| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 3.0, | |
| "eval_steps": 900, | |
| "global_step": 13500, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.00022222222222222223, | |
| "grad_norm": 22.625, | |
| "learning_rate": 2.222222222222222e-09, | |
| "logits/chosen": -3.4569907188415527, | |
| "logits/rejected": -3.426312208175659, | |
| "logps/chosen": -237.60638427734375, | |
| "logps/rejected": -149.21551513671875, | |
| "loss": 0.6931, | |
| "rewards/accuracies": 0.0, | |
| "rewards/chosen": 0.0, | |
| "rewards/margins": 0.0, | |
| "rewards/rejected": 0.0, | |
| "step": 1 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "grad_norm": 10.625, | |
| "learning_rate": 9.655172413793103e-07, | |
| "logits/chosen": -3.2259411811828613, | |
| "logits/rejected": -3.2373013496398926, | |
| "logps/chosen": -157.755859375, | |
| "logps/rejected": -135.1826171875, | |
| "loss": 0.7093, | |
| "rewards/accuracies": 0.5077864527702332, | |
| "rewards/chosen": -0.0009299311204813421, | |
| "rewards/margins": 0.0015665229875594378, | |
| "rewards/rejected": -0.0024964541662484407, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "eval_logits/chosen": -3.013110399246216, | |
| "eval_logits/rejected": -3.0272324085235596, | |
| "eval_logps/chosen": -157.49517822265625, | |
| "eval_logps/rejected": -135.0393829345703, | |
| "eval_loss": 0.707986056804657, | |
| "eval_rewards/accuracies": 0.5138888955116272, | |
| "eval_rewards/chosen": 0.009778406471014023, | |
| "eval_rewards/margins": 0.0047084130346775055, | |
| "eval_rewards/rejected": 0.005069993901997805, | |
| "eval_runtime": 621.3179, | |
| "eval_samples_per_second": 1.609, | |
| "eval_steps_per_second": 0.101, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "grad_norm": 9.125, | |
| "learning_rate": 8.96551724137931e-07, | |
| "logits/chosen": -3.2368197441101074, | |
| "logits/rejected": -3.2453553676605225, | |
| "logps/chosen": -156.8467254638672, | |
| "logps/rejected": -132.40830993652344, | |
| "loss": 0.7103, | |
| "rewards/accuracies": 0.49944445490837097, | |
| "rewards/chosen": 0.0005948151228949428, | |
| "rewards/margins": -6.089528324082494e-05, | |
| "rewards/rejected": 0.0006557104061357677, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "eval_logits/chosen": -3.013019323348999, | |
| "eval_logits/rejected": -3.0271592140197754, | |
| "eval_logps/chosen": -157.4865264892578, | |
| "eval_logps/rejected": -135.03961181640625, | |
| "eval_loss": 0.7029208540916443, | |
| "eval_rewards/accuracies": 0.5059523582458496, | |
| "eval_rewards/chosen": 0.01844022050499916, | |
| "eval_rewards/margins": 0.013589809648692608, | |
| "eval_rewards/rejected": 0.004850410390645266, | |
| "eval_runtime": 618.8158, | |
| "eval_samples_per_second": 1.616, | |
| "eval_steps_per_second": 0.102, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "grad_norm": 9.5625, | |
| "learning_rate": 8.275862068965517e-07, | |
| "logits/chosen": -3.228179693222046, | |
| "logits/rejected": -3.2370007038116455, | |
| "logps/chosen": -155.81236267089844, | |
| "logps/rejected": -132.52044677734375, | |
| "loss": 0.7057, | |
| "rewards/accuracies": 0.49444442987442017, | |
| "rewards/chosen": 0.01380106434226036, | |
| "rewards/margins": 0.007251843344420195, | |
| "rewards/rejected": 0.006549220532178879, | |
| "step": 2700 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "eval_logits/chosen": -3.012927293777466, | |
| "eval_logits/rejected": -3.0271010398864746, | |
| "eval_logps/chosen": -157.47120666503906, | |
