sigmoid_ds_chat_rmsprop_iter1 / trainer_state.json
yiran-wang3's picture
End of training
d9f44d5 verified
{
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 1.0,
"eval_steps": 100,
"global_step": 55,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"debug/policy_chosen_logits": 1.3768084049224854,
"debug/policy_chosen_logps": -161.25889587402344,
"debug/policy_rejected_logits": 1.3094204664230347,
"debug/policy_rejected_logps": -151.26834106445312,
"debug/reference_chosen_logps": -161.25889587402344,
"debug/reference_rejected_logps": -151.26834106445312,
"epoch": 0.01818181818181818,
"grad_norm": 2.5769607314165515,
"learning_rate": 5e-07,
"logits/chosen": 1.3768084049224854,
"logits/rejected": 1.3094204664230347,
"logps/chosen": -161.25889587402344,
"logps/rejected": -151.26834106445312,
"loss": 0.6931,
"rewards/accuracies": 0.0,
"rewards/chosen": 0.0,
"rewards/margins": 0.0,
"rewards/rejected": 0.0,
"step": 1
},
{
"debug/policy_chosen_logits": 0.889292299747467,
"debug/policy_chosen_logps": -130.50076293945312,
"debug/policy_rejected_logits": 1.1603094339370728,
"debug/policy_rejected_logps": -134.81027221679688,
"debug/reference_chosen_logps": -130.64517211914062,
"debug/reference_rejected_logps": -135.13914489746094,
"epoch": 0.03636363636363636,
"grad_norm": 2.472634188308473,
"learning_rate": 5e-07,
"logits/chosen": 0.889292299747467,
"logits/rejected": 1.1603094339370728,
"logps/chosen": -130.50076293945312,
"logps/rejected": -134.81027221679688,
"loss": 0.6925,
"rewards/accuracies": 0.5,
"rewards/chosen": 0.0014441299717873335,
"rewards/margins": -0.0018445965833961964,
"rewards/rejected": 0.0032887267880141735,
"step": 2
},
{
"debug/policy_chosen_logits": 0.6971374154090881,
"debug/policy_chosen_logps": -137.5977783203125,
"debug/policy_rejected_logits": 1.2003397941589355,
"debug/policy_rejected_logps": -122.20698547363281,
"debug/reference_chosen_logps": -138.00997924804688,
"debug/reference_rejected_logps": -122.32334899902344,
"epoch": 0.05454545454545454,
"grad_norm": 2.960555045135284,
"learning_rate": 5e-07,
"logits/chosen": 0.6971374154090881,
"logits/rejected": 1.2003397941589355,
"logps/chosen": -137.5977783203125,
"logps/rejected": -122.20698547363281,
"loss": 0.6908,
"rewards/accuracies": 0.5,
"rewards/chosen": 0.004121971316635609,
"rewards/margins": 0.0029583165887743235,
"rewards/rejected": 0.001163654262199998,
"step": 3
},
{
"debug/policy_chosen_logits": 0.4667843282222748,
"debug/policy_chosen_logps": -124.55257415771484,
"debug/policy_rejected_logits": 1.1448464393615723,
"debug/policy_rejected_logps": -149.21067810058594,
"debug/reference_chosen_logps": -125.31111907958984,
"debug/reference_rejected_logps": -148.5081787109375,
"epoch": 0.07272727272727272,
"grad_norm": 2.6975903263061793,
"learning_rate": 5e-07,
"logits/chosen": 0.4667843282222748,
"logits/rejected": 1.1448464393615723,
"logps/chosen": -124.55257415771484,
"logps/rejected": -149.21067810058594,
"loss": 0.6893,
"rewards/accuracies": 0.625,
"rewards/chosen": 0.007585420273244381,
"rewards/margins": 0.014610327780246735,
"rewards/rejected": -0.007024907972663641,
"step": 4
},
{
"debug/policy_chosen_logits": 0.7746908664703369,
"debug/policy_chosen_logps": -121.77815246582031,
"debug/policy_rejected_logits": 1.0104609727859497,
"debug/policy_rejected_logps": -127.0907211303711,
"debug/reference_chosen_logps": -122.27123260498047,
"debug/reference_rejected_logps": -126.88908386230469,
"epoch": 0.09090909090909091,
"grad_norm": 2.3195072990326535,
"learning_rate": 5e-07,
"logits/chosen": 0.7746908664703369,
"logits/rejected": 1.0104609727859497,
"logps/chosen": -121.77815246582031,
"logps/rejected": -127.0907211303711,
"loss": 0.6902,
"rewards/accuracies": 0.5,
"rewards/chosen": 0.004930791910737753,
"rewards/margins": 0.006947212386876345,
"rewards/rejected": -0.0020164204761385918,
"step": 5
},
{
"debug/policy_chosen_logits": 0.4436347484588623,
"debug/policy_chosen_logps": -126.17998504638672,
"debug/policy_rejected_logits": 0.6884921789169312,
"debug/policy_rejected_logps": -126.66116333007812,
"debug/reference_chosen_logps": -126.64784240722656,
"debug/reference_rejected_logps": -126.05181884765625,
"epoch": 0.10909090909090909,
"grad_norm": 2.7552991032792282,
"learning_rate": 5e-07,
"logits/chosen": 0.4436347484588623,
"logits/rejected": 0.6884921789169312,
"logps/chosen": -126.17998504638672,
"logps/rejected": -126.66116333007812,
"loss": 0.6891,
"rewards/accuracies": 0.875,
"rewards/chosen": 0.004678516648709774,
"rewards/margins": 0.01077202707529068,
"rewards/rejected": -0.006093511823564768,
"step": 6
},
{
"debug/policy_chosen_logits": 0.3949795365333557,
"debug/policy_chosen_logps": -120.72560119628906,
"debug/policy_rejected_logits": 1.16127610206604,
"debug/policy_rejected_logps": -166.54122924804688,
"debug/reference_chosen_logps": -122.05653381347656,
"debug/reference_rejected_logps": -166.2891387939453,
"epoch": 0.12727272727272726,
"grad_norm": 2.618269906477997,
"learning_rate": 5e-07,
"logits/chosen": 0.3949795365333557,
"logits/rejected": 1.16127610206604,
