| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 0.09452689290103035, | |
| "eval_steps": 1000, | |
| "global_step": 500, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 4e-05, | |
| "loss": 2.0452, | |
| "step": 1 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 8e-05, | |
| "loss": 2.1339, | |
| "step": 2 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 0.00012, | |
| "loss": 2.1537, | |
| "step": 3 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 0.00016, | |
| "loss": 1.9245, | |
| "step": 4 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 0.0002, | |
| "loss": 1.6188, | |
| "step": 5 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 0.0001995967741935484, | |
| "loss": 1.9631, | |
| "step": 6 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 0.0001991935483870968, | |
| "loss": 1.8202, | |
| "step": 7 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 0.00019879032258064515, | |
| "loss": 1.7829, | |
| "step": 8 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 0.00019838709677419357, | |
| "loss": 1.6106, | |
| "step": 9 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 0.00019798387096774193, | |
| "loss": 1.8472, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 0.00019758064516129035, | |
| "loss": 1.752, | |
| "step": 11 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 0.0001971774193548387, | |
| "loss": 1.8356, | |
| "step": 12 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 0.0001967741935483871, | |
| "loss": 1.7975, | |
| "step": 13 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 0.0001963709677419355, | |
| "loss": 1.6779, | |
| "step": 14 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 0.00019596774193548388, | |
| "loss": 1.6186, | |
| "step": 15 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 0.00019556451612903227, | |
| "loss": 1.6386, | |
| "step": 16 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 0.00019556451612903227, | |
| "loss": 1.5367, | |
| "step": 17 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 0.00019516129032258066, | |
| "loss": 1.7151, | |
| "step": 18 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 0.00019475806451612905, | |
| "loss": 1.7181, | |
| "step": 19 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 0.00019435483870967744, | |
| "loss": 1.8218, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 0.0001939516129032258, | |
| "loss": 1.3944, | |
| "step": 21 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 0.0001939516129032258, | |
| "loss": 1.5311, | |
| "step": 22 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 0.00019354838709677422, | |
| "loss": 1.8437, | |
| "step": 23 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 0.00019314516129032258, | |
| "loss": 1.7631, | |
| "step": 24 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 0.000192741935483871, | |
| "loss": 1.6397, | |
| "step": 25 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 0.00019233870967741936, | |
| "loss": 1.6021, | |
| "step": 26 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00019193548387096775, | |
| "loss": 1.5021, | |
| "step": 27 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00019153225806451614, | |
| "loss": 1.4953, | |
| "step": 28 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00019112903225806453, | |
| "loss": 1.8371, | |
| "step": 29 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00019072580645161291, | |
| "loss": 1.9104, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0001903225806451613, | |
| "loss": 1.8631, | |
| "step": 31 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0001899193548387097, | |
| "loss": 1.6603, | |
| "step": 32 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00018951612903225808, | |
| "loss": 1.713, | |
| "step": 33 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00018911290322580645, | |
| "loss": 1.6329, | |
| "step": 34 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00018870967741935486, | |
| "loss": 1.5225, | |
| "step": 35 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00018830645161290322, | |
| "loss": 1.7089, | |
| "step": 36 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00018790322580645164, | |
| "loss": 1.497, | |
| "step": 37 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0001875, | |
| "loss": 1.6165, | |
| "step": 38 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0001870967741935484, | |
| "loss": 1.6202, | |
| "step": 39 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00018669354838709678, | |
| "loss": 1.712, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00018629032258064517, | |
| "loss": 1.7133, | |
| "step": 41 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00018588709677419356, | |
| "loss": 1.6171, | |
| "step": 42 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00018548387096774192, | |
| "loss": 1.4604, | |
| "step": 43 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00018508064516129034, | |
| "loss": 1.4851, | |
| "step": 44 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0001846774193548387, | |
| "loss": 1.6546, | |
| "step": 45 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00018427419354838712, | |
| "loss": 1.7893, | |
| "step": 46 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00018387096774193548, | |
| "loss": 1.7062, | |
| "step": 47 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00018346774193548387, | |
| "loss": 1.623, | |
| "step": 48 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00018306451612903226, | |
| "loss": 1.6374, | |
| "step": 49 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00018266129032258065, | |
| "loss": 1.6614, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00018225806451612904, | |
| "loss": 1.4882, | |
| "step": 51 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00018185483870967743, | |
| "loss": 1.6825, | |
| "step": 52 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00018145161290322582, | |
| "loss": 1.7217, | |
| "step": 53 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0001810483870967742, | |
| "loss": 1.6998, | |
| "step": 54 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00018064516129032257, | |
| "loss": 1.501, | |
| "step": 55 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.000180241935483871, | |
| "loss": 1.5068, | |
| "step": 56 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00017983870967741935, | |
| "loss": 1.5879, | |
| "step": 57 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00017943548387096777, | |
| "loss": 1.583, | |
| "step": 58 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00017903225806451613, | |
| "loss": 1.6105, | |
| "step": 59 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00017862903225806452, | |
| "loss": 1.6217, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0001782258064516129, | |
| "loss": 1.5975, | |
| "step": 61 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0001778225806451613, | |
| "loss": 1.4892, | |
| "step": 62 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0001774193548387097, | |
| "loss": 1.6847, | |
| "step": 63 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00017701612903225808, | |
