| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 1.2053878141384922, | |
| "global_step": 15000, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 1e-05, | |
| "loss": 1.9456, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 1.3010299956639812e-05, | |
| "loss": 1.6437, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 1.4771212547196626e-05, | |
| "loss": 1.6438, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 1.6020599913279625e-05, | |
| "loss": 1.493, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 1.6989700043360187e-05, | |
| "loss": 1.5275, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 1.7781512503836436e-05, | |
| "loss": 1.4028, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 1.845098040014257e-05, | |
| "loss": 1.4608, | |
| "step": 70 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 1.9030899869919437e-05, | |
| "loss": 1.2957, | |
| "step": 80 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 1.954242509439325e-05, | |
| "loss": 1.3968, | |
| "step": 90 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 2e-05, | |
| "loss": 1.3591, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 2.0413926851582255e-05, | |
| "loss": 1.3462, | |
| "step": 110 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 2.0791812460476247e-05, | |
| "loss": 1.3115, | |
| "step": 120 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 2.1139433523068368e-05, | |
| "loss": 1.2937, | |
| "step": 130 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 2.1461280356782382e-05, | |
| "loss": 1.2266, | |
| "step": 140 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 2.1760912590556812e-05, | |
| "loss": 1.1994, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 2.204119982655925e-05, | |
| "loss": 1.2548, | |
| "step": 160 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 2.2304489213782743e-05, | |
| "loss": 1.1843, | |
| "step": 170 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 2.255272505103306e-05, | |
| "loss": 1.2799, | |
| "step": 180 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 2.278753600952829e-05, | |
| "loss": 1.167, | |
| "step": 190 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 2.301029995663981e-05, | |
| "loss": 1.129, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 2.3222192947339193e-05, | |
| "loss": 1.2135, | |
| "step": 210 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 2.3424226808222067e-05, | |
| "loss": 1.2332, | |
| "step": 220 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 2.361727836017593e-05, | |
| "loss": 1.1461, | |
| "step": 230 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 2.380211241711606e-05, | |
| "loss": 1.1849, | |
| "step": 240 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 2.3979400086720374e-05, | |
| "loss": 1.2342, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 2.414973347970818e-05, | |
| "loss": 1.2376, | |
| "step": 260 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 2.4313637641589878e-05, | |
| "loss": 1.1966, | |
| "step": 270 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 2.4471580313422194e-05, | |
| "loss": 1.2426, | |
| "step": 280 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 2.462397997898956e-05, | |
| "loss": 1.1851, | |
| "step": 290 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 2.4771212547196627e-05, | |
| "loss": 1.1906, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 2.491361693834273e-05, | |
| "loss": 1.1677, | |
| "step": 310 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 2.505149978319906e-05, | |
| "loss": 1.2344, | |
| "step": 320 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 2.5185139398778874e-05, | |
| "loss": 1.2005, | |
| "step": 330 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 2.5314789170422555e-05, | |
| "loss": 1.2777, | |
| "step": 340 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 2.5440680443502755e-05, | |
| "loss": 1.2101, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 2.5563025007672876e-05, | |
| "loss": 1.1983, | |
| "step": 360 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 2.568201724066995e-05, | |
| "loss": 1.2566, | |
| "step": 370 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 2.5797835966168102e-05, | |
| "loss": 1.1816, | |
| "step": 380 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 2.5910646070264996e-05, | |
| "loss": 1.2346, | |
| "step": 390 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 2.6020599913279625e-05, | |
| "loss": 1.2128, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 2.612783856719736e-05, | |
| "loss": 1.1537, | |
| "step": 410 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 2.6232492903979004e-05, | |
| "loss": 1.1787, | |
| "step": 420 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 2.6334684555795868e-05, | |
| "loss": 1.1315, | |
| "step": 430 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 2.6434526764861875e-05, | |
| "loss": 1.2758, | |
| "step": 440 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 2.653212513775344e-05, | |
| "loss": 1.1824, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 2.6627578316815746e-05, | |
| "loss": 1.193, | |
| "step": 460 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 2.6720978579357177e-05, | |
| "loss": 1.2263, | |
| "step": 470 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 2.6812412373755874e-05, | |
| "loss": 1.1467, | |
| "step": 480 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 2.690196080028514e-05, | |
| "loss": 1.1846, | |
| "step": 490 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 2.698970004336019e-05, | |
| "loss": 1.1736, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "eval_get_denotation_accuracy": 0.4383124846701006, | |
| "eval_loss": 0.9204946160316467, | |
| "eval_runtime": 236.4251, | |
| "eval_samples_per_second": 17.244, | |
| "eval_steps_per_second": 1.079, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 2.7075701760979362e-05, | |
| "loss": 1.1186, | |
| "step": 510 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 2.7160033436347994e-05, | |
| "loss": 1.0984, | |
| "step": 520 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 2.7242758696007892e-05, | |
| "loss": 1.175, | |
| "step": 530 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 2.732393759822969e-05, | |
| "loss": 1.0451, | |
| "step": 540 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 2.7403626894942436e-05, | |
| "loss": 1.1391, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 2.7481880270062005e-05, | |
| "loss": 1.2073, | |
| "step": 560 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 2.755874855672492e-05, | |
| "loss": 1.091, | |
| "step": 570 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 2.7634279935629375e-05, | |
| "loss": 1.19, | |
| "step": 580 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 2.7708520116421442e-05, | |
| "loss": 1.0771, | |
| "step": 590 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 2.7781512503836438e-05, | |
| "loss": 1.1513, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 2.7853298350107673e-05, | |
| "loss": 1.0998, | |
| "step": 610 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 2.7923916894982543e-05, | |
| "loss": 1.2006, | |
| "step": 620 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 2.799340549453582e-05, | |
| "loss": 1.1671, | |
| "step": 630 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 2.8061799739838875e-05, | |
| "loss": 1.0936, | |
| "step": 640 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 2.812913356642856e-05, | |
| "loss": 1.1308, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 2.819543935541869e-05, | |
| "loss": 1.1227, | |
| "step": 660 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 2.8260748027008266e-05, | |
| "loss": 1.1403, | |
| "step": 670 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 2.8325089127062363e-05, | |
| "loss": 1.1926, | |
| "step": 680 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 2.8388490907372553e-05, | |
| "loss": 1.1912, | |
| "step": 690 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 2.8450980400142567e-05, | |
| "loss": 1.1237, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 2.8512583487190755e-05, | |
| "loss": 1.0667, | |
| "step": 710 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 2.8573324964312687e-05, | |
| "loss": 1.1059, | |
| "step": 720 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 2.863322860120456e-05, | |
| "loss": 1.0878, | |
| "step": 730 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 2.8692317197309762e-05, | |
| "loss": 1.0756, | |
| "step": 740 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 2.8750612633917003e-05, | |
| "loss": 1.082, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 2.8808135922807917e-05, | |
| "loss": 1.0643, | |
| "step": 760 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 2.886490725172482e-05, | |
| "loss": 1.1299, | |
| "step": 770 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 2.8920946026904808e-05, | |
| "loss": 1.1062, | |
| "step": 780 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 2.8976270912904416e-05, | |
| "loss": 1.1176, | |
| "step": 790 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 2.9030899869919436e-05, | |
| "loss": 1.1123, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 2.9084850188786496e-05, | |
| "loss": 1.2128, | |
| "step": 810 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 2.9138138523837167e-05, | |
| "loss": 1.1562, | |
| "step": 820 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 2.919078092376074e-05, | |
| "loss": 1.1277, | |
| "step": 830 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 2.924279286061882e-05, | |
| "loss": 1.0851, | |
| "step": 840 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 2.929418925714293e-05, | |
| "loss": 1.0564, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 2.934498451243568e-05, | |
| "loss": 1.1252, | |
| "step": 860 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 2.939519252618619e-05, | |
| "loss": 1.0855, | |
| "step": 870 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 2.9444826721501686e-05, | |
| "loss": 1.1313, | |
| "step": 880 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 2.9493900066449127e-05, | |
| "loss": 1.1188, | |
| "step": 890 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 2.9542425094393252e-05, | |
| "loss": 1.1184, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 2.9590413923210936e-05, | |
| "loss": 1.1324, | |
| "step": 910 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 2.9637878273455557e-05, | |
| "loss": 1.0961, | |
| "step": 920 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 2.968482948553935e-05, | |
| "loss": 1.1489, | |
| "step": 930 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 2.973127853599699e-05, | |
| "loss": 1.1094, | |
| "step": 940 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 2.977723605288848e-05, | |
| "loss": 1.0903, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 2.9822712330395685e-05, | |
| "loss": 1.0927, | |
| "step": 960 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 2.9867717342662454e-05, | |
| "loss": 1.0697, | |
| "step": 970 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 2.991226075692495e-05, | |
| "loss": 1.1705, | |
| "step": 980 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 2.9956351945975502e-05, | |
| "loss": 1.0569, | |
| "step": 990 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 3e-05, | |
| "loss": 1.08, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "eval_get_denotation_accuracy": 0.4383124846701006, | |
| "eval_loss": 0.9220618009567261, | |
| "eval_runtime": 267.6287, | |
| "eval_samples_per_second": 15.234, | |
| "eval_steps_per_second": 0.953, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 2.9985789473684212e-05, | |
| "loss": 1.0803, | |
| "step": 1010 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 2.997e-05, | |
| "loss": 1.0548, | |
| "step": 1020 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 2.995421052631579e-05, | |
| "loss": 1.0935, | |
| "step": 1030 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 2.9938421052631577e-05, | |
| "loss": 1.1247, | |
| "step": 1040 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 2.992263157894737e-05, | |
| "loss": 1.0353, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 2.990684210526316e-05, | |
| "loss": 1.1308, | |
| "step": 1060 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 2.9891052631578948e-05, | |
| "loss": 1.0504, | |
| "step": 1070 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 2.9875263157894738e-05, | |
| "loss": 1.123, | |
| "step": 1080 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 2.985947368421053e-05, | |
| "loss": 1.1322, | |
| "step": 1090 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 2.984368421052632e-05, | |
| "loss": 0.9887, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 2.9827894736842106e-05, | |
| "loss": 1.1453, | |
| "step": 1110 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 2.9812105263157896e-05, | |
| "loss": 1.0818, | |
| "step": 1120 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 2.9796315789473683e-05, | |
| "loss": 1.0601, | |
| "step": 1130 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 2.9780526315789477e-05, | |
| "loss": 1.1313, | |
| "step": 1140 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 2.9764736842105264e-05, | |
| "loss": 1.1357, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 2.9748947368421054e-05, | |
| "loss": 1.0524, | |
| "step": 1160 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 2.973315789473684e-05, | |
| "loss": 1.0806, | |
| "step": 1170 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 2.971736842105263e-05, | |
| "loss": 1.0605, | |
| "step": 1180 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 2.970157894736842e-05, | |
| "loss": 1.1875, | |
| "step": 1190 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 2.9685789473684212e-05, | |
| "loss": 1.1273, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 2.967e-05, | |
| "loss": 1.1252, | |
| "step": 1210 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 2.965421052631579e-05, | |
| "loss": 1.0654, | |
| "step": 1220 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 2.9638421052631583e-05, | |
| "loss": 1.0421, | |
| "step": 1230 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 2.962263157894737e-05, | |
| "loss": 1.117, | |
| "step": 1240 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 2.960684210526316e-05, | |
| "loss": 1.0849, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 2.9591052631578947e-05, | |
| "loss": 1.1126, | |
| "step": 1260 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 2.9575263157894737e-05, | |
| "loss": 1.039, | |
| "step": 1270 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 2.9559473684210528e-05, | |
| "loss": 1.084, | |
| "step": 1280 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 2.9543684210526318e-05, | |
| "loss": 1.1559, | |
| "step": 1290 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 2.9527894736842105e-05, | |
| "loss": 1.1139, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 2.9512105263157895e-05, | |
| "loss": 0.9822, | |
| "step": 1310 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 2.9496315789473682e-05, | |
| "loss": 1.0209, | |
| "step": 1320 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 2.9480526315789476e-05, | |
| "loss": 1.0462, | |
| "step": 1330 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 2.9464736842105263e-05, | |
| "loss": 1.1401, | |
| "step": 1340 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 2.9448947368421053e-05, | |
| "loss": 1.0874, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 2.943315789473684e-05, | |
| "loss": 1.1206, | |
| "step": 1360 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 2.9417368421052634e-05, | |
| "loss": 1.0679, | |
| "step": 1370 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 2.940157894736842e-05, | |
| "loss": 1.101, | |
| "step": 1380 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 2.938578947368421e-05, | |
| "loss": 0.9896, | |
| "step": 1390 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 2.9370000000000002e-05, | |
| "loss": 1.1539, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 2.935421052631579e-05, | |
| "loss": 1.0817, | |
| "step": 1410 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 2.9338421052631582e-05, | |
| "loss": 1.0821, | |
| "step": 1420 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 2.932263157894737e-05, | |
| "loss": 1.1255, | |
| "step": 1430 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 2.930684210526316e-05, | |
| "loss": 1.1148, | |
| "step": 1440 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 2.9291052631578947e-05, | |
| "loss": 1.0495, | |
| "step": 1450 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 2.9275263157894737e-05, | |
| "loss": 1.1169, | |
| "step": 1460 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 2.9259473684210527e-05, | |
| "loss": 1.1052, | |
| "step": 1470 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 2.9243684210526318e-05, | |
| "loss": 1.1417, | |
| "step": 1480 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 2.9227894736842105e-05, | |
| "loss": 0.9703, | |
| "step": 1490 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 2.9212105263157895e-05, | |
| "loss": 1.0687, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "eval_get_denotation_accuracy": 0.4432180524895757, | |
| "eval_loss": 0.8963027000427246, | |
| "eval_runtime": 270.68, | |
| "eval_samples_per_second": 15.062, | |
| "eval_steps_per_second": 0.942, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 2.9196315789473685e-05, | |
| "loss": 1.0476, | |
| "step": 1510 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 2.9180526315789476e-05, | |
| "loss": 1.1054, | |
| "step": 1520 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 2.9164736842105263e-05, | |
| "loss": 1.0612, | |
| "step": 1530 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 2.9148947368421053e-05, | |
| "loss": 1.0617, | |
| "step": 1540 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 2.9133157894736843e-05, | |
| "loss": 1.1465, | |
| "step": 1550 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 2.9117368421052634e-05, | |
| "loss": 1.0275, | |
| "step": 1560 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 2.9101578947368424e-05, | |
| "loss": 1.1222, | |
| "step": 1570 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 2.908578947368421e-05, | |
| "loss": 1.0845, | |
| "step": 1580 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 2.907e-05, | |
| "loss": 1.0929, | |
| "step": 1590 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 2.9054210526315788e-05, | |
| "loss": 1.1213, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 2.9038421052631582e-05, | |
| "loss": 1.0569, | |
| "step": 1610 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 2.902263157894737e-05, | |
| "loss": 1.0698, | |
| "step": 1620 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 2.900684210526316e-05, | |
| "loss": 1.0309, | |
| "step": 1630 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 2.8991052631578946e-05, | |
| "loss": 1.0961, | |
| "step": 1640 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 2.897526315789474e-05, | |
| "loss": 1.1316, | |
| "step": 1650 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 2.8959473684210527e-05, | |
| "loss": 1.073, | |
| "step": 1660 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 2.8943684210526317e-05, | |
| "loss": 1.077, | |
| "step": 1670 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 2.8927894736842104e-05, | |
| "loss": 1.0776, | |
| "step": 1680 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 2.8912105263157894e-05, | |
| "loss": 1.0831, | |
| "step": 1690 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 2.8896315789473685e-05, | |
| "loss": 1.0339, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 2.8880526315789475e-05, | |
| "loss": 1.0182, | |
| "step": 1710 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 2.8864736842105265e-05, | |
| "loss": 1.1317, | |
| "step": 1720 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 2.8848947368421052e-05, | |
| "loss": 1.0502, | |
| "step": 1730 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 2.8833157894736843e-05, | |
| "loss": 1.0111, | |
| "step": 1740 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 2.8817368421052633e-05, | |
| "loss": 1.0059, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 2.8801578947368423e-05, | |
| "loss": 1.0628, | |
| "step": 1760 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 2.878578947368421e-05, | |
| "loss": 1.0761, | |
| "step": 1770 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 2.877e-05, | |
| "loss": 1.0605, | |
| "step": 1780 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 2.875421052631579e-05, | |
| "loss": 1.0272, | |
| "step": 1790 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 2.873842105263158e-05, | |
| "loss": 1.0903, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 2.8722631578947368e-05, | |
| "loss": 1.1161, | |
| "step": 1810 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 2.870684210526316e-05, | |
| "loss": 1.0112, | |
| "step": 1820 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 2.8691052631578946e-05, | |
| "loss": 1.042, | |
| "step": 1830 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 2.867526315789474e-05, | |
| "loss": 1.0366, | |
| "step": 1840 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 2.8659473684210526e-05, | |
| "loss": 1.0757, | |
| "step": 1850 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 2.8643684210526317e-05, | |
| "loss": 1.0347, | |
| "step": 1860 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 2.8627894736842104e-05, | |
| "loss": 1.0715, | |
| "step": 1870 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 2.8612105263157894e-05, | |
| "loss": 0.9993, | |
| "step": 1880 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 2.8596315789473688e-05, | |
| "loss": 1.0156, | |
| "step": 1890 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 2.8580526315789475e-05, | |
| "loss": 1.0539, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 2.8564736842105265e-05, | |
| "loss": 1.0708, | |
| "step": 1910 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 2.8548947368421052e-05, | |
| "loss": 1.1075, | |
| "step": 1920 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 2.8533157894736846e-05, | |
| "loss": 1.1084, | |
| "step": 1930 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 2.8517368421052633e-05, | |
| "loss": 1.0305, | |
| "step": 1940 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 2.8501578947368423e-05, | |
| "loss": 1.0903, | |
| "step": 1950 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 2.848578947368421e-05, | |
