hZzy's picture
Model save
1eb0db0 verified
{
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 1.995276334435522,
"eval_steps": 50,
"global_step": 352,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"dpo_loss": 0.6931471824645996,
"epoch": 0.005668398677373642,
"grad_norm": 13413.646860161565,
"learning_rate": 1.3888888888888888e-07,
"logits": -1.3147305250167847,
"logps": -88.0877456665039,
"loss": 0.6931,
"objective": 0.6931471824645996,
"ranking_idealized": 0.5208333134651184,
"ranking_idealized_expo": 0.5208333134651184,
"ranking_simple": 0.5208333134651184,
"regularize": 0.6931471824645996,
"step": 1
},
{
"dpo_loss": 6.691188812255859,
"epoch": 0.02834199338686821,
"grad_norm": 25871.0368186516,
"learning_rate": 6.944444444444446e-07,
"logits": -1.3677693605422974,
"logps": -84.30184173583984,
"loss": 5.8145,
"objective": 6.691188812255859,
"ranking_idealized": 0.546875,
"ranking_idealized_expo": 0.546875,
"ranking_simple": 0.546875,
"regularize": 6.691188812255859,
"step": 5
},
{
"dpo_loss": 10.607693672180176,
"epoch": 0.05668398677373642,
"grad_norm": 23078.123366524644,
"learning_rate": 1.3888888888888892e-06,
"logits": -1.4513399600982666,
"logps": -83.46287536621094,
"loss": 12.4676,
"objective": 10.607693672180176,
"ranking_idealized": 0.5083333253860474,
"ranking_idealized_expo": 0.5083333253860474,
"ranking_simple": 0.512499988079071,
"regularize": 10.607693672180176,
"step": 10
},
{
"dpo_loss": 32.00202941894531,
"epoch": 0.08502598016060463,
"grad_norm": 18694.454058265223,
"learning_rate": 2.0833333333333334e-06,
"logits": -1.4168579578399658,
"logps": -83.43496704101562,
"loss": 22.9959,
"objective": 32.00202941894531,
"ranking_idealized": 0.5041666626930237,
"ranking_idealized_expo": 0.5041666626930237,
"ranking_simple": 0.5083333253860474,
"regularize": 32.00202941894531,
"step": 15
},
{
"dpo_loss": 45.9811897277832,
"epoch": 0.11336797354747284,
"grad_norm": 15943.429295919892,
"learning_rate": 2.7777777777777783e-06,
"logits": -1.3946782350540161,
"logps": -84.17158508300781,
"loss": 40.0365,
"objective": 45.9811897277832,
"ranking_idealized": 0.5375000238418579,
"ranking_idealized_expo": 0.5375000238418579,
"ranking_simple": 0.550000011920929,
"regularize": 45.9811897277832,
"step": 20
},
{
"dpo_loss": 47.39955520629883,
"epoch": 0.14170996693434104,
"grad_norm": 18118.727737282403,
"learning_rate": 3.4722222222222224e-06,
"logits": -1.4082823991775513,
"logps": -83.22772979736328,
"loss": 54.5353,
"objective": 47.39955520629883,
"ranking_idealized": 0.4958333373069763,
"ranking_idealized_expo": 0.4958333373069763,
"ranking_simple": 0.512499988079071,
"regularize": 47.39955520629883,
"step": 25
},
{
"dpo_loss": 78.01228332519531,
"epoch": 0.17005196032120926,
"grad_norm": 15166.879840931451,
"learning_rate": 4.166666666666667e-06,
"logits": -1.3809435367584229,
"logps": -83.88168334960938,
"loss": 66.4783,
"objective": 78.01228332519531,
"ranking_idealized": 0.5416666865348816,
"ranking_idealized_expo": 0.5416666865348816,
"ranking_simple": 0.5375000238418579,
"regularize": 78.01228332519531,
"step": 30
},
{
"dpo_loss": 79.99754333496094,
"epoch": 0.19839395370807747,
"grad_norm": 16063.368260477022,
"learning_rate": 4.861111111111111e-06,
"logits": -1.3859208822250366,
"logps": -82.78713989257812,
"loss": 66.6027,
"objective": 79.99754333496094,
"ranking_idealized": 0.5041666626930237,
"ranking_idealized_expo": 0.5041666626930237,
"ranking_simple": 0.5166666507720947,
"regularize": 79.99754333496094,
"step": 35
},
{
"dpo_loss": 67.1819076538086,
"epoch": 0.22673594709494568,
"grad_norm": 12400.159237202879,
"learning_rate": 4.998023493068255e-06,
"logits": -1.3994922637939453,
"logps": -82.4559555053711,
"loss": 81.1307,
"objective": 67.1819076538086,
"ranking_idealized": 0.5208333134651184,
"ranking_idealized_expo": 0.5208333134651184,
"ranking_simple": 0.5625,
"regularize": 67.1819076538086,
"step": 40
},
{
"dpo_loss": 89.6316146850586,
"epoch": 0.25507794048181387,
"grad_norm": 13052.93003597411,
"learning_rate": 4.989999289644993e-06,
"logits": -1.5318328142166138,
"logps": -87.26154327392578,
"loss": 93.1317,
"objective": 89.6316146850586,
"ranking_idealized": 0.5375000238418579,