"logps/chosen": -120.72560119628906,
"logps/rejected": -166.54122924804688,
"loss": 0.6896,
"rewards/accuracies": 0.75,
"rewards/chosen": 0.013309288769960403,
"rewards/margins": 0.015830213204026222,
"rewards/rejected": -0.002520923502743244,
"step": 7
},
{
"debug/policy_chosen_logits": 0.7735803127288818,
"debug/policy_chosen_logps": -135.4464569091797,
"debug/policy_rejected_logits": 0.9391681551933289,
"debug/policy_rejected_logps": -139.32675170898438,
"debug/reference_chosen_logps": -136.41879272460938,
"debug/reference_rejected_logps": -139.55023193359375,
"epoch": 0.14545454545454545,
"grad_norm": 2.5281133640812117,
"learning_rate": 5e-07,
"logits/chosen": 0.7735803127288818,
"logits/rejected": 0.9391681551933289,
"logps/chosen": -135.4464569091797,
"logps/rejected": -139.32675170898438,
"loss": 0.6896,
"rewards/accuracies": 0.75,
"rewards/chosen": 0.00972352921962738,
"rewards/margins": 0.007488718256354332,
"rewards/rejected": 0.002234811894595623,
"step": 8
},
{
"debug/policy_chosen_logits": 0.5073169469833374,
"debug/policy_chosen_logps": -135.45155334472656,
"debug/policy_rejected_logits": 1.3236225843429565,
"debug/policy_rejected_logps": -130.271728515625,
"debug/reference_chosen_logps": -135.94186401367188,
"debug/reference_rejected_logps": -129.65411376953125,
"epoch": 0.16363636363636364,
"grad_norm": 2.5354121562315766,
"learning_rate": 5e-07,
"logits/chosen": 0.5073169469833374,
"logits/rejected": 1.3236225843429565,
"logps/chosen": -135.45155334472656,
"logps/rejected": -130.271728515625,
"loss": 0.6887,
"rewards/accuracies": 0.75,
"rewards/chosen": 0.004903106484562159,
"rewards/margins": 0.011079207062721252,
"rewards/rejected": -0.006176099181175232,
"step": 9
},
{
"debug/policy_chosen_logits": 0.9053974151611328,
"debug/policy_chosen_logps": -129.9651641845703,
"debug/policy_rejected_logits": 1.6513068675994873,
"debug/policy_rejected_logps": -157.7940673828125,
"debug/reference_chosen_logps": -130.6460418701172,
"debug/reference_rejected_logps": -158.63949584960938,
"epoch": 0.18181818181818182,
"grad_norm": 2.810542042119807,
"learning_rate": 5e-07,
"logits/chosen": 0.9053974151611328,
"logits/rejected": 1.6513068675994873,
"logps/chosen": -129.9651641845703,
"logps/rejected": -157.7940673828125,
"loss": 0.6851,
"rewards/accuracies": 0.375,
"rewards/chosen": 0.006808795966207981,
"rewards/margins": -0.0016453838907182217,
"rewards/rejected": 0.008454179391264915,
"step": 10
},
{
"debug/policy_chosen_logits": 0.792403519153595,
"debug/policy_chosen_logps": -141.16513061523438,
"debug/policy_rejected_logits": 1.2816684246063232,
"debug/policy_rejected_logps": -152.61949157714844,
"debug/reference_chosen_logps": -141.37376403808594,
"debug/reference_rejected_logps": -153.21987915039062,
"epoch": 0.2,
"grad_norm": 2.488423861858769,
"learning_rate": 5e-07,
"logits/chosen": 0.792403519153595,
"logits/rejected": 1.2816684246063232,
"logps/chosen": -141.16513061523438,
"logps/rejected": -152.61949157714844,
"loss": 0.6883,
"rewards/accuracies": 0.625,
"rewards/chosen": 0.0020863343961536884,
"rewards/margins": -0.00391742680221796,
"rewards/rejected": 0.0060037607327103615,
"step": 11
},
{
"debug/policy_chosen_logits": 1.0979342460632324,
"debug/policy_chosen_logps": -127.43941497802734,
"debug/policy_rejected_logits": 0.8406213521957397,
"debug/policy_rejected_logps": -133.73110961914062,
"debug/reference_chosen_logps": -128.33518981933594,
"debug/reference_rejected_logps": -133.88735961914062,
"epoch": 0.21818181818181817,
"grad_norm": 2.5700027163746975,
"learning_rate": 5e-07,
"logits/chosen": 1.0979342460632324,
"logits/rejected": 0.8406213521957397,
"logps/chosen": -127.43941497802734,
"logps/rejected": -133.73110961914062,
"loss": 0.6829,
"rewards/accuracies": 0.75,
"rewards/chosen": 0.008957691490650177,
"rewards/margins": 0.007395162247121334,
"rewards/rejected": 0.0015625286614522338,
"step": 12
},
{
"debug/policy_chosen_logits": 0.8384308218955994,
"debug/policy_chosen_logps": -129.3570556640625,
"debug/policy_rejected_logits": 1.2891650199890137,
"debug/policy_rejected_logps": -140.60781860351562,
"debug/reference_chosen_logps": -130.777099609375,
"debug/reference_rejected_logps": -137.655029296875,
"epoch": 0.23636363636363636,
"grad_norm": 2.5021787713492976,
"learning_rate": 5e-07,
"logits/chosen": 0.8384308218955994,
"logits/rejected": 1.2891650199890137,
"logps/chosen": -129.3570556640625,
"logps/rejected": -140.60781860351562,
"loss": 0.6844,
"rewards/accuracies": 1.0,
"rewards/chosen": 0.014200429432094097,
"rewards/margins": 0.043728265911340714,
"rewards/rejected": -0.029527835547924042,
"step": 13
},
{
"debug/policy_chosen_logits": 0.7953339219093323,
"debug/policy_chosen_logps": -123.55655670166016,
"debug/policy_rejected_logits": 1.151228427886963,
"debug/policy_rejected_logps": -121.413330078125,
"debug/reference_chosen_logps": -122.65122985839844,
"debug/reference_rejected_logps": -119.04695892333984,
"epoch": 0.2545454545454545,
"grad_norm": 2.645883535930893,
"learning_rate": 5e-07,
"logits/chosen": 0.7953339219093323,
"logits/rejected": 1.151228427886963,
