Efficient-Large-Model/VILA1.5-3b
3619k
1{2 "best_metric": null,3 "best_model_checkpoint": null,4 "epoch": 1.0,5 "eval_steps": 500,6 "global_step": 2241,7 "is_hyper_param_search": false,8 "is_local_process_zero": true,9 "is_world_process_zero": true,10 "log_history": [11 {12 "epoch": 0.0,13 "learning_rate": 1.4705882352941177e-06,14 "loss": 2.4788,15 "step": 116 },17 {18 "epoch": 0.0,19 "learning_rate": 2.9411764705882355e-06,20 "loss": 2.1742,21 "step": 222 },23 {24 "epoch": 0.0,25 "learning_rate": 4.411764705882353e-06,26 "loss": 2.3452,27 "step": 328 },29 {30 "epoch": 0.0,31 "learning_rate": 5.882352941176471e-06,32 "loss": 2.4186,33 "step": 434 },35 {36 "epoch": 0.0,37 "learning_rate": 7.3529411764705884e-06,38 "loss": 2.0701,39 "step": 540 },41 {42 "epoch": 0.0,43 "learning_rate": 8.823529411764707e-06,44 "loss": 2.1071,45 "step": 646 },47 {48 "epoch": 0.0,49 "learning_rate": 1.0294117647058824e-05,50 "loss": 1.7251,51 "step": 752 },53 {54 "epoch": 0.0,55 "learning_rate": 1.1764705882352942e-05,56 "loss": 1.7863,57 "step": 858 },59 {60 "epoch": 0.0,61 "learning_rate": 1.323529411764706e-05,62 "loss": 1.8111,63 "step": 964 },65 {66 "epoch": 0.0,67 "learning_rate": 1.4705882352941177e-05,68 "loss": 1.7325,69 "step": 1070 },71 {72 "epoch": 0.0,73 "learning_rate": 1.6176470588235296e-05,74 "loss": 1.7824,75 "step": 1176 },77 {78 "epoch": 0.01,79 "learning_rate": 1.7647058823529414e-05,80 "loss": 1.7706,81 "step": 1282 },83 {84 "epoch": 0.01,85 "learning_rate": 1.9117647058823528e-05,86 "loss": 1.724,87 "step": 1388 },89 {90 "epoch": 0.01,91 "learning_rate": 2.058823529411765e-05,92 "loss": 1.7277,93 "step": 1494 },95 {96 "epoch": 0.01,97 "learning_rate": 2.2058823529411766e-05,98 "loss": 1.7394,99 "step": 15100 },101 {102 "epoch": 0.01,103 "learning_rate": 2.3529411764705884e-05,104 "loss": 1.6897,105 "step": 16106 },107 {108 "epoch": 0.01,109 "learning_rate": 2.5e-05,110 "loss": 1.6762,111 "step": 17112 },113 {114 "epoch": 0.01,115 "learning_rate": 2.647058823529412e-05,116 "loss": 1.6358,117 "step": 18118 },119 {120 "epoch": 0.01,121 "learning_rate": 2.7941176470588236e-05,122 "loss": 1.5994,123 "step": 19124 },125 {126 "epoch": 0.01,127 "learning_rate": 2.9411764705882354e-05,128 "loss": 1.6197,129 "step": 20130 },131 {132 "epoch": 0.01,133 "learning_rate": 3.0882352941176475e-05,134 "loss": 1.5702,135 "step": 21136 },137 {138 "epoch": 0.01,139 "learning_rate": 3.235294117647059e-05,140 "loss": 1.6441,141 "step": 22142 },143 {144 "epoch": 0.01,145 "learning_rate": 3.382352941176471e-05,146 "loss": 1.6177,147 "step": 23148 },149 {150 "epoch": 0.01,151 "learning_rate": 3.529411764705883e-05,152 "loss": 1.493,153 "step": 24154 },155 {156 "epoch": 0.01,157 "learning_rate": 3.6764705882352945e-05,158 "loss": 1.6278,159 "step": 25160 },161 {162 "epoch": 0.01,163 "learning_rate": 3.8235294117647055e-05,164 "loss": 1.583,165 "step": 26166 },167 {168 "epoch": 0.01,169 "learning_rate": 3.970588235294117e-05,170 "loss": 1.5539,171 "step": 27172 },173 {174 "epoch": 0.01,175 "learning_rate": 4.11764705882353e-05,176 "loss": 1.4853,177 "step": 28178 },179 {180 "epoch": 0.01,181 "learning_rate": 4.2647058823529415e-05,182 "loss": 1.5598,183 "step": 29184 },185 {186 "epoch": 0.01,187 "learning_rate": 4.411764705882353e-05,188 "loss": 1.4251,189 "step": 30190 },191 {192 "epoch": 0.01,193 "learning_rate": 4.558823529411765e-05,194 "loss": 1.5344,195 "step": 31196 },197 {198 "epoch": 0.01,199 "learning_rate": 4.705882352941177e-05,200 "loss": 1.5459,201 "step": 32202 },203 {204 "epoch": 0.01,205 "learning_rate": 4.8529411764705885e-05,206 "loss": 1.4635,207 "step": 33208 },209 {210 "epoch": 0.02,211 "learning_rate": 5e-05,212 "loss": 1.4967,213 "step": 34214 },215 {216 "epoch": 0.02,217 "learning_rate": 5.147058823529411e-05,218 "loss": 1.4587,219 "step": 35220 },221 {222 "epoch": 0.02,223 "learning_rate": 5.294117647058824e-05,224 "loss": 1.4996,225 "step": 36226 },227 {228 "epoch": 0.02,229 "learning_rate": 5.441176470588235e-05,230 "loss": 1.57,231 "step": 37232 },233 {234 "epoch": 0.02,235 "learning_rate": 5.588235294117647e-05,236 "loss": 1.5212,237 "step": 38238 },239 {240 "epoch": 0.02,241 "learning_rate": 5.735294117647059e-05,242 "loss": 1.4115,243 "step": 39244 },245 {246 "epoch": 0.02,247 "learning_rate": 5.882352941176471e-05,248 "loss": 1.5373,249 "step": 40250 },251 {252 "epoch": 0.02,253 "learning_rate": 6.0294117647058825e-05,254 "loss": 1.4239,255 "step": 41256 },257 {258 "epoch": 0.02,259 "learning_rate": 6.176470588235295e-05,260 "loss": 1.4767,261 "step": 42262 },263 {264 "epoch": 0.02,265 "learning_rate": 6.323529411764705e-05,266 "loss": 1.5196,267 "step": 43268 },269 {270 "epoch": 0.02,271 "learning_rate": 6.470588235294118e-05,272 "loss": 1.3984,273 "step": 44274 },275 {276 "epoch": 0.02,277 "learning_rate": 6.61764705882353e-05,278 "loss": 1.62,279 "step": 45280 },281 {282 "epoch": 0.02,283 "learning_rate": 6.764705882352942e-05,284 "loss": 1.3954,285 "step": 46286 },287 {288 "epoch": 0.02,289 "learning_rate": 6.911764705882354e-05,290 "loss": 1.4746,291 "step": 47292 },293 {294 "epoch": 0.02,295 "learning_rate": 7.058823529411765e-05,296 "loss": 1.4916,297 "step": 48298 },299 {300 "epoch": 0.02,301 "learning_rate": 7.205882352941177e-05,302 "loss": 1.4281,303 "step": 49304 },305 {306 "epoch": 0.02,307 "learning_rate": 7.352941176470589e-05,308 "loss": 1.4423,309 "step": 50310 },311 {312 "epoch": 0.02,313 "learning_rate": 7.500000000000001e-05,314 "loss": 1.4676,315 "step": 51316 },317 {318 "epoch": 0.02,319 "learning_rate": 7.647058823529411e-05,320 "loss": 1.4503,321 "step": 52322 },323 {324 "epoch": 0.02,325 "learning_rate": 7.794117647058824e-05,326 "loss": 1.4357,327 "step": 53328 },329 {330 "epoch": 0.02,331 "learning_rate": 7.941176470588235e-05,332 "loss": 1.4552,333 "step": 54334 },335 {336 "epoch": 0.02,337 "learning_rate": 8.088235294117648e-05,338 "loss": 1.3965,339 "step": 55340 },341 {342 "epoch": 0.02,343 "learning_rate": 8.23529411764706e-05,344 "loss": 1.4523,345 "step": 56346 },347 {348 "epoch": 0.03,349 "learning_rate": 8.382352941176471e-05,350 "loss": 1.5022,351 "step": 57352 },353 {354 "epoch": 0.03,355 "learning_rate": 8.529411764705883e-05,356 "loss": 1.3563,357 "step": 58358 },359 {360 "epoch": 0.03,361 "learning_rate": 8.676470588235295e-05,362 "loss": 1.4362,363 "step": 59364 },365 {366 "epoch": 0.03,367 "learning_rate": 8.823529411764706e-05,368 "loss": 1.4223,369 "step": 60370 },371 {372 "epoch": 0.03,373 "learning_rate": 8.970588235294118e-05,374 "loss": 1.3593,375 "step": 61376 },377 {378 "epoch": 0.03,379 "learning_rate": 9.11764705882353e-05,380 "loss": 1.4579,381 "step": 62382 },383 {384 "epoch": 0.03,385 "learning_rate": 9.264705882352942e-05,386 "loss": 1.4635,387 "step": 63388 },389 {390 "epoch": 0.03,391 "learning_rate": 9.411764705882353e-05,392 "loss": 1.3896,393 "step": 64394 },395 {396 "epoch": 0.03,397 "learning_rate": 9.558823529411765e-05,398 "loss": 1.4864,399 "step": 65400 },401 {402 "epoch": 0.03,403 "learning_rate": 9.705882352941177e-05,404 "loss": 1.3461,405 "step": 66406 },407 {408 "epoch": 0.03,409 "learning_rate": 9.852941176470589e-05,410 "loss": 1.4454,411 "step": 67412 },413 {414 "epoch": 0.03,415 "learning_rate": 0.0001,416 "loss": 1.4292,417 "step": 68418 },419 {420 "epoch": 0.03,421 "learning_rate": 9.999994774591761e-05,422 "loss": 1.3922,423 "step": 69424 },425 {426 "epoch": 0.03,427 "learning_rate": 9.999979098377963e-05,428 "loss": 1.4472,429 "step": 70430 },431 {432 "epoch": 0.03,433 "learning_rate": 9.999952971391373e-05,434 "loss": 1.4291,435 "step": 71436 },437 {438 "epoch": 0.03,439 "learning_rate": 9.9999163936866e-05,440 "loss": 1.438,441 "step": 72442 },443 {444 "epoch": 0.03,445 "learning_rate": 9.9998693653401e-05,446 "loss": 1.372,447 "step": 73448 },449 {450 "epoch": 0.03,451 "learning_rate": 9.999811886450166e-05,452 "loss": 1.4195,453 "step": 74454 },455 {456 "epoch": 0.03,457 "learning_rate": 9.99974395713694e-05,458 "loss": 1.3846,459 "step": 75460 },461 {462 "epoch": 0.03,463 "learning_rate": 9.999665577542406e-05,464 "loss": 1.4186,465 "step": 76466 },467 {468 "epoch": 0.03,469 "learning_rate": 9.99957674783039e-05,470 "loss": 1.4086,471 "step": 77472 },473 {474 "epoch": 0.03,475 "learning_rate": 9.99947746818656e-05,476 "loss": 1.3691,477 "step": 78478 },479 {480 "epoch": 0.04,481 "learning_rate": 9.999367738818428e-05,482 "loss": 1.4126,483 "step": 79484 },485 {486 "epoch": 0.04,487 "learning_rate": 9.999247559955345e-05,488 "loss": 1.3972,489 "step": 80490 },491 {492 "epoch": 0.04,493 "learning_rate": 9.999116931848504e-05,494 "loss": 1.3279,495 "step": 81496 },497 {498 "epoch": 0.04,499 "learning_rate": 9.99897585477094e-05,500 "loss": 1.4977,501 "step": 82502 },503 {504 "epoch": 0.04,505 "learning_rate": 9.998824329017527e-05,506 "loss": 1.4041,507 "step": 83508 },509 {510 "epoch": 0.04,511 "learning_rate": 9.998662354904977e-05,512 "loss": 1.3215,513 "step": 84514 },515 {516 "epoch": 0.04,517 "learning_rate": 9.998489932771845e-05,518 "loss": 1.4292,519 "step": 85520 },521 {522 "epoch": 0.04,523 "learning_rate": 9.99830706297852e-05,524 "loss": 1.3729,525 "step": 86526 },527 {528 "epoch": 0.04,529 "learning_rate": 9.99811374590723e-05,530 "loss": 1.3613,531 "step": 87532 },533 {534 "epoch": 0.04,535 "learning_rate": 9.99790998196204e-05,536 "loss": 1.3718,537 "step": 88538 },539 {540 "epoch": 0.04,541 "learning_rate": 9.997695771568849e-05,542 "loss": 1.3346,543 "step": 89544 },545 {546 "epoch": 0.04,547 "learning_rate": 9.99747111517539e-05,548 "loss": 1.366,549 "step": 90550 },551 {552 "epoch": 0.04,553 "learning_rate": 9.997236013251234e-05,554 "loss": 1.4396,555 "step": 91556 },557 {558 "epoch": 0.04,559 "learning_rate": 9.996990466287783e-05,560 "loss": 1.3603,561 "step": 92562 },563 {564 "epoch": 0.04,565 "learning_rate": 9.996734474798268e-05,566 "loss": 1.4104,567 "step": 93568 },569 {570 "epoch": 0.04,571 "learning_rate": 9.996468039317754e-05,572 "loss": 1.4893,573 "step": 94574 },575 {576 "epoch": 0.04,577 "learning_rate": 9.996191160403137e-05,578 "loss": 1.3518,579 "step": 95580 },581 {582 "epoch": 0.04,583 "learning_rate": 9.995903838633134e-05,584 "loss": 1.3501,585 "step": 96586 },587 {588 "epoch": 0.04,589 "learning_rate": 9.995606074608299e-05,590 "loss": 1.4145,591 "step": 97592 },593 {594 "epoch": 0.04,595 "learning_rate": 9.995297868951006e-05,596 "loss": 1.3133,597 "step": 98598 },599 {600 "epoch": 0.04,601 "learning_rate": 9.994979222305453e-05,602 "loss": 1.4353,603 "step": 99604 },605 {606 "epoch": 0.04,607 "learning_rate": 9.994650135337667e-05,608 "loss": 1.3907,609 "step": 100610 },611 {612 "epoch": 0.05,613 "learning_rate": 9.994310608735491e-05,614 "loss": 1.2955,615 "step": 101616 },617 {618 "epoch": 0.05,619 "learning_rate": 9.993960643208591e-05,620 "loss": 1.5013,621 "step": 102622 },623 {624 "epoch": 0.05,625 "learning_rate": 9.993600239488454e-05,626 "loss": 1.394,627 "step": 103628 },629 {630 "epoch": 0.05,631 "learning_rate": 9.993229398328382e-05,632 "loss": 1.3382,633 "step": 104634 },635 {636 "epoch": 0.05,637 "learning_rate": 9.992848120503493e-05,638 "loss": 1.4251,639 "step": 105640 },641 {642 "epoch": 0.05,643 "learning_rate": 9.992456406810719e-05,644 "loss": 1.3131,645 "step": 106646 },647 {648 "epoch": 0.05,649 "learning_rate": 9.99205425806881e-05,650 "loss": 1.4099,651 "step": 107652 },653 {654 "epoch": 0.05,655 "learning_rate": 9.991641675118316e-05,656 "loss": 1.4179,657 "step": 108658 },659 {660 "epoch": 0.05,661 "learning_rate": 9.991218658821608e-05,662 "loss": 1.2979,663 "step": 109664 },665 {666 "epoch": 0.05,667 "learning_rate": 9.990785210062856e-05,668 "loss": 1.4283,669 "step": 110670 },671 {672 "epoch": 0.05,673 "learning_rate": 9.99034132974804e-05,674 "loss": 1.4327,675 "step": 111676 },677 {678 "epoch": 0.05,679 "learning_rate": 9.989887018804942e-05,680 "loss": 1.2713,681 "step": 112682 },683 {684 "epoch": 0.05,685 "learning_rate": 9.989422278183146e-05,686 "loss": 1.3652,687 "step": 113688 },689 {690 "epoch": 0.05,691 "learning_rate": 9.988947108854037e-05,692 "loss": 1.4147,693 "step": 114694 },695 {696 "epoch": 0.05,697 "learning_rate": 9.988461511810796e-05,698 "loss": 1.3238,699 "step": 115700 },701 {702 "epoch": 0.05,703 "learning_rate": 9.987965488068398e-05,704 "loss": 1.3624,705 "step": 116706 },707 {708 "epoch": 0.05,709 "learning_rate": 9.987459038663616e-05,710 "loss": 1.3728,711 "step": 117712 },713 {714 "epoch": 0.05,715 "learning_rate": 9.986942164655012e-05,716 "loss": 1.3642,717 "step": 118718 },719 {720 "epoch": 0.05,721 "learning_rate": 9.986414867122936e-05,722 "loss": 1.339,723 "step": 119724 },725 {726 "epoch": 0.05,727 "learning_rate": 9.985877147169525e-05,728 "loss": 1.3893,729 "step": 120730 },731 {732 "epoch": 0.05,733 "learning_rate": 9.985329005918702e-05,734 "loss": 1.3028,735 "step": 121736 },737 {738 "epoch": 0.05,739 "learning_rate": 9.984770444516174e-05,740 "loss": 1.4613,741 "step": 122742 },743 {744 "epoch": 0.05,745 "learning_rate": 9.984201464129425e-05,746 "loss": 1.3835,747 "step": 123748 },749 {750 "epoch": 0.06,751 "learning_rate": 9.983622065947714e-05,752 "loss": 1.3367,753 "step": 124754 },755 {756 "epoch": 0.06,757 "learning_rate": 9.983032251182081e-05,758 "loss": 1.3554,759 "step": 125760 },761 {762 "epoch": 0.06,763 "learning_rate": 9.982432021065333e-05,764 "loss": 1.3255,765 "step": 126766 },767 {768 "epoch": 0.06,769 "learning_rate": 9.98182137685205e-05,770 "loss": 1.3448,771 "step": 127772 },773 {774 "epoch": 0.06,775 "learning_rate": 9.98120031981858e-05,776 "loss": 1.3574,777 "step": 128778 },779 {780 "epoch": 0.06,781 "learning_rate": 9.980568851263029e-05,782 "loss": 1.3003,783 "step": 129784 },785 {786 "epoch": 0.06,787 "learning_rate": 9.979926972505273e-05,788 "loss": 1.4174,789 "step": 130790 },791 {792 "epoch": 0.06,793 "learning_rate": 9.979274684886943e-05,794 "loss": 1.4257,795 "step": 131796 },797 {798 "epoch": 0.06,799 "learning_rate": 9.978611989771424e-05,800 "loss": 1.2841,801 "step": 132802 },803 {804 "epoch": 0.06,805 "learning_rate": 9.977938888543862e-05,806 "loss": 1.318,807 "step": 133808 },809 {810 "epoch": 0.06,811 "learning_rate": 9.977255382611144e-05,812 "loss": 1.3566,813 "step": 134814 },815 {816 "epoch": 0.06,817 "learning_rate": 9.97656147340191e-05,818 "loss": 1.367,819 "step": 135820 },821 {822 "epoch": 0.06,823 "learning_rate": 9.975857162366546e-05,824 "loss": 1.37,825 "step": 136826 },827 {828 "epoch": 0.06,829 "learning_rate": 9.975142450977174e-05,830 "loss": 1.3842,831 "step": 137832 },833 {834 "epoch": 0.06,835 "learning_rate": 9.974417340727659e-05,836 "loss": 1.3769,837 "step": 138838 },839 {840 "epoch": 0.06,841 "learning_rate": 9.973681833133598e-05,842 "loss": 1.4379,843 "step": 139844 },845 {846 "epoch": 0.06,847 "learning_rate": 9.972935929732325e-05,848 "loss": 1.299,849 "step": 140850 },851 {852 "epoch": 0.06,853 "learning_rate": 9.972179632082898e-05,854 "loss": 1.4021,855 "step": 141856 },857 {858 "epoch": 0.06,859 "learning_rate": 9.971412941766105e-05,860 "loss": 1.3592,861 "step": 142862 },863 {864 "epoch": 0.06,865 "learning_rate": 9.970635860384449e-05,866 "loss": 1.3067,867 "step": 143868 },869 {870 "epoch": 0.06,871 "learning_rate": 9.969848389562162e-05,872 "loss": 1.3637,873 "step": 144874 },875 {876 "epoch": 0.06,877 "learning_rate": 9.969050530945184e-05,878 "loss": 1.3433,879 "step": 145880 },881 {882 "epoch": 0.07,883 "learning_rate": 9.968242286201171e-05,884 "loss": 1.2948,885 "step": 146886 },887 {888 "epoch": 0.07,889 "learning_rate": 9.967423657019485e-05,890 "loss": 1.4001,891 "step": 147892 },893 {894 "epoch": 0.07,895 "learning_rate": 9.966594645111196e-05,896 "loss": 1.3833,897 "step": 148898 },899 {900 "epoch": 0.07,901 "learning_rate": 9.965755252209073e-05,902 "loss": 1.3291,903 "step": 149904 },905 {906 "epoch": 0.07,907 "learning_rate": 9.964905480067586e-05,908 "loss": 1.446,909 "step": 150910 },911 {912 "epoch": 0.07,913 "learning_rate": 9.964045330462896e-05,914 "loss": 1.3754,915 "step": 151916 },917 {918 "epoch": 0.07,919 "learning_rate": 9.963174805192857e-05,920 "loss": 1.2877,921 "step": 152922 },923 {924 "epoch": 0.07,925 "learning_rate": 9.962293906077006e-05,926 "loss": 1.3607,927 "step": 153928 },929 {930 "epoch": 0.07,931 "learning_rate": 9.961402634956573e-05,932 "loss": 1.2588,933 "step": 154934 },935 {936 "epoch": 0.07,937 "learning_rate": 9.960500993694454e-05,938 "loss": 1.3904,939 "step": 155940 },941 {942 "epoch": 0.07,943 "learning_rate": 9.959588984175228e-05,944 "loss": 1.3443,945 "step": 156946 },947 {948 "epoch": 0.07,949 "learning_rate": 9.958666608305145e-05,950 "loss": 1.3054,951 "step": 157952 },953 {954 "epoch": 0.07,955 "learning_rate": 9.95773386801212e-05,956 "loss": 1.3815,957 "step": 158958 },959 {960 "epoch": 0.07,961 "learning_rate": 9.956790765245732e-05,962 "loss": 1.3279,963 "step": 159964 },965 {966 "epoch": 0.07,967 "learning_rate": 9.955837301977223e-05,968 "loss": 1.2615,969 "step": 160970 },971 {972 "epoch": 0.07,973 "learning_rate": 9.954873480199482e-05,974 "loss": 1.3867,975 "step": 161976 },977 {978 "epoch": 0.07,979 "learning_rate": 9.953899301927057e-05,980 "loss": 1.2997,981 "step": 162982 },983 {984 "epoch": 0.07,985 "learning_rate": 9.95291476919614e-05,986 "loss": 1.3289,987 "step": 163988 },989 {990 "epoch": 0.07,991 "learning_rate": 9.951919884064564e-05,992 "loss": 1.3931,993 "step": 164994 },995 {996 "epoch": 0.07,997 "learning_rate": 9.950914648611802e-05,998 "loss": 1.3387,999 "step": 1651000 },1001 {1002 "epoch": 0.07,1003 "learning_rate": 9.949899064938959e-05,1004 "loss": 1.2911,1005 "step": 1661006 },1007 {1008 "epoch": 0.07,1009 "learning_rate": 9.948873135168773e-05,1010 "loss": 1.3589,1011 "step": 1671012 },1013 {1014 "epoch": 0.07,1015 "learning_rate": 9.947836861445603e-05,1016 "loss": 1.3822,1017 "step": 1681018 },1019 {1020 "epoch": 0.08,1021 "learning_rate": 9.946790245935429e-05,1022 "loss": 1.2447,1023 "step": 1691024 },1025 {1026 "epoch": 0.08,1027 "learning_rate": 9.945733290825852e-05,1028 "loss": 1.3806,1029 "step": 1701030 },1031 {1032 "epoch": 0.08,1033 "learning_rate": 9.944665998326079e-05,1034 "loss": 1.2989,1035 "step": 1711036 },1037 {1038 "epoch": 0.08,1039 "learning_rate": 9.943588370666926e-05,1040 "loss": 1.3644,1041 "step": 1721042 },1043 {1044 "epoch": 0.08,1045 "learning_rate": 9.942500410100808e-05,1046 "loss": 1.3536,1047 "step": 1731048 },1049 {1050 "epoch": 0.08,1051 "learning_rate": 9.941402118901744e-05,1052 "loss": 1.301,1053 "step": 1741054 },1055 {1056 "epoch": 0.08,1057 "learning_rate": 9.940293499365339e-05,1058 "loss": 1.3621,1059 "step": 1751060 },1061 {1062 "epoch": 0.08,1063 "learning_rate": 9.939174553808793e-05,1064 "loss": 1.3381,1065 "step": 1761066 },1067 {1068 "epoch": 0.08,1069 "learning_rate": 9.938045284570878e-05,1070 "loss": 1.3471,1071 "step": 1771072 },1073 {1074 "epoch": 0.08,1075 "learning_rate": 9.936905694011959e-05,1076 "loss": 1.2256,1077 "step": 1781078 },1079 {1080 "epoch": 0.08,1081 "learning_rate": 9.93575578451396e-05,1082 "loss": 1.3558,1083 "step": 1791084 },1085 {1086 "epoch": 0.08,1087 "learning_rate": 9.934595558480384e-05,1088 "loss": 1.3343,1089 "step": 1801090 },1091 {1092 "epoch": 0.08,1093 "learning_rate": 9.933425018336293e-05,1094 "loss": 1.3473,1095 "step": 1811096 },1097 {1098 "epoch": 0.08,1099 "learning_rate": 9.932244166528302e-05,1100 "loss": 1.3312,1101 "step": 1821102 },1103 {1104 "epoch": 0.08,1105 "learning_rate": 9.93105300552459e-05,1106 "loss": 1.286,1107 "step": 1831108 },1109 {1110 "epoch": 0.08,1111 "learning_rate": 9.929851537814871e-05,1112 "loss": 1.3196,1113 "step": 1841114 },1115 {1116 "epoch": 0.08,1117 "learning_rate": 9.928639765910417e-05,1118 "loss": 1.2862,1119 "step": 1851120 },1121 {1122 "epoch": 0.08,1123 "learning_rate": 9.927417692344024e-05,1124 "loss": 1.3695,1125 "step": 1861126 },1127 {1128 "epoch": 0.08,1129 "learning_rate": 9.926185319670027e-05,1130 "loss": 1.3424,1131 "step": 1871132 },1133 {1134 "epoch": 0.08,1135 "learning_rate": 9.924942650464287e-05,1136 "loss": 1.3259,1137 "step": 1881138 },1139 {1140 "epoch": 0.08,1141 "learning_rate": 9.923689687324181e-05,1142 "loss": 1.3039,1143 "step": 1891144 },1145 {1146 "epoch": 0.08,1147 "learning_rate": 9.92242643286861e-05,1148 "loss": 1.3714,1149 "step": 1901150 },1151 {1152 "epoch": 0.09,1153 "learning_rate": 9.921152889737984e-05,1154 "loss": 1.3822,1155 "step": 1911156 },1157 {1158 "epoch": 0.09,1159 "learning_rate": 9.919869060594214e-05,1160 "loss": 1.2961,1161 "step": 1921162 },1163 {1164 "epoch": 0.09,1165 "learning_rate": 9.918574948120711e-05,1166 "loss": 1.3482,1167 "step": 1931168 },1169 {1170 "epoch": 0.09,1171 "learning_rate": 9.917270555022384e-05,1172 "loss": 1.2113,1173 "step": 1941174 },1175 {1176 "epoch": 0.09,1177 "learning_rate": 9.915955884025626e-05,1178 "loss": 1.3763,1179 "step": 1951180 },1181 {1182 "epoch": 0.09,1183 "learning_rate": 9.914630937878315e-05,1184 "loss": 1.342,1185 "step": 1961186 },1187 {1188 "epoch": 0.09,1189 "learning_rate": 9.913295719349805e-05,1190 "loss": 1.3451,1191 "step": 1971192 },1193 {1194 "epoch": 0.09,1195 "learning_rate": 9.911950231230919e-05,1196 "loss": 1.3275,1197 "step": 1981198 },1199 {1200 "epoch": 0.09,