{ "best_metric": null, "best_model_checkpoint": null, "epoch": 0.9915492957746479, "eval_steps": 18, "global_step": 88, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.011267605633802818, "grad_norm": 13.436114311218262, "learning_rate": 8e-07, "log_odds_chosen": 0.3692830204963684, "log_odds_ratio": -0.6691731214523315, "logits/chosen": -2.073309898376465, "logits/rejected": -2.0528674125671387, "logps/chosen": -6.679821491241455, "logps/rejected": -7.049269676208496, "loss": 7.3222, "nll_loss": 7.255331993103027, "rewards/accuracies": 0.5, "rewards/chosen": -0.6679821610450745, "rewards/margins": 0.03694486618041992, "rewards/rejected": -0.7049269676208496, "step": 1 }, { "epoch": 0.022535211267605635, "grad_norm": 12.660109519958496, "learning_rate": 1.6e-06, "log_odds_chosen": 0.05790233612060547, "log_odds_ratio": -0.7999902963638306, "logits/chosen": -2.122316360473633, "logits/rejected": -2.1929054260253906, "logps/chosen": -6.760974884033203, "logps/rejected": -6.818158149719238, "loss": 6.7429, "nll_loss": 6.662899971008301, "rewards/accuracies": 0.5, "rewards/chosen": -0.6760974526405334, "rewards/margins": 0.005718342959880829, "rewards/rejected": -0.6818158626556396, "step": 2 }, { "epoch": 0.03380281690140845, "grad_norm": 15.815394401550293, "learning_rate": 2.4e-06, "log_odds_chosen": 0.5895991325378418, "log_odds_ratio": -0.5308127999305725, "logits/chosen": -2.167480945587158, "logits/rejected": -2.164055347442627, "logps/chosen": -6.8017683029174805, "logps/rejected": -7.390748977661133, "loss": 7.2921, "nll_loss": 7.239068031311035, "rewards/accuracies": 0.75, "rewards/chosen": -0.680176854133606, "rewards/margins": 0.058898091316223145, "rewards/rejected": -0.7390749454498291, "step": 3 }, { "epoch": 0.04507042253521127, "grad_norm": 15.963722229003906, "learning_rate": 3.2e-06, "log_odds_chosen": -0.2698395252227783, "log_odds_ratio": -0.8873093128204346, "logits/chosen": -2.084567070007324, "logits/rejected": -2.0936269760131836, "logps/chosen": -6.373527526855469, "logps/rejected": -6.103606224060059, "loss": 7.1739, "nll_loss": 7.0851545333862305, "rewards/accuracies": 0.25, "rewards/chosen": -0.637352705001831, "rewards/margins": -0.026992086321115494, "rewards/rejected": -0.6103606224060059, "step": 4 }, { "epoch": 0.056338028169014086, "grad_norm": 14.874364852905273, "learning_rate": 4e-06, "log_odds_chosen": -0.19485323131084442, "log_odds_ratio": -0.9221639633178711, "logits/chosen": -2.071868896484375, "logits/rejected": -2.113917827606201, "logps/chosen": -6.983762741088867, "logps/rejected": -6.790091037750244, "loss": 7.3473, "nll_loss": 7.25510835647583, "rewards/accuracies": 0.5, "rewards/chosen": -0.6983763575553894, "rewards/margins": -0.019367169588804245, "rewards/rejected": -0.6790091395378113, "step": 5 }, { "epoch": 0.0676056338028169, "grad_norm": 12.149578094482422, "learning_rate": 4.8e-06, "log_odds_chosen": 1.0678657293319702, "log_odds_ratio": -0.3254351019859314, "logits/chosen": -2.24808669090271, "logits/rejected": -2.239896774291992, "logps/chosen": -6.008720397949219, "logps/rejected": -7.073472023010254, "loss": 6.3764, "nll_loss": 6.343899250030518, "rewards/accuracies": 1.0, "rewards/chosen": -0.6008720993995667, "rewards/margins": 0.10647513717412949, "rewards/rejected": -0.7073472142219543, "step": 6 }, { "epoch": 0.07887323943661972, "grad_norm": 17.523408889770508, "learning_rate": 5.6e-06, "log_odds_chosen": -0.13575804233551025, "log_odds_ratio": -0.8457558751106262, "logits/chosen": -2.143648386001587, "logits/rejected": -2.1670777797698975, "logps/chosen": -6.814686298370361, "logps/rejected": -6.678918838500977, "loss": 7.1252, "nll_loss": 7.040646553039551, "rewards/accuracies": 0.375, "rewards/chosen": -0.6814686059951782, "rewards/margins": -0.01357676088809967, "rewards/rejected": -0.6678918600082397, "step": 7 }, { "epoch": 0.09014084507042254, "grad_norm": 14.829681396484375, "learning_rate": 6.4e-06, "log_odds_chosen": -0.01214347779750824, "log_odds_ratio": -0.7939789295196533, "logits/chosen": -2.1688995361328125, "logits/rejected": -2.2231764793395996, "logps/chosen": -6.625088691711426, "logps/rejected": -6.613010406494141, "loss": 7.1343, "nll_loss": 7.054868221282959, "rewards/accuracies": 0.5, "rewards/chosen": -0.6625088453292847, "rewards/margins": -0.001207888126373291, "rewards/rejected": -0.6613010168075562, "step": 8 }, { "epoch": 0.10140845070422536, "grad_norm": 13.42191219329834, "learning_rate": 7.2e-06, "log_odds_chosen": -0.026338055729866028, "log_odds_ratio": -0.9790022969245911, "logits/chosen": -2.1700892448425293, "logits/rejected": -2.2076737880706787, "logps/chosen": -6.157310485839844, "logps/rejected": -6.1309003829956055, "loss": 6.7773, "nll_loss": 6.679416656494141, "rewards/accuracies": 0.375, "rewards/chosen": -0.6157310605049133, "rewards/margins": -0.0026410408318042755, "rewards/rejected": -0.6130900382995605, "step": 9 }, { "epoch": 0.11267605633802817, "grad_norm": 13.783522605895996, "learning_rate": 8e-06, "log_odds_chosen": 0.06254176795482635, "log_odds_ratio": -0.7391232252120972, "logits/chosen": -2.1378564834594727, "logits/rejected": -2.195751667022705, "logps/chosen": -6.942321300506592, "logps/rejected": -7.00494384765625, "loss": 7.045, "nll_loss": 6.971048355102539, "rewards/accuracies": 0.5, "rewards/chosen": -0.694232165813446, "rewards/margins": 0.006262235343456268, "rewards/rejected": -0.7004944086074829, "step": 10 }, { "epoch": 0.12394366197183099, "grad_norm": 16.540300369262695, "learning_rate": 7.897435897435898e-06, "log_odds_chosen": -0.39260566234588623, "log_odds_ratio": -0.9909158945083618, "logits/chosen": -2.095216751098633, "logits/rejected": -2.0956408977508545, "logps/chosen": -6.841823101043701, "logps/rejected": -6.4507737159729, "loss": 7.197, "nll_loss": 7.097899436950684, "rewards/accuracies": 0.25, "rewards/chosen": -0.6841824054718018, "rewards/margins": -0.03910496458411217, "rewards/rejected": -0.6450773477554321, "step": 11 }, { "epoch": 0.1352112676056338, "grad_norm": 13.459568977355957, "learning_rate": 7.794871794871794e-06, "log_odds_chosen": 0.20161187648773193, "log_odds_ratio": -0.661760687828064, "logits/chosen": -2.1810107231140137, "logits/rejected": -2.2119851112365723, "logps/chosen": -5.996310234069824, "logps/rejected": -6.197461128234863, "loss": 6.3979, "nll_loss": 