mapo80/DeQA-Doc-Color
05
1{2 "best_metric": null,3 "best_model_checkpoint": null,4 "epoch": 3.0,5 "eval_steps": 500,6 "global_step": 330,7 "is_hyper_param_search": false,8 "is_local_process_zero": true,9 "is_world_process_zero": true,10 "log_history": [11 {12 "epoch": 0.01,13 "learning_rate": 2.0000000000000003e-06,14 "loss": 0.1218,15 "step": 116 },17 {18 "epoch": 0.02,19 "learning_rate": 4.000000000000001e-06,20 "loss": 0.1411,21 "step": 222 },23 {24 "epoch": 0.03,25 "learning_rate": 6e-06,26 "loss": 0.1144,27 "step": 328 },29 {30 "epoch": 0.04,31 "learning_rate": 8.000000000000001e-06,32 "loss": 0.1014,33 "step": 434 },35 {36 "epoch": 0.05,37 "learning_rate": 1e-05,38 "loss": 0.0579,39 "step": 540 },41 {42 "epoch": 0.05,43 "learning_rate": 1.2e-05,44 "loss": 0.0477,45 "step": 646 },47 {48 "epoch": 0.06,49 "learning_rate": 1.4e-05,50 "loss": 0.0924,51 "step": 752 },53 {54 "epoch": 0.07,55 "learning_rate": 1.6000000000000003e-05,56 "loss": 0.056,57 "step": 858 },59 {60 "epoch": 0.08,61 "learning_rate": 1.8e-05,62 "loss": 0.0583,63 "step": 964 },65 {66 "epoch": 0.09,67 "learning_rate": 2e-05,68 "loss": 0.0393,69 "step": 1070 },71 {72 "epoch": 0.1,73 "learning_rate": 1.999951808959328e-05,74 "loss": 0.055,75 "step": 1176 },77 {78 "epoch": 0.11,79 "learning_rate": 1.9998072404820648e-05,80 "loss": 0.0339,81 "step": 1282 },83 {84 "epoch": 0.12,85 "learning_rate": 1.9995663085020215e-05,86 "loss": 0.0491,87 "step": 1388 },89 {90 "epoch": 0.13,91 "learning_rate": 1.9992290362407232e-05,92 "loss": 0.0519,93 "step": 1494 },95 {96 "epoch": 0.14,97 "learning_rate": 1.9987954562051724e-05,98 "loss": 0.0747,99 "step": 15100 },101 {102 "epoch": 0.15,103 "learning_rate": 1.998265610184716e-05,104 "loss": 0.0449,105 "step": 16106 },107 {108 "epoch": 0.15,109 "learning_rate": 1.997639549247016e-05,110 "loss": 0.0546,111 "step": 17112 },113 {114 "epoch": 0.16,115 "learning_rate": 1.9969173337331283e-05,116 "loss": 0.0371,117 "step": 18118 },119 {120 "epoch": 0.17,121 "learning_rate": 1.9960990332516875e-05,122 "loss": 0.0331,123 "step": 19124 },125 {126 "epoch": 0.18,127 "learning_rate": 1.995184726672197e-05,128 "loss": 0.0295,129 "step": 20130 },131 {132 "epoch": 0.19,133 "learning_rate": 1.9941745021174284e-05,134 "loss": 0.0425,135 "step": 21136 },137 {138 "epoch": 0.2,139 "learning_rate": 1.9930684569549265e-05,140 "loss": 0.0341,141 "step": 22142 },143 {144 "epoch": 0.21,145 "learning_rate": 1.991866697787626e-05,146 "loss": 0.0249,147 "step": 23148 },149 {150 "epoch": 0.22,151 "learning_rate": 1.990569340443577e-05,152 "loss": 0.0453,153 "step": 24154 },155 {156 "epoch": 0.23,157 "learning_rate": 1.989176509964781e-05,158 "loss": 0.0383,159 "step": 25160 },161 {162 "epoch": 0.24,163 "learning_rate": 1.9876883405951378e-05,164 "loss": 0.0401,165 "step": 26166 },167 {168 "epoch": 0.25,169 "learning_rate": 1.9861049757675087e-05,170 "loss": 0.0169,171 "step": 27172 },173 {174 "epoch": 0.25,175 "learning_rate": 1.9844265680898917e-05,176 "loss": 0.0196,177 "step": 28178 },179 {180 "epoch": 0.26,181 "learning_rate": 1.982653279330712e-05,182 "loss": 0.0287,183 "step": 29184 },185 {186 "epoch": 0.27,187 "learning_rate": 1.9807852804032306e-05,188 "loss": 0.0645,189 "step": 30190 },191 {192 "epoch": 0.28,193 "learning_rate": 1.9788227513490724e-05,194 "loss": 0.0276,195 "step": 31196 },197 {198 "epoch": 0.29,199 "learning_rate": 1.9767658813208725e-05,200 "loss": 0.0257,201 "step": 32202 },203 {204 "epoch": 0.3,205 "learning_rate": 1.974614868564045e-05,206 "loss": 0.0373,207 "step": 33208 },209 {210 "epoch": 0.31,211 "learning_rate": 1.9723699203976768e-05,212 "loss": 0.0453,213 "step": 34214 },215 {216 "epoch": 0.32,217 "learning_rate": 1.9700312531945444e-05,218 "loss": 0.041,219 "step": 35220 },221 {222 "epoch": 0.33,223 "learning_rate": 1.96759909236026e-05,224 "loss": 0.0425,225 "step": 36226 },227 {228 "epoch": 0.34,229 "learning_rate": 1.9650736723115476e-05,230 "loss": 0.0271,231 "step": 37232 },233 {234 "epoch": 0.35,235 "learning_rate": 1.9624552364536472e-05,236 "loss": 0.0339,237 "step": 38238 },239 {240 "epoch": 0.35,241 "learning_rate": 1.9597440371568576e-05,242 "loss": 0.0466,243 "step": 39244 },245 {246 "epoch": 0.36,247 "learning_rate": 1.956940335732209e-05,248 "loss": 0.0281,249 "step": 40250 },251 {252 "epoch": 0.37,253 "learning_rate": 1.9540444024062807e-05,254 "loss": 0.0382,255 "step": 41256 },257 {258 "epoch": 0.38,259 "learning_rate": 1.9510565162951538e-05,260 "loss": 0.0273,261 "step": 42262 },263 {264 "epoch": 0.39,265 "learning_rate": 1.9479769653775107e-05,266 "loss": 0.0246,267 "step": 43268 },269 {270 "epoch": 0.4,271 "learning_rate": 1.944806046466878e-05,272 "loss": 0.0252,273 "step": 44274 },275 {276 "epoch": 0.41,277 "learning_rate": 1.941544065183021e-05,278 "loss": 0.0216,279 "step": 45280 },281 {282 "epoch": 0.42,283 "learning_rate": 1.9381913359224844e-05,284 "loss": 0.0299,285 "step": 46286 },287 {288 "epoch": 0.43,289 "learning_rate": 1.9347481818282927e-05,290 "loss": 0.0254,291 "step": 47292 },293 {294 "epoch": 0.44,295 "learning_rate": 1.9312149347588035e-05,296 "loss": 0.0276,297 "step": 48298 },299 {300 "epoch": 0.45,301 "learning_rate": 1.9275919352557242e-05,302 "loss": 0.0308,303 "step": 49304 },305 {306 "epoch": 0.45,307 "learning_rate": 1.9238795325112867e-05,308 "loss": 0.0425,309 "step": 50310 },311 {312 "epoch": 0.46,313 "learning_rate": 1.920078084334595e-05,314 "loss": 0.0211,315 "step": 51316 },317 {318 "epoch": 0.47,319 "learning_rate": 1.916187957117136e-05,320 "loss": 0.0207,321 "step": 52322 },323 {324 "epoch": 0.48,325 "learning_rate": 1.9122095257974676e-05,326 "loss": 0.0191,327 "step": 53328 },329 {330 "epoch": 0.49,331 "learning_rate": 1.9081431738250815e-05,332 "loss": 0.0201,333 "step": 54334 },335 {336 "epoch": 0.5,337 "learning_rate": 1.9039892931234434e-05,338 "loss": 0.0155,339 "step": 55340 },341 {342 "epoch": 0.51,343 "learning_rate": 1.8997482840522218e-05,344 "loss": 0.0363,345 "step": 56346 },347 {348 "epoch": 0.52,349 "learning_rate": 1.895420555368697e-05,350 "loss": 0.0198,351 "step": 57352 },353 {354 "epoch": 0.53,355 "learning_rate": 1.891006524188368e-05,356 "loss": 0.0167,357 "step": 58358 },359 {360 "epoch": 0.54,361 "learning_rate": 1.8865066159447468e-05,362 "loss": 0.0185,363 "step": 59364 },365 {366 "epoch": 0.55,367 "learning_rate": 1.881921264348355e-05,368 "loss": 0.0185,369 "step": 60370 },371 {372 "epoch": 0.55,373 "learning_rate": 1.8772509113449243e-05,374 "loss": 0.0125,375 "step": 61376 },377 {378 "epoch": 0.56,379 "learning_rate": 1.8724960070727974e-05,380 "loss": 0.0158,381 "step": 62382 },383 {384 "epoch": 0.57,385 "learning_rate": 1.8676570098195443e-05,386 "loss": 0.0212,387 "step": 63388 },389 {390 "epoch": 0.58,391 "learning_rate": 1.862734385977792e-05,392 "loss": 0.0123,393 "step": 64394 },395 {396 "epoch": 0.59,397 "learning_rate": 1.8577286100002723e-05,398 "loss": 0.0136,399 "step": 65400 },401 {402 "epoch": 0.6,403 "learning_rate": 1.8526401643540924e-05,404 "loss": 0.0109,405 "step": 66406 },407 {408 "epoch": 0.61,409 "learning_rate": 1.8474695394742345e-05,410 "loss": 0.0135,411 "step": 67412 },413 {414 "epoch": 0.62,415 "learning_rate": 1.8422172337162865e-05,416 "loss": 0.0159,417 "step": 68418 },419 {420 "epoch": 0.63,421 "learning_rate": 1.8368837533084092e-05,422 "loss": 0.0148,423 "step": 69424 },425 {426 "epoch": 0.64,427 "learning_rate": 1.8314696123025456e-05,428 "loss": 0.0154,429 "step": 70430 },431 {432 "epoch": 0.65,433 "learning_rate": 1.825975332524873e-05,434 "loss": 0.0183,435 "step": 71436 },437 {438 "epoch": 0.65,439 "learning_rate": 1.8204014435255136e-05,440 "loss": 0.0174,441 "step": 72442 },443 {444 "epoch": 0.66,445 "learning_rate": 1.8147484825274895e-05,446 "loss": 0.0279,447 "step": 73448 },449 {450 "epoch": 0.67,451 "learning_rate": 1.8090169943749477e-05,452 "loss": 0.014,453 "step": 74454 },455 {456 "epoch": 0.68,457 "learning_rate": 1.803207531480645e-05,458 "loss": 0.018,459 "step": 75460 },461 {462 "epoch": 0.69,463 "learning_rate": 1.797320653772707e-05,464 "loss": 0.026,465 "step": 76466 },467 {468 "epoch": 0.7,469 "learning_rate": 1.7913569286406606e-05,470 "loss": 0.0255,471 "step": 77472 },473 {474 "epoch": 0.71,475 "learning_rate": 1.785316930880745e-05,476 "loss": 0.0294,477 "step": 78478 },479 {480 "epoch": 0.72,481 "learning_rate": 1.779201242640517e-05,482 "loss": 0.0214,483 "step": 79484 },485 {486 "epoch": 0.73,487 "learning_rate": 1.773010453362737e-05,488 "loss": 0.0208,489 "step": 80490 },491 {492 "epoch": 0.74,493 "learning_rate": 1.7667451597285617e-05,494 "loss": 0.0174,495 "step": 81496 },497 {498 "epoch": 0.75,499 "learning_rate": 1.7604059656000313e-05,500 "loss": 0.0206,501 "step": 82502 },503 {504 "epoch": 0.75,505 "learning_rate": 1.7539934819618696e-05,506 "loss": 0.0191,507 "step": 83508 },509 {510 "epoch": 0.76,511 "learning_rate": 1.747508326862597e-05,512 "loss": 0.0258,513 "step": 84514 },515 {516 "epoch": 0.77,517 "learning_rate": 1.7409511253549592e-05,518 "loss": 0.0191,519 "step": 85520 },521 {522 "epoch": 0.78,523 "learning_rate": 1.7343225094356857e-05,524 "loss": 0.0409,525 "step": 86526 },527 {528 "epoch": 0.79,529 "learning_rate": 1.727623117984575e-05,530 "loss": 0.034,531 "step": 87532 },533 {534 "epoch": 0.8,535 "learning_rate": 1.720853596702919e-05,536 "loss": 0.0431,537 "step": 88538 },539 {540 "epoch": 0.81,541 "learning_rate": 1.7140145980512684e-05,542 "loss": 0.0427,543 "step": 89544 },545 {546 "epoch": 0.82,547 "learning_rate": 1.7071067811865477e-05,548 "loss": 0.022,549 "step": 90550 },551 {552 "epoch": 0.83,553 "learning_rate": 1.7001308118985237e-05,554 "loss": 0.023,555 "step": 91556 },557 {558 "epoch": 0.84,559 "learning_rate": 1.6930873625456362e-05,560 "loss": 0.0351,561 "step": 92562 },563 {564 "epoch": 0.85,565 "learning_rate": 1.685977111990193e-05,566 "loss": 0.0295,567 "step": 93568 },569 {570 "epoch": 0.85,571 "learning_rate": 1.678800745532942e-05,572 "loss": 0.0157,573 "step": 94574 },575 {576 "epoch": 0.86,577 "learning_rate": 1.6715589548470187e-05,578 "loss": 0.0165,579 "step": 95580 },581 {582 "epoch": 0.87,583 "learning_rate": 1.664252437911282e-05,584 "loss": 0.0193,585 "step": 96586 },587 {588 "epoch": 0.88,589 "learning_rate": 1.6568818989430416e-05,590 "loss": 0.0305,591 "step": 97592 },593 {594 "epoch": 0.89,595 "learning_rate": 1.6494480483301836e-05,596 "loss": 0.0324,597 "step": 98598 },599 {600 "epoch": 0.9,601 "learning_rate": 1.641951602562703e-05,602 "loss": 0.0117,603 "step": 99604 },605 {606 "epoch": 0.91,607 "learning_rate": 1.6343932841636455e-05,608 "loss": 0.0228,609 "step": 100610 },611 {612 "epoch": 0.92,613 "learning_rate": 1.6267738216194698e-05,614 "loss": 0.0181,615 "step": 101616 },617 {618 "epoch": 0.93,619 "learning_rate": 1.6190939493098344e-05,620 "loss": 0.0296,621 "step": 102622 },623 {624 "epoch": 0.94,625 "learning_rate": 1.6113544074368166e-05,626 "loss": 0.0275,627 "step": 103628 },629 {630 "epoch": 0.95,631 "learning_rate": 1.6035559419535714e-05,632 "loss": 0.0208,633 "step": 104634 },635 {636 "epoch": 0.95,637 "learning_rate": 1.5956993044924334e-05,638 "loss": 0.0161,639 "step": 105640 },641 {642 "epoch": 0.96,643 "learning_rate": 1.5877852522924733e-05,644 "loss": 0.0152,645 "step": 106646 },647 {648 "epoch": 0.97,649 "learning_rate": 1.579814548126514e-05,650 "loss": 0.0204,651 "step": 107652 },653 {654 "epoch": 0.98,655 "learning_rate": 1.5717879602276123e-05,656 "loss": 0.0225,657 "step": 108658 },659 {660 "epoch": 0.99,661 "learning_rate": 1.5637062622150168e-05,662 "loss": 0.0258,663 "step": 109664 },665 {666 "epoch": 1.0,667 "learning_rate": 1.5555702330196024e-05,668 "loss": 0.0127,669 "step": 110670 },671 {672 "epoch": 1.01,673 "learning_rate": 1.547380656808797e-05,674 "loss": 0.0246,675 "step": 111676 },677 {678 "epoch": 1.02,679 "learning_rate": 1.5391383229110005e-05,680 "loss": 0.0242,681 "step": 112682 },683 {684 "epoch": 1.03,685 "learning_rate": 1.5308440257395095e-05,686 "loss": 0.0187,687 "step": 113688 },689 {690 "epoch": 1.04,691 "learning_rate": 1.5224985647159489e-05,692 "loss": 0.0232,693 "step": 114694 },695 {696 "epoch": 1.05,697 "learning_rate": 1.5141027441932217e-05,698 "loss": 0.019,699 "step": 115700 },701 {702 "epoch": 1.05,703 "learning_rate": 1.5056573733779848e-05,704 "loss": 0.0219,705 "step": 116706 },707 {708 "epoch": 1.06,709 "learning_rate": 1.4971632662526545e-05,710 "loss": 0.0194,711 "step": 117712 },713 {714 "epoch": 1.07,715 "learning_rate": 1.4886212414969551e-05,716 "loss": 0.029,717 "step": 118718 },719 {720 "epoch": 1.08,721 "learning_rate": 1.4800321224090114e-05,722 "loss": 0.035,723 "step": 119724 },725 {726 "epoch": 1.09,727 "learning_rate": 1.4713967368259981e-05,728 "loss": 0.0165,729 "step": 120730 },731 {732 "epoch": 1.1,733 "learning_rate": 1.4627159170443504e-05,734 "loss": 0.014,735 "step": 121736 },737 {738 "epoch": 1.11,739 "learning_rate": 1.4539904997395468e-05,740 "loss": 0.0151,741 "step": 122742 },743 {744 "epoch": 1.12,745 "learning_rate": 1.4452213258854684e-05,746 "loss": 0.017,747 "step": 123748 },749 {750 "epoch": 1.13,751 "learning_rate": 1.436409240673342e-05,752 "loss": 0.0188,753 "step": 124754 },755 {756 "epoch": 1.14,757 "learning_rate": 1.4275550934302822e-05,758 "loss": 0.0128,759 "step": 125760 },761 {762 "epoch": 1.15,763 "learning_rate": 1.4186597375374283e-05,764 "loss": 0.0164,765 "step": 126766 },767 {768 "epoch": 1.15,769 "learning_rate": 1.4097240303476955e-05,770 "loss": 0.0113,771 "step": 127772 },773 {774 "epoch": 1.16,775 "learning_rate": 1.4007488331031409e-05,776 "loss": 0.0082,777 "step": 128778 },779 {780 "epoch": 1.17,781 "learning_rate": 1.391735010851956e-05,782 "loss": 0.015,783 "step": 129784 },785 {786 "epoch": 1.18,787 "learning_rate": 1.3826834323650899e-05,788 "loss": 0.0132,789 "step": 130790 },791 {792 "epoch": 1.19,793 "learning_rate": 1.3735949700525164e-05,794 "loss": 0.0127,795 "step": 131796 },797 {798 "epoch": 1.2,799 "learning_rate": 1.3644704998791501e-05,800 "loss": 0.0135,801 "step": 132802 },803 {804 "epoch": 1.21,805 "learning_rate": 1.3553109012804162e-05,806 "loss": 0.0172,807 "step": 133808 },809 {810 "epoch": 1.22,811 "learning_rate": 1.346117057077493e-05,812 "loss": 0.017,813 "step": 134814 },815 {816 "epoch": 1.23,817 "learning_rate": 1.3368898533922202e-05,818 "loss": 0.0206,819 "step": 135820 },821 {822 "epoch": 1.24,823 "learning_rate": 1.3276301795616937e-05,824 "loss": 0.0201,825 "step": 136826 },827 {828 "epoch": 1.25,829 "learning_rate": 1.3183389280525497e-05,830 "loss": 0.0103,831 "step": 137832 },833 {834 "epoch": 1.25,835 "learning_rate": 1.3090169943749475e-05,836 "loss": 0.0098,837 "step": 138838 },839 {840 "epoch": 1.26,841 "learning_rate": 1.2996652769962567e-05,842 "loss": 0.016,843 "step": 139844 },845 {846 "epoch": 1.27,847 "learning_rate": 1.2902846772544625e-05,848 "loss": 0.0195,849 "step": 140850 },851 {852 "epoch": 1.28,853 "learning_rate": 1.2808760992712923e-05,854 "loss": 0.0208,855 "step": 141856 },857 {858 "epoch": 1.29,859 "learning_rate": 1.2714404498650743e-05,860 "loss": 0.0159,861 "step": 142862 },863 {864 "epoch": 1.3,865 "learning_rate": 1.2619786384633374e-05,866 "loss": 0.0121,867 "step": 143868 },869 {870 "epoch": 1.31,871 "learning_rate": 1.252491577015158e-05,872 "loss": 0.016,873 "step": 144874 },875 {876 "epoch": 1.32,877 "learning_rate": 1.242980179903264e-05,878 "loss": 0.0134,879 "step": 145880 },881 {882 "epoch": 1.33,883 "learning_rate": 1.2334453638559057e-05,884 "loss": 0.015,885 "step": 146886 },887 {888 "epoch": 1.34,889 "learning_rate": 1.2238880478584987e-05,890 "loss": 0.0132,891 "step": 147892 },893 {894 "epoch": 1.35,895 "learning_rate": 1.2143091530650508e-05,896 "loss": 0.0132,897 "step": 148898 },899 {900 "epoch": 1.35,901 "learning_rate": 1.2047096027093798e-05,902 "loss": 0.0228,903 "step": 149904 },905 {906 "epoch": 1.36,907 "learning_rate": 1.1950903220161286e-05,908 "loss": 0.0157,909 "step": 150910 },911 {912 "epoch": 1.37,913 "learning_rate": 1.185452238111591e-05,914 "loss": 0.0105,915 "step": 151916 },917 {918 "epoch": 1.38,919 "learning_rate": 1.1757962799343548e-05,920 "loss": 0.0183,921 "step": 152922 },923 {924 "epoch": 1.39,925 "learning_rate": 1.1661233781457655e-05,926 "loss": 0.0192,927 "step": 153928 },929 {930 "epoch": 1.4,931 "learning_rate": 1.156434465040231e-05,932 "loss": 0.0117,933 "step": 154934 },935 {936 "epoch": 1.41,937 "learning_rate": 1.1467304744553618e-05,938 "loss": 0.0155,939 "step": 155940 },941 {942 "epoch": 1.42,943 "learning_rate": 1.1370123416819683e-05,944 "loss": 0.0183,945 "step": 156946 },947 {948 "epoch": 1.43,949 "learning_rate": 1.1272810033739134e-05,950 "loss": 0.0122,951 "step": 157952 },953 {954 "epoch": 1.44,955 "learning_rate": 1.1175373974578378e-05,956 "loss": 0.0159,957 "step": 158958 },959 {960 "epoch": 1.45,961 "learning_rate": 1.1077824630427593e-05,962 "loss": 0.0083,963 "step": 159964 },965 {966 "epoch": 1.45,967 "learning_rate": 1.098017140329561e-05,968 "loss": 0.0121,969 "step": 160970 },971 {972 "epoch": 1.46,973 "learning_rate": 1.0882423705203698e-05,974 "loss": 0.0142,975 "step": 161976 },977 {978 "epoch": 1.47,979 "learning_rate": 1.0784590957278452e-05,980 "loss": 0.0099,981 "step": 162982 },983 {984 "epoch": 1.48,985 "learning_rate": 1.0686682588843737e-05,986 "loss": 0.0114,987 "step": 163988 },989 {990 "epoch": 1.49,991 "learning_rate": 1.058870803651189e-05,992 "loss": 0.0156,993 "step": 164994 },995 {996 "epoch": 1.5,997 "learning_rate": 1.0490676743274181e-05,998 "loss": 0.0091,999 "step": 1651000 },1001 {1002 "epoch": 1.51,1003 "learning_rate": 1.0392598157590687e-05,1004 "loss": 0.0098,1005 "step": 1661006 },1007 {1008 "epoch": 1.52,1009 "learning_rate": 1.0294481732479635e-05,1010 "loss": 0.0087,1011 "step": 1671012 },1013 {1014 "epoch": 1.53,1015 "learning_rate": 1.0196336924606282e-05,1016 "loss": 0.0118,1017 "step": 1681018 },1019 {1020 "epoch": 1.54,1021 "learning_rate": 1.0098173193371498e-05,1022 "loss": 0.0116,1023 "step": 1691024 },1025 {1026 "epoch": 1.55,1027 "learning_rate": 1e-05,1028 "loss": 0.0096,1029 "step": 1701030 },1031 {1032 "epoch": 1.55,1033 "learning_rate": 9.901826806628505e-06,1034 "loss": 0.01,1035 "step": 1711036 },1037 {1038 "epoch": 1.56,1039 "learning_rate": 9.80366307539372e-06,1040 "loss": 0.0076,1041 "step": 1721042 },1043 {1044 "epoch": 1.57,1045 "learning_rate": 9.705518267520369e-06,1046 "loss": 0.0106,1047 "step": 1731048 },1049 {1050 "epoch": 1.58,1051 "learning_rate": 9.607401842409318e-06,1052 "loss": 0.0111,1053 "step": 1741054 },1055 {1056 "epoch": 1.59,1057 "learning_rate": 9.50932325672582e-06,1058 "loss": 0.0085,1059 "step": 1751060 },1061 {1062 "epoch": 1.6,1063 "learning_rate": 9.41129196348811e-06,1064 "loss": 0.0094,1065 "step": 1761066 },1067 {1068 "epoch": 1.61,1069 "learning_rate": 9.313317411156265e-06,1070 "loss": 0.0138,1071 "step": 1771072 },1073 {1074 "epoch": 1.62,1075 "learning_rate": 9.215409042721553e-06,1076 "loss": 0.0114,1077 "step": 1781078 },1079 {1080 "epoch": 1.63,1081 "learning_rate": 9.117576294796307e-06,1082 "loss": 0.0112,1083 "step": 1791084 },1085 {1086 "epoch": 1.64,1087 "learning_rate": 9.019828596704394e-06,1088 "loss": 0.0119,1089 "step": 1801090 },1091 {1092 "epoch": 1.65,1093 "learning_rate": 8.922175369572407e-06,1094 "loss": 0.0099,1095 "step": 1811096 },1097 {1098 "epoch": 1.65,1099 "learning_rate": 8.824626025421625e-06,1100 "loss": 0.0089,1101 "step": 1821102 },1103 {1104 "epoch": 1.66,1105 "learning_rate": 8.72718996626087e-06,1106 "loss": 0.0115,1107 "step": 1831108 },1109 {1110 "epoch": 1.67,1111 "learning_rate": 8.629876583180322e-06,1112 "loss": 0.012,1113 "step": 1841114 },1115 {1116 "epoch": 1.68,1117 "learning_rate": 8.532695255446384e-06,1118 "loss": 0.0107,1119 "step": 1851120 },1121 {1122 "epoch": 1.69,1123 "learning_rate": 8.43565534959769e-06,1124 "loss": 0.0142,1125 "step": 1861126 },1127 {1128 "epoch": 1.7,1129 "learning_rate": 8.338766218542348e-06,1130 "loss": 0.0082,1131 "step": 1871132 },1133 {1134 "epoch": 1.71,1135 "learning_rate": 8.242037200656455e-06,1136 "loss": 0.0248,1137 "step": 1881138 },1139 {1140 "epoch": 1.72,1141 "learning_rate": 8.145477618884092e-06,1142 "loss": 0.0103,1143 "step": 1891144 },1145 {1146 "epoch": 1.73,1147 "learning_rate": 8.04909677983872e-06,1148 "loss": 0.015,1149 "step": 1901150 },1151 {1152 "epoch": 1.74,1153 "learning_rate": 7.952903972906205e-06,1154 "loss": 0.0098,1155 "step": 1911156 },1157 {1158 "epoch": 1.75,1159 "learning_rate": 7.856908469349495e-06,1160 "loss": 0.0097,1161 "step": 1921162 },1163 {1164 "epoch": 1.75,1165 "learning_rate": 7.761119521415017e-06,1166 "loss": 0.0078,1167 "step": 1931168 },1169 {1170 "epoch": 1.76,1171 "learning_rate": 7.66554636144095e-06,1172 "loss": 0.0075,1173 "step": 1941174 },1175 {1176 "epoch": 1.77,1177 "learning_rate": 7.570198200967363e-06,1178 "loss": 0.0266,1179 "step": 1951180 },1181 {1182 "epoch": 1.78,1183 "learning_rate": 7.4750842298484205e-06,1184 "loss": 0.0095,1185 "step": 1961186 },1187 {1188 "epoch": 1.79,1189 "learning_rate": 7.380213615366627e-06,1190 "loss": 0.0156,1191 "step": 1971192 },1193 {1194 "epoch": 1.8,1195 "learning_rate": 7.285595501349259e-06,1196 "loss": 0.0208,1197 "step": 1981198 },1199 {1200 "epoch": 1.81,