| { | |
| "best_metric": 48.83268482490272, | |
| "best_model_checkpoint": "/netscratch/duynguyen/Research/Nghiem_LLaVA-Med/LVLM-Med/weights_finetuned/CoT-100_exgra_med_on_CoT_dataset_epochs9_batchsize32_lr2e-5_prompt_mode_simple/checkpoint-1515", | |
| "epoch": 9.0, | |
| "global_step": 1521, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 4.347826086956522e-07, | |
| "loss": 1.0365, | |
| "step": 1 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 8.695652173913044e-07, | |
| "loss": 1.0558, | |
| "step": 2 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.3043478260869566e-06, | |
| "loss": 0.682, | |
| "step": 3 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.7391304347826088e-06, | |
| "loss": 0.3229, | |
| "step": 4 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 2.173913043478261e-06, | |
| "loss": 0.1299, | |
| "step": 5 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 2.6086956521739132e-06, | |
| "loss": 0.0726, | |
| "step": 6 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 3.043478260869566e-06, | |
| "loss": 0.0795, | |
| "step": 7 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 3.4782608695652175e-06, | |
| "loss": 0.0596, | |
| "step": 8 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 3.91304347826087e-06, | |
| "loss": 0.0858, | |
| "step": 9 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 4.347826086956522e-06, | |
| "loss": 0.0782, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 4.782608695652174e-06, | |
| "loss": 0.0661, | |
| "step": 11 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 5.2173913043478265e-06, | |
| "loss": 0.0788, | |
| "step": 12 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 5.652173913043479e-06, | |
| "loss": 0.0763, | |
| "step": 13 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 6.086956521739132e-06, | |
| "loss": 0.06, | |
| "step": 14 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 6.521739130434783e-06, | |
| "loss": 0.0832, | |
| "step": 15 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 6.956521739130435e-06, | |
| "loss": 0.0658, | |
| "step": 16 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 7.391304347826087e-06, | |
| "loss": 0.0674, | |
| "step": 17 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 7.82608695652174e-06, | |
| "loss": 0.0771, | |
| "step": 18 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 8.260869565217392e-06, | |
| "loss": 0.0545, | |
| "step": 19 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 8.695652173913044e-06, | |
| "loss": 0.0565, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.130434782608697e-06, | |
| "loss": 0.0633, | |
| "step": 21 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 9.565217391304349e-06, | |
| "loss": 0.0582, | |
| "step": 22 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1e-05, | |
| "loss": 0.0565, | |
| "step": 23 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.0434782608695653e-05, | |
| "loss": 0.0523, | |
| "step": 24 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.0869565217391305e-05, | |
| "loss": 0.0471, | |
| "step": 25 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.1304347826086957e-05, | |
| "loss": 0.0722, | |
| "step": 26 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.1739130434782611e-05, | |
| "loss": 0.0434, | |
| "step": 27 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.2173913043478263e-05, | |
| "loss": 0.0442, | |
| "step": 28 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.2608695652173915e-05, | |
| "loss": 0.0648, | |
| "step": 29 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.3043478260869566e-05, | |
| "loss": 0.0605, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.3478260869565218e-05, | |
| "loss": 0.0839, | |
| "step": 31 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.391304347826087e-05, | |
| "loss": 0.0673, | |
| "step": 32 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.4347826086956522e-05, | |
| "loss": 0.0569, | |
| "step": 33 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.4782608695652174e-05, | |
| "loss": 0.0475, | |
| "step": 34 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.5217391304347828e-05, | |
| "loss": 0.0484, | |
| "step": 35 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.565217391304348e-05, | |
| "loss": 0.0388, | |
| "step": 36 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.6086956521739132e-05, | |
| "loss": 0.0555, | |
| "step": 37 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.6521739130434785e-05, | |
| "loss": 0.0415, | |
| "step": 38 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.6956521739130437e-05, | |
| "loss": 0.0553, | |
| "step": 39 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.739130434782609e-05, | |
| "loss": 0.0532, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.782608695652174e-05, | |
| "loss": 0.0606, | |
| "step": 41 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.8260869565217393e-05, | |
| "loss": 0.0529, | |
| "step": 42 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.8695652173913045e-05, | |
| "loss": 0.0443, | |
| "step": 43 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.9130434782608697e-05, | |
| "loss": 0.0456, | |
| "step": 44 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.956521739130435e-05, | |
| "loss": 0.0456, | |
| "step": 45 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 2e-05, | |
| "loss": 0.0346, | |
| "step": 46 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.999997731778071e-05, | |
| "loss": 0.0337, | |
| "step": 47 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.9999909271225726e-05, | |
| "loss": 0.0463, | |
| "step": 48 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.999979586064374e-05, | |
| "loss": 0.0357, | |
| "step": 49 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.9999637086549236e-05, | |
| "loss": 0.0448, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.9999432949662483e-05, | |
| "loss": 0.0429, | |
| "step": 51 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.9999183450909533e-05, | |
| "loss": 0.0391, | |
| "step": 52 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.999888859142223e-05, | |
| "loss": 0.0383, | |
| "step": 53 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.9998548372538175e-05, | |
| "loss": 0.0329, | |
| "step": 54 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.9998162795800766e-05, | |
| "loss": 0.0356, | |
| "step": 55 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.9997731862959143e-05, | |
| "loss": 0.0363, | |
| "step": 56 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.9997255575968207e-05, | |
| "loss": 0.0265, | |
| "step": 57 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.9996733936988613e-05, | |
| "loss": 0.0232, | |
| "step": 58 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.9996166948386745e-05, | |
| "loss": 0.0444, | |
| "step": 59 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.9995554612734713e-05, | |
| "loss": 0.031, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.999489693281034e-05, | |
| "loss": 0.0362, | |
| "step": 61 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.999419391159716e-05, | |
| "loss": 0.0347, | |
| "step": 62 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.9993445552284388e-05, | |
| "loss": 0.0336, | |
| "step": 63 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.9992651858266913e-05, | |
| "loss": 0.0571, | |
| "step": 64 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.9991812833145287e-05, | |
| "loss": 0.0217, | |
| "step": 65 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.9990928480725694e-05, | |
| "loss": 0.0312, | |
| "step": 66 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.998999880501995e-05, | |
| "loss": 0.0331, | |
| "step": 67 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.9989023810245484e-05, | |
| "loss": 0.0319, | |
| "step": 68 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.9988003500825293e-05, | |
| "loss": 0.0415, | |
| "step": 69 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.9986937881387968e-05, | |
| "loss": 0.0234, | |
| "step": 70 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.998582695676762e-05, | |
| "loss": 0.0192, | |
| "step": 71 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.99846707320039e-05, | |
| "loss": 0.0286, | |
| "step": 72 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.998346921234196e-05, | |
| "loss": 0.0407, | |
| "step": 73 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.998222240323242e-05, | |
| "loss": 0.0411, | |
| "step": 74 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.9980930310331367e-05, | |
| "loss": 0.027, | |
| "step": 75 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.99795929395003e-05, | |
| "loss": 0.0309, | |
| "step": 76 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.9978210296806134e-05, | |
| "loss": 0.032, | |
| "step": 77 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.9976782388521144e-05, | |
| "loss": 0.0286, | |
| "step": 78 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.9975309221122958e-05, | |
| "loss": 0.0333, | |
| "step": 79 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.9973790801294518e-05, | |
| "loss": 0.0316, | |
| "step": 80 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.997222713592405e-05, | |
| "loss": 0.0289, | |
| "step": 81 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.9970618232105036e-05, | |
| "loss": 0.0324, | |
| "step": 82 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.9968964097136173e-05, | |
| "loss": 0.0203, | |
| "step": 83 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.9967264738521358e-05, | |
| "loss": 0.034, | |
| "step": 84 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.996552016396963e-05, | |
| "loss": 0.0286, | |
| "step": 85 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.9963730381395154e-05, | |
| "loss": 0.0317, | |
| "step": 86 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.9961895398917182e-05, | |
| "loss": 0.0208, | |
| "step": 87 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 1.9960015224860007e-05, | |
| "loss": 0.0324, | |
| "step": 88 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 1.995808986775293e-05, | |
| "loss": 0.0312, | |
| "step": 89 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 1.9956119336330233e-05, | |
| "loss": 0.023, | |
| "step": 90 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 1.9954103639531116e-05, | |
| "loss": 0.0268, | |
| "step": 91 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 1.995204278649967e-05, | |
| "loss": 0.0502, | |
| "step": 92 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 1.9949936786584846e-05, | |
| "loss": 0.0253, | |
| "step": 93 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 1.9947785649340395e-05, | |
| "loss": 0.0366, | |
| "step": 94 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 1.9945589384524827e-05, | |
| "loss": 0.022, | |
| "step": 95 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 1.9943348002101374e-05, | |
| "loss": 0.0195, | |
| "step": 96 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 1.994106151223794e-05, | |
| "loss": 0.0285, | |
| "step": 97 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 1.9938729925307064e-05, | |
| "loss": 0.0215, | |
| "step": 98 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 1.9936353251885853e-05, | |
| "loss": 0.0179, | |
| "step": 99 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 1.9933931502755956e-05, | |
| "loss": 0.0256, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 1.9931464688903502e-05, | |
| "loss": 0.0241, | |
| "step": 101 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 1.9928952821519048e-05, | |
| "loss": 0.0228, | |
| "step": 102 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 1.9926395911997548e-05, | |
| "loss": 0.0165, | |
| "step": 103 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 1.9923793971938274e-05, | |
| "loss": 0.0235, | |
| "step": 104 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 1.9921147013144782e-05, | |
| "loss": 0.0219, | |
| "step": 105 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 1.9918455047624847e-05, | |
| "loss": 0.0222, | |
| "step": 106 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 1.991571808759043e-05, | |
| "loss": 0.0208, | |
| "step": 107 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 1.9912936145457586e-05, | |
| "loss": 0.0231, | |
| "step": 108 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 1.9910109233846443e-05, | |
| "loss": 0.024, | |
| "step": 109 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 1.990723736558113e-05, | |
| "loss": 0.0293, | |
| "step": 110 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 1.990432055368971e-05, | |
| "loss": 0.0155, | |
| "step": 111 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 1.990135881140414e-05, | |
| "loss": 0.0245, | |
| "step": 112 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 1.9898352152160204e-05, | |
| "loss": 0.0133, | |
| "step": 113 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 1.989530058959743e-05, | |
| "loss": 0.0158, | |
| "step": 114 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 1.989220413755907e-05, | |
| "loss": 0.0135, | |
| "step": 115 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 1.9889062810092002e-05, | |
| "loss": 0.0225, | |
| "step": 116 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 1.9885876621446677e-05, | |
| "loss": 0.0581, | |
| "step": 117 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 1.988264558607707e-05, | |
| "loss": 0.0288, | |
| "step": 118 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 1.987936971864058e-05, | |
| "loss": 0.0227, | |
| "step": 119 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 1.9876049033998007e-05, | |
| "loss": 0.0364, | |
| "step": 120 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 1.9872683547213446e-05, | |
| "loss": 0.0274, | |
| "step": 121 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 1.986927327355424e-05, | |
| "loss": 0.0208, | |
| "step": 122 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 1.98658182284909e-05, | |
| "loss": 0.0435, | |
| "step": 123 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 1.986231842769705e-05, | |
| "loss": 0.0219, | |
| "step": 124 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 1.985877388704933e-05, | |
| "loss": 0.0222, | |
| "step": 125 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 1.9855184622627362e-05, | |
| "loss": 0.0265, | |
| "step": 126 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 1.9851550650713632e-05, | |
| "loss": 0.0307, | |
| "step": 127 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 1.984787198779346e-05, | |
| "loss": 0.0167, | |
| "step": 128 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 1.9844148650554883e-05, | |
| "loss": 0.0194, | |
| "step": 129 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 1.9840380655888622e-05, | |
| "loss": 0.0271, | |
| "step": 130 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 1.9836568020887963e-05, | |
| "loss": 0.0231, | |
| "step": 131 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 1.9832710762848716e-05, | |
| "loss": 0.0187, | |
| "step": 132 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 1.9828808899269114e-05, | |
| "loss": 0.0182, | |
| "step": 133 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 1.9824862447849745e-05, | |
| "loss": 0.0392, | |
| "step": 134 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 1.9820871426493458e-05, | |
| "loss": 0.0258, | |
| "step": 135 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 1.9816835853305306e-05, | |
| "loss": 0.0222, | |
| "step": 136 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.9812755746592436e-05, | |
| "loss": 0.0198, | |
| "step": 137 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.980863112486402e-05, | |
| "loss": 0.0184, | |
| "step": 138 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.9804462006831174e-05, | |
| "loss": 0.03, | |
| "step": 139 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.9800248411406876e-05, | |
| "loss": 0.016, | |
| "step": 140 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.9795990357705853e-05, | |
| "loss": 0.0235, | |
| "step": 141 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.9791687865044537e-05, | |
| "loss": 0.0147, | |
| "step": 142 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.9787340952940935e-05, | |
| "loss": 0.0186, | |
| "step": 143 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.9782949641114576e-05, | |
| "loss": 0.02, | |
| "step": 144 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.97785139494864e-05, | |
| "loss": 0.0175, | |
| "step": 145 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.9774033898178668e-05, | |
| "loss": 0.0147, | |
| "step": 146 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.9769509507514887e-05, | |
| "loss": 0.0162, | |
| "step": 147 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.9764940798019695e-05, | |
| "loss": 0.0186, | |
| "step": 148 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.976032779041879e-05, | |
| "loss": 0.0145, | |
| "step": 149 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.9755670505638824e-05, | |
| "loss": 0.0206, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.9750968964807305e-05, | |
| "loss": 0.0231, | |
| "step": 151 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.974622318925251e-05, | |
| "loss": 0.0158, | |
| "step": 152 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 1.9741433200503382e-05, | |
| "loss": 0.0223, | |
| "step": 153 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 1.9736599020289438e-05, | |
| "loss": 0.0208, | |
| "step": 154 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 1.973172067054066e-05, | |
| "loss": 0.0227, | |
| "step": 155 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 1.9726798173387417e-05, | |
| "loss": 0.0237, | |
| "step": 156 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 1.9721831551160334e-05, | |
| "loss": 0.0354, | |
| "step": 157 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 1.9716820826390214e-05, | |
| "loss": 0.0201, | |
| "step": 158 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.971176602180793e-05, | |
| "loss": 0.0317, | |
| "step": 159 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.9706667160344325e-05, | |
| "loss": 0.0332, | |
| "step": 160 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.9701524265130088e-05, | |
| "loss": 0.0127, | |
| "step": 161 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.9696337359495678e-05, | |
| "loss": 0.009, | |
| "step": 162 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.9691106466971204e-05, | |
| "loss": 0.027, | |
| "step": 163 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 1.9685831611286312e-05, | |
| "loss": 0.0198, | |
| "step": 164 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.9680512816370096e-05, | |
| "loss": 0.0183, | |
| "step": 165 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.9675150106350957e-05, | |
| "loss": 0.0172, | |
| "step": 166 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.9669743505556544e-05, | |
| "loss": 0.0128, | |
| "step": 167 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.9664293038513587e-05, | |
| "loss": 0.0223, | |
| "step": 168 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 1.9658798729947827e-05, | |
| "loss": 0.0164, | |
| "step": 169 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 1.9653260604783887e-05, | |
| "loss": 0.0159, | |
| "step": 170 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 1.9647678688145163e-05, | |
| "loss": 0.0189, | |
| "step": 171 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 1.96420530053537e-05, | |
| "loss": 0.0128, | |
| "step": 172 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 1.9636383581930097e-05, | |
| "loss": 0.0118, | |
| "step": 173 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 1.9630670443593373e-05, | |
| "loss": 0.0099, | |
| "step": 174 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 1.9624913616260862e-05, | |
| "loss": 0.0193, | |
| "step": 175 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 1.9619113126048086e-05, | |
| "loss": 0.0117, | |
| "step": 176 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 1.9613268999268644e-05, | |
| "loss": 0.0152, | |
| "step": 177 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 1.9607381262434086e-05, | |
| "loss": 0.0153, | |
| "step": 178 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 1.9601449942253805e-05, | |
| "loss": 0.0074, | |
| "step": 179 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 1.9595475065634896e-05, | |
| "loss": 0.0123, | |
| "step": 180 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 1.958945665968206e-05, | |
| "loss": 0.0216, | |
| "step": 181 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 1.9583394751697447e-05, | |
| "loss": 0.0075, | |
| "step": 182 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 1.957728936918057e-05, | |
| "loss": 0.0143, | |
| "step": 183 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 1.957114053982815e-05, | |
| "loss": 0.0135, | |
| "step": 184 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 1.956494829153401e-05, | |
| "loss": 0.0195, | |
| "step": 185 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 1.9558712652388932e-05, | |
| "loss": 0.0096, | |
| "step": 186 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 1.9552433650680545e-05, | |
| "loss": 0.0062, | |
| "step": 187 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 1.954611131489319e-05, | |
| "loss": 0.0106, | |
| "step": 188 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 1.9539745673707785e-05, | |
| "loss": 0.0159, | |
| "step": 189 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 1.9533336756001707e-05, | |
| "loss": 0.0159, | |
| "step": 190 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 1.9526884590848646e-05, | |
| "loss": 0.0103, | |
| "step": 191 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 1.9520389207518495e-05, | |
| "loss": 0.0201, | |
| "step": 192 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 1.9513850635477188e-05, | |
| "loss": 0.0169, | |
| "step": 193 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 1.9507268904386596e-05, | |
| "loss": 0.0095, | |
| "step": 194 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 1.950064404410437e-05, | |
| "loss": 0.0114, | |
| "step": 195 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 1.9493976084683814e-05, | |
| "loss": 0.0263, | |
| "step": 196 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 1.9487265056373756e-05, | |
| "loss": 0.01, | |
| "step": 197 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 1.9480510989618396e-05, | |
| "loss": 0.0181, | |
| "step": 198 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 1.9473713915057183e-05, | |
| "loss": 0.019, | |
| "step": 199 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 1.946687386352466e-05, | |
| "loss": 0.0069, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 1.9459990866050337e-05, | |
| "loss": 0.0204, | |