| "loss": 1.0329, | |
| "step": 1960 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 2.847e-05, | |
| "loss": 1.092, | |
| "step": 1970 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 2.845421052631579e-05, | |
| "loss": 1.0535, | |
| "step": 1980 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 2.843842105263158e-05, | |
| "loss": 0.9838, | |
| "step": 1990 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 2.8422631578947368e-05, | |
| "loss": 0.9118, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "eval_get_denotation_accuracy": 0.4255580083394653, | |
| "eval_loss": 0.943108856678009, | |
| "eval_runtime": 233.6145, | |
| "eval_samples_per_second": 17.452, | |
| "eval_steps_per_second": 1.092, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 2.8406842105263158e-05, | |
| "loss": 1.0671, | |
| "step": 2010 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 2.8391052631578945e-05, | |
| "loss": 1.0006, | |
| "step": 2020 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 2.837526315789474e-05, | |
| "loss": 1.0514, | |
| "step": 2030 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 2.8359473684210526e-05, | |
| "loss": 1.0597, | |
| "step": 2040 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 2.8343684210526316e-05, | |
| "loss": 1.0267, | |
| "step": 2050 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 2.8327894736842106e-05, | |
| "loss": 0.9874, | |
| "step": 2060 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 2.8312105263157897e-05, | |
| "loss": 1.0139, | |
| "step": 2070 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 2.8296315789473687e-05, | |
| "loss": 0.9834, | |
| "step": 2080 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 2.8280526315789474e-05, | |
| "loss": 1.1119, | |
| "step": 2090 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 2.8264736842105264e-05, | |
| "loss": 1.0442, | |
| "step": 2100 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 2.824894736842105e-05, | |
| "loss": 1.007, | |
| "step": 2110 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 2.8233157894736845e-05, | |
| "loss": 1.0275, | |
| "step": 2120 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 2.8217368421052632e-05, | |
| "loss": 1.0664, | |
| "step": 2130 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 2.8201578947368422e-05, | |
| "loss": 1.0957, | |
| "step": 2140 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 2.818578947368421e-05, | |
| "loss": 1.034, | |
| "step": 2150 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 2.817e-05, | |
| "loss": 1.0778, | |
| "step": 2160 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 2.815421052631579e-05, | |
| "loss": 1.0958, | |
| "step": 2170 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 2.813842105263158e-05, | |
| "loss": 1.0358, | |
| "step": 2180 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 2.8122631578947367e-05, | |
| "loss": 1.0436, | |
| "step": 2190 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 2.8106842105263158e-05, | |
| "loss": 1.0577, | |
| "step": 2200 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 2.8091052631578948e-05, | |
| "loss": 1.0337, | |
| "step": 2210 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 2.8075263157894738e-05, | |
| "loss": 1.0033, | |
| "step": 2220 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 2.805947368421053e-05, | |
| "loss": 1.1064, | |
| "step": 2230 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 2.8043684210526316e-05, | |
| "loss": 0.9971, | |
| "step": 2240 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 2.8027894736842106e-05, | |
| "loss": 1.0335, | |
| "step": 2250 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 2.8012105263157896e-05, | |
| "loss": 0.9978, | |
| "step": 2260 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 2.7996315789473687e-05, | |
| "loss": 0.9603, | |
| "step": 2270 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 2.7980526315789474e-05, | |
| "loss": 1.03, | |
| "step": 2280 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 2.7964736842105264e-05, | |
| "loss": 1.0269, | |
| "step": 2290 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 2.794894736842105e-05, | |
| "loss": 1.1107, | |
| "step": 2300 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 2.7933157894736845e-05, | |
| "loss": 1.0932, | |
| "step": 2310 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 2.791736842105263e-05, | |
| "loss": 1.0937, | |
| "step": 2320 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 2.7901578947368422e-05, | |
| "loss": 1.0952, | |
| "step": 2330 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 2.788578947368421e-05, | |
| "loss": 1.0542, | |
| "step": 2340 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 2.7870000000000003e-05, | |
| "loss": 1.0937, | |
| "step": 2350 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 2.785421052631579e-05, | |
| "loss": 1.1312, | |
| "step": 2360 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 2.783842105263158e-05, | |
| "loss": 1.0614, | |
| "step": 2370 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 2.7822631578947367e-05, | |
| "loss": 1.0843, | |
| "step": 2380 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 2.7806842105263157e-05, | |
| "loss": 1.0934, | |
| "step": 2390 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 2.779105263157895e-05, | |
| "loss": 1.0319, | |
| "step": 2400 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 2.7775263157894738e-05, | |
| "loss": 1.0297, | |
| "step": 2410 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 2.7759473684210528e-05, | |
| "loss": 0.9999, | |
| "step": 2420 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 2.7743684210526315e-05, | |
| "loss": 1.0612, | |
| "step": 2430 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 2.772789473684211e-05, | |
| "loss": 1.022, | |
| "step": 2440 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 2.7712105263157896e-05, | |
| "loss": 1.0553, | |
| "step": 2450 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 2.7696315789473686e-05, | |
| "loss": 1.0087, | |
| "step": 2460 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 2.7680526315789473e-05, | |
| "loss": 1.0865, | |
| "step": 2470 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 2.7664736842105263e-05, | |
| "loss": 1.0538, | |
| "step": 2480 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 2.7648947368421054e-05, | |
| "loss": 1.0232, | |
| "step": 2490 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 2.7633157894736844e-05, | |
| "loss": 1.0017, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "eval_get_denotation_accuracy": 0.45719892077507973, | |
| "eval_loss": 0.9026257395744324, | |
| "eval_runtime": 232.0877, | |
| "eval_samples_per_second": 17.567, | |
| "eval_steps_per_second": 1.099, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 2.761736842105263e-05, | |
| "loss": 1.0373, | |
| "step": 2510 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 2.760157894736842e-05, | |
| "loss": 1.0501, | |
| "step": 2520 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 2.7585789473684208e-05, | |
| "loss": 1.0102, | |
| "step": 2530 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 2.7570000000000002e-05, | |
| "loss": 1.0003, | |
| "step": 2540 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 2.755421052631579e-05, | |
| "loss": 1.0132, | |
| "step": 2550 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 2.753842105263158e-05, | |
| "loss": 1.1177, | |
| "step": 2560 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 2.752263157894737e-05, | |
| "loss": 0.9424, | |
| "step": 2570 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 2.750684210526316e-05, | |
| "loss": 1.0153, | |
| "step": 2580 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 2.749105263157895e-05, | |
| "loss": 1.0635, | |
| "step": 2590 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 2.7475263157894737e-05, | |
| "loss": 0.9576, | |
| "step": 2600 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 2.7459473684210528e-05, | |
| "loss": 1.0302, | |
| "step": 2610 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 2.7443684210526315e-05, | |
| "loss": 1.0201, | |
| "step": 2620 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 2.7427894736842108e-05, | |
| "loss": 1.0524, | |
| "step": 2630 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 2.7412105263157895e-05, | |
| "loss": 1.0559, | |
| "step": 2640 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 2.7396315789473686e-05, | |
| "loss": 1.018, | |
| "step": 2650 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 2.7380526315789473e-05, | |
| "loss": 1.0135, | |
| "step": 2660 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 2.7364736842105263e-05, | |
| "loss": 1.0447, | |
| "step": 2670 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 2.7348947368421053e-05, | |
| "loss": 1.0601, | |
| "step": 2680 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 2.7333157894736844e-05, | |
| "loss": 1.0182, | |
| "step": 2690 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 2.731736842105263e-05, | |
| "loss": 1.0494, | |
| "step": 2700 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 2.730157894736842e-05, | |
| "loss": 1.0011, | |
| "step": 2710 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 2.728578947368421e-05, | |
| "loss": 1.0838, | |
| "step": 2720 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 2.727e-05, | |
| "loss": 0.9707, | |
| "step": 2730 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 2.7254210526315792e-05, | |
| "loss": 0.9892, | |
| "step": 2740 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 2.723842105263158e-05, | |
| "loss": 1.0623, | |
| "step": 2750 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 2.722263157894737e-05, | |
| "loss": 1.0185, | |
| "step": 2760 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 2.720684210526316e-05, | |
| "loss": 1.0887, | |
| "step": 2770 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 2.719105263157895e-05, | |
| "loss": 1.0204, | |
| "step": 2780 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 2.7175263157894737e-05, | |
| "loss": 1.015, | |
| "step": 2790 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 2.7159473684210527e-05, | |
| "loss": 1.0167, | |
| "step": 2800 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 2.7143684210526314e-05, | |
| "loss": 1.0941, | |
| "step": 2810 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 2.7127894736842108e-05, | |
| "loss": 1.1229, | |
| "step": 2820 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 2.7112105263157895e-05, | |
| "loss": 1.0106, | |
| "step": 2830 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 2.7096315789473685e-05, | |
| "loss": 1.1508, | |
| "step": 2840 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 2.7080526315789472e-05, | |
| "loss": 1.0192, | |
| "step": 2850 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 2.7064736842105266e-05, | |
| "loss": 1.0381, | |
| "step": 2860 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 2.7048947368421053e-05, | |
| "loss": 0.9954, | |
| "step": 2870 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 2.7033157894736843e-05, | |
| "loss": 1.0687, | |
| "step": 2880 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 2.7017368421052633e-05, | |
| "loss": 1.0403, | |
| "step": 2890 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 2.700157894736842e-05, | |
| "loss": 1.0272, | |
| "step": 2900 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 2.6985789473684214e-05, | |
| "loss": 1.0096, | |
| "step": 2910 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 2.697e-05, | |
| "loss": 0.9641, | |
| "step": 2920 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 2.695421052631579e-05, | |
| "loss": 0.975, | |
| "step": 2930 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 2.6938421052631578e-05, | |
| "loss": 0.9384, | |
| "step": 2940 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 2.692263157894737e-05, | |
| "loss": 0.9905, | |
| "step": 2950 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 2.690684210526316e-05, | |
| "loss": 0.9804, | |
| "step": 2960 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 2.689105263157895e-05, | |
| "loss": 1.0942, | |
| "step": 2970 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 2.6875263157894736e-05, | |
| "loss": 1.0254, | |
| "step": 2980 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 2.6859473684210527e-05, | |
| "loss": 1.0575, | |
| "step": 2990 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 2.6843684210526317e-05, | |
| "loss": 1.0196, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "eval_get_denotation_accuracy": 0.457934755948001, | |
| "eval_loss": 0.876184344291687, | |
| "eval_runtime": 244.0361, | |
| "eval_samples_per_second": 16.707, | |
| "eval_steps_per_second": 1.045, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 2.6827894736842107e-05, | |
| "loss": 1.0563, | |
| "step": 3010 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 2.6812105263157894e-05, | |
| "loss": 1.0039, | |
| "step": 3020 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 2.6796315789473685e-05, | |
| "loss": 0.9485, | |
| "step": 3030 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 2.678052631578947e-05, | |
| "loss": 1.0705, | |
| "step": 3040 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 2.6764736842105265e-05, | |
| "loss": 1.0055, | |
| "step": 3050 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 2.6748947368421056e-05, | |
| "loss": 1.0035, | |
| "step": 3060 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 2.6733157894736842e-05, | |
| "loss": 1.0492, | |
| "step": 3070 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 2.6717368421052633e-05, | |
| "loss": 1.1027, | |
| "step": 3080 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 2.670157894736842e-05, | |
| "loss": 1.0883, | |
| "step": 3090 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 2.6685789473684214e-05, | |
| "loss": 1.0351, | |
| "step": 3100 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 2.667e-05, | |
| "loss": 1.0188, | |
| "step": 3110 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 2.665421052631579e-05, | |
| "loss": 0.9927, | |
| "step": 3120 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 2.6638421052631578e-05, | |
| "loss": 0.931, | |
| "step": 3130 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 2.662263157894737e-05, | |
| "loss": 1.0144, | |
| "step": 3140 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 2.660684210526316e-05, | |
| "loss": 1.0399, | |
| "step": 3150 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 2.659105263157895e-05, | |
| "loss": 1.0065, | |
| "step": 3160 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 2.6575263157894736e-05, | |
| "loss": 0.9821, | |
| "step": 3170 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 2.6559473684210526e-05, | |
| "loss": 1.0147, | |
| "step": 3180 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 2.6543684210526316e-05, | |
| "loss": 1.0128, | |
| "step": 3190 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 2.6527894736842107e-05, | |
| "loss": 1.1112, | |
| "step": 3200 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 2.6512105263157894e-05, | |
| "loss": 1.005, | |
| "step": 3210 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 2.6496315789473684e-05, | |
| "loss": 1.0822, | |
| "step": 3220 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 2.6480526315789474e-05, | |
| "loss": 1.0297, | |
| "step": 3230 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 2.6464736842105265e-05, | |
| "loss": 1.0213, | |
| "step": 3240 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 2.6448947368421055e-05, | |
| "loss": 1.0078, | |
| "step": 3250 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 2.6433157894736842e-05, | |
| "loss": 1.0381, | |
| "step": 3260 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 2.6417368421052632e-05, | |
| "loss": 0.9792, | |
| "step": 3270 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 2.6401578947368423e-05, | |
| "loss": 0.9916, | |
| "step": 3280 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 2.6385789473684213e-05, | |
| "loss": 1.1386, | |
| "step": 3290 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 2.637e-05, | |
| "loss": 0.9853, | |
| "step": 3300 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 2.635421052631579e-05, | |
| "loss": 0.9641, | |
| "step": 3310 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 2.6338421052631577e-05, | |
| "loss": 1.0471, | |
| "step": 3320 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 2.632263157894737e-05, | |
| "loss": 1.0737, | |
| "step": 3330 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 2.6306842105263158e-05, | |
| "loss": 1.0703, | |
| "step": 3340 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 2.6291052631578948e-05, | |
| "loss": 0.9755, | |
| "step": 3350 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 2.6275263157894735e-05, | |
| "loss": 0.9964, | |
| "step": 3360 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 2.6259473684210526e-05, | |
| "loss": 1.0181, | |
| "step": 3370 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 2.6243684210526316e-05, | |
| "loss": 1.1007, | |
| "step": 3380 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 2.6227894736842106e-05, | |
| "loss": 1.0173, | |
| "step": 3390 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 2.6212105263157897e-05, | |
| "loss": 0.9296, | |
| "step": 3400 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 2.6196315789473683e-05, | |
| "loss": 0.9584, | |
| "step": 3410 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 2.6180526315789477e-05, | |
| "loss": 1.0229, | |
| "step": 3420 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 2.6164736842105264e-05, | |
| "loss": 1.0355, | |
| "step": 3430 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 2.6148947368421055e-05, | |
| "loss": 1.0586, | |
| "step": 3440 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 2.613315789473684e-05, | |
| "loss": 1.0587, | |
| "step": 3450 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 2.6117368421052632e-05, | |
| "loss": 0.9826, | |
| "step": 3460 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 2.6101578947368422e-05, | |
| "loss": 0.9878, | |
| "step": 3470 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 2.6085789473684212e-05, | |
| "loss": 1.0297, | |
| "step": 3480 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 2.607e-05, | |
| "loss": 0.9465, | |
| "step": 3490 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 2.605421052631579e-05, | |
| "loss": 1.0392, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "eval_get_denotation_accuracy": 0.4650478292862399, | |
| "eval_loss": 0.897882878780365, | |
| "eval_runtime": 223.6292, | |
| "eval_samples_per_second": 18.231, | |
| "eval_steps_per_second": 1.14, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 2.6038421052631577e-05, | |
| "loss": 1.0521, | |
| "step": 3510 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 2.602263157894737e-05, | |
| "loss": 1.0067, | |
| "step": 3520 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 2.6006842105263157e-05, | |
| "loss": 1.071, | |
| "step": 3530 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 2.5991052631578948e-05, | |
| "loss": 1.0192, | |
| "step": 3540 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 2.5975263157894738e-05, | |
| "loss": 1.0119, | |
| "step": 3550 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 2.595947368421053e-05, | |
| "loss": 1.0436, | |
| "step": 3560 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 2.594368421052632e-05, | |
| "loss": 0.9783, | |
| "step": 3570 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 2.5927894736842106e-05, | |
| "loss": 0.927, | |
| "step": 3580 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 2.5912105263157896e-05, | |
| "loss": 0.9704, | |
| "step": 3590 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 2.5896315789473683e-05, | |
| "loss": 1.0329, | |
| "step": 3600 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 2.5880526315789477e-05, | |
| "loss": 0.9621, | |
| "step": 3610 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 2.5864736842105264e-05, | |
| "loss": 0.9888, | |
| "step": 3620 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 2.5848947368421054e-05, | |
| "loss": 1.0215, | |
| "step": 3630 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 2.583315789473684e-05, | |
| "loss": 1.0729, | |
| "step": 3640 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 2.581736842105263e-05, | |
| "loss": 1.0184, | |
| "step": 3650 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 2.580157894736842e-05, | |
| "loss": 0.9363, | |
| "step": 3660 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 2.5785789473684212e-05, | |
| "loss": 1.0606, | |
| "step": 3670 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 2.577e-05, | |
| "loss": 0.9232, | |
| "step": 3680 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 2.575421052631579e-05, | |
| "loss": 1.0499, | |
| "step": 3690 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 2.573842105263158e-05, | |
| "loss": 1.0413, | |
| "step": 3700 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 2.572263157894737e-05, | |
| "loss": 0.9717, | |
| "step": 3710 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 2.570684210526316e-05, | |
| "loss": 1.0151, | |
| "step": 3720 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 2.5691052631578947e-05, | |
| "loss": 1.0533, | |
| "step": 3730 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 2.5675263157894738e-05, | |
| "loss": 1.022, | |
| "step": 3740 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 2.5659473684210528e-05, | |
| "loss": 1.0458, | |
| "step": 3750 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 2.5643684210526318e-05, | |
| "loss": 1.0996, | |
| "step": 3760 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 2.5627894736842105e-05, | |
| "loss": 0.971, | |
| "step": 3770 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 2.5612105263157896e-05, | |
| "loss": 1.0158, | |
| "step": 3780 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 2.5596315789473682e-05, | |
| "loss": 0.9443, | |
| "step": 3790 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 2.5580526315789476e-05, | |
| "loss": 0.9103, | |
| "step": 3800 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 2.5564736842105263e-05, | |
| "loss": 1.0915, | |
| "step": 3810 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 2.5548947368421053e-05, | |
| "loss": 1.043, | |
| "step": 3820 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 2.553315789473684e-05, | |
| "loss": 0.9867, | |