6.331758975982666, "rewards/accuracies": 0.5, "rewards/chosen": -0.5996310710906982, "rewards/margins": 0.020115099847316742, "rewards/rejected": -0.6197460889816284, "step": 12 }, { "epoch": 0.14647887323943662, "grad_norm": 11.37065601348877, "learning_rate": 7.692307692307692e-06, "log_odds_chosen": 0.2524498701095581, "log_odds_ratio": -0.6112524271011353, "logits/chosen": -2.2797930240631104, "logits/rejected": -2.2433085441589355, "logps/chosen": -5.814953327178955, "logps/rejected": -6.066402435302734, "loss": 6.0765, "nll_loss": 6.015379428863525, "rewards/accuracies": 0.75, "rewards/chosen": -0.5814952850341797, "rewards/margins": 0.02514498308300972, "rewards/rejected": -0.6066402792930603, "step": 13 }, { "epoch": 0.15774647887323945, "grad_norm": 12.900912284851074, "learning_rate": 7.589743589743589e-06, "log_odds_chosen": -0.4836970567703247, "log_odds_ratio": -1.0226668119430542, "logits/chosen": -2.215272903442383, "logits/rejected": -2.2382960319519043, "logps/chosen": -6.522824287414551, "logps/rejected": -6.040271759033203, "loss": 6.7429, "nll_loss": 6.640633583068848, "rewards/accuracies": 0.125, "rewards/chosen": -0.6522824764251709, "rewards/margins": -0.0482553206384182, "rewards/rejected": -0.6040271520614624, "step": 14 }, { "epoch": 0.16901408450704225, "grad_norm": 11.437642097473145, "learning_rate": 7.487179487179487e-06, "log_odds_chosen": 0.15005001425743103, "log_odds_ratio": -0.6865947842597961, "logits/chosen": -2.3359432220458984, "logits/rejected": -2.382951021194458, "logps/chosen": -6.031307220458984, "logps/rejected": -6.181303024291992, "loss": 6.4206, "nll_loss": 6.3519721031188965, "rewards/accuracies": 0.5, "rewards/chosen": -0.6031307578086853, "rewards/margins": 0.014999527484178543, "rewards/rejected": -0.6181303262710571, "step": 15 }, { "epoch": 0.18028169014084508, "grad_norm": 11.808122634887695, "learning_rate": 7.384615384615385e-06, "log_odds_chosen": -0.15265953540802002, "log_odds_ratio": -0.979753315448761, "logits/chosen": -2.309048652648926, "logits/rejected": -2.3273842334747314, "logps/chosen": -6.232041358947754, "logps/rejected": -6.080443859100342, "loss": 6.235, "nll_loss": 6.136982440948486, "rewards/accuracies": 0.25, "rewards/chosen": -0.6232041120529175, "rewards/margins": -0.015159666538238525, "rewards/rejected": -0.608044445514679, "step": 16 }, { "epoch": 0.19154929577464788, "grad_norm": 13.229806900024414, "learning_rate": 7.282051282051281e-06, "log_odds_chosen": -0.13886889815330505, "log_odds_ratio": -0.8053479194641113, "logits/chosen": -2.339905261993408, "logits/rejected": -2.3555707931518555, "logps/chosen": -5.892672061920166, "logps/rejected": -5.755575180053711, "loss": 6.0915, "nll_loss": 6.010932445526123, "rewards/accuracies": 0.375, "rewards/chosen": -0.5892672538757324, "rewards/margins": -0.0137096606194973, "rewards/rejected": -0.5755575895309448, "step": 17 }, { "epoch": 0.2028169014084507, "grad_norm": 12.478164672851562, "learning_rate": 7.179487179487179e-06, "log_odds_chosen": -0.1651839315891266, "log_odds_ratio": -0.8343566656112671, "logits/chosen": -2.440002918243408, "logits/rejected": -2.462096691131592, "logps/chosen": -6.1440839767456055, "logps/rejected": -5.97899055480957, "loss": 6.1604, "nll_loss": 6.076998233795166, "rewards/accuracies": 0.375, "rewards/chosen": -0.6144084930419922, "rewards/margins": -0.01650938391685486, "rewards/rejected": -0.5978990793228149, "step": 18 }, { "epoch": 0.2028169014084507, "eval_log_odds_chosen": 0.23532851040363312, "eval_log_odds_ratio": -0.7237306833267212, "eval_logits/chosen": -2.3824925422668457, "eval_logits/rejected": -2.3515429496765137, "eval_logps/chosen": -5.896577835083008, "eval_logps/rejected": -6.131969451904297, "eval_loss": 6.009965896606445, "eval_nll_loss": 5.93759298324585, "eval_rewards/accuracies": 0.625, "eval_rewards/chosen": -0.5896577835083008, "eval_rewards/margins": 0.02353917807340622, "eval_rewards/rejected": -0.6131969690322876, "eval_runtime": 5.3524, "eval_samples_per_second": 1.495, "eval_steps_per_second": 0.747, "step": 18 }, { "epoch": 0.2140845070422535, "grad_norm": 14.38143253326416, "learning_rate": 7.076923076923076e-06, "log_odds_chosen": 0.41149377822875977, "log_odds_ratio": -0.5679665207862854, "logits/chosen": -2.2283358573913574, "logits/rejected": -2.2696075439453125, "logps/chosen": -5.357198715209961, "logps/rejected": -5.767269611358643, "loss": 5.4298, "nll_loss": 5.373036861419678, "rewards/accuracies": 0.75, "rewards/chosen": -0.5357198715209961, "rewards/margins": 0.04100704938173294, "rewards/rejected": -0.5767269730567932, "step": 19 }, { "epoch": 0.22535211267605634, "grad_norm": 13.76131534576416, "learning_rate": 6.974358974358974e-06, "log_odds_chosen": -0.2197912484407425, "log_odds_ratio": -0.8451388478279114, "logits/chosen": -2.3156001567840576, "logits/rejected": -2.3646183013916016, "logps/chosen": -5.359432697296143, "logps/rejected": -5.1403656005859375, "loss": 5.3198, "nll_loss": 5.235289096832275, "rewards/accuracies": 0.375, "rewards/chosen": -0.5359432697296143, "rewards/margins": -0.021906737238168716, "rewards/rejected": -0.5140365362167358, "step": 20 }, { "epoch": 0.23661971830985915, "grad_norm": 13.503231048583984, "learning_rate": 6.871794871794871e-06, "log_odds_chosen": -0.38248211145401, "log_odds_ratio": -1.001150369644165, "logits/chosen": -2.4095211029052734, "logits/rejected": -2.393695592880249, "logps/chosen": -5.506905555725098, "logps/rejected": -5.1283159255981445, "loss": 5.4066, "nll_loss": 5.306459903717041, "rewards/accuracies": 0.375, "rewards/chosen": -0.5506905913352966, "rewards/margins": -0.037858996540308, "rewards/rejected": -0.5128315687179565, "step": 21 }, { "epoch": 0.24788732394366197, "grad_norm": 11.429216384887695, "learning_rate": 6.769230769230769e-06, "log_odds_chosen": 0.3094485402107239, "log_odds_ratio": -0.5879026651382446, "logits/chosen": -2.545229434967041, "logits/rejected": -2.479078531265259, "logps/chosen": -4.822021484375, "logps/rejected": -5.128696918487549, "loss": 5.1873, "nll_loss": 5.128541469573975, "rewards/accuracies": 0.625, "rewards/chosen": -0.4822021722793579, "rewards/margins": 0.03066752851009369, "rewards/rejected": -0.5128697156906128, "step": 22 }, { "epoch": 0.2591549295774648, "grad_norm": 25.22710609436035, "learning_rate": 