loss": 0.0017, "step": 33775 }, { "epoch": 6.19, "learning_rate": 2.555897113004877e-06, "loss": 0.0022, "step": 33776 }, { "epoch": 6.19, "learning_rate": 2.555401396296856e-06, "loss": 0.0025, "step": 33777 }, { "epoch": 6.19, "learning_rate": 2.554905720623637e-06, "loss": 0.0024, "step": 33778 }, { "epoch": 6.19, "learning_rate": 2.554410085987952e-06, "loss": 0.0013, "step": 33779 }, { "epoch": 6.19, "learning_rate": 2.553914492392533e-06, "loss": 0.0017, "step": 33780 }, { "epoch": 6.19, "learning_rate": 2.553418939840112e-06, "loss": 0.0023, "step": 33781 }, { "epoch": 6.2, "learning_rate": 2.55292342833342e-06, "loss": 0.0018, "step": 33782 }, { "epoch": 6.2, "learning_rate": 2.55242795787519e-06, "loss": 0.0031, "step": 33783 }, { "epoch": 6.2, "learning_rate": 2.551932528468146e-06, "loss": 0.0014, "step": 33784 }, { "epoch": 6.2, "learning_rate": 2.551437140115033e-06, "loss": 0.0018, "step": 33785 }, { "epoch": 6.2, "learning_rate": 2.5509417928185675e-06, "loss": 0.0023, "step": 33786 }, { "epoch": 6.2, "learning_rate": 2.550446486581487e-06, "loss": 0.0037, "step": 33787 }, { "epoch": 6.2, "learning_rate": 2.549951221406517e-06, "loss": 0.0021, "step": 33788 }, { "epoch": 6.2, "learning_rate": 2.5494559972963928e-06, "loss": 0.0017, "step": 33789 }, { "epoch": 6.2, "learning_rate": 2.5489608142538414e-06, "loss": 0.0024, "step": 33790 }, { "epoch": 6.2, "learning_rate": 2.548465672281594e-06, "loss": 0.0023, "step": 33791 }, { "epoch": 6.2, "learning_rate": 2.54797057138238e-06, "loss": 0.0012, "step": 33792 }, { "epoch": 6.2, "learning_rate": 2.547475511558921e-06, "loss": 0.0019, "step": 33793 }, { "epoch": 6.2, "learning_rate": 2.546980492813954e-06, "loss": 0.002, "step": 33794 }, { "epoch": 6.2, "learning_rate": 2.5464855151502066e-06, "loss": 0.0025, "step": 33795 }, { "epoch": 6.2, "learning_rate": 2.545990578570404e-06, "loss": 0.0023, "step": 33796 }, { "epoch": 6.2, "learning_rate": 2.545495683077277e-06, "loss": 0.0027, "step": 33797 }, { "epoch": 6.2, "learning_rate": 2.545000828673553e-06, "loss": 0.0021, "step": 33798 }, { "epoch": 6.2, "learning_rate": 2.544506015361958e-06, "loss": 0.0013, "step": 33799 }, { "epoch": 6.2, "learning_rate": 2.544011243145221e-06, "loss": 0.003, "step": 33800 }, { "epoch": 6.2, "learning_rate": 2.5435165120260683e-06, "loss": 0.0015, "step": 33801 }, { "epoch": 6.2, "learning_rate": 2.5430218220072254e-06, "loss": 0.0021, "step": 33802 }, { "epoch": 6.2, "learning_rate": 2.5425271730914258e-06, "loss": 0.0013, "step": 33803 }, { "epoch": 6.2, "learning_rate": 2.542032565281388e-06, "loss": 0.005, "step": 33804 }, { "epoch": 6.2, "learning_rate": 2.541537998579843e-06, "loss": 0.0018, "step": 33805 }, { "epoch": 6.2, "learning_rate": 2.5410434729895107e-06, "loss": 0.0016, "step": 33806 }, { "epoch": 6.2, "learning_rate": 2.5405489885131253e-06, "loss": 0.0009, "step": 33807 }, { "epoch": 6.2, "learning_rate": 2.5400545451534076e-06, "loss": 0.003, "step": 33808 }, { "epoch": 6.2, "learning_rate": 2.539560142913088e-06, "loss": 0.0029, "step": 33809 }, { "epoch": 6.2, "learning_rate": 2.539065781794882e-06, "loss": 0.0021, "step": 33810 }, { "epoch": 6.2, "learning_rate": 2.538571461801522e-06, "loss": 0.0017, "step": 33811 }, { "epoch": 6.2, "learning_rate": 2.5380771829357308e-06, "loss": 0.0015, "step": 33812 }, { "epoch": 6.2, "learning_rate": 2.5375829452002343e-06, "loss": 0.0013, "step": 33813 }, { "epoch": 6.2, "learning_rate": 2.5370887485977546e-06, "loss": 0.0029, "step": 33814 }, { "epoch": 6.2, "learning_rate": 2.5365945931310167e-06, "loss": 0.0012, "step": 33815 }, { "epoch": 6.2, "learning_rate": 2.5361004788027446e-06, "loss": 0.0036, "step": 33816 }, { "epoch": 6.2, "learning_rate": 2.5356064056156614e-06, "loss": 0.0011, "step": 33817 }, { "epoch": 6.2, "learning_rate": 2.5351123735724905e-06, "loss": 0.0039, "step": 33818 }, { "epoch": 6.2, "learning_rate": 2.534618382675952e-06, "loss": 0.0017, "step": 33819 }, { "epoch": 6.2, "learning_rate": 2.534124432928774e-06, "loss": 0.0034, "step": 33820 }, { "epoch": 6.2, "learning_rate": 2.5336305243336814e-06, "loss": 0.0029, "step": 33821 }, { "epoch": 6.2, "learning_rate": 2.533136656893389e-06, "loss": 0.0026, "step": 33822 }, { "epoch": 6.2, "learning_rate": 2.532642830610622e-06, "loss": 0.0033, "step": 33823 }, { "epoch": 6.2, "learning_rate": 2.532149045488099e-06, "loss": 0.0014, "step": 33824 }, { "epoch": 6.2, "learning_rate": 2.5316553015285496e-06, "loss": 0.0038, "step": 33825 }, { "epoch": 6.2, "learning_rate": 2.5311615987346904e-06, "loss": 0.0027, "step": 33826 }, { "epoch": 6.2, "learning_rate": 2.5306679371092467e-06, "loss": 0.0019, "step": 33827 }, { "epoch": 6.2, "learning_rate": 2.530174316654931e-06, "loss": 0.0028, "step": 33828 }, { "epoch": 6.2, "learning_rate": 2.5296807373744716e-06, "loss": 0.0024, "step": 33829 }, { "epoch": 6.2, "learning_rate": 2.5291871992705865e-06, "loss": 0.0032, "step": 33830 }, { "epoch": 6.2, "learning_rate": 2.528693702345998e-06, "loss": 0.0021, "step": 33831 }, { "epoch": 6.2, "learning_rate": 2.5282002466034238e-06, "loss": 0.0026, "step": 33832 }, { "epoch": 6.2, "learning_rate": 2.5277068320455845e-06, "loss": 0.0012, "step": 33833 }, { "epoch": 6.2, "learning_rate": 2.5272134586752018e-06, "loss": 0.0015, "step": 33834 }, { "epoch": 6.2, "learning_rate": 2.5267201264949923e-06, "loss": 0.0016, "step": 33835 }, { "epoch": 6.21, "learning_rate": 2.526226835507677e-06, "loss": 0.0021, "step": 33836 }, { "epoch": 6.21, "learning_rate": 2.5257335857159715e-06, "loss": 0.0029, "step": 33837 }, { "epoch": 6.21, "learning_rate": 2.5252403771225998e-06, "loss": 0.0017, "step": 33838 }, { "epoch": 6.21, "learning_rate": 2.5247472097302815e-06, "loss": 0.0017, "step": 33839 }, { "epoch": 6.21, "learning_rate": 2.5242540835417293e-06, "loss": 0.0027, "step": 33840 }, { "epoch": 6.21, "learning_rate": 2.523760998559661e-06, "loss": 0.0017, "step": 33841 }, { "epoch": 6.21, "learning_rate": 2.5232679547868e-06, "loss": 0.0019, "step": 33842 }, { "epoch": 6.21, "learning_rate": 2.5227749522258605e-06, "loss": 0.0051, "step": 33843 }, { "epoch": 6.21, "learning_rate": 2.5222819908795616e-06, "loss": 0.0024, "step": 33844 }, { "epoch": 6.21, "learning_rate": 2.521789070750622e-06, "loss": 0.0022, "step": 33845 }, { "epoch": 6.21, "learning_rate": 2.52129619184175e-06, "loss": 0.0024, "step": 33846 }, { "epoch": 6.21, "learning_rate": 2.520803354155672e-06, "loss": 0.0041, "step": 33847 }, { "epoch": 6.21, "learning_rate": 2.5203105576951014e-06, "loss": 0.0023, "step": 33848 }, { "epoch": 6.21, "learning_rate": 2.5198178024627527e-06, "loss": 0.0025, "step": 33849 }, { "epoch": 6.21, "learning_rate": 2.519325088461345e-06, "loss": 0.0024, "step": 33850 }, { "epoch": 6.21, "learning_rate": 2.518832415693593e-06, "loss": 0.0018, "step": 33851 }, { "epoch": 6.21, "learning_rate": 2.5183397841622115e-06, "loss": 0.002, "step": 33852 }, { "epoch": 6.21, "learning_rate": 2.517847193869916e-06, "loss": 0.0018, "step": 33853 }, { "epoch": 6.21, "learning_rate": 2.517354644819422e-06, "loss": 0.0027, "step": 33854 }, { "epoch": 6.21, "learning_rate": 2.5168621370134417e-06, "loss": 0.0015, "step": 33855 }, { "epoch": 6.21, "learning_rate": 2.516369670454699e-06, "loss": 0.002, "step": 33856 }, { "epoch": 6.21, "learning_rate": 2.515877245145898e-06, "loss": 0.0029, "step": 33857 }, { "epoch": 6.21, "learning_rate": 2.515384861089758e-06, "loss": 0.0027, "step": 33858 }, { "epoch": 6.21, "learning_rate": 2.514892518288988e-06, "loss": 0.0025, "step": 33859 }, { "epoch": 6.21, "learning_rate": 2.51440021674631e-06, "loss": 0.0018, "step": 33860 }, { "epoch": 6.21, "learning_rate": 2.513907956464433e-06, "loss": 0.0023, "step": 33861 }, { "epoch": 6.21, "learning_rate": 2.5134157374460697e-06, "loss": 0.0027, "step": 33862 }, { "epoch": 6.21, "learning_rate": 2.512923559693936e-06, "loss": 0.0025, "step": 33863 }, { "epoch": 6.21, "learning_rate": 2.512431423210742e-06, "loss": 0.0014, "step": 33864 }, { "epoch": 6.21, "learning_rate": 2.5119393279992012e-06, "loss": 0.0025, "step": 33865 }, { "epoch": 6.21, "learning_rate": 2.5114472740620265e-06, "loss": 0.0027, "step": 33866 }, { "epoch": 6.21, "learning_rate": 2.5109552614019307e-06, "loss": 0.0036, "step": 33867 }, { "epoch": 6.21, "learning_rate": 2.5104632900216243e-06, "loss": 0.0027, "step": 33868 }, { "epoch": 6.21, "learning_rate": 2.5099713599238195e-06, "loss": 0.0029, "step": 33869 }, { "epoch": 6.21, "learning_rate": 2.509479471111228e-06, "loss": 0.0021, "step": 33870 }, { "epoch": 6.21, "learning_rate": 2.5089876235865607e-06, "loss": 0.0039, "step": 33871 }, { "epoch": 6.21, "learning_rate": 2.5084958173525264e-06, "loss": 0.002, "step": 33872 }, { "epoch": 6.21, "learning_rate": 2.508004052411842e-06, "loss": 0.002, "step": 33873 }, { "epoch": 6.21, "learning_rate": 2.5075123287672174e-06, "loss": 0.0035, "step": 33874 }, { "epoch": 6.21, "learning_rate": 2.5070206464213587e-06, "loss": 0.0029, "step": 33875 }, { "epoch": 6.21, "learning_rate": 2.5065290053769764e-06, "loss": 0.0013, "step": 33876 }, { "epoch": 6.21, "learning_rate": 2.5060374056367798e-06, "loss": 0.0034, "step": 33877 }, { "epoch": 6.21, "learning_rate": 2.5055458472034823e-06, "loss": 0.0023, "step": 33878 }, { "epoch": 6.21, "learning_rate": 2.5050543300797923e-06, "loss": 0.0013, "step": 33879 }, { "epoch": 6.21, "learning_rate": 2.5045628542684187e-06, "loss": 0.0016, "step": 33880 }, { "epoch": 6.21, "learning_rate": 2.5040714197720693e-06, "loss": 0.0023, "step": 33881 }, { "epoch": 6.21, "learning_rate": 2.5035800265934552e-06, "loss": 0.0029, "step": 33882 }, { "epoch": 6.21, "learning_rate": 2.503088674735282e-06, "loss": 0.0023, "step": 33883 }, { "epoch": 6.21, "learning_rate": 2.502597364200261e-06, "loss": 0.0033, "step": 33884 }, { "epoch": 6.21, "learning_rate": 2.502106094991096e-06, "loss": 0.002, "step": 33885 }, { "epoch": 6.21, "learning_rate": 2.5016148671105032e-06, "loss": 0.0021, "step": 33886 }, { "epoch": 6.21, "learning_rate": 2.5011236805611818e-06, "loss": 0.0014, "step": 33887 }, { "epoch": 6.21, "learning_rate": 2.500632535345843e-06, "loss": 0.0025, "step": 33888 }, { "epoch": 6.21, "learning_rate": 2.5001414314671925e-06, "loss": 0.0026, "step": 33889 }, { "epoch": 6.21, "learning_rate": 2.499650368927936e-06, "loss": 0.0014, "step": 33890 }, { "epoch": 6.22, "learning_rate": 2.499159347730785e-06, "loss": 0.0025, "step": 33891 }, { "epoch": 6.22, "learning_rate": 2.498668367878446e-06, "loss": 0.0019, "step": 33892 }, { "epoch": 6.22, "learning_rate": 2.4981774293736206e-06, "loss": 0.0018, "step": 33893 }, { "epoch": 6.22, "learning_rate": 2.4976865322190136e-06, "loss": 0.0016, "step": 33894 }, { "epoch": 6.22, "learning_rate": 2.4971956764173365e-06, "loss": 0.0028, "step": 33895 }, { "epoch": 6.22, "learning_rate": 2.4967048619712943e-06, "loss": 0.0027, "step": 33896 }, { "epoch": 6.22, "learning_rate": 2.4962140888835885e-06, "loss": 0.003, "step": 33897 }, { "epoch": 6.22, "learning_rate": 2.4957233571569285e-06, "loss": 0.0019, "step": 33898 }, { "epoch": 6.22, "learning_rate": 2.495232666794015e-06, "loss": 0.0024, "step": 33899 }, { "epoch": 6.22, "learning_rate": 2.4947420177975568e-06, "loss": 0.0019, "step": 33900 }, { "epoch": 6.22, "learning_rate": 2.4942514101702554e-06, "loss": 0.0023, "step": 33901 }, { "epoch": 6.22, "learning_rate": 2.4937608439148164e-06, "loss": 0.0018, "step": 33902 }, { "epoch": 6.22, "learning_rate": 2.4932703190339435e-06, "loss": 0.0013, "step": 33903 }, { "epoch": 6.22, "learning_rate": 2.49277983553034e-06, "loss": 0.0035, "step": 33904 }, { "epoch": 6.22, "learning_rate": 2.4922893934067107e-06, "loss": 0.0036, "step": 33905 }, { "epoch": 6.22, "learning_rate": 2.491798992665757e-06, "loss": 0.0022, "step": 33906 }, { "epoch": 6.22, "learning_rate": 2.4913086333101844e-06, "loss": 0.0019, "step": 33907 }, { "epoch": 6.22, "learning_rate": 2.4908183153426913e-06, "loss": 0.0022, "step": 33908 }, { "epoch": 6.22, "learning_rate": 2.4903280387659854e-06, "loss": 0.0014, "step": 33909 }, { "epoch": 6.22, "learning_rate": 2.489837803582772e-06, "loss": 0.0024, "step": 33910 }, { "epoch": 6.22, "learning_rate": 2.489347609795745e-06, "loss": 0.0014, "step": 33911 }, { "epoch": 6.22, "learning_rate": 2.4888574574076064e-06, "loss": 0.0033, "step": 33912 }, { "epoch": 6.22, "learning_rate": 2.4883673464210644e-06, "loss": 0.0017, "step": 33913 }, { "epoch": 6.22, "learning_rate": 2.487877276838818e-06, "loss": 0.0018, "step": 33914 }, { "epoch": 6.22, "learning_rate": 2.4873872486635673e-06, "loss": 0.0023, "step": 33915 }, { "epoch": 6.22, "learning_rate": 2.486897261898016e-06, "loss": 0.0013, "step": 33916 }, { "epoch": 6.22, "learning_rate": 2.486407316544861e-06, "loss": 0.0021, "step": 33917 }, { "epoch": 6.22, "learning_rate": 2.4859174126068053e-06, "loss": 0.0028, "step": 33918 }, { "epoch": 6.22, "learning_rate": 2.48542755008655e-06, "loss": 0.002, "step": 33919 }, { "epoch": 6.22, "learning_rate": 2.4849377289867927e-06, "loss": 0.0019, "step": 33920 }, { "epoch": 6.22, "learning_rate": 2.4844479493102356e-06, "loss": 0.0026, "step": 33921 }, { "epoch": 6.22, "learning_rate": 2.4839582110595773e-06, "loss": 0.002, "step": 33922 }, { "epoch": 6.22, "learning_rate": 2.483468514237517e-06, "loss": 0.0033, "step": 33923 }, { "epoch": 6.22, "learning_rate": 2.482978858846755e-06, "loss": 0.002, "step": 33924 }, { "epoch": 6.22, "learning_rate": 2.4824892448899863e-06, "loss": 0.0019, "step": 33925 }, { "epoch": 6.22, "learning_rate": 2.4819996723699156e-06, "loss": 0.0026, "step": 33926 }, { "epoch": 6.22, "learning_rate": 2.4815101412892394e-06, "loss": 0.0025, "step": 33927 }, { "epoch": 6.22, "learning_rate": 2.4810206516506574e-06, "loss": 0.0015, "step": 33928 }, { "epoch": 6.22, "learning_rate": 2.480531203456864e-06, "loss": 0.0031, "step": 33929 }, { "epoch": 6.22, "learning_rate": 2.480041796710555e-06, "loss": 0.0019, "step": 33930 }, { "epoch": 6.22, "learning_rate": 2.4795524314144357e-06, "loss": 0.0022, "step": 33931 }, { "epoch": 6.22, "learning_rate": 2.479063107571199e-06, "loss": 0.0022, "step": 33932 }, { "epoch": 6.22, "learning_rate": 2.4785738251835424e-06, "loss": 0.0019, "step": 33933 }, { "epoch": 6.22, "learning_rate": 2.4780845842541633e-06, "loss": 0.0024, "step": 33934 }, { "epoch": 6.22, "learning_rate": 2.477595384785758e-06, "loss": 0.0021, "step": 33935 }, { "epoch": 6.22, "learning_rate": 2.477106226781024e-06, "loss": 0.0014, "step": 33936 }, { "epoch": 6.22, "learning_rate": 2.476617110242656e-06, "loss": 0.002, "step": 33937 }, { "epoch": 6.22, "learning_rate": 2.4761280351733475e-06, "loss": 0.0021, "step": 33938 }, { "epoch": 6.22, "learning_rate": 2.4756390015758037e-06, "loss": 0.0018, "step": 33939 }, { "epoch": 6.22, "learning_rate": 2.4751500094527114e-06, "loss": 0.0018, "step": 33940 }, { "epoch": 6.22, "learning_rate": 2.4746610588067686e-06, "loss": 0.0034, "step": 33941 }, { "epoch": 6.22, "learning_rate": 2.47417214964067e-06, "loss": 0.0019, "step": 33942 }, { "epoch": 6.22, "learning_rate": 2.4736832819571078e-06, "loss": 0.0019, "step": 33943 }, { "epoch": 6.22, "learning_rate": 2.473194455758783e-06, "loss": 0.0023, "step": 33944 }, { "epoch": 6.23, "learning_rate": 2.4727056710483867e-06, "loss": 0.0022, "step": 33945 }, { "epoch": 6.23, "learning_rate": 2.472216927828616e-06, "loss": 0.0018, "step": 33946 }, { "epoch": 6.23, "learning_rate": 2.471728226102155e-06, "loss": 0.0013, "step": 33947 }, { "epoch": 6.23, "learning_rate": 2.4712395658717083e-06, "loss": 0.0025, "step": 33948 }, { "epoch": 6.23, "learning_rate": 2.4707509471399647e-06, "loss": 0.0028, "step": 33949 }, { "epoch": 6.23, "learning_rate": 2.4702623699096194e-06, "loss": 0.0023, "step": 33950 }, { "epoch": 6.23, "learning_rate": 2.4697738341833642e-06, "loss": 0.0021, "step": 33951 }, { "epoch": 6.23, "learning_rate": 2.469285339963892e-06, "loss": 0.0025, "step": 33952 }, { "epoch": 6.23, "learning_rate": 2.468796887253895e-06, "loss": 0.0018, "step": 33953 }, { "epoch": 6.23, "learning_rate": 2.468308476056066e-06, "loss": 0.0021, "step": 33954 }, { "epoch": 6.23, "learning_rate": 2.4678201063730966e-06, "loss": 0.0022, "step": 33955 }, { "epoch": 6.23, "learning_rate": 2.4673317782076776e-06, "loss": 0.0014, "step": 33956 }, { "epoch": 6.23, "learning_rate": 2.466843491562507e-06, "loss": 0.0017, "step": 33957 }, { "epoch": 6.23, "learning_rate": 2.4663552464402694e-06, "loss": 0.0015, "step": 33958 }, { "epoch": 6.23, "learning_rate": 2.465867042843658e-06, "loss": 0.0027, "step": 33959 }, { "epoch": 6.23, "learning_rate": 2.4653788807753644e-06, "loss": 0.0023, "step": 33960 }, { "epoch": 6.23, "learning_rate": 2.4648907602380745e-06, "loss": 0.0027, "step": 33961 }, { "epoch": 6.23, "learning_rate": 2.464402681234488e-06, "loss": 0.0021, "step": 33962 }, { "epoch": 6.23, "learning_rate": 2.463914643767292e-06, "loss": 0.0023, "step": 33963 }, { "epoch": 6.23, "learning_rate": 2.463426647839173e-06, "loss": 0.0023, "step": 33964 }, { "epoch": 6.23, "learning_rate": 2.46293869345282e-06, "loss": 0.0015, "step": 33965 }, { "epoch": 6.23, "learning_rate": 2.4624507806109275e-06, "loss": 0.0028, "step": 33966 }, { "epoch": 6.23, "learning_rate": 2.4619629093161833e-06, "loss": 0.0022, "step": 33967 }, { "epoch": 6.23, "learning_rate": 2.4614750795712763e-06, "loss": 0.0027, "step": 33968 }, { "epoch": 6.23, "learning_rate": 2.4609872913788956e-06, "loss": 0.0022, "step": 33969 }, { "epoch": 6.23, "learning_rate": 2.460499544741729e-06, "loss": 0.0018, "step": 33970 }, { "epoch": 6.23, "learning_rate": 2.460011839662466e-06, "loss": 0.0013, "step": 33971 }, { "epoch": 6.23, "learning_rate": 2.459524176143794e-06, "loss": 0.002, "step": 33972 }, { "epoch": 6.23, "learning_rate": 2.459036554188401e-06, "loss": 0.0024, "step": 33973 }, { "epoch": 6.23, "learning_rate": 2.458548973798973e-06, "loss": 0.0015, "step": 33974 }, { "epoch": 6.23, "learning_rate": 2.458061434978205e-06, "loss": 0.0017, "step": 33975 }, { "epoch": 6.23, "learning_rate": 2.4575739377287768e-06, "loss": 0.0027, "step": 33976 }, { "epoch": 6.23, "learning_rate": 2.457086482053378e-06, "loss": 0.0013, "step": 33977 }, { "epoch": 6.23, "learning_rate": 2.4565990679546913e-06, "loss": 0.002, "step": 33978 }, { "epoch": 6.23, "learning_rate": 2.456111695435411e-06, "loss": 0.0026, "step": 33979 }, { "epoch": 6.23, "learning_rate": 2.455624364498219e-06, "loss": 0.0031, "step": 33980 }, { "epoch": 6.23, "learning_rate": 2.455137075145806e-06, "loss": 0.0026, "step": 33981 }, { "epoch": 6.23, "learning_rate": 2.4546498273808503e-06, "loss": 0.0015, "step": 33982 }, { "epoch": 6.23, "learning_rate": 2.45416262120604e-06, "loss": 0.0022, "step": 33983 }, { "epoch": 6.23, "learning_rate": 2.453675456624064e-06, "loss": 0.0012, "step": 33984 }, { "epoch": 6.23, "learning_rate": 2.453188333637606e-06, "loss": 0.002, "step": 33985 }, { "epoch": 6.23, "learning_rate": 2.4527012522493497e-06, "loss": 0.0035, "step": 33986 }, { "epoch": 6.23, "learning_rate": 2.452214212461982e-06, "loss": 0.0028, "step": 33987 }, { "epoch": 6.23, "learning_rate": 2.451727214278187e-06, "loss": 0.0019, "step": 33988 }, { "epoch": 6.23, "learning_rate": 2.4512402577006468e-06, "loss": 0.0034, "step": 33989 }, { "epoch": 6.23, "learning_rate": 2.4507533427320485e-06, "loss": 0.0027, "step": 33990 }, { "epoch": 6.23, "learning_rate": 2.450266469375071e-06, "loss": 0.0027, "step": 33991 }, { "epoch": 6.23, "learning_rate": 2.4497796376324045e-06, "loss": 0.0017, "step": 33992 }, { "epoch": 6.23, "learning_rate": 2.4492928475067324e-06, "loss": 0.0013, "step": 33993 }, { "epoch": 6.23, "learning_rate": 2.4488060990007322e-06, "loss": 0.0029, "step": 33994 }, { "epoch": 6.23, "learning_rate": 2.44831939211709e-06, "loss": 0.0019, "step": 33995 }, { "epoch": 6.23, "learning_rate": 2.4478327268584857e-06, "loss": 0.0026, "step": 33996 }, { "epoch": 6.23, "learning_rate": 2.447346103227607e-06, "loss": 0.0022, "step": 33997 }, { "epoch": 6.23, "learning_rate": 2.446859521227133e-06, "loss": 0.0025, "step": 33998 }, { "epoch": 6.23, "learning_rate": 2.4463729808597493e-06, "loss": 0.0024, "step": 33999 }, { "epoch": 6.24, "learning_rate": 2.445886482128129e-06, "loss": 0.0019, "step": 34000 }, { "epoch": 6.24, "learning_rate": 2.445400025034962e-06, "loss": 0.0025, "step": 34001 }, { "epoch": 6.24, "learning_rate": 2.4449136095829274e-06, "loss": 0.0018, "step": 34002 }, { "epoch": 6.24, "learning_rate": 2.4444272357747057e-06, "loss": 0.0031, "step": 34003 }, { "epoch": 6.24, "learning_rate": 2.443940903612978e-06, "loss": 0.0019, "step": 34004 }, { "epoch": 6.24, "learning_rate": 2.4434546131004255e-06, "loss": 0.0017, "step": 34005 }, { "epoch": 6.24, "learning_rate": 2.4429683642397283e-06, "loss": 0.0018, "step": 34006 }, { "epoch": 6.24, "learning_rate": 2.4424821570335657e-06, "loss": 0.0021, "step": 34007 }, { "epoch": 6.24, "learning_rate": 2.4419959914846194e-06, "loss": 0.0026, "step": 34008 }, { "epoch": 6.24, "learning_rate": 2.441509867595565e-06, "loss": 0.0027, "step": 34009 }, { "epoch": 6.24, "learning_rate": 2.441023785369091e-06, "loss": 0.0025, "step": 34010 }, { "epoch": 6.24, "learning_rate": 2.440537744807868e-06, "loss": 0.0018, "step": 34011 }, { "epoch": 6.24, "learning_rate": 2.440051745914579e-06, "loss": 0.0025, "step": 34012 }, { "epoch": 6.24, "learning_rate": 2.439565788691898e-06, "loss": 0.003, "step": 34013 }, { "epoch": 6.24, "learning_rate": 2.43907987314251e-06, "loss": 0.002, "step": 34014 }, { "epoch": 6.24, "learning_rate": 2.438593999269092e-06, "loss": 0.0032, "step": 34015 }, { "epoch": 6.24, "learning_rate": 2.43810816707432e-06, "loss": 0.0025, "step": 34016 }, { "epoch": 6.24, "learning_rate": 2.4376223765608764e-06, "loss": 0.0029, "step": 34017 }, { "epoch": 6.24, "learning_rate": 2.4371366277314302e-06, "loss": 0.0022, "step": 34018 }, { "epoch": 6.24, "learning_rate": 2.4366509205886656e-06, "loss": 0.0037, "step": 34019 }, { "epoch": 6.24, "learning_rate": 2.436165255135259e-06, "loss": 0.001, "step": 34020 }, { "epoch": 6.24, "learning_rate": 2.4356796313738873e-06, "loss": 0.0036, "step": 34021 }, { "epoch": 6.24, "learning_rate": 2.4351940493072255e-06, "loss": 0.0029, "step": 34022 }, { "epoch": 6.24, "learning_rate": 2.434708508937953e-06, "loss": 0.0028, "step": 34023 }, { "epoch": 6.24, "learning_rate": 2.4342230102687425e-06, "loss": 0.0024, "step": 34024 }, { "epoch": 6.24, "learning_rate": 2.4337375533022733e-06, "loss": 0.0021, "step": 34025 }, { "epoch": 6.24, "learning_rate": 2.43325213804122e-06, "loss": 0.0018, "step": 34026 }, { "epoch": 6.24, "learning_rate": 2.4327667644882547e-06, "loss": 0.0016, "step": 34027 }, { "epoch": 6.24, "learning_rate": 2.432281432646062e-06, "loss": 0.0027, "step": 34028 }, { "epoch": 6.24, "learning_rate": 2.4317961425173088e-06, "loss": 0.002, "step": 34029 }, { "epoch": 6.24, "learning_rate": 2.4313108941046728e-06, "loss": 0.0016, "step": 34030 }, { "epoch": 6.24, "learning_rate": 2.4308256874108247e-06, "loss": 0.0021, "step": 34031 }, { "epoch": 6.24, "learning_rate": 2.4303405224384468e-06, "loss": 0.002, "step": 34032 }, { "epoch": 6.24, "learning_rate": 2.4298553991902084e-06, "loss": 0.0019, "step": 34033 }, { "epoch": 6.24, "learning_rate": 2.4293703176687844e-06, "loss": 0.0026, "step": 34034 }, { "epoch": 6.24, "learning_rate": 2.4288852778768523e-06, "loss": 0.0019, "step": 34035 }, { "epoch": 6.24, "learning_rate": 2.428400279817075e-06, "loss": 0.0018, "step": 34036 }, { "epoch": 6.24, "learning_rate": 2.427915323492136e-06, "loss": 0.003, "step": 34037 }, { "epoch": 6.24, "learning_rate": 2.4274304089047053e-06, "loss": 0.0021, "step": 34038 }, { "epoch": 6.24, "learning_rate": 2.4269455360574555e-06, "loss": 0.0026, "step": 34039 }, { "epoch": 6.24, "learning_rate": 2.4264607049530596e-06, "loss": 0.0013, "step": 34040 }, { "epoch": 6.24, "learning_rate": 2.42597591559419e-06, "loss": 0.0029, "step": 34041 }, { "epoch": 6.24, "learning_rate": 2.425491167983518e-06, "loss": 0.0022, "step": 34042 }, { "epoch": 6.24, "learning_rate": 2.4250064621237155e-06, "loss": 0.0019, "step": 34043 }, { "epoch": 6.24, "learning_rate": 2.424521798017453e-06, "loss": 0.0015, "step": 34044 }, { "epoch": 6.24, "learning_rate": 2.4240371756674063e-06, "loss": 0.002, "step": 34045 }, { "epoch": 6.24, "learning_rate": 2.4235525950762474e-06, "loss": 0.0023, "step": 34046 }, { "epoch": 6.24, "learning_rate": 2.4230680562466415e-06, "loss": 0.0021, "step": 34047 }, { "epoch": 6.24, "learning_rate": 2.4225835591812617e-06, "loss": 0.0013, "step": 34048 }, { "epoch": 6.24, "learning_rate": 2.4220991038827755e-06, "loss": 0.0024, "step": 34049 }, { "epoch": 6.24, "learning_rate": 2.4216146903538594e-06, "loss": 0.0017, "step": 34050 }, { "epoch": 6.24, "learning_rate": 2.421130318597181e-06, "loss": 0.0016, "step": 34051 }, { "epoch": 6.24, "learning_rate": 2.42064598861541e-06, "loss": 0.0019, "step": 34052 }, { "epoch": 6.24, "learning_rate": 2.4201617004112156e-06, "loss": 0.0015, "step": 34053 }, { "epoch": 6.25, "learning_rate": 2.419677453987268e-06, "loss": 0.0025, "step": 34054 }, { "epoch": 6.25, "learning_rate": 2.4191932493462358e-06, "loss": 0.0016, "step": 34055 }, { "epoch": 6.25, "learning_rate": 2.418709086490787e-06, "loss": 0.0012, "step": 34056 }, { "epoch": 6.25, "learning_rate": 2.4182249654235925e-06, "loss": 0.0037, "step": 34057 }, { "epoch": 6.25, "learning_rate": 2.4177408861473185e-06, "loss": 0.0031, "step": 34058 }, { "epoch": 6.25, "learning_rate": 2.4172568486646353e-06, "loss": 0.0013, "step": 34059 }, { "epoch": 6.25, "learning_rate": 2.4167728529782096e-06, "loss": 0.002, "step": 34060 }, { "epoch": 6.25, "learning_rate": 2.4162888990907097e-06, "loss": 0.0015, "step": 34061 }, { "epoch": 6.25, "learning_rate": 2.4158049870047995e-06, "loss": 0.0029, "step": 34062 }, { "epoch": 6.25, "learning_rate": 2.415321116723154e-06, "loss": 0.0037, "step": 34063 }, { "epoch": 6.25, "learning_rate": 2.4148372882484385e-06, "loss": 0.0015, "step": 34064 }, { "epoch": 6.25, "learning_rate": 2.4143535015833142e-06, "loss": 0.0019, "step": 34065 }, { "epoch": 6.25, "learning_rate": 2.4138697567304493e-06, "loss": 0.0014, "step": 34066 }, { "epoch": 6.25, "learning_rate": 2.4133860536925136e-06, "loss": 0.003, "step": 34067 }, { "epoch": 6.25, "learning_rate": 2.412902392472173e-06, "loss": 0.0012, "step": 34068 }, { "epoch": 6.25, "learning_rate": 2.4124187730720916e-06, "loss": 0.0026, "step": 34069 }, { "epoch": 6.25, "learning_rate": 2.4119351954949354e-06, "loss": 0.002, "step": 34070 }, { "epoch": 6.25, "learning_rate": 2.4114516597433703e-06, "loss": 0.003, "step": 34071 }, { "epoch": 6.25, "learning_rate": 2.410968165820061e-06, "loss": 0.0024, "step": 34072 }, { "epoch": 6.25, "learning_rate": 2.4104847137276745e-06, "loss": 0.0016, "step": 34073 }, { "epoch": 6.25, "learning_rate": 2.410001303468873e-06, "loss": 0.0023, "step": 34074 }, { "epoch": 6.25, "learning_rate": 2.4095179350463216e-06, "loss": 0.0019, "step": 34075 }, { "epoch": 6.25, "learning_rate": 2.4090346084626857e-06, "loss": 0.0032, "step": 34076 }, { "epoch": 6.25, "learning_rate": 2.408551323720629e-06, "loss": 0.0037, "step": 34077 }, { "epoch": 6.25, "learning_rate": 2.408068080822815e-06, "loss": 0.0021, "step": 34078 }, { "epoch": 6.25, "learning_rate": 2.4075848797719083e-06, "loss": 0.0027, "step": 34079 }, { "epoch": 6.25, "learning_rate": 2.4071017205705683e-06, "loss": 0.0033, "step": 34080 }, { "epoch": 6.25, "learning_rate": 2.4066186032214646e-06, "loss": 0.0033, "step": 34081 }, { "epoch": 6.25, "learning_rate": 2.406135527727259e-06, "loss": 0.0021, "step": 34082 }, { "epoch": 6.25, "learning_rate": 2.40565249409061e-06, "loss": 0.0016, "step": 34083 }, { "epoch": 6.25, "learning_rate": 2.4051695023141797e-06, "loss": 0.0022, "step": 34084 }, { "epoch": 6.25, "learning_rate": 2.404686552400636e-06, "loss": 0.0025, "step": 34085 }, { "epoch": 6.25, "learning_rate": 2.4042036443526375e-06, "loss": 0.002, "step": 34086 }, { "epoch": 6.25, "learning_rate": 2.4037207781728465e-06, "loss": 0.0018, "step": 34087 }, { "epoch": 6.25, "learning_rate": 2.403237953863927e-06, "loss": 0.003, "step": 34088 }, { "epoch": 6.25, "learning_rate": 2.4027551714285337e-06, "loss": 0.0015, "step": 34089 }, { "epoch": 6.25, "learning_rate": 2.4022724308693336e-06, "loss": 0.0043, "step": 34090 }, { "epoch": 6.25, "learning_rate": 2.4017897321889862e-06, "loss": 0.0015, "step": 34091 }, { "epoch": 6.25, "learning_rate": 2.401307075390151e-06, "loss": 0.0023, "step": 34092 }, { "epoch": 6.25, "learning_rate": 2.4008244604754893e-06, "loss": 0.0015, "step": 34093 }, { "epoch": 6.25, "learning_rate": 2.4003418874476626e-06, "loss": 0.0021, "step": 34094 }, { "epoch": 6.25, "learning_rate": 2.399859356309329e-06, "loss": 0.0021, "step": 34095 }, { "epoch": 6.25, "learning_rate": 2.399376867063148e-06, "loss": 0.0012, "step": 34096 }, { "epoch": 6.25, "learning_rate": 2.398894419711777e-06, "loss": 0.0023, "step": 34097 }, { "epoch": 6.25, "learning_rate": 2.3984120142578804e-06, "loss": 0.0016, "step": 34098 }, { "epoch": 6.25, "learning_rate": 2.397929650704116e-06, "loss": 0.0022, "step": 34099 }, { "epoch": 6.25, "learning_rate": 2.3974473290531443e-06, "loss": 0.002, "step": 34100 }, { "epoch": 6.25, "learning_rate": 2.3969650493076168e-06, "loss": 0.002, "step": 34101 }, { "epoch": 6.25, "learning_rate": 2.396482811470194e-06, "loss": 0.0029, "step": 34102 }, { "epoch": 6.25, "learning_rate": 2.39600061554354e-06, "loss": 0.0027, "step": 34103 }, { "epoch": 6.25, "learning_rate": 2.395518461530307e-06, "loss": 0.0017, "step": 34104 }, { "epoch": 6.25, "learning_rate": 2.3950363494331553e-06, "loss": 0.0025, "step": 34105 }, { "epoch": 6.25, "learning_rate": 2.3945542792547415e-06, "loss": 0.0012, "step": 34106 }, { "epoch": 6.25, "learning_rate": 2.3940722509977222e-06, "loss": 0.0017, "step": 34107 }, { "epoch": 6.25, "learning_rate": 2.3935902646647558e-06, "loss": 0.0032, "step": 34108 }, { "epoch": 6.26, "learning_rate": 2.393108320258497e-06, "loss": 0.0017, "step": 34109 }, { "epoch": 6.26, "learning_rate": 2.392626417781604e-06, "loss": 0.0018, "step": 34110 }, { "epoch": 6.26, "learning_rate": 2.392144557236732e-06, "loss": 0.0019, "step": 34111 }, { "epoch": 6.26, "learning_rate": 2.391662738626538e-06, "loss": 0.0025, "step": 34112 }, { "epoch": 6.26, "learning_rate": 2.3911809619536765e-06, "loss": 0.0021, "step": 34113 }, { "epoch": 6.26, "learning_rate": 2.3906992272208042e-06, "loss": 0.0021, "step": 34114 }, { "epoch": 6.26, "learning_rate": 2.3902175344305734e-06, "loss": 0.0018, "step": 34115 }, { "epoch": 6.26, "learning_rate": 2.3897358835856445e-06, "loss": 0.0019, "step": 34116 }, { "epoch": 6.26, "learning_rate": 2.3892542746886717e-06, "loss": 0.0018, "step": 34117 }, { "epoch": 6.26, "learning_rate": 2.3887727077423063e-06, "loss": 0.0021, "step": 34118 }, { "epoch": 6.26, "learning_rate": 2.3882911827491994e-06, "loss": 0.0011, "step": 34119 }, { "epoch": 6.26, "learning_rate": 2.3878096997120137e-06, "loss": 0.0013, "step": 34120 }, { "epoch": 6.26, "learning_rate": 2.3873282586334e-06, "loss": 0.0018, "step": 34121 }, { "epoch": 6.26, "learning_rate": 2.38684685951601e-06, "loss": 0.0015, "step": 34122 }, { "epoch": 6.26, "learning_rate": 2.386365502362499e-06, "loss": 0.0017, "step": 34123 }, { "epoch": 6.26, "learning_rate": 2.38588418717552e-06, "loss": 0.003, "step": 34124 }, { "epoch": 6.26, "learning_rate": 2.385402913957725e-06, "loss": 0.0019, "step": 34125 }, { "epoch": 6.26, "learning_rate": 2.3849216827117685e-06, "loss": 0.0016, "step": 34126 }, { "epoch": 6.26, "learning_rate": 2.384440493440302e-06, "loss": 0.0018, "step": 34127 }, { "epoch": 6.26, "learning_rate": 2.3839593461459744e-06, "loss": 0.0026, "step": 34128 }, { "epoch": 6.26, "learning_rate": 2.383478240831447e-06, "loss": 0.0022, "step": 34129 }, { "epoch": 6.26, "learning_rate": 2.3829971774993633e-06, "loss": 0.0022, "step": 34130 }, { "epoch": 6.26, "learning_rate": 2.382516156152378e-06, "loss": 0.0028, "step": 34131 }, { "epoch": 6.26, "learning_rate": 2.3820351767931417e-06, "loss": 0.0013, "step": 34132 }, { "epoch": 6.26, "learning_rate": 2.381554239424303e-06, "loss": 0.0018, "step": 34133 }, { "epoch": 6.26, "learning_rate": 2.381073344048519e-06, "loss": 0.0012, "step": 34134 }, { "epoch": 6.26, "learning_rate": 2.3805924906684396e-06, "loss": 0.0018, "step": 34135 }, { "epoch": 6.26, "learning_rate": 2.380111679286711e-06, "loss": 0.0017, "step": 34136 }, { "epoch": 6.26, "learning_rate": 2.379630909905981e-06, "loss": 0.0018, "step": 34137 }, { "epoch": 6.26, "learning_rate": 2.379150182528909e-06, "loss": 0.0036, "step": 34138 }, { "epoch": 6.26, "learning_rate": 2.378669497158138e-06, "loss": 0.0023, "step": 34139 }, { "epoch": 6.26, "learning_rate": 2.3781888537963195e-06, "loss": 0.003, "step": 34140 }, { "epoch": 6.26, "learning_rate": 2.3777082524461026e-06, "loss": 0.0015, "step": 34141 }, { "epoch": 6.26, "learning_rate": 2.377227693110137e-06, "loss": 0.0017, "step": 34142 }, { "epoch": 6.26, "learning_rate": 2.3767471757910697e-06, "loss": 0.001, "step": 34143 }, { "epoch": 6.26, "learning_rate": 2.3762667004915518e-06, "loss": 0.002, "step": 34144 }, { "epoch": 6.26, "learning_rate": 2.3757862672142295e-06, "loss": 0.0017, "step": 34145 }, { "epoch": 6.26, "learning_rate": 2.3753058759617498e-06, "loss": 0.0032, "step": 34146 }, { "epoch": 6.26, "learning_rate": 2.374825526736768e-06, "loss": 0.0026, "step": 34147 }, { "epoch": 6.26, "learning_rate": 2.3743452195419246e-06, "loss": 0.0018, "step": 34148 }, { "epoch": 6.26, "learning_rate": 2.373864954379869e-06, "loss": 0.0011, "step": 34149 }, { "epoch": 6.26, "learning_rate": 2.3733847312532444e-06, "loss": 0.0013, "step": 34150 }, { "epoch": 6.26, "learning_rate": 2.372904550164705e-06, "loss": 0.0016, "step": 34151 }, { "epoch": 6.26, "learning_rate": 2.372424411116895e-06, "loss": 0.0019, "step": 34152 }, { "epoch": 6.26, "learning_rate": 2.371944314112462e-06, "loss": 0.0014, "step": 34153 }, { "epoch": 6.26, "learning_rate": 2.371464259154049e-06, "loss": 0.0026, "step": 34154 }, { "epoch": 6.26, "learning_rate": 2.3709842462443e-06, "loss": 0.0023, "step": 34155 }, { "epoch": 6.26, "learning_rate": 2.3705042753858688e-06, "loss": 0.0021, "step": 34156 }, { "epoch": 6.26, "learning_rate": 2.3700243465813956e-06, "loss": 0.0021, "step": 34157 }, { "epoch": 6.26, "learning_rate": 2.369544459833527e-06, "loss": 0.0017, "step": 34158 }, { "epoch": 6.26, "learning_rate": 2.3690646151449083e-06, "loss": 0.0028, "step": 34159 }, { "epoch": 6.26, "learning_rate": 2.3685848125181843e-06, "loss": 0.0016, "step": 34160 }, { "epoch": 6.26, "learning_rate": 2.368105051955999e-06, "loss": 0.0018, "step": 34161 }, { "epoch": 6.26, "learning_rate": 2.3676253334609976e-06, "loss": 0.0022, "step": 34162 }, { "epoch": 6.26, "learning_rate": 2.3671456570358254e-06, "loss": 0.0023, "step": 34163 }, { "epoch": 6.27, "learning_rate": 2.366666022683124e-06, "loss": 0.0023, "step": 34164 }, { "epoch": 6.27, "learning_rate": 2.366186430405538e-06, "loss": 0.0017, "step": 34165 }, { "epoch": 6.27, "learning_rate": 2.3657068802057126e-06, "loss": 0.0028, "step": 34166 }, { "epoch": 6.27, "learning_rate": 2.3652273720862885e-06, "loss": 0.0027, "step": 34167 }, { "epoch": 6.27, "learning_rate": 2.364747906049908e-06, "loss": 0.0023, "step": 34168 }, { "epoch": 6.27, "learning_rate": 2.364268482099218e-06, "loss": 0.0015, "step": 34169 }, { "epoch": 6.27, "learning_rate": 2.3637891002368596e-06, "loss": 0.0045, "step": 34170 }, { "epoch": 6.27, "learning_rate": 2.3633097604654774e-06, "loss": 0.0015, "step": 34171 }, { "epoch": 6.27, "learning_rate": 2.362830462787704e-06, "loss": 0.0025, "step": 34172 }, { "epoch": 6.27, "learning_rate": 2.3623512072061926e-06, "loss": 0.0032, "step": 34173 }, { "epoch": 6.27, "learning_rate": 2.361871993723579e-06, "loss": 0.0021, "step": 34174 }, { "epoch": 6.27, "learning_rate": 2.3613928223425066e-06, "loss": 0.0026, "step": 34175 }, { "epoch": 6.27, "learning_rate": 2.360913693065615e-06, "loss": 0.0022, "step": 34176 }, { "epoch": 6.27, "learning_rate": 2.3604346058955463e-06, "loss": 0.0025, "step": 34177 }, { "epoch": 6.27, "learning_rate": 2.3599555608349424e-06, "loss": 0.0018, "step": 34178 }, { "epoch": 6.27, "learning_rate": 2.3594765578864407e-06, "loss": 0.001, "step": 34179 }, { "epoch": 6.27, "learning_rate": 2.3589975970526844e-06, "loss": 0.0017, "step": 34180 }, { "epoch": 6.27, "learning_rate": 2.358518678336309e-06, "loss": 0.002, "step": 34181 }, { "epoch": 6.27, "learning_rate": 2.358039801739963e-06, "loss": 0.0025, "step": 34182 }, { "epoch": 6.27, "learning_rate": 2.357560967266278e-06, "loss": 0.0026, "step": 34183 }, { "epoch": 6.27, "learning_rate": 2.3570821749178963e-06, "loss": 0.0016, "step": 34184 }, { "epoch": 6.27, "learning_rate": 2.356603424697457e-06, "loss": 0.0025, "step": 34185 }, { "epoch": 6.27, "learning_rate": 2.356124716607595e-06, "loss": 0.001, "step": 34186 }, { "epoch": 6.27, "learning_rate": 2.355646050650956e-06, "loss": 0.002, "step": 34187 }, { "epoch": 6.27, "learning_rate": 2.355167426830175e-06, "loss": 0.0024, "step": 34188 }, { "epoch": 6.27, "learning_rate": 2.3546888451478934e-06, "loss": 0.0018, "step": 34189 }, { "epoch": 6.27, "learning_rate": 2.35421030560674e-06, "loss": 0.0017, "step": 34190 }, { "epoch": 6.27, "learning_rate": 2.3537318082093618e-06, "loss": 0.0021, "step": 34191 }, { "epoch": 6.27, "learning_rate": 2.353253352958392e-06, "loss": 0.0035, "step": 34192 }, { "epoch": 6.27, "learning_rate": 2.3527749398564693e-06, "loss": 0.002, "step": 34193 }, { "epoch": 6.27, "learning_rate": 2.3522965689062305e-06, "loss": 0.0028, "step": 34194 }, { "epoch": 6.27, "learning_rate": 2.351818240110312e-06, "loss": 0.0016, "step": 34195 }, { "epoch": 6.27, "learning_rate": 2.351339953471351e-06, "loss": 0.0024, "step": 34196 }, { "epoch": 6.27, "learning_rate": 2.3508617089919825e-06, "loss": 0.0021, "step": 34197 }, { "epoch": 6.27, "learning_rate": 2.350383506674844e-06, "loss": 0.0021, "step": 34198 }, { "epoch": 6.27, "learning_rate": 2.3499053465225684e-06, "loss": 0.0015, "step": 34199 }, { "epoch": 6.27, "learning_rate": 2.3494272285377986e-06, "loss": 0.001, "step": 34200 }, { "epoch": 6.27, "learning_rate": 2.348949152723162e-06, "loss": 0.002, "step": 34201 }, { "epoch": 6.27, "learning_rate": 2.3484711190812982e-06, "loss": 0.0025, "step": 34202 }, { "epoch": 6.27, "learning_rate": 2.347993127614837e-06, "loss": 0.0026, "step": 34203 }, { "epoch": 6.27, "learning_rate": 2.3475151783264204e-06, "loss": 0.0034, "step": 34204 }, { "epoch": 6.27, "learning_rate": 2.3470372712186797e-06, "loss": 0.0047, "step": 34205 }, { "epoch": 6.27, "learning_rate": 2.3465594062942474e-06, "loss": 0.0024, "step": 34206 }, { "epoch": 6.27, "learning_rate": 2.346081583555764e-06, "loss": 0.0021, "step": 34207 }, { "epoch": 6.27, "learning_rate": 2.3456038030058504e-06, "loss": 0.0016, "step": 34208 }, { "epoch": 6.27, "learning_rate": 2.3451260646471517e-06, "loss": 0.0012, "step": 34209 }, { "epoch": 6.27, "learning_rate": 2.3446483684822973e-06, "loss": 0.0016, "step": 34210 }, { "epoch": 6.27, "learning_rate": 2.3441707145139214e-06, "loss": 0.0025, "step": 34211 }, { "epoch": 6.27, "learning_rate": 2.3436931027446553e-06, "loss": 0.0014, "step": 34212 }, { "epoch": 6.27, "learning_rate": 2.3432155331771313e-06, "loss": 0.0023, "step": 34213 }, { "epoch": 6.27, "learning_rate": 2.3427380058139836e-06, "loss": 0.0017, "step": 34214 }, { "epoch": 6.27, "learning_rate": 2.342260520657843e-06, "loss": 0.0015, "step": 34215 }, { "epoch": 6.27, "learning_rate": 2.3417830777113413e-06, "loss": 0.0018, "step": 34216 }, { "epoch": 6.27, "learning_rate": 2.341305676977109e-06, "loss": 0.002, "step": 34217 }, { "epoch": 6.28, "learning_rate": 2.340828318457784e-06, "loss": 0.0029, "step": 34218 }, { "epoch": 6.28, "learning_rate": 2.3403510021559904e-06, "loss": 0.0023, "step": 34219 }, { "epoch": 6.28, "learning_rate": 2.339873728074362e-06, "loss": 0.0014, "step": 34220 }, { "epoch": 6.28, "learning_rate": 2.339396496215526e-06, "loss": 0.0026, "step": 34221 }, { "epoch": 6.28, "learning_rate": 2.338919306582118e-06, "loss": 0.0033, "step": 34222 }, { "epoch": 6.28, "learning_rate": 2.338442159176767e-06, "loss": 0.0023, "step": 34223 }, { "epoch": 6.28, "learning_rate": 2.3379650540021015e-06, "loss": 0.0021, "step": 34224 }, { "epoch": 6.28, "learning_rate": 2.337487991060753e-06, "loss": 0.0018, "step": 34225 }, { "epoch": 6.28, "learning_rate": 2.3370109703553502e-06, "loss": 0.0015, "step": 34226 }, { "epoch": 6.28, "learning_rate": 2.336533991888522e-06, "loss": 0.0011, "step": 34227 }, { "epoch": 6.28, "learning_rate": 2.3360570556628983e-06, "loss": 0.002, "step": 34228 }, { "epoch": 6.28, "learning_rate": 2.335580161681107e-06, "loss": 0.0022, "step": 34229 }, { "epoch": 6.28, "learning_rate": 2.3351033099457774e-06, "loss": 0.0018, "step": 34230 }, { "epoch": 6.28, "learning_rate": 2.334626500459539e-06, "loss": 0.0013, "step": 34231 }, { "epoch": 6.28, "learning_rate": 2.3341497332250187e-06, "loss": 0.0027, "step": 34232 }, { "epoch": 6.28, "learning_rate": 2.3336730082448435e-06, "loss": 0.0026, "step": 34233 }, { "epoch": 6.28, "learning_rate": 2.3331963255216407e-06, "loss": 0.0018, "step": 34234 }, { "epoch": 6.28, "learning_rate": 2.3327196850580415e-06, "loss": 0.0023, "step": 34235 }, { "epoch": 6.28, "learning_rate": 2.3322430868566736e-06, "loss": 0.0028, "step": 34236 }, { "epoch": 6.28, "learning_rate": 2.3317665309201596e-06, "loss": 0.0029, "step": 34237 }, { "epoch": 6.28, "learning_rate": 2.3312900172511276e-06, "loss": 0.0024, "step": 34238 }, { "epoch": 6.28, "learning_rate": 2.330813545852202e-06, "loss": 0.0028, "step": 34239 }, { "epoch": 6.28, "learning_rate": 2.3303371167260147e-06, "loss": 0.0026, "step": 34240 }, { "epoch": 6.28, "learning_rate": 2.3298607298751884e-06, "loss": 0.002, "step": 34241 }, { "epoch": 6.28, "learning_rate": 2.3293843853023523e-06, "loss": 0.002, "step": 34242 }, { "epoch": 6.28, "learning_rate": 2.328908083010124e-06, "loss": 0.002, "step": 34243 }, { "epoch": 6.28, "learning_rate": 2.328431823001136e-06, "loss": 0.0025, "step": 34244 }, { "epoch": 6.28, "learning_rate": 2.327955605278013e-06, "loss": 0.0014, "step": 34245 }, { "epoch": 6.28, "learning_rate": 2.327479429843378e-06, "loss": 0.0031, "step": 34246 }, { "epoch": 6.28, "learning_rate": 2.3270032966998546e-06, "loss": 0.0015, "step": 34247 }, { "epoch": 6.28, "learning_rate": 2.32652720585007e-06, "loss": 0.0015, "step": 34248 }, { "epoch": 6.28, "learning_rate": 2.3260511572966473e-06, "loss": 0.0015, "step": 34249 }, { "epoch": 6.28, "learning_rate": 2.325575151042211e-06, "loss": 0.0022, "step": 34250 }, { "epoch": 6.28, "learning_rate": 2.3250991870893837e-06, "loss": 0.0024, "step": 34251 }, { "epoch": 6.28, "learning_rate": 2.3246232654407864e-06, "loss": 0.0017, "step": 34252 }, { "epoch": 6.28, "learning_rate": 2.3241473860990483e-06, "loss": 0.0012, "step": 34253 }, { "epoch": 6.28, "learning_rate": 2.3236715490667914e-06, "loss": 0.0015, "step": 34254 }, { "epoch": 6.28, "learning_rate": 2.3231957543466356e-06, "loss": 0.0012, "step": 34255 }, { "epoch": 6.28, "learning_rate": 2.3227200019412012e-06, "loss": 0.0017, "step": 34256 }, { "epoch": 6.28, "learning_rate": 2.3222442918531174e-06, "loss": 0.0022, "step": 34257 }, { "epoch": 6.28, "learning_rate": 2.3217686240850024e-06, "loss": 0.0023, "step": 34258 }, { "epoch": 6.28, "learning_rate": 2.3212929986394785e-06, "loss": 0.002, "step": 34259 }, { "epoch": 6.28, "learning_rate": 2.320817415519171e-06, "loss": 0.0017, "step": 34260 }, { "epoch": 6.28, "learning_rate": 2.3203418747266913e-06, "loss": 0.0033, "step": 34261 }, { "epoch": 6.28, "learning_rate": 2.31986637626467e-06, "loss": 0.0043, "step": 34262 }, { "epoch": 6.28, "learning_rate": 2.3193909201357255e-06, "loss": 0.0024, "step": 34263 }, { "epoch": 6.28, "learning_rate": 2.3189155063424784e-06, "loss": 0.0032, "step": 34264 }, { "epoch": 6.28, "learning_rate": 2.3184401348875484e-06, "loss": 0.0032, "step": 34265 }, { "epoch": 6.28, "learning_rate": 2.3179648057735573e-06, "loss": 0.0019, "step": 34266 }, { "epoch": 6.28, "learning_rate": 2.317489519003123e-06, "loss": 0.0018, "step": 34267 }, { "epoch": 6.28, "learning_rate": 2.317014274578867e-06, "loss": 0.0017, "step": 34268 }, { "epoch": 6.28, "learning_rate": 2.3165390725034053e-06, "loss": 0.0023, "step": 34269 }, { "epoch": 6.28, "learning_rate": 2.316063912779364e-06, "loss": 0.0021, "step": 34270 }, { "epoch": 6.28, "learning_rate": 2.315588795409357e-06, "loss": 0.0021, "step": 34271 }, { "epoch": 6.28, "learning_rate": 2.3151137203960085e-06, "loss": 0.0031, "step": 34272 }, { "epoch": 6.29, "learning_rate": 2.314638687741929e-06, "loss": 0.0014, "step": 34273 }, { "epoch": 6.29, "learning_rate": 2.3141636974497395e-06, "loss": 0.0026, "step": 34274 }, { "epoch": 6.29, "learning_rate": 2.313688749522063e-06, "loss": 0.0018, "step": 34275 }, { "epoch": 6.29, "learning_rate": 2.3132138439615125e-06, "loss": 0.004, "step": 34276 }, { "epoch": 6.29, "learning_rate": 2.312738980770709e-06, "loss": 0.0015, "step": 34277 }, { "epoch": 6.29, "learning_rate": 2.312264159952268e-06, "loss": 0.0028, "step": 34278 }, { "epoch": 6.29, "learning_rate": 2.3117893815088067e-06, "loss": 0.0034, "step": 34279 }, { "epoch": 6.29, "learning_rate": 2.3113146454429426e-06, "loss": 0.0019, "step": 34280 }, { "epoch": 6.29, "learning_rate": 2.310839951757292e-06, "loss": 0.001, "step": 34281 }, { "epoch": 6.29, "learning_rate": 2.310365300454471e-06, "loss": 0.0016, "step": 34282 }, { "epoch": 6.29, "learning_rate": 2.3098906915370976e-06, "loss": 0.0022, "step": 34283 }, { "epoch": 6.29, "learning_rate": 2.3094161250077874e-06, "loss": 0.0026, "step": 34284 }, { "epoch": 6.29, "learning_rate": 2.3089416008691547e-06, "loss": 0.0021, "step": 34285 }, { "epoch": 6.29, "learning_rate": 2.3084671191238174e-06, "loss": 0.0018, "step": 34286 }, { "epoch": 6.29, "learning_rate": 2.3079926797743846e-06, "loss": 0.0018, "step": 34287 }, { "epoch": 6.29, "learning_rate": 2.307518282823481e-06, "loss": 0.0021, "step": 34288 }, { "epoch": 6.29, "learning_rate": 2.307043928273719e-06, "loss": 0.0035, "step": 34289 }, { "epoch": 6.29, "learning_rate": 2.3065696161277084e-06, "loss": 0.003, "step": 34290 }, { "epoch": 6.29, "learning_rate": 2.3060953463880665e-06, "loss": 0.0028, "step": 34291 }, { "epoch": 6.29, "learning_rate": 2.305621119057404e-06, "loss": 0.0035, "step": 34292 }, { "epoch": 6.29, "learning_rate": 2.3051469341383403e-06, "loss": 0.0017, "step": 34293 }, { "epoch": 6.29, "learning_rate": 2.304672791633489e-06, "loss": 0.0018, "step": 34294 }, { "epoch": 6.29, "learning_rate": 2.3041986915454595e-06, "loss": 0.0017, "step": 34295 }, { "epoch": 6.29, "learning_rate": 2.3037246338768684e-06, "loss": 0.0016, "step": 34296 }, { "epoch": 6.29, "learning_rate": 2.303250618630327e-06, "loss": 0.0026, "step": 34297 }, { "epoch": 6.29, "learning_rate": 2.3027766458084477e-06, "loss": 0.0018, "step": 34298 }, { "epoch": 6.29, "learning_rate": 2.3023027154138457e-06, "loss": 0.0027, "step": 34299 }, { "epoch": 6.29, "learning_rate": 2.3018288274491276e-06, "loss": 0.0015, "step": 34300 }, { "epoch": 6.29, "learning_rate": 2.3013549819169146e-06, "loss": 0.0021, "step": 34301 }, { "epoch": 6.29, "learning_rate": 2.3008811788198106e-06, "loss": 0.0019, "step": 34302 }, { "epoch": 6.29, "learning_rate": 2.3004074181604306e-06, "loss": 0.0012, "step": 34303 }, { "epoch": 6.29, "learning_rate": 2.2999336999413845e-06, "loss": 0.0018, "step": 34304 }, { "epoch": 6.29, "learning_rate": 2.299460024165282e-06, "loss": 0.0022, "step": 34305 }, { "epoch": 6.29, "learning_rate": 2.2989863908347386e-06, "loss": 0.0022, "step": 34306 }, { "epoch": 6.29, "learning_rate": 2.2985127999523658e-06, "loss": 0.0036, "step": 34307 }, { "epoch": 6.29, "learning_rate": 2.298039251520768e-06, "loss": 0.0031, "step": 34308 }, { "epoch": 6.29, "learning_rate": 2.297565745542556e-06, "loss": 0.0016, "step": 34309 }, { "epoch": 6.29, "learning_rate": 2.2970922820203446e-06, "loss": 0.0024, "step": 34310 }, { "epoch": 6.29, "learning_rate": 2.2966188609567398e-06, "loss": 0.0019, "step": 34311 }, { "epoch": 6.29, "learning_rate": 2.296145482354354e-06, "loss": 0.0026, "step": 34312 }, { "epoch": 6.29, "learning_rate": 2.2956721462157938e-06, "loss": 0.0017, "step": 34313 }, { "epoch": 6.29, "learning_rate": 2.2951988525436695e-06, "loss": 0.0016, "step": 34314 }, { "epoch": 6.29, "learning_rate": 2.2947256013405892e-06, "loss": 0.0023, "step": 34315 }, { "epoch": 6.29, "learning_rate": 2.294252392609162e-06, "loss": 0.0016, "step": 34316 }, { "epoch": 6.29, "learning_rate": 2.2937792263519966e-06, "loss": 0.0012, "step": 34317 }, { "epoch": 6.29, "learning_rate": 2.2933061025717007e-06, "loss": 0.0026, "step": 34318 }, { "epoch": 6.29, "learning_rate": 2.2928330212708817e-06, "loss": 0.0021, "step": 34319 }, { "epoch": 6.29, "learning_rate": 2.292359982452148e-06, "loss": 0.0025, "step": 34320 }, { "epoch": 6.29, "learning_rate": 2.291886986118107e-06, "loss": 0.0024, "step": 34321 }, { "epoch": 6.29, "learning_rate": 2.2914140322713617e-06, "loss": 0.0026, "step": 34322 }, { "epoch": 6.29, "learning_rate": 2.290941120914526e-06, "loss": 0.0015, "step": 34323 }, { "epoch": 6.29, "learning_rate": 2.290468252050204e-06, "loss": 0.0021, "step": 34324 }, { "epoch": 6.29, "learning_rate": 2.2899954256810042e-06, "loss": 0.002, "step": 34325 }, { "epoch": 6.29, "learning_rate": 2.289522641809526e-06, "loss": 0.0029, "step": 34326 }, { "epoch": 6.3, "learning_rate": 2.289049900438379e-06, "loss": 0.0034, "step": 34327 }, { "epoch": 6.3, "learning_rate": 2.288577201570171e-06, "loss": 0.0025, "step": 34328 }, { "epoch": 6.3, "learning_rate": 2.288104545207507e-06, "loss": 0.0016, "step": 34329 }, { "epoch": 6.3, "learning_rate": 2.2876319313529906e-06, "loss": 0.0019, "step": 34330 }, { "epoch": 6.3, "learning_rate": 2.287159360009228e-06, "loss": 0.0024, "step": 34331 }, { "epoch": 6.3, "learning_rate": 2.2866868311788237e-06, "loss": 0.0019, "step": 34332 }, { "epoch": 6.3, "learning_rate": 2.2862143448643825e-06, "loss": 0.0015, "step": 34333 }, { "epoch": 6.3, "learning_rate": 2.2857419010685087e-06, "loss": 0.0022, "step": 34334 }, { "epoch": 6.3, "learning_rate": 2.285269499793806e-06, "loss": 0.0018, "step": 34335 }, { "epoch": 6.3, "learning_rate": 2.2847971410428783e-06, "loss": 0.0012, "step": 34336 }, { "epoch": 6.3, "learning_rate": 2.28432482481833e-06, "loss": 0.0035, "step": 34337 }, { "epoch": 6.3, "learning_rate": 2.2838525511227637e-06, "loss": 0.0022, "step": 34338 }, { "epoch": 6.3, "learning_rate": 2.2833803199587836e-06, "loss": 0.0018, "step": 34339 }, { "epoch": 6.3, "learning_rate": 2.282908131328988e-06, "loss": 0.0025, "step": 34340 }, { "epoch": 6.3, "learning_rate": 2.2824359852359867e-06, "loss": 0.0016, "step": 34341 }, { "epoch": 6.3, "learning_rate": 2.2819638816823796e-06, "loss": 0.0035, "step": 34342 }, { "epoch": 6.3, "learning_rate": 2.2814918206707705e-06, "loss": 0.0018, "step": 34343 }, { "epoch": 6.3, "learning_rate": 2.2810198022037566e-06, "loss": 0.0015, "step": 34344 }, { "epoch": 6.3, "learning_rate": 2.28054782628394e-06, "loss": 0.0024, "step": 34345 }, { "epoch": 6.3, "learning_rate": 2.2800758929139276e-06, "loss": 0.002, "step": 34346 }, { "epoch": 6.3, "learning_rate": 2.2796040020963175e-06, "loss": 0.0026, "step": 34347 }, { "epoch": 6.3, "learning_rate": 2.279132153833711e-06, "loss": 0.0021, "step": 34348 }, { "epoch": 6.3, "learning_rate": 2.2786603481287084e-06, "loss": 0.0017, "step": 34349 }, { "epoch": 6.3, "learning_rate": 2.2781885849839115e-06, "loss": 0.0021, "step": 34350 }, { "epoch": 6.3, "learning_rate": 2.2777168644019197e-06, "loss": 0.0026, "step": 34351 }, { "epoch": 6.3, "learning_rate": 2.277245186385334e-06, "loss": 0.0025, "step": 34352 }, { "epoch": 6.3, "learning_rate": 2.2767735509367506e-06, "loss": 0.0027, "step": 34353 }, { "epoch": 6.3, "learning_rate": 2.2763019580587776e-06, "loss": 0.002, "step": 34354 }, { "epoch": 6.3, "learning_rate": 2.275830407754006e-06, "loss": 0.0033, "step": 34355 }, { "epoch": 6.3, "learning_rate": 2.27535890002504e-06, "loss": 0.0021, "step": 34356 }, { "epoch": 6.3, "learning_rate": 2.2748874348744755e-06, "loss": 0.0014, "step": 34357 }, { "epoch": 6.3, "learning_rate": 2.2744160123049084e-06, "loss": 0.0029, "step": 34358 }, { "epoch": 6.3, "learning_rate": 2.273944632318946e-06, "loss": 0.0028, "step": 34359 }, { "epoch": 6.3, "learning_rate": 2.2734732949191805e-06, "loss": 0.0019, "step": 34360 }, { "epoch": 6.3, "learning_rate": 2.2730020001082154e-06, "loss": 0.002, "step": 34361 }, { "epoch": 6.3, "learning_rate": 2.272530747888637e-06, "loss": 0.0029, "step": 34362 }, { "epoch": 6.3, "learning_rate": 2.272059538263054e-06, "loss": 0.0015, "step": 34363 }, { "epoch": 6.3, "learning_rate": 2.2715883712340604e-06, "loss": 0.003, "step": 34364 }, { "epoch": 6.3, "learning_rate": 2.271117246804252e-06, "loss": 0.0014, "step": 34365 }, { "epoch": 6.3, "learning_rate": 2.270646164976226e-06, "loss": 0.0024, "step": 34366 }, { "epoch": 6.3, "learning_rate": 2.2701751257525796e-06, "loss": 0.0022, "step": 34367 }, { "epoch": 6.3, "learning_rate": 2.26970412913591e-06, "loss": 0.0017, "step": 34368 }, { "epoch": 6.3, "learning_rate": 2.269233175128811e-06, "loss": 0.0025, "step": 34369 }, { "epoch": 6.3, "learning_rate": 2.2687622637338814e-06, "loss": 0.0013, "step": 34370 }, { "epoch": 6.3, "learning_rate": 2.268291394953711e-06, "loss": 0.0021, "step": 34371 }, { "epoch": 6.3, "learning_rate": 2.267820568790906e-06, "loss": 0.0028, "step": 34372 }, { "epoch": 6.3, "learning_rate": 2.267349785248052e-06, "loss": 0.0027, "step": 34373 }, { "epoch": 6.3, "learning_rate": 2.2668790443277468e-06, "loss": 0.0027, "step": 34374 }, { "epoch": 6.3, "learning_rate": 2.2664083460325837e-06, "loss": 0.0021, "step": 34375 }, { "epoch": 6.3, "learning_rate": 2.265937690365161e-06, "loss": 0.0024, "step": 34376 }, { "epoch": 6.3, "learning_rate": 2.2654670773280707e-06, "loss": 0.0031, "step": 34377 }, { "epoch": 6.3, "learning_rate": 2.2649965069239077e-06, "loss": 0.0023, "step": 34378 }, { "epoch": 6.3, "learning_rate": 2.2645259791552675e-06, "loss": 0.0022, "step": 34379 }, { "epoch": 6.3, "learning_rate": 2.264055494024736e-06, "loss": 0.0021, "step": 34380 }, { "epoch": 6.3, "learning_rate": 2.2635850515349156e-06, "loss": 0.0026, "step": 34381 }, { "epoch": 6.31, "learning_rate": 2.263114651688394e-06, "loss": 0.0028, "step": 34382 }, { "epoch": 6.31, "learning_rate": 2.262644294487767e-06, "loss": 0.002, "step": 34383 }, { "epoch": 6.31, "learning_rate": 2.2621739799356244e-06, "loss": 0.0013, "step": 34384 }, { "epoch": 6.31, "learning_rate": 2.2617037080345617e-06, "loss": 0.0019, "step": 34385 }, { "epoch": 6.31, "learning_rate": 2.261233478787169e-06, "loss": 0.003, "step": 34386 }, { "epoch": 6.31, "learning_rate": 2.260763292196039e-06, "loss": 0.0057, "step": 34387 }, { "epoch": 6.31, "learning_rate": 2.2602931482637613e-06, "loss": 0.0025, "step": 34388 }, { "epoch": 6.31, "learning_rate": 2.2598230469929283e-06, "loss": 0.0014, "step": 34389 }, { "epoch": 6.31, "learning_rate": 2.2593529883861365e-06, "loss": 0.0012, "step": 34390 }, { "epoch": 6.31, "learning_rate": 2.2588829724459694e-06, "loss": 0.0038, "step": 34391 }, { "epoch": 6.31, "learning_rate": 2.2584129991750212e-06, "loss": 0.002, "step": 34392 }, { "epoch": 6.31, "learning_rate": 2.2579430685758795e-06, "loss": 0.0017, "step": 34393 }, { "epoch": 6.31, "learning_rate": 2.2574731806511397e-06, "loss": 0.0033, "step": 34394 }, { "epoch": 6.31, "learning_rate": 2.257003335403388e-06, "loss": 0.0022, "step": 34395 }, { "epoch": 6.31, "learning_rate": 2.256533532835219e-06, "loss": 0.0029, "step": 34396 }, { "epoch": 6.31, "learning_rate": 2.256063772949213e-06, "loss": 0.0029, "step": 34397 }, { "epoch": 6.31, "learning_rate": 2.255594055747967e-06, "loss": 0.0023, "step": 34398 }, { "epoch": 6.31, "learning_rate": 2.2551243812340696e-06, "loss": 0.0014, "step": 34399 }, { "epoch": 6.31, "learning_rate": 2.254654749410107e-06, "loss": 0.0019, "step": 34400 }, { "epoch": 6.31, "learning_rate": 2.254185160278669e-06, "loss": 0.0022, "step": 34401 }, { "epoch": 6.31, "learning_rate": 2.2537156138423433e-06, "loss": 0.0023, "step": 34402 }, { "epoch": 6.31, "learning_rate": 2.2532461101037197e-06, "loss": 0.0024, "step": 34403 }, { "epoch": 6.31, "learning_rate": 2.2527766490653835e-06, "loss": 0.0015, "step": 34404 }, { "epoch": 6.31, "learning_rate": 2.2523072307299253e-06, "loss": 0.0021, "step": 34405 }, { "epoch": 6.31, "learning_rate": 2.2518378550999274e-06, "loss": 0.001, "step": 34406 }, { "epoch": 6.31, "learning_rate": 2.2513685221779846e-06, "loss": 0.0023, "step": 34407 }, { "epoch": 6.31, "learning_rate": 2.250899231966682e-06, "loss": 0.0019, "step": 34408 }, { "epoch": 6.31, "learning_rate": 2.250429984468602e-06, "loss": 0.0015, "step": 34409 }, { "epoch": 6.31, "learning_rate": 2.249960779686333e-06, "loss": 0.002, "step": 34410 }, { "epoch": 6.31, "learning_rate": 2.249491617622459e-06, "loss": 0.0029, "step": 34411 }, { "epoch": 6.31, "learning_rate": 2.2490224982795718e-06, "loss": 0.0017, "step": 34412 }, { "epoch": 6.31, "learning_rate": 2.2485534216602544e-06, "loss": 0.0024, "step": 34413 }, { "epoch": 6.31, "learning_rate": 2.2480843877670954e-06, "loss": 0.0036, "step": 34414 }, { "epoch": 6.31, "learning_rate": 2.2476153966026705e-06, "loss": 0.0032, "step": 34415 }, { "epoch": 6.31, "learning_rate": 2.247146448169574e-06, "loss": 0.0016, "step": 34416 }, { "epoch": 6.31, "learning_rate": 2.246677542470388e-06, "loss": 0.0024, "step": 34417 }, { "epoch": 6.31, "learning_rate": 2.2462086795076963e-06, "loss": 0.0023, "step": 34418 }, { "epoch": 6.31, "learning_rate": 2.245739859284085e-06, "loss": 0.0015, "step": 34419 }, { "epoch": 6.31, "learning_rate": 2.245271081802137e-06, "loss": 0.002, "step": 34420 }, { "epoch": 6.31, "learning_rate": 2.2448023470644366e-06, "loss": 0.0015, "step": 34421 }, { "epoch": 6.31, "learning_rate": 2.2443336550735674e-06, "loss": 0.0014, "step": 34422 }, { "epoch": 6.31, "learning_rate": 2.2438650058321123e-06, "loss": 0.0034, "step": 34423 }, { "epoch": 6.31, "learning_rate": 2.243396399342652e-06, "loss": 0.0031, "step": 34424 }, { "epoch": 6.31, "learning_rate": 2.2429278356077766e-06, "loss": 0.0021, "step": 34425 }, { "epoch": 6.31, "learning_rate": 2.2424593146300664e-06, "loss": 0.001, "step": 34426 }, { "epoch": 6.31, "learning_rate": 2.241990836412099e-06, "loss": 0.0018, "step": 34427 }, { "epoch": 6.31, "learning_rate": 2.2415224009564586e-06, "loss": 0.0017, "step": 34428 }, { "epoch": 6.31, "learning_rate": 2.2410540082657295e-06, "loss": 0.0011, "step": 34429 }, { "epoch": 6.31, "learning_rate": 2.2405856583424936e-06, "loss": 0.0039, "step": 34430 }, { "epoch": 6.31, "learning_rate": 2.2401173511893317e-06, "loss": 0.0013, "step": 34431 }, { "epoch": 6.31, "learning_rate": 2.239649086808826e-06, "loss": 0.0024, "step": 34432 }, { "epoch": 6.31, "learning_rate": 2.239180865203552e-06, "loss": 0.0029, "step": 34433 }, { "epoch": 6.31, "learning_rate": 2.2387126863760978e-06, "loss": 0.0016, "step": 34434 }, { "epoch": 6.31, "learning_rate": 2.2382445503290406e-06, "loss": 0.0015, "step": 34435 }, { "epoch": 6.32, "learning_rate": 2.237776457064962e-06, "loss": 0.0012, "step": 34436 }, { "epoch": 6.32, "learning_rate": 2.2373084065864402e-06, "loss": 0.0013, "step": 34437 }, { "epoch": 6.32, "learning_rate": 2.2368403988960564e-06, "loss": 0.0026, "step": 34438 }, { "epoch": 6.32, "learning_rate": 2.236372433996391e-06, "loss": 0.0029, "step": 34439 }, { "epoch": 6.32, "learning_rate": 2.2359045118900224e-06, "loss": 0.0024, "step": 34440 }, { "epoch": 6.32, "learning_rate": 2.2354366325795295e-06, "loss": 0.0014, "step": 34441 }, { "epoch": 6.32, "learning_rate": 2.2349687960674904e-06, "loss": 0.0023, "step": 34442 }, { "epoch": 6.32, "learning_rate": 2.2345010023564895e-06, "loss": 0.0015, "step": 34443 }, { "epoch": 6.32, "learning_rate": 2.2340332514490983e-06, "loss": 0.0031, "step": 34444 }, { "epoch": 6.32, "learning_rate": 2.2335655433478978e-06, "loss": 0.0009, "step": 34445 }, { "epoch": 6.32, "learning_rate": 2.2330978780554637e-06, "loss": 0.0031, "step": 34446 }, { "epoch": 6.32, "learning_rate": 2.2326302555743783e-06, "loss": 0.0018, "step": 34447 }, { "epoch": 6.32, "learning_rate": 2.232162675907217e-06, "loss": 0.0037, "step": 34448 }, { "epoch": 6.32, "learning_rate": 2.2316951390565577e-06, "loss": 0.0024, "step": 34449 }, { "epoch": 6.32, "learning_rate": 2.2312276450249758e-06, "loss": 0.0024, "step": 34450 }, { "epoch": 6.32, "learning_rate": 2.230760193815049e-06, "loss": 0.0023, "step": 34451 }, { "epoch": 6.32, "learning_rate": 2.2302927854293555e-06, "loss": 0.0025, "step": 34452 }, { "epoch": 6.32, "learning_rate": 2.2298254198704693e-06, "loss": 0.0017, "step": 34453 }, { "epoch": 6.32, "learning_rate": 2.2293580971409667e-06, "loss": 0.0019, "step": 34454 }, { "epoch": 6.32, "learning_rate": 2.2288908172434254e-06, "loss": 0.0024, "step": 34455 }, { "epoch": 6.32, "learning_rate": 2.22842358018042e-06, "loss": 0.0021, "step": 34456 }, { "epoch": 6.32, "learning_rate": 2.227956385954525e-06, "loss": 0.0022, "step": 34457 }, { "epoch": 6.32, "learning_rate": 2.2274892345683176e-06, "loss": 0.0024, "step": 34458 }, { "epoch": 6.32, "learning_rate": 2.2270221260243675e-06, "loss": 0.0019, "step": 34459 }, { "epoch": 6.32, "learning_rate": 2.2265550603252573e-06, "loss": 0.0018, "step": 34460 }, { "epoch": 6.32, "learning_rate": 2.2260880374735603e-06, "loss": 0.0021, "step": 34461 }, { "epoch": 6.32, "learning_rate": 2.225621057471845e-06, "loss": 0.0013, "step": 34462 }, { "epoch": 6.32, "learning_rate": 2.225154120322689e-06, "loss": 0.0016, "step": 34463 }, { "epoch": 6.32, "learning_rate": 2.2246872260286624e-06, "loss": 0.0019, "step": 34464 }, { "epoch": 6.32, "learning_rate": 2.224220374592345e-06, "loss": 0.0026, "step": 34465 }, { "epoch": 6.32, "learning_rate": 2.223753566016308e-06, "loss": 0.0016, "step": 34466 }, { "epoch": 6.32, "learning_rate": 2.2232868003031217e-06, "loss": 0.0014, "step": 34467 }, { "epoch": 6.32, "learning_rate": 2.222820077455362e-06, "loss": 0.0011, "step": 34468 }, { "epoch": 6.32, "learning_rate": 2.2223533974755996e-06, "loss": 0.0017, "step": 34469 }, { "epoch": 6.32, "learning_rate": 2.2218867603664083e-06, "loss": 0.002, "step": 34470 }, { "epoch": 6.32, "learning_rate": 2.2214201661303593e-06, "loss": 0.0021, "step": 34471 }, { "epoch": 6.32, "learning_rate": 2.2209536147700238e-06, "loss": 0.0026, "step": 34472 }, { "epoch": 6.32, "learning_rate": 2.220487106287974e-06, "loss": 0.0012, "step": 34473 }, { "epoch": 6.32, "learning_rate": 2.220020640686783e-06, "loss": 0.0021, "step": 34474 }, { "epoch": 6.32, "learning_rate": 2.219554217969019e-06, "loss": 0.0033, "step": 34475 }, { "epoch": 6.32, "learning_rate": 2.219087838137255e-06, "loss": 0.0025, "step": 34476 }, { "epoch": 6.32, "learning_rate": 2.2186215011940593e-06, "loss": 0.0019, "step": 34477 }, { "epoch": 6.32, "learning_rate": 2.2181552071420055e-06, "loss": 0.0023, "step": 34478 }, { "epoch": 6.32, "learning_rate": 2.2176889559836655e-06, "loss": 0.0025, "step": 34479 }, { "epoch": 6.32, "learning_rate": 2.2172227477216045e-06, "loss": 0.0024, "step": 34480 }, { "epoch": 6.32, "learning_rate": 2.21675658235839e-06, "loss": 0.0024, "step": 34481 }, { "epoch": 6.32, "learning_rate": 2.2162904598965997e-06, "loss": 0.0024, "step": 34482 }, { "epoch": 6.32, "learning_rate": 2.215824380338798e-06, "loss": 0.0029, "step": 34483 }, { "epoch": 6.32, "learning_rate": 2.2153583436875546e-06, "loss": 0.0017, "step": 34484 }, { "epoch": 6.32, "learning_rate": 2.2148923499454377e-06, "loss": 0.0023, "step": 34485 }, { "epoch": 6.32, "learning_rate": 2.2144263991150183e-06, "loss": 0.0023, "step": 34486 }, { "epoch": 6.32, "learning_rate": 2.213960491198862e-06, "loss": 0.0019, "step": 34487 }, { "epoch": 6.32, "learning_rate": 2.213494626199537e-06, "loss": 0.0018, "step": 34488 }, { "epoch": 6.32, "learning_rate": 2.2130288041196135e-06, "loss": 0.002, "step": 34489 }, { "epoch": 6.32, "learning_rate": 2.212563024961657e-06, "loss": 0.0019, "step": 34490 }, { "epoch": 6.33, "learning_rate": 2.2120972887282356e-06, "loss": 0.0028, "step": 34491 }, { "epoch": 6.33, "learning_rate": 2.211631595421917e-06, "loss": 0.0012, "step": 34492 }, { "epoch": 6.33, "learning_rate": 2.211165945045266e-06, "loss": 0.0029, "step": 34493 }, { "epoch": 6.33, "learning_rate": 2.210700337600852e-06, "loss": 0.003, "step": 34494 }, { "epoch": 6.33, "learning_rate": 2.2102347730912376e-06, "loss": 0.0016, "step": 34495 }, { "epoch": 6.33, "learning_rate": 2.209769251518994e-06, "loss": 0.0016, "step": 34496 }, { "epoch": 6.33, "learning_rate": 2.209303772886687e-06, "loss": 0.0024, "step": 34497 }, { "epoch": 6.33, "learning_rate": 2.208838337196878e-06, "loss": 0.0023, "step": 34498 }, { "epoch": 6.33, "learning_rate": 2.208372944452132e-06, "loss": 0.0028, "step": 34499 }, { "epoch": 6.33, "learning_rate": 2.207907594655019e-06, "loss": 0.002, "step": 34500 }, { "epoch": 6.33, "learning_rate": 2.2074422878081025e-06, "loss": 0.0021, "step": 34501 }, { "epoch": 6.33, "learning_rate": 2.2069770239139464e-06, "loss": 0.0031, "step": 34502 }, { "epoch": 6.33, "learning_rate": 2.2065118029751156e-06, "loss": 0.0023, "step": 34503 }, { "epoch": 6.33, "learning_rate": 2.2060466249941747e-06, "loss": 0.002, "step": 34504 }, { "epoch": 6.33, "learning_rate": 2.205581489973686e-06, "loss": 0.0014, "step": 34505 }, { "epoch": 6.33, "learning_rate": 2.205116397916216e-06, "loss": 0.0013, "step": 34506 }, { "epoch": 6.33, "learning_rate": 2.204651348824327e-06, "loss": 0.0021, "step": 34507 }, { "epoch": 6.33, "learning_rate": 2.2041863427005816e-06, "loss": 0.0026, "step": 34508 }, { "epoch": 6.33, "learning_rate": 2.2037213795475445e-06, "loss": 0.0021, "step": 34509 }, { "epoch": 6.33, "learning_rate": 2.2032564593677773e-06, "loss": 0.0047, "step": 34510 }, { "epoch": 6.33, "learning_rate": 2.2027915821638433e-06, "loss": 0.0014, "step": 34511 }, { "epoch": 6.33, "learning_rate": 2.2023267479383027e-06, "loss": 0.0029, "step": 34512 }, { "epoch": 6.33, "learning_rate": 2.2018619566937218e-06, "loss": 0.002, "step": 34513 }, { "epoch": 6.33, "learning_rate": 2.201397208432661e-06, "loss": 0.003, "step": 34514 }, { "epoch": 6.33, "learning_rate": 2.2009325031576835e-06, "loss": 0.0017, "step": 34515 }, { "epoch": 6.33, "learning_rate": 2.2004678408713463e-06, "loss": 0.0015, "step": 34516 }, { "epoch": 6.33, "learning_rate": 2.20000322157621e-06, "loss": 0.0017, "step": 34517 }, { "epoch": 6.33, "learning_rate": 2.1995386452748425e-06, "loss": 0.0017, "step": 34518 }, { "epoch": 6.33, "learning_rate": 2.1990741119698e-06, "loss": 0.0021, "step": 34519 }, { "epoch": 6.33, "learning_rate": 2.198609621663643e-06, "loss": 0.0024, "step": 34520 }, { "epoch": 6.33, "learning_rate": 2.198145174358933e-06, "loss": 0.002, "step": 34521 }, { "epoch": 6.33, "learning_rate": 2.19768077005823e-06, "loss": 0.0026, "step": 34522 }, { "epoch": 6.33, "learning_rate": 2.1972164087640933e-06, "loss": 0.0017, "step": 34523 }, { "epoch": 6.33, "learning_rate": 2.196752090479083e-06, "loss": 0.0023, "step": 34524 }, { "epoch": 6.33, "learning_rate": 2.1962878152057545e-06, "loss": 0.0016, "step": 34525 }, { "epoch": 6.33, "learning_rate": 2.1958235829466766e-06, "loss": 0.0017, "step": 34526 }, { "epoch": 6.33, "learning_rate": 2.1953593937043983e-06, "loss": 0.0013, "step": 34527 }, { "epoch": 6.33, "learning_rate": 2.194895247481482e-06, "loss": 0.0023, "step": 34528 }, { "epoch": 6.33, "learning_rate": 2.1944311442804865e-06, "loss": 0.0025, "step": 34529 }, { "epoch": 6.33, "learning_rate": 2.1939670841039675e-06, "loss": 0.003, "step": 34530 }, { "epoch": 6.33, "learning_rate": 2.1935030669544856e-06, "loss": 0.0019, "step": 34531 }, { "epoch": 6.33, "learning_rate": 2.193039092834599e-06, "loss": 0.0024, "step": 34532 }, { "epoch": 6.33, "learning_rate": 2.192575161746866e-06, "loss": 0.0019, "step": 34533 }, { "epoch": 6.33, "learning_rate": 2.1921112736938367e-06, "loss": 0.002, "step": 34534 }, { "epoch": 6.33, "learning_rate": 2.1916474286780744e-06, "loss": 0.0015, "step": 34535 }, { "epoch": 6.33, "learning_rate": 2.1911836267021357e-06, "loss": 0.0018, "step": 34536 }, { "epoch": 6.33, "learning_rate": 2.1907198677685746e-06, "loss": 0.0019, "step": 34537 }, { "epoch": 6.33, "learning_rate": 2.1902561518799502e-06, "loss": 0.0027, "step": 34538 }, { "epoch": 6.33, "learning_rate": 2.189792479038816e-06, "loss": 0.0016, "step": 34539 }, { "epoch": 6.33, "learning_rate": 2.1893288492477282e-06, "loss": 0.002, "step": 34540 }, { "epoch": 6.33, "learning_rate": 2.1888652625092433e-06, "loss": 0.0021, "step": 34541 }, { "epoch": 6.33, "learning_rate": 2.1884017188259153e-06, "loss": 0.0018, "step": 34542 }, { "epoch": 6.33, "learning_rate": 2.1879382182002985e-06, "loss": 0.0018, "step": 34543 }, { "epoch": 6.33, "learning_rate": 2.1874747606349543e-06, "loss": 0.0016, "step": 34544 }, { "epoch": 6.34, "learning_rate": 2.1870113461324292e-06, "loss": 0.0011, "step": 34545 }, { "epoch": 6.34, "learning_rate": 2.18654797469528e-06, "loss": 0.0031, "step": 34546 }, { "epoch": 6.34, "learning_rate": 2.1860846463260633e-06, "loss": 0.0017, "step": 34547 }, { "epoch": 6.34, "learning_rate": 2.1856213610273267e-06, "loss": 0.0016, "step": 34548 }, { "epoch": 6.34, "learning_rate": 2.1851581188016313e-06, "loss": 0.002, "step": 34549 }, { "epoch": 6.34, "learning_rate": 2.18469491965153e-06, "loss": 0.0016, "step": 34550 }, { "epoch": 6.34, "learning_rate": 2.1842317635795716e-06, "loss": 0.0022, "step": 34551 }, { "epoch": 6.34, "learning_rate": 2.1837686505883073e-06, "loss": 0.0035, "step": 34552 }, { "epoch": 6.34, "learning_rate": 2.183305580680296e-06, "loss": 0.0023, "step": 34553 }, { "epoch": 6.34, "learning_rate": 2.1828425538580878e-06, "loss": 0.0026, "step": 34554 }, { "epoch": 6.34, "learning_rate": 2.1823795701242346e-06, "loss": 0.002, "step": 34555 }, { "epoch": 6.34, "learning_rate": 2.1819166294812887e-06, "loss": 0.0014, "step": 34556 }, { "epoch": 6.34, "learning_rate": 2.181453731931801e-06, "loss": 0.0015, "step": 34557 }, { "epoch": 6.34, "learning_rate": 2.1809908774783227e-06, "loss": 0.0014, "step": 34558 }, { "epoch": 6.34, "learning_rate": 2.1805280661234073e-06, "loss": 0.0029, "step": 34559 }, { "epoch": 6.34, "learning_rate": 2.1800652978696037e-06, "loss": 0.0019, "step": 34560 }, { "epoch": 6.34, "learning_rate": 2.1796025727194603e-06, "loss": 0.0032, "step": 34561 }, { "epoch": 6.34, "learning_rate": 2.179139890675537e-06, "loss": 0.0014, "step": 34562 }, { "epoch": 6.34, "learning_rate": 2.1786772517403756e-06, "loss": 0.0019, "step": 34563 }, { "epoch": 6.34, "learning_rate": 2.1782146559165274e-06, "loss": 0.0014, "step": 34564 }, { "epoch": 6.34, "learning_rate": 2.17775210320654e-06, "loss": 0.0021, "step": 34565 }, { "epoch": 6.34, "learning_rate": 2.17728959361297e-06, "loss": 0.0022, "step": 34566 }, { "epoch": 6.34, "learning_rate": 2.1768271271383623e-06, "loss": 0.0037, "step": 34567 }, { "epoch": 6.34, "learning_rate": 2.17636470378527e-06, "loss": 0.0031, "step": 34568 }, { "epoch": 6.34, "learning_rate": 2.175902323556236e-06, "loss": 0.0025, "step": 34569 }, { "epoch": 6.34, "learning_rate": 2.175439986453809e-06, "loss": 0.0017, "step": 34570 }, { "epoch": 6.34, "learning_rate": 2.174977692480542e-06, "loss": 0.0015, "step": 34571 }, { "epoch": 6.34, "learning_rate": 2.1745154416389823e-06, "loss": 0.0027, "step": 34572 }, { "epoch": 6.34, "learning_rate": 2.1740532339316767e-06, "loss": 0.0019, "step": 34573 }, { "epoch": 6.34, "learning_rate": 2.1735910693611727e-06, "loss": 0.003, "step": 34574 }, { "epoch": 6.34, "learning_rate": 2.173128947930019e-06, "loss": 0.0019, "step": 34575 }, { "epoch": 6.34, "learning_rate": 2.1726668696407605e-06, "loss": 0.002, "step": 34576 }, { "epoch": 6.34, "learning_rate": 2.1722048344959468e-06, "loss": 0.002, "step": 34577 }, { "epoch": 6.34, "learning_rate": 2.1717428424981203e-06, "loss": 0.0021, "step": 34578 }, { "epoch": 6.34, "learning_rate": 2.171280893649832e-06, "loss": 0.0027, "step": 34579 }, { "epoch": 6.34, "learning_rate": 2.170818987953631e-06, "loss": 0.0014, "step": 34580 }, { "epoch": 6.34, "learning_rate": 2.170357125412057e-06, "loss": 0.0029, "step": 34581 }, { "epoch": 6.34, "learning_rate": 2.1698953060276564e-06, "loss": 0.002, "step": 34582 }, { "epoch": 6.34, "learning_rate": 2.1694335298029744e-06, "loss": 0.0015, "step": 34583 }, { "epoch": 6.34, "learning_rate": 2.168971796740561e-06, "loss": 0.0026, "step": 34584 }, { "epoch": 6.34, "learning_rate": 2.168510106842958e-06, "loss": 0.0022, "step": 34585 }, { "epoch": 6.34, "learning_rate": 2.1680484601127137e-06, "loss": 0.0045, "step": 34586 }, { "epoch": 6.34, "learning_rate": 2.167586856552365e-06, "loss": 0.0019, "step": 34587 }, { "epoch": 6.34, "learning_rate": 2.1671252961644627e-06, "loss": 0.0013, "step": 34588 }, { "epoch": 6.34, "learning_rate": 2.16666377895155e-06, "loss": 0.0027, "step": 34589 }, { "epoch": 6.34, "learning_rate": 2.1662023049161707e-06, "loss": 0.0028, "step": 34590 }, { "epoch": 6.34, "learning_rate": 2.1657408740608666e-06, "loss": 0.0027, "step": 34591 }, { "epoch": 6.34, "learning_rate": 2.165279486388183e-06, "loss": 0.0032, "step": 34592 }, { "epoch": 6.34, "learning_rate": 2.164818141900662e-06, "loss": 0.0022, "step": 34593 }, { "epoch": 6.34, "learning_rate": 2.1643568406008476e-06, "loss": 0.0021, "step": 34594 }, { "epoch": 6.34, "learning_rate": 2.1638955824912812e-06, "loss": 0.0047, "step": 34595 }, { "epoch": 6.34, "learning_rate": 2.163434367574503e-06, "loss": 0.0028, "step": 34596 }, { "epoch": 6.34, "learning_rate": 2.1629731958530644e-06, "loss": 0.0019, "step": 34597 }, { "epoch": 6.34, "learning_rate": 2.1625120673294974e-06, "loss": 0.0013, "step": 34598 }, { "epoch": 6.34, "learning_rate": 2.1620509820063484e-06, "loss": 0.0015, "step": 34599 }, { "epoch": 6.35, "learning_rate": 2.1615899398861585e-06, "loss": 0.0012, "step": 34600 }, { "epoch": 6.35, "learning_rate": 2.161128940971464e-06, "loss": 0.0025, "step": 34601 }, { "epoch": 6.35, "learning_rate": 2.160667985264815e-06, "loss": 0.002, "step": 34602 }, { "epoch": 6.35, "learning_rate": 2.1602070727687463e-06, "loss": 0.0028, "step": 34603 }, { "epoch": 6.35, "learning_rate": 2.159746203485803e-06, "loss": 0.0016, "step": 34604 }, { "epoch": 6.35, "learning_rate": 2.1592853774185173e-06, "loss": 0.0021, "step": 34605 }, { "epoch": 6.35, "learning_rate": 2.1588245945694366e-06, "loss": 0.0019, "step": 34606 }, { "epoch": 6.35, "learning_rate": 2.158363854941099e-06, "loss": 0.0022, "step": 34607 }, { "epoch": 6.35, "learning_rate": 2.1579031585360435e-06, "loss": 0.0026, "step": 34608 }, { "epoch": 6.35, "learning_rate": 2.157442505356808e-06, "loss": 0.0027, "step": 34609 }, { "epoch": 6.35, "learning_rate": 2.156981895405935e-06, "loss": 0.0016, "step": 34610 }, { "epoch": 6.35, "learning_rate": 2.156521328685961e-06, "loss": 0.0015, "step": 34611 }, { "epoch": 6.35, "learning_rate": 2.1560608051994246e-06, "loss": 0.0013, "step": 34612 }, { "epoch": 6.35, "learning_rate": 2.1556003249488654e-06, "loss": 0.0018, "step": 34613 }, { "epoch": 6.35, "learning_rate": 2.155139887936818e-06, "loss": 0.0007, "step": 34614 }, { "epoch": 6.35, "learning_rate": 2.154679494165829e-06, "loss": 0.0022, "step": 34615 }, { "epoch": 6.35, "learning_rate": 2.1542191436384275e-06, "loss": 0.0018, "step": 34616 }, { "epoch": 6.35, "learning_rate": 2.1537588363571538e-06, "loss": 0.0019, "step": 34617 }, { "epoch": 6.35, "learning_rate": 2.1532985723245425e-06, "loss": 0.0011, "step": 34618 }, { "epoch": 6.35, "learning_rate": 2.152838351543136e-06, "loss": 0.0016, "step": 34619 }, { "epoch": 6.35, "learning_rate": 2.1523781740154683e-06, "loss": 0.0017, "step": 34620 }, { "epoch": 6.35, "learning_rate": 2.1519180397440765e-06, "loss": 0.0029, "step": 34621 }, { "epoch": 6.35, "learning_rate": 2.151457948731499e-06, "loss": 0.0019, "step": 34622 }, { "epoch": 6.35, "learning_rate": 2.150997900980263e-06, "loss": 0.0024, "step": 34623 }, { "epoch": 6.35, "learning_rate": 2.1505378964929137e-06, "loss": 0.0021, "step": 34624 }, { "epoch": 6.35, "learning_rate": 2.1500779352719836e-06, "loss": 0.0018, "step": 34625 }, { "epoch": 6.35, "learning_rate": 2.1496180173200075e-06, "loss": 0.0025, "step": 34626 }, { "epoch": 6.35, "learning_rate": 2.14915814263952e-06, "loss": 0.002, "step": 34627 }, { "epoch": 6.35, "learning_rate": 2.1486983112330583e-06, "loss": 0.0013, "step": 34628 }, { "epoch": 6.35, "learning_rate": 2.1482385231031543e-06, "loss": 0.0012, "step": 34629 }, { "epoch": 6.35, "learning_rate": 2.147778778252344e-06, "loss": 0.0023, "step": 34630 }, { "epoch": 6.35, "learning_rate": 2.1473190766831587e-06, "loss": 0.0014, "step": 34631 }, { "epoch": 6.35, "learning_rate": 2.146859418398136e-06, "loss": 0.0021, "step": 34632 }, { "epoch": 6.35, "learning_rate": 2.1463998033998124e-06, "loss": 0.0019, "step": 34633 }, { "epoch": 6.35, "learning_rate": 2.145940231690713e-06, "loss": 0.0026, "step": 34634 }, { "epoch": 6.35, "learning_rate": 2.145480703273376e-06, "loss": 0.0024, "step": 34635 }, { "epoch": 6.35, "learning_rate": 2.1450212181503305e-06, "loss": 0.0031, "step": 34636 }, { "epoch": 6.35, "learning_rate": 2.1445617763241146e-06, "loss": 0.0023, "step": 34637 }, { "epoch": 6.35, "learning_rate": 2.144102377797258e-06, "loss": 0.0023, "step": 34638 }, { "epoch": 6.35, "learning_rate": 2.1436430225722937e-06, "loss": 0.002, "step": 34639 }, { "epoch": 6.35, "learning_rate": 2.143183710651753e-06, "loss": 0.0023, "step": 34640 }, { "epoch": 6.35, "learning_rate": 2.1427244420381675e-06, "loss": 0.0019, "step": 34641 }, { "epoch": 6.35, "learning_rate": 2.1422652167340695e-06, "loss": 0.002, "step": 34642 }, { "epoch": 6.35, "learning_rate": 2.141806034741989e-06, "loss": 0.0024, "step": 34643 }, { "epoch": 6.35, "learning_rate": 2.141346896064458e-06, "loss": 0.0029, "step": 34644 }, { "epoch": 6.35, "learning_rate": 2.1408878007040067e-06, "loss": 0.0034, "step": 34645 }, { "epoch": 6.35, "learning_rate": 2.140428748663167e-06, "loss": 0.0019, "step": 34646 }, { "epoch": 6.35, "learning_rate": 2.1399697399444674e-06, "loss": 0.0019, "step": 34647 }, { "epoch": 6.35, "learning_rate": 2.1395107745504384e-06, "loss": 0.0025, "step": 34648 }, { "epoch": 6.35, "learning_rate": 2.1390518524836078e-06, "loss": 0.0019, "step": 34649 }, { "epoch": 6.35, "learning_rate": 2.1385929737465104e-06, "loss": 0.0011, "step": 34650 }, { "epoch": 6.35, "learning_rate": 2.1381341383416756e-06, "loss": 0.0022, "step": 34651 }, { "epoch": 6.35, "learning_rate": 2.137675346271626e-06, "loss": 0.0019, "step": 34652 }, { "epoch": 6.35, "learning_rate": 2.1372165975388925e-06, "loss": 0.0023, "step": 34653 }, { "epoch": 6.36, "learning_rate": 2.1367578921460075e-06, "loss": 0.0017, "step": 34654 }, { "epoch": 6.36, "learning_rate": 2.1362992300954976e-06, "loss": 0.0041, "step": 34655 }, { "epoch": 6.36, "learning_rate": 2.13584061138989e-06, "loss": 0.0028, "step": 34656 }, { "epoch": 6.36, "learning_rate": 2.135382036031716e-06, "loss": 0.0019, "step": 34657 }, { "epoch": 6.36, "learning_rate": 2.1349235040234952e-06, "loss": 0.0018, "step": 34658 }, { "epoch": 6.36, "learning_rate": 2.1344650153677625e-06, "loss": 0.0021, "step": 34659 }, { "epoch": 6.36, "learning_rate": 2.134006570067044e-06, "loss": 0.0032, "step": 34660 }, { "epoch": 6.36, "learning_rate": 2.1335481681238644e-06, "loss": 0.0035, "step": 34661 }, { "epoch": 6.36, "learning_rate": 2.1330898095407528e-06, "loss": 0.0023, "step": 34662 }, { "epoch": 6.36, "learning_rate": 2.132631494320233e-06, "loss": 0.0014, "step": 34663 }, { "epoch": 6.36, "learning_rate": 2.132173222464834e-06, "loss": 0.0021, "step": 34664 }, { "epoch": 6.36, "learning_rate": 2.13171499397708e-06, "loss": 0.0017, "step": 34665 }, { "epoch": 6.36, "learning_rate": 2.1312568088594977e-06, "loss": 0.0015, "step": 34666 }, { "epoch": 6.36, "learning_rate": 2.1307986671146086e-06, "loss": 0.0019, "step": 34667 }, { "epoch": 6.36, "learning_rate": 2.130340568744945e-06, "loss": 0.0013, "step": 34668 }, { "epoch": 6.36, "learning_rate": 2.129882513753031e-06, "loss": 0.0037, "step": 34669 }, { "epoch": 6.36, "learning_rate": 2.1294245021413864e-06, "loss": 0.0024, "step": 34670 }, { "epoch": 6.36, "learning_rate": 2.128966533912535e-06, "loss": 0.0024, "step": 34671 }, { "epoch": 6.36, "learning_rate": 2.128508609069009e-06, "loss": 0.0008, "step": 34672 }, { "epoch": 6.36, "learning_rate": 2.1280507276133265e-06, "loss": 0.0014, "step": 34673 }, { "epoch": 6.36, "learning_rate": 2.1275928895480126e-06, "loss": 0.0029, "step": 34674 }, { "epoch": 6.36, "learning_rate": 2.127135094875594e-06, "loss": 0.0026, "step": 34675 }, { "epoch": 6.36, "learning_rate": 2.1266773435985867e-06, "loss": 0.0014, "step": 34676 }, { "epoch": 6.36, "learning_rate": 2.1262196357195196e-06, "loss": 0.0013, "step": 34677 }, { "epoch": 6.36, "learning_rate": 2.125761971240916e-06, "loss": 0.0027, "step": 34678 }, { "epoch": 6.36, "learning_rate": 2.125304350165296e-06, "loss": 0.0015, "step": 34679 }, { "epoch": 6.36, "learning_rate": 2.1248467724951825e-06, "loss": 0.0027, "step": 34680 }, { "epoch": 6.36, "learning_rate": 2.124389238233099e-06, "loss": 0.0024, "step": 34681 }, { "epoch": 6.36, "learning_rate": 2.1239317473815657e-06, "loss": 0.0026, "step": 34682 }, { "epoch": 6.36, "learning_rate": 2.1234742999431057e-06, "loss": 0.002, "step": 34683 }, { "epoch": 6.36, "learning_rate": 2.1230168959202367e-06, "loss": 0.0014, "step": 34684 }, { "epoch": 6.36, "learning_rate": 2.122559535315486e-06, "loss": 0.0023, "step": 34685 }, { "epoch": 6.36, "learning_rate": 2.122102218131371e-06, "loss": 0.0013, "step": 34686 }, { "epoch": 6.36, "learning_rate": 2.121644944370417e-06, "loss": 0.002, "step": 34687 }, { "epoch": 6.36, "learning_rate": 2.1211877140351377e-06, "loss": 0.0009, "step": 34688 }, { "epoch": 6.36, "learning_rate": 2.1207305271280533e-06, "loss": 0.0022, "step": 34689 }, { "epoch": 6.36, "learning_rate": 2.1202733836516897e-06, "loss": 0.001, "step": 34690 }, { "epoch": 6.36, "learning_rate": 2.1198162836085634e-06, "loss": 0.0016, "step": 34691 }, { "epoch": 6.36, "learning_rate": 2.119359227001195e-06, "loss": 0.0014, "step": 34692 }, { "epoch": 6.36, "learning_rate": 2.1189022138321023e-06, "loss": 0.0021, "step": 34693 }, { "epoch": 6.36, "learning_rate": 2.118445244103806e-06, "loss": 0.0025, "step": 34694 }, { "epoch": 6.36, "learning_rate": 2.1179883178188244e-06, "loss": 0.0027, "step": 34695 }, { "epoch": 6.36, "learning_rate": 2.117531434979675e-06, "loss": 0.0021, "step": 34696 }, { "epoch": 6.36, "learning_rate": 2.117074595588877e-06, "loss": 0.0026, "step": 34697 }, { "epoch": 6.36, "learning_rate": 2.116617799648949e-06, "loss": 0.0013, "step": 34698 }, { "epoch": 6.36, "learning_rate": 2.1161610471624084e-06, "loss": 0.0015, "step": 34699 }, { "epoch": 6.36, "learning_rate": 2.115704338131773e-06, "loss": 0.0013, "step": 34700 }, { "epoch": 6.36, "learning_rate": 2.1152476725595595e-06, "loss": 0.002, "step": 34701 }, { "epoch": 6.36, "learning_rate": 2.114791050448283e-06, "loss": 0.0029, "step": 34702 }, { "epoch": 6.36, "learning_rate": 2.1143344718004656e-06, "loss": 0.0013, "step": 34703 }, { "epoch": 6.36, "learning_rate": 2.113877936618625e-06, "loss": 0.0019, "step": 34704 }, { "epoch": 6.36, "learning_rate": 2.1134214449052705e-06, "loss": 0.0017, "step": 34705 }, { "epoch": 6.36, "learning_rate": 2.1129649966629185e-06, "loss": 0.0033, "step": 34706 }, { "epoch": 6.36, "learning_rate": 2.112508591894091e-06, "loss": 0.0051, "step": 34707 }, { "epoch": 6.36, "learning_rate": 2.1120522306013013e-06, "loss": 0.002, "step": 34708 }, { "epoch": 6.37, "learning_rate": 2.1115959127870645e-06, "loss": 0.0026, "step": 34709 }, { "epoch": 6.37, "learning_rate": 2.111139638453896e-06, "loss": 0.0022, "step": 34710 }, { "epoch": 6.37, "learning_rate": 2.11068340760431e-06, "loss": 0.0018, "step": 34711 }, { "epoch": 6.37, "learning_rate": 2.1102272202408227e-06, "loss": 0.0021, "step": 34712 }, { "epoch": 6.37, "learning_rate": 2.1097710763659473e-06, "loss": 0.0026, "step": 34713 }, { "epoch": 6.37, "learning_rate": 2.1093149759821983e-06, "loss": 0.003, "step": 34714 }, { "epoch": 6.37, "learning_rate": 2.108858919092088e-06, "loss": 0.0015, "step": 34715 }, { "epoch": 6.37, "learning_rate": 2.108402905698138e-06, "loss": 0.0032, "step": 34716 }, { "epoch": 6.37, "learning_rate": 2.1079469358028524e-06, "loss": 0.0018, "step": 34717 }, { "epoch": 6.37, "learning_rate": 2.1074910094087485e-06, "loss": 0.0025, "step": 34718 }, { "epoch": 6.37, "learning_rate": 2.1070351265183385e-06, "loss": 0.0053, "step": 34719 }, { "epoch": 6.37, "learning_rate": 2.106579287134134e-06, "loss": 0.0016, "step": 34720 }, { "epoch": 6.37, "learning_rate": 2.1061234912586515e-06, "loss": 0.0022, "step": 34721 }, { "epoch": 6.37, "learning_rate": 2.1056677388944046e-06, "loss": 0.0017, "step": 34722 }, { "epoch": 6.37, "learning_rate": 2.1052120300438983e-06, "loss": 0.0009, "step": 34723 }, { "epoch": 6.37, "learning_rate": 2.104756364709647e-06, "loss": 0.0025, "step": 34724 }, { "epoch": 6.37, "learning_rate": 2.104300742894165e-06, "loss": 0.0018, "step": 34725 }, { "epoch": 6.37, "learning_rate": 2.1038451645999623e-06, "loss": 0.0035, "step": 34726 }, { "epoch": 6.37, "learning_rate": 2.103389629829551e-06, "loss": 0.0039, "step": 34727 }, { "epoch": 6.37, "learning_rate": 2.1029341385854406e-06, "loss": 0.0014, "step": 34728 }, { "epoch": 6.37, "learning_rate": 2.1024786908701423e-06, "loss": 0.0021, "step": 34729 }, { "epoch": 6.37, "learning_rate": 2.102023286686167e-06, "loss": 0.0016, "step": 34730 }, { "epoch": 6.37, "learning_rate": 2.1015679260360233e-06, "loss": 0.0015, "step": 34731 }, { "epoch": 6.37, "learning_rate": 2.101112608922222e-06, "loss": 0.0018, "step": 34732 }, { "epoch": 6.37, "learning_rate": 2.1006573353472716e-06, "loss": 0.0011, "step": 34733 }, { "epoch": 6.37, "learning_rate": 2.1002021053136878e-06, "loss": 0.0024, "step": 34734 }, { "epoch": 6.37, "learning_rate": 2.0997469188239726e-06, "loss": 0.0023, "step": 34735 }, { "epoch": 6.37, "learning_rate": 2.0992917758806376e-06, "loss": 0.003, "step": 34736 }, { "epoch": 6.37, "learning_rate": 2.0988366764861878e-06, "loss": 0.0017, "step": 34737 }, { "epoch": 6.37, "learning_rate": 2.0983816206431385e-06, "loss": 0.0018, "step": 34738 }, { "epoch": 6.37, "learning_rate": 2.097926608353995e-06, "loss": 0.0019, "step": 34739 }, { "epoch": 6.37, "learning_rate": 2.097471639621268e-06, "loss": 0.002, "step": 34740 }, { "epoch": 6.37, "learning_rate": 2.0970167144474594e-06, "loss": 0.0015, "step": 34741 }, { "epoch": 6.37, "learning_rate": 2.096561832835078e-06, "loss": 0.0014, "step": 34742 }, { "epoch": 6.37, "learning_rate": 2.0961069947866343e-06, "loss": 0.0018, "step": 34743 }, { "epoch": 6.37, "learning_rate": 2.095652200304634e-06, "loss": 0.0017, "step": 34744 }, { "epoch": 6.37, "learning_rate": 2.095197449391585e-06, "loss": 0.0013, "step": 34745 }, { "epoch": 6.37, "learning_rate": 2.0947427420499934e-06, "loss": 0.0018, "step": 34746 }, { "epoch": 6.37, "learning_rate": 2.0942880782823637e-06, "loss": 0.0027, "step": 34747 }, { "epoch": 6.37, "learning_rate": 2.0938334580912032e-06, "loss": 0.0031, "step": 34748 }, { "epoch": 6.37, "learning_rate": 2.093378881479019e-06, "loss": 0.002, "step": 34749 }, { "epoch": 6.37, "learning_rate": 2.0929243484483143e-06, "loss": 0.0017, "step": 34750 }, { "epoch": 6.37, "learning_rate": 2.092469859001597e-06, "loss": 0.0014, "step": 34751 }, { "epoch": 6.37, "learning_rate": 2.0920154131413706e-06, "loss": 0.0008, "step": 34752 }, { "epoch": 6.37, "learning_rate": 2.0915610108701403e-06, "loss": 0.0021, "step": 34753 }, { "epoch": 6.37, "learning_rate": 2.091106652190411e-06, "loss": 0.0018, "step": 34754 }, { "epoch": 6.37, "learning_rate": 2.0906523371046848e-06, "loss": 0.0015, "step": 34755 }, { "epoch": 6.37, "learning_rate": 2.0901980656154706e-06, "loss": 0.0023, "step": 34756 }, { "epoch": 6.37, "learning_rate": 2.089743837725271e-06, "loss": 0.0016, "step": 34757 }, { "epoch": 6.37, "learning_rate": 2.08928965343659e-06, "loss": 0.0015, "step": 34758 }, { "epoch": 6.37, "learning_rate": 2.0888355127519254e-06, "loss": 0.002, "step": 34759 }, { "epoch": 6.37, "learning_rate": 2.088381415673787e-06, "loss": 0.0019, "step": 34760 }, { "epoch": 6.37, "learning_rate": 2.087927362204675e-06, "loss": 0.003, "step": 34761 }, { "epoch": 6.37, "learning_rate": 2.0874733523470936e-06, "loss": 0.003, "step": 34762 }, { "epoch": 6.38, "learning_rate": 2.0870193861035437e-06, "loss": 0.0019, "step": 34763 }, { "epoch": 6.38, "learning_rate": 2.0865654634765287e-06, "loss": 0.0022, "step": 34764 }, { "epoch": 6.38, "learning_rate": 2.086111584468551e-06, "loss": 0.0029, "step": 34765 }, { "epoch": 6.38, "learning_rate": 2.0856577490821107e-06, "loss": 0.0017, "step": 34766 }, { "epoch": 6.38, "learning_rate": 2.0852039573197103e-06, "loss": 0.0029, "step": 34767 }, { "epoch": 6.38, "learning_rate": 2.084750209183849e-06, "loss": 0.0021, "step": 34768 }, { "epoch": 6.38, "learning_rate": 2.084296504677035e-06, "loss": 0.0013, "step": 34769 }, { "epoch": 6.38, "learning_rate": 2.0838428438017624e-06, "loss": 0.0018, "step": 34770 }, { "epoch": 6.38, "learning_rate": 2.0833892265605316e-06, "loss": 0.0018, "step": 34771 }, { "epoch": 6.38, "learning_rate": 2.0829356529558463e-06, "loss": 0.0033, "step": 34772 }, { "epoch": 6.38, "learning_rate": 2.0824821229902015e-06, "loss": 0.0033, "step": 34773 }, { "epoch": 6.38, "learning_rate": 2.082028636666105e-06, "loss": 0.0025, "step": 34774 }, { "epoch": 6.38, "learning_rate": 2.0815751939860508e-06, "loss": 0.0018, "step": 34775 }, { "epoch": 6.38, "learning_rate": 2.081121794952542e-06, "loss": 0.0025, "step": 34776 }, { "epoch": 6.38, "learning_rate": 2.080668439568071e-06, "loss": 0.0021, "step": 34777 }, { "epoch": 6.38, "learning_rate": 2.0802151278351435e-06, "loss": 0.0022, "step": 34778 }, { "epoch": 6.38, "learning_rate": 2.0797618597562554e-06, "loss": 0.0023, "step": 34779 }, { "epoch": 6.38, "learning_rate": 2.0793086353339053e-06, "loss": 0.0013, "step": 34780 }, { "epoch": 6.38, "learning_rate": 2.078855454570592e-06, "loss": 0.0021, "step": 34781 }, { "epoch": 6.38, "learning_rate": 2.078402317468813e-06, "loss": 0.0024, "step": 34782 }, { "epoch": 6.38, "learning_rate": 2.077949224031066e-06, "loss": 0.0031, "step": 34783 }, { "epoch": 6.38, "learning_rate": 2.077496174259849e-06, "loss": 0.0023, "step": 34784 }, { "epoch": 6.38, "learning_rate": 2.0770431681576575e-06, "loss": 0.0018, "step": 34785 }, { "epoch": 6.38, "learning_rate": 2.076590205726988e-06, "loss": 0.0026, "step": 34786 }, { "epoch": 6.38, "learning_rate": 2.0761372869703443e-06, "loss": 0.0018, "step": 34787 }, { "epoch": 6.38, "learning_rate": 2.075684411890214e-06, "loss": 0.0026, "step": 34788 }, { "epoch": 6.38, "learning_rate": 2.075231580489098e-06, "loss": 0.0015, "step": 34789 }, { "epoch": 6.38, "learning_rate": 2.0747787927694873e-06, "loss": 0.0026, "step": 34790 }, { "epoch": 6.38, "learning_rate": 2.0743260487338844e-06, "loss": 0.001, "step": 34791 }, { "epoch": 6.38, "learning_rate": 2.073873348384783e-06, "loss": 0.0041, "step": 34792 }, { "epoch": 6.38, "learning_rate": 2.073420691724677e-06, "loss": 0.0022, "step": 34793 }, { "epoch": 6.38, "learning_rate": 2.0729680787560645e-06, "loss": 0.0021, "step": 34794 }, { "epoch": 6.38, "learning_rate": 2.072515509481432e-06, "loss": 0.0032, "step": 34795 }, { "epoch": 6.38, "learning_rate": 2.0720629839032824e-06, "loss": 0.0023, "step": 34796 }, { "epoch": 6.38, "learning_rate": 2.0716105020241074e-06, "loss": 0.0028, "step": 34797 }, { "epoch": 6.38, "learning_rate": 2.0711580638464013e-06, "loss": 0.0022, "step": 34798 }, { "epoch": 6.38, "learning_rate": 2.070705669372657e-06, "loss": 0.0021, "step": 34799 }, { "epoch": 6.38, "learning_rate": 2.0702533186053695e-06, "loss": 0.0015, "step": 34800 }, { "epoch": 6.38, "learning_rate": 2.0698010115470313e-06, "loss": 0.0019, "step": 34801 }, { "epoch": 6.38, "learning_rate": 2.069348748200136e-06, "loss": 0.0016, "step": 34802 }, { "epoch": 6.38, "learning_rate": 2.0688965285671757e-06, "loss": 0.002, "step": 34803 }, { "epoch": 6.38, "learning_rate": 2.0684443526506415e-06, "loss": 0.0018, "step": 34804 }, { "epoch": 6.38, "learning_rate": 2.0679922204530323e-06, "loss": 0.0015, "step": 34805 }, { "epoch": 6.38, "learning_rate": 2.067540131976834e-06, "loss": 0.0033, "step": 34806 }, { "epoch": 6.38, "learning_rate": 2.06708808722454e-06, "loss": 0.0015, "step": 34807 }, { "epoch": 6.38, "learning_rate": 2.06663608619864e-06, "loss": 0.0022, "step": 34808 }, { "epoch": 6.38, "learning_rate": 2.066184128901629e-06, "loss": 0.0021, "step": 34809 }, { "epoch": 6.38, "learning_rate": 2.065732215335998e-06, "loss": 0.0017, "step": 34810 }, { "epoch": 6.38, "learning_rate": 2.06528034550424e-06, "loss": 0.0022, "step": 34811 }, { "epoch": 6.38, "learning_rate": 2.064828519408837e-06, "loss": 0.0027, "step": 34812 }, { "epoch": 6.38, "learning_rate": 2.0643767370522872e-06, "loss": 0.0017, "step": 34813 }, { "epoch": 6.38, "learning_rate": 2.06392499843708e-06, "loss": 0.0029, "step": 34814 }, { "epoch": 6.38, "learning_rate": 2.063473303565703e-06, "loss": 0.0025, "step": 34815 }, { "epoch": 6.38, "learning_rate": 2.0630216524406475e-06, "loss": 0.0018, "step": 34816 }, { "epoch": 6.38, "learning_rate": 2.0625700450644025e-06, "loss": 0.002, "step": 34817 }, { "epoch": 6.39, "learning_rate": 2.062118481439457e-06, "loss": 0.0017, "step": 34818 }, { "epoch": 6.39, "learning_rate": 2.0616669615683007e-06, "loss": 0.0022, "step": 34819 }, { "epoch": 6.39, "learning_rate": 2.0612154854534227e-06, "loss": 0.0035, "step": 34820 }, { "epoch": 6.39, "learning_rate": 2.0607640530973073e-06, "loss": 0.0042, "step": 34821 }, { "epoch": 6.39, "learning_rate": 2.0603126645024495e-06, "loss": 0.0021, "step": 34822 }, { "epoch": 6.39, "learning_rate": 2.0598613196713367e-06, "loss": 0.0011, "step": 34823 }, { "epoch": 6.39, "learning_rate": 2.0594100186064526e-06, "loss": 0.0015, "step": 34824 }, { "epoch": 6.39, "learning_rate": 2.058958761310286e-06, "loss": 0.0012, "step": 34825 }, { "epoch": 6.39, "learning_rate": 2.0585075477853213e-06, "loss": 0.002, "step": 34826 }, { "epoch": 6.39, "learning_rate": 2.058056378034052e-06, "loss": 0.0014, "step": 34827 }, { "epoch": 6.39, "learning_rate": 2.057605252058963e-06, "loss": 0.0018, "step": 34828 }, { "epoch": 6.39, "learning_rate": 2.0571541698625418e-06, "loss": 0.0024, "step": 34829 }, { "epoch": 6.39, "learning_rate": 2.0567031314472675e-06, "loss": 0.0024, "step": 34830 }, { "epoch": 6.39, "learning_rate": 2.056252136815634e-06, "loss": 0.0021, "step": 34831 }, { "epoch": 6.39, "learning_rate": 2.0558011859701253e-06, "loss": 0.003, "step": 34832 }, { "epoch": 6.39, "learning_rate": 2.055350278913226e-06, "loss": 0.0025, "step": 34833 }, { "epoch": 6.39, "learning_rate": 2.0548994156474234e-06, "loss": 0.0029, "step": 34834 }, { "epoch": 6.39, "learning_rate": 2.0544485961752e-06, "loss": 0.0012, "step": 34835 }, { "epoch": 6.39, "learning_rate": 2.053997820499043e-06, "loss": 0.0027, "step": 34836 }, { "epoch": 6.39, "learning_rate": 2.053547088621436e-06, "loss": 0.0015, "step": 34837 }, { "epoch": 6.39, "learning_rate": 2.053096400544864e-06, "loss": 0.0014, "step": 34838 }, { "epoch": 6.39, "learning_rate": 2.0526457562718074e-06, "loss": 0.0015, "step": 34839 }, { "epoch": 6.39, "learning_rate": 2.0521951558047558e-06, "loss": 0.0023, "step": 34840 }, { "epoch": 6.39, "learning_rate": 2.051744599146195e-06, "loss": 0.0009, "step": 34841 }, { "epoch": 6.39, "learning_rate": 2.0512940862985995e-06, "loss": 0.0024, "step": 34842 }, { "epoch": 6.39, "learning_rate": 2.0508436172644564e-06, "loss": 0.0016, "step": 34843 }, { "epoch": 6.39, "learning_rate": 2.0503931920462516e-06, "loss": 0.0021, "step": 34844 }, { "epoch": 6.39, "learning_rate": 2.0499428106464646e-06, "loss": 0.0023, "step": 34845 }, { "epoch": 6.39, "learning_rate": 2.0494924730675802e-06, "loss": 0.0019, "step": 34846 }, { "epoch": 6.39, "learning_rate": 2.0490421793120817e-06, "loss": 0.0024, "step": 34847 }, { "epoch": 6.39, "learning_rate": 2.0485919293824443e-06, "loss": 0.0012, "step": 34848 }, { "epoch": 6.39, "learning_rate": 2.0481417232811575e-06, "loss": 0.0019, "step": 34849 }, { "epoch": 6.39, "learning_rate": 2.0476915610106985e-06, "loss": 0.0015, "step": 34850 }, { "epoch": 6.39, "learning_rate": 2.047241442573551e-06, "loss": 0.0015, "step": 34851 }, { "epoch": 6.39, "learning_rate": 2.046791367972194e-06, "loss": 0.0029, "step": 34852 }, { "epoch": 6.39, "learning_rate": 2.04634133720911e-06, "loss": 0.0019, "step": 34853 }, { "epoch": 6.39, "learning_rate": 2.045891350286778e-06, "loss": 0.0032, "step": 34854 }, { "epoch": 6.39, "learning_rate": 2.045441407207681e-06, "loss": 0.0019, "step": 34855 }, { "epoch": 6.39, "learning_rate": 2.0449915079742955e-06, "loss": 0.0041, "step": 34856 }, { "epoch": 6.39, "learning_rate": 2.0445416525891016e-06, "loss": 0.0032, "step": 34857 }, { "epoch": 6.39, "learning_rate": 2.0440918410545853e-06, "loss": 0.0012, "step": 34858 }, { "epoch": 6.39, "learning_rate": 2.0436420733732187e-06, "loss": 0.0023, "step": 34859 }, { "epoch": 6.39, "learning_rate": 2.043192349547484e-06, "loss": 0.0024, "step": 34860 }, { "epoch": 6.39, "learning_rate": 2.042742669579856e-06, "loss": 0.0018, "step": 34861 }, { "epoch": 6.39, "learning_rate": 2.0422930334728197e-06, "loss": 0.0028, "step": 34862 }, { "epoch": 6.39, "learning_rate": 2.04184344122885e-06, "loss": 0.0012, "step": 34863 }, { "epoch": 6.39, "learning_rate": 2.041393892850426e-06, "loss": 0.0026, "step": 34864 }, { "epoch": 6.39, "learning_rate": 2.040944388340026e-06, "loss": 0.0018, "step": 34865 }, { "epoch": 6.39, "learning_rate": 2.040494927700126e-06, "loss": 0.002, "step": 34866 }, { "epoch": 6.39, "learning_rate": 2.040045510933205e-06, "loss": 0.0018, "step": 34867 }, { "epoch": 6.39, "learning_rate": 2.0395961380417406e-06, "loss": 0.0022, "step": 34868 }, { "epoch": 6.39, "learning_rate": 2.0391468090282074e-06, "loss": 0.0023, "step": 34869 }, { "epoch": 6.39, "learning_rate": 2.0386975238950835e-06, "loss": 0.0027, "step": 34870 }, { "epoch": 6.39, "learning_rate": 2.0382482826448456e-06, "loss": 0.0028, "step": 34871 }, { "epoch": 6.4, "learning_rate": 2.03779908527997e-06, "loss": 0.0011, "step": 34872 }, { "epoch": 6.4, "learning_rate": 2.0373499318029334e-06, "loss": 0.0018, "step": 34873 }, { "epoch": 6.4, "learning_rate": 2.0369008222162057e-06, "loss": 0.0013, "step": 34874 }, { "epoch": 6.4, "learning_rate": 2.0364517565222706e-06, "loss": 0.0016, "step": 34875 }, { "epoch": 6.4, "learning_rate": 2.0360027347236033e-06, "loss": 0.0014, "step": 34876 }, { "epoch": 6.4, "learning_rate": 2.0355537568226723e-06, "loss": 0.0022, "step": 34877 }, { "epoch": 6.4, "learning_rate": 2.0351048228219562e-06, "loss": 0.0025, "step": 34878 }, { "epoch": 6.4, "learning_rate": 2.0346559327239256e-06, "loss": 0.0011, "step": 34879 }, { "epoch": 6.4, "learning_rate": 2.0342070865310613e-06, "loss": 0.0019, "step": 34880 }, { "epoch": 6.4, "learning_rate": 2.0337582842458337e-06, "loss": 0.002, "step": 34881 }, { "epoch": 6.4, "learning_rate": 2.033309525870717e-06, "loss": 0.0034, "step": 34882 }, { "epoch": 6.4, "learning_rate": 2.032860811408185e-06, "loss": 0.0016, "step": 34883 }, { "epoch": 6.4, "learning_rate": 2.032412140860711e-06, "loss": 0.0019, "step": 34884 }, { "epoch": 6.4, "learning_rate": 2.031963514230768e-06, "loss": 0.0024, "step": 34885 }, { "epoch": 6.4, "learning_rate": 2.0315149315208283e-06, "loss": 0.0015, "step": 34886 }, { "epoch": 6.4, "learning_rate": 2.031066392733363e-06, "loss": 0.0012, "step": 34887 }, { "epoch": 6.4, "learning_rate": 2.030617897870851e-06, "loss": 0.0031, "step": 34888 }, { "epoch": 6.4, "learning_rate": 2.030169446935758e-06, "loss": 0.0034, "step": 34889 }, { "epoch": 6.4, "learning_rate": 2.0297210399305568e-06, "loss": 0.0016, "step": 34890 }, { "epoch": 6.4, "learning_rate": 2.029272676857721e-06, "loss": 0.0026, "step": 34891 }, { "epoch": 6.4, "learning_rate": 2.0288243577197177e-06, "loss": 0.0022, "step": 34892 }, { "epoch": 6.4, "learning_rate": 2.0283760825190234e-06, "loss": 0.0027, "step": 34893 }, { "epoch": 6.4, "learning_rate": 2.02792785125811e-06, "loss": 0.0009, "step": 34894 }, { "epoch": 6.4, "learning_rate": 2.0274796639394435e-06, "loss": 0.0024, "step": 34895 }, { "epoch": 6.4, "learning_rate": 2.0270315205654913e-06, "loss": 0.0023, "step": 34896 }, { "epoch": 6.4, "learning_rate": 2.0265834211387327e-06, "loss": 0.0017, "step": 34897 }, { "epoch": 6.4, "learning_rate": 2.0261353656616312e-06, "loss": 0.0026, "step": 34898 }, { "epoch": 6.4, "learning_rate": 2.0256873541366595e-06, "loss": 0.0021, "step": 34899 }, { "epoch": 6.4, "learning_rate": 2.0252393865662856e-06, "loss": 0.0026, "step": 34900 }, { "epoch": 6.4, "learning_rate": 2.0247914629529796e-06, "loss": 0.0025, "step": 34901 }, { "epoch": 6.4, "learning_rate": 2.0243435832992085e-06, "loss": 0.0025, "step": 34902 }, { "epoch": 6.4, "learning_rate": 2.023895747607444e-06, "loss": 0.0013, "step": 34903 }, { "epoch": 6.4, "learning_rate": 2.0234479558801523e-06, "loss": 0.0034, "step": 34904 }, { "epoch": 6.4, "learning_rate": 2.023000208119802e-06, "loss": 0.0019, "step": 34905 }, { "epoch": 6.4, "learning_rate": 2.022552504328862e-06, "loss": 0.0024, "step": 34906 }, { "epoch": 6.4, "learning_rate": 2.022104844509799e-06, "loss": 0.0026, "step": 34907 }, { "epoch": 6.4, "learning_rate": 2.0216572286650817e-06, "loss": 0.0029, "step": 34908 }, { "epoch": 6.4, "learning_rate": 2.021209656797174e-06, "loss": 0.0023, "step": 34909 }, { "epoch": 6.4, "learning_rate": 2.020762128908548e-06, "loss": 0.0022, "step": 34910 }, { "epoch": 6.4, "learning_rate": 2.0203146450016674e-06, "loss": 0.0029, "step": 34911 }, { "epoch": 6.4, "learning_rate": 2.0198672050790026e-06, "loss": 0.0009, "step": 34912 }, { "epoch": 6.4, "learning_rate": 2.0194198091430137e-06, "loss": 0.0012, "step": 34913 }, { "epoch": 6.4, "learning_rate": 2.0189724571961677e-06, "loss": 0.0024, "step": 34914 }, { "epoch": 6.4, "learning_rate": 2.0185251492409354e-06, "loss": 0.0024, "step": 34915 }, { "epoch": 6.4, "learning_rate": 2.0180778852797787e-06, "loss": 0.0027, "step": 34916 }, { "epoch": 6.4, "learning_rate": 2.0176306653151646e-06, "loss": 0.0028, "step": 34917 }, { "epoch": 6.4, "learning_rate": 2.0171834893495557e-06, "loss": 0.0019, "step": 34918 }, { "epoch": 6.4, "learning_rate": 2.0167363573854205e-06, "loss": 0.0015, "step": 34919 }, { "epoch": 6.4, "learning_rate": 2.01628926942522e-06, "loss": 0.002, "step": 34920 }, { "epoch": 6.4, "learning_rate": 2.0158422254714204e-06, "loss": 0.002, "step": 34921 }, { "epoch": 6.4, "learning_rate": 2.0153952255264853e-06, "loss": 0.0018, "step": 34922 }, { "epoch": 6.4, "learning_rate": 2.014948269592879e-06, "loss": 0.0021, "step": 34923 }, { "epoch": 6.4, "learning_rate": 2.0145013576730656e-06, "loss": 0.0027, "step": 34924 }, { "epoch": 6.4, "learning_rate": 2.014054489769507e-06, "loss": 0.0022, "step": 34925 }, { "epoch": 6.4, "learning_rate": 2.013607665884667e-06, "loss": 0.0017, "step": 34926 }, { "epoch": 6.41, "learning_rate": 2.0131608860210063e-06, "loss": 0.0012, "step": 34927 }, { "epoch": 6.41, "learning_rate": 2.0127141501809933e-06, "loss": 0.001, "step": 34928 }, { "epoch": 6.41, "learning_rate": 2.0122674583670864e-06, "loss": 0.001, "step": 34929 }, { "epoch": 6.41, "learning_rate": 2.0118208105817505e-06, "loss": 0.0016, "step": 34930 }, { "epoch": 6.41, "learning_rate": 2.0113742068274434e-06, "loss": 0.0019, "step": 34931 }, { "epoch": 6.41, "learning_rate": 2.0109276471066263e-06, "loss": 0.0013, "step": 34932 }, { "epoch": 6.41, "learning_rate": 2.0104811314217656e-06, "loss": 0.003, "step": 34933 }, { "epoch": 6.41, "learning_rate": 2.0100346597753196e-06, "loss": 0.0016, "step": 34934 }, { "epoch": 6.41, "learning_rate": 2.0095882321697504e-06, "loss": 0.0016, "step": 34935 }, { "epoch": 6.41, "learning_rate": 2.0091418486075176e-06, "loss": 0.0023, "step": 34936 }, { "epoch": 6.41, "learning_rate": 2.008695509091082e-06, "loss": 0.0034, "step": 34937 }, { "epoch": 6.41, "learning_rate": 2.008249213622905e-06, "loss": 0.0028, "step": 34938 }, { "epoch": 6.41, "learning_rate": 2.0078029622054453e-06, "loss": 0.0034, "step": 34939 }, { "epoch": 6.41, "learning_rate": 2.007356754841159e-06, "loss": 0.0023, "step": 34940 }, { "epoch": 6.41, "learning_rate": 2.006910591532516e-06, "loss": 0.0017, "step": 34941 }, { "epoch": 6.41, "learning_rate": 2.006464472281966e-06, "loss": 0.0024, "step": 34942 }, { "epoch": 6.41, "learning_rate": 2.006018397091971e-06, "loss": 0.0015, "step": 34943 }, { "epoch": 6.41, "learning_rate": 2.0055723659649907e-06, "loss": 0.0013, "step": 34944 }, { "epoch": 6.41, "learning_rate": 2.005126378903479e-06, "loss": 0.0032, "step": 34945 }, { "epoch": 6.41, "learning_rate": 2.0046804359099013e-06, "loss": 0.0024, "step": 34946 }, { "epoch": 6.41, "learning_rate": 2.004234536986712e-06, "loss": 0.0026, "step": 34947 }, { "epoch": 6.41, "learning_rate": 2.0037886821363727e-06, "loss": 0.002, "step": 34948 }, { "epoch": 6.41, "learning_rate": 2.0033428713613313e-06, "loss": 0.0012, "step": 34949 }, { "epoch": 6.41, "learning_rate": 2.0028971046640543e-06, "loss": 0.0023, "step": 34950 }, { "epoch": 6.41, "learning_rate": 2.002451382046996e-06, "loss": 0.0027, "step": 34951 }, { "epoch": 6.41, "learning_rate": 2.002005703512613e-06, "loss": 0.0016, "step": 34952 }, { "epoch": 6.41, "learning_rate": 2.001560069063362e-06, "loss": 0.0017, "step": 34953 }, { "epoch": 6.41, "learning_rate": 2.001114478701699e-06, "loss": 0.0014, "step": 34954 }, { "epoch": 6.41, "learning_rate": 2.000668932430081e-06, "loss": 0.0019, "step": 34955 }, { "epoch": 6.41, "learning_rate": 2.0002234302509627e-06, "loss": 0.0023, "step": 34956 }, { "epoch": 6.41, "learning_rate": 1.9997779721668e-06, "loss": 0.0023, "step": 34957 }, { "epoch": 6.41, "learning_rate": 1.999332558180046e-06, "loss": 0.0031, "step": 34958 }, { "epoch": 6.41, "learning_rate": 1.9988871882931638e-06, "loss": 0.0031, "step": 34959 }, { "epoch": 6.41, "learning_rate": 1.9984418625086e-06, "loss": 0.0009, "step": 34960 }, { "epoch": 6.41, "learning_rate": 1.9979965808288125e-06, "loss": 0.0019, "step": 34961 }, { "epoch": 6.41, "learning_rate": 1.9975513432562532e-06, "loss": 0.0027, "step": 34962 }, { "epoch": 6.41, "learning_rate": 1.99710614979338e-06, "loss": 0.002, "step": 34963 }, { "epoch": 6.41, "learning_rate": 1.9966610004426457e-06, "loss": 0.0023, "step": 34964 }, { "epoch": 6.41, "learning_rate": 1.996215895206506e-06, "loss": 0.002, "step": 34965 }, { "epoch": 6.41, "learning_rate": 1.9957708340874095e-06, "loss": 0.0012, "step": 34966 }, { "epoch": 6.41, "learning_rate": 1.9953258170878087e-06, "loss": 0.0019, "step": 34967 }, { "epoch": 6.41, "learning_rate": 1.994880844210162e-06, "loss": 0.0019, "step": 34968 }, { "epoch": 6.41, "learning_rate": 1.994435915456919e-06, "loss": 0.0021, "step": 34969 }, { "epoch": 6.41, "learning_rate": 1.993991030830534e-06, "loss": 0.0033, "step": 34970 }, { "epoch": 6.41, "learning_rate": 1.9935461903334575e-06, "loss": 0.0019, "step": 34971 }, { "epoch": 6.41, "learning_rate": 1.993101393968142e-06, "loss": 0.002, "step": 34972 }, { "epoch": 6.41, "learning_rate": 1.9926566417370395e-06, "loss": 0.0029, "step": 34973 }, { "epoch": 6.41, "learning_rate": 1.9922119336426004e-06, "loss": 0.0014, "step": 34974 }, { "epoch": 6.41, "learning_rate": 1.991767269687278e-06, "loss": 0.0011, "step": 34975 }, { "epoch": 6.41, "learning_rate": 1.991322649873518e-06, "loss": 0.0022, "step": 34976 }, { "epoch": 6.41, "learning_rate": 1.9908780742037815e-06, "loss": 0.0013, "step": 34977 }, { "epoch": 6.41, "learning_rate": 1.99043354268051e-06, "loss": 0.0017, "step": 34978 }, { "epoch": 6.41, "learning_rate": 1.9899890553061565e-06, "loss": 0.002, "step": 34979 }, { "epoch": 6.41, "learning_rate": 1.9895446120831674e-06, "loss": 0.0025, "step": 34980 }, { "epoch": 6.42, "learning_rate": 1.9891002130139993e-06, "loss": 0.0018, "step": 34981 }, { "epoch": 6.42, "learning_rate": 1.9886558581010983e-06, "loss": 0.0011, "step": 34982 }, { "epoch": 6.42, "learning_rate": 1.9882115473469167e-06, "loss": 0.0023, "step": 34983 }, { "epoch": 6.42, "learning_rate": 1.9877672807538984e-06, "loss": 0.0013, "step": 34984 }, { "epoch": 6.42, "learning_rate": 1.9873230583244907e-06, "loss": 0.0013, "step": 34985 }, { "epoch": 6.42, "learning_rate": 1.9868788800611503e-06, "loss": 0.0019, "step": 34986 }, { "epoch": 6.42, "learning_rate": 1.98643474596632e-06, "loss": 0.0021, "step": 34987 }, { "epoch": 6.42, "learning_rate": 1.98599065604245e-06, "loss": 0.0017, "step": 34988 }, { "epoch": 6.42, "learning_rate": 1.9855466102919864e-06, "loss": 0.0027, "step": 34989 }, { "epoch": 6.42, "learning_rate": 1.9851026087173785e-06, "loss": 0.0016, "step": 34990 }, { "epoch": 6.42, "learning_rate": 1.984658651321072e-06, "loss": 0.0023, "step": 34991 }, { "epoch": 6.42, "learning_rate": 1.9842147381055156e-06, "loss": 0.0021, "step": 34992 }, { "epoch": 6.42, "learning_rate": 1.9837708690731515e-06, "loss": 0.002, "step": 34993 }, { "epoch": 6.42, "learning_rate": 1.983327044226434e-06, "loss": 0.0019, "step": 34994 }, { "epoch": 6.42, "learning_rate": 1.9828832635678086e-06, "loss": 0.0039, "step": 34995 }, { "epoch": 6.42, "learning_rate": 1.982439527099714e-06, "loss": 0.0031, "step": 34996 }, { "epoch": 6.42, "learning_rate": 1.981995834824603e-06, "loss": 0.0026, "step": 34997 }, { "epoch": 6.42, "learning_rate": 1.981552186744914e-06, "loss": 0.0021, "step": 34998 }, { "epoch": 6.42, "learning_rate": 1.9811085828631003e-06, "loss": 0.0015, "step": 34999 }, { "epoch": 6.42, "learning_rate": 1.980665023181605e-06, "loss": 0.0019, "step": 35000 }, { "epoch": 6.42, "learning_rate": 1.980221507702874e-06, "loss": 0.0014, "step": 35001 }, { "epoch": 6.42, "learning_rate": 1.979778036429344e-06, "loss": 0.0016, "step": 35002 }, { "epoch": 6.42, "learning_rate": 1.979334609363469e-06, "loss": 0.0012, "step": 35003 }, { "epoch": 6.42, "learning_rate": 1.9788912265076886e-06, "loss": 0.0016, "step": 35004 }, { "epoch": 6.42, "learning_rate": 1.9784478878644485e-06, "loss": 0.0012, "step": 35005 }, { "epoch": 6.42, "learning_rate": 1.9780045934361915e-06, "loss": 0.0022, "step": 35006 }, { "epoch": 