| "step": 3830 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 2.5517368421052634e-05, | |
| "loss": 1.0856, | |
| "step": 3840 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 2.550157894736842e-05, | |
| "loss": 1.0218, | |
| "step": 3850 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 2.548578947368421e-05, | |
| "loss": 0.9604, | |
| "step": 3860 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 2.547e-05, | |
| "loss": 1.0484, | |
| "step": 3870 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 2.545421052631579e-05, | |
| "loss": 0.9075, | |
| "step": 3880 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 2.5438421052631582e-05, | |
| "loss": 1.0553, | |
| "step": 3890 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 2.542263157894737e-05, | |
| "loss": 1.0331, | |
| "step": 3900 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 2.540684210526316e-05, | |
| "loss": 0.9569, | |
| "step": 3910 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 2.5391052631578947e-05, | |
| "loss": 1.0463, | |
| "step": 3920 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 2.5375263157894737e-05, | |
| "loss": 1.0559, | |
| "step": 3930 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 2.5359473684210527e-05, | |
| "loss": 1.0503, | |
| "step": 3940 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 2.5343684210526318e-05, | |
| "loss": 0.9853, | |
| "step": 3950 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 2.5327894736842105e-05, | |
| "loss": 1.0049, | |
| "step": 3960 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 2.5312105263157895e-05, | |
| "loss": 1.0293, | |
| "step": 3970 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 2.5296315789473685e-05, | |
| "loss": 1.0387, | |
| "step": 3980 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 2.5280526315789476e-05, | |
| "loss": 1.0535, | |
| "step": 3990 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 2.5264736842105263e-05, | |
| "loss": 0.9677, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "eval_get_denotation_accuracy": 0.4665194996320824, | |
| "eval_loss": 0.889651358127594, | |
| "eval_runtime": 260.5569, | |
| "eval_samples_per_second": 15.647, | |
| "eval_steps_per_second": 0.979, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 2.5248947368421053e-05, | |
| "loss": 1.0439, | |
| "step": 4010 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 2.523315789473684e-05, | |
| "loss": 0.9977, | |
| "step": 4020 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 2.5217368421052634e-05, | |
| "loss": 1.0159, | |
| "step": 4030 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 2.520157894736842e-05, | |
| "loss": 1.0015, | |
| "step": 4040 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 2.518578947368421e-05, | |
| "loss": 1.0032, | |
| "step": 4050 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 2.517e-05, | |
| "loss": 1.0141, | |
| "step": 4060 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 2.515421052631579e-05, | |
| "loss": 1.0196, | |
| "step": 4070 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 2.5138421052631582e-05, | |
| "loss": 1.0079, | |
| "step": 4080 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 2.512263157894737e-05, | |
| "loss": 1.101, | |
| "step": 4090 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 2.510684210526316e-05, | |
| "loss": 1.0585, | |
| "step": 4100 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 2.5091052631578946e-05, | |
| "loss": 1.0242, | |
| "step": 4110 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 2.507526315789474e-05, | |
| "loss": 0.9472, | |
| "step": 4120 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 2.5059473684210527e-05, | |
| "loss": 1.0035, | |
| "step": 4130 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 2.5043684210526317e-05, | |
| "loss": 1.0103, | |
| "step": 4140 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 2.5027894736842104e-05, | |
| "loss": 1.0457, | |
| "step": 4150 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 2.5012105263157894e-05, | |
| "loss": 1.0429, | |
| "step": 4160 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 2.4996315789473685e-05, | |
| "loss": 1.0362, | |
| "step": 4170 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 2.4980526315789475e-05, | |
| "loss": 1.0056, | |
| "step": 4180 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 2.4964736842105262e-05, | |
| "loss": 1.014, | |
| "step": 4190 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 2.4948947368421052e-05, | |
| "loss": 0.9206, | |
| "step": 4200 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 2.4933157894736843e-05, | |
| "loss": 1.0915, | |
| "step": 4210 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 2.4917368421052633e-05, | |
| "loss": 0.9606, | |
| "step": 4220 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 2.4901578947368423e-05, | |
| "loss": 1.0557, | |
| "step": 4230 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 2.488578947368421e-05, | |
| "loss": 0.9718, | |
| "step": 4240 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 2.487e-05, | |
| "loss": 0.967, | |
| "step": 4250 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 2.485421052631579e-05, | |
| "loss": 1.0016, | |
| "step": 4260 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 2.483842105263158e-05, | |
| "loss": 1.0913, | |
| "step": 4270 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 2.482263157894737e-05, | |
| "loss": 0.9326, | |
| "step": 4280 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 2.480684210526316e-05, | |
| "loss": 1.0497, | |
| "step": 4290 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 2.4791052631578946e-05, | |
| "loss": 0.9351, | |
| "step": 4300 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 2.477526315789474e-05, | |
| "loss": 1.024, | |
| "step": 4310 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 2.4759473684210526e-05, | |
| "loss": 0.9998, | |
| "step": 4320 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 2.4743684210526317e-05, | |
| "loss": 1.0417, | |
| "step": 4330 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 2.4727894736842104e-05, | |
| "loss": 0.9122, | |
| "step": 4340 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 2.4712105263157897e-05, | |
| "loss": 1.1297, | |
| "step": 4350 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 2.4696315789473684e-05, | |
| "loss": 1.0099, | |
| "step": 4360 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 2.4680526315789475e-05, | |
| "loss": 1.1269, | |
| "step": 4370 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 2.466473684210526e-05, | |
| "loss": 1.0236, | |
| "step": 4380 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 2.4648947368421052e-05, | |
| "loss": 1.0352, | |
| "step": 4390 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 2.4633157894736846e-05, | |
| "loss": 0.9973, | |
| "step": 4400 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 2.4617368421052633e-05, | |
| "loss": 1.0016, | |
| "step": 4410 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 2.4601578947368423e-05, | |
| "loss": 0.9684, | |
| "step": 4420 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 2.458578947368421e-05, | |
| "loss": 1.0226, | |
| "step": 4430 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 2.457e-05, | |
| "loss": 0.9922, | |
| "step": 4440 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 2.455421052631579e-05, | |
| "loss": 1.0112, | |
| "step": 4450 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 2.453842105263158e-05, | |
| "loss": 0.977, | |
| "step": 4460 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 2.4522631578947368e-05, | |
| "loss": 0.9721, | |
| "step": 4470 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 2.4506842105263158e-05, | |
| "loss": 1.0738, | |
| "step": 4480 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 2.449105263157895e-05, | |
| "loss": 0.9887, | |
| "step": 4490 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 2.447526315789474e-05, | |
| "loss": 1.055, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "eval_get_denotation_accuracy": 0.47633063527103264, | |
| "eval_loss": 0.8773789405822754, | |
| "eval_runtime": 254.7458, | |
| "eval_samples_per_second": 16.004, | |
| "eval_steps_per_second": 1.001, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 2.4459473684210526e-05, | |
| "loss": 0.953, | |
| "step": 4510 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 2.4443684210526316e-05, | |
| "loss": 1.0545, | |
| "step": 4520 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 2.4427894736842103e-05, | |
| "loss": 1.0534, | |
| "step": 4530 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 2.4412105263157897e-05, | |
| "loss": 1.0373, | |
| "step": 4540 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 2.4396315789473684e-05, | |
| "loss": 1.002, | |
| "step": 4550 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 2.4380526315789474e-05, | |
| "loss": 1.0608, | |
| "step": 4560 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 2.4364736842105264e-05, | |
| "loss": 1.0014, | |
| "step": 4570 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 2.434894736842105e-05, | |
| "loss": 1.032, | |
| "step": 4580 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 2.4333157894736845e-05, | |
| "loss": 1.0434, | |
| "step": 4590 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 2.4317368421052632e-05, | |
| "loss": 1.057, | |
| "step": 4600 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 2.4301578947368422e-05, | |
| "loss": 0.9441, | |
| "step": 4610 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 2.428578947368421e-05, | |
| "loss": 1.0191, | |
| "step": 4620 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 2.4270000000000003e-05, | |
| "loss": 1.1007, | |
| "step": 4630 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 2.425421052631579e-05, | |
| "loss": 1.0085, | |
| "step": 4640 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 2.423842105263158e-05, | |
| "loss": 1.0426, | |
| "step": 4650 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 2.4222631578947367e-05, | |
| "loss": 1.0341, | |
| "step": 4660 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 2.4206842105263158e-05, | |
| "loss": 1.0701, | |
| "step": 4670 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 2.4191052631578948e-05, | |
| "loss": 1.1301, | |
| "step": 4680 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 2.417526315789474e-05, | |
| "loss": 1.0237, | |
| "step": 4690 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 2.4159473684210525e-05, | |
| "loss": 0.9653, | |
| "step": 4700 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 2.4143684210526316e-05, | |
| "loss": 0.9588, | |
| "step": 4710 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 2.4127894736842106e-05, | |
| "loss": 0.9547, | |
| "step": 4720 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 2.4112105263157896e-05, | |
| "loss": 0.9591, | |
| "step": 4730 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 2.4096315789473687e-05, | |
| "loss": 1.0202, | |
| "step": 4740 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 2.4080526315789474e-05, | |
| "loss": 1.0633, | |
| "step": 4750 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 2.4064736842105264e-05, | |
| "loss": 0.9712, | |
| "step": 4760 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 2.4048947368421054e-05, | |
| "loss": 0.9645, | |
| "step": 4770 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 2.4033157894736845e-05, | |
| "loss": 0.9802, | |
| "step": 4780 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 2.401736842105263e-05, | |
| "loss": 1.0162, | |
| "step": 4790 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 2.4001578947368422e-05, | |
| "loss": 1.0307, | |
| "step": 4800 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 2.398578947368421e-05, | |
| "loss": 1.0398, | |
| "step": 4810 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 2.3970000000000003e-05, | |
| "loss": 1.0292, | |
| "step": 4820 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 2.395421052631579e-05, | |
| "loss": 0.9821, | |
| "step": 4830 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 2.393842105263158e-05, | |
| "loss": 1.0242, | |
| "step": 4840 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 2.3922631578947367e-05, | |
| "loss": 0.9971, | |
| "step": 4850 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 2.3906842105263157e-05, | |
| "loss": 0.9161, | |
| "step": 4860 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 2.3891052631578948e-05, | |
| "loss": 0.9727, | |
| "step": 4870 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 2.3875263157894738e-05, | |
| "loss": 0.98, | |
| "step": 4880 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 2.3859473684210528e-05, | |
| "loss": 1.0035, | |
| "step": 4890 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 2.3843684210526315e-05, | |
| "loss": 0.9578, | |
| "step": 4900 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 2.382789473684211e-05, | |
| "loss": 1.0698, | |
| "step": 4910 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 2.3812105263157896e-05, | |
| "loss": 0.9207, | |
| "step": 4920 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 2.3796315789473686e-05, | |
| "loss": 0.9419, | |
| "step": 4930 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 2.3780526315789473e-05, | |
| "loss": 1.0114, | |
| "step": 4940 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 2.3764736842105263e-05, | |
| "loss": 0.9258, | |
| "step": 4950 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 2.3748947368421054e-05, | |
| "loss": 0.9218, | |
| "step": 4960 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 2.3733157894736844e-05, | |
| "loss": 1.0667, | |
| "step": 4970 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 2.371736842105263e-05, | |
| "loss": 1.057, | |
| "step": 4980 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 2.370157894736842e-05, | |
| "loss": 0.9798, | |
| "step": 4990 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 2.368578947368421e-05, | |
| "loss": 1.0044, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "eval_get_denotation_accuracy": 0.46431199411331864, | |
| "eval_loss": 0.9005911350250244, | |
| "eval_runtime": 253.4522, | |
| "eval_samples_per_second": 16.086, | |
| "eval_steps_per_second": 1.006, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 2.3670000000000002e-05, | |
| "loss": 0.984, | |
| "step": 5010 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 2.365421052631579e-05, | |
| "loss": 0.9185, | |
| "step": 5020 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 2.363842105263158e-05, | |
| "loss": 1.0296, | |
| "step": 5030 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 2.3622631578947366e-05, | |
| "loss": 1.0351, | |
| "step": 5040 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 2.360684210526316e-05, | |
| "loss": 0.9271, | |
| "step": 5050 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 2.359105263157895e-05, | |
| "loss": 1.0033, | |
| "step": 5060 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 2.3575263157894737e-05, | |
| "loss": 0.9761, | |
| "step": 5070 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 2.3559473684210528e-05, | |
| "loss": 0.9967, | |
| "step": 5080 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 2.3543684210526315e-05, | |
| "loss": 1.0026, | |
| "step": 5090 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 2.352789473684211e-05, | |
| "loss": 1.0534, | |
| "step": 5100 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 2.3512105263157895e-05, | |
| "loss": 1.0229, | |
| "step": 5110 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 2.3496315789473686e-05, | |
| "loss": 0.9959, | |
| "step": 5120 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 2.3480526315789473e-05, | |
| "loss": 1.0232, | |
| "step": 5130 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 2.3464736842105263e-05, | |
| "loss": 0.9669, | |
| "step": 5140 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 2.3448947368421053e-05, | |
| "loss": 1.0366, | |
| "step": 5150 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 2.3433157894736844e-05, | |
| "loss": 0.991, | |
| "step": 5160 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 2.341736842105263e-05, | |
| "loss": 0.9509, | |
| "step": 5170 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 2.340157894736842e-05, | |
| "loss": 1.1378, | |
| "step": 5180 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 2.338578947368421e-05, | |
| "loss": 1.0201, | |
| "step": 5190 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 2.337e-05, | |
| "loss": 0.9643, | |
| "step": 5200 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 2.335421052631579e-05, | |
| "loss": 0.9404, | |
| "step": 5210 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 2.333842105263158e-05, | |
| "loss": 0.9785, | |
| "step": 5220 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 2.332263157894737e-05, | |
| "loss": 0.9684, | |
| "step": 5230 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 2.330684210526316e-05, | |
| "loss": 1.0068, | |
| "step": 5240 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 2.329105263157895e-05, | |
| "loss": 1.043, | |
| "step": 5250 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 2.3275263157894737e-05, | |
| "loss": 0.9789, | |
| "step": 5260 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 2.3259473684210527e-05, | |
| "loss": 0.9395, | |
| "step": 5270 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 2.3243684210526314e-05, | |
| "loss": 0.9646, | |
| "step": 5280 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 2.3227894736842108e-05, | |
| "loss": 1.0677, | |
| "step": 5290 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 2.3212105263157895e-05, | |
| "loss": 1.0751, | |
| "step": 5300 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 2.3196315789473685e-05, | |
| "loss": 0.9828, | |
| "step": 5310 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 2.3180526315789472e-05, | |
| "loss": 1.007, | |
| "step": 5320 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 2.3164736842105266e-05, | |
| "loss": 1.0219, | |
| "step": 5330 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 2.3148947368421053e-05, | |
| "loss": 1.0095, | |
| "step": 5340 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 2.3133157894736843e-05, | |
| "loss": 1.0138, | |
| "step": 5350 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 2.311736842105263e-05, | |
| "loss": 0.9335, | |
| "step": 5360 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 2.310157894736842e-05, | |
| "loss": 0.9974, | |
| "step": 5370 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 2.308578947368421e-05, | |
| "loss": 0.9958, | |
| "step": 5380 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 2.307e-05, | |
| "loss": 1.0, | |
| "step": 5390 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 2.305421052631579e-05, | |
| "loss": 0.9487, | |
| "step": 5400 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 2.303842105263158e-05, | |
| "loss": 0.9297, | |
| "step": 5410 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 2.302263157894737e-05, | |
| "loss": 0.9855, | |
| "step": 5420 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 2.300684210526316e-05, | |
| "loss": 0.9853, | |
| "step": 5430 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 2.299105263157895e-05, | |
| "loss": 1.0419, | |
| "step": 5440 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 2.2975263157894736e-05, | |
| "loss": 1.0321, | |
| "step": 5450 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 2.2959473684210527e-05, | |
| "loss": 1.0031, | |
| "step": 5460 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 2.2943684210526317e-05, | |
| "loss": 0.9801, | |
| "step": 5470 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 2.2927894736842107e-05, | |
| "loss": 0.9084, | |
| "step": 5480 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 2.2912105263157894e-05, | |
| "loss": 0.9604, | |
| "step": 5490 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 2.2896315789473685e-05, | |
| "loss": 0.9234, | |
| "step": 5500 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "eval_get_denotation_accuracy": 0.46701005641402993, | |
| "eval_loss": 0.9108076095581055, | |
| "eval_runtime": 268.8446, | |
| "eval_samples_per_second": 15.165, | |
| "eval_steps_per_second": 0.949, | |
| "step": 5500 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 2.288052631578947e-05, | |
| "loss": 0.9886, | |
| "step": 5510 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 2.2864736842105265e-05, | |
| "loss": 0.9723, | |
| "step": 5520 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 2.2848947368421052e-05, | |
| "loss": 1.0236, | |
| "step": 5530 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 2.2833157894736843e-05, | |
| "loss": 1.0294, | |
| "step": 5540 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 2.2817368421052633e-05, | |
| "loss": 0.9711, | |
| "step": 5550 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 2.280157894736842e-05, | |
| "loss": 0.9831, | |
| "step": 5560 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 2.2785789473684214e-05, | |
| "loss": 1.0243, | |
| "step": 5570 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 2.277e-05, | |
| "loss": 0.9906, | |
| "step": 5580 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 2.275421052631579e-05, | |
| "loss": 1.007, | |
| "step": 5590 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 2.2738421052631578e-05, | |
| "loss": 0.9364, | |
| "step": 5600 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 2.272263157894737e-05, | |
| "loss": 1.0508, | |
| "step": 5610 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 2.270684210526316e-05, | |
| "loss": 1.0433, | |
| "step": 5620 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 2.269105263157895e-05, | |
| "loss": 1.0452, | |
| "step": 5630 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 2.2675263157894736e-05, | |
| "loss": 1.0136, | |
| "step": 5640 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 2.2659473684210526e-05, | |
| "loss": 1.0191, | |
| "step": 5650 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 2.2643684210526316e-05, | |
| "loss": 0.9874, | |
| "step": 5660 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 2.2627894736842107e-05, | |
| "loss": 1.044, | |
| "step": 5670 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 2.2612105263157894e-05, | |
| "loss": 1.0112, | |
| "step": 5680 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 2.2596315789473684e-05, | |
| "loss": 1.0305, | |
| "step": 5690 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 2.2580526315789474e-05, | |
| "loss": 1.0982, | |
| "step": 5700 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 2.2564736842105265e-05, | |
| "loss": 0.9984, | |
| "step": 5710 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 2.2548947368421055e-05, | |
| "loss": 1.0257, | |
| "step": 5720 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 2.2533157894736842e-05, | |
| "loss": 0.979, | |
| "step": 5730 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 2.2517368421052632e-05, | |
| "loss": 1.0156, | |
| "step": 5740 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 2.2501578947368423e-05, | |
| "loss": 1.0505, | |
| "step": 5750 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 2.2485789473684213e-05, | |
| "loss": 0.994, | |
| "step": 5760 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 2.247e-05, | |
| "loss": 0.9947, | |
| "step": 5770 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 2.245421052631579e-05, | |
| "loss": 1.0242, | |
| "step": 5780 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 2.2438421052631577e-05, | |
| "loss": 1.0089, | |
| "step": 5790 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 2.242263157894737e-05, | |
| "loss": 0.9797, | |
| "step": 5800 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 2.2406842105263158e-05, | |
| "loss": 0.9705, | |
| "step": 5810 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 2.239105263157895e-05, | |
| "loss": 0.9908, | |
| "step": 5820 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 2.2375263157894735e-05, | |
| "loss": 0.9629, | |
| "step": 5830 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 2.235947368421053e-05, | |
| "loss": 0.9659, | |
| "step": 5840 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 2.2343684210526316e-05, | |
| "loss": 1.0484, | |
| "step": 5850 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 2.2327894736842106e-05, | |
| "loss": 1.0439, | |
| "step": 5860 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 2.2312105263157893e-05, | |
| "loss": 0.9348, | |
| "step": 5870 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 2.2296315789473684e-05, | |
| "loss": 0.956, | |
| "step": 5880 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 2.2280526315789477e-05, | |
| "loss": 1.0675, | |
| "step": 5890 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 2.2264736842105264e-05, | |
| "loss": 0.9814, | |
| "step": 5900 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 2.2248947368421055e-05, | |
| "loss": 0.9494, | |
| "step": 5910 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 2.223315789473684e-05, | |
| "loss": 0.9037, | |
| "step": 5920 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 2.2217368421052632e-05, | |
| "loss": 1.0346, | |
| "step": 5930 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 2.2201578947368422e-05, | |
| "loss": 0.9887, | |
| "step": 5940 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 2.2185789473684213e-05, | |
| "loss": 0.9585, | |
| "step": 5950 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 2.217e-05, | |
| "loss": 0.9893, | |
| "step": 5960 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 2.215421052631579e-05, | |
| "loss": 1.0578, | |
| "step": 5970 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 2.213842105263158e-05, | |
| "loss": 0.9491, | |
| "step": 5980 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 2.212263157894737e-05, | |
| "loss": 0.9802, | |
| "step": 5990 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 2.2106842105263157e-05, | |
| "loss": 0.9155, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "eval_get_denotation_accuracy": 0.4527839097375521, | |
| "eval_loss": 0.9444504976272583, | |
| "eval_runtime": 257.8608, | |
| "eval_samples_per_second": 15.811, | |
| "eval_steps_per_second": 0.989, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 2.2091052631578948e-05, | |
| "loss": 1.0323, | |
| "step": 6010 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 2.2075263157894735e-05, | |
| "loss": 0.9507, | |
| "step": 6020 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 2.205947368421053e-05, | |
| "loss": 1.0078, | |
| "step": 6030 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 2.2043684210526315e-05, | |
| "loss": 1.0729, | |
| "step": 6040 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 2.2027894736842106e-05, | |
| "loss": 1.0941, | |
| "step": 6050 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 2.2012105263157896e-05, | |
| "loss": 1.0244, | |
| "step": 6060 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 2.1996315789473683e-05, | |
| "loss": 1.0042, | |
| "step": 6070 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 2.1980526315789477e-05, | |
| "loss": 0.93, | |
| "step": 6080 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 2.1964736842105264e-05, | |
| "loss": 1.0543, | |
| "step": 6090 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 2.1948947368421054e-05, | |
| "loss": 1.0498, | |
| "step": 6100 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 2.193315789473684e-05, | |
| "loss": 0.9669, | |
| "step": 6110 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 2.1917368421052635e-05, | |
| "loss": 0.9927, | |
| "step": 6120 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 2.190157894736842e-05, | |
| "loss": 0.9735, | |
| "step": 6130 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 2.1885789473684212e-05, | |
| "loss": 0.9773, | |
| "step": 6140 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 2.187e-05, | |
| "loss": 0.9675, | |
| "step": 6150 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 2.185421052631579e-05, | |
| "loss": 1.062, | |
| "step": 6160 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 2.183842105263158e-05, | |
| "loss": 0.9643, | |
| "step": 6170 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 2.182263157894737e-05, | |
| "loss": 0.9739, | |
| "step": 6180 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 2.1806842105263157e-05, | |
| "loss": 0.9491, | |
| "step": 6190 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 2.1791052631578947e-05, | |
| "loss": 0.9716, | |
| "step": 6200 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 2.1775263157894734e-05, | |
| "loss": 0.9303, | |
| "step": 6210 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 2.1759473684210528e-05, | |
| "loss": 0.9936, | |
| "step": 6220 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 2.1743684210526318e-05, | |
| "loss": 0.9463, | |
| "step": 6230 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 2.1727894736842105e-05, | |
| "loss": 0.9315, | |
| "step": 6240 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 2.1712105263157896e-05, | |
| "loss": 1.0116, | |
| "step": 6250 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 2.1696315789473686e-05, | |
| "loss": 0.9661, | |
| "step": 6260 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 2.1680526315789476e-05, | |
| "loss": 0.9708, | |
| "step": 6270 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 2.1664736842105263e-05, | |
| "loss": 0.9284, | |
| "step": 6280 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 2.1648947368421054e-05, | |
| "loss": 0.969, | |
| "step": 6290 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 2.163315789473684e-05, | |
| "loss": 0.9366, | |
| "step": 6300 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 2.1617368421052634e-05, | |
| "loss": 0.9587, | |
| "step": 6310 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 2.160157894736842e-05, | |
| "loss": 1.031, | |
| "step": 6320 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 2.158578947368421e-05, | |
| "loss": 0.9711, | |
| "step": 6330 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 2.157e-05, | |
| "loss": 0.9997, | |
| "step": 6340 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 2.155421052631579e-05, | |
| "loss": 1.0411, | |
| "step": 6350 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 2.153842105263158e-05, | |
| "loss": 1.0691, | |
| "step": 6360 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 2.152263157894737e-05, | |
| "loss": 1.0019, | |
| "step": 6370 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 2.1506842105263156e-05, | |
| "loss": 0.9699, | |
| "step": 6380 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 2.1491052631578947e-05, | |
| "loss": 0.9594, | |
| "step": 6390 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 2.147526315789474e-05, | |
| "loss": 0.9815, | |
| "step": 6400 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 2.1459473684210527e-05, | |
| "loss": 1.0267, | |
| "step": 6410 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 2.1443684210526318e-05, | |
| "loss": 1.0074, | |
| "step": 6420 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 2.1427894736842105e-05, | |
| "loss": 0.9747, | |
| "step": 6430 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 2.1412105263157895e-05, | |
| "loss": 0.9391, | |
| "step": 6440 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 2.1396315789473685e-05, | |
| "loss": 1.0059, | |
| "step": 6450 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 2.1380526315789476e-05, | |
| "loss": 0.9702, | |
| "step": 6460 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 2.1364736842105263e-05, | |
| "loss": 0.9454, | |
| "step": 6470 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 2.1348947368421053e-05, | |
| "loss": 0.9945, | |
| "step": 6480 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 2.133315789473684e-05, | |
| "loss": 0.9342, | |
| "step": 6490 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 2.1317368421052634e-05, | |
| "loss": 0.9979, | |
| "step": 6500 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "eval_get_denotation_accuracy": 0.47387785136129507, | |
| "eval_loss": 0.8795158863067627, | |
| "eval_runtime": 249.8736, | |
| "eval_samples_per_second": 16.316, | |
| "eval_steps_per_second": 1.021, | |
| "step": 6500 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 2.130157894736842e-05, | |
| "loss": 0.9221, | |
| "step": 6510 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 2.128578947368421e-05, | |
| "loss": 0.9871, | |
| "step": 6520 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 2.1269999999999998e-05, | |
| "loss": 0.9627, | |
| "step": 6530 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 2.125421052631579e-05, | |
| "loss": 0.9246, | |
| "step": 6540 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 2.123842105263158e-05, | |
| "loss": 1.0387, | |
| "step": 6550 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 2.122263157894737e-05, | |
| "loss": 0.9356, | |
| "step": 6560 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 2.120684210526316e-05, | |
| "loss": 0.9432, | |
| "step": 6570 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 2.1191052631578946e-05, | |
| "loss": 0.9311, | |
| "step": 6580 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 2.117526315789474e-05, | |
| "loss": 0.9183, | |
| "step": 6590 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 2.1159473684210527e-05, | |
| "loss": 0.9224, | |
| "step": 6600 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 2.1143684210526317e-05, | |
| "loss": 0.9364, | |
| "step": 6610 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 2.1127894736842104e-05, | |
| "loss": 1.0059, | |
| "step": 6620 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 2.1112105263157895e-05, | |
| "loss": 0.9316, | |
| "step": 6630 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 2.1096315789473685e-05, | |
| "loss": 0.9801, | |
| "step": 6640 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 2.1080526315789475e-05, | |
| "loss": 0.9428, | |
| "step": 6650 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 2.1064736842105262e-05, | |
| "loss": 0.9937, | |
| "step": 6660 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 2.1048947368421053e-05, | |
| "loss": 0.9906, | |
| "step": 6670 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 2.1033157894736843e-05, | |
| "loss": 1.0346, | |
| "step": 6680 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 2.1017368421052633e-05, | |
| "loss": 0.9906, | |
| "step": 6690 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 2.100157894736842e-05, | |
| "loss": 1.0524, | |
| "step": 6700 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 2.098578947368421e-05, | |
| "loss": 0.961, | |
| "step": 6710 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 2.097e-05, | |
| "loss": 0.9522, | |
| "step": 6720 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 2.095421052631579e-05, | |
| "loss": 0.9366, | |
| "step": 6730 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 2.093842105263158e-05, | |
| "loss": 1.0735, | |
| "step": 6740 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 2.092263157894737e-05, | |
| "loss": 0.8852, | |
| "step": 6750 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 2.090684210526316e-05, | |
| "loss": 0.9389, | |
| "step": 6760 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 2.0891052631578946e-05, | |
| "loss": 1.0394, | |
| "step": 6770 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 2.087526315789474e-05, | |
| "loss": 1.0329, | |
| "step": 6780 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 2.0859473684210526e-05, | |
| "loss": 0.9396, | |
| "step": 6790 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 2.0843684210526317e-05, | |
| "loss": 0.9846, | |
| "step": 6800 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 2.0827894736842104e-05, | |
| "loss": 1.0176, | |
| "step": 6810 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 2.0812105263157897e-05, | |
| "loss": 0.9905, | |
| "step": 6820 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 2.0796315789473684e-05, | |
| "loss": 0.9654, | |
| "step": 6830 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 2.0780526315789475e-05, | |
| "loss": 1.002, | |
| "step": 6840 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 2.076473684210526e-05, | |
| "loss": 0.9142, | |
| "step": 6850 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 2.0748947368421052e-05, | |
| "loss": 0.9308, | |
| "step": 6860 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 2.0733157894736842e-05, | |
| "loss": 0.9985, | |
| "step": 6870 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 2.0717368421052633e-05, | |
| "loss": 1.0112, | |
| "step": 6880 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 2.0701578947368423e-05, | |
| "loss": 0.9749, | |
| "step": 6890 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 2.068578947368421e-05, | |
| "loss": 0.9981, | |
| "step": 6900 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 2.067e-05, | |
| "loss": 0.9432, | |
| "step": 6910 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 2.065421052631579e-05, | |
| "loss": 0.9547, | |
| "step": 6920 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 2.063842105263158e-05, | |
| "loss": 1.0422, | |
| "step": 6930 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 2.0622631578947368e-05, | |
| "loss": 1.0602, | |
| "step": 6940 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 2.0606842105263158e-05, | |
| "loss": 0.9757, | |
| "step": 6950 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 2.059105263157895e-05, | |
| "loss": 1.0051, | |
| "step": 6960 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 2.057526315789474e-05, | |
| "loss": 0.9267, | |
| "step": 6970 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 2.0559473684210526e-05, | |
| "loss": 0.9647, | |
| "step": 6980 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 2.0543684210526316e-05, | |
| "loss": 1.0339, | |
| "step": 6990 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 2.0527894736842103e-05, | |
| "loss": 0.9185, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "eval_get_denotation_accuracy": 0.47412312975226883, | |
| "eval_loss": 0.8895652294158936, | |
| "eval_runtime": 240.0315, | |
| "eval_samples_per_second": 16.985, | |
| "eval_steps_per_second": 1.062, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 2.0512105263157897e-05, | |
| "loss": 0.9818, | |
| "step": 7010 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 2.0496315789473684e-05, | |
| "loss": 0.9737, | |
| "step": 7020 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 2.0480526315789474e-05, | |
| "loss": 1.0036, | |
| "step": 7030 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 2.046473684210526e-05, | |
| "loss": 0.9822, | |
| "step": 7040 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 2.044894736842105e-05, | |
| "loss": 0.9819, | |
| "step": 7050 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 2.0433157894736845e-05, | |
| "loss": 0.9024, | |
| "step": 7060 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 2.0417368421052632e-05, | |
| "loss": 1.0799, | |
| "step": 7070 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 2.0401578947368422e-05, | |
| "loss": 0.9576, | |
| "step": 7080 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 2.038578947368421e-05, | |
| "loss": 0.9659, | |
| "step": 7090 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 2.0370000000000003e-05, | |
| "loss": 0.9756, | |
| "step": 7100 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 2.035421052631579e-05, | |
| "loss": 1.0247, | |
| "step": 7110 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 2.033842105263158e-05, | |
| "loss": 0.9834, | |
| "step": 7120 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 2.0322631578947367e-05, | |
| "loss": 0.9607, | |
| "step": 7130 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 2.0306842105263158e-05, | |
| "loss": 0.8982, | |
| "step": 7140 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 2.0291052631578948e-05, | |
| "loss": 0.9772, | |
| "step": 7150 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 2.027526315789474e-05, | |
| "loss": 0.9037, | |
| "step": 7160 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 2.0259473684210525e-05, | |
| "loss": 0.9059, | |
| "step": 7170 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 2.0243684210526316e-05, | |
| "loss": 0.9981, | |
| "step": 7180 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 2.0227894736842106e-05, | |
| "loss": 1.0669, | |
| "step": 7190 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 2.0212105263157896e-05, | |
| "loss": 0.9879, | |
| "step": 7200 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 2.0196315789473683e-05, | |
| "loss": 1.0508, | |
| "step": 7210 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 2.0180526315789474e-05, | |
| "loss": 0.9472, | |
| "step": 7220 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 2.0164736842105264e-05, | |
| "loss": 0.9703, | |
| "step": 7230 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 2.0148947368421054e-05, | |
| "loss": 1.0772, | |
| "step": 7240 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 2.0133157894736845e-05, | |
| "loss": 1.0144, | |
| "step": 7250 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 2.011736842105263e-05, | |
| "loss": 1.0318, | |
| "step": 7260 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 2.0101578947368422e-05, | |
| "loss": 0.9503, | |
| "step": 7270 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 2.008578947368421e-05, | |
| "loss": 1.0449, | |
| "step": 7280 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 2.0070000000000003e-05, | |
| "loss": 1.0206, | |
| "step": 7290 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 2.005421052631579e-05, | |
| "loss": 0.9528, | |
| "step": 7300 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 2.003842105263158e-05, | |
| "loss": 0.9294, | |
| "step": 7310 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 2.0022631578947367e-05, | |
| "loss": 1.0417, | |
| "step": 7320 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 2.000684210526316e-05, | |
| "loss": 1.0689, | |
| "step": 7330 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 1.9991052631578948e-05, | |
| "loss": 0.8967, | |
| "step": 7340 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 1.9975263157894738e-05, | |
| "loss": 0.9687, | |
| "step": 7350 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 1.9959473684210525e-05, | |
| "loss": 1.003, | |
| "step": 7360 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 1.9943684210526315e-05, | |
| "loss": 0.9772, | |
| "step": 7370 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 1.9927894736842106e-05, | |
| "loss": 0.9683, | |
| "step": 7380 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 1.9912105263157896e-05, | |
| "loss": 0.9457, | |
| "step": 7390 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 1.9896315789473686e-05, | |
| "loss": 0.9867, | |
| "step": 7400 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 1.9880526315789473e-05, | |
| "loss": 0.9552, | |
| "step": 7410 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 1.9864736842105264e-05, | |
| "loss": 0.9849, | |
| "step": 7420 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 1.9848947368421054e-05, | |
| "loss": 0.9488, | |
| "step": 7430 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 1.9833157894736844e-05, | |
| "loss": 1.0298, | |
| "step": 7440 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 1.981736842105263e-05, | |
| "loss": 0.9471, | |
| "step": 7450 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 1.980157894736842e-05, | |
| "loss": 0.9553, | |
| "step": 7460 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 1.9785789473684212e-05, | |
| "loss": 0.9433, | |
| "step": 7470 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 1.9770000000000002e-05, | |
| "loss": 1.0108, | |
| "step": 7480 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 1.975421052631579e-05, | |
| "loss": 0.982, | |
| "step": 7490 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 1.973842105263158e-05, | |
| "loss": 0.9941, | |
| "step": 7500 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "eval_get_denotation_accuracy": 0.4638214373313711, | |
| "eval_loss": 0.9287481904029846, | |
| "eval_runtime": 246.9862, | |
| "eval_samples_per_second": 16.507, | |
| "eval_steps_per_second": 1.032, | |
| "step": 7500 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 1.9722631578947366e-05, | |
| "loss": 0.9911, | |
| "step": 7510 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 1.970684210526316e-05, | |
| "loss": 0.8973, | |
| "step": 7520 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 1.9691052631578947e-05, | |
| "loss": 0.9224, | |
| "step": 7530 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 1.9675263157894737e-05, | |
| "loss": 0.9358, | |
| "step": 7540 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 1.9659473684210528e-05, | |
| "loss": 0.9473, | |
| "step": 7550 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 1.9643684210526315e-05, | |
| "loss": 0.9648, | |
| "step": 7560 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 1.962789473684211e-05, | |
| "loss": 0.9403, | |