6.666666666666667e-06, "log_odds_chosen": -0.8079856634140015, "log_odds_ratio": -1.24208664894104, "logits/chosen": -2.232295036315918, "logits/rejected": -2.3469386100769043, "logps/chosen": -5.6021294593811035, "logps/rejected": -4.799851417541504, "loss": 5.6475, "nll_loss": 5.523254871368408, "rewards/accuracies": 0.125, "rewards/chosen": -0.5602129697799683, "rewards/margins": -0.08022782951593399, "rewards/rejected": -0.4799851179122925, "step": 23 }, { "epoch": 0.2704225352112676, "grad_norm": 14.158093452453613, "learning_rate": 6.564102564102564e-06, "log_odds_chosen": -0.18114782869815826, "log_odds_ratio": -0.8821460008621216, "logits/chosen": -2.4146053791046143, "logits/rejected": -2.373107671737671, "logps/chosen": -5.3170366287231445, "logps/rejected": -5.139774799346924, "loss": 5.4074, "nll_loss": 5.319136619567871, "rewards/accuracies": 0.625, "rewards/chosen": -0.5317037105560303, "rewards/margins": -0.017726212739944458, "rewards/rejected": -0.5139775276184082, "step": 24 }, { "epoch": 0.28169014084507044, "grad_norm": 16.50591468811035, "learning_rate": 6.461538461538462e-06, "log_odds_chosen": -0.2912433445453644, "log_odds_ratio": -0.9506481885910034, "logits/chosen": -2.361240863800049, "logits/rejected": -2.369182586669922, "logps/chosen": -5.270952224731445, "logps/rejected": -4.982378959655762, "loss": 5.3692, "nll_loss": 5.274152755737305, "rewards/accuracies": 0.625, "rewards/chosen": -0.5270952582359314, "rewards/margins": -0.028857309371232986, "rewards/rejected": -0.4982379674911499, "step": 25 }, { "epoch": 0.29295774647887324, "grad_norm": 14.906917572021484, "learning_rate": 6.358974358974358e-06, "log_odds_chosen": -0.3039167523384094, "log_odds_ratio": -0.9283076524734497, "logits/chosen": -2.428100109100342, "logits/rejected": -2.4127683639526367, "logps/chosen": -4.490434169769287, "logps/rejected": -4.19234037399292, "loss": 4.7795, "nll_loss": 4.686664581298828, "rewards/accuracies": 0.5, "rewards/chosen": -0.4490433931350708, "rewards/margins": -0.02980935573577881, "rewards/rejected": -0.4192340672016144, "step": 26 }, { "epoch": 0.30422535211267604, "grad_norm": 12.076257705688477, "learning_rate": 6.256410256410256e-06, "log_odds_chosen": -0.006805986166000366, "log_odds_ratio": -0.8016737699508667, "logits/chosen": -2.520240068435669, "logits/rejected": -2.4479568004608154, "logps/chosen": -4.917964458465576, "logps/rejected": -4.9105448722839355, "loss": 4.8248, "nll_loss": 4.744668483734131, "rewards/accuracies": 0.625, "rewards/chosen": -0.49179643392562866, "rewards/margins": -0.0007419846951961517, "rewards/rejected": -0.4910544455051422, "step": 27 }, { "epoch": 0.3154929577464789, "grad_norm": 13.027565002441406, "learning_rate": 6.153846153846154e-06, "log_odds_chosen": -0.18745584785938263, "log_odds_ratio": -0.8756581544876099, "logits/chosen": -2.5150811672210693, "logits/rejected": -2.483858108520508, "logps/chosen": -4.473248481750488, "logps/rejected": -4.290022373199463, "loss": 4.7801, "nll_loss": 4.692525386810303, "rewards/accuracies": 0.375, "rewards/chosen": -0.44732481241226196, "rewards/margins": -0.018322546035051346, "rewards/rejected": -0.4290022850036621, "step": 28 }, { "epoch": 0.3267605633802817, "grad_norm": 11.076545715332031, "learning_rate": 6.051282051282051e-06, "log_odds_chosen": -0.3999772071838379, "log_odds_ratio": -0.9861010909080505, "logits/chosen": -2.4857661724090576, "logits/rejected": -2.4484825134277344, "logps/chosen": -5.028776168823242, "logps/rejected": -4.632839202880859, "loss": 5.1706, "nll_loss": 5.071954727172852, "rewards/accuracies": 0.25, "rewards/chosen": -0.5028775930404663, "rewards/margins": -0.03959362953901291, "rewards/rejected": -0.4632839560508728, "step": 29 }, { "epoch": 0.3380281690140845, "grad_norm": 14.6617431640625, "learning_rate": 5.948717948717948e-06, "log_odds_chosen": -0.5616844892501831, "log_odds_ratio": -1.1660798788070679, "logits/chosen": -2.510787010192871, "logits/rejected": -2.4491658210754395, "logps/chosen": -4.740803241729736, "logps/rejected": -4.18714714050293, "loss": 4.4952, "nll_loss": 4.37855339050293, "rewards/accuracies": 0.375, "rewards/chosen": -0.47408032417297363, "rewards/margins": -0.055365610867738724, "rewards/rejected": -0.418714702129364, "step": 30 }, { "epoch": 0.3492957746478873, "grad_norm": 12.136735916137695, "learning_rate": 5.8461538461538454e-06, "log_odds_chosen": -0.5842780470848083, "log_odds_ratio": -1.139350175857544, "logits/chosen": -2.523271083831787, "logits/rejected": -2.4937047958374023, "logps/chosen": -4.513538837432861, "logps/rejected": -3.9382987022399902, "loss": 4.4507, "nll_loss": 4.336762428283691, "rewards/accuracies": 0.375, "rewards/chosen": -0.45135390758514404, "rewards/margins": -0.057523999363183975, "rewards/rejected": -0.393829882144928, "step": 31 }, { "epoch": 0.36056338028169016, "grad_norm": 13.099180221557617, "learning_rate": 5.743589743589743e-06, "log_odds_chosen": -0.37855735421180725, "log_odds_ratio": -0.9612523317337036, "logits/chosen": -2.5555570125579834, "logits/rejected": -2.4845659732818604, "logps/chosen": -4.688088417053223, "logps/rejected": -4.315810680389404, "loss": 4.7878, "nll_loss": 4.69166374206543, "rewards/accuracies": 0.375, "rewards/chosen": -0.4688088893890381, "rewards/margins": -0.0372278057038784, "rewards/rejected": -0.4315810799598694, "step": 32 }, { "epoch": 0.37183098591549296, "grad_norm": 13.548160552978516, "learning_rate": 5.641025641025641e-06, "log_odds_chosen": -0.6679523587226868, "log_odds_ratio": -1.1635067462921143, "logits/chosen": -2.4850564002990723, "logits/rejected": -2.4568276405334473, "logps/chosen": -4.625258445739746, "logps/rejected": -3.966517925262451, "loss": 4.3309, "nll_loss": 4.214511871337891, "rewards/accuracies": 0.25, "rewards/chosen": -0.4625258445739746, "rewards/margins": -0.06587404012680054, "rewards/rejected": -0.3966518044471741, "step": 33 }, { "epoch": 0.38309859154929576, "grad_norm": 11.71092700958252, "learning_rate": 5.538461538461538e-06, "log_odds_chosen": 0.24815137684345245, "log_odds_ratio": -0.642487645149231, "logits/chosen": -2.511573076248169, "logits/rejected": -2.466554880142212, "logps/chosen": -4.356619834899902, "logps/rejected": -4.603775501251221, "loss": 4.2485, "nll_loss": 4.18422794342041, "rewards/accuracies": 0.5, "rewards/chosen": -0.4356619715690613, "rewards/margins": 0.024715568870306015, "rewards/rejected": -0.4603775143623352, "step": 34 }, { "epoch": 0.39436619718309857, "grad_norm": 17.99422836303711, "learning_rate": 5.435897435897436e-06, "log_odds_chosen": -0.4279804825782776, "log_odds_ratio": -1.0195913314819336, "logits/chosen": -2.4478812217712402, "logits/rejected": -2.4185688495635986, "logps/chosen": -4.629607200622559, "logps/rejected": -4.212224960327148, "loss": 4.3901, "nll_loss": 4.288104057312012, "rewards/accuracies": 0.375, "rewards/chosen": -0.46296072006225586, "rewards/margins": -0.041738200932741165, "rewards/rejected": -0.4212225377559662, "step": 35 }, { "epoch": 0.4056338028169014, "grad_norm": 26.7180118560791, "learning_rate": 5.333333333333333e-06, "log_odds_chosen": -0.49325498938560486, "log_odds_ratio": -1.1076762676239014, "logits/chosen": -2.384472608566284, "logits/rejected": -2.384478807449341, "logps/chosen": -4.394374370574951, "logps/rejected": -3.9007925987243652, "loss": 4.5105, "nll_loss": 4.399693489074707, "rewards/accuracies": 0.25, "rewards/chosen": -0.4394374489784241, "rewards/margins": -0.04935820400714874, "rewards/rejected": -0.39007923007011414, "step": 36 }, { "epoch": 0.4056338028169014, "eval_log_odds_chosen": -0.021530766040086746, "eval_log_odds_ratio": -0.7920652627944946, "eval_logits/chosen": -2.437562942504883, "eval_logits/rejected": -2.4226341247558594, "eval_logps/chosen": -4.752222061157227, "eval_logps/rejected": -4.731179237365723, "eval_loss": 5.0522871017456055, "eval_nll_loss": 4.973080635070801, "eval_rewards/accuracies": 0.5, "eval_rewards/chosen": -0.47522225975990295, "eval_rewards/margins": -0.0021043792366981506, "eval_rewards/rejected": -0.4731178879737854, "eval_runtime": 5.3465, "eval_samples_per_second": 1.496, "eval_steps_per_second": 0.748, "step": 36 }, { "epoch": 0.4169014084507042, "grad_norm": 25.756877899169922, "learning_rate": 5.230769230769231e-06, "log_odds_chosen": -0.04049903154373169, "log_odds_ratio": -0.8640150427818298, "logits/chosen": -2.536158800125122, "logits/rejected": -2.481358051300049, "logps/chosen": -4.724617958068848, "logps/rejected": -4.686985492706299, "loss": 5.0137, "nll_loss": 4.927271366119385, "rewards/accuracies": 0.375, "rewards/chosen": -0.4724617600440979, "rewards/margins": -0.0037632212042808533, "rewards/rejected": -0.46869853138923645, "step": 37 }, { "epoch": 0.428169014084507, "grad_norm": 15.886479377746582, "learning_rate": 5.128205128205129e-06, "log_odds_chosen": -0.37556567788124084, "log_odds_ratio": -0.9566004276275635, "logits/chosen": -2.4243788719177246, "logits/rejected": -2.399319887161255, "logps/chosen": -4.280425071716309, "logps/rejected": -3.9121856689453125, "loss": 4.3575, "nll_loss": 4.2618303298950195, "rewards/accuracies": 0.25, "rewards/chosen": -0.428042471408844, "rewards/margins": -0.03682393580675125, "rewards/rejected": -0.3912185728549957, "step": 38 }, { "epoch": 0.4394366197183099, "grad_norm": 17.65472412109375, "learning_rate": 5.025641025641025e-06, "log_odds_chosen": 0.2230280339717865, "log_odds_ratio": -0.6618625521659851, "logits/chosen": -2.3576548099517822, "logits/rejected": -2.333857297897339, "logps/chosen": -3.9965877532958984, "logps/rejected": -4.214459419250488, "loss": 4.3671, "nll_loss": 4.300962448120117, "rewards/accuracies": 0.625, "rewards/chosen": -0.399658739566803, "rewards/margins": 0.02178717404603958, "rewards/rejected": -0.42144593596458435, "step": 39 }, { "epoch": 0.4507042253521127, "grad_norm": 11.106135368347168, "learning_rate": 4.923076923076923e-06, "log_odds_chosen": 0.0009728074073791504, "log_odds_ratio": -0.7869582772254944, "logits/chosen": -2.4159560203552246, "logits/rejected": -2.4208855628967285, "logps/chosen": -3.9074864387512207, "logps/rejected": -3.903223991394043, "loss": 4.1296, "nll_loss": 4.050935745239258, "rewards/accuracies": 0.5, "rewards/chosen": -0.39074867963790894, "rewards/margins": -0.0004262477159500122, "rewards/rejected": -0.39032241702079773, "step": 40 }, { "epoch": 0.4619718309859155, "grad_norm": 12.57712459564209, "learning_rate": 4.82051282051282e-06, "log_odds_chosen": -0.2918354868888855, "log_odds_ratio": -0.8778151869773865, "logits/chosen": -2.3547985553741455, "logits/rejected": -2.322421073913574, "logps/chosen": -4.1212158203125, "logps/rejected": -3.8306081295013428, "loss": 4.0642, "nll_loss": 3.9763693809509277, "rewards/accuracies": 0.375, "rewards/chosen": -0.41212156414985657, "rewards/margins": -0.029060743749141693, "rewards/rejected": -0.3830608129501343, "step": 41 }, { "epoch": 0.4732394366197183, "grad_norm": 12.822919845581055, "learning_rate": 4.717948717948718e-06, "log_odds_chosen": -0.810235321521759, "log_odds_ratio": -1.2506808042526245, "logits/chosen": -2.2839574813842773, "logits/rejected": -2.288829803466797, "logps/chosen": -4.443515300750732, "logps/rejected": -3.658328056335449, "loss": 4.3262, "nll_loss": 4.201174259185791, "rewards/accuracies": 0.125, "rewards/chosen": -0.44435155391693115, "rewards/margins": -0.07851871848106384, "rewards/rejected": -0.3658328056335449, "step": 42 }, { "epoch": 0.48450704225352115, "grad_norm": 12.537949562072754, "learning_rate": 4.6153846153846145e-06, "log_odds_chosen": -0.9749997854232788, "log_odds_ratio": -1.3443421125411987, "logits/chosen": -2.3435025215148926, "logits/rejected": -2.321125030517578, "logps/chosen": -4.51336145401001, "logps/rejected": -3.566725254058838, "loss": 4.5357, "nll_loss": 4.401299476623535, "rewards/accuracies": 0.125, "rewards/chosen": -0.45133620500564575, "rewards/margins": -0.09466366469860077, "rewards/rejected": -0.3566725254058838, "step": 43 }, { "epoch": 0.49577464788732395, "grad_norm": 10.948065757751465, "learning_rate": 4.5128205128205124e-06, "log_odds_chosen": -0.7950465083122253, "log_odds_ratio": -1.22953200340271, "logits/chosen": -2.319894790649414, "logits/rejected": -2.3266992568969727, "logps/chosen": -4.4190673828125, "logps/rejected": -3.646463394165039, "loss": 4.4846, "nll_loss": 4.361649513244629, "rewards/accuracies": 0.125, "rewards/chosen": -0.44190675020217896, "rewards/margins": -0.07726040482521057, "rewards/rejected": -0.3646463453769684, "step": 44 }, { "epoch": 0.5070422535211268, "grad_norm": 10.234577178955078, "learning_rate": 4.4102564102564104e-06, "log_odds_chosen": -0.03596124053001404, "log_odds_ratio": -0.7954179048538208, "logits/chosen": -2.3405230045318604, "logits/rejected": -2.3773298263549805, "logps/chosen": -3.984051465988159, "logps/rejected": -3.9488115310668945, "loss": 4.3053, "nll_loss": 4.225733280181885, "rewards/accuracies": 0.375, "rewards/chosen": -0.39840513467788696, "rewards/margins": -0.003523964434862137, "rewards/rejected": -0.3948811888694763, "step": 45 }, { "epoch": 0.5183098591549296, "grad_norm": 11.955574035644531, "learning_rate": 4.3076923076923076e-06, "log_odds_chosen": -0.47261345386505127, "log_odds_ratio": -1.0186831951141357, "logits/chosen": -2.2264914512634277, "logits/rejected": -2.2124805450439453, "logps/chosen": -4.316086769104004, "logps/rejected": -3.8509907722473145, "loss": 4.2714, "nll_loss": 4.169527053833008, "rewards/accuracies": 0.25, "rewards/chosen": -0.4316087067127228, "rewards/margins": -0.046509623527526855, "rewards/rejected": -0.3850990831851959, "step": 46 }, { "epoch": 0.5295774647887324, "grad_norm": 14.009207725524902, "learning_rate": 4.205128205128205e-06, "log_odds_chosen": -0.5015394687652588, "log_odds_ratio": -1.0697970390319824, "logits/chosen": -2.216539144515991, "logits/rejected": -2.236360549926758, "logps/chosen": -4.524508476257324, "logps/rejected": -4.033357620239258, "loss": 4.2457, "nll_loss": 4.138679504394531, "rewards/accuracies": 0.125, "rewards/chosen": -0.45245084166526794, "rewards/margins": -0.04911504685878754, "rewards/rejected": -0.4033357799053192, "step": 47 }, { "epoch": 0.5408450704225352, "grad_norm": 11.495881080627441, "learning_rate": 4.102564102564102e-06, "log_odds_chosen": -0.0002641826868057251, "log_odds_ratio": -0.7958545088768005, "logits/chosen": -2.2024898529052734, "logits/rejected": -2.2096405029296875, "logps/chosen": -3.650709629058838, "logps/rejected": -3.6475670337677, "loss": 3.7327, "nll_loss": 3.653069019317627, "rewards/accuracies": 0.375, "rewards/chosen": -0.3650709390640259, "rewards/margins": -0.0003142599016427994, "rewards/rejected": -0.36475670337677, "step": 48 }, { "epoch": 0.5521126760563381, "grad_norm": 10.599851608276367, "learning_rate": 4e-06, "log_odds_chosen": -0.40274348855018616, "log_odds_ratio": -0.9958802461624146, "logits/chosen": -2.1248817443847656, "logits/rejected": -2.1678757667541504, "logps/chosen": -3.9321091175079346, "logps/rejected": -3.5374057292938232, "loss": 3.7393, "nll_loss": 3.6396658420562744, "rewards/accuracies": 0.375, "rewards/chosen": -0.39321091771125793, "rewards/margins": -0.03947034850716591, "rewards/rejected": -0.3537406027317047, "step": 49 }, { "epoch": 0.5633802816901409, "grad_norm": 14.113234519958496, "learning_rate": 3.897435897435897e-06, "log_odds_chosen": -0.15864649415016174, "log_odds_ratio": -0.8118338584899902, "logits/chosen": -2.182473659515381, "logits/rejected": -2.2479374408721924, "logps/chosen": -4.00465202331543, "logps/rejected": -3.8499021530151367, "loss": 3.984, "nll_loss": 3.902796745300293, "rewards/accuracies": 0.5, "rewards/chosen": -0.4004652500152588, "rewards/margins": -0.015475019812583923, "rewards/rejected": -0.38499021530151367, "step": 50 }, { "epoch": 0.5746478873239437, "grad_norm": 15.240550994873047, "learning_rate": 3.7948717948717945e-06, "log_odds_chosen": -0.3683556020259857, "log_odds_ratio": -1.0493247509002686, "logits/chosen": -2.142444133758545, "logits/rejected": -2.1910910606384277, "logps/chosen": -4.096914768218994, "logps/rejected": -3.7346315383911133, "loss": 3.9369, "nll_loss": 3.8319671154022217, "rewards/accuracies": 0.5, "rewards/chosen": -0.4096915125846863, "rewards/margins": -0.036228351294994354, "rewards/rejected": -0.37346315383911133, "step": 51 }, { "epoch": 0.5859154929577465, "grad_norm": 13.516912460327148, "learning_rate": 3.6923076923076925e-06, "log_odds_chosen": -0.14869263768196106, "log_odds_ratio": -0.8158061504364014, "logits/chosen": -2.1389424800872803, "logits/rejected": -2.1591155529022217, "logps/chosen": -3.914469003677368, "logps/rejected": -3.7656819820404053, "loss": 3.6203, "nll_loss": 3.538712978363037, "rewards/accuracies": 0.375, "rewards/chosen": -0.39144688844680786, "rewards/margins": -0.014878695830702782, "rewards/rejected": -0.3765681982040405, "step": 52 }, { "epoch": 0.5971830985915493, "grad_norm": 12.95617961883545, "learning_rate": 3.5897435897435896e-06, "log_odds_chosen": -0.24592818319797516, "log_odds_ratio": -0.9381157159805298, "logits/chosen": -2.0287747383117676, "logits/rejected": -2.1300435066223145, "logps/chosen": -3.8074216842651367, "logps/rejected": -3.571408748626709, "loss": 3.6729, "nll_loss": 3.5790419578552246, "rewards/accuracies": 0.375, "rewards/chosen": -0.3807421326637268, "rewards/margins": -0.023601267486810684, "rewards/rejected": -0.3571408987045288, "step": 53 }, { "epoch": 0.6084507042253521, "grad_norm": 11.833647727966309, "learning_rate": 3.487179487179487e-06, "log_odds_chosen": -0.5012875199317932, "log_odds_ratio": -1.0330029726028442, "logits/chosen": -2.141413688659668, "logits/rejected": -2.1845943927764893, "logps/chosen": -4.043262958526611, "logps/rejected": -3.5469398498535156, "loss": 3.9015, "nll_loss": 3.798215627670288, "rewards/accuracies": 0.25, "rewards/chosen": -0.40432626008987427, "rewards/margins": -0.049632273614406586, "rewards/rejected": -0.3546940088272095, "step": 54 }, { "epoch": 0.6084507042253521, "eval_log_odds_chosen": -0.0070199985057115555, "eval_log_odds_ratio": -0.8004560470581055, "eval_logits/chosen": -2.2082390785217285, "eval_logits/rejected": -2.2092061042785645, "eval_logps/chosen": -4.265551567077637, "eval_logps/rejected": -4.259222030639648, "eval_loss": 4.50450325012207, "eval_nll_loss": 4.42445707321167, "eval_rewards/accuracies": 0.5, "eval_rewards/chosen": -0.42655515670776367, "eval_rewards/margins": -0.0006329789757728577, "eval_rewards/rejected": -0.4259222149848938, "eval_runtime": 5.3539, "eval_samples_per_second": 1.494, "eval_steps_per_second": 0.747, "step": 54 }, { "epoch": 0.6197183098591549, "grad_norm": 13.9421968460083, "learning_rate": 3.3846153846153843e-06, "log_odds_chosen": -0.17257481813430786, "log_odds_ratio": -0.8791506886482239, "logits/chosen": -2.223339080810547, "logits/rejected": -2.183260440826416, "logps/chosen": -4.611689567565918, "logps/rejected": -4.4401044845581055, "loss": 4.3277, "nll_loss": 4.239750862121582, "rewards/accuracies": 0.375, "rewards/chosen": -0.46116903424263, "rewards/margins": -0.017158538103103638, "rewards/rejected": -0.444010466337204, "step": 55 }, { "epoch": 0.6309859154929578, "grad_norm": 11.743844032287598, "learning_rate": 3.282051282051282e-06, "log_odds_chosen": -0.662411630153656, "log_odds_ratio": -1.2490286827087402, "logits/chosen": -1.9998000860214233, "logits/rejected": -2.092249631881714, "logps/chosen": -3.955559730529785, "logps/rejected": -3.31760311126709, "loss": 3.5195, "nll_loss": 3.394559383392334, "rewards/accuracies": 0.375, "rewards/chosen": -0.3955560028553009, "rewards/margins": -0.06379567086696625, "rewards/rejected": -0.33176031708717346, "step": 56 }, { "epoch": 0.6422535211267606, "grad_norm": 8.948927879333496, "learning_rate": 3.179487179487179e-06, "log_odds_chosen": -0.0703490823507309, "log_odds_ratio": -0.8050236105918884, "logits/chosen": -2.2200632095336914, "logits/rejected": -2.263523817062378, "logps/chosen": -3.7273356914520264, "logps/rejected": -3.654778480529785, "loss": 4.0303, "nll_loss": 3.9498226642608643, "rewards/accuracies": 0.375, "rewards/chosen": -0.37273356318473816, "rewards/margins": -0.00725569948554039, "rewards/rejected": -0.36547785997390747, "step": 57 }, { "epoch": 0.6535211267605634, "grad_norm": 9.45151138305664, "learning_rate": 3.076923076923077e-06, "log_odds_chosen": -0.11648232489824295, "log_odds_ratio": -0.8264567852020264, "logits/chosen": -2.264216184616089, "logits/rejected": -2.270479202270508, "logps/chosen": -3.7268123626708984, "logps/rejected": -3.617461681365967, "loss": 4.3405, "nll_loss": 4.257853984832764, "rewards/accuracies": 0.5, "rewards/chosen": -0.37268123030662537, "rewards/margins": -0.010935075581073761, "rewards/rejected": -0.3617461919784546, "step": 58 }, { "epoch": 0.6647887323943662, "grad_norm": 13.517334938049316, "learning_rate": 2.974358974358974e-06, "log_odds_chosen": -0.04971995949745178, "log_odds_ratio": -0.8288277387619019, "logits/chosen": -2.0212604999542236, "logits/rejected": -2.118274211883545, "logps/chosen": -3.4767675399780273, "logps/rejected": -3.431030035018921, "loss": 3.3954, "nll_loss": 3.312514066696167, "rewards/accuracies": 0.625, "rewards/chosen": -0.34767675399780273, "rewards/margins": -0.004573751240968704, "rewards/rejected": -0.34310299158096313, "step": 59 }, { "epoch": 0.676056338028169, "grad_norm": 10.096856117248535, "learning_rate": 2.8717948717948717e-06, "log_odds_chosen": 0.0826490968465805, "log_odds_ratio": -0.8434144258499146, "logits/chosen": -2.0799660682678223, "logits/rejected": -2.1399242877960205, "logps/chosen": -3.7268831729888916, "logps/rejected": -3.8236966133117676, "loss": 3.7636, "nll_loss": 3.679297685623169, "rewards/accuracies": 0.5, "rewards/chosen": -0.372688353061676, "rewards/margins": 0.009681336581707, "rewards/rejected": -0.38236966729164124, "step": 60 }, { "epoch": 0.6873239436619718, "grad_norm": 11.962920188903809, "learning_rate": 2.769230769230769e-06, "log_odds_chosen": 0.11713685840368271, "log_odds_ratio": -0.7304390072822571, "logits/chosen": -2.0528836250305176, "logits/rejected": -2.1740870475769043, "logps/chosen": -3.7169508934020996, "logps/rejected": -3.827744960784912, "loss": 3.7201, "nll_loss": 3.6470143795013428, "rewards/accuracies": 0.5, "rewards/chosen": -0.3716951012611389, "rewards/margins": 0.011079397052526474, "rewards/rejected": -0.3827744722366333, "step": 61 }, { "epoch": 0.6985915492957746, "grad_norm": 9.109440803527832, "learning_rate": 2.6666666666666664e-06, "log_odds_chosen": -0.16476285457611084, "log_odds_ratio": -0.902513325214386, "logits/chosen": -2.2049403190612793, "logits/rejected": -2.266833543777466, "logps/chosen": -3.761657476425171, "logps/rejected": -3.6014232635498047, "loss": 3.9372, "nll_loss": 3.846921443939209, "rewards/accuracies": 0.5, "rewards/chosen": -0.3761657476425171, "rewards/margins": -0.016023436561226845, "rewards/rejected": -0.3601423501968384, "step": 62 }, { "epoch": 0.7098591549295775, "grad_norm": 10.382341384887695, "learning_rate": 2.5641025641025644e-06, "log_odds_chosen": -0.4335900843143463, "log_odds_ratio": -1.0794651508331299, "logits/chosen": -2.0837039947509766, "logits/rejected": -2.1245696544647217, "logps/chosen": -3.4378552436828613, "logps/rejected": -3.0201077461242676, "loss": 3.6002, "nll_loss": 3.492260456085205, "rewards/accuracies": 0.25, "rewards/chosen": -0.34378552436828613, "rewards/margins": -0.04177476093173027, "rewards/rejected": -0.30201077461242676, "step": 63 }, { "epoch": 0.7211267605633803, "grad_norm": 12.012513160705566, "learning_rate": 2.4615384615384615e-06, "log_odds_chosen": -0.21123048663139343, "log_odds_ratio": -0.8581590056419373, "logits/chosen": -2.011561155319214, "logits/rejected": -2.095082998275757, "logps/chosen": -3.706972599029541, "logps/rejected": -3.507680654525757, "loss": 3.5491, "nll_loss": 3.4632492065429688, "rewards/accuracies": 0.25, "rewards/chosen": -0.3706972599029541, "rewards/margins": -0.01992921158671379, "rewards/rejected": -0.3507680296897888, "step": 64 }, { "epoch": 0.7323943661971831, "grad_norm": 10.67837142944336, "learning_rate": 2.358974358974359e-06, "log_odds_chosen": 0.16844214498996735, "log_odds_ratio": -0.6445605754852295, "logits/chosen": -2.1053593158721924, "logits/rejected": -2.14455509185791, "logps/chosen": -3.4388720989227295, "logps/rejected": -3.601630210876465, "loss": 3.4626, "nll_loss": 3.398159980773926, "rewards/accuracies": 0.625, "rewards/chosen": -0.34388723969459534, "rewards/margins": 0.016275810077786446, "rewards/rejected": -0.36016303300857544, "step": 65 }, { "epoch": 0.7436619718309859, "grad_norm": 14.22414493560791, "learning_rate": 2.2564102564102562e-06, "log_odds_chosen": -0.837117075920105, "log_odds_ratio": -1.3559041023254395, "logits/chosen": -2.200282573699951, "logits/rejected": -2.2180466651916504, "logps/chosen": -4.45404052734375, "logps/rejected": -3.649200916290283, "loss": 4.4569, "nll_loss": 4.321296215057373, "rewards/accuracies": 0.25, "rewards/chosen": -0.4454040825366974, "rewards/margins": -0.08048398792743683, "rewards/rejected": -0.36492010951042175, "step": 66 }, { "epoch": 0.7549295774647887, "grad_norm": 12.72874641418457, "learning_rate": 2.1538461538461538e-06, "log_odds_chosen": -0.9214501976966858, "log_odds_ratio": -1.3725779056549072, "logits/chosen": -2.1959099769592285, "logits/rejected": -2.1905465126037598, "logps/chosen": -4.213864803314209, "logps/rejected": -3.309765577316284, "loss": 4.0375, "nll_loss": 3.900239944458008, "rewards/accuracies": 0.25, "rewards/chosen": -0.4213864803314209, "rewards/margins": -0.09040991961956024, "rewards/rejected": -0.33097654581069946, "step": 67 }, { "epoch": 0.7661971830985915, "grad_norm": 11.578482627868652, "learning_rate": 2.051282051282051e-06, "log_odds_chosen": -0.3243113160133362, "log_odds_ratio": -0.9195367097854614, "logits/chosen": -2.1235108375549316, "logits/rejected": -2.168004035949707, "logps/chosen": -3.7418808937072754, "logps/rejected": -3.425140142440796, "loss": 3.4577, "nll_loss": 3.365786075592041, "rewards/accuracies": 0.375, "rewards/chosen": -0.374188095331192, "rewards/margins": -0.031674087047576904, "rewards/rejected": -0.3425140380859375, "step": 68 }, { "epoch": 0.7774647887323943, "grad_norm": 12.581503868103027, "learning_rate": 1.9487179487179485e-06, "log_odds_chosen": -0.03559496998786926, "log_odds_ratio": -0.8679975271224976, "logits/chosen": -1.9304691553115845, "logits/rejected": -2.009279727935791, "logps/chosen": -3.82916522026062, "logps/rejected": -3.7897675037384033, "loss": 3.6078, "nll_loss": 3.520955801010132, "rewards/accuracies": 0.5, "rewards/chosen": -0.38291651010513306, "rewards/margins": -0.00393977016210556, "rewards/rejected": -0.3789767622947693, "step": 69 }, { "epoch": 0.7887323943661971, "grad_norm": 10.235701560974121, "learning_rate": 1.8461538461538462e-06, "log_odds_chosen": 0.7963358759880066, "log_odds_ratio": -0.4532206356525421, "logits/chosen": -2.097158908843994, "logits/rejected": -2.1584880352020264, "logps/chosen": -3.211224317550659, "logps/rejected": -3.9823241233825684, "loss": 3.4753, "nll_loss": 3.4299583435058594, "rewards/accuracies": 0.75, "rewards/chosen": -0.321122407913208, "rewards/margins": 0.07711000740528107, "rewards/rejected": -0.39823243021965027, "step": 70 }, { "epoch": 0.8, "grad_norm": 10.014704704284668, "learning_rate": 1.7435897435897436e-06, "log_odds_chosen": -0.49982893466949463, "log_odds_ratio": -1.097215175628662, "logits/chosen": -2.100175380706787, "logits/rejected": -2.1770012378692627, "logps/chosen": -3.92124080657959, "logps/rejected": -3.4407193660736084, "loss": 3.9242, "nll_loss": 3.8145065307617188, "rewards/accuracies": 0.25, "rewards/chosen": -0.39212408661842346, "rewards/margins": -0.04805214703083038, "rewards/rejected": -0.3440719544887543, "step": 71 }, { "epoch": 0.8112676056338028, "grad_norm": 11.128293991088867, "learning_rate": 1.641025641025641e-06, "log_odds_chosen": -0.03216097131371498, "log_odds_ratio": -0.7252599000930786, "logits/chosen": -1.9201287031173706, "logits/rejected": -2.0173075199127197, "logps/chosen": -3.436187744140625, "logps/rejected": -3.4034676551818848, "loss": 3.3486, "nll_loss": 3.2760744094848633, "rewards/accuracies": 0.75, "rewards/chosen": -0.3436187505722046, "rewards/margins": -0.0032719969749450684, "rewards/rejected": -0.3403467535972595, "step": 72 }, { "epoch": 0.8112676056338028, "eval_log_odds_chosen": 0.06351960450410843, "eval_log_odds_ratio": -0.7638184428215027, "eval_logits/chosen": -2.0969061851501465, "eval_logits/rejected": -2.098085880279541, "eval_logps/chosen": -3.975425958633423, "eval_logps/rejected": -4.037297248840332, "eval_loss": 4.211479187011719, "eval_nll_loss": 4.135097503662109, "eval_rewards/accuracies": 0.5, "eval_rewards/chosen": -0.3975425958633423, "eval_rewards/margins": 0.006187114864587784, "eval_rewards/rejected": -0.40372970700263977, "eval_runtime": 5.3655, "eval_samples_per_second": 1.491, "eval_steps_per_second": 0.746, "step": 72 }, { "epoch": 0.8225352112676056, "grad_norm": 13.299338340759277, "learning_rate": 1.5384615384615385e-06, "log_odds_chosen": -0.08484344184398651, "log_odds_ratio": -0.794756293296814, "logits/chosen": -1.9263019561767578, "logits/rejected": -1.9672499895095825, "logps/chosen": -3.4308979511260986, "logps/rejected": -3.347562789916992, "loss": 3.225, "nll_loss": 3.145517349243164, "rewards/accuracies": 0.625, "rewards/chosen": -0.3430897891521454, "rewards/margins": -0.008333522826433182, "rewards/rejected": -0.3347562849521637, "step": 73 }, { "epoch": 0.8338028169014085, "grad_norm": 10.876466751098633, "learning_rate": 1.4358974358974359e-06, "log_odds_chosen": -0.16019976139068604, "log_odds_ratio": -0.8210447430610657, "logits/chosen": -1.9499791860580444, "logits/rejected": -2.0576226711273193, "logps/chosen": -3.4298863410949707, "logps/rejected": -3.2793407440185547, "loss": 3.2026, "nll_loss": 3.1205201148986816, "rewards/accuracies": 0.375, "rewards/chosen": -0.34298866987228394, "rewards/margins": -0.015054592862725258, "rewards/rejected": -0.32793405652046204, "step": 74 }, { "epoch": 0.8450704225352113, "grad_norm": 11.573190689086914, "learning_rate": 1.3333333333333332e-06, "log_odds_chosen": -0.4442010223865509, "log_odds_ratio": -0.9989137053489685, "logits/chosen": -2.0589771270751953, "logits/rejected": -2.151106595993042, "logps/chosen": -3.764126777648926, "logps/rejected": -3.340824604034424, "loss": 3.6574, "nll_loss": 3.557495594024658, "rewards/accuracies": 0.25, "rewards/chosen": -0.37641265988349915, "rewards/margins": -0.042330194264650345, "rewards/rejected": -0.3340824842453003, "step": 75 }, { "epoch": 0.856338028169014, "grad_norm": 10.903554916381836, "learning_rate": 1.2307692307692308e-06, "log_odds_chosen": -0.5089474320411682, "log_odds_ratio": -1.0521657466888428, "logits/chosen": -2.061066150665283, "logits/rejected": -2.086852550506592, "logps/chosen": -3.958143711090088, "logps/rejected": -3.4646599292755127, "loss": 3.9915, "nll_loss": 3.886314868927002, "rewards/accuracies": 0.125, "rewards/chosen": -0.39581435918807983, "rewards/margins": -0.049348361790180206, "rewards/rejected": -0.3464660048484802, "step": 76 }, { "epoch": 0.8676056338028169, "grad_norm": 13.457832336425781, "learning_rate": 1.1282051282051281e-06, "log_odds_chosen": 0.047474101185798645, "log_odds_ratio": -0.7185592651367188, "logits/chosen": -1.970052719116211, "logits/rejected": -2.0962562561035156, "logps/chosen": -3.3302972316741943, "logps/rejected": -3.3730051517486572, "loss": 3.4269, "nll_loss": 3.3550257682800293, "rewards/accuracies": 0.625, "rewards/chosen": -0.33302971720695496, "rewards/margins": 0.004270805045962334, "rewards/rejected": -0.33730053901672363, "step": 77 }, { "epoch": 0.8788732394366198, "grad_norm": 13.770342826843262, "learning_rate": 1.0256410256410255e-06, "log_odds_chosen": 0.29137304425239563, "log_odds_ratio": -0.7174768447875977, "logits/chosen": -1.9147363901138306, "logits/rejected": -2.074281692504883, "logps/chosen": -3.6085116863250732, "logps/rejected": -3.9022023677825928, "loss": 3.6132, "nll_loss": 3.541477680206299, "rewards/accuracies": 0.625, "rewards/chosen": -0.3608511686325073, "rewards/margins": 0.02936907671391964, "rewards/rejected": -0.3902202248573303, "step": 78 }, { "epoch": 0.8901408450704226, "grad_norm": 11.069844245910645, "learning_rate": 9.230769230769231e-07, "log_odds_chosen": -0.5195303559303284, "log_odds_ratio": -1.1289513111114502, "logits/chosen": -2.013373851776123, "logits/rejected": -2.119136333465576, "logps/chosen": -4.341013431549072, "logps/rejected": -3.829310178756714, "loss": 4.1473, "nll_loss": 4.034441947937012, "rewards/accuracies": 0.375, "rewards/chosen": -0.43410131335258484, "rewards/margins": -0.05117032676935196, "rewards/rejected": -0.3829309940338135, "step": 79 }, { "epoch": 0.9014084507042254, "grad_norm": 11.87746810913086, "learning_rate": 8.205128205128205e-07, "log_odds_chosen": -0.13725686073303223, "log_odds_ratio": -0.8136627674102783, "logits/chosen": -1.9058239459991455, "logits/rejected": -2.0132644176483154, "logps/chosen": -3.670649528503418, "logps/rejected": -3.545762062072754, "loss": 3.5501, "nll_loss": 3.46868634223938, "rewards/accuracies": 0.25, "rewards/chosen": -0.3670649826526642, "rewards/margins": -0.01248876191675663, "rewards/rejected": -0.3545762300491333, "step": 80 }, { "epoch": 0.9126760563380282, "grad_norm": 14.643924713134766, "learning_rate": 7.179487179487179e-07, "log_odds_chosen": 0.27706030011177063, "log_odds_ratio": -0.6745563745498657, "logits/chosen": -1.9193668365478516, "logits/rejected": -2.0037789344787598, "logps/chosen": -3.511770725250244, "logps/rejected": -3.7823686599731445, "loss": 3.4836, "nll_loss": 3.416135311126709, "rewards/accuracies": 0.75, "rewards/chosen": -0.3511770963668823, "rewards/margins": 0.027059808373451233, "rewards/rejected": -0.37823688983917236, "step": 81 }, { "epoch": 0.923943661971831, "grad_norm": 11.816194534301758, "learning_rate": 6.153846153846154e-07, "log_odds_chosen": -0.5967518091201782, "log_odds_ratio": -1.1597626209259033, "logits/chosen": -2.0161638259887695, "logits/rejected": -2.067549228668213, "logps/chosen": -3.8716623783111572, "logps/rejected": -3.2700958251953125, "loss": 4.0015, "nll_loss": 3.8855390548706055, "rewards/accuracies": 0.25, "rewards/chosen": -0.38716626167297363, "rewards/margins": -0.06015666574239731, "rewards/rejected": -0.3270095884799957, "step": 82 }, { "epoch": 0.9352112676056338, "grad_norm": 11.173247337341309, "learning_rate": 5.128205128205127e-07, "log_odds_chosen": -0.45643556118011475, "log_odds_ratio": -1.0668988227844238, "logits/chosen": -2.018862009048462, "logits/rejected": -2.0817437171936035, "logps/chosen": -4.08735466003418, "logps/rejected": -3.6331050395965576, "loss": 3.8471, "nll_loss": 3.740403890609741, "rewards/accuracies": 0.5, "rewards/chosen": -0.4087355136871338, "rewards/margins": -0.04542497545480728, "rewards/rejected": -0.3633105158805847, "step": 83 }, { "epoch": 0.9464788732394366, "grad_norm": 12.033004760742188, "learning_rate": 4.1025641025641024e-07, "log_odds_chosen": -0.16924908757209778, "log_odds_ratio": -0.8661594390869141, "logits/chosen": -1.8825433254241943, "logits/rejected": -1.9633598327636719, "logps/chosen": -3.092277765274048, "logps/rejected": -2.939509868621826, "loss": 2.9276, "nll_loss": 2.840937614440918, "rewards/accuracies": 0.5, "rewards/chosen": -0.3092277944087982, "rewards/margins": -0.015276828780770302, "rewards/rejected": -0.29395097494125366, "step": 84 }, { "epoch": 0.9577464788732394, "grad_norm": 11.308173179626465, "learning_rate": 3.076923076923077e-07, "log_odds_chosen": -0.5136163830757141, "log_odds_ratio": -1.1063024997711182, "logits/chosen": -1.9425265789031982, "logits/rejected": -2.0414481163024902, "logps/chosen": -3.515744924545288, "logps/rejected": -3.021228790283203, "loss": 3.4749, "nll_loss": 3.364316940307617, "rewards/accuracies": 0.5, "rewards/chosen": -0.35157451033592224, "rewards/margins": -0.04945161193609238, "rewards/rejected": -0.30212289094924927, "step": 85 }, { "epoch": 0.9690140845070423, "grad_norm": 11.616045951843262, "learning_rate": 2.0512820512820512e-07, "log_odds_chosen": -0.21424178779125214, "log_odds_ratio": -0.8350873589515686, "logits/chosen": -1.8886396884918213, "logits/rejected": -1.9696400165557861, "logps/chosen": -3.475510358810425, "logps/rejected": -3.2696402072906494, "loss": 3.5142, "nll_loss": 3.430680990219116, "rewards/accuracies": 0.375, "rewards/chosen": -0.34755104780197144, "rewards/margins": -0.020586999133229256, "rewards/rejected": -0.32696402072906494, "step": 86 }, { "epoch": 0.9802816901408451, "grad_norm": 11.651595115661621, "learning_rate": 1.0256410256410256e-07, "log_odds_chosen": 0.28179410099983215, "log_odds_ratio": -0.655161440372467, "logits/chosen": -1.94129478931427, "logits/rejected": -1.9500186443328857, "logps/chosen": -2.9875540733337402, "logps/rejected": -3.257594108581543, "loss": 2.9425, "nll_loss": 2.8769569396972656, "rewards/accuracies": 0.625, "rewards/chosen": -0.298755407333374, "rewards/margins": 0.027004040777683258, "rewards/rejected": -0.3257594108581543, "step": 87 }, { "epoch": 0.9915492957746479, "grad_norm": 11.723467826843262, "learning_rate": 0.0, "log_odds_chosen": -0.272980272769928, "log_odds_ratio": -0.9368327856063843, "logits/chosen": -1.8292593955993652, "logits/rejected": -1.9691557884216309, "logps/chosen": -3.174346923828125, "logps/rejected": -2.9293620586395264, "loss": 2.9493, "nll_loss": 2.855593204498291, "rewards/accuracies": 0.375, "rewards/chosen": -0.3174346685409546, "rewards/margins": -0.024498462677001953, "rewards/rejected": -0.292936235666275, "step": 88 } ], "logging_steps": 1, "max_steps": 88, "num_input_tokens_seen": 0, "num_train_epochs": 1, "save_steps": 500, "stateful_callbacks": { "TrainerControl": { "args": { "should_epoch_stop": false, "should_evaluate": false, "should_log": false, "should_save": true, "should_training_stop": true }, "attributes": {} } }, "total_flos": 0.0, "train_batch_size": 2, "trial_name": null, "trial_params": null }