6.42, "learning_rate": 1.9775613432253616e-06, "loss": 0.003, "step": 35007 }, { "epoch": 6.42, "learning_rate": 1.9771181372344004e-06, "loss": 0.0023, "step": 35008 }, { "epoch": 6.42, "learning_rate": 1.976674975465752e-06, "loss": 0.0011, "step": 35009 }, { "epoch": 6.42, "learning_rate": 1.97623185792186e-06, "loss": 0.0024, "step": 35010 }, { "epoch": 6.42, "learning_rate": 1.9757887846051624e-06, "loss": 0.0015, "step": 35011 }, { "epoch": 6.42, "learning_rate": 1.975345755518111e-06, "loss": 0.0016, "step": 35012 }, { "epoch": 6.42, "learning_rate": 1.9749027706631383e-06, "loss": 0.0029, "step": 35013 }, { "epoch": 6.42, "learning_rate": 1.974459830042691e-06, "loss": 0.0021, "step": 35014 }, { "epoch": 6.42, "learning_rate": 1.974016933659205e-06, "loss": 0.0026, "step": 35015 }, { "epoch": 6.42, "learning_rate": 1.9735740815151294e-06, "loss": 0.0021, "step": 35016 }, { "epoch": 6.42, "learning_rate": 1.9731312736129004e-06, "loss": 0.0015, "step": 35017 }, { "epoch": 6.42, "learning_rate": 1.972688509954961e-06, "loss": 0.0018, "step": 35018 }, { "epoch": 6.42, "learning_rate": 1.972245790543753e-06, "loss": 0.0016, "step": 35019 }, { "epoch": 6.42, "learning_rate": 1.9718031153817085e-06, "loss": 0.002, "step": 35020 }, { "epoch": 6.42, "learning_rate": 1.9713604844712765e-06, "loss": 0.0027, "step": 35021 }, { "epoch": 6.42, "learning_rate": 1.9709178978148924e-06, "loss": 0.0015, "step": 35022 }, { "epoch": 6.42, "learning_rate": 1.970475355414998e-06, "loss": 0.0017, "step": 35023 }, { "epoch": 6.42, "learning_rate": 1.970032857274031e-06, "loss": 0.002, "step": 35024 }, { "epoch": 6.42, "learning_rate": 1.9695904033944325e-06, "loss": 0.002, "step": 35025 }, { "epoch": 6.42, "learning_rate": 1.9691479937786383e-06, "loss": 0.002, "step": 35026 }, { "epoch": 6.42, "learning_rate": 1.9687056284290882e-06, "loss": 0.0033, "step": 35027 }, { "epoch": 6.42, "learning_rate": 1.9682633073482227e-06, "loss": 0.0015, "step": 35028 }, { "epoch": 6.42, "learning_rate": 1.9678210305384735e-06, "loss": 0.0016, "step": 35029 }, { "epoch": 6.42, "learning_rate": 1.9673787980022895e-06, "loss": 0.0022, "step": 35030 }, { "epoch": 6.42, "learning_rate": 1.966936609742098e-06, "loss": 0.0017, "step": 35031 }, { "epoch": 6.42, "learning_rate": 1.966494465760341e-06, "loss": 0.0022, "step": 35032 }, { "epoch": 6.42, "learning_rate": 1.9660523660594523e-06, "loss": 0.0015, "step": 35033 }, { "epoch": 6.42, "learning_rate": 1.965610310641872e-06, "loss": 0.0018, "step": 35034 }, { "epoch": 6.42, "learning_rate": 1.9651682995100372e-06, "loss": 0.0024, "step": 35035 }, { "epoch": 6.43, "learning_rate": 1.9647263326663835e-06, "loss": 0.0022, "step": 35036 }, { "epoch": 6.43, "learning_rate": 1.964284410113346e-06, "loss": 0.0013, "step": 35037 }, { "epoch": 6.43, "learning_rate": 1.9638425318533607e-06, "loss": 0.0028, "step": 35038 }, { "epoch": 6.43, "learning_rate": 1.963400697888864e-06, "loss": 0.003, "step": 35039 }, { "epoch": 6.43, "learning_rate": 1.962958908222291e-06, "loss": 0.0024, "step": 35040 }, { "epoch": 6.43, "learning_rate": 1.9625171628560768e-06, "loss": 0.0021, "step": 35041 }, { "epoch": 6.43, "learning_rate": 1.9620754617926565e-06, "loss": 0.0018, "step": 35042 }, { "epoch": 6.43, "learning_rate": 1.961633805034464e-06, "loss": 0.0017, "step": 35043 }, { "epoch": 6.43, "learning_rate": 1.961192192583934e-06, "loss": 0.0024, "step": 35044 }, { "epoch": 6.43, "learning_rate": 1.9607506244435013e-06, "loss": 0.0042, "step": 35045 }, { "epoch": 6.43, "learning_rate": 1.9603091006155974e-06, "loss": 0.0014, "step": 35046 }, { "epoch": 6.43, "learning_rate": 1.9598676211026602e-06, "loss": 0.0017, "step": 35047 }, { "epoch": 6.43, "learning_rate": 1.959426185907124e-06, "loss": 0.0019, "step": 35048 }, { "epoch": 6.43, "learning_rate": 1.9589847950314167e-06, "loss": 0.0013, "step": 35049 }, { "epoch": 6.43, "learning_rate": 1.958543448477973e-06, "loss": 0.0024, "step": 35050 }, { "epoch": 6.43, "learning_rate": 1.9581021462492232e-06, "loss": 0.0023, "step": 35051 }, { "epoch": 6.43, "learning_rate": 1.9576608883476066e-06, "loss": 0.001, "step": 35052 }, { "epoch": 6.43, "learning_rate": 1.9572196747755502e-06, "loss": 0.0015, "step": 35053 }, { "epoch": 6.43, "learning_rate": 1.956778505535488e-06, "loss": 0.0027, "step": 35054 }, { "epoch": 6.43, "learning_rate": 1.9563373806298515e-06, "loss": 0.0015, "step": 35055 }, { "epoch": 6.43, "learning_rate": 1.955896300061072e-06, "loss": 0.0035, "step": 35056 }, { "epoch": 6.43, "learning_rate": 1.9554552638315795e-06, "loss": 0.0027, "step": 35057 }, { "epoch": 6.43, "learning_rate": 1.955014271943807e-06, "loss": 0.0018, "step": 35058 }, { "epoch": 6.43, "learning_rate": 1.954573324400183e-06, "loss": 0.001, "step": 35059 }, { "epoch": 6.43, "learning_rate": 1.9541324212031408e-06, "loss": 0.0015, "step": 35060 }, { "epoch": 6.43, "learning_rate": 1.953691562355109e-06, "loss": 0.0015, "step": 35061 }, { "epoch": 6.43, "learning_rate": 1.953250747858517e-06, "loss": 0.0015, "step": 35062 }, { "epoch": 6.43, "learning_rate": 1.952809977715796e-06, "loss": 0.0014, "step": 35063 }, { "epoch": 6.43, "learning_rate": 1.9523692519293714e-06, "loss": 0.0023, "step": 35064 }, { "epoch": 6.43, "learning_rate": 1.95192857050168e-06, "loss": 0.0016, "step": 35065 }, { "epoch": 6.43, "learning_rate": 1.951487933435149e-06, "loss": 0.0014, "step": 35066 }, { "epoch": 6.43, "learning_rate": 1.9510473407322018e-06, "loss": 0.003, "step": 35067 }, { "epoch": 6.43, "learning_rate": 1.9506067923952676e-06, "loss": 0.0029, "step": 35068 }, { "epoch": 6.43, "learning_rate": 1.9501662884267814e-06, "loss": 0.0032, "step": 35069 }, { "epoch": 6.43, "learning_rate": 1.9497258288291655e-06, "loss": 0.0016, "step": 35070 }, { "epoch": 6.43, "learning_rate": 1.94928541360485e-06, "loss": 0.0017, "step": 35071 }, { "epoch": 6.43, "learning_rate": 1.9488450427562624e-06, "loss": 0.0022, "step": 35072 }, { "epoch": 6.43, "learning_rate": 1.9484047162858287e-06, "loss": 0.0027, "step": 35073 }, { "epoch": 6.43, "learning_rate": 1.947964434195978e-06, "loss": 0.0019, "step": 35074 }, { "epoch": 6.43, "learning_rate": 1.9475241964891358e-06, "loss": 0.0015, "step": 35075 }, { "epoch": 6.43, "learning_rate": 1.9470840031677274e-06, "loss": 0.0018, "step": 35076 }, { "epoch": 6.43, "learning_rate": 1.9466438542341828e-06, "loss": 0.0025, "step": 35077 }, { "epoch": 6.43, "learning_rate": 1.9462037496909235e-06, "loss": 0.0011, "step": 35078 }, { "epoch": 6.43, "learning_rate": 1.94576368954038e-06, "loss": 0.0022, "step": 35079 }, { "epoch": 6.43, "learning_rate": 1.9453236737849747e-06, "loss": 0.0025, "step": 35080 }, { "epoch": 6.43, "learning_rate": 1.944883702427134e-06, "loss": 0.0027, "step": 35081 }, { "epoch": 6.43, "learning_rate": 1.94444377546928e-06, "loss": 0.0033, "step": 35082 }, { "epoch": 6.43, "learning_rate": 1.944003892913844e-06, "loss": 0.003, "step": 35083 }, { "epoch": 6.43, "learning_rate": 1.9435640547632497e-06, "loss": 0.0022, "step": 35084 }, { "epoch": 6.43, "learning_rate": 1.943124261019916e-06, "loss": 0.002, "step": 35085 }, { "epoch": 6.43, "learning_rate": 1.9426845116862694e-06, "loss": 0.0014, "step": 35086 }, { "epoch": 6.43, "learning_rate": 1.9422448067647347e-06, "loss": 0.0016, "step": 35087 }, { "epoch": 6.43, "learning_rate": 1.941805146257737e-06, "loss": 0.0023, "step": 35088 }, { "epoch": 6.43, "learning_rate": 1.941365530167697e-06, "loss": 0.0021, "step": 35089 }, { "epoch": 6.43, "learning_rate": 1.9409259584970396e-06, "loss": 0.0016, "step": 35090 }, { "epoch": 6.44, "learning_rate": 1.9404864312481873e-06, "loss": 0.002, "step": 35091 }, { "epoch": 6.44, "learning_rate": 1.9400469484235627e-06, "loss": 0.0022, "step": 35092 }, { "epoch": 6.44, "learning_rate": 1.939607510025587e-06, "loss": 0.0015, "step": 35093 }, { "epoch": 6.44, "learning_rate": 1.9391681160566843e-06, "loss": 0.0022, "step": 35094 }, { "epoch": 6.44, "learning_rate": 1.9387287665192766e-06, "loss": 0.0016, "step": 35095 }, { "epoch": 6.44, "learning_rate": 1.9382894614157843e-06, "loss": 0.0025, "step": 35096 }, { "epoch": 6.44, "learning_rate": 1.9378502007486288e-06, "loss": 0.0012, "step": 35097 }, { "epoch": 6.44, "learning_rate": 1.9374109845202314e-06, "loss": 0.0027, "step": 35098 }, { "epoch": 6.44, "learning_rate": 1.936971812733012e-06, "loss": 0.0022, "step": 35099 }, { "epoch": 6.44, "learning_rate": 1.936532685389395e-06, "loss": 0.0023, "step": 35100 }, { "epoch": 6.44, "learning_rate": 1.9360936024918e-06, "loss": 0.0026, "step": 35101 }, { "epoch": 6.44, "learning_rate": 1.935654564042647e-06, "loss": 0.0023, "step": 35102 }, { "epoch": 6.44, "learning_rate": 1.9352155700443533e-06, "loss": 0.0022, "step": 35103 }, { "epoch": 6.44, "learning_rate": 1.934776620499337e-06, "loss": 0.0024, "step": 35104 }, { "epoch": 6.44, "learning_rate": 1.9343377154100228e-06, "loss": 0.0027, "step": 35105 }, { "epoch": 6.44, "learning_rate": 1.9338988547788295e-06, "loss": 0.0019, "step": 35106 }, { "epoch": 6.44, "learning_rate": 1.9334600386081746e-06, "loss": 0.0013, "step": 35107 }, { "epoch": 6.44, "learning_rate": 1.9330212669004767e-06, "loss": 0.0038, "step": 35108 }, { "epoch": 6.44, "learning_rate": 1.932582539658154e-06, "loss": 0.0027, "step": 35109 }, { "epoch": 6.44, "learning_rate": 1.932143856883626e-06, "loss": 0.002, "step": 35110 }, { "epoch": 6.44, "learning_rate": 1.9317052185793093e-06, "loss": 0.0017, "step": 35111 }, { "epoch": 6.44, "learning_rate": 1.931266624747623e-06, "loss": 0.0021, "step": 35112 }, { "epoch": 6.44, "learning_rate": 1.9308280753909846e-06, "loss": 0.002, "step": 35113 }, { "epoch": 6.44, "learning_rate": 1.93038957051181e-06, "loss": 0.0014, "step": 35114 }, { "epoch": 6.44, "learning_rate": 1.929951110112518e-06, "loss": 0.0019, "step": 35115 }, { "epoch": 6.44, "learning_rate": 1.9295126941955244e-06, "loss": 0.0016, "step": 35116 }, { "epoch": 6.44, "learning_rate": 1.9290743227632424e-06, "loss": 0.0013, "step": 35117 }, { "epoch": 6.44, "learning_rate": 1.9286359958180956e-06, "loss": 0.0011, "step": 35118 }, { "epoch": 6.44, "learning_rate": 1.928197713362495e-06, "loss": 0.0016, "step": 35119 }, { "epoch": 6.44, "learning_rate": 1.9277594753988615e-06, "loss": 0.0013, "step": 35120 }, { "epoch": 6.44, "learning_rate": 1.9273212819296006e-06, "loss": 0.0022, "step": 35121 }, { "epoch": 6.44, "learning_rate": 1.926883132957137e-06, "loss": 0.0017, "step": 35122 }, { "epoch": 6.44, "learning_rate": 1.926445028483883e-06, "loss": 0.003, "step": 35123 }, { "epoch": 6.44, "learning_rate": 1.926006968512253e-06, "loss": 0.0012, "step": 35124 }, { "epoch": 6.44, "learning_rate": 1.925568953044662e-06, "loss": 0.0012, "step": 35125 }, { "epoch": 6.44, "learning_rate": 1.925130982083524e-06, "loss": 0.0022, "step": 35126 }, { "epoch": 6.44, "learning_rate": 1.9246930556312524e-06, "loss": 0.0017, "step": 35127 }, { "epoch": 6.44, "learning_rate": 1.9242551736902627e-06, "loss": 0.0015, "step": 35128 }, { "epoch": 6.44, "learning_rate": 1.9238173362629676e-06, "loss": 0.002, "step": 35129 }, { "epoch": 6.44, "learning_rate": 1.9233795433517775e-06, "loss": 0.0023, "step": 35130 }, { "epoch": 6.44, "learning_rate": 1.9229417949591145e-06, "loss": 0.0022, "step": 35131 }, { "epoch": 6.44, "learning_rate": 1.9225040910873815e-06, "loss": 0.0022, "step": 35132 }, { "epoch": 6.44, "learning_rate": 1.922066431738997e-06, "loss": 0.0017, "step": 35133 }, { "epoch": 6.44, "learning_rate": 1.92162881691637e-06, "loss": 0.0022, "step": 35134 }, { "epoch": 6.44, "learning_rate": 1.921191246621913e-06, "loss": 0.0026, "step": 35135 }, { "epoch": 6.44, "learning_rate": 1.9207537208580407e-06, "loss": 0.0021, "step": 35136 }, { "epoch": 6.44, "learning_rate": 1.9203162396271667e-06, "loss": 0.0025, "step": 35137 }, { "epoch": 6.44, "learning_rate": 1.919878802931695e-06, "loss": 0.0016, "step": 35138 }, { "epoch": 6.44, "learning_rate": 1.919441410774039e-06, "loss": 0.0015, "step": 35139 }, { "epoch": 6.44, "learning_rate": 1.919004063156614e-06, "loss": 0.002, "step": 35140 }, { "epoch": 6.44, "learning_rate": 1.918566760081827e-06, "loss": 0.0027, "step": 35141 }, { "epoch": 6.44, "learning_rate": 1.9181295015520906e-06, "loss": 0.0018, "step": 35142 }, { "epoch": 6.44, "learning_rate": 1.917692287569812e-06, "loss": 0.0023, "step": 35143 }, { "epoch": 6.44, "learning_rate": 1.9172551181374043e-06, "loss": 0.0018, "step": 35144 }, { "epoch": 6.45, "learning_rate": 1.916817993257275e-06, "loss": 0.0016, "step": 35145 }, { "epoch": 6.45, "learning_rate": 1.9163809129318343e-06, "loss": 0.0012, "step": 35146 }, { "epoch": 6.45, "learning_rate": 1.915943877163492e-06, "loss": 0.0019, "step": 35147 }, { "epoch": 6.45, "learning_rate": 1.9155068859546544e-06, "loss": 0.0021, "step": 35148 }, { "epoch": 6.45, "learning_rate": 1.9150699393077364e-06, "loss": 0.0021, "step": 35149 }, { "epoch": 6.45, "learning_rate": 1.91463303722514e-06, "loss": 0.0023, "step": 35150 }, { "epoch": 6.45, "learning_rate": 1.9141961797092758e-06, "loss": 0.0019, "step": 35151 }, { "epoch": 6.45, "learning_rate": 1.9137593667625486e-06, "loss": 0.0031, "step": 35152 }, { "epoch": 6.45, "learning_rate": 1.9133225983873714e-06, "loss": 0.0023, "step": 35153 }, { "epoch": 6.45, "learning_rate": 1.91288587458615e-06, "loss": 0.0016, "step": 35154 }, { "epoch": 6.45, "learning_rate": 1.9124491953612946e-06, "loss": 0.0014, "step": 35155 }, { "epoch": 6.45, "learning_rate": 1.9120125607152053e-06, "loss": 0.002, "step": 35156 }, { "epoch": 6.45, "learning_rate": 1.91157597065029e-06, "loss": 0.0018, "step": 35157 }, { "epoch": 6.45, "learning_rate": 1.9111394251689598e-06, "loss": 0.0028, "step": 35158 }, { "epoch": 6.45, "learning_rate": 1.9107029242736185e-06, "loss": 0.0015, "step": 35159 }, { "epoch": 6.45, "learning_rate": 1.9102664679666714e-06, "loss": 0.0025, "step": 35160 }, { "epoch": 6.45, "learning_rate": 1.9098300562505266e-06, "loss": 0.0012, "step": 35161 }, { "epoch": 6.45, "learning_rate": 1.909393689127588e-06, "loss": 0.0014, "step": 35162 }, { "epoch": 6.45, "learning_rate": 1.908957366600259e-06, "loss": 0.0018, "step": 35163 }, { "epoch": 6.45, "learning_rate": 1.9085210886709482e-06, "loss": 0.002, "step": 35164 }, { "epoch": 6.45, "learning_rate": 1.9080848553420563e-06, "loss": 0.0014, "step": 35165 }, { "epoch": 6.45, "learning_rate": 1.9076486666159944e-06, "loss": 0.0024, "step": 35166 }, { "epoch": 6.45, "learning_rate": 1.9072125224951609e-06, "loss": 0.0022, "step": 35167 }, { "epoch": 6.45, "learning_rate": 1.9067764229819608e-06, "loss": 0.0019, "step": 35168 }, { "epoch": 6.45, "learning_rate": 1.9063403680787994e-06, "loss": 0.0019, "step": 35169 }, { "epoch": 6.45, "learning_rate": 1.9059043577880766e-06, "loss": 0.0031, "step": 35170 }, { "epoch": 6.45, "learning_rate": 1.9054683921122009e-06, "loss": 0.0034, "step": 35171 }, { "epoch": 6.45, "learning_rate": 1.905032471053574e-06, "loss": 0.0023, "step": 35172 }, { "epoch": 6.45, "learning_rate": 1.9045965946145994e-06, "loss": 0.0022, "step": 35173 }, { "epoch": 6.45, "learning_rate": 1.9041607627976733e-06, "loss": 0.0026, "step": 35174 }, { "epoch": 6.45, "learning_rate": 1.9037249756052056e-06, "loss": 0.0021, "step": 35175 }, { "epoch": 6.45, "learning_rate": 1.9032892330395946e-06, "loss": 0.0033, "step": 35176 }, { "epoch": 6.45, "learning_rate": 1.902853535103244e-06, "loss": 0.0018, "step": 35177 }, { "epoch": 6.45, "learning_rate": 1.9024178817985538e-06, "loss": 0.0025, "step": 35178 }, { "epoch": 6.45, "learning_rate": 1.9019822731279258e-06, "loss": 0.003, "step": 35179 }, { "epoch": 6.45, "learning_rate": 1.901546709093761e-06, "loss": 0.0017, "step": 35180 }, { "epoch": 6.45, "learning_rate": 1.9011111896984614e-06, "loss": 0.0023, "step": 35181 }, { "epoch": 6.45, "learning_rate": 1.9006757149444266e-06, "loss": 0.0024, "step": 35182 }, { "epoch": 6.45, "learning_rate": 1.900240284834054e-06, "loss": 0.0019, "step": 35183 }, { "epoch": 6.45, "learning_rate": 1.8998048993697516e-06, "loss": 0.0015, "step": 35184 }, { "epoch": 6.45, "learning_rate": 1.8993695585539128e-06, "loss": 0.0023, "step": 35185 }, { "epoch": 6.45, "learning_rate": 1.8989342623889373e-06, "loss": 0.0021, "step": 35186 }, { "epoch": 6.45, "learning_rate": 1.8984990108772273e-06, "loss": 0.0016, "step": 35187 }, { "epoch": 6.45, "learning_rate": 1.8980638040211774e-06, "loss": 0.0028, "step": 35188 }, { "epoch": 6.45, "learning_rate": 1.8976286418231916e-06, "loss": 0.0025, "step": 35189 }, { "epoch": 6.45, "learning_rate": 1.8971935242856676e-06, "loss": 0.0022, "step": 35190 }, { "epoch": 6.45, "learning_rate": 1.8967584514110049e-06, "loss": 0.0018, "step": 35191 }, { "epoch": 6.45, "learning_rate": 1.8963234232015937e-06, "loss": 0.001, "step": 35192 }, { "epoch": 6.45, "learning_rate": 1.895888439659841e-06, "loss": 0.0019, "step": 35193 }, { "epoch": 6.45, "learning_rate": 1.8954535007881415e-06, "loss": 0.0022, "step": 35194 }, { "epoch": 6.45, "learning_rate": 1.895018606588891e-06, "loss": 0.0026, "step": 35195 }, { "epoch": 6.45, "learning_rate": 1.8945837570644887e-06, "loss": 0.0015, "step": 35196 }, { "epoch": 6.45, "learning_rate": 1.8941489522173295e-06, "loss": 0.0018, "step": 35197 }, { "epoch": 6.45, "learning_rate": 1.8937141920498125e-06, "loss": 0.0027, "step": 35198 }, { "epoch": 6.45, "learning_rate": 1.8932794765643326e-06, "loss": 0.0031, "step": 35199 }, { "epoch": 6.46, "learning_rate": 1.8928448057632854e-06, "loss": 0.0014, "step": 35200 }, { "epoch": 6.46, "learning_rate": 1.8924101796490657e-06, "loss": 0.0029, "step": 35201 }, { "epoch": 6.46, "learning_rate": 1.891975598224075e-06, "loss": 0.003, "step": 35202 }, { "epoch": 6.46, "learning_rate": 1.8915410614907038e-06, "loss": 0.0009, "step": 35203 }, { "epoch": 6.46, "learning_rate": 1.8911065694513474e-06, "loss": 0.0019, "step": 35204 }, { "epoch": 6.46, "learning_rate": 1.890672122108399e-06, "loss": 0.0015, "step": 35205 }, { "epoch": 6.46, "learning_rate": 1.8902377194642584e-06, "loss": 0.0021, "step": 35206 }, { "epoch": 6.46, "learning_rate": 1.8898033615213173e-06, "loss": 0.003, "step": 35207 }, { "epoch": 6.46, "learning_rate": 1.8893690482819705e-06, "loss": 0.0016, "step": 35208 }, { "epoch": 6.46, "learning_rate": 1.8889347797486135e-06, "loss": 0.0024, "step": 35209 }, { "epoch": 6.46, "learning_rate": 1.8885005559236335e-06, "loss": 0.0011, "step": 35210 }, { "epoch": 6.46, "learning_rate": 1.888066376809431e-06, "loss": 0.0018, "step": 35211 }, { "epoch": 6.46, "learning_rate": 1.8876322424083959e-06, "loss": 0.0027, "step": 35212 }, { "epoch": 6.46, "learning_rate": 1.8871981527229223e-06, "loss": 0.003, "step": 35213 }, { "epoch": 6.46, "learning_rate": 1.8867641077554033e-06, "loss": 0.0029, "step": 35214 }, { "epoch": 6.46, "learning_rate": 1.8863301075082296e-06, "loss": 0.003, "step": 35215 }, { "epoch": 6.46, "learning_rate": 1.8858961519837948e-06, "loss": 0.0029, "step": 35216 }, { "epoch": 6.46, "learning_rate": 1.8854622411844913e-06, "loss": 0.0028, "step": 35217 }, { "epoch": 6.46, "learning_rate": 1.8850283751127063e-06, "loss": 0.0013, "step": 35218 }, { "epoch": 6.46, "learning_rate": 1.8845945537708376e-06, "loss": 0.0028, "step": 35219 }, { "epoch": 6.46, "learning_rate": 1.8841607771612768e-06, "loss": 0.0024, "step": 35220 }, { "epoch": 6.46, "learning_rate": 1.8837270452864098e-06, "loss": 0.0021, "step": 35221 }, { "epoch": 6.46, "learning_rate": 1.883293358148629e-06, "loss": 0.0023, "step": 35222 }, { "epoch": 6.46, "learning_rate": 1.8828597157503237e-06, "loss": 0.0022, "step": 35223 }, { "epoch": 6.46, "learning_rate": 1.8824261180938875e-06, "loss": 0.0012, "step": 35224 }, { "epoch": 6.46, "learning_rate": 1.8819925651817084e-06, "loss": 0.0031, "step": 35225 }, { "epoch": 6.46, "learning_rate": 1.8815590570161768e-06, "loss": 0.0025, "step": 35226 }, { "epoch": 6.46, "learning_rate": 1.881125593599682e-06, "loss": 0.0018, "step": 35227 }, { "epoch": 6.46, "learning_rate": 1.880692174934613e-06, "loss": 0.0021, "step": 35228 }, { "epoch": 6.46, "learning_rate": 1.880258801023359e-06, "loss": 0.003, "step": 35229 }, { "epoch": 6.46, "learning_rate": 1.8798254718683085e-06, "loss": 0.0022, "step": 35230 }, { "epoch": 6.46, "learning_rate": 1.8793921874718502e-06, "loss": 0.002, "step": 35231 }, { "epoch": 6.46, "learning_rate": 1.8789589478363724e-06, "loss": 0.0019, "step": 35232 }, { "epoch": 6.46, "learning_rate": 1.8785257529642641e-06, "loss": 0.0025, "step": 35233 }, { "epoch": 6.46, "learning_rate": 1.8780926028579115e-06, "loss": 0.0013, "step": 35234 }, { "epoch": 6.46, "learning_rate": 1.8776594975197028e-06, "loss": 0.0013, "step": 35235 }, { "epoch": 6.46, "learning_rate": 1.877226436952022e-06, "loss": 0.0022, "step": 35236 }, { "epoch": 6.46, "learning_rate": 1.8767934211572624e-06, "loss": 0.0017, "step": 35237 }, { "epoch": 6.46, "learning_rate": 1.8763604501378108e-06, "loss": 0.0013, "step": 35238 }, { "epoch": 6.46, "learning_rate": 1.8759275238960473e-06, "loss": 0.0013, "step": 35239 }, { "epoch": 6.46, "learning_rate": 1.8754946424343622e-06, "loss": 0.0027, "step": 35240 }, { "epoch": 6.46, "learning_rate": 1.8750618057551384e-06, "loss": 0.0026, "step": 35241 }, { "epoch": 6.46, "learning_rate": 1.8746290138607658e-06, "loss": 0.0018, "step": 35242 }, { "epoch": 6.46, "learning_rate": 1.8741962667536296e-06, "loss": 0.0019, "step": 35243 }, { "epoch": 6.46, "learning_rate": 1.8737635644361162e-06, "loss": 0.0022, "step": 35244 }, { "epoch": 6.46, "learning_rate": 1.8733309069106043e-06, "loss": 0.002, "step": 35245 }, { "epoch": 6.46, "learning_rate": 1.8728982941794838e-06, "loss": 0.0015, "step": 35246 }, { "epoch": 6.46, "learning_rate": 1.8724657262451384e-06, "loss": 0.0011, "step": 35247 }, { "epoch": 6.46, "learning_rate": 1.872033203109953e-06, "loss": 0.0024, "step": 35248 }, { "epoch": 6.46, "learning_rate": 1.8716007247763113e-06, "loss": 0.0018, "step": 35249 }, { "epoch": 6.46, "learning_rate": 1.8711682912465967e-06, "loss": 0.0027, "step": 35250 }, { "epoch": 6.46, "learning_rate": 1.8707359025231919e-06, "loss": 0.0018, "step": 35251 }, { "epoch": 6.46, "learning_rate": 1.8703035586084817e-06, "loss": 0.0025, "step": 35252 }, { "epoch": 6.46, "learning_rate": 1.8698712595048495e-06, "loss": 0.0015, "step": 35253 }, { "epoch": 6.47, "learning_rate": 1.8694390052146737e-06, "loss": 0.0019, "step": 35254 }, { "epoch": 6.47, "learning_rate": 1.8690067957403447e-06, "loss": 0.0018, "step": 35255 }, { "epoch": 6.47, "learning_rate": 1.8685746310842423e-06, "loss": 0.003, "step": 35256 }, { "epoch": 6.47, "learning_rate": 1.8681425112487438e-06, "loss": 0.002, "step": 35257 }, { "epoch": 6.47, "learning_rate": 1.8677104362362331e-06, "loss": 0.0022, "step": 35258 }, { "epoch": 6.47, "learning_rate": 1.8672784060490946e-06, "loss": 0.0027, "step": 35259 }, { "epoch": 6.47, "learning_rate": 1.8668464206897085e-06, "loss": 0.0012, "step": 35260 }, { "epoch": 6.47, "learning_rate": 1.8664144801604556e-06, "loss": 0.0015, "step": 35261 }, { "epoch": 6.47, "learning_rate": 1.865982584463718e-06, "loss": 0.0015, "step": 35262 }, { "epoch": 6.47, "learning_rate": 1.8655507336018708e-06, "loss": 0.0021, "step": 35263 }, { "epoch": 6.47, "learning_rate": 1.8651189275773018e-06, "loss": 0.0015, "step": 35264 }, { "epoch": 6.47, "learning_rate": 1.8646871663923871e-06, "loss": 0.0018, "step": 35265 }, { "epoch": 6.47, "learning_rate": 1.864255450049507e-06, "loss": 0.0015, "step": 35266 }, { "epoch": 6.47, "learning_rate": 1.8638237785510416e-06, "loss": 0.0024, "step": 35267 }, { "epoch": 6.47, "learning_rate": 1.8633921518993703e-06, "loss": 0.0034, "step": 35268 }, { "epoch": 6.47, "learning_rate": 1.8629605700968723e-06, "loss": 0.0016, "step": 35269 }, { "epoch": 6.47, "learning_rate": 1.8625290331459255e-06, "loss": 0.0017, "step": 35270 }, { "epoch": 6.47, "learning_rate": 1.8620975410489083e-06, "loss": 0.0019, "step": 35271 }, { "epoch": 6.47, "learning_rate": 1.861666093808201e-06, "loss": 0.001, "step": 35272 }, { "epoch": 6.47, "learning_rate": 1.8612346914261815e-06, "loss": 0.0024, "step": 35273 }, { "epoch": 6.47, "learning_rate": 1.86080333390523e-06, "loss": 0.0023, "step": 35274 }, { "epoch": 6.47, "learning_rate": 1.8603720212477182e-06, "loss": 0.0017, "step": 35275 }, { "epoch": 6.47, "learning_rate": 1.8599407534560242e-06, "loss": 0.0029, "step": 35276 }, { "epoch": 6.47, "learning_rate": 1.8595095305325294e-06, "loss": 0.0032, "step": 35277 }, { "epoch": 6.47, "learning_rate": 1.8590783524796097e-06, "loss": 0.0016, "step": 35278 }, { "epoch": 6.47, "learning_rate": 1.8586472192996408e-06, "loss": 0.0013, "step": 35279 }, { "epoch": 6.47, "learning_rate": 1.8582161309949987e-06, "loss": 0.0015, "step": 35280 }, { "epoch": 6.47, "learning_rate": 1.8577850875680593e-06, "loss": 0.0016, "step": 35281 }, { "epoch": 6.47, "learning_rate": 1.8573540890212006e-06, "loss": 0.0027, "step": 35282 }, { "epoch": 6.47, "learning_rate": 1.8569231353567962e-06, "loss": 0.0015, "step": 35283 }, { "epoch": 6.47, "learning_rate": 1.8564922265772224e-06, "loss": 0.0011, "step": 35284 }, { "epoch": 6.47, "learning_rate": 1.8560613626848544e-06, "loss": 0.0017, "step": 35285 }, { "epoch": 6.47, "learning_rate": 1.8556305436820665e-06, "loss": 0.0022, "step": 35286 }, { "epoch": 6.47, "learning_rate": 1.8551997695712343e-06, "loss": 0.0018, "step": 35287 }, { "epoch": 6.47, "learning_rate": 1.8547690403547314e-06, "loss": 0.0026, "step": 35288 }, { "epoch": 6.47, "learning_rate": 1.8543383560349304e-06, "loss": 0.0026, "step": 35289 }, { "epoch": 6.47, "learning_rate": 1.8539077166142094e-06, "loss": 0.0016, "step": 35290 }, { "epoch": 6.47, "learning_rate": 1.8534771220949422e-06, "loss": 0.0019, "step": 35291 }, { "epoch": 6.47, "learning_rate": 1.8530465724794978e-06, "loss": 0.0014, "step": 35292 }, { "epoch": 6.47, "learning_rate": 1.8526160677702488e-06, "loss": 0.0037, "step": 35293 }, { "epoch": 6.47, "learning_rate": 1.8521856079695732e-06, "loss": 0.0022, "step": 35294 }, { "epoch": 6.47, "learning_rate": 1.8517551930798417e-06, "loss": 0.0018, "step": 35295 }, { "epoch": 6.47, "learning_rate": 1.8513248231034264e-06, "loss": 0.0021, "step": 35296 }, { "epoch": 6.47, "learning_rate": 1.8508944980427e-06, "loss": 0.0019, "step": 35297 }, { "epoch": 6.47, "learning_rate": 1.8504642179000344e-06, "loss": 0.0015, "step": 35298 }, { "epoch": 6.47, "learning_rate": 1.8500339826778003e-06, "loss": 0.0025, "step": 35299 }, { "epoch": 6.47, "learning_rate": 1.8496037923783694e-06, "loss": 0.0023, "step": 35300 }, { "epoch": 6.47, "learning_rate": 1.8491736470041132e-06, "loss": 0.001, "step": 35301 }, { "epoch": 6.47, "learning_rate": 1.848743546557401e-06, "loss": 0.0015, "step": 35302 }, { "epoch": 6.47, "learning_rate": 1.8483134910406109e-06, "loss": 0.0026, "step": 35303 }, { "epoch": 6.47, "learning_rate": 1.847883480456104e-06, "loss": 0.0016, "step": 35304 }, { "epoch": 6.47, "learning_rate": 1.8474535148062555e-06, "loss": 0.0015, "step": 35305 }, { "epoch": 6.47, "learning_rate": 1.8470235940934333e-06, "loss": 0.0008, "step": 35306 }, { "epoch": 6.47, "learning_rate": 1.8465937183200056e-06, "loss": 0.0013, "step": 35307 }, { "epoch": 6.47, "learning_rate": 1.8461638874883469e-06, "loss": 0.0035, "step": 35308 }, { "epoch": 6.48, "learning_rate": 1.8457341016008268e-06, "loss": 0.0024, "step": 35309 }, { "epoch": 6.48, "learning_rate": 1.8453043606598087e-06, "loss": 0.0021, "step": 35310 }, { "epoch": 6.48, "learning_rate": 1.8448746646676618e-06, "loss": 0.0032, "step": 35311 }, { "epoch": 6.48, "learning_rate": 1.8444450136267588e-06, "loss": 0.0017, "step": 35312 }, { "epoch": 6.48, "learning_rate": 1.8440154075394667e-06, "loss": 0.0013, "step": 35313 }, { "epoch": 6.48, "learning_rate": 1.8435858464081525e-06, "loss": 0.0021, "step": 35314 }, { "epoch": 6.48, "learning_rate": 1.843156330235183e-06, "loss": 0.001, "step": 35315 }, { "epoch": 6.48, "learning_rate": 1.8427268590229286e-06, "loss": 0.0027, "step": 35316 }, { "epoch": 6.48, "learning_rate": 1.8422974327737541e-06, "loss": 0.0013, "step": 35317 }, { "epoch": 6.48, "learning_rate": 1.8418680514900267e-06, "loss": 0.0015, "step": 35318 }, { "epoch": 6.48, "learning_rate": 1.8414387151741142e-06, "loss": 0.002, "step": 35319 }, { "epoch": 6.48, "learning_rate": 1.8410094238283805e-06, "loss": 0.0029, "step": 35320 }, { "epoch": 6.48, "learning_rate": 1.840580177455199e-06, "loss": 0.0026, "step": 35321 }, { "epoch": 6.48, "learning_rate": 1.840150976056927e-06, "loss": 0.002, "step": 35322 }, { "epoch": 6.48, "learning_rate": 1.8397218196359356e-06, "loss": 0.0015, "step": 35323 }, { "epoch": 6.48, "learning_rate": 1.8392927081945854e-06, "loss": 0.0025, "step": 35324 }, { "epoch": 6.48, "learning_rate": 1.8388636417352468e-06, "loss": 0.0015, "step": 35325 }, { "epoch": 6.48, "learning_rate": 1.8384346202602844e-06, "loss": 0.0025, "step": 35326 }, { "epoch": 6.48, "learning_rate": 1.838005643772064e-06, "loss": 0.0015, "step": 35327 }, { "epoch": 6.48, "learning_rate": 1.8375767122729437e-06, "loss": 0.0016, "step": 35328 }, { "epoch": 6.48, "learning_rate": 1.8371478257652908e-06, "loss": 0.0025, "step": 35329 }, { "epoch": 6.48, "learning_rate": 1.8367189842514722e-06, "loss": 0.0026, "step": 35330 }, { "epoch": 6.48, "learning_rate": 1.83629018773385e-06, "loss": 0.0021, "step": 35331 }, { "epoch": 6.48, "learning_rate": 1.8358614362147876e-06, "loss": 0.0016, "step": 35332 }, { "epoch": 6.48, "learning_rate": 1.8354327296966478e-06, "loss": 0.0019, "step": 35333 }, { "epoch": 6.48, "learning_rate": 1.8350040681817938e-06, "loss": 0.0019, "step": 35334 }, { "epoch": 6.48, "learning_rate": 1.83457545167259e-06, "loss": 0.001, "step": 35335 }, { "epoch": 6.48, "learning_rate": 1.8341468801713958e-06, "loss": 0.002, "step": 35336 }, { "epoch": 6.48, "learning_rate": 1.8337183536805769e-06, "loss": 0.0031, "step": 35337 }, { "epoch": 6.48, "learning_rate": 1.8332898722024927e-06, "loss": 0.0027, "step": 35338 }, { "epoch": 6.48, "learning_rate": 1.832861435739507e-06, "loss": 0.0015, "step": 35339 }, { "epoch": 6.48, "learning_rate": 1.8324330442939798e-06, "loss": 0.0014, "step": 35340 }, { "epoch": 6.48, "learning_rate": 1.8320046978682727e-06, "loss": 0.0022, "step": 35341 }, { "epoch": 6.48, "learning_rate": 1.8315763964647448e-06, "loss": 0.0016, "step": 35342 }, { "epoch": 6.48, "learning_rate": 1.8311481400857622e-06, "loss": 0.003, "step": 35343 }, { "epoch": 6.48, "learning_rate": 1.830719928733682e-06, "loss": 0.0019, "step": 35344 }, { "epoch": 6.48, "learning_rate": 1.8302917624108674e-06, "loss": 0.0027, "step": 35345 }, { "epoch": 6.48, "learning_rate": 1.8298636411196713e-06, "loss": 0.0021, "step": 35346 }, { "epoch": 6.48, "learning_rate": 1.8294355648624607e-06, "loss": 0.0021, "step": 35347 }, { "epoch": 6.48, "learning_rate": 1.8290075336415924e-06, "loss": 0.0017, "step": 35348 }, { "epoch": 6.48, "learning_rate": 1.828579547459426e-06, "loss": 0.0009, "step": 35349 }, { "epoch": 6.48, "learning_rate": 1.8281516063183202e-06, "loss": 0.0013, "step": 35350 }, { "epoch": 6.48, "learning_rate": 1.8277237102206346e-06, "loss": 0.0016, "step": 35351 }, { "epoch": 6.48, "learning_rate": 1.8272958591687272e-06, "loss": 0.0023, "step": 35352 }, { "epoch": 6.48, "learning_rate": 1.8268680531649574e-06, "loss": 0.0022, "step": 35353 }, { "epoch": 6.48, "learning_rate": 1.826440292211682e-06, "loss": 0.0006, "step": 35354 }, { "epoch": 6.48, "learning_rate": 1.8260125763112557e-06, "loss": 0.0021, "step": 35355 }, { "epoch": 6.48, "learning_rate": 1.8255849054660457e-06, "loss": 0.0015, "step": 35356 }, { "epoch": 6.48, "learning_rate": 1.8251572796784e-06, "loss": 0.0028, "step": 35357 }, { "epoch": 6.48, "learning_rate": 1.824729698950679e-06, "loss": 0.0023, "step": 35358 }, { "epoch": 6.48, "learning_rate": 1.8243021632852398e-06, "loss": 0.003, "step": 35359 }, { "epoch": 6.48, "learning_rate": 1.823874672684436e-06, "loss": 0.002, "step": 35360 }, { "epoch": 6.48, "learning_rate": 1.8234472271506288e-06, "loss": 0.0032, "step": 35361 }, { "epoch": 6.48, "learning_rate": 1.8230198266861722e-06, "loss": 0.0017, "step": 35362 }, { "epoch": 6.49, "learning_rate": 1.8225924712934239e-06, "loss": 0.0024, "step": 35363 }, { "epoch": 6.49, "learning_rate": 1.8221651609747337e-06, "loss": 0.0026, "step": 35364 }, { "epoch": 6.49, "learning_rate": 1.8217378957324616e-06, "loss": 0.0017, "step": 35365 }, { "epoch": 6.49, "learning_rate": 1.8213106755689636e-06, "loss": 0.0015, "step": 35366 }, { "epoch": 6.49, "learning_rate": 1.820883500486591e-06, "loss": 0.0022, "step": 35367 }, { "epoch": 6.49, "learning_rate": 1.820456370487701e-06, "loss": 0.0023, "step": 35368 }, { "epoch": 6.49, "learning_rate": 1.820029285574647e-06, "loss": 0.0019, "step": 35369 }, { "epoch": 6.49, "learning_rate": 1.8196022457497841e-06, "loss": 0.0025, "step": 35370 }, { "epoch": 6.49, "learning_rate": 1.8191752510154636e-06, "loss": 0.0022, "step": 35371 }, { "epoch": 6.49, "learning_rate": 1.8187483013740426e-06, "loss": 0.0013, "step": 35372 }, { "epoch": 6.49, "learning_rate": 1.818321396827869e-06, "loss": 0.002, "step": 35373 }, { "epoch": 6.49, "learning_rate": 1.8178945373793045e-06, "loss": 0.0008, "step": 35374 }, { "epoch": 6.49, "learning_rate": 1.8174677230306947e-06, "loss": 0.0024, "step": 35375 }, { "epoch": 6.49, "learning_rate": 1.8170409537843948e-06, "loss": 0.0012, "step": 35376 }, { "epoch": 6.49, "learning_rate": 1.8166142296427535e-06, "loss": 0.0027, "step": 35377 }, { "epoch": 6.49, "learning_rate": 1.8161875506081294e-06, "loss": 0.0018, "step": 35378 }, { "epoch": 6.49, "learning_rate": 1.8157609166828715e-06, "loss": 0.0017, "step": 35379 }, { "epoch": 6.49, "learning_rate": 1.8153343278693314e-06, "loss": 0.0024, "step": 35380 }, { "epoch": 6.49, "learning_rate": 1.8149077841698615e-06, "loss": 0.0012, "step": 35381 }, { "epoch": 6.49, "learning_rate": 1.8144812855868076e-06, "loss": 0.001, "step": 35382 }, { "epoch": 6.49, "learning_rate": 1.814054832122527e-06, "loss": 0.0015, "step": 35383 }, { "epoch": 6.49, "learning_rate": 1.8136284237793677e-06, "loss": 0.0048, "step": 35384 }, { "epoch": 6.49, "learning_rate": 1.8132020605596801e-06, "loss": 0.0022, "step": 35385 }, { "epoch": 6.49, "learning_rate": 1.8127757424658155e-06, "loss": 0.0014, "step": 35386 }, { "epoch": 6.49, "learning_rate": 1.8123494695001221e-06, "loss": 0.0009, "step": 35387 }, { "epoch": 6.49, "learning_rate": 1.8119232416649502e-06, "loss": 0.0025, "step": 35388 }, { "epoch": 6.49, "learning_rate": 1.811497058962649e-06, "loss": 0.0034, "step": 35389 }, { "epoch": 6.49, "learning_rate": 1.811070921395569e-06, "loss": 0.0017, "step": 35390 }, { "epoch": 6.49, "learning_rate": 1.8106448289660538e-06, "loss": 0.0015, "step": 35391 }, { "epoch": 6.49, "learning_rate": 1.8102187816764626e-06, "loss": 0.0018, "step": 35392 }, { "epoch": 6.49, "learning_rate": 1.8097927795291336e-06, "loss": 0.0029, "step": 35393 }, { "epoch": 6.49, "learning_rate": 1.8093668225264195e-06, "loss": 0.0025, "step": 35394 }, { "epoch": 6.49, "learning_rate": 1.8089409106706645e-06, "loss": 0.0019, "step": 35395 }, { "epoch": 6.49, "learning_rate": 1.8085150439642218e-06, "loss": 0.002, "step": 35396 }, { "epoch": 6.49, "learning_rate": 1.808089222409436e-06, "loss": 0.0037, "step": 35397 }, { "epoch": 6.49, "learning_rate": 1.8076634460086573e-06, "loss": 0.0021, "step": 35398 }, { "epoch": 6.49, "learning_rate": 1.8072377147642239e-06, "loss": 0.0015, "step": 35399 }, { "epoch": 6.49, "learning_rate": 1.8068120286784908e-06, "loss": 0.0017, "step": 35400 }, { "epoch": 6.49, "learning_rate": 1.8063863877538024e-06, "loss": 0.0016, "step": 35401 }, { "epoch": 6.49, "learning_rate": 1.8059607919925038e-06, "loss": 0.0017, "step": 35402 }, { "epoch": 6.49, "learning_rate": 1.8055352413969408e-06, "loss": 0.0015, "step": 35403 }, { "epoch": 6.49, "learning_rate": 1.8051097359694603e-06, "loss": 0.0018, "step": 35404 }, { "epoch": 6.49, "learning_rate": 1.804684275712406e-06, "loss": 0.0016, "step": 35405 }, { "epoch": 6.49, "learning_rate": 1.8042588606281241e-06, "loss": 0.002, "step": 35406 }, { "epoch": 6.49, "learning_rate": 1.8038334907189602e-06, "loss": 0.0017, "step": 35407 }, { "epoch": 6.49, "learning_rate": 1.8034081659872548e-06, "loss": 0.0028, "step": 35408 }, { "epoch": 6.49, "learning_rate": 1.8029828864353583e-06, "loss": 0.0017, "step": 35409 }, { "epoch": 6.49, "learning_rate": 1.8025576520656152e-06, "loss": 0.0017, "step": 35410 }, { "epoch": 6.49, "learning_rate": 1.8021324628803638e-06, "loss": 0.0024, "step": 35411 }, { "epoch": 6.49, "learning_rate": 1.80170731888195e-06, "loss": 0.0019, "step": 35412 }, { "epoch": 6.49, "learning_rate": 1.8012822200727153e-06, "loss": 0.0022, "step": 35413 }, { "epoch": 6.49, "learning_rate": 1.8008571664550067e-06, "loss": 0.0005, "step": 35414 }, { "epoch": 6.49, "learning_rate": 1.8004321580311667e-06, "loss": 0.0015, "step": 35415 }, { "epoch": 6.49, "learning_rate": 1.800007194803538e-06, "loss": 0.0025, "step": 35416 }, { "epoch": 6.49, "learning_rate": 1.7995822767744574e-06, "loss": 0.0026, "step": 35417 }, { "epoch": 6.5, "learning_rate": 1.7991574039462745e-06, "loss": 0.0016, "step": 35418 }, { "epoch": 6.5, "learning_rate": 1.798732576321327e-06, "loss": 0.0012, "step": 35419 }, { "epoch": 6.5, "learning_rate": 1.7983077939019577e-06, "loss": 0.003, "step": 35420 }, { "epoch": 6.5, "learning_rate": 1.7978830566905092e-06, "loss": 0.0016, "step": 35421 }, { "epoch": 6.5, "learning_rate": 1.7974583646893206e-06, "loss": 0.0021, "step": 35422 }, { "epoch": 6.5, "learning_rate": 1.7970337179007335e-06, "loss": 0.0033, "step": 35423 }, { "epoch": 6.5, "learning_rate": 1.7966091163270893e-06, "loss": 0.0018, "step": 35424 }, { "epoch": 6.5, "learning_rate": 1.7961845599707273e-06, "loss": 0.0019, "step": 35425 }, { "epoch": 6.5, "learning_rate": 1.7957600488339854e-06, "loss": 0.002, "step": 35426 }, { "epoch": 6.5, "learning_rate": 1.7953355829192098e-06, "loss": 0.0014, "step": 35427 }, { "epoch": 6.5, "learning_rate": 1.7949111622287385e-06, "loss": 0.0017, "step": 35428 }, { "epoch": 6.5, "learning_rate": 1.7944867867649074e-06, "loss": 0.0022, "step": 35429 }, { "epoch": 6.5, "learning_rate": 1.7940624565300547e-06, "loss": 0.0015, "step": 35430 }, { "epoch": 6.5, "learning_rate": 1.7936381715265239e-06, "loss": 0.0013, "step": 35431 }, { "epoch": 6.5, "learning_rate": 1.793213931756651e-06, "loss": 0.001, "step": 35432 }, { "epoch": 6.5, "learning_rate": 1.7927897372227764e-06, "loss": 0.0018, "step": 35433 }, { "epoch": 6.5, "learning_rate": 1.7923655879272395e-06, "loss": 0.0015, "step": 35434 }, { "epoch": 6.5, "learning_rate": 1.7919414838723704e-06, "loss": 0.0025, "step": 35435 }, { "epoch": 6.5, "learning_rate": 1.791517425060515e-06, "loss": 0.0016, "step": 35436 }, { "epoch": 6.5, "learning_rate": 1.791093411494007e-06, "loss": 0.0019, "step": 35437 }, { "epoch": 6.5, "learning_rate": 1.790669443175186e-06, "loss": 0.0026, "step": 35438 }, { "epoch": 6.5, "learning_rate": 1.7902455201063863e-06, "loss": 0.002, "step": 35439 }, { "epoch": 6.5, "learning_rate": 1.789821642289945e-06, "loss": 0.0026, "step": 35440 }, { "epoch": 6.5, "learning_rate": 1.7893978097282005e-06, "loss": 0.0022, "step": 35441 }, { "epoch": 6.5, "learning_rate": 1.7889740224234863e-06, "loss": 0.0039, "step": 35442 }, { "epoch": 6.5, "learning_rate": 1.7885502803781407e-06, "loss": 0.0012, "step": 35443 }, { "epoch": 6.5, "learning_rate": 1.788126583594495e-06, "loss": 0.0012, "step": 35444 }, { "epoch": 6.5, "learning_rate": 1.787702932074894e-06, "loss": 0.0023, "step": 35445 }, { "epoch": 6.5, "learning_rate": 1.7872793258216636e-06, "loss": 0.0021, "step": 35446 }, { "epoch": 6.5, "learning_rate": 1.786855764837142e-06, "loss": 0.0011, "step": 35447 }, { "epoch": 6.5, "learning_rate": 1.7864322491236619e-06, "loss": 0.0008, "step": 35448 }, { "epoch": 6.5, "learning_rate": 1.786008778683561e-06, "loss": 0.0017, "step": 35449 }, { "epoch": 6.5, "learning_rate": 1.7855853535191736e-06, "loss": 0.0019, "step": 35450 }, { "epoch": 6.5, "learning_rate": 1.7851619736328307e-06, "loss": 0.0027, "step": 35451 }, { "epoch": 6.5, "learning_rate": 1.7847386390268684e-06, "loss": 0.002, "step": 35452 }, { "epoch": 6.5, "learning_rate": 1.7843153497036191e-06, "loss": 0.0022, "step": 35453 }, { "epoch": 6.5, "learning_rate": 1.7838921056654158e-06, "loss": 0.003, "step": 35454 }, { "epoch": 6.5, "learning_rate": 1.7834689069145917e-06, "loss": 0.0023, "step": 35455 }, { "epoch": 6.5, "learning_rate": 1.7830457534534796e-06, "loss": 0.0015, "step": 35456 }, { "epoch": 6.5, "learning_rate": 1.782622645284412e-06, "loss": 0.0025, "step": 35457 }, { "epoch": 6.5, "learning_rate": 1.7821995824097205e-06, "loss": 0.0013, "step": 35458 }, { "epoch": 6.5, "learning_rate": 1.7817765648317376e-06, "loss": 0.0012, "step": 35459 }, { "epoch": 6.5, "learning_rate": 1.7813535925527957e-06, "loss": 0.0029, "step": 35460 }, { "epoch": 6.5, "learning_rate": 1.7809306655752213e-06, "loss": 0.0026, "step": 35461 }, { "epoch": 6.5, "learning_rate": 1.7805077839013529e-06, "loss": 0.0015, "step": 35462 }, { "epoch": 6.5, "learning_rate": 1.7800849475335213e-06, "loss": 0.0017, "step": 35463 }, { "epoch": 6.5, "learning_rate": 1.77966215647405e-06, "loss": 0.0023, "step": 35464 }, { "epoch": 6.5, "learning_rate": 1.7792394107252742e-06, "loss": 0.0024, "step": 35465 }, { "epoch": 6.5, "learning_rate": 1.7788167102895214e-06, "loss": 0.0019, "step": 35466 }, { "epoch": 6.5, "learning_rate": 1.778394055169126e-06, "loss": 0.001, "step": 35467 }, { "epoch": 6.5, "learning_rate": 1.7779714453664142e-06, "loss": 0.0016, "step": 35468 }, { "epoch": 6.5, "learning_rate": 1.777548880883717e-06, "loss": 0.0034, "step": 35469 }, { "epoch": 6.5, "learning_rate": 1.7771263617233624e-06, "loss": 0.0008, "step": 35470 }, { "epoch": 6.5, "learning_rate": 1.7767038878876809e-06, "loss": 0.0019, "step": 35471 }, { "epoch": 6.51, "learning_rate": 1.776281459379e-06, "loss": 0.0012, "step": 35472 }, { "epoch": 6.51, "learning_rate": 1.7758590761996486e-06, "loss": 0.0018, "step": 35473 }, { "epoch": 6.51, "learning_rate": 1.7754367383519511e-06, "loss": 0.0027, "step": 35474 }, { "epoch": 6.51, "learning_rate": 1.7750144458382445e-06, "loss": 0.0015, "step": 35475 }, { "epoch": 6.51, "learning_rate": 1.7745921986608483e-06, "loss": 0.0026, "step": 35476 }, { "epoch": 6.51, "learning_rate": 1.7741699968220939e-06, "loss": 0.0018, "step": 35477 }, { "epoch": 6.51, "learning_rate": 1.7737478403243058e-06, "loss": 0.0024, "step": 35478 }, { "epoch": 6.51, "learning_rate": 1.7733257291698103e-06, "loss": 0.0017, "step": 35479 }, { "epoch": 6.51, "learning_rate": 1.7729036633609375e-06, "loss": 0.002, "step": 35480 }, { "epoch": 6.51, "learning_rate": 1.7724816429000168e-06, "loss": 0.0017, "step": 35481 }, { "epoch": 6.51, "learning_rate": 1.7720596677893664e-06, "loss": 0.0018, "step": 35482 }, { "epoch": 6.51, "learning_rate": 1.7716377380313133e-06, "loss": 0.0028, "step": 35483 }, { "epoch": 6.51, "learning_rate": 1.7712158536281887e-06, "loss": 0.0022, "step": 35484 }, { "epoch": 6.51, "learning_rate": 1.7707940145823154e-06, "loss": 0.0025, "step": 35485 }, { "epoch": 6.51, "learning_rate": 1.7703722208960173e-06, "loss": 0.0017, "step": 35486 }, { "epoch": 6.51, "learning_rate": 1.769950472571621e-06, "loss": 0.0015, "step": 35487 }, { "epoch": 6.51, "learning_rate": 1.7695287696114505e-06, "loss": 0.0024, "step": 35488 }, { "epoch": 6.51, "learning_rate": 1.7691071120178306e-06, "loss": 0.0022, "step": 35489 }, { "epoch": 6.51, "learning_rate": 1.7686854997930847e-06, "loss": 0.0017, "step": 35490 }, { "epoch": 6.51, "learning_rate": 1.7682639329395369e-06, "loss": 0.0019, "step": 35491 }, { "epoch": 6.51, "learning_rate": 1.7678424114595105e-06, "loss": 0.0016, "step": 35492 }, { "epoch": 6.51, "learning_rate": 1.7674209353553306e-06, "loss": 0.0015, "step": 35493 }, { "epoch": 6.51, "learning_rate": 1.7669995046293197e-06, "loss": 0.0019, "step": 35494 }, { "epoch": 6.51, "learning_rate": 1.7665781192837993e-06, "loss": 0.0031, "step": 35495 }, { "epoch": 6.51, "learning_rate": 1.7661567793210943e-06, "loss": 0.0016, "step": 35496 }, { "epoch": 6.51, "learning_rate": 1.7657354847435226e-06, "loss": 0.0017, "step": 35497 }, { "epoch": 6.51, "learning_rate": 1.7653142355534136e-06, "loss": 0.0023, "step": 35498 }, { "epoch": 6.51, "learning_rate": 1.7648930317530867e-06, "loss": 0.0028, "step": 35499 }, { "epoch": 6.51, "learning_rate": 1.7644718733448607e-06, "loss": 0.0024, "step": 35500 }, { "epoch": 6.51, "learning_rate": 1.7640507603310553e-06, "loss": 0.0013, "step": 35501 }, { "epoch": 6.51, "learning_rate": 1.7636296927139984e-06, "loss": 0.0019, "step": 35502 }, { "epoch": 6.51, "learning_rate": 1.7632086704960073e-06, "loss": 0.0013, "step": 35503 }, { "epoch": 6.51, "learning_rate": 1.762787693679403e-06, "loss": 0.0016, "step": 35504 }, { "epoch": 6.51, "learning_rate": 1.7623667622665053e-06, "loss": 0.0013, "step": 35505 }, { "epoch": 6.51, "learning_rate": 1.7619458762596353e-06, "loss": 0.002, "step": 35506 }, { "epoch": 6.51, "learning_rate": 1.7615250356611125e-06, "loss": 0.0023, "step": 35507 }, { "epoch": 6.51, "learning_rate": 1.761104240473257e-06, "loss": 0.002, "step": 35508 }, { "epoch": 6.51, "learning_rate": 1.7606834906983884e-06, "loss": 0.0012, "step": 35509 }, { "epoch": 6.51, "learning_rate": 1.7602627863388255e-06, "loss": 0.003, "step": 35510 }, { "epoch": 6.51, "learning_rate": 1.7598421273968857e-06, "loss": 0.002, "step": 35511 }, { "epoch": 6.51, "learning_rate": 1.7594215138748904e-06, "loss": 0.0021, "step": 35512 }, { "epoch": 6.51, "learning_rate": 1.7590009457751566e-06, "loss": 0.0016, "step": 35513 }, { "epoch": 6.51, "learning_rate": 1.7585804231000003e-06, "loss": 0.0025, "step": 35514 }, { "epoch": 6.51, "learning_rate": 1.758159945851744e-06, "loss": 0.001, "step": 35515 }, { "epoch": 6.51, "learning_rate": 1.7577395140327026e-06, "loss": 0.003, "step": 35516 }, { "epoch": 6.51, "learning_rate": 1.7573191276451972e-06, "loss": 0.002, "step": 35517 }, { "epoch": 6.51, "learning_rate": 1.7568987866915398e-06, "loss": 0.0013, "step": 35518 }, { "epoch": 6.51, "learning_rate": 1.7564784911740463e-06, "loss": 0.0027, "step": 35519 }, { "epoch": 6.51, "learning_rate": 1.7560582410950388e-06, "loss": 0.0024, "step": 35520 }, { "epoch": 6.51, "learning_rate": 1.7556380364568314e-06, "loss": 0.0019, "step": 35521 }, { "epoch": 6.51, "learning_rate": 1.7552178772617411e-06, "loss": 0.0016, "step": 35522 }, { "epoch": 6.51, "learning_rate": 1.7547977635120828e-06, "loss": 0.0026, "step": 35523 }, { "epoch": 6.51, "learning_rate": 1.7543776952101721e-06, "loss": 0.0023, "step": 35524 }, { "epoch": 6.51, "learning_rate": 1.753957672358324e-06, "loss": 0.0027, "step": 35525 }, { "epoch": 6.51, "learning_rate": 1.7535376949588557e-06, "loss": 0.0012, "step": 35526 }, { "epoch": 6.52, "learning_rate": 1.7531177630140772e-06, "loss": 0.0034, "step": 35527 }, { "epoch": 6.52, "learning_rate": 1.7526978765263125e-06, "loss": 0.002, "step": 35528 }, { "epoch": 6.52, "learning_rate": 1.7522780354978675e-06, "loss": 0.0013, "step": 35529 }, { "epoch": 6.52, "learning_rate": 1.7518582399310592e-06, "loss": 0.003, "step": 35530 }, { "epoch": 6.52, "learning_rate": 1.7514384898282022e-06, "loss": 0.0019, "step": 35531 }, { "epoch": 6.52, "learning_rate": 1.7510187851916062e-06, "loss": 0.0021, "step": 35532 }, { "epoch": 6.52, "learning_rate": 1.750599126023591e-06, "loss": 0.0021, "step": 35533 }, { "epoch": 6.52, "learning_rate": 1.7501795123264664e-06, "loss": 0.0016, "step": 35534 }, { "epoch": 6.52, "learning_rate": 1.749759944102548e-06, "loss": 0.0025, "step": 35535 }, { "epoch": 6.52, "learning_rate": 1.749340421354141e-06, "loss": 0.0023, "step": 35536 }, { "epoch": 6.52, "learning_rate": 1.7489209440835654e-06, "loss": 0.0023, "step": 35537 }, { "epoch": 6.52, "learning_rate": 1.7485015122931304e-06, "loss": 0.0016, "step": 35538 }, { "epoch": 6.52, "learning_rate": 1.7480821259851488e-06, "loss": 0.0026, "step": 35539 }, { "epoch": 6.52, "learning_rate": 1.7476627851619322e-06, "loss": 0.0027, "step": 35540 }, { "epoch": 6.52, "learning_rate": 1.7472434898257918e-06, "loss": 0.0011, "step": 35541 }, { "epoch": 6.52, "learning_rate": 1.746824239979038e-06, "loss": 0.001, "step": 35542 }, { "epoch": 6.52, "learning_rate": 1.7464050356239826e-06, "loss": 0.0025, "step": 35543 }, { "epoch": 6.52, "learning_rate": 1.7459858767629367e-06, "loss": 0.0014, "step": 35544 }, { "epoch": 6.52, "learning_rate": 1.7455667633982066e-06, "loss": 0.0014, "step": 35545 }, { "epoch": 6.52, "learning_rate": 1.7451476955321111e-06, "loss": 0.0018, "step": 35546 }, { "epoch": 6.52, "learning_rate": 1.7447286731669522e-06, "loss": 0.003, "step": 35547 }, { "epoch": 6.52, "learning_rate": 1.7443096963050422e-06, "loss": 0.0022, "step": 35548 }, { "epoch": 6.52, "learning_rate": 1.743890764948688e-06, "loss": 0.0018, "step": 35549 }, { "epoch": 6.52, "learning_rate": 1.7434718791002047e-06, "loss": 0.0018, "step": 35550 }, { "epoch": 6.52, "learning_rate": 1.7430530387618972e-06, "loss": 0.0024, "step": 35551 }, { "epoch": 6.52, "learning_rate": 1.7426342439360778e-06, "loss": 0.0011, "step": 35552 }, { "epoch": 6.52, "learning_rate": 1.742215494625048e-06, "loss": 0.0016, "step": 35553 }, { "epoch": 6.52, "learning_rate": 1.7417967908311185e-06, "loss": 0.0028, "step": 35554 }, { "epoch": 6.52, "learning_rate": 1.7413781325566015e-06, "loss": 0.0012, "step": 35555 }, { "epoch": 6.52, "learning_rate": 1.740959519803801e-06, "loss": 0.0025, "step": 35556 }, { "epoch": 6.52, "learning_rate": 1.740540952575025e-06, "loss": 0.0019, "step": 35557 }, { "epoch": 6.52, "learning_rate": 1.7401224308725806e-06, "loss": 0.0029, "step": 35558 }, { "epoch": 6.52, "learning_rate": 1.739703954698776e-06, "loss": 0.0023, "step": 35559 }, { "epoch": 6.52, "learning_rate": 1.7392855240559158e-06, "loss": 0.0022, "step": 35560 }, { "epoch": 6.52, "learning_rate": 1.7388671389463073e-06, "loss": 0.0019, "step": 35561 }, { "epoch": 6.52, "learning_rate": 1.7384487993722566e-06, "loss": 0.0018, "step": 35562 }, { "epoch": 6.52, "learning_rate": 1.738030505336068e-06, "loss": 0.0013, "step": 35563 }, { "epoch": 6.52, "learning_rate": 1.7376122568400533e-06, "loss": 0.0015, "step": 35564 }, { "epoch": 6.52, "learning_rate": 1.7371940538865106e-06, "loss": 0.0013, "step": 35565 }, { "epoch": 6.52, "learning_rate": 1.7367758964777482e-06, "loss": 0.0022, "step": 35566 }, { "epoch": 6.52, "learning_rate": 1.7363577846160674e-06, "loss": 0.0021, "step": 35567 }, { "epoch": 6.52, "learning_rate": 1.7359397183037795e-06, "loss": 0.002, "step": 35568 }, { "epoch": 6.52, "learning_rate": 1.7355216975431855e-06, "loss": 0.0022, "step": 35569 }, { "epoch": 6.52, "learning_rate": 1.735103722336592e-06, "loss": 0.0012, "step": 35570 }, { "epoch": 6.52, "learning_rate": 1.734685792686297e-06, "loss": 0.0015, "step": 35571 }, { "epoch": 6.52, "learning_rate": 1.7342679085946056e-06, "loss": 0.0019, "step": 35572 }, { "epoch": 6.52, "learning_rate": 1.733850070063826e-06, "loss": 0.002, "step": 35573 }, { "epoch": 6.52, "learning_rate": 1.7334322770962575e-06, "loss": 0.0018, "step": 35574 }, { "epoch": 6.52, "learning_rate": 1.7330145296942047e-06, "loss": 0.0015, "step": 35575 }, { "epoch": 6.52, "learning_rate": 1.7325968278599703e-06, "loss": 0.0014, "step": 35576 }, { "epoch": 6.52, "learning_rate": 1.7321791715958547e-06, "loss": 0.003, "step": 35577 }, { "epoch": 6.52, "learning_rate": 1.7317615609041615e-06, "loss": 0.0011, "step": 35578 }, { "epoch": 6.52, "learning_rate": 1.7313439957871925e-06, "loss": 0.0019, "step": 35579 }, { "epoch": 6.52, "learning_rate": 1.7309264762472455e-06, "loss": 0.002, "step": 35580 }, { "epoch": 6.53, "learning_rate": 1.7305090022866288e-06, "loss": 0.002, "step": 35581 }, { "epoch": 6.53, "learning_rate": 1.730091573907643e-06, "loss": 0.0027, "step": 35582 }, { "epoch": 6.53, "learning_rate": 1.7296741911125836e-06, "loss": 0.0026, "step": 35583 }, { "epoch": 6.53, "learning_rate": 1.7292568539037534e-06, "loss": 0.0014, "step": 35584 }, { "epoch": 6.53, "learning_rate": 1.728839562283451e-06, "loss": 0.0013, "step": 35585 }, { "epoch": 6.53, "learning_rate": 1.7284223162539815e-06, "loss": 0.0019, "step": 35586 }, { "epoch": 6.53, "learning_rate": 1.7280051158176414e-06, "loss": 0.0023, "step": 35587 }, { "epoch": 6.53, "learning_rate": 1.7275879609767344e-06, "loss": 0.0025, "step": 35588 }, { "epoch": 6.53, "learning_rate": 1.7271708517335496e-06, "loss": 0.0013, "step": 35589 }, { "epoch": 6.53, "learning_rate": 1.7267537880903962e-06, "loss": 0.0023, "step": 35590 }, { "epoch": 6.53, "learning_rate": 1.7263367700495703e-06, "loss": 0.0012, "step": 35591 }, { "epoch": 6.53, "learning_rate": 1.7259197976133702e-06, "loss": 0.0019, "step": 35592 }, { "epoch": 6.53, "learning_rate": 1.7255028707840925e-06, "loss": 0.002, "step": 35593 }, { "epoch": 6.53, "learning_rate": 1.725085989564038e-06, "loss": 0.0017, "step": 35594 }, { "epoch": 6.53, "learning_rate": 1.7246691539555027e-06, "loss": 0.0024, "step": 35595 }, { "epoch": 6.53, "learning_rate": 1.7242523639607855e-06, "loss": 0.0015, "step": 35596 }, { "epoch": 6.53, "learning_rate": 1.7238356195821838e-06, "loss": 0.0017, "step": 35597 }, { "epoch": 6.53, "learning_rate": 1.72341892082199e-06, "loss": 0.0015, "step": 35598 }, { "epoch": 6.53, "learning_rate": 1.723002267682511e-06, "loss": 0.0017, "step": 35599 }, { "epoch": 6.53, "learning_rate": 1.722585660166034e-06, "loss": 0.0012, "step": 35600 }, { "epoch": 6.53, "learning_rate": 1.7221690982748596e-06, "loss": 0.004, "step": 35601 }, { "epoch": 6.53, "learning_rate": 1.72175258201128e-06, "loss": 0.0017, "step": 35602 }, { "epoch": 6.53, "learning_rate": 1.7213361113775973e-06, "loss": 0.0011, "step": 35603 }, { "epoch": 6.53, "learning_rate": 1.7209196863761035e-06, "loss": 0.003, "step": 35604 }, { "epoch": 6.53, "learning_rate": 1.720503307009095e-06, "loss": 0.0011, "step": 35605 }, { "epoch": 6.53, "learning_rate": 1.7200869732788673e-06, "loss": 0.0017, "step": 35606 }, { "epoch": 6.53, "learning_rate": 1.7196706851877098e-06, "loss": 0.0024, "step": 35607 }, { "epoch": 6.53, "learning_rate": 1.7192544427379242e-06, "loss": 0.0029, "step": 35608 }, { "epoch": 6.53, "learning_rate": 1.7188382459318033e-06, "loss": 0.0018, "step": 35609 }, { "epoch": 6.53, "learning_rate": 1.718422094771638e-06, "loss": 0.0012, "step": 35610 }, { "epoch": 6.53, "learning_rate": 1.7180059892597246e-06, "loss": 0.0019, "step": 35611 }, { "epoch": 6.53, "learning_rate": 1.7175899293983567e-06, "loss": 0.0018, "step": 35612 }, { "epoch": 6.53, "learning_rate": 1.717173915189827e-06, "loss": 0.0013, "step": 35613 }, { "epoch": 6.53, "learning_rate": 1.7167579466364292e-06, "loss": 0.0024, "step": 35614 }, { "epoch": 6.53, "learning_rate": 1.7163420237404549e-06, "loss": 0.0022, "step": 35615 }, { "epoch": 6.53, "learning_rate": 1.7159261465041954e-06, "loss": 0.0024, "step": 35616 }, { "epoch": 6.53, "learning_rate": 1.7155103149299501e-06, "loss": 0.0008, "step": 35617 }, { "epoch": 6.53, "learning_rate": 1.7150945290200027e-06, "loss": 0.0026, "step": 35618 }, { "epoch": 6.53, "learning_rate": 1.7146787887766492e-06, "loss": 0.0027, "step": 35619 }, { "epoch": 6.53, "learning_rate": 1.7142630942021777e-06, "loss": 0.0023, "step": 35620 }, { "epoch": 6.53, "learning_rate": 1.713847445298884e-06, "loss": 0.0014, "step": 35621 }, { "epoch": 6.53, "learning_rate": 1.7134318420690576e-06, "loss": 0.0018, "step": 35622 }, { "epoch": 6.53, "learning_rate": 1.71301628451499e-06, "loss": 0.0012, "step": 35623 }, { "epoch": 6.53, "learning_rate": 1.7126007726389716e-06, "loss": 0.0022, "step": 35624 }, { "epoch": 6.53, "learning_rate": 1.712185306443288e-06, "loss": 0.0016, "step": 35625 }, { "epoch": 6.53, "learning_rate": 1.7117698859302357e-06, "loss": 0.0019, "step": 35626 }, { "epoch": 6.53, "learning_rate": 1.7113545111021012e-06, "loss": 0.0015, "step": 35627 }, { "epoch": 6.53, "learning_rate": 1.7109391819611754e-06, "loss": 0.0019, "step": 35628 }, { "epoch": 6.53, "learning_rate": 1.7105238985097472e-06, "loss": 0.0018, "step": 35629 }, { "epoch": 6.53, "learning_rate": 1.7101086607501049e-06, "loss": 0.0027, "step": 35630 }, { "epoch": 6.53, "learning_rate": 1.709693468684539e-06, "loss": 0.0015, "step": 35631 }, { "epoch": 6.53, "learning_rate": 1.7092783223153353e-06, "loss": 0.0025, "step": 35632 }, { "epoch": 6.53, "learning_rate": 1.708863221644782e-06, "loss": 0.002, "step": 35633 }, { "epoch": 6.53, "learning_rate": 1.7084481666751717e-06, "loss": 0.0026, "step": 35634 }, { "epoch": 6.53, "learning_rate": 1.7080331574087915e-06, "loss": 0.0014, "step": 35635 }, { "epoch": 6.54, "learning_rate": 1.7076181938479242e-06, "loss": 0.0035, "step": 35636 }, { "epoch": 6.54, "learning_rate": 1.7072032759948597e-06, "loss": 0.0022, "step": 35637 }, { "epoch": 6.54, "learning_rate": 1.7067884038518834e-06, "loss": 0.0022, "step": 35638 }, { "epoch": 6.54, "learning_rate": 1.7063735774212852e-06, "loss": 0.0018, "step": 35639 }, { "epoch": 6.54, "learning_rate": 1.7059587967053503e-06, "loss": 0.0018, "step": 35640 }, { "epoch": 6.54, "learning_rate": 1.7055440617063646e-06, "loss": 0.0013, "step": 35641 }, { "epoch": 6.54, "learning_rate": 1.705129372426615e-06, "loss": 0.0015, "step": 35642 }, { "epoch": 6.54, "learning_rate": 1.7047147288683862e-06, "loss": 0.0023, "step": 35643 }, { "epoch": 6.54, "learning_rate": 1.7043001310339646e-06, "loss": 0.0018, "step": 35644 }, { "epoch": 6.54, "learning_rate": 1.7038855789256347e-06, "loss": 0.0043, "step": 35645 }, { "epoch": 6.54, "learning_rate": 1.7034710725456815e-06, "loss": 0.0035, "step": 35646 }, { "epoch": 6.54, "learning_rate": 1.7030566118963899e-06, "loss": 0.0017, "step": 35647 }, { "epoch": 6.54, "learning_rate": 1.7026421969800455e-06, "loss": 0.0024, "step": 35648 }, { "epoch": 6.54, "learning_rate": 1.7022278277989312e-06, "loss": 0.0023, "step": 35649 }, { "epoch": 6.54, "learning_rate": 1.7018135043553318e-06, "loss": 0.0023, "step": 35650 }, { "epoch": 6.54, "learning_rate": 1.7013992266515279e-06, "loss": 0.0023, "step": 35651 }, { "epoch": 6.54, "learning_rate": 1.7009849946898093e-06, "loss": 0.0011, "step": 35652 }, { "epoch": 6.54, "learning_rate": 1.700570808472457e-06, "loss": 0.003, "step": 35653 }, { "epoch": 6.54, "learning_rate": 1.7001566680017512e-06, "loss": 0.0013, "step": 35654 }, { "epoch": 6.54, "learning_rate": 1.6997425732799734e-06, "loss": 0.0021, "step": 35655 }, { "epoch": 6.54, "learning_rate": 1.6993285243094115e-06, "loss": 0.002, "step": 35656 }, { "epoch": 6.54, "learning_rate": 1.6989145210923463e-06, "loss": 0.0022, "step": 35657 }, { "epoch": 6.54, "learning_rate": 1.698500563631057e-06, "loss": 0.0016, "step": 35658 }, { "epoch": 6.54, "learning_rate": 1.6980866519278304e-06, "loss": 0.003, "step": 35659 }, { "epoch": 6.54, "learning_rate": 1.6976727859849406e-06, "loss": 0.0027, "step": 35660 }, { "epoch": 6.54, "learning_rate": 1.6972589658046745e-06, "loss": 0.003, "step": 35661 }, { "epoch": 6.54, "learning_rate": 1.6968451913893113e-06, "loss": 0.0014, "step": 35662 }, { "epoch": 6.54, "learning_rate": 1.6964314627411316e-06, "loss": 0.0019, "step": 35663 }, { "epoch": 6.54, "learning_rate": 1.6960177798624167e-06, "loss": 0.0023, "step": 35664 }, { "epoch": 6.54, "learning_rate": 1.695604142755447e-06, "loss": 0.0017, "step": 35665 }, { "epoch": 6.54, "learning_rate": 1.695190551422501e-06, "loss": 0.0028, "step": 35666 }, { "epoch": 6.54, "learning_rate": 1.6947770058658597e-06, "loss": 0.0019, "step": 35667 }, { "epoch": 6.54, "learning_rate": 1.6943635060878017e-06, "loss": 0.0028, "step": 35668 }, { "epoch": 6.54, "learning_rate": 1.6939500520906038e-06, "loss": 0.0023, "step": 35669 }, { "epoch": 6.54, "learning_rate": 1.693536643876551e-06, "loss": 0.0013, "step": 35670 }, { "epoch": 6.54, "learning_rate": 1.6931232814479214e-06, "loss": 0.0019, "step": 35671 }, { "epoch": 6.54, "learning_rate": 1.6927099648069878e-06, "loss": 0.0012, "step": 35672 }, { "epoch": 6.54, "learning_rate": 1.6922966939560292e-06, "loss": 0.0025, "step": 35673 }, { "epoch": 6.54, "learning_rate": 1.6918834688973285e-06, "loss": 0.0024, "step": 35674 }, { "epoch": 6.54, "learning_rate": 1.6914702896331602e-06, "loss": 0.0032, "step": 35675 }, { "epoch": 6.54, "learning_rate": 1.691057156165803e-06, "loss": 0.0015, "step": 35676 }, { "epoch": 6.54, "learning_rate": 1.6906440684975333e-06, "loss": 0.0021, "step": 35677 }, { "epoch": 6.54, "learning_rate": 1.6902310266306277e-06, "loss": 0.003, "step": 35678 }, { "epoch": 6.54, "learning_rate": 1.689818030567364e-06, "loss": 0.0017, "step": 35679 }, { "epoch": 6.54, "learning_rate": 1.6894050803100182e-06, "loss": 0.0016, "step": 35680 }, { "epoch": 6.54, "learning_rate": 1.6889921758608651e-06, "loss": 0.0031, "step": 35681 }, { "epoch": 6.54, "learning_rate": 1.6885793172221832e-06, "loss": 0.0016, "step": 35682 }, { "epoch": 6.54, "learning_rate": 1.6881665043962459e-06, "loss": 0.0018, "step": 35683 }, { "epoch": 6.54, "learning_rate": 1.6877537373853303e-06, "loss": 0.0025, "step": 35684 }, { "epoch": 6.54, "learning_rate": 1.6873410161917103e-06, "loss": 0.0017, "step": 35685 }, { "epoch": 6.54, "learning_rate": 1.6869283408176585e-06, "loss": 0.0011, "step": 35686 }, { "epoch": 6.54, "learning_rate": 1.686515711265455e-06, "loss": 0.001, "step": 35687 }, { "epoch": 6.54, "learning_rate": 1.686103127537373e-06, "loss": 0.002, "step": 35688 }, { "epoch": 6.54, "learning_rate": 1.6856905896356869e-06, "loss": 0.0026, "step": 35689 }, { "epoch": 6.55, "learning_rate": 1.6852780975626658e-06, "loss": 0.0014, "step": 35690 }, { "epoch": 6.55, "learning_rate": 1.6848656513205852e-06, "loss": 0.0024, "step": 35691 }, { "epoch": 6.55, "learning_rate": 1.6844532509117217e-06, "loss": 0.0016, "step": 35692 }, { "epoch": 6.55, "learning_rate": 1.6840408963383469e-06, "loss": 0.0021, "step": 35693 }, { "epoch": 6.55, "learning_rate": 1.6836285876027337e-06, "loss": 0.002, "step": 35694 }, { "epoch": 6.55, "learning_rate": 1.6832163247071543e-06, "loss": 0.0015, "step": 35695 }, { "epoch": 6.55, "learning_rate": 1.6828041076538815e-06, "loss": 0.0016, "step": 35696 }, { "epoch": 6.55, "learning_rate": 1.682391936445188e-06, "loss": 0.0012, "step": 35697 }, { "epoch": 6.55, "learning_rate": 1.681979811083344e-06, "loss": 0.0018, "step": 35698 }, { "epoch": 6.55, "learning_rate": 1.6815677315706224e-06, "loss": 0.0027, "step": 35699 }, { "epoch": 6.55, "learning_rate": 1.6811556979092947e-06, "loss": 0.0042, "step": 35700 }, { "epoch": 6.55, "learning_rate": 1.6807437101016321e-06, "loss": 0.0024, "step": 35701 }, { "epoch": 6.55, "learning_rate": 1.6803317681499044e-06, "loss": 0.002, "step": 35702 }, { "epoch": 6.55, "learning_rate": 1.6799198720563824e-06, "loss": 0.0012, "step": 35703 }, { "epoch": 6.55, "learning_rate": 1.679508021823335e-06, "loss": 0.0014, "step": 35704 }, { "epoch": 6.55, "learning_rate": 1.6790962174530367e-06, "loss": 0.0029, "step": 35705 }, { "epoch": 6.55, "learning_rate": 1.678684458947758e-06, "loss": 0.0022, "step": 35706 }, { "epoch": 6.55, "learning_rate": 1.6782727463097626e-06, "loss": 0.0016, "step": 35707 }, { "epoch": 6.55, "learning_rate": 1.6778610795413208e-06, "loss": 0.0024, "step": 35708 }, { "epoch": 6.55, "learning_rate": 1.6774494586447054e-06, "loss": 0.0028, "step": 35709 }, { "epoch": 6.55, "learning_rate": 1.6770378836221835e-06, "loss": 0.0016, "step": 35710 }, { "epoch": 6.55, "learning_rate": 1.6766263544760241e-06, "loss": 0.0023, "step": 35711 }, { "epoch": 6.55, "learning_rate": 1.6762148712084959e-06, "loss": 0.0023, "step": 35712 }, { "epoch": 6.55, "learning_rate": 1.6758034338218655e-06, "loss": 0.0015, "step": 35713 }, { "epoch": 6.55, "learning_rate": 1.6753920423184022e-06, "loss": 0.002, "step": 35714 }, { "epoch": 6.55, "learning_rate": 1.6749806967003723e-06, "loss": 0.0022, "step": 35715 }, { "epoch": 6.55, "learning_rate": 1.674569396970044e-06, "loss": 0.0019, "step": 35716 }, { "epoch": 6.55, "learning_rate": 1.6741581431296817e-06, "loss": 0.001, "step": 35717 }, { "epoch": 6.55, "learning_rate": 1.6737469351815605e-06, "loss": 0.0025, "step": 35718 }, { "epoch": 6.55, "learning_rate": 1.6733357731279375e-06, "loss": 0.0021, "step": 35719 }, { "epoch": 6.55, "learning_rate": 1.6729246569710832e-06, "loss": 0.0015, "step": 35720 }, { "epoch": 6.55, "learning_rate": 1.6725135867132636e-06, "loss": 0.0013, "step": 35721 }, { "epoch": 6.55, "learning_rate": 1.6721025623567421e-06, "loss": 0.0029, "step": 35722 }, { "epoch": 6.55, "learning_rate": 1.6716915839037885e-06, "loss": 0.0023, "step": 35723 }, { "epoch": 6.55, "learning_rate": 1.6712806513566682e-06, "loss": 0.0027, "step": 35724 }, { "epoch": 6.55, "learning_rate": 1.6708697647176419e-06, "loss": 0.001, "step": 35725 }, { "epoch": 6.55, "learning_rate": 1.6704589239889746e-06, "loss": 0.0034, "step": 35726 }, { "epoch": 6.55, "learning_rate": 1.6700481291729342e-06, "loss": 0.0018, "step": 35727 }, { "epoch": 6.55, "learning_rate": 1.6696373802717846e-06, "loss": 0.0035, "step": 35728 }, { "epoch": 6.55, "learning_rate": 1.6692266772877885e-06, "loss": 0.002, "step": 35729 }, { "epoch": 6.55, "learning_rate": 1.6688160202232096e-06, "loss": 0.0019, "step": 35730 }, { "epoch": 6.55, "learning_rate": 1.6684054090803125e-06, "loss": 0.0014, "step": 35731 }, { "epoch": 6.55, "learning_rate": 1.6679948438613603e-06, "loss": 0.0018, "step": 35732 }, { "epoch": 6.55, "learning_rate": 1.6675843245686151e-06, "loss": 0.0027, "step": 35733 }, { "epoch": 6.55, "learning_rate": 1.6671738512043412e-06, "loss": 0.0016, "step": 35734 }, { "epoch": 6.55, "learning_rate": 1.6667634237707964e-06, "loss": 0.0024, "step": 35735 }, { "epoch": 6.55, "learning_rate": 1.6663530422702534e-06, "loss": 0.0015, "step": 35736 }, { "epoch": 6.55, "learning_rate": 1.6659427067049638e-06, "loss": 0.0012, "step": 35737 }, { "epoch": 6.55, "learning_rate": 1.6655324170771937e-06, "loss": 0.0022, "step": 35738 }, { "epoch": 6.55, "learning_rate": 1.665122173389202e-06, "loss": 0.0021, "step": 35739 }, { "epoch": 6.55, "learning_rate": 1.664711975643255e-06, "loss": 0.0016, "step": 35740 }, { "epoch": 6.55, "learning_rate": 1.6643018238416099e-06, "loss": 0.0023, "step": 35741 }, { "epoch": 6.55, "learning_rate": 1.663891717986531e-06, "loss": 0.0017, "step": 35742 }, { "epoch": 6.55, "learning_rate": 1.6634816580802737e-06, "loss": 0.0014, "step": 35743 }, { "epoch": 6.55, "learning_rate": 1.6630716441250994e-06, "loss": 0.0013, "step": 35744 }, { "epoch": 6.56, "learning_rate": 1.6626616761232718e-06, "loss": 0.0025, "step": 35745 }, { "epoch": 6.56, "learning_rate": 1.6622517540770477e-06, "loss": 0.0011, "step": 35746 }, { "epoch": 6.56, "learning_rate": 1.6618418779886868e-06, "loss": 0.0021, "step": 35747 }, { "epoch": 6.56, "learning_rate": 1.6614320478604495e-06, "loss": 0.0011, "step": 35748 }, { "epoch": 6.56, "learning_rate": 1.661022263694594e-06, "loss": 0.0017, "step": 35749 }, { "epoch": 6.56, "learning_rate": 1.6606125254933792e-06, "loss": 0.0016, "step": 35750 }, { "epoch": 6.56, "learning_rate": 1.6602028332590636e-06, "loss": 0.0022, "step": 35751 }, { "epoch": 6.56, "learning_rate": 1.6597931869939055e-06, "loss": 0.0022, "step": 35752 }, { "epoch": 6.56, "learning_rate": 1.6593835867001628e-06, "loss": 0.0019, "step": 35753 }, { "epoch": 6.56, "learning_rate": 1.6589740323800929e-06, "loss": 0.0016, "step": 35754 }, { "epoch": 6.56, "learning_rate": 1.6585645240359527e-06, "loss": 0.002, "step": 35755 }, { "epoch": 6.56, "learning_rate": 1.6581550616700016e-06, "loss": 0.0028, "step": 35756 }, { "epoch": 6.56, "learning_rate": 1.657745645284492e-06, "loss": 0.0019, "step": 35757 }, { "epoch": 6.56, "learning_rate": 1.6573362748816868e-06, "loss": 0.0023, "step": 35758 }, { "epoch": 6.56, "learning_rate": 1.6569269504638396e-06, "loss": 0.0032, "step": 35759 }, { "epoch": 6.56, "learning_rate": 1.6565176720332076e-06, "loss": 0.0037, "step": 35760 }, { "epoch": 6.56, "learning_rate": 1.6561084395920423e-06, "loss": 0.0017, "step": 35761 }, { "epoch": 6.56, "learning_rate": 1.6556992531426052e-06, "loss": 0.0019, "step": 35762 }, { "epoch": 6.56, "learning_rate": 1.6552901126871478e-06, "loss": 0.0015, "step": 35763 }, { "epoch": 6.56, "learning_rate": 1.6548810182279285e-06, "loss": 0.0023, "step": 35764 }, { "epoch": 6.56, "learning_rate": 1.6544719697671995e-06, "loss": 0.0022, "step": 35765 }, { "epoch": 6.56, "learning_rate": 1.6540629673072174e-06, "loss": 0.003, "step": 35766 }, { "epoch": 6.56, "learning_rate": 1.6536540108502353e-06, "loss": 0.001, "step": 35767 }, { "epoch": 6.56, "learning_rate": 1.6532451003985072e-06, "loss": 0.0024, "step": 35768 }, { "epoch": 6.56, "learning_rate": 1.6528362359542882e-06, "loss": 0.0024, "step": 35769 }, { "epoch": 6.56, "learning_rate": 1.6524274175198284e-06, "loss": 0.0017, "step": 35770 }, { "epoch": 6.56, "learning_rate": 1.6520186450973897e-06, "loss": 0.0023, "step": 35771 }, { "epoch": 6.56, "learning_rate": 1.6516099186892176e-06, "loss": 0.002, "step": 35772 }, { "epoch": 6.56, "learning_rate": 1.6512012382975673e-06, "loss": 0.0021, "step": 35773 }, { "epoch": 6.56, "learning_rate": 1.6507926039246914e-06, "loss": 0.0017, "step": 35774 }, { "epoch": 6.56, "learning_rate": 1.6503840155728391e-06, "loss": 0.0007, "step": 35775 }, { "epoch": 6.56, "learning_rate": 1.6499754732442686e-06, "loss": 0.0028, "step": 35776 }, { "epoch": 6.56, "learning_rate": 1.6495669769412282e-06, "loss": 0.0014, "step": 35777 }, { "epoch": 6.56, "learning_rate": 1.6491585266659739e-06, "loss": 0.0016, "step": 35778 }, { "epoch": 6.56, "learning_rate": 1.648750122420748e-06, "loss": 0.0016, "step": 35779 }, { "epoch": 6.56, "learning_rate": 1.6483417642078092e-06, "loss": 0.0018, "step": 35780 }, { "epoch": 6.56, "learning_rate": 1.6479334520294066e-06, "loss": 0.0017, "step": 35781 }, { "epoch": 6.56, "learning_rate": 1.6475251858877906e-06, "loss": 0.0022, "step": 35782 }, { "epoch": 6.56, "learning_rate": 1.6471169657852104e-06, "loss": 0.0018, "step": 35783 }, { "epoch": 6.56, "learning_rate": 1.6467087917239189e-06, "loss": 0.0023, "step": 35784 }, { "epoch": 6.56, "learning_rate": 1.6463006637061629e-06, "loss": 0.0026, "step": 35785 }, { "epoch": 6.56, "learning_rate": 1.6458925817341943e-06, "loss": 0.0026, "step": 35786 }, { "epoch": 6.56, "learning_rate": 1.6454845458102608e-06, "loss": 0.0021, "step": 35787 }, { "epoch": 6.56, "learning_rate": 1.64507655593661e-06, "loss": 0.002, "step": 35788 }, { "epoch": 6.56, "learning_rate": 1.6446686121154976e-06, "loss": 0.0014, "step": 35789 }, { "epoch": 6.56, "learning_rate": 1.644260714349165e-06, "loss": 0.0018, "step": 35790 }, { "epoch": 6.56, "learning_rate": 1.6438528626398643e-06, "loss": 0.0023, "step": 35791 }, { "epoch": 6.56, "learning_rate": 1.6434450569898387e-06, "loss": 0.0016, "step": 35792 }, { "epoch": 6.56, "learning_rate": 1.643037297401342e-06, "loss": 0.0018, "step": 35793 }, { "epoch": 6.56, "learning_rate": 1.6426295838766205e-06, "loss": 0.0016, "step": 35794 }, { "epoch": 6.56, "learning_rate": 1.64222191641792e-06, "loss": 0.0022, "step": 35795 }, { "epoch": 6.56, "learning_rate": 1.6418142950274906e-06, "loss": 0.0039, "step": 35796 }, { "epoch": 6.56, "learning_rate": 1.641406719707571e-06, "loss": 0.0015, "step": 35797 }, { "epoch": 6.56, "learning_rate": 1.6409991904604173e-06, "loss": 0.0018, "step": 35798 }, { "epoch": 6.57, "learning_rate": 1.6405917072882715e-06, "loss": 0.0012, "step": 35799 }, { "epoch": 6.57, "learning_rate": 1.640184270193379e-06, "loss": 0.0015, "step": 35800 }, { "epoch": 6.57, "learning_rate": 1.6397768791779877e-06, "loss": 0.0021, "step": 35801 }, { "epoch": 6.57, "learning_rate": 1.6393695342443416e-06, "loss": 0.0019, "step": 35802 }, { "epoch": 6.57, "learning_rate": 1.6389622353946867e-06, "loss": 0.0009, "step": 35803 }, { "epoch": 6.57, "learning_rate": 1.6385549826312674e-06, "loss": 0.0011, "step": 35804 }, { "epoch": 6.57, "learning_rate": 1.638147775956327e-06, "loss": 0.0011, "step": 35805 }, { "epoch": 6.57, "learning_rate": 1.6377406153721143e-06, "loss": 0.0017, "step": 35806 }, { "epoch": 6.57, "learning_rate": 1.6373335008808744e-06, "loss": 0.0024, "step": 35807 }, { "epoch": 6.57, "learning_rate": 1.6369264324848443e-06, "loss": 0.0026, "step": 35808 }, { "epoch": 6.57, "learning_rate": 1.6365194101862725e-06, "loss": 0.0023, "step": 35809 }, { "epoch": 6.57, "learning_rate": 1.636112433987399e-06, "loss": 0.0014, "step": 35810 }, { "epoch": 6.57, "learning_rate": 1.6357055038904724e-06, "loss": 0.0012, "step": 35811 }, { "epoch": 6.57, "learning_rate": 1.6352986198977327e-06, "loss": 0.0015, "step": 35812 }, { "epoch": 6.57, "learning_rate": 1.6348917820114253e-06, "loss": 0.0015, "step": 35813 }, { "epoch": 6.57, "learning_rate": 1.6344849902337855e-06, "loss": 0.0016, "step": 35814 }, { "epoch": 6.57, "learning_rate": 1.6340782445670634e-06, "loss": 0.0027, "step": 35815 }, { "epoch": 6.57, "learning_rate": 1.6336715450134976e-06, "loss": 0.0016, "step": 35816 }, { "epoch": 6.57, "learning_rate": 1.63326489157533e-06, "loss": 0.0023, "step": 35817 }, { "epoch": 6.57, "learning_rate": 1.6328582842548025e-06, "loss": 0.0022, "step": 35818 }, { "epoch": 6.57, "learning_rate": 1.6324517230541558e-06, "loss": 0.0022, "step": 35819 }, { "epoch": 6.57, "learning_rate": 1.6320452079756322e-06, "loss": 0.0024, "step": 35820 }, { "epoch": 6.57, "learning_rate": 1.6316387390214716e-06, "loss": 0.0027, "step": 35821 }, { "epoch": 6.57, "learning_rate": 1.6312323161939136e-06, "loss": 0.0023, "step": 35822 }, { "epoch": 6.57, "learning_rate": 1.6308259394951963e-06, "loss": 0.0016, "step": 35823 }, { "epoch": 6.57, "learning_rate": 1.630419608927566e-06, "loss": 0.0022, "step": 35824 }, { "epoch": 6.57, "learning_rate": 1.6300133244932604e-06, "loss": 0.0026, "step": 35825 }, { "epoch": 6.57, "learning_rate": 1.6296070861945147e-06, "loss": 0.0016, "step": 35826 }, { "epoch": 6.57, "learning_rate": 1.6292008940335724e-06, "loss": 0.0013, "step": 35827 }, { "epoch": 6.57, "learning_rate": 1.6287947480126676e-06, "loss": 0.0017, "step": 35828 }, { "epoch": 6.57, "learning_rate": 1.6283886481340439e-06, "loss": 0.002, "step": 35829 }, { "epoch": 6.57, "learning_rate": 1.6279825943999394e-06, "loss": 0.0016, "step": 35830 }, { "epoch": 6.57, "learning_rate": 1.6275765868125927e-06, "loss": 0.0011, "step": 35831 }, { "epoch": 6.57, "learning_rate": 1.6271706253742347e-06, "loss": 0.0013, "step": 35832 }, { "epoch": 6.57, "learning_rate": 1.626764710087112e-06, "loss": 0.0016, "step": 35833 }, { "epoch": 6.57, "learning_rate": 1.626358840953457e-06, "loss": 0.0021, "step": 35834 }, { "epoch": 6.57, "learning_rate": 1.625953017975509e-06, "loss": 0.0017, "step": 35835 }, { "epoch": 6.57, "learning_rate": 1.625547241155504e-06, "loss": 0.0019, "step": 35836 }, { "epoch": 6.57, "learning_rate": 1.6251415104956792e-06, "loss": 0.0018, "step": 35837 }, { "epoch": 6.57, "learning_rate": 1.6247358259982704e-06, "loss": 0.0023, "step": 35838 }, { "epoch": 6.57, "learning_rate": 1.6243301876655138e-06, "loss": 0.0012, "step": 35839 }, { "epoch": 6.57, "learning_rate": 1.6239245954996452e-06, "loss": 0.0019, "step": 35840 }, { "epoch": 6.57, "learning_rate": 1.6235190495028985e-06, "loss": 0.0023, "step": 35841 }, { "epoch": 6.57, "learning_rate": 1.623113549677514e-06, "loss": 0.0017, "step": 35842 }, { "epoch": 6.57, "learning_rate": 1.6227080960257247e-06, "loss": 0.0027, "step": 35843 }, { "epoch": 6.57, "learning_rate": 1.622302688549764e-06, "loss": 0.002, "step": 35844 }, { "epoch": 6.57, "learning_rate": 1.6218973272518634e-06, "loss": 0.0023, "step": 35845 }, { "epoch": 6.57, "learning_rate": 1.6214920121342637e-06, "loss": 0.0017, "step": 35846 }, { "epoch": 6.57, "learning_rate": 1.6210867431991962e-06, "loss": 0.0017, "step": 35847 }, { "epoch": 6.57, "learning_rate": 1.6206815204488958e-06, "loss": 0.0012, "step": 35848 }, { "epoch": 6.57, "learning_rate": 1.6202763438855963e-06, "loss": 0.0017, "step": 35849 }, { "epoch": 6.57, "learning_rate": 1.6198712135115258e-06, "loss": 0.0032, "step": 35850 }, { "epoch": 6.57, "learning_rate": 1.6194661293289238e-06, "loss": 0.0019, "step": 35851 }, { "epoch": 6.57, "learning_rate": 1.6190610913400207e-06, "loss": 0.0013, "step": 35852 }, { "epoch": 6.57, "learning_rate": 1.6186560995470502e-06, "loss": 0.0019, "step": 35853 }, { "epoch": 6.58, "learning_rate": 1.6182511539522427e-06, "loss": 0.0023, "step": 35854 }, { "epoch": 6.58, "learning_rate": 1.617846254557831e-06, "loss": 0.0024, "step": 35855 }, { "epoch": 6.58, "learning_rate": 1.6174414013660477e-06, "loss": 0.0016, "step": 35856 }, { "epoch": 6.58, "learning_rate": 1.6170365943791234e-06, "loss": 0.0024, "step": 35857 }, { "epoch": 6.58, "learning_rate": 1.616631833599287e-06, "loss": 0.0023, "step": 35858 }, { "epoch": 6.58, "learning_rate": 1.6162271190287749e-06, "loss": 0.003, "step": 35859 }, { "epoch": 6.58, "learning_rate": 1.6158224506698184e-06, "loss": 0.0022, "step": 35860 }, { "epoch": 6.58, "learning_rate": 1.6154178285246424e-06, "loss": 0.0017, "step": 35861 }, { "epoch": 6.58, "learning_rate": 1.6150132525954798e-06, "loss": 0.0012, "step": 35862 }, { "epoch": 6.58, "learning_rate": 1.6146087228845574e-06, "loss": 0.0017, "step": 35863 }, { "epoch": 6.58, "learning_rate": 1.6142042393941115e-06, "loss": 0.0025, "step": 35864 }, { "epoch": 6.58, "learning_rate": 1.6137998021263678e-06, "loss": 0.0023, "step": 35865 }, { "epoch": 6.58, "learning_rate": 1.6133954110835558e-06, "loss": 0.0026, "step": 35866 }, { "epoch": 6.58, "learning_rate": 1.612991066267905e-06, "loss": 0.0025, "step": 35867 }, { "epoch": 6.58, "learning_rate": 1.6125867676816442e-06, "loss": 0.0015, "step": 35868 }, { "epoch": 6.58, "learning_rate": 1.6121825153270021e-06, "loss": 0.0016, "step": 35869 }, { "epoch": 6.58, "learning_rate": 1.6117783092062057e-06, "loss": 0.0014, "step": 35870 }, { "epoch": 6.58, "learning_rate": 1.611374149321484e-06, "loss": 0.0017, "step": 35871 }, { "epoch": 6.58, "learning_rate": 1.6109700356750647e-06, "loss": 0.0026, "step": 35872 }, { "epoch": 6.58, "learning_rate": 1.6105659682691742e-06, "loss": 0.0019, "step": 35873 }, { "epoch": 6.58, "learning_rate": 1.6101619471060415e-06, "loss": 0.0014, "step": 35874 }, { "epoch": 6.58, "learning_rate": 1.609757972187892e-06, "loss": 0.0016, "step": 35875 }, { "epoch": 6.58, "learning_rate": 1.6093540435169519e-06, "loss": 0.0015, "step": 35876 }, { "epoch": 6.58, "learning_rate": 1.6089501610954506e-06, "loss": 0.0016, "step": 35877 }, { "epoch": 6.58, "learning_rate": 1.608546324925615e-06, "loss": 0.0013, "step": 35878 }, { "epoch": 6.58, "learning_rate": 1.6081425350096658e-06, "loss": 0.0017, "step": 35879 }, { "epoch": 6.58, "learning_rate": 1.607738791349832e-06, "loss": 0.0015, "step": 35880 }, { "epoch": 6.58, "learning_rate": 1.6073350939483368e-06, "loss": 0.0017, "step": 35881 }, { "epoch": 6.58, "learning_rate": 1.606931442807409e-06, "loss": 0.0026, "step": 35882 }, { "epoch": 6.58, "learning_rate": 1.6065278379292714e-06, "loss": 0.0023, "step": 35883 }, { "epoch": 6.58, "learning_rate": 1.6061242793161492e-06, "loss": 0.0025, "step": 35884 }, { "epoch": 6.58, "learning_rate": 1.6057207669702668e-06, "loss": 0.0023, "step": 35885 }, { "epoch": 6.58, "learning_rate": 1.6053173008938482e-06, "loss": 0.0015, "step": 35886 }, { "epoch": 6.58, "learning_rate": 1.6049138810891173e-06, "loss": 0.0012, "step": 35887 }, { "epoch": 6.58, "learning_rate": 1.6045105075582967e-06, "loss": 0.0018, "step": 35888 }, { "epoch": 6.58, "learning_rate": 1.60410718030361e-06, "loss": 0.0014, "step": 35889 }, { "epoch": 6.58, "learning_rate": 1.6037038993272858e-06, "loss": 0.0014, "step": 35890 }, { "epoch": 6.58, "learning_rate": 1.6033006646315398e-06, "loss": 0.0015, "step": 35891 }, { "epoch": 6.58, "learning_rate": 1.6028974762185966e-06, "loss": 0.0015, "step": 35892 }, { "epoch": 6.58, "learning_rate": 1.6024943340906796e-06, "loss": 0.002, "step": 35893 }, { "epoch": 6.58, "learning_rate": 1.6020912382500086e-06, "loss": 0.0014, "step": 35894 }, { "epoch": 6.58, "learning_rate": 1.601688188698809e-06, "loss": 0.0024, "step": 35895 }, { "epoch": 6.58, "learning_rate": 1.601285185439304e-06, "loss": 0.0021, "step": 35896 }, { "epoch": 6.58, "learning_rate": 1.600882228473709e-06, "loss": 0.0019, "step": 35897 }, { "epoch": 6.58, "learning_rate": 1.6004793178042455e-06, "loss": 0.002, "step": 35898 }, { "epoch": 6.58, "learning_rate": 1.6000764534331404e-06, "loss": 0.0028, "step": 35899 }, { "epoch": 6.58, "learning_rate": 1.5996736353626096e-06, "loss": 0.002, "step": 35900 }, { "epoch": 6.58, "learning_rate": 1.599270863594874e-06, "loss": 0.002, "step": 35901 }, { "epoch": 6.58, "learning_rate": 1.5988681381321546e-06, "loss": 0.0029, "step": 35902 }, { "epoch": 6.58, "learning_rate": 1.5984654589766712e-06, "loss": 0.0022, "step": 35903 }, { "epoch": 6.58, "learning_rate": 1.598062826130643e-06, "loss": 0.0016, "step": 35904 }, { "epoch": 6.58, "learning_rate": 1.5976602395962892e-06, "loss": 0.0023, "step": 35905 }, { "epoch": 6.58, "learning_rate": 1.5972576993758293e-06, "loss": 0.0018, "step": 35906 }, { "epoch": 6.58, "learning_rate": 1.5968552054714813e-06, "loss": 0.0014, "step": 35907 }, { "epoch": 6.58, "learning_rate": 1.5964527578854649e-06, "loss": 0.0016, "step": 35908 }, { "epoch": 6.59, "learning_rate": 1.596050356619997e-06, "loss": 0.0015, "step": 35909 }, { "epoch": 6.59, "learning_rate": 1.5956480016772968e-06, "loss": 0.0037, "step": 35910 }, { "epoch": 6.59, "learning_rate": 1.5952456930595795e-06, "loss": 0.0009, "step": 35911 }, { "epoch": 6.59, "learning_rate": 1.5948434307690675e-06, "loss": 0.0026, "step": 35912 }, { "epoch": 6.59, "learning_rate": 1.5944412148079747e-06, "loss": 0.0027, "step": 35913 }, { "epoch": 6.59, "learning_rate": 1.5940390451785215e-06, "loss": 0.0013, "step": 35914 }, { "epoch": 6.59, "learning_rate": 1.5936369218829195e-06, "loss": 0.0025, "step": 35915 }, { "epoch": 6.59, "learning_rate": 1.5932348449233858e-06, "loss": 0.002, "step": 35916 }, { "epoch": 6.59, "learning_rate": 1.5928328143021399e-06, "loss": 0.0023, "step": 35917 }, { "epoch": 6.59, "learning_rate": 1.5924308300213976e-06, "loss": 0.0026, "step": 35918 }, { "epoch": 6.59, "learning_rate": 1.5920288920833726e-06, "loss": 0.0026, "step": 35919 }, { "epoch": 6.59, "learning_rate": 1.5916270004902822e-06, "loss": 0.0011, "step": 35920 }, { "epoch": 6.59, "learning_rate": 1.5912251552443402e-06, "loss": 0.0017, "step": 35921 }, { "epoch": 6.59, "learning_rate": 1.5908233563477626e-06, "loss": 0.0017, "step": 35922 }, { "epoch": 6.59, "learning_rate": 1.5904216038027642e-06, "loss": 0.0015, "step": 35923 }, { "epoch": 6.59, "learning_rate": 1.5900198976115577e-06, "loss": 0.0021, "step": 35924 }, { "epoch": 6.59, "learning_rate": 1.5896182377763592e-06, "loss": 0.0028, "step": 35925 }, { "epoch": 6.59, "learning_rate": 1.5892166242993812e-06, "loss": 0.0023, "step": 35926 }, { "epoch": 6.59, "learning_rate": 1.5888150571828397e-06, "loss": 0.002, "step": 35927 }, { "epoch": 6.59, "learning_rate": 1.5884135364289454e-06, "loss": 0.0016, "step": 35928 }, { "epoch": 6.59, "learning_rate": 1.588012062039911e-06, "loss": 0.001, "step": 35929 }, { "epoch": 6.59, "learning_rate": 1.5876106340179532e-06, "loss": 0.0023, "step": 35930 }, { "epoch": 6.59, "learning_rate": 1.5872092523652827e-06, "loss": 0.0016, "step": 35931 }, { "epoch": 6.59, "learning_rate": 1.5868079170841144e-06, "loss": 0.0016, "step": 35932 }, { "epoch": 6.59, "learning_rate": 1.5864066281766532e-06, "loss": 0.0022, "step": 35933 }, { "epoch": 6.59, "learning_rate": 1.5860053856451186e-06, "loss": 0.0011, "step": 35934 }, { "epoch": 6.59, "learning_rate": 1.5856041894917185e-06, "loss": 0.0016, "step": 35935 }, { "epoch": 6.59, "learning_rate": 1.5852030397186657e-06, "loss": 0.003, "step": 35936 }, { "epoch": 6.59, "learning_rate": 1.5848019363281707e-06, "loss": 0.0029, "step": 35937 }, { "epoch": 6.59, "learning_rate": 1.5844008793224453e-06, "loss": 0.0022, "step": 35938 }, { "epoch": 6.59, "learning_rate": 1.5839998687036984e-06, "loss": 0.0016, "step": 35939 }, { "epoch": 6.59, "learning_rate": 1.5835989044741417e-06, "loss": 0.0013, "step": 35940 }, { "epoch": 6.59, "learning_rate": 1.5831979866359858e-06, "loss": 0.001, "step": 35941 }, { "epoch": 6.59, "learning_rate": 1.5827971151914368e-06, "loss": 0.0017, "step": 35942 }, { "epoch": 6.59, "learning_rate": 1.5823962901427115e-06, "loss": 0.0015, "step": 35943 }, { "epoch": 6.59, "learning_rate": 1.5819955114920139e-06, "loss": 0.0019, "step": 35944 }, { "epoch": 6.59, "learning_rate": 1.581594779241553e-06, "loss": 0.0024, "step": 35945 }, { "epoch": 6.59, "learning_rate": 1.5811940933935398e-06, "loss": 0.0013, "step": 35946 }, { "epoch": 6.59, "learning_rate": 1.580793453950179e-06, "loss": 0.0025, "step": 35947 }, { "epoch": 6.59, "learning_rate": 1.5803928609136844e-06, "loss": 0.0021, "step": 35948 }, { "epoch": 6.59, "learning_rate": 1.5799923142862618e-06, "loss": 0.0011, "step": 35949 }, { "epoch": 6.59, "learning_rate": 1.5795918140701217e-06, "loss": 0.0014, "step": 35950 }, { "epoch": 6.59, "learning_rate": 1.5791913602674636e-06, "loss": 0.0019, "step": 35951 }, { "epoch": 6.59, "learning_rate": 1.5787909528805024e-06, "loss": 0.0018, "step": 35952 }, { "epoch": 6.59, "learning_rate": 1.5783905919114418e-06, "loss": 0.0021, "step": 35953 }, { "epoch": 6.59, "learning_rate": 1.577990277362491e-06, "loss": 0.0017, "step": 35954 }, { "epoch": 6.59, "learning_rate": 1.5775900092358542e-06, "loss": 0.0021, "step": 35955 }, { "epoch": 6.59, "learning_rate": 1.5771897875337394e-06, "loss": 0.0013, "step": 35956 }, { "epoch": 6.59, "learning_rate": 1.5767896122583515e-06, "loss": 0.0026, "step": 35957 }, { "epoch": 6.59, "learning_rate": 1.5763894834118953e-06, "loss": 0.0013, "step": 35958 }, { "epoch": 6.59, "learning_rate": 1.5759894009965793e-06, "loss": 0.0035, "step": 35959 }, { "epoch": 6.59, "learning_rate": 1.5755893650146038e-06, "loss": 0.0008, "step": 35960 }, { "epoch": 6.59, "learning_rate": 1.5751893754681825e-06, "loss": 0.0019, "step": 35961 }, { "epoch": 6.59, "learning_rate": 1.5747894323595114e-06, "loss": 0.0024, "step": 35962 }, { "epoch": 6.6, "learning_rate": 1.574389535690799e-06, "loss": 0.0011, "step": 35963 }, { "epoch": 6.6, "learning_rate": 1.5739896854642467e-06, "loss": 0.0023, "step": 35964 }, { "epoch": 6.6, "learning_rate": 1.5735898816820628e-06, "loss": 0.0017, "step": 35965 }, { "epoch": 6.6, "learning_rate": 1.5731901243464488e-06, "loss": 0.0025, "step": 35966 }, { "epoch": 6.6, "learning_rate": 1.5727904134596084e-06, "loss": 0.0016, "step": 35967 }, { "epoch": 6.6, "learning_rate": 1.5723907490237468e-06, "loss": 0.0014, "step": 35968 }, { "epoch": 6.6, "learning_rate": 1.5719911310410606e-06, "loss": 0.0014, "step": 35969 }, { "epoch": 6.6, "learning_rate": 1.5715915595137588e-06, "loss": 0.002, "step": 35970 }, { "epoch": 6.6, "learning_rate": 1.5711920344440411e-06, "loss": 0.0031, "step": 35971 }, { "epoch": 6.6, "learning_rate": 1.5707925558341109e-06, "loss": 0.002, "step": 35972 }, { "epoch": 6.6, "learning_rate": 1.5703931236861703e-06, "loss": 0.0029, "step": 35973 }, { "epoch": 6.6, "learning_rate": 1.56999373800242e-06, "loss": 0.0031, "step": 35974 }, { "epoch": 6.6, "learning_rate": 1.5695943987850604e-06, "loss": 0.0026, "step": 35975 }, { "epoch": 6.6, "learning_rate": 1.5691951060362953e-06, "loss": 0.0018, "step": 35976 }, { "epoch": 6.6, "learning_rate": 1.568795859758324e-06, "loss": 0.0018, "step": 35977 }, { "epoch": 6.6, "learning_rate": 1.568396659953345e-06, "loss": 0.0025, "step": 35978 }, { "epoch": 6.6, "learning_rate": 1.5679975066235665e-06, "loss": 0.0023, "step": 35979 }, { "epoch": 6.6, "learning_rate": 1.5675983997711797e-06, "loss": 0.0019, "step": 35980 }, { "epoch": 6.6, "learning_rate": 1.5671993393983886e-06, "loss": 0.0022, "step": 35981 }, { "epoch": 6.6, "learning_rate": 1.5668003255073905e-06, "loss": 0.0019, "step": 35982 }, { "epoch": 6.6, "learning_rate": 1.5664013581003878e-06, "loss": 0.002, "step": 35983 }, { "epoch": 6.6, "learning_rate": 1.566002437179579e-06, "loss": 0.0015, "step": 35984 }, { "epoch": 6.6, "learning_rate": 1.5656035627471655e-06, "loss": 0.0036, "step": 35985 }, { "epoch": 6.6, "learning_rate": 1.5652047348053368e-06, "loss": 0.0032, "step": 35986 }, { "epoch": 6.6, "learning_rate": 1.5648059533562999e-06, "loss": 0.0015, "step": 35987 }, { "epoch": 6.6, "learning_rate": 1.5644072184022508e-06, "loss": 0.0021, "step": 35988 }, { "epoch": 6.6, "learning_rate": 1.5640085299453867e-06, "loss": 0.0013, "step": 35989 }, { "epoch": 6.6, "learning_rate": 1.5636098879879046e-06, "loss": 0.0016, "step": 35990 }, { "epoch": 6.6, "learning_rate": 1.563211292532002e-06, "loss": 0.0021, "step": 35991 }, { "epoch": 6.6, "learning_rate": 1.562812743579878e-06, "loss": 0.0031, "step": 35992 }, { "epoch": 6.6, "learning_rate": 1.5624142411337261e-06, "loss": 0.0014, "step": 35993 }, { "epoch": 6.6, "learning_rate": 1.5620157851957462e-06, "loss": 0.0023, "step": 35994 }, { "epoch": 6.6, "learning_rate": 1.5616173757681297e-06, "loss": 0.0024, "step": 35995 }, { "epoch": 6.6, "learning_rate": 1.561219012853078e-06, "loss": 0.0011, "step": 35996 }, { "epoch": 6.6, "learning_rate": 1.5608206964527872e-06, "loss": 0.0021, "step": 35997 }, { "epoch": 6.6, "learning_rate": 1.560422426569449e-06, "loss": 0.0006, "step": 35998 }, { "epoch": 6.6, "learning_rate": 1.5600242032052593e-06, "loss": 0.0015, "step": 35999 }, { "epoch": 6.6, "learning_rate": 1.559626026362412e-06, "loss": 0.0031, "step": 36000 }, { "epoch": 6.6, "learning_rate": 1.559227896043105e-06, "loss": 0.0013, "step": 36001 }, { "epoch": 6.6, "learning_rate": 1.5588298122495327e-06, "loss": 0.002, "step": 36002 }, { "epoch": 6.6, "learning_rate": 1.5584317749838906e-06, "loss": 0.0028, "step": 36003 }, { "epoch": 6.6, "learning_rate": 1.558033784248365e-06, "loss": 0.002, "step": 36004 }, { "epoch": 6.6, "learning_rate": 1.5576358400451586e-06, "loss": 0.0012, "step": 36005 }, { "epoch": 6.6, "learning_rate": 1.5572379423764594e-06, "loss": 0.002, "step": 36006 }, { "epoch": 6.6, "learning_rate": 1.5568400912444636e-06, "loss": 0.0014, "step": 36007 }, { "epoch": 6.6, "learning_rate": 1.556442286651363e-06, "loss": 0.0017, "step": 36008 }, { "epoch": 6.6, "learning_rate": 1.5560445285993497e-06, "loss": 0.0017, "step": 36009 }, { "epoch": 6.6, "learning_rate": 1.555646817090618e-06, "loss": 0.0015, "step": 36010 }, { "epoch": 6.6, "learning_rate": 1.5552491521273583e-06, "loss": 0.0027, "step": 36011 }, { "epoch": 6.6, "learning_rate": 1.554851533711762e-06, "loss": 0.0014, "step": 36012 }, { "epoch": 6.6, "learning_rate": 1.5544539618460208e-06, "loss": 0.001, "step": 36013 }, { "epoch": 6.6, "learning_rate": 1.5540564365323319e-06, "loss": 0.0024, "step": 36014 }, { "epoch": 6.6, "learning_rate": 1.553658957772879e-06, "loss": 0.0012, "step": 36015 }, { "epoch": 6.6, "learning_rate": 1.5532615255698557e-06, "loss": 0.0026, "step": 36016 }, { "epoch": 6.6, "learning_rate": 1.5528641399254507e-06, "loss": 0.0012, "step": 36017 }, { "epoch": 6.61, "learning_rate": 1.5524668008418587e-06, "loss": 0.0014, "step": 36018 }, { "epoch": 6.61, "learning_rate": 1.5520695083212678e-06, "loss": 0.0018, "step": 36019 }, { "epoch": 6.61, "learning_rate": 1.5516722623658675e-06, "loss": 0.0012, "step": 36020 }, { "epoch": 6.61, "learning_rate": 1.5512750629778494e-06, "loss": 0.0014, "step": 36021 }, { "epoch": 6.61, "learning_rate": 1.5508779101593974e-06, "loss": 0.0015, "step": 36022 }, { "epoch": 6.61, "learning_rate": 1.5504808039127061e-06, "loss": 0.0025, "step": 36023 }, { "epoch": 6.61, "learning_rate": 1.550083744239963e-06, "loss": 0.0014, "step": 36024 }, { "epoch": 6.61, "learning_rate": 1.549686731143355e-06, "loss": 0.0012, "step": 36025 }, { "epoch": 6.61, "learning_rate": 1.5492897646250727e-06, "loss": 0.0015, "step": 36026 }, { "epoch": 6.61, "learning_rate": 1.548892844687303e-06, "loss": 0.0027, "step": 36027 }, { "epoch": 6.61, "learning_rate": 1.5484959713322345e-06, "loss": 0.002, "step": 36028 }, { "epoch": 6.61, "learning_rate": 1.5480991445620541e-06, "loss": 0.0023, "step": 36029 }, { "epoch": 6.61, "learning_rate": 1.5477023643789491e-06, "loss": 0.0014, "step": 36030 }, { "epoch": 6.61, "learning_rate": 1.547305630785103e-06, "loss": 0.0009, "step": 36031 }, { "epoch": 6.61, "learning_rate": 1.5469089437827123e-06, "loss": 0.002, "step": 36032 }, { "epoch": 6.61, "learning_rate": 1.5465123033739549e-06, "loss": 0.0013, "step": 36033 }, { "epoch": 6.61, "learning_rate": 1.5461157095610202e-06, "loss": 0.0019, "step": 36034 }, { "epoch": 6.61, "learning_rate": 1.545719162346091e-06, "loss": 0.0011, "step": 36035 }, { "epoch": 6.61, "learning_rate": 1.5453226617313577e-06, "loss": 0.0022, "step": 36036 }, { "epoch": 6.61, "learning_rate": 1.5449262077190053e-06, "loss": 0.0012, "step": 36037 }, { "epoch": 6.61, "learning_rate": 1.5445298003112163e-06, "loss": 0.002, "step": 36038 }, { "epoch": 6.61, "learning_rate": 1.5441334395101781e-06, "loss": 0.0013, "step": 36039 }, { "epoch": 6.61, "learning_rate": 1.5437371253180745e-06, "loss": 0.0021, "step": 36040 }, { "epoch": 6.61, "learning_rate": 1.543340857737089e-06, "loss": 0.0019, "step": 36041 }, { "epoch": 6.61, "learning_rate": 1.5429446367694079e-06, "loss": 0.0023, "step": 36042 }, { "epoch": 6.61, "learning_rate": 1.5425484624172127e-06, "loss": 0.0016, "step": 36043 }, { "epoch": 6.61, "learning_rate": 1.5421523346826894e-06, "loss": 0.0016, "step": 36044 }, { "epoch": 6.61, "learning_rate": 1.5417562535680207e-06, "loss": 0.0016, "step": 36045 }, { "epoch": 6.61, "learning_rate": 1.5413602190753895e-06, "loss": 0.0016, "step": 36046 }, { "epoch": 6.61, "learning_rate": 1.5409642312069794e-06, "loss": 0.0019, "step": 36047 }, { "epoch": 6.61, "learning_rate": 1.54056828996497e-06, "loss": 0.0023, "step": 36048 }, { "epoch": 6.61, "learning_rate": 1.540172395351549e-06, "loss": 0.002, "step": 36049 }, { "epoch": 6.61, "learning_rate": 1.5397765473688975e-06, "loss": 0.0019, "step": 36050 }, { "epoch": 6.61, "learning_rate": 1.5393807460191934e-06, "loss": 0.0018, "step": 36051 }, { "epoch": 6.61, "learning_rate": 1.5389849913046217e-06, "loss": 0.0015, "step": 36052 }, { "epoch": 6.61, "learning_rate": 1.5385892832273608e-06, "loss": 0.0021, "step": 36053 }, { "epoch": 6.61, "learning_rate": 1.5381936217895954e-06, "loss": 0.0016, "step": 36054 }, { "epoch": 6.61, "learning_rate": 1.537798006993505e-06, "loss": 0.0024, "step": 36055 }, { "epoch": 6.61, "learning_rate": 1.5374024388412713e-06, "loss": 0.0031, "step": 36056 }, { "epoch": 6.61, "learning_rate": 1.5370069173350721e-06, "loss": 0.0017, "step": 36057 }, { "epoch": 6.61, "learning_rate": 1.5366114424770905e-06, "loss": 0.0026, "step": 36058 }, { "epoch": 6.61, "learning_rate": 1.5362160142695037e-06, "loss": 0.0016, "step": 36059 }, { "epoch": 6.61, "learning_rate": 1.5358206327144942e-06, "loss": 0.0019, "step": 36060 }, { "epoch": 6.61, "learning_rate": 1.535425297814238e-06, "loss": 0.0015, "step": 36061 }, { "epoch": 6.61, "learning_rate": 1.535030009570917e-06, "loss": 0.001, "step": 36062 }, { "epoch": 6.61, "learning_rate": 1.534634767986708e-06, "loss": 0.0011, "step": 36063 }, { "epoch": 6.61, "learning_rate": 1.5342395730637904e-06, "loss": 0.0032, "step": 36064 }, { "epoch": 6.61, "learning_rate": 1.5338444248043438e-06, "loss": 0.0021, "step": 36065 }, { "epoch": 6.61, "learning_rate": 1.5334493232105418e-06, "loss": 0.0015, "step": 36066 }, { "epoch": 6.61, "learning_rate": 1.5330542682845672e-06, "loss": 0.0016, "step": 36067 }, { "epoch": 6.61, "learning_rate": 1.5326592600285994e-06, "loss": 0.0023, "step": 36068 }, { "epoch": 6.61, "learning_rate": 1.5322642984448088e-06, "loss": 0.0014, "step": 36069 }, { "epoch": 6.61, "learning_rate": 1.531869383535375e-06, "loss": 0.0027, "step": 36070 }, { "epoch": 6.61, "learning_rate": 1.5314745153024768e-06, "loss": 0.001, "step": 36071 }, { "epoch": 6.62, "learning_rate": 1.5310796937482886e-06, "loss": 0.0013, "step": 36072 }, { "epoch": 6.62, "learning_rate": 1.5306849188749884e-06, "loss": 0.0013, "step": 36073 }, { "epoch": 6.62, "learning_rate": 1.5302901906847522e-06, "loss": 0.0013, "step": 36074 }, { "epoch": 6.62, "learning_rate": 1.529895509179754e-06, "loss": 0.0029, "step": 36075 }, { "epoch": 6.62, "learning_rate": 1.5295008743621698e-06, "loss": 0.0018, "step": 36076 }, { "epoch": 6.62, "learning_rate": 1.5291062862341754e-06, "loss": 0.0018, "step": 36077 }, { "epoch": 6.62, "learning_rate": 1.528711744797946e-06, "loss": 0.001, "step": 36078 }, { "epoch": 6.62, "learning_rate": 1.5283172500556564e-06, "loss": 0.0018, "step": 36079 }, { "epoch": 6.62, "learning_rate": 1.5279228020094795e-06, "loss": 0.0018, "step": 36080 }, { "epoch": 6.62, "learning_rate": 1.5275284006615921e-06, "loss": 0.0024, "step": 36081 }, { "epoch": 6.62, "learning_rate": 1.5271340460141647e-06, "loss": 0.0027, "step": 36082 }, { "epoch": 6.62, "learning_rate": 1.5267397380693738e-06, "loss": 0.0015, "step": 36083 }, { "epoch": 6.62, "learning_rate": 1.5263454768293906e-06, "loss": 0.0017, "step": 36084 }, { "epoch": 6.62, "learning_rate": 1.52595126229639e-06, "loss": 0.0018, "step": 36085 }, { "epoch": 6.62, "learning_rate": 1.5255570944725483e-06, "loss": 0.0015, "step": 36086 }, { "epoch": 6.62, "learning_rate": 1.5251629733600327e-06, "loss": 0.003, "step": 36087 }, { "epoch": 6.62, "learning_rate": 1.5247688989610131e-06, "loss": 0.0013, "step": 36088 }, { "epoch": 6.62, "learning_rate": 1.5243748712776697e-06, "loss": 0.002, "step": 36089 }, { "epoch": 6.62, "learning_rate": 1.52398089031217e-06, "loss": 0.0014, "step": 36090 }, { "epoch": 6.62, "learning_rate": 1.523586956066686e-06, "loss": 0.0013, "step": 36091 }, { "epoch": 6.62, "learning_rate": 1.5231930685433904e-06, "loss": 0.0016, "step": 36092 }, { "epoch": 6.62, "learning_rate": 1.5227992277444525e-06, "loss": 0.0023, "step": 36093 }, { "epoch": 6.62, "learning_rate": 1.5224054336720429e-06, "loss": 0.0014, "step": 36094 }, { "epoch": 6.62, "learning_rate": 1.5220116863283341e-06, "loss": 0.0015, "step": 36095 }, { "epoch": 6.62, "learning_rate": 1.5216179857154956e-06, "loss": 0.0014, "step": 36096 }, { "epoch": 6.62, "learning_rate": 1.5212243318356967e-06, "loss": 0.0025, "step": 36097 }, { "epoch": 6.62, "learning_rate": 1.5208307246911092e-06, "loss": 0.0012, "step": 36098 }, { "epoch": 6.62, "learning_rate": 1.5204371642839e-06, "loss": 0.0017, "step": 36099 }, { "epoch": 6.62, "learning_rate": 1.5200436506162408e-06, "loss": 0.002, "step": 36100 }, { "epoch": 6.62, "learning_rate": 1.5196501836902965e-06, "loss": 0.0016, "step": 36101 }, { "epoch": 6.62, "learning_rate": 1.5192567635082422e-06, "loss": 0.0013, "step": 36102 }, { "epoch": 6.62, "learning_rate": 1.5188633900722428e-06, "loss": 0.0019, "step": 36103 }, { "epoch": 6.62, "learning_rate": 1.5184700633844696e-06, "loss": 0.0016, "step": 36104 }, { "epoch": 6.62, "learning_rate": 1.5180767834470855e-06, "loss": 0.0013, "step": 36105 }, { "epoch": 6.62, "learning_rate": 1.517683550262259e-06, "loss": 0.0012, "step": 36106 }, { "epoch": 6.62, "learning_rate": 1.5172903638321613e-06, "loss": 0.0015, "step": 36107 }, { "epoch": 6.62, "learning_rate": 1.5168972241589586e-06, "loss": 0.0066, "step": 36108 }, { "epoch": 6.62, "learning_rate": 1.516504131244817e-06, "loss": 0.0024, "step": 36109 }, { "epoch": 6.62, "learning_rate": 1.5161110850919025e-06, "loss": 0.0025, "step": 36110 }, { "epoch": 6.62, "learning_rate": 1.5157180857023834e-06, "loss": 0.0018, "step": 36111 }, { "epoch": 6.62, "learning_rate": 1.5153251330784246e-06, "loss": 0.0015, "step": 36112 }, { "epoch": 6.62, "learning_rate": 1.5149322272221922e-06, "loss": 0.0009, "step": 36113 }, { "epoch": 6.62, "learning_rate": 1.51453936813585e-06, "loss": 0.0023, "step": 36114 }, { "epoch": 6.62, "learning_rate": 1.5141465558215708e-06, "loss": 0.0019, "step": 36115 }, { "epoch": 6.62, "learning_rate": 1.5137537902815113e-06, "loss": 0.0029, "step": 36116 }, { "epoch": 6.62, "learning_rate": 1.5133610715178405e-06, "loss": 0.0011, "step": 36117 }, { "epoch": 6.62, "learning_rate": 1.5129683995327217e-06, "loss": 0.0022, "step": 36118 }, { "epoch": 6.62, "learning_rate": 1.5125757743283175e-06, "loss": 0.0024, "step": 36119 }, { "epoch": 6.62, "learning_rate": 1.512183195906798e-06, "loss": 0.0019, "step": 36120 }, { "epoch": 6.62, "learning_rate": 1.5117906642703218e-06, "loss": 0.0026, "step": 36121 }, { "epoch": 6.62, "learning_rate": 1.5113981794210574e-06, "loss": 0.002, "step": 36122 }, { "epoch": 6.62, "learning_rate": 1.5110057413611613e-06, "loss": 0.001, "step": 36123 }, { "epoch": 6.62, "learning_rate": 1.5106133500928021e-06, "loss": 0.0022, "step": 36124 }, { "epoch": 6.62, "learning_rate": 1.510221005618141e-06, "loss": 0.0037, "step": 36125 }, { "epoch": 6.62, "learning_rate": 1.5098287079393403e-06, "loss": 0.0035, "step": 36126 }, { "epoch": 6.63, "learning_rate": 1.509436457058563e-06, "loss": 0.0011, "step": 36127 }, { "epoch": 6.63, "learning_rate": 1.5090442529779704e-06, "loss": 0.0021, "step": 36128 }, { "epoch": 6.63, "learning_rate": 1.5086520956997253e-06, "loss": 0.0015, "step": 36129 }, { "epoch": 6.63, "learning_rate": 1.5082599852259895e-06, "loss": 0.0016, "step": 36130 }, { "epoch": 6.63, "learning_rate": 1.507867921558922e-06, "loss": 0.0014, "step": 36131 }, { "epoch": 6.63, "learning_rate": 1.5074759047006837e-06, "loss": 0.002, "step": 36132 }, { "epoch": 6.63, "learning_rate": 1.5070839346534426e-06, "loss": 0.0016, "step": 36133 }, { "epoch": 6.63, "learning_rate": 1.5066920114193506e-06, "loss": 0.0018, "step": 36134 }, { "epoch": 6.63, "learning_rate": 1.5063001350005724e-06, "loss": 0.0018, "step": 36135 }, { "epoch": 6.63, "learning_rate": 1.505908305399265e-06, "loss": 0.0023, "step": 36136 }, { "epoch": 6.63, "learning_rate": 1.5055165226175893e-06, "loss": 0.0007, "step": 36137 }, { "epoch": 6.63, "learning_rate": 1.5051247866577068e-06, "loss": 0.0007, "step": 36138 }, { "epoch": 6.63, "learning_rate": 1.504733097521779e-06, "loss": 0.0015, "step": 36139 }, { "epoch": 6.63, "learning_rate": 1.5043414552119584e-06, "loss": 0.0023, "step": 36140 }, { "epoch": 6.63, "learning_rate": 1.5039498597304035e-06, "loss": 0.0016, "step": 36141 }, { "epoch": 6.63, "learning_rate": 1.5035583110792795e-06, "loss": 0.0015, "step": 36142 }, { "epoch": 6.63, "learning_rate": 1.503166809260741e-06, "loss": 0.0011, "step": 36143 }, { "epoch": 6.63, "learning_rate": 1.5027753542769463e-06, "loss": 0.0013, "step": 36144 }, { "epoch": 6.63, "learning_rate": 1.5023839461300526e-06, "loss": 0.002, "step": 36145 }, { "epoch": 6.63, "learning_rate": 1.5019925848222172e-06, "loss": 0.0027, "step": 36146 }, { "epoch": 6.63, "learning_rate": 1.5016012703555983e-06, "loss": 0.0011, "step": 36147 }, { "epoch": 6.63, "learning_rate": 1.5012100027323517e-06, "loss": 0.0018, "step": 36148 }, { "epoch": 6.63, "learning_rate": 1.5008187819546349e-06, "loss": 0.0016, "step": 36149 }, { "epoch": 6.63, "learning_rate": 1.5004276080246028e-06, "loss": 0.0027, "step": 36150 }, { "epoch": 6.63, "learning_rate": 1.5000364809444168e-06, "loss": 0.0017, "step": 36151 }, { "epoch": 6.63, "learning_rate": 1.4996454007162264e-06, "loss": 0.0019, "step": 36152 }, { "epoch": 6.63, "learning_rate": 1.499254367342189e-06, "loss": 0.0032, "step": 36153 }, { "epoch": 6.63, "learning_rate": 1.498863380824459e-06, "loss": 0.0018, "step": 36154 }, { "epoch": 6.63, "learning_rate": 1.4984724411651962e-06, "loss": 0.0023, "step": 36155 }, { "epoch": 6.63, "learning_rate": 1.4980815483665513e-06, "loss": 0.0019, "step": 36156 }, { "epoch": 6.63, "learning_rate": 1.497690702430683e-06, "loss": 0.002, "step": 36157 }, { "epoch": 6.63, "learning_rate": 1.4972999033597402e-06, "loss": 0.0023, "step": 36158 }, { "epoch": 6.63, "learning_rate": 1.4969091511558776e-06, "loss": 0.0025, "step": 36159 }, { "epoch": 6.63, "learning_rate": 1.4965184458212534e-06, "loss": 0.003, "step": 36160 }, { "epoch": 6.63, "learning_rate": 1.4961277873580182e-06, "loss": 0.0017, "step": 36161 }, { "epoch": 6.63, "learning_rate": 1.4957371757683258e-06, "loss": 0.0028, "step": 36162 }, { "epoch": 6.63, "learning_rate": 1.4953466110543303e-06, "loss": 0.0022, "step": 36163 }, { "epoch": 6.63, "learning_rate": 1.4949560932181827e-06, "loss": 0.0024, "step": 36164 }, { "epoch": 6.63, "learning_rate": 1.4945656222620363e-06, "loss": 0.0012, "step": 36165 }, { "epoch": 6.63, "learning_rate": 1.4941751981880437e-06, "loss": 0.002, "step": 36166 }, { "epoch": 6.63, "learning_rate": 1.493784820998354e-06, "loss": 0.0011, "step": 36167 }, { "epoch": 6.63, "learning_rate": 1.4933944906951247e-06, "loss": 0.0018, "step": 36168 }, { "epoch": 6.63, "learning_rate": 1.4930042072805062e-06, "loss": 0.0018, "step": 36169 }, { "epoch": 6.63, "learning_rate": 1.4926139707566456e-06, "loss": 0.0007, "step": 36170 }, { "epoch": 6.63, "learning_rate": 1.4922237811256957e-06, "loss": 0.0013, "step": 36171 }, { "epoch": 6.63, "learning_rate": 1.4918336383898057e-06, "loss": 0.0022, "step": 36172 }, { "epoch": 6.63, "learning_rate": 1.4914435425511298e-06, "loss": 0.0023, "step": 36173 }, { "epoch": 6.63, "learning_rate": 1.4910534936118172e-06, "loss": 0.0014, "step": 36174 }, { "epoch": 6.63, "learning_rate": 1.4906634915740192e-06, "loss": 0.0025, "step": 36175 }, { "epoch": 6.63, "learning_rate": 1.4902735364398778e-06, "loss": 0.0023, "step": 36176 }, { "epoch": 6.63, "learning_rate": 1.4898836282115514e-06, "loss": 0.0013, "step": 36177 }, { "epoch": 6.63, "learning_rate": 1.4894937668911857e-06, "loss": 0.0014, "step": 36178 }, { "epoch": 6.63, "learning_rate": 1.4891039524809292e-06, "loss": 0.0014, "step": 36179 }, { "epoch": 6.63, "learning_rate": 1.4887141849829322e-06, "loss": 0.0023, "step": 36180 }, { "epoch": 6.64, "learning_rate": 1.4883244643993411e-06, "loss": 0.0011, "step": 36181 }, { "epoch": 6.64, "learning_rate": 1.487934790732305e-06, "loss": 0.0034, "step": 36182 }, { "epoch": 6.64, "learning_rate": 1.4875451639839721e-06, "loss": 0.0013, "step": 36183 }, { "epoch": 6.64, "learning_rate": 1.4871555841564889e-06, "loss": 0.0016, "step": 36184 }, { "epoch": 6.64, "learning_rate": 1.4867660512520021e-06, "loss": 0.0013, "step": 36185 }, { "epoch": 6.64, "learning_rate": 1.4863765652726647e-06, "loss": 0.0022, "step": 36186 }, { "epoch": 6.64, "learning_rate": 1.485987126220617e-06, "loss": 0.0024, "step": 36187 }, { "epoch": 6.64, "learning_rate": 1.4855977340980066e-06, "loss": 0.0017, "step": 36188 }, { "epoch": 6.64, "learning_rate": 1.4852083889069801e-06, "loss": 0.0025, "step": 36189 }, { "epoch": 6.64, "learning_rate": 1.4848190906496863e-06, "loss": 0.0016, "step": 36190 }, { "epoch": 6.64, "learning_rate": 1.4844298393282685e-06, "loss": 0.0011, "step": 36191 }, { "epoch": 6.64, "learning_rate": 1.4840406349448732e-06, "loss": 0.0016, "step": 36192 }, { "epoch": 6.64, "learning_rate": 1.4836514775016486e-06, "loss": 0.0013, "step": 36193 }, { "epoch": 6.64, "learning_rate": 1.4832623670007306e-06, "loss": 0.0025, "step": 36194 }, { "epoch": 6.64, "learning_rate": 1.4828733034442733e-06, "loss": 0.0028, "step": 36195 }, { "epoch": 6.64, "learning_rate": 1.482484286834418e-06, "loss": 0.0013, "step": 36196 }, { "epoch": 6.64, "learning_rate": 1.4820953171733076e-06, "loss": 0.0016, "step": 36197 }, { "epoch": 6.64, "learning_rate": 1.4817063944630882e-06, "loss": 0.002, "step": 36198 }, { "epoch": 6.64, "learning_rate": 1.4813175187059036e-06, "loss": 0.0022, "step": 36199 }, { "epoch": 6.64, "learning_rate": 1.4809286899038954e-06, "loss": 0.0016, "step": 36200 }, { "epoch": 6.64, "learning_rate": 1.4805399080592076e-06, "loss": 0.002, "step": 36201 }, { "epoch": 6.64, "learning_rate": 1.4801511731739838e-06, "loss": 0.0019, "step": 36202 }, { "epoch": 6.64, "learning_rate": 1.4797624852503645e-06, "loss": 0.0016, "step": 36203 }, { "epoch": 6.64, "learning_rate": 1.4793738442904982e-06, "loss": 0.0021, "step": 36204 }, { "epoch": 6.64, "learning_rate": 1.4789852502965196e-06, "loss": 0.0032, "step": 36205 }, { "epoch": 6.64, "learning_rate": 1.478596703270575e-06, "loss": 0.0016, "step": 36206 }, { "epoch": 6.64, "learning_rate": 1.4782082032148026e-06, "loss": 0.0016, "step": 36207 }, { "epoch": 6.64, "learning_rate": 1.4778197501313485e-06, "loss": 0.0014, "step": 36208 }, { "epoch": 6.64, "learning_rate": 1.477431344022352e-06, "loss": 0.0016, "step": 36209 }, { "epoch": 6.64, "learning_rate": 1.4770429848899526e-06, "loss": 0.0019, "step": 36210 }, { "epoch": 6.64, "learning_rate": 1.4766546727362951e-06, "loss": 0.0023, "step": 36211 }, { "epoch": 6.64, "learning_rate": 1.4762664075635113e-06, "loss": 0.0037, "step": 36212 }, { "epoch": 6.64, "learning_rate": 1.4758781893737496e-06, "loss": 0.0018, "step": 36213 }, { "epoch": 6.64, "learning_rate": 1.4754900181691467e-06, "loss": 0.0014, "step": 36214 }, { "epoch": 6.64, "learning_rate": 1.4751018939518436e-06, "loss": 0.0016, "step": 36215 }, { "epoch": 6.64, "learning_rate": 1.4747138167239772e-06, "loss": 0.0026, "step": 36216 }, { "epoch": 6.64, "learning_rate": 1.4743257864876891e-06, "loss": 0.0021, "step": 36217 }, { "epoch": 6.64, "learning_rate": 1.4739378032451157e-06, "loss": 0.0016, "step": 36218 }, { "epoch": 6.64, "learning_rate": 1.4735498669983984e-06, "loss": 0.001, "step": 36219 }, { "epoch": 6.64, "learning_rate": 1.473161977749671e-06, "loss": 0.0019, "step": 36220 }, { "epoch": 6.64, "learning_rate": 1.472774135501076e-06, "loss": 0.0017, "step": 36221 }, { "epoch": 6.64, "learning_rate": 1.4723863402547533e-06, "loss": 0.0014, "step": 36222 }, { "epoch": 6.64, "learning_rate": 1.4719985920128343e-06, "loss": 0.0011, "step": 36223 }, { "epoch": 6.64, "learning_rate": 1.4716108907774584e-06, "loss": 0.0019, "step": 36224 }, { "epoch": 6.64, "learning_rate": 1.4712232365507605e-06, "loss": 0.0016, "step": 36225 }, { "epoch": 6.64, "learning_rate": 1.4708356293348824e-06, "loss": 0.0025, "step": 36226 }, { "epoch": 6.64, "learning_rate": 1.470448069131959e-06, "loss": 0.0015, "step": 36227 }, { "epoch": 6.64, "learning_rate": 1.470060555944125e-06, "loss": 0.0015, "step": 36228 }, { "epoch": 6.64, "learning_rate": 1.4696730897735168e-06, "loss": 0.0027, "step": 36229 }, { "epoch": 6.64, "learning_rate": 1.4692856706222714e-06, "loss": 0.0012, "step": 36230 }, { "epoch": 6.64, "learning_rate": 1.4688982984925226e-06, "loss": 0.0011, "step": 36231 }, { "epoch": 6.64, "learning_rate": 1.4685109733864055e-06, "loss": 0.0017, "step": 36232 }, { "epoch": 6.64, "learning_rate": 1.4681236953060574e-06, "loss": 0.0024, "step": 36233 }, { "epoch": 6.64, "learning_rate": 1.4677364642536096e-06, "loss": 0.0015, "step": 36234 }, { "epoch": 6.64, "learning_rate": 1.4673492802311995e-06, "loss": 0.0011, "step": 36235 }, { "epoch": 6.65, "learning_rate": 1.4669621432409598e-06, "loss": 0.0017, "step": 36236 }, { "epoch": 6.65, "learning_rate": 1.4665750532850241e-06, "loss": 0.0015, "step": 36237 }, { "epoch": 6.65, "learning_rate": 1.4661880103655258e-06, "loss": 0.0019, "step": 36238 }, { "epoch": 6.65, "learning_rate": 1.4658010144846001e-06, "loss": 0.0019, "step": 36239 }, { "epoch": 6.65, "learning_rate": 1.4654140656443817e-06, "loss": 0.0028, "step": 36240 }, { "epoch": 6.65, "learning_rate": 1.4650271638469981e-06, "loss": 0.0012, "step": 36241 }, { "epoch": 6.65, "learning_rate": 1.4646403090945839e-06, "loss": 0.0019, "step": 36242 }, { "epoch": 6.65, "learning_rate": 1.4642535013892734e-06, "loss": 0.002, "step": 36243 }, { "epoch": 6.65, "learning_rate": 1.4638667407331974e-06, "loss": 0.0025, "step": 36244 }, { "epoch": 6.65, "learning_rate": 1.4634800271284889e-06, "loss": 0.0018, "step": 36245 }, { "epoch": 6.65, "learning_rate": 1.4630933605772801e-06, "loss": 0.0021, "step": 36246 }, { "epoch": 6.65, "learning_rate": 1.4627067410816963e-06, "loss": 0.0012, "step": 36247 }, { "epoch": 6.65, "learning_rate": 1.462320168643875e-06, "loss": 0.0012, "step": 36248 }, { "epoch": 6.65, "learning_rate": 1.4619336432659447e-06, "loss": 0.0012, "step": 36249 }, { "epoch": 6.65, "learning_rate": 1.4615471649500367e-06, "loss": 0.0017, "step": 36250 }, { "epoch": 6.65, "learning_rate": 1.4611607336982813e-06, "loss": 0.0015, "step": 36251 }, { "epoch": 6.65, "learning_rate": 1.4607743495128067e-06, "loss": 0.002, "step": 36252 }, { "epoch": 6.65, "learning_rate": 1.4603880123957448e-06, "loss": 0.0018, "step": 36253 }, { "epoch": 6.65, "learning_rate": 1.4600017223492246e-06, "loss": 0.001, "step": 36254 }, { "epoch": 6.65, "learning_rate": 1.4596154793753748e-06, "loss": 0.0021, "step": 36255 }, { "epoch": 6.65, "learning_rate": 1.4592292834763211e-06, "loss": 0.0015, "step": 36256 }, { "epoch": 6.65, "learning_rate": 1.4588431346541988e-06, "loss": 0.0013, "step": 36257 }, { "epoch": 6.65, "learning_rate": 1.4584570329111358e-06, "loss": 0.0011, "step": 36258 }, { "epoch": 6.65, "learning_rate": 1.4580709782492551e-06, "loss": 0.0013, "step": 36259 }, { "epoch": 6.65, "learning_rate": 1.4576849706706853e-06, "loss": 0.0026, "step": 36260 }, { "epoch": 6.65, "learning_rate": 1.4572990101775574e-06, "loss": 0.0016, "step": 36261 }, { "epoch": 6.65, "learning_rate": 1.456913096771999e-06, "loss": 0.0013, "step": 36262 }, { "epoch": 6.65, "learning_rate": 1.456527230456134e-06, "loss": 0.0018, "step": 36263 }, { "epoch": 6.65, "learning_rate": 1.4561414112320948e-06, "loss": 0.0022, "step": 36264 }, { "epoch": 6.65, "learning_rate": 1.4557556391019989e-06, "loss": 0.0012, "step": 36265 }, { "epoch": 6.65, "learning_rate": 1.4553699140679812e-06, "loss": 0.0011, "step": 36266 }, { "epoch": 6.65, "learning_rate": 1.4549842361321643e-06, "loss": 0.0019, "step": 36267 }, { "epoch": 6.65, "learning_rate": 1.4545986052966743e-06, "loss": 0.0019, "step": 36268 }, { "epoch": 6.65, "learning_rate": 1.4542130215636364e-06, "loss": 0.0011, "step": 36269 }, { "epoch": 6.65, "learning_rate": 1.4538274849351774e-06, "loss": 0.0029, "step": 36270 }, { "epoch": 6.65, "learning_rate": 1.4534419954134216e-06, "loss": 0.002, "step": 36271 }, { "epoch": 6.65, "learning_rate": 1.4530565530004936e-06, "loss": 0.0014, "step": 36272 }, { "epoch": 6.65, "learning_rate": 1.4526711576985152e-06, "loss": 0.0023, "step": 36273 }, { "epoch": 6.65, "learning_rate": 1.452285809509617e-06, "loss": 0.0018, "step": 36274 }, { "epoch": 6.65, "learning_rate": 1.4519005084359183e-06, "loss": 0.004, "step": 36275 }, { "epoch": 6.65, "learning_rate": 1.4515152544795473e-06, "loss": 0.002, "step": 36276 }, { "epoch": 6.65, "learning_rate": 1.4511300476426227e-06, "loss": 0.0017, "step": 36277 }, { "epoch": 6.65, "learning_rate": 1.4507448879272667e-06, "loss": 0.0019, "step": 36278 }, { "epoch": 6.65, "learning_rate": 1.450359775335608e-06, "loss": 0.0022, "step": 36279 }, { "epoch": 6.65, "learning_rate": 1.4499747098697658e-06, "loss": 0.0012, "step": 36280 }, { "epoch": 6.65, "learning_rate": 1.449589691531863e-06, "loss": 0.0012, "step": 36281 }, { "epoch": 6.65, "learning_rate": 1.4492047203240234e-06, "loss": 0.0023, "step": 36282 }, { "epoch": 6.65, "learning_rate": 1.4488197962483664e-06, "loss": 0.0011, "step": 36283 }, { "epoch": 6.65, "learning_rate": 1.4484349193070157e-06, "loss": 0.0015, "step": 36284 }, { "epoch": 6.65, "learning_rate": 1.4480500895020922e-06, "loss": 0.0034, "step": 36285 }, { "epoch": 6.65, "learning_rate": 1.447665306835717e-06, "loss": 0.0021, "step": 36286 }, { "epoch": 6.65, "learning_rate": 1.4472805713100103e-06, "loss": 0.0018, "step": 36287 }, { "epoch": 6.65, "learning_rate": 1.446895882927094e-06, "loss": 0.0014, "step": 36288 }, { "epoch": 6.65, "learning_rate": 1.446511241689088e-06, "loss": 0.0016, "step": 36289 }, { "epoch": 6.66, "learning_rate": 1.4461266475981118e-06, "loss": 0.0021, "step": 36290 }, { "epoch": 6.66, "learning_rate": 1.4457421006562843e-06, "loss": 0.0015, "step": 36291 }, { "epoch": 6.66, "learning_rate": 1.4453576008657288e-06, "loss": 0.001, "step": 36292 }, { "epoch": 6.66, "learning_rate": 1.4449731482285646e-06, "loss": 0.0008, "step": 36293 }, { "epoch": 6.66, "learning_rate": 1.4445887427469074e-06, "loss": 0.0019, "step": 36294 }, { "epoch": 6.66, "learning_rate": 1.4442043844228747e-06, "loss": 0.0018, "step": 36295 }, { "epoch": 6.66, "learning_rate": 1.4438200732585906e-06, "loss": 0.001, "step": 36296 }, { "epoch": 6.66, "learning_rate": 1.4434358092561694e-06, "loss": 0.0014, "step": 36297 }, { "epoch": 6.66, "learning_rate": 1.4430515924177314e-06, "loss": 0.0018, "step": 36298 }, { "epoch": 6.66, "learning_rate": 1.442667422745394e-06, "loss": 0.0019, "step": 36299 }, { "epoch": 6.66, "learning_rate": 1.4422833002412728e-06, "loss": 0.0018, "step": 36300 }, { "epoch": 6.66, "learning_rate": 1.441899224907488e-06, "loss": 0.0022, "step": 36301 }, { "epoch": 6.66, "learning_rate": 1.4415151967461549e-06, "loss": 0.0029, "step": 36302 }, { "epoch": 6.66, "learning_rate": 1.4411312157593893e-06, "loss": 0.002, "step": 36303 }, { "epoch": 6.66, "learning_rate": 1.4407472819493085e-06, "loss": 0.0019, "step": 36304 }, { "epoch": 6.66, "learning_rate": 1.4403633953180318e-06, "loss": 0.0014, "step": 36305 }, { "epoch": 6.66, "learning_rate": 1.4399795558676722e-06, "loss": 0.0017, "step": 36306 }, { "epoch": 6.66, "learning_rate": 1.4395957636003443e-06, "loss": 0.0019, "step": 36307 }, { "epoch": 6.66, "learning_rate": 1.4392120185181657e-06, "loss": 0.0022, "step": 36308 }, { "epoch": 6.66, "learning_rate": 1.4388283206232489e-06, "loss": 0.0024, "step": 36309 }, { "epoch": 6.66, "learning_rate": 1.4384446699177134e-06, "loss": 0.0016, "step": 36310 }, { "epoch": 6.66, "learning_rate": 1.438061066403673e-06, "loss": 0.0022, "step": 36311 }, { "epoch": 6.66, "learning_rate": 1.4376775100832396e-06, "loss": 0.002, "step": 36312 }, { "epoch": 6.66, "learning_rate": 1.4372940009585245e-06, "loss": 0.0017, "step": 36313 }, { "epoch": 6.66, "learning_rate": 1.4369105390316496e-06, "loss": 0.0013, "step": 36314 }, { "epoch": 6.66, "learning_rate": 1.436527124304723e-06, "loss": 0.0011, "step": 36315 }, { "epoch": 6.66, "learning_rate": 1.4361437567798596e-06, "loss": 0.002, "step": 36316 }, { "epoch": 6.66, "learning_rate": 1.4357604364591727e-06, "loss": 0.0035, "step": 36317 }, { "epoch": 6.66, "learning_rate": 1.4353771633447756e-06, "loss": 0.0019, "step": 36318 }, { "epoch": 6.66, "learning_rate": 1.434993937438779e-06, "loss": 0.0022, "step": 36319 }, { "epoch": 6.66, "learning_rate": 1.4346107587432967e-06, "loss": 0.0024, "step": 36320 }, { "epoch": 6.66, "learning_rate": 1.4342276272604416e-06, "loss": 0.0029, "step": 36321 }, { "epoch": 6.66, "learning_rate": 1.4338445429923208e-06, "loss": 0.0023, "step": 36322 }, { "epoch": 6.66, "learning_rate": 1.433461505941055e-06, "loss": 0.0009, "step": 36323 }, { "epoch": 6.66, "learning_rate": 1.433078516108748e-06, "loss": 0.0024, "step": 36324 }, { "epoch": 6.66, "learning_rate": 1.4326955734975124e-06, "loss": 0.0015, "step": 36325 }, { "epoch": 6.66, "learning_rate": 1.4323126781094577e-06, "loss": 0.0017, "step": 36326 }, { "epoch": 6.66, "learning_rate": 1.4319298299466977e-06, "loss": 0.0016, "step": 36327 }, { "epoch": 6.66, "learning_rate": 1.431547029011342e-06, "loss": 0.0032, "step": 36328 }, { "epoch": 6.66, "learning_rate": 1.4311642753055011e-06, "loss": 0.0019, "step": 36329 }, { "epoch": 6.66, "learning_rate": 1.4307815688312821e-06, "loss": 0.0025, "step": 36330 }, { "epoch": 6.66, "learning_rate": 1.4303989095907933e-06, "loss": 0.0012, "step": 36331 }, { "epoch": 6.66, "learning_rate": 1.4300162975861487e-06, "loss": 0.0019, "step": 36332 }, { "epoch": 6.66, "learning_rate": 1.4296337328194555e-06, "loss": 0.0035, "step": 36333 }, { "epoch": 6.66, "learning_rate": 1.4292512152928218e-06, "loss": 0.0018, "step": 36334 }, { "epoch": 6.66, "learning_rate": 1.428868745008355e-06, "loss": 0.0015, "step": 36335 }, { "epoch": 6.66, "learning_rate": 1.4284863219681654e-06, "loss": 0.0017, "step": 36336 }, { "epoch": 6.66, "learning_rate": 1.4281039461743607e-06, "loss": 0.0013, "step": 36337 }, { "epoch": 6.66, "learning_rate": 1.4277216176290465e-06, "loss": 0.0021, "step": 36338 }, { "epoch": 6.66, "learning_rate": 1.4273393363343324e-06, "loss": 0.0018, "step": 36339 }, { "epoch": 6.66, "learning_rate": 1.4269571022923245e-06, "loss": 0.002, "step": 36340 }, { "epoch": 6.66, "learning_rate": 1.42657491550513e-06, "loss": 0.0014, "step": 36341 }, { "epoch": 6.66, "learning_rate": 1.426192775974856e-06, "loss": 0.0017, "step": 36342 }, { "epoch": 6.66, "learning_rate": 1.425810683703608e-06, "loss": 0.002, "step": 36343 }, { "epoch": 6.66, "learning_rate": 1.425428638693489e-06, "loss": 0.0009, "step": 36344 }, { "epoch": 6.67, "learning_rate": 1.4250466409466113e-06, "loss": 0.0007, "step": 36345 }, { "epoch": 6.67, "learning_rate": 1.4246646904650773e-06, "loss": 0.0017, "step": 36346 }, { "epoch": 6.67, "learning_rate": 1.4242827872509958e-06, "loss": 0.0019, "step": 36347 }, { "epoch": 6.67, "learning_rate": 1.4239009313064622e-06, "loss": 0.0011, "step": 36348 }, { "epoch": 6.67, "learning_rate": 1.423519122633591e-06, "loss": 0.0018, "step": 36349 }, { "epoch": 6.67, "learning_rate": 1.4231373612344823e-06, "loss": 0.0027, "step": 36350 }, { "epoch": 6.67, "learning_rate": 1.4227556471112424e-06, "loss": 0.0013, "step": 36351 }, { "epoch": 6.67, "learning_rate": 1.4223739802659743e-06, "loss": 0.0015, "step": 36352 }, { "epoch": 6.67, "learning_rate": 1.4219923607007812e-06, "loss": 0.0015, "step": 36353 }, { "epoch": 6.67, "learning_rate": 1.4216107884177677e-06, "loss": 0.0009, "step": 36354 }, { "epoch": 6.67, "learning_rate": 1.421229263419036e-06, "loss": 0.0027, "step": 36355 }, { "epoch": 6.67, "learning_rate": 1.4208477857066894e-06, "loss": 0.0013, "step": 36356 }, { "epoch": 6.67, "learning_rate": 1.420466355282829e-06, "loss": 0.0012, "step": 36357 }, { "epoch": 6.67, "learning_rate": 1.4200849721495634e-06, "loss": 0.0017, "step": 36358 }, { "epoch": 6.67, "learning_rate": 1.4197036363089879e-06, "loss": 0.0011, "step": 36359 }, { "epoch": 6.67, "learning_rate": 1.4193223477632078e-06, "loss": 0.0018, "step": 36360 }, { "epoch": 6.67, "learning_rate": 1.4189411065143232e-06, "loss": 0.002, "step": 36361 }, { "epoch": 6.67, "learning_rate": 1.4185599125644345e-06, "loss": 0.0022, "step": 36362 }, { "epoch": 6.67, "learning_rate": 1.4181787659156466e-06, "loss": 0.0027, "step": 36363 }, { "epoch": 6.67, "learning_rate": 1.4177976665700576e-06, "loss": 0.0008, "step": 36364 }, { "epoch": 6.67, "learning_rate": 1.417416614529773e-06, "loss": 0.0016, "step": 36365 }, { "epoch": 6.67, "learning_rate": 1.417035609796883e-06, "loss": 0.0015, "step": 36366 }, { "epoch": 6.67, "learning_rate": 1.416654652373497e-06, "loss": 0.0018, "step": 36367 }, { "epoch": 6.67, "learning_rate": 1.4162737422617113e-06, "loss": 0.0014, "step": 36368 }, { "epoch": 6.67, "learning_rate": 1.4158928794636251e-06, "loss": 0.0032, "step": 36369 }, { "epoch": 6.67, "learning_rate": 1.4155120639813392e-06, "loss": 0.0022, "step": 36370 }, { "epoch": 6.67, "learning_rate": 1.4151312958169516e-06, "loss": 0.0011, "step": 36371 }, { "epoch": 6.67, "learning_rate": 1.414750574972562e-06, "loss": 0.0014, "step": 36372 }, { "epoch": 6.67, "learning_rate": 1.4143699014502676e-06, "loss": 0.0025, "step": 36373 }, { "epoch": 6.67, "learning_rate": 1.4139892752521688e-06, "loss": 0.0017, "step": 36374 }, { "epoch": 6.67, "learning_rate": 1.4136086963803586e-06, "loss": 0.0011, "step": 36375 }, { "epoch": 6.67, "learning_rate": 1.413228164836944e-06, "loss": 0.0014, "step": 36376 }, { "epoch": 6.67, "learning_rate": 1.4128476806240132e-06, "loss": 0.002, "step": 36377 }, { "epoch": 6.67, "learning_rate": 1.4124672437436681e-06, "loss": 0.0021, "step": 36378 }, { "epoch": 6.67, "learning_rate": 1.4120868541980026e-06, "loss": 0.0023, "step": 36379 }, { "epoch": 6.67, "learning_rate": 1.4117065119891183e-06, "loss": 0.0018, "step": 36380 }, { "epoch": 6.67, "learning_rate": 1.4113262171191077e-06, "loss": 0.0021, "step": 36381 }, { "epoch": 6.67, "learning_rate": 1.4109459695900685e-06, "loss": 0.0021, "step": 36382 }, { "epoch": 6.67, "learning_rate": 1.4105657694040986e-06, "loss": 0.0016, "step": 36383 }, { "epoch": 6.67, "learning_rate": 1.4101856165632877e-06, "loss": 0.003, "step": 36384 }, { "epoch": 6.67, "learning_rate": 1.4098055110697362e-06, "loss": 0.0017, "step": 36385 }, { "epoch": 6.67, "learning_rate": 1.4094254529255391e-06, "loss": 0.0015, "step": 36386 }, { "epoch": 6.67, "learning_rate": 1.4090454421327892e-06, "loss": 0.002, "step": 36387 }, { "epoch": 6.67, "learning_rate": 1.4086654786935826e-06, "loss": 0.0026, "step": 36388 }, { "epoch": 6.67, "learning_rate": 1.4082855626100122e-06, "loss": 0.002, "step": 36389 }, { "epoch": 6.67, "learning_rate": 1.407905693884174e-06, "loss": 0.0022, "step": 36390 }, { "epoch": 6.67, "learning_rate": 1.4075258725181607e-06, "loss": 0.0019, "step": 36391 }, { "epoch": 6.67, "learning_rate": 1.4071460985140662e-06, "loss": 0.0019, "step": 36392 }, { "epoch": 6.67, "learning_rate": 1.406766371873981e-06, "loss": 0.002, "step": 36393 }, { "epoch": 6.67, "learning_rate": 1.4063866926000057e-06, "loss": 0.0011, "step": 36394 }, { "epoch": 6.67, "learning_rate": 1.4060070606942256e-06, "loss": 0.0019, "step": 36395 }, { "epoch": 6.67, "learning_rate": 1.4056274761587364e-06, "loss": 0.0026, "step": 36396 }, { "epoch": 6.67, "learning_rate": 1.4052479389956275e-06, "loss": 0.0027, "step": 36397 }, { "epoch": 6.67, "learning_rate": 1.4048684492069963e-06, "loss": 0.0022, "step": 36398 }, { "epoch": 6.68, "learning_rate": 1.4044890067949302e-06, "loss": 0.0043, "step": 36399 }, { "epoch": 6.68, "learning_rate": 1.404109611761526e-06, "loss": 0.0011, "step": 36400 }, { "epoch": 6.68, "learning_rate": 1.4037302641088646e-06, "loss": 0.0023, "step": 36401 }, { "epoch": 6.68, "learning_rate": 1.4033509638390474e-06, "loss": 0.0017, "step": 36402 }, { "epoch": 6.68, "learning_rate": 1.4029717109541596e-06, "loss": 0.0016, "step": 36403 }, { "epoch": 6.68, "learning_rate": 1.4025925054562938e-06, "loss": 0.0018, "step": 36404 }, { "epoch": 6.68, "learning_rate": 1.4022133473475397e-06, "loss": 0.0019, "step": 36405 }, { "epoch": 6.68, "learning_rate": 1.4018342366299864e-06, "loss": 0.0011, "step": 36406 }, { "epoch": 6.68, "learning_rate": 1.4014551733057246e-06, "loss": 0.0021, "step": 36407 }, { "epoch": 6.68, "learning_rate": 1.401076157376844e-06, "loss": 0.0018, "step": 36408 }, { "epoch": 6.68, "learning_rate": 1.4006971888454324e-06, "loss": 0.0014, "step": 36409 }, { "epoch": 6.68, "learning_rate": 1.4003182677135774e-06, "loss": 0.0024, "step": 36410 }, { "epoch": 6.68, "learning_rate": 1.3999393939833716e-06, "loss": 0.002, "step": 36411 }, { "epoch": 6.68, "learning_rate": 1.3995605676569036e-06, "loss": 0.0017, "step": 36412 }, { "epoch": 6.68, "learning_rate": 1.3991817887362568e-06, "loss": 0.0028, "step": 36413 }, { "epoch": 6.68, "learning_rate": 1.3988030572235212e-06, "loss": 0.0013, "step": 36414 }, { "epoch": 6.68, "learning_rate": 1.3984243731207835e-06, "loss": 0.0019, "step": 36415 }, { "epoch": 6.68, "learning_rate": 1.3980457364301336e-06, "loss": 0.0018, "step": 36416 }, { "epoch": 6.68, "learning_rate": 1.3976671471536574e-06, "loss": 0.0028, "step": 36417 }, { "epoch": 6.68, "learning_rate": 1.3972886052934431e-06, "loss": 0.0018, "step": 36418 }, { "epoch": 6.68, "learning_rate": 1.3969101108515715e-06, "loss": 0.0018, "step": 36419 }, { "epoch": 6.68, "learning_rate": 1.3965316638301351e-06, "loss": 0.0028, "step": 36420 }, { "epoch": 6.68, "learning_rate": 1.3961532642312181e-06, "loss": 0.0033, "step": 36421 }, { "epoch": 6.68, "learning_rate": 1.3957749120569052e-06, "loss": 0.0018, "step": 36422 }, { "epoch": 6.68, "learning_rate": 1.3953966073092827e-06, "loss": 0.002, "step": 36423 }, { "epoch": 6.68, "learning_rate": 1.3950183499904368e-06, "loss": 0.0015, "step": 36424 }, { "epoch": 6.68, "learning_rate": 1.39464014010245e-06, "loss": 0.0027, "step": 36425 }, { "epoch": 6.68, "learning_rate": 1.3942619776474086e-06, "loss": 0.0015, "step": 36426 }, { "epoch": 6.68, "learning_rate": 1.3938838626273975e-06, "loss": 0.0025, "step": 36427 }, { "epoch": 6.68, "learning_rate": 1.3935057950444974e-06, "loss": 0.0018, "step": 36428 }, { "epoch": 6.68, "learning_rate": 1.3931277749007977e-06, "loss": 0.0011, "step": 36429 }, { "epoch": 6.68, "learning_rate": 1.3927498021983809e-06, "loss": 0.0018, "step": 36430 }, { "epoch": 6.68, "learning_rate": 1.3923718769393268e-06, "loss": 0.0024, "step": 36431 }, { "epoch": 6.68, "learning_rate": 1.391993999125718e-06, "loss": 0.0031, "step": 36432 }, { "epoch": 6.68, "learning_rate": 1.391616168759643e-06, "loss": 0.0017, "step": 36433 }, { "epoch": 6.68, "learning_rate": 1.391238385843181e-06, "loss": 0.0016, "step": 36434 }, { "epoch": 6.68, "learning_rate": 1.390860650378414e-06, "loss": 0.001, "step": 36435 }, { "epoch": 6.68, "learning_rate": 1.3904829623674275e-06, "loss": 0.0013, "step": 36436 }, { "epoch": 6.68, "learning_rate": 1.390105321812296e-06, "loss": 0.002, "step": 36437 }, { "epoch": 6.68, "learning_rate": 1.389727728715109e-06, "loss": 0.0018, "step": 36438 }, { "epoch": 6.68, "learning_rate": 1.3893501830779431e-06, "loss": 0.002, "step": 36439 }, { "epoch": 6.68, "learning_rate": 1.3889726849028818e-06, "loss": 0.0035, "step": 36440 }, { "epoch": 6.68, "learning_rate": 1.388595234192004e-06, "loss": 0.003, "step": 36441 }, { "epoch": 6.68, "learning_rate": 1.3882178309473915e-06, "loss": 0.001, "step": 36442 }, { "epoch": 6.68, "learning_rate": 1.3878404751711238e-06, "loss": 0.0019, "step": 36443 }, { "epoch": 6.68, "learning_rate": 1.3874631668652817e-06, "loss": 0.0015, "step": 36444 }, { "epoch": 6.68, "learning_rate": 1.387085906031941e-06, "loss": 0.0023, "step": 36445 }, { "epoch": 6.68, "learning_rate": 1.3867086926731877e-06, "loss": 0.0015, "step": 36446 }, { "epoch": 6.68, "learning_rate": 1.3863315267911004e-06, "loss": 0.0019, "step": 36447 }, { "epoch": 6.68, "learning_rate": 1.385954408387752e-06, "loss": 0.0014, "step": 36448 }, { "epoch": 6.68, "learning_rate": 1.3855773374652258e-06, "loss": 0.0025, "step": 36449 }, { "epoch": 6.68, "learning_rate": 1.3852003140255965e-06, "loss": 0.0013, "step": 36450 }, { "epoch": 6.68, "learning_rate": 1.3848233380709475e-06, "loss": 0.0022, "step": 36451 }, { "epoch": 6.68, "learning_rate": 1.3844464096033538e-06, "loss": 0.0018, "step": 36452 }, { "epoch": 6.68, "learning_rate": 1.3840695286248929e-06, "loss": 0.0011, "step": 36453 }, { "epoch": 6.69, "learning_rate": 1.383692695137643e-06, "loss": 0.0033, "step": 36454 }, { "epoch": 6.69, "learning_rate": 1.3833159091436809e-06, "loss": 0.0012, "step": 36455 }, { "epoch": 6.69, "learning_rate": 1.3829391706450835e-06, "loss": 0.0017, "step": 36456 }, { "epoch": 6.69, "learning_rate": 1.3825624796439274e-06, "loss": 0.001, "step": 36457 }, { "epoch": 6.69, "learning_rate": 1.382185836142288e-06, "loss": 0.0018, "step": 36458 }, { "epoch": 6.69, "learning_rate": 1.3818092401422424e-06, "loss": 0.0013, "step": 36459 }, { "epoch": 6.69, "learning_rate": 1.3814326916458666e-06, "loss": 0.0009, "step": 36460 }, { "epoch": 6.69, "learning_rate": 1.3810561906552356e-06, "loss": 0.0011, "step": 36461 }, { "epoch": 6.69, "learning_rate": 1.3806797371724246e-06, "loss": 0.0018, "step": 36462 }, { "epoch": 6.69, "learning_rate": 1.3803033311995072e-06, "loss": 0.0016, "step": 36463 }, { "epoch": 6.69, "learning_rate": 1.3799269727385612e-06, "loss": 0.0012, "step": 36464 }, { "epoch": 6.69, "learning_rate": 1.3795506617916621e-06, "loss": 0.0027, "step": 36465 }, { "epoch": 6.69, "learning_rate": 1.3791743983608796e-06, "loss": 0.001, "step": 36466 }, { "epoch": 6.69, "learning_rate": 1.3787981824482899e-06, "loss": 0.0012, "step": 36467 }, { "epoch": 6.69, "learning_rate": 1.3784220140559646e-06, "loss": 0.0011, "step": 36468 }, { "epoch": 6.69, "learning_rate": 1.3780458931859808e-06, "loss": 0.0013, "step": 36469 }, { "epoch": 6.69, "learning_rate": 1.3776698198404103e-06, "loss": 0.0014, "step": 36470 }, { "epoch": 6.69, "learning_rate": 1.377293794021326e-06, "loss": 0.0026, "step": 36471 }, { "epoch": 6.69, "learning_rate": 1.3769178157307993e-06, "loss": 0.002, "step": 36472 }, { "epoch": 6.69, "learning_rate": 1.3765418849709045e-06, "loss": 0.001, "step": 36473 }, { "epoch": 6.69, "learning_rate": 1.3761660017437128e-06, "loss": 0.0014, "step": 36474 }, { "epoch": 6.69, "learning_rate": 1.3757901660512962e-06, "loss": 0.002, "step": 36475 }, { "epoch": 6.69, "learning_rate": 1.3754143778957242e-06, "loss": 0.0018, "step": 36476 }, { "epoch": 6.69, "learning_rate": 1.3750386372790737e-06, "loss": 0.0013, "step": 36477 }, { "epoch": 6.69, "learning_rate": 1.374662944203411e-06, "loss": 0.0017, "step": 36478 }, { "epoch": 6.69, "learning_rate": 1.3742872986708077e-06, "loss": 0.0019, "step": 36479 }, { "epoch": 6.69, "learning_rate": 1.3739117006833347e-06, "loss": 0.0019, "step": 36480 }, { "epoch": 6.69, "learning_rate": 1.373536150243061e-06, "loss": 0.0015, "step": 36481 }, { "epoch": 6.69, "learning_rate": 1.3731606473520598e-06, "loss": 0.0027, "step": 36482 }, { "epoch": 6.69, "learning_rate": 1.3727851920124025e-06, "loss": 0.0014, "step": 36483 }, { "epoch": 6.69, "learning_rate": 1.372409784226152e-06, "loss": 0.0012, "step": 36484 }, { "epoch": 6.69, "learning_rate": 1.3720344239953787e-06, "loss": 0.0008, "step": 36485 }, { "epoch": 6.69, "learning_rate": 1.3716591113221566e-06, "loss": 0.001, "step": 36486 }, { "epoch": 6.69, "learning_rate": 1.371283846208551e-06, "loss": 0.0008, "step": 36487 }, { "epoch": 6.69, "learning_rate": 1.3709086286566308e-06, "loss": 0.0034, "step": 36488 }, { "epoch": 6.69, "learning_rate": 1.370533458668465e-06, "loss": 0.0024, "step": 36489 }, { "epoch": 6.69, "learning_rate": 1.3701583362461212e-06, "loss": 0.0013, "step": 36490 }, { "epoch": 6.69, "learning_rate": 1.369783261391666e-06, "loss": 0.0027, "step": 36491 }, { "epoch": 6.69, "learning_rate": 1.3694082341071679e-06, "loss": 0.001, "step": 36492 }, { "epoch": 6.69, "learning_rate": 1.369033254394695e-06, "loss": 0.0021, "step": 36493 }, { "epoch": 6.69, "learning_rate": 1.368658322256311e-06, "loss": 0.0016, "step": 36494 }, { "epoch": 6.69, "learning_rate": 1.3682834376940857e-06, "loss": 0.0016, "step": 36495 }, { "epoch": 6.69, "learning_rate": 1.367908600710084e-06, "loss": 0.0006, "step": 36496 }, { "epoch": 6.69, "learning_rate": 1.367533811306373e-06, "loss": 0.0021, "step": 36497 }, { "epoch": 6.69, "learning_rate": 1.3671590694850145e-06, "loss": 0.003, "step": 36498 }, { "epoch": 6.69, "learning_rate": 1.3667843752480802e-06, "loss": 0.0019, "step": 36499 }, { "epoch": 6.69, "learning_rate": 1.366409728597633e-06, "loss": 0.0019, "step": 36500 }, { "epoch": 6.69, "learning_rate": 1.3660351295357398e-06, "loss": 0.0025, "step": 36501 }, { "epoch": 6.69, "learning_rate": 1.3656605780644605e-06, "loss": 0.002, "step": 36502 }, { "epoch": 6.69, "learning_rate": 1.3652860741858608e-06, "loss": 0.0019, "step": 36503 }, { "epoch": 6.69, "learning_rate": 1.3649116179020083e-06, "loss": 0.0012, "step": 36504 }, { "epoch": 6.69, "learning_rate": 1.3645372092149655e-06, "loss": 0.0019, "step": 36505 }, { "epoch": 6.69, "learning_rate": 1.3641628481267965e-06, "loss": 0.001, "step": 36506 }, { "epoch": 6.69, "learning_rate": 1.363788534639563e-06, "loss": 0.0021, "step": 36507 }, { "epoch": 6.7, "learning_rate": 1.3634142687553298e-06, "loss": 0.0017, "step": 36508 }, { "epoch": 6.7, "learning_rate": 1.36304005047616e-06, "loss": 0.0016, "step": 36509 }, { "epoch": 6.7, "learning_rate": 1.3626658798041147e-06, "loss": 0.003, "step": 36510 }, { "epoch": 6.7, "learning_rate": 1.3622917567412587e-06, "loss": 0.0018, "step": 36511 }, { "epoch": 6.7, "learning_rate": 1.3619176812896528e-06, "loss": 0.0024, "step": 36512 }, { "epoch": 6.7, "learning_rate": 1.361543653451358e-06, "loss": 0.0012, "step": 36513 }, { "epoch": 6.7, "learning_rate": 1.3611696732284374e-06, "loss": 0.0021, "step": 36514 }, { "epoch": 6.7, "learning_rate": 1.3607957406229522e-06, "loss": 0.0027, "step": 36515 }, { "epoch": 6.7, "learning_rate": 1.3604218556369608e-06, "loss": 0.0016, "step": 36516 }, { "epoch": 6.7, "learning_rate": 1.3600480182725295e-06, "loss": 0.0015, "step": 36517 }, { "epoch": 6.7, "learning_rate": 1.3596742285317143e-06, "loss": 0.0011, "step": 36518 }, { "epoch": 6.7, "learning_rate": 1.3593004864165805e-06, "loss": 0.0015, "step": 36519 }, { "epoch": 6.7, "learning_rate": 1.3589267919291826e-06, "loss": 0.0023, "step": 36520 }, { "epoch": 6.7, "learning_rate": 1.3585531450715806e-06, "loss": 0.0021, "step": 36521 }, { "epoch": 6.7, "learning_rate": 1.358179545845838e-06, "loss": 0.0021, "step": 36522 }, { "epoch": 6.7, "learning_rate": 1.3578059942540124e-06, "loss": 0.0011, "step": 36523 }, { "epoch": 6.7, "learning_rate": 1.3574324902981628e-06, "loss": 0.0018, "step": 36524 }, { "epoch": 6.7, "learning_rate": 1.3570590339803469e-06, "loss": 0.0017, "step": 36525 }, { "epoch": 6.7, "learning_rate": 1.356685625302625e-06, "loss": 0.0015, "step": 36526 }, { "epoch": 6.7, "learning_rate": 1.3563122642670545e-06, "loss": 0.001, "step": 36527 }, { "epoch": 6.7, "learning_rate": 1.3559389508756926e-06, "loss": 0.0035, "step": 36528 }, { "epoch": 6.7, "learning_rate": 1.3555656851305955e-06, "loss": 0.0008, "step": 36529 }, { "epoch": 6.7, "learning_rate": 1.3551924670338278e-06, "loss": 0.0025, "step": 36530 }, { "epoch": 6.7, "learning_rate": 1.3548192965874396e-06, "loss": 0.0015, "step": 36531 }, { "epoch": 6.7, "learning_rate": 1.3544461737934899e-06, "loss": 0.0029, "step": 36532 }, { "epoch": 6.7, "learning_rate": 1.3540730986540351e-06, "loss": 0.003, "step": 36533 }, { "epoch": 6.7, "learning_rate": 1.35370007117113e-06, "loss": 0.0026, "step": 36534 }, { "epoch": 6.7, "learning_rate": 1.3533270913468343e-06, "loss": 0.0009, "step": 36535 }, { "epoch": 6.7, "learning_rate": 1.352954159183203e-06, "loss": 0.0019, "step": 36536 }, { "epoch": 6.7, "learning_rate": 1.3525812746822931e-06, "loss": 0.0017, "step": 36537 }, { "epoch": 6.7, "learning_rate": 1.352208437846153e-06, "loss": 0.0017, "step": 36538 }, { "epoch": 6.7, "learning_rate": 1.3518356486768446e-06, "loss": 0.002, "step": 36539 }, { "epoch": 6.7, "learning_rate": 1.3514629071764208e-06, "loss": 0.0025, "step": 36540 }, { "epoch": 6.7, "learning_rate": 1.351090213346936e-06, "loss": 0.0023, "step": 36541 }, { "epoch": 6.7, "learning_rate": 1.350717567190445e-06, "loss": 0.0026, "step": 36542 }, { "epoch": 6.7, "learning_rate": 1.350344968709002e-06, "loss": 0.0019, "step": 36543 }, { "epoch": 6.7, "learning_rate": 1.3499724179046591e-06, "loss": 0.0011, "step": 36544 }, { "epoch": 6.7, "learning_rate": 1.3495999147794713e-06, "loss": 0.002, "step": 36545 }, { "epoch": 6.7, "learning_rate": 1.3492274593354914e-06, "loss": 0.0026, "step": 36546 }, { "epoch": 6.7, "learning_rate": 1.34885505157477e-06, "loss": 0.0014, "step": 36547 }, { "epoch": 6.7, "learning_rate": 1.3484826914993665e-06, "loss": 0.0017, "step": 36548 }, { "epoch": 6.7, "learning_rate": 1.348110379111327e-06, "loss": 0.0012, "step": 36549 }, { "epoch": 6.7, "learning_rate": 1.3477381144127067e-06, "loss": 0.0025, "step": 36550 }, { "epoch": 6.7, "learning_rate": 1.3473658974055537e-06, "loss": 0.0015, "step": 36551 }, { "epoch": 6.7, "learning_rate": 1.3469937280919243e-06, "loss": 0.0011, "step": 36552 }, { "epoch": 6.7, "learning_rate": 1.3466216064738692e-06, "loss": 0.0021, "step": 36553 }, { "epoch": 6.7, "learning_rate": 1.3462495325534397e-06, "loss": 0.0024, "step": 36554 }, { "epoch": 6.7, "learning_rate": 1.3458775063326845e-06, "loss": 0.0009, "step": 36555 }, { "epoch": 6.7, "learning_rate": 1.345505527813652e-06, "loss": 0.0018, "step": 36556 }, { "epoch": 6.7, "learning_rate": 1.345133596998398e-06, "loss": 0.0022, "step": 36557 }, { "epoch": 6.7, "learning_rate": 1.3447617138889712e-06, "loss": 0.0028, "step": 36558 }, { "epoch": 6.7, "learning_rate": 1.34438987848742e-06, "loss": 0.0031, "step": 36559 }, { "epoch": 6.7, "learning_rate": 1.3440180907957944e-06, "loss": 0.0021, "step": 36560 }, { "epoch": 6.7, "learning_rate": 1.3436463508161446e-06, "loss": 0.0018, "step": 36561 }, { "epoch": 6.7, "learning_rate": 1.3432746585505186e-06, "loss": 0.0029, "step": 36562 }, { "epoch": 6.71, "learning_rate": 1.3429030140009646e-06, "loss": 0.0012, "step": 36563 }, { "epoch": 6.71, "learning_rate": 1.3425314171695326e-06, "loss": 0.0013, "step": 36564 }, { "epoch": 6.71, "learning_rate": 1.3421598680582682e-06, "loss": 0.0024, "step": 36565 }, { "epoch": 6.71, "learning_rate": 1.3417883666692254e-06, "loss": 0.0013, "step": 36566 }, { "epoch": 6.71, "learning_rate": 1.3414169130044463e-06, "loss": 0.0029, "step": 36567 }, { "epoch": 6.71, "learning_rate": 1.34104550706598e-06, "loss": 0.0019, "step": 36568 }, { "epoch": 6.71, "learning_rate": 1.3406741488558717e-06, "loss": 0.0023, "step": 36569 }, { "epoch": 6.71, "learning_rate": 1.3403028383761729e-06, "loss": 0.0013, "step": 36570 }, { "epoch": 6.71, "learning_rate": 1.3399315756289266e-06, "loss": 0.0017, "step": 36571 }, { "epoch": 6.71, "learning_rate": 1.3395603606161833e-06, "loss": 0.0023, "step": 36572 }, { "epoch": 6.71, "learning_rate": 1.3391891933399813e-06, "loss": 0.0007, "step": 36573 }, { "epoch": 6.71, "learning_rate": 1.3388180738023738e-06, "loss": 0.0024, "step": 36574 }, { "epoch": 6.71, "learning_rate": 1.338447002005404e-06, "loss": 0.0025, "step": 36575 }, { "epoch": 6.71, "learning_rate": 1.3380759779511176e-06, "loss": 0.0019, "step": 36576 }, { "epoch": 6.71, "learning_rate": 1.3377050016415582e-06, "loss": 0.0014, "step": 36577 }, { "epoch": 6.71, "learning_rate": 1.3373340730787731e-06, "loss": 0.0022, "step": 36578 }, { "epoch": 6.71, "learning_rate": 1.3369631922648051e-06, "loss": 0.0013, "step": 36579 }, { "epoch": 6.71, "learning_rate": 1.3365923592016982e-06, "loss": 0.0007, "step": 36580 }, { "epoch": 6.71, "learning_rate": 1.3362215738914986e-06, "loss": 0.0017, "step": 36581 }, { "epoch": 6.71, "learning_rate": 1.3358508363362456e-06, "loss": 0.002, "step": 36582 }, { "epoch": 6.71, "learning_rate": 1.3354801465379874e-06, "loss": 0.0013, "step": 36583 }, { "epoch": 6.71, "learning_rate": 1.3351095044987683e-06, "loss": 0.0025, "step": 36584 }, { "epoch": 6.71, "learning_rate": 1.3347389102206265e-06, "loss": 0.0013, "step": 36585 }, { "epoch": 6.71, "learning_rate": 1.334368363705607e-06, "loss": 0.0012, "step": 36586 }, { "epoch": 6.71, "learning_rate": 1.3339978649557506e-06, "loss": 0.0013, "step": 36587 }, { "epoch": 6.71, "learning_rate": 1.333627413973102e-06, "loss": 0.0016, "step": 36588 }, { "epoch": 6.71, "learning_rate": 1.333257010759702e-06, "loss": 0.002, "step": 36589 }, { "epoch": 6.71, "learning_rate": 1.3328866553175956e-06, "loss": 0.0012, "step": 36590 }, { "epoch": 6.71, "learning_rate": 1.3325163476488157e-06, "loss": 0.0011, "step": 36591 }, { "epoch": 6.71, "learning_rate": 1.3321460877554116e-06, "loss": 0.002, "step": 36592 }, { "epoch": 6.71, "learning_rate": 1.3317758756394217e-06, "loss": 0.0015, "step": 36593 }, { "epoch": 6.71, "learning_rate": 1.3314057113028856e-06, "loss": 0.0012, "step": 36594 }, { "epoch": 6.71, "learning_rate": 1.3310355947478437e-06, "loss": 0.002, "step": 36595 }, { "epoch": 6.71, "learning_rate": 1.3306655259763378e-06, "loss": 0.0031, "step": 36596 }, { "epoch": 6.71, "learning_rate": 1.3302955049904053e-06, "loss": 0.0009, "step": 36597 }, { "epoch": 6.71, "learning_rate": 1.329925531792089e-06, "loss": 0.0022, "step": 36598 }, { "epoch": 6.71, "learning_rate": 1.3295556063834248e-06, "loss": 0.001, "step": 36599 }, { "epoch": 6.71, "learning_rate": 1.3291857287664511e-06, "loss": 0.0019, "step": 36600 }, { "epoch": 6.71, "learning_rate": 1.328815898943213e-06, "loss": 0.0009, "step": 36601 }, { "epoch": 6.71, "learning_rate": 1.3284461169157437e-06, "loss": 0.0021, "step": 36602 }, { "epoch": 6.71, "learning_rate": 1.328076382686082e-06, "loss": 0.0016, "step": 36603 }, { "epoch": 6.71, "learning_rate": 1.3277066962562647e-06, "loss": 0.0018, "step": 36604 }, { "epoch": 6.71, "learning_rate": 1.327337057628333e-06, "loss": 0.0021, "step": 36605 }, { "epoch": 6.71, "learning_rate": 1.326967466804322e-06, "loss": 0.0008, "step": 36606 }, { "epoch": 6.71, "learning_rate": 1.3265979237862703e-06, "loss": 0.0023, "step": 36607 }, { "epoch": 6.71, "learning_rate": 1.326228428576216e-06, "loss": 0.0028, "step": 36608 }, { "epoch": 6.71, "learning_rate": 1.32585898117619e-06, "loss": 0.0021, "step": 36609 }, { "epoch": 6.71, "learning_rate": 1.325489581588234e-06, "loss": 0.0015, "step": 36610 }, { "epoch": 6.71, "learning_rate": 1.3251202298143828e-06, "loss": 0.0013, "step": 36611 }, { "epoch": 6.71, "learning_rate": 1.3247509258566726e-06, "loss": 0.0015, "step": 36612 }, { "epoch": 6.71, "learning_rate": 1.3243816697171373e-06, "loss": 0.0017, "step": 36613 }, { "epoch": 6.71, "learning_rate": 1.3240124613978145e-06, "loss": 0.0033, "step": 36614 }, { "epoch": 6.71, "learning_rate": 1.3236433009007377e-06, "loss": 0.0018, "step": 36615 }, { "epoch": 6.71, "learning_rate": 1.3232741882279432e-06, "loss": 0.0024, "step": 36616 }, { "epoch": 6.72, "learning_rate": 1.3229051233814637e-06, "loss": 0.0018, "step": 36617 }, { "epoch": 6.72, "learning_rate": 1.3225361063633324e-06, "loss": 0.0016, "step": 36618 }, { "epoch": 6.72, "learning_rate": 1.3221671371755907e-06, "loss": 0.0013, "step": 36619 }, { "epoch": 6.72, "learning_rate": 1.3217982158202635e-06, "loss": 0.0014, "step": 36620 }, { "epoch": 6.72, "learning_rate": 1.3214293422993885e-06, "loss": 0.0032, "step": 36621 }, { "epoch": 6.72, "learning_rate": 1.321060516614997e-06, "loss": 0.0013, "step": 36622 }, { "epoch": 6.72, "learning_rate": 1.3206917387691243e-06, "loss": 0.0012, "step": 36623 }, { "epoch": 6.72, "learning_rate": 1.320323008763802e-06, "loss": 0.0021, "step": 36624 }, { "epoch": 6.72, "learning_rate": 1.3199543266010628e-06, "loss": 0.0013, "step": 36625 }, { "epoch": 6.72, "learning_rate": 1.3195856922829387e-06, "loss": 0.0021, "step": 36626 }, { "epoch": 6.72, "learning_rate": 1.3192171058114621e-06, "loss": 0.0017, "step": 36627 }, { "epoch": 6.72, "learning_rate": 1.3188485671886642e-06, "loss": 0.0017, "step": 36628 }, { "epoch": 6.72, "learning_rate": 1.3184800764165762e-06, "loss": 0.0012, "step": 36629 }, { "epoch": 6.72, "learning_rate": 1.3181116334972288e-06, "loss": 0.0023, "step": 36630 }, { "epoch": 6.72, "learning_rate": 1.3177432384326539e-06, "loss": 0.0036, "step": 36631 }, { "epoch": 6.72, "learning_rate": 1.317374891224882e-06, "loss": 0.0015, "step": 36632 }, { "epoch": 6.72, "learning_rate": 1.3170065918759423e-06, "loss": 0.0017, "step": 36633 }, { "epoch": 6.72, "learning_rate": 1.3166383403878657e-06, "loss": 0.001, "step": 36634 }, { "epoch": 6.72, "learning_rate": 1.3162701367626806e-06, "loss": 0.002, "step": 36635 }, { "epoch": 6.72, "learning_rate": 1.3159019810024198e-06, "loss": 0.0015, "step": 36636 }, { "epoch": 6.72, "learning_rate": 1.3155338731091128e-06, "loss": 0.0014, "step": 36637 }, { "epoch": 6.72, "learning_rate": 1.3151658130847845e-06, "loss": 0.0012, "step": 36638 }, { "epoch": 6.72, "learning_rate": 1.3147978009314656e-06, "loss": 0.002, "step": 36639 }, { "epoch": 6.72, "learning_rate": 1.3144298366511831e-06, "loss": 0.0011, "step": 36640 }, { "epoch": 6.72, "learning_rate": 1.3140619202459682e-06, "loss": 0.0019, "step": 36641 }, { "epoch": 6.72, "learning_rate": 1.3136940517178476e-06, "loss": 0.0019, "step": 36642 }, { "epoch": 6.72, "learning_rate": 1.313326231068849e-06, "loss": 0.0022, "step": 36643 }, { "epoch": 6.72, "learning_rate": 1.3129584583010002e-06, "loss": 0.002, "step": 36644 }, { "epoch": 6.72, "learning_rate": 1.3125907334163267e-06, "loss": 0.0014, "step": 36645 }, { "epoch": 6.72, "learning_rate": 1.312223056416858e-06, "loss": 0.0015, "step": 36646 }, { "epoch": 6.72, "learning_rate": 1.3118554273046192e-06, "loss": 0.0019, "step": 36647 }, { "epoch": 6.72, "learning_rate": 1.311487846081636e-06, "loss": 0.0026, "step": 36648 }, { "epoch": 6.72, "learning_rate": 1.311120312749935e-06, "loss": 0.0019, "step": 36649 }, { "epoch": 6.72, "learning_rate": 1.3107528273115433e-06, "loss": 0.0022, "step": 36650 }, { "epoch": 6.72, "learning_rate": 1.310385389768486e-06, "loss": 0.0011, "step": 36651 }, { "epoch": 6.72, "learning_rate": 1.310018000122787e-06, "loss": 0.002, "step": 36652 }, { "epoch": 6.72, "learning_rate": 1.3096506583764713e-06, "loss": 0.0014, "step": 36653 }, { "epoch": 6.72, "learning_rate": 1.3092833645315662e-06, "loss": 0.0016, "step": 36654 }, { "epoch": 6.72, "learning_rate": 1.3089161185900966e-06, "loss": 0.0019, "step": 36655 }, { "epoch": 6.72, "learning_rate": 1.3085489205540824e-06, "loss": 0.003, "step": 36656 }, { "epoch": 6.72, "learning_rate": 1.3081817704255495e-06, "loss": 0.0016, "step": 36657 }, { "epoch": 6.72, "learning_rate": 1.307814668206524e-06, "loss": 0.0028, "step": 36658 }, { "epoch": 6.72, "learning_rate": 1.3074476138990266e-06, "loss": 0.0019, "step": 36659 }, { "epoch": 6.72, "learning_rate": 1.3070806075050824e-06, "loss": 0.0019, "step": 36660 }, { "epoch": 6.72, "learning_rate": 1.3067136490267162e-06, "loss": 0.0008, "step": 36661 }, { "epoch": 6.72, "learning_rate": 1.3063467384659433e-06, "loss": 0.0018, "step": 36662 }, { "epoch": 6.72, "learning_rate": 1.3059798758247933e-06, "loss": 0.0009, "step": 36663 }, { "epoch": 6.72, "learning_rate": 1.3056130611052852e-06, "loss": 0.0018, "step": 36664 }, { "epoch": 6.72, "learning_rate": 1.3052462943094423e-06, "loss": 0.0013, "step": 36665 }, { "epoch": 6.72, "learning_rate": 1.3048795754392852e-06, "loss": 0.0017, "step": 36666 }, { "epoch": 6.72, "learning_rate": 1.3045129044968352e-06, "loss": 0.0011, "step": 36667 }, { "epoch": 6.72, "learning_rate": 1.3041462814841145e-06, "loss": 0.0015, "step": 36668 }, { "epoch": 6.72, "learning_rate": 1.3037797064031421e-06, "loss": 0.0011, "step": 36669 }, { "epoch": 6.72, "learning_rate": 1.3034131792559413e-06, "loss": 0.0018, "step": 36670 }, { "epoch": 6.72, "learning_rate": 1.3030467000445269e-06, "loss": 0.0013, "step": 36671 }, { "epoch": 6.73, "learning_rate": 1.3026802687709262e-06, "loss": 0.0007, "step": 36672 }, { "epoch": 6.73, "learning_rate": 1.3023138854371576e-06, "loss": 0.0019, "step": 36673 }, { "epoch": 6.73, "learning_rate": 1.301947550045236e-06, "loss": 0.0015, "step": 36674 }, { "epoch": 6.73, "learning_rate": 1.3015812625971813e-06, "loss": 0.0018, "step": 36675 }, { "epoch": 6.73, "learning_rate": 1.301215023095017e-06, "loss": 0.0014, "step": 36676 }, { "epoch": 6.73, "learning_rate": 1.3008488315407597e-06, "loss": 0.0032, "step": 36677 }, { "epoch": 6.73, "learning_rate": 1.3004826879364263e-06, "loss": 0.0039, "step": 36678 }, { "epoch": 6.73, "learning_rate": 1.3001165922840365e-06, "loss": 0.0019, "step": 36679 }, { "epoch": 6.73, "learning_rate": 1.2997505445856085e-06, "loss": 0.0018, "step": 36680 }, { "epoch": 6.73, "learning_rate": 1.2993845448431587e-06, "loss": 0.0014, "step": 36681 }, { "epoch": 6.73, "learning_rate": 1.299018593058705e-06, "loss": 0.0014, "step": 36682 }, { "epoch": 6.73, "learning_rate": 1.2986526892342643e-06, "loss": 0.0018, "step": 36683 }, { "epoch": 6.73, "learning_rate": 1.2982868333718546e-06, "loss": 0.0016, "step": 36684 }, { "epoch": 6.73, "learning_rate": 1.2979210254734919e-06, "loss": 0.0031, "step": 36685 }, { "epoch": 6.73, "learning_rate": 1.2975552655411917e-06, "loss": 0.0009, "step": 36686 }, { "epoch": 6.73, "learning_rate": 1.297189553576972e-06, "loss": 0.0017, "step": 36687 }, { "epoch": 6.73, "learning_rate": 1.2968238895828433e-06, "loss": 0.0009, "step": 36688 }, { "epoch": 6.73, "learning_rate": 1.296458273560829e-06, "loss": 0.0027, "step": 36689 }, { "epoch": 6.73, "learning_rate": 1.2960927055129402e-06, "loss": 0.003, "step": 36690 }, { "epoch": 6.73, "learning_rate": 1.2957271854411934e-06, "loss": 0.0018, "step": 36691 }, { "epoch": 6.73, "learning_rate": 1.2953617133476016e-06, "loss": 0.0011, "step": 36692 }, { "epoch": 6.73, "learning_rate": 1.2949962892341772e-06, "loss": 0.0024, "step": 36693 }, { "epoch": 6.73, "learning_rate": 1.294630913102939e-06, "loss": 0.0017, "step": 36694 }, { "epoch": 6.73, "learning_rate": 1.2942655849558995e-06, "loss": 0.0022, "step": 36695 }, { "epoch": 6.73, "learning_rate": 1.2939003047950716e-06, "loss": 0.0014, "step": 36696 }, { "epoch": 6.73, "learning_rate": 1.2935350726224705e-06, "loss": 0.0019, "step": 36697 }, { "epoch": 6.73, "learning_rate": 1.2931698884401067e-06, "loss": 0.0013, "step": 36698 }, { "epoch": 6.73, "learning_rate": 1.2928047522499942e-06, "loss": 0.0019, "step": 36699 }, { "epoch": 6.73, "learning_rate": 1.292439664054147e-06, "loss": 0.0028, "step": 36700 }, { "epoch": 6.73, "learning_rate": 1.2920746238545735e-06, "loss": 0.0024, "step": 36701 }, { "epoch": 6.73, "learning_rate": 1.291709631653294e-06, "loss": 0.0022, "step": 36702 }, { "epoch": 6.73, "learning_rate": 1.2913446874523116e-06, "loss": 0.0017, "step": 36703 }, { "epoch": 6.73, "learning_rate": 1.2909797912536426e-06, "loss": 0.0026, "step": 36704 }, { "epoch": 6.73, "learning_rate": 1.2906149430592962e-06, "loss": 0.0018, "step": 36705 }, { "epoch": 6.73, "learning_rate": 1.2902501428712822e-06, "loss": 0.001, "step": 36706 }, { "epoch": 6.73, "learning_rate": 1.2898853906916154e-06, "loss": 0.0024, "step": 36707 }, { "epoch": 6.73, "learning_rate": 1.2895206865223065e-06, "loss": 0.0014, "step": 36708 }, { "epoch": 6.73, "learning_rate": 1.2891560303653616e-06, "loss": 0.0015, "step": 36709 }, { "epoch": 6.73, "learning_rate": 1.2887914222227905e-06, "loss": 0.0021, "step": 36710 }, { "epoch": 6.73, "learning_rate": 1.2884268620966078e-06, "loss": 0.0028, "step": 36711 }, { "epoch": 6.73, "learning_rate": 1.28806234998882e-06, "loss": 0.0021, "step": 36712 }, { "epoch": 6.73, "learning_rate": 1.2876978859014366e-06, "loss": 0.0021, "step": 36713 }, { "epoch": 6.73, "learning_rate": 1.287333469836467e-06, "loss": 0.0018, "step": 36714 }, { "epoch": 6.73, "learning_rate": 1.2869691017959195e-06, "loss": 0.0021, "step": 36715 }, { "epoch": 6.73, "learning_rate": 1.2866047817818016e-06, "loss": 0.0025, "step": 36716 }, { "epoch": 6.73, "learning_rate": 1.2862405097961228e-06, "loss": 0.0011, "step": 36717 }, { "epoch": 6.73, "learning_rate": 1.2858762858408902e-06, "loss": 0.0016, "step": 36718 }, { "epoch": 6.73, "learning_rate": 1.28551210991811e-06, "loss": 0.001, "step": 36719 }, { "epoch": 6.73, "learning_rate": 1.2851479820297952e-06, "loss": 0.0018, "step": 36720 }, { "epoch": 6.73, "learning_rate": 1.2847839021779462e-06, "loss": 0.0015, "step": 36721 }, { "epoch": 6.73, "learning_rate": 1.2844198703645727e-06, "loss": 0.0012, "step": 36722 }, { "epoch": 6.73, "learning_rate": 1.284055886591682e-06, "loss": 0.0019, "step": 36723 }, { "epoch": 6.73, "learning_rate": 1.2836919508612766e-06, "loss": 0.0018, "step": 36724 }, { "epoch": 6.73, "learning_rate": 1.2833280631753675e-06, "loss": 0.0017, "step": 36725 }, { "epoch": 6.74, "learning_rate": 1.2829642235359607e-06, "loss": 0.001, "step": 36726 }, { "epoch": 6.74, "learning_rate": 1.2826004319450569e-06, "loss": 0.0011, "step": 36727 }, { "epoch": 6.74, "learning_rate": 1.2822366884046622e-06, "loss": 0.002, "step": 36728 }, { "epoch": 6.74, "learning_rate": 1.2818729929167838e-06, "loss": 0.0019, "step": 36729 }, { "epoch": 6.74, "learning_rate": 1.2815093454834272e-06, "loss": 0.0019, "step": 36730 }, { "epoch": 6.74, "learning_rate": 1.2811457461065947e-06, "loss": 0.0007, "step": 36731 }, { "epoch": 6.74, "learning_rate": 1.2807821947882915e-06, "loss": 0.002, "step": 36732 }, { "epoch": 6.74, "learning_rate": 1.2804186915305207e-06, "loss": 0.002, "step": 36733 }, { "epoch": 6.74, "learning_rate": 1.2800552363352868e-06, "loss": 0.0017, "step": 36734 }, { "epoch": 6.74, "learning_rate": 1.2796918292045922e-06, "loss": 0.0025, "step": 36735 }, { "epoch": 6.74, "learning_rate": 1.2793284701404407e-06, "loss": 0.0013, "step": 36736 }, { "epoch": 6.74, "learning_rate": 1.2789651591448326e-06, "loss": 0.0015, "step": 36737 }, { "epoch": 6.74, "learning_rate": 1.2786018962197778e-06, "loss": 0.0012, "step": 36738 }, { "epoch": 6.74, "learning_rate": 1.2782386813672709e-06, "loss": 0.0015, "step": 36739 }, { "epoch": 6.74, "learning_rate": 1.2778755145893174e-06, "loss": 0.002, "step": 36740 }, { "epoch": 6.74, "learning_rate": 1.2775123958879154e-06, "loss": 0.0015, "step": 36741 }, { "epoch": 6.74, "learning_rate": 1.2771493252650723e-06, "loss": 0.0013, "step": 36742 }, { "epoch": 6.74, "learning_rate": 1.2767863027227856e-06, "loss": 0.0018, "step": 36743 }, { "epoch": 6.74, "learning_rate": 1.27642332826306e-06, "loss": 0.0012, "step": 36744 }, { "epoch": 6.74, "learning_rate": 1.2760604018878919e-06, "loss": 0.0024, "step": 36745 }, { "epoch": 6.74, "learning_rate": 1.2756975235992798e-06, "loss": 0.0017, "step": 36746 }, { "epoch": 6.74, "learning_rate": 1.2753346933992306e-06, "loss": 0.0014, "step": 36747 }, { "epoch": 6.74, "learning_rate": 1.2749719112897396e-06, "loss": 0.0028, "step": 36748 }, { "epoch": 6.74, "learning_rate": 1.2746091772728085e-06, "loss": 0.0016, "step": 36749 }, { "epoch": 6.74, "learning_rate": 1.274246491350436e-06, "loss": 0.0013, "step": 36750 }, { "epoch": 6.74, "learning_rate": 1.2738838535246223e-06, "loss": 0.0014, "step": 36751 }, { "epoch": 6.74, "learning_rate": 1.2735212637973637e-06, "loss": 0.0026, "step": 36752 }, { "epoch": 6.74, "learning_rate": 1.273158722170661e-06, "loss": 0.0019, "step": 36753 }, { "epoch": 6.74, "learning_rate": 1.2727962286465102e-06, "loss": 0.0036, "step": 36754 }, { "epoch": 6.74, "learning_rate": 1.2724337832269141e-06, "loss": 0.0018, "step": 36755 }, { "epoch": 6.74, "learning_rate": 1.2720713859138667e-06, "loss": 0.0016, "step": 36756 }, { "epoch": 6.74, "learning_rate": 1.2717090367093653e-06, "loss": 0.0014, "step": 36757 }, { "epoch": 6.74, "learning_rate": 1.2713467356154086e-06, "loss": 0.0019, "step": 36758 }, { "epoch": 6.74, "learning_rate": 1.2709844826339913e-06, "loss": 0.0009, "step": 36759 }, { "epoch": 6.74, "learning_rate": 1.2706222777671128e-06, "loss": 0.0025, "step": 36760 }, { "epoch": 6.74, "learning_rate": 1.2702601210167708e-06, "loss": 0.0009, "step": 36761 }, { "epoch": 6.74, "learning_rate": 1.2698980123849603e-06, "loss": 0.0007, "step": 36762 }, { "epoch": 6.74, "learning_rate": 1.2695359518736727e-06, "loss": 0.0023, "step": 36763 }, { "epoch": 6.74, "learning_rate": 1.2691739394849089e-06, "loss": 0.002, "step": 36764 }, { "epoch": 6.74, "learning_rate": 1.2688119752206641e-06, "loss": 0.0031, "step": 36765 }, { "epoch": 6.74, "learning_rate": 1.2684500590829317e-06, "loss": 0.002, "step": 36766 }, { "epoch": 6.74, "learning_rate": 1.2680881910737065e-06, "loss": 0.0008, "step": 36767 }, { "epoch": 6.74, "learning_rate": 1.267726371194985e-06, "loss": 0.0034, "step": 36768 }, { "epoch": 6.74, "learning_rate": 1.2673645994487593e-06, "loss": 0.0033, "step": 36769 }, { "epoch": 6.74, "learning_rate": 1.2670028758370245e-06, "loss": 0.0014, "step": 36770 }, { "epoch": 6.74, "learning_rate": 1.2666412003617756e-06, "loss": 0.0026, "step": 36771 }, { "epoch": 6.74, "learning_rate": 1.266279573025001e-06, "loss": 0.0019, "step": 36772 }, { "epoch": 6.74, "learning_rate": 1.2659179938287035e-06, "loss": 0.0012, "step": 36773 }, { "epoch": 6.74, "learning_rate": 1.2655564627748685e-06, "loss": 0.0025, "step": 36774 }, { "epoch": 6.74, "learning_rate": 1.2651949798654905e-06, "loss": 0.0041, "step": 36775 }, { "epoch": 6.74, "learning_rate": 1.2648335451025627e-06, "loss": 0.0035, "step": 36776 }, { "epoch": 6.74, "learning_rate": 1.2644721584880747e-06, "loss": 0.0016, "step": 36777 }, { "epoch": 6.74, "learning_rate": 1.2641108200240237e-06, "loss": 0.0013, "step": 36778 }, { "epoch": 6.74, "learning_rate": 1.263749529712398e-06, "loss": 0.0021, "step": 36779 }, { "epoch": 6.74, "learning_rate": 1.2633882875551918e-06, "loss": 0.0011, "step": 36780 }, { "epoch": 6.75, "learning_rate": 1.2630270935543898e-06, "loss": 0.0021, "step": 36781 }, { "epoch": 6.75, "learning_rate": 1.2626659477119884e-06, "loss": 0.002, "step": 36782 }, { "epoch": 6.75, "learning_rate": 1.2623048500299773e-06, "loss": 0.001, "step": 36783 }, { "epoch": 6.75, "learning_rate": 1.261943800510348e-06, "loss": 0.0016, "step": 36784 }, { "epoch": 6.75, "learning_rate": 1.2615827991550878e-06, "loss": 0.005, "step": 36785 }, { "epoch": 6.75, "learning_rate": 1.2612218459661885e-06, "loss": 0.0031, "step": 36786 }, { "epoch": 6.75, "learning_rate": 1.2608609409456384e-06, "loss": 0.0029, "step": 36787 }, { "epoch": 6.75, "learning_rate": 1.2605000840954273e-06, "loss": 0.0027, "step": 36788 }, { "epoch": 6.75, "learning_rate": 1.2601392754175456e-06, "loss": 0.0021, "step": 36789 }, { "epoch": 6.75, "learning_rate": 1.2597785149139786e-06, "loss": 0.0013, "step": 36790 }, { "epoch": 6.75, "learning_rate": 1.259417802586721e-06, "loss": 0.0017, "step": 36791 }, { "epoch": 6.75, "learning_rate": 1.2590571384377558e-06, "loss": 0.0021, "step": 36792 }, { "epoch": 6.75, "learning_rate": 1.2586965224690716e-06, "loss": 0.0013, "step": 36793 }, { "epoch": 6.75, "learning_rate": 1.2583359546826557e-06, "loss": 0.0016, "step": 36794 }, { "epoch": 6.75, "learning_rate": 1.2579754350804984e-06, "loss": 0.002, "step": 36795 }, { "epoch": 6.75, "learning_rate": 1.2576149636645852e-06, "loss": 0.0021, "step": 36796 }, { "epoch": 6.75, "learning_rate": 1.257254540436903e-06, "loss": 0.0023, "step": 36797 }, { "epoch": 6.75, "learning_rate": 1.2568941653994416e-06, "loss": 0.0007, "step": 36798 }, { "epoch": 6.75, "learning_rate": 1.2565338385541792e-06, "loss": 0.0028, "step": 36799 }, { "epoch": 6.75, "learning_rate": 1.25617355990311e-06, "loss": 0.0014, "step": 36800 }, { "epoch": 6.75, "learning_rate": 1.2558133294482156e-06, "loss": 0.0012, "step": 36801 }, { "epoch": 6.75, "learning_rate": 1.2554531471914844e-06, "loss": 0.0026, "step": 36802 }, { "epoch": 6.75, "learning_rate": 1.2550930131349004e-06, "loss": 0.001, "step": 36803 }, { "epoch": 6.75, "learning_rate": 1.2547329272804477e-06, "loss": 0.001, "step": 36804 }, { "epoch": 6.75, "learning_rate": 1.254372889630112e-06, "loss": 0.0031, "step": 36805 }, { "epoch": 6.75, "learning_rate": 1.254012900185878e-06, "loss": 0.0023, "step": 36806 }, { "epoch": 6.75, "learning_rate": 1.253652958949728e-06, "loss": 0.0015, "step": 36807 }, { "epoch": 6.75, "learning_rate": 1.2532930659236497e-06, "loss": 0.0025, "step": 36808 }, { "epoch": 6.75, "learning_rate": 1.252933221109628e-06, "loss": 0.0016, "step": 36809 }, { "epoch": 6.75, "learning_rate": 1.2525734245096399e-06, "loss": 0.0021, "step": 36810 }, { "epoch": 6.75, "learning_rate": 1.252213676125672e-06, "loss": 0.002, "step": 36811 }, { "epoch": 6.75, "learning_rate": 1.251853975959706e-06, "loss": 0.0017, "step": 36812 }, { "epoch": 6.75, "learning_rate": 1.2514943240137268e-06, "loss": 0.002, "step": 36813 }, { "epoch": 6.75, "learning_rate": 1.2511347202897174e-06, "loss": 0.0014, "step": 36814 }, { "epoch": 6.75, "learning_rate": 1.2507751647896571e-06, "loss": 0.0027, "step": 36815 }, { "epoch": 6.75, "learning_rate": 1.250415657515529e-06, "loss": 0.0016, "step": 36816 }, { "epoch": 6.75, "learning_rate": 1.250056198469316e-06, "loss": 0.001, "step": 36817 }, { "epoch": 6.75, "learning_rate": 1.2496967876529976e-06, "loss": 0.0017, "step": 36818 }, { "epoch": 6.75, "learning_rate": 1.2493374250685552e-06, "loss": 0.0017, "step": 36819 }, { "epoch": 6.75, "learning_rate": 1.24897811071797e-06, "loss": 0.0018, "step": 36820 }, { "epoch": 6.75, "learning_rate": 1.248618844603222e-06, "loss": 0.0018, "step": 36821 }, { "epoch": 6.75, "learning_rate": 1.2482596267262925e-06, "loss": 0.0018, "step": 36822 }, { "epoch": 6.75, "learning_rate": 1.2479004570891617e-06, "loss": 0.0032, "step": 36823 }, { "epoch": 6.75, "learning_rate": 1.2475413356938081e-06, "loss": 0.0033, "step": 36824 }, { "epoch": 6.75, "learning_rate": 1.2471822625422092e-06, "loss": 0.0016, "step": 36825 }, { "epoch": 6.75, "learning_rate": 1.2468232376363487e-06, "loss": 0.0018, "step": 36826 }, { "epoch": 6.75, "learning_rate": 1.2464642609782064e-06, "loss": 0.0016, "step": 36827 }, { "epoch": 6.75, "learning_rate": 1.246105332569756e-06, "loss": 0.0013, "step": 36828 }, { "epoch": 6.75, "learning_rate": 1.2457464524129747e-06, "loss": 0.0021, "step": 36829 }, { "epoch": 6.75, "learning_rate": 1.2453876205098481e-06, "loss": 0.0017, "step": 36830 }, { "epoch": 6.75, "learning_rate": 1.2450288368623487e-06, "loss": 0.0014, "step": 36831 }, { "epoch": 6.75, "learning_rate": 1.2446701014724561e-06, "loss": 0.0013, "step": 36832 }, { "epoch": 6.75, "learning_rate": 1.2443114143421486e-06, "loss": 0.0021, "step": 36833 }, { "epoch": 6.75, "learning_rate": 1.2439527754733972e-06, "loss": 0.0016, "step": 36834 }, { "epoch": 6.75, "learning_rate": 1.2435941848681864e-06, "loss": 0.003, "step": 36835 }, { "epoch": 6.76, "learning_rate": 1.2432356425284886e-06, "loss": 0.0019, "step": 36836 }, { "epoch": 6.76, "learning_rate": 1.2428771484562817e-06, "loss": 0.002, "step": 36837 }, { "epoch": 6.76, "learning_rate": 1.2425187026535412e-06, "loss": 0.0018, "step": 36838 }, { "epoch": 6.76, "learning_rate": 1.242160305122243e-06, "loss": 0.0014, "step": 36839 }, { "epoch": 6.76, "learning_rate": 1.2418019558643612e-06, "loss": 0.0017, "step": 36840 }, { "epoch": 6.76, "learning_rate": 1.2414436548818732e-06, "loss": 0.0015, "step": 36841 }, { "epoch": 6.76, "learning_rate": 1.2410854021767527e-06, "loss": 0.0014, "step": 36842 }, { "epoch": 6.76, "learning_rate": 1.2407271977509728e-06, "loss": 0.0022, "step": 36843 }, { "epoch": 6.76, "learning_rate": 1.2403690416065116e-06, "loss": 0.0015, "step": 36844 }, { "epoch": 6.76, "learning_rate": 1.2400109337453436e-06, "loss": 0.0016, "step": 36845 }, { "epoch": 6.76, "learning_rate": 1.239652874169438e-06, "loss": 0.001, "step": 36846 }, { "epoch": 6.76, "learning_rate": 1.23929486288077e-06, "loss": 0.0009, "step": 36847 }, { "epoch": 6.76, "learning_rate": 1.2389368998813156e-06, "loss": 0.002, "step": 36848 }, { "epoch": 6.76, "learning_rate": 1.2385789851730456e-06, "loss": 0.0019, "step": 36849 }, { "epoch": 6.76, "learning_rate": 1.238221118757934e-06, "loss": 0.003, "step": 36850 }, { "epoch": 6.76, "learning_rate": 1.2378633006379548e-06, "loss": 0.001, "step": 36851 }, { "epoch": 6.76, "learning_rate": 1.2375055308150741e-06, "loss": 0.0013, "step": 36852 }, { "epoch": 6.76, "learning_rate": 1.2371478092912704e-06, "loss": 0.0028, "step": 36853 }, { "epoch": 6.76, "learning_rate": 1.2367901360685141e-06, "loss": 0.0014, "step": 36854 }, { "epoch": 6.76, "learning_rate": 1.2364325111487752e-06, "loss": 0.0016, "step": 36855 }, { "epoch": 6.76, "learning_rate": 1.236074934534025e-06, "loss": 0.0015, "step": 36856 }, { "epoch": 6.76, "learning_rate": 1.2357174062262356e-06, "loss": 0.0017, "step": 36857 }, { "epoch": 6.76, "learning_rate": 1.2353599262273775e-06, "loss": 0.0021, "step": 36858 }, { "epoch": 6.76, "learning_rate": 1.2350024945394202e-06, "loss": 0.003, "step": 36859 }, { "epoch": 6.76, "learning_rate": 1.2346451111643331e-06, "loss": 0.0017, "step": 36860 }, { "epoch": 6.76, "learning_rate": 1.2342877761040884e-06, "loss": 0.0023, "step": 36861 }, { "epoch": 6.76, "learning_rate": 1.2339304893606585e-06, "loss": 0.0014, "step": 36862 }, { "epoch": 6.76, "learning_rate": 1.2335732509360054e-06, "loss": 0.0019, "step": 36863 }, { "epoch": 6.76, "learning_rate": 1.233216060832103e-06, "loss": 0.002, "step": 36864 }, { "epoch": 6.76, "learning_rate": 1.2328589190509167e-06, "loss": 0.0008, "step": 36865 }, { "epoch": 6.76, "learning_rate": 1.23250182559442e-06, "loss": 0.0016, "step": 36866 }, { "epoch": 6.76, "learning_rate": 1.2321447804645781e-06, "loss": 0.0022, "step": 36867 }, { "epoch": 6.76, "learning_rate": 1.2317877836633596e-06, "loss": 0.0012, "step": 36868 }, { "epoch": 6.76, "learning_rate": 1.231430835192733e-06, "loss": 0.0008, "step": 36869 }, { "epoch": 6.76, "learning_rate": 1.231073935054664e-06, "loss": 0.0016, "step": 36870 }, { "epoch": 6.76, "learning_rate": 1.2307170832511207e-06, "loss": 0.0013, "step": 36871 }, { "epoch": 6.76, "learning_rate": 1.2303602797840708e-06, "loss": 0.0009, "step": 36872 }, { "epoch": 6.76, "learning_rate": 1.2300035246554808e-06, "loss": 0.0015, "step": 36873 }, { "epoch": 6.76, "learning_rate": 1.229646817867317e-06, "loss": 0.002, "step": 36874 }, { "epoch": 6.76, "learning_rate": 1.2292901594215446e-06, "loss": 0.0012, "step": 36875 }, { "epoch": 6.76, "learning_rate": 1.2289335493201304e-06, "loss": 0.0018, "step": 36876 }, { "epoch": 6.76, "learning_rate": 1.228576987565041e-06, "loss": 0.0016, "step": 36877 }, { "epoch": 6.76, "learning_rate": 1.2282204741582381e-06, "loss": 0.0018, "step": 36878 }, { "epoch": 6.76, "learning_rate": 1.2278640091016913e-06, "loss": 0.0021, "step": 36879 }, { "epoch": 6.76, "learning_rate": 1.2275075923973655e-06, "loss": 0.0019, "step": 36880 }, { "epoch": 6.76, "learning_rate": 1.2271512240472216e-06, "loss": 0.0018, "step": 36881 }, { "epoch": 6.76, "learning_rate": 1.2267949040532235e-06, "loss": 0.0022, "step": 36882 }, { "epoch": 6.76, "learning_rate": 1.2264386324173395e-06, "loss": 0.0021, "step": 36883 }, { "epoch": 6.76, "learning_rate": 1.2260824091415314e-06, "loss": 0.0031, "step": 36884 }, { "epoch": 6.76, "learning_rate": 1.2257262342277621e-06, "loss": 0.0019, "step": 36885 }, { "epoch": 6.76, "learning_rate": 1.2253701076779955e-06, "loss": 0.0027, "step": 36886 }, { "epoch": 6.76, "learning_rate": 1.2250140294941947e-06, "loss": 0.0011, "step": 36887 }, { "epoch": 6.76, "learning_rate": 1.2246579996783225e-06, "loss": 0.0018, "step": 36888 }, { "epoch": 6.76, "learning_rate": 1.2243020182323407e-06, "loss": 0.0023, "step": 36889 }, { "epoch": 6.77, "learning_rate": 1.2239460851582118e-06, "loss": 0.002, "step": 36890 }, { "epoch": 6.77, "learning_rate": 1.223590200457896e-06, "loss": 0.0019, "step": 36891 }, { "epoch": 6.77, "learning_rate": 1.22323436413336e-06, "loss": 0.0013, "step": 36892 }, { "epoch": 6.77, "learning_rate": 1.2228785761865602e-06, "loss": 0.0024, "step": 36893 }, { "epoch": 6.77, "learning_rate": 1.2225228366194587e-06, "loss": 0.0018, "step": 36894 }, { "epoch": 6.77, "learning_rate": 1.2221671454340178e-06, "loss": 0.0012, "step": 36895 }, { "epoch": 6.77, "learning_rate": 1.2218115026321942e-06, "loss": 0.0017, "step": 36896 }, { "epoch": 6.77, "learning_rate": 1.2214559082159538e-06, "loss": 0.0018, "step": 36897 }, { "epoch": 6.77, "learning_rate": 1.2211003621872564e-06, "loss": 0.0018, "step": 36898 }, { "epoch": 6.77, "learning_rate": 1.220744864548058e-06, "loss": 0.0028, "step": 36899 }, { "epoch": 6.77, "learning_rate": 1.220389415300316e-06, "loss": 0.0009, "step": 36900 }, { "epoch": 6.77, "learning_rate": 1.2200340144459966e-06, "loss": 0.0027, "step": 36901 }, { "epoch": 6.77, "learning_rate": 1.2196786619870548e-06, "loss": 0.0012, "step": 36902 }, { "epoch": 6.77, "learning_rate": 1.2193233579254505e-06, "loss": 0.0014, "step": 36903 }, { "epoch": 6.77, "learning_rate": 1.2189681022631405e-06, "loss": 0.0019, "step": 36904 }, { "epoch": 6.77, "learning_rate": 1.2186128950020847e-06, "loss": 0.0017, "step": 36905 }, { "epoch": 6.77, "learning_rate": 1.2182577361442404e-06, "loss": 0.0025, "step": 36906 }, { "epoch": 6.77, "learning_rate": 1.2179026256915648e-06, "loss": 0.0016, "step": 36907 }, { "epoch": 6.77, "learning_rate": 1.2175475636460153e-06, "loss": 0.0014, "step": 36908 }, { "epoch": 6.77, "learning_rate": 1.2171925500095506e-06, "loss": 0.0027, "step": 36909 }, { "epoch": 6.77, "learning_rate": 1.2168375847841252e-06, "loss": 0.0014, "step": 36910 }, { "epoch": 6.77, "learning_rate": 1.2164826679716978e-06, "loss": 0.0025, "step": 36911 }, { "epoch": 6.77, "learning_rate": 1.2161277995742227e-06, "loss": 0.0017, "step": 36912 }, { "epoch": 6.77, "learning_rate": 1.2157729795936557e-06, "loss": 0.0026, "step": 36913 }, { "epoch": 6.77, "learning_rate": 1.2154182080319554e-06, "loss": 0.0008, "step": 36914 }, { "epoch": 6.77, "learning_rate": 1.2150634848910758e-06, "loss": 0.002, "step": 36915 }, { "epoch": 6.77, "learning_rate": 1.2147088101729743e-06, "loss": 0.0019, "step": 36916 }, { "epoch": 6.77, "learning_rate": 1.2143541838796025e-06, "loss": 0.0017, "step": 36917 }, { "epoch": 6.77, "learning_rate": 1.2139996060129122e-06, "loss": 0.0023, "step": 36918 }, { "epoch": 6.77, "learning_rate": 1.2136450765748653e-06, "loss": 0.0015, "step": 36919 }, { "epoch": 6.77, "learning_rate": 1.2132905955674123e-06, "loss": 0.0025, "step": 36920 }, { "epoch": 6.77, "learning_rate": 1.2129361629925074e-06, "loss": 0.0028, "step": 36921 }, { "epoch": 6.77, "learning_rate": 1.2125817788521044e-06, "loss": 0.0019, "step": 36922 }, { "epoch": 6.77, "learning_rate": 1.2122274431481562e-06, "loss": 0.002, "step": 36923 }, { "epoch": 6.77, "learning_rate": 1.211873155882617e-06, "loss": 0.0023, "step": 36924 }, { "epoch": 6.77, "learning_rate": 1.2115189170574382e-06, "loss": 0.0018, "step": 36925 }, { "epoch": 6.77, "learning_rate": 1.211164726674573e-06, "loss": 0.0018, "step": 36926 }, { "epoch": 6.77, "learning_rate": 1.210810584735973e-06, "loss": 0.001, "step": 36927 }, { "epoch": 6.77, "learning_rate": 1.2104564912435924e-06, "loss": 0.0018, "step": 36928 }, { "epoch": 6.77, "learning_rate": 1.2101024461993804e-06, "loss": 0.0012, "step": 36929 }, { "epoch": 6.77, "learning_rate": 1.2097484496052903e-06, "loss": 0.0028, "step": 36930 }, { "epoch": 6.77, "learning_rate": 1.2093945014632701e-06, "loss": 0.0024, "step": 36931 }, { "epoch": 6.77, "learning_rate": 1.2090406017752765e-06, "loss": 0.0019, "step": 36932 }, { "epoch": 6.77, "learning_rate": 1.2086867505432553e-06, "loss": 0.0016, "step": 36933 }, { "epoch": 6.77, "learning_rate": 1.208332947769162e-06, "loss": 0.0015, "step": 36934 }, { "epoch": 6.77, "learning_rate": 1.207979193454939e-06, "loss": 0.0008, "step": 36935 }, { "epoch": 6.77, "learning_rate": 1.207625487602543e-06, "loss": 0.0016, "step": 36936 }, { "epoch": 6.77, "learning_rate": 1.2072718302139209e-06, "loss": 0.003, "step": 36937 }, { "epoch": 6.77, "learning_rate": 1.2069182212910225e-06, "loss": 0.0015, "step": 36938 }, { "epoch": 6.77, "learning_rate": 1.2065646608357972e-06, "loss": 0.0014, "step": 36939 }, { "epoch": 6.77, "learning_rate": 1.2062111488501938e-06, "loss": 0.0017, "step": 36940 }, { "epoch": 6.77, "learning_rate": 1.2058576853361604e-06, "loss": 0.0012, "step": 36941 }, { "epoch": 6.77, "learning_rate": 1.2055042702956443e-06, "loss": 0.0014, "step": 36942 }, { "epoch": 6.77, "learning_rate": 1.2051509037305954e-06, "loss": 0.0015, "step": 36943 }, { "epoch": 6.77, "learning_rate": 1.2047975856429595e-06, "loss": 0.0022, "step": 36944 }, { "epoch": 6.78, "learning_rate": 1.2044443160346886e-06, "loss": 0.0032, "step": 36945 }, { "epoch": 6.78, "learning_rate": 1.2040910949077245e-06, "loss": 0.001, "step": 36946 }, { "epoch": 6.78, "learning_rate": 1.2037379222640167e-06, "loss": 0.0011, "step": 36947 }, { "epoch": 6.78, "learning_rate": 1.2033847981055114e-06, "loss": 0.0014, "step": 36948 }, { "epoch": 6.78, "learning_rate": 1.2030317224341538e-06, "loss": 0.0011, "step": 36949 }, { "epoch": 6.78, "learning_rate": 1.2026786952518932e-06, "loss": 0.0018, "step": 36950 }, { "epoch": 6.78, "learning_rate": 1.2023257165606727e-06, "loss": 0.0013, "step": 36951 }, { "epoch": 6.78, "learning_rate": 1.201972786362442e-06, "loss": 0.0027, "step": 36952 }, { "epoch": 6.78, "learning_rate": 1.2016199046591393e-06, "loss": 0.0016, "step": 36953 }, { "epoch": 6.78, "learning_rate": 1.2012670714527153e-06, "loss": 0.0029, "step": 36954 }, { "epoch": 6.78, "learning_rate": 1.200914286745114e-06, "loss": 0.0018, "step": 36955 }, { "epoch": 6.78, "learning_rate": 1.2005615505382795e-06, "loss": 0.0014, "step": 36956 }, { "epoch": 6.78, "learning_rate": 1.2002088628341546e-06, "loss": 0.0014, "step": 36957 }, { "epoch": 6.78, "learning_rate": 1.1998562236346857e-06, "loss": 0.0026, "step": 36958 }, { "epoch": 6.78, "learning_rate": 1.1995036329418153e-06, "loss": 0.0011, "step": 36959 }, { "epoch": 6.78, "learning_rate": 1.1991510907574867e-06, "loss": 0.0022, "step": 36960 }, { "epoch": 6.78, "learning_rate": 1.1987985970836436e-06, "loss": 0.0016, "step": 36961 }, { "epoch": 6.78, "learning_rate": 1.1984461519222269e-06, "loss": 0.0017, "step": 36962 }, { "epoch": 6.78, "learning_rate": 1.1980937552751848e-06, "loss": 0.0024, "step": 36963 }, { "epoch": 6.78, "learning_rate": 1.1977414071444549e-06, "loss": 0.0011, "step": 36964 }, { "epoch": 6.78, "learning_rate": 1.1973891075319798e-06, "loss": 0.0021, "step": 36965 }, { "epoch": 6.78, "learning_rate": 1.1970368564397005e-06, "loss": 0.0022, "step": 36966 }, { "epoch": 6.78, "learning_rate": 1.1966846538695632e-06, "loss": 0.0018, "step": 36967 }, { "epoch": 6.78, "learning_rate": 1.1963324998235048e-06, "loss": 0.0013, "step": 36968 }, { "epoch": 6.78, "learning_rate": 1.1959803943034687e-06, "loss": 0.0021, "step": 36969 }, { "epoch": 6.78, "learning_rate": 1.1956283373113964e-06, "loss": 0.0019, "step": 36970 }, { "epoch": 6.78, "learning_rate": 1.1952763288492241e-06, "loss": 0.0017, "step": 36971 }, { "epoch": 6.78, "learning_rate": 1.1949243689188961e-06, "loss": 0.0019, "step": 36972 }, { "epoch": 6.78, "learning_rate": 1.1945724575223516e-06, "loss": 0.002, "step": 36973 }, { "epoch": 6.78, "learning_rate": 1.1942205946615292e-06, "loss": 0.0023, "step": 36974 }, { "epoch": 6.78, "learning_rate": 1.1938687803383697e-06, "loss": 0.0026, "step": 36975 }, { "epoch": 6.78, "learning_rate": 1.1935170145548125e-06, "loss": 0.0013, "step": 36976 }, { "epoch": 6.78, "learning_rate": 1.193165297312795e-06, "loss": 0.0031, "step": 36977 }, { "epoch": 6.78, "learning_rate": 1.1928136286142566e-06, "loss": 0.0012, "step": 36978 }, { "epoch": 6.78, "learning_rate": 1.192462008461136e-06, "loss": 0.0017, "step": 36979 }, { "epoch": 6.78, "learning_rate": 1.1921104368553694e-06, "loss": 0.0024, "step": 36980 }, { "epoch": 6.78, "learning_rate": 1.1917589137989006e-06, "loss": 0.0015, "step": 36981 }, { "epoch": 6.78, "learning_rate": 1.1914074392936603e-06, "loss": 0.0023, "step": 36982 }, { "epoch": 6.78, "learning_rate": 1.1910560133415895e-06, "loss": 0.001, "step": 36983 }, { "epoch": 6.78, "learning_rate": 1.1907046359446205e-06, "loss": 0.0013, "step": 36984 }, { "epoch": 6.78, "learning_rate": 1.190353307104698e-06, "loss": 0.0009, "step": 36985 }, { "epoch": 6.78, "learning_rate": 1.1900020268237533e-06, "loss": 0.0013, "step": 36986 }, { "epoch": 6.78, "learning_rate": 1.1896507951037261e-06, "loss": 0.0021, "step": 36987 }, { "epoch": 6.78, "learning_rate": 1.1892996119465461e-06, "loss": 0.0012, "step": 36988 }, { "epoch": 6.78, "learning_rate": 1.188948477354156e-06, "loss": 0.0019, "step": 36989 }, { "epoch": 6.78, "learning_rate": 1.1885973913284865e-06, "loss": 0.0014, "step": 36990 }, { "epoch": 6.78, "learning_rate": 1.1882463538714762e-06, "loss": 0.0027, "step": 36991 }, { "epoch": 6.78, "learning_rate": 1.1878953649850578e-06, "loss": 0.0019, "step": 36992 }, { "epoch": 6.78, "learning_rate": 1.1875444246711665e-06, "loss": 0.0012, "step": 36993 }, { "epoch": 6.78, "learning_rate": 1.1871935329317363e-06, "loss": 0.0017, "step": 36994 }, { "epoch": 6.78, "learning_rate": 1.1868426897687025e-06, "loss": 0.0017, "step": 36995 }, { "epoch": 6.78, "learning_rate": 1.1864918951839988e-06, "loss": 0.0012, "step": 36996 }, { "epoch": 6.78, "learning_rate": 1.186141149179555e-06, "loss": 0.0032, "step": 36997 }, { "epoch": 6.78, "learning_rate": 1.1857904517573105e-06, "loss": 0.0015, "step": 36998 }, { "epoch": 6.79, "learning_rate": 1.185439802919197e-06, "loss": 0.0016, "step": 36999 }, { "epoch": 6.79, "learning_rate": 1.1850892026671446e-06, "loss": 0.0024, "step": 37000 }, { "epoch": 6.79, "learning_rate": 1.1847386510030868e-06, "loss": 0.0015, "step": 37001 }, { "epoch": 6.79, "learning_rate": 1.1843881479289532e-06, "loss": 0.0012, "step": 37002 }, { "epoch": 6.79, "learning_rate": 1.1840376934466812e-06, "loss": 0.0012, "step": 37003 }, { "epoch": 6.79, "learning_rate": 1.1836872875582006e-06, "loss": 0.0015, "step": 37004 }, { "epoch": 6.79, "learning_rate": 1.1833369302654429e-06, "loss": 0.002, "step": 37005 }, { "epoch": 6.79, "learning_rate": 1.1829866215703357e-06, "loss": 0.0016, "step": 37006 }, { "epoch": 6.79, "learning_rate": 1.182636361474815e-06, "loss": 0.002, "step": 37007 }, { "epoch": 6.79, "learning_rate": 1.182286149980808e-06, "loss": 0.0014, "step": 37008 }, { "epoch": 6.79, "learning_rate": 1.181935987090247e-06, "loss": 0.0018, "step": 37009 }, { "epoch": 6.79, "learning_rate": 1.1815858728050623e-06, "loss": 0.0013, "step": 37010 }, { "epoch": 6.79, "learning_rate": 1.1812358071271811e-06, "loss": 0.0015, "step": 37011 }, { "epoch": 6.79, "learning_rate": 1.1808857900585358e-06, "loss": 0.0016, "step": 37012 }, { "epoch": 6.79, "learning_rate": 1.1805358216010544e-06, "loss": 0.001, "step": 37013 }, { "epoch": 6.79, "learning_rate": 1.1801859017566663e-06, "loss": 0.0017, "step": 37014 }, { "epoch": 6.79, "learning_rate": 1.1798360305272983e-06, "loss": 0.0029, "step": 37015 }, { "epoch": 6.79, "learning_rate": 1.1794862079148828e-06, "loss": 0.0011, "step": 37016 }, { "epoch": 6.79, "learning_rate": 1.1791364339213474e-06, "loss": 0.0015, "step": 37017 }, { "epoch": 6.79, "learning_rate": 1.178786708548617e-06, "loss": 0.001, "step": 37018 }, { "epoch": 6.79, "learning_rate": 1.1784370317986182e-06, "loss": 0.0018, "step": 37019 }, { "epoch": 6.79, "learning_rate": 1.1780874036732838e-06, "loss": 0.0012, "step": 37020 }, { "epoch": 6.79, "learning_rate": 1.1777378241745385e-06, "loss": 0.0026, "step": 37021 }, { "epoch": 6.79, "learning_rate": 1.177388293304309e-06, "loss": 0.0025, "step": 37022 }, { "epoch": 6.79, "learning_rate": 1.1770388110645236e-06, "loss": 0.0017, "step": 37023 }, { "epoch": 6.79, "learning_rate": 1.176689377457103e-06, "loss": 0.0022, "step": 37024 }, { "epoch": 6.79, "learning_rate": 1.176339992483979e-06, "loss": 0.0011, "step": 37025 }, { "epoch": 6.79, "learning_rate": 1.175990656147077e-06, "loss": 0.0016, "step": 37026 }, { "epoch": 6.79, "learning_rate": 1.1756413684483192e-06, "loss": 0.0011, "step": 37027 }, { "epoch": 6.79, "learning_rate": 1.1752921293896347e-06, "loss": 0.0012, "step": 37028 }, { "epoch": 6.79, "learning_rate": 1.1749429389729462e-06, "loss": 0.0014, "step": 37029 }, { "epoch": 6.79, "learning_rate": 1.1745937972001787e-06, "loss": 0.0012, "step": 37030 }, { "epoch": 6.79, "learning_rate": 1.1742447040732564e-06, "loss": 0.0029, "step": 37031 }, { "epoch": 6.79, "learning_rate": 1.1738956595941054e-06, "loss": 0.0022, "step": 37032 }, { "epoch": 6.79, "learning_rate": 1.1735466637646453e-06, "loss": 0.0017, "step": 37033 }, { "epoch": 6.79, "learning_rate": 1.1731977165868069e-06, "loss": 0.0016, "step": 37034 }, { "epoch": 6.79, "learning_rate": 1.1728488180625075e-06, "loss": 0.0015, "step": 37035 }, { "epoch": 6.79, "learning_rate": 1.1724999681936723e-06, "loss": 0.0016, "step": 37036 }, { "epoch": 6.79, "learning_rate": 1.1721511669822216e-06, "loss": 0.0012, "step": 37037 }, { "epoch": 6.79, "learning_rate": 1.1718024144300833e-06, "loss": 0.0021, "step": 37038 }, { "epoch": 6.79, "learning_rate": 1.1714537105391766e-06, "loss": 0.0022, "step": 37039 }, { "epoch": 6.79, "learning_rate": 1.1711050553114234e-06, "loss": 0.0024, "step": 37040 }, { "epoch": 6.79, "learning_rate": 1.1707564487487467e-06, "loss": 0.0016, "step": 37041 }, { "epoch": 6.79, "learning_rate": 1.1704078908530669e-06, "loss": 0.0017, "step": 37042 }, { "epoch": 6.79, "learning_rate": 1.170059381626306e-06, "loss": 0.0013, "step": 37043 }, { "epoch": 6.79, "learning_rate": 1.169710921070385e-06, "loss": 0.0013, "step": 37044 }, { "epoch": 6.79, "learning_rate": 1.169362509187223e-06, "loss": 0.0027, "step": 37045 }, { "epoch": 6.79, "learning_rate": 1.1690141459787429e-06, "loss": 0.0013, "step": 37046 }, { "epoch": 6.79, "learning_rate": 1.1686658314468634e-06, "loss": 0.0014, "step": 37047 }, { "epoch": 6.79, "learning_rate": 1.168317565593504e-06, "loss": 0.0016, "step": 37048 }, { "epoch": 6.79, "learning_rate": 1.1679693484205856e-06, "loss": 0.0024, "step": 37049 }, { "epoch": 6.79, "learning_rate": 1.1676211799300251e-06, "loss": 0.0022, "step": 37050 }, { "epoch": 6.79, "learning_rate": 1.167273060123746e-06, "loss": 0.0016, "step": 37051 }, { "epoch": 6.79, "learning_rate": 1.166924989003666e-06, "loss": 0.002, "step": 37052 }, { "epoch": 6.79, "learning_rate": 1.166576966571701e-06, "loss": 0.0015, "step": 37053 }, { "epoch": 6.8, "learning_rate": 1.166228992829771e-06, "loss": 0.002, "step": 37054 }, { "epoch": 6.8, "learning_rate": 1.1658810677797906e-06, "loss": 0.0022, "step": 37055 }, { "epoch": 6.8, "learning_rate": 1.1655331914236835e-06, "loss": 0.0019, "step": 37056 }, { "epoch": 6.8, "learning_rate": 1.1651853637633637e-06, "loss": 0.0018, "step": 37057 }, { "epoch": 6.8, "learning_rate": 1.1648375848007498e-06, "loss": 0.0014, "step": 37058 }, { "epoch": 6.8, "learning_rate": 1.164489854537758e-06, "loss": 0.0014, "step": 37059 }, { "epoch": 6.8, "learning_rate": 1.1641421729763057e-06, "loss": 0.0024, "step": 37060 }, { "epoch": 6.8, "learning_rate": 1.1637945401183081e-06, "loss": 0.002, "step": 37061 }, { "epoch": 6.8, "learning_rate": 1.1634469559656814e-06, "loss": 0.0015, "step": 37062 }, { "epoch": 6.8, "learning_rate": 1.1630994205203427e-06, "loss": 0.0015, "step": 37063 }, { "epoch": 6.8, "learning_rate": 1.1627519337842074e-06, "loss": 0.0028, "step": 37064 }, { "epoch": 6.8, "learning_rate": 1.1624044957591907e-06, "loss": 0.0014, "step": 37065 }, { "epoch": 6.8, "learning_rate": 1.1620571064472064e-06, "loss": 0.0016, "step": 37066 }, { "epoch": 6.8, "learning_rate": 1.1617097658501708e-06, "loss": 0.0027, "step": 37067 }, { "epoch": 6.8, "learning_rate": 1.1613624739699958e-06, "loss": 0.0021, "step": 37068 }, { "epoch": 6.8, "learning_rate": 1.161015230808601e-06, "loss": 0.0018, "step": 37069 }, { "epoch": 6.8, "learning_rate": 1.160668036367898e-06, "loss": 0.0008, "step": 37070 }, { "epoch": 6.8, "learning_rate": 1.1603208906497987e-06, "loss": 0.002, "step": 37071 }, { "epoch": 6.8, "learning_rate": 1.159973793656215e-06, "loss": 0.0006, "step": 37072 }, { "epoch": 6.8, "learning_rate": 1.1596267453890653e-06, "loss": 0.0012, "step": 37073 }, { "epoch": 6.8, "learning_rate": 1.1592797458502603e-06, "loss": 0.002, "step": 37074 }, { "epoch": 6.8, "learning_rate": 1.1589327950417129e-06, "loss": 0.0011, "step": 37075 }, { "epoch": 6.8, "learning_rate": 1.1585858929653337e-06, "loss": 0.0018, "step": 37076 }, { "epoch": 6.8, "learning_rate": 1.158239039623037e-06, "loss": 0.0029, "step": 37077 }, { "epoch": 6.8, "learning_rate": 1.1578922350167343e-06, "loss": 0.0016, "step": 37078 }, { "epoch": 6.8, "learning_rate": 1.1575454791483365e-06, "loss": 0.0039, "step": 37079 }, { "epoch": 6.8, "learning_rate": 1.1571987720197542e-06, "loss": 0.0022, "step": 37080 }, { "epoch": 6.8, "learning_rate": 1.1568521136329002e-06, "loss": 0.0019, "step": 37081 }, { "epoch": 6.8, "learning_rate": 1.1565055039896843e-06, "loss": 0.0014, "step": 37082 }, { "epoch": 6.8, "learning_rate": 1.1561589430920173e-06, "loss": 0.0016, "step": 37083 }, { "epoch": 6.8, "learning_rate": 1.1558124309418096e-06, "loss": 0.0017, "step": 37084 }, { "epoch": 6.8, "learning_rate": 1.1554659675409685e-06, "loss": 0.0014, "step": 37085 }, { "epoch": 6.8, "learning_rate": 1.1551195528914094e-06, "loss": 0.0022, "step": 37086 }, { "epoch": 6.8, "learning_rate": 1.1547731869950374e-06, "loss": 0.0011, "step": 37087 }, { "epoch": 6.8, "learning_rate": 1.1544268698537652e-06, "loss": 0.0016, "step": 37088 }, { "epoch": 6.8, "learning_rate": 1.1540806014694972e-06, "loss": 0.0013, "step": 37089 }, { "epoch": 6.8, "learning_rate": 1.1537343818441426e-06, "loss": 0.0012, "step": 37090 }, { "epoch": 6.8, "learning_rate": 1.1533882109796125e-06, "loss": 0.0027, "step": 37091 }, { "epoch": 6.8, "learning_rate": 1.1530420888778137e-06, "loss": 0.0015, "step": 37092 }, { "epoch": 6.8, "learning_rate": 1.1526960155406552e-06, "loss": 0.002, "step": 37093 }, { "epoch": 6.8, "learning_rate": 1.1523499909700431e-06, "loss": 0.0017, "step": 37094 }, { "epoch": 6.8, "learning_rate": 1.1520040151678847e-06, "loss": 0.0021, "step": 37095 }, { "epoch": 6.8, "learning_rate": 1.1516580881360885e-06, "loss": 0.0017, "step": 37096 }, { "epoch": 6.8, "learning_rate": 1.1513122098765595e-06, "loss": 0.001, "step": 37097 }, { "epoch": 6.8, "learning_rate": 1.1509663803912052e-06, "loss": 0.0013, "step": 37098 }, { "epoch": 6.8, "learning_rate": 1.1506205996819309e-06, "loss": 0.0014, "step": 37099 }, { "epoch": 6.8, "learning_rate": 1.1502748677506436e-06, "loss": 0.0011, "step": 37100 }, { "epoch": 6.8, "learning_rate": 1.149929184599249e-06, "loss": 0.0013, "step": 37101 }, { "epoch": 6.8, "learning_rate": 1.1495835502296516e-06, "loss": 0.0015, "step": 37102 }, { "epoch": 6.8, "learning_rate": 1.1492379646437546e-06, "loss": 0.0024, "step": 37103 }, { "epoch": 6.8, "learning_rate": 1.1488924278434677e-06, "loss": 0.002, "step": 37104 }, { "epoch": 6.8, "learning_rate": 1.1485469398306926e-06, "loss": 0.0012, "step": 37105 }, { "epoch": 6.8, "learning_rate": 1.1482015006073366e-06, "loss": 0.0024, "step": 37106 }, { "epoch": 6.8, "learning_rate": 1.1478561101752995e-06, "loss": 0.0019, "step": 37107 }, { "epoch": 6.81, "learning_rate": 1.147510768536485e-06, "loss": 0.0013, "step": 37108 }, { "epoch": 6.81, "learning_rate": 1.1471654756927997e-06, "loss": 0.0024, "step": 37109 }, { "epoch": 6.81, "learning_rate": 1.1468202316461464e-06, "loss": 0.0016, "step": 37110 }, { "epoch": 6.81, "learning_rate": 1.1464750363984278e-06, "loss": 0.002, "step": 37111 }, { "epoch": 6.81, "learning_rate": 1.146129889951545e-06, "loss": 0.0024, "step": 37112 }, { "epoch": 6.81, "learning_rate": 1.1457847923074029e-06, "loss": 0.0017, "step": 37113 }, { "epoch": 6.81, "learning_rate": 1.1454397434679022e-06, "loss": 0.0018, "step": 37114 }, { "epoch": 6.81, "learning_rate": 1.1450947434349446e-06, "loss": 0.0007, "step": 37115 }, { "epoch": 6.81, "learning_rate": 1.14474979221043e-06, "loss": 0.002, "step": 37116 }, { "epoch": 6.81, "learning_rate": 1.1444048897962667e-06, "loss": 0.0024, "step": 37117 }, { "epoch": 6.81, "learning_rate": 1.1440600361943488e-06, "loss": 0.0014, "step": 37118 }, { "epoch": 6.81, "learning_rate": 1.1437152314065792e-06, "loss": 0.0015, "step": 37119 }, { "epoch": 6.81, "learning_rate": 1.1433704754348596e-06, "loss": 0.0021, "step": 37120 }, { "epoch": 6.81, "learning_rate": 1.1430257682810864e-06, "loss": 0.0019, "step": 37121 }, { "epoch": 6.81, "learning_rate": 1.1426811099471646e-06, "loss": 0.0011, "step": 37122 }, { "epoch": 6.81, "learning_rate": 1.1423365004349928e-06, "loss": 0.0012, "step": 37123 }, { "epoch": 6.81, "learning_rate": 1.1419919397464718e-06, "loss": 0.0008, "step": 37124 }, { "epoch": 6.81, "learning_rate": 1.1416474278834943e-06, "loss": 0.0018, "step": 37125 }, { "epoch": 6.81, "learning_rate": 1.1413029648479657e-06, "loss": 0.001, "step": 37126 }, { "epoch": 6.81, "learning_rate": 1.140958550641783e-06, "loss": 0.0013, "step": 37127 }, { "epoch": 6.81, "learning_rate": 1.140614185266844e-06, "loss": 0.0017, "step": 37128 }, { "epoch": 6.81, "learning_rate": 1.140269868725047e-06, "loss": 0.0024, "step": 37129 }, { "epoch": 6.81, "learning_rate": 1.1399256010182903e-06, "loss": 0.004, "step": 37130 }, { "epoch": 6.81, "learning_rate": 1.1395813821484714e-06, "loss": 0.0015, "step": 37131 }, { "epoch": 6.81, "learning_rate": 1.1392372121174876e-06, "loss": 0.0015, "step": 37132 }, { "epoch": 6.81, "learning_rate": 1.1388930909272355e-06, "loss": 0.0012, "step": 37133 }, { "epoch": 6.81, "learning_rate": 1.13854901857961e-06, "loss": 0.0016, "step": 37134 }, { "epoch": 6.81, "learning_rate": 1.138204995076514e-06, "loss": 0.0018, "step": 37135 }, { "epoch": 6.81, "learning_rate": 1.1378610204198382e-06, "loss": 0.0018, "step": 37136 }, { "epoch": 6.81, "learning_rate": 1.13751709461148e-06, "loss": 0.0015, "step": 37137 }, { "epoch": 6.81, "learning_rate": 1.1371732176533335e-06, "loss": 0.0023, "step": 37138 }, { "epoch": 6.81, "learning_rate": 1.136829389547297e-06, "loss": 0.002, "step": 37139 }, { "epoch": 6.81, "learning_rate": 1.1364856102952648e-06, "loss": 0.0023, "step": 37140 }, { "epoch": 6.81, "learning_rate": 1.1361418798991342e-06, "loss": 0.0021, "step": 37141 }, { "epoch": 6.81, "learning_rate": 1.1357981983607946e-06, "loss": 0.0023, "step": 37142 }, { "epoch": 6.81, "learning_rate": 1.1354545656821415e-06, "loss": 0.0021, "step": 37143 }, { "epoch": 6.81, "learning_rate": 1.1351109818650719e-06, "loss": 0.001, "step": 37144 }, { "epoch": 6.81, "learning_rate": 1.1347674469114788e-06, "loss": 0.0031, "step": 37145 }, { "epoch": 6.81, "learning_rate": 1.1344239608232544e-06, "loss": 0.0012, "step": 37146 }, { "epoch": 6.81, "learning_rate": 1.1340805236022934e-06, "loss": 0.0015, "step": 37147 }, { "epoch": 6.81, "learning_rate": 1.1337371352504877e-06, "loss": 0.0017, "step": 37148 }, { "epoch": 6.81, "learning_rate": 1.1333937957697317e-06, "loss": 0.0021, "step": 37149 }, { "epoch": 6.81, "learning_rate": 1.1330505051619166e-06, "loss": 0.002, "step": 37150 }, { "epoch": 6.81, "learning_rate": 1.1327072634289338e-06, "loss": 0.0018, "step": 37151 }, { "epoch": 6.81, "learning_rate": 1.1323640705726756e-06, "loss": 0.0031, "step": 37152 }, { "epoch": 6.81, "learning_rate": 1.1320209265950377e-06, "loss": 0.0015, "step": 37153 }, { "epoch": 6.81, "learning_rate": 1.131677831497906e-06, "loss": 0.003, "step": 37154 }, { "epoch": 6.81, "learning_rate": 1.1313347852831747e-06, "loss": 0.0015, "step": 37155 }, { "epoch": 6.81, "learning_rate": 1.130991787952731e-06, "loss": 0.0021, "step": 37156 }, { "epoch": 6.81, "learning_rate": 1.1306488395084703e-06, "loss": 0.0024, "step": 37157 }, { "epoch": 6.81, "learning_rate": 1.130305939952282e-06, "loss": 0.0014, "step": 37158 }, { "epoch": 6.81, "learning_rate": 1.1299630892860558e-06, "loss": 0.0013, "step": 37159 }, { "epoch": 6.81, "learning_rate": 1.1296202875116803e-06, "loss": 0.003, "step": 37160 }, { "epoch": 6.81, "learning_rate": 1.1292775346310425e-06, "loss": 0.0023, "step": 37161 }, { "epoch": 6.81, "learning_rate": 1.1289348306460367e-06, "loss": 0.002, "step": 37162 }, { "epoch": 6.82, "learning_rate": 1.1285921755585504e-06, "loss": 0.0029, "step": 37163 }, { "epoch": 6.82, "learning_rate": 1.128249569370472e-06, "loss": 0.001, "step": 37164 }, { "epoch": 6.82, "learning_rate": 1.1279070120836899e-06, "loss": 0.0009, "step": 37165 }, { "epoch": 6.82, "learning_rate": 1.1275645037000915e-06, "loss": 0.0021, "step": 37166 }, { "epoch": 6.82, "learning_rate": 1.1272220442215664e-06, "loss": 0.0017, "step": 37167 }, { "epoch": 6.82, "learning_rate": 1.1268796336499998e-06, "loss": 0.0017, "step": 37168 }, { "epoch": 6.82, "learning_rate": 1.1265372719872803e-06, "loss": 0.0013, "step": 37169 }, { "epoch": 6.82, "learning_rate": 1.1261949592352962e-06, "loss": 0.0017, "step": 37170 }, { "epoch": 6.82, "learning_rate": 1.1258526953959349e-06, "loss": 0.0024, "step": 37171 }, { "epoch": 6.82, "learning_rate": 1.1255104804710803e-06, "loss": 0.0011, "step": 37172 }, { "epoch": 6.82, "learning_rate": 1.125168314462619e-06, "loss": 0.0017, "step": 37173 }, { "epoch": 6.82, "learning_rate": 1.1248261973724373e-06, "loss": 0.0017, "step": 37174 }, { "epoch": 6.82, "learning_rate": 1.1244841292024222e-06, "loss": 0.0015, "step": 37175 }, { "epoch": 6.82, "learning_rate": 1.124142109954459e-06, "loss": 0.0018, "step": 37176 }, { "epoch": 6.82, "learning_rate": 1.1238001396304343e-06, "loss": 0.0014, "step": 37177 }, { "epoch": 6.82, "learning_rate": 1.1234582182322274e-06, "loss": 0.0021, "step": 37178 }, { "epoch": 6.82, "learning_rate": 1.1231163457617289e-06, "loss": 0.0016, "step": 37179 }, { "epoch": 6.82, "learning_rate": 1.1227745222208208e-06, "loss": 0.0017, "step": 37180 }, { "epoch": 6.82, "learning_rate": 1.1224327476113873e-06, "loss": 0.0021, "step": 37181 }, { "epoch": 6.82, "learning_rate": 1.1220910219353132e-06, "loss": 0.0011, "step": 37182 }, { "epoch": 6.82, "learning_rate": 1.1217493451944806e-06, "loss": 0.0008, "step": 37183 }, { "epoch": 6.82, "learning_rate": 1.1214077173907745e-06, "loss": 0.0013, "step": 37184 }, { "epoch": 6.82, "learning_rate": 1.1210661385260758e-06, "loss": 0.0012, "step": 37185 }, { "epoch": 6.82, "learning_rate": 1.1207246086022695e-06, "loss": 0.0014, "step": 37186 }, { "epoch": 6.82, "learning_rate": 1.1203831276212351e-06, "loss": 0.0029, "step": 37187 }, { "epoch": 6.82, "learning_rate": 1.1200416955848604e-06, "loss": 0.0022, "step": 37188 }, { "epoch": 6.82, "learning_rate": 1.1197003124950223e-06, "loss": 0.0017, "step": 37189 }, { "epoch": 6.82, "learning_rate": 1.1193589783536042e-06, "loss": 0.0024, "step": 37190 }, { "epoch": 6.82, "learning_rate": 1.1190176931624852e-06, "loss": 0.0017, "step": 37191 }, { "epoch": 6.82, "learning_rate": 1.1186764569235509e-06, "loss": 0.0018, "step": 37192 }, { "epoch": 6.82, "learning_rate": 1.1183352696386795e-06, "loss": 0.002, "step": 37193 }, { "epoch": 6.82, "learning_rate": 1.117994131309753e-06, "loss": 0.0013, "step": 37194 }, { "epoch": 6.82, "learning_rate": 1.1176530419386522e-06, "loss": 0.0016, "step": 37195 }, { "epoch": 6.82, "learning_rate": 1.117312001527252e-06, "loss": 0.0021, "step": 37196 }, { "epoch": 6.82, "learning_rate": 1.1169710100774389e-06, "loss": 0.0016, "step": 37197 }, { "epoch": 6.82, "learning_rate": 1.116630067591089e-06, "loss": 0.0021, "step": 37198 }, { "epoch": 6.82, "learning_rate": 1.1162891740700832e-06, "loss": 0.002, "step": 37199 }, { "epoch": 6.82, "learning_rate": 1.1159483295162986e-06, "loss": 0.0014, "step": 37200 }, { "epoch": 6.82, "learning_rate": 1.1156075339316152e-06, "loss": 0.0021, "step": 37201 }, { "epoch": 6.82, "learning_rate": 1.1152667873179112e-06, "loss": 0.0012, "step": 37202 }, { "epoch": 6.82, "learning_rate": 1.1149260896770652e-06, "loss": 0.0013, "step": 37203 }, { "epoch": 6.82, "learning_rate": 1.1145854410109546e-06, "loss": 0.0018, "step": 37204 }, { "epoch": 6.82, "learning_rate": 1.1142448413214547e-06, "loss": 0.0016, "step": 37205 }, { "epoch": 6.82, "learning_rate": 1.1139042906104502e-06, "loss": 0.0016, "step": 37206 }, { "epoch": 6.82, "learning_rate": 1.1135637888798101e-06, "loss": 0.0019, "step": 37207 }, { "epoch": 6.82, "learning_rate": 1.113223336131416e-06, "loss": 0.0016, "step": 37208 }, { "epoch": 6.82, "learning_rate": 1.1128829323671408e-06, "loss": 0.0021, "step": 37209 }, { "epoch": 6.82, "learning_rate": 1.1125425775888642e-06, "loss": 0.0015, "step": 37210 }, { "epoch": 6.82, "learning_rate": 1.1122022717984615e-06, "loss": 0.0015, "step": 37211 }, { "epoch": 6.82, "learning_rate": 1.1118620149978088e-06, "loss": 0.0011, "step": 37212 }, { "epoch": 6.82, "learning_rate": 1.11152180718878e-06, "loss": 0.0011, "step": 37213 }, { "epoch": 6.82, "learning_rate": 1.1111816483732518e-06, "loss": 0.002, "step": 37214 }, { "epoch": 6.82, "learning_rate": 1.110841538553099e-06, "loss": 0.0023, "step": 37215 }, { "epoch": 6.82, "learning_rate": 1.1105014777301949e-06, "loss": 0.0026, "step": 37216 }, { "epoch": 6.83, "learning_rate": 1.1101614659064153e-06, "loss": 0.002, "step": 37217 }, { "epoch": 6.83, "learning_rate": 1.1098215030836345e-06, "loss": 0.0015, "step": 37218 }, { "epoch": 6.83, "learning_rate": 1.1094815892637256e-06, "loss": 0.0032, "step": 37219 }, { "epoch": 6.83, "learning_rate": 1.1091417244485624e-06, "loss": 0.0018, "step": 37220 }, { "epoch": 6.83, "learning_rate": 1.1088019086400181e-06, "loss": 0.002, "step": 37221 }, { "epoch": 6.83, "learning_rate": 1.1084621418399632e-06, "loss": 0.0015, "step": 37222 }, { "epoch": 6.83, "learning_rate": 1.1081224240502763e-06, "loss": 0.0016, "step": 37223 }, { "epoch": 6.83, "learning_rate": 1.1077827552728293e-06, "loss": 0.0021, "step": 37224 }, { "epoch": 6.83, "learning_rate": 1.1074431355094895e-06, "loss": 0.0017, "step": 37225 }, { "epoch": 6.83, "learning_rate": 1.1071035647621308e-06, "loss": 0.0018, "step": 37226 }, { "epoch": 6.83, "learning_rate": 1.1067640430326244e-06, "loss": 0.0016, "step": 37227 }, { "epoch": 6.83, "learning_rate": 1.1064245703228448e-06, "loss": 0.0016, "step": 37228 }, { "epoch": 6.83, "learning_rate": 1.1060851466346612e-06, "loss": 0.0023, "step": 37229 }, { "epoch": 6.83, "learning_rate": 1.105745771969945e-06, "loss": 0.0014, "step": 37230 }, { "epoch": 6.83, "learning_rate": 1.105406446330567e-06, "loss": 0.0026, "step": 37231 }, { "epoch": 6.83, "learning_rate": 1.1050671697183956e-06, "loss": 0.0018, "step": 37232 }, { "epoch": 6.83, "learning_rate": 1.104727942135303e-06, "loss": 0.0014, "step": 37233 }, { "epoch": 6.83, "learning_rate": 1.1043887635831584e-06, "loss": 0.0024, "step": 37234 }, { "epoch": 6.83, "learning_rate": 1.1040496340638318e-06, "loss": 0.0026, "step": 37235 }, { "epoch": 6.83, "learning_rate": 1.1037105535791914e-06, "loss": 0.0012, "step": 37236 }, { "epoch": 6.83, "learning_rate": 1.103371522131107e-06, "loss": 0.0025, "step": 37237 }, { "epoch": 6.83, "learning_rate": 1.103032539721447e-06, "loss": 0.001, "step": 37238 }, { "epoch": 6.83, "learning_rate": 1.10269360635208e-06, "loss": 0.0015, "step": 37239 }, { "epoch": 6.83, "learning_rate": 1.1023547220248732e-06, "loss": 0.0024, "step": 37240 }, { "epoch": 6.83, "learning_rate": 1.1020158867416964e-06, "loss": 0.0018, "step": 37241 }, { "epoch": 6.83, "learning_rate": 1.101677100504419e-06, "loss": 0.0033, "step": 37242 }, { "epoch": 6.83, "learning_rate": 1.1013383633149045e-06, "loss": 0.0014, "step": 37243 }, { "epoch": 6.83, "learning_rate": 1.1009996751750185e-06, "loss": 0.0022, "step": 37244 }, { "epoch": 6.83, "learning_rate": 1.1006610360866333e-06, "loss": 0.0013, "step": 37245 }, { "epoch": 6.83, "learning_rate": 1.1003224460516127e-06, "loss": 0.0017, "step": 37246 }, { "epoch": 6.83, "learning_rate": 1.0999839050718242e-06, "loss": 0.0018, "step": 37247 }, { "epoch": 6.83, "learning_rate": 1.099645413149134e-06, "loss": 0.0015, "step": 37248 }, { "epoch": 6.83, "learning_rate": 1.099306970285402e-06, "loss": 0.0019, "step": 37249 }, { "epoch": 6.83, "learning_rate": 1.098968576482502e-06, "loss": 0.0013, "step": 37250 }, { "epoch": 6.83, "learning_rate": 1.0986302317422948e-06, "loss": 0.0011, "step": 37251 }, { "epoch": 6.83, "learning_rate": 1.0982919360666466e-06, "loss": 0.0025, "step": 37252 }, { "epoch": 6.83, "learning_rate": 1.0979536894574217e-06, "loss": 0.0011, "step": 37253 }, { "epoch": 6.83, "learning_rate": 1.097615491916485e-06, "loss": 0.0014, "step": 37254 }, { "epoch": 6.83, "learning_rate": 1.0972773434457006e-06, "loss": 0.002, "step": 37255 }, { "epoch": 6.83, "learning_rate": 1.0969392440469306e-06, "loss": 0.0016, "step": 37256 }, { "epoch": 6.83, "learning_rate": 1.0966011937220412e-06, "loss": 0.0024, "step": 37257 }, { "epoch": 6.83, "learning_rate": 1.0962631924728929e-06, "loss": 0.0023, "step": 37258 }, { "epoch": 6.83, "learning_rate": 1.0959252403013521e-06, "loss": 0.0013, "step": 37259 }, { "epoch": 6.83, "learning_rate": 1.0955873372092818e-06, "loss": 0.0015, "step": 37260 }, { "epoch": 6.83, "learning_rate": 1.0952494831985404e-06, "loss": 0.0019, "step": 37261 }, { "epoch": 6.83, "learning_rate": 1.09491167827099e-06, "loss": 0.0013, "step": 37262 }, { "epoch": 6.83, "learning_rate": 1.094573922428498e-06, "loss": 0.0011, "step": 37263 }, { "epoch": 6.83, "learning_rate": 1.0942362156729236e-06, "loss": 0.0022, "step": 37264 }, { "epoch": 6.83, "learning_rate": 1.0938985580061267e-06, "loss": 0.0014, "step": 37265 }, { "epoch": 6.83, "learning_rate": 1.0935609494299703e-06, "loss": 0.0016, "step": 37266 }, { "epoch": 6.83, "learning_rate": 1.093223389946314e-06, "loss": 0.0016, "step": 37267 }, { "epoch": 6.83, "learning_rate": 1.0928858795570197e-06, "loss": 0.0021, "step": 37268 }, { "epoch": 6.83, "learning_rate": 1.0925484182639467e-06, "loss": 0.0017, "step": 37269 }, { "epoch": 6.83, "learning_rate": 1.0922110060689561e-06, "loss": 0.002, "step": 37270 }, { "epoch": 6.83, "learning_rate": 1.0918736429739064e-06, "loss": 0.002, "step": 37271 }, { "epoch": 6.84, "learning_rate": 1.091536328980658e-06, "loss": 0.0017, "step": 37272 }, { "epoch": 6.84, "learning_rate": 1.0911990640910709e-06, "loss": 0.0019, "step": 37273 }, { "epoch": 6.84, "learning_rate": 1.0908618483070021e-06, "loss": 0.0022, "step": 37274 }, { "epoch": 6.84, "learning_rate": 1.0905246816303105e-06, "loss": 0.0017, "step": 37275 }, { "epoch": 6.84, "learning_rate": 1.0901875640628568e-06, "loss": 0.0011, "step": 37276 }, { "epoch": 6.84, "learning_rate": 1.089850495606498e-06, "loss": 0.0019, "step": 37277 }, { "epoch": 6.84, "learning_rate": 1.0895134762630942e-06, "loss": 0.0016, "step": 37278 }, { "epoch": 6.84, "learning_rate": 1.0891765060344994e-06, "loss": 0.0018, "step": 37279 }, { "epoch": 6.84, "learning_rate": 1.0888395849225697e-06, "loss": 0.0025, "step": 37280 }, { "epoch": 6.84, "learning_rate": 1.088502712929167e-06, "loss": 0.0023, "step": 37281 }, { "epoch": 6.84, "learning_rate": 1.0881658900561464e-06, "loss": 0.0015, "step": 37282 }, { "epoch": 6.84, "learning_rate": 1.0878291163053644e-06, "loss": 0.0023, "step": 37283 }, { "epoch": 6.84, "learning_rate": 1.0874923916786773e-06, "loss": 0.0018, "step": 37284 }, { "epoch": 6.84, "learning_rate": 1.08715571617794e-06, "loss": 0.0005, "step": 37285 }, { "epoch": 6.84, "learning_rate": 1.0868190898050102e-06, "loss": 0.0015, "step": 37286 }, { "epoch": 6.84, "learning_rate": 1.0864825125617419e-06, "loss": 0.0028, "step": 37287 }, { "epoch": 6.84, "learning_rate": 1.08614598444999e-06, "loss": 0.0016, "step": 37288 }, { "epoch": 6.84, "learning_rate": 1.0858095054716111e-06, "loss": 0.0016, "step": 37289 }, { "epoch": 6.84, "learning_rate": 1.0854730756284594e-06, "loss": 0.0011, "step": 37290 }, { "epoch": 6.84, "learning_rate": 1.0851366949223885e-06, "loss": 0.003, "step": 37291 }, { "epoch": 6.84, "learning_rate": 1.084800363355253e-06, "loss": 0.0012, "step": 37292 }, { "epoch": 6.84, "learning_rate": 1.0844640809289053e-06, "loss": 0.0014, "step": 37293 }, { "epoch": 6.84, "learning_rate": 1.0841278476452011e-06, "loss": 0.0016, "step": 37294 }, { "epoch": 6.84, "learning_rate": 1.0837916635059963e-06, "loss": 0.0019, "step": 37295 }, { "epoch": 6.84, "learning_rate": 1.0834555285131388e-06, "loss": 0.0016, "step": 37296 }, { "epoch": 6.84, "learning_rate": 1.083119442668481e-06, "loss": 0.0026, "step": 37297 }, { "epoch": 6.84, "learning_rate": 1.0827834059738796e-06, "loss": 0.002, "step": 37298 }, { "epoch": 6.84, "learning_rate": 1.0824474184311852e-06, "loss": 0.0018, "step": 37299 }, { "epoch": 6.84, "learning_rate": 1.0821114800422482e-06, "loss": 0.0014, "step": 37300 }, { "epoch": 6.84, "learning_rate": 1.0817755908089234e-06, "loss": 0.0025, "step": 37301 }, { "epoch": 6.84, "learning_rate": 1.08143975073306e-06, "loss": 0.0027, "step": 37302 }, { "epoch": 6.84, "learning_rate": 1.0811039598165097e-06, "loss": 0.0014, "step": 37303 }, { "epoch": 6.84, "learning_rate": 1.0807682180611234e-06, "loss": 0.0018, "step": 37304 }, { "epoch": 6.84, "learning_rate": 1.0804325254687508e-06, "loss": 0.0007, "step": 37305 }, { "epoch": 6.84, "learning_rate": 1.0800968820412416e-06, "loss": 0.0021, "step": 37306 }, { "epoch": 6.84, "learning_rate": 1.079761287780452e-06, "loss": 0.0013, "step": 37307 }, { "epoch": 6.84, "learning_rate": 1.0794257426882238e-06, "loss": 0.003, "step": 37308 }, { "epoch": 6.84, "learning_rate": 1.079090246766411e-06, "loss": 0.0011, "step": 37309 }, { "epoch": 6.84, "learning_rate": 1.0787548000168613e-06, "loss": 0.0026, "step": 37310 }, { "epoch": 6.84, "learning_rate": 1.078419402441422e-06, "loss": 0.0021, "step": 37311 }, { "epoch": 6.84, "learning_rate": 1.0780840540419468e-06, "loss": 0.0016, "step": 37312 }, { "epoch": 6.84, "learning_rate": 1.0777487548202825e-06, "loss": 0.0013, "step": 37313 }, { "epoch": 6.84, "learning_rate": 1.0774135047782741e-06, "loss": 0.0023, "step": 37314 }, { "epoch": 6.84, "learning_rate": 1.0770783039177701e-06, "loss": 0.0036, "step": 37315 }, { "epoch": 6.84, "learning_rate": 1.0767431522406214e-06, "loss": 0.001, "step": 37316 }, { "epoch": 6.84, "learning_rate": 1.0764080497486728e-06, "loss": 0.001, "step": 37317 }, { "epoch": 6.84, "learning_rate": 1.0760729964437722e-06, "loss": 0.0012, "step": 37318 }, { "epoch": 6.84, "learning_rate": 1.0757379923277667e-06, "loss": 0.0011, "step": 37319 }, { "epoch": 6.84, "learning_rate": 1.0754030374025027e-06, "loss": 0.001, "step": 37320 }, { "epoch": 6.84, "learning_rate": 1.0750681316698265e-06, "loss": 0.0016, "step": 37321 }, { "epoch": 6.84, "learning_rate": 1.0747332751315831e-06, "loss": 0.0009, "step": 37322 }, { "epoch": 6.84, "learning_rate": 1.074398467789619e-06, "loss": 0.0013, "step": 37323 }, { "epoch": 6.84, "learning_rate": 1.0740637096457773e-06, "loss": 0.0007, "step": 37324 }, { "epoch": 6.84, "learning_rate": 1.0737290007019107e-06, "loss": 0.0013, "step": 37325 }, { "epoch": 6.85, "learning_rate": 1.0733943409598558e-06, "loss": 0.0024, "step": 37326 }, { "epoch": 6.85, "learning_rate": 1.0730597304214618e-06, "loss": 0.0007, "step": 37327 }, { "epoch": 6.85, "learning_rate": 1.0727251690885688e-06, "loss": 0.0017, "step": 37328 }, { "epoch": 6.85, "learning_rate": 1.072390656963026e-06, "loss": 0.0014, "step": 37329 }, { "epoch": 6.85, "learning_rate": 1.0720561940466745e-06, "loss": 0.0008, "step": 37330 }, { "epoch": 6.85, "learning_rate": 1.0717217803413605e-06, "loss": 0.0016, "step": 37331 }, { "epoch": 6.85, "learning_rate": 1.0713874158489235e-06, "loss": 0.0019, "step": 37332 }, { "epoch": 6.85, "learning_rate": 1.0710531005712065e-06, "loss": 0.001, "step": 37333 }, { "epoch": 6.85, "learning_rate": 1.070718834510056e-06, "loss": 0.0011, "step": 37334 }, { "epoch": 6.85, "learning_rate": 1.0703846176673126e-06, "loss": 0.0013, "step": 37335 }, { "epoch": 6.85, "learning_rate": 1.0700504500448173e-06, "loss": 0.0007, "step": 37336 }, { "epoch": 6.85, "learning_rate": 1.0697163316444148e-06, "loss": 0.0012, "step": 37337 }, { "epoch": 6.85, "learning_rate": 1.0693822624679439e-06, "loss": 0.0021, "step": 37338 }, { "epoch": 6.85, "learning_rate": 1.0690482425172478e-06, "loss": 0.0013, "step": 37339 }, { "epoch": 6.85, "learning_rate": 1.0687142717941656e-06, "loss": 0.0008, "step": 37340 }, { "epoch": 6.85, "learning_rate": 1.0683803503005397e-06, "loss": 0.0025, "step": 37341 }, { "epoch": 6.85, "learning_rate": 1.0680464780382126e-06, "loss": 0.0021, "step": 37342 }, { "epoch": 6.85, "learning_rate": 1.0677126550090222e-06, "loss": 0.0017, "step": 37343 }, { "epoch": 6.85, "learning_rate": 1.0673788812148079e-06, "loss": 0.0022, "step": 37344 }, { "epoch": 6.85, "learning_rate": 1.0670451566574102e-06, "loss": 0.0023, "step": 37345 }, { "epoch": 6.85, "learning_rate": 1.066711481338668e-06, "loss": 0.0016, "step": 37346 }, { "epoch": 6.85, "learning_rate": 1.0663778552604231e-06, "loss": 0.0016, "step": 37347 }, { "epoch": 6.85, "learning_rate": 1.0660442784245117e-06, "loss": 0.0032, "step": 37348 }, { "epoch": 6.85, "learning_rate": 1.0657107508327768e-06, "loss": 0.0026, "step": 37349 }, { "epoch": 6.85, "learning_rate": 1.0653772724870492e-06, "loss": 0.0016, "step": 37350 }, { "epoch": 6.85, "learning_rate": 1.065043843389173e-06, "loss": 0.0011, "step": 37351 }, { "epoch": 6.85, "learning_rate": 1.0647104635409844e-06, "loss": 0.0021, "step": 37352 }, { "epoch": 6.85, "learning_rate": 1.064377132944322e-06, "loss": 0.0009, "step": 37353 }, { "epoch": 6.85, "learning_rate": 1.064043851601021e-06, "loss": 0.002, "step": 37354 }, { "epoch": 6.85, "learning_rate": 1.063710619512921e-06, "loss": 0.0022, "step": 37355 }, { "epoch": 6.85, "learning_rate": 1.0633774366818562e-06, "loss": 0.0014, "step": 37356 }, { "epoch": 6.85, "learning_rate": 1.0630443031096648e-06, "loss": 0.0011, "step": 37357 }, { "epoch": 6.85, "learning_rate": 1.0627112187981825e-06, "loss": 0.0014, "step": 37358 }, { "epoch": 6.85, "learning_rate": 1.0623781837492431e-06, "loss": 0.002, "step": 37359 }, { "epoch": 6.85, "learning_rate": 1.0620451979646895e-06, "loss": 0.0014, "step": 37360 }, { "epoch": 6.85, "learning_rate": 1.0617122614463494e-06, "loss": 0.0018, "step": 37361 }, { "epoch": 6.85, "learning_rate": 1.0613793741960609e-06, "loss": 0.0014, "step": 37362 }, { "epoch": 6.85, "learning_rate": 1.0610465362156585e-06, "loss": 0.0009, "step": 37363 }, { "epoch": 6.85, "learning_rate": 1.0607137475069751e-06, "loss": 0.0014, "step": 37364 }, { "epoch": 6.85, "learning_rate": 1.0603810080718491e-06, "loss": 0.0012, "step": 37365 }, { "epoch": 6.85, "learning_rate": 1.0600483179121112e-06, "loss": 0.0018, "step": 37366 }, { "epoch": 6.85, "learning_rate": 1.0597156770296002e-06, "loss": 0.0014, "step": 37367 }, { "epoch": 6.85, "learning_rate": 1.0593830854261411e-06, "loss": 0.0024, "step": 37368 }, { "epoch": 6.85, "learning_rate": 1.0590505431035735e-06, "loss": 0.0023, "step": 37369 }, { "epoch": 6.85, "learning_rate": 1.0587180500637294e-06, "loss": 0.0022, "step": 37370 }, { "epoch": 6.85, "learning_rate": 1.0583856063084407e-06, "loss": 0.0019, "step": 37371 }, { "epoch": 6.85, "learning_rate": 1.0580532118395404e-06, "loss": 0.0021, "step": 37372 }, { "epoch": 6.85, "learning_rate": 1.057720866658859e-06, "loss": 0.0019, "step": 37373 }, { "epoch": 6.85, "learning_rate": 1.0573885707682308e-06, "loss": 0.0021, "step": 37374 }, { "epoch": 6.85, "learning_rate": 1.0570563241694864e-06, "loss": 0.0012, "step": 37375 }, { "epoch": 6.85, "learning_rate": 1.0567241268644569e-06, "loss": 0.0025, "step": 37376 }, { "epoch": 6.85, "learning_rate": 1.0563919788549704e-06, "loss": 0.0013, "step": 37377 }, { "epoch": 6.85, "learning_rate": 1.0560598801428668e-06, "loss": 0.001, "step": 37378 }, { "epoch": 6.85, "learning_rate": 1.0557278307299667e-06, "loss": 0.0021, "step": 37379 }, { "epoch": 6.85, "learning_rate": 1.0553958306181056e-06, "loss": 0.0018, "step": 37380 }, { "epoch": 6.86, "learning_rate": 1.0550638798091096e-06, "loss": 0.0025, "step": 37381 }, { "epoch": 6.86, "learning_rate": 1.0547319783048137e-06, "loss": 0.0017, "step": 37382 }, { "epoch": 6.86, "learning_rate": 1.0544001261070447e-06, "loss": 0.0015, "step": 37383 }, { "epoch": 6.86, "learning_rate": 1.0540683232176307e-06, "loss": 0.0012, "step": 37384 }, { "epoch": 6.86, "learning_rate": 1.053736569638405e-06, "loss": 0.0023, "step": 37385 }, { "epoch": 6.86, "learning_rate": 1.0534048653711882e-06, "loss": 0.0019, "step": 37386 }, { "epoch": 6.86, "learning_rate": 1.0530732104178155e-06, "loss": 0.0012, "step": 37387 }, { "epoch": 6.86, "learning_rate": 1.0527416047801131e-06, "loss": 0.0018, "step": 37388 }, { "epoch": 6.86, "learning_rate": 1.0524100484599077e-06, "loss": 0.0016, "step": 37389 }, { "epoch": 6.86, "learning_rate": 1.0520785414590284e-06, "loss": 0.0012, "step": 37390 }, { "epoch": 6.86, "learning_rate": 1.0517470837793021e-06, "loss": 0.0024, "step": 37391 }, { "epoch": 6.86, "learning_rate": 1.0514156754225557e-06, "loss": 0.0022, "step": 37392 }, { "epoch": 6.86, "learning_rate": 1.0510843163906148e-06, "loss": 0.0017, "step": 37393 }, { "epoch": 6.86, "learning_rate": 1.0507530066853055e-06, "loss": 0.0013, "step": 37394 }, { "epoch": 6.86, "learning_rate": 1.0504217463084564e-06, "loss": 0.0013, "step": 37395 }, { "epoch": 6.86, "learning_rate": 1.0500905352618952e-06, "loss": 0.0015, "step": 37396 }, { "epoch": 6.86, "learning_rate": 1.049759373547442e-06, "loss": 0.0017, "step": 37397 }, { "epoch": 6.86, "learning_rate": 1.049428261166926e-06, "loss": 0.0023, "step": 37398 }, { "epoch": 6.86, "learning_rate": 1.0490971981221676e-06, "loss": 0.0014, "step": 37399 }, { "epoch": 6.86, "learning_rate": 1.0487661844149989e-06, "loss": 0.0015, "step": 37400 }, { "epoch": 6.86, "learning_rate": 1.048435220047239e-06, "loss": 0.0029, "step": 37401 }, { "epoch": 6.86, "learning_rate": 1.0481043050207163e-06, "loss": 0.002, "step": 37402 }, { "epoch": 6.86, "learning_rate": 1.0477734393372485e-06, "loss": 0.0023, "step": 37403 }, { "epoch": 6.86, "learning_rate": 1.0474426229986656e-06, "loss": 0.0018, "step": 37404 }, { "epoch": 6.86, "learning_rate": 1.0471118560067882e-06, "loss": 0.0012, "step": 37405 }, { "epoch": 6.86, "learning_rate": 1.0467811383634407e-06, "loss": 0.0021, "step": 37406 }, { "epoch": 6.86, "learning_rate": 1.0464504700704447e-06, "loss": 0.0019, "step": 37407 }, { "epoch": 6.86, "learning_rate": 1.0461198511296233e-06, "loss": 0.0009, "step": 37408 }, { "epoch": 6.86, "learning_rate": 1.0457892815427994e-06, "loss": 0.0023, "step": 37409 }, { "epoch": 6.86, "learning_rate": 1.045458761311795e-06, "loss": 0.0017, "step": 37410 }, { "epoch": 6.86, "learning_rate": 1.045128290438432e-06, "loss": 0.0011, "step": 37411 }, { "epoch": 6.86, "learning_rate": 1.0447978689245287e-06, "loss": 0.0012, "step": 37412 }, { "epoch": 6.86, "learning_rate": 1.0444674967719115e-06, "loss": 0.0016, "step": 37413 }, { "epoch": 6.86, "learning_rate": 1.0441371739824012e-06, "loss": 0.0022, "step": 37414 }, { "epoch": 6.86, "learning_rate": 1.0438069005578156e-06, "loss": 0.0017, "step": 37415 }, { "epoch": 6.86, "learning_rate": 1.0434766764999749e-06, "loss": 0.0011, "step": 37416 }, { "epoch": 6.86, "learning_rate": 1.043146501810699e-06, "loss": 0.0021, "step": 37417 }, { "epoch": 6.86, "learning_rate": 1.042816376491812e-06, "loss": 0.001, "step": 37418 }, { "epoch": 6.86, "learning_rate": 1.0424863005451303e-06, "loss": 0.0025, "step": 37419 }, { "epoch": 6.86, "learning_rate": 1.0421562739724767e-06, "loss": 0.002, "step": 37420 }, { "epoch": 6.86, "learning_rate": 1.0418262967756632e-06, "loss": 0.0016, "step": 37421 }, { "epoch": 6.86, "learning_rate": 1.041496368956515e-06, "loss": 0.0018, "step": 37422 }, { "epoch": 6.86, "learning_rate": 1.0411664905168484e-06, "loss": 0.0019, "step": 37423 }, { "epoch": 6.86, "learning_rate": 1.040836661458482e-06, "loss": 0.0026, "step": 37424 }, { "epoch": 6.86, "learning_rate": 1.0405068817832343e-06, "loss": 0.0024, "step": 37425 }, { "epoch": 6.86, "learning_rate": 1.0401771514929226e-06, "loss": 0.0014, "step": 37426 }, { "epoch": 6.86, "learning_rate": 1.0398474705893634e-06, "loss": 0.0011, "step": 37427 }, { "epoch": 6.86, "learning_rate": 1.0395178390743754e-06, "loss": 0.0017, "step": 37428 }, { "epoch": 6.86, "learning_rate": 1.0391882569497758e-06, "loss": 0.0026, "step": 37429 }, { "epoch": 6.86, "learning_rate": 1.0388587242173776e-06, "loss": 0.0018, "step": 37430 }, { "epoch": 6.86, "learning_rate": 1.0385292408790015e-06, "loss": 0.0023, "step": 37431 }, { "epoch": 6.86, "learning_rate": 1.0381998069364652e-06, "loss": 0.0006, "step": 37432 }, { "epoch": 6.86, "learning_rate": 1.0378704223915792e-06, "loss": 0.0012, "step": 37433 }, { "epoch": 6.86, "learning_rate": 1.0375410872461589e-06, "loss": 0.0022, "step": 37434 }, { "epoch": 6.87, "learning_rate": 1.037211801502025e-06, "loss": 0.003, "step": 37435 }, { "epoch": 6.87, "learning_rate": 1.0368825651609893e-06, "loss": 0.0017, "step": 37436 }, { "epoch": 6.87, "learning_rate": 1.0365533782248672e-06, "loss": 0.0013, "step": 37437 }, { "epoch": 6.87, "learning_rate": 1.0362242406954737e-06, "loss": 0.0013, "step": 37438 }, { "epoch": 6.87, "learning_rate": 1.0358951525746186e-06, "loss": 0.0019, "step": 37439 }, { "epoch": 6.87, "learning_rate": 1.0355661138641226e-06, "loss": 0.001, "step": 37440 }, { "epoch": 6.87, "learning_rate": 1.0352371245657956e-06, "loss": 0.0009, "step": 37441 }, { "epoch": 6.87, "learning_rate": 1.0349081846814502e-06, "loss": 0.0029, "step": 37442 }, { "epoch": 6.87, "learning_rate": 1.0345792942129029e-06, "loss": 0.0028, "step": 37443 }, { "epoch": 6.87, "learning_rate": 1.0342504531619634e-06, "loss": 0.0016, "step": 37444 }, { "epoch": 6.87, "learning_rate": 1.033921661530446e-06, "loss": 0.002, "step": 37445 }, { "epoch": 6.87, "learning_rate": 1.0335929193201621e-06, "loss": 0.0017, "step": 37446 }, { "epoch": 6.87, "learning_rate": 1.0332642265329217e-06, "loss": 0.0016, "step": 37447 }, { "epoch": 6.87, "learning_rate": 1.0329355831705424e-06, "loss": 0.0011, "step": 37448 }, { "epoch": 6.87, "learning_rate": 1.0326069892348322e-06, "loss": 0.0008, "step": 37449 }, { "epoch": 6.87, "learning_rate": 1.0322784447276024e-06, "loss": 0.0015, "step": 37450 }, { "epoch": 6.87, "learning_rate": 1.0319499496506624e-06, "loss": 0.0015, "step": 37451 }, { "epoch": 6.87, "learning_rate": 1.031621504005823e-06, "loss": 0.0011, "step": 37452 }, { "epoch": 6.87, "learning_rate": 1.0312931077948983e-06, "loss": 0.0011, "step": 37453 }, { "epoch": 6.87, "learning_rate": 1.030964761019696e-06, "loss": 0.0034, "step": 37454 }, { "epoch": 6.87, "learning_rate": 1.0306364636820254e-06, "loss": 0.0017, "step": 37455 }, { "epoch": 6.87, "learning_rate": 1.0303082157836975e-06, "loss": 0.0019, "step": 37456 }, { "epoch": 6.87, "learning_rate": 1.0299800173265195e-06, "loss": 0.0012, "step": 37457 }, { "epoch": 6.87, "learning_rate": 1.0296518683123035e-06, "loss": 0.0014, "step": 37458 }, { "epoch": 6.87, "learning_rate": 1.0293237687428558e-06, "loss": 0.0013, "step": 37459 }, { "epoch": 6.87, "learning_rate": 1.028995718619985e-06, "loss": 0.0025, "step": 37460 }, { "epoch": 6.87, "learning_rate": 1.0286677179455006e-06, "loss": 0.0014, "step": 37461 }, { "epoch": 6.87, "learning_rate": 