"logps/chosen": -123.55655670166016,
"logps/rejected": -121.413330078125,
"loss": 0.69,
"rewards/accuracies": 0.625,
"rewards/chosen": -0.00905330665409565,
"rewards/margins": 0.014610443264245987,
"rewards/rejected": -0.023663748055696487,
"step": 14
},
{
"debug/policy_chosen_logits": 0.7620865106582642,
"debug/policy_chosen_logps": -142.4312286376953,
"debug/policy_rejected_logits": 0.9396656155586243,
"debug/policy_rejected_logps": -148.87515258789062,
"debug/reference_chosen_logps": -139.45565795898438,
"debug/reference_rejected_logps": -145.11453247070312,
"epoch": 0.2727272727272727,
"grad_norm": 2.9294875671025875,
"learning_rate": 5e-07,
"logits/chosen": 0.7620865106582642,
"logits/rejected": 0.9396656155586243,
"logps/chosen": -142.4312286376953,
"logps/rejected": -148.87515258789062,
"loss": 0.6889,
"rewards/accuracies": 0.75,
"rewards/chosen": -0.02975570783019066,
"rewards/margins": 0.007850446738302708,
"rewards/rejected": -0.03760614991188049,
"step": 15
},
{
"debug/policy_chosen_logits": 1.0136598348617554,
"debug/policy_chosen_logps": -130.98876953125,
"debug/policy_rejected_logits": 0.7458176612854004,
"debug/policy_rejected_logps": -134.5099639892578,
"debug/reference_chosen_logps": -131.6763153076172,
"debug/reference_rejected_logps": -133.0446014404297,
"epoch": 0.2909090909090909,
"grad_norm": 2.2028450675710602,
"learning_rate": 5e-07,
"logits/chosen": 1.0136598348617554,
"logits/rejected": 0.7458176612854004,
"logps/chosen": -130.98876953125,
"logps/rejected": -134.5099639892578,
"loss": 0.6842,
"rewards/accuracies": 0.625,
"rewards/chosen": 0.006875476334244013,
"rewards/margins": 0.021529102697968483,
"rewards/rejected": -0.014653624035418034,
"step": 16
},
{
"debug/policy_chosen_logits": 0.8773481249809265,
"debug/policy_chosen_logps": -134.43736267089844,
"debug/policy_rejected_logits": 0.8306017518043518,
"debug/policy_rejected_logps": -131.41659545898438,
"debug/reference_chosen_logps": -134.44261169433594,
"debug/reference_rejected_logps": -130.02728271484375,
"epoch": 0.3090909090909091,
"grad_norm": 2.8510973680896896,
"learning_rate": 5e-07,
"logits/chosen": 0.8773481249809265,
"logits/rejected": 0.8306017518043518,
"logps/chosen": -134.43736267089844,
"logps/rejected": -131.41659545898438,
"loss": 0.6817,
"rewards/accuracies": 0.625,
"rewards/chosen": 5.249073728919029e-05,
"rewards/margins": 0.01394571177661419,
"rewards/rejected": -0.013893223367631435,
"step": 17
},
{
"debug/policy_chosen_logits": 0.32443925738334656,
"debug/policy_chosen_logps": -117.34490966796875,
"debug/policy_rejected_logits": 0.6942044496536255,
"debug/policy_rejected_logps": -133.7316436767578,
"debug/reference_chosen_logps": -118.01707458496094,
"debug/reference_rejected_logps": -132.0999755859375,
"epoch": 0.32727272727272727,
"grad_norm": 2.9624915134206953,
"learning_rate": 5e-07,
"logits/chosen": 0.32443925738334656,
"logits/rejected": 0.6942044496536255,
"logps/chosen": -117.34490966796875,
"logps/rejected": -133.7316436767578,
"loss": 0.6789,
"rewards/accuracies": 0.5,
"rewards/chosen": 0.006721640005707741,
"rewards/margins": 0.023038387298583984,
"rewards/rejected": -0.016316747292876244,
"step": 18
},
{
"debug/policy_chosen_logits": 0.6036683320999146,
"debug/policy_chosen_logps": -139.1854248046875,
"debug/policy_rejected_logits": 1.0925521850585938,
"debug/policy_rejected_logps": -129.66262817382812,
"debug/reference_chosen_logps": -139.90301513671875,
"debug/reference_rejected_logps": -128.35787963867188,
"epoch": 0.34545454545454546,
"grad_norm": 3.0283747066503546,
"learning_rate": 5e-07,
"logits/chosen": 0.6036683320999146,
"logits/rejected": 1.0925521850585938,
"logps/chosen": -139.1854248046875,
"logps/rejected": -129.66262817382812,
"loss": 0.6771,
"rewards/accuracies": 0.875,
"rewards/chosen": 0.007175750564783812,
"rewards/margins": 0.02022336982190609,
"rewards/rejected": -0.013047618791460991,
"step": 19
},
{
"debug/policy_chosen_logits": 0.19202691316604614,
"debug/policy_chosen_logps": -125.41830444335938,
"debug/policy_rejected_logits": 0.6729949116706848,
"debug/policy_rejected_logps": -136.26675415039062,
"debug/reference_chosen_logps": -129.2891845703125,
"debug/reference_rejected_logps": -131.7703857421875,
"epoch": 0.36363636363636365,
"grad_norm": 2.5690310299484906,
"learning_rate": 5e-07,
"logits/chosen": 0.19202691316604614,
"logits/rejected": 0.6729949116706848,
"logps/chosen": -125.41830444335938,
"logps/rejected": -136.26675415039062,
"loss": 0.6813,
"rewards/accuracies": 0.875,
"rewards/chosen": 0.03870879113674164,
"rewards/margins": 0.08367244899272919,
"rewards/rejected": -0.04496365413069725,
"step": 20
},
{
"debug/policy_chosen_logits": 0.7662903070449829,
"debug/policy_chosen_logps": -124.8802261352539,
"debug/policy_rejected_logits": 1.350595474243164,
"debug/policy_rejected_logps": -131.6766357421875,
"debug/reference_chosen_logps": -125.80413818359375,
"debug/reference_rejected_logps": -129.85723876953125,
"epoch": 0.38181818181818183,
"grad_norm": 2.9390962336045523,
"learning_rate": 5e-07,
"logits/chosen": 0.7662903070449829,