| "eval_logps/rejected": -135.03591918945312, | |
| "eval_loss": 0.6978173851966858, | |
| "eval_rewards/accuracies": 0.5128968358039856, | |
| "eval_rewards/chosen": 0.0337512232363224, | |
| "eval_rewards/margins": 0.025207480415701866, | |
| "eval_rewards/rejected": 0.008543744683265686, | |
| "eval_runtime": 616.9849, | |
| "eval_samples_per_second": 1.621, | |
| "eval_steps_per_second": 0.102, | |
| "step": 2700 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "grad_norm": 9.4375, | |
| "learning_rate": 7.586206896551724e-07, | |
| "logits/chosen": -3.2351443767547607, | |
| "logits/rejected": -3.2431368827819824, | |
| "logps/chosen": -156.77496337890625, | |
| "logps/rejected": -133.58193969726562, | |
| "loss": 0.7032, | |
| "rewards/accuracies": 0.5188888907432556, | |
| "rewards/chosen": 0.02398551069200039, | |
| "rewards/margins": 0.01356814056634903, | |
| "rewards/rejected": 0.01041737012565136, | |
| "step": 3600 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "eval_logits/chosen": -3.013091802597046, | |
| "eval_logits/rejected": -3.0272812843322754, | |
| "eval_logps/chosen": -157.4769744873047, | |
| "eval_logps/rejected": -135.029052734375, | |
| "eval_loss": 0.7038853764533997, | |
| "eval_rewards/accuracies": 0.5, | |
| "eval_rewards/chosen": 0.02797560952603817, | |
| "eval_rewards/margins": 0.01257056463509798, | |
| "eval_rewards/rejected": 0.015405045822262764, | |
| "eval_runtime": 617.4916, | |
| "eval_samples_per_second": 1.619, | |
| "eval_steps_per_second": 0.102, | |
| "step": 3600 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "grad_norm": 15.4375, | |
| "learning_rate": 6.896551724137931e-07, | |
| "logits/chosen": -3.2322237491607666, | |
| "logits/rejected": -3.2422921657562256, | |
| "logps/chosen": -158.39981079101562, | |
| "logps/rejected": -134.93020629882812, | |
| "loss": 0.7054, | |
| "rewards/accuracies": 0.5116666555404663, | |
| "rewards/chosen": 0.03937062993645668, | |
| "rewards/margins": 0.009623720310628414, | |
| "rewards/rejected": 0.02974691055715084, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_logits/chosen": -3.013054609298706, | |
| "eval_logits/rejected": -3.0272459983825684, | |
| "eval_logps/chosen": -157.4499053955078, | |
| "eval_logps/rejected": -135.0231170654297, | |
| "eval_loss": 0.6935561299324036, | |
| "eval_rewards/accuracies": 0.538690447807312, | |
| "eval_rewards/chosen": 0.05505846440792084, | |
| "eval_rewards/margins": 0.03372044861316681, | |
| "eval_rewards/rejected": 0.021338019520044327, | |
| "eval_runtime": 618.0893, | |
| "eval_samples_per_second": 1.618, | |
| "eval_steps_per_second": 0.102, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "grad_norm": 10.25, | |
| "learning_rate": 6.206896551724138e-07, | |
| "logits/chosen": -3.23718523979187, | |
| "logits/rejected": -3.246558666229248, | |
| "logps/chosen": -157.82701110839844, | |
| "logps/rejected": -135.48013305664062, | |
| "loss": 0.7033, | |
| "rewards/accuracies": 0.5244444608688354, | |
| "rewards/chosen": 0.043098967522382736, | |
| "rewards/margins": 0.014880964532494545, | |
| "rewards/rejected": 0.028218001127243042, | |
| "step": 5400 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "eval_logits/chosen": -3.01303768157959, | |
| "eval_logits/rejected": -3.0271825790405273, | |
| "eval_logps/chosen": -157.44830322265625, | |
| "eval_logps/rejected": -135.01748657226562, | |
| "eval_loss": 0.6953737139701843, | |
| "eval_rewards/accuracies": 0.528769850730896, | |
| "eval_rewards/chosen": 0.05664879083633423, | |
