HuggingFaceTB/SmolLM2-135M-Instruct
4231.6m
1{2 "best_metric": null,3 "best_model_checkpoint": null,4 "epoch": 1.9973828840617638,5 "eval_steps": 100,6 "global_step": 954,7 "is_hyper_param_search": false,8 "is_local_process_zero": true,9 "is_world_process_zero": true,10 "log_history": [11 {12 "epoch": 0.002093692750588851,13 "grad_norm": 38.333577570579074,14 "learning_rate": 1.0416666666666666e-08,15 "logits/chosen": 5.468747138977051,16 "logits/rejected": 5.353150367736816,17 "logps/chosen": -399.0700988769531,18 "logps/rejected": -414.2703857421875,19 "loss": 0.6931,20 "rewards/accuracies": 0.0,21 "rewards/chosen": 0.0,22 "rewards/margins": 0.0,23 "rewards/rejected": 0.0,24 "step": 125 },26 {27 "epoch": 0.02093692750588851,28 "grad_norm": 36.45097456473781,29 "learning_rate": 1.0416666666666667e-07,30 "logits/chosen": 4.634159088134766,31 "logits/rejected": 4.8650617599487305,32 "logps/chosen": -481.9865417480469,33 "logps/rejected": -402.9172668457031,34 "loss": 0.7192,35 "rewards/accuracies": 0.4236111044883728,36 "rewards/chosen": 0.036201052367687225,37 "rewards/margins": 0.05521820858120918,38 "rewards/rejected": -0.01901715248823166,39 "step": 1040 },41 {42 "epoch": 0.04187385501177702,43 "grad_norm": 38.31461130932718,44 "learning_rate": 2.0833333333333333e-07,45 "logits/chosen": 4.8017449378967285,46 "logits/rejected": 5.193596363067627,47 "logps/chosen": -428.74591064453125,48 "logps/rejected": -379.7098693847656,49 "loss": 0.7525,50 "rewards/accuracies": 0.512499988079071,51 "rewards/chosen": 0.014457901008427143,52 "rewards/margins": -0.03727109357714653,53 "rewards/rejected": 0.0517289862036705,54 "step": 2055 },56 {57 "epoch": 0.06281078251766553,58 "grad_norm": 38.39033659648525,59 "learning_rate": 3.1249999999999997e-07,60 "logits/chosen": 4.625308513641357,61 "logits/rejected": 4.913487434387207,62 "logps/chosen": -459.8934631347656,63 "logps/rejected": -365.87176513671875,64 "loss": 0.7389,65 "rewards/accuracies": 0.543749988079071,66 "rewards/chosen": 0.032367587089538574,67 "rewards/margins": 0.07106685638427734,68 "rewards/rejected": -0.03869926929473877,69 "step": 3070 },71 {72 "epoch": 0.08374771002355404,73 "grad_norm": 34.959297165636315,74 "learning_rate": 4.1666666666666667e-07,75 "logits/chosen": 5.135643005371094,76 "logits/rejected": 5.29467248916626,77 "logps/chosen": -388.5003662109375,78 "logps/rejected": -341.11138916015625,79 "loss": 0.7521,80 "rewards/accuracies": 0.44999998807907104,81 "rewards/chosen": -0.061884332448244095,82 "rewards/margins": -0.051958512514829636,83 "rewards/rejected": -0.009925814345479012,84 "step": 4085 },86 {87 "epoch": 0.10468463752944256,88 "grad_norm": 40.59746593571697,89 "learning_rate": 5.208333333333334e-07,90 "logits/chosen": 4.794947147369385,91 "logits/rejected": 5.206262111663818,92 "logps/chosen": -418.7637634277344,93 "logps/rejected": -366.21783447265625,94 "loss": 0.7539,95 "rewards/accuracies": 0.5249999761581421,96 "rewards/chosen": -0.0046131848357617855,97 "rewards/margins": 0.009517465718090534,98 "rewards/rejected": -0.014130651950836182,99 "step": 50100 },101 {102 "epoch": 0.12562156503533106,103 "grad_norm": 35.116262927070615,104 "learning_rate": 6.249999999999999e-07,105 "logits/chosen": 4.984349250793457,106 "logits/rejected": 5.210784435272217,107 "logps/chosen": -389.5479431152344,108 "logps/rejected": -355.3258361816406,109 "loss": 0.7337,110 "rewards/accuracies": 0.581250011920929,111 "rewards/chosen": 0.09027661383152008,112 "rewards/margins": 0.13924987614154816,113 "rewards/rejected": -0.048973266035318375,114 "step": 60115 },116 {117 "epoch": 0.14655849254121958,118 "grad_norm": 37.65630163274615,119 "learning_rate": 7.291666666666666e-07,120 "logits/chosen": 5.079476356506348,121 "logits/rejected": 5.1062331199646,122 "logps/chosen": -472.6788635253906,123 "logps/rejected": -410.6566467285156,124 "loss": 0.7532,125 "rewards/accuracies": 0.48750001192092896,126 "rewards/chosen": 0.01137494295835495,127 "rewards/margins": -0.019245151430368423,128 "rewards/rejected": 0.030620098114013672,129 "step": 70130 },131 {132 "epoch": 0.16749542004710807,133 "grad_norm": 36.35209638887961,134 "learning_rate": 8.333333333333333e-07,135 "logits/chosen": 4.831971645355225,136 "logits/rejected": 5.179555892944336,137 "logps/chosen": -465.8661193847656,138 "logps/rejected": -352.46063232421875,139 "loss": 0.7337,140 "rewards/accuracies": 0.5249999761581421,141 "rewards/chosen": 0.029347699135541916,142 "rewards/margins": 0.011580700054764748,143 "rewards/rejected": 0.017767000943422318,144 "step": 80145 },146 {147 "epoch": 0.1884323475529966,148 "grad_norm": 40.247342074541066,149 "learning_rate": 9.374999999999999e-07,150 "logits/chosen": 4.667853355407715,151 "logits/rejected": 5.083367347717285,152 "logps/chosen": -410.145263671875,153 "logps/rejected": -347.24871826171875,154 "loss": 0.7325,155 "rewards/accuracies": 0.5625,156 "rewards/chosen": 0.04706032946705818,157 "rewards/margins": 0.05596155673265457,158 "rewards/rejected": -0.008901228196918964,159 "step": 90160 },161 {162 "epoch": 0.2093692750588851,163 "grad_norm": 36.88448509348576,164 "learning_rate": 9.999463737538052e-07,165 "logits/chosen": 5.017066955566406,166 "logits/rejected": 5.157826900482178,167 "logps/chosen": -453.6114196777344,168 "logps/rejected": -376.13214111328125,169 "loss": 0.7296,170 "rewards/accuracies": 0.543749988079071,171 "rewards/chosen": 0.03123999759554863,172 "rewards/margins": 0.038888636976480484,173 "rewards/rejected": -0.007648637983947992,174 "step": 100175 },176 {177 "epoch": 0.2093692750588851,178 "eval_logits/chosen": 4.755386829376221,179 "eval_logits/rejected": 5.127224445343018,180 "eval_logps/chosen": -443.48101806640625,181 "eval_logps/rejected": -377.5273742675781,182 "eval_loss": 0.7357296347618103,183 "eval_rewards/accuracies": 0.5515872836112976,184 "eval_rewards/chosen": 0.01168334111571312,185 "eval_rewards/margins": 0.03692733868956566,186 "eval_rewards/rejected": -0.025244001299142838,187 "eval_runtime": 21.3186,188 "eval_samples_per_second": 93.815,189 "eval_steps_per_second": 2.955,190 "step": 100191 },192 {193 "epoch": 0.23030620256477363,194 "grad_norm": 36.21924654761057,195 "learning_rate": 9.993432105822034e-07,196 "logits/chosen": 4.768385887145996,197 "logits/rejected": 5.076653957366943,198 "logps/chosen": -449.16375732421875,199 "logps/rejected": -369.2919006347656,200 "loss": 0.7211,201 "rewards/accuracies": 0.42500001192092896,202 "rewards/chosen": -0.04003235697746277,203 "rewards/margins": -0.059906214475631714,204 "rewards/rejected": 0.019873863086104393,205 "step": 110206 },207 {208 "epoch": 0.2512431300706621,209 "grad_norm": 36.12599439727971,210 "learning_rate": 9.980706626858607e-07,211 "logits/chosen": 5.0697174072265625,212 "logits/rejected": 5.350961208343506,213 "logps/chosen": -392.9084777832031,214 "logps/rejected": -342.9964294433594,215 "loss": 0.7213,216 "rewards/accuracies": 0.543749988079071,217 "rewards/chosen": 0.005513651762157679,218 "rewards/margins": 0.07320307195186615,219 "rewards/rejected": -0.06768941879272461,220 "step": 120221 },222 {223 "epoch": 0.2721800575765506,224 "grad_norm": 43.81008737879554,225 "learning_rate": 9.961304359538434e-07,226 "logits/chosen": 4.7396440505981445,227 "logits/rejected": 5.110291957855225,228 "logps/chosen": -445.08209228515625,229 "logps/rejected": -356.9689636230469,230 "loss": 0.7319,231 "rewards/accuracies": 0.543749988079071,232 "rewards/chosen": 0.03245898336172104,233 "rewards/margins": 0.0837341919541359,234 "rewards/rejected": -0.05127520486712456,235 "step": 130236 },237 {238 "epoch": 0.29311698508243916,239 "grad_norm": 37.12091478913465,240 "learning_rate": 9.935251313189563e-07,241 "logits/chosen": 4.5339274406433105,242 "logits/rejected": 5.020459175109863,243 "logps/chosen": -473.4126892089844,244 "logps/rejected": -364.12939453125,245 "loss": 0.7193,246 "rewards/accuracies": 0.5562499761581421,247 "rewards/chosen": 0.008073748089373112,248 "rewards/margins": 0.035515300929546356,249 "rewards/rejected": -0.02744155190885067,250 "step": 140251 },252 {253 "epoch": 0.31405391258832765,254 "grad_norm": 36.18564480374988,255 "learning_rate": 9.902582412711118e-07,256 "logits/chosen": 4.540812969207764,257 "logits/rejected": 4.964258193969727,258 "logps/chosen": -426.5033264160156,259 "logps/rejected": -353.1463317871094,260 "loss": 0.7232,261 "rewards/accuracies": 0.512499988079071,262 "rewards/chosen": -0.01480414904654026,263 "rewards/margins": 0.06241898611187935,264 "rewards/rejected": -0.07722313702106476,265 "step": 150266 },267 {268 "epoch": 0.33499084009421615,269 "grad_norm": 32.90233228487631,270 "learning_rate": 9.86334145175542e-07,271 "logits/chosen": 4.807779788970947,272 "logits/rejected": 5.042156219482422,273 "logps/chosen": -396.0440673828125,274 "logps/rejected": -360.52886962890625,275 "loss": 0.7013,276 "rewards/accuracies": 0.5249999761581421,277 "rewards/chosen": 0.010227044112980366,278 "rewards/margins": 0.055318038910627365,279 "rewards/rejected": -0.045090995728969574,280 "step": 160281 },282 {283 "epoch": 0.3559277676001047,284 "grad_norm": 37.311964741290105,285 "learning_rate": 9.817581034021272e-07,286 "logits/chosen": 4.897703170776367,287 "logits/rejected": 5.062272071838379,288 "logps/chosen": -389.55810546875,289 "logps/rejected": -329.748779296875,290 "loss": 0.7043,291 "rewards/accuracies": 0.518750011920929,292 "rewards/chosen": -0.008063828572630882,293 "rewards/margins": 0.049024712294340134,294 "rewards/rejected": -0.05708853527903557,295 "step": 170296 },297 {298 "epoch": 0.3768646951059932,299 "grad_norm": 34.14102924438106,300 "learning_rate": 9.765362502737097e-07,301 "logits/chosen": 5.039429187774658,302 "logits/rejected": 5.049492835998535,303 "logps/chosen": -384.9471130371094,304 "logps/rejected": -381.6601257324219,305 "loss": 0.7091,306 "rewards/accuracies": 0.5562499761581421,307 "rewards/chosen": 0.004380516707897186,308 "rewards/margins": 0.038515396416187286,309 "rewards/rejected": -0.0341348834335804,310 "step": 180311 },312 {313 "epoch": 0.39780162261188173,314 "grad_norm": 35.88568083270778,315 "learning_rate": 9.706755858428485e-07,316 "logits/chosen": 5.025214195251465,317 "logits/rejected": 5.097342014312744,318 "logps/chosen": -397.56402587890625,319 "logps/rejected": -396.12799072265625,320 "loss": 0.7161,321 "rewards/accuracies": 0.518750011920929,322 "rewards/chosen": -0.056682147085666656,323 "rewards/margins": 0.023842817172408104,324 "rewards/rejected": -0.08052496612071991,325 "step": 190326 },327 {328 "epoch": 0.4187385501177702,329 "grad_norm": 38.60485416145746,330 "learning_rate": 9.641839665080363e-07,331 "logits/chosen": 5.1590471267700195,332 "logits/rejected": 5.290652275085449,333 "logps/chosen": -399.9522399902344,334 "logps/rejected": -363.53936767578125,335 "loss": 0.7062,336 "rewards/accuracies": 0.53125,337 "rewards/chosen": 0.04297986626625061,338 "rewards/margins": 0.06266864389181137,339 "rewards/rejected": -0.019688773900270462,340 "step": 200341 },342 {343 "epoch": 0.4187385501177702,344 "eval_logits/chosen": 4.725487232208252,345 "eval_logits/rejected": 5.087867736816406,346 "eval_logps/chosen": -443.55450439453125,347 "eval_logps/rejected": -377.6705627441406,348 "eval_loss": 0.6988219022750854,349 "eval_rewards/accuracies": 0.567460298538208,350 "eval_rewards/chosen": -0.025081120431423187,351 "eval_rewards/margins": 0.07174728065729141,352 "eval_rewards/rejected": -0.0968284010887146,353 "eval_runtime": 21.5315,354 "eval_samples_per_second": 92.887,355 "eval_steps_per_second": 2.926,356 "step": 200357 },358 {359 "epoch": 0.4396754776236587,360 "grad_norm": 36.10684145537435,361 "learning_rate": 9.570700944819582e-07,362 "logits/chosen": 4.827897548675537,363 "logits/rejected": 5.154609680175781,364 "logps/chosen": -451.2969665527344,365 "logps/rejected": -372.116455078125,366 "loss": 0.7141,367 "rewards/accuracies": 0.543749988079071,368 "rewards/chosen": -0.07117662578821182,369 "rewards/margins": 0.02444976009428501,370 "rewards/rejected": -0.09562637656927109,371 "step": 210372 },373 {374 "epoch": 0.46061240512954726,375 "grad_norm": 33.13953697631782,376 "learning_rate": 9.493435061259129e-07,377 "logits/chosen": 5.24191427230835,378 "logits/rejected": 5.477172374725342,379 "logps/chosen": -365.3572692871094,380 "logps/rejected": -345.34814453125,381 "loss": 0.7138,382 "rewards/accuracies": 0.48750001192092896,383 "rewards/chosen": -0.07684006541967392,384 "rewards/margins": -0.033464811742305756,385 "rewards/rejected": -0.043375253677368164,386 "step": 220387 },388 {389 "epoch": 0.48154933263543576,390 "grad_norm": 36.46869252662201,391 "learning_rate": 9.4101455916603e-07,392 "logits/chosen": 4.996638298034668,393 "logits/rejected": 5.203185081481934,394 "logps/chosen": -390.2169494628906,395 "logps/rejected": -381.74090576171875,396 "loss": 0.7027,397 "rewards/accuracies": 0.574999988079071,398 "rewards/chosen": -0.006761978380382061,399 "rewards/margins": 0.11298926174640656,400 "rewards/rejected": -0.11975125223398209,401 "step": 230402 },403 {404 "epoch": 0.5024862601413242,405 "grad_norm": 32.465417090707724,406 "learning_rate": 9.320944188084241e-07,407 "logits/chosen": 4.961588382720947,408 "logits/rejected": 5.104936122894287,409 "logps/chosen": -405.73651123046875,410 "logps/rejected": -368.39312744140625,411 "loss": 0.6966,412 "rewards/accuracies": 0.612500011920929,413 "rewards/chosen": 0.04590672254562378,414 "rewards/margins": 0.21882851421833038,415 "rewards/rejected": -0.1729217916727066,416 "step": 240417 },418 {419 "epoch": 0.5234231876472127,420 "grad_norm": 32.56446548910322,421 "learning_rate": 9.225950427718974e-07,422 "logits/chosen": 4.295259475708008,423 "logits/rejected": 4.731950759887695,424 "logps/chosen": -457.6085510253906,425 "logps/rejected": -378.92083740234375,426 "loss": 0.6895,427 "rewards/accuracies": 0.5249999761581421,428 "rewards/chosen": -0.08796132355928421,429 "rewards/margins": 0.07854396849870682,430 "rewards/rejected": -0.16650527715682983,431 "step": 250432 },433 {434 "epoch": 0.5443601151531012,435 "grad_norm": 37.3555609307497,436 "learning_rate": 9.125291652582547e-07,437 "logits/chosen": 4.772681713104248,438 "logits/rejected": 4.774602890014648,439 "logps/chosen": -429.554931640625,440 "logps/rejected": -350.5638122558594,441 "loss": 0.697,442 "rewards/accuracies": 0.574999988079071,443 "rewards/chosen": 0.017600687220692635,444 "rewards/margins": 0.15435068309307098,445 "rewards/rejected": -0.1367500126361847,446 "step": 260447 },448 {449 "epoch": 0.5652970426589898,450 "grad_norm": 33.42195670184768,451 "learning_rate": 9.019102798817195e-07,452 "logits/chosen": 4.580355644226074,453 "logits/rejected": 5.011557102203369,454 "logps/chosen": -446.68438720703125,455 "logps/rejected": -380.5400390625,456 "loss": 0.6778,457 "rewards/accuracies": 0.5625,458 "rewards/chosen": 0.019351882860064507,459 "rewards/margins": 0.15824225544929504,460 "rewards/rejected": -0.13889038562774658,461 "step": 270462 },463 {464 "epoch": 0.5862339701648783,465 "grad_norm": 39.119947805234894,466 "learning_rate": 8.90752621580335e-07,467 "logits/chosen": 5.025314807891846,468 "logits/rejected": 5.18468713760376,469 "logps/chosen": -424.27191162109375,470 "logps/rejected": -344.4115295410156,471 "loss": 0.7073,472 "rewards/accuracies": 0.625,473 "rewards/chosen": 0.023305395618081093,474 "rewards/margins": 0.14927226305007935,475 "rewards/rejected": -0.125966876745224,476 "step": 280477 },478 {479 "epoch": 0.6071708976707668,480 "grad_norm": 33.472133866310024,481 "learning_rate": 8.79071147533597e-07,482 "logits/chosen": 4.961835861206055,483 "logits/rejected": 5.123082637786865,484 "logps/chosen": -400.4245300292969,485 "logps/rejected": -388.8963623046875,486 "loss": 0.6687,487 "rewards/accuracies": 0.643750011920929,488 "rewards/chosen": 0.08449111878871918,489 "rewards/margins": 0.2874522805213928,490 "rewards/rejected": -0.20296116173267365,491 "step": 290492 },493 {494 "epoch": 0.6281078251766553,495 "grad_norm": 38.29458361274397,496 "learning_rate": 8.668815171119019e-07,497 "logits/chosen": 4.6071085929870605,498 "logits/rejected": 4.85768985748291,499 "logps/chosen": -445.59393310546875,500 "logps/rejected": -373.83636474609375,501 "loss": 0.6782,502 "rewards/accuracies": 0.643750011920929,503 "rewards/chosen": 0.043323811143636703,504 "rewards/margins": 0.2540794312953949,505 "rewards/rejected": -0.2107556313276291,506 "step": 300507 },508 {509 "epoch": 0.6281078251766553,510 "eval_logits/chosen": 4.662118434906006,511 "eval_logits/rejected": 5.016141414642334,512 "eval_logps/chosen": -443.56884765625,513 "eval_logps/rejected": -377.883056640625,514 "eval_loss": 0.6942777037620544,515 "eval_rewards/accuracies": 0.567460298538208,516 "eval_rewards/chosen": -0.03225937858223915,517 "eval_rewards/margins": 0.17080551385879517,518 "eval_rewards/rejected": -0.20306488871574402,519 "eval_runtime": 21.6344,520 "eval_samples_per_second": 92.445,521 "eval_steps_per_second": 2.912,522 "step": 300523 },524 {525 "epoch": 0.6490447526825438,526 "grad_norm": 35.277636776310345,527 "learning_rate": 8.54200070884685e-07,528 "logits/chosen": 4.7398271560668945,529 "logits/rejected": 5.0438690185546875,530 "logps/chosen": -455.08074951171875,531 "logps/rejected": -346.21905517578125,532 "loss": 0.6648,533 "rewards/accuracies": 0.5874999761581421,534 "rewards/chosen": -0.00532907247543335,535 "rewards/margins": 0.22818481922149658,536 "rewards/rejected": -0.23351387679576874,537 "step": 310538 },539 {540 "epoch": 0.6699816801884323,541 "grad_norm": 36.1242464612453,542 "learning_rate": 8.410438087153911e-07,543 "logits/chosen": 4.823008060455322,544 "logits/rejected": 4.949624538421631,545 "logps/chosen": -420.04150390625,546 "logps/rejected": -346.31134033203125,547 "loss": 0.6633,548 "rewards/accuracies": 0.5625,549 "rewards/chosen": -0.02057427167892456,550 "rewards/margins": 0.1925923228263855,551 "rewards/rejected": -0.21316656470298767,552 "step": 320553 },554 {555 "epoch": 0.6909186076943209,556 "grad_norm": 34.485635067716444,557 "learning_rate": 8.274303669726426e-07,558 "logits/chosen": 4.866278171539307,559 "logits/rejected": 5.084838390350342,560 "logps/chosen": -413.07647705078125,561 "logps/rejected": -359.72637939453125,562 "loss": 0.6964,563 "rewards/accuracies": 0.5562499761581421,564 "rewards/chosen": -0.0661710649728775,565 "rewards/margins": 0.11275775730609894,566 "rewards/rejected": -0.17892882227897644,567 "step": 330568 },569 {570 "epoch": 0.7118555352002094,571 "grad_norm": 36.021204090397475,572 "learning_rate": 8.133779948881513e-07,573 "logits/chosen": 4.962647914886475,574 "logits/rejected": 5.274256229400635,575 "logps/chosen": -423.34796142578125,576 "logps/rejected": -374.83831787109375,577 "loss": 0.6843,578 "rewards/accuracies": 0.581250011920929,579 "rewards/chosen": -0.06308840215206146,580 "rewards/margins": 0.14286582171916962,581 "rewards/rejected": -0.20595422387123108,582 "step": 340583 },584 {585 "epoch": 0.7327924627060979,586 "grad_norm": 40.80633953486743,587 "learning_rate": 7.989055300930704e-07,588 "logits/chosen": 4.9410552978515625,589 "logits/rejected": 5.171365737915039,590 "logps/chosen": -401.3800048828125,591 "logps/rejected": -339.8207702636719,592 "loss": 0.6799,593 "rewards/accuracies": 0.581250011920929,594 "rewards/chosen": -0.05132729932665825,595 "rewards/margins": 0.14917483925819397,596 "rewards/rejected": -0.20050212740898132,597 "step": 350598 },599 {600 "epoch": 0.7537293902119864,601 "grad_norm": 32.290367261199215,602 "learning_rate": 7.840323733655778e-07,603 "logits/chosen": 4.760105609893799,604 "logits/rejected": 4.936800956726074,605 "logps/chosen": -475.94305419921875,606 "logps/rejected": -373.4317626953125,607 "loss": 0.6723,608 "rewards/accuracies": 0.65625,609 "rewards/chosen": 0.040363796055316925,610 "rewards/margins": 0.26596716046333313,611 "rewards/rejected": -0.2256033718585968,612 "step": 360613 },614 {615 "epoch": 0.7746663177178749,616 "grad_norm": 33.44911142948228,617 "learning_rate": 7.687784626235447e-07,618 "logits/chosen": 4.649796485900879,619 "logits/rejected": 4.882054328918457,620 "logps/chosen": -437.54791259765625,621 "logps/rejected": -343.1330871582031,622 "loss": 0.6722,623 "rewards/accuracies": 0.6000000238418579,624 "rewards/chosen": -0.03584844991564751,625 "rewards/margins": 0.2080194056034088,626 "rewards/rejected": -0.2438678741455078,627 "step": 370628 },629 {630 "epoch": 0.7956032452237635,631 "grad_norm": 32.64019199720913,632 "learning_rate": 7.531642461971514e-07,633 "logits/chosen": 4.7331953048706055,634 "logits/rejected": 5.047934532165527,635 "logps/chosen": -434.0751953125,636 "logps/rejected": -363.1179504394531,637 "loss": 0.673,638 "rewards/accuracies": 0.581250011920929,639 "rewards/chosen": -0.03724004700779915,640 "rewards/margins": 0.23209133744239807,641 "rewards/rejected": -0.2693313956260681,642 "step": 380643 },644 {645 "epoch": 0.816540172729652,646 "grad_norm": 33.38515796622506,647 "learning_rate": 7.372106554172801e-07,648 "logits/chosen": 4.660643577575684,649 "logits/rejected": 4.7719621658325195,650 "logps/chosen": -434.41015625,651 "logps/rejected": -394.9471130371094,652 "loss": 0.6778,653 "rewards/accuracies": 0.59375,654 "rewards/chosen": 0.021881069988012314,655 "rewards/margins": 0.2415298968553543,656 "rewards/rejected": -0.2196488082408905,657 "step": 390658 },659 {660 "epoch": 0.8374771002355405,661 "grad_norm": 32.60957530709497,662 "learning_rate": 7.209390765564318e-07,663 "logits/chosen": 4.807684421539307,664 "logits/rejected": 5.217709541320801,665 "logps/chosen": -368.07122802734375,666 "logps/rejected": -328.12066650390625,667 "loss": 0.6863,668 "rewards/accuracies": 0.59375,669 "rewards/chosen": -0.03271085396409035,670 "rewards/margins": 0.1988353729248047,671 "rewards/rejected": -0.23154623806476593,672 "step": 400673 },674 {675 "epoch": 0.8374771002355405,676 "eval_logits/chosen": 4.645900249481201,677 "eval_logits/rejected": 4.999230861663818,678 "eval_logps/chosen": -443.68084716796875,679 "eval_logps/rejected": -378.0348205566406,680 "eval_loss": 0.6756832003593445,681 "eval_rewards/accuracies": 0.5992063283920288,682 "eval_rewards/chosen": -0.08822782337665558,683 "eval_rewards/margins": 0.19070643186569214,684 "eval_rewards/rejected": -0.2789342403411865,685 "eval_runtime": 21.4973,686 "eval_samples_per_second": 93.035,687 "eval_steps_per_second": 2.931,688 "step": 400689 },690 {691 "epoch": 0.8584140277414289,692 "grad_norm": 80.32344495768098,693 "learning_rate": 7.043713221597773e-07,694 "logits/chosen": 4.9558234214782715,695 "logits/rejected": 5.171336650848389,696 "logps/chosen": -464.4634704589844,697 "logps/rejected": -378.52130126953125,698 "loss": 0.6691,699 "rewards/accuracies": 0.5687500238418579,700 "rewards/chosen": -0.028427015990018845,701 "rewards/margins": 0.16248683631420135,702 "rewards/rejected": -0.1909138560295105,703 "step": 410704 },705 {706 "epoch": 0.8793509552473174,707 "grad_norm": 38.003974257278905,708 "learning_rate": 6.875296018047809e-07,709 "logits/chosen": 5.062918663024902,710 "logits/rejected": 5.093894958496094,711 "logps/chosen": -414.597900390625,712 "logps/rejected": -392.76422119140625,713 "loss": 0.6778,714 "rewards/accuracies": 0.612500011920929,715 "rewards/chosen": -0.05205658823251724,716 "rewards/margins": 0.166357159614563,717 "rewards/rejected": -0.21841374039649963,718 "step": 420719 },720 {721 "epoch": 0.9002878827532059,722 "grad_norm": 33.39326538286459,723 "learning_rate": 6.704364923285857e-07,724 "logits/chosen": 4.783626556396484,725 "logits/rejected": 5.061443328857422,726 "logps/chosen": -454.7694396972656,727 "logps/rejected": -349.71099853515625,728 "loss": 0.6613,729 "rewards/accuracies": 0.6312500238418579,730 "rewards/chosen": -0.03129550814628601,731 "rewards/margins": 0.2672887146472931,732 "rewards/rejected": -0.2985842227935791,733 "step": 430734 },735 {736 "epoch": 0.9212248102590945,737 "grad_norm": 33.88594593881104,738 "learning_rate": 6.531149075630796e-07,739 "logits/chosen": 4.762629985809326,740 "logits/rejected": 4.992688179016113,741 "logps/chosen": -422.49639892578125,742 "logps/rejected": -342.6626892089844,743 "loss": 0.6829,744 "rewards/accuracies": 0.574999988079071,745 "rewards/chosen": -0.08267354220151901,746 "rewards/margins": 0.18921074271202087,747 "rewards/rejected": -0.2718842923641205,748 "step": 440749 },750 {751 "epoch": 0.942161737764983,752 "grad_norm": 34.194378360359096,753 "learning_rate": 6.355880676182085e-07,754 "logits/chosen": 4.86130952835083,755 "logits/rejected": 5.088041305541992,756 "logps/chosen": -423.82366943359375,757 "logps/rejected": -386.20172119140625,758 "loss": 0.6777,759 "rewards/accuracies": 0.637499988079071,760 "rewards/chosen": -0.029518108814954758,761 "rewards/margins": 0.2812921106815338,762 "rewards/rejected": -0.3108102083206177,763 "step": 450764 },765 {766 "epoch": 0.9630986652708715,767 "grad_norm": 35.220161431379815,768 "learning_rate": 6.178794677547137e-07,769 "logits/chosen": 4.96859073638916,770 "logits/rejected": 5.295912265777588,771 "logps/chosen": -408.28228759765625,772 "logps/rejected": -337.6819763183594,773 "loss": 0.6573,774 "rewards/accuracies": 0.65625,775 "rewards/chosen": -0.02989841438829899,776 "rewards/margins": 0.3131854832172394,777 "rewards/rejected": -0.343083918094635,778 "step": 460779 },780 {781 "epoch": 0.98403559277676,782 "grad_norm": 36.11741005068747,783 "learning_rate": 6.000128468880222e-07,784 "logits/chosen": 4.616504669189453,785 "logits/rejected": 4.935946464538574,786 "logps/chosen": -435.3017578125,787 "logps/rejected": -375.13800048828125,788 "loss": 0.6647,789 "rewards/accuracies": 0.606249988079071,790 "rewards/chosen": -0.11524273455142975,791 "rewards/margins": 0.2546766698360443,792 "rewards/rejected": -0.3699193596839905,793 "step": 470794 },795 {796 "epoch": 1.0049725202826485,797 "grad_norm": 32.471857091487834,798 "learning_rate": 5.820121557655108e-07,799 "logits/chosen": 4.9493536949157715,800 "logits/rejected": 5.226868152618408,801 "logps/chosen": -423.6285095214844,802 "logps/rejected": -362.1949768066406,803 "loss": 0.6629,804 "rewards/accuracies": 0.637499988079071,805 "rewards/chosen": 0.05261852219700813,806 "rewards/margins": 0.3280298113822937,807 "rewards/rejected": -0.27541130781173706,808 "step": 480809 },810 {811 "epoch": 1.025909447788537,812 "grad_norm": 39.51652905408664,813 "learning_rate": 5.639015248598023e-07,814 "logits/chosen": 4.762259006500244,815 "logits/rejected": 5.021244525909424,816 "logps/chosen": -424.96697998046875,817 "logps/rejected": -342.76666259765625,818 "loss": 0.6644,819 "rewards/accuracies": 0.6000000238418579,820 "rewards/chosen": -0.03716667741537094,821 "rewards/margins": 0.2011403739452362,822 "rewards/rejected": -0.23830704391002655,823 "step": 490824 },825 {826 "epoch": 1.0468463752944255,827 "grad_norm": 34.43579926142672,828 "learning_rate": 5.457052320211339e-07,829 "logits/chosen": 4.543593406677246,830 "logits/rejected": 4.786489009857178,831 "logps/chosen": -434.46746826171875,832 "logps/rejected": -367.75689697265625,833 "loss": 0.6836,834 "rewards/accuracies": 0.606249988079071,835 "rewards/chosen": -0.08059108257293701,836 "rewards/margins": 0.23346829414367676,837 "rewards/rejected": -0.3140593469142914,838 "step": 500839 },840 {841 "epoch": 1.0468463752944255,842 "eval_logits/chosen": 4.617003440856934,843 "eval_logits/rejected": 4.9695563316345215,844 "eval_logps/chosen": -443.69580078125,845 "eval_logps/rejected": -378.1418762207031,846 "eval_loss": 0.6708300113677979,847 "eval_rewards/accuracies": 0.6349206566810608,848 "eval_rewards/chosen": -0.09571509808301926,849 "eval_rewards/margins": 0.23677198588848114,850 "eval_rewards/rejected": -0.3324871063232422,851 "eval_runtime": 20.9626,852 "eval_samples_per_second": 95.408,853 "eval_steps_per_second": 3.005,854 "step": 500855 },856 {857 "epoch": 1.067783302800314,858 "grad_norm": 33.3201987416808,859 "learning_rate": 5.274476699321637e-07,860 "logits/chosen": 4.583409786224365,861 "logits/rejected": 4.803020477294922,862 "logps/chosen": -390.48565673828125,863 "logps/rejected": -351.6776123046875,864 "loss": 0.6779,865 "rewards/accuracies": 0.6312500238418579,866 "rewards/chosen": -0.0327589213848114,867 "rewards/margins": 0.32544368505477905,868 "rewards/rejected": -0.35820263624191284,869 "step": 510870 },871 {872 "epoch": 1.0887202303062025,873 "grad_norm": 30.742730814185435,874 "learning_rate": 5.091533134088387e-07,875 "logits/chosen": 4.493949890136719,876 "logits/rejected": 4.9839911460876465,877 "logps/chosen": -383.7958984375,878 "logps/rejected": -354.36480712890625,879 "loss": 0.656,880 "rewards/accuracies": 0.5375000238418579,881 "rewards/chosen": -0.07341745495796204,882 "rewards/margins": 0.19709812104701996,883 "rewards/rejected": -0.2705155909061432,884 "step": 520885 },886 {887 "epoch": 1.109657157812091,888 "grad_norm": 34.05900047947194,889 "learning_rate": 4.908466865911614e-07,890 "logits/chosen": 4.793222904205322,891 "logits/rejected": 5.078155517578125,892 "logps/chosen": -401.0002746582031,893 "logps/rejected": -340.4061279296875,894 "loss": 0.6618,895 "rewards/accuracies": 0.6499999761581421,896 "rewards/chosen": -0.05051114410161972,897 "rewards/margins": 0.27152642607688904,898 "rewards/rejected": -0.32203757762908936,899 "step": 530900 },901 {902 "epoch": 1.1305940853179797,903 "grad_norm": 30.483486401054424,904 "learning_rate": 4.7255233006783624e-07,905 "logits/chosen": 4.857717990875244,906 "logits/rejected": 5.0497636795043945,907 "logps/chosen": -375.65362548828125,908 "logps/rejected": -330.26165771484375,909 "loss": 0.6544,910 "rewards/accuracies": 0.7124999761581421,911 "rewards/chosen": -0.02352207899093628,912 "rewards/margins": 0.35729408264160156,913 "rewards/rejected": -0.38081610202789307,914 "step": 540915 },916 {917 "epoch": 1.151531012823868,918 "grad_norm": 35.09603470685652,919 "learning_rate": 4.5429476797886617e-07,920 "logits/chosen": 4.932369232177734,921 "logits/rejected": 5.050224781036377,922 "logps/chosen": -430.0126953125,923 "logps/rejected": -331.1691589355469,924 "loss": 0.6599,925 "rewards/accuracies": 0.6312500238418579,926 "rewards/chosen": -0.0060789333656430244,927 "rewards/margins": 0.25288811326026917,928 "rewards/rejected": -0.25896701216697693,929 "step": 550930 },931 {932 "epoch": 1.1724679403297567,933 "grad_norm": 40.64422646125966,934 "learning_rate": 4.3609847514019763e-07,935 "logits/chosen": 4.637743949890137,936 "logits/rejected": 5.000674724578857,937 "logps/chosen": -420.3258361816406,938 "logps/rejected": -362.2751159667969,939 "loss": 0.6718,940 "rewards/accuracies": 0.574999988079071,941 "rewards/chosen": -0.0094971414655447,942 "rewards/margins": 0.22678783535957336,943 "rewards/rejected": -0.23628497123718262,944 "step": 560945 },946 {947 "epoch": 1.193404867835645,948 "grad_norm": 32.638009640148645,949 "learning_rate": 4.179878442344892e-07,950 "logits/chosen": 4.855754375457764,951 "logits/rejected": 4.871184349060059,952 "logps/chosen": -384.08660888671875,953 "logps/rejected": -371.4262390136719,954 "loss": 0.6766,955 "rewards/accuracies": 0.625,956 "rewards/chosen": -0.05095939710736275,957 "rewards/margins": 0.28148993849754333,958 "rewards/rejected": -0.332449346780777,959 "step": 570960 },961 {962 "epoch": 1.2143417953415336,963 "grad_norm": 35.519971577107064,964 "learning_rate": 3.9998715311197783e-07,965 "logits/chosen": 4.73850679397583,966 "logits/rejected": 5.173120021820068,967 "logps/chosen": -414.8775329589844,968 "logps/rejected": -341.5818786621094,969 "loss": 0.6508,970 "rewards/accuracies": 0.6000000238418579,971 "rewards/chosen": -0.09668377041816711,972 "rewards/margins": 0.25211262702941895,973 "rewards/rejected": -0.34879642724990845,974 "step": 580975 },976 {977 "epoch": 1.235278722847422,978 "grad_norm": 34.20580765627037,979 "learning_rate": 3.821205322452863e-07,980 "logits/chosen": 4.916988372802734,981 "logits/rejected": 5.1998610496521,982 "logps/chosen": -448.5626525878906,983 "logps/rejected": -367.84027099609375,984 "loss": 0.644,985 "rewards/accuracies": 0.625,986 "rewards/chosen": -0.07886572182178497,987 "rewards/margins": 0.3578983247280121,988 "rewards/rejected": -0.43676406145095825,989 "step": 590990 },991 {992 "epoch": 1.2562156503533106,993 "grad_norm": 33.854286929995176,994 "learning_rate": 3.6441193238179146e-07,995 "logits/chosen": 4.852269649505615,996 "logits/rejected": 4.903324127197266,997 "logps/chosen": -446.4149475097656,998 "logps/rejected": -423.3356018066406,999 "loss": 0.6349,1000 "rewards/accuracies": 0.581250011920929,1001 "rewards/chosen": -0.14010193943977356,1002 "rewards/margins": 0.15067996084690094,1003 "rewards/rejected": -0.2907818853855133,1004 "step": 6001005 },1006 {1007 "epoch": 1.2562156503533106,1008 "eval_logits/chosen": 4.62031364440918,1009 "eval_logits/rejected": 4.9707465171813965,1010 "eval_logps/chosen": -443.61212158203125,1011 "eval_logps/rejected": -378.1197204589844,1012 "eval_loss": 0.6720485091209412,1013 "eval_rewards/accuracies": 0.5992063283920288,1014 "eval_rewards/chosen": -0.053870752453804016,1015 "eval_rewards/margins": 0.267531156539917,1016 "eval_rewards/rejected": -0.321401983499527,1017 "eval_runtime": 20.8046,1018 "eval_samples_per_second": 96.133,1019 "eval_steps_per_second": 3.028,1020 "step": 6001021 },1022 {1023 "epoch": 1.2771525778591992,1024 "grad_norm": 36.085842973391074,1025 "learning_rate": 3.4688509243692034e-07,1026 "logits/chosen": 4.767918586730957,1027 "logits/rejected": 4.757430553436279,1028 "logps/chosen": -407.41668701171875,1029 "logps/rejected": -317.3873596191406,1030 "loss": 0.6402,1031 "rewards/accuracies": 0.643750011920929,1032 "rewards/chosen": -0.08183420449495316,1033 "rewards/margins": 0.33883604407310486,1034 "rewards/rejected": -0.42067021131515503,1035 "step": 6101036 },1037 {1038 "epoch": 1.2980895053650876,1039 "grad_norm": 29.698333183198105,1040 "learning_rate": 3.295635076714144e-07,1041 "logits/chosen": 5.085806846618652,1042 "logits/rejected": 5.415268898010254,1043 "logps/chosen": -395.627685546875,1044 "logps/rejected": -331.7653503417969,1045 "loss": 0.6266,1046 "rewards/accuracies": 0.65625,1047 "rewards/chosen": -0.09060301631689072,1048 "rewards/margins": 0.3094441294670105,1049 "rewards/rejected": -0.4000471234321594,1050 "step": 6201051 },1052 {1053 "epoch": 1.3190264328709762,1054 "grad_norm": 35.208773349468885,1055 "learning_rate": 3.12470398195219e-07,1056 "logits/chosen": 4.828533172607422,1057 "logits/rejected": 4.925856113433838,1058 "logps/chosen": -418.5848083496094,1059 "logps/rejected": -376.3353576660156,1060 "loss": 0.6486,1061 "rewards/accuracies": 0.6312500238418579,1062 "rewards/chosen": 0.06391973793506622,1063 "rewards/margins": 0.44674786925315857,1064 "rewards/rejected": -0.3828281760215759,1065 "step": 6301066 },1067 {1068 "epoch": 1.3399633603768648,1069 "grad_norm": 29.673309842493335,1070 "learning_rate": 2.956286778402226e-07,1071 "logits/chosen": 4.896113872528076,1072 "logits/rejected": 5.183098793029785,1073 "logps/chosen": -394.4980773925781,1074 "logps/rejected": -374.76422119140625,1075 "loss": 0.6394,1076 "rewards/accuracies": 0.6312500238418579,1077 "rewards/chosen": -0.04267222806811333,1078 "rewards/margins": 0.2913575768470764,1079 "rewards/rejected": -0.33402982354164124,1080 "step": 6401081 },1082 {1083 "epoch": 1.3609002878827532,1084 "grad_norm": 35.03684415648848,1085 "learning_rate": 2.7906092344356826e-07,1086 "logits/chosen": 4.610795021057129,1087 "logits/rejected": 4.8373188972473145,1088 "logps/chosen": -379.4288024902344,1089 "logps/rejected": -345.05596923828125,1090 "loss": 0.6646,1091 "rewards/accuracies": 0.581250011920929,1092 "rewards/chosen": -0.1311950385570526,1093 "rewards/margins": 0.2357216328382492,1094 "rewards/rejected": -0.366916686296463,1095 "step": 6501096 },1097 {1098 "epoch": 1.3818372153886418,1099 "grad_norm": 33.06984951084542,1100 "learning_rate": 2.6278934458271996e-07,1101 "logits/chosen": 4.830328941345215,1102 "logits/rejected": 5.017812252044678,1103 "logps/chosen": -377.4278564453125,1104 "logps/rejected": -343.86529541015625,1105 "loss": 0.6613,1106 "rewards/accuracies": 0.5625,1107 "rewards/chosen": -0.10503290593624115,1108 "rewards/margins": 0.11762680858373642,1109 "rewards/rejected": -0.22265975177288055,1110 "step": 6601111 },1112 {1113 "epoch": 1.4027741428945302,1114 "grad_norm": 31.761556922593446,1115 "learning_rate": 2.468357538028487e-07,1116 "logits/chosen": 4.728631496429443,1117 "logits/rejected": 4.90619421005249,1118 "logps/chosen": -413.2724609375,1119 "logps/rejected": -346.9877624511719,1120 "loss": 0.6393,1121 "rewards/accuracies": 0.637499988079071,1122 "rewards/chosen": -0.03663766756653786,1123 "rewards/margins": 0.2855125069618225,1124 "rewards/rejected": -0.32215017080307007,1125 "step": 6701126 },1127 {1128 "epoch": 1.4237110704004188,1129 "grad_norm": 34.93162849349177,1130 "learning_rate": 2.312215373764551e-07,1131 "logits/chosen": 4.728277206420898,1132 "logits/rejected": 5.018845558166504,1133 "logps/chosen": -421.8961486816406,1134 "logps/rejected": -403.57354736328125,1135 "loss": 0.6533,1136 "rewards/accuracies": 0.606249988079071,1137 "rewards/chosen": -0.1260446161031723,1138 "rewards/margins": 0.2200162708759308,1139 "rewards/rejected": -0.3460609018802643,1140 "step": 6801141 },1142 {1143 "epoch": 1.4446479979063072,1144 "grad_norm": 33.66523822793528,1145 "learning_rate": 2.1596762663442213e-07,1146 "logits/chosen": 4.863284111022949,1147 "logits/rejected": 4.840500354766846,1148 "logps/chosen": -422.4331970214844,1149 "logps/rejected": -355.96868896484375,1150 "loss": 0.6477,1151 "rewards/accuracies": 0.625,1152 "rewards/chosen": -0.10065089166164398,1153 "rewards/margins": 0.24511468410491943,1154 "rewards/rejected": -0.3457655906677246,1155 "step": 6901156 },1157 {1158 "epoch": 1.4655849254121958,1159 "grad_norm": 34.48257400044076,1160 "learning_rate": 2.0109446990692963e-07,1161 "logits/chosen": 4.709015846252441,1162 "logits/rejected": 4.914425849914551,1163 "logps/chosen": -452.9461364746094,1164 "logps/rejected": -442.56658935546875,1165 "loss": 0.6427,1166 "rewards/accuracies": 0.65625,1167 "rewards/chosen": 0.08032918721437454,1168 "rewards/margins": 0.39584842324256897,1169 "rewards/rejected": -0.3155192732810974,1170 "step": 7001171 },1172 {1173 "epoch": 1.4655849254121958,1174 "eval_logits/chosen": 4.592012882232666,1175 "eval_logits/rejected": 4.943046569824219,1176 "eval_logps/chosen": -443.6796875,1177 "eval_logps/rejected": -378.1680908203125,1178 "eval_loss": 0.6795812845230103,1179 "eval_rewards/accuracies": 0.60317462682724,1180 "eval_rewards/chosen": -0.08766676485538483,1181 "eval_rewards/margins": 0.25792673230171204,1182 "eval_rewards/rejected": -0.34559354186058044,1183 "eval_runtime": 21.1978,1184 "eval_samples_per_second": 94.349,1185 "eval_steps_per_second": 2.972,1186 "step": 7001187 },1188 {1189 "epoch": 1.4865218529180844,1190 "grad_norm": 36.350524448045036,1191 "learning_rate": 1.8662200511184872e-07,1192 "logits/chosen": 4.871232509613037,1193 "logits/rejected": 4.886293411254883,1194 "logps/chosen": -417.8133850097656,1195 "logps/rejected": -384.177490234375,1196 "loss": 0.6669,1197 "rewards/accuracies": 0.6000000238418579,1198 "rewards/chosen": -0.12399878352880478,1199 "rewards/margins": 0.27855515480041504,1200 "rewards/rejected": -0.40255388617515564,