1.0283397667212102e-06, "loss": 0.0014, "step": 37462 }, { "epoch": 6.87, "learning_rate": 1.0280118649489213e-06, "loss": 0.0017, "step": 37463 }, { "epoch": 6.87, "learning_rate": 1.027684012630441e-06, "loss": 0.0025, "step": 37464 }, { "epoch": 6.87, "learning_rate": 1.0273562097675738e-06, "loss": 0.0017, "step": 37465 }, { "epoch": 6.87, "learning_rate": 1.0270284563621324e-06, "loss": 0.0018, "step": 37466 }, { "epoch": 6.87, "learning_rate": 1.0267007524159222e-06, "loss": 0.0017, "step": 37467 }, { "epoch": 6.87, "learning_rate": 1.0263730979307452e-06, "loss": 0.0028, "step": 37468 }, { "epoch": 6.87, "learning_rate": 1.0260454929084086e-06, "loss": 0.0016, "step": 37469 }, { "epoch": 6.87, "learning_rate": 1.025717937350721e-06, "loss": 0.0016, "step": 37470 }, { "epoch": 6.87, "learning_rate": 1.0253904312594865e-06, "loss": 0.0018, "step": 37471 }, { "epoch": 6.87, "learning_rate": 1.0250629746365094e-06, "loss": 0.0013, "step": 37472 }, { "epoch": 6.87, "learning_rate": 1.0247355674835968e-06, "loss": 0.0012, "step": 37473 }, { "epoch": 6.87, "learning_rate": 1.024408209802551e-06, "loss": 0.0017, "step": 37474 }, { "epoch": 6.87, "learning_rate": 1.024080901595178e-06, "loss": 0.002, "step": 37475 }, { "epoch": 6.87, "learning_rate": 1.023753642863281e-06, "loss": 0.0021, "step": 37476 }, { "epoch": 6.87, "learning_rate": 1.0234264336086642e-06, "loss": 0.0015, "step": 37477 }, { "epoch": 6.87, "learning_rate": 1.0230992738331302e-06, "loss": 0.0018, "step": 37478 }, { "epoch": 6.87, "learning_rate": 1.0227721635384856e-06, "loss": 0.002, "step": 37479 }, { "epoch": 6.87, "learning_rate": 1.0224451027265303e-06, "loss": 0.0022, "step": 37480 }, { "epoch": 6.87, "learning_rate": 1.022118091399068e-06, "loss": 0.0012, "step": 37481 }, { "epoch": 6.87, "learning_rate": 1.0217911295579008e-06, "loss": 0.0027, "step": 37482 }, { "epoch": 6.87, "learning_rate": 1.0214642172048295e-06, "loss": 0.0016, "step": 37483 }, { "epoch": 6.87, "learning_rate": 1.0211373543416592e-06, "loss": 0.0016, "step": 37484 }, { "epoch": 6.87, "learning_rate": 1.020810540970192e-06, "loss": 0.001, "step": 37485 }, { "epoch": 6.87, "learning_rate": 1.020483777092226e-06, "loss": 0.0021, "step": 37486 }, { "epoch": 6.87, "learning_rate": 1.0201570627095613e-06, "loss": 0.0017, "step": 37487 }, { "epoch": 6.87, "learning_rate": 1.0198303978240042e-06, "loss": 0.0012, "step": 37488 }, { "epoch": 6.87, "learning_rate": 1.0195037824373522e-06, "loss": 0.0017, "step": 37489 }, { "epoch": 6.88, "learning_rate": 1.0191772165514047e-06, "loss": 0.0017, "step": 37490 }, { "epoch": 6.88, "learning_rate": 1.0188507001679637e-06, "loss": 0.0016, "step": 37491 }, { "epoch": 6.88, "learning_rate": 1.0185242332888279e-06, "loss": 0.0017, "step": 37492 }, { "epoch": 6.88, "learning_rate": 1.0181978159157968e-06, "loss": 0.0016, "step": 37493 }, { "epoch": 6.88, "learning_rate": 1.017871448050669e-06, "loss": 0.0018, "step": 37494 }, { "epoch": 6.88, "learning_rate": 1.0175451296952454e-06, "loss": 0.002, "step": 37495 }, { "epoch": 6.88, "learning_rate": 1.0172188608513234e-06, "loss": 0.0008, "step": 37496 }, { "epoch": 6.88, "learning_rate": 1.0168926415207004e-06, "loss": 0.0012, "step": 37497 }, { "epoch": 6.88, "learning_rate": 1.016566471705176e-06, "loss": 0.0033, "step": 37498 }, { "epoch": 6.88, "learning_rate": 1.0162403514065488e-06, "loss": 0.0017, "step": 37499 }, { "epoch": 6.88, "learning_rate": 1.0159142806266121e-06, "loss": 0.0017, "step": 37500 }, { "epoch": 6.88, "learning_rate": 1.0155882593671684e-06, "loss": 0.0017, "step": 37501 }, { "epoch": 6.88, "learning_rate": 1.0152622876300133e-06, "loss": 0.0018, "step": 37502 }, { "epoch": 6.88, "learning_rate": 1.0149363654169442e-06, "loss": 0.0035, "step": 37503 }, { "epoch": 6.88, "learning_rate": 1.014610492729755e-06, "loss": 0.0016, "step": 37504 }, { "epoch": 6.88, "learning_rate": 1.0142846695702414e-06, "loss": 0.0016, "step": 37505 }, { "epoch": 6.88, "learning_rate": 1.0139588959402026e-06, "loss": 0.0019, "step": 37506 }, { "epoch": 6.88, "learning_rate": 1.013633171841434e-06, "loss": 0.0015, "step": 37507 }, { "epoch": 6.88, "learning_rate": 1.0133074972757284e-06, "loss": 0.0023, "step": 37508 }, { "epoch": 6.88, "learning_rate": 1.0129818722448825e-06, "loss": 0.0018, "step": 37509 }, { "epoch": 6.88, "learning_rate": 1.0126562967506926e-06, "loss": 0.0012, "step": 37510 }, { "epoch": 6.88, "learning_rate": 1.0123307707949504e-06, "loss": 0.0018, "step": 37511 }, { "epoch": 6.88, "learning_rate": 1.0120052943794523e-06, "loss": 0.0012, "step": 37512 }, { "epoch": 6.88, "learning_rate": 1.0116798675059913e-06, "loss": 0.0022, "step": 37513 }, { "epoch": 6.88, "learning_rate": 1.0113544901763606e-06, "loss": 0.0028, "step": 37514 }, { "epoch": 6.88, "learning_rate": 1.0110291623923552e-06, "loss": 0.0015, "step": 37515 }, { "epoch": 6.88, "learning_rate": 1.0107038841557681e-06, "loss": 0.0019, "step": 37516 }, { "epoch": 6.88, "learning_rate": 1.0103786554683915e-06, "loss": 0.002, "step": 37517 }, { "epoch": 6.88, "learning_rate": 1.0100534763320158e-06, "loss": 0.002, "step": 37518 }, { "epoch": 6.88, "learning_rate": 1.0097283467484387e-06, "loss": 0.0007, "step": 37519 }, { "epoch": 6.88, "learning_rate": 1.0094032667194497e-06, "loss": 0.0013, "step": 37520 }, { "epoch": 6.88, "learning_rate": 1.0090782362468422e-06, "loss": 0.002, "step": 37521 }, { "epoch": 6.88, "learning_rate": 1.008753255332402e-06, "loss": 0.0008, "step": 37522 }, { "epoch": 6.88, "learning_rate": 1.0084283239779257e-06, "loss": 0.0015, "step": 37523 }, { "epoch": 6.88, "learning_rate": 1.0081034421852043e-06, "loss": 0.0025, "step": 37524 }, { "epoch": 6.88, "learning_rate": 1.0077786099560272e-06, "loss": 0.0011, "step": 37525 }, { "epoch": 6.88, "learning_rate": 1.0074538272921853e-06, "loss": 0.0022, "step": 37526 }, { "epoch": 6.88, "learning_rate": 1.007129094195468e-06, "loss": 0.0007, "step": 37527 }, { "epoch": 6.88, "learning_rate": 1.0068044106676656e-06, "loss": 0.0019, "step": 37528 }, { "epoch": 6.88, "learning_rate": 1.0064797767105682e-06, "loss": 0.0016, "step": 37529 }, { "epoch": 6.88, "learning_rate": 1.0061551923259648e-06, "loss": 0.0015, "step": 37530 }, { "epoch": 6.88, "learning_rate": 1.0058306575156441e-06, "loss": 0.0035, "step": 37531 }, { "epoch": 6.88, "learning_rate": 1.0055061722813974e-06, "loss": 0.0012, "step": 37532 }, { "epoch": 6.88, "learning_rate": 1.0051817366250105e-06, "loss": 0.0012, "step": 37533 }, { "epoch": 6.88, "learning_rate": 1.0048573505482728e-06, "loss": 0.0016, "step": 37534 }, { "epoch": 6.88, "learning_rate": 1.004533014052972e-06, "loss": 0.0017, "step": 37535 }, { "epoch": 6.88, "learning_rate": 1.004208727140894e-06, "loss": 0.0016, "step": 37536 }, { "epoch": 6.88, "learning_rate": 1.00388448981383e-06, "loss": 0.0008, "step": 37537 }, { "epoch": 6.88, "learning_rate": 1.0035603020735652e-06, "loss": 0.0029, "step": 37538 }, { "epoch": 6.88, "learning_rate": 1.003236163921889e-06, "loss": 0.0019, "step": 37539 }, { "epoch": 6.88, "learning_rate": 1.0029120753605826e-06, "loss": 0.0014, "step": 37540 }, { "epoch": 6.88, "learning_rate": 1.0025880363914363e-06, "loss": 0.0015, "step": 37541 }, { "epoch": 6.88, "learning_rate": 1.002264047016236e-06, "loss": 0.0015, "step": 37542 }, { "epoch": 6.88, "learning_rate": 1.0019401072367673e-06, "loss": 0.0018, "step": 37543 }, { "epoch": 6.89, "learning_rate": 1.001616217054816e-06, "loss": 0.0014, "step": 37544 }, { "epoch": 6.89, "learning_rate": 1.001292376472166e-06, "loss": 0.0024, "step": 37545 }, { "epoch": 6.89, "learning_rate": 1.0009685854906037e-06, "loss": 0.0019, "step": 37546 }, { "epoch": 6.89, "learning_rate": 1.0006448441119132e-06, "loss": 0.0013, "step": 37547 }, { "epoch": 6.89, "learning_rate": 1.0003211523378798e-06, "loss": 0.001, "step": 37548 }, { "epoch": 6.89, "learning_rate": 9.999975101702841e-07, "loss": 0.0018, "step": 37549 }, { "epoch": 6.89, "learning_rate": 9.99673917610917e-07, "loss": 0.0016, "step": 37550 }, { "epoch": 6.89, "learning_rate": 9.993503746615574e-07, "loss": 0.001, "step": 37551 }, { "epoch": 6.89, "learning_rate": 9.990268813239879e-07, "loss": 0.0017, "step": 37552 }, { "epoch": 6.89, "learning_rate": 9.987034375999915e-07, "loss": 0.0013, "step": 37553 }, { "epoch": 6.89, "learning_rate": 9.983800434913549e-07, "loss": 0.0018, "step": 37554 }, { "epoch": 6.89, "learning_rate": 9.980566989998585e-07, "loss": 0.0016, "step": 37555 }, { "epoch": 6.89, "learning_rate": 9.977334041272869e-07, "loss": 0.0021, "step": 37556 }, { "epoch": 6.89, "learning_rate": 9.97410158875417e-07, "loss": 0.001, "step": 37557 }, { "epoch": 6.89, "learning_rate": 9.97086963246031e-07, "loss": 0.0027, "step": 37558 }, { "epoch": 6.89, "learning_rate": 9.967638172409144e-07, "loss": 0.0017, "step": 37559 }, { "epoch": 6.89, "learning_rate": 9.964407208618476e-07, "loss": 0.0011, "step": 37560 }, { "epoch": 6.89, "learning_rate": 9.961176741106094e-07, "loss": 0.0012, "step": 37561 }, { "epoch": 6.89, "learning_rate": 9.957946769889826e-07, "loss": 0.0014, "step": 37562 }, { "epoch": 6.89, "learning_rate": 9.95471729498746e-07, "loss": 0.0015, "step": 37563 }, { "epoch": 6.89, "learning_rate": 9.951488316416791e-07, "loss": 0.0023, "step": 37564 }, { "epoch": 6.89, "learning_rate": 9.94825983419564e-07, "loss": 0.0015, "step": 37565 }, { "epoch": 6.89, "learning_rate": 9.94503184834179e-07, "loss": 0.0017, "step": 37566 }, { "epoch": 6.89, "learning_rate": 9.94180435887302e-07, "loss": 0.001, "step": 37567 }, { "epoch": 6.89, "learning_rate": 9.938577365807155e-07, "loss": 0.0011, "step": 37568 }, { "epoch": 6.89, "learning_rate": 9.935350869161965e-07, "loss": 0.0014, "step": 37569 }, { "epoch": 6.89, "learning_rate": 9.932124868955218e-07, "loss": 0.002, "step": 37570 }, { "epoch": 6.89, "learning_rate": 9.92889936520468e-07, "loss": 0.002, "step": 37571 }, { "epoch": 6.89, "learning_rate": 9.925674357928194e-07, "loss": 0.0015, "step": 37572 }, { "epoch": 6.89, "learning_rate": 9.9224498471435e-07, "loss": 0.0012, "step": 37573 }, { "epoch": 6.89, "learning_rate": 9.919225832868384e-07, "loss": 0.0012, "step": 37574 }, { "epoch": 6.89, "learning_rate": 9.916002315120577e-07, "loss": 0.001, "step": 37575 }, { "epoch": 6.89, "learning_rate": 9.912779293917885e-07, "loss": 0.0017, "step": 37576 }, { "epoch": 6.89, "learning_rate": 9.90955676927806e-07, "loss": 0.0007, "step": 37577 }, { "epoch": 6.89, "learning_rate": 9.906334741218882e-07, "loss": 0.0014, "step": 37578 }, { "epoch": 6.89, "learning_rate": 9.903113209758098e-07, "loss": 0.0012, "step": 37579 }, { "epoch": 6.89, "learning_rate": 9.89989217491345e-07, "loss": 0.0015, "step": 37580 }, { "epoch": 6.89, "learning_rate": 9.896671636702716e-07, "loss": 0.0016, "step": 37581 }, { "epoch": 6.89, "learning_rate": 9.893451595143644e-07, "loss": 0.0012, "step": 37582 }, { "epoch": 6.89, "learning_rate": 9.890232050253968e-07, "loss": 0.0025, "step": 37583 }, { "epoch": 6.89, "learning_rate": 9.887013002051438e-07, "loss": 0.0018, "step": 37584 }, { "epoch": 6.89, "learning_rate": 9.883794450553807e-07, "loss": 0.0013, "step": 37585 }, { "epoch": 6.89, "learning_rate": 9.88057639577884e-07, "loss": 0.0013, "step": 37586 }, { "epoch": 6.89, "learning_rate": 9.877358837744222e-07, "loss": 0.0018, "step": 37587 }, { "epoch": 6.89, "learning_rate": 9.874141776467727e-07, "loss": 0.002, "step": 37588 }, { "epoch": 6.89, "learning_rate": 9.870925211967043e-07, "loss": 0.0013, "step": 37589 }, { "epoch": 6.89, "learning_rate": 9.86770914425995e-07, "loss": 0.0012, "step": 37590 }, { "epoch": 6.89, "learning_rate": 9.864493573364165e-07, "loss": 0.0013, "step": 37591 }, { "epoch": 6.89, "learning_rate": 9.861278499297411e-07, "loss": 0.0029, "step": 37592 }, { "epoch": 6.89, "learning_rate": 9.858063922077366e-07, "loss": 0.0016, "step": 37593 }, { "epoch": 6.89, "learning_rate": 9.854849841721803e-07, "loss": 0.0021, "step": 37594 }, { "epoch": 6.89, "learning_rate": 9.851636258248432e-07, "loss": 0.0019, "step": 37595 }, { "epoch": 6.89, "learning_rate": 9.848423171674937e-07, "loss": 0.0031, "step": 37596 }, { "epoch": 6.89, "learning_rate": 9.84521058201905e-07, "loss": 0.002, "step": 37597 }, { "epoch": 6.89, "learning_rate": 9.841998489298477e-07, "loss": 0.0017, "step": 37598 }, { "epoch": 6.9, "learning_rate": 9.838786893530928e-07, "loss": 0.0008, "step": 37599 }, { "epoch": 6.9, "learning_rate": 9.835575794734088e-07, "loss": 0.0025, "step": 37600 }, { "epoch": 6.9, "learning_rate": 9.832365192925664e-07, "loss": 0.0017, "step": 37601 }, { "epoch": 6.9, "learning_rate": 9.829155088123343e-07, "loss": 0.0015, "step": 37602 }, { "epoch": 6.9, "learning_rate": 9.825945480344868e-07, "loss": 0.0011, "step": 37603 }, { "epoch": 6.9, "learning_rate": 9.822736369607878e-07, "loss": 0.0022, "step": 37604 }, { "epoch": 6.9, "learning_rate": 9.81952775593007e-07, "loss": 0.0012, "step": 37605 }, { "epoch": 6.9, "learning_rate": 9.81631963932913e-07, "loss": 0.0021, "step": 37606 }, { "epoch": 6.9, "learning_rate": 9.813112019822756e-07, "loss": 0.0014, "step": 37607 }, { "epoch": 6.9, "learning_rate": 9.809904897428635e-07, "loss": 0.0022, "step": 37608 }, { "epoch": 6.9, "learning_rate": 9.806698272164416e-07, "loss": 0.0017, "step": 37609 }, { "epoch": 6.9, "learning_rate": 9.803492144047822e-07, "loss": 0.0014, "step": 37610 }, { "epoch": 6.9, "learning_rate": 9.800286513096445e-07, "loss": 0.002, "step": 37611 }, { "epoch": 6.9, "learning_rate": 9.797081379328021e-07, "loss": 0.0014, "step": 37612 }, { "epoch": 6.9, "learning_rate": 9.7938767427602e-07, "loss": 0.0011, "step": 37613 }, { "epoch": 6.9, "learning_rate": 9.790672603410645e-07, "loss": 0.0015, "step": 37614 }, { "epoch": 6.9, "learning_rate": 9.787468961297008e-07, "loss": 0.0012, "step": 37615 }, { "epoch": 6.9, "learning_rate": 9.784265816436965e-07, "loss": 0.0016, "step": 37616 }, { "epoch": 6.9, "learning_rate": 9.78106316884816e-07, "loss": 0.0012, "step": 37617 }, { "epoch": 6.9, "learning_rate": 9.77786101854825e-07, "loss": 0.0015, "step": 37618 }, { "epoch": 6.9, "learning_rate": 9.774659365554883e-07, "loss": 0.0013, "step": 37619 }, { "epoch": 6.9, "learning_rate": 9.771458209885687e-07, "loss": 0.0019, "step": 37620 }, { "epoch": 6.9, "learning_rate": 9.76825755155837e-07, "loss": 0.001, "step": 37621 }, { "epoch": 6.9, "learning_rate": 9.765057390590505e-07, "loss": 0.0017, "step": 37622 }, { "epoch": 6.9, "learning_rate": 9.761857726999757e-07, "loss": 0.0015, "step": 37623 }, { "epoch": 6.9, "learning_rate": 9.758658560803746e-07, "loss": 0.0018, "step": 37624 }, { "epoch": 6.9, "learning_rate": 9.755459892020147e-07, "loss": 0.0019, "step": 37625 }, { "epoch": 6.9, "learning_rate": 9.752261720666567e-07, "loss": 0.0018, "step": 37626 }, { "epoch": 6.9, "learning_rate": 9.749064046760626e-07, "loss": 0.003, "step": 37627 }, { "epoch": 6.9, "learning_rate": 9.745866870319965e-07, "loss": 0.0011, "step": 37628 }, { "epoch": 6.9, "learning_rate": 9.742670191362203e-07, "loss": 0.0027, "step": 37629 }, { "epoch": 6.9, "learning_rate": 9.739474009904948e-07, "loss": 0.0016, "step": 37630 }, { "epoch": 6.9, "learning_rate": 9.736278325965831e-07, "loss": 0.0014, "step": 37631 }, { "epoch": 6.9, "learning_rate": 9.73308313956247e-07, "loss": 0.002, "step": 37632 }, { "epoch": 6.9, "learning_rate": 9.729888450712466e-07, "loss": 0.0015, "step": 37633 }, { "epoch": 6.9, "learning_rate": 9.726694259433423e-07, "loss": 0.001, "step": 37634 }, { "epoch": 6.9, "learning_rate": 9.72350056574296e-07, "loss": 0.0015, "step": 37635 }, { "epoch": 6.9, "learning_rate": 9.720307369658677e-07, "loss": 0.0007, "step": 37636 }, { "epoch": 6.9, "learning_rate": 9.717114671198157e-07, "loss": 0.0018, "step": 37637 }, { "epoch": 6.9, "learning_rate": 9.713922470379044e-07, "loss": 0.0012, "step": 37638 }, { "epoch": 6.9, "learning_rate": 9.710730767218913e-07, "loss": 0.0031, "step": 37639 }, { "epoch": 6.9, "learning_rate": 9.707539561735335e-07, "loss": 0.0018, "step": 37640 }, { "epoch": 6.9, "learning_rate": 9.70434885394591e-07, "loss": 0.0025, "step": 37641 }, { "epoch": 6.9, "learning_rate": 9.701158643868213e-07, "loss": 0.0016, "step": 37642 }, { "epoch": 6.9, "learning_rate": 9.697968931519864e-07, "loss": 0.0026, "step": 37643 }, { "epoch": 6.9, "learning_rate": 9.694779716918434e-07, "loss": 0.0016, "step": 37644 }, { "epoch": 6.9, "learning_rate": 9.691591000081479e-07, "loss": 0.0018, "step": 37645 }, { "epoch": 6.9, "learning_rate": 9.688402781026595e-07, "loss": 0.0013, "step": 37646 }, { "epoch": 6.9, "learning_rate": 9.685215059771358e-07, "loss": 0.0017, "step": 37647 }, { "epoch": 6.9, "learning_rate": 9.68202783633333e-07, "loss": 0.0015, "step": 37648 }, { "epoch": 6.9, "learning_rate": 9.678841110730074e-07, "loss": 0.0024, "step": 37649 }, { "epoch": 6.9, "learning_rate": 9.675654882979168e-07, "loss": 0.0019, "step": 37650 }, { "epoch": 6.9, "learning_rate": 9.672469153098163e-07, "loss": 0.0015, "step": 37651 }, { "epoch": 6.9, "learning_rate": 9.669283921104623e-07, "loss": 0.0017, "step": 37652 }, { "epoch": 6.91, "learning_rate": 9.66609918701611e-07, "loss": 0.0009, "step": 37653 }, { "epoch": 6.91, "learning_rate": 9.662914950850177e-07, "loss": 0.0006, "step": 37654 }, { "epoch": 6.91, "learning_rate": 9.659731212624347e-07, "loss": 0.0014, "step": 37655 }, { "epoch": 6.91, "learning_rate": 9.656547972356223e-07, "loss": 0.0012, "step": 37656 }, { "epoch": 6.91, "learning_rate": 9.653365230063338e-07, "loss": 0.0017, "step": 37657 }, { "epoch": 6.91, "learning_rate": 9.650182985763201e-07, "loss": 0.0025, "step": 37658 }, { "epoch": 6.91, "learning_rate": 9.647001239473352e-07, "loss": 0.0011, "step": 37659 }, { "epoch": 6.91, "learning_rate": 9.643819991211368e-07, "loss": 0.0018, "step": 37660 }, { "epoch": 6.91, "learning_rate": 9.640639240994764e-07, "loss": 0.0021, "step": 37661 }, { "epoch": 6.91, "learning_rate": 9.637458988841076e-07, "loss": 0.0011, "step": 37662 }, { "epoch": 6.91, "learning_rate": 9.63427923476783e-07, "loss": 0.0029, "step": 37663 }, { "epoch": 6.91, "learning_rate": 9.631099978792557e-07, "loss": 0.0016, "step": 37664 }, { "epoch": 6.91, "learning_rate": 9.627921220932767e-07, "loss": 0.0012, "step": 37665 }, { "epoch": 6.91, "learning_rate": 9.624742961206002e-07, "loss": 0.0017, "step": 37666 }, { "epoch": 6.91, "learning_rate": 9.621565199629767e-07, "loss": 0.0014, "step": 37667 }, { "epoch": 6.91, "learning_rate": 9.618387936221585e-07, "loss": 0.0025, "step": 37668 }, { "epoch": 6.91, "learning_rate": 9.615211170998972e-07, "loss": 0.0023, "step": 37669 }, { "epoch": 6.91, "learning_rate": 9.612034903979428e-07, "loss": 0.0015, "step": 37670 }, { "epoch": 6.91, "learning_rate": 9.60885913518046e-07, "loss": 0.0025, "step": 37671 }, { "epoch": 6.91, "learning_rate": 9.605683864619574e-07, "loss": 0.0016, "step": 37672 }, { "epoch": 6.91, "learning_rate": 9.602509092314272e-07, "loss": 0.0014, "step": 37673 }, { "epoch": 6.91, "learning_rate": 9.59933481828207e-07, "loss": 0.0029, "step": 37674 }, { "epoch": 6.91, "learning_rate": 9.596161042540464e-07, "loss": 0.0025, "step": 37675 }, { "epoch": 6.91, "learning_rate": 9.592987765106932e-07, "loss": 0.0018, "step": 37676 }, { "epoch": 6.91, "learning_rate": 9.589814985998936e-07, "loss": 0.0023, "step": 37677 }, { "epoch": 6.91, "learning_rate": 9.586642705234028e-07, "loss": 0.0009, "step": 37678 }, { "epoch": 6.91, "learning_rate": 9.583470922829653e-07, "loss": 0.0012, "step": 37679 }, { "epoch": 6.91, "learning_rate": 9.580299638803314e-07, "loss": 0.0017, "step": 37680 }, { "epoch": 6.91, "learning_rate": 9.57712885317248e-07, "loss": 0.0012, "step": 37681 }, { "epoch": 6.91, "learning_rate": 9.573958565954622e-07, "loss": 0.0024, "step": 37682 }, { "epoch": 6.91, "learning_rate": 9.57078877716724e-07, "loss": 0.0017, "step": 37683 }, { "epoch": 6.91, "learning_rate": 9.567619486827773e-07, "loss": 0.0013, "step": 37684 }, { "epoch": 6.91, "learning_rate": 9.564450694953709e-07, "loss": 0.0012, "step": 37685 }, { "epoch": 6.91, "learning_rate": 9.56128240156251e-07, "loss": 0.0009, "step": 37686 }, { "epoch": 6.91, "learning_rate": 9.558114606671654e-07, "loss": 0.001, "step": 37687 }, { "epoch": 6.91, "learning_rate": 9.55494731029858e-07, "loss": 0.0011, "step": 37688 }, { "epoch": 6.91, "learning_rate": 9.55178051246075e-07, "loss": 0.0013, "step": 37689 }, { "epoch": 6.91, "learning_rate": 9.54861421317561e-07, "loss": 0.0011, "step": 37690 }, { "epoch": 6.91, "learning_rate": 9.545448412460644e-07, "loss": 0.0019, "step": 37691 }, { "epoch": 6.91, "learning_rate": 9.542283110333295e-07, "loss": 0.0013, "step": 37692 }, { "epoch": 6.91, "learning_rate": 9.539118306811013e-07, "loss": 0.0014, "step": 37693 }, { "epoch": 6.91, "learning_rate": 9.535954001911207e-07, "loss": 0.0024, "step": 37694 }, { "epoch": 6.91, "learning_rate": 9.532790195651331e-07, "loss": 0.0021, "step": 37695 }, { "epoch": 6.91, "learning_rate": 9.529626888048849e-07, "loss": 0.0012, "step": 37696 }, { "epoch": 6.91, "learning_rate": 9.526464079121178e-07, "loss": 0.0017, "step": 37697 }, { "epoch": 6.91, "learning_rate": 9.523301768885773e-07, "loss": 0.0035, "step": 37698 }, { "epoch": 6.91, "learning_rate": 9.52013995736003e-07, "loss": 0.0021, "step": 37699 }, { "epoch": 6.91, "learning_rate": 9.516978644561403e-07, "loss": 0.0012, "step": 37700 }, { "epoch": 6.91, "learning_rate": 9.51381783050731e-07, "loss": 0.0013, "step": 37701 }, { "epoch": 6.91, "learning_rate": 9.510657515215172e-07, "loss": 0.0014, "step": 37702 }, { "epoch": 6.91, "learning_rate": 9.507497698702395e-07, "loss": 0.001, "step": 37703 }, { "epoch": 6.91, "learning_rate": 9.504338380986445e-07, "loss": 0.0014, "step": 37704 }, { "epoch": 6.91, "learning_rate": 9.501179562084684e-07, "loss": 0.0013, "step": 37705 }, { "epoch": 6.91, "learning_rate": 9.498021242014544e-07, "loss": 0.0017, "step": 37706 }, { "epoch": 6.91, "learning_rate": 9.494863420793421e-07, "loss": 0.0012, "step": 37707 }, { "epoch": 6.92, "learning_rate": 9.491706098438724e-07, "loss": 0.001, "step": 37708 }, { "epoch": 6.92, "learning_rate": 9.488549274967873e-07, "loss": 0.0022, "step": 37709 }, { "epoch": 6.92, "learning_rate": 9.485392950398287e-07, "loss": 0.0021, "step": 37710 }, { "epoch": 6.92, "learning_rate": 9.482237124747306e-07, "loss": 0.0022, "step": 37711 }, { "epoch": 6.92, "learning_rate": 9.479081798032341e-07, "loss": 0.0026, "step": 37712 }, { "epoch": 6.92, "learning_rate": 9.475926970270821e-07, "loss": 0.0017, "step": 37713 }, { "epoch": 6.92, "learning_rate": 9.47277264148011e-07, "loss": 0.0016, "step": 37714 }, { "epoch": 6.92, "learning_rate": 9.469618811677594e-07, "loss": 0.0016, "step": 37715 }, { "epoch": 6.92, "learning_rate": 9.46646548088066e-07, "loss": 0.0009, "step": 37716 }, { "epoch": 6.92, "learning_rate": 9.463312649106693e-07, "loss": 0.0015, "step": 37717 }, { "epoch": 6.92, "learning_rate": 9.460160316373068e-07, "loss": 0.0022, "step": 37718 }, { "epoch": 6.92, "learning_rate": 9.45700848269716e-07, "loss": 0.001, "step": 37719 }, { "epoch": 6.92, "learning_rate": 9.453857148096335e-07, "loss": 0.001, "step": 37720 }, { "epoch": 6.92, "learning_rate": 9.450706312587954e-07, "loss": 0.0022, "step": 37721 }, { "epoch": 6.92, "learning_rate": 9.447555976189449e-07, "loss": 0.0019, "step": 37722 }, { "epoch": 6.92, "learning_rate": 9.444406138918106e-07, "loss": 0.0008, "step": 37723 }, { "epoch": 6.92, "learning_rate": 9.441256800791321e-07, "loss": 0.0011, "step": 37724 }, { "epoch": 6.92, "learning_rate": 9.438107961826437e-07, "loss": 0.0023, "step": 37725 }, { "epoch": 6.92, "learning_rate": 9.43495962204084e-07, "loss": 0.0015, "step": 37726 }, { "epoch": 6.92, "learning_rate": 9.43181178145186e-07, "loss": 0.0019, "step": 37727 }, { "epoch": 6.92, "learning_rate": 9.428664440076885e-07, "loss": 0.0008, "step": 37728 }, { "epoch": 6.92, "learning_rate": 9.425517597933209e-07, "loss": 0.0008, "step": 37729 }, { "epoch": 6.92, "learning_rate": 9.422371255038187e-07, "loss": 0.0018, "step": 37730 }, { "epoch": 6.92, "learning_rate": 9.419225411409195e-07, "loss": 0.0017, "step": 37731 }, { "epoch": 6.92, "learning_rate": 9.416080067063549e-07, "loss": 0.0026, "step": 37732 }, { "epoch": 6.92, "learning_rate": 9.412935222018593e-07, "loss": 0.0019, "step": 37733 }, { "epoch": 6.92, "learning_rate": 9.409790876291658e-07, "loss": 0.0009, "step": 37734 }, { "epoch": 6.92, "learning_rate": 9.406647029900084e-07, "loss": 0.0013, "step": 37735 }, { "epoch": 6.92, "learning_rate": 9.403503682861192e-07, "loss": 0.0011, "step": 37736 }, { "epoch": 6.92, "learning_rate": 9.400360835192312e-07, "loss": 0.0015, "step": 37737 }, { "epoch": 6.92, "learning_rate": 9.397218486910764e-07, "loss": 0.0008, "step": 37738 }, { "epoch": 6.92, "learning_rate": 9.394076638033856e-07, "loss": 0.0028, "step": 37739 }, { "epoch": 6.92, "learning_rate": 9.39093528857895e-07, "loss": 0.0015, "step": 37740 }, { "epoch": 6.92, "learning_rate": 9.387794438563313e-07, "loss": 0.0011, "step": 37741 }, { "epoch": 6.92, "learning_rate": 9.384654088004275e-07, "loss": 0.002, "step": 37742 }, { "epoch": 6.92, "learning_rate": 9.381514236919131e-07, "loss": 0.002, "step": 37743 }, { "epoch": 6.92, "learning_rate": 9.378374885325225e-07, "loss": 0.0015, "step": 37744 }, { "epoch": 6.92, "learning_rate": 9.375236033239843e-07, "loss": 0.0026, "step": 37745 }, { "epoch": 6.92, "learning_rate": 9.372097680680292e-07, "loss": 0.003, "step": 37746 }, { "epoch": 6.92, "learning_rate": 9.368959827663848e-07, "loss": 0.0015, "step": 37747 }, { "epoch": 6.92, "learning_rate": 9.365822474207797e-07, "loss": 0.0019, "step": 37748 }, { "epoch": 6.92, "learning_rate": 9.362685620329481e-07, "loss": 0.0009, "step": 37749 }, { "epoch": 6.92, "learning_rate": 9.359549266046164e-07, "loss": 0.0018, "step": 37750 }, { "epoch": 6.92, "learning_rate": 9.356413411375132e-07, "loss": 0.0021, "step": 37751 }, { "epoch": 6.92, "learning_rate": 9.35327805633367e-07, "loss": 0.0011, "step": 37752 }, { "epoch": 6.92, "learning_rate": 9.350143200939077e-07, "loss": 0.0016, "step": 37753 }, { "epoch": 6.92, "learning_rate": 9.347008845208605e-07, "loss": 0.001, "step": 37754 }, { "epoch": 6.92, "learning_rate": 9.343874989159551e-07, "loss": 0.0023, "step": 37755 }, { "epoch": 6.92, "learning_rate": 9.340741632809158e-07, "loss": 0.0011, "step": 37756 }, { "epoch": 6.92, "learning_rate": 9.337608776174756e-07, "loss": 0.0007, "step": 37757 }, { "epoch": 6.92, "learning_rate": 9.334476419273564e-07, "loss": 0.0015, "step": 37758 }, { "epoch": 6.92, "learning_rate": 9.331344562122868e-07, "loss": 0.0016, "step": 37759 }, { "epoch": 6.92, "learning_rate": 9.328213204739923e-07, "loss": 0.0017, "step": 37760 }, { "epoch": 6.92, "learning_rate": 9.325082347141967e-07, "loss": 0.0023, "step": 37761 }, { "epoch": 6.92, "learning_rate": 9.3219519893463e-07, "loss": 0.0018, "step": 37762 }, { "epoch": 6.93, "learning_rate": 9.318822131370153e-07, "loss": 0.0006, "step": 37763 }, { "epoch": 6.93, "learning_rate": 9.315692773230811e-07, "loss": 0.0013, "step": 37764 }, { "epoch": 6.93, "learning_rate": 9.312563914945461e-07, "loss": 0.0015, "step": 37765 }, { "epoch": 6.93, "learning_rate": 9.309435556531387e-07, "loss": 0.0014, "step": 37766 }, { "epoch": 6.93, "learning_rate": 9.306307698005846e-07, "loss": 0.0016, "step": 37767 }, { "epoch": 6.93, "learning_rate": 9.303180339386054e-07, "loss": 0.002, "step": 37768 }, { "epoch": 6.93, "learning_rate": 9.300053480689253e-07, "loss": 0.0021, "step": 37769 }, { "epoch": 6.93, "learning_rate": 9.296927121932687e-07, "loss": 0.0017, "step": 37770 }, { "epoch": 6.93, "learning_rate": 9.293801263133584e-07, "loss": 0.002, "step": 37771 }, { "epoch": 6.93, "learning_rate": 9.290675904309166e-07, "loss": 0.0013, "step": 37772 }, { "epoch": 6.93, "learning_rate": 9.287551045476672e-07, "loss": 0.0018, "step": 37773 }, { "epoch": 6.93, "learning_rate": 9.284426686653303e-07, "loss": 0.0008, "step": 37774 }, { "epoch": 6.93, "learning_rate": 9.281302827856342e-07, "loss": 0.0017, "step": 37775 }, { "epoch": 6.93, "learning_rate": 9.278179469102944e-07, "loss": 0.0013, "step": 37776 }, { "epoch": 6.93, "learning_rate": 9.275056610410338e-07, "loss": 0.0014, "step": 37777 }, { "epoch": 6.93, "learning_rate": 9.271934251795734e-07, "loss": 0.0015, "step": 37778 }, { "epoch": 6.93, "learning_rate": 9.268812393276372e-07, "loss": 0.0028, "step": 37779 }, { "epoch": 6.93, "learning_rate": 9.26569103486944e-07, "loss": 0.0018, "step": 37780 }, { "epoch": 6.93, "learning_rate": 9.262570176592145e-07, "loss": 0.0012, "step": 37781 }, { "epoch": 6.93, "learning_rate": 9.259449818461708e-07, "loss": 0.002, "step": 37782 }, { "epoch": 6.93, "learning_rate": 9.256329960495281e-07, "loss": 0.0012, "step": 37783 }, { "epoch": 6.93, "learning_rate": 9.253210602710095e-07, "loss": 0.0019, "step": 37784 }, { "epoch": 6.93, "learning_rate": 9.250091745123346e-07, "loss": 0.0016, "step": 37785 }, { "epoch": 6.93, "learning_rate": 9.246973387752212e-07, "loss": 0.002, "step": 37786 }, { "epoch": 6.93, "learning_rate": 9.243855530613888e-07, "loss": 0.0022, "step": 37787 }, { "epoch": 6.93, "learning_rate": 9.240738173725561e-07, "loss": 0.0016, "step": 37788 }, { "epoch": 6.93, "learning_rate": 9.237621317104406e-07, "loss": 0.0013, "step": 37789 }, { "epoch": 6.93, "learning_rate": 9.23450496076761e-07, "loss": 0.0023, "step": 37790 }, { "epoch": 6.93, "learning_rate": 9.231389104732358e-07, "loss": 0.0022, "step": 37791 }, { "epoch": 6.93, "learning_rate": 9.228273749015793e-07, "loss": 0.0027, "step": 37792 }, { "epoch": 6.93, "learning_rate": 9.225158893635145e-07, "loss": 0.0019, "step": 37793 }, { "epoch": 6.93, "learning_rate": 9.222044538607522e-07, "loss": 0.0015, "step": 37794 }, { "epoch": 6.93, "learning_rate": 9.218930683950123e-07, "loss": 0.0013, "step": 37795 }, { "epoch": 6.93, "learning_rate": 9.215817329680087e-07, "loss": 0.0018, "step": 37796 }, { "epoch": 6.93, "learning_rate": 9.212704475814616e-07, "loss": 0.0017, "step": 37797 }, { "epoch": 6.93, "learning_rate": 9.209592122370836e-07, "loss": 0.0019, "step": 37798 }, { "epoch": 6.93, "learning_rate": 9.206480269365914e-07, "loss": 0.0013, "step": 37799 }, { "epoch": 6.93, "learning_rate": 9.203368916817012e-07, "loss": 0.0009, "step": 37800 }, { "epoch": 6.93, "learning_rate": 9.200258064741241e-07, "loss": 0.0022, "step": 37801 }, { "epoch": 6.93, "learning_rate": 9.197147713155797e-07, "loss": 0.0014, "step": 37802 }, { "epoch": 6.93, "learning_rate": 9.194037862077787e-07, "loss": 0.0015, "step": 37803 }, { "epoch": 6.93, "learning_rate": 9.190928511524377e-07, "loss": 0.0012, "step": 37804 }, { "epoch": 6.93, "learning_rate": 9.187819661512698e-07, "loss": 0.0014, "step": 37805 }, { "epoch": 6.93, "learning_rate": 9.18471131205988e-07, "loss": 0.0019, "step": 37806 }, { "epoch": 6.93, "learning_rate": 9.181603463183064e-07, "loss": 0.0016, "step": 37807 }, { "epoch": 6.93, "learning_rate": 9.178496114899371e-07, "loss": 0.0017, "step": 37808 }, { "epoch": 6.93, "learning_rate": 9.17538926722592e-07, "loss": 0.0019, "step": 37809 }, { "epoch": 6.93, "learning_rate": 9.172282920179876e-07, "loss": 0.0017, "step": 37810 }, { "epoch": 6.93, "learning_rate": 9.169177073778346e-07, "loss": 0.0012, "step": 37811 }, { "epoch": 6.93, "learning_rate": 9.166071728038429e-07, "loss": 0.002, "step": 37812 }, { "epoch": 6.93, "learning_rate": 9.162966882977254e-07, "loss": 0.0013, "step": 37813 }, { "epoch": 6.93, "learning_rate": 9.159862538611908e-07, "loss": 0.0018, "step": 37814 }, { "epoch": 6.93, "learning_rate": 9.156758694959544e-07, "loss": 0.0019, "step": 37815 }, { "epoch": 6.93, "learning_rate": 9.153655352037261e-07, "loss": 0.0009, "step": 37816 }, { "epoch": 6.94, "learning_rate": 9.150552509862154e-07, "loss": 0.0016, "step": 37817 }, { "epoch": 6.94, "learning_rate": 9.147450168451333e-07, "loss": 0.0015, "step": 37818 }, { "epoch": 6.94, "learning_rate": 9.144348327821895e-07, "loss": 0.0022, "step": 37819 }, { "epoch": 6.94, "learning_rate": 9.141246987990937e-07, "loss": 0.0015, "step": 37820 }, { "epoch": 6.94, "learning_rate": 9.138146148975547e-07, "loss": 0.0023, "step": 37821 }, { "epoch": 6.94, "learning_rate": 9.13504581079283e-07, "loss": 0.0014, "step": 37822 }, { "epoch": 6.94, "learning_rate": 9.131945973459877e-07, "loss": 0.0014, "step": 37823 }, { "epoch": 6.94, "learning_rate": 9.12884663699376e-07, "loss": 0.0026, "step": 37824 }, { "epoch": 6.94, "learning_rate": 9.125747801411566e-07, "loss": 0.0018, "step": 37825 }, { "epoch": 6.94, "learning_rate": 9.122649466730382e-07, "loss": 0.0014, "step": 37826 }, { "epoch": 6.94, "learning_rate": 9.11955163296726e-07, "loss": 0.0025, "step": 37827 }, { "epoch": 6.94, "learning_rate": 9.116454300139321e-07, "loss": 0.0017, "step": 37828 }, { "epoch": 6.94, "learning_rate": 9.113357468263639e-07, "loss": 0.0009, "step": 37829 }, { "epoch": 6.94, "learning_rate": 9.110261137357235e-07, "loss": 0.0034, "step": 37830 }, { "epoch": 6.94, "learning_rate": 9.107165307437183e-07, "loss": 0.0021, "step": 37831 }, { "epoch": 6.94, "learning_rate": 9.104069978520591e-07, "loss": 0.0011, "step": 37832 }, { "epoch": 6.94, "learning_rate": 9.10097515062448e-07, "loss": 0.0016, "step": 37833 }, { "epoch": 6.94, "learning_rate": 9.097880823765936e-07, "loss": 0.0025, "step": 37834 }, { "epoch": 6.94, "learning_rate": 9.094786997962013e-07, "loss": 0.0011, "step": 37835 }, { "epoch": 6.94, "learning_rate": 9.091693673229729e-07, "loss": 0.0019, "step": 37836 }, { "epoch": 6.94, "learning_rate": 9.08860084958616e-07, "loss": 0.0016, "step": 37837 }, { "epoch": 6.94, "learning_rate": 9.08550852704837e-07, "loss": 0.002, "step": 37838 }, { "epoch": 6.94, "learning_rate": 9.082416705633379e-07, "loss": 0.0018, "step": 37839 }, { "epoch": 6.94, "learning_rate": 9.079325385358229e-07, "loss": 0.0026, "step": 37840 }, { "epoch": 6.94, "learning_rate": 9.076234566239983e-07, "loss": 0.0016, "step": 37841 }, { "epoch": 6.94, "learning_rate": 9.073144248295651e-07, "loss": 0.0018, "step": 37842 }, { "epoch": 6.94, "learning_rate": 9.070054431542285e-07, "loss": 0.0016, "step": 37843 }, { "epoch": 6.94, "learning_rate": 9.066965115996906e-07, "loss": 0.0014, "step": 37844 }, { "epoch": 6.94, "learning_rate": 9.063876301676522e-07, "loss": 0.0018, "step": 37845 }, { "epoch": 6.94, "learning_rate": 9.060787988598208e-07, "loss": 0.0017, "step": 37846 }, { "epoch": 6.94, "learning_rate": 9.057700176778983e-07, "loss": 0.0016, "step": 37847 }, { "epoch": 6.94, "learning_rate": 9.054612866235824e-07, "loss": 0.0015, "step": 37848 }, { "epoch": 6.94, "learning_rate": 9.051526056985737e-07, "loss": 0.0013, "step": 37849 }, { "epoch": 6.94, "learning_rate": 9.048439749045801e-07, "loss": 0.0028, "step": 37850 }, { "epoch": 6.94, "learning_rate": 9.045353942433e-07, "loss": 0.0011, "step": 37851 }, { "epoch": 6.94, "learning_rate": 9.042268637164331e-07, "loss": 0.0017, "step": 37852 }, { "epoch": 6.94, "learning_rate": 9.039183833256804e-07, "loss": 0.0025, "step": 37853 }, { "epoch": 6.94, "learning_rate": 9.036099530727438e-07, "loss": 0.0016, "step": 37854 }, { "epoch": 6.94, "learning_rate": 9.03301572959322e-07, "loss": 0.0017, "step": 37855 }, { "epoch": 6.94, "learning_rate": 9.029932429871146e-07, "loss": 0.0019, "step": 37856 }, { "epoch": 6.94, "learning_rate": 9.026849631578217e-07, "loss": 0.0015, "step": 37857 }, { "epoch": 6.94, "learning_rate": 9.023767334731426e-07, "loss": 0.0014, "step": 37858 }, { "epoch": 6.94, "learning_rate": 9.020685539347752e-07, "loss": 0.0014, "step": 37859 }, { "epoch": 6.94, "learning_rate": 9.0176042454442e-07, "loss": 0.0013, "step": 37860 }, { "epoch": 6.94, "learning_rate": 9.014523453037749e-07, "loss": 0.0024, "step": 37861 }, { "epoch": 6.94, "learning_rate": 9.01144316214535e-07, "loss": 0.002, "step": 37862 }, { "epoch": 6.94, "learning_rate": 9.008363372784024e-07, "loss": 0.0017, "step": 37863 }, { "epoch": 6.94, "learning_rate": 9.005284084970734e-07, "loss": 0.0015, "step": 37864 }, { "epoch": 6.94, "learning_rate": 9.002205298722466e-07, "loss": 0.0012, "step": 37865 }, { "epoch": 6.94, "learning_rate": 8.999127014056164e-07, "loss": 0.002, "step": 37866 }, { "epoch": 6.94, "learning_rate": 8.99604923098878e-07, "loss": 0.0016, "step": 37867 }, { "epoch": 6.94, "learning_rate": 8.992971949537333e-07, "loss": 0.0014, "step": 37868 }, { "epoch": 6.94, "learning_rate": 8.989895169718744e-07, "loss": 0.0016, "step": 37869 }, { "epoch": 6.94, "learning_rate": 8.986818891549987e-07, "loss": 0.0005, "step": 37870 }, { "epoch": 6.94, "learning_rate": 8.983743115048027e-07, "loss": 0.001, "step": 37871 }, { "epoch": 6.95, "learning_rate": 8.980667840229795e-07, "loss": 0.0008, "step": 37872 }, { "epoch": 6.95, "learning_rate": 8.977593067112256e-07, "loss": 0.0013, "step": 37873 }, { "epoch": 6.95, "learning_rate": 8.974518795712362e-07, "loss": 0.0008, "step": 37874 }, { "epoch": 6.95, "learning_rate": 8.971445026047054e-07, "loss": 0.0015, "step": 37875 }, { "epoch": 6.95, "learning_rate": 8.968371758133265e-07, "loss": 0.0012, "step": 37876 }, { "epoch": 6.95, "learning_rate": 8.965298991987948e-07, "loss": 0.0026, "step": 37877 }, { "epoch": 6.95, "learning_rate": 8.962226727628032e-07, "loss": 0.0023, "step": 37878 }, { "epoch": 6.95, "learning_rate": 8.959154965070461e-07, "loss": 0.0019, "step": 37879 }, { "epoch": 6.95, "learning_rate": 8.956083704332142e-07, "loss": 0.0026, "step": 37880 }, { "epoch": 6.95, "learning_rate": 8.95301294543004e-07, "loss": 0.0021, "step": 37881 }, { "epoch": 6.95, "learning_rate": 8.949942688381086e-07, "loss": 0.0028, "step": 37882 }, { "epoch": 6.95, "learning_rate": 8.946872933202156e-07, "loss": 0.0015, "step": 37883 }, { "epoch": 6.95, "learning_rate": 8.943803679910179e-07, "loss": 0.0018, "step": 37884 }, { "epoch": 6.95, "learning_rate": 8.94073492852211e-07, "loss": 0.0014, "step": 37885 }, { "epoch": 6.95, "learning_rate": 8.937666679054835e-07, "loss": 0.0015, "step": 37886 }, { "epoch": 6.95, "learning_rate": 8.934598931525285e-07, "loss": 0.0014, "step": 37887 }, { "epoch": 6.95, "learning_rate": 8.931531685950356e-07, "loss": 0.0024, "step": 37888 }, { "epoch": 6.95, "learning_rate": 8.928464942346949e-07, "loss": 0.0015, "step": 37889 }, { "epoch": 6.95, "learning_rate": 8.925398700731991e-07, "loss": 0.0012, "step": 37890 }, { "epoch": 6.95, "learning_rate": 8.92233296112236e-07, "loss": 0.0016, "step": 37891 }, { "epoch": 6.95, "learning_rate": 8.919267723534963e-07, "loss": 0.0018, "step": 37892 }, { "epoch": 6.95, "learning_rate": 8.916202987986689e-07, "loss": 0.0016, "step": 37893 }, { "epoch": 6.95, "learning_rate": 8.913138754494466e-07, "loss": 0.0021, "step": 37894 }, { "epoch": 6.95, "learning_rate": 8.910075023075138e-07, "loss": 0.0029, "step": 37895 }, { "epoch": 6.95, "learning_rate": 8.907011793745613e-07, "loss": 0.0012, "step": 37896 }, { "epoch": 6.95, "learning_rate": 8.903949066522766e-07, "loss": 0.001, "step": 37897 }, { "epoch": 6.95, "learning_rate": 8.900886841423473e-07, "loss": 0.0012, "step": 37898 }, { "epoch": 6.95, "learning_rate": 8.897825118464643e-07, "loss": 0.0014, "step": 37899 }, { "epoch": 6.95, "learning_rate": 8.89476389766315e-07, "loss": 0.001, "step": 37900 }, { "epoch": 6.95, "learning_rate": 8.891703179035837e-07, "loss": 0.0009, "step": 37901 }, { "epoch": 6.95, "learning_rate": 8.888642962599581e-07, "loss": 0.0019, "step": 37902 }, { "epoch": 6.95, "learning_rate": 8.885583248371265e-07, "loss": 0.0013, "step": 37903 }, { "epoch": 6.95, "learning_rate": 8.882524036367757e-07, "loss": 0.001, "step": 37904 }, { "epoch": 6.95, "learning_rate": 8.879465326605908e-07, "loss": 0.0023, "step": 37905 }, { "epoch": 6.95, "learning_rate": 8.876407119102581e-07, "loss": 0.0028, "step": 37906 }, { "epoch": 6.95, "learning_rate": 8.873349413874632e-07, "loss": 0.003, "step": 37907 }, { "epoch": 6.95, "learning_rate": 8.870292210938925e-07, "loss": 0.0025, "step": 37908 }, { "epoch": 6.95, "learning_rate": 8.867235510312288e-07, "loss": 0.0017, "step": 37909 }, { "epoch": 6.95, "learning_rate": 8.864179312011589e-07, "loss": 0.0013, "step": 37910 }, { "epoch": 6.95, "learning_rate": 8.861123616053669e-07, "loss": 0.0013, "step": 37911 }, { "epoch": 6.95, "learning_rate": 8.858068422455379e-07, "loss": 0.0017, "step": 37912 }, { "epoch": 6.95, "learning_rate": 8.855013731233542e-07, "loss": 0.0011, "step": 37913 }, { "epoch": 6.95, "learning_rate": 8.851959542405009e-07, "loss": 0.0012, "step": 37914 }, { "epoch": 6.95, "learning_rate": 8.848905855986589e-07, "loss": 0.0016, "step": 37915 }, { "epoch": 6.95, "learning_rate": 8.845852671995148e-07, "loss": 0.0019, "step": 37916 }, { "epoch": 6.95, "learning_rate": 8.842799990447515e-07, "loss": 0.0009, "step": 37917 }, { "epoch": 6.95, "learning_rate": 8.839747811360511e-07, "loss": 0.0016, "step": 37918 }, { "epoch": 6.95, "learning_rate": 8.836696134750944e-07, "loss": 0.0017, "step": 37919 }, { "epoch": 6.95, "learning_rate": 8.833644960635624e-07, "loss": 0.0012, "step": 37920 }, { "epoch": 6.95, "learning_rate": 8.830594289031403e-07, "loss": 0.0017, "step": 37921 }, { "epoch": 6.95, "learning_rate": 8.827544119955078e-07, "loss": 0.0008, "step": 37922 }, { "epoch": 6.95, "learning_rate": 8.824494453423471e-07, "loss": 0.0014, "step": 37923 }, { "epoch": 6.95, "learning_rate": 8.82144528945339e-07, "loss": 0.0008, "step": 37924 }, { "epoch": 6.95, "learning_rate": 8.818396628061631e-07, "loss": 0.0019, "step": 37925 }, { "epoch": 6.96, "learning_rate": 8.815348469265017e-07, "loss": 0.0013, "step": 37926 }, { "epoch": 6.96, "learning_rate": 8.812300813080321e-07, "loss": 0.0017, "step": 37927 }, { "epoch": 6.96, "learning_rate": 8.809253659524375e-07, "loss": 0.0016, "step": 37928 }, { "epoch": 6.96, "learning_rate": 8.806207008613953e-07, "loss": 0.0022, "step": 37929 }, { "epoch": 6.96, "learning_rate": 8.803160860365856e-07, "loss": 0.002, "step": 37930 }, { "epoch": 6.96, "learning_rate": 8.800115214796878e-07, "loss": 0.0013, "step": 37931 }, { "epoch": 6.96, "learning_rate": 8.797070071923786e-07, "loss": 0.0028, "step": 37932 }, { "epoch": 6.96, "learning_rate": 8.794025431763376e-07, "loss": 0.0017, "step": 37933 }, { "epoch": 6.96, "learning_rate": 8.790981294332446e-07, "loss": 0.0023, "step": 37934 }, { "epoch": 6.96, "learning_rate": 8.787937659647772e-07, "loss": 0.001, "step": 37935 }, { "epoch": 6.96, "learning_rate": 8.784894527726128e-07, "loss": 0.0012, "step": 37936 }, { "epoch": 6.96, "learning_rate": 8.781851898584259e-07, "loss": 0.0016, "step": 37937 }, { "epoch": 6.96, "learning_rate": 8.778809772238972e-07, "loss": 0.0015, "step": 37938 }, { "epoch": 6.96, "learning_rate": 8.77576814870702e-07, "loss": 0.0011, "step": 37939 }, { "epoch": 6.96, "learning_rate": 8.77272702800518e-07, "loss": 0.0021, "step": 37940 }, { "epoch": 6.96, "learning_rate": 8.769686410150192e-07, "loss": 0.001, "step": 37941 }, { "epoch": 6.96, "learning_rate": 8.766646295158843e-07, "loss": 0.0013, "step": 37942 }, { "epoch": 6.96, "learning_rate": 8.763606683047865e-07, "loss": 0.0013, "step": 37943 }, { "epoch": 6.96, "learning_rate": 8.760567573834033e-07, "loss": 0.0012, "step": 37944 }, { "epoch": 6.96, "learning_rate": 8.75752896753409e-07, "loss": 0.0013, "step": 37945 }, { "epoch": 6.96, "learning_rate": 8.754490864164767e-07, "loss": 0.0022, "step": 37946 }, { "epoch": 6.96, "learning_rate": 8.75145326374286e-07, "loss": 0.0008, "step": 37947 }, { "epoch": 6.96, "learning_rate": 8.748416166285056e-07, "loss": 0.0017, "step": 37948 }, { "epoch": 6.96, "learning_rate": 8.745379571808133e-07, "loss": 0.0013, "step": 37949 }, { "epoch": 6.96, "learning_rate": 8.742343480328808e-07, "loss": 0.002, "step": 37950 }, { "epoch": 6.96, "learning_rate": 8.739307891863813e-07, "loss": 0.0009, "step": 37951 }, { "epoch": 6.96, "learning_rate": 8.736272806429902e-07, "loss": 0.0021, "step": 37952 }, { "epoch": 6.96, "learning_rate": 8.733238224043794e-07, "loss": 0.0021, "step": 37953 }, { "epoch": 6.96, "learning_rate": 8.730204144722232e-07, "loss": 0.002, "step": 37954 }, { "epoch": 6.96, "learning_rate": 8.727170568481891e-07, "loss": 0.0014, "step": 37955 }, { "epoch": 6.96, "learning_rate": 8.724137495339546e-07, "loss": 0.0012, "step": 37956 }, { "epoch": 6.96, "learning_rate": 8.721104925311885e-07, "loss": 0.001, "step": 37957 }, { "epoch": 6.96, "learning_rate": 8.718072858415638e-07, "loss": 0.0016, "step": 37958 }, { "epoch": 6.96, "learning_rate": 8.715041294667503e-07, "loss": 0.0018, "step": 37959 }, { "epoch": 6.96, "learning_rate": 8.712010234084212e-07, "loss": 0.0012, "step": 37960 }, { "epoch": 6.96, "learning_rate": 8.70897967668245e-07, "loss": 0.0017, "step": 37961 }, { "epoch": 6.96, "learning_rate": 8.705949622478927e-07, "loss": 0.0016, "step": 37962 }, { "epoch": 6.96, "learning_rate": 8.702920071490362e-07, "loss": 0.0016, "step": 37963 }, { "epoch": 6.96, "learning_rate": 8.699891023733409e-07, "loss": 0.0017, "step": 37964 }, { "epoch": 6.96, "learning_rate": 8.696862479224843e-07, "loss": 0.0019, "step": 37965 }, { "epoch": 6.96, "learning_rate": 8.693834437981285e-07, "loss": 0.0012, "step": 37966 }, { "epoch": 6.96, "learning_rate": 8.690806900019455e-07, "loss": 0.0013, "step": 37967 }, { "epoch": 6.96, "learning_rate": 8.687779865356016e-07, "loss": 0.0026, "step": 37968 }, { "epoch": 6.96, "learning_rate": 8.68475333400769e-07, "loss": 0.0017, "step": 37969 }, { "epoch": 6.96, "learning_rate": 8.681727305991139e-07, "loss": 0.0018, "step": 37970 }, { "epoch": 6.96, "learning_rate": 8.678701781323051e-07, "loss": 0.0024, "step": 37971 }, { "epoch": 6.96, "learning_rate": 8.675676760020125e-07, "loss": 0.0021, "step": 37972 }, { "epoch": 6.96, "learning_rate": 8.672652242098956e-07, "loss": 0.0018, "step": 37973 }, { "epoch": 6.96, "learning_rate": 8.669628227576298e-07, "loss": 0.0024, "step": 37974 }, { "epoch": 6.96, "learning_rate": 8.666604716468785e-07, "loss": 0.0013, "step": 37975 }, { "epoch": 6.96, "learning_rate": 8.663581708793078e-07, "loss": 0.0018, "step": 37976 }, { "epoch": 6.96, "learning_rate": 8.660559204565866e-07, "loss": 0.0015, "step": 37977 }, { "epoch": 6.96, "learning_rate": 8.657537203803779e-07, "loss": 0.0014, "step": 37978 }, { "epoch": 6.96, "learning_rate": 8.654515706523492e-07, "loss": 0.0019, "step": 37979 }, { "epoch": 6.96, "learning_rate": 8.651494712741648e-07, "loss": 0.0023, "step": 37980 }, { "epoch": 6.97, "learning_rate": 8.64847422247489e-07, "loss": 0.0012, "step": 37981 }, { "epoch": 6.97, "learning_rate": 8.645454235739903e-07, "loss": 0.0012, "step": 37982 }, { "epoch": 6.97, "learning_rate": 8.642434752553331e-07, "loss": 0.0035, "step": 37983 }, { "epoch": 6.97, "learning_rate": 8.639415772931781e-07, "loss": 0.002, "step": 37984 }, { "epoch": 6.97, "learning_rate": 8.636397296891908e-07, "loss": 0.0024, "step": 37985 }, { "epoch": 6.97, "learning_rate": 8.63337932445033e-07, "loss": 0.0013, "step": 37986 }, { "epoch": 6.97, "learning_rate": 8.630361855623737e-07, "loss": 0.0015, "step": 37987 }, { "epoch": 6.97, "learning_rate": 8.627344890428723e-07, "loss": 0.0013, "step": 37988 }, { "epoch": 6.97, "learning_rate": 8.624328428881945e-07, "loss": 0.0017, "step": 37989 }, { "epoch": 6.97, "learning_rate": 8.621312470999965e-07, "loss": 0.0015, "step": 37990 }, { "epoch": 6.97, "learning_rate": 8.618297016799482e-07, "loss": 0.0011, "step": 37991 }, { "epoch": 6.97, "learning_rate": 8.615282066297082e-07, "loss": 0.0023, "step": 37992 }, { "epoch": 6.97, "learning_rate": 8.612267619509385e-07, "loss": 0.0016, "step": 37993 }, { "epoch": 6.97, "learning_rate": 8.609253676453011e-07, "loss": 0.0007, "step": 37994 }, { "epoch": 6.97, "learning_rate": 8.60624023714457e-07, "loss": 0.0013, "step": 37995 }, { "epoch": 6.97, "learning_rate": 8.603227301600681e-07, "loss": 0.002, "step": 37996 }, { "epoch": 6.97, "learning_rate": 8.600214869837942e-07, "loss": 0.001, "step": 37997 }, { "epoch": 6.97, "learning_rate": 8.597202941872951e-07, "loss": 0.0012, "step": 37998 }, { "epoch": 6.97, "learning_rate": 8.594191517722306e-07, "loss": 0.0012, "step": 37999 }, { "epoch": 6.97, "learning_rate": 8.591180597402649e-07, "loss": 0.0038, "step": 38000 }, { "epoch": 6.97, "learning_rate": 8.588170180930544e-07, "loss": 0.0009, "step": 38001 }, { "epoch": 6.97, "learning_rate": 8.585160268322579e-07, "loss": 0.0021, "step": 38002 }, { "epoch": 6.97, "learning_rate": 8.582150859595351e-07, "loss": 0.001, "step": 38003 }, { "epoch": 6.97, "learning_rate": 8.579141954765435e-07, "loss": 0.0014, "step": 38004 }, { "epoch": 6.97, "learning_rate": 8.576133553849442e-07, "loss": 0.0024, "step": 38005 }, { "epoch": 6.97, "learning_rate": 8.573125656863945e-07, "loss": 0.0024, "step": 38006 }, { "epoch": 6.97, "learning_rate": 8.570118263825544e-07, "loss": 0.0015, "step": 38007 }, { "epoch": 6.97, "learning_rate": 8.567111374750747e-07, "loss": 0.0016, "step": 38008 }, { "epoch": 6.97, "learning_rate": 8.564104989656197e-07, "loss": 0.0014, "step": 38009 }, { "epoch": 6.97, "learning_rate": 8.561099108558446e-07, "loss": 0.0012, "step": 38010 }, { "epoch": 6.97, "learning_rate": 8.55809373147406e-07, "loss": 0.0029, "step": 38011 }, { "epoch": 6.97, "learning_rate": 8.555088858419592e-07, "loss": 0.0024, "step": 38012 }, { "epoch": 6.97, "learning_rate": 8.552084489411627e-07, "loss": 0.0028, "step": 38013 }, { "epoch": 6.97, "learning_rate": 8.549080624466711e-07, "loss": 0.0019, "step": 38014 }, { "epoch": 6.97, "learning_rate": 8.546077263601404e-07, "loss": 0.0016, "step": 38015 }, { "epoch": 6.97, "learning_rate": 8.543074406832263e-07, "loss": 0.0017, "step": 38016 }, { "epoch": 6.97, "learning_rate": 8.54007205417583e-07, "loss": 0.0024, "step": 38017 }, { "epoch": 6.97, "learning_rate": 8.537070205648667e-07, "loss": 0.001, "step": 38018 }, { "epoch": 6.97, "learning_rate": 8.534068861267342e-07, "loss": 0.0019, "step": 38019 }, { "epoch": 6.97, "learning_rate": 8.53106802104835e-07, "loss": 0.0032, "step": 38020 }, { "epoch": 6.97, "learning_rate": 8.528067685008235e-07, "loss": 0.0015, "step": 38021 }, { "epoch": 6.97, "learning_rate": 8.525067853163571e-07, "loss": 0.0023, "step": 38022 }, { "epoch": 6.97, "learning_rate": 8.52206852553088e-07, "loss": 0.0013, "step": 38023 }, { "epoch": 6.97, "learning_rate": 8.519069702126693e-07, "loss": 0.0013, "step": 38024 }, { "epoch": 6.97, "learning_rate": 8.51607138296755e-07, "loss": 0.0013, "step": 38025 }, { "epoch": 6.97, "learning_rate": 8.51307356806993e-07, "loss": 0.0014, "step": 38026 }, { "epoch": 6.97, "learning_rate": 8.510076257450406e-07, "loss": 0.0015, "step": 38027 }, { "epoch": 6.97, "learning_rate": 8.507079451125489e-07, "loss": 0.004, "step": 38028 }, { "epoch": 6.97, "learning_rate": 8.504083149111697e-07, "loss": 0.001, "step": 38029 }, { "epoch": 6.97, "learning_rate": 8.501087351425529e-07, "loss": 0.001, "step": 38030 }, { "epoch": 6.97, "learning_rate": 8.498092058083529e-07, "loss": 0.0015, "step": 38031 }, { "epoch": 6.97, "learning_rate": 8.495097269102181e-07, "loss": 0.0022, "step": 38032 }, { "epoch": 6.97, "learning_rate": 8.492102984497996e-07, "loss": 0.0015, "step": 38033 }, { "epoch": 6.97, "learning_rate": 8.489109204287472e-07, "loss": 0.0016, "step": 38034 }, { "epoch": 6.98, "learning_rate": 8.486115928487137e-07, "loss": 0.002, "step": 38035 }, { "epoch": 6.98, "learning_rate": 8.483123157113493e-07, "loss": 0.001, "step": 38036 }, { "epoch": 6.98, "learning_rate": 8.480130890183003e-07, "loss": 0.0012, "step": 38037 }, { "epoch": 6.98, "learning_rate": 8.477139127712175e-07, "loss": 0.0022, "step": 38038 }, { "epoch": 6.98, "learning_rate": 8.474147869717475e-07, "loss": 0.0019, "step": 38039 }, { "epoch": 6.98, "learning_rate": 8.471157116215444e-07, "loss": 0.0011, "step": 38040 }, { "epoch": 6.98, "learning_rate": 8.468166867222527e-07, "loss": 0.0009, "step": 38041 }, { "epoch": 6.98, "learning_rate": 8.46517712275523e-07, "loss": 0.0014, "step": 38042 }, { "epoch": 6.98, "learning_rate": 8.46218788283002e-07, "loss": 0.001, "step": 38043 }, { "epoch": 6.98, "learning_rate": 8.459199147463371e-07, "loss": 0.0009, "step": 38044 }, { "epoch": 6.98, "learning_rate": 8.456210916671758e-07, "loss": 0.0007, "step": 38045 }, { "epoch": 6.98, "learning_rate": 8.45322319047166e-07, "loss": 0.0019, "step": 38046 }, { "epoch": 6.98, "learning_rate": 8.450235968879539e-07, "loss": 0.0014, "step": 38047 }, { "epoch": 6.98, "learning_rate": 8.447249251911871e-07, "loss": 0.0019, "step": 38048 }, { "epoch": 6.98, "learning_rate": 8.444263039585099e-07, "loss": 0.0009, "step": 38049 }, { "epoch": 6.98, "learning_rate": 8.441277331915699e-07, "loss": 0.0011, "step": 38050 }, { "epoch": 6.98, "learning_rate": 8.438292128920111e-07, "loss": 0.0007, "step": 38051 }, { "epoch": 6.98, "learning_rate": 8.435307430614803e-07, "loss": 0.0019, "step": 38052 }, { "epoch": 6.98, "learning_rate": 8.432323237016227e-07, "loss": 0.0017, "step": 38053 }, { "epoch": 6.98, "learning_rate": 8.429339548140858e-07, "loss": 0.0019, "step": 38054 }, { "epoch": 6.98, "learning_rate": 8.426356364005094e-07, "loss": 0.0015, "step": 38055 }, { "epoch": 6.98, "learning_rate": 8.423373684625402e-07, "loss": 0.0011, "step": 38056 }, { "epoch": 6.98, "learning_rate": 8.420391510018199e-07, "loss": 0.0023, "step": 38057 }, { "epoch": 6.98, "learning_rate": 8.417409840199964e-07, "loss": 0.002, "step": 38058 }, { "epoch": 6.98, "learning_rate": 8.414428675187114e-07, "loss": 0.0013, "step": 38059 }, { "epoch": 6.98, "learning_rate": 8.411448014996071e-07, "loss": 0.0013, "step": 38060 }, { "epoch": 6.98, "learning_rate": 8.408467859643277e-07, "loss": 0.0016, "step": 38061 }, { "epoch": 6.98, "learning_rate": 8.405488209145152e-07, "loss": 0.0012, "step": 38062 }, { "epoch": 6.98, "learning_rate": 8.402509063518116e-07, "loss": 0.0015, "step": 38063 }, { "epoch": 6.98, "learning_rate": 8.399530422778612e-07, "loss": 0.0015, "step": 38064 }, { "epoch": 6.98, "learning_rate": 8.396552286943005e-07, "loss": 0.0013, "step": 38065 }, { "epoch": 6.98, "learning_rate": 8.393574656027803e-07, "loss": 0.0013, "step": 38066 }, { "epoch": 6.98, "learning_rate": 8.390597530049328e-07, "loss": 0.0023, "step": 38067 }, { "epoch": 6.98, "learning_rate": 8.387620909024042e-07, "loss": 0.0013, "step": 38068 }, { "epoch": 6.98, "learning_rate": 8.384644792968322e-07, "loss": 0.0017, "step": 38069 }, { "epoch": 6.98, "learning_rate": 8.381669181898577e-07, "loss": 0.0012, "step": 38070 }, { "epoch": 6.98, "learning_rate": 8.378694075831228e-07, "loss": 0.0031, "step": 38071 }, { "epoch": 6.98, "learning_rate": 8.375719474782684e-07, "loss": 0.002, "step": 38072 }, { "epoch": 6.98, "learning_rate": 8.37274537876931e-07, "loss": 0.0011, "step": 38073 }, { "epoch": 6.98, "learning_rate": 8.36977178780749e-07, "loss": 0.0016, "step": 38074 }, { "epoch": 6.98, "learning_rate": 8.366798701913648e-07, "loss": 0.0025, "step": 38075 }, { "epoch": 6.98, "learning_rate": 8.363826121104157e-07, "loss": 0.0016, "step": 38076 }, { "epoch": 6.98, "learning_rate": 8.360854045395405e-07, "loss": 0.0019, "step": 38077 }, { "epoch": 6.98, "learning_rate": 8.35788247480378e-07, "loss": 0.0014, "step": 38078 }, { "epoch": 6.98, "learning_rate": 8.354911409345645e-07, "loss": 0.001, "step": 38079 }, { "epoch": 6.98, "learning_rate": 8.351940849037387e-07, "loss": 0.0015, "step": 38080 }, { "epoch": 6.98, "learning_rate": 8.348970793895373e-07, "loss": 0.0023, "step": 38081 }, { "epoch": 6.98, "learning_rate": 8.346001243935986e-07, "loss": 0.0014, "step": 38082 }, { "epoch": 6.98, "learning_rate": 8.343032199175594e-07, "loss": 0.0017, "step": 38083 }, { "epoch": 6.98, "learning_rate": 8.340063659630548e-07, "loss": 0.0011, "step": 38084 }, { "epoch": 6.98, "learning_rate": 8.337095625317216e-07, "loss": 0.0025, "step": 38085 }, { "epoch": 6.98, "learning_rate": 8.334128096251959e-07, "loss": 0.0017, "step": 38086 }, { "epoch": 6.98, "learning_rate": 8.331161072451122e-07, "loss": 0.002, "step": 38087 }, { "epoch": 6.98, "learning_rate": 8.32819455393109e-07, "loss": 0.0014, "step": 38088 }, { "epoch": 6.98, "learning_rate": 8.325228540708197e-07, "loss": 0.0011, "step": 38089 }, { "epoch": 6.99, "learning_rate": 8.322263032798817e-07, "loss": 0.0015, "step": 38090 }, { "epoch": 6.99, "learning_rate": 8.319298030219236e-07, "loss": 0.0018, "step": 38091 }, { "epoch": 6.99, "learning_rate": 8.316333532985832e-07, "loss": 0.0019, "step": 38092 }, { "epoch": 6.99, "learning_rate": 8.313369541114957e-07, "loss": 0.001, "step": 38093 }, { "epoch": 6.99, "learning_rate": 8.310406054622933e-07, "loss": 0.0014, "step": 38094 }, { "epoch": 6.99, "learning_rate": 8.3074430735261e-07, "loss": 0.0018, "step": 38095 }, { "epoch": 6.99, "learning_rate": 8.304480597840792e-07, "loss": 0.0014, "step": 38096 }, { "epoch": 6.99, "learning_rate": 8.301518627583339e-07, "loss": 0.0011, "step": 38097 }, { "epoch": 6.99, "learning_rate": 8.298557162770049e-07, "loss": 0.0012, "step": 38098 }, { "epoch": 6.99, "learning_rate": 8.295596203417277e-07, "loss": 0.0009, "step": 38099 }, { "epoch": 6.99, "learning_rate": 8.292635749541323e-07, "loss": 0.0024, "step": 38100 }, { "epoch": 6.99, "learning_rate": 8.289675801158503e-07, "loss": 0.0011, "step": 38101 }, { "epoch": 6.99, "learning_rate": 8.286716358285141e-07, "loss": 0.0013, "step": 38102 }, { "epoch": 6.99, "learning_rate": 8.283757420937543e-07, "loss": 0.0015, "step": 38103 }, { "epoch": 6.99, "learning_rate": 8.280798989132033e-07, "loss": 0.001, "step": 38104 }, { "epoch": 6.99, "learning_rate": 8.277841062884884e-07, "loss": 0.001, "step": 38105 }, { "epoch": 6.99, "learning_rate": 8.274883642212439e-07, "loss": 0.0017, "step": 38106 }, { "epoch": 6.99, "learning_rate": 8.271926727130986e-07, "loss": 0.0017, "step": 38107 }, { "epoch": 6.99, "learning_rate": 8.268970317656843e-07, "loss": 0.0008, "step": 38108 }, { "epoch": 6.99, "learning_rate": 8.266014413806245e-07, "loss": 0.001, "step": 38109 }, { "epoch": 6.99, "learning_rate": 8.263059015595543e-07, "loss": 0.0016, "step": 38110 }, { "epoch": 6.99, "learning_rate": 8.260104123041013e-07, "loss": 0.001, "step": 38111 }, { "epoch": 6.99, "learning_rate": 8.257149736158931e-07, "loss": 0.0013, "step": 38112 }, { "epoch": 6.99, "learning_rate": 8.254195854965597e-07, "loss": 0.0008, "step": 38113 }, { "epoch": 6.99, "learning_rate": 8.251242479477283e-07, "loss": 0.001, "step": 38114 }, { "epoch": 6.99, "learning_rate": 8.24828960971027e-07, "loss": 0.0018, "step": 38115 }, { "epoch": 6.99, "learning_rate": 8.245337245680829e-07, "loss": 0.0015, "step": 38116 }, { "epoch": 6.99, "learning_rate": 8.242385387405249e-07, "loss": 0.0008, "step": 38117 }, { "epoch": 6.99, "learning_rate": 8.239434034899762e-07, "loss": 0.003, "step": 38118 }, { "epoch": 6.99, "learning_rate": 8.236483188180699e-07, "loss": 0.0011, "step": 38119 }, { "epoch": 6.99, "learning_rate": 8.23353284726428e-07, "loss": 0.0011, "step": 38120 }, { "epoch": 6.99, "learning_rate": 8.230583012166782e-07, "loss": 0.0009, "step": 38121 }, { "epoch": 6.99, "learning_rate": 8.227633682904446e-07, "loss": 0.002, "step": 38122 }, { "epoch": 6.99, "learning_rate": 8.224684859493537e-07, "loss": 0.0026, "step": 38123 }, { "epoch": 6.99, "learning_rate": 8.221736541950331e-07, "loss": 0.0023, "step": 38124 }, { "epoch": 6.99, "learning_rate": 8.218788730291061e-07, "loss": 0.0009, "step": 38125 }, { "epoch": 6.99, "learning_rate": 8.215841424531989e-07, "loss": 0.0015, "step": 38126 }, { "epoch": 6.99, "learning_rate": 8.212894624689327e-07, "loss": 0.0014, "step": 38127 }, { "epoch": 6.99, "learning_rate": 8.209948330779349e-07, "loss": 0.0014, "step": 38128 }, { "epoch": 6.99, "learning_rate": 8.207002542818287e-07, "loss": 0.001, "step": 38129 }, { "epoch": 6.99, "learning_rate": 8.204057260822373e-07, "loss": 0.0011, "step": 38130 }, { "epoch": 6.99, "learning_rate": 8.201112484807861e-07, "loss": 0.0014, "step": 38131 }, { "epoch": 6.99, "learning_rate": 8.198168214790958e-07, "loss": 0.0013, "step": 38132 }, { "epoch": 6.99, "learning_rate": 8.195224450787898e-07, "loss": 0.0018, "step": 38133 }, { "epoch": 6.99, "learning_rate": 8.192281192814922e-07, "loss": 0.0013, "step": 38134 }, { "epoch": 6.99, "learning_rate": 8.18933844088824e-07, "loss": 0.0013, "step": 38135 }, { "epoch": 6.99, "learning_rate": 8.186396195024049e-07, "loss": 0.0015, "step": 38136 }, { "epoch": 6.99, "learning_rate": 8.183454455238638e-07, "loss": 0.0022, "step": 38137 }, { "epoch": 6.99, "learning_rate": 8.180513221548159e-07, "loss": 0.0015, "step": 38138 }, { "epoch": 6.99, "learning_rate": 8.177572493968833e-07, "loss": 0.0013, "step": 38139 }, { "epoch": 6.99, "learning_rate": 8.17463227251687e-07, "loss": 0.0009, "step": 38140 }, { "epoch": 6.99, "learning_rate": 8.1716925572085e-07, "loss": 0.0011, "step": 38141 }, { "epoch": 6.99, "learning_rate": 8.16875334805991e-07, "loss": 0.0015, "step": 38142 }, { "epoch": 6.99, "learning_rate": 8.165814645087323e-07, "loss": 0.0025, "step": 38143 }, { "epoch": 7.0, "learning_rate": 8.1628764483069e-07, "loss": 0.0027, "step": 38144 }, { "epoch": 7.0, "learning_rate": 8.159938757734831e-07, "loss": 0.0012, "step": 38145 }, { "epoch": 7.0, "learning_rate": 8.157001573387358e-07, "loss": 0.0011, "step": 38146 }, { "epoch": 7.0, "learning_rate": 8.154064895280633e-07, "loss": 0.0019, "step": 38147 }, { "epoch": 7.0, "learning_rate": 8.151128723430868e-07, "loss": 0.001, "step": 38148 }, { "epoch": 7.0, "learning_rate": 8.148193057854226e-07, "loss": 0.0028, "step": 38149 }, { "epoch": 7.0, "learning_rate": 8.145257898566894e-07, "loss": 0.0017, "step": 38150 }, { "epoch": 7.0, "learning_rate": 8.142323245585049e-07, "loss": 0.0012, "step": 38151 }, { "epoch": 7.0, "learning_rate": 8.139389098924877e-07, "loss": 0.0018, "step": 38152 }, { "epoch": 7.0, "learning_rate": 8.136455458602544e-07, "loss": 0.0019, "step": 38153 }, { "epoch": 7.0, "learning_rate": 8.133522324634191e-07, "loss": 0.0016, "step": 38154 }, { "epoch": 7.0, "learning_rate": 8.130589697036063e-07, "loss": 0.001, "step": 38155 }, { "epoch": 7.0, "learning_rate": 8.127657575824255e-07, "loss": 0.0013, "step": 38156 }, { "epoch": 7.0, "learning_rate": 8.124725961014957e-07, "loss": 0.0019, "step": 38157 }, { "epoch": 7.0, "learning_rate": 8.121794852624298e-07, "loss": 0.0014, "step": 38158 }, { "epoch": 7.0, "learning_rate": 8.118864250668478e-07, "loss": 0.0016, "step": 38159 }, { "epoch": 7.0, "learning_rate": 8.11593415516363e-07, "loss": 0.0027, "step": 38160 }, { "epoch": 7.0, "learning_rate": 8.113004566125926e-07, "loss": 0.0014, "step": 38161 }, { "epoch": 7.0, "learning_rate": 8.110075483571455e-07, "loss": 0.0008, "step": 38162 }, { "epoch": 7.0, "learning_rate": 8.107146907516416e-07, "loss": 0.0021, "step": 38163 }, { "epoch": 7.0, "learning_rate": 8.10421883797694e-07, "loss": 0.0018, "step": 38164 }, { "epoch": 7.0, "learning_rate": 8.101291274969159e-07, "loss": 0.0015, "step": 38165 }, { "epoch": 7.0, "learning_rate": 8.098364218509214e-07, "loss": 0.0013, "step": 38166 }, { "epoch": 7.0, "learning_rate": 8.095437668613248e-07, "loss": 0.001, "step": 38167 }, { "epoch": 7.0, "learning_rate": 8.092511625297373e-07, "loss": 0.0012, "step": 38168 }, { "epoch": 7.0, "learning_rate": 8.089586088577728e-07, "loss": 0.0012, "step": 38169 }, { "epoch": 7.0, "learning_rate": 8.086661058470435e-07, "loss": 0.0015, "step": 38170 }, { "epoch": 7.0, "learning_rate": 8.083736534991615e-07, "loss": 0.0017, "step": 38171 }, { "epoch": 7.0, "learning_rate": 8.0808125181574e-07, "loss": 0.0006, "step": 38172 }, { "epoch": 7.0, "learning_rate": 8.077889007983919e-07, "loss": 0.0005, "step": 38173 }, { "epoch": 7.0, "learning_rate": 8.074966004487251e-07, "loss": 0.0003, "step": 38174 }, { "epoch": 7.0, "learning_rate": 8.072043507683514e-07, "loss": 0.0009, "step": 38175 }, { "epoch": 7.0, "learning_rate": 8.069121517588818e-07, "loss": 0.0005, "step": 38176 }, { "epoch": 7.0, "learning_rate": 8.066200034219296e-07, "loss": 0.0003, "step": 38177 }, { "epoch": 7.0, "learning_rate": 8.063279057591033e-07, "loss": 0.0005, "step": 38178 }, { "epoch": 7.0, "learning_rate": 8.060358587720152e-07, "loss": 0.0006, "step": 38179 }, { "epoch": 7.0, "learning_rate": 8.057438624622682e-07, "loss": 0.0006, "step": 38180 }, { "epoch": 7.0, "learning_rate": 8.054519168314789e-07, "loss": 0.0003, "step": 38181 }, { "epoch": 7.0, "learning_rate": 8.051600218812539e-07, "loss": 0.0003, "step": 38182 }, { "epoch": 7.0, "learning_rate": 8.048681776132017e-07, "loss": 0.0005, "step": 38183 }, { "epoch": 7.0, "learning_rate": 8.045763840289322e-07, "loss": 0.0016, "step": 38184 }, { "epoch": 7.0, "learning_rate": 8.042846411300531e-07, "loss": 0.0007, "step": 38185 }, { "epoch": 7.0, "learning_rate": 8.03992948918173e-07, "loss": 0.0007, "step": 38186 }, { "epoch": 7.0, "learning_rate": 8.037013073948985e-07, "loss": 0.0011, "step": 38187 }, { "epoch": 7.0, "learning_rate": 8.034097165618382e-07, "loss": 0.0005, "step": 38188 }, { "epoch": 7.0, "learning_rate": 8.031181764205964e-07, "loss": 0.0004, "step": 38189 }, { "epoch": 7.0, "learning_rate": 8.028266869727863e-07, "loss": 0.0008, "step": 38190 }, { "epoch": 7.0, "learning_rate": 8.025352482200099e-07, "loss": 0.0007, "step": 38191 }, { "epoch": 7.0, "learning_rate": 8.022438601638749e-07, "loss": 0.0004, "step": 38192 }, { "epoch": 7.0, "learning_rate": 8.019525228059843e-07, "loss": 0.0004, "step": 38193 }, { "epoch": 7.0, "learning_rate": 8.016612361479504e-07, "loss": 0.0005, "step": 38194 }, { "epoch": 7.0, "learning_rate": 8.013700001913737e-07, "loss": 0.0003, "step": 38195 }, { "epoch": 7.0, "learning_rate": 8.010788149378612e-07, "loss": 0.0004, "step": 38196 }, { "epoch": 7.0, "learning_rate": 8.007876803890202e-07, "loss": 0.0006, "step": 38197 }, { "epoch": 7.0, "learning_rate": 8.004965965464495e-07, "loss": 0.0006, "step": 38198 }, { "epoch": 7.01, "learning_rate": 8.002055634117578e-07, "loss": 0.0004, "step": 38199 }, { "epoch": 7.01, "learning_rate": 7.999145809865493e-07, "loss": 0.0011, "step": 38200 }, { "epoch": 7.01, "learning_rate": 7.996236492724275e-07, "loss": 0.0004, "step": 38201 }, { "epoch": 7.01, "learning_rate": 7.99332768270995e-07, "loss": 0.0007, "step": 38202 }, { "epoch": 7.01, "learning_rate": 7.990419379838566e-07, "loss": 0.0007, "step": 38203 }, { "epoch": 7.01, "learning_rate": 7.987511584126139e-07, "loss": 0.0004, "step": 38204 }, { "epoch": 7.01, "learning_rate": 7.984604295588716e-07, "loss": 0.0003, "step": 38205 }, { "epoch": 7.01, "learning_rate": 7.981697514242293e-07, "loss": 0.0003, "step": 38206 }, { "epoch": 7.01, "learning_rate": 7.978791240102901e-07, "loss": 0.0011, "step": 38207 }, { "epoch": 7.01, "learning_rate": 7.975885473186596e-07, "loss": 0.0004, "step": 38208 }, { "epoch": 7.01, "learning_rate": 7.972980213509362e-07, "loss": 0.0006, "step": 38209 }, { "epoch": 7.01, "learning_rate": 7.970075461087201e-07, "loss": 0.0007, "step": 38210 }, { "epoch": 7.01, "learning_rate": 7.967171215936132e-07, "loss": 0.0008, "step": 38211 }, { "epoch": 7.01, "learning_rate": 7.964267478072185e-07, "loss": 0.0007, "step": 38212 }, { "epoch": 7.01, "learning_rate": 7.961364247511361e-07, "loss": 0.0005, "step": 38213 }, { "epoch": 7.01, "learning_rate": 7.958461524269645e-07, "loss": 0.0007, "step": 38214 }, { "epoch": 7.01, "learning_rate": 7.955559308363048e-07, "loss": 0.0003, "step": 38215 }, { "epoch": 7.01, "learning_rate": 7.952657599807556e-07, "loss": 0.0003, "step": 38216 }, { "epoch": 7.01, "learning_rate": 7.949756398619179e-07, "loss": 0.0003, "step": 38217 }, { "epoch": 7.01, "learning_rate": 7.946855704813905e-07, "loss": 0.0005, "step": 38218 }, { "epoch": 7.01, "learning_rate": 7.943955518407709e-07, "loss": 0.0011, "step": 38219 }, { "epoch": 7.01, "learning_rate": 7.941055839416601e-07, "loss": 0.0004, "step": 38220 }, { "epoch": 7.01, "learning_rate": 7.938156667856533e-07, "loss": 0.0004, "step": 38221 }, { "epoch": 7.01, "learning_rate": 7.935258003743518e-07, "loss": 0.0005, "step": 38222 }, { "epoch": 7.01, "learning_rate": 7.932359847093507e-07, "loss": 0.0005, "step": 38223 }, { "epoch": 7.01, "learning_rate": 7.929462197922467e-07, "loss": 0.0003, "step": 38224 }, { "epoch": 7.01, "learning_rate": 7.926565056246416e-07, "loss": 0.0004, "step": 38225 }, { "epoch": 7.01, "learning_rate": 7.92366842208131e-07, "loss": 0.0004, "step": 38226 }, { "epoch": 7.01, "learning_rate": 7.92077229544308e-07, "loss": 0.0005, "step": 38227 }, { "epoch": 7.01, "learning_rate": 7.917876676347702e-07, "loss": 0.0005, "step": 38228 }, { "epoch": 7.01, "learning_rate": 7.914981564811141e-07, "loss": 0.0004, "step": 38229 }, { "epoch": 7.01, "learning_rate": 7.912086960849374e-07, "loss": 0.0004, "step": 38230 }, { "epoch": 7.01, "learning_rate": 7.909192864478332e-07, "loss": 0.0006, "step": 38231 }, { "epoch": 7.01, "learning_rate": 7.906299275713991e-07, "loss": 0.0003, "step": 38232 }, { "epoch": 7.01, "learning_rate": 7.903406194572272e-07, "loss": 0.0004, "step": 38233 }, { "epoch": 7.01, "learning_rate": 7.900513621069151e-07, "loss": 0.0006, "step": 38234 }, { "epoch": 7.01, "learning_rate": 7.897621555220547e-07, "loss": 0.0004, "step": 38235 }, { "epoch": 7.01, "learning_rate": 7.894729997042405e-07, "loss": 0.0014, "step": 38236 }, { "epoch": 7.01, "learning_rate": 7.891838946550679e-07, "loss": 0.0016, "step": 38237 }, { "epoch": 7.01, "learning_rate": 7.888948403761287e-07, "loss": 0.0006, "step": 38238 }, { "epoch": 7.01, "learning_rate": 7.886058368690175e-07, "loss": 0.0006, "step": 38239 }, { "epoch": 7.01, "learning_rate": 7.883168841353261e-07, "loss": 0.0008, "step": 38240 }, { "epoch": 7.01, "learning_rate": 7.88027982176649e-07, "loss": 0.0004, "step": 38241 }, { "epoch": 7.01, "learning_rate": 7.877391309945748e-07, "loss": 0.0003, "step": 38242 }, { "epoch": 7.01, "learning_rate": 7.874503305906999e-07, "loss": 0.0008, "step": 38243 }, { "epoch": 7.01, "learning_rate": 7.871615809666167e-07, "loss": 0.0005, "step": 38244 }, { "epoch": 7.01, "learning_rate": 7.868728821239135e-07, "loss": 0.0006, "step": 38245 }, { "epoch": 7.01, "learning_rate": 7.865842340641805e-07, "loss": 0.0015, "step": 38246 }, { "epoch": 7.01, "learning_rate": 7.862956367890129e-07, "loss": 0.0003, "step": 38247 }, { "epoch": 7.01, "learning_rate": 7.860070902999994e-07, "loss": 0.0003, "step": 38248 }, { "epoch": 7.01, "learning_rate": 7.857185945987311e-07, "loss": 0.0004, "step": 38249 }, { "epoch": 7.01, "learning_rate": 7.854301496867989e-07, "loss": 0.0004, "step": 38250 }, { "epoch": 7.01, "learning_rate": 7.851417555657881e-07, "loss": 0.0004, "step": 38251 }, { "epoch": 7.01, "learning_rate": 7.848534122372941e-07, "loss": 0.0006, "step": 38252 }, { "epoch": 7.02, "learning_rate": 7.845651197029036e-07, "loss": 0.0005, "step": 38253 }, { "epoch": 7.02, "learning_rate": 7.842768779642062e-07, "loss": 0.0013, "step": 38254 }, { "epoch": 7.02, "learning_rate": 7.839886870227909e-07, "loss": 0.0006, "step": 38255 }, { "epoch": 7.02, "learning_rate": 7.837005468802461e-07, "loss": 0.0005, "step": 38256 }, { "epoch": 7.02, "learning_rate": 7.834124575381585e-07, "loss": 0.0009, "step": 38257 }, { "epoch": 7.02, "learning_rate": 7.831244189981179e-07, "loss": 0.0003, "step": 38258 }, { "epoch": 7.02, "learning_rate": 7.828364312617121e-07, "loss": 0.0004, "step": 38259 }, { "epoch": 7.02, "learning_rate": 7.825484943305262e-07, "loss": 0.0009, "step": 38260 }, { "epoch": 7.02, "learning_rate": 7.822606082061501e-07, "loss": 0.0006, "step": 38261 }, { "epoch": 7.02, "learning_rate": 7.819727728901704e-07, "loss": 0.0004, "step": 38262 }, { "epoch": 7.02, "learning_rate": 7.816849883841726e-07, "loss": 0.0004, "step": 38263 }, { "epoch": 7.02, "learning_rate": 7.813972546897408e-07, "loss": 0.0003, "step": 38264 }, { "epoch": 7.02, "learning_rate": 7.811095718084638e-07, "loss": 0.0006, "step": 38265 }, { "epoch": 7.02, "learning_rate": 7.808219397419281e-07, "loss": 0.001, "step": 38266 }, { "epoch": 7.02, "learning_rate": 7.80534358491718e-07, "loss": 0.0004, "step": 38267 }, { "epoch": 7.02, "learning_rate": 7.802468280594178e-07, "loss": 0.0007, "step": 38268 }, { "epoch": 7.02, "learning_rate": 7.799593484466139e-07, "loss": 0.0003, "step": 38269 }, { "epoch": 7.02, "learning_rate": 7.796719196548897e-07, "loss": 0.0004, "step": 38270 }, { "epoch": 7.02, "learning_rate": 7.793845416858292e-07, "loss": 0.0005, "step": 38271 }, { "epoch": 7.02, "learning_rate": 7.790972145410181e-07, "loss": 0.0006, "step": 38272 }, { "epoch": 7.02, "learning_rate": 7.788099382220382e-07, "loss": 0.0008, "step": 38273 }, { "epoch": 7.02, "learning_rate": 7.785227127304751e-07, "loss": 0.0004, "step": 38274 }, { "epoch": 7.02, "learning_rate": 7.782355380679097e-07, "loss": 0.0008, "step": 38275 }, { "epoch": 7.02, "learning_rate": 7.779484142359272e-07, "loss": 0.0006, "step": 38276 }, { "epoch": 7.02, "learning_rate": 7.776613412361067e-07, "loss": 0.0004, "step": 38277 }, { "epoch": 7.02, "learning_rate": 7.773743190700345e-07, "loss": 0.0004, "step": 38278 }, { "epoch": 7.02, "learning_rate": 7.770873477392915e-07, "loss": 0.0007, "step": 38279 }, { "epoch": 7.02, "learning_rate": 7.76800427245461e-07, "loss": 0.0006, "step": 38280 }, { "epoch": 7.02, "learning_rate": 7.765135575901206e-07, "loss": 0.0003, "step": 38281 }, { "epoch": 7.02, "learning_rate": 7.762267387748512e-07, "loss": 0.0009, "step": 38282 }, { "epoch": 7.02, "learning_rate": 7.759399708012383e-07, "loss": 0.0005, "step": 38283 }, { "epoch": 7.02, "learning_rate": 7.756532536708606e-07, "loss": 0.0007, "step": 38284 }, { "epoch": 7.02, "learning_rate": 7.753665873852989e-07, "loss": 0.0004, "step": 38285 }, { "epoch": 7.02, "learning_rate": 7.75079971946131e-07, "loss": 0.0005, "step": 38286 }, { "epoch": 7.02, "learning_rate": 7.747934073549401e-07, "loss": 0.0005, "step": 38287 }, { "epoch": 7.02, "learning_rate": 7.745068936133026e-07, "loss": 0.0003, "step": 38288 }, { "epoch": 7.02, "learning_rate": 7.742204307228008e-07, "loss": 0.0004, "step": 38289 }, { "epoch": 7.02, "learning_rate": 7.739340186850097e-07, "loss": 0.0004, "step": 38290 }, { "epoch": 7.02, "learning_rate": 7.736476575015128e-07, "loss": 0.0006, "step": 38291 }, { "epoch": 7.02, "learning_rate": 7.733613471738854e-07, "loss": 0.0005, "step": 38292 }, { "epoch": 7.02, "learning_rate": 7.730750877037075e-07, "loss": 0.0006, "step": 38293 }, { "epoch": 7.02, "learning_rate": 7.727888790925542e-07, "loss": 0.0011, "step": 38294 }, { "epoch": 7.02, "learning_rate": 7.725027213420034e-07, "loss": 0.0006, "step": 38295 }, { "epoch": 7.02, "learning_rate": 7.722166144536358e-07, "loss": 0.0005, "step": 38296 }, { "epoch": 7.02, "learning_rate": 7.719305584290282e-07, "loss": 0.0003, "step": 38297 }, { "epoch": 7.02, "learning_rate": 7.716445532697536e-07, "loss": 0.0003, "step": 38298 }, { "epoch": 7.02, "learning_rate": 7.713585989773887e-07, "loss": 0.0004, "step": 38299 }, { "epoch": 7.02, "learning_rate": 7.71072695553513e-07, "loss": 0.0005, "step": 38300 }, { "epoch": 7.02, "learning_rate": 7.707868429997012e-07, "loss": 0.0004, "step": 38301 }, { "epoch": 7.02, "learning_rate": 7.705010413175284e-07, "loss": 0.0007, "step": 38302 }, { "epoch": 7.02, "learning_rate": 7.702152905085703e-07, "loss": 0.0003, "step": 38303 }, { "epoch": 7.02, "learning_rate": 7.69929590574402e-07, "loss": 0.0004, "step": 38304 }, { "epoch": 7.02, "learning_rate": 7.696439415165969e-07, "loss": 0.0004, "step": 38305 }, { "epoch": 7.02, "learning_rate": 7.693583433367314e-07, "loss": 0.0003, "step": 38306 }, { "epoch": 7.02, "learning_rate": 7.690727960363787e-07, "loss": 0.0004, "step": 38307 }, { "epoch": 7.03, "learning_rate": 7.687872996171119e-07, "loss": 0.0005, "step": 38308 }, { "epoch": 7.03, "learning_rate": 7.685018540805089e-07, "loss": 0.0007, "step": 38309 }, { "epoch": 7.03, "learning_rate": 7.682164594281371e-07, "loss": 0.0004, "step": 38310 }, { "epoch": 7.03, "learning_rate": 7.679311156615743e-07, "loss": 0.0005, "step": 38311 }, { "epoch": 7.03, "learning_rate": 7.676458227823902e-07, "loss": 0.0004, "step": 38312 }, { "epoch": 7.03, "learning_rate": 7.67360580792157e-07, "loss": 0.0005, "step": 38313 }, { "epoch": 7.03, "learning_rate": 7.67075389692451e-07, "loss": 0.0005, "step": 38314 }, { "epoch": 7.03, "learning_rate": 7.667902494848434e-07, "loss": 0.0006, "step": 38315 }, { "epoch": 7.03, "learning_rate": 7.665051601709017e-07, "loss": 0.0009, "step": 38316 }, { "epoch": 7.03, "learning_rate": 7.662201217521991e-07, "loss": 0.0003, "step": 38317 }, { "epoch": 7.03, "learning_rate": 7.659351342303089e-07, "loss": 0.0004, "step": 38318 }, { "epoch": 7.03, "learning_rate": 7.656501976067998e-07, "loss": 0.0006, "step": 38319 }, { "epoch": 7.03, "learning_rate": 7.653653118832438e-07, "loss": 0.0006, "step": 38320 }, { "epoch": 7.03, "learning_rate": 7.650804770612108e-07, "loss": 0.0004, "step": 38321 }, { "epoch": 7.03, "learning_rate": 7.647956931422706e-07, "loss": 0.0003, "step": 38322 }, { "epoch": 7.03, "learning_rate": 7.645109601279921e-07, "loss": 0.0004, "step": 38323 }, { "epoch": 7.03, "learning_rate": 7.642262780199461e-07, "loss": 0.0004, "step": 38324 }, { "epoch": 7.03, "learning_rate": 7.639416468197014e-07, "loss": 0.0005, "step": 38325 }, { "epoch": 7.03, "learning_rate": 7.636570665288246e-07, "loss": 0.0006, "step": 38326 }, { "epoch": 7.03, "learning_rate": 7.633725371488898e-07, "loss": 0.0004, "step": 38327 }, { "epoch": 7.03, "learning_rate": 7.630880586814615e-07, "loss": 0.0004, "step": 38328 }, { "epoch": 7.03, "learning_rate": 7.628036311281062e-07, "loss": 0.0007, "step": 38329 }, { "epoch": 7.03, "learning_rate": 7.625192544903936e-07, "loss": 0.0006, "step": 38330 }, { "epoch": 7.03, "learning_rate": 7.622349287698927e-07, "loss": 0.0005, "step": 38331 }, { "epoch": 7.03, "learning_rate": 7.619506539681687e-07, "loss": 0.0008, "step": 38332 }, { "epoch": 7.03, "learning_rate": 7.616664300867915e-07, "loss": 0.0005, "step": 38333 }, { "epoch": 7.03, "learning_rate": 7.613822571273233e-07, "loss": 0.0009, "step": 38334 }, { "epoch": 7.03, "learning_rate": 7.610981350913305e-07, "loss": 0.0006, "step": 38335 }, { "epoch": 7.03, "learning_rate": 7.608140639803829e-07, "loss": 0.0008, "step": 38336 }, { "epoch": 7.03, "learning_rate": 7.60530043796045e-07, "loss": 0.0004, "step": 38337 }, { "epoch": 7.03, "learning_rate": 7.602460745398821e-07, "loss": 0.0023, "step": 38338 }, { "epoch": 7.03, "learning_rate": 7.599621562134596e-07, "loss": 0.0005, "step": 38339 }, { "epoch": 7.03, "learning_rate": 7.596782888183407e-07, "loss": 0.0005, "step": 38340 }, { "epoch": 7.03, "learning_rate": 7.593944723560931e-07, "loss": 0.0008, "step": 38341 }, { "epoch": 7.03, "learning_rate": 7.591107068282788e-07, "loss": 0.0004, "step": 38342 }, { "epoch": 7.03, "learning_rate": 7.588269922364622e-07, "loss": 0.0004, "step": 38343 }, { "epoch": 7.03, "learning_rate": 7.585433285822096e-07, "loss": 0.0003, "step": 38344 }, { "epoch": 7.03, "learning_rate": 7.582597158670823e-07, "loss": 0.0005, "step": 38345 }, { "epoch": 7.03, "learning_rate": 7.579761540926434e-07, "loss": 0.0004, "step": 38346 }, { "epoch": 7.03, "learning_rate": 7.576926432604559e-07, "loss": 0.0004, "step": 38347 }, { "epoch": 7.03, "learning_rate": 7.57409183372082e-07, "loss": 0.0005, "step": 38348 }, { "epoch": 7.03, "learning_rate": 7.571257744290861e-07, "loss": 0.0004, "step": 38349 }, { "epoch": 7.03, "learning_rate": 7.568424164330302e-07, "loss": 0.0004, "step": 38350 }, { "epoch": 7.03, "learning_rate": 7.565591093854774e-07, "loss": 0.0002, "step": 38351 }, { "epoch": 7.03, "learning_rate": 7.562758532879833e-07, "loss": 0.0003, "step": 38352 }, { "epoch": 7.03, "learning_rate": 7.559926481421154e-07, "loss": 0.0005, "step": 38353 }, { "epoch": 7.03, "learning_rate": 7.557094939494325e-07, "loss": 0.0003, "step": 38354 }, { "epoch": 7.03, "learning_rate": 7.554263907114956e-07, "loss": 0.0008, "step": 38355 }, { "epoch": 7.03, "learning_rate": 7.551433384298645e-07, "loss": 0.0006, "step": 38356 }, { "epoch": 7.03, "learning_rate": 7.548603371061003e-07, "loss": 0.0003, "step": 38357 }, { "epoch": 7.03, "learning_rate": 7.545773867417627e-07, "loss": 0.0005, "step": 38358 }, { "epoch": 7.03, "learning_rate": 7.542944873384106e-07, "loss": 0.0005, "step": 38359 }, { "epoch": 7.03, "learning_rate": 7.540116388976049e-07, "loss": 0.0007, "step": 38360 }, { "epoch": 7.03, "learning_rate": 7.537288414209009e-07, "loss": 0.0006, "step": 38361 }, { "epoch": 7.04, "learning_rate": 7.534460949098643e-07, "loss": 0.0003, "step": 38362 }, { "epoch": 7.04, "learning_rate": 7.53163399366047e-07, "loss": 0.0003, "step": 38363 }, { "epoch": 7.04, "learning_rate": 7.5288075479101e-07, "loss": 0.0003, "step": 38364 }, { "epoch": 7.04, "learning_rate": 7.525981611863087e-07, "loss": 0.0002, "step": 38365 }, { "epoch": 7.04, "learning_rate": 7.523156185535052e-07, "loss": 0.0004, "step": 38366 }, { "epoch": 7.04, "learning_rate": 7.520331268941549e-07, "loss": 0.0007, "step": 38367 }, { "epoch": 7.04, "learning_rate": 7.517506862098145e-07, "loss": 0.0005, "step": 38368 }, { "epoch": 7.04, "learning_rate": 7.514682965020426e-07, "loss": 0.0004, "step": 38369 }, { "epoch": 7.04, "learning_rate": 7.511859577723901e-07, "loss": 0.0006, "step": 38370 }, { "epoch": 7.04, "learning_rate": 7.509036700224204e-07, "loss": 0.0005, "step": 38371 }, { "epoch": 7.04, "learning_rate": 7.506214332536854e-07, "loss": 0.0011, "step": 38372 }, { "epoch": 7.04, "learning_rate": 7.503392474677418e-07, "loss": 0.0005, "step": 38373 }, { "epoch": 7.04, "learning_rate": 7.500571126661449e-07, "loss": 0.0003, "step": 38374 }, { "epoch": 7.04, "learning_rate": 7.497750288504501e-07, "loss": 0.0011, "step": 38375 }, { "epoch": 7.04, "learning_rate": 7.494929960222119e-07, "loss": 0.0005, "step": 38376 }, { "epoch": 7.04, "learning_rate": 7.492110141829856e-07, "loss": 0.0005, "step": 38377 }, { "epoch": 7.04, "learning_rate": 7.489290833343243e-07, "loss": 0.0005, "step": 38378 }, { "epoch": 7.04, "learning_rate": 7.486472034777803e-07, "loss": 0.0005, "step": 38379 }, { "epoch": 7.04, "learning_rate": 7.483653746149144e-07, "loss": 0.0009, "step": 38380 }, { "epoch": 7.04, "learning_rate": 7.480835967472733e-07, "loss": 0.0006, "step": 38381 }, { "epoch": 7.04, "learning_rate": 7.478018698764111e-07, "loss": 0.0005, "step": 38382 }, { "epoch": 7.04, "learning_rate": 7.475201940038812e-07, "loss": 0.0004, "step": 38383 }, { "epoch": 7.04, "learning_rate": 7.472385691312379e-07, "loss": 0.0014, "step": 38384 }, { "epoch": 7.04, "learning_rate": 7.46956995260033e-07, "loss": 0.0004, "step": 38385 }, { "epoch": 7.04, "learning_rate": 7.466754723918168e-07, "loss": 0.0004, "step": 38386 }, { "epoch": 7.04, "learning_rate": 7.463940005281445e-07, "loss": 0.0003, "step": 38387 }, { "epoch": 7.04, "learning_rate": 7.461125796705626e-07, "loss": 0.0003, "step": 38388 }, { "epoch": 7.04, "learning_rate": 7.458312098206255e-07, "loss": 0.0004, "step": 38389 }, { "epoch": 7.04, "learning_rate": 7.455498909798842e-07, "loss": 0.0004, "step": 38390 }, { "epoch": 7.04, "learning_rate": 7.452686231498885e-07, "loss": 0.0004, "step": 38391 }, { "epoch": 7.04, "learning_rate": 7.449874063321894e-07, "loss": 0.0006, "step": 38392 }, { "epoch": 7.04, "learning_rate": 7.447062405283357e-07, "loss": 0.0004, "step": 38393 }, { "epoch": 7.04, "learning_rate": 7.444251257398782e-07, "loss": 0.0004, "step": 38394 }, { "epoch": 7.04, "learning_rate": 7.441440619683671e-07, "loss": 0.0005, "step": 38395 }, { "epoch": 7.04, "learning_rate": 7.438630492153487e-07, "loss": 0.0004, "step": 38396 }, { "epoch": 7.04, "learning_rate": 7.435820874823752e-07, "loss": 0.0006, "step": 38397 }, { "epoch": 7.04, "learning_rate": 7.433011767709964e-07, "loss": 0.0003, "step": 38398 }, { "epoch": 7.04, "learning_rate": 7.430203170827555e-07, "loss": 0.0015, "step": 38399 }, { "epoch": 7.04, "learning_rate": 7.427395084192046e-07, "loss": 0.0011, "step": 38400 }, { "epoch": 7.04, "learning_rate": 7.424587507818892e-07, "loss": 0.0006, "step": 38401 }, { "epoch": 7.04, "learning_rate": 7.42178044172358e-07, "loss": 0.0009, "step": 38402 }, { "epoch": 7.04, "learning_rate": 7.418973885921599e-07, "loss": 0.0005, "step": 38403 }, { "epoch": 7.04, "learning_rate": 7.416167840428413e-07, "loss": 0.0003, "step": 38404 }, { "epoch": 7.04, "learning_rate": 7.413362305259442e-07, "loss": 0.0015, "step": 38405 }, { "epoch": 7.04, "learning_rate": 7.410557280430208e-07, "loss": 0.0004, "step": 38406 }, { "epoch": 7.04, "learning_rate": 7.407752765956155e-07, "loss": 0.0005, "step": 38407 }, { "epoch": 7.04, "learning_rate": 7.404948761852737e-07, "loss": 0.0005, "step": 38408 }, { "epoch": 7.04, "learning_rate": 7.402145268135408e-07, "loss": 0.0008, "step": 38409 }, { "epoch": 7.04, "learning_rate": 7.399342284819621e-07, "loss": 0.0004, "step": 38410 }, { "epoch": 7.04, "learning_rate": 7.396539811920833e-07, "loss": 0.0006, "step": 38411 }, { "epoch": 7.04, "learning_rate": 7.393737849454485e-07, "loss": 0.0009, "step": 38412 }, { "epoch": 7.04, "learning_rate": 7.390936397436021e-07, "loss": 0.0006, "step": 38413 }, { "epoch": 7.04, "learning_rate": 7.388135455880874e-07, "loss": 0.0006, "step": 38414 }, { "epoch": 7.04, "learning_rate": 7.385335024804507e-07, "loss": 0.0002, "step": 38415 }, { "epoch": 7.04, "learning_rate": 7.382535104222366e-07, "loss": 0.0006, "step": 38416 }, { "epoch": 7.05, "learning_rate": 7.379735694149837e-07, "loss": 0.0006, "step": 38417 }, { "epoch": 7.05, "learning_rate": 7.376936794602363e-07, "loss": 0.0012, "step": 38418 }, { "epoch": 7.05, "learning_rate": 7.374138405595388e-07, "loss": 0.0004, "step": 38419 }, { "epoch": 7.05, "learning_rate": 7.371340527144344e-07, "loss": 0.0008, "step": 38420 }, { "epoch": 7.05, "learning_rate": 7.368543159264641e-07, "loss": 0.0005, "step": 38421 }, { "epoch": 7.05, "learning_rate": 7.36574630197171e-07, "loss": 0.0002, "step": 38422 }, { "epoch": 7.05, "learning_rate": 7.36294995528093e-07, "loss": 0.0003, "step": 38423 }, { "epoch": 7.05, "learning_rate": 7.360154119207752e-07, "loss": 0.0003, "step": 38424 }, { "epoch": 7.05, "learning_rate": 7.357358793767577e-07, "loss": 0.0003, "step": 38425 }, { "epoch": 7.05, "learning_rate": 7.354563978975804e-07, "loss": 0.0007, "step": 38426 }, { "epoch": 7.05, "learning_rate": 7.351769674847853e-07, "loss": 0.0007, "step": 38427 }, { "epoch": 7.05, "learning_rate": 7.348975881399112e-07, "loss": 0.0004, "step": 38428 }, { "epoch": 7.05, "learning_rate": 7.346182598644991e-07, "loss": 0.0004, "step": 38429 }, { "epoch": 7.05, "learning_rate": 7.343389826600878e-07, "loss": 0.0004, "step": 38430 }, { "epoch": 7.05, "learning_rate": 7.340597565282171e-07, "loss": 0.0004, "step": 38431 }, { "epoch": 7.05, "learning_rate": 7.337805814704247e-07, "loss": 0.0004, "step": 38432 }, { "epoch": 7.05, "learning_rate": 7.335014574882526e-07, "loss": 0.0006, "step": 38433 }, { "epoch": 7.05, "learning_rate": 7.332223845832387e-07, "loss": 0.0006, "step": 38434 }, { "epoch": 7.05, "learning_rate": 7.329433627569182e-07, "loss": 0.0004, "step": 38435 }, { "epoch": 7.05, "learning_rate": 7.3266439201083e-07, "loss": 0.0004, "step": 38436 }, { "epoch": 7.05, "learning_rate": 7.32385472346514e-07, "loss": 0.0005, "step": 38437 }, { "epoch": 7.05, "learning_rate": 7.321066037655078e-07, "loss": 0.0002, "step": 38438 }, { "epoch": 7.05, "learning_rate": 7.318277862693457e-07, "loss": 0.0004, "step": 38439 }, { "epoch": 7.05, "learning_rate": 7.315490198595676e-07, "loss": 0.0002, "step": 38440 }, { "epoch": 7.05, "learning_rate": 7.312703045377057e-07, "loss": 0.0003, "step": 38441 }, { "epoch": 7.05, "learning_rate": 7.309916403053008e-07, "loss": 0.0009, "step": 38442 }, { "epoch": 7.05, "learning_rate": 7.307130271638862e-07, "loss": 0.0006, "step": 38443 }, { "epoch": 7.05, "learning_rate": 7.304344651149998e-07, "loss": 0.0006, "step": 38444 }, { "epoch": 7.05, "learning_rate": 7.301559541601744e-07, "loss": 0.0004, "step": 38445 }, { "epoch": 7.05, "learning_rate": 7.298774943009479e-07, "loss": 0.0005, "step": 38446 }, { "epoch": 7.05, "learning_rate": 7.295990855388535e-07, "loss": 0.0008, "step": 38447 }, { "epoch": 7.05, "learning_rate": 7.293207278754256e-07, "loss": 0.001, "step": 38448 }, { "epoch": 7.05, "learning_rate": 7.290424213121971e-07, "loss": 0.0004, "step": 38449 }, { "epoch": 7.05, "learning_rate": 7.28764165850706e-07, "loss": 0.0003, "step": 38450 }, { "epoch": 7.05, "learning_rate": 7.284859614924867e-07, "loss": 0.0007, "step": 38451 }, { "epoch": 7.05, "learning_rate": 7.282078082390676e-07, "loss": 0.0011, "step": 38452 }, { "epoch": 7.05, "learning_rate": 7.279297060919833e-07, "loss": 0.0011, "step": 38453 }, { "epoch": 7.05, "learning_rate": 7.276516550527668e-07, "loss": 0.0006, "step": 38454 }, { "epoch": 7.05, "learning_rate": 7.273736551229538e-07, "loss": 0.0003, "step": 38455 }, { "epoch": 7.05, "learning_rate": 7.27095706304074e-07, "loss": 0.0006, "step": 38456 }, { "epoch": 7.05, "learning_rate": 7.268178085976597e-07, "loss": 0.0003, "step": 38457 }, { "epoch": 7.05, "learning_rate": 7.265399620052427e-07, "loss": 0.0005, "step": 38458 }, { "epoch": 7.05, "learning_rate": 7.262621665283553e-07, "loss": 0.0005, "step": 38459 }, { "epoch": 7.05, "learning_rate": 7.259844221685286e-07, "loss": 0.0003, "step": 38460 }, { "epoch": 7.05, "learning_rate": 7.257067289272923e-07, "loss": 0.0003, "step": 38461 }, { "epoch": 7.05, "learning_rate": 7.254290868061775e-07, "loss": 0.0006, "step": 38462 }, { "epoch": 7.05, "learning_rate": 7.251514958067152e-07, "loss": 0.0018, "step": 38463 }, { "epoch": 7.05, "learning_rate": 7.248739559304352e-07, "loss": 0.0006, "step": 38464 }, { "epoch": 7.05, "learning_rate": 7.245964671788664e-07, "loss": 0.0004, "step": 38465 }, { "epoch": 7.05, "learning_rate": 7.243190295535407e-07, "loss": 0.0003, "step": 38466 }, { "epoch": 7.05, "learning_rate": 7.240416430559827e-07, "loss": 0.0005, "step": 38467 }, { "epoch": 7.05, "learning_rate": 7.237643076877266e-07, "loss": 0.0003, "step": 38468 }, { "epoch": 7.05, "learning_rate": 7.234870234503011e-07, "loss": 0.0003, "step": 38469 }, { "epoch": 7.05, "learning_rate": 7.232097903452307e-07, "loss": 0.0003, "step": 38470 }, { "epoch": 7.06, "learning_rate": 7.229326083740429e-07, "loss": 0.0007, "step": 38471 }, { "epoch": 7.06, "learning_rate": 7.226554775382699e-07, "loss": 0.0005, "step": 38472 }, { "epoch": 7.06, "learning_rate": 7.223783978394383e-07, "loss": 0.0009, "step": 38473 }, { "epoch": 7.06, "learning_rate": 7.221013692790735e-07, "loss": 0.0006, "step": 38474 }, { "epoch": 7.06, "learning_rate": 7.218243918587031e-07, "loss": 0.0012, "step": 38475 }, { "epoch": 7.06, "learning_rate": 7.215474655798538e-07, "loss": 0.0009, "step": 38476 }, { "epoch": 7.06, "learning_rate": 7.21270590444052e-07, "loss": 0.0006, "step": 38477 }, { "epoch": 7.06, "learning_rate": 7.209937664528255e-07, "loss": 0.0008, "step": 38478 }, { "epoch": 7.06, "learning_rate": 7.207169936076974e-07, "loss": 0.0006, "step": 38479 }, { "epoch": 7.06, "learning_rate": 7.204402719101933e-07, "loss": 0.0002, "step": 38480 }, { "epoch": 7.06, "learning_rate": 7.20163601361844e-07, "loss": 0.0006, "step": 38481 }, { "epoch": 7.06, "learning_rate": 7.198869819641674e-07, "loss": 0.0005, "step": 38482 }, { "epoch": 7.06, "learning_rate": 7.196104137186921e-07, "loss": 0.0009, "step": 38483 }, { "epoch": 7.06, "learning_rate": 7.193338966269415e-07, "loss": 0.0004, "step": 38484 }, { "epoch": 7.06, "learning_rate": 7.190574306904374e-07, "loss": 0.0002, "step": 38485 }, { "epoch": 7.06, "learning_rate": 7.187810159107089e-07, "loss": 0.0004, "step": 38486 }, { "epoch": 7.06, "learning_rate": 7.18504652289278e-07, "loss": 0.0004, "step": 38487 }, { "epoch": 7.06, "learning_rate": 7.182283398276657e-07, "loss": 0.0004, "step": 38488 }, { "epoch": 7.06, "learning_rate": 7.179520785273941e-07, "loss": 0.0005, "step": 38489 }, { "epoch": 7.06, "learning_rate": 7.176758683899898e-07, "loss": 0.0005, "step": 38490 }, { "epoch": 7.06, "learning_rate": 7.173997094169738e-07, "loss": 0.0005, "step": 38491 }, { "epoch": 7.06, "learning_rate": 7.17123601609867e-07, "loss": 0.0002, "step": 38492 }, { "epoch": 7.06, "learning_rate": 7.168475449701939e-07, "loss": 0.0006, "step": 38493 }, { "epoch": 7.06, "learning_rate": 7.165715394994743e-07, "loss": 0.0006, "step": 38494 }, { "epoch": 7.06, "learning_rate": 7.162955851992292e-07, "loss": 0.0006, "step": 38495 }, { "epoch": 7.06, "learning_rate": 7.160196820709809e-07, "loss": 0.0008, "step": 38496 }, { "epoch": 7.06, "learning_rate": 7.15743830116249e-07, "loss": 0.0007, "step": 38497 }, { "epoch": 7.06, "learning_rate": 7.154680293365557e-07, "loss": 0.0003, "step": 38498 }, { "epoch": 7.06, "learning_rate": 7.151922797334187e-07, "loss": 0.0006, "step": 38499 }, { "epoch": 7.06, "learning_rate": 7.149165813083603e-07, "loss": 0.0003, "step": 38500 }, { "epoch": 7.06, "learning_rate": 7.14640934062899e-07, "loss": 0.0004, "step": 38501 }, { "epoch": 7.06, "learning_rate": 7.143653379985527e-07, "loss": 0.0006, "step": 38502 }, { "epoch": 7.06, "learning_rate": 7.140897931168445e-07, "loss": 0.0003, "step": 38503 }, { "epoch": 7.06, "learning_rate": 7.138142994192898e-07, "loss": 0.0007, "step": 38504 }, { "epoch": 7.06, "learning_rate": 7.135388569074109e-07, "loss": 0.0003, "step": 38505 }, { "epoch": 7.06, "learning_rate": 7.132634655827208e-07, "loss": 0.0005, "step": 38506 }, { "epoch": 7.06, "learning_rate": 7.129881254467385e-07, "loss": 0.0005, "step": 38507 }, { "epoch": 7.06, "learning_rate": 7.12712836500986e-07, "loss": 0.0005, "step": 38508 }, { "epoch": 7.06, "learning_rate": 7.124375987469767e-07, "loss": 0.0004, "step": 38509 }, { "epoch": 7.06, "learning_rate": 7.121624121862291e-07, "loss": 0.0003, "step": 38510 }, { "epoch": 7.06, "learning_rate": 7.118872768202611e-07, "loss": 0.0004, "step": 38511 }, { "epoch": 7.06, "learning_rate": 7.11612192650587e-07, "loss": 0.0004, "step": 38512 }, { "epoch": 7.06, "learning_rate": 7.113371596787244e-07, "loss": 0.001, "step": 38513 }, { "epoch": 7.06, "learning_rate": 7.110621779061889e-07, "loss": 0.0004, "step": 38514 }, { "epoch": 7.06, "learning_rate": 7.107872473344957e-07, "loss": 0.001, "step": 38515 }, { "epoch": 7.06, "learning_rate": 7.105123679651615e-07, "loss": 0.0006, "step": 38516 }, { "epoch": 7.06, "learning_rate": 7.102375397997008e-07, "loss": 0.0006, "step": 38517 }, { "epoch": 7.06, "learning_rate": 7.099627628396278e-07, "loss": 0.0003, "step": 38518 }, { "epoch": 7.06, "learning_rate": 7.096880370864578e-07, "loss": 0.0004, "step": 38519 }, { "epoch": 7.06, "learning_rate": 7.094133625417032e-07, "loss": 0.0003, "step": 38520 }, { "epoch": 7.06, "learning_rate": 7.091387392068826e-07, "loss": 0.0004, "step": 38521 }, { "epoch": 7.06, "learning_rate": 7.088641670835062e-07, "loss": 0.0008, "step": 38522 }, { "epoch": 7.06, "learning_rate": 7.08589646173089e-07, "loss": 0.0007, "step": 38523 }, { "epoch": 7.06, "learning_rate": 7.083151764771412e-07, "loss": 0.0004, "step": 38524 }, { "epoch": 7.06, "learning_rate": 7.080407579971794e-07, "loss": 0.0006, "step": 38525 }, { "epoch": 7.07, "learning_rate": 7.077663907347143e-07, "loss": 0.0014, "step": 38526 }, { "epoch": 7.07, "learning_rate": 7.074920746912583e-07, "loss": 0.0002, "step": 38527 }, { "epoch": 7.07, "learning_rate": 7.072178098683247e-07, "loss": 0.0003, "step": 38528 }, { "epoch": 7.07, "learning_rate": 7.069435962674231e-07, "loss": 0.0003, "step": 38529 }, { "epoch": 7.07, "learning_rate": 7.06669433890067e-07, "loss": 0.0007, "step": 38530 }, { "epoch": 7.07, "learning_rate": 7.063953227377662e-07, "loss": 0.0007, "step": 38531 }, { "epoch": 7.07, "learning_rate": 7.061212628120318e-07, "loss": 0.0012, "step": 38532 }, { "epoch": 7.07, "learning_rate": 7.058472541143735e-07, "loss": 0.0003, "step": 38533 }, { "epoch": 7.07, "learning_rate": 7.055732966463058e-07, "loss": 0.001, "step": 38534 }, { "epoch": 7.07, "learning_rate": 7.052993904093353e-07, "loss": 0.0006, "step": 38535 }, { "epoch": 7.07, "learning_rate": 7.050255354049707e-07, "loss": 0.0003, "step": 38536 }, { "epoch": 7.07, "learning_rate": 7.047517316347241e-07, "loss": 0.0006, "step": 38537 }, { "epoch": 7.07, "learning_rate": 7.044779791001022e-07, "loss": 0.001, "step": 38538 }, { "epoch": 7.07, "learning_rate": 7.04204277802617e-07, "loss": 0.0006, "step": 38539 }, { "epoch": 7.07, "learning_rate": 7.039306277437752e-07, "loss": 0.001, "step": 38540 }, { "epoch": 7.07, "learning_rate": 7.036570289250866e-07, "loss": 0.0007, "step": 38541 }, { "epoch": 7.07, "learning_rate": 7.033834813480567e-07, "loss": 0.0006, "step": 38542 }, { "epoch": 7.07, "learning_rate": 7.031099850141953e-07, "loss": 0.0003, "step": 38543 }, { "epoch": 7.07, "learning_rate": 7.028365399250093e-07, "loss": 0.0004, "step": 38544 }, { "epoch": 7.07, "learning_rate": 7.025631460820071e-07, "loss": 0.0003, "step": 38545 }, { "epoch": 7.07, "learning_rate": 7.022898034866932e-07, "loss": 0.0004, "step": 38546 }, { "epoch": 7.07, "learning_rate": 7.020165121405775e-07, "loss": 0.0007, "step": 38547 }, { "epoch": 7.07, "learning_rate": 7.017432720451633e-07, "loss": 0.0004, "step": 38548 }, { "epoch": 7.07, "learning_rate": 7.014700832019583e-07, "loss": 0.0003, "step": 38549 }, { "epoch": 7.07, "learning_rate": 7.011969456124679e-07, "loss": 0.0004, "step": 38550 }, { "epoch": 7.07, "learning_rate": 7.009238592781964e-07, "loss": 0.0007, "step": 38551 }, { "epoch": 7.07, "learning_rate": 7.006508242006538e-07, "loss": 0.0004, "step": 38552 }, { "epoch": 7.07, "learning_rate": 7.003778403813399e-07, "loss": 0.0005, "step": 38553 }, { "epoch": 7.07, "learning_rate": 7.001049078217614e-07, "loss": 0.0003, "step": 38554 }, { "epoch": 7.07, "learning_rate": 6.998320265234216e-07, "loss": 0.0005, "step": 38555 }, { "epoch": 7.07, "learning_rate": 6.995591964878257e-07, "loss": 0.0005, "step": 38556 }, { "epoch": 7.07, "learning_rate": 6.992864177164782e-07, "loss": 0.0008, "step": 38557 }, { "epoch": 7.07, "learning_rate": 6.990136902108835e-07, "loss": 0.0007, "step": 38558 }, { "epoch": 7.07, "learning_rate": 6.987410139725414e-07, "loss": 0.0008, "step": 38559 }, { "epoch": 7.07, "learning_rate": 6.984683890029554e-07, "loss": 0.0005, "step": 38560 }, { "epoch": 7.07, "learning_rate": 6.981958153036306e-07, "loss": 0.0003, "step": 38561 }, { "epoch": 7.07, "learning_rate": 6.979232928760693e-07, "loss": 0.0007, "step": 38562 }, { "epoch": 7.07, "learning_rate": 6.976508217217714e-07, "loss": 0.0006, "step": 38563 }, { "epoch": 7.07, "learning_rate": 6.973784018422414e-07, "loss": 0.0004, "step": 38564 }, { "epoch": 7.07, "learning_rate": 6.971060332389789e-07, "loss": 0.0007, "step": 38565 }, { "epoch": 7.07, "learning_rate": 6.968337159134863e-07, "loss": 0.0006, "step": 38566 }, { "epoch": 7.07, "learning_rate": 6.965614498672635e-07, "loss": 0.0006, "step": 38567 }, { "epoch": 7.07, "learning_rate": 6.962892351018124e-07, "loss": 0.0009, "step": 38568 }, { "epoch": 7.07, "learning_rate": 6.960170716186321e-07, "loss": 0.0005, "step": 38569 }, { "epoch": 7.07, "learning_rate": 6.957449594192256e-07, "loss": 0.0008, "step": 38570 }, { "epoch": 7.07, "learning_rate": 6.954728985050896e-07, "loss": 0.0005, "step": 38571 }, { "epoch": 7.07, "learning_rate": 6.952008888777262e-07, "loss": 0.0004, "step": 38572 }, { "epoch": 7.07, "learning_rate": 6.949289305386308e-07, "loss": 0.0004, "step": 38573 }, { "epoch": 7.07, "learning_rate": 6.946570234893068e-07, "loss": 0.0005, "step": 38574 }, { "epoch": 7.07, "learning_rate": 6.943851677312508e-07, "loss": 0.0005, "step": 38575 }, { "epoch": 7.07, "learning_rate": 6.941133632659636e-07, "loss": 0.0002, "step": 38576 }, { "epoch": 7.07, "learning_rate": 6.938416100949385e-07, "loss": 0.0006, "step": 38577 }, { "epoch": 7.07, "learning_rate": 6.93569908219679e-07, "loss": 0.0007, "step": 38578 }, { "epoch": 7.07, "learning_rate": 6.932982576416792e-07, "loss": 0.001, "step": 38579 }, { "epoch": 7.08, "learning_rate": 6.930266583624379e-07, "loss": 0.0005, "step": 38580 }, { "epoch": 7.08, "learning_rate": 6.927551103834507e-07, "loss": 0.0007, "step": 38581 }, { "epoch": 7.08, "learning_rate": 6.924836137062163e-07, "loss": 0.0006, "step": 38582 }, { "epoch": 7.08, "learning_rate": 6.922121683322303e-07, "loss": 0.0005, "step": 38583 }, { "epoch": 7.08, "learning_rate": 6.919407742629891e-07, "loss": 0.0003, "step": 38584 }, { "epoch": 7.08, "learning_rate": 6.916694314999872e-07, "loss": 0.0003, "step": 38585 }, { "epoch": 7.08, "learning_rate": 6.913981400447212e-07, "loss": 0.0004, "step": 38586 }, { "epoch": 7.08, "learning_rate": 6.911268998986876e-07, "loss": 0.0002, "step": 38587 }, { "epoch": 7.08, "learning_rate": 6.908557110633818e-07, "loss": 0.0002, "step": 38588 }, { "epoch": 7.08, "learning_rate": 6.905845735402972e-07, "loss": 0.0003, "step": 38589 }, { "epoch": 7.08, "learning_rate": 6.90313487330927e-07, "loss": 0.0008, "step": 38590 }, { "epoch": 7.08, "learning_rate": 6.900424524367666e-07, "loss": 0.0005, "step": 38591 }, { "epoch": 7.08, "learning_rate": 6.897714688593115e-07, "loss": 0.0006, "step": 38592 }, { "epoch": 7.08, "learning_rate": 6.89500536600054e-07, "loss": 0.0016, "step": 38593 }, { "epoch": 7.08, "learning_rate": 6.892296556604894e-07, "loss": 0.0008, "step": 38594 }, { "epoch": 7.08, "learning_rate": 6.889588260421054e-07, "loss": 0.0008, "step": 38595 }, { "epoch": 7.08, "learning_rate": 6.886880477464009e-07, "loss": 0.0003, "step": 38596 }, { "epoch": 7.08, "learning_rate": 6.884173207748656e-07, "loss": 0.0004, "step": 38597 }, { "epoch": 7.08, "learning_rate": 6.88146645128992e-07, "loss": 0.0013, "step": 38598 }, { "epoch": 7.08, "learning_rate": 6.87876020810272e-07, "loss": 0.0007, "step": 38599 }, { "epoch": 7.08, "learning_rate": 6.876054478201988e-07, "loss": 0.0005, "step": 38600 }, { "epoch": 7.08, "learning_rate": 6.873349261602613e-07, "loss": 0.0003, "step": 38601 }, { "epoch": 7.08, "learning_rate": 6.870644558319528e-07, "loss": 0.0004, "step": 38602 }, { "epoch": 7.08, "learning_rate": 6.86794036836762e-07, "loss": 0.0006, "step": 38603 }, { "epoch": 7.08, "learning_rate": 6.865236691761801e-07, "loss": 0.0004, "step": 38604 }, { "epoch": 7.08, "learning_rate": 6.862533528517002e-07, "loss": 0.0004, "step": 38605 }, { "epoch": 7.08, "learning_rate": 6.859830878648088e-07, "loss": 0.0003, "step": 38606 }, { "epoch": 7.08, "learning_rate": 6.857128742169972e-07, "loss": 0.0002, "step": 38607 }, { "epoch": 7.08, "learning_rate": 6.854427119097528e-07, "loss": 0.0009, "step": 38608 }, { "epoch": 7.08, "learning_rate": 6.851726009445681e-07, "loss": 0.0014, "step": 38609 }, { "epoch": 7.08, "learning_rate": 6.849025413229293e-07, "loss": 0.0003, "step": 38610 }, { "epoch": 7.08, "learning_rate": 6.846325330463266e-07, "loss": 0.0003, "step": 38611 }, { "epoch": 7.08, "learning_rate": 6.843625761162498e-07, "loss": 0.0004, "step": 38612 }, { "epoch": 7.08, "learning_rate": 6.84092670534181e-07, "loss": 0.0003, "step": 38613 }, { "epoch": 7.08, "learning_rate": 6.838228163016125e-07, "loss": 0.0004, "step": 38614 }, { "epoch": 7.08, "learning_rate": 6.835530134200319e-07, "loss": 0.0006, "step": 38615 }, { "epoch": 7.08, "learning_rate": 6.832832618909246e-07, "loss": 0.0008, "step": 38616 }, { "epoch": 7.08, "learning_rate": 6.830135617157796e-07, "loss": 0.0006, "step": 38617 }, { "epoch": 7.08, "learning_rate": 6.827439128960811e-07, "loss": 0.0005, "step": 38618 }, { "epoch": 7.08, "learning_rate": 6.824743154333157e-07, "loss": 0.0005, "step": 38619 }, { "epoch": 7.08, "learning_rate": 6.822047693289712e-07, "loss": 0.0005, "step": 38620 }, { "epoch": 7.08, "learning_rate": 6.819352745845309e-07, "loss": 0.0006, "step": 38621 }, { "epoch": 7.08, "learning_rate": 6.816658312014824e-07, "loss": 0.0005, "step": 38622 }, { "epoch": 7.08, "learning_rate": 6.813964391813121e-07, "loss": 0.0002, "step": 38623 }, { "epoch": 7.08, "learning_rate": 6.811270985255014e-07, "loss": 0.0004, "step": 38624 }, { "epoch": 7.08, "learning_rate": 6.808578092355366e-07, "loss": 0.0003, "step": 38625 }, { "epoch": 7.08, "learning_rate": 6.805885713129002e-07, "loss": 0.0007, "step": 38626 }, { "epoch": 7.08, "learning_rate": 6.803193847590795e-07, "loss": 0.0005, "step": 38627 }, { "epoch": 7.08, "learning_rate": 6.800502495755568e-07, "loss": 0.0005, "step": 38628 }, { "epoch": 7.08, "learning_rate": 6.797811657638154e-07, "loss": 0.0006, "step": 38629 }, { "epoch": 7.08, "learning_rate": 6.795121333253385e-07, "loss": 0.0005, "step": 38630 }, { "epoch": 7.08, "learning_rate": 6.792431522616094e-07, "loss": 0.0007, "step": 38631 }, { "epoch": 7.08, "learning_rate": 6.789742225741092e-07, "loss": 0.0011, "step": 38632 }, { "epoch": 7.08, "learning_rate": 6.787053442643233e-07, "loss": 0.0004, "step": 38633 }, { "epoch": 7.08, "learning_rate": 6.784365173337304e-07, "loss": 0.0002, "step": 38634 }, { "epoch": 7.09, "learning_rate": 6.781677417838139e-07, "loss": 0.0007, "step": 38635 }, { "epoch": 7.09, "learning_rate": 6.77899017616056e-07, "loss": 0.0006, "step": 38636 }, { "epoch": 7.09, "learning_rate": 6.776303448319365e-07, "loss": 0.0003, "step": 38637 }, { "epoch": 7.09, "learning_rate": 6.773617234329377e-07, "loss": 0.0005, "step": 38638 }, { "epoch": 7.09, "learning_rate": 6.770931534205371e-07, "loss": 0.0003, "step": 38639 }, { "epoch": 7.09, "learning_rate": 6.768246347962193e-07, "loss": 0.0002, "step": 38640 }, { "epoch": 7.09, "learning_rate": 6.76556167561464e-07, "loss": 0.001, "step": 38641 }, { "epoch": 7.09, "learning_rate": 6.76287751717748e-07, "loss": 0.0005, "step": 38642 }, { "epoch": 7.09, "learning_rate": 6.760193872665522e-07, "loss": 0.0005, "step": 38643 }, { "epoch": 7.09, "learning_rate": 6.757510742093553e-07, "loss": 0.0004, "step": 38644 }, { "epoch": 7.09, "learning_rate": 6.754828125476376e-07, "loss": 0.0004, "step": 38645 }, { "epoch": 7.09, "learning_rate": 6.752146022828776e-07, "loss": 0.0004, "step": 38646 }, { "epoch": 7.09, "learning_rate": 6.749464434165531e-07, "loss": 0.0006, "step": 38647 }, { "epoch": 7.09, "learning_rate": 6.746783359501419e-07, "loss": 0.0005, "step": 38648 }, { "epoch": 7.09, "learning_rate": 6.744102798851216e-07, "loss": 0.0006, "step": 38649 }, { "epoch": 7.09, "learning_rate": 6.741422752229709e-07, "loss": 0.0008, "step": 38650 }, { "epoch": 7.09, "learning_rate": 6.738743219651667e-07, "loss": 0.0003, "step": 38651 }, { "epoch": 7.09, "learning_rate": 6.736064201131865e-07, "loss": 0.0006, "step": 38652 }, { "epoch": 7.09, "learning_rate": 6.733385696685047e-07, "loss": 0.0007, "step": 38653 }, { "epoch": 7.09, "learning_rate": 6.730707706326001e-07, "loss": 0.0003, "step": 38654 }, { "epoch": 7.09, "learning_rate": 6.728030230069471e-07, "loss": 0.0007, "step": 38655 }, { "epoch": 7.09, "learning_rate": 6.725353267930234e-07, "loss": 0.0003, "step": 38656 }, { "epoch": 7.09, "learning_rate": 6.722676819923013e-07, "loss": 0.0003, "step": 38657 }, { "epoch": 7.09, "learning_rate": 6.720000886062605e-07, "loss": 0.0004, "step": 38658 }, { "epoch": 7.09, "learning_rate": 6.717325466363744e-07, "loss": 0.0008, "step": 38659 }, { "epoch": 7.09, "learning_rate": 6.714650560841162e-07, "loss": 0.0007, "step": 38660 }, { "epoch": 7.09, "learning_rate": 6.711976169509593e-07, "loss": 0.0003, "step": 38661 }, { "epoch": 7.09, "learning_rate": 6.709302292383813e-07, "loss": 0.0004, "step": 38662 }, { "epoch": 7.09, "learning_rate": 6.706628929478543e-07, "loss": 0.0006, "step": 38663 }, { "epoch": 7.09, "learning_rate": 6.703956080808515e-07, "loss": 0.0005, "step": 38664 }, { "epoch": 7.09, "learning_rate": 6.701283746388477e-07, "loss": 0.0002, "step": 38665 }, { "epoch": 7.09, "learning_rate": 6.698611926233134e-07, "loss": 0.0007, "step": 38666 }, { "epoch": 7.09, "learning_rate": 6.695940620357244e-07, "loss": 0.0003, "step": 38667 }, { "epoch": 7.09, "learning_rate": 6.693269828775506e-07, "loss": 0.0005, "step": 38668 }, { "epoch": 7.09, "learning_rate": 6.690599551502652e-07, "loss": 0.0006, "step": 38669 }, { "epoch": 7.09, "learning_rate": 6.687929788553404e-07, "loss": 0.0005, "step": 38670 }, { "epoch": 7.09, "learning_rate": 6.685260539942473e-07, "loss": 0.0006, "step": 38671 }, { "epoch": 7.09, "learning_rate": 6.682591805684568e-07, "loss": 0.0009, "step": 38672 }, { "epoch": 7.09, "learning_rate": 6.679923585794401e-07, "loss": 0.0004, "step": 38673 }, { "epoch": 7.09, "learning_rate": 6.67725588028667e-07, "loss": 0.0004, "step": 38674 }, { "epoch": 7.09, "learning_rate": 6.67458868917611e-07, "loss": 0.0006, "step": 38675 }, { "epoch": 7.09, "learning_rate": 6.671922012477394e-07, "loss": 0.0003, "step": 38676 }, { "epoch": 7.09, "learning_rate": 6.669255850205247e-07, "loss": 0.0003, "step": 38677 }, { "epoch": 7.09, "learning_rate": 6.666590202374334e-07, "loss": 0.0004, "step": 38678 }, { "epoch": 7.09, "learning_rate": 6.663925068999355e-07, "loss": 0.0006, "step": 38679 }, { "epoch": 7.09, "learning_rate": 6.661260450095019e-07, "loss": 0.0006, "step": 38680 }, { "epoch": 7.09, "learning_rate": 6.658596345675994e-07, "loss": 0.0007, "step": 38681 }, { "epoch": 7.09, "learning_rate": 6.655932755756988e-07, "loss": 0.0004, "step": 38682 }, { "epoch": 7.09, "learning_rate": 6.653269680352659e-07, "loss": 0.0005, "step": 38683 }, { "epoch": 7.09, "learning_rate": 6.650607119477692e-07, "loss": 0.0003, "step": 38684 }, { "epoch": 7.09, "learning_rate": 6.647945073146766e-07, "loss": 0.0006, "step": 38685 }, { "epoch": 7.09, "learning_rate": 6.645283541374559e-07, "loss": 0.0005, "step": 38686 }, { "epoch": 7.09, "learning_rate": 6.642622524175723e-07, "loss": 0.0005, "step": 38687 }, { "epoch": 7.09, "learning_rate": 6.63996202156495e-07, "loss": 0.0005, "step": 38688 }, { "epoch": 7.09, "learning_rate": 6.637302033556891e-07, "loss": 0.0007, "step": 38689 }, { "epoch": 7.1, "learning_rate": 6.634642560166216e-07, "loss": 0.0004, "step": 38690 }, { "epoch": 7.1, "learning_rate": 6.631983601407576e-07, "loss": 0.0007, "step": 38691 }, { "epoch": 7.1, "learning_rate": 6.629325157295607e-07, "loss": 0.0003, "step": 38692 }, { "epoch": 7.1, "learning_rate": 6.626667227845008e-07, "loss": 0.0004, "step": 38693 }, { "epoch": 7.1, "learning_rate": 6.624009813070409e-07, "loss": 0.0005, "step": 38694 }, { "epoch": 7.1, "learning_rate": 6.621352912986468e-07, "loss": 0.0005, "step": 38695 }, { "epoch": 7.1, "learning_rate": 6.618696527607793e-07, "loss": 0.0005, "step": 38696 }, { "epoch": 7.1, "learning_rate": 6.616040656949041e-07, "loss": 0.0004, "step": 38697 }, { "epoch": 7.1, "learning_rate": 6.613385301024877e-07, "loss": 0.0004, "step": 38698 }, { "epoch": 7.1, "learning_rate": 6.610730459849923e-07, "loss": 0.0006, "step": 38699 }, { "epoch": 7.1, "learning_rate": 6.608076133438813e-07, "loss": 0.0006, "step": 38700 }, { "epoch": 