| "eval_rewards/margins": 0.029674744233489037, | |
| "eval_rewards/rejected": 0.02697405032813549, | |
| "eval_runtime": 619.6669, | |
| "eval_samples_per_second": 1.614, | |
| "eval_steps_per_second": 0.102, | |
| "step": 5400 | |
| }, | |
| { | |
| "epoch": 1.4, | |
| "grad_norm": 15.375, | |
| "learning_rate": 5.517241379310344e-07, | |
| "logits/chosen": -3.2300009727478027, | |
| "logits/rejected": -3.239253044128418, | |
| "logps/chosen": -157.90548706054688, | |
| "logps/rejected": -133.99058532714844, | |
| "loss": 0.7052, | |
| "rewards/accuracies": 0.5138888955116272, | |
| "rewards/chosen": 0.043052662163972855, | |
| "rewards/margins": 0.010972147807478905, | |
| "rewards/rejected": 0.0320805124938488, | |
| "step": 6300 | |
| }, | |
| { | |
| "epoch": 1.4, | |
| "eval_logits/chosen": -3.0130200386047363, | |
| "eval_logits/rejected": -3.0271594524383545, | |
| "eval_logps/chosen": -157.44041442871094, | |
| "eval_logps/rejected": -135.01556396484375, | |
| "eval_loss": 0.6925241351127625, | |
| "eval_rewards/accuracies": 0.5396825671195984, | |
| "eval_rewards/chosen": 0.06454168260097504, | |
| "eval_rewards/margins": 0.03564530983567238, | |
| "eval_rewards/rejected": 0.02889636531472206, | |
| "eval_runtime": 619.0036, | |
| "eval_samples_per_second": 1.615, | |
| "eval_steps_per_second": 0.102, | |
| "step": 6300 | |
| }, | |
| { | |
| "epoch": 1.6, | |
| "grad_norm": 13.8125, | |
| "learning_rate": 4.827586206896552e-07, | |
| "logits/chosen": -3.227038860321045, | |
| "logits/rejected": -3.237276792526245, | |
| "logps/chosen": -155.9982452392578, | |
| "logps/rejected": -133.17002868652344, | |
| "loss": 0.6945, | |
| "rewards/accuracies": 0.5294444561004639, | |
| "rewards/chosen": 0.06239644065499306, | |
| "rewards/margins": 0.032080113887786865, | |
| "rewards/rejected": 0.030316324904561043, | |
| "step": 7200 | |
| }, | |
| { | |
| "epoch": 1.6, | |
| "eval_logits/chosen": -3.0130317211151123, | |
| "eval_logits/rejected": -3.0272297859191895, | |
| "eval_logps/chosen": -157.42637634277344, | |
| "eval_logps/rejected": -135.01864624023438, | |
| "eval_loss": 0.6855846643447876, | |
| "eval_rewards/accuracies": 0.5664682388305664, | |
| "eval_rewards/chosen": 0.07857942581176758, | |
| "eval_rewards/margins": 0.05277930945158005, | |
| "eval_rewards/rejected": 0.025800110772252083, | |
| "eval_runtime": 617.3442, | |
| "eval_samples_per_second": 1.62, | |
| "eval_steps_per_second": 0.102, | |
| "step": 7200 | |
| }, | |
| { | |
| "epoch": 1.8, | |
| "grad_norm": 7.625, | |
| "learning_rate": 4.1379310344827586e-07, | |
| "logits/chosen": -3.233175754547119, | |
| "logits/rejected": -3.2411041259765625, | |
| "logps/chosen": -156.3334503173828, | |
| "logps/rejected": -131.23414611816406, | |
| "loss": 0.7011, | |
| "rewards/accuracies": 0.5183333158493042, | |
| "rewards/chosen": 0.058114588260650635, | |
| "rewards/margins": 0.01674296148121357, | |
| "rewards/rejected": 0.041371628642082214, | |
| "step": 8100 | |
| }, | |
| { | |
| "epoch": 1.8, | |
| "eval_logits/chosen": -3.01287579536438, | |
| "eval_logits/rejected": -3.027003288269043, | |
| "eval_logps/chosen": -157.41868591308594, | |
| "eval_logps/rejected": -135.01287841796875, | |
| "eval_loss": 0.6833264827728271, | |
| "eval_rewards/accuracies": 0.5605158805847168, | |
| "eval_rewards/chosen": 0.08628029376268387, | |
| "eval_rewards/margins": 0.05471265688538551, | |
| "eval_rewards/rejected": 0.031567640602588654, | |
| "eval_runtime": 617.0438, | |
