b1dd7b86f3
Model: ali-elganzory/Baguettotron-longsft_16k-DPO-Tulu3-decontaminated Source: Original Platform
3239 lines
108 KiB
JSON
3239 lines
108 KiB
JSON
{
|
|
"best_global_step": null,
|
|
"best_metric": null,
|
|
"best_model_checkpoint": null,
|
|
"epoch": 1.0,
|
|
"eval_steps": 500,
|
|
"global_step": 2130,
|
|
"is_hyper_param_search": false,
|
|
"is_local_process_zero": true,
|
|
"is_world_process_zero": true,
|
|
"log_history": [
|
|
{
|
|
"epoch": 0.004694835680751174,
|
|
"grad_norm": 520.142591451342,
|
|
"learning_rate": 2.1126760563380282e-08,
|
|
"logits/chosen": -2.4390625953674316,
|
|
"logits/rejected": -2.137500047683716,
|
|
"logps/chosen": -696.7999877929688,
|
|
"logps/rejected": -814.0,
|
|
"loss": 1.997,
|
|
"rewards/accuracies": 0.09296874701976776,
|
|
"rewards/chosen": -0.338623046875,
|
|
"rewards/margins": -0.05322265625,
|
|
"rewards/rejected": -0.2852539122104645,
|
|
"step": 10
|
|
},
|
|
{
|
|
"epoch": 0.009389671361502348,
|
|
"grad_norm": 718.6985853103332,
|
|
"learning_rate": 4.460093896713615e-08,
|
|
"logits/chosen": -2.362499952316284,
|
|
"logits/rejected": -2.2249999046325684,
|
|
"logps/chosen": -732.4000244140625,
|
|
"logps/rejected": -813.2000122070312,
|
|
"loss": 3.6606,
|
|
"rewards/accuracies": 0.23906250298023224,
|
|
"rewards/chosen": 0.20595702528953552,
|
|
"rewards/margins": 0.1640625,
|
|
"rewards/rejected": 0.04326171800494194,
|
|
"step": 20
|
|
},
|
|
{
|
|
"epoch": 0.014084507042253521,
|
|
"grad_norm": 1552.2132968157528,
|
|
"learning_rate": 6.807511737089202e-08,
|
|
"logits/chosen": -2.2523436546325684,
|
|
"logits/rejected": -2.0,
|
|
"logps/chosen": -778.7999877929688,
|
|
"logps/rejected": -908.7999877929688,
|
|
"loss": 4.2634,
|
|
"rewards/accuracies": 0.26171875,
|
|
"rewards/chosen": -0.35332030057907104,
|
|
"rewards/margins": -0.07529296725988388,
|
|
"rewards/rejected": -0.2796874940395355,
|
|
"step": 30
|
|
},
|
|
{
|
|
"epoch": 0.018779342723004695,
|
|
"grad_norm": 8657.18713257464,
|
|
"learning_rate": 9.154929577464789e-08,
|
|
"logits/chosen": -2.339062452316284,
|
|
"logits/rejected": -2.137500047683716,
|
|
"logps/chosen": -770.4000244140625,
|
|
"logps/rejected": -930.7999877929688,
|
|
"loss": 3.8689,
|
|
"rewards/accuracies": 0.28515625,
|
|
"rewards/chosen": 0.4447265565395355,
|
|
"rewards/margins": 0.9095703363418579,
|
|
"rewards/rejected": -0.46601563692092896,
|
|
"step": 40
|
|
},
|
|
{
|
|
"epoch": 0.023474178403755867,
|
|
"grad_norm": 565.9183355419589,
|
|
"learning_rate": 1.1502347417840374e-07,
|
|
"logits/chosen": -2.268749952316284,
|
|
"logits/rejected": -2.1328125,
|
|
"logps/chosen": -753.2000122070312,
|
|
"logps/rejected": -833.5999755859375,
|
|
"loss": 4.4475,
|
|
"rewards/accuracies": 0.2718749940395355,
|
|
"rewards/chosen": -0.08486328274011612,
|
|
"rewards/margins": -0.04443359375,
|
|
"rewards/rejected": -0.04072265699505806,
|
|
"step": 50
|
|
},
|
|
{
|
|
"epoch": 0.028169014084507043,
|
|
"grad_norm": 697.3566349630747,
|
|
"learning_rate": 1.384976525821596e-07,
|
|
"logits/chosen": -2.2906250953674316,
|
|
"logits/rejected": -2.1039061546325684,
|
|
"logps/chosen": -779.5999755859375,
|
|
"logps/rejected": -908.4000244140625,
|
|
"loss": 4.4893,
|
|
"rewards/accuracies": 0.26171875,
|
|
"rewards/chosen": -0.203125,
|
|
"rewards/margins": -0.654125988483429,
|
|
"rewards/rejected": 0.4522460997104645,
|
|
"step": 60
|
|
},
|
|
{
|
|
"epoch": 0.03286384976525822,
|
|
"grad_norm": 10766.834300261338,
|
|
"learning_rate": 1.619718309859155e-07,
|
|
"logits/chosen": -2.1617188453674316,
|
|
"logits/rejected": -1.9695312976837158,
|
|
"logps/chosen": -745.5999755859375,
|
|
"logps/rejected": -816.7999877929688,
|
|
"loss": 4.8299,
|
|
"rewards/accuracies": 0.2867187559604645,
|
|
"rewards/chosen": 0.15312500298023224,
|
|
"rewards/margins": -0.07851562649011612,
|
|
"rewards/rejected": 0.23310546576976776,
|
|
"step": 70
|
|
},
|
|
{
|
|
"epoch": 0.03755868544600939,
|
|
"grad_norm": 505.2891052099527,
|
|
"learning_rate": 1.8544600938967138e-07,
|
|
"logits/chosen": -2.3765625953674316,
|
|
"logits/rejected": -2.1953125,
|
|
"logps/chosen": -740.7999877929688,
|
|
"logps/rejected": -854.7999877929688,
|
|
"loss": 4.1491,
|
|
"rewards/accuracies": 0.2718749940395355,
|
|
"rewards/chosen": 0.2137451171875,
|
|
"rewards/margins": 0.07695312798023224,
|
|
"rewards/rejected": 0.13845214247703552,
|
|
"step": 80
|
|
},
|
|
{
|
|
"epoch": 0.04225352112676056,
|
|
"grad_norm": 668.2554945973214,
|
|
"learning_rate": 2.089201877934272e-07,
|
|
"logits/chosen": -2.3890624046325684,
|
|
"logits/rejected": -2.2265625,
|
|
"logps/chosen": -718.7999877929688,
|
|
"logps/rejected": -816.4000244140625,
|
|
"loss": 3.908,
|
|
"rewards/accuracies": 0.2578125,
|
|
"rewards/chosen": 0.4122680723667145,
|
|
"rewards/margins": -0.19423827528953552,
|
|
"rewards/rejected": 0.6078125238418579,
|
|
"step": 90
|
|
},
|
|
{
|
|
"epoch": 0.046948356807511735,
|
|
"grad_norm": 776.6929187550726,
|
|
"learning_rate": 2.323943661971831e-07,
|
|
"logits/chosen": -2.307812452316284,
|
|
"logits/rejected": -2.211718797683716,
|
|
"logps/chosen": -764.4000244140625,
|
|
"logps/rejected": -798.4000244140625,
|
|
"loss": 3.9806,
|
|
"rewards/accuracies": 0.2632812559604645,
|
|
"rewards/chosen": -0.03876953199505806,
|
|
"rewards/margins": -0.2451171875,
|
|
"rewards/rejected": 0.20556640625,
|
|
"step": 100
|
|
},
|
|
{
|
|
"epoch": 0.051643192488262914,
|
|
"grad_norm": 717.5365230161353,
|
|
"learning_rate": 2.5586854460093895e-07,
|
|
"logits/chosen": -2.4203124046325684,
|
|
"logits/rejected": -2.2671875953674316,
|
|
"logps/chosen": -742.0,
|
|
"logps/rejected": -870.4000244140625,
|
|
"loss": 4.6294,
|
|
"rewards/accuracies": 0.27812498807907104,
|
|
"rewards/chosen": 0.15576171875,
|
|
"rewards/margins": -0.35624998807907104,
|
|
"rewards/rejected": 0.5127929449081421,
|
|
"step": 110
|
|
},
|
|
{
|
|
"epoch": 0.056338028169014086,
|
|
"grad_norm": 2977.9502504703755,
|
|
"learning_rate": 2.7934272300469483e-07,
|
|
"logits/chosen": -2.2554688453674316,
|
|
"logits/rejected": -2.1578125953674316,
|
|
"logps/chosen": -768.0,
|
|
"logps/rejected": -862.4000244140625,
|
|
"loss": 4.0131,
|
|
"rewards/accuracies": 0.28125,
|
|
"rewards/chosen": 0.18720702826976776,
|
|
"rewards/margins": -0.06640625,
|
|
"rewards/rejected": 0.25371092557907104,
|
|
"step": 120
|
|
},
|
|
{
|
|
"epoch": 0.06103286384976526,
|
|
"grad_norm": 2509.124943319802,
|
|
"learning_rate": 3.0281690140845066e-07,
|
|
"logits/chosen": -2.235156297683716,
|
|
"logits/rejected": -2.1156249046325684,
|
|
"logps/chosen": -781.2000122070312,
|
|
"logps/rejected": -900.7999877929688,
|
|
"loss": 6.0719,
|
|
"rewards/accuracies": 0.27421873807907104,
|
|
"rewards/chosen": -0.826171875,
|
|
"rewards/margins": -1.742456078529358,
|
|
"rewards/rejected": 0.917285144329071,
|
|
"step": 130
|
|
},
|
|
{
|
|
"epoch": 0.06572769953051644,
|
|
"grad_norm": 690.7196142418027,
|
|
"learning_rate": 3.2629107981220654e-07,
|
|
"logits/chosen": -2.234375,
|
|
"logits/rejected": -2.11328125,
|
|
"logps/chosen": -766.7999877929688,
|
|
"logps/rejected": -898.0,
|
|
"loss": 5.058,
|
|
"rewards/accuracies": 0.25703126192092896,
|
|
"rewards/chosen": -0.03217773512005806,
|
|
"rewards/margins": -0.32109373807907104,
|
|
"rewards/rejected": 0.2876953184604645,
|
|
"step": 140
|
|
},
|
|
{
|
|
"epoch": 0.07042253521126761,
|
|
"grad_norm": 669.3896973423242,
|
|
"learning_rate": 3.497652582159624e-07,
|
|
"logits/chosen": -2.305468797683716,
|
|
"logits/rejected": -2.282031297683716,
|
|
"logps/chosen": -746.7999877929688,
|
|
"logps/rejected": -861.5999755859375,
|
|
"loss": 3.5748,
|
|
"rewards/accuracies": 0.2757812440395355,
|
|
"rewards/chosen": 0.6036132574081421,
|
|
"rewards/margins": 0.8902343511581421,
|
|
"rewards/rejected": -0.28557127714157104,
|
|
"step": 150
|
|
},
|
|
{
|
|
"epoch": 0.07511737089201878,
|
|
"grad_norm": 721.4120567673536,
|
|
"learning_rate": 3.732394366197183e-07,
|
|
"logits/chosen": -2.2750000953674316,
|
|
"logits/rejected": -2.120312452316284,
|
|
"logps/chosen": -770.7999877929688,
|
|
"logps/rejected": -874.4000244140625,
|
|
"loss": 4.8792,
|
|
"rewards/accuracies": 0.29296875,
|
|
"rewards/chosen": -0.11386718600988388,
|
|
"rewards/margins": 0.2789062559604645,
|
|
"rewards/rejected": -0.38984376192092896,
|
|
"step": 160
|
|
},
|
|
{
|
|
"epoch": 0.07981220657276995,
|
|
"grad_norm": 1753.633484672201,
|
|
"learning_rate": 3.967136150234742e-07,
|
|
"logits/chosen": -2.2515625953674316,
|
|
"logits/rejected": -2.065624952316284,
|
|
"logps/chosen": -728.7999877929688,
|
|
"logps/rejected": -867.5999755859375,
|
|
"loss": 5.735,
|
|
"rewards/accuracies": 0.29609376192092896,
|
|
"rewards/chosen": 0.771484375,
|
|
"rewards/margins": -0.755175769329071,
|
|
"rewards/rejected": 1.525048851966858,
|
|
"step": 170
|
|
},
|
|
{
|
|
"epoch": 0.08450704225352113,
|
|
"grad_norm": 778.6604227743157,
|
|
"learning_rate": 4.2018779342723e-07,
|
|
"logits/chosen": -2.2890625,
|
|
"logits/rejected": -2.090625047683716,
|
|
"logps/chosen": -700.0,
|
|
"logps/rejected": -869.2000122070312,
|
|
"loss": 3.5314,
|
|
"rewards/accuracies": 0.2757812440395355,
|
|
"rewards/chosen": 0.5699707269668579,
|
|
"rewards/margins": 0.905102550983429,
|
|
"rewards/rejected": -0.3370117247104645,
|
|
"step": 180
|
|
},
|
|
{
|
|
"epoch": 0.0892018779342723,
|
|
"grad_norm": 742.4443494266026,
|
|
"learning_rate": 4.436619718309859e-07,
|
|
"logits/chosen": -2.21875,
|
|
"logits/rejected": -2.149218797683716,
|
|
"logps/chosen": -752.0,
|
|
"logps/rejected": -814.4000244140625,
|
|
"loss": 4.7805,
|
|
"rewards/accuracies": 0.28828126192092896,
|
|
"rewards/chosen": 0.2789062559604645,
|
|
"rewards/margins": -0.45097655057907104,
|
|
"rewards/rejected": 0.729296863079071,
|
|
"step": 190
|
|
},
|
|
{
|
|
"epoch": 0.09389671361502347,
|
|
"grad_norm": 542.7728573196438,
|
|
"learning_rate": 4.671361502347418e-07,
|
|
"logits/chosen": -2.2554688453674316,
|
|
"logits/rejected": -2.0648436546325684,
|
|
"logps/chosen": -729.5999755859375,
|
|
"logps/rejected": -824.4000244140625,
|
|
"loss": 3.6234,
|
|
"rewards/accuracies": 0.27812498807907104,
|
|
"rewards/chosen": 0.767382800579071,
|
|
"rewards/margins": 1.3642578125,
|
|
"rewards/rejected": -0.5965331792831421,
|
|
"step": 200
|
|
},
|
|
{
|
|
"epoch": 0.09859154929577464,
|
|
"grad_norm": 1461.4933077149913,
|
|
"learning_rate": 4.906103286384976e-07,
|
|
"logits/chosen": -2.2671875953674316,
|
|
"logits/rejected": -2.1195311546325684,
|
|
"logps/chosen": -748.7999877929688,
|
|
"logps/rejected": -891.5999755859375,
|
|
"loss": 3.7303,
|
|
"rewards/accuracies": 0.2757812440395355,
|
|
"rewards/chosen": 0.4224853515625,
|
|
"rewards/margins": 0.23105469346046448,
|
|
"rewards/rejected": 0.19394531846046448,
|
|
"step": 210
|
|
},
|
|
{
|
|
"epoch": 0.10328638497652583,
|
|
"grad_norm": 756.737522385797,
|
|
"learning_rate": 4.984350547730829e-07,
|
|
"logits/chosen": -2.3921875953674316,
|
|
"logits/rejected": -2.1703124046325684,
|
|
"logps/chosen": -779.2000122070312,
|
|
"logps/rejected": -852.7999877929688,
|
|
"loss": 4.1088,
|
|
"rewards/accuracies": 0.26875001192092896,
|
|
"rewards/chosen": 0.364013671875,
|
|
"rewards/margins": 0.3695312440395355,
|
|
"rewards/rejected": -0.00791015662252903,
|
|
"step": 220
|
|
},
|
|
{
|
|
"epoch": 0.107981220657277,
|
|
"grad_norm": 2242.980863525319,
|
|
"learning_rate": 4.958268127282212e-07,
|
|
"logits/chosen": -2.2867188453674316,
|
|
"logits/rejected": -2.1851563453674316,
|
|
"logps/chosen": -755.2000122070312,
|
|
"logps/rejected": -878.7999877929688,
|
|
"loss": 3.9552,
|
|
"rewards/accuracies": 0.3062500059604645,
|
|
"rewards/chosen": 1.0666015148162842,
|
|
"rewards/margins": 2.1900391578674316,
|
|
"rewards/rejected": -1.1203124523162842,
|
|
"step": 230
|
|
},
|
|
{
|
|
"epoch": 0.11267605633802817,
|
|
"grad_norm": 455.8500781083609,
|
|
"learning_rate": 4.932185706833594e-07,
|
|
"logits/chosen": -2.2164063453674316,
|
|
"logits/rejected": -2.168750047683716,
|
|
"logps/chosen": -693.5999755859375,
|
|
"logps/rejected": -736.4000244140625,
|
|
"loss": 4.3035,
|
|
"rewards/accuracies": 0.29609376192092896,
|
|
"rewards/chosen": 0.21464844048023224,
|
|
"rewards/margins": 0.4082275331020355,
|
|
"rewards/rejected": -0.19091796875,
|
|
"step": 240
|
|
},
|
|
{
|
|
"epoch": 0.11737089201877934,
|
|
"grad_norm": 9707.277894991463,
|
|
"learning_rate": 4.906103286384976e-07,
|
|
"logits/chosen": -2.317187547683716,
|
|
"logits/rejected": -2.1796875,
|
|
"logps/chosen": -742.4000244140625,
|
|
"logps/rejected": -878.0,
|
|
"loss": 5.258,
|
|
"rewards/accuracies": 0.28593748807907104,
|
|
"rewards/chosen": -1.098413109779358,
|
|
"rewards/margins": -0.5367676019668579,
|
|
"rewards/rejected": -0.5621093511581421,
|
|
"step": 250
|
|
},
|
|
{
|
|
"epoch": 0.12206572769953052,
|
|
"grad_norm": 636.3943955298223,
|
|
"learning_rate": 4.880020865936358e-07,
|
|
"logits/chosen": -2.2265625,
|
|
"logits/rejected": -2.200000047683716,
|
|
"logps/chosen": -700.7999877929688,
|
|
"logps/rejected": -814.4000244140625,
|
|
"loss": 4.1654,
|
|
"rewards/accuracies": 0.296875,
|
|
"rewards/chosen": 0.925488293170929,
|
|
"rewards/margins": 0.37675780057907104,
|
|
"rewards/rejected": 0.548291027545929,
|
|
"step": 260
|
|
},
|
|
{
|
|
"epoch": 0.1267605633802817,
|
|
"grad_norm": 628.6083377328596,
|
|
"learning_rate": 4.853938445487741e-07,
|
|
"logits/chosen": -2.1937499046325684,
|
|
"logits/rejected": -1.8914062976837158,
|
|
"logps/chosen": -742.4000244140625,
|
|
"logps/rejected": -864.7999877929688,
|
|
"loss": 4.3146,
|
|
"rewards/accuracies": 0.3046875,
|
|
"rewards/chosen": 0.3232421875,
|
|
"rewards/margins": 0.791210949420929,
|
|
"rewards/rejected": -0.46650391817092896,
|
|
"step": 270
|
|
},
|
|
{
|
|
"epoch": 0.13145539906103287,
|
|
"grad_norm": 518.911626956217,
|
|
"learning_rate": 4.827856025039123e-07,
|
|
"logits/chosen": -2.3062500953674316,
|
|
"logits/rejected": -2.207812547683716,
|
|
"logps/chosen": -752.0,
|
|
"logps/rejected": -876.7999877929688,
|
|
"loss": 4.545,
|
|
"rewards/accuracies": 0.3031249940395355,
|
|
"rewards/chosen": 1.2041015625,
|
|
"rewards/margins": 1.7902343273162842,
|
|
"rewards/rejected": -0.5816406011581421,
|
|
"step": 280
|
|
},
|
|
{
|
|
"epoch": 0.13615023474178403,
|
|
"grad_norm": 466.3217272893134,
|
|
"learning_rate": 4.801773604590506e-07,
|
|
"logits/chosen": -2.221874952316284,
|
|
"logits/rejected": -2.1812500953674316,
|
|
"logps/chosen": -797.2000122070312,
|
|
"logps/rejected": -874.7999877929688,
|
|
"loss": 3.4628,
|
|
"rewards/accuracies": 0.2992187440395355,
|
|
"rewards/chosen": 0.5323241949081421,
|
|
"rewards/margins": 1.7703125476837158,
|
|
"rewards/rejected": -1.240966796875,
|
|
"step": 290
|
|
},
|
|
{
|
|
"epoch": 0.14084507042253522,
|
|
"grad_norm": 1099.1881265876816,
|
|
"learning_rate": 4.775691184141888e-07,
|
|
"logits/chosen": -2.3578124046325684,
|
|
"logits/rejected": -2.2203125953674316,
|
|
"logps/chosen": -760.4000244140625,
|
|
"logps/rejected": -860.0,
|
|
"loss": 4.6967,
|
|
"rewards/accuracies": 0.2945312559604645,
|
|
"rewards/chosen": 0.25957030057907104,
|
|
"rewards/margins": 3.1707520484924316,
|
|
"rewards/rejected": -2.9110350608825684,
|
|
"step": 300
|
|
},
|
|
{
|
|
"epoch": 0.14553990610328638,
|
|
"grad_norm": 967.9292035390623,
|
|
"learning_rate": 4.749608763693271e-07,
|
|
"logits/chosen": -2.3109374046325684,
|
|
"logits/rejected": -2.1226563453674316,
|
|
"logps/chosen": -725.2000122070312,
|
|
"logps/rejected": -836.4000244140625,
|
|
"loss": 3.5246,
|
|
"rewards/accuracies": 0.3148437440395355,
|
|
"rewards/chosen": 1.117578148841858,
|
|
"rewards/margins": 1.4765625,
|
|
"rewards/rejected": -0.36152344942092896,
|
|
"step": 310
|
|
},
|
|
{
|
|
"epoch": 0.15023474178403756,
|
|
"grad_norm": 798.8789598637936,
|
|
"learning_rate": 4.7235263432446533e-07,
|
|
"logits/chosen": -2.3187499046325684,
|
|
"logits/rejected": -2.12890625,
|
|
"logps/chosen": -739.5999755859375,
|
|
"logps/rejected": -856.7999877929688,
|
|
"loss": 3.8969,
|
|
"rewards/accuracies": 0.3179687559604645,
|
|
"rewards/chosen": 0.737060546875,
|
|
"rewards/margins": 2.250781297683716,
|
|
"rewards/rejected": -1.516015648841858,
|
|
"step": 320
|
|
},
|
|
{
|
|
"epoch": 0.15492957746478872,
|
|
"grad_norm": 718.0201290330298,
|
|
"learning_rate": 4.6974439227960353e-07,
|
|
"logits/chosen": -2.28125,
|
|
"logits/rejected": -2.059375047683716,
|
|
"logps/chosen": -770.0,
|
|
"logps/rejected": -887.2000122070312,
|
|
"loss": 3.9144,
|
|
"rewards/accuracies": 0.30546873807907104,
|
|
"rewards/chosen": 1.103124976158142,
|
|
"rewards/margins": 1.5705077648162842,
|
|
"rewards/rejected": -0.46699219942092896,
|
|
"step": 330
|
|
},
|
|
{
|
|
"epoch": 0.1596244131455399,
|
|
"grad_norm": 519.8108487600896,
|
|
"learning_rate": 4.671361502347418e-07,
|
|
"logits/chosen": -2.3984375,
|
|
"logits/rejected": -2.2796874046325684,
|
|
"logps/chosen": -720.7999877929688,
|
|
"logps/rejected": -777.2000122070312,
|
|
"loss": 5.8592,
|
|
"rewards/accuracies": 0.30859375,
|
|
"rewards/chosen": 0.3306640684604645,
|
|
"rewards/margins": -0.9193359613418579,
|
|
"rewards/rejected": 1.249609351158142,
|
|
"step": 340
|
|
},
|
|
{
|
|
"epoch": 0.1643192488262911,
|
|
"grad_norm": 619.4060279377454,
|
|
"learning_rate": 4.6452790818988004e-07,
|
|
"logits/chosen": -2.3335938453674316,
|
|
"logits/rejected": -2.171093702316284,
|
|
"logps/chosen": -773.2000122070312,
|
|
"logps/rejected": -869.5999755859375,
|
|
"loss": 3.5465,
|
|
"rewards/accuracies": 0.3148437440395355,
|
|
"rewards/chosen": 0.9961913824081421,
|
|
"rewards/margins": 1.6785156726837158,
|
|
"rewards/rejected": -0.6844726800918579,
|
|
"step": 350
|
|
},
|
|
{
|
|
"epoch": 0.16901408450704225,
|
|
"grad_norm": 10870.44571613044,
|
|
"learning_rate": 4.6191966614501824e-07,
|
|
"logits/chosen": -2.31640625,
|
|
"logits/rejected": -2.110156297683716,
|
|
"logps/chosen": -731.2000122070312,
|
|
"logps/rejected": -841.5999755859375,
|
|
"loss": 6.2137,
|
|
"rewards/accuracies": 0.32343751192092896,
|
|
"rewards/chosen": 0.8446289300918579,
|
|
"rewards/margins": 0.35224610567092896,
|
|
"rewards/rejected": 0.49506837129592896,
|
|
"step": 360
|
|
},
|
|
{
|
|
"epoch": 0.17370892018779344,
|
|
"grad_norm": 6196.223136121422,
|
|
"learning_rate": 4.593114241001565e-07,
|
|
"logits/chosen": -2.2249999046325684,
|
|
"logits/rejected": -2.133593797683716,
|
|
"logps/chosen": -795.2000122070312,
|
|
"logps/rejected": -871.2000122070312,
|
|
"loss": 4.3763,
|
|
"rewards/accuracies": 0.30937498807907104,
|
|
"rewards/chosen": 0.8076171875,
|
|
"rewards/margins": 2.6605467796325684,
|
|
"rewards/rejected": -1.8560059070587158,
|
|
"step": 370
|
|
},
|
|
{
|
|
"epoch": 0.1784037558685446,
|
|
"grad_norm": 956.6319306542155,
|
|
"learning_rate": 4.5670318205529474e-07,
|
|
"logits/chosen": -2.2015624046325684,
|
|
"logits/rejected": -2.020312547683716,
|
|
"logps/chosen": -750.7999877929688,
|
|
"logps/rejected": -871.5999755859375,
|
|
"loss": 3.8167,
|
|
"rewards/accuracies": 0.3304687440395355,
|
|
"rewards/chosen": 1.7366943359375,
|
|
"rewards/margins": 3.37890625,
|
|
"rewards/rejected": -1.638818383216858,
|
|
"step": 380
|
|
},
|
|
{
|
|
"epoch": 0.18309859154929578,
|
|
"grad_norm": 8776.661255480909,
|
|
"learning_rate": 4.54094940010433e-07,
|
|
"logits/chosen": -2.2578125,
|
|
"logits/rejected": -2.11328125,
|
|
"logps/chosen": -763.5999755859375,
|
|
"logps/rejected": -902.4000244140625,
|
|
"loss": 4.024,
|
|
"rewards/accuracies": 0.3125,
|
|
"rewards/chosen": 0.804882824420929,
|
|
"rewards/margins": 2.1529297828674316,
|
|
"rewards/rejected": -1.345312476158142,
|
|
"step": 390
|
|
},
|
|
{
|
|
"epoch": 0.18779342723004694,
|
|
"grad_norm": 719.0136384818551,
|
|
"learning_rate": 4.514866979655712e-07,
|
|
"logits/chosen": -2.3140625953674316,
|
|
"logits/rejected": -2.1226563453674316,
|
|
"logps/chosen": -755.5999755859375,
|
|
"logps/rejected": -854.0,
|
|
"loss": 3.6524,
|
|
"rewards/accuracies": 0.3304687440395355,
|
|
"rewards/chosen": 1.326562523841858,
|
|
"rewards/margins": 5.921875,
|
|
"rewards/rejected": -4.58935546875,
|
|
"step": 400
|
|
},
|
|
{
|
|
"epoch": 0.19248826291079812,
|
|
"grad_norm": 642.6361751965826,
|
|
"learning_rate": 4.4887845592070945e-07,
|
|
"logits/chosen": -2.2421875,
|
|
"logits/rejected": -2.06640625,
|
|
"logps/chosen": -722.4000244140625,
|
|
"logps/rejected": -879.2000122070312,
|
|
"loss": 3.4391,
|
|
"rewards/accuracies": 0.3382812440395355,
|
|
"rewards/chosen": 1.3244140148162842,
|
|
"rewards/margins": 2.5455079078674316,
|
|
"rewards/rejected": -1.2185547351837158,
|
|
"step": 410
|
|
},
|
|
{
|
|
"epoch": 0.19718309859154928,
|
|
"grad_norm": 497.8577583794182,
|
|
"learning_rate": 4.462702138758477e-07,
|
|
"logits/chosen": -2.20703125,
|
|
"logits/rejected": -2.0960936546325684,
|
|
"logps/chosen": -778.4000244140625,
|
|
"logps/rejected": -868.0,
|
|
"loss": 2.8947,
|
|
"rewards/accuracies": 0.34062498807907104,
|
|
"rewards/chosen": 1.898046851158142,
|
|
"rewards/margins": 5.805468559265137,
|
|
"rewards/rejected": -3.9066405296325684,
|
|
"step": 420
|
|
},
|
|
{
|
|
"epoch": 0.20187793427230047,
|
|
"grad_norm": 890.313950643353,
|
|
"learning_rate": 4.436619718309859e-07,
|
|
"logits/chosen": -2.2109375,
|
|
"logits/rejected": -1.9265625476837158,
|
|
"logps/chosen": -750.0,
|
|
"logps/rejected": -900.7999877929688,
|
|
"loss": 3.8911,
|
|
"rewards/accuracies": 0.3515625,
|
|
"rewards/chosen": 1.587011694908142,
|
|
"rewards/margins": 2.849804639816284,
|
|
"rewards/rejected": -1.2580077648162842,
|
|
"step": 430
|
|
},
|
|
{
|
|
"epoch": 0.20657276995305165,
|
|
"grad_norm": 2009.4647327356392,
|
|
"learning_rate": 4.4105372978612415e-07,
|
|
"logits/chosen": -2.3203125,
|
|
"logits/rejected": -2.2328124046325684,
|
|
"logps/chosen": -769.2000122070312,
|
|
"logps/rejected": -882.4000244140625,
|
|
"loss": 4.6993,
|
|
"rewards/accuracies": 0.33281248807907104,
|
|
"rewards/chosen": 0.58251953125,
|
|
"rewards/margins": 1.8867676258087158,
|
|
"rewards/rejected": -1.3054687976837158,
|
|
"step": 440
|
|
},
|
|
{
|
|
"epoch": 0.2112676056338028,
|
|
"grad_norm": 989.3531885201104,
|
|
"learning_rate": 4.384454877412624e-07,
|
|
"logits/chosen": -2.38671875,
|
|
"logits/rejected": -2.067187547683716,
|
|
"logps/chosen": -738.0,
|
|
"logps/rejected": -975.2000122070312,
|
|
"loss": 3.7969,
|
|
"rewards/accuracies": 0.3414062559604645,
|
|
"rewards/chosen": 1.134667992591858,
|
|
"rewards/margins": 3.767578125,
|
|
"rewards/rejected": -2.6279296875,
|
|
"step": 450
|
|
},
|
|
{
|
|
"epoch": 0.215962441314554,
|
|
"grad_norm": 545.8369253651695,
|
|
"learning_rate": 4.358372456964006e-07,
|
|
"logits/chosen": -2.260937452316284,
|
|
"logits/rejected": -2.1070313453674316,
|
|
"logps/chosen": -742.0,
|
|
"logps/rejected": -873.5999755859375,
|
|
"loss": 3.7957,
|
|
"rewards/accuracies": 0.3453125059604645,
|
|
"rewards/chosen": 1.5535156726837158,
|
|
"rewards/margins": 3.942187547683716,
|
|
"rewards/rejected": -2.3919677734375,
|
|
"step": 460
|
|
},
|
|
{
|
|
"epoch": 0.22065727699530516,
|
|
"grad_norm": 527.7080770261233,
|
|
"learning_rate": 4.3322900365153886e-07,
|
|
"logits/chosen": -2.370312452316284,
|
|
"logits/rejected": -2.196093797683716,
|
|
"logps/chosen": -799.2000122070312,
|
|
"logps/rejected": -863.2000122070312,
|
|
"loss": 4.7648,
|
|
"rewards/accuracies": 0.3265624940395355,
|
|
"rewards/chosen": 0.34296876192092896,
|
|
"rewards/margins": 2.1605467796325684,
|
|
"rewards/rejected": -1.819580078125,
|
|
"step": 470
|
|
},
|
|
{
|
|
"epoch": 0.22535211267605634,
|
|
"grad_norm": 406.09742225980256,
|
|
"learning_rate": 4.306207616066771e-07,
|
|
"logits/chosen": -2.359375,
|
|
"logits/rejected": -2.190624952316284,
|
|
"logps/chosen": -735.2000122070312,
|
|
"logps/rejected": -839.5999755859375,
|
|
"loss": 4.1495,
|
|
"rewards/accuracies": 0.33515626192092896,
|
|
"rewards/chosen": 1.804296851158142,
|
|
"rewards/margins": 2.5394530296325684,
|
|
"rewards/rejected": -0.736889660358429,
|
|
"step": 480
|
|
},
|
|
{
|
|
"epoch": 0.2300469483568075,
|
|
"grad_norm": 5490.36357087384,
|
|
"learning_rate": 4.280125195618153e-07,
|
|
"logits/chosen": -2.325000047683716,
|
|
"logits/rejected": -2.061718702316284,
|
|
"logps/chosen": -743.5999755859375,
|
|
"logps/rejected": -869.5999755859375,
|
|
"loss": 3.8629,
|
|
"rewards/accuracies": 0.3382812440395355,
|
|
"rewards/chosen": 0.9203125238418579,
|
|
"rewards/margins": 2.908886671066284,
|
|
"rewards/rejected": -1.987890601158142,
|
|
"step": 490
|
|
},
|
|
{
|
|
"epoch": 0.2347417840375587,
|
|
"grad_norm": 997.4028144857181,
|
|
"learning_rate": 4.2540427751695357e-07,
|
|
"logits/chosen": -2.292187452316284,
|
|
"logits/rejected": -2.1773438453674316,
|
|
"logps/chosen": -766.7999877929688,
|
|
"logps/rejected": -850.7999877929688,
|
|
"loss": 4.0899,
|
|
"rewards/accuracies": 0.33515626192092896,
|
|
"rewards/chosen": 1.191796898841858,
|
|
"rewards/margins": 2.816601514816284,
|
|
"rewards/rejected": -1.6278808116912842,
|
|
"step": 500
|
|
},
|
|
{
|
|
"epoch": 0.23943661971830985,
|
|
"grad_norm": 561.7245894831859,
|
|
"learning_rate": 4.227960354720918e-07,
|
|
"logits/chosen": -2.3539061546325684,
|
|
"logits/rejected": -2.186718702316284,
|
|
"logps/chosen": -741.2000122070312,
|
|
"logps/rejected": -897.5999755859375,
|
|
"loss": 3.5953,
|
|
"rewards/accuracies": 0.3765625059604645,
|
|
"rewards/chosen": 2.7421875,
|
|
"rewards/margins": 4.625,
|
|
"rewards/rejected": -1.880468726158142,
|
|
"step": 510
|
|
},
|
|
{
|
|
"epoch": 0.24413145539906103,
|
|
"grad_norm": 499.4831548868703,
|
|
"learning_rate": 4.2018779342723e-07,
|
|
"logits/chosen": -2.2593750953674316,
|
|
"logits/rejected": -2.258593797683716,
|
|
"logps/chosen": -762.7999877929688,
|
|
"logps/rejected": -837.2000122070312,
|
|
"loss": 3.7502,
|
|
"rewards/accuracies": 0.3492187559604645,
|
|
"rewards/chosen": 1.6145508289337158,
|
|
"rewards/margins": 3.1888670921325684,
|
|
"rewards/rejected": -1.576171875,
|
|
"step": 520
|
|
},
|
|
{
|
|
"epoch": 0.24882629107981222,
|
|
"grad_norm": 455.056965245539,
|
|
"learning_rate": 4.1757955138236827e-07,
|
|
"logits/chosen": -2.319531202316284,
|
|
"logits/rejected": -2.0843749046325684,
|
|
"logps/chosen": -731.5999755859375,
|
|
"logps/rejected": -871.5999755859375,
|
|
"loss": 3.7579,
|
|
"rewards/accuracies": 0.3609375059604645,
|
|
"rewards/chosen": 1.85546875,
|
|
"rewards/margins": 4.625,
|
|
"rewards/rejected": -2.7668700218200684,
|
|
"step": 530
|
|
},
|
|
{
|
|
"epoch": 0.2535211267605634,
|
|
"grad_norm": 446.0832378408931,
|
|
"learning_rate": 4.149713093375065e-07,
|
|
"logits/chosen": -2.375,
|
|
"logits/rejected": -2.121875047683716,
|
|
"logps/chosen": -734.7999877929688,
|
|
"logps/rejected": -792.0,
|
|
"loss": 3.6603,
|
|
"rewards/accuracies": 0.3453125059604645,
|
|
"rewards/chosen": 1.8771483898162842,
|
|
"rewards/margins": 3.235156297683716,
|
|
"rewards/rejected": -1.357031226158142,
|
|
"step": 540
|
|
},
|
|
{
|
|
"epoch": 0.25821596244131456,
|
|
"grad_norm": 313.2168391651203,
|
|
"learning_rate": 4.123630672926447e-07,
|
|
"logits/chosen": -2.362499952316284,
|
|
"logits/rejected": -2.19140625,
|
|
"logps/chosen": -717.2000122070312,
|
|
"logps/rejected": -854.4000244140625,
|
|
"loss": 3.8799,
|
|
"rewards/accuracies": 0.37421876192092896,
|
|
"rewards/chosen": 2.354687452316284,
|
|
"rewards/margins": 5.006249904632568,
|
|
"rewards/rejected": -2.6578125953674316,
|
|
"step": 550
|
|
},
|
|
{
|
|
"epoch": 0.26291079812206575,
|
|
"grad_norm": 380.2190323980337,
|
|
"learning_rate": 4.09754825247783e-07,
|
|
"logits/chosen": -2.284374952316284,
|
|
"logits/rejected": -2.0414061546325684,
|
|
"logps/chosen": -749.2000122070312,
|
|
"logps/rejected": -858.0,
|
|
"loss": 3.701,
|
|
"rewards/accuracies": 0.3617187440395355,
|
|
"rewards/chosen": 2.235156297683716,
|
|
"rewards/margins": 4.295312404632568,
|
|
"rewards/rejected": -2.0572266578674316,
|
|
"step": 560
|
|
},
|
|
{
|
|
"epoch": 0.2676056338028169,
|
|
"grad_norm": 414.74954225225156,
|
|
"learning_rate": 4.0714658320292123e-07,
|
|
"logits/chosen": -2.448437452316284,
|
|
"logits/rejected": -2.260937452316284,
|
|
"logps/chosen": -726.0,
|
|
"logps/rejected": -850.4000244140625,
|
|
"loss": 3.8754,
|
|
"rewards/accuracies": 0.37968748807907104,
|
|
"rewards/chosen": 1.817968726158142,
|
|
"rewards/margins": 5.8125,
|
|
"rewards/rejected": -3.9976563453674316,
|
|
"step": 570
|
|
},
|
|
{
|
|
"epoch": 0.27230046948356806,
|
|
"grad_norm": 556.6860073670566,
|
|
"learning_rate": 4.045383411580595e-07,
|
|
"logits/chosen": -2.3218750953674316,
|
|
"logits/rejected": -2.110156297683716,
|
|
"logps/chosen": -725.2000122070312,
|
|
"logps/rejected": -862.0,
|
|
"loss": 4.1742,
|
|
"rewards/accuracies": 0.3656249940395355,
|
|
"rewards/chosen": 2.0453124046325684,
|
|
"rewards/margins": 4.602734565734863,
|
|
"rewards/rejected": -2.5546875,
|
|
"step": 580
|
|
},
|
|
{
|
|
"epoch": 0.27699530516431925,
|
|
"grad_norm": 514.6027456989814,
|
|
"learning_rate": 4.019300991131977e-07,
|
|
"logits/chosen": -2.190624952316284,
|
|
"logits/rejected": -2.106250047683716,
|
|
"logps/chosen": -732.7999877929688,
|
|
"logps/rejected": -848.7999877929688,
|
|
"loss": 3.3683,
|
|
"rewards/accuracies": 0.37890625,
|
|
"rewards/chosen": 2.628124952316284,
|
|
"rewards/margins": 4.798437595367432,
|
|
"rewards/rejected": -2.1703124046325684,
|
|
"step": 590
|
|
},
|
|
{
|
|
"epoch": 0.28169014084507044,
|
|
"grad_norm": 346.10997799515474,
|
|
"learning_rate": 3.9932185706833594e-07,
|
|
"logits/chosen": -2.315624952316284,
|
|
"logits/rejected": -2.0875000953674316,
|
|
"logps/chosen": -760.0,
|
|
"logps/rejected": -837.2000122070312,
|
|
"loss": 3.4386,
|
|
"rewards/accuracies": 0.37031251192092896,
|
|
"rewards/chosen": 2.828906297683716,
|
|
"rewards/margins": 5.220703125,
|
|
"rewards/rejected": -2.400097608566284,
|
|
"step": 600
|
|
},
|
|
{
|
|
"epoch": 0.2863849765258216,
|
|
"grad_norm": 499.7878762613954,
|
|
"learning_rate": 3.967136150234742e-07,
|
|
"logits/chosen": -2.225781202316284,
|
|
"logits/rejected": -2.104687452316284,
|
|
"logps/chosen": -771.5999755859375,
|
|
"logps/rejected": -854.0,
|
|
"loss": 4.0253,
|
|
"rewards/accuracies": 0.38203126192092896,
|
|
"rewards/chosen": 1.9755859375,
|
|
"rewards/margins": 4.127343654632568,
|
|
"rewards/rejected": -2.1502928733825684,
|
|
"step": 610
|
|
},
|
|
{
|
|
"epoch": 0.29107981220657275,
|
|
"grad_norm": 443.15171808264427,
|
|
"learning_rate": 3.941053729786124e-07,
|
|
"logits/chosen": -2.253124952316284,
|
|
"logits/rejected": -2.0484375953674316,
|
|
"logps/chosen": -768.7999877929688,
|
|
"logps/rejected": -872.7999877929688,
|
|
"loss": 3.6086,
|
|
"rewards/accuracies": 0.3984375,
|
|
"rewards/chosen": 2.6519532203674316,
|
|
"rewards/margins": 6.217187404632568,
|
|
"rewards/rejected": -3.5660157203674316,
|
|
"step": 620
|
|
},
|
|
{
|
|
"epoch": 0.29577464788732394,
|
|
"grad_norm": 848.1849647479538,
|
|
"learning_rate": 3.9149713093375064e-07,
|
|
"logits/chosen": -2.229687452316284,
|
|
"logits/rejected": -2.05078125,
|
|
"logps/chosen": -748.4000244140625,
|
|
"logps/rejected": -893.5999755859375,
|
|
"loss": 3.7016,
|
|
"rewards/accuracies": 0.37968748807907104,
|
|
"rewards/chosen": 2.3148436546325684,
|
|
"rewards/margins": 6.423437595367432,
|
|
"rewards/rejected": -4.104687690734863,
|
|
"step": 630
|
|
},
|
|
{
|
|
"epoch": 0.3004694835680751,
|
|
"grad_norm": 1575.8528024836098,
|
|
"learning_rate": 3.888888888888889e-07,
|
|
"logits/chosen": -2.2953124046325684,
|
|
"logits/rejected": -2.086718797683716,
|
|
"logps/chosen": -732.0,
|
|
"logps/rejected": -898.0,
|
|
"loss": 4.2902,
|
|
"rewards/accuracies": 0.3929687440395355,
|
|
"rewards/chosen": 2.0843749046325684,
|
|
"rewards/margins": 6.378125190734863,
|
|
"rewards/rejected": -4.292578220367432,
|
|
"step": 640
|
|
},
|
|
{
|
|
"epoch": 0.3051643192488263,
|
|
"grad_norm": 537.2576474944893,
|
|
"learning_rate": 3.862806468440271e-07,
|
|
"logits/chosen": -2.430468797683716,
|
|
"logits/rejected": -2.3203125,
|
|
"logps/chosen": -684.7999877929688,
|
|
"logps/rejected": -791.5999755859375,
|
|
"loss": 3.7996,
|
|
"rewards/accuracies": 0.3812499940395355,
|
|
"rewards/chosen": 1.9363281726837158,
|
|
"rewards/margins": 4.251562595367432,
|
|
"rewards/rejected": -2.314648389816284,
|
|
"step": 650
|
|
},
|
|
{
|
|
"epoch": 0.30985915492957744,
|
|
"grad_norm": 3342.812121912648,
|
|
"learning_rate": 3.8367240479916535e-07,
|
|
"logits/chosen": -2.3046875,
|
|
"logits/rejected": -2.260937452316284,
|
|
"logps/chosen": -733.2000122070312,
|
|
"logps/rejected": -809.5999755859375,
|
|
"loss": 3.9199,
|
|
"rewards/accuracies": 0.38359373807907104,
|
|
"rewards/chosen": 2.831249952316284,
|
|
"rewards/margins": 4.091406345367432,
|
|
"rewards/rejected": -1.263085961341858,
|
|
"step": 660
|
|
},
|
|
{
|
|
"epoch": 0.3145539906103286,
|
|
"grad_norm": 464.3621079865818,
|
|
"learning_rate": 3.810641627543036e-07,
|
|
"logits/chosen": -2.4375,
|
|
"logits/rejected": -2.171093702316284,
|
|
"logps/chosen": -734.4000244140625,
|
|
"logps/rejected": -854.4000244140625,
|
|
"loss": 3.6537,
|
|
"rewards/accuracies": 0.39453125,
|
|
"rewards/chosen": 2.4437499046325684,
|
|
"rewards/margins": 6.0390625,
|
|
"rewards/rejected": -3.594531297683716,
|
|
"step": 670
|
|
},
|
|
{
|
|
"epoch": 0.3192488262910798,
|
|
"grad_norm": 433.46451540511623,
|
|
"learning_rate": 3.784559207094418e-07,
|
|
"logits/chosen": -2.339062452316284,
|
|
"logits/rejected": -2.167187452316284,
|
|
"logps/chosen": -758.4000244140625,
|
|
"logps/rejected": -833.5999755859375,
|
|
"loss": 4.2747,
|
|
"rewards/accuracies": 0.38203126192092896,
|
|
"rewards/chosen": 2.720703125,
|
|
"rewards/margins": 4.618750095367432,
|
|
"rewards/rejected": -1.903173804283142,
|
|
"step": 680
|
|
},
|
|
{
|
|
"epoch": 0.323943661971831,
|
|
"grad_norm": 361.46426282793567,
|
|
"learning_rate": 3.7584767866458005e-07,
|
|
"logits/chosen": -2.2890625,
|
|
"logits/rejected": -2.128124952316284,
|
|
"logps/chosen": -749.2000122070312,
|
|
"logps/rejected": -830.0,
|
|
"loss": 4.0208,
|
|
"rewards/accuracies": 0.41093748807907104,
|
|
"rewards/chosen": 2.555859327316284,
|
|
"rewards/margins": 5.395703315734863,
|
|
"rewards/rejected": -2.8382811546325684,
|
|
"step": 690
|
|
},
|
|
{
|
|
"epoch": 0.3286384976525822,
|
|
"grad_norm": 485.60510363341695,
|
|
"learning_rate": 3.732394366197183e-07,
|
|
"logits/chosen": -2.342968702316284,
|
|
"logits/rejected": -2.120312452316284,
|
|
"logps/chosen": -780.7999877929688,
|
|
"logps/rejected": -820.7999877929688,
|
|
"loss": 4.662,
|
|
"rewards/accuracies": 0.40703123807907104,
|
|
"rewards/chosen": 2.749682664871216,
|
|
"rewards/margins": 5.974999904632568,
|
|
"rewards/rejected": -3.227343797683716,
|
|
"step": 700
|
|
},
|
|
{
|
|
"epoch": 0.3333333333333333,
|
|
"grad_norm": 361.06030358227264,
|
|
"learning_rate": 3.706311945748565e-07,
|
|
"logits/chosen": -2.3265624046325684,
|
|
"logits/rejected": -2.0859375,
|
|
"logps/chosen": -796.7999877929688,
|
|
"logps/rejected": -838.4000244140625,
|
|
"loss": 3.8436,
|
|
"rewards/accuracies": 0.39921873807907104,
|
|
"rewards/chosen": 2.5707030296325684,
|
|
"rewards/margins": 5.237109184265137,
|
|
"rewards/rejected": -2.668652296066284,
|
|
"step": 710
|
|
},
|
|
{
|
|
"epoch": 0.3380281690140845,
|
|
"grad_norm": 805.4099665248119,
|
|
"learning_rate": 3.6802295252999476e-07,
|
|
"logits/chosen": -2.25,
|
|
"logits/rejected": -1.954687476158142,
|
|
"logps/chosen": -771.5999755859375,
|
|
"logps/rejected": -909.5999755859375,
|
|
"loss": 4.5502,
|
|
"rewards/accuracies": 0.40546876192092896,
|
|
"rewards/chosen": 1.332421898841858,
|
|
"rewards/margins": 5.862500190734863,
|
|
"rewards/rejected": -4.526123046875,
|
|
"step": 720
|
|
},
|
|
{
|
|
"epoch": 0.3427230046948357,
|
|
"grad_norm": 836.6120700173155,
|
|
"learning_rate": 3.65414710485133e-07,
|
|
"logits/chosen": -2.3046875,
|
|
"logits/rejected": -2.2046875953674316,
|
|
"logps/chosen": -765.5999755859375,
|
|
"logps/rejected": -828.7999877929688,
|
|
"loss": 4.277,
|
|
"rewards/accuracies": 0.41874998807907104,
|
|
"rewards/chosen": 1.676367163658142,
|
|
"rewards/margins": 5.441601753234863,
|
|
"rewards/rejected": -3.758007764816284,
|
|
"step": 730
|
|
},
|
|
{
|
|
"epoch": 0.3474178403755869,
|
|
"grad_norm": 495.30095761214017,
|
|
"learning_rate": 3.6280646844027127e-07,
|
|
"logits/chosen": -2.2750000953674316,
|
|
"logits/rejected": -2.1953125,
|
|
"logps/chosen": -744.4000244140625,
|
|
"logps/rejected": -912.4000244140625,
|
|
"loss": 3.7158,
|
|
"rewards/accuracies": 0.4039062559604645,
|
|
"rewards/chosen": 2.78515625,
|
|
"rewards/margins": 6.276562690734863,
|
|
"rewards/rejected": -3.493945360183716,
|
|
"step": 740
|
|
},
|
|
{
|
|
"epoch": 0.352112676056338,
|
|
"grad_norm": 505.136183106453,
|
|
"learning_rate": 3.6019822639540947e-07,
|
|
"logits/chosen": -2.2421875,
|
|
"logits/rejected": -2.035937547683716,
|
|
"logps/chosen": -778.7999877929688,
|
|
"logps/rejected": -818.4000244140625,
|
|
"loss": 5.3975,
|
|
"rewards/accuracies": 0.40625,
|
|
"rewards/chosen": 0.9281250238418579,
|
|
"rewards/margins": 5.040625095367432,
|
|
"rewards/rejected": -4.099218845367432,
|
|
"step": 750
|
|
},
|
|
{
|
|
"epoch": 0.3568075117370892,
|
|
"grad_norm": 660.5281059191414,
|
|
"learning_rate": 3.575899843505477e-07,
|
|
"logits/chosen": -2.260937452316284,
|
|
"logits/rejected": -2.120312452316284,
|
|
"logps/chosen": -749.2000122070312,
|
|
"logps/rejected": -799.5999755859375,
|
|
"loss": 4.2791,
|
|
"rewards/accuracies": 0.4085937440395355,
|
|
"rewards/chosen": 2.5106444358825684,
|
|
"rewards/margins": 5.396874904632568,
|
|
"rewards/rejected": -2.874706983566284,
|
|
"step": 760
|
|
},
|
|
{
|
|
"epoch": 0.3615023474178404,
|
|
"grad_norm": 625.8963095469089,
|
|
"learning_rate": 3.5498174230568597e-07,
|
|
"logits/chosen": -2.2328124046325684,
|
|
"logits/rejected": -2.08984375,
|
|
"logps/chosen": -735.2000122070312,
|
|
"logps/rejected": -838.7999877929688,
|
|
"loss": 3.6004,
|
|
"rewards/accuracies": 0.36640626192092896,
|
|
"rewards/chosen": 2.1490235328674316,
|
|
"rewards/margins": 5.450781345367432,
|
|
"rewards/rejected": -3.302734375,
|
|
"step": 770
|
|
},
|
|
{
|
|
"epoch": 0.36619718309859156,
|
|
"grad_norm": 2236.100732457622,
|
|
"learning_rate": 3.5237350026082417e-07,
|
|
"logits/chosen": -2.3499999046325684,
|
|
"logits/rejected": -2.1421875953674316,
|
|
"logps/chosen": -751.2000122070312,
|
|
"logps/rejected": -796.7999877929688,
|
|
"loss": 4.5821,
|
|
"rewards/accuracies": 0.39140623807907104,
|
|
"rewards/chosen": 1.852148413658142,
|
|
"rewards/margins": 4.368750095367432,
|
|
"rewards/rejected": -2.520703077316284,
|
|
"step": 780
|
|
},
|
|
{
|
|
"epoch": 0.37089201877934275,
|
|
"grad_norm": 10858.709315184346,
|
|
"learning_rate": 3.497652582159624e-07,
|
|
"logits/chosen": -2.305468797683716,
|
|
"logits/rejected": -2.104687452316284,
|
|
"logps/chosen": -743.2000122070312,
|
|
"logps/rejected": -882.0,
|
|
"loss": 4.5831,
|
|
"rewards/accuracies": 0.40546876192092896,
|
|
"rewards/chosen": 3.220703125,
|
|
"rewards/margins": 5.546875,
|
|
"rewards/rejected": -2.3187499046325684,
|
|
"step": 790
|
|
},
|
|
{
|
|
"epoch": 0.3755868544600939,
|
|
"grad_norm": 521.9508389090934,
|
|
"learning_rate": 3.471570161711007e-07,
|
|
"logits/chosen": -2.2562499046325684,
|
|
"logits/rejected": -2.039843797683716,
|
|
"logps/chosen": -792.4000244140625,
|
|
"logps/rejected": -906.7999877929688,
|
|
"loss": 4.3751,
|
|
"rewards/accuracies": 0.3929687440395355,
|
|
"rewards/chosen": 2.5406250953674316,
|
|
"rewards/margins": 7.356249809265137,
|
|
"rewards/rejected": -4.815625190734863,
|
|
"step": 800
|
|
},
|
|
{
|
|
"epoch": 0.38028169014084506,
|
|
"grad_norm": 10833.720938010409,
|
|
"learning_rate": 3.445487741262389e-07,
|
|
"logits/chosen": -2.3421874046325684,
|
|
"logits/rejected": -2.176562547683716,
|
|
"logps/chosen": -728.7999877929688,
|
|
"logps/rejected": -824.4000244140625,
|
|
"loss": 4.7887,
|
|
"rewards/accuracies": 0.4085937440395355,
|
|
"rewards/chosen": 2.8812499046325684,
|
|
"rewards/margins": 6.7578125,
|
|
"rewards/rejected": -3.8828125,
|
|
"step": 810
|
|
},
|
|
{
|
|
"epoch": 0.38497652582159625,
|
|
"grad_norm": 486.2322033900268,
|
|
"learning_rate": 3.4194053208137713e-07,
|
|
"logits/chosen": -2.3375000953674316,
|
|
"logits/rejected": -2.203125,
|
|
"logps/chosen": -741.2000122070312,
|
|
"logps/rejected": -824.4000244140625,
|
|
"loss": 3.8345,
|
|
"rewards/accuracies": 0.39921873807907104,
|
|
"rewards/chosen": 2.8296875953674316,
|
|
"rewards/margins": 6.026171684265137,
|
|
"rewards/rejected": -3.198046922683716,
|
|
"step": 820
|
|
},
|
|
{
|
|
"epoch": 0.38967136150234744,
|
|
"grad_norm": 549.5772895819349,
|
|
"learning_rate": 3.393322900365154e-07,
|
|
"logits/chosen": -2.284374952316284,
|
|
"logits/rejected": -2.09375,
|
|
"logps/chosen": -696.0,
|
|
"logps/rejected": -800.4000244140625,
|
|
"loss": 3.6164,
|
|
"rewards/accuracies": 0.4124999940395355,
|
|
"rewards/chosen": 2.98046875,
|
|
"rewards/margins": 6.065625190734863,
|
|
"rewards/rejected": -3.092578172683716,
|
|
"step": 830
|
|
},
|
|
{
|
|
"epoch": 0.39436619718309857,
|
|
"grad_norm": 471.91907103487586,
|
|
"learning_rate": 3.367240479916536e-07,
|
|
"logits/chosen": -2.3304686546325684,
|
|
"logits/rejected": -2.128124952316284,
|
|
"logps/chosen": -785.5999755859375,
|
|
"logps/rejected": -877.5999755859375,
|
|
"loss": 4.2607,
|
|
"rewards/accuracies": 0.4156250059604645,
|
|
"rewards/chosen": 3.482617139816284,
|
|
"rewards/margins": 6.278124809265137,
|
|
"rewards/rejected": -2.8047852516174316,
|
|
"step": 840
|
|
},
|
|
{
|
|
"epoch": 0.39906103286384975,
|
|
"grad_norm": 416.94331200444964,
|
|
"learning_rate": 3.3411580594679184e-07,
|
|
"logits/chosen": -2.340625047683716,
|
|
"logits/rejected": -2.15625,
|
|
"logps/chosen": -728.7999877929688,
|
|
"logps/rejected": -836.0,
|
|
"loss": 4.2115,
|
|
"rewards/accuracies": 0.4078125059604645,
|
|
"rewards/chosen": 2.883984327316284,
|
|
"rewards/margins": 5.96484375,
|
|
"rewards/rejected": -3.0833983421325684,
|
|
"step": 850
|
|
},
|
|
{
|
|
"epoch": 0.40375586854460094,
|
|
"grad_norm": 410.68613216293846,
|
|
"learning_rate": 3.315075639019301e-07,
|
|
"logits/chosen": -2.2671875953674316,
|
|
"logits/rejected": -2.159374952316284,
|
|
"logps/chosen": -725.2000122070312,
|
|
"logps/rejected": -842.0,
|
|
"loss": 3.8122,
|
|
"rewards/accuracies": 0.42656248807907104,
|
|
"rewards/chosen": 3.440624952316284,
|
|
"rewards/margins": 7.651562690734863,
|
|
"rewards/rejected": -4.215624809265137,
|
|
"step": 860
|
|
},
|
|
{
|
|
"epoch": 0.4084507042253521,
|
|
"grad_norm": 348.24347301593184,
|
|
"learning_rate": 3.288993218570683e-07,
|
|
"logits/chosen": -2.266406297683716,
|
|
"logits/rejected": -2.034374952316284,
|
|
"logps/chosen": -738.4000244140625,
|
|
"logps/rejected": -829.5999755859375,
|
|
"loss": 4.2795,
|
|
"rewards/accuracies": 0.4164062440395355,
|
|
"rewards/chosen": 2.887500047683716,
|
|
"rewards/margins": 5.710156440734863,
|
|
"rewards/rejected": -2.8257813453674316,
|
|
"step": 870
|
|
},
|
|
{
|
|
"epoch": 0.4131455399061033,
|
|
"grad_norm": 462.9151988995226,
|
|
"learning_rate": 3.2629107981220654e-07,
|
|
"logits/chosen": -2.32421875,
|
|
"logits/rejected": -2.0875000953674316,
|
|
"logps/chosen": -764.4000244140625,
|
|
"logps/rejected": -850.4000244140625,
|
|
"loss": 3.855,
|
|
"rewards/accuracies": 0.4273437559604645,
|
|
"rewards/chosen": 2.641796827316284,
|
|
"rewards/margins": 7.3359375,
|
|
"rewards/rejected": -4.700585842132568,
|
|
"step": 880
|
|
},
|
|
{
|
|
"epoch": 0.41784037558685444,
|
|
"grad_norm": 633.2921618059665,
|
|
"learning_rate": 3.236828377673448e-07,
|
|
"logits/chosen": -2.2945313453674316,
|
|
"logits/rejected": -2.19921875,
|
|
"logps/chosen": -793.5999755859375,
|
|
"logps/rejected": -836.0,
|
|
"loss": 5.2967,
|
|
"rewards/accuracies": 0.40937501192092896,
|
|
"rewards/chosen": 2.6263670921325684,
|
|
"rewards/margins": 6.115624904632568,
|
|
"rewards/rejected": -3.4859375953674316,
|
|
"step": 890
|
|
},
|
|
{
|
|
"epoch": 0.4225352112676056,
|
|
"grad_norm": 370.9120195694783,
|
|
"learning_rate": 3.2107459572248305e-07,
|
|
"logits/chosen": -2.280468702316284,
|
|
"logits/rejected": -2.1578125953674316,
|
|
"logps/chosen": -759.5999755859375,
|
|
"logps/rejected": -836.7999877929688,
|
|
"loss": 3.9937,
|
|
"rewards/accuracies": 0.39921873807907104,
|
|
"rewards/chosen": 2.9664063453674316,
|
|
"rewards/margins": 6.659375190734863,
|
|
"rewards/rejected": -3.6904296875,
|
|
"step": 900
|
|
},
|
|
{
|
|
"epoch": 0.4272300469483568,
|
|
"grad_norm": 420.0254816718112,
|
|
"learning_rate": 3.1846635367762125e-07,
|
|
"logits/chosen": -2.19140625,
|
|
"logits/rejected": -2.0648436546325684,
|
|
"logps/chosen": -762.0,
|
|
"logps/rejected": -865.2000122070312,
|
|
"loss": 4.0513,
|
|
"rewards/accuracies": 0.41328126192092896,
|
|
"rewards/chosen": 2.8734374046325684,
|
|
"rewards/margins": 7.330859184265137,
|
|
"rewards/rejected": -4.443359375,
|
|
"step": 910
|
|
},
|
|
{
|
|
"epoch": 0.431924882629108,
|
|
"grad_norm": 516.9297080870588,
|
|
"learning_rate": 3.158581116327595e-07,
|
|
"logits/chosen": -2.203906297683716,
|
|
"logits/rejected": -2.1617188453674316,
|
|
"logps/chosen": -769.2000122070312,
|
|
"logps/rejected": -916.0,
|
|
"loss": 4.3426,
|
|
"rewards/accuracies": 0.41874998807907104,
|
|
"rewards/chosen": 3.044140577316284,
|
|
"rewards/margins": 7.994531154632568,
|
|
"rewards/rejected": -4.952343940734863,
|
|
"step": 920
|
|
},
|
|
{
|
|
"epoch": 0.43661971830985913,
|
|
"grad_norm": 309.58605122753823,
|
|
"learning_rate": 3.1324986958789775e-07,
|
|
"logits/chosen": -2.3031249046325684,
|
|
"logits/rejected": -2.19140625,
|
|
"logps/chosen": -708.0,
|
|
"logps/rejected": -827.2000122070312,
|
|
"loss": 3.6086,
|
|
"rewards/accuracies": 0.4203124940395355,
|
|
"rewards/chosen": 2.8515625,
|
|
"rewards/margins": 7.5078125,
|
|
"rewards/rejected": -4.663177490234375,
|
|
"step": 930
|
|
},
|
|
{
|
|
"epoch": 0.4413145539906103,
|
|
"grad_norm": 618.6834777646845,
|
|
"learning_rate": 3.1064162754303595e-07,
|
|
"logits/chosen": -2.223437547683716,
|
|
"logits/rejected": -2.096874952316284,
|
|
"logps/chosen": -779.2000122070312,
|
|
"logps/rejected": -881.2000122070312,
|
|
"loss": 4.2693,
|
|
"rewards/accuracies": 0.4281249940395355,
|
|
"rewards/chosen": 2.9593749046325684,
|
|
"rewards/margins": 7.512499809265137,
|
|
"rewards/rejected": -4.546875,
|
|
"step": 940
|
|
},
|
|
{
|
|
"epoch": 0.4460093896713615,
|
|
"grad_norm": 368.69666501852373,
|
|
"learning_rate": 3.080333854981742e-07,
|
|
"logits/chosen": -2.276562452316284,
|
|
"logits/rejected": -2.1421875953674316,
|
|
"logps/chosen": -751.2000122070312,
|
|
"logps/rejected": -802.7999877929688,
|
|
"loss": 4.4111,
|
|
"rewards/accuracies": 0.4039062559604645,
|
|
"rewards/chosen": 3.410937547683716,
|
|
"rewards/margins": 4.789843559265137,
|
|
"rewards/rejected": -1.383203148841858,
|
|
"step": 950
|
|
},
|
|
{
|
|
"epoch": 0.4507042253521127,
|
|
"grad_norm": 746.5695982997314,
|
|
"learning_rate": 3.0542514345331246e-07,
|
|
"logits/chosen": -2.2640624046325684,
|
|
"logits/rejected": -2.0609374046325684,
|
|
"logps/chosen": -750.7999877929688,
|
|
"logps/rejected": -913.5999755859375,
|
|
"loss": 4.3995,
|
|
"rewards/accuracies": 0.4312500059604645,
|
|
"rewards/chosen": 3.2171874046325684,
|
|
"rewards/margins": 7.142578125,
|
|
"rewards/rejected": -3.922656297683716,
|
|
"step": 960
|
|
},
|
|
{
|
|
"epoch": 0.45539906103286387,
|
|
"grad_norm": 760.3386454959006,
|
|
"learning_rate": 3.0281690140845066e-07,
|
|
"logits/chosen": -2.1781249046325684,
|
|
"logits/rejected": -2.125,
|
|
"logps/chosen": -806.7999877929688,
|
|
"logps/rejected": -877.5999755859375,
|
|
"loss": 4.693,
|
|
"rewards/accuracies": 0.42890626192092896,
|
|
"rewards/chosen": 3.000781297683716,
|
|
"rewards/margins": 5.631249904632568,
|
|
"rewards/rejected": -2.630859375,
|
|
"step": 970
|
|
},
|
|
{
|
|
"epoch": 0.460093896713615,
|
|
"grad_norm": 1085.1427603923041,
|
|
"learning_rate": 3.002086593635889e-07,
|
|
"logits/chosen": -2.3531250953674316,
|
|
"logits/rejected": -2.1343750953674316,
|
|
"logps/chosen": -760.4000244140625,
|
|
"logps/rejected": -890.7999877929688,
|
|
"loss": 4.008,
|
|
"rewards/accuracies": 0.44062501192092896,
|
|
"rewards/chosen": 3.5023436546325684,
|
|
"rewards/margins": 8.532812118530273,
|
|
"rewards/rejected": -5.0234375,
|
|
"step": 980
|
|
},
|
|
{
|
|
"epoch": 0.4647887323943662,
|
|
"grad_norm": 573.0020487487674,
|
|
"learning_rate": 2.9760041731872716e-07,
|
|
"logits/chosen": -2.4312500953674316,
|
|
"logits/rejected": -2.2718749046325684,
|
|
"logps/chosen": -692.4000244140625,
|
|
"logps/rejected": -826.7999877929688,
|
|
"loss": 3.5181,
|
|
"rewards/accuracies": 0.43046873807907104,
|
|
"rewards/chosen": 3.774218797683716,
|
|
"rewards/margins": 7.478125095367432,
|
|
"rewards/rejected": -3.699023485183716,
|
|
"step": 990
|
|
},
|
|
{
|
|
"epoch": 0.4694835680751174,
|
|
"grad_norm": 532.002820648586,
|
|
"learning_rate": 2.9499217527386536e-07,
|
|
"logits/chosen": -2.2906250953674316,
|
|
"logits/rejected": -2.1890625953674316,
|
|
"logps/chosen": -736.4000244140625,
|
|
"logps/rejected": -849.2000122070312,
|
|
"loss": 4.0951,
|
|
"rewards/accuracies": 0.40937501192092896,
|
|
"rewards/chosen": 2.7989745140075684,
|
|
"rewards/margins": 5.766797065734863,
|
|
"rewards/rejected": -2.973828077316284,
|
|
"step": 1000
|
|
},
|
|
{
|
|
"epoch": 0.47417840375586856,
|
|
"grad_norm": 747.5917247217351,
|
|
"learning_rate": 2.923839332290036e-07,
|
|
"logits/chosen": -2.25390625,
|
|
"logits/rejected": -2.061718702316284,
|
|
"logps/chosen": -732.0,
|
|
"logps/rejected": -876.0,
|
|
"loss": 3.9824,
|
|
"rewards/accuracies": 0.40703123807907104,
|
|
"rewards/chosen": 2.9765625,
|
|
"rewards/margins": 9.293749809265137,
|
|
"rewards/rejected": -6.31884765625,
|
|
"step": 1010
|
|
},
|
|
{
|
|
"epoch": 0.4788732394366197,
|
|
"grad_norm": 426.966215090325,
|
|
"learning_rate": 2.8977569118414187e-07,
|
|
"logits/chosen": -2.401562452316284,
|
|
"logits/rejected": -2.25390625,
|
|
"logps/chosen": -714.0,
|
|
"logps/rejected": -848.7999877929688,
|
|
"loss": 3.8824,
|
|
"rewards/accuracies": 0.4468750059604645,
|
|
"rewards/chosen": 2.970263719558716,
|
|
"rewards/margins": 9.743749618530273,
|
|
"rewards/rejected": -6.774316310882568,
|
|
"step": 1020
|
|
},
|
|
{
|
|
"epoch": 0.4835680751173709,
|
|
"grad_norm": 365.5440615260703,
|
|
"learning_rate": 2.8716744913928007e-07,
|
|
"logits/chosen": -2.3062500953674316,
|
|
"logits/rejected": -2.0921874046325684,
|
|
"logps/chosen": -740.4000244140625,
|
|
"logps/rejected": -818.0,
|
|
"loss": 4.139,
|
|
"rewards/accuracies": 0.4296875,
|
|
"rewards/chosen": 2.978710889816284,
|
|
"rewards/margins": 6.871874809265137,
|
|
"rewards/rejected": -3.887500047683716,
|
|
"step": 1030
|
|
},
|
|
{
|
|
"epoch": 0.48826291079812206,
|
|
"grad_norm": 447.88692510897215,
|
|
"learning_rate": 2.845592070944183e-07,
|
|
"logits/chosen": -2.2953124046325684,
|
|
"logits/rejected": -2.1304688453674316,
|
|
"logps/chosen": -746.4000244140625,
|
|
"logps/rejected": -881.5999755859375,
|
|
"loss": 3.8399,
|
|
"rewards/accuracies": 0.43828123807907104,
|
|
"rewards/chosen": 4.042187690734863,
|
|
"rewards/margins": 7.153124809265137,
|
|
"rewards/rejected": -3.12255859375,
|
|
"step": 1040
|
|
},
|
|
{
|
|
"epoch": 0.49295774647887325,
|
|
"grad_norm": 6937.32676845602,
|
|
"learning_rate": 2.819509650495566e-07,
|
|
"logits/chosen": -2.3499999046325684,
|
|
"logits/rejected": -2.0999999046325684,
|
|
"logps/chosen": -725.5999755859375,
|
|
"logps/rejected": -826.4000244140625,
|
|
"loss": 4.1211,
|
|
"rewards/accuracies": 0.42890626192092896,
|
|
"rewards/chosen": 3.339062452316284,
|
|
"rewards/margins": 7.1875,
|
|
"rewards/rejected": -3.845410108566284,
|
|
"step": 1050
|
|
},
|
|
{
|
|
"epoch": 0.49765258215962443,
|
|
"grad_norm": 354.53169803091913,
|
|
"learning_rate": 2.7934272300469483e-07,
|
|
"logits/chosen": -2.328125,
|
|
"logits/rejected": -2.065624952316284,
|
|
"logps/chosen": -773.5999755859375,
|
|
"logps/rejected": -839.5999755859375,
|
|
"loss": 3.6095,
|
|
"rewards/accuracies": 0.44453126192092896,
|
|
"rewards/chosen": 3.508593797683716,
|
|
"rewards/margins": 9.635937690734863,
|
|
"rewards/rejected": -6.143359184265137,
|
|
"step": 1060
|
|
},
|
|
{
|
|
"epoch": 0.5023474178403756,
|
|
"grad_norm": 456.8381360686394,
|
|
"learning_rate": 2.7673448095983303e-07,
|
|
"logits/chosen": -2.2671875953674316,
|
|
"logits/rejected": -2.109375,
|
|
"logps/chosen": -742.0,
|
|
"logps/rejected": -838.7999877929688,
|
|
"loss": 4.201,
|
|
"rewards/accuracies": 0.43671876192092896,
|
|
"rewards/chosen": 3.332812547683716,
|
|
"rewards/margins": 6.599023342132568,
|
|
"rewards/rejected": -3.263964891433716,
|
|
"step": 1070
|
|
},
|
|
{
|
|
"epoch": 0.5070422535211268,
|
|
"grad_norm": 418.36525536731284,
|
|
"learning_rate": 2.741262389149713e-07,
|
|
"logits/chosen": -2.2281250953674316,
|
|
"logits/rejected": -1.998437523841858,
|
|
"logps/chosen": -765.5999755859375,
|
|
"logps/rejected": -866.0,
|
|
"loss": 3.8887,
|
|
"rewards/accuracies": 0.44921875,
|
|
"rewards/chosen": 3.53125,
|
|
"rewards/margins": 8.162500381469727,
|
|
"rewards/rejected": -4.622656345367432,
|
|
"step": 1080
|
|
},
|
|
{
|
|
"epoch": 0.5117370892018779,
|
|
"grad_norm": 2210.2364862666204,
|
|
"learning_rate": 2.7151799687010953e-07,
|
|
"logits/chosen": -2.2640624046325684,
|
|
"logits/rejected": -1.997656226158142,
|
|
"logps/chosen": -756.0,
|
|
"logps/rejected": -870.0,
|
|
"loss": 4.5425,
|
|
"rewards/accuracies": 0.4242187440395355,
|
|
"rewards/chosen": 2.4625000953674316,
|
|
"rewards/margins": 7.379687309265137,
|
|
"rewards/rejected": -4.918749809265137,
|
|
"step": 1090
|
|
},
|
|
{
|
|
"epoch": 0.5164319248826291,
|
|
"grad_norm": 443.0955421475496,
|
|
"learning_rate": 2.6890975482524773e-07,
|
|
"logits/chosen": -2.246875047683716,
|
|
"logits/rejected": -2.121875047683716,
|
|
"logps/chosen": -792.4000244140625,
|
|
"logps/rejected": -862.7999877929688,
|
|
"loss": 4.1742,
|
|
"rewards/accuracies": 0.41484373807907104,
|
|
"rewards/chosen": 2.030078172683716,
|
|
"rewards/margins": 7.120312690734863,
|
|
"rewards/rejected": -5.096289157867432,
|
|
"step": 1100
|
|
},
|
|
{
|
|
"epoch": 0.5211267605633803,
|
|
"grad_norm": 518.8473672890916,
|
|
"learning_rate": 2.66301512780386e-07,
|
|
"logits/chosen": -2.3375000953674316,
|
|
"logits/rejected": -2.0015625953674316,
|
|
"logps/chosen": -702.0,
|
|
"logps/rejected": -856.4000244140625,
|
|
"loss": 4.1225,
|
|
"rewards/accuracies": 0.41796875,
|
|
"rewards/chosen": 2.481640577316284,
|
|
"rewards/margins": 9.381250381469727,
|
|
"rewards/rejected": -6.911913871765137,
|
|
"step": 1110
|
|
},
|
|
{
|
|
"epoch": 0.5258215962441315,
|
|
"grad_norm": 363.22129286758627,
|
|
"learning_rate": 2.6369327073552424e-07,
|
|
"logits/chosen": -2.26953125,
|
|
"logits/rejected": -2.1773438453674316,
|
|
"logps/chosen": -738.0,
|
|
"logps/rejected": -804.7999877929688,
|
|
"loss": 3.7051,
|
|
"rewards/accuracies": 0.4312500059604645,
|
|
"rewards/chosen": 3.3125,
|
|
"rewards/margins": 6.903124809265137,
|
|
"rewards/rejected": -3.5972657203674316,
|
|
"step": 1120
|
|
},
|
|
{
|
|
"epoch": 0.5305164319248826,
|
|
"grad_norm": 536.6899668393712,
|
|
"learning_rate": 2.6108502869066244e-07,
|
|
"logits/chosen": -2.268749952316284,
|
|
"logits/rejected": -2.147656202316284,
|
|
"logps/chosen": -779.5999755859375,
|
|
"logps/rejected": -889.2000122070312,
|
|
"loss": 5.0237,
|
|
"rewards/accuracies": 0.43437498807907104,
|
|
"rewards/chosen": 2.6285157203674316,
|
|
"rewards/margins": 9.725000381469727,
|
|
"rewards/rejected": -7.099999904632568,
|
|
"step": 1130
|
|
},
|
|
{
|
|
"epoch": 0.5352112676056338,
|
|
"grad_norm": 424.48547696453915,
|
|
"learning_rate": 2.584767866458007e-07,
|
|
"logits/chosen": -2.3359375,
|
|
"logits/rejected": -2.114062547683716,
|
|
"logps/chosen": -741.5999755859375,
|
|
"logps/rejected": -851.2000122070312,
|
|
"loss": 4.3225,
|
|
"rewards/accuracies": 0.43828123807907104,
|
|
"rewards/chosen": 3.164843797683716,
|
|
"rewards/margins": 8.550000190734863,
|
|
"rewards/rejected": -5.394384860992432,
|
|
"step": 1140
|
|
},
|
|
{
|
|
"epoch": 0.539906103286385,
|
|
"grad_norm": 335.36613670796896,
|
|
"learning_rate": 2.5586854460093895e-07,
|
|
"logits/chosen": -2.2421875,
|
|
"logits/rejected": -2.1343750953674316,
|
|
"logps/chosen": -756.4000244140625,
|
|
"logps/rejected": -864.0,
|
|
"loss": 3.6961,
|
|
"rewards/accuracies": 0.43671876192092896,
|
|
"rewards/chosen": 3.1851563453674316,
|
|
"rewards/margins": 7.917187690734863,
|
|
"rewards/rejected": -4.73046875,
|
|
"step": 1150
|
|
},
|
|
{
|
|
"epoch": 0.5446009389671361,
|
|
"grad_norm": 423.8242216632141,
|
|
"learning_rate": 2.5326030255607715e-07,
|
|
"logits/chosen": -2.3375000953674316,
|
|
"logits/rejected": -2.198437452316284,
|
|
"logps/chosen": -748.7999877929688,
|
|
"logps/rejected": -840.4000244140625,
|
|
"loss": 4.3461,
|
|
"rewards/accuracies": 0.4242187440395355,
|
|
"rewards/chosen": 3.552734375,
|
|
"rewards/margins": 7.80859375,
|
|
"rewards/rejected": -4.260156154632568,
|
|
"step": 1160
|
|
},
|
|
{
|
|
"epoch": 0.5492957746478874,
|
|
"grad_norm": 421.6433201369265,
|
|
"learning_rate": 2.506520605112154e-07,
|
|
"logits/chosen": -2.2874999046325684,
|
|
"logits/rejected": -2.171875,
|
|
"logps/chosen": -746.4000244140625,
|
|
"logps/rejected": -847.2000122070312,
|
|
"loss": 4.2687,
|
|
"rewards/accuracies": 0.4195312559604645,
|
|
"rewards/chosen": 3.234375,
|
|
"rewards/margins": 8.793749809265137,
|
|
"rewards/rejected": -5.568749904632568,
|
|
"step": 1170
|
|
},
|
|
{
|
|
"epoch": 0.5539906103286385,
|
|
"grad_norm": 360.7665284965791,
|
|
"learning_rate": 2.4804381846635365e-07,
|
|
"logits/chosen": -2.276562452316284,
|
|
"logits/rejected": -2.160937547683716,
|
|
"logps/chosen": -773.5999755859375,
|
|
"logps/rejected": -845.5999755859375,
|
|
"loss": 4.4898,
|
|
"rewards/accuracies": 0.4476562440395355,
|
|
"rewards/chosen": 2.789904832839966,
|
|
"rewards/margins": 7.818749904632568,
|
|
"rewards/rejected": -5.0234375,
|
|
"step": 1180
|
|
},
|
|
{
|
|
"epoch": 0.5586854460093896,
|
|
"grad_norm": 324.5642211557035,
|
|
"learning_rate": 2.454355764214919e-07,
|
|
"logits/chosen": -2.1953125,
|
|
"logits/rejected": -2.098437547683716,
|
|
"logps/chosen": -728.4000244140625,
|
|
"logps/rejected": -878.7999877929688,
|
|
"loss": 4.0787,
|
|
"rewards/accuracies": 0.4312500059604645,
|
|
"rewards/chosen": 3.4945311546325684,
|
|
"rewards/margins": 7.643750190734863,
|
|
"rewards/rejected": -4.148193359375,
|
|
"step": 1190
|
|
},
|
|
{
|
|
"epoch": 0.5633802816901409,
|
|
"grad_norm": 405.78556055418204,
|
|
"learning_rate": 2.4282733437663016e-07,
|
|
"logits/chosen": -2.28515625,
|
|
"logits/rejected": -2.2640624046325684,
|
|
"logps/chosen": -752.4000244140625,
|
|
"logps/rejected": -815.5999755859375,
|
|
"loss": 3.5846,
|
|
"rewards/accuracies": 0.4078125059604645,
|
|
"rewards/chosen": 3.9476561546325684,
|
|
"rewards/margins": 7.10546875,
|
|
"rewards/rejected": -3.1617188453674316,
|
|
"step": 1200
|
|
},
|
|
{
|
|
"epoch": 0.568075117370892,
|
|
"grad_norm": 373.50694707333025,
|
|
"learning_rate": 2.4021909233176836e-07,
|
|
"logits/chosen": -2.249218702316284,
|
|
"logits/rejected": -2.0703125,
|
|
"logps/chosen": -714.4000244140625,
|
|
"logps/rejected": -846.4000244140625,
|
|
"loss": 3.9689,
|
|
"rewards/accuracies": 0.4234375059604645,
|
|
"rewards/chosen": 3.852343797683716,
|
|
"rewards/margins": 8.003125190734863,
|
|
"rewards/rejected": -4.151757717132568,
|
|
"step": 1210
|
|
},
|
|
{
|
|
"epoch": 0.5727699530516432,
|
|
"grad_norm": 361.61775798266564,
|
|
"learning_rate": 2.376108502869066e-07,
|
|
"logits/chosen": -2.3921875953674316,
|
|
"logits/rejected": -2.3203125,
|
|
"logps/chosen": -757.2000122070312,
|
|
"logps/rejected": -862.7999877929688,
|
|
"loss": 5.3978,
|
|
"rewards/accuracies": 0.4351562559604645,
|
|
"rewards/chosen": 3.139453172683716,
|
|
"rewards/margins": 7.23046875,
|
|
"rewards/rejected": -4.089648246765137,
|
|
"step": 1220
|
|
},
|
|
{
|
|
"epoch": 0.5774647887323944,
|
|
"grad_norm": 420.8433662870921,
|
|
"learning_rate": 2.3500260824204484e-07,
|
|
"logits/chosen": -2.2914061546325684,
|
|
"logits/rejected": -2.1343750953674316,
|
|
"logps/chosen": -790.0,
|
|
"logps/rejected": -889.5999755859375,
|
|
"loss": 5.1055,
|
|
"rewards/accuracies": 0.44453126192092896,
|
|
"rewards/chosen": 2.4339842796325684,
|
|
"rewards/margins": 7.3828125,
|
|
"rewards/rejected": -4.946093559265137,
|
|
"step": 1230
|
|
},
|
|
{
|
|
"epoch": 0.5821596244131455,
|
|
"grad_norm": 382.45656168082024,
|
|
"learning_rate": 2.323943661971831e-07,
|
|
"logits/chosen": -2.385937452316284,
|
|
"logits/rejected": -2.1976561546325684,
|
|
"logps/chosen": -747.5999755859375,
|
|
"logps/rejected": -843.5999755859375,
|
|
"loss": 4.4922,
|
|
"rewards/accuracies": 0.4296875,
|
|
"rewards/chosen": 2.762500047683716,
|
|
"rewards/margins": 6.120312690734863,
|
|
"rewards/rejected": -3.3580079078674316,
|
|
"step": 1240
|
|
},
|
|
{
|
|
"epoch": 0.5868544600938967,
|
|
"grad_norm": 515.3933772313463,
|
|
"learning_rate": 2.2978612415232132e-07,
|
|
"logits/chosen": -2.332812547683716,
|
|
"logits/rejected": -2.149218797683716,
|
|
"logps/chosen": -803.5999755859375,
|
|
"logps/rejected": -885.2000122070312,
|
|
"loss": 3.8058,
|
|
"rewards/accuracies": 0.4359374940395355,
|
|
"rewards/chosen": 3.7613282203674316,
|
|
"rewards/margins": 9.2890625,
|
|
"rewards/rejected": -5.537499904632568,
|
|
"step": 1250
|
|
},
|
|
{
|
|
"epoch": 0.5915492957746479,
|
|
"grad_norm": 511.2143919673897,
|
|
"learning_rate": 2.2717788210745957e-07,
|
|
"logits/chosen": -2.229687452316284,
|
|
"logits/rejected": -2.0062499046325684,
|
|
"logps/chosen": -753.5999755859375,
|
|
"logps/rejected": -888.7999877929688,
|
|
"loss": 4.3063,
|
|
"rewards/accuracies": 0.41093748807907104,
|
|
"rewards/chosen": 2.5609374046325684,
|
|
"rewards/margins": 8.8359375,
|
|
"rewards/rejected": -6.276953220367432,
|
|
"step": 1260
|
|
},
|
|
{
|
|
"epoch": 0.596244131455399,
|
|
"grad_norm": 644.6610888055669,
|
|
"learning_rate": 2.245696400625978e-07,
|
|
"logits/chosen": -2.299999952316284,
|
|
"logits/rejected": -2.1695313453674316,
|
|
"logps/chosen": -763.5999755859375,
|
|
"logps/rejected": -851.2000122070312,
|
|
"loss": 5.2695,
|
|
"rewards/accuracies": 0.4351562559604645,
|
|
"rewards/chosen": 2.975390672683716,
|
|
"rewards/margins": 5.928124904632568,
|
|
"rewards/rejected": -2.943554639816284,
|
|
"step": 1270
|
|
},
|
|
{
|
|
"epoch": 0.6009389671361502,
|
|
"grad_norm": 321.7455984861024,
|
|
"learning_rate": 2.2196139801773602e-07,
|
|
"logits/chosen": -2.3125,
|
|
"logits/rejected": -2.089062452316284,
|
|
"logps/chosen": -718.0,
|
|
"logps/rejected": -816.7999877929688,
|
|
"loss": 3.8524,
|
|
"rewards/accuracies": 0.45390623807907104,
|
|
"rewards/chosen": 3.815624952316284,
|
|
"rewards/margins": 9.934374809265137,
|
|
"rewards/rejected": -6.112500190734863,
|
|
"step": 1280
|
|
},
|
|
{
|
|
"epoch": 0.6056338028169014,
|
|
"grad_norm": 590.758641860248,
|
|
"learning_rate": 2.1935315597287428e-07,
|
|
"logits/chosen": -2.2054686546325684,
|
|
"logits/rejected": -2.067187547683716,
|
|
"logps/chosen": -763.5999755859375,
|
|
"logps/rejected": -828.0,
|
|
"loss": 4.7133,
|
|
"rewards/accuracies": 0.4312500059604645,
|
|
"rewards/chosen": 3.06591796875,
|
|
"rewards/margins": 6.80078125,
|
|
"rewards/rejected": -3.7359375953674316,
|
|
"step": 1290
|
|
},
|
|
{
|
|
"epoch": 0.6103286384976526,
|
|
"grad_norm": 461.2463674049622,
|
|
"learning_rate": 2.167449139280125e-07,
|
|
"logits/chosen": -2.2320313453674316,
|
|
"logits/rejected": -2.065624952316284,
|
|
"logps/chosen": -764.7999877929688,
|
|
"logps/rejected": -869.2000122070312,
|
|
"loss": 4.3558,
|
|
"rewards/accuracies": 0.4085937440395355,
|
|
"rewards/chosen": 3.3890624046325684,
|
|
"rewards/margins": 7.753125190734863,
|
|
"rewards/rejected": -4.361132621765137,
|
|
"step": 1300
|
|
},
|
|
{
|
|
"epoch": 0.6150234741784038,
|
|
"grad_norm": 280.5584514541023,
|
|
"learning_rate": 2.1413667188315073e-07,
|
|
"logits/chosen": -2.3765625953674316,
|
|
"logits/rejected": -2.3765625953674316,
|
|
"logps/chosen": -736.4000244140625,
|
|
"logps/rejected": -794.7999877929688,
|
|
"loss": 4.0469,
|
|
"rewards/accuracies": 0.44999998807907104,
|
|
"rewards/chosen": 3.831249952316284,
|
|
"rewards/margins": 7.349218845367432,
|
|
"rewards/rejected": -3.514453172683716,
|
|
"step": 1310
|
|
},
|
|
{
|
|
"epoch": 0.6197183098591549,
|
|
"grad_norm": 495.67213017482453,
|
|
"learning_rate": 2.1152842983828898e-07,
|
|
"logits/chosen": -2.307812452316284,
|
|
"logits/rejected": -2.1429686546325684,
|
|
"logps/chosen": -758.0,
|
|
"logps/rejected": -843.2000122070312,
|
|
"loss": 4.2716,
|
|
"rewards/accuracies": 0.43437498807907104,
|
|
"rewards/chosen": 3.620312452316284,
|
|
"rewards/margins": 6.261328220367432,
|
|
"rewards/rejected": -2.6399903297424316,
|
|
"step": 1320
|
|
},
|
|
{
|
|
"epoch": 0.6244131455399061,
|
|
"grad_norm": 608.4249803472775,
|
|
"learning_rate": 2.089201877934272e-07,
|
|
"logits/chosen": -2.249218702316284,
|
|
"logits/rejected": -1.9968750476837158,
|
|
"logps/chosen": -788.0,
|
|
"logps/rejected": -903.5999755859375,
|
|
"loss": 4.9686,
|
|
"rewards/accuracies": 0.44843751192092896,
|
|
"rewards/chosen": 2.845898389816284,
|
|
"rewards/margins": 8.354687690734863,
|
|
"rewards/rejected": -5.509521484375,
|
|
"step": 1330
|
|
},
|
|
{
|
|
"epoch": 0.6291079812206573,
|
|
"grad_norm": 299.32455814618487,
|
|
"learning_rate": 2.0631194574856543e-07,
|
|
"logits/chosen": -2.4124999046325684,
|
|
"logits/rejected": -2.229687452316284,
|
|
"logps/chosen": -720.0,
|
|
"logps/rejected": -778.4000244140625,
|
|
"loss": 4.9123,
|
|
"rewards/accuracies": 0.4281249940395355,
|
|
"rewards/chosen": 3.137500047683716,
|
|
"rewards/margins": 7.143359184265137,
|
|
"rewards/rejected": -4.014062404632568,
|
|
"step": 1340
|
|
},
|
|
{
|
|
"epoch": 0.6338028169014085,
|
|
"grad_norm": 452.9560817569427,
|
|
"learning_rate": 2.0370370370370369e-07,
|
|
"logits/chosen": -2.190624952316284,
|
|
"logits/rejected": -2.05078125,
|
|
"logps/chosen": -761.5999755859375,
|
|
"logps/rejected": -901.5999755859375,
|
|
"loss": 4.2726,
|
|
"rewards/accuracies": 0.42890626192092896,
|
|
"rewards/chosen": 3.0523438453674316,
|
|
"rewards/margins": 8.350781440734863,
|
|
"rewards/rejected": -5.297436714172363,
|
|
"step": 1350
|
|
},
|
|
{
|
|
"epoch": 0.6384976525821596,
|
|
"grad_norm": 449.42369709396064,
|
|
"learning_rate": 2.010954616588419e-07,
|
|
"logits/chosen": -2.3203125,
|
|
"logits/rejected": -2.125,
|
|
"logps/chosen": -747.5999755859375,
|
|
"logps/rejected": -842.0,
|
|
"loss": 3.7139,
|
|
"rewards/accuracies": 0.45234376192092896,
|
|
"rewards/chosen": 4.084374904632568,
|
|
"rewards/margins": 7.849999904632568,
|
|
"rewards/rejected": -3.768749952316284,
|
|
"step": 1360
|
|
},
|
|
{
|
|
"epoch": 0.6431924882629108,
|
|
"grad_norm": 332.49243374349845,
|
|
"learning_rate": 1.9848721961398017e-07,
|
|
"logits/chosen": -2.3515625,
|
|
"logits/rejected": -2.2421875,
|
|
"logps/chosen": -755.2000122070312,
|
|
"logps/rejected": -823.5999755859375,
|
|
"loss": 4.3203,
|
|
"rewards/accuracies": 0.42265623807907104,
|
|
"rewards/chosen": 3.0523438453674316,
|
|
"rewards/margins": 8.746874809265137,
|
|
"rewards/rejected": -5.705468654632568,
|
|
"step": 1370
|
|
},
|
|
{
|
|
"epoch": 0.647887323943662,
|
|
"grad_norm": 475.9502903148322,
|
|
"learning_rate": 1.958789775691184e-07,
|
|
"logits/chosen": -2.3062500953674316,
|
|
"logits/rejected": -2.1273436546325684,
|
|
"logps/chosen": -734.7999877929688,
|
|
"logps/rejected": -838.4000244140625,
|
|
"loss": 4.0781,
|
|
"rewards/accuracies": 0.4078125059604645,
|
|
"rewards/chosen": 3.0570311546325684,
|
|
"rewards/margins": 8.162500381469727,
|
|
"rewards/rejected": -5.091864109039307,
|
|
"step": 1380
|
|
},
|
|
{
|
|
"epoch": 0.6525821596244131,
|
|
"grad_norm": 5461.133518431073,
|
|
"learning_rate": 1.9327073552425662e-07,
|
|
"logits/chosen": -2.340625047683716,
|
|
"logits/rejected": -2.125,
|
|
"logps/chosen": -716.4000244140625,
|
|
"logps/rejected": -803.5999755859375,
|
|
"loss": 3.605,
|
|
"rewards/accuracies": 0.46015626192092896,
|
|
"rewards/chosen": 3.429394483566284,
|
|
"rewards/margins": 8.985156059265137,
|
|
"rewards/rejected": -5.564843654632568,
|
|
"step": 1390
|
|
},
|
|
{
|
|
"epoch": 0.6572769953051644,
|
|
"grad_norm": 6885.443450742423,
|
|
"learning_rate": 1.906624934793949e-07,
|
|
"logits/chosen": -2.239062547683716,
|
|
"logits/rejected": -2.196093797683716,
|
|
"logps/chosen": -748.0,
|
|
"logps/rejected": -885.5999755859375,
|
|
"loss": 4.3522,
|
|
"rewards/accuracies": 0.4664062559604645,
|
|
"rewards/chosen": 2.715625047683716,
|
|
"rewards/margins": 9.5234375,
|
|
"rewards/rejected": -6.817968845367432,
|
|
"step": 1400
|
|
},
|
|
{
|
|
"epoch": 0.6619718309859155,
|
|
"grad_norm": 423.958432658112,
|
|
"learning_rate": 1.8805425143453312e-07,
|
|
"logits/chosen": -2.364062547683716,
|
|
"logits/rejected": -2.0406250953674316,
|
|
"logps/chosen": -725.2000122070312,
|
|
"logps/rejected": -925.5999755859375,
|
|
"loss": 3.816,
|
|
"rewards/accuracies": 0.44999998807907104,
|
|
"rewards/chosen": 3.6527342796325684,
|
|
"rewards/margins": 9.598437309265137,
|
|
"rewards/rejected": -5.954297065734863,
|
|
"step": 1410
|
|
},
|
|
{
|
|
"epoch": 0.6666666666666666,
|
|
"grad_norm": 452.36410328926013,
|
|
"learning_rate": 1.8544600938967138e-07,
|
|
"logits/chosen": -2.3226561546325684,
|
|
"logits/rejected": -2.225781202316284,
|
|
"logps/chosen": -734.7999877929688,
|
|
"logps/rejected": -816.0,
|
|
"loss": 3.5225,
|
|
"rewards/accuracies": 0.4375,
|
|
"rewards/chosen": 2.813281297683716,
|
|
"rewards/margins": 7.532812595367432,
|
|
"rewards/rejected": -4.720312595367432,
|
|
"step": 1420
|
|
},
|
|
{
|
|
"epoch": 0.6713615023474179,
|
|
"grad_norm": 361.3010901392399,
|
|
"learning_rate": 1.828377673448096e-07,
|
|
"logits/chosen": -2.2210936546325684,
|
|
"logits/rejected": -2.192187547683716,
|
|
"logps/chosen": -785.2000122070312,
|
|
"logps/rejected": -850.7999877929688,
|
|
"loss": 4.9125,
|
|
"rewards/accuracies": 0.4515624940395355,
|
|
"rewards/chosen": 1.88671875,
|
|
"rewards/margins": 7.845312595367432,
|
|
"rewards/rejected": -5.946875095367432,
|
|
"step": 1430
|
|
},
|
|
{
|
|
"epoch": 0.676056338028169,
|
|
"grad_norm": 426.6786872090034,
|
|
"learning_rate": 1.8022952529994783e-07,
|
|
"logits/chosen": -2.4078125953674316,
|
|
"logits/rejected": -2.2093749046325684,
|
|
"logps/chosen": -738.4000244140625,
|
|
"logps/rejected": -838.4000244140625,
|
|
"loss": 4.1243,
|
|
"rewards/accuracies": 0.43359375,
|
|
"rewards/chosen": 3.356250047683716,
|
|
"rewards/margins": 10.160937309265137,
|
|
"rewards/rejected": -6.804931640625,
|
|
"step": 1440
|
|
},
|
|
{
|
|
"epoch": 0.6807511737089202,
|
|
"grad_norm": 399.76098537479686,
|
|
"learning_rate": 1.7762128325508608e-07,
|
|
"logits/chosen": -2.354687452316284,
|
|
"logits/rejected": -2.19140625,
|
|
"logps/chosen": -769.2000122070312,
|
|
"logps/rejected": -922.4000244140625,
|
|
"loss": 4.9205,
|
|
"rewards/accuracies": 0.43828123807907104,
|
|
"rewards/chosen": 2.5992188453674316,
|
|
"rewards/margins": 9.056249618530273,
|
|
"rewards/rejected": -6.466015815734863,
|
|
"step": 1450
|
|
},
|
|
{
|
|
"epoch": 0.6854460093896714,
|
|
"grad_norm": 479.2473483667506,
|
|
"learning_rate": 1.750130412102243e-07,
|
|
"logits/chosen": -2.1734375953674316,
|
|
"logits/rejected": -2.106250047683716,
|
|
"logps/chosen": -813.5999755859375,
|
|
"logps/rejected": -881.5999755859375,
|
|
"loss": 5.3232,
|
|
"rewards/accuracies": 0.43671876192092896,
|
|
"rewards/chosen": 2.576171875,
|
|
"rewards/margins": 8.268750190734863,
|
|
"rewards/rejected": -5.686718940734863,
|
|
"step": 1460
|
|
},
|
|
{
|
|
"epoch": 0.6901408450704225,
|
|
"grad_norm": 415.9604208306485,
|
|
"learning_rate": 1.7240479916536254e-07,
|
|
"logits/chosen": -2.203125,
|
|
"logits/rejected": -2.145312547683716,
|
|
"logps/chosen": -739.5999755859375,
|
|
"logps/rejected": -874.7999877929688,
|
|
"loss": 4.774,
|
|
"rewards/accuracies": 0.4296875,
|
|
"rewards/chosen": 2.839062452316284,
|
|
"rewards/margins": 10.199999809265137,
|
|
"rewards/rejected": -7.369531154632568,
|
|
"step": 1470
|
|
},
|
|
{
|
|
"epoch": 0.6948356807511737,
|
|
"grad_norm": 297.0991105900083,
|
|
"learning_rate": 1.697965571205008e-07,
|
|
"logits/chosen": -2.4281249046325684,
|
|
"logits/rejected": -2.2562499046325684,
|
|
"logps/chosen": -732.0,
|
|
"logps/rejected": -794.7999877929688,
|
|
"loss": 3.9867,
|
|
"rewards/accuracies": 0.45390623807907104,
|
|
"rewards/chosen": 4.060156345367432,
|
|
"rewards/margins": 9.248437881469727,
|
|
"rewards/rejected": -5.195703029632568,
|
|
"step": 1480
|
|
},
|
|
{
|
|
"epoch": 0.6995305164319249,
|
|
"grad_norm": 559.399278140405,
|
|
"learning_rate": 1.6718831507563902e-07,
|
|
"logits/chosen": -2.364062547683716,
|
|
"logits/rejected": -2.270312547683716,
|
|
"logps/chosen": -746.4000244140625,
|
|
"logps/rejected": -792.0,
|
|
"loss": 3.8338,
|
|
"rewards/accuracies": 0.4296875,
|
|
"rewards/chosen": 3.2894530296325684,
|
|
"rewards/margins": 6.504687309265137,
|
|
"rewards/rejected": -3.2171874046325684,
|
|
"step": 1490
|
|
},
|
|
{
|
|
"epoch": 0.704225352112676,
|
|
"grad_norm": 318.92530656132584,
|
|
"learning_rate": 1.6458007303077727e-07,
|
|
"logits/chosen": -2.3125,
|
|
"logits/rejected": -2.234375,
|
|
"logps/chosen": -727.2000122070312,
|
|
"logps/rejected": -800.0,
|
|
"loss": 4.4147,
|
|
"rewards/accuracies": 0.4546875059604645,
|
|
"rewards/chosen": 3.581249952316284,
|
|
"rewards/margins": 7.589062690734863,
|
|
"rewards/rejected": -4.009961128234863,
|
|
"step": 1500
|
|
},
|
|
{
|
|
"epoch": 0.7089201877934272,
|
|
"grad_norm": 389.24309598289756,
|
|
"learning_rate": 1.619718309859155e-07,
|
|
"logits/chosen": -2.3890624046325684,
|
|
"logits/rejected": -2.116406202316284,
|
|
"logps/chosen": -756.7999877929688,
|
|
"logps/rejected": -879.5999755859375,
|
|
"loss": 4.2559,
|
|
"rewards/accuracies": 0.4585937559604645,
|
|
"rewards/chosen": 3.3070311546325684,
|
|
"rewards/margins": 9.251562118530273,
|
|
"rewards/rejected": -5.9453125,
|
|
"step": 1510
|
|
},
|
|
{
|
|
"epoch": 0.7136150234741784,
|
|
"grad_norm": 747.4741107805338,
|
|
"learning_rate": 1.5936358894105372e-07,
|
|
"logits/chosen": -2.3656249046325684,
|
|
"logits/rejected": -2.2109375,
|
|
"logps/chosen": -752.0,
|
|
"logps/rejected": -842.7999877929688,
|
|
"loss": 4.1086,
|
|
"rewards/accuracies": 0.46171873807907104,
|
|
"rewards/chosen": 4.715624809265137,
|
|
"rewards/margins": 8.232812881469727,
|
|
"rewards/rejected": -3.5199217796325684,
|
|
"step": 1520
|
|
},
|
|
{
|
|
"epoch": 0.7183098591549296,
|
|
"grad_norm": 1212.1237860042627,
|
|
"learning_rate": 1.5675534689619197e-07,
|
|
"logits/chosen": -2.370312452316284,
|
|
"logits/rejected": -2.2523436546325684,
|
|
"logps/chosen": -760.4000244140625,
|
|
"logps/rejected": -783.5999755859375,
|
|
"loss": 4.6827,
|
|
"rewards/accuracies": 0.421875,
|
|
"rewards/chosen": 2.787304639816284,
|
|
"rewards/margins": 5.881249904632568,
|
|
"rewards/rejected": -3.096874952316284,
|
|
"step": 1530
|
|
},
|
|
{
|
|
"epoch": 0.7230046948356808,
|
|
"grad_norm": 404.49052723472374,
|
|
"learning_rate": 1.541471048513302e-07,
|
|
"logits/chosen": -2.2578125,
|
|
"logits/rejected": -2.141406297683716,
|
|
"logps/chosen": -761.5999755859375,
|
|
"logps/rejected": -847.2000122070312,
|
|
"loss": 3.9635,
|
|
"rewards/accuracies": 0.46562498807907104,
|
|
"rewards/chosen": 4.581250190734863,
|
|
"rewards/margins": 8.944531440734863,
|
|
"rewards/rejected": -4.373242378234863,
|
|
"step": 1540
|
|
},
|
|
{
|
|
"epoch": 0.7276995305164319,
|
|
"grad_norm": 368.63816383458794,
|
|
"learning_rate": 1.5153886280646843e-07,
|
|
"logits/chosen": -2.265625,
|
|
"logits/rejected": -2.12109375,
|
|
"logps/chosen": -718.0,
|
|
"logps/rejected": -819.2000122070312,
|
|
"loss": 3.908,
|
|
"rewards/accuracies": 0.4312500059604645,
|
|
"rewards/chosen": 3.5171875953674316,
|
|
"rewards/margins": 7.840624809265137,
|
|
"rewards/rejected": -4.328515529632568,
|
|
"step": 1550
|
|
},
|
|
{
|
|
"epoch": 0.7323943661971831,
|
|
"grad_norm": 370.59996660930864,
|
|
"learning_rate": 1.4893062076160668e-07,
|
|
"logits/chosen": -2.4234375953674316,
|
|
"logits/rejected": -2.3453125953674316,
|
|
"logps/chosen": -737.2000122070312,
|
|
"logps/rejected": -839.2000122070312,
|
|
"loss": 4.1557,
|
|
"rewards/accuracies": 0.4351562559604645,
|
|
"rewards/chosen": 3.2367186546325684,
|
|
"rewards/margins": 7.862500190734863,
|
|
"rewards/rejected": -4.623046875,
|
|
"step": 1560
|
|
},
|
|
{
|
|
"epoch": 0.7370892018779343,
|
|
"grad_norm": 387.72493562249076,
|
|
"learning_rate": 1.463223787167449e-07,
|
|
"logits/chosen": -2.3046875,
|
|
"logits/rejected": -2.178906202316284,
|
|
"logps/chosen": -767.5999755859375,
|
|
"logps/rejected": -842.4000244140625,
|
|
"loss": 4.2895,
|
|
"rewards/accuracies": 0.44140625,
|
|
"rewards/chosen": 3.366406202316284,
|
|
"rewards/margins": 7.634375095367432,
|
|
"rewards/rejected": -4.2734375,
|
|
"step": 1570
|
|
},
|
|
{
|
|
"epoch": 0.7417840375586855,
|
|
"grad_norm": 498.16996435367645,
|
|
"learning_rate": 1.4371413667188313e-07,
|
|
"logits/chosen": -2.3984375,
|
|
"logits/rejected": -2.241406202316284,
|
|
"logps/chosen": -738.0,
|
|
"logps/rejected": -871.5999755859375,
|
|
"loss": 4.1729,
|
|
"rewards/accuracies": 0.4585937559604645,
|
|
"rewards/chosen": 4.318749904632568,
|
|
"rewards/margins": 8.706250190734863,
|
|
"rewards/rejected": -4.389062404632568,
|
|
"step": 1580
|
|
},
|
|
{
|
|
"epoch": 0.7464788732394366,
|
|
"grad_norm": 422.9669286102344,
|
|
"learning_rate": 1.4110589462702139e-07,
|
|
"logits/chosen": -2.4140625,
|
|
"logits/rejected": -2.2953124046325684,
|
|
"logps/chosen": -741.5999755859375,
|
|
"logps/rejected": -805.2000122070312,
|
|
"loss": 4.2679,
|
|
"rewards/accuracies": 0.4351562559604645,
|
|
"rewards/chosen": 2.90234375,
|
|
"rewards/margins": 6.705468654632568,
|
|
"rewards/rejected": -3.8003907203674316,
|
|
"step": 1590
|
|
},
|
|
{
|
|
"epoch": 0.7511737089201878,
|
|
"grad_norm": 334.8000258754459,
|
|
"learning_rate": 1.384976525821596e-07,
|
|
"logits/chosen": -2.4234375953674316,
|
|
"logits/rejected": -2.1500000953674316,
|
|
"logps/chosen": -754.7999877929688,
|
|
"logps/rejected": -885.5999755859375,
|
|
"loss": 3.4982,
|
|
"rewards/accuracies": 0.4585937559604645,
|
|
"rewards/chosen": 4.109375,
|
|
"rewards/margins": 10.159375190734863,
|
|
"rewards/rejected": -6.046093940734863,
|
|
"step": 1600
|
|
},
|
|
{
|
|
"epoch": 0.755868544600939,
|
|
"grad_norm": 450.17417017313414,
|
|
"learning_rate": 1.3588941053729787e-07,
|
|
"logits/chosen": -2.2203125953674316,
|
|
"logits/rejected": -2.180468797683716,
|
|
"logps/chosen": -767.2000122070312,
|
|
"logps/rejected": -880.4000244140625,
|
|
"loss": 4.1285,
|
|
"rewards/accuracies": 0.46015626192092896,
|
|
"rewards/chosen": 3.856250047683716,
|
|
"rewards/margins": 8.818750381469727,
|
|
"rewards/rejected": -4.962500095367432,
|
|
"step": 1610
|
|
},
|
|
{
|
|
"epoch": 0.7605633802816901,
|
|
"grad_norm": 394.65988882766004,
|
|
"learning_rate": 1.332811684924361e-07,
|
|
"logits/chosen": -2.2828125953674316,
|
|
"logits/rejected": -2.1703124046325684,
|
|
"logps/chosen": -784.0,
|
|
"logps/rejected": -851.5999755859375,
|
|
"loss": 4.6866,
|
|
"rewards/accuracies": 0.4625000059604645,
|
|
"rewards/chosen": 3.389843702316284,
|
|
"rewards/margins": 8.590624809265137,
|
|
"rewards/rejected": -5.199999809265137,
|
|
"step": 1620
|
|
},
|
|
{
|
|
"epoch": 0.7652582159624414,
|
|
"grad_norm": 424.3996351789177,
|
|
"learning_rate": 1.3067292644757432e-07,
|
|
"logits/chosen": -2.051562547683716,
|
|
"logits/rejected": -1.9343750476837158,
|
|
"logps/chosen": -787.5999755859375,
|
|
"logps/rejected": -906.4000244140625,
|
|
"loss": 3.7363,
|
|
"rewards/accuracies": 0.44921875,
|
|
"rewards/chosen": 4.043749809265137,
|
|
"rewards/margins": 9.721875190734863,
|
|
"rewards/rejected": -5.680468559265137,
|
|
"step": 1630
|
|
},
|
|
{
|
|
"epoch": 0.7699530516431925,
|
|
"grad_norm": 399.16976696133713,
|
|
"learning_rate": 1.2806468440271257e-07,
|
|
"logits/chosen": -2.301562547683716,
|
|
"logits/rejected": -2.171093702316284,
|
|
"logps/chosen": -730.4000244140625,
|
|
"logps/rejected": -824.4000244140625,
|
|
"loss": 3.8615,
|
|
"rewards/accuracies": 0.43281251192092896,
|
|
"rewards/chosen": 3.6468749046325684,
|
|
"rewards/margins": 7.654687404632568,
|
|
"rewards/rejected": -4.010156154632568,
|
|
"step": 1640
|
|
},
|
|
{
|
|
"epoch": 0.7746478873239436,
|
|
"grad_norm": 394.3409840047376,
|
|
"learning_rate": 1.254564423578508e-07,
|
|
"logits/chosen": -2.2992186546325684,
|
|
"logits/rejected": -2.0835938453674316,
|
|
"logps/chosen": -746.4000244140625,
|
|
"logps/rejected": -870.4000244140625,
|
|
"loss": 4.1122,
|
|
"rewards/accuracies": 0.4453125,
|
|
"rewards/chosen": 3.388671875,
|
|
"rewards/margins": 8.792187690734863,
|
|
"rewards/rejected": -5.399804592132568,
|
|
"step": 1650
|
|
},
|
|
{
|
|
"epoch": 0.7793427230046949,
|
|
"grad_norm": 409.67174776858445,
|
|
"learning_rate": 1.2284820031298902e-07,
|
|
"logits/chosen": -2.274218797683716,
|
|
"logits/rejected": -2.11328125,
|
|
"logps/chosen": -789.5999755859375,
|
|
"logps/rejected": -873.2000122070312,
|
|
"loss": 4.4931,
|
|
"rewards/accuracies": 0.48515623807907104,
|
|
"rewards/chosen": 3.9164061546325684,
|
|
"rewards/margins": 11.112500190734863,
|
|
"rewards/rejected": -7.189062595367432,
|
|
"step": 1660
|
|
},
|
|
{
|
|
"epoch": 0.784037558685446,
|
|
"grad_norm": 456.6688094864312,
|
|
"learning_rate": 1.2023995826812728e-07,
|
|
"logits/chosen": -2.34765625,
|
|
"logits/rejected": -2.1875,
|
|
"logps/chosen": -746.7999877929688,
|
|
"logps/rejected": -847.2000122070312,
|
|
"loss": 4.8528,
|
|
"rewards/accuracies": 0.44843751192092896,
|
|
"rewards/chosen": 3.4117188453674316,
|
|
"rewards/margins": 7.196875095367432,
|
|
"rewards/rejected": -3.789843797683716,
|
|
"step": 1670
|
|
},
|
|
{
|
|
"epoch": 0.7887323943661971,
|
|
"grad_norm": 1310.523103957443,
|
|
"learning_rate": 1.176317162232655e-07,
|
|
"logits/chosen": -2.3968749046325684,
|
|
"logits/rejected": -2.278125047683716,
|
|
"logps/chosen": -735.5999755859375,
|
|
"logps/rejected": -870.0,
|
|
"loss": 3.953,
|
|
"rewards/accuracies": 0.43359375,
|
|
"rewards/chosen": 3.5853514671325684,
|
|
"rewards/margins": 8.796875,
|
|
"rewards/rejected": -5.217187404632568,
|
|
"step": 1680
|
|
},
|
|
{
|
|
"epoch": 0.7934272300469484,
|
|
"grad_norm": 1724.0070361409773,
|
|
"learning_rate": 1.1502347417840374e-07,
|
|
"logits/chosen": -2.2890625,
|
|
"logits/rejected": -2.073437452316284,
|
|
"logps/chosen": -737.5999755859375,
|
|
"logps/rejected": -845.2000122070312,
|
|
"loss": 4.1449,
|
|
"rewards/accuracies": 0.4398437440395355,
|
|
"rewards/chosen": 3.5546875,
|
|
"rewards/margins": 8.342187881469727,
|
|
"rewards/rejected": -4.786035060882568,
|
|
"step": 1690
|
|
},
|
|
{
|
|
"epoch": 0.7981220657276995,
|
|
"grad_norm": 622.6768320134224,
|
|
"learning_rate": 1.1241523213354198e-07,
|
|
"logits/chosen": -2.3343749046325684,
|
|
"logits/rejected": -2.203906297683716,
|
|
"logps/chosen": -745.5999755859375,
|
|
"logps/rejected": -874.0,
|
|
"loss": 4.8334,
|
|
"rewards/accuracies": 0.40546876192092896,
|
|
"rewards/chosen": 3.487499952316284,
|
|
"rewards/margins": 7.410937309265137,
|
|
"rewards/rejected": -3.9242186546325684,
|
|
"step": 1700
|
|
},
|
|
{
|
|
"epoch": 0.8028169014084507,
|
|
"grad_norm": 439.70596258416913,
|
|
"learning_rate": 1.0980699008868022e-07,
|
|
"logits/chosen": -2.21484375,
|
|
"logits/rejected": -2.069531202316284,
|
|
"logps/chosen": -784.7999877929688,
|
|
"logps/rejected": -828.4000244140625,
|
|
"loss": 4.4736,
|
|
"rewards/accuracies": 0.4296875,
|
|
"rewards/chosen": 2.723437547683716,
|
|
"rewards/margins": 6.446875095367432,
|
|
"rewards/rejected": -3.7225584983825684,
|
|
"step": 1710
|
|
},
|
|
{
|
|
"epoch": 0.8075117370892019,
|
|
"grad_norm": 564.5328064267883,
|
|
"learning_rate": 1.0719874804381846e-07,
|
|
"logits/chosen": -2.3734374046325684,
|
|
"logits/rejected": -2.2874999046325684,
|
|
"logps/chosen": -732.7999877929688,
|
|
"logps/rejected": -771.5999755859375,
|
|
"loss": 4.1398,
|
|
"rewards/accuracies": 0.43671876192092896,
|
|
"rewards/chosen": 3.6968750953674316,
|
|
"rewards/margins": 5.946875095367432,
|
|
"rewards/rejected": -2.2494139671325684,
|
|
"step": 1720
|
|
},
|
|
{
|
|
"epoch": 0.812206572769953,
|
|
"grad_norm": 409.65102881247816,
|
|
"learning_rate": 1.045905059989567e-07,
|
|
"logits/chosen": -2.336718797683716,
|
|
"logits/rejected": -2.2398438453674316,
|
|
"logps/chosen": -711.2000122070312,
|
|
"logps/rejected": -772.4000244140625,
|
|
"loss": 4.5148,
|
|
"rewards/accuracies": 0.46875,
|
|
"rewards/chosen": 3.5531249046325684,
|
|
"rewards/margins": 9.015625,
|
|
"rewards/rejected": -5.461718559265137,
|
|
"step": 1730
|
|
},
|
|
{
|
|
"epoch": 0.8169014084507042,
|
|
"grad_norm": 748.089847869853,
|
|
"learning_rate": 1.0198226395409494e-07,
|
|
"logits/chosen": -2.176562547683716,
|
|
"logits/rejected": -2.0570311546325684,
|
|
"logps/chosen": -722.7999877929688,
|
|
"logps/rejected": -810.0,
|
|
"loss": 4.262,
|
|
"rewards/accuracies": 0.44062501192092896,
|
|
"rewards/chosen": 3.4253907203674316,
|
|
"rewards/margins": 7.412499904632568,
|
|
"rewards/rejected": -3.9896483421325684,
|
|
"step": 1740
|
|
},
|
|
{
|
|
"epoch": 0.8215962441314554,
|
|
"grad_norm": 328.6171942963381,
|
|
"learning_rate": 9.937402190923318e-08,
|
|
"logits/chosen": -2.225781202316284,
|
|
"logits/rejected": -2.0328125953674316,
|
|
"logps/chosen": -720.7999877929688,
|
|
"logps/rejected": -867.2000122070312,
|
|
"loss": 3.3953,
|
|
"rewards/accuracies": 0.46562498807907104,
|
|
"rewards/chosen": 4.143750190734863,
|
|
"rewards/margins": 10.809374809265137,
|
|
"rewards/rejected": -6.6640625,
|
|
"step": 1750
|
|
},
|
|
{
|
|
"epoch": 0.8262910798122066,
|
|
"grad_norm": 396.428360485738,
|
|
"learning_rate": 9.676577986437141e-08,
|
|
"logits/chosen": -2.2328124046325684,
|
|
"logits/rejected": -2.0335936546325684,
|
|
"logps/chosen": -781.5999755859375,
|
|
"logps/rejected": -849.5999755859375,
|
|
"loss": 4.593,
|
|
"rewards/accuracies": 0.4671874940395355,
|
|
"rewards/chosen": 3.284960985183716,
|
|
"rewards/margins": 9.949999809265137,
|
|
"rewards/rejected": -6.658715724945068,
|
|
"step": 1760
|
|
},
|
|
{
|
|
"epoch": 0.8309859154929577,
|
|
"grad_norm": 707.3465488581495,
|
|
"learning_rate": 9.415753781950965e-08,
|
|
"logits/chosen": -2.3203125,
|
|
"logits/rejected": -2.129687547683716,
|
|
"logps/chosen": -727.5999755859375,
|
|
"logps/rejected": -838.0,
|
|
"loss": 3.6212,
|
|
"rewards/accuracies": 0.4593749940395355,
|
|
"rewards/chosen": 4.284375190734863,
|
|
"rewards/margins": 9.942187309265137,
|
|
"rewards/rejected": -5.672265529632568,
|
|
"step": 1770
|
|
},
|
|
{
|
|
"epoch": 0.8356807511737089,
|
|
"grad_norm": 395.9535365243982,
|
|
"learning_rate": 9.154929577464789e-08,
|
|
"logits/chosen": -2.340625047683716,
|
|
"logits/rejected": -2.1484375,
|
|
"logps/chosen": -744.7999877929688,
|
|
"logps/rejected": -850.0,
|
|
"loss": 3.3361,
|
|
"rewards/accuracies": 0.4476562440395355,
|
|
"rewards/chosen": 3.8828125,
|
|
"rewards/margins": 12.431249618530273,
|
|
"rewards/rejected": -8.540624618530273,
|
|
"step": 1780
|
|
},
|
|
{
|
|
"epoch": 0.8403755868544601,
|
|
"grad_norm": 330.81263976850016,
|
|
"learning_rate": 8.894105372978613e-08,
|
|
"logits/chosen": -2.2320313453674316,
|
|
"logits/rejected": -2.2249999046325684,
|
|
"logps/chosen": -781.5999755859375,
|
|
"logps/rejected": -858.0,
|
|
"loss": 3.7592,
|
|
"rewards/accuracies": 0.45703125,
|
|
"rewards/chosen": 3.5062499046325684,
|
|
"rewards/margins": 8.868749618530273,
|
|
"rewards/rejected": -5.375,
|
|
"step": 1790
|
|
},
|
|
{
|
|
"epoch": 0.8450704225352113,
|
|
"grad_norm": 499.877437455524,
|
|
"learning_rate": 8.633281168492435e-08,
|
|
"logits/chosen": -2.3375000953674316,
|
|
"logits/rejected": -2.1929688453674316,
|
|
"logps/chosen": -744.4000244140625,
|
|
"logps/rejected": -806.7999877929688,
|
|
"loss": 4.4666,
|
|
"rewards/accuracies": 0.44843751192092896,
|
|
"rewards/chosen": 3.653515577316284,
|
|
"rewards/margins": 7.83203125,
|
|
"rewards/rejected": -4.1796875,
|
|
"step": 1800
|
|
},
|
|
{
|
|
"epoch": 0.8497652582159625,
|
|
"grad_norm": 396.3397631474569,
|
|
"learning_rate": 8.372456964006259e-08,
|
|
"logits/chosen": -2.2320313453674316,
|
|
"logits/rejected": -2.073437452316284,
|
|
"logps/chosen": -748.4000244140625,
|
|
"logps/rejected": -866.4000244140625,
|
|
"loss": 4.7405,
|
|
"rewards/accuracies": 0.4632812440395355,
|
|
"rewards/chosen": 2.959277391433716,
|
|
"rewards/margins": 9.856249809265137,
|
|
"rewards/rejected": -6.900000095367432,
|
|
"step": 1810
|
|
},
|
|
{
|
|
"epoch": 0.8544600938967136,
|
|
"grad_norm": 530.220629633677,
|
|
"learning_rate": 8.111632759520083e-08,
|
|
"logits/chosen": -2.200000047683716,
|
|
"logits/rejected": -2.078125,
|
|
"logps/chosen": -801.5999755859375,
|
|
"logps/rejected": -862.0,
|
|
"loss": 4.4322,
|
|
"rewards/accuracies": 0.46015626192092896,
|
|
"rewards/chosen": 3.389453172683716,
|
|
"rewards/margins": 7.340624809265137,
|
|
"rewards/rejected": -3.950000047683716,
|
|
"step": 1820
|
|
},
|
|
{
|
|
"epoch": 0.8591549295774648,
|
|
"grad_norm": 440.65956743164475,
|
|
"learning_rate": 7.850808555033907e-08,
|
|
"logits/chosen": -2.3421874046325684,
|
|
"logits/rejected": -2.1929688453674316,
|
|
"logps/chosen": -743.5999755859375,
|
|
"logps/rejected": -870.0,
|
|
"loss": 3.9501,
|
|
"rewards/accuracies": 0.44921875,
|
|
"rewards/chosen": 3.938281297683716,
|
|
"rewards/margins": 10.178125381469727,
|
|
"rewards/rejected": -6.231249809265137,
|
|
"step": 1830
|
|
},
|
|
{
|
|
"epoch": 0.863849765258216,
|
|
"grad_norm": 442.2402627617055,
|
|
"learning_rate": 7.58998435054773e-08,
|
|
"logits/chosen": -2.332812547683716,
|
|
"logits/rejected": -2.0648436546325684,
|
|
"logps/chosen": -751.2000122070312,
|
|
"logps/rejected": -862.7999877929688,
|
|
"loss": 4.3231,
|
|
"rewards/accuracies": 0.4476562440395355,
|
|
"rewards/chosen": 3.4214844703674316,
|
|
"rewards/margins": 9.412500381469727,
|
|
"rewards/rejected": -6.0,
|
|
"step": 1840
|
|
},
|
|
{
|
|
"epoch": 0.8685446009389671,
|
|
"grad_norm": 10384.624792404133,
|
|
"learning_rate": 7.329160146061554e-08,
|
|
"logits/chosen": -2.370312452316284,
|
|
"logits/rejected": -2.3296875953674316,
|
|
"logps/chosen": -734.0,
|
|
"logps/rejected": -836.7999877929688,
|
|
"loss": 4.7064,
|
|
"rewards/accuracies": 0.4710937440395355,
|
|
"rewards/chosen": 4.026562690734863,
|
|
"rewards/margins": 8.247655868530273,
|
|
"rewards/rejected": -4.221875190734863,
|
|
"step": 1850
|
|
},
|
|
{
|
|
"epoch": 0.8732394366197183,
|
|
"grad_norm": 673.9242811565833,
|
|
"learning_rate": 7.068335941575378e-08,
|
|
"logits/chosen": -2.1734375953674316,
|
|
"logits/rejected": -2.153125047683716,
|
|
"logps/chosen": -740.0,
|
|
"logps/rejected": -806.7999877929688,
|
|
"loss": 4.3453,
|
|
"rewards/accuracies": 0.45234376192092896,
|
|
"rewards/chosen": 3.1578125953674316,
|
|
"rewards/margins": 7.625,
|
|
"rewards/rejected": -4.46240234375,
|
|
"step": 1860
|
|
},
|
|
{
|
|
"epoch": 0.8779342723004695,
|
|
"grad_norm": 607.3789650567315,
|
|
"learning_rate": 6.807511737089202e-08,
|
|
"logits/chosen": -2.4046874046325684,
|
|
"logits/rejected": -2.196093797683716,
|
|
"logps/chosen": -739.2000122070312,
|
|
"logps/rejected": -842.0,
|
|
"loss": 4.1078,
|
|
"rewards/accuracies": 0.44921875,
|
|
"rewards/chosen": 3.664843797683716,
|
|
"rewards/margins": 9.100000381469727,
|
|
"rewards/rejected": -5.440625190734863,
|
|
"step": 1870
|
|
},
|
|
{
|
|
"epoch": 0.8826291079812206,
|
|
"grad_norm": 381.2094754272133,
|
|
"learning_rate": 6.546687532603024e-08,
|
|
"logits/chosen": -2.278125047683716,
|
|
"logits/rejected": -2.155468702316284,
|
|
"logps/chosen": -735.2000122070312,
|
|
"logps/rejected": -841.5999755859375,
|
|
"loss": 4.8184,
|
|
"rewards/accuracies": 0.45234376192092896,
|
|
"rewards/chosen": 3.410937547683716,
|
|
"rewards/margins": 7.385937690734863,
|
|
"rewards/rejected": -3.982617139816284,
|
|
"step": 1880
|
|
},
|
|
{
|
|
"epoch": 0.8873239436619719,
|
|
"grad_norm": 299.44549442697485,
|
|
"learning_rate": 6.285863328116848e-08,
|
|
"logits/chosen": -2.3125,
|
|
"logits/rejected": -2.11328125,
|
|
"logps/chosen": -682.0,
|
|
"logps/rejected": -856.7999877929688,
|
|
"loss": 3.7071,
|
|
"rewards/accuracies": 0.46875,
|
|
"rewards/chosen": 4.295312404632568,
|
|
"rewards/margins": 10.567187309265137,
|
|
"rewards/rejected": -6.278124809265137,
|
|
"step": 1890
|
|
},
|
|
{
|
|
"epoch": 0.892018779342723,
|
|
"grad_norm": 398.4727420176694,
|
|
"learning_rate": 6.025039123630672e-08,
|
|
"logits/chosen": -2.3140625953674316,
|
|
"logits/rejected": -2.067187547683716,
|
|
"logps/chosen": -762.0,
|
|
"logps/rejected": -814.0,
|
|
"loss": 3.5143,
|
|
"rewards/accuracies": 0.453125,
|
|
"rewards/chosen": 4.451562404632568,
|
|
"rewards/margins": 9.012499809265137,
|
|
"rewards/rejected": -4.559668064117432,
|
|
"step": 1900
|
|
},
|
|
{
|
|
"epoch": 0.8967136150234741,
|
|
"grad_norm": 337.5599153360449,
|
|
"learning_rate": 5.764214919144496e-08,
|
|
"logits/chosen": -2.35546875,
|
|
"logits/rejected": -2.22265625,
|
|
"logps/chosen": -770.7999877929688,
|
|
"logps/rejected": -814.4000244140625,
|
|
"loss": 3.8246,
|
|
"rewards/accuracies": 0.43828123807907104,
|
|
"rewards/chosen": 3.7601561546325684,
|
|
"rewards/margins": 8.521875381469727,
|
|
"rewards/rejected": -4.768456935882568,
|
|
"step": 1910
|
|
},
|
|
{
|
|
"epoch": 0.9014084507042254,
|
|
"grad_norm": 505.2498825423291,
|
|
"learning_rate": 5.50339071465832e-08,
|
|
"logits/chosen": -2.284374952316284,
|
|
"logits/rejected": -2.3359375,
|
|
"logps/chosen": -759.2000122070312,
|
|
"logps/rejected": -774.0,
|
|
"loss": 3.8311,
|
|
"rewards/accuracies": 0.4320312440395355,
|
|
"rewards/chosen": 3.6195311546325684,
|
|
"rewards/margins": 7.840624809265137,
|
|
"rewards/rejected": -4.220410346984863,
|
|
"step": 1920
|
|
},
|
|
{
|
|
"epoch": 0.9061032863849765,
|
|
"grad_norm": 327.8189786562515,
|
|
"learning_rate": 5.2425665101721436e-08,
|
|
"logits/chosen": -2.328125,
|
|
"logits/rejected": -2.1312499046325684,
|
|
"logps/chosen": -751.5999755859375,
|
|
"logps/rejected": -880.0,
|
|
"loss": 4.4301,
|
|
"rewards/accuracies": 0.4453125,
|
|
"rewards/chosen": 3.4945311546325684,
|
|
"rewards/margins": 9.09375,
|
|
"rewards/rejected": -5.605273246765137,
|
|
"step": 1930
|
|
},
|
|
{
|
|
"epoch": 0.9107981220657277,
|
|
"grad_norm": 347.13831647070737,
|
|
"learning_rate": 4.9817423056859675e-08,
|
|
"logits/chosen": -2.3531250953674316,
|
|
"logits/rejected": -2.1273436546325684,
|
|
"logps/chosen": -756.7999877929688,
|
|
"logps/rejected": -864.4000244140625,
|
|
"loss": 4.8472,
|
|
"rewards/accuracies": 0.44140625,
|
|
"rewards/chosen": 3.235058546066284,
|
|
"rewards/margins": 8.7421875,
|
|
"rewards/rejected": -5.513281345367432,
|
|
"step": 1940
|
|
},
|
|
{
|
|
"epoch": 0.9154929577464789,
|
|
"grad_norm": 298.01909557922994,
|
|
"learning_rate": 4.720918101199791e-08,
|
|
"logits/chosen": -2.139843702316284,
|
|
"logits/rejected": -1.9132812023162842,
|
|
"logps/chosen": -776.0,
|
|
"logps/rejected": -940.4000244140625,
|
|
"loss": 3.9666,
|
|
"rewards/accuracies": 0.4859375059604645,
|
|
"rewards/chosen": 4.353125095367432,
|
|
"rewards/margins": 11.556249618530273,
|
|
"rewards/rejected": -7.199999809265137,
|
|
"step": 1950
|
|
},
|
|
{
|
|
"epoch": 0.92018779342723,
|
|
"grad_norm": 294.48518836380975,
|
|
"learning_rate": 4.460093896713615e-08,
|
|
"logits/chosen": -2.3609375953674316,
|
|
"logits/rejected": -2.0921874046325684,
|
|
"logps/chosen": -745.2000122070312,
|
|
"logps/rejected": -877.2000122070312,
|
|
"loss": 4.4285,
|
|
"rewards/accuracies": 0.4554687440395355,
|
|
"rewards/chosen": 4.26953125,
|
|
"rewards/margins": 8.362500190734863,
|
|
"rewards/rejected": -4.088281154632568,
|
|
"step": 1960
|
|
},
|
|
{
|
|
"epoch": 0.9248826291079812,
|
|
"grad_norm": 445.20212474569666,
|
|
"learning_rate": 4.199269692227438e-08,
|
|
"logits/chosen": -2.3031249046325684,
|
|
"logits/rejected": -2.2281250953674316,
|
|
"logps/chosen": -743.5999755859375,
|
|
"logps/rejected": -786.7999877929688,
|
|
"loss": 4.7986,
|
|
"rewards/accuracies": 0.43281251192092896,
|
|
"rewards/chosen": 3.094531297683716,
|
|
"rewards/margins": 7.203125,
|
|
"rewards/rejected": -4.1083984375,
|
|
"step": 1970
|
|
},
|
|
{
|
|
"epoch": 0.9295774647887324,
|
|
"grad_norm": 454.23834025841336,
|
|
"learning_rate": 3.938445487741262e-08,
|
|
"logits/chosen": -2.1875,
|
|
"logits/rejected": -1.9148437976837158,
|
|
"logps/chosen": -774.0,
|
|
"logps/rejected": -861.2000122070312,
|
|
"loss": 4.2181,
|
|
"rewards/accuracies": 0.44453126192092896,
|
|
"rewards/chosen": 3.535937547683716,
|
|
"rewards/margins": 9.074999809265137,
|
|
"rewards/rejected": -5.548047065734863,
|
|
"step": 1980
|
|
},
|
|
{
|
|
"epoch": 0.9342723004694836,
|
|
"grad_norm": 392.17307200423585,
|
|
"learning_rate": 3.677621283255086e-08,
|
|
"logits/chosen": -2.4078125953674316,
|
|
"logits/rejected": -2.2710938453674316,
|
|
"logps/chosen": -766.7999877929688,
|
|
"logps/rejected": -832.7999877929688,
|
|
"loss": 3.8059,
|
|
"rewards/accuracies": 0.4429687559604645,
|
|
"rewards/chosen": 3.9078125953674316,
|
|
"rewards/margins": 7.943749904632568,
|
|
"rewards/rejected": -4.034375190734863,
|
|
"step": 1990
|
|
},
|
|
{
|
|
"epoch": 0.9389671361502347,
|
|
"grad_norm": 339.44226418791203,
|
|
"learning_rate": 3.41679707876891e-08,
|
|
"logits/chosen": -2.3687500953674316,
|
|
"logits/rejected": -2.2007813453674316,
|
|
"logps/chosen": -732.0,
|
|
"logps/rejected": -894.7999877929688,
|
|
"loss": 3.7702,
|
|
"rewards/accuracies": 0.45703125,
|
|
"rewards/chosen": 3.6273436546325684,
|
|
"rewards/margins": 11.837499618530273,
|
|
"rewards/rejected": -8.2109375,
|
|
"step": 2000
|
|
},
|
|
{
|
|
"epoch": 0.9436619718309859,
|
|
"grad_norm": 390.26770671950203,
|
|
"learning_rate": 3.155972874282733e-08,
|
|
"logits/chosen": -2.348437547683716,
|
|
"logits/rejected": -2.1382813453674316,
|
|
"logps/chosen": -724.0,
|
|
"logps/rejected": -818.4000244140625,
|
|
"loss": 4.599,
|
|
"rewards/accuracies": 0.43437498807907104,
|
|
"rewards/chosen": 3.7381348609924316,
|
|
"rewards/margins": 6.074999809265137,
|
|
"rewards/rejected": -2.3402342796325684,
|
|
"step": 2010
|
|
},
|
|
{
|
|
"epoch": 0.9483568075117371,
|
|
"grad_norm": 445.6413313616932,
|
|
"learning_rate": 2.8951486697965573e-08,
|
|
"logits/chosen": -2.297656297683716,
|
|
"logits/rejected": -2.19140625,
|
|
"logps/chosen": -767.5999755859375,
|
|
"logps/rejected": -858.4000244140625,
|
|
"loss": 4.0664,
|
|
"rewards/accuracies": 0.4437499940395355,
|
|
"rewards/chosen": 3.754687547683716,
|
|
"rewards/margins": 8.571874618530273,
|
|
"rewards/rejected": -4.814745903015137,
|
|
"step": 2020
|
|
},
|
|
{
|
|
"epoch": 0.9530516431924883,
|
|
"grad_norm": 443.53300480170367,
|
|
"learning_rate": 2.634324465310381e-08,
|
|
"logits/chosen": -2.375,
|
|
"logits/rejected": -2.143749952316284,
|
|
"logps/chosen": -735.2000122070312,
|
|
"logps/rejected": -838.0,
|
|
"loss": 3.9934,
|
|
"rewards/accuracies": 0.4546875059604645,
|
|
"rewards/chosen": 3.706249952316284,
|
|
"rewards/margins": 9.168749809265137,
|
|
"rewards/rejected": -5.470312595367432,
|
|
"step": 2030
|
|
},
|
|
{
|
|
"epoch": 0.9577464788732394,
|
|
"grad_norm": 548.3499825235165,
|
|
"learning_rate": 2.3735002608242045e-08,
|
|
"logits/chosen": -2.2359375953674316,
|
|
"logits/rejected": -2.0875000953674316,
|
|
"logps/chosen": -777.2000122070312,
|
|
"logps/rejected": -815.2000122070312,
|
|
"loss": 5.05,
|
|
"rewards/accuracies": 0.4546875059604645,
|
|
"rewards/chosen": 3.278515577316284,
|
|
"rewards/margins": 8.28125,
|
|
"rewards/rejected": -5.006640434265137,
|
|
"step": 2040
|
|
},
|
|
{
|
|
"epoch": 0.9624413145539906,
|
|
"grad_norm": 407.4024447296721,
|
|
"learning_rate": 2.1126760563380282e-08,
|
|
"logits/chosen": -2.3578124046325684,
|
|
"logits/rejected": -2.2281250953674316,
|
|
"logps/chosen": -746.7999877929688,
|
|
"logps/rejected": -779.5999755859375,
|
|
"loss": 4.1102,
|
|
"rewards/accuracies": 0.45703125,
|
|
"rewards/chosen": 3.909374952316284,
|
|
"rewards/margins": 8.640625,
|
|
"rewards/rejected": -4.732226371765137,
|
|
"step": 2050
|
|
},
|
|
{
|
|
"epoch": 0.9671361502347418,
|
|
"grad_norm": 332.7928982783245,
|
|
"learning_rate": 1.8518518518518518e-08,
|
|
"logits/chosen": -2.28515625,
|
|
"logits/rejected": -2.0492186546325684,
|
|
"logps/chosen": -773.2000122070312,
|
|
"logps/rejected": -853.5999755859375,
|
|
"loss": 4.4431,
|
|
"rewards/accuracies": 0.44062501192092896,
|
|
"rewards/chosen": 3.1953125,
|
|
"rewards/margins": 8.565625190734863,
|
|
"rewards/rejected": -5.377343654632568,
|
|
"step": 2060
|
|
},
|
|
{
|
|
"epoch": 0.971830985915493,
|
|
"grad_norm": 363.9987243736904,
|
|
"learning_rate": 1.5910276473656755e-08,
|
|
"logits/chosen": -2.2718749046325684,
|
|
"logits/rejected": -2.030468702316284,
|
|
"logps/chosen": -745.5999755859375,
|
|
"logps/rejected": -845.5999755859375,
|
|
"loss": 4.1346,
|
|
"rewards/accuracies": 0.4546875059604645,
|
|
"rewards/chosen": 4.180468559265137,
|
|
"rewards/margins": 8.756250381469727,
|
|
"rewards/rejected": -4.577734470367432,
|
|
"step": 2070
|
|
},
|
|
{
|
|
"epoch": 0.9765258215962441,
|
|
"grad_norm": 282.5830371548437,
|
|
"learning_rate": 1.3302034428794991e-08,
|
|
"logits/chosen": -2.4046874046325684,
|
|
"logits/rejected": -2.3187499046325684,
|
|
"logps/chosen": -708.0,
|
|
"logps/rejected": -786.0,
|
|
"loss": 4.9213,
|
|
"rewards/accuracies": 0.47265625,
|
|
"rewards/chosen": 3.958203077316284,
|
|
"rewards/margins": 8.301562309265137,
|
|
"rewards/rejected": -4.33984375,
|
|
"step": 2080
|
|
},
|
|
{
|
|
"epoch": 0.9812206572769953,
|
|
"grad_norm": 454.09493922137517,
|
|
"learning_rate": 1.0693792383933229e-08,
|
|
"logits/chosen": -2.344531297683716,
|
|
"logits/rejected": -2.282031297683716,
|
|
"logps/chosen": -765.2000122070312,
|
|
"logps/rejected": -849.2000122070312,
|
|
"loss": 4.1305,
|
|
"rewards/accuracies": 0.4749999940395355,
|
|
"rewards/chosen": 4.190625190734863,
|
|
"rewards/margins": 8.904687881469727,
|
|
"rewards/rejected": -4.71044921875,
|
|
"step": 2090
|
|
},
|
|
{
|
|
"epoch": 0.9859154929577465,
|
|
"grad_norm": 978.3858800511712,
|
|
"learning_rate": 8.085550339071465e-09,
|
|
"logits/chosen": -2.28125,
|
|
"logits/rejected": -2.147656202316284,
|
|
"logps/chosen": -782.7999877929688,
|
|
"logps/rejected": -878.7999877929688,
|
|
"loss": 4.6955,
|
|
"rewards/accuracies": 0.44062501192092896,
|
|
"rewards/chosen": 2.861328125,
|
|
"rewards/margins": 7.328125,
|
|
"rewards/rejected": -4.464062690734863,
|
|
"step": 2100
|
|
},
|
|
{
|
|
"epoch": 0.9906103286384976,
|
|
"grad_norm": 3846.6990654292445,
|
|
"learning_rate": 5.4773082942097025e-09,
|
|
"logits/chosen": -2.296875,
|
|
"logits/rejected": -2.1875,
|
|
"logps/chosen": -728.7999877929688,
|
|
"logps/rejected": -802.0,
|
|
"loss": 4.1061,
|
|
"rewards/accuracies": 0.45625001192092896,
|
|
"rewards/chosen": 3.919140577316284,
|
|
"rewards/margins": 10.006250381469727,
|
|
"rewards/rejected": -6.084741115570068,
|
|
"step": 2110
|
|
},
|
|
{
|
|
"epoch": 0.9953051643192489,
|
|
"grad_norm": 435.0445407316125,
|
|
"learning_rate": 2.8690662493479393e-09,
|
|
"logits/chosen": -2.2750000953674316,
|
|
"logits/rejected": -2.01953125,
|
|
"logps/chosen": -753.5999755859375,
|
|
"logps/rejected": -842.4000244140625,
|
|
"loss": 4.6775,
|
|
"rewards/accuracies": 0.4554687440395355,
|
|
"rewards/chosen": 3.551562547683716,
|
|
"rewards/margins": 8.192187309265137,
|
|
"rewards/rejected": -4.649218559265137,
|
|
"step": 2120
|
|
},
|
|
{
|
|
"epoch": 1.0,
|
|
"grad_norm": 379.7605837883577,
|
|
"learning_rate": 2.608242044861763e-10,
|
|
"logits/chosen": -2.31640625,
|
|
"logits/rejected": -2.184375047683716,
|
|
"logps/chosen": -742.7999877929688,
|
|
"logps/rejected": -854.7999877929688,
|
|
"loss": 3.895,
|
|
"rewards/accuracies": 0.47061213850975037,
|
|
"rewards/chosen": 4.150390625,
|
|
"rewards/margins": 12.193750381469727,
|
|
"rewards/rejected": -8.0390625,
|
|
"step": 2130
|
|
},
|
|
{
|
|
"epoch": 1.0,
|
|
"step": 2130,
|
|
"total_flos": 0.0,
|
|
"train_loss": 4.194636241930751,
|
|
"train_runtime": 7092.6143,
|
|
"train_samples_per_second": 38.432,
|
|
"train_steps_per_second": 0.3
|
|
}
|
|
],
|
|
"logging_steps": 10,
|
|
"max_steps": 2130,
|
|
"num_input_tokens_seen": 0,
|
|
"num_train_epochs": 1,
|
|
"save_steps": 200,
|
|
"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": 32,
|
|
"trial_name": null,
|
|
"trial_params": null
|
|
}
|