cllm/deepseekcoder_7b_codesearch_net_python
134
1{2 "best_metric": null,3 "best_model_checkpoint": null,4 "epoch": 2.9943685455195057,5 "eval_steps": 500,6 "global_step": 44000,7 "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": 2.0000000000000003e-06,14 "loss": 0.9909,15 "step": 116 },17 {18 "epoch": 0.0,19 "learning_rate": 4.000000000000001e-06,20 "loss": 0.9824,21 "step": 222 },23 {24 "epoch": 0.0,25 "learning_rate": 6e-06,26 "loss": 0.9979,27 "step": 328 },29 {30 "epoch": 0.0,31 "learning_rate": 8.000000000000001e-06,32 "loss": 0.9784,33 "step": 434 },35 {36 "epoch": 0.0,37 "learning_rate": 1e-05,38 "loss": 1.0243,39 "step": 540 },41 {42 "epoch": 0.0,43 "learning_rate": 1.2e-05,44 "loss": 0.7993,45 "step": 646 },47 {48 "epoch": 0.0,49 "learning_rate": 1.4e-05,50 "loss": 0.8254,51 "step": 752 },53 {54 "epoch": 0.0,55 "learning_rate": 1.6000000000000003e-05,56 "loss": 0.8486,57 "step": 858 },59 {60 "epoch": 0.0,61 "learning_rate": 1.8e-05,62 "loss": 0.8454,63 "step": 964 },65 {66 "epoch": 0.0,67 "learning_rate": 2e-05,68 "loss": 0.6792,69 "step": 1070 },71 {72 "epoch": 0.0,73 "learning_rate": 1.9999999974593536e-05,74 "loss": 0.7772,75 "step": 1176 },77 {78 "epoch": 0.0,79 "learning_rate": 1.9999999898374144e-05,80 "loss": 0.7329,81 "step": 1282 },83 {84 "epoch": 0.0,85 "learning_rate": 1.9999999771341824e-05,86 "loss": 0.6691,87 "step": 1388 },89 {90 "epoch": 0.0,91 "learning_rate": 1.9999999593496574e-05,92 "loss": 0.668,93 "step": 1494 },95 {96 "epoch": 0.0,97 "learning_rate": 1.99999993648384e-05,98 "loss": 0.6422,99 "step": 15100 },101 {102 "epoch": 0.0,103 "learning_rate": 1.99999990853673e-05,104 "loss": 0.6852,105 "step": 16106 },107 {108 "epoch": 0.0,109 "learning_rate": 1.9999998755083275e-05,110 "loss": 0.6932,111 "step": 17112 },113 {114 "epoch": 0.0,115 "learning_rate": 1.999999837398633e-05,116 "loss": 0.6604,117 "step": 18118 },119 {120 "epoch": 0.0,121 "learning_rate": 1.999999794207646e-05,122 "loss": 0.6675,123 "step": 19124 },125 {126 "epoch": 0.0,127 "learning_rate": 1.999999745935368e-05,128 "loss": 0.6175,129 "step": 20130 },131 {132 "epoch": 0.0,133 "learning_rate": 1.9999996925817976e-05,134 "loss": 0.6721,135 "step": 21136 },137 {138 "epoch": 0.0,139 "learning_rate": 1.999999634146936e-05,140 "loss": 0.7605,141 "step": 22142 },143 {144 "epoch": 0.0,145 "learning_rate": 1.9999995706307837e-05,146 "loss": 0.7594,147 "step": 23148 },149 {150 "epoch": 0.0,151 "learning_rate": 1.9999995020333407e-05,152 "loss": 0.6646,153 "step": 24154 },155 {156 "epoch": 0.0,157 "learning_rate": 1.9999994283546073e-05,158 "loss": 0.6854,159 "step": 25160 },161 {162 "epoch": 0.0,163 "learning_rate": 1.999999349594584e-05,164 "loss": 0.5974,165 "step": 26166 },167 {168 "epoch": 0.0,169 "learning_rate": 1.999999265753271e-05,170 "loss": 0.5894,171 "step": 27172 },173 {174 "epoch": 0.0,175 "learning_rate": 1.9999991768306687e-05,176 "loss": 0.6329,177 "step": 28178 },179 {180 "epoch": 0.0,181 "learning_rate": 1.999999082826778e-05,182 "loss": 0.6758,183 "step": 29184 },185 {186 "epoch": 0.0,187 "learning_rate": 1.999998983741599e-05,188 "loss": 0.6067,189 "step": 30190 },191 {192 "epoch": 0.0,193 "learning_rate": 1.9999988795751325e-05,194 "loss": 0.6483,195 "step": 31196 },197 {198 "epoch": 0.0,199 "learning_rate": 1.9999987703273786e-05,200 "loss": 0.6174,201 "step": 32202 },203 {204 "epoch": 0.0,205 "learning_rate": 1.999998655998338e-05,206 "loss": 0.6921,207 "step": 33208 },209 {210 "epoch": 0.0,211 "learning_rate": 1.9999985365880117e-05,212 "loss": 0.628,213 "step": 34214 },215 {216 "epoch": 0.0,217 "learning_rate": 1.9999984120964e-05,218 "loss": 0.5899,219 "step": 35220 },221 {222 "epoch": 0.0,223 "learning_rate": 1.999998282523503e-05,224 "loss": 0.621,225 "step": 36226 },227 {228 "epoch": 0.0,229 "learning_rate": 1.9999981478693226e-05,230 "loss": 0.7353,231 "step": 37232 },233 {234 "epoch": 0.0,235 "learning_rate": 1.999998008133858e-05,236 "loss": 0.6697,237 "step": 38238 },239 {240 "epoch": 0.0,241 "learning_rate": 1.999997863317111e-05,242 "loss": 0.6902,243 "step": 39244 },245 {246 "epoch": 0.0,247 "learning_rate": 1.999997713419082e-05,248 "loss": 0.6301,249 "step": 40250 },251 {252 "epoch": 0.0,253 "learning_rate": 1.9999975584397717e-05,254 "loss": 0.6609,255 "step": 41256 },257 {258 "epoch": 0.0,259 "learning_rate": 1.9999973983791808e-05,260 "loss": 0.6304,261 "step": 42262 },263 {264 "epoch": 0.0,265 "learning_rate": 1.9999972332373104e-05,266 "loss": 0.6619,267 "step": 43268 },269 {270 "epoch": 0.0,271 "learning_rate": 1.9999970630141616e-05,272 "loss": 0.7103,273 "step": 44274 },275 {276 "epoch": 0.0,277 "learning_rate": 1.9999968877097342e-05,278 "loss": 0.6727,279 "step": 45280 },281 {282 "epoch": 0.0,283 "learning_rate": 1.99999670732403e-05,284 "loss": 0.6565,285 "step": 46286 },287 {288 "epoch": 0.0,289 "learning_rate": 1.99999652185705e-05,290 "loss": 0.6822,291 "step": 47292 },293 {294 "epoch": 0.0,295 "learning_rate": 1.999996331308794e-05,296 "loss": 0.6721,297 "step": 48298 },299 {300 "epoch": 0.0,301 "learning_rate": 1.9999961356792647e-05,302 "loss": 0.5848,303 "step": 49304 },305 {306 "epoch": 0.0,307 "learning_rate": 1.9999959349684616e-05,308 "loss": 0.6848,309 "step": 50310 },311 {312 "epoch": 0.0,313 "learning_rate": 1.9999957291763864e-05,314 "loss": 0.6204,315 "step": 51316 },317 {318 "epoch": 0.0,319 "learning_rate": 1.99999551830304e-05,320 "loss": 0.6667,321 "step": 52322 },323 {324 "epoch": 0.0,325 "learning_rate": 1.999995302348424e-05,326 "loss": 0.7393,327 "step": 53328 },329 {330 "epoch": 0.0,331 "learning_rate": 1.9999950813125383e-05,332 "loss": 0.6167,333 "step": 54334 },335 {336 "epoch": 0.0,337 "learning_rate": 1.999994855195385e-05,338 "loss": 0.6454,339 "step": 55340 },341 {342 "epoch": 0.0,343 "learning_rate": 1.999994623996965e-05,344 "loss": 0.6647,345 "step": 56346 },347 {348 "epoch": 0.0,349 "learning_rate": 1.9999943877172798e-05,350 "loss": 0.6818,351 "step": 57352 },353 {354 "epoch": 0.0,355 "learning_rate": 1.9999941463563297e-05,356 "loss": 0.66,357 "step": 58358 },359 {360 "epoch": 0.0,361 "learning_rate": 1.9999938999141168e-05,362 "loss": 0.6213,363 "step": 59364 },365 {366 "epoch": 0.0,367 "learning_rate": 1.9999936483906416e-05,368 "loss": 0.6715,369 "step": 60370 },371 {372 "epoch": 0.0,373 "learning_rate": 1.9999933917859066e-05,374 "loss": 0.7579,375 "step": 61376 },377 {378 "epoch": 0.0,379 "learning_rate": 1.9999931300999117e-05,380 "loss": 0.6687,381 "step": 62382 },383 {384 "epoch": 0.0,385 "learning_rate": 1.999992863332659e-05,386 "loss": 0.635,387 "step": 63388 },389 {390 "epoch": 0.0,391 "learning_rate": 1.9999925914841495e-05,392 "loss": 0.6594,393 "step": 64394 },395 {396 "epoch": 0.0,397 "learning_rate": 1.9999923145543853e-05,398 "loss": 0.6305,399 "step": 65400 },401 {402 "epoch": 0.0,403 "learning_rate": 1.999992032543367e-05,404 "loss": 0.6036,405 "step": 66406 },407 {408 "epoch": 0.0,409 "learning_rate": 1.9999917454510962e-05,410 "loss": 0.6631,411 "step": 67412 },413 {414 "epoch": 0.0,415 "learning_rate": 1.9999914532775744e-05,416 "loss": 0.6624,417 "step": 68418 },419 {420 "epoch": 0.0,421 "learning_rate": 1.9999911560228033e-05,422 "loss": 0.6977,423 "step": 69424 },425 {426 "epoch": 0.0,427 "learning_rate": 1.9999908536867846e-05,428 "loss": 0.5775,429 "step": 70430 },431 {432 "epoch": 0.0,433 "learning_rate": 1.999990546269519e-05,434 "loss": 0.6821,435 "step": 71436 },437 {438 "epoch": 0.0,439 "learning_rate": 1.9999902337710088e-05,440 "loss": 0.6577,441 "step": 72442 },443 {444 "epoch": 0.0,445 "learning_rate": 1.9999899161912552e-05,446 "loss": 0.6624,447 "step": 73448 },449 {450 "epoch": 0.01,451 "learning_rate": 1.99998959353026e-05,452 "loss": 0.6595,453 "step": 74454 },455 {456 "epoch": 0.01,457 "learning_rate": 1.9999892657880254e-05,458 "loss": 0.6295,459 "step": 75460 },461 {462 "epoch": 0.01,463 "learning_rate": 1.999988932964552e-05,464 "loss": 0.6837,465 "step": 76466 },467 {468 "epoch": 0.01,469 "learning_rate": 1.999988595059842e-05,470 "loss": 0.6006,471 "step": 77472 },473 {474 "epoch": 0.01,475 "learning_rate": 1.999988252073897e-05,476 "loss": 0.6364,477 "step": 78478 },479 {480 "epoch": 0.01,481 "learning_rate": 1.999987904006719e-05,482 "loss": 0.6456,483 "step": 79484 },485 {486 "epoch": 0.01,487 "learning_rate": 1.9999875508583094e-05,488 "loss": 0.7056,489 "step": 80490 },491 {492 "epoch": 0.01,493 "learning_rate": 1.9999871926286702e-05,494 "loss": 0.6716,495 "step": 81496 },497 {498 "epoch": 0.01,499 "learning_rate": 1.9999868293178033e-05,500 "loss": 0.6725,501 "step": 82502 },503 {504 "epoch": 0.01,505 "learning_rate": 1.9999864609257104e-05,506 "loss": 0.7592,507 "step": 83508 },509 {510 "epoch": 0.01,511 "learning_rate": 1.999986087452394e-05,512 "loss": 0.6453,513 "step": 84514 },515 {516 "epoch": 0.01,517 "learning_rate": 1.9999857088978548e-05,518 "loss": 0.6452,519 "step": 85520 },521 {522 "epoch": 0.01,523 "learning_rate": 1.9999853252620952e-05,524 "loss": 0.6337,525 "step": 86526 },527 {528 "epoch": 0.01,529 "learning_rate": 1.9999849365451178e-05,530 "loss": 0.6689,531 "step": 87532 },533 {534 "epoch": 0.01,535 "learning_rate": 1.999984542746924e-05,536 "loss": 0.5882,537 "step": 88538 },539 {540 "epoch": 0.01,541 "learning_rate": 1.9999841438675155e-05,542 "loss": 0.6507,543 "step": 89544 },545 {546 "epoch": 0.01,547 "learning_rate": 1.999983739906895e-05,548 "loss": 0.6544,549 "step": 90550 },551 {552 "epoch": 0.01,553 "learning_rate": 1.9999833308650642e-05,554 "loss": 0.6458,555 "step": 91556 },557 {558 "epoch": 0.01,559 "learning_rate": 1.9999829167420254e-05,560 "loss": 0.6677,561 "step": 92562 },563 {564 "epoch": 0.01,565 "learning_rate": 1.9999824975377805e-05,566 "loss": 0.6789,567 "step": 93568 },569 {570 "epoch": 0.01,571 "learning_rate": 1.9999820732523318e-05,572 "loss": 0.6714,573 "step": 94574 },575 {576 "epoch": 0.01,577 "learning_rate": 1.999981643885681e-05,578 "loss": 0.6836,579 "step": 95580 },581 {582 "epoch": 0.01,583 "learning_rate": 1.9999812094378302e-05,584 "loss": 0.7371,585 "step": 96586 },587 {588 "epoch": 0.01,589 "learning_rate": 1.999980769908783e-05,590 "loss": 0.6447,591 "step": 97592 },593 {594 "epoch": 0.01,595 "learning_rate": 1.99998032529854e-05,596 "loss": 0.6473,597 "step": 98598 },599 {600 "epoch": 0.01,601 "learning_rate": 1.9999798756071044e-05,602 "loss": 0.6628,603 "step": 99604 },605 {606 "epoch": 0.01,607 "learning_rate": 1.9999794208344777e-05,608 "loss": 0.6186,609 "step": 100610 },611 {612 "epoch": 0.01,613 "learning_rate": 1.9999789609806633e-05,614 "loss": 0.642,615 "step": 101616 },617 {618 "epoch": 0.01,619 "learning_rate": 1.999978496045663e-05,620 "loss": 0.6727,621 "step": 102622 },623 {624 "epoch": 0.01,625 "learning_rate": 1.9999780260294786e-05,626 "loss": 0.655,627 "step": 103628 },629 {630 "epoch": 0.01,631 "learning_rate": 1.9999775509321134e-05,632 "loss": 0.6585,633 "step": 104634 },635 {636 "epoch": 0.01,637 "learning_rate": 1.999977070753569e-05,638 "loss": 0.669,639 "step": 105640 },641 {642 "epoch": 0.01,643 "learning_rate": 1.9999765854938484e-05,644 "loss": 0.6407,645 "step": 106646 },647 {648 "epoch": 0.01,649 "learning_rate": 1.9999760951529544e-05,650 "loss": 0.6889,651 "step": 107652 },653 {654 "epoch": 0.01,655 "learning_rate": 1.9999755997308882e-05,656 "loss": 0.6441,657 "step": 108658 },659 {660 "epoch": 0.01,661 "learning_rate": 1.9999750992276537e-05,662 "loss": 0.6605,663 "step": 109664 },665 {666 "epoch": 0.01,667 "learning_rate": 1.999974593643253e-05,668 "loss": 0.6136,669 "step": 110670 },671 {672 "epoch": 0.01,673 "learning_rate": 1.999974082977688e-05,674 "loss": 0.6536,675 "step": 111676 },677 {678 "epoch": 0.01,679 "learning_rate": 1.999973567230962e-05,680 "loss": 0.6944,681 "step": 112682 },683 {684 "epoch": 0.01,685 "learning_rate": 1.9999730464030774e-05,686 "loss": 0.6415,687 "step": 113688 },689 {690 "epoch": 0.01,691 "learning_rate": 1.999972520494037e-05,692 "loss": 0.6477,693 "step": 114694 },695 {696 "epoch": 0.01,697 "learning_rate": 1.9999719895038434e-05,698 "loss": 0.653,699 "step": 115700 },701 {702 "epoch": 0.01,703 "learning_rate": 1.999971453432499e-05,704 "loss": 0.6666,705 "step": 116706 },707 {708 "epoch": 0.01,709 "learning_rate": 1.9999709122800074e-05,710 "loss": 0.6148,711 "step": 117712 },713 {714 "epoch": 0.01,715 "learning_rate": 1.99997036604637e-05,716 "loss": 0.6832,717 "step": 118718 },719 {720 "epoch": 0.01,721 "learning_rate": 1.9999698147315908e-05,722 "loss": 0.6611,723 "step": 119724 },725 {726 "epoch": 0.01,727 "learning_rate": 1.999969258335672e-05,728 "loss": 0.6612,729 "step": 120730 },731 {732 "epoch": 0.01,733 "learning_rate": 1.9999686968586166e-05,734 "loss": 0.7165,735 "step": 121736 },737 {738 "epoch": 0.01,739 "learning_rate": 1.9999681303004274e-05,740 "loss": 0.6566,741 "step": 122742 },743 {744 "epoch": 0.01,745 "learning_rate": 1.999967558661107e-05,746 "loss": 0.6613,747 "step": 123748 },749 {750 "epoch": 0.01,751 "learning_rate": 1.9999669819406588e-05,752 "loss": 0.6075,753 "step": 124754 },755 {756 "epoch": 0.01,757 "learning_rate": 1.9999664001390858e-05,758 "loss": 0.6626,759 "step": 125760 },761 {762 "epoch": 0.01,763 "learning_rate": 1.99996581325639e-05,764 "loss": 0.6032,765 "step": 126766 },767 {768 "epoch": 0.01,769 "learning_rate": 1.9999652212925756e-05,770 "loss": 0.6149,771 "step": 127772 },773 {774 "epoch": 0.01,775 "learning_rate": 1.9999646242476446e-05,776 "loss": 0.6232,777 "step": 128778 },779 {780 "epoch": 0.01,781 "learning_rate": 1.999964022121601e-05,782 "loss": 0.6603,783 "step": 129784 },785 {786 "epoch": 0.01,787 "learning_rate": 1.999963414914447e-05,788 "loss": 0.6051,789 "step": 130790 },791 {792 "epoch": 0.01,793 "learning_rate": 1.999962802626186e-05,794 "loss": 0.6183,795 "step": 131796 },797 {798 "epoch": 0.01,799 "learning_rate": 1.9999621852568218e-05,800 "loss": 0.7117,801 "step": 132802 },803 {804 "epoch": 0.01,805 "learning_rate": 1.999961562806356e-05,806 "loss": 0.6306,807 "step": 133808 },809 {810 "epoch": 0.01,811 "learning_rate": 1.999960935274793e-05,812 "loss": 0.6286,813 "step": 134814 },815 {816 "epoch": 0.01,817 "learning_rate": 1.9999603026621357e-05,818 "loss": 0.6584,819 "step": 135820 },821 {822 "epoch": 0.01,823 "learning_rate": 1.9999596649683872e-05,824 "loss": 0.6351,825 "step": 136826 },827 {828 "epoch": 0.01,829 "learning_rate": 1.9999590221935507e-05,830 "loss": 0.6882,831 "step": 137832 },833 {834 "epoch": 0.01,835 "learning_rate": 1.99995837433763e-05,836 "loss": 0.6423,837 "step": 138838 },839 {840 "epoch": 0.01,841 "learning_rate": 1.9999577214006277e-05,842 "loss": 0.6378,843 "step": 139844 },845 {846 "epoch": 0.01,847 "learning_rate": 1.9999570633825473e-05,848 "loss": 0.6231,849 "step": 140850 },851 {852 "epoch": 0.01,853 "learning_rate": 1.999956400283392e-05,854 "loss": 0.6098,855 "step": 141856 },857 {858 "epoch": 0.01,859 "learning_rate": 1.9999557321031657e-05,860 "loss": 0.6274,861 "step": 142862 },863 {864 "epoch": 0.01,865 "learning_rate": 1.999955058841871e-05,866 "loss": 0.666,867 "step": 143868 },869 {870 "epoch": 0.01,871 "learning_rate": 1.9999543804995124e-05,872 "loss": 0.5958,873 "step": 144874 },875 {876 "epoch": 0.01,877 "learning_rate": 1.9999536970760922e-05,878 "loss": 0.6623,879 "step": 145880 },881 {882 "epoch": 0.01,883 "learning_rate": 1.999953008571615e-05,884 "loss": 0.6809,885 "step": 146886 },887 {888 "epoch": 0.01,889 "learning_rate": 1.999952314986083e-05,890 "loss": 0.6326,891 "step": 147892 },893 {894 "epoch": 0.01,895 "learning_rate": 1.9999516163195013e-05,896 "loss": 0.6444,897 "step": 148898 },899 {900 "epoch": 0.01,901 "learning_rate": 1.999950912571872e-05,902 "loss": 0.6047,903 "step": 149904 },905 {906 "epoch": 0.01,907 "learning_rate": 1.999950203743199e-05,908 "loss": 0.6825,909 "step": 150910 },911 {912 "epoch": 0.01,913 "learning_rate": 1.9999494898334867e-05,914 "loss": 0.6323,915 "step": 151916 },917 {918 "epoch": 0.01,919 "learning_rate": 1.999948770842738e-05,920 "loss": 0.6542,921 "step": 152922 },923 {924 "epoch": 0.01,925 "learning_rate": 1.999948046770957e-05,926 "loss": 0.6996,927 "step": 153928 },929 {930 "epoch": 0.01,931 "learning_rate": 1.9999473176181467e-05,932 "loss": 0.6612,933 "step": 154934 },935 {936 "epoch": 0.01,937 "learning_rate": 1.9999465833843118e-05,938 "loss": 0.6072,939 "step": 155940 },941 {942 "epoch": 0.01,943 "learning_rate": 1.999945844069455e-05,944 "loss": 0.6298,945 "step": 156946 },947 {948 "epoch": 0.01,949 "learning_rate": 1.9999450996735806e-05,950 "loss": 0.6273,951 "step": 157952 },953 {954 "epoch": 0.01,955 "learning_rate": 1.9999443501966925e-05,956 "loss": 0.6236,957 "step": 158958 },959 {960 "epoch": 0.01,961 "learning_rate": 1.999943595638794e-05,962 "loss": 0.6336,963 "step": 159964 },965 {966 "epoch": 0.01,967 "learning_rate": 1.9999428359998893e-05,968 "loss": 0.6489,969 "step": 160970 },971 {972 "epoch": 0.01,973 "learning_rate": 1.9999420712799823e-05,974 "loss": 0.6568,975 "step": 161976 },977 {978 "epoch": 0.01,979 "learning_rate": 1.9999413014790767e-05,980 "loss": 0.6712,981 "step": 162982 },983 {984 "epoch": 0.01,985 "learning_rate": 1.9999405265971768e-05,986 "loss": 0.6558,987 "step": 163988 },989 {990 "epoch": 0.01,991 "learning_rate": 1.999939746634286e-05,992 "loss": 0.6958,993 "step": 164994 },995 {996 "epoch": 0.01,997 "learning_rate": 1.9999389615904087e-05,998 "loss": 0.6978,999 "step": 1651000 },1001 {1002 "epoch": 0.01,1003 "learning_rate": 1.9999381714655482e-05,1004 "loss": 0.5999,1005 "step": 1661006 },1007 {1008 "epoch": 0.01,1009 "learning_rate": 1.9999373762597094e-05,1010 "loss": 0.5917,1011 "step": 1671012 },1013 {1014 "epoch": 0.01,1015 "learning_rate": 1.9999365759728958e-05,1016 "loss": 0.6307,1017 "step": 1681018 },1019 {1020 "epoch": 0.01,1021 "learning_rate": 1.9999357706051116e-05,1022 "loss": 0.6465,1023 "step": 1691024 },1025 {1026 "epoch": 0.01,1027 "learning_rate": 1.999934960156361e-05,1028 "loss": 0.7325,1029 "step": 1701030 },1031 {1032 "epoch": 0.01,1033 "learning_rate": 1.9999341446266483e-05,1034 "loss": 0.674,1035 "step": 1711036 },1037 {1038 "epoch": 0.01,1039 "learning_rate": 1.9999333240159767e-05,1040 "loss": 0.6234,1041 "step": 1721042 },1043 {1044 "epoch": 0.01,1045 "learning_rate": 1.9999324983243515e-05,1046 "loss": 0.6656,1047 "step": 1731048 },1049 {1050 "epoch": 0.01,1051 "learning_rate": 1.9999316675517765e-05,1052 "loss": 0.7162,1053 "step": 1741054 },1055 {1056 "epoch": 0.01,1057 "learning_rate": 1.9999308316982557e-05,1058 "loss": 0.701,1059 "step": 1751060 },1061 {1062 "epoch": 0.01,1063 "learning_rate": 1.9999299907637935e-05,1064 "loss": 0.689,1065 "step": 1761066 },1067 {1068 "epoch": 0.01,1069 "learning_rate": 1.999929144748394e-05,1070 "loss": 0.724,1071 "step": 1771072 },1073 {1074 "epoch": 0.01,1075 "learning_rate": 1.999928293652062e-05,1076 "loss": 0.623,1077 "step": 1781078 },1079 {1080 "epoch": 0.01,1081 "learning_rate": 1.9999274374748015e-05,1082 "loss": 0.6698,1083 "step": 1791084 },1085 {1086 "epoch": 0.01,1087 "learning_rate": 1.9999265762166166e-05,1088 "loss": 0.7361,1089 "step": 1801090 },1091 {1092 "epoch": 0.01,1093 "learning_rate": 1.9999257098775117e-05,1094 "loss": 0.6837,1095 "step": 1811096 },1097 {1098 "epoch": 0.01,1099 "learning_rate": 1.999924838457492e-05,1100 "loss": 0.6037,1101 "step": 1821102 },1103 {1104 "epoch": 0.01,1105 "learning_rate": 1.999923961956561e-05,1106 "loss": 0.6448,1107 "step": 1831108 },1109 {1110 "epoch": 0.01,1111 "learning_rate": 1.9999230803747235e-05,1112 "loss": 0.6281,1113 "step": 1841114 },1115 {1116 "epoch": 0.01,1117 "learning_rate": 1.9999221937119842e-05,1118 "loss": 0.6935,1119 "step": 1851120 },1121 {1122 "epoch": 0.01,1123 "learning_rate": 1.9999213019683476e-05,1124 "loss": 0.6443,1125 "step": 1861126 },1127 {1128 "epoch": 0.01,1129 "learning_rate": 1.9999204051438174e-05,1130 "loss": 0.625,1131 "step": 1871132 },1133 {1134 "epoch": 0.01,1135 "learning_rate": 1.999919503238399e-05,1136 "loss": 0.6192,1137 "step": 1881138 },1139 {1140 "epoch": 0.01,1141 "learning_rate": 1.999918596252097e-05,1142 "loss": 0.5939,1143 "step": 1891144 },1145 {1146 "epoch": 0.01,1147 "learning_rate": 1.9999176841849154e-05,1148 "loss": 0.6487,1149 "step": 1901150 },1151 {1152 "epoch": 0.01,1153 "learning_rate": 1.9999167670368596e-05,1154 "loss": 0.6334,1155 "step": 1911156 },1157 {1158 "epoch": 0.01,1159 "learning_rate": 1.9999158448079336e-05,1160 "loss": 0.6681,1161 "step": 1921162 },1163 {1164 "epoch": 0.01,1165 "learning_rate": 1.9999149174981425e-05,1166 "loss": 0.6713,1167 "step": 1931168 },1169 {1170 "epoch": 0.01,1171 "learning_rate": 1.9999139851074907e-05,1172 "loss": 0.5888,1173 "step": 1941174 },1175 {1176 "epoch": 0.01,1177 "learning_rate": 1.9999130476359835e-05,1178 "loss": 0.6544,1179 "step": 1951180 },1181 {1182 "epoch": 0.01,1183 "learning_rate": 1.9999121050836248e-05,1184 "loss": 0.6688,1185 "step": 1961186 },1187 {1188 "epoch": 0.01,1189 "learning_rate": 1.99991115745042e-05,1190 "loss": 0.6771,1191 "step": 1971192 },1193 {1194 "epoch": 0.01,1195 "learning_rate": 1.999910204736374e-05,1196 "loss": 0.662,1197 "step": 1981198 },1199 {1200 "epoch": 0.01,