Files
ModelHub XC c6a30a8d26 初始化项目,由ModelHub XC社区提供模型
Model: HCY123902/llama-3-8b-inst-dpo-on-p-tw31-beta-2.5e-0-ift
Source: Original Platform
2026-05-22 13:24:16 +08:00

1128 lines
44 KiB
JSON

{
"best_global_step": null,
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 0.26721891909947226,
"eval_steps": 500,
"global_step": 500,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.0005344378381989445,
"grad_norm": 31.37949562072754,
"learning_rate": 0.0,
"logits/chosen": -0.2252655029296875,
"logits/rejected": -0.192626953125,
"logps/chosen": -110.828125,
"logps/rejected": -115.515625,
"logps/weighted_chosen": -0.31903076171875,
"logps/weighted_rejected": -0.333709716796875,
"loss": 0.6914,
"rewards/accuracies": 0.0,
"rewards/chosen": 0.0,
"rewards/margins": 0.0,
"rewards/rejected": 0.0,
"rewards/weighted_accuracies": 0.0,
"rewards/weighted_chosen": 0.0,
"rewards/weighted_margins": 0.0,
"rewards/weighted_rejected": 0.0,
"step": 1
},
{
"epoch": 0.005344378381989445,
"grad_norm": 75.76042938232422,
"learning_rate": 4.787234042553191e-08,
"logits/chosen": -0.3284708559513092,
"logits/rejected": -0.3214448392391205,
"logps/chosen": -134.46006774902344,
"logps/rejected": -138.06597900390625,
"logps/weighted_chosen": -0.3890923261642456,
"logps/weighted_rejected": -0.4067721962928772,
"loss": 0.6953,
"rewards/accuracies": 0.2951388955116272,
"rewards/chosen": -0.2072482705116272,
"rewards/margins": -0.1790364533662796,
"rewards/rejected": -0.0282118059694767,
"rewards/weighted_accuracies": 0.3472222089767456,
"rewards/weighted_chosen": -0.0032717387657612562,
"rewards/weighted_margins": -0.0047516291961073875,
"rewards/weighted_rejected": 0.0014813741436228156,
"step": 10
},
{
"epoch": 0.01068875676397889,
"grad_norm": 24.94420623779297,
"learning_rate": 1.0106382978723404e-07,
"logits/chosen": -0.2780090272426605,
"logits/rejected": -0.2689048647880554,
"logps/chosen": -115.3070297241211,
"logps/rejected": -114.8101577758789,
"logps/weighted_chosen": -0.354583740234375,
"logps/weighted_rejected": -0.36929017305374146,
"loss": 0.6925,
"rewards/accuracies": 0.3687500059604645,
"rewards/chosen": 0.0087890625,
"rewards/margins": -0.02734375,
"rewards/rejected": 0.0361328125,
"rewards/weighted_accuracies": 0.4312500059604645,
"rewards/weighted_chosen": 0.0016719817649573088,
"rewards/weighted_margins": 0.0011638641590252519,
"rewards/weighted_rejected": 0.0005052566411904991,
"step": 20
},
{
"epoch": 0.016033135145968335,
"grad_norm": 26.90618133544922,
"learning_rate": 1.5425531914893615e-07,
"logits/chosen": -0.26707762479782104,
"logits/rejected": -0.2697288393974304,
"logps/chosen": -122.49687194824219,
"logps/rejected": -128.2218780517578,
"logps/weighted_chosen": -0.364663690328598,
"logps/weighted_rejected": -0.40430909395217896,
"loss": 0.6919,
"rewards/accuracies": 0.3687500059604645,
"rewards/chosen": 0.099609375,
"rewards/margins": 0.2177734375,
"rewards/rejected": -0.1181640625,
"rewards/weighted_accuracies": 0.4468750059604645,
"rewards/weighted_chosen": 0.0011037830263376236,
"rewards/weighted_margins": 0.0029600143898278475,
"rewards/weighted_rejected": -0.0018524170154705644,
"step": 30
},
{
"epoch": 0.02137751352795778,
"grad_norm": 19.056455612182617,
"learning_rate": 2.074468085106383e-07,
"logits/chosen": -0.31552428007125854,
"logits/rejected": -0.309671014547348,
"logps/chosen": -126.5132827758789,
"logps/rejected": -127.7515640258789,
"logps/weighted_chosen": -0.3717803955078125,
"logps/weighted_rejected": -0.36720579862594604,
"loss": 0.6927,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": 0.041015625,
"rewards/margins": 0.150390625,
"rewards/rejected": -0.109375,
"rewards/weighted_accuracies": 0.44062501192092896,
"rewards/weighted_chosen": 0.0009314537164755166,
"rewards/weighted_margins": 0.0009433746454305947,
"rewards/weighted_rejected": -1.831054760259576e-05,
"step": 40
},
{
"epoch": 0.026721891909947223,
"grad_norm": 94.1146469116211,
"learning_rate": 2.6063829787234044e-07,
"logits/chosen": -0.2799697816371918,
"logits/rejected": -0.2664199769496918,
"logps/chosen": -120.34375,
"logps/rejected": -120.19062805175781,
"logps/weighted_chosen": -0.367788702249527,
"logps/weighted_rejected": -0.37299805879592896,
"loss": 0.6934,
"rewards/accuracies": 0.3499999940395355,
"rewards/chosen": -0.01806640625,
"rewards/margins": -0.11865234375,
"rewards/rejected": 0.1005859375,
"rewards/weighted_accuracies": 0.421875,
"rewards/weighted_chosen": 0.0011091709602624178,
"rewards/weighted_margins": -0.0005058288807049394,
"rewards/weighted_rejected": 0.00161571498028934,
"step": 50
},
{
"epoch": 0.03206627029193667,
"grad_norm": 47.161922454833984,
"learning_rate": 3.1382978723404253e-07,
"logits/chosen": -0.22172394394874573,
"logits/rejected": -0.2157600373029709,
"logps/chosen": -115.8382797241211,