7.1, "learning_rate": 6.605422321806165e-07, "loss": 0.0005, "step": 38701 }, { "epoch": 7.1, "learning_rate": 6.602769024966627e-07, "loss": 0.0003, "step": 38702 }, { "epoch": 7.1, "learning_rate": 6.600116242934817e-07, "loss": 0.0006, "step": 38703 }, { "epoch": 7.1, "learning_rate": 6.597463975725348e-07, "loss": 0.0008, "step": 38704 }, { "epoch": 7.1, "learning_rate": 6.594812223352831e-07, "loss": 0.0002, "step": 38705 }, { "epoch": 7.1, "learning_rate": 6.59216098583192e-07, "loss": 0.0003, "step": 38706 }, { "epoch": 7.1, "learning_rate": 6.589510263177201e-07, "loss": 0.0003, "step": 38707 }, { "epoch": 7.1, "learning_rate": 6.586860055403288e-07, "loss": 0.0004, "step": 38708 }, { "epoch": 7.1, "learning_rate": 6.58421036252479e-07, "loss": 0.0003, "step": 38709 }, { "epoch": 7.1, "learning_rate": 6.581561184556296e-07, "loss": 0.0011, "step": 38710 }, { "epoch": 7.1, "learning_rate": 6.578912521512448e-07, "loss": 0.0008, "step": 38711 }, { "epoch": 7.1, "learning_rate": 6.576264373407815e-07, "loss": 0.0003, "step": 38712 }, { "epoch": 7.1, "learning_rate": 6.573616740257016e-07, "loss": 0.0003, "step": 38713 }, { "epoch": 7.1, "learning_rate": 6.570969622074597e-07, "loss": 0.0002, "step": 38714 }, { "epoch": 7.1, "learning_rate": 6.568323018875211e-07, "loss": 0.0006, "step": 38715 }, { "epoch": 7.1, "learning_rate": 6.565676930673404e-07, "loss": 0.0005, "step": 38716 }, { "epoch": 7.1, "learning_rate": 6.563031357483785e-07, "loss": 0.0003, "step": 38717 }, { "epoch": 7.1, "learning_rate": 6.56038629932092e-07, "loss": 0.0008, "step": 38718 }, { "epoch": 7.1, "learning_rate": 6.557741756199399e-07, "loss": 0.0004, "step": 38719 }, { "epoch": 7.1, "learning_rate": 6.555097728133786e-07, "loss": 0.0008, "step": 38720 }, { "epoch": 7.1, "learning_rate": 6.55245421513867e-07, "loss": 0.0004, "step": 38721 }, { "epoch": 7.1, "learning_rate": 6.549811217228618e-07, "loss": 0.0007, "step": 38722 }, { "epoch": 7.1, "learning_rate": 6.547168734418174e-07, "loss": 0.0006, "step": 38723 }, { "epoch": 7.1, "learning_rate": 6.544526766721959e-07, "loss": 0.0013, "step": 38724 }, { "epoch": 7.1, "learning_rate": 6.541885314154484e-07, "loss": 0.0005, "step": 38725 }, { "epoch": 7.1, "learning_rate": 6.539244376730336e-07, "loss": 0.0005, "step": 38726 }, { "epoch": 7.1, "learning_rate": 6.536603954464038e-07, "loss": 0.0003, "step": 38727 }, { "epoch": 7.1, "learning_rate": 6.53396404737019e-07, "loss": 0.0006, "step": 38728 }, { "epoch": 7.1, "learning_rate": 6.531324655463311e-07, "loss": 0.0003, "step": 38729 }, { "epoch": 7.1, "learning_rate": 6.528685778757982e-07, "loss": 0.0008, "step": 38730 }, { "epoch": 7.1, "learning_rate": 6.52604741726871e-07, "loss": 0.0002, "step": 38731 }, { "epoch": 7.1, "learning_rate": 6.523409571010042e-07, "loss": 0.0003, "step": 38732 }, { "epoch": 7.1, "learning_rate": 6.520772239996553e-07, "loss": 0.0003, "step": 38733 }, { "epoch": 7.1, "learning_rate": 6.518135424242744e-07, "loss": 0.0006, "step": 38734 }, { "epoch": 7.1, "learning_rate": 6.51549912376317e-07, "loss": 0.0007, "step": 38735 }, { "epoch": 7.1, "learning_rate": 6.51286333857235e-07, "loss": 0.0004, "step": 38736 }, { "epoch": 7.1, "learning_rate": 6.510228068684821e-07, "loss": 0.0003, "step": 38737 }, { "epoch": 7.1, "learning_rate": 6.507593314115113e-07, "loss": 0.0007, "step": 38738 }, { "epoch": 7.1, "learning_rate": 6.504959074877725e-07, "loss": 0.0003, "step": 38739 }, { "epoch": 7.1, "learning_rate": 6.502325350987204e-07, "loss": 0.0005, "step": 38740 }, { "epoch": 7.1, "learning_rate": 6.499692142458036e-07, "loss": 0.0008, "step": 38741 }, { "epoch": 7.1, "learning_rate": 6.497059449304788e-07, "loss": 0.0005, "step": 38742 }, { "epoch": 7.1, "learning_rate": 6.494427271541926e-07, "loss": 0.0003, "step": 38743 }, { "epoch": 7.11, "learning_rate": 6.491795609183971e-07, "loss": 0.0005, "step": 38744 }, { "epoch": 7.11, "learning_rate": 6.489164462245401e-07, "loss": 0.0001, "step": 38745 }, { "epoch": 7.11, "learning_rate": 6.486533830740782e-07, "loss": 0.001, "step": 38746 }, { "epoch": 7.11, "learning_rate": 6.48390371468457e-07, "loss": 0.0003, "step": 38747 }, { "epoch": 7.11, "learning_rate": 6.481274114091285e-07, "loss": 0.0002, "step": 38748 }, { "epoch": 7.11, "learning_rate": 6.478645028975372e-07, "loss": 0.0004, "step": 38749 }, { "epoch": 7.11, "learning_rate": 6.476016459351387e-07, "loss": 0.0012, "step": 38750 }, { "epoch": 7.11, "learning_rate": 6.473388405233783e-07, "loss": 0.0006, "step": 38751 }, { "epoch": 7.11, "learning_rate": 6.47076086663706e-07, "loss": 0.0003, "step": 38752 }, { "epoch": 7.11, "learning_rate": 6.468133843575686e-07, "loss": 0.0006, "step": 38753 }, { "epoch": 7.11, "learning_rate": 6.465507336064147e-07, "loss": 0.0003, "step": 38754 }, { "epoch": 7.11, "learning_rate": 6.462881344116934e-07, "loss": 0.0005, "step": 38755 }, { "epoch": 7.11, "learning_rate": 6.4602558677485e-07, "loss": 0.0004, "step": 38756 }, { "epoch": 7.11, "learning_rate": 6.457630906973333e-07, "loss": 0.0004, "step": 38757 }, { "epoch": 7.11, "learning_rate": 6.45500646180588e-07, "loss": 0.0003, "step": 38758 }, { "epoch": 7.11, "learning_rate": 6.452382532260659e-07, "loss": 0.0005, "step": 38759 }, { "epoch": 7.11, "learning_rate": 6.449759118352072e-07, "loss": 0.0003, "step": 38760 }, { "epoch": 7.11, "learning_rate": 6.447136220094607e-07, "loss": 0.0004, "step": 38761 }, { "epoch": 7.11, "learning_rate": 6.444513837502719e-07, "loss": 0.0002, "step": 38762 }, { "epoch": 7.11, "learning_rate": 6.441891970590852e-07, "loss": 0.0004, "step": 38763 }, { "epoch": 7.11, "learning_rate": 6.439270619373472e-07, "loss": 0.0008, "step": 38764 }, { "epoch": 7.11, "learning_rate": 6.436649783865034e-07, "loss": 0.0004, "step": 38765 }, { "epoch": 7.11, "learning_rate": 6.434029464079994e-07, "loss": 0.0008, "step": 38766 }, { "epoch": 7.11, "learning_rate": 6.431409660032739e-07, "loss": 0.0002, "step": 38767 }, { "epoch": 7.11, "learning_rate": 6.428790371737759e-07, "loss": 0.0014, "step": 38768 }, { "epoch": 7.11, "learning_rate": 6.426171599209485e-07, "loss": 0.0005, "step": 38769 }, { "epoch": 7.11, "learning_rate": 6.423553342462341e-07, "loss": 0.0005, "step": 38770 }, { "epoch": 7.11, "learning_rate": 6.420935601510758e-07, "loss": 0.0007, "step": 38771 }, { "epoch": 7.11, "learning_rate": 6.41831837636917e-07, "loss": 0.0004, "step": 38772 }, { "epoch": 7.11, "learning_rate": 6.41570166705201e-07, "loss": 0.0004, "step": 38773 }, { "epoch": 7.11, "learning_rate": 6.4130854735737e-07, "loss": 0.001, "step": 38774 }, { "epoch": 7.11, "learning_rate": 6.410469795948648e-07, "loss": 0.0005, "step": 38775 }, { "epoch": 7.11, "learning_rate": 6.407854634191268e-07, "loss": 0.0008, "step": 38776 }, { "epoch": 7.11, "learning_rate": 6.405239988316015e-07, "loss": 0.0006, "step": 38777 }, { "epoch": 7.11, "learning_rate": 6.402625858337253e-07, "loss": 0.0003, "step": 38778 }, { "epoch": 7.11, "learning_rate": 6.400012244269405e-07, "loss": 0.0004, "step": 38779 }, { "epoch": 7.11, "learning_rate": 6.397399146126881e-07, "loss": 0.001, "step": 38780 }, { "epoch": 7.11, "learning_rate": 6.394786563924093e-07, "loss": 0.0009, "step": 38781 }, { "epoch": 7.11, "learning_rate": 6.39217449767543e-07, "loss": 0.0004, "step": 38782 }, { "epoch": 7.11, "learning_rate": 6.389562947395301e-07, "loss": 0.0004, "step": 38783 }, { "epoch": 7.11, "learning_rate": 6.386951913098117e-07, "loss": 0.0003, "step": 38784 }, { "epoch": 7.11, "learning_rate": 6.384341394798199e-07, "loss": 0.0004, "step": 38785 }, { "epoch": 7.11, "learning_rate": 6.381731392510015e-07, "loss": 0.0007, "step": 38786 }, { "epoch": 7.11, "learning_rate": 6.379121906247909e-07, "loss": 0.0003, "step": 38787 }, { "epoch": 7.11, "learning_rate": 6.37651293602628e-07, "loss": 0.0002, "step": 38788 }, { "epoch": 7.11, "learning_rate": 6.373904481859505e-07, "loss": 0.0007, "step": 38789 }, { "epoch": 7.11, "learning_rate": 6.371296543761973e-07, "loss": 0.0003, "step": 38790 }, { "epoch": 7.11, "learning_rate": 6.36868912174804e-07, "loss": 0.0003, "step": 38791 }, { "epoch": 7.11, "learning_rate": 6.366082215832082e-07, "loss": 0.0005, "step": 38792 }, { "epoch": 7.11, "learning_rate": 6.363475826028476e-07, "loss": 0.0004, "step": 38793 }, { "epoch": 7.11, "learning_rate": 6.360869952351568e-07, "loss": 0.0004, "step": 38794 }, { "epoch": 7.11, "learning_rate": 6.358264594815766e-07, "loss": 0.0004, "step": 38795 }, { "epoch": 7.11, "learning_rate": 6.355659753435394e-07, "loss": 0.0004, "step": 38796 }, { "epoch": 7.11, "learning_rate": 6.353055428224819e-07, "loss": 0.0006, "step": 38797 }, { "epoch": 7.11, "learning_rate": 6.350451619198372e-07, "loss": 0.0002, "step": 38798 }, { "epoch": 7.12, "learning_rate": 6.347848326370454e-07, "loss": 0.0003, "step": 38799 }, { "epoch": 7.12, "learning_rate": 6.345245549755386e-07, "loss": 0.0003, "step": 38800 }, { "epoch": 7.12, "learning_rate": 6.342643289367523e-07, "loss": 0.0003, "step": 38801 }, { "epoch": 7.12, "learning_rate": 6.34004154522121e-07, "loss": 0.0002, "step": 38802 }, { "epoch": 7.12, "learning_rate": 6.337440317330768e-07, "loss": 0.0004, "step": 38803 }, { "epoch": 7.12, "learning_rate": 6.334839605710552e-07, "loss": 0.0004, "step": 38804 }, { "epoch": 7.12, "learning_rate": 6.332239410374908e-07, "loss": 0.0004, "step": 38805 }, { "epoch": 7.12, "learning_rate": 6.329639731338145e-07, "loss": 0.0003, "step": 38806 }, { "epoch": 7.12, "learning_rate": 6.327040568614595e-07, "loss": 0.0003, "step": 38807 }, { "epoch": 7.12, "learning_rate": 6.324441922218605e-07, "loss": 0.0005, "step": 38808 }, { "epoch": 7.12, "learning_rate": 6.321843792164472e-07, "loss": 0.0006, "step": 38809 }, { "epoch": 7.12, "learning_rate": 6.319246178466543e-07, "loss": 0.0003, "step": 38810 }, { "epoch": 7.12, "learning_rate": 6.316649081139104e-07, "loss": 0.0005, "step": 38811 }, { "epoch": 7.12, "learning_rate": 6.31405250019651e-07, "loss": 0.0006, "step": 38812 }, { "epoch": 7.12, "learning_rate": 6.311456435653062e-07, "loss": 0.0005, "step": 38813 }, { "epoch": 7.12, "learning_rate": 6.308860887523049e-07, "loss": 0.0005, "step": 38814 }, { "epoch": 7.12, "learning_rate": 6.306265855820781e-07, "loss": 0.0004, "step": 38815 }, { "epoch": 7.12, "learning_rate": 6.303671340560569e-07, "loss": 0.0003, "step": 38816 }, { "epoch": 7.12, "learning_rate": 6.301077341756723e-07, "loss": 0.0007, "step": 38817 }, { "epoch": 7.12, "learning_rate": 6.298483859423532e-07, "loss": 0.0004, "step": 38818 }, { "epoch": 7.12, "learning_rate": 6.295890893575296e-07, "loss": 0.0009, "step": 38819 }, { "epoch": 7.12, "learning_rate": 6.293298444226304e-07, "loss": 0.0009, "step": 38820 }, { "epoch": 7.12, "learning_rate": 6.290706511390853e-07, "loss": 0.0007, "step": 38821 }, { "epoch": 7.12, "learning_rate": 6.288115095083214e-07, "loss": 0.0005, "step": 38822 }, { "epoch": 7.12, "learning_rate": 6.285524195317682e-07, "loss": 0.0007, "step": 38823 }, { "epoch": 7.12, "learning_rate": 6.282933812108538e-07, "loss": 0.0005, "step": 38824 }, { "epoch": 7.12, "learning_rate": 6.280343945470069e-07, "loss": 0.0005, "step": 38825 }, { "epoch": 7.12, "learning_rate": 6.277754595416519e-07, "loss": 0.0007, "step": 38826 }, { "epoch": 7.12, "learning_rate": 6.275165761962199e-07, "loss": 0.0004, "step": 38827 }, { "epoch": 7.12, "learning_rate": 6.272577445121353e-07, "loss": 0.0007, "step": 38828 }, { "epoch": 7.12, "learning_rate": 6.269989644908237e-07, "loss": 0.0004, "step": 38829 }, { "epoch": 7.12, "learning_rate": 6.26740236133716e-07, "loss": 0.0008, "step": 38830 }, { "epoch": 7.12, "learning_rate": 6.264815594422358e-07, "loss": 0.0005, "step": 38831 }, { "epoch": 7.12, "learning_rate": 6.262229344178083e-07, "loss": 0.0006, "step": 38832 }, { "epoch": 7.12, "learning_rate": 6.259643610618582e-07, "loss": 0.0004, "step": 38833 }, { "epoch": 7.12, "learning_rate": 6.257058393758131e-07, "loss": 0.0004, "step": 38834 }, { "epoch": 7.12, "learning_rate": 6.254473693610974e-07, "loss": 0.0005, "step": 38835 }, { "epoch": 7.12, "learning_rate": 6.251889510191355e-07, "loss": 0.0007, "step": 38836 }, { "epoch": 7.12, "learning_rate": 6.249305843513531e-07, "loss": 0.0004, "step": 38837 }, { "epoch": 7.12, "learning_rate": 6.246722693591689e-07, "loss": 0.0005, "step": 38838 }, { "epoch": 7.12, "learning_rate": 6.24414006044014e-07, "loss": 0.001, "step": 38839 }, { "epoch": 7.12, "learning_rate": 6.241557944073074e-07, "loss": 0.001, "step": 38840 }, { "epoch": 7.12, "learning_rate": 6.238976344504743e-07, "loss": 0.0004, "step": 38841 }, { "epoch": 7.12, "learning_rate": 6.236395261749373e-07, "loss": 0.0003, "step": 38842 }, { "epoch": 7.12, "learning_rate": 6.233814695821194e-07, "loss": 0.0005, "step": 38843 }, { "epoch": 7.12, "learning_rate": 6.231234646734419e-07, "loss": 0.0004, "step": 38844 }, { "epoch": 7.12, "learning_rate": 6.228655114503268e-07, "loss": 0.0011, "step": 38845 }, { "epoch": 7.12, "learning_rate": 6.226076099141975e-07, "loss": 0.0011, "step": 38846 }, { "epoch": 7.12, "learning_rate": 6.22349760066473e-07, "loss": 0.0005, "step": 38847 }, { "epoch": 7.12, "learning_rate": 6.220919619085786e-07, "loss": 0.0005, "step": 38848 }, { "epoch": 7.12, "learning_rate": 6.218342154419332e-07, "loss": 0.0003, "step": 38849 }, { "epoch": 7.12, "learning_rate": 6.215765206679569e-07, "loss": 0.0003, "step": 38850 }, { "epoch": 7.12, "learning_rate": 6.213188775880686e-07, "loss": 0.0003, "step": 38851 }, { "epoch": 7.12, "learning_rate": 6.210612862036913e-07, "loss": 0.0003, "step": 38852 }, { "epoch": 7.13, "learning_rate": 6.208037465162442e-07, "loss": 0.0002, "step": 38853 }, { "epoch": 7.13, "learning_rate": 6.205462585271472e-07, "loss": 0.0005, "step": 38854 }, { "epoch": 7.13, "learning_rate": 6.202888222378189e-07, "loss": 0.0005, "step": 38855 }, { "epoch": 7.13, "learning_rate": 6.200314376496774e-07, "loss": 0.0004, "step": 38856 }, { "epoch": 7.13, "learning_rate": 6.197741047641436e-07, "loss": 0.0006, "step": 38857 }, { "epoch": 7.13, "learning_rate": 6.195168235826332e-07, "loss": 0.0007, "step": 38858 }, { "epoch": 7.13, "learning_rate": 6.192595941065671e-07, "loss": 0.001, "step": 38859 }, { "epoch": 7.13, "learning_rate": 6.19002416337362e-07, "loss": 0.0005, "step": 38860 }, { "epoch": 7.13, "learning_rate": 6.187452902764346e-07, "loss": 0.0003, "step": 38861 }, { "epoch": 7.13, "learning_rate": 6.184882159252037e-07, "loss": 0.0006, "step": 38862 }, { "epoch": 7.13, "learning_rate": 6.182311932850849e-07, "loss": 0.0003, "step": 38863 }, { "epoch": 7.13, "learning_rate": 6.179742223574936e-07, "loss": 0.0005, "step": 38864 }, { "epoch": 7.13, "learning_rate": 6.177173031438499e-07, "loss": 0.0002, "step": 38865 }, { "epoch": 7.13, "learning_rate": 6.174604356455693e-07, "loss": 0.0003, "step": 38866 }, { "epoch": 7.13, "learning_rate": 6.172036198640674e-07, "loss": 0.0005, "step": 38867 }, { "epoch": 7.13, "learning_rate": 6.169468558007574e-07, "loss": 0.0006, "step": 38868 }, { "epoch": 7.13, "learning_rate": 6.166901434570549e-07, "loss": 0.0008, "step": 38869 }, { "epoch": 7.13, "learning_rate": 6.164334828343777e-07, "loss": 0.0004, "step": 38870 }, { "epoch": 7.13, "learning_rate": 6.161768739341378e-07, "loss": 0.0014, "step": 38871 }, { "epoch": 7.13, "learning_rate": 6.159203167577521e-07, "loss": 0.0005, "step": 38872 }, { "epoch": 7.13, "learning_rate": 6.156638113066338e-07, "loss": 0.0002, "step": 38873 }, { "epoch": 7.13, "learning_rate": 6.15407357582195e-07, "loss": 0.0006, "step": 38874 }, { "epoch": 7.13, "learning_rate": 6.151509555858515e-07, "loss": 0.0004, "step": 38875 }, { "epoch": 7.13, "learning_rate": 6.148946053190164e-07, "loss": 0.0008, "step": 38876 }, { "epoch": 7.13, "learning_rate": 6.146383067830997e-07, "loss": 0.0006, "step": 38877 }, { "epoch": 7.13, "learning_rate": 6.143820599795192e-07, "loss": 0.0003, "step": 38878 }, { "epoch": 7.13, "learning_rate": 6.141258649096837e-07, "loss": 0.0005, "step": 38879 }, { "epoch": 7.13, "learning_rate": 6.138697215750066e-07, "loss": 0.0001, "step": 38880 }, { "epoch": 7.13, "learning_rate": 6.136136299768991e-07, "loss": 0.0003, "step": 38881 }, { "epoch": 7.13, "learning_rate": 6.133575901167721e-07, "loss": 0.0003, "step": 38882 }, { "epoch": 7.13, "learning_rate": 6.13101601996039e-07, "loss": 0.0003, "step": 38883 }, { "epoch": 7.13, "learning_rate": 6.128456656161119e-07, "loss": 0.0007, "step": 38884 }, { "epoch": 7.13, "learning_rate": 6.125897809783977e-07, "loss": 0.0011, "step": 38885 }, { "epoch": 7.13, "learning_rate": 6.12333948084306e-07, "loss": 0.0005, "step": 38886 }, { "epoch": 7.13, "learning_rate": 6.120781669352527e-07, "loss": 0.0003, "step": 38887 }, { "epoch": 7.13, "learning_rate": 6.118224375326431e-07, "loss": 0.0004, "step": 38888 }, { "epoch": 7.13, "learning_rate": 6.115667598778885e-07, "loss": 0.0003, "step": 38889 }, { "epoch": 7.13, "learning_rate": 6.113111339723987e-07, "loss": 0.0004, "step": 38890 }, { "epoch": 7.13, "learning_rate": 6.110555598175805e-07, "loss": 0.0002, "step": 38891 }, { "epoch": 7.13, "learning_rate": 6.108000374148448e-07, "loss": 0.0004, "step": 38892 }, { "epoch": 7.13, "learning_rate": 6.105445667655996e-07, "loss": 0.0003, "step": 38893 }, { "epoch": 7.13, "learning_rate": 6.102891478712525e-07, "loss": 0.0006, "step": 38894 }, { "epoch": 7.13, "learning_rate": 6.100337807332091e-07, "loss": 0.0009, "step": 38895 }, { "epoch": 7.13, "learning_rate": 6.097784653528838e-07, "loss": 0.0002, "step": 38896 }, { "epoch": 7.13, "learning_rate": 6.095232017316777e-07, "loss": 0.0011, "step": 38897 }, { "epoch": 7.13, "learning_rate": 6.092679898709997e-07, "loss": 0.0002, "step": 38898 }, { "epoch": 7.13, "learning_rate": 6.090128297722564e-07, "loss": 0.0006, "step": 38899 }, { "epoch": 7.13, "learning_rate": 6.087577214368523e-07, "loss": 0.0004, "step": 38900 }, { "epoch": 7.13, "learning_rate": 6.085026648661984e-07, "loss": 0.0014, "step": 38901 }, { "epoch": 7.13, "learning_rate": 6.082476600616982e-07, "loss": 0.0007, "step": 38902 }, { "epoch": 7.13, "learning_rate": 6.07992707024756e-07, "loss": 0.0005, "step": 38903 }, { "epoch": 7.13, "learning_rate": 6.077378057567762e-07, "loss": 0.0005, "step": 38904 }, { "epoch": 7.13, "learning_rate": 6.074829562591677e-07, "loss": 0.0001, "step": 38905 }, { "epoch": 7.13, "learning_rate": 6.072281585333329e-07, "loss": 0.0002, "step": 38906 }, { "epoch": 7.13, "learning_rate": 6.06973412580677e-07, "loss": 0.0011, "step": 38907 }, { "epoch": 7.14, "learning_rate": 6.067187184026035e-07, "loss": 0.0003, "step": 38908 }, { "epoch": 7.14, "learning_rate": 6.064640760005158e-07, "loss": 0.0009, "step": 38909 }, { "epoch": 7.14, "learning_rate": 6.062094853758183e-07, "loss": 0.0005, "step": 38910 }, { "epoch": 7.14, "learning_rate": 6.059549465299153e-07, "loss": 0.0003, "step": 38911 }, { "epoch": 7.14, "learning_rate": 6.05700459464208e-07, "loss": 0.0005, "step": 38912 }, { "epoch": 7.14, "learning_rate": 6.054460241800975e-07, "loss": 0.001, "step": 38913 }, { "epoch": 7.14, "learning_rate": 6.051916406789926e-07, "loss": 0.0004, "step": 38914 }, { "epoch": 7.14, "learning_rate": 6.0493730896229e-07, "loss": 0.0004, "step": 38915 }, { "epoch": 7.14, "learning_rate": 6.046830290313921e-07, "loss": 0.0003, "step": 38916 }, { "epoch": 7.14, "learning_rate": 6.044288008877009e-07, "loss": 0.0003, "step": 38917 }, { "epoch": 7.14, "learning_rate": 6.041746245326196e-07, "loss": 0.0002, "step": 38918 }, { "epoch": 7.14, "learning_rate": 6.039204999675486e-07, "loss": 0.0007, "step": 38919 }, { "epoch": 7.14, "learning_rate": 6.036664271938886e-07, "loss": 0.0004, "step": 38920 }, { "epoch": 7.14, "learning_rate": 6.034124062130375e-07, "loss": 0.0005, "step": 38921 }, { "epoch": 7.14, "learning_rate": 6.031584370263965e-07, "loss": 0.0013, "step": 38922 }, { "epoch": 7.14, "learning_rate": 6.029045196353677e-07, "loss": 0.0004, "step": 38923 }, { "epoch": 7.14, "learning_rate": 6.026506540413491e-07, "loss": 0.0005, "step": 38924 }, { "epoch": 7.14, "learning_rate": 6.023968402457403e-07, "loss": 0.0003, "step": 38925 }, { "epoch": 7.14, "learning_rate": 6.021430782499405e-07, "loss": 0.0005, "step": 38926 }, { "epoch": 7.14, "learning_rate": 6.018893680553484e-07, "loss": 0.0012, "step": 38927 }, { "epoch": 7.14, "learning_rate": 6.016357096633618e-07, "loss": 0.0004, "step": 38928 }, { "epoch": 7.14, "learning_rate": 6.013821030753797e-07, "loss": 0.0007, "step": 38929 }, { "epoch": 7.14, "learning_rate": 6.011285482927975e-07, "loss": 0.0006, "step": 38930 }, { "epoch": 7.14, "learning_rate": 6.008750453170186e-07, "loss": 0.0007, "step": 38931 }, { "epoch": 7.14, "learning_rate": 6.006215941494354e-07, "loss": 0.0008, "step": 38932 }, { "epoch": 7.14, "learning_rate": 6.003681947914453e-07, "loss": 0.0005, "step": 38933 }, { "epoch": 7.14, "learning_rate": 6.001148472444462e-07, "loss": 0.0003, "step": 38934 }, { "epoch": 7.14, "learning_rate": 5.998615515098327e-07, "loss": 0.0008, "step": 38935 }, { "epoch": 7.14, "learning_rate": 5.996083075890047e-07, "loss": 0.0009, "step": 38936 }, { "epoch": 7.14, "learning_rate": 5.993551154833555e-07, "loss": 0.0007, "step": 38937 }, { "epoch": 7.14, "learning_rate": 5.991019751942829e-07, "loss": 0.0005, "step": 38938 }, { "epoch": 7.14, "learning_rate": 5.988488867231768e-07, "loss": 0.0003, "step": 38939 }, { "epoch": 7.14, "learning_rate": 5.985958500714384e-07, "loss": 0.0006, "step": 38940 }, { "epoch": 7.14, "learning_rate": 5.983428652404588e-07, "loss": 0.0003, "step": 38941 }, { "epoch": 7.14, "learning_rate": 5.980899322316347e-07, "loss": 0.0004, "step": 38942 }, { "epoch": 7.14, "learning_rate": 5.978370510463583e-07, "loss": 0.0007, "step": 38943 }, { "epoch": 7.14, "learning_rate": 5.975842216860239e-07, "loss": 0.0002, "step": 38944 }, { "epoch": 7.14, "learning_rate": 5.973314441520261e-07, "loss": 0.0004, "step": 38945 }, { "epoch": 7.14, "learning_rate": 5.970787184457582e-07, "loss": 0.0008, "step": 38946 }, { "epoch": 7.14, "learning_rate": 5.968260445686125e-07, "loss": 0.0004, "step": 38947 }, { "epoch": 7.14, "learning_rate": 5.965734225219799e-07, "loss": 0.001, "step": 38948 }, { "epoch": 7.14, "learning_rate": 5.963208523072583e-07, "loss": 0.0004, "step": 38949 }, { "epoch": 7.14, "learning_rate": 5.960683339258355e-07, "loss": 0.0003, "step": 38950 }, { "epoch": 7.14, "learning_rate": 5.958158673791037e-07, "loss": 0.0002, "step": 38951 }, { "epoch": 7.14, "learning_rate": 5.95563452668455e-07, "loss": 0.0006, "step": 38952 }, { "epoch": 7.14, "learning_rate": 5.953110897952796e-07, "loss": 0.0006, "step": 38953 }, { "epoch": 7.14, "learning_rate": 5.950587787609707e-07, "loss": 0.0007, "step": 38954 }, { "epoch": 7.14, "learning_rate": 5.948065195669195e-07, "loss": 0.0002, "step": 38955 }, { "epoch": 7.14, "learning_rate": 5.945543122145159e-07, "loss": 0.0007, "step": 38956 }, { "epoch": 7.14, "learning_rate": 5.943021567051455e-07, "loss": 0.0006, "step": 38957 }, { "epoch": 7.14, "learning_rate": 5.940500530402049e-07, "loss": 0.0004, "step": 38958 }, { "epoch": 7.14, "learning_rate": 5.937980012210798e-07, "loss": 0.0008, "step": 38959 }, { "epoch": 7.14, "learning_rate": 5.935460012491612e-07, "loss": 0.0004, "step": 38960 }, { "epoch": 7.14, "learning_rate": 5.932940531258368e-07, "loss": 0.0006, "step": 38961 }, { "epoch": 7.15, "learning_rate": 5.930421568524958e-07, "loss": 0.0003, "step": 38962 }, { "epoch": 7.15, "learning_rate": 5.927903124305268e-07, "loss": 0.0004, "step": 38963 }, { "epoch": 7.15, "learning_rate": 5.925385198613187e-07, "loss": 0.0003, "step": 38964 }, { "epoch": 7.15, "learning_rate": 5.922867791462583e-07, "loss": 0.0007, "step": 38965 }, { "epoch": 7.15, "learning_rate": 5.920350902867322e-07, "loss": 0.0003, "step": 38966 }, { "epoch": 7.15, "learning_rate": 5.917834532841316e-07, "loss": 0.0002, "step": 38967 }, { "epoch": 7.15, "learning_rate": 5.915318681398397e-07, "loss": 0.0009, "step": 38968 }, { "epoch": 7.15, "learning_rate": 5.912803348552443e-07, "loss": 0.0005, "step": 38969 }, { "epoch": 7.15, "learning_rate": 5.91028853431731e-07, "loss": 0.0003, "step": 38970 }, { "epoch": 7.15, "learning_rate": 5.907774238706887e-07, "loss": 0.001, "step": 38971 }, { "epoch": 7.15, "learning_rate": 5.905260461735018e-07, "loss": 0.0007, "step": 38972 }, { "epoch": 7.15, "learning_rate": 5.90274720341556e-07, "loss": 0.0005, "step": 38973 }, { "epoch": 7.15, "learning_rate": 5.900234463762367e-07, "loss": 0.0004, "step": 38974 }, { "epoch": 7.15, "learning_rate": 5.897722242789272e-07, "loss": 0.0005, "step": 38975 }, { "epoch": 7.15, "learning_rate": 5.895210540510143e-07, "loss": 0.0005, "step": 38976 }, { "epoch": 7.15, "learning_rate": 5.892699356938814e-07, "loss": 0.0001, "step": 38977 }, { "epoch": 7.15, "learning_rate": 5.890188692089138e-07, "loss": 0.001, "step": 38978 }, { "epoch": 7.15, "learning_rate": 5.88767854597494e-07, "loss": 0.0002, "step": 38979 }, { "epoch": 7.15, "learning_rate": 5.885168918610063e-07, "loss": 0.0008, "step": 38980 }, { "epoch": 7.15, "learning_rate": 5.882659810008351e-07, "loss": 0.0002, "step": 38981 }, { "epoch": 7.15, "learning_rate": 5.880151220183616e-07, "loss": 0.0007, "step": 38982 }, { "epoch": 7.15, "learning_rate": 5.877643149149669e-07, "loss": 0.0005, "step": 38983 }, { "epoch": 7.15, "learning_rate": 5.87513559692039e-07, "loss": 0.0006, "step": 38984 }, { "epoch": 7.15, "learning_rate": 5.872628563509575e-07, "loss": 0.0003, "step": 38985 }, { "epoch": 7.15, "learning_rate": 5.870122048931015e-07, "loss": 0.0003, "step": 38986 }, { "epoch": 7.15, "learning_rate": 5.867616053198555e-07, "loss": 0.0005, "step": 38987 }, { "epoch": 7.15, "learning_rate": 5.865110576325983e-07, "loss": 0.0004, "step": 38988 }, { "epoch": 7.15, "learning_rate": 5.862605618327144e-07, "loss": 0.0003, "step": 38989 }, { "epoch": 7.15, "learning_rate": 5.860101179215827e-07, "loss": 0.0006, "step": 38990 }, { "epoch": 7.15, "learning_rate": 5.857597259005854e-07, "loss": 0.0005, "step": 38991 }, { "epoch": 7.15, "learning_rate": 5.85509385771098e-07, "loss": 0.0026, "step": 38992 }, { "epoch": 7.15, "learning_rate": 5.852590975345052e-07, "loss": 0.0005, "step": 38993 }, { "epoch": 7.15, "learning_rate": 5.850088611921844e-07, "loss": 0.0004, "step": 38994 }, { "epoch": 7.15, "learning_rate": 5.847586767455149e-07, "loss": 0.0005, "step": 38995 }, { "epoch": 7.15, "learning_rate": 5.845085441958764e-07, "loss": 0.0005, "step": 38996 }, { "epoch": 7.15, "learning_rate": 5.84258463544648e-07, "loss": 0.0007, "step": 38997 }, { "epoch": 7.15, "learning_rate": 5.840084347932073e-07, "loss": 0.0005, "step": 38998 }, { "epoch": 7.15, "learning_rate": 5.837584579429323e-07, "loss": 0.0004, "step": 38999 }, { "epoch": 7.15, "learning_rate": 5.835085329952017e-07, "loss": 0.0003, "step": 39000 }, { "epoch": 7.15, "learning_rate": 5.832586599513912e-07, "loss": 0.0006, "step": 39001 }, { "epoch": 7.15, "learning_rate": 5.830088388128807e-07, "loss": 0.0003, "step": 39002 }, { "epoch": 7.15, "learning_rate": 5.827590695810481e-07, "loss": 0.0007, "step": 39003 }, { "epoch": 7.15, "learning_rate": 5.825093522572666e-07, "loss": 0.0004, "step": 39004 }, { "epoch": 7.15, "learning_rate": 5.82259686842912e-07, "loss": 0.0007, "step": 39005 }, { "epoch": 7.15, "learning_rate": 5.820100733393641e-07, "loss": 0.0005, "step": 39006 }, { "epoch": 7.15, "learning_rate": 5.817605117479985e-07, "loss": 0.0003, "step": 39007 }, { "epoch": 7.15, "learning_rate": 5.815110020701886e-07, "loss": 0.0006, "step": 39008 }, { "epoch": 7.15, "learning_rate": 5.812615443073122e-07, "loss": 0.0002, "step": 39009 }, { "epoch": 7.15, "learning_rate": 5.810121384607403e-07, "loss": 0.0002, "step": 39010 }, { "epoch": 7.15, "learning_rate": 5.80762784531852e-07, "loss": 0.0003, "step": 39011 }, { "epoch": 7.15, "learning_rate": 5.805134825220193e-07, "loss": 0.0004, "step": 39012 }, { "epoch": 7.15, "learning_rate": 5.802642324326168e-07, "loss": 0.0003, "step": 39013 }, { "epoch": 7.15, "learning_rate": 5.80015034265019e-07, "loss": 0.0007, "step": 39014 }, { "epoch": 7.15, "learning_rate": 5.797658880205981e-07, "loss": 0.0005, "step": 39015 }, { "epoch": 7.15, "learning_rate": 5.795167937007296e-07, "loss": 0.0003, "step": 39016 }, { "epoch": 7.16, "learning_rate": 5.792677513067846e-07, "loss": 0.0004, "step": 39017 }, { "epoch": 7.16, "learning_rate": 5.790187608401366e-07, "loss": 0.0007, "step": 39018 }, { "epoch": 7.16, "learning_rate": 5.787698223021565e-07, "loss": 0.0007, "step": 39019 }, { "epoch": 7.16, "learning_rate": 5.78520935694219e-07, "loss": 0.0003, "step": 39020 }, { "epoch": 7.16, "learning_rate": 5.782721010176962e-07, "loss": 0.0006, "step": 39021 }, { "epoch": 7.16, "learning_rate": 5.780233182739569e-07, "loss": 0.0005, "step": 39022 }, { "epoch": 7.16, "learning_rate": 5.777745874643725e-07, "loss": 0.0008, "step": 39023 }, { "epoch": 7.16, "learning_rate": 5.775259085903173e-07, "loss": 0.0006, "step": 39024 }, { "epoch": 7.16, "learning_rate": 5.772772816531591e-07, "loss": 0.0003, "step": 39025 }, { "epoch": 7.16, "learning_rate": 5.770287066542702e-07, "loss": 0.0005, "step": 39026 }, { "epoch": 7.16, "learning_rate": 5.767801835950205e-07, "loss": 0.0003, "step": 39027 }, { "epoch": 7.16, "learning_rate": 5.765317124767766e-07, "loss": 0.0006, "step": 39028 }, { "epoch": 7.16, "learning_rate": 5.762832933009121e-07, "loss": 0.0005, "step": 39029 }, { "epoch": 7.16, "learning_rate": 5.760349260687948e-07, "loss": 0.0008, "step": 39030 }, { "epoch": 7.16, "learning_rate": 5.757866107817933e-07, "loss": 0.0006, "step": 39031 }, { "epoch": 7.16, "learning_rate": 5.755383474412779e-07, "loss": 0.0002, "step": 39032 }, { "epoch": 7.16, "learning_rate": 5.752901360486152e-07, "loss": 0.0003, "step": 39033 }, { "epoch": 7.16, "learning_rate": 5.75041976605174e-07, "loss": 0.0004, "step": 39034 }, { "epoch": 7.16, "learning_rate": 5.747938691123223e-07, "loss": 0.0002, "step": 39035 }, { "epoch": 7.16, "learning_rate": 5.745458135714266e-07, "loss": 0.0003, "step": 39036 }, { "epoch": 7.16, "learning_rate": 5.742978099838559e-07, "loss": 0.0004, "step": 39037 }, { "epoch": 7.16, "learning_rate": 5.740498583509791e-07, "loss": 0.0002, "step": 39038 }, { "epoch": 7.16, "learning_rate": 5.738019586741573e-07, "loss": 0.0008, "step": 39039 }, { "epoch": 7.16, "learning_rate": 5.735541109547615e-07, "loss": 0.0003, "step": 39040 }, { "epoch": 7.16, "learning_rate": 5.733063151941531e-07, "loss": 0.0003, "step": 39041 }, { "epoch": 7.16, "learning_rate": 5.730585713937042e-07, "loss": 0.0003, "step": 39042 }, { "epoch": 7.16, "learning_rate": 5.728108795547771e-07, "loss": 0.0005, "step": 39043 }, { "epoch": 7.16, "learning_rate": 5.725632396787362e-07, "loss": 0.0009, "step": 39044 }, { "epoch": 7.16, "learning_rate": 5.723156517669481e-07, "loss": 0.0003, "step": 39045 }, { "epoch": 7.16, "learning_rate": 5.720681158207775e-07, "loss": 0.0002, "step": 39046 }, { "epoch": 7.16, "learning_rate": 5.718206318415875e-07, "loss": 0.0002, "step": 39047 }, { "epoch": 7.16, "learning_rate": 5.715731998307439e-07, "loss": 0.0007, "step": 39048 }, { "epoch": 7.16, "learning_rate": 5.713258197896099e-07, "loss": 0.0003, "step": 39049 }, { "epoch": 7.16, "learning_rate": 5.710784917195477e-07, "loss": 0.0008, "step": 39050 }, { "epoch": 7.16, "learning_rate": 5.70831215621922e-07, "loss": 0.0004, "step": 39051 }, { "epoch": 7.16, "learning_rate": 5.70583991498096e-07, "loss": 0.0005, "step": 39052 }, { "epoch": 7.16, "learning_rate": 5.703368193494318e-07, "loss": 0.0007, "step": 39053 }, { "epoch": 7.16, "learning_rate": 5.700896991772908e-07, "loss": 0.0004, "step": 39054 }, { "epoch": 7.16, "learning_rate": 5.698426309830374e-07, "loss": 0.0009, "step": 39055 }, { "epoch": 7.16, "learning_rate": 5.695956147680337e-07, "loss": 0.0003, "step": 39056 }, { "epoch": 7.16, "learning_rate": 5.693486505336388e-07, "loss": 0.0006, "step": 39057 }, { "epoch": 7.16, "learning_rate": 5.691017382812136e-07, "loss": 0.0003, "step": 39058 }, { "epoch": 7.16, "learning_rate": 5.688548780121217e-07, "loss": 0.0005, "step": 39059 }, { "epoch": 7.16, "learning_rate": 5.68608069727723e-07, "loss": 0.0004, "step": 39060 }, { "epoch": 7.16, "learning_rate": 5.683613134293775e-07, "loss": 0.0004, "step": 39061 }, { "epoch": 7.16, "learning_rate": 5.681146091184453e-07, "loss": 0.0002, "step": 39062 }, { "epoch": 7.16, "learning_rate": 5.678679567962864e-07, "loss": 0.0005, "step": 39063 }, { "epoch": 7.16, "learning_rate": 5.676213564642597e-07, "loss": 0.0005, "step": 39064 }, { "epoch": 7.16, "learning_rate": 5.673748081237263e-07, "loss": 0.0003, "step": 39065 }, { "epoch": 7.16, "learning_rate": 5.671283117760428e-07, "loss": 0.0008, "step": 39066 }, { "epoch": 7.16, "learning_rate": 5.668818674225684e-07, "loss": 0.0003, "step": 39067 }, { "epoch": 7.16, "learning_rate": 5.666354750646641e-07, "loss": 0.0011, "step": 39068 }, { "epoch": 7.16, "learning_rate": 5.663891347036854e-07, "loss": 0.0004, "step": 39069 }, { "epoch": 7.16, "learning_rate": 5.661428463409901e-07, "loss": 0.0008, "step": 39070 }, { "epoch": 7.17, "learning_rate": 5.658966099779373e-07, "loss": 0.0004, "step": 39071 }, { "epoch": 7.17, "learning_rate": 5.656504256158812e-07, "loss": 0.0003, "step": 39072 }, { "epoch": 7.17, "learning_rate": 5.654042932561831e-07, "loss": 0.0008, "step": 39073 }, { "epoch": 7.17, "learning_rate": 5.651582129001987e-07, "loss": 0.0003, "step": 39074 }, { "epoch": 7.17, "learning_rate": 5.649121845492822e-07, "loss": 0.0004, "step": 39075 }, { "epoch": 7.17, "learning_rate": 5.646662082047894e-07, "loss": 0.0005, "step": 39076 }, { "epoch": 7.17, "learning_rate": 5.644202838680779e-07, "loss": 0.0004, "step": 39077 }, { "epoch": 7.17, "learning_rate": 5.641744115405034e-07, "loss": 0.0002, "step": 39078 }, { "epoch": 7.17, "learning_rate": 5.639285912234205e-07, "loss": 0.0002, "step": 39079 }, { "epoch": 7.17, "learning_rate": 5.636828229181845e-07, "loss": 0.0004, "step": 39080 }, { "epoch": 7.17, "learning_rate": 5.634371066261502e-07, "loss": 0.0012, "step": 39081 }, { "epoch": 7.17, "learning_rate": 5.631914423486707e-07, "loss": 0.0007, "step": 39082 }, { "epoch": 7.17, "learning_rate": 5.629458300871016e-07, "loss": 0.0004, "step": 39083 }, { "epoch": 7.17, "learning_rate": 5.627002698427964e-07, "loss": 0.0005, "step": 39084 }, { "epoch": 7.17, "learning_rate": 5.624547616171083e-07, "loss": 0.0007, "step": 39085 }, { "epoch": 7.17, "learning_rate": 5.622093054113908e-07, "loss": 0.0003, "step": 39086 }, { "epoch": 7.17, "learning_rate": 5.619639012269961e-07, "loss": 0.0009, "step": 39087 }, { "epoch": 7.17, "learning_rate": 5.617185490652777e-07, "loss": 0.0006, "step": 39088 }, { "epoch": 7.17, "learning_rate": 5.614732489275865e-07, "loss": 0.0006, "step": 39089 }, { "epoch": 7.17, "learning_rate": 5.612280008152771e-07, "loss": 0.0004, "step": 39090 }, { "epoch": 7.17, "learning_rate": 5.609828047297006e-07, "loss": 0.0004, "step": 39091 }, { "epoch": 7.17, "learning_rate": 5.607376606722093e-07, "loss": 0.0003, "step": 39092 }, { "epoch": 7.17, "learning_rate": 5.604925686441509e-07, "loss": 0.0004, "step": 39093 }, { "epoch": 7.17, "learning_rate": 5.602475286468778e-07, "loss": 0.0004, "step": 39094 }, { "epoch": 7.17, "learning_rate": 5.600025406817433e-07, "loss": 0.0004, "step": 39095 }, { "epoch": 7.17, "learning_rate": 5.597576047500952e-07, "loss": 0.0004, "step": 39096 }, { "epoch": 7.17, "learning_rate": 5.595127208532847e-07, "loss": 0.0005, "step": 39097 }, { "epoch": 7.17, "learning_rate": 5.592678889926617e-07, "loss": 0.0004, "step": 39098 }, { "epoch": 7.17, "learning_rate": 5.590231091695741e-07, "loss": 0.0006, "step": 39099 }, { "epoch": 7.17, "learning_rate": 5.587783813853731e-07, "loss": 0.0002, "step": 39100 }, { "epoch": 7.17, "learning_rate": 5.585337056414064e-07, "loss": 0.0009, "step": 39101 }, { "epoch": 7.17, "learning_rate": 5.582890819390241e-07, "loss": 0.0003, "step": 39102 }, { "epoch": 7.17, "learning_rate": 5.580445102795728e-07, "loss": 0.0005, "step": 39103 }, { "epoch": 7.17, "learning_rate": 5.577999906644016e-07, "loss": 0.0004, "step": 39104 }, { "epoch": 7.17, "learning_rate": 5.575555230948581e-07, "loss": 0.0005, "step": 39105 }, { "epoch": 7.17, "learning_rate": 5.573111075722892e-07, "loss": 0.0005, "step": 39106 }, { "epoch": 7.17, "learning_rate": 5.570667440980415e-07, "loss": 0.0003, "step": 39107 }, { "epoch": 7.17, "learning_rate": 5.568224326734651e-07, "loss": 0.0005, "step": 39108 }, { "epoch": 7.17, "learning_rate": 5.565781732999043e-07, "loss": 0.0007, "step": 39109 }, { "epoch": 7.17, "learning_rate": 5.56333965978707e-07, "loss": 0.0003, "step": 39110 }, { "epoch": 7.17, "learning_rate": 5.560898107112156e-07, "loss": 0.0005, "step": 39111 }, { "epoch": 7.17, "learning_rate": 5.558457074987799e-07, "loss": 0.0018, "step": 39112 }, { "epoch": 7.17, "learning_rate": 5.556016563427435e-07, "loss": 0.0004, "step": 39113 }, { "epoch": 7.17, "learning_rate": 5.55357657244453e-07, "loss": 0.0005, "step": 39114 }, { "epoch": 7.17, "learning_rate": 5.551137102052517e-07, "loss": 0.0006, "step": 39115 }, { "epoch": 7.17, "learning_rate": 5.548698152264842e-07, "loss": 0.0003, "step": 39116 }, { "epoch": 7.17, "learning_rate": 5.546259723094971e-07, "loss": 0.0007, "step": 39117 }, { "epoch": 7.17, "learning_rate": 5.543821814556316e-07, "loss": 0.0003, "step": 39118 }, { "epoch": 7.17, "learning_rate": 5.541384426662333e-07, "loss": 0.0005, "step": 39119 }, { "epoch": 7.17, "learning_rate": 5.538947559426433e-07, "loss": 0.0005, "step": 39120 }, { "epoch": 7.17, "learning_rate": 5.536511212862106e-07, "loss": 0.0004, "step": 39121 }, { "epoch": 7.17, "learning_rate": 5.534075386982718e-07, "loss": 0.0004, "step": 39122 }, { "epoch": 7.17, "learning_rate": 5.531640081801725e-07, "loss": 0.0005, "step": 39123 }, { "epoch": 7.17, "learning_rate": 5.52920529733254e-07, "loss": 0.0003, "step": 39124 }, { "epoch": 7.17, "learning_rate": 5.526771033588585e-07, "loss": 0.0003, "step": 39125 }, { "epoch": 7.18, "learning_rate": 5.524337290583281e-07, "loss": 0.0004, "step": 39126 }, { "epoch": 7.18, "learning_rate": 5.521904068330052e-07, "loss": 0.0006, "step": 39127 }, { "epoch": 7.18, "learning_rate": 5.519471366842322e-07, "loss": 0.0004, "step": 39128 }, { "epoch": 7.18, "learning_rate": 5.517039186133433e-07, "loss": 0.0006, "step": 39129 }, { "epoch": 7.18, "learning_rate": 5.514607526216864e-07, "loss": 0.0003, "step": 39130 }, { "epoch": 7.18, "learning_rate": 5.512176387105994e-07, "loss": 0.0005, "step": 39131 }, { "epoch": 7.18, "learning_rate": 5.509745768814212e-07, "loss": 0.0008, "step": 39132 }, { "epoch": 7.18, "learning_rate": 5.507315671354929e-07, "loss": 0.0008, "step": 39133 }, { "epoch": 7.18, "learning_rate": 5.504886094741546e-07, "loss": 0.0002, "step": 39134 }, { "epoch": 7.18, "learning_rate": 5.50245703898743e-07, "loss": 0.0005, "step": 39135 }, { "epoch": 7.18, "learning_rate": 5.500028504105992e-07, "loss": 0.0004, "step": 39136 }, { "epoch": 7.18, "learning_rate": 5.49760049011061e-07, "loss": 0.0004, "step": 39137 }, { "epoch": 7.18, "learning_rate": 5.495172997014653e-07, "loss": 0.0004, "step": 39138 }, { "epoch": 7.18, "learning_rate": 5.492746024831541e-07, "loss": 0.0002, "step": 39139 }, { "epoch": 7.18, "learning_rate": 5.49031957357462e-07, "loss": 0.0004, "step": 39140 }, { "epoch": 7.18, "learning_rate": 5.487893643257269e-07, "loss": 0.0002, "step": 39141 }, { "epoch": 7.18, "learning_rate": 5.485468233892843e-07, "loss": 0.0007, "step": 39142 }, { "epoch": 7.18, "learning_rate": 5.483043345494742e-07, "loss": 0.0007, "step": 39143 }, { "epoch": 7.18, "learning_rate": 5.480618978076335e-07, "loss": 0.0006, "step": 39144 }, { "epoch": 7.18, "learning_rate": 5.478195131650977e-07, "loss": 0.0005, "step": 39145 }, { "epoch": 7.18, "learning_rate": 5.475771806232e-07, "loss": 0.0003, "step": 39146 }, { "epoch": 7.18, "learning_rate": 5.473349001832783e-07, "loss": 0.0005, "step": 39147 }, { "epoch": 7.18, "learning_rate": 5.470926718466685e-07, "loss": 0.0002, "step": 39148 }, { "epoch": 7.18, "learning_rate": 5.468504956147058e-07, "loss": 0.0005, "step": 39149 }, { "epoch": 7.18, "learning_rate": 5.46608371488725e-07, "loss": 0.0009, "step": 39150 }, { "epoch": 7.18, "learning_rate": 5.463662994700591e-07, "loss": 0.0017, "step": 39151 }, { "epoch": 7.18, "learning_rate": 5.461242795600452e-07, "loss": 0.0005, "step": 39152 }, { "epoch": 7.18, "learning_rate": 5.458823117600143e-07, "loss": 0.0003, "step": 39153 }, { "epoch": 7.18, "learning_rate": 5.45640396071302e-07, "loss": 0.0002, "step": 39154 }, { "epoch": 7.18, "learning_rate": 5.453985324952415e-07, "loss": 0.0004, "step": 39155 }, { "epoch": 7.18, "learning_rate": 5.451567210331643e-07, "loss": 0.0008, "step": 39156 }, { "epoch": 7.18, "learning_rate": 5.44914961686408e-07, "loss": 0.0002, "step": 39157 }, { "epoch": 7.18, "learning_rate": 5.446732544563005e-07, "loss": 0.0005, "step": 39158 }, { "epoch": 7.18, "learning_rate": 5.444315993441751e-07, "loss": 0.0003, "step": 39159 }, { "epoch": 7.18, "learning_rate": 5.441899963513631e-07, "loss": 0.0003, "step": 39160 }, { "epoch": 7.18, "learning_rate": 5.439484454791988e-07, "loss": 0.0003, "step": 39161 }, { "epoch": 7.18, "learning_rate": 5.437069467290123e-07, "loss": 0.0003, "step": 39162 }, { "epoch": 7.18, "learning_rate": 5.43465500102136e-07, "loss": 0.0009, "step": 39163 }, { "epoch": 7.18, "learning_rate": 5.432241055998954e-07, "loss": 0.0004, "step": 39164 }, { "epoch": 7.18, "learning_rate": 5.429827632236284e-07, "loss": 0.0003, "step": 39165 }, { "epoch": 7.18, "learning_rate": 5.427414729746605e-07, "loss": 0.0005, "step": 39166 }, { "epoch": 7.18, "learning_rate": 5.42500234854324e-07, "loss": 0.0003, "step": 39167 }, { "epoch": 7.18, "learning_rate": 5.422590488639479e-07, "loss": 0.0003, "step": 39168 }, { "epoch": 7.18, "learning_rate": 5.420179150048611e-07, "loss": 0.0007, "step": 39169 }, { "epoch": 7.18, "learning_rate": 5.417768332783924e-07, "loss": 0.001, "step": 39170 }, { "epoch": 7.18, "learning_rate": 5.41535803685872e-07, "loss": 0.0006, "step": 39171 }, { "epoch": 7.18, "learning_rate": 5.412948262286288e-07, "loss": 0.0004, "step": 39172 }, { "epoch": 7.18, "learning_rate": 5.410539009079874e-07, "loss": 0.0003, "step": 39173 }, { "epoch": 7.18, "learning_rate": 5.408130277252799e-07, "loss": 0.0007, "step": 39174 }, { "epoch": 7.18, "learning_rate": 5.405722066818353e-07, "loss": 0.0004, "step": 39175 }, { "epoch": 7.18, "learning_rate": 5.403314377789759e-07, "loss": 0.0004, "step": 39176 }, { "epoch": 7.18, "learning_rate": 5.400907210180318e-07, "loss": 0.0007, "step": 39177 }, { "epoch": 7.18, "learning_rate": 5.398500564003273e-07, "loss": 0.0002, "step": 39178 }, { "epoch": 7.18, "learning_rate": 5.396094439271926e-07, "loss": 0.0004, "step": 39179 }, { "epoch": 7.19, "learning_rate": 5.393688835999511e-07, "loss": 0.001, "step": 39180 }, { "epoch": 7.19, "learning_rate": 5.391283754199327e-07, "loss": 0.0006, "step": 39181 }, { "epoch": 7.19, "learning_rate": 5.388879193884577e-07, "loss": 0.0003, "step": 39182 }, { "epoch": 7.19, "learning_rate": 5.386475155068549e-07, "loss": 0.0005, "step": 39183 }, { "epoch": 7.19, "learning_rate": 5.384071637764488e-07, "loss": 0.0004, "step": 39184 }, { "epoch": 7.19, "learning_rate": 5.381668641985638e-07, "loss": 0.0005, "step": 39185 }, { "epoch": 7.19, "learning_rate": 5.379266167745256e-07, "loss": 0.0005, "step": 39186 }, { "epoch": 7.19, "learning_rate": 5.376864215056566e-07, "loss": 0.0008, "step": 39187 }, { "epoch": 7.19, "learning_rate": 5.374462783932821e-07, "loss": 0.0006, "step": 39188 }, { "epoch": 7.19, "learning_rate": 5.372061874387258e-07, "loss": 0.0007, "step": 39189 }, { "epoch": 7.19, "learning_rate": 5.369661486433098e-07, "loss": 0.0006, "step": 39190 }, { "epoch": 7.19, "learning_rate": 5.367261620083575e-07, "loss": 0.0004, "step": 39191 }, { "epoch": 7.19, "learning_rate": 5.364862275351946e-07, "loss": 0.0004, "step": 39192 }, { "epoch": 7.19, "learning_rate": 5.362463452251399e-07, "loss": 0.0005, "step": 39193 }, { "epoch": 7.19, "learning_rate": 5.360065150795179e-07, "loss": 0.0004, "step": 39194 }, { "epoch": 7.19, "learning_rate": 5.357667370996478e-07, "loss": 0.0007, "step": 39195 }, { "epoch": 7.19, "learning_rate": 5.355270112868538e-07, "loss": 0.001, "step": 39196 }, { "epoch": 7.19, "learning_rate": 5.352873376424572e-07, "loss": 0.0002, "step": 39197 }, { "epoch": 7.19, "learning_rate": 5.350477161677792e-07, "loss": 0.0003, "step": 39198 }, { "epoch": 7.19, "learning_rate": 5.348081468641408e-07, "loss": 0.0005, "step": 39199 }, { "epoch": 7.19, "learning_rate": 5.345686297328578e-07, "loss": 0.0005, "step": 39200 }, { "epoch": 7.19, "learning_rate": 5.343291647752569e-07, "loss": 0.0005, "step": 39201 }, { "epoch": 7.19, "learning_rate": 5.340897519926558e-07, "loss": 0.0002, "step": 39202 }, { "epoch": 7.19, "learning_rate": 5.338503913863724e-07, "loss": 0.0004, "step": 39203 }, { "epoch": 7.19, "learning_rate": 5.336110829577279e-07, "loss": 0.0011, "step": 39204 }, { "epoch": 7.19, "learning_rate": 5.333718267080412e-07, "loss": 0.0005, "step": 39205 }, { "epoch": 7.19, "learning_rate": 5.331326226386313e-07, "loss": 0.0003, "step": 39206 }, { "epoch": 7.19, "learning_rate": 5.328934707508149e-07, "loss": 0.001, "step": 39207 }, { "epoch": 7.19, "learning_rate": 5.32654371045912e-07, "loss": 0.0002, "step": 39208 }, { "epoch": 7.19, "learning_rate": 5.324153235252394e-07, "loss": 0.0004, "step": 39209 }, { "epoch": 7.19, "learning_rate": 5.321763281901171e-07, "loss": 0.0003, "step": 39210 }, { "epoch": 7.19, "learning_rate": 5.319373850418597e-07, "loss": 0.0006, "step": 39211 }, { "epoch": 7.19, "learning_rate": 5.316984940817849e-07, "loss": 0.0006, "step": 39212 }, { "epoch": 7.19, "learning_rate": 5.314596553112094e-07, "loss": 0.0004, "step": 39213 }, { "epoch": 7.19, "learning_rate": 5.312208687314502e-07, "loss": 0.0004, "step": 39214 }, { "epoch": 7.19, "learning_rate": 5.309821343438238e-07, "loss": 0.0004, "step": 39215 }, { "epoch": 7.19, "learning_rate": 5.307434521496458e-07, "loss": 0.0006, "step": 39216 }, { "epoch": 7.19, "learning_rate": 5.305048221502318e-07, "loss": 0.0003, "step": 39217 }, { "epoch": 7.19, "learning_rate": 5.302662443468976e-07, "loss": 0.0004, "step": 39218 }, { "epoch": 7.19, "learning_rate": 5.300277187409575e-07, "loss": 0.0004, "step": 39219 }, { "epoch": 7.19, "learning_rate": 5.297892453337262e-07, "loss": 0.0008, "step": 39220 }, { "epoch": 7.19, "learning_rate": 5.295508241265179e-07, "loss": 0.0007, "step": 39221 }, { "epoch": 7.19, "learning_rate": 5.293124551206485e-07, "loss": 0.0007, "step": 39222 }, { "epoch": 7.19, "learning_rate": 5.290741383174303e-07, "loss": 0.0003, "step": 39223 }, { "epoch": 7.19, "learning_rate": 5.288358737181765e-07, "loss": 0.0007, "step": 39224 }, { "epoch": 7.19, "learning_rate": 5.285976613242016e-07, "loss": 0.0012, "step": 39225 }, { "epoch": 7.19, "learning_rate": 5.28359501136817e-07, "loss": 0.0006, "step": 39226 }, { "epoch": 7.19, "learning_rate": 5.281213931573392e-07, "loss": 0.0009, "step": 39227 }, { "epoch": 7.19, "learning_rate": 5.278833373870785e-07, "loss": 0.0005, "step": 39228 }, { "epoch": 7.19, "learning_rate": 5.276453338273457e-07, "loss": 0.0005, "step": 39229 }, { "epoch": 7.19, "learning_rate": 5.274073824794535e-07, "loss": 0.0006, "step": 39230 }, { "epoch": 7.19, "learning_rate": 5.271694833447117e-07, "loss": 0.0004, "step": 39231 }, { "epoch": 7.19, "learning_rate": 5.269316364244359e-07, "loss": 0.0001, "step": 39232 }, { "epoch": 7.19, "learning_rate": 5.266938417199352e-07, "loss": 0.0006, "step": 39233 }, { "epoch": 7.19, "learning_rate": 5.264560992325185e-07, "loss": 0.0004, "step": 39234 }, { "epoch": 7.2, "learning_rate": 5.262184089634991e-07, "loss": 0.0003, "step": 39235 }, { "epoch": 7.2, "learning_rate": 5.259807709141851e-07, "loss": 0.0004, "step": 39236 }, { "epoch": 7.2, "learning_rate": 5.257431850858874e-07, "loss": 0.0004, "step": 39237 }, { "epoch": 7.2, "learning_rate": 5.25505651479915e-07, "loss": 0.0004, "step": 39238 }, { "epoch": 7.2, "learning_rate": 5.252681700975781e-07, "loss": 0.0005, "step": 39239 }, { "epoch": 7.2, "learning_rate": 5.250307409401845e-07, "loss": 0.0002, "step": 39240 }, { "epoch": 7.2, "learning_rate": 5.24793364009043e-07, "loss": 0.0006, "step": 39241 }, { "epoch": 7.2, "learning_rate": 5.245560393054628e-07, "loss": 0.0004, "step": 39242 }, { "epoch": 7.2, "learning_rate": 5.243187668307515e-07, "loss": 0.0013, "step": 39243 }, { "epoch": 7.2, "learning_rate": 5.240815465862159e-07, "loss": 0.0008, "step": 39244 }, { "epoch": 7.2, "learning_rate": 5.238443785731661e-07, "loss": 0.0003, "step": 39245 }, { "epoch": 7.2, "learning_rate": 5.236072627929101e-07, "loss": 0.0004, "step": 39246 }, { "epoch": 7.2, "learning_rate": 5.23370199246751e-07, "loss": 0.0003, "step": 39247 }, { "epoch": 7.2, "learning_rate": 5.231331879359958e-07, "loss": 0.0005, "step": 39248 }, { "epoch": 7.2, "learning_rate": 5.228962288619544e-07, "loss": 0.0006, "step": 39249 }, { "epoch": 7.2, "learning_rate": 5.226593220259301e-07, "loss": 0.0011, "step": 39250 }, { "epoch": 7.2, "learning_rate": 5.224224674292311e-07, "loss": 0.0004, "step": 39251 }, { "epoch": 7.2, "learning_rate": 5.221856650731616e-07, "loss": 0.0003, "step": 39252 }, { "epoch": 7.2, "learning_rate": 5.219489149590251e-07, "loss": 0.0009, "step": 39253 }, { "epoch": 7.2, "learning_rate": 5.217122170881284e-07, "loss": 0.0006, "step": 39254 }, { "epoch": 7.2, "learning_rate": 5.21475571461777e-07, "loss": 0.0013, "step": 39255 }, { "epoch": 7.2, "learning_rate": 5.212389780812733e-07, "loss": 0.0006, "step": 39256 }, { "epoch": 7.2, "learning_rate": 5.210024369479227e-07, "loss": 0.0002, "step": 39257 }, { "epoch": 7.2, "learning_rate": 5.2076594806303e-07, "loss": 0.0005, "step": 39258 }, { "epoch": 7.2, "learning_rate": 5.205295114278963e-07, "loss": 0.0005, "step": 39259 }, { "epoch": 7.2, "learning_rate": 5.202931270438261e-07, "loss": 0.0004, "step": 39260 }, { "epoch": 7.2, "learning_rate": 5.200567949121204e-07, "loss": 0.0009, "step": 39261 }, { "epoch": 7.2, "learning_rate": 5.198205150340863e-07, "loss": 0.0002, "step": 39262 }, { "epoch": 7.2, "learning_rate": 5.195842874110224e-07, "loss": 0.0002, "step": 39263 }, { "epoch": 7.2, "learning_rate": 5.193481120442345e-07, "loss": 0.0005, "step": 39264 }, { "epoch": 7.2, "learning_rate": 5.191119889350205e-07, "loss": 0.0007, "step": 39265 }, { "epoch": 7.2, "learning_rate": 5.188759180846814e-07, "loss": 0.0012, "step": 39266 }, { "epoch": 7.2, "learning_rate": 5.186398994945219e-07, "loss": 0.0004, "step": 39267 }, { "epoch": 7.2, "learning_rate": 5.184039331658409e-07, "loss": 0.0004, "step": 39268 }, { "epoch": 7.2, "learning_rate": 5.181680190999394e-07, "loss": 0.0008, "step": 39269 }, { "epoch": 7.2, "learning_rate": 5.179321572981189e-07, "loss": 0.0003, "step": 39270 }, { "epoch": 7.2, "learning_rate": 5.176963477616781e-07, "loss": 0.0005, "step": 39271 }, { "epoch": 7.2, "learning_rate": 5.174605904919172e-07, "loss": 0.0011, "step": 39272 }, { "epoch": 7.2, "learning_rate": 5.172248854901363e-07, "loss": 0.0007, "step": 39273 }, { "epoch": 7.2, "learning_rate": 5.169892327576332e-07, "loss": 0.0009, "step": 39274 }, { "epoch": 7.2, "learning_rate": 5.167536322957079e-07, "loss": 0.0004, "step": 39275 }, { "epoch": 7.2, "learning_rate": 5.165180841056583e-07, "loss": 0.0002, "step": 39276 }, { "epoch": 7.2, "learning_rate": 5.162825881887834e-07, "loss": 0.0003, "step": 39277 }, { "epoch": 7.2, "learning_rate": 5.160471445463811e-07, "loss": 0.0002, "step": 39278 }, { "epoch": 7.2, "learning_rate": 5.15811753179748e-07, "loss": 0.0004, "step": 39279 }, { "epoch": 7.2, "learning_rate": 5.155764140901832e-07, "loss": 0.0008, "step": 39280 }, { "epoch": 7.2, "learning_rate": 5.153411272789843e-07, "loss": 0.0002, "step": 39281 }, { "epoch": 7.2, "learning_rate": 5.151058927474484e-07, "loss": 0.0002, "step": 39282 }, { "epoch": 7.2, "learning_rate": 5.148707104968697e-07, "loss": 0.0007, "step": 39283 }, { "epoch": 7.2, "learning_rate": 5.146355805285452e-07, "loss": 0.0008, "step": 39284 }, { "epoch": 7.2, "learning_rate": 5.144005028437715e-07, "loss": 0.0004, "step": 39285 }, { "epoch": 7.2, "learning_rate": 5.141654774438454e-07, "loss": 0.0001, "step": 39286 }, { "epoch": 7.2, "learning_rate": 5.139305043300613e-07, "loss": 0.0004, "step": 39287 }, { "epoch": 7.2, "learning_rate": 5.13695583503715e-07, "loss": 0.0002, "step": 39288 }, { "epoch": 7.21, "learning_rate": 5.134607149660998e-07, "loss": 0.0009, "step": 39289 }, { "epoch": 7.21, "learning_rate": 5.132258987185124e-07, "loss": 0.0004, "step": 39290 }, { "epoch": 7.21, "learning_rate": 5.129911347622452e-07, "loss": 0.0004, "step": 39291 }, { "epoch": 7.21, "learning_rate": 5.127564230985926e-07, "loss": 0.0004, "step": 39292 }, { "epoch": 7.21, "learning_rate": 5.125217637288504e-07, "loss": 0.0006, "step": 39293 }, { "epoch": 7.21, "learning_rate": 5.122871566543097e-07, "loss": 0.0002, "step": 39294 }, { "epoch": 7.21, "learning_rate": 5.120526018762651e-07, "loss": 0.0004, "step": 39295 }, { "epoch": 7.21, "learning_rate": 5.118180993960087e-07, "loss": 0.0004, "step": 39296 }, { "epoch": 7.21, "learning_rate": 5.115836492148307e-07, "loss": 0.0003, "step": 39297 }, { "epoch": 7.21, "learning_rate": 5.11349251334029e-07, "loss": 0.0009, "step": 39298 }, { "epoch": 7.21, "learning_rate": 5.111149057548925e-07, "loss": 0.0004, "step": 39299 }, { "epoch": 7.21, "learning_rate": 5.108806124787125e-07, "loss": 0.0003, "step": 39300 }, { "epoch": 7.21, "learning_rate": 5.106463715067789e-07, "loss": 0.0007, "step": 39301 }, { "epoch": 7.21, "learning_rate": 5.104121828403863e-07, "loss": 0.0005, "step": 39302 }, { "epoch": 7.21, "learning_rate": 5.101780464808237e-07, "loss": 0.0006, "step": 39303 }, { "epoch": 7.21, "learning_rate": 5.099439624293833e-07, "loss": 0.0005, "step": 39304 }, { "epoch": 7.21, "learning_rate": 5.097099306873531e-07, "loss": 0.0008, "step": 39305 }, { "epoch": 7.21, "learning_rate": 5.094759512560243e-07, "loss": 0.0005, "step": 39306 }, { "epoch": 7.21, "learning_rate": 5.092420241366858e-07, "loss": 0.0004, "step": 39307 }, { "epoch": 7.21, "learning_rate": 5.090081493306287e-07, "loss": 0.0004, "step": 39308 }, { "epoch": 7.21, "learning_rate": 5.08774326839141e-07, "loss": 0.0007, "step": 39309 }, { "epoch": 7.21, "learning_rate": 5.085405566635093e-07, "loss": 0.0006, "step": 39310 }, { "epoch": 7.21, "learning_rate": 5.083068388050272e-07, "loss": 0.0005, "step": 39311 }, { "epoch": 7.21, "learning_rate": 5.080731732649791e-07, "loss": 0.0004, "step": 39312 }, { "epoch": 7.21, "learning_rate": 5.07839560044654e-07, "loss": 0.0008, "step": 39313 }, { "epoch": 7.21, "learning_rate": 5.076059991453385e-07, "loss": 0.0006, "step": 39314 }, { "epoch": 7.21, "learning_rate": 5.073724905683231e-07, "loss": 0.0006, "step": 39315 }, { "epoch": 7.21, "learning_rate": 5.071390343148918e-07, "loss": 0.0003, "step": 39316 }, { "epoch": 7.21, "learning_rate": 5.06905630386334e-07, "loss": 0.0002, "step": 39317 }, { "epoch": 7.21, "learning_rate": 5.066722787839339e-07, "loss": 0.0004, "step": 39318 }, { "epoch": 7.21, "learning_rate": 5.064389795089764e-07, "loss": 0.0003, "step": 39319 }, { "epoch": 7.21, "learning_rate": 5.062057325627523e-07, "loss": 0.0003, "step": 39320 }, { "epoch": 7.21, "learning_rate": 5.059725379465441e-07, "loss": 0.0002, "step": 39321 }, { "epoch": 7.21, "learning_rate": 5.057393956616374e-07, "loss": 0.0004, "step": 39322 }, { "epoch": 7.21, "learning_rate": 5.05506305709318e-07, "loss": 0.0005, "step": 39323 }, { "epoch": 7.21, "learning_rate": 5.052732680908689e-07, "loss": 0.0006, "step": 39324 }, { "epoch": 7.21, "learning_rate": 5.050402828075774e-07, "loss": 0.0005, "step": 39325 }, { "epoch": 7.21, "learning_rate": 5.048073498607264e-07, "loss": 0.0003, "step": 39326 }, { "epoch": 7.21, "learning_rate": 5.045744692515997e-07, "loss": 0.0002, "step": 39327 }, { "epoch": 7.21, "learning_rate": 5.043416409814783e-07, "loss": 0.0006, "step": 39328 }, { "epoch": 7.21, "learning_rate": 5.041088650516523e-07, "loss": 0.0003, "step": 39329 }, { "epoch": 7.21, "learning_rate": 5.038761414633986e-07, "loss": 0.0004, "step": 39330 }, { "epoch": 7.21, "learning_rate": 5.036434702180026e-07, "loss": 0.0012, "step": 39331 }, { "epoch": 7.21, "learning_rate": 5.034108513167446e-07, "loss": 0.0004, "step": 39332 }, { "epoch": 7.21, "learning_rate": 5.031782847609101e-07, "loss": 0.0006, "step": 39333 }, { "epoch": 7.21, "learning_rate": 5.029457705517793e-07, "loss": 0.0007, "step": 39334 }, { "epoch": 7.21, "learning_rate": 5.027133086906355e-07, "loss": 0.0003, "step": 39335 }, { "epoch": 7.21, "learning_rate": 5.024808991787567e-07, "loss": 0.0005, "step": 39336 }, { "epoch": 7.21, "learning_rate": 5.022485420174239e-07, "loss": 0.0003, "step": 39337 }, { "epoch": 7.21, "learning_rate": 5.020162372079218e-07, "loss": 0.0004, "step": 39338 }, { "epoch": 7.21, "learning_rate": 5.017839847515282e-07, "loss": 0.0006, "step": 39339 }, { "epoch": 7.21, "learning_rate": 5.015517846495243e-07, "loss": 0.0008, "step": 39340 }, { "epoch": 7.21, "learning_rate": 5.013196369031892e-07, "loss": 0.0003, "step": 39341 }, { "epoch": 7.21, "learning_rate": 5.010875415138027e-07, "loss": 0.0002, "step": 39342 }, { "epoch": 7.21, "learning_rate": 5.008554984826441e-07, "loss": 0.0004, "step": 39343 }, { "epoch": 7.22, "learning_rate": 5.006235078109933e-07, "loss": 0.0001, "step": 39344 }, { "epoch": 7.22, "learning_rate": 5.00391569500126e-07, "loss": 0.0002, "step": 39345 }, { "epoch": 7.22, "learning_rate": 5.001596835513256e-07, "loss": 0.0005, "step": 39346 }, { "epoch": 7.22, "learning_rate": 4.999278499658667e-07, "loss": 0.0011, "step": 39347 }, { "epoch": 7.22, "learning_rate": 4.99696068745028e-07, "loss": 0.0007, "step": 39348 }, { "epoch": 7.22, "learning_rate": 4.994643398900878e-07, "loss": 0.0003, "step": 39349 }, { "epoch": 7.22, "learning_rate": 4.992326634023204e-07, "loss": 0.001, "step": 39350 }, { "epoch": 7.22, "learning_rate": 4.990010392830069e-07, "loss": 0.0005, "step": 39351 }, { "epoch": 7.22, "learning_rate": 4.987694675334221e-07, "loss": 0.0003, "step": 39352 }, { "epoch": 7.22, "learning_rate": 4.985379481548435e-07, "loss": 0.0013, "step": 39353 }, { "epoch": 7.22, "learning_rate": 4.983064811485438e-07, "loss": 0.0002, "step": 39354 }, { "epoch": 7.22, "learning_rate": 4.98075066515803e-07, "loss": 0.0004, "step": 39355 }, { "epoch": 7.22, "learning_rate": 4.978437042578954e-07, "loss": 0.0005, "step": 39356 }, { "epoch": 7.22, "learning_rate": 4.976123943760957e-07, "loss": 0.0009, "step": 39357 }, { "epoch": 7.22, "learning_rate": 4.973811368716785e-07, "loss": 0.0005, "step": 39358 }, { "epoch": 7.22, "learning_rate": 4.971499317459205e-07, "loss": 0.0018, "step": 39359 }, { "epoch": 7.22, "learning_rate": 4.969187790000929e-07, "loss": 0.0004, "step": 39360 }, { "epoch": 7.22, "learning_rate": 4.966876786354724e-07, "loss": 0.0004, "step": 39361 }, { "epoch": 7.22, "learning_rate": 4.964566306533325e-07, "loss": 0.0005, "step": 39362 }, { "epoch": 7.22, "learning_rate": 4.962256350549443e-07, "loss": 0.0008, "step": 39363 }, { "epoch": 7.22, "learning_rate": 4.95994691841587e-07, "loss": 0.0003, "step": 39364 }, { "epoch": 7.22, "learning_rate": 4.957638010145271e-07, "loss": 0.0012, "step": 39365 }, { "epoch": 7.22, "learning_rate": 4.955329625750393e-07, "loss": 0.0002, "step": 39366 }, { "epoch": 7.22, "learning_rate": 4.95302176524396e-07, "loss": 0.0003, "step": 39367 }, { "epoch": 7.22, "learning_rate": 4.950714428638714e-07, "loss": 0.0004, "step": 39368 }, { "epoch": 7.22, "learning_rate": 4.948407615947348e-07, "loss": 0.0003, "step": 39369 }, { "epoch": 7.22, "learning_rate": 4.946101327182595e-07, "loss": 0.0003, "step": 39370 }, { "epoch": 7.22, "learning_rate": 4.943795562357167e-07, "loss": 0.0007, "step": 39371 }, { "epoch": 7.22, "learning_rate": 4.941490321483733e-07, "loss": 0.0003, "step": 39372 }, { "epoch": 7.22, "learning_rate": 4.939185604575048e-07, "loss": 0.0004, "step": 39373 }, { "epoch": 7.22, "learning_rate": 4.936881411643801e-07, "loss": 0.0007, "step": 39374 }, { "epoch": 7.22, "learning_rate": 4.934577742702695e-07, "loss": 0.0003, "step": 39375 }, { "epoch": 7.22, "learning_rate": 4.932274597764419e-07, "loss": 0.0005, "step": 39376 }, { "epoch": 7.22, "learning_rate": 4.929971976841674e-07, "loss": 0.0005, "step": 39377 }, { "epoch": 7.22, "learning_rate": 4.92766987994715e-07, "loss": 0.0002, "step": 39378 }, { "epoch": 7.22, "learning_rate": 4.925368307093537e-07, "loss": 0.0004, "step": 39379 }, { "epoch": 7.22, "learning_rate": 4.923067258293524e-07, "loss": 0.0004, "step": 39380 }, { "epoch": 7.22, "learning_rate": 4.920766733559767e-07, "loss": 0.0004, "step": 39381 }, { "epoch": 7.22, "learning_rate": 4.918466732905014e-07, "loss": 0.0005, "step": 39382 }, { "epoch": 7.22, "learning_rate": 4.916167256341875e-07, "loss": 0.0005, "step": 39383 }, { "epoch": 7.22, "learning_rate": 4.913868303883052e-07, "loss": 0.0003, "step": 39384 }, { "epoch": 7.22, "learning_rate": 4.9115698755412e-07, "loss": 0.0004, "step": 39385 }, { "epoch": 7.22, "learning_rate": 4.909271971329021e-07, "loss": 0.0006, "step": 39386 }, { "epoch": 7.22, "learning_rate": 4.906974591259161e-07, "loss": 0.0005, "step": 39387 }, { "epoch": 7.22, "learning_rate": 4.904677735344277e-07, "loss": 0.0003, "step": 39388 }, { "epoch": 7.22, "learning_rate": 4.902381403597046e-07, "loss": 0.0003, "step": 39389 }, { "epoch": 7.22, "learning_rate": 4.900085596030113e-07, "loss": 0.0004, "step": 39390 }, { "epoch": 7.22, "learning_rate": 4.897790312656137e-07, "loss": 0.0003, "step": 39391 }, { "epoch": 7.22, "learning_rate": 4.895495553487773e-07, "loss": 0.0004, "step": 39392 }, { "epoch": 7.22, "learning_rate": 4.893201318537666e-07, "loss": 0.0006, "step": 39393 }, { "epoch": 7.22, "learning_rate": 4.890907607818451e-07, "loss": 0.0005, "step": 39394 }, { "epoch": 7.22, "learning_rate": 4.888614421342797e-07, "loss": 0.0004, "step": 39395 }, { "epoch": 7.22, "learning_rate": 4.886321759123314e-07, "loss": 0.0004, "step": 39396 }, { "epoch": 7.22, "learning_rate": 4.884029621172659e-07, "loss": 0.0007, "step": 39397 }, { "epoch": 7.23, "learning_rate": 4.881738007503445e-07, "loss": 0.0006, "step": 39398 }, { "epoch": 7.23, "learning_rate": 4.879446918128339e-07, "loss": 0.0004, "step": 39399 }, { "epoch": 7.23, "learning_rate": 4.877156353059964e-07, "loss": 0.0005, "step": 39400 }, { "epoch": 7.23, "learning_rate": 4.874866312310911e-07, "loss": 0.0004, "step": 39401 }, { "epoch": 7.23, "learning_rate": 4.872576795893824e-07, "loss": 0.0005, "step": 39402 }, { "epoch": 7.23, "learning_rate": 4.870287803821317e-07, "loss": 0.0007, "step": 39403 }, { "epoch": 7.23, "learning_rate": 4.867999336106021e-07, "loss": 0.0011, "step": 39404 }, { "epoch": 7.23, "learning_rate": 4.865711392760541e-07, "loss": 0.0006, "step": 39405 }, { "epoch": 7.23, "learning_rate": 4.863423973797498e-07, "loss": 0.0002, "step": 39406 }, { "epoch": 7.23, "learning_rate": 4.861137079229472e-07, "loss": 0.0004, "step": 39407 }, { "epoch": 7.23, "learning_rate": 4.858850709069096e-07, "loss": 0.0002, "step": 39408 }, { "epoch": 7.23, "learning_rate": 4.856564863328961e-07, "loss": 0.0008, "step": 39409 }, { "epoch": 7.23, "learning_rate": 4.854279542021678e-07, "loss": 0.0003, "step": 39410 }, { "epoch": 7.23, "learning_rate": 4.851994745159827e-07, "loss": 0.0002, "step": 39411 }, { "epoch": 7.23, "learning_rate": 4.849710472756008e-07, "loss": 0.0003, "step": 39412 }, { "epoch": 7.23, "learning_rate": 4.847426724822823e-07, "loss": 0.0003, "step": 39413 }, { "epoch": 7.23, "learning_rate": 4.845143501372851e-07, "loss": 0.0002, "step": 39414 }, { "epoch": 7.23, "learning_rate": 4.84286080241867e-07, "loss": 0.0002, "step": 39415 }, { "epoch": 7.23, "learning_rate": 4.840578627972858e-07, "loss": 0.0005, "step": 39416 }, { "epoch": 7.23, "learning_rate": 4.838296978048019e-07, "loss": 0.0006, "step": 39417 }, { "epoch": 7.23, "learning_rate": 4.836015852656739e-07, "loss": 0.0003, "step": 39418 }, { "epoch": 7.23, "learning_rate": 4.833735251811544e-07, "loss": 0.0004, "step": 39419 }, { "epoch": 7.23, "learning_rate": 4.831455175525024e-07, "loss": 0.0006, "step": 39420 }, { "epoch": 7.23, "learning_rate": 4.829175623809757e-07, "loss": 0.0003, "step": 39421 }, { "epoch": 7.23, "learning_rate": 4.82689659667831e-07, "loss": 0.0017, "step": 39422 }, { "epoch": 7.23, "learning_rate": 4.82461809414324e-07, "loss": 0.0009, "step": 39423 }, { "epoch": 7.23, "learning_rate": 4.822340116217116e-07, "loss": 0.0006, "step": 39424 }, { "epoch": 7.23, "learning_rate": 4.820062662912461e-07, "loss": 0.0003, "step": 39425 }, { "epoch": 7.23, "learning_rate": 4.817785734241853e-07, "loss": 0.0006, "step": 39426 }, { "epoch": 7.23, "learning_rate": 4.815509330217849e-07, "loss": 0.0009, "step": 39427 }, { "epoch": 7.23, "learning_rate": 4.813233450852984e-07, "loss": 0.0003, "step": 39428 }, { "epoch": 7.23, "learning_rate": 4.810958096159812e-07, "loss": 0.0003, "step": 39429 }, { "epoch": 7.23, "learning_rate": 4.80868326615086e-07, "loss": 0.0005, "step": 39430 }, { "epoch": 7.23, "learning_rate": 4.806408960838682e-07, "loss": 0.0008, "step": 39431 }, { "epoch": 7.23, "learning_rate": 4.804135180235802e-07, "loss": 0.0002, "step": 39432 }, { "epoch": 7.23, "learning_rate": 4.801861924354768e-07, "loss": 0.0004, "step": 39433 }, { "epoch": 7.23, "learning_rate": 4.799589193208076e-07, "loss": 0.0006, "step": 39434 }, { "epoch": 7.23, "learning_rate": 4.797316986808298e-07, "loss": 0.0005, "step": 39435 }, { "epoch": 7.23, "learning_rate": 4.795045305167956e-07, "loss": 0.0007, "step": 39436 }, { "epoch": 7.23, "learning_rate": 4.792774148299539e-07, "loss": 0.0015, "step": 39437 }, { "epoch": 7.23, "learning_rate": 4.790503516215572e-07, "loss": 0.0002, "step": 39438 }, { "epoch": 7.23, "learning_rate": 4.788233408928588e-07, "loss": 0.0004, "step": 39439 }, { "epoch": 7.23, "learning_rate": 4.78596382645109e-07, "loss": 0.0004, "step": 39440 }, { "epoch": 7.23, "learning_rate": 4.783694768795599e-07, "loss": 0.0003, "step": 39441 }, { "epoch": 7.23, "learning_rate": 4.781426235974607e-07, "loss": 0.0004, "step": 39442 }, { "epoch": 7.23, "learning_rate": 4.779158228000624e-07, "loss": 0.0005, "step": 39443 }, { "epoch": 7.23, "learning_rate": 4.776890744886154e-07, "loss": 0.0005, "step": 39444 }, { "epoch": 7.23, "learning_rate": 4.774623786643684e-07, "loss": 0.0002, "step": 39445 }, { "epoch": 7.23, "learning_rate": 4.772357353285729e-07, "loss": 0.0002, "step": 39446 }, { "epoch": 7.23, "learning_rate": 4.770091444824765e-07, "loss": 0.0004, "step": 39447 }, { "epoch": 7.23, "learning_rate": 4.7678260612732953e-07, "loss": 0.0006, "step": 39448 }, { "epoch": 7.23, "learning_rate": 4.765561202643798e-07, "loss": 0.0004, "step": 39449 }, { "epoch": 7.23, "learning_rate": 4.763296868948752e-07, "loss": 0.0005, "step": 39450 }, { "epoch": 7.23, "learning_rate": 4.7610330602006373e-07, "loss": 0.0006, "step": 39451 }, { "epoch": 7.23, "learning_rate": 4.7587697764119535e-07, "loss": 0.0006, "step": 39452 }, { "epoch": 7.24, "learning_rate": 4.7565070175951687e-07, "loss": 0.0002, "step": 39453 }, { "epoch": 7.24, "learning_rate": 4.7542447837627405e-07, "loss": 0.0005, "step": 39454 }, { "epoch": 7.24, "learning_rate": 4.751983074927147e-07, "loss": 0.0013, "step": 39455 }, { "epoch": 7.24, "learning_rate": 4.7497218911008334e-07, "loss": 0.0004, "step": 39456 }, { "epoch": 7.24, "learning_rate": 4.747461232296302e-07, "loss": 0.0004, "step": 39457 }, { "epoch": 7.24, "learning_rate": 4.7452010985259977e-07, "loss": 0.0008, "step": 39458 }, { "epoch": 7.24, "learning_rate": 4.742941489802377e-07, "loss": 0.0004, "step": 39459 }, { "epoch": 7.24, "learning_rate": 4.7406824061378975e-07, "loss": 0.0002, "step": 39460 }, { "epoch": 7.24, "learning_rate": 4.738423847545004e-07, "loss": 0.0019, "step": 39461 }, { "epoch": 7.24, "learning_rate": 4.7361658140361534e-07, "loss": 0.0006, "step": 39462 }, { "epoch": 7.24, "learning_rate": 4.73390830562378e-07, "loss": 0.0002, "step": 39463 }, { "epoch": 7.24, "learning_rate": 4.731651322320341e-07, "loss": 0.0002, "step": 39464 }, { "epoch": 7.24, "learning_rate": 4.7293948641382704e-07, "loss": 0.0006, "step": 39465 }, { "epoch": 7.24, "learning_rate": 4.727138931090014e-07, "loss": 0.0005, "step": 39466 }, { "epoch": 7.24, "learning_rate": 4.724883523187995e-07, "loss": 0.0006, "step": 39467 }, { "epoch": 7.24, "learning_rate": 4.7226286404446487e-07, "loss": 0.0002, "step": 39468 }, { "epoch": 7.24, "learning_rate": 4.7203742828723864e-07, "loss": 0.0006, "step": 39469 }, { "epoch": 7.24, "learning_rate": 4.7181204504836764e-07, "loss": 0.0003, "step": 39470 }, { "epoch": 7.24, "learning_rate": 4.715867143290931e-07, "loss": 0.0005, "step": 39471 }, { "epoch": 7.24, "learning_rate": 4.71361436130654e-07, "loss": 0.0003, "step": 39472 }, { "epoch": 7.24, "learning_rate": 4.711362104542927e-07, "loss": 0.0003, "step": 39473 }, { "epoch": 7.24, "learning_rate": 4.7091103730125376e-07, "loss": 0.0002, "step": 39474 }, { "epoch": 7.24, "learning_rate": 4.706859166727762e-07, "loss": 0.0005, "step": 39475 }, { "epoch": 7.24, "learning_rate": 4.7046084857010013e-07, "loss": 0.0005, "step": 39476 }, { "epoch": 7.24, "learning_rate": 4.7023583299446785e-07, "loss": 0.0003, "step": 39477 }, { "epoch": 7.24, "learning_rate": 4.700108699471195e-07, "loss": 0.0005, "step": 39478 }, { "epoch": 7.24, "learning_rate": 4.6978595942929417e-07, "loss": 0.0006, "step": 39479 }, { "epoch": 7.24, "learning_rate": 4.6956110144223077e-07, "loss": 0.0003, "step": 39480 }, { "epoch": 7.24, "learning_rate": 4.693362959871717e-07, "loss": 0.0007, "step": 39481 }, { "epoch": 7.24, "learning_rate": 4.6911154306535143e-07, "loss": 0.0013, "step": 39482 }, { "epoch": 7.24, "learning_rate": 4.6888684267801466e-07, "loss": 0.0002, "step": 39483 }, { "epoch": 7.24, "learning_rate": 4.6866219482639587e-07, "loss": 0.001, "step": 39484 }, { "epoch": 7.24, "learning_rate": 4.6843759951173406e-07, "loss": 0.0004, "step": 39485 }, { "epoch": 7.24, "learning_rate": 4.682130567352683e-07, "loss": 0.0009, "step": 39486 }, { "epoch": 7.24, "learning_rate": 4.679885664982331e-07, "loss": 0.0009, "step": 39487 }, { "epoch": 7.24, "learning_rate": 4.677641288018697e-07, "loss": 0.0002, "step": 39488 }, { "epoch": 7.24, "learning_rate": 4.6753974364741606e-07, "loss": 0.0005, "step": 39489 }, { "epoch": 7.24, "learning_rate": 4.6731541103610447e-07, "loss": 0.0008, "step": 39490 }, { "epoch": 7.24, "learning_rate": 4.6709113096917283e-07, "loss": 0.0009, "step": 39491 }, { "epoch": 7.24, "learning_rate": 4.66866903447859e-07, "loss": 0.0003, "step": 39492 }, { "epoch": 7.24, "learning_rate": 4.6664272847339766e-07, "loss": 0.0006, "step": 39493 }, { "epoch": 7.24, "learning_rate": 4.664186060470266e-07, "loss": 0.0003, "step": 39494 }, { "epoch": 7.24, "learning_rate": 4.6619453616997824e-07, "loss": 0.0002, "step": 39495 }, { "epoch": 7.24, "learning_rate": 4.6597051884348934e-07, "loss": 0.0002, "step": 39496 }, { "epoch": 7.24, "learning_rate": 4.657465540687944e-07, "loss": 0.0005, "step": 39497 }, { "epoch": 7.24, "learning_rate": 4.65522641847127e-07, "loss": 0.0003, "step": 39498 }, { "epoch": 7.24, "learning_rate": 4.652987821797228e-07, "loss": 0.0003, "step": 39499 }, { "epoch": 7.24, "learning_rate": 4.6507497506781517e-07, "loss": 0.0002, "step": 39500 }, { "epoch": 7.24, "learning_rate": 4.6485122051263764e-07, "loss": 0.0005, "step": 39501 }, { "epoch": 7.24, "learning_rate": 4.6462751851542367e-07, "loss": 0.0007, "step": 39502 }, { "epoch": 7.24, "learning_rate": 4.6440386907740553e-07, "loss": 0.0002, "step": 39503 }, { "epoch": 7.24, "learning_rate": 4.6418027219981677e-07, "loss": 0.0004, "step": 39504 }, { "epoch": 7.24, "learning_rate": 4.639567278838897e-07, "loss": 0.0005, "step": 39505 }, { "epoch": 7.24, "learning_rate": 4.6373323613085666e-07, "loss": 0.0007, "step": 39506 }, { "epoch": 7.25, "learning_rate": 4.635097969419511e-07, "loss": 0.0005, "step": 39507 }, { "epoch": 7.25, "learning_rate": 4.6328641031840203e-07, "loss": 0.0002, "step": 39508 }, { "epoch": 7.25, "learning_rate": 4.6306307626143964e-07, "loss": 0.0007, "step": 39509 }, { "epoch": 7.25, "learning_rate": 4.6283979477229846e-07, "loss": 0.0003, "step": 39510 }, { "epoch": 7.25, "learning_rate": 4.626165658522086e-07, "loss": 0.0007, "step": 39511 }, { "epoch": 7.25, "learning_rate": 4.623933895023991e-07, "loss": 0.0005, "step": 39512 }, { "epoch": 7.25, "learning_rate": 4.6217026572410006e-07, "loss": 0.0003, "step": 39513 }, { "epoch": 7.25, "learning_rate": 4.6194719451854273e-07, "loss": 0.0005, "step": 39514 }, { "epoch": 7.25, "learning_rate": 4.617241758869562e-07, "loss": 0.0002, "step": 39515 }, { "epoch": 7.25, "learning_rate": 4.6150120983056826e-07, "loss": 0.0006, "step": 39516 }, { "epoch": 7.25, "learning_rate": 4.6127829635060904e-07, "loss": 0.0002, "step": 39517 }, { "epoch": 7.25, "learning_rate": 4.610554354483088e-07, "loss": 0.0012, "step": 39518 }, { "epoch": 7.25, "learning_rate": 4.608326271248942e-07, "loss": 0.0006, "step": 39519 }, { "epoch": 7.25, "learning_rate": 4.6060987138159316e-07, "loss": 0.0003, "step": 39520 }, { "epoch": 7.25, "learning_rate": 4.603871682196337e-07, "loss": 0.0005, "step": 39521 }, { "epoch": 7.25, "learning_rate": 4.601645176402425e-07, "loss": 0.001, "step": 39522 }, { "epoch": 7.25, "learning_rate": 4.599419196446497e-07, "loss": 0.0005, "step": 39523 }, { "epoch": 7.25, "learning_rate": 4.597193742340811e-07, "loss": 0.0005, "step": 39524 }, { "epoch": 7.25, "learning_rate": 4.594968814097633e-07, "loss": 0.0007, "step": 39525 }, { "epoch": 7.25, "learning_rate": 4.5927444117291884e-07, "loss": 0.0003, "step": 39526 }, { "epoch": 7.25, "learning_rate": 4.5905205352478e-07, "loss": 0.0002, "step": 39527 }, { "epoch": 7.25, "learning_rate": 4.588297184665691e-07, "loss": 0.0004, "step": 39528 }, { "epoch": 7.25, "learning_rate": 4.5860743599951186e-07, "loss": 0.0005, "step": 39529 }, { "epoch": 7.25, "learning_rate": 4.5838520612483396e-07, "loss": 0.0005, "step": 39530 }, { "epoch": 7.25, "learning_rate": 4.5816302884376106e-07, "loss": 0.0004, "step": 39531 }, { "epoch": 7.25, "learning_rate": 4.5794090415751666e-07, "loss": 0.0005, "step": 39532 }, { "epoch": 7.25, "learning_rate": 4.5771883206732535e-07, "loss": 0.0005, "step": 39533 }, { "epoch": 7.25, "learning_rate": 4.574968125744117e-07, "loss": 0.0003, "step": 39534 }, { "epoch": 7.25, "learning_rate": 4.57274845679998e-07, "loss": 0.0007, "step": 39535 }, { "epoch": 7.25, "learning_rate": 4.5705293138531113e-07, "loss": 0.0015, "step": 39536 }, { "epoch": 7.25, "learning_rate": 4.5683106969157123e-07, "loss": 0.0007, "step": 39537 }, { "epoch": 7.25, "learning_rate": 4.5660926060000164e-07, "loss": 0.0009, "step": 39538 }, { "epoch": 7.25, "learning_rate": 4.5638750411182595e-07, "loss": 0.0004, "step": 39539 }, { "epoch": 7.25, "learning_rate": 4.561658002282643e-07, "loss": 0.0004, "step": 39540 }, { "epoch": 7.25, "learning_rate": 4.5594414895054117e-07, "loss": 0.0011, "step": 39541 }, { "epoch": 7.25, "learning_rate": 4.557225502798779e-07, "loss": 0.0002, "step": 39542 }, { "epoch": 7.25, "learning_rate": 4.555010042174968e-07, "loss": 0.0007, "step": 39543 }, { "epoch": 7.25, "learning_rate": 4.5527951076461575e-07, "loss": 0.0007, "step": 39544 }, { "epoch": 7.25, "learning_rate": 4.5505806992245825e-07, "loss": 0.0004, "step": 39545 }, { "epoch": 7.25, "learning_rate": 4.548366816922445e-07, "loss": 0.0003, "step": 39546 }, { "epoch": 7.25, "learning_rate": 4.5461534607519454e-07, "loss": 0.0005, "step": 39547 }, { "epoch": 7.25, "learning_rate": 4.543940630725285e-07, "loss": 0.0003, "step": 39548 }, { "epoch": 7.25, "learning_rate": 4.5417283268546554e-07, "loss": 0.0004, "step": 39549 }, { "epoch": 7.25, "learning_rate": 4.5395165491522564e-07, "loss": 0.0004, "step": 39550 }, { "epoch": 7.25, "learning_rate": 4.53730529763029e-07, "loss": 0.0004, "step": 39551 }, { "epoch": 7.25, "learning_rate": 4.5350945723009245e-07, "loss": 0.0001, "step": 39552 }, { "epoch": 7.25, "learning_rate": 4.532884373176338e-07, "loss": 0.0005, "step": 39553 }, { "epoch": 7.25, "learning_rate": 4.5306747002687556e-07, "loss": 0.0007, "step": 39554 }, { "epoch": 7.25, "learning_rate": 4.528465553590322e-07, "loss": 0.0005, "step": 39555 }, { "epoch": 7.25, "learning_rate": 4.526256933153217e-07, "loss": 0.0002, "step": 39556 }, { "epoch": 7.25, "learning_rate": 4.524048838969619e-07, "loss": 0.0005, "step": 39557 }, { "epoch": 7.25, "learning_rate": 4.521841271051708e-07, "loss": 0.0005, "step": 39558 }, { "epoch": 7.25, "learning_rate": 4.5196342294116404e-07, "loss": 0.0005, "step": 39559 }, { "epoch": 7.25, "learning_rate": 4.5174277140615954e-07, "loss": 0.0004, "step": 39560 }, { "epoch": 7.25, "learning_rate": 4.51522172501373e-07, "loss": 0.0002, "step": 39561 }, { "epoch": 7.26, "learning_rate": 4.5130162622801676e-07, "loss": 0.0005, "step": 39562 }, { "epoch": 7.26, "learning_rate": 4.510811325873121e-07, "loss": 0.0004, "step": 39563 }, { "epoch": 7.26, "learning_rate": 4.5086069158047143e-07, "loss": 0.0003, "step": 39564 }, { "epoch": 7.26, "learning_rate": 4.5064030320871146e-07, "loss": 0.0004, "step": 39565 }, { "epoch": 7.26, "learning_rate": 4.504199674732446e-07, "loss": 0.0007, "step": 39566 }, { "epoch": 7.26, "learning_rate": 4.501996843752865e-07, "loss": 0.0004, "step": 39567 }, { "epoch": 7.26, "learning_rate": 4.499794539160518e-07, "loss": 0.0006, "step": 39568 }, { "epoch": 7.26, "learning_rate": 4.497592760967551e-07, "loss": 0.0009, "step": 39569 }, { "epoch": 7.26, "learning_rate": 4.4953915091860646e-07, "loss": 0.0002, "step": 39570 }, { "epoch": 7.26, "learning_rate": 4.4931907838282273e-07, "loss": 0.0004, "step": 39571 }, { "epoch": 7.26, "learning_rate": 4.490990584906185e-07, "loss": 0.0004, "step": 39572 }, { "epoch": 7.26, "learning_rate": 4.488790912432017e-07, "loss": 0.0012, "step": 39573 }, { "epoch": 7.26, "learning_rate": 4.4865917664178805e-07, "loss": 0.0004, "step": 39574 }, { "epoch": 7.26, "learning_rate": 4.484393146875865e-07, "loss": 0.0003, "step": 39575 }, { "epoch": 7.26, "learning_rate": 4.482195053818139e-07, "loss": 0.0002, "step": 39576 }, { "epoch": 7.26, "learning_rate": 4.479997487256782e-07, "loss": 0.0006, "step": 39577 }, { "epoch": 7.26, "learning_rate": 4.4778004472039284e-07, "loss": 0.0003, "step": 39578 }, { "epoch": 7.26, "learning_rate": 4.475603933671657e-07, "loss": 0.0008, "step": 39579 }, { "epoch": 7.26, "learning_rate": 4.4734079466721034e-07, "loss": 0.0004, "step": 39580 }, { "epoch": 7.26, "learning_rate": 4.4712124862173576e-07, "loss": 0.0004, "step": 39581 }, { "epoch": 7.26, "learning_rate": 4.469017552319532e-07, "loss": 0.0003, "step": 39582 }, { "epoch": 7.26, "learning_rate": 4.466823144990717e-07, "loss": 0.0005, "step": 39583 }, { "epoch": 7.26, "learning_rate": 4.464629264243014e-07, "loss": 0.0004, "step": 39584 }, { "epoch": 7.26, "learning_rate": 4.462435910088514e-07, "loss": 0.0008, "step": 39585 }, { "epoch": 7.26, "learning_rate": 4.460243082539306e-07, "loss": 0.0003, "step": 39586 }, { "epoch": 7.26, "learning_rate": 4.4580507816074703e-07, "loss": 0.0004, "step": 39587 }, { "epoch": 7.26, "learning_rate": 4.4558590073050965e-07, "loss": 0.0002, "step": 39588 }, { "epoch": 7.26, "learning_rate": 4.4536677596442644e-07, "loss": 0.0012, "step": 39589 }, { "epoch": 7.26, "learning_rate": 4.4514770386370755e-07, "loss": 0.0005, "step": 39590 }, { "epoch": 7.26, "learning_rate": 4.449286844295564e-07, "loss": 0.0004, "step": 39591 }, { "epoch": 7.26, "learning_rate": 4.4470971766318317e-07, "loss": 0.0006, "step": 39592 }, { "epoch": 7.26, "learning_rate": 4.444908035657924e-07, "loss": 0.0003, "step": 39593 }, { "epoch": 7.26, "learning_rate": 4.4427194213859216e-07, "loss": 0.0005, "step": 39594 }, { "epoch": 7.26, "learning_rate": 4.440531333827902e-07, "loss": 0.0005, "step": 39595 }, { "epoch": 7.26, "learning_rate": 4.4383437729959234e-07, "loss": 0.0003, "step": 39596 }, { "epoch": 7.26, "learning_rate": 4.436156738901998e-07, "loss": 0.0004, "step": 39597 }, { "epoch": 7.26, "learning_rate": 4.433970231558238e-07, "loss": 0.0006, "step": 39598 }, { "epoch": 7.26, "learning_rate": 4.4317842509766674e-07, "loss": 0.0009, "step": 39599 }, { "epoch": 7.26, "learning_rate": 4.429598797169343e-07, "loss": 0.0006, "step": 39600 }, { "epoch": 7.26, "learning_rate": 4.427413870148312e-07, "loss": 0.0006, "step": 39601 }, { "epoch": 7.26, "learning_rate": 4.425229469925618e-07, "loss": 0.0008, "step": 39602 }, { "epoch": 7.26, "learning_rate": 4.423045596513287e-07, "loss": 0.0006, "step": 39603 }, { "epoch": 7.26, "learning_rate": 4.4208622499233745e-07, "loss": 0.0002, "step": 39604 }, { "epoch": 7.26, "learning_rate": 4.418679430167905e-07, "loss": 0.0005, "step": 39605 }, { "epoch": 7.26, "learning_rate": 4.4164971372589014e-07, "loss": 0.0005, "step": 39606 }, { "epoch": 7.26, "learning_rate": 4.414315371208433e-07, "loss": 0.0007, "step": 39607 }, { "epoch": 7.26, "learning_rate": 4.4121341320284894e-07, "loss": 0.0009, "step": 39608 }, { "epoch": 7.26, "learning_rate": 4.4099534197310944e-07, "loss": 0.0004, "step": 39609 }, { "epoch": 7.26, "learning_rate": 4.407773234328261e-07, "loss": 0.0003, "step": 39610 }, { "epoch": 7.26, "learning_rate": 4.4055935758320343e-07, "loss": 0.0009, "step": 39611 }, { "epoch": 7.26, "learning_rate": 4.403414444254428e-07, "loss": 0.0006, "step": 39612 }, { "epoch": 7.26, "learning_rate": 4.401235839607432e-07, "loss": 0.0003, "step": 39613 }, { "epoch": 7.26, "learning_rate": 4.399057761903069e-07, "loss": 0.0005, "step": 39614 }, { "epoch": 7.26, "learning_rate": 4.3968802111533207e-07, "loss": 0.0002, "step": 39615 }, { "epoch": 7.26, "learning_rate": 4.3947031873702194e-07, "loss": 0.0003, "step": 39616 }, { "epoch": 7.27, "learning_rate": 4.392526690565757e-07, "loss": 0.0003, "step": 39617 }, { "epoch": 7.27, "learning_rate": 4.3903507207519234e-07, "loss": 0.0009, "step": 39618 }, { "epoch": 7.27, "learning_rate": 4.3881752779407204e-07, "loss": 0.0004, "step": 39619 }, { "epoch": 7.27, "learning_rate": 4.3860003621441384e-07, "loss": 0.0003, "step": 39620 }, { "epoch": 7.27, "learning_rate": 4.383825973374156e-07, "loss": 0.0003, "step": 39621 }, { "epoch": 7.27, "learning_rate": 4.381652111642765e-07, "loss": 0.0003, "step": 39622 }, { "epoch": 7.27, "learning_rate": 4.379478776961943e-07, "loss": 0.0005, "step": 39623 }, { "epoch": 7.27, "learning_rate": 4.3773059693436817e-07, "loss": 0.0002, "step": 39624 }, { "epoch": 7.27, "learning_rate": 4.3751336887999597e-07, "loss": 0.0004, "step": 39625 }, { "epoch": 7.27, "learning_rate": 4.3729619353427343e-07, "loss": 0.0005, "step": 39626 }, { "epoch": 7.27, "learning_rate": 4.370790708983974e-07, "loss": 0.0003, "step": 39627 }, { "epoch": 7.27, "learning_rate": 4.368620009735647e-07, "loss": 0.0003, "step": 39628 }, { "epoch": 7.27, "learning_rate": 4.366449837609743e-07, "loss": 0.0003, "step": 39629 }, { "epoch": 7.27, "learning_rate": 4.3642801926182086e-07, "loss": 0.0002, "step": 39630 }, { "epoch": 7.27, "learning_rate": 4.36211107477299e-07, "loss": 0.0003, "step": 39631 }, { "epoch": 7.27, "learning_rate": 4.359942484086066e-07, "loss": 0.0003, "step": 39632 }, { "epoch": 7.27, "learning_rate": 4.357774420569383e-07, "loss": 0.0005, "step": 39633 }, { "epoch": 7.27, "learning_rate": 4.3556068842348865e-07, "loss": 0.0013, "step": 39634 }, { "epoch": 7.27, "learning_rate": 4.353439875094523e-07, "loss": 0.0008, "step": 39635 }, { "epoch": 7.27, "learning_rate": 4.351273393160238e-07, "loss": 0.0005, "step": 39636 }, { "epoch": 7.27, "learning_rate": 4.349107438443978e-07, "loss": 0.0002, "step": 39637 }, { "epoch": 7.27, "learning_rate": 4.3469420109576774e-07, "loss": 0.0007, "step": 39638 }, { "epoch": 7.27, "learning_rate": 4.344777110713272e-07, "loss": 0.0003, "step": 39639 }, { "epoch": 7.27, "learning_rate": 4.3426127377226957e-07, "loss": 0.0004, "step": 39640 }, { "epoch": 7.27, "learning_rate": 4.3404488919978615e-07, "loss": 0.0003, "step": 39641 }, { "epoch": 7.27, "learning_rate": 4.338285573550738e-07, "loss": 0.0008, "step": 39642 }, { "epoch": 7.27, "learning_rate": 4.3361227823932374e-07, "loss": 0.0009, "step": 39643 }, { "epoch": 7.27, "learning_rate": 4.33396051853725e-07, "loss": 0.0006, "step": 39644 }, { "epoch": 7.27, "learning_rate": 4.3317987819947006e-07, "loss": 0.0005, "step": 39645 }, { "epoch": 7.27, "learning_rate": 4.3296375727775344e-07, "loss": 0.001, "step": 39646 }, { "epoch": 7.27, "learning_rate": 4.3274768908976415e-07, "loss": 0.0006, "step": 39647 }, { "epoch": 7.27, "learning_rate": 4.3253167363669466e-07, "loss": 0.001, "step": 39648 }, { "epoch": 7.27, "learning_rate": 4.32315710919734e-07, "loss": 0.0005, "step": 39649 }, { "epoch": 7.27, "learning_rate": 4.320998009400734e-07, "loss": 0.0004, "step": 39650 }, { "epoch": 7.27, "learning_rate": 4.3188394369890307e-07, "loss": 0.0004, "step": 39651 }, { "epoch": 7.27, "learning_rate": 4.31668139197412e-07, "loss": 0.0003, "step": 39652 }, { "epoch": 7.27, "learning_rate": 4.3145238743679154e-07, "loss": 0.0005, "step": 39653 }, { "epoch": 7.27, "learning_rate": 4.312366884182284e-07, "loss": 0.0006, "step": 39654 }, { "epoch": 7.27, "learning_rate": 4.3102104214291397e-07, "loss": 0.0003, "step": 39655 }, { "epoch": 7.27, "learning_rate": 4.3080544861203497e-07, "loss": 0.0003, "step": 39656 }, { "epoch": 7.27, "learning_rate": 4.305899078267806e-07, "loss": 0.0005, "step": 39657 }, { "epoch": 7.27, "learning_rate": 4.303744197883386e-07, "loss": 0.0003, "step": 39658 }, { "epoch": 7.27, "learning_rate": 4.30158984497896e-07, "loss": 0.0005, "step": 39659 }, { "epoch": 7.27, "learning_rate": 4.2994360195664275e-07, "loss": 0.0004, "step": 39660 }, { "epoch": 7.27, "learning_rate": 4.2972827216576583e-07, "loss": 0.0003, "step": 39661 }, { "epoch": 7.27, "learning_rate": 4.2951299512644873e-07, "loss": 0.0003, "step": 39662 }, { "epoch": 7.27, "learning_rate": 4.2929777083988047e-07, "loss": 0.0004, "step": 39663 }, { "epoch": 7.27, "learning_rate": 4.2908259930724673e-07, "loss": 0.0004, "step": 39664 }, { "epoch": 7.27, "learning_rate": 4.288674805297355e-07, "loss": 0.0006, "step": 39665 }, { "epoch": 7.27, "learning_rate": 4.2865241450853024e-07, "loss": 0.0006, "step": 39666 }, { "epoch": 7.27, "learning_rate": 4.284374012448167e-07, "loss": 0.0008, "step": 39667 }, { "epoch": 7.27, "learning_rate": 4.282224407397806e-07, "loss": 0.0008, "step": 39668 }, { "epoch": 7.27, "learning_rate": 4.280075329946076e-07, "loss": 0.0004, "step": 39669 }, { "epoch": 7.27, "learning_rate": 4.277926780104802e-07, "loss": 0.0004, "step": 39670 }, { "epoch": 7.28, "learning_rate": 4.27577875788584e-07, "loss": 0.0006, "step": 39671 }, { "epoch": 7.28, "learning_rate": 4.2736312633010255e-07, "loss": 0.0003, "step": 39672 }, { "epoch": 7.28, "learning_rate": 4.271484296362205e-07, "loss": 0.0002, "step": 39673 }, { "epoch": 7.28, "learning_rate": 4.2693378570812017e-07, "loss": 0.0009, "step": 39674 }, { "epoch": 7.28, "learning_rate": 4.267191945469862e-07, "loss": 0.0003, "step": 39675 }, { "epoch": 7.28, "learning_rate": 4.265046561539976e-07, "loss": 0.0003, "step": 39676 }, { "epoch": 7.28, "learning_rate": 4.2629017053034127e-07, "loss": 0.0005, "step": 39677 }, { "epoch": 7.28, "learning_rate": 4.2607573767719846e-07, "loss": 0.0002, "step": 39678 }, { "epoch": 7.28, "learning_rate": 4.2586135759575043e-07, "loss": 0.0008, "step": 39679 }, { "epoch": 7.28, "learning_rate": 4.2564703028717845e-07, "loss": 0.0006, "step": 39680 }, { "epoch": 7.28, "learning_rate": 4.2543275575266385e-07, "loss": 0.001, "step": 39681 }, { "epoch": 7.28, "learning_rate": 4.2521853399338787e-07, "loss": 0.0003, "step": 39682 }, { "epoch": 7.28, "learning_rate": 4.2500436501053287e-07, "loss": 0.0002, "step": 39683 }, { "epoch": 7.28, "learning_rate": 4.2479024880527796e-07, "loss": 0.0004, "step": 39684 }, { "epoch": 7.28, "learning_rate": 4.245761853788033e-07, "loss": 0.0007, "step": 39685 }, { "epoch": 7.28, "learning_rate": 4.243621747322879e-07, "loss": 0.0004, "step": 39686 }, { "epoch": 7.28, "learning_rate": 4.241482168669142e-07, "loss": 0.0006, "step": 39687 }, { "epoch": 7.28, "learning_rate": 4.2393431178385904e-07, "loss": 0.0003, "step": 39688 }, { "epoch": 7.28, "learning_rate": 4.2372045948430253e-07, "loss": 0.0009, "step": 39689 }, { "epoch": 7.28, "learning_rate": 4.2350665996942377e-07, "loss": 0.0002, "step": 39690 }, { "epoch": 7.28, "learning_rate": 4.2329291324039955e-07, "loss": 0.0003, "step": 39691 }, { "epoch": 7.28, "learning_rate": 4.230792192984101e-07, "loss": 0.0002, "step": 39692 }, { "epoch": 7.28, "learning_rate": 4.228655781446311e-07, "loss": 0.0002, "step": 39693 }, { "epoch": 7.28, "learning_rate": 4.226519897802417e-07, "loss": 0.0013, "step": 39694 }, { "epoch": 7.28, "learning_rate": 4.224384542064197e-07, "loss": 0.0005, "step": 39695 }, { "epoch": 7.28, "learning_rate": 4.222249714243409e-07, "loss": 0.0004, "step": 39696 }, { "epoch": 7.28, "learning_rate": 4.220115414351833e-07, "loss": 0.0003, "step": 39697 }, { "epoch": 7.28, "learning_rate": 4.2179816424012033e-07, "loss": 0.0002, "step": 39698 }, { "epoch": 7.28, "learning_rate": 4.2158483984033215e-07, "loss": 0.0006, "step": 39699 }, { "epoch": 7.28, "learning_rate": 4.2137156823699234e-07, "loss": 0.0003, "step": 39700 }, { "epoch": 7.28, "learning_rate": 4.2115834943127655e-07, "loss": 0.0004, "step": 39701 }, { "epoch": 7.28, "learning_rate": 4.209451834243605e-07, "loss": 0.0004, "step": 39702 }, { "epoch": 7.28, "learning_rate": 4.2073207021742004e-07, "loss": 0.0002, "step": 39703 }, { "epoch": 7.28, "learning_rate": 4.205190098116274e-07, "loss": 0.001, "step": 39704 }, { "epoch": 7.28, "learning_rate": 4.2030600220815844e-07, "loss": 0.0006, "step": 39705 }, { "epoch": 7.28, "learning_rate": 4.200930474081877e-07, "loss": 0.0005, "step": 39706 }, { "epoch": 7.28, "learning_rate": 4.1988014541288756e-07, "loss": 0.0003, "step": 39707 }, { "epoch": 7.28, "learning_rate": 4.1966729622343494e-07, "loss": 0.0004, "step": 39708 }, { "epoch": 7.28, "learning_rate": 4.194544998409988e-07, "loss": 0.0004, "step": 39709 }, { "epoch": 7.28, "learning_rate": 4.192417562667539e-07, "loss": 0.0005, "step": 39710 }, { "epoch": 7.28, "learning_rate": 4.190290655018736e-07, "loss": 0.0006, "step": 39711 }, { "epoch": 7.28, "learning_rate": 4.188164275475282e-07, "loss": 0.0003, "step": 39712 }, { "epoch": 7.28, "learning_rate": 4.1860384240489214e-07, "loss": 0.0007, "step": 39713 }, { "epoch": 7.28, "learning_rate": 4.1839131007513576e-07, "loss": 0.0002, "step": 39714 }, { "epoch": 7.28, "learning_rate": 4.1817883055943254e-07, "loss": 0.0005, "step": 39715 }, { "epoch": 7.28, "learning_rate": 4.179664038589493e-07, "loss": 0.0006, "step": 39716 }, { "epoch": 7.28, "learning_rate": 4.177540299748617e-07, "loss": 0.0005, "step": 39717 }, { "epoch": 7.28, "learning_rate": 4.1754170890833777e-07, "loss": 0.0002, "step": 39718 }, { "epoch": 7.28, "learning_rate": 4.173294406605488e-07, "loss": 0.0002, "step": 39719 }, { "epoch": 7.28, "learning_rate": 4.1711722523266387e-07, "loss": 0.0002, "step": 39720 }, { "epoch": 7.28, "learning_rate": 4.169050626258542e-07, "loss": 0.0002, "step": 39721 }, { "epoch": 7.28, "learning_rate": 4.1669295284128776e-07, "loss": 0.0003, "step": 39722 }, { "epoch": 7.28, "learning_rate": 4.1648089588013473e-07, "loss": 0.0009, "step": 39723 }, { "epoch": 7.28, "learning_rate": 4.162688917435631e-07, "loss": 0.0007, "step": 39724 }, { "epoch": 7.28, "learning_rate": 4.160569404327408e-07, "loss": 0.0004, "step": 39725 }, { "epoch": 7.29, "learning_rate": 4.158450419488391e-07, "loss": 0.0005, "step": 39726 }, { "epoch": 7.29, "learning_rate": 4.1563319629302377e-07, "loss": 0.0008, "step": 39727 }, { "epoch": 7.29, "learning_rate": 4.1542140346646163e-07, "loss": 0.0003, "step": 39728 }, { "epoch": 7.29, "learning_rate": 4.152096634703207e-07, "loss": 0.0007, "step": 39729 }, { "epoch": 7.29, "learning_rate": 4.1499797630576985e-07, "loss": 0.0002, "step": 39730 }, { "epoch": 7.29, "learning_rate": 4.1478634197397506e-07, "loss": 0.0008, "step": 39731 }, { "epoch": 7.29, "learning_rate": 4.14574760476103e-07, "loss": 0.0004, "step": 39732 }, { "epoch": 7.29, "learning_rate": 4.1436323181331837e-07, "loss": 0.0003, "step": 39733 }, { "epoch": 7.29, "learning_rate": 4.141517559867869e-07, "loss": 0.0004, "step": 39734 }, { "epoch": 7.29, "learning_rate": 4.1394033299767765e-07, "loss": 0.0002, "step": 39735 }, { "epoch": 7.29, "learning_rate": 4.1372896284715305e-07, "loss": 0.0004, "step": 39736 }, { "epoch": 7.29, "learning_rate": 4.1351764553637874e-07, "loss": 0.0004, "step": 39737 }, { "epoch": 7.29, "learning_rate": 4.1330638106652056e-07, "loss": 0.0006, "step": 39738 }, { "epoch": 7.29, "learning_rate": 4.1309516943874196e-07, "loss": 0.0002, "step": 39739 }, { "epoch": 7.29, "learning_rate": 4.1288401065420755e-07, "loss": 0.0003, "step": 39740 }, { "epoch": 7.29, "learning_rate": 4.1267290471407985e-07, "loss": 0.0004, "step": 39741 }, { "epoch": 7.29, "learning_rate": 4.124618516195256e-07, "loss": 0.0005, "step": 39742 }, { "epoch": 7.29, "learning_rate": 4.1225085137170386e-07, "loss": 0.0004, "step": 39743 }, { "epoch": 7.29, "learning_rate": 4.1203990397178263e-07, "loss": 0.0006, "step": 39744 }, { "epoch": 7.29, "learning_rate": 4.1182900942092096e-07, "loss": 0.0003, "step": 39745 }, { "epoch": 7.29, "learning_rate": 4.116181677202824e-07, "loss": 0.0002, "step": 39746 }, { "epoch": 7.29, "learning_rate": 4.114073788710271e-07, "loss": 0.0005, "step": 39747 }, { "epoch": 7.29, "learning_rate": 4.111966428743208e-07, "loss": 0.0005, "step": 39748 }, { "epoch": 7.29, "learning_rate": 4.109859597313237e-07, "loss": 0.0003, "step": 39749 }, { "epoch": 7.29, "learning_rate": 4.107753294431982e-07, "loss": 0.0005, "step": 39750 }, { "epoch": 7.29, "learning_rate": 4.105647520111e-07, "loss": 0.0005, "step": 39751 }, { "epoch": 7.29, "learning_rate": 4.1035422743619603e-07, "loss": 0.0007, "step": 39752 }, { "epoch": 7.29, "learning_rate": 4.1014375571964306e-07, "loss": 0.0004, "step": 39753 }, { "epoch": 7.29, "learning_rate": 4.099333368626024e-07, "loss": 0.001, "step": 39754 }, { "epoch": 7.29, "learning_rate": 4.0972297086623536e-07, "loss": 0.0008, "step": 39755 }, { "epoch": 7.29, "learning_rate": 4.095126577316988e-07, "loss": 0.0008, "step": 39756 }, { "epoch": 7.29, "learning_rate": 4.093023974601529e-07, "loss": 0.0003, "step": 39757 }, { "epoch": 7.29, "learning_rate": 4.090921900527578e-07, "loss": 0.0003, "step": 39758 }, { "epoch": 7.29, "learning_rate": 4.088820355106704e-07, "loss": 0.0006, "step": 39759 }, { "epoch": 7.29, "learning_rate": 4.0867193383504864e-07, "loss": 0.0005, "step": 39760 }, { "epoch": 7.29, "learning_rate": 4.0846188502705386e-07, "loss": 0.0009, "step": 39761 }, { "epoch": 7.29, "learning_rate": 4.0825188908784286e-07, "loss": 0.0002, "step": 39762 }, { "epoch": 7.29, "learning_rate": 4.0804194601857026e-07, "loss": 0.0002, "step": 39763 }, { "epoch": 7.29, "learning_rate": 4.078320558203963e-07, "loss": 0.0005, "step": 39764 }, { "epoch": 7.29, "learning_rate": 4.076222184944745e-07, "loss": 0.0003, "step": 39765 }, { "epoch": 7.29, "learning_rate": 4.0741243404196604e-07, "loss": 0.0003, "step": 39766 }, { "epoch": 7.29, "learning_rate": 4.072027024640235e-07, "loss": 0.0003, "step": 39767 }, { "epoch": 7.29, "learning_rate": 4.0699302376180695e-07, "loss": 0.0002, "step": 39768 }, { "epoch": 7.29, "learning_rate": 4.067833979364666e-07, "loss": 0.0008, "step": 39769 }, { "epoch": 7.29, "learning_rate": 4.0657382498916266e-07, "loss": 0.0003, "step": 39770 }, { "epoch": 7.29, "learning_rate": 4.063643049210475e-07, "loss": 0.0015, "step": 39771 }, { "epoch": 7.29, "learning_rate": 4.06154837733278e-07, "loss": 0.0006, "step": 39772 }, { "epoch": 7.29, "learning_rate": 4.059454234270077e-07, "loss": 0.0004, "step": 39773 }, { "epoch": 7.29, "learning_rate": 4.0573606200339013e-07, "loss": 0.0002, "step": 39774 }, { "epoch": 7.29, "learning_rate": 4.0552675346358095e-07, "loss": 0.0004, "step": 39775 }, { "epoch": 7.29, "learning_rate": 4.0531749780873266e-07, "loss": 0.0004, "step": 39776 }, { "epoch": 7.29, "learning_rate": 4.051082950399998e-07, "loss": 0.0004, "step": 39777 }, { "epoch": 7.29, "learning_rate": 4.048991451585327e-07, "loss": 0.0003, "step": 39778 }, { "epoch": 7.29, "learning_rate": 4.046900481654892e-07, "loss": 0.0004, "step": 39779 }, { "epoch": 7.3, "learning_rate": 4.044810040620184e-07, "loss": 0.0002, "step": 39780 }, { "epoch": 7.3, "learning_rate": 4.0427201284927165e-07, "loss": 0.0008, "step": 39781 }, { "epoch": 7.3, "learning_rate": 4.040630745284024e-07, "loss": 0.0005, "step": 39782 }, { "epoch": 7.3, "learning_rate": 4.038541891005632e-07, "loss": 0.0007, "step": 39783 }, { "epoch": 7.3, "learning_rate": 4.0364535656690404e-07, "loss": 0.0004, "step": 39784 }, { "epoch": 7.3, "learning_rate": 4.034365769285764e-07, "loss": 0.0003, "step": 39785 }, { "epoch": 7.3, "learning_rate": 4.0322785018673373e-07, "loss": 0.0004, "step": 39786 }, { "epoch": 7.3, "learning_rate": 4.0301917634252064e-07, "loss": 0.0002, "step": 39787 }, { "epoch": 7.3, "learning_rate": 4.028105553970918e-07, "loss": 0.0005, "step": 39788 }, { "epoch": 7.3, "learning_rate": 4.0260198735159514e-07, "loss": 0.0004, "step": 39789 }, { "epoch": 7.3, "learning_rate": 4.023934722071821e-07, "loss": 0.0003, "step": 39790 }, { "epoch": 7.3, "learning_rate": 4.0218500996500153e-07, "loss": 0.0002, "step": 39791 }, { "epoch": 7.3, "learning_rate": 4.0197660062620156e-07, "loss": 0.0006, "step": 39792 }, { "epoch": 7.3, "learning_rate": 4.0176824419193127e-07, "loss": 0.0003, "step": 39793 }, { "epoch": 7.3, "learning_rate": 4.015599406633397e-07, "loss": 0.0003, "step": 39794 }, { "epoch": 7.3, "learning_rate": 4.0135169004157483e-07, "loss": 0.0003, "step": 39795 }, { "epoch": 7.3, "learning_rate": 4.011434923277824e-07, "loss": 0.0004, "step": 39796 }, { "epoch": 7.3, "learning_rate": 4.009353475231148e-07, "loss": 0.0003, "step": 39797 }, { "epoch": 7.3, "learning_rate": 4.007272556287145e-07, "loss": 0.0006, "step": 39798 }, { "epoch": 7.3, "learning_rate": 4.0051921664573165e-07, "loss": 0.0002, "step": 39799 }, { "epoch": 7.3, "learning_rate": 4.0031123057530984e-07, "loss": 0.0006, "step": 39800 }, { "epoch": 7.3, "learning_rate": 4.001032974185981e-07, "loss": 0.0003, "step": 39801 }, { "epoch": 7.3, "learning_rate": 3.998954171767422e-07, "loss": 0.0005, "step": 39802 }, { "epoch": 7.3, "learning_rate": 3.9968758985088784e-07, "loss": 0.0005, "step": 39803 }, { "epoch": 7.3, "learning_rate": 3.9947981544218084e-07, "loss": 0.0004, "step": 39804 }, { "epoch": 7.3, "learning_rate": 3.992720939517658e-07, "loss": 0.001, "step": 39805 }, { "epoch": 7.3, "learning_rate": 3.9906442538078735e-07, "loss": 0.0005, "step": 39806 }, { "epoch": 7.3, "learning_rate": 3.988568097303913e-07, "loss": 0.0005, "step": 39807 }, { "epoch": 7.3, "learning_rate": 3.986492470017211e-07, "loss": 0.0007, "step": 39808 }, { "epoch": 7.3, "learning_rate": 3.984417371959215e-07, "loss": 0.0004, "step": 39809 }, { "epoch": 7.3, "learning_rate": 3.9823428031413593e-07, "loss": 0.0002, "step": 39810 }, { "epoch": 7.3, "learning_rate": 3.980268763575079e-07, "loss": 0.0004, "step": 39811 }, { "epoch": 7.3, "learning_rate": 3.9781952532718106e-07, "loss": 0.0004, "step": 39812 }, { "epoch": 7.3, "learning_rate": 3.976122272242966e-07, "loss": 0.0004, "step": 39813 }, { "epoch": 7.3, "learning_rate": 3.974049820499992e-07, "loss": 0.0007, "step": 39814 }, { "epoch": 7.3, "learning_rate": 3.9719778980543136e-07, "loss": 0.0005, "step": 39815 }, { "epoch": 7.3, "learning_rate": 3.969906504917342e-07, "loss": 0.0002, "step": 39816 }, { "epoch": 7.3, "learning_rate": 3.9678356411004927e-07, "loss": 0.0007, "step": 39817 }, { "epoch": 7.3, "learning_rate": 3.9657653066151657e-07, "loss": 0.0006, "step": 39818 }, { "epoch": 7.3, "learning_rate": 3.963695501472808e-07, "loss": 0.0004, "step": 39819 }, { "epoch": 7.3, "learning_rate": 3.9616262256848e-07, "loss": 0.0005, "step": 39820 }, { "epoch": 7.3, "learning_rate": 3.9595574792625656e-07, "loss": 0.0005, "step": 39821 }, { "epoch": 7.3, "learning_rate": 3.9574892622175063e-07, "loss": 0.0004, "step": 39822 }, { "epoch": 7.3, "learning_rate": 3.9554215745610024e-07, "loss": 0.0008, "step": 39823 }, { "epoch": 7.3, "learning_rate": 3.953354416304478e-07, "loss": 0.0003, "step": 39824 }, { "epoch": 7.3, "learning_rate": 3.9512877874593123e-07, "loss": 0.0009, "step": 39825 }, { "epoch": 7.3, "learning_rate": 3.949221688036897e-07, "loss": 0.0005, "step": 39826 }, { "epoch": 7.3, "learning_rate": 3.947156118048612e-07, "loss": 0.0009, "step": 39827 }, { "epoch": 7.3, "learning_rate": 3.945091077505869e-07, "loss": 0.0007, "step": 39828 }, { "epoch": 7.3, "learning_rate": 3.943026566420027e-07, "loss": 0.0003, "step": 39829 }, { "epoch": 7.3, "learning_rate": 3.9409625848024655e-07, "loss": 0.0003, "step": 39830 }, { "epoch": 7.3, "learning_rate": 3.9388991326645644e-07, "loss": 0.0003, "step": 39831 }, { "epoch": 7.3, "learning_rate": 3.936836210017714e-07, "loss": 0.0008, "step": 39832 }, { "epoch": 7.3, "learning_rate": 3.934773816873283e-07, "loss": 0.0003, "step": 39833 }, { "epoch": 7.3, "learning_rate": 3.932711953242618e-07, "loss": 0.0009, "step": 39834 }, { "epoch": 7.31, "learning_rate": 3.930650619137089e-07, "loss": 0.0003, "step": 39835 }, { "epoch": 7.31, "learning_rate": 3.9285898145680624e-07, "loss": 0.0012, "step": 39836 }, { "epoch": 7.31, "learning_rate": 3.926529539546908e-07, "loss": 0.0007, "step": 39837 }, { "epoch": 7.31, "learning_rate": 3.924469794084973e-07, "loss": 0.0007, "step": 39838 }, { "epoch": 7.31, "learning_rate": 3.922410578193625e-07, "loss": 0.0008, "step": 39839 }, { "epoch": 7.31, "learning_rate": 3.9203518918841774e-07, "loss": 0.0003, "step": 39840 }, { "epoch": 7.31, "learning_rate": 3.9182937351680104e-07, "loss": 0.0004, "step": 39841 }, { "epoch": 7.31, "learning_rate": 3.9162361080564594e-07, "loss": 0.0009, "step": 39842 }, { "epoch": 7.31, "learning_rate": 3.9141790105608703e-07, "loss": 0.0005, "step": 39843 }, { "epoch": 7.31, "learning_rate": 3.9121224426925675e-07, "loss": 0.0004, "step": 39844 }, { "epoch": 7.31, "learning_rate": 3.910066404462909e-07, "loss": 0.0004, "step": 39845 }, { "epoch": 7.31, "learning_rate": 3.908010895883208e-07, "loss": 0.0009, "step": 39846 }, { "epoch": 7.31, "learning_rate": 3.90595591696481e-07, "loss": 0.0004, "step": 39847 }, { "epoch": 7.31, "learning_rate": 3.903901467719029e-07, "loss": 0.0004, "step": 39848 }, { "epoch": 7.31, "learning_rate": 3.901847548157178e-07, "loss": 0.0005, "step": 39849 }, { "epoch": 7.31, "learning_rate": 3.899794158290604e-07, "loss": 0.0008, "step": 39850 }, { "epoch": 7.31, "learning_rate": 3.897741298130642e-07, "loss": 0.0003, "step": 39851 }, { "epoch": 7.31, "learning_rate": 3.89568896768856e-07, "loss": 0.0003, "step": 39852 }, { "epoch": 7.31, "learning_rate": 3.8936371669756834e-07, "loss": 0.0006, "step": 39853 }, { "epoch": 7.31, "learning_rate": 3.891585896003336e-07, "loss": 0.0005, "step": 39854 }, { "epoch": 7.31, "learning_rate": 3.8895351547828196e-07, "loss": 0.0006, "step": 39855 }, { "epoch": 7.31, "learning_rate": 3.887484943325448e-07, "loss": 0.0009, "step": 39856 }, { "epoch": 7.31, "learning_rate": 3.8854352616425005e-07, "loss": 0.0007, "step": 39857 }, { "epoch": 7.31, "learning_rate": 3.8833861097452907e-07, "loss": 0.0007, "step": 39858 }, { "epoch": 7.31, "learning_rate": 3.8813374876451093e-07, "loss": 0.0003, "step": 39859 }, { "epoch": 7.31, "learning_rate": 3.8792893953532363e-07, "loss": 0.0002, "step": 39860 }, { "epoch": 7.31, "learning_rate": 3.8772418328809847e-07, "loss": 0.0009, "step": 39861 }, { "epoch": 7.31, "learning_rate": 3.8751948002396234e-07, "loss": 0.0003, "step": 39862 }, { "epoch": 7.31, "learning_rate": 3.873148297440443e-07, "loss": 0.0003, "step": 39863 }, { "epoch": 7.31, "learning_rate": 3.8711023244947133e-07, "loss": 0.0007, "step": 39864 }, { "epoch": 7.31, "learning_rate": 3.8690568814137355e-07, "loss": 0.0003, "step": 39865 }, { "epoch": 7.31, "learning_rate": 3.867011968208745e-07, "loss": 0.0002, "step": 39866 }, { "epoch": 7.31, "learning_rate": 3.8649675848910437e-07, "loss": 0.0003, "step": 39867 }, { "epoch": 7.31, "learning_rate": 3.8629237314719016e-07, "loss": 0.0003, "step": 39868 }, { "epoch": 7.31, "learning_rate": 3.8608804079625864e-07, "loss": 0.0004, "step": 39869 }, { "epoch": 7.31, "learning_rate": 3.8588376143743333e-07, "loss": 0.0003, "step": 39870 }, { "epoch": 7.31, "learning_rate": 3.856795350718412e-07, "loss": 0.0005, "step": 39871 }, { "epoch": 7.31, "learning_rate": 3.8547536170061016e-07, "loss": 0.0004, "step": 39872 }, { "epoch": 7.31, "learning_rate": 3.8527124132486383e-07, "loss": 0.001, "step": 39873 }, { "epoch": 7.31, "learning_rate": 3.850671739457279e-07, "loss": 0.0003, "step": 39874 }, { "epoch": 7.31, "learning_rate": 3.8486315956432706e-07, "loss": 0.0005, "step": 39875 }, { "epoch": 7.31, "learning_rate": 3.84659198181786e-07, "loss": 0.0006, "step": 39876 }, { "epoch": 7.31, "learning_rate": 3.844552897992282e-07, "loss": 0.0004, "step": 39877 }, { "epoch": 7.31, "learning_rate": 3.8425143441777833e-07, "loss": 0.0003, "step": 39878 }, { "epoch": 7.31, "learning_rate": 3.840476320385578e-07, "loss": 0.0005, "step": 39879 }, { "epoch": 7.31, "learning_rate": 3.838438826626956e-07, "loss": 0.0004, "step": 39880 }, { "epoch": 7.31, "learning_rate": 3.836401862913086e-07, "loss": 0.0002, "step": 39881 }, { "epoch": 7.31, "learning_rate": 3.834365429255238e-07, "loss": 0.0002, "step": 39882 }, { "epoch": 7.31, "learning_rate": 3.832329525664613e-07, "loss": 0.0003, "step": 39883 }, { "epoch": 7.31, "learning_rate": 3.830294152152425e-07, "loss": 0.0004, "step": 39884 }, { "epoch": 7.31, "learning_rate": 3.82825930872992e-07, "loss": 0.0005, "step": 39885 }, { "epoch": 7.31, "learning_rate": 3.826224995408323e-07, "loss": 0.0003, "step": 39886 }, { "epoch": 7.31, "learning_rate": 3.8241912121988133e-07, "loss": 0.0005, "step": 39887 }, { "epoch": 7.31, "learning_rate": 3.822157959112605e-07, "loss": 0.0009, "step": 39888 }, { "epoch": 7.32, "learning_rate": 3.8201252361609207e-07, "loss": 0.0008, "step": 39889 }, { "epoch": 7.32, "learning_rate": 3.8180930433549537e-07, "loss": 0.0004, "step": 39890 }, { "epoch": 7.32, "learning_rate": 3.8160613807059263e-07, "loss": 0.0021, "step": 39891 }, { "epoch": 7.32, "learning_rate": 3.814030248225009e-07, "loss": 0.0004, "step": 39892 }, { "epoch": 7.32, "learning_rate": 3.8119996459234144e-07, "loss": 0.0003, "step": 39893 }, { "epoch": 7.32, "learning_rate": 3.8099695738123334e-07, "loss": 0.0009, "step": 39894 }, { "epoch": 7.32, "learning_rate": 3.8079400319029567e-07, "loss": 0.0009, "step": 39895 }, { "epoch": 7.32, "learning_rate": 3.8059110202064656e-07, "loss": 0.0002, "step": 39896 }, { "epoch": 7.32, "learning_rate": 3.803882538734027e-07, "loss": 0.0003, "step": 39897 }, { "epoch": 7.32, "learning_rate": 3.801854587496878e-07, "loss": 0.0002, "step": 39898 }, { "epoch": 7.32, "learning_rate": 3.7998271665061426e-07, "loss": 0.0004, "step": 39899 }, { "epoch": 7.32, "learning_rate": 3.7978002757730115e-07, "loss": 0.0002, "step": 39900 }, { "epoch": 7.32, "learning_rate": 3.7957739153086427e-07, "loss": 0.0005, "step": 39901 }, { "epoch": 7.32, "learning_rate": 3.7937480851242493e-07, "loss": 0.0003, "step": 39902 }, { "epoch": 7.32, "learning_rate": 3.7917227852309555e-07, "loss": 0.0003, "step": 39903 }, { "epoch": 7.32, "learning_rate": 3.7896980156399533e-07, "loss": 0.0003, "step": 39904 }, { "epoch": 7.32, "learning_rate": 3.7876737763623886e-07, "loss": 0.0008, "step": 39905 }, { "epoch": 7.32, "learning_rate": 3.7856500674093966e-07, "loss": 0.0004, "step": 39906 }, { "epoch": 7.32, "learning_rate": 3.783626888792169e-07, "loss": 0.0004, "step": 39907 }, { "epoch": 7.32, "learning_rate": 3.781604240521852e-07, "loss": 0.0004, "step": 39908 }, { "epoch": 7.32, "learning_rate": 3.779582122609582e-07, "loss": 0.0006, "step": 39909 }, { "epoch": 7.32, "learning_rate": 3.7775605350665047e-07, "loss": 0.0002, "step": 39910 }, { "epoch": 7.32, "learning_rate": 3.775539477903778e-07, "loss": 0.0004, "step": 39911 }, { "epoch": 7.32, "learning_rate": 3.7735189511325267e-07, "loss": 0.0005, "step": 39912 }, { "epoch": 7.32, "learning_rate": 3.771498954763886e-07, "loss": 0.0003, "step": 39913 }, { "epoch": 7.32, "learning_rate": 3.7694794888090025e-07, "loss": 0.0003, "step": 39914 }, { "epoch": 7.32, "learning_rate": 3.7674605532789897e-07, "loss": 0.0005, "step": 39915 }, { "epoch": 7.32, "learning_rate": 3.765442148185017e-07, "loss": 0.0005, "step": 39916 }, { "epoch": 7.32, "learning_rate": 3.763424273538174e-07, "loss": 0.0005, "step": 39917 }, { "epoch": 7.32, "learning_rate": 3.761406929349587e-07, "loss": 0.0003, "step": 39918 }, { "epoch": 7.32, "learning_rate": 3.759390115630357e-07, "loss": 0.0002, "step": 39919 }, { "epoch": 7.32, "learning_rate": 3.757373832391642e-07, "loss": 0.0003, "step": 39920 }, { "epoch": 7.32, "learning_rate": 3.7553580796445445e-07, "loss": 0.0003, "step": 39921 }, { "epoch": 7.32, "learning_rate": 3.7533428574001774e-07, "loss": 0.0002, "step": 39922 }, { "epoch": 7.32, "learning_rate": 3.751328165669621e-07, "loss": 0.0004, "step": 39923 }, { "epoch": 7.32, "learning_rate": 3.749314004463989e-07, "loss": 0.0005, "step": 39924 }, { "epoch": 7.32, "learning_rate": 3.747300373794405e-07, "loss": 0.0004, "step": 39925 }, { "epoch": 7.32, "learning_rate": 3.7452872736719603e-07, "loss": 0.0002, "step": 39926 }, { "epoch": 7.32, "learning_rate": 3.743274704107747e-07, "loss": 0.0004, "step": 39927 }, { "epoch": 7.32, "learning_rate": 3.741262665112855e-07, "loss": 0.0003, "step": 39928 }, { "epoch": 7.32, "learning_rate": 3.739251156698376e-07, "loss": 0.0004, "step": 39929 }, { "epoch": 7.32, "learning_rate": 3.737240178875401e-07, "loss": 0.0012, "step": 39930 }, { "epoch": 7.32, "learning_rate": 3.7352297316550215e-07, "loss": 0.0003, "step": 39931 }, { "epoch": 7.32, "learning_rate": 3.7332198150482946e-07, "loss": 0.0006, "step": 39932 }, { "epoch": 7.32, "learning_rate": 3.7312104290663343e-07, "loss": 0.0013, "step": 39933 }, { "epoch": 7.32, "learning_rate": 3.729201573720187e-07, "loss": 0.0003, "step": 39934 }, { "epoch": 7.32, "learning_rate": 3.7271932490209327e-07, "loss": 0.0002, "step": 39935 }, { "epoch": 7.32, "learning_rate": 3.725185454979652e-07, "loss": 0.0014, "step": 39936 }, { "epoch": 7.32, "learning_rate": 3.723178191607391e-07, "loss": 0.0004, "step": 39937 }, { "epoch": 7.32, "learning_rate": 3.7211714589152403e-07, "loss": 0.0006, "step": 39938 }, { "epoch": 7.32, "learning_rate": 3.719165256914248e-07, "loss": 0.0002, "step": 39939 }, { "epoch": 7.32, "learning_rate": 3.7171595856154817e-07, "loss": 0.0002, "step": 39940 }, { "epoch": 7.32, "learning_rate": 3.715154445029967e-07, "loss": 0.0004, "step": 39941 }, { "epoch": 7.32, "learning_rate": 3.7131498351687833e-07, "loss": 0.0003, "step": 39942 }, { "epoch": 7.32, "learning_rate": 3.711145756042978e-07, "loss": 0.0004, "step": 39943 }, { "epoch": 7.33, "learning_rate": 3.709142207663585e-07, "loss": 0.0005, "step": 39944 }, { "epoch": 7.33, "learning_rate": 3.7071391900416644e-07, "loss": 0.0002, "step": 39945 }, { "epoch": 7.33, "learning_rate": 3.7051367031882503e-07, "loss": 0.0003, "step": 39946 }, { "epoch": 7.33, "learning_rate": 3.7031347471143784e-07, "loss": 0.0006, "step": 39947 }, { "epoch": 7.33, "learning_rate": 3.701133321831074e-07, "loss": 0.0003, "step": 39948 }, { "epoch": 7.33, "learning_rate": 3.699132427349383e-07, "loss": 0.0006, "step": 39949 }, { "epoch": 7.33, "learning_rate": 3.69713206368032e-07, "loss": 0.0009, "step": 39950 }, { "epoch": 7.33, "learning_rate": 3.6951322308349524e-07, "loss": 0.0009, "step": 39951 }, { "epoch": 7.33, "learning_rate": 3.69313292882425e-07, "loss": 0.0008, "step": 39952 }, { "epoch": 7.33, "learning_rate": 3.6911341576592594e-07, "loss": 0.0005, "step": 39953 }, { "epoch": 7.33, "learning_rate": 3.6891359173509833e-07, "loss": 0.0006, "step": 39954 }, { "epoch": 7.33, "learning_rate": 3.6871382079104567e-07, "loss": 0.0002, "step": 39955 }, { "epoch": 7.33, "learning_rate": 3.685141029348671e-07, "loss": 0.0004, "step": 39956 }, { "epoch": 7.33, "learning_rate": 3.683144381676662e-07, "loss": 0.0004, "step": 39957 }, { "epoch": 7.33, "learning_rate": 3.681148264905421e-07, "loss": 0.0006, "step": 39958 }, { "epoch": 7.33, "learning_rate": 3.6791526790459165e-07, "loss": 0.0003, "step": 39959 }, { "epoch": 7.33, "learning_rate": 3.6771576241091955e-07, "loss": 0.0003, "step": 39960 }, { "epoch": 7.33, "learning_rate": 3.675163100106238e-07, "loss": 0.0006, "step": 39961 }, { "epoch": 7.33, "learning_rate": 3.673169107048047e-07, "loss": 0.0004, "step": 39962 }, { "epoch": 7.33, "learning_rate": 3.6711756449456017e-07, "loss": 0.0003, "step": 39963 }, { "epoch": 7.33, "learning_rate": 3.6691827138098825e-07, "loss": 0.0006, "step": 39964 }, { "epoch": 7.33, "learning_rate": 3.6671903136518917e-07, "loss": 0.0006, "step": 39965 }, { "epoch": 7.33, "learning_rate": 3.6651984444826095e-07, "loss": 0.0018, "step": 39966 }, { "epoch": 7.33, "learning_rate": 3.663207106313005e-07, "loss": 0.0006, "step": 39967 }, { "epoch": 7.33, "learning_rate": 3.661216299154047e-07, "loss": 0.0005, "step": 39968 }, { "epoch": 7.33, "learning_rate": 3.6592260230167486e-07, "loss": 0.001, "step": 39969 }, { "epoch": 7.33, "learning_rate": 3.657236277912035e-07, "loss": 0.0006, "step": 39970 }, { "epoch": 7.33, "learning_rate": 3.6552470638508974e-07, "loss": 0.0004, "step": 39971 }, { "epoch": 7.33, "learning_rate": 3.653258380844282e-07, "loss": 0.0007, "step": 39972 }, { "epoch": 7.33, "learning_rate": 3.651270228903181e-07, "loss": 0.0002, "step": 39973 }, { "epoch": 7.33, "learning_rate": 3.649282608038529e-07, "loss": 0.0008, "step": 39974 }, { "epoch": 7.33, "learning_rate": 3.6472955182612846e-07, "loss": 0.0003, "step": 39975 }, { "epoch": 7.33, "learning_rate": 3.645308959582428e-07, "loss": 0.0005, "step": 39976 }, { "epoch": 7.33, "learning_rate": 3.643322932012849e-07, "loss": 0.0005, "step": 39977 }, { "epoch": 7.33, "learning_rate": 3.641337435563552e-07, "loss": 0.0007, "step": 39978 }, { "epoch": 7.33, "learning_rate": 3.639352470245461e-07, "loss": 0.0006, "step": 39979 }, { "epoch": 7.33, "learning_rate": 3.637368036069522e-07, "loss": 0.0003, "step": 39980 }, { "epoch": 7.33, "learning_rate": 3.63538413304666e-07, "loss": 0.0011, "step": 39981 }, { "epoch": 7.33, "learning_rate": 3.633400761187811e-07, "loss": 0.0006, "step": 39982 }, { "epoch": 7.33, "learning_rate": 3.631417920503921e-07, "loss": 0.0004, "step": 39983 }, { "epoch": 7.33, "learning_rate": 3.629435611005916e-07, "loss": 0.0003, "step": 39984 }, { "epoch": 7.33, "learning_rate": 3.627453832704708e-07, "loss": 0.0006, "step": 39985 }, { "epoch": 7.33, "learning_rate": 3.6254725856112447e-07, "loss": 0.0002, "step": 39986 }, { "epoch": 7.33, "learning_rate": 3.6234918697364396e-07, "loss": 0.0003, "step": 39987 }, { "epoch": 7.33, "learning_rate": 3.621511685091206e-07, "loss": 0.0004, "step": 39988 }, { "epoch": 7.33, "learning_rate": 3.619532031686446e-07, "loss": 0.0003, "step": 39989 }, { "epoch": 7.33, "learning_rate": 3.617552909533073e-07, "loss": 0.0004, "step": 39990 }, { "epoch": 7.33, "learning_rate": 3.615574318642023e-07, "loss": 0.0003, "step": 39991 }, { "epoch": 7.33, "learning_rate": 3.613596259024188e-07, "loss": 0.0003, "step": 39992 }, { "epoch": 7.33, "learning_rate": 3.6116187306904695e-07, "loss": 0.0006, "step": 39993 }, { "epoch": 7.33, "learning_rate": 3.6096417336517477e-07, "loss": 0.0002, "step": 39994 }, { "epoch": 7.33, "learning_rate": 3.6076652679189583e-07, "loss": 0.0003, "step": 39995 }, { "epoch": 7.33, "learning_rate": 3.60568933350296e-07, "loss": 0.0005, "step": 39996 }, { "epoch": 7.33, "learning_rate": 3.603713930414676e-07, "loss": 0.0004, "step": 39997 }, { "epoch": 7.34, "learning_rate": 3.601739058664966e-07, "loss": 0.0009, "step": 39998 }, { "epoch": 7.34, "learning_rate": 3.599764718264742e-07, "loss": 0.0003, "step": 39999 }, { "epoch": 7.34, "learning_rate": 3.597790909224863e-07, "loss": 0.0003, "step": 40000 }, { "epoch": 7.34, "learning_rate": 3.595817631556231e-07, "loss": 0.0002, "step": 40001 }, { "epoch": 7.34, "learning_rate": 3.5938448852696927e-07, "loss": 0.0002, "step": 40002 }, { "epoch": 7.34, "learning_rate": 3.59187267037614e-07, "loss": 0.0004, "step": 40003 }, { "epoch": 7.34, "learning_rate": 3.589900986886452e-07, "loss": 0.0004, "step": 40004 }, { "epoch": 7.34, "learning_rate": 3.58792983481151e-07, "loss": 0.0014, "step": 40005 }, { "epoch": 7.34, "learning_rate": 3.585959214162138e-07, "loss": 0.0003, "step": 40006 }, { "epoch": 7.34, "learning_rate": 3.5839891249492056e-07, "loss": 0.0006, "step": 40007 }, { "epoch": 7.34, "learning_rate": 3.5820195671835924e-07, "loss": 0.0003, "step": 40008 }, { "epoch": 7.34, "learning_rate": 3.5800505408761457e-07, "loss": 0.0004, "step": 40009 }, { "epoch": 7.34, "learning_rate": 3.578082046037723e-07, "loss": 0.0005, "step": 40010 }, { "epoch": 7.34, "learning_rate": 3.576114082679183e-07, "loss": 0.0014, "step": 40011 }, { "epoch": 7.34, "learning_rate": 3.5741466508113385e-07, "loss": 0.0002, "step": 40012 }, { "epoch": 7.34, "learning_rate": 3.5721797504450706e-07, "loss": 0.0004, "step": 40013 }, { "epoch": 7.34, "learning_rate": 3.570213381591203e-07, "loss": 0.0002, "step": 40014 }, { "epoch": 7.34, "learning_rate": 3.5682475442605723e-07, "loss": 0.0006, "step": 40015 }, { "epoch": 7.34, "learning_rate": 3.566282238464036e-07, "loss": 0.001, "step": 40016 }, { "epoch": 7.34, "learning_rate": 3.5643174642123965e-07, "loss": 0.0008, "step": 40017 }, { "epoch": 7.34, "learning_rate": 3.562353221516501e-07, "loss": 0.0001, "step": 40018 }, { "epoch": 7.34, "learning_rate": 3.560389510387186e-07, "loss": 0.0004, "step": 40019 }, { "epoch": 7.34, "learning_rate": 3.558426330835252e-07, "loss": 0.0004, "step": 40020 }, { "epoch": 7.34, "learning_rate": 3.5564636828715136e-07, "loss": 0.0002, "step": 40021 }, { "epoch": 7.34, "learning_rate": 3.5545015665068293e-07, "loss": 0.0005, "step": 40022 }, { "epoch": 7.34, "learning_rate": 3.5525399817520014e-07, "loss": 0.0004, "step": 40023 }, { "epoch": 7.34, "learning_rate": 3.5505789286178205e-07, "loss": 0.0003, "step": 40024 }, { "epoch": 7.34, "learning_rate": 3.548618407115101e-07, "loss": 0.0007, "step": 40025 }, { "epoch": 7.34, "learning_rate": 3.546658417254667e-07, "loss": 0.0011, "step": 40026 }, { "epoch": 7.34, "learning_rate": 3.5446989590472993e-07, "loss": 0.0005, "step": 40027 }, { "epoch": 7.34, "learning_rate": 3.5427400325038217e-07, "loss": 0.0006, "step": 40028 }, { "epoch": 7.34, "learning_rate": 3.540781637635016e-07, "loss": 0.0003, "step": 40029 }, { "epoch": 7.34, "learning_rate": 3.538823774451683e-07, "loss": 0.0002, "step": 40030 }, { "epoch": 7.34, "learning_rate": 3.5368664429646147e-07, "loss": 0.0001, "step": 40031 }, { "epoch": 7.34, "learning_rate": 3.5349096431846033e-07, "loss": 0.0008, "step": 40032 }, { "epoch": 7.34, "learning_rate": 3.5329533751224274e-07, "loss": 0.0004, "step": 40033 }, { "epoch": 7.34, "learning_rate": 3.530997638788869e-07, "loss": 0.0006, "step": 40034 }, { "epoch": 7.34, "learning_rate": 3.529042434194707e-07, "loss": 0.0005, "step": 40035 }, { "epoch": 7.34, "learning_rate": 3.5270877613507337e-07, "loss": 0.0007, "step": 40036 }, { "epoch": 7.34, "learning_rate": 3.525133620267707e-07, "loss": 0.0003, "step": 40037 }, { "epoch": 7.34, "learning_rate": 3.523180010956395e-07, "loss": 0.0004, "step": 40038 }, { "epoch": 7.34, "learning_rate": 3.5212269334275795e-07, "loss": 0.0005, "step": 40039 }, { "epoch": 7.34, "learning_rate": 3.51927438769204e-07, "loss": 0.0007, "step": 40040 }, { "epoch": 7.34, "learning_rate": 3.5173223737605124e-07, "loss": 0.0004, "step": 40041 }, { "epoch": 7.34, "learning_rate": 3.5153708916437544e-07, "loss": 0.0003, "step": 40042 }, { "epoch": 7.34, "learning_rate": 3.513419941352525e-07, "loss": 0.0002, "step": 40043 }, { "epoch": 7.34, "learning_rate": 3.5114695228976036e-07, "loss": 0.0002, "step": 40044 }, { "epoch": 7.34, "learning_rate": 3.5095196362897154e-07, "loss": 0.0007, "step": 40045 }, { "epoch": 7.34, "learning_rate": 3.5075702815396075e-07, "loss": 0.0008, "step": 40046 }, { "epoch": 7.34, "learning_rate": 3.505621458658048e-07, "loss": 0.0012, "step": 40047 }, { "epoch": 7.34, "learning_rate": 3.503673167655752e-07, "loss": 0.0005, "step": 40048 }, { "epoch": 7.34, "learning_rate": 3.5017254085434773e-07, "loss": 0.0004, "step": 40049 }, { "epoch": 7.34, "learning_rate": 3.4997781813319475e-07, "loss": 0.0004, "step": 40050 }, { "epoch": 7.34, "learning_rate": 3.4978314860319104e-07, "loss": 0.0003, "step": 40051 }, { "epoch": 7.34, "learning_rate": 3.49588532265408e-07, "loss": 0.0003, "step": 40052 }, { "epoch": 7.35, "learning_rate": 3.4939396912091917e-07, "loss": 0.0003, "step": 40053 }, { "epoch": 7.35, "learning_rate": 3.49199459170797e-07, "loss": 0.0003, "step": 40054 }, { "epoch": 7.35, "learning_rate": 3.490050024161129e-07, "loss": 0.001, "step": 40055 }, { "epoch": 7.35, "learning_rate": 3.4881059885793935e-07, "loss": 0.0009, "step": 40056 }, { "epoch": 7.35, "learning_rate": 3.486162484973477e-07, "loss": 0.0022, "step": 40057 }, { "epoch": 7.35, "learning_rate": 3.4842195133541147e-07, "loss": 0.0007, "step": 40058 }, { "epoch": 7.35, "learning_rate": 3.482277073731988e-07, "loss": 0.0006, "step": 40059 }, { "epoch": 7.35, "learning_rate": 3.480335166117788e-07, "loss": 0.0003, "step": 40060 }, { "epoch": 7.35, "learning_rate": 3.478393790522261e-07, "loss": 0.0006, "step": 40061 }, { "epoch": 7.35, "learning_rate": 3.476452946956088e-07, "loss": 0.0002, "step": 40062 }, { "epoch": 7.35, "learning_rate": 3.4745126354299606e-07, "loss": 0.0002, "step": 40063 }, { "epoch": 7.35, "learning_rate": 3.472572855954581e-07, "loss": 0.0006, "step": 40064 }, { "epoch": 7.35, "learning_rate": 3.470633608540652e-07, "loss": 0.0004, "step": 40065 }, { "epoch": 7.35, "learning_rate": 3.468694893198843e-07, "loss": 0.0002, "step": 40066 }, { "epoch": 7.35, "learning_rate": 3.466756709939856e-07, "loss": 0.0003, "step": 40067 }, { "epoch": 7.35, "learning_rate": 3.46481905877436e-07, "loss": 0.0004, "step": 40068 }, { "epoch": 7.35, "learning_rate": 3.4628819397130366e-07, "loss": 0.0002, "step": 40069 }, { "epoch": 7.35, "learning_rate": 3.460945352766598e-07, "loss": 0.0003, "step": 40070 }, { "epoch": 7.35, "learning_rate": 3.45900929794567e-07, "loss": 0.0003, "step": 40071 }, { "epoch": 7.35, "learning_rate": 3.457073775260944e-07, "loss": 0.0003, "step": 40072 }, { "epoch": 7.35, "learning_rate": 3.4551387847230887e-07, "loss": 0.0007, "step": 40073 }, { "epoch": 7.35, "learning_rate": 3.453204326342763e-07, "loss": 0.0002, "step": 40074 }, { "epoch": 7.35, "learning_rate": 3.451270400130646e-07, "loss": 0.0004, "step": 40075 }, { "epoch": 7.35, "learning_rate": 3.4493370060974083e-07, "loss": 0.0008, "step": 40076 }, { "epoch": 7.35, "learning_rate": 3.447404144253663e-07, "loss": 0.0009, "step": 40077 }, { "epoch": 7.35, "learning_rate": 3.4454718146100794e-07, "loss": 0.0006, "step": 40078 }, { "epoch": 7.35, "learning_rate": 3.4435400171773384e-07, "loss": 0.0018, "step": 40079 }, { "epoch": 7.35, "learning_rate": 3.441608751966052e-07, "loss": 0.0007, "step": 40080 }, { "epoch": 7.35, "learning_rate": 3.439678018986892e-07, "loss": 0.0005, "step": 40081 }, { "epoch": 7.35, "learning_rate": 3.437747818250481e-07, "loss": 0.0004, "step": 40082 }, { "epoch": 7.35, "learning_rate": 3.435818149767478e-07, "loss": 0.0004, "step": 40083 }, { "epoch": 7.35, "learning_rate": 3.433889013548497e-07, "loss": 0.0003, "step": 40084 }, { "epoch": 7.35, "learning_rate": 3.4319604096041846e-07, "loss": 0.0004, "step": 40085 }, { "epoch": 7.35, "learning_rate": 3.430032337945166e-07, "loss": 0.0005, "step": 40086 }, { "epoch": 7.35, "learning_rate": 3.428104798582077e-07, "loss": 0.0003, "step": 40087 }, { "epoch": 7.35, "learning_rate": 3.4261777915255315e-07, "loss": 0.0004, "step": 40088 }, { "epoch": 7.35, "learning_rate": 3.424251316786165e-07, "loss": 0.0004, "step": 40089 }, { "epoch": 7.35, "learning_rate": 3.42232537437458e-07, "loss": 0.0006, "step": 40090 }, { "epoch": 7.35, "learning_rate": 3.4203999643013905e-07, "loss": 0.0003, "step": 40091 }, { "epoch": 7.35, "learning_rate": 3.4184750865772333e-07, "loss": 0.0002, "step": 40092 }, { "epoch": 7.35, "learning_rate": 3.4165507412126984e-07, "loss": 0.0006, "step": 40093 }, { "epoch": 7.35, "learning_rate": 3.414626928218412e-07, "loss": 0.0003, "step": 40094 }, { "epoch": 7.35, "learning_rate": 3.4127036476049533e-07, "loss": 0.0006, "step": 40095 }, { "epoch": 7.35, "learning_rate": 3.4107808993829263e-07, "loss": 0.0008, "step": 40096 }, { "epoch": 7.35, "learning_rate": 3.408858683562943e-07, "loss": 0.0002, "step": 40097 }, { "epoch": 7.35, "learning_rate": 3.4069370001555967e-07, "loss": 0.0004, "step": 40098 }, { "epoch": 7.35, "learning_rate": 3.405015849171478e-07, "loss": 0.0002, "step": 40099 }, { "epoch": 7.35, "learning_rate": 3.403095230621178e-07, "loss": 0.0005, "step": 40100 }, { "epoch": 7.35, "learning_rate": 3.4011751445152784e-07, "loss": 0.0004, "step": 40101 }, { "epoch": 7.35, "learning_rate": 3.399255590864359e-07, "loss": 0.0005, "step": 40102 }, { "epoch": 7.35, "learning_rate": 3.3973365696790107e-07, "loss": 0.0003, "step": 40103 }, { "epoch": 7.35, "learning_rate": 3.395418080969803e-07, "loss": 0.0011, "step": 40104 }, { "epoch": 7.35, "learning_rate": 3.3935001247473176e-07, "loss": 0.0006, "step": 40105 }, { "epoch": 7.35, "learning_rate": 3.3915827010221226e-07, "loss": 0.0008, "step": 40106 }, { "epoch": 7.36, "learning_rate": 3.3896658098047876e-07, "loss": 0.0005, "step": 40107 }, { "epoch": 7.36, "learning_rate": 3.3877494511058815e-07, "loss": 0.0002, "step": 40108 }, { "epoch": 7.36, "learning_rate": 3.3858336249359413e-07, "loss": 0.0005, "step": 40109 }, { "epoch": 7.36, "learning_rate": 3.3839183313055693e-07, "loss": 0.0005, "step": 40110 }, { "epoch": 7.36, "learning_rate": 3.382003570225301e-07, "loss": 0.0008, "step": 40111 }, { "epoch": 7.36, "learning_rate": 3.3800893417057056e-07, "loss": 0.0005, "step": 40112 }, { "epoch": 7.36, "learning_rate": 3.3781756457572977e-07, "loss": 0.0008, "step": 40113 }, { "epoch": 7.36, "learning_rate": 3.3762624823906577e-07, "loss": 0.0002, "step": 40114 }, { "epoch": 7.36, "learning_rate": 3.374349851616321e-07, "loss": 0.0005, "step": 40115 }, { "epoch": 7.36, "learning_rate": 3.372437753444835e-07, "loss": 0.0006, "step": 40116 }, { "epoch": 7.36, "learning_rate": 3.370526187886736e-07, "loss": 0.0003, "step": 40117 }, { "epoch": 7.36, "learning_rate": 3.3686151549525593e-07, "loss": 0.0003, "step": 40118 }, { "epoch": 7.36, "learning_rate": 3.3667046546528415e-07, "loss": 0.0004, "step": 40119 }, { "epoch": 7.36, "learning_rate": 3.364794686998107e-07, "loss": 0.0006, "step": 40120 }, { "epoch": 7.36, "learning_rate": 3.362885251998893e-07, "loss": 0.0003, "step": 40121 }, { "epoch": 7.36, "learning_rate": 3.360976349665712e-07, "loss": 0.0002, "step": 40122 }, { "epoch": 7.36, "learning_rate": 3.3590679800091117e-07, "loss": 0.0004, "step": 40123 }, { "epoch": 7.36, "learning_rate": 3.357160143039584e-07, "loss": 0.0008, "step": 40124 }, { "epoch": 7.36, "learning_rate": 3.355252838767642e-07, "loss": 0.0002, "step": 40125 }, { "epoch": 7.36, "learning_rate": 3.353346067203822e-07, "loss": 0.0005, "step": 40126 }, { "epoch": 7.36, "learning_rate": 3.351439828358605e-07, "loss": 0.0003, "step": 40127 }, { "epoch": 7.36, "learning_rate": 3.3495341222425373e-07, "loss": 0.0004, "step": 40128 }, { "epoch": 7.36, "learning_rate": 3.3476289488660885e-07, "loss": 0.0003, "step": 40129 }, { "epoch": 7.36, "learning_rate": 3.345724308239795e-07, "loss": 0.0003, "step": 40130 }, { "epoch": 7.36, "learning_rate": 3.343820200374104e-07, "loss": 0.0004, "step": 40131 }, { "epoch": 7.36, "learning_rate": 3.341916625279562e-07, "loss": 0.0004, "step": 40132 }, { "epoch": 7.36, "learning_rate": 3.340013582966628e-07, "loss": 0.0006, "step": 40133 }, { "epoch": 7.36, "learning_rate": 3.3381110734458046e-07, "loss": 0.0004, "step": 40134 }, { "epoch": 7.36, "learning_rate": 3.336209096727583e-07, "loss": 0.0004, "step": 40135 }, { "epoch": 7.36, "learning_rate": 3.3343076528224436e-07, "loss": 0.0002, "step": 40136 }, { "epoch": 7.36, "learning_rate": 3.332406741740857e-07, "loss": 0.0004, "step": 40137 }, { "epoch": 7.36, "learning_rate": 3.3305063634933133e-07, "loss": 0.0002, "step": 40138 }, { "epoch": 7.36, "learning_rate": 3.328606518090283e-07, "loss": 0.0003, "step": 40139 }, { "epoch": 7.36, "learning_rate": 3.3267072055422344e-07, "loss": 0.0004, "step": 40140 }, { "epoch": 7.36, "learning_rate": 3.324808425859649e-07, "loss": 0.0003, "step": 40141 }, { "epoch": 7.36, "learning_rate": 3.322910179052985e-07, "loss": 0.0004, "step": 40142 }, { "epoch": 7.36, "learning_rate": 3.3210124651327e-07, "loss": 0.0003, "step": 40143 }, { "epoch": 7.36, "learning_rate": 3.319115284109242e-07, "loss": 0.0002, "step": 40144 }, { "epoch": 7.36, "learning_rate": 3.3172186359931026e-07, "loss": 0.0004, "step": 40145 }, { "epoch": 7.36, "learning_rate": 3.3153225207947283e-07, "loss": 0.0003, "step": 40146 }, { "epoch": 7.36, "learning_rate": 3.313426938524555e-07, "loss": 0.0004, "step": 40147 }, { "epoch": 7.36, "learning_rate": 3.3115318891930425e-07, "loss": 0.0002, "step": 40148 }, { "epoch": 7.36, "learning_rate": 3.3096373728106146e-07, "loss": 0.0006, "step": 40149 }, { "epoch": 7.36, "learning_rate": 3.307743389387741e-07, "loss": 0.0002, "step": 40150 }, { "epoch": 7.36, "learning_rate": 3.3058499389348465e-07, "loss": 0.0004, "step": 40151 }, { "epoch": 7.36, "learning_rate": 3.303957021462378e-07, "loss": 0.0002, "step": 40152 }, { "epoch": 7.36, "learning_rate": 3.302064636980762e-07, "loss": 0.0002, "step": 40153 }, { "epoch": 7.36, "learning_rate": 3.300172785500422e-07, "loss": 0.0003, "step": 40154 }, { "epoch": 7.36, "learning_rate": 3.2982814670318053e-07, "loss": 0.0005, "step": 40155 }, { "epoch": 7.36, "learning_rate": 3.2963906815853263e-07, "loss": 0.0004, "step": 40156 }, { "epoch": 7.36, "learning_rate": 3.294500429171399e-07, "loss": 0.0002, "step": 40157 }, { "epoch": 7.36, "learning_rate": 3.292610709800459e-07, "loss": 0.0006, "step": 40158 }, { "epoch": 7.36, "learning_rate": 3.2907215234829205e-07, "loss": 0.0007, "step": 40159 }, { "epoch": 7.36, "learning_rate": 3.288832870229186e-07, "loss": 0.0009, "step": 40160 }, { "epoch": 7.36, "learning_rate": 3.28694475004967e-07, "loss": 0.0008, "step": 40161 }, { "epoch": 7.37, "learning_rate": 3.285057162954763e-07, "loss": 0.0003, "step": 40162 }, { "epoch": 7.37, "learning_rate": 3.283170108954903e-07, "loss": 0.0007, "step": 40163 }, { "epoch": 7.37, "learning_rate": 3.28128358806048e-07, "loss": 0.0008, "step": 40164 }, { "epoch": 7.37, "learning_rate": 3.279397600281886e-07, "loss": 0.0006, "step": 40165 }, { "epoch": 7.37, "learning_rate": 3.2775121456295024e-07, "loss": 0.0005, "step": 40166 }, { "epoch": 7.37, "learning_rate": 3.2756272241137535e-07, "loss": 0.0005, "step": 40167 }, { "epoch": 7.37, "learning_rate": 3.2737428357450086e-07, "loss": 0.0006, "step": 40168 }, { "epoch": 7.37, "learning_rate": 3.27185898053366e-07, "loss": 0.0002, "step": 40169 }, { "epoch": 7.37, "learning_rate": 3.269975658490099e-07, "loss": 0.0008, "step": 40170 }, { "epoch": 7.37, "learning_rate": 3.268092869624684e-07, "loss": 0.0003, "step": 40171 }, { "epoch": 7.37, "learning_rate": 3.2662106139478176e-07, "loss": 0.0002, "step": 40172 }, { "epoch": 7.37, "learning_rate": 3.264328891469859e-07, "loss": 0.0004, "step": 40173 }, { "epoch": 7.37, "learning_rate": 3.262447702201188e-07, "loss": 0.0005, "step": 40174 }, { "epoch": 7.37, "learning_rate": 3.2605670461521634e-07, "loss": 0.0008, "step": 40175 }, { "epoch": 7.37, "learning_rate": 3.2586869233331653e-07, "loss": 0.0008, "step": 40176 }, { "epoch": 7.37, "learning_rate": 3.2568073337545635e-07, "loss": 0.0004, "step": 40177 }, { "epoch": 7.37, "learning_rate": 3.2549282774266946e-07, "loss": 0.0006, "step": 40178 }, { "epoch": 7.37, "learning_rate": 3.253049754359927e-07, "loss": 0.0005, "step": 40179 }, { "epoch": 7.37, "learning_rate": 3.251171764564609e-07, "loss": 0.0006, "step": 40180 }, { "epoch": 7.37, "learning_rate": 3.24929430805111e-07, "loss": 0.0007, "step": 40181 }, { "epoch": 7.37, "learning_rate": 3.2474173848297653e-07, "loss": 0.0004, "step": 40182 }, { "epoch": 7.37, "learning_rate": 3.245540994910934e-07, "loss": 0.0006, "step": 40183 }, { "epoch": 7.37, "learning_rate": 3.2436651383049187e-07, "loss": 0.0002, "step": 40184 }, { "epoch": 7.37, "learning_rate": 3.2417898150221117e-07, "loss": 0.0005, "step": 40185 }, { "epoch": 7.37, "learning_rate": 3.239915025072815e-07, "loss": 0.0002, "step": 40186 }, { "epoch": 7.37, "learning_rate": 3.2380407684673764e-07, "loss": 0.0002, "step": 40187 }, { "epoch": 7.37, "learning_rate": 3.236167045216121e-07, "loss": 0.0002, "step": 40188 }, { "epoch": 7.37, "learning_rate": 3.2342938553293844e-07, "loss": 0.0007, "step": 40189 }, { "epoch": 7.37, "learning_rate": 3.2324211988174927e-07, "loss": 0.0009, "step": 40190 }, { "epoch": 7.37, "learning_rate": 3.230549075690759e-07, "loss": 0.0003, "step": 40191 }, { "epoch": 7.37, "learning_rate": 3.2286774859594974e-07, "loss": 0.0005, "step": 40192 }, { "epoch": 7.37, "learning_rate": 3.226806429634033e-07, "loss": 0.0004, "step": 40193 }, { "epoch": 7.37, "learning_rate": 3.224935906724691e-07, "loss": 0.0003, "step": 40194 }, { "epoch": 7.37, "learning_rate": 3.223065917241763e-07, "loss": 0.0005, "step": 40195 }, { "epoch": 7.37, "learning_rate": 3.221196461195564e-07, "loss": 0.0002, "step": 40196 }, { "epoch": 7.37, "learning_rate": 3.219327538596384e-07, "loss": 0.0005, "step": 40197 }, { "epoch": 7.37, "learning_rate": 3.2174591494545493e-07, "loss": 0.0006, "step": 40198 }, { "epoch": 7.37, "learning_rate": 3.2155912937803514e-07, "loss": 0.0023, "step": 40199 }, { "epoch": 7.37, "learning_rate": 3.2137239715840706e-07, "loss": 0.0006, "step": 40200 }, { "epoch": 7.37, "learning_rate": 3.2118571828760327e-07, "loss": 0.0003, "step": 40201 }, { "epoch": 7.37, "learning_rate": 3.209990927666484e-07, "loss": 0.0024, "step": 40202 }, { "epoch": 7.37, "learning_rate": 3.20812520596574e-07, "loss": 0.0003, "step": 40203 }, { "epoch": 7.37, "learning_rate": 3.2062600177840685e-07, "loss": 0.0007, "step": 40204 }, { "epoch": 7.37, "learning_rate": 3.2043953631317734e-07, "loss": 0.0008, "step": 40205 }, { "epoch": 7.37, "learning_rate": 3.2025312420191135e-07, "loss": 0.0004, "step": 40206 }, { "epoch": 7.37, "learning_rate": 3.2006676544563684e-07, "loss": 0.0004, "step": 40207 }, { "epoch": 7.37, "learning_rate": 3.1988046004538085e-07, "loss": 0.0005, "step": 40208 }, { "epoch": 7.37, "learning_rate": 3.196942080021714e-07, "loss": 0.0004, "step": 40209 }, { "epoch": 7.37, "learning_rate": 3.1950800931703216e-07, "loss": 0.0004, "step": 40210 }, { "epoch": 7.37, "learning_rate": 3.193218639909934e-07, "loss": 0.0012, "step": 40211 }, { "epoch": 7.37, "learning_rate": 3.1913577202507985e-07, "loss": 0.0006, "step": 40212 }, { "epoch": 7.37, "learning_rate": 3.189497334203162e-07, "loss": 0.0005, "step": 40213 }, { "epoch": 7.37, "learning_rate": 3.187637481777284e-07, "loss": 0.0003, "step": 40214 }, { "epoch": 7.37, "learning_rate": 3.1857781629833994e-07, "loss": 0.0002, "step": 40215 }, { "epoch": 7.38, "learning_rate": 3.18391937783179e-07, "loss": 0.0005, "step": 40216 }, { "epoch": 7.38, "learning_rate": 3.1820611263326806e-07, "loss": 0.0004, "step": 40217 }, { "epoch": 7.38, "learning_rate": 3.1802034084963186e-07, "loss": 0.0002, "step": 40218 }, { "epoch": 7.38, "learning_rate": 3.178346224332951e-07, "loss": 0.0002, "step": 40219 }, { "epoch": 7.38, "learning_rate": 3.1764895738528036e-07, "loss": 0.0004, "step": 40220 }, { "epoch": 7.38, "learning_rate": 3.174633457066112e-07, "loss": 0.0007, "step": 40221 }, { "epoch": 7.38, "learning_rate": 3.1727778739831127e-07, "loss": 0.0003, "step": 40222 }, { "epoch": 7.38, "learning_rate": 3.17092282461402e-07, "loss": 0.0006, "step": 40223 }, { "epoch": 7.38, "learning_rate": 3.169068308969081e-07, "loss": 0.0005, "step": 40224 }, { "epoch": 7.38, "learning_rate": 3.1672143270584987e-07, "loss": 0.0003, "step": 40225 }, { "epoch": 7.38, "learning_rate": 3.165360878892509e-07, "loss": 0.0015, "step": 40226 }, { "epoch": 7.38, "learning_rate": 3.163507964481305e-07, "loss": 0.0008, "step": 40227 }, { "epoch": 7.38, "learning_rate": 3.1616555838351104e-07, "loss": 0.0002, "step": 40228 }, { "epoch": 7.38, "learning_rate": 3.159803736964151e-07, "loss": 0.0011, "step": 40229 }, { "epoch": 7.38, "learning_rate": 3.1579524238786296e-07, "loss": 0.0005, "step": 40230 }, { "epoch": 7.38, "learning_rate": 3.156101644588738e-07, "loss": 0.0007, "step": 40231 }, { "epoch": 7.38, "learning_rate": 3.15425139910468e-07, "loss": 0.0004, "step": 40232 }, { "epoch": 7.38, "learning_rate": 3.152401687436646e-07, "loss": 0.0002, "step": 40233 }, { "epoch": 7.38, "learning_rate": 3.150552509594851e-07, "loss": 0.0003, "step": 40234 }, { "epoch": 7.38, "learning_rate": 3.1487038655894866e-07, "loss": 0.0006, "step": 40235 }, { "epoch": 7.38, "learning_rate": 3.146855755430722e-07, "loss": 0.0005, "step": 40236 }, { "epoch": 7.38, "learning_rate": 3.1450081791287723e-07, "loss": 0.001, "step": 40237 }, { "epoch": 7.38, "learning_rate": 3.1431611366937955e-07, "loss": 0.0003, "step": 40238 }, { "epoch": 7.38, "learning_rate": 3.141314628135994e-07, "loss": 0.0006, "step": 40239 }, { "epoch": 7.38, "learning_rate": 3.139468653465527e-07, "loss": 0.0005, "step": 40240 }, { "epoch": 7.38, "learning_rate": 3.1376232126925867e-07, "loss": 0.0003, "step": 40241 }, { "epoch": 7.38, "learning_rate": 3.1357783058273304e-07, "loss": 0.0009, "step": 40242 }, { "epoch": 7.38, "learning_rate": 3.133933932879929e-07, "loss": 0.0009, "step": 40243 }, { "epoch": 7.38, "learning_rate": 3.1320900938605627e-07, "loss": 0.0004, "step": 40244 }, { "epoch": 7.38, "learning_rate": 3.1302467887793785e-07, "loss": 0.0004, "step": 40245 }, { "epoch": 7.38, "learning_rate": 3.128404017646525e-07, "loss": 0.0004, "step": 40246 }, { "epoch": 7.38, "learning_rate": 3.126561780472204e-07, "loss": 0.0003, "step": 40247 }, { "epoch": 7.38, "learning_rate": 3.124720077266552e-07, "loss": 0.0005, "step": 40248 }, { "epoch": 7.38, "learning_rate": 3.122878908039695e-07, "loss": 0.0003, "step": 40249 }, { "epoch": 7.38, "learning_rate": 3.121038272801791e-07, "loss": 0.0004, "step": 40250 }, { "epoch": 7.38, "learning_rate": 3.1191981715630095e-07, "loss": 0.0004, "step": 40251 }, { "epoch": 7.38, "learning_rate": 3.117358604333465e-07, "loss": 0.0006, "step": 40252 }, { "epoch": 7.38, "learning_rate": 3.115519571123327e-07, "loss": 0.0008, "step": 40253 }, { "epoch": 7.38, "learning_rate": 3.113681071942698e-07, "loss": 0.0008, "step": 40254 }, { "epoch": 7.38, "learning_rate": 3.1118431068017373e-07, "loss": 0.0003, "step": 40255 }, { "epoch": 7.38, "learning_rate": 3.11000567571057e-07, "loss": 0.0003, "step": 40256 }, { "epoch": 7.38, "learning_rate": 3.10816877867931e-07, "loss": 0.0003, "step": 40257 }, { "epoch": 7.38, "learning_rate": 3.1063324157181053e-07, "loss": 0.0003, "step": 40258 }, { "epoch": 7.38, "learning_rate": 3.104496586837069e-07, "loss": 0.0002, "step": 40259 }, { "epoch": 7.38, "learning_rate": 3.1026612920463053e-07, "loss": 0.0007, "step": 40260 }, { "epoch": 7.38, "learning_rate": 3.100826531355949e-07, "loss": 0.0006, "step": 40261 }, { "epoch": 7.38, "learning_rate": 3.098992304776105e-07, "loss": 0.0004, "step": 40262 }, { "epoch": 7.38, "learning_rate": 3.097158612316875e-07, "loss": 0.0004, "step": 40263 }, { "epoch": 7.38, "learning_rate": 3.095325453988385e-07, "loss": 0.0004, "step": 40264 }, { "epoch": 7.38, "learning_rate": 3.093492829800737e-07, "loss": 0.0005, "step": 40265 }, { "epoch": 7.38, "learning_rate": 3.091660739764035e-07, "loss": 0.001, "step": 40266 }, { "epoch": 7.38, "learning_rate": 3.08982918388836e-07, "loss": 0.0002, "step": 40267 }, { "epoch": 7.38, "learning_rate": 3.0879981621838027e-07, "loss": 0.0003, "step": 40268 }, { "epoch": 7.38, "learning_rate": 3.0861676746604895e-07, "loss": 0.0004, "step": 40269 }, { "epoch": 7.38, "learning_rate": 3.0843377213284784e-07, "loss": 0.0015, "step": 40270 }, { "epoch": 7.39, "learning_rate": 3.082508302197873e-07, "loss": 0.0003, "step": 40271 }, { "epoch": 7.39, "learning_rate": 3.0806794172787534e-07, "loss": 0.0002, "step": 40272 }, { "epoch": 7.39, "learning_rate": 3.078851066581201e-07, "loss": 0.0004, "step": 40273 }, { "epoch": 7.39, "learning_rate": 3.0770232501152965e-07, "loss": 0.0012, "step": 40274 }, { "epoch": 7.39, "learning_rate": 3.075195967891098e-07, "loss": 0.0005, "step": 40275 }, { "epoch": 7.39, "learning_rate": 3.073369219918698e-07, "loss": 0.0003, "step": 40276 }, { "epoch": 7.39, "learning_rate": 3.071543006208155e-07, "loss": 0.0004, "step": 40277 }, { "epoch": 7.39, "learning_rate": 3.069717326769539e-07, "loss": 0.0003, "step": 40278 }, { "epoch": 7.39, "learning_rate": 3.0678921816129194e-07, "loss": 0.0004, "step": 40279 }, { "epoch": 7.39, "learning_rate": 3.0660675707483325e-07, "loss": 0.0004, "step": 40280 }, { "epoch": 7.39, "learning_rate": 3.064243494185859e-07, "loss": 0.0004, "step": 40281 }, { "epoch": 7.39, "learning_rate": 3.062419951935547e-07, "loss": 0.0007, "step": 40282 }, { "epoch": 7.39, "learning_rate": 3.0605969440074434e-07, "loss": 0.0005, "step": 40283 }, { "epoch": 7.39, "learning_rate": 3.0587744704116187e-07, "loss": 0.0008, "step": 40284 }, { "epoch": 7.39, "learning_rate": 3.056952531158075e-07, "loss": 0.0002, "step": 40285 }, { "epoch": 7.39, "learning_rate": 3.0551311262568937e-07, "loss": 0.0003, "step": 40286 }, { "epoch": 7.39, "learning_rate": 3.0533102557181003e-07, "loss": 0.0002, "step": 40287 }, { "epoch": 7.39, "learning_rate": 3.051489919551731e-07, "loss": 0.0004, "step": 40288 }, { "epoch": 7.39, "learning_rate": 3.049670117767822e-07, "loss": 0.0005, "step": 40289 }, { "epoch": 7.39, "learning_rate": 3.0478508503763995e-07, "loss": 0.0003, "step": 40290 }, { "epoch": 7.39, "learning_rate": 3.0460321173874986e-07, "loss": 0.0007, "step": 40291 }, { "epoch": 7.39, "learning_rate": 3.044213918811134e-07, "loss": 0.0004, "step": 40292 }, { "epoch": 7.39, "learning_rate": 3.042396254657343e-07, "loss": 0.0005, "step": 40293 }, { "epoch": 7.39, "learning_rate": 3.040579124936116e-07, "loss": 0.0003, "step": 40294 }, { "epoch": 7.39, "learning_rate": 3.038762529657513e-07, "loss": 0.0003, "step": 40295 }, { "epoch": 7.39, "learning_rate": 3.0369464688315144e-07, "loss": 0.001, "step": 40296 }, { "epoch": 7.39, "learning_rate": 3.0351309424681343e-07, "loss": 0.001, "step": 40297 }, { "epoch": 7.39, "learning_rate": 3.0333159505773866e-07, "loss": 0.0006, "step": 40298 }, { "epoch": 7.39, "learning_rate": 3.0315014931692534e-07, "loss": 0.0006, "step": 40299 }, { "epoch": 7.39, "learning_rate": 3.0296875702537807e-07, "loss": 0.0008, "step": 40300 }, { "epoch": 7.39, "learning_rate": 3.0278741818409396e-07, "loss": 0.0006, "step": 40301 }, { "epoch": 7.39, "learning_rate": 3.0260613279407325e-07, "loss": 0.0009, "step": 40302 }, { "epoch": 7.39, "learning_rate": 3.0242490085631184e-07, "loss": 0.0008, "step": 40303 }, { "epoch": 7.39, "learning_rate": 3.022437223718144e-07, "loss": 0.0004, "step": 40304 }, { "epoch": 7.39, "learning_rate": 3.020625973415758e-07, "loss": 0.0002, "step": 40305 }, { "epoch": 7.39, "learning_rate": 3.0188152576659523e-07, "loss": 0.0006, "step": 40306 }, { "epoch": 7.39, "learning_rate": 3.017005076478707e-07, "loss": 0.0008, "step": 40307 }, { "epoch": 7.39, "learning_rate": 3.015195429864004e-07, "loss": 0.0008, "step": 40308 }, { "epoch": 7.39, "learning_rate": 3.0133863178318237e-07, "loss": 0.0017, "step": 40309 }, { "epoch": 7.39, "learning_rate": 3.0115777403921244e-07, "loss": 0.0018, "step": 40310 }, { "epoch": 7.39, "learning_rate": 3.0097696975548764e-07, "loss": 0.0004, "step": 40311 }, { "epoch": 7.39, "learning_rate": 3.007962189330049e-07, "loss": 0.0005, "step": 40312 }, { "epoch": 7.39, "learning_rate": 3.006155215727624e-07, "loss": 0.0005, "step": 40313 }, { "epoch": 7.39, "learning_rate": 3.004348776757526e-07, "loss": 0.0006, "step": 40314 }, { "epoch": 7.39, "learning_rate": 3.0025428724297367e-07, "loss": 0.0004, "step": 40315 }, { "epoch": 7.39, "learning_rate": 3.000737502754192e-07, "loss": 0.0007, "step": 40316 }, { "epoch": 7.39, "learning_rate": 2.9989326677408727e-07, "loss": 0.0008, "step": 40317 }, { "epoch": 7.39, "learning_rate": 2.997128367399693e-07, "loss": 0.0003, "step": 40318 }, { "epoch": 7.39, "learning_rate": 2.9953246017406343e-07, "loss": 0.0006, "step": 40319 }, { "epoch": 7.39, "learning_rate": 2.993521370773611e-07, "loss": 0.0002, "step": 40320 }, { "epoch": 7.39, "learning_rate": 2.9917186745085594e-07, "loss": 0.0005, "step": 40321 }, { "epoch": 7.39, "learning_rate": 2.989916512955426e-07, "loss": 0.0003, "step": 40322 }, { "epoch": 7.39, "learning_rate": 2.9881148861241606e-07, "loss": 0.0003, "step": 40323 }, { "epoch": 7.39, "learning_rate": 2.9863137940246643e-07, "loss": 0.0004, "step": 40324 }, { "epoch": 7.4, "learning_rate": 2.9845132366668974e-07, "loss": 0.0004, "step": 40325 }, { "epoch": 7.4, "learning_rate": 2.9827132140607507e-07, "loss": 0.0003, "step": 40326 }, { "epoch": 7.4, "learning_rate": 2.9809137262161614e-07, "loss": 0.0004, "step": 40327 }, { "epoch": 7.4, "learning_rate": 2.9791147731430545e-07, "loss": 0.0002, "step": 40328 }, { "epoch": 7.4, "learning_rate": 2.9773163548513453e-07, "loss": 0.0004, "step": 40329 }, { "epoch": 7.4, "learning_rate": 2.9755184713509243e-07, "loss": 0.0003, "step": 40330 }, { "epoch": 7.4, "learning_rate": 2.9737211226517405e-07, "loss": 0.0008, "step": 40331 }, { "epoch": 7.4, "learning_rate": 2.971924308763663e-07, "loss": 0.0011, "step": 40332 }, { "epoch": 7.4, "learning_rate": 2.970128029696617e-07, "loss": 0.0002, "step": 40333 }, { "epoch": 7.4, "learning_rate": 2.968332285460496e-07, "loss": 0.001, "step": 40334 }, { "epoch": 7.4, "learning_rate": 2.966537076065201e-07, "loss": 0.0007, "step": 40335 }, { "epoch": 7.4, "learning_rate": 2.964742401520626e-07, "loss": 0.0006, "step": 40336 }, { "epoch": 7.4, "learning_rate": 2.9629482618366845e-07, "loss": 0.0005, "step": 40337 }, { "epoch": 7.4, "learning_rate": 2.961154657023213e-07, "loss": 0.0002, "step": 40338 }, { "epoch": 7.4, "learning_rate": 2.9593615870901483e-07, "loss": 0.0009, "step": 40339 }, { "epoch": 7.4, "learning_rate": 2.957569052047349e-07, "loss": 0.001, "step": 40340 }, { "epoch": 7.4, "learning_rate": 2.955777051904707e-07, "loss": 0.0006, "step": 40341 }, { "epoch": 7.4, "learning_rate": 2.9539855866720925e-07, "loss": 0.0003, "step": 40342 }, { "epoch": 7.4, "learning_rate": 2.952194656359386e-07, "loss": 0.0004, "step": 40343 }, { "epoch": 7.4, "learning_rate": 2.9504042609764584e-07, "loss": 0.0005, "step": 40344 }, { "epoch": 7.4, "learning_rate": 2.9486144005331783e-07, "loss": 0.0006, "step": 40345 }, { "epoch": 7.4, "learning_rate": 2.946825075039406e-07, "loss": 0.0009, "step": 40346 }, { "epoch": 7.4, "learning_rate": 2.9450362845049874e-07, "loss": 0.0004, "step": 40347 }, { "epoch": 7.4, "learning_rate": 2.943248028939838e-07, "loss": 0.0003, "step": 40348 }, { "epoch": 7.4, "learning_rate": 2.9414603083537605e-07, "loss": 0.0002, "step": 40349 }, { "epoch": 7.4, "learning_rate": 2.9396731227566255e-07, "loss": 0.0004, "step": 40350 }, { "epoch": 7.4, "learning_rate": 2.9378864721582913e-07, "loss": 0.0003, "step": 40351 }, { "epoch": 7.4, "learning_rate": 2.936100356568594e-07, "loss": 0.0003, "step": 40352 }, { "epoch": 7.4, "learning_rate": 2.934314775997393e-07, "loss": 0.0004, "step": 40353 }, { "epoch": 7.4, "learning_rate": 2.932529730454514e-07, "loss": 0.0006, "step": 40354 }, { "epoch": 7.4, "learning_rate": 2.9307452199498266e-07, "loss": 0.0004, "step": 40355 }, { "epoch": 7.4, "learning_rate": 2.928961244493123e-07, "loss": 0.0006, "step": 40356 }, { "epoch": 7.4, "learning_rate": 2.9271778040942723e-07, "loss": 0.0009, "step": 40357 }, { "epoch": 7.4, "learning_rate": 2.925394898763079e-07, "loss": 0.0005, "step": 40358 }, { "epoch": 7.4, "learning_rate": 2.92361252850939e-07, "loss": 0.0005, "step": 40359 }, { "epoch": 7.4, "learning_rate": 2.921830693343031e-07, "loss": 0.0008, "step": 40360 }, { "epoch": 7.4, "learning_rate": 2.920049393273805e-07, "loss": 0.0004, "step": 40361 }, { "epoch": 7.4, "learning_rate": 2.9182686283115493e-07, "loss": 0.0003, "step": 40362 }, { "epoch": 7.4, "learning_rate": 2.9164883984660664e-07, "loss": 0.0005, "step": 40363 }, { "epoch": 7.4, "learning_rate": 2.914708703747171e-07, "loss": 0.0005, "step": 40364 }, { "epoch": 7.4, "learning_rate": 2.9129295441646665e-07, "loss": 0.0002, "step": 40365 }, { "epoch": 7.4, "learning_rate": 2.9111509197284006e-07, "loss": 0.0006, "step": 40366 }, { "epoch": 7.4, "learning_rate": 2.9093728304481204e-07, "loss": 0.0004, "step": 40367 }, { "epoch": 7.4, "learning_rate": 2.9075952763336637e-07, "loss": 0.0003, "step": 40368 }, { "epoch": 7.4, "learning_rate": 2.905818257394799e-07, "loss": 0.0005, "step": 40369 }, { "epoch": 7.4, "learning_rate": 2.904041773641342e-07, "loss": 0.0003, "step": 40370 }, { "epoch": 7.4, "learning_rate": 2.9022658250830946e-07, "loss": 0.0009, "step": 40371 }, { "epoch": 7.4, "learning_rate": 2.900490411729828e-07, "loss": 0.0005, "step": 40372 }, { "epoch": 7.4, "learning_rate": 2.898715533591345e-07, "loss": 0.0005, "step": 40373 }, { "epoch": 7.4, "learning_rate": 2.8969411906773935e-07, "loss": 0.0005, "step": 40374 }, { "epoch": 7.4, "learning_rate": 2.8951673829977876e-07, "loss": 0.0007, "step": 40375 }, { "epoch": 7.4, "learning_rate": 2.8933941105622976e-07, "loss": 0.0006, "step": 40376 }, { "epoch": 7.4, "learning_rate": 2.891621373380682e-07, "loss": 0.0002, "step": 40377 }, { "epoch": 7.4, "learning_rate": 2.889849171462744e-07, "loss": 0.0005, "step": 40378 }, { "epoch": 7.4, "learning_rate": 2.888077504818221e-07, "loss": 0.0003, "step": 40379 }, { "epoch": 7.41, "learning_rate": 2.8863063734568817e-07, "loss": 0.0004, "step": 40380 }, { "epoch": 7.41, "learning_rate": 2.884535777388509e-07, "loss": 0.0002, "step": 40381 }, { "epoch": 7.41, "learning_rate": 2.8827657166228483e-07, "loss": 0.0013, "step": 40382 }, { "epoch": 7.41, "learning_rate": 2.880996191169649e-07, "loss": 0.0008, "step": 40383 }, { "epoch": 7.41, "learning_rate": 2.8792272010386924e-07, "loss": 0.0005, "step": 40384 }, { "epoch": 7.41, "learning_rate": 2.877458746239703e-07, "loss": 0.0004, "step": 40385 }, { "epoch": 7.41, "learning_rate": 2.875690826782429e-07, "loss": 0.0002, "step": 40386 }, { "epoch": 7.41, "learning_rate": 2.8739234426766184e-07, "loss": 0.0004, "step": 40387 }, { "epoch": 7.41, "learning_rate": 2.8721565939320294e-07, "loss": 0.0003, "step": 40388 }, { "epoch": 7.41, "learning_rate": 2.870390280558377e-07, "loss": 0.0006, "step": 40389 }, { "epoch": 7.41, "learning_rate": 2.8686245025654204e-07, "loss": 0.0012, "step": 40390 }, { "epoch": 7.41, "learning_rate": 2.8668592599628843e-07, "loss": 0.0003, "step": 40391 }, { "epoch": 7.41, "learning_rate": 2.865094552760483e-07, "loss": 0.0007, "step": 40392 }, { "epoch": 7.41, "learning_rate": 2.863330380967966e-07, "loss": 0.0003, "step": 40393 }, { "epoch": 7.41, "learning_rate": 2.8615667445950454e-07, "loss": 0.0003, "step": 40394 }, { "epoch": 7.41, "learning_rate": 2.8598036436514375e-07, "loss": 0.0006, "step": 40395 }, { "epoch": 7.41, "learning_rate": 2.8580410781468783e-07, "loss": 0.0002, "step": 40396 }, { "epoch": 7.41, "learning_rate": 2.8562790480910706e-07, "loss": 0.0003, "step": 40397 }, { "epoch": 7.41, "learning_rate": 2.854517553493719e-07, "loss": 0.0005, "step": 40398 }, { "epoch": 7.41, "learning_rate": 2.8527565943645474e-07, "loss": 0.0005, "step": 40399 }, { "epoch": 7.41, "learning_rate": 2.8509961707132496e-07, "loss": 0.0002, "step": 40400 }, { "epoch": 7.41, "learning_rate": 2.8492362825495503e-07, "loss": 0.0003, "step": 40401 }, { "epoch": 7.41, "learning_rate": 2.847476929883142e-07, "loss": 0.0005, "step": 40402 }, { "epoch": 7.41, "learning_rate": 2.8457181127237056e-07, "loss": 0.0005, "step": 40403 }, { "epoch": 7.41, "learning_rate": 2.8439598310809556e-07, "loss": 0.0007, "step": 40404 }, { "epoch": 7.41, "learning_rate": 2.8422020849645517e-07, "loss": 0.0004, "step": 40405 }, { "epoch": 7.41, "learning_rate": 2.8404448743842295e-07, "loss": 0.0003, "step": 40406 }, { "epoch": 7.41, "learning_rate": 2.838688199349637e-07, "loss": 0.0006, "step": 40407 }, { "epoch": 7.41, "learning_rate": 2.8369320598704786e-07, "loss": 0.0005, "step": 40408 }, { "epoch": 7.41, "learning_rate": 2.8351764559564344e-07, "loss": 0.0009, "step": 40409 }, { "epoch": 7.41, "learning_rate": 2.8334213876171637e-07, "loss": 0.0004, "step": 40410 }, { "epoch": 7.41, "learning_rate": 2.8316668548623583e-07, "loss": 0.0005, "step": 40411 }, { "epoch": 7.41, "learning_rate": 2.8299128577016775e-07, "loss": 0.0004, "step": 40412 }, { "epoch": 7.41, "learning_rate": 2.828159396144792e-07, "loss": 0.0002, "step": 40413 }, { "epoch": 7.41, "learning_rate": 2.8264064702013706e-07, "loss": 0.0006, "step": 40414 }, { "epoch": 7.41, "learning_rate": 2.8246540798810727e-07, "loss": 0.0005, "step": 40415 }, { "epoch": 7.41, "learning_rate": 2.822902225193558e-07, "loss": 0.0005, "step": 40416 }, { "epoch": 7.41, "learning_rate": 2.821150906148484e-07, "loss": 0.0004, "step": 40417 }, { "epoch": 7.41, "learning_rate": 2.819400122755489e-07, "loss": 0.0007, "step": 40418 }, { "epoch": 7.41, "learning_rate": 2.817649875024242e-07, "loss": 0.0007, "step": 40419 }, { "epoch": 7.41, "learning_rate": 2.8159001629644134e-07, "loss": 0.0004, "step": 40420 }, { "epoch": 7.41, "learning_rate": 2.8141509865855953e-07, "loss": 0.0003, "step": 40421 }, { "epoch": 7.41, "learning_rate": 2.812402345897447e-07, "loss": 0.0006, "step": 40422 }, { "epoch": 7.41, "learning_rate": 2.810654240909627e-07, "loss": 0.0005, "step": 40423 }, { "epoch": 7.41, "learning_rate": 2.8089066716317503e-07, "loss": 0.0002, "step": 40424 }, { "epoch": 7.41, "learning_rate": 2.8071596380734644e-07, "loss": 0.0006, "step": 40425 }, { "epoch": 7.41, "learning_rate": 2.805413140244406e-07, "loss": 0.0005, "step": 40426 }, { "epoch": 7.41, "learning_rate": 2.803667178154157e-07, "loss": 0.0008, "step": 40427 }, { "epoch": 7.41, "learning_rate": 2.8019217518123866e-07, "loss": 0.0009, "step": 40428 }, { "epoch": 7.41, "learning_rate": 2.800176861228698e-07, "loss": 0.0007, "step": 40429 }, { "epoch": 7.41, "learning_rate": 2.7984325064127184e-07, "loss": 0.0009, "step": 40430 }, { "epoch": 7.41, "learning_rate": 2.7966886873740495e-07, "loss": 0.0017, "step": 40431 }, { "epoch": 7.41, "learning_rate": 2.794945404122307e-07, "loss": 0.0008, "step": 40432 }, { "epoch": 7.41, "learning_rate": 2.7932026566671154e-07, "loss": 0.0004, "step": 40433 }, { "epoch": 7.42, "learning_rate": 2.791460445018057e-07, "loss": 0.0004, "step": 40434 }, { "epoch": 7.42, "learning_rate": 2.789718769184746e-07, "loss": 0.0007, "step": 40435 }, { "epoch": 7.42, "learning_rate": 2.7879776291767747e-07, "loss": 0.0002, "step": 40436 }, { "epoch": 7.42, "learning_rate": 2.7862370250037574e-07, "loss": 0.0011, "step": 40437 }, { "epoch": 7.42, "learning_rate": 2.784496956675298e-07, "loss": 0.0005, "step": 40438 }, { "epoch": 7.42, "learning_rate": 2.7827574242009434e-07, "loss": 0.0006, "step": 40439 }, { "epoch": 7.42, "learning_rate": 2.78101842759031e-07, "loss": 0.0004, "step": 40440 }, { "epoch": 7.42, "learning_rate": 2.779279966852988e-07, "loss": 0.0003, "step": 40441 }, { "epoch": 7.42, "learning_rate": 2.777542041998549e-07, "loss": 0.0001, "step": 40442 }, { "epoch": 7.42, "learning_rate": 2.7758046530365736e-07, "loss": 0.0004, "step": 40443 }, { "epoch": 7.42, "learning_rate": 2.774067799976643e-07, "loss": 0.0006, "step": 40444 }, { "epoch": 7.42, "learning_rate": 2.7723314828283276e-07, "loss": 0.0004, "step": 40445 }, { "epoch": 7.42, "learning_rate": 2.770595701601209e-07, "loss": 0.0003, "step": 40446 }, { "epoch": 7.42, "learning_rate": 2.768860456304834e-07, "loss": 0.0004, "step": 40447 }, { "epoch": 7.42, "learning_rate": 2.767125746948784e-07, "loss": 0.0003, "step": 40448 }, { "epoch": 7.42, "learning_rate": 2.7653915735426086e-07, "loss": 0.0013, "step": 40449 }, { "epoch": 7.42, "learning_rate": 2.763657936095876e-07, "loss": 0.0002, "step": 40450 }, { "epoch": 7.42, "learning_rate": 2.761924834618135e-07, "loss": 0.0006, "step": 40451 }, { "epoch": 7.42, "learning_rate": 2.7601922691189556e-07, "loss": 0.0005, "step": 40452 }, { "epoch": 7.42, "learning_rate": 2.758460239607852e-07, "loss": 0.0005, "step": 40453 }, { "epoch": 7.42, "learning_rate": 2.756728746094406e-07, "loss": 0.0005, "step": 40454 }, { "epoch": 7.42, "learning_rate": 2.7549977885881543e-07, "loss": 0.0007, "step": 40455 }, { "epoch": 7.42, "learning_rate": 2.753267367098633e-07, "loss": 0.0006, "step": 40456 }, { "epoch": 7.42, "learning_rate": 2.75153748163538e-07, "loss": 0.0006, "step": 40457 }, { "epoch": 7.42, "learning_rate": 2.7498081322079097e-07, "loss": 0.0004, "step": 40458 }, { "epoch": 7.42, "learning_rate": 2.7480793188257913e-07, "loss": 0.0005, "step": 40459 }, { "epoch": 7.42, "learning_rate": 2.7463510414985406e-07, "loss": 0.0005, "step": 40460 }, { "epoch": 7.42, "learning_rate": 2.7446233002356823e-07, "loss": 0.0004, "step": 40461 }, { "epoch": 7.42, "learning_rate": 2.742896095046732e-07, "loss": 0.0002, "step": 40462 }, { "epoch": 7.42, "learning_rate": 2.7411694259412256e-07, "loss": 0.0003, "step": 40463 }, { "epoch": 7.42, "learning_rate": 2.7394432929286673e-07, "loss": 0.0003, "step": 40464 }, { "epoch": 7.42, "learning_rate": 2.7377176960185826e-07, "loss": 0.0003, "step": 40465 }, { "epoch": 7.42, "learning_rate": 2.7359926352204524e-07, "loss": 0.0002, "step": 40466 }, { "epoch": 7.42, "learning_rate": 2.734268110543836e-07, "loss": 0.0004, "step": 40467 }, { "epoch": 7.42, "learning_rate": 2.732544121998204e-07, "loss": 0.0014, "step": 40468 }, { "epoch": 7.42, "learning_rate": 2.7308206695930704e-07, "loss": 0.0004, "step": 40469 }, { "epoch": 7.42, "learning_rate": 2.729097753337928e-07, "loss": 0.0004, "step": 40470 }, { "epoch": 7.42, "learning_rate": 2.727375373242269e-07, "loss": 0.0003, "step": 40471 }, { "epoch": 7.42, "learning_rate": 2.7256535293156084e-07, "loss": 0.0003, "step": 40472 }, { "epoch": 7.42, "learning_rate": 2.723932221567427e-07, "loss": 0.0002, "step": 40473 }, { "epoch": 7.42, "learning_rate": 2.722211450007206e-07, "loss": 0.0004, "step": 40474 }, { "epoch": 7.42, "learning_rate": 2.720491214644427e-07, "loss": 0.0005, "step": 40475 }, { "epoch": 7.42, "learning_rate": 2.718771515488583e-07, "loss": 0.0006, "step": 40476 }, { "epoch": 7.42, "learning_rate": 2.7170523525491543e-07, "loss": 0.0006, "step": 40477 }, { "epoch": 7.42, "learning_rate": 2.715333725835612e-07, "loss": 0.0002, "step": 40478 }, { "epoch": 7.42, "learning_rate": 2.713615635357425e-07, "loss": 0.0009, "step": 40479 }, { "epoch": 7.42, "learning_rate": 2.711898081124065e-07, "loss": 0.0007, "step": 40480 }, { "epoch": 7.42, "learning_rate": 2.7101810631450123e-07, "loss": 0.0003, "step": 40481 }, { "epoch": 7.42, "learning_rate": 2.708464581429715e-07, "loss": 0.0004, "step": 40482 }, { "epoch": 7.42, "learning_rate": 2.706748635987644e-07, "loss": 0.0009, "step": 40483 }, { "epoch": 7.42, "learning_rate": 2.705033226828235e-07, "loss": 0.0006, "step": 40484 }, { "epoch": 7.42, "learning_rate": 2.703318353960993e-07, "loss": 0.0006, "step": 40485 }, { "epoch": 7.42, "learning_rate": 2.701604017395321e-07, "loss": 0.0016, "step": 40486 }, { "epoch": 7.42, "learning_rate": 2.6998902171406883e-07, "loss": 0.0003, "step": 40487 }, { "epoch": 7.42, "learning_rate": 2.698176953206544e-07, "loss": 0.0012, "step": 40488 }, { "epoch": 7.43, "learning_rate": 2.696464225602313e-07, "loss": 0.0006, "step": 40489 }, { "epoch": 7.43, "learning_rate": 2.694752034337467e-07, "loss": 0.0003, "step": 40490 }, { "epoch": 7.43, "learning_rate": 2.69304037942143e-07, "loss": 0.0011, "step": 40491 }, { "epoch": 7.43, "learning_rate": 2.691329260863629e-07, "loss": 0.0006, "step": 40492 }, { "epoch": 7.43, "learning_rate": 2.689618678673489e-07, "loss": 0.0006, "step": 40493 }, { "epoch": 7.43, "learning_rate": 2.6879086328604696e-07, "loss": 0.0002, "step": 40494 }, { "epoch": 7.43, "learning_rate": 2.6861991234339635e-07, "loss": 0.0003, "step": 40495 }, { "epoch": 7.43, "learning_rate": 2.684490150403418e-07, "loss": 0.0003, "step": 40496 }, { "epoch": 7.43, "learning_rate": 2.6827817137782376e-07, "loss": 0.0003, "step": 40497 }, { "epoch": 7.43, "learning_rate": 2.681073813567847e-07, "loss": 0.0005, "step": 40498 }, { "epoch": 7.43, "learning_rate": 2.679366449781662e-07, "loss": 0.0005, "step": 40499 }, { "epoch": 7.43, "learning_rate": 2.677659622429074e-07, "loss": 0.0007, "step": 40500 }, { "epoch": 7.43, "learning_rate": 2.675953331519521e-07, "loss": 0.0004, "step": 40501 }, { "epoch": 7.43, "learning_rate": 2.6742475770623833e-07, "loss": 0.0002, "step": 40502 }, { "epoch": 7.43, "learning_rate": 2.672542359067076e-07, "loss": 0.0015, "step": 40503 }, { "epoch": 7.43, "learning_rate": 2.6708376775430035e-07, "loss": 0.0005, "step": 40504 }, { "epoch": 7.43, "learning_rate": 2.669133532499546e-07, "loss": 0.0001, "step": 40505 }, { "epoch": 7.43, "learning_rate": 2.6674299239460964e-07, "loss": 0.0004, "step": 40506 }, { "epoch": 7.43, "learning_rate": 2.6657268518920586e-07, "loss": 0.0006, "step": 40507 }, { "epoch": 7.43, "learning_rate": 2.664024316346825e-07, "loss": 0.0002, "step": 40508 }, { "epoch": 7.43, "learning_rate": 2.6623223173197656e-07, "loss": 0.0003, "step": 40509 }, { "epoch": 7.43, "learning_rate": 2.6606208548202727e-07, "loss": 0.0003, "step": 40510 }, { "epoch": 7.43, "learning_rate": 2.658919928857695e-07, "loss": 0.0005, "step": 40511 }, { "epoch": 7.43, "learning_rate": 2.657219539441447e-07, "loss": 0.0008, "step": 40512 }, { "epoch": 7.43, "learning_rate": 2.655519686580876e-07, "loss": 0.0007, "step": 40513 }, { "epoch": 7.43, "learning_rate": 2.653820370285376e-07, "loss": 0.0008, "step": 40514 }, { "epoch": 7.43, "learning_rate": 2.6521215905642936e-07, "loss": 0.0001, "step": 40515 }, { "epoch": 7.43, "learning_rate": 2.650423347426989e-07, "loss": 0.0004, "step": 40516 }, { "epoch": 7.43, "learning_rate": 2.6487256408828434e-07, "loss": 0.0007, "step": 40517 }, { "epoch": 7.43, "learning_rate": 2.647028470941193e-07, "loss": 0.0005, "step": 40518 }, { "epoch": 7.43, "learning_rate": 2.645331837611398e-07, "loss": 0.0011, "step": 40519 }, { "epoch": 7.43, "learning_rate": 2.643635740902839e-07, "loss": 0.0006, "step": 40520 }, { "epoch": 7.43, "learning_rate": 2.64194018082482e-07, "loss": 0.0003, "step": 40521 }, { "epoch": 7.43, "learning_rate": 2.6402451573867115e-07, "loss": 0.0008, "step": 40522 }, { "epoch": 7.43, "learning_rate": 2.6385506705978505e-07, "loss": 0.0002, "step": 40523 }, { "epoch": 7.43, "learning_rate": 2.636856720467573e-07, "loss": 0.0006, "step": 40524 }, { "epoch": 7.43, "learning_rate": 2.63516330700524e-07, "loss": 0.0004, "step": 40525 }, { "epoch": 7.43, "learning_rate": 2.633470430220153e-07, "loss": 0.0002, "step": 40526 }, { "epoch": 7.43, "learning_rate": 2.6317780901216726e-07, "loss": 0.0004, "step": 40527 }, { "epoch": 7.43, "learning_rate": 2.6300862867191023e-07, "loss": 0.0005, "step": 40528 }, { "epoch": 7.43, "learning_rate": 2.628395020021779e-07, "loss": 0.0004, "step": 40529 }, { "epoch": 7.43, "learning_rate": 2.6267042900390173e-07, "loss": 0.0003, "step": 40530 }, { "epoch": 7.43, "learning_rate": 2.625014096780154e-07, "loss": 0.001, "step": 40531 }, { "epoch": 7.43, "learning_rate": 2.6233244402544934e-07, "loss": 0.0002, "step": 40532 }, { "epoch": 7.43, "learning_rate": 2.6216353204713385e-07, "loss": 0.0004, "step": 40533 }, { "epoch": 7.43, "learning_rate": 2.6199467374400265e-07, "loss": 0.0003, "step": 40534 }, { "epoch": 7.43, "learning_rate": 2.6182586911698394e-07, "loss": 0.0007, "step": 40535 }, { "epoch": 7.43, "learning_rate": 2.616571181670091e-07, "loss": 0.0004, "step": 40536 }, { "epoch": 7.43, "learning_rate": 2.614884208950075e-07, "loss": 0.0007, "step": 40537 }, { "epoch": 7.43, "learning_rate": 2.613197773019127e-07, "loss": 0.0002, "step": 40538 }, { "epoch": 7.43, "learning_rate": 2.6115118738864963e-07, "loss": 0.0004, "step": 40539 }, { "epoch": 7.43, "learning_rate": 2.609826511561497e-07, "loss": 0.0007, "step": 40540 }, { "epoch": 7.43, "learning_rate": 2.6081416860534003e-07, "loss": 0.0004, "step": 40541 }, { "epoch": 7.43, "learning_rate": 2.60645739737152e-07, "loss": 0.0005, "step": 40542 }, { "epoch": 7.43, "learning_rate": 2.604773645525138e-07, "loss": 0.0002, "step": 40543 }, { "epoch": 7.44, "learning_rate": 2.603090430523514e-07, "loss": 0.0005, "step": 40544 }, { "epoch": 7.44, "learning_rate": 2.601407752375951e-07, "loss": 0.0003, "step": 40545 }, { "epoch": 7.44, "learning_rate": 2.5997256110916856e-07, "loss": 0.0008, "step": 40546 }, { "epoch": 7.44, "learning_rate": 2.598044006680034e-07, "loss": 0.0003, "step": 40547 }, { "epoch": 7.44, "learning_rate": 2.596362939150243e-07, "loss": 0.0005, "step": 40548 }, { "epoch": 7.44, "learning_rate": 2.594682408511573e-07, "loss": 0.0004, "step": 40549 }, { "epoch": 7.44, "learning_rate": 2.5930024147733046e-07, "loss": 0.0005, "step": 40550 }, { "epoch": 7.44, "learning_rate": 2.5913229579446866e-07, "loss": 0.0008, "step": 40551 }, { "epoch": 7.44, "learning_rate": 2.5896440380349775e-07, "loss": 0.0003, "step": 40552 }, { "epoch": 7.44, "learning_rate": 2.587965655053437e-07, "loss": 0.0003, "step": 40553 }, { "epoch": 7.44, "learning_rate": 2.5862878090093024e-07, "loss": 0.0003, "step": 40554 }, { "epoch": 7.44, "learning_rate": 2.584610499911833e-07, "loss": 0.0002, "step": 40555 }, { "epoch": 7.44, "learning_rate": 2.582933727770287e-07, "loss": 0.0007, "step": 40556 }, { "epoch": 7.44, "learning_rate": 2.5812574925938805e-07, "loss": 0.0006, "step": 40557 }, { "epoch": 7.44, "learning_rate": 2.5795817943918613e-07, "loss": 0.0003, "step": 40558 }, { "epoch": 7.44, "learning_rate": 2.577906633173466e-07, "loss": 0.0003, "step": 40559 }, { "epoch": 7.44, "learning_rate": 2.576232008947943e-07, "loss": 0.0004, "step": 40560 }, { "epoch": 7.44, "learning_rate": 2.5745579217245073e-07, "loss": 0.0012, "step": 40561 }, { "epoch": 7.44, "learning_rate": 2.5728843715123963e-07, "loss": 0.0013, "step": 40562 }, { "epoch": 7.44, "learning_rate": 2.5712113583208356e-07, "loss": 0.0003, "step": 40563 }, { "epoch": 7.44, "learning_rate": 2.569538882159017e-07, "loss": 0.0002, "step": 40564 }, { "epoch": 7.44, "learning_rate": 2.567866943036201e-07, "loss": 0.0003, "step": 40565 }, { "epoch": 7.44, "learning_rate": 2.5661955409615804e-07, "loss": 0.0012, "step": 40566 }, { "epoch": 7.44, "learning_rate": 2.5645246759443687e-07, "loss": 0.0008, "step": 40567 }, { "epoch": 7.44, "learning_rate": 2.5628543479937817e-07, "loss": 0.0003, "step": 40568 }, { "epoch": 7.44, "learning_rate": 2.561184557119023e-07, "loss": 0.0006, "step": 40569 }, { "epoch": 7.44, "learning_rate": 2.5595153033293075e-07, "loss": 0.0007, "step": 40570 }, { "epoch": 7.44, "learning_rate": 2.5578465866338163e-07, "loss": 0.0005, "step": 40571 }, { "epoch": 7.44, "learning_rate": 2.556178407041754e-07, "loss": 0.0008, "step": 40572 }, { "epoch": 7.44, "learning_rate": 2.5545107645623234e-07, "loss": 0.0005, "step": 40573 }, { "epoch": 7.44, "learning_rate": 2.5528436592047287e-07, "loss": 0.0007, "step": 40574 }, { "epoch": 7.44, "learning_rate": 2.5511770909781295e-07, "loss": 0.0004, "step": 40575 }, { "epoch": 7.44, "learning_rate": 2.5495110598917183e-07, "loss": 0.0005, "step": 40576 }, { "epoch": 7.44, "learning_rate": 2.5478455659546873e-07, "loss": 0.0005, "step": 40577 }, { "epoch": 7.44, "learning_rate": 2.5461806091762186e-07, "loss": 0.0004, "step": 40578 }, { "epoch": 7.44, "learning_rate": 2.544516189565482e-07, "loss": 0.0005, "step": 40579 }, { "epoch": 7.44, "learning_rate": 2.542852307131671e-07, "loss": 0.0003, "step": 40580 }, { "epoch": 7.44, "learning_rate": 2.5411889618839227e-07, "loss": 0.0002, "step": 40581 }, { "epoch": 7.44, "learning_rate": 2.539526153831429e-07, "loss": 0.0003, "step": 40582 }, { "epoch": 7.44, "learning_rate": 2.5378638829833493e-07, "loss": 0.0009, "step": 40583 }, { "epoch": 7.44, "learning_rate": 2.536202149348854e-07, "loss": 0.0007, "step": 40584 }, { "epoch": 7.44, "learning_rate": 2.534540952937092e-07, "loss": 0.0004, "step": 40585 }, { "epoch": 7.44, "learning_rate": 2.532880293757223e-07, "loss": 0.0003, "step": 40586 }, { "epoch": 7.44, "learning_rate": 2.531220171818394e-07, "loss": 0.0003, "step": 40587 }, { "epoch": 7.44, "learning_rate": 2.5295605871297755e-07, "loss": 0.0007, "step": 40588 }, { "epoch": 7.44, "learning_rate": 2.5279015397004945e-07, "loss": 0.0003, "step": 40589 }, { "epoch": 7.44, "learning_rate": 2.526243029539699e-07, "loss": 0.0003, "step": 40590 }, { "epoch": 7.44, "learning_rate": 2.5245850566565365e-07, "loss": 0.0003, "step": 40591 }, { "epoch": 7.44, "learning_rate": 2.522927621060167e-07, "loss": 0.0006, "step": 40592 }, { "epoch": 7.44, "learning_rate": 2.5212707227596835e-07, "loss": 0.0006, "step": 40593 }, { "epoch": 7.44, "learning_rate": 2.519614361764233e-07, "loss": 0.0006, "step": 40594 }, { "epoch": 7.44, "learning_rate": 2.517958538082965e-07, "loss": 0.0007, "step": 40595 }, { "epoch": 7.44, "learning_rate": 2.516303251724994e-07, "loss": 0.0006, "step": 40596 }, { "epoch": 7.44, "learning_rate": 2.514648502699446e-07, "loss": 0.0005, "step": 40597 }, { "epoch": 7.45, "learning_rate": 2.512994291015436e-07, "loss": 0.0002, "step": 40598 }, { "epoch": 7.45, "learning_rate": 2.5113406166820784e-07, "loss": 0.0004, "step": 40599 }, { "epoch": 7.45, "learning_rate": 2.5096874797085004e-07, "loss": 0.0003, "step": 40600 }, { "epoch": 7.45, "learning_rate": 2.5080348801038044e-07, "loss": 0.0005, "step": 40601 }, { "epoch": 7.45, "learning_rate": 2.5063828178771177e-07, "loss": 0.0009, "step": 40602 }, { "epoch": 7.45, "learning_rate": 2.504731293037521e-07, "loss": 0.0004, "step": 40603 }, { "epoch": 7.45, "learning_rate": 2.50308030559413e-07, "loss": 0.0003, "step": 40604 }, { "epoch": 7.45, "learning_rate": 2.5014298555560477e-07, "loss": 0.0002, "step": 40605 }, { "epoch": 7.45, "learning_rate": 2.4997799429323677e-07, "loss": 0.0005, "step": 40606 }, { "epoch": 7.45, "learning_rate": 2.498130567732182e-07, "loss": 0.0004, "step": 40607 }, { "epoch": 7.45, "learning_rate": 2.496481729964584e-07, "loss": 0.0004, "step": 40608 }, { "epoch": 7.45, "learning_rate": 2.4948334296386657e-07, "loss": 0.0005, "step": 40609 }, { "epoch": 7.45, "learning_rate": 2.49318566676352e-07, "loss": 0.0005, "step": 40610 }, { "epoch": 7.45, "learning_rate": 2.491538441348207e-07, "loss": 0.0005, "step": 40611 }, { "epoch": 7.45, "learning_rate": 2.4898917534018075e-07, "loss": 0.0007, "step": 40612 }, { "epoch": 7.45, "learning_rate": 2.4882456029334254e-07, "loss": 0.0006, "step": 40613 }, { "epoch": 7.45, "learning_rate": 2.48659998995211e-07, "loss": 0.0005, "step": 40614 }, { "epoch": 7.45, "learning_rate": 2.4849549144669525e-07, "loss": 0.0004, "step": 40615 }, { "epoch": 7.45, "learning_rate": 2.4833103764870135e-07, "loss": 0.0004, "step": 40616 }, { "epoch": 7.45, "learning_rate": 2.48166637602133e-07, "loss": 0.0002, "step": 40617 }, { "epoch": 7.45, "learning_rate": 2.4800229130789944e-07, "loss": 0.0003, "step": 40618 }, { "epoch": 7.45, "learning_rate": 2.478379987669055e-07, "loss": 0.0003, "step": 40619 }, { "epoch": 7.45, "learning_rate": 2.4767375998005717e-07, "loss": 0.0002, "step": 40620 }, { "epoch": 7.45, "learning_rate": 2.4750957494826033e-07, "loss": 0.0003, "step": 40621 }, { "epoch": 7.45, "learning_rate": 2.473454436724176e-07, "loss": 0.0008, "step": 40622 }, { "epoch": 7.45, "learning_rate": 2.4718136615343615e-07, "loss": 0.0003, "step": 40623 }, { "epoch": 7.45, "learning_rate": 2.4701734239221955e-07, "loss": 0.0005, "step": 40624 }, { "epoch": 7.45, "learning_rate": 2.4685337238967154e-07, "loss": 0.0002, "step": 40625 }, { "epoch": 7.45, "learning_rate": 2.466894561466959e-07, "loss": 0.0005, "step": 40626 }, { "epoch": 7.45, "learning_rate": 2.465255936641986e-07, "loss": 0.0006, "step": 40627 }, { "epoch": 7.45, "learning_rate": 2.463617849430788e-07, "loss": 0.0004, "step": 40628 }, { "epoch": 7.45, "learning_rate": 2.4619802998424256e-07, "loss": 0.0004, "step": 40629 }, { "epoch": 7.45, "learning_rate": 2.460343287885891e-07, "loss": 0.0006, "step": 40630 }, { "epoch": 7.45, "learning_rate": 2.458706813570244e-07, "loss": 0.0008, "step": 40631 }, { "epoch": 7.45, "learning_rate": 2.457070876904488e-07, "loss": 0.0004, "step": 40632 }, { "epoch": 7.45, "learning_rate": 2.4554354778976496e-07, "loss": 0.0003, "step": 40633 }, { "epoch": 7.45, "learning_rate": 2.4538006165587323e-07, "loss": 0.0002, "step": 40634 }, { "epoch": 7.45, "learning_rate": 2.452166292896752e-07, "loss": 0.0002, "step": 40635 }, { "epoch": 7.45, "learning_rate": 2.450532506920711e-07, "loss": 0.0002, "step": 40636 }, { "epoch": 7.45, "learning_rate": 2.4488992586396255e-07, "loss": 0.0004, "step": 40637 }, { "epoch": 7.45, "learning_rate": 2.447266548062488e-07, "loss": 0.0004, "step": 40638 }, { "epoch": 7.45, "learning_rate": 2.4456343751983025e-07, "loss": 0.0003, "step": 40639 }, { "epoch": 7.45, "learning_rate": 2.444002740056073e-07, "loss": 0.0006, "step": 40640 }, { "epoch": 7.45, "learning_rate": 2.442371642644781e-07, "loss": 0.0004, "step": 40641 }, { "epoch": 7.45, "learning_rate": 2.4407410829734193e-07, "loss": 0.0005, "step": 40642 }, { "epoch": 7.45, "learning_rate": 2.4391110610509694e-07, "loss": 0.0003, "step": 40643 }, { "epoch": 7.45, "learning_rate": 2.437481576886436e-07, "loss": 0.0008, "step": 40644 }, { "epoch": 7.45, "learning_rate": 2.4358526304888e-07, "loss": 0.0005, "step": 40645 }, { "epoch": 7.45, "learning_rate": 2.434224221867021e-07, "loss": 0.0006, "step": 40646 }, { "epoch": 7.45, "learning_rate": 2.432596351030081e-07, "loss": 0.0004, "step": 40647 }, { "epoch": 7.45, "learning_rate": 2.4309690179869503e-07, "loss": 0.0003, "step": 40648 }, { "epoch": 7.45, "learning_rate": 2.429342222746611e-07, "loss": 0.0004, "step": 40649 }, { "epoch": 7.45, "learning_rate": 2.4277159653180335e-07, "loss": 0.0004, "step": 40650 }, { "epoch": 7.45, "learning_rate": 2.4260902457101554e-07, "loss": 0.0012, "step": 40651 }, { "epoch": 7.45, "learning_rate": 2.424465063931958e-07, "loss": 0.0004, "step": 40652 }, { "epoch": 7.46, "learning_rate": 2.4228404199924007e-07, "loss": 0.0006, "step": 40653 }, { "epoch": 7.46, "learning_rate": 2.421216313900432e-07, "loss": 0.0001, "step": 40654 }, { "epoch": 7.46, "learning_rate": 2.419592745665e-07, "loss": 0.0004, "step": 40655 }, { "epoch": 7.46, "learning_rate": 2.4179697152950653e-07, "loss": 0.001, "step": 40656 }, { "epoch": 7.46, "learning_rate": 2.416347222799564e-07, "loss": 0.0004, "step": 40657 }, { "epoch": 7.46, "learning_rate": 2.4147252681874454e-07, "loss": 0.0002, "step": 40658 }, { "epoch": 7.46, "learning_rate": 2.4131038514676466e-07, "loss": 0.0001, "step": 40659 }, { "epoch": 7.46, "learning_rate": 2.411482972649104e-07, "loss": 0.0007, "step": 40660 }, { "epoch": 7.46, "learning_rate": 2.4098626317407447e-07, "loss": 0.0005, "step": 40661 }, { "epoch": 7.46, "learning_rate": 2.408242828751517e-07, "loss": 0.0002, "step": 40662 }, { "epoch": 7.46, "learning_rate": 2.406623563690347e-07, "loss": 0.0008, "step": 40663 }, { "epoch": 7.46, "learning_rate": 2.4050048365661493e-07, "loss": 0.0002, "step": 40664 }, { "epoch": 7.46, "learning_rate": 2.4033866473878395e-07, "loss": 0.0008, "step": 40665 }, { "epoch": 7.46, "learning_rate": 2.4017689961643664e-07, "loss": 0.0009, "step": 40666 }, { "epoch": 7.46, "learning_rate": 2.400151882904622e-07, "loss": 0.0006, "step": 40667 }, { "epoch": 7.46, "learning_rate": 2.398535307617522e-07, "loss": 0.0005, "step": 40668 }, { "epoch": 7.46, "learning_rate": 2.3969192703119924e-07, "loss": 0.0003, "step": 40669 }, { "epoch": 7.46, "learning_rate": 2.395303770996926e-07, "loss": 0.0004, "step": 40670 }, { "epoch": 7.46, "learning_rate": 2.3936888096812384e-07, "loss": 0.0007, "step": 40671 }, { "epoch": 7.46, "learning_rate": 2.392074386373811e-07, "loss": 0.0005, "step": 40672 }, { "epoch": 7.46, "learning_rate": 2.3904605010835693e-07, "loss": 0.0006, "step": 40673 }, { "epoch": 7.46, "learning_rate": 2.3888471538193847e-07, "loss": 0.0006, "step": 40674 }, { "epoch": 7.46, "learning_rate": 2.3872343445901724e-07, "loss": 0.0007, "step": 40675 }, { "epoch": 7.46, "learning_rate": 2.385622073404803e-07, "loss": 0.0002, "step": 40676 }, { "epoch": 7.46, "learning_rate": 2.3840103402721804e-07, "loss": 0.0003, "step": 40677 }, { "epoch": 7.46, "learning_rate": 2.382399145201164e-07, "loss": 0.0005, "step": 40678 }, { "epoch": 7.46, "learning_rate": 2.3807884882006583e-07, "loss": 0.0004, "step": 40679 }, { "epoch": 7.46, "learning_rate": 2.3791783692795333e-07, "loss": 0.0007, "step": 40680 }, { "epoch": 7.46, "learning_rate": 2.3775687884466714e-07, "loss": 0.0003, "step": 40681 }, { "epoch": 7.46, "learning_rate": 2.375959745710932e-07, "loss": 0.0011, "step": 40682 }, { "epoch": 7.46, "learning_rate": 2.3743512410811852e-07, "loss": 0.0006, "step": 40683 }, { "epoch": 7.46, "learning_rate": 2.3727432745663025e-07, "loss": 0.0002, "step": 40684 }, { "epoch": 7.46, "learning_rate": 2.3711358461751544e-07, "loss": 0.0002, "step": 40685 }, { "epoch": 7.46, "learning_rate": 2.369528955916589e-07, "loss": 0.0008, "step": 40686 }, { "epoch": 7.46, "learning_rate": 2.3679226037994662e-07, "loss": 0.001, "step": 40687 }, { "epoch": 7.46, "learning_rate": 2.3663167898326345e-07, "loss": 0.0002, "step": 40688 }, { "epoch": 7.46, "learning_rate": 2.3647115140249644e-07, "loss": 0.0008, "step": 40689 }, { "epoch": 7.46, "learning_rate": 2.3631067763852821e-07, "loss": 0.0005, "step": 40690 }, { "epoch": 7.46, "learning_rate": 2.3615025769224476e-07, "loss": 0.0006, "step": 40691 }, { "epoch": 7.46, "learning_rate": 2.359898915645298e-07, "loss": 0.0003, "step": 40692 }, { "epoch": 7.46, "learning_rate": 2.3582957925626704e-07, "loss": 0.0005, "step": 40693 }, { "epoch": 7.46, "learning_rate": 2.3566932076834138e-07, "loss": 0.0004, "step": 40694 }, { "epoch": 7.46, "learning_rate": 2.3550911610163428e-07, "loss": 0.0002, "step": 40695 }, { "epoch": 7.46, "learning_rate": 2.3534896525702845e-07, "loss": 0.0005, "step": 40696 }, { "epoch": 7.46, "learning_rate": 2.3518886823540975e-07, "loss": 0.0007, "step": 40697 }, { "epoch": 7.46, "learning_rate": 2.3502882503765756e-07, "loss": 0.0002, "step": 40698 }, { "epoch": 7.46, "learning_rate": 2.348688356646578e-07, "loss": 0.0004, "step": 40699 }, { "epoch": 7.46, "learning_rate": 2.3470890011728642e-07, "loss": 0.0006, "step": 40700 }, { "epoch": 7.46, "learning_rate": 2.345490183964294e-07, "loss": 0.0006, "step": 40701 }, { "epoch": 7.46, "learning_rate": 2.3438919050296827e-07, "loss": 0.0004, "step": 40702 }, { "epoch": 7.46, "learning_rate": 2.342294164377812e-07, "loss": 0.0002, "step": 40703 }, { "epoch": 7.46, "learning_rate": 2.340696962017519e-07, "loss": 0.0008, "step": 40704 }, { "epoch": 7.46, "learning_rate": 2.339100297957575e-07, "loss": 0.0007, "step": 40705 }, { "epoch": 7.46, "learning_rate": 2.337504172206817e-07, "loss": 0.0003, "step": 40706 }, { "epoch": 7.47, "learning_rate": 2.3359085847740048e-07, "loss": 0.0005, "step": 40707 }, { "epoch": 7.47, "learning_rate": 2.3343135356679647e-07, "loss": 0.0005, "step": 40708 }, { "epoch": 7.47, "learning_rate": 2.3327190248974674e-07, "loss": 0.0003, "step": 40709 }, { "epoch": 7.47, "learning_rate": 2.331125052471328e-07, "loss": 0.0008, "step": 40710 }, { "epoch": 7.47, "learning_rate": 2.3295316183983064e-07, "loss": 0.0005, "step": 40711 }, { "epoch": 7.47, "learning_rate": 2.3279387226871842e-07, "loss": 0.0005, "step": 40712 }, { "epoch": 7.47, "learning_rate": 2.3263463653467655e-07, "loss": 0.0014, "step": 40713 }, { "epoch": 7.47, "learning_rate": 2.3247545463857991e-07, "loss": 0.0003, "step": 40714 }, { "epoch": 7.47, "learning_rate": 2.3231632658130777e-07, "loss": 0.0004, "step": 40715 }, { "epoch": 7.47, "learning_rate": 2.321572523637383e-07, "loss": 0.0004, "step": 40716 }, { "epoch": 7.47, "learning_rate": 2.3199823198674643e-07, "loss": 0.0005, "step": 40717 }, { "epoch": 7.47, "learning_rate": 2.3183926545120804e-07, "loss": 0.0003, "step": 40718 }, { "epoch": 7.47, "learning_rate": 2.316803527580014e-07, "loss": 0.0006, "step": 40719 }, { "epoch": 7.47, "learning_rate": 2.3152149390800128e-07, "loss": 0.0005, "step": 40720 }, { "epoch": 7.47, "learning_rate": 2.313626889020837e-07, "loss": 0.0003, "step": 40721 }, { "epoch": 7.47, "learning_rate": 2.312039377411235e-07, "loss": 0.0005, "step": 40722 }, { "epoch": 7.47, "learning_rate": 2.3104524042599662e-07, "loss": 0.0011, "step": 40723 }, { "epoch": 7.47, "learning_rate": 2.3088659695757687e-07, "loss": 0.0003, "step": 40724 }, { "epoch": 7.47, "learning_rate": 2.3072800733673906e-07, "loss": 0.0003, "step": 40725 }, { "epoch": 7.47, "learning_rate": 2.3056947156435695e-07, "loss": 0.0003, "step": 40726 }, { "epoch": 7.47, "learning_rate": 2.3041098964130426e-07, "loss": 0.0003, "step": 40727 }, { "epoch": 7.47, "learning_rate": 2.3025256156845698e-07, "loss": 0.0005, "step": 40728 }, { "epoch": 7.47, "learning_rate": 2.3009418734668554e-07, "loss": 0.0002, "step": 40729 }, { "epoch": 7.47, "learning_rate": 2.2993586697686366e-07, "loss": 0.0002, "step": 40730 }, { "epoch": 7.47, "learning_rate": 2.2977760045986397e-07, "loss": 0.0005, "step": 40731 }, { "epoch": 7.47, "learning_rate": 2.2961938779655911e-07, "loss": 0.0003, "step": 40732 }, { "epoch": 7.47, "learning_rate": 2.2946122898782065e-07, "loss": 0.0011, "step": 40733 }, { "epoch": 7.47, "learning_rate": 2.2930312403452226e-07, "loss": 0.0003, "step": 40734 }, { "epoch": 7.47, "learning_rate": 2.291450729375333e-07, "loss": 0.0003, "step": 40735 }, { "epoch": 7.47, "learning_rate": 2.2898707569772528e-07, "loss": 0.0005, "step": 40736 }, { "epoch": 7.47, "learning_rate": 2.288291323159697e-07, "loss": 0.0005, "step": 40737 }, { "epoch": 7.47, "learning_rate": 2.286712427931359e-07, "loss": 0.0007, "step": 40738 }, { "epoch": 7.47, "learning_rate": 2.285134071300965e-07, "loss": 0.0006, "step": 40739 }, { "epoch": 7.47, "learning_rate": 2.2835562532771972e-07, "loss": 0.0002, "step": 40740 }, { "epoch": 7.47, "learning_rate": 2.2819789738687482e-07, "loss": 0.0006, "step": 40741 }, { "epoch": 7.47, "learning_rate": 2.2804022330843335e-07, "loss": 0.0003, "step": 40742 }, { "epoch": 7.47, "learning_rate": 2.2788260309326237e-07, "loss": 0.0006, "step": 40743 }, { "epoch": 7.47, "learning_rate": 2.2772503674223124e-07, "loss": 0.0005, "step": 40744 }, { "epoch": 7.47, "learning_rate": 2.275675242562081e-07, "loss": 0.0007, "step": 40745 }, { "epoch": 7.47, "learning_rate": 2.2741006563606337e-07, "loss": 0.0003, "step": 40746 }, { "epoch": 7.47, "learning_rate": 2.2725266088266306e-07, "loss": 0.0004, "step": 40747 }, { "epoch": 7.47, "learning_rate": 2.2709530999687423e-07, "loss": 0.0003, "step": 40748 }, { "epoch": 7.47, "learning_rate": 2.2693801297956396e-07, "loss": 0.0003, "step": 40749 }, { "epoch": 7.47, "learning_rate": 2.2678076983160158e-07, "loss": 0.0003, "step": 40750 }, { "epoch": 7.47, "learning_rate": 2.2662358055385302e-07, "loss": 0.0004, "step": 40751 }, { "epoch": 7.47, "learning_rate": 2.264664451471843e-07, "loss": 0.0004, "step": 40752 }, { "epoch": 7.47, "learning_rate": 2.2630936361246026e-07, "loss": 0.0008, "step": 40753 }, { "epoch": 7.47, "learning_rate": 2.2615233595054908e-07, "loss": 0.0003, "step": 40754 }, { "epoch": 7.47, "learning_rate": 2.2599536216231454e-07, "loss": 0.0003, "step": 40755 }, { "epoch": 7.47, "learning_rate": 2.2583844224862373e-07, "loss": 0.0005, "step": 40756 }, { "epoch": 7.47, "learning_rate": 2.2568157621033926e-07, "loss": 0.0005, "step": 40757 }, { "epoch": 7.47, "learning_rate": 2.2552476404832711e-07, "loss": 0.0005, "step": 40758 }, { "epoch": 7.47, "learning_rate": 2.2536800576345107e-07, "loss": 0.0004, "step": 40759 }, { "epoch": 7.47, "learning_rate": 2.2521130135657599e-07, "loss": 0.0005, "step": 40760 }, { "epoch": 7.47, "learning_rate": 2.250546508285656e-07, "loss": 0.0002, "step": 40761 }, { "epoch": 7.48, "learning_rate": 2.2489805418028142e-07, "loss": 0.0002, "step": 40762 }, { "epoch": 7.48, "learning_rate": 2.247415114125895e-07, "loss": 0.0002, "step": 40763 }, { "epoch": 7.48, "learning_rate": 2.2458502252635238e-07, "loss": 0.0007, "step": 40764 }, { "epoch": 7.48, "learning_rate": 2.244285875224306e-07, "loss": 0.0002, "step": 40765 }, { "epoch": 7.48, "learning_rate": 2.242722064016878e-07, "loss": 0.0006, "step": 40766 }, { "epoch": 7.48, "learning_rate": 2.2411587916498445e-07, "loss": 0.0011, "step": 40767 }, { "epoch": 7.48, "learning_rate": 2.2395960581318432e-07, "loss": 0.0005, "step": 40768 }, { "epoch": 7.48, "learning_rate": 2.2380338634714783e-07, "loss": 0.0004, "step": 40769 }, { "epoch": 7.48, "learning_rate": 2.236472207677365e-07, "loss": 0.0008, "step": 40770 }, { "epoch": 7.48, "learning_rate": 2.2349110907580963e-07, "loss": 0.0005, "step": 40771 }, { "epoch": 7.48, "learning_rate": 2.2333505127222877e-07, "loss": 0.0008, "step": 40772 }, { "epoch": 7.48, "learning_rate": 2.2317904735785546e-07, "loss": 0.0004, "step": 40773 }, { "epoch": 7.48, "learning_rate": 2.2302309733354676e-07, "loss": 0.0004, "step": 40774 }, { "epoch": 7.48, "learning_rate": 2.2286720120016537e-07, "loss": 0.0004, "step": 40775 }, { "epoch": 7.48, "learning_rate": 2.2271135895856833e-07, "loss": 0.0004, "step": 40776 }, { "epoch": 7.48, "learning_rate": 2.2255557060961496e-07, "loss": 0.0003, "step": 40777 }, { "epoch": 7.48, "learning_rate": 2.223998361541646e-07, "loss": 0.0004, "step": 40778 }, { "epoch": 7.48, "learning_rate": 2.2224415559307544e-07, "loss": 0.0003, "step": 40779 }, { "epoch": 7.48, "learning_rate": 2.2208852892720457e-07, "loss": 0.0006, "step": 40780 }, { "epoch": 7.48, "learning_rate": 2.219329561574135e-07, "loss": 0.0006, "step": 40781 }, { "epoch": 7.48, "learning_rate": 2.217774372845549e-07, "loss": 0.0014, "step": 40782 }, { "epoch": 7.48, "learning_rate": 2.2162197230948922e-07, "loss": 0.0006, "step": 40783 }, { "epoch": 7.48, "learning_rate": 2.214665612330702e-07, "loss": 0.0003, "step": 40784 }, { "epoch": 7.48, "learning_rate": 2.2131120405615825e-07, "loss": 0.0004, "step": 40785 }, { "epoch": 7.48, "learning_rate": 2.2115590077960826e-07, "loss": 0.0002, "step": 40786 }, { "epoch": 7.48, "learning_rate": 2.2100065140427619e-07, "loss": 0.0004, "step": 40787 }, { "epoch": 7.48, "learning_rate": 2.2084545593101804e-07, "loss": 0.0003, "step": 40788 }, { "epoch": 7.48, "learning_rate": 2.2069031436068643e-07, "loss": 0.0003, "step": 40789 }, { "epoch": 7.48, "learning_rate": 2.2053522669414073e-07, "loss": 0.0004, "step": 40790 }, { "epoch": 7.48, "learning_rate": 2.2038019293223356e-07, "loss": 0.0008, "step": 40791 }, { "epoch": 7.48, "learning_rate": 2.2022521307581978e-07, "loss": 0.0003, "step": 40792 }, { "epoch": 7.48, "learning_rate": 2.200702871257543e-07, "loss": 0.0003, "step": 40793 }, { "epoch": 7.48, "learning_rate": 2.199154150828897e-07, "loss": 0.0004, "step": 40794 }, { "epoch": 7.48, "learning_rate": 2.1976059694808093e-07, "loss": 0.0005, "step": 40795 }, { "epoch": 7.48, "learning_rate": 2.1960583272218062e-07, "loss": 0.0003, "step": 40796 }, { "epoch": 7.48, "learning_rate": 2.1945112240604138e-07, "loss": 0.0003, "step": 40797 }, { "epoch": 7.48, "learning_rate": 2.1929646600051812e-07, "loss": 0.0005, "step": 40798 }, { "epoch": 7.48, "learning_rate": 2.1914186350646238e-07, "loss": 0.0007, "step": 40799 }, { "epoch": 7.48, "learning_rate": 2.1898731492472458e-07, "loss": 0.0002, "step": 40800 }, { "epoch": 7.48, "learning_rate": 2.1883282025615848e-07, "loss": 0.0015, "step": 40801 }, { "epoch": 7.48, "learning_rate": 2.186783795016134e-07, "loss": 0.0003, "step": 40802 }, { "epoch": 7.48, "learning_rate": 2.1852399266194312e-07, "loss": 0.0004, "step": 40803 }, { "epoch": 7.48, "learning_rate": 2.1836965973799807e-07, "loss": 0.0008, "step": 40804 }, { "epoch": 7.48, "learning_rate": 2.1821538073062865e-07, "loss": 0.0003, "step": 40805 }, { "epoch": 7.48, "learning_rate": 2.180611556406853e-07, "loss": 0.0005, "step": 40806 }, { "epoch": 7.48, "learning_rate": 2.179069844690185e-07, "loss": 0.0003, "step": 40807 }, { "epoch": 7.48, "learning_rate": 2.1775286721647748e-07, "loss": 0.0003, "step": 40808 }, { "epoch": 7.48, "learning_rate": 2.1759880388391053e-07, "loss": 0.0005, "step": 40809 }, { "epoch": 7.48, "learning_rate": 2.1744479447216914e-07, "loss": 0.0006, "step": 40810 }, { "epoch": 7.48, "learning_rate": 2.1729083898210157e-07, "loss": 0.0003, "step": 40811 }, { "epoch": 7.48, "learning_rate": 2.171369374145549e-07, "loss": 0.0006, "step": 40812 }, { "epoch": 7.48, "learning_rate": 2.1698308977037952e-07, "loss": 0.0003, "step": 40813 }, { "epoch": 7.48, "learning_rate": 2.1682929605042258e-07, "loss": 0.0006, "step": 40814 }, { "epoch": 7.48, "learning_rate": 2.1667555625553005e-07, "loss": 0.0007, "step": 40815 }, { "epoch": 7.49, "learning_rate": 2.1652187038655348e-07, "loss": 0.0007, "step": 40816 }, { "epoch": 7.49, "learning_rate": 2.1636823844433663e-07, "loss": 0.0003, "step": 40817 }, { "epoch": 7.49, "learning_rate": 2.1621466042972773e-07, "loss": 0.0004, "step": 40818 }, { "epoch": 7.49, "learning_rate": 2.1606113634357272e-07, "loss": 0.0004, "step": 40819 }, { "epoch": 7.49, "learning_rate": 2.1590766618671654e-07, "loss": 0.0005, "step": 40820 }, { "epoch": 7.49, "learning_rate": 2.157542499600074e-07, "loss": 0.0002, "step": 40821 }, { "epoch": 7.49, "learning_rate": 2.156008876642901e-07, "loss": 0.0001, "step": 40822 }, { "epoch": 7.49, "learning_rate": 2.154475793004096e-07, "loss": 0.0005, "step": 40823 }, { "epoch": 7.49, "learning_rate": 2.1529432486921188e-07, "loss": 0.0006, "step": 40824 }, { "epoch": 7.49, "learning_rate": 2.1514112437154066e-07, "loss": 0.0002, "step": 40825 }, { "epoch": 7.49, "learning_rate": 2.1498797780824086e-07, "loss": 0.0003, "step": 40826 }, { "epoch": 7.49, "learning_rate": 2.1483488518015738e-07, "loss": 0.0004, "step": 40827 }, { "epoch": 7.49, "learning_rate": 2.1468184648813173e-07, "loss": 0.0013, "step": 40828 }, { "epoch": 7.49, "learning_rate": 2.145288617330099e-07, "loss": 0.0003, "step": 40829 }, { "epoch": 7.49, "learning_rate": 2.1437593091563457e-07, "loss": 0.0005, "step": 40830 }, { "epoch": 7.49, "learning_rate": 2.142230540368484e-07, "loss": 0.0004, "step": 40831 }, { "epoch": 7.49, "learning_rate": 2.140702310974929e-07, "loss": 0.0004, "step": 40832 }, { "epoch": 7.49, "learning_rate": 2.1391746209841192e-07, "loss": 0.0004, "step": 40833 }, { "epoch": 7.49, "learning_rate": 2.1376474704044693e-07, "loss": 0.0005, "step": 40834 }, { "epoch": 7.49, "learning_rate": 2.136120859244417e-07, "loss": 0.0001, "step": 40835 }, { "epoch": 7.49, "learning_rate": 2.1345947875123452e-07, "loss": 0.0002, "step": 40836 }, { "epoch": 7.49, "learning_rate": 2.1330692552166688e-07, "loss": 0.0003, "step": 40837 }, { "epoch": 7.49, "learning_rate": 2.1315442623658256e-07, "loss": 0.0003, "step": 40838 }, { "epoch": 7.49, "learning_rate": 2.130019808968198e-07, "loss": 0.0002, "step": 40839 }, { "epoch": 7.49, "learning_rate": 2.1284958950321899e-07, "loss": 0.0004, "step": 40840 }, { "epoch": 7.49, "learning_rate": 2.1269725205662172e-07, "loss": 0.0003, "step": 40841 }, { "epoch": 7.49, "learning_rate": 2.125449685578651e-07, "loss": 0.0003, "step": 40842 }, { "epoch": 7.49, "learning_rate": 2.1239273900779066e-07, "loss": 0.0003, "step": 40843 }, { "epoch": 7.49, "learning_rate": 2.1224056340723664e-07, "loss": 0.0005, "step": 40844 }, { "epoch": 7.49, "learning_rate": 2.1208844175704125e-07, "loss": 0.0005, "step": 40845 }, { "epoch": 7.49, "learning_rate": 2.119363740580449e-07, "loss": 0.0013, "step": 40846 }, { "epoch": 7.49, "learning_rate": 2.1178436031108362e-07, "loss": 0.0006, "step": 40847 }, { "epoch": 7.49, "learning_rate": 2.1163240051699674e-07, "loss": 0.0011, "step": 40848 }, { "epoch": 7.49, "learning_rate": 2.1148049467662136e-07, "loss": 0.0002, "step": 40849 }, { "epoch": 7.49, "learning_rate": 2.1132864279079345e-07, "loss": 0.0008, "step": 40850 }, { "epoch": 7.49, "learning_rate": 2.1117684486035238e-07, "loss": 0.0003, "step": 40851 }, { "epoch": 7.49, "learning_rate": 2.1102510088613415e-07, "loss": 0.0008, "step": 40852 }, { "epoch": 7.49, "learning_rate": 2.1087341086897583e-07, "loss": 0.0003, "step": 40853 }, { "epoch": 7.49, "learning_rate": 2.1072177480971233e-07, "loss": 0.0006, "step": 40854 }, { "epoch": 7.49, "learning_rate": 2.1057019270917746e-07, "loss": 0.0006, "step": 40855 }, { "epoch": 7.49, "learning_rate": 2.1041866456821048e-07, "loss": 0.0002, "step": 40856 }, { "epoch": 7.49, "learning_rate": 2.102671903876452e-07, "loss": 0.0008, "step": 40857 }, { "epoch": 7.49, "learning_rate": 2.1011577016831764e-07, "loss": 0.0004, "step": 40858 }, { "epoch": 7.49, "learning_rate": 2.0996440391105931e-07, "loss": 0.0012, "step": 40859 }, { "epoch": 7.49, "learning_rate": 2.0981309161670848e-07, "loss": 0.0005, "step": 40860 }, { "epoch": 7.49, "learning_rate": 2.0966183328609556e-07, "loss": 0.0002, "step": 40861 }, { "epoch": 7.49, "learning_rate": 2.095106289200566e-07, "loss": 0.0005, "step": 40862 }, { "epoch": 7.49, "learning_rate": 2.0935947851942528e-07, "loss": 0.0004, "step": 40863 }, { "epoch": 7.49, "learning_rate": 2.0920838208503325e-07, "loss": 0.0003, "step": 40864 }, { "epoch": 7.49, "learning_rate": 2.0905733961771313e-07, "loss": 0.0004, "step": 40865 }, { "epoch": 7.49, "learning_rate": 2.0890635111829983e-07, "loss": 0.0006, "step": 40866 }, { "epoch": 7.49, "learning_rate": 2.0875541658762267e-07, "loss": 0.0008, "step": 40867 }, { "epoch": 7.49, "learning_rate": 2.0860453602651543e-07, "loss": 0.0003, "step": 40868 }, { "epoch": 7.49, "learning_rate": 2.0845370943580967e-07, "loss": 0.0006, "step": 40869 }, { "epoch": 7.49, "learning_rate": 2.0830293681633585e-07, "loss": 0.0003, "step": 40870 }, { "epoch": 7.5, "learning_rate": 2.0815221816892772e-07, "loss": 0.0015, "step": 40871 }, { "epoch": 7.5, "learning_rate": 2.0800155349441242e-07, "loss": 0.0008, "step": 40872 }, { "epoch": 7.5, "learning_rate": 2.078509427936204e-07, "loss": 0.001, "step": 40873 }, { "epoch": 7.5, "learning_rate": 2.077003860673854e-07, "loss": 0.0003, "step": 40874 }, { "epoch": 7.5, "learning_rate": 2.0754988331653457e-07, "loss": 0.0006, "step": 40875 }, { "epoch": 7.5, "learning_rate": 2.0739943454189838e-07, "loss": 0.0008, "step": 40876 }, { "epoch": 7.5, "learning_rate": 2.072490397443061e-07, "loss": 0.0002, "step": 40877 }, { "epoch": 7.5, "learning_rate": 2.070986989245871e-07, "loss": 0.0007, "step": 40878 }, { "epoch": 7.5, "learning_rate": 2.0694841208356853e-07, "loss": 0.0006, "step": 40879 }, { "epoch": 7.5, "learning_rate": 2.067981792220797e-07, "loss": 0.0002, "step": 40880 }, { "epoch": 7.5, "learning_rate": 2.0664800034094877e-07, "loss": 0.0006, "step": 40881 }, { "epoch": 7.5, "learning_rate": 2.0649787544100519e-07, "loss": 0.0001, "step": 40882 }, { "epoch": 7.5, "learning_rate": 2.0634780452307267e-07, "loss": 0.0002, "step": 40883 }, { "epoch": 7.5, "learning_rate": 2.061977875879817e-07, "loss": 0.0001, "step": 40884 }, { "epoch": 7.5, "learning_rate": 2.0604782463655713e-07, "loss": 0.0003, "step": 40885 }, { "epoch": 7.5, "learning_rate": 2.0589791566962612e-07, "loss": 0.0005, "step": 40886 }, { "epoch": 7.5, "learning_rate": 2.0574806068801578e-07, "loss": 0.0003, "step": 40887 }, { "epoch": 7.5, "learning_rate": 2.055982596925532e-07, "loss": 0.0002, "step": 40888 }, { "epoch": 7.5, "learning_rate": 2.054485126840611e-07, "loss": 0.0004, "step": 40889 }, { "epoch": 7.5, "learning_rate": 2.0529881966336428e-07, "loss": 0.0003, "step": 40890 }, { "epoch": 7.5, "learning_rate": 2.051491806312922e-07, "loss": 0.0004, "step": 40891 }, { "epoch": 7.5, "learning_rate": 2.0499959558866633e-07, "loss": 0.0002, "step": 40892 }, { "epoch": 7.5, "learning_rate": 2.0485006453631273e-07, "loss": 0.0004, "step": 40893 }, { "epoch": 7.5, "learning_rate": 2.0470058747505516e-07, "loss": 0.0007, "step": 40894 }, { "epoch": 7.5, "learning_rate": 2.0455116440571743e-07, "loss": 0.0008, "step": 40895 }, { "epoch": 7.5, "learning_rate": 2.044017953291233e-07, "loss": 0.0005, "step": 40896 }, { "epoch": 7.5, "learning_rate": 2.0425248024609546e-07, "loss": 0.0003, "step": 40897 }, { "epoch": 7.5, "learning_rate": 2.0410321915745767e-07, "loss": 0.0004, "step": 40898 }, { "epoch": 7.5, "learning_rate": 2.0395401206403153e-07, "loss": 0.0004, "step": 40899 }, { "epoch": 7.5, "learning_rate": 2.0380485896664192e-07, "loss": 0.0003, "step": 40900 }, { "epoch": 7.5, "learning_rate": 2.036557598661082e-07, "loss": 0.0002, "step": 40901 }, { "epoch": 7.5, "learning_rate": 2.0350671476325413e-07, "loss": 0.0005, "step": 40902 }, { "epoch": 7.5, "learning_rate": 2.0335772365890017e-07, "loss": 0.0003, "step": 40903 }, { "epoch": 7.5, "learning_rate": 2.0320878655386788e-07, "loss": 0.0003, "step": 40904 }, { "epoch": 7.5, "learning_rate": 2.0305990344897774e-07, "loss": 0.0003, "step": 40905 }, { "epoch": 7.5, "learning_rate": 2.0291107434505352e-07, "loss": 0.0004, "step": 40906 }, { "epoch": 7.5, "learning_rate": 2.0276229924291123e-07, "loss": 0.0004, "step": 40907 }, { "epoch": 7.5, "learning_rate": 2.0261357814337136e-07, "loss": 0.0005, "step": 40908 }, { "epoch": 7.5, "learning_rate": 2.0246491104725653e-07, "loss": 0.0005, "step": 40909 }, { "epoch": 7.5, "learning_rate": 2.0231629795538388e-07, "loss": 0.0003, "step": 40910 }, { "epoch": 7.5, "learning_rate": 2.0216773886857389e-07, "loss": 0.0007, "step": 40911 }, { "epoch": 7.5, "learning_rate": 2.0201923378764477e-07, "loss": 0.0005, "step": 40912 }, { "epoch": 7.5, "learning_rate": 2.0187078271341588e-07, "loss": 0.0003, "step": 40913 }, { "epoch": 7.5, "learning_rate": 2.017223856467032e-07, "loss": 0.0007, "step": 40914 }, { "epoch": 7.5, "learning_rate": 2.0157404258832724e-07, "loss": 0.0004, "step": 40915 }, { "epoch": 7.5, "learning_rate": 2.014257535391051e-07, "loss": 0.0003, "step": 40916 }, { "epoch": 7.5, "learning_rate": 2.0127751849985279e-07, "loss": 0.0005, "step": 40917 }, { "epoch": 7.5, "learning_rate": 2.0112933747138962e-07, "loss": 0.0003, "step": 40918 }, { "epoch": 7.5, "learning_rate": 2.0098121045452946e-07, "loss": 0.0005, "step": 40919 }, { "epoch": 7.5, "learning_rate": 2.0083313745009158e-07, "loss": 0.0003, "step": 40920 }, { "epoch": 7.5, "learning_rate": 2.0068511845888982e-07, "loss": 0.0003, "step": 40921 }, { "epoch": 7.5, "learning_rate": 2.005371534817424e-07, "loss": 0.0004, "step": 40922 }, { "epoch": 7.5, "learning_rate": 2.0038924251946422e-07, "loss": 0.0004, "step": 40923 }, { "epoch": 7.5, "learning_rate": 2.002413855728702e-07, "loss": 0.0003, "step": 40924 }, { "epoch": 7.51, "learning_rate": 2.00093582642773e-07, "loss": 0.0003, "step": 40925 }, { "epoch": 7.51, "learning_rate": 1.9994583372999087e-07, "loss": 0.0006, "step": 40926 }, { "epoch": 7.51, "learning_rate": 1.9979813883533762e-07, "loss": 0.0013, "step": 40927 }, { "epoch": 7.51, "learning_rate": 1.9965049795962588e-07, "loss": 0.0004, "step": 40928 }, { "epoch": 7.51, "learning_rate": 1.995029111036706e-07, "loss": 0.0007, "step": 40929 }, { "epoch": 7.51, "learning_rate": 1.9935537826828445e-07, "loss": 0.0004, "step": 40930 }, { "epoch": 7.51, "learning_rate": 1.992078994542812e-07, "loss": 0.0003, "step": 40931 }, { "epoch": 7.51, "learning_rate": 1.990604746624747e-07, "loss": 0.0004, "step": 40932 }, { "epoch": 7.51, "learning_rate": 1.9891310389367535e-07, "loss": 0.0002, "step": 40933 }, { "epoch": 7.51, "learning_rate": 1.9876578714869588e-07, "loss": 0.0004, "step": 40934 }, { "epoch": 7.51, "learning_rate": 1.9861852442835005e-07, "loss": 0.0002, "step": 40935 }, { "epoch": 7.51, "learning_rate": 1.9847131573344835e-07, "loss": 0.0003, "step": 40936 }, { "epoch": 7.51, "learning_rate": 1.9832416106480233e-07, "loss": 0.0005, "step": 40937 }, { "epoch": 7.51, "learning_rate": 1.981770604232236e-07, "loss": 0.0003, "step": 40938 }, { "epoch": 7.51, "learning_rate": 1.9803001380952146e-07, "loss": 0.0004, "step": 40939 }, { "epoch": 7.51, "learning_rate": 1.9788302122450753e-07, "loss": 0.0005, "step": 40940 }, { "epoch": 7.51, "learning_rate": 1.9773608266899337e-07, "loss": 0.0002, "step": 40941 }, { "epoch": 7.51, "learning_rate": 1.975891981437872e-07, "loss": 0.0005, "step": 40942 }, { "epoch": 7.51, "learning_rate": 1.9744236764969837e-07, "loss": 0.0005, "step": 40943 }, { "epoch": 7.51, "learning_rate": 1.9729559118753738e-07, "loss": 0.0006, "step": 40944 }, { "epoch": 7.51, "learning_rate": 1.9714886875811245e-07, "loss": 0.0003, "step": 40945 }, { "epoch": 7.51, "learning_rate": 1.9700220036223294e-07, "loss": 0.0008, "step": 40946 }, { "epoch": 7.51, "learning_rate": 1.9685558600070598e-07, "loss": 0.0005, "step": 40947 }, { "epoch": 7.51, "learning_rate": 1.9670902567434202e-07, "loss": 0.0016, "step": 40948 }, { "epoch": 7.51, "learning_rate": 1.965625193839471e-07, "loss": 0.0002, "step": 40949 }, { "epoch": 7.51, "learning_rate": 1.9641606713032945e-07, "loss": 0.0001, "step": 40950 }, { "epoch": 7.51, "learning_rate": 1.962696689142951e-07, "loss": 0.0012, "step": 40951 }, { "epoch": 7.51, "learning_rate": 1.961233247366523e-07, "loss": 0.0004, "step": 40952 }, { "epoch": 7.51, "learning_rate": 1.9597703459820926e-07, "loss": 0.0008, "step": 40953 }, { "epoch": 7.51, "learning_rate": 1.9583079849976872e-07, "loss": 0.0003, "step": 40954 }, { "epoch": 7.51, "learning_rate": 1.9568461644213887e-07, "loss": 0.0006, "step": 40955 }, { "epoch": 7.51, "learning_rate": 1.9553848842612466e-07, "loss": 0.0004, "step": 40956 }, { "epoch": 7.51, "learning_rate": 1.953924144525321e-07, "loss": 0.0002, "step": 40957 }, { "epoch": 7.51, "learning_rate": 1.9524639452216609e-07, "loss": 0.0004, "step": 40958 }, { "epoch": 7.51, "learning_rate": 1.9510042863583268e-07, "loss": 0.0006, "step": 40959 }, { "epoch": 7.51, "learning_rate": 1.9495451679433565e-07, "loss": 0.0004, "step": 40960 }, { "epoch": 7.51, "learning_rate": 1.9480865899847768e-07, "loss": 0.0003, "step": 40961 }, { "epoch": 7.51, "learning_rate": 1.946628552490648e-07, "loss": 0.0005, "step": 40962 }, { "epoch": 7.51, "learning_rate": 1.9451710554689974e-07, "loss": 0.0003, "step": 40963 }, { "epoch": 7.51, "learning_rate": 1.9437140989278624e-07, "loss": 0.0003, "step": 40964 }, { "epoch": 7.51, "learning_rate": 1.9422576828752704e-07, "loss": 0.0008, "step": 40965 }, { "epoch": 7.51, "learning_rate": 1.9408018073192482e-07, "loss": 0.0002, "step": 40966 }, { "epoch": 7.51, "learning_rate": 1.9393464722678334e-07, "loss": 0.0005, "step": 40967 }, { "epoch": 7.51, "learning_rate": 1.9378916777290312e-07, "loss": 0.0005, "step": 40968 }, { "epoch": 7.51, "learning_rate": 1.9364374237108686e-07, "loss": 0.0003, "step": 40969 }, { "epoch": 7.51, "learning_rate": 1.934983710221361e-07, "loss": 0.0008, "step": 40970 }, { "epoch": 7.51, "learning_rate": 1.9335305372685242e-07, "loss": 0.0006, "step": 40971 }, { "epoch": 7.51, "learning_rate": 1.9320779048603632e-07, "loss": 0.0005, "step": 40972 }, { "epoch": 7.51, "learning_rate": 1.9306258130048828e-07, "loss": 0.0004, "step": 40973 }, { "epoch": 7.51, "learning_rate": 1.9291742617100872e-07, "loss": 0.0002, "step": 40974 }, { "epoch": 7.51, "learning_rate": 1.9277232509839816e-07, "loss": 0.0006, "step": 40975 }, { "epoch": 7.51, "learning_rate": 1.9262727808345704e-07, "loss": 0.001, "step": 40976 }, { "epoch": 7.51, "learning_rate": 1.9248228512698365e-07, "loss": 0.0006, "step": 40977 }, { "epoch": 7.51, "learning_rate": 1.923373462297784e-07, "loss": 0.0004, "step": 40978 }, { "epoch": 7.51, "learning_rate": 1.9219246139263848e-07, "loss": 0.0005, "step": 40979 }, { "epoch": 7.52, "learning_rate": 1.9204763061636434e-07, "loss": 0.0003, "step": 40980 }, { "epoch": 7.52, "learning_rate": 1.91902853901752e-07, "loss": 0.0005, "step": 40981 }, { "epoch": 7.52, "learning_rate": 1.9175813124960196e-07, "loss": 0.0005, "step": 40982 }, { "epoch": 7.52, "learning_rate": 1.9161346266071134e-07, "loss": 0.0009, "step": 40983 }, { "epoch": 7.52, "learning_rate": 1.914688481358762e-07, "loss": 0.0004, "step": 40984 }, { "epoch": 7.52, "learning_rate": 1.9132428767589471e-07, "loss": 0.0005, "step": 40985 }, { "epoch": 7.52, "learning_rate": 1.9117978128156412e-07, "loss": 0.0004, "step": 40986 }, { "epoch": 7.52, "learning_rate": 1.9103532895367925e-07, "loss": 0.0002, "step": 40987 }, { "epoch": 7.52, "learning_rate": 1.9089093069303844e-07, "loss": 0.0008, "step": 40988 }, { "epoch": 7.52, "learning_rate": 1.9074658650043764e-07, "loss": 0.0012, "step": 40989 }, { "epoch": 7.52, "learning_rate": 1.906022963766696e-07, "loss": 0.0013, "step": 40990 }, { "epoch": 7.52, "learning_rate": 1.9045806032253255e-07, "loss": 0.0005, "step": 40991 }, { "epoch": 7.52, "learning_rate": 1.9031387833881921e-07, "loss": 0.0005, "step": 40992 }, { "epoch": 7.52, "learning_rate": 1.9016975042632667e-07, "loss": 0.0005, "step": 40993 }, { "epoch": 7.52, "learning_rate": 1.900256765858477e-07, "loss": 0.0003, "step": 40994 }, { "epoch": 7.52, "learning_rate": 1.8988165681817827e-07, "loss": 0.0002, "step": 40995 }, { "epoch": 7.52, "learning_rate": 1.897376911241089e-07, "loss": 0.0017, "step": 40996 }, { "epoch": 7.52, "learning_rate": 1.895937795044367e-07, "loss": 0.0003, "step": 40997 }, { "epoch": 7.52, "learning_rate": 1.8944992195995326e-07, "loss": 0.0003, "step": 40998 }, { "epoch": 7.52, "learning_rate": 1.8930611849145131e-07, "loss": 0.0007, "step": 40999 }, { "epoch": 7.52, "learning_rate": 1.8916236909972463e-07, "loss": 0.0003, "step": 41000 }, { "epoch": 7.52, "learning_rate": 1.890186737855637e-07, "loss": 0.0004, "step": 41001 }, { "epoch": 7.52, "learning_rate": 1.8887503254976235e-07, "loss": 0.0004, "step": 41002 }, { "epoch": 7.52, "learning_rate": 1.8873144539311217e-07, "loss": 0.0007, "step": 41003 }, { "epoch": 7.52, "learning_rate": 1.885879123164025e-07, "loss": 0.0004, "step": 41004 }, { "epoch": 7.52, "learning_rate": 1.8844443332042716e-07, "loss": 0.0008, "step": 41005 }, { "epoch": 7.52, "learning_rate": 1.8830100840597555e-07, "loss": 0.0004, "step": 41006 }, { "epoch": 7.52, "learning_rate": 1.881576375738403e-07, "loss": 0.0005, "step": 41007 }, { "epoch": 7.52, "learning_rate": 1.8801432082480863e-07, "loss": 0.0012, "step": 41008 }, { "epoch": 7.52, "learning_rate": 1.87871058159671e-07, "loss": 0.0003, "step": 41009 }, { "epoch": 7.52, "learning_rate": 1.8772784957921898e-07, "loss": 0.0003, "step": 41010 }, { "epoch": 7.52, "learning_rate": 1.8758469508424082e-07, "loss": 0.0004, "step": 41011 }, { "epoch": 7.52, "learning_rate": 1.8744159467552593e-07, "loss": 0.0008, "step": 41012 }, { "epoch": 7.52, "learning_rate": 1.8729854835386362e-07, "loss": 0.0009, "step": 41013 }, { "epoch": 7.52, "learning_rate": 1.8715555612004e-07, "loss": 0.0003, "step": 41014 }, { "epoch": 7.52, "learning_rate": 1.8701261797484548e-07, "loss": 0.0003, "step": 41015 }, { "epoch": 7.52, "learning_rate": 1.8686973391906727e-07, "loss": 0.0005, "step": 41016 }, { "epoch": 7.52, "learning_rate": 1.8672690395349357e-07, "loss": 0.0008, "step": 41017 }, { "epoch": 7.52, "learning_rate": 1.8658412807891048e-07, "loss": 0.0003, "step": 41018 }, { "epoch": 7.52, "learning_rate": 1.8644140629610619e-07, "loss": 0.0003, "step": 41019 }, { "epoch": 7.52, "learning_rate": 1.8629873860586567e-07, "loss": 0.0005, "step": 41020 }, { "epoch": 7.52, "learning_rate": 1.8615612500897717e-07, "loss": 0.0002, "step": 41021 }, { "epoch": 7.52, "learning_rate": 1.860135655062256e-07, "loss": 0.0004, "step": 41022 }, { "epoch": 7.52, "learning_rate": 1.8587106009839705e-07, "loss": 0.0007, "step": 41023 }, { "epoch": 7.52, "learning_rate": 1.857286087862775e-07, "loss": 0.0007, "step": 41024 }, { "epoch": 7.52, "learning_rate": 1.8558621157065303e-07, "loss": 0.001, "step": 41025 }, { "epoch": 7.52, "learning_rate": 1.8544386845230633e-07, "loss": 0.0004, "step": 41026 }, { "epoch": 7.52, "learning_rate": 1.853015794320212e-07, "loss": 0.0002, "step": 41027 }, { "epoch": 7.52, "learning_rate": 1.8515934451058593e-07, "loss": 0.0003, "step": 41028 }, { "epoch": 7.52, "learning_rate": 1.8501716368878097e-07, "loss": 0.0004, "step": 41029 }, { "epoch": 7.52, "learning_rate": 1.848750369673924e-07, "loss": 0.0002, "step": 41030 }, { "epoch": 7.52, "learning_rate": 1.847329643472018e-07, "loss": 0.0006, "step": 41031 }, { "epoch": 7.52, "learning_rate": 1.8459094582899406e-07, "loss": 0.0004, "step": 41032 }, { "epoch": 7.52, "learning_rate": 1.8444898141354972e-07, "loss": 0.0007, "step": 41033 }, { "epoch": 7.53, "learning_rate": 1.843070711016537e-07, "loss": 0.0003, "step": 41034 }, { "epoch": 7.53, "learning_rate": 1.8416521489408647e-07, "loss": 0.0004, "step": 41035 }, { "epoch": 7.53, "learning_rate": 1.840234127916307e-07, "loss": 0.0002, "step": 41036 }, { "epoch": 7.53, "learning_rate": 1.8388166479506698e-07, "loss": 0.0002, "step": 41037 }, { "epoch": 7.53, "learning_rate": 1.8373997090517903e-07, "loss": 0.0002, "step": 41038 }, { "epoch": 7.53, "learning_rate": 1.8359833112274517e-07, "loss": 0.0004, "step": 41039 }, { "epoch": 7.53, "learning_rate": 1.8345674544854696e-07, "loss": 0.0004, "step": 41040 }, { "epoch": 7.53, "learning_rate": 1.8331521388336603e-07, "loss": 0.0003, "step": 41041 }, { "epoch": 7.53, "learning_rate": 1.8317373642798176e-07, "loss": 0.0002, "step": 41042 }, { "epoch": 7.53, "learning_rate": 1.830323130831735e-07, "loss": 0.0003, "step": 41043 }, { "epoch": 7.53, "learning_rate": 1.8289094384972172e-07, "loss": 0.0005, "step": 41044 }, { "epoch": 7.53, "learning_rate": 1.8274962872840362e-07, "loss": 0.0004, "step": 41045 }, { "epoch": 7.53, "learning_rate": 1.8260836771999968e-07, "loss": 0.0002, "step": 41046 }, { "epoch": 7.53, "learning_rate": 1.8246716082528925e-07, "loss": 0.0006, "step": 41047 }, { "epoch": 7.53, "learning_rate": 1.8232600804504952e-07, "loss": 0.0004, "step": 41048 }, { "epoch": 7.53, "learning_rate": 1.8218490938005985e-07, "loss": 0.0002, "step": 41049 }, { "epoch": 7.53, "learning_rate": 1.8204386483109514e-07, "loss": 0.0004, "step": 41050 }, { "epoch": 7.53, "learning_rate": 1.8190287439893595e-07, "loss": 0.0003, "step": 41051 }, { "epoch": 7.53, "learning_rate": 1.8176193808435826e-07, "loss": 0.0003, "step": 41052 }, { "epoch": 7.53, "learning_rate": 1.8162105588813704e-07, "loss": 0.0004, "step": 41053 }, { "epoch": 7.53, "learning_rate": 1.814802278110528e-07, "loss": 0.0003, "step": 41054 }, { "epoch": 7.53, "learning_rate": 1.813394538538793e-07, "loss": 0.0002, "step": 41055 }, { "epoch": 7.53, "learning_rate": 1.8119873401739151e-07, "loss": 0.0002, "step": 41056 }, { "epoch": 7.53, "learning_rate": 1.810580683023666e-07, "loss": 0.0007, "step": 41057 }, { "epoch": 7.53, "learning_rate": 1.809174567095795e-07, "loss": 0.0005, "step": 41058 }, { "epoch": 7.53, "learning_rate": 1.8077689923980623e-07, "loss": 0.0002, "step": 41059 }, { "epoch": 7.53, "learning_rate": 1.8063639589382065e-07, "loss": 0.0002, "step": 41060 }, { "epoch": 7.53, "learning_rate": 1.8049594667239767e-07, "loss": 0.0004, "step": 41061 }, { "epoch": 7.53, "learning_rate": 1.8035555157631e-07, "loss": 0.0004, "step": 41062 }, { "epoch": 7.53, "learning_rate": 1.8021521060633262e-07, "loss": 0.0004, "step": 41063 }, { "epoch": 7.53, "learning_rate": 1.8007492376323933e-07, "loss": 0.0003, "step": 41064 }, { "epoch": 7.53, "learning_rate": 1.799346910478028e-07, "loss": 0.0003, "step": 41065 }, { "epoch": 7.53, "learning_rate": 1.7979451246079694e-07, "loss": 0.0006, "step": 41066 }, { "epoch": 7.53, "learning_rate": 1.796543880029944e-07, "loss": 0.0003, "step": 41067 }, { "epoch": 7.53, "learning_rate": 1.795143176751657e-07, "loss": 0.0004, "step": 41068 }, { "epoch": 7.53, "learning_rate": 1.7937430147808465e-07, "loss": 0.0002, "step": 41069 }, { "epoch": 7.53, "learning_rate": 1.7923433941252287e-07, "loss": 0.0005, "step": 41070 }, { "epoch": 7.53, "learning_rate": 1.7909443147924976e-07, "loss": 0.0003, "step": 41071 }, { "epoch": 7.53, "learning_rate": 1.7895457767904022e-07, "loss": 0.0001, "step": 41072 }, { "epoch": 7.53, "learning_rate": 1.7881477801266255e-07, "loss": 0.0004, "step": 41073 }, { "epoch": 7.53, "learning_rate": 1.7867503248088836e-07, "loss": 0.0002, "step": 41074 }, { "epoch": 7.53, "learning_rate": 1.78535341084487e-07, "loss": 0.0003, "step": 41075 }, { "epoch": 7.53, "learning_rate": 1.783957038242279e-07, "loss": 0.0004, "step": 41076 }, { "epoch": 7.53, "learning_rate": 1.7825612070088262e-07, "loss": 0.0005, "step": 41077 }, { "epoch": 7.53, "learning_rate": 1.7811659171522054e-07, "loss": 0.0003, "step": 41078 }, { "epoch": 7.53, "learning_rate": 1.7797711686801e-07, "loss": 0.0004, "step": 41079 }, { "epoch": 7.53, "learning_rate": 1.7783769616001812e-07, "loss": 0.0002, "step": 41080 }, { "epoch": 7.53, "learning_rate": 1.7769832959201537e-07, "loss": 0.0003, "step": 41081 }, { "epoch": 7.53, "learning_rate": 1.7755901716477008e-07, "loss": 0.0004, "step": 41082 }, { "epoch": 7.53, "learning_rate": 1.774197588790494e-07, "loss": 0.0005, "step": 41083 }, { "epoch": 7.53, "learning_rate": 1.7728055473562044e-07, "loss": 0.0005, "step": 41084 }, { "epoch": 7.53, "learning_rate": 1.7714140473525265e-07, "loss": 0.0004, "step": 41085 }, { "epoch": 7.53, "learning_rate": 1.7700230887871094e-07, "loss": 0.0004, "step": 41086 }, { "epoch": 7.53, "learning_rate": 1.768632671667625e-07, "loss": 0.0006, "step": 41087 }, { "epoch": 7.53, "learning_rate": 1.7672427960017334e-07, "loss": 0.0002, "step": 41088 }, { "epoch": 7.54, "learning_rate": 1.765853461797107e-07, "loss": 0.0005, "step": 41089 }, { "epoch": 7.54, "learning_rate": 1.7644646690613942e-07, "loss": 0.0005, "step": 41090 }, { "epoch": 7.54, "learning_rate": 1.7630764178022676e-07, "loss": 0.0005, "step": 41091 }, { "epoch": 7.54, "learning_rate": 1.7616887080273537e-07, "loss": 0.0005, "step": 41092 }, { "epoch": 7.54, "learning_rate": 1.7603015397443024e-07, "loss": 0.0005, "step": 41093 }, { "epoch": 7.54, "learning_rate": 1.758914912960774e-07, "loss": 0.0006, "step": 41094 }, { "epoch": 7.54, "learning_rate": 1.7575288276844184e-07, "loss": 0.0004, "step": 41095 }, { "epoch": 7.54, "learning_rate": 1.7561432839228732e-07, "loss": 0.0004, "step": 41096 }, { "epoch": 7.54, "learning_rate": 1.754758281683755e-07, "loss": 0.0004, "step": 41097 }, { "epoch": 7.54, "learning_rate": 1.753373820974713e-07, "loss": 0.0005, "step": 41098 }, { "epoch": 7.54, "learning_rate": 1.751989901803375e-07, "loss": 0.0005, "step": 41099 }, { "epoch": 7.54, "learning_rate": 1.7506065241773673e-07, "loss": 0.0002, "step": 41100 }, { "epoch": 7.54, "learning_rate": 1.7492236881043289e-07, "loss": 0.0008, "step": 41101 }, { "epoch": 7.54, "learning_rate": 1.7478413935918648e-07, "loss": 0.0005, "step": 41102 }, { "epoch": 7.54, "learning_rate": 1.746459640647602e-07, "loss": 0.0006, "step": 41103 }, { "epoch": 7.54, "learning_rate": 1.7450784292791456e-07, "loss": 0.0003, "step": 41104 }, { "epoch": 7.54, "learning_rate": 1.743697759494134e-07, "loss": 0.0002, "step": 41105 }, { "epoch": 7.54, "learning_rate": 1.7423176313001499e-07, "loss": 0.0005, "step": 41106 }, { "epoch": 7.54, "learning_rate": 1.7409380447048207e-07, "loss": 0.0006, "step": 41107 }, { "epoch": 7.54, "learning_rate": 1.7395589997157403e-07, "loss": 0.0003, "step": 41108 }, { "epoch": 7.54, "learning_rate": 1.7381804963405135e-07, "loss": 0.0005, "step": 41109 }, { "epoch": 7.54, "learning_rate": 1.7368025345867456e-07, "loss": 0.0005, "step": 41110 }, { "epoch": 7.54, "learning_rate": 1.7354251144620082e-07, "loss": 0.0004, "step": 41111 }, { "epoch": 7.54, "learning_rate": 1.7340482359739175e-07, "loss": 0.0003, "step": 41112 }, { "epoch": 7.54, "learning_rate": 1.7326718991300563e-07, "loss": 0.0004, "step": 41113 }, { "epoch": 7.54, "learning_rate": 1.7312961039380182e-07, "loss": 0.0002, "step": 41114 }, { "epoch": 7.54, "learning_rate": 1.7299208504053533e-07, "loss": 0.0008, "step": 41115 }, { "epoch": 7.54, "learning_rate": 1.7285461385396884e-07, "loss": 0.0003, "step": 41116 }, { "epoch": 7.54, "learning_rate": 1.7271719683485732e-07, "loss": 0.0006, "step": 41117 }, { "epoch": 7.54, "learning_rate": 1.7257983398395906e-07, "loss": 0.0002, "step": 41118 }, { "epoch": 7.54, "learning_rate": 1.7244252530203009e-07, "loss": 0.0004, "step": 41119 }, { "epoch": 7.54, "learning_rate": 1.7230527078982872e-07, "loss": 0.0005, "step": 41120 }, { "epoch": 7.54, "learning_rate": 1.721680704481121e-07, "loss": 0.0003, "step": 41121 }, { "epoch": 7.54, "learning_rate": 1.7203092427763414e-07, "loss": 0.0003, "step": 41122 }, { "epoch": 7.54, "learning_rate": 1.718938322791519e-07, "loss": 0.0005, "step": 41123 }, { "epoch": 7.54, "learning_rate": 1.717567944534204e-07, "loss": 0.0008, "step": 41124 }, { "epoch": 7.54, "learning_rate": 1.7161981080119793e-07, "loss": 0.0006, "step": 41125 }, { "epoch": 7.54, "learning_rate": 1.714828813232361e-07, "loss": 0.0005, "step": 41126 }, { "epoch": 7.54, "learning_rate": 1.7134600602028983e-07, "loss": 0.0005, "step": 41127 }, { "epoch": 7.54, "learning_rate": 1.7120918489311634e-07, "loss": 0.0003, "step": 41128 }, { "epoch": 7.54, "learning_rate": 1.710724179424661e-07, "loss": 0.0004, "step": 41129 }, { "epoch": 7.54, "learning_rate": 1.709357051690963e-07, "loss": 0.0004, "step": 41130 }, { "epoch": 7.54, "learning_rate": 1.7079904657375968e-07, "loss": 0.0003, "step": 41131 }, { "epoch": 7.54, "learning_rate": 1.7066244215720896e-07, "loss": 0.0008, "step": 41132 }, { "epoch": 7.54, "learning_rate": 1.7052589192019576e-07, "loss": 0.0005, "step": 41133 }, { "epoch": 7.54, "learning_rate": 1.7038939586347503e-07, "loss": 0.0003, "step": 41134 }, { "epoch": 7.54, "learning_rate": 1.7025295398779839e-07, "loss": 0.0002, "step": 41135 }, { "epoch": 7.54, "learning_rate": 1.701165662939175e-07, "loss": 0.0005, "step": 41136 }, { "epoch": 7.54, "learning_rate": 1.6998023278258392e-07, "loss": 0.0002, "step": 41137 }, { "epoch": 7.54, "learning_rate": 1.698439534545504e-07, "loss": 0.0003, "step": 41138 }, { "epoch": 7.54, "learning_rate": 1.6970772831056637e-07, "loss": 0.0007, "step": 41139 }, { "epoch": 7.54, "learning_rate": 1.6957155735138454e-07, "loss": 0.0007, "step": 41140 }, { "epoch": 7.54, "learning_rate": 1.694354405777543e-07, "loss": 0.0007, "step": 41141 }, { "epoch": 7.54, "learning_rate": 1.6929937799042507e-07, "loss": 0.0006, "step": 41142 }, { "epoch": 7.55, "learning_rate": 1.6916336959015068e-07, "loss": 0.0003, "step": 41143 }, { "epoch": 7.55, "learning_rate": 1.6902741537767607e-07, "loss": 0.0009, "step": 41144 }, { "epoch": 7.55, "learning_rate": 1.6889151535375292e-07, "loss": 0.0003, "step": 41145 }, { "epoch": 7.55, "learning_rate": 1.6875566951912947e-07, "loss": 0.0004, "step": 41146 }, { "epoch": 7.55, "learning_rate": 1.6861987787455624e-07, "loss": 0.0003, "step": 41147 }, { "epoch": 7.55, "learning_rate": 1.6848414042078042e-07, "loss": 0.0003, "step": 41148 }, { "epoch": 7.55, "learning_rate": 1.683484571585503e-07, "loss": 0.0002, "step": 41149 }, { "epoch": 7.55, "learning_rate": 1.6821282808861416e-07, "loss": 0.0005, "step": 41150 }, { "epoch": 7.55, "learning_rate": 1.6807725321171808e-07, "loss": 0.0006, "step": 41151 }, { "epoch": 7.55, "learning_rate": 1.6794173252861147e-07, "loss": 0.0004, "step": 41152 }, { "epoch": 7.55, "learning_rate": 1.6780626604003925e-07, "loss": 0.0012, "step": 41153 }, { "epoch": 7.55, "learning_rate": 1.6767085374675085e-07, "loss": 0.0005, "step": 41154 }, { "epoch": 7.55, "learning_rate": 1.6753549564949011e-07, "loss": 0.0006, "step": 41155 }, { "epoch": 7.55, "learning_rate": 1.6740019174900424e-07, "loss": 0.0004, "step": 41156 }, { "epoch": 7.55, "learning_rate": 1.6726494204603927e-07, "loss": 0.0005, "step": 41157 }, { "epoch": 7.55, "learning_rate": 1.6712974654133907e-07, "loss": 0.0002, "step": 41158 }, { "epoch": 7.55, "learning_rate": 1.669946052356508e-07, "loss": 0.0003, "step": 41159 }, { "epoch": 7.55, "learning_rate": 1.6685951812971836e-07, "loss": 0.0004, "step": 41160 }, { "epoch": 7.55, "learning_rate": 1.6672448522428775e-07, "loss": 0.0002, "step": 41161 }, { "epoch": 7.55, "learning_rate": 1.6658950652010175e-07, "loss": 0.0005, "step": 41162 }, { "epoch": 7.55, "learning_rate": 1.6645458201790422e-07, "loss": 0.0002, "step": 41163 }, { "epoch": 7.55, "learning_rate": 1.6631971171843898e-07, "loss": 0.0009, "step": 41164 }, { "epoch": 7.55, "learning_rate": 1.6618489562244988e-07, "loss": 0.0002, "step": 41165 }, { "epoch": 7.55, "learning_rate": 1.6605013373068078e-07, "loss": 0.0004, "step": 41166 }, { "epoch": 7.55, "learning_rate": 1.6591542604387445e-07, "loss": 0.0004, "step": 41167 }, { "epoch": 7.55, "learning_rate": 1.6578077256277135e-07, "loss": 0.0007, "step": 41168 }, { "epoch": 7.55, "learning_rate": 1.656461732881154e-07, "loss": 0.0006, "step": 41169 }, { "epoch": 7.55, "learning_rate": 1.6551162822064815e-07, "loss": 0.0002, "step": 41170 }, { "epoch": 7.55, "learning_rate": 1.6537713736111127e-07, "loss": 0.0002, "step": 41171 }, { "epoch": 7.55, "learning_rate": 1.652427007102453e-07, "loss": 0.0003, "step": 41172 }, { "epoch": 7.55, "learning_rate": 1.6510831826879292e-07, "loss": 0.0003, "step": 41173 }, { "epoch": 7.55, "learning_rate": 1.649739900374936e-07, "loss": 0.0003, "step": 41174 }, { "epoch": 7.55, "learning_rate": 1.6483971601708782e-07, "loss": 0.0003, "step": 41175 }, { "epoch": 7.55, "learning_rate": 1.647054962083161e-07, "loss": 0.0016, "step": 41176 }, { "epoch": 7.55, "learning_rate": 1.645713306119179e-07, "loss": 0.0005, "step": 41177 }, { "epoch": 7.55, "learning_rate": 1.6443721922863365e-07, "loss": 0.0004, "step": 41178 }, { "epoch": 7.55, "learning_rate": 1.6430316205920172e-07, "loss": 0.0006, "step": 41179 }, { "epoch": 7.55, "learning_rate": 1.641691591043615e-07, "loss": 0.0003, "step": 41180 }, { "epoch": 7.55, "learning_rate": 1.6403521036484903e-07, "loss": 0.0004, "step": 41181 }, { "epoch": 7.55, "learning_rate": 1.6390131584140712e-07, "loss": 0.0005, "step": 41182 }, { "epoch": 7.55, "learning_rate": 1.637674755347707e-07, "loss": 0.0002, "step": 41183 }, { "epoch": 7.55, "learning_rate": 1.6363368944567915e-07, "loss": 0.0003, "step": 41184 }, { "epoch": 7.55, "learning_rate": 1.634999575748697e-07, "loss": 0.0003, "step": 41185 }, { "epoch": 7.55, "learning_rate": 1.633662799230773e-07, "loss": 0.0004, "step": 41186 }, { "epoch": 7.55, "learning_rate": 1.6323265649104137e-07, "loss": 0.0005, "step": 41187 }, { "epoch": 7.55, "learning_rate": 1.6309908727949686e-07, "loss": 0.0008, "step": 41188 }, { "epoch": 7.55, "learning_rate": 1.6296557228918098e-07, "loss": 0.0004, "step": 41189 }, { "epoch": 7.55, "learning_rate": 1.6283211152082978e-07, "loss": 0.0003, "step": 41190 }, { "epoch": 7.55, "learning_rate": 1.6269870497517715e-07, "loss": 0.0008, "step": 41191 }, { "epoch": 7.55, "learning_rate": 1.6256535265296137e-07, "loss": 0.0014, "step": 41192 }, { "epoch": 7.55, "learning_rate": 1.6243205455491518e-07, "loss": 0.0005, "step": 41193 }, { "epoch": 7.55, "learning_rate": 1.6229881068177355e-07, "loss": 0.0008, "step": 41194 }, { "epoch": 7.55, "learning_rate": 1.6216562103427037e-07, "loss": 0.0009, "step": 41195 }, { "epoch": 7.55, "learning_rate": 1.620324856131428e-07, "loss": 0.0003, "step": 41196 }, { "epoch": 7.55, "learning_rate": 1.6189940441912243e-07, "loss": 0.0003, "step": 41197 }, { "epoch": 7.56, "learning_rate": 1.617663774529421e-07, "loss": 0.0004, "step": 41198 }, { "epoch": 7.56, "learning_rate": 1.6163340471533561e-07, "loss": 0.0006, "step": 41199 }, { "epoch": 7.56, "learning_rate": 1.6150048620703685e-07, "loss": 0.0004, "step": 41200 }, { "epoch": 7.56, "learning_rate": 1.6136762192877853e-07, "loss": 0.0004, "step": 41201 }, { "epoch": 7.56, "learning_rate": 1.612348118812923e-07, "loss": 0.0005, "step": 41202 }, { "epoch": 7.56, "learning_rate": 1.6110205606530981e-07, "loss": 0.0008, "step": 41203 }, { "epoch": 7.56, "learning_rate": 1.609693544815627e-07, "loss": 0.0008, "step": 41204 }, { "epoch": 7.56, "learning_rate": 1.608367071307837e-07, "loss": 0.0002, "step": 41205 }, { "epoch": 7.56, "learning_rate": 1.6070411401370335e-07, "loss": 0.0003, "step": 41206 }, { "epoch": 7.56, "learning_rate": 1.6057157513105215e-07, "loss": 0.0002, "step": 41207 }, { "epoch": 7.56, "learning_rate": 1.6043909048356177e-07, "loss": 0.0011, "step": 41208 }, { "epoch": 7.56, "learning_rate": 1.603066600719605e-07, "loss": 0.0002, "step": 41209 }, { "epoch": 7.56, "learning_rate": 1.6017428389697998e-07, "loss": 0.0003, "step": 41210 }, { "epoch": 7.56, "learning_rate": 1.600419619593496e-07, "loss": 0.0004, "step": 41211 }, { "epoch": 7.56, "learning_rate": 1.5990969425979775e-07, "loss": 0.0005, "step": 41212 }, { "epoch": 7.56, "learning_rate": 1.5977748079905487e-07, "loss": 0.0004, "step": 41213 }, { "epoch": 7.56, "learning_rate": 1.596453215778493e-07, "loss": 0.0006, "step": 41214 }, { "epoch": 7.56, "learning_rate": 1.5951321659690822e-07, "loss": 0.0005, "step": 41215 }, { "epoch": 7.56, "learning_rate": 1.5938116585696216e-07, "loss": 0.0004, "step": 41216 }, { "epoch": 7.56, "learning_rate": 1.5924916935873614e-07, "loss": 0.0006, "step": 41217 }, { "epoch": 7.56, "learning_rate": 1.591172271029595e-07, "loss": 0.0005, "step": 41218 }, { "epoch": 7.56, "learning_rate": 1.5898533909035952e-07, "loss": 0.0003, "step": 41219 }, { "epoch": 7.56, "learning_rate": 1.5885350532166333e-07, "loss": 0.0001, "step": 41220 }, { "epoch": 7.56, "learning_rate": 1.5872172579759703e-07, "loss": 0.0006, "step": 41221 }, { "epoch": 7.56, "learning_rate": 1.5859000051888674e-07, "loss": 0.0002, "step": 41222 }, { "epoch": 7.56, "learning_rate": 1.5845832948625962e-07, "loss": 0.0007, "step": 41223 }, { "epoch": 7.56, "learning_rate": 1.5832671270043952e-07, "loss": 0.0004, "step": 41224 }, { "epoch": 7.56, "learning_rate": 1.581951501621548e-07, "loss": 0.0016, "step": 41225 }, { "epoch": 7.56, "learning_rate": 1.580636418721282e-07, "loss": 0.0006, "step": 41226 }, { "epoch": 7.56, "learning_rate": 1.5793218783108467e-07, "loss": 0.0008, "step": 41227 }, { "epoch": 7.56, "learning_rate": 1.578007880397503e-07, "loss": 0.0008, "step": 41228 }, { "epoch": 7.56, "learning_rate": 1.5766944249884896e-07, "loss": 0.0004, "step": 41229 }, { "epoch": 7.56, "learning_rate": 1.5753815120910232e-07, "loss": 0.0004, "step": 41230 }, { "epoch": 7.56, "learning_rate": 1.5740691417123754e-07, "loss": 0.0002, "step": 41231 }, { "epoch": 7.56, "learning_rate": 1.572757313859774e-07, "loss": 0.0003, "step": 41232 }, { "epoch": 7.56, "learning_rate": 1.5714460285404242e-07, "loss": 0.0009, "step": 41233 }, { "epoch": 7.56, "learning_rate": 1.5701352857615648e-07, "loss": 0.0006, "step": 41234 }, { "epoch": 7.56, "learning_rate": 1.5688250855304343e-07, "loss": 0.0004, "step": 41235 }, { "epoch": 7.56, "learning_rate": 1.5675154278542494e-07, "loss": 0.0004, "step": 41236 }, { "epoch": 7.56, "learning_rate": 1.5662063127402262e-07, "loss": 0.0007, "step": 41237 }, { "epoch": 7.56, "learning_rate": 1.56489774019557e-07, "loss": 0.0005, "step": 41238 }, { "epoch": 7.56, "learning_rate": 1.56358971022752e-07, "loss": 0.0003, "step": 41239 }, { "epoch": 7.56, "learning_rate": 1.5622822228432588e-07, "loss": 0.0005, "step": 41240 }, { "epoch": 7.56, "learning_rate": 1.5609752780500032e-07, "loss": 0.0003, "step": 41241 }, { "epoch": 7.56, "learning_rate": 1.5596688758549582e-07, "loss": 0.001, "step": 41242 }, { "epoch": 7.56, "learning_rate": 1.5583630162653296e-07, "loss": 0.0006, "step": 41243 }, { "epoch": 7.56, "learning_rate": 1.5570576992883e-07, "loss": 0.0007, "step": 41244 }, { "epoch": 7.56, "learning_rate": 1.5557529249310866e-07, "loss": 0.0003, "step": 41245 }, { "epoch": 7.56, "learning_rate": 1.5544486932008608e-07, "loss": 0.0003, "step": 41246 }, { "epoch": 7.56, "learning_rate": 1.553145004104828e-07, "loss": 0.0003, "step": 41247 }, { "epoch": 7.56, "learning_rate": 1.5518418576501494e-07, "loss": 0.0006, "step": 41248 }, { "epoch": 7.56, "learning_rate": 1.5505392538440412e-07, "loss": 0.0002, "step": 41249 }, { "epoch": 7.56, "learning_rate": 1.5492371926936755e-07, "loss": 0.0009, "step": 41250 }, { "epoch": 7.56, "learning_rate": 1.5479356742062025e-07, "loss": 0.0002, "step": 41251 }, { "epoch": 7.57, "learning_rate": 1.546634698388816e-07, "loss": 0.0004, "step": 41252 }, { "epoch": 7.57, "learning_rate": 1.545334265248688e-07, "loss": 0.0006, "step": 41253 }, { "epoch": 7.57, "learning_rate": 1.5440343747929908e-07, "loss": 0.0004, "step": 41254 }, { "epoch": 7.57, "learning_rate": 1.5427350270288744e-07, "loss": 0.0005, "step": 41255 }, { "epoch": 7.57, "learning_rate": 1.5414362219635104e-07, "loss": 0.0004, "step": 41256 }, { "epoch": 7.57, "learning_rate": 1.54013795960406e-07, "loss": 0.0002, "step": 41257 }, { "epoch": 7.57, "learning_rate": 1.538840239957684e-07, "loss": 0.0006, "step": 41258 }, { "epoch": 7.57, "learning_rate": 1.5375430630315101e-07, "loss": 0.0003, "step": 41259 }, { "epoch": 7.57, "learning_rate": 1.5362464288327217e-07, "loss": 0.0002, "step": 41260 }, { "epoch": 7.57, "learning_rate": 1.534950337368435e-07, "loss": 0.0005, "step": 41261 }, { "epoch": 7.57, "learning_rate": 1.5336547886458218e-07, "loss": 0.0006, "step": 41262 }, { "epoch": 7.57, "learning_rate": 1.5323597826719994e-07, "loss": 0.0005, "step": 41263 }, { "epoch": 7.57, "learning_rate": 1.5310653194541168e-07, "loss": 0.0004, "step": 41264 }, { "epoch": 7.57, "learning_rate": 1.5297713989993135e-07, "loss": 0.0003, "step": 41265 }, { "epoch": 7.57, "learning_rate": 1.5284780213147165e-07, "loss": 0.0006, "step": 41266 }, { "epoch": 7.57, "learning_rate": 1.5271851864074538e-07, "loss": 0.0004, "step": 41267 }, { "epoch": 7.57, "learning_rate": 1.525892894284664e-07, "loss": 0.0003, "step": 41268 }, { "epoch": 7.57, "learning_rate": 1.5246011449534526e-07, "loss": 0.0003, "step": 41269 }, { "epoch": 7.57, "learning_rate": 1.523309938420936e-07, "loss": 0.0002, "step": 41270 }, { "epoch": 7.57, "learning_rate": 1.522019274694253e-07, "loss": 0.0007, "step": 41271 }, { "epoch": 7.57, "learning_rate": 1.5207291537805092e-07, "loss": 0.0004, "step": 41272 }, { "epoch": 7.57, "learning_rate": 1.5194395756868096e-07, "loss": 0.0003, "step": 41273 }, { "epoch": 7.57, "learning_rate": 1.5181505404202601e-07, "loss": 0.0004, "step": 41274 }, { "epoch": 7.57, "learning_rate": 1.516862047987988e-07, "loss": 0.0002, "step": 41275 }, { "epoch": 7.57, "learning_rate": 1.5155740983970658e-07, "loss": 0.0008, "step": 41276 }, { "epoch": 7.57, "learning_rate": 1.5142866916546094e-07, "loss": 0.0006, "step": 41277 }, { "epoch": 7.57, "learning_rate": 1.5129998277677138e-07, "loss": 0.0004, "step": 41278 }, { "epoch": 7.57, "learning_rate": 1.511713506743473e-07, "loss": 0.0004, "step": 41279 }, { "epoch": 7.57, "learning_rate": 1.5104277285889702e-07, "loss": 0.0007, "step": 41280 }, { "epoch": 7.57, "learning_rate": 1.5091424933112998e-07, "loss": 0.0005, "step": 41281 }, { "epoch": 7.57, "learning_rate": 1.5078578009175337e-07, "loss": 0.0003, "step": 41282 }, { "epoch": 7.57, "learning_rate": 1.5065736514147665e-07, "loss": 0.0003, "step": 41283 }, { "epoch": 7.57, "learning_rate": 1.5052900448100815e-07, "loss": 0.0004, "step": 41284 }, { "epoch": 7.57, "learning_rate": 1.5040069811105395e-07, "loss": 0.0013, "step": 41285 }, { "epoch": 7.57, "learning_rate": 1.502724460323224e-07, "loss": 0.0004, "step": 41286 }, { "epoch": 7.57, "learning_rate": 1.5014424824551954e-07, "loss": 0.0005, "step": 41287 }, { "epoch": 7.57, "learning_rate": 1.5001610475135154e-07, "loss": 0.0008, "step": 41288 }, { "epoch": 7.57, "learning_rate": 1.4988801555052668e-07, "loss": 0.0002, "step": 41289 }, { "epoch": 7.57, "learning_rate": 1.4975998064374998e-07, "loss": 0.0008, "step": 41290 }, { "epoch": 7.57, "learning_rate": 1.4963200003172639e-07, "loss": 0.0002, "step": 41291 }, { "epoch": 7.57, "learning_rate": 1.4950407371516208e-07, "loss": 0.0007, "step": 41292 }, { "epoch": 7.57, "learning_rate": 1.4937620169476198e-07, "loss": 0.0004, "step": 41293 }, { "epoch": 7.57, "learning_rate": 1.4924838397123109e-07, "loss": 0.0003, "step": 41294 }, { "epoch": 7.57, "learning_rate": 1.4912062054527333e-07, "loss": 0.0004, "step": 41295 }, { "epoch": 7.57, "learning_rate": 1.4899291141759363e-07, "loss": 0.0004, "step": 41296 }, { "epoch": 7.57, "learning_rate": 1.4886525658889707e-07, "loss": 0.0002, "step": 41297 }, { "epoch": 7.57, "learning_rate": 1.4873765605988522e-07, "loss": 0.0003, "step": 41298 }, { "epoch": 7.57, "learning_rate": 1.4861010983126202e-07, "loss": 0.0004, "step": 41299 }, { "epoch": 7.57, "learning_rate": 1.484826179037313e-07, "loss": 0.0005, "step": 41300 }, { "epoch": 7.57, "learning_rate": 1.4835518027799367e-07, "loss": 0.0005, "step": 41301 }, { "epoch": 7.57, "learning_rate": 1.482277969547541e-07, "loss": 0.0003, "step": 41302 }, { "epoch": 7.57, "learning_rate": 1.4810046793471532e-07, "loss": 0.0002, "step": 41303 }, { "epoch": 7.57, "learning_rate": 1.4797319321857572e-07, "loss": 0.0003, "step": 41304 }, { "epoch": 7.57, "learning_rate": 1.4784597280703805e-07, "loss": 0.0002, "step": 41305 }, { "epoch": 7.57, "learning_rate": 1.4771880670080508e-07, "loss": 0.0003, "step": 41306 }, { "epoch": 7.58, "learning_rate": 1.4759169490057733e-07, "loss": 0.0006, "step": 41307 }, { "epoch": 7.58, "learning_rate": 1.474646374070543e-07, "loss": 0.0004, "step": 41308 }, { "epoch": 7.58, "learning_rate": 1.473376342209376e-07, "loss": 0.0003, "step": 41309 }, { "epoch": 7.58, "learning_rate": 1.472106853429256e-07, "loss": 0.0003, "step": 41310 }, { "epoch": 7.58, "learning_rate": 1.4708379077371992e-07, "loss": 0.0004, "step": 41311 }, { "epoch": 7.58, "learning_rate": 1.4695695051402003e-07, "loss": 0.0008, "step": 41312 }, { "epoch": 7.58, "learning_rate": 1.4683016456452314e-07, "loss": 0.0004, "step": 41313 }, { "epoch": 7.58, "learning_rate": 1.467034329259287e-07, "loss": 0.001, "step": 41314 }, { "epoch": 7.58, "learning_rate": 1.4657675559893725e-07, "loss": 0.0005, "step": 41315 }, { "epoch": 7.58, "learning_rate": 1.4645013258424489e-07, "loss": 0.0003, "step": 41316 }, { "epoch": 7.58, "learning_rate": 1.4632356388254998e-07, "loss": 0.0004, "step": 41317 }, { "epoch": 7.58, "learning_rate": 1.461970494945497e-07, "loss": 0.0005, "step": 41318 }, { "epoch": 7.58, "learning_rate": 1.4607058942094353e-07, "loss": 0.0007, "step": 41319 }, { "epoch": 7.58, "learning_rate": 1.4594418366242646e-07, "loss": 0.0004, "step": 41320 }, { "epoch": 7.58, "learning_rate": 1.458178322196968e-07, "loss": 0.0008, "step": 41321 }, { "epoch": 7.58, "learning_rate": 1.4569153509344958e-07, "loss": 0.0006, "step": 41322 }, { "epoch": 7.58, "learning_rate": 1.4556529228438088e-07, "loss": 0.0004, "step": 41323 }, { "epoch": 7.58, "learning_rate": 1.4543910379318792e-07, "loss": 0.0004, "step": 41324 }, { "epoch": 7.58, "learning_rate": 1.4531296962056463e-07, "loss": 0.0004, "step": 41325 }, { "epoch": 7.58, "learning_rate": 1.4518688976720817e-07, "loss": 0.0003, "step": 41326 }, { "epoch": 7.58, "learning_rate": 1.450608642338114e-07, "loss": 0.0005, "step": 41327 }, { "epoch": 7.58, "learning_rate": 1.4493489302107146e-07, "loss": 0.0013, "step": 41328 }, { "epoch": 7.58, "learning_rate": 1.4480897612968003e-07, "loss": 0.0002, "step": 41329 }, { "epoch": 7.58, "learning_rate": 1.4468311356033216e-07, "loss": 0.0006, "step": 41330 }, { "epoch": 7.58, "learning_rate": 1.445573053137228e-07, "loss": 0.0008, "step": 41331 }, { "epoch": 7.58, "learning_rate": 1.4443155139054364e-07, "loss": 0.0003, "step": 41332 }, { "epoch": 7.58, "learning_rate": 1.443058517914897e-07, "loss": 0.0004, "step": 41333 }, { "epoch": 7.58, "learning_rate": 1.4418020651725262e-07, "loss": 0.0005, "step": 41334 }, { "epoch": 7.58, "learning_rate": 1.4405461556852408e-07, "loss": 0.0007, "step": 41335 }, { "epoch": 7.58, "learning_rate": 1.4392907894599796e-07, "loss": 0.0007, "step": 41336 }, { "epoch": 7.58, "learning_rate": 1.4380359665036592e-07, "loss": 0.0003, "step": 41337 }, { "epoch": 7.58, "learning_rate": 1.4367816868231855e-07, "loss": 0.0004, "step": 41338 }, { "epoch": 7.58, "learning_rate": 1.435527950425497e-07, "loss": 0.0003, "step": 41339 }, { "epoch": 7.58, "learning_rate": 1.4342747573174775e-07, "loss": 0.0011, "step": 41340 }, { "epoch": 7.58, "learning_rate": 1.4330221075060435e-07, "loss": 0.0009, "step": 41341 }, { "epoch": 7.58, "learning_rate": 1.4317700009981006e-07, "loss": 0.0003, "step": 41342 }, { "epoch": 7.58, "learning_rate": 1.4305184378005542e-07, "loss": 0.0004, "step": 41343 }, { "epoch": 7.58, "learning_rate": 1.4292674179202992e-07, "loss": 0.0007, "step": 41344 }, { "epoch": 7.58, "learning_rate": 1.4280169413642298e-07, "loss": 0.0007, "step": 41345 }, { "epoch": 7.58, "learning_rate": 1.4267670081392404e-07, "loss": 0.0008, "step": 41346 }, { "epoch": 7.58, "learning_rate": 1.4255176182522145e-07, "loss": 0.0004, "step": 41347 }, { "epoch": 7.58, "learning_rate": 1.4242687717100466e-07, "loss": 0.0007, "step": 41348 }, { "epoch": 7.58, "learning_rate": 1.4230204685196202e-07, "loss": 0.0007, "step": 41349 }, { "epoch": 7.58, "learning_rate": 1.4217727086878297e-07, "loss": 0.0004, "step": 41350 }, { "epoch": 7.58, "learning_rate": 1.420525492221525e-07, "loss": 0.0003, "step": 41351 }, { "epoch": 7.58, "learning_rate": 1.4192788191275896e-07, "loss": 0.0005, "step": 41352 }, { "epoch": 7.58, "learning_rate": 1.418032689412896e-07, "loss": 0.0003, "step": 41353 }, { "epoch": 7.58, "learning_rate": 1.416787103084316e-07, "loss": 0.0004, "step": 41354 }, { "epoch": 7.58, "learning_rate": 1.4155420601487224e-07, "loss": 0.0003, "step": 41355 }, { "epoch": 7.58, "learning_rate": 1.414297560612976e-07, "loss": 0.0003, "step": 41356 }, { "epoch": 7.58, "learning_rate": 1.413053604483927e-07, "loss": 0.0002, "step": 41357 }, { "epoch": 7.58, "learning_rate": 1.4118101917684257e-07, "loss": 0.0007, "step": 41358 }, { "epoch": 7.58, "learning_rate": 1.410567322473344e-07, "loss": 0.0002, "step": 41359 }, { "epoch": 7.58, "learning_rate": 1.4093249966055212e-07, "loss": 0.0002, "step": 41360 }, { "epoch": 7.58, "learning_rate": 1.408083214171818e-07, "loss": 0.0004, "step": 41361 }, { "epoch": 7.59, "learning_rate": 1.406841975179063e-07, "loss": 0.0005, "step": 41362 }, { "epoch": 7.59, "learning_rate": 1.4056012796341056e-07, "loss": 0.0003, "step": 41363 }, { "epoch": 7.59, "learning_rate": 1.404361127543785e-07, "loss": 0.0003, "step": 41364 }, { "epoch": 7.59, "learning_rate": 1.4031215189149406e-07, "loss": 0.0006, "step": 41365 }, { "epoch": 7.59, "learning_rate": 1.4018824537543886e-07, "loss": 0.0002, "step": 41366 }, { "epoch": 7.59, "learning_rate": 1.400643932068968e-07, "loss": 0.0005, "step": 41367 }, { "epoch": 7.59, "learning_rate": 1.3994059538655292e-07, "loss": 0.0006, "step": 41368 }, { "epoch": 7.59, "learning_rate": 1.3981685191508554e-07, "loss": 0.0003, "step": 41369 }, { "epoch": 7.59, "learning_rate": 1.396931627931797e-07, "loss": 0.0004, "step": 41370 }, { "epoch": 7.59, "learning_rate": 1.395695280215148e-07, "loss": 0.0003, "step": 41371 }, { "epoch": 7.59, "learning_rate": 1.3944594760077478e-07, "loss": 0.0003, "step": 41372 }, { "epoch": 7.59, "learning_rate": 1.3932242153164023e-07, "loss": 0.0003, "step": 41373 }, { "epoch": 7.59, "learning_rate": 1.3919894981479053e-07, "loss": 0.0004, "step": 41374 }, { "epoch": 7.59, "learning_rate": 1.3907553245090856e-07, "loss": 0.0003, "step": 41375 }, { "epoch": 7.59, "learning_rate": 1.3895216944067147e-07, "loss": 0.0007, "step": 41376 }, { "epoch": 7.59, "learning_rate": 1.38828860784761e-07, "loss": 0.0003, "step": 41377 }, { "epoch": 7.59, "learning_rate": 1.3870560648385878e-07, "loss": 0.0002, "step": 41378 }, { "epoch": 7.59, "learning_rate": 1.3858240653864096e-07, "loss": 0.0003, "step": 41379 }, { "epoch": 7.59, "learning_rate": 1.3845926094978813e-07, "loss": 0.0006, "step": 41380 }, { "epoch": 7.59, "learning_rate": 1.383361697179786e-07, "loss": 0.0005, "step": 41381 }, { "epoch": 7.59, "learning_rate": 1.3821313284389183e-07, "loss": 0.0005, "step": 41382 }, { "epoch": 7.59, "learning_rate": 1.3809015032820505e-07, "loss": 0.0011, "step": 41383 }, { "epoch": 7.59, "learning_rate": 1.3796722217159664e-07, "loss": 0.0007, "step": 41384 }, { "epoch": 7.59, "learning_rate": 1.378443483747427e-07, "loss": 0.001, "step": 41385 }, { "epoch": 7.59, "learning_rate": 1.377215289383227e-07, "loss": 0.0009, "step": 41386 }, { "epoch": 7.59, "learning_rate": 1.3759876386301275e-07, "loss": 0.0003, "step": 41387 }, { "epoch": 7.59, "learning_rate": 1.3747605314949008e-07, "loss": 0.0009, "step": 41388 }, { "epoch": 7.59, "learning_rate": 1.3735339679842864e-07, "loss": 0.0012, "step": 41389 }, { "epoch": 7.59, "learning_rate": 1.3723079481050781e-07, "loss": 0.0003, "step": 41390 }, { "epoch": 7.59, "learning_rate": 1.371082471864016e-07, "loss": 0.0003, "step": 41391 }, { "epoch": 7.59, "learning_rate": 1.3698575392678492e-07, "loss": 0.001, "step": 41392 }, { "epoch": 7.59, "learning_rate": 1.3686331503233506e-07, "loss": 0.0005, "step": 41393 }, { "epoch": 7.59, "learning_rate": 1.3674093050372483e-07, "loss": 0.0002, "step": 41394 }, { "epoch": 7.59, "learning_rate": 1.3661860034163032e-07, "loss": 0.0003, "step": 41395 }, { "epoch": 7.59, "learning_rate": 1.3649632454672545e-07, "loss": 0.0005, "step": 41396 }, { "epoch": 7.59, "learning_rate": 1.3637410311968303e-07, "loss": 0.0003, "step": 41397 }, { "epoch": 7.59, "learning_rate": 1.3625193606117804e-07, "loss": 0.0002, "step": 41398 }, { "epoch": 7.59, "learning_rate": 1.3612982337188329e-07, "loss": 0.0007, "step": 41399 }, { "epoch": 7.59, "learning_rate": 1.3600776505247272e-07, "loss": 0.0008, "step": 41400 }, { "epoch": 7.59, "learning_rate": 1.3588576110361795e-07, "loss": 0.0007, "step": 41401 }, { "epoch": 7.59, "learning_rate": 1.3576381152599073e-07, "loss": 0.0002, "step": 41402 }, { "epoch": 7.59, "learning_rate": 1.3564191632026602e-07, "loss": 0.0006, "step": 41403 }, { "epoch": 7.59, "learning_rate": 1.3552007548711443e-07, "loss": 0.0004, "step": 41404 }, { "epoch": 7.59, "learning_rate": 1.3539828902720653e-07, "loss": 0.0009, "step": 41405 }, { "epoch": 7.59, "learning_rate": 1.35276556941214e-07, "loss": 0.0003, "step": 41406 }, { "epoch": 7.59, "learning_rate": 1.3515487922980851e-07, "loss": 0.0002, "step": 41407 }, { "epoch": 7.59, "learning_rate": 1.3503325589366067e-07, "loss": 0.0016, "step": 41408 }, { "epoch": 7.59, "learning_rate": 1.3491168693343993e-07, "loss": 0.0003, "step": 41409 }, { "epoch": 7.59, "learning_rate": 1.3479017234981796e-07, "loss": 0.0003, "step": 41410 }, { "epoch": 7.59, "learning_rate": 1.346687121434642e-07, "loss": 0.0001, "step": 41411 }, { "epoch": 7.59, "learning_rate": 1.3454730631504598e-07, "loss": 0.0003, "step": 41412 }, { "epoch": 7.59, "learning_rate": 1.34425954865236e-07, "loss": 0.0004, "step": 41413 }, { "epoch": 7.59, "learning_rate": 1.3430465779469937e-07, "loss": 0.0015, "step": 41414 }, { "epoch": 7.59, "learning_rate": 1.341834151041077e-07, "loss": 0.0007, "step": 41415 }, { "epoch": 7.6, "learning_rate": 1.3406222679412827e-07, "loss": 0.0002, "step": 41416 }, { "epoch": 7.6, "learning_rate": 1.3394109286542834e-07, "loss": 0.0003, "step": 41417 }, { "epoch": 7.6, "learning_rate": 1.3382001331867734e-07, "loss": 0.0004, "step": 41418 }, { "epoch": 7.6, "learning_rate": 1.336989881545403e-07, "loss": 0.0002, "step": 41419 }, { "epoch": 7.6, "learning_rate": 1.335780173736856e-07, "loss": 0.0004, "step": 41420 }, { "epoch": 7.6, "learning_rate": 1.3345710097678044e-07, "loss": 0.0006, "step": 41421 }, { "epoch": 7.6, "learning_rate": 1.333362389644921e-07, "loss": 0.0011, "step": 41422 }, { "epoch": 7.6, "learning_rate": 1.332154313374845e-07, "loss": 0.0003, "step": 41423 }, { "epoch": 7.6, "learning_rate": 1.3309467809642373e-07, "loss": 0.0008, "step": 41424 }, { "epoch": 7.6, "learning_rate": 1.3297397924197708e-07, "loss": 0.0005, "step": 41425 }, { "epoch": 7.6, "learning_rate": 1.3285333477480843e-07, "loss": 0.0003, "step": 41426 }, { "epoch": 7.6, "learning_rate": 1.3273274469558396e-07, "loss": 0.0004, "step": 41427 }, { "epoch": 7.6, "learning_rate": 1.3261220900496863e-07, "loss": 0.0003, "step": 41428 }, { "epoch": 7.6, "learning_rate": 1.3249172770362416e-07, "loss": 0.0002, "step": 41429 }, { "epoch": 7.6, "learning_rate": 1.3237130079221784e-07, "loss": 0.0006, "step": 41430 }, { "epoch": 7.6, "learning_rate": 1.322509282714113e-07, "loss": 0.0007, "step": 41431 }, { "epoch": 7.6, "learning_rate": 1.3213061014186957e-07, "loss": 0.0003, "step": 41432 }, { "epoch": 7.6, "learning_rate": 1.3201034640425438e-07, "loss": 0.0005, "step": 41433 }, { "epoch": 7.6, "learning_rate": 1.318901370592296e-07, "loss": 0.0003, "step": 41434 }, { "epoch": 7.6, "learning_rate": 1.3176998210745805e-07, "loss": 0.0006, "step": 41435 }, { "epoch": 7.6, "learning_rate": 1.3164988154960035e-07, "loss": 0.0006, "step": 41436 }, { "epoch": 7.6, "learning_rate": 1.3152983538632037e-07, "loss": 0.0002, "step": 41437 }, { "epoch": 7.6, "learning_rate": 1.314098436182787e-07, "loss": 0.0008, "step": 41438 }, { "epoch": 7.6, "learning_rate": 1.3128990624613812e-07, "loss": 0.0012, "step": 41439 }, { "epoch": 7.6, "learning_rate": 1.3117002327055927e-07, "loss": 0.0006, "step": 41440 }, { "epoch": 7.6, "learning_rate": 1.3105019469220047e-07, "loss": 0.0004, "step": 41441 }, { "epoch": 7.6, "learning_rate": 1.3093042051172456e-07, "loss": 0.0005, "step": 41442 }, { "epoch": 7.6, "learning_rate": 1.3081070072979206e-07, "loss": 0.0008, "step": 41443 }, { "epoch": 7.6, "learning_rate": 1.3069103534706252e-07, "loss": 0.0004, "step": 41444 }, { "epoch": 7.6, "learning_rate": 1.3057142436419533e-07, "loss": 0.0002, "step": 41445 }, { "epoch": 7.6, "learning_rate": 1.3045186778184892e-07, "loss": 0.0004, "step": 41446 }, { "epoch": 7.6, "learning_rate": 1.3033236560068385e-07, "loss": 0.0004, "step": 41447 }, { "epoch": 7.6, "learning_rate": 1.3021291782135736e-07, "loss": 0.0004, "step": 41448 }, { "epoch": 7.6, "learning_rate": 1.3009352444452783e-07, "loss": 0.0003, "step": 41449 }, { "epoch": 7.6, "learning_rate": 1.2997418547085472e-07, "loss": 0.0006, "step": 41450 }, { "epoch": 7.6, "learning_rate": 1.2985490090099527e-07, "loss": 0.0002, "step": 41451 }, { "epoch": 7.6, "learning_rate": 1.2973567073560678e-07, "loss": 0.0005, "step": 41452 }, { "epoch": 7.6, "learning_rate": 1.2961649497534645e-07, "loss": 0.0006, "step": 41453 }, { "epoch": 7.6, "learning_rate": 1.2949737362087156e-07, "loss": 0.0006, "step": 41454 }, { "epoch": 7.6, "learning_rate": 1.2937830667283713e-07, "loss": 0.0004, "step": 41455 }, { "epoch": 7.6, "learning_rate": 1.2925929413190263e-07, "loss": 0.0004, "step": 41456 }, { "epoch": 7.6, "learning_rate": 1.2914033599872088e-07, "loss": 0.0004, "step": 41457 }, { "epoch": 7.6, "learning_rate": 1.2902143227395026e-07, "loss": 0.0004, "step": 41458 }, { "epoch": 7.6, "learning_rate": 1.2890258295824355e-07, "loss": 0.0005, "step": 41459 }, { "epoch": 7.6, "learning_rate": 1.287837880522558e-07, "loss": 0.0003, "step": 41460 }, { "epoch": 7.6, "learning_rate": 1.2866504755664534e-07, "loss": 0.0002, "step": 41461 }, { "epoch": 7.6, "learning_rate": 1.285463614720628e-07, "loss": 0.0004, "step": 41462 }, { "epoch": 7.6, "learning_rate": 1.284277297991654e-07, "loss": 0.0004, "step": 41463 }, { "epoch": 7.6, "learning_rate": 1.283091525386049e-07, "loss": 0.0004, "step": 41464 }, { "epoch": 7.6, "learning_rate": 1.2819062969103512e-07, "loss": 0.0011, "step": 41465 }, { "epoch": 7.6, "learning_rate": 1.280721612571101e-07, "loss": 0.0002, "step": 41466 }, { "epoch": 7.6, "learning_rate": 1.2795374723748365e-07, "loss": 0.0007, "step": 41467 }, { "epoch": 7.6, "learning_rate": 1.2783538763280535e-07, "loss": 0.0003, "step": 41468 }, { "epoch": 7.6, "learning_rate": 1.2771708244373238e-07, "loss": 0.0004, "step": 41469 }, { "epoch": 7.6, "learning_rate": 1.2759883167091203e-07, "loss": 0.0003, "step": 41470 }, { "epoch": 7.61, "learning_rate": 1.2748063531499933e-07, "loss": 0.0004, "step": 41471 }, { "epoch": 7.61, "learning_rate": 1.2736249337664375e-07, "loss": 0.0008, "step": 41472 }, { "epoch": 7.61, "learning_rate": 1.272444058564981e-07, "loss": 0.0013, "step": 41473 }, { "epoch": 7.61, "learning_rate": 1.271263727552119e-07, "loss": 0.0007, "step": 41474 }, { "epoch": 7.61, "learning_rate": 1.270083940734379e-07, "loss": 0.0005, "step": 41475 }, { "epoch": 7.61, "learning_rate": 1.268904698118234e-07, "loss": 0.0006, "step": 41476 }, { "epoch": 7.61, "learning_rate": 1.267725999710201e-07, "loss": 0.0003, "step": 41477 }, { "epoch": 7.61, "learning_rate": 1.2665478455167746e-07, "loss": 0.0004, "step": 41478 }, { "epoch": 7.61, "learning_rate": 1.2653702355444608e-07, "loss": 0.0006, "step": 41479 }, { "epoch": 7.61, "learning_rate": 1.2641931697997323e-07, "loss": 0.0005, "step": 41480 }, { "epoch": 7.61, "learning_rate": 1.263016648289084e-07, "loss": 0.0005, "step": 41481 }, { "epoch": 7.61, "learning_rate": 1.261840671018999e-07, "loss": 0.0005, "step": 41482 }, { "epoch": 7.61, "learning_rate": 1.2606652379959617e-07, "loss": 0.0003, "step": 41483 }, { "epoch": 7.61, "learning_rate": 1.2594903492264554e-07, "loss": 0.0003, "step": 41484 }, { "epoch": 7.61, "learning_rate": 1.258316004716953e-07, "loss": 0.0004, "step": 41485 }, { "epoch": 7.61, "learning_rate": 1.2571422044739157e-07, "loss": 0.0004, "step": 41486 }, { "epoch": 7.61, "learning_rate": 1.2559689485038383e-07, "loss": 0.0007, "step": 41487 }, { "epoch": 7.61, "learning_rate": 1.2547962368131605e-07, "loss": 0.0007, "step": 41488 }, { "epoch": 7.61, "learning_rate": 1.2536240694083658e-07, "loss": 0.0002, "step": 41489 }, { "epoch": 7.61, "learning_rate": 1.2524524462959042e-07, "loss": 0.0003, "step": 41490 }, { "epoch": 7.61, "learning_rate": 1.2512813674822377e-07, "loss": 0.0003, "step": 41491 }, { "epoch": 7.61, "learning_rate": 1.2501108329738275e-07, "loss": 0.0003, "step": 41492 }, { "epoch": 7.61, "learning_rate": 1.2489408427771242e-07, "loss": 0.0004, "step": 41493 }, { "epoch": 7.61, "learning_rate": 1.2477713968985562e-07, "loss": 0.0006, "step": 41494 }, { "epoch": 7.61, "learning_rate": 1.2466024953445953e-07, "loss": 0.0005, "step": 41495 }, { "epoch": 7.61, "learning_rate": 1.2454341381216705e-07, "loss": 0.0004, "step": 41496 }, { "epoch": 7.61, "learning_rate": 1.244266325236232e-07, "loss": 0.0007, "step": 41497 }, { "epoch": 7.61, "learning_rate": 1.2430990566947076e-07, "loss": 0.0006, "step": 41498 }, { "epoch": 7.61, "learning_rate": 1.241932332503537e-07, "loss": 0.0002, "step": 41499 }, { "epoch": 7.61, "learning_rate": 1.2407661526691484e-07, "loss": 0.0002, "step": 41500 }, { "epoch": 7.61, "learning_rate": 1.2396005171979807e-07, "loss": 0.0005, "step": 41501 }, { "epoch": 7.61, "learning_rate": 1.2384354260964292e-07, "loss": 0.0006, "step": 41502 }, { "epoch": 7.61, "learning_rate": 1.2372708793709553e-07, "loss": 0.0006, "step": 41503 }, { "epoch": 7.61, "learning_rate": 1.2361068770279428e-07, "loss": 0.0003, "step": 41504 }, { "epoch": 7.61, "learning_rate": 1.2349434190738307e-07, "loss": 0.0002, "step": 41505 }, { "epoch": 7.61, "learning_rate": 1.2337805055150144e-07, "loss": 0.0003, "step": 41506 }, { "epoch": 7.61, "learning_rate": 1.2326181363579216e-07, "loss": 0.0004, "step": 41507 }, { "epoch": 7.61, "learning_rate": 1.2314563116089472e-07, "loss": 0.0004, "step": 41508 }, { "epoch": 7.61, "learning_rate": 1.2302950312745088e-07, "loss": 0.0003, "step": 41509 }, { "epoch": 7.61, "learning_rate": 1.2291342953609896e-07, "loss": 0.0002, "step": 41510 }, { "epoch": 7.61, "learning_rate": 1.2279741038748072e-07, "loss": 0.0009, "step": 41511 }, { "epoch": 7.61, "learning_rate": 1.2268144568223338e-07, "loss": 0.0003, "step": 41512 }, { "epoch": 7.61, "learning_rate": 1.2256553542099758e-07, "loss": 0.0006, "step": 41513 }, { "epoch": 7.61, "learning_rate": 1.2244967960441168e-07, "loss": 0.0007, "step": 41514 }, { "epoch": 7.61, "learning_rate": 1.2233387823311516e-07, "loss": 0.0002, "step": 41515 }, { "epoch": 7.61, "learning_rate": 1.222181313077453e-07, "loss": 0.0005, "step": 41516 }, { "epoch": 7.61, "learning_rate": 1.2210243882894047e-07, "loss": 0.0005, "step": 41517 }, { "epoch": 7.61, "learning_rate": 1.2198680079733904e-07, "loss": 0.0006, "step": 41518 }, { "epoch": 7.61, "learning_rate": 1.218712172135772e-07, "loss": 0.0003, "step": 41519 }, { "epoch": 7.61, "learning_rate": 1.217556880782933e-07, "loss": 0.0002, "step": 41520 }, { "epoch": 7.61, "learning_rate": 1.2164021339212128e-07, "loss": 0.0004, "step": 41521 }, { "epoch": 7.61, "learning_rate": 1.2152479315570288e-07, "loss": 0.0005, "step": 41522 }, { "epoch": 7.61, "learning_rate": 1.2140942736966975e-07, "loss": 0.0007, "step": 41523 }, { "epoch": 7.61, "learning_rate": 1.2129411603465924e-07, "loss": 0.0003, "step": 41524 }, { "epoch": 7.62, "learning_rate": 1.2117885915130745e-07, "loss": 0.0003, "step": 41525 }, { "epoch": 7.62, "learning_rate": 1.2106365672024834e-07, "loss": 0.0004, "step": 41526 }, { "epoch": 7.62, "learning_rate": 1.2094850874211807e-07, "loss": 0.0004, "step": 41527 }, { "epoch": 7.62, "learning_rate": 1.2083341521755166e-07, "loss": 0.0004, "step": 41528 }, { "epoch": 7.62, "learning_rate": 1.2071837614718306e-07, "loss": 0.0004, "step": 41529 }, { "epoch": 7.62, "learning_rate": 1.2060339153164513e-07, "loss": 0.0005, "step": 41530 }, { "epoch": 7.62, "learning_rate": 1.20488461371574e-07, "loss": 0.0003, "step": 41531 }, { "epoch": 7.62, "learning_rate": 1.2037358566760137e-07, "loss": 0.0004, "step": 41532 }, { "epoch": 7.62, "learning_rate": 1.202587644203612e-07, "loss": 0.0003, "step": 41533 }, { "epoch": 7.62, "learning_rate": 1.2014399763048522e-07, "loss": 0.0009, "step": 41534 }, { "epoch": 7.62, "learning_rate": 1.2002928529860847e-07, "loss": 0.0006, "step": 41535 }, { "epoch": 7.62, "learning_rate": 1.199146274253604e-07, "loss": 0.0003, "step": 41536 }, { "epoch": 7.62, "learning_rate": 1.1980002401137503e-07, "loss": 0.0006, "step": 41537 }, { "epoch": 7.62, "learning_rate": 1.19685475057284e-07, "loss": 0.0005, "step": 41538 }, { "epoch": 7.62, "learning_rate": 1.1957098056371685e-07, "loss": 0.0002, "step": 41539 }, { "epoch": 7.62, "learning_rate": 1.194565405313064e-07, "loss": 0.0008, "step": 41540 }, { "epoch": 7.62, "learning_rate": 1.1934215496068324e-07, "loss": 0.0004, "step": 41541 }, { "epoch": 7.62, "learning_rate": 1.19227823852478e-07, "loss": 0.0004, "step": 41542 }, { "epoch": 7.62, "learning_rate": 1.1911354720731904e-07, "loss": 0.0004, "step": 41543 }, { "epoch": 7.62, "learning_rate": 1.1899932502583811e-07, "loss": 0.0002, "step": 41544 }, { "epoch": 7.62, "learning_rate": 1.1888515730866579e-07, "loss": 0.0002, "step": 41545 }, { "epoch": 7.62, "learning_rate": 1.1877104405642825e-07, "loss": 0.0004, "step": 41546 }, { "epoch": 7.62, "learning_rate": 1.1865698526975833e-07, "loss": 0.0004, "step": 41547 }, { "epoch": 7.62, "learning_rate": 1.1854298094927996e-07, "loss": 0.0004, "step": 41548 }, { "epoch": 7.62, "learning_rate": 1.1842903109562598e-07, "loss": 0.0004, "step": 41549 }, { "epoch": 7.62, "learning_rate": 1.1831513570942254e-07, "loss": 0.0005, "step": 41550 }, { "epoch": 7.62, "learning_rate": 1.1820129479129694e-07, "loss": 0.0009, "step": 41551 }, { "epoch": 7.62, "learning_rate": 1.1808750834187754e-07, "loss": 0.0004, "step": 41552 }, { "epoch": 7.62, "learning_rate": 1.1797377636179164e-07, "loss": 0.0003, "step": 41553 }, { "epoch": 7.62, "learning_rate": 1.1786009885166538e-07, "loss": 0.0005, "step": 41554 }, { "epoch": 7.62, "learning_rate": 1.1774647581212606e-07, "loss": 0.0005, "step": 41555 }, { "epoch": 7.62, "learning_rate": 1.1763290724379984e-07, "loss": 0.0004, "step": 41556 }, { "epoch": 7.62, "learning_rate": 1.1751939314731064e-07, "loss": 0.0007, "step": 41557 }, { "epoch": 7.62, "learning_rate": 1.174059335232891e-07, "loss": 0.0005, "step": 41558 }, { "epoch": 7.62, "learning_rate": 1.1729252837235583e-07, "loss": 0.0006, "step": 41559 }, { "epoch": 7.62, "learning_rate": 1.1717917769513809e-07, "loss": 0.0004, "step": 41560 }, { "epoch": 7.62, "learning_rate": 1.1706588149225983e-07, "loss": 0.0005, "step": 41561 }, { "epoch": 7.62, "learning_rate": 1.1695263976434612e-07, "loss": 0.0003, "step": 41562 }, { "epoch": 7.62, "learning_rate": 1.1683945251202088e-07, "loss": 0.0003, "step": 41563 }, { "epoch": 7.62, "learning_rate": 1.1672631973590809e-07, "loss": 0.0005, "step": 41564 }, { "epoch": 7.62, "learning_rate": 1.1661324143663278e-07, "loss": 0.0002, "step": 41565 }, { "epoch": 7.62, "learning_rate": 1.1650021761481555e-07, "loss": 0.0006, "step": 41566 }, { "epoch": 7.62, "learning_rate": 1.1638724827108038e-07, "loss": 0.0002, "step": 41567 }, { "epoch": 7.62, "learning_rate": 1.1627433340605121e-07, "loss": 0.0003, "step": 41568 }, { "epoch": 7.62, "learning_rate": 1.1616147302034864e-07, "loss": 0.0003, "step": 41569 }, { "epoch": 7.62, "learning_rate": 1.160486671145955e-07, "loss": 0.0003, "step": 41570 }, { "epoch": 7.62, "learning_rate": 1.1593591568941464e-07, "loss": 0.0005, "step": 41571 }, { "epoch": 7.62, "learning_rate": 1.1582321874542558e-07, "loss": 0.0002, "step": 41572 }, { "epoch": 7.62, "learning_rate": 1.157105762832511e-07, "loss": 0.0005, "step": 41573 }, { "epoch": 7.62, "learning_rate": 1.1559798830351076e-07, "loss": 0.0004, "step": 41574 }, { "epoch": 7.62, "learning_rate": 1.1548545480682628e-07, "loss": 0.0004, "step": 41575 }, { "epoch": 7.62, "learning_rate": 1.1537297579381823e-07, "loss": 0.0003, "step": 41576 }, { "epoch": 7.62, "learning_rate": 1.1526055126510616e-07, "loss": 0.0007, "step": 41577 }, { "epoch": 7.62, "learning_rate": 1.1514818122130844e-07, "loss": 0.0004, "step": 41578 }, { "epoch": 7.62, "learning_rate": 1.1503586566304459e-07, "loss": 0.0003, "step": 41579 }, { "epoch": 7.63, "learning_rate": 1.1492360459093632e-07, "loss": 0.0005, "step": 41580 }, { "epoch": 7.63, "learning_rate": 1.1481139800559981e-07, "loss": 0.0006, "step": 41581 }, { "epoch": 7.63, "learning_rate": 1.1469924590765569e-07, "loss": 0.0004, "step": 41582 }, { "epoch": 7.63, "learning_rate": 1.145871482977201e-07, "loss": 0.0002, "step": 41583 }, { "epoch": 7.63, "learning_rate": 1.1447510517641147e-07, "loss": 0.0003, "step": 41584 }, { "epoch": 7.63, "learning_rate": 1.1436311654434817e-07, "loss": 0.0003, "step": 41585 }, { "epoch": 7.63, "learning_rate": 1.1425118240214638e-07, "loss": 0.0004, "step": 41586 }, { "epoch": 7.63, "learning_rate": 1.1413930275042451e-07, "loss": 0.0002, "step": 41587 }, { "epoch": 7.63, "learning_rate": 1.1402747758979759e-07, "loss": 0.0006, "step": 41588 }, { "epoch": 7.63, "learning_rate": 1.1391570692088294e-07, "loss": 0.0005, "step": 41589 }, { "epoch": 7.63, "learning_rate": 1.1380399074429671e-07, "loss": 0.0004, "step": 41590 }, { "epoch": 7.63, "learning_rate": 1.1369232906065509e-07, "loss": 0.0003, "step": 41591 }, { "epoch": 7.63, "learning_rate": 1.13580721870572e-07, "loss": 0.0007, "step": 41592 }, { "epoch": 7.63, "learning_rate": 1.1346916917466366e-07, "loss": 0.0006, "step": 41593 }, { "epoch": 7.63, "learning_rate": 1.1335767097354622e-07, "loss": 0.0005, "step": 41594 }, { "epoch": 7.63, "learning_rate": 1.132462272678314e-07, "loss": 0.0006, "step": 41595 }, { "epoch": 7.63, "learning_rate": 1.131348380581343e-07, "loss": 0.0006, "step": 41596 }, { "epoch": 7.63, "learning_rate": 1.1302350334507106e-07, "loss": 0.0002, "step": 41597 }, { "epoch": 7.63, "learning_rate": 1.1291222312925343e-07, "loss": 0.0002, "step": 41598 }, { "epoch": 7.63, "learning_rate": 1.1280099741129647e-07, "loss": 0.0002, "step": 41599 }, { "epoch": 7.63, "learning_rate": 1.1268982619181079e-07, "loss": 0.0003, "step": 41600 }, { "epoch": 7.63, "learning_rate": 1.1257870947141036e-07, "loss": 0.0012, "step": 41601 }, { "epoch": 7.63, "learning_rate": 1.1246764725070803e-07, "loss": 0.0006, "step": 41602 }, { "epoch": 7.63, "learning_rate": 1.123566395303155e-07, "loss": 0.0004, "step": 41603 }, { "epoch": 7.63, "learning_rate": 1.1224568631084564e-07, "loss": 0.0004, "step": 41604 }, { "epoch": 7.63, "learning_rate": 1.1213478759290796e-07, "loss": 0.0011, "step": 41605 }, { "epoch": 7.63, "learning_rate": 1.1202394337711531e-07, "loss": 0.0003, "step": 41606 }, { "epoch": 7.63, "learning_rate": 1.1191315366407939e-07, "loss": 0.0003, "step": 41607 }, { "epoch": 7.63, "learning_rate": 1.1180241845440864e-07, "loss": 0.0002, "step": 41608 }, { "epoch": 7.63, "learning_rate": 1.1169173774871478e-07, "loss": 0.0004, "step": 41609 }, { "epoch": 7.63, "learning_rate": 1.1158111154760731e-07, "loss": 0.0003, "step": 41610 }, { "epoch": 7.63, "learning_rate": 1.1147053985169687e-07, "loss": 0.0008, "step": 41611 }, { "epoch": 7.63, "learning_rate": 1.1136002266159408e-07, "loss": 0.0005, "step": 41612 }, { "epoch": 7.63, "learning_rate": 1.1124955997790509e-07, "loss": 0.0007, "step": 41613 }, { "epoch": 7.63, "learning_rate": 1.1113915180123947e-07, "loss": 0.0006, "step": 41614 }, { "epoch": 7.63, "learning_rate": 1.1102879813220668e-07, "loss": 0.0004, "step": 41615 }, { "epoch": 7.63, "learning_rate": 1.1091849897141515e-07, "loss": 0.0011, "step": 41616 }, { "epoch": 7.63, "learning_rate": 1.1080825431947329e-07, "loss": 0.0001, "step": 41617 }, { "epoch": 7.63, "learning_rate": 1.1069806417698725e-07, "loss": 0.0006, "step": 41618 }, { "epoch": 7.63, "learning_rate": 1.1058792854456546e-07, "loss": 0.0003, "step": 41619 }, { "epoch": 7.63, "learning_rate": 1.1047784742281409e-07, "loss": 0.0002, "step": 41620 }, { "epoch": 7.63, "learning_rate": 1.1036782081234154e-07, "loss": 0.0005, "step": 41621 }, { "epoch": 7.63, "learning_rate": 1.102578487137529e-07, "loss": 0.0003, "step": 41622 }, { "epoch": 7.63, "learning_rate": 1.1014793112765431e-07, "loss": 0.0004, "step": 41623 }, { "epoch": 7.63, "learning_rate": 1.10038068054652e-07, "loss": 0.0008, "step": 41624 }, { "epoch": 7.63, "learning_rate": 1.0992825949535212e-07, "loss": 0.0002, "step": 41625 }, { "epoch": 7.63, "learning_rate": 1.0981850545035866e-07, "loss": 0.0005, "step": 41626 }, { "epoch": 7.63, "learning_rate": 1.0970880592027777e-07, "loss": 0.0006, "step": 41627 }, { "epoch": 7.63, "learning_rate": 1.0959916090571343e-07, "loss": 0.0011, "step": 41628 }, { "epoch": 7.63, "learning_rate": 1.0948957040727071e-07, "loss": 0.001, "step": 41629 }, { "epoch": 7.63, "learning_rate": 1.0938003442555356e-07, "loss": 0.0003, "step": 41630 }, { "epoch": 7.63, "learning_rate": 1.0927055296116373e-07, "loss": 0.0003, "step": 41631 }, { "epoch": 7.63, "learning_rate": 1.091611260147074e-07, "loss": 0.0005, "step": 41632 }, { "epoch": 7.63, "learning_rate": 1.0905175358678633e-07, "loss": 0.0003, "step": 41633 }, { "epoch": 7.64, "learning_rate": 1.0894243567800333e-07, "loss": 0.0008, "step": 41634 }, { "epoch": 7.64, "learning_rate": 1.088331722889624e-07, "loss": 0.0006, "step": 41635 }, { "epoch": 7.64, "learning_rate": 1.0872396342026414e-07, "loss": 0.0001, "step": 41636 }, { "epoch": 7.64, "learning_rate": 1.0861480907251143e-07, "loss": 0.0004, "step": 41637 }, { "epoch": 7.64, "learning_rate": 1.0850570924630488e-07, "loss": 0.0003, "step": 41638 }, { "epoch": 7.64, "learning_rate": 1.0839666394224735e-07, "loss": 0.0005, "step": 41639 }, { "epoch": 7.64, "learning_rate": 1.0828767316093946e-07, "loss": 0.0012, "step": 41640 }, { "epoch": 7.64, "learning_rate": 1.0817873690298076e-07, "loss": 0.0004, "step": 41641 }, { "epoch": 7.64, "learning_rate": 1.0806985516897295e-07, "loss": 0.0002, "step": 41642 }, { "epoch": 7.64, "learning_rate": 1.0796102795951669e-07, "loss": 0.0018, "step": 41643 }, { "epoch": 7.64, "learning_rate": 1.0785225527520927e-07, "loss": 0.0002, "step": 41644 }, { "epoch": 7.64, "learning_rate": 1.0774353711665241e-07, "loss": 0.0002, "step": 41645 }, { "epoch": 7.64, "learning_rate": 1.0763487348444568e-07, "loss": 0.0002, "step": 41646 }, { "epoch": 7.64, "learning_rate": 1.0752626437918745e-07, "loss": 0.0003, "step": 41647 }, { "epoch": 7.64, "learning_rate": 1.0741770980147615e-07, "loss": 0.0005, "step": 41648 }, { "epoch": 7.64, "learning_rate": 1.0730920975190906e-07, "loss": 0.0004, "step": 41649 }, { "epoch": 7.64, "learning_rate": 1.0720076423108571e-07, "loss": 0.0002, "step": 41650 }, { "epoch": 7.64, "learning_rate": 1.0709237323960453e-07, "loss": 0.0004, "step": 41651 }, { "epoch": 7.64, "learning_rate": 1.0698403677806057e-07, "loss": 0.0003, "step": 41652 }, { "epoch": 7.64, "learning_rate": 1.0687575484705337e-07, "loss": 0.0003, "step": 41653 }, { "epoch": 7.64, "learning_rate": 1.067675274471791e-07, "loss": 0.0006, "step": 41654 }, { "epoch": 7.64, "learning_rate": 1.0665935457903286e-07, "loss": 0.0002, "step": 41655 }, { "epoch": 7.64, "learning_rate": 1.0655123624321306e-07, "loss": 0.0006, "step": 41656 }, { "epoch": 7.64, "learning_rate": 1.0644317244031366e-07, "loss": 0.0008, "step": 41657 }, { "epoch": 7.64, "learning_rate": 1.0633516317093084e-07, "loss": 0.0005, "step": 41658 }, { "epoch": 7.64, "learning_rate": 1.0622720843566192e-07, "loss": 0.0005, "step": 41659 }, { "epoch": 7.64, "learning_rate": 1.0611930823510085e-07, "loss": 0.0005, "step": 41660 }, { "epoch": 7.64, "learning_rate": 1.0601146256984052e-07, "loss": 0.0002, "step": 41661 }, { "epoch": 7.64, "learning_rate": 1.0590367144047708e-07, "loss": 0.0008, "step": 41662 }, { "epoch": 7.64, "learning_rate": 1.0579593484760343e-07, "loss": 0.0006, "step": 41663 }, { "epoch": 7.64, "learning_rate": 1.0568825279181572e-07, "loss": 0.0007, "step": 41664 }, { "epoch": 7.64, "learning_rate": 1.0558062527370683e-07, "loss": 0.0004, "step": 41665 }, { "epoch": 7.64, "learning_rate": 1.0547305229386851e-07, "loss": 0.0007, "step": 41666 }, { "epoch": 7.64, "learning_rate": 1.0536553385289361e-07, "loss": 0.0004, "step": 41667 }, { "epoch": 7.64, "learning_rate": 1.052580699513761e-07, "loss": 0.0004, "step": 41668 }, { "epoch": 7.64, "learning_rate": 1.0515066058990885e-07, "loss": 0.0007, "step": 41669 }, { "epoch": 7.64, "learning_rate": 1.0504330576908251e-07, "loss": 0.0005, "step": 41670 }, { "epoch": 7.64, "learning_rate": 1.0493600548948879e-07, "loss": 0.0005, "step": 41671 }, { "epoch": 7.64, "learning_rate": 1.0482875975172058e-07, "loss": 0.0002, "step": 41672 }, { "epoch": 7.64, "learning_rate": 1.047215685563674e-07, "loss": 0.0008, "step": 41673 }, { "epoch": 7.64, "learning_rate": 1.0461443190402098e-07, "loss": 0.0004, "step": 41674 }, { "epoch": 7.64, "learning_rate": 1.0450734979527089e-07, "loss": 0.0002, "step": 41675 }, { "epoch": 7.64, "learning_rate": 1.0440032223070884e-07, "loss": 0.0006, "step": 41676 }, { "epoch": 7.64, "learning_rate": 1.0429334921092437e-07, "loss": 0.0003, "step": 41677 }, { "epoch": 7.64, "learning_rate": 1.0418643073650702e-07, "loss": 0.0002, "step": 41678 }, { "epoch": 7.64, "learning_rate": 1.040795668080452e-07, "loss": 0.0004, "step": 41679 }, { "epoch": 7.64, "learning_rate": 1.0397275742612734e-07, "loss": 0.0003, "step": 41680 }, { "epoch": 7.64, "learning_rate": 1.0386600259134517e-07, "loss": 0.0004, "step": 41681 }, { "epoch": 7.64, "learning_rate": 1.0375930230428488e-07, "loss": 0.0004, "step": 41682 }, { "epoch": 7.64, "learning_rate": 1.0365265656553603e-07, "loss": 0.0002, "step": 41683 }, { "epoch": 7.64, "learning_rate": 1.0354606537568479e-07, "loss": 0.0002, "step": 41684 }, { "epoch": 7.64, "learning_rate": 1.034395287353196e-07, "loss": 0.0002, "step": 41685 }, { "epoch": 7.64, "learning_rate": 1.0333304664502775e-07, "loss": 0.0003, "step": 41686 }, { "epoch": 7.64, "learning_rate": 1.0322661910539545e-07, "loss": 0.0003, "step": 41687 }, { "epoch": 7.64, "learning_rate": 1.0312024611701111e-07, "loss": 0.0005, "step": 41688 }, { "epoch": 7.65, "learning_rate": 1.030139276804587e-07, "loss": 0.0002, "step": 41689 }, { "epoch": 7.65, "learning_rate": 1.0290766379632555e-07, "loss": 0.0005, "step": 41690 }, { "epoch": 7.65, "learning_rate": 1.0280145446519784e-07, "loss": 0.0009, "step": 41691 }, { "epoch": 7.65, "learning_rate": 1.0269529968765956e-07, "loss": 0.0002, "step": 41692 }, { "epoch": 7.65, "learning_rate": 1.025891994642969e-07, "loss": 0.0008, "step": 41693 }, { "epoch": 7.65, "learning_rate": 1.0248315379569496e-07, "loss": 0.0006, "step": 41694 }, { "epoch": 7.65, "learning_rate": 1.0237716268243659e-07, "loss": 0.0003, "step": 41695 }, { "epoch": 7.65, "learning_rate": 1.0227122612510799e-07, "loss": 0.0003, "step": 41696 }, { "epoch": 7.65, "learning_rate": 1.0216534412429313e-07, "loss": 0.0002, "step": 41697 }, { "epoch": 7.65, "learning_rate": 1.0205951668057268e-07, "loss": 0.0005, "step": 41698 }, { "epoch": 7.65, "learning_rate": 1.0195374379453393e-07, "loss": 0.0004, "step": 41699 }, { "epoch": 7.65, "learning_rate": 1.0184802546675643e-07, "loss": 0.0007, "step": 41700 }, { "epoch": 7.65, "learning_rate": 1.0174236169782636e-07, "loss": 0.0004, "step": 41701 }, { "epoch": 7.65, "learning_rate": 1.0163675248832328e-07, "loss": 0.0006, "step": 41702 }, { "epoch": 7.65, "learning_rate": 1.0153119783883003e-07, "loss": 0.0002, "step": 41703 }, { "epoch": 7.65, "learning_rate": 1.014256977499295e-07, "loss": 0.0005, "step": 41704 }, { "epoch": 7.65, "learning_rate": 1.0132025222220121e-07, "loss": 0.0005, "step": 41705 }, { "epoch": 7.65, "learning_rate": 1.0121486125622914e-07, "loss": 0.0006, "step": 41706 }, { "epoch": 7.65, "learning_rate": 1.0110952485259174e-07, "loss": 0.0002, "step": 41707 }, { "epoch": 7.65, "learning_rate": 1.0100424301187073e-07, "loss": 0.0003, "step": 41708 }, { "epoch": 7.65, "learning_rate": 1.0089901573464566e-07, "loss": 0.0004, "step": 41709 }, { "epoch": 7.65, "learning_rate": 1.007938430214983e-07, "loss": 0.0004, "step": 41710 }, { "epoch": 7.65, "learning_rate": 1.0068872487300596e-07, "loss": 0.0008, "step": 41711 }, { "epoch": 7.65, "learning_rate": 1.005836612897504e-07, "loss": 0.0006, "step": 41712 }, { "epoch": 7.65, "learning_rate": 1.0047865227230891e-07, "loss": 0.0003, "step": 41713 }, { "epoch": 7.65, "learning_rate": 1.0037369782126105e-07, "loss": 0.0002, "step": 41714 }, { "epoch": 7.65, "learning_rate": 1.0026879793718525e-07, "loss": 0.0002, "step": 41715 }, { "epoch": 7.65, "learning_rate": 1.0016395262065992e-07, "loss": 0.0003, "step": 41716 }, { "epoch": 7.65, "learning_rate": 1.0005916187226239e-07, "loss": 0.0004, "step": 41717 }, { "epoch": 7.65, "learning_rate": 9.995442569257108e-08, "loss": 0.0003, "step": 41718 }, { "epoch": 7.65, "learning_rate": 9.984974408216441e-08, "loss": 0.0002, "step": 41719 }, { "epoch": 7.65, "learning_rate": 9.974511704161527e-08, "loss": 0.0007, "step": 41720 }, { "epoch": 7.65, "learning_rate": 9.96405445715043e-08, "loss": 0.0002, "step": 41721 }, { "epoch": 7.65, "learning_rate": 9.953602667240658e-08, "loss": 0.0008, "step": 41722 }, { "epoch": 7.65, "learning_rate": 9.943156334489835e-08, "loss": 0.0003, "step": 41723 }, { "epoch": 7.65, "learning_rate": 9.932715458955467e-08, "loss": 0.0006, "step": 41724 }, { "epoch": 7.65, "learning_rate": 9.922280040695176e-08, "loss": 0.0001, "step": 41725 }, { "epoch": 7.65, "learning_rate": 9.911850079766583e-08, "loss": 0.0002, "step": 41726 }, { "epoch": 7.65, "learning_rate": 9.901425576226975e-08, "loss": 0.0003, "step": 41727 }, { "epoch": 7.65, "learning_rate": 9.891006530133862e-08, "loss": 0.0004, "step": 41728 }, { "epoch": 7.65, "learning_rate": 9.880592941544753e-08, "loss": 0.0006, "step": 41729 }, { "epoch": 7.65, "learning_rate": 9.870184810517047e-08, "loss": 0.0005, "step": 41730 }, { "epoch": 7.65, "learning_rate": 9.859782137108142e-08, "loss": 0.0004, "step": 41731 }, { "epoch": 7.65, "learning_rate": 9.849384921375216e-08, "loss": 0.0005, "step": 41732 }, { "epoch": 7.65, "learning_rate": 9.838993163375554e-08, "loss": 0.0009, "step": 41733 }, { "epoch": 7.65, "learning_rate": 9.828606863166779e-08, "loss": 0.0006, "step": 41734 }, { "epoch": 7.65, "learning_rate": 9.818226020805844e-08, "loss": 0.0002, "step": 41735 }, { "epoch": 7.65, "learning_rate": 9.807850636350036e-08, "loss": 0.0008, "step": 41736 }, { "epoch": 7.65, "learning_rate": 9.797480709856644e-08, "loss": 0.0003, "step": 41737 }, { "epoch": 7.65, "learning_rate": 9.78711624138251e-08, "loss": 0.0004, "step": 41738 }, { "epoch": 7.65, "learning_rate": 9.776757230985146e-08, "loss": 0.0005, "step": 41739 }, { "epoch": 7.65, "learning_rate": 9.766403678721504e-08, "loss": 0.0006, "step": 41740 }, { "epoch": 7.65, "learning_rate": 9.756055584648649e-08, "loss": 0.0002, "step": 41741 }, { "epoch": 7.65, "learning_rate": 9.745712948823538e-08, "loss": 0.0003, "step": 41742 }, { "epoch": 7.66, "learning_rate": 9.735375771303346e-08, "loss": 0.0009, "step": 41743 }, { "epoch": 7.66, "learning_rate": 9.725044052144916e-08, "loss": 0.0005, "step": 41744 }, { "epoch": 7.66, "learning_rate": 9.714717791405204e-08, "loss": 0.0004, "step": 41745 }, { "epoch": 7.66, "learning_rate": 9.704396989141162e-08, "loss": 0.0008, "step": 41746 }, { "epoch": 7.66, "learning_rate": 9.694081645409748e-08, "loss": 0.0003, "step": 41747 }, { "epoch": 7.66, "learning_rate": 9.683771760267801e-08, "loss": 0.0004, "step": 41748 }, { "epoch": 7.66, "learning_rate": 9.673467333772058e-08, "loss": 0.0003, "step": 41749 }, { "epoch": 7.66, "learning_rate": 9.663168365979359e-08, "loss": 0.0005, "step": 41750 }, { "epoch": 7.66, "learning_rate": 9.652874856946326e-08, "loss": 0.0003, "step": 41751 }, { "epoch": 7.66, "learning_rate": 9.642586806730025e-08, "loss": 0.0003, "step": 41752 }, { "epoch": 7.66, "learning_rate": 9.632304215386968e-08, "loss": 0.0007, "step": 41753 }, { "epoch": 7.66, "learning_rate": 9.622027082973883e-08, "loss": 0.0005, "step": 41754 }, { "epoch": 7.66, "learning_rate": 9.611755409547285e-08, "loss": 0.0007, "step": 41755 }, { "epoch": 7.66, "learning_rate": 9.601489195163905e-08, "loss": 0.0003, "step": 41756 }, { "epoch": 7.66, "learning_rate": 9.591228439880363e-08, "loss": 0.001, "step": 41757 }, { "epoch": 7.66, "learning_rate": 9.580973143753169e-08, "loss": 0.0004, "step": 41758 }, { "epoch": 7.66, "learning_rate": 9.570723306838836e-08, "loss": 0.0004, "step": 41759 }, { "epoch": 7.66, "learning_rate": 9.560478929193984e-08, "loss": 0.0002, "step": 41760 }, { "epoch": 7.66, "learning_rate": 9.550240010874901e-08, "loss": 0.0005, "step": 41761 }, { "epoch": 7.66, "learning_rate": 9.540006551938097e-08, "loss": 0.0003, "step": 41762 }, { "epoch": 7.66, "learning_rate": 9.529778552439972e-08, "loss": 0.0006, "step": 41763 }, { "epoch": 7.66, "learning_rate": 9.519556012436815e-08, "loss": 0.0006, "step": 41764 }, { "epoch": 7.66, "learning_rate": 9.509338931985135e-08, "loss": 0.0004, "step": 41765 }, { "epoch": 7.66, "learning_rate": 9.499127311141332e-08, "loss": 0.0004, "step": 41766 }, { "epoch": 7.66, "learning_rate": 9.48892114996136e-08, "loss": 0.0006, "step": 41767 }, { "epoch": 7.66, "learning_rate": 9.478720448501732e-08, "loss": 0.0007, "step": 41768 }, { "epoch": 7.66, "learning_rate": 9.468525206818624e-08, "loss": 0.0004, "step": 41769 }, { "epoch": 7.66, "learning_rate": 9.458335424968101e-08, "loss": 0.0008, "step": 41770 }, { "epoch": 7.66, "learning_rate": 9.448151103006564e-08, "loss": 0.0004, "step": 41771 }, { "epoch": 7.66, "learning_rate": 9.437972240990079e-08, "loss": 0.0005, "step": 41772 }, { "epoch": 7.66, "learning_rate": 9.427798838974489e-08, "loss": 0.0002, "step": 41773 }, { "epoch": 7.66, "learning_rate": 9.417630897016195e-08, "loss": 0.0005, "step": 41774 }, { "epoch": 7.66, "learning_rate": 9.407468415171151e-08, "loss": 0.0003, "step": 41775 }, { "epoch": 7.66, "learning_rate": 9.397311393495312e-08, "loss": 0.0002, "step": 41776 }, { "epoch": 7.66, "learning_rate": 9.387159832044746e-08, "loss": 0.001, "step": 41777 }, { "epoch": 7.66, "learning_rate": 9.377013730875295e-08, "loss": 0.0002, "step": 41778 }, { "epoch": 7.66, "learning_rate": 9.366873090042917e-08, "loss": 0.0004, "step": 41779 }, { "epoch": 7.66, "learning_rate": 9.356737909603675e-08, "loss": 0.0005, "step": 41780 }, { "epoch": 7.66, "learning_rate": 9.346608189613193e-08, "loss": 0.0004, "step": 41781 }, { "epoch": 7.66, "learning_rate": 9.336483930127427e-08, "loss": 0.0007, "step": 41782 }, { "epoch": 7.66, "learning_rate": 9.32636513120222e-08, "loss": 0.0004, "step": 41783 }, { "epoch": 7.66, "learning_rate": 9.316251792893305e-08, "loss": 0.0003, "step": 41784 }, { "epoch": 7.66, "learning_rate": 9.306143915256415e-08, "loss": 0.0002, "step": 41785 }, { "epoch": 7.66, "learning_rate": 9.296041498347174e-08, "loss": 0.0011, "step": 41786 }, { "epoch": 7.66, "learning_rate": 9.285944542221425e-08, "loss": 0.0006, "step": 41787 }, { "epoch": 7.66, "learning_rate": 9.275853046934791e-08, "loss": 0.0002, "step": 41788 }, { "epoch": 7.66, "learning_rate": 9.265767012542781e-08, "loss": 0.0003, "step": 41789 }, { "epoch": 7.66, "learning_rate": 9.255686439101241e-08, "loss": 0.001, "step": 41790 }, { "epoch": 7.66, "learning_rate": 9.245611326665349e-08, "loss": 0.0005, "step": 41791 }, { "epoch": 7.66, "learning_rate": 9.23554167529106e-08, "loss": 0.0003, "step": 41792 }, { "epoch": 7.66, "learning_rate": 9.225477485033552e-08, "loss": 0.0007, "step": 41793 }, { "epoch": 7.66, "learning_rate": 9.215418755948446e-08, "loss": 0.0002, "step": 41794 }, { "epoch": 7.66, "learning_rate": 9.205365488091145e-08, "loss": 0.001, "step": 41795 }, { "epoch": 7.66, "learning_rate": 9.195317681517157e-08, "loss": 0.0004, "step": 41796 }, { "epoch": 7.66, "learning_rate": 9.185275336281773e-08, "loss": 0.0005, "step": 41797 }, { "epoch": 7.67, "learning_rate": 9.175238452440283e-08, "loss": 0.0002, "step": 41798 }, { "epoch": 7.67, "learning_rate": 9.165207030048084e-08, "loss": 0.0005, "step": 41799 }, { "epoch": 7.67, "learning_rate": 9.155181069160579e-08, "loss": 0.0002, "step": 41800 }, { "epoch": 7.67, "learning_rate": 9.145160569832944e-08, "loss": 0.0003, "step": 41801 }, { "epoch": 7.67, "learning_rate": 9.13514553212036e-08, "loss": 0.0003, "step": 41802 }, { "epoch": 7.67, "learning_rate": 9.125135956078002e-08, "loss": 0.0002, "step": 41803 }, { "epoch": 7.67, "learning_rate": 9.11513184176116e-08, "loss": 0.0002, "step": 41804 }, { "epoch": 7.67, "learning_rate": 9.105133189224902e-08, "loss": 0.0006, "step": 41805 }, { "epoch": 7.67, "learning_rate": 9.095139998524405e-08, "loss": 0.0006, "step": 41806 }, { "epoch": 7.67, "learning_rate": 9.085152269714626e-08, "loss": 0.0004, "step": 41807 }, { "epoch": 7.67, "learning_rate": 9.075170002850741e-08, "loss": 0.0005, "step": 41808 }, { "epoch": 7.67, "learning_rate": 9.06519319798782e-08, "loss": 0.0004, "step": 41809 }, { "epoch": 7.67, "learning_rate": 9.055221855180706e-08, "loss": 0.0003, "step": 41810 }, { "epoch": 7.67, "learning_rate": 9.045255974484357e-08, "loss": 0.0006, "step": 41811 }, { "epoch": 7.67, "learning_rate": 9.035295555953838e-08, "loss": 0.0009, "step": 41812 }, { "epoch": 7.67, "learning_rate": 9.025340599643994e-08, "loss": 0.0006, "step": 41813 }, { "epoch": 7.67, "learning_rate": 9.015391105609784e-08, "loss": 0.0008, "step": 41814 }, { "epoch": 7.67, "learning_rate": 9.005447073905826e-08, "loss": 0.0004, "step": 41815 }, { "epoch": 7.67, "learning_rate": 8.995508504587081e-08, "loss": 0.0004, "step": 41816 }, { "epoch": 7.67, "learning_rate": 8.98557539770828e-08, "loss": 0.0003, "step": 41817 }, { "epoch": 7.67, "learning_rate": 8.975647753324157e-08, "loss": 0.0008, "step": 41818 }, { "epoch": 7.67, "learning_rate": 8.96572557148967e-08, "loss": 0.0004, "step": 41819 }, { "epoch": 7.67, "learning_rate": 8.955808852259219e-08, "loss": 0.0003, "step": 41820 }, { "epoch": 7.67, "learning_rate": 8.945897595687536e-08, "loss": 0.0002, "step": 41821 }, { "epoch": 7.67, "learning_rate": 8.935991801829247e-08, "loss": 0.0008, "step": 41822 }, { "epoch": 7.67, "learning_rate": 8.926091470739084e-08, "loss": 0.001, "step": 41823 }, { "epoch": 7.67, "learning_rate": 8.91619660247145e-08, "loss": 0.0005, "step": 41824 }, { "epoch": 7.67, "learning_rate": 8.906307197081076e-08, "loss": 0.0005, "step": 41825 }, { "epoch": 7.67, "learning_rate": 8.896423254622255e-08, "loss": 0.0002, "step": 41826 }, { "epoch": 7.67, "learning_rate": 8.886544775149608e-08, "loss": 0.001, "step": 41827 }, { "epoch": 7.67, "learning_rate": 8.876671758717537e-08, "loss": 0.0008, "step": 41828 }, { "epoch": 7.67, "learning_rate": 8.866804205380441e-08, "loss": 0.0005, "step": 41829 }, { "epoch": 7.67, "learning_rate": 8.856942115192724e-08, "loss": 0.0007, "step": 41830 }, { "epoch": 7.67, "learning_rate": 8.847085488208895e-08, "loss": 0.0003, "step": 41831 }, { "epoch": 7.67, "learning_rate": 8.837234324483023e-08, "loss": 0.0003, "step": 41832 }, { "epoch": 7.67, "learning_rate": 8.827388624069621e-08, "loss": 0.0011, "step": 41833 }, { "epoch": 7.67, "learning_rate": 8.817548387022757e-08, "loss": 0.0005, "step": 41834 }, { "epoch": 7.67, "learning_rate": 8.807713613396828e-08, "loss": 0.0002, "step": 41835 }, { "epoch": 7.67, "learning_rate": 8.797884303246018e-08, "loss": 0.0003, "step": 41836 }, { "epoch": 7.67, "learning_rate": 8.788060456624614e-08, "loss": 0.0006, "step": 41837 }, { "epoch": 7.67, "learning_rate": 8.778242073586462e-08, "loss": 0.0004, "step": 41838 }, { "epoch": 7.67, "learning_rate": 8.768429154185853e-08, "loss": 0.0002, "step": 41839 }, { "epoch": 7.67, "learning_rate": 8.758621698476965e-08, "loss": 0.0006, "step": 41840 }, { "epoch": 7.67, "learning_rate": 8.748819706513756e-08, "loss": 0.0002, "step": 41841 }, { "epoch": 7.67, "learning_rate": 8.739023178350403e-08, "loss": 0.0006, "step": 41842 }, { "epoch": 7.67, "learning_rate": 8.729232114040642e-08, "loss": 0.0004, "step": 41843 }, { "epoch": 7.67, "learning_rate": 8.71944651363854e-08, "loss": 0.0002, "step": 41844 }, { "epoch": 7.67, "learning_rate": 8.709666377198167e-08, "loss": 0.0004, "step": 41845 }, { "epoch": 7.67, "learning_rate": 8.699891704773256e-08, "loss": 0.0004, "step": 41846 }, { "epoch": 7.67, "learning_rate": 8.690122496417874e-08, "loss": 0.0003, "step": 41847 }, { "epoch": 7.67, "learning_rate": 8.680358752185647e-08, "loss": 0.0002, "step": 41848 }, { "epoch": 7.67, "learning_rate": 8.670600472130641e-08, "loss": 0.0003, "step": 41849 }, { "epoch": 7.67, "learning_rate": 8.66084765630637e-08, "loss": 0.0013, "step": 41850 }, { "epoch": 7.67, "learning_rate": 8.65110030476679e-08, "loss": 0.0011, "step": 41851 }, { "epoch": 7.68, "learning_rate": 8.641358417565527e-08, "loss": 0.0002, "step": 41852 }, { "epoch": 7.68, "learning_rate": 8.631621994756423e-08, "loss": 0.0002, "step": 41853 }, { "epoch": 7.68, "learning_rate": 8.621891036392882e-08, "loss": 0.0006, "step": 41854 }, { "epoch": 7.68, "learning_rate": 8.612165542528861e-08, "loss": 0.0002, "step": 41855 }, { "epoch": 7.68, "learning_rate": 8.602445513217761e-08, "loss": 0.0002, "step": 41856 }, { "epoch": 7.68, "learning_rate": 8.592730948513205e-08, "loss": 0.0005, "step": 41857 }, { "epoch": 7.68, "learning_rate": 8.58302184846871e-08, "loss": 0.0001, "step": 41858 }, { "epoch": 7.68, "learning_rate": 8.573318213137893e-08, "loss": 0.0006, "step": 41859 }, { "epoch": 7.68, "learning_rate": 8.563620042574161e-08, "loss": 0.0008, "step": 41860 }, { "epoch": 7.68, "learning_rate": 8.553927336830914e-08, "loss": 0.0005, "step": 41861 }, { "epoch": 7.68, "learning_rate": 8.544240095961665e-08, "loss": 0.0002, "step": 41862 }, { "epoch": 7.68, "learning_rate": 8.534558320019814e-08, "loss": 0.0009, "step": 41863 }, { "epoch": 7.68, "learning_rate": 8.524882009058765e-08, "loss": 0.0007, "step": 41864 }, { "epoch": 7.68, "learning_rate": 8.515211163131697e-08, "loss": 0.0005, "step": 41865 }, { "epoch": 7.68, "learning_rate": 8.505545782292012e-08, "loss": 0.0006, "step": 41866 }, { "epoch": 7.68, "learning_rate": 8.495885866593001e-08, "loss": 0.0002, "step": 41867 }, { "epoch": 7.68, "learning_rate": 8.486231416087953e-08, "loss": 0.0005, "step": 41868 }, { "epoch": 7.68, "learning_rate": 8.476582430830049e-08, "loss": 0.0007, "step": 41869 }, { "epoch": 7.68, "learning_rate": 8.466938910872247e-08, "loss": 0.0003, "step": 41870 }, { "epoch": 7.68, "learning_rate": 8.457300856268059e-08, "loss": 0.0002, "step": 41871 }, { "epoch": 7.68, "learning_rate": 8.447668267070441e-08, "loss": 0.0006, "step": 41872 }, { "epoch": 7.68, "learning_rate": 8.438041143332688e-08, "loss": 0.001, "step": 41873 }, { "epoch": 7.68, "learning_rate": 8.42841948510742e-08, "loss": 0.0005, "step": 41874 }, { "epoch": 7.68, "learning_rate": 8.418803292448041e-08, "loss": 0.0011, "step": 41875 }, { "epoch": 7.68, "learning_rate": 8.409192565407399e-08, "loss": 0.0003, "step": 41876 }, { "epoch": 7.68, "learning_rate": 8.399587304038559e-08, "loss": 0.0005, "step": 41877 }, { "epoch": 7.68, "learning_rate": 8.38998750839437e-08, "loss": 0.0004, "step": 41878 }, { "epoch": 7.68, "learning_rate": 8.380393178527902e-08, "loss": 0.0003, "step": 41879 }, { "epoch": 7.68, "learning_rate": 8.370804314491887e-08, "loss": 0.0004, "step": 41880 }, { "epoch": 7.68, "learning_rate": 8.361220916339174e-08, "loss": 0.0005, "step": 41881 }, { "epoch": 7.68, "learning_rate": 8.351642984122721e-08, "loss": 0.0002, "step": 41882 }, { "epoch": 7.68, "learning_rate": 8.342070517895152e-08, "loss": 0.0004, "step": 41883 }, { "epoch": 7.68, "learning_rate": 8.332503517709423e-08, "loss": 0.0002, "step": 41884 }, { "epoch": 7.68, "learning_rate": 8.32294198361816e-08, "loss": 0.001, "step": 41885 }, { "epoch": 7.68, "learning_rate": 8.3133859156741e-08, "loss": 0.0005, "step": 41886 }, { "epoch": 7.68, "learning_rate": 8.303835313929865e-08, "loss": 0.0004, "step": 41887 }, { "epoch": 7.68, "learning_rate": 8.29429017843797e-08, "loss": 0.0006, "step": 41888 }, { "epoch": 7.68, "learning_rate": 8.284750509251371e-08, "loss": 0.0004, "step": 41889 }, { "epoch": 7.68, "learning_rate": 8.275216306422362e-08, "loss": 0.0003, "step": 41890 }, { "epoch": 7.68, "learning_rate": 8.265687570003678e-08, "loss": 0.0011, "step": 41891 }, { "epoch": 7.68, "learning_rate": 8.256164300047608e-08, "loss": 0.0006, "step": 41892 }, { "epoch": 7.68, "learning_rate": 8.246646496606892e-08, "loss": 0.0003, "step": 41893 }, { "epoch": 7.68, "learning_rate": 8.237134159733928e-08, "loss": 0.0003, "step": 41894 }, { "epoch": 7.68, "learning_rate": 8.227627289481121e-08, "loss": 0.0009, "step": 41895 }, { "epoch": 7.68, "learning_rate": 8.218125885900763e-08, "loss": 0.0002, "step": 41896 }, { "epoch": 7.68, "learning_rate": 8.208629949045477e-08, "loss": 0.0002, "step": 41897 }, { "epoch": 7.68, "learning_rate": 8.199139478967333e-08, "loss": 0.0002, "step": 41898 }, { "epoch": 7.68, "learning_rate": 8.189654475718845e-08, "loss": 0.0005, "step": 41899 }, { "epoch": 7.68, "learning_rate": 8.180174939352192e-08, "loss": 0.0003, "step": 41900 }, { "epoch": 7.68, "learning_rate": 8.170700869919557e-08, "loss": 0.0002, "step": 41901 }, { "epoch": 7.68, "learning_rate": 8.16123226747334e-08, "loss": 0.0011, "step": 41902 }, { "epoch": 7.68, "learning_rate": 8.151769132065723e-08, "loss": 0.0005, "step": 41903 }, { "epoch": 7.68, "learning_rate": 8.142311463748665e-08, "loss": 0.0007, "step": 41904 }, { "epoch": 7.68, "learning_rate": 8.132859262574455e-08, "loss": 0.0003, "step": 41905 }, { "epoch": 7.68, "learning_rate": 8.123412528595165e-08, "loss": 0.0003, "step": 41906 }, { "epoch": 7.69, "learning_rate": 8.113971261862863e-08, "loss": 0.0003, "step": 41907 }, { "epoch": 7.69, "learning_rate": 8.104535462429619e-08, "loss": 0.0002, "step": 41908 }, { "epoch": 7.69, "learning_rate": 8.095105130347503e-08, "loss": 0.0006, "step": 41909 }, { "epoch": 7.69, "learning_rate": 8.085680265668249e-08, "loss": 0.0006, "step": 41910 }, { "epoch": 7.69, "learning_rate": 8.076260868443931e-08, "loss": 0.0007, "step": 41911 }, { "epoch": 7.69, "learning_rate": 8.066846938726614e-08, "loss": 0.0006, "step": 41912 }, { "epoch": 7.69, "learning_rate": 8.057438476568036e-08, "loss": 0.0002, "step": 41913 }, { "epoch": 7.69, "learning_rate": 8.048035482020155e-08, "loss": 0.0003, "step": 41914 }, { "epoch": 7.69, "learning_rate": 8.038637955134709e-08, "loss": 0.0004, "step": 41915 }, { "epoch": 7.69, "learning_rate": 8.029245895963434e-08, "loss": 0.0007, "step": 41916 }, { "epoch": 7.69, "learning_rate": 8.019859304558286e-08, "loss": 0.0001, "step": 41917 }, { "epoch": 7.69, "learning_rate": 8.010478180970893e-08, "loss": 0.0004, "step": 41918 }, { "epoch": 7.69, "learning_rate": 8.00110252525299e-08, "loss": 0.0002, "step": 41919 }, { "epoch": 7.69, "learning_rate": 7.991732337456315e-08, "loss": 0.0007, "step": 41920 }, { "epoch": 7.69, "learning_rate": 7.982367617632381e-08, "loss": 0.0006, "step": 41921 }, { "epoch": 7.69, "learning_rate": 7.973008365832812e-08, "loss": 0.001, "step": 41922 }, { "epoch": 7.69, "learning_rate": 7.963654582109348e-08, "loss": 0.0002, "step": 41923 }, { "epoch": 7.69, "learning_rate": 7.954306266513389e-08, "loss": 0.0006, "step": 41924 }, { "epoch": 7.69, "learning_rate": 7.944963419096674e-08, "loss": 0.0003, "step": 41925 }, { "epoch": 7.69, "learning_rate": 7.935626039910493e-08, "loss": 0.0009, "step": 41926 }, { "epoch": 7.69, "learning_rate": 7.926294129006362e-08, "loss": 0.0002, "step": 41927 }, { "epoch": 7.69, "learning_rate": 7.916967686435684e-08, "loss": 0.0002, "step": 41928 }, { "epoch": 7.69, "learning_rate": 7.907646712249972e-08, "loss": 0.001, "step": 41929 }, { "epoch": 7.69, "learning_rate": 7.898331206500632e-08, "loss": 0.0004, "step": 41930 }, { "epoch": 7.69, "learning_rate": 7.889021169238842e-08, "loss": 0.0001, "step": 41931 }, { "epoch": 7.69, "learning_rate": 7.879716600516008e-08, "loss": 0.0004, "step": 41932 }, { "epoch": 7.69, "learning_rate": 7.870417500383531e-08, "loss": 0.0005, "step": 41933 }, { "epoch": 7.69, "learning_rate": 7.861123868892484e-08, "loss": 0.0004, "step": 41934 }, { "epoch": 7.69, "learning_rate": 7.851835706094157e-08, "loss": 0.0003, "step": 41935 }, { "epoch": 7.69, "learning_rate": 7.842553012039622e-08, "loss": 0.0006, "step": 41936 }, { "epoch": 7.69, "learning_rate": 7.833275786780392e-08, "loss": 0.0003, "step": 41937 }, { "epoch": 7.69, "learning_rate": 7.824004030367316e-08, "loss": 0.0003, "step": 41938 }, { "epoch": 7.69, "learning_rate": 7.814737742851575e-08, "loss": 0.0003, "step": 41939 }, { "epoch": 7.69, "learning_rate": 7.805476924284239e-08, "loss": 0.0005, "step": 41940 }, { "epoch": 7.69, "learning_rate": 7.796221574716267e-08, "loss": 0.0003, "step": 41941 }, { "epoch": 7.69, "learning_rate": 7.786971694198842e-08, "loss": 0.0016, "step": 41942 }, { "epoch": 7.69, "learning_rate": 7.77772728278292e-08, "loss": 0.0004, "step": 41943 }, { "epoch": 7.69, "learning_rate": 7.768488340519464e-08, "loss": 0.0002, "step": 41944 }, { "epoch": 7.69, "learning_rate": 7.75925486745921e-08, "loss": 0.0009, "step": 41945 }, { "epoch": 7.69, "learning_rate": 7.750026863653226e-08, "loss": 0.0002, "step": 41946 }, { "epoch": 7.69, "learning_rate": 7.740804329152363e-08, "loss": 0.0004, "step": 41947 }, { "epoch": 7.69, "learning_rate": 7.731587264007356e-08, "loss": 0.0005, "step": 41948 }, { "epoch": 7.69, "learning_rate": 7.722375668269166e-08, "loss": 0.0003, "step": 41949 }, { "epoch": 7.69, "learning_rate": 7.71316954198853e-08, "loss": 0.0005, "step": 41950 }, { "epoch": 7.69, "learning_rate": 7.703968885216073e-08, "loss": 0.0003, "step": 41951 }, { "epoch": 7.69, "learning_rate": 7.694773698002533e-08, "loss": 0.0003, "step": 41952 }, { "epoch": 7.69, "learning_rate": 7.685583980398759e-08, "loss": 0.0005, "step": 41953 }, { "epoch": 7.69, "learning_rate": 7.676399732455264e-08, "loss": 0.0005, "step": 41954 }, { "epoch": 7.69, "learning_rate": 7.667220954222676e-08, "loss": 0.0007, "step": 41955 }, { "epoch": 7.69, "learning_rate": 7.658047645751621e-08, "loss": 0.0006, "step": 41956 }, { "epoch": 7.69, "learning_rate": 7.648879807092724e-08, "loss": 0.0008, "step": 41957 }, { "epoch": 7.69, "learning_rate": 7.639717438296279e-08, "loss": 0.0004, "step": 41958 }, { "epoch": 7.69, "learning_rate": 7.630560539413134e-08, "loss": 0.0005, "step": 41959 }, { "epoch": 7.69, "learning_rate": 7.621409110493471e-08, "loss": 0.0004, "step": 41960 }, { "epoch": 7.7, "learning_rate": 7.612263151587918e-08, "loss": 0.0006, "step": 41961 }, { "epoch": 7.7, "learning_rate": 7.603122662746876e-08, "loss": 0.0003, "step": 41962 }, { "epoch": 7.7, "learning_rate": 7.593987644020529e-08, "loss": 0.0007, "step": 41963 }, { "epoch": 7.7, "learning_rate": 7.584858095459502e-08, "loss": 0.0006, "step": 41964 }, { "epoch": 7.7, "learning_rate": 7.575734017113868e-08, "loss": 0.0004, "step": 41965 }, { "epoch": 7.7, "learning_rate": 7.56661540903414e-08, "loss": 0.0003, "step": 41966 }, { "epoch": 7.7, "learning_rate": 7.557502271270501e-08, "loss": 0.0003, "step": 41967 }, { "epoch": 7.7, "learning_rate": 7.548394603873022e-08, "loss": 0.0004, "step": 41968 }, { "epoch": 7.7, "learning_rate": 7.539292406892107e-08, "loss": 0.0002, "step": 41969 }, { "epoch": 7.7, "learning_rate": 7.530195680377939e-08, "loss": 0.0004, "step": 41970 }, { "epoch": 7.7, "learning_rate": 7.521104424380476e-08, "loss": 0.0008, "step": 41971 }, { "epoch": 7.7, "learning_rate": 7.512018638949902e-08, "loss": 0.001, "step": 41972 }, { "epoch": 7.7, "learning_rate": 7.502938324136511e-08, "loss": 0.0003, "step": 41973 }, { "epoch": 7.7, "learning_rate": 7.493863479990038e-08, "loss": 0.0005, "step": 41974 }, { "epoch": 7.7, "learning_rate": 7.484794106560667e-08, "loss": 0.0002, "step": 41975 }, { "epoch": 7.7, "learning_rate": 7.475730203898358e-08, "loss": 0.0001, "step": 41976 }, { "epoch": 7.7, "learning_rate": 7.466671772053069e-08, "loss": 0.0006, "step": 41977 }, { "epoch": 7.7, "learning_rate": 7.457618811074762e-08, "loss": 0.0004, "step": 41978 }, { "epoch": 7.7, "learning_rate": 7.448571321013399e-08, "loss": 0.0005, "step": 41979 }, { "epoch": 7.7, "learning_rate": 7.439529301918603e-08, "loss": 0.0006, "step": 41980 }, { "epoch": 7.7, "learning_rate": 7.430492753840446e-08, "loss": 0.0004, "step": 41981 }, { "epoch": 7.7, "learning_rate": 7.421461676828668e-08, "loss": 0.0006, "step": 41982 }, { "epoch": 7.7, "learning_rate": 7.412436070933115e-08, "loss": 0.0003, "step": 41983 }, { "epoch": 7.7, "learning_rate": 7.403415936203417e-08, "loss": 0.0004, "step": 41984 }, { "epoch": 7.7, "learning_rate": 7.394401272689311e-08, "loss": 0.0004, "step": 41985 }, { "epoch": 7.7, "learning_rate": 7.385392080440535e-08, "loss": 0.0007, "step": 41986 }, { "epoch": 7.7, "learning_rate": 7.376388359506714e-08, "loss": 0.001, "step": 41987 }, { "epoch": 7.7, "learning_rate": 7.367390109937478e-08, "loss": 0.0006, "step": 41988 }, { "epoch": 7.7, "learning_rate": 7.358397331782452e-08, "loss": 0.0004, "step": 41989 }, { "epoch": 7.7, "learning_rate": 7.349410025091264e-08, "loss": 0.0003, "step": 41990 }, { "epoch": 7.7, "learning_rate": 7.340428189913429e-08, "loss": 0.0013, "step": 41991 }, { "epoch": 7.7, "learning_rate": 7.331451826298241e-08, "loss": 0.0002, "step": 41992 }, { "epoch": 7.7, "learning_rate": 7.322480934295439e-08, "loss": 0.0005, "step": 41993 }, { "epoch": 7.7, "learning_rate": 7.313515513954316e-08, "loss": 0.0007, "step": 41994 }, { "epoch": 7.7, "learning_rate": 7.30455556532439e-08, "loss": 0.0002, "step": 41995 }, { "epoch": 7.7, "learning_rate": 7.295601088455062e-08, "loss": 0.0006, "step": 41996 }, { "epoch": 7.7, "learning_rate": 7.286652083395629e-08, "loss": 0.0005, "step": 41997 }, { "epoch": 7.7, "learning_rate": 7.277708550195384e-08, "loss": 0.0002, "step": 41998 }, { "epoch": 7.7, "learning_rate": 7.26877048890362e-08, "loss": 0.0003, "step": 41999 }, { "epoch": 7.7, "learning_rate": 7.259837899569744e-08, "loss": 0.0002, "step": 42000 }, { "epoch": 7.7, "learning_rate": 7.250910782242826e-08, "loss": 0.0002, "step": 42001 }, { "epoch": 7.7, "learning_rate": 7.241989136972272e-08, "loss": 0.0003, "step": 42002 }, { "epoch": 7.7, "learning_rate": 7.233072963807042e-08, "loss": 0.0004, "step": 42003 }, { "epoch": 7.7, "learning_rate": 7.224162262796319e-08, "loss": 0.0004, "step": 42004 }, { "epoch": 7.7, "learning_rate": 7.215257033989398e-08, "loss": 0.0004, "step": 42005 }, { "epoch": 7.7, "learning_rate": 7.206357277435127e-08, "loss": 0.0004, "step": 42006 }, { "epoch": 7.7, "learning_rate": 7.197462993182691e-08, "loss": 0.0008, "step": 42007 }, { "epoch": 7.7, "learning_rate": 7.18857418128105e-08, "loss": 0.0012, "step": 42008 }, { "epoch": 7.7, "learning_rate": 7.179690841779385e-08, "loss": 0.0002, "step": 42009 }, { "epoch": 7.7, "learning_rate": 7.170812974726438e-08, "loss": 0.0004, "step": 42010 }, { "epoch": 7.7, "learning_rate": 7.161940580171278e-08, "loss": 0.0003, "step": 42011 }, { "epoch": 7.7, "learning_rate": 7.153073658162646e-08, "loss": 0.0004, "step": 42012 }, { "epoch": 7.7, "learning_rate": 7.144212208749613e-08, "loss": 0.0006, "step": 42013 }, { "epoch": 7.7, "learning_rate": 7.135356231981028e-08, "loss": 0.0002, "step": 42014 }, { "epoch": 7.7, "learning_rate": 7.12650572790552e-08, "loss": 0.0003, "step": 42015 }, { "epoch": 7.71, "learning_rate": 7.117660696571938e-08, "loss": 0.0003, "step": 42016 }, { "epoch": 7.71, "learning_rate": 7.108821138029132e-08, "loss": 0.0004, "step": 42017 }, { "epoch": 7.71, "learning_rate": 7.099987052325841e-08, "loss": 0.0006, "step": 42018 }, { "epoch": 7.71, "learning_rate": 7.091158439510582e-08, "loss": 0.0005, "step": 42019 }, { "epoch": 7.71, "learning_rate": 7.082335299632092e-08, "loss": 0.0003, "step": 42020 }, { "epoch": 7.71, "learning_rate": 7.073517632739113e-08, "loss": 0.0003, "step": 42021 }, { "epoch": 7.71, "learning_rate": 7.064705438880048e-08, "loss": 0.0009, "step": 42022 }, { "epoch": 7.71, "learning_rate": 7.055898718103749e-08, "loss": 0.0006, "step": 42023 }, { "epoch": 7.71, "learning_rate": 7.047097470458508e-08, "loss": 0.0007, "step": 42024 }, { "epoch": 7.71, "learning_rate": 7.038301695992844e-08, "loss": 0.0003, "step": 42025 }, { "epoch": 7.71, "learning_rate": 7.029511394755495e-08, "loss": 0.0003, "step": 42026 }, { "epoch": 7.71, "learning_rate": 7.020726566794645e-08, "loss": 0.0004, "step": 42027 }, { "epoch": 7.71, "learning_rate": 7.01194721215881e-08, "loss": 0.0002, "step": 42028 }, { "epoch": 7.71, "learning_rate": 7.003173330896396e-08, "loss": 0.0003, "step": 42029 }, { "epoch": 7.71, "learning_rate": 6.994404923055808e-08, "loss": 0.0002, "step": 42030 }, { "epoch": 7.71, "learning_rate": 6.985641988685232e-08, "loss": 0.0002, "step": 42031 }, { "epoch": 7.71, "learning_rate": 6.976884527833184e-08, "loss": 0.0009, "step": 42032 }, { "epoch": 7.71, "learning_rate": 6.968132540547735e-08, "loss": 0.0002, "step": 42033 }, { "epoch": 7.71, "learning_rate": 6.95938602687718e-08, "loss": 0.0002, "step": 42034 }, { "epoch": 7.71, "learning_rate": 6.950644986869815e-08, "loss": 0.0004, "step": 42035 }, { "epoch": 7.71, "learning_rate": 6.941909420573711e-08, "loss": 0.0002, "step": 42036 }, { "epoch": 7.71, "learning_rate": 6.933179328037054e-08, "loss": 0.0003, "step": 42037 }, { "epoch": 7.71, "learning_rate": 6.924454709308026e-08, "loss": 0.0006, "step": 42038 }, { "epoch": 7.71, "learning_rate": 6.91573556443459e-08, "loss": 0.0002, "step": 42039 }, { "epoch": 7.71, "learning_rate": 6.907021893465039e-08, "loss": 0.0004, "step": 42040 }, { "epoch": 7.71, "learning_rate": 6.898313696447112e-08, "loss": 0.0005, "step": 42041 }, { "epoch": 7.71, "learning_rate": 6.889610973428885e-08, "loss": 0.0002, "step": 42042 }, { "epoch": 7.71, "learning_rate": 6.880913724458538e-08, "loss": 0.0004, "step": 42043 }, { "epoch": 7.71, "learning_rate": 6.872221949583924e-08, "loss": 0.0004, "step": 42044 }, { "epoch": 7.71, "learning_rate": 6.863535648852781e-08, "loss": 0.0003, "step": 42045 }, { "epoch": 7.71, "learning_rate": 6.854854822313073e-08, "loss": 0.0004, "step": 42046 }, { "epoch": 7.71, "learning_rate": 6.846179470012648e-08, "loss": 0.0002, "step": 42047 }, { "epoch": 7.71, "learning_rate": 6.837509591999359e-08, "loss": 0.0004, "step": 42048 }, { "epoch": 7.71, "learning_rate": 6.828845188321054e-08, "loss": 0.0002, "step": 42049 }, { "epoch": 7.71, "learning_rate": 6.820186259025363e-08, "loss": 0.0004, "step": 42050 }, { "epoch": 7.71, "learning_rate": 6.811532804160137e-08, "loss": 0.0006, "step": 42051 }, { "epoch": 7.71, "learning_rate": 6.802884823772893e-08, "loss": 0.0003, "step": 42052 }, { "epoch": 7.71, "learning_rate": 6.794242317911593e-08, "loss": 0.0007, "step": 42053 }, { "epoch": 7.71, "learning_rate": 6.785605286623532e-08, "loss": 0.0002, "step": 42054 }, { "epoch": 7.71, "learning_rate": 6.776973729956559e-08, "loss": 0.0003, "step": 42055 }, { "epoch": 7.71, "learning_rate": 6.768347647958196e-08, "loss": 0.0005, "step": 42056 }, { "epoch": 7.71, "learning_rate": 6.759727040675956e-08, "loss": 0.0002, "step": 42057 }, { "epoch": 7.71, "learning_rate": 6.75111190815736e-08, "loss": 0.0003, "step": 42058 }, { "epoch": 7.71, "learning_rate": 6.742502250449811e-08, "loss": 0.0002, "step": 42059 }, { "epoch": 7.71, "learning_rate": 6.733898067600941e-08, "loss": 0.0002, "step": 42060 }, { "epoch": 7.71, "learning_rate": 6.725299359658155e-08, "loss": 0.0007, "step": 42061 }, { "epoch": 7.71, "learning_rate": 6.716706126668748e-08, "loss": 0.0007, "step": 42062 }, { "epoch": 7.71, "learning_rate": 6.708118368680128e-08, "loss": 0.0005, "step": 42063 }, { "epoch": 7.71, "learning_rate": 6.699536085739589e-08, "loss": 0.0003, "step": 42064 }, { "epoch": 7.71, "learning_rate": 6.690959277894427e-08, "loss": 0.0008, "step": 42065 }, { "epoch": 7.71, "learning_rate": 6.682387945192048e-08, "loss": 0.0008, "step": 42066 }, { "epoch": 7.71, "learning_rate": 6.673822087679638e-08, "loss": 0.0005, "step": 42067 }, { "epoch": 7.71, "learning_rate": 6.66526170540438e-08, "loss": 0.0005, "step": 42068 }, { "epoch": 7.71, "learning_rate": 6.65670679841346e-08, "loss": 0.0005, "step": 42069 }, { "epoch": 7.72, "learning_rate": 6.64815736675406e-08, "loss": 0.0007, "step": 42070 }, { "epoch": 7.72, "learning_rate": 6.639613410473256e-08, "loss": 0.0006, "step": 42071 }, { "epoch": 7.72, "learning_rate": 6.631074929618231e-08, "loss": 0.0004, "step": 42072 }, { "epoch": 7.72, "learning_rate": 6.622541924235837e-08, "loss": 0.0006, "step": 42073 }, { "epoch": 7.72, "learning_rate": 6.61401439437348e-08, "loss": 0.0003, "step": 42074 }, { "epoch": 7.72, "learning_rate": 6.60549234007779e-08, "loss": 0.0004, "step": 42075 }, { "epoch": 7.72, "learning_rate": 6.596975761395952e-08, "loss": 0.0003, "step": 42076 }, { "epoch": 7.72, "learning_rate": 6.588464658374816e-08, "loss": 0.0004, "step": 42077 }, { "epoch": 7.72, "learning_rate": 6.579959031061455e-08, "loss": 0.0007, "step": 42078 }, { "epoch": 7.72, "learning_rate": 6.571458879502501e-08, "loss": 0.0005, "step": 42079 }, { "epoch": 7.72, "learning_rate": 6.562964203745137e-08, "loss": 0.0007, "step": 42080 }, { "epoch": 7.72, "learning_rate": 6.554475003835769e-08, "loss": 0.0005, "step": 42081 }, { "epoch": 7.72, "learning_rate": 6.545991279821474e-08, "loss": 0.0013, "step": 42082 }, { "epoch": 7.72, "learning_rate": 6.537513031748988e-08, "loss": 0.0004, "step": 42083 }, { "epoch": 7.72, "learning_rate": 6.529040259664943e-08, "loss": 0.0007, "step": 42084 }, { "epoch": 7.72, "learning_rate": 6.52057296361619e-08, "loss": 0.0004, "step": 42085 }, { "epoch": 7.72, "learning_rate": 6.512111143649247e-08, "loss": 0.0003, "step": 42086 }, { "epoch": 7.72, "learning_rate": 6.503654799810854e-08, "loss": 0.0008, "step": 42087 }, { "epoch": 7.72, "learning_rate": 6.495203932147531e-08, "loss": 0.0008, "step": 42088 }, { "epoch": 7.72, "learning_rate": 6.486758540706017e-08, "loss": 0.0003, "step": 42089 }, { "epoch": 7.72, "learning_rate": 6.478318625532721e-08, "loss": 0.0005, "step": 42090 }, { "epoch": 7.72, "learning_rate": 6.469884186674159e-08, "loss": 0.0003, "step": 42091 }, { "epoch": 7.72, "learning_rate": 6.46145522417696e-08, "loss": 0.0006, "step": 42092 }, { "epoch": 7.72, "learning_rate": 6.453031738087424e-08, "loss": 0.0007, "step": 42093 }, { "epoch": 7.72, "learning_rate": 6.444613728452064e-08, "loss": 0.0003, "step": 42094 }, { "epoch": 7.72, "learning_rate": 6.436201195317182e-08, "loss": 0.0003, "step": 42095 }, { "epoch": 7.72, "learning_rate": 6.427794138729404e-08, "loss": 0.0003, "step": 42096 }, { "epoch": 7.72, "learning_rate": 6.419392558734806e-08, "loss": 0.0005, "step": 42097 }, { "epoch": 7.72, "learning_rate": 6.410996455379903e-08, "loss": 0.0005, "step": 42098 }, { "epoch": 7.72, "learning_rate": 6.402605828710773e-08, "loss": 0.001, "step": 42099 }, { "epoch": 7.72, "learning_rate": 6.39422067877371e-08, "loss": 0.001, "step": 42100 }, { "epoch": 7.72, "learning_rate": 6.385841005615013e-08, "loss": 0.0005, "step": 42101 }, { "epoch": 7.72, "learning_rate": 6.377466809280863e-08, "loss": 0.0005, "step": 42102 }, { "epoch": 7.72, "learning_rate": 6.36909808981745e-08, "loss": 0.0006, "step": 42103 }, { "epoch": 7.72, "learning_rate": 6.360734847270733e-08, "loss": 0.0001, "step": 42104 }, { "epoch": 7.72, "learning_rate": 6.352377081687011e-08, "loss": 0.0006, "step": 42105 }, { "epoch": 7.72, "learning_rate": 6.344024793112246e-08, "loss": 0.0002, "step": 42106 }, { "epoch": 7.72, "learning_rate": 6.335677981592514e-08, "loss": 0.0002, "step": 42107 }, { "epoch": 7.72, "learning_rate": 6.327336647173665e-08, "loss": 0.0001, "step": 42108 }, { "epoch": 7.72, "learning_rate": 6.319000789901996e-08, "loss": 0.0004, "step": 42109 }, { "epoch": 7.72, "learning_rate": 6.31067040982325e-08, "loss": 0.0002, "step": 42110 }, { "epoch": 7.72, "learning_rate": 6.302345506983276e-08, "loss": 0.0003, "step": 42111 }, { "epoch": 7.72, "learning_rate": 6.29402608142815e-08, "loss": 0.0001, "step": 42112 }, { "epoch": 7.72, "learning_rate": 6.285712133203503e-08, "loss": 0.0002, "step": 42113 }, { "epoch": 7.72, "learning_rate": 6.277403662355408e-08, "loss": 0.0003, "step": 42114 }, { "epoch": 7.72, "learning_rate": 6.269100668929495e-08, "loss": 0.0002, "step": 42115 }, { "epoch": 7.72, "learning_rate": 6.260803152971617e-08, "loss": 0.0011, "step": 42116 }, { "epoch": 7.72, "learning_rate": 6.252511114527404e-08, "loss": 0.0002, "step": 42117 }, { "epoch": 7.72, "learning_rate": 6.244224553642708e-08, "loss": 0.0004, "step": 42118 }, { "epoch": 7.72, "learning_rate": 6.235943470363049e-08, "loss": 0.0003, "step": 42119 }, { "epoch": 7.72, "learning_rate": 6.227667864734277e-08, "loss": 0.0007, "step": 42120 }, { "epoch": 7.72, "learning_rate": 6.219397736801802e-08, "loss": 0.0002, "step": 42121 }, { "epoch": 7.72, "learning_rate": 6.211133086611254e-08, "loss": 0.001, "step": 42122 }, { "epoch": 7.72, "learning_rate": 6.202873914208263e-08, "loss": 0.0005, "step": 42123 }, { "epoch": 7.72, "learning_rate": 6.194620219638236e-08, "loss": 0.0004, "step": 42124 }, { "epoch": 7.73, "learning_rate": 6.186372002946917e-08, "loss": 0.0005, "step": 42125 }, { "epoch": 7.73, "learning_rate": 6.178129264179378e-08, "loss": 0.001, "step": 42126 }, { "epoch": 7.73, "learning_rate": 6.169892003381472e-08, "loss": 0.0008, "step": 42127 }, { "epoch": 7.73, "learning_rate": 6.161660220598276e-08, "loss": 0.0006, "step": 42128 }, { "epoch": 7.73, "learning_rate": 6.153433915875417e-08, "loss": 0.0003, "step": 42129 }, { "epoch": 7.73, "learning_rate": 6.145213089257973e-08, "loss": 0.0003, "step": 42130 }, { "epoch": 7.73, "learning_rate": 6.136997740791461e-08, "loss": 0.0003, "step": 42131 }, { "epoch": 7.73, "learning_rate": 6.12878787052107e-08, "loss": 0.0003, "step": 42132 }, { "epoch": 7.73, "learning_rate": 6.120583478492093e-08, "loss": 0.0006, "step": 42133 }, { "epoch": 7.73, "learning_rate": 6.112384564749829e-08, "loss": 0.0006, "step": 42134 }, { "epoch": 7.73, "learning_rate": 6.104191129339243e-08, "loss": 0.0003, "step": 42135 }, { "epoch": 7.73, "learning_rate": 6.096003172305742e-08, "loss": 0.0007, "step": 42136 }, { "epoch": 7.73, "learning_rate": 6.087820693694402e-08, "loss": 0.0007, "step": 42137 }, { "epoch": 7.73, "learning_rate": 6.079643693550186e-08, "loss": 0.0007, "step": 42138 }, { "epoch": 7.73, "learning_rate": 6.07147217191828e-08, "loss": 0.0004, "step": 42139 }, { "epoch": 7.73, "learning_rate": 6.06330612884376e-08, "loss": 0.0004, "step": 42140 }, { "epoch": 7.73, "learning_rate": 6.055145564371478e-08, "loss": 0.0002, "step": 42141 }, { "epoch": 7.73, "learning_rate": 6.046990478546622e-08, "loss": 0.0002, "step": 42142 }, { "epoch": 7.73, "learning_rate": 6.038840871413931e-08, "loss": 0.0004, "step": 42143 }, { "epoch": 7.73, "learning_rate": 6.030696743018483e-08, "loss": 0.0002, "step": 42144 }, { "epoch": 7.73, "learning_rate": 6.022558093405129e-08, "loss": 0.0005, "step": 42145 }, { "epoch": 7.73, "learning_rate": 6.014424922618722e-08, "loss": 0.0004, "step": 42146 }, { "epoch": 7.73, "learning_rate": 6.006297230704116e-08, "loss": 0.0002, "step": 42147 }, { "epoch": 7.73, "learning_rate": 5.998175017706053e-08, "loss": 0.0002, "step": 42148 }, { "epoch": 7.73, "learning_rate": 5.990058283669276e-08, "loss": 0.0003, "step": 42149 }, { "epoch": 7.73, "learning_rate": 5.981947028638746e-08, "loss": 0.0003, "step": 42150 }, { "epoch": 7.73, "learning_rate": 5.973841252658874e-08, "loss": 0.0004, "step": 42151 }, { "epoch": 7.73, "learning_rate": 5.965740955774624e-08, "loss": 0.0003, "step": 42152 }, { "epoch": 7.73, "learning_rate": 5.957646138030293e-08, "loss": 0.0011, "step": 42153 }, { "epoch": 7.73, "learning_rate": 5.949556799470846e-08, "loss": 0.001, "step": 42154 }, { "epoch": 7.73, "learning_rate": 5.941472940140691e-08, "loss": 0.0003, "step": 42155 }, { "epoch": 7.73, "learning_rate": 5.93339456008446e-08, "loss": 0.0007, "step": 42156 }, { "epoch": 7.73, "learning_rate": 5.925321659346561e-08, "loss": 0.001, "step": 42157 }, { "epoch": 7.73, "learning_rate": 5.917254237971626e-08, "loss": 0.0004, "step": 42158 }, { "epoch": 7.73, "learning_rate": 5.909192296004063e-08, "loss": 0.0003, "step": 42159 }, { "epoch": 7.73, "learning_rate": 5.901135833488392e-08, "loss": 0.0004, "step": 42160 }, { "epoch": 7.73, "learning_rate": 5.8930848504687996e-08, "loss": 0.0005, "step": 42161 }, { "epoch": 7.73, "learning_rate": 5.885039346989918e-08, "loss": 0.0011, "step": 42162 }, { "epoch": 7.73, "learning_rate": 5.8769993230960445e-08, "loss": 0.0003, "step": 42163 }, { "epoch": 7.73, "learning_rate": 5.8689647788313654e-08, "loss": 0.0004, "step": 42164 }, { "epoch": 7.73, "learning_rate": 5.86093571424029e-08, "loss": 0.0002, "step": 42165 }, { "epoch": 7.73, "learning_rate": 5.8529121293668946e-08, "loss": 0.0009, "step": 42166 }, { "epoch": 7.73, "learning_rate": 5.844894024255699e-08, "loss": 0.0002, "step": 42167 }, { "epoch": 7.73, "learning_rate": 5.836881398950667e-08, "loss": 0.0002, "step": 42168 }, { "epoch": 7.73, "learning_rate": 5.828874253496097e-08, "loss": 0.0004, "step": 42169 }, { "epoch": 7.73, "learning_rate": 5.820872587935955e-08, "loss": 0.0006, "step": 42170 }, { "epoch": 7.73, "learning_rate": 5.812876402314538e-08, "loss": 0.0003, "step": 42171 }, { "epoch": 7.73, "learning_rate": 5.80488569667581e-08, "loss": 0.0006, "step": 42172 }, { "epoch": 7.73, "learning_rate": 5.796900471063849e-08, "loss": 0.0007, "step": 42173 }, { "epoch": 7.73, "learning_rate": 5.7889207255226174e-08, "loss": 0.0002, "step": 42174 }, { "epoch": 7.73, "learning_rate": 5.780946460096193e-08, "loss": 0.0006, "step": 42175 }, { "epoch": 7.73, "learning_rate": 5.772977674828428e-08, "loss": 0.0002, "step": 42176 }, { "epoch": 7.73, "learning_rate": 5.7650143697633997e-08, "loss": 0.0004, "step": 42177 }, { "epoch": 7.73, "learning_rate": 5.75705654494485e-08, "loss": 0.0006, "step": 42178 }, { "epoch": 7.74, "learning_rate": 5.749104200416633e-08, "loss": 0.0002, "step": 42179 }, { "epoch": 7.74, "learning_rate": 5.741157336222713e-08, "loss": 0.0004, "step": 42180 }, { "epoch": 7.74, "learning_rate": 5.733215952406945e-08, "loss": 0.0003, "step": 42181 }, { "epoch": 7.74, "learning_rate": 5.7252800490128494e-08, "loss": 0.0005, "step": 42182 }, { "epoch": 7.74, "learning_rate": 5.717349626084279e-08, "loss": 0.0003, "step": 42183 }, { "epoch": 7.74, "learning_rate": 5.7094246836649767e-08, "loss": 0.0005, "step": 42184 }, { "epoch": 7.74, "learning_rate": 5.701505221798687e-08, "loss": 0.0005, "step": 42185 }, { "epoch": 7.74, "learning_rate": 5.6935912405289284e-08, "loss": 0.0003, "step": 42186 }, { "epoch": 7.74, "learning_rate": 5.685682739899445e-08, "loss": 0.0002, "step": 42187 }, { "epoch": 7.74, "learning_rate": 5.6777797199537575e-08, "loss": 0.0004, "step": 42188 }, { "epoch": 7.74, "learning_rate": 5.6698821807354975e-08, "loss": 0.0007, "step": 42189 }, { "epoch": 7.74, "learning_rate": 5.6619901222880744e-08, "loss": 0.0005, "step": 42190 }, { "epoch": 7.74, "learning_rate": 5.6541035446550096e-08, "loss": 0.0004, "step": 42191 }, { "epoch": 7.74, "learning_rate": 5.646222447879934e-08, "loss": 0.0003, "step": 42192 }, { "epoch": 7.74, "learning_rate": 5.6383468320060365e-08, "loss": 0.0003, "step": 42193 }, { "epoch": 7.74, "learning_rate": 5.630476697076948e-08, "loss": 0.0003, "step": 42194 }, { "epoch": 7.74, "learning_rate": 5.622612043135966e-08, "loss": 0.0004, "step": 42195 }, { "epoch": 7.74, "learning_rate": 5.6147528702265033e-08, "loss": 0.0004, "step": 42196 }, { "epoch": 7.74, "learning_rate": 5.606899178391745e-08, "loss": 0.0005, "step": 42197 }, { "epoch": 7.74, "learning_rate": 5.5990509676751016e-08, "loss": 0.0002, "step": 42198 }, { "epoch": 7.74, "learning_rate": 5.591208238119761e-08, "loss": 0.0006, "step": 42199 }, { "epoch": 7.74, "learning_rate": 5.58337098976891e-08, "loss": 0.0008, "step": 42200 }, { "epoch": 7.74, "learning_rate": 5.575539222665849e-08, "loss": 0.0004, "step": 42201 }, { "epoch": 7.74, "learning_rate": 5.5677129368537644e-08, "loss": 0.0007, "step": 42202 }, { "epoch": 7.74, "learning_rate": 5.5598921323757325e-08, "loss": 0.0003, "step": 42203 }, { "epoch": 7.74, "learning_rate": 5.552076809274831e-08, "loss": 0.0002, "step": 42204 }, { "epoch": 7.74, "learning_rate": 5.5442669675941364e-08, "loss": 0.001, "step": 42205 }, { "epoch": 7.74, "learning_rate": 5.536462607376836e-08, "loss": 0.0007, "step": 42206 }, { "epoch": 7.74, "learning_rate": 5.528663728665895e-08, "loss": 0.0005, "step": 42207 }, { "epoch": 7.74, "learning_rate": 5.520870331504169e-08, "loss": 0.0004, "step": 42208 }, { "epoch": 7.74, "learning_rate": 5.513082415934734e-08, "loss": 0.0005, "step": 42209 }, { "epoch": 7.74, "learning_rate": 5.505299982000445e-08, "loss": 0.0002, "step": 42210 }, { "epoch": 7.74, "learning_rate": 5.4975230297442675e-08, "loss": 0.0002, "step": 42211 }, { "epoch": 7.74, "learning_rate": 5.4897515592090556e-08, "loss": 0.0007, "step": 42212 }, { "epoch": 7.74, "learning_rate": 5.4819855704376645e-08, "loss": 0.0003, "step": 42213 }, { "epoch": 7.74, "learning_rate": 5.474225063472727e-08, "loss": 0.0004, "step": 42214 }, { "epoch": 7.74, "learning_rate": 5.466470038357208e-08, "loss": 0.0004, "step": 42215 }, { "epoch": 7.74, "learning_rate": 5.4587204951339625e-08, "loss": 0.0003, "step": 42216 }, { "epoch": 7.74, "learning_rate": 5.4509764338454004e-08, "loss": 0.0011, "step": 42217 }, { "epoch": 7.74, "learning_rate": 5.4432378545343775e-08, "loss": 0.0005, "step": 42218 }, { "epoch": 7.74, "learning_rate": 5.435504757243526e-08, "loss": 0.0001, "step": 42219 }, { "epoch": 7.74, "learning_rate": 5.427777142015478e-08, "loss": 0.0003, "step": 42220 }, { "epoch": 7.74, "learning_rate": 5.420055008892866e-08, "loss": 0.0002, "step": 42221 }, { "epoch": 7.74, "learning_rate": 5.412338357918212e-08, "loss": 0.0007, "step": 42222 }, { "epoch": 7.74, "learning_rate": 5.404627189134037e-08, "loss": 0.0004, "step": 42223 }, { "epoch": 7.74, "learning_rate": 5.396921502582864e-08, "loss": 0.0003, "step": 42224 }, { "epoch": 7.74, "learning_rate": 5.389221298307101e-08, "loss": 0.0004, "step": 42225 }, { "epoch": 7.74, "learning_rate": 5.381526576349272e-08, "loss": 0.0001, "step": 42226 }, { "epoch": 7.74, "learning_rate": 5.373837336751786e-08, "loss": 0.0002, "step": 42227 }, { "epoch": 7.74, "learning_rate": 5.366153579557054e-08, "loss": 0.0002, "step": 42228 }, { "epoch": 7.74, "learning_rate": 5.3584753048073756e-08, "loss": 0.0001, "step": 42229 }, { "epoch": 7.74, "learning_rate": 5.35080251254505e-08, "loss": 0.0005, "step": 42230 }, { "epoch": 7.74, "learning_rate": 5.3431352028123774e-08, "loss": 0.0005, "step": 42231 }, { "epoch": 7.74, "learning_rate": 5.3354733756516566e-08, "loss": 0.0003, "step": 42232 }, { "epoch": 7.74, "learning_rate": 5.327817031105076e-08, "loss": 0.0005, "step": 42233 }, { "epoch": 7.75, "learning_rate": 5.3201661692149356e-08, "loss": 0.001, "step": 42234 }, { "epoch": 7.75, "learning_rate": 5.3125207900233125e-08, "loss": 0.0004, "step": 42235 }, { "epoch": 7.75, "learning_rate": 5.304880893572395e-08, "loss": 0.0002, "step": 42236 }, { "epoch": 7.75, "learning_rate": 5.297246479904261e-08, "loss": 0.0003, "step": 42237 }, { "epoch": 7.75, "learning_rate": 5.289617549060988e-08, "loss": 0.0003, "step": 42238 }, { "epoch": 7.75, "learning_rate": 5.2819941010847643e-08, "loss": 0.0003, "step": 42239 }, { "epoch": 7.75, "learning_rate": 5.274376136017334e-08, "loss": 0.0002, "step": 42240 }, { "epoch": 7.75, "learning_rate": 5.266763653900997e-08, "loss": 0.0006, "step": 42241 }, { "epoch": 7.75, "learning_rate": 5.2591566547774975e-08, "loss": 0.0004, "step": 42242 }, { "epoch": 7.75, "learning_rate": 5.251555138688802e-08, "loss": 0.0009, "step": 42243 }, { "epoch": 7.75, "learning_rate": 5.243959105676877e-08, "loss": 0.0005, "step": 42244 }, { "epoch": 7.75, "learning_rate": 5.2363685557834664e-08, "loss": 0.0008, "step": 42245 }, { "epoch": 7.75, "learning_rate": 5.2287834890505376e-08, "loss": 0.0005, "step": 42246 }, { "epoch": 7.75, "learning_rate": 5.221203905519945e-08, "loss": 0.0004, "step": 42247 }, { "epoch": 7.75, "learning_rate": 5.213629805233322e-08, "loss": 0.0002, "step": 42248 }, { "epoch": 7.75, "learning_rate": 5.2060611882325254e-08, "loss": 0.0007, "step": 42249 }, { "epoch": 7.75, "learning_rate": 5.1984980545590755e-08, "loss": 0.0004, "step": 42250 }, { "epoch": 7.75, "learning_rate": 5.1909404042549405e-08, "loss": 0.0002, "step": 42251 }, { "epoch": 7.75, "learning_rate": 5.183388237361642e-08, "loss": 0.0005, "step": 42252 }, { "epoch": 7.75, "learning_rate": 5.175841553920813e-08, "loss": 0.0006, "step": 42253 }, { "epoch": 7.75, "learning_rate": 5.168300353973976e-08, "loss": 0.0004, "step": 42254 }, { "epoch": 7.75, "learning_rate": 5.1607646375627654e-08, "loss": 0.0005, "step": 42255 }, { "epoch": 7.75, "learning_rate": 5.153234404728813e-08, "loss": 0.001, "step": 42256 }, { "epoch": 7.75, "learning_rate": 5.145709655513531e-08, "loss": 0.0002, "step": 42257 }, { "epoch": 7.75, "learning_rate": 5.1381903899584415e-08, "loss": 0.0002, "step": 42258 }, { "epoch": 7.75, "learning_rate": 5.1306766081048456e-08, "loss": 0.0003, "step": 42259 }, { "epoch": 7.75, "learning_rate": 5.1231683099943753e-08, "loss": 0.0004, "step": 42260 }, { "epoch": 7.75, "learning_rate": 5.1156654956682204e-08, "loss": 0.0003, "step": 42261 }, { "epoch": 7.75, "learning_rate": 5.108168165167904e-08, "loss": 0.0003, "step": 42262 }, { "epoch": 7.75, "learning_rate": 5.1006763185346143e-08, "loss": 0.0002, "step": 42263 }, { "epoch": 7.75, "learning_rate": 5.093189955809652e-08, "loss": 0.0006, "step": 42264 }, { "epoch": 7.75, "learning_rate": 5.0857090770344284e-08, "loss": 0.0006, "step": 42265 }, { "epoch": 7.75, "learning_rate": 5.078233682250133e-08, "loss": 0.0003, "step": 42266 }, { "epoch": 7.75, "learning_rate": 5.0707637714977334e-08, "loss": 0.0004, "step": 42267 }, { "epoch": 7.75, "learning_rate": 5.063299344818751e-08, "loss": 0.0006, "step": 42268 }, { "epoch": 7.75, "learning_rate": 5.055840402254153e-08, "loss": 0.0002, "step": 42269 }, { "epoch": 7.75, "learning_rate": 5.0483869438450186e-08, "loss": 0.0003, "step": 42270 }, { "epoch": 7.75, "learning_rate": 5.0409389696325364e-08, "loss": 0.0003, "step": 42271 }, { "epoch": 7.75, "learning_rate": 5.033496479657673e-08, "loss": 0.0005, "step": 42272 }, { "epoch": 7.75, "learning_rate": 5.0260594739615086e-08, "loss": 0.0004, "step": 42273 }, { "epoch": 7.75, "learning_rate": 5.018627952585009e-08, "loss": 0.0004, "step": 42274 }, { "epoch": 7.75, "learning_rate": 5.0112019155691414e-08, "loss": 0.0002, "step": 42275 }, { "epoch": 7.75, "learning_rate": 5.003781362954763e-08, "loss": 0.0007, "step": 42276 }, { "epoch": 7.75, "learning_rate": 4.9963662947829505e-08, "loss": 0.0009, "step": 42277 }, { "epoch": 7.75, "learning_rate": 4.9889567110944504e-08, "loss": 0.0004, "step": 42278 }, { "epoch": 7.75, "learning_rate": 4.9815526119301185e-08, "loss": 0.0006, "step": 42279 }, { "epoch": 7.75, "learning_rate": 4.974153997330811e-08, "loss": 0.0002, "step": 42280 }, { "epoch": 7.75, "learning_rate": 4.966760867337272e-08, "loss": 0.0004, "step": 42281 }, { "epoch": 7.75, "learning_rate": 4.959373221990249e-08, "loss": 0.0004, "step": 42282 }, { "epoch": 7.75, "learning_rate": 4.9519910613304853e-08, "loss": 0.0004, "step": 42283 }, { "epoch": 7.75, "learning_rate": 4.944614385398616e-08, "loss": 0.0005, "step": 42284 }, { "epoch": 7.75, "learning_rate": 4.937243194235275e-08, "loss": 0.0004, "step": 42285 }, { "epoch": 7.75, "learning_rate": 4.9298774878813184e-08, "loss": 0.0003, "step": 42286 }, { "epoch": 7.75, "learning_rate": 4.922517266377158e-08, "loss": 0.0003, "step": 42287 }, { "epoch": 7.75, "learning_rate": 4.9151625297633175e-08, "loss": 0.0007, "step": 42288 }, { "epoch": 7.76, "learning_rate": 4.907813278080431e-08, "loss": 0.0004, "step": 42289 }, { "epoch": 7.76, "learning_rate": 4.900469511369022e-08, "loss": 0.0008, "step": 42290 }, { "epoch": 7.76, "learning_rate": 4.893131229669501e-08, "loss": 0.0004, "step": 42291 }, { "epoch": 7.76, "learning_rate": 4.885798433022393e-08, "loss": 0.0005, "step": 42292 }, { "epoch": 7.76, "learning_rate": 4.878471121468109e-08, "loss": 0.0011, "step": 42293 }, { "epoch": 7.76, "learning_rate": 4.871149295046951e-08, "loss": 0.0005, "step": 42294 }, { "epoch": 7.76, "learning_rate": 4.86383295379933e-08, "loss": 0.0005, "step": 42295 }, { "epoch": 7.76, "learning_rate": 4.8565220977656594e-08, "loss": 0.0003, "step": 42296 }, { "epoch": 7.76, "learning_rate": 4.849216726986017e-08, "loss": 0.0003, "step": 42297 }, { "epoch": 7.76, "learning_rate": 4.8419168415009264e-08, "loss": 0.0004, "step": 42298 }, { "epoch": 7.76, "learning_rate": 4.834622441350467e-08, "loss": 0.0002, "step": 42299 }, { "epoch": 7.76, "learning_rate": 4.827333526574829e-08, "loss": 0.0007, "step": 42300 }, { "epoch": 7.76, "learning_rate": 4.820050097214313e-08, "loss": 0.0003, "step": 42301 }, { "epoch": 7.76, "learning_rate": 4.812772153308998e-08, "loss": 0.0003, "step": 42302 }, { "epoch": 7.76, "learning_rate": 4.805499694898852e-08, "loss": 0.0007, "step": 42303 }, { "epoch": 7.76, "learning_rate": 4.7982327220242874e-08, "loss": 0.0004, "step": 42304 }, { "epoch": 7.76, "learning_rate": 4.7909712347250505e-08, "loss": 0.0006, "step": 42305 }, { "epoch": 7.76, "learning_rate": 4.783715233041441e-08, "loss": 0.0003, "step": 42306 }, { "epoch": 7.76, "learning_rate": 4.7764647170132074e-08, "loss": 0.0005, "step": 42307 }, { "epoch": 7.76, "learning_rate": 4.769219686680426e-08, "loss": 0.0008, "step": 42308 }, { "epoch": 7.76, "learning_rate": 4.761980142082956e-08, "loss": 0.0003, "step": 42309 }, { "epoch": 7.76, "learning_rate": 4.754746083260875e-08, "loss": 0.0002, "step": 42310 }, { "epoch": 7.76, "learning_rate": 4.7475175102539293e-08, "loss": 0.0004, "step": 42311 }, { "epoch": 7.76, "learning_rate": 4.7402944231019766e-08, "loss": 0.0004, "step": 42312 }, { "epoch": 7.76, "learning_rate": 4.733076821844762e-08, "loss": 0.0002, "step": 42313 }, { "epoch": 7.76, "learning_rate": 4.725864706522254e-08, "loss": 0.0005, "step": 42314 }, { "epoch": 7.76, "learning_rate": 4.718658077174087e-08, "loss": 0.0003, "step": 42315 }, { "epoch": 7.76, "learning_rate": 4.7114569338398973e-08, "loss": 0.0008, "step": 42316 }, { "epoch": 7.76, "learning_rate": 4.704261276559541e-08, "loss": 0.0002, "step": 42317 }, { "epoch": 7.76, "learning_rate": 4.697071105372653e-08, "loss": 0.0002, "step": 42318 }, { "epoch": 7.76, "learning_rate": 4.6898864203187565e-08, "loss": 0.0004, "step": 42319 }, { "epoch": 7.76, "learning_rate": 4.682707221437488e-08, "loss": 0.0007, "step": 42320 }, { "epoch": 7.76, "learning_rate": 4.6755335087685924e-08, "loss": 0.0004, "step": 42321 }, { "epoch": 7.76, "learning_rate": 4.6683652823513725e-08, "loss": 0.0003, "step": 42322 }, { "epoch": 7.76, "learning_rate": 4.6612025422254625e-08, "loss": 0.0008, "step": 42323 }, { "epoch": 7.76, "learning_rate": 4.654045288430386e-08, "loss": 0.0007, "step": 42324 }, { "epoch": 7.76, "learning_rate": 4.646893521005447e-08, "loss": 0.0003, "step": 42325 }, { "epoch": 7.76, "learning_rate": 4.6397472399901665e-08, "loss": 0.0003, "step": 42326 }, { "epoch": 7.76, "learning_rate": 4.63260644542396e-08, "loss": 0.0003, "step": 42327 }, { "epoch": 7.76, "learning_rate": 4.625471137346016e-08, "loss": 0.0005, "step": 42328 }, { "epoch": 7.76, "learning_rate": 4.6183413157959713e-08, "loss": 0.0002, "step": 42329 }, { "epoch": 7.76, "learning_rate": 4.611216980812794e-08, "loss": 0.0002, "step": 42330 }, { "epoch": 7.76, "learning_rate": 4.6040981324360076e-08, "loss": 0.0005, "step": 42331 }, { "epoch": 7.76, "learning_rate": 4.596984770704693e-08, "loss": 0.0004, "step": 42332 }, { "epoch": 7.76, "learning_rate": 4.589876895658152e-08, "loss": 0.0004, "step": 42333 }, { "epoch": 7.76, "learning_rate": 4.5827745073355746e-08, "loss": 0.0004, "step": 42334 }, { "epoch": 7.76, "learning_rate": 4.575677605776041e-08, "loss": 0.0004, "step": 42335 }, { "epoch": 7.76, "learning_rate": 4.5685861910186316e-08, "loss": 0.0003, "step": 42336 }, { "epoch": 7.76, "learning_rate": 4.561500263102536e-08, "loss": 0.0006, "step": 42337 }, { "epoch": 7.76, "learning_rate": 4.554419822066725e-08, "loss": 0.0003, "step": 42338 }, { "epoch": 7.76, "learning_rate": 4.5473448679503876e-08, "loss": 0.0004, "step": 42339 }, { "epoch": 7.76, "learning_rate": 4.540275400792271e-08, "loss": 0.0003, "step": 42340 }, { "epoch": 7.76, "learning_rate": 4.533211420631567e-08, "loss": 0.0002, "step": 42341 }, { "epoch": 7.76, "learning_rate": 4.5261529275071326e-08, "loss": 0.0004, "step": 42342 }, { "epoch": 7.77, "learning_rate": 4.5190999214578255e-08, "loss": 0.0002, "step": 42343 }, { "epoch": 7.77, "learning_rate": 4.512052402522504e-08, "loss": 0.0007, "step": 42344 }, { "epoch": 7.77, "learning_rate": 4.505010370740137e-08, "loss": 0.0002, "step": 42345 }, { "epoch": 7.77, "learning_rate": 4.497973826149471e-08, "loss": 0.0004, "step": 42346 }, { "epoch": 7.77, "learning_rate": 4.490942768789364e-08, "loss": 0.0003, "step": 42347 }, { "epoch": 7.77, "learning_rate": 4.483917198698451e-08, "loss": 0.0004, "step": 42348 }, { "epoch": 7.77, "learning_rate": 4.4768971159154794e-08, "loss": 0.0004, "step": 42349 }, { "epoch": 7.77, "learning_rate": 4.469882520479196e-08, "loss": 0.0003, "step": 42350 }, { "epoch": 7.77, "learning_rate": 4.4628734124281257e-08, "loss": 0.0003, "step": 42351 }, { "epoch": 7.77, "learning_rate": 4.455869791801237e-08, "loss": 0.0002, "step": 42352 }, { "epoch": 7.77, "learning_rate": 4.448871658636833e-08, "loss": 0.0003, "step": 42353 }, { "epoch": 7.77, "learning_rate": 4.441879012973549e-08, "loss": 0.0004, "step": 42354 }, { "epoch": 7.77, "learning_rate": 4.43489185484991e-08, "loss": 0.0002, "step": 42355 }, { "epoch": 7.77, "learning_rate": 4.4279101843044406e-08, "loss": 0.0006, "step": 42356 }, { "epoch": 7.77, "learning_rate": 4.4209340013757765e-08, "loss": 0.0006, "step": 42357 }, { "epoch": 7.77, "learning_rate": 4.413963306102109e-08, "loss": 0.0003, "step": 42358 }, { "epoch": 7.77, "learning_rate": 4.406998098522186e-08, "loss": 0.0005, "step": 42359 }, { "epoch": 7.77, "learning_rate": 4.4000383786740874e-08, "loss": 0.0002, "step": 42360 }, { "epoch": 7.77, "learning_rate": 4.393084146596338e-08, "loss": 0.0003, "step": 42361 }, { "epoch": 7.77, "learning_rate": 4.386135402327241e-08, "loss": 0.0004, "step": 42362 }, { "epoch": 7.77, "learning_rate": 4.379192145904987e-08, "loss": 0.0004, "step": 42363 }, { "epoch": 7.77, "learning_rate": 4.3722543773681016e-08, "loss": 0.0002, "step": 42364 }, { "epoch": 7.77, "learning_rate": 4.365322096754554e-08, "loss": 0.0002, "step": 42365 }, { "epoch": 7.77, "learning_rate": 4.358395304102758e-08, "loss": 0.0004, "step": 42366 }, { "epoch": 7.77, "learning_rate": 4.351473999450795e-08, "loss": 0.0001, "step": 42367 }, { "epoch": 7.77, "learning_rate": 4.3445581828368556e-08, "loss": 0.001, "step": 42368 }, { "epoch": 7.77, "learning_rate": 4.33764785429891e-08, "loss": 0.0003, "step": 42369 }, { "epoch": 7.77, "learning_rate": 4.3307430138753715e-08, "loss": 0.0004, "step": 42370 }, { "epoch": 7.77, "learning_rate": 4.323843661603988e-08, "loss": 0.0001, "step": 42371 }, { "epoch": 7.77, "learning_rate": 4.3169497975229515e-08, "loss": 0.0004, "step": 42372 }, { "epoch": 7.77, "learning_rate": 4.3100614216701196e-08, "loss": 0.0003, "step": 42373 }, { "epoch": 7.77, "learning_rate": 4.303178534083574e-08, "loss": 0.0004, "step": 42374 }, { "epoch": 7.77, "learning_rate": 4.296301134801284e-08, "loss": 0.0004, "step": 42375 }, { "epoch": 7.77, "learning_rate": 4.289429223861108e-08, "loss": 0.0003, "step": 42376 }, { "epoch": 7.77, "learning_rate": 4.282562801301016e-08, "loss": 0.0009, "step": 42377 }, { "epoch": 7.77, "learning_rate": 4.275701867158643e-08, "loss": 0.0002, "step": 42378 }, { "epoch": 7.77, "learning_rate": 4.268846421471962e-08, "loss": 0.0005, "step": 42379 }, { "epoch": 7.77, "learning_rate": 4.2619964642787174e-08, "loss": 0.0004, "step": 42380 }, { "epoch": 7.77, "learning_rate": 4.2551519956166576e-08, "loss": 0.0006, "step": 42381 }, { "epoch": 7.77, "learning_rate": 4.248313015523531e-08, "loss": 0.0005, "step": 42382 }, { "epoch": 7.77, "learning_rate": 4.241479524037084e-08, "loss": 0.0015, "step": 42383 }, { "epoch": 7.77, "learning_rate": 4.234651521194844e-08, "loss": 0.0006, "step": 42384 }, { "epoch": 7.77, "learning_rate": 4.227829007034667e-08, "loss": 0.0008, "step": 42385 }, { "epoch": 7.77, "learning_rate": 4.2210119815938586e-08, "loss": 0.0002, "step": 42386 }, { "epoch": 7.77, "learning_rate": 4.214200444910277e-08, "loss": 0.0002, "step": 42387 }, { "epoch": 7.77, "learning_rate": 4.207394397021336e-08, "loss": 0.0007, "step": 42388 }, { "epoch": 7.77, "learning_rate": 4.200593837964562e-08, "loss": 0.0004, "step": 42389 }, { "epoch": 7.77, "learning_rate": 4.1937987677774794e-08, "loss": 0.0003, "step": 42390 }, { "epoch": 7.77, "learning_rate": 4.187009186497393e-08, "loss": 0.0005, "step": 42391 }, { "epoch": 7.77, "learning_rate": 4.180225094161938e-08, "loss": 0.0004, "step": 42392 }, { "epoch": 7.77, "learning_rate": 4.173446490808308e-08, "loss": 0.0001, "step": 42393 }, { "epoch": 7.77, "learning_rate": 4.1666733764740285e-08, "loss": 0.0005, "step": 42394 }, { "epoch": 7.77, "learning_rate": 4.159905751196403e-08, "loss": 0.0004, "step": 42395 }, { "epoch": 7.77, "learning_rate": 4.1531436150127335e-08, "loss": 0.0006, "step": 42396 }, { "epoch": 7.77, "learning_rate": 4.146386967960214e-08, "loss": 0.0004, "step": 42397 }, { "epoch": 7.78, "learning_rate": 4.139635810076148e-08, "loss": 0.0003, "step": 42398 }, { "epoch": 7.78, "learning_rate": 4.132890141397727e-08, "loss": 0.0003, "step": 42399 }, { "epoch": 7.78, "learning_rate": 4.126149961962256e-08, "loss": 0.0005, "step": 42400 }, { "epoch": 7.78, "learning_rate": 4.119415271806704e-08, "loss": 0.0002, "step": 42401 }, { "epoch": 7.78, "learning_rate": 4.112686070968153e-08, "loss": 0.0005, "step": 42402 }, { "epoch": 7.78, "learning_rate": 4.105962359483906e-08, "loss": 0.0002, "step": 42403 }, { "epoch": 7.78, "learning_rate": 4.099244137390934e-08, "loss": 0.0014, "step": 42404 }, { "epoch": 7.78, "learning_rate": 4.0925314047262074e-08, "loss": 0.0001, "step": 42405 }, { "epoch": 7.78, "learning_rate": 4.085824161526919e-08, "loss": 0.0002, "step": 42406 }, { "epoch": 7.78, "learning_rate": 4.079122407829817e-08, "loss": 0.0002, "step": 42407 }, { "epoch": 7.78, "learning_rate": 4.07242614367187e-08, "loss": 0.0003, "step": 42408 }, { "epoch": 7.78, "learning_rate": 4.065735369090052e-08, "loss": 0.0002, "step": 42409 }, { "epoch": 7.78, "learning_rate": 4.0590500841212185e-08, "loss": 0.0002, "step": 42410 }, { "epoch": 7.78, "learning_rate": 4.052370288802343e-08, "loss": 0.002, "step": 42411 }, { "epoch": 7.78, "learning_rate": 4.045695983170062e-08, "loss": 0.0001, "step": 42412 }, { "epoch": 7.78, "learning_rate": 4.0390271672612336e-08, "loss": 0.0003, "step": 42413 }, { "epoch": 7.78, "learning_rate": 4.032363841112608e-08, "loss": 0.0003, "step": 42414 }, { "epoch": 7.78, "learning_rate": 4.025706004760932e-08, "loss": 0.0003, "step": 42415 }, { "epoch": 7.78, "learning_rate": 4.019053658242955e-08, "loss": 0.0004, "step": 42416 }, { "epoch": 7.78, "learning_rate": 4.012406801595314e-08, "loss": 0.0017, "step": 42417 }, { "epoch": 7.78, "learning_rate": 4.005765434854536e-08, "loss": 0.0002, "step": 42418 }, { "epoch": 7.78, "learning_rate": 3.999129558057258e-08, "loss": 0.0003, "step": 42419 }, { "epoch": 7.78, "learning_rate": 3.992499171240227e-08, "loss": 0.0003, "step": 42420 }, { "epoch": 7.78, "learning_rate": 3.9858742744398604e-08, "loss": 0.0003, "step": 42421 }, { "epoch": 7.78, "learning_rate": 3.979254867692572e-08, "loss": 0.0002, "step": 42422 }, { "epoch": 7.78, "learning_rate": 3.972640951035112e-08, "loss": 0.0004, "step": 42423 }, { "epoch": 7.78, "learning_rate": 3.9660325245037824e-08, "loss": 0.0004, "step": 42424 }, { "epoch": 7.78, "learning_rate": 3.959429588134889e-08, "loss": 0.0006, "step": 42425 }, { "epoch": 7.78, "learning_rate": 3.952832141965068e-08, "loss": 0.0004, "step": 42426 }, { "epoch": 7.78, "learning_rate": 3.9462401860305145e-08, "loss": 0.0005, "step": 42427 }, { "epoch": 7.78, "learning_rate": 3.9396537203676424e-08, "loss": 0.0003, "step": 42428 }, { "epoch": 7.78, "learning_rate": 3.933072745012756e-08, "loss": 0.0003, "step": 42429 }, { "epoch": 7.78, "learning_rate": 3.926497260002049e-08, "loss": 0.0005, "step": 42430 }, { "epoch": 7.78, "learning_rate": 3.919927265371826e-08, "loss": 0.0011, "step": 42431 }, { "epoch": 7.78, "learning_rate": 3.91336276115839e-08, "loss": 0.0006, "step": 42432 }, { "epoch": 7.78, "learning_rate": 3.9068037473977136e-08, "loss": 0.0003, "step": 42433 }, { "epoch": 7.78, "learning_rate": 3.900250224126212e-08, "loss": 0.0003, "step": 42434 }, { "epoch": 7.78, "learning_rate": 3.8937021913797444e-08, "loss": 0.0006, "step": 42435 }, { "epoch": 7.78, "learning_rate": 3.8871596491946164e-08, "loss": 0.0004, "step": 42436 }, { "epoch": 7.78, "learning_rate": 3.880622597606687e-08, "loss": 0.0006, "step": 42437 }, { "epoch": 7.78, "learning_rate": 3.87409103665215e-08, "loss": 0.0004, "step": 42438 }, { "epoch": 7.78, "learning_rate": 3.867564966366977e-08, "loss": 0.0004, "step": 42439 }, { "epoch": 7.78, "learning_rate": 3.8610443867871386e-08, "loss": 0.0004, "step": 42440 }, { "epoch": 7.78, "learning_rate": 3.854529297948606e-08, "loss": 0.0005, "step": 42441 }, { "epoch": 7.78, "learning_rate": 3.8480196998871287e-08, "loss": 0.0005, "step": 42442 }, { "epoch": 7.78, "learning_rate": 3.8415155926387894e-08, "loss": 0.0004, "step": 42443 }, { "epoch": 7.78, "learning_rate": 3.8350169762393366e-08, "loss": 0.0004, "step": 42444 }, { "epoch": 7.78, "learning_rate": 3.82852385072463e-08, "loss": 0.0005, "step": 42445 }, { "epoch": 7.78, "learning_rate": 3.8220362161303095e-08, "loss": 0.0002, "step": 42446 }, { "epoch": 7.78, "learning_rate": 3.815554072492456e-08, "loss": 0.0003, "step": 42447 }, { "epoch": 7.78, "learning_rate": 3.809077419846485e-08, "loss": 0.0012, "step": 42448 }, { "epoch": 7.78, "learning_rate": 3.802606258228259e-08, "loss": 0.0007, "step": 42449 }, { "epoch": 7.78, "learning_rate": 3.7961405876734136e-08, "loss": 0.0009, "step": 42450 }, { "epoch": 7.78, "learning_rate": 3.789680408217478e-08, "loss": 0.0004, "step": 42451 }, { "epoch": 7.79, "learning_rate": 3.7832257198963106e-08, "loss": 0.0003, "step": 42452 }, { "epoch": 7.79, "learning_rate": 3.776776522745218e-08, "loss": 0.0005, "step": 42453 }, { "epoch": 7.79, "learning_rate": 3.7703328167999485e-08, "loss": 0.0007, "step": 42454 }, { "epoch": 7.79, "learning_rate": 3.763894602095919e-08, "loss": 0.0002, "step": 42455 }, { "epoch": 7.79, "learning_rate": 3.7574618786686556e-08, "loss": 0.0008, "step": 42456 }, { "epoch": 7.79, "learning_rate": 3.7510346465535754e-08, "loss": 0.0004, "step": 42457 }, { "epoch": 7.79, "learning_rate": 3.7446129057860934e-08, "loss": 0.0002, "step": 42458 }, { "epoch": 7.79, "learning_rate": 3.738196656401738e-08, "loss": 0.0005, "step": 42459 }, { "epoch": 7.79, "learning_rate": 3.7317858984358135e-08, "loss": 0.0003, "step": 42460 }, { "epoch": 7.79, "learning_rate": 3.725380631923514e-08, "loss": 0.0004, "step": 42461 }, { "epoch": 7.79, "learning_rate": 3.7189808569002564e-08, "loss": 0.0004, "step": 42462 }, { "epoch": 7.79, "learning_rate": 3.7125865734013445e-08, "loss": 0.0003, "step": 42463 }, { "epoch": 7.79, "learning_rate": 3.706197781462084e-08, "loss": 0.0011, "step": 42464 }, { "epoch": 7.79, "learning_rate": 3.6998144811174474e-08, "loss": 0.0004, "step": 42465 }, { "epoch": 7.79, "learning_rate": 3.6934366724028506e-08, "loss": 0.0003, "step": 42466 }, { "epoch": 7.79, "learning_rate": 3.6870643553533756e-08, "loss": 0.0003, "step": 42467 }, { "epoch": 7.79, "learning_rate": 3.680697530004218e-08, "loss": 0.0007, "step": 42468 }, { "epoch": 7.79, "learning_rate": 3.674336196390238e-08, "loss": 0.0002, "step": 42469 }, { "epoch": 7.79, "learning_rate": 3.667980354546741e-08, "loss": 0.0007, "step": 42470 }, { "epoch": 7.79, "learning_rate": 3.6616300045086986e-08, "loss": 0.0003, "step": 42471 }, { "epoch": 7.79, "learning_rate": 3.655285146311083e-08, "loss": 0.0005, "step": 42472 }, { "epoch": 7.79, "learning_rate": 3.648945779988866e-08, "loss": 0.0008, "step": 42473 }, { "epoch": 7.79, "learning_rate": 3.6426119055770204e-08, "loss": 0.0003, "step": 42474 }, { "epoch": 7.79, "learning_rate": 3.636283523110407e-08, "loss": 0.0007, "step": 42475 }, { "epoch": 7.79, "learning_rate": 3.629960632623997e-08, "loss": 0.0004, "step": 42476 }, { "epoch": 7.79, "learning_rate": 3.6236432341525405e-08, "loss": 0.0002, "step": 42477 }, { "epoch": 7.79, "learning_rate": 3.6173313277309e-08, "loss": 0.0002, "step": 42478 }, { "epoch": 7.79, "learning_rate": 3.6110249133938237e-08, "loss": 0.0004, "step": 42479 }, { "epoch": 7.79, "learning_rate": 3.6047239911762846e-08, "loss": 0.0003, "step": 42480 }, { "epoch": 7.79, "learning_rate": 3.598428561112699e-08, "loss": 0.0004, "step": 42481 }, { "epoch": 7.79, "learning_rate": 3.5921386232379285e-08, "loss": 0.0003, "step": 42482 }, { "epoch": 7.79, "learning_rate": 3.585854177586723e-08, "loss": 0.0005, "step": 42483 }, { "epoch": 7.79, "learning_rate": 3.57957522419361e-08, "loss": 0.0006, "step": 42484 }, { "epoch": 7.79, "learning_rate": 3.573301763093118e-08, "loss": 0.0002, "step": 42485 }, { "epoch": 7.79, "learning_rate": 3.567033794319996e-08, "loss": 0.0002, "step": 42486 }, { "epoch": 7.79, "learning_rate": 3.5607713179087734e-08, "loss": 0.0003, "step": 42487 }, { "epoch": 7.79, "learning_rate": 3.554514333893866e-08, "loss": 0.0002, "step": 42488 }, { "epoch": 7.79, "learning_rate": 3.548262842309913e-08, "loss": 0.0003, "step": 42489 }, { "epoch": 7.79, "learning_rate": 3.542016843191221e-08, "loss": 0.0004, "step": 42490 }, { "epoch": 7.79, "learning_rate": 3.535776336572205e-08, "loss": 0.0014, "step": 42491 }, { "epoch": 7.79, "learning_rate": 3.5295413224873955e-08, "loss": 0.0004, "step": 42492 }, { "epoch": 7.79, "learning_rate": 3.523311800971097e-08, "loss": 0.0002, "step": 42493 }, { "epoch": 7.79, "learning_rate": 3.5170877720576144e-08, "loss": 0.0009, "step": 42494 }, { "epoch": 7.79, "learning_rate": 3.5108692357813664e-08, "loss": 0.0003, "step": 42495 }, { "epoch": 7.79, "learning_rate": 3.504656192176548e-08, "loss": 0.0003, "step": 42496 }, { "epoch": 7.79, "learning_rate": 3.498448641277352e-08, "loss": 0.0005, "step": 42497 }, { "epoch": 7.79, "learning_rate": 3.492246583117975e-08, "loss": 0.0003, "step": 42498 }, { "epoch": 7.79, "learning_rate": 3.4860500177328336e-08, "loss": 0.0003, "step": 42499 }, { "epoch": 7.79, "learning_rate": 3.4798589451559006e-08, "loss": 0.0004, "step": 42500 }, { "epoch": 7.79, "learning_rate": 3.473673365421259e-08, "loss": 0.0006, "step": 42501 }, { "epoch": 7.79, "learning_rate": 3.4674932785631055e-08, "loss": 0.0002, "step": 42502 }, { "epoch": 7.79, "learning_rate": 3.461318684615522e-08, "loss": 0.0002, "step": 42503 }, { "epoch": 7.79, "learning_rate": 3.455149583612483e-08, "loss": 0.0004, "step": 42504 }, { "epoch": 7.79, "learning_rate": 3.4489859755879594e-08, "loss": 0.0002, "step": 42505 }, { "epoch": 7.79, "learning_rate": 3.442827860576037e-08, "loss": 0.0004, "step": 42506 }, { "epoch": 7.8, "learning_rate": 3.436675238610576e-08, "loss": 0.0004, "step": 42507 }, { "epoch": 7.8, "learning_rate": 3.430528109725439e-08, "loss": 0.0002, "step": 42508 }, { "epoch": 7.8, "learning_rate": 3.424386473954711e-08, "loss": 0.0002, "step": 42509 }, { "epoch": 7.8, "learning_rate": 3.418250331331918e-08, "loss": 0.0003, "step": 42510 }, { "epoch": 7.8, "learning_rate": 3.412119681891257e-08, "loss": 0.0006, "step": 42511 }, { "epoch": 7.8, "learning_rate": 3.4059945256662564e-08, "loss": 0.0005, "step": 42512 }, { "epoch": 7.8, "learning_rate": 3.3998748626907777e-08, "loss": 0.0004, "step": 42513 }, { "epoch": 7.8, "learning_rate": 3.393760692998571e-08, "loss": 0.0004, "step": 42514 }, { "epoch": 7.8, "learning_rate": 3.3876520166231664e-08, "loss": 0.0002, "step": 42515 }, { "epoch": 7.8, "learning_rate": 3.381548833598536e-08, "loss": 0.0003, "step": 42516 }, { "epoch": 7.8, "learning_rate": 3.375451143958097e-08, "loss": 0.0004, "step": 42517 }, { "epoch": 7.8, "learning_rate": 3.36935894773549e-08, "loss": 0.0005, "step": 42518 }, { "epoch": 7.8, "learning_rate": 3.3632722449643554e-08, "loss": 0.0005, "step": 42519 }, { "epoch": 7.8, "learning_rate": 3.35719103567822e-08, "loss": 0.0005, "step": 42520 }, { "epoch": 7.8, "learning_rate": 3.351115319910614e-08, "loss": 0.0005, "step": 42521 }, { "epoch": 7.8, "learning_rate": 3.345045097694955e-08, "loss": 0.0018, "step": 42522 }, { "epoch": 7.8, "learning_rate": 3.3389803690647706e-08, "loss": 0.0003, "step": 42523 }, { "epoch": 7.8, "learning_rate": 3.33292113405348e-08, "loss": 0.0004, "step": 42524 }, { "epoch": 7.8, "learning_rate": 3.3268673926945e-08, "loss": 0.0012, "step": 42525 }, { "epoch": 7.8, "learning_rate": 3.320819145021137e-08, "loss": 0.0004, "step": 42526 }, { "epoch": 7.8, "learning_rate": 3.31477639106681e-08, "loss": 0.001, "step": 42527 }, { "epoch": 7.8, "learning_rate": 3.3087391308648244e-08, "loss": 0.0007, "step": 42528 }, { "epoch": 7.8, "learning_rate": 3.3027073644482655e-08, "loss": 0.001, "step": 42529 }, { "epoch": 7.8, "learning_rate": 3.296681091850662e-08, "loss": 0.0004, "step": 42530 }, { "epoch": 7.8, "learning_rate": 3.2906603131050986e-08, "loss": 0.0006, "step": 42531 }, { "epoch": 7.8, "learning_rate": 3.284645028244771e-08, "loss": 0.0004, "step": 42532 }, { "epoch": 7.8, "learning_rate": 3.278635237302763e-08, "loss": 0.0007, "step": 42533 }, { "epoch": 7.8, "learning_rate": 3.2726309403122713e-08, "loss": 0.001, "step": 42534 }, { "epoch": 7.8, "learning_rate": 3.266632137306491e-08, "loss": 0.0007, "step": 42535 }, { "epoch": 7.8, "learning_rate": 3.260638828318285e-08, "loss": 0.0006, "step": 42536 }, { "epoch": 7.8, "learning_rate": 3.25465101338085e-08, "loss": 0.0005, "step": 42537 }, { "epoch": 7.8, "learning_rate": 3.248668692527046e-08, "loss": 0.0005, "step": 42538 }, { "epoch": 7.8, "learning_rate": 3.242691865790071e-08, "loss": 0.0003, "step": 42539 }, { "epoch": 7.8, "learning_rate": 3.236720533202564e-08, "loss": 0.0005, "step": 42540 }, { "epoch": 7.8, "learning_rate": 3.230754694797722e-08, "loss": 0.0005, "step": 42541 }, { "epoch": 7.8, "learning_rate": 3.224794350608296e-08, "loss": 0.0004, "step": 42542 }, { "epoch": 7.8, "learning_rate": 3.218839500667148e-08, "loss": 0.0003, "step": 42543 }, { "epoch": 7.8, "learning_rate": 3.21289014500703e-08, "loss": 0.0002, "step": 42544 }, { "epoch": 7.8, "learning_rate": 3.2069462836609164e-08, "loss": 0.0008, "step": 42545 }, { "epoch": 7.8, "learning_rate": 3.201007916661447e-08, "loss": 0.0004, "step": 42546 }, { "epoch": 7.8, "learning_rate": 3.1950750440413734e-08, "loss": 0.0002, "step": 42547 }, { "epoch": 7.8, "learning_rate": 3.1891476658333366e-08, "loss": 0.0002, "step": 42548 }, { "epoch": 7.8, "learning_rate": 3.1832257820701984e-08, "loss": 0.0003, "step": 42549 }, { "epoch": 7.8, "learning_rate": 3.1773093927842666e-08, "loss": 0.0002, "step": 42550 }, { "epoch": 7.8, "learning_rate": 3.171398498008516e-08, "loss": 0.0005, "step": 42551 }, { "epoch": 7.8, "learning_rate": 3.165493097775363e-08, "loss": 0.0003, "step": 42552 }, { "epoch": 7.8, "learning_rate": 3.1595931921173385e-08, "loss": 0.0003, "step": 42553 }, { "epoch": 7.8, "learning_rate": 3.1536987810669716e-08, "loss": 0.0003, "step": 42554 }, { "epoch": 7.8, "learning_rate": 3.147809864656792e-08, "loss": 0.0001, "step": 42555 }, { "epoch": 7.8, "learning_rate": 3.1419264429192186e-08, "loss": 0.0002, "step": 42556 }, { "epoch": 7.8, "learning_rate": 3.13604851588678e-08, "loss": 0.0004, "step": 42557 }, { "epoch": 7.8, "learning_rate": 3.130176083591674e-08, "loss": 0.0003, "step": 42558 }, { "epoch": 7.8, "learning_rate": 3.1243091460664285e-08, "loss": 0.0002, "step": 42559 }, { "epoch": 7.8, "learning_rate": 3.1184477033434633e-08, "loss": 0.0005, "step": 42560 }, { "epoch": 7.81, "learning_rate": 3.112591755454863e-08, "loss": 0.0002, "step": 42561 }, { "epoch": 7.81, "learning_rate": 3.106741302433047e-08, "loss": 0.0006, "step": 42562 }, { "epoch": 7.81, "learning_rate": 3.1008963443101e-08, "loss": 0.0002, "step": 42563 }, { "epoch": 7.81, "learning_rate": 3.095056881118552e-08, "loss": 0.0005, "step": 42564 }, { "epoch": 7.81, "learning_rate": 3.0892229128902665e-08, "loss": 0.0006, "step": 42565 }, { "epoch": 7.81, "learning_rate": 3.083394439657661e-08, "loss": 0.0006, "step": 42566 }, { "epoch": 7.81, "learning_rate": 3.0775714614525995e-08, "loss": 0.0003, "step": 42567 }, { "epoch": 7.81, "learning_rate": 3.071753978307501e-08, "loss": 0.0005, "step": 42568 }, { "epoch": 7.81, "learning_rate": 3.0659419902541175e-08, "loss": 0.0004, "step": 42569 }, { "epoch": 7.81, "learning_rate": 3.060135497324645e-08, "loss": 0.0003, "step": 42570 }, { "epoch": 7.81, "learning_rate": 3.054334499551059e-08, "loss": 0.0008, "step": 42571 }, { "epoch": 7.81, "learning_rate": 3.0485389969653334e-08, "loss": 0.0004, "step": 42572 }, { "epoch": 7.81, "learning_rate": 3.042748989599442e-08, "loss": 0.0004, "step": 42573 }, { "epoch": 7.81, "learning_rate": 3.036964477485249e-08, "loss": 0.0004, "step": 42574 }, { "epoch": 7.81, "learning_rate": 3.031185460654618e-08, "loss": 0.0002, "step": 42575 }, { "epoch": 7.81, "learning_rate": 3.025411939139522e-08, "loss": 0.0005, "step": 42576 }, { "epoch": 7.81, "learning_rate": 3.0196439129717146e-08, "loss": 0.0002, "step": 42577 }, { "epoch": 7.81, "learning_rate": 3.013881382182948e-08, "loss": 0.0004, "step": 42578 }, { "epoch": 7.81, "learning_rate": 3.0081243468050857e-08, "loss": 0.0002, "step": 42579 }, { "epoch": 7.81, "learning_rate": 3.002372806869769e-08, "loss": 0.0004, "step": 42580 }, { "epoch": 7.81, "learning_rate": 2.9966267624086385e-08, "loss": 0.0002, "step": 42581 }, { "epoch": 7.81, "learning_rate": 2.990886213453559e-08, "loss": 0.0002, "step": 42582 }, { "epoch": 7.81, "learning_rate": 2.9851511600360596e-08, "loss": 0.0003, "step": 42583 }, { "epoch": 7.81, "learning_rate": 2.979421602187782e-08, "loss": 0.0002, "step": 42584 }, { "epoch": 7.81, "learning_rate": 2.973697539940257e-08, "loss": 0.0005, "step": 42585 }, { "epoch": 7.81, "learning_rate": 2.9679789733250142e-08, "loss": 0.0008, "step": 42586 }, { "epoch": 7.81, "learning_rate": 2.9622659023736954e-08, "loss": 0.0011, "step": 42587 }, { "epoch": 7.81, "learning_rate": 2.9565583271177202e-08, "loss": 0.0005, "step": 42588 }, { "epoch": 7.81, "learning_rate": 2.950856247588618e-08, "loss": 0.0002, "step": 42589 }, { "epoch": 7.81, "learning_rate": 2.945159663817698e-08, "loss": 0.0003, "step": 42590 }, { "epoch": 7.81, "learning_rate": 2.9394685758363793e-08, "loss": 0.0008, "step": 42591 }, { "epoch": 7.81, "learning_rate": 2.9337829836760813e-08, "loss": 0.0005, "step": 42592 }, { "epoch": 7.81, "learning_rate": 2.9281028873682227e-08, "loss": 0.0003, "step": 42593 }, { "epoch": 7.81, "learning_rate": 2.9224282869438902e-08, "loss": 0.0002, "step": 42594 }, { "epoch": 7.81, "learning_rate": 2.9167591824345033e-08, "loss": 0.0005, "step": 42595 }, { "epoch": 7.81, "learning_rate": 2.9110955738713696e-08, "loss": 0.0003, "step": 42596 }, { "epoch": 7.81, "learning_rate": 2.9054374612856872e-08, "loss": 0.0003, "step": 42597 }, { "epoch": 7.81, "learning_rate": 2.899784844708431e-08, "loss": 0.0007, "step": 42598 }, { "epoch": 7.81, "learning_rate": 2.89413772417102e-08, "loss": 0.0007, "step": 42599 }, { "epoch": 7.81, "learning_rate": 2.88849609970443e-08, "loss": 0.0005, "step": 42600 }, { "epoch": 7.81, "learning_rate": 2.8828599713398575e-08, "loss": 0.0006, "step": 42601 }, { "epoch": 7.81, "learning_rate": 2.8772293391083895e-08, "loss": 0.0003, "step": 42602 }, { "epoch": 7.81, "learning_rate": 2.8716042030408897e-08, "loss": 0.0002, "step": 42603 }, { "epoch": 7.81, "learning_rate": 2.8659845631684446e-08, "loss": 0.0007, "step": 42604 }, { "epoch": 7.81, "learning_rate": 2.8603704195221405e-08, "loss": 0.0004, "step": 42605 }, { "epoch": 7.81, "learning_rate": 2.8547617721327304e-08, "loss": 0.0002, "step": 42606 }, { "epoch": 7.81, "learning_rate": 2.8491586210313005e-08, "loss": 0.0006, "step": 42607 }, { "epoch": 7.81, "learning_rate": 2.843560966248715e-08, "loss": 0.0006, "step": 42608 }, { "epoch": 7.81, "learning_rate": 2.8379688078157275e-08, "loss": 0.0009, "step": 42609 }, { "epoch": 7.81, "learning_rate": 2.8323821457632018e-08, "loss": 0.001, "step": 42610 }, { "epoch": 7.81, "learning_rate": 2.826800980122002e-08, "loss": 0.0003, "step": 42611 }, { "epoch": 7.81, "learning_rate": 2.8212253109227704e-08, "loss": 0.0008, "step": 42612 }, { "epoch": 7.81, "learning_rate": 2.8156551381962603e-08, "loss": 0.0004, "step": 42613 }, { "epoch": 7.81, "learning_rate": 2.810090461973336e-08, "loss": 0.0006, "step": 42614 }, { "epoch": 7.81, "learning_rate": 2.8045312822845284e-08, "loss": 0.0004, "step": 42615 }, { "epoch": 7.82, "learning_rate": 2.7989775991604794e-08, "loss": 0.0003, "step": 42616 }, { "epoch": 7.82, "learning_rate": 2.7934294126318316e-08, "loss": 0.0002, "step": 42617 }, { "epoch": 7.82, "learning_rate": 2.7878867227292273e-08, "loss": 0.0009, "step": 42618 }, { "epoch": 7.82, "learning_rate": 2.782349529483086e-08, "loss": 0.0005, "step": 42619 }, { "epoch": 7.82, "learning_rate": 2.7768178329240504e-08, "loss": 0.0003, "step": 42620 }, { "epoch": 7.82, "learning_rate": 2.771291633082429e-08, "loss": 0.0003, "step": 42621 }, { "epoch": 7.82, "learning_rate": 2.7657709299889757e-08, "loss": 0.0004, "step": 42622 }, { "epoch": 7.82, "learning_rate": 2.760255723673888e-08, "loss": 0.0004, "step": 42623 }, { "epoch": 7.82, "learning_rate": 2.7547460141675865e-08, "loss": 0.0003, "step": 42624 }, { "epoch": 7.82, "learning_rate": 2.7492418015004907e-08, "loss": 0.0003, "step": 42625 }, { "epoch": 7.82, "learning_rate": 2.7437430857030213e-08, "loss": 0.0003, "step": 42626 }, { "epoch": 7.82, "learning_rate": 2.7382498668053757e-08, "loss": 0.0005, "step": 42627 }, { "epoch": 7.82, "learning_rate": 2.7327621448378638e-08, "loss": 0.0005, "step": 42628 }, { "epoch": 7.82, "learning_rate": 2.7272799198306832e-08, "loss": 0.0005, "step": 42629 }, { "epoch": 7.82, "learning_rate": 2.721803191814032e-08, "loss": 0.0007, "step": 42630 }, { "epoch": 7.82, "learning_rate": 2.7163319608183304e-08, "loss": 0.0004, "step": 42631 }, { "epoch": 7.82, "learning_rate": 2.7108662268735542e-08, "loss": 0.0002, "step": 42632 }, { "epoch": 7.82, "learning_rate": 2.7054059900097908e-08, "loss": 0.0002, "step": 42633 }, { "epoch": 7.82, "learning_rate": 2.6999512502571266e-08, "loss": 0.0003, "step": 42634 }, { "epoch": 7.82, "learning_rate": 2.6945020076457605e-08, "loss": 0.0003, "step": 42635 }, { "epoch": 7.82, "learning_rate": 2.6890582622056683e-08, "loss": 0.0004, "step": 42636 }, { "epoch": 7.82, "learning_rate": 2.6836200139668256e-08, "loss": 0.0009, "step": 42637 }, { "epoch": 7.82, "learning_rate": 2.6781872629593197e-08, "loss": 0.0005, "step": 42638 }, { "epoch": 7.82, "learning_rate": 2.6727600092129048e-08, "loss": 0.0003, "step": 42639 }, { "epoch": 7.82, "learning_rate": 2.6673382527576674e-08, "loss": 0.0002, "step": 42640 }, { "epoch": 7.82, "learning_rate": 2.661921993623362e-08, "loss": 0.0002, "step": 42641 }, { "epoch": 7.82, "learning_rate": 2.6565112318399644e-08, "loss": 0.0002, "step": 42642 }, { "epoch": 7.82, "learning_rate": 2.6511059674372286e-08, "loss": 0.0007, "step": 42643 }, { "epoch": 7.82, "learning_rate": 2.6457062004449087e-08, "loss": 0.0004, "step": 42644 }, { "epoch": 7.82, "learning_rate": 2.640311930892869e-08, "loss": 0.0003, "step": 42645 }, { "epoch": 7.82, "learning_rate": 2.6349231588108648e-08, "loss": 0.0003, "step": 42646 }, { "epoch": 7.82, "learning_rate": 2.629539884228427e-08, "loss": 0.0007, "step": 42647 }, { "epoch": 7.82, "learning_rate": 2.62416210717531e-08, "loss": 0.0004, "step": 42648 }, { "epoch": 7.82, "learning_rate": 2.6187898276813785e-08, "loss": 0.0003, "step": 42649 }, { "epoch": 7.82, "learning_rate": 2.6134230457758315e-08, "loss": 0.0006, "step": 42650 }, { "epoch": 7.82, "learning_rate": 2.608061761488534e-08, "loss": 0.0004, "step": 42651 }, { "epoch": 7.82, "learning_rate": 2.602705974848907e-08, "loss": 0.0003, "step": 42652 }, { "epoch": 7.82, "learning_rate": 2.5973556858865934e-08, "loss": 0.0002, "step": 42653 }, { "epoch": 7.82, "learning_rate": 2.592010894631014e-08, "loss": 0.0004, "step": 42654 }, { "epoch": 7.82, "learning_rate": 2.586671601111701e-08, "loss": 0.0005, "step": 42655 }, { "epoch": 7.82, "learning_rate": 2.581337805357964e-08, "loss": 0.0003, "step": 42656 }, { "epoch": 7.82, "learning_rate": 2.5760095073992243e-08, "loss": 0.0005, "step": 42657 }, { "epoch": 7.82, "learning_rate": 2.5706867072649024e-08, "loss": 0.0004, "step": 42658 }, { "epoch": 7.82, "learning_rate": 2.565369404984419e-08, "loss": 0.0003, "step": 42659 }, { "epoch": 7.82, "learning_rate": 2.5600576005868627e-08, "loss": 0.0003, "step": 42660 }, { "epoch": 7.82, "learning_rate": 2.5547512941016538e-08, "loss": 0.0004, "step": 42661 }, { "epoch": 7.82, "learning_rate": 2.5494504855581027e-08, "loss": 0.0007, "step": 42662 }, { "epoch": 7.82, "learning_rate": 2.5441551749852966e-08, "loss": 0.0006, "step": 42663 }, { "epoch": 7.82, "learning_rate": 2.5388653624124347e-08, "loss": 0.0005, "step": 42664 }, { "epoch": 7.82, "learning_rate": 2.5335810478687162e-08, "loss": 0.0005, "step": 42665 }, { "epoch": 7.82, "learning_rate": 2.5283022313833396e-08, "loss": 0.0002, "step": 42666 }, { "epoch": 7.82, "learning_rate": 2.5230289129853924e-08, "loss": 0.0006, "step": 42667 }, { "epoch": 7.82, "learning_rate": 2.517761092703741e-08, "loss": 0.0005, "step": 42668 }, { "epoch": 7.82, "learning_rate": 2.512498770567584e-08, "loss": 0.0003, "step": 42669 }, { "epoch": 7.83, "learning_rate": 2.5072419466058984e-08, "loss": 0.0009, "step": 42670 }, { "epoch": 7.83, "learning_rate": 2.5019906208476608e-08, "loss": 0.0002, "step": 42671 }, { "epoch": 7.83, "learning_rate": 2.496744793321848e-08, "loss": 0.0006, "step": 42672 }, { "epoch": 7.83, "learning_rate": 2.491504464057326e-08, "loss": 0.001, "step": 42673 }, { "epoch": 7.83, "learning_rate": 2.4862696330830715e-08, "loss": 0.0007, "step": 42674 }, { "epoch": 7.83, "learning_rate": 2.4810403004278394e-08, "loss": 0.0006, "step": 42675 }, { "epoch": 7.83, "learning_rate": 2.4758164661203843e-08, "loss": 0.0005, "step": 42676 }, { "epoch": 7.83, "learning_rate": 2.470598130189683e-08, "loss": 0.0003, "step": 42677 }, { "epoch": 7.83, "learning_rate": 2.4653852926644905e-08, "loss": 0.0003, "step": 42678 }, { "epoch": 7.83, "learning_rate": 2.4601779535733394e-08, "loss": 0.0012, "step": 42679 }, { "epoch": 7.83, "learning_rate": 2.4549761129450955e-08, "loss": 0.0003, "step": 42680 }, { "epoch": 7.83, "learning_rate": 2.4497797708084025e-08, "loss": 0.0003, "step": 42681 }, { "epoch": 7.83, "learning_rate": 2.4445889271917932e-08, "loss": 0.0007, "step": 42682 }, { "epoch": 7.83, "learning_rate": 2.4394035821241334e-08, "loss": 0.0004, "step": 42683 }, { "epoch": 7.83, "learning_rate": 2.434223735633734e-08, "loss": 0.0003, "step": 42684 }, { "epoch": 7.83, "learning_rate": 2.4290493877493492e-08, "loss": 0.0009, "step": 42685 }, { "epoch": 7.83, "learning_rate": 2.4238805384994014e-08, "loss": 0.0005, "step": 42686 }, { "epoch": 7.83, "learning_rate": 2.4187171879124228e-08, "loss": 0.001, "step": 42687 }, { "epoch": 7.83, "learning_rate": 2.413559336016724e-08, "loss": 0.0003, "step": 42688 }, { "epoch": 7.83, "learning_rate": 2.4084069828409497e-08, "loss": 0.0006, "step": 42689 }, { "epoch": 7.83, "learning_rate": 2.4032601284134095e-08, "loss": 0.0005, "step": 42690 }, { "epoch": 7.83, "learning_rate": 2.3981187727624143e-08, "loss": 0.0002, "step": 42691 }, { "epoch": 7.83, "learning_rate": 2.3929829159163865e-08, "loss": 0.0004, "step": 42692 }, { "epoch": 7.83, "learning_rate": 2.387852557903636e-08, "loss": 0.0002, "step": 42693 }, { "epoch": 7.83, "learning_rate": 2.3827276987524738e-08, "loss": 0.0004, "step": 42694 }, { "epoch": 7.83, "learning_rate": 2.3776083384908775e-08, "loss": 0.0009, "step": 42695 }, { "epoch": 7.83, "learning_rate": 2.3724944771474913e-08, "loss": 0.0005, "step": 42696 }, { "epoch": 7.83, "learning_rate": 2.3673861147501807e-08, "loss": 0.0003, "step": 42697 }, { "epoch": 7.83, "learning_rate": 2.3622832513272577e-08, "loss": 0.0002, "step": 42698 }, { "epoch": 7.83, "learning_rate": 2.3571858869068098e-08, "loss": 0.0002, "step": 42699 }, { "epoch": 7.83, "learning_rate": 2.3520940215168153e-08, "loss": 0.0004, "step": 42700 }, { "epoch": 7.83, "learning_rate": 2.3470076551855847e-08, "loss": 0.0004, "step": 42701 }, { "epoch": 7.83, "learning_rate": 2.3419267879409847e-08, "loss": 0.0008, "step": 42702 }, { "epoch": 7.83, "learning_rate": 2.3368514198109925e-08, "loss": 0.0002, "step": 42703 }, { "epoch": 7.83, "learning_rate": 2.3317815508236975e-08, "loss": 0.0008, "step": 42704 }, { "epoch": 7.83, "learning_rate": 2.3267171810069654e-08, "loss": 0.001, "step": 42705 }, { "epoch": 7.83, "learning_rate": 2.3216583103887745e-08, "loss": 0.0003, "step": 42706 }, { "epoch": 7.83, "learning_rate": 2.3166049389969912e-08, "loss": 0.0003, "step": 42707 }, { "epoch": 7.83, "learning_rate": 2.311557066859482e-08, "loss": 0.0005, "step": 42708 }, { "epoch": 7.83, "learning_rate": 2.306514694004003e-08, "loss": 0.0002, "step": 42709 }, { "epoch": 7.83, "learning_rate": 2.3014778204583087e-08, "loss": 0.0008, "step": 42710 }, { "epoch": 7.83, "learning_rate": 2.2964464462503778e-08, "loss": 0.0009, "step": 42711 }, { "epoch": 7.83, "learning_rate": 2.291420571407743e-08, "loss": 0.0005, "step": 42712 }, { "epoch": 7.83, "learning_rate": 2.2864001959580494e-08, "loss": 0.0004, "step": 42713 }, { "epoch": 7.83, "learning_rate": 2.2813853199292745e-08, "loss": 0.0006, "step": 42714 }, { "epoch": 7.83, "learning_rate": 2.2763759433488408e-08, "loss": 0.0002, "step": 42715 }, { "epoch": 7.83, "learning_rate": 2.2713720662443928e-08, "loss": 0.0004, "step": 42716 }, { "epoch": 7.83, "learning_rate": 2.266373688643464e-08, "loss": 0.0004, "step": 42717 }, { "epoch": 7.83, "learning_rate": 2.261380810573699e-08, "loss": 0.0004, "step": 42718 }, { "epoch": 7.83, "learning_rate": 2.2563934320625203e-08, "loss": 0.0004, "step": 42719 }, { "epoch": 7.83, "learning_rate": 2.2514115531375723e-08, "loss": 0.0005, "step": 42720 }, { "epoch": 7.83, "learning_rate": 2.246435173826167e-08, "loss": 0.0003, "step": 42721 }, { "epoch": 7.83, "learning_rate": 2.2414642941557262e-08, "loss": 0.0005, "step": 42722 }, { "epoch": 7.83, "learning_rate": 2.236498914153784e-08, "loss": 0.0002, "step": 42723 }, { "epoch": 7.83, "learning_rate": 2.2315390338476517e-08, "loss": 0.0005, "step": 42724 }, { "epoch": 7.84, "learning_rate": 2.2265846532646408e-08, "loss": 0.0008, "step": 42725 }, { "epoch": 7.84, "learning_rate": 2.2216357724320624e-08, "loss": 0.0003, "step": 42726 }, { "epoch": 7.84, "learning_rate": 2.2166923913772286e-08, "loss": 0.0003, "step": 42727 }, { "epoch": 7.84, "learning_rate": 2.2117545101273397e-08, "loss": 0.0003, "step": 42728 }, { "epoch": 7.84, "learning_rate": 2.2068221287095958e-08, "loss": 0.0003, "step": 42729 }, { "epoch": 7.84, "learning_rate": 2.201895247151198e-08, "loss": 0.0005, "step": 42730 }, { "epoch": 7.84, "learning_rate": 2.196973865479346e-08, "loss": 0.0007, "step": 42731 }, { "epoch": 7.84, "learning_rate": 2.1920579837212407e-08, "loss": 0.0003, "step": 42732 }, { "epoch": 7.84, "learning_rate": 2.1871476019038608e-08, "loss": 0.0005, "step": 42733 }, { "epoch": 7.84, "learning_rate": 2.182242720054295e-08, "loss": 0.0004, "step": 42734 }, { "epoch": 7.84, "learning_rate": 2.1773433381995224e-08, "loss": 0.0002, "step": 42735 }, { "epoch": 7.84, "learning_rate": 2.1724494563666322e-08, "loss": 0.0002, "step": 42736 }, { "epoch": 7.84, "learning_rate": 2.1675610745826026e-08, "loss": 0.0003, "step": 42737 }, { "epoch": 7.84, "learning_rate": 2.1626781928744123e-08, "loss": 0.0003, "step": 42738 }, { "epoch": 7.84, "learning_rate": 2.1578008112688177e-08, "loss": 0.0003, "step": 42739 }, { "epoch": 7.84, "learning_rate": 2.152928929792797e-08, "loss": 0.0003, "step": 42740 }, { "epoch": 7.84, "learning_rate": 2.1480625484732175e-08, "loss": 0.0004, "step": 42741 }, { "epoch": 7.84, "learning_rate": 2.1432016673369473e-08, "loss": 0.0002, "step": 42742 }, { "epoch": 7.84, "learning_rate": 2.138346286410742e-08, "loss": 0.0004, "step": 42743 }, { "epoch": 7.84, "learning_rate": 2.133496405721247e-08, "loss": 0.0005, "step": 42744 }, { "epoch": 7.84, "learning_rate": 2.1286520252954412e-08, "loss": 0.0002, "step": 42745 }, { "epoch": 7.84, "learning_rate": 2.1238131451597475e-08, "loss": 0.0006, "step": 42746 }, { "epoch": 7.84, "learning_rate": 2.118979765341034e-08, "loss": 0.0002, "step": 42747 }, { "epoch": 7.84, "learning_rate": 2.114151885865834e-08, "loss": 0.0008, "step": 42748 }, { "epoch": 7.84, "learning_rate": 2.1093295067609042e-08, "loss": 0.0002, "step": 42749 }, { "epoch": 7.84, "learning_rate": 2.1045126280526686e-08, "loss": 0.0002, "step": 42750 }, { "epoch": 7.84, "learning_rate": 2.0997012497677715e-08, "loss": 0.0003, "step": 42751 }, { "epoch": 7.84, "learning_rate": 2.094895371932637e-08, "loss": 0.0004, "step": 42752 }, { "epoch": 7.84, "learning_rate": 2.090094994573799e-08, "loss": 0.0003, "step": 42753 }, { "epoch": 7.84, "learning_rate": 2.0853001177177923e-08, "loss": 0.0002, "step": 42754 }, { "epoch": 7.84, "learning_rate": 2.0805107413910396e-08, "loss": 0.0007, "step": 42755 }, { "epoch": 7.84, "learning_rate": 2.0757268656198536e-08, "loss": 0.0003, "step": 42756 }, { "epoch": 7.84, "learning_rate": 2.0709484904305466e-08, "loss": 0.0006, "step": 42757 }, { "epoch": 7.84, "learning_rate": 2.0661756158496525e-08, "loss": 0.0003, "step": 42758 }, { "epoch": 7.84, "learning_rate": 2.0614082419033732e-08, "loss": 0.0006, "step": 42759 }, { "epoch": 7.84, "learning_rate": 2.0566463686180205e-08, "loss": 0.0006, "step": 42760 }, { "epoch": 7.84, "learning_rate": 2.0518899960196848e-08, "loss": 0.0004, "step": 42761 }, { "epoch": 7.84, "learning_rate": 2.0471391241347894e-08, "loss": 0.0004, "step": 42762 }, { "epoch": 7.84, "learning_rate": 2.0423937529894244e-08, "loss": 0.0011, "step": 42763 }, { "epoch": 7.84, "learning_rate": 2.0376538826097912e-08, "loss": 0.0009, "step": 42764 }, { "epoch": 7.84, "learning_rate": 2.0329195130219803e-08, "loss": 0.0003, "step": 42765 }, { "epoch": 7.84, "learning_rate": 2.0281906442520815e-08, "loss": 0.0001, "step": 42766 }, { "epoch": 7.84, "learning_rate": 2.0234672763261852e-08, "loss": 0.0003, "step": 42767 }, { "epoch": 7.84, "learning_rate": 2.0187494092703818e-08, "loss": 0.0003, "step": 42768 }, { "epoch": 7.84, "learning_rate": 2.0140370431105395e-08, "loss": 0.0006, "step": 42769 }, { "epoch": 7.84, "learning_rate": 2.0093301778727482e-08, "loss": 0.0005, "step": 42770 }, { "epoch": 7.84, "learning_rate": 2.0046288135828763e-08, "loss": 0.0003, "step": 42771 }, { "epoch": 7.84, "learning_rate": 1.999932950266792e-08, "loss": 0.0012, "step": 42772 }, { "epoch": 7.84, "learning_rate": 1.9952425879505853e-08, "loss": 0.0003, "step": 42773 }, { "epoch": 7.84, "learning_rate": 1.990557726660014e-08, "loss": 0.0003, "step": 42774 }, { "epoch": 7.84, "learning_rate": 1.9858783664207238e-08, "loss": 0.0002, "step": 42775 }, { "epoch": 7.84, "learning_rate": 1.981204507258694e-08, "loss": 0.0003, "step": 42776 }, { "epoch": 7.84, "learning_rate": 1.9765361491996816e-08, "loss": 0.0004, "step": 42777 }, { "epoch": 7.84, "learning_rate": 1.971873292269444e-08, "loss": 0.0003, "step": 42778 }, { "epoch": 7.85, "learning_rate": 1.9672159364935162e-08, "loss": 0.0002, "step": 42779 }, { "epoch": 7.85, "learning_rate": 1.9625640818976553e-08, "loss": 0.0007, "step": 42780 }, { "epoch": 7.85, "learning_rate": 1.957917728507619e-08, "loss": 0.0003, "step": 42781 }, { "epoch": 7.85, "learning_rate": 1.953276876348942e-08, "loss": 0.0002, "step": 42782 }, { "epoch": 7.85, "learning_rate": 1.9486415254470482e-08, "loss": 0.0003, "step": 42783 }, { "epoch": 7.85, "learning_rate": 1.944011675827695e-08, "loss": 0.0003, "step": 42784 }, { "epoch": 7.85, "learning_rate": 1.9393873275164176e-08, "loss": 0.0008, "step": 42785 }, { "epoch": 7.85, "learning_rate": 1.934768480538529e-08, "loss": 0.0003, "step": 42786 }, { "epoch": 7.85, "learning_rate": 1.9301551349195648e-08, "loss": 0.0008, "step": 42787 }, { "epoch": 7.85, "learning_rate": 1.9255472906849482e-08, "loss": 0.0005, "step": 42788 }, { "epoch": 7.85, "learning_rate": 1.920944947860215e-08, "loss": 0.0003, "step": 42789 }, { "epoch": 7.85, "learning_rate": 1.9163481064704558e-08, "loss": 0.0003, "step": 42790 }, { "epoch": 7.85, "learning_rate": 1.9117567665413172e-08, "loss": 0.0003, "step": 42791 }, { "epoch": 7.85, "learning_rate": 1.90717092809789e-08, "loss": 0.0004, "step": 42792 }, { "epoch": 7.85, "learning_rate": 1.9025905911654875e-08, "loss": 0.0008, "step": 42793 }, { "epoch": 7.85, "learning_rate": 1.8980157557694224e-08, "loss": 0.0005, "step": 42794 }, { "epoch": 7.85, "learning_rate": 1.893446421934897e-08, "loss": 0.0007, "step": 42795 }, { "epoch": 7.85, "learning_rate": 1.8888825896871133e-08, "loss": 0.0007, "step": 42796 }, { "epoch": 7.85, "learning_rate": 1.8843242590510514e-08, "loss": 0.0008, "step": 42797 }, { "epoch": 7.85, "learning_rate": 1.8797714300521353e-08, "loss": 0.0003, "step": 42798 }, { "epoch": 7.85, "learning_rate": 1.875224102715234e-08, "loss": 0.0004, "step": 42799 }, { "epoch": 7.85, "learning_rate": 1.8706822770654388e-08, "loss": 0.0003, "step": 42800 }, { "epoch": 7.85, "learning_rate": 1.8661459531278402e-08, "loss": 0.0004, "step": 42801 }, { "epoch": 7.85, "learning_rate": 1.8616151309275298e-08, "loss": 0.0005, "step": 42802 }, { "epoch": 7.85, "learning_rate": 1.8570898104892654e-08, "loss": 0.0003, "step": 42803 }, { "epoch": 7.85, "learning_rate": 1.8525699918381378e-08, "loss": 0.0003, "step": 42804 }, { "epoch": 7.85, "learning_rate": 1.8480556749991274e-08, "loss": 0.0004, "step": 42805 }, { "epoch": 7.85, "learning_rate": 1.843546859996881e-08, "loss": 0.0007, "step": 42806 }, { "epoch": 7.85, "learning_rate": 1.8390435468563784e-08, "loss": 0.0004, "step": 42807 }, { "epoch": 7.85, "learning_rate": 1.8345457356025997e-08, "loss": 0.0006, "step": 42808 }, { "epoch": 7.85, "learning_rate": 1.830053426260081e-08, "loss": 0.0003, "step": 42809 }, { "epoch": 7.85, "learning_rate": 1.8255666188536912e-08, "loss": 0.0003, "step": 42810 }, { "epoch": 7.85, "learning_rate": 1.8210853134081884e-08, "loss": 0.0002, "step": 42811 }, { "epoch": 7.85, "learning_rate": 1.8166095099482196e-08, "loss": 0.0001, "step": 42812 }, { "epoch": 7.85, "learning_rate": 1.8121392084985423e-08, "loss": 0.0004, "step": 42813 }, { "epoch": 7.85, "learning_rate": 1.807674409083693e-08, "loss": 0.0005, "step": 42814 }, { "epoch": 7.85, "learning_rate": 1.803215111728318e-08, "loss": 0.0012, "step": 42815 }, { "epoch": 7.85, "learning_rate": 1.798761316457065e-08, "loss": 0.0003, "step": 42816 }, { "epoch": 7.85, "learning_rate": 1.7943130232943585e-08, "loss": 0.0005, "step": 42817 }, { "epoch": 7.85, "learning_rate": 1.7898702322648453e-08, "loss": 0.0002, "step": 42818 }, { "epoch": 7.85, "learning_rate": 1.7854329433929507e-08, "loss": 0.0003, "step": 42819 }, { "epoch": 7.85, "learning_rate": 1.7810011567030995e-08, "loss": 0.0003, "step": 42820 }, { "epoch": 7.85, "learning_rate": 1.7765748722197163e-08, "loss": 0.0008, "step": 42821 }, { "epoch": 7.85, "learning_rate": 1.7721540899673372e-08, "loss": 0.0004, "step": 42822 }, { "epoch": 7.85, "learning_rate": 1.7677388099701652e-08, "loss": 0.0006, "step": 42823 }, { "epoch": 7.85, "learning_rate": 1.7633290322526254e-08, "loss": 0.0004, "step": 42824 }, { "epoch": 7.85, "learning_rate": 1.758924756839031e-08, "loss": 0.0003, "step": 42825 }, { "epoch": 7.85, "learning_rate": 1.7545259837535854e-08, "loss": 0.0006, "step": 42826 }, { "epoch": 7.85, "learning_rate": 1.7501327130204916e-08, "loss": 0.0003, "step": 42827 }, { "epoch": 7.85, "learning_rate": 1.745744944664174e-08, "loss": 0.0003, "step": 42828 }, { "epoch": 7.85, "learning_rate": 1.7413626787087244e-08, "loss": 0.0008, "step": 42829 }, { "epoch": 7.85, "learning_rate": 1.7369859151782352e-08, "loss": 0.0004, "step": 42830 }, { "epoch": 7.85, "learning_rate": 1.732614654096798e-08, "loss": 0.0006, "step": 42831 }, { "epoch": 7.85, "learning_rate": 1.7282488954886158e-08, "loss": 0.0003, "step": 42832 }, { "epoch": 7.85, "learning_rate": 1.72388863937778e-08, "loss": 0.0005, "step": 42833 }, { "epoch": 7.86, "learning_rate": 1.7195338857881604e-08, "loss": 0.0005, "step": 42834 }, { "epoch": 7.86, "learning_rate": 1.7151846347438493e-08, "loss": 0.0004, "step": 42835 }, { "epoch": 7.86, "learning_rate": 1.7108408862689386e-08, "loss": 0.0006, "step": 42836 }, { "epoch": 7.86, "learning_rate": 1.7065026403871866e-08, "loss": 0.0004, "step": 42837 }, { "epoch": 7.86, "learning_rate": 1.7021698971225743e-08, "loss": 0.0002, "step": 42838 }, { "epoch": 7.86, "learning_rate": 1.6978426564989715e-08, "loss": 0.0004, "step": 42839 }, { "epoch": 7.86, "learning_rate": 1.693520918540359e-08, "loss": 0.0008, "step": 42840 }, { "epoch": 7.86, "learning_rate": 1.6892046832702736e-08, "loss": 0.0006, "step": 42841 }, { "epoch": 7.86, "learning_rate": 1.684893950712807e-08, "loss": 0.0002, "step": 42842 }, { "epoch": 7.86, "learning_rate": 1.6805887208916073e-08, "loss": 0.0003, "step": 42843 }, { "epoch": 7.86, "learning_rate": 1.6762889938303216e-08, "loss": 0.0002, "step": 42844 }, { "epoch": 7.86, "learning_rate": 1.6719947695527095e-08, "loss": 0.0004, "step": 42845 }, { "epoch": 7.86, "learning_rate": 1.667706048082529e-08, "loss": 0.0005, "step": 42846 }, { "epoch": 7.86, "learning_rate": 1.6634228294433174e-08, "loss": 0.0005, "step": 42847 }, { "epoch": 7.86, "learning_rate": 1.6591451136586112e-08, "loss": 0.001, "step": 42848 }, { "epoch": 7.86, "learning_rate": 1.654872900752169e-08, "loss": 0.0003, "step": 42849 }, { "epoch": 7.86, "learning_rate": 1.650606190747417e-08, "loss": 0.0002, "step": 42850 }, { "epoch": 7.86, "learning_rate": 1.646344983668002e-08, "loss": 0.0004, "step": 42851 }, { "epoch": 7.86, "learning_rate": 1.6420892795372402e-08, "loss": 0.0003, "step": 42852 }, { "epoch": 7.86, "learning_rate": 1.637839078378667e-08, "loss": 0.0006, "step": 42853 }, { "epoch": 7.86, "learning_rate": 1.6335943802158194e-08, "loss": 0.0004, "step": 42854 }, { "epoch": 7.86, "learning_rate": 1.6293551850719012e-08, "loss": 0.0003, "step": 42855 }, { "epoch": 7.86, "learning_rate": 1.6251214929703386e-08, "loss": 0.0007, "step": 42856 }, { "epoch": 7.86, "learning_rate": 1.6208933039345565e-08, "loss": 0.0005, "step": 42857 }, { "epoch": 7.86, "learning_rate": 1.6166706179877588e-08, "loss": 0.0002, "step": 42858 }, { "epoch": 7.86, "learning_rate": 1.6124534351532607e-08, "loss": 0.0004, "step": 42859 }, { "epoch": 7.86, "learning_rate": 1.608241755454376e-08, "loss": 0.0013, "step": 42860 }, { "epoch": 7.86, "learning_rate": 1.604035578914198e-08, "loss": 0.0003, "step": 42861 }, { "epoch": 7.86, "learning_rate": 1.5998349055559304e-08, "loss": 0.0003, "step": 42862 }, { "epoch": 7.86, "learning_rate": 1.5956397354027765e-08, "loss": 0.0004, "step": 42863 }, { "epoch": 7.86, "learning_rate": 1.59145006847794e-08, "loss": 0.0004, "step": 42864 }, { "epoch": 7.86, "learning_rate": 1.5872659048042914e-08, "loss": 0.0005, "step": 42865 }, { "epoch": 7.86, "learning_rate": 1.583087244405146e-08, "loss": 0.0003, "step": 42866 }, { "epoch": 7.86, "learning_rate": 1.5789140873033737e-08, "loss": 0.0003, "step": 42867 }, { "epoch": 7.86, "learning_rate": 1.5747464335219563e-08, "loss": 0.0003, "step": 42868 }, { "epoch": 7.86, "learning_rate": 1.570584283083987e-08, "loss": 0.0007, "step": 42869 }, { "epoch": 7.86, "learning_rate": 1.5664276360123352e-08, "loss": 0.0006, "step": 42870 }, { "epoch": 7.86, "learning_rate": 1.5622764923299838e-08, "loss": 0.0004, "step": 42871 }, { "epoch": 7.86, "learning_rate": 1.5581308520595804e-08, "loss": 0.0003, "step": 42872 }, { "epoch": 7.86, "learning_rate": 1.5539907152243295e-08, "loss": 0.0002, "step": 42873 }, { "epoch": 7.86, "learning_rate": 1.5498560818467678e-08, "loss": 0.0004, "step": 42874 }, { "epoch": 7.86, "learning_rate": 1.5457269519498773e-08, "loss": 0.0009, "step": 42875 }, { "epoch": 7.86, "learning_rate": 1.5416033255563066e-08, "loss": 0.0004, "step": 42876 }, { "epoch": 7.86, "learning_rate": 1.5374852026887045e-08, "loss": 0.0003, "step": 42877 }, { "epoch": 7.86, "learning_rate": 1.533372583369941e-08, "loss": 0.0003, "step": 42878 }, { "epoch": 7.86, "learning_rate": 1.5292654676226648e-08, "loss": 0.0006, "step": 42879 }, { "epoch": 7.86, "learning_rate": 1.5251638554694137e-08, "loss": 0.0006, "step": 42880 }, { "epoch": 7.86, "learning_rate": 1.521067746932947e-08, "loss": 0.0005, "step": 42881 }, { "epoch": 7.86, "learning_rate": 1.5169771420356914e-08, "loss": 0.0004, "step": 42882 }, { "epoch": 7.86, "learning_rate": 1.5128920408001845e-08, "loss": 0.0003, "step": 42883 }, { "epoch": 7.86, "learning_rate": 1.5088124432490747e-08, "loss": 0.0003, "step": 42884 }, { "epoch": 7.86, "learning_rate": 1.5047383494047883e-08, "loss": 0.0011, "step": 42885 }, { "epoch": 7.86, "learning_rate": 1.5006697592897522e-08, "loss": 0.0006, "step": 42886 }, { "epoch": 7.86, "learning_rate": 1.496606672926393e-08, "loss": 0.0003, "step": 42887 }, { "epoch": 7.87, "learning_rate": 1.4925490903371365e-08, "loss": 0.0003, "step": 42888 }, { "epoch": 7.87, "learning_rate": 1.4884970115444097e-08, "loss": 0.0002, "step": 42889 }, { "epoch": 7.87, "learning_rate": 1.4844504365704171e-08, "loss": 0.0003, "step": 42890 }, { "epoch": 7.87, "learning_rate": 1.480409365437585e-08, "loss": 0.0002, "step": 42891 }, { "epoch": 7.87, "learning_rate": 1.476373798168118e-08, "loss": 0.0007, "step": 42892 }, { "epoch": 7.87, "learning_rate": 1.4723437347842207e-08, "loss": 0.0002, "step": 42893 }, { "epoch": 7.87, "learning_rate": 1.4683191753080972e-08, "loss": 0.0007, "step": 42894 }, { "epoch": 7.87, "learning_rate": 1.4643001197620633e-08, "loss": 0.0002, "step": 42895 }, { "epoch": 7.87, "learning_rate": 1.4602865681682122e-08, "loss": 0.0005, "step": 42896 }, { "epoch": 7.87, "learning_rate": 1.4562785205487484e-08, "loss": 0.0004, "step": 42897 }, { "epoch": 7.87, "learning_rate": 1.4522759769255435e-08, "loss": 0.0002, "step": 42898 }, { "epoch": 7.87, "learning_rate": 1.4482789373208017e-08, "loss": 0.0005, "step": 42899 }, { "epoch": 7.87, "learning_rate": 1.4442874017565056e-08, "loss": 0.0004, "step": 42900 }, { "epoch": 7.87, "learning_rate": 1.4403013702547486e-08, "loss": 0.0003, "step": 42901 }, { "epoch": 7.87, "learning_rate": 1.436320842837513e-08, "loss": 0.0003, "step": 42902 }, { "epoch": 7.87, "learning_rate": 1.4323458195265594e-08, "loss": 0.0012, "step": 42903 }, { "epoch": 7.87, "learning_rate": 1.428376300343981e-08, "loss": 0.0003, "step": 42904 }, { "epoch": 7.87, "learning_rate": 1.4244122853115383e-08, "loss": 0.0009, "step": 42905 }, { "epoch": 7.87, "learning_rate": 1.4204537744511026e-08, "loss": 0.0003, "step": 42906 }, { "epoch": 7.87, "learning_rate": 1.4165007677845455e-08, "loss": 0.0003, "step": 42907 }, { "epoch": 7.87, "learning_rate": 1.4125532653336272e-08, "loss": 0.0003, "step": 42908 }, { "epoch": 7.87, "learning_rate": 1.408611267120219e-08, "loss": 0.0004, "step": 42909 }, { "epoch": 7.87, "learning_rate": 1.4046747731658595e-08, "loss": 0.0009, "step": 42910 }, { "epoch": 7.87, "learning_rate": 1.4007437834923087e-08, "loss": 0.0002, "step": 42911 }, { "epoch": 7.87, "learning_rate": 1.3968182981212164e-08, "loss": 0.0004, "step": 42912 }, { "epoch": 7.87, "learning_rate": 1.3928983170743427e-08, "loss": 0.0006, "step": 42913 }, { "epoch": 7.87, "learning_rate": 1.3889838403731149e-08, "loss": 0.0006, "step": 42914 }, { "epoch": 7.87, "learning_rate": 1.3850748680392933e-08, "loss": 0.0008, "step": 42915 }, { "epoch": 7.87, "learning_rate": 1.3811714000943055e-08, "loss": 0.0004, "step": 42916 }, { "epoch": 7.87, "learning_rate": 1.3772734365596897e-08, "loss": 0.0003, "step": 42917 }, { "epoch": 7.87, "learning_rate": 1.3733809774568729e-08, "loss": 0.0005, "step": 42918 }, { "epoch": 7.87, "learning_rate": 1.3694940228073938e-08, "loss": 0.0003, "step": 42919 }, { "epoch": 7.87, "learning_rate": 1.3656125726327907e-08, "loss": 0.0009, "step": 42920 }, { "epoch": 7.87, "learning_rate": 1.3617366269541577e-08, "loss": 0.0004, "step": 42921 }, { "epoch": 7.87, "learning_rate": 1.357866185793144e-08, "loss": 0.0004, "step": 42922 }, { "epoch": 7.87, "learning_rate": 1.3540012491709554e-08, "loss": 0.0003, "step": 42923 }, { "epoch": 7.87, "learning_rate": 1.3501418171089075e-08, "loss": 0.0003, "step": 42924 }, { "epoch": 7.87, "learning_rate": 1.346287889628206e-08, "loss": 0.0003, "step": 42925 }, { "epoch": 7.87, "learning_rate": 1.3424394667501672e-08, "loss": 0.0003, "step": 42926 }, { "epoch": 7.87, "learning_rate": 1.3385965484961073e-08, "loss": 0.0006, "step": 42927 }, { "epoch": 7.87, "learning_rate": 1.3347591348870092e-08, "loss": 0.0012, "step": 42928 }, { "epoch": 7.87, "learning_rate": 1.3309272259440786e-08, "loss": 0.0003, "step": 42929 }, { "epoch": 7.87, "learning_rate": 1.3271008216885206e-08, "loss": 0.0004, "step": 42930 }, { "epoch": 7.87, "learning_rate": 1.3232799221414294e-08, "loss": 0.0002, "step": 42931 }, { "epoch": 7.87, "learning_rate": 1.3194645273237883e-08, "loss": 0.0003, "step": 42932 }, { "epoch": 7.87, "learning_rate": 1.3156546372565803e-08, "loss": 0.0004, "step": 42933 }, { "epoch": 7.87, "learning_rate": 1.311850251961011e-08, "loss": 0.0006, "step": 42934 }, { "epoch": 7.87, "learning_rate": 1.3080513714578413e-08, "loss": 0.001, "step": 42935 }, { "epoch": 7.87, "learning_rate": 1.3042579957680546e-08, "loss": 0.0004, "step": 42936 }, { "epoch": 7.87, "learning_rate": 1.300470124912634e-08, "loss": 0.0005, "step": 42937 }, { "epoch": 7.87, "learning_rate": 1.2966877589124516e-08, "loss": 0.0003, "step": 42938 }, { "epoch": 7.87, "learning_rate": 1.2929108977882687e-08, "loss": 0.0006, "step": 42939 }, { "epoch": 7.87, "learning_rate": 1.2891395415610686e-08, "loss": 0.0005, "step": 42940 }, { "epoch": 7.87, "learning_rate": 1.2853736902513902e-08, "loss": 0.0004, "step": 42941 }, { "epoch": 7.87, "learning_rate": 1.2816133438802169e-08, "loss": 0.0004, "step": 42942 }, { "epoch": 7.88, "learning_rate": 1.2778585024680879e-08, "loss": 0.0003, "step": 42943 }, { "epoch": 7.88, "learning_rate": 1.2741091660358751e-08, "loss": 0.0003, "step": 42944 }, { "epoch": 7.88, "learning_rate": 1.2703653346042288e-08, "loss": 0.0003, "step": 42945 }, { "epoch": 7.88, "learning_rate": 1.2666270081936883e-08, "loss": 0.0005, "step": 42946 }, { "epoch": 7.88, "learning_rate": 1.2628941868247924e-08, "loss": 0.0002, "step": 42947 }, { "epoch": 7.88, "learning_rate": 1.2591668705183024e-08, "loss": 0.0004, "step": 42948 }, { "epoch": 7.88, "learning_rate": 1.2554450592946466e-08, "loss": 0.0005, "step": 42949 }, { "epoch": 7.88, "learning_rate": 1.2517287531744748e-08, "loss": 0.0003, "step": 42950 }, { "epoch": 7.88, "learning_rate": 1.2480179521781044e-08, "loss": 0.0007, "step": 42951 }, { "epoch": 7.88, "learning_rate": 1.2443126563260743e-08, "loss": 0.0005, "step": 42952 }, { "epoch": 7.88, "learning_rate": 1.2406128656387018e-08, "loss": 0.0004, "step": 42953 }, { "epoch": 7.88, "learning_rate": 1.2369185801366368e-08, "loss": 0.0002, "step": 42954 }, { "epoch": 7.88, "learning_rate": 1.2332297998399745e-08, "loss": 0.0002, "step": 42955 }, { "epoch": 7.88, "learning_rate": 1.2295465247691429e-08, "loss": 0.0004, "step": 42956 }, { "epoch": 7.88, "learning_rate": 1.2258687549444593e-08, "loss": 0.0004, "step": 42957 }, { "epoch": 7.88, "learning_rate": 1.2221964903861295e-08, "loss": 0.0002, "step": 42958 }, { "epoch": 7.88, "learning_rate": 1.2185297311145815e-08, "loss": 0.0005, "step": 42959 }, { "epoch": 7.88, "learning_rate": 1.2148684771497999e-08, "loss": 0.0004, "step": 42960 }, { "epoch": 7.88, "learning_rate": 1.2112127285121011e-08, "loss": 0.0004, "step": 42961 }, { "epoch": 7.88, "learning_rate": 1.2075624852215806e-08, "loss": 0.0006, "step": 42962 }, { "epoch": 7.88, "learning_rate": 1.2039177472983331e-08, "loss": 0.0004, "step": 42963 }, { "epoch": 7.88, "learning_rate": 1.2002785147625651e-08, "loss": 0.0009, "step": 42964 }, { "epoch": 7.88, "learning_rate": 1.1966447876342601e-08, "loss": 0.0004, "step": 42965 }, { "epoch": 7.88, "learning_rate": 1.1930165659334026e-08, "loss": 0.0004, "step": 42966 }, { "epoch": 7.88, "learning_rate": 1.1893938496800871e-08, "loss": 0.0003, "step": 42967 }, { "epoch": 7.88, "learning_rate": 1.1857766388941872e-08, "loss": 0.0004, "step": 42968 }, { "epoch": 7.88, "learning_rate": 1.1821649335956864e-08, "loss": 0.0006, "step": 42969 }, { "epoch": 7.88, "learning_rate": 1.178558733804458e-08, "loss": 0.0004, "step": 42970 }, { "epoch": 7.88, "learning_rate": 1.1749580395404857e-08, "loss": 0.0002, "step": 42971 }, { "epoch": 7.88, "learning_rate": 1.171362850823532e-08, "loss": 0.0003, "step": 42972 }, { "epoch": 7.88, "learning_rate": 1.1677731676733584e-08, "loss": 0.0004, "step": 42973 }, { "epoch": 7.88, "learning_rate": 1.164188990109949e-08, "loss": 0.0003, "step": 42974 }, { "epoch": 7.88, "learning_rate": 1.1606103181528438e-08, "loss": 0.0004, "step": 42975 }, { "epoch": 7.88, "learning_rate": 1.1570371518219159e-08, "loss": 0.0003, "step": 42976 }, { "epoch": 7.88, "learning_rate": 1.153469491136705e-08, "loss": 0.0002, "step": 42977 }, { "epoch": 7.88, "learning_rate": 1.1499073361169733e-08, "loss": 0.0004, "step": 42978 }, { "epoch": 7.88, "learning_rate": 1.1463506867824826e-08, "loss": 0.0004, "step": 42979 }, { "epoch": 7.88, "learning_rate": 1.1427995431525508e-08, "loss": 0.0007, "step": 42980 }, { "epoch": 7.88, "learning_rate": 1.1392539052470508e-08, "loss": 0.0008, "step": 42981 }, { "epoch": 7.88, "learning_rate": 1.1357137730853008e-08, "loss": 0.0002, "step": 42982 }, { "epoch": 7.88, "learning_rate": 1.1321791466868404e-08, "loss": 0.0004, "step": 42983 }, { "epoch": 7.88, "learning_rate": 1.1286500260713207e-08, "loss": 0.0005, "step": 42984 }, { "epoch": 7.88, "learning_rate": 1.1251264112580596e-08, "loss": 0.0003, "step": 42985 }, { "epoch": 7.88, "learning_rate": 1.1216083022664859e-08, "loss": 0.0011, "step": 42986 }, { "epoch": 7.88, "learning_rate": 1.1180956991160285e-08, "loss": 0.0002, "step": 42987 }, { "epoch": 7.88, "learning_rate": 1.1145886018260054e-08, "loss": 0.0007, "step": 42988 }, { "epoch": 7.88, "learning_rate": 1.1110870104157345e-08, "loss": 0.0004, "step": 42989 }, { "epoch": 7.88, "learning_rate": 1.1075909249046445e-08, "loss": 0.0003, "step": 42990 }, { "epoch": 7.88, "learning_rate": 1.1041003453118315e-08, "loss": 0.0004, "step": 42991 }, { "epoch": 7.88, "learning_rate": 1.1006152716566132e-08, "loss": 0.0003, "step": 42992 }, { "epoch": 7.88, "learning_rate": 1.0971357039583075e-08, "loss": 0.0005, "step": 42993 }, { "epoch": 7.88, "learning_rate": 1.0936616422358992e-08, "loss": 0.0004, "step": 42994 }, { "epoch": 7.88, "learning_rate": 1.0901930865085952e-08, "loss": 0.0004, "step": 42995 }, { "epoch": 7.88, "learning_rate": 1.0867300367956024e-08, "loss": 0.0007, "step": 42996 }, { "epoch": 7.89, "learning_rate": 1.0832724931160165e-08, "loss": 0.0009, "step": 42997 }, { "epoch": 7.89, "learning_rate": 1.0798204554888225e-08, "loss": 0.0004, "step": 42998 }, { "epoch": 7.89, "learning_rate": 1.0763739239330052e-08, "loss": 0.0005, "step": 42999 }, { "epoch": 7.89, "learning_rate": 1.0729328984675492e-08, "loss": 0.0006, "step": 43000 }, { "epoch": 7.89, "learning_rate": 1.0694973791116614e-08, "loss": 0.0004, "step": 43001 }, { "epoch": 7.89, "learning_rate": 1.0660673658839937e-08, "loss": 0.0007, "step": 43002 }, { "epoch": 7.89, "learning_rate": 1.0626428588035309e-08, "loss": 0.0006, "step": 43003 }, { "epoch": 7.89, "learning_rate": 1.0592238578892577e-08, "loss": 0.0003, "step": 43004 }, { "epoch": 7.89, "learning_rate": 1.055810363159937e-08, "loss": 0.0004, "step": 43005 }, { "epoch": 7.89, "learning_rate": 1.0524023746344425e-08, "loss": 0.0005, "step": 43006 }, { "epoch": 7.89, "learning_rate": 1.0489998923314259e-08, "loss": 0.0007, "step": 43007 }, { "epoch": 7.89, "learning_rate": 1.0456029162697612e-08, "loss": 0.0004, "step": 43008 }, { "epoch": 7.89, "learning_rate": 1.0422114464680998e-08, "loss": 0.0003, "step": 43009 }, { "epoch": 7.89, "learning_rate": 1.0388254829452049e-08, "loss": 0.0002, "step": 43010 }, { "epoch": 7.89, "learning_rate": 1.0354450257197279e-08, "loss": 0.0007, "step": 43011 }, { "epoch": 7.89, "learning_rate": 1.0320700748102097e-08, "loss": 0.0003, "step": 43012 }, { "epoch": 7.89, "learning_rate": 1.028700630235413e-08, "loss": 0.0004, "step": 43013 }, { "epoch": 7.89, "learning_rate": 1.0253366920137674e-08, "loss": 0.0003, "step": 43014 }, { "epoch": 7.89, "learning_rate": 1.021978260163814e-08, "loss": 0.0003, "step": 43015 }, { "epoch": 7.89, "learning_rate": 1.0186253347042042e-08, "loss": 0.0003, "step": 43016 }, { "epoch": 7.89, "learning_rate": 1.0152779156533676e-08, "loss": 0.0005, "step": 43017 }, { "epoch": 7.89, "learning_rate": 1.0119360030296233e-08, "loss": 0.0003, "step": 43018 }, { "epoch": 7.89, "learning_rate": 1.0085995968515116e-08, "loss": 0.0002, "step": 43019 }, { "epoch": 7.89, "learning_rate": 1.0052686971374626e-08, "loss": 0.0003, "step": 43020 }, { "epoch": 7.89, "learning_rate": 1.0019433039057946e-08, "loss": 0.0005, "step": 43021 }, { "epoch": 7.89, "learning_rate": 9.986234171747156e-09, "loss": 0.0004, "step": 43022 }, { "epoch": 7.89, "learning_rate": 9.95309036962655e-09, "loss": 0.0002, "step": 43023 }, { "epoch": 7.89, "learning_rate": 9.920001632879317e-09, "loss": 0.0004, "step": 43024 }, { "epoch": 7.89, "learning_rate": 9.886967961686423e-09, "loss": 0.0003, "step": 43025 }, { "epoch": 7.89, "learning_rate": 9.853989356231053e-09, "loss": 0.0002, "step": 43026 }, { "epoch": 7.89, "learning_rate": 9.821065816694176e-09, "loss": 0.0006, "step": 43027 }, { "epoch": 7.89, "learning_rate": 9.788197343257866e-09, "loss": 0.0003, "step": 43028 }, { "epoch": 7.89, "learning_rate": 9.75538393610309e-09, "loss": 0.0003, "step": 43029 }, { "epoch": 7.89, "learning_rate": 9.722625595409707e-09, "loss": 0.0002, "step": 43030 }, { "epoch": 7.89, "learning_rate": 9.68992232135979e-09, "loss": 0.0005, "step": 43031 }, { "epoch": 7.89, "learning_rate": 9.657274114133198e-09, "loss": 0.0003, "step": 43032 }, { "epoch": 7.89, "learning_rate": 9.624680973910893e-09, "loss": 0.0007, "step": 43033 }, { "epoch": 7.89, "learning_rate": 9.592142900869405e-09, "loss": 0.0003, "step": 43034 }, { "epoch": 7.89, "learning_rate": 9.55965989519192e-09, "loss": 0.0004, "step": 43035 }, { "epoch": 7.89, "learning_rate": 9.52723195705385e-09, "loss": 0.0002, "step": 43036 }, { "epoch": 7.89, "learning_rate": 9.494859086637276e-09, "loss": 0.0003, "step": 43037 }, { "epoch": 7.89, "learning_rate": 9.46254128411872e-09, "loss": 0.0003, "step": 43038 }, { "epoch": 7.89, "learning_rate": 9.430278549675819e-09, "loss": 0.0003, "step": 43039 }, { "epoch": 7.89, "learning_rate": 9.39807088348843e-09, "loss": 0.0014, "step": 43040 }, { "epoch": 7.89, "learning_rate": 9.365918285731968e-09, "loss": 0.0004, "step": 43041 }, { "epoch": 7.89, "learning_rate": 9.333820756585178e-09, "loss": 0.0004, "step": 43042 }, { "epoch": 7.89, "learning_rate": 9.301778296224584e-09, "loss": 0.0005, "step": 43043 }, { "epoch": 7.89, "learning_rate": 9.269790904825604e-09, "loss": 0.0005, "step": 43044 }, { "epoch": 7.89, "learning_rate": 9.237858582566983e-09, "loss": 0.0007, "step": 43045 }, { "epoch": 7.89, "learning_rate": 9.205981329623026e-09, "loss": 0.0004, "step": 43046 }, { "epoch": 7.89, "learning_rate": 9.174159146169148e-09, "loss": 0.0003, "step": 43047 }, { "epoch": 7.89, "learning_rate": 9.142392032381875e-09, "loss": 0.0004, "step": 43048 }, { "epoch": 7.89, "learning_rate": 9.11067998843551e-09, "loss": 0.0003, "step": 43049 }, { "epoch": 7.89, "learning_rate": 9.079023014504363e-09, "loss": 0.0003, "step": 43050 }, { "epoch": 7.89, "learning_rate": 9.047421110764953e-09, "loss": 0.0005, "step": 43051 }, { "epoch": 7.9, "learning_rate": 9.01587427738937e-09, "loss": 0.0007, "step": 43052 }, { "epoch": 7.9, "learning_rate": 8.984382514553024e-09, "loss": 0.0004, "step": 43053 }, { "epoch": 7.9, "learning_rate": 8.952945822429115e-09, "loss": 0.0002, "step": 43054 }, { "epoch": 7.9, "learning_rate": 8.921564201189724e-09, "loss": 0.0008, "step": 43055 }, { "epoch": 7.9, "learning_rate": 8.89023765100916e-09, "loss": 0.0002, "step": 43056 }, { "epoch": 7.9, "learning_rate": 8.858966172059502e-09, "loss": 0.0003, "step": 43057 }, { "epoch": 7.9, "learning_rate": 8.82774976451395e-09, "loss": 0.0005, "step": 43058 }, { "epoch": 7.9, "learning_rate": 8.796588428543474e-09, "loss": 0.0003, "step": 43059 }, { "epoch": 7.9, "learning_rate": 8.76548216432016e-09, "loss": 0.0004, "step": 43060 }, { "epoch": 7.9, "learning_rate": 8.734430972014985e-09, "loss": 0.0003, "step": 43061 }, { "epoch": 7.9, "learning_rate": 8.70343485180114e-09, "loss": 0.0002, "step": 43062 }, { "epoch": 7.9, "learning_rate": 8.67249380384627e-09, "loss": 0.0004, "step": 43063 }, { "epoch": 7.9, "learning_rate": 8.641607828324682e-09, "loss": 0.0003, "step": 43064 }, { "epoch": 7.9, "learning_rate": 8.610776925402909e-09, "loss": 0.0005, "step": 43065 }, { "epoch": 7.9, "learning_rate": 8.580001095253032e-09, "loss": 0.0003, "step": 43066 }, { "epoch": 7.9, "learning_rate": 8.54928033804492e-09, "loss": 0.0006, "step": 43067 }, { "epoch": 7.9, "learning_rate": 8.518614653946211e-09, "loss": 0.0003, "step": 43068 }, { "epoch": 7.9, "learning_rate": 8.488004043127884e-09, "loss": 0.0004, "step": 43069 }, { "epoch": 7.9, "learning_rate": 8.457448505757582e-09, "loss": 0.0004, "step": 43070 }, { "epoch": 7.9, "learning_rate": 8.426948042004058e-09, "loss": 0.0002, "step": 43071 }, { "epoch": 7.9, "learning_rate": 8.396502652034955e-09, "loss": 0.0006, "step": 43072 }, { "epoch": 7.9, "learning_rate": 8.366112336017917e-09, "loss": 0.0008, "step": 43073 }, { "epoch": 7.9, "learning_rate": 8.3357770941217e-09, "loss": 0.0006, "step": 43074 }, { "epoch": 7.9, "learning_rate": 8.305496926512835e-09, "loss": 0.0004, "step": 43075 }, { "epoch": 7.9, "learning_rate": 8.275271833357857e-09, "loss": 0.0012, "step": 43076 }, { "epoch": 7.9, "learning_rate": 8.245101814823298e-09, "loss": 0.0002, "step": 43077 }, { "epoch": 7.9, "learning_rate": 8.214986871076803e-09, "loss": 0.0004, "step": 43078 }, { "epoch": 7.9, "learning_rate": 8.184927002282683e-09, "loss": 0.0006, "step": 43079 }, { "epoch": 7.9, "learning_rate": 8.154922208607474e-09, "loss": 0.0003, "step": 43080 }, { "epoch": 7.9, "learning_rate": 8.124972490215488e-09, "loss": 0.0005, "step": 43081 }, { "epoch": 7.9, "learning_rate": 8.095077847274369e-09, "loss": 0.0007, "step": 43082 }, { "epoch": 7.9, "learning_rate": 8.065238279946208e-09, "loss": 0.0003, "step": 43083 }, { "epoch": 7.9, "learning_rate": 8.03545378839643e-09, "loss": 0.001, "step": 43084 }, { "epoch": 7.9, "learning_rate": 8.005724372790458e-09, "loss": 0.0002, "step": 43085 }, { "epoch": 7.9, "learning_rate": 7.976050033290384e-09, "loss": 0.0003, "step": 43086 }, { "epoch": 7.9, "learning_rate": 7.94643077006052e-09, "loss": 0.0004, "step": 43087 }, { "epoch": 7.9, "learning_rate": 7.916866583262961e-09, "loss": 0.0004, "step": 43088 }, { "epoch": 7.9, "learning_rate": 7.88735747306313e-09, "loss": 0.0006, "step": 43089 }, { "epoch": 7.9, "learning_rate": 7.857903439622005e-09, "loss": 0.0003, "step": 43090 }, { "epoch": 7.9, "learning_rate": 7.828504483101684e-09, "loss": 0.0002, "step": 43091 }, { "epoch": 7.9, "learning_rate": 7.799160603664257e-09, "loss": 0.0009, "step": 43092 }, { "epoch": 7.9, "learning_rate": 7.769871801472927e-09, "loss": 0.0003, "step": 43093 }, { "epoch": 7.9, "learning_rate": 7.740638076688677e-09, "loss": 0.0011, "step": 43094 }, { "epoch": 7.9, "learning_rate": 7.71145942947027e-09, "loss": 0.0004, "step": 43095 }, { "epoch": 7.9, "learning_rate": 7.682335859982015e-09, "loss": 0.0012, "step": 43096 }, { "epoch": 7.9, "learning_rate": 7.653267368381567e-09, "loss": 0.0002, "step": 43097 }, { "epoch": 7.9, "learning_rate": 7.624253954832128e-09, "loss": 0.0003, "step": 43098 }, { "epoch": 7.9, "learning_rate": 7.595295619490239e-09, "loss": 0.0005, "step": 43099 }, { "epoch": 7.9, "learning_rate": 7.566392362517994e-09, "loss": 0.0003, "step": 43100 }, { "epoch": 7.9, "learning_rate": 7.537544184074152e-09, "loss": 0.0003, "step": 43101 }, { "epoch": 7.9, "learning_rate": 7.508751084316367e-09, "loss": 0.0008, "step": 43102 }, { "epoch": 7.9, "learning_rate": 7.480013063405623e-09, "loss": 0.0005, "step": 43103 }, { "epoch": 7.9, "learning_rate": 7.451330121498457e-09, "loss": 0.0006, "step": 43104 }, { "epoch": 7.9, "learning_rate": 7.422702258753634e-09, "loss": 0.0003, "step": 43105 }, { "epoch": 7.91, "learning_rate": 7.394129475328804e-09, "loss": 0.0008, "step": 43106 }, { "epoch": 7.91, "learning_rate": 7.36561177138273e-09, "loss": 0.0008, "step": 43107 }, { "epoch": 7.91, "learning_rate": 7.337149147070844e-09, "loss": 0.0004, "step": 43108 }, { "epoch": 7.91, "learning_rate": 7.308741602550795e-09, "loss": 0.0002, "step": 43109 }, { "epoch": 7.91, "learning_rate": 7.280389137978016e-09, "loss": 0.0003, "step": 43110 }, { "epoch": 7.91, "learning_rate": 7.25209175351127e-09, "loss": 0.0003, "step": 43111 }, { "epoch": 7.91, "learning_rate": 7.223849449304876e-09, "loss": 0.0002, "step": 43112 }, { "epoch": 7.91, "learning_rate": 7.195662225514266e-09, "loss": 0.0002, "step": 43113 }, { "epoch": 7.91, "learning_rate": 7.1675300822948715e-09, "loss": 0.0002, "step": 43114 }, { "epoch": 7.91, "learning_rate": 7.139453019803233e-09, "loss": 0.0002, "step": 43115 }, { "epoch": 7.91, "learning_rate": 7.111431038192563e-09, "loss": 0.0002, "step": 43116 }, { "epoch": 7.91, "learning_rate": 7.083464137617179e-09, "loss": 0.0005, "step": 43117 }, { "epoch": 7.91, "learning_rate": 7.0555523182325165e-09, "loss": 0.0004, "step": 43118 }, { "epoch": 7.91, "learning_rate": 7.0276955801917845e-09, "loss": 0.0007, "step": 43119 }, { "epoch": 7.91, "learning_rate": 6.999893923648194e-09, "loss": 0.0002, "step": 43120 }, { "epoch": 7.91, "learning_rate": 6.9721473487560644e-09, "loss": 0.0004, "step": 43121 }, { "epoch": 7.91, "learning_rate": 6.944455855666388e-09, "loss": 0.0003, "step": 43122 }, { "epoch": 7.91, "learning_rate": 6.916819444533485e-09, "loss": 0.0006, "step": 43123 }, { "epoch": 7.91, "learning_rate": 6.889238115509456e-09, "loss": 0.0004, "step": 43124 }, { "epoch": 7.91, "learning_rate": 6.861711868746401e-09, "loss": 0.0004, "step": 43125 }, { "epoch": 7.91, "learning_rate": 6.834240704395312e-09, "loss": 0.0003, "step": 43126 }, { "epoch": 7.91, "learning_rate": 6.8068246226071775e-09, "loss": 0.0003, "step": 43127 }, { "epoch": 7.91, "learning_rate": 6.7794636235352095e-09, "loss": 0.0007, "step": 43128 }, { "epoch": 7.91, "learning_rate": 6.752157707328177e-09, "loss": 0.0005, "step": 43129 }, { "epoch": 7.91, "learning_rate": 6.7249068741370715e-09, "loss": 0.0003, "step": 43130 }, { "epoch": 7.91, "learning_rate": 6.697711124112883e-09, "loss": 0.0004, "step": 43131 }, { "epoch": 7.91, "learning_rate": 6.670570457405489e-09, "loss": 0.0003, "step": 43132 }, { "epoch": 7.91, "learning_rate": 6.643484874163664e-09, "loss": 0.0005, "step": 43133 }, { "epoch": 7.91, "learning_rate": 6.616454374537284e-09, "loss": 0.0005, "step": 43134 }, { "epoch": 7.91, "learning_rate": 6.589478958674012e-09, "loss": 0.0002, "step": 43135 }, { "epoch": 7.91, "learning_rate": 6.5625586267248355e-09, "loss": 0.0007, "step": 43136 }, { "epoch": 7.91, "learning_rate": 6.535693378836305e-09, "loss": 0.0005, "step": 43137 }, { "epoch": 7.91, "learning_rate": 6.508883215158301e-09, "loss": 0.0003, "step": 43138 }, { "epoch": 7.91, "learning_rate": 6.482128135836263e-09, "loss": 0.0004, "step": 43139 }, { "epoch": 7.91, "learning_rate": 6.45542814101896e-09, "loss": 0.0003, "step": 43140 }, { "epoch": 7.91, "learning_rate": 6.428783230854052e-09, "loss": 0.0003, "step": 43141 }, { "epoch": 7.91, "learning_rate": 6.402193405486978e-09, "loss": 0.0003, "step": 43142 }, { "epoch": 7.91, "learning_rate": 6.375658665065398e-09, "loss": 0.0006, "step": 43143 }, { "epoch": 7.91, "learning_rate": 6.3491790097358615e-09, "loss": 0.0002, "step": 43144 }, { "epoch": 7.91, "learning_rate": 6.322754439643808e-09, "loss": 0.0004, "step": 43145 }, { "epoch": 7.91, "learning_rate": 6.296384954934676e-09, "loss": 0.0005, "step": 43146 }, { "epoch": 7.91, "learning_rate": 6.270070555753904e-09, "loss": 0.0004, "step": 43147 }, { "epoch": 7.91, "learning_rate": 6.243811242245823e-09, "loss": 0.0004, "step": 43148 }, { "epoch": 7.91, "learning_rate": 6.217607014556981e-09, "loss": 0.0002, "step": 43149 }, { "epoch": 7.91, "learning_rate": 6.191457872830597e-09, "loss": 0.0008, "step": 43150 }, { "epoch": 7.91, "learning_rate": 6.165363817211001e-09, "loss": 0.0005, "step": 43151 }, { "epoch": 7.91, "learning_rate": 6.139324847842521e-09, "loss": 0.0015, "step": 43152 }, { "epoch": 7.91, "learning_rate": 6.113340964867265e-09, "loss": 0.0002, "step": 43153 }, { "epoch": 7.91, "learning_rate": 6.0874121684295626e-09, "loss": 0.0006, "step": 43154 }, { "epoch": 7.91, "learning_rate": 6.061538458672633e-09, "loss": 0.0004, "step": 43155 }, { "epoch": 7.91, "learning_rate": 6.035719835738585e-09, "loss": 0.0005, "step": 43156 }, { "epoch": 7.91, "learning_rate": 6.009956299769526e-09, "loss": 0.001, "step": 43157 }, { "epoch": 7.91, "learning_rate": 5.984247850908676e-09, "loss": 0.0003, "step": 43158 }, { "epoch": 7.91, "learning_rate": 5.958594489295921e-09, "loss": 0.0006, "step": 43159 }, { "epoch": 7.91, "learning_rate": 5.932996215073372e-09, "loss": 0.0004, "step": 43160 }, { "epoch": 7.92, "learning_rate": 5.907453028382026e-09, "loss": 0.0006, "step": 43161 }, { "epoch": 7.92, "learning_rate": 5.8819649293639925e-09, "loss": 0.0005, "step": 43162 }, { "epoch": 7.92, "learning_rate": 5.856531918158048e-09, "loss": 0.0002, "step": 43163 }, { "epoch": 7.92, "learning_rate": 5.8311539949051925e-09, "loss": 0.0005, "step": 43164 }, { "epoch": 7.92, "learning_rate": 5.805831159745312e-09, "loss": 0.0005, "step": 43165 }, { "epoch": 7.92, "learning_rate": 5.780563412817186e-09, "loss": 0.0002, "step": 43166 }, { "epoch": 7.92, "learning_rate": 5.7553507542607025e-09, "loss": 0.0005, "step": 43167 }, { "epoch": 7.92, "learning_rate": 5.730193184215749e-09, "loss": 0.0003, "step": 43168 }, { "epoch": 7.92, "learning_rate": 5.705090702819993e-09, "loss": 0.0002, "step": 43169 }, { "epoch": 7.92, "learning_rate": 5.680043310211103e-09, "loss": 0.0005, "step": 43170 }, { "epoch": 7.92, "learning_rate": 5.655051006528967e-09, "loss": 0.0002, "step": 43171 }, { "epoch": 7.92, "learning_rate": 5.630113791909031e-09, "loss": 0.0005, "step": 43172 }, { "epoch": 7.92, "learning_rate": 5.6052316664911845e-09, "loss": 0.0004, "step": 43173 }, { "epoch": 7.92, "learning_rate": 5.580404630410874e-09, "loss": 0.0007, "step": 43174 }, { "epoch": 7.92, "learning_rate": 5.555632683805767e-09, "loss": 0.0017, "step": 43175 }, { "epoch": 7.92, "learning_rate": 5.530915826812422e-09, "loss": 0.0006, "step": 43176 }, { "epoch": 7.92, "learning_rate": 5.506254059566285e-09, "loss": 0.0004, "step": 43177 }, { "epoch": 7.92, "learning_rate": 5.481647382203914e-09, "loss": 0.0005, "step": 43178 }, { "epoch": 7.92, "learning_rate": 5.457095794861866e-09, "loss": 0.0003, "step": 43179 }, { "epoch": 7.92, "learning_rate": 5.432599297673369e-09, "loss": 0.0006, "step": 43180 }, { "epoch": 7.92, "learning_rate": 5.408157890773868e-09, "loss": 0.0006, "step": 43181 }, { "epoch": 7.92, "learning_rate": 5.383771574299923e-09, "loss": 0.0006, "step": 43182 }, { "epoch": 7.92, "learning_rate": 5.3594403483847594e-09, "loss": 0.0003, "step": 43183 }, { "epoch": 7.92, "learning_rate": 5.335164213162714e-09, "loss": 0.0004, "step": 43184 }, { "epoch": 7.92, "learning_rate": 5.310943168767013e-09, "loss": 0.0005, "step": 43185 }, { "epoch": 7.92, "learning_rate": 5.286777215330885e-09, "loss": 0.0002, "step": 43186 }, { "epoch": 7.92, "learning_rate": 5.262666352988666e-09, "loss": 0.0003, "step": 43187 }, { "epoch": 7.92, "learning_rate": 5.238610581872472e-09, "loss": 0.0004, "step": 43188 }, { "epoch": 7.92, "learning_rate": 5.214609902116641e-09, "loss": 0.0005, "step": 43189 }, { "epoch": 7.92, "learning_rate": 5.190664313851068e-09, "loss": 0.0014, "step": 43190 }, { "epoch": 7.92, "learning_rate": 5.1667738172078704e-09, "loss": 0.0003, "step": 43191 }, { "epoch": 7.92, "learning_rate": 5.142938412320275e-09, "loss": 0.0003, "step": 43192 }, { "epoch": 7.92, "learning_rate": 5.1191580993192876e-09, "loss": 0.0002, "step": 43193 }, { "epoch": 7.92, "learning_rate": 5.095432878335915e-09, "loss": 0.0003, "step": 43194 }, { "epoch": 7.92, "learning_rate": 5.071762749500053e-09, "loss": 0.0003, "step": 43195 }, { "epoch": 7.92, "learning_rate": 5.048147712942708e-09, "loss": 0.0005, "step": 43196 }, { "epoch": 7.92, "learning_rate": 5.024587768793776e-09, "loss": 0.0007, "step": 43197 }, { "epoch": 7.92, "learning_rate": 5.001082917184263e-09, "loss": 0.0002, "step": 43198 }, { "epoch": 7.92, "learning_rate": 4.977633158242956e-09, "loss": 0.0007, "step": 43199 }, { "epoch": 7.92, "learning_rate": 4.95423849209864e-09, "loss": 0.0007, "step": 43200 }, { "epoch": 7.92, "learning_rate": 4.93089891888121e-09, "loss": 0.0003, "step": 43201 }, { "epoch": 7.92, "learning_rate": 4.9076144387183445e-09, "loss": 0.0009, "step": 43202 }, { "epoch": 7.92, "learning_rate": 4.884385051738827e-09, "loss": 0.0002, "step": 43203 }, { "epoch": 7.92, "learning_rate": 4.861210758071444e-09, "loss": 0.0002, "step": 43204 }, { "epoch": 7.92, "learning_rate": 4.838091557842761e-09, "loss": 0.0007, "step": 43205 }, { "epoch": 7.92, "learning_rate": 4.815027451180454e-09, "loss": 0.0004, "step": 43206 }, { "epoch": 7.92, "learning_rate": 4.792018438212198e-09, "loss": 0.0002, "step": 43207 }, { "epoch": 7.92, "learning_rate": 4.769064519064559e-09, "loss": 0.0002, "step": 43208 }, { "epoch": 7.92, "learning_rate": 4.746165693862992e-09, "loss": 0.0006, "step": 43209 }, { "epoch": 7.92, "learning_rate": 4.723321962736282e-09, "loss": 0.0005, "step": 43210 }, { "epoch": 7.92, "learning_rate": 4.700533325807666e-09, "loss": 0.0007, "step": 43211 }, { "epoch": 7.92, "learning_rate": 4.677799783204817e-09, "loss": 0.0002, "step": 43212 }, { "epoch": 7.92, "learning_rate": 4.655121335050972e-09, "loss": 0.0003, "step": 43213 }, { "epoch": 7.92, "learning_rate": 4.632497981472694e-09, "loss": 0.0002, "step": 43214 }, { "epoch": 7.92, "learning_rate": 4.60992972259433e-09, "loss": 0.0002, "step": 43215 }, { "epoch": 7.93, "learning_rate": 4.587416558540225e-09, "loss": 0.0008, "step": 43216 }, { "epoch": 7.93, "learning_rate": 4.564958489434723e-09, "loss": 0.0005, "step": 43217 }, { "epoch": 7.93, "learning_rate": 4.542555515401059e-09, "loss": 0.0007, "step": 43218 }, { "epoch": 7.93, "learning_rate": 4.5202076365635785e-09, "loss": 0.0004, "step": 43219 }, { "epoch": 7.93, "learning_rate": 4.497914853044405e-09, "loss": 0.0015, "step": 43220 }, { "epoch": 7.93, "learning_rate": 4.475677164966774e-09, "loss": 0.0002, "step": 43221 }, { "epoch": 7.93, "learning_rate": 4.45349457245281e-09, "loss": 0.0008, "step": 43222 }, { "epoch": 7.93, "learning_rate": 4.431367075625747e-09, "loss": 0.0004, "step": 43223 }, { "epoch": 7.93, "learning_rate": 4.409294674607711e-09, "loss": 0.0003, "step": 43224 }, { "epoch": 7.93, "learning_rate": 4.387277369519716e-09, "loss": 0.0008, "step": 43225 }, { "epoch": 7.93, "learning_rate": 4.365315160482775e-09, "loss": 0.0006, "step": 43226 }, { "epoch": 7.93, "learning_rate": 4.343408047619013e-09, "loss": 0.0003, "step": 43227 }, { "epoch": 7.93, "learning_rate": 4.321556031048335e-09, "loss": 0.0005, "step": 43228 }, { "epoch": 7.93, "learning_rate": 4.299759110890645e-09, "loss": 0.0002, "step": 43229 }, { "epoch": 7.93, "learning_rate": 4.278017287268066e-09, "loss": 0.0004, "step": 43230 }, { "epoch": 7.93, "learning_rate": 4.256330560298283e-09, "loss": 0.0005, "step": 43231 }, { "epoch": 7.93, "learning_rate": 4.23469893010231e-09, "loss": 0.0006, "step": 43232 }, { "epoch": 7.93, "learning_rate": 4.213122396798941e-09, "loss": 0.0001, "step": 43233 }, { "epoch": 7.93, "learning_rate": 4.1916009605058594e-09, "loss": 0.0002, "step": 43234 }, { "epoch": 7.93, "learning_rate": 4.17013462134408e-09, "loss": 0.0005, "step": 43235 }, { "epoch": 7.93, "learning_rate": 4.1487233794301754e-09, "loss": 0.0008, "step": 43236 }, { "epoch": 7.93, "learning_rate": 4.12736723488294e-09, "loss": 0.0003, "step": 43237 }, { "epoch": 7.93, "learning_rate": 4.106066187820057e-09, "loss": 0.0003, "step": 43238 }, { "epoch": 7.93, "learning_rate": 4.0848202383581e-09, "loss": 0.0005, "step": 43239 }, { "epoch": 7.93, "learning_rate": 4.063629386615864e-09, "loss": 0.0003, "step": 43240 }, { "epoch": 7.93, "learning_rate": 4.0424936327077e-09, "loss": 0.0008, "step": 43241 }, { "epoch": 7.93, "learning_rate": 4.0214129767524034e-09, "loss": 0.0004, "step": 43242 }, { "epoch": 7.93, "learning_rate": 4.000387418865437e-09, "loss": 0.0002, "step": 43243 }, { "epoch": 7.93, "learning_rate": 3.979416959161153e-09, "loss": 0.0004, "step": 43244 }, { "epoch": 7.93, "learning_rate": 3.958501597758346e-09, "loss": 0.0004, "step": 43245 }, { "epoch": 7.93, "learning_rate": 3.937641334769149e-09, "loss": 0.0004, "step": 43246 }, { "epoch": 7.93, "learning_rate": 3.916836170310134e-09, "loss": 0.0007, "step": 43247 }, { "epoch": 7.93, "learning_rate": 3.8960861044956554e-09, "loss": 0.0003, "step": 43248 }, { "epoch": 7.93, "learning_rate": 3.875391137440065e-09, "loss": 0.0004, "step": 43249 }, { "epoch": 7.93, "learning_rate": 3.854751269257717e-09, "loss": 0.0002, "step": 43250 }, { "epoch": 7.93, "learning_rate": 3.834166500062963e-09, "loss": 0.0004, "step": 43251 }, { "epoch": 7.93, "learning_rate": 3.8136368299668266e-09, "loss": 0.0006, "step": 43252 }, { "epoch": 7.93, "learning_rate": 3.793162259084771e-09, "loss": 0.0006, "step": 43253 }, { "epoch": 7.93, "learning_rate": 3.772742787530037e-09, "loss": 0.0003, "step": 43254 }, { "epoch": 7.93, "learning_rate": 3.752378415412539e-09, "loss": 0.0003, "step": 43255 }, { "epoch": 7.93, "learning_rate": 3.732069142847739e-09, "loss": 0.0003, "step": 43256 }, { "epoch": 7.93, "learning_rate": 3.7118149699444385e-09, "loss": 0.0006, "step": 43257 }, { "epoch": 7.93, "learning_rate": 3.6916158968169913e-09, "loss": 0.0004, "step": 43258 }, { "epoch": 7.93, "learning_rate": 3.6714719235741993e-09, "loss": 0.0006, "step": 43259 }, { "epoch": 7.93, "learning_rate": 3.6513830503293047e-09, "loss": 0.0003, "step": 43260 }, { "epoch": 7.93, "learning_rate": 3.6313492771922198e-09, "loss": 0.0003, "step": 43261 }, { "epoch": 7.93, "learning_rate": 3.6113706042739673e-09, "loss": 0.0003, "step": 43262 }, { "epoch": 7.93, "learning_rate": 3.5914470316833484e-09, "loss": 0.0004, "step": 43263 }, { "epoch": 7.93, "learning_rate": 3.5715785595302755e-09, "loss": 0.0003, "step": 43264 }, { "epoch": 7.93, "learning_rate": 3.551765187925771e-09, "loss": 0.0003, "step": 43265 }, { "epoch": 7.93, "learning_rate": 3.5320069169775263e-09, "loss": 0.0005, "step": 43266 }, { "epoch": 7.93, "learning_rate": 3.5123037467954534e-09, "loss": 0.0003, "step": 43267 }, { "epoch": 7.93, "learning_rate": 3.4926556774872445e-09, "loss": 0.0014, "step": 43268 }, { "epoch": 7.93, "learning_rate": 3.4730627091617007e-09, "loss": 0.0003, "step": 43269 }, { "epoch": 7.94, "learning_rate": 3.4535248419276244e-09, "loss": 0.0008, "step": 43270 }, { "epoch": 7.94, "learning_rate": 3.434042075891597e-09, "loss": 0.0014, "step": 43271 }, { "epoch": 7.94, "learning_rate": 3.41461441116131e-09, "loss": 0.0003, "step": 43272 }, { "epoch": 7.94, "learning_rate": 3.3952418478433446e-09, "loss": 0.0002, "step": 43273 }, { "epoch": 7.94, "learning_rate": 3.3759243860453926e-09, "loss": 0.0003, "step": 43274 }, { "epoch": 7.94, "learning_rate": 3.3566620258729254e-09, "loss": 0.0006, "step": 43275 }, { "epoch": 7.94, "learning_rate": 3.3374547674336343e-09, "loss": 0.0003, "step": 43276 }, { "epoch": 7.94, "learning_rate": 3.3183026108318805e-09, "loss": 0.0009, "step": 43277 }, { "epoch": 7.94, "learning_rate": 3.2992055561742453e-09, "loss": 0.0004, "step": 43278 }, { "epoch": 7.94, "learning_rate": 3.2801636035662e-09, "loss": 0.0009, "step": 43279 }, { "epoch": 7.94, "learning_rate": 3.261176753110995e-09, "loss": 0.0003, "step": 43280 }, { "epoch": 7.94, "learning_rate": 3.242245004915212e-09, "loss": 0.0002, "step": 43281 }, { "epoch": 7.94, "learning_rate": 3.2233683590809915e-09, "loss": 0.001, "step": 43282 }, { "epoch": 7.94, "learning_rate": 3.2045468157149153e-09, "loss": 0.0006, "step": 43283 }, { "epoch": 7.94, "learning_rate": 3.185780374920233e-09, "loss": 0.0002, "step": 43284 }, { "epoch": 7.94, "learning_rate": 3.167069036797976e-09, "loss": 0.0005, "step": 43285 }, { "epoch": 7.94, "learning_rate": 3.1484128014547257e-09, "loss": 0.0002, "step": 43286 }, { "epoch": 7.94, "learning_rate": 3.1298116689904014e-09, "loss": 0.0015, "step": 43287 }, { "epoch": 7.94, "learning_rate": 3.111265639509364e-09, "loss": 0.0003, "step": 43288 }, { "epoch": 7.94, "learning_rate": 3.092774713112645e-09, "loss": 0.0003, "step": 43289 }, { "epoch": 7.94, "learning_rate": 3.0743388899034943e-09, "loss": 0.0001, "step": 43290 }, { "epoch": 7.94, "learning_rate": 3.055958169981832e-09, "loss": 0.0005, "step": 43291 }, { "epoch": 7.94, "learning_rate": 3.037632553450909e-09, "loss": 0.0003, "step": 43292 }, { "epoch": 7.94, "learning_rate": 3.019362040410645e-09, "loss": 0.0004, "step": 43293 }, { "epoch": 7.94, "learning_rate": 3.0011466309609603e-09, "loss": 0.0002, "step": 43294 }, { "epoch": 7.94, "learning_rate": 2.982986325203996e-09, "loss": 0.0007, "step": 43295 }, { "epoch": 7.94, "learning_rate": 2.9648811232396713e-09, "loss": 0.0003, "step": 43296 }, { "epoch": 7.94, "learning_rate": 2.946831025165686e-09, "loss": 0.0006, "step": 43297 }, { "epoch": 7.94, "learning_rate": 2.9288360310841814e-09, "loss": 0.0006, "step": 43298 }, { "epoch": 7.94, "learning_rate": 2.9108961410917458e-09, "loss": 0.0005, "step": 43299 }, { "epoch": 7.94, "learning_rate": 2.8930113552894103e-09, "loss": 0.001, "step": 43300 }, { "epoch": 7.94, "learning_rate": 2.875181673775984e-09, "loss": 0.0003, "step": 43301 }, { "epoch": 7.94, "learning_rate": 2.8574070966469467e-09, "loss": 0.0003, "step": 43302 }, { "epoch": 7.94, "learning_rate": 2.839687624003329e-09, "loss": 0.0011, "step": 43303 }, { "epoch": 7.94, "learning_rate": 2.8220232559417193e-09, "loss": 0.0005, "step": 43304 }, { "epoch": 7.94, "learning_rate": 2.804413992558708e-09, "loss": 0.0002, "step": 43305 }, { "epoch": 7.94, "learning_rate": 2.786859833951994e-09, "loss": 0.0003, "step": 43306 }, { "epoch": 7.94, "learning_rate": 2.7693607802181666e-09, "loss": 0.0002, "step": 43307 }, { "epoch": 7.94, "learning_rate": 2.7519168314538157e-09, "loss": 0.0007, "step": 43308 }, { "epoch": 7.94, "learning_rate": 2.734527987755531e-09, "loss": 0.0003, "step": 43309 }, { "epoch": 7.94, "learning_rate": 2.7171942492187907e-09, "loss": 0.0004, "step": 43310 }, { "epoch": 7.94, "learning_rate": 2.699915615939075e-09, "loss": 0.0002, "step": 43311 }, { "epoch": 7.94, "learning_rate": 2.682692088010752e-09, "loss": 0.0002, "step": 43312 }, { "epoch": 7.94, "learning_rate": 2.6655236655304116e-09, "loss": 0.0003, "step": 43313 }, { "epoch": 7.94, "learning_rate": 2.648410348592423e-09, "loss": 0.0003, "step": 43314 }, { "epoch": 7.94, "learning_rate": 2.6313521372900443e-09, "loss": 0.0007, "step": 43315 }, { "epoch": 7.94, "learning_rate": 2.6143490317176445e-09, "loss": 0.0003, "step": 43316 }, { "epoch": 7.94, "learning_rate": 2.597401031969593e-09, "loss": 0.0004, "step": 43317 }, { "epoch": 7.94, "learning_rate": 2.580508138138038e-09, "loss": 0.0002, "step": 43318 }, { "epoch": 7.94, "learning_rate": 2.563670350317349e-09, "loss": 0.0004, "step": 43319 }, { "epoch": 7.94, "learning_rate": 2.5468876686007837e-09, "loss": 0.0005, "step": 43320 }, { "epoch": 7.94, "learning_rate": 2.5301600930793813e-09, "loss": 0.0003, "step": 43321 }, { "epoch": 7.94, "learning_rate": 2.51348762384529e-09, "loss": 0.0003, "step": 43322 }, { "epoch": 7.94, "learning_rate": 2.4968702609917683e-09, "loss": 0.0007, "step": 43323 }, { "epoch": 7.94, "learning_rate": 2.4803080046098547e-09, "loss": 0.0005, "step": 43324 }, { "epoch": 7.95, "learning_rate": 2.463800854790588e-09, "loss": 0.0016, "step": 43325 }, { "epoch": 7.95, "learning_rate": 2.4473488116250055e-09, "loss": 0.0006, "step": 43326 }, { "epoch": 7.95, "learning_rate": 2.430951875203036e-09, "loss": 0.0004, "step": 43327 }, { "epoch": 7.95, "learning_rate": 2.4146100456168276e-09, "loss": 0.0004, "step": 43328 }, { "epoch": 7.95, "learning_rate": 2.3983233229551982e-09, "loss": 0.0006, "step": 43329 }, { "epoch": 7.95, "learning_rate": 2.3820917073091864e-09, "loss": 0.0003, "step": 43330 }, { "epoch": 7.95, "learning_rate": 2.3659151987664997e-09, "loss": 0.0012, "step": 43331 }, { "epoch": 7.95, "learning_rate": 2.349793797417066e-09, "loss": 0.0003, "step": 43332 }, { "epoch": 7.95, "learning_rate": 2.333727503349703e-09, "loss": 0.0002, "step": 43333 }, { "epoch": 7.95, "learning_rate": 2.317716316653229e-09, "loss": 0.0002, "step": 43334 }, { "epoch": 7.95, "learning_rate": 2.3017602374153513e-09, "loss": 0.0006, "step": 43335 }, { "epoch": 7.95, "learning_rate": 2.285859265724888e-09, "loss": 0.0004, "step": 43336 }, { "epoch": 7.95, "learning_rate": 2.270013401668436e-09, "loss": 0.001, "step": 43337 }, { "epoch": 7.95, "learning_rate": 2.2542226453348135e-09, "loss": 0.0003, "step": 43338 }, { "epoch": 7.95, "learning_rate": 2.2384869968095078e-09, "loss": 0.0003, "step": 43339 }, { "epoch": 7.95, "learning_rate": 2.2228064561791163e-09, "loss": 0.0004, "step": 43340 }, { "epoch": 7.95, "learning_rate": 2.2071810235324564e-09, "loss": 0.0007, "step": 43341 }, { "epoch": 7.95, "learning_rate": 2.1916106989527954e-09, "loss": 0.0005, "step": 43342 }, { "epoch": 7.95, "learning_rate": 2.17609548252784e-09, "loss": 0.0004, "step": 43343 }, { "epoch": 7.95, "learning_rate": 2.160635374341968e-09, "loss": 0.0007, "step": 43344 }, { "epoch": 7.95, "learning_rate": 2.145230374481777e-09, "loss": 0.0007, "step": 43345 }, { "epoch": 7.95, "learning_rate": 2.129880483030533e-09, "loss": 0.0004, "step": 43346 }, { "epoch": 7.95, "learning_rate": 2.1145857000737234e-09, "loss": 0.0002, "step": 43347 }, { "epoch": 7.95, "learning_rate": 2.0993460256957252e-09, "loss": 0.0003, "step": 43348 }, { "epoch": 7.95, "learning_rate": 2.0841614599798055e-09, "loss": 0.0004, "step": 43349 }, { "epoch": 7.95, "learning_rate": 2.0690320030114508e-09, "loss": 0.0003, "step": 43350 }, { "epoch": 7.95, "learning_rate": 2.053957654871708e-09, "loss": 0.0007, "step": 43351 }, { "epoch": 7.95, "learning_rate": 2.038938415644953e-09, "loss": 0.0004, "step": 43352 }, { "epoch": 7.95, "learning_rate": 2.023974285414454e-09, "loss": 0.0004, "step": 43353 }, { "epoch": 7.95, "learning_rate": 2.0090652642612564e-09, "loss": 0.0006, "step": 43354 }, { "epoch": 7.95, "learning_rate": 1.9942113522686267e-09, "loss": 0.0004, "step": 43355 }, { "epoch": 7.95, "learning_rate": 1.979412549517612e-09, "loss": 0.0003, "step": 43356 }, { "epoch": 7.95, "learning_rate": 1.9646688560914784e-09, "loss": 0.0004, "step": 43357 }, { "epoch": 7.95, "learning_rate": 1.9499802720690518e-09, "loss": 0.0003, "step": 43358 }, { "epoch": 7.95, "learning_rate": 1.9353467975335994e-09, "loss": 0.0004, "step": 43359 }, { "epoch": 7.95, "learning_rate": 1.9207684325639464e-09, "loss": 0.0006, "step": 43360 }, { "epoch": 7.95, "learning_rate": 1.9062451772422495e-09, "loss": 0.0005, "step": 43361 }, { "epoch": 7.95, "learning_rate": 1.8917770316462247e-09, "loss": 0.0005, "step": 43362 }, { "epoch": 7.95, "learning_rate": 1.877363995858028e-09, "loss": 0.0004, "step": 43363 }, { "epoch": 7.95, "learning_rate": 1.8630060699553755e-09, "loss": 0.0002, "step": 43364 }, { "epoch": 7.95, "learning_rate": 1.8487032540182025e-09, "loss": 0.0004, "step": 43365 }, { "epoch": 7.95, "learning_rate": 1.8344555481253357e-09, "loss": 0.0005, "step": 43366 }, { "epoch": 7.95, "learning_rate": 1.8202629523556003e-09, "loss": 0.0005, "step": 43367 }, { "epoch": 7.95, "learning_rate": 1.8061254667867122e-09, "loss": 0.0003, "step": 43368 }, { "epoch": 7.95, "learning_rate": 1.7920430914963872e-09, "loss": 0.0005, "step": 43369 }, { "epoch": 7.95, "learning_rate": 1.7780158265623404e-09, "loss": 0.001, "step": 43370 }, { "epoch": 7.95, "learning_rate": 1.7640436720622877e-09, "loss": 0.0003, "step": 43371 }, { "epoch": 7.95, "learning_rate": 1.7501266280728345e-09, "loss": 0.0006, "step": 43372 }, { "epoch": 7.95, "learning_rate": 1.7362646946716967e-09, "loss": 0.0006, "step": 43373 }, { "epoch": 7.95, "learning_rate": 1.7224578719332585e-09, "loss": 0.001, "step": 43374 }, { "epoch": 7.95, "learning_rate": 1.7087061599352361e-09, "loss": 0.0002, "step": 43375 }, { "epoch": 7.95, "learning_rate": 1.6950095587542348e-09, "loss": 0.0003, "step": 43376 }, { "epoch": 7.95, "learning_rate": 1.6813680684635292e-09, "loss": 0.0006, "step": 43377 }, { "epoch": 7.95, "learning_rate": 1.6677816891397248e-09, "loss": 0.0003, "step": 43378 }, { "epoch": 7.96, "learning_rate": 1.6542504208572064e-09, "loss": 0.0017, "step": 43379 }, { "epoch": 7.96, "learning_rate": 1.640774263690359e-09, "loss": 0.001, "step": 43380 }, { "epoch": 7.96, "learning_rate": 1.6273532177146779e-09, "loss": 0.0006, "step": 43381 }, { "epoch": 7.96, "learning_rate": 1.6139872830023273e-09, "loss": 0.0002, "step": 43382 }, { "epoch": 7.96, "learning_rate": 1.6006764596288027e-09, "loss": 0.0007, "step": 43383 }, { "epoch": 7.96, "learning_rate": 1.5874207476651581e-09, "loss": 0.0005, "step": 43384 }, { "epoch": 7.96, "learning_rate": 1.574220147186889e-09, "loss": 0.0004, "step": 43385 }, { "epoch": 7.96, "learning_rate": 1.5610746582661595e-09, "loss": 0.0008, "step": 43386 }, { "epoch": 7.96, "learning_rate": 1.5479842809740243e-09, "loss": 0.0002, "step": 43387 }, { "epoch": 7.96, "learning_rate": 1.534949015384868e-09, "loss": 0.0005, "step": 43388 }, { "epoch": 7.96, "learning_rate": 1.5219688615686345e-09, "loss": 0.0002, "step": 43389 }, { "epoch": 7.96, "learning_rate": 1.5090438195974888e-09, "loss": 0.0003, "step": 43390 }, { "epoch": 7.96, "learning_rate": 1.496173889542485e-09, "loss": 0.0002, "step": 43391 }, { "epoch": 7.96, "learning_rate": 1.483359071474677e-09, "loss": 0.0002, "step": 43392 }, { "epoch": 7.96, "learning_rate": 1.4705993654651196e-09, "loss": 0.0005, "step": 43393 }, { "epoch": 7.96, "learning_rate": 1.4578947715848669e-09, "loss": 0.0003, "step": 43394 }, { "epoch": 7.96, "learning_rate": 1.4452452899016423e-09, "loss": 0.0007, "step": 43395 }, { "epoch": 7.96, "learning_rate": 1.4326509204865003e-09, "loss": 0.0004, "step": 43396 }, { "epoch": 7.96, "learning_rate": 1.4201116634093849e-09, "loss": 0.0006, "step": 43397 }, { "epoch": 7.96, "learning_rate": 1.4076275187391298e-09, "loss": 0.0004, "step": 43398 }, { "epoch": 7.96, "learning_rate": 1.3951984865434587e-09, "loss": 0.0005, "step": 43399 }, { "epoch": 7.96, "learning_rate": 1.3828245668923156e-09, "loss": 0.0006, "step": 43400 }, { "epoch": 7.96, "learning_rate": 1.3705057598534244e-09, "loss": 0.0002, "step": 43401 }, { "epoch": 7.96, "learning_rate": 1.3582420654933982e-09, "loss": 0.0002, "step": 43402 }, { "epoch": 7.96, "learning_rate": 1.3460334838821809e-09, "loss": 0.0003, "step": 43403 }, { "epoch": 7.96, "learning_rate": 1.333880015085276e-09, "loss": 0.0008, "step": 43404 }, { "epoch": 7.96, "learning_rate": 1.3217816591704068e-09, "loss": 0.0003, "step": 43405 }, { "epoch": 7.96, "learning_rate": 1.3097384162041871e-09, "loss": 0.0003, "step": 43406 }, { "epoch": 7.96, "learning_rate": 1.2977502862532298e-09, "loss": 0.0005, "step": 43407 }, { "epoch": 7.96, "learning_rate": 1.2858172693830385e-09, "loss": 0.0002, "step": 43408 }, { "epoch": 7.96, "learning_rate": 1.273939365659116e-09, "loss": 0.0003, "step": 43409 }, { "epoch": 7.96, "learning_rate": 1.2621165751480758e-09, "loss": 0.0006, "step": 43410 }, { "epoch": 7.96, "learning_rate": 1.2503488979143108e-09, "loss": 0.0003, "step": 43411 }, { "epoch": 7.96, "learning_rate": 1.2386363340233244e-09, "loss": 0.001, "step": 43412 }, { "epoch": 7.96, "learning_rate": 1.2269788835383989e-09, "loss": 0.0003, "step": 43413 }, { "epoch": 7.96, "learning_rate": 1.2153765465250378e-09, "loss": 0.0003, "step": 43414 }, { "epoch": 7.96, "learning_rate": 1.2038293230465236e-09, "loss": 0.0002, "step": 43415 }, { "epoch": 7.96, "learning_rate": 1.1923372131672496e-09, "loss": 0.0008, "step": 43416 }, { "epoch": 7.96, "learning_rate": 1.180900216950498e-09, "loss": 0.0002, "step": 43417 }, { "epoch": 7.96, "learning_rate": 1.1695183344573313e-09, "loss": 0.0013, "step": 43418 }, { "epoch": 7.96, "learning_rate": 1.1581915657532527e-09, "loss": 0.0003, "step": 43419 }, { "epoch": 7.96, "learning_rate": 1.1469199108993244e-09, "loss": 0.0002, "step": 43420 }, { "epoch": 7.96, "learning_rate": 1.1357033699577191e-09, "loss": 0.0003, "step": 43421 }, { "epoch": 7.96, "learning_rate": 1.1245419429906091e-09, "loss": 0.0004, "step": 43422 }, { "epoch": 7.96, "learning_rate": 1.1134356300579463e-09, "loss": 0.0006, "step": 43423 }, { "epoch": 7.96, "learning_rate": 1.1023844312241239e-09, "loss": 0.0002, "step": 43424 }, { "epoch": 7.96, "learning_rate": 1.0913883465468733e-09, "loss": 0.0005, "step": 43425 }, { "epoch": 7.96, "learning_rate": 1.0804473760883671e-09, "loss": 0.0004, "step": 43426 }, { "epoch": 7.96, "learning_rate": 1.0695615199096677e-09, "loss": 0.0002, "step": 43427 }, { "epoch": 7.96, "learning_rate": 1.0587307780685063e-09, "loss": 0.0006, "step": 43428 }, { "epoch": 7.96, "learning_rate": 1.0479551506259456e-09, "loss": 0.0004, "step": 43429 }, { "epoch": 7.96, "learning_rate": 1.0372346376419373e-09, "loss": 0.0007, "step": 43430 }, { "epoch": 7.96, "learning_rate": 1.0265692391753235e-09, "loss": 0.0006, "step": 43431 }, { "epoch": 7.96, "learning_rate": 1.0159589552838355e-09, "loss": 0.0003, "step": 43432 }, { "epoch": 7.96, "learning_rate": 1.0054037860263155e-09, "loss": 0.0011, "step": 43433 }, { "epoch": 7.97, "learning_rate": 9.94903731461605e-10, "loss": 0.0003, "step": 43434 }, { "epoch": 7.97, "learning_rate": 9.844587916474358e-10, "loss": 0.0003, "step": 43435 }, { "epoch": 7.97, "learning_rate": 9.740689666415393e-10, "loss": 0.0007, "step": 43436 }, { "epoch": 7.97, "learning_rate": 9.63734256500537e-10, "loss": 0.0008, "step": 43437 }, { "epoch": 7.97, "learning_rate": 9.534546612810502e-10, "loss": 0.0002, "step": 43438 }, { "epoch": 7.97, "learning_rate": 9.432301810408106e-10, "loss": 0.0001, "step": 43439 }, { "epoch": 7.97, "learning_rate": 9.330608158364395e-10, "loss": 0.0001, "step": 43440 }, { "epoch": 7.97, "learning_rate": 9.22946565722338e-10, "loss": 0.0005, "step": 43441 }, { "epoch": 7.97, "learning_rate": 9.128874307551273e-10, "loss": 0.0003, "step": 43442 }, { "epoch": 7.97, "learning_rate": 9.028834109914286e-10, "loss": 0.0004, "step": 43443 }, { "epoch": 7.97, "learning_rate": 8.929345064845329e-10, "loss": 0.0017, "step": 43444 }, { "epoch": 7.97, "learning_rate": 8.830407172899514e-10, "loss": 0.0009, "step": 43445 }, { "epoch": 7.97, "learning_rate": 8.732020434620847e-10, "loss": 0.0007, "step": 43446 }, { "epoch": 7.97, "learning_rate": 8.63418485055334e-10, "loss": 0.0004, "step": 43447 }, { "epoch": 7.97, "learning_rate": 8.536900421241001e-10, "loss": 0.0004, "step": 43448 }, { "epoch": 7.97, "learning_rate": 8.440167147216738e-10, "loss": 0.0002, "step": 43449 }, { "epoch": 7.97, "learning_rate": 8.343985029013457e-10, "loss": 0.0003, "step": 43450 }, { "epoch": 7.97, "learning_rate": 8.248354067164066e-10, "loss": 0.0002, "step": 43451 }, { "epoch": 7.97, "learning_rate": 8.153274262179267e-10, "loss": 0.0002, "step": 43452 }, { "epoch": 7.97, "learning_rate": 8.05874561460307e-10, "loss": 0.0003, "step": 43453 }, { "epoch": 7.97, "learning_rate": 7.964768124957279e-10, "loss": 0.0002, "step": 43454 }, { "epoch": 7.97, "learning_rate": 7.871341793741494e-10, "loss": 0.0005, "step": 43455 }, { "epoch": 7.97, "learning_rate": 7.778466621488623e-10, "loss": 0.0005, "step": 43456 }, { "epoch": 7.97, "learning_rate": 7.686142608698266e-10, "loss": 0.0006, "step": 43457 }, { "epoch": 7.97, "learning_rate": 7.594369755892228e-10, "loss": 0.0003, "step": 43458 }, { "epoch": 7.97, "learning_rate": 7.503148063559007e-10, "loss": 0.0003, "step": 43459 }, { "epoch": 7.97, "learning_rate": 7.412477532209306e-10, "loss": 0.0004, "step": 43460 }, { "epoch": 7.97, "learning_rate": 7.322358162353827e-10, "loss": 0.0003, "step": 43461 }, { "epoch": 7.97, "learning_rate": 7.232789954481068e-10, "loss": 0.0002, "step": 43462 }, { "epoch": 7.97, "learning_rate": 7.143772909079527e-10, "loss": 0.0005, "step": 43463 }, { "epoch": 7.97, "learning_rate": 7.055307026637703e-10, "loss": 0.0003, "step": 43464 }, { "epoch": 7.97, "learning_rate": 6.967392307655197e-10, "loss": 0.0005, "step": 43465 }, { "epoch": 7.97, "learning_rate": 6.880028752609402e-10, "loss": 0.0004, "step": 43466 }, { "epoch": 7.97, "learning_rate": 6.793216361988819e-10, "loss": 0.0003, "step": 43467 }, { "epoch": 7.97, "learning_rate": 6.706955136259741e-10, "loss": 0.0003, "step": 43468 }, { "epoch": 7.97, "learning_rate": 6.621245075910665e-10, "loss": 0.0009, "step": 43469 }, { "epoch": 7.97, "learning_rate": 6.536086181407886e-10, "loss": 0.0004, "step": 43470 }, { "epoch": 7.97, "learning_rate": 6.451478453217696e-10, "loss": 0.0004, "step": 43471 }, { "epoch": 7.97, "learning_rate": 6.367421891806391e-10, "loss": 0.0003, "step": 43472 }, { "epoch": 7.97, "learning_rate": 6.283916497651366e-10, "loss": 0.0004, "step": 43473 }, { "epoch": 7.97, "learning_rate": 6.200962271185606e-10, "loss": 0.0005, "step": 43474 }, { "epoch": 7.97, "learning_rate": 6.118559212897612e-10, "loss": 0.0006, "step": 43475 }, { "epoch": 7.97, "learning_rate": 6.03670732322037e-10, "loss": 0.0004, "step": 43476 }, { "epoch": 7.97, "learning_rate": 5.955406602609071e-10, "loss": 0.0004, "step": 43477 }, { "epoch": 7.97, "learning_rate": 5.874657051518906e-10, "loss": 0.0002, "step": 43478 }, { "epoch": 7.97, "learning_rate": 5.794458670393965e-10, "loss": 0.0007, "step": 43479 }, { "epoch": 7.97, "learning_rate": 5.714811459667235e-10, "loss": 0.0003, "step": 43480 }, { "epoch": 7.97, "learning_rate": 5.635715419771703e-10, "loss": 0.0006, "step": 43481 }, { "epoch": 7.97, "learning_rate": 5.557170551173663e-10, "loss": 0.0015, "step": 43482 }, { "epoch": 7.97, "learning_rate": 5.479176854272795e-10, "loss": 0.0003, "step": 43483 }, { "epoch": 7.97, "learning_rate": 5.401734329524289e-10, "loss": 0.0007, "step": 43484 }, { "epoch": 7.97, "learning_rate": 5.324842977338928e-10, "loss": 0.0005, "step": 43485 }, { "epoch": 7.97, "learning_rate": 5.248502798149701e-10, "loss": 0.0006, "step": 43486 }, { "epoch": 7.97, "learning_rate": 5.17271379236739e-10, "loss": 0.0003, "step": 43487 }, { "epoch": 7.98, "learning_rate": 5.097475960413878e-10, "loss": 0.0003, "step": 43488 }, { "epoch": 7.98, "learning_rate": 5.022789302711051e-10, "loss": 0.0005, "step": 43489 }, { "epoch": 7.98, "learning_rate": 4.948653819669691e-10, "loss": 0.0004, "step": 43490 }, { "epoch": 7.98, "learning_rate": 4.875069511689478e-10, "loss": 0.0003, "step": 43491 }, { "epoch": 7.98, "learning_rate": 4.802036379181196e-10, "loss": 0.0007, "step": 43492 }, { "epoch": 7.98, "learning_rate": 4.729554422544525e-10, "loss": 0.0003, "step": 43493 }, { "epoch": 7.98, "learning_rate": 4.657623642179143e-10, "loss": 0.0003, "step": 43494 }, { "epoch": 7.98, "learning_rate": 4.586244038484733e-10, "loss": 0.0008, "step": 43495 }, { "epoch": 7.98, "learning_rate": 4.515415611860974e-10, "loss": 0.0005, "step": 43496 }, { "epoch": 7.98, "learning_rate": 4.4451383626853417e-10, "loss": 0.0006, "step": 43497 }, { "epoch": 7.98, "learning_rate": 4.3754122913464146e-10, "loss": 0.0006, "step": 43498 }, { "epoch": 7.98, "learning_rate": 4.306237398232771e-10, "loss": 0.0003, "step": 43499 }, { "epoch": 7.98, "learning_rate": 4.2376136837329886e-10, "loss": 0.0003, "step": 43500 }, { "epoch": 7.98, "learning_rate": 4.1695411482134405e-10, "loss": 0.0006, "step": 43501 }, { "epoch": 7.98, "learning_rate": 4.102019792062706e-10, "loss": 0.0005, "step": 43502 }, { "epoch": 7.98, "learning_rate": 4.035049615636055e-10, "loss": 0.0002, "step": 43503 }, { "epoch": 7.98, "learning_rate": 3.9686306193109646e-10, "loss": 0.0008, "step": 43504 }, { "epoch": 7.98, "learning_rate": 3.902762803453808e-10, "loss": 0.0003, "step": 43505 }, { "epoch": 7.98, "learning_rate": 3.837446168430958e-10, "loss": 0.0006, "step": 43506 }, { "epoch": 7.98, "learning_rate": 3.7726807145865853e-10, "loss": 0.0005, "step": 43507 }, { "epoch": 7.98, "learning_rate": 3.7084664423092667e-10, "loss": 0.0003, "step": 43508 }, { "epoch": 7.98, "learning_rate": 3.644803351920967e-10, "loss": 0.0004, "step": 43509 }, { "epoch": 7.98, "learning_rate": 3.58169144378806e-10, "loss": 0.0002, "step": 43510 }, { "epoch": 7.98, "learning_rate": 3.519130718254715e-10, "loss": 0.0002, "step": 43511 }, { "epoch": 7.98, "learning_rate": 3.4571211756651014e-10, "loss": 0.0011, "step": 43512 }, { "epoch": 7.98, "learning_rate": 3.3956628163633874e-10, "loss": 0.0002, "step": 43513 }, { "epoch": 7.98, "learning_rate": 3.334755640693743e-10, "loss": 0.0002, "step": 43514 }, { "epoch": 7.98, "learning_rate": 3.2743996489781327e-10, "loss": 0.0006, "step": 43515 }, { "epoch": 7.98, "learning_rate": 3.214594841560725e-10, "loss": 0.0005, "step": 43516 }, { "epoch": 7.98, "learning_rate": 3.155341218763486e-10, "loss": 0.0002, "step": 43517 }, { "epoch": 7.98, "learning_rate": 3.096638780919481e-10, "loss": 0.0004, "step": 43518 }, { "epoch": 7.98, "learning_rate": 3.038487528350675e-10, "loss": 0.0003, "step": 43519 }, { "epoch": 7.98, "learning_rate": 2.9808874613679317e-10, "loss": 0.0003, "step": 43520 }, { "epoch": 7.98, "learning_rate": 2.923838580304317e-10, "loss": 0.0008, "step": 43521 }, { "epoch": 7.98, "learning_rate": 2.867340885470693e-10, "loss": 0.0003, "step": 43522 }, { "epoch": 7.98, "learning_rate": 2.811394377166821e-10, "loss": 0.0002, "step": 43523 }, { "epoch": 7.98, "learning_rate": 2.755999055714664e-10, "loss": 0.0004, "step": 43524 }, { "epoch": 7.98, "learning_rate": 2.701154921413984e-10, "loss": 0.0002, "step": 43525 }, { "epoch": 7.98, "learning_rate": 2.6468619745645407e-10, "loss": 0.0007, "step": 43526 }, { "epoch": 7.98, "learning_rate": 2.593120215466094e-10, "loss": 0.0002, "step": 43527 }, { "epoch": 7.98, "learning_rate": 2.539929644418404e-10, "loss": 0.0003, "step": 43528 }, { "epoch": 7.98, "learning_rate": 2.487290261721231e-10, "loss": 0.0002, "step": 43529 }, { "epoch": 7.98, "learning_rate": 2.4352020676521313e-10, "loss": 0.0004, "step": 43530 }, { "epoch": 7.98, "learning_rate": 2.3836650624997627e-10, "loss": 0.0003, "step": 43531 }, { "epoch": 7.98, "learning_rate": 2.3326792465527824e-10, "loss": 0.0003, "step": 43532 }, { "epoch": 7.98, "learning_rate": 2.282244620088747e-10, "loss": 0.0002, "step": 43533 }, { "epoch": 7.98, "learning_rate": 2.2323611833963143e-10, "loss": 0.0003, "step": 43534 }, { "epoch": 7.98, "learning_rate": 2.1830289367308355e-10, "loss": 0.0013, "step": 43535 }, { "epoch": 7.98, "learning_rate": 2.1342478803809686e-10, "loss": 0.0006, "step": 43536 }, { "epoch": 7.98, "learning_rate": 2.086018014613167e-10, "loss": 0.0007, "step": 43537 }, { "epoch": 7.98, "learning_rate": 2.038339339682782e-10, "loss": 0.0002, "step": 43538 }, { "epoch": 7.98, "learning_rate": 1.9912118558673698e-10, "loss": 0.0002, "step": 43539 }, { "epoch": 7.98, "learning_rate": 1.9446355634111791e-10, "loss": 0.0005, "step": 43540 }, { "epoch": 7.98, "learning_rate": 1.8986104625806634e-10, "loss": 0.0004, "step": 43541 }, { "epoch": 7.98, "learning_rate": 1.853136553631174e-10, "loss": 0.0018, "step": 43542 }, { "epoch": 7.99, "learning_rate": 1.80821383680696e-10, "loss": 0.0003, "step": 43543 }, { "epoch": 7.99, "learning_rate": 1.763842312363373e-10, "loss": 0.0009, "step": 43544 }, { "epoch": 7.99, "learning_rate": 1.7200219805335593e-10, "loss": 0.0004, "step": 43545 }, { "epoch": 7.99, "learning_rate": 1.6767528415617684e-10, "loss": 0.0005, "step": 43546 }, { "epoch": 7.99, "learning_rate": 1.6340348957033513e-10, "loss": 0.0009, "step": 43547 }, { "epoch": 7.99, "learning_rate": 1.5918681431692505e-10, "loss": 0.0003, "step": 43548 }, { "epoch": 7.99, "learning_rate": 1.5502525842148176e-10, "loss": 0.0002, "step": 43549 }, { "epoch": 7.99, "learning_rate": 1.5091882190509944e-10, "loss": 0.0004, "step": 43550 }, { "epoch": 7.99, "learning_rate": 1.468675047910928e-10, "loss": 0.0005, "step": 43551 }, { "epoch": 7.99, "learning_rate": 1.428713071016663e-10, "loss": 0.0006, "step": 43552 }, { "epoch": 7.99, "learning_rate": 1.3893022886013464e-10, "loss": 0.0009, "step": 43553 }, { "epoch": 7.99, "learning_rate": 1.3504427008537158e-10, "loss": 0.0004, "step": 43554 }, { "epoch": 7.99, "learning_rate": 1.3121343080180205e-10, "loss": 0.0004, "step": 43555 }, { "epoch": 7.99, "learning_rate": 1.2743771102941005e-10, "loss": 0.0002, "step": 43556 }, { "epoch": 7.99, "learning_rate": 1.237171107881796e-10, "loss": 0.0011, "step": 43557 }, { "epoch": 7.99, "learning_rate": 1.2005163009920495e-10, "loss": 0.0003, "step": 43558 }, { "epoch": 7.99, "learning_rate": 1.164412689824701e-10, "loss": 0.0004, "step": 43559 }, { "epoch": 7.99, "learning_rate": 1.1288602745906929e-10, "loss": 0.0002, "step": 43560 }, { "epoch": 7.99, "learning_rate": 1.0938590554787631e-10, "loss": 0.0005, "step": 43561 }, { "epoch": 7.99, "learning_rate": 1.0594090326665473e-10, "loss": 0.0004, "step": 43562 }, { "epoch": 7.99, "learning_rate": 1.0255102063760902e-10, "loss": 0.0003, "step": 43563 }, { "epoch": 7.99, "learning_rate": 9.921625767628229e-11, "loss": 0.0002, "step": 43564 }, { "epoch": 7.99, "learning_rate": 9.593661440265856e-11, "loss": 0.0008, "step": 43565 }, { "epoch": 7.99, "learning_rate": 9.27120908345014e-11, "loss": 0.0003, "step": 43566 }, { "epoch": 7.99, "learning_rate": 8.954268698957436e-11, "loss": 0.0003, "step": 43567 }, { "epoch": 7.99, "learning_rate": 8.64284028845308e-11, "loss": 0.0002, "step": 43568 }, { "epoch": 7.99, "learning_rate": 8.336923853824453e-11, "loss": 0.0003, "step": 43569 }, { "epoch": 7.99, "learning_rate": 8.036519396736886e-11, "loss": 0.0004, "step": 43570 }, { "epoch": 7.99, "learning_rate": 7.741626918633672e-11, "loss": 0.0005, "step": 43571 }, { "epoch": 7.99, "learning_rate": 7.452246421402187e-11, "loss": 0.0004, "step": 43572 }, { "epoch": 7.99, "learning_rate": 7.168377906485724e-11, "loss": 0.0004, "step": 43573 }, { "epoch": 7.99, "learning_rate": 6.890021375438594e-11, "loss": 0.0005, "step": 43574 }, { "epoch": 7.99, "learning_rate": 6.617176829926131e-11, "loss": 0.0006, "step": 43575 }, { "epoch": 7.99, "learning_rate": 6.349844271280603e-11, "loss": 0.0006, "step": 43576 }, { "epoch": 7.99, "learning_rate": 6.088023701167345e-11, "loss": 0.0004, "step": 43577 }, { "epoch": 7.99, "learning_rate": 5.831715120807602e-11, "loss": 0.0007, "step": 43578 }, { "epoch": 7.99, "learning_rate": 5.580918531755686e-11, "loss": 0.0004, "step": 43579 }, { "epoch": 7.99, "learning_rate": 5.335633935454887e-11, "loss": 0.0003, "step": 43580 }, { "epoch": 7.99, "learning_rate": 5.095861333126451e-11, "loss": 0.0005, "step": 43581 }, { "epoch": 7.99, "learning_rate": 4.8616007261026443e-11, "loss": 0.0007, "step": 43582 }, { "epoch": 7.99, "learning_rate": 4.632852115715736e-11, "loss": 0.0003, "step": 43583 }, { "epoch": 7.99, "learning_rate": 4.409615503297993e-11, "loss": 0.0006, "step": 43584 }, { "epoch": 7.99, "learning_rate": 4.191890889848615e-11, "loss": 0.0009, "step": 43585 }, { "epoch": 7.99, "learning_rate": 3.979678276810894e-11, "loss": 0.0008, "step": 43586 }, { "epoch": 7.99, "learning_rate": 3.772977665295052e-11, "loss": 0.0009, "step": 43587 }, { "epoch": 7.99, "learning_rate": 3.571789056411312e-11, "loss": 0.0006, "step": 43588 }, { "epoch": 7.99, "learning_rate": 3.376112451158875e-11, "loss": 0.0005, "step": 43589 }, { "epoch": 7.99, "learning_rate": 3.185947850870008e-11, "loss": 0.0003, "step": 43590 }, { "epoch": 7.99, "learning_rate": 3.001295256321868e-11, "loss": 0.0008, "step": 43591 }, { "epoch": 7.99, "learning_rate": 2.8221546687356992e-11, "loss": 0.0003, "step": 43592 }, { "epoch": 7.99, "learning_rate": 2.6485260889996813e-11, "loss": 0.0004, "step": 43593 }, { "epoch": 7.99, "learning_rate": 2.480409518113014e-11, "loss": 0.0002, "step": 43594 }, { "epoch": 7.99, "learning_rate": 2.3178049569638762e-11, "loss": 0.0005, "step": 43595 }, { "epoch": 7.99, "learning_rate": 2.160712406440446e-11, "loss": 0.0003, "step": 43596 }, { "epoch": 8.0, "learning_rate": 2.0091318674309024e-11, "loss": 0.0003, "step": 43597 }, { "epoch": 8.0, "learning_rate": 1.8630633408234233e-11, "loss": 0.0006, "step": 43598 }, { "epoch": 8.0, "learning_rate": 1.722506827395165e-11, "loss": 0.0002, "step": 43599 }, { "epoch": 8.0, "learning_rate": 1.5874623278122613e-11, "loss": 0.0003, "step": 43600 }, { "epoch": 8.0, "learning_rate": 1.4579298429628907e-11, "loss": 0.0004, "step": 43601 }, { "epoch": 8.0, "learning_rate": 1.3339093735131869e-11, "loss": 0.0006, "step": 43602 }, { "epoch": 8.0, "learning_rate": 1.2154009201292839e-11, "loss": 0.0011, "step": 43603 }, { "epoch": 8.0, "learning_rate": 1.1024044834773151e-11, "loss": 0.0006, "step": 43604 }, { "epoch": 8.0, "learning_rate": 9.949200641123923e-12, "loss": 0.0006, "step": 43605 }, { "epoch": 8.0, "learning_rate": 8.929476628116717e-12, "loss": 0.0004, "step": 43606 }, { "epoch": 8.0, "learning_rate": 7.964872799082202e-12, "loss": 0.0003, "step": 43607 }, { "epoch": 8.0, "learning_rate": 7.055389160681714e-12, "loss": 0.0003, "step": 43608 }, { "epoch": 8.0, "learning_rate": 6.201025718466369e-12, "loss": 0.0003, "step": 43609 }, { "epoch": 8.0, "learning_rate": 5.401782474656614e-12, "loss": 0.0004, "step": 43610 }, { "epoch": 8.0, "learning_rate": 4.657659437024009e-12, "loss": 0.0006, "step": 43611 }, { "epoch": 8.0, "learning_rate": 3.968656606678778e-12, "loss": 0.0005, "step": 43612 }, { "epoch": 8.0, "learning_rate": 3.3347739902822586e-12, "loss": 0.0007, "step": 43613 }, { "epoch": 8.0, "learning_rate": 2.7560115889446736e-12, "loss": 0.0004, "step": 43614 }, { "epoch": 8.0, "learning_rate": 2.232369405996693e-12, "loss": 0.0002, "step": 43615 }, { "epoch": 8.0, "learning_rate": 1.7638474458792077e-12, "loss": 0.0005, "step": 43616 }, { "epoch": 8.0, "learning_rate": 1.3504457097024416e-12, "loss": 0.0005, "step": 43617 }, { "epoch": 8.0, "learning_rate": 9.921642007970633e-13, "loss": 0.0006, "step": 43618 }, { "epoch": 8.0, "learning_rate": 6.890029213835192e-13, "loss": 0.0002, "step": 43619 }, { "epoch": 8.0, "learning_rate": 4.4096187146180914e-13, "loss": 0.0006, "step": 43620 }, { "epoch": 8.0, "learning_rate": 2.4804105325237917e-13, "loss": 0.0003, "step": 43621 }, { "epoch": 8.0, "learning_rate": 1.1024046786545228e-13, "loss": 0.0002, "step": 43622 }, { "epoch": 8.0, "learning_rate": 2.7560117521474582e-14, "loss": 0.0003, "step": 43623 }, { "epoch": 8.0, "learning_rate": 0.0, "loss": 0.0003, "step": 43624 }, { "epoch": 8.0, "step": 43624, "total_flos": 1.4325804459704608e+20, "train_loss": 0.019492357890355163, "train_runtime": 300283.8796, "train_samples_per_second": 18.594, "train_steps_per_second": 0.145 } ], "logging_steps": 1.0, "max_steps": 43624, "num_input_tokens_seen": 0, "num_train_epochs": 8, "save_steps": 500, "total_flos": 1.4325804459704608e+20, "train_batch_size": 8, "trial_name": null, "trial_params": null }