| "loss": 1.6837, | |
| "step": 64 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00017661290322580647, | |
| "loss": 1.6239, | |
| "step": 65 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00017620967741935486, | |
| "loss": 1.7792, | |
| "step": 66 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00017580645161290325, | |
| "loss": 1.7764, | |
| "step": 67 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00017540322580645164, | |
| "loss": 1.5387, | |
| "step": 68 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.000175, | |
| "loss": 1.6544, | |
| "step": 69 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0001745967741935484, | |
| "loss": 1.6512, | |
| "step": 70 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00017419354838709678, | |
| "loss": 1.9079, | |
| "step": 71 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00017379032258064517, | |
| "loss": 1.6991, | |
| "step": 72 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00017338709677419356, | |
| "loss": 1.7934, | |
| "step": 73 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00017298387096774195, | |
| "loss": 1.6651, | |
| "step": 74 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00017258064516129034, | |
| "loss": 1.7393, | |
| "step": 75 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0001721774193548387, | |
| "loss": 1.7071, | |
| "step": 76 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00017177419354838711, | |
| "loss": 1.5009, | |
| "step": 77 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00017137096774193548, | |
| "loss": 1.658, | |
| "step": 78 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0001709677419354839, | |
| "loss": 1.6428, | |
| "step": 79 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00017056451612903226, | |
| "loss": 1.7405, | |
| "step": 80 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00017016129032258065, | |
| "loss": 1.7162, | |
| "step": 81 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00016975806451612904, | |
| "loss": 1.6236, | |
| "step": 82 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00016935483870967742, | |
| "loss": 1.5788, | |
| "step": 83 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00016895161290322581, | |
| "loss": 1.5383, | |
| "step": 84 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0001685483870967742, | |
| "loss": 1.8091, | |
| "step": 85 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0001681451612903226, | |
| "loss": 1.4177, | |
| "step": 86 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00016774193548387098, | |
| "loss": 1.6378, | |
| "step": 87 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00016733870967741935, | |
| "loss": 1.5325, | |
| "step": 88 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00016693548387096776, | |
| "loss": 1.4395, | |
| "step": 89 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00016653225806451612, | |
| "loss": 1.577, | |
| "step": 90 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00016612903225806454, | |
| "loss": 1.7508, | |
| "step": 91 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0001657258064516129, | |
| "loss": 1.5263, | |
| "step": 92 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00016532258064516132, | |
| "loss": 1.6134, | |
| "step": 93 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00016491935483870968, | |
| "loss": 1.9328, | |
| "step": 94 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00016451612903225807, | |
| "loss": 1.7819, | |
| "step": 95 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00016411290322580646, | |
| "loss": 1.8635, | |
| "step": 96 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00016370967741935485, | |
| "loss": 1.4804, | |
| "step": 97 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00016330645161290324, | |
| "loss": 1.6171, | |
| "step": 98 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00016290322580645163, | |
| "loss": 1.6722, | |
| "step": 99 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00016250000000000002, | |
| "loss": 1.5711, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00016209677419354838, | |
| "loss": 1.6117, | |
| "step": 101 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00016169354838709677, | |
| "loss": 1.6901, | |
| "step": 102 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00016129032258064516, | |
| "loss": 1.8487, | |
| "step": 103 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00016088709677419355, | |
| "loss": 1.6705, | |
| "step": 104 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00016048387096774194, | |
| "loss": 1.5581, | |
| "step": 105 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00016008064516129033, | |
| "loss": 1.4694, | |
| "step": 106 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00015967741935483872, | |
| "loss": 1.4355, | |
| "step": 107 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0001592741935483871, | |
| "loss": 1.4237, | |
| "step": 108 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00015887096774193547, | |
| "loss": 1.6782, | |
| "step": 109 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0001584677419354839, | |
| "loss": 1.4701, | |
| "step": 110 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00015806451612903225, | |
| "loss": 1.7806, | |
| "step": 111 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00015766129032258067, | |
| "loss": 1.5857, | |
| "step": 112 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00015725806451612903, | |
| "loss": 1.7187, | |
| "step": 113 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00015685483870967742, | |
| "loss": 1.6985, | |
| "step": 114 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0001564516129032258, | |
| "loss": 1.809, | |
| "step": 115 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0001560483870967742, | |
| "loss": 1.6915, | |
| "step": 116 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0001556451612903226, | |
| "loss": 1.6854, | |
| "step": 117 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00015524193548387098, | |
| "loss": 1.6484, | |
| "step": 118 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00015483870967741937, | |
| "loss": 1.6628, | |
| "step": 119 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00015443548387096776, | |
| "loss": 1.6994, | |
| "step": 120 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00015403225806451615, | |
| "loss": 1.7657, | |
| "step": 121 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00015362903225806454, | |
| "loss": 1.5611, | |
| "step": 122 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0001532258064516129, | |
| "loss": 1.5338, | |
| "step": 123 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00015282258064516131, | |
| "loss": 1.7617, | |
| "step": 124 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00015241935483870968, | |
| "loss": 1.67, | |
| "step": 125 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0001520161290322581, | |
| "loss": 1.7283, | |
| "step": 126 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00015161290322580646, | |
| "loss": 1.7582, | |
| "step": 127 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00015120967741935485, | |
| "loss": 1.3957, | |
| "step": 128 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00015080645161290323, | |
| "loss": 1.6221, | |
| "step": 129 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00015040322580645162, | |
| "loss": 1.7759, | |
| "step": 130 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00015000000000000001, | |
| "loss": 1.7331, | |
| "step": 131 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00014959677419354838, | |
| "loss": 1.6689, | |
| "step": 132 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0001491935483870968, | |
| "loss": 1.5734, | |
| "step": 133 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00014879032258064516, | |
| "loss": 1.6629, | |
| "step": 134 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00014838709677419355, | |
| "loss": 1.6368, | |
| "step": 135 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00014798387096774193, | |
| "loss": 1.742, | |
| "step": 136 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00014758064516129032, | |
| "loss": 1.4901, | |
| "step": 137 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00014717741935483871, | |
| "loss": 1.6417, | |
| "step": 138 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0001467741935483871, | |
| "loss": 1.6693, | |
| "step": 139 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0001463709677419355, | |
| "loss": 1.588, | |
| "step": 140 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00014596774193548388, | |
| "loss": 1.6614, | |
| "step": 141 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00014556451612903224, | |
| "loss": 1.7408, | |
| "step": 142 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00014516129032258066, | |
| "loss": 1.7049, | |
| "step": 143 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00014475806451612902, | |
| "loss": 1.2828, | |
| "step": 144 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00014435483870967744, | |
| "loss": 1.7891, | |
| "step": 145 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0001439516129032258, | |
| "loss": 1.7558, | |
| "step": 146 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00014354838709677422, | |
| "loss": 1.8506, | |
| "step": 147 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00014314516129032258, | |
| "loss": 1.7095, | |
| "step": 148 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00014274193548387097, | |
| "loss": 1.7061, | |
| "step": 149 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00014233870967741936, | |
| "loss": 1.4998, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00014193548387096775, | |
| "loss": 1.6332, | |
| "step": 151 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00014153225806451614, | |
| "loss": 1.7242, | |
| "step": 152 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00014112903225806453, | |
| "loss": 1.9499, | |
| "step": 153 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00014072580645161292, | |
| "loss": 1.6425, | |
| "step": 154 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0001403225806451613, | |
| "loss": 1.7308, | |
| "step": 155 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00013991935483870967, | |
| "loss": 1.8306, | |
| "step": 156 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0001395161290322581, | |
| "loss": 1.6522, | |
| "step": 157 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00013911290322580645, | |
| "loss": 1.432, | |
| "step": 158 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00013870967741935487, | |
| "loss": 1.3832, | |
| "step": 159 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00013830645161290323, | |
| "loss": 1.4807, | |
| "step": 160 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00013790322580645162, | |
| "loss": 1.4447, | |
| "step": 161 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0001375, | |
| "loss": 1.6328, | |
| "step": 162 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00013709677419354837, | |
| "loss": 1.5871, | |
| "step": 163 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0001366935483870968, | |
| "loss": 1.5617, | |
| "step": 164 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00013629032258064515, | |
| "loss": 1.5104, | |
| "step": 165 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00013588709677419357, | |
| "loss": 1.7455, | |
| "step": 166 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00013548387096774193, | |
| "loss": 1.5425, | |
| "step": 167 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00013508064516129032, | |
| "loss": 1.5297, | |
| "step": 168 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0001346774193548387, | |
| "loss": 1.6117, | |
| "step": 169 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0001342741935483871, | |
| "loss": 1.4543, | |
| "step": 170 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0001338709677419355, | |
| "loss": 1.6787, | |
| "step": 171 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00013346774193548388, | |
| "loss": 1.4363, | |
| "step": 172 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00013306451612903227, | |
| "loss": 1.7276, | |
| "step": 173 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00013266129032258066, | |
| "loss": 1.6509, | |
| "step": 174 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00013225806451612905, | |
| "loss": 1.7243, | |
| "step": 175 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00013185483870967743, | |
| "loss": 1.6976, | |
| "step": 176 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0001314516129032258, | |
| "loss": 1.3966, | |
| "step": 177 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00013104838709677421, | |
| "loss": 1.5175, | |
| "step": 178 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00013064516129032258, | |
| "loss": 1.6467, | |
| "step": 179 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.000130241935483871, | |
| "loss": 1.449, | |
| "step": 180 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00012983870967741936, | |
| "loss": 1.6027, | |
| "step": 181 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00012943548387096774, | |
| "loss": 1.596, | |
| "step": 182 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00012903225806451613, | |
| "loss": 1.5356, | |
| "step": 183 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00012862903225806452, | |
| "loss": 1.4891, | |
| "step": 184 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0001282258064516129, | |
| "loss": 1.7449, | |
| "step": 185 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0001278225806451613, | |
| "loss": 1.5996, | |
| "step": 186 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0001274193548387097, | |
| "loss": 1.5612, | |
| "step": 187 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00012701612903225808, | |
| "loss": 1.752, | |
| "step": 188 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00012661290322580644, | |
| "loss": 1.4981, | |
| "step": 189 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00012620967741935486, | |
| "loss": 1.6575, | |
| "step": 190 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00012580645161290322, | |
| "loss": 1.6475, | |
| "step": 191 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00012540322580645164, | |
| "loss": 1.7674, | |
| "step": 192 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.000125, | |
| "loss": 1.6106, | |
| "step": 193 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0001245967741935484, | |
| "loss": 1.6557, | |
| "step": 194 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00012419354838709678, | |
| "loss": 1.6838, | |
| "step": 195 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00012379032258064514, | |
| "loss": 1.5761, | |
| "step": 196 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00012338709677419356, | |
| "loss": 1.6403, | |
| "step": 197 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00012298387096774192, | |
| "loss": 1.5898, | |
| "step": 198 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00012258064516129034, | |
| "loss": 1.7031, | |
| "step": 199 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0001221774193548387, | |
| "loss": 1.6444, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0001217741935483871, | |
| "loss": 1.6056, | |
| "step": 201 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00012137096774193548, | |
| "loss": 1.7643, | |
| "step": 202 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00012096774193548388, | |
| "loss": 1.6614, | |
| "step": 203 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00012056451612903226, | |
| "loss": 1.4906, | |
| "step": 204 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00012016129032258065, | |
| "loss": 1.598, | |
| "step": 205 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00011975806451612903, | |
| "loss": 1.6326, | |
| "step": 206 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00011935483870967743, | |
| "loss": 1.6153, | |
| "step": 207 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0001189516129032258, | |
| "loss": 1.6332, | |
| "step": 208 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00011854838709677421, | |
| "loss": 1.6477, | |
| "step": 209 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00011814516129032258, | |
| "loss": 1.6652, | |
| "step": 210 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00011774193548387097, | |
| "loss": 1.412, | |
| "step": 211 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00011733870967741935, | |
| "loss": 1.5335, | |
| "step": 212 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00011693548387096775, | |
| "loss": 1.6941, | |
| "step": 213 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00011653225806451613, | |
| "loss": 1.5277, | |
| "step": 214 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00011612903225806453, | |
| "loss": 1.5259, | |
| "step": 215 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00011572580645161291, | |
| "loss": 1.8913, | |
| "step": 216 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00011532258064516131, | |
| "loss": 1.5655, | |
| "step": 217 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00011491935483870969, | |
| "loss": 1.4732, | |
| "step": 218 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00011451612903225808, | |
| "loss": 1.6836, | |
| "step": 219 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00011411290322580645, | |
| "loss": 1.7453, | |
| "step": 220 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00011370967741935486, | |
| "loss": 1.801, | |
| "step": 221 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00011330645161290323, | |
| "loss": 1.6863, | |
| "step": 222 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00011290322580645163, | |
| "loss": 1.7664, | |
| "step": 223 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00011250000000000001, | |
| "loss": 1.7441, | |
| "step": 224 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00011209677419354839, | |
| "loss": 1.7277, | |
| "step": 225 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00011169354838709678, | |
| "loss": 1.5249, | |
| "step": 226 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00011129032258064515, | |
| "loss": 1.5534, | |
| "step": 227 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00011088709677419356, | |
| "loss": 1.6429, | |
| "step": 228 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00011048387096774193, | |
| "loss": 1.5998, | |
| "step": 229 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00011008064516129033, | |
| "loss": 1.7443, | |
| "step": 230 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00010967741935483871, | |
| "loss": 1.6395, | |
| "step": 231 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0001092741935483871, | |
| "loss": 1.7052, | |
| "step": 232 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00010887096774193548, | |
| "loss": 1.4872, | |
| "step": 233 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00010846774193548388, | |
| "loss": 1.9156, | |
| "step": 234 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00010806451612903225, | |
| "loss": 1.5576, | |
| "step": 235 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00010766129032258066, | |
| "loss": 1.5047, | |
| "step": 236 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00010725806451612903, | |
| "loss": 1.5351, | |
| "step": 237 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00010685483870967742, | |
| "loss": 1.4936, | |
| "step": 238 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0001064516129032258, | |
| "loss": 1.4739, | |
| "step": 239 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0001060483870967742, | |
| "loss": 1.7144, | |
| "step": 240 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.00010564516129032258, | |
| "loss": 1.4508, | |
| "step": 241 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.00010524193548387098, | |
| "loss": 1.5671, | |
| "step": 242 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.00010483870967741936, | |
| "loss": 1.6563, | |
| "step": 243 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.00010443548387096776, | |
| "loss": 1.4779, | |
| "step": 244 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.00010403225806451614, | |
| "loss": 1.6882, | |
| "step": 245 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.00010362903225806453, | |
| "loss": 1.2259, | |
| "step": 246 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0001032258064516129, | |
| "loss": 1.4727, | |
| "step": 247 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0001028225806451613, | |
| "loss": 1.4654, | |
| "step": 248 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.00010241935483870968, | |
| "loss": 1.6736, | |
| "step": 249 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.00010201612903225808, | |
| "loss": 1.5656, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.00010161290322580646, | |
| "loss": 1.6925, | |
| "step": 251 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.00010120967741935485, | |
| "loss": 1.5055, | |
| "step": 252 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.00010080645161290323, | |
| "loss": 1.5479, | |
| "step": 253 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.00010040322580645163, | |
| "loss": 1.6352, | |
| "step": 254 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0001, | |
| "loss": 1.6747, | |
| "step": 255 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.95967741935484e-05, | |
| "loss": 1.7197, | |
| "step": 256 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.919354838709678e-05, | |
| "loss": 1.5599, | |
| "step": 257 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.879032258064517e-05, | |
| "loss": 1.4343, | |
| "step": 258 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.838709677419355e-05, | |
| "loss": 1.6037, | |
| "step": 259 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.798387096774194e-05, | |
| "loss": 1.6565, | |
| "step": 260 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.758064516129033e-05, | |
| "loss": 1.6615, | |
| "step": 261 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.717741935483872e-05, | |
| "loss": 1.6193, | |
| "step": 262 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.677419354838711e-05, | |
| "loss": 1.5369, | |
| "step": 263 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.63709677419355e-05, | |
| "loss": 1.7713, | |
| "step": 264 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.596774193548387e-05, | |
| "loss": 1.6298, | |
| "step": 265 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.556451612903226e-05, | |
| "loss": 1.4293, | |
| "step": 266 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.516129032258065e-05, | |
| "loss": 1.5491, | |
| "step": 267 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.475806451612904e-05, | |
| "loss": 1.6733, | |
| "step": 268 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.435483870967743e-05, | |
| "loss": 1.6588, | |
| "step": 269 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.395161290322582e-05, | |
| "loss": 1.5837, | |
| "step": 270 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.35483870967742e-05, | |
| "loss": 1.3857, | |
| "step": 271 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.314516129032259e-05, | |
| "loss": 1.6731, | |
| "step": 272 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.274193548387096e-05, | |
| "loss": 1.5275, | |
| "step": 273 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.233870967741935e-05, | |
| "loss": 1.5816, | |
| "step": 274 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.193548387096774e-05, | |
| "loss": 1.7432, | |
| "step": 275 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.153225806451613e-05, | |
| "loss": 1.3871, | |
| "step": 276 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.112903225806452e-05, | |
| "loss": 1.4472, | |
| "step": 277 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.072580645161291e-05, | |
| "loss": 1.6518, | |
| "step": 278 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.032258064516129e-05, | |
| "loss": 1.5679, | |
| "step": 279 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 8.991935483870968e-05, | |
| "loss": 1.5579, | |
| "step": 280 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 8.951612903225806e-05, | |
| "loss": 1.7382, | |
| "step": 281 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 8.911290322580645e-05, | |
| "loss": 1.6818, | |
| "step": 282 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 8.870967741935484e-05, | |
| "loss": 1.8847, | |
| "step": 283 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 8.830645161290323e-05, | |
| "loss": 1.7741, | |
| "step": 284 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 8.790322580645162e-05, | |
| "loss": 1.3626, | |
| "step": 285 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 8.75e-05, | |
| "loss": 1.7136, | |
| "step": 286 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 8.709677419354839e-05, | |
| "loss": 1.5743, | |
| "step": 287 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 8.669354838709678e-05, | |
| "loss": 1.6282, | |
| "step": 288 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 8.629032258064517e-05, | |
| "loss": 1.6719, | |
| "step": 289 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 8.588709677419356e-05, | |
| "loss": 1.7067, | |
| "step": 290 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 8.548387096774195e-05, | |
| "loss": 1.5314, | |
| "step": 291 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 8.508064516129032e-05, | |
| "loss": 1.5345, | |
| "step": 292 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 8.467741935483871e-05, | |
| "loss": 1.7922, | |
| "step": 293 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 8.42741935483871e-05, | |
| "loss": 1.4014, | |
| "step": 294 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 8.387096774193549e-05, | |
| "loss": 1.7632, | |
| "step": 295 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 8.346774193548388e-05, | |
| "loss": 1.8397, | |
| "step": 296 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 8.306451612903227e-05, | |
| "loss": 1.6557, | |
| "step": 297 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 8.266129032258066e-05, | |
| "loss": 1.4892, | |
| "step": 298 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 8.225806451612904e-05, | |
| "loss": 1.6108, | |
| "step": 299 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 8.185483870967743e-05, | |
| "loss": 1.6695, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 8.145161290322582e-05, | |
| "loss": 1.6221, | |
| "step": 301 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 8.104838709677419e-05, | |
| "loss": 1.5938, | |
| "step": 302 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 8.064516129032258e-05, | |
| "loss": 1.7077, | |
| "step": 303 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 8.024193548387097e-05, | |
| "loss": 1.3318, | |
| "step": 304 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 7.983870967741936e-05, | |
| "loss": 1.4808, | |
| "step": 305 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 7.943548387096774e-05, | |
| "loss": 1.6729, | |
| "step": 306 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 7.903225806451613e-05, | |
| "loss": 1.6133, | |
| "step": 307 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 7.862903225806451e-05, | |
| "loss": 1.6705, | |
| "step": 308 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 7.82258064516129e-05, | |
| "loss": 1.5732, | |
| "step": 309 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 7.78225806451613e-05, | |
| "loss": 1.3341, | |
| "step": 310 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 7.741935483870968e-05, | |
| "loss": 1.7081, | |
| "step": 311 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 7.701612903225807e-05, | |
| "loss": 1.6325, | |
| "step": 312 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 7.661290322580645e-05, | |
| "loss": 1.5734, | |
| "step": 313 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 7.620967741935484e-05, | |
| "loss": 1.5356, | |
| "step": 314 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 7.580645161290323e-05, | |
| "loss": 1.6806, | |
| "step": 315 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 7.540322580645162e-05, | |
| "loss": 1.5687, | |
| "step": 316 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 7.500000000000001e-05, | |
| "loss": 1.6411, | |
| "step": 317 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 7.45967741935484e-05, | |
| "loss": 1.6724, | |
| "step": 318 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 7.419354838709677e-05, | |
| "loss": 1.5039, | |
| "step": 319 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 7.379032258064516e-05, | |
| "loss": 1.6331, | |
| "step": 320 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 7.338709677419355e-05, | |
| "loss": 1.629, | |
| "step": 321 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 7.298387096774194e-05, | |
| "loss": 1.6863, | |
| "step": 322 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 7.258064516129033e-05, | |
| "loss": 1.5211, | |
| "step": 323 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 7.217741935483872e-05, | |
| "loss": 1.6081, | |
| "step": 324 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 7.177419354838711e-05, | |
| "loss": 1.6189, | |
| "step": 325 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 7.137096774193549e-05, | |
| "loss": 1.8124, | |
| "step": 326 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 7.096774193548388e-05, | |
| "loss": 1.7381, | |
| "step": 327 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 7.056451612903226e-05, | |
| "loss": 1.7527, | |
| "step": 328 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 7.016129032258065e-05, | |
| "loss": 1.6601, | |
| "step": 329 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 6.975806451612904e-05, | |
| "loss": 1.55, | |
| "step": 330 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 6.935483870967743e-05, | |
| "loss": 1.7738, | |
| "step": 331 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 6.895161290322581e-05, | |
| "loss": 1.5447, | |
| "step": 332 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 6.854838709677419e-05, | |
| "loss": 1.5738, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 6.814516129032257e-05, | |
| "loss": 1.6678, | |
| "step": 334 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 6.774193548387096e-05, | |
| "loss": 1.8589, | |
| "step": 335 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 6.733870967741935e-05, | |
| "loss": 1.6991, | |
| "step": 336 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 6.693548387096774e-05, | |
| "loss": 1.5882, | |
| "step": 337 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 6.653225806451613e-05, | |
| "loss": 1.6036, | |
| "step": 338 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 6.612903225806452e-05, | |
| "loss": 1.6684, | |
| "step": 339 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 6.57258064516129e-05, | |
| "loss": 1.4674, | |
| "step": 340 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 6.532258064516129e-05, | |
| "loss": 1.7046, | |
| "step": 341 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 6.491935483870968e-05, | |
| "loss": 1.7792, | |
| "step": 342 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 6.451612903225807e-05, | |
| "loss": 1.5975, | |
| "step": 343 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 6.411290322580646e-05, | |
| "loss": 1.4108, | |
| "step": 344 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 6.370967741935485e-05, | |
| "loss": 1.6642, | |
| "step": 345 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 6.330645161290322e-05, | |
| "loss": 1.608, | |
| "step": 346 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 6.290322580645161e-05, | |
| "loss": 1.6593, | |
| "step": 347 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 6.25e-05, | |
| "loss": 1.6623, | |
| "step": 348 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 6.209677419354839e-05, | |
| "loss": 1.6213, | |
| "step": 349 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 6.169354838709678e-05, | |
| "loss": 1.6196, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 6.129032258064517e-05, | |
| "loss": 1.5282, | |
| "step": 351 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 6.088709677419355e-05, | |
| "loss": 1.3838, | |
| "step": 352 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 6.048387096774194e-05, | |
| "loss": 1.6048, | |
| "step": 353 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 6.0080645161290325e-05, | |
| "loss": 1.6077, | |
| "step": 354 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 5.9677419354838715e-05, | |
| "loss": 1.4637, | |
| "step": 355 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 5.9274193548387104e-05, | |
| "loss": 1.7545, | |
| "step": 356 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 5.887096774193549e-05, | |
| "loss": 1.7271, | |
| "step": 357 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 5.8467741935483876e-05, | |
| "loss": 1.5428, | |
| "step": 358 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 5.8064516129032266e-05, | |
| "loss": 1.5648, | |
| "step": 359 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 5.7661290322580655e-05, | |
| "loss": 1.6512, | |
| "step": 360 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 5.725806451612904e-05, | |
| "loss": 1.644, | |
| "step": 361 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 5.685483870967743e-05, | |
| "loss": 1.7342, | |
| "step": 362 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 5.645161290322582e-05, | |
| "loss": 1.82, | |
| "step": 363 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 5.604838709677419e-05, | |
| "loss": 1.6958, | |
| "step": 364 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 5.5645161290322576e-05, | |
| "loss": 1.7386, | |
| "step": 365 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 5.5241935483870966e-05, | |
| "loss": 1.46, | |
| "step": 366 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 5.4838709677419355e-05, | |
| "loss": 1.4443, | |
| "step": 367 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 5.443548387096774e-05, | |
| "loss": 1.5522, | |
| "step": 368 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 5.403225806451613e-05, | |
| "loss": 1.6077, | |
| "step": 369 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 5.362903225806452e-05, | |
| "loss": 1.6593, | |
| "step": 370 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 5.32258064516129e-05, | |
| "loss": 1.5183, | |
| "step": 371 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 5.282258064516129e-05, | |
| "loss": 1.5314, | |
| "step": 372 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 5.241935483870968e-05, | |
| "loss": 1.589, | |
| "step": 373 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 5.201612903225807e-05, | |
| "loss": 1.5632, | |
| "step": 374 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 5.161290322580645e-05, | |
| "loss": 1.6664, | |
| "step": 375 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 5.120967741935484e-05, | |
| "loss": 1.6375, | |
| "step": 376 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 5.080645161290323e-05, | |
| "loss": 1.5933, | |
| "step": 377 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 5.040322580645161e-05, | |
| "loss": 1.6126, | |
| "step": 378 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 5e-05, | |
| "loss": 1.645, | |
| "step": 379 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 4.959677419354839e-05, | |
| "loss": 1.5649, | |
| "step": 380 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 4.9193548387096775e-05, | |
| "loss": 1.5601, | |
| "step": 381 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 4.8790322580645164e-05, | |
| "loss": 1.6028, | |
| "step": 382 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 4.8387096774193554e-05, | |
| "loss": 1.675, | |
| "step": 383 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 4.7983870967741937e-05, | |
| "loss": 1.5058, | |
| "step": 384 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 4.7580645161290326e-05, | |
| "loss": 1.7313, | |
| "step": 385 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 4.7177419354838716e-05, | |
| "loss": 1.609, | |
| "step": 386 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 4.67741935483871e-05, | |
| "loss": 1.5993, | |
| "step": 387 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 4.637096774193548e-05, | |
| "loss": 1.7103, | |
| "step": 388 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 4.596774193548387e-05, | |
| "loss": 1.6556, | |
| "step": 389 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 4.556451612903226e-05, | |
| "loss": 1.6981, | |
| "step": 390 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 4.516129032258064e-05, | |
| "loss": 1.6413, | |
| "step": 391 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 4.475806451612903e-05, | |
| "loss": 1.5337, | |
| "step": 392 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 4.435483870967742e-05, | |
| "loss": 1.7622, | |
| "step": 393 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 4.395161290322581e-05, | |
| "loss": 1.6373, | |
| "step": 394 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 4.3548387096774194e-05, | |
| "loss": 1.7476, | |
| "step": 395 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 4.3145161290322584e-05, | |
| "loss": 1.7093, | |
| "step": 396 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 4.2741935483870973e-05, | |
| "loss": 1.225, | |
| "step": 397 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 4.2338709677419356e-05, | |
| "loss": 1.8272, | |
| "step": 398 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 4.1935483870967746e-05, | |
| "loss": 1.6983, | |
| "step": 399 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 4.1532258064516135e-05, | |
| "loss": 1.5777, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 4.112903225806452e-05, | |
| "loss": 1.549, | |
| "step": 401 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 4.072580645161291e-05, | |
| "loss": 1.7277, | |
| "step": 402 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 4.032258064516129e-05, | |
| "loss": 1.447, | |
| "step": 403 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 3.991935483870968e-05, | |
| "loss": 1.4048, | |
| "step": 404 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 3.951612903225806e-05, | |
| "loss": 1.4329, | |
| "step": 405 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 3.911290322580645e-05, | |
| "loss": 1.6236, | |
| "step": 406 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 3.870967741935484e-05, | |
| "loss": 1.6898, | |
| "step": 407 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 3.8306451612903224e-05, | |
| "loss": 1.6663, | |
| "step": 408 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 3.7903225806451614e-05, | |
| "loss": 1.8015, | |
| "step": 409 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 3.7500000000000003e-05, | |
| "loss": 1.74, | |
| "step": 410 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 3.7096774193548386e-05, | |
| "loss": 1.7129, | |
| "step": 411 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 3.6693548387096776e-05, | |
| "loss": 1.8136, | |
| "step": 412 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 3.6290322580645165e-05, | |
| "loss": 1.805, | |
| "step": 413 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 3.5887096774193555e-05, | |
| "loss": 1.5013, | |
| "step": 414 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 3.548387096774194e-05, | |
| "loss": 1.6694, | |
| "step": 415 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 3.508064516129033e-05, | |
| "loss": 1.6166, | |
| "step": 416 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 3.467741935483872e-05, | |
| "loss": 1.7731, | |
| "step": 417 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 3.427419354838709e-05, | |
| "loss": 1.5846, | |
| "step": 418 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 3.387096774193548e-05, | |
| "loss": 1.6882, | |
| "step": 419 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 3.346774193548387e-05, | |
| "loss": 1.585, | |
| "step": 420 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 3.306451612903226e-05, | |
| "loss": 1.5087, | |
| "step": 421 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 3.2661290322580644e-05, | |
| "loss": 1.726, | |
| "step": 422 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 3.2258064516129034e-05, | |
| "loss": 1.6284, | |
| "step": 423 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 3.185483870967742e-05, | |
| "loss": 1.4476, | |
| "step": 424 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 3.1451612903225806e-05, | |
| "loss": 1.6885, | |
| "step": 425 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 3.1048387096774195e-05, | |
| "loss": 1.7086, | |
| "step": 426 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 3.0645161290322585e-05, | |
| "loss": 1.6276, | |
| "step": 427 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 3.024193548387097e-05, | |
| "loss": 1.4247, | |
| "step": 428 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 2.9838709677419357e-05, | |
| "loss": 1.5147, | |
| "step": 429 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 2.9435483870967743e-05, | |
| "loss": 1.5369, | |
| "step": 430 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 2.9032258064516133e-05, | |
| "loss": 1.6678, | |
| "step": 431 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 2.862903225806452e-05, | |
| "loss": 1.6508, | |
| "step": 432 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 2.822580645161291e-05, | |
| "loss": 1.7007, | |
| "step": 433 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 2.7822580645161288e-05, | |
| "loss": 1.5613, | |
| "step": 434 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 2.7419354838709678e-05, | |
| "loss": 1.6609, | |
| "step": 435 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 2.7016129032258064e-05, | |
| "loss": 1.6067, | |
| "step": 436 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 2.661290322580645e-05, | |
| "loss": 1.6103, | |
| "step": 437 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 2.620967741935484e-05, | |
| "loss": 1.8243, | |
| "step": 438 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 2.5806451612903226e-05, | |
| "loss": 1.744, | |
| "step": 439 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 2.5403225806451615e-05, | |
| "loss": 1.7083, | |
| "step": 440 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 2.5e-05, | |
| "loss": 1.4637, | |
| "step": 441 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 2.4596774193548387e-05, | |
| "loss": 1.7277, | |
| "step": 442 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 2.4193548387096777e-05, | |
| "loss": 1.7535, | |
| "step": 443 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 2.3790322580645163e-05, | |
| "loss": 1.713, | |
| "step": 444 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 2.338709677419355e-05, | |
| "loss": 1.6814, | |
| "step": 445 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 2.2983870967741935e-05, | |
| "loss": 1.6429, | |
| "step": 446 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 2.258064516129032e-05, | |
| "loss": 1.6736, | |
| "step": 447 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 2.217741935483871e-05, | |
| "loss": 1.5785, | |
| "step": 448 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 2.1774193548387097e-05, | |
| "loss": 1.7577, | |
| "step": 449 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 2.1370967741935487e-05, | |
| "loss": 1.5823, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 2.0967741935483873e-05, | |
| "loss": 1.502, | |
| "step": 451 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 2.056451612903226e-05, | |
| "loss": 1.4476, | |
| "step": 452 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 2.0161290322580645e-05, | |
| "loss": 1.3888, | |
| "step": 453 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.975806451612903e-05, | |
| "loss": 1.4592, | |
| "step": 454 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.935483870967742e-05, | |
| "loss": 1.489, | |
| "step": 455 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.8951612903225807e-05, | |
| "loss": 1.6278, | |
| "step": 456 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.8548387096774193e-05, | |
| "loss": 1.7016, | |
| "step": 457 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.8145161290322583e-05, | |
| "loss": 1.622, | |
| "step": 458 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.774193548387097e-05, | |
| "loss": 1.5188, | |
| "step": 459 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.733870967741936e-05, | |
| "loss": 1.4665, | |
| "step": 460 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.693548387096774e-05, | |
| "loss": 1.6974, | |
| "step": 461 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.653225806451613e-05, | |
| "loss": 1.6466, | |
| "step": 462 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.6129032258064517e-05, | |
| "loss": 1.6057, | |
| "step": 463 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.5725806451612903e-05, | |
| "loss": 1.489, | |
| "step": 464 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.5322580645161292e-05, | |
| "loss": 1.7375, | |
| "step": 465 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.4919354838709679e-05, | |
| "loss": 1.5686, | |
| "step": 466 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.4516129032258066e-05, | |
| "loss": 1.382, | |
| "step": 467 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.4112903225806454e-05, | |
| "loss": 1.7037, | |
| "step": 468 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.3709677419354839e-05, | |
| "loss": 1.5521, | |
| "step": 469 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.3306451612903225e-05, | |
| "loss": 1.3172, | |
| "step": 470 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.2903225806451613e-05, | |
| "loss": 1.4997, | |
| "step": 471 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.25e-05, | |
| "loss": 1.7076, | |
| "step": 472 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.2096774193548388e-05, | |
| "loss": 1.432, | |
| "step": 473 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.1693548387096775e-05, | |
| "loss": 1.8895, | |
| "step": 474 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.129032258064516e-05, | |
| "loss": 1.536, | |
| "step": 475 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.0887096774193549e-05, | |
| "loss": 1.7395, | |
| "step": 476 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.0483870967741936e-05, | |
| "loss": 1.4813, | |
| "step": 477 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.0080645161290323e-05, | |
| "loss": 1.4942, | |
| "step": 478 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 9.67741935483871e-06, | |
| "loss": 1.5411, | |
| "step": 479 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 9.274193548387097e-06, | |
| "loss": 1.6977, | |
| "step": 480 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 8.870967741935484e-06, | |
| "loss": 1.5042, | |
| "step": 481 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 8.46774193548387e-06, | |
| "loss": 1.9896, | |
| "step": 482 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 8.064516129032258e-06, | |
| "loss": 1.4309, | |
| "step": 483 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 7.661290322580646e-06, | |
| "loss": 1.6371, | |
| "step": 484 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 7.258064516129033e-06, | |
| "loss": 1.3355, | |
| "step": 485 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 6.854838709677419e-06, | |
| "loss": 1.7125, | |
| "step": 486 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 6.451612903225806e-06, | |
| "loss": 1.5899, | |
| "step": 487 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 6.048387096774194e-06, | |
| "loss": 1.3346, | |
| "step": 488 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 5.64516129032258e-06, | |
| "loss": 1.5943, | |
| "step": 489 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 5.241935483870968e-06, | |
| "loss": 1.6136, | |
| "step": 490 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 4.838709677419355e-06, | |
| "loss": 1.8624, | |
| "step": 491 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 4.435483870967742e-06, | |
| "loss": 1.5963, | |
| "step": 492 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 4.032258064516129e-06, | |
| "loss": 1.5722, | |
| "step": 493 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 3.6290322580645166e-06, | |
| "loss": 1.7267, | |
| "step": 494 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 3.225806451612903e-06, | |
| "loss": 1.4455, | |
| "step": 495 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 2.82258064516129e-06, | |
| "loss": 1.6525, | |
| "step": 496 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 2.4193548387096776e-06, | |
| "loss": 1.5588, | |
| "step": 497 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 2.0161290322580646e-06, | |
| "loss": 1.7389, | |
| "step": 498 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.6129032258064516e-06, | |
| "loss": 1.4058, | |
| "step": 499 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.2096774193548388e-06, | |
| "loss": 1.518, | |
| "step": 500 | |
| } | |
| ], | |
| "logging_steps": 1, | |
| "max_steps": 501, | |
| "num_train_epochs": 1, | |
| "save_steps": 25, | |
| "total_flos": 3.0694656581369856e+16, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |