| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 2.035964035964036, | |
| "global_step": 2038, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 0.0012782749814122844, | |
| "loss": 1.6279, | |
| "step": 5 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0018077538151554684, | |
| "loss": 1.2509, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0022140372138502386, | |
| "loss": 1.4568, | |
| "step": 15 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0025565499628245687, | |
| "loss": 1.4557, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0028583097523751473, | |
| "loss": 1.1046, | |
| "step": 25 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0031311214554257476, | |
| "loss": 1.3079, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.003381997707972616, | |
| "loss": 1.1913, | |
| "step": 35 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0036155076303109367, | |
| "loss": 1.415, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.003834824944236853, | |
| "loss": 1.2117, | |
| "step": 45 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0040422604172722166, | |
| "loss": 1.3286, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.004239558492243069, | |
| "loss": 1.2645, | |
| "step": 55 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.004428074427700477, | |
| "loss": 1.3262, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0046088859896247685, | |
| "loss": 1.3905, | |
| "step": 65 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.004782867026529596, | |
| "loss": 1.2985, | |
| "step": 70 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.004950737714883372, | |
| "loss": 1.1367, | |
| "step": 75 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0051130999256491375, | |
| "loss": 1.3175, | |
| "step": 80 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.005270462766947299, | |
| "loss": 1.4444, | |
| "step": 85 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0054232614454664055, | |
| "loss": 1.2882, | |
| "step": 90 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.005571871466032479, | |
| "loss": 1.4907, | |
| "step": 95 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0057166195047502946, | |
| "loss": 1.4369, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.005857791861290061, | |
| "loss": 1.2819, | |
| "step": 105 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.005995641118204179, | |
| "loss": 1.3525, | |
| "step": 110 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.006130391451319013, | |
| "loss": 1.3266, | |
| "step": 115 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.006262242910851495, | |
| "loss": 1.4452, | |
| "step": 120 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.006391374907061422, | |
| "loss": 1.5723, | |
| "step": 125 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.006517949073958691, | |
| "loss": 1.4178, | |
| "step": 130 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0066421116415507145, | |
| "loss": 1.4234, | |
| "step": 135 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.006763995415945232, | |
| "loss": 1.5096, | |
| "step": 140 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.006883721443741945, | |
| "loss": 1.3332, | |
| "step": 145 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.007001400420140049, | |
| "loss": 1.5342, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.007117133887404524, | |
| "loss": 1.5825, | |
| "step": 155 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.007231015260621873, | |
| "loss": 1.3028, | |
| "step": 160 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.007343130710225101, | |
| "loss": 1.5338, | |
| "step": 165 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0074535599249993005, | |
| "loss": 1.4564, | |
| "step": 170 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.007562376774775252, | |
| "loss": 1.319, | |
| "step": 175 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.007669649888473706, | |
| "loss": 1.6363, | |
| "step": 180 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.007775443160352296, | |
| "loss": 1.4293, | |
| "step": 185 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.007879816195062792, | |
| "loss": 1.4431, | |
| "step": 190 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.007982824700322464, | |
| "loss": 1.3983, | |
| "step": 195 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.008084520834544433, | |
| "loss": 1.448, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.008184953515585021, | |
| "loss": 1.6799, | |
| "step": 205 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.008284168695795142, | |
| "loss": 1.5823, | |
| "step": 210 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.008382209607764337, | |
| "loss": 1.4564, | |
| "step": 215 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.008479116984486139, | |
| "loss": 1.576, | |
| "step": 220 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.008574929257125443, | |
| "loss": 1.5307, | |
| "step": 225 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.00866968273311143, | |
| "loss": 1.4519, | |
| "step": 230 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.008763411756896685, | |
| "loss": 1.6421, | |
| "step": 235 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.008856148855400954, | |
| "loss": 1.321, | |
| "step": 240 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.008947924869885989, | |
| "loss": 1.3645, | |
| "step": 245 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.009038769075777341, | |
| "loss": 1.6528, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.00912870929175277, | |
| "loss": 1.5315, | |
| "step": 255 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.009217771979249537, | |
| "loss": 1.6313, | |
| "step": 260 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.00930598233339944, | |
| "loss": 1.6657, | |
| "step": 265 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.009393364366277242, | |
| "loss": 1.602, | |
| "step": 270 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.00947994098324202, | |
| "loss": 1.5175, | |
| "step": 275 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.009565734053059192, | |
| "loss": 1.4379, | |
| "step": 280 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.009650764472411541, | |
| "loss": 1.5895, | |
| "step": 285 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.009735052225338362, | |
| "loss": 1.9034, | |
| "step": 290 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.00981861643808179, | |
| "loss": 1.5616, | |
| "step": 295 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.009901475429766745, | |
| "loss": 1.5318, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.009983646759294876, | |
| "loss": 1.6532, | |
| "step": 305 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.009976905542725173, | |
| "loss": 1.6601, | |
| "step": 310 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.00994803747113164, | |
| "loss": 1.9683, | |
| "step": 315 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.009919169399538106, | |
| "loss": 1.694, | |
| "step": 320 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.009890301327944573, | |
| "loss": 1.4985, | |
| "step": 325 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.009861433256351039, | |
| "loss": 1.7896, | |
| "step": 330 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.009832565184757506, | |
| "loss": 1.6075, | |
| "step": 335 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.009803697113163973, | |
| "loss": 1.7854, | |
| "step": 340 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.009774829041570439, | |
| "loss": 1.6752, | |
| "step": 345 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.009745960969976906, | |
| "loss": 1.6439, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.009717092898383371, | |
| "loss": 1.9989, | |
| "step": 355 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.009688224826789838, | |
| "loss": 1.8198, | |
| "step": 360 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.009659356755196306, | |
| "loss": 1.8466, | |
| "step": 365 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.009630488683602771, | |
| "loss": 1.7034, | |
| "step": 370 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.009601620612009238, | |
| "loss": 1.6235, | |
| "step": 375 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.009572752540415704, | |
| "loss": 1.5958, | |
| "step": 380 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.009543884468822171, | |
| "loss": 1.7292, | |
| "step": 385 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.009515016397228638, | |
| "loss": 1.6004, | |
| "step": 390 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.009486148325635104, | |
| "loss": 1.8302, | |
| "step": 395 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.009457280254041571, | |
| "loss": 1.6322, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.009428412182448036, | |
| "loss": 1.4802, | |
| "step": 405 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.009399544110854504, | |
| "loss": 1.5544, | |
| "step": 410 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.009370676039260971, | |
| "loss": 1.6265, | |
| "step": 415 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.009341807967667436, | |
| "loss": 1.7169, | |
| "step": 420 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.009312939896073904, | |
| "loss": 1.511, | |
| "step": 425 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.009284071824480369, | |
| "loss": 1.63, | |
| "step": 430 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.009255203752886836, | |
| "loss": 1.6076, | |
| "step": 435 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.009226335681293303, | |
| "loss": 1.5973, | |
| "step": 440 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.009197467609699769, | |
| "loss": 1.6757, | |
| "step": 445 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.009168599538106236, | |
| "loss": 1.5146, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.009139731466512702, | |
| "loss": 1.459, | |
| "step": 455 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.009110863394919169, | |
| "loss": 1.4292, | |
| "step": 460 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.009081995323325636, | |
| "loss": 1.7206, | |
| "step": 465 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.009053127251732102, | |
| "loss": 1.491, | |
| "step": 470 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.009024259180138567, | |
| "loss": 1.5426, | |
| "step": 475 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.008995391108545034, | |
| "loss": 1.6884, | |
| "step": 480 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.008966523036951501, | |
| "loss": 1.6555, | |
| "step": 485 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.008937654965357967, | |
| "loss": 1.6036, | |
| "step": 490 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.008908786893764434, | |
| "loss": 1.4405, | |
| "step": 495 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.008879918822170901, | |
| "loss": 1.5134, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.008851050750577369, | |
| "loss": 1.6087, | |
| "step": 505 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.008822182678983834, | |
| "loss": 1.6898, | |
| "step": 510 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.0087933146073903, | |
| "loss": 1.7172, | |
| "step": 515 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.008764446535796767, | |
| "loss": 1.6621, | |
| "step": 520 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.008735578464203234, | |
| "loss": 1.4144, | |
| "step": 525 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.008706710392609701, | |
| "loss": 1.685, | |
| "step": 530 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.008677842321016167, | |
| "loss": 1.6112, | |
| "step": 535 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.008648974249422632, | |
| "loss": 1.5041, | |
| "step": 540 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.0086201061778291, | |
| "loss": 1.5515, | |
| "step": 545 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.008591238106235567, | |
| "loss": 1.711, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.008562370034642034, | |
| "loss": 1.6166, | |
| "step": 555 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.0085335019630485, | |
| "loss": 1.5324, | |
| "step": 560 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.008504633891454965, | |
| "loss": 1.4422, | |
| "step": 565 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.008475765819861432, | |
| "loss": 1.6069, | |
| "step": 570 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.008446897748267899, | |
| "loss": 1.6231, | |
| "step": 575 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.008418029676674366, | |
| "loss": 1.3719, | |
| "step": 580 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.008389161605080832, | |
| "loss": 1.6578, | |
| "step": 585 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.008360293533487297, | |
| "loss": 1.6525, | |
| "step": 590 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.008331425461893764, | |
| "loss": 1.6094, | |
| "step": 595 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.008302557390300232, | |
| "loss": 1.633, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.008273689318706697, | |
| "loss": 1.7865, | |
| "step": 605 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.008244821247113164, | |
| "loss": 1.602, | |
| "step": 610 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.00821595317551963, | |
| "loss": 1.6723, | |
| "step": 615 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.008187085103926097, | |
| "loss": 1.5495, | |
| "step": 620 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.008158217032332564, | |
| "loss": 1.5716, | |
| "step": 625 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.00812934896073903, | |
| "loss": 1.6036, | |
| "step": 630 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.008100480889145497, | |
| "loss": 1.4638, | |
| "step": 635 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.008071612817551962, | |
| "loss": 1.6912, | |
| "step": 640 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.00804274474595843, | |
| "loss": 1.6037, | |
| "step": 645 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.008013876674364897, | |
| "loss": 1.5457, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.007985008602771362, | |
| "loss": 1.4561, | |
| "step": 655 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.00795614053117783, | |
| "loss": 1.4871, | |
| "step": 660 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.007927272459584295, | |
| "loss": 1.5253, | |
| "step": 665 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.007898404387990762, | |
| "loss": 1.5775, | |
| "step": 670 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.00786953631639723, | |
| "loss": 1.724, | |
| "step": 675 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.007840668244803695, | |
| "loss": 1.6629, | |
| "step": 680 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.007811800173210163, | |
| "loss": 1.519, | |
| "step": 685 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 0.00778293210161663, | |
| "loss": 1.5684, | |
| "step": 690 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 0.007754064030023096, | |
| "loss": 1.6109, | |
| "step": 695 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 0.007725195958429562, | |
| "loss": 1.6378, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 0.007696327886836028, | |
| "loss": 1.4673, | |
| "step": 705 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 0.007667459815242496, | |
| "loss": 1.3671, | |
| "step": 710 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 0.007638591743648962, | |
| "loss": 1.6971, | |
| "step": 715 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 0.007609723672055428, | |
| "loss": 1.5656, | |
| "step": 720 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 0.007580855600461894, | |
| "loss": 1.6135, | |
| "step": 725 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 0.007551987528868362, | |
| "loss": 1.6247, | |
| "step": 730 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 0.007523119457274827, | |
| "loss": 1.4691, | |
| "step": 735 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 0.007494251385681295, | |
| "loss": 1.5108, | |
| "step": 740 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 0.007465383314087762, | |
| "loss": 1.7138, | |
| "step": 745 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 0.007436515242494226, | |
| "loss": 1.4466, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 0.007407647170900693, | |
| "loss": 1.6503, | |
| "step": 755 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 0.00737877909930716, | |
| "loss": 1.5161, | |
| "step": 760 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 0.007349911027713627, | |
| "loss": 1.5638, | |
| "step": 765 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 0.007321042956120093, | |
| "loss": 1.6321, | |
| "step": 770 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 0.007292174884526559, | |
| "loss": 1.5957, | |
| "step": 775 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 0.0072633068129330245, | |
| "loss": 1.4813, | |
| "step": 780 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 0.0072344387413394926, | |
| "loss": 1.6003, | |
| "step": 785 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 0.00720557066974596, | |
| "loss": 1.474, | |
| "step": 790 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 0.007176702598152426, | |
| "loss": 1.6341, | |
| "step": 795 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 0.007147834526558892, | |
| "loss": 1.6502, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 0.007118966454965358, | |
| "loss": 1.4565, | |
| "step": 805 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 0.007090098383371824, | |
| "loss": 1.4588, | |
| "step": 810 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 0.007061230311778292, | |
| "loss": 1.5407, | |
| "step": 815 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 0.007032362240184759, | |
| "loss": 1.3778, | |
| "step": 820 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 0.007003494168591224, | |
| "loss": 1.5602, | |
| "step": 825 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 0.0069746260969976905, | |
| "loss": 1.6861, | |
| "step": 830 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 0.006945758025404158, | |
| "loss": 1.5511, | |
| "step": 835 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 0.006916889953810624, | |
| "loss": 1.44, | |
| "step": 840 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 0.006888021882217091, | |
| "loss": 1.4159, | |
| "step": 845 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 0.006859153810623558, | |
| "loss": 1.6642, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 0.006830285739030023, | |
| "loss": 1.461, | |
| "step": 855 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 0.00680141766743649, | |
| "loss": 1.6532, | |
| "step": 860 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 0.0067725495958429575, | |
| "loss": 1.5106, | |
| "step": 865 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 0.006743681524249424, | |
| "loss": 1.564, | |
| "step": 870 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 0.00671481345265589, | |
| "loss": 1.5954, | |
| "step": 875 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 0.006685945381062356, | |
| "loss": 1.5075, | |
| "step": 880 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 0.006657077309468823, | |
| "loss": 1.4821, | |
| "step": 885 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 0.006628209237875289, | |
| "loss": 1.6993, | |
| "step": 890 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 0.006599341166281756, | |
| "loss": 1.3872, | |
| "step": 895 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 0.006570473094688223, | |
| "loss": 1.684, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 0.006541605023094689, | |
| "loss": 1.3419, | |
| "step": 905 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 0.006512736951501155, | |
| "loss": 1.3974, | |
| "step": 910 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 0.006483868879907622, | |
| "loss": 1.5703, | |
| "step": 915 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 0.006455000808314089, | |
| "loss": 1.5572, | |
| "step": 920 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 0.006426132736720555, | |
| "loss": 1.5941, | |
| "step": 925 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 0.006397264665127021, | |
| "loss": 1.706, | |
| "step": 930 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 0.006368396593533487, | |
| "loss": 1.5107 |