"logps/rejected": -118.75859069824219,
"logps/weighted_chosen": -0.36602783203125,
"logps/weighted_rejected": -0.369253545999527,
"loss": 0.6908,
"rewards/accuracies": 0.3499999940395355,
"rewards/chosen": -0.10546875,
"rewards/margins": -0.1259765625,
"rewards/rejected": 0.0205078125,
"rewards/weighted_accuracies": 0.4000000059604645,
"rewards/weighted_chosen": 0.008862781338393688,
"rewards/weighted_margins": 0.00811080913990736,
"rewards/weighted_rejected": 0.0007405281066894531,
"step": 60
},
{
"epoch": 0.037410648673926114,
"grad_norm": 42.38877868652344,
"learning_rate": 3.6702127659574467e-07,
"logits/chosen": -0.299722284078598,
"logits/rejected": -0.29665374755859375,
"logps/chosen": -114.1656265258789,
"logps/rejected": -118.0765609741211,
"logps/weighted_chosen": -0.35313719511032104,
"logps/weighted_rejected": -0.3739013671875,
"loss": 0.6949,
"rewards/accuracies": 0.375,
"rewards/chosen": -0.0693359375,
"rewards/margins": -0.0458984375,
"rewards/rejected": -0.0234375,
"rewards/weighted_accuracies": 0.390625,
"rewards/weighted_chosen": 0.0011390686267986894,
"rewards/weighted_margins": -0.0028884888160973787,
"rewards/weighted_rejected": 0.0040260315872728825,
"step": 70
},
{
"epoch": 0.04275502705591556,
"grad_norm": 35.451927185058594,
"learning_rate": 4.202127659574468e-07,
"logits/chosen": -0.30262452363967896,
"logits/rejected": -0.24024733901023865,
"logps/chosen": -112.3812484741211,
"logps/rejected": -111.8375015258789,
"logps/weighted_chosen": -0.37481385469436646,
"logps/weighted_rejected": -0.38435667753219604,
"loss": 0.694,
"rewards/accuracies": 0.390625,
"rewards/chosen": -0.1552734375,
"rewards/margins": -0.0341796875,
"rewards/rejected": -0.12109375,
"rewards/weighted_accuracies": 0.48750001192092896,
"rewards/weighted_chosen": 0.01006317138671875,
"rewards/weighted_margins": 0.0036018372047692537,
"rewards/weighted_rejected": 0.006462156772613525,
"step": 80
},
{
"epoch": 0.048099405437905,
"grad_norm": 42.264678955078125,
"learning_rate": 4.734042553191489e-07,
"logits/chosen": -0.2891853451728821,
"logits/rejected": -0.23835448920726776,
"logps/chosen": -118.0296859741211,
"logps/rejected": -116.484375,
"logps/weighted_chosen": -0.376077264547348,
"logps/weighted_rejected": -0.3848114013671875,
"loss": 0.6935,
"rewards/accuracies": 0.4625000059604645,
"rewards/chosen": 0.0458984375,
"rewards/margins": 0.267578125,
"rewards/rejected": -0.2216796875,
"rewards/weighted_accuracies": 0.46562498807907104,
"rewards/weighted_chosen": 0.016840171068906784,
"rewards/weighted_margins": 0.014923477545380592,
"rewards/weighted_rejected": 0.00187511439435184,
"step": 90
},
{
"epoch": 0.053443783819894446,
"grad_norm": 36.87267303466797,
"learning_rate": 5.26595744680851e-07,
"logits/chosen": -0.3333755433559418,
"logits/rejected": -0.28821104764938354,
"logps/chosen": -124.0367202758789,
"logps/rejected": -124.33906555175781,
"logps/weighted_chosen": -0.39268797636032104,
"logps/weighted_rejected": -0.4093261659145355,
"loss": 0.6875,
"rewards/accuracies": 0.44062501192092896,
"rewards/chosen": -0.0302734375,
"rewards/margins": 0.36835938692092896,
"rewards/rejected": -0.39863282442092896,
"rewards/weighted_accuracies": 0.484375,
"rewards/weighted_chosen": 0.012582575902342796,
"rewards/weighted_margins": 0.026942063122987747,
"rewards/weighted_rejected": -0.014329910278320312,
"step": 100
},
{
"epoch": 0.058788162201883896,
"grad_norm": 17.8848876953125,
"learning_rate": 5.797872340425531e-07,
"logits/chosen": -0.3335327208042145,
"logits/rejected": -0.32384032011032104,
"logps/chosen": -117.6968765258789,
"logps/rejected": -119.85859680175781,
"logps/weighted_chosen": -0.35866087675094604,
"logps/weighted_rejected": -0.37585145235061646,
"loss": 0.7015,
"rewards/accuracies": 0.3968749940395355,
"rewards/chosen": -0.524609386920929,
"rewards/margins": -0.01328125037252903,
"rewards/rejected": -0.511523425579071,
"rewards/weighted_accuracies": 0.4281249940395355,
"rewards/weighted_chosen": -0.0057319640181958675,
"rewards/weighted_margins": 0.0012493133544921875,
"rewards/weighted_rejected": -0.0070056915283203125,
"step": 110
},
{
"epoch": 0.06413254058387334,
"grad_norm": 37.190059661865234,
"learning_rate": 6.329787234042553e-07,
"logits/chosen": -0.29607391357421875,
"logits/rejected": -0.2735137939453125,
"logps/chosen": -120.81718444824219,
"logps/rejected": -127.04219055175781,
"logps/weighted_chosen": -0.41831666231155396,
"logps/weighted_rejected": -0.42036741971969604,
"loss": 0.7443,
"rewards/accuracies": 0.42500001192092896,
"rewards/chosen": -0.43964844942092896,
"rewards/margins": 0.690234363079071,
"rewards/rejected": -1.1298828125,
"rewards/weighted_accuracies": 0.45625001192092896,
"rewards/weighted_chosen": -0.07387389987707138,
"rewards/weighted_margins": -0.04258232191205025,
"rewards/weighted_rejected": -0.031409453600645065,
"step": 120
},
{
"epoch": 0.06947691896586278,
"grad_norm": 22.76742172241211,
"learning_rate": 6.861702127659574e-07,
"logits/chosen": -0.30709609389305115,
"logits/rejected": -0.29381561279296875,
"logps/chosen": -120.3140640258789,
"logps/rejected": -121.73515319824219,
"logps/weighted_chosen": -0.38916015625,
"logps/weighted_rejected": -0.38067322969436646,
"loss": 0.6906,
"rewards/accuracies": 0.4312500059604645,
"rewards/chosen": -0.8837890625,
"rewards/margins": 0.4273437559604645,
"rewards/rejected": -1.310937523841858,
"rewards/weighted_accuracies": 0.46875,
"rewards/weighted_chosen": -0.0019147873390465975,
"rewards/weighted_margins": 0.025261688977479935,
"rewards/weighted_rejected": -0.027143806219100952,
"step": 130
},
{
"epoch": 0.07482129734785223,
"grad_norm": 25.72498321533203,
"learning_rate": 7.393617021276596e-07,
"logits/chosen": -0.2590804994106293,
"logits/rejected": -0.25146180391311646,
"logps/chosen": -114.2992172241211,
"logps/rejected": -118.26094055175781,
"logps/weighted_chosen": -0.376434326171875,
"logps/weighted_rejected": -0.39961546659469604,
"loss": 0.6968,
"rewards/accuracies": 0.4906249940395355,
"rewards/chosen": -1.7428710460662842,
"rewards/margins": 1.043554663658142,
"rewards/rejected": -2.7867188453674316,
"rewards/weighted_accuracies": 0.46562498807907104,
"rewards/weighted_chosen": -0.01796722412109375,
"rewards/weighted_margins": 0.044054411351680756,
"rewards/weighted_rejected": -0.062059782445430756,
"step": 140
},
{
"epoch": 0.08016567572984168,
"grad_norm": 27.819217681884766,
"learning_rate": 7.925531914893616e-07,
"logits/chosen": -0.260824590921402,
"logits/rejected": -0.2456924468278885,
"logps/chosen": -118.3109359741211,
"logps/rejected": -116.1898422241211,
"logps/weighted_chosen": -0.3584175109863281,
"logps/weighted_rejected": -0.389230340719223,
"loss": 0.693,
"rewards/accuracies": 0.518750011920929,
"rewards/chosen": -2.796679735183716,
"rewards/margins": 1.470312476158142,
"rewards/rejected": -4.267773628234863,
"rewards/weighted_accuracies": 0.5406249761581421,
"rewards/weighted_chosen": 0.02875671349465847,
"rewards/weighted_margins": 0.08138389885425568,
"rewards/weighted_rejected": -0.05276889726519585,
"step": 150
},
{
"epoch": 0.08551005411183112,
"grad_norm": 20.792280197143555,
"learning_rate": 8.457446808510637e-07,
"logits/chosen": -0.27181702852249146,
"logits/rejected": -0.26198044419288635,
"logps/chosen": -119.7906265258789,
"logps/rejected": -120.09687805175781,
"logps/weighted_chosen": -0.3836608827114105,
"logps/weighted_rejected": -0.40611571073532104,
"loss": 0.6795,
"rewards/accuracies": 0.503125011920929,
"rewards/chosen": -3.6748046875,
"rewards/margins": 1.8689453601837158,
"rewards/rejected": -5.542382717132568,
"rewards/weighted_accuracies": 0.543749988079071,
"rewards/weighted_chosen": -0.013747024349868298,
"rewards/weighted_margins": 0.10787200927734375,
"rewards/weighted_rejected": -0.12159118801355362,
"step": 160
},
{
"epoch": 0.09085443249382057,
"grad_norm": 28.162086486816406,
"learning_rate": 8.989361702127659e-07,
"logits/chosen": -0.30354803800582886,
"logits/rejected": -0.28291016817092896,
"logps/chosen": -118.81172180175781,
"logps/rejected": -123.3851547241211,
"logps/weighted_chosen": -0.364785760641098,
"logps/weighted_rejected": -0.40638428926467896,
"loss": 0.7104,
"rewards/accuracies": 0.528124988079071,
"rewards/chosen": -4.450781345367432,
"rewards/margins": 1.540624976158142,
"rewards/rejected": -5.989843845367432,
"rewards/weighted_accuracies": 0.49687498807907104,
"rewards/weighted_chosen": -0.05272483825683594,
"rewards/weighted_margins": 0.03959999233484268,
"rewards/weighted_rejected": -0.0922950729727745,
"step": 170
},
{
"epoch": 0.09619881087581,
"grad_norm": 62.7450065612793,
"learning_rate": 9.52127659574468e-07,
"logits/chosen": -0.3086685240268707,
"logits/rejected": -0.29756468534469604,
"logps/chosen": -120.9000015258789,
"logps/rejected": -121.6031265258789,
"logps/weighted_chosen": -0.4021057188510895,
"logps/weighted_rejected": -0.43016356229782104,
"loss": 0.6902,
"rewards/accuracies": 0.5062500238418579,
"rewards/chosen": -5.364843845367432,
"rewards/margins": 0.95654296875,
"rewards/rejected": -6.323437690734863,
"rewards/weighted_accuracies": 0.515625,
"rewards/weighted_chosen": -0.07762374728918076,
"rewards/weighted_margins": 0.042090605944395065,
"rewards/weighted_rejected": -0.11983337253332138,
"step": 180
},
{
"epoch": 0.10154318925779945,
"grad_norm": 107.93684387207031,
"learning_rate": 9.99999129927477e-07,
"logits/chosen": -0.2998809814453125,
"logits/rejected": -0.2729034423828125,
"logps/chosen": -121.68046569824219,
"logps/rejected": -122.90547180175781,
"logps/weighted_chosen": -0.4000488221645355,
"logps/weighted_rejected": -0.49153441190719604,
"loss": 0.6846,
"rewards/accuracies": 0.4906249940395355,
"rewards/chosen": -7.084570407867432,
"rewards/margins": 1.763671875,
"rewards/rejected": -8.850390434265137,
"rewards/weighted_accuracies": 0.5625,
"rewards/weighted_chosen": -0.06708984076976776,
"rewards/weighted_margins": 0.19685058295726776,
"rewards/weighted_rejected": -0.2636512815952301,
"step": 190
},
{
"epoch": 0.10688756763978889,
"grad_norm": 42.62810516357422,
"learning_rate": 9.99894724888679e-07,
"logits/chosen": -0.2994216978549957,
"logits/rejected": -0.2775813937187195,
"logps/chosen": -129.10311889648438,
"logps/rejected": -127.71015930175781,
"logps/weighted_chosen": -0.40317994356155396,
"logps/weighted_rejected": -0.4718689024448395,
"loss": 0.7052,
"rewards/accuracies": 0.4937500059604645,
"rewards/chosen": -8.422460556030273,
"rewards/margins": 1.215234398841858,
"rewards/rejected": -9.638280868530273,
"rewards/weighted_accuracies": 0.518750011920929,
"rewards/weighted_chosen": -0.10227356106042862,
"rewards/weighted_margins": 0.08837012946605682,
"rewards/weighted_rejected": -0.19074249267578125,
"step": 200
},
{
"epoch": 0.11223194602177834,
"grad_norm": 56.36786651611328,
"learning_rate": 9.996163469793475e-07,
"logits/chosen": -0.3200393617153168,
"logits/rejected": -0.28080445528030396,
"logps/chosen": -127.65625,
"logps/rejected": -122.5406265258789,
"logps/weighted_chosen": -0.402487188577652,
"logps/weighted_rejected": -0.4784179627895355,
"loss": 0.6855,
"rewards/accuracies": 0.559374988079071,
"rewards/chosen": -9.696874618530273,
"rewards/margins": 3.138671875,
"rewards/rejected": -12.8359375,
"rewards/weighted_accuracies": 0.559374988079071,
"rewards/weighted_chosen": -0.07228164374828339,
"rewards/weighted_margins": 0.1660926789045334,
"rewards/weighted_rejected": -0.23847046494483948,
"step": 210
},
{
"epoch": 0.11757632440376779,
"grad_norm": 24.45851707458496,
"learning_rate": 9.991640930802883e-07,
"logits/chosen": -0.30699461698532104,
"logits/rejected": -0.3066558837890625,
"logps/chosen": -125.8734359741211,
"logps/rejected": -129.6999969482422,
"logps/weighted_chosen": -0.41710203886032104,
"logps/weighted_rejected": -0.4827117919921875,
"loss": 0.6884,
"rewards/accuracies": 0.546875,
"rewards/chosen": -12.277539253234863,
"rewards/margins": 2.571484327316284,
"rewards/rejected": -14.851171493530273,
"rewards/weighted_accuracies": 0.5218750238418579,
"rewards/weighted_chosen": -0.12371826171875,
"rewards/weighted_margins": 0.1390731781721115,
"rewards/weighted_rejected": -0.26273268461227417,
"step": 220
},
{
"epoch": 0.12292070278575723,
"grad_norm": 27.215944290161133,
"learning_rate": 9.98538120584459e-07,
"logits/chosen": -0.3107505738735199,
"logits/rejected": -0.283193975687027,
"logps/chosen": -134.8015594482422,
"logps/rejected": -138.5890655517578,
"logps/weighted_chosen": -0.4462524354457855,
"logps/weighted_rejected": -0.511853039264679,
"loss": 0.6939,
"rewards/accuracies": 0.5562499761581421,
"rewards/chosen": -15.346875190734863,
"rewards/margins": 2.830273389816284,
"rewards/rejected": -18.179492950439453,
"rewards/weighted_accuracies": 0.5562499761581421,
"rewards/weighted_chosen": -0.1530204713344574,
"rewards/weighted_margins": 0.1789344847202301,
"rewards/weighted_rejected": -0.3320491909980774,
"step": 230
},
{
"epoch": 0.12826508116774668,
"grad_norm": 17.173702239990234,
"learning_rate": 9.977386473421917e-07,
"logits/chosen": -0.27986279129981995,
"logits/rejected": -0.2772073745727539,
"logps/chosen": -121.0453109741211,
"logps/rejected": -124.65312194824219,
"logps/weighted_chosen": -0.4528869688510895,
"logps/weighted_rejected": -0.501416027545929,
"loss": 0.7222,
"rewards/accuracies": 0.5687500238418579,
"rewards/chosen": -18.59375,
"rewards/margins": 2.5132813453674316,
"rewards/rejected": -21.106054306030273,
"rewards/weighted_accuracies": 0.578125,
"rewards/weighted_chosen": -0.2343955934047699,
"rewards/weighted_margins": 0.07334060966968536,
"rewards/weighted_rejected": -0.30777662992477417,
"step": 240
},
{
"epoch": 0.13360945954973613,
"grad_norm": 24.67556381225586,
"learning_rate": 9.96765951585378e-07,
"logits/chosen": -0.262664794921875,
"logits/rejected": -0.24544373154640198,
"logps/chosen": -121.7421875,
"logps/rejected": -126.42500305175781,
"logps/weighted_chosen": -0.46705931425094604,
"logps/weighted_rejected": -0.547576904296875,
"loss": 0.6683,
"rewards/accuracies": 0.5625,
"rewards/chosen": -21.9228515625,
"rewards/margins": 2.5238280296325684,
"rewards/rejected": -24.447071075439453,
"rewards/weighted_accuracies": 0.5687500238418579,
"rewards/weighted_chosen": -0.2455238401889801,
"rewards/weighted_margins": 0.1756332367658615,
"rewards/weighted_rejected": -0.4210983216762543,
"step": 250
},
{
"epoch": 0.13895383793172555,
"grad_norm": 33.55352783203125,
"learning_rate": 9.956203718306388e-07,
"logits/chosen": -0.18781813979148865,
"logits/rejected": -0.156982421875,
"logps/chosen": -127.8578109741211,
"logps/rejected": -132.35546875,
"logps/weighted_chosen": -0.4968322813510895,
"logps/weighted_rejected": -0.5148254632949829,
"loss": 0.7213,
"rewards/accuracies": 0.5531250238418579,
"rewards/chosen": -26.109960556030273,
"rewards/margins": 2.660937547683716,
"rewards/rejected": -28.761327743530273,
"rewards/weighted_accuracies": 0.534375011920929,
"rewards/weighted_chosen": -0.29717254638671875,
"rewards/weighted_margins": 0.04170074313879013,
"rewards/weighted_rejected": -0.3386779725551605,
"step": 260
},
{
"epoch": 0.144298216313715,
"grad_norm": 242.56521606445312,
"learning_rate": 9.943023067615136e-07,
"logits/chosen": -0.17297974228858948,
"logits/rejected": -0.1584724485874176,
"logps/chosen": -139.23046875,
"logps/rejected": -142.3390655517578,
"logps/weighted_chosen": -0.48270875215530396,
"logps/weighted_rejected": -0.560772716999054,
"loss": 0.7059,
"rewards/accuracies": 0.5375000238418579,
"rewards/chosen": -33.55195236206055,
"rewards/margins": 3.075976610183716,
"rewards/rejected": -36.6328125,
"rewards/weighted_accuracies": 0.4937500059604645,
"rewards/weighted_chosen": -0.321615606546402,
"rewards/weighted_margins": 0.13364562392234802,
"rewards/weighted_rejected": -0.45517730712890625,
"step": 270
},
{
"epoch": 0.14964259469570446,
"grad_norm": 16.8142147064209,
"learning_rate": 9.928122150897112e-07,
"logits/chosen": -0.21183013916015625,
"logits/rejected": -0.168986514210701,
"logps/chosen": -130.66250610351562,
"logps/rejected": -132.0812530517578,
"logps/weighted_chosen": -0.48672789335250854,
"logps/weighted_rejected": -0.578961193561554,
"loss": 0.6614,
"rewards/accuracies": 0.590624988079071,
"rewards/chosen": -34.4287109375,
"rewards/margins": 3.7144532203674316,
"rewards/rejected": -38.127342224121094,
"rewards/weighted_accuracies": 0.5687500238418579,
"rewards/weighted_chosen": -0.28246229887008667,
"rewards/weighted_margins": 0.23281364142894745,
"rewards/weighted_rejected": -0.5149310827255249,
"step": 280
},
{
"epoch": 0.1549869730776939,
"grad_norm": 17.110448837280273,
"learning_rate": 9.9115061539547e-07,
"logits/chosen": -0.20588979125022888,
"logits/rejected": -0.18258285522460938,
"logps/chosen": -141.0203094482422,
"logps/rejected": -142.50625610351562,
"logps/weighted_chosen": -0.4897003173828125,
"logps/weighted_rejected": -0.5554351806640625,
"loss": 0.7051,
"rewards/accuracies": 0.574999988079071,
"rewards/chosen": -38.395896911621094,
"rewards/margins": 3.002734422683716,
"rewards/rejected": -41.394920349121094,
"rewards/weighted_accuracies": 0.5406249761581421,
"rewards/weighted_chosen": -0.3098343014717102,
"rewards/weighted_margins": 0.1397857666015625,
"rewards/weighted_rejected": -0.4498863220214844,
"step": 290
},
{
"epoch": 0.16033135145968336,
"grad_norm": 21.960878372192383,
"learning_rate": 9.893180859470818e-07,
"logits/chosen": -0.1905662566423416,
"logits/rejected": -0.16956177353858948,
"logps/chosen": -131.609375,
"logps/rejected": -133.328125,
"logps/weighted_chosen": -0.5007995367050171,
"logps/weighted_rejected": -0.5999816656112671,
"loss": 0.6676,
"rewards/accuracies": 0.559374988079071,
"rewards/chosen": -38.5810546875,
"rewards/margins": 5.082226753234863,
"rewards/rejected": -43.662498474121094,
"rewards/weighted_accuracies": 0.590624988079071,
"rewards/weighted_chosen": -0.27146607637405396,
"rewards/weighted_margins": 0.2570602297782898,
"rewards/weighted_rejected": -0.5287536382675171,
"step": 300
},
{
"epoch": 0.16567572984167278,
"grad_norm": 45.954952239990234,
"learning_rate": 9.873152644996424e-07,
"logits/chosen": -0.23566055297851562,
"logits/rejected": -0.23574523627758026,
"logps/chosen": -134.5734405517578,
"logps/rejected": -136.5500030517578,
"logps/weighted_chosen": -0.565338134765625,
"logps/weighted_rejected": -0.620849609375,
"loss": 0.7314,
"rewards/accuracies": 0.59375,
"rewards/chosen": -44.349021911621094,
"rewards/margins": 6.559765815734863,
"rewards/rejected": -50.908592224121094,
"rewards/weighted_accuracies": 0.606249988079071,
"rewards/weighted_chosen": -0.44298553466796875,
"rewards/weighted_margins": 0.1432647705078125,
"rewards/weighted_rejected": -0.5857940912246704,
"step": 310
},
{
"epoch": 0.17102010822366223,
"grad_norm": 22.280086517333984,
"learning_rate": 9.85142848073103e-07,
"logits/chosen": -0.2385093718767166,
"logits/rejected": -0.21721191704273224,
"logps/chosen": -138.27499389648438,
"logps/rejected": -138.13516235351562,
"logps/weighted_chosen": -0.5242675542831421,
"logps/weighted_rejected": -0.591705322265625,
"loss": 0.6903,
"rewards/accuracies": 0.543749988079071,
"rewards/chosen": -48.820899963378906,
"rewards/margins": 1.562109351158142,
"rewards/rejected": -50.394920349121094,
"rewards/weighted_accuracies": 0.5718749761581421,
"rewards/weighted_chosen": -0.3802032470703125,
"rewards/weighted_margins": 0.14343567192554474,
"rewards/weighted_rejected": -0.523608386516571,
"step": 320
},
{
"epoch": 0.17636448660565168,
"grad_norm": 19.125673294067383,
"learning_rate": 9.828015927096914e-07,
"logits/chosen": -0.2693939208984375,
"logits/rejected": -0.23780974745750427,
"logps/chosen": -148.56875610351562,
"logps/rejected": -147.24453735351562,
"logps/weighted_chosen": -0.521636962890625,
"logps/weighted_rejected": -0.5547729730606079,
"loss": 0.6703,
"rewards/accuracies": 0.528124988079071,
"rewards/chosen": -50.613868713378906,
"rewards/margins": 3.5126953125,
"rewards/rejected": -54.113670349121094,
"rewards/weighted_accuracies": 0.550000011920929,
"rewards/weighted_chosen": -0.321145623922348,
"rewards/weighted_margins": 0.18524780869483948,
"rewards/weighted_rejected": -0.5064395666122437,
"step": 330
},
{
"epoch": 0.18170886498764113,
"grad_norm": 14.343570709228516,
"learning_rate": 9.802923132107968e-07,
"logits/chosen": -0.25108033418655396,
"logits/rejected": -0.2313240021467209,
"logps/chosen": -149.78671264648438,
"logps/rejected": -153.48046875,
"logps/weighted_chosen": -0.562329113483429,
"logps/weighted_rejected": -0.583233654499054,
"loss": 0.741,
"rewards/accuracies": 0.5562499761581421,
"rewards/chosen": -58.06855392456055,
"rewards/margins": 4.334374904632568,
"rewards/rejected": -62.40312576293945,
"rewards/weighted_accuracies": 0.512499988079071,
"rewards/weighted_chosen": -0.47095948457717896,
"rewards/weighted_margins": 0.02723388746380806,
"rewards/weighted_rejected": -0.498382568359375,
"step": 340
},
{
"epoch": 0.18705324336963056,
"grad_norm": 66.20745849609375,
"learning_rate": 9.776158828534024e-07,
"logits/chosen": -0.2837265133857727,
"logits/rejected": -0.2537124752998352,
"logps/chosen": -149.5359344482422,
"logps/rejected": -194.4765625,
"logps/weighted_chosen": -0.540771484375,
"logps/weighted_rejected": -0.589801013469696,
"loss": 0.6779,
"rewards/accuracies": 0.612500011920929,
"rewards/chosen": -59.75273513793945,
"rewards/margins": 7.256640434265137,
"rewards/rejected": -66.99531555175781,
"rewards/weighted_accuracies": 0.5531250238418579,
"rewards/weighted_chosen": -0.4245468080043793,
"rewards/weighted_margins": 0.146717831492424,
"rewards/weighted_rejected": -0.571148693561554,
"step": 350
},
{
"epoch": 0.19239762175162,
"grad_norm": 13.690327644348145,
"learning_rate": 9.747732330861695e-07,
"logits/chosen": -0.18397827446460724,
"logits/rejected": -0.15533828735351562,
"logps/chosen": -143.4562530517578,
"logps/rejected": -145.44686889648438,
"logps/weighted_chosen": -0.5499817132949829,
"logps/weighted_rejected": -0.6820312738418579,
"loss": 0.6644,
"rewards/accuracies": 0.581250011920929,
"rewards/chosen": -63.58203125,
"rewards/margins": 7.447851657867432,
"rewards/rejected": -71.0308609008789,
"rewards/weighted_accuracies": 0.6031249761581421,
"rewards/weighted_chosen": -0.43936461210250854,
"rewards/weighted_margins": 0.3239502012729645,
"rewards/weighted_rejected": -0.7634918093681335,
"step": 360
},
{
"epoch": 0.19774200013360946,
"grad_norm": 20.364688873291016,
"learning_rate": 9.717653532052742e-07,
"logits/chosen": -0.16991272568702698,
"logits/rejected": -0.16076354682445526,
"logps/chosen": -137.11172485351562,
"logps/rejected": -146.09375,
"logps/weighted_chosen": -0.609234631061554,
"logps/weighted_rejected": -0.658831775188446,
"loss": 0.7108,
"rewards/accuracies": 0.606249988079071,
"rewards/chosen": -63.02363204956055,
"rewards/margins": 9.1494140625,
"rewards/rejected": -72.1488265991211,
"rewards/weighted_accuracies": 0.59375,
"rewards/weighted_chosen": -0.5135604739189148,
"rewards/weighted_margins": 0.1833236664533615,
"rewards/weighted_rejected": -0.6966766119003296,
"step": 370
},
{
"epoch": 0.2030863785155989,
"grad_norm": 15.459892272949219,
"learning_rate": 9.685932900101146e-07,
"logits/chosen": -0.17396697402000427,
"logits/rejected": -0.1525276154279709,
"logps/chosen": -143.62655639648438,
"logps/rejected": -147.03515625,
"logps/weighted_chosen": -0.584460437297821,
"logps/weighted_rejected": -0.642120361328125,
"loss": 0.7159,
"rewards/accuracies": 0.590624988079071,
"rewards/chosen": -68.6537094116211,
"rewards/margins": 5.928124904632568,
"rewards/rejected": -74.57109069824219,
"rewards/weighted_accuracies": 0.59375,
"rewards/weighted_chosen": -0.528277575969696,
"rewards/weighted_margins": 0.11352996528148651,
"rewards/weighted_rejected": -0.6417190432548523,
"step": 380
},
{
"epoch": 0.20843075689758836,
"grad_norm": 25.710723876953125,
"learning_rate": 9.652581474390043e-07,
"logits/chosen": -0.17167052626609802,
"logits/rejected": -0.14335784316062927,
"logps/chosen": -147.3156280517578,
"logps/rejected": -151.1484375,
"logps/weighted_chosen": -0.5953735113143921,
"logps/weighted_rejected": -0.649151623249054,
"loss": 0.7014,
"rewards/accuracies": 0.59375,
"rewards/chosen": -70.93769836425781,
"rewards/margins": 9.441015243530273,
"rewards/rejected": -80.35664367675781,
"rewards/weighted_accuracies": 0.612500011920929,
"rewards/weighted_chosen": -0.5777953863143921,
"rewards/weighted_margins": 0.10624237358570099,
"rewards/weighted_rejected": -0.6836212277412415,
"step": 390
},
{
"epoch": 0.21377513527957778,
"grad_norm": 42.97126007080078,
"learning_rate": 9.61761086184981e-07,
"logits/chosen": -0.192851260304451,
"logits/rejected": -0.16070251166820526,
"logps/chosen": -148.39688110351562,
"logps/rejected": -149.6046905517578,
"logps/weighted_chosen": -0.6315368413925171,
"logps/weighted_rejected": -0.7087768316268921,
"loss": 0.7113,
"rewards/accuracies": 0.565625011920929,
"rewards/chosen": -73.91288757324219,
"rewards/margins": 6.164453029632568,
"rewards/rejected": -80.05976867675781,
"rewards/weighted_accuracies": 0.559374988079071,
"rewards/weighted_chosen": -0.6400848627090454,
"rewards/weighted_margins": 0.11194305121898651,
"rewards/weighted_rejected": -0.752105712890625,
"step": 400
},
{
"epoch": 0.21911951366156723,
"grad_norm": 17.99481773376465,
"learning_rate": 9.581033232918629e-07,
"logits/chosen": -0.14135894179344177,
"logits/rejected": -0.11229457706212997,
"logps/chosen": -145.88827514648438,
"logps/rejected": -149.74063110351562,
"logps/weighted_chosen": -0.6018310785293579,
"logps/weighted_rejected": -0.7620849609375,
"loss": 0.6764,
"rewards/accuracies": 0.6031249761581421,
"rewards/chosen": -74.9345703125,
"rewards/margins": 8.443944931030273,
"rewards/rejected": -83.36601257324219,
"rewards/weighted_accuracies": 0.578125,
"rewards/weighted_chosen": -0.578625500202179,
"rewards/weighted_margins": 0.35536497831344604,
"rewards/weighted_rejected": -0.9342681765556335,
"step": 410
},
{
"epoch": 0.22446389204355668,
"grad_norm": 19.236024856567383,
"learning_rate": 9.542861317306952e-07,
"logits/chosen": -0.1445457488298416,
"logits/rejected": -0.1324237883090973,
"logps/chosen": -148.76718139648438,
"logps/rejected": -150.97811889648438,
"logps/weighted_chosen": -0.5991576910018921,
"logps/weighted_rejected": -0.674072265625,
"loss": 0.6735,
"rewards/accuracies": 0.5562499761581421,
"rewards/chosen": -80.28242492675781,
"rewards/margins": 4.427929878234863,
"rewards/rejected": -84.7109375,
"rewards/weighted_accuracies": 0.559374988079071,
"rewards/weighted_chosen": -0.543652355670929,
"rewards/weighted_margins": 0.184315487742424,
"rewards/weighted_rejected": -0.727569580078125,
"step": 420
},
{
"epoch": 0.22980827042554614,
"grad_norm": 14.300553321838379,
"learning_rate": 9.503108399567308e-07,
"logits/chosen": -0.14830398559570312,
"logits/rejected": -0.09484557807445526,
"logps/chosen": -162.40625,
"logps/rejected": -167.7195281982422,
"logps/weighted_chosen": -0.5840820074081421,
"logps/weighted_rejected": -0.647021472454071,
"loss": 0.672,
"rewards/accuracies": 0.5687500238418579,
"rewards/chosen": -89.32051086425781,
"rewards/margins": 8.4365234375,
"rewards/rejected": -97.79023742675781,
"rewards/weighted_accuracies": 0.596875011920929,
"rewards/weighted_chosen": -0.545318603515625,
"rewards/weighted_margins": 0.13585510849952698,
"rewards/weighted_rejected": -0.681243896484375,
"step": 430
},
{
"epoch": 0.23515264880753559,
"grad_norm": 21.087541580200195,
"learning_rate": 9.461788314471034e-07,
"logits/chosen": -0.10236664116382599,
"logits/rejected": -0.05696678161621094,
"logps/chosen": -155.13827514648438,
"logps/rejected": -159.828125,
"logps/weighted_chosen": -0.6568237543106079,
"logps/weighted_rejected": -0.7309814691543579,
"loss": 0.6898,
"rewards/accuracies": 0.5562499761581421,
"rewards/chosen": -92.36328125,
"rewards/margins": 7.942968845367432,
"rewards/rejected": -100.3109359741211,
"rewards/weighted_accuracies": 0.559374988079071,
"rewards/weighted_chosen": -0.6760101318359375,
"rewards/weighted_margins": 0.16096191108226776,
"rewards/weighted_rejected": -0.836883544921875,
"step": 440
},
{
"epoch": 0.240497027189525,
"grad_norm": 19.655607223510742,
"learning_rate": 9.418915442193509e-07,
"logits/chosen": -0.10150299221277237,
"logits/rejected": -0.05734825134277344,
"logps/chosen": -152.95858764648438,
"logps/rejected": -165.30078125,
"logps/weighted_chosen": -0.6426635980606079,
"logps/weighted_rejected": -0.69622802734375,
"loss": 0.7073,
"rewards/accuracies": 0.612500011920929,
"rewards/chosen": -89.14433288574219,
"rewards/margins": 14.668359756469727,
"rewards/rejected": -103.8412094116211,
"rewards/weighted_accuracies": 0.565625011920929,
"rewards/weighted_chosen": -0.694122314453125,
"rewards/weighted_margins": 0.11862488090991974,
"rewards/weighted_rejected": -0.8125030398368835,
"step": 450
},
{
"epoch": 0.24584140557151446,
"grad_norm": 35.64816665649414,
"learning_rate": 9.374504703309579e-07,
"logits/chosen": -0.1730697602033615,
"logits/rejected": -0.1376514434814453,
"logps/chosen": -149.97811889648438,
"logps/rejected": -152.8171844482422,
"logps/weighted_chosen": -0.638531506061554,
"logps/weighted_rejected": -0.77386474609375,
"loss": 0.6563,
"rewards/accuracies": 0.5687500238418579,
"rewards/chosen": -86.3251953125,
"rewards/margins": 8.056055068969727,
"rewards/rejected": -94.4281234741211,
"rewards/weighted_accuracies": 0.5625,
"rewards/weighted_chosen": -0.6749175786972046,
"rewards/weighted_margins": 0.30900877714157104,
"rewards/weighted_rejected": -0.983563244342804,
"step": 460
},
{
"epoch": 0.2511857839535039,
"grad_norm": 20.12373161315918,
"learning_rate": 9.328571553600915e-07,
"logits/chosen": -0.14519290626049042,
"logits/rejected": -0.11081619560718536,
"logps/chosen": -154.73828125,
"logps/rejected": -158.703125,
"logps/weighted_chosen": -0.656390368938446,
"logps/weighted_rejected": -0.7307983636856079,
"loss": 0.7214,
"rewards/accuracies": 0.5843750238418579,
"rewards/chosen": -92.1263656616211,
"rewards/margins": 7.731054782867432,
"rewards/rejected": -99.86836242675781,
"rewards/weighted_accuracies": 0.565625011920929,
"rewards/weighted_chosen": -0.6984283328056335,
"rewards/weighted_margins": 0.14166870713233948,
"rewards/weighted_rejected": -0.8403259515762329,
"step": 470
},
{
"epoch": 0.25653016233549336,
"grad_norm": 19.16153907775879,
"learning_rate": 9.281131978677106e-07,
"logits/chosen": -0.1819503754377365,
"logits/rejected": -0.14701232314109802,
"logps/chosen": -160.57968139648438,
"logps/rejected": -164.2414093017578,
"logps/weighted_chosen": -0.609344482421875,
"logps/weighted_rejected": -0.750903308391571,
"loss": 0.6525,
"rewards/accuracies": 0.6156250238418579,
"rewards/chosen": -93.19140625,
"rewards/margins": 9.908788681030273,
"rewards/rejected": -103.0589828491211,
"rewards/weighted_accuracies": 0.590624988079071,
"rewards/weighted_chosen": -0.6482604742050171,
"rewards/weighted_margins": 0.2539626955986023,
"rewards/weighted_rejected": -0.9027160406112671,
"step": 480
},
{
"epoch": 0.2618745407174828,
"grad_norm": 47.42090606689453,
"learning_rate": 9.232202488412361e-07,
"logits/chosen": -0.18560639023780823,
"logits/rejected": -0.152149960398674,
"logps/chosen": -150.15078735351562,
"logps/rejected": -156.22109985351562,
"logps/weighted_chosen": -0.6249145269393921,
"logps/weighted_rejected": -0.755505383014679,
"loss": 0.6793,
"rewards/accuracies": 0.612500011920929,
"rewards/chosen": -89.3853530883789,
"rewards/margins": 8.602734565734863,
"rewards/rejected": -97.9830093383789,
"rewards/weighted_accuracies": 0.5531250238418579,
"rewards/weighted_chosen": -0.6580413579940796,
"rewards/weighted_margins": 0.2508483827114105,
"rewards/weighted_rejected": -0.909197986125946,
"step": 490
},
{
"epoch": 0.26721891909947226,
"grad_norm": 21.92582893371582,
"learning_rate": 9.181800111199766e-07,
"logits/chosen": -0.2139892578125,
"logits/rejected": -0.185211181640625,
"logps/chosen": -153.05624389648438,
"logps/rejected": -155.6671905517578,
"logps/weighted_chosen": -0.660186767578125,
"logps/weighted_rejected": -0.740765392780304,
"loss": 0.6631,
"rewards/accuracies": 0.596875011920929,
"rewards/chosen": -91.49101257324219,
"rewards/margins": 12.721094131469727,
"rewards/rejected": -104.1937484741211,
"rewards/weighted_accuracies": 0.6156250238418579,
"rewards/weighted_chosen": -0.679455578327179,
"rewards/weighted_margins": 0.2384185791015625,
"rewards/weighted_rejected": -0.9178100824356079,
"step": 500
},
{
"epoch": 0.26721891909947226,
"eval_logits/chosen": -0.27163267135620117,
"eval_logits/rejected": -0.24348750710487366,
"eval_logps/chosen": -159.72760009765625,
"eval_logps/rejected": -165.05091857910156,
"eval_logps/weighted_chosen": -0.6352449059486389,
"eval_logps/weighted_rejected": -0.7405111789703369,
"eval_loss": 0.6863088607788086,
"eval_rewards/accuracies": 0.5992871522903442,
"eval_rewards/chosen": -97.6285629272461,
"eval_rewards/margins": 10.883528709411621,
"eval_rewards/rejected": -108.50712585449219,
"eval_rewards/weighted_accuracies": 0.5972505211830139,
"eval_rewards/weighted_chosen": -0.6794247031211853,
"eval_rewards/weighted_margins": 0.20837070047855377,
"eval_rewards/weighted_rejected": -0.8877954483032227,
"eval_runtime": 1137.126,
"eval_samples_per_second": 1.725,
"eval_steps_per_second": 0.432,
"step": 500
}
],
"logging_steps": 10,
"max_steps": 1872,
"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": false
},
"attributes": {}
}
},
"total_flos": 0.0,
"train_batch_size": 1,
"trial_name": null,
"trial_params": null
}