| "step": 7570 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 1.9612105263157895e-05, | |
| "loss": 1.1487, | |
| "step": 7580 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 1.9596315789473686e-05, | |
| "loss": 1.009, | |
| "step": 7590 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 1.9580526315789473e-05, | |
| "loss": 0.915, | |
| "step": 7600 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 1.9564736842105266e-05, | |
| "loss": 0.92, | |
| "step": 7610 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 1.9548947368421053e-05, | |
| "loss": 0.9518, | |
| "step": 7620 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 1.9533157894736844e-05, | |
| "loss": 0.9731, | |
| "step": 7630 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 1.951736842105263e-05, | |
| "loss": 0.9952, | |
| "step": 7640 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 1.950157894736842e-05, | |
| "loss": 0.9623, | |
| "step": 7650 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 1.948578947368421e-05, | |
| "loss": 0.9247, | |
| "step": 7660 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 1.947e-05, | |
| "loss": 1.0254, | |
| "step": 7670 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 1.945421052631579e-05, | |
| "loss": 0.9958, | |
| "step": 7680 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 1.943842105263158e-05, | |
| "loss": 0.9872, | |
| "step": 7690 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 1.9422631578947366e-05, | |
| "loss": 0.9981, | |
| "step": 7700 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 1.940684210526316e-05, | |
| "loss": 0.9812, | |
| "step": 7710 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 1.939105263157895e-05, | |
| "loss": 0.9483, | |
| "step": 7720 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 1.9375263157894737e-05, | |
| "loss": 0.9892, | |
| "step": 7730 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 1.9359473684210527e-05, | |
| "loss": 1.034, | |
| "step": 7740 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 1.9343684210526318e-05, | |
| "loss": 0.9821, | |
| "step": 7750 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 1.9327894736842108e-05, | |
| "loss": 0.9924, | |
| "step": 7760 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 1.9312105263157895e-05, | |
| "loss": 0.9663, | |
| "step": 7770 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 1.9296315789473685e-05, | |
| "loss": 1.0233, | |
| "step": 7780 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 1.9280526315789472e-05, | |
| "loss": 1.0078, | |
| "step": 7790 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 1.9264736842105266e-05, | |
| "loss": 0.9736, | |
| "step": 7800 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 1.9248947368421053e-05, | |
| "loss": 0.9986, | |
| "step": 7810 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 1.9233157894736843e-05, | |
| "loss": 1.006, | |
| "step": 7820 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 1.921736842105263e-05, | |
| "loss": 0.9502, | |
| "step": 7830 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 1.920157894736842e-05, | |
| "loss": 0.9941, | |
| "step": 7840 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 1.918578947368421e-05, | |
| "loss": 0.9671, | |
| "step": 7850 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 1.917e-05, | |
| "loss": 0.9958, | |
| "step": 7860 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 1.9154210526315788e-05, | |
| "loss": 0.937, | |
| "step": 7870 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 1.913842105263158e-05, | |
| "loss": 0.912, | |
| "step": 7880 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 1.9122631578947372e-05, | |
| "loss": 0.9927, | |
| "step": 7890 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 1.910684210526316e-05, | |
| "loss": 1.0253, | |
| "step": 7900 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 1.909105263157895e-05, | |
| "loss": 0.9658, | |
| "step": 7910 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 1.9075263157894736e-05, | |
| "loss": 1.0006, | |
| "step": 7920 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 1.9059473684210527e-05, | |
| "loss": 0.9963, | |
| "step": 7930 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 1.9043684210526317e-05, | |
| "loss": 0.9182, | |
| "step": 7940 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 1.9027894736842107e-05, | |
| "loss": 1.0314, | |
| "step": 7950 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 1.9012105263157894e-05, | |
| "loss": 0.9469, | |
| "step": 7960 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 1.8996315789473685e-05, | |
| "loss": 0.9323, | |
| "step": 7970 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 1.898052631578947e-05, | |
| "loss": 0.8863, | |
| "step": 7980 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 1.8964736842105265e-05, | |
| "loss": 0.9131, | |
| "step": 7990 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 1.8948947368421052e-05, | |
| "loss": 0.9525, | |
| "step": 8000 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "eval_get_denotation_accuracy": 0.4677458915869512, | |
| "eval_loss": 0.9251512289047241, | |
| "eval_runtime": 248.4984, | |
| "eval_samples_per_second": 16.407, | |
| "eval_steps_per_second": 1.026, | |
| "step": 8000 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 1.8933157894736843e-05, | |
| "loss": 1.0018, | |
| "step": 8010 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 1.891736842105263e-05, | |
| "loss": 0.9985, | |
| "step": 8020 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 1.8901578947368423e-05, | |
| "loss": 0.93, | |
| "step": 8030 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 1.888578947368421e-05, | |
| "loss": 0.9973, | |
| "step": 8040 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 1.887e-05, | |
| "loss": 1.058, | |
| "step": 8050 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 1.885421052631579e-05, | |
| "loss": 1.0463, | |
| "step": 8060 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 1.8838421052631578e-05, | |
| "loss": 1.0016, | |
| "step": 8070 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 1.882263157894737e-05, | |
| "loss": 0.9876, | |
| "step": 8080 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 1.880684210526316e-05, | |
| "loss": 0.9702, | |
| "step": 8090 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 1.879105263157895e-05, | |
| "loss": 0.9736, | |
| "step": 8100 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 1.8775263157894736e-05, | |
| "loss": 0.9938, | |
| "step": 8110 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 1.8759473684210526e-05, | |
| "loss": 0.9937, | |
| "step": 8120 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 1.8743684210526317e-05, | |
| "loss": 0.9345, | |
| "step": 8130 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 1.8727894736842107e-05, | |
| "loss": 1.0055, | |
| "step": 8140 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 1.8712105263157894e-05, | |
| "loss": 0.9901, | |
| "step": 8150 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 1.8696315789473684e-05, | |
| "loss": 1.0098, | |
| "step": 8160 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 1.8680526315789474e-05, | |
| "loss": 0.8845, | |
| "step": 8170 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 1.8664736842105265e-05, | |
| "loss": 0.9496, | |
| "step": 8180 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 1.8648947368421052e-05, | |
| "loss": 0.9832, | |
| "step": 8190 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 1.8633157894736842e-05, | |
| "loss": 1.0295, | |
| "step": 8200 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 1.861736842105263e-05, | |
| "loss": 0.9603, | |
| "step": 8210 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 1.8601578947368423e-05, | |
| "loss": 0.9739, | |
| "step": 8220 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 1.8585789473684213e-05, | |
| "loss": 1.033, | |
| "step": 8230 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 1.857e-05, | |
| "loss": 0.9615, | |
| "step": 8240 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 1.855421052631579e-05, | |
| "loss": 0.9258, | |
| "step": 8250 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 1.8538421052631577e-05, | |
| "loss": 1.0015, | |
| "step": 8260 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 1.852263157894737e-05, | |
| "loss": 0.9972, | |
| "step": 8270 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 1.8506842105263158e-05, | |
| "loss": 1.0129, | |
| "step": 8280 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 1.849105263157895e-05, | |
| "loss": 1.0155, | |
| "step": 8290 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 1.8475263157894735e-05, | |
| "loss": 1.0267, | |
| "step": 8300 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 1.845947368421053e-05, | |
| "loss": 0.9773, | |
| "step": 8310 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 1.8443684210526316e-05, | |
| "loss": 0.9881, | |
| "step": 8320 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 1.8427894736842106e-05, | |
| "loss": 0.9909, | |
| "step": 8330 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 1.8412105263157893e-05, | |
| "loss": 0.9886, | |
| "step": 8340 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 1.8396315789473684e-05, | |
| "loss": 0.9674, | |
| "step": 8350 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 1.8380526315789474e-05, | |
| "loss": 0.9482, | |
| "step": 8360 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 1.8364736842105264e-05, | |
| "loss": 0.9497, | |
| "step": 8370 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 1.834894736842105e-05, | |
| "loss": 0.9422, | |
| "step": 8380 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 1.833315789473684e-05, | |
| "loss": 0.8988, | |
| "step": 8390 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 1.8317368421052632e-05, | |
| "loss": 0.9614, | |
| "step": 8400 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 1.8301578947368422e-05, | |
| "loss": 0.9971, | |
| "step": 8410 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 1.8285789473684213e-05, | |
| "loss": 0.8827, | |
| "step": 8420 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 1.827e-05, | |
| "loss": 0.9785, | |
| "step": 8430 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 1.825421052631579e-05, | |
| "loss": 0.993, | |
| "step": 8440 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 1.823842105263158e-05, | |
| "loss": 0.9771, | |
| "step": 8450 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 1.822263157894737e-05, | |
| "loss": 0.9741, | |
| "step": 8460 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 1.8206842105263158e-05, | |
| "loss": 0.9621, | |
| "step": 8470 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 1.8191052631578948e-05, | |
| "loss": 0.9828, | |
| "step": 8480 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 1.8175263157894735e-05, | |
| "loss": 0.9377, | |
| "step": 8490 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 1.815947368421053e-05, | |
| "loss": 1.0148, | |
| "step": 8500 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "eval_get_denotation_accuracy": 0.4616139318126073, | |
| "eval_loss": 0.9519839882850647, | |
| "eval_runtime": 259.0893, | |
| "eval_samples_per_second": 15.736, | |
| "eval_steps_per_second": 0.984, | |
| "step": 8500 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 1.8143684210526315e-05, | |
| "loss": 0.9269, | |
| "step": 8510 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 1.8127894736842106e-05, | |
| "loss": 1.024, | |
| "step": 8520 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 1.8112105263157893e-05, | |
| "loss": 0.8278, | |
| "step": 8530 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 1.8096315789473683e-05, | |
| "loss": 0.9145, | |
| "step": 8540 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 1.8080526315789473e-05, | |
| "loss": 0.9236, | |
| "step": 8550 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 1.8064736842105264e-05, | |
| "loss": 0.9723, | |
| "step": 8560 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 1.8048947368421054e-05, | |
| "loss": 0.9367, | |
| "step": 8570 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 1.803315789473684e-05, | |
| "loss": 1.0005, | |
| "step": 8580 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 1.8017368421052635e-05, | |
| "loss": 1.0059, | |
| "step": 8590 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 1.8001578947368422e-05, | |
| "loss": 1.0237, | |
| "step": 8600 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 1.7985789473684212e-05, | |
| "loss": 0.9098, | |
| "step": 8610 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 1.797e-05, | |
| "loss": 0.9344, | |
| "step": 8620 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 1.795421052631579e-05, | |
| "loss": 1.0177, | |
| "step": 8630 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 1.793842105263158e-05, | |
| "loss": 0.9476, | |
| "step": 8640 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 1.792263157894737e-05, | |
| "loss": 0.9419, | |
| "step": 8650 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 1.7906842105263157e-05, | |
| "loss": 1.0584, | |
| "step": 8660 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 1.7891052631578947e-05, | |
| "loss": 0.9816, | |
| "step": 8670 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 1.7875263157894734e-05, | |
| "loss": 0.8823, | |
| "step": 8680 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 1.7859473684210528e-05, | |
| "loss": 0.8585, | |
| "step": 8690 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 1.7843684210526315e-05, | |
| "loss": 1.0386, | |
| "step": 8700 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 1.7827894736842105e-05, | |
| "loss": 1.0144, | |
| "step": 8710 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 1.7812105263157896e-05, | |
| "loss": 0.8985, | |
| "step": 8720 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 1.7796315789473686e-05, | |
| "loss": 0.9835, | |
| "step": 8730 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 1.7780526315789476e-05, | |
| "loss": 1.004, | |
| "step": 8740 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 1.7764736842105263e-05, | |
| "loss": 1.0201, | |
| "step": 8750 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 1.7748947368421054e-05, | |
| "loss": 1.0519, | |
| "step": 8760 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 1.773315789473684e-05, | |
| "loss": 0.9284, | |
| "step": 8770 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 1.7717368421052634e-05, | |
| "loss": 0.9925, | |
| "step": 8780 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 1.770157894736842e-05, | |
| "loss": 0.9654, | |
| "step": 8790 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 1.768578947368421e-05, | |
| "loss": 0.9423, | |
| "step": 8800 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 1.767e-05, | |
| "loss": 0.9918, | |
| "step": 8810 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 1.7654210526315792e-05, | |
| "loss": 0.9739, | |
| "step": 8820 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 1.763842105263158e-05, | |
| "loss": 0.9541, | |
| "step": 8830 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 1.762263157894737e-05, | |
| "loss": 0.9928, | |
| "step": 8840 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 1.7606842105263156e-05, | |
| "loss": 0.8675, | |
| "step": 8850 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 1.7591052631578947e-05, | |
| "loss": 1.0327, | |
| "step": 8860 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 1.7575263157894737e-05, | |
| "loss": 0.9771, | |
| "step": 8870 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 1.7559473684210528e-05, | |
| "loss": 0.9863, | |
| "step": 8880 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 1.7543684210526318e-05, | |
| "loss": 0.9929, | |
| "step": 8890 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 1.7527894736842105e-05, | |
| "loss": 1.0211, | |
| "step": 8900 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 1.7512105263157895e-05, | |
| "loss": 0.9558, | |
| "step": 8910 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 1.7496315789473685e-05, | |
| "loss": 1.0041, | |
| "step": 8920 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 1.7480526315789476e-05, | |
| "loss": 0.9815, | |
| "step": 8930 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 1.7464736842105263e-05, | |
| "loss": 0.9055, | |
| "step": 8940 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 1.7448947368421053e-05, | |
| "loss": 0.962, | |
| "step": 8950 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 1.7433157894736843e-05, | |
| "loss": 0.9739, | |
| "step": 8960 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 1.7417368421052634e-05, | |
| "loss": 0.9835, | |
| "step": 8970 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 1.740157894736842e-05, | |
| "loss": 0.9523, | |
| "step": 8980 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 1.738578947368421e-05, | |
| "loss": 0.9809, | |
| "step": 8990 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 1.7369999999999998e-05, | |
| "loss": 0.8853, | |
| "step": 9000 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "eval_get_denotation_accuracy": 0.46455727250429235, | |
| "eval_loss": 0.926638126373291, | |
| "eval_runtime": 236.6855, | |
| "eval_samples_per_second": 17.225, | |
| "eval_steps_per_second": 1.077, | |
| "step": 9000 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 1.7354210526315792e-05, | |
| "loss": 0.9512, | |
| "step": 9010 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 1.733842105263158e-05, | |
| "loss": 0.9229, | |
| "step": 9020 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 1.732263157894737e-05, | |
| "loss": 0.9306, | |
| "step": 9030 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 1.7306842105263156e-05, | |
| "loss": 0.9765, | |
| "step": 9040 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 1.7291052631578946e-05, | |
| "loss": 0.9626, | |
| "step": 9050 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 1.727526315789474e-05, | |
| "loss": 0.9705, | |
| "step": 9060 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 1.7259473684210527e-05, | |
| "loss": 0.9375, | |
| "step": 9070 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 1.7243684210526317e-05, | |
| "loss": 0.9996, | |
| "step": 9080 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 1.7227894736842104e-05, | |
| "loss": 0.9406, | |
| "step": 9090 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 1.7212105263157898e-05, | |
| "loss": 0.975, | |
| "step": 9100 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 1.7196315789473685e-05, | |
| "loss": 1.0174, | |
| "step": 9110 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 1.7180526315789475e-05, | |
| "loss": 0.9311, | |
| "step": 9120 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 1.7164736842105262e-05, | |
| "loss": 0.9646, | |
| "step": 9130 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 1.7148947368421053e-05, | |
| "loss": 1.0456, | |
| "step": 9140 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 1.7133157894736843e-05, | |
| "loss": 0.9987, | |
| "step": 9150 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 1.7117368421052633e-05, | |
| "loss": 0.8931, | |
| "step": 9160 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 1.710157894736842e-05, | |
| "loss": 0.921, | |
| "step": 9170 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 1.708578947368421e-05, | |
| "loss": 1.0058, | |
| "step": 9180 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 1.7069999999999998e-05, | |
| "loss": 0.9869, | |
| "step": 9190 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 1.705421052631579e-05, | |
| "loss": 0.9661, | |
| "step": 9200 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 1.7038421052631578e-05, | |
| "loss": 0.9359, | |
| "step": 9210 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 1.702263157894737e-05, | |
| "loss": 0.9785, | |
| "step": 9220 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 1.700684210526316e-05, | |
| "loss": 0.9785, | |
| "step": 9230 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 1.699105263157895e-05, | |
| "loss": 0.9692, | |
| "step": 9240 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 1.697526315789474e-05, | |
| "loss": 0.9677, | |
| "step": 9250 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 1.6959473684210526e-05, | |
| "loss": 0.9376, | |
| "step": 9260 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 1.6943684210526317e-05, | |
| "loss": 0.9206, | |
| "step": 9270 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 1.6927894736842104e-05, | |
| "loss": 0.9357, | |
| "step": 9280 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 1.6912105263157897e-05, | |
| "loss": 0.9246, | |
| "step": 9290 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 1.6896315789473684e-05, | |
| "loss": 1.0461, | |
| "step": 9300 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 1.6880526315789475e-05, | |
| "loss": 0.9399, | |
| "step": 9310 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 1.6864736842105262e-05, | |
| "loss": 0.9181, | |
| "step": 9320 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 1.6848947368421052e-05, | |
| "loss": 1.0124, | |
| "step": 9330 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 1.6833157894736842e-05, | |
| "loss": 1.0314, | |
| "step": 9340 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 1.6817368421052633e-05, | |
| "loss": 0.9966, | |
| "step": 9350 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 1.680157894736842e-05, | |
| "loss": 0.8777, | |
| "step": 9360 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 1.678578947368421e-05, | |
| "loss": 0.9791, | |
| "step": 9370 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 1.677e-05, | |
| "loss": 0.9, | |
| "step": 9380 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 1.675421052631579e-05, | |
| "loss": 1.0118, | |
| "step": 9390 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 1.673842105263158e-05, | |
| "loss": 0.9493, | |
| "step": 9400 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 1.6722631578947368e-05, | |
| "loss": 0.9875, | |
| "step": 9410 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 1.670684210526316e-05, | |
| "loss": 0.981, | |
| "step": 9420 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 1.669105263157895e-05, | |
| "loss": 0.9798, | |
| "step": 9430 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 1.667526315789474e-05, | |
| "loss": 0.909, | |
| "step": 9440 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 1.6659473684210526e-05, | |
| "loss": 0.9592, | |
| "step": 9450 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 1.6643684210526316e-05, | |