"ranking_idealized_expo": 0.5375000238418579,
"ranking_simple": 0.5833333134651184,
"regularize": 89.6316146850586,
"step": 45
},
{
"dpo_loss": 78.09004974365234,
"epoch": 0.2834199338686821,
"grad_norm": 11250.619998937018,
"learning_rate": 4.975823666181256e-06,
"logits": -1.4500072002410889,
"logps": -85.9083480834961,
"loss": 89.5677,
"objective": 78.09004974365234,
"ranking_idealized": 0.42500001192092896,
"ranking_idealized_expo": 0.42500001192092896,
"ranking_simple": 0.5041666626930237,
"regularize": 78.09004974365234,
"step": 50
},
{
"epoch": 0.2834199338686821,
"eval_dpo_loss": 103.5480728149414,
"eval_logits": -1.4669513702392578,
"eval_logps": -93.47566986083984,
"eval_loss": 97.00978088378906,
"eval_objective": 103.5480728149414,
"eval_ranking_idealized": 0.5092975497245789,
"eval_ranking_idealized_expo": 0.5092975497245789,
"eval_ranking_simple": 0.5072314143180847,
"eval_regularize": 103.5480728149414,
"eval_runtime": 260.1107,
"eval_samples_per_second": 22.26,
"eval_steps_per_second": 0.93,
"step": 50
},
{
"dpo_loss": 119.75566101074219,
"epoch": 0.3117619272555503,
"grad_norm": 11967.354493984145,
"learning_rate": 4.955531642853404e-06,
"logits": -1.4773657321929932,
"logps": -87.53749084472656,
"loss": 103.6358,
"objective": 119.75566101074219,
"ranking_idealized": 0.5416666865348816,
"ranking_idealized_expo": 0.5416666865348816,
"ranking_simple": 0.5625,
"regularize": 119.75566101074219,
"step": 55
},
{
"dpo_loss": 103.13274383544922,
"epoch": 0.3401039206424185,
"grad_norm": 13660.537457669956,
"learning_rate": 4.929173350101025e-06,
"logits": -1.3649991750717163,
"logps": -84.92853546142578,
"loss": 108.801,
"objective": 103.13274383544922,
"ranking_idealized": 0.512499988079071,
"ranking_idealized_expo": 0.512499988079071,
"ranking_simple": 0.5625,
"regularize": 103.13274383544922,
"step": 60
},
{
"dpo_loss": 113.21337127685547,
"epoch": 0.3684459140292867,
"grad_norm": 11209.264550394939,
"learning_rate": 4.896813904782162e-06,
"logits": -1.261940836906433,
"logps": -87.40038299560547,
"loss": 120.3211,
"objective": 113.21337127685547,
"ranking_idealized": 0.5083333253860474,
"ranking_idealized_expo": 0.5083333253860474,
"ranking_simple": 0.5583333373069763,
"regularize": 113.21337127685547,
"step": 65
},
{
"dpo_loss": 95.49691009521484,
"epoch": 0.39678790741615494,
"grad_norm": 11680.213261821276,
"learning_rate": 4.858533249305337e-06,
"logits": -1.2914222478866577,
"logps": -89.73137664794922,
"loss": 116.7305,
"objective": 95.49691009521484,
"ranking_idealized": 0.5166666507720947,
"ranking_idealized_expo": 0.5166666507720947,
"ranking_simple": 0.5291666388511658,
"regularize": 95.49691009521484,
"step": 70
},
{
"dpo_loss": 102.52694702148438,
"epoch": 0.42512990080302315,
"grad_norm": 10698.837873548197,
"learning_rate": 4.814425954135786e-06,
"logits": -1.3049275875091553,
"logps": -86.52586364746094,
"loss": 118.3255,
"objective": 102.52694702148438,
"ranking_idealized": 0.5458333492279053,
"ranking_idealized_expo": 0.5458333492279053,
"ranking_simple": 0.5874999761581421,
"regularize": 102.52694702148438,
"step": 75
},
{
"dpo_loss": 117.94725036621094,
"epoch": 0.45347189418989137,
"grad_norm": 11104.97125441522,
"learning_rate": 4.764600984163809e-06,
"logits": -1.389280080795288,
"logps": -79.9697036743164,
"loss": 116.8339,
"objective": 117.94725036621094,
"ranking_idealized": 0.5166666507720947,
"ranking_idealized_expo": 0.5166666507720947,
"ranking_simple": 0.5791666507720947,
"regularize": 117.94725036621094,
"step": 80
},
{
"dpo_loss": 81.52342987060547,
"epoch": 0.4818138875767596,
"grad_norm": 9437.095124097401,
"learning_rate": 4.709181429512391e-06,
"logits": -1.402341365814209,
"logps": -74.89615631103516,
"loss": 96.3803,
"objective": 81.52342987060547,
"ranking_idealized": 0.5249999761581421,
"ranking_idealized_expo": 0.5249999761581421,
"ranking_simple": 0.5833333134651184,
"regularize": 81.52342987060547,
"step": 85
},
{
"dpo_loss": 123.97755432128906,
"epoch": 0.5101558809636277,
"grad_norm": 10670.5277421999,
"learning_rate": 4.648304201449153e-06,
"logits": -1.3095799684524536,
"logps": -75.67063903808594,
"loss": 113.352,
"objective": 123.97755432128906,
"ranking_idealized": 0.4791666567325592,
"ranking_idealized_expo": 0.4791666567325592,
"ranking_simple": 0.48750001192092896,
"regularize": 123.97755432128906,
"step": 90
},
{
"dpo_loss": 99.67278289794922,
"epoch": 0.538497874350496,
"grad_norm": 10466.418227348957,
"learning_rate": 4.582119694153834e-06,
"logits": -1.3693902492523193,
"logps": -73.55167388916016,
"loss": 100.734,
"objective": 99.67278289794922,
"ranking_idealized": 0.5416666865348816,
"ranking_idealized_expo": 0.5416666865348816,
"ranking_simple": 0.612500011920929,
"regularize": 99.67278289794922,
"step": 95
},
{
"dpo_loss": 81.45591735839844,
"epoch": 0.5668398677373642,
"grad_norm": 9021.660348254312,
"learning_rate": 4.510791413176912e-06,
"logits": -1.430596947669983,
"logps": -75.68968200683594,
"loss": 102.7372,
"objective": 81.45591735839844,
"ranking_idealized": 0.5416666865348816,
"ranking_idealized_expo": 0.5416666865348816,
"ranking_simple": 0.6041666865348816,
"regularize": 81.45591735839844,
"step": 100
},
{
"epoch": 0.5668398677373642,
"eval_dpo_loss": 169.08367919921875,
"eval_logits": -1.415890097618103,
"eval_logps": -79.38496398925781,
"eval_loss": 164.44805908203125,
"eval_objective": 169.08367919921875,
"eval_ranking_idealized": 0.5092975497245789,
"eval_ranking_idealized_expo": 0.5092975497245789,
"eval_ranking_simple": 0.5237603187561035,
"eval_regularize": 169.08367919921875,
"eval_runtime": 259.898,
"eval_samples_per_second": 22.278,
"eval_steps_per_second": 0.931,
"step": 100
},
{
"dpo_loss": 73.47309875488281,
"epoch": 0.5951818611242324,
"grad_norm": 8300.394502250436,
"learning_rate": 4.4344955715072344e-06,
"logits": -1.2836483716964722,
"logps": -74.31696319580078,
"loss": 98.1306,
"objective": 73.47309875488281,
"ranking_idealized": 0.5333333611488342,
"ranking_idealized_expo": 0.5333333611488342,
"ranking_simple": 0.6208333373069763,
"regularize": 73.47309875488281,
"step": 105
},
{
"dpo_loss": 113.51387023925781,
"epoch": 0.6235238545111006,
"grad_norm": 9297.62702465533,
"learning_rate": 4.353420654246546e-06,
"logits": -1.3452988862991333,
"logps": -70.39832305908203,
"loss": 101.6881,
"objective": 113.51387023925781,
"ranking_idealized": 0.44999998807907104,
"ranking_idealized_expo": 0.44999998807907104,
"ranking_simple": 0.5458333492279053,
"regularize": 113.51387023925781,
"step": 110
},
{
"dpo_loss": 100.13522338867188,
"epoch": 0.6518658478979689,
"grad_norm": 8180.786427844167,
"learning_rate": 4.267766952966369e-06,
"logits": -1.3274390697479248,
"logps": -69.5139389038086,
"loss": 95.7434,
"objective": 100.13522338867188,
"ranking_idealized": 0.4833333194255829,
"ranking_idealized_expo": 0.4833333194255829,
"ranking_simple": 0.5458333492279053,
"regularize": 100.13522338867188,
"step": 115
},
{
"dpo_loss": 135.6210174560547,
"epoch": 0.680207841284837,
"grad_norm": 8871.326596420884,
"learning_rate": 4.177746070897593e-06,
"logits": -1.2849870920181274,
"logps": -69.76378631591797,
"loss": 95.0051,
"objective": 135.6210174560547,
"ranking_idealized": 0.5333333611488342,
"ranking_idealized_expo": 0.5333333611488342,
"ranking_simple": 0.6000000238418579,
"regularize": 135.6210174560547,
"step": 120
},
{
"dpo_loss": 112.95771789550781,
"epoch": 0.7085498346717053,
"grad_norm": 8888.276865587854,
"learning_rate": 4.083580400175153e-06,
"logits": -1.2603411674499512,
"logps": -68.24909210205078,
"loss": 104.944,
"objective": 112.95771789550781,
"ranking_idealized": 0.550000011920929,
"ranking_idealized_expo": 0.550000011920929,
"ranking_simple": 0.6166666746139526,
"regularize": 112.95771789550781,
"step": 125
},
{
"dpo_loss": 91.826171875,
"epoch": 0.7368918280585735,
"grad_norm": 8626.309200082072,
"learning_rate": 3.985502572429277e-06,
"logits": -1.3430557250976562,
"logps": -67.98188018798828,
"loss": 94.7868,
"objective": 91.826171875,
"ranking_idealized": 0.44999998807907104,
"ranking_idealized_expo": 0.44999998807907104,
"ranking_simple": 0.5416666865348816,
"regularize": 91.826171875,
"step": 130
},
{
"dpo_loss": 79.87511444091797,
"epoch": 0.7652338214454416,
"grad_norm": 9413.731102456015,
"learning_rate": 3.8837548840805395e-06,
"logits": -1.2686209678649902,
"logps": -72.34845733642578,
"loss": 86.2141,
"objective": 79.87511444091797,
"ranking_idealized": 0.5291666388511658,
"ranking_idealized_expo": 0.5291666388511658,
"ranking_simple": 0.5791666507720947,
"regularize": 79.87511444091797,
"step": 135
},
{
"dpo_loss": 84.28844451904297,
"epoch": 0.7935758148323099,
"grad_norm": 7250.694243922807,
"learning_rate": 3.7785886977585562e-06,
"logits": -1.1709845066070557,
"logps": -72.76610565185547,
"loss": 82.4447,
"objective": 84.28844451904297,
"ranking_idealized": 0.5041666626930237,
"ranking_idealized_expo": 0.5041666626930237,
"ranking_simple": 0.574999988079071,
"regularize": 84.28844451904297,
"step": 140
},
{
"dpo_loss": 59.73074722290039,
"epoch": 0.821917808219178,
"grad_norm": 8021.4859376597515,
"learning_rate": 3.6702638213230344e-06,
"logits": -1.2125686407089233,
"logps": -73.49073791503906,
"loss": 76.9856,
"objective": 59.73074722290039,
"ranking_idealized": 0.4833333194255829,
"ranking_idealized_expo": 0.4833333194255829,
"ranking_simple": 0.550000011920929,
"regularize": 59.73074722290039,
"step": 145
},
{
"dpo_loss": 85.42176818847656,
"epoch": 0.8502598016060463,
"grad_norm": 7404.685943657088,
"learning_rate": 3.5590478660213214e-06,
"logits": -1.2063896656036377,
"logps": -72.71475982666016,
"loss": 86.6457,
"objective": 85.42176818847656,
"ranking_idealized": 0.5375000238418579,
"ranking_idealized_expo": 0.5375000238418579,
"ranking_simple": 0.5791666507720947,
"regularize": 85.42176818847656,
"step": 150
},
{
"epoch": 0.8502598016060463,
"eval_dpo_loss": 155.21026611328125,
"eval_logits": -1.2164032459259033,
"eval_logps": -80.36211395263672,
"eval_loss": 159.7296905517578,
"eval_objective": 155.21026611328125,
"eval_ranking_idealized": 0.5092975497245789,
"eval_ranking_idealized_expo": 0.5092975497245789,
"eval_ranking_simple": 0.5278925895690918,
"eval_regularize": 155.21026611328125,
"eval_runtime": 259.4907,
"eval_samples_per_second": 22.313,
"eval_steps_per_second": 0.933,
"step": 150
},
{
"dpo_loss": 90.06107330322266,
"epoch": 0.8786017949929145,
"grad_norm": 7592.300325394891,
"learning_rate": 3.4452155853680457e-06,
"logits": -1.129014253616333,
"logps": -74.393798828125,
"loss": 74.8455,
"objective": 90.06107330322266,
"ranking_idealized": 0.512499988079071,
"ranking_idealized_expo": 0.512499988079071,
"ranking_simple": 0.612500011920929,
"regularize": 90.06107330322266,
"step": 155
},
{
"dpo_loss": 80.76229095458984,
"epoch": 0.9069437883797827,
"grad_norm": 8007.48855908563,
"learning_rate": 3.32904819638017e-06,
"logits": -0.9675666093826294,
"logps": -77.41276550292969,
"loss": 83.7281,
"objective": 80.76229095458984,
"ranking_idealized": 0.4958333373069763,
"ranking_idealized_expo": 0.4958333373069763,
"ranking_simple": 0.5958333611488342,
"regularize": 80.76229095458984,
"step": 160
},
{
"dpo_loss": 66.69430541992188,
"epoch": 0.9352857817666509,
"grad_norm": 6958.931367168085,
"learning_rate": 3.2108326848442507e-06,
"logits": -1.0013041496276855,
"logps": -77.83885192871094,
"loss": 79.6961,
"objective": 66.69430541992188,
"ranking_idealized": 0.5208333134651184,
"ranking_idealized_expo": 0.5208333134651184,
"ranking_simple": 0.6000000238418579,
"regularize": 66.69430541992188,
"step": 165
},
{
"dpo_loss": 88.68855285644531,
"epoch": 0.9636277751535192,
"grad_norm": 7579.2896683801755,
"learning_rate": 3.090861096332263e-06,
"logits": -1.0412068367004395,
"logps": -77.83744812011719,
"loss": 81.023,
"objective": 88.68855285644531,
"ranking_idealized": 0.5083333253860474,
"ranking_idealized_expo": 0.5083333253860474,
"ranking_simple": 0.5791666507720947,
"regularize": 88.68855285644531,
"step": 170
},
{
"dpo_loss": 71.89635467529297,
"epoch": 0.9919697685403873,
"grad_norm": 7064.218511171905,
"learning_rate": 2.9694298147174566e-06,
"logits": -1.030120849609375,
"logps": -77.39608001708984,
"loss": 67.6747,
"objective": 71.89635467529297,
"ranking_idealized": 0.574999988079071,
"ranking_idealized_expo": 0.574999988079071,
"ranking_simple": 0.6458333134651184,
"regularize": 71.89635467529297,
"step": 175
},
{
"dpo_loss": 47.506874084472656,
"epoch": 1.0203117619272555,
"grad_norm": 5387.152679525541,
"learning_rate": 2.8468388299726714e-06,
"logits": -1.1444560289382935,
"logps": -77.81489562988281,
"loss": 45.6813,
"objective": 47.506874084472656,
"ranking_idealized": 0.5416666865348816,
"ranking_idealized_expo": 0.5416666865348816,
"ranking_simple": 0.625,
"regularize": 47.506874084472656,
"step": 180
},
{
"dpo_loss": 31.202707290649414,
"epoch": 1.0486537553141237,
"grad_norm": 6114.056834531026,
"learning_rate": 2.723390997059943e-06,
"logits": -1.0663727521896362,
"logps": -76.52796936035156,
"loss": 39.8195,
"objective": 31.202707290649414,
"ranking_idealized": 0.4833333194255829,
"ranking_idealized_expo": 0.4833333194255829,
"ranking_simple": 0.5791666507720947,
"regularize": 31.202707290649414,
"step": 185
},
{
"dpo_loss": 44.269371032714844,
"epoch": 1.076995748700992,
"grad_norm": 5270.211981458209,
"learning_rate": 2.599391287742315e-06,
"logits": -1.1319305896759033,
"logps": -76.61626434326172,
"loss": 40.779,
"objective": 44.269371032714844,
"ranking_idealized": 0.550000011920929,
"ranking_idealized_expo": 0.550000011920929,
"ranking_simple": 0.6583333611488342,
"regularize": 44.269371032714844,
"step": 190
},
{
"dpo_loss": 26.442119598388672,
"epoch": 1.10533774208786,
"grad_norm": 5494.940123258311,
"learning_rate": 2.4751460371661763e-06,
"logits": -1.1119343042373657,
"logps": -75.08609771728516,
"loss": 34.1025,
"objective": 26.442119598388672,
"ranking_idealized": 0.5666666626930237,
"ranking_idealized_expo": 0.5666666626930237,
"ranking_simple": 0.6625000238418579,
"regularize": 26.442119598388672,
"step": 195
},
{
"dpo_loss": 38.491580963134766,
"epoch": 1.1336797354747283,
"grad_norm": 5360.53985158099,
"learning_rate": 2.3509621870754505e-06,
"logits": -1.1571154594421387,
"logps": -74.1915283203125,
"loss": 40.1205,
"objective": 38.491580963134766,
"ranking_idealized": 0.5083333253860474,
"ranking_idealized_expo": 0.5083333253860474,
"ranking_simple": 0.6041666865348816,
"regularize": 38.491580963134766,
"step": 200
},
{
"epoch": 1.1336797354747283,
"eval_dpo_loss": 161.01707458496094,
"eval_logits": -1.175844430923462,
"eval_logps": -78.84459686279297,
"eval_loss": 164.80186462402344,
"eval_objective": 161.01707458496094,
"eval_ranking_idealized": 0.5092975497245789,
"eval_ranking_idealized_expo": 0.5092975497245789,
"eval_ranking_simple": 0.5247933864593506,
"eval_regularize": 161.01707458496094,
"eval_runtime": 263.5743,
"eval_samples_per_second": 21.967,
"eval_steps_per_second": 0.918,
"step": 200
},
{
"dpo_loss": 40.98146057128906,
"epoch": 1.1620217288615966,
"grad_norm": 4948.541109752384,
"learning_rate": 2.2271465275271985e-06,
"logits": -1.1139543056488037,
"logps": -74.31828308105469,
"loss": 36.4184,
"objective": 40.98146057128906,
"ranking_idealized": 0.48750001192092896,
"ranking_idealized_expo": 0.48750001192092896,
"ranking_simple": 0.5583333373069763,
"regularize": 40.98146057128906,
"step": 205
},
{
"dpo_loss": 33.11323165893555,
"epoch": 1.1903637222484649,
"grad_norm": 5703.102811198386,
"learning_rate": 2.1040049389819628e-06,
"logits": -1.088984727859497,
"logps": -74.35527038574219,
"loss": 37.3029,
"objective": 33.11323165893555,
"ranking_idealized": 0.4958333373069763,
"ranking_idealized_expo": 0.4958333373069763,
"ranking_simple": 0.5833333134651184,
"regularize": 33.11323165893555,
"step": 210
},
{
"dpo_loss": 39.365211486816406,
"epoch": 1.2187057156353331,
"grad_norm": 6043.230876979848,
"learning_rate": 1.9818416366412277e-06,
"logits": -1.0735052824020386,
"logps": -74.10447692871094,
"loss": 37.505,
"objective": 39.365211486816406,
"ranking_idealized": 0.4833333194255829,
"ranking_idealized_expo": 0.4833333194255829,
"ranking_simple": 0.612500011920929,
"regularize": 39.365211486816406,
"step": 215
},
{
"dpo_loss": 31.49414825439453,
"epoch": 1.2470477090222012,
"grad_norm": 5895.953296177759,
"learning_rate": 1.8609584188988135e-06,
"logits": -0.9674416780471802,
"logps": -74.39078521728516,
"loss": 39.9075,
"objective": 31.49414825439453,
"ranking_idealized": 0.6000000238418579,
"ranking_idealized_expo": 0.6000000238418579,
"ranking_simple": 0.6833333373069763,
"regularize": 31.49414825439453,
"step": 220
},
{
"dpo_loss": 44.06827926635742,
"epoch": 1.2753897024090695,
"grad_norm": 5452.729795480165,
"learning_rate": 1.7416539217628792e-06,
"logits": -1.0521718263626099,
"logps": -76.1507797241211,
"loss": 39.9429,
"objective": 44.06827926635742,
"ranking_idealized": 0.5291666388511658,
"ranking_idealized_expo": 0.5291666388511658,
"ranking_simple": 0.6583333611488342,
"regularize": 44.06827926635742,
"step": 225
},
{
"dpo_loss": 37.31355285644531,
"epoch": 1.3037316957959377,
"grad_norm": 5721.574963742396,
"learning_rate": 1.6242228810904393e-06,
"logits": -1.0492419004440308,
"logps": -75.18516540527344,
"loss": 41.1066,
"objective": 37.31355285644531,
"ranking_idealized": 0.5083333253860474,
"ranking_idealized_expo": 0.5083333253860474,
"ranking_simple": 0.5958333611488342,
"regularize": 37.31355285644531,
"step": 230
},
{
"dpo_loss": 19.191679000854492,
"epoch": 1.3320736891828058,
"grad_norm": 5669.232860831048,
"learning_rate": 1.508955404457015e-06,
"logits": -1.1627520322799683,
"logps": -72.39102172851562,
"loss": 34.805,
"objective": 19.191679000854492,
"ranking_idealized": 0.4833333194255829,
"ranking_idealized_expo": 0.4833333194255829,
"ranking_simple": 0.6291666626930237,
"regularize": 19.191679000854492,
"step": 235
},
{
"dpo_loss": 34.50796890258789,
"epoch": 1.360415682569674,
"grad_norm": 5948.081274276107,
"learning_rate": 1.3961362544602215e-06,
"logits": -1.1354502439498901,
"logps": -74.62174224853516,
"loss": 34.725,
"objective": 34.50796890258789,
"ranking_idealized": 0.4791666567325592,
"ranking_idealized_expo": 0.4791666567325592,
"ranking_simple": 0.6041666865348816,
"regularize": 34.50796890258789,
"step": 240
},
{
"dpo_loss": 40.89650344848633,
"epoch": 1.3887576759565423,
"grad_norm": 5428.624790806261,
"learning_rate": 1.2860441452278578e-06,
"logits": -1.1472022533416748,
"logps": -74.75707244873047,
"loss": 32.9306,
"objective": 40.89650344848633,
"ranking_idealized": 0.5166666507720947,
"ranking_idealized_expo": 0.5166666507720947,
"ranking_simple": 0.6083333492279053,
"regularize": 40.89650344848633,
"step": 245
},
{
"dpo_loss": 27.386281967163086,
"epoch": 1.4170996693434104,
"grad_norm": 5252.261360047395,
"learning_rate": 1.1789510538684524e-06,
"logits": -1.1399437189102173,
"logps": -74.77461242675781,
"loss": 40.2475,
"objective": 27.386281967163086,
"ranking_idealized": 0.47083333134651184,
"ranking_idealized_expo": 0.47083333134651184,
"ranking_simple": 0.6000000238418579,
"regularize": 27.386281967163086,
"step": 250
},
{
"epoch": 1.4170996693434104,
"eval_dpo_loss": 156.97756958007812,
"eval_logits": -1.2420457601547241,
"eval_logps": -80.06932830810547,
"eval_loss": 156.89581298828125,
"eval_objective": 156.97756958007812,
"eval_ranking_idealized": 0.5092975497245789,
"eval_ranking_idealized_expo": 0.5092975497245789,
"eval_ranking_simple": 0.5278925895690918,
"eval_regularize": 156.97756958007812,
"eval_runtime": 259.9252,
"eval_samples_per_second": 22.276,
"eval_steps_per_second": 0.931,
"step": 250
},
{
"dpo_loss": 32.30294418334961,
"epoch": 1.4454416627302786,
"grad_norm": 5037.556682913565,
"learning_rate": 1.0751215485652644e-06,
"logits": -1.146283745765686,
"logps": -75.65272521972656,
"loss": 32.8874,
"objective": 32.30294418334961,
"ranking_idealized": 0.5458333492279053,
"ranking_idealized_expo": 0.5458333492279053,
"ranking_simple": 0.6458333134651184,
"regularize": 32.30294418334961,
"step": 255
},
{
"dpo_loss": 31.96709632873535,
"epoch": 1.473783656117147,
"grad_norm": 5067.355799390521,
"learning_rate": 9.74812134973689e-07,
"logits": -1.1747548580169678,
"logps": -76.0610580444336,
"loss": 26.7271,
"objective": 31.96709632873535,
"ranking_idealized": 0.48750001192092896,
"ranking_idealized_expo": 0.48750001192092896,
"ranking_simple": 0.6041666865348816,
"regularize": 31.96709632873535,
"step": 260
},
{
"dpo_loss": 37.10657501220703,
"epoch": 1.5021256495040152,
"grad_norm": 6284.794884577889,
"learning_rate": 8.78270622536716e-07,
"logits": -1.1707472801208496,
"logps": -74.64857482910156,
"loss": 32.6747,
"objective": 37.10657501220703,
"ranking_idealized": 0.5041666626930237,
"ranking_idealized_expo": 0.5041666626930237,
"ranking_simple": 0.6208333373069763,
"regularize": 37.10657501220703,
"step": 265
},
{
"dpo_loss": 23.847713470458984,
"epoch": 1.5304676428908834,
"grad_norm": 5807.154647298787,
"learning_rate": 7.857355122839674e-07,
"logits": -1.1641870737075806,
"logps": -75.7414321899414,
"loss": 26.4138,
"objective": 23.847713470458984,
"ranking_idealized": 0.5041666626930237,
"ranking_idealized_expo": 0.5041666626930237,
"ranking_simple": 0.612500011920929,
"regularize": 23.847713470458984,
"step": 270
},
{
"dpo_loss": 44.59386444091797,
"epoch": 1.5588096362777515,
"grad_norm": 5794.329572664694,
"learning_rate": 6.974354076267081e-07,
"logits": -1.1526013612747192,
"logps": -75.73735046386719,
"loss": 29.7657,
"objective": 44.59386444091797,
"ranking_idealized": 0.5249999761581421,
"ranking_idealized_expo": 0.5249999761581421,
"ranking_simple": 0.6416666507720947,
"regularize": 44.59386444091797,
"step": 275
},
{
"dpo_loss": 17.95917320251465,
"epoch": 1.5871516296646198,
"grad_norm": 5034.884137017343,
"learning_rate": 6.135884496044245e-07,
"logits": -1.077419400215149,
"logps": -74.3064956665039,
"loss": 30.0354,
"objective": 17.95917320251465,
"ranking_idealized": 0.5041666626930237,
"ranking_idealized_expo": 0.5041666626930237,
"ranking_simple": 0.6041666865348816,
"regularize": 17.95917320251465,
"step": 280
},
{
"dpo_loss": 24.92406463623047,
"epoch": 1.615493623051488,
"grad_norm": 5064.875903206014,
"learning_rate": 5.344017779781835e-07,
"logits": -1.1750729084014893,
"logps": -75.65376281738281,
"loss": 28.5295,
"objective": 24.92406463623047,
"ranking_idealized": 0.5166666507720947,
"ranking_idealized_expo": 0.5166666507720947,
"ranking_simple": 0.6333333253860474,
"regularize": 24.92406463623047,
"step": 285
},
{
"dpo_loss": 33.37800598144531,
"epoch": 1.643835616438356,
"grad_norm": 5057.655323901423,
"learning_rate": 4.6007101950209827e-07,
"logits": -1.1915957927703857,
"logps": -73.43339538574219,
"loss": 27.6645,
"objective": 33.37800598144531,
"ranking_idealized": 0.5625,
"ranking_idealized_expo": 0.5625,
"ranking_simple": 0.6291666626930237,
"regularize": 33.37800598144531,
"step": 290
},
{
"dpo_loss": 17.80238151550293,
"epoch": 1.6721776098252243,
"grad_norm": 5184.197441213302,
"learning_rate": 3.907798046371139e-07,
"logits": -1.1149822473526,
"logps": -72.55225372314453,
"loss": 25.6518,
"objective": 17.80238151550293,
"ranking_idealized": 0.4791666567325592,
"ranking_idealized_expo": 0.4791666567325592,
"ranking_simple": 0.6041666865348816,
"regularize": 17.80238151550293,
"step": 295
},
{
"dpo_loss": 16.620241165161133,
"epoch": 1.7005196032120926,
"grad_norm": 4977.226734001169,
"learning_rate": 3.266993139010438e-07,
"logits": -1.1575456857681274,
"logps": -74.20482635498047,
"loss": 24.0056,
"objective": 16.620241165161133,
"ranking_idealized": 0.574999988079071,
"ranking_idealized_expo": 0.574999988079071,
"ranking_simple": 0.6791666746139526,
"regularize": 16.620241165161133,
"step": 300
},
{
"epoch": 1.7005196032120926,
"eval_dpo_loss": 153.7111358642578,
"eval_logits": -1.1970975399017334,
"eval_logps": -79.46780395507812,
"eval_loss": 154.3220672607422,
"eval_objective": 153.7111358642578,
"eval_ranking_idealized": 0.5092975497245789,
"eval_ranking_idealized_expo": 0.5092975497245789,
"eval_ranking_simple": 0.5237603187561035,
"eval_regularize": 153.7111358642578,
"eval_runtime": 260.0103,
"eval_samples_per_second": 22.268,
"eval_steps_per_second": 0.931,
"step": 300
},
{
"dpo_loss": 26.987300872802734,
"epoch": 1.7288615965989607,
"grad_norm": 5666.8865170633535,
"learning_rate": 2.6798785497556184e-07,
"logits": -1.1298245191574097,
"logps": -75.2568588256836,
"loss": 25.3362,
"objective": 26.987300872802734,
"ranking_idealized": 0.5041666626930237,
"ranking_idealized_expo": 0.5041666626930237,
"ranking_simple": 0.6000000238418579,
"regularize": 26.987300872802734,
"step": 305
},
{
"dpo_loss": 27.362499237060547,
"epoch": 1.7572035899858292,
"grad_norm": 6726.005866778725,
"learning_rate": 2.1479047161491351e-07,
"logits": -1.1657706499099731,
"logps": -73.20048522949219,
"loss": 34.4912,
"objective": 27.362499237060547,
"ranking_idealized": 0.5208333134651184,
"ranking_idealized_expo": 0.5208333134651184,
"ranking_simple": 0.6291666626930237,
"regularize": 27.362499237060547,
"step": 310
},
{
"dpo_loss": 26.551549911499023,
"epoch": 1.7855455833726972,
"grad_norm": 4920.254333292925,
"learning_rate": 1.6723858532249783e-07,
"logits": -1.1162974834442139,
"logps": -73.76349639892578,
"loss": 28.333,
"objective": 26.551549911499023,
"ranking_idealized": 0.5083333253860474,
"ranking_idealized_expo": 0.5083333253860474,
"ranking_simple": 0.625,
"regularize": 26.551549911499023,
"step": 315
},
{
"dpo_loss": 27.110830307006836,
"epoch": 1.8138875767595655,
"grad_norm": 5055.149103387853,
"learning_rate": 1.2544967068054332e-07,
"logits": -1.1586967706680298,
"logps": -75.58198547363281,
"loss": 29.2707,
"objective": 27.110830307006836,
"ranking_idealized": 0.5041666626930237,
"ranking_idealized_expo": 0.5041666626930237,
"ranking_simple": 0.6208333373069763,
"regularize": 27.110830307006836,
"step": 320
},
{
"dpo_loss": 21.96232032775879,
"epoch": 1.8422295701464337,
"grad_norm": 5590.846808187711,
"learning_rate": 8.952696513496756e-08,
"logits": -1.190901279449463,
"logps": -76.7601318359375,
"loss": 31.3297,
"objective": 21.96232032775879,
"ranking_idealized": 0.5166666507720947,
"ranking_idealized_expo": 0.5166666507720947,
"ranking_simple": 0.6583333611488342,
"regularize": 21.96232032775879,
"step": 325
},
{
"dpo_loss": 15.405153274536133,
"epoch": 1.8705715635333018,
"grad_norm": 4532.986973110623,
"learning_rate": 5.955921395237318e-08,
"logits": -1.0123432874679565,
"logps": -73.34371948242188,
"loss": 25.3935,
"objective": 15.405153274536133,
"ranking_idealized": 0.47083333134651184,
"ranking_idealized_expo": 0.47083333134651184,
"ranking_simple": 0.5833333134651184,
"regularize": 15.405153274536133,
"step": 330
},
{
"dpo_loss": 27.14239501953125,
"epoch": 1.89891355692017,
"grad_norm": 4025.6636432059317,
"learning_rate": 3.5620450979259034e-08,
"logits": -1.1433656215667725,
"logps": -73.34452056884766,
"loss": 28.8551,
"objective": 27.14239501953125,
"ranking_idealized": 0.4958333373069763,
"ranking_idealized_expo": 0.4958333373069763,
"ranking_simple": 0.6333333253860474,
"regularize": 27.14239501953125,
"step": 335
},
{
"dpo_loss": 27.192691802978516,
"epoch": 1.9272555503070383,
"grad_norm": 4979.044810332174,
"learning_rate": 1.7769815745066476e-08,
"logits": -1.1102468967437744,
"logps": -75.71009063720703,
"loss": 28.1441,
"objective": 27.192691802978516,
"ranking_idealized": 0.5416666865348816,
"ranking_idealized_expo": 0.5416666865348816,
"ranking_simple": 0.6625000238418579,
"regularize": 27.192691802978516,
"step": 340
},
{
"dpo_loss": 20.978731155395508,
"epoch": 1.9555975436939064,
"grad_norm": 4629.432890304778,
"learning_rate": 6.051407360895822e-09,
"logits": -1.1233471632003784,
"logps": -74.21357727050781,
"loss": 26.2298,
"objective": 20.978731155395508,
"ranking_idealized": 0.5333333611488342,
"ranking_idealized_expo": 0.5333333611488342,
"ranking_simple": 0.6083333492279053,
"regularize": 20.978731155395508,
"step": 345
},
{
"dpo_loss": 24.83169174194336,
"epoch": 1.9839395370807746,
"grad_norm": 5527.549670581212,
"learning_rate": 4.941755748361088e-10,
"logits": -1.0784180164337158,
"logps": -74.90613555908203,
"loss": 25.1496,
"objective": 24.83169174194336,
"ranking_idealized": 0.48750001192092896,
"ranking_idealized_expo": 0.48750001192092896,
"ranking_simple": 0.6041666865348816,
"regularize": 24.83169174194336,
"step": 350
},
{
"epoch": 1.9839395370807746,
"eval_dpo_loss": 152.31163024902344,
"eval_logits": -1.1890522241592407,
"eval_logps": -79.3233642578125,
"eval_loss": 153.95770263671875,
"eval_objective": 152.31163024902344,
"eval_ranking_idealized": 0.5092975497245789,
"eval_ranking_idealized_expo": 0.5092975497245789,
"eval_ranking_simple": 0.5227272510528564,
"eval_regularize": 152.31163024902344,
"eval_runtime": 259.9659,
"eval_samples_per_second": 22.272,
"eval_steps_per_second": 0.931,
"step": 350
},
{
"epoch": 1.995276334435522,
"step": 352,
"total_flos": 0.0,
"train_loss": 58.09165079349821,
"train_runtime": 14332.9867,
"train_samples_per_second": 7.089,
"train_steps_per_second": 0.025
}
],
"logging_steps": 5,
"max_steps": 352,
"num_input_tokens_seen": 0,
"num_train_epochs": 2,
"save_steps": 500,
"stateful_callbacks": {
"TrainerControl": {
"args": {
"should_epoch_stop": false,
"should_evaluate": false,
"should_log": false,
"should_save": false,
"should_training_stop": false
},
"attributes": {}
}
},
"total_flos": 0.0,
"train_batch_size": 4,
"trial_name": null,
"trial_params": null
}