| "eval_samples_per_second": 1.621, | |
| "eval_steps_per_second": 0.102, | |
| "step": 8100 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "grad_norm": 6.0, | |
| "learning_rate": 3.4482758620689656e-07, | |
| "logits/chosen": -3.2326903343200684, | |
| "logits/rejected": -3.241744041442871, | |
| "logps/chosen": -157.41111755371094, | |
| "logps/rejected": -134.63653564453125, | |
| "loss": 0.6943, | |
| "rewards/accuracies": 0.5327777862548828, | |
| "rewards/chosen": 0.07450559735298157, | |
| "rewards/margins": 0.033994805067777634, | |
| "rewards/rejected": 0.04051079601049423, | |
| "step": 9000 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_logits/chosen": -3.012874126434326, | |
| "eval_logits/rejected": -3.027069091796875, | |
| "eval_logps/chosen": -157.4245147705078, | |
| "eval_logps/rejected": -135.01060485839844, | |
| "eval_loss": 0.6874573826789856, | |
| "eval_rewards/accuracies": 0.5416666865348816, | |
| "eval_rewards/chosen": 0.08044072240591049, | |
| "eval_rewards/margins": 0.04658132046461105, | |
| "eval_rewards/rejected": 0.033859409391880035, | |
| "eval_runtime": 617.9074, | |
| "eval_samples_per_second": 1.618, | |
| "eval_steps_per_second": 0.102, | |
| "step": 9000 | |
| }, | |
| { | |
| "epoch": 2.2, | |
| "grad_norm": 11.0625, | |
| "learning_rate": 2.758620689655172e-07, | |
| "logits/chosen": -3.2308294773101807, | |
| "logits/rejected": -3.241821765899658, | |
| "logps/chosen": -156.53921508789062, | |
| "logps/rejected": -133.07757568359375, | |
| "loss": 0.7004, | |
| "rewards/accuracies": 0.5172222256660461, | |
| "rewards/chosen": 0.058275189250707626, | |
| "rewards/margins": 0.019040387123823166, | |
| "rewards/rejected": 0.03923480585217476, | |
| "step": 9900 | |
| }, | |
| { | |
| "epoch": 2.2, | |
| "eval_logits/chosen": -3.012793779373169, | |
| "eval_logits/rejected": -3.026986837387085, | |
| "eval_logps/chosen": -157.421142578125, | |
| "eval_logps/rejected": -135.0035400390625, | |
| "eval_loss": 0.6900457739830017, | |
| "eval_rewards/accuracies": 0.5476190447807312, | |
| "eval_rewards/chosen": 0.08382151275873184, | |
| "eval_rewards/margins": 0.04290608689188957, | |
| "eval_rewards/rejected": 0.04091542959213257, | |
| "eval_runtime": 617.2416, | |
| "eval_samples_per_second": 1.62, | |
| "eval_steps_per_second": 0.102, | |
| "step": 9900 | |
| }, | |
| { | |
| "epoch": 2.4, | |
| "grad_norm": 9.75, | |
| "learning_rate": 2.0689655172413793e-07, | |
| "logits/chosen": -3.2308058738708496, | |
| "logits/rejected": -3.2401750087738037, | |
| "logps/chosen": -159.013427734375, | |
| "logps/rejected": -133.7303924560547, | |
| "loss": 0.6946, | |
| "rewards/accuracies": 0.5322222113609314, | |
| "rewards/chosen": 0.07445600628852844, | |
| "rewards/margins": 0.03202791139483452, | |
| "rewards/rejected": 0.04242809861898422, | |
| "step": 10800 | |
| }, | |
| { | |
| "epoch": 2.4, | |
| "eval_logits/chosen": -3.0131027698516846, | |
| "eval_logits/rejected": -3.0273118019104004, | |
| "eval_logps/chosen": -157.4429473876953, | |
| "eval_logps/rejected": -135.01034545898438, | |
| "eval_loss": 0.6979788541793823, | |
| "eval_rewards/accuracies": 0.538690447807312, | |
| "eval_rewards/chosen": 0.06201785430312157, | |
| "eval_rewards/margins": 0.027890633791685104, | |
| "eval_rewards/rejected": 0.03412722051143646, | |
| "eval_runtime": 617.1551, | |
| "eval_samples_per_second": 1.62, | |
| "eval_steps_per_second": 0.102, | |
| "step": 10800 | |
| }, | |
| { | |
| "epoch": 2.6, | |
| "grad_norm": 9.4375, | |