"logits/rejected": 1.350595474243164,
"logps/chosen": -124.8802261352539,
"logps/rejected": -131.6766357421875,
"loss": 0.6739,
"rewards/accuracies": 0.75,
"rewards/chosen": 0.009239129722118378,
"rewards/margins": 0.027433060109615326,
"rewards/rejected": -0.018193930387496948,
"step": 21
},
{
"debug/policy_chosen_logits": 1.21824312210083,
"debug/policy_chosen_logps": -118.64605712890625,
"debug/policy_rejected_logits": 1.3743467330932617,
"debug/policy_rejected_logps": -130.24822998046875,
"debug/reference_chosen_logps": -119.62400817871094,
"debug/reference_rejected_logps": -129.10247802734375,
"epoch": 0.4,
"grad_norm": 2.4995079289310267,
"learning_rate": 5e-07,
"logits/chosen": 1.21824312210083,
"logits/rejected": 1.3743467330932617,
"logps/chosen": -118.64605712890625,
"logps/rejected": -130.24822998046875,
"loss": 0.6711,
"rewards/accuracies": 0.5,
"rewards/chosen": 0.009779386222362518,
"rewards/margins": 0.021236935630440712,
"rewards/rejected": -0.011457548476755619,
"step": 22
},
{
"debug/policy_chosen_logits": 0.32036519050598145,
"debug/policy_chosen_logps": -137.20065307617188,
"debug/policy_rejected_logits": 0.4707115590572357,
"debug/policy_rejected_logps": -141.1231231689453,
"debug/reference_chosen_logps": -137.7662353515625,
"debug/reference_rejected_logps": -140.11264038085938,
"epoch": 0.41818181818181815,
"grad_norm": 2.726999800833551,
"learning_rate": 5e-07,
"logits/chosen": 0.32036519050598145,
"logits/rejected": 0.4707115590572357,
"logps/chosen": -137.20065307617188,
"logps/rejected": -141.1231231689453,
"loss": 0.6804,
"rewards/accuracies": 0.875,
"rewards/chosen": 0.00565571803599596,
"rewards/margins": 0.01576041243970394,
"rewards/rejected": -0.010104694403707981,
"step": 23
},
{
"debug/policy_chosen_logits": 1.174080491065979,
"debug/policy_chosen_logps": -130.7235870361328,
"debug/policy_rejected_logits": 1.6448463201522827,
"debug/policy_rejected_logps": -149.97439575195312,
"debug/reference_chosen_logps": -131.06661987304688,
"debug/reference_rejected_logps": -147.88232421875,
"epoch": 0.43636363636363634,
"grad_norm": 2.652353634767768,
"learning_rate": 5e-07,
"logits/chosen": 1.174080491065979,
"logits/rejected": 1.6448463201522827,
"logps/chosen": -130.7235870361328,
"logps/rejected": -149.97439575195312,
"loss": 0.6831,
"rewards/accuracies": 0.75,
"rewards/chosen": 0.0034304135479032993,
"rewards/margins": 0.02435128018260002,
"rewards/rejected": -0.02092086710035801,
"step": 24
},
{
"debug/policy_chosen_logits": 0.5194191336631775,
"debug/policy_chosen_logps": -125.35334014892578,
"debug/policy_rejected_logits": 0.8994636535644531,
"debug/policy_rejected_logps": -129.91224670410156,
"debug/reference_chosen_logps": -127.37261962890625,
"debug/reference_rejected_logps": -127.31116485595703,
"epoch": 0.45454545454545453,
"grad_norm": 2.970456164285477,
"learning_rate": 5e-07,
"logits/chosen": 0.5194191336631775,
"logits/rejected": 0.8994636535644531,
"logps/chosen": -125.35334014892578,
"logps/rejected": -129.91224670410156,
"loss": 0.682,
"rewards/accuracies": 0.5,
"rewards/chosen": 0.02019280195236206,
"rewards/margins": 0.04620352387428284,
"rewards/rejected": -0.026010721921920776,
"step": 25
},
{
"debug/policy_chosen_logits": 0.6115962862968445,
"debug/policy_chosen_logps": -150.66409301757812,
"debug/policy_rejected_logits": 0.6578989028930664,
"debug/policy_rejected_logps": -153.54840087890625,
"debug/reference_chosen_logps": -151.63339233398438,
"debug/reference_rejected_logps": -151.82666015625,
"epoch": 0.4727272727272727,
"grad_norm": 2.580941309384753,
"learning_rate": 5e-07,
"logits/chosen": 0.6115962862968445,
"logits/rejected": 0.6578989028930664,
"logps/chosen": -150.66409301757812,
"logps/rejected": -153.54840087890625,
"loss": 0.6722,
"rewards/accuracies": 0.875,
"rewards/chosen": 0.009693136438727379,
"rewards/margins": 0.026910629123449326,
"rewards/rejected": -0.017217492684721947,
"step": 26
},
{
"debug/policy_chosen_logits": 0.41199997067451477,
"debug/policy_chosen_logps": -148.89768981933594,
"debug/policy_rejected_logits": 0.782092273235321,
"debug/policy_rejected_logps": -147.73719787597656,
"debug/reference_chosen_logps": -149.99037170410156,
"debug/reference_rejected_logps": -146.23422241210938,
"epoch": 0.4909090909090909,
"grad_norm": 2.7525802515180575,
"learning_rate": 5e-07,
"logits/chosen": 0.41199997067451477,
"logits/rejected": 0.782092273235321,
"logps/chosen": -148.89768981933594,
"logps/rejected": -147.73719787597656,
"loss": 0.6753,
"rewards/accuracies": 0.75,
"rewards/chosen": 0.010926798917353153,
"rewards/margins": 0.02595655433833599,
"rewards/rejected": -0.015029754489660263,
"step": 27
},
{
"debug/policy_chosen_logits": 0.34080997109413147,
"debug/policy_chosen_logps": -126.34262084960938,
"debug/policy_rejected_logits": 0.36349666118621826,
"debug/policy_rejected_logps": -134.22015380859375,
"debug/reference_chosen_logps": -125.65828704833984,
"debug/reference_rejected_logps": -130.0347442626953,
"epoch": 0.509090909090909,
"grad_norm": 2.680704105439303,
"learning_rate": 5e-07,
"logits/chosen": 0.34080997109413147,
"logits/rejected": 0.36349666118621826,
"logps/chosen": -126.34262084960938,
"logps/rejected": -134.22015380859375,
"loss": 0.6773,
"rewards/accuracies": 0.625,
"rewards/chosen": -0.006843318697065115,
"rewards/margins": 0.03501085191965103,
"rewards/rejected": -0.04185417294502258,
"step": 28
},
{
"debug/policy_chosen_logits": 0.7369721531867981,
"debug/policy_chosen_logps": -137.26649475097656,
"debug/policy_rejected_logits": 0.7872332334518433,
"debug/policy_rejected_logps": -143.5257568359375,
"debug/reference_chosen_logps": -137.09490966796875,
"debug/reference_rejected_logps": -140.14822387695312,
"epoch": 0.5272727272727272,
"grad_norm": 2.8512448335063185,
"learning_rate": 5e-07,
"logits/chosen": 0.7369721531867981,
"logits/rejected": 0.7872332334518433,
"logps/chosen": -137.26649475097656,
"logps/rejected": -143.5257568359375,
"loss": 0.6764,
"rewards/accuracies": 0.625,
"rewards/chosen": -0.0017157741822302341,
"rewards/margins": 0.03205941990017891,
"rewards/rejected": -0.033775195479393005,
"step": 29
},
{
"debug/policy_chosen_logits": 0.7467297911643982,
"debug/policy_chosen_logps": -142.2055206298828,
"debug/policy_rejected_logits": 0.8480007648468018,
"debug/policy_rejected_logps": -145.74615478515625,
"debug/reference_chosen_logps": -141.9608154296875,
"debug/reference_rejected_logps": -141.98846435546875,
"epoch": 0.5454545454545454,
"grad_norm": 3.2448790034806456,
"learning_rate": 5e-07,
"logits/chosen": 0.7467297911643982,
"logits/rejected": 0.8480007648468018,
"logps/chosen": -142.2055206298828,
"logps/rejected": -145.74615478515625,
"loss": 0.6655,
"rewards/accuracies": 0.625,
"rewards/chosen": -0.0024469951167702675,
"rewards/margins": 0.03513010963797569,
"rewards/rejected": -0.03757710009813309,
"step": 30
},
{
"debug/policy_chosen_logits": 0.45751431584358215,
"debug/policy_chosen_logps": -119.06462097167969,
"debug/policy_rejected_logits": 1.1633354425430298,
"debug/policy_rejected_logps": -134.27395629882812,
"debug/reference_chosen_logps": -118.83071899414062,
"debug/reference_rejected_logps": -132.58956909179688,
"epoch": 0.5636363636363636,
"grad_norm": 2.604825578516903,
"learning_rate": 5e-07,
"logits/chosen": 0.45751431584358215,
"logits/rejected": 1.1633354425430298,
"logps/chosen": -119.06462097167969,
"logps/rejected": -134.27395629882812,
"loss": 0.6774,
"rewards/accuracies": 0.875,
"rewards/chosen": -0.002338953083381057,
"rewards/margins": 0.01450504269450903,
"rewards/rejected": -0.016843995079398155,
"step": 31
},
{
"debug/policy_chosen_logits": 0.47755831480026245,
"debug/policy_chosen_logps": -113.85858154296875,
"debug/policy_rejected_logits": 0.9192254543304443,
"debug/policy_rejected_logps": -142.87522888183594,
"debug/reference_chosen_logps": -114.60741424560547,
"debug/reference_rejected_logps": -141.06182861328125,
"epoch": 0.5818181818181818,
"grad_norm": 2.4614920126129927,
"learning_rate": 5e-07,
"logits/chosen": 0.47755831480026245,
"logits/rejected": 0.9192254543304443,
"logps/chosen": -113.85858154296875,
"logps/rejected": -142.87522888183594,
"loss": 0.6833,
"rewards/accuracies": 1.0,
"rewards/chosen": 0.007488393224775791,
"rewards/margins": 0.02562214806675911,
"rewards/rejected": -0.018133753910660744,
"step": 32
},
{
"debug/policy_chosen_logits": 0.7646514773368835,
"debug/policy_chosen_logps": -140.6458740234375,
"debug/policy_rejected_logits": 1.0664713382720947,
"debug/policy_rejected_logps": -154.0317840576172,
"debug/reference_chosen_logps": -141.03515625,
"debug/reference_rejected_logps": -152.28282165527344,
"epoch": 0.6,
"grad_norm": 2.892367660321636,
"learning_rate": 5e-07,
"logits/chosen": 0.7646514773368835,
"logits/rejected": 1.0664713382720947,
"logps/chosen": -140.6458740234375,
"logps/rejected": -154.0317840576172,
"loss": 0.6726,
"rewards/accuracies": 0.75,
"rewards/chosen": 0.0038927835412323475,
"rewards/margins": 0.021382424980401993,
"rewards/rejected": -0.017489641904830933,
"step": 33
},
{
"debug/policy_chosen_logits": 1.0055601596832275,
"debug/policy_chosen_logps": -130.6605224609375,
"debug/policy_rejected_logits": 0.745191216468811,
"debug/policy_rejected_logps": -126.47428894042969,
"debug/reference_chosen_logps": -129.04339599609375,
"debug/reference_rejected_logps": -119.90361022949219,
"epoch": 0.6181818181818182,
"grad_norm": 2.9093181223713334,
"learning_rate": 5e-07,
"logits/chosen": 1.0055601596832275,
"logits/rejected": 0.745191216468811,
"logps/chosen": -130.6605224609375,
"logps/rejected": -126.47428894042969,
"loss": 0.6777,
"rewards/accuracies": 0.625,
"rewards/chosen": -0.01617136038839817,
"rewards/margins": 0.049535419791936874,
"rewards/rejected": -0.0657067745923996,
"step": 34
},
{
"debug/policy_chosen_logits": 0.665949285030365,
"debug/policy_chosen_logps": -167.20567321777344,
"debug/policy_rejected_logits": 0.8378525972366333,
"debug/policy_rejected_logps": -176.95294189453125,
"debug/reference_chosen_logps": -166.0945281982422,
"debug/reference_rejected_logps": -174.29322814941406,
"epoch": 0.6363636363636364,
"grad_norm": 3.164591716357413,
"learning_rate": 5e-07,
"logits/chosen": 0.665949285030365,
"logits/rejected": 0.8378525972366333,
"logps/chosen": -167.20567321777344,
"logps/rejected": -176.95294189453125,
"loss": 0.6686,
"rewards/accuracies": 0.875,
"rewards/chosen": -0.011111450381577015,
"rewards/margins": 0.015485725365579128,
"rewards/rejected": -0.026597173884510994,
"step": 35
},
{
"debug/policy_chosen_logits": 0.9554933905601501,
"debug/policy_chosen_logps": -136.39425659179688,
"debug/policy_rejected_logits": 0.49876803159713745,
"debug/policy_rejected_logps": -153.6390380859375,
"debug/reference_chosen_logps": -136.20895385742188,
"debug/reference_rejected_logps": -146.877197265625,
"epoch": 0.6545454545454545,
"grad_norm": 3.098324286117154,
"learning_rate": 5e-07,
"logits/chosen": 0.9554933905601501,
"logits/rejected": 0.49876803159713745,
"logps/chosen": -136.39425659179688,
"logps/rejected": -153.6390380859375,
"loss": 0.6738,
"rewards/accuracies": 0.5,
"rewards/chosen": -0.0018529421649873257,
"rewards/margins": 0.06576552242040634,
"rewards/rejected": -0.0676184669137001,
"step": 36
},
{
"debug/policy_chosen_logits": 0.2520717680454254,
"debug/policy_chosen_logps": -121.43704223632812,
"debug/policy_rejected_logits": 0.5053534507751465,
"debug/policy_rejected_logps": -158.01513671875,
"debug/reference_chosen_logps": -120.23490142822266,
"debug/reference_rejected_logps": -155.57327270507812,
"epoch": 0.6727272727272727,
"grad_norm": 3.1244720060016986,
"learning_rate": 5e-07,
"logits/chosen": 0.2520717680454254,
"logits/rejected": 0.5053534507751465,
"logps/chosen": -121.43704223632812,
"logps/rejected": -158.01513671875,
"loss": 0.6738,
"rewards/accuracies": 0.625,
"rewards/chosen": -0.012021454982459545,
"rewards/margins": 0.012397251091897488,
"rewards/rejected": -0.024418707937002182,
"step": 37
},
{
"debug/policy_chosen_logits": 0.35868680477142334,
"debug/policy_chosen_logps": -139.0185546875,
"debug/policy_rejected_logits": 0.39101022481918335,
"debug/policy_rejected_logps": -153.45266723632812,
"debug/reference_chosen_logps": -138.5296630859375,
"debug/reference_rejected_logps": -147.90438842773438,
"epoch": 0.6909090909090909,
"grad_norm": 5.757112422371425,
"learning_rate": 5e-07,
"logits/chosen": 0.35868680477142334,
"logits/rejected": 0.39101022481918335,
"logps/chosen": -139.0185546875,
"logps/rejected": -153.45266723632812,
"loss": 0.6783,
"rewards/accuracies": 0.75,
"rewards/chosen": -0.0048888688907027245,
"rewards/margins": 0.050593726336956024,
"rewards/rejected": -0.05548259615898132,
"step": 38
},
{
"debug/policy_chosen_logits": 0.6587602496147156,
"debug/policy_chosen_logps": -140.99093627929688,
"debug/policy_rejected_logits": 0.5164349675178528,
"debug/policy_rejected_logps": -140.2601318359375,
"debug/reference_chosen_logps": -136.22140502929688,
"debug/reference_rejected_logps": -129.1555633544922,
"epoch": 0.7090909090909091,
"grad_norm": 3.224110720714856,
"learning_rate": 5e-07,
"logits/chosen": 0.6587602496147156,
"logits/rejected": 0.5164349675178528,
"logps/chosen": -140.99093627929688,
"logps/rejected": -140.2601318359375,
"loss": 0.6744,
"rewards/accuracies": 0.625,
"rewards/chosen": -0.04769532382488251,
"rewards/margins": 0.0633503645658493,
"rewards/rejected": -0.11104568839073181,
"step": 39
},
{
"debug/policy_chosen_logits": 0.36872461438179016,
"debug/policy_chosen_logps": -153.08731079101562,
"debug/policy_rejected_logits": 0.7442142367362976,
"debug/policy_rejected_logps": -148.73162841796875,
"debug/reference_chosen_logps": -152.52410888671875,
"debug/reference_rejected_logps": -143.73696899414062,
"epoch": 0.7272727272727273,
"grad_norm": 3.030981160478689,
"learning_rate": 5e-07,
"logits/chosen": 0.36872461438179016,
"logits/rejected": 0.7442142367362976,
"logps/chosen": -153.08731079101562,
"logps/rejected": -148.73162841796875,
"loss": 0.6692,
"rewards/accuracies": 0.875,
"rewards/chosen": -0.0056318375281989574,
"rewards/margins": 0.044314805418252945,
"rewards/rejected": -0.04994663968682289,
"step": 40
},
{
"debug/policy_chosen_logits": 0.9497181177139282,
"debug/policy_chosen_logps": -147.80657958984375,
"debug/policy_rejected_logits": 1.0767797231674194,
"debug/policy_rejected_logps": -147.0999755859375,
"debug/reference_chosen_logps": -144.19863891601562,
"debug/reference_rejected_logps": -141.43292236328125,
"epoch": 0.7454545454545455,
"grad_norm": 2.594544171794356,
"learning_rate": 5e-07,
"logits/chosen": 0.9497181177139282,
"logits/rejected": 1.0767797231674194,
"logps/chosen": -147.80657958984375,
"logps/rejected": -147.0999755859375,
"loss": 0.6729,
"rewards/accuracies": 0.5,
"rewards/chosen": -0.0360795296728611,
"rewards/margins": 0.020590826869010925,
"rewards/rejected": -0.05667036026716232,
"step": 41
},
{
"debug/policy_chosen_logits": 0.2258227914571762,
"debug/policy_chosen_logps": -125.74639129638672,
"debug/policy_rejected_logits": 0.5640020966529846,
"debug/policy_rejected_logps": -163.89407348632812,
"debug/reference_chosen_logps": -124.42830657958984,
"debug/reference_rejected_logps": -156.05613708496094,
"epoch": 0.7636363636363637,
"grad_norm": 2.6702311112671437,
"learning_rate": 5e-07,
"logits/chosen": 0.2258227914571762,
"logits/rejected": 0.5640020966529846,
"logps/chosen": -125.74639129638672,
"logps/rejected": -163.89407348632812,
"loss": 0.6707,
"rewards/accuracies": 0.75,
"rewards/chosen": -0.013180827721953392,
"rewards/margins": 0.06519865989685059,
"rewards/rejected": -0.07837948203086853,
"step": 42
},
{
"debug/policy_chosen_logits": 0.5549467206001282,
"debug/policy_chosen_logps": -146.0997314453125,
"debug/policy_rejected_logits": 0.725266695022583,
"debug/policy_rejected_logps": -147.44781494140625,
"debug/reference_chosen_logps": -145.01519775390625,
"debug/reference_rejected_logps": -139.47506713867188,
"epoch": 0.7818181818181819,
"grad_norm": 3.292126654382054,
"learning_rate": 5e-07,
"logits/chosen": 0.5549467206001282,
"logits/rejected": 0.725266695022583,
"logps/chosen": -146.0997314453125,
"logps/rejected": -147.44781494140625,
"loss": 0.6715,
"rewards/accuracies": 0.625,
"rewards/chosen": -0.010845517739653587,
"rewards/margins": 0.06888195872306824,
"rewards/rejected": -0.07972747832536697,
"step": 43
},
{
"debug/policy_chosen_logits": -0.019000336527824402,
"debug/policy_chosen_logps": -190.69863891601562,
"debug/policy_rejected_logits": 0.052988436073064804,
"debug/policy_rejected_logps": -200.59532165527344,
"debug/reference_chosen_logps": -116.88156127929688,
"debug/reference_rejected_logps": -129.81407165527344,
"epoch": 0.8,
"grad_norm": 6.5675247662578915,
"learning_rate": 5e-07,
"logits/chosen": -0.019000336527824402,
"logits/rejected": 0.052988436073064804,
"logps/chosen": -190.69863891601562,
"logps/rejected": -200.59532165527344,
"loss": 0.6624,
"rewards/accuracies": 0.625,
"rewards/chosen": -0.7381708025932312,
"rewards/margins": -0.030358243733644485,
"rewards/rejected": -0.7078125476837158,
"step": 44
},
{
"debug/policy_chosen_logits": -0.10405106842517853,
"debug/policy_chosen_logps": -157.44500732421875,
"debug/policy_rejected_logits": 0.3257129192352295,
"debug/policy_rejected_logps": -170.44888305664062,
"debug/reference_chosen_logps": -139.64646911621094,
"debug/reference_rejected_logps": -142.53656005859375,
"epoch": 0.8181818181818182,
"grad_norm": 4.692245763612819,
"learning_rate": 5e-07,
"logits/chosen": -0.10405106842517853,
"logits/rejected": 0.3257129192352295,
"logps/chosen": -157.44500732421875,
"logps/rejected": -170.44888305664062,
"loss": 0.6487,
"rewards/accuracies": 0.625,
"rewards/chosen": -0.17798534035682678,
"rewards/margins": 0.10113789886236191,
"rewards/rejected": -0.2791232466697693,
"step": 45
},
{
"debug/policy_chosen_logits": 0.4848513901233673,
"debug/policy_chosen_logps": -132.71551513671875,
"debug/policy_rejected_logits": 0.5532675981521606,
"debug/policy_rejected_logps": -146.56289672851562,
"debug/reference_chosen_logps": -128.2412109375,
"debug/reference_rejected_logps": -136.38479614257812,
"epoch": 0.8363636363636363,
"grad_norm": 3.7539173764574945,
"learning_rate": 5e-07,
"logits/chosen": 0.4848513901233673,
"logits/rejected": 0.5532675981521606,
"logps/chosen": -132.71551513671875,
"logps/rejected": -146.56289672851562,
"loss": 0.6673,
"rewards/accuracies": 0.375,
"rewards/chosen": -0.04474309831857681,
"rewards/margins": 0.05703797936439514,
"rewards/rejected": -0.10178108513355255,
"step": 46
},
{
"debug/policy_chosen_logits": 1.0264931917190552,
"debug/policy_chosen_logps": -127.01493072509766,
"debug/policy_rejected_logits": 1.2245111465454102,
"debug/policy_rejected_logps": -139.18240356445312,
"debug/reference_chosen_logps": -120.42625427246094,
"debug/reference_rejected_logps": -133.286865234375,
"epoch": 0.8545454545454545,
"grad_norm": 3.509455580959997,
"learning_rate": 5e-07,
"logits/chosen": 1.0264931917190552,
"logits/rejected": 1.2245111465454102,
"logps/chosen": -127.01493072509766,
"logps/rejected": -139.18240356445312,
"loss": 0.6752,
"rewards/accuracies": 0.5,
"rewards/chosen": -0.06588685512542725,
"rewards/margins": -0.0069314176216721535,
"rewards/rejected": -0.05895543843507767,
"step": 47
},
{
"debug/policy_chosen_logits": 0.8054248094558716,
"debug/policy_chosen_logps": -141.4749755859375,
"debug/policy_rejected_logits": 0.9782359004020691,
"debug/policy_rejected_logps": -170.6127471923828,
"debug/reference_chosen_logps": -137.53990173339844,
"debug/reference_rejected_logps": -161.91265869140625,
"epoch": 0.8727272727272727,
"grad_norm": 2.906297231648741,
"learning_rate": 5e-07,
"logits/chosen": 0.8054248094558716,
"logits/rejected": 0.9782359004020691,
"logps/chosen": -141.4749755859375,
"logps/rejected": -170.6127471923828,
"loss": 0.6715,
"rewards/accuracies": 0.875,
"rewards/chosen": -0.039350755512714386,
"rewards/margins": 0.04765000939369202,
"rewards/rejected": -0.087000772356987,
"step": 48
},
{
"debug/policy_chosen_logits": 0.7787018418312073,
"debug/policy_chosen_logps": -144.430419921875,
"debug/policy_rejected_logits": 1.1615475416183472,
"debug/policy_rejected_logps": -154.2880859375,
"debug/reference_chosen_logps": -143.35760498046875,
"debug/reference_rejected_logps": -149.20867919921875,
"epoch": 0.8909090909090909,
"grad_norm": 3.916602695749529,
"learning_rate": 5e-07,
"logits/chosen": 0.7787018418312073,
"logits/rejected": 1.1615475416183472,
"logps/chosen": -144.430419921875,
"logps/rejected": -154.2880859375,
"loss": 0.6714,
"rewards/accuracies": 0.75,
"rewards/chosen": -0.010728321969509125,
"rewards/margins": 0.04006559029221535,
"rewards/rejected": -0.05079391598701477,
"step": 49
},
{
"debug/policy_chosen_logits": 0.4613797962665558,
"debug/policy_chosen_logps": -136.51910400390625,
"debug/policy_rejected_logits": 0.8358896970748901,
"debug/policy_rejected_logps": -144.94639587402344,
"debug/reference_chosen_logps": -134.311279296875,
"debug/reference_rejected_logps": -135.14649963378906,
"epoch": 0.9090909090909091,
"grad_norm": 3.261755038151918,
"learning_rate": 5e-07,
"logits/chosen": 0.4613797962665558,
"logits/rejected": 0.8358896970748901,
"logps/chosen": -136.51910400390625,
"logps/rejected": -144.94639587402344,
"loss": 0.6625,
"rewards/accuracies": 0.75,
"rewards/chosen": -0.022078264504671097,
"rewards/margins": 0.07592067122459412,
"rewards/rejected": -0.09799893200397491,
"step": 50
},
{
"debug/policy_chosen_logits": 0.2970171570777893,
"debug/policy_chosen_logps": -131.04600524902344,
"debug/policy_rejected_logits": 0.780690610408783,
"debug/policy_rejected_logps": -150.49591064453125,
"debug/reference_chosen_logps": -130.48240661621094,
"debug/reference_rejected_logps": -142.41004943847656,
"epoch": 0.9272727272727272,
"grad_norm": 2.7673993762098075,
"learning_rate": 5e-07,
"logits/chosen": 0.2970171570777893,
"logits/rejected": 0.780690610408783,
"logps/chosen": -131.04600524902344,
"logps/rejected": -150.49591064453125,
"loss": 0.6769,
"rewards/accuracies": 0.875,
"rewards/chosen": -0.005636042915284634,
"rewards/margins": 0.07522260397672653,
"rewards/rejected": -0.08085864782333374,
"step": 51
},
{
"debug/policy_chosen_logits": 0.5221680998802185,
"debug/policy_chosen_logps": -143.44683837890625,
"debug/policy_rejected_logits": 1.1260162591934204,
"debug/policy_rejected_logps": -153.56045532226562,
"debug/reference_chosen_logps": -140.38705444335938,
"debug/reference_rejected_logps": -145.02029418945312,
"epoch": 0.9454545454545454,
"grad_norm": 2.977638074764286,
"learning_rate": 5e-07,
"logits/chosen": 0.5221680998802185,
"logits/rejected": 1.1260162591934204,
"logps/chosen": -143.44683837890625,
"logps/rejected": -153.56045532226562,
"loss": 0.6616,
"rewards/accuracies": 0.75,
"rewards/chosen": -0.0305978674441576,
"rewards/margins": 0.05480368435382843,
"rewards/rejected": -0.08540154993534088,
"step": 52
},
{
"debug/policy_chosen_logits": 0.917246401309967,
"debug/policy_chosen_logps": -144.25767517089844,
"debug/policy_rejected_logits": 1.1883795261383057,
"debug/policy_rejected_logps": -146.60035705566406,
"debug/reference_chosen_logps": -142.2523193359375,
"debug/reference_rejected_logps": -136.43551635742188,
"epoch": 0.9636363636363636,
"grad_norm": 2.933035711917657,
"learning_rate": 5e-07,
"logits/chosen": 0.917246401309967,
"logits/rejected": 1.1883795261383057,
"logps/chosen": -144.25767517089844,
"logps/rejected": -146.60035705566406,
"loss": 0.6687,
"rewards/accuracies": 0.625,
"rewards/chosen": -0.02005346305668354,
"rewards/margins": 0.08159493654966354,
"rewards/rejected": -0.10164839029312134,
"step": 53
},
{
"debug/policy_chosen_logits": 0.44729718565940857,
"debug/policy_chosen_logps": -159.4932098388672,
"debug/policy_rejected_logits": 0.5373156666755676,
"debug/policy_rejected_logps": -145.55308532714844,
"debug/reference_chosen_logps": -154.90708923339844,
"debug/reference_rejected_logps": -141.65737915039062,
"epoch": 0.9818181818181818,
"grad_norm": 3.1066391658700105,
"learning_rate": 5e-07,
"logits/chosen": 0.44729718565940857,
"logits/rejected": 0.5373156666755676,
"logps/chosen": -159.4932098388672,
"logps/rejected": -145.55308532714844,
"loss": 0.6641,
"rewards/accuracies": 0.375,
"rewards/chosen": -0.04586116969585419,
"rewards/margins": -0.006904135458171368,
"rewards/rejected": -0.038957033306360245,
"step": 54
},
{
"debug/policy_chosen_logits": 0.4478115737438202,
"debug/policy_chosen_logps": -131.2347412109375,
"debug/policy_rejected_logits": 1.390880823135376,
"debug/policy_rejected_logps": -158.60877990722656,
"debug/reference_chosen_logps": -128.52542114257812,
"debug/reference_rejected_logps": -150.23410034179688,
"epoch": 1.0,
"grad_norm": 3.235468961904597,
"learning_rate": 5e-07,
"logits/chosen": 0.4478115737438202,
"logits/rejected": 1.390880823135376,
"logps/chosen": -131.2347412109375,
"logps/rejected": -158.60877990722656,
"loss": 0.6763,
"rewards/accuracies": 0.875,
"rewards/chosen": -0.02709309384226799,
"rewards/margins": 0.056653641164302826,
"rewards/rejected": -0.08374673128128052,
"step": 55
},
{
"epoch": 1.0,
"step": 55,
"total_flos": 0.0,
"train_loss": 0.6775240507992831,
"train_runtime": 440.7264,
"train_samples_per_second": 7.907,
"train_steps_per_second": 0.125
}
],
"logging_steps": 1,
"max_steps": 55,
"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": 8,
"trial_name": null,
"trial_params": null
}