Files
ModelHub XC 6547139a3d 初始化项目,由ModelHub XC社区提供模型
Model: HCY123902/llama-3-8b-dpo-tw31-beta-1e-0-ift
Source: Original Platform
2026-05-06 09:44:46 +08:00

1128 lines
44 KiB
JSON

{
"best_global_step": null,
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 0.26171159382360637,
"eval_steps": 500,
"global_step": 500,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.0005234231876472127,
"grad_norm": 136.4956817626953,
"learning_rate": 0.0,
"logits/chosen": -0.40118408203125,
"logits/rejected": -0.41802978515625,
"logps/chosen": -297.609375,
"logps/rejected": -247.84375,
"logps/weighted_chosen": -4.50634765625,
"logps/weighted_rejected": -3.43408203125,
"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": 0.005234231876472127,
"grad_norm": 269.8279113769531,
"learning_rate": 4.6875e-08,
"logits/chosen": -0.3176456093788147,
"logits/rejected": -0.3530849814414978,
"logps/chosen": -275.5694580078125,
"logps/rejected": -255.875,
"logps/weighted_chosen": -2.478325843811035,
"logps/weighted_rejected": -2.635009765625,
"loss": 0.6911,
"rewards/accuracies": 0.2604166567325592,
"rewards/chosen": -0.0588107630610466,
"rewards/margins": -0.0334201380610466,
"rewards/rejected": -0.025390625,
"rewards/weighted_accuracies": 0.3229166567325592,
"rewards/weighted_chosen": 0.0010899438057094812,
"rewards/weighted_margins": 0.0014837053604424,
"rewards/weighted_rejected": -0.00039333768654614687,
"step": 10
},
{
"epoch": 0.010468463752944255,
"grad_norm": 12.950738906860352,
"learning_rate": 9.895833333333332e-08,
"logits/chosen": -0.2977302670478821,
"logits/rejected": -0.3139175474643707,
"logps/chosen": -294.37579345703125,
"logps/rejected": -272.7203063964844,
"logps/weighted_chosen": -2.30804443359375,
"logps/weighted_rejected": -2.247570753097534,
"loss": 0.692,
"rewards/accuracies": 0.296875,
"rewards/chosen": -0.05908203125,
"rewards/margins": -0.01513671875,
"rewards/rejected": -0.0439453125,
"rewards/weighted_accuracies": 0.34687501192092896,
"rewards/weighted_chosen": 7.171630568336695e-05,
"rewards/weighted_margins": 3.662109520519152e-05,
"rewards/weighted_rejected": 3.509521411615424e-05,
"step": 20
},
{
"epoch": 0.015702695629416383,
"grad_norm": 65.14871215820312,
"learning_rate": 1.5104166666666664e-07,
"logits/chosen": -0.2917991578578949,
"logits/rejected": -0.3373458981513977,
"logps/chosen": -297.9375,
"logps/rejected": -268.0062561035156,
"logps/weighted_chosen": -2.3156495094299316,
"logps/weighted_rejected": -2.529247999191284,
"loss": 0.693,
"rewards/accuracies": 0.28125,
"rewards/chosen": 0.02177734300494194,
"rewards/margins": -0.06005859375,
"rewards/rejected": 0.08183594048023224,
"rewards/weighted_accuracies": 0.34062498807907104,
"rewards/weighted_chosen": 0.0012060165172442794,
"rewards/weighted_margins": -0.0021120072342455387,
"rewards/weighted_rejected": 0.0033180236350744963,
"step": 30
},
{
"epoch": 0.02093692750588851,
"grad_norm": 30.699859619140625,
"learning_rate": 2.03125e-07,
"logits/chosen": -0.301095575094223,
"logits/rejected": -0.3433547914028168,
"logps/chosen": -278.7124938964844,
"logps/rejected": -253.84530639648438,
"logps/weighted_chosen": -2.3508667945861816,
"logps/weighted_rejected": -2.55126953125,
"loss": 0.6909,
"rewards/accuracies": 0.2718749940395355,
"rewards/chosen": -0.014892578125,
"rewards/margins": -0.105224609375,
"rewards/rejected": 0.09033203125,
"rewards/weighted_accuracies": 0.359375,
"rewards/weighted_chosen": 0.0031726837623864412,
"rewards/weighted_margins": 0.0010789871448650956,
"rewards/weighted_rejected": 0.002093696501106024,
"step": 40
},
{
"epoch": 0.02617115938236064,
"grad_norm": 15.275498390197754,
"learning_rate": 2.552083333333333e-07,
"logits/chosen": -0.28184205293655396,
"logits/rejected": -0.32042425870895386,
"logps/chosen": -280.19140625,
"logps/rejected": -267.4398498535156,
"logps/weighted_chosen": -2.281604051589966,
"logps/weighted_rejected": -2.278125047683716,
"loss": 0.691,
"rewards/accuracies": 0.3125,
"rewards/chosen": 0.009082031436264515,
"rewards/margins": -0.11552734673023224,
"rewards/rejected": 0.12460937350988388,
"rewards/weighted_accuracies": 0.375,
"rewards/weighted_chosen": 0.0036834715865552425,
"rewards/weighted_margins": 0.0022117614280432463,
"rewards/weighted_rejected": 0.0014717101585119963,
"step": 50
},
{
"epoch": 0.031405391258832765,
"grad_norm": 42.36075973510742,
"learning_rate": 3.0729166666666665e-07,
"logits/chosen": -0.3147949278354645,
"logits/rejected": -0.3081672787666321,
"logps/chosen": -277.5492248535156,
"logps/rejected": -261.63201904296875,
"logps/weighted_chosen": -2.6925292015075684,
"logps/weighted_rejected": -2.5440917015075684,
"loss": 0.6888,
"rewards/accuracies": 0.39375001192092896,
"rewards/chosen": 0.15625,
"rewards/margins": 0.06132812425494194,
"rewards/rejected": 0.09492187201976776,
"rewards/weighted_accuracies": 0.43437498807907104,
"rewards/weighted_chosen": 0.0014663696056231856,
"rewards/weighted_margins": 0.0070289610885083675,
"rewards/weighted_rejected": -0.005612182430922985,
"step": 60
},
{
"epoch": 0.036639623135304895,
"grad_norm": 70.38953399658203,
"learning_rate": 3.59375e-07,
"logits/chosen": -0.31798094511032104,
"logits/rejected": -0.32500457763671875,
"logps/chosen": -289.6695251464844,
"logps/rejected": -244.7351531982422,
"logps/weighted_chosen": -2.19110107421875,
"logps/weighted_rejected": -2.427661180496216,
"loss": 0.6869,
"rewards/accuracies": 0.4375,
"rewards/chosen": 0.36865234375,
"rewards/margins": 0.34619140625,
"rewards/rejected": 0.0224609375,
"rewards/weighted_accuracies": 0.5375000238418579,
"rewards/weighted_chosen": 0.02108154259622097,
"rewards/weighted_margins": 0.011628913693130016,
"rewards/weighted_rejected": 0.009454727172851562,
"step": 70
},
{
"epoch": 0.04187385501177702,
"grad_norm": 54.81600570678711,
"learning_rate": 4.114583333333333e-07,
"logits/chosen": -0.2848251461982727,
"logits/rejected": -0.3061889708042145,
"logps/chosen": -289.08203125,
"logps/rejected": -269.95782470703125,
"logps/weighted_chosen": -2.3393311500549316,
"logps/weighted_rejected": -2.5093994140625,
"loss": 0.6827,
"rewards/accuracies": 0.44999998807907104,
"rewards/chosen": 0.679980456829071,
"rewards/margins": 0.3672851622104645,
"rewards/rejected": 0.31269532442092896,
"rewards/weighted_accuracies": 0.5062500238418579,
"rewards/weighted_chosen": 0.048927437514066696,
"rewards/weighted_margins": 0.025808846578001976,
"rewards/weighted_rejected": 0.02311401441693306,
"step": 80
},
{
"epoch": 0.04710808688824915,
"grad_norm": 30.603567123413086,
"learning_rate": 4.6354166666666664e-07,
"logits/chosen": -0.312631219625473,
"logits/rejected": -0.30586013197898865,
"logps/chosen": -279.92578125,
"logps/rejected": -257.5234375,
"logps/weighted_chosen": -2.5583739280700684,
"logps/weighted_rejected": -2.6037840843200684,
"loss": 0.6752,
"rewards/accuracies": 0.49687498807907104,
"rewards/chosen": 0.757617175579071,
"rewards/margins": 0.5347656011581421,
"rewards/rejected": 0.22285155951976776,
"rewards/weighted_accuracies": 0.546875,
"rewards/weighted_chosen": 0.08651771396398544,
"rewards/weighted_margins": 0.05735473707318306,
"rewards/weighted_rejected": 0.02917785570025444,
"step": 90
},
{
"epoch": 0.05234231876472128,
"grad_norm": 99.52488708496094,
"learning_rate": 5.156249999999999e-07,
"logits/chosen": -0.2793869078159332,
"logits/rejected": -0.3509994447231293,
"logps/chosen": -290.19061279296875,
"logps/rejected": -285.94842529296875,
"logps/weighted_chosen": -2.013537645339966,
"logps/weighted_rejected": -2.5231690406799316,
"loss": 0.6984,
"rewards/accuracies": 0.512499988079071,
"rewards/chosen": 0.72607421875,
"rewards/margins": 0.80712890625,
"rewards/rejected": -0.0810546875,
"rewards/weighted_accuracies": 0.6187499761581421,
"rewards/weighted_chosen": 0.07289886474609375,
"rewards/weighted_margins": 0.019171524792909622,
"rewards/weighted_rejected": 0.05372200161218643,
"step": 100
},
{
"epoch": 0.05757655064119341,
"grad_norm": 55.06239318847656,
"learning_rate": 5.677083333333333e-07,
"logits/chosen": -0.3277145326137543,
"logits/rejected": -0.3189544677734375,
"logps/chosen": -296.17889404296875,
"logps/rejected": -261.51953125,
"logps/weighted_chosen": -2.647631883621216,
"logps/weighted_rejected": -2.6458740234375,
"loss": 0.6844,
"rewards/accuracies": 0.518750011920929,
"rewards/chosen": 0.1376953125,
"rewards/margins": 0.789746105670929,
"rewards/rejected": -0.6522461175918579,
"rewards/weighted_accuracies": 0.565625011920929,
"rewards/weighted_chosen": 0.05432138592004776,
"rewards/weighted_margins": 0.03696594387292862,
"rewards/weighted_rejected": 0.01735992357134819,
"step": 110
},
{
"epoch": 0.06281078251766553,
"grad_norm": 42.45036315917969,
"learning_rate": 6.197916666666666e-07,
"logits/chosen": -0.33232802152633667,
"logits/rejected": -0.35865744948387146,
"logps/chosen": -294.6078186035156,
"logps/rejected": -255.01327514648438,
"logps/weighted_chosen": -1.970422387123108,
"logps/weighted_rejected": -2.1345458030700684,
"loss": 0.6837,
"rewards/accuracies": 0.609375,
"rewards/chosen": 0.0013671874767169356,
"rewards/margins": 1.2595703601837158,
"rewards/rejected": -1.258203148841858,
"rewards/weighted_accuracies": 0.59375,
"rewards/weighted_chosen": 0.048947714269161224,
"rewards/weighted_margins": 0.03674011304974556,
"rewards/weighted_rejected": 0.012198830023407936,
"step": 120
},
{
"epoch": 0.06804501439413765,
"grad_norm": 28.637327194213867,
"learning_rate": 6.718749999999999e-07,
"logits/chosen": -0.2933105528354645,
"logits/rejected": -0.29020920395851135,
"logps/chosen": -305.86602783203125,
"logps/rejected": -279.09844970703125,
"logps/weighted_chosen": -1.9946777820587158,
"logps/weighted_rejected": -2.2933349609375,
"loss": 0.6754,
"rewards/accuracies": 0.625,
"rewards/chosen": 0.285400390625,
"rewards/margins": 1.7852051258087158,
"rewards/rejected": -1.4998047351837158,
"rewards/weighted_accuracies": 0.6187499761581421,
"rewards/weighted_chosen": 0.06223602220416069,
"rewards/weighted_margins": 0.05499267578125,
"rewards/weighted_rejected": 0.0072036744095385075,
"step": 130
},
{
"epoch": 0.07327924627060979,
"grad_norm": 20.866477966308594,
"learning_rate": 7.239583333333333e-07,
"logits/chosen": -0.33005982637405396,
"logits/rejected": -0.34638214111328125,
"logps/chosen": -300.7281188964844,
"logps/rejected": -276.9156188964844,
"logps/weighted_chosen": -2.294506788253784,
"logps/weighted_rejected": -2.561084032058716,
"loss": 0.6606,
"rewards/accuracies": 0.628125011920929,
"rewards/chosen": -1.998437523841858,
"rewards/margins": 2.3173828125,
"rewards/rejected": -4.315820217132568,
"rewards/weighted_accuracies": 0.671875,
"rewards/weighted_chosen": 0.04616241529583931,
"rewards/weighted_margins": 0.08864898979663849,
"rewards/weighted_rejected": -0.04253540188074112,
"step": 140
},
{
"epoch": 0.07851347814708191,
"grad_norm": 43.497623443603516,
"learning_rate": 7.760416666666666e-07,
"logits/chosen": -0.2890525758266449,
"logits/rejected": -0.30130767822265625,
"logps/chosen": -287.3109436035156,
"logps/rejected": -256.72967529296875,
"logps/weighted_chosen": -2.883862257003784,
"logps/weighted_rejected": -2.796630859375,
"loss": 0.664,
"rewards/accuracies": 0.628125011920929,
"rewards/chosen": -3.108203172683716,
"rewards/margins": 2.7275390625,
"rewards/rejected": -5.835741996765137,
"rewards/weighted_accuracies": 0.6343749761581421,
"rewards/weighted_chosen": 0.04738159105181694,
"rewards/weighted_margins": 0.09773864597082138,
"rewards/weighted_rejected": -0.05034027248620987,
"step": 150
},
{
"epoch": 0.08374771002355404,
"grad_norm": 12.989988327026367,
"learning_rate": 8.28125e-07,
"logits/chosen": -0.3334007263183594,
"logits/rejected": -0.3831237852573395,
"logps/chosen": -308.2320251464844,
"logps/rejected": -281.5335998535156,
"logps/weighted_chosen": -2.378857374191284,
"logps/weighted_rejected": -2.3824095726013184,
"loss": 0.6464,
"rewards/accuracies": 0.6499999761581421,
"rewards/chosen": -4.347460746765137,
"rewards/margins": 3.98388671875,
"rewards/rejected": -8.331347465515137,
"rewards/weighted_accuracies": 0.65625,
"rewards/weighted_chosen": 0.04017028957605362,
"rewards/weighted_margins": 0.15716858208179474,
"rewards/weighted_rejected": -0.11710510402917862,
"step": 160
},
{
"epoch": 0.08898194190002617,
"grad_norm": 691.15625,
"learning_rate": 8.802083333333333e-07,
"logits/chosen": -0.37783128023147583,
"logits/rejected": -0.37789613008499146,
"logps/chosen": -310.76953125,
"logps/rejected": -270.8374938964844,
"logps/weighted_chosen": -2.6180419921875,
"logps/weighted_rejected": -2.8051514625549316,
"loss": 0.683,
"rewards/accuracies": 0.6156250238418579,
"rewards/chosen": -7.780859470367432,
"rewards/margins": 4.9013671875,
"rewards/rejected": -12.683398246765137,
"rewards/weighted_accuracies": 0.643750011920929,
"rewards/weighted_chosen": 0.06108245998620987,
"rewards/weighted_margins": 0.17230224609375,
"rewards/weighted_rejected": -0.11125946044921875,
"step": 170
},
{
"epoch": 0.0942161737764983,
"grad_norm": 65.56312561035156,
"learning_rate": 9.322916666666666e-07,
"logits/chosen": -0.3433074951171875,
"logits/rejected": -0.363290399312973,
"logps/chosen": -285.8812561035156,
"logps/rejected": -274.94842529296875,
"logps/weighted_chosen": -2.50494384765625,
"logps/weighted_rejected": -2.9246826171875,
"loss": 0.6821,
"rewards/accuracies": 0.6343749761581421,
"rewards/chosen": -12.179101943969727,
"rewards/margins": 6.432421684265137,
"rewards/rejected": -18.608592987060547,
"rewards/weighted_accuracies": 0.59375,
"rewards/weighted_chosen": -0.0216827392578125,
"rewards/weighted_margins": 0.19813232123851776,
"rewards/weighted_rejected": -0.2197723388671875,
"step": 180
},
{
"epoch": 0.09945040565297043,
"grad_norm": 26.828750610351562,
"learning_rate": 9.84375e-07,
"logits/chosen": -0.36560744047164917,
"logits/rejected": -0.40380173921585083,
"logps/chosen": -321.97967529296875,
"logps/rejected": -287.5570373535156,
"logps/weighted_chosen": -2.3037109375,
"logps/weighted_rejected": -2.64697265625,
"loss": 0.6489,
"rewards/accuracies": 0.596875011920929,
"rewards/chosen": -16.246679306030273,
"rewards/margins": 7.440039157867432,
"rewards/rejected": -23.685937881469727,
"rewards/weighted_accuracies": 0.640625,
"rewards/weighted_chosen": -0.11536254733800888,
"rewards/weighted_margins": 0.19368895888328552,
"rewards/weighted_rejected": -0.309121698141098,
"step": 190
},
{
"epoch": 0.10468463752944256,
"grad_norm": 117.84883880615234,
"learning_rate": 9.99959085414323e-07,
"logits/chosen": -0.4036270081996918,
"logits/rejected": -0.43666380643844604,
"logps/chosen": -327.8265686035156,
"logps/rejected": -284.4234313964844,
"logps/weighted_chosen": -2.6368165016174316,
"logps/weighted_rejected": -3.0156006813049316,
"loss": 0.6383,
"rewards/accuracies": 0.6156250238418579,
"rewards/chosen": -18.538379669189453,
"rewards/margins": 8.212695121765137,
"rewards/rejected": -26.748046875,
"rewards/weighted_accuracies": 0.6468750238418579,
"rewards/weighted_chosen": -0.12391586601734161,
"rewards/weighted_margins": 0.28039854764938354,
"rewards/weighted_rejected": -0.404205322265625,
"step": 200
},
{
"epoch": 0.10991886940591468,
"grad_norm": 15.525572776794434,
"learning_rate": 9.997587035630105e-07,
"logits/chosen": -0.41632765531539917,
"logits/rejected": -0.4597106873989105,
"logps/chosen": -308.75469970703125,
"logps/rejected": -316.69842529296875,
"logps/weighted_chosen": -2.439282178878784,
"logps/weighted_rejected": -3.152050733566284,
"loss": 0.6437,
"rewards/accuracies": 0.653124988079071,
"rewards/chosen": -21.632617950439453,
"rewards/margins": 10.560155868530273,
"rewards/rejected": -32.18867111206055,
"rewards/weighted_accuracies": 0.671875,
"rewards/weighted_chosen": -0.2110244780778885,
"rewards/weighted_margins": 0.23976440727710724,
"rewards/weighted_rejected": -0.4508819580078125,
"step": 210
},
{
"epoch": 0.11515310128238682,
"grad_norm": 31.583986282348633,
"learning_rate": 9.99391406364405e-07,
"logits/chosen": -0.37994080781936646,
"logits/rejected": -0.384744256734848,
"logps/chosen": -318.63592529296875,
"logps/rejected": -305.0,
"logps/weighted_chosen": -2.9573731422424316,
"logps/weighted_rejected": -3.207348585128784,
"loss": 0.7259,
"rewards/accuracies": 0.606249988079071,
"rewards/chosen": -26.522266387939453,
"rewards/margins": 9.657129287719727,
"rewards/rejected": -36.18437576293945,
"rewards/weighted_accuracies": 0.625,
"rewards/weighted_chosen": -0.3494918942451477,
"rewards/weighted_margins": 0.17980042099952698,
"rewards/weighted_rejected": -0.5295776128768921,
"step": 220
},
{
"epoch": 0.12038733315885894,
"grad_norm": 611.2006225585938,
"learning_rate": 9.988573164927884e-07,
"logits/chosen": -0.3254447877407074,
"logits/rejected": -0.36442261934280396,
"logps/chosen": -297.16717529296875,
"logps/rejected": -296.48907470703125,
"logps/weighted_chosen": -2.35888671875,
"logps/weighted_rejected": -2.804126024246216,
"loss": 0.6637,
"rewards/accuracies": 0.6656249761581421,
"rewards/chosen": -28.293750762939453,
"rewards/margins": 16.522851943969727,
"rewards/rejected": -44.817970275878906,
"rewards/weighted_accuracies": 0.6812499761581421,
"rewards/weighted_chosen": -0.1478603333234787,
"rewards/weighted_margins": 0.3882904052734375,
"rewards/weighted_rejected": -0.5360336303710938,
"step": 230
},
{
"epoch": 0.12562156503533106,
"grad_norm": 34.738059997558594,
"learning_rate": 9.98156612329838e-07,
"logits/chosen": -0.3876676559448242,
"logits/rejected": -0.4364303648471832,
"logps/chosen": -312.5921936035156,
"logps/rejected": -348.60626220703125,
"logps/weighted_chosen": -2.594775438308716,
"logps/weighted_rejected": -3.1068115234375,
"loss": 0.7482,
"rewards/accuracies": 0.637499988079071,
"rewards/chosen": -47.75507736206055,
"rewards/margins": 18.762109756469727,
"rewards/rejected": -66.5289077758789,
"rewards/weighted_accuracies": 0.6499999761581421,
"rewards/weighted_chosen": -0.319406121969223,
"rewards/weighted_margins": 0.204498291015625,
"rewards/weighted_rejected": -0.5241454839706421,
"step": 240
},
{
"epoch": 0.13085579691180318,
"grad_norm": 141.2784423828125,
"learning_rate": 9.97289527905053e-07,
"logits/chosen": -0.4159797728061676,
"logits/rejected": -0.4382568299770355,
"logps/chosen": -308.44842529296875,
"logps/rejected": -313.49530029296875,
"logps/weighted_chosen": -2.802807569503784,
"logps/weighted_rejected": -2.9874024391174316,
"loss": 0.7583,
"rewards/accuracies": 0.609375,
"rewards/chosen": -44.16621017456055,
"rewards/margins": 13.382226943969727,
"rewards/rejected": -57.54296875,
"rewards/weighted_accuracies": 0.6187499761581421,
"rewards/weighted_chosen": -0.20551452040672302,
"rewards/weighted_margins": 0.22150878608226776,
"rewards/weighted_rejected": -0.42730408906936646,
"step": 250
},
{
"epoch": 0.1360900287882753,
"grad_norm": 75.63247680664062,
"learning_rate": 9.962563528175875e-07,
"logits/chosen": -0.4156021177768707,
"logits/rejected": -0.4493362307548523,
"logps/chosen": -342.1312561035156,
"logps/rejected": -317.84063720703125,
"logps/weighted_chosen": -2.56585693359375,
"logps/weighted_rejected": -3.4276366233825684,
"loss": 0.6673,
"rewards/accuracies": 0.5625,
"rewards/chosen": -43.44648361206055,
"rewards/margins": 13.459375381469727,
"rewards/rejected": -56.90898513793945,
"rewards/weighted_accuracies": 0.6156250238418579,
"rewards/weighted_chosen": -0.310464471578598,
"rewards/weighted_margins": 0.32973939180374146,
"rewards/weighted_rejected": -0.6404876708984375,
"step": 260
},
{
"epoch": 0.14132426066474746,
"grad_norm": 32.991607666015625,
"learning_rate": 9.950574321395277e-07,
"logits/chosen": -0.507153332233429,
"logits/rejected": -0.5264068841934204,
"logps/chosen": -317.0335998535156,
"logps/rejected": -300.5718688964844,
"logps/weighted_chosen": -2.669018507003784,
"logps/weighted_rejected": -2.8846678733825684,
"loss": 0.6774,
"rewards/accuracies": 0.609375,
"rewards/chosen": -27.629491806030273,
"rewards/margins": 9.762109756469727,
"rewards/rejected": -37.39765548706055,
"rewards/weighted_accuracies": 0.6000000238418579,
"rewards/weighted_chosen": -0.23316803574562073,
"rewards/weighted_margins": 0.2321929931640625,
"rewards/weighted_rejected": -0.46527099609375,
"step": 270
},
{
"epoch": 0.14655849254121958,
"grad_norm": 41.87471389770508,
"learning_rate": 9.936931663006413e-07,
"logits/chosen": -0.5015045404434204,
"logits/rejected": -0.49525564908981323,
"logps/chosen": -319.9296875,
"logps/rejected": -312.28826904296875,
"logps/weighted_chosen": -2.667797803878784,
"logps/weighted_rejected": -3.121386766433716,
"loss": 0.6458,
"rewards/accuracies": 0.6625000238418579,
"rewards/chosen": -17.077342987060547,
"rewards/margins": 15.751367568969727,
"rewards/rejected": -32.82636642456055,
"rewards/weighted_accuracies": 0.703125,
"rewards/weighted_chosen": -0.10239715874195099,
"rewards/weighted_margins": 0.3067169189453125,
"rewards/weighted_rejected": -0.409027099609375,
"step": 280
},
{
"epoch": 0.1517927244176917,
"grad_norm": 22.655261993408203,
"learning_rate": 9.921640109546357e-07,
"logits/chosen": -0.4936012327671051,
"logits/rejected": -0.5640270113945007,
"logps/chosen": -286.4117126464844,
"logps/rejected": -286.15313720703125,
"logps/weighted_chosen": -2.640380859375,
"logps/weighted_rejected": -3.774829149246216,
"loss": 0.6369,
"rewards/accuracies": 0.6343749761581421,
"rewards/chosen": -19.629491806030273,
"rewards/margins": 14.98291015625,
"rewards/rejected": -34.61503982543945,
"rewards/weighted_accuracies": 0.668749988079071,
"rewards/weighted_chosen": -0.21005859971046448,
"rewards/weighted_margins": 0.3854110836982727,
"rewards/weighted_rejected": -0.5955780148506165,
"step": 290
},
{
"epoch": 0.15702695629416383,
"grad_norm": 33.83458709716797,
"learning_rate": 9.90470476826975e-07,
"logits/chosen": -0.5687958002090454,
"logits/rejected": -0.5700439214706421,
"logps/chosen": -298.06561279296875,
"logps/rejected": -311.8359375,
"logps/weighted_chosen": -2.555248975753784,
"logps/weighted_rejected": -3.0333251953125,
"loss": 0.6326,
"rewards/accuracies": 0.6625000238418579,
"rewards/chosen": -28.754297256469727,
"rewards/margins": 15.813085556030273,
"rewards/rejected": -44.567771911621094,
"rewards/weighted_accuracies": 0.6468750238418579,
"rewards/weighted_chosen": -0.25576478242874146,
"rewards/weighted_margins": 0.3327178955078125,
"rewards/weighted_rejected": -0.5886474847793579,
"step": 300
},
{
"epoch": 0.16226118817063595,
"grad_norm": 29.578954696655273,
"learning_rate": 9.886131295443002e-07,
"logits/chosen": -0.6218292117118835,
"logits/rejected": -0.6808746457099915,
"logps/chosen": -313.44061279296875,
"logps/rejected": -297.9156188964844,
"logps/weighted_chosen": -2.719403028488159,
"logps/weighted_rejected": -3.1142334938049316,
"loss": 0.6221,
"rewards/accuracies": 0.653124988079071,
"rewards/chosen": -32.1875,
"rewards/margins": 15.571874618530273,
"rewards/rejected": -47.7666015625,
"rewards/weighted_accuracies": 0.6499999761581421,
"rewards/weighted_chosen": -0.19769592583179474,
"rewards/weighted_margins": 0.521374523639679,
"rewards/weighted_rejected": -0.7188171148300171,
"step": 310
},
{
"epoch": 0.16749542004710807,
"grad_norm": 15.948070526123047,
"learning_rate": 9.865925894455166e-07,
"logits/chosen": -0.6636382937431335,
"logits/rejected": -0.6817939877510071,
"logps/chosen": -336.1507873535156,
"logps/rejected": -306.62890625,
"logps/weighted_chosen": -2.83941650390625,
"logps/weighted_rejected": -3.273486375808716,
"loss": 0.7046,
"rewards/accuracies": 0.637499988079071,
"rewards/chosen": -36.21074295043945,
"rewards/margins": 12.9765625,
"rewards/rejected": -49.17656326293945,
"rewards/weighted_accuracies": 0.6343749761581421,
"rewards/weighted_chosen": -0.3679946959018707,
"rewards/weighted_margins": 0.278738409280777,
"rewards/weighted_rejected": -0.6464904546737671,
"step": 320
},
{
"epoch": 0.17272965192358022,
"grad_norm": 53.10931396484375,
"learning_rate": 9.84409531374603e-07,
"logits/chosen": -0.6222923398017883,
"logits/rejected": -0.607867419719696,
"logps/chosen": -344.19061279296875,
"logps/rejected": -315.84686279296875,
"logps/weighted_chosen": -2.915234327316284,
"logps/weighted_rejected": -3.346630811691284,
"loss": 0.6334,
"rewards/accuracies": 0.668749988079071,
"rewards/chosen": -39.39531326293945,
"rewards/margins": 13.606640815734863,
"rewards/rejected": -53.00273513793945,
"rewards/weighted_accuracies": 0.65625,
"rewards/weighted_chosen": -0.33662718534469604,
"rewards/weighted_margins": 0.3510375916957855,
"rewards/weighted_rejected": -0.6879855990409851,
"step": 330
},
{
"epoch": 0.17796388380005235,
"grad_norm": 13.994344711303711,
"learning_rate": 9.820646844552219e-07,
"logits/chosen": -0.6615959405899048,
"logits/rejected": -0.714038074016571,
"logps/chosen": -312.7328186035156,
"logps/rejected": -323.3359375,
"logps/weighted_chosen": -2.910736083984375,
"logps/weighted_rejected": -3.2298340797424316,
"loss": 0.6376,
"rewards/accuracies": 0.675000011920929,
"rewards/chosen": -36.327537536621094,
"rewards/margins": 21.465625762939453,
"rewards/rejected": -57.79179763793945,
"rewards/weighted_accuracies": 0.675000011920929,
"rewards/weighted_chosen": -0.34371185302734375,
"rewards/weighted_margins": 0.390716552734375,
"rewards/weighted_rejected": -0.734301745891571,
"step": 340
},
{
"epoch": 0.18319811567652447,
"grad_norm": 13.682384490966797,
"learning_rate": 9.795588318471964e-07,
"logits/chosen": -0.774444580078125,
"logits/rejected": -0.7887184023857117,
"logps/chosen": -302.76251220703125,
"logps/rejected": -336.90313720703125,
"logps/weighted_chosen": -2.718432664871216,
"logps/weighted_rejected": -3.0622315406799316,
"loss": 0.6451,
"rewards/accuracies": 0.6343749761581421,
"rewards/chosen": -43.29413986206055,
"rewards/margins": 17.229297637939453,
"rewards/rejected": -60.52812576293945,
"rewards/weighted_accuracies": 0.637499988079071,
"rewards/weighted_chosen": -0.33645325899124146,
"rewards/weighted_margins": 0.33460694551467896,
"rewards/weighted_rejected": -0.6709426641464233,
"step": 350
},
{
"epoch": 0.1884323475529966,
"grad_norm": 19.35496711730957,
"learning_rate": 9.768928104849415e-07,
"logits/chosen": -0.815447986125946,
"logits/rejected": -0.8216186761856079,
"logps/chosen": -330.66015625,
"logps/rejected": -313.8687438964844,
"logps/weighted_chosen": -2.970752000808716,
"logps/weighted_rejected": -3.1889405250549316,
"loss": 0.6763,
"rewards/accuracies": 0.621874988079071,
"rewards/chosen": -47.90937423706055,
"rewards/margins": 17.064844131469727,
"rewards/rejected": -64.98515319824219,
"rewards/weighted_accuracies": 0.6499999761581421,
"rewards/weighted_chosen": -0.3897460997104645,
"rewards/weighted_margins": 0.384918212890625,
"rewards/weighted_rejected": -0.7745300531387329,
"step": 360
},
{
"epoch": 0.19366657942946872,
"grad_norm": 21.70383071899414,
"learning_rate": 9.740675107979355e-07,
"logits/chosen": -0.7939224243164062,
"logits/rejected": -0.827624499797821,
"logps/chosen": -373.0406188964844,
"logps/rejected": -349.3343811035156,
"logps/weighted_chosen": -2.506054639816284,
"logps/weighted_rejected": -3.43896484375,
"loss": 0.6215,
"rewards/accuracies": 0.596875011920929,
"rewards/chosen": -58.082420349121094,
"rewards/margins": 16.594335556030273,
"rewards/rejected": -74.671875,
"rewards/weighted_accuracies": 0.671875,
"rewards/weighted_chosen": -0.48696595430374146,
"rewards/weighted_margins": 0.412750244140625,
"rewards/weighted_rejected": -0.90020751953125,
"step": 370
},
{
"epoch": 0.19890081130594087,
"grad_norm": 21.682790756225586,
"learning_rate": 9.71083876413323e-07,
"logits/chosen": -0.77825927734375,
"logits/rejected": -0.7921401858329773,
"logps/chosen": -367.0062561035156,
"logps/rejected": -355.78594970703125,
"logps/weighted_chosen": -2.644775390625,
"logps/weighted_rejected": -3.5374755859375,
"loss": 0.6609,
"rewards/accuracies": 0.628125011920929,
"rewards/chosen": -64.59492492675781,
"rewards/margins": 21.614843368530273,
"rewards/rejected": -86.1988296508789,
"rewards/weighted_accuracies": 0.6625000238418579,
"rewards/weighted_chosen": -0.5685027837753296,
"rewards/weighted_margins": 0.3586792051792145,
"rewards/weighted_rejected": -0.9269439578056335,
"step": 380
},
{
"epoch": 0.204135043182413,
"grad_norm": 30.26679229736328,
"learning_rate": 9.67942903840751e-07,
"logits/chosen": -0.8401764035224915,
"logits/rejected": -0.898754894733429,
"logps/chosen": -362.0328063964844,
"logps/rejected": -362.3687438964844,
"logps/weighted_chosen": -2.8179688453674316,
"logps/weighted_rejected": -3.4515624046325684,
"loss": 0.5992,
"rewards/accuracies": 0.668749988079071,
"rewards/chosen": -57.87226486206055,
"rewards/margins": 30.196680068969727,
"rewards/rejected": -88.0804672241211,
"rewards/weighted_accuracies": 0.715624988079071,
"rewards/weighted_chosen": -0.456369012594223,
"rewards/weighted_margins": 0.495849609375,
"rewards/weighted_rejected": -0.951983630657196,
"step": 390
},
{
"epoch": 0.2093692750588851,
"grad_norm": 37.81236267089844,
"learning_rate": 9.646456421395447e-07,
"logits/chosen": -0.8902359008789062,
"logits/rejected": -0.9107666015625,
"logps/chosen": -380.6343688964844,
"logps/rejected": -399.7281188964844,
"logps/weighted_chosen": -2.709228515625,
"logps/weighted_rejected": -3.455004930496216,
"loss": 0.6557,
"rewards/accuracies": 0.6812499761581421,
"rewards/chosen": -62.43437576293945,
"rewards/margins": 33.634765625,
"rewards/rejected": -96.0941390991211,
"rewards/weighted_accuracies": 0.653124988079071,
"rewards/weighted_chosen": -0.4688262939453125,
"rewards/weighted_margins": 0.3546081483364105,
"rewards/weighted_rejected": -0.8232513666152954,
"step": 400
},
{
"epoch": 0.21460350693535724,
"grad_norm": 21.761716842651367,
"learning_rate": 9.611931925683266e-07,
"logits/chosen": -0.849993109703064,
"logits/rejected": -0.8886367678642273,
"logps/chosen": -368.8374938964844,
"logps/rejected": -353.73126220703125,
"logps/weighted_chosen": -2.533374071121216,
"logps/weighted_rejected": -3.233935594558716,
"loss": 0.591,
"rewards/accuracies": 0.6656249761581421,
"rewards/chosen": -62.575782775878906,
"rewards/margins": 27.423242568969727,
"rewards/rejected": -90.0132827758789,
"rewards/weighted_accuracies": 0.6937500238418579,
"rewards/weighted_chosen": -0.4288879334926605,
"rewards/weighted_margins": 0.507708728313446,
"rewards/weighted_rejected": -0.9364013671875,
"step": 410
},
{
"epoch": 0.21983773881182936,
"grad_norm": 28.842378616333008,
"learning_rate": 9.575867082172085e-07,
"logits/chosen": -0.836225152015686,
"logits/rejected": -0.8854034543037415,
"logps/chosen": -371.90625,
"logps/rejected": -369.4125061035156,
"logps/weighted_chosen": -2.9190917015075684,
"logps/weighted_rejected": -3.118579149246216,
"loss": 0.6198,
"rewards/accuracies": 0.668749988079071,
"rewards/chosen": -66.98905944824219,
"rewards/margins": 32.337890625,
"rewards/rejected": -99.3402328491211,
"rewards/weighted_accuracies": 0.6625000238418579,
"rewards/weighted_chosen": -0.4911743104457855,
"rewards/weighted_margins": 0.5335143804550171,
"rewards/weighted_rejected": -1.0242736339569092,
"step": 420
},
{
"epoch": 0.22507197068830148,
"grad_norm": 16.805343627929688,
"learning_rate": 9.538273936226673e-07,
"logits/chosen": -0.8646026849746704,
"logits/rejected": -0.898516833782196,
"logps/chosen": -328.94921875,
"logps/rejected": -353.66717529296875,
"logps/weighted_chosen": -3.1983399391174316,
"logps/weighted_rejected": -3.6509766578674316,
"loss": 0.6173,
"rewards/accuracies": 0.625,
"rewards/chosen": -63.52910232543945,
"rewards/margins": 25.6201171875,
"rewards/rejected": -89.1246109008789,
"rewards/weighted_accuracies": 0.6499999761581421,
"rewards/weighted_chosen": -0.436279296875,
"rewards/weighted_margins": 0.438516229391098,
"rewards/weighted_rejected": -0.8747711181640625,
"step": 430
},
{
"epoch": 0.23030620256477363,
"grad_norm": 23.78999137878418,
"learning_rate": 9.499165043652391e-07,
"logits/chosen": -0.8759521245956421,
"logits/rejected": -0.8865814208984375,
"logps/chosen": -358.69140625,
"logps/rejected": -358.76251220703125,
"logps/weighted_chosen": -3.2625732421875,
"logps/weighted_rejected": -3.359301805496216,
"loss": 0.6366,
"rewards/accuracies": 0.6187499761581421,
"rewards/chosen": -67.4326171875,
"rewards/margins": 24.610157012939453,
"rewards/rejected": -92.03788757324219,
"rewards/weighted_accuracies": 0.668749988079071,
"rewards/weighted_chosen": -0.5896576046943665,
"rewards/weighted_margins": 0.34770506620407104,
"rewards/weighted_rejected": -0.9376128911972046,
"step": 440
},
{
"epoch": 0.23554043444124576,
"grad_norm": 49.646263122558594,
"learning_rate": 9.458553466501665e-07,
"logits/chosen": -0.9158996343612671,
"logits/rejected": -0.9590820074081421,
"logps/chosen": -355.3109436035156,
"logps/rejected": -341.9765625,
"logps/weighted_chosen": -3.231823682785034,
"logps/weighted_rejected": -3.4568848609924316,
"loss": 0.657,
"rewards/accuracies": 0.6312500238418579,
"rewards/chosen": -69.14453125,
"rewards/margins": 28.884180068969727,
"rewards/rejected": -98.0492172241211,
"rewards/weighted_accuracies": 0.6875,
"rewards/weighted_chosen": -0.6470184326171875,
"rewards/weighted_margins": 0.4533935487270355,
"rewards/weighted_rejected": -1.100683569908142,
"step": 450
},
{
"epoch": 0.24077466631771788,
"grad_norm": 38.45325469970703,
"learning_rate": 9.416452768711366e-07,
"logits/chosen": -0.9281005859375,
"logits/rejected": -0.9545837640762329,
"logps/chosen": -375.6390686035156,
"logps/rejected": -370.203125,
"logps/weighted_chosen": -2.9860596656799316,
"logps/weighted_rejected": -3.635498046875,
"loss": 0.6457,
"rewards/accuracies": 0.612500011920929,
"rewards/chosen": -83.3871078491211,
"rewards/margins": 26.872264862060547,
"rewards/rejected": -110.26094055175781,
"rewards/weighted_accuracies": 0.6625000238418579,
"rewards/weighted_chosen": -0.6367782354354858,
"rewards/weighted_margins": 0.515301525592804,
"rewards/weighted_rejected": -1.152197241783142,
"step": 460
},
{
"epoch": 0.24600889819419,
"grad_norm": 59.67642593383789,
"learning_rate": 9.372877011572557e-07,
"logits/chosen": -0.8797394037246704,
"logits/rejected": -0.904949963092804,
"logps/chosen": -398.8531188964844,
"logps/rejected": -388.22186279296875,
"logps/weighted_chosen": -3.0103516578674316,
"logps/weighted_rejected": -3.409374952316284,
"loss": 0.6181,
"rewards/accuracies": 0.6031249761581421,
"rewards/chosen": -88.2437515258789,
"rewards/margins": 25.542577743530273,
"rewards/rejected": -113.7890625,
"rewards/weighted_accuracies": 0.6781250238418579,
"rewards/weighted_chosen": -0.6517273187637329,
"rewards/weighted_margins": 0.482818603515625,
"rewards/weighted_rejected": -1.1340011358261108,
"step": 470
},
{
"epoch": 0.2512431300706621,
"grad_norm": 27.465656280517578,
"learning_rate": 9.327840749034141e-07,
"logits/chosen": -0.910723865032196,
"logits/rejected": -0.9439147710800171,
"logps/chosen": -364.4750061035156,
"logps/rejected": -392.03436279296875,
"logps/weighted_chosen": -2.8880372047424316,
"logps/weighted_rejected": -4.069311618804932,
"loss": 0.6276,
"rewards/accuracies": 0.6812499761581421,
"rewards/chosen": -77.44355773925781,
"rewards/margins": 38.147850036621094,
"rewards/rejected": -115.59648132324219,
"rewards/weighted_accuracies": 0.684374988079071,
"rewards/weighted_chosen": -0.5768798589706421,
"rewards/weighted_margins": 0.541027843952179,
"rewards/weighted_rejected": -1.1178162097930908,
"step": 480
},
{
"epoch": 0.2564773619471343,
"grad_norm": 34.35914993286133,
"learning_rate": 9.281359022841965e-07,
"logits/chosen": -0.80584716796875,
"logits/rejected": -0.8291991949081421,
"logps/chosen": -349.86798095703125,
"logps/rejected": -354.75701904296875,
"logps/weighted_chosen": -2.914257764816284,
"logps/weighted_rejected": -4.178680419921875,
"loss": 0.63,
"rewards/accuracies": 0.625,
"rewards/chosen": -74.94511413574219,
"rewards/margins": 34.957618713378906,
"rewards/rejected": -109.90644836425781,
"rewards/weighted_accuracies": 0.690625011920929,
"rewards/weighted_chosen": -0.564099133014679,
"rewards/weighted_margins": 0.658795177936554,
"rewards/weighted_rejected": -1.2229950428009033,
"step": 490
},
{
"epoch": 0.26171159382360637,
"grad_norm": 41.01088333129883,
"learning_rate": 9.233447357514989e-07,
"logits/chosen": -0.775347888469696,
"logits/rejected": -0.8226684331893921,
"logps/chosen": -369.1148376464844,
"logps/rejected": -373.8374938964844,
"logps/weighted_chosen": -3.288012742996216,
"logps/weighted_rejected": -3.7696776390075684,
"loss": 0.6225,
"rewards/accuracies": 0.643750011920929,
"rewards/chosen": -71.1283187866211,
"rewards/margins": 32.867576599121094,
"rewards/rejected": -104.0218734741211,
"rewards/weighted_accuracies": 0.6937500238418579,
"rewards/weighted_chosen": -0.5540069341659546,
"rewards/weighted_margins": 0.598828136920929,
"rewards/weighted_rejected": -1.1535155773162842,
"step": 500
},
{
"epoch": 0.26171159382360637,
"eval_logits/chosen": -0.8767929673194885,
"eval_logits/rejected": -0.8968105316162109,
"eval_logps/chosen": -369.21099853515625,
"eval_logps/rejected": -379.3609924316406,
"eval_logps/weighted_chosen": -2.957455635070801,
"eval_logps/weighted_rejected": -3.6658689975738525,
"eval_loss": 0.6115986108779907,
"eval_rewards/accuracies": 0.6294999718666077,
"eval_rewards/chosen": -80.43468475341797,
"eval_rewards/margins": 31.297624588012695,
"eval_rewards/rejected": -111.73987579345703,
"eval_rewards/weighted_accuracies": 0.6759999990463257,
"eval_rewards/weighted_chosen": -0.5706117153167725,
"eval_rewards/weighted_margins": 0.5400464534759521,
"eval_rewards/weighted_rejected": -1.1106581687927246,
"eval_runtime": 1361.4222,
"eval_samples_per_second": 1.469,
"eval_steps_per_second": 0.367,
"step": 500
}
],
"logging_steps": 10,
"max_steps": 1911,
"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
}