| "learning_rate": 1.379310344827586e-07, | |
| "logits/chosen": -3.236497640609741, | |
| "logits/rejected": -3.243530035018921, | |
| "logps/chosen": -157.6371307373047, | |
| "logps/rejected": -134.19334411621094, | |
| "loss": 0.6898, | |
| "rewards/accuracies": 0.5533333420753479, | |
| "rewards/chosen": 0.0833386555314064, | |
| "rewards/margins": 0.043027978390455246, | |
| "rewards/rejected": 0.04031067714095116, | |
| "step": 11700 | |
| }, | |
| { | |
| "epoch": 2.6, | |
| "eval_logits/chosen": -3.0131278038024902, | |
| "eval_logits/rejected": -3.0272884368896484, | |
| "eval_logps/chosen": -157.4271240234375, | |
| "eval_logps/rejected": -135.00433349609375, | |
| "eval_loss": 0.6915609240531921, | |
| "eval_rewards/accuracies": 0.5426587462425232, | |
| "eval_rewards/chosen": 0.07781906425952911, | |
| "eval_rewards/margins": 0.03770728409290314, | |
| "eval_rewards/rejected": 0.04011178016662598, | |
| "eval_runtime": 617.4019, | |
| "eval_samples_per_second": 1.62, | |
| "eval_steps_per_second": 0.102, | |
| "step": 11700 | |
| }, | |
| { | |
| "epoch": 2.8, | |
| "grad_norm": 12.875, | |
| "learning_rate": 6.89655172413793e-08, | |
| "logits/chosen": -3.2290825843811035, | |
| "logits/rejected": -3.238636016845703, | |
| "logps/chosen": -155.427978515625, | |
| "logps/rejected": -132.6368865966797, | |
| "loss": 0.6911, | |
| "rewards/accuracies": 0.5477777719497681, | |
| "rewards/chosen": 0.07133558392524719, | |
| "rewards/margins": 0.04007010534405708, | |
| "rewards/rejected": 0.03126548230648041, | |
| "step": 12600 | |
| }, | |
| { | |
| "epoch": 2.8, | |
| "eval_logits/chosen": -3.0130114555358887, | |
| "eval_logits/rejected": -3.0271997451782227, | |
| "eval_logps/chosen": -157.41952514648438, | |
| "eval_logps/rejected": -135.01426696777344, | |
| "eval_loss": 0.6846425533294678, | |
| "eval_rewards/accuracies": 0.5436508059501648, | |
| "eval_rewards/chosen": 0.0854325145483017, | |
| "eval_rewards/margins": 0.055243875831365585, | |
| "eval_rewards/rejected": 0.030188636854290962, | |
| "eval_runtime": 617.22, | |
| "eval_samples_per_second": 1.62, | |
| "eval_steps_per_second": 0.102, | |
| "step": 12600 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "grad_norm": 10.0, | |
| "learning_rate": 0.0, | |
| "logits/chosen": -3.229140520095825, | |
| "logits/rejected": -3.2376339435577393, | |
| "logps/chosen": -156.7906036376953, | |
| "logps/rejected": -134.84449768066406, | |
| "loss": 0.7003, | |
| "rewards/accuracies": 0.528333306312561, | |
| "rewards/chosen": 0.06633351743221283, | |
| "rewards/margins": 0.02000141702592373, | |
| "rewards/rejected": 0.04633209481835365, | |
| "step": 13500 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_logits/chosen": -3.0129826068878174, | |
| "eval_logits/rejected": -3.027216911315918, | |
| "eval_logps/chosen": -157.42532348632812, | |
| "eval_logps/rejected": -135.01596069335938, | |
| "eval_loss": 0.6864064931869507, | |
| "eval_rewards/accuracies": 0.5426587462425232, | |
| "eval_rewards/chosen": 0.07963090389966965, | |
| "eval_rewards/margins": 0.051132071763277054, | |
| "eval_rewards/rejected": 0.028498835861682892, | |
| "eval_runtime": 617.15, | |
| "eval_samples_per_second": 1.62, | |
| "eval_steps_per_second": 0.102, | |
| "step": 13500 | |
| } | |
| ], | |
| "logging_steps": 900, | |
| "max_steps": 13500, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 3, | |
| "save_steps": 900, | |
| "total_flos": 0.0, | |
| "train_batch_size": 2, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |