tosin/LLaDoc
024
1{2 "best_metric": null,3 "best_model_checkpoint": null,4 "epoch": 12.0,5 "global_step": 3480,6 "is_hyper_param_search": false,7 "is_local_process_zero": true,8 "is_world_process_zero": true,9 "log_history": [10 {11 "epoch": 0.0,12 "learning_rate": 1.904761904761905e-07,13 "loss": 2.6677,14 "step": 115 },16 {17 "epoch": 0.01,18 "learning_rate": 3.80952380952381e-07,19 "loss": 2.6284,20 "step": 221 },22 {23 "epoch": 0.01,24 "learning_rate": 5.714285714285715e-07,25 "loss": 2.5856,26 "step": 327 },28 {29 "epoch": 0.01,30 "learning_rate": 7.61904761904762e-07,31 "loss": 2.5541,32 "step": 433 },34 {35 "epoch": 0.02,36 "learning_rate": 9.523809523809525e-07,37 "loss": 2.5865,38 "step": 539 },40 {41 "epoch": 0.02,42 "learning_rate": 1.142857142857143e-06,43 "loss": 2.632,44 "step": 645 },46 {47 "epoch": 0.02,48 "learning_rate": 1.3333333333333334e-06,49 "loss": 2.4835,50 "step": 751 },52 {53 "epoch": 0.03,54 "learning_rate": 1.523809523809524e-06,55 "loss": 2.401,56 "step": 857 },58 {59 "epoch": 0.03,60 "learning_rate": 1.7142857142857145e-06,61 "loss": 2.1792,62 "step": 963 },64 {65 "epoch": 0.03,66 "learning_rate": 1.904761904761905e-06,67 "loss": 2.1515,68 "step": 1069 },70 {71 "epoch": 0.04,72 "learning_rate": 2.0952380952380955e-06,73 "loss": 1.88,74 "step": 1175 },76 {77 "epoch": 0.04,78 "learning_rate": 2.285714285714286e-06,79 "loss": 1.819,80 "step": 1281 },82 {83 "epoch": 0.04,84 "learning_rate": 2.4761904761904764e-06,85 "loss": 1.8622,86 "step": 1387 },88 {89 "epoch": 0.05,90 "learning_rate": 2.666666666666667e-06,91 "loss": 1.7666,92 "step": 1493 },94 {95 "epoch": 0.05,96 "learning_rate": 2.8571428571428573e-06,97 "loss": 1.6463,98 "step": 1599 },100 {101 "epoch": 0.06,102 "learning_rate": 3.047619047619048e-06,103 "loss": 1.6241,104 "step": 16105 },106 {107 "epoch": 0.06,108 "learning_rate": 3.2380952380952385e-06,109 "loss": 1.6968,110 "step": 17111 },112 {113 "epoch": 0.06,114 "learning_rate": 3.428571428571429e-06,115 "loss": 1.827,116 "step": 18117 },118 {119 "epoch": 0.07,120 "learning_rate": 3.6190476190476194e-06,121 "loss": 1.7846,122 "step": 19123 },124 {125 "epoch": 0.07,126 "learning_rate": 3.80952380952381e-06,127 "loss": 1.5993,128 "step": 20129 },130 {131 "epoch": 0.07,132 "learning_rate": 4.000000000000001e-06,133 "loss": 1.6619,134 "step": 21135 },136 {137 "epoch": 0.08,138 "learning_rate": 4.190476190476191e-06,139 "loss": 1.5851,140 "step": 22141 },142 {143 "epoch": 0.08,144 "learning_rate": 4.3809523809523815e-06,145 "loss": 1.5793,146 "step": 23147 },148 {149 "epoch": 0.08,150 "learning_rate": 4.571428571428572e-06,151 "loss": 1.6287,152 "step": 24153 },154 {155 "epoch": 0.09,156 "learning_rate": 4.761904761904762e-06,157 "loss": 1.5258,158 "step": 25159 },160 {161 "epoch": 0.09,162 "learning_rate": 4.952380952380953e-06,163 "loss": 1.4903,164 "step": 26165 },166 {167 "epoch": 0.09,168 "learning_rate": 5.142857142857142e-06,169 "loss": 1.4062,170 "step": 27171 },172 {173 "epoch": 0.1,174 "learning_rate": 5.333333333333334e-06,175 "loss": 1.5041,176 "step": 28177 },178 {179 "epoch": 0.1,180 "learning_rate": 5.523809523809525e-06,181 "loss": 1.6663,182 "step": 29183 },184 {185 "epoch": 0.1,186 "learning_rate": 5.7142857142857145e-06,187 "loss": 1.4352,188 "step": 30189 },190 {191 "epoch": 0.11,192 "learning_rate": 5.904761904761905e-06,193 "loss": 1.6159,194 "step": 31195 },196 {197 "epoch": 0.11,198 "learning_rate": 6.095238095238096e-06,199 "loss": 1.4788,200 "step": 32201 },202 {203 "epoch": 0.11,204 "learning_rate": 6.285714285714286e-06,205 "loss": 1.7855,206 "step": 33207 },208 {209 "epoch": 0.12,210 "learning_rate": 6.476190476190477e-06,211 "loss": 1.5745,212 "step": 34213 },214 {215 "epoch": 0.12,216 "learning_rate": 6.666666666666667e-06,217 "loss": 1.3839,218 "step": 35219 },220 {221 "epoch": 0.12,222 "learning_rate": 6.857142857142858e-06,223 "loss": 1.5843,224 "step": 36225 },226 {227 "epoch": 0.13,228 "learning_rate": 7.047619047619048e-06,229 "loss": 1.5361,230 "step": 37231 },232 {233 "epoch": 0.13,234 "learning_rate": 7.238095238095239e-06,235 "loss": 1.5176,236 "step": 38237 },238 {239 "epoch": 0.13,240 "learning_rate": 7.428571428571429e-06,241 "loss": 1.3972,242 "step": 39243 },244 {245 "epoch": 0.14,246 "learning_rate": 7.61904761904762e-06,247 "loss": 1.4516,248 "step": 40249 },250 {251 "epoch": 0.14,252 "learning_rate": 7.809523809523811e-06,253 "loss": 1.4685,254 "step": 41255 },256 {257 "epoch": 0.14,258 "learning_rate": 8.000000000000001e-06,259 "loss": 1.4953,260 "step": 42261 },262 {263 "epoch": 0.15,264 "learning_rate": 8.190476190476192e-06,265 "loss": 1.4643,266 "step": 43267 },268 {269 "epoch": 0.15,270 "learning_rate": 8.380952380952382e-06,271 "loss": 1.5037,272 "step": 44273 },274 {275 "epoch": 0.16,276 "learning_rate": 8.571428571428571e-06,277 "loss": 1.4227,278 "step": 45279 },280 {281 "epoch": 0.16,282 "learning_rate": 8.761904761904763e-06,283 "loss": 1.2973,284 "step": 46285 },286 {287 "epoch": 0.16,288 "learning_rate": 8.952380952380953e-06,289 "loss": 1.4621,290 "step": 47291 },292 {293 "epoch": 0.17,294 "learning_rate": 9.142857142857144e-06,295 "loss": 1.3477,296 "step": 48297 },298 {299 "epoch": 0.17,300 "learning_rate": 9.333333333333334e-06,301 "loss": 1.3225,302 "step": 49303 },304 {305 "epoch": 0.17,306 "learning_rate": 9.523809523809525e-06,307 "loss": 1.3506,308 "step": 50309 },310 {311 "epoch": 0.18,312 "learning_rate": 9.714285714285715e-06,313 "loss": 1.3848,314 "step": 51315 },316 {317 "epoch": 0.18,318 "learning_rate": 9.904761904761906e-06,319 "loss": 1.2545,320 "step": 52321 },322 {323 "epoch": 0.18,324 "learning_rate": 1.0095238095238096e-05,325 "loss": 1.3143,326 "step": 53327 },328 {329 "epoch": 0.19,330 "learning_rate": 1.0285714285714285e-05,331 "loss": 1.5702,332 "step": 54333 },334 {335 "epoch": 0.19,336 "learning_rate": 1.0476190476190477e-05,337 "loss": 1.4382,338 "step": 55339 },340 {341 "epoch": 0.19,342 "learning_rate": 1.0666666666666667e-05,343 "loss": 1.4677,344 "step": 56345 },346 {347 "epoch": 0.2,348 "learning_rate": 1.0857142857142858e-05,349 "loss": 1.3729,350 "step": 57351 },352 {353 "epoch": 0.2,354 "learning_rate": 1.104761904761905e-05,355 "loss": 1.3802,356 "step": 58357 },358 {359 "epoch": 0.2,360 "learning_rate": 1.1238095238095239e-05,361 "loss": 1.3365,362 "step": 59363 },364 {365 "epoch": 0.21,366 "learning_rate": 1.1428571428571429e-05,367 "loss": 1.3134,368 "step": 60369 },370 {371 "epoch": 0.21,372 "learning_rate": 1.1619047619047621e-05,373 "loss": 1.394,374 "step": 61375 },376 {377 "epoch": 0.21,378 "learning_rate": 1.180952380952381e-05,379 "loss": 1.2911,380 "step": 62381 },382 {383 "epoch": 0.22,384 "learning_rate": 1.2e-05,385 "loss": 1.3986,386 "step": 63387 },388 {389 "epoch": 0.22,390 "learning_rate": 1.2190476190476192e-05,391 "loss": 1.3676,392 "step": 64393 },394 {395 "epoch": 0.22,396 "learning_rate": 1.2380952380952383e-05,397 "loss": 1.4231,398 "step": 65399 },400 {401 "epoch": 0.23,402 "learning_rate": 1.2571428571428572e-05,403 "loss": 1.1589,404 "step": 66405 },406 {407 "epoch": 0.23,408 "learning_rate": 1.2761904761904762e-05,409 "loss": 1.5108,410 "step": 67411 },412 {413 "epoch": 0.23,414 "learning_rate": 1.2952380952380954e-05,415 "loss": 1.3699,416 "step": 68417 },418 {419 "epoch": 0.24,420 "learning_rate": 1.3142857142857145e-05,421 "loss": 1.1858,422 "step": 69423 },424 {425 "epoch": 0.24,426 "learning_rate": 1.3333333333333333e-05,427 "loss": 1.4158,428 "step": 70429 },430 {431 "epoch": 0.24,432 "learning_rate": 1.3523809523809525e-05,433 "loss": 1.1927,434 "step": 71435 },436 {437 "epoch": 0.25,438 "learning_rate": 1.3714285714285716e-05,439 "loss": 1.2916,440 "step": 72441 },442 {443 "epoch": 0.25,444 "learning_rate": 1.3904761904761905e-05,445 "loss": 1.367,446 "step": 73447 },448 {449 "epoch": 0.26,450 "learning_rate": 1.4095238095238097e-05,451 "loss": 1.3799,452 "step": 74453 },454 {455 "epoch": 0.26,456 "learning_rate": 1.4285714285714287e-05,457 "loss": 1.342,458 "step": 75459 },460 {461 "epoch": 0.26,462 "learning_rate": 1.4476190476190478e-05,463 "loss": 1.3428,464 "step": 76465 },466 {467 "epoch": 0.27,468 "learning_rate": 1.4666666666666666e-05,469 "loss": 1.3155,470 "step": 77471 },472 {473 "epoch": 0.27,474 "learning_rate": 1.4857142857142858e-05,475 "loss": 1.3361,476 "step": 78477 },478 {479 "epoch": 0.27,480 "learning_rate": 1.5047619047619049e-05,481 "loss": 1.2827,482 "step": 79483 },484 {485 "epoch": 0.28,486 "learning_rate": 1.523809523809524e-05,487 "loss": 1.3093,488 "step": 80489 },490 {491 "epoch": 0.28,492 "learning_rate": 1.542857142857143e-05,493 "loss": 1.3409,494 "step": 81495 },496 {497 "epoch": 0.28,498 "learning_rate": 1.5619047619047622e-05,499 "loss": 1.3673,500 "step": 82501 },502 {503 "epoch": 0.29,504 "learning_rate": 1.580952380952381e-05,505 "loss": 1.4359,506 "step": 83507 },508 {509 "epoch": 0.29,510 "learning_rate": 1.6000000000000003e-05,511 "loss": 1.2387,512 "step": 84513 },514 {515 "epoch": 0.29,516 "learning_rate": 1.6190476190476193e-05,517 "loss": 1.4272,518 "step": 85519 },520 {521 "epoch": 0.3,522 "learning_rate": 1.6380952380952384e-05,523 "loss": 1.4308,524 "step": 86525 },526 {527 "epoch": 0.3,528 "learning_rate": 1.6571428571428574e-05,529 "loss": 1.4227,530 "step": 87531 },532 {533 "epoch": 0.3,534 "learning_rate": 1.6761904761904764e-05,535 "loss": 1.3882,536 "step": 88537 },538 {539 "epoch": 0.31,540 "learning_rate": 1.6952380952380955e-05,541 "loss": 1.2234,542 "step": 89543 },544 {545 "epoch": 0.31,546 "learning_rate": 1.7142857142857142e-05,547 "loss": 1.3993,548 "step": 90549 },550 {551 "epoch": 0.31,552 "learning_rate": 1.7333333333333336e-05,553 "loss": 1.2739,554 "step": 91555 },556 {557 "epoch": 0.32,558 "learning_rate": 1.7523809523809526e-05,559 "loss": 1.2929,560 "step": 92561 },562 {563 "epoch": 0.32,564 "learning_rate": 1.7714285714285717e-05,565 "loss": 1.3247,566 "step": 93567 },568 {569 "epoch": 0.32,570 "learning_rate": 1.7904761904761907e-05,571 "loss": 1.3627,572 "step": 94573 },574 {575 "epoch": 0.33,576 "learning_rate": 1.8095238095238097e-05,577 "loss": 1.2955,578 "step": 95579 },580 {581 "epoch": 0.33,582 "learning_rate": 1.8285714285714288e-05,583 "loss": 1.3403,584 "step": 96585 },586 {587 "epoch": 0.33,588 "learning_rate": 1.8476190476190478e-05,589 "loss": 1.3244,590 "step": 97591 },592 {593 "epoch": 0.34,594 "learning_rate": 1.866666666666667e-05,595 "loss": 1.2245,596 "step": 98597 },598 {599 "epoch": 0.34,600 "learning_rate": 1.885714285714286e-05,601 "loss": 1.3188,602 "step": 99603 },604 {605 "epoch": 0.34,606 "learning_rate": 1.904761904761905e-05,607 "loss": 1.3082,608 "step": 100609 },610 {611 "epoch": 0.35,612 "learning_rate": 1.923809523809524e-05,613 "loss": 1.2401,614 "step": 101615 },616 {617 "epoch": 0.35,618 "learning_rate": 1.942857142857143e-05,619 "loss": 1.2955,620 "step": 102621 },622 {623 "epoch": 0.36,624 "learning_rate": 1.961904761904762e-05,625 "loss": 1.3467,626 "step": 103627 },628 {629 "epoch": 0.36,630 "learning_rate": 1.980952380952381e-05,631 "loss": 1.2793,632 "step": 104633 },634 {635 "epoch": 0.36,636 "learning_rate": 2e-05,637 "loss": 1.282,638 "step": 105639 },640 {641 "epoch": 0.37,642 "learning_rate": 1.9999995667663676e-05,643 "loss": 1.2427,644 "step": 106645 },646 {647 "epoch": 0.37,648 "learning_rate": 1.9999982670658457e-05,649 "loss": 1.3705,650 "step": 107651 },652 {653 "epoch": 0.37,654 "learning_rate": 1.9999961008995607e-05,655 "loss": 1.2572,656 "step": 108657 },658 {659 "epoch": 0.38,660 "learning_rate": 1.9999930682693892e-05,661 "loss": 1.3193,662 "step": 109663 },664 {665 "epoch": 0.38,666 "learning_rate": 1.999989169177959e-05,667 "loss": 1.3349,668 "step": 110669 },670 {671 "epoch": 0.38,672 "learning_rate": 1.9999844036286483e-05,673 "loss": 1.307,674 "step": 111675 },676 {677 "epoch": 0.39,678 "learning_rate": 1.9999787716255867e-05,679 "loss": 1.3953,680 "step": 112681 },682 {683 "epoch": 0.39,684 "learning_rate": 1.999972273173654e-05,685 "loss": 1.3307,686 "step": 113687 },688 {689 "epoch": 0.39,690 "learning_rate": 1.9999649082784807e-05,691 "loss": 1.3176,692 "step": 114693 },694 {695 "epoch": 0.4,696 "learning_rate": 1.9999566769464483e-05,697 "loss": 1.3142,698 "step": 115699 },700 {701 "epoch": 0.4,702 "learning_rate": 1.9999475791846896e-05,703 "loss": 1.2715,704 "step": 116705 },706 {707 "epoch": 0.4,708 "learning_rate": 1.9999376150010868e-05,709 "loss": 1.3684,710 "step": 117711 },712 {713 "epoch": 0.41,714 "learning_rate": 1.9999267844042734e-05,715 "loss": 1.2952,716 "step": 118717 },718 {719 "epoch": 0.41,720 "learning_rate": 1.9999150874036344e-05,721 "loss": 1.0749,722 "step": 119723 },724 {725 "epoch": 0.41,726 "learning_rate": 1.9999025240093045e-05,727 "loss": 1.1719,728 "step": 120729 },730 {731 "epoch": 0.42,732 "learning_rate": 1.9998890942321693e-05,733 "loss": 1.3597,734 "step": 121735 },736 {737 "epoch": 0.42,738 "learning_rate": 1.9998747980838655e-05,739 "loss": 1.2331,740 "step": 122741 },742 {743 "epoch": 0.42,744 "learning_rate": 1.9998596355767805e-05,745 "loss": 1.1528,746 "step": 123747 },748 {749 "epoch": 0.43,750 "learning_rate": 1.9998436067240518e-05,751 "loss": 1.3303,752 "step": 124753 },754 {755 "epoch": 0.43,756 "learning_rate": 1.999826711539568e-05,757 "loss": 1.3337,758 "step": 125759 },760 {761 "epoch": 0.43,762 "learning_rate": 1.999808950037968e-05,763 "loss": 1.1301,764 "step": 126765 },766 {767 "epoch": 0.44,768 "learning_rate": 1.9997903222346418e-05,769 "loss": 1.4194,770 "step": 127771 },772 {773 "epoch": 0.44,774 "learning_rate": 1.9997708281457293e-05,775 "loss": 1.3396,776 "step": 128777 },778 {779 "epoch": 0.44,780 "learning_rate": 1.9997504677881224e-05,781 "loss": 1.2695,782 "step": 129783 },784 {785 "epoch": 0.45,786 "learning_rate": 1.999729241179462e-05,787 "loss": 1.2832,788 "step": 130789 },790 {791 "epoch": 0.45,792 "learning_rate": 1.99970714833814e-05,793 "loss": 1.2267,794 "step": 131795 },796 {797 "epoch": 0.46,798 "learning_rate": 1.9996841892833e-05,799 "loss": 1.3401,800 "step": 132801 },802 {803 "epoch": 0.46,804 "learning_rate": 1.999660364034835e-05,805 "loss": 1.3008,806 "step": 133807 },808 {809 "epoch": 0.46,810 "learning_rate": 1.999635672613388e-05,811 "loss": 1.3397,812 "step": 134813 },814 {815 "epoch": 0.47,816 "learning_rate": 1.9996101150403543e-05,817 "loss": 1.2069,818 "step": 135819 },820 {821 "epoch": 0.47,822 "learning_rate": 1.9995836913378786e-05,823 "loss": 1.3959,824 "step": 136825 },826 {827 "epoch": 0.47,828 "learning_rate": 1.9995564015288555e-05,829 "loss": 1.3376,830 "step": 137831 },832 {833 "epoch": 0.48,834 "learning_rate": 1.9995282456369313e-05,835 "loss": 1.3011,836 "step": 138837 },838 {839 "epoch": 0.48,840 "learning_rate": 1.9994992236865017e-05,841 "loss": 1.4596,842 "step": 139843 },844 {845 "epoch": 0.48,846 "learning_rate": 1.999469335702714e-05,847 "loss": 1.2458,848 "step": 140849 },850 {851 "epoch": 0.49,852 "learning_rate": 1.9994385817114644e-05,853 "loss": 1.3178,854 "step": 141855 },856 {857 "epoch": 0.49,858 "learning_rate": 1.999406961739401e-05,859 "loss": 1.2166,860 "step": 142861 },862 {863 "epoch": 0.49,864 "learning_rate": 1.9993744758139203e-05,865 "loss": 1.4431,866 "step": 143867 },868 {869 "epoch": 0.5,870 "learning_rate": 1.9993411239631713e-05,871 "loss": 1.2514,872 "step": 144873 },874 {875 "epoch": 0.5,876 "learning_rate": 1.999306906216052e-05,877 "loss": 1.2483,878 "step": 145879 },880 {881 "epoch": 0.5,882 "learning_rate": 1.999271822602211e-05,883 "loss": 1.3231,884 "step": 146885 },886 {887 "epoch": 0.51,888 "learning_rate": 1.999235873152047e-05,889 "loss": 1.2121,890 "step": 147891 },892 {893 "epoch": 0.51,894 "learning_rate": 1.999199057896709e-05,895 "loss": 1.1671,896 "step": 148897 },898 {899 "epoch": 0.51,900 "learning_rate": 1.9991613768680965e-05,901 "loss": 1.2598,902 "step": 149903 },904 {905 "epoch": 0.52,906 "learning_rate": 1.9991228300988586e-05,907 "loss": 1.2223,908 "step": 150909 },910 {911 "epoch": 0.52,912 "learning_rate": 1.999083417622395e-05,913 "loss": 1.3249,914 "step": 151915 },916 {917 "epoch": 0.52,918 "learning_rate": 1.9990431394728547e-05,919 "loss": 1.4458,920 "step": 152921 },922 {923 "epoch": 0.53,924 "learning_rate": 1.9990019956851384e-05,925 "loss": 1.1747,926 "step": 153927 },928 {929 "epoch": 0.53,930 "learning_rate": 1.998959986294895e-05,931 "loss": 1.1985,932 "step": 154933 },934 {935 "epoch": 0.53,936 "learning_rate": 1.998917111338525e-05,937 "loss": 1.358,938 "step": 155939 },940 {941 "epoch": 0.54,942 "learning_rate": 1.9988733708531772e-05,943 "loss": 1.338,944 "step": 156945 },946 {947 "epoch": 0.54,948 "learning_rate": 1.9988287648767523e-05,949 "loss": 1.2886,950 "step": 157951 },952 {953 "epoch": 0.54,954 "learning_rate": 1.9987832934478992e-05,955 "loss": 1.2624,956 "step": 158957 },958 {959 "epoch": 0.55,960 "learning_rate": 1.998736956606018e-05,961 "loss": 1.1608,962 "step": 159963 },964 {965 "epoch": 0.55,966 "learning_rate": 1.998689754391257e-05,967 "loss": 1.3064,968 "step": 160969 },970 {971 "epoch": 0.56,972 "learning_rate": 1.9986416868445167e-05,973 "loss": 1.2455,974 "step": 161975 },976 {977 "epoch": 0.56,978 "learning_rate": 1.9985927540074453e-05,979 "loss": 1.1564,980 "step": 162981 },982 {983 "epoch": 0.56,984 "learning_rate": 1.998542955922442e-05,985 "loss": 1.3674,986 "step": 163987 },988 {989 "epoch": 0.57,990 "learning_rate": 1.9984922926326543e-05,991 "loss": 1.1626,992 "step": 164993 },994 {995 "epoch": 0.57,996 "learning_rate": 1.9984407641819812e-05,997 "loss": 1.2604,998 "step": 165999 },1000 {1001 "epoch": 0.57,1002 "learning_rate": 1.99838837061507e-05,1003 "loss": 1.301,1004 "step": 1661005 },1006 {1007 "epoch": 0.58,1008 "learning_rate": 1.998335111977318e-05,1009 "loss": 1.3197,1010 "step": 1671011 },1012 {1013 "epoch": 0.58,1014 "learning_rate": 1.998280988314872e-05,1015 "loss": 1.2998,1016 "step": 1681017 },1018 {1019 "epoch": 0.58,1020 "learning_rate": 1.998225999674629e-05,1021 "loss": 1.3085,1022 "step": 1691023 },1024 {1025 "epoch": 0.59,1026 "learning_rate": 1.998170146104234e-05,1027 "loss": 1.4856,1028 "step": 1701029 },1030 {1031 "epoch": 0.59,1032 "learning_rate": 1.9981134276520828e-05,1033 "loss": 1.2919,1034 "step": 1711035 },1036 {1037 "epoch": 0.59,1038 "learning_rate": 1.99805584436732e-05,1039 "loss": 1.2869,1040 "step": 1721041 },1042 {1043 "epoch": 0.6,1044 "learning_rate": 1.9979973962998396e-05,1045 "loss": 1.1668,1046 "step": 1731047 },1048 {1049 "epoch": 0.6,1050 "learning_rate": 1.9979380835002846e-05,1051 "loss": 1.2876,1052 "step": 1741053 },1054 {1055 "epoch": 0.6,1056 "learning_rate": 1.9978779060200483e-05,1057 "loss": 1.4127,1058 "step": 1751059 },1060 {1061 "epoch": 0.61,1062 "learning_rate": 1.997816863911272e-05,1063 "loss": 1.1729,1064 "step": 1761065 },1066 {1067 "epoch": 0.61,1068 "learning_rate": 1.997754957226847e-05,1069 "loss": 1.2415,1070 "step": 1771071 },1072 {1073 "epoch": 0.61,1074 "learning_rate": 1.997692186020413e-05,1075 "loss": 1.3111,1076 "step": 1781077 },1078 {1079 "epoch": 0.62,1080 "learning_rate": 1.9976285503463593e-05,1081 "loss": 1.3244,1082 "step": 1791083 },1084 {1085 "epoch": 0.62,1086 "learning_rate": 1.9975640502598243e-05,1087 "loss": 1.1277,1088 "step": 1801089 },1090 {1091 "epoch": 0.62,1092 "learning_rate": 1.9974986858166954e-05,1093 "loss": 1.2408,1094 "step": 1811095 },1096 {1097 "epoch": 0.63,1098 "learning_rate": 1.9974324570736087e-05,1099 "loss": 1.1875,1100 "step": 1821101 },1102 {1103 "epoch": 0.63,1104 "learning_rate": 1.9973653640879486e-05,1105 "loss": 1.2715,1106 "step": 1831107 },1108 {1109 "epoch": 0.63,1110 "learning_rate": 1.9972974069178498e-05,1111 "loss": 1.2486,1112 "step": 1841113 },1114 {1115 "epoch": 0.64,1116 "learning_rate": 1.9972285856221944e-05,1117 "loss": 1.1248,1118 "step": 1851119 },1120 {1121 "epoch": 0.64,1122 "learning_rate": 1.997158900260614e-05,1123 "loss": 1.3108,1124 "step": 1861125 },1126 {1127 "epoch": 0.64,1128 "learning_rate": 1.997088350893489e-05,1129 "loss": 1.3548,1130 "step": 1871131 },1132 {1133 "epoch": 0.65,1134 "learning_rate": 1.9970169375819475e-05,1135 "loss": 1.2499,1136 "step": 1881137 },1138 {1139 "epoch": 0.65,1140 "learning_rate": 1.9969446603878673e-05,1141 "loss": 1.2364,1142 "step": 1891143 },1144 {1145 "epoch": 0.66,1146 "learning_rate": 1.9968715193738738e-05,1147 "loss": 1.0131,1148 "step": 1901149 },1150 {1151 "epoch": 0.66,1152 "learning_rate": 1.9967975146033417e-05,1153 "loss": 1.3815,1154 "step": 1911155 },1156 {1157 "epoch": 0.66,1158 "learning_rate": 1.9967226461403934e-05,1159 "loss": 1.2856,1160 "step": 1921161 },1162 {1163 "epoch": 0.67,1164 "learning_rate": 1.9966469140499e-05,1165 "loss": 1.1579,1166 "step": 1931167 },1168 {1169 "epoch": 0.67,1170 "learning_rate": 1.9965703183974813e-05,1171 "loss": 1.2069,1172 "step": 1941173 },1174 {1175 "epoch": 0.67,1176 "learning_rate": 1.9964928592495046e-05,1177 "loss": 1.2146,1178 "step": 1951179 },1180 {1181 "epoch": 0.68,1182 "learning_rate": 1.9964145366730854e-05,1183 "loss": 1.0996,1184 "step": 1961185 },1186 {1187 "epoch": 0.68,1188 "learning_rate": 1.9963353507360884e-05,1189 "loss": 1.2706,1190 "step": 1971191 },1192 {1193 "epoch": 0.68,1194 "learning_rate": 1.996255301507125e-05,1195 "loss": 1.2786,1196 "step": 1981197 },1198 {1199 "epoch": 0.69,1200 "learning_rate": 1.9961743890555553e-05,