| "loss": 0.9579, | |
| "step": 9460 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 1.6627894736842103e-05, | |
| "loss": 0.9824, | |
| "step": 9470 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 1.6612105263157897e-05, | |
| "loss": 0.9537, | |
| "step": 9480 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 1.6596315789473684e-05, | |
| "loss": 0.9407, | |
| "step": 9490 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 1.6580526315789474e-05, | |
| "loss": 0.9297, | |
| "step": 9500 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "eval_get_denotation_accuracy": 0.4782928623988227, | |
| "eval_loss": 0.9149366021156311, | |
| "eval_runtime": 246.4757, | |
| "eval_samples_per_second": 16.541, | |
| "eval_steps_per_second": 1.035, | |
| "step": 9500 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 1.656473684210526e-05, | |
| "loss": 0.9581, | |
| "step": 9510 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 1.6548947368421055e-05, | |
| "loss": 1.0649, | |
| "step": 9520 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 1.6533157894736842e-05, | |
| "loss": 0.8826, | |
| "step": 9530 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 1.6517368421052632e-05, | |
| "loss": 0.9945, | |
| "step": 9540 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 1.6501578947368423e-05, | |
| "loss": 0.9891, | |
| "step": 9550 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 1.648578947368421e-05, | |
| "loss": 0.9985, | |
| "step": 9560 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 1.6470000000000003e-05, | |
| "loss": 0.919, | |
| "step": 9570 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 1.645421052631579e-05, | |
| "loss": 0.9338, | |
| "step": 9580 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 1.643842105263158e-05, | |
| "loss": 0.957, | |
| "step": 9590 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 1.6422631578947367e-05, | |
| "loss": 1.0452, | |
| "step": 9600 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 1.6406842105263158e-05, | |
| "loss": 1.0478, | |
| "step": 9610 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 1.6391052631578948e-05, | |
| "loss": 1.0028, | |
| "step": 9620 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 1.637526315789474e-05, | |
| "loss": 0.9938, | |
| "step": 9630 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 1.6359473684210525e-05, | |
| "loss": 0.9639, | |
| "step": 9640 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 1.6343684210526316e-05, | |
| "loss": 1.0243, | |
| "step": 9650 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 1.6327894736842106e-05, | |
| "loss": 1.0092, | |
| "step": 9660 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 1.6312105263157896e-05, | |
| "loss": 0.9481, | |
| "step": 9670 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 1.6296315789473683e-05, | |
| "loss": 1.0041, | |
| "step": 9680 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 1.6280526315789474e-05, | |
| "loss": 0.8919, | |
| "step": 9690 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 1.626473684210526e-05, | |
| "loss": 0.923, | |
| "step": 9700 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 1.6248947368421054e-05, | |
| "loss": 0.971, | |
| "step": 9710 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 1.6233157894736845e-05, | |
| "loss": 0.9875, | |
| "step": 9720 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 1.6217368421052632e-05, | |
| "loss": 0.9286, | |
| "step": 9730 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 1.6201578947368422e-05, | |
| "loss": 0.915, | |
| "step": 9740 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 1.618578947368421e-05, | |
| "loss": 0.9622, | |
| "step": 9750 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 1.6170000000000003e-05, | |
| "loss": 1.0241, | |
| "step": 9760 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 1.615421052631579e-05, | |
| "loss": 0.9866, | |
| "step": 9770 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 1.613842105263158e-05, | |
| "loss": 1.0167, | |
| "step": 9780 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 1.6122631578947367e-05, | |
| "loss": 0.9456, | |
| "step": 9790 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 1.610684210526316e-05, | |
| "loss": 0.9988, | |
| "step": 9800 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 1.6091052631578948e-05, | |
| "loss": 0.9432, | |
| "step": 9810 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 1.6075263157894738e-05, | |
| "loss": 0.9822, | |
| "step": 9820 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 1.6059473684210525e-05, | |
| "loss": 0.9149, | |
| "step": 9830 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 1.6043684210526315e-05, | |
| "loss": 0.8641, | |
| "step": 9840 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 1.6027894736842106e-05, | |
| "loss": 0.9149, | |
| "step": 9850 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 1.6012105263157896e-05, | |
| "loss": 0.9868, | |
| "step": 9860 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 1.5996315789473683e-05, | |
| "loss": 0.9498, | |
| "step": 9870 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 1.5980526315789473e-05, | |
| "loss": 0.9213, | |
| "step": 9880 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 1.5964736842105264e-05, | |
| "loss": 0.9884, | |
| "step": 9890 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 1.5948947368421054e-05, | |
| "loss": 1.0028, | |
| "step": 9900 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 1.5933157894736844e-05, | |
| "loss": 0.9571, | |
| "step": 9910 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 1.591736842105263e-05, | |
| "loss": 0.9484, | |
| "step": 9920 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 1.590157894736842e-05, | |
| "loss": 0.9944, | |
| "step": 9930 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 1.5885789473684212e-05, | |
| "loss": 0.9915, | |
| "step": 9940 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 1.5870000000000002e-05, | |
| "loss": 1.0459, | |
| "step": 9950 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 1.585421052631579e-05, | |
| "loss": 1.0178, | |
| "step": 9960 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 1.583842105263158e-05, | |
| "loss": 0.9805, | |
| "step": 9970 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 1.5822631578947366e-05, | |
| "loss": 0.9694, | |
| "step": 9980 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 1.580684210526316e-05, | |
| "loss": 0.9671, | |
| "step": 9990 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 1.5791052631578947e-05, | |
| "loss": 0.9269, | |
| "step": 10000 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "eval_get_denotation_accuracy": 0.4726514594064263, | |
| "eval_loss": 0.9314199686050415, | |
| "eval_runtime": 229.6675, | |
| "eval_samples_per_second": 17.752, | |
| "eval_steps_per_second": 1.11, | |
| "step": 10000 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 1.5775263157894737e-05, | |
| "loss": 0.9701, | |
| "step": 10010 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.5759473684210524e-05, | |
| "loss": 0.9517, | |
| "step": 10020 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.5743684210526315e-05, | |
| "loss": 0.9066, | |
| "step": 10030 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.5727894736842105e-05, | |
| "loss": 1.0189, | |
| "step": 10040 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.5712105263157895e-05, | |
| "loss": 1.101, | |
| "step": 10050 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.5696315789473686e-05, | |
| "loss": 0.9877, | |
| "step": 10060 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.5680526315789473e-05, | |
| "loss": 1.002, | |
| "step": 10070 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.5664736842105266e-05, | |
| "loss": 0.9769, | |
| "step": 10080 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.5648947368421053e-05, | |
| "loss": 0.9049, | |
| "step": 10090 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.5633157894736844e-05, | |
| "loss": 0.9607, | |
| "step": 10100 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.561736842105263e-05, | |
| "loss": 0.9429, | |
| "step": 10110 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.560157894736842e-05, | |
| "loss": 0.9807, | |
| "step": 10120 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.558578947368421e-05, | |
| "loss": 0.8555, | |
| "step": 10130 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.5570000000000002e-05, | |
| "loss": 0.9652, | |
| "step": 10140 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.555421052631579e-05, | |
| "loss": 0.9268, | |
| "step": 10150 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.553842105263158e-05, | |
| "loss": 1.0229, | |
| "step": 10160 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.5522631578947366e-05, | |
| "loss": 0.9754, | |
| "step": 10170 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.550684210526316e-05, | |
| "loss": 1.0279, | |
| "step": 10180 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.5491052631578947e-05, | |
| "loss": 0.9685, | |
| "step": 10190 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.5475263157894737e-05, | |
| "loss": 0.9277, | |
| "step": 10200 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.5459473684210524e-05, | |
| "loss": 0.8861, | |
| "step": 10210 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.5443684210526318e-05, | |
| "loss": 0.9403, | |
| "step": 10220 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.5427894736842108e-05, | |
| "loss": 0.9502, | |
| "step": 10230 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.5412105263157895e-05, | |
| "loss": 0.9233, | |
| "step": 10240 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.5396315789473685e-05, | |
| "loss": 0.9991, | |
| "step": 10250 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.5380526315789472e-05, | |
| "loss": 0.9546, | |
| "step": 10260 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.5364736842105266e-05, | |
| "loss": 0.9074, | |
| "step": 10270 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.5348947368421053e-05, | |
| "loss": 0.9613, | |
| "step": 10280 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.5333157894736843e-05, | |
| "loss": 0.9134, | |
| "step": 10290 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.531736842105263e-05, | |
| "loss": 0.919, | |
| "step": 10300 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.530157894736842e-05, | |
| "loss": 0.9763, | |
| "step": 10310 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.528578947368421e-05, | |
| "loss": 0.9336, | |
| "step": 10320 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.527e-05, | |
| "loss": 0.9394, | |
| "step": 10330 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.525421052631579e-05, | |
| "loss": 1.007, | |
| "step": 10340 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.5238421052631578e-05, | |
| "loss": 0.8863, | |
| "step": 10350 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.522263157894737e-05, | |
| "loss": 1.0601, | |
| "step": 10360 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.5206842105263159e-05, | |
| "loss": 0.9999, | |
| "step": 10370 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.5191052631578948e-05, | |
| "loss": 1.0112, | |
| "step": 10380 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.5175263157894736e-05, | |
| "loss": 0.8849, | |
| "step": 10390 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.5159473684210525e-05, | |
| "loss": 0.9973, | |
| "step": 10400 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.5143684210526317e-05, | |
| "loss": 1.0451, | |
| "step": 10410 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.5127894736842106e-05, | |
| "loss": 0.9296, | |
| "step": 10420 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.5112105263157894e-05, | |
| "loss": 0.9863, | |
| "step": 10430 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.5096315789473683e-05, | |
| "loss": 0.9274, | |
| "step": 10440 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.5080526315789475e-05, | |
| "loss": 0.989, | |
| "step": 10450 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.5064736842105264e-05, | |
| "loss": 1.0079, | |
| "step": 10460 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.5048947368421054e-05, | |
| "loss": 0.8703, | |
| "step": 10470 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.5033157894736843e-05, | |
| "loss": 0.9883, | |
| "step": 10480 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.5017368421052631e-05, | |
| "loss": 0.9986, | |
| "step": 10490 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.5001578947368423e-05, | |
| "loss": 0.9734, | |
| "step": 10500 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "eval_get_denotation_accuracy": 0.4677458915869512, | |
| "eval_loss": 0.9230586886405945, | |
| "eval_runtime": 241.6832, | |
| "eval_samples_per_second": 16.869, | |
| "eval_steps_per_second": 1.055, | |
| "step": 10500 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.4985789473684212e-05, | |
| "loss": 0.9085, | |
| "step": 10510 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.497e-05, | |
| "loss": 1.0361, | |
| "step": 10520 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.495421052631579e-05, | |
| "loss": 0.9538, | |
| "step": 10530 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.493842105263158e-05, | |
| "loss": 1.0394, | |
| "step": 10540 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.4922631578947368e-05, | |
| "loss": 0.9146, | |
| "step": 10550 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.4906842105263159e-05, | |
| "loss": 0.8953, | |
| "step": 10560 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.4891052631578947e-05, | |
| "loss": 1.0349, | |
| "step": 10570 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.4875263157894738e-05, | |
| "loss": 0.9213, | |
| "step": 10580 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.4859473684210526e-05, | |
| "loss": 0.9519, | |
| "step": 10590 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.4843684210526315e-05, | |
| "loss": 0.9537, | |
| "step": 10600 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.4827894736842105e-05, | |
| "loss": 0.9251, | |
| "step": 10610 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.4812105263157894e-05, | |
| "loss": 0.9124, | |
| "step": 10620 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.4796315789473684e-05, | |
| "loss": 0.9608, | |
| "step": 10630 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.4780526315789475e-05, | |
| "loss": 0.9014, | |
| "step": 10640 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.4764736842105265e-05, | |
| "loss": 0.9642, | |
| "step": 10650 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.4748947368421054e-05, | |
| "loss": 1.0173, | |
| "step": 10660 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.4733157894736842e-05, | |
| "loss": 0.9941, | |
| "step": 10670 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.4717368421052633e-05, | |
| "loss": 0.9867, | |
| "step": 10680 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.4701578947368421e-05, | |
| "loss": 0.9687, | |
| "step": 10690 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.4685789473684212e-05, | |
| "loss": 0.9218, | |
| "step": 10700 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.467e-05, | |
| "loss": 0.9298, | |
| "step": 10710 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.465421052631579e-05, | |
| "loss": 0.9505, | |
| "step": 10720 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.4638421052631579e-05, | |
| "loss": 0.9584, | |
| "step": 10730 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.4622631578947368e-05, | |
| "loss": 0.8743, | |
| "step": 10740 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.4606842105263158e-05, | |
| "loss": 0.9154, | |
| "step": 10750 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.4591052631578947e-05, | |
| "loss": 0.9337, | |
| "step": 10760 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.4575263157894737e-05, | |
| "loss": 0.8526, | |
| "step": 10770 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.4559473684210526e-05, | |
| "loss": 0.9251, | |
| "step": 10780 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.4543684210526316e-05, | |
| "loss": 0.9347, | |
| "step": 10790 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.4527894736842105e-05, | |
| "loss": 0.9755, | |
| "step": 10800 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.4512105263157895e-05, | |
| "loss": 0.9797, | |
| "step": 10810 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.4496315789473685e-05, | |
| "loss": 0.9268, | |
| "step": 10820 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.4480526315789474e-05, | |
| "loss": 1.0137, | |
| "step": 10830 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.4464736842105264e-05, | |
| "loss": 0.9812, | |
| "step": 10840 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.4448947368421053e-05, | |
| "loss": 0.9803, | |
| "step": 10850 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.4433157894736843e-05, | |
| "loss": 0.9592, | |
| "step": 10860 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.4417368421052632e-05, | |
| "loss": 0.9835, | |
| "step": 10870 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.440157894736842e-05, | |
| "loss": 0.961, | |
| "step": 10880 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.4385789473684211e-05, | |
| "loss": 1.0207, | |
| "step": 10890 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.437e-05, | |
| "loss": 0.9549, | |
| "step": 10900 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.435421052631579e-05, | |
| "loss": 0.9783, | |
| "step": 10910 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.4338421052631579e-05, | |
| "loss": 0.8843, | |
| "step": 10920 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.4322631578947369e-05, | |
| "loss": 1.0533, | |
| "step": 10930 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.4306842105263158e-05, | |
| "loss": 0.9515, | |
| "step": 10940 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.4291052631578946e-05, | |
| "loss": 0.9166, | |
| "step": 10950 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.4275263157894737e-05, | |
| "loss": 0.9588, | |
| "step": 10960 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.4259473684210527e-05, | |
| "loss": 0.9285, | |
| "step": 10970 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.4243684210526317e-05, | |
| "loss": 1.013, | |
| "step": 10980 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.4227894736842106e-05, | |
| "loss": 0.9996, | |
| "step": 10990 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.4212105263157896e-05, | |
| "loss": 0.9477, | |
| "step": 11000 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "eval_get_denotation_accuracy": 0.4795192543536914, | |
| "eval_loss": 0.8999451994895935, | |
| "eval_runtime": 241.3624, | |
| "eval_samples_per_second": 16.892, | |
| "eval_steps_per_second": 1.057, | |
| "step": 11000 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.4196315789473685e-05, | |
| "loss": 0.9456, | |
| "step": 11010 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.4180526315789474e-05, | |
| "loss": 0.9384, | |
| "step": 11020 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.4164736842105264e-05, | |
| "loss": 0.9776, | |
| "step": 11030 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.4148947368421053e-05, | |
| "loss": 0.9382, | |
| "step": 11040 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.4133157894736843e-05, | |
| "loss": 1.0121, | |
| "step": 11050 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.4117368421052631e-05, | |
| "loss": 0.9032, | |
| "step": 11060 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.4101578947368422e-05, | |
| "loss": 0.9912, | |
| "step": 11070 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.408578947368421e-05, | |
| "loss": 0.9444, | |
| "step": 11080 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.4069999999999999e-05, | |
| "loss": 1.0416, | |
| "step": 11090 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.405421052631579e-05, | |
| "loss": 0.8772, | |
| "step": 11100 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.4038421052631578e-05, | |
| "loss": 0.896, | |
| "step": 11110 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.4022631578947368e-05, | |
| "loss": 0.9684, | |
| "step": 11120 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.4006842105263157e-05, | |
| "loss": 0.9427, | |
| "step": 11130 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.3991052631578949e-05, | |
| "loss": 0.9574, | |
| "step": 11140 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.3975263157894738e-05, | |
| "loss": 0.9848, | |
| "step": 11150 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.3959473684210528e-05, | |
| "loss": 0.9153, | |
| "step": 11160 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.3943684210526317e-05, | |
| "loss": 0.8842, | |
| "step": 11170 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.3927894736842105e-05, | |
| "loss": 0.9311, | |
| "step": 11180 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.3912105263157896e-05, | |
| "loss": 0.9831, | |
| "step": 11190 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.3896315789473684e-05, | |
| "loss": 0.9475, | |
| "step": 11200 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.3880526315789475e-05, | |
| "loss": 1.005, | |
| "step": 11210 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.3864736842105263e-05, | |
| "loss": 0.8998, | |
| "step": 11220 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.3848947368421054e-05, | |
| "loss": 0.9842, | |
| "step": 11230 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.3833157894736842e-05, | |
| "loss": 0.9932, | |
| "step": 11240 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.3817368421052631e-05, | |
| "loss": 0.9029, | |
| "step": 11250 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.3801578947368421e-05, | |
| "loss": 1.0074, | |
| "step": 11260 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 1.378578947368421e-05, | |
| "loss": 1.033, | |
| "step": 11270 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 1.377e-05, | |
| "loss": 0.9659, | |
| "step": 11280 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 1.3754210526315789e-05, | |
| "loss": 0.9908, | |
| "step": 11290 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 1.373842105263158e-05, | |
| "loss": 0.9173, | |
| "step": 11300 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 1.372263157894737e-05, | |
| "loss": 0.946, | |
| "step": 11310 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 1.3706842105263158e-05, | |
| "loss": 0.9739, | |
| "step": 11320 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 1.3691052631578949e-05, | |
| "loss": 0.9435, | |
| "step": 11330 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 1.3675263157894737e-05, | |
| "loss": 0.9885, | |
| "step": 11340 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 1.3659473684210528e-05, | |
| "loss": 0.893, | |
| "step": 11350 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 1.3643684210526316e-05, | |
| "loss": 1.0282, | |
| "step": 11360 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 1.3627894736842107e-05, | |
| "loss": 0.9377, | |
| "step": 11370 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 1.3612105263157895e-05, | |
| "loss": 1.0104, | |
| "step": 11380 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 1.3596315789473684e-05, | |
| "loss": 1.0209, | |
| "step": 11390 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 1.3580526315789474e-05, | |
| "loss": 0.973, | |
| "step": 11400 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 1.3564736842105263e-05, | |
| "loss": 0.9634, | |
| "step": 11410 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 1.3548947368421053e-05, | |
| "loss": 0.9921, | |
| "step": 11420 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 1.3533157894736842e-05, | |
| "loss": 1.0186, | |
| "step": 11430 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 1.3517368421052632e-05, | |
| "loss": 0.9819, | |
| "step": 11440 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 1.350157894736842e-05, | |
| "loss": 0.927, | |
| "step": 11450 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 1.348578947368421e-05, | |
| "loss": 0.9331, | |
| "step": 11460 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 1.3470000000000001e-05, | |
| "loss": 0.9586, | |
| "step": 11470 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 1.345421052631579e-05, | |
| "loss": 0.9382, | |
| "step": 11480 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 1.343842105263158e-05, | |
| "loss": 0.9198, | |
| "step": 11490 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 1.3422631578947369e-05, | |
| "loss": 0.9324, | |
| "step": 11500 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "eval_get_denotation_accuracy": 0.47902869757174393, | |
| "eval_loss": 0.8948884010314941, | |
| "eval_runtime": 232.3445, | |
| "eval_samples_per_second": 17.547, | |
| "eval_steps_per_second": 1.098, | |
| "step": 11500 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 1.340684210526316e-05, | |
| "loss": 0.9529, | |
| "step": 11510 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 1.3391052631578948e-05, | |
| "loss": 0.9848, | |
| "step": 11520 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 1.3375263157894737e-05, | |
| "loss": 0.968, | |
| "step": 11530 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 1.3359473684210527e-05, | |
| "loss": 0.9602, | |
| "step": 11540 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 1.3343684210526316e-05, | |
| "loss": 0.9568, | |
| "step": 11550 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 1.3327894736842106e-05, | |
| "loss": 1.0069, | |
| "step": 11560 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 1.3312105263157895e-05, | |
| "loss": 0.9421, | |
| "step": 11570 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 1.3296315789473685e-05, | |
| "loss": 0.952, | |
| "step": 11580 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 1.3280526315789474e-05, | |
| "loss": 0.9882, | |
| "step": 11590 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 1.3264736842105262e-05, | |
| "loss": 1.028, | |
| "step": 11600 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 1.3248947368421053e-05, | |
| "loss": 0.8531, | |
| "step": 11610 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 1.3233157894736841e-05, | |
| "loss": 1.0165, | |
| "step": 11620 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 1.3217368421052632e-05, | |
| "loss": 0.8926, | |
| "step": 11630 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.3201578947368422e-05, | |
| "loss": 0.9704, | |
| "step": 11640 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.3185789473684212e-05, | |
| "loss": 0.9456, | |
| "step": 11650 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.3170000000000001e-05, | |
| "loss": 0.9441, | |
| "step": 11660 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.315421052631579e-05, | |
| "loss": 0.9143, | |
| "step": 11670 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.313842105263158e-05, | |
| "loss": 0.9443, | |
| "step": 11680 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.3122631578947369e-05, | |
| "loss": 0.9467, | |
| "step": 11690 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.3106842105263159e-05, | |
| "loss": 0.9355, | |
| "step": 11700 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.3091052631578948e-05, | |
| "loss": 0.9463, | |
| "step": 11710 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.3075263157894738e-05, | |
| "loss": 1.0012, | |
| "step": 11720 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.3059473684210527e-05, | |
| "loss": 0.9949, | |
| "step": 11730 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.3043684210526315e-05, | |
| "loss": 0.9865, | |
| "step": 11740 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.3027894736842106e-05, | |
| "loss": 0.935, | |
| "step": 11750 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.3012105263157894e-05, | |
| "loss": 0.9432, | |
| "step": 11760 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.2996315789473685e-05, | |
| "loss": 0.9185, | |
| "step": 11770 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.2980526315789473e-05, | |
| "loss": 0.9587, | |
| "step": 11780 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.2964736842105263e-05, | |
| "loss": 0.9803, | |
| "step": 11790 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.2948947368421052e-05, | |
| "loss": 1.0278, | |
| "step": 11800 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.2933157894736842e-05, | |
| "loss": 0.9842, | |
| "step": 11810 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.2917368421052633e-05, | |
| "loss": 0.8908, | |
| "step": 11820 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.2901578947368421e-05, | |
| "loss": 0.9412, | |
| "step": 11830 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.2885789473684212e-05, | |
| "loss": 0.8609, | |
| "step": 11840 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.287e-05, | |
| "loss": 0.9532, | |
| "step": 11850 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.285421052631579e-05, | |
| "loss": 0.9701, | |
| "step": 11860 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.283842105263158e-05, | |
| "loss": 0.9302, | |
| "step": 11870 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.2822631578947368e-05, | |
| "loss": 1.0146, | |
| "step": 11880 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.2806842105263158e-05, | |
| "loss": 0.9609, | |
| "step": 11890 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.2791052631578947e-05, | |
| "loss": 0.8539, | |
| "step": 11900 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.2775263157894737e-05, | |
| "loss": 1.0032, | |
| "step": 11910 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.2759473684210526e-05, | |
| "loss": 0.9099, | |
| "step": 11920 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.2743684210526316e-05, | |
| "loss": 0.937, | |
| "step": 11930 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.2727894736842105e-05, | |
| "loss": 0.9219, | |
| "step": 11940 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.2712105263157894e-05, | |
| "loss": 0.9502, | |
| "step": 11950 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.2696315789473684e-05, | |
| "loss": 0.9979, | |
| "step": 11960 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.2680526315789474e-05, | |
| "loss": 0.9364, | |
| "step": 11970 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.2664736842105265e-05, | |
| "loss": 0.9401, | |
| "step": 11980 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.2648947368421053e-05, | |
| "loss": 0.9467, | |
| "step": 11990 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.2633157894736844e-05, | |
| "loss": 0.9598, | |
| "step": 12000 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "eval_get_denotation_accuracy": 0.46701005641402993, | |
| "eval_loss": 0.9270404577255249, | |
| "eval_runtime": 235.7621, | |
| "eval_samples_per_second": 17.293, | |
| "eval_steps_per_second": 1.082, | |
| "step": 12000 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 1.2617368421052632e-05, | |
| "loss": 0.9844, | |
| "step": 12010 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 1.2601578947368421e-05, | |
| "loss": 0.9077, | |
| "step": 12020 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 1.2585789473684211e-05, | |
| "loss": 0.9628, | |
| "step": 12030 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 1.257e-05, | |
| "loss": 0.9696, | |
| "step": 12040 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 1.255421052631579e-05, | |
| "loss": 0.9577, | |
| "step": 12050 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 1.2538421052631579e-05, | |
| "loss": 0.9052, | |
| "step": 12060 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 1.252263157894737e-05, | |
| "loss": 0.9735, | |
| "step": 12070 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 1.2506842105263158e-05, | |
| "loss": 0.9496, | |
| "step": 12080 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 1.2491052631578947e-05, | |
| "loss": 0.8934, | |
| "step": 12090 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 1.2475263157894737e-05, | |
| "loss": 0.9669, | |
| "step": 12100 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 1.2459473684210526e-05, | |
| "loss": 0.8894, | |
| "step": 12110 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 1.2443684210526316e-05, | |
| "loss": 0.9323, | |
| "step": 12120 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 1.2427894736842105e-05, | |
| "loss": 0.8627, | |
| "step": 12130 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.2412105263157897e-05, | |
| "loss": 0.9324, | |
| "step": 12140 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.2396315789473685e-05, | |
| "loss": 0.9671, | |
| "step": 12150 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.2380526315789474e-05, | |
| "loss": 0.9995, | |
| "step": 12160 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.2364736842105264e-05, | |
| "loss": 0.9526, | |
| "step": 12170 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.2348947368421053e-05, | |
| "loss": 0.9975, | |
| "step": 12180 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.2333157894736843e-05, | |
| "loss": 0.8614, | |
| "step": 12190 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.2317368421052632e-05, | |
| "loss": 0.9937, | |
| "step": 12200 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.2301578947368422e-05, | |
| "loss": 0.997, | |
| "step": 12210 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.228578947368421e-05, | |
| "loss": 0.9323, | |
| "step": 12220 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.227e-05, | |
| "loss": 0.9201, | |
| "step": 12230 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.225421052631579e-05, | |
| "loss": 1.0284, | |
| "step": 12240 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.2238421052631578e-05, | |
| "loss": 0.9643, | |
| "step": 12250 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.2222631578947369e-05, | |
| "loss": 0.9623, | |
| "step": 12260 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.2206842105263157e-05, | |
| "loss": 0.9669, | |
| "step": 12270 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.2191052631578948e-05, | |
| "loss": 0.9241, | |
| "step": 12280 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.2175263157894736e-05, | |
| "loss": 0.8715, | |
| "step": 12290 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.2159473684210527e-05, | |
| "loss": 0.9781, | |
| "step": 12300 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.2143684210526317e-05, | |
| "loss": 0.8883, | |
| "step": 12310 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.2127894736842106e-05, | |
| "loss": 1.0113, | |
| "step": 12320 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.2112105263157896e-05, | |
| "loss": 0.9311, | |
| "step": 12330 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.2096315789473685e-05, | |
| "loss": 0.8775, | |
| "step": 12340 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.2080526315789475e-05, | |
| "loss": 0.9003, | |
| "step": 12350 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.2064736842105264e-05, | |
| "loss": 0.9946, | |
| "step": 12360 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.2048947368421052e-05, | |
| "loss": 0.954, | |
| "step": 12370 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.2033157894736843e-05, | |
| "loss": 0.9033, | |
| "step": 12380 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 1.2017368421052631e-05, | |
| "loss": 1.013, | |
| "step": 12390 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 1.2001578947368422e-05, | |
| "loss": 0.9724, | |
| "step": 12400 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 1.198578947368421e-05, | |
| "loss": 0.9102, | |
| "step": 12410 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 1.197e-05, | |
| "loss": 0.8131, | |
| "step": 12420 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 1.195421052631579e-05, | |
| "loss": 0.9199, | |
| "step": 12430 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 1.1938421052631578e-05, | |
| "loss": 0.9115, | |
| "step": 12440 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 1.1922631578947368e-05, | |
| "loss": 1.0054, | |
| "step": 12450 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 1.1906842105263157e-05, | |
| "loss": 0.8474, | |
| "step": 12460 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 1.1891052631578949e-05, | |
| "loss": 0.8419, | |
| "step": 12470 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 1.1875263157894738e-05, | |
| "loss": 0.8427, | |
| "step": 12480 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 1.1859473684210528e-05, | |
| "loss": 0.8966, | |
| "step": 12490 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 1.1843684210526317e-05, | |
| "loss": 0.8734, | |
| "step": 12500 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_get_denotation_accuracy": 0.47780230561687514, | |
| "eval_loss": 0.9081824421882629, | |
| "eval_runtime": 227.871, | |
| "eval_samples_per_second": 17.892, | |
| "eval_steps_per_second": 1.119, | |
| "step": 12500 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 1.1827894736842105e-05, | |
| "loss": 0.9143, | |
| "step": 12510 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 1.1812105263157896e-05, | |
| "loss": 0.839, | |
| "step": 12520 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 1.1796315789473684e-05, | |
| "loss": 0.8653, | |
| "step": 12530 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 1.1780526315789474e-05, | |
| "loss": 0.8873, | |
| "step": 12540 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 1.1764736842105263e-05, | |
| "loss": 0.8528, | |
| "step": 12550 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 1.1748947368421053e-05, | |
| "loss": 0.8685, | |
| "step": 12560 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 1.1733157894736842e-05, | |
| "loss": 0.8989, | |
| "step": 12570 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 1.171736842105263e-05, | |
| "loss": 0.8823, | |
| "step": 12580 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 1.1701578947368421e-05, | |
| "loss": 0.9632, | |
| "step": 12590 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 1.168578947368421e-05, | |
| "loss": 0.9019, | |
| "step": 12600 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 1.167e-05, | |
| "loss": 0.9031, | |
| "step": 12610 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 1.1654210526315789e-05, | |
| "loss": 0.8701, | |
| "step": 12620 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 1.1638421052631579e-05, | |
| "loss": 0.9202, | |
| "step": 12630 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 1.162263157894737e-05, | |
| "loss": 0.9147, | |
| "step": 12640 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 1.1606842105263158e-05, | |
| "loss": 0.8666, | |
| "step": 12650 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 1.1591052631578948e-05, | |
| "loss": 0.9134, | |
| "step": 12660 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 1.1575263157894737e-05, | |
| "loss": 0.8758, | |
| "step": 12670 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 1.1559473684210527e-05, | |
| "loss": 0.8481, | |
| "step": 12680 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 1.1543684210526316e-05, | |
| "loss": 0.8792, | |
| "step": 12690 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 1.1527894736842106e-05, | |
| "loss": 0.899, | |
| "step": 12700 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 1.1512105263157895e-05, | |
| "loss": 0.8806, | |
| "step": 12710 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 1.1496315789473685e-05, | |
| "loss": 0.7982, | |
| "step": 12720 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 1.1480526315789474e-05, | |
| "loss": 0.8807, | |
| "step": 12730 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 1.1464736842105263e-05, | |
| "loss": 0.8392, | |
| "step": 12740 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 1.1448947368421053e-05, | |
| "loss": 0.9382, | |
| "step": 12750 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 1.1433157894736842e-05, | |
| "loss": 0.9007, | |
| "step": 12760 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 1.1417368421052632e-05, | |
| "loss": 0.9237, | |
| "step": 12770 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 1.140157894736842e-05, | |
| "loss": 0.8415, | |
| "step": 12780 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 1.1385789473684211e-05, | |
| "loss": 0.8427, | |
| "step": 12790 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 1.137e-05, | |
| "loss": 0.8881, | |
| "step": 12800 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 1.135421052631579e-05, | |
| "loss": 0.8442, | |
| "step": 12810 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 1.133842105263158e-05, | |
| "loss": 0.8685, | |
| "step": 12820 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 1.1322631578947369e-05, | |
| "loss": 0.8822, | |
| "step": 12830 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 1.130684210526316e-05, | |
| "loss": 0.8655, | |
| "step": 12840 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 1.1291052631578948e-05, | |
| "loss": 0.8487, | |
| "step": 12850 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 1.1275263157894738e-05, | |
| "loss": 0.9276, | |
| "step": 12860 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 1.1259473684210527e-05, | |
| "loss": 0.8529, | |
| "step": 12870 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 1.1243684210526315e-05, | |
| "loss": 0.826, | |
| "step": 12880 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 1.1227894736842106e-05, | |
| "loss": 0.8792, | |
| "step": 12890 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 1.1212105263157894e-05, | |
| "loss": 0.8831, | |
| "step": 12900 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 1.1196315789473685e-05, | |
| "loss": 0.9012, | |
| "step": 12910 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 1.1180526315789473e-05, | |
| "loss": 0.8368, | |
| "step": 12920 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 1.1164736842105264e-05, | |
| "loss": 0.8694, | |
| "step": 12930 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 1.1148947368421052e-05, | |
| "loss": 0.9176, | |
| "step": 12940 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 1.1133157894736841e-05, | |
| "loss": 0.8446, | |
| "step": 12950 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 1.1117368421052631e-05, | |
| "loss": 0.8766, | |
| "step": 12960 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 1.1101578947368422e-05, | |
| "loss": 0.8434, | |
| "step": 12970 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 1.1085789473684212e-05, | |
| "loss": 0.8835, | |
| "step": 12980 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 1.107e-05, | |
| "loss": 0.7653, | |
| "step": 12990 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 1.1054210526315791e-05, | |
| "loss": 0.9371, | |
| "step": 13000 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "eval_get_denotation_accuracy": 0.46848172675987243, | |
| "eval_loss": 0.9864445924758911, | |
| "eval_runtime": 239.8054, | |
| "eval_samples_per_second": 17.001, | |
| "eval_steps_per_second": 1.063, | |
| "step": 13000 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 1.103842105263158e-05, | |
| "loss": 0.8836, | |
| "step": 13010 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 1.1022631578947368e-05, | |
| "loss": 0.8513, | |
| "step": 13020 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 1.1006842105263159e-05, | |
| "loss": 0.8696, | |
| "step": 13030 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 1.0991052631578947e-05, | |
| "loss": 0.8958, | |
| "step": 13040 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 1.0975263157894738e-05, | |
| "loss": 0.8015, | |
| "step": 13050 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 1.0959473684210526e-05, | |
| "loss": 0.8608, | |
| "step": 13060 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 1.0943684210526317e-05, | |
| "loss": 0.8399, | |
| "step": 13070 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 1.0927894736842105e-05, | |
| "loss": 0.9047, | |
| "step": 13080 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 1.0912105263157894e-05, | |
| "loss": 0.9121, | |
| "step": 13090 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 1.0896315789473684e-05, | |
| "loss": 0.9045, | |
| "step": 13100 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 1.0880526315789473e-05, | |
| "loss": 0.8474, | |
| "step": 13110 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 1.0864736842105263e-05, | |
| "loss": 0.8228, | |
| "step": 13120 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 1.0848947368421052e-05, | |
| "loss": 0.819, | |
| "step": 13130 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 1.0833157894736844e-05, | |
| "loss": 0.9136, | |
| "step": 13140 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 1.0817368421052633e-05, | |
| "loss": 0.8869, | |
| "step": 13150 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 1.0801578947368421e-05, | |
| "loss": 0.8792, | |
| "step": 13160 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 1.0785789473684212e-05, | |
| "loss": 0.9296, | |
| "step": 13170 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 1.077e-05, | |