| "step": 201 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "learning_rate": 1.945306495385855e-05, | |
| "loss": 0.0167, | |
| "step": 202 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "learning_rate": 1.9446096158368303e-05, | |
| "loss": 0.0096, | |
| "step": 203 | |
| }, | |
| { | |
| "epoch": 1.21, | |
| "learning_rate": 1.9439084511193153e-05, | |
| "loss": 0.0053, | |
| "step": 204 | |
| }, | |
| { | |
| "epoch": 1.21, | |
| "learning_rate": 1.943203004414104e-05, | |
| "loss": 0.0113, | |
| "step": 205 | |
| }, | |
| { | |
| "epoch": 1.22, | |
| "learning_rate": 1.9424932789214158e-05, | |
| "loss": 0.0129, | |
| "step": 206 | |
| }, | |
| { | |
| "epoch": 1.22, | |
| "learning_rate": 1.9417792778608806e-05, | |
| "loss": 0.0198, | |
| "step": 207 | |
| }, | |
| { | |
| "epoch": 1.23, | |
| "learning_rate": 1.9410610044715243e-05, | |
| "loss": 0.0073, | |
| "step": 208 | |
| }, | |
| { | |
| "epoch": 1.24, | |
| "learning_rate": 1.9403384620117534e-05, | |
| "loss": 0.0126, | |
| "step": 209 | |
| }, | |
| { | |
| "epoch": 1.24, | |
| "learning_rate": 1.9396116537593413e-05, | |
| "loss": 0.0099, | |
| "step": 210 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "learning_rate": 1.9388805830114132e-05, | |
| "loss": 0.0131, | |
| "step": 211 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "learning_rate": 1.9381452530844306e-05, | |
| "loss": 0.0041, | |
| "step": 212 | |
| }, | |
| { | |
| "epoch": 1.26, | |
| "learning_rate": 1.9374056673141756e-05, | |
| "loss": 0.0087, | |
| "step": 213 | |
| }, | |
| { | |
| "epoch": 1.27, | |
| "learning_rate": 1.9366618290557384e-05, | |
| "loss": 0.0114, | |
| "step": 214 | |
| }, | |
| { | |
| "epoch": 1.27, | |
| "learning_rate": 1.9359137416834987e-05, | |
| "loss": 0.012, | |
| "step": 215 | |
| }, | |
| { | |
| "epoch": 1.28, | |
| "learning_rate": 1.9351614085911134e-05, | |
| "loss": 0.0149, | |
| "step": 216 | |
| }, | |
| { | |
| "epoch": 1.28, | |
| "learning_rate": 1.9344048331914998e-05, | |
| "loss": 0.0115, | |
| "step": 217 | |
| }, | |
| { | |
| "epoch": 1.29, | |
| "learning_rate": 1.9336440189168186e-05, | |
| "loss": 0.0073, | |
| "step": 218 | |
| }, | |
| { | |
| "epoch": 1.3, | |
| "learning_rate": 1.9328789692184618e-05, | |
| "loss": 0.009, | |
| "step": 219 | |
| }, | |
| { | |
| "epoch": 1.3, | |
| "learning_rate": 1.932109687567034e-05, | |
| "loss": 0.0022, | |
| "step": 220 | |
| }, | |
| { | |
| "epoch": 1.31, | |
| "learning_rate": 1.9313361774523387e-05, | |
| "loss": 0.0085, | |
| "step": 221 | |
| }, | |
| { | |
| "epoch": 1.31, | |
| "learning_rate": 1.9305584423833606e-05, | |
| "loss": 0.0199, | |
| "step": 222 | |
| }, | |
| { | |
| "epoch": 1.32, | |
| "learning_rate": 1.9297764858882516e-05, | |
| "loss": 0.0126, | |
| "step": 223 | |
| }, | |
| { | |
| "epoch": 1.33, | |
| "learning_rate": 1.9289903115143133e-05, | |
| "loss": 0.0098, | |
| "step": 224 | |
| }, | |
| { | |
| "epoch": 1.33, | |
| "learning_rate": 1.9281999228279814e-05, | |
| "loss": 0.0158, | |
| "step": 225 | |
| }, | |
| { | |
| "epoch": 1.34, | |
| "learning_rate": 1.92740532341481e-05, | |
| "loss": 0.0068, | |
| "step": 226 | |
| }, | |
| { | |
| "epoch": 1.34, | |
| "learning_rate": 1.9266065168794546e-05, | |
| "loss": 0.0077, | |
| "step": 227 | |
| }, | |
| { | |
| "epoch": 1.35, | |
| "learning_rate": 1.9258035068456566e-05, | |
| "loss": 0.0243, | |
| "step": 228 | |
| }, | |
| { | |
| "epoch": 1.36, | |
| "learning_rate": 1.9249962969562256e-05, | |
| "loss": 0.0198, | |
| "step": 229 | |
| }, | |
| { | |
| "epoch": 1.36, | |
| "learning_rate": 1.9241848908730238e-05, | |
| "loss": 0.0118, | |
| "step": 230 | |
| }, | |
| { | |
| "epoch": 1.37, | |
| "learning_rate": 1.9233692922769497e-05, | |
| "loss": 0.0213, | |
| "step": 231 | |
| }, | |
| { | |
| "epoch": 1.37, | |
| "learning_rate": 1.92254950486792e-05, | |
| "loss": 0.0115, | |
| "step": 232 | |
| }, | |
| { | |
| "epoch": 1.38, | |
| "learning_rate": 1.9217255323648548e-05, | |
| "loss": 0.0104, | |
| "step": 233 | |
| }, | |
| { | |
| "epoch": 1.38, | |
| "learning_rate": 1.920897378505659e-05, | |
| "loss": 0.0096, | |
| "step": 234 | |
| }, | |
| { | |
| "epoch": 1.39, | |
| "learning_rate": 1.920065047047206e-05, | |
| "loss": 0.0106, | |
| "step": 235 | |
| }, | |
| { | |
| "epoch": 1.4, | |
| "learning_rate": 1.9192285417653208e-05, | |
| "loss": 0.0093, | |
| "step": 236 | |
| }, | |
| { | |
| "epoch": 1.4, | |
| "learning_rate": 1.9183878664547628e-05, | |
| "loss": 0.0054, | |
| "step": 237 | |
| }, | |
| { | |
| "epoch": 1.41, | |
| "learning_rate": 1.917543024929208e-05, | |
| "loss": 0.0087, | |
| "step": 238 | |
| }, | |
| { | |
| "epoch": 1.41, | |
| "learning_rate": 1.9166940210212324e-05, | |
| "loss": 0.0064, | |
| "step": 239 | |
| }, | |
| { | |
| "epoch": 1.42, | |
| "learning_rate": 1.915840858582295e-05, | |
| "loss": 0.0056, | |
| "step": 240 | |
| }, | |
| { | |
| "epoch": 1.43, | |
| "learning_rate": 1.9149835414827193e-05, | |
| "loss": 0.0082, | |
| "step": 241 | |
| }, | |
| { | |
| "epoch": 1.43, | |
| "learning_rate": 1.914122073611676e-05, | |
| "loss": 0.0165, | |
| "step": 242 | |
| }, | |
| { | |
| "epoch": 1.44, | |
| "learning_rate": 1.913256458877166e-05, | |
| "loss": 0.0039, | |
| "step": 243 | |
| }, | |
| { | |
| "epoch": 1.44, | |
| "learning_rate": 1.9123867012060016e-05, | |
| "loss": 0.0121, | |
| "step": 244 | |
| }, | |
| { | |
| "epoch": 1.45, | |
| "learning_rate": 1.9115128045437896e-05, | |
| "loss": 0.0119, | |
| "step": 245 | |
| }, | |
| { | |
| "epoch": 1.46, | |
| "learning_rate": 1.9106347728549134e-05, | |
| "loss": 0.0087, | |
| "step": 246 | |
| }, | |
| { | |
| "epoch": 1.46, | |
| "learning_rate": 1.9097526101225146e-05, | |
| "loss": 0.0057, | |
| "step": 247 | |
| }, | |
| { | |
| "epoch": 1.47, | |
| "learning_rate": 1.9088663203484742e-05, | |
| "loss": 0.0038, | |
| "step": 248 | |
| }, | |
| { | |
| "epoch": 1.47, | |
| "learning_rate": 1.907975907553397e-05, | |
| "loss": 0.0068, | |
| "step": 249 | |
| }, | |
| { | |
| "epoch": 1.48, | |
| "learning_rate": 1.90708137577659e-05, | |
| "loss": 0.0069, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 1.49, | |
| "learning_rate": 1.9061827290760466e-05, | |
| "loss": 0.007, | |
| "step": 251 | |
| }, | |
| { | |
| "epoch": 1.49, | |
| "learning_rate": 1.905279971528427e-05, | |
| "loss": 0.0049, | |
| "step": 252 | |
| }, | |
| { | |
| "epoch": 1.5, | |
| "learning_rate": 1.90437310722904e-05, | |
| "loss": 0.0068, | |
| "step": 253 | |
| }, | |
| { | |
| "epoch": 1.5, | |
| "learning_rate": 1.903462140291825e-05, | |
| "loss": 0.0124, | |
| "step": 254 | |
| }, | |
| { | |
| "epoch": 1.51, | |
| "learning_rate": 1.9025470748493323e-05, | |
| "loss": 0.0061, | |
| "step": 255 | |
| }, | |
| { | |
| "epoch": 1.51, | |
| "learning_rate": 1.9016279150527044e-05, | |
| "loss": 0.0168, | |
| "step": 256 | |
| }, | |
| { | |
| "epoch": 1.52, | |
| "learning_rate": 1.9007046650716584e-05, | |
| "loss": 0.012, | |
| "step": 257 | |
| }, | |
| { | |
| "epoch": 1.53, | |
| "learning_rate": 1.8997773290944663e-05, | |
| "loss": 0.0048, | |
| "step": 258 | |
| }, | |
| { | |
| "epoch": 1.53, | |
| "learning_rate": 1.8988459113279355e-05, | |
| "loss": 0.0239, | |
| "step": 259 | |
| }, | |
| { | |
| "epoch": 1.54, | |
| "learning_rate": 1.8979104159973905e-05, | |
| "loss": 0.0037, | |
| "step": 260 | |
| }, | |
| { | |
| "epoch": 1.54, | |
| "learning_rate": 1.896970847346653e-05, | |
| "loss": 0.0105, | |
| "step": 261 | |
| }, | |
| { | |
| "epoch": 1.55, | |
| "learning_rate": 1.8960272096380236e-05, | |
| "loss": 0.0093, | |
| "step": 262 | |
| }, | |
| { | |
| "epoch": 1.56, | |
| "learning_rate": 1.895079507152262e-05, | |
| "loss": 0.0076, | |
| "step": 263 | |
| }, | |
| { | |
| "epoch": 1.56, | |
| "learning_rate": 1.8941277441885672e-05, | |
| "loss": 0.0155, | |
| "step": 264 | |
| }, | |
| { | |
| "epoch": 1.57, | |
| "learning_rate": 1.8931719250645583e-05, | |
| "loss": 0.0078, | |
| "step": 265 | |
| }, | |
| { | |
| "epoch": 1.57, | |
| "learning_rate": 1.892212054116255e-05, | |
| "loss": 0.0079, | |
| "step": 266 | |
| }, | |
| { | |
| "epoch": 1.58, | |
| "learning_rate": 1.8912481356980582e-05, | |
| "loss": 0.0101, | |
| "step": 267 | |
| }, | |
| { | |
| "epoch": 1.59, | |
| "learning_rate": 1.89028017418273e-05, | |
| "loss": 0.0085, | |
| "step": 268 | |
| }, | |
| { | |
| "epoch": 1.59, | |
| "learning_rate": 1.8893081739613728e-05, | |
| "loss": 0.0135, | |
| "step": 269 | |
| }, | |
| { | |
| "epoch": 1.6, | |
| "learning_rate": 1.8883321394434116e-05, | |
| "loss": 0.0081, | |
| "step": 270 | |
| }, | |
| { | |
| "epoch": 1.6, | |
| "learning_rate": 1.8873520750565716e-05, | |
| "loss": 0.0057, | |
| "step": 271 | |
| }, | |
| { | |
| "epoch": 1.61, | |
| "learning_rate": 1.8863679852468605e-05, | |
| "loss": 0.0133, | |
| "step": 272 | |
| }, | |
| { | |
| "epoch": 1.62, | |
| "learning_rate": 1.8853798744785466e-05, | |
| "loss": 0.0114, | |
| "step": 273 | |
| }, | |
| { | |
| "epoch": 1.62, | |
| "learning_rate": 1.8843877472341382e-05, | |
| "loss": 0.0069, | |
| "step": 274 | |
| }, | |
| { | |
| "epoch": 1.63, | |
| "learning_rate": 1.883391608014365e-05, | |
| "loss": 0.0072, | |
| "step": 275 | |
| }, | |
| { | |
| "epoch": 1.63, | |
| "learning_rate": 1.8823914613381568e-05, | |
| "loss": 0.0049, | |
| "step": 276 | |
| }, | |
| { | |
| "epoch": 1.64, | |
| "learning_rate": 1.881387311742623e-05, | |
| "loss": 0.0036, | |
| "step": 277 | |
| }, | |
| { | |
| "epoch": 1.64, | |
| "learning_rate": 1.8803791637830317e-05, | |
| "loss": 0.0054, | |
| "step": 278 | |
| }, | |
| { | |
| "epoch": 1.65, | |
| "learning_rate": 1.8793670220327897e-05, | |
| "loss": 0.0062, | |
| "step": 279 | |
| }, | |
| { | |
| "epoch": 1.66, | |
| "learning_rate": 1.8783508910834212e-05, | |
| "loss": 0.0036, | |
| "step": 280 | |
| }, | |
| { | |
| "epoch": 1.66, | |
| "learning_rate": 1.8773307755445468e-05, | |
| "loss": 0.0095, | |
| "step": 281 | |
| }, | |
| { | |
| "epoch": 1.67, | |
| "learning_rate": 1.8763066800438638e-05, | |
| "loss": 0.0111, | |
| "step": 282 | |
| }, | |
| { | |
| "epoch": 1.67, | |
| "learning_rate": 1.875278609227124e-05, | |
| "loss": 0.0054, | |
| "step": 283 | |
| }, | |
| { | |
| "epoch": 1.68, | |
| "learning_rate": 1.874246567758112e-05, | |
| "loss": 0.0085, | |
| "step": 284 | |
| }, | |
| { | |
| "epoch": 1.69, | |
| "learning_rate": 1.8732105603186273e-05, | |
| "loss": 0.0054, | |
| "step": 285 | |
| }, | |
| { | |
| "epoch": 1.69, | |
| "learning_rate": 1.872170591608459e-05, | |
| "loss": 0.0082, | |
| "step": 286 | |
| }, | |
| { | |
| "epoch": 1.7, | |
| "learning_rate": 1.8711266663453666e-05, | |
| "loss": 0.0047, | |
| "step": 287 | |
| }, | |
| { | |
| "epoch": 1.7, | |
| "learning_rate": 1.8700787892650584e-05, | |
| "loss": 0.0059, | |
| "step": 288 | |
| }, | |
| { | |
| "epoch": 1.71, | |
| "learning_rate": 1.8690269651211705e-05, | |
| "loss": 0.0038, | |
| "step": 289 | |
| }, | |
| { | |
| "epoch": 1.72, | |
| "learning_rate": 1.8679711986852432e-05, | |
| "loss": 0.0056, | |
| "step": 290 | |
| }, | |
| { | |
| "epoch": 1.72, | |
| "learning_rate": 1.866911494746702e-05, | |
| "loss": 0.0072, | |
| "step": 291 | |
| }, | |
| { | |
| "epoch": 1.73, | |
| "learning_rate": 1.8658478581128347e-05, | |
| "loss": 0.0062, | |
| "step": 292 | |
| }, | |
| { | |
| "epoch": 1.73, | |
| "learning_rate": 1.864780293608769e-05, | |
| "loss": 0.0109, | |
| "step": 293 | |
| }, | |
| { | |
| "epoch": 1.74, | |
| "learning_rate": 1.8637088060774514e-05, | |
| "loss": 0.0105, | |
| "step": 294 | |
| }, | |
| { | |
| "epoch": 1.75, | |
| "learning_rate": 1.8626334003796244e-05, | |
| "loss": 0.0105, | |
| "step": 295 | |
| }, | |
| { | |
| "epoch": 1.75, | |
| "learning_rate": 1.8615540813938063e-05, | |
| "loss": 0.0169, | |
| "step": 296 | |
| }, | |
| { | |
| "epoch": 1.76, | |
| "learning_rate": 1.8604708540162665e-05, | |
| "loss": 0.0113, | |
| "step": 297 | |
| }, | |
| { | |
| "epoch": 1.76, | |
| "learning_rate": 1.8593837231610056e-05, | |
| "loss": 0.0031, | |
| "step": 298 | |
| }, | |
| { | |
| "epoch": 1.77, | |
| "learning_rate": 1.8582926937597316e-05, | |
| "loss": 0.0116, | |
| "step": 299 | |
| }, | |
| { | |
| "epoch": 1.78, | |
| "learning_rate": 1.857197770761838e-05, | |
| "loss": 0.0118, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 1.78, | |
| "learning_rate": 1.856098959134381e-05, | |
| "loss": 0.0163, | |
| "step": 301 | |
| }, | |
| { | |
| "epoch": 1.79, | |
| "learning_rate": 1.8549962638620592e-05, | |
| "loss": 0.005, | |
| "step": 302 | |
| }, | |
| { | |
| "epoch": 1.79, | |
| "learning_rate": 1.8538896899471868e-05, | |
| "loss": 0.009, | |
| "step": 303 | |
| }, | |
| { | |
| "epoch": 1.8, | |
| "learning_rate": 1.8527792424096743e-05, | |
| "loss": 0.0092, | |
| "step": 304 | |
| }, | |
| { | |
| "epoch": 1.8, | |
| "learning_rate": 1.851664926287005e-05, | |
| "loss": 0.0115, | |
| "step": 305 | |
| }, | |
| { | |
| "epoch": 1.81, | |
| "learning_rate": 1.850546746634211e-05, | |
| "loss": 0.0129, | |
| "step": 306 | |
| }, | |
| { | |
| "epoch": 1.82, | |
| "learning_rate": 1.8494247085238516e-05, | |
| "loss": 0.0128, | |
| "step": 307 | |
| }, | |
| { | |
| "epoch": 1.82, | |
| "learning_rate": 1.84829881704599e-05, | |
| "loss": 0.0098, | |
| "step": 308 | |
| }, | |
| { | |
| "epoch": 1.83, | |
| "learning_rate": 1.84716907730817e-05, | |
| "loss": 0.0107, | |
| "step": 309 | |
| }, | |
| { | |
| "epoch": 1.83, | |
| "learning_rate": 1.846035494435391e-05, | |
| "loss": 0.0183, | |
| "step": 310 | |
| }, | |
| { | |
| "epoch": 1.84, | |
| "learning_rate": 1.84489807357009e-05, | |
| "loss": 0.0137, | |
| "step": 311 | |
| }, | |
| { | |
| "epoch": 1.85, | |
| "learning_rate": 1.8437568198721113e-05, | |
| "loss": 0.0073, | |
| "step": 312 | |
| }, | |
| { | |
| "epoch": 1.85, | |
| "learning_rate": 1.842611738518689e-05, | |
| "loss": 0.0031, | |
| "step": 313 | |
| }, | |
| { | |
| "epoch": 1.86, | |
| "learning_rate": 1.8414628347044203e-05, | |
| "loss": 0.0179, | |
| "step": 314 | |
| }, | |
| { | |
| "epoch": 1.86, | |
| "learning_rate": 1.840310113641243e-05, | |
| "loss": 0.0108, | |
| "step": 315 | |
| }, | |
| { | |
| "epoch": 1.87, | |
| "learning_rate": 1.839153580558411e-05, | |
| "loss": 0.0093, | |
| "step": 316 | |
| }, | |
| { | |
| "epoch": 1.88, | |
| "learning_rate": 1.8379932407024723e-05, | |
| "loss": 0.0167, | |
| "step": 317 | |
| }, | |
| { | |
| "epoch": 1.88, | |
| "learning_rate": 1.836829099337243e-05, | |
| "loss": 0.0066, | |
| "step": 318 | |
| }, | |
| { | |
| "epoch": 1.89, | |
| "learning_rate": 1.8356611617437853e-05, | |
| "loss": 0.007, | |
| "step": 319 | |
| }, | |
| { | |
| "epoch": 1.89, | |
| "learning_rate": 1.8344894332203823e-05, | |
| "loss": 0.0098, | |
| "step": 320 | |
| }, | |
| { | |
| "epoch": 1.9, | |
| "learning_rate": 1.833313919082515e-05, | |
| "loss": 0.0095, | |
| "step": 321 | |
| }, | |
| { | |
| "epoch": 1.91, | |
| "learning_rate": 1.8321346246628374e-05, | |
| "loss": 0.0074, | |
| "step": 322 | |
| }, | |
| { | |
| "epoch": 1.91, | |
| "learning_rate": 1.8309515553111518e-05, | |
| "loss": 0.0074, | |
| "step": 323 | |
| }, | |
| { | |
| "epoch": 1.92, | |
| "learning_rate": 1.8297647163943865e-05, | |
| "loss": 0.0035, | |
| "step": 324 | |
| }, | |
| { | |
| "epoch": 1.92, | |
| "learning_rate": 1.8285741132965693e-05, | |
| "loss": 0.0142, | |
| "step": 325 | |
| }, | |
| { | |
| "epoch": 1.93, | |
| "learning_rate": 1.8273797514188043e-05, | |
| "loss": 0.0133, | |
| "step": 326 | |
| }, | |
| { | |
| "epoch": 1.93, | |
| "learning_rate": 1.8261816361792474e-05, | |
| "loss": 0.0038, | |
| "step": 327 | |
| }, | |
| { | |
| "epoch": 1.94, | |
| "learning_rate": 1.8249797730130804e-05, | |
| "loss": 0.0083, | |
| "step": 328 | |
| }, | |
| { | |
| "epoch": 1.95, | |
| "learning_rate": 1.8237741673724892e-05, | |
| "loss": 0.0028, | |
| "step": 329 | |
| }, | |
| { | |
| "epoch": 1.95, | |
| "learning_rate": 1.822564824726635e-05, | |
| "loss": 0.0118, | |
| "step": 330 | |
| }, | |
| { | |
| "epoch": 1.96, | |
| "learning_rate": 1.821351750561634e-05, | |
| "loss": 0.0076, | |
| "step": 331 | |
| }, | |
| { | |
| "epoch": 1.96, | |
| "learning_rate": 1.8201349503805275e-05, | |
| "loss": 0.0079, | |
| "step": 332 | |
| }, | |
| { | |
| "epoch": 1.97, | |
| "learning_rate": 1.8189144297032627e-05, | |
| "loss": 0.0051, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 1.98, | |
| "learning_rate": 1.8176901940666624e-05, | |
| "loss": 0.021, | |
| "step": 334 | |
| }, | |
| { | |
| "epoch": 1.98, | |
| "learning_rate": 1.8164622490244027e-05, | |
| "loss": 0.0065, | |
| "step": 335 | |
| }, | |
| { | |
| "epoch": 1.99, | |
| "learning_rate": 1.8152306001469875e-05, | |
| "loss": 0.011, | |
| "step": 336 | |
| }, | |
| { | |
| "epoch": 1.99, | |
| "learning_rate": 1.8139952530217232e-05, | |
| "loss": 0.0059, | |
| "step": 337 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 1.8127562132526922e-05, | |
| "loss": 0.0075, | |
| "step": 338 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 1.811513486460729e-05, | |
| "loss": 0.0045, | |
| "step": 339 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 1.8102670782833938e-05, | |
| "loss": 0.0027, | |
| "step": 340 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "learning_rate": 1.8090169943749477e-05, | |
| "loss": 0.0021, | |
| "step": 341 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "learning_rate": 1.8077632404063256e-05, | |
| "loss": 0.0027, | |
| "step": 342 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "learning_rate": 1.8065058220651127e-05, | |
| "loss": 0.0029, | |
| "step": 343 | |
| }, | |
| { | |
| "epoch": 2.04, | |
| "learning_rate": 1.8052447450555157e-05, | |
| "loss": 0.0023, | |
| "step": 344 | |
| }, | |
| { | |
| "epoch": 2.04, | |
| "learning_rate": 1.803980015098341e-05, | |
| "loss": 0.0027, | |
| "step": 345 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "learning_rate": 1.8027116379309637e-05, | |
| "loss": 0.0035, | |
| "step": 346 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "learning_rate": 1.8014396193073067e-05, | |
| "loss": 0.0008, | |
| "step": 347 | |
| }, | |
| { | |
| "epoch": 2.06, | |
| "learning_rate": 1.8001639649978102e-05, | |
| "loss": 0.0015, | |
| "step": 348 | |
| }, | |
| { | |
| "epoch": 2.07, | |
| "learning_rate": 1.798884680789409e-05, | |
| "loss": 0.0025, | |
| "step": 349 | |
| }, | |
| { | |
| "epoch": 2.07, | |
| "learning_rate": 1.797601772485504e-05, | |
| "loss": 0.0029, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 2.08, | |
| "learning_rate": 1.796315245905936e-05, | |
| "loss": 0.0038, | |
| "step": 351 | |
| }, | |
| { | |
| "epoch": 2.08, | |
| "learning_rate": 1.795025106886962e-05, | |
| "loss": 0.001, | |
| "step": 352 | |
| }, | |
| { | |
| "epoch": 2.09, | |
| "learning_rate": 1.7937313612812234e-05, | |
| "loss": 0.0015, | |
| "step": 353 | |
| }, | |
| { | |
| "epoch": 2.09, | |
| "learning_rate": 1.7924340149577263e-05, | |
| "loss": 0.0042, | |
| "step": 354 | |
| }, | |
| { | |
| "epoch": 2.1, | |
| "learning_rate": 1.7911330738018084e-05, | |
| "loss": 0.0009, | |
| "step": 355 | |
| }, | |
| { | |
| "epoch": 2.11, | |
| "learning_rate": 1.7898285437151163e-05, | |
| "loss": 0.0009, | |
| "step": 356 | |
| }, | |
| { | |
| "epoch": 2.11, | |
| "learning_rate": 1.7885204306155777e-05, | |
| "loss": 0.0005, | |
| "step": 357 | |
| }, | |
| { | |
| "epoch": 2.12, | |
| "learning_rate": 1.787208740437374e-05, | |
| "loss": 0.0022, | |
| "step": 358 | |
| }, | |
| { | |
| "epoch": 2.12, | |
| "learning_rate": 1.7858934791309146e-05, | |
| "loss": 0.0019, | |
| "step": 359 | |
| }, | |
| { | |
| "epoch": 2.13, | |
| "learning_rate": 1.784574652662808e-05, | |
| "loss": 0.0032, | |
| "step": 360 | |
| }, | |
| { | |
| "epoch": 2.14, | |
| "learning_rate": 1.783252267015837e-05, | |
| "loss": 0.0007, | |
| "step": 361 | |
| }, | |
| { | |
| "epoch": 2.14, | |
| "learning_rate": 1.781926328188929e-05, | |
| "loss": 0.003, | |
| "step": 362 | |
| }, | |
| { | |
| "epoch": 2.15, | |
| "learning_rate": 1.780596842197132e-05, | |
| "loss": 0.0049, | |
| "step": 363 | |
| }, | |
| { | |
| "epoch": 2.15, | |
| "learning_rate": 1.779263815071584e-05, | |
| "loss": 0.0073, | |
| "step": 364 | |
| }, | |
| { | |
| "epoch": 2.16, | |
| "learning_rate": 1.7779272528594876e-05, | |
| "loss": 0.002, | |
| "step": 365 | |
| }, | |
| { | |
| "epoch": 2.17, | |
| "learning_rate": 1.776587161624083e-05, | |
| "loss": 0.0005, | |
| "step": 366 | |
| }, | |
| { | |
| "epoch": 2.17, | |
| "learning_rate": 1.7752435474446178e-05, | |
| "loss": 0.0017, | |
| "step": 367 | |
| }, | |
| { | |
| "epoch": 2.18, | |
| "learning_rate": 1.773896416416323e-05, | |
| "loss": 0.0061, | |
| "step": 368 | |
| }, | |
| { | |
| "epoch": 2.18, | |
| "learning_rate": 1.772545774650383e-05, | |
| "loss": 0.0019, | |
| "step": 369 | |
| }, | |
| { | |
| "epoch": 2.19, | |
| "learning_rate": 1.7711916282739075e-05, | |
| "loss": 0.0006, | |
| "step": 370 | |
| }, | |
| { | |
| "epoch": 2.2, | |
| "learning_rate": 1.7698339834299064e-05, | |
| "loss": 0.0005, | |
| "step": 371 | |
| }, | |
| { | |
| "epoch": 2.2, | |
| "learning_rate": 1.768472846277259e-05, | |
| "loss": 0.0005, | |
| "step": 372 | |
| }, | |
| { | |
| "epoch": 2.21, | |
| "learning_rate": 1.767108222990687e-05, | |
| "loss": 0.0018, | |
| "step": 373 | |
| }, | |
| { | |
| "epoch": 2.21, | |
| "learning_rate": 1.765740119760729e-05, | |
| "loss": 0.0023, | |
| "step": 374 | |
| }, | |
| { | |
| "epoch": 2.22, | |
| "learning_rate": 1.7643685427937065e-05, | |
| "loss": 0.0018, | |
| "step": 375 | |
| }, | |
| { | |
| "epoch": 2.22, | |
| "learning_rate": 1.7629934983117025e-05, | |
| "loss": 0.0058, | |
| "step": 376 | |
| }, | |
| { | |
| "epoch": 2.23, | |
| "learning_rate": 1.761614992552529e-05, | |
| "loss": 0.0014, | |
| "step": 377 | |
| }, | |
| { | |
| "epoch": 2.24, | |
| "learning_rate": 1.7602330317696998e-05, | |
| "loss": 0.0005, | |
| "step": 378 | |
| }, | |
| { | |
| "epoch": 2.24, | |
| "learning_rate": 1.758847622232402e-05, | |
| "loss": 0.0039, | |
| "step": 379 | |
| }, | |
| { | |
| "epoch": 2.25, | |
| "learning_rate": 1.7574587702254692e-05, | |
| "loss": 0.0016, | |
| "step": 380 | |
| }, | |
| { | |
| "epoch": 2.25, | |
| "learning_rate": 1.7560664820493502e-05, | |
| "loss": 0.001, | |
| "step": 381 | |
| }, | |
| { | |
| "epoch": 2.26, | |
| "learning_rate": 1.7546707640200815e-05, | |
| "loss": 0.0015, | |
| "step": 382 | |
| }, | |
| { | |
| "epoch": 2.27, | |
| "learning_rate": 1.7532716224692602e-05, | |
| "loss": 0.0016, | |
| "step": 383 | |
| }, | |
| { | |
| "epoch": 2.27, | |
| "learning_rate": 1.7518690637440135e-05, | |
| "loss": 0.0012, | |
| "step": 384 | |
| }, | |
| { | |
| "epoch": 2.28, | |
| "learning_rate": 1.7504630942069695e-05, | |
| "loss": 0.0006, | |
| "step": 385 | |
| }, | |
| { | |
| "epoch": 2.28, | |
| "learning_rate": 1.7490537202362313e-05, | |
| "loss": 0.0003, | |
| "step": 386 | |
| }, | |
| { | |
| "epoch": 2.29, | |
| "learning_rate": 1.7476409482253433e-05, | |
| "loss": 0.002, | |
| "step": 387 | |
| }, | |
| { | |
| "epoch": 2.3, | |
| "learning_rate": 1.746224784583268e-05, | |
| "loss": 0.0028, | |
| "step": 388 | |
| }, | |
| { | |
| "epoch": 2.3, | |
| "learning_rate": 1.7448052357343514e-05, | |
| "loss": 0.0002, | |
| "step": 389 | |
| }, | |
| { | |
| "epoch": 2.31, | |
| "learning_rate": 1.7433823081182967e-05, | |
| "loss": 0.0009, | |
| "step": 390 | |
| }, | |
| { | |
| "epoch": 2.31, | |
| "learning_rate": 1.741956008190136e-05, | |
| "loss": 0.002, | |
| "step": 391 | |
| }, | |
| { | |
| "epoch": 2.32, | |
| "learning_rate": 1.7405263424201987e-05, | |
| "loss": 0.0004, | |
| "step": 392 | |
| }, | |
| { | |
| "epoch": 2.33, | |
| "learning_rate": 1.739093317294083e-05, | |
| "loss": 0.0023, | |
| "step": 393 | |
| }, | |
| { | |
| "epoch": 2.33, | |
| "learning_rate": 1.7376569393126274e-05, | |
| "loss": 0.0008, | |
| "step": 394 | |
| }, | |
| { | |
| "epoch": 2.34, | |
| "learning_rate": 1.736217214991879e-05, | |
| "loss": 0.0032, | |
| "step": 395 | |
| }, | |
| { | |
| "epoch": 2.34, | |
| "learning_rate": 1.7347741508630673e-05, | |
| "loss": 0.0005, | |
| "step": 396 | |
| }, | |
| { | |
| "epoch": 2.35, | |
| "learning_rate": 1.7333277534725716e-05, | |
| "loss": 0.0037, | |
| "step": 397 | |
| }, | |
| { | |
| "epoch": 2.36, | |
| "learning_rate": 1.731878029381892e-05, | |
| "loss": 0.0004, | |
| "step": 398 | |
| }, | |
| { | |
| "epoch": 2.36, | |
| "learning_rate": 1.730424985167621e-05, | |
| "loss": 0.0004, | |
| "step": 399 | |
| }, | |
| { | |
| "epoch": 2.37, | |
| "learning_rate": 1.7289686274214116e-05, | |
| "loss": 0.0012, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 2.37, | |
| "learning_rate": 1.7275089627499493e-05, | |
| "loss": 0.002, | |
| "step": 401 | |
| }, | |
| { | |
| "epoch": 2.38, | |
| "learning_rate": 1.7260459977749206e-05, | |
| "loss": 0.0035, | |
| "step": 402 | |
| }, | |
| { | |
| "epoch": 2.38, | |
| "learning_rate": 1.7245797391329847e-05, | |
| "loss": 0.0021, | |
| "step": 403 | |
| }, | |
| { | |
| "epoch": 2.39, | |
| "learning_rate": 1.7231101934757406e-05, | |
| "loss": 0.0077, | |
| "step": 404 | |
| }, | |
| { | |
| "epoch": 2.4, | |
| "learning_rate": 1.7216373674697002e-05, | |
| "loss": 0.0065, | |
| "step": 405 | |
| }, | |
| { | |
| "epoch": 2.4, | |
| "learning_rate": 1.720161267796256e-05, | |
| "loss": 0.0008, | |
| "step": 406 | |
| }, | |
| { | |
| "epoch": 2.41, | |
| "learning_rate": 1.7186819011516512e-05, | |
| "loss": 0.0072, | |
| "step": 407 | |
| }, | |
| { | |
| "epoch": 2.41, | |
| "learning_rate": 1.71719927424695e-05, | |
| "loss": 0.0062, | |
| "step": 408 | |
| }, | |
| { | |
| "epoch": 2.42, | |
| "learning_rate": 1.7157133938080054e-05, | |
| "loss": 0.0064, | |
| "step": 409 | |
| }, | |
| { | |
| "epoch": 2.43, | |
| "learning_rate": 1.7142242665754308e-05, | |
| "loss": 0.0007, | |
| "step": 410 | |
| }, | |
| { | |
| "epoch": 2.43, | |
| "learning_rate": 1.7127318993045686e-05, | |
| "loss": 0.0096, | |
| "step": 411 | |
| }, | |
| { | |
| "epoch": 2.44, | |
| "learning_rate": 1.7112362987654587e-05, | |
| "loss": 0.0012, | |
| "step": 412 | |
| }, | |
| { | |
| "epoch": 2.44, | |
| "learning_rate": 1.7097374717428096e-05, | |
| "loss": 0.0011, | |
| "step": 413 | |
| }, | |
| { | |
| "epoch": 2.45, | |
| "learning_rate": 1.708235425035965e-05, | |
| "loss": 0.0031, | |
| "step": 414 | |
| }, | |
| { | |
| "epoch": 2.46, | |
| "learning_rate": 1.7067301654588764e-05, | |
| "loss": 0.0029, | |
| "step": 415 | |
| }, | |
| { | |
| "epoch": 2.46, | |
| "learning_rate": 1.705221699840069e-05, | |
| "loss": 0.0012, | |
| "step": 416 | |
| }, | |
| { | |
| "epoch": 2.47, | |
| "learning_rate": 1.7037100350226118e-05, | |
| "loss": 0.0002, | |
| "step": 417 | |
| }, | |
| { | |
| "epoch": 2.47, | |
| "learning_rate": 1.7021951778640885e-05, | |
| "loss": 0.0073, | |
| "step": 418 | |
| }, | |
| { | |
| "epoch": 2.48, | |
| "learning_rate": 1.7006771352365627e-05, | |
| "loss": 0.0006, | |
| "step": 419 | |
| }, | |
| { | |
| "epoch": 2.49, | |
| "learning_rate": 1.6991559140265498e-05, | |
| "loss": 0.0003, | |
| "step": 420 | |
| }, | |
| { | |
| "epoch": 2.49, | |
| "learning_rate": 1.6976315211349848e-05, | |
| "loss": 0.0017, | |
| "step": 421 | |
| }, | |
| { | |
| "epoch": 2.5, | |
| "learning_rate": 1.69610396347719e-05, | |
| "loss": 0.0047, | |
| "step": 422 | |
| }, | |
| { | |
| "epoch": 2.5, | |
| "learning_rate": 1.694573247982845e-05, | |
| "loss": 0.0022, | |
| "step": 423 | |
| }, | |
| { | |
| "epoch": 2.51, | |
| "learning_rate": 1.693039381595955e-05, | |
| "loss": 0.0031, | |
| "step": 424 | |
| }, | |
| { | |
| "epoch": 2.51, | |
| "learning_rate": 1.6915023712748185e-05, | |
| "loss": 0.0086, | |
| "step": 425 | |
| }, | |
| { | |
| "epoch": 2.52, | |
| "learning_rate": 1.6899622239919965e-05, | |
| "loss": 0.0062, | |
| "step": 426 | |
| }, | |
| { | |
| "epoch": 2.53, | |
| "learning_rate": 1.6884189467342805e-05, | |
| "loss": 0.0038, | |
| "step": 427 | |
| }, | |
| { | |
| "epoch": 2.53, | |
| "learning_rate": 1.6868725465026616e-05, | |
| "loss": 0.01, | |
| "step": 428 | |
| }, | |
| { | |
| "epoch": 2.54, | |
| "learning_rate": 1.6853230303122977e-05, | |
| "loss": 0.0092, | |
| "step": 429 | |
| }, | |
| { | |
| "epoch": 2.54, | |
| "learning_rate": 1.6837704051924814e-05, | |
| "loss": 0.0027, | |
| "step": 430 | |
| }, | |
| { | |
| "epoch": 2.55, | |
| "learning_rate": 1.6822146781866097e-05, | |
| "loss": 0.0088, | |
| "step": 431 | |
| }, | |
| { | |
| "epoch": 2.56, | |
| "learning_rate": 1.680655856352151e-05, | |
| "loss": 0.0078, | |
| "step": 432 | |
| }, | |
| { | |
| "epoch": 2.56, | |
| "learning_rate": 1.679093946760613e-05, | |
| "loss": 0.0016, | |
| "step": 433 | |
| }, | |
| { | |
| "epoch": 2.57, | |
| "learning_rate": 1.6775289564975104e-05, | |
| "loss": 0.0022, | |
| "step": 434 | |
| }, | |
| { | |
| "epoch": 2.57, | |
| "learning_rate": 1.6759608926623344e-05, | |
| "loss": 0.0074, | |
| "step": 435 | |
| }, | |
| { | |
| "epoch": 2.58, | |
| "learning_rate": 1.6743897623685178e-05, | |
| "loss": 0.0104, | |
| "step": 436 | |
| }, | |
| { | |
| "epoch": 2.59, | |
| "learning_rate": 1.6728155727434057e-05, | |
| "loss": 0.0073, | |
| "step": 437 | |
| }, | |
| { | |
| "epoch": 2.59, | |
| "learning_rate": 1.6712383309282208e-05, | |
| "loss": 0.0056, | |
| "step": 438 | |
| }, | |
| { | |
| "epoch": 2.6, | |
| "learning_rate": 1.6696580440780314e-05, | |
| "loss": 0.0021, | |
| "step": 439 | |
| }, | |
| { | |
| "epoch": 2.6, | |
| "learning_rate": 1.6680747193617206e-05, | |
| "loss": 0.0107, | |
| "step": 440 | |
| }, | |
| { | |
| "epoch": 2.61, | |
| "learning_rate": 1.666488363961952e-05, | |
| "loss": 0.0066, | |
| "step": 441 | |
| }, | |
| { | |
| "epoch": 2.62, | |
| "learning_rate": 1.664898985075138e-05, | |
| "loss": 0.0047, | |
| "step": 442 | |
| }, | |
| { | |
| "epoch": 2.62, | |
| "learning_rate": 1.6633065899114065e-05, | |
| "loss": 0.0031, | |
| "step": 443 | |
| }, | |
| { | |
| "epoch": 2.63, | |
| "learning_rate": 1.6617111856945686e-05, | |
| "loss": 0.0043, | |
| "step": 444 | |
| }, | |
| { | |
| "epoch": 2.63, | |
| "learning_rate": 1.6601127796620863e-05, | |
| "loss": 0.006, | |
| "step": 445 | |
| }, | |
| { | |
| "epoch": 2.64, | |
| "learning_rate": 1.658511379065039e-05, | |
| "loss": 0.0036, | |
| "step": 446 | |
| }, | |
| { | |
| "epoch": 2.64, | |
| "learning_rate": 1.6569069911680896e-05, | |
| "loss": 0.0052, | |
| "step": 447 | |
| }, | |
| { | |
| "epoch": 2.65, | |
| "learning_rate": 1.655299623249455e-05, | |
| "loss": 0.0015, | |
| "step": 448 | |
| }, | |
| { | |
| "epoch": 2.66, | |
| "learning_rate": 1.6536892826008687e-05, | |
| "loss": 0.0009, | |
| "step": 449 | |
| }, | |
| { | |
| "epoch": 2.66, | |
| "learning_rate": 1.652075976527551e-05, | |
| "loss": 0.0033, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 2.67, | |
| "learning_rate": 1.6504597123481737e-05, | |
| "loss": 0.003, | |
| "step": 451 | |
| }, | |
| { | |
| "epoch": 2.67, | |
| "learning_rate": 1.6488404973948292e-05, | |
| "loss": 0.009, | |
| "step": 452 | |
| }, | |
| { | |
| "epoch": 2.68, | |
| "learning_rate": 1.6472183390129953e-05, | |
| "loss": 0.0026, | |
| "step": 453 | |
| }, | |
| { | |
| "epoch": 2.69, | |
| "learning_rate": 1.645593244561502e-05, | |
| "loss": 0.0039, | |
| "step": 454 | |
| }, | |
| { | |
| "epoch": 2.69, | |
| "learning_rate": 1.6439652214124992e-05, | |
| "loss": 0.0016, | |
| "step": 455 | |
| }, | |
| { | |
| "epoch": 2.7, | |
| "learning_rate": 1.6423342769514227e-05, | |
| "loss": 0.0014, | |
| "step": 456 | |
| }, | |
| { | |
| "epoch": 2.7, | |
| "learning_rate": 1.64070041857696e-05, | |
| "loss": 0.0012, | |
| "step": 457 | |
| }, | |
| { | |
| "epoch": 2.71, | |
| "learning_rate": 1.6390636537010188e-05, | |
| "loss": 0.0083, | |
| "step": 458 | |
| }, | |
| { | |
| "epoch": 2.72, | |
| "learning_rate": 1.63742398974869e-05, | |
| "loss": 0.0032, | |
| "step": 459 | |
| }, | |
| { | |
| "epoch": 2.72, | |
| "learning_rate": 1.6357814341582173e-05, | |
| "loss": 0.002, | |
| "step": 460 | |
| }, | |
| { | |
| "epoch": 2.73, | |
| "learning_rate": 1.6341359943809626e-05, | |
| "loss": 0.0037, | |
| "step": 461 | |
| }, | |
| { | |
| "epoch": 2.73, | |
| "learning_rate": 1.6324876778813704e-05, | |
| "loss": 0.0019, | |
| "step": 462 | |
| }, | |
| { | |
| "epoch": 2.74, | |
| "learning_rate": 1.6308364921369365e-05, | |
| "loss": 0.0031, | |
| "step": 463 | |
| }, | |
| { | |
| "epoch": 2.75, | |
| "learning_rate": 1.629182444638172e-05, | |
| "loss": 0.0025, | |
| "step": 464 | |
| }, | |
| { | |
| "epoch": 2.75, | |
| "learning_rate": 1.62752554288857e-05, | |
| "loss": 0.0002, | |
| "step": 465 | |
| }, | |
| { | |
| "epoch": 2.76, | |
| "learning_rate": 1.625865794404573e-05, | |
| "loss": 0.0008, | |
| "step": 466 | |
| }, | |
| { | |
| "epoch": 2.76, | |
| "learning_rate": 1.624203206715537e-05, | |
| "loss": 0.0009, | |
| "step": 467 | |
| }, | |
| { | |
| "epoch": 2.77, | |
| "learning_rate": 1.622537787363697e-05, | |
| "loss": 0.0008, | |
| "step": 468 | |
| }, | |
| { | |
| "epoch": 2.78, | |
| "learning_rate": 1.620869543904135e-05, | |
| "loss": 0.0048, | |
| "step": 469 | |
| }, | |
| { | |
| "epoch": 2.78, | |
| "learning_rate": 1.6191984839047435e-05, | |
| "loss": 0.003, | |
| "step": 470 | |
| }, | |
| { | |
| "epoch": 2.79, | |
| "learning_rate": 1.617524614946192e-05, | |
| "loss": 0.0043, | |
| "step": 471 | |
| }, | |
| { | |
| "epoch": 2.79, | |
| "learning_rate": 1.615847944621894e-05, | |
| "loss": 0.0069, | |
| "step": 472 | |
| }, | |
| { | |
| "epoch": 2.8, | |
| "learning_rate": 1.6141684805379693e-05, | |
| "loss": 0.0011, | |
| "step": 473 | |
| }, | |
| { | |
| "epoch": 2.8, | |
| "learning_rate": 1.612486230313213e-05, | |
| "loss": 0.0057, | |
| "step": 474 | |
| }, | |
| { | |
| "epoch": 2.81, | |
| "learning_rate": 1.6108012015790587e-05, | |
| "loss": 0.0015, | |
| "step": 475 | |
| }, | |
| { | |
| "epoch": 2.82, | |
| "learning_rate": 1.6091134019795447e-05, | |
| "loss": 0.0038, | |
| "step": 476 | |
| }, | |
| { | |
| "epoch": 2.82, | |
| "learning_rate": 1.6074228391712788e-05, | |
| "loss": 0.001, | |
| "step": 477 | |
| }, | |
| { | |
| "epoch": 2.83, | |
| "learning_rate": 1.605729520823405e-05, | |
| "loss": 0.0035, | |
| "step": 478 | |
| }, | |
| { | |
| "epoch": 2.83, | |
| "learning_rate": 1.6040334546175663e-05, | |
| "loss": 0.0008, | |
| "step": 479 | |
| }, | |
| { | |
| "epoch": 2.84, | |
| "learning_rate": 1.6023346482478714e-05, | |
| "loss": 0.0023, | |
| "step": 480 | |
| }, | |
| { | |
| "epoch": 2.85, | |
| "learning_rate": 1.600633109420861e-05, | |
| "loss": 0.0004, | |
| "step": 481 | |
| }, | |
| { | |
| "epoch": 2.85, | |
| "learning_rate": 1.5989288458554703e-05, | |
| "loss": 0.0037, | |
| "step": 482 | |
| }, | |
| { | |
| "epoch": 2.86, | |
| "learning_rate": 1.5972218652829948e-05, | |
| "loss": 0.0021, | |
| "step": 483 | |
| }, | |
| { | |
| "epoch": 2.86, | |
| "learning_rate": 1.5955121754470562e-05, | |
| "loss": 0.0011, | |
| "step": 484 | |
| }, | |
| { | |
| "epoch": 2.87, | |
| "learning_rate": 1.5937997841035667e-05, | |
| "loss": 0.0016, | |
| "step": 485 | |
| }, | |
| { | |
| "epoch": 2.88, | |
| "learning_rate": 1.5920846990206934e-05, | |
| "loss": 0.0007, | |
| "step": 486 | |
| }, | |
| { | |
| "epoch": 2.88, | |
| "learning_rate": 1.5903669279788234e-05, | |
| "loss": 0.0009, | |
| "step": 487 | |
| }, | |
| { | |
| "epoch": 2.89, | |
| "learning_rate": 1.5886464787705285e-05, | |
| "loss": 0.0034, | |
| "step": 488 | |
| }, | |
| { | |
| "epoch": 2.89, | |
| "learning_rate": 1.58692335920053e-05, | |
| "loss": 0.0004, | |
| "step": 489 | |
| }, | |
| { | |
| "epoch": 2.9, | |
| "learning_rate": 1.585197577085663e-05, | |
| "loss": 0.0004, | |
| "step": 490 | |
| }, | |
| { | |
| "epoch": 2.91, | |
| "learning_rate": 1.5834691402548415e-05, | |
| "loss": 0.0007, | |
| "step": 491 | |
| }, | |
| { | |
| "epoch": 2.91, | |
| "learning_rate": 1.581738056549022e-05, | |
| "loss": 0.0006, | |
| "step": 492 | |
| }, | |
| { | |
| "epoch": 2.92, | |
| "learning_rate": 1.5800043338211683e-05, | |
| "loss": 0.0025, | |
| "step": 493 | |
| }, | |
| { | |
| "epoch": 2.92, | |
| "learning_rate": 1.578267979936217e-05, | |
| "loss": 0.0019, | |
| "step": 494 | |
| }, | |
| { | |
| "epoch": 2.93, | |
| "learning_rate": 1.576529002771039e-05, | |
| "loss": 0.0006, | |
| "step": 495 | |
| }, | |
| { | |
| "epoch": 2.93, | |
| "learning_rate": 1.5747874102144073e-05, | |
| "loss": 0.0012, | |
| "step": 496 | |
| }, | |
| { | |
| "epoch": 2.94, | |
| "learning_rate": 1.573043210166958e-05, | |
| "loss": 0.0013, | |
| "step": 497 | |
| }, | |
| { | |
| "epoch": 2.95, | |
| "learning_rate": 1.5712964105411575e-05, | |
| "loss": 0.0023, | |
| "step": 498 | |
| }, | |
| { | |
| "epoch": 2.95, | |
| "learning_rate": 1.569547019261264e-05, | |
| "loss": 0.0002, | |
| "step": 499 | |
| }, | |
| { | |
| "epoch": 2.96, | |
| "learning_rate": 1.5677950442632924e-05, | |
| "loss": 0.0004, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 2.96, | |
| "learning_rate": 1.5660404934949798e-05, | |
| "loss": 0.0057, | |
| "step": 501 | |
| }, | |
| { | |
| "epoch": 2.97, | |
| "learning_rate": 1.5642833749157466e-05, | |
| "loss": 0.0012, | |
| "step": 502 | |
| }, | |
| { | |
| "epoch": 2.98, | |
| "learning_rate": 1.5625236964966624e-05, | |
| "loss": 0.0019, | |
| "step": 503 | |
| }, | |
| { | |
| "epoch": 2.98, | |
| "learning_rate": 1.56076146622041e-05, | |
| "loss": 0.0048, | |
| "step": 504 | |
| }, | |
| { | |
| "epoch": 2.99, | |
| "learning_rate": 1.5589966920812483e-05, | |
| "loss": 0.0001, | |
| "step": 505 | |
| }, | |
| { | |
| "epoch": 2.99, | |
| "eval_test_accuracy": 47.016861219195846, | |
| "eval_test_average": 47.016861219195846, | |
| "eval_test_loss": 1.3707863092422485, | |
| "eval_test_runtime": 1249.8978, | |
| "eval_test_samples_per_second": 1.234, | |
| "eval_test_steps_per_second": 0.01, | |
| "step": 505 | |
| }, | |
| { | |
| "epoch": 2.99, | |
| "learning_rate": 1.5572293820849754e-05, | |
| "loss": 0.0003, | |
| "step": 506 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "learning_rate": 1.5554595442488946e-05, | |
| "loss": 0.0059, | |
| "step": 507 | |
| }, | |
| { | |
| "epoch": 3.01, | |
| "learning_rate": 1.5536871866017752e-05, | |
| "loss": 0.0004, | |
| "step": 508 | |
| }, | |
| { | |
| "epoch": 3.01, | |
| "learning_rate": 1.5519123171838185e-05, | |
| "loss": 0.0009, | |
| "step": 509 | |
| }, | |
| { | |
| "epoch": 3.02, | |
| "learning_rate": 1.5501349440466203e-05, | |
| "loss": 0.0008, | |
| "step": 510 | |
| }, | |
| { | |
| "epoch": 3.02, | |
| "learning_rate": 1.5483550752531337e-05, | |
| "loss": 0.0029, | |
| "step": 511 | |
| }, | |
| { | |
| "epoch": 3.03, | |
| "learning_rate": 1.5465727188776333e-05, | |
| "loss": 0.0032, | |
| "step": 512 | |
| }, | |
| { | |
| "epoch": 3.04, | |
| "learning_rate": 1.5447878830056792e-05, | |
| "loss": 0.0104, | |
| "step": 513 | |
| }, | |
| { | |
| "epoch": 3.04, | |
| "learning_rate": 1.5430005757340794e-05, | |
| "loss": 0.0002, | |
| "step": 514 | |
| }, | |
| { | |
| "epoch": 3.05, | |
| "learning_rate": 1.5412108051708522e-05, | |
| "loss": 0.0002, | |
| "step": 515 | |
| }, | |
| { | |
| "epoch": 3.05, | |
| "learning_rate": 1.5394185794351914e-05, | |
| "loss": 0.0008, | |
| "step": 516 | |
| }, | |
| { | |
| "epoch": 3.06, | |
| "learning_rate": 1.537623906657429e-05, | |
| "loss": 0.001, | |
| "step": 517 | |
| }, | |
| { | |
| "epoch": 3.07, | |
| "learning_rate": 1.5358267949789968e-05, | |
| "loss": 0.002, | |
| "step": 518 | |
| }, | |
| { | |
| "epoch": 3.07, | |
| "learning_rate": 1.5340272525523912e-05, | |
| "loss": 0.0013, | |
| "step": 519 | |
| }, | |
| { | |
| "epoch": 3.08, | |
| "learning_rate": 1.5322252875411356e-05, | |
| "loss": 0.0035, | |
| "step": 520 | |
| }, | |
| { | |
| "epoch": 3.08, | |
| "learning_rate": 1.5304209081197425e-05, | |
| "loss": 0.0005, | |
| "step": 521 | |
| }, | |
| { | |
| "epoch": 3.09, | |
| "learning_rate": 1.5286141224736785e-05, | |
| "loss": 0.0003, | |
| "step": 522 | |
| }, | |
| { | |
| "epoch": 3.09, | |
| "learning_rate": 1.526804938799325e-05, | |
| "loss": 0.0006, | |
| "step": 523 | |
| }, | |
| { | |
| "epoch": 3.1, | |
| "learning_rate": 1.5249933653039424e-05, | |
| "loss": 0.0067, | |
| "step": 524 | |
| }, | |
| { | |
| "epoch": 3.11, | |
| "learning_rate": 1.5231794102056317e-05, | |
| "loss": 0.0001, | |
| "step": 525 | |
| }, | |
| { | |
| "epoch": 3.11, | |
| "learning_rate": 1.5213630817332985e-05, | |
| "loss": 0.0006, | |
| "step": 526 | |
| }, | |
| { | |
| "epoch": 3.12, | |
| "learning_rate": 1.5195443881266153e-05, | |
| "loss": 0.0012, | |
| "step": 527 | |
| }, | |
| { | |
| "epoch": 3.12, | |
| "learning_rate": 1.517723337635983e-05, | |
| "loss": 0.0009, | |
| "step": 528 | |
| }, | |
| { | |
| "epoch": 3.13, | |
| "learning_rate": 1.5158999385224956e-05, | |
| "loss": 0.0007, | |
| "step": 529 | |
| }, | |
| { | |
| "epoch": 3.14, | |
| "learning_rate": 1.5140741990578997e-05, | |
| "loss": 0.0001, | |
| "step": 530 | |
| }, | |
| { | |
| "epoch": 3.14, | |
| "learning_rate": 1.512246127524561e-05, | |
| "loss": 0.0003, | |
| "step": 531 | |
| }, | |
| { | |
| "epoch": 3.15, | |
| "learning_rate": 1.5104157322154232e-05, | |
| "loss": 0.0003, | |
| "step": 532 | |
| }, | |
| { | |
| "epoch": 3.15, | |
| "learning_rate": 1.5085830214339715e-05, | |
| "loss": 0.0022, | |
| "step": 533 | |
| }, | |
| { | |
| "epoch": 3.16, | |
| "learning_rate": 1.5067480034941952e-05, | |
| "loss": 0.0047, | |
| "step": 534 | |
| }, | |
| { | |
| "epoch": 3.17, | |
| "learning_rate": 1.5049106867205509e-05, | |
| "loss": 0.0005, | |
| "step": 535 | |
| }, | |
| { | |
| "epoch": 3.17, | |
| "learning_rate": 1.5030710794479226e-05, | |
| "loss": 0.0007, | |
| "step": 536 | |
| }, | |
| { | |
| "epoch": 3.18, | |
| "learning_rate": 1.5012291900215855e-05, | |
| "loss": 0.0007, | |
| "step": 537 | |
| }, | |
| { | |
| "epoch": 3.18, | |
| "learning_rate": 1.4993850267971672e-05, | |
| "loss": 0.0013, | |
| "step": 538 | |
| }, | |
| { | |
| "epoch": 3.19, | |
| "learning_rate": 1.4975385981406115e-05, | |
| "loss": 0.0036, | |
| "step": 539 | |
| }, | |
| { | |
| "epoch": 3.2, | |
| "learning_rate": 1.4956899124281376e-05, | |
| "loss": 0.0005, | |
| "step": 540 | |
| }, | |
| { | |
| "epoch": 3.2, | |
| "learning_rate": 1.4938389780462044e-05, | |
| "loss": 0.0056, | |
| "step": 541 | |
| }, | |
| { | |
| "epoch": 3.21, | |
| "learning_rate": 1.4919858033914723e-05, | |
| "loss": 0.0003, | |
| "step": 542 | |
| }, | |
| { | |
| "epoch": 3.21, | |
| "learning_rate": 1.4901303968707638e-05, | |
| "loss": 0.0025, | |
| "step": 543 | |
| }, | |
| { | |
| "epoch": 3.22, | |
| "learning_rate": 1.4882727669010269e-05, | |
| "loss": 0.0036, | |
| "step": 544 | |
| }, | |
| { | |
| "epoch": 3.22, | |
| "learning_rate": 1.4864129219092946e-05, | |
| "loss": 0.0018, | |
| "step": 545 | |
| }, | |
| { | |
| "epoch": 3.23, | |
| "learning_rate": 1.4845508703326504e-05, | |
| "loss": 0.0005, | |
| "step": 546 | |
| }, | |
| { | |
| "epoch": 3.24, | |
| "learning_rate": 1.482686620618186e-05, | |
| "loss": 0.0002, | |
| "step": 547 | |
| }, | |
| { | |
| "epoch": 3.24, | |
| "learning_rate": 1.4808201812229659e-05, | |
| "loss": 0.0005, | |
| "step": 548 | |
| }, | |
| { | |
| "epoch": 3.25, | |
| "learning_rate": 1.4789515606139878e-05, | |
| "loss": 0.0007, | |
| "step": 549 | |
| }, | |
| { | |
| "epoch": 3.25, | |
| "learning_rate": 1.4770807672681435e-05, | |
| "loss": 0.0003, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 3.26, | |
| "learning_rate": 1.4752078096721827e-05, | |
| "loss": 0.0004, | |
| "step": 551 | |
| }, | |
| { | |
| "epoch": 3.27, | |
| "learning_rate": 1.4733326963226722e-05, | |
| "loss": 0.0001, | |
| "step": 552 | |
| }, | |
| { | |
| "epoch": 3.27, | |
| "learning_rate": 1.471455435725958e-05, | |
| "loss": 0.0012, | |
| "step": 553 | |
| }, | |
| { | |
| "epoch": 3.28, | |
| "learning_rate": 1.469576036398128e-05, | |
| "loss": 0.0001, | |
| "step": 554 | |
| }, | |
| { | |
| "epoch": 3.28, | |
| "learning_rate": 1.4676945068649718e-05, | |
| "loss": 0.0012, | |
| "step": 555 | |
| }, | |
| { | |
| "epoch": 3.29, | |
| "learning_rate": 1.4658108556619417e-05, | |
| "loss": 0.0017, | |
| "step": 556 | |
| }, | |
| { | |
| "epoch": 3.3, | |
| "learning_rate": 1.4639250913341164e-05, | |
| "loss": 0.0001, | |
| "step": 557 | |
| }, | |
| { | |
| "epoch": 3.3, | |
| "learning_rate": 1.4620372224361594e-05, | |
| "loss": 0.0001, | |
| "step": 558 | |
| }, | |
| { | |
| "epoch": 3.31, | |
| "learning_rate": 1.4601472575322825e-05, | |
| "loss": 0.0002, | |
| "step": 559 | |
| }, | |
| { | |
| "epoch": 3.31, | |
| "learning_rate": 1.4582552051962051e-05, | |
| "loss": 0.0, | |
| "step": 560 | |
| }, | |
| { | |
| "epoch": 3.32, | |
| "learning_rate": 1.4563610740111163e-05, | |
| "loss": 0.0005, | |
| "step": 561 | |
| }, | |
| { | |
| "epoch": 3.33, | |
| "learning_rate": 1.454464872569636e-05, | |
| "loss": 0.0004, | |
| "step": 562 | |
| }, | |
| { | |
| "epoch": 3.33, | |
| "learning_rate": 1.4525666094737751e-05, | |
| "loss": 0.0002, | |
| "step": 563 | |
| }, | |
| { | |
| "epoch": 3.34, | |
| "learning_rate": 1.4506662933348987e-05, | |
| "loss": 0.0006, | |
| "step": 564 | |
| }, | |
| { | |
| "epoch": 3.34, | |
| "learning_rate": 1.448763932773683e-05, | |
| "loss": 0.0003, | |
| "step": 565 | |
| }, | |
| { | |
| "epoch": 3.35, | |
| "learning_rate": 1.4468595364200808e-05, | |
| "loss": 0.0001, | |
| "step": 566 | |
| }, | |
| { | |
| "epoch": 3.36, | |
| "learning_rate": 1.444953112913279e-05, | |
| "loss": 0.0002, | |
| "step": 567 | |
| }, | |
| { | |
| "epoch": 3.36, | |
| "learning_rate": 1.4430446709016605e-05, | |
| "loss": 0.0043, | |
| "step": 568 | |
| }, | |
| { | |
| "epoch": 3.37, | |
| "learning_rate": 1.4411342190427657e-05, | |
| "loss": 0.003, | |
| "step": 569 | |
| }, | |
| { | |
| "epoch": 3.37, | |
| "learning_rate": 1.4392217660032522e-05, | |
| "loss": 0.0, | |
| "step": 570 | |
| }, | |
| { | |
| "epoch": 3.38, | |
| "learning_rate": 1.4373073204588556e-05, | |
| "loss": 0.0, | |
| "step": 571 | |
| }, | |
| { | |
| "epoch": 3.38, | |
| "learning_rate": 1.4353908910943508e-05, | |
| "loss": 0.0024, | |
| "step": 572 | |
| }, | |
| { | |
| "epoch": 3.39, | |
| "learning_rate": 1.433472486603512e-05, | |
| "loss": 0.0037, | |
| "step": 573 | |
| }, | |
| { | |
| "epoch": 3.4, | |
| "learning_rate": 1.431552115689073e-05, | |
| "loss": 0.0009, | |
| "step": 574 | |
| }, | |
| { | |
| "epoch": 3.4, | |
| "learning_rate": 1.4296297870626897e-05, | |
| "loss": 0.0002, | |
| "step": 575 | |
| }, | |
| { | |
| "epoch": 3.41, | |
| "learning_rate": 1.427705509444897e-05, | |
| "loss": 0.0002, | |
| "step": 576 | |
| }, | |
| { | |
| "epoch": 3.41, | |
| "learning_rate": 1.4257792915650728e-05, | |
| "loss": 0.0003, | |
| "step": 577 | |
| }, | |
| { | |
| "epoch": 3.42, | |
| "learning_rate": 1.4238511421613962e-05, | |
| "loss": 0.0002, | |
| "step": 578 | |
| }, | |
| { | |
| "epoch": 3.43, | |
| "learning_rate": 1.4219210699808087e-05, | |
| "loss": 0.0014, | |
| "step": 579 | |
| }, | |
| { | |
| "epoch": 3.43, | |
| "learning_rate": 1.4199890837789746e-05, | |
| "loss": 0.001, | |
| "step": 580 | |
| }, | |
| { | |
| "epoch": 3.44, | |
| "learning_rate": 1.4180551923202406e-05, | |
| "loss": 0.0008, | |
| "step": 581 | |
| }, | |
| { | |
| "epoch": 3.44, | |
| "learning_rate": 1.4161194043775968e-05, | |
| "loss": 0.0002, | |
| "step": 582 | |
| }, | |
| { | |
| "epoch": 3.45, | |
| "learning_rate": 1.4141817287326366e-05, | |
| "loss": 0.0, | |
| "step": 583 | |
| }, | |
| { | |
| "epoch": 3.46, | |
| "learning_rate": 1.4122421741755166e-05, | |
| "loss": 0.0009, | |
| "step": 584 | |
| }, | |
| { | |
| "epoch": 3.46, | |
| "learning_rate": 1.4103007495049173e-05, | |
| "loss": 0.0005, | |
| "step": 585 | |
| }, | |
| { | |
| "epoch": 3.47, | |
| "learning_rate": 1.4083574635280029e-05, | |
| "loss": 0.0033, | |
| "step": 586 | |
| }, | |
| { | |
| "epoch": 3.47, | |
| "learning_rate": 1.4064123250603807e-05, | |
| "loss": 0.0005, | |
| "step": 587 | |
| }, | |
| { | |
| "epoch": 3.48, | |
| "learning_rate": 1.4044653429260627e-05, | |
| "loss": 0.0002, | |
| "step": 588 | |
| }, | |
| { | |
| "epoch": 3.49, | |
| "learning_rate": 1.4025165259574234e-05, | |
| "loss": 0.0001, | |
| "step": 589 | |
| }, | |
| { | |
| "epoch": 3.49, | |
| "learning_rate": 1.4005658829951619e-05, | |
| "loss": 0.0002, | |
| "step": 590 | |
| }, | |
| { | |
| "epoch": 3.5, | |
| "learning_rate": 1.3986134228882607e-05, | |
| "loss": 0.0007, | |
| "step": 591 | |
| }, | |
| { | |
| "epoch": 3.5, | |
| "learning_rate": 1.3966591544939451e-05, | |
| "loss": 0.001, | |
| "step": 592 | |
| }, | |
| { | |
| "epoch": 3.51, | |
| "learning_rate": 1.394703086677644e-05, | |
| "loss": 0.0002, | |
| "step": 593 | |
| }, | |
| { | |
| "epoch": 3.51, | |
| "learning_rate": 1.3927452283129496e-05, | |
| "loss": 0.0008, | |
| "step": 594 | |
| }, | |
| { | |
| "epoch": 3.52, | |
| "learning_rate": 1.3907855882815762e-05, | |
| "loss": 0.0001, | |
| "step": 595 | |
| }, | |
| { | |
| "epoch": 3.53, | |
| "learning_rate": 1.388824175473321e-05, | |
| "loss": 0.0001, | |
| "step": 596 | |
| }, | |
| { | |
| "epoch": 3.53, | |
| "learning_rate": 1.3868609987860224e-05, | |
| "loss": 0.0002, | |
| "step": 597 | |
| }, | |
| { | |
| "epoch": 3.54, | |
| "learning_rate": 1.3848960671255218e-05, | |
| "loss": 0.0005, | |
| "step": 598 | |
| }, | |
| { | |
| "epoch": 3.54, | |
| "learning_rate": 1.3829293894056214e-05, | |
| "loss": 0.0012, | |
| "step": 599 | |
| }, | |
| { | |
| "epoch": 3.55, | |
| "learning_rate": 1.3809609745480441e-05, | |
| "loss": 0.0017, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 3.56, | |
| "learning_rate": 1.3789908314823932e-05, | |
| "loss": 0.0004, | |
| "step": 601 | |
| }, | |
| { | |
| "epoch": 3.56, | |
| "learning_rate": 1.3770189691461127e-05, | |
| "loss": 0.0057, | |
| "step": 602 | |
| }, | |
| { | |
| "epoch": 3.57, | |
| "learning_rate": 1.375045396484445e-05, | |
| "loss": 0.0002, | |
| "step": 603 | |
| }, | |
| { | |
| "epoch": 3.57, | |
| "learning_rate": 1.3730701224503916e-05, | |
| "loss": 0.0001, | |
| "step": 604 | |
| }, | |
| { | |
| "epoch": 3.58, | |
| "learning_rate": 1.371093156004672e-05, | |
| "loss": 0.0, | |
| "step": 605 | |
| }, | |
| { | |
| "epoch": 3.59, | |
| "learning_rate": 1.3691145061156843e-05, | |
| "loss": 0.0009, | |
| "step": 606 | |
| }, | |
| { | |
| "epoch": 3.59, | |
| "learning_rate": 1.3671341817594622e-05, | |
| "loss": 0.0038, | |
| "step": 607 | |
| }, | |
| { | |
| "epoch": 3.6, | |
| "learning_rate": 1.3651521919196358e-05, | |
| "loss": 0.0003, | |
| "step": 608 | |
| }, | |
| { | |
| "epoch": 3.6, | |
| "learning_rate": 1.3631685455873906e-05, | |
| "loss": 0.0008, | |
| "step": 609 | |
| }, | |
| { | |
| "epoch": 3.61, | |
| "learning_rate": 1.3611832517614273e-05, | |
| "loss": 0.0005, | |
| "step": 610 | |
| }, | |
| { | |
| "epoch": 3.62, | |
| "learning_rate": 1.3591963194479198e-05, | |
| "loss": 0.0041, | |
| "step": 611 | |
| }, | |
| { | |
| "epoch": 3.62, | |
| "learning_rate": 1.3572077576604752e-05, | |
| "loss": 0.0002, | |
| "step": 612 | |
| }, | |
| { | |
| "epoch": 3.63, | |
| "learning_rate": 1.3552175754200917e-05, | |
| "loss": 0.0001, | |
| "step": 613 | |
| }, | |
| { | |
| "epoch": 3.63, | |
| "learning_rate": 1.3532257817551195e-05, | |
| "loss": 0.0001, | |
| "step": 614 | |
| }, | |
| { | |
| "epoch": 3.64, | |
| "learning_rate": 1.3512323857012192e-05, | |
| "loss": 0.0018, | |
| "step": 615 | |
| }, | |
| { | |
| "epoch": 3.64, | |
| "learning_rate": 1.3492373963013199e-05, | |
| "loss": 0.0006, | |
| "step": 616 | |
| }, | |
| { | |
| "epoch": 3.65, | |
| "learning_rate": 1.3472408226055787e-05, | |
| "loss": 0.0013, | |
| "step": 617 | |
| }, | |
| { | |
| "epoch": 3.66, | |
| "learning_rate": 1.3452426736713404e-05, | |
| "loss": 0.0003, | |
| "step": 618 | |
| }, | |
| { | |
| "epoch": 3.66, | |
| "learning_rate": 1.3432429585630958e-05, | |
| "loss": 0.0004, | |
| "step": 619 | |
| }, | |
| { | |
| "epoch": 3.67, | |
| "learning_rate": 1.3412416863524392e-05, | |
| "loss": 0.0004, | |
| "step": 620 | |
| }, | |
| { | |
| "epoch": 3.67, | |
| "learning_rate": 1.3392388661180303e-05, | |
| "loss": 0.0015, | |
| "step": 621 | |
| }, | |
| { | |
| "epoch": 3.68, | |
| "learning_rate": 1.3372345069455508e-05, | |
| "loss": 0.0005, | |
| "step": 622 | |
| }, | |
| { | |
| "epoch": 3.69, | |
| "learning_rate": 1.3352286179276632e-05, | |
| "loss": 0.0002, | |
| "step": 623 | |
| }, | |
| { | |
| "epoch": 3.69, | |
| "learning_rate": 1.3332212081639709e-05, | |
| "loss": 0.0007, | |
| "step": 624 | |
| }, | |
| { | |
| "epoch": 3.7, | |
| "learning_rate": 1.3312122867609749e-05, | |
| "loss": 0.0022, | |
| "step": 625 | |
| }, | |
| { | |
| "epoch": 3.7, | |
| "learning_rate": 1.3292018628320346e-05, | |
| "loss": 0.0, | |
| "step": 626 | |
| }, | |
| { | |
| "epoch": 3.71, | |
| "learning_rate": 1.3271899454973256e-05, | |
| "loss": 0.0006, | |
| "step": 627 | |
| }, | |
| { | |
| "epoch": 3.72, | |
| "learning_rate": 1.3251765438837978e-05, | |
| "loss": 0.0006, | |
| "step": 628 | |
| }, | |
| { | |
| "epoch": 3.72, | |
| "learning_rate": 1.3231616671251341e-05, | |
| "loss": 0.0016, | |
| "step": 629 | |
| }, | |
| { | |
| "epoch": 3.73, | |
| "learning_rate": 1.3211453243617108e-05, | |
| "loss": 0.0004, | |
| "step": 630 | |
| }, | |
| { | |
| "epoch": 3.73, | |
| "learning_rate": 1.3191275247405525e-05, | |
| "loss": 0.0082, | |
| "step": 631 | |
| }, | |
| { | |
| "epoch": 3.74, | |
| "learning_rate": 1.317108277415295e-05, | |
| "loss": 0.0002, | |
| "step": 632 | |
| }, | |
| { | |
| "epoch": 3.75, | |
| "learning_rate": 1.3150875915461395e-05, | |
| "loss": 0.0003, | |
| "step": 633 | |
| }, | |
| { | |
| "epoch": 3.75, | |
| "learning_rate": 1.3130654762998145e-05, | |
| "loss": 0.0001, | |
| "step": 634 | |
| }, | |
| { | |
| "epoch": 3.76, | |
| "learning_rate": 1.3110419408495323e-05, | |
| "loss": 0.0004, | |
| "step": 635 | |
| }, | |
| { | |
| "epoch": 3.76, | |
| "learning_rate": 1.3090169943749475e-05, | |
| "loss": 0.0009, | |
| "step": 636 | |
| }, | |
| { | |
| "epoch": 3.77, | |
| "learning_rate": 1.3069906460621165e-05, | |
| "loss": 0.0034, | |
| "step": 637 | |
| }, | |
| { | |
| "epoch": 3.78, | |
| "learning_rate": 1.3049629051034545e-05, | |
| "loss": 0.0003, | |
| "step": 638 | |
| }, | |
| { | |
| "epoch": 3.78, | |
| "learning_rate": 1.3029337806976944e-05, | |
| "loss": 0.0073, | |
| "step": 639 | |
| }, | |
| { | |
| "epoch": 3.79, | |
| "learning_rate": 1.3009032820498457e-05, | |
| "loss": 0.0003, | |
| "step": 640 | |
| }, | |
| { | |
| "epoch": 3.79, | |
| "learning_rate": 1.2988714183711504e-05, | |
| "loss": 0.0015, | |
| "step": 641 | |
| }, | |
| { | |
| "epoch": 3.8, | |
| "learning_rate": 1.296838198879045e-05, | |
| "loss": 0.0037, | |
| "step": 642 | |
| }, | |
| { | |
| "epoch": 3.8, | |
| "learning_rate": 1.2948036327971156e-05, | |
| "loss": 0.0033, | |
| "step": 643 | |
| }, | |
| { | |
| "epoch": 3.81, | |
| "learning_rate": 1.2927677293550564e-05, | |
| "loss": 0.007, | |
| "step": 644 | |
| }, | |
| { | |
| "epoch": 3.82, | |
| "learning_rate": 1.2907304977886295e-05, | |
| "loss": 0.0029, | |
| "step": 645 | |
| }, | |
| { | |
| "epoch": 3.82, | |
| "learning_rate": 1.2886919473396212e-05, | |
| "loss": 0.0041, | |
| "step": 646 | |
| }, | |
| { | |
| "epoch": 3.83, | |
| "learning_rate": 1.2866520872558013e-05, | |
| "loss": 0.007, | |
| "step": 647 | |
| }, | |
| { | |
| "epoch": 3.83, | |
| "learning_rate": 1.2846109267908808e-05, | |
| "loss": 0.0028, | |
| "step": 648 | |
| }, | |
| { | |
| "epoch": 3.84, | |
| "learning_rate": 1.2825684752044692e-05, | |
| "loss": 0.0008, | |
| "step": 649 | |
| }, | |
| { | |
| "epoch": 3.85, | |
| "learning_rate": 1.2805247417620337e-05, | |
| "loss": 0.0046, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 3.85, | |
| "learning_rate": 1.2784797357348562e-05, | |
| "loss": 0.0015, | |
| "step": 651 | |
| }, | |
| { | |
| "epoch": 3.86, | |
| "learning_rate": 1.2764334663999917e-05, | |
| "loss": 0.0006, | |
| "step": 652 | |
| }, | |
| { | |
| "epoch": 3.86, | |
| "learning_rate": 1.2743859430402262e-05, | |
| "loss": 0.0008, | |
| "step": 653 | |
| }, | |
| { | |
| "epoch": 3.87, | |
| "learning_rate": 1.2723371749440347e-05, | |
| "loss": 0.0002, | |
| "step": 654 | |
| }, | |
| { | |
| "epoch": 3.88, | |
| "learning_rate": 1.2702871714055382e-05, | |
| "loss": 0.0007, | |
| "step": 655 | |
| }, | |
| { | |
| "epoch": 3.88, | |
| "learning_rate": 1.268235941724463e-05, | |
| "loss": 0.0022, | |
| "step": 656 | |
| }, | |
| { | |
| "epoch": 3.89, | |
| "learning_rate": 1.2661834952060971e-05, | |
| "loss": 0.0004, | |
| "step": 657 | |
| }, | |
| { | |
| "epoch": 3.89, | |
| "learning_rate": 1.2641298411612494e-05, | |
| "loss": 0.001, | |
| "step": 658 | |
| }, | |
| { | |
| "epoch": 3.9, | |
| "learning_rate": 1.2620749889062058e-05, | |
| "loss": 0.001, | |
| "step": 659 | |
| }, | |
| { | |
| "epoch": 3.91, | |
| "learning_rate": 1.2600189477626885e-05, | |
| "loss": 0.0005, | |
| "step": 660 | |
| }, | |
| { | |
| "epoch": 3.91, | |
| "learning_rate": 1.257961727057812e-05, | |
| "loss": 0.0, | |
| "step": 661 | |
| }, | |
| { | |
| "epoch": 3.92, | |
| "learning_rate": 1.2559033361240431e-05, | |
| "loss": 0.0001, | |
| "step": 662 | |
| }, | |
| { | |
| "epoch": 3.92, | |
| "learning_rate": 1.2538437842991568e-05, | |
| "loss": 0.0038, | |
| "step": 663 | |
| }, | |
| { | |
| "epoch": 3.93, | |
| "learning_rate": 1.2517830809261942e-05, | |
| "loss": 0.0, | |
| "step": 664 | |
| }, | |
| { | |
| "epoch": 3.93, | |
| "learning_rate": 1.2497212353534202e-05, | |
| "loss": 0.0007, | |
| "step": 665 | |
| }, | |
| { | |
| "epoch": 3.94, | |
| "learning_rate": 1.2476582569342819e-05, | |
| "loss": 0.0005, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 3.95, | |
| "learning_rate": 1.2455941550273647e-05, | |
| "loss": 0.0014, | |
| "step": 667 | |
| }, | |
| { | |
| "epoch": 3.95, | |
| "learning_rate": 1.2435289389963514e-05, | |
| "loss": 0.0005, | |
| "step": 668 | |
| }, | |
| { | |
| "epoch": 3.96, | |
| "learning_rate": 1.2414626182099782e-05, | |
| "loss": 0.0002, | |
| "step": 669 | |
| }, | |
| { | |
| "epoch": 3.96, | |
| "learning_rate": 1.2393952020419937e-05, | |
| "loss": 0.0002, | |
| "step": 670 | |
| }, | |
| { | |
| "epoch": 3.97, | |
| "learning_rate": 1.2373266998711152e-05, | |
| "loss": 0.0006, | |
| "step": 671 | |
| }, | |
| { | |
| "epoch": 3.98, | |
| "learning_rate": 1.235257121080986e-05, | |
| "loss": 0.0016, | |
| "step": 672 | |
| }, | |
| { | |
| "epoch": 3.98, | |
| "learning_rate": 1.233186475060135e-05, | |
| "loss": 0.0003, | |
| "step": 673 | |
| }, | |
| { | |
| "epoch": 3.99, | |
| "learning_rate": 1.2311147712019316e-05, | |
| "loss": 0.0001, | |
| "step": 674 | |
| }, | |
| { | |
| "epoch": 3.99, | |
| "learning_rate": 1.2290420189045436e-05, | |
| "loss": 0.0002, | |
| "step": 675 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "learning_rate": 1.2269682275708951e-05, | |
| "loss": 0.0, | |
| "step": 676 | |
| }, | |
| { | |
| "epoch": 4.01, | |
| "learning_rate": 1.2248934066086251e-05, | |
| "loss": 0.0001, | |
| "step": 677 | |
| }, | |
| { | |
| "epoch": 4.01, | |
| "learning_rate": 1.2228175654300413e-05, | |
| "loss": 0.0007, | |
| "step": 678 | |
| }, | |
| { | |
| "epoch": 4.02, | |
| "learning_rate": 1.2207407134520816e-05, | |
| "loss": 0.0002, | |
| "step": 679 | |
| }, | |
| { | |
| "epoch": 4.02, | |
| "learning_rate": 1.2186628600962682e-05, | |
| "loss": 0.0005, | |
| "step": 680 | |
| }, | |
| { | |
| "epoch": 4.03, | |
| "learning_rate": 1.2165840147886656e-05, | |
| "loss": 0.0003, | |
| "step": 681 | |
| }, | |
| { | |
| "epoch": 4.04, | |
| "learning_rate": 1.2145041869598393e-05, | |
| "loss": 0.0003, | |
| "step": 682 | |
| }, | |
| { | |
| "epoch": 4.04, | |
| "learning_rate": 1.2124233860448113e-05, | |
| "loss": 0.0002, | |
| "step": 683 | |
| }, | |
| { | |
| "epoch": 4.05, | |
| "learning_rate": 1.2103416214830185e-05, | |
| "loss": 0.0012, | |
| "step": 684 | |
| }, | |
| { | |
| "epoch": 4.05, | |
| "learning_rate": 1.2082589027182685e-05, | |
| "loss": 0.0, | |
| "step": 685 | |
| }, | |
| { | |
| "epoch": 4.06, | |
| "learning_rate": 1.2061752391986982e-05, | |
| "loss": 0.0003, | |
| "step": 686 | |
| }, | |
| { | |
| "epoch": 4.07, | |
| "learning_rate": 1.2040906403767306e-05, | |
| "loss": 0.0002, | |
| "step": 687 | |
| }, | |
| { | |
| "epoch": 4.07, | |
| "learning_rate": 1.2020051157090305e-05, | |
| "loss": 0.0007, | |
| "step": 688 | |
| }, | |
| { | |
| "epoch": 4.08, | |
| "learning_rate": 1.1999186746564637e-05, | |
| "loss": 0.0001, | |
| "step": 689 | |
| }, | |
| { | |
| "epoch": 4.08, | |
| "learning_rate": 1.1978313266840536e-05, | |
| "loss": 0.0, | |
| "step": 690 | |
| }, | |
| { | |
| "epoch": 4.09, | |
| "learning_rate": 1.1957430812609361e-05, | |
| "loss": 0.0007, | |
| "step": 691 | |
| }, | |
| { | |
| "epoch": 4.09, | |
| "learning_rate": 1.1936539478603197e-05, | |
| "loss": 0.0002, | |
| "step": 692 | |
| }, | |
| { | |
| "epoch": 4.1, | |
| "learning_rate": 1.1915639359594407e-05, | |
| "loss": 0.0, | |
| "step": 693 | |
| }, | |
| { | |
| "epoch": 4.11, | |
| "learning_rate": 1.1894730550395213e-05, | |
| "loss": 0.0, | |
| "step": 694 | |
| }, | |
| { | |
| "epoch": 4.11, | |
| "learning_rate": 1.187381314585725e-05, | |
| "loss": 0.0, | |
| "step": 695 | |
| }, | |
| { | |
| "epoch": 4.12, | |
| "learning_rate": 1.1852887240871145e-05, | |
| "loss": 0.0001, | |
| "step": 696 | |
| }, | |
| { | |
| "epoch": 4.12, | |
| "learning_rate": 1.1831952930366097e-05, | |
| "loss": 0.0001, | |
| "step": 697 | |
| }, | |
| { | |
| "epoch": 4.13, | |
| "learning_rate": 1.1811010309309427e-05, | |
| "loss": 0.0001, | |
| "step": 698 | |
| }, | |
| { | |
| "epoch": 4.14, | |
| "learning_rate": 1.1790059472706166e-05, | |
| "loss": 0.0, | |
| "step": 699 | |
| }, | |
| { | |
| "epoch": 4.14, | |
| "learning_rate": 1.1769100515598602e-05, | |
| "loss": 0.0001, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 4.15, | |
| "learning_rate": 1.1748133533065864e-05, | |
| "loss": 0.0, | |
| "step": 701 | |
| }, | |
| { | |
| "epoch": 4.15, | |
| "learning_rate": 1.1727158620223503e-05, | |
| "loss": 0.0002, | |
| "step": 702 | |
| }, | |
| { | |
| "epoch": 4.16, | |
| "learning_rate": 1.170617587222302e-05, | |
| "loss": 0.0001, | |
| "step": 703 | |
| }, | |
| { | |
| "epoch": 4.17, | |
| "learning_rate": 1.1685185384251482e-05, | |
| "loss": 0.0, | |
| "step": 704 | |
| }, | |
| { | |
| "epoch": 4.17, | |
| "learning_rate": 1.1664187251531055e-05, | |
| "loss": 0.0, | |
| "step": 705 | |
| }, | |
| { | |
| "epoch": 4.18, | |
| "learning_rate": 1.1643181569318596e-05, | |
| "loss": 0.0, | |
| "step": 706 | |
| }, | |
| { | |
| "epoch": 4.18, | |
| "learning_rate": 1.1622168432905196e-05, | |
| "loss": 0.0001, | |
| "step": 707 | |
| }, | |
| { | |
| "epoch": 4.19, | |
| "learning_rate": 1.1601147937615769e-05, | |
| "loss": 0.0002, | |
| "step": 708 | |
| }, | |
| { | |
| "epoch": 4.2, | |
| "learning_rate": 1.1580120178808614e-05, | |
| "loss": 0.0, | |
| "step": 709 | |
| }, | |
| { | |
| "epoch": 4.2, | |
| "learning_rate": 1.1559085251874979e-05, | |
| "loss": 0.0, | |
| "step": 710 | |
| }, | |
| { | |
| "epoch": 4.21, | |
| "learning_rate": 1.1538043252238629e-05, | |
| "loss": 0.0002, | |
| "step": 711 | |
| }, | |
| { | |
| "epoch": 4.21, | |
| "learning_rate": 1.151699427535541e-05, | |
| "loss": 0.0, | |
| "step": 712 | |
| }, | |
| { | |
| "epoch": 4.22, | |
| "learning_rate": 1.1495938416712829e-05, | |
| "loss": 0.0, | |
| "step": 713 | |
| }, | |
| { | |
| "epoch": 4.22, | |
| "learning_rate": 1.1474875771829604e-05, | |
| "loss": 0.0021, | |
| "step": 714 | |
| }, | |
| { | |
| "epoch": 4.23, | |
| "learning_rate": 1.1453806436255241e-05, | |
| "loss": 0.0, | |
| "step": 715 | |
| }, | |
| { | |
| "epoch": 4.24, | |
| "learning_rate": 1.1432730505569597e-05, | |
| "loss": 0.0006, | |
| "step": 716 | |
| }, | |
| { | |
| "epoch": 4.24, | |
| "learning_rate": 1.1411648075382455e-05, | |
| "loss": 0.0, | |
| "step": 717 | |
| }, | |
| { | |
| "epoch": 4.25, | |
| "learning_rate": 1.139055924133307e-05, | |
| "loss": 0.0001, | |
| "step": 718 | |
| }, | |
| { | |
| "epoch": 4.25, | |
| "learning_rate": 1.1369464099089751e-05, | |
| "loss": 0.0005, | |
| "step": 719 | |
| }, | |
| { | |
| "epoch": 4.26, | |
| "learning_rate": 1.1348362744349432e-05, | |
| "loss": 0.0001, | |
| "step": 720 | |
| }, | |
| { | |
| "epoch": 4.27, | |
| "learning_rate": 1.1327255272837221e-05, | |
| "loss": 0.0003, | |
| "step": 721 | |
| }, | |
| { | |
| "epoch": 4.27, | |
| "learning_rate": 1.130614178030598e-05, | |
| "loss": 0.0, | |
| "step": 722 | |
| }, | |
| { | |
| "epoch": 4.28, | |
| "learning_rate": 1.1285022362535883e-05, | |
| "loss": 0.0002, | |
| "step": 723 | |
| }, | |
| { | |
| "epoch": 4.28, | |
| "learning_rate": 1.1263897115333977e-05, | |
| "loss": 0.0001, | |
| "step": 724 | |
| }, | |
| { | |
| "epoch": 4.29, | |
| "learning_rate": 1.1242766134533764e-05, | |
| "loss": 0.0, | |
| "step": 725 | |
| }, | |
| { | |
| "epoch": 4.3, | |
| "learning_rate": 1.1221629515994754e-05, | |
| "loss": 0.0001, | |
| "step": 726 | |
| }, | |
| { | |
| "epoch": 4.3, | |
| "learning_rate": 1.120048735560203e-05, | |
| "loss": 0.0001, | |
| "step": 727 | |
| }, | |
| { | |
| "epoch": 4.31, | |
| "learning_rate": 1.1179339749265812e-05, | |
| "loss": 0.0039, | |
| "step": 728 | |
| }, | |
| { | |
| "epoch": 4.31, | |
| "learning_rate": 1.1158186792921031e-05, | |
| "loss": 0.0, | |
| "step": 729 | |
| }, | |
| { | |
| "epoch": 4.32, | |
| "learning_rate": 1.1137028582526888e-05, | |
| "loss": 0.0011, | |
| "step": 730 | |
| }, | |
| { | |
| "epoch": 4.33, | |
| "learning_rate": 1.1115865214066414e-05, | |
| "loss": 0.0001, | |
| "step": 731 | |
| }, | |
| { | |
| "epoch": 4.33, | |
| "learning_rate": 1.1094696783546042e-05, | |
| "loss": 0.0, | |
| "step": 732 | |
| }, | |
| { | |
| "epoch": 4.34, | |
| "learning_rate": 1.107352338699517e-05, | |
| "loss": 0.0, | |
| "step": 733 | |
| }, | |
| { | |
| "epoch": 4.34, | |
| "learning_rate": 1.1052345120465724e-05, | |
| "loss": 0.0, | |
| "step": 734 | |
| }, | |
| { | |
| "epoch": 4.35, | |
| "learning_rate": 1.1031162080031715e-05, | |
| "loss": 0.0008, | |
| "step": 735 | |
| }, | |
| { | |
| "epoch": 4.36, | |
| "learning_rate": 1.1009974361788822e-05, | |
| "loss": 0.0, | |
| "step": 736 | |
| }, | |
| { | |
| "epoch": 4.36, | |
| "learning_rate": 1.098878206185394e-05, | |
| "loss": 0.0064, | |
| "step": 737 | |
| }, | |
| { | |
| "epoch": 4.37, | |
| "learning_rate": 1.0967585276364745e-05, | |
| "loss": 0.0009, | |
| "step": 738 | |
| }, | |
| { | |
| "epoch": 4.37, | |
| "learning_rate": 1.0946384101479265e-05, | |
| "loss": 0.0005, | |
| "step": 739 | |
| }, | |
| { | |
| "epoch": 4.38, | |
| "learning_rate": 1.0925178633375437e-05, | |
| "loss": 0.0, | |
| "step": 740 | |
| }, | |
| { | |
| "epoch": 4.38, | |
| "learning_rate": 1.0903968968250682e-05, | |
| "loss": 0.0037, | |
| "step": 741 | |
| }, | |
| { | |
| "epoch": 4.39, | |
| "learning_rate": 1.0882755202321453e-05, | |
| "loss": 0.0005, | |
| "step": 742 | |
| }, | |
| { | |
| "epoch": 4.4, | |
| "learning_rate": 1.0861537431822808e-05, | |
| "loss": 0.0003, | |
| "step": 743 | |
| }, | |
| { | |
| "epoch": 4.4, | |
| "learning_rate": 1.084031575300797e-05, | |
| "loss": 0.0001, | |
| "step": 744 | |
| }, | |
| { | |
| "epoch": 4.41, | |
| "learning_rate": 1.0819090262147895e-05, | |
| "loss": 0.0008, | |
| "step": 745 | |
| }, | |
| { | |
| "epoch": 4.41, | |
| "learning_rate": 1.0797861055530832e-05, | |
| "loss": 0.0, | |
| "step": 746 | |
| }, | |
| { | |
| "epoch": 4.42, | |
| "learning_rate": 1.0776628229461882e-05, | |
| "loss": 0.0002, | |
| "step": 747 | |
| }, | |
| { | |
| "epoch": 4.43, | |
| "learning_rate": 1.075539188026257e-05, | |
| "loss": 0.0, | |
| "step": 748 | |
| }, | |
| { | |
| "epoch": 4.43, | |
| "learning_rate": 1.0734152104270406e-05, | |
| "loss": 0.001, | |
| "step": 749 | |
| }, | |
| { | |
| "epoch": 4.44, | |
| "learning_rate": 1.0712908997838431e-05, | |
| "loss": 0.0, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 4.44, | |
| "learning_rate": 1.0691662657334815e-05, | |
| "loss": 0.0077, | |
| "step": 751 | |
| }, | |
| { | |
| "epoch": 4.45, | |
| "learning_rate": 1.0670413179142382e-05, | |
| "loss": 0.001, | |
| "step": 752 | |
| }, | |
| { | |
| "epoch": 4.46, | |
| "learning_rate": 1.0649160659658204e-05, | |
| "loss": 0.0002, | |
| "step": 753 | |
| }, | |
| { | |
| "epoch": 4.46, | |
| "learning_rate": 1.0627905195293135e-05, | |
| "loss": 0.0002, | |
| "step": 754 | |
| }, | |
| { | |
| "epoch": 4.47, | |
| "learning_rate": 1.0606646882471398e-05, | |
| "loss": 0.0009, | |
| "step": 755 | |
| }, | |
| { | |
| "epoch": 4.47, | |
| "learning_rate": 1.0585385817630137e-05, | |
| "loss": 0.0004, | |
| "step": 756 | |
| }, | |
| { | |
| "epoch": 4.48, | |
| "learning_rate": 1.0564122097218984e-05, | |
| "loss": 0.0042, | |
| "step": 757 | |
| }, | |
| { | |
| "epoch": 4.49, | |
| "learning_rate": 1.0542855817699602e-05, | |
| "loss": 0.0014, | |
| "step": 758 | |
| }, | |
| { | |
| "epoch": 4.49, | |
| "learning_rate": 1.0521587075545283e-05, | |
| "loss": 0.0, | |
| "step": 759 | |
| }, | |
| { | |
| "epoch": 4.5, | |
| "learning_rate": 1.0500315967240476e-05, | |
| "loss": 0.001, | |
| "step": 760 | |
| }, | |
| { | |
| "epoch": 4.5, | |
| "learning_rate": 1.047904258928037e-05, | |
| "loss": 0.0001, | |
| "step": 761 | |
| }, | |
| { | |
| "epoch": 4.51, | |
| "learning_rate": 1.0457767038170457e-05, | |
| "loss": 0.0, | |
| "step": 762 | |
| }, | |
| { | |
| "epoch": 4.51, | |
| "learning_rate": 1.0436489410426071e-05, | |
| "loss": 0.0002, | |
| "step": 763 | |
| }, | |
| { | |
| "epoch": 4.52, | |
| "learning_rate": 1.0415209802571982e-05, | |
| "loss": 0.0008, | |
| "step": 764 | |
| }, | |
| { | |
| "epoch": 4.53, | |
| "learning_rate": 1.0393928311141934e-05, | |
| "loss": 0.0004, | |
| "step": 765 | |
| }, | |
| { | |
| "epoch": 4.53, | |
| "learning_rate": 1.0372645032678215e-05, | |
| "loss": 0.0025, | |
| "step": 766 | |
| }, | |
| { | |
| "epoch": 4.54, | |
| "learning_rate": 1.0351360063731231e-05, | |
| "loss": 0.0001, | |
| "step": 767 | |
| }, | |
| { | |
| "epoch": 4.54, | |
| "learning_rate": 1.033007350085904e-05, | |
| "loss": 0.0009, | |
| "step": 768 | |
| }, | |
| { | |
| "epoch": 4.55, | |
| "learning_rate": 1.0308785440626943e-05, | |
| "loss": 0.0001, | |
| "step": 769 | |
| }, | |
| { | |
| "epoch": 4.56, | |
| "learning_rate": 1.0287495979607035e-05, | |
| "loss": 0.0004, | |
| "step": 770 | |
| }, | |
| { | |
| "epoch": 4.56, | |
| "learning_rate": 1.026620521437775e-05, | |
| "loss": 0.0002, | |
| "step": 771 | |
| }, | |
| { | |
| "epoch": 4.57, | |
| "learning_rate": 1.0244913241523455e-05, | |
| "loss": 0.0001, | |
| "step": 772 | |
| }, | |
| { | |
| "epoch": 4.57, | |
| "learning_rate": 1.0223620157633993e-05, | |
| "loss": 0.0001, | |
| "step": 773 | |
| }, | |
| { | |
| "epoch": 4.58, | |
| "learning_rate": 1.020232605930424e-05, | |
| "loss": 0.0, | |
| "step": 774 | |
| }, | |
| { | |
| "epoch": 4.59, | |
| "learning_rate": 1.0181031043133678e-05, | |
| "loss": 0.0026, | |
| "step": 775 | |
| }, | |
| { | |
| "epoch": 4.59, | |
| "learning_rate": 1.0159735205725949e-05, | |
| "loss": 0.0005, | |
| "step": 776 | |
| }, | |
| { | |
| "epoch": 4.6, | |
| "learning_rate": 1.013843864368843e-05, | |
| "loss": 0.0001, | |
| "step": 777 | |
| }, | |
| { | |
| "epoch": 4.6, | |
| "learning_rate": 1.0117141453631775e-05, | |
| "loss": 0.0004, | |
| "step": 778 | |
| }, | |
| { | |
| "epoch": 4.61, | |
| "learning_rate": 1.0095843732169493e-05, | |
| "loss": 0.0002, | |
| "step": 779 | |
| }, | |
| { | |
| "epoch": 4.62, | |
| "learning_rate": 1.00745455759175e-05, | |
| "loss": 0.0001, | |
| "step": 780 | |
| }, | |
| { | |
| "epoch": 4.62, | |
| "learning_rate": 1.0053247081493684e-05, | |
| "loss": 0.0, | |
| "step": 781 | |
| }, | |
| { | |
| "epoch": 4.63, | |
| "learning_rate": 1.0031948345517478e-05, | |
| "loss": 0.0001, | |
| "step": 782 | |
| }, | |
| { | |
| "epoch": 4.63, | |
| "learning_rate": 1.0010649464609394e-05, | |
| "loss": 0.0, | |
| "step": 783 | |
| }, | |
| { | |
| "epoch": 4.64, | |
| "learning_rate": 9.98935053539061e-06, | |
| "loss": 0.0, | |
| "step": 784 | |
| }, | |
| { | |
| "epoch": 4.64, | |
| "learning_rate": 9.968051654482526e-06, | |
| "loss": 0.001, | |
| "step": 785 | |
| }, | |
| { | |
| "epoch": 4.65, | |
| "learning_rate": 9.946752918506319e-06, | |
| "loss": 0.0002, | |
| "step": 786 | |
| }, | |
| { | |
| "epoch": 4.66, | |
| "learning_rate": 9.925454424082503e-06, | |
| "loss": 0.0002, | |
| "step": 787 | |
| }, | |
| { | |
| "epoch": 4.66, | |
| "learning_rate": 9.904156267830512e-06, | |
| "loss": 0.0001, | |
| "step": 788 | |
| }, | |
| { | |
| "epoch": 4.67, | |
| "learning_rate": 9.882858546368226e-06, | |
| "loss": 0.0, | |
| "step": 789 | |
| }, | |
| { | |
| "epoch": 4.67, | |
| "learning_rate": 9.861561356311572e-06, | |
| "loss": 0.0001, | |
| "step": 790 | |
| }, | |
| { | |
| "epoch": 4.68, | |
| "learning_rate": 9.840264794274053e-06, | |
| "loss": 0.0, | |
| "step": 791 | |
| }, | |
| { | |
| "epoch": 4.69, | |
| "learning_rate": 9.818968956866325e-06, | |
| "loss": 0.0, | |
| "step": 792 | |
| }, | |
| { | |
| "epoch": 4.69, | |
| "learning_rate": 9.797673940695762e-06, | |
| "loss": 0.0001, | |
| "step": 793 | |
| }, | |
| { | |
| "epoch": 4.7, | |
| "learning_rate": 9.77637984236601e-06, | |
| "loss": 0.0001, | |
| "step": 794 | |
| }, | |
| { | |
| "epoch": 4.7, | |
| "learning_rate": 9.755086758476548e-06, | |
| "loss": 0.0023, | |
| "step": 795 | |
| }, | |
| { | |
| "epoch": 4.71, | |
| "learning_rate": 9.733794785622254e-06, | |
| "loss": 0.0, | |
| "step": 796 | |
| }, | |
| { | |
| "epoch": 4.72, | |
| "learning_rate": 9.712504020392972e-06, | |
| "loss": 0.0003, | |
| "step": 797 | |
| }, | |
| { | |
| "epoch": 4.72, | |
| "learning_rate": 9.691214559373062e-06, | |
| "loss": 0.0026, | |
| "step": 798 | |
| }, | |
| { | |
| "epoch": 4.73, | |
| "learning_rate": 9.669926499140964e-06, | |
| "loss": 0.0001, | |
| "step": 799 | |
| }, | |
| { | |
| "epoch": 4.73, | |
| "learning_rate": 9.648639936268774e-06, | |
| "loss": 0.0, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 4.74, | |
| "learning_rate": 9.627354967321785e-06, | |
| "loss": 0.0007, | |
| "step": 801 | |
| }, | |
| { | |
| "epoch": 4.75, | |
| "learning_rate": 9.606071688858069e-06, | |
| "loss": 0.0, | |
| "step": 802 | |
| }, | |
| { | |
| "epoch": 4.75, | |
| "learning_rate": 9.584790197428023e-06, | |
| "loss": 0.0, | |
| "step": 803 | |
| }, | |
| { | |
| "epoch": 4.76, | |
| "learning_rate": 9.56351058957393e-06, | |
| "loss": 0.0001, | |
| "step": 804 | |
| }, | |
| { | |
| "epoch": 4.76, | |
| "learning_rate": 9.542232961829548e-06, | |
| "loss": 0.0001, | |
| "step": 805 | |
| }, | |
| { | |
| "epoch": 4.77, | |
| "learning_rate": 9.520957410719632e-06, | |
| "loss": 0.0004, | |
| "step": 806 | |
| }, | |
| { | |
| "epoch": 4.78, | |
| "learning_rate": 9.499684032759527e-06, | |
| "loss": 0.0, | |
| "step": 807 | |
| }, | |
| { | |
| "epoch": 4.78, | |
| "learning_rate": 9.478412924454722e-06, | |
| "loss": 0.0008, | |
| "step": 808 | |
| }, | |
| { | |
| "epoch": 4.79, | |
| "learning_rate": 9.457144182300401e-06, | |
| "loss": 0.0003, | |
| "step": 809 | |
| }, | |
| { | |
| "epoch": 4.79, | |
| "learning_rate": 9.435877902781018e-06, | |
| "loss": 0.001, | |
| "step": 810 | |
| }, | |
| { | |
| "epoch": 4.8, | |
| "learning_rate": 9.414614182369862e-06, | |
| "loss": 0.002, | |
| "step": 811 | |
| }, | |
| { | |
| "epoch": 4.8, | |
| "learning_rate": 9.3933531175286e-06, | |
| "loss": 0.0001, | |
| "step": 812 | |
| }, | |
| { | |
| "epoch": 4.81, | |
| "learning_rate": 9.372094804706867e-06, | |
| "loss": 0.0, | |
| "step": 813 | |
| }, | |
| { | |
| "epoch": 4.82, | |
| "learning_rate": 9.3508393403418e-06, | |
| "loss": 0.0004, | |
| "step": 814 | |
| }, | |
| { | |
| "epoch": 4.82, | |
| "learning_rate": 9.32958682085762e-06, | |
| "loss": 0.0041, | |
| "step": 815 | |
| }, | |
| { | |
| "epoch": 4.83, | |
| "learning_rate": 9.308337342665188e-06, | |
| "loss": 0.0014, | |
| "step": 816 | |
| }, | |
| { | |
| "epoch": 4.83, | |
| "learning_rate": 9.28709100216157e-06, | |
| "loss": 0.0017, | |
| "step": 817 | |
| }, | |
| { | |
| "epoch": 4.84, | |
| "learning_rate": 9.2658478957296e-06, | |
| "loss": 0.0014, | |
| "step": 818 | |
| }, | |
| { | |
| "epoch": 4.85, | |
| "learning_rate": 9.244608119737431e-06, | |
| "loss": 0.0, | |
| "step": 819 | |
| }, | |
| { | |
| "epoch": 4.85, | |
| "learning_rate": 9.22337177053812e-06, | |
| "loss": 0.0, | |
| "step": 820 | |
| }, | |
| { | |
| "epoch": 4.86, | |
| "learning_rate": 9.202138944469168e-06, | |
| "loss": 0.0002, | |
| "step": 821 | |
| }, | |
| { | |
| "epoch": 4.86, | |
| "learning_rate": 9.180909737852106e-06, | |
| "loss": 0.0, | |
| "step": 822 | |
| }, | |
| { | |
| "epoch": 4.87, | |
| "learning_rate": 9.159684246992032e-06, | |
| "loss": 0.0, | |
| "step": 823 | |
| }, | |
| { | |
| "epoch": 4.88, | |
| "learning_rate": 9.138462568177194e-06, | |
| "loss": 0.0, | |
| "step": 824 | |
| }, | |
| { | |
| "epoch": 4.88, | |
| "learning_rate": 9.11724479767855e-06, | |
| "loss": 0.0, | |
| "step": 825 | |
| }, | |
| { | |
| "epoch": 4.89, | |
| "learning_rate": 9.096031031749321e-06, | |
| "loss": 0.0, | |
| "step": 826 | |
| }, | |
| { | |
| "epoch": 4.89, | |
| "learning_rate": 9.074821366624565e-06, | |
| "loss": 0.0003, | |
| "step": 827 | |
| }, | |
| { | |
| "epoch": 4.9, | |
| "learning_rate": 9.05361589852074e-06, | |
| "loss": 0.0001, | |
| "step": 828 | |
| }, | |
| { | |
| "epoch": 4.91, | |
| "learning_rate": 9.032414723635261e-06, | |
| "loss": 0.0009, | |
| "step": 829 | |
| }, | |
| { | |
| "epoch": 4.91, | |
| "learning_rate": 9.011217938146066e-06, | |
| "loss": 0.0002, | |
| "step": 830 | |
| }, | |
| { | |
| "epoch": 4.92, | |
| "learning_rate": 8.99002563821118e-06, | |
| "loss": 0.0, | |
| "step": 831 | |
| }, | |
| { | |
| "epoch": 4.92, | |
| "learning_rate": 8.968837919968285e-06, | |
| "loss": 0.0, | |
| "step": 832 | |
| }, | |
| { | |
| "epoch": 4.93, | |
| "learning_rate": 8.947654879534279e-06, | |
| "loss": 0.0006, | |
| "step": 833 | |
| }, | |
| { | |
| "epoch": 4.93, | |
| "learning_rate": 8.926476613004831e-06, | |
| "loss": 0.0006, | |
| "step": 834 | |
| }, | |
| { | |
| "epoch": 4.94, | |
| "learning_rate": 8.90530321645396e-06, | |
| "loss": 0.0001, | |
| "step": 835 | |
| }, | |
| { | |
| "epoch": 4.95, | |
| "learning_rate": 8.884134785933588e-06, | |
| "loss": 0.0005, | |
| "step": 836 | |
| }, | |
| { | |
| "epoch": 4.95, | |
| "learning_rate": 8.862971417473114e-06, | |
| "loss": 0.0, | |
| "step": 837 | |
| }, | |
| { | |
| "epoch": 4.96, | |
| "learning_rate": 8.841813207078972e-06, | |
| "loss": 0.0006, | |
| "step": 838 | |
| }, | |
| { | |
| "epoch": 4.96, | |
| "learning_rate": 8.82066025073419e-06, | |
| "loss": 0.0001, | |
| "step": 839 | |
| }, | |
| { | |
| "epoch": 4.97, | |
| "learning_rate": 8.799512644397974e-06, | |
| "loss": 0.0, | |
| "step": 840 | |
| }, | |
| { | |
| "epoch": 4.98, | |
| "learning_rate": 8.778370484005245e-06, | |
| "loss": 0.0028, | |
| "step": 841 | |
| }, | |
| { | |
| "epoch": 4.98, | |
| "learning_rate": 8.757233865466238e-06, | |
| "loss": 0.0, | |
| "step": 842 | |
| }, | |
| { | |
| "epoch": 4.99, | |
| "learning_rate": 8.736102884666026e-06, | |
| "loss": 0.0001, | |
| "step": 843 | |
| }, | |
| { | |
| "epoch": 4.99, | |
| "learning_rate": 8.714977637464122e-06, | |
| "loss": 0.0004, | |
| "step": 844 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "learning_rate": 8.693858219694022e-06, | |
| "loss": 0.0005, | |
| "step": 845 | |
| }, | |
| { | |
| "epoch": 5.01, | |
| "learning_rate": 8.672744727162782e-06, | |
| "loss": 0.0003, | |
| "step": 846 | |
| }, | |
| { | |
| "epoch": 5.01, | |
| "learning_rate": 8.651637255650571e-06, | |
| "loss": 0.0002, | |
| "step": 847 | |
| }, | |
| { | |
| "epoch": 5.02, | |
| "learning_rate": 8.630535900910252e-06, | |
| "loss": 0.0003, | |
| "step": 848 | |
| }, | |
| { | |
| "epoch": 5.02, | |
| "learning_rate": 8.609440758666937e-06, | |
| "loss": 0.0002, | |
| "step": 849 | |
| }, | |
| { | |
| "epoch": 5.03, | |
| "learning_rate": 8.58835192461755e-06, | |
| "loss": 0.0006, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 5.04, | |
| "learning_rate": 8.567269494430404e-06, | |
| "loss": 0.0006, | |
| "step": 851 | |
| }, | |
| { | |
| "epoch": 5.04, | |
| "learning_rate": 8.546193563744757e-06, | |
| "loss": 0.0, | |
| "step": 852 | |
| }, | |
| { | |
| "epoch": 5.05, | |
| "learning_rate": 8.525124228170398e-06, | |
| "loss": 0.0, | |
| "step": 853 | |
| }, | |
| { | |
| "epoch": 5.05, | |
| "learning_rate": 8.504061583287173e-06, | |
| "loss": 0.0005, | |
| "step": 854 | |
| }, | |
| { | |
| "epoch": 5.06, | |
| "learning_rate": 8.483005724644593e-06, | |
| "loss": 0.0, | |
| "step": 855 | |
| }, | |
| { | |
| "epoch": 5.07, | |
| "learning_rate": 8.461956747761375e-06, | |
| "loss": 0.0, | |
| "step": 856 | |
| }, | |
| { | |
| "epoch": 5.07, | |
| "learning_rate": 8.440914748125024e-06, | |
| "loss": 0.0, | |
| "step": 857 | |
| }, | |
| { | |
| "epoch": 5.08, | |
| "learning_rate": 8.41987982119139e-06, | |
| "loss": 0.0, | |
| "step": 858 | |
| }, | |
| { | |
| "epoch": 5.08, | |
| "learning_rate": 8.398852062384235e-06, | |
| "loss": 0.0002, | |
| "step": 859 | |
| }, | |
| { | |
| "epoch": 5.09, | |
| "learning_rate": 8.37783156709481e-06, | |
| "loss": 0.0, | |
| "step": 860 | |
| }, | |
| { | |
| "epoch": 5.09, | |
| "learning_rate": 8.356818430681409e-06, | |
| "loss": 0.0, | |
| "step": 861 | |
| }, | |
| { | |
| "epoch": 5.1, | |
| "learning_rate": 8.335812748468945e-06, | |
| "loss": 0.0003, | |
| "step": 862 | |
| }, | |
| { | |
| "epoch": 5.11, | |
| "learning_rate": 8.31481461574852e-06, | |
| "loss": 0.0, | |
| "step": 863 | |
| }, | |
| { | |
| "epoch": 5.11, | |
| "learning_rate": 8.293824127776981e-06, | |
| "loss": 0.0002, | |
| "step": 864 | |
| }, | |
| { | |
| "epoch": 5.12, | |
| "learning_rate": 8.272841379776502e-06, | |
| "loss": 0.0001, | |
| "step": 865 | |
| }, | |
| { | |
| "epoch": 5.12, | |
| "learning_rate": 8.251866466934137e-06, | |
| "loss": 0.0001, | |
| "step": 866 | |
| }, | |
| { | |
| "epoch": 5.13, | |
| "learning_rate": 8.230899484401402e-06, | |
| "loss": 0.0, | |
| "step": 867 | |
| }, | |
| { | |
| "epoch": 5.14, | |
| "learning_rate": 8.209940527293838e-06, | |
| "loss": 0.0, | |
| "step": 868 | |
| }, | |
| { | |
| "epoch": 5.14, | |
| "learning_rate": 8.188989690690574e-06, | |
| "loss": 0.0, | |
| "step": 869 | |
| }, | |
| { | |
| "epoch": 5.15, | |
| "learning_rate": 8.168047069633908e-06, | |
| "loss": 0.0, | |
| "step": 870 | |
| }, | |
| { | |
| "epoch": 5.15, | |
| "learning_rate": 8.147112759128859e-06, | |
| "loss": 0.0, | |
| "step": 871 | |
| }, | |
| { | |
| "epoch": 5.16, | |
| "learning_rate": 8.126186854142752e-06, | |
| "loss": 0.0, | |
| "step": 872 | |
| }, | |
| { | |
| "epoch": 5.17, | |
| "learning_rate": 8.105269449604787e-06, | |
| "loss": 0.0, | |
| "step": 873 | |
| }, | |
| { | |
| "epoch": 5.17, | |
| "learning_rate": 8.084360640405593e-06, | |
| "loss": 0.0001, | |
| "step": 874 | |
| }, | |
| { | |
| "epoch": 5.18, | |
| "learning_rate": 8.063460521396806e-06, | |
| "loss": 0.0015, | |
| "step": 875 | |
| }, | |
| { | |
| "epoch": 5.18, | |
| "learning_rate": 8.042569187390642e-06, | |
| "loss": 0.0, | |
| "step": 876 | |
| }, | |
| { | |
| "epoch": 5.19, | |
| "learning_rate": 8.021686733159469e-06, | |
| "loss": 0.0, | |
| "step": 877 | |
| }, | |
| { | |
| "epoch": 5.2, | |
| "learning_rate": 8.000813253435365e-06, | |
| "loss": 0.0, | |
| "step": 878 | |
| }, | |
| { | |
| "epoch": 5.2, | |
| "learning_rate": 7.979948842909699e-06, | |
| "loss": 0.0, | |
| "step": 879 | |
| }, | |
| { | |
| "epoch": 5.21, | |
| "learning_rate": 7.9590935962327e-06, | |
| "loss": 0.0, | |
| "step": 880 | |
| }, | |
| { | |
| "epoch": 5.21, | |
| "learning_rate": 7.938247608013021e-06, | |
| "loss": 0.0, | |
| "step": 881 | |
| }, | |
| { | |
| "epoch": 5.22, | |
| "learning_rate": 7.917410972817322e-06, | |
| "loss": 0.0, | |
| "step": 882 | |
| }, | |
| { | |
| "epoch": 5.22, | |
| "learning_rate": 7.896583785169818e-06, | |
| "loss": 0.0, | |
| "step": 883 | |
| }, | |
| { | |
| "epoch": 5.23, | |
| "learning_rate": 7.875766139551889e-06, | |
| "loss": 0.0, | |
| "step": 884 | |
| }, | |
| { | |
| "epoch": 5.24, | |
| "learning_rate": 7.85495813040161e-06, | |
| "loss": 0.0004, | |
| "step": 885 | |
| }, | |
| { | |
| "epoch": 5.24, | |
| "learning_rate": 7.834159852113347e-06, | |
| "loss": 0.0001, | |
| "step": 886 | |
| }, | |
| { | |
| "epoch": 5.25, | |
| "learning_rate": 7.81337139903732e-06, | |
| "loss": 0.0, | |
| "step": 887 | |
| }, | |
| { | |
| "epoch": 5.25, | |
| "learning_rate": 7.792592865479186e-06, | |
| "loss": 0.0, | |
| "step": 888 | |
| }, | |
| { | |
| "epoch": 5.26, | |
| "learning_rate": 7.771824345699589e-06, | |
| "loss": 0.0, | |
| "step": 889 | |
| }, | |
| { | |
| "epoch": 5.27, | |
| "learning_rate": 7.751065933913756e-06, | |
| "loss": 0.0001, | |
| "step": 890 | |
| }, | |
| { | |
| "epoch": 5.27, | |
| "learning_rate": 7.73031772429105e-06, | |
| "loss": 0.0001, | |
| "step": 891 | |
| }, | |
| { | |
| "epoch": 5.28, | |
| "learning_rate": 7.70957981095457e-06, | |
| "loss": 0.0, | |
| "step": 892 | |
| }, | |
| { | |
| "epoch": 5.28, | |
| "learning_rate": 7.688852287980686e-06, | |
| "loss": 0.0001, | |
| "step": 893 | |
| }, | |
| { | |
| "epoch": 5.29, | |
| "learning_rate": 7.668135249398649e-06, | |
| "loss": 0.0, | |
| "step": 894 | |
| }, | |
| { | |
| "epoch": 5.3, | |
| "learning_rate": 7.647428789190142e-06, | |
| "loss": 0.0, | |
| "step": 895 | |
| }, | |
| { | |
| "epoch": 5.3, | |
| "learning_rate": 7.626733001288852e-06, | |
| "loss": 0.0, | |
| "step": 896 | |
| }, | |
| { | |
| "epoch": 5.31, | |
| "learning_rate": 7.606047979580065e-06, | |
| "loss": 0.0, | |
| "step": 897 | |
| }, | |
| { | |
| "epoch": 5.31, | |
| "learning_rate": 7.585373817900221e-06, | |
| "loss": 0.0, | |
| "step": 898 | |
| }, | |
| { | |
| "epoch": 5.32, | |
| "learning_rate": 7.564710610036488e-06, | |
| "loss": 0.0, | |
| "step": 899 | |
| }, | |
| { | |
| "epoch": 5.33, | |
| "learning_rate": 7.544058449726356e-06, | |
| "loss": 0.0, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 5.33, | |
| "learning_rate": 7.523417430657186e-06, | |
| "loss": 0.0, | |
| "step": 901 | |
| }, | |
| { | |
| "epoch": 5.34, | |
| "learning_rate": 7.502787646465804e-06, | |
| "loss": 0.0, | |
| "step": 902 | |
| }, | |
| { | |
| "epoch": 5.34, | |
| "learning_rate": 7.48216919073806e-06, | |
| "loss": 0.0, | |
| "step": 903 | |
| }, | |
| { | |
| "epoch": 5.35, | |
| "learning_rate": 7.461562157008433e-06, | |
| "loss": 0.0, | |
| "step": 904 | |
| }, | |
| { | |
| "epoch": 5.36, | |
| "learning_rate": 7.44096663875957e-06, | |
| "loss": 0.0, | |
| "step": 905 | |
| }, | |
| { | |
| "epoch": 5.36, | |
| "learning_rate": 7.420382729421883e-06, | |
| "loss": 0.0001, | |
| "step": 906 | |
| }, | |
| { | |
| "epoch": 5.37, | |
| "learning_rate": 7.3998105223731186e-06, | |
| "loss": 0.0, | |
| "step": 907 | |
| }, | |
| { | |
| "epoch": 5.37, | |
| "learning_rate": 7.3792501109379445e-06, | |
| "loss": 0.0, | |
| "step": 908 | |
| }, | |
| { | |
| "epoch": 5.38, | |
| "learning_rate": 7.358701588387509e-06, | |
| "loss": 0.0001, | |
| "step": 909 | |
| }, | |
| { | |
| "epoch": 5.38, | |
| "learning_rate": 7.338165047939032e-06, | |
| "loss": 0.0, | |
| "step": 910 | |
| }, | |
| { | |
| "epoch": 5.39, | |
| "learning_rate": 7.317640582755373e-06, | |
| "loss": 0.0001, | |
| "step": 911 | |
| }, | |
| { | |
| "epoch": 5.4, | |
| "learning_rate": 7.297128285944621e-06, | |
| "loss": 0.0, | |
| "step": 912 | |
| }, | |
| { | |
| "epoch": 5.4, | |
| "learning_rate": 7.276628250559659e-06, | |
| "loss": 0.0, | |
| "step": 913 | |
| }, | |
| { | |
| "epoch": 5.41, | |
| "learning_rate": 7.256140569597738e-06, | |
| "loss": 0.0, | |
| "step": 914 | |
| }, | |
| { | |
| "epoch": 5.41, | |
| "learning_rate": 7.235665336000086e-06, | |
| "loss": 0.0, | |
| "step": 915 | |
| }, | |
| { | |
| "epoch": 5.42, | |
| "learning_rate": 7.2152026426514395e-06, | |
| "loss": 0.0, | |
| "step": 916 | |
| }, | |
| { | |
| "epoch": 5.43, | |
| "learning_rate": 7.194752582379665e-06, | |
| "loss": 0.0, | |
| "step": 917 | |
| }, | |
| { | |
| "epoch": 5.43, | |
| "learning_rate": 7.174315247955312e-06, | |
| "loss": 0.0, | |
| "step": 918 | |
| }, | |
| { | |
| "epoch": 5.44, | |
| "learning_rate": 7.153890732091194e-06, | |
| "loss": 0.0, | |
| "step": 919 | |
| }, | |
| { | |
| "epoch": 5.44, | |
| "learning_rate": 7.133479127441989e-06, | |
| "loss": 0.0, | |
| "step": 920 | |
| }, | |
| { | |
| "epoch": 5.45, | |
| "learning_rate": 7.113080526603793e-06, | |
| "loss": 0.0, | |
| "step": 921 | |
| }, | |
| { | |
| "epoch": 5.46, | |
| "learning_rate": 7.0926950221137114e-06, | |
| "loss": 0.0, | |
| "step": 922 | |
| }, | |
| { | |
| "epoch": 5.46, | |
| "learning_rate": 7.072322706449439e-06, | |
| "loss": 0.0002, | |
| "step": 923 | |
| }, | |
| { | |
| "epoch": 5.47, | |
| "learning_rate": 7.051963672028844e-06, | |
| "loss": 0.0, | |
| "step": 924 | |
| }, | |
| { | |
| "epoch": 5.47, | |
| "learning_rate": 7.031618011209549e-06, | |
| "loss": 0.0, | |
| "step": 925 | |
| }, | |
| { | |
| "epoch": 5.48, | |
| "learning_rate": 7.011285816288496e-06, | |
| "loss": 0.0, | |
| "step": 926 | |
| }, | |
| { | |
| "epoch": 5.49, | |
| "learning_rate": 6.990967179501547e-06, | |
| "loss": 0.0, | |
| "step": 927 | |
| }, | |
| { | |
| "epoch": 5.49, | |
| "learning_rate": 6.970662193023057e-06, | |
| "loss": 0.0, | |
| "step": 928 | |
| }, | |
| { | |
| "epoch": 5.5, | |
| "learning_rate": 6.950370948965457e-06, | |
| "loss": 0.0, | |
| "step": 929 | |
| }, | |
| { | |
| "epoch": 5.5, | |
| "learning_rate": 6.930093539378838e-06, | |
| "loss": 0.0, | |
| "step": 930 | |
| }, | |
| { | |
| "epoch": 5.51, | |
| "learning_rate": 6.909830056250527e-06, | |
| "loss": 0.0, | |
| "step": 931 | |
| }, | |
| { | |
| "epoch": 5.51, | |
| "learning_rate": 6.889580591504681e-06, | |
| "loss": 0.0, | |
| "step": 932 | |
| }, | |
| { | |
| "epoch": 5.52, | |
| "learning_rate": 6.8693452370018585e-06, | |
| "loss": 0.0, | |
| "step": 933 | |
| }, | |
| { | |
| "epoch": 5.53, | |
| "learning_rate": 6.849124084538608e-06, | |
| "loss": 0.0, | |
| "step": 934 | |
| }, | |
| { | |
| "epoch": 5.53, | |
| "learning_rate": 6.828917225847053e-06, | |
| "loss": 0.0, | |
| "step": 935 | |
| }, | |
| { | |
| "epoch": 5.54, | |
| "learning_rate": 6.8087247525944745e-06, | |
| "loss": 0.0, | |
| "step": 936 | |
| }, | |
| { | |
| "epoch": 5.54, | |
| "learning_rate": 6.7885467563828946e-06, | |
| "loss": 0.0, | |
| "step": 937 | |
| }, | |
| { | |
| "epoch": 5.55, | |
| "learning_rate": 6.76838332874866e-06, | |
| "loss": 0.0, | |
| "step": 938 | |
| }, | |
| { | |
| "epoch": 5.56, | |
| "learning_rate": 6.748234561162025e-06, | |
| "loss": 0.0001, | |
| "step": 939 | |
| }, | |
| { | |
| "epoch": 5.56, | |
| "learning_rate": 6.7281005450267465e-06, | |
| "loss": 0.0, | |
| "step": 940 | |
| }, | |
| { | |
| "epoch": 5.57, | |
| "learning_rate": 6.707981371679657e-06, | |
| "loss": 0.0, | |
| "step": 941 | |
| }, | |
| { | |
| "epoch": 5.57, | |
| "learning_rate": 6.687877132390257e-06, | |
| "loss": 0.0, | |
| "step": 942 | |
| }, | |
| { | |
| "epoch": 5.58, | |
| "learning_rate": 6.667787918360296e-06, | |
| "loss": 0.0, | |
| "step": 943 | |
| }, | |
| { | |
| "epoch": 5.59, | |
| "learning_rate": 6.647713820723372e-06, | |
| "loss": 0.0002, | |
| "step": 944 | |
| }, | |
| { | |
| "epoch": 5.59, | |
| "learning_rate": 6.627654930544493e-06, | |
| "loss": 0.0, | |
| "step": 945 | |
| }, | |
| { | |
| "epoch": 5.6, | |
| "learning_rate": 6.607611338819697e-06, | |
| "loss": 0.0007, | |
| "step": 946 | |
| }, | |
| { | |
| "epoch": 5.6, | |
| "learning_rate": 6.587583136475608e-06, | |
| "loss": 0.0, | |
| "step": 947 | |
| }, | |
| { | |
| "epoch": 5.61, | |
| "learning_rate": 6.567570414369046e-06, | |
| "loss": 0.0, | |
| "step": 948 | |
| }, | |
| { | |
| "epoch": 5.62, | |
| "learning_rate": 6.547573263286598e-06, | |
| "loss": 0.0, | |
| "step": 949 | |
| }, | |
| { | |
| "epoch": 5.62, | |
| "learning_rate": 6.527591773944215e-06, | |
| "loss": 0.0001, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 5.63, | |
| "learning_rate": 6.507626036986804e-06, | |
| "loss": 0.0, | |
| "step": 951 | |
| }, | |
| { | |
| "epoch": 5.63, | |
| "learning_rate": 6.487676142987811e-06, | |
| "loss": 0.0001, | |
| "step": 952 | |
| }, | |
| { | |
| "epoch": 5.64, | |
| "learning_rate": 6.467742182448809e-06, | |
| "loss": 0.0, | |
| "step": 953 | |
| }, | |
| { | |
| "epoch": 5.64, | |
| "learning_rate": 6.4478242457990884e-06, | |
| "loss": 0.0, | |
| "step": 954 | |
| }, | |
| { | |
| "epoch": 5.65, | |
| "learning_rate": 6.427922423395252e-06, | |
| "loss": 0.0006, | |
| "step": 955 | |
| }, | |
| { | |
| "epoch": 5.66, | |
| "learning_rate": 6.408036805520801e-06, | |
| "loss": 0.0, | |
| "step": 956 | |
| }, | |
| { | |
| "epoch": 5.66, | |
| "learning_rate": 6.3881674823857275e-06, | |
| "loss": 0.0, | |
| "step": 957 | |
| }, | |
| { | |
| "epoch": 5.67, | |
| "learning_rate": 6.368314544126095e-06, | |
| "loss": 0.0, | |
| "step": 958 | |
| }, | |
| { | |
| "epoch": 5.67, | |
| "learning_rate": 6.348478080803646e-06, | |
| "loss": 0.0, | |
| "step": 959 | |
| }, | |
| { | |
| "epoch": 5.68, | |
| "learning_rate": 6.3286581824053826e-06, | |
| "loss": 0.0, | |
| "step": 960 | |
| }, | |
| { | |
| "epoch": 5.69, | |
| "learning_rate": 6.308854938843161e-06, | |
| "loss": 0.0, | |
| "step": 961 | |
| }, | |
| { | |
| "epoch": 5.69, | |
| "learning_rate": 6.289068439953283e-06, | |
| "loss": 0.0, | |
| "step": 962 | |
| }, | |
| { | |
| "epoch": 5.7, | |
| "learning_rate": 6.269298775496089e-06, | |
| "loss": 0.0, | |
| "step": 963 | |
| }, | |
| { | |
| "epoch": 5.7, | |
| "learning_rate": 6.249546035155556e-06, | |
| "loss": 0.0, | |
| "step": 964 | |
| }, | |
| { | |
| "epoch": 5.71, | |
| "learning_rate": 6.229810308538877e-06, | |
| "loss": 0.0, | |
| "step": 965 | |
| }, | |
| { | |
| "epoch": 5.72, | |
| "learning_rate": 6.210091685176067e-06, | |
| "loss": 0.0001, | |
| "step": 966 | |
| }, | |
| { | |
| "epoch": 5.72, | |
| "learning_rate": 6.1903902545195595e-06, | |
| "loss": 0.0, | |
| "step": 967 | |
| }, | |
| { | |
| "epoch": 5.73, | |
| "learning_rate": 6.170706105943787e-06, | |
| "loss": 0.0, | |
| "step": 968 | |
| }, | |
| { | |
| "epoch": 5.73, | |
| "learning_rate": 6.151039328744784e-06, | |
| "loss": 0.0001, | |
| "step": 969 | |
| }, | |
| { | |
| "epoch": 5.74, | |
| "learning_rate": 6.13139001213978e-06, | |
| "loss": 0.0, | |
| "step": 970 | |
| }, | |
| { | |
| "epoch": 5.75, | |
| "learning_rate": 6.111758245266795e-06, | |
| "loss": 0.0, | |
| "step": 971 | |
| }, | |
| { | |
| "epoch": 5.75, | |
| "learning_rate": 6.09214411718424e-06, | |
| "loss": 0.0, | |
| "step": 972 | |
| }, | |
| { | |
| "epoch": 5.76, | |
| "learning_rate": 6.072547716870507e-06, | |
| "loss": 0.0, | |
| "step": 973 | |
| }, | |
| { | |
| "epoch": 5.76, | |
| "learning_rate": 6.052969133223561e-06, | |
| "loss": 0.0, | |
| "step": 974 | |
| }, | |
| { | |
| "epoch": 5.77, | |
| "learning_rate": 6.033408455060552e-06, | |
| "loss": 0.0, | |
| "step": 975 | |
| }, | |
| { | |
| "epoch": 5.78, | |
| "learning_rate": 6.013865771117394e-06, | |
| "loss": 0.0, | |
| "step": 976 | |
| }, | |
| { | |
| "epoch": 5.78, | |
| "learning_rate": 5.994341170048382e-06, | |
| "loss": 0.0, | |
| "step": 977 | |
| }, | |
| { | |
| "epoch": 5.79, | |
| "learning_rate": 5.97483474042577e-06, | |
| "loss": 0.0, | |
| "step": 978 | |
| }, | |
| { | |
| "epoch": 5.79, | |
| "learning_rate": 5.9553465707393765e-06, | |
| "loss": 0.0, | |
| "step": 979 | |
| }, | |
| { | |
| "epoch": 5.8, | |
| "learning_rate": 5.935876749396194e-06, | |
| "loss": 0.0, | |
| "step": 980 | |
| }, | |
| { | |
| "epoch": 5.8, | |
| "learning_rate": 5.916425364719975e-06, | |
| "loss": 0.0005, | |
| "step": 981 | |
| }, | |
| { | |
| "epoch": 5.81, | |
| "learning_rate": 5.8969925049508304e-06, | |
| "loss": 0.0, | |
| "step": 982 | |
| }, | |
| { | |
| "epoch": 5.82, | |
| "learning_rate": 5.877578258244836e-06, | |
| "loss": 0.0, | |
| "step": 983 | |
| }, | |
| { | |
| "epoch": 5.82, | |
| "learning_rate": 5.8581827126736394e-06, | |
| "loss": 0.0, | |
| "step": 984 | |
| }, | |
| { | |
| "epoch": 5.83, | |
| "learning_rate": 5.838805956224036e-06, | |
| "loss": 0.0, | |
| "step": 985 | |
| }, | |
| { | |
| "epoch": 5.83, | |
| "learning_rate": 5.8194480767976e-06, | |
| "loss": 0.0, | |
| "step": 986 | |
| }, | |
| { | |
| "epoch": 5.84, | |
| "learning_rate": 5.8001091622102545e-06, | |
| "loss": 0.0, | |
| "step": 987 | |
| }, | |
| { | |
| "epoch": 5.85, | |
| "learning_rate": 5.780789300191916e-06, | |
| "loss": 0.0, | |
| "step": 988 | |
| }, | |
| { | |
| "epoch": 5.85, | |
| "learning_rate": 5.761488578386041e-06, | |
| "loss": 0.0001, | |
| "step": 989 | |
| }, | |
| { | |
| "epoch": 5.86, | |
| "learning_rate": 5.742207084349274e-06, | |
| "loss": 0.0001, | |
| "step": 990 | |
| }, | |
| { | |
| "epoch": 5.86, | |
| "learning_rate": 5.7229449055510335e-06, | |
| "loss": 0.0001, | |
| "step": 991 | |
| }, | |
| { | |
| "epoch": 5.87, | |
| "learning_rate": 5.7037021293731075e-06, | |
| "loss": 0.0, | |
| "step": 992 | |
| }, | |
| { | |
| "epoch": 5.88, | |
| "learning_rate": 5.68447884310927e-06, | |
| "loss": 0.0, | |
| "step": 993 | |
| }, | |
| { | |
| "epoch": 5.88, | |
| "learning_rate": 5.665275133964887e-06, | |
| "loss": 0.0005, | |
| "step": 994 | |
| }, | |
| { | |
| "epoch": 5.89, | |
| "learning_rate": 5.6460910890564965e-06, | |
| "loss": 0.0, | |
| "step": 995 | |
| }, | |
| { | |
| "epoch": 5.89, | |
| "learning_rate": 5.626926795411447e-06, | |
| "loss": 0.0, | |
| "step": 996 | |
| }, | |
| { | |
| "epoch": 5.9, | |
| "learning_rate": 5.6077823399674805e-06, | |
| "loss": 0.0001, | |
| "step": 997 | |
| }, | |
| { | |
| "epoch": 5.91, | |
| "learning_rate": 5.588657809572342e-06, | |
| "loss": 0.0, | |
| "step": 998 | |
| }, | |
| { | |
| "epoch": 5.91, | |
| "learning_rate": 5.569553290983397e-06, | |
| "loss": 0.0, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 5.92, | |
| "learning_rate": 5.550468870867211e-06, | |
| "loss": 0.0, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 5.92, | |
| "learning_rate": 5.531404635799191e-06, | |
| "loss": 0.0, | |
| "step": 1001 | |
| }, | |
| { | |
| "epoch": 5.93, | |
| "learning_rate": 5.512360672263173e-06, | |
| "loss": 0.0, | |
| "step": 1002 | |
| }, | |
| { | |
| "epoch": 5.93, | |
| "learning_rate": 5.493337066651017e-06, | |
| "loss": 0.0, | |
| "step": 1003 | |
| }, | |
| { | |
| "epoch": 5.94, | |
| "learning_rate": 5.474333905262248e-06, | |
| "loss": 0.0, | |
| "step": 1004 | |
| }, | |
| { | |
| "epoch": 5.95, | |
| "learning_rate": 5.4553512743036464e-06, | |
| "loss": 0.0, | |
| "step": 1005 | |
| }, | |
| { | |
| "epoch": 5.95, | |
| "learning_rate": 5.436389259888841e-06, | |
| "loss": 0.0001, | |
| "step": 1006 | |
| }, | |
| { | |
| "epoch": 5.96, | |
| "learning_rate": 5.417447948037953e-06, | |
| "loss": 0.0, | |
| "step": 1007 | |
| }, | |
| { | |
| "epoch": 5.96, | |
| "learning_rate": 5.398527424677175e-06, | |
| "loss": 0.0002, | |
| "step": 1008 | |
| }, | |
| { | |
| "epoch": 5.97, | |
| "learning_rate": 5.379627775638404e-06, | |
| "loss": 0.0, | |
| "step": 1009 | |
| }, | |
| { | |
| "epoch": 5.98, | |
| "learning_rate": 5.360749086658839e-06, | |
| "loss": 0.0, | |
| "step": 1010 | |
| }, | |
| { | |
| "epoch": 5.98, | |
| "eval_test_accuracy": 48.7029831387808, | |
| "eval_test_average": 48.7029831387808, | |
| "eval_test_loss": 1.2541236877441406, | |
| "eval_test_runtime": 1246.4837, | |
| "eval_test_samples_per_second": 1.237, | |
| "eval_test_steps_per_second": 0.01, | |
| "step": 1010 | |
| }, | |
| { | |
| "epoch": 5.98, | |
| "learning_rate": 5.341891443380585e-06, | |
| "loss": 0.0, | |
| "step": 1011 | |
| }, | |
| { | |
| "epoch": 5.99, | |
| "learning_rate": 5.323054931350289e-06, | |
| "loss": 0.0, | |
| "step": 1012 | |
| }, | |
| { | |
| "epoch": 5.99, | |
| "learning_rate": 5.304239636018722e-06, | |
| "loss": 0.0, | |
| "step": 1013 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "learning_rate": 5.285445642740421e-06, | |
| "loss": 0.0, | |
| "step": 1014 | |
| }, | |
| { | |
| "epoch": 6.01, | |
| "learning_rate": 5.266673036773284e-06, | |
| "loss": 0.0, | |
| "step": 1015 | |
| }, | |
| { | |
| "epoch": 6.01, | |
| "learning_rate": 5.247921903278177e-06, | |
| "loss": 0.0, | |
| "step": 1016 | |
| }, | |
| { | |
| "epoch": 6.02, | |
| "learning_rate": 5.229192327318569e-06, | |
| "loss": 0.0, | |
| "step": 1017 | |
| }, | |
| { | |
| "epoch": 6.02, | |
| "learning_rate": 5.210484393860127e-06, | |
| "loss": 0.0, | |
| "step": 1018 | |
| }, | |
| { | |
| "epoch": 6.03, | |
| "learning_rate": 5.191798187770343e-06, | |
| "loss": 0.0, | |
| "step": 1019 | |
| }, | |
| { | |
| "epoch": 6.04, | |
| "learning_rate": 5.173133793818144e-06, | |
| "loss": 0.0, | |
| "step": 1020 | |
| }, | |
| { | |
| "epoch": 6.04, | |
| "learning_rate": 5.1544912966735e-06, | |
| "loss": 0.0, | |
| "step": 1021 | |
| }, | |
| { | |
| "epoch": 6.05, | |
| "learning_rate": 5.135870780907055e-06, | |
| "loss": 0.0, | |
| "step": 1022 | |
| }, | |
| { | |
| "epoch": 6.05, | |
| "learning_rate": 5.117272330989738e-06, | |
| "loss": 0.0, | |
| "step": 1023 | |
| }, | |
| { | |
| "epoch": 6.06, | |
| "learning_rate": 5.098696031292362e-06, | |
| "loss": 0.0, | |
| "step": 1024 | |
| }, | |
| { | |
| "epoch": 6.07, | |
| "learning_rate": 5.080141966085278e-06, | |
| "loss": 0.0, | |
| "step": 1025 | |
| }, | |
| { | |
| "epoch": 6.07, | |
| "learning_rate": 5.06161021953796e-06, | |
| "loss": 0.0, | |
| "step": 1026 | |
| }, | |
| { | |
| "epoch": 6.08, | |
| "learning_rate": 5.0431008757186295e-06, | |
| "loss": 0.0, | |
| "step": 1027 | |
| }, | |
| { | |
| "epoch": 6.08, | |
| "learning_rate": 5.024614018593889e-06, | |
| "loss": 0.0, | |
| "step": 1028 | |
| }, | |
| { | |
| "epoch": 6.09, | |
| "learning_rate": 5.006149732028328e-06, | |
| "loss": 0.0, | |
| "step": 1029 | |
| }, | |
| { | |
| "epoch": 6.09, | |
| "learning_rate": 4.987708099784148e-06, | |
| "loss": 0.0, | |
| "step": 1030 | |
| }, | |
| { | |
| "epoch": 6.1, | |
| "learning_rate": 4.9692892055207784e-06, | |
| "loss": 0.0, | |
| "step": 1031 | |
| }, | |
| { | |
| "epoch": 6.11, | |
| "learning_rate": 4.9508931327944944e-06, | |
| "loss": 0.0, | |
| "step": 1032 | |
| }, | |
| { | |
| "epoch": 6.11, | |
| "learning_rate": 4.932519965058049e-06, | |
| "loss": 0.0, | |
| "step": 1033 | |
| }, | |
| { | |
| "epoch": 6.12, | |
| "learning_rate": 4.914169785660292e-06, | |
| "loss": 0.0, | |
| "step": 1034 | |
| }, | |
| { | |
| "epoch": 6.12, | |
| "learning_rate": 4.895842677845772e-06, | |
| "loss": 0.0, | |
| "step": 1035 | |
| }, | |
| { | |
| "epoch": 6.13, | |
| "learning_rate": 4.877538724754392e-06, | |
| "loss": 0.0, | |
| "step": 1036 | |
| }, | |
| { | |
| "epoch": 6.14, | |
| "learning_rate": 4.8592580094210065e-06, | |
| "loss": 0.0, | |
| "step": 1037 | |
| }, | |
| { | |
| "epoch": 6.14, | |
| "learning_rate": 4.841000614775047e-06, | |
| "loss": 0.0, | |
| "step": 1038 | |
| }, | |
| { | |
| "epoch": 6.15, | |
| "learning_rate": 4.822766623640173e-06, | |
| "loss": 0.0, | |
| "step": 1039 | |
| }, | |
| { | |
| "epoch": 6.15, | |
| "learning_rate": 4.80455611873385e-06, | |
| "loss": 0.0, | |
| "step": 1040 | |
| }, | |
| { | |
| "epoch": 6.16, | |
| "learning_rate": 4.7863691826670146e-06, | |
| "loss": 0.0, | |
| "step": 1041 | |
| }, | |
| { | |
| "epoch": 6.17, | |
| "learning_rate": 4.768205897943686e-06, | |
| "loss": 0.0, | |
| "step": 1042 | |
| }, | |
| { | |
| "epoch": 6.17, | |
| "learning_rate": 4.750066346960579e-06, | |
| "loss": 0.0, | |
| "step": 1043 | |
| }, | |
| { | |
| "epoch": 6.18, | |
| "learning_rate": 4.731950612006749e-06, | |
| "loss": 0.0, | |
| "step": 1044 | |
| }, | |
| { | |
| "epoch": 6.18, | |
| "learning_rate": 4.713858775263217e-06, | |
| "loss": 0.0, | |
| "step": 1045 | |
| }, | |
| { | |
| "epoch": 6.19, | |
| "learning_rate": 4.695790918802577e-06, | |
| "loss": 0.0, | |
| "step": 1046 | |
| }, | |
| { | |
| "epoch": 6.2, | |
| "learning_rate": 4.677747124588651e-06, | |
| "loss": 0.0, | |
| "step": 1047 | |
| }, | |
| { | |
| "epoch": 6.2, | |
| "learning_rate": 4.659727474476093e-06, | |
| "loss": 0.0, | |
| "step": 1048 | |
| }, | |
| { | |
| "epoch": 6.21, | |
| "learning_rate": 4.641732050210032e-06, | |
| "loss": 0.0, | |
| "step": 1049 | |
| }, | |
| { | |
| "epoch": 6.21, | |
| "learning_rate": 4.623760933425713e-06, | |
| "loss": 0.0, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 6.22, | |
| "learning_rate": 4.605814205648087e-06, | |
| "loss": 0.0, | |
| "step": 1051 | |
| }, | |
| { | |
| "epoch": 6.22, | |
| "learning_rate": 4.587891948291484e-06, | |
| "loss": 0.0, | |
| "step": 1052 | |
| }, | |
| { | |
| "epoch": 6.23, | |
| "learning_rate": 4.5699942426592105e-06, | |
| "loss": 0.0, | |
| "step": 1053 | |
| }, | |
| { | |
| "epoch": 6.24, | |
| "learning_rate": 4.552121169943208e-06, | |
| "loss": 0.0, | |
| "step": 1054 | |
| }, | |
| { | |
| "epoch": 6.24, | |
| "learning_rate": 4.534272811223669e-06, | |
| "loss": 0.0, | |
| "step": 1055 | |
| }, | |
| { | |
| "epoch": 6.25, | |
| "learning_rate": 4.516449247468666e-06, | |
| "loss": 0.0, | |
| "step": 1056 | |
| }, | |
| { | |
| "epoch": 6.25, | |
| "learning_rate": 4.498650559533799e-06, | |
| "loss": 0.0, | |
| "step": 1057 | |
| }, | |
| { | |
| "epoch": 6.26, | |
| "learning_rate": 4.480876828161819e-06, | |
| "loss": 0.0, | |
| "step": 1058 | |
| }, | |
| { | |
| "epoch": 6.27, | |
| "learning_rate": 4.463128133982248e-06, | |
| "loss": 0.0, | |
| "step": 1059 | |
| }, | |
| { | |
| "epoch": 6.27, | |
| "learning_rate": 4.445404557511058e-06, | |
| "loss": 0.0, | |
| "step": 1060 | |
| }, | |
| { | |
| "epoch": 6.28, | |
| "learning_rate": 4.427706179150247e-06, | |
| "loss": 0.0, | |
| "step": 1061 | |
| }, | |
| { | |
| "epoch": 6.28, | |
| "learning_rate": 4.410033079187519e-06, | |
| "loss": 0.0, | |
| "step": 1062 | |
| }, | |
| { | |
| "epoch": 6.29, | |
| "learning_rate": 4.392385337795903e-06, | |
| "loss": 0.0, | |
| "step": 1063 | |
| }, | |
| { | |
| "epoch": 6.3, | |
| "learning_rate": 4.374763035033378e-06, | |
| "loss": 0.0, | |
| "step": 1064 | |
| }, | |
| { | |
| "epoch": 6.3, | |
| "learning_rate": 4.3571662508425375e-06, | |
| "loss": 0.0, | |
| "step": 1065 | |
| }, | |
| { | |
| "epoch": 6.31, | |
| "learning_rate": 4.339595065050206e-06, | |
| "loss": 0.0, | |
| "step": 1066 | |
| }, | |
| { | |
| "epoch": 6.31, | |
| "learning_rate": 4.3220495573670765e-06, | |
| "loss": 0.0, | |
| "step": 1067 | |
| }, | |
| { | |
| "epoch": 6.32, | |
| "learning_rate": 4.304529807387362e-06, | |
| "loss": 0.0, | |
| "step": 1068 | |
| }, | |
| { | |
| "epoch": 6.33, | |
| "learning_rate": 4.287035894588429e-06, | |
| "loss": 0.0, | |
| "step": 1069 | |
| }, | |
| { | |
| "epoch": 6.33, | |
| "learning_rate": 4.26956789833042e-06, | |
| "loss": 0.0, | |
| "step": 1070 | |
| }, | |
| { | |
| "epoch": 6.34, | |
| "learning_rate": 4.2521258978559324e-06, | |
| "loss": 0.0, | |
| "step": 1071 | |
| }, | |
| { | |
| "epoch": 6.34, | |
| "learning_rate": 4.234709972289612e-06, | |
| "loss": 0.0, | |
| "step": 1072 | |
| }, | |
| { | |
| "epoch": 6.35, | |
| "learning_rate": 4.217320200637833e-06, | |
| "loss": 0.0, | |
| "step": 1073 | |
| }, | |
| { | |
| "epoch": 6.36, | |
| "learning_rate": 4.199956661788318e-06, | |
| "loss": 0.0, | |
| "step": 1074 | |
| }, | |
| { | |
| "epoch": 6.36, | |
| "learning_rate": 4.182619434509783e-06, | |
| "loss": 0.0, | |
| "step": 1075 | |
| }, | |
| { | |
| "epoch": 6.37, | |
| "learning_rate": 4.165308597451586e-06, | |
| "loss": 0.0, | |
| "step": 1076 | |
| }, | |
| { | |
| "epoch": 6.37, | |
| "learning_rate": 4.1480242291433735e-06, | |
| "loss": 0.0, | |
| "step": 1077 | |
| }, | |
| { | |
| "epoch": 6.38, | |
| "learning_rate": 4.130766407994705e-06, | |
| "loss": 0.0, | |
| "step": 1078 | |
| }, | |
| { | |
| "epoch": 6.38, | |
| "learning_rate": 4.113535212294719e-06, | |
| "loss": 0.0, | |
| "step": 1079 | |
| }, | |
| { | |
| "epoch": 6.39, | |
| "learning_rate": 4.09633072021177e-06, | |
| "loss": 0.0, | |
| "step": 1080 | |
| }, | |
| { | |
| "epoch": 6.4, | |
| "learning_rate": 4.079153009793068e-06, | |
| "loss": 0.0, | |
| "step": 1081 | |
| }, | |
| { | |
| "epoch": 6.4, | |
| "learning_rate": 4.062002158964336e-06, | |
| "loss": 0.0, | |
| "step": 1082 | |
| }, | |
| { | |
| "epoch": 6.41, | |
| "learning_rate": 4.0448782455294386e-06, | |
| "loss": 0.0, | |
| "step": 1083 | |
| }, | |
| { | |
| "epoch": 6.41, | |
| "learning_rate": 4.027781347170054e-06, | |
| "loss": 0.0, | |
| "step": 1084 | |
| }, | |
| { | |
| "epoch": 6.42, | |
| "learning_rate": 4.0107115414453e-06, | |
| "loss": 0.0, | |
| "step": 1085 | |
| }, | |
| { | |
| "epoch": 6.43, | |
| "learning_rate": 3.99366890579139e-06, | |
| "loss": 0.0, | |
| "step": 1086 | |
| }, | |
| { | |
| "epoch": 6.43, | |
| "learning_rate": 3.97665351752129e-06, | |
| "loss": 0.0, | |
| "step": 1087 | |
| }, | |
| { | |
| "epoch": 6.44, | |
| "learning_rate": 3.959665453824344e-06, | |
| "loss": 0.0, | |
| "step": 1088 | |
| }, | |
| { | |
| "epoch": 6.44, | |
| "learning_rate": 3.942704791765955e-06, | |
| "loss": 0.0, | |
| "step": 1089 | |
| }, | |
| { | |
| "epoch": 6.45, | |
| "learning_rate": 3.925771608287213e-06, | |
| "loss": 0.0, | |
| "step": 1090 | |
| }, | |
| { | |
| "epoch": 6.46, | |
| "learning_rate": 3.908865980204555e-06, | |
| "loss": 0.0, | |
| "step": 1091 | |
| }, | |
| { | |
| "epoch": 6.46, | |
| "learning_rate": 3.891987984209417e-06, | |
| "loss": 0.0, | |
| "step": 1092 | |
| }, | |
| { | |
| "epoch": 6.47, | |
| "learning_rate": 3.875137696867873e-06, | |
| "loss": 0.0, | |
| "step": 1093 | |
| }, | |
| { | |
| "epoch": 6.47, | |
| "learning_rate": 3.858315194620309e-06, | |
| "loss": 0.0, | |
| "step": 1094 | |
| }, | |
| { | |
| "epoch": 6.48, | |
| "learning_rate": 3.841520553781066e-06, | |
| "loss": 0.0, | |
| "step": 1095 | |
| }, | |
| { | |
| "epoch": 6.49, | |
| "learning_rate": 3.824753850538082e-06, | |
| "loss": 0.0, | |
| "step": 1096 | |
| }, | |
| { | |
| "epoch": 6.49, | |
| "learning_rate": 3.8080151609525683e-06, | |
| "loss": 0.0, | |
| "step": 1097 | |
| }, | |
| { | |
| "epoch": 6.5, | |
| "learning_rate": 3.791304560958655e-06, | |
| "loss": 0.0, | |
| "step": 1098 | |
| }, | |
| { | |
| "epoch": 6.5, | |
| "learning_rate": 3.7746221263630322e-06, | |
| "loss": 0.0, | |
| "step": 1099 | |
| }, | |
| { | |
| "epoch": 6.51, | |
| "learning_rate": 3.757967932844633e-06, | |
| "loss": 0.0, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 6.51, | |
| "learning_rate": 3.741342055954269e-06, | |
| "loss": 0.0, | |
| "step": 1101 | |
| }, | |
| { | |
| "epoch": 6.52, | |
| "learning_rate": 3.7247445711143004e-06, | |
| "loss": 0.0, | |
| "step": 1102 | |
| }, | |
| { | |
| "epoch": 6.53, | |
| "learning_rate": 3.7081755536182863e-06, | |
| "loss": 0.0, | |
| "step": 1103 | |
| }, | |
| { | |
| "epoch": 6.53, | |
| "learning_rate": 3.6916350786306375e-06, | |
| "loss": 0.0, | |
| "step": 1104 | |
| }, | |
| { | |
| "epoch": 6.54, | |
| "learning_rate": 3.6751232211862965e-06, | |
| "loss": 0.0, | |
| "step": 1105 | |
| }, | |
| { | |
| "epoch": 6.54, | |
| "learning_rate": 3.658640056190378e-06, | |
| "loss": 0.0, | |
| "step": 1106 | |
| }, | |
| { | |
| "epoch": 6.55, | |
| "learning_rate": 3.6421856584178285e-06, | |
| "loss": 0.0, | |
| "step": 1107 | |
| }, | |
| { | |
| "epoch": 6.56, | |
| "learning_rate": 3.625760102513103e-06, | |
| "loss": 0.0, | |
| "step": 1108 | |
| }, | |
| { | |
| "epoch": 6.56, | |
| "learning_rate": 3.609363462989819e-06, | |
| "loss": 0.0, | |
| "step": 1109 | |
| }, | |
| { | |
| "epoch": 6.57, | |
| "learning_rate": 3.592995814230401e-06, | |
| "loss": 0.0, | |
| "step": 1110 | |
| }, | |
| { | |
| "epoch": 6.57, | |
| "learning_rate": 3.576657230485775e-06, | |
| "loss": 0.0, | |
| "step": 1111 | |
| }, | |
| { | |
| "epoch": 6.58, | |
| "learning_rate": 3.560347785875008e-06, | |
| "loss": 0.0, | |
| "step": 1112 | |
| }, | |
| { | |
| "epoch": 6.59, | |
| "learning_rate": 3.54406755438498e-06, | |
| "loss": 0.0, | |
| "step": 1113 | |
| }, | |
| { | |
| "epoch": 6.59, | |
| "learning_rate": 3.5278166098700493e-06, | |
| "loss": 0.0, | |
| "step": 1114 | |
| }, | |
| { | |
| "epoch": 6.6, | |
| "learning_rate": 3.5115950260517083e-06, | |
| "loss": 0.0, | |
| "step": 1115 | |
| }, | |
| { | |
| "epoch": 6.6, | |
| "learning_rate": 3.4954028765182633e-06, | |
| "loss": 0.0, | |
| "step": 1116 | |
| }, | |
| { | |
| "epoch": 6.61, | |
| "learning_rate": 3.479240234724497e-06, | |
| "loss": 0.0, | |
| "step": 1117 | |
| }, | |
| { | |
| "epoch": 6.62, | |
| "learning_rate": 3.463107173991317e-06, | |
| "loss": 0.0, | |
| "step": 1118 | |
| }, | |
| { | |
| "epoch": 6.62, | |
| "learning_rate": 3.4470037675054525e-06, | |
| "loss": 0.0, | |
| "step": 1119 | |
| }, | |
| { | |
| "epoch": 6.63, | |
| "learning_rate": 3.4309300883191065e-06, | |
| "loss": 0.0, | |
| "step": 1120 | |
| }, | |
| { | |
| "epoch": 6.63, | |
| "learning_rate": 3.414886209349615e-06, | |
| "loss": 0.0, | |
| "step": 1121 | |
| }, | |
| { | |
| "epoch": 6.64, | |
| "learning_rate": 3.3988722033791377e-06, | |
| "loss": 0.0, | |
| "step": 1122 | |
| }, | |
| { | |
| "epoch": 6.64, | |
| "learning_rate": 3.3828881430543136e-06, | |
| "loss": 0.0, | |
| "step": 1123 | |
| }, | |
| { | |
| "epoch": 6.65, | |
| "learning_rate": 3.366934100885936e-06, | |
| "loss": 0.0, | |
| "step": 1124 | |
| }, | |
| { | |
| "epoch": 6.66, | |
| "learning_rate": 3.3510101492486224e-06, | |
| "loss": 0.0, | |
| "step": 1125 | |
| }, | |
| { | |
| "epoch": 6.66, | |
| "learning_rate": 3.3351163603804805e-06, | |
| "loss": 0.0, | |
| "step": 1126 | |
| }, | |
| { | |
| "epoch": 6.67, | |
| "learning_rate": 3.3192528063827977e-06, | |
| "loss": 0.0, | |
| "step": 1127 | |
| }, | |
| { | |
| "epoch": 6.67, | |
| "learning_rate": 3.3034195592196906e-06, | |
| "loss": 0.0, | |
| "step": 1128 | |
| }, | |
| { | |
| "epoch": 6.68, | |
| "learning_rate": 3.2876166907177963e-06, | |
| "loss": 0.0, | |
| "step": 1129 | |
| }, | |
| { | |
| "epoch": 6.69, | |
| "learning_rate": 3.271844272565947e-06, | |
| "loss": 0.0, | |
| "step": 1130 | |
| }, | |
| { | |
| "epoch": 6.69, | |
| "learning_rate": 3.2561023763148237e-06, | |
| "loss": 0.0, | |
| "step": 1131 | |
| }, | |
| { | |
| "epoch": 6.7, | |
| "learning_rate": 3.2403910733766595e-06, | |
| "loss": 0.0, | |
| "step": 1132 | |
| }, | |
| { | |
| "epoch": 6.7, | |
| "learning_rate": 3.224710435024897e-06, | |
| "loss": 0.0, | |
| "step": 1133 | |
| }, | |
| { | |
| "epoch": 6.71, | |
| "learning_rate": 3.209060532393872e-06, | |
| "loss": 0.0, | |
| "step": 1134 | |
| }, | |
| { | |
| "epoch": 6.72, | |
| "learning_rate": 3.193441436478493e-06, | |
| "loss": 0.0, | |
| "step": 1135 | |
| }, | |
| { | |
| "epoch": 6.72, | |
| "learning_rate": 3.177853218133905e-06, | |
| "loss": 0.0, | |
| "step": 1136 | |
| }, | |
| { | |
| "epoch": 6.73, | |
| "learning_rate": 3.162295948075188e-06, | |
| "loss": 0.0, | |
| "step": 1137 | |
| }, | |
| { | |
| "epoch": 6.73, | |
| "learning_rate": 3.1467696968770288e-06, | |
| "loss": 0.0, | |
| "step": 1138 | |
| }, | |
| { | |
| "epoch": 6.74, | |
| "learning_rate": 3.1312745349733863e-06, | |
| "loss": 0.0, | |
| "step": 1139 | |
| }, | |
| { | |
| "epoch": 6.75, | |
| "learning_rate": 3.1158105326571963e-06, | |
| "loss": 0.0, | |
| "step": 1140 | |
| }, | |
| { | |
| "epoch": 6.75, | |
| "learning_rate": 3.100377760080041e-06, | |
| "loss": 0.0, | |
| "step": 1141 | |
| }, | |
| { | |
| "epoch": 6.76, | |
| "learning_rate": 3.0849762872518176e-06, | |
| "loss": 0.0, | |
| "step": 1142 | |
| }, | |
| { | |
| "epoch": 6.76, | |
| "learning_rate": 3.0696061840404532e-06, | |
| "loss": 0.0, | |
| "step": 1143 | |
| }, | |
| { | |
| "epoch": 6.77, | |
| "learning_rate": 3.054267520171551e-06, | |
| "loss": 0.0, | |
| "step": 1144 | |
| }, | |
| { | |
| "epoch": 6.78, | |
| "learning_rate": 3.0389603652281006e-06, | |
| "loss": 0.0, | |
| "step": 1145 | |
| }, | |
| { | |
| "epoch": 6.78, | |
| "learning_rate": 3.023684788650154e-06, | |
| "loss": 0.0, | |
| "step": 1146 | |
| }, | |
| { | |
| "epoch": 6.79, | |
| "learning_rate": 3.0084408597345016e-06, | |
| "loss": 0.0, | |
| "step": 1147 | |
| }, | |
| { | |
| "epoch": 6.79, | |
| "learning_rate": 2.993228647634374e-06, | |
| "loss": 0.0, | |
| "step": 1148 | |
| }, | |
| { | |
| "epoch": 6.8, | |
| "learning_rate": 2.9780482213591187e-06, | |
| "loss": 0.0, | |
| "step": 1149 | |
| }, | |
| { | |
| "epoch": 6.8, | |
| "learning_rate": 2.962899649773884e-06, | |
| "loss": 0.0, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 6.81, | |
| "learning_rate": 2.947783001599315e-06, | |
| "loss": 0.0, | |
| "step": 1151 | |
| }, | |
| { | |
| "epoch": 6.82, | |
| "learning_rate": 2.9326983454112414e-06, | |
| "loss": 0.0, | |
| "step": 1152 | |
| }, | |
| { | |
| "epoch": 6.82, | |
| "learning_rate": 2.9176457496403498e-06, | |
| "loss": 0.0, | |
| "step": 1153 | |
| }, | |
| { | |
| "epoch": 6.83, | |
| "learning_rate": 2.9026252825719094e-06, | |
| "loss": 0.0, | |
| "step": 1154 | |
| }, | |
| { | |
| "epoch": 6.83, | |
| "learning_rate": 2.8876370123454146e-06, | |
| "loss": 0.0, | |
| "step": 1155 | |
| }, | |
| { | |
| "epoch": 6.84, | |
| "learning_rate": 2.8726810069543156e-06, | |
| "loss": 0.0, | |
| "step": 1156 | |
| }, | |
| { | |
| "epoch": 6.85, | |
| "learning_rate": 2.857757334245694e-06, | |
| "loss": 0.0, | |
| "step": 1157 | |
| }, | |
| { | |
| "epoch": 6.85, | |
| "learning_rate": 2.842866061919949e-06, | |
| "loss": 0.0, | |
| "step": 1158 | |
| }, | |
| { | |
| "epoch": 6.86, | |
| "learning_rate": 2.8280072575305016e-06, | |
| "loss": 0.0, | |
| "step": 1159 | |
| }, | |
| { | |
| "epoch": 6.86, | |
| "learning_rate": 2.8131809884834895e-06, | |
| "loss": 0.0, | |
| "step": 1160 | |
| }, | |
| { | |
| "epoch": 6.87, | |
| "learning_rate": 2.7983873220374415e-06, | |
| "loss": 0.0, | |
| "step": 1161 | |
| }, | |
| { | |
| "epoch": 6.88, | |
| "learning_rate": 2.7836263253030015e-06, | |
| "loss": 0.0, | |
| "step": 1162 | |
| }, | |
| { | |
| "epoch": 6.88, | |
| "learning_rate": 2.7688980652425957e-06, | |
| "loss": 0.0, | |
| "step": 1163 | |
| }, | |
| { | |
| "epoch": 6.89, | |
| "learning_rate": 2.7542026086701546e-06, | |
| "loss": 0.0, | |
| "step": 1164 | |
| }, | |
| { | |
| "epoch": 6.89, | |
| "learning_rate": 2.739540022250794e-06, | |
| "loss": 0.0, | |
| "step": 1165 | |
| }, | |
| { | |
| "epoch": 6.9, | |
| "learning_rate": 2.724910372500508e-06, | |
| "loss": 0.0, | |
| "step": 1166 | |
| }, | |
| { | |
| "epoch": 6.91, | |
| "learning_rate": 2.7103137257858867e-06, | |
| "loss": 0.0, | |
| "step": 1167 | |
| }, | |
| { | |
| "epoch": 6.91, | |
| "learning_rate": 2.6957501483237924e-06, | |
| "loss": 0.0, | |
| "step": 1168 | |
| }, | |
| { | |
| "epoch": 6.92, | |
| "learning_rate": 2.6812197061810808e-06, | |
| "loss": 0.0, | |
| "step": 1169 | |
| }, | |
| { | |
| "epoch": 6.92, | |
| "learning_rate": 2.6667224652742886e-06, | |
| "loss": 0.0, | |
| "step": 1170 | |
| }, | |
| { | |
| "epoch": 6.93, | |
| "learning_rate": 2.6522584913693295e-06, | |
| "loss": 0.0, | |
| "step": 1171 | |
| }, | |
| { | |
| "epoch": 6.93, | |
| "learning_rate": 2.6378278500812114e-06, | |
| "loss": 0.0, | |
| "step": 1172 | |
| }, | |
| { | |
| "epoch": 6.94, | |
| "learning_rate": 2.6234306068737336e-06, | |
| "loss": 0.0, | |
| "step": 1173 | |
| }, | |
| { | |
| "epoch": 6.95, | |
| "learning_rate": 2.6090668270591714e-06, | |
| "loss": 0.0, | |
| "step": 1174 | |
| }, | |
| { | |
| "epoch": 6.95, | |
| "learning_rate": 2.5947365757980157e-06, | |
| "loss": 0.0, | |
| "step": 1175 | |
| }, | |
| { | |
| "epoch": 6.96, | |
| "learning_rate": 2.5804399180986417e-06, | |
| "loss": 0.0, | |
| "step": 1176 | |
| }, | |
| { | |
| "epoch": 6.96, | |
| "learning_rate": 2.5661769188170336e-06, | |
| "loss": 0.0, | |
| "step": 1177 | |
| }, | |
| { | |
| "epoch": 6.97, | |
| "learning_rate": 2.5519476426564928e-06, | |
| "loss": 0.0, | |
| "step": 1178 | |
| }, | |
| { | |
| "epoch": 6.98, | |
| "learning_rate": 2.5377521541673233e-06, | |
| "loss": 0.0, | |
| "step": 1179 | |
| }, | |
| { | |
| "epoch": 6.98, | |
| "learning_rate": 2.523590517746567e-06, | |
| "loss": 0.0, | |
| "step": 1180 | |
| }, | |
| { | |
| "epoch": 6.99, | |
| "learning_rate": 2.509462797637693e-06, | |
| "loss": 0.0, | |
| "step": 1181 | |
| }, | |
| { | |
| "epoch": 6.99, | |
| "learning_rate": 2.495369057930307e-06, | |
| "loss": 0.0, | |
| "step": 1182 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "learning_rate": 2.4813093625598694e-06, | |
| "loss": 0.0, | |
| "step": 1183 | |
| }, | |
| { | |
| "epoch": 7.01, | |
| "learning_rate": 2.4672837753073986e-06, | |
| "loss": 0.0, | |
| "step": 1184 | |
| }, | |
| { | |
| "epoch": 7.01, | |
| "learning_rate": 2.4532923597991855e-06, | |
| "loss": 0.0, | |
| "step": 1185 | |
| }, | |
| { | |
| "epoch": 7.02, | |
| "learning_rate": 2.4393351795065023e-06, | |
| "loss": 0.0, | |
| "step": 1186 | |
| }, | |
| { | |
| "epoch": 7.02, | |
| "learning_rate": 2.4254122977453087e-06, | |
| "loss": 0.0, | |
| "step": 1187 | |
| }, | |
| { | |
| "epoch": 7.03, | |
| "learning_rate": 2.4115237776759783e-06, | |
| "loss": 0.0, | |
| "step": 1188 | |
| }, | |
| { | |
| "epoch": 7.04, | |
| "learning_rate": 2.3976696823030064e-06, | |
| "loss": 0.0, | |
| "step": 1189 | |
| }, | |
| { | |
| "epoch": 7.04, | |
| "learning_rate": 2.383850074474714e-06, | |
| "loss": 0.0, | |
| "step": 1190 | |
| }, | |
| { | |
| "epoch": 7.05, | |
| "learning_rate": 2.3700650168829765e-06, | |
| "loss": 0.0, | |
| "step": 1191 | |
| }, | |
| { | |
| "epoch": 7.05, | |
| "learning_rate": 2.356314572062939e-06, | |
| "loss": 0.0, | |
| "step": 1192 | |
| }, | |
| { | |
| "epoch": 7.06, | |
| "learning_rate": 2.342598802392716e-06, | |
| "loss": 0.0, | |
| "step": 1193 | |
| }, | |
| { | |
| "epoch": 7.07, | |
| "learning_rate": 2.328917770093129e-06, | |
| "loss": 0.0, | |
| "step": 1194 | |
| }, | |
| { | |
| "epoch": 7.07, | |
| "learning_rate": 2.3152715372274128e-06, | |
| "loss": 0.0, | |
| "step": 1195 | |
| }, | |
| { | |
| "epoch": 7.08, | |
| "learning_rate": 2.3016601657009364e-06, | |
| "loss": 0.0, | |
| "step": 1196 | |
| }, | |
| { | |
| "epoch": 7.08, | |
| "learning_rate": 2.288083717260927e-06, | |
| "loss": 0.0, | |
| "step": 1197 | |
| }, | |
| { | |
| "epoch": 7.09, | |
| "learning_rate": 2.2745422534961735e-06, | |
| "loss": 0.0, | |
| "step": 1198 | |
| }, | |
| { | |
| "epoch": 7.09, | |
| "learning_rate": 2.26103583583677e-06, | |
| "loss": 0.0, | |
| "step": 1199 | |
| }, | |
| { | |
| "epoch": 7.1, | |
| "learning_rate": 2.2475645255538247e-06, | |
| "loss": 0.0, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 7.11, | |
| "learning_rate": 2.234128383759174e-06, | |
| "loss": 0.0, | |
| "step": 1201 | |
| }, | |
| { | |
| "epoch": 7.11, | |
| "learning_rate": 2.2207274714051265e-06, | |
| "loss": 0.0, | |
| "step": 1202 | |
| }, | |
| { | |
| "epoch": 7.12, | |
| "learning_rate": 2.207361849284164e-06, | |
| "loss": 0.0, | |
| "step": 1203 | |
| }, | |
| { | |
| "epoch": 7.12, | |
| "learning_rate": 2.1940315780286823e-06, | |
| "loss": 0.0, | |
| "step": 1204 | |
| }, | |
| { | |
| "epoch": 7.13, | |
| "learning_rate": 2.1807367181107107e-06, | |
| "loss": 0.0, | |
| "step": 1205 | |
| }, | |
| { | |
| "epoch": 7.14, | |
| "learning_rate": 2.167477329841633e-06, | |
| "loss": 0.0, | |
| "step": 1206 | |
| }, | |
| { | |
| "epoch": 7.14, | |
| "learning_rate": 2.1542534733719224e-06, | |
| "loss": 0.0, | |
| "step": 1207 | |
| }, | |
| { | |
| "epoch": 7.15, | |
| "learning_rate": 2.1410652086908567e-06, | |
| "loss": 0.0, | |
| "step": 1208 | |
| }, | |
| { | |
| "epoch": 7.15, | |
| "learning_rate": 2.12791259562626e-06, | |
| "loss": 0.0, | |
| "step": 1209 | |
| }, | |
| { | |
| "epoch": 7.16, | |
| "learning_rate": 2.1147956938442284e-06, | |
| "loss": 0.0, | |
| "step": 1210 | |
| }, | |
| { | |
| "epoch": 7.17, | |
| "learning_rate": 2.101714562848841e-06, | |
| "loss": 0.0, | |
| "step": 1211 | |
| }, | |
| { | |
| "epoch": 7.17, | |
| "learning_rate": 2.0886692619819206e-06, | |
| "loss": 0.0, | |
| "step": 1212 | |
| }, | |
| { | |
| "epoch": 7.18, | |
| "learning_rate": 2.075659850422742e-06, | |
| "loss": 0.0, | |
| "step": 1213 | |
| }, | |
| { | |
| "epoch": 7.18, | |
| "learning_rate": 2.0626863871877677e-06, | |
| "loss": 0.0, | |
| "step": 1214 | |
| }, | |
| { | |
| "epoch": 7.19, | |
| "learning_rate": 2.0497489311303852e-06, | |
| "loss": 0.0, | |
| "step": 1215 | |
| }, | |
| { | |
| "epoch": 7.2, | |
| "learning_rate": 2.0368475409406396e-06, | |
| "loss": 0.0, | |
| "step": 1216 | |
| }, | |
| { | |
| "epoch": 7.2, | |
| "learning_rate": 2.0239822751449623e-06, | |
| "loss": 0.0, | |
| "step": 1217 | |
| }, | |
| { | |
| "epoch": 7.21, | |
| "learning_rate": 2.0111531921059125e-06, | |
| "loss": 0.0, | |
| "step": 1218 | |
| }, | |
| { | |
| "epoch": 7.21, | |
| "learning_rate": 1.9983603500218994e-06, | |
| "loss": 0.0, | |
| "step": 1219 | |
| }, | |
| { | |
| "epoch": 7.22, | |
| "learning_rate": 1.985603806926936e-06, | |
| "loss": 0.0, | |
| "step": 1220 | |
| }, | |
| { | |
| "epoch": 7.22, | |
| "learning_rate": 1.972883620690366e-06, | |
| "loss": 0.0, | |
| "step": 1221 | |
| }, | |
| { | |
| "epoch": 7.23, | |
| "learning_rate": 1.9601998490165943e-06, | |
| "loss": 0.0, | |
| "step": 1222 | |
| }, | |
| { | |
| "epoch": 7.24, | |
| "learning_rate": 1.9475525494448434e-06, | |
| "loss": 0.0, | |
| "step": 1223 | |
| }, | |
| { | |
| "epoch": 7.24, | |
| "learning_rate": 1.934941779348879e-06, | |
| "loss": 0.0, | |
| "step": 1224 | |
| }, | |
| { | |
| "epoch": 7.25, | |
| "learning_rate": 1.9223675959367438e-06, | |
| "loss": 0.0, | |
| "step": 1225 | |
| }, | |
| { | |
| "epoch": 7.25, | |
| "learning_rate": 1.9098300562505266e-06, | |
| "loss": 0.0, | |
| "step": 1226 | |
| }, | |
| { | |
| "epoch": 7.26, | |
| "learning_rate": 1.897329217166063e-06, | |
| "loss": 0.0, | |
| "step": 1227 | |
| }, | |
| { | |
| "epoch": 7.27, | |
| "learning_rate": 1.8848651353927116e-06, | |
| "loss": 0.0, | |
| "step": 1228 | |
| }, | |
| { | |
| "epoch": 7.27, | |
| "learning_rate": 1.8724378674730813e-06, | |
| "loss": 0.0, | |
| "step": 1229 | |
| }, | |
| { | |
| "epoch": 7.28, | |
| "learning_rate": 1.8600474697827708e-06, | |
| "loss": 0.0, | |
| "step": 1230 | |
| }, | |
| { | |
| "epoch": 7.28, | |
| "learning_rate": 1.8476939985301257e-06, | |
| "loss": 0.0, | |
| "step": 1231 | |
| }, | |
| { | |
| "epoch": 7.29, | |
| "learning_rate": 1.8353775097559778e-06, | |
| "loss": 0.0, | |
| "step": 1232 | |
| }, | |
| { | |
| "epoch": 7.3, | |
| "learning_rate": 1.8230980593333814e-06, | |
| "loss": 0.0, | |
| "step": 1233 | |
| }, | |
| { | |
| "epoch": 7.3, | |
| "learning_rate": 1.810855702967379e-06, | |
| "loss": 0.0, | |
| "step": 1234 | |
| }, | |
| { | |
| "epoch": 7.31, | |
| "learning_rate": 1.7986504961947272e-06, | |
| "loss": 0.0, | |
| "step": 1235 | |
| }, | |
| { | |
| "epoch": 7.31, | |
| "learning_rate": 1.7864824943836633e-06, | |
| "loss": 0.0, | |
| "step": 1236 | |
| }, | |
| { | |
| "epoch": 7.32, | |
| "learning_rate": 1.7743517527336506e-06, | |
| "loss": 0.0, | |
| "step": 1237 | |
| }, | |
| { | |
| "epoch": 7.33, | |
| "learning_rate": 1.7622583262751102e-06, | |
| "loss": 0.0, | |
| "step": 1238 | |
| }, | |
| { | |
| "epoch": 7.33, | |
| "learning_rate": 1.7502022698691956e-06, | |
| "loss": 0.0, | |
| "step": 1239 | |
| }, | |
| { | |
| "epoch": 7.34, | |
| "learning_rate": 1.7381836382075301e-06, | |
| "loss": 0.0, | |
| "step": 1240 | |
| }, | |
| { | |
| "epoch": 7.34, | |
| "learning_rate": 1.7262024858119597e-06, | |
| "loss": 0.0, | |
| "step": 1241 | |
| }, | |
| { | |
| "epoch": 7.35, | |
| "learning_rate": 1.7142588670343118e-06, | |
| "loss": 0.0, | |
| "step": 1242 | |
| }, | |
| { | |
| "epoch": 7.36, | |
| "learning_rate": 1.7023528360561382e-06, | |
| "loss": 0.0, | |
| "step": 1243 | |
| }, | |
| { | |
| "epoch": 7.36, | |
| "learning_rate": 1.690484446888484e-06, | |
| "loss": 0.0, | |
| "step": 1244 | |
| }, | |
| { | |
| "epoch": 7.37, | |
| "learning_rate": 1.6786537533716308e-06, | |
| "loss": 0.0, | |
| "step": 1245 | |
| }, | |
| { | |
| "epoch": 7.37, | |
| "learning_rate": 1.6668608091748495e-06, | |
| "loss": 0.0, | |
| "step": 1246 | |
| }, | |
| { | |
| "epoch": 7.38, | |
| "learning_rate": 1.6551056677961785e-06, | |
| "loss": 0.0, | |
| "step": 1247 | |
| }, | |
| { | |
| "epoch": 7.38, | |
| "learning_rate": 1.64338838256215e-06, | |
| "loss": 0.0, | |
| "step": 1248 | |
| }, | |
| { | |
| "epoch": 7.39, | |
| "learning_rate": 1.631709006627572e-06, | |
| "loss": 0.0, | |
| "step": 1249 | |
| }, | |
| { | |
| "epoch": 7.4, | |
| "learning_rate": 1.620067592975282e-06, | |
| "loss": 0.0, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 7.4, | |
| "learning_rate": 1.6084641944158918e-06, | |
| "loss": 0.0, | |
| "step": 1251 | |
| }, | |
| { | |
| "epoch": 7.41, | |
| "learning_rate": 1.5968988635875727e-06, | |
| "loss": 0.0, | |
| "step": 1252 | |
| }, | |
| { | |
| "epoch": 7.41, | |
| "learning_rate": 1.5853716529558006e-06, | |
| "loss": 0.0, | |
| "step": 1253 | |
| }, | |
| { | |
| "epoch": 7.42, | |
| "learning_rate": 1.5738826148131126e-06, | |
| "loss": 0.0, | |
| "step": 1254 | |
| }, | |
| { | |
| "epoch": 7.43, | |
| "learning_rate": 1.5624318012788897e-06, | |
| "loss": 0.0, | |
| "step": 1255 | |
| }, | |
| { | |
| "epoch": 7.43, | |
| "learning_rate": 1.5510192642991073e-06, | |
| "loss": 0.0, | |
| "step": 1256 | |
| }, | |
| { | |
| "epoch": 7.44, | |
| "learning_rate": 1.5396450556460895e-06, | |
| "loss": 0.0, | |
| "step": 1257 | |
| }, | |
| { | |
| "epoch": 7.44, | |
| "learning_rate": 1.5283092269183064e-06, | |
| "loss": 0.0, | |
| "step": 1258 | |
| }, | |
| { | |
| "epoch": 7.45, | |
| "learning_rate": 1.5170118295401005e-06, | |
| "loss": 0.0, | |
| "step": 1259 | |
| }, | |
| { | |
| "epoch": 7.46, | |
| "learning_rate": 1.5057529147614836e-06, | |
| "loss": 0.0, | |
| "step": 1260 | |
| }, | |
| { | |
| "epoch": 7.46, | |
| "learning_rate": 1.494532533657893e-06, | |
| "loss": 0.0, | |
| "step": 1261 | |
| }, | |
| { | |
| "epoch": 7.47, | |
| "learning_rate": 1.4833507371299539e-06, | |
| "loss": 0.0, | |
| "step": 1262 | |
| }, | |
| { | |
| "epoch": 7.47, | |
| "learning_rate": 1.4722075759032583e-06, | |
| "loss": 0.0, | |
| "step": 1263 | |
| }, | |
| { | |
| "epoch": 7.48, | |
| "learning_rate": 1.4611031005281362e-06, | |
| "loss": 0.0, | |
| "step": 1264 | |
| }, | |
| { | |
| "epoch": 7.49, | |
| "learning_rate": 1.4500373613794106e-06, | |
| "loss": 0.0, | |
| "step": 1265 | |
| }, | |
| { | |
| "epoch": 7.49, | |
| "learning_rate": 1.4390104086561886e-06, | |
| "loss": 0.0, | |
| "step": 1266 | |
| }, | |
| { | |
| "epoch": 7.5, | |
| "learning_rate": 1.4280222923816234e-06, | |
| "loss": 0.0, | |
| "step": 1267 | |
| }, | |
| { | |
| "epoch": 7.5, | |
| "learning_rate": 1.4170730624026863e-06, | |
| "loss": 0.0, | |
| "step": 1268 | |
| }, | |
| { | |
| "epoch": 7.51, | |
| "learning_rate": 1.4061627683899458e-06, | |
| "loss": 0.0, | |
| "step": 1269 | |
| }, | |
| { | |
| "epoch": 7.51, | |
| "learning_rate": 1.3952914598373357e-06, | |
| "loss": 0.0, | |
| "step": 1270 | |
| }, | |
| { | |
| "epoch": 7.52, | |
| "learning_rate": 1.3844591860619382e-06, | |
| "loss": 0.0, | |
| "step": 1271 | |
| }, | |
| { | |
| "epoch": 7.53, | |
| "learning_rate": 1.373665996203758e-06, | |
| "loss": 0.0, | |
| "step": 1272 | |
| }, | |
| { | |
| "epoch": 7.53, | |
| "learning_rate": 1.3629119392254885e-06, | |
| "loss": 0.0, | |
| "step": 1273 | |
| }, | |
| { | |
| "epoch": 7.54, | |
| "learning_rate": 1.3521970639123127e-06, | |
| "loss": 0.0, | |
| "step": 1274 | |
| }, | |
| { | |
| "epoch": 7.54, | |
| "learning_rate": 1.341521418871654e-06, | |
| "loss": 0.0, | |
| "step": 1275 | |
| }, | |
| { | |
| "epoch": 7.55, | |
| "learning_rate": 1.330885052532981e-06, | |
| "loss": 0.0, | |
| "step": 1276 | |
| }, | |
| { | |
| "epoch": 7.56, | |
| "learning_rate": 1.3202880131475704e-06, | |
| "loss": 0.0, | |
| "step": 1277 | |
| }, | |
| { | |
| "epoch": 7.56, | |
| "learning_rate": 1.3097303487882983e-06, | |
| "loss": 0.0, | |
| "step": 1278 | |
| }, | |
| { | |
| "epoch": 7.57, | |
| "learning_rate": 1.2992121073494157e-06, | |
| "loss": 0.0, | |
| "step": 1279 | |
| }, | |
| { | |
| "epoch": 7.57, | |
| "learning_rate": 1.288733336546336e-06, | |
| "loss": 0.0, | |
| "step": 1280 | |
| }, | |
| { | |
| "epoch": 7.58, | |
| "learning_rate": 1.2782940839154113e-06, | |
| "loss": 0.0, | |
| "step": 1281 | |
| }, | |
| { | |
| "epoch": 7.59, | |
| "learning_rate": 1.267894396813728e-06, | |
| "loss": 0.0, | |
| "step": 1282 | |
| }, | |
| { | |
| "epoch": 7.59, | |
| "learning_rate": 1.2575343224188796e-06, | |
| "loss": 0.0, | |
| "step": 1283 | |
| }, | |
| { | |
| "epoch": 7.6, | |
| "learning_rate": 1.2472139077287647e-06, | |
| "loss": 0.0, | |
| "step": 1284 | |
| }, | |
| { | |
| "epoch": 7.6, | |
| "learning_rate": 1.2369331995613664e-06, | |
| "loss": 0.0, | |
| "step": 1285 | |
| }, | |
| { | |
| "epoch": 7.61, | |
| "learning_rate": 1.2266922445545348e-06, | |
| "loss": 0.0, | |
| "step": 1286 | |
| }, | |
| { | |
| "epoch": 7.62, | |
| "learning_rate": 1.2164910891657911e-06, | |
| "loss": 0.0, | |
| "step": 1287 | |
| }, | |
| { | |
| "epoch": 7.62, | |
| "learning_rate": 1.206329779672104e-06, | |
| "loss": 0.0, | |
| "step": 1288 | |
| }, | |
| { | |
| "epoch": 7.63, | |
| "learning_rate": 1.1962083621696829e-06, | |
| "loss": 0.0, | |
| "step": 1289 | |
| }, | |
| { | |
| "epoch": 7.63, | |
| "learning_rate": 1.1861268825737715e-06, | |
| "loss": 0.0, | |
| "step": 1290 | |
| }, | |
| { | |
| "epoch": 7.64, | |
| "learning_rate": 1.176085386618434e-06, | |
| "loss": 0.0, | |
| "step": 1291 | |
| }, | |
| { | |
| "epoch": 7.64, | |
| "learning_rate": 1.1660839198563524e-06, | |
| "loss": 0.0, | |
| "step": 1292 | |
| }, | |
| { | |
| "epoch": 7.65, | |
| "learning_rate": 1.1561225276586241e-06, | |
| "loss": 0.0, | |
| "step": 1293 | |
| }, | |
| { | |
| "epoch": 7.66, | |
| "learning_rate": 1.1462012552145385e-06, | |
| "loss": 0.0, | |
| "step": 1294 | |
| }, | |
| { | |
| "epoch": 7.66, | |
| "learning_rate": 1.1363201475313956e-06, | |
| "loss": 0.0, | |
| "step": 1295 | |
| }, | |
| { | |
| "epoch": 7.67, | |
| "learning_rate": 1.1264792494342858e-06, | |
| "loss": 0.0, | |
| "step": 1296 | |
| }, | |
| { | |
| "epoch": 7.67, | |
| "learning_rate": 1.1166786055658886e-06, | |
| "loss": 0.0, | |
| "step": 1297 | |
| }, | |
| { | |
| "epoch": 7.68, | |
| "learning_rate": 1.1069182603862749e-06, | |
| "loss": 0.0, | |
| "step": 1298 | |
| }, | |
| { | |
| "epoch": 7.69, | |
| "learning_rate": 1.0971982581727035e-06, | |
| "loss": 0.0, | |
| "step": 1299 | |
| }, | |
| { | |
| "epoch": 7.69, | |
| "learning_rate": 1.0875186430194196e-06, | |
| "loss": 0.0, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 7.7, | |
| "learning_rate": 1.0778794588374542e-06, | |
| "loss": 0.0, | |
| "step": 1301 | |
| }, | |
| { | |
| "epoch": 7.7, | |
| "learning_rate": 1.0682807493544222e-06, | |
| "loss": 0.0, | |
| "step": 1302 | |
| }, | |
| { | |
| "epoch": 7.71, | |
| "learning_rate": 1.0587225581143323e-06, | |
| "loss": 0.0, | |
| "step": 1303 | |
| }, | |
| { | |
| "epoch": 7.72, | |
| "learning_rate": 1.049204928477384e-06, | |
| "loss": 0.0, | |
| "step": 1304 | |
| }, | |
| { | |
| "epoch": 7.72, | |
| "learning_rate": 1.039727903619767e-06, | |
| "loss": 0.0, | |
| "step": 1305 | |
| }, | |
| { | |
| "epoch": 7.73, | |
| "learning_rate": 1.0302915265334722e-06, | |
| "loss": 0.0, | |
| "step": 1306 | |
| }, | |
| { | |
| "epoch": 7.73, | |
| "learning_rate": 1.0208958400260994e-06, | |
| "loss": 0.0, | |
| "step": 1307 | |
| }, | |
| { | |
| "epoch": 7.74, | |
| "learning_rate": 1.0115408867206466e-06, | |
| "loss": 0.0, | |
| "step": 1308 | |
| }, | |
| { | |
| "epoch": 7.75, | |
| "learning_rate": 1.0022267090553372e-06, | |
| "loss": 0.0, | |
| "step": 1309 | |
| }, | |
| { | |
| "epoch": 7.75, | |
| "learning_rate": 9.929533492834155e-07, | |
| "loss": 0.0, | |
| "step": 1310 | |
| }, | |
| { | |
| "epoch": 7.76, | |
| "learning_rate": 9.837208494729567e-07, | |
| "loss": 0.0, | |
| "step": 1311 | |
| }, | |
| { | |
| "epoch": 7.76, | |
| "learning_rate": 9.745292515066806e-07, | |
| "loss": 0.0, | |
| "step": 1312 | |
| }, | |
| { | |
| "epoch": 7.77, | |
| "learning_rate": 9.653785970817508e-07, | |
| "loss": 0.0, | |
| "step": 1313 | |
| }, | |
| { | |
| "epoch": 7.78, | |
| "learning_rate": 9.562689277096015e-07, | |
| "loss": 0.0, | |
| "step": 1314 | |
| }, | |
| { | |
| "epoch": 7.78, | |
| "learning_rate": 9.472002847157335e-07, | |
| "loss": 0.0, | |
| "step": 1315 | |
| }, | |
| { | |
| "epoch": 7.79, | |
| "learning_rate": 9.381727092395365e-07, | |
| "loss": 0.0, | |
| "step": 1316 | |
| }, | |
| { | |
| "epoch": 7.79, | |
| "learning_rate": 9.291862422341036e-07, | |
| "loss": 0.0, | |
| "step": 1317 | |
| }, | |
| { | |
| "epoch": 7.8, | |
| "learning_rate": 9.202409244660327e-07, | |
| "loss": 0.0, | |
| "step": 1318 | |
| }, | |
| { | |
| "epoch": 7.8, | |
| "learning_rate": 9.113367965152575e-07, | |
| "loss": 0.0, | |
| "step": 1319 | |
| }, | |
| { | |
| "epoch": 7.81, | |
| "learning_rate": 9.024738987748582e-07, | |
| "loss": 0.0, | |
| "step": 1320 | |
| }, | |
| { | |
| "epoch": 7.82, | |
| "learning_rate": 8.936522714508678e-07, | |
| "loss": 0.0, | |
| "step": 1321 | |
| }, | |
| { | |
| "epoch": 7.82, | |
| "learning_rate": 8.848719545621076e-07, | |
| "loss": 0.0, | |
| "step": 1322 | |
| }, | |
| { | |
| "epoch": 7.83, | |
| "learning_rate": 8.761329879399882e-07, | |
| "loss": 0.0, | |
| "step": 1323 | |
| }, | |
| { | |
| "epoch": 7.83, | |
| "learning_rate": 8.674354112283434e-07, | |
| "loss": 0.0, | |
| "step": 1324 | |
| }, | |
| { | |
| "epoch": 7.84, | |
| "learning_rate": 8.587792638832426e-07, | |
| "loss": 0.0, | |
| "step": 1325 | |
| }, | |
| { | |
| "epoch": 7.85, | |
| "learning_rate": 8.501645851728091e-07, | |
| "loss": 0.0, | |
| "step": 1326 | |
| }, | |
| { | |
| "epoch": 7.85, | |
| "learning_rate": 8.41591414177051e-07, | |
| "loss": 0.0, | |
| "step": 1327 | |
| }, | |
| { | |
| "epoch": 7.86, | |
| "learning_rate": 8.330597897876791e-07, | |
| "loss": 0.0, | |
| "step": 1328 | |
| }, | |
| { | |
| "epoch": 7.86, | |
| "learning_rate": 8.245697507079231e-07, | |
| "loss": 0.0, | |
| "step": 1329 | |
| }, | |
| { | |
| "epoch": 7.87, | |
| "learning_rate": 8.161213354523756e-07, | |
| "loss": 0.0, | |
| "step": 1330 | |
| }, | |
| { | |
| "epoch": 7.88, | |
| "learning_rate": 8.077145823467924e-07, | |
| "loss": 0.0, | |
| "step": 1331 | |
| }, | |
| { | |
| "epoch": 7.88, | |
| "learning_rate": 7.993495295279397e-07, | |
| "loss": 0.0, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 7.89, | |
| "learning_rate": 7.910262149434111e-07, | |
| "loss": 0.0, | |
| "step": 1333 | |
| }, | |
| { | |
| "epoch": 7.89, | |
| "learning_rate": 7.827446763514524e-07, | |
| "loss": 0.0, | |
| "step": 1334 | |
| }, | |
| { | |
| "epoch": 7.9, | |
| "learning_rate": 7.745049513208014e-07, | |
| "loss": 0.0, | |
| "step": 1335 | |
| }, | |
| { | |
| "epoch": 7.91, | |
| "learning_rate": 7.663070772305081e-07, | |
| "loss": 0.0, | |
| "step": 1336 | |
| }, | |
| { | |
| "epoch": 7.91, | |
| "learning_rate": 7.581510912697654e-07, | |
| "loss": 0.0, | |
| "step": 1337 | |
| }, | |
| { | |
| "epoch": 7.92, | |
| "learning_rate": 7.50037030437748e-07, | |
| "loss": 0.0, | |
| "step": 1338 | |
| }, | |
| { | |
| "epoch": 7.92, | |
| "learning_rate": 7.419649315434374e-07, | |
| "loss": 0.0, | |
| "step": 1339 | |
| }, | |
| { | |
| "epoch": 7.93, | |
| "learning_rate": 7.33934831205454e-07, | |
| "loss": 0.0, | |
| "step": 1340 | |
| }, | |
| { | |
| "epoch": 7.93, | |
| "learning_rate": 7.259467658519026e-07, | |
| "loss": 0.0, | |
| "step": 1341 | |
| }, | |
| { | |
| "epoch": 7.94, | |
| "learning_rate": 7.180007717201887e-07, | |
| "loss": 0.0, | |
| "step": 1342 | |
| }, | |
| { | |
| "epoch": 7.95, | |
| "learning_rate": 7.10096884856869e-07, | |
| "loss": 0.0, | |
| "step": 1343 | |
| }, | |
| { | |
| "epoch": 7.95, | |
| "learning_rate": 7.022351411174866e-07, | |
| "loss": 0.0, | |
| "step": 1344 | |
| }, | |
| { | |
| "epoch": 7.96, | |
| "learning_rate": 6.944155761663962e-07, | |
| "loss": 0.0, | |
| "step": 1345 | |
| }, | |
| { | |
| "epoch": 7.96, | |
| "learning_rate": 6.866382254766158e-07, | |
| "loss": 0.0, | |
| "step": 1346 | |
| }, | |
| { | |
| "epoch": 7.97, | |
| "learning_rate": 6.789031243296629e-07, | |
| "loss": 0.0, | |
| "step": 1347 | |
| }, | |
| { | |
| "epoch": 7.98, | |
| "learning_rate": 6.712103078153853e-07, | |
| "loss": 0.0, | |
| "step": 1348 | |
| }, | |
| { | |
| "epoch": 7.98, | |
| "learning_rate": 6.635598108318176e-07, | |
| "loss": 0.0, | |
| "step": 1349 | |
| }, | |
| { | |
| "epoch": 7.99, | |
| "learning_rate": 6.559516680850053e-07, | |
| "loss": 0.0, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 7.99, | |
| "learning_rate": 6.483859140888648e-07, | |
| "loss": 0.0, | |
| "step": 1351 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "learning_rate": 6.40862583165014e-07, | |
| "loss": 0.0, | |
| "step": 1352 | |
| }, | |
| { | |
| "epoch": 8.01, | |
| "learning_rate": 6.333817094426198e-07, | |
| "loss": 0.0, | |
| "step": 1353 | |
| }, | |
| { | |
| "epoch": 8.01, | |
| "learning_rate": 6.25943326858246e-07, | |
| "loss": 0.0, | |
| "step": 1354 | |
| }, | |
| { | |
| "epoch": 8.02, | |
| "learning_rate": 6.185474691556981e-07, | |
| "loss": 0.0, | |
| "step": 1355 | |
| }, | |
| { | |
| "epoch": 8.02, | |
| "learning_rate": 6.111941698858681e-07, | |
| "loss": 0.0, | |
| "step": 1356 | |
| }, | |
| { | |
| "epoch": 8.03, | |
| "learning_rate": 6.038834624065881e-07, | |
| "loss": 0.0, | |
| "step": 1357 | |
| }, | |
| { | |
| "epoch": 8.04, | |
| "learning_rate": 5.966153798824692e-07, | |
| "loss": 0.0, | |
| "step": 1358 | |
| }, | |
| { | |
| "epoch": 8.04, | |
| "learning_rate": 5.893899552847604e-07, | |
| "loss": 0.0, | |
| "step": 1359 | |
| }, | |
| { | |
| "epoch": 8.05, | |
| "learning_rate": 5.822072213911956e-07, | |
| "loss": 0.0, | |
| "step": 1360 | |
| }, | |
| { | |
| "epoch": 8.05, | |
| "learning_rate": 5.750672107858435e-07, | |
| "loss": 0.0, | |
| "step": 1361 | |
| }, | |
| { | |
| "epoch": 8.06, | |
| "learning_rate": 5.679699558589624e-07, | |
| "loss": 0.0, | |
| "step": 1362 | |
| }, | |
| { | |
| "epoch": 8.07, | |
| "learning_rate": 5.609154888068491e-07, | |
| "loss": 0.0, | |
| "step": 1363 | |
| }, | |
| { | |
| "epoch": 8.07, | |
| "learning_rate": 5.539038416316978e-07, | |
| "loss": 0.0, | |
| "step": 1364 | |
| }, | |
| { | |
| "epoch": 8.08, | |
| "learning_rate": 5.469350461414535e-07, | |
| "loss": 0.0, | |
| "step": 1365 | |
| }, | |
| { | |
| "epoch": 8.08, | |
| "learning_rate": 5.400091339496638e-07, | |
| "loss": 0.0, | |
| "step": 1366 | |
| }, | |
| { | |
| "epoch": 8.09, | |
| "learning_rate": 5.331261364753415e-07, | |
| "loss": 0.0, | |
| "step": 1367 | |
| }, | |
| { | |
| "epoch": 8.09, | |
| "learning_rate": 5.262860849428197e-07, | |
| "loss": 0.0, | |
| "step": 1368 | |
| }, | |
| { | |
| "epoch": 8.1, | |
| "learning_rate": 5.194890103816041e-07, | |
| "loss": 0.0, | |
| "step": 1369 | |
| }, | |
| { | |
| "epoch": 8.11, | |
| "learning_rate": 5.127349436262452e-07, | |
| "loss": 0.0, | |
| "step": 1370 | |
| }, | |
| { | |
| "epoch": 8.11, | |
| "learning_rate": 5.060239153161872e-07, | |
| "loss": 0.0, | |
| "step": 1371 | |
| }, | |
| { | |
| "epoch": 8.12, | |
| "learning_rate": 4.993559558956318e-07, | |
| "loss": 0.0, | |
| "step": 1372 | |
| }, | |
| { | |
| "epoch": 8.12, | |
| "learning_rate": 4.927310956134057e-07, | |
| "loss": 0.0, | |
| "step": 1373 | |
| }, | |
| { | |
| "epoch": 8.13, | |
| "learning_rate": 4.861493645228132e-07, | |
| "loss": 0.0, | |
| "step": 1374 | |
| }, | |
| { | |
| "epoch": 8.14, | |
| "learning_rate": 4.796107924815074e-07, | |
| "loss": 0.0, | |
| "step": 1375 | |
| }, | |
| { | |
| "epoch": 8.14, | |
| "learning_rate": 4.731154091513546e-07, | |
| "loss": 0.0, | |
| "step": 1376 | |
| }, | |
| { | |
| "epoch": 8.15, | |
| "learning_rate": 4.6666324399829655e-07, | |
| "loss": 0.0, | |
| "step": 1377 | |
| }, | |
| { | |
| "epoch": 8.15, | |
| "learning_rate": 4.6025432629221655e-07, | |
| "loss": 0.0, | |
| "step": 1378 | |
| }, | |
| { | |
| "epoch": 8.16, | |
| "learning_rate": 4.538886851068136e-07, | |
| "loss": 0.0, | |
| "step": 1379 | |
| }, | |
| { | |
| "epoch": 8.17, | |
| "learning_rate": 4.4756634931945707e-07, | |
| "loss": 0.0, | |
| "step": 1380 | |
| }, | |
| { | |
| "epoch": 8.17, | |
| "learning_rate": 4.412873476110702e-07, | |
| "loss": 0.0, | |
| "step": 1381 | |
| }, | |
| { | |
| "epoch": 8.18, | |
| "learning_rate": 4.350517084659933e-07, | |
| "loss": 0.0, | |
| "step": 1382 | |
| }, | |
| { | |
| "epoch": 8.18, | |
| "learning_rate": 4.2885946017185096e-07, | |
| "loss": 0.0, | |
| "step": 1383 | |
| }, | |
| { | |
| "epoch": 8.19, | |
| "learning_rate": 4.2271063081943287e-07, | |
| "loss": 0.0, | |
| "step": 1384 | |
| }, | |
| { | |
| "epoch": 8.2, | |
| "learning_rate": 4.1660524830255513e-07, | |
| "loss": 0.0, | |
| "step": 1385 | |
| }, | |
| { | |
| "epoch": 8.2, | |
| "learning_rate": 4.1054334031794373e-07, | |
| "loss": 0.0, | |
| "step": 1386 | |
| }, | |
| { | |
| "epoch": 8.21, | |
| "learning_rate": 4.0452493436510474e-07, | |
| "loss": 0.0, | |
| "step": 1387 | |
| }, | |
| { | |
| "epoch": 8.21, | |
| "learning_rate": 3.985500577461976e-07, | |
| "loss": 0.0, | |
| "step": 1388 | |
| }, | |
| { | |
| "epoch": 8.22, | |
| "learning_rate": 3.9261873756591627e-07, | |
| "loss": 0.0, | |
| "step": 1389 | |
| }, | |
| { | |
| "epoch": 8.22, | |
| "learning_rate": 3.867310007313596e-07, | |
| "loss": 0.0, | |
| "step": 1390 | |
| }, | |
| { | |
| "epoch": 8.23, | |
| "learning_rate": 3.808868739519167e-07, | |
| "loss": 0.0, | |
| "step": 1391 | |
| }, | |
| { | |
| "epoch": 8.24, | |
| "learning_rate": 3.750863837391405e-07, | |
| "loss": 0.0, | |
| "step": 1392 | |
| }, | |
| { | |
| "epoch": 8.24, | |
| "learning_rate": 3.693295564066279e-07, | |
| "loss": 0.0, | |
| "step": 1393 | |
| }, | |
| { | |
| "epoch": 8.25, | |
| "learning_rate": 3.636164180699064e-07, | |
| "loss": 0.0, | |
| "step": 1394 | |
| }, | |
| { | |
| "epoch": 8.25, | |
| "learning_rate": 3.5794699464630324e-07, | |
| "loss": 0.0, | |
| "step": 1395 | |
| }, | |
| { | |
| "epoch": 8.26, | |
| "learning_rate": 3.5232131185484075e-07, | |
| "loss": 0.0, | |
| "step": 1396 | |
| }, | |
| { | |
| "epoch": 8.27, | |
| "learning_rate": 3.4673939521611466e-07, | |
| "loss": 0.0, | |
| "step": 1397 | |
| }, | |
| { | |
| "epoch": 8.27, | |
| "learning_rate": 3.4120127005217497e-07, | |
| "loss": 0.0, | |
| "step": 1398 | |
| }, | |
| { | |
| "epoch": 8.28, | |
| "learning_rate": 3.3570696148641504e-07, | |
| "loss": 0.0, | |
| "step": 1399 | |
| }, | |
| { | |
| "epoch": 8.28, | |
| "learning_rate": 3.302564944434594e-07, | |
| "loss": 0.0, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 8.29, | |
| "learning_rate": 3.2484989364904295e-07, | |
| "loss": 0.0, | |
| "step": 1401 | |
| }, | |
| { | |
| "epoch": 8.3, | |
| "learning_rate": 3.194871836299085e-07, | |
| "loss": 0.0, | |
| "step": 1402 | |
| }, | |
| { | |
| "epoch": 8.3, | |
| "learning_rate": 3.1416838871368925e-07, | |
| "loss": 0.0, | |
| "step": 1403 | |
| }, | |
| { | |
| "epoch": 8.31, | |
| "learning_rate": 3.0889353302879786e-07, | |
| "loss": 0.0, | |
| "step": 1404 | |
| }, | |
| { | |
| "epoch": 8.31, | |
| "learning_rate": 3.0366264050432416e-07, | |
| "loss": 0.0, | |
| "step": 1405 | |
| }, | |
| { | |
| "epoch": 8.32, | |
| "learning_rate": 2.984757348699152e-07, | |
| "loss": 0.0, | |
| "step": 1406 | |
| }, | |
| { | |
| "epoch": 8.33, | |
| "learning_rate": 2.9333283965567896e-07, | |
| "loss": 0.0, | |
| "step": 1407 | |
| }, | |
| { | |
| "epoch": 8.33, | |
| "learning_rate": 2.8823397819207066e-07, | |
| "loss": 0.0, | |
| "step": 1408 | |
| }, | |
| { | |
| "epoch": 8.34, | |
| "learning_rate": 2.8317917360978774e-07, | |
| "loss": 0.0, | |
| "step": 1409 | |
| }, | |
| { | |
| "epoch": 8.34, | |
| "learning_rate": 2.7816844883966963e-07, | |
| "loss": 0.0, | |
| "step": 1410 | |
| }, | |
| { | |
| "epoch": 8.35, | |
| "learning_rate": 2.7320182661258687e-07, | |
| "loss": 0.0, | |
| "step": 1411 | |
| }, | |
| { | |
| "epoch": 8.36, | |
| "learning_rate": 2.6827932945934e-07, | |
| "loss": 0.0, | |
| "step": 1412 | |
| }, | |
| { | |
| "epoch": 8.36, | |
| "learning_rate": 2.6340097971056635e-07, | |
| "loss": 0.0, | |
| "step": 1413 | |
| }, | |
| { | |
| "epoch": 8.37, | |
| "learning_rate": 2.5856679949662124e-07, | |
| "loss": 0.0, | |
| "step": 1414 | |
| }, | |
| { | |
| "epoch": 8.37, | |
| "learning_rate": 2.537768107474925e-07, | |
| "loss": 0.0, | |
| "step": 1415 | |
| }, | |
| { | |
| "epoch": 8.38, | |
| "learning_rate": 2.4903103519269724e-07, | |
| "loss": 0.0, | |
| "step": 1416 | |
| }, | |
| { | |
| "epoch": 8.38, | |
| "learning_rate": 2.4432949436117737e-07, | |
| "loss": 0.0, | |
| "step": 1417 | |
| }, | |
| { | |
| "epoch": 8.39, | |
| "learning_rate": 2.3967220958120986e-07, | |
| "loss": 0.0, | |
| "step": 1418 | |
| }, | |
| { | |
| "epoch": 8.4, | |
| "learning_rate": 2.3505920198030662e-07, | |
| "loss": 0.0, | |
| "step": 1419 | |
| }, | |
| { | |
| "epoch": 8.4, | |
| "learning_rate": 2.30490492485117e-07, | |
| "loss": 0.0, | |
| "step": 1420 | |
| }, | |
| { | |
| "epoch": 8.41, | |
| "learning_rate": 2.2596610182133328e-07, | |
| "loss": 0.0, | |
| "step": 1421 | |
| }, | |
| { | |
| "epoch": 8.41, | |
| "learning_rate": 2.2148605051360295e-07, | |
| "loss": 0.0, | |
| "step": 1422 | |
| }, | |
| { | |
| "epoch": 8.42, | |
| "learning_rate": 2.1705035888542448e-07, | |
| "loss": 0.0, | |
| "step": 1423 | |
| }, | |
| { | |
| "epoch": 8.43, | |
| "learning_rate": 2.1265904705906616e-07, | |
| "loss": 0.0, | |
| "step": 1424 | |
| }, | |
| { | |
| "epoch": 8.43, | |
| "learning_rate": 2.0831213495546622e-07, | |
| "loss": 0.0, | |
| "step": 1425 | |
| }, | |
| { | |
| "epoch": 8.44, | |
| "learning_rate": 2.0400964229414732e-07, | |
| "loss": 0.0, | |
| "step": 1426 | |
| }, | |
| { | |
| "epoch": 8.44, | |
| "learning_rate": 1.997515885931267e-07, | |
| "loss": 0.0, | |
| "step": 1427 | |
| }, | |
| { | |
| "epoch": 8.45, | |
| "learning_rate": 1.9553799316882504e-07, | |
| "loss": 0.0, | |
| "step": 1428 | |
| }, | |
| { | |
| "epoch": 8.46, | |
| "learning_rate": 1.9136887513598213e-07, | |
| "loss": 0.0, | |
| "step": 1429 | |
| }, | |
| { | |
| "epoch": 8.46, | |
| "learning_rate": 1.8724425340756692e-07, | |
| "loss": 0.0, | |
| "step": 1430 | |
| }, | |
| { | |
| "epoch": 8.47, | |
| "learning_rate": 1.831641466946954e-07, | |
| "loss": 0.0, | |
| "step": 1431 | |
| }, | |
| { | |
| "epoch": 8.47, | |
| "learning_rate": 1.7912857350654288e-07, | |
| "loss": 0.0, | |
| "step": 1432 | |
| }, | |
| { | |
| "epoch": 8.48, | |
| "learning_rate": 1.751375521502574e-07, | |
| "loss": 0.0, | |
| "step": 1433 | |
| }, | |
| { | |
| "epoch": 8.49, | |
| "learning_rate": 1.7119110073088862e-07, | |
| "loss": 0.0, | |
| "step": 1434 | |
| }, | |
| { | |
| "epoch": 8.49, | |
| "learning_rate": 1.6728923715128686e-07, | |
| "loss": 0.0, | |
| "step": 1435 | |
| }, | |
| { | |
| "epoch": 8.5, | |
| "learning_rate": 1.6343197911203978e-07, | |
| "loss": 0.0, | |
| "step": 1436 | |
| }, | |
| { | |
| "epoch": 8.5, | |
| "learning_rate": 1.5961934411138248e-07, | |
| "loss": 0.0, | |
| "step": 1437 | |
| }, | |
| { | |
| "epoch": 8.51, | |
| "learning_rate": 1.558513494451175e-07, | |
| "loss": 0.0, | |
| "step": 1438 | |
| }, | |
| { | |
| "epoch": 8.51, | |
| "learning_rate": 1.5212801220654273e-07, | |
| "loss": 0.0, | |
| "step": 1439 | |
| }, | |
| { | |
| "epoch": 8.52, | |
| "learning_rate": 1.484493492863681e-07, | |
| "loss": 0.0, | |
| "step": 1440 | |
| }, | |
| { | |
| "epoch": 8.53, | |
| "learning_rate": 1.448153773726402e-07, | |
| "loss": 0.0, | |
| "step": 1441 | |
| }, | |
| { | |
| "epoch": 8.53, | |
| "learning_rate": 1.412261129506698e-07, | |
| "loss": 0.0, | |
| "step": 1442 | |
| }, | |
| { | |
| "epoch": 8.54, | |
| "learning_rate": 1.3768157230295453e-07, | |
| "loss": 0.0, | |
| "step": 1443 | |
| }, | |
| { | |
| "epoch": 8.54, | |
| "learning_rate": 1.3418177150910094e-07, | |
| "loss": 0.0, | |
| "step": 1444 | |
| }, | |
| { | |
| "epoch": 8.55, | |
| "learning_rate": 1.3072672644576235e-07, | |
| "loss": 0.0, | |
| "step": 1445 | |
| }, | |
| { | |
| "epoch": 8.56, | |
| "learning_rate": 1.2731645278655448e-07, | |
| "loss": 0.0, | |
| "step": 1446 | |
| }, | |
| { | |
| "epoch": 8.56, | |
| "learning_rate": 1.239509660019922e-07, | |
| "loss": 0.0, | |
| "step": 1447 | |
| }, | |
| { | |
| "epoch": 8.57, | |
| "learning_rate": 1.2063028135941956e-07, | |
| "loss": 0.0, | |
| "step": 1448 | |
| }, | |
| { | |
| "epoch": 8.57, | |
| "learning_rate": 1.1735441392293433e-07, | |
| "loss": 0.0, | |
| "step": 1449 | |
| }, | |
| { | |
| "epoch": 8.58, | |
| "learning_rate": 1.1412337855332467e-07, | |
| "loss": 0.0, | |
| "step": 1450 | |
| }, | |
| { | |
| "epoch": 8.59, | |
| "learning_rate": 1.109371899080025e-07, | |
| "loss": 0.0, | |
| "step": 1451 | |
| }, | |
| { | |
| "epoch": 8.59, | |
| "learning_rate": 1.0779586244093254e-07, | |
| "loss": 0.0, | |
| "step": 1452 | |
| }, | |
| { | |
| "epoch": 8.6, | |
| "learning_rate": 1.0469941040257114e-07, | |
| "loss": 0.0, | |
| "step": 1453 | |
| }, | |
| { | |
| "epoch": 8.6, | |
| "learning_rate": 1.0164784783979864e-07, | |
| "loss": 0.0, | |
| "step": 1454 | |
| }, | |
| { | |
| "epoch": 8.61, | |
| "learning_rate": 9.864118859585825e-08, | |
| "loss": 0.0, | |
| "step": 1455 | |
| }, | |
| { | |
| "epoch": 8.62, | |
| "learning_rate": 9.56794463102917e-08, | |
| "loss": 0.0, | |
| "step": 1456 | |
| }, | |
| { | |
| "epoch": 8.62, | |
| "learning_rate": 9.276263441887368e-08, | |
| "loss": 0.0, | |
| "step": 1457 | |
| }, | |
| { | |
| "epoch": 8.63, | |
| "learning_rate": 8.989076615355864e-08, | |
| "loss": 0.0, | |
| "step": 1458 | |
| }, | |
| { | |
| "epoch": 8.63, | |
| "learning_rate": 8.706385454241739e-08, | |
| "loss": 0.0, | |
| "step": 1459 | |
| }, | |
| { | |
| "epoch": 8.64, | |
| "learning_rate": 8.428191240957506e-08, | |
| "loss": 0.0, | |
| "step": 1460 | |
| }, | |
| { | |
| "epoch": 8.64, | |
| "learning_rate": 8.154495237515436e-08, | |
| "loss": 0.0, | |
| "step": 1461 | |
| }, | |
| { | |
| "epoch": 8.65, | |
| "learning_rate": 7.885298685522235e-08, | |
| "loss": 0.0, | |
| "step": 1462 | |
| }, | |
| { | |
| "epoch": 8.66, | |
| "learning_rate": 7.620602806172828e-08, | |
| "loss": 0.0, | |
| "step": 1463 | |
| }, | |
| { | |
| "epoch": 8.66, | |
| "learning_rate": 7.360408800245355e-08, | |
| "loss": 0.0, | |
| "step": 1464 | |
| }, | |
| { | |
| "epoch": 8.67, | |
| "learning_rate": 7.104717848095188e-08, | |
| "loss": 0.0, | |
| "step": 1465 | |
| }, | |
| { | |
| "epoch": 8.67, | |
| "learning_rate": 6.853531109650147e-08, | |
| "loss": 0.0, | |
| "step": 1466 | |
| }, | |
| { | |
| "epoch": 8.68, | |
| "learning_rate": 6.606849724404618e-08, | |
| "loss": 0.0, | |
| "step": 1467 | |
| }, | |
| { | |
| "epoch": 8.69, | |
| "learning_rate": 6.364674811414895e-08, | |
| "loss": 0.0, | |
| "step": 1468 | |
| }, | |
| { | |
| "epoch": 8.69, | |
| "learning_rate": 6.127007469293845e-08, | |
| "loss": 0.0, | |
| "step": 1469 | |
| }, | |
| { | |
| "epoch": 8.7, | |
| "learning_rate": 5.893848776206024e-08, | |
| "loss": 0.0, | |
| "step": 1470 | |
| }, | |
| { | |
| "epoch": 8.7, | |
| "learning_rate": 5.665199789862907e-08, | |
| "loss": 0.0, | |
| "step": 1471 | |
| }, | |
| { | |
| "epoch": 8.71, | |
| "learning_rate": 5.441061547517557e-08, | |
| "loss": 0.0, | |
| "step": 1472 | |
| }, | |
| { | |
| "epoch": 8.72, | |
| "learning_rate": 5.221435065960734e-08, | |
| "loss": 0.0, | |
| "step": 1473 | |
| }, | |
| { | |
| "epoch": 8.72, | |
| "learning_rate": 5.006321341515463e-08, | |
| "loss": 0.0, | |
| "step": 1474 | |
| }, | |
| { | |
| "epoch": 8.73, | |
| "learning_rate": 4.7957213500331445e-08, | |
| "loss": 0.0, | |
| "step": 1475 | |
| }, | |
| { | |
| "epoch": 8.73, | |
| "learning_rate": 4.589636046888779e-08, | |
| "loss": 0.0, | |
| "step": 1476 | |
| }, | |
| { | |
| "epoch": 8.74, | |
| "learning_rate": 4.388066366976973e-08, | |
| "loss": 0.0, | |
| "step": 1477 | |
| }, | |
| { | |
| "epoch": 8.75, | |
| "learning_rate": 4.191013224707052e-08, | |
| "loss": 0.0, | |
| "step": 1478 | |
| }, | |
| { | |
| "epoch": 8.75, | |
| "learning_rate": 3.9984775139995106e-08, | |
| "loss": 0.0, | |
| "step": 1479 | |
| }, | |
| { | |
| "epoch": 8.76, | |
| "learning_rate": 3.81046010828201e-08, | |
| "loss": 0.0, | |
| "step": 1480 | |
| }, | |
| { | |
| "epoch": 8.76, | |
| "learning_rate": 3.626961860484723e-08, | |
| "loss": 0.0, | |
| "step": 1481 | |
| }, | |
| { | |
| "epoch": 8.77, | |
| "learning_rate": 3.4479836030372195e-08, | |
| "loss": 0.0, | |
| "step": 1482 | |
| }, | |
| { | |
| "epoch": 8.78, | |
| "learning_rate": 3.2735261478644695e-08, | |
| "loss": 0.0, | |
| "step": 1483 | |
| }, | |
| { | |
| "epoch": 8.78, | |
| "learning_rate": 3.10359028638274e-08, | |
| "loss": 0.0, | |
| "step": 1484 | |
| }, | |
| { | |
| "epoch": 8.79, | |
| "learning_rate": 2.938176789496483e-08, | |
| "loss": 0.0, | |
| "step": 1485 | |
| }, | |
| { | |
| "epoch": 8.79, | |
| "learning_rate": 2.7772864075950036e-08, | |
| "loss": 0.0, | |
| "step": 1486 | |
| }, | |
| { | |
| "epoch": 8.8, | |
| "learning_rate": 2.620919870548133e-08, | |
| "loss": 0.0, | |
| "step": 1487 | |
| }, | |
| { | |
| "epoch": 8.8, | |
| "learning_rate": 2.46907788770423e-08, | |
| "loss": 0.0, | |
| "step": 1488 | |
| }, | |
| { | |
| "epoch": 8.81, | |
| "learning_rate": 2.3217611478857373e-08, | |
| "loss": 0.0, | |
| "step": 1489 | |
| }, | |
| { | |
| "epoch": 8.82, | |
| "learning_rate": 2.1789703193868527e-08, | |
| "loss": 0.0, | |
| "step": 1490 | |
| }, | |
| { | |
| "epoch": 8.82, | |
| "learning_rate": 2.040706049970087e-08, | |
| "loss": 0.0, | |
| "step": 1491 | |
| }, | |
| { | |
| "epoch": 8.83, | |
| "learning_rate": 1.9069689668634873e-08, | |
| "loss": 0.0, | |
| "step": 1492 | |
| }, | |
| { | |
| "epoch": 8.83, | |
| "learning_rate": 1.777759676757973e-08, | |
| "loss": 0.0, | |
| "step": 1493 | |
| }, | |
| { | |
| "epoch": 8.84, | |
| "learning_rate": 1.653078765804117e-08, | |
| "loss": 0.0, | |
| "step": 1494 | |
| }, | |
| { | |
| "epoch": 8.85, | |
| "learning_rate": 1.532926799609924e-08, | |
| "loss": 0.0, | |
| "step": 1495 | |
| }, | |
| { | |
| "epoch": 8.85, | |
| "learning_rate": 1.4173043232380557e-08, | |
| "loss": 0.0, | |
| "step": 1496 | |
| }, | |
| { | |
| "epoch": 8.86, | |
| "learning_rate": 1.3062118612033881e-08, | |
| "loss": 0.0, | |
| "step": 1497 | |
| }, | |
| { | |
| "epoch": 8.86, | |
| "learning_rate": 1.199649917470569e-08, | |
| "loss": 0.0, | |
| "step": 1498 | |
| }, | |
| { | |
| "epoch": 8.87, | |
| "learning_rate": 1.0976189754519085e-08, | |
| "loss": 0.0, | |
| "step": 1499 | |
| }, | |
| { | |
| "epoch": 8.88, | |
| "learning_rate": 1.0001194980050477e-08, | |
| "loss": 0.0, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 8.88, | |
| "learning_rate": 9.071519274308494e-09, | |
| "loss": 0.0, | |
| "step": 1501 | |
| }, | |
| { | |
| "epoch": 8.89, | |
| "learning_rate": 8.18716685471621e-09, | |
| "loss": 0.0, | |
| "step": 1502 | |
| }, | |
| { | |
| "epoch": 8.89, | |
| "learning_rate": 7.348141733087844e-09, | |
| "loss": 0.0, | |
| "step": 1503 | |
| }, | |
| { | |
| "epoch": 8.9, | |
| "learning_rate": 6.5544477156131994e-09, | |
| "loss": 0.0, | |
| "step": 1504 | |
| }, | |
| { | |
| "epoch": 8.91, | |
| "learning_rate": 5.8060884028410255e-09, | |
| "loss": 0.0, | |
| "step": 1505 | |
| }, | |
| { | |
| "epoch": 8.91, | |
| "learning_rate": 5.1030671896623585e-09, | |
| "loss": 0.0, | |
| "step": 1506 | |
| }, | |
| { | |
| "epoch": 8.92, | |
| "learning_rate": 4.445387265291645e-09, | |
| "loss": 0.0, | |
| "step": 1507 | |
| }, | |
| { | |
| "epoch": 8.92, | |
| "learning_rate": 3.833051613257866e-09, | |
| "loss": 0.0, | |
| "step": 1508 | |
| }, | |
| { | |
| "epoch": 8.93, | |
| "learning_rate": 3.2660630113878765e-09, | |
| "loss": 0.0, | |
| "step": 1509 | |
| }, | |
| { | |
| "epoch": 8.93, | |
| "learning_rate": 2.7444240317930913e-09, | |
| "loss": 0.0, | |
| "step": 1510 | |
| }, | |
| { | |
| "epoch": 8.94, | |
| "learning_rate": 2.268137040859486e-09, | |
| "loss": 0.0, | |
| "step": 1511 | |
| }, | |
| { | |
| "epoch": 8.95, | |
| "learning_rate": 1.8372041992353872e-09, | |
| "loss": 0.0, | |
| "step": 1512 | |
| }, | |
| { | |
| "epoch": 8.95, | |
| "learning_rate": 1.4516274618248117e-09, | |
| "loss": 0.0, | |
| "step": 1513 | |
| }, | |
| { | |
| "epoch": 8.96, | |
| "learning_rate": 1.1114085777741422e-09, | |
| "loss": 0.0, | |
| "step": 1514 | |
| }, | |
| { | |
| "epoch": 8.96, | |
| "learning_rate": 8.165490904676887e-10, | |
| "loss": 0.0, | |
| "step": 1515 | |
| }, | |
| { | |
| "epoch": 8.96, | |
| "eval_test_accuracy": 48.83268482490272, | |
| "eval_test_average": 48.83268482490272, | |
| "eval_test_loss": 1.2939674854278564, | |
| "eval_test_runtime": 1273.1058, | |
| "eval_test_samples_per_second": 1.211, | |
| "eval_test_steps_per_second": 0.01, | |
| "step": 1515 | |
| }, | |
| { | |
| "epoch": 8.97, | |
| "learning_rate": 5.670503375188041e-10, | |
| "loss": 0.0, | |
| "step": 1516 | |
| }, | |
| { | |
| "epoch": 8.98, | |
| "learning_rate": 3.6291345076433505e-10, | |
| "loss": 0.0, | |
| "step": 1517 | |
| }, | |
| { | |
| "epoch": 8.98, | |
| "learning_rate": 2.0413935626129032e-10, | |
| "loss": 0.0, | |
| "step": 1518 | |
| }, | |
| { | |
| "epoch": 8.99, | |
| "learning_rate": 9.072877427684923e-11, | |
| "loss": 0.0, | |
| "step": 1519 | |
| }, | |
| { | |
| "epoch": 8.99, | |
| "learning_rate": 2.268221929280223e-11, | |
| "loss": 0.0, | |
| "step": 1520 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "learning_rate": 0.0, | |
| "loss": 0.0, | |
| "step": 1521 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "step": 1521, | |
| "total_flos": 6.940909639827456e+17, | |
| "train_loss": 0.007590809710388884, | |
| "train_runtime": 10567.5001, | |
| "train_samples_per_second": 9.184, | |
| "train_steps_per_second": 0.144 | |
| } | |
| ], | |
| "max_steps": 1521, | |
| "num_train_epochs": 9, | |
| "total_flos": 6.940909639827456e+17, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |