| { |
| "best_global_step": null, |
| "best_metric": null, |
| "best_model_checkpoint": null, |
| "epoch": 7.0, |
| "eval_steps": 500, |
| "global_step": 4004, |
| "is_hyper_param_search": false, |
| "is_local_process_zero": true, |
| "is_world_process_zero": true, |
| "log_history": [ |
| { |
| "epoch": 0.008741258741258742, |
| "grad_norm": 6.891807865886709, |
| "learning_rate": 3.990024937655861e-07, |
| "loss": 0.7806, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.688166618347168, |
| "step": 5, |
| "valid_targets_mean": 4434.8, |
| "valid_targets_min": 1164 |
| }, |
| { |
| "epoch": 0.017482517482517484, |
| "grad_norm": 10.07235768114182, |
| "learning_rate": 8.977556109725687e-07, |
| "loss": 0.7949, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.6993734836578369, |
| "step": 10, |
| "valid_targets_mean": 2489.4, |
| "valid_targets_min": 760 |
| }, |
| { |
| "epoch": 0.026223776223776224, |
| "grad_norm": 9.371443623670519, |
| "learning_rate": 1.396508728179551e-06, |
| "loss": 0.8104, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.8695156574249268, |
| "step": 15, |
| "valid_targets_mean": 3041.5, |
| "valid_targets_min": 1109 |
| }, |
| { |
| "epoch": 0.03496503496503497, |
| "grad_norm": 5.8828558878845, |
| "learning_rate": 1.8952618453865338e-06, |
| "loss": 0.7645, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.820339024066925, |
| "step": 20, |
| "valid_targets_mean": 5423.1, |
| "valid_targets_min": 1087 |
| }, |
| { |
| "epoch": 0.043706293706293704, |
| "grad_norm": 4.921738857282242, |
| "learning_rate": 2.3940149625935166e-06, |
| "loss": 0.7558, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.75657057762146, |
| "step": 25, |
| "valid_targets_mean": 4129.1, |
| "valid_targets_min": 1235 |
| }, |
| { |
| "epoch": 0.05244755244755245, |
| "grad_norm": 3.1722389939987297, |
| "learning_rate": 2.892768079800499e-06, |
| "loss": 0.7613, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.7961629033088684, |
| "step": 30, |
| "valid_targets_mean": 5228.2, |
| "valid_targets_min": 1258 |
| }, |
| { |
| "epoch": 0.06118881118881119, |
| "grad_norm": 2.483826973866041, |
| "learning_rate": 3.391521197007482e-06, |
| "loss": 0.6789, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.6847758293151855, |
| "step": 35, |
| "valid_targets_mean": 3565.4, |
| "valid_targets_min": 1135 |
| }, |
| { |
| "epoch": 0.06993006993006994, |
| "grad_norm": 1.7214635926067359, |
| "learning_rate": 3.890274314214464e-06, |
| "loss": 0.6961, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.6839813590049744, |
| "step": 40, |
| "valid_targets_mean": 4050.9, |
| "valid_targets_min": 1743 |
| }, |
| { |
| "epoch": 0.07867132867132867, |
| "grad_norm": 1.4291208613040705, |
| "learning_rate": 4.389027431421446e-06, |
| "loss": 0.6516, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.7264044284820557, |
| "step": 45, |
| "valid_targets_mean": 4053.4, |
| "valid_targets_min": 795 |
| }, |
| { |
| "epoch": 0.08741258741258741, |
| "grad_norm": 1.1744063794805446, |
| "learning_rate": 4.887780548628429e-06, |
| "loss": 0.6135, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5638917684555054, |
| "step": 50, |
| "valid_targets_mean": 2825.8, |
| "valid_targets_min": 1378 |
| }, |
| { |
| "epoch": 0.09615384615384616, |
| "grad_norm": 1.27848851631508, |
| "learning_rate": 5.386533665835411e-06, |
| "loss": 0.6698, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.8109359741210938, |
| "step": 55, |
| "valid_targets_mean": 6348.3, |
| "valid_targets_min": 1256 |
| }, |
| { |
| "epoch": 0.1048951048951049, |
| "grad_norm": 0.880099323037957, |
| "learning_rate": 5.885286783042394e-06, |
| "loss": 0.6445, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.6418143510818481, |
| "step": 60, |
| "valid_targets_mean": 5170.8, |
| "valid_targets_min": 775 |
| }, |
| { |
| "epoch": 0.11363636363636363, |
| "grad_norm": 0.8423522806449167, |
| "learning_rate": 6.384039900249377e-06, |
| "loss": 0.5967, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.6989394426345825, |
| "step": 65, |
| "valid_targets_mean": 4292.8, |
| "valid_targets_min": 1137 |
| }, |
| { |
| "epoch": 0.12237762237762238, |
| "grad_norm": 0.8774344604276414, |
| "learning_rate": 6.882793017456359e-06, |
| "loss": 0.5712, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5156315565109253, |
| "step": 70, |
| "valid_targets_mean": 3363.1, |
| "valid_targets_min": 1000 |
| }, |
| { |
| "epoch": 0.13111888111888112, |
| "grad_norm": 0.7471525308366247, |
| "learning_rate": 7.381546134663342e-06, |
| "loss": 0.5889, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.524969220161438, |
| "step": 75, |
| "valid_targets_mean": 3267.2, |
| "valid_targets_min": 437 |
| }, |
| { |
| "epoch": 0.13986013986013987, |
| "grad_norm": 0.6351105968482974, |
| "learning_rate": 7.880299251870324e-06, |
| "loss": 0.5783, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5628579258918762, |
| "step": 80, |
| "valid_targets_mean": 4445.7, |
| "valid_targets_min": 1269 |
| }, |
| { |
| "epoch": 0.1486013986013986, |
| "grad_norm": 0.8435486850711871, |
| "learning_rate": 8.379052369077308e-06, |
| "loss": 0.5584, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5862073302268982, |
| "step": 85, |
| "valid_targets_mean": 3406.3, |
| "valid_targets_min": 1195 |
| }, |
| { |
| "epoch": 0.15734265734265734, |
| "grad_norm": 0.7027259715932884, |
| "learning_rate": 8.87780548628429e-06, |
| "loss": 0.5716, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5122252702713013, |
| "step": 90, |
| "valid_targets_mean": 3020.4, |
| "valid_targets_min": 816 |
| }, |
| { |
| "epoch": 0.1660839160839161, |
| "grad_norm": 0.7143768069579088, |
| "learning_rate": 9.376558603491272e-06, |
| "loss": 0.5652, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5643905401229858, |
| "step": 95, |
| "valid_targets_mean": 3280.6, |
| "valid_targets_min": 1458 |
| }, |
| { |
| "epoch": 0.17482517482517482, |
| "grad_norm": 0.6466673146750979, |
| "learning_rate": 9.875311720698254e-06, |
| "loss": 0.5271, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5161752700805664, |
| "step": 100, |
| "valid_targets_mean": 3466.4, |
| "valid_targets_min": 785 |
| }, |
| { |
| "epoch": 0.18356643356643357, |
| "grad_norm": 0.660607431222328, |
| "learning_rate": 1.0374064837905238e-05, |
| "loss": 0.5565, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.49148479104042053, |
| "step": 105, |
| "valid_targets_mean": 2937.3, |
| "valid_targets_min": 967 |
| }, |
| { |
| "epoch": 0.19230769230769232, |
| "grad_norm": 0.6418923782026298, |
| "learning_rate": 1.087281795511222e-05, |
| "loss": 0.4982, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.53751540184021, |
| "step": 110, |
| "valid_targets_mean": 4915.1, |
| "valid_targets_min": 753 |
| }, |
| { |
| "epoch": 0.20104895104895104, |
| "grad_norm": 0.5772065562262413, |
| "learning_rate": 1.1371571072319202e-05, |
| "loss": 0.4941, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5409708023071289, |
| "step": 115, |
| "valid_targets_mean": 4977.4, |
| "valid_targets_min": 1187 |
| }, |
| { |
| "epoch": 0.2097902097902098, |
| "grad_norm": 0.6801379276243148, |
| "learning_rate": 1.1870324189526186e-05, |
| "loss": 0.4966, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5502436757087708, |
| "step": 120, |
| "valid_targets_mean": 4482.5, |
| "valid_targets_min": 771 |
| }, |
| { |
| "epoch": 0.21853146853146854, |
| "grad_norm": 0.6133787753502925, |
| "learning_rate": 1.2369077306733168e-05, |
| "loss": 0.5088, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.45697876811027527, |
| "step": 125, |
| "valid_targets_mean": 3453.1, |
| "valid_targets_min": 1168 |
| }, |
| { |
| "epoch": 0.22727272727272727, |
| "grad_norm": 0.7592541102494601, |
| "learning_rate": 1.286783042394015e-05, |
| "loss": 0.4966, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4945298135280609, |
| "step": 130, |
| "valid_targets_mean": 5048.1, |
| "valid_targets_min": 1529 |
| }, |
| { |
| "epoch": 0.23601398601398602, |
| "grad_norm": 0.6366182167764396, |
| "learning_rate": 1.3366583541147134e-05, |
| "loss": 0.4939, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5059365034103394, |
| "step": 135, |
| "valid_targets_mean": 4097.1, |
| "valid_targets_min": 1112 |
| }, |
| { |
| "epoch": 0.24475524475524477, |
| "grad_norm": 0.6564205667008085, |
| "learning_rate": 1.3865336658354116e-05, |
| "loss": 0.4725, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.44329914450645447, |
| "step": 140, |
| "valid_targets_mean": 2867.4, |
| "valid_targets_min": 1213 |
| }, |
| { |
| "epoch": 0.2534965034965035, |
| "grad_norm": 0.6724554034077109, |
| "learning_rate": 1.4364089775561098e-05, |
| "loss": 0.538, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4999472200870514, |
| "step": 145, |
| "valid_targets_mean": 4216.0, |
| "valid_targets_min": 1142 |
| }, |
| { |
| "epoch": 0.26223776223776224, |
| "grad_norm": 0.7059030471238626, |
| "learning_rate": 1.486284289276808e-05, |
| "loss": 0.4688, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.48053932189941406, |
| "step": 150, |
| "valid_targets_mean": 2904.5, |
| "valid_targets_min": 954 |
| }, |
| { |
| "epoch": 0.270979020979021, |
| "grad_norm": 0.7125138132476857, |
| "learning_rate": 1.5361596009975064e-05, |
| "loss": 0.499, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.581404983997345, |
| "step": 155, |
| "valid_targets_mean": 4037.4, |
| "valid_targets_min": 428 |
| }, |
| { |
| "epoch": 0.27972027972027974, |
| "grad_norm": 0.728117797766979, |
| "learning_rate": 1.5860349127182046e-05, |
| "loss": 0.4432, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4301885962486267, |
| "step": 160, |
| "valid_targets_mean": 2771.1, |
| "valid_targets_min": 1129 |
| }, |
| { |
| "epoch": 0.28846153846153844, |
| "grad_norm": 0.715988515093435, |
| "learning_rate": 1.635910224438903e-05, |
| "loss": 0.4658, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4865019917488098, |
| "step": 165, |
| "valid_targets_mean": 3270.8, |
| "valid_targets_min": 783 |
| }, |
| { |
| "epoch": 0.2972027972027972, |
| "grad_norm": 0.6269860955107454, |
| "learning_rate": 1.685785536159601e-05, |
| "loss": 0.4362, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4705418646335602, |
| "step": 170, |
| "valid_targets_mean": 4284.6, |
| "valid_targets_min": 1122 |
| }, |
| { |
| "epoch": 0.30594405594405594, |
| "grad_norm": 0.5592398328135262, |
| "learning_rate": 1.7356608478802993e-05, |
| "loss": 0.4659, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3761853873729706, |
| "step": 175, |
| "valid_targets_mean": 3507.1, |
| "valid_targets_min": 711 |
| }, |
| { |
| "epoch": 0.3146853146853147, |
| "grad_norm": 0.6467612787611225, |
| "learning_rate": 1.7855361596009978e-05, |
| "loss": 0.4857, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5388031005859375, |
| "step": 180, |
| "valid_targets_mean": 4472.5, |
| "valid_targets_min": 1251 |
| }, |
| { |
| "epoch": 0.32342657342657344, |
| "grad_norm": 0.7059874120353821, |
| "learning_rate": 1.835411471321696e-05, |
| "loss": 0.4413, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4448562562465668, |
| "step": 185, |
| "valid_targets_mean": 3762.6, |
| "valid_targets_min": 317 |
| }, |
| { |
| "epoch": 0.3321678321678322, |
| "grad_norm": 0.647197969313476, |
| "learning_rate": 1.8852867830423942e-05, |
| "loss": 0.4846, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5174081921577454, |
| "step": 190, |
| "valid_targets_mean": 4593.6, |
| "valid_targets_min": 848 |
| }, |
| { |
| "epoch": 0.3409090909090909, |
| "grad_norm": 0.758407351151443, |
| "learning_rate": 1.9351620947630925e-05, |
| "loss": 0.4638, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4045215845108032, |
| "step": 195, |
| "valid_targets_mean": 2296.8, |
| "valid_targets_min": 870 |
| }, |
| { |
| "epoch": 0.34965034965034963, |
| "grad_norm": 0.6647262744061739, |
| "learning_rate": 1.9850374064837907e-05, |
| "loss": 0.452, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4018682539463043, |
| "step": 200, |
| "valid_targets_mean": 3484.7, |
| "valid_targets_min": 1374 |
| }, |
| { |
| "epoch": 0.3583916083916084, |
| "grad_norm": 0.8052379949927564, |
| "learning_rate": 2.034912718204489e-05, |
| "loss": 0.5022, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4757809638977051, |
| "step": 205, |
| "valid_targets_mean": 2731.2, |
| "valid_targets_min": 1124 |
| }, |
| { |
| "epoch": 0.36713286713286714, |
| "grad_norm": 0.5730191757062828, |
| "learning_rate": 2.084788029925187e-05, |
| "loss": 0.4553, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.49919718503952026, |
| "step": 210, |
| "valid_targets_mean": 5203.2, |
| "valid_targets_min": 1011 |
| }, |
| { |
| "epoch": 0.3758741258741259, |
| "grad_norm": 0.6122316092766614, |
| "learning_rate": 2.1346633416458853e-05, |
| "loss": 0.461, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5063073635101318, |
| "step": 215, |
| "valid_targets_mean": 6397.0, |
| "valid_targets_min": 1352 |
| }, |
| { |
| "epoch": 0.38461538461538464, |
| "grad_norm": 0.7049095990494619, |
| "learning_rate": 2.1845386533665835e-05, |
| "loss": 0.416, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5339525938034058, |
| "step": 220, |
| "valid_targets_mean": 4158.1, |
| "valid_targets_min": 823 |
| }, |
| { |
| "epoch": 0.39335664335664333, |
| "grad_norm": 0.698438091999467, |
| "learning_rate": 2.2344139650872817e-05, |
| "loss": 0.4574, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.47370338439941406, |
| "step": 225, |
| "valid_targets_mean": 3506.1, |
| "valid_targets_min": 1179 |
| }, |
| { |
| "epoch": 0.4020979020979021, |
| "grad_norm": 0.7054393747626264, |
| "learning_rate": 2.28428927680798e-05, |
| "loss": 0.4607, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4713825285434723, |
| "step": 230, |
| "valid_targets_mean": 3467.7, |
| "valid_targets_min": 769 |
| }, |
| { |
| "epoch": 0.41083916083916083, |
| "grad_norm": 0.6398523387870552, |
| "learning_rate": 2.3341645885286785e-05, |
| "loss": 0.4693, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.49769261479377747, |
| "step": 235, |
| "valid_targets_mean": 5616.6, |
| "valid_targets_min": 1339 |
| }, |
| { |
| "epoch": 0.4195804195804196, |
| "grad_norm": 0.5779595979100703, |
| "learning_rate": 2.3840399002493767e-05, |
| "loss": 0.4295, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.42991095781326294, |
| "step": 240, |
| "valid_targets_mean": 4796.9, |
| "valid_targets_min": 1399 |
| }, |
| { |
| "epoch": 0.42832167832167833, |
| "grad_norm": 0.7018424514021866, |
| "learning_rate": 2.433915211970075e-05, |
| "loss": 0.4287, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.416814386844635, |
| "step": 245, |
| "valid_targets_mean": 3066.3, |
| "valid_targets_min": 1091 |
| }, |
| { |
| "epoch": 0.4370629370629371, |
| "grad_norm": 0.7070083860196137, |
| "learning_rate": 2.483790523690773e-05, |
| "loss": 0.4089, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.42583611607551575, |
| "step": 250, |
| "valid_targets_mean": 3095.8, |
| "valid_targets_min": 836 |
| }, |
| { |
| "epoch": 0.4458041958041958, |
| "grad_norm": 0.6542255655325461, |
| "learning_rate": 2.5336658354114713e-05, |
| "loss": 0.4766, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4974190890789032, |
| "step": 255, |
| "valid_targets_mean": 4516.5, |
| "valid_targets_min": 1431 |
| }, |
| { |
| "epoch": 0.45454545454545453, |
| "grad_norm": 0.6057280619187766, |
| "learning_rate": 2.5835411471321695e-05, |
| "loss": 0.4346, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.494473934173584, |
| "step": 260, |
| "valid_targets_mean": 4743.2, |
| "valid_targets_min": 1768 |
| }, |
| { |
| "epoch": 0.4632867132867133, |
| "grad_norm": 0.6576098745294991, |
| "learning_rate": 2.6334164588528677e-05, |
| "loss": 0.4003, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.39938753843307495, |
| "step": 265, |
| "valid_targets_mean": 3158.4, |
| "valid_targets_min": 1073 |
| }, |
| { |
| "epoch": 0.47202797202797203, |
| "grad_norm": 0.6101915400033449, |
| "learning_rate": 2.6832917705735663e-05, |
| "loss": 0.4526, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4094006419181824, |
| "step": 270, |
| "valid_targets_mean": 4226.7, |
| "valid_targets_min": 1197 |
| }, |
| { |
| "epoch": 0.4807692307692308, |
| "grad_norm": 0.692943572411887, |
| "learning_rate": 2.7331670822942645e-05, |
| "loss": 0.4704, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.42553043365478516, |
| "step": 275, |
| "valid_targets_mean": 3468.1, |
| "valid_targets_min": 866 |
| }, |
| { |
| "epoch": 0.48951048951048953, |
| "grad_norm": 0.625051535199143, |
| "learning_rate": 2.7830423940149627e-05, |
| "loss": 0.4102, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4416755139827728, |
| "step": 280, |
| "valid_targets_mean": 4131.4, |
| "valid_targets_min": 274 |
| }, |
| { |
| "epoch": 0.4982517482517482, |
| "grad_norm": 0.6202235346478298, |
| "learning_rate": 2.832917705735661e-05, |
| "loss": 0.4464, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.38613349199295044, |
| "step": 285, |
| "valid_targets_mean": 3556.7, |
| "valid_targets_min": 1975 |
| }, |
| { |
| "epoch": 0.506993006993007, |
| "grad_norm": 0.6373806668498091, |
| "learning_rate": 2.882793017456359e-05, |
| "loss": 0.4648, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3658190369606018, |
| "step": 290, |
| "valid_targets_mean": 3424.1, |
| "valid_targets_min": 985 |
| }, |
| { |
| "epoch": 0.5157342657342657, |
| "grad_norm": 0.6404836592194673, |
| "learning_rate": 2.9326683291770573e-05, |
| "loss": 0.3999, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4423462748527527, |
| "step": 295, |
| "valid_targets_mean": 3966.9, |
| "valid_targets_min": 699 |
| }, |
| { |
| "epoch": 0.5244755244755245, |
| "grad_norm": 0.6065384985472947, |
| "learning_rate": 2.982543640897756e-05, |
| "loss": 0.402, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.478157639503479, |
| "step": 300, |
| "valid_targets_mean": 4951.2, |
| "valid_targets_min": 794 |
| }, |
| { |
| "epoch": 0.5332167832167832, |
| "grad_norm": 0.7363666663898127, |
| "learning_rate": 3.032418952618454e-05, |
| "loss": 0.4199, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.44249227643013, |
| "step": 305, |
| "valid_targets_mean": 3142.7, |
| "valid_targets_min": 350 |
| }, |
| { |
| "epoch": 0.541958041958042, |
| "grad_norm": 0.7419524259349666, |
| "learning_rate": 3.082294264339152e-05, |
| "loss": 0.4473, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3787216246128082, |
| "step": 310, |
| "valid_targets_mean": 2529.6, |
| "valid_targets_min": 593 |
| }, |
| { |
| "epoch": 0.5506993006993007, |
| "grad_norm": 0.6373826151069447, |
| "learning_rate": 3.13216957605985e-05, |
| "loss": 0.4175, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.48238128423690796, |
| "step": 315, |
| "valid_targets_mean": 4511.8, |
| "valid_targets_min": 1360 |
| }, |
| { |
| "epoch": 0.5594405594405595, |
| "grad_norm": 0.6890339528272231, |
| "learning_rate": 3.182044887780549e-05, |
| "loss": 0.4172, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4121536612510681, |
| "step": 320, |
| "valid_targets_mean": 3423.4, |
| "valid_targets_min": 361 |
| }, |
| { |
| "epoch": 0.5681818181818182, |
| "grad_norm": 0.6709193485672122, |
| "learning_rate": 3.231920199501247e-05, |
| "loss": 0.4201, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4338151216506958, |
| "step": 325, |
| "valid_targets_mean": 3460.5, |
| "valid_targets_min": 1175 |
| }, |
| { |
| "epoch": 0.5769230769230769, |
| "grad_norm": 0.5691860103331635, |
| "learning_rate": 3.281795511221945e-05, |
| "loss": 0.4425, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4696723520755768, |
| "step": 330, |
| "valid_targets_mean": 4651.1, |
| "valid_targets_min": 1325 |
| }, |
| { |
| "epoch": 0.5856643356643356, |
| "grad_norm": 0.6081921585343787, |
| "learning_rate": 3.331670822942644e-05, |
| "loss": 0.4207, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.46315813064575195, |
| "step": 335, |
| "valid_targets_mean": 4476.0, |
| "valid_targets_min": 1014 |
| }, |
| { |
| "epoch": 0.5944055944055944, |
| "grad_norm": 0.6569386219652944, |
| "learning_rate": 3.3815461346633416e-05, |
| "loss": 0.4259, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4307020902633667, |
| "step": 340, |
| "valid_targets_mean": 3833.4, |
| "valid_targets_min": 942 |
| }, |
| { |
| "epoch": 0.6031468531468531, |
| "grad_norm": 0.5939332314153571, |
| "learning_rate": 3.43142144638404e-05, |
| "loss": 0.4181, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4571363925933838, |
| "step": 345, |
| "valid_targets_mean": 4945.1, |
| "valid_targets_min": 970 |
| }, |
| { |
| "epoch": 0.6118881118881119, |
| "grad_norm": 0.5969183422034433, |
| "learning_rate": 3.481296758104738e-05, |
| "loss": 0.4408, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4501033127307892, |
| "step": 350, |
| "valid_targets_mean": 4973.1, |
| "valid_targets_min": 1044 |
| }, |
| { |
| "epoch": 0.6206293706293706, |
| "grad_norm": 0.5757241278002854, |
| "learning_rate": 3.5311720698254365e-05, |
| "loss": 0.4381, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.43396443128585815, |
| "step": 355, |
| "valid_targets_mean": 4245.8, |
| "valid_targets_min": 1144 |
| }, |
| { |
| "epoch": 0.6293706293706294, |
| "grad_norm": 0.6280624127149288, |
| "learning_rate": 3.581047381546135e-05, |
| "loss": 0.4111, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3709952235221863, |
| "step": 360, |
| "valid_targets_mean": 3450.9, |
| "valid_targets_min": 1038 |
| }, |
| { |
| "epoch": 0.6381118881118881, |
| "grad_norm": 0.6272407181268593, |
| "learning_rate": 3.630922693266833e-05, |
| "loss": 0.4238, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4287092685699463, |
| "step": 365, |
| "valid_targets_mean": 3760.8, |
| "valid_targets_min": 1263 |
| }, |
| { |
| "epoch": 0.6468531468531469, |
| "grad_norm": 0.6303444859879423, |
| "learning_rate": 3.6807980049875315e-05, |
| "loss": 0.4292, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4412212371826172, |
| "step": 370, |
| "valid_targets_mean": 3570.2, |
| "valid_targets_min": 632 |
| }, |
| { |
| "epoch": 0.6555944055944056, |
| "grad_norm": 0.5953844799248368, |
| "learning_rate": 3.7306733167082294e-05, |
| "loss": 0.3885, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.43160712718963623, |
| "step": 375, |
| "valid_targets_mean": 4197.4, |
| "valid_targets_min": 1396 |
| }, |
| { |
| "epoch": 0.6643356643356644, |
| "grad_norm": 0.6082707171096963, |
| "learning_rate": 3.780548628428928e-05, |
| "loss": 0.4052, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3460230529308319, |
| "step": 380, |
| "valid_targets_mean": 3363.0, |
| "valid_targets_min": 872 |
| }, |
| { |
| "epoch": 0.6730769230769231, |
| "grad_norm": 0.5998396388984467, |
| "learning_rate": 3.8304239401496265e-05, |
| "loss": 0.3892, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4442782998085022, |
| "step": 385, |
| "valid_targets_mean": 4488.0, |
| "valid_targets_min": 396 |
| }, |
| { |
| "epoch": 0.6818181818181818, |
| "grad_norm": 0.6607949283944162, |
| "learning_rate": 3.8802992518703244e-05, |
| "loss": 0.3803, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35880744457244873, |
| "step": 390, |
| "valid_targets_mean": 2970.6, |
| "valid_targets_min": 1071 |
| }, |
| { |
| "epoch": 0.6905594405594405, |
| "grad_norm": 0.6900570931097955, |
| "learning_rate": 3.930174563591023e-05, |
| "loss": 0.4064, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36735618114471436, |
| "step": 395, |
| "valid_targets_mean": 3114.1, |
| "valid_targets_min": 1095 |
| }, |
| { |
| "epoch": 0.6993006993006993, |
| "grad_norm": 0.6914449468836832, |
| "learning_rate": 3.980049875311721e-05, |
| "loss": 0.3688, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2671054005622864, |
| "step": 400, |
| "valid_targets_mean": 2244.2, |
| "valid_targets_min": 875 |
| }, |
| { |
| "epoch": 0.708041958041958, |
| "grad_norm": 0.7160530816069505, |
| "learning_rate": 3.999993157520847e-05, |
| "loss": 0.413, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5022163391113281, |
| "step": 405, |
| "valid_targets_mean": 5052.9, |
| "valid_targets_min": 659 |
| }, |
| { |
| "epoch": 0.7167832167832168, |
| "grad_norm": 0.8302002805431459, |
| "learning_rate": 3.999951342540017e-05, |
| "loss": 0.4308, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.48281511664390564, |
| "step": 410, |
| "valid_targets_mean": 4652.4, |
| "valid_targets_min": 913 |
| }, |
| { |
| "epoch": 0.7255244755244755, |
| "grad_norm": 0.5546189149897846, |
| "learning_rate": 3.999871514749473e-05, |
| "loss": 0.4438, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3389652669429779, |
| "step": 415, |
| "valid_targets_mean": 4212.8, |
| "valid_targets_min": 758 |
| }, |
| { |
| "epoch": 0.7342657342657343, |
| "grad_norm": 0.5634394054527263, |
| "learning_rate": 3.999753675666491e-05, |
| "loss": 0.4093, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.41193175315856934, |
| "step": 420, |
| "valid_targets_mean": 5020.0, |
| "valid_targets_min": 885 |
| }, |
| { |
| "epoch": 0.743006993006993, |
| "grad_norm": 0.7004352845216515, |
| "learning_rate": 3.9995978275308226e-05, |
| "loss": 0.4024, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4134072959423065, |
| "step": 425, |
| "valid_targets_mean": 3430.8, |
| "valid_targets_min": 1156 |
| }, |
| { |
| "epoch": 0.7517482517482518, |
| "grad_norm": 0.6196320289647665, |
| "learning_rate": 3.999403973304654e-05, |
| "loss": 0.4477, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3583190441131592, |
| "step": 430, |
| "valid_targets_mean": 3033.9, |
| "valid_targets_min": 1023 |
| }, |
| { |
| "epoch": 0.7604895104895105, |
| "grad_norm": 0.5773913116859302, |
| "learning_rate": 3.999172116672545e-05, |
| "loss": 0.4322, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.397724449634552, |
| "step": 435, |
| "valid_targets_mean": 4305.8, |
| "valid_targets_min": 857 |
| }, |
| { |
| "epoch": 0.7692307692307693, |
| "grad_norm": 0.9368889840219, |
| "learning_rate": 3.9989022620413646e-05, |
| "loss": 0.3866, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.43847835063934326, |
| "step": 440, |
| "valid_targets_mean": 4250.9, |
| "valid_targets_min": 1030 |
| }, |
| { |
| "epoch": 0.777972027972028, |
| "grad_norm": 0.6324915742383318, |
| "learning_rate": 3.998594414540204e-05, |
| "loss": 0.4222, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2982059121131897, |
| "step": 445, |
| "valid_targets_mean": 2808.2, |
| "valid_targets_min": 1264 |
| }, |
| { |
| "epoch": 0.7867132867132867, |
| "grad_norm": 0.7217580076816469, |
| "learning_rate": 3.99824858002028e-05, |
| "loss": 0.3705, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.43571165204048157, |
| "step": 450, |
| "valid_targets_mean": 3194.4, |
| "valid_targets_min": 765 |
| }, |
| { |
| "epoch": 0.7954545454545454, |
| "grad_norm": 0.6237816031652519, |
| "learning_rate": 3.9978647650548225e-05, |
| "loss": 0.3564, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3463501036167145, |
| "step": 455, |
| "valid_targets_mean": 3051.4, |
| "valid_targets_min": 911 |
| }, |
| { |
| "epoch": 0.8041958041958042, |
| "grad_norm": 0.6681795392176033, |
| "learning_rate": 3.9974429769389524e-05, |
| "loss": 0.3839, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3593922555446625, |
| "step": 460, |
| "valid_targets_mean": 3530.2, |
| "valid_targets_min": 1062 |
| }, |
| { |
| "epoch": 0.8129370629370629, |
| "grad_norm": 0.6468833105433451, |
| "learning_rate": 3.996983223689541e-05, |
| "loss": 0.4033, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3399583101272583, |
| "step": 465, |
| "valid_targets_mean": 2962.4, |
| "valid_targets_min": 941 |
| }, |
| { |
| "epoch": 0.8216783216783217, |
| "grad_norm": 0.5620199704651588, |
| "learning_rate": 3.9964855140450556e-05, |
| "loss": 0.4159, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.43714678287506104, |
| "step": 470, |
| "valid_targets_mean": 5177.6, |
| "valid_targets_min": 400 |
| }, |
| { |
| "epoch": 0.8304195804195804, |
| "grad_norm": 0.6050684953003372, |
| "learning_rate": 3.995949857465399e-05, |
| "loss": 0.3659, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.40957096219062805, |
| "step": 475, |
| "valid_targets_mean": 3481.8, |
| "valid_targets_min": 907 |
| }, |
| { |
| "epoch": 0.8391608391608392, |
| "grad_norm": 0.5881967704006619, |
| "learning_rate": 3.995376264131723e-05, |
| "loss": 0.4496, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4586448073387146, |
| "step": 480, |
| "valid_targets_mean": 4609.4, |
| "valid_targets_min": 875 |
| }, |
| { |
| "epoch": 0.8479020979020979, |
| "grad_norm": 0.5322379740150757, |
| "learning_rate": 3.994764744946242e-05, |
| "loss": 0.3688, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3586678206920624, |
| "step": 485, |
| "valid_targets_mean": 4380.9, |
| "valid_targets_min": 988 |
| }, |
| { |
| "epoch": 0.8566433566433567, |
| "grad_norm": 0.5525711150756555, |
| "learning_rate": 3.994115311532018e-05, |
| "loss": 0.3889, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3657156229019165, |
| "step": 490, |
| "valid_targets_mean": 4059.6, |
| "valid_targets_min": 776 |
| }, |
| { |
| "epoch": 0.8653846153846154, |
| "grad_norm": 0.6513120586241753, |
| "learning_rate": 3.993427976232748e-05, |
| "loss": 0.3798, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.41130882501602173, |
| "step": 495, |
| "valid_targets_mean": 3188.6, |
| "valid_targets_min": 1226 |
| }, |
| { |
| "epoch": 0.8741258741258742, |
| "grad_norm": 0.6127418213610771, |
| "learning_rate": 3.99270275211252e-05, |
| "loss": 0.4165, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4457509219646454, |
| "step": 500, |
| "valid_targets_mean": 4294.4, |
| "valid_targets_min": 1023 |
| }, |
| { |
| "epoch": 0.8828671328671329, |
| "grad_norm": 0.6484091198476236, |
| "learning_rate": 3.9919396529555754e-05, |
| "loss": 0.4581, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.44694262742996216, |
| "step": 505, |
| "valid_targets_mean": 3536.8, |
| "valid_targets_min": 1392 |
| }, |
| { |
| "epoch": 0.8916083916083916, |
| "grad_norm": 0.630930600673457, |
| "learning_rate": 3.991138693266037e-05, |
| "loss": 0.3405, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3340379595756531, |
| "step": 510, |
| "valid_targets_mean": 2815.2, |
| "valid_targets_min": 1008 |
| }, |
| { |
| "epoch": 0.9003496503496503, |
| "grad_norm": 0.6303251542828444, |
| "learning_rate": 3.990299888267639e-05, |
| "loss": 0.4278, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4999693036079407, |
| "step": 515, |
| "valid_targets_mean": 5249.6, |
| "valid_targets_min": 1123 |
| }, |
| { |
| "epoch": 0.9090909090909091, |
| "grad_norm": 0.6382900314128872, |
| "learning_rate": 3.9894232539034376e-05, |
| "loss": 0.4116, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.39463382959365845, |
| "step": 520, |
| "valid_targets_mean": 3450.0, |
| "valid_targets_min": 721 |
| }, |
| { |
| "epoch": 0.9178321678321678, |
| "grad_norm": 0.5751517540128765, |
| "learning_rate": 3.9885088068355046e-05, |
| "loss": 0.4414, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4512973427772522, |
| "step": 525, |
| "valid_targets_mean": 5112.5, |
| "valid_targets_min": 1534 |
| }, |
| { |
| "epoch": 0.9265734265734266, |
| "grad_norm": 0.7081469703641994, |
| "learning_rate": 3.9875565644446165e-05, |
| "loss": 0.4415, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4004817306995392, |
| "step": 530, |
| "valid_targets_mean": 3352.9, |
| "valid_targets_min": 1239 |
| }, |
| { |
| "epoch": 0.9353146853146853, |
| "grad_norm": 0.644292943390024, |
| "learning_rate": 3.986566544829915e-05, |
| "loss": 0.3654, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3852795958518982, |
| "step": 535, |
| "valid_targets_mean": 3494.4, |
| "valid_targets_min": 1279 |
| }, |
| { |
| "epoch": 0.9440559440559441, |
| "grad_norm": 0.5674453210768845, |
| "learning_rate": 3.9855387668085733e-05, |
| "loss": 0.3847, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3958588242530823, |
| "step": 540, |
| "valid_targets_mean": 4133.1, |
| "valid_targets_min": 1497 |
| }, |
| { |
| "epoch": 0.9527972027972028, |
| "grad_norm": 0.6005492272047183, |
| "learning_rate": 3.984473249915432e-05, |
| "loss": 0.3475, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4054728150367737, |
| "step": 545, |
| "valid_targets_mean": 3535.9, |
| "valid_targets_min": 1910 |
| }, |
| { |
| "epoch": 0.9615384615384616, |
| "grad_norm": 0.5958464349073712, |
| "learning_rate": 3.983370014402629e-05, |
| "loss": 0.3792, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3395378589630127, |
| "step": 550, |
| "valid_targets_mean": 4115.0, |
| "valid_targets_min": 924 |
| }, |
| { |
| "epoch": 0.9702797202797203, |
| "grad_norm": 0.5549227602894283, |
| "learning_rate": 3.982229081239214e-05, |
| "loss": 0.4286, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4644068479537964, |
| "step": 555, |
| "valid_targets_mean": 5402.5, |
| "valid_targets_min": 860 |
| }, |
| { |
| "epoch": 0.9790209790209791, |
| "grad_norm": 0.5798046099495293, |
| "learning_rate": 3.981050472110753e-05, |
| "loss": 0.3703, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.38240253925323486, |
| "step": 560, |
| "valid_targets_mean": 3724.3, |
| "valid_targets_min": 1237 |
| }, |
| { |
| "epoch": 0.9877622377622378, |
| "grad_norm": 0.6238998765833637, |
| "learning_rate": 3.979834209418914e-05, |
| "loss": 0.4027, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.42617619037628174, |
| "step": 565, |
| "valid_targets_mean": 4521.6, |
| "valid_targets_min": 740 |
| }, |
| { |
| "epoch": 0.9965034965034965, |
| "grad_norm": 0.6024006428537647, |
| "learning_rate": 3.97858031628104e-05, |
| "loss": 0.4082, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.49112826585769653, |
| "step": 570, |
| "valid_targets_mean": 5095.2, |
| "valid_targets_min": 1146 |
| }, |
| { |
| "epoch": 1.0052447552447552, |
| "grad_norm": 0.6342997589440736, |
| "learning_rate": 3.977288816529711e-05, |
| "loss": 0.3258, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3129345178604126, |
| "step": 575, |
| "valid_targets_mean": 2719.0, |
| "valid_targets_min": 968 |
| }, |
| { |
| "epoch": 1.013986013986014, |
| "grad_norm": 0.629458660323066, |
| "learning_rate": 3.975959734712293e-05, |
| "loss": 0.406, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3763844668865204, |
| "step": 580, |
| "valid_targets_mean": 3271.1, |
| "valid_targets_min": 402 |
| }, |
| { |
| "epoch": 1.0227272727272727, |
| "grad_norm": 0.656404561523582, |
| "learning_rate": 3.974593096090465e-05, |
| "loss": 0.3476, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31588298082351685, |
| "step": 585, |
| "valid_targets_mean": 2591.9, |
| "valid_targets_min": 659 |
| }, |
| { |
| "epoch": 1.0314685314685315, |
| "grad_norm": 0.7256823083367988, |
| "learning_rate": 3.973188926639749e-05, |
| "loss": 0.3918, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.45462703704833984, |
| "step": 590, |
| "valid_targets_mean": 3205.2, |
| "valid_targets_min": 1217 |
| }, |
| { |
| "epoch": 1.0402097902097902, |
| "grad_norm": 0.5895144169036058, |
| "learning_rate": 3.971747253049004e-05, |
| "loss": 0.381, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3712193965911865, |
| "step": 595, |
| "valid_targets_mean": 3694.0, |
| "valid_targets_min": 1084 |
| }, |
| { |
| "epoch": 1.048951048951049, |
| "grad_norm": 0.5690400741815232, |
| "learning_rate": 3.9702681027199315e-05, |
| "loss": 0.3814, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.427038311958313, |
| "step": 600, |
| "valid_targets_mean": 4761.9, |
| "valid_targets_min": 1074 |
| }, |
| { |
| "epoch": 1.0576923076923077, |
| "grad_norm": 0.638796668052871, |
| "learning_rate": 3.968751503766544e-05, |
| "loss": 0.3895, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3525705337524414, |
| "step": 605, |
| "valid_targets_mean": 3294.1, |
| "valid_targets_min": 1100 |
| }, |
| { |
| "epoch": 1.0664335664335665, |
| "grad_norm": 0.5612529799321073, |
| "learning_rate": 3.967197485014637e-05, |
| "loss": 0.3967, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.40732014179229736, |
| "step": 610, |
| "valid_targets_mean": 4946.8, |
| "valid_targets_min": 829 |
| }, |
| { |
| "epoch": 1.0751748251748252, |
| "grad_norm": 0.5332511823970404, |
| "learning_rate": 3.9656060760012386e-05, |
| "loss": 0.3887, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3988763093948364, |
| "step": 615, |
| "valid_targets_mean": 4778.2, |
| "valid_targets_min": 1161 |
| }, |
| { |
| "epoch": 1.083916083916084, |
| "grad_norm": 0.9907300506794291, |
| "learning_rate": 3.9639773069740484e-05, |
| "loss": 0.3474, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.348149836063385, |
| "step": 620, |
| "valid_targets_mean": 3704.6, |
| "valid_targets_min": 1498 |
| }, |
| { |
| "epoch": 1.0926573426573427, |
| "grad_norm": 0.6489040626377737, |
| "learning_rate": 3.9623112088908627e-05, |
| "loss": 0.3851, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22918564081192017, |
| "step": 625, |
| "valid_targets_mean": 2000.0, |
| "valid_targets_min": 864 |
| }, |
| { |
| "epoch": 1.1013986013986015, |
| "grad_norm": 0.5178066275559241, |
| "learning_rate": 3.960607813418987e-05, |
| "loss": 0.3858, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4411054849624634, |
| "step": 630, |
| "valid_targets_mean": 6213.5, |
| "valid_targets_min": 1224 |
| }, |
| { |
| "epoch": 1.1101398601398602, |
| "grad_norm": 0.586299273251672, |
| "learning_rate": 3.9588671529346345e-05, |
| "loss": 0.3361, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4047016203403473, |
| "step": 635, |
| "valid_targets_mean": 4316.3, |
| "valid_targets_min": 1014 |
| }, |
| { |
| "epoch": 1.118881118881119, |
| "grad_norm": 0.6006735013024012, |
| "learning_rate": 3.957089260522306e-05, |
| "loss": 0.347, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3549342155456543, |
| "step": 640, |
| "valid_targets_mean": 3763.6, |
| "valid_targets_min": 1124 |
| }, |
| { |
| "epoch": 1.1276223776223777, |
| "grad_norm": 0.6863014700797007, |
| "learning_rate": 3.9552741699741674e-05, |
| "loss": 0.3996, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.452499657869339, |
| "step": 645, |
| "valid_targets_mean": 3432.2, |
| "valid_targets_min": 281 |
| }, |
| { |
| "epoch": 1.1363636363636362, |
| "grad_norm": 0.6285631579650135, |
| "learning_rate": 3.953421915789403e-05, |
| "loss": 0.3744, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3433941602706909, |
| "step": 650, |
| "valid_targets_mean": 2940.2, |
| "valid_targets_min": 727 |
| }, |
| { |
| "epoch": 1.145104895104895, |
| "grad_norm": 0.6198040249923603, |
| "learning_rate": 3.9515325331735635e-05, |
| "loss": 0.3474, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36902672052383423, |
| "step": 655, |
| "valid_targets_mean": 3038.7, |
| "valid_targets_min": 1422 |
| }, |
| { |
| "epoch": 1.1538461538461537, |
| "grad_norm": 0.5020755508723698, |
| "learning_rate": 3.949606058037893e-05, |
| "loss": 0.3811, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3350529670715332, |
| "step": 660, |
| "valid_targets_mean": 4884.1, |
| "valid_targets_min": 1339 |
| }, |
| { |
| "epoch": 1.1625874125874125, |
| "grad_norm": 0.5538973265752763, |
| "learning_rate": 3.947642526998649e-05, |
| "loss": 0.357, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.38185301423072815, |
| "step": 665, |
| "valid_targets_mean": 4053.4, |
| "valid_targets_min": 711 |
| }, |
| { |
| "epoch": 1.1713286713286712, |
| "grad_norm": 0.6390899954407651, |
| "learning_rate": 3.945641977376404e-05, |
| "loss": 0.3731, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4220157265663147, |
| "step": 670, |
| "valid_targets_mean": 4628.8, |
| "valid_targets_min": 1091 |
| }, |
| { |
| "epoch": 1.18006993006993, |
| "grad_norm": 0.6411330384913675, |
| "learning_rate": 3.94360444719534e-05, |
| "loss": 0.4001, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4042211174964905, |
| "step": 675, |
| "valid_targets_mean": 3807.4, |
| "valid_targets_min": 1606 |
| }, |
| { |
| "epoch": 1.1888111888111887, |
| "grad_norm": 0.6039498968928956, |
| "learning_rate": 3.941529975182524e-05, |
| "loss": 0.3622, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.40398502349853516, |
| "step": 680, |
| "valid_targets_mean": 4085.2, |
| "valid_targets_min": 831 |
| }, |
| { |
| "epoch": 1.1975524475524475, |
| "grad_norm": 0.5877952077296495, |
| "learning_rate": 3.939418600767168e-05, |
| "loss": 0.3757, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4611789584159851, |
| "step": 685, |
| "valid_targets_mean": 4750.3, |
| "valid_targets_min": 1151 |
| }, |
| { |
| "epoch": 1.2062937062937062, |
| "grad_norm": 0.6355903559851636, |
| "learning_rate": 3.937270364079886e-05, |
| "loss": 0.3513, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31703072786331177, |
| "step": 690, |
| "valid_targets_mean": 3618.4, |
| "valid_targets_min": 1137 |
| }, |
| { |
| "epoch": 1.215034965034965, |
| "grad_norm": 0.6128675990146162, |
| "learning_rate": 3.935085305951929e-05, |
| "loss": 0.3937, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4473511874675751, |
| "step": 695, |
| "valid_targets_mean": 3876.3, |
| "valid_targets_min": 1234 |
| }, |
| { |
| "epoch": 1.2237762237762237, |
| "grad_norm": 0.6112543941773279, |
| "learning_rate": 3.932863467914405e-05, |
| "loss": 0.359, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36574071645736694, |
| "step": 700, |
| "valid_targets_mean": 3475.4, |
| "valid_targets_min": 898 |
| }, |
| { |
| "epoch": 1.2325174825174825, |
| "grad_norm": 0.6808146656666811, |
| "learning_rate": 3.930604892197496e-05, |
| "loss": 0.3789, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4020303785800934, |
| "step": 705, |
| "valid_targets_mean": 3257.4, |
| "valid_targets_min": 904 |
| }, |
| { |
| "epoch": 1.2412587412587412, |
| "grad_norm": 0.6143751077595933, |
| "learning_rate": 3.9283096217296496e-05, |
| "loss": 0.3558, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.38672855496406555, |
| "step": 710, |
| "valid_targets_mean": 3559.6, |
| "valid_targets_min": 933 |
| }, |
| { |
| "epoch": 1.25, |
| "grad_norm": 0.638690578066305, |
| "learning_rate": 3.925977700136768e-05, |
| "loss": 0.3882, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.39257192611694336, |
| "step": 715, |
| "valid_targets_mean": 3589.1, |
| "valid_targets_min": 940 |
| }, |
| { |
| "epoch": 1.2587412587412588, |
| "grad_norm": 0.5457598428008629, |
| "learning_rate": 3.923609171741374e-05, |
| "loss": 0.3654, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3970078229904175, |
| "step": 720, |
| "valid_targets_mean": 4277.0, |
| "valid_targets_min": 286 |
| }, |
| { |
| "epoch": 1.2674825174825175, |
| "grad_norm": 0.502276044700275, |
| "learning_rate": 3.921204081561774e-05, |
| "loss": 0.3799, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.30176183581352234, |
| "step": 725, |
| "valid_targets_mean": 4076.9, |
| "valid_targets_min": 875 |
| }, |
| { |
| "epoch": 1.2762237762237763, |
| "grad_norm": 0.6472659228715367, |
| "learning_rate": 3.918762475311197e-05, |
| "loss": 0.4084, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4843234419822693, |
| "step": 730, |
| "valid_targets_mean": 4021.4, |
| "valid_targets_min": 1101 |
| }, |
| { |
| "epoch": 1.284965034965035, |
| "grad_norm": 0.5283842836058, |
| "learning_rate": 3.91628439939693e-05, |
| "loss": 0.3659, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3728162348270416, |
| "step": 735, |
| "valid_targets_mean": 4352.0, |
| "valid_targets_min": 954 |
| }, |
| { |
| "epoch": 1.2937062937062938, |
| "grad_norm": 0.5966704473150931, |
| "learning_rate": 3.913769900919431e-05, |
| "loss": 0.3566, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.38558197021484375, |
| "step": 740, |
| "valid_targets_mean": 5053.8, |
| "valid_targets_min": 1163 |
| }, |
| { |
| "epoch": 1.3024475524475525, |
| "grad_norm": 0.6138554581865869, |
| "learning_rate": 3.91121902767144e-05, |
| "loss": 0.3435, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3052510619163513, |
| "step": 745, |
| "valid_targets_mean": 3518.2, |
| "valid_targets_min": 1127 |
| }, |
| { |
| "epoch": 1.3111888111888113, |
| "grad_norm": 0.6487462123963363, |
| "learning_rate": 3.908631828137067e-05, |
| "loss": 0.4114, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4444257318973541, |
| "step": 750, |
| "valid_targets_mean": 4018.6, |
| "valid_targets_min": 836 |
| }, |
| { |
| "epoch": 1.31993006993007, |
| "grad_norm": 0.508254488406239, |
| "learning_rate": 3.9060083514908695e-05, |
| "loss": 0.4277, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.42669036984443665, |
| "step": 755, |
| "valid_targets_mean": 5763.0, |
| "valid_targets_min": 1080 |
| }, |
| { |
| "epoch": 1.3286713286713288, |
| "grad_norm": 0.48404850385049064, |
| "learning_rate": 3.903348647596919e-05, |
| "loss": 0.3673, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.42117491364479065, |
| "step": 760, |
| "valid_targets_mean": 6734.8, |
| "valid_targets_min": 1439 |
| }, |
| { |
| "epoch": 1.3374125874125875, |
| "grad_norm": 0.5928130214677121, |
| "learning_rate": 3.900652767007855e-05, |
| "loss": 0.363, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3812492787837982, |
| "step": 765, |
| "valid_targets_mean": 5822.9, |
| "valid_targets_min": 844 |
| }, |
| { |
| "epoch": 1.3461538461538463, |
| "grad_norm": 0.6239373645782617, |
| "learning_rate": 3.8979207609639225e-05, |
| "loss": 0.3715, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4304284155368805, |
| "step": 770, |
| "valid_targets_mean": 4057.4, |
| "valid_targets_min": 816 |
| }, |
| { |
| "epoch": 1.354895104895105, |
| "grad_norm": 0.5195351816200816, |
| "learning_rate": 3.8951526813919975e-05, |
| "loss": 0.3717, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.30584895610809326, |
| "step": 775, |
| "valid_targets_mean": 4133.8, |
| "valid_targets_min": 1393 |
| }, |
| { |
| "epoch": 1.3636363636363638, |
| "grad_norm": 0.602430063013215, |
| "learning_rate": 3.8923485809046006e-05, |
| "loss": 0.3618, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.43198758363723755, |
| "step": 780, |
| "valid_targets_mean": 5221.6, |
| "valid_targets_min": 865 |
| }, |
| { |
| "epoch": 1.3723776223776225, |
| "grad_norm": 0.5793853580026894, |
| "learning_rate": 3.889508512798898e-05, |
| "loss": 0.3733, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.375804603099823, |
| "step": 785, |
| "valid_targets_mean": 3890.5, |
| "valid_targets_min": 1119 |
| }, |
| { |
| "epoch": 1.381118881118881, |
| "grad_norm": 0.6283176928865934, |
| "learning_rate": 3.886632531055687e-05, |
| "loss": 0.3507, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36522990465164185, |
| "step": 790, |
| "valid_targets_mean": 2879.0, |
| "valid_targets_min": 914 |
| }, |
| { |
| "epoch": 1.3898601398601398, |
| "grad_norm": 0.5586170363792717, |
| "learning_rate": 3.883720690338372e-05, |
| "loss": 0.3915, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36750155687332153, |
| "step": 795, |
| "valid_targets_mean": 3684.4, |
| "valid_targets_min": 1034 |
| }, |
| { |
| "epoch": 1.3986013986013985, |
| "grad_norm": 0.5971605319742396, |
| "learning_rate": 3.880773045991921e-05, |
| "loss": 0.3822, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31041955947875977, |
| "step": 800, |
| "valid_targets_mean": 3070.0, |
| "valid_targets_min": 405 |
| }, |
| { |
| "epoch": 1.4073426573426573, |
| "grad_norm": 0.5992895435680253, |
| "learning_rate": 3.8777896540418204e-05, |
| "loss": 0.3407, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3061218559741974, |
| "step": 805, |
| "valid_targets_mean": 2962.3, |
| "valid_targets_min": 1200 |
| }, |
| { |
| "epoch": 1.416083916083916, |
| "grad_norm": 0.5163991467972393, |
| "learning_rate": 3.874770571193004e-05, |
| "loss": 0.3634, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4339129328727722, |
| "step": 810, |
| "valid_targets_mean": 6174.8, |
| "valid_targets_min": 2106 |
| }, |
| { |
| "epoch": 1.4248251748251748, |
| "grad_norm": 0.5915515016899537, |
| "learning_rate": 3.871715854828779e-05, |
| "loss": 0.3775, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3776523470878601, |
| "step": 815, |
| "valid_targets_mean": 4111.8, |
| "valid_targets_min": 986 |
| }, |
| { |
| "epoch": 1.4335664335664335, |
| "grad_norm": 0.5746894134419273, |
| "learning_rate": 3.8686255630097346e-05, |
| "loss": 0.3695, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3666810989379883, |
| "step": 820, |
| "valid_targets_mean": 3390.7, |
| "valid_targets_min": 1340 |
| }, |
| { |
| "epoch": 1.4423076923076923, |
| "grad_norm": 0.6144270484363288, |
| "learning_rate": 3.8654997544726374e-05, |
| "loss": 0.3715, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.38752689957618713, |
| "step": 825, |
| "valid_targets_mean": 3446.4, |
| "valid_targets_min": 1126 |
| }, |
| { |
| "epoch": 1.451048951048951, |
| "grad_norm": 0.5922060872547307, |
| "learning_rate": 3.862338488629314e-05, |
| "loss": 0.3757, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34740880131721497, |
| "step": 830, |
| "valid_targets_mean": 4006.9, |
| "valid_targets_min": 991 |
| }, |
| { |
| "epoch": 1.4597902097902098, |
| "grad_norm": 0.6002301065901973, |
| "learning_rate": 3.859141825565525e-05, |
| "loss": 0.3843, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35890963673591614, |
| "step": 835, |
| "valid_targets_mean": 3065.4, |
| "valid_targets_min": 1571 |
| }, |
| { |
| "epoch": 1.4685314685314685, |
| "grad_norm": 0.6051255681472264, |
| "learning_rate": 3.855909826039822e-05, |
| "loss": 0.3499, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3590480089187622, |
| "step": 840, |
| "valid_targets_mean": 2918.1, |
| "valid_targets_min": 1159 |
| }, |
| { |
| "epoch": 1.4772727272727273, |
| "grad_norm": 0.46918016708384985, |
| "learning_rate": 3.852642551482391e-05, |
| "loss": 0.3998, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3891509175300598, |
| "step": 845, |
| "valid_targets_mean": 5518.4, |
| "valid_targets_min": 794 |
| }, |
| { |
| "epoch": 1.486013986013986, |
| "grad_norm": 0.5672184897056678, |
| "learning_rate": 3.849340063993884e-05, |
| "loss": 0.3841, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.46270751953125, |
| "step": 850, |
| "valid_targets_mean": 5028.8, |
| "valid_targets_min": 1013 |
| }, |
| { |
| "epoch": 1.4947552447552448, |
| "grad_norm": 0.6032267971221293, |
| "learning_rate": 3.8460024263442446e-05, |
| "loss": 0.3907, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36976176500320435, |
| "step": 855, |
| "valid_targets_mean": 3296.6, |
| "valid_targets_min": 295 |
| }, |
| { |
| "epoch": 1.5034965034965035, |
| "grad_norm": 0.5937950046708127, |
| "learning_rate": 3.842629701971507e-05, |
| "loss": 0.3838, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3807605504989624, |
| "step": 860, |
| "valid_targets_mean": 3674.1, |
| "valid_targets_min": 866 |
| }, |
| { |
| "epoch": 1.512237762237762, |
| "grad_norm": 0.5638487866108519, |
| "learning_rate": 3.839221954980596e-05, |
| "loss": 0.3611, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35325905680656433, |
| "step": 865, |
| "valid_targets_mean": 3803.4, |
| "valid_targets_min": 397 |
| }, |
| { |
| "epoch": 1.5209790209790208, |
| "grad_norm": 0.5702323841541543, |
| "learning_rate": 3.835779250142108e-05, |
| "loss": 0.3512, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34167003631591797, |
| "step": 870, |
| "valid_targets_mean": 3481.8, |
| "valid_targets_min": 992 |
| }, |
| { |
| "epoch": 1.5297202797202796, |
| "grad_norm": 0.5291286455210686, |
| "learning_rate": 3.8323016528910746e-05, |
| "loss": 0.3859, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3847964107990265, |
| "step": 875, |
| "valid_targets_mean": 4120.1, |
| "valid_targets_min": 1221 |
| }, |
| { |
| "epoch": 1.5384615384615383, |
| "grad_norm": 0.6275254868035214, |
| "learning_rate": 3.828789229325726e-05, |
| "loss": 0.3926, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.42605069279670715, |
| "step": 880, |
| "valid_targets_mean": 4082.6, |
| "valid_targets_min": 865 |
| }, |
| { |
| "epoch": 1.547202797202797, |
| "grad_norm": 0.5396398856381337, |
| "learning_rate": 3.8252420462062315e-05, |
| "loss": 0.3914, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36167097091674805, |
| "step": 885, |
| "valid_targets_mean": 4174.7, |
| "valid_targets_min": 1224 |
| }, |
| { |
| "epoch": 1.5559440559440558, |
| "grad_norm": 0.47910903259168275, |
| "learning_rate": 3.8216601709534284e-05, |
| "loss": 0.3922, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3376033306121826, |
| "step": 890, |
| "valid_targets_mean": 5025.0, |
| "valid_targets_min": 1179 |
| }, |
| { |
| "epoch": 1.5646853146853146, |
| "grad_norm": 0.5073648171270649, |
| "learning_rate": 3.818043671647545e-05, |
| "loss": 0.368, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.39315563440322876, |
| "step": 895, |
| "valid_targets_mean": 5091.6, |
| "valid_targets_min": 286 |
| }, |
| { |
| "epoch": 1.5734265734265733, |
| "grad_norm": 0.6352472398575295, |
| "learning_rate": 3.814392617026904e-05, |
| "loss": 0.3474, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.42686229944229126, |
| "step": 900, |
| "valid_targets_mean": 4175.8, |
| "valid_targets_min": 783 |
| }, |
| { |
| "epoch": 1.582167832167832, |
| "grad_norm": 0.6935854438042858, |
| "learning_rate": 3.810707076486616e-05, |
| "loss": 0.3934, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27806347608566284, |
| "step": 905, |
| "valid_targets_mean": 2418.1, |
| "valid_targets_min": 745 |
| }, |
| { |
| "epoch": 1.5909090909090908, |
| "grad_norm": 0.6454560477570744, |
| "learning_rate": 3.80698712007726e-05, |
| "loss": 0.3393, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.40068015456199646, |
| "step": 910, |
| "valid_targets_mean": 3282.2, |
| "valid_targets_min": 1009 |
| }, |
| { |
| "epoch": 1.5996503496503496, |
| "grad_norm": 0.71255907929696, |
| "learning_rate": 3.8032328185035535e-05, |
| "loss": 0.3491, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.38632285594940186, |
| "step": 915, |
| "valid_targets_mean": 2531.9, |
| "valid_targets_min": 1153 |
| }, |
| { |
| "epoch": 1.6083916083916083, |
| "grad_norm": 0.6583779603016164, |
| "learning_rate": 3.7994442431230096e-05, |
| "loss": 0.3722, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3876379430294037, |
| "step": 920, |
| "valid_targets_mean": 2792.9, |
| "valid_targets_min": 721 |
| }, |
| { |
| "epoch": 1.617132867132867, |
| "grad_norm": 0.548264282270815, |
| "learning_rate": 3.7956214659445764e-05, |
| "loss": 0.332, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3522208034992218, |
| "step": 925, |
| "valid_targets_mean": 3671.8, |
| "valid_targets_min": 1283 |
| }, |
| { |
| "epoch": 1.6258741258741258, |
| "grad_norm": 0.5866905542029235, |
| "learning_rate": 3.7917645596272734e-05, |
| "loss": 0.3874, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3922966420650482, |
| "step": 930, |
| "valid_targets_mean": 3553.9, |
| "valid_targets_min": 753 |
| }, |
| { |
| "epoch": 1.6346153846153846, |
| "grad_norm": 0.48891096645703447, |
| "learning_rate": 3.7878735974788056e-05, |
| "loss": 0.3726, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4203706383705139, |
| "step": 935, |
| "valid_targets_mean": 5460.0, |
| "valid_targets_min": 1274 |
| }, |
| { |
| "epoch": 1.6433566433566433, |
| "grad_norm": 0.5973489726719953, |
| "learning_rate": 3.783948653454176e-05, |
| "loss": 0.3664, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3803213834762573, |
| "step": 940, |
| "valid_targets_mean": 3419.8, |
| "valid_targets_min": 1239 |
| }, |
| { |
| "epoch": 1.652097902097902, |
| "grad_norm": 0.5931207818844045, |
| "learning_rate": 3.779989802154273e-05, |
| "loss": 0.3846, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4405151903629303, |
| "step": 945, |
| "valid_targets_mean": 4785.6, |
| "valid_targets_min": 703 |
| }, |
| { |
| "epoch": 1.6608391608391608, |
| "grad_norm": 0.49894822325713817, |
| "learning_rate": 3.7759971188244594e-05, |
| "loss": 0.3649, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3540881276130676, |
| "step": 950, |
| "valid_targets_mean": 4723.8, |
| "valid_targets_min": 1461 |
| }, |
| { |
| "epoch": 1.6695804195804196, |
| "grad_norm": 0.5865370598515728, |
| "learning_rate": 3.771970679353135e-05, |
| "loss": 0.3366, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2708732485771179, |
| "step": 955, |
| "valid_targets_mean": 2668.8, |
| "valid_targets_min": 1174 |
| }, |
| { |
| "epoch": 1.6783216783216783, |
| "grad_norm": 0.5791850182023949, |
| "learning_rate": 3.767910560270303e-05, |
| "loss": 0.3523, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3443334102630615, |
| "step": 960, |
| "valid_targets_mean": 3261.9, |
| "valid_targets_min": 317 |
| }, |
| { |
| "epoch": 1.687062937062937, |
| "grad_norm": 0.6093901195497682, |
| "learning_rate": 3.763816838746107e-05, |
| "loss": 0.3395, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3926513195037842, |
| "step": 965, |
| "valid_targets_mean": 3403.4, |
| "valid_targets_min": 860 |
| }, |
| { |
| "epoch": 1.6958041958041958, |
| "grad_norm": 0.5919210341830557, |
| "learning_rate": 3.759689592589367e-05, |
| "loss": 0.3628, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31532764434814453, |
| "step": 970, |
| "valid_targets_mean": 2884.4, |
| "valid_targets_min": 1038 |
| }, |
| { |
| "epoch": 1.7045454545454546, |
| "grad_norm": 0.5817284643320272, |
| "learning_rate": 3.755528900246106e-05, |
| "loss": 0.3591, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3037341237068176, |
| "step": 975, |
| "valid_targets_mean": 2825.8, |
| "valid_targets_min": 1070 |
| }, |
| { |
| "epoch": 1.7132867132867133, |
| "grad_norm": 0.5981859182760877, |
| "learning_rate": 3.7513348407980483e-05, |
| "loss": 0.4053, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34905627369880676, |
| "step": 980, |
| "valid_targets_mean": 3172.6, |
| "valid_targets_min": 1304 |
| }, |
| { |
| "epoch": 1.722027972027972, |
| "grad_norm": 0.5243877796252934, |
| "learning_rate": 3.747107493961126e-05, |
| "loss": 0.3649, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.44003045558929443, |
| "step": 985, |
| "valid_targets_mean": 6281.9, |
| "valid_targets_min": 1602 |
| }, |
| { |
| "epoch": 1.7307692307692308, |
| "grad_norm": 0.6888200590905601, |
| "learning_rate": 3.7428469400839604e-05, |
| "loss": 0.3172, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27428072690963745, |
| "step": 990, |
| "valid_targets_mean": 2553.9, |
| "valid_targets_min": 908 |
| }, |
| { |
| "epoch": 1.7395104895104896, |
| "grad_norm": 0.6562142049594524, |
| "learning_rate": 3.738553260146332e-05, |
| "loss": 0.3421, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36127781867980957, |
| "step": 995, |
| "valid_targets_mean": 3028.3, |
| "valid_targets_min": 938 |
| }, |
| { |
| "epoch": 1.7482517482517483, |
| "grad_norm": 0.6242645620698037, |
| "learning_rate": 3.734226535757647e-05, |
| "loss": 0.3666, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4239537715911865, |
| "step": 1000, |
| "valid_targets_mean": 3935.4, |
| "valid_targets_min": 1225 |
| }, |
| { |
| "epoch": 1.756993006993007, |
| "grad_norm": 0.48728283464574973, |
| "learning_rate": 3.729866849155381e-05, |
| "loss": 0.326, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.388911634683609, |
| "step": 1005, |
| "valid_targets_mean": 5385.4, |
| "valid_targets_min": 1104 |
| }, |
| { |
| "epoch": 1.7657342657342658, |
| "grad_norm": 0.6342231197385638, |
| "learning_rate": 3.7254742832035174e-05, |
| "loss": 0.3811, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3924434185028076, |
| "step": 1010, |
| "valid_targets_mean": 3746.3, |
| "valid_targets_min": 1110 |
| }, |
| { |
| "epoch": 1.7744755244755246, |
| "grad_norm": 0.5552439867083838, |
| "learning_rate": 3.721048921390975e-05, |
| "loss": 0.3546, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32628345489501953, |
| "step": 1015, |
| "valid_targets_mean": 2988.6, |
| "valid_targets_min": 1117 |
| }, |
| { |
| "epoch": 1.7832167832167833, |
| "grad_norm": 0.5174881529638417, |
| "learning_rate": 3.716590847830019e-05, |
| "loss": 0.3405, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4475651979446411, |
| "step": 1020, |
| "valid_targets_mean": 5184.8, |
| "valid_targets_min": 1395 |
| }, |
| { |
| "epoch": 1.791958041958042, |
| "grad_norm": 0.5747806673287248, |
| "learning_rate": 3.712100147254662e-05, |
| "loss": 0.3569, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31613489985466003, |
| "step": 1025, |
| "valid_targets_mean": 3144.9, |
| "valid_targets_min": 1323 |
| }, |
| { |
| "epoch": 1.8006993006993008, |
| "grad_norm": 0.5455829057478265, |
| "learning_rate": 3.707576905019053e-05, |
| "loss": 0.3623, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33471807837486267, |
| "step": 1030, |
| "valid_targets_mean": 3088.8, |
| "valid_targets_min": 857 |
| }, |
| { |
| "epoch": 1.8094405594405596, |
| "grad_norm": 0.6517636899008046, |
| "learning_rate": 3.703021207095856e-05, |
| "loss": 0.3776, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.38492679595947266, |
| "step": 1035, |
| "valid_targets_mean": 2833.6, |
| "valid_targets_min": 894 |
| }, |
| { |
| "epoch": 1.8181818181818183, |
| "grad_norm": 0.5801054629631718, |
| "learning_rate": 3.6984331400746184e-05, |
| "loss": 0.3721, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3238530457019806, |
| "step": 1040, |
| "valid_targets_mean": 3322.8, |
| "valid_targets_min": 985 |
| }, |
| { |
| "epoch": 1.8269230769230769, |
| "grad_norm": 0.5586861505293964, |
| "learning_rate": 3.693812791160121e-05, |
| "loss": 0.3558, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3880930542945862, |
| "step": 1045, |
| "valid_targets_mean": 3823.8, |
| "valid_targets_min": 1197 |
| }, |
| { |
| "epoch": 1.8356643356643356, |
| "grad_norm": 0.6145308306637935, |
| "learning_rate": 3.6891602481707207e-05, |
| "loss": 0.3266, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3544296622276306, |
| "step": 1050, |
| "valid_targets_mean": 3499.2, |
| "valid_targets_min": 1000 |
| }, |
| { |
| "epoch": 1.8444055944055944, |
| "grad_norm": 0.58907820308691, |
| "learning_rate": 3.684475599536687e-05, |
| "loss": 0.3682, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4255306124687195, |
| "step": 1055, |
| "valid_targets_mean": 5237.4, |
| "valid_targets_min": 1340 |
| }, |
| { |
| "epoch": 1.8531468531468531, |
| "grad_norm": 0.5015996234360681, |
| "learning_rate": 3.6797589342985126e-05, |
| "loss": 0.3616, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.37960782647132874, |
| "step": 1060, |
| "valid_targets_mean": 4262.1, |
| "valid_targets_min": 910 |
| }, |
| { |
| "epoch": 1.8618881118881119, |
| "grad_norm": 0.5260945935702037, |
| "learning_rate": 3.6750103421052286e-05, |
| "loss": 0.3365, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.37209346890449524, |
| "step": 1065, |
| "valid_targets_mean": 4495.8, |
| "valid_targets_min": 1369 |
| }, |
| { |
| "epoch": 1.8706293706293706, |
| "grad_norm": 0.5542537902859789, |
| "learning_rate": 3.670229913212696e-05, |
| "loss": 0.38, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.42989593744277954, |
| "step": 1070, |
| "valid_targets_mean": 5143.3, |
| "valid_targets_min": 1686 |
| }, |
| { |
| "epoch": 1.8793706293706294, |
| "grad_norm": 0.5321301926401418, |
| "learning_rate": 3.6654177384818936e-05, |
| "loss": 0.3785, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4028235673904419, |
| "step": 1075, |
| "valid_targets_mean": 4950.6, |
| "valid_targets_min": 1291 |
| }, |
| { |
| "epoch": 1.8881118881118881, |
| "grad_norm": 0.584447522220945, |
| "learning_rate": 3.6605739093771854e-05, |
| "loss": 0.3795, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.37142544984817505, |
| "step": 1080, |
| "valid_targets_mean": 3263.2, |
| "valid_targets_min": 795 |
| }, |
| { |
| "epoch": 1.8968531468531469, |
| "grad_norm": 0.5350619514142159, |
| "learning_rate": 3.6556985179645905e-05, |
| "loss": 0.3577, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.41316014528274536, |
| "step": 1085, |
| "valid_targets_mean": 4889.9, |
| "valid_targets_min": 1222 |
| }, |
| { |
| "epoch": 1.9055944055944056, |
| "grad_norm": 0.5814069068694546, |
| "learning_rate": 3.6507916569100255e-05, |
| "loss": 0.3466, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3615434169769287, |
| "step": 1090, |
| "valid_targets_mean": 2938.7, |
| "valid_targets_min": 1106 |
| }, |
| { |
| "epoch": 1.9143356643356644, |
| "grad_norm": 0.6849481252444816, |
| "learning_rate": 3.645853419477547e-05, |
| "loss": 0.3651, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.258215069770813, |
| "step": 1095, |
| "valid_targets_mean": 2303.6, |
| "valid_targets_min": 344 |
| }, |
| { |
| "epoch": 1.9230769230769231, |
| "grad_norm": 0.5454940887177128, |
| "learning_rate": 3.640883899527579e-05, |
| "loss": 0.3889, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.38337862491607666, |
| "step": 1100, |
| "valid_targets_mean": 4225.1, |
| "valid_targets_min": 1207 |
| }, |
| { |
| "epoch": 1.9318181818181817, |
| "grad_norm": 0.5879298330740542, |
| "learning_rate": 3.6358831915151276e-05, |
| "loss": 0.3781, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5195822715759277, |
| "step": 1105, |
| "valid_targets_mean": 5530.6, |
| "valid_targets_min": 896 |
| }, |
| { |
| "epoch": 1.9405594405594404, |
| "grad_norm": 0.5232477468919542, |
| "learning_rate": 3.630851390487985e-05, |
| "loss": 0.3618, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32636207342147827, |
| "step": 1110, |
| "valid_targets_mean": 3888.2, |
| "valid_targets_min": 1052 |
| }, |
| { |
| "epoch": 1.9493006993006992, |
| "grad_norm": 0.5356916033918182, |
| "learning_rate": 3.625788592084926e-05, |
| "loss": 0.3587, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3277405798435211, |
| "step": 1115, |
| "valid_targets_mean": 3823.6, |
| "valid_targets_min": 408 |
| }, |
| { |
| "epoch": 1.958041958041958, |
| "grad_norm": 0.5544913344651682, |
| "learning_rate": 3.620694892533889e-05, |
| "loss": 0.3694, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.42964452505111694, |
| "step": 1120, |
| "valid_targets_mean": 4196.4, |
| "valid_targets_min": 1248 |
| }, |
| { |
| "epoch": 1.9667832167832167, |
| "grad_norm": 0.558219067439635, |
| "learning_rate": 3.615570388650144e-05, |
| "loss": 0.3371, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.38968485593795776, |
| "step": 1125, |
| "valid_targets_mean": 3462.7, |
| "valid_targets_min": 870 |
| }, |
| { |
| "epoch": 1.9755244755244754, |
| "grad_norm": 0.5275189202715219, |
| "learning_rate": 3.610415177834457e-05, |
| "loss": 0.3589, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.42595839500427246, |
| "step": 1130, |
| "valid_targets_mean": 4739.6, |
| "valid_targets_min": 880 |
| }, |
| { |
| "epoch": 1.9842657342657342, |
| "grad_norm": 0.5691050784475108, |
| "learning_rate": 3.605229358071236e-05, |
| "loss": 0.3618, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.39064157009124756, |
| "step": 1135, |
| "valid_targets_mean": 3912.1, |
| "valid_targets_min": 1217 |
| }, |
| { |
| "epoch": 1.993006993006993, |
| "grad_norm": 0.5388771565097371, |
| "learning_rate": 3.600013027926667e-05, |
| "loss": 0.3487, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2591717839241028, |
| "step": 1140, |
| "valid_targets_mean": 3237.1, |
| "valid_targets_min": 1415 |
| }, |
| { |
| "epoch": 2.0017482517482517, |
| "grad_norm": 0.5553503636171387, |
| "learning_rate": 3.594766286546847e-05, |
| "loss": 0.3646, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32958346605300903, |
| "step": 1145, |
| "valid_targets_mean": 3299.1, |
| "valid_targets_min": 890 |
| }, |
| { |
| "epoch": 2.0104895104895104, |
| "grad_norm": 0.6240329555824098, |
| "learning_rate": 3.589489233655891e-05, |
| "loss": 0.3581, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3101613521575928, |
| "step": 1150, |
| "valid_targets_mean": 3488.4, |
| "valid_targets_min": 910 |
| }, |
| { |
| "epoch": 2.019230769230769, |
| "grad_norm": 0.5394247078220722, |
| "learning_rate": 3.5841819695540446e-05, |
| "loss": 0.35, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3253103196620941, |
| "step": 1155, |
| "valid_targets_mean": 3960.2, |
| "valid_targets_min": 1113 |
| }, |
| { |
| "epoch": 2.027972027972028, |
| "grad_norm": 0.5536721946245187, |
| "learning_rate": 3.5788445951157705e-05, |
| "loss": 0.2966, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2637212872505188, |
| "step": 1160, |
| "valid_targets_mean": 3050.2, |
| "valid_targets_min": 1087 |
| }, |
| { |
| "epoch": 2.0367132867132867, |
| "grad_norm": 0.5495101580283069, |
| "learning_rate": 3.5734772117878376e-05, |
| "loss": 0.3411, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.286973237991333, |
| "step": 1165, |
| "valid_targets_mean": 3388.9, |
| "valid_targets_min": 1085 |
| }, |
| { |
| "epoch": 2.0454545454545454, |
| "grad_norm": 0.7159443153475205, |
| "learning_rate": 3.568079921587388e-05, |
| "loss": 0.3789, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.396413117647171, |
| "step": 1170, |
| "valid_targets_mean": 4071.5, |
| "valid_targets_min": 1321 |
| }, |
| { |
| "epoch": 2.054195804195804, |
| "grad_norm": 0.5091694275857868, |
| "learning_rate": 3.5626528271000004e-05, |
| "loss": 0.3309, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26522356271743774, |
| "step": 1175, |
| "valid_targets_mean": 3710.6, |
| "valid_targets_min": 1186 |
| }, |
| { |
| "epoch": 2.062937062937063, |
| "grad_norm": 0.5494408643701906, |
| "learning_rate": 3.557196031477739e-05, |
| "loss": 0.3245, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26861274242401123, |
| "step": 1180, |
| "valid_targets_mean": 3189.2, |
| "valid_targets_min": 444 |
| }, |
| { |
| "epoch": 2.0716783216783217, |
| "grad_norm": 0.5642588272002523, |
| "learning_rate": 3.551709638437193e-05, |
| "loss": 0.3679, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4018964469432831, |
| "step": 1185, |
| "valid_targets_mean": 4603.7, |
| "valid_targets_min": 1275 |
| }, |
| { |
| "epoch": 2.0804195804195804, |
| "grad_norm": 0.6029948726748001, |
| "learning_rate": 3.546193752257509e-05, |
| "loss": 0.3367, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4047200679779053, |
| "step": 1190, |
| "valid_targets_mean": 3797.2, |
| "valid_targets_min": 1575 |
| }, |
| { |
| "epoch": 2.089160839160839, |
| "grad_norm": 0.6765562020479058, |
| "learning_rate": 3.540648477778401e-05, |
| "loss": 0.3233, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2479398250579834, |
| "step": 1195, |
| "valid_targets_mean": 2208.5, |
| "valid_targets_min": 864 |
| }, |
| { |
| "epoch": 2.097902097902098, |
| "grad_norm": 0.6399857417560751, |
| "learning_rate": 3.535073920398167e-05, |
| "loss": 0.3357, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3365986943244934, |
| "step": 1200, |
| "valid_targets_mean": 4008.9, |
| "valid_targets_min": 780 |
| }, |
| { |
| "epoch": 2.1066433566433567, |
| "grad_norm": 0.5423206989158699, |
| "learning_rate": 3.5294701860716773e-05, |
| "loss": 0.3223, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36652547121047974, |
| "step": 1205, |
| "valid_targets_mean": 4254.9, |
| "valid_targets_min": 1298 |
| }, |
| { |
| "epoch": 2.1153846153846154, |
| "grad_norm": 0.5548106930603924, |
| "learning_rate": 3.523837381308369e-05, |
| "loss": 0.3519, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.28331395983695984, |
| "step": 1210, |
| "valid_targets_mean": 3385.2, |
| "valid_targets_min": 814 |
| }, |
| { |
| "epoch": 2.124125874125874, |
| "grad_norm": 0.5212164148919388, |
| "learning_rate": 3.518175613170212e-05, |
| "loss": 0.3265, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.29387807846069336, |
| "step": 1215, |
| "valid_targets_mean": 3507.8, |
| "valid_targets_min": 286 |
| }, |
| { |
| "epoch": 2.132867132867133, |
| "grad_norm": 0.5973253477387896, |
| "learning_rate": 3.512484989269683e-05, |
| "loss": 0.3086, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33061522245407104, |
| "step": 1220, |
| "valid_targets_mean": 3305.0, |
| "valid_targets_min": 1372 |
| }, |
| { |
| "epoch": 2.1416083916083917, |
| "grad_norm": 0.6309218234838987, |
| "learning_rate": 3.506765617767715e-05, |
| "loss": 0.3305, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3732965588569641, |
| "step": 1225, |
| "valid_targets_mean": 3301.4, |
| "valid_targets_min": 721 |
| }, |
| { |
| "epoch": 2.1503496503496504, |
| "grad_norm": 0.5574659734718239, |
| "learning_rate": 3.501017607371644e-05, |
| "loss": 0.3393, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31696557998657227, |
| "step": 1230, |
| "valid_targets_mean": 3856.6, |
| "valid_targets_min": 785 |
| }, |
| { |
| "epoch": 2.159090909090909, |
| "grad_norm": 0.5329453873399369, |
| "learning_rate": 3.4952410673331394e-05, |
| "loss": 0.3115, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3600800335407257, |
| "step": 1235, |
| "valid_targets_mean": 4775.6, |
| "valid_targets_min": 749 |
| }, |
| { |
| "epoch": 2.167832167832168, |
| "grad_norm": 0.8114793310805336, |
| "learning_rate": 3.489436107446131e-05, |
| "loss": 0.3408, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3316611051559448, |
| "step": 1240, |
| "valid_targets_mean": 4332.1, |
| "valid_targets_min": 1436 |
| }, |
| { |
| "epoch": 2.1765734265734267, |
| "grad_norm": 0.6006878212399688, |
| "learning_rate": 3.483602838044722e-05, |
| "loss": 0.3194, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3003237247467041, |
| "step": 1245, |
| "valid_targets_mean": 3266.6, |
| "valid_targets_min": 1159 |
| }, |
| { |
| "epoch": 2.1853146853146854, |
| "grad_norm": 0.6353674050757039, |
| "learning_rate": 3.477741370001088e-05, |
| "loss": 0.3026, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36900120973587036, |
| "step": 1250, |
| "valid_targets_mean": 3345.0, |
| "valid_targets_min": 1196 |
| }, |
| { |
| "epoch": 2.194055944055944, |
| "grad_norm": 0.5737740743381646, |
| "learning_rate": 3.471851814723375e-05, |
| "loss": 0.3521, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3949390649795532, |
| "step": 1255, |
| "valid_targets_mean": 3930.3, |
| "valid_targets_min": 1197 |
| }, |
| { |
| "epoch": 2.202797202797203, |
| "grad_norm": 0.552663490793486, |
| "learning_rate": 3.4659342841535795e-05, |
| "loss": 0.3548, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.371884286403656, |
| "step": 1260, |
| "valid_targets_mean": 4117.4, |
| "valid_targets_min": 1068 |
| }, |
| { |
| "epoch": 2.2115384615384617, |
| "grad_norm": 0.6384782410157998, |
| "learning_rate": 3.459988890765418e-05, |
| "loss": 0.369, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3499044179916382, |
| "step": 1265, |
| "valid_targets_mean": 4232.2, |
| "valid_targets_min": 1426 |
| }, |
| { |
| "epoch": 2.2202797202797204, |
| "grad_norm": 0.5746738157396588, |
| "learning_rate": 3.454015747562194e-05, |
| "loss": 0.2951, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34578144550323486, |
| "step": 1270, |
| "valid_targets_mean": 3794.6, |
| "valid_targets_min": 1139 |
| }, |
| { |
| "epoch": 2.229020979020979, |
| "grad_norm": 0.5054542749499512, |
| "learning_rate": 3.448014968074648e-05, |
| "loss": 0.3199, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2880406081676483, |
| "step": 1275, |
| "valid_targets_mean": 4538.5, |
| "valid_targets_min": 1731 |
| }, |
| { |
| "epoch": 2.237762237762238, |
| "grad_norm": 0.5818466658704541, |
| "learning_rate": 3.4419866663587985e-05, |
| "loss": 0.3404, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3284459114074707, |
| "step": 1280, |
| "valid_targets_mean": 3571.0, |
| "valid_targets_min": 1487 |
| }, |
| { |
| "epoch": 2.2465034965034967, |
| "grad_norm": 0.4995232658017769, |
| "learning_rate": 3.4359309569937766e-05, |
| "loss": 0.3373, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.297016978263855, |
| "step": 1285, |
| "valid_targets_mean": 4168.3, |
| "valid_targets_min": 1398 |
| }, |
| { |
| "epoch": 2.2552447552447554, |
| "grad_norm": 0.5808033797273252, |
| "learning_rate": 3.429847955079646e-05, |
| "loss": 0.3706, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32967421412467957, |
| "step": 1290, |
| "valid_targets_mean": 3232.4, |
| "valid_targets_min": 933 |
| }, |
| { |
| "epoch": 2.263986013986014, |
| "grad_norm": 0.551203626753292, |
| "learning_rate": 3.4237377762352166e-05, |
| "loss": 0.3053, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.30884337425231934, |
| "step": 1295, |
| "valid_targets_mean": 3447.6, |
| "valid_targets_min": 1581 |
| }, |
| { |
| "epoch": 2.2727272727272725, |
| "grad_norm": 0.6298591338426267, |
| "learning_rate": 3.4176005365958484e-05, |
| "loss": 0.3233, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3573980927467346, |
| "step": 1300, |
| "valid_targets_mean": 3034.1, |
| "valid_targets_min": 870 |
| }, |
| { |
| "epoch": 2.2814685314685317, |
| "grad_norm": 0.543847122853011, |
| "learning_rate": 3.4114363528112376e-05, |
| "loss": 0.361, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35683882236480713, |
| "step": 1305, |
| "valid_targets_mean": 4214.3, |
| "valid_targets_min": 250 |
| }, |
| { |
| "epoch": 2.29020979020979, |
| "grad_norm": 0.5493972333043645, |
| "learning_rate": 3.40524534204321e-05, |
| "loss": 0.3097, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3217867314815521, |
| "step": 1310, |
| "valid_targets_mean": 4069.3, |
| "valid_targets_min": 1209 |
| }, |
| { |
| "epoch": 2.298951048951049, |
| "grad_norm": 0.5790482571574604, |
| "learning_rate": 3.399027621963484e-05, |
| "loss": 0.3362, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3702848255634308, |
| "step": 1315, |
| "valid_targets_mean": 3910.4, |
| "valid_targets_min": 1270 |
| }, |
| { |
| "epoch": 2.3076923076923075, |
| "grad_norm": 0.5521276510177341, |
| "learning_rate": 3.392783310751441e-05, |
| "loss": 0.342, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3059583306312561, |
| "step": 1320, |
| "valid_targets_mean": 3604.1, |
| "valid_targets_min": 783 |
| }, |
| { |
| "epoch": 2.3164335664335667, |
| "grad_norm": 0.6269207847914563, |
| "learning_rate": 3.3865125270918735e-05, |
| "loss": 0.3392, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.30475157499313354, |
| "step": 1325, |
| "valid_targets_mean": 2917.2, |
| "valid_targets_min": 340 |
| }, |
| { |
| "epoch": 2.325174825174825, |
| "grad_norm": 0.6239842100481497, |
| "learning_rate": 3.380215390172736e-05, |
| "loss": 0.3013, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23520943522453308, |
| "step": 1330, |
| "valid_targets_mean": 2649.5, |
| "valid_targets_min": 902 |
| }, |
| { |
| "epoch": 2.3339160839160837, |
| "grad_norm": 0.786197683842576, |
| "learning_rate": 3.373892019682874e-05, |
| "loss": 0.3944, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4770873785018921, |
| "step": 1335, |
| "valid_targets_mean": 5682.6, |
| "valid_targets_min": 857 |
| }, |
| { |
| "epoch": 2.3426573426573425, |
| "grad_norm": 0.5897989134852589, |
| "learning_rate": 3.36754253580975e-05, |
| "loss": 0.3396, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.41615045070648193, |
| "step": 1340, |
| "valid_targets_mean": 4481.4, |
| "valid_targets_min": 1026 |
| }, |
| { |
| "epoch": 2.3513986013986012, |
| "grad_norm": 0.5033226934658113, |
| "learning_rate": 3.361167059237162e-05, |
| "loss": 0.4161, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3581768870353699, |
| "step": 1345, |
| "valid_targets_mean": 4554.1, |
| "valid_targets_min": 1684 |
| }, |
| { |
| "epoch": 2.36013986013986, |
| "grad_norm": 0.6251621961865819, |
| "learning_rate": 3.354765711142946e-05, |
| "loss": 0.3507, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3330328166484833, |
| "step": 1350, |
| "valid_targets_mean": 3935.3, |
| "valid_targets_min": 826 |
| }, |
| { |
| "epoch": 2.3688811188811187, |
| "grad_norm": 0.543210210337472, |
| "learning_rate": 3.348338613196678e-05, |
| "loss": 0.3605, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.37580981850624084, |
| "step": 1355, |
| "valid_targets_mean": 4101.4, |
| "valid_targets_min": 1062 |
| }, |
| { |
| "epoch": 2.3776223776223775, |
| "grad_norm": 1.263978205574014, |
| "learning_rate": 3.341885887557353e-05, |
| "loss": 0.3516, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3562210500240326, |
| "step": 1360, |
| "valid_targets_mean": 3485.2, |
| "valid_targets_min": 1030 |
| }, |
| { |
| "epoch": 2.3863636363636362, |
| "grad_norm": 0.5891510875039799, |
| "learning_rate": 3.3354076568710715e-05, |
| "loss": 0.3245, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.29343318939208984, |
| "step": 1365, |
| "valid_targets_mean": 3082.8, |
| "valid_targets_min": 745 |
| }, |
| { |
| "epoch": 2.395104895104895, |
| "grad_norm": 0.6049119497188193, |
| "learning_rate": 3.328904044268705e-05, |
| "loss": 0.3184, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3105580806732178, |
| "step": 1370, |
| "valid_targets_mean": 3080.9, |
| "valid_targets_min": 774 |
| }, |
| { |
| "epoch": 2.4038461538461537, |
| "grad_norm": 0.5211424331279453, |
| "learning_rate": 3.322375173363554e-05, |
| "loss": 0.3088, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2609102129936218, |
| "step": 1375, |
| "valid_targets_mean": 3569.1, |
| "valid_targets_min": 1052 |
| }, |
| { |
| "epoch": 2.4125874125874125, |
| "grad_norm": 0.4570911075451314, |
| "learning_rate": 3.315821168249002e-05, |
| "loss": 0.3583, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34511393308639526, |
| "step": 1380, |
| "valid_targets_mean": 5819.4, |
| "valid_targets_min": 1308 |
| }, |
| { |
| "epoch": 2.4213286713286712, |
| "grad_norm": 0.6915681486907906, |
| "learning_rate": 3.309242153496154e-05, |
| "loss": 0.3087, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31641054153442383, |
| "step": 1385, |
| "valid_targets_mean": 2650.2, |
| "valid_targets_min": 1219 |
| }, |
| { |
| "epoch": 2.43006993006993, |
| "grad_norm": 0.5044835103913676, |
| "learning_rate": 3.3026382541514706e-05, |
| "loss": 0.36, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3088311553001404, |
| "step": 1390, |
| "valid_targets_mean": 4296.8, |
| "valid_targets_min": 1167 |
| }, |
| { |
| "epoch": 2.4388111888111887, |
| "grad_norm": 0.5352841239256858, |
| "learning_rate": 3.29600959573439e-05, |
| "loss": 0.3652, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3440842032432556, |
| "step": 1395, |
| "valid_targets_mean": 4173.2, |
| "valid_targets_min": 1070 |
| }, |
| { |
| "epoch": 2.4475524475524475, |
| "grad_norm": 0.5386197966301163, |
| "learning_rate": 3.289356304234942e-05, |
| "loss": 0.3521, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2736024558544159, |
| "step": 1400, |
| "valid_targets_mean": 3539.0, |
| "valid_targets_min": 942 |
| }, |
| { |
| "epoch": 2.4562937062937062, |
| "grad_norm": 0.5671969479135336, |
| "learning_rate": 3.2826785061113564e-05, |
| "loss": 0.3211, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3724323809146881, |
| "step": 1405, |
| "valid_targets_mean": 3864.0, |
| "valid_targets_min": 1121 |
| }, |
| { |
| "epoch": 2.465034965034965, |
| "grad_norm": 0.5827298416516434, |
| "learning_rate": 3.275976328287654e-05, |
| "loss": 0.3485, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.29571932554244995, |
| "step": 1410, |
| "valid_targets_mean": 3138.4, |
| "valid_targets_min": 915 |
| }, |
| { |
| "epoch": 2.4737762237762237, |
| "grad_norm": 0.5047580511144616, |
| "learning_rate": 3.269249898151241e-05, |
| "loss": 0.3293, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3644856810569763, |
| "step": 1415, |
| "valid_targets_mean": 4450.2, |
| "valid_targets_min": 1031 |
| }, |
| { |
| "epoch": 2.4825174825174825, |
| "grad_norm": 0.5540884026688192, |
| "learning_rate": 3.262499343550481e-05, |
| "loss": 0.333, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.30576711893081665, |
| "step": 1420, |
| "valid_targets_mean": 3474.9, |
| "valid_targets_min": 1176 |
| }, |
| { |
| "epoch": 2.4912587412587412, |
| "grad_norm": 0.6282355274552387, |
| "learning_rate": 3.25572479279227e-05, |
| "loss": 0.3045, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2186950147151947, |
| "step": 1425, |
| "valid_targets_mean": 2447.4, |
| "valid_targets_min": 1368 |
| }, |
| { |
| "epoch": 2.5, |
| "grad_norm": 0.46532594589288007, |
| "learning_rate": 3.2489263746395946e-05, |
| "loss": 0.3, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2158913016319275, |
| "step": 1430, |
| "valid_targets_mean": 4137.2, |
| "valid_targets_min": 1300 |
| }, |
| { |
| "epoch": 2.5087412587412588, |
| "grad_norm": 0.515624368170962, |
| "learning_rate": 3.242104218309087e-05, |
| "loss": 0.2997, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.311441570520401, |
| "step": 1435, |
| "valid_targets_mean": 4209.9, |
| "valid_targets_min": 1573 |
| }, |
| { |
| "epoch": 2.5174825174825175, |
| "grad_norm": 0.4745417190069869, |
| "learning_rate": 3.235258453468566e-05, |
| "loss": 0.3833, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4018992781639099, |
| "step": 1440, |
| "valid_targets_mean": 5670.8, |
| "valid_targets_min": 913 |
| }, |
| { |
| "epoch": 2.5262237762237763, |
| "grad_norm": 0.5077390838148781, |
| "learning_rate": 3.2283892102345753e-05, |
| "loss": 0.3265, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.335680216550827, |
| "step": 1445, |
| "valid_targets_mean": 4140.2, |
| "valid_targets_min": 1754 |
| }, |
| { |
| "epoch": 2.534965034965035, |
| "grad_norm": 0.6332018264920871, |
| "learning_rate": 3.2214966191699103e-05, |
| "loss": 0.3041, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21595916152000427, |
| "step": 1450, |
| "valid_targets_mean": 2705.9, |
| "valid_targets_min": 1154 |
| }, |
| { |
| "epoch": 2.5437062937062938, |
| "grad_norm": 0.48385050928663215, |
| "learning_rate": 3.2145808112811346e-05, |
| "loss": 0.3308, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2742727994918823, |
| "step": 1455, |
| "valid_targets_mean": 4247.0, |
| "valid_targets_min": 745 |
| }, |
| { |
| "epoch": 2.5524475524475525, |
| "grad_norm": 0.5577007979771508, |
| "learning_rate": 3.207641918016089e-05, |
| "loss": 0.3286, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3233272135257721, |
| "step": 1460, |
| "valid_targets_mean": 3605.4, |
| "valid_targets_min": 1827 |
| }, |
| { |
| "epoch": 2.5611888111888113, |
| "grad_norm": 0.5872688946896231, |
| "learning_rate": 3.200680071261399e-05, |
| "loss": 0.3435, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3883779048919678, |
| "step": 1465, |
| "valid_targets_mean": 4236.3, |
| "valid_targets_min": 1237 |
| }, |
| { |
| "epoch": 2.56993006993007, |
| "grad_norm": 0.5103628984175301, |
| "learning_rate": 3.1936954033399604e-05, |
| "loss": 0.3338, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3114474415779114, |
| "step": 1470, |
| "valid_targets_mean": 4111.7, |
| "valid_targets_min": 1671 |
| }, |
| { |
| "epoch": 2.5786713286713288, |
| "grad_norm": 0.540599739633964, |
| "learning_rate": 3.1866880470084295e-05, |
| "loss": 0.3184, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.346377968788147, |
| "step": 1475, |
| "valid_targets_mean": 3915.7, |
| "valid_targets_min": 1226 |
| }, |
| { |
| "epoch": 2.5874125874125875, |
| "grad_norm": 0.5182722420558209, |
| "learning_rate": 3.1796581354546986e-05, |
| "loss": 0.3422, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.365431010723114, |
| "step": 1480, |
| "valid_targets_mean": 4607.2, |
| "valid_targets_min": 1553 |
| }, |
| { |
| "epoch": 2.5961538461538463, |
| "grad_norm": 0.5325618536519481, |
| "learning_rate": 3.172605802295363e-05, |
| "loss": 0.3141, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3091772794723511, |
| "step": 1485, |
| "valid_targets_mean": 3804.9, |
| "valid_targets_min": 1206 |
| }, |
| { |
| "epoch": 2.604895104895105, |
| "grad_norm": 0.5080776916603383, |
| "learning_rate": 3.165531181573183e-05, |
| "loss": 0.3577, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.29155874252319336, |
| "step": 1490, |
| "valid_targets_mean": 3623.9, |
| "valid_targets_min": 1234 |
| }, |
| { |
| "epoch": 2.6136363636363638, |
| "grad_norm": 0.5493133795695918, |
| "learning_rate": 3.158434407754536e-05, |
| "loss": 0.3224, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.37722280621528625, |
| "step": 1495, |
| "valid_targets_mean": 4509.4, |
| "valid_targets_min": 1005 |
| }, |
| { |
| "epoch": 2.6223776223776225, |
| "grad_norm": 0.4760962427482821, |
| "learning_rate": 3.15131561572686e-05, |
| "loss": 0.3008, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33894044160842896, |
| "step": 1500, |
| "valid_targets_mean": 5020.0, |
| "valid_targets_min": 1178 |
| }, |
| { |
| "epoch": 2.6311188811188813, |
| "grad_norm": 0.5909610512631924, |
| "learning_rate": 3.14417494079609e-05, |
| "loss": 0.324, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2644537389278412, |
| "step": 1505, |
| "valid_targets_mean": 2790.8, |
| "valid_targets_min": 1458 |
| }, |
| { |
| "epoch": 2.63986013986014, |
| "grad_norm": 0.4867929709499423, |
| "learning_rate": 3.137012518684087e-05, |
| "loss": 0.335, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3499234914779663, |
| "step": 1510, |
| "valid_targets_mean": 5411.1, |
| "valid_targets_min": 844 |
| }, |
| { |
| "epoch": 2.6486013986013988, |
| "grad_norm": 0.6312052243485106, |
| "learning_rate": 3.129828485526056e-05, |
| "loss": 0.3713, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35967153310775757, |
| "step": 1515, |
| "valid_targets_mean": 4242.8, |
| "valid_targets_min": 1373 |
| }, |
| { |
| "epoch": 2.6573426573426575, |
| "grad_norm": 0.4836028803777569, |
| "learning_rate": 3.122622977867963e-05, |
| "loss": 0.3551, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3933538794517517, |
| "step": 1520, |
| "valid_targets_mean": 6016.8, |
| "valid_targets_min": 778 |
| }, |
| { |
| "epoch": 2.666083916083916, |
| "grad_norm": 0.48888711002835405, |
| "learning_rate": 3.1153961326639364e-05, |
| "loss": 0.3507, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3524247705936432, |
| "step": 1525, |
| "valid_targets_mean": 4931.2, |
| "valid_targets_min": 1005 |
| }, |
| { |
| "epoch": 2.674825174825175, |
| "grad_norm": 0.6089357100371079, |
| "learning_rate": 3.1081480872736625e-05, |
| "loss": 0.3383, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.44115176796913147, |
| "step": 1530, |
| "valid_targets_mean": 4462.6, |
| "valid_targets_min": 1632 |
| }, |
| { |
| "epoch": 2.6835664335664333, |
| "grad_norm": 0.4578837646445128, |
| "learning_rate": 3.1008789794597776e-05, |
| "loss": 0.338, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.351039856672287, |
| "step": 1535, |
| "valid_targets_mean": 5812.1, |
| "valid_targets_min": 1240 |
| }, |
| { |
| "epoch": 2.6923076923076925, |
| "grad_norm": 0.41101789198349736, |
| "learning_rate": 3.0935889473852504e-05, |
| "loss": 0.3161, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27018117904663086, |
| "step": 1540, |
| "valid_targets_mean": 5820.7, |
| "valid_targets_min": 937 |
| }, |
| { |
| "epoch": 2.701048951048951, |
| "grad_norm": 0.532122890158766, |
| "learning_rate": 3.086278129610752e-05, |
| "loss": 0.3155, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.39900821447372437, |
| "step": 1545, |
| "valid_targets_mean": 5061.4, |
| "valid_targets_min": 939 |
| }, |
| { |
| "epoch": 2.70979020979021, |
| "grad_norm": 0.558531282366923, |
| "learning_rate": 3.078946665092026e-05, |
| "loss": 0.3141, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24225279688835144, |
| "step": 1550, |
| "valid_targets_mean": 2957.4, |
| "valid_targets_min": 1131 |
| }, |
| { |
| "epoch": 2.7185314685314683, |
| "grad_norm": 0.5109631555447813, |
| "learning_rate": 3.071594693177245e-05, |
| "loss": 0.3675, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4015381336212158, |
| "step": 1555, |
| "valid_targets_mean": 4465.8, |
| "valid_targets_min": 828 |
| }, |
| { |
| "epoch": 2.7272727272727275, |
| "grad_norm": 0.536961938759069, |
| "learning_rate": 3.064222353604364e-05, |
| "loss": 0.3344, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.278196781873703, |
| "step": 1560, |
| "valid_targets_mean": 3323.1, |
| "valid_targets_min": 1201 |
| }, |
| { |
| "epoch": 2.736013986013986, |
| "grad_norm": 0.6246699552933006, |
| "learning_rate": 3.056829786498462e-05, |
| "loss": 0.3385, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3288848400115967, |
| "step": 1565, |
| "valid_targets_mean": 3083.0, |
| "valid_targets_min": 430 |
| }, |
| { |
| "epoch": 2.744755244755245, |
| "grad_norm": 0.606461060216546, |
| "learning_rate": 3.0494171323690806e-05, |
| "loss": 0.3127, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2629588842391968, |
| "step": 1570, |
| "valid_targets_mean": 2680.1, |
| "valid_targets_min": 1377 |
| }, |
| { |
| "epoch": 2.7534965034965033, |
| "grad_norm": 0.5680548667160676, |
| "learning_rate": 3.041984532107554e-05, |
| "loss": 0.3302, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23354268074035645, |
| "step": 1575, |
| "valid_targets_mean": 2878.9, |
| "valid_targets_min": 1391 |
| }, |
| { |
| "epoch": 2.762237762237762, |
| "grad_norm": 0.4888929142446612, |
| "learning_rate": 3.0345321269843277e-05, |
| "loss": 0.3359, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3905789256095886, |
| "step": 1580, |
| "valid_targets_mean": 5311.2, |
| "valid_targets_min": 894 |
| }, |
| { |
| "epoch": 2.770979020979021, |
| "grad_norm": 0.4913581565807214, |
| "learning_rate": 3.0270600586462778e-05, |
| "loss": 0.335, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3499818742275238, |
| "step": 1585, |
| "valid_targets_mean": 5184.2, |
| "valid_targets_min": 1497 |
| }, |
| { |
| "epoch": 2.7797202797202796, |
| "grad_norm": 0.6229148339552479, |
| "learning_rate": 3.019568469114015e-05, |
| "loss": 0.3432, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4142291843891144, |
| "step": 1590, |
| "valid_targets_mean": 3948.3, |
| "valid_targets_min": 1096 |
| }, |
| { |
| "epoch": 2.7884615384615383, |
| "grad_norm": 0.4869929533107764, |
| "learning_rate": 3.012057500779187e-05, |
| "loss": 0.3461, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3251883387565613, |
| "step": 1595, |
| "valid_targets_mean": 4624.6, |
| "valid_targets_min": 1304 |
| }, |
| { |
| "epoch": 2.797202797202797, |
| "grad_norm": 0.5826642414041437, |
| "learning_rate": 3.0045272964017716e-05, |
| "loss": 0.343, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36511659622192383, |
| "step": 1600, |
| "valid_targets_mean": 4380.1, |
| "valid_targets_min": 278 |
| }, |
| { |
| "epoch": 2.805944055944056, |
| "grad_norm": 0.519436625480804, |
| "learning_rate": 2.996977999107365e-05, |
| "loss": 0.3224, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34158188104629517, |
| "step": 1605, |
| "valid_targets_mean": 4013.6, |
| "valid_targets_min": 1457 |
| }, |
| { |
| "epoch": 2.8146853146853146, |
| "grad_norm": 0.5696670222694973, |
| "learning_rate": 2.989409752384458e-05, |
| "loss": 0.3291, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3094896972179413, |
| "step": 1610, |
| "valid_targets_mean": 3533.7, |
| "valid_targets_min": 991 |
| }, |
| { |
| "epoch": 2.8234265734265733, |
| "grad_norm": 0.5347638782980102, |
| "learning_rate": 2.9818227000817118e-05, |
| "loss": 0.3289, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.28824296593666077, |
| "step": 1615, |
| "valid_targets_mean": 3428.8, |
| "valid_targets_min": 1488 |
| }, |
| { |
| "epoch": 2.832167832167832, |
| "grad_norm": 0.5142694431369366, |
| "learning_rate": 2.9742169864052228e-05, |
| "loss": 0.3058, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33047711849212646, |
| "step": 1620, |
| "valid_targets_mean": 3887.8, |
| "valid_targets_min": 1383 |
| }, |
| { |
| "epoch": 2.840909090909091, |
| "grad_norm": 0.5251206238823921, |
| "learning_rate": 2.9665927559157806e-05, |
| "loss": 0.3305, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36617588996887207, |
| "step": 1625, |
| "valid_targets_mean": 5211.3, |
| "valid_targets_min": 900 |
| }, |
| { |
| "epoch": 2.8496503496503496, |
| "grad_norm": 0.6372098783365453, |
| "learning_rate": 2.9589501535261225e-05, |
| "loss": 0.3314, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3497511148452759, |
| "step": 1630, |
| "valid_targets_mean": 2946.0, |
| "valid_targets_min": 867 |
| }, |
| { |
| "epoch": 2.8583916083916083, |
| "grad_norm": 0.50501773040429, |
| "learning_rate": 2.9512893244981787e-05, |
| "loss": 0.3475, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32178816199302673, |
| "step": 1635, |
| "valid_targets_mean": 4174.2, |
| "valid_targets_min": 1037 |
| }, |
| { |
| "epoch": 2.867132867132867, |
| "grad_norm": 0.6092809892942609, |
| "learning_rate": 2.9436104144403077e-05, |
| "loss": 0.3566, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.37956398725509644, |
| "step": 1640, |
| "valid_targets_mean": 3658.4, |
| "valid_targets_min": 1280 |
| }, |
| { |
| "epoch": 2.875874125874126, |
| "grad_norm": 0.49333402495352063, |
| "learning_rate": 2.935913569304535e-05, |
| "loss": 0.3413, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31394535303115845, |
| "step": 1645, |
| "valid_targets_mean": 4416.7, |
| "valid_targets_min": 1183 |
| }, |
| { |
| "epoch": 2.8846153846153846, |
| "grad_norm": 0.6289070209634252, |
| "learning_rate": 2.9281989353837746e-05, |
| "loss": 0.3338, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2899363338947296, |
| "step": 1650, |
| "valid_targets_mean": 2902.4, |
| "valid_targets_min": 1459 |
| }, |
| { |
| "epoch": 2.8933566433566433, |
| "grad_norm": 0.889407629295206, |
| "learning_rate": 2.9204666593090497e-05, |
| "loss": 0.3377, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.40736573934555054, |
| "step": 1655, |
| "valid_targets_mean": 3440.2, |
| "valid_targets_min": 1080 |
| }, |
| { |
| "epoch": 2.902097902097902, |
| "grad_norm": 0.5530560757360455, |
| "learning_rate": 2.9127168880467046e-05, |
| "loss": 0.3024, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23936739563941956, |
| "step": 1660, |
| "valid_targets_mean": 2988.8, |
| "valid_targets_min": 1192 |
| }, |
| { |
| "epoch": 2.910839160839161, |
| "grad_norm": 0.4627722312735592, |
| "learning_rate": 2.904949768895613e-05, |
| "loss": 0.3629, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.38242220878601074, |
| "step": 1665, |
| "valid_targets_mean": 5788.4, |
| "valid_targets_min": 1672 |
| }, |
| { |
| "epoch": 2.9195804195804196, |
| "grad_norm": 0.5282100795632996, |
| "learning_rate": 2.8971654494843787e-05, |
| "loss": 0.3452, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3152114748954773, |
| "step": 1670, |
| "valid_targets_mean": 3547.1, |
| "valid_targets_min": 1059 |
| }, |
| { |
| "epoch": 2.9283216783216783, |
| "grad_norm": 0.5555470512379443, |
| "learning_rate": 2.8893640777685262e-05, |
| "loss": 0.3549, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2666780948638916, |
| "step": 1675, |
| "valid_targets_mean": 2986.2, |
| "valid_targets_min": 1327 |
| }, |
| { |
| "epoch": 2.937062937062937, |
| "grad_norm": 0.73170673641065, |
| "learning_rate": 2.8815458020276926e-05, |
| "loss": 0.3402, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34276720881462097, |
| "step": 1680, |
| "valid_targets_mean": 3467.1, |
| "valid_targets_min": 1189 |
| }, |
| { |
| "epoch": 2.945804195804196, |
| "grad_norm": 0.5551173357470257, |
| "learning_rate": 2.873710770862808e-05, |
| "loss": 0.2935, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.39830976724624634, |
| "step": 1685, |
| "valid_targets_mean": 4390.9, |
| "valid_targets_min": 1355 |
| }, |
| { |
| "epoch": 2.9545454545454546, |
| "grad_norm": 0.5429468572998486, |
| "learning_rate": 2.865859133193269e-05, |
| "loss": 0.3495, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35579365491867065, |
| "step": 1690, |
| "valid_targets_mean": 4104.4, |
| "valid_targets_min": 867 |
| }, |
| { |
| "epoch": 2.9632867132867133, |
| "grad_norm": 0.588815970444589, |
| "learning_rate": 2.857991038254111e-05, |
| "loss": 0.3557, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.29141175746917725, |
| "step": 1695, |
| "valid_targets_mean": 2931.9, |
| "valid_targets_min": 831 |
| }, |
| { |
| "epoch": 2.972027972027972, |
| "grad_norm": 0.530558546752831, |
| "learning_rate": 2.8501066355931706e-05, |
| "loss": 0.3188, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2682480812072754, |
| "step": 1700, |
| "valid_targets_mean": 3245.7, |
| "valid_targets_min": 923 |
| }, |
| { |
| "epoch": 2.980769230769231, |
| "grad_norm": 0.5237411811728068, |
| "learning_rate": 2.8422060750682415e-05, |
| "loss": 0.3317, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32749104499816895, |
| "step": 1705, |
| "valid_targets_mean": 3853.8, |
| "valid_targets_min": 1482 |
| }, |
| { |
| "epoch": 2.9895104895104896, |
| "grad_norm": 0.5573031127069092, |
| "learning_rate": 2.8342895068442294e-05, |
| "loss": 0.3687, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4529867172241211, |
| "step": 1710, |
| "valid_targets_mean": 5293.9, |
| "valid_targets_min": 880 |
| }, |
| { |
| "epoch": 2.9982517482517483, |
| "grad_norm": 0.5866186073439474, |
| "learning_rate": 2.8263570813902964e-05, |
| "loss": 0.3628, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4375874400138855, |
| "step": 1715, |
| "valid_targets_mean": 4802.0, |
| "valid_targets_min": 1091 |
| }, |
| { |
| "epoch": 3.006993006993007, |
| "grad_norm": 0.5593741909162024, |
| "learning_rate": 2.8184089494769996e-05, |
| "loss": 0.336, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.30689311027526855, |
| "step": 1720, |
| "valid_targets_mean": 3870.9, |
| "valid_targets_min": 1023 |
| }, |
| { |
| "epoch": 3.015734265734266, |
| "grad_norm": 0.6875702773552215, |
| "learning_rate": 2.8104452621734276e-05, |
| "loss": 0.2966, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24441856145858765, |
| "step": 1725, |
| "valid_targets_mean": 2328.2, |
| "valid_targets_min": 785 |
| }, |
| { |
| "epoch": 3.0244755244755246, |
| "grad_norm": 0.6701000204338885, |
| "learning_rate": 2.8024661708443283e-05, |
| "loss": 0.3066, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27847230434417725, |
| "step": 1730, |
| "valid_targets_mean": 2640.6, |
| "valid_targets_min": 1097 |
| }, |
| { |
| "epoch": 3.0332167832167833, |
| "grad_norm": 0.5561404654820966, |
| "learning_rate": 2.7944718271472307e-05, |
| "loss": 0.3239, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2966724634170532, |
| "step": 1735, |
| "valid_targets_mean": 3719.1, |
| "valid_targets_min": 1431 |
| }, |
| { |
| "epoch": 3.041958041958042, |
| "grad_norm": 0.5474562888952275, |
| "learning_rate": 2.7864623830295644e-05, |
| "loss": 0.2829, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2625727951526642, |
| "step": 1740, |
| "valid_targets_mean": 3559.9, |
| "valid_targets_min": 1339 |
| }, |
| { |
| "epoch": 3.050699300699301, |
| "grad_norm": 0.5467537174092281, |
| "learning_rate": 2.778437990725772e-05, |
| "loss": 0.3275, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32522091269493103, |
| "step": 1745, |
| "valid_targets_mean": 3952.1, |
| "valid_targets_min": 1306 |
| }, |
| { |
| "epoch": 3.0594405594405596, |
| "grad_norm": 0.5837961644003972, |
| "learning_rate": 2.770398802754411e-05, |
| "loss": 0.3025, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27994534373283386, |
| "step": 1750, |
| "valid_targets_mean": 3050.2, |
| "valid_targets_min": 1634 |
| }, |
| { |
| "epoch": 3.0681818181818183, |
| "grad_norm": 0.5722932789937027, |
| "learning_rate": 2.762344971915261e-05, |
| "loss": 0.2907, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3312224745750427, |
| "step": 1755, |
| "valid_targets_mean": 4157.8, |
| "valid_targets_min": 1501 |
| }, |
| { |
| "epoch": 3.076923076923077, |
| "grad_norm": 0.5310388929296251, |
| "learning_rate": 2.754276651286417e-05, |
| "loss": 0.3043, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2896951735019684, |
| "step": 1760, |
| "valid_targets_mean": 3732.2, |
| "valid_targets_min": 765 |
| }, |
| { |
| "epoch": 3.085664335664336, |
| "grad_norm": 0.5520914848544317, |
| "learning_rate": 2.7461939942213764e-05, |
| "loss": 0.3191, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2980251908302307, |
| "step": 1765, |
| "valid_targets_mean": 4002.0, |
| "valid_targets_min": 1123 |
| }, |
| { |
| "epoch": 3.0944055944055946, |
| "grad_norm": 0.6542055095053531, |
| "learning_rate": 2.7380971543461297e-05, |
| "loss": 0.3217, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.29861879348754883, |
| "step": 1770, |
| "valid_targets_mean": 5110.6, |
| "valid_targets_min": 1737 |
| }, |
| { |
| "epoch": 3.1031468531468533, |
| "grad_norm": 0.5594852097545516, |
| "learning_rate": 2.7299862855562388e-05, |
| "loss": 0.344, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35483282804489136, |
| "step": 1775, |
| "valid_targets_mean": 4066.8, |
| "valid_targets_min": 1142 |
| }, |
| { |
| "epoch": 3.111888111888112, |
| "grad_norm": 0.5118972220091799, |
| "learning_rate": 2.7218615420139087e-05, |
| "loss": 0.3169, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26075172424316406, |
| "step": 1780, |
| "valid_targets_mean": 3815.4, |
| "valid_targets_min": 1091 |
| }, |
| { |
| "epoch": 3.120629370629371, |
| "grad_norm": 0.5806431136298092, |
| "learning_rate": 2.7137230781450622e-05, |
| "loss": 0.2855, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22283409535884857, |
| "step": 1785, |
| "valid_targets_mean": 2682.1, |
| "valid_targets_min": 1586 |
| }, |
| { |
| "epoch": 3.129370629370629, |
| "grad_norm": 0.6888048962906062, |
| "learning_rate": 2.705571048636402e-05, |
| "loss": 0.324, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3880237936973572, |
| "step": 1790, |
| "valid_targets_mean": 4161.1, |
| "valid_targets_min": 865 |
| }, |
| { |
| "epoch": 3.138111888111888, |
| "grad_norm": 0.5897865617319268, |
| "learning_rate": 2.6974056084324714e-05, |
| "loss": 0.2689, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.19622839987277985, |
| "step": 1795, |
| "valid_targets_mean": 2572.1, |
| "valid_targets_min": 1043 |
| }, |
| { |
| "epoch": 3.1468531468531467, |
| "grad_norm": 0.5139881969438773, |
| "learning_rate": 2.689226912732708e-05, |
| "loss": 0.3195, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.310763955116272, |
| "step": 1800, |
| "valid_targets_mean": 4432.8, |
| "valid_targets_min": 1740 |
| }, |
| { |
| "epoch": 3.1555944055944054, |
| "grad_norm": 0.6327514647943825, |
| "learning_rate": 2.681035116988495e-05, |
| "loss": 0.2854, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2328225076198578, |
| "step": 1805, |
| "valid_targets_mean": 2745.8, |
| "valid_targets_min": 1203 |
| }, |
| { |
| "epoch": 3.164335664335664, |
| "grad_norm": 0.5786087357328902, |
| "learning_rate": 2.672830376900207e-05, |
| "loss": 0.2758, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21883933246135712, |
| "step": 1810, |
| "valid_targets_mean": 2616.9, |
| "valid_targets_min": 791 |
| }, |
| { |
| "epoch": 3.173076923076923, |
| "grad_norm": 0.5982326471685516, |
| "learning_rate": 2.6646128484142496e-05, |
| "loss": 0.2805, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2592069208621979, |
| "step": 1815, |
| "valid_targets_mean": 2841.3, |
| "valid_targets_min": 769 |
| }, |
| { |
| "epoch": 3.1818181818181817, |
| "grad_norm": 0.5077262037530513, |
| "learning_rate": 2.6563826877200973e-05, |
| "loss": 0.3337, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.305428147315979, |
| "step": 1820, |
| "valid_targets_mean": 4644.9, |
| "valid_targets_min": 870 |
| }, |
| { |
| "epoch": 3.1905594405594404, |
| "grad_norm": 0.5277249384140703, |
| "learning_rate": 2.6481400512473216e-05, |
| "loss": 0.3314, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3702384829521179, |
| "step": 1825, |
| "valid_targets_mean": 4858.6, |
| "valid_targets_min": 1258 |
| }, |
| { |
| "epoch": 3.199300699300699, |
| "grad_norm": 0.5934305740478432, |
| "learning_rate": 2.6398850956626185e-05, |
| "loss": 0.3159, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3127822279930115, |
| "step": 1830, |
| "valid_targets_mean": 3075.4, |
| "valid_targets_min": 1248 |
| }, |
| { |
| "epoch": 3.208041958041958, |
| "grad_norm": 0.5556627340017823, |
| "learning_rate": 2.6316179778668342e-05, |
| "loss": 0.2654, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27353787422180176, |
| "step": 1835, |
| "valid_targets_mean": 3647.0, |
| "valid_targets_min": 875 |
| }, |
| { |
| "epoch": 3.2167832167832167, |
| "grad_norm": 0.6769529052233687, |
| "learning_rate": 2.62333885499198e-05, |
| "loss": 0.2963, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2421294003725052, |
| "step": 1840, |
| "valid_targets_mean": 2436.8, |
| "valid_targets_min": 795 |
| }, |
| { |
| "epoch": 3.2255244755244754, |
| "grad_norm": 0.5674592394416891, |
| "learning_rate": 2.6150478843982442e-05, |
| "loss": 0.276, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2963680922985077, |
| "step": 1845, |
| "valid_targets_mean": 3300.4, |
| "valid_targets_min": 1277 |
| }, |
| { |
| "epoch": 3.234265734265734, |
| "grad_norm": 0.4952708038181234, |
| "learning_rate": 2.606745223671005e-05, |
| "loss": 0.3533, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4648565649986267, |
| "step": 1850, |
| "valid_targets_mean": 6469.8, |
| "valid_targets_min": 869 |
| }, |
| { |
| "epoch": 3.243006993006993, |
| "grad_norm": 0.503186087950448, |
| "learning_rate": 2.5984310306178326e-05, |
| "loss": 0.3144, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3806156516075134, |
| "step": 1855, |
| "valid_targets_mean": 5322.2, |
| "valid_targets_min": 1201 |
| }, |
| { |
| "epoch": 3.2517482517482517, |
| "grad_norm": 0.5743530413508324, |
| "learning_rate": 2.5901054632654915e-05, |
| "loss": 0.3232, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3227553963661194, |
| "step": 1860, |
| "valid_targets_mean": 3967.1, |
| "valid_targets_min": 911 |
| }, |
| { |
| "epoch": 3.2604895104895104, |
| "grad_norm": 0.5653915692401206, |
| "learning_rate": 2.5817686798569344e-05, |
| "loss": 0.3773, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.45867377519607544, |
| "step": 1865, |
| "valid_targets_mean": 5937.1, |
| "valid_targets_min": 1105 |
| }, |
| { |
| "epoch": 3.269230769230769, |
| "grad_norm": 0.5756883469118854, |
| "learning_rate": 2.5734208388482978e-05, |
| "loss": 0.3006, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24555999040603638, |
| "step": 1870, |
| "valid_targets_mean": 2718.0, |
| "valid_targets_min": 956 |
| }, |
| { |
| "epoch": 3.277972027972028, |
| "grad_norm": 0.5740807872776593, |
| "learning_rate": 2.565062098905887e-05, |
| "loss": 0.3458, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2976549565792084, |
| "step": 1875, |
| "valid_targets_mean": 3536.9, |
| "valid_targets_min": 1558 |
| }, |
| { |
| "epoch": 3.2867132867132867, |
| "grad_norm": 0.5351814557204507, |
| "learning_rate": 2.556692618903163e-05, |
| "loss": 0.2878, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23402386903762817, |
| "step": 1880, |
| "valid_targets_mean": 3553.0, |
| "valid_targets_min": 1245 |
| }, |
| { |
| "epoch": 3.2954545454545454, |
| "grad_norm": 0.5347095217029892, |
| "learning_rate": 2.548312557917721e-05, |
| "loss": 0.3092, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.37455683946609497, |
| "step": 1885, |
| "valid_targets_mean": 4354.1, |
| "valid_targets_min": 1197 |
| }, |
| { |
| "epoch": 3.304195804195804, |
| "grad_norm": 0.5087532787953783, |
| "learning_rate": 2.539922075228269e-05, |
| "loss": 0.3, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.28234240412712097, |
| "step": 1890, |
| "valid_targets_mean": 4121.2, |
| "valid_targets_min": 1014 |
| }, |
| { |
| "epoch": 3.312937062937063, |
| "grad_norm": 0.6003931803373139, |
| "learning_rate": 2.5315213303115965e-05, |
| "loss": 0.3228, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31622016429901123, |
| "step": 1895, |
| "valid_targets_mean": 3575.1, |
| "valid_targets_min": 1184 |
| }, |
| { |
| "epoch": 3.3216783216783217, |
| "grad_norm": 0.5324200090699127, |
| "learning_rate": 2.5231104828395482e-05, |
| "loss": 0.3376, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.43272459506988525, |
| "step": 1900, |
| "valid_targets_mean": 4967.6, |
| "valid_targets_min": 970 |
| }, |
| { |
| "epoch": 3.3304195804195804, |
| "grad_norm": 0.4644604979173125, |
| "learning_rate": 2.5146896926759846e-05, |
| "loss": 0.3405, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26517045497894287, |
| "step": 1905, |
| "valid_targets_mean": 4802.2, |
| "valid_targets_min": 465 |
| }, |
| { |
| "epoch": 3.339160839160839, |
| "grad_norm": 0.5567107972295677, |
| "learning_rate": 2.506259119873747e-05, |
| "loss": 0.2874, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2603760361671448, |
| "step": 1910, |
| "valid_targets_mean": 3453.2, |
| "valid_targets_min": 1046 |
| }, |
| { |
| "epoch": 3.347902097902098, |
| "grad_norm": 0.5824511685174325, |
| "learning_rate": 2.4978189246716138e-05, |
| "loss": 0.3317, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3032684624195099, |
| "step": 1915, |
| "valid_targets_mean": 3675.1, |
| "valid_targets_min": 1394 |
| }, |
| { |
| "epoch": 3.3566433566433567, |
| "grad_norm": 0.562852289563815, |
| "learning_rate": 2.4893692674912546e-05, |
| "loss": 0.3148, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27050256729125977, |
| "step": 1920, |
| "valid_targets_mean": 3446.0, |
| "valid_targets_min": 1239 |
| }, |
| { |
| "epoch": 3.3653846153846154, |
| "grad_norm": 0.4985979591826673, |
| "learning_rate": 2.480910308934182e-05, |
| "loss": 0.3332, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3425469994544983, |
| "step": 1925, |
| "valid_targets_mean": 5447.1, |
| "valid_targets_min": 1395 |
| }, |
| { |
| "epoch": 3.374125874125874, |
| "grad_norm": 0.5319686306642291, |
| "learning_rate": 2.4724422097787e-05, |
| "loss": 0.3131, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33209139108657837, |
| "step": 1930, |
| "valid_targets_mean": 4106.9, |
| "valid_targets_min": 1241 |
| }, |
| { |
| "epoch": 3.382867132867133, |
| "grad_norm": 0.5452257512149177, |
| "learning_rate": 2.463965130976843e-05, |
| "loss": 0.3127, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25301462411880493, |
| "step": 1935, |
| "valid_targets_mean": 3712.9, |
| "valid_targets_min": 1248 |
| }, |
| { |
| "epoch": 3.3916083916083917, |
| "grad_norm": 0.5994620335296995, |
| "learning_rate": 2.455479233651324e-05, |
| "loss": 0.3094, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3276280164718628, |
| "step": 1940, |
| "valid_targets_mean": 3709.6, |
| "valid_targets_min": 1363 |
| }, |
| { |
| "epoch": 3.4003496503496504, |
| "grad_norm": 0.6461501673593458, |
| "learning_rate": 2.4469846790924665e-05, |
| "loss": 0.3061, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.29547345638275146, |
| "step": 1945, |
| "valid_targets_mean": 2801.9, |
| "valid_targets_min": 942 |
| }, |
| { |
| "epoch": 3.409090909090909, |
| "grad_norm": 0.6637610589856805, |
| "learning_rate": 2.4384816287551417e-05, |
| "loss": 0.3183, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2376343160867691, |
| "step": 1950, |
| "valid_targets_mean": 3314.5, |
| "valid_targets_min": 844 |
| }, |
| { |
| "epoch": 3.417832167832168, |
| "grad_norm": 0.5781629714916717, |
| "learning_rate": 2.429970244255699e-05, |
| "loss": 0.3443, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4136093854904175, |
| "step": 1955, |
| "valid_targets_mean": 4240.6, |
| "valid_targets_min": 904 |
| }, |
| { |
| "epoch": 3.4265734265734267, |
| "grad_norm": 0.6348556096389305, |
| "learning_rate": 2.421450687368893e-05, |
| "loss": 0.3334, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25327175855636597, |
| "step": 1960, |
| "valid_targets_mean": 2663.6, |
| "valid_targets_min": 1247 |
| }, |
| { |
| "epoch": 3.4353146853146854, |
| "grad_norm": 0.5470950021821281, |
| "learning_rate": 2.4129231200248127e-05, |
| "loss": 0.2898, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31240352988243103, |
| "step": 1965, |
| "valid_targets_mean": 3655.2, |
| "valid_targets_min": 844 |
| }, |
| { |
| "epoch": 3.444055944055944, |
| "grad_norm": 0.545174136340182, |
| "learning_rate": 2.4043877043057967e-05, |
| "loss": 0.3037, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3076508343219757, |
| "step": 1970, |
| "valid_targets_mean": 3656.6, |
| "valid_targets_min": 1120 |
| }, |
| { |
| "epoch": 3.452797202797203, |
| "grad_norm": 0.548675450396162, |
| "learning_rate": 2.3958446024433594e-05, |
| "loss": 0.2763, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26705825328826904, |
| "step": 1975, |
| "valid_targets_mean": 3810.5, |
| "valid_targets_min": 1198 |
| }, |
| { |
| "epoch": 3.4615384615384617, |
| "grad_norm": 0.5451879746136353, |
| "learning_rate": 2.387293976815104e-05, |
| "loss": 0.2967, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31972768902778625, |
| "step": 1980, |
| "valid_targets_mean": 4197.5, |
| "valid_targets_min": 1083 |
| }, |
| { |
| "epoch": 3.4702797202797204, |
| "grad_norm": 0.5581621250486566, |
| "learning_rate": 2.3787359899416368e-05, |
| "loss": 0.3082, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3593602776527405, |
| "step": 1985, |
| "valid_targets_mean": 4509.0, |
| "valid_targets_min": 1340 |
| }, |
| { |
| "epoch": 3.479020979020979, |
| "grad_norm": 0.7071170505697403, |
| "learning_rate": 2.370170804483477e-05, |
| "loss": 0.3123, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.363255113363266, |
| "step": 1990, |
| "valid_targets_mean": 4484.3, |
| "valid_targets_min": 988 |
| }, |
| { |
| "epoch": 3.487762237762238, |
| "grad_norm": 0.5953191456261493, |
| "learning_rate": 2.3615985832379712e-05, |
| "loss": 0.3395, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.39135444164276123, |
| "step": 1995, |
| "valid_targets_mean": 4099.6, |
| "valid_targets_min": 838 |
| }, |
| { |
| "epoch": 3.4965034965034967, |
| "grad_norm": 0.6266903019082694, |
| "learning_rate": 2.353019489136188e-05, |
| "loss": 0.3184, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35590970516204834, |
| "step": 2000, |
| "valid_targets_mean": 3606.8, |
| "valid_targets_min": 934 |
| }, |
| { |
| "epoch": 3.505244755244755, |
| "grad_norm": 0.4764377794762096, |
| "learning_rate": 2.344433685239831e-05, |
| "loss": 0.329, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3431362509727478, |
| "step": 2005, |
| "valid_targets_mean": 6622.6, |
| "valid_targets_min": 1105 |
| }, |
| { |
| "epoch": 3.513986013986014, |
| "grad_norm": 0.6304170909454058, |
| "learning_rate": 2.335841334738136e-05, |
| "loss": 0.3164, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.29519712924957275, |
| "step": 2010, |
| "valid_targets_mean": 4467.6, |
| "valid_targets_min": 1355 |
| }, |
| { |
| "epoch": 3.5227272727272725, |
| "grad_norm": 0.490952302275404, |
| "learning_rate": 2.3272426009447672e-05, |
| "loss": 0.2906, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2910092771053314, |
| "step": 2015, |
| "valid_targets_mean": 5033.2, |
| "valid_targets_min": 1287 |
| }, |
| { |
| "epoch": 3.5314685314685317, |
| "grad_norm": 0.6155323428675108, |
| "learning_rate": 2.3186376472947168e-05, |
| "loss": 0.3064, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2902764678001404, |
| "step": 2020, |
| "valid_targets_mean": 2894.0, |
| "valid_targets_min": 740 |
| }, |
| { |
| "epoch": 3.54020979020979, |
| "grad_norm": 0.6025591175683395, |
| "learning_rate": 2.3100266373411974e-05, |
| "loss": 0.3453, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4039961099624634, |
| "step": 2025, |
| "valid_targets_mean": 4974.3, |
| "valid_targets_min": 1085 |
| }, |
| { |
| "epoch": 3.548951048951049, |
| "grad_norm": 0.6541300455848044, |
| "learning_rate": 2.3014097347525318e-05, |
| "loss": 0.3228, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3308113217353821, |
| "step": 2030, |
| "valid_targets_mean": 2943.2, |
| "valid_targets_min": 370 |
| }, |
| { |
| "epoch": 3.5576923076923075, |
| "grad_norm": 0.5772386978918906, |
| "learning_rate": 2.292787103309044e-05, |
| "loss": 0.313, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34256115555763245, |
| "step": 2035, |
| "valid_targets_mean": 3433.4, |
| "valid_targets_min": 1081 |
| }, |
| { |
| "epoch": 3.5664335664335667, |
| "grad_norm": 0.5104371056143377, |
| "learning_rate": 2.2841589068999458e-05, |
| "loss": 0.3112, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2974053621292114, |
| "step": 2040, |
| "valid_targets_mean": 4280.6, |
| "valid_targets_min": 1377 |
| }, |
| { |
| "epoch": 3.575174825174825, |
| "grad_norm": 0.537050676241048, |
| "learning_rate": 2.2755253095202197e-05, |
| "loss": 0.3329, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3554688096046448, |
| "step": 2045, |
| "valid_targets_mean": 4372.0, |
| "valid_targets_min": 1054 |
| }, |
| { |
| "epoch": 3.583916083916084, |
| "grad_norm": 0.5208129965095871, |
| "learning_rate": 2.2668864752675067e-05, |
| "loss": 0.2975, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36728352308273315, |
| "step": 2050, |
| "valid_targets_mean": 4518.3, |
| "valid_targets_min": 1222 |
| }, |
| { |
| "epoch": 3.5926573426573425, |
| "grad_norm": 0.5363774636487812, |
| "learning_rate": 2.2582425683389815e-05, |
| "loss": 0.3007, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2853674292564392, |
| "step": 2055, |
| "valid_targets_mean": 3562.8, |
| "valid_targets_min": 632 |
| }, |
| { |
| "epoch": 3.6013986013986012, |
| "grad_norm": 0.5865618188816509, |
| "learning_rate": 2.2495937530282368e-05, |
| "loss": 0.3391, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34035515785217285, |
| "step": 2060, |
| "valid_targets_mean": 4121.7, |
| "valid_targets_min": 778 |
| }, |
| { |
| "epoch": 3.61013986013986, |
| "grad_norm": 0.6378564206491916, |
| "learning_rate": 2.240940193722157e-05, |
| "loss": 0.3218, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33596569299697876, |
| "step": 2065, |
| "valid_targets_mean": 3544.1, |
| "valid_targets_min": 295 |
| }, |
| { |
| "epoch": 3.6188811188811187, |
| "grad_norm": 0.8804204695385472, |
| "learning_rate": 2.232282054897795e-05, |
| "loss": 0.3141, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3685414791107178, |
| "step": 2070, |
| "valid_targets_mean": 3786.0, |
| "valid_targets_min": 948 |
| }, |
| { |
| "epoch": 3.6276223776223775, |
| "grad_norm": 4.564250479592216, |
| "learning_rate": 2.2236195011192465e-05, |
| "loss": 0.309, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2899073362350464, |
| "step": 2075, |
| "valid_targets_mean": 3357.1, |
| "valid_targets_min": 1336 |
| }, |
| { |
| "epoch": 3.6363636363636362, |
| "grad_norm": 0.635900091215057, |
| "learning_rate": 2.2149526970345206e-05, |
| "loss": 0.3144, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.30170345306396484, |
| "step": 2080, |
| "valid_targets_mean": 2948.2, |
| "valid_targets_min": 730 |
| }, |
| { |
| "epoch": 3.645104895104895, |
| "grad_norm": 0.5582750079594184, |
| "learning_rate": 2.2062818073724132e-05, |
| "loss": 0.2988, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24713920056819916, |
| "step": 2085, |
| "valid_targets_mean": 3688.3, |
| "valid_targets_min": 1570 |
| }, |
| { |
| "epoch": 3.6538461538461537, |
| "grad_norm": 0.5603245581882423, |
| "learning_rate": 2.197606996939373e-05, |
| "loss": 0.3181, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3602690100669861, |
| "step": 2090, |
| "valid_targets_mean": 5358.8, |
| "valid_targets_min": 1278 |
| }, |
| { |
| "epoch": 3.6625874125874125, |
| "grad_norm": 0.4738594919894474, |
| "learning_rate": 2.1889284306163705e-05, |
| "loss": 0.2993, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.30577272176742554, |
| "step": 2095, |
| "valid_targets_mean": 4808.5, |
| "valid_targets_min": 408 |
| }, |
| { |
| "epoch": 3.6713286713286712, |
| "grad_norm": 0.6366766240922407, |
| "learning_rate": 2.1802462733557633e-05, |
| "loss": 0.3392, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3579673767089844, |
| "step": 2100, |
| "valid_targets_mean": 5506.0, |
| "valid_targets_min": 897 |
| }, |
| { |
| "epoch": 3.68006993006993, |
| "grad_norm": 0.6139697875303248, |
| "learning_rate": 2.1715606901781632e-05, |
| "loss": 0.3237, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33977657556533813, |
| "step": 2105, |
| "valid_targets_mean": 4469.7, |
| "valid_targets_min": 813 |
| }, |
| { |
| "epoch": 3.6888111888111887, |
| "grad_norm": 0.5475120095818063, |
| "learning_rate": 2.1628718461692958e-05, |
| "loss": 0.2856, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2815313935279846, |
| "step": 2110, |
| "valid_targets_mean": 3470.9, |
| "valid_targets_min": 1034 |
| }, |
| { |
| "epoch": 3.6975524475524475, |
| "grad_norm": 0.64206791153299, |
| "learning_rate": 2.1541799064768657e-05, |
| "loss": 0.2768, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27947911620140076, |
| "step": 2115, |
| "valid_targets_mean": 2544.1, |
| "valid_targets_min": 1200 |
| }, |
| { |
| "epoch": 3.7062937062937062, |
| "grad_norm": 0.5642192836007337, |
| "learning_rate": 2.1454850363074185e-05, |
| "loss": 0.3429, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26924845576286316, |
| "step": 2120, |
| "valid_targets_mean": 3421.4, |
| "valid_targets_min": 700 |
| }, |
| { |
| "epoch": 3.715034965034965, |
| "grad_norm": 0.5687312412041607, |
| "learning_rate": 2.1367874009231952e-05, |
| "loss": 0.3134, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3756811022758484, |
| "step": 2125, |
| "valid_targets_mean": 4092.6, |
| "valid_targets_min": 1091 |
| }, |
| { |
| "epoch": 3.7237762237762237, |
| "grad_norm": 0.5254319030668086, |
| "learning_rate": 2.128087165638999e-05, |
| "loss": 0.3382, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.295890212059021, |
| "step": 2130, |
| "valid_targets_mean": 4654.0, |
| "valid_targets_min": 1198 |
| }, |
| { |
| "epoch": 3.7325174825174825, |
| "grad_norm": 0.507350094446985, |
| "learning_rate": 2.119384495819047e-05, |
| "loss": 0.3169, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34149473905563354, |
| "step": 2135, |
| "valid_targets_mean": 4297.9, |
| "valid_targets_min": 286 |
| }, |
| { |
| "epoch": 3.7412587412587412, |
| "grad_norm": 0.5874873776278813, |
| "learning_rate": 2.1106795568738282e-05, |
| "loss": 0.3242, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.29470622539520264, |
| "step": 2140, |
| "valid_targets_mean": 3362.8, |
| "valid_targets_min": 1310 |
| }, |
| { |
| "epoch": 3.75, |
| "grad_norm": 0.6344178553130709, |
| "learning_rate": 2.1019725142569624e-05, |
| "loss": 0.3454, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3341315686702728, |
| "step": 2145, |
| "valid_targets_mean": 4382.1, |
| "valid_targets_min": 363 |
| }, |
| { |
| "epoch": 3.7587412587412588, |
| "grad_norm": 1.224454305795279, |
| "learning_rate": 2.0932635334620538e-05, |
| "loss": 0.3285, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3387570083141327, |
| "step": 2150, |
| "valid_targets_mean": 3175.1, |
| "valid_targets_min": 1080 |
| }, |
| { |
| "epoch": 3.7674825174825175, |
| "grad_norm": 0.4946950905437756, |
| "learning_rate": 2.084552780019543e-05, |
| "loss": 0.2918, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.261683851480484, |
| "step": 2155, |
| "valid_targets_mean": 4516.0, |
| "valid_targets_min": 740 |
| }, |
| { |
| "epoch": 3.7762237762237763, |
| "grad_norm": 0.5650252780634637, |
| "learning_rate": 2.075840419493565e-05, |
| "loss": 0.2844, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26833733916282654, |
| "step": 2160, |
| "valid_targets_mean": 3376.2, |
| "valid_targets_min": 885 |
| }, |
| { |
| "epoch": 3.784965034965035, |
| "grad_norm": 0.5530808312399973, |
| "learning_rate": 2.0671266174787993e-05, |
| "loss": 0.3636, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3434409201145172, |
| "step": 2165, |
| "valid_targets_mean": 3988.6, |
| "valid_targets_min": 1309 |
| }, |
| { |
| "epoch": 3.7937062937062938, |
| "grad_norm": 0.5785171812014174, |
| "learning_rate": 2.0584115395973255e-05, |
| "loss": 0.3515, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.41139060258865356, |
| "step": 2170, |
| "valid_targets_mean": 4776.5, |
| "valid_targets_min": 1094 |
| }, |
| { |
| "epoch": 3.8024475524475525, |
| "grad_norm": 0.45557991298431744, |
| "learning_rate": 2.049695351495472e-05, |
| "loss": 0.3466, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3612552285194397, |
| "step": 2175, |
| "valid_targets_mean": 5962.0, |
| "valid_targets_min": 1323 |
| }, |
| { |
| "epoch": 3.8111888111888113, |
| "grad_norm": 0.62555155861987, |
| "learning_rate": 2.0409782188406683e-05, |
| "loss": 0.3349, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27812838554382324, |
| "step": 2180, |
| "valid_targets_mean": 2643.1, |
| "valid_targets_min": 396 |
| }, |
| { |
| "epoch": 3.81993006993007, |
| "grad_norm": 0.6924523922525574, |
| "learning_rate": 2.032260307318298e-05, |
| "loss": 0.3308, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.29945462942123413, |
| "step": 2185, |
| "valid_targets_mean": 3498.6, |
| "valid_targets_min": 1210 |
| }, |
| { |
| "epoch": 3.8286713286713288, |
| "grad_norm": 0.6015653998125049, |
| "learning_rate": 2.0235417826285488e-05, |
| "loss": 0.2836, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2896122336387634, |
| "step": 2190, |
| "valid_targets_mean": 2937.8, |
| "valid_targets_min": 938 |
| }, |
| { |
| "epoch": 3.8374125874125875, |
| "grad_norm": 0.5861252074236171, |
| "learning_rate": 2.0148228104832614e-05, |
| "loss": 0.3066, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2773173153400421, |
| "step": 2195, |
| "valid_targets_mean": 3218.1, |
| "valid_targets_min": 1289 |
| }, |
| { |
| "epoch": 3.8461538461538463, |
| "grad_norm": 0.531620088389809, |
| "learning_rate": 2.0061035566027828e-05, |
| "loss": 0.323, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.29950347542762756, |
| "step": 2200, |
| "valid_targets_mean": 3860.1, |
| "valid_targets_min": 1100 |
| }, |
| { |
| "epoch": 3.854895104895105, |
| "grad_norm": 0.4886301358801487, |
| "learning_rate": 1.9973841867128137e-05, |
| "loss": 0.3281, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36873340606689453, |
| "step": 2205, |
| "valid_targets_mean": 5673.1, |
| "valid_targets_min": 437 |
| }, |
| { |
| "epoch": 3.8636363636363638, |
| "grad_norm": 0.5784230025210767, |
| "learning_rate": 1.9886648665412612e-05, |
| "loss": 0.3036, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36677759885787964, |
| "step": 2210, |
| "valid_targets_mean": 3799.9, |
| "valid_targets_min": 885 |
| }, |
| { |
| "epoch": 3.8723776223776225, |
| "grad_norm": 0.6969081154876361, |
| "learning_rate": 1.9799457618150847e-05, |
| "loss": 0.3197, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35714465379714966, |
| "step": 2215, |
| "valid_targets_mean": 3011.3, |
| "valid_targets_min": 857 |
| }, |
| { |
| "epoch": 3.8811188811188813, |
| "grad_norm": 0.597796061214715, |
| "learning_rate": 1.9712270382571527e-05, |
| "loss": 0.3395, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3203423023223877, |
| "step": 2220, |
| "valid_targets_mean": 3129.9, |
| "valid_targets_min": 783 |
| }, |
| { |
| "epoch": 3.88986013986014, |
| "grad_norm": 0.5086711274830522, |
| "learning_rate": 1.9625088615830852e-05, |
| "loss": 0.2971, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.30308815836906433, |
| "step": 2225, |
| "valid_targets_mean": 4242.3, |
| "valid_targets_min": 870 |
| }, |
| { |
| "epoch": 3.8986013986013988, |
| "grad_norm": 0.5822116633636045, |
| "learning_rate": 1.9537913974981093e-05, |
| "loss": 0.3211, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33314594626426697, |
| "step": 2230, |
| "valid_targets_mean": 3820.7, |
| "valid_targets_min": 2027 |
| }, |
| { |
| "epoch": 3.9073426573426575, |
| "grad_norm": 0.5836236703850111, |
| "learning_rate": 1.9450748116939076e-05, |
| "loss": 0.303, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3218787610530853, |
| "step": 2235, |
| "valid_targets_mean": 3663.4, |
| "valid_targets_min": 902 |
| }, |
| { |
| "epoch": 3.916083916083916, |
| "grad_norm": 0.5924349660287784, |
| "learning_rate": 1.936359269845471e-05, |
| "loss": 0.295, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.291001558303833, |
| "step": 2240, |
| "valid_targets_mean": 2916.3, |
| "valid_targets_min": 836 |
| }, |
| { |
| "epoch": 3.924825174825175, |
| "grad_norm": 0.5459663099084542, |
| "learning_rate": 1.927644937607944e-05, |
| "loss": 0.3512, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33792153000831604, |
| "step": 2245, |
| "valid_targets_mean": 4093.5, |
| "valid_targets_min": 1395 |
| }, |
| { |
| "epoch": 3.9335664335664333, |
| "grad_norm": 0.5193027045932389, |
| "learning_rate": 1.9189319806134847e-05, |
| "loss": 0.3393, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.28675055503845215, |
| "step": 2250, |
| "valid_targets_mean": 3711.7, |
| "valid_targets_min": 1220 |
| }, |
| { |
| "epoch": 3.9423076923076925, |
| "grad_norm": 0.5883356076136136, |
| "learning_rate": 1.91022056446811e-05, |
| "loss": 0.3205, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3391038179397583, |
| "step": 2255, |
| "valid_targets_mean": 3419.8, |
| "valid_targets_min": 840 |
| }, |
| { |
| "epoch": 3.951048951048951, |
| "grad_norm": 0.5586868273562965, |
| "learning_rate": 1.9015108547485494e-05, |
| "loss": 0.3069, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.262579083442688, |
| "step": 2260, |
| "valid_targets_mean": 3207.1, |
| "valid_targets_min": 919 |
| }, |
| { |
| "epoch": 3.95979020979021, |
| "grad_norm": 0.5554339434479842, |
| "learning_rate": 1.8928030169991006e-05, |
| "loss": 0.3223, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24667473137378693, |
| "step": 2265, |
| "valid_targets_mean": 3612.5, |
| "valid_targets_min": 1378 |
| }, |
| { |
| "epoch": 3.9685314685314683, |
| "grad_norm": 0.505303802479737, |
| "learning_rate": 1.8840972167284797e-05, |
| "loss": 0.3254, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35376450419425964, |
| "step": 2270, |
| "valid_targets_mean": 5061.0, |
| "valid_targets_min": 1318 |
| }, |
| { |
| "epoch": 3.9772727272727275, |
| "grad_norm": 0.5448435256913293, |
| "learning_rate": 1.8753936194066756e-05, |
| "loss": 0.2858, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31435316801071167, |
| "step": 2275, |
| "valid_targets_mean": 3888.8, |
| "valid_targets_min": 904 |
| }, |
| { |
| "epoch": 3.986013986013986, |
| "grad_norm": 0.5264911091118978, |
| "learning_rate": 1.8666923904618098e-05, |
| "loss": 0.3012, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27534744143486023, |
| "step": 2280, |
| "valid_targets_mean": 3623.6, |
| "valid_targets_min": 1612 |
| }, |
| { |
| "epoch": 3.994755244755245, |
| "grad_norm": 0.7118360207165516, |
| "learning_rate": 1.8579936952769842e-05, |
| "loss": 0.3431, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.30096369981765747, |
| "step": 2285, |
| "valid_targets_mean": 3708.7, |
| "valid_targets_min": 1137 |
| }, |
| { |
| "epoch": 4.003496503496503, |
| "grad_norm": 0.5351169874979324, |
| "learning_rate": 1.8492976991871432e-05, |
| "loss": 0.2969, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2072516679763794, |
| "step": 2290, |
| "valid_targets_mean": 2993.5, |
| "valid_targets_min": 1274 |
| }, |
| { |
| "epoch": 4.0122377622377625, |
| "grad_norm": 0.47553857464505833, |
| "learning_rate": 1.840604567475932e-05, |
| "loss": 0.2913, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2974964380264282, |
| "step": 2295, |
| "valid_targets_mean": 5104.4, |
| "valid_targets_min": 310 |
| }, |
| { |
| "epoch": 4.020979020979021, |
| "grad_norm": 0.4899707307146761, |
| "learning_rate": 1.8319144653725496e-05, |
| "loss": 0.2847, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23744121193885803, |
| "step": 2300, |
| "valid_targets_mean": 4422.4, |
| "valid_targets_min": 1360 |
| }, |
| { |
| "epoch": 4.02972027972028, |
| "grad_norm": 0.6081271010604331, |
| "learning_rate": 1.8232275580486132e-05, |
| "loss": 0.2903, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2675149440765381, |
| "step": 2305, |
| "valid_targets_mean": 3177.8, |
| "valid_targets_min": 758 |
| }, |
| { |
| "epoch": 4.038461538461538, |
| "grad_norm": 0.5181437168972942, |
| "learning_rate": 1.8145440106150193e-05, |
| "loss": 0.307, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31278491020202637, |
| "step": 2310, |
| "valid_targets_mean": 5048.5, |
| "valid_targets_min": 1347 |
| }, |
| { |
| "epoch": 4.0472027972027975, |
| "grad_norm": 0.6091159221103379, |
| "learning_rate": 1.8058639881188e-05, |
| "loss": 0.2788, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32723408937454224, |
| "step": 2315, |
| "valid_targets_mean": 3590.0, |
| "valid_targets_min": 1321 |
| }, |
| { |
| "epoch": 4.055944055944056, |
| "grad_norm": 0.5773222384218777, |
| "learning_rate": 1.7971876555399897e-05, |
| "loss": 0.2736, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3435593247413635, |
| "step": 2320, |
| "valid_targets_mean": 3994.2, |
| "valid_targets_min": 1003 |
| }, |
| { |
| "epoch": 4.064685314685315, |
| "grad_norm": 0.5617194273901538, |
| "learning_rate": 1.7885151777884923e-05, |
| "loss": 0.2886, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3240012526512146, |
| "step": 2325, |
| "valid_targets_mean": 4003.6, |
| "valid_targets_min": 1523 |
| }, |
| { |
| "epoch": 4.073426573426573, |
| "grad_norm": 0.5511985618275491, |
| "learning_rate": 1.77984671970094e-05, |
| "loss": 0.2761, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.29591992497444153, |
| "step": 2330, |
| "valid_targets_mean": 5243.3, |
| "valid_targets_min": 1455 |
| }, |
| { |
| "epoch": 4.0821678321678325, |
| "grad_norm": 0.593007324714538, |
| "learning_rate": 1.7711824460375652e-05, |
| "loss": 0.3108, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34284406900405884, |
| "step": 2335, |
| "valid_targets_mean": 3934.2, |
| "valid_targets_min": 1201 |
| }, |
| { |
| "epoch": 4.090909090909091, |
| "grad_norm": 0.4665478716876811, |
| "learning_rate": 1.7625225214790676e-05, |
| "loss": 0.319, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3358380198478699, |
| "step": 2340, |
| "valid_targets_mean": 6468.1, |
| "valid_targets_min": 1338 |
| }, |
| { |
| "epoch": 4.09965034965035, |
| "grad_norm": 0.6081324344603725, |
| "learning_rate": 1.7538671106234844e-05, |
| "loss": 0.2765, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22779546678066254, |
| "step": 2345, |
| "valid_targets_mean": 2993.1, |
| "valid_targets_min": 1087 |
| }, |
| { |
| "epoch": 4.108391608391608, |
| "grad_norm": 0.5694494355739524, |
| "learning_rate": 1.7452163779830594e-05, |
| "loss": 0.2949, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27797362208366394, |
| "step": 2350, |
| "valid_targets_mean": 3926.6, |
| "valid_targets_min": 1238 |
| }, |
| { |
| "epoch": 4.1171328671328675, |
| "grad_norm": 0.692642453035784, |
| "learning_rate": 1.7365704879811218e-05, |
| "loss": 0.2881, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.29587244987487793, |
| "step": 2355, |
| "valid_targets_mean": 3438.6, |
| "valid_targets_min": 894 |
| }, |
| { |
| "epoch": 4.125874125874126, |
| "grad_norm": 0.5884637093456763, |
| "learning_rate": 1.7279296049489534e-05, |
| "loss": 0.287, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.29776161909103394, |
| "step": 2360, |
| "valid_targets_mean": 3834.5, |
| "valid_targets_min": 1711 |
| }, |
| { |
| "epoch": 4.134615384615385, |
| "grad_norm": 0.8095854580575237, |
| "learning_rate": 1.7192938931226718e-05, |
| "loss": 0.3166, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3704269826412201, |
| "step": 2365, |
| "valid_targets_mean": 4851.2, |
| "valid_targets_min": 1429 |
| }, |
| { |
| "epoch": 4.143356643356643, |
| "grad_norm": 0.6470687556702546, |
| "learning_rate": 1.7106635166401055e-05, |
| "loss": 0.3019, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22572281956672668, |
| "step": 2370, |
| "valid_targets_mean": 2469.3, |
| "valid_targets_min": 812 |
| }, |
| { |
| "epoch": 4.1520979020979025, |
| "grad_norm": 0.5299244250517481, |
| "learning_rate": 1.7020386395376754e-05, |
| "loss": 0.2931, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3037104904651642, |
| "step": 2375, |
| "valid_targets_mean": 4737.6, |
| "valid_targets_min": 1518 |
| }, |
| { |
| "epoch": 4.160839160839161, |
| "grad_norm": 0.5734004469644143, |
| "learning_rate": 1.6934194257472742e-05, |
| "loss": 0.2802, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3412582576274872, |
| "step": 2380, |
| "valid_targets_mean": 4164.4, |
| "valid_targets_min": 904 |
| }, |
| { |
| "epoch": 4.16958041958042, |
| "grad_norm": 0.522302186747358, |
| "learning_rate": 1.6848060390931563e-05, |
| "loss": 0.2963, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31243306398391724, |
| "step": 2385, |
| "valid_targets_mean": 5573.2, |
| "valid_targets_min": 1552 |
| }, |
| { |
| "epoch": 4.178321678321678, |
| "grad_norm": 0.6004156669696765, |
| "learning_rate": 1.6761986432888176e-05, |
| "loss": 0.2997, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2502054274082184, |
| "step": 2390, |
| "valid_targets_mean": 3042.6, |
| "valid_targets_min": 765 |
| }, |
| { |
| "epoch": 4.187062937062937, |
| "grad_norm": 0.5233158597298586, |
| "learning_rate": 1.667597401933887e-05, |
| "loss": 0.2746, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.28923219442367554, |
| "step": 2395, |
| "valid_targets_mean": 4870.8, |
| "valid_targets_min": 1237 |
| }, |
| { |
| "epoch": 4.195804195804196, |
| "grad_norm": 0.5483627771622933, |
| "learning_rate": 1.6590024785110185e-05, |
| "loss": 0.3103, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3161337971687317, |
| "step": 2400, |
| "valid_targets_mean": 5001.0, |
| "valid_targets_min": 941 |
| }, |
| { |
| "epoch": 4.204545454545454, |
| "grad_norm": 0.5871866958014593, |
| "learning_rate": 1.6504140363827806e-05, |
| "loss": 0.302, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3478851020336151, |
| "step": 2405, |
| "valid_targets_mean": 3951.9, |
| "valid_targets_min": 882 |
| }, |
| { |
| "epoch": 4.213286713286713, |
| "grad_norm": 0.6006732063938374, |
| "learning_rate": 1.6418322387885516e-05, |
| "loss": 0.3246, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.29331815242767334, |
| "step": 2410, |
| "valid_targets_mean": 3227.8, |
| "valid_targets_min": 1113 |
| }, |
| { |
| "epoch": 4.222027972027972, |
| "grad_norm": 0.5349202137195906, |
| "learning_rate": 1.6332572488414217e-05, |
| "loss": 0.3081, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34923678636550903, |
| "step": 2415, |
| "valid_targets_mean": 4982.9, |
| "valid_targets_min": 852 |
| }, |
| { |
| "epoch": 4.230769230769231, |
| "grad_norm": 0.5312131062229603, |
| "learning_rate": 1.624689229525085e-05, |
| "loss": 0.3054, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3073064088821411, |
| "step": 2420, |
| "valid_targets_mean": 4661.3, |
| "valid_targets_min": 1189 |
| }, |
| { |
| "epoch": 4.239510489510489, |
| "grad_norm": 0.5799290945851939, |
| "learning_rate": 1.6161283436907466e-05, |
| "loss": 0.2829, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3483518362045288, |
| "step": 2425, |
| "valid_targets_mean": 4749.2, |
| "valid_targets_min": 1558 |
| }, |
| { |
| "epoch": 4.248251748251748, |
| "grad_norm": 0.5006339545460906, |
| "learning_rate": 1.6075747540540297e-05, |
| "loss": 0.3054, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2663540840148926, |
| "step": 2430, |
| "valid_targets_mean": 4513.2, |
| "valid_targets_min": 1177 |
| }, |
| { |
| "epoch": 4.256993006993007, |
| "grad_norm": 0.542580678933771, |
| "learning_rate": 1.599028623191875e-05, |
| "loss": 0.2768, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.274956613779068, |
| "step": 2435, |
| "valid_targets_mean": 4109.2, |
| "valid_targets_min": 1575 |
| }, |
| { |
| "epoch": 4.265734265734266, |
| "grad_norm": 0.5326860859143445, |
| "learning_rate": 1.5904901135394585e-05, |
| "loss": 0.2689, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2595234215259552, |
| "step": 2440, |
| "valid_targets_mean": 3658.4, |
| "valid_targets_min": 703 |
| }, |
| { |
| "epoch": 4.274475524475524, |
| "grad_norm": 0.5394753907754853, |
| "learning_rate": 1.5819593873871e-05, |
| "loss": 0.3204, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2550932466983795, |
| "step": 2445, |
| "valid_targets_mean": 3616.3, |
| "valid_targets_min": 1453 |
| }, |
| { |
| "epoch": 4.283216783216783, |
| "grad_norm": 0.5560535542529935, |
| "learning_rate": 1.5734366068771784e-05, |
| "loss": 0.2792, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27691853046417236, |
| "step": 2450, |
| "valid_targets_mean": 4063.1, |
| "valid_targets_min": 1196 |
| }, |
| { |
| "epoch": 4.291958041958042, |
| "grad_norm": 0.6200097958741395, |
| "learning_rate": 1.5649219340010526e-05, |
| "loss": 0.3189, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.326045423746109, |
| "step": 2455, |
| "valid_targets_mean": 3681.8, |
| "valid_targets_min": 1163 |
| }, |
| { |
| "epoch": 4.300699300699301, |
| "grad_norm": 0.5628210227789493, |
| "learning_rate": 1.5564155305959796e-05, |
| "loss": 0.2985, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31680306792259216, |
| "step": 2460, |
| "valid_targets_mean": 3843.2, |
| "valid_targets_min": 878 |
| }, |
| { |
| "epoch": 4.309440559440559, |
| "grad_norm": 0.5604933048616457, |
| "learning_rate": 1.547917558342039e-05, |
| "loss": 0.3369, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32571083307266235, |
| "step": 2465, |
| "valid_targets_mean": 4244.7, |
| "valid_targets_min": 1078 |
| }, |
| { |
| "epoch": 4.318181818181818, |
| "grad_norm": 0.5778075405917112, |
| "learning_rate": 1.5394281787590625e-05, |
| "loss": 0.2894, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2553318440914154, |
| "step": 2470, |
| "valid_targets_mean": 3426.7, |
| "valid_targets_min": 1366 |
| }, |
| { |
| "epoch": 4.326923076923077, |
| "grad_norm": 0.5995882800981643, |
| "learning_rate": 1.5309475532035602e-05, |
| "loss": 0.2853, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3515400290489197, |
| "step": 2475, |
| "valid_targets_mean": 4027.1, |
| "valid_targets_min": 986 |
| }, |
| { |
| "epoch": 4.335664335664336, |
| "grad_norm": 0.521959423096504, |
| "learning_rate": 1.5224758428656565e-05, |
| "loss": 0.2798, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2591817378997803, |
| "step": 2480, |
| "valid_targets_mean": 4302.8, |
| "valid_targets_min": 1045 |
| }, |
| { |
| "epoch": 4.344405594405594, |
| "grad_norm": 0.6208431532511489, |
| "learning_rate": 1.5140132087660267e-05, |
| "loss": 0.3251, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2892325520515442, |
| "step": 2485, |
| "valid_targets_mean": 3343.4, |
| "valid_targets_min": 824 |
| }, |
| { |
| "epoch": 4.353146853146853, |
| "grad_norm": 0.6314479692815503, |
| "learning_rate": 1.5055598117528327e-05, |
| "loss": 0.293, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34228602051734924, |
| "step": 2490, |
| "valid_targets_mean": 4748.9, |
| "valid_targets_min": 1244 |
| }, |
| { |
| "epoch": 4.361888111888112, |
| "grad_norm": 0.6147592728755509, |
| "learning_rate": 1.497115812498669e-05, |
| "loss": 0.3161, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26770299673080444, |
| "step": 2495, |
| "valid_targets_mean": 3033.6, |
| "valid_targets_min": 1366 |
| }, |
| { |
| "epoch": 4.370629370629371, |
| "grad_norm": 0.6046397551554065, |
| "learning_rate": 1.4886813714975098e-05, |
| "loss": 0.2505, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2695387601852417, |
| "step": 2500, |
| "valid_targets_mean": 3264.9, |
| "valid_targets_min": 1345 |
| }, |
| { |
| "epoch": 4.379370629370629, |
| "grad_norm": 0.500859303620456, |
| "learning_rate": 1.480256649061655e-05, |
| "loss": 0.3122, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3223142623901367, |
| "step": 2505, |
| "valid_targets_mean": 5378.9, |
| "valid_targets_min": 1083 |
| }, |
| { |
| "epoch": 4.388111888111888, |
| "grad_norm": 0.6865694931163415, |
| "learning_rate": 1.4718418053186854e-05, |
| "loss": 0.2789, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24780099093914032, |
| "step": 2510, |
| "valid_targets_mean": 2477.6, |
| "valid_targets_min": 816 |
| }, |
| { |
| "epoch": 4.396853146853147, |
| "grad_norm": 0.5867839077278701, |
| "learning_rate": 1.4634370002084205e-05, |
| "loss": 0.2746, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31326526403427124, |
| "step": 2515, |
| "valid_targets_mean": 3771.1, |
| "valid_targets_min": 838 |
| }, |
| { |
| "epoch": 4.405594405594406, |
| "grad_norm": 0.5978502595499985, |
| "learning_rate": 1.4550423934798747e-05, |
| "loss": 0.2771, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27199578285217285, |
| "step": 2520, |
| "valid_targets_mean": 3424.1, |
| "valid_targets_min": 1490 |
| }, |
| { |
| "epoch": 4.414335664335664, |
| "grad_norm": 0.5735035291809039, |
| "learning_rate": 1.4466581446882231e-05, |
| "loss": 0.2983, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34619852900505066, |
| "step": 2525, |
| "valid_targets_mean": 4088.4, |
| "valid_targets_min": 1412 |
| }, |
| { |
| "epoch": 4.423076923076923, |
| "grad_norm": 0.5692035115010636, |
| "learning_rate": 1.438284413191771e-05, |
| "loss": 0.2816, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2908937633037567, |
| "step": 2530, |
| "valid_targets_mean": 3680.6, |
| "valid_targets_min": 835 |
| }, |
| { |
| "epoch": 4.431818181818182, |
| "grad_norm": 0.6108968071141647, |
| "learning_rate": 1.4299213581489203e-05, |
| "loss": 0.3469, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.333668977022171, |
| "step": 2535, |
| "valid_targets_mean": 3268.6, |
| "valid_targets_min": 389 |
| }, |
| { |
| "epoch": 4.440559440559441, |
| "grad_norm": 0.6778129262816379, |
| "learning_rate": 1.421569138515149e-05, |
| "loss": 0.2812, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2540547549724579, |
| "step": 2540, |
| "valid_targets_mean": 2597.3, |
| "valid_targets_min": 1052 |
| }, |
| { |
| "epoch": 4.449300699300699, |
| "grad_norm": 0.4807213661570069, |
| "learning_rate": 1.4132279130399866e-05, |
| "loss": 0.2841, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.19758757948875427, |
| "step": 2545, |
| "valid_targets_mean": 3986.0, |
| "valid_targets_min": 851 |
| }, |
| { |
| "epoch": 4.458041958041958, |
| "grad_norm": 0.6287083054789264, |
| "learning_rate": 1.4048978402639995e-05, |
| "loss": 0.2608, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2614743709564209, |
| "step": 2550, |
| "valid_targets_mean": 2846.9, |
| "valid_targets_min": 396 |
| }, |
| { |
| "epoch": 4.466783216783217, |
| "grad_norm": 0.5079659942405071, |
| "learning_rate": 1.3965790785157733e-05, |
| "loss": 0.304, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2736564874649048, |
| "step": 2555, |
| "valid_targets_mean": 4273.1, |
| "valid_targets_min": 1275 |
| }, |
| { |
| "epoch": 4.475524475524476, |
| "grad_norm": 0.623480452513458, |
| "learning_rate": 1.388271785908911e-05, |
| "loss": 0.2822, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2957373559474945, |
| "step": 2560, |
| "valid_targets_mean": 3718.1, |
| "valid_targets_min": 923 |
| }, |
| { |
| "epoch": 4.484265734265734, |
| "grad_norm": 0.6577985176467535, |
| "learning_rate": 1.379976120339019e-05, |
| "loss": 0.3132, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.270229697227478, |
| "step": 2565, |
| "valid_targets_mean": 2898.3, |
| "valid_targets_min": 969 |
| }, |
| { |
| "epoch": 4.493006993006993, |
| "grad_norm": 0.5465264693819658, |
| "learning_rate": 1.3716922394807108e-05, |
| "loss": 0.3135, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.323173463344574, |
| "step": 2570, |
| "valid_targets_mean": 4383.8, |
| "valid_targets_min": 1359 |
| }, |
| { |
| "epoch": 4.501748251748252, |
| "grad_norm": 0.5661164401178533, |
| "learning_rate": 1.3634203007846123e-05, |
| "loss": 0.3073, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3571394085884094, |
| "step": 2575, |
| "valid_targets_mean": 4352.5, |
| "valid_targets_min": 1213 |
| }, |
| { |
| "epoch": 4.510489510489511, |
| "grad_norm": 0.6173442841080252, |
| "learning_rate": 1.3551604614743638e-05, |
| "loss": 0.2779, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2723095417022705, |
| "step": 2580, |
| "valid_targets_mean": 3077.6, |
| "valid_targets_min": 1212 |
| }, |
| { |
| "epoch": 4.519230769230769, |
| "grad_norm": 0.5285171225530177, |
| "learning_rate": 1.3469128785436336e-05, |
| "loss": 0.3152, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.39151817560195923, |
| "step": 2585, |
| "valid_targets_mean": 5290.2, |
| "valid_targets_min": 1008 |
| }, |
| { |
| "epoch": 4.527972027972028, |
| "grad_norm": 0.6424931675333562, |
| "learning_rate": 1.3386777087531378e-05, |
| "loss": 0.2614, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2340153455734253, |
| "step": 2590, |
| "valid_targets_mean": 3059.4, |
| "valid_targets_min": 787 |
| }, |
| { |
| "epoch": 4.536713286713287, |
| "grad_norm": 0.5536096774127396, |
| "learning_rate": 1.3304551086276542e-05, |
| "loss": 0.3235, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.360897421836853, |
| "step": 2595, |
| "valid_targets_mean": 4667.6, |
| "valid_targets_min": 904 |
| }, |
| { |
| "epoch": 4.545454545454545, |
| "grad_norm": 0.5522304667699259, |
| "learning_rate": 1.3222452344530514e-05, |
| "loss": 0.309, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3527144193649292, |
| "step": 2600, |
| "valid_targets_mean": 5043.0, |
| "valid_targets_min": 769 |
| }, |
| { |
| "epoch": 4.554195804195804, |
| "grad_norm": 0.5930207853002226, |
| "learning_rate": 1.3140482422733198e-05, |
| "loss": 0.2897, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21184858679771423, |
| "step": 2605, |
| "valid_targets_mean": 2679.9, |
| "valid_targets_min": 360 |
| }, |
| { |
| "epoch": 4.562937062937063, |
| "grad_norm": 0.4523824968127589, |
| "learning_rate": 1.3058642878876005e-05, |
| "loss": 0.305, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.400458425283432, |
| "step": 2610, |
| "valid_targets_mean": 6793.4, |
| "valid_targets_min": 794 |
| }, |
| { |
| "epoch": 4.571678321678322, |
| "grad_norm": 0.6660598737030216, |
| "learning_rate": 1.2976935268472282e-05, |
| "loss": 0.3031, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31761932373046875, |
| "step": 2615, |
| "valid_targets_mean": 2962.8, |
| "valid_targets_min": 788 |
| }, |
| { |
| "epoch": 4.58041958041958, |
| "grad_norm": 0.570251827846613, |
| "learning_rate": 1.289536114452775e-05, |
| "loss": 0.3376, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3196180462837219, |
| "step": 2620, |
| "valid_targets_mean": 4478.8, |
| "valid_targets_min": 1368 |
| }, |
| { |
| "epoch": 4.589160839160839, |
| "grad_norm": 0.7146176161503408, |
| "learning_rate": 1.2813922057510941e-05, |
| "loss": 0.3078, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35588014125823975, |
| "step": 2625, |
| "valid_targets_mean": 3962.6, |
| "valid_targets_min": 1574 |
| }, |
| { |
| "epoch": 4.597902097902098, |
| "grad_norm": 0.5910225666256981, |
| "learning_rate": 1.2732619555323771e-05, |
| "loss": 0.2962, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3604264259338379, |
| "step": 2630, |
| "valid_targets_mean": 4157.6, |
| "valid_targets_min": 836 |
| }, |
| { |
| "epoch": 4.606643356643357, |
| "grad_norm": 0.5876064972193442, |
| "learning_rate": 1.2651455183272123e-05, |
| "loss": 0.2977, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24652817845344543, |
| "step": 2635, |
| "valid_targets_mean": 3293.5, |
| "valid_targets_min": 1311 |
| }, |
| { |
| "epoch": 4.615384615384615, |
| "grad_norm": 0.6316151890865688, |
| "learning_rate": 1.2570430484036428e-05, |
| "loss": 0.2873, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3414528965950012, |
| "step": 2640, |
| "valid_targets_mean": 3320.4, |
| "valid_targets_min": 1186 |
| }, |
| { |
| "epoch": 4.624125874125874, |
| "grad_norm": 0.6231043964765773, |
| "learning_rate": 1.2489546997642394e-05, |
| "loss": 0.2988, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24330928921699524, |
| "step": 2645, |
| "valid_targets_mean": 3018.8, |
| "valid_targets_min": 1173 |
| }, |
| { |
| "epoch": 4.632867132867133, |
| "grad_norm": 0.5459532756435473, |
| "learning_rate": 1.2408806261431719e-05, |
| "loss": 0.307, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2778834402561188, |
| "step": 2650, |
| "valid_targets_mean": 3856.0, |
| "valid_targets_min": 841 |
| }, |
| { |
| "epoch": 4.641608391608392, |
| "grad_norm": 0.5138719298724235, |
| "learning_rate": 1.2328209810032852e-05, |
| "loss": 0.3059, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3052661418914795, |
| "step": 2655, |
| "valid_targets_mean": 4457.4, |
| "valid_targets_min": 1094 |
| }, |
| { |
| "epoch": 4.65034965034965, |
| "grad_norm": 0.7468748014300275, |
| "learning_rate": 1.2247759175331843e-05, |
| "loss": 0.3214, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2765350639820099, |
| "step": 2660, |
| "valid_targets_mean": 3678.1, |
| "valid_targets_min": 397 |
| }, |
| { |
| "epoch": 4.659090909090909, |
| "grad_norm": 0.568068341055875, |
| "learning_rate": 1.2167455886443244e-05, |
| "loss": 0.32, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33724042773246765, |
| "step": 2665, |
| "valid_targets_mean": 4406.8, |
| "valid_targets_min": 274 |
| }, |
| { |
| "epoch": 4.6678321678321675, |
| "grad_norm": 0.5369284700537394, |
| "learning_rate": 1.2087301469680992e-05, |
| "loss": 0.312, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3213573098182678, |
| "step": 2670, |
| "valid_targets_mean": 5050.0, |
| "valid_targets_min": 721 |
| }, |
| { |
| "epoch": 4.676573426573427, |
| "grad_norm": 0.5418193994362802, |
| "learning_rate": 1.2007297448529451e-05, |
| "loss": 0.3102, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35388627648353577, |
| "step": 2675, |
| "valid_targets_mean": 4565.1, |
| "valid_targets_min": 315 |
| }, |
| { |
| "epoch": 4.685314685314685, |
| "grad_norm": 0.5017207500392754, |
| "learning_rate": 1.1927445343614436e-05, |
| "loss": 0.2663, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27704697847366333, |
| "step": 2680, |
| "valid_targets_mean": 4477.9, |
| "valid_targets_min": 1565 |
| }, |
| { |
| "epoch": 4.694055944055944, |
| "grad_norm": 0.5363032891756204, |
| "learning_rate": 1.1847746672674308e-05, |
| "loss": 0.2716, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24790403246879578, |
| "step": 2685, |
| "valid_targets_mean": 3882.2, |
| "valid_targets_min": 984 |
| }, |
| { |
| "epoch": 4.7027972027972025, |
| "grad_norm": 0.5760992816533218, |
| "learning_rate": 1.1768202950531116e-05, |
| "loss": 0.2885, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3482862114906311, |
| "step": 2690, |
| "valid_targets_mean": 3853.1, |
| "valid_targets_min": 866 |
| }, |
| { |
| "epoch": 4.711538461538462, |
| "grad_norm": 0.5278073637246373, |
| "learning_rate": 1.1688815689061845e-05, |
| "loss": 0.3242, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3284091353416443, |
| "step": 2695, |
| "valid_targets_mean": 4456.2, |
| "valid_targets_min": 711 |
| }, |
| { |
| "epoch": 4.72027972027972, |
| "grad_norm": 0.46225936004312645, |
| "learning_rate": 1.160958639716963e-05, |
| "loss": 0.3021, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3566405773162842, |
| "step": 2700, |
| "valid_targets_mean": 6233.1, |
| "valid_targets_min": 1931 |
| }, |
| { |
| "epoch": 4.729020979020979, |
| "grad_norm": 0.6393556958936994, |
| "learning_rate": 1.1530516580755094e-05, |
| "loss": 0.2669, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23386213183403015, |
| "step": 2705, |
| "valid_targets_mean": 2735.7, |
| "valid_targets_min": 1235 |
| }, |
| { |
| "epoch": 4.7377622377622375, |
| "grad_norm": 0.7313544858772298, |
| "learning_rate": 1.145160774268776e-05, |
| "loss": 0.2866, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2925187945365906, |
| "step": 2710, |
| "valid_targets_mean": 3208.3, |
| "valid_targets_min": 760 |
| }, |
| { |
| "epoch": 4.746503496503497, |
| "grad_norm": 0.593415721475349, |
| "learning_rate": 1.1372861382777436e-05, |
| "loss": 0.3431, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27762988209724426, |
| "step": 2715, |
| "valid_targets_mean": 3455.1, |
| "valid_targets_min": 894 |
| }, |
| { |
| "epoch": 4.755244755244755, |
| "grad_norm": 0.6896039425675504, |
| "learning_rate": 1.1294278997745717e-05, |
| "loss": 0.2662, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27165350317955017, |
| "step": 2720, |
| "valid_targets_mean": 3064.6, |
| "valid_targets_min": 1453 |
| }, |
| { |
| "epoch": 4.763986013986014, |
| "grad_norm": 0.653532801402401, |
| "learning_rate": 1.1215862081197588e-05, |
| "loss": 0.2444, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22621586918830872, |
| "step": 2725, |
| "valid_targets_mean": 2338.2, |
| "valid_targets_min": 659 |
| }, |
| { |
| "epoch": 4.7727272727272725, |
| "grad_norm": 0.5579565197706448, |
| "learning_rate": 1.1137612123592955e-05, |
| "loss": 0.3407, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3474411964416504, |
| "step": 2730, |
| "valid_targets_mean": 4759.7, |
| "valid_targets_min": 1250 |
| }, |
| { |
| "epoch": 4.781468531468532, |
| "grad_norm": 0.6008079310769544, |
| "learning_rate": 1.1059530612218384e-05, |
| "loss": 0.2778, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24356213212013245, |
| "step": 2735, |
| "valid_targets_mean": 3257.1, |
| "valid_targets_min": 1154 |
| }, |
| { |
| "epoch": 4.79020979020979, |
| "grad_norm": 0.5625591929943424, |
| "learning_rate": 1.0981619031158805e-05, |
| "loss": 0.3002, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25547558069229126, |
| "step": 2740, |
| "valid_targets_mean": 3769.6, |
| "valid_targets_min": 1096 |
| }, |
| { |
| "epoch": 4.798951048951049, |
| "grad_norm": 0.5564650872223581, |
| "learning_rate": 1.0903878861269283e-05, |
| "loss": 0.285, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.39255663752555847, |
| "step": 2745, |
| "valid_targets_mean": 4813.8, |
| "valid_targets_min": 1190 |
| }, |
| { |
| "epoch": 4.8076923076923075, |
| "grad_norm": 0.5465111718947036, |
| "learning_rate": 1.0826311580146926e-05, |
| "loss": 0.2817, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31452226638793945, |
| "step": 2750, |
| "valid_targets_mean": 4456.3, |
| "valid_targets_min": 1608 |
| }, |
| { |
| "epoch": 4.816433566433567, |
| "grad_norm": 0.6002285372123841, |
| "learning_rate": 1.0748918662102738e-05, |
| "loss": 0.2903, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24080748856067657, |
| "step": 2755, |
| "valid_targets_mean": 2933.1, |
| "valid_targets_min": 776 |
| }, |
| { |
| "epoch": 4.825174825174825, |
| "grad_norm": 0.6031634744630441, |
| "learning_rate": 1.0671701578133638e-05, |
| "loss": 0.3093, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2960241734981537, |
| "step": 2760, |
| "valid_targets_mean": 3287.8, |
| "valid_targets_min": 1340 |
| }, |
| { |
| "epoch": 4.833916083916084, |
| "grad_norm": 0.5648244176335662, |
| "learning_rate": 1.0594661795894506e-05, |
| "loss": 0.3288, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3948041498661041, |
| "step": 2765, |
| "valid_targets_mean": 4577.2, |
| "valid_targets_min": 400 |
| }, |
| { |
| "epoch": 4.8426573426573425, |
| "grad_norm": 0.5897183525483226, |
| "learning_rate": 1.0517800779670246e-05, |
| "loss": 0.2831, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2534165680408478, |
| "step": 2770, |
| "valid_targets_mean": 2986.2, |
| "valid_targets_min": 885 |
| }, |
| { |
| "epoch": 4.851398601398602, |
| "grad_norm": 0.599383167273539, |
| "learning_rate": 1.044111999034799e-05, |
| "loss": 0.2741, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31281739473342896, |
| "step": 2775, |
| "valid_targets_mean": 3864.4, |
| "valid_targets_min": 599 |
| }, |
| { |
| "epoch": 4.86013986013986, |
| "grad_norm": 0.621675092661028, |
| "learning_rate": 1.0364620885389332e-05, |
| "loss": 0.2779, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2532220780849457, |
| "step": 2780, |
| "valid_targets_mean": 2882.8, |
| "valid_targets_min": 1129 |
| }, |
| { |
| "epoch": 4.868881118881119, |
| "grad_norm": 0.5174219008544854, |
| "learning_rate": 1.0288304918802601e-05, |
| "loss": 0.2974, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3275800347328186, |
| "step": 2785, |
| "valid_targets_mean": 5350.1, |
| "valid_targets_min": 1528 |
| }, |
| { |
| "epoch": 4.8776223776223775, |
| "grad_norm": 0.5564725618436285, |
| "learning_rate": 1.0212173541115238e-05, |
| "loss": 0.3181, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3417879641056061, |
| "step": 2790, |
| "valid_targets_mean": 4438.7, |
| "valid_targets_min": 1031 |
| }, |
| { |
| "epoch": 4.886363636363637, |
| "grad_norm": 0.5202279923981036, |
| "learning_rate": 1.013622819934625e-05, |
| "loss": 0.2873, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31056395173072815, |
| "step": 2795, |
| "valid_targets_mean": 5017.1, |
| "valid_targets_min": 1366 |
| }, |
| { |
| "epoch": 4.895104895104895, |
| "grad_norm": 0.8272229183082167, |
| "learning_rate": 1.0060470336978647e-05, |
| "loss": 0.2918, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2937367260456085, |
| "step": 2800, |
| "valid_targets_mean": 3412.6, |
| "valid_targets_min": 1314 |
| }, |
| { |
| "epoch": 4.903846153846154, |
| "grad_norm": 0.6023369863061019, |
| "learning_rate": 9.984901393932079e-06, |
| "loss": 0.3024, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3329635262489319, |
| "step": 2805, |
| "valid_targets_mean": 3690.2, |
| "valid_targets_min": 1103 |
| }, |
| { |
| "epoch": 4.9125874125874125, |
| "grad_norm": 0.5347626155440187, |
| "learning_rate": 9.909522806535419e-06, |
| "loss": 0.3002, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34003421664237976, |
| "step": 2810, |
| "valid_targets_mean": 4804.3, |
| "valid_targets_min": 1241 |
| }, |
| { |
| "epoch": 4.921328671328672, |
| "grad_norm": 0.5810834195037871, |
| "learning_rate": 9.834336007499466e-06, |
| "loss": 0.2925, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2992189824581146, |
| "step": 2815, |
| "valid_targets_mean": 3655.9, |
| "valid_targets_min": 363 |
| }, |
| { |
| "epoch": 4.93006993006993, |
| "grad_norm": 0.7516655603694848, |
| "learning_rate": 9.759342425889719e-06, |
| "loss": 0.3299, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26004546880722046, |
| "step": 2820, |
| "valid_targets_mean": 2960.8, |
| "valid_targets_min": 775 |
| }, |
| { |
| "epoch": 4.938811188811189, |
| "grad_norm": 0.5474762483923421, |
| "learning_rate": 9.684543487099252e-06, |
| "loss": 0.2925, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31097865104675293, |
| "step": 2825, |
| "valid_targets_mean": 4025.8, |
| "valid_targets_min": 1372 |
| }, |
| { |
| "epoch": 4.9475524475524475, |
| "grad_norm": 0.603868266770557, |
| "learning_rate": 9.609940612821553e-06, |
| "loss": 0.2902, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21819128096103668, |
| "step": 2830, |
| "valid_targets_mean": 2892.9, |
| "valid_targets_min": 843 |
| }, |
| { |
| "epoch": 4.956293706293707, |
| "grad_norm": 0.6179932621413688, |
| "learning_rate": 9.535535221023547e-06, |
| "loss": 0.2902, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2900503873825073, |
| "step": 2835, |
| "valid_targets_mean": 3740.0, |
| "valid_targets_min": 1399 |
| }, |
| { |
| "epoch": 4.965034965034965, |
| "grad_norm": 0.5718727937808842, |
| "learning_rate": 9.461328725918652e-06, |
| "loss": 0.3472, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32516181468963623, |
| "step": 2840, |
| "valid_targets_mean": 3534.2, |
| "valid_targets_min": 1482 |
| }, |
| { |
| "epoch": 4.973776223776223, |
| "grad_norm": 0.5820685648597708, |
| "learning_rate": 9.387322537939867e-06, |
| "loss": 0.2878, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23170748353004456, |
| "step": 2845, |
| "valid_targets_mean": 2971.9, |
| "valid_targets_min": 1151 |
| }, |
| { |
| "epoch": 4.9825174825174825, |
| "grad_norm": 0.6160327533376939, |
| "learning_rate": 9.313518063712977e-06, |
| "loss": 0.2766, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2837224304676056, |
| "step": 2850, |
| "valid_targets_mean": 4007.0, |
| "valid_targets_min": 977 |
| }, |
| { |
| "epoch": 4.991258741258742, |
| "grad_norm": 0.6590612654521167, |
| "learning_rate": 9.239916706029837e-06, |
| "loss": 0.2644, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2656211256980896, |
| "step": 2855, |
| "valid_targets_mean": 2824.0, |
| "valid_targets_min": 867 |
| }, |
| { |
| "epoch": 5.0, |
| "grad_norm": 0.487778840287654, |
| "learning_rate": 9.166519863821682e-06, |
| "loss": 0.2977, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3223125636577606, |
| "step": 2860, |
| "valid_targets_mean": 5592.4, |
| "valid_targets_min": 1554 |
| }, |
| { |
| "epoch": 5.008741258741258, |
| "grad_norm": 0.5515870319188418, |
| "learning_rate": 9.093328932132536e-06, |
| "loss": 0.3178, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.29514843225479126, |
| "step": 2865, |
| "valid_targets_mean": 3846.6, |
| "valid_targets_min": 977 |
| }, |
| { |
| "epoch": 5.0174825174825175, |
| "grad_norm": 0.6027515247820308, |
| "learning_rate": 9.02034530209274e-06, |
| "loss": 0.2657, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21801328659057617, |
| "step": 2870, |
| "valid_targets_mean": 3039.9, |
| "valid_targets_min": 868 |
| }, |
| { |
| "epoch": 5.026223776223776, |
| "grad_norm": 0.6604761384788594, |
| "learning_rate": 8.94757036089246e-06, |
| "loss": 0.2915, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.29287010431289673, |
| "step": 2875, |
| "valid_targets_mean": 3395.0, |
| "valid_targets_min": 703 |
| }, |
| { |
| "epoch": 5.034965034965035, |
| "grad_norm": 0.5186334303195224, |
| "learning_rate": 8.875005491755336e-06, |
| "loss": 0.3228, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.39983949065208435, |
| "step": 2880, |
| "valid_targets_mean": 5769.5, |
| "valid_targets_min": 780 |
| }, |
| { |
| "epoch": 5.043706293706293, |
| "grad_norm": 0.6726806564828863, |
| "learning_rate": 8.802652073912217e-06, |
| "loss": 0.2941, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2766971290111542, |
| "step": 2885, |
| "valid_targets_mean": 2916.9, |
| "valid_targets_min": 1247 |
| }, |
| { |
| "epoch": 5.0524475524475525, |
| "grad_norm": 0.6558280884607233, |
| "learning_rate": 8.730511482574905e-06, |
| "loss": 0.2988, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2496262788772583, |
| "step": 2890, |
| "valid_targets_mean": 2870.6, |
| "valid_targets_min": 1260 |
| }, |
| { |
| "epoch": 5.061188811188811, |
| "grad_norm": 0.5997253722708635, |
| "learning_rate": 8.658585088910034e-06, |
| "loss": 0.2845, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.30124104022979736, |
| "step": 2895, |
| "valid_targets_mean": 4088.7, |
| "valid_targets_min": 771 |
| }, |
| { |
| "epoch": 5.06993006993007, |
| "grad_norm": 0.6511353534883123, |
| "learning_rate": 8.586874260013043e-06, |
| "loss": 0.2655, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24419128894805908, |
| "step": 2900, |
| "valid_targets_mean": 3125.4, |
| "valid_targets_min": 882 |
| }, |
| { |
| "epoch": 5.078671328671328, |
| "grad_norm": 0.5500028690151533, |
| "learning_rate": 8.515380358882125e-06, |
| "loss": 0.2648, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.1970711648464203, |
| "step": 2905, |
| "valid_targets_mean": 3520.8, |
| "valid_targets_min": 844 |
| }, |
| { |
| "epoch": 5.0874125874125875, |
| "grad_norm": 1.1741959939280981, |
| "learning_rate": 8.444104744392368e-06, |
| "loss": 0.29, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3086763918399811, |
| "step": 2910, |
| "valid_targets_mean": 3857.3, |
| "valid_targets_min": 1625 |
| }, |
| { |
| "epoch": 5.096153846153846, |
| "grad_norm": 0.6282809967830072, |
| "learning_rate": 8.373048771269925e-06, |
| "loss": 0.2705, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32048705220222473, |
| "step": 2915, |
| "valid_targets_mean": 3851.1, |
| "valid_targets_min": 1160 |
| }, |
| { |
| "epoch": 5.104895104895105, |
| "grad_norm": 0.6350201505584189, |
| "learning_rate": 8.30221379006624e-06, |
| "loss": 0.2971, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2668588161468506, |
| "step": 2920, |
| "valid_targets_mean": 3230.1, |
| "valid_targets_min": 1067 |
| }, |
| { |
| "epoch": 5.113636363636363, |
| "grad_norm": 0.537977318451259, |
| "learning_rate": 8.231601147132395e-06, |
| "loss": 0.3058, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31173622608184814, |
| "step": 2925, |
| "valid_targets_mean": 4844.6, |
| "valid_targets_min": 833 |
| }, |
| { |
| "epoch": 5.1223776223776225, |
| "grad_norm": 0.6525436356749689, |
| "learning_rate": 8.161212184593533e-06, |
| "loss": 0.274, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2828228175640106, |
| "step": 2930, |
| "valid_targets_mean": 2888.5, |
| "valid_targets_min": 1244 |
| }, |
| { |
| "epoch": 5.131118881118881, |
| "grad_norm": 0.7670147857668765, |
| "learning_rate": 8.091048240323309e-06, |
| "loss": 0.2898, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3218318223953247, |
| "step": 2935, |
| "valid_targets_mean": 4409.9, |
| "valid_targets_min": 783 |
| }, |
| { |
| "epoch": 5.13986013986014, |
| "grad_norm": 0.5832403428104913, |
| "learning_rate": 8.021110647918506e-06, |
| "loss": 0.2927, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.28175467252731323, |
| "step": 2940, |
| "valid_targets_mean": 4289.4, |
| "valid_targets_min": 1588 |
| }, |
| { |
| "epoch": 5.148601398601398, |
| "grad_norm": 0.5913795547664308, |
| "learning_rate": 7.951400736673666e-06, |
| "loss": 0.2929, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2117207646369934, |
| "step": 2945, |
| "valid_targets_mean": 3373.2, |
| "valid_targets_min": 1135 |
| }, |
| { |
| "epoch": 5.1573426573426575, |
| "grad_norm": 0.6195775292187421, |
| "learning_rate": 7.881919831555812e-06, |
| "loss": 0.2425, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.19947922229766846, |
| "step": 2950, |
| "valid_targets_mean": 2951.7, |
| "valid_targets_min": 1245 |
| }, |
| { |
| "epoch": 5.166083916083916, |
| "grad_norm": 0.5629505536805914, |
| "learning_rate": 7.812669253179284e-06, |
| "loss": 0.2957, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2782512903213501, |
| "step": 2955, |
| "valid_targets_mean": 4214.4, |
| "valid_targets_min": 1349 |
| }, |
| { |
| "epoch": 5.174825174825175, |
| "grad_norm": 0.6510672774482675, |
| "learning_rate": 7.74365031778064e-06, |
| "loss": 0.2817, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2591198682785034, |
| "step": 2960, |
| "valid_targets_mean": 3180.0, |
| "valid_targets_min": 1215 |
| }, |
| { |
| "epoch": 5.183566433566433, |
| "grad_norm": 0.5026506652919864, |
| "learning_rate": 7.67486433719362e-06, |
| "loss": 0.2555, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2523643374443054, |
| "step": 2965, |
| "valid_targets_mean": 4377.1, |
| "valid_targets_min": 1551 |
| }, |
| { |
| "epoch": 5.1923076923076925, |
| "grad_norm": 0.5981654940081627, |
| "learning_rate": 7.606312618824221e-06, |
| "loss": 0.2983, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3656559884548187, |
| "step": 2970, |
| "valid_targets_mean": 4603.1, |
| "valid_targets_min": 1128 |
| }, |
| { |
| "epoch": 5.201048951048951, |
| "grad_norm": 5.001470852750849, |
| "learning_rate": 7.53799646562587e-06, |
| "loss": 0.2451, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26559704542160034, |
| "step": 2975, |
| "valid_targets_mean": 3457.1, |
| "valid_targets_min": 1008 |
| }, |
| { |
| "epoch": 5.20979020979021, |
| "grad_norm": 0.6154061160480875, |
| "learning_rate": 7.469917176074615e-06, |
| "loss": 0.2797, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.284348726272583, |
| "step": 2980, |
| "valid_targets_mean": 3541.6, |
| "valid_targets_min": 1072 |
| }, |
| { |
| "epoch": 5.218531468531468, |
| "grad_norm": 0.5767457725736881, |
| "learning_rate": 7.402076044144468e-06, |
| "loss": 0.2949, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.288425475358963, |
| "step": 2985, |
| "valid_targets_mean": 3888.5, |
| "valid_targets_min": 754 |
| }, |
| { |
| "epoch": 5.2272727272727275, |
| "grad_norm": 0.9484072946363189, |
| "learning_rate": 7.334474359282835e-06, |
| "loss": 0.3025, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2538584768772125, |
| "step": 2990, |
| "valid_targets_mean": 3242.8, |
| "valid_targets_min": 1171 |
| }, |
| { |
| "epoch": 5.236013986013986, |
| "grad_norm": 0.6182933263467078, |
| "learning_rate": 7.267113406385957e-06, |
| "loss": 0.2808, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.30061790347099304, |
| "step": 2995, |
| "valid_targets_mean": 3523.0, |
| "valid_targets_min": 700 |
| }, |
| { |
| "epoch": 5.244755244755245, |
| "grad_norm": 0.6353976287717851, |
| "learning_rate": 7.199994465774525e-06, |
| "loss": 0.2776, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2508491277694702, |
| "step": 3000, |
| "valid_targets_mean": 3439.9, |
| "valid_targets_min": 948 |
| }, |
| { |
| "epoch": 5.253496503496503, |
| "grad_norm": 0.5482075343729362, |
| "learning_rate": 7.133118813169333e-06, |
| "loss": 0.2788, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2493390440940857, |
| "step": 3005, |
| "valid_targets_mean": 4017.3, |
| "valid_targets_min": 281 |
| }, |
| { |
| "epoch": 5.2622377622377625, |
| "grad_norm": 0.6796942212521933, |
| "learning_rate": 7.066487719667048e-06, |
| "loss": 0.3051, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3333054184913635, |
| "step": 3010, |
| "valid_targets_mean": 3150.6, |
| "valid_targets_min": 1401 |
| }, |
| { |
| "epoch": 5.270979020979021, |
| "grad_norm": 0.5316665562656377, |
| "learning_rate": 7.000102451716002e-06, |
| "loss": 0.254, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.19675248861312866, |
| "step": 3015, |
| "valid_targets_mean": 4021.2, |
| "valid_targets_min": 1108 |
| }, |
| { |
| "epoch": 5.27972027972028, |
| "grad_norm": 0.601665131189074, |
| "learning_rate": 6.9339642710921865e-06, |
| "loss": 0.249, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2753317058086395, |
| "step": 3020, |
| "valid_targets_mean": 3542.6, |
| "valid_targets_min": 878 |
| }, |
| { |
| "epoch": 5.288461538461538, |
| "grad_norm": 0.6016387836746144, |
| "learning_rate": 6.868074434875207e-06, |
| "loss": 0.3308, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3341265916824341, |
| "step": 3025, |
| "valid_targets_mean": 4177.6, |
| "valid_targets_min": 1125 |
| }, |
| { |
| "epoch": 5.2972027972027975, |
| "grad_norm": 0.6319032564526934, |
| "learning_rate": 6.802434195424452e-06, |
| "loss": 0.3055, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.29447418451309204, |
| "step": 3030, |
| "valid_targets_mean": 3719.3, |
| "valid_targets_min": 913 |
| }, |
| { |
| "epoch": 5.305944055944056, |
| "grad_norm": 0.5916637424592456, |
| "learning_rate": 6.73704480035523e-06, |
| "loss": 0.3015, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2799910306930542, |
| "step": 3035, |
| "valid_targets_mean": 3499.6, |
| "valid_targets_min": 1323 |
| }, |
| { |
| "epoch": 5.314685314685315, |
| "grad_norm": 0.6561281929032633, |
| "learning_rate": 6.6719074925150885e-06, |
| "loss": 0.2858, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22917607426643372, |
| "step": 3040, |
| "valid_targets_mean": 3229.9, |
| "valid_targets_min": 1366 |
| }, |
| { |
| "epoch": 5.323426573426573, |
| "grad_norm": 0.5631601957129813, |
| "learning_rate": 6.607023509960207e-06, |
| "loss": 0.2806, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.30073410272598267, |
| "step": 3045, |
| "valid_targets_mean": 4102.4, |
| "valid_targets_min": 902 |
| }, |
| { |
| "epoch": 5.3321678321678325, |
| "grad_norm": 0.5939420651573594, |
| "learning_rate": 6.54239408593182e-06, |
| "loss": 0.2826, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.30191031098365784, |
| "step": 3050, |
| "valid_targets_mean": 4112.1, |
| "valid_targets_min": 1112 |
| }, |
| { |
| "epoch": 5.340909090909091, |
| "grad_norm": 0.62297242809634, |
| "learning_rate": 6.478020448832807e-06, |
| "loss": 0.2842, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26781389117240906, |
| "step": 3055, |
| "valid_targets_mean": 3714.4, |
| "valid_targets_min": 1393 |
| }, |
| { |
| "epoch": 5.34965034965035, |
| "grad_norm": 0.5551628821680462, |
| "learning_rate": 6.41390382220436e-06, |
| "loss": 0.3012, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3098997473716736, |
| "step": 3060, |
| "valid_targets_mean": 4681.3, |
| "valid_targets_min": 699 |
| }, |
| { |
| "epoch": 5.358391608391608, |
| "grad_norm": 0.5574581153427564, |
| "learning_rate": 6.350045424702678e-06, |
| "loss": 0.3141, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34914591908454895, |
| "step": 3065, |
| "valid_targets_mean": 5659.1, |
| "valid_targets_min": 1391 |
| }, |
| { |
| "epoch": 5.3671328671328675, |
| "grad_norm": 0.5862085004980788, |
| "learning_rate": 6.286446470075844e-06, |
| "loss": 0.2653, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24797584116458893, |
| "step": 3070, |
| "valid_targets_mean": 3695.2, |
| "valid_targets_min": 1317 |
| }, |
| { |
| "epoch": 5.375874125874126, |
| "grad_norm": 0.5191297805074992, |
| "learning_rate": 6.223108167140754e-06, |
| "loss": 0.292, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2748684585094452, |
| "step": 3075, |
| "valid_targets_mean": 4941.7, |
| "valid_targets_min": 1381 |
| }, |
| { |
| "epoch": 5.384615384615385, |
| "grad_norm": 0.5045700473089122, |
| "learning_rate": 6.1600317197601315e-06, |
| "loss": 0.3003, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3563288748264313, |
| "step": 3080, |
| "valid_targets_mean": 5475.4, |
| "valid_targets_min": 1462 |
| }, |
| { |
| "epoch": 5.393356643356643, |
| "grad_norm": 0.5378903669303158, |
| "learning_rate": 6.0972183268196226e-06, |
| "loss": 0.2657, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2671928405761719, |
| "step": 3085, |
| "valid_targets_mean": 4020.6, |
| "valid_targets_min": 984 |
| }, |
| { |
| "epoch": 5.402097902097902, |
| "grad_norm": 0.5843880944609243, |
| "learning_rate": 6.034669182205069e-06, |
| "loss": 0.2687, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27627307176589966, |
| "step": 3090, |
| "valid_targets_mean": 3800.9, |
| "valid_targets_min": 1235 |
| }, |
| { |
| "epoch": 5.410839160839161, |
| "grad_norm": 0.6602300244309561, |
| "learning_rate": 5.972385474779749e-06, |
| "loss": 0.2679, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2647150754928589, |
| "step": 3095, |
| "valid_targets_mean": 2855.0, |
| "valid_targets_min": 843 |
| }, |
| { |
| "epoch": 5.41958041958042, |
| "grad_norm": 0.6462995998231272, |
| "learning_rate": 5.910368388361818e-06, |
| "loss": 0.2926, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26018214225769043, |
| "step": 3100, |
| "valid_targets_mean": 3224.8, |
| "valid_targets_min": 1357 |
| }, |
| { |
| "epoch": 5.428321678321678, |
| "grad_norm": 0.4891544819132749, |
| "learning_rate": 5.84861910170182e-06, |
| "loss": 0.315, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2529061734676361, |
| "step": 3105, |
| "valid_targets_mean": 5225.0, |
| "valid_targets_min": 1269 |
| }, |
| { |
| "epoch": 5.437062937062937, |
| "grad_norm": 0.5575392553853251, |
| "learning_rate": 5.787138788460243e-06, |
| "loss": 0.2638, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27974453568458557, |
| "step": 3110, |
| "valid_targets_mean": 4549.4, |
| "valid_targets_min": 1333 |
| }, |
| { |
| "epoch": 5.445804195804196, |
| "grad_norm": 0.9169550213827106, |
| "learning_rate": 5.725928617185232e-06, |
| "loss": 0.3004, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36411306262016296, |
| "step": 3115, |
| "valid_targets_mean": 4993.6, |
| "valid_targets_min": 465 |
| }, |
| { |
| "epoch": 5.454545454545454, |
| "grad_norm": 0.6333429716490787, |
| "learning_rate": 5.664989751290404e-06, |
| "loss": 0.2704, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2409021258354187, |
| "step": 3120, |
| "valid_targets_mean": 2938.7, |
| "valid_targets_min": 1156 |
| }, |
| { |
| "epoch": 5.463286713286713, |
| "grad_norm": 0.6164781154770047, |
| "learning_rate": 5.604323349032688e-06, |
| "loss": 0.2872, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3192230761051178, |
| "step": 3125, |
| "valid_targets_mean": 3691.0, |
| "valid_targets_min": 1022 |
| }, |
| { |
| "epoch": 5.472027972027972, |
| "grad_norm": 0.673990836348822, |
| "learning_rate": 5.543930563490329e-06, |
| "loss": 0.3229, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2710210084915161, |
| "step": 3130, |
| "valid_targets_mean": 2789.3, |
| "valid_targets_min": 1063 |
| }, |
| { |
| "epoch": 5.480769230769231, |
| "grad_norm": 0.6255943702870599, |
| "learning_rate": 5.483812542541005e-06, |
| "loss": 0.2877, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26079046726226807, |
| "step": 3135, |
| "valid_targets_mean": 4804.1, |
| "valid_targets_min": 1268 |
| }, |
| { |
| "epoch": 5.489510489510489, |
| "grad_norm": 0.6091977606764691, |
| "learning_rate": 5.423970428839944e-06, |
| "loss": 0.2692, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27790728211402893, |
| "step": 3140, |
| "valid_targets_mean": 3833.6, |
| "valid_targets_min": 1655 |
| }, |
| { |
| "epoch": 5.498251748251748, |
| "grad_norm": 0.6667888999428736, |
| "learning_rate": 5.364405359798273e-06, |
| "loss": 0.2559, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25091731548309326, |
| "step": 3145, |
| "valid_targets_mean": 2849.9, |
| "valid_targets_min": 1083 |
| }, |
| { |
| "epoch": 5.506993006993007, |
| "grad_norm": 0.7128463727312214, |
| "learning_rate": 5.305118467561337e-06, |
| "loss": 0.2655, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24288232624530792, |
| "step": 3150, |
| "valid_targets_mean": 2881.0, |
| "valid_targets_min": 841 |
| }, |
| { |
| "epoch": 5.515734265734266, |
| "grad_norm": 0.5707716004157022, |
| "learning_rate": 5.246110878987243e-06, |
| "loss": 0.2725, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26704850792884827, |
| "step": 3155, |
| "valid_targets_mean": 4202.4, |
| "valid_targets_min": 1416 |
| }, |
| { |
| "epoch": 5.524475524475524, |
| "grad_norm": 0.5654249158794692, |
| "learning_rate": 5.187383715625382e-06, |
| "loss": 0.2795, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24396446347236633, |
| "step": 3160, |
| "valid_targets_mean": 3794.2, |
| "valid_targets_min": 970 |
| }, |
| { |
| "epoch": 5.533216783216783, |
| "grad_norm": 0.5007607833940395, |
| "learning_rate": 5.128938093695166e-06, |
| "loss": 0.2723, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3158084750175476, |
| "step": 3165, |
| "valid_targets_mean": 5185.2, |
| "valid_targets_min": 1196 |
| }, |
| { |
| "epoch": 5.541958041958042, |
| "grad_norm": 0.6987253561806125, |
| "learning_rate": 5.0707751240647665e-06, |
| "loss": 0.2632, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2553308308124542, |
| "step": 3170, |
| "valid_targets_mean": 2831.2, |
| "valid_targets_min": 1078 |
| }, |
| { |
| "epoch": 5.550699300699301, |
| "grad_norm": 0.6693739024014761, |
| "learning_rate": 5.012895912230021e-06, |
| "loss": 0.2662, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2554604411125183, |
| "step": 3175, |
| "valid_targets_mean": 3050.4, |
| "valid_targets_min": 437 |
| }, |
| { |
| "epoch": 5.559440559440559, |
| "grad_norm": 0.5729662992883009, |
| "learning_rate": 4.955301558293439e-06, |
| "loss": 0.2752, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.237440288066864, |
| "step": 3180, |
| "valid_targets_mean": 3831.5, |
| "valid_targets_min": 1266 |
| }, |
| { |
| "epoch": 5.568181818181818, |
| "grad_norm": 0.6086761990951153, |
| "learning_rate": 4.897993156943257e-06, |
| "loss": 0.2772, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26315420866012573, |
| "step": 3185, |
| "valid_targets_mean": 3630.5, |
| "valid_targets_min": 1483 |
| }, |
| { |
| "epoch": 5.576923076923077, |
| "grad_norm": 0.6680147625247186, |
| "learning_rate": 4.840971797432643e-06, |
| "loss": 0.2636, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2638908922672272, |
| "step": 3190, |
| "valid_targets_mean": 3079.1, |
| "valid_targets_min": 769 |
| }, |
| { |
| "epoch": 5.585664335664336, |
| "grad_norm": 0.6033920894084822, |
| "learning_rate": 4.784238563559025e-06, |
| "loss": 0.2754, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3316216468811035, |
| "step": 3195, |
| "valid_targets_mean": 4097.1, |
| "valid_targets_min": 1321 |
| }, |
| { |
| "epoch": 5.594405594405594, |
| "grad_norm": 0.6247334828081322, |
| "learning_rate": 4.727794533643442e-06, |
| "loss": 0.2731, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2612161636352539, |
| "step": 3200, |
| "valid_targets_mean": 3408.8, |
| "valid_targets_min": 805 |
| }, |
| { |
| "epoch": 5.603146853146853, |
| "grad_norm": 0.5988968477062824, |
| "learning_rate": 4.671640780510069e-06, |
| "loss": 0.2707, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2181917130947113, |
| "step": 3205, |
| "valid_targets_mean": 2906.1, |
| "valid_targets_min": 1078 |
| }, |
| { |
| "epoch": 5.611888111888112, |
| "grad_norm": 0.46415277622783324, |
| "learning_rate": 4.615778371465856e-06, |
| "loss": 0.2921, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3689541816711426, |
| "step": 3210, |
| "valid_targets_mean": 6330.0, |
| "valid_targets_min": 402 |
| }, |
| { |
| "epoch": 5.620629370629371, |
| "grad_norm": 0.5613877412055053, |
| "learning_rate": 4.560208368280183e-06, |
| "loss": 0.3161, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3414204716682434, |
| "step": 3215, |
| "valid_targets_mean": 4572.8, |
| "valid_targets_min": 1159 |
| }, |
| { |
| "epoch": 5.629370629370629, |
| "grad_norm": 0.5560487965185203, |
| "learning_rate": 4.5049318271647295e-06, |
| "loss": 0.2626, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2097901999950409, |
| "step": 3220, |
| "valid_targets_mean": 3893.2, |
| "valid_targets_min": 915 |
| }, |
| { |
| "epoch": 5.638111888111888, |
| "grad_norm": 0.7296638177669432, |
| "learning_rate": 4.4499497987533816e-06, |
| "loss": 0.3103, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.229948490858078, |
| "step": 3225, |
| "valid_targets_mean": 2531.2, |
| "valid_targets_min": 807 |
| }, |
| { |
| "epoch": 5.646853146853147, |
| "grad_norm": 0.6166550888524173, |
| "learning_rate": 4.395263328082251e-06, |
| "loss": 0.2568, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.29370614886283875, |
| "step": 3230, |
| "valid_targets_mean": 3795.5, |
| "valid_targets_min": 917 |
| }, |
| { |
| "epoch": 5.655594405594406, |
| "grad_norm": 0.5808842651980372, |
| "learning_rate": 4.340873454569821e-06, |
| "loss": 0.2863, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3381272554397583, |
| "step": 3235, |
| "valid_targets_mean": 4567.4, |
| "valid_targets_min": 1063 |
| }, |
| { |
| "epoch": 5.664335664335664, |
| "grad_norm": 0.548523750131149, |
| "learning_rate": 4.286781211997208e-06, |
| "loss": 0.262, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2726270854473114, |
| "step": 3240, |
| "valid_targets_mean": 4212.8, |
| "valid_targets_min": 1200 |
| }, |
| { |
| "epoch": 5.673076923076923, |
| "grad_norm": 0.6767114674832129, |
| "learning_rate": 4.232987628488483e-06, |
| "loss": 0.269, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20798838138580322, |
| "step": 3245, |
| "valid_targets_mean": 2656.3, |
| "valid_targets_min": 1177 |
| }, |
| { |
| "epoch": 5.681818181818182, |
| "grad_norm": 0.6061394916537475, |
| "learning_rate": 4.179493726491144e-06, |
| "loss": 0.297, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.28469932079315186, |
| "step": 3250, |
| "valid_targets_mean": 3790.7, |
| "valid_targets_min": 1247 |
| }, |
| { |
| "epoch": 5.690559440559441, |
| "grad_norm": 0.5514714164196605, |
| "learning_rate": 4.126300522756692e-06, |
| "loss": 0.2945, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2420085370540619, |
| "step": 3255, |
| "valid_targets_mean": 4023.1, |
| "valid_targets_min": 1003 |
| }, |
| { |
| "epoch": 5.699300699300699, |
| "grad_norm": 0.7165447495050303, |
| "learning_rate": 4.073409028321289e-06, |
| "loss": 0.2789, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2511775493621826, |
| "step": 3260, |
| "valid_targets_mean": 2425.6, |
| "valid_targets_min": 836 |
| }, |
| { |
| "epoch": 5.708041958041958, |
| "grad_norm": 0.5309431458422739, |
| "learning_rate": 4.020820248486544e-06, |
| "loss": 0.2863, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.312569260597229, |
| "step": 3265, |
| "valid_targets_mean": 4707.6, |
| "valid_targets_min": 1005 |
| }, |
| { |
| "epoch": 5.716783216783217, |
| "grad_norm": 0.6205803951689329, |
| "learning_rate": 3.968535182800423e-06, |
| "loss": 0.3001, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.283012330532074, |
| "step": 3270, |
| "valid_targets_mean": 3780.8, |
| "valid_targets_min": 840 |
| }, |
| { |
| "epoch": 5.725524475524476, |
| "grad_norm": 0.5702696584364179, |
| "learning_rate": 3.916554825038221e-06, |
| "loss": 0.2865, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.29808011651039124, |
| "step": 3275, |
| "valid_targets_mean": 3887.8, |
| "valid_targets_min": 1299 |
| }, |
| { |
| "epoch": 5.734265734265734, |
| "grad_norm": 0.6299800822954458, |
| "learning_rate": 3.864880163183709e-06, |
| "loss": 0.2915, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.17876118421554565, |
| "step": 3280, |
| "valid_targets_mean": 2609.9, |
| "valid_targets_min": 1104 |
| }, |
| { |
| "epoch": 5.743006993006993, |
| "grad_norm": 0.5991337075244536, |
| "learning_rate": 3.813512179410317e-06, |
| "loss": 0.2775, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.28599756956100464, |
| "step": 3285, |
| "valid_targets_mean": 3688.8, |
| "valid_targets_min": 1165 |
| }, |
| { |
| "epoch": 5.751748251748252, |
| "grad_norm": 0.5243673827463061, |
| "learning_rate": 3.762451850062501e-06, |
| "loss": 0.2918, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2722562551498413, |
| "step": 3290, |
| "valid_targets_mean": 4477.4, |
| "valid_targets_min": 1031 |
| }, |
| { |
| "epoch": 5.760489510489511, |
| "grad_norm": 0.6658579790305317, |
| "learning_rate": 3.711700145637156e-06, |
| "loss": 0.3113, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33141064643859863, |
| "step": 3295, |
| "valid_targets_mean": 3457.7, |
| "valid_targets_min": 1188 |
| }, |
| { |
| "epoch": 5.769230769230769, |
| "grad_norm": 0.5388723983560816, |
| "learning_rate": 3.6612580307652024e-06, |
| "loss": 0.293, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.309293270111084, |
| "step": 3300, |
| "valid_targets_mean": 5008.1, |
| "valid_targets_min": 350 |
| }, |
| { |
| "epoch": 5.777972027972028, |
| "grad_norm": 0.5300494283794329, |
| "learning_rate": 3.611126464193211e-06, |
| "loss": 0.2842, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3127099871635437, |
| "step": 3305, |
| "valid_targets_mean": 5129.1, |
| "valid_targets_min": 1776 |
| }, |
| { |
| "epoch": 5.786713286713287, |
| "grad_norm": 0.5734939334413017, |
| "learning_rate": 3.561306398765225e-06, |
| "loss": 0.2787, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2457369565963745, |
| "step": 3310, |
| "valid_targets_mean": 3891.1, |
| "valid_targets_min": 1186 |
| }, |
| { |
| "epoch": 5.795454545454545, |
| "grad_norm": 0.5898212051083919, |
| "learning_rate": 3.5117987814046074e-06, |
| "loss": 0.2943, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22512537240982056, |
| "step": 3315, |
| "valid_targets_mean": 3410.9, |
| "valid_targets_min": 1094 |
| }, |
| { |
| "epoch": 5.804195804195804, |
| "grad_norm": 0.6207274470955104, |
| "learning_rate": 3.4626045530960696e-06, |
| "loss": 0.2832, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.30430057644844055, |
| "step": 3320, |
| "valid_targets_mean": 4685.4, |
| "valid_targets_min": 1113 |
| }, |
| { |
| "epoch": 5.812937062937063, |
| "grad_norm": 0.608742829096707, |
| "learning_rate": 3.4137246488677866e-06, |
| "loss": 0.2673, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3098594546318054, |
| "step": 3325, |
| "valid_targets_mean": 3769.6, |
| "valid_targets_min": 1045 |
| }, |
| { |
| "epoch": 5.821678321678322, |
| "grad_norm": 0.5846165269760552, |
| "learning_rate": 3.365159997773606e-06, |
| "loss": 0.2942, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22525198757648468, |
| "step": 3330, |
| "valid_targets_mean": 3777.8, |
| "valid_targets_min": 1666 |
| }, |
| { |
| "epoch": 5.83041958041958, |
| "grad_norm": 0.6111370595067496, |
| "learning_rate": 3.3169115228754014e-06, |
| "loss": 0.309, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36875244975090027, |
| "step": 3335, |
| "valid_targets_mean": 4112.8, |
| "valid_targets_min": 753 |
| }, |
| { |
| "epoch": 5.839160839160839, |
| "grad_norm": 0.5387375942451661, |
| "learning_rate": 3.268980141225537e-06, |
| "loss": 0.3289, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31299394369125366, |
| "step": 3340, |
| "valid_targets_mean": 5053.8, |
| "valid_targets_min": 1449 |
| }, |
| { |
| "epoch": 5.847902097902098, |
| "grad_norm": 0.6240330539437785, |
| "learning_rate": 3.221366763849423e-06, |
| "loss": 0.2531, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26625147461891174, |
| "step": 3345, |
| "valid_targets_mean": 3334.1, |
| "valid_targets_min": 1340 |
| }, |
| { |
| "epoch": 5.856643356643357, |
| "grad_norm": 0.7043735642239611, |
| "learning_rate": 3.1740722957281923e-06, |
| "loss": 0.2938, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3169170618057251, |
| "step": 3350, |
| "valid_targets_mean": 3898.0, |
| "valid_targets_min": 974 |
| }, |
| { |
| "epoch": 5.865384615384615, |
| "grad_norm": 0.4821674121878327, |
| "learning_rate": 3.127097635781533e-06, |
| "loss": 0.2709, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2442520707845688, |
| "step": 3355, |
| "valid_targets_mean": 5463.6, |
| "valid_targets_min": 1323 |
| }, |
| { |
| "epoch": 5.874125874125874, |
| "grad_norm": 0.59032654806126, |
| "learning_rate": 3.080443676850575e-06, |
| "loss": 0.3114, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3039432168006897, |
| "step": 3360, |
| "valid_targets_mean": 4510.3, |
| "valid_targets_min": 1476 |
| }, |
| { |
| "epoch": 5.882867132867133, |
| "grad_norm": 0.5711335653285038, |
| "learning_rate": 3.0341113056809113e-06, |
| "loss": 0.2657, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3196086883544922, |
| "step": 3365, |
| "valid_targets_mean": 4369.4, |
| "valid_targets_min": 1253 |
| }, |
| { |
| "epoch": 5.891608391608392, |
| "grad_norm": 0.6147801856977855, |
| "learning_rate": 2.9881014029057853e-06, |
| "loss": 0.2907, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34057629108428955, |
| "step": 3370, |
| "valid_targets_mean": 4283.4, |
| "valid_targets_min": 1334 |
| }, |
| { |
| "epoch": 5.90034965034965, |
| "grad_norm": 0.5972285440857505, |
| "learning_rate": 2.942414843029302e-06, |
| "loss": 0.2705, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2990839183330536, |
| "step": 3375, |
| "valid_targets_mean": 4576.2, |
| "valid_targets_min": 885 |
| }, |
| { |
| "epoch": 5.909090909090909, |
| "grad_norm": 0.5952021910403875, |
| "learning_rate": 2.89705249440984e-06, |
| "loss": 0.2577, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27879855036735535, |
| "step": 3380, |
| "valid_targets_mean": 3924.9, |
| "valid_targets_min": 758 |
| }, |
| { |
| "epoch": 5.9178321678321675, |
| "grad_norm": 0.6227318532091863, |
| "learning_rate": 2.8520152192435468e-06, |
| "loss": 0.241, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22561070322990417, |
| "step": 3385, |
| "valid_targets_mean": 3241.6, |
| "valid_targets_min": 835 |
| }, |
| { |
| "epoch": 5.926573426573427, |
| "grad_norm": 0.6036831338415529, |
| "learning_rate": 2.807303873547931e-06, |
| "loss": 0.2875, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.269223153591156, |
| "step": 3390, |
| "valid_targets_mean": 3960.1, |
| "valid_targets_min": 1308 |
| }, |
| { |
| "epoch": 5.935314685314685, |
| "grad_norm": 0.6198564976381323, |
| "learning_rate": 2.7629193071455974e-06, |
| "loss": 0.2978, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2761087417602539, |
| "step": 3395, |
| "valid_targets_mean": 3745.8, |
| "valid_targets_min": 1250 |
| }, |
| { |
| "epoch": 5.944055944055944, |
| "grad_norm": 0.5769269392848372, |
| "learning_rate": 2.718862363648118e-06, |
| "loss": 0.2963, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.28240329027175903, |
| "step": 3400, |
| "valid_targets_mean": 4084.9, |
| "valid_targets_min": 838 |
| }, |
| { |
| "epoch": 5.9527972027972025, |
| "grad_norm": 0.5159296065059323, |
| "learning_rate": 2.6751338804399685e-06, |
| "loss": 0.3087, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33804747462272644, |
| "step": 3405, |
| "valid_targets_mean": 6029.0, |
| "valid_targets_min": 1526 |
| }, |
| { |
| "epoch": 5.961538461538462, |
| "grad_norm": 0.5627855303755237, |
| "learning_rate": 2.6317346886626192e-06, |
| "loss": 0.299, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.29597389698028564, |
| "step": 3410, |
| "valid_targets_mean": 4527.5, |
| "valid_targets_min": 1235 |
| }, |
| { |
| "epoch": 5.97027972027972, |
| "grad_norm": 0.5297648101917568, |
| "learning_rate": 2.5886656131987575e-06, |
| "loss": 0.2732, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2557566165924072, |
| "step": 3415, |
| "valid_targets_mean": 4257.1, |
| "valid_targets_min": 851 |
| }, |
| { |
| "epoch": 5.979020979020979, |
| "grad_norm": 0.6191215344301044, |
| "learning_rate": 2.5459274726565752e-06, |
| "loss": 0.2773, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23104539513587952, |
| "step": 3420, |
| "valid_targets_mean": 2965.7, |
| "valid_targets_min": 1289 |
| }, |
| { |
| "epoch": 5.9877622377622375, |
| "grad_norm": 0.7122036934147264, |
| "learning_rate": 2.5035210793542387e-06, |
| "loss": 0.3233, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31105631589889526, |
| "step": 3425, |
| "valid_targets_mean": 4595.3, |
| "valid_targets_min": 923 |
| }, |
| { |
| "epoch": 5.996503496503497, |
| "grad_norm": 0.6248335921345303, |
| "learning_rate": 2.461447239304444e-06, |
| "loss": 0.2838, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31339454650878906, |
| "step": 3430, |
| "valid_targets_mean": 3528.1, |
| "valid_targets_min": 839 |
| }, |
| { |
| "epoch": 6.005244755244755, |
| "grad_norm": 0.6546736256801129, |
| "learning_rate": 2.41970675219908e-06, |
| "loss": 0.2626, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.1906014084815979, |
| "step": 3435, |
| "valid_targets_mean": 2545.0, |
| "valid_targets_min": 915 |
| }, |
| { |
| "epoch": 6.013986013986014, |
| "grad_norm": 0.546331877599847, |
| "learning_rate": 2.378300411394041e-06, |
| "loss": 0.2693, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21837225556373596, |
| "step": 3440, |
| "valid_targets_mean": 3521.7, |
| "valid_targets_min": 844 |
| }, |
| { |
| "epoch": 6.0227272727272725, |
| "grad_norm": 0.5676929839026588, |
| "learning_rate": 2.337229003894157e-06, |
| "loss": 0.3256, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36682766675949097, |
| "step": 3445, |
| "valid_targets_mean": 4742.9, |
| "valid_targets_min": 1237 |
| }, |
| { |
| "epoch": 6.031468531468532, |
| "grad_norm": 0.5669674632220835, |
| "learning_rate": 2.2964933103382148e-06, |
| "loss": 0.2658, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.28727206587791443, |
| "step": 3450, |
| "valid_targets_mean": 4004.6, |
| "valid_targets_min": 1310 |
| }, |
| { |
| "epoch": 6.04020979020979, |
| "grad_norm": 0.6309898614135956, |
| "learning_rate": 2.2560941049841346e-06, |
| "loss": 0.2761, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20886394381523132, |
| "step": 3455, |
| "valid_targets_mean": 2680.5, |
| "valid_targets_min": 1059 |
| }, |
| { |
| "epoch": 6.048951048951049, |
| "grad_norm": 0.60928327793459, |
| "learning_rate": 2.2160321556942544e-06, |
| "loss": 0.2691, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.30445602536201477, |
| "step": 3460, |
| "valid_targets_mean": 4046.9, |
| "valid_targets_min": 915 |
| }, |
| { |
| "epoch": 6.0576923076923075, |
| "grad_norm": 0.7833054139185534, |
| "learning_rate": 2.1763082239207266e-06, |
| "loss": 0.253, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21612828969955444, |
| "step": 3465, |
| "valid_targets_mean": 2976.7, |
| "valid_targets_min": 1566 |
| }, |
| { |
| "epoch": 6.066433566433567, |
| "grad_norm": 0.6497425853670982, |
| "learning_rate": 2.136923064691041e-06, |
| "loss": 0.2543, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.15709246695041656, |
| "step": 3470, |
| "valid_targets_mean": 2333.2, |
| "valid_targets_min": 858 |
| }, |
| { |
| "epoch": 6.075174825174825, |
| "grad_norm": 0.6725133904006579, |
| "learning_rate": 2.097877426593704e-06, |
| "loss": 0.2637, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20932920277118683, |
| "step": 3475, |
| "valid_targets_mean": 2497.1, |
| "valid_targets_min": 815 |
| }, |
| { |
| "epoch": 6.083916083916084, |
| "grad_norm": 0.6290252855540327, |
| "learning_rate": 2.0591720517639713e-06, |
| "loss": 0.3105, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2524038553237915, |
| "step": 3480, |
| "valid_targets_mean": 3004.4, |
| "valid_targets_min": 823 |
| }, |
| { |
| "epoch": 6.0926573426573425, |
| "grad_norm": 0.6862544212215417, |
| "learning_rate": 2.020807675869758e-06, |
| "loss": 0.2694, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2957307696342468, |
| "step": 3485, |
| "valid_targets_mean": 4532.2, |
| "valid_targets_min": 361 |
| }, |
| { |
| "epoch": 6.101398601398602, |
| "grad_norm": 0.6485079485932239, |
| "learning_rate": 1.9827850280976733e-06, |
| "loss": 0.2767, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32891833782196045, |
| "step": 3490, |
| "valid_targets_mean": 4233.2, |
| "valid_targets_min": 278 |
| }, |
| { |
| "epoch": 6.11013986013986, |
| "grad_norm": 0.6921512870006977, |
| "learning_rate": 1.9451048311391375e-06, |
| "loss": 0.2905, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26255372166633606, |
| "step": 3495, |
| "valid_targets_mean": 3512.6, |
| "valid_targets_min": 894 |
| }, |
| { |
| "epoch": 6.118881118881119, |
| "grad_norm": 0.6208115294079365, |
| "learning_rate": 1.9077678011766455e-06, |
| "loss": 0.3224, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.29941844940185547, |
| "step": 3500, |
| "valid_targets_mean": 3561.2, |
| "valid_targets_min": 1191 |
| }, |
| { |
| "epoch": 6.1276223776223775, |
| "grad_norm": 0.5916073069732769, |
| "learning_rate": 1.8707746478701794e-06, |
| "loss": 0.3104, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.296660840511322, |
| "step": 3505, |
| "valid_targets_mean": 3865.4, |
| "valid_targets_min": 1120 |
| }, |
| { |
| "epoch": 6.136363636363637, |
| "grad_norm": 0.5576227926843184, |
| "learning_rate": 1.8341260743436873e-06, |
| "loss": 0.2724, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2193690687417984, |
| "step": 3510, |
| "valid_targets_mean": 3824.8, |
| "valid_targets_min": 865 |
| }, |
| { |
| "epoch": 6.145104895104895, |
| "grad_norm": 0.5853811367031623, |
| "learning_rate": 1.7978227771717405e-06, |
| "loss": 0.2832, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3498114347457886, |
| "step": 3515, |
| "valid_targets_mean": 4405.4, |
| "valid_targets_min": 1504 |
| }, |
| { |
| "epoch": 6.153846153846154, |
| "grad_norm": 0.6053214238879862, |
| "learning_rate": 1.7618654463662954e-06, |
| "loss": 0.2799, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.28211668133735657, |
| "step": 3520, |
| "valid_targets_mean": 3949.8, |
| "valid_targets_min": 1235 |
| }, |
| { |
| "epoch": 6.1625874125874125, |
| "grad_norm": 0.566264654468671, |
| "learning_rate": 1.7262547653635641e-06, |
| "loss": 0.2842, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2645253539085388, |
| "step": 3525, |
| "valid_targets_mean": 3999.5, |
| "valid_targets_min": 968 |
| }, |
| { |
| "epoch": 6.171328671328672, |
| "grad_norm": 0.6113838491407106, |
| "learning_rate": 1.6909914110110293e-06, |
| "loss": 0.2567, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24374493956565857, |
| "step": 3530, |
| "valid_targets_mean": 3655.2, |
| "valid_targets_min": 1315 |
| }, |
| { |
| "epoch": 6.18006993006993, |
| "grad_norm": 0.5494925652722575, |
| "learning_rate": 1.656076053554594e-06, |
| "loss": 0.2606, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2420923113822937, |
| "step": 3535, |
| "valid_targets_mean": 4067.6, |
| "valid_targets_min": 801 |
| }, |
| { |
| "epoch": 6.188811188811189, |
| "grad_norm": 0.6071906510133728, |
| "learning_rate": 1.621509356625819e-06, |
| "loss": 0.2674, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26888033747673035, |
| "step": 3540, |
| "valid_targets_mean": 4184.9, |
| "valid_targets_min": 1107 |
| }, |
| { |
| "epoch": 6.1975524475524475, |
| "grad_norm": 0.6406113556865684, |
| "learning_rate": 1.5872919772293215e-06, |
| "loss": 0.2574, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24279576539993286, |
| "step": 3545, |
| "valid_targets_mean": 3324.6, |
| "valid_targets_min": 1078 |
| }, |
| { |
| "epoch": 6.206293706293707, |
| "grad_norm": 0.6003844929874508, |
| "learning_rate": 1.5534245657302994e-06, |
| "loss": 0.2954, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.28774210810661316, |
| "step": 3550, |
| "valid_targets_mean": 4073.2, |
| "valid_targets_min": 1522 |
| }, |
| { |
| "epoch": 6.215034965034965, |
| "grad_norm": 0.6802616494670375, |
| "learning_rate": 1.5199077658421368e-06, |
| "loss": 0.2862, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24521152675151825, |
| "step": 3555, |
| "valid_targets_mean": 2741.4, |
| "valid_targets_min": 897 |
| }, |
| { |
| "epoch": 6.223776223776224, |
| "grad_norm": 0.5311261915809192, |
| "learning_rate": 1.486742214614205e-06, |
| "loss": 0.296, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.29008519649505615, |
| "step": 3560, |
| "valid_targets_mean": 4434.4, |
| "valid_targets_min": 1423 |
| }, |
| { |
| "epoch": 6.2325174825174825, |
| "grad_norm": 0.574513939046325, |
| "learning_rate": 1.4539285424197401e-06, |
| "loss": 0.2807, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32967033982276917, |
| "step": 3565, |
| "valid_targets_mean": 4577.4, |
| "valid_targets_min": 1109 |
| }, |
| { |
| "epoch": 6.241258741258742, |
| "grad_norm": 0.49392742782010396, |
| "learning_rate": 1.42146737294385e-06, |
| "loss": 0.2759, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3063806891441345, |
| "step": 3570, |
| "valid_targets_mean": 5692.9, |
| "valid_targets_min": 1531 |
| }, |
| { |
| "epoch": 6.25, |
| "grad_norm": 0.5578568105254652, |
| "learning_rate": 1.3893593231716685e-06, |
| "loss": 0.2481, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2709795832633972, |
| "step": 3575, |
| "valid_targets_mean": 4448.0, |
| "valid_targets_min": 1255 |
| }, |
| { |
| "epoch": 6.258741258741258, |
| "grad_norm": 0.6101064186593799, |
| "learning_rate": 1.3576050033766431e-06, |
| "loss": 0.2661, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3113132119178772, |
| "step": 3580, |
| "valid_targets_mean": 3706.6, |
| "valid_targets_min": 1213 |
| }, |
| { |
| "epoch": 6.2674825174825175, |
| "grad_norm": 0.6295757030791121, |
| "learning_rate": 1.3262050171089125e-06, |
| "loss": 0.2887, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2644977271556854, |
| "step": 3585, |
| "valid_targets_mean": 3470.4, |
| "valid_targets_min": 315 |
| }, |
| { |
| "epoch": 6.276223776223776, |
| "grad_norm": 0.5663034758396543, |
| "learning_rate": 1.2951599611838494e-06, |
| "loss": 0.2756, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26824554800987244, |
| "step": 3590, |
| "valid_targets_mean": 3999.2, |
| "valid_targets_min": 838 |
| }, |
| { |
| "epoch": 6.284965034965035, |
| "grad_norm": 0.6036200941709391, |
| "learning_rate": 1.264470425670712e-06, |
| "loss": 0.2949, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3287750482559204, |
| "step": 3595, |
| "valid_targets_mean": 5751.5, |
| "valid_targets_min": 1197 |
| }, |
| { |
| "epoch": 6.293706293706293, |
| "grad_norm": 0.5810804752661107, |
| "learning_rate": 1.234136993881423e-06, |
| "loss": 0.2526, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2942957282066345, |
| "step": 3600, |
| "valid_targets_mean": 4172.7, |
| "valid_targets_min": 912 |
| }, |
| { |
| "epoch": 6.3024475524475525, |
| "grad_norm": 0.657214863100275, |
| "learning_rate": 1.2041602423595e-06, |
| "loss": 0.2634, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.18747061491012573, |
| "step": 3605, |
| "valid_targets_mean": 2734.8, |
| "valid_targets_min": 985 |
| }, |
| { |
| "epoch": 6.311188811188811, |
| "grad_norm": 0.6501996468721265, |
| "learning_rate": 1.174540740869079e-06, |
| "loss": 0.2772, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32021254301071167, |
| "step": 3610, |
| "valid_targets_mean": 3323.9, |
| "valid_targets_min": 749 |
| }, |
| { |
| "epoch": 6.31993006993007, |
| "grad_norm": 0.5803676942137402, |
| "learning_rate": 1.1452790523840895e-06, |
| "loss": 0.2754, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.281261146068573, |
| "step": 3615, |
| "valid_targets_mean": 4163.1, |
| "valid_targets_min": 771 |
| }, |
| { |
| "epoch": 6.328671328671328, |
| "grad_norm": 0.7055138161863601, |
| "learning_rate": 1.1163757330775683e-06, |
| "loss": 0.3053, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.18472999334335327, |
| "step": 3620, |
| "valid_targets_mean": 2490.1, |
| "valid_targets_min": 1264 |
| }, |
| { |
| "epoch": 6.3374125874125875, |
| "grad_norm": 0.6313383239462723, |
| "learning_rate": 1.0878313323110624e-06, |
| "loss": 0.3055, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2720009684562683, |
| "step": 3625, |
| "valid_targets_mean": 3518.5, |
| "valid_targets_min": 828 |
| }, |
| { |
| "epoch": 6.346153846153846, |
| "grad_norm": 0.6561844097817968, |
| "learning_rate": 1.0596463926242117e-06, |
| "loss": 0.2733, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23926571011543274, |
| "step": 3630, |
| "valid_targets_mean": 2988.6, |
| "valid_targets_min": 897 |
| }, |
| { |
| "epoch": 6.354895104895105, |
| "grad_norm": 0.5665310024576693, |
| "learning_rate": 1.0318214497244194e-06, |
| "loss": 0.3238, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3651200830936432, |
| "step": 3635, |
| "valid_targets_mean": 5215.1, |
| "valid_targets_min": 1625 |
| }, |
| { |
| "epoch": 6.363636363636363, |
| "grad_norm": 0.6554284589319097, |
| "learning_rate": 1.0043570324766905e-06, |
| "loss": 0.2894, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24973063170909882, |
| "step": 3640, |
| "valid_targets_mean": 2915.8, |
| "valid_targets_min": 777 |
| }, |
| { |
| "epoch": 6.3723776223776225, |
| "grad_norm": 0.7008129809196576, |
| "learning_rate": 9.772536628935492e-07, |
| "loss": 0.283, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24499808251857758, |
| "step": 3645, |
| "valid_targets_mean": 2670.0, |
| "valid_targets_min": 1221 |
| }, |
| { |
| "epoch": 6.381118881118881, |
| "grad_norm": 0.7024012851039138, |
| "learning_rate": 9.505118561251514e-07, |
| "loss": 0.2798, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2695966362953186, |
| "step": 3650, |
| "valid_targets_mean": 2731.5, |
| "valid_targets_min": 866 |
| }, |
| { |
| "epoch": 6.38986013986014, |
| "grad_norm": 0.5845704369992565, |
| "learning_rate": 9.241321204494658e-07, |
| "loss": 0.2456, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25256311893463135, |
| "step": 3655, |
| "valid_targets_mean": 4473.8, |
| "valid_targets_min": 1264 |
| }, |
| { |
| "epoch": 6.398601398601398, |
| "grad_norm": 0.5813539539242096, |
| "learning_rate": 8.981149572626214e-07, |
| "loss": 0.2553, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.152411550283432, |
| "step": 3660, |
| "valid_targets_mean": 2954.5, |
| "valid_targets_min": 840 |
| }, |
| { |
| "epoch": 6.4073426573426575, |
| "grad_norm": 0.8400642322428409, |
| "learning_rate": 8.724608610693952e-07, |
| "loss": 0.2364, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2882179915904999, |
| "step": 3665, |
| "valid_targets_mean": 3022.2, |
| "valid_targets_min": 1200 |
| }, |
| { |
| "epoch": 6.416083916083916, |
| "grad_norm": 0.7360779762095463, |
| "learning_rate": 8.471703194737846e-07, |
| "loss": 0.2653, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2204851359128952, |
| "step": 3670, |
| "valid_targets_mean": 2568.9, |
| "valid_targets_min": 1137 |
| }, |
| { |
| "epoch": 6.424825174825175, |
| "grad_norm": 0.5801639551802711, |
| "learning_rate": 8.222438131697541e-07, |
| "loss": 0.2705, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25313958525657654, |
| "step": 3675, |
| "valid_targets_mean": 3821.8, |
| "valid_targets_min": 938 |
| }, |
| { |
| "epoch": 6.433566433566433, |
| "grad_norm": 0.6509857982198833, |
| "learning_rate": 7.976818159321031e-07, |
| "loss": 0.2622, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24882566928863525, |
| "step": 3680, |
| "valid_targets_mean": 3058.3, |
| "valid_targets_min": 867 |
| }, |
| { |
| "epoch": 6.4423076923076925, |
| "grad_norm": 0.5859694732821341, |
| "learning_rate": 7.734847946074531e-07, |
| "loss": 0.2945, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23909510672092438, |
| "step": 3685, |
| "valid_targets_mean": 3752.7, |
| "valid_targets_min": 1217 |
| }, |
| { |
| "epoch": 6.451048951048951, |
| "grad_norm": 0.4932257801456551, |
| "learning_rate": 7.496532091053765e-07, |
| "loss": 0.2555, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3305898904800415, |
| "step": 3690, |
| "valid_targets_mean": 5985.6, |
| "valid_targets_min": 1330 |
| }, |
| { |
| "epoch": 6.45979020979021, |
| "grad_norm": 0.6428645373849343, |
| "learning_rate": 7.261875123896533e-07, |
| "loss": 0.2696, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33091801404953003, |
| "step": 3695, |
| "valid_targets_mean": 4158.1, |
| "valid_targets_min": 836 |
| }, |
| { |
| "epoch": 6.468531468531468, |
| "grad_norm": 0.6419446507900263, |
| "learning_rate": 7.030881504696708e-07, |
| "loss": 0.2879, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.28880882263183594, |
| "step": 3700, |
| "valid_targets_mean": 4430.9, |
| "valid_targets_min": 1343 |
| }, |
| { |
| "epoch": 6.4772727272727275, |
| "grad_norm": 0.6349223749579598, |
| "learning_rate": 6.803555623919322e-07, |
| "loss": 0.2867, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22542111575603485, |
| "step": 3705, |
| "valid_targets_mean": 3429.0, |
| "valid_targets_min": 1425 |
| }, |
| { |
| "epoch": 6.486013986013986, |
| "grad_norm": 0.5586249125018691, |
| "learning_rate": 6.579901802317223e-07, |
| "loss": 0.326, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3446670174598694, |
| "step": 3710, |
| "valid_targets_mean": 5845.7, |
| "valid_targets_min": 1794 |
| }, |
| { |
| "epoch": 6.494755244755245, |
| "grad_norm": 0.5654988182171874, |
| "learning_rate": 6.359924290848929e-07, |
| "loss": 0.309, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.37433212995529175, |
| "step": 3715, |
| "valid_targets_mean": 4621.6, |
| "valid_targets_min": 1213 |
| }, |
| { |
| "epoch": 6.503496503496503, |
| "grad_norm": 0.5671227364089624, |
| "learning_rate": 6.143627270597785e-07, |
| "loss": 0.2255, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.17292237281799316, |
| "step": 3720, |
| "valid_targets_mean": 3233.8, |
| "valid_targets_min": 1087 |
| }, |
| { |
| "epoch": 6.5122377622377625, |
| "grad_norm": 0.6425112452389675, |
| "learning_rate": 5.931014852692541e-07, |
| "loss": 0.2728, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26524510979652405, |
| "step": 3725, |
| "valid_targets_mean": 3356.4, |
| "valid_targets_min": 882 |
| }, |
| { |
| "epoch": 6.520979020979021, |
| "grad_norm": 0.5756543070733119, |
| "learning_rate": 5.72209107822923e-07, |
| "loss": 0.274, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.29785022139549255, |
| "step": 3730, |
| "valid_targets_mean": 4071.3, |
| "valid_targets_min": 1156 |
| }, |
| { |
| "epoch": 6.52972027972028, |
| "grad_norm": 0.6271284223823304, |
| "learning_rate": 5.516859918194262e-07, |
| "loss": 0.307, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.29389727115631104, |
| "step": 3735, |
| "valid_targets_mean": 3572.2, |
| "valid_targets_min": 968 |
| }, |
| { |
| "epoch": 6.538461538461538, |
| "grad_norm": 0.5863542188132574, |
| "learning_rate": 5.315325273389094e-07, |
| "loss": 0.2827, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26336008310317993, |
| "step": 3740, |
| "valid_targets_mean": 4096.9, |
| "valid_targets_min": 347 |
| }, |
| { |
| "epoch": 6.5472027972027975, |
| "grad_norm": 0.6259632998483117, |
| "learning_rate": 5.117490974355943e-07, |
| "loss": 0.3089, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.30066174268722534, |
| "step": 3745, |
| "valid_targets_mean": 3837.0, |
| "valid_targets_min": 1444 |
| }, |
| { |
| "epoch": 6.555944055944056, |
| "grad_norm": 0.5762913607782765, |
| "learning_rate": 4.923360781305087e-07, |
| "loss": 0.2835, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.30665403604507446, |
| "step": 3750, |
| "valid_targets_mean": 4425.5, |
| "valid_targets_min": 400 |
| }, |
| { |
| "epoch": 6.564685314685315, |
| "grad_norm": 0.604842563995092, |
| "learning_rate": 4.7329383840433395e-07, |
| "loss": 0.2822, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24494880437850952, |
| "step": 3755, |
| "valid_targets_mean": 3657.3, |
| "valid_targets_min": 1207 |
| }, |
| { |
| "epoch": 6.573426573426573, |
| "grad_norm": 0.6627025272241348, |
| "learning_rate": 4.546227401903913e-07, |
| "loss": 0.3036, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3197290599346161, |
| "step": 3760, |
| "valid_targets_mean": 3325.8, |
| "valid_targets_min": 779 |
| }, |
| { |
| "epoch": 6.5821678321678325, |
| "grad_norm": 0.539869528159097, |
| "learning_rate": 4.363231383677691e-07, |
| "loss": 0.2681, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2869723439216614, |
| "step": 3765, |
| "valid_targets_mean": 5366.8, |
| "valid_targets_min": 1013 |
| }, |
| { |
| "epoch": 6.590909090909091, |
| "grad_norm": 0.5724917365681328, |
| "learning_rate": 4.1839538075456597e-07, |
| "loss": 0.2615, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2898808717727661, |
| "step": 3770, |
| "valid_targets_mean": 4081.4, |
| "valid_targets_min": 1344 |
| }, |
| { |
| "epoch": 6.59965034965035, |
| "grad_norm": 0.6592760685566884, |
| "learning_rate": 4.0083980810129876e-07, |
| "loss": 0.3064, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2967371940612793, |
| "step": 3775, |
| "valid_targets_mean": 3181.4, |
| "valid_targets_min": 872 |
| }, |
| { |
| "epoch": 6.608391608391608, |
| "grad_norm": 0.606456437191496, |
| "learning_rate": 3.8365675408439827e-07, |
| "loss": 0.2328, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24623604118824005, |
| "step": 3780, |
| "valid_targets_mean": 3611.8, |
| "valid_targets_min": 1044 |
| }, |
| { |
| "epoch": 6.617132867132867, |
| "grad_norm": 0.5829659808615845, |
| "learning_rate": 3.6684654529989704e-07, |
| "loss": 0.2572, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2534424066543579, |
| "step": 3785, |
| "valid_targets_mean": 3794.0, |
| "valid_targets_min": 665 |
| }, |
| { |
| "epoch": 6.625874125874126, |
| "grad_norm": 0.586621741689212, |
| "learning_rate": 3.504095012572006e-07, |
| "loss": 0.2773, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2955080270767212, |
| "step": 3790, |
| "valid_targets_mean": 4248.8, |
| "valid_targets_min": 1314 |
| }, |
| { |
| "epoch": 6.634615384615385, |
| "grad_norm": 0.6326737534702042, |
| "learning_rate": 3.3434593437302155e-07, |
| "loss": 0.315, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3392415940761566, |
| "step": 3795, |
| "valid_targets_mean": 3692.1, |
| "valid_targets_min": 1609 |
| }, |
| { |
| "epoch": 6.643356643356643, |
| "grad_norm": 0.561790153450764, |
| "learning_rate": 3.1865614996544834e-07, |
| "loss": 0.2749, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.322073757648468, |
| "step": 3800, |
| "valid_targets_mean": 4551.2, |
| "valid_targets_min": 785 |
| }, |
| { |
| "epoch": 6.652097902097902, |
| "grad_norm": 0.5628076005463111, |
| "learning_rate": 3.0334044624812596e-07, |
| "loss": 0.2849, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2843906581401825, |
| "step": 3805, |
| "valid_targets_mean": 4424.9, |
| "valid_targets_min": 1046 |
| }, |
| { |
| "epoch": 6.660839160839161, |
| "grad_norm": 0.5160597220001579, |
| "learning_rate": 2.883991143246023e-07, |
| "loss": 0.251, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2697199583053589, |
| "step": 3810, |
| "valid_targets_mean": 4928.8, |
| "valid_targets_min": 1110 |
| }, |
| { |
| "epoch": 6.66958041958042, |
| "grad_norm": 0.5931743945808303, |
| "learning_rate": 2.73832438182795e-07, |
| "loss": 0.2617, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2316768765449524, |
| "step": 3815, |
| "valid_targets_mean": 3766.6, |
| "valid_targets_min": 1408 |
| }, |
| { |
| "epoch": 6.678321678321678, |
| "grad_norm": 0.5442096900994562, |
| "learning_rate": 2.59640694689578e-07, |
| "loss": 0.3055, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2675340175628662, |
| "step": 3820, |
| "valid_targets_mean": 4535.4, |
| "valid_targets_min": 994 |
| }, |
| { |
| "epoch": 6.687062937062937, |
| "grad_norm": 1.8394774855479727, |
| "learning_rate": 2.458241535855388e-07, |
| "loss": 0.269, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26005691289901733, |
| "step": 3825, |
| "valid_targets_mean": 3036.9, |
| "valid_targets_min": 1137 |
| }, |
| { |
| "epoch": 6.695804195804196, |
| "grad_norm": 0.6113943146191173, |
| "learning_rate": 2.32383077479843e-07, |
| "loss": 0.3076, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33162689208984375, |
| "step": 3830, |
| "valid_targets_mean": 4674.4, |
| "valid_targets_min": 906 |
| }, |
| { |
| "epoch": 6.704545454545455, |
| "grad_norm": 0.6602273502952561, |
| "learning_rate": 2.1931772184523358e-07, |
| "loss": 0.2879, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.309783935546875, |
| "step": 3835, |
| "valid_targets_mean": 3429.7, |
| "valid_targets_min": 1079 |
| }, |
| { |
| "epoch": 6.713286713286713, |
| "grad_norm": 0.5736876331423237, |
| "learning_rate": 2.0662833501319702e-07, |
| "loss": 0.2892, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32325828075408936, |
| "step": 3840, |
| "valid_targets_mean": 4180.1, |
| "valid_targets_min": 1209 |
| }, |
| { |
| "epoch": 6.722027972027972, |
| "grad_norm": 0.6126276173990889, |
| "learning_rate": 1.9431515816922042e-07, |
| "loss": 0.2824, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2594378888607025, |
| "step": 3845, |
| "valid_targets_mean": 3517.1, |
| "valid_targets_min": 1114 |
| }, |
| { |
| "epoch": 6.730769230769231, |
| "grad_norm": 0.58394492206243, |
| "learning_rate": 1.8237842534822636e-07, |
| "loss": 0.3023, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2913021743297577, |
| "step": 3850, |
| "valid_targets_mean": 4055.2, |
| "valid_targets_min": 1327 |
| }, |
| { |
| "epoch": 6.739510489510489, |
| "grad_norm": 0.7192431832478173, |
| "learning_rate": 1.7081836343010748e-07, |
| "loss": 0.2643, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22975113987922668, |
| "step": 3855, |
| "valid_targets_mean": 2542.4, |
| "valid_targets_min": 1030 |
| }, |
| { |
| "epoch": 6.748251748251748, |
| "grad_norm": 0.5767349334206501, |
| "learning_rate": 1.5963519213542555e-07, |
| "loss": 0.302, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.30083373188972473, |
| "step": 3860, |
| "valid_targets_mean": 4077.9, |
| "valid_targets_min": 1071 |
| }, |
| { |
| "epoch": 6.756993006993007, |
| "grad_norm": 0.6081230019210105, |
| "learning_rate": 1.4882912402123696e-07, |
| "loss": 0.2919, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3036191463470459, |
| "step": 3865, |
| "valid_targets_mean": 3846.3, |
| "valid_targets_min": 1381 |
| }, |
| { |
| "epoch": 6.765734265734266, |
| "grad_norm": 0.5891570706997271, |
| "learning_rate": 1.3840036447704042e-07, |
| "loss": 0.2691, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20486155152320862, |
| "step": 3870, |
| "valid_targets_mean": 3797.2, |
| "valid_targets_min": 363 |
| }, |
| { |
| "epoch": 6.774475524475524, |
| "grad_norm": 0.6322766168986574, |
| "learning_rate": 1.2834911172088683e-07, |
| "loss": 0.2843, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31095296144485474, |
| "step": 3875, |
| "valid_targets_mean": 3419.8, |
| "valid_targets_min": 788 |
| }, |
| { |
| "epoch": 6.783216783216783, |
| "grad_norm": 0.6409732017990925, |
| "learning_rate": 1.1867555679560217e-07, |
| "loss": 0.2421, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2345668077468872, |
| "step": 3880, |
| "valid_targets_mean": 3109.5, |
| "valid_targets_min": 792 |
| }, |
| { |
| "epoch": 6.791958041958042, |
| "grad_norm": 0.6299652829670256, |
| "learning_rate": 1.093798835651616e-07, |
| "loss": 0.3018, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20473377406597137, |
| "step": 3885, |
| "valid_targets_mean": 3211.1, |
| "valid_targets_min": 1014 |
| }, |
| { |
| "epoch": 6.800699300699301, |
| "grad_norm": 0.6490254006725523, |
| "learning_rate": 1.0046226871119002e-07, |
| "loss": 0.2729, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.28175103664398193, |
| "step": 3890, |
| "valid_targets_mean": 3185.8, |
| "valid_targets_min": 915 |
| }, |
| { |
| "epoch": 6.809440559440559, |
| "grad_norm": 0.623847951653859, |
| "learning_rate": 9.192288172961139e-08, |
| "loss": 0.2862, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24998612701892853, |
| "step": 3895, |
| "valid_targets_mean": 3280.8, |
| "valid_targets_min": 1436 |
| }, |
| { |
| "epoch": 6.818181818181818, |
| "grad_norm": 0.5647723189271752, |
| "learning_rate": 8.376188492742021e-08, |
| "loss": 0.2383, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.28807640075683594, |
| "step": 3900, |
| "valid_targets_mean": 4534.2, |
| "valid_targets_min": 875 |
| }, |
| { |
| "epoch": 6.826923076923077, |
| "grad_norm": 0.5941267825959952, |
| "learning_rate": 7.597943341959957e-08, |
| "loss": 0.2683, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2007588893175125, |
| "step": 3905, |
| "valid_targets_mean": 3297.1, |
| "valid_targets_min": 1462 |
| }, |
| { |
| "epoch": 6.835664335664336, |
| "grad_norm": 0.6353613704744627, |
| "learning_rate": 6.857567512617236e-08, |
| "loss": 0.2994, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26361724734306335, |
| "step": 3910, |
| "valid_targets_mean": 3440.0, |
| "valid_targets_min": 1558 |
| }, |
| { |
| "epoch": 6.844405594405594, |
| "grad_norm": 0.5656132531908707, |
| "learning_rate": 6.155075076939244e-08, |
| "loss": 0.2823, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.30361640453338623, |
| "step": 3915, |
| "valid_targets_mean": 4592.8, |
| "valid_targets_min": 1209 |
| }, |
| { |
| "epoch": 6.853146853146853, |
| "grad_norm": 0.6310367427887863, |
| "learning_rate": 5.490479387106007e-08, |
| "loss": 0.2793, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2142931967973709, |
| "step": 3920, |
| "valid_targets_mean": 3075.1, |
| "valid_targets_min": 304 |
| }, |
| { |
| "epoch": 6.861888111888112, |
| "grad_norm": 0.6165604607896683, |
| "learning_rate": 4.863793074999512e-08, |
| "loss": 0.2996, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24841853976249695, |
| "step": 3925, |
| "valid_targets_mean": 3701.6, |
| "valid_targets_min": 819 |
| }, |
| { |
| "epoch": 6.870629370629371, |
| "grad_norm": 0.5061668762750048, |
| "learning_rate": 4.275028051963892e-08, |
| "loss": 0.2888, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24676506221294403, |
| "step": 3930, |
| "valid_targets_mean": 4298.7, |
| "valid_targets_min": 754 |
| }, |
| { |
| "epoch": 6.879370629370629, |
| "grad_norm": 0.6960871564818618, |
| "learning_rate": 3.7241955085773884e-08, |
| "loss": 0.2763, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22061927616596222, |
| "step": 3935, |
| "valid_targets_mean": 2478.1, |
| "valid_targets_min": 1043 |
| }, |
| { |
| "epoch": 6.888111888111888, |
| "grad_norm": 0.6033700639408716, |
| "learning_rate": 3.211305914441187e-08, |
| "loss": 0.2829, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2619025707244873, |
| "step": 3940, |
| "valid_targets_mean": 3731.0, |
| "valid_targets_min": 777 |
| }, |
| { |
| "epoch": 6.896853146853147, |
| "grad_norm": 0.5730719084236546, |
| "learning_rate": 2.736369017979579e-08, |
| "loss": 0.2622, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25423169136047363, |
| "step": 3945, |
| "valid_targets_mean": 3761.8, |
| "valid_targets_min": 1318 |
| }, |
| { |
| "epoch": 6.905594405594406, |
| "grad_norm": 0.5714109307509674, |
| "learning_rate": 2.2993938462554377e-08, |
| "loss": 0.2683, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31169724464416504, |
| "step": 3950, |
| "valid_targets_mean": 4320.3, |
| "valid_targets_min": 1288 |
| }, |
| { |
| "epoch": 6.914335664335664, |
| "grad_norm": 0.7039855872325952, |
| "learning_rate": 1.9003887047972515e-08, |
| "loss": 0.2542, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2564855217933655, |
| "step": 3955, |
| "valid_targets_mean": 2501.0, |
| "valid_targets_min": 258 |
| }, |
| { |
| "epoch": 6.923076923076923, |
| "grad_norm": 0.5977123819404943, |
| "learning_rate": 1.5393611774434657e-08, |
| "loss": 0.2882, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.28014615178108215, |
| "step": 3960, |
| "valid_targets_mean": 3639.7, |
| "valid_targets_min": 1208 |
| }, |
| { |
| "epoch": 6.931818181818182, |
| "grad_norm": 0.8262715036574306, |
| "learning_rate": 1.2163181261959368e-08, |
| "loss": 0.2998, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3190450668334961, |
| "step": 3965, |
| "valid_targets_mean": 4795.8, |
| "valid_targets_min": 1200 |
| }, |
| { |
| "epoch": 6.940559440559441, |
| "grad_norm": 0.6693168848678276, |
| "learning_rate": 9.312656910911432e-09, |
| "loss": 0.2707, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.28227052092552185, |
| "step": 3970, |
| "valid_targets_mean": 3110.6, |
| "valid_targets_min": 860 |
| }, |
| { |
| "epoch": 6.949300699300699, |
| "grad_norm": 0.7086238255439974, |
| "learning_rate": 6.8420929008317094e-09, |
| "loss": 0.2721, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23945844173431396, |
| "step": 3975, |
| "valid_targets_mean": 2615.2, |
| "valid_targets_min": 1022 |
| }, |
| { |
| "epoch": 6.958041958041958, |
| "grad_norm": 0.6720432782965369, |
| "learning_rate": 4.751536189397943e-09, |
| "loss": 0.2495, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.1829328089952469, |
| "step": 3980, |
| "valid_targets_mean": 2545.9, |
| "valid_targets_min": 1284 |
| }, |
| { |
| "epoch": 6.966783216783217, |
| "grad_norm": 0.6533371659227324, |
| "learning_rate": 3.041026511545475e-09, |
| "loss": 0.2582, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.17008773982524872, |
| "step": 3985, |
| "valid_targets_mean": 2636.8, |
| "valid_targets_min": 1230 |
| }, |
| { |
| "epoch": 6.975524475524476, |
| "grad_norm": 0.5251441298018588, |
| "learning_rate": 1.7105963787078517e-09, |
| "loss": 0.3031, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33975541591644287, |
| "step": 3990, |
| "valid_targets_mean": 5142.2, |
| "valid_targets_min": 1149 |
| }, |
| { |
| "epoch": 6.984265734265734, |
| "grad_norm": 0.6173950688495906, |
| "learning_rate": 7.602710781884348e-10, |
| "loss": 0.2756, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27754876017570496, |
| "step": 3995, |
| "valid_targets_mean": 3394.9, |
| "valid_targets_min": 1074 |
| }, |
| { |
| "epoch": 6.993006993006993, |
| "grad_norm": 0.6366022888173039, |
| "learning_rate": 1.9006867270077168e-10, |
| "loss": 0.243, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24022555351257324, |
| "step": 4000, |
| "valid_targets_mean": 3167.0, |
| "valid_targets_min": 1252 |
| }, |
| { |
| "epoch": 7.0, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23146472871303558, |
| "step": 4004, |
| "total_flos": 735767221108736.0, |
| "train_loss": 0.3356110471766907, |
| "train_runtime": 16919.046, |
| "train_samples_per_second": 3.782, |
| "train_steps_per_second": 0.237, |
| "valid_targets_mean": 3788.2, |
| "valid_targets_min": 1011 |
| } |
| ], |
| "logging_steps": 5, |
| "max_steps": 4004, |
| "num_input_tokens_seen": 0, |
| "num_train_epochs": 7, |
| "save_steps": 1500, |
| "stateful_callbacks": { |
| "TrainerControl": { |
| "args": { |
| "should_epoch_stop": false, |
| "should_evaluate": false, |
| "should_log": false, |
| "should_save": true, |
| "should_training_stop": true |
| }, |
| "attributes": {} |
| } |
| }, |
| "total_flos": 735767221108736.0, |
| "train_batch_size": 1, |
| "trial_name": null, |
| "trial_params": null |
| } |
|
|