| "loss": 0.9138, | |
| "step": 13180 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 1.075421052631579e-05, | |
| "loss": 0.8427, | |
| "step": 13190 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 1.073842105263158e-05, | |
| "loss": 0.8654, | |
| "step": 13200 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 1.072263157894737e-05, | |
| "loss": 0.8661, | |
| "step": 13210 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 1.0706842105263158e-05, | |
| "loss": 0.8933, | |
| "step": 13220 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 1.0691052631578947e-05, | |
| "loss": 0.9463, | |
| "step": 13230 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 1.0675263157894737e-05, | |
| "loss": 0.8524, | |
| "step": 13240 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 1.0659473684210526e-05, | |
| "loss": 0.9295, | |
| "step": 13250 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 1.0643684210526316e-05, | |
| "loss": 0.8591, | |
| "step": 13260 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 1.0627894736842105e-05, | |
| "loss": 0.8902, | |
| "step": 13270 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 1.0612105263157895e-05, | |
| "loss": 0.8687, | |
| "step": 13280 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 1.0596315789473684e-05, | |
| "loss": 0.833, | |
| "step": 13290 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 1.0580526315789474e-05, | |
| "loss": 0.8154, | |
| "step": 13300 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 1.0564736842105264e-05, | |
| "loss": 0.84, | |
| "step": 13310 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 1.0548947368421053e-05, | |
| "loss": 0.8698, | |
| "step": 13320 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 1.0533157894736843e-05, | |
| "loss": 0.8736, | |
| "step": 13330 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 1.0517368421052632e-05, | |
| "loss": 0.8258, | |
| "step": 13340 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 1.0501578947368422e-05, | |
| "loss": 0.8427, | |
| "step": 13350 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 1.0485789473684211e-05, | |
| "loss": 0.8363, | |
| "step": 13360 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 1.047e-05, | |
| "loss": 0.8662, | |
| "step": 13370 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 1.045421052631579e-05, | |
| "loss": 0.8508, | |
| "step": 13380 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 1.0438421052631579e-05, | |
| "loss": 0.868, | |
| "step": 13390 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 1.0422631578947369e-05, | |
| "loss": 0.8478, | |
| "step": 13400 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 1.0406842105263158e-05, | |
| "loss": 0.8924, | |
| "step": 13410 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 1.0391052631578948e-05, | |
| "loss": 0.8992, | |
| "step": 13420 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 1.0375263157894737e-05, | |
| "loss": 0.9264, | |
| "step": 13430 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 1.0359473684210525e-05, | |
| "loss": 0.9336, | |
| "step": 13440 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 1.0343684210526316e-05, | |
| "loss": 0.83, | |
| "step": 13450 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 1.0327894736842104e-05, | |
| "loss": 0.8183, | |
| "step": 13460 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 1.0312105263157896e-05, | |
| "loss": 0.903, | |
| "step": 13470 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 1.0296315789473685e-05, | |
| "loss": 0.8715, | |
| "step": 13480 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 1.0280526315789475e-05, | |
| "loss": 0.8773, | |
| "step": 13490 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 1.0264736842105264e-05, | |
| "loss": 0.8945, | |
| "step": 13500 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "eval_get_denotation_accuracy": 0.47657591366200636, | |
| "eval_loss": 0.9642069339752197, | |
| "eval_runtime": 237.1275, | |
| "eval_samples_per_second": 17.193, | |
| "eval_steps_per_second": 1.075, | |
| "step": 13500 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 1.0248947368421053e-05, | |
| "loss": 0.8971, | |
| "step": 13510 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 1.0233157894736843e-05, | |
| "loss": 0.8717, | |
| "step": 13520 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 1.0217368421052632e-05, | |
| "loss": 0.9304, | |
| "step": 13530 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 1.0201578947368422e-05, | |
| "loss": 0.974, | |
| "step": 13540 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 1.018578947368421e-05, | |
| "loss": 0.8908, | |
| "step": 13550 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 1.0170000000000001e-05, | |
| "loss": 0.8936, | |
| "step": 13560 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 1.015421052631579e-05, | |
| "loss": 0.8176, | |
| "step": 13570 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 1.0138421052631578e-05, | |
| "loss": 0.839, | |
| "step": 13580 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 1.0122631578947369e-05, | |
| "loss": 0.9415, | |
| "step": 13590 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 1.0106842105263157e-05, | |
| "loss": 0.8843, | |
| "step": 13600 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 1.0091052631578947e-05, | |
| "loss": 0.877, | |
| "step": 13610 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 1.0075263157894736e-05, | |
| "loss": 0.9242, | |
| "step": 13620 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 1.0059473684210526e-05, | |
| "loss": 0.9133, | |
| "step": 13630 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 1.0043684210526317e-05, | |
| "loss": 0.9046, | |
| "step": 13640 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 1.0027894736842105e-05, | |
| "loss": 0.9339, | |
| "step": 13650 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 1.0012105263157896e-05, | |
| "loss": 0.8824, | |
| "step": 13660 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 9.996315789473684e-06, | |
| "loss": 0.8547, | |
| "step": 13670 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 9.980526315789475e-06, | |
| "loss": 0.8779, | |
| "step": 13680 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 9.964736842105263e-06, | |
| "loss": 0.8072, | |
| "step": 13690 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 9.948947368421054e-06, | |
| "loss": 0.927, | |
| "step": 13700 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 9.933157894736842e-06, | |
| "loss": 0.8547, | |
| "step": 13710 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 9.917368421052631e-06, | |
| "loss": 0.9088, | |
| "step": 13720 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 9.901578947368421e-06, | |
| "loss": 0.9095, | |
| "step": 13730 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 9.88578947368421e-06, | |
| "loss": 0.8842, | |
| "step": 13740 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 9.87e-06, | |
| "loss": 0.8726, | |
| "step": 13750 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 9.854210526315789e-06, | |
| "loss": 0.8749, | |
| "step": 13760 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 9.83842105263158e-06, | |
| "loss": 0.8684, | |
| "step": 13770 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 9.822631578947368e-06, | |
| "loss": 0.9212, | |
| "step": 13780 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 9.806842105263157e-06, | |
| "loss": 0.8891, | |
| "step": 13790 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 9.791052631578947e-06, | |
| "loss": 0.7833, | |
| "step": 13800 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 9.775263157894737e-06, | |
| "loss": 0.9201, | |
| "step": 13810 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 9.759473684210528e-06, | |
| "loss": 0.8358, | |
| "step": 13820 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 9.743684210526316e-06, | |
| "loss": 0.8905, | |
| "step": 13830 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 9.727894736842107e-06, | |
| "loss": 0.8893, | |
| "step": 13840 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 9.712105263157895e-06, | |
| "loss": 0.853, | |
| "step": 13850 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 9.696315789473684e-06, | |
| "loss": 0.9108, | |
| "step": 13860 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 9.680526315789474e-06, | |
| "loss": 0.8194, | |
| "step": 13870 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 9.664736842105263e-06, | |
| "loss": 0.8473, | |
| "step": 13880 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 9.648947368421053e-06, | |
| "loss": 0.8879, | |
| "step": 13890 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 9.633157894736842e-06, | |
| "loss": 0.8999, | |
| "step": 13900 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 9.617368421052632e-06, | |
| "loss": 0.9429, | |
| "step": 13910 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 9.601578947368421e-06, | |
| "loss": 0.8843, | |
| "step": 13920 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 9.58578947368421e-06, | |
| "loss": 0.8704, | |
| "step": 13930 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 9.57e-06, | |
| "loss": 0.9064, | |
| "step": 13940 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 9.554210526315788e-06, | |
| "loss": 0.8679, | |
| "step": 13950 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 9.538421052631579e-06, | |
| "loss": 0.8678, | |
| "step": 13960 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 9.52263157894737e-06, | |
| "loss": 0.9391, | |
| "step": 13970 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 9.50684210526316e-06, | |
| "loss": 0.8387, | |
| "step": 13980 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 9.491052631578948e-06, | |
| "loss": 0.8408, | |
| "step": 13990 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 9.475263157894737e-06, | |
| "loss": 0.9746, | |
| "step": 14000 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "eval_get_denotation_accuracy": 0.4736325729703213, | |
| "eval_loss": 0.96991366147995, | |
| "eval_runtime": 250.0702, | |
| "eval_samples_per_second": 16.303, | |
| "eval_steps_per_second": 1.02, | |
| "step": 14000 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 9.459473684210527e-06, | |
| "loss": 0.8949, | |
| "step": 14010 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 9.443684210526316e-06, | |
| "loss": 0.8376, | |
| "step": 14020 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 9.427894736842106e-06, | |
| "loss": 0.848, | |
| "step": 14030 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 9.412105263157895e-06, | |
| "loss": 0.8501, | |
| "step": 14040 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 9.396315789473685e-06, | |
| "loss": 0.8243, | |
| "step": 14050 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 9.380526315789474e-06, | |
| "loss": 0.8642, | |
| "step": 14060 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 9.364736842105262e-06, | |
| "loss": 0.8485, | |
| "step": 14070 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 9.348947368421053e-06, | |
| "loss": 0.9008, | |
| "step": 14080 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 9.333157894736841e-06, | |
| "loss": 0.8867, | |
| "step": 14090 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 9.317368421052632e-06, | |
| "loss": 0.9288, | |
| "step": 14100 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 9.30157894736842e-06, | |
| "loss": 0.7863, | |
| "step": 14110 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 9.28578947368421e-06, | |
| "loss": 0.9072, | |
| "step": 14120 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 9.27e-06, | |
| "loss": 0.8691, | |
| "step": 14130 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 9.25421052631579e-06, | |
| "loss": 0.8931, | |
| "step": 14140 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 9.23842105263158e-06, | |
| "loss": 0.9736, | |
| "step": 14150 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 9.222631578947369e-06, | |
| "loss": 0.8218, | |
| "step": 14160 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 9.206842105263159e-06, | |
| "loss": 0.9309, | |
| "step": 14170 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 9.191052631578948e-06, | |
| "loss": 0.8458, | |
| "step": 14180 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 9.175263157894738e-06, | |
| "loss": 0.8162, | |
| "step": 14190 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 9.159473684210527e-06, | |
| "loss": 0.8747, | |
| "step": 14200 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 9.143684210526315e-06, | |
| "loss": 0.8609, | |
| "step": 14210 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 9.127894736842106e-06, | |
| "loss": 0.8826, | |
| "step": 14220 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 9.112105263157894e-06, | |
| "loss": 0.8272, | |
| "step": 14230 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 9.096315789473685e-06, | |
| "loss": 0.8725, | |
| "step": 14240 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 9.080526315789473e-06, | |
| "loss": 0.8752, | |
| "step": 14250 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 9.064736842105264e-06, | |
| "loss": 0.8297, | |
| "step": 14260 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 9.048947368421052e-06, | |
| "loss": 0.9066, | |
| "step": 14270 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 9.033157894736841e-06, | |
| "loss": 0.8478, | |
| "step": 14280 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 9.017368421052631e-06, | |
| "loss": 0.8775, | |
| "step": 14290 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 9.001578947368422e-06, | |
| "loss": 0.8973, | |
| "step": 14300 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 8.985789473684212e-06, | |
| "loss": 0.842, | |
| "step": 14310 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 8.97e-06, | |
| "loss": 0.8286, | |
| "step": 14320 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 8.954210526315791e-06, | |
| "loss": 0.8781, | |
| "step": 14330 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 8.93842105263158e-06, | |
| "loss": 0.8913, | |
| "step": 14340 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 8.92263157894737e-06, | |
| "loss": 0.8063, | |
| "step": 14350 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 8.906842105263158e-06, | |
| "loss": 0.8941, | |
| "step": 14360 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 8.891052631578947e-06, | |
| "loss": 0.8975, | |
| "step": 14370 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 8.875263157894737e-06, | |
| "loss": 0.8663, | |
| "step": 14380 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 8.859473684210526e-06, | |
| "loss": 0.8858, | |
| "step": 14390 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 8.843684210526316e-06, | |
| "loss": 0.9231, | |
| "step": 14400 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 8.827894736842105e-06, | |
| "loss": 0.8422, | |
| "step": 14410 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 8.812105263157895e-06, | |
| "loss": 0.8881, | |
| "step": 14420 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 8.796315789473684e-06, | |
| "loss": 0.8195, | |
| "step": 14430 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 8.780526315789473e-06, | |
| "loss": 0.899, | |
| "step": 14440 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 8.764736842105263e-06, | |
| "loss": 0.8601, | |
| "step": 14450 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 8.748947368421052e-06, | |
| "loss": 0.9079, | |
| "step": 14460 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 8.733157894736844e-06, | |
| "loss": 0.8702, | |
| "step": 14470 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 8.717368421052632e-06, | |
| "loss": 0.8858, | |
| "step": 14480 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 8.701578947368423e-06, | |
| "loss": 0.9535, | |
| "step": 14490 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 8.685789473684211e-06, | |
| "loss": 0.8338, | |
| "step": 14500 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "eval_get_denotation_accuracy": 0.4780475840078489, | |
| "eval_loss": 0.9859474301338196, | |
| "eval_runtime": 240.5029, | |
| "eval_samples_per_second": 16.952, | |
| "eval_steps_per_second": 1.06, | |
| "step": 14500 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 8.67e-06, | |
| "loss": 0.822, | |
| "step": 14510 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 8.65421052631579e-06, | |
| "loss": 0.8478, | |
| "step": 14520 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 8.638421052631579e-06, | |
| "loss": 0.86, | |
| "step": 14530 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 8.62263157894737e-06, | |
| "loss": 0.9104, | |
| "step": 14540 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 8.606842105263158e-06, | |
| "loss": 0.9088, | |
| "step": 14550 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 8.591052631578948e-06, | |
| "loss": 0.9169, | |
| "step": 14560 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 8.575263157894737e-06, | |
| "loss": 0.8256, | |
| "step": 14570 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 8.559473684210526e-06, | |
| "loss": 0.9072, | |
| "step": 14580 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 8.543684210526316e-06, | |
| "loss": 0.9258, | |
| "step": 14590 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 8.527894736842105e-06, | |
| "loss": 0.8833, | |
| "step": 14600 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 8.512105263157895e-06, | |
| "loss": 0.7958, | |
| "step": 14610 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 8.496315789473684e-06, | |
| "loss": 0.8463, | |
| "step": 14620 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 8.480526315789474e-06, | |
| "loss": 0.8489, | |
| "step": 14630 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 8.464736842105264e-06, | |
| "loss": 0.8576, | |
| "step": 14640 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 8.448947368421053e-06, | |
| "loss": 0.8827, | |
| "step": 14650 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 8.433157894736843e-06, | |
| "loss": 0.9005, | |
| "step": 14660 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 8.417368421052632e-06, | |
| "loss": 0.932, | |
| "step": 14670 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 8.401578947368422e-06, | |
| "loss": 0.8623, | |
| "step": 14680 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 8.38578947368421e-06, | |
| "loss": 0.9182, | |
| "step": 14690 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 8.370000000000001e-06, | |
| "loss": 0.8711, | |
| "step": 14700 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 8.35421052631579e-06, | |
| "loss": 0.8164, | |
| "step": 14710 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 8.338421052631578e-06, | |
| "loss": 0.8631, | |
| "step": 14720 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 8.322631578947369e-06, | |
| "loss": 0.9048, | |
| "step": 14730 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 8.306842105263157e-06, | |
| "loss": 0.9186, | |
| "step": 14740 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 8.291052631578948e-06, | |
| "loss": 0.9031, | |
| "step": 14750 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 8.275263157894736e-06, | |
| "loss": 0.8455, | |
| "step": 14760 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 8.259473684210527e-06, | |
| "loss": 0.8546, | |
| "step": 14770 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 8.243684210526315e-06, | |
| "loss": 0.833, | |
| "step": 14780 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 8.227894736842104e-06, | |
| "loss": 0.9135, | |
| "step": 14790 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 8.212105263157894e-06, | |
| "loss": 0.8699, | |
| "step": 14800 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 8.196315789473685e-06, | |
| "loss": 0.8699, | |
| "step": 14810 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 8.180526315789475e-06, | |
| "loss": 0.8913, | |
| "step": 14820 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 8.164736842105264e-06, | |
| "loss": 0.9578, | |
| "step": 14830 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 8.148947368421054e-06, | |
| "loss": 0.874, | |
| "step": 14840 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 8.133157894736843e-06, | |
| "loss": 0.9424, | |
| "step": 14850 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 8.117368421052631e-06, | |
| "loss": 0.8701, | |
| "step": 14860 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 8.101578947368422e-06, | |
| "loss": 0.7962, | |
| "step": 14870 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "learning_rate": 8.08578947368421e-06, | |
| "loss": 0.8808, | |
| "step": 14880 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "learning_rate": 8.07e-06, | |
| "loss": 0.8857, | |
| "step": 14890 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "learning_rate": 8.05421052631579e-06, | |
| "loss": 0.9239, | |
| "step": 14900 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "learning_rate": 8.03842105263158e-06, | |
| "loss": 0.9216, | |
| "step": 14910 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "learning_rate": 8.022631578947368e-06, | |
| "loss": 0.8288, | |
| "step": 14920 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "learning_rate": 8.006842105263157e-06, | |
| "loss": 0.868, | |
| "step": 14930 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "learning_rate": 7.991052631578947e-06, | |
| "loss": 0.8339, | |
| "step": 14940 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "learning_rate": 7.975263157894736e-06, | |
| "loss": 0.9501, | |
| "step": 14950 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "learning_rate": 7.959473684210526e-06, | |
| "loss": 0.9601, | |
| "step": 14960 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "learning_rate": 7.943684210526317e-06, | |
| "loss": 0.8391, | |
| "step": 14970 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "learning_rate": 7.927894736842107e-06, | |
| "loss": 0.8242, | |
| "step": 14980 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "learning_rate": 7.912105263157896e-06, | |
| "loss": 0.8619, | |
| "step": 14990 | |
| }, | |
| { | |
| "epoch": 1.21, | |
| "learning_rate": 7.896315789473684e-06, | |
| "loss": 0.9129, | |
| "step": 15000 | |
| }, | |
| { | |
| "epoch": 1.21, | |
| "eval_get_denotation_accuracy": 0.4890851116016679, | |
| "eval_loss": 0.9386902451515198, | |
| "eval_runtime": 230.2048, | |
| "eval_samples_per_second": 17.71, | |
| "eval_steps_per_second": 1.108, | |
| "step": 15000 | |
| } | |
| ], | |
| "max_steps": 20000, | |
| "num_train_epochs": 2, | |
| "total_flos": 5.687599974087393e+18, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |