gemma-7b-orpo / trainer_state.json
silviasapora's picture
Model save
e8658cf verified
raw
history blame
168 kB
{
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 1.0,
"eval_steps": 500,
"global_step": 1259,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.003971405877680699,
"grad_norm": 1986.144287109375,
"learning_rate": 2.5000000000000004e-07,
"log_odds_chosen": -0.8231229782104492,
"log_odds_ratio": -1.391985297203064,
"logits/chosen": 102.16714477539062,
"logits/rejected": -12.402770042419434,
"logps/chosen": -16.669206619262695,
"logps/rejected": -15.846084594726562,
"loss": 14.8236,
"nll_loss": 15.787309646606445,
"rewards/accuracies": 0.20000000298023224,
"rewards/chosen": -0.8334604501724243,
"rewards/margins": -0.041156161576509476,
"rewards/rejected": -0.7923042178153992,
"step": 5
},
{
"epoch": 0.007942811755361398,
"grad_norm": 1124.8104248046875,
"learning_rate": 5.000000000000001e-07,
"log_odds_chosen": -1.4938147068023682,
"log_odds_ratio": -2.2117180824279785,
"logits/chosen": 69.36089324951172,
"logits/rejected": 133.50851440429688,
"logps/chosen": -14.369367599487305,
"logps/rejected": -12.875558853149414,
"loss": 12.448,
"nll_loss": 12.069561958312988,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.7184683680534363,
"rewards/margins": -0.07469038665294647,
"rewards/rejected": -0.6437779664993286,
"step": 10
},
{
"epoch": 0.011914217633042097,
"grad_norm": 571.815673828125,
"learning_rate": 7.5e-07,
"log_odds_chosen": 0.07529473304748535,
"log_odds_ratio": -0.9602964520454407,
"logits/chosen": 194.84005737304688,
"logits/rejected": 170.63455200195312,
"logps/chosen": -8.528478622436523,
"logps/rejected": -8.603917121887207,
"loss": 8.6647,
"nll_loss": 8.767313003540039,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.42642393708229065,
"rewards/margins": 0.00377195468172431,
"rewards/rejected": -0.4301958680152893,
"step": 15
},
{
"epoch": 0.015885623510722795,
"grad_norm": 294.7610778808594,
"learning_rate": 1.0000000000000002e-06,
"log_odds_chosen": 1.0600534677505493,
"log_odds_ratio": -0.5760771632194519,
"logits/chosen": 130.64846801757812,
"logits/rejected": 219.7195281982422,
"logps/chosen": -5.179438591003418,
"logps/rejected": -6.236131191253662,
"loss": 5.951,
"nll_loss": 5.380393028259277,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.2589719295501709,
"rewards/margins": 0.05283462256193161,
"rewards/rejected": -0.3118065893650055,
"step": 20
},
{
"epoch": 0.019857029388403495,
"grad_norm": 175.71804809570312,
"learning_rate": 1.25e-06,
"log_odds_chosen": -0.23735050857067108,
"log_odds_ratio": -0.8487253189086914,
"logits/chosen": 127.70460510253906,
"logits/rejected": 241.3983154296875,
"logps/chosen": -3.850130796432495,
"logps/rejected": -3.6309711933135986,
"loss": 4.3668,
"nll_loss": 4.186649322509766,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.19250653684139252,
"rewards/margins": -0.010957981459796429,
"rewards/rejected": -0.18154855072498322,
"step": 25
},
{
"epoch": 0.023828435266084195,
"grad_norm": 124.54273986816406,
"learning_rate": 1.5e-06,
"log_odds_chosen": 0.1647549569606781,
"log_odds_ratio": -1.391825795173645,
"logits/chosen": 265.0519714355469,
"logits/rejected": 185.30813598632812,
"logps/chosen": -3.637852430343628,
"logps/rejected": -3.7852470874786377,
"loss": 3.5575,
"nll_loss": 3.7124857902526855,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.18189263343811035,
"rewards/margins": 0.007369739003479481,
"rewards/rejected": -0.18926236033439636,
"step": 30
},
{
"epoch": 0.02779984114376489,
"grad_norm": 197.60816955566406,
"learning_rate": 1.75e-06,
"log_odds_chosen": 0.08467637002468109,
"log_odds_ratio": -0.7092531323432922,
"logits/chosen": 209.2491455078125,
"logits/rejected": 306.3213806152344,
"logps/chosen": -2.6641006469726562,
"logps/rejected": -2.7374536991119385,
"loss": 3.3245,
"nll_loss": 2.6615092754364014,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.13320502638816833,
"rewards/margins": 0.0036676630843430758,
"rewards/rejected": -0.13687269389629364,
"step": 35
},
{
"epoch": 0.03177124702144559,
"grad_norm": 185.7123260498047,
"learning_rate": 2.0000000000000003e-06,
"log_odds_chosen": 0.11347303539514542,
"log_odds_ratio": -0.7381674647331238,
"logits/chosen": 297.0364990234375,
"logits/rejected": 257.74261474609375,
"logps/chosen": -2.3121635913848877,
"logps/rejected": -2.4389572143554688,
"loss": 2.4557,
"nll_loss": 3.096728801727295,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.11560817062854767,
"rewards/margins": 0.006339688785374165,
"rewards/rejected": -0.12194786220788956,
"step": 40
},
{
"epoch": 0.035742652899126294,
"grad_norm": 110.63021087646484,
"learning_rate": 2.25e-06,
"log_odds_chosen": -0.3320659101009369,
"log_odds_ratio": -0.9123506546020508,
"logits/chosen": 301.091552734375,
"logits/rejected": 277.54205322265625,
"logps/chosen": -1.4611170291900635,
"logps/rejected": -1.2663248777389526,
"loss": 2.169,
"nll_loss": 1.8063217401504517,
"rewards/accuracies": 0.20000000298023224,
"rewards/chosen": -0.07305584847927094,
"rewards/margins": -0.009739604778587818,
"rewards/rejected": -0.0633162409067154,
"step": 45
},
{
"epoch": 0.03971405877680699,
"grad_norm": 178.21774291992188,
"learning_rate": 2.5e-06,
"log_odds_chosen": 0.13006296753883362,
"log_odds_ratio": -0.6751025915145874,
"logits/chosen": 348.32208251953125,
"logits/rejected": 229.12075805664062,
"logps/chosen": -1.464900255203247,
"logps/rejected": -1.5294269323349,
"loss": 2.2912,
"nll_loss": 2.135530710220337,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.07324501872062683,
"rewards/margins": 0.003226341214030981,
"rewards/rejected": -0.07647135108709335,
"step": 50
},
{
"epoch": 0.043685464654487687,
"grad_norm": 245.26560974121094,
"learning_rate": 2.7500000000000004e-06,
"log_odds_chosen": 0.5194543600082397,
"log_odds_ratio": -0.47707730531692505,
"logits/chosen": 271.95819091796875,
"logits/rejected": 315.78558349609375,
"logps/chosen": -1.3941795825958252,
"logps/rejected": -1.8157556056976318,
"loss": 1.9179,
"nll_loss": 1.9026470184326172,
"rewards/accuracies": 1.0,
"rewards/chosen": -0.06970898061990738,
"rewards/margins": 0.021078798919916153,
"rewards/rejected": -0.09078778326511383,
"step": 55
},
{
"epoch": 0.04765687053216839,
"grad_norm": 50.196834564208984,
"learning_rate": 3e-06,
"log_odds_chosen": -0.3250153362751007,
"log_odds_ratio": -1.0242193937301636,
"logits/chosen": 358.44586181640625,
"logits/rejected": 318.2178955078125,
"logps/chosen": -1.6527748107910156,
"logps/rejected": -1.3916703462600708,
"loss": 2.1109,
"nll_loss": 1.7649085521697998,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.08263873308897018,
"rewards/margins": -0.013055220246315002,
"rewards/rejected": -0.06958352029323578,
"step": 60
},
{
"epoch": 0.051628276409849086,
"grad_norm": 108.7151107788086,
"learning_rate": 3.2500000000000002e-06,
"log_odds_chosen": 0.29782918095588684,
"log_odds_ratio": -0.6089301705360413,
"logits/chosen": 285.428466796875,
"logits/rejected": 288.96624755859375,
"logps/chosen": -1.7433933019638062,
"logps/rejected": -2.007410764694214,
"loss": 2.4942,
"nll_loss": 2.303445339202881,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.08716966956853867,
"rewards/margins": 0.013200879096984863,
"rewards/rejected": -0.10037054866552353,
"step": 65
},
{
"epoch": 0.05559968228752978,
"grad_norm": 232.15228271484375,
"learning_rate": 3.5e-06,
"log_odds_chosen": -0.16374030709266663,
"log_odds_ratio": -0.8438912630081177,
"logits/chosen": 316.9410400390625,
"logits/rejected": 329.91485595703125,
"logps/chosen": -1.642735481262207,
"logps/rejected": -1.506341814994812,
"loss": 2.2097,
"nll_loss": 1.9609638452529907,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.08213677257299423,
"rewards/margins": -0.006819679401814938,
"rewards/rejected": -0.07531709969043732,
"step": 70
},
{
"epoch": 0.059571088165210485,
"grad_norm": 78.16633605957031,
"learning_rate": 3.7500000000000005e-06,
"log_odds_chosen": 0.08518339693546295,
"log_odds_ratio": -0.6790117025375366,
"logits/chosen": 318.6166076660156,
"logits/rejected": 272.7696533203125,
"logps/chosen": -1.5736838579177856,
"logps/rejected": -1.6470115184783936,
"loss": 2.0889,
"nll_loss": 2.043640375137329,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.07868418842554092,
"rewards/margins": 0.0036663957871496677,
"rewards/rejected": -0.08235058933496475,
"step": 75
},
{
"epoch": 0.06354249404289118,
"grad_norm": 107.8170166015625,
"learning_rate": 4.000000000000001e-06,
"log_odds_chosen": 0.7262557744979858,
"log_odds_ratio": -0.4617268443107605,
"logits/chosen": 258.40740966796875,
"logits/rejected": 339.81085205078125,
"logps/chosen": -1.2588051557540894,
"logps/rejected": -1.8158533573150635,
"loss": 2.0223,
"nll_loss": 1.9233391284942627,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.06294026225805283,
"rewards/margins": 0.027852404862642288,
"rewards/rejected": -0.09079267084598541,
"step": 80
},
{
"epoch": 0.06751389992057188,
"grad_norm": 94.64545440673828,
"learning_rate": 4.25e-06,
"log_odds_chosen": 0.24036984145641327,
"log_odds_ratio": -0.6297720670700073,
"logits/chosen": 322.71148681640625,
"logits/rejected": 335.28228759765625,
"logps/chosen": -1.28748619556427,
"logps/rejected": -1.4782545566558838,
"loss": 1.7358,
"nll_loss": 1.7002407312393188,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.06437431275844574,
"rewards/margins": 0.009538417682051659,
"rewards/rejected": -0.07391272485256195,
"step": 85
},
{
"epoch": 0.07148530579825259,
"grad_norm": 133.2477264404297,
"learning_rate": 4.5e-06,
"log_odds_chosen": 0.11964414268732071,
"log_odds_ratio": -0.6827563643455505,
"logits/chosen": 325.7275695800781,
"logits/rejected": 273.9080505371094,
"logps/chosen": -1.321221947669983,
"logps/rejected": -1.4350178241729736,
"loss": 2.0792,
"nll_loss": 2.0664451122283936,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.0660611018538475,
"rewards/margins": 0.005689795129001141,
"rewards/rejected": -0.07175089418888092,
"step": 90
},
{
"epoch": 0.07545671167593328,
"grad_norm": 75.57556915283203,
"learning_rate": 4.75e-06,
"log_odds_chosen": 0.1924123615026474,
"log_odds_ratio": -0.7588543891906738,
"logits/chosen": 268.33111572265625,
"logits/rejected": 284.4769287109375,
"logps/chosen": -1.2349271774291992,
"logps/rejected": -1.3220834732055664,
"loss": 1.8031,
"nll_loss": 2.01653790473938,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.06174635887145996,
"rewards/margins": 0.0043578073382377625,
"rewards/rejected": -0.06610416620969772,
"step": 95
},
{
"epoch": 0.07942811755361398,
"grad_norm": 82.01164245605469,
"learning_rate": 5e-06,
"log_odds_chosen": 0.22679157555103302,
"log_odds_ratio": -0.6229124069213867,
"logits/chosen": 347.00042724609375,
"logits/rejected": 280.84796142578125,
"logps/chosen": -1.0023002624511719,
"logps/rejected": -1.172823429107666,
"loss": 1.9771,
"nll_loss": 1.778070092201233,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.050115011632442474,
"rewards/margins": 0.008526156656444073,
"rewards/rejected": -0.05864117294549942,
"step": 100
},
{
"epoch": 0.08339952343129468,
"grad_norm": 144.94651794433594,
"learning_rate": 4.99977039769305e-06,
"log_odds_chosen": -0.6747381091117859,
"log_odds_ratio": -1.3916146755218506,
"logits/chosen": 285.270751953125,
"logits/rejected": 392.5542907714844,
"logps/chosen": -2.4368515014648438,
"logps/rejected": -1.8863086700439453,
"loss": 2.1097,
"nll_loss": 2.505847454071045,
"rewards/accuracies": 0.20000000298023224,
"rewards/chosen": -0.12184257805347443,
"rewards/margins": -0.02752714790403843,
"rewards/rejected": -0.09431543946266174,
"step": 105
},
{
"epoch": 0.08737092930897537,
"grad_norm": 907.4935302734375,
"learning_rate": 4.9990816329459744e-06,
"log_odds_chosen": 0.6644043326377869,
"log_odds_ratio": -0.6343256235122681,
"logits/chosen": 355.09014892578125,
"logits/rejected": 328.604736328125,
"logps/chosen": -3.468595027923584,
"logps/rejected": -4.0634002685546875,
"loss": 2.2136,
"nll_loss": 2.77046275138855,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.17342975735664368,
"rewards/margins": 0.02974027954041958,
"rewards/rejected": -0.2031700611114502,
"step": 110
},
{
"epoch": 0.09134233518665608,
"grad_norm": 55.67234802246094,
"learning_rate": 4.997933832272354e-06,
"log_odds_chosen": 0.5580138564109802,
"log_odds_ratio": -0.507127046585083,
"logits/chosen": 261.1326599121094,
"logits/rejected": 371.37396240234375,
"logps/chosen": -1.3323694467544556,
"logps/rejected": -1.7812392711639404,
"loss": 1.7648,
"nll_loss": 1.5408798456192017,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.06661847233772278,
"rewards/margins": 0.02244349755346775,
"rewards/rejected": -0.08906197547912598,
"step": 115
},
{
"epoch": 0.09531374106433678,
"grad_norm": 155.13243103027344,
"learning_rate": 4.996327206502335e-06,
"log_odds_chosen": 0.4205778241157532,
"log_odds_ratio": -0.6450524926185608,
"logits/chosen": 249.2978973388672,
"logits/rejected": 313.87274169921875,
"logps/chosen": -1.5952913761138916,
"logps/rejected": -1.9669711589813232,
"loss": 2.1811,
"nll_loss": 2.7610068321228027,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.07976456731557846,
"rewards/margins": 0.018583994358778,
"rewards/rejected": -0.09834857285022736,
"step": 120
},
{
"epoch": 0.09928514694201747,
"grad_norm": 56.40447235107422,
"learning_rate": 4.994262050743902e-06,
"log_odds_chosen": -0.344782292842865,
"log_odds_ratio": -0.9613167643547058,
"logits/chosen": 297.97796630859375,
"logits/rejected": 316.97796630859375,
"logps/chosen": -1.9141199588775635,
"logps/rejected": -1.645132064819336,
"loss": 2.097,
"nll_loss": 2.657578945159912,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.09570600092411041,
"rewards/margins": -0.013449391350150108,
"rewards/rejected": -0.08225660771131516,
"step": 125
},
{
"epoch": 0.10325655281969817,
"grad_norm": 63.89404296875,
"learning_rate": 4.991738744328679e-06,
"log_odds_chosen": -0.5421128869056702,
"log_odds_ratio": -1.0545024871826172,
"logits/chosen": 358.645751953125,
"logits/rejected": 298.12786865234375,
"logps/chosen": -1.4808099269866943,
"logps/rejected": -1.1260448694229126,
"loss": 2.0496,
"nll_loss": 2.101855993270874,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.0740404948592186,
"rewards/margins": -0.017738252878189087,
"rewards/rejected": -0.05630224198102951,
"step": 130
},
{
"epoch": 0.10722795869737888,
"grad_norm": 94.53604125976562,
"learning_rate": 4.988757750742243e-06,
"log_odds_chosen": -0.7190758585929871,
"log_odds_ratio": -1.2133655548095703,
"logits/chosen": 339.0290832519531,
"logits/rejected": 269.4937438964844,
"logps/chosen": -1.472083568572998,
"logps/rejected": -1.066962718963623,
"loss": 1.8894,
"nll_loss": 2.1049113273620605,
"rewards/accuracies": 0.20000000298023224,
"rewards/chosen": -0.07360417395830154,
"rewards/margins": -0.020256036892533302,
"rewards/rejected": -0.05334814265370369,
"step": 135
},
{
"epoch": 0.11119936457505956,
"grad_norm": 71.55030059814453,
"learning_rate": 4.985319617538998e-06,
"log_odds_chosen": 0.4117642343044281,
"log_odds_ratio": -0.6588890552520752,
"logits/chosen": 288.08966064453125,
"logits/rejected": 382.9283142089844,
"logps/chosen": -1.4075809717178345,
"logps/rejected": -1.670000433921814,
"loss": 2.1323,
"nll_loss": 1.7244634628295898,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.07037904858589172,
"rewards/margins": 0.013120980933308601,
"rewards/rejected": -0.08350002765655518,
"step": 140
},
{
"epoch": 0.11517077045274027,
"grad_norm": 34.86751174926758,
"learning_rate": 4.981424976241598e-06,
"log_odds_chosen": 0.4058244228363037,
"log_odds_ratio": -0.6222633123397827,
"logits/chosen": 308.98876953125,
"logits/rejected": 273.16015625,
"logps/chosen": -1.11463463306427,
"logps/rejected": -1.428716778755188,
"loss": 1.7015,
"nll_loss": 1.501156210899353,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.05573173239827156,
"rewards/margins": 0.01570410653948784,
"rewards/rejected": -0.0714358389377594,
"step": 145
},
{
"epoch": 0.11914217633042097,
"grad_norm": 52.555850982666016,
"learning_rate": 4.977074542224941e-06,
"log_odds_chosen": -0.05959262698888779,
"log_odds_ratio": -0.7522531151771545,
"logits/chosen": 290.1683349609375,
"logits/rejected": 313.0419616699219,
"logps/chosen": -1.5048227310180664,
"logps/rejected": -1.4883638620376587,
"loss": 1.814,
"nll_loss": 1.6905310153961182,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.07524113357067108,
"rewards/margins": -0.0008229411905631423,
"rewards/rejected": -0.07441819459199905,
"step": 150
},
{
"epoch": 0.12311358220810167,
"grad_norm": 108.9931869506836,
"learning_rate": 4.972269114584779e-06,
"log_odds_chosen": -0.10580176115036011,
"log_odds_ratio": -0.794634997844696,
"logits/chosen": 386.6126403808594,
"logits/rejected": 284.34765625,
"logps/chosen": -1.1588705778121948,
"logps/rejected": -1.1435911655426025,
"loss": 1.9924,
"nll_loss": 2.0270378589630127,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.05794353038072586,
"rewards/margins": -0.0007639724644832313,
"rewards/rejected": -0.05717955902218819,
"step": 155
},
{
"epoch": 0.12708498808578236,
"grad_norm": 61.099525451660156,
"learning_rate": 4.9670095759909275e-06,
"log_odds_chosen": 0.3500244915485382,
"log_odds_ratio": -0.577224612236023,
"logits/chosen": 328.20562744140625,
"logits/rejected": 250.172607421875,
"logps/chosen": -1.0037747621536255,
"logps/rejected": -1.1577932834625244,
"loss": 1.8237,
"nll_loss": 1.7924268245697021,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.050188738852739334,
"rewards/margins": 0.007700921501964331,
"rewards/rejected": -0.0578896589577198,
"step": 160
},
{
"epoch": 0.13105639396346305,
"grad_norm": 50.632930755615234,
"learning_rate": 4.961296892525144e-06,
"log_odds_chosen": 0.3148774206638336,
"log_odds_ratio": -0.7541528940200806,
"logits/chosen": 326.50323486328125,
"logits/rejected": 326.8573913574219,
"logps/chosen": -1.0643494129180908,
"logps/rejected": -1.2868483066558838,
"loss": 2.0091,
"nll_loss": 1.6609458923339844,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.05321747064590454,
"rewards/margins": 0.011124944314360619,
"rewards/rejected": -0.06434241682291031,
"step": 165
},
{
"epoch": 0.13502779984114377,
"grad_norm": 44.31376266479492,
"learning_rate": 4.95513211350367e-06,
"log_odds_chosen": 0.7684804797172546,
"log_odds_ratio": -0.424949586391449,
"logits/chosen": 315.80316162109375,
"logits/rejected": 317.47760009765625,
"logps/chosen": -1.0722829103469849,
"logps/rejected": -1.50656259059906,
"loss": 1.5922,
"nll_loss": 1.819898009300232,
"rewards/accuracies": 1.0,
"rewards/chosen": -0.05361414700746536,
"rewards/margins": 0.0217139795422554,
"rewards/rejected": -0.07532812654972076,
"step": 170
},
{
"epoch": 0.13899920571882446,
"grad_norm": 37.67052459716797,
"learning_rate": 4.948516371284493e-06,
"log_odds_chosen": -0.38165563344955444,
"log_odds_ratio": -0.9263311624526978,
"logits/chosen": 399.5979919433594,
"logits/rejected": 327.8373107910156,
"logps/chosen": -1.0562649965286255,
"logps/rejected": -0.7994272112846375,
"loss": 1.7336,
"nll_loss": 1.5048203468322754,
"rewards/accuracies": 0.20000000298023224,
"rewards/chosen": -0.052813250571489334,
"rewards/margins": -0.012841887772083282,
"rewards/rejected": -0.03997135907411575,
"step": 175
},
{
"epoch": 0.14297061159650518,
"grad_norm": 39.37586212158203,
"learning_rate": 4.941450881059354e-06,
"log_odds_chosen": 0.14676916599273682,
"log_odds_ratio": -0.6898115873336792,
"logits/chosen": 288.27630615234375,
"logits/rejected": 291.10162353515625,
"logps/chosen": -1.1020549535751343,
"logps/rejected": -1.1935259103775024,
"loss": 1.5256,
"nll_loss": 1.1978670358657837,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.05510275438427925,
"rewards/margins": 0.00457354262471199,
"rewards/rejected": -0.05967629700899124,
"step": 180
},
{
"epoch": 0.14694201747418587,
"grad_norm": 57.68708419799805,
"learning_rate": 4.933936940630537e-06,
"log_odds_chosen": -0.36938825249671936,
"log_odds_ratio": -0.9125706553459167,
"logits/chosen": 301.7081298828125,
"logits/rejected": 338.01690673828125,
"logps/chosen": -1.5262658596038818,
"logps/rejected": -1.2626394033432007,
"loss": 1.7968,
"nll_loss": 1.8017867803573608,
"rewards/accuracies": 0.20000000298023224,
"rewards/chosen": -0.07631329447031021,
"rewards/margins": -0.013181325979530811,
"rewards/rejected": -0.06313197314739227,
"step": 185
},
{
"epoch": 0.15091342335186655,
"grad_norm": 61.880126953125,
"learning_rate": 4.925975930172489e-06,
"log_odds_chosen": 0.2720580995082855,
"log_odds_ratio": -0.5754180550575256,
"logits/chosen": 240.9505157470703,
"logits/rejected": 291.4943542480469,
"logps/chosen": -1.128051519393921,
"logps/rejected": -1.3102426528930664,
"loss": 1.6575,
"nll_loss": 1.4304395914077759,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.056402575224637985,
"rewards/margins": 0.009109559468925,
"rewards/rejected": -0.06551213562488556,
"step": 190
},
{
"epoch": 0.15488482922954727,
"grad_norm": 128.1157684326172,
"learning_rate": 4.917569311978301e-06,
"log_odds_chosen": -0.16586491465568542,
"log_odds_ratio": -0.8022063374519348,
"logits/chosen": 339.9231872558594,
"logits/rejected": 310.51458740234375,
"logps/chosen": -1.4521989822387695,
"logps/rejected": -1.3116356134414673,
"loss": 1.8756,
"nll_loss": 1.6476871967315674,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.07260995358228683,
"rewards/margins": -0.007028169929981232,
"rewards/rejected": -0.0655817836523056,
"step": 195
},
{
"epoch": 0.15885623510722796,
"grad_norm": 360.7728271484375,
"learning_rate": 4.9087186301911196e-06,
"log_odds_chosen": -0.15912006795406342,
"log_odds_ratio": -0.8308361172676086,
"logits/chosen": 348.33721923828125,
"logits/rejected": 337.98748779296875,
"logps/chosen": -1.4099957942962646,
"logps/rejected": -1.3773982524871826,
"loss": 1.7693,
"nll_loss": 2.075568914413452,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.07049979269504547,
"rewards/margins": -0.001629872596822679,
"rewards/rejected": -0.06886991858482361,
"step": 200
},
{
"epoch": 0.16282764098490865,
"grad_norm": 55.327293395996094,
"learning_rate": 4.89942551052051e-06,
"log_odds_chosen": 0.21036644279956818,
"log_odds_ratio": -0.6468175053596497,
"logits/chosen": 343.9696960449219,
"logits/rejected": 264.9131164550781,
"logps/chosen": -1.5017131567001343,
"logps/rejected": -1.672141432762146,
"loss": 1.8711,
"nll_loss": 1.952646255493164,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.07508565485477448,
"rewards/margins": 0.008521410636603832,
"rewards/rejected": -0.08360707014799118,
"step": 205
},
{
"epoch": 0.16679904686258937,
"grad_norm": 44.57875442504883,
"learning_rate": 4.889691659943842e-06,
"log_odds_chosen": 0.6125321388244629,
"log_odds_ratio": -0.5584419965744019,
"logits/chosen": 301.97369384765625,
"logits/rejected": 392.7519836425781,
"logps/chosen": -1.1298048496246338,
"logps/rejected": -1.5633180141448975,
"loss": 1.9275,
"nll_loss": 1.591841220855713,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.056490249931812286,
"rewards/margins": 0.021675655618309975,
"rewards/rejected": -0.07816590368747711,
"step": 210
},
{
"epoch": 0.17077045274027006,
"grad_norm": 40.897430419921875,
"learning_rate": 4.879518866392757e-06,
"log_odds_chosen": 0.31865325570106506,
"log_odds_ratio": -0.5695816278457642,
"logits/chosen": 289.89654541015625,
"logits/rejected": 337.4901428222656,
"logps/chosen": -1.0857855081558228,
"logps/rejected": -1.3082635402679443,
"loss": 1.8816,
"nll_loss": 1.8894774913787842,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.05428927391767502,
"rewards/margins": 0.011123904958367348,
"rewards/rejected": -0.06541318446397781,
"step": 215
},
{
"epoch": 0.17474185861795075,
"grad_norm": 57.284912109375,
"learning_rate": 4.868908998424749e-06,
"log_odds_chosen": 1.3685444593429565,
"log_odds_ratio": -0.32077115774154663,
"logits/chosen": 461.255859375,
"logits/rejected": 236.81936645507812,
"logps/chosen": -1.125225305557251,
"logps/rejected": -2.0914998054504395,
"loss": 1.8968,
"nll_loss": 2.5089685916900635,
"rewards/accuracies": 1.0,
"rewards/chosen": -0.05626126378774643,
"rewards/margins": 0.04831372946500778,
"rewards/rejected": -0.10457499325275421,
"step": 220
},
{
"epoch": 0.17871326449563146,
"grad_norm": 48.706504821777344,
"learning_rate": 4.8578640048799435e-06,
"log_odds_chosen": 0.48141545057296753,
"log_odds_ratio": -0.5349053740501404,
"logits/chosen": 331.5611877441406,
"logits/rejected": 310.23876953125,
"logps/chosen": -1.1479623317718506,
"logps/rejected": -1.398716688156128,
"loss": 1.6831,
"nll_loss": 2.0783326625823975,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.05739812180399895,
"rewards/margins": 0.012537715956568718,
"rewards/rejected": -0.06993584334850311,
"step": 225
},
{
"epoch": 0.18268467037331215,
"grad_norm": 40.74349594116211,
"learning_rate": 4.846385914523143e-06,
"log_odds_chosen": -0.004401213023811579,
"log_odds_ratio": -0.7065997123718262,
"logits/chosen": 321.564453125,
"logits/rejected": 321.5721130371094,
"logps/chosen": -1.444544792175293,
"logps/rejected": -1.465693473815918,
"loss": 1.6282,
"nll_loss": 1.7196967601776123,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.07222724705934525,
"rewards/margins": 0.0010574304033070803,
"rewards/rejected": -0.07328467816114426,
"step": 230
},
{
"epoch": 0.18665607625099284,
"grad_norm": 124.04603576660156,
"learning_rate": 4.834476835671166e-06,
"log_odds_chosen": -0.40913257002830505,
"log_odds_ratio": -1.0309641361236572,
"logits/chosen": 303.4057312011719,
"logits/rejected": 345.175048828125,
"logps/chosen": -1.1636359691619873,
"logps/rejected": -0.9079948663711548,
"loss": 1.8382,
"nll_loss": 1.716036081314087,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.058181799948215485,
"rewards/margins": -0.012782062403857708,
"rewards/rejected": -0.0453997440636158,
"step": 235
},
{
"epoch": 0.19062748212867356,
"grad_norm": 354.8912048339844,
"learning_rate": 4.822138955805595e-06,
"log_odds_chosen": -0.03981683775782585,
"log_odds_ratio": -0.7560557126998901,
"logits/chosen": 365.21771240234375,
"logits/rejected": 281.8612060546875,
"logps/chosen": -1.5758720636367798,
"logps/rejected": -1.6159578561782837,
"loss": 1.7801,
"nll_loss": 1.9230833053588867,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.07879360020160675,
"rewards/margins": 0.0020043007098138332,
"rewards/rejected": -0.08079790323972702,
"step": 240
},
{
"epoch": 0.19459888800635425,
"grad_norm": 112.47100067138672,
"learning_rate": 4.809374541170974e-06,
"log_odds_chosen": 0.4120238423347473,
"log_odds_ratio": -0.5282896161079407,
"logits/chosen": 280.23223876953125,
"logits/rejected": 381.760986328125,
"logps/chosen": -1.43479323387146,
"logps/rejected": -1.7411903142929077,
"loss": 1.705,
"nll_loss": 1.610185980796814,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.07173965871334076,
"rewards/margins": 0.015319856815040112,
"rewards/rejected": -0.08705951273441315,
"step": 245
},
{
"epoch": 0.19857029388403494,
"grad_norm": 89.64398193359375,
"learning_rate": 4.796185936358543e-06,
"log_odds_chosen": -0.09865443408489227,
"log_odds_ratio": -0.8241540789604187,
"logits/chosen": 375.30975341796875,
"logits/rejected": 277.07080078125,
"logps/chosen": -1.1246061325073242,
"logps/rejected": -1.0761168003082275,
"loss": 1.5139,
"nll_loss": 1.5120210647583008,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.05623030662536621,
"rewards/margins": -0.002424471778795123,
"rewards/rejected": -0.05380583554506302,
"step": 250
},
{
"epoch": 0.20254169976171565,
"grad_norm": 71.554443359375,
"learning_rate": 4.78257556387557e-06,
"log_odds_chosen": 0.17121019959449768,
"log_odds_ratio": -0.8539883494377136,
"logits/chosen": 271.27740478515625,
"logits/rejected": 308.79962158203125,
"logps/chosen": -1.4543489217758179,
"logps/rejected": -1.6729555130004883,
"loss": 1.7579,
"nll_loss": 1.7364709377288818,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.07271744310855865,
"rewards/margins": 0.010930335149168968,
"rewards/rejected": -0.08364777266979218,
"step": 255
},
{
"epoch": 0.20651310563939634,
"grad_norm": 38.314876556396484,
"learning_rate": 4.7685459237003954e-06,
"log_odds_chosen": 0.2899346947669983,
"log_odds_ratio": -0.614177942276001,
"logits/chosen": 298.9414367675781,
"logits/rejected": 384.4459228515625,
"logps/chosen": -1.6951465606689453,
"logps/rejected": -1.8937695026397705,
"loss": 1.8883,
"nll_loss": 1.8541686534881592,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.08475733548402786,
"rewards/margins": 0.009931142441928387,
"rewards/rejected": -0.09468847513198853,
"step": 260
},
{
"epoch": 0.21048451151707703,
"grad_norm": 72.83179473876953,
"learning_rate": 4.754099592823216e-06,
"log_odds_chosen": 0.8651946783065796,
"log_odds_ratio": -0.5348523855209351,
"logits/chosen": 358.8797912597656,
"logits/rejected": 280.5845031738281,
"logps/chosen": -1.027199387550354,
"logps/rejected": -1.6986967325210571,
"loss": 1.512,
"nll_loss": 1.2993109226226807,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.05135997012257576,
"rewards/margins": 0.033574867993593216,
"rewards/rejected": -0.08493484556674957,
"step": 265
},
{
"epoch": 0.21445591739475775,
"grad_norm": 44.400272369384766,
"learning_rate": 4.739239224772751e-06,
"log_odds_chosen": -0.05065007135272026,
"log_odds_ratio": -0.8082035183906555,
"logits/chosen": 298.37335205078125,
"logits/rejected": 344.8133239746094,
"logps/chosen": -1.5913331508636475,
"logps/rejected": -1.4946712255477905,
"loss": 1.6712,
"nll_loss": 1.572104811668396,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.07956665754318237,
"rewards/margins": -0.004833097103983164,
"rewards/rejected": -0.07473356276750565,
"step": 270
},
{
"epoch": 0.21842732327243844,
"grad_norm": 59.13341522216797,
"learning_rate": 4.7239675491288285e-06,
"log_odds_chosen": 0.4572678506374359,
"log_odds_ratio": -0.5575239062309265,
"logits/chosen": 366.19207763671875,
"logits/rejected": 318.4181213378906,
"logps/chosen": -1.4838532209396362,
"logps/rejected": -1.8985198736190796,
"loss": 1.6193,
"nll_loss": 1.7495654821395874,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.07419265806674957,
"rewards/margins": 0.020733339712023735,
"rewards/rejected": -0.09492600709199905,
"step": 275
},
{
"epoch": 0.22239872915011913,
"grad_norm": 153.78656005859375,
"learning_rate": 4.708287371021016e-06,
"log_odds_chosen": 0.4286623001098633,
"log_odds_ratio": -0.6032269597053528,
"logits/chosen": 264.1981506347656,
"logits/rejected": 437.3480529785156,
"logps/chosen": -1.4498531818389893,
"logps/rejected": -1.8169384002685547,
"loss": 1.6785,
"nll_loss": 1.932142972946167,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.07249265164136887,
"rewards/margins": 0.018354270607233047,
"rewards/rejected": -0.09084691852331161,
"step": 280
},
{
"epoch": 0.22637013502779985,
"grad_norm": 55.48463821411133,
"learning_rate": 4.692201570613367e-06,
"log_odds_chosen": -0.38779932260513306,
"log_odds_ratio": -0.9519031643867493,
"logits/chosen": 315.96405029296875,
"logits/rejected": 441.1253967285156,
"logps/chosen": -1.785449743270874,
"logps/rejected": -1.507177472114563,
"loss": 1.9603,
"nll_loss": 1.9527504444122314,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.08927249163389206,
"rewards/margins": -0.013913619332015514,
"rewards/rejected": -0.07535887509584427,
"step": 285
},
{
"epoch": 0.23034154090548054,
"grad_norm": 196.46524047851562,
"learning_rate": 4.675713102575389e-06,
"log_odds_chosen": 0.13416634500026703,
"log_odds_ratio": -0.6926567554473877,
"logits/chosen": 325.7991943359375,
"logits/rejected": 359.30426025390625,
"logps/chosen": -1.287474274635315,
"logps/rejected": -1.4152565002441406,
"loss": 1.8866,
"nll_loss": 1.7071374654769897,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.06437370926141739,
"rewards/margins": 0.006389112211763859,
"rewards/rejected": -0.07076282799243927,
"step": 290
},
{
"epoch": 0.23431294678316125,
"grad_norm": 40.833988189697266,
"learning_rate": 4.658824995539318e-06,
"log_odds_chosen": -0.3848091959953308,
"log_odds_ratio": -0.9408342242240906,
"logits/chosen": 466.2362365722656,
"logits/rejected": 305.64935302734375,
"logps/chosen": -1.5434811115264893,
"logps/rejected": -1.2438277006149292,
"loss": 1.7838,
"nll_loss": 1.8352413177490234,
"rewards/accuracies": 0.20000000298023224,
"rewards/chosen": -0.07717405259609222,
"rewards/margins": -0.01498266588896513,
"rewards/rejected": -0.06219138950109482,
"step": 295
},
{
"epoch": 0.23828435266084194,
"grad_norm": 85.50627899169922,
"learning_rate": 4.641540351543826e-06,
"log_odds_chosen": -0.06552493572235107,
"log_odds_ratio": -0.8229795694351196,
"logits/chosen": 306.17236328125,
"logits/rejected": 333.0214538574219,
"logps/chosen": -1.408279538154602,
"logps/rejected": -1.4727718830108643,
"loss": 1.7687,
"nll_loss": 1.7104421854019165,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.0704139694571495,
"rewards/margins": 0.0032246210612356663,
"rewards/rejected": -0.07363860309123993,
"step": 300
},
{
"epoch": 0.24225575853852263,
"grad_norm": 52.15938186645508,
"learning_rate": 4.6238623454642215e-06,
"log_odds_chosen": -0.7005519270896912,
"log_odds_ratio": -1.180597186088562,
"logits/chosen": 393.58575439453125,
"logits/rejected": 291.7729187011719,
"logps/chosen": -1.5035789012908936,
"logps/rejected": -1.0074363946914673,
"loss": 1.9541,
"nll_loss": 1.8733670711517334,
"rewards/accuracies": 0.20000000298023224,
"rewards/chosen": -0.07517894357442856,
"rewards/margins": -0.02480713650584221,
"rewards/rejected": -0.050371818244457245,
"step": 305
},
{
"epoch": 0.24622716441620335,
"grad_norm": 71.7739028930664,
"learning_rate": 4.605794224429286e-06,
"log_odds_chosen": 0.5956661105155945,
"log_odds_ratio": -0.5032454133033752,
"logits/chosen": 287.00799560546875,
"logits/rejected": 278.5837707519531,
"logps/chosen": -1.3277604579925537,
"logps/rejected": -1.8034179210662842,
"loss": 2.0805,
"nll_loss": 2.333066463470459,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.06638802587985992,
"rewards/margins": 0.023782875388860703,
"rewards/rejected": -0.09017090499401093,
"step": 310
},
{
"epoch": 0.25019857029388404,
"grad_norm": 95.7601089477539,
"learning_rate": 4.587339307224837e-06,
"log_odds_chosen": 0.4633815884590149,
"log_odds_ratio": -0.5474061965942383,
"logits/chosen": 285.03057861328125,
"logits/rejected": 313.64971923828125,
"logps/chosen": -1.064753532409668,
"logps/rejected": -1.4205200672149658,
"loss": 1.674,
"nll_loss": 1.6759742498397827,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.0532376766204834,
"rewards/margins": 0.01778833009302616,
"rewards/rejected": -0.07102601230144501,
"step": 315
},
{
"epoch": 0.2541699761715647,
"grad_norm": 35.280120849609375,
"learning_rate": 4.568500983684127e-06,
"log_odds_chosen": 1.0717216730117798,
"log_odds_ratio": -0.3502858281135559,
"logits/chosen": 316.9742431640625,
"logits/rejected": 361.8272705078125,
"logps/chosen": -1.0445092916488647,
"logps/rejected": -1.8099143505096436,
"loss": 1.5821,
"nll_loss": 1.5527000427246094,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.05222546309232712,
"rewards/margins": 0.0382702499628067,
"rewards/rejected": -0.09049571305513382,
"step": 320
},
{
"epoch": 0.2581413820492454,
"grad_norm": 56.177520751953125,
"learning_rate": 4.54928271406519e-06,
"log_odds_chosen": 0.08797760307788849,
"log_odds_ratio": -0.6692509055137634,
"logits/chosen": 311.7930603027344,
"logits/rejected": 325.4112854003906,
"logps/chosen": -0.8228706121444702,
"logps/rejected": -0.8611005544662476,
"loss": 1.6019,
"nll_loss": 1.4109102487564087,
"rewards/accuracies": 0.20000000298023224,
"rewards/chosen": -0.04114353656768799,
"rewards/margins": 0.0019115001196041703,
"rewards/rejected": -0.043055035173892975,
"step": 325
},
{
"epoch": 0.2621127879269261,
"grad_norm": 49.081207275390625,
"learning_rate": 4.529688028415259e-06,
"log_odds_chosen": 0.5036702156066895,
"log_odds_ratio": -0.5119932293891907,
"logits/chosen": 349.2168273925781,
"logits/rejected": 417.1556091308594,
"logps/chosen": -1.1619529724121094,
"logps/rejected": -1.4640512466430664,
"loss": 1.7083,
"nll_loss": 1.5884101390838623,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.05809764936566353,
"rewards/margins": 0.015104919672012329,
"rewards/rejected": -0.07320256531238556,
"step": 330
},
{
"epoch": 0.26608419380460685,
"grad_norm": 59.9859619140625,
"learning_rate": 4.509720525922356e-06,
"log_odds_chosen": -0.2589249014854431,
"log_odds_ratio": -0.8613080978393555,
"logits/chosen": 319.08282470703125,
"logits/rejected": 374.44512939453125,
"logps/chosen": -1.1872944831848145,
"logps/rejected": -1.0131120681762695,
"loss": 1.6996,
"nll_loss": 1.4569952487945557,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.05936472490429878,
"rewards/margins": -0.008709125220775604,
"rewards/rejected": -0.05065560340881348,
"step": 335
},
{
"epoch": 0.27005559968228754,
"grad_norm": 34.96027755737305,
"learning_rate": 4.489383874254193e-06,
"log_odds_chosen": 0.19578817486763,
"log_odds_ratio": -0.6897228360176086,
"logits/chosen": 308.80645751953125,
"logits/rejected": 363.35272216796875,
"logps/chosen": -1.0954322814941406,
"logps/rejected": -1.26204514503479,
"loss": 1.8048,
"nll_loss": 1.4526374340057373,
"rewards/accuracies": 0.20000000298023224,
"rewards/chosen": -0.05477161332964897,
"rewards/margins": 0.008330638520419598,
"rewards/rejected": -0.06310225278139114,
"step": 340
},
{
"epoch": 0.27402700555996823,
"grad_norm": 88.31672668457031,
"learning_rate": 4.46868180888448e-06,
"log_odds_chosen": 0.5558279752731323,
"log_odds_ratio": -0.6389625072479248,
"logits/chosen": 343.12237548828125,
"logits/rejected": 301.82843017578125,
"logps/chosen": -2.0307748317718506,
"logps/rejected": -2.624824047088623,
"loss": 1.8654,
"nll_loss": 2.4084858894348145,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.1015387549996376,
"rewards/margins": 0.029702449217438698,
"rewards/rejected": -0.13124120235443115,
"step": 345
},
{
"epoch": 0.2779984114376489,
"grad_norm": 36.35618209838867,
"learning_rate": 4.447618132406788e-06,
"log_odds_chosen": 0.05257995054125786,
"log_odds_ratio": -0.7060452699661255,
"logits/chosen": 395.55926513671875,
"logits/rejected": 324.0714416503906,
"logps/chosen": -1.2713950872421265,
"logps/rejected": -1.3762328624725342,
"loss": 1.807,
"nll_loss": 1.739328145980835,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.06356975436210632,
"rewards/margins": 0.005241888575255871,
"rewards/rejected": -0.06881164014339447,
"step": 350
},
{
"epoch": 0.2819698173153296,
"grad_norm": 56.2636604309082,
"learning_rate": 4.4261967138360905e-06,
"log_odds_chosen": -0.501254141330719,
"log_odds_ratio": -1.0499690771102905,
"logits/chosen": 375.1532897949219,
"logits/rejected": 253.8148193359375,
"logps/chosen": -1.6551625728607178,
"logps/rejected": -1.2964580059051514,
"loss": 1.5264,
"nll_loss": 1.8501193523406982,
"rewards/accuracies": 0.20000000298023224,
"rewards/chosen": -0.08275812864303589,
"rewards/margins": -0.017935223877429962,
"rewards/rejected": -0.06482290476560593,
"step": 355
},
{
"epoch": 0.28594122319301035,
"grad_norm": 87.0551528930664,
"learning_rate": 4.404421487898083e-06,
"log_odds_chosen": -0.14473596215248108,
"log_odds_ratio": -0.7785184383392334,
"logits/chosen": 331.86968994140625,
"logits/rejected": 319.239013671875,
"logps/chosen": -1.191341519355774,
"logps/rejected": -1.0848348140716553,
"loss": 1.6036,
"nll_loss": 1.8287181854248047,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.059567082673311234,
"rewards/margins": -0.005325344391167164,
"rewards/rejected": -0.054241735488176346,
"step": 360
},
{
"epoch": 0.28991262907069104,
"grad_norm": 42.729244232177734,
"learning_rate": 4.3822964543064536e-06,
"log_odds_chosen": -0.40415963530540466,
"log_odds_ratio": -0.9470674395561218,
"logits/chosen": 336.14984130859375,
"logits/rejected": 290.5699157714844,
"logps/chosen": -1.3225958347320557,
"logps/rejected": -0.9997372627258301,
"loss": 1.6299,
"nll_loss": 1.5339610576629639,
"rewards/accuracies": 0.20000000298023224,
"rewards/chosen": -0.06612979620695114,
"rewards/margins": -0.01614293083548546,
"rewards/rejected": -0.04998686537146568,
"step": 365
},
{
"epoch": 0.29388403494837173,
"grad_norm": 46.94746780395508,
"learning_rate": 4.359825677028206e-06,
"log_odds_chosen": -0.8854487538337708,
"log_odds_ratio": -1.2963764667510986,
"logits/chosen": 375.13153076171875,
"logits/rejected": 305.2864990234375,
"logps/chosen": -1.3724980354309082,
"logps/rejected": -0.7736121416091919,
"loss": 1.7241,
"nll_loss": 1.8799304962158203,
"rewards/accuracies": 0.20000000298023224,
"rewards/chosen": -0.06862489879131317,
"rewards/margins": -0.029944291338324547,
"rewards/rejected": -0.038680605590343475,
"step": 370
},
{
"epoch": 0.2978554408260524,
"grad_norm": 55.18174743652344,
"learning_rate": 4.337013283537182e-06,
"log_odds_chosen": 0.2510035037994385,
"log_odds_ratio": -0.6310834884643555,
"logits/chosen": 368.63214111328125,
"logits/rejected": 276.0588684082031,
"logps/chosen": -1.1369847059249878,
"logps/rejected": -1.3218457698822021,
"loss": 1.7533,
"nll_loss": 1.9112876653671265,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.05684923380613327,
"rewards/margins": 0.009243053384125233,
"rewards/rejected": -0.06609228998422623,
"step": 375
},
{
"epoch": 0.3018268467037331,
"grad_norm": 41.9847526550293,
"learning_rate": 4.3138634640559185e-06,
"log_odds_chosen": 0.3149817883968353,
"log_odds_ratio": -0.5802162289619446,
"logits/chosen": 320.53692626953125,
"logits/rejected": 413.7608337402344,
"logps/chosen": -1.536766767501831,
"logps/rejected": -1.7898311614990234,
"loss": 1.7084,
"nll_loss": 1.9358304738998413,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.07683834433555603,
"rewards/margins": 0.012653221376240253,
"rewards/rejected": -0.08949156850576401,
"step": 380
},
{
"epoch": 0.3057982525814138,
"grad_norm": 36.58580780029297,
"learning_rate": 4.290380470785984e-06,
"log_odds_chosen": -0.44627267122268677,
"log_odds_ratio": -0.9523780941963196,
"logits/chosen": 384.55609130859375,
"logits/rejected": 288.72259521484375,
"logps/chosen": -1.3820741176605225,
"logps/rejected": -1.060802698135376,
"loss": 1.7563,
"nll_loss": 1.8211044073104858,
"rewards/accuracies": 0.20000000298023224,
"rewards/chosen": -0.06910370290279388,
"rewards/margins": -0.016063563525676727,
"rewards/rejected": -0.05304013937711716,
"step": 385
},
{
"epoch": 0.30976965845909454,
"grad_norm": 41.187171936035156,
"learning_rate": 4.266568617126919e-06,
"log_odds_chosen": -0.3663569390773773,
"log_odds_ratio": -1.1484845876693726,
"logits/chosen": 353.6982727050781,
"logits/rejected": 327.0970764160156,
"logps/chosen": -1.3744146823883057,
"logps/rejected": -1.1000757217407227,
"loss": 1.7097,
"nll_loss": 1.7982555627822876,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.0687207281589508,
"rewards/margins": -0.013716941699385643,
"rewards/rejected": -0.05500379204750061,
"step": 390
},
{
"epoch": 0.31374106433677523,
"grad_norm": 50.873226165771484,
"learning_rate": 4.2424322768839534e-06,
"log_odds_chosen": 0.5074528455734253,
"log_odds_ratio": -0.5399994254112244,
"logits/chosen": 336.0807800292969,
"logits/rejected": 341.9483947753906,
"logps/chosen": -1.1645238399505615,
"logps/rejected": -1.5138776302337646,
"loss": 1.6867,
"nll_loss": 1.4324095249176025,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.05822619050741196,
"rewards/margins": 0.017467692494392395,
"rewards/rejected": -0.07569388300180435,
"step": 395
},
{
"epoch": 0.3177124702144559,
"grad_norm": 31.38866424560547,
"learning_rate": 4.217975883464604e-06,
"log_odds_chosen": 0.4223383367061615,
"log_odds_ratio": -0.557848334312439,
"logits/chosen": 341.1311950683594,
"logits/rejected": 330.12530517578125,
"logps/chosen": -0.972659707069397,
"logps/rejected": -1.1718895435333252,
"loss": 1.5047,
"nll_loss": 1.4151824712753296,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.04863298311829567,
"rewards/margins": 0.009961498901247978,
"rewards/rejected": -0.0585944838821888,
"step": 400
},
{
"epoch": 0.3216838760921366,
"grad_norm": 113.73686218261719,
"learning_rate": 4.1932039290643534e-06,
"log_odds_chosen": -0.6433423161506653,
"log_odds_ratio": -1.1012665033340454,
"logits/chosen": 340.7366638183594,
"logits/rejected": 359.9513854980469,
"logps/chosen": -1.2754091024398804,
"logps/rejected": -0.9146528244018555,
"loss": 1.5981,
"nll_loss": 1.4592812061309814,
"rewards/accuracies": 0.20000000298023224,
"rewards/chosen": -0.06377045810222626,
"rewards/margins": -0.018037814646959305,
"rewards/rejected": -0.04573264718055725,
"step": 405
},
{
"epoch": 0.3256552819698173,
"grad_norm": 27.97637367248535,
"learning_rate": 4.168120963841501e-06,
"log_odds_chosen": 0.515986979007721,
"log_odds_ratio": -0.4868692457675934,
"logits/chosen": 322.2940979003906,
"logits/rejected": 315.605224609375,
"logps/chosen": -1.0631706714630127,
"logps/rejected": -1.4311316013336182,
"loss": 1.5618,
"nll_loss": 1.2970093488693237,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.053158532828092575,
"rewards/margins": 0.018398040905594826,
"rewards/rejected": -0.07155657559633255,
"step": 410
},
{
"epoch": 0.329626687847498,
"grad_norm": 111.9078598022461,
"learning_rate": 4.142731595081394e-06,
"log_odds_chosen": -0.7611304521560669,
"log_odds_ratio": -1.230291485786438,
"logits/chosen": 472.66253662109375,
"logits/rejected": 277.21435546875,
"logps/chosen": -2.0160393714904785,
"logps/rejected": -1.417152762413025,
"loss": 1.6858,
"nll_loss": 2.240736484527588,
"rewards/accuracies": 0.20000000298023224,
"rewards/chosen": -0.1008019670844078,
"rewards/margins": -0.029944339767098427,
"rewards/rejected": -0.07085762917995453,
"step": 415
},
{
"epoch": 0.33359809372517873,
"grad_norm": 55.40426254272461,
"learning_rate": 4.117040486350141e-06,
"log_odds_chosen": -0.27019426226615906,
"log_odds_ratio": -0.9864810705184937,
"logits/chosen": 320.2743225097656,
"logits/rejected": 390.2867736816406,
"logps/chosen": -1.4704816341400146,
"logps/rejected": -1.3386101722717285,
"loss": 2.1069,
"nll_loss": 1.8192126750946045,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.07352408021688461,
"rewards/margins": -0.00659357151016593,
"rewards/rejected": -0.06693051010370255,
"step": 420
},
{
"epoch": 0.3375694996028594,
"grad_norm": 80.17561340332031,
"learning_rate": 4.0910523566380115e-06,
"log_odds_chosen": 0.49502259492874146,
"log_odds_ratio": -0.5022796392440796,
"logits/chosen": 338.86505126953125,
"logits/rejected": 331.8837890625,
"logps/chosen": -1.1411818265914917,
"logps/rejected": -1.491014003753662,
"loss": 1.5704,
"nll_loss": 1.140134572982788,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.057059090584516525,
"rewards/margins": 0.01749161258339882,
"rewards/rejected": -0.07455070316791534,
"step": 425
},
{
"epoch": 0.3415409054805401,
"grad_norm": 40.142581939697266,
"learning_rate": 4.06477197949263e-06,
"log_odds_chosen": 0.31735625863075256,
"log_odds_ratio": -0.6405702233314514,
"logits/chosen": 370.8393859863281,
"logits/rejected": 353.68585205078125,
"logps/chosen": -1.5020195245742798,
"logps/rejected": -1.7753969430923462,
"loss": 1.6606,
"nll_loss": 2.0026352405548096,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.07510097324848175,
"rewards/margins": 0.01366887241601944,
"rewards/rejected": -0.08876985311508179,
"step": 430
},
{
"epoch": 0.3455123113582208,
"grad_norm": 39.17948913574219,
"learning_rate": 4.038204182142174e-06,
"log_odds_chosen": -0.04409918934106827,
"log_odds_ratio": -1.0017715692520142,
"logits/chosen": 365.0375061035156,
"logits/rejected": 287.4947814941406,
"logps/chosen": -1.5504963397979736,
"logps/rejected": -1.3872708082199097,
"loss": 1.754,
"nll_loss": 1.8789138793945312,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.0775248184800148,
"rewards/margins": -0.008161274716258049,
"rewards/rejected": -0.0693635419011116,
"step": 435
},
{
"epoch": 0.3494837172359015,
"grad_norm": 65.36585998535156,
"learning_rate": 4.011353844608695e-06,
"log_odds_chosen": 0.25077468156814575,
"log_odds_ratio": -0.5847761034965515,
"logits/chosen": 398.1842346191406,
"logits/rejected": 302.77557373046875,
"logps/chosen": -1.1319334506988525,
"logps/rejected": -1.2919623851776123,
"loss": 1.5668,
"nll_loss": 1.5274730920791626,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.056596674025058746,
"rewards/margins": 0.008001448586583138,
"rewards/rejected": -0.06459812819957733,
"step": 440
},
{
"epoch": 0.3534551231135822,
"grad_norm": 66.42986297607422,
"learning_rate": 3.9842258988117435e-06,
"log_odds_chosen": 0.39154669642448425,
"log_odds_ratio": -0.5791794657707214,
"logits/chosen": 292.7413635253906,
"logits/rejected": 333.33306884765625,
"logps/chosen": -1.3780268430709839,
"logps/rejected": -1.5879228115081787,
"loss": 1.6821,
"nll_loss": 1.878339409828186,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.06890134513378143,
"rewards/margins": 0.010494804009795189,
"rewards/rejected": -0.07939615100622177,
"step": 445
},
{
"epoch": 0.3574265289912629,
"grad_norm": 41.643070220947266,
"learning_rate": 3.9568253276624704e-06,
"log_odds_chosen": 0.3447812795639038,
"log_odds_ratio": -0.6143133044242859,
"logits/chosen": 308.4149169921875,
"logits/rejected": 373.98712158203125,
"logps/chosen": -1.146469235420227,
"logps/rejected": -1.3937455415725708,
"loss": 1.7364,
"nll_loss": 1.5407047271728516,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.05732346326112747,
"rewards/margins": 0.012363811954855919,
"rewards/rejected": -0.06968727707862854,
"step": 450
},
{
"epoch": 0.3613979348689436,
"grad_norm": 51.8876953125,
"learning_rate": 3.929157164148352e-06,
"log_odds_chosen": 0.25782614946365356,
"log_odds_ratio": -0.6060336828231812,
"logits/chosen": 289.9158020019531,
"logits/rejected": 339.3231201171875,
"logps/chosen": -0.9036432504653931,
"logps/rejected": -1.0764741897583008,
"loss": 1.831,
"nll_loss": 1.5824403762817383,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.045182161033153534,
"rewards/margins": 0.008641545660793781,
"rewards/rejected": -0.05382370948791504,
"step": 455
},
{
"epoch": 0.3653693407466243,
"grad_norm": 54.782958984375,
"learning_rate": 3.901226490408728e-06,
"log_odds_chosen": 0.2922573685646057,
"log_odds_ratio": -0.5908970236778259,
"logits/chosen": 307.7389221191406,
"logits/rejected": 323.6591796875,
"logps/chosen": -1.3589437007904053,
"logps/rejected": -1.5952913761138916,
"loss": 1.5749,
"nll_loss": 1.6173921823501587,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.06794719398021698,
"rewards/margins": 0.011817372404038906,
"rewards/rejected": -0.07976455986499786,
"step": 460
},
{
"epoch": 0.369340746624305,
"grad_norm": 30.29743194580078,
"learning_rate": 3.873038436801298e-06,
"log_odds_chosen": 0.310823529958725,
"log_odds_ratio": -0.662110447883606,
"logits/chosen": 300.14251708984375,
"logits/rejected": 391.355712890625,
"logps/chosen": -1.3028764724731445,
"logps/rejected": -1.4683092832565308,
"loss": 1.714,
"nll_loss": 1.549576997756958,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.06514382362365723,
"rewards/margins": 0.008271644823253155,
"rewards/rejected": -0.07341547310352325,
"step": 465
},
{
"epoch": 0.3733121525019857,
"grad_norm": 59.415611267089844,
"learning_rate": 3.8445981809597715e-06,
"log_odds_chosen": -0.1661299765110016,
"log_odds_ratio": -0.8740431070327759,
"logits/chosen": 272.77264404296875,
"logits/rejected": 408.80975341796875,
"logps/chosen": -1.0485327243804932,
"logps/rejected": -1.0229226350784302,
"loss": 1.4859,
"nll_loss": 1.2975164651870728,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.05242663621902466,
"rewards/margins": -0.0012805074220523238,
"rewards/rejected": -0.05114613100886345,
"step": 470
},
{
"epoch": 0.37728355837966643,
"grad_norm": 38.312618255615234,
"learning_rate": 3.815910946842828e-06,
"log_odds_chosen": -0.014984751120209694,
"log_odds_ratio": -0.8228279948234558,
"logits/chosen": 292.1250305175781,
"logits/rejected": 363.7527770996094,
"logps/chosen": -1.3335845470428467,
"logps/rejected": -1.5042054653167725,
"loss": 1.5739,
"nll_loss": 1.5051485300064087,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.0666792243719101,
"rewards/margins": 0.008531046099960804,
"rewards/rejected": -0.07521027326583862,
"step": 475
},
{
"epoch": 0.3812549642573471,
"grad_norm": 56.91576385498047,
"learning_rate": 3.7869820037745773e-06,
"log_odds_chosen": -0.030622828751802444,
"log_odds_ratio": -0.7795476913452148,
"logits/chosen": 385.81536865234375,
"logits/rejected": 331.4559631347656,
"logps/chosen": -1.2507215738296509,
"logps/rejected": -1.279594898223877,
"loss": 1.6582,
"nll_loss": 1.5629332065582275,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.0625360757112503,
"rewards/margins": 0.0014436636120080948,
"rewards/rejected": -0.06397974491119385,
"step": 480
},
{
"epoch": 0.3852263701350278,
"grad_norm": 147.40060424804688,
"learning_rate": 3.7578166654766695e-06,
"log_odds_chosen": 0.012112426571547985,
"log_odds_ratio": -0.8280073404312134,
"logits/chosen": 324.9056091308594,
"logits/rejected": 301.1451721191406,
"logps/chosen": -1.0767914056777954,
"logps/rejected": -0.936957836151123,
"loss": 1.6,
"nll_loss": 1.675865888595581,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.05383957177400589,
"rewards/margins": -0.006991674192249775,
"rewards/rejected": -0.04684789478778839,
"step": 485
},
{
"epoch": 0.3891977760127085,
"grad_norm": 62.930694580078125,
"learning_rate": 3.7284202890922707e-06,
"log_odds_chosen": -0.3624054789543152,
"log_odds_ratio": -0.9134989976882935,
"logits/chosen": 387.71923828125,
"logits/rejected": 309.9226989746094,
"logps/chosen": -1.203176498413086,
"logps/rejected": -0.9530506134033203,
"loss": 1.6178,
"nll_loss": 1.5490739345550537,
"rewards/accuracies": 0.0,
"rewards/chosen": -0.060158826410770416,
"rewards/margins": -0.012506293132901192,
"rewards/rejected": -0.047652535140514374,
"step": 490
},
{
"epoch": 0.3931691818903892,
"grad_norm": 44.98346710205078,
"learning_rate": 3.698798274202048e-06,
"log_odds_chosen": 1.106994867324829,
"log_odds_ratio": -0.4969615340232849,
"logits/chosen": 416.2843322753906,
"logits/rejected": 322.34722900390625,
"logps/chosen": -1.0825506448745728,
"logps/rejected": -1.9832751750946045,
"loss": 1.4863,
"nll_loss": 1.3532053232192993,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.0541275329887867,
"rewards/margins": 0.045036230236291885,
"rewards/rejected": -0.09916376322507858,
"step": 495
},
{
"epoch": 0.3971405877680699,
"grad_norm": 71.20710754394531,
"learning_rate": 3.668956061832365e-06,
"log_odds_chosen": 0.18296189606189728,
"log_odds_ratio": -0.6548603773117065,
"logits/chosen": 375.1370544433594,
"logits/rejected": 340.265869140625,
"logps/chosen": -1.1238301992416382,
"logps/rejected": -1.242456078529358,
"loss": 1.7543,
"nll_loss": 2.0999672412872314,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.05619151517748833,
"rewards/margins": 0.005931290797889233,
"rewards/rejected": -0.062122803181409836,
"step": 500
},
{
"epoch": 0.4011119936457506,
"grad_norm": 43.03315353393555,
"learning_rate": 3.6388991334558665e-06,
"log_odds_chosen": -0.3676120638847351,
"log_odds_ratio": -0.9084262847900391,
"logits/chosen": 356.4194030761719,
"logits/rejected": 317.63275146484375,
"logps/chosen": -1.1387929916381836,
"logps/rejected": -0.8903687596321106,
"loss": 1.4812,
"nll_loss": 1.3678486347198486,
"rewards/accuracies": 0.0,
"rewards/chosen": -0.05693964287638664,
"rewards/margins": -0.012421206571161747,
"rewards/rejected": -0.04451843723654747,
"step": 505
},
{
"epoch": 0.4050833995234313,
"grad_norm": 27.506656646728516,
"learning_rate": 3.6086330099846274e-06,
"log_odds_chosen": -0.5120627880096436,
"log_odds_ratio": -1.0131856203079224,
"logits/chosen": 349.39093017578125,
"logits/rejected": 328.4683532714844,
"logps/chosen": -1.0289686918258667,
"logps/rejected": -0.724139392375946,
"loss": 1.7849,
"nll_loss": 1.7454910278320312,
"rewards/accuracies": 0.20000000298023224,
"rewards/chosen": -0.05144844204187393,
"rewards/margins": -0.015241468325257301,
"rewards/rejected": -0.03620696812868118,
"step": 510
},
{
"epoch": 0.409054805401112,
"grad_norm": 61.25779724121094,
"learning_rate": 3.578163250756065e-06,
"log_odds_chosen": 0.4315710663795471,
"log_odds_ratio": -0.5859761834144592,
"logits/chosen": 317.2684631347656,
"logits/rejected": 341.7981262207031,
"logps/chosen": -1.1744743585586548,
"logps/rejected": -1.5218207836151123,
"loss": 1.6567,
"nll_loss": 1.4415868520736694,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.05872371047735214,
"rewards/margins": 0.017367318272590637,
"rewards/rejected": -0.07609103620052338,
"step": 515
},
{
"epoch": 0.4130262112787927,
"grad_norm": 74.8428955078125,
"learning_rate": 3.5474954525117887e-06,
"log_odds_chosen": 0.05642819404602051,
"log_odds_ratio": -0.8421560525894165,
"logits/chosen": 321.3084716796875,
"logits/rejected": 286.5706787109375,
"logps/chosen": -1.519960641860962,
"logps/rejected": -1.6510608196258545,
"loss": 1.7797,
"nll_loss": 1.7759168148040771,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.07599803060293198,
"rewards/margins": 0.006555011961609125,
"rewards/rejected": -0.08255304396152496,
"step": 520
},
{
"epoch": 0.4169976171564734,
"grad_norm": 50.956214904785156,
"learning_rate": 3.5166352483695803e-06,
"log_odds_chosen": 0.43155431747436523,
"log_odds_ratio": -0.5425541400909424,
"logits/chosen": 367.09210205078125,
"logits/rejected": 334.6671447753906,
"logps/chosen": -1.14950430393219,
"logps/rejected": -1.4378228187561035,
"loss": 1.7118,
"nll_loss": 1.758131742477417,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.05747520923614502,
"rewards/margins": 0.014415934681892395,
"rewards/rejected": -0.07189114391803741,
"step": 525
},
{
"epoch": 0.42096902303415407,
"grad_norm": 91.34703826904297,
"learning_rate": 3.4855883067886888e-06,
"log_odds_chosen": 0.24526679515838623,
"log_odds_ratio": -0.6680157780647278,
"logits/chosen": 316.42547607421875,
"logits/rejected": 285.8825378417969,
"logps/chosen": -0.9753881692886353,
"logps/rejected": -1.2148548364639282,
"loss": 1.4501,
"nll_loss": 1.2995529174804688,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.04876940697431564,
"rewards/margins": 0.011973332613706589,
"rewards/rejected": -0.060742735862731934,
"step": 530
},
{
"epoch": 0.4249404289118348,
"grad_norm": 110.02918243408203,
"learning_rate": 3.4543603305286432e-06,
"log_odds_chosen": 0.0072197020053863525,
"log_odds_ratio": -0.7949660420417786,
"logits/chosen": 320.6426696777344,
"logits/rejected": 313.28411865234375,
"logps/chosen": -1.0177392959594727,
"logps/rejected": -0.8970460891723633,
"loss": 1.6384,
"nll_loss": 1.6469953060150146,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.05088697001338005,
"rewards/margins": -0.006034668534994125,
"rewards/rejected": -0.044852301478385925,
"step": 535
},
{
"epoch": 0.4289118347895155,
"grad_norm": 39.02376937866211,
"learning_rate": 3.422957055601758e-06,
"log_odds_chosen": -0.32208532094955444,
"log_odds_ratio": -0.9446969032287598,
"logits/chosen": 281.58331298828125,
"logits/rejected": 433.8438415527344,
"logps/chosen": -1.2129267454147339,
"logps/rejected": -0.957917332649231,
"loss": 1.5224,
"nll_loss": 1.474649429321289,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.060646336525678635,
"rewards/margins": -0.012750471010804176,
"rewards/rejected": -0.04789586737751961,
"step": 540
},
{
"epoch": 0.4328832406671962,
"grad_norm": 72.89534759521484,
"learning_rate": 3.3913842502195256e-06,
"log_odds_chosen": 0.10887251049280167,
"log_odds_ratio": -0.6846013069152832,
"logits/chosen": 337.80560302734375,
"logits/rejected": 373.49957275390625,
"logps/chosen": -1.1243635416030884,
"logps/rejected": -1.1805238723754883,
"loss": 1.573,
"nll_loss": 1.2677637338638306,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.05621817708015442,
"rewards/margins": 0.0028080150950700045,
"rewards/rejected": -0.059026192873716354,
"step": 545
},
{
"epoch": 0.4368546465448769,
"grad_norm": 72.7591323852539,
"learning_rate": 3.3596477137331106e-06,
"log_odds_chosen": 0.3823426365852356,
"log_odds_ratio": -0.5422973036766052,
"logits/chosen": 345.8101806640625,
"logits/rejected": 315.0772399902344,
"logps/chosen": -1.0160396099090576,
"logps/rejected": -1.2651389837265015,
"loss": 1.6836,
"nll_loss": 1.2323284149169922,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.05080198124051094,
"rewards/margins": 0.01245497353374958,
"rewards/rejected": -0.06325694918632507,
"step": 550
},
{
"epoch": 0.44082605242255757,
"grad_norm": 90.90532684326172,
"learning_rate": 3.327753275568105e-06,
"log_odds_chosen": 0.33174973726272583,
"log_odds_ratio": -0.5770066380500793,
"logits/chosen": 357.4688415527344,
"logits/rejected": 283.86163330078125,
"logps/chosen": -1.4533777236938477,
"logps/rejected": -1.7138382196426392,
"loss": 1.9812,
"nll_loss": 1.9684550762176514,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.0726688951253891,
"rewards/margins": 0.013023021630942822,
"rewards/rejected": -0.0856919139623642,
"step": 555
},
{
"epoch": 0.44479745830023826,
"grad_norm": 50.66121292114258,
"learning_rate": 3.2957067941537745e-06,
"log_odds_chosen": -0.82317054271698,
"log_odds_ratio": -1.2737700939178467,
"logits/chosen": 260.4927062988281,
"logits/rejected": 432.49237060546875,
"logps/chosen": -1.632800817489624,
"logps/rejected": -1.045261025428772,
"loss": 1.8113,
"nll_loss": 1.9882593154907227,
"rewards/accuracies": 0.20000000298023224,
"rewards/chosen": -0.08164004981517792,
"rewards/margins": -0.02937699481844902,
"rewards/rejected": -0.0522630512714386,
"step": 560
},
{
"epoch": 0.448768864177919,
"grad_norm": 74.89939880371094,
"learning_rate": 3.263514155846969e-06,
"log_odds_chosen": 0.328817218542099,
"log_odds_ratio": -0.5914020538330078,
"logits/chosen": 342.4215393066406,
"logits/rejected": 324.5187683105469,
"logps/chosen": -0.9976029396057129,
"logps/rejected": -1.1394035816192627,
"loss": 1.5494,
"nll_loss": 1.7881031036376953,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.04988015443086624,
"rewards/margins": 0.007090026047080755,
"rewards/rejected": -0.056970179080963135,
"step": 565
},
{
"epoch": 0.4527402700555997,
"grad_norm": 79.3864974975586,
"learning_rate": 3.2311812738509062e-06,
"log_odds_chosen": 0.5359792709350586,
"log_odds_ratio": -0.582006573677063,
"logits/chosen": 312.7106018066406,
"logits/rejected": 337.06732177734375,
"logps/chosen": -1.0472519397735596,
"logps/rejected": -1.3570013046264648,
"loss": 1.7019,
"nll_loss": 1.3205798864364624,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.0523625984787941,
"rewards/margins": 0.015487474389374256,
"rewards/rejected": -0.06785006821155548,
"step": 570
},
{
"epoch": 0.4567116759332804,
"grad_norm": 45.589111328125,
"learning_rate": 3.198714087129024e-06,
"log_odds_chosen": 0.5243362188339233,
"log_odds_ratio": -0.5980243682861328,
"logits/chosen": 344.4964904785156,
"logits/rejected": 442.112548828125,
"logps/chosen": -1.115337610244751,
"logps/rejected": -1.4375159740447998,
"loss": 1.4739,
"nll_loss": 1.505415678024292,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.055766891688108444,
"rewards/margins": 0.016108911484479904,
"rewards/rejected": -0.07187579572200775,
"step": 575
},
{
"epoch": 0.46068308181096107,
"grad_norm": 82.72159576416016,
"learning_rate": 3.1661185593140986e-06,
"log_odds_chosen": -0.06218218803405762,
"log_odds_ratio": -0.7745502591133118,
"logits/chosen": 349.07562255859375,
"logits/rejected": 288.5042724609375,
"logps/chosen": -1.5691912174224854,
"logps/rejected": -1.5143083333969116,
"loss": 1.7202,
"nll_loss": 1.9988605976104736,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.07845956087112427,
"rewards/margins": -0.0027441338170319796,
"rewards/rejected": -0.07571543008089066,
"step": 580
},
{
"epoch": 0.46465448768864176,
"grad_norm": 36.90455627441406,
"learning_rate": 3.133400677612836e-06,
"log_odds_chosen": -0.509355366230011,
"log_odds_ratio": -1.0258004665374756,
"logits/chosen": 336.6009216308594,
"logits/rejected": 255.05517578125,
"logps/chosen": -1.52445387840271,
"logps/rejected": -1.1118301153182983,
"loss": 1.7538,
"nll_loss": 2.2724270820617676,
"rewards/accuracies": 0.20000000298023224,
"rewards/chosen": -0.07622268795967102,
"rewards/margins": -0.020631182938814163,
"rewards/rejected": -0.05559150502085686,
"step": 585
},
{
"epoch": 0.4686258935663225,
"grad_norm": 50.35420608520508,
"learning_rate": 3.100566451706132e-06,
"log_odds_chosen": -0.455788791179657,
"log_odds_ratio": -0.9897588491439819,
"logits/chosen": 316.55255126953125,
"logits/rejected": 320.16485595703125,
"logps/chosen": -1.5558842420578003,
"logps/rejected": -1.1951056718826294,
"loss": 1.8191,
"nll_loss": 1.6975910663604736,
"rewards/accuracies": 0.20000000298023224,
"rewards/chosen": -0.07779420912265778,
"rewards/margins": -0.018038932234048843,
"rewards/rejected": -0.05975528433918953,
"step": 590
},
{
"epoch": 0.4725972994440032,
"grad_norm": 51.966094970703125,
"learning_rate": 3.067621912645195e-06,
"log_odds_chosen": 0.5225323438644409,
"log_odds_ratio": -0.5115421414375305,
"logits/chosen": 318.2131042480469,
"logits/rejected": 299.44970703125,
"logps/chosen": -1.1214492321014404,
"logps/rejected": -1.5328994989395142,
"loss": 1.609,
"nll_loss": 1.4310369491577148,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.05607246607542038,
"rewards/margins": 0.020572511479258537,
"rewards/rejected": -0.07664497196674347,
"step": 595
},
{
"epoch": 0.4765687053216839,
"grad_norm": 43.6855354309082,
"learning_rate": 3.0345731117437636e-06,
"log_odds_chosen": -0.1032852903008461,
"log_odds_ratio": -0.8357051610946655,
"logits/chosen": 371.5491638183594,
"logits/rejected": 289.3468322753906,
"logps/chosen": -1.2020314931869507,
"logps/rejected": -1.056931734085083,
"loss": 1.8538,
"nll_loss": 2.0933589935302734,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.060101576149463654,
"rewards/margins": -0.007254990749061108,
"rewards/rejected": -0.05284658074378967,
"step": 600
},
{
"epoch": 0.4805401111993646,
"grad_norm": 65.06558990478516,
"learning_rate": 3.001426119466581e-06,
"log_odds_chosen": 0.4957882761955261,
"log_odds_ratio": -0.4822470247745514,
"logits/chosen": 320.2109680175781,
"logits/rejected": 344.2149353027344,
"logps/chosen": -1.1690635681152344,
"logps/rejected": -1.5483322143554688,
"loss": 1.541,
"nll_loss": 1.4401990175247192,
"rewards/accuracies": 1.0,
"rewards/chosen": -0.05845317989587784,
"rewards/margins": 0.018963433802127838,
"rewards/rejected": -0.07741661369800568,
"step": 605
},
{
"epoch": 0.48451151707704526,
"grad_norm": 42.35622024536133,
"learning_rate": 2.9681870243143616e-06,
"log_odds_chosen": 0.8287761807441711,
"log_odds_ratio": -0.434671550989151,
"logits/chosen": 405.56640625,
"logits/rejected": 303.19744873046875,
"logps/chosen": -1.0719507932662964,
"logps/rejected": -1.627996802330017,
"loss": 1.7037,
"nll_loss": 1.929690957069397,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.053597547113895416,
"rewards/margins": 0.027802307158708572,
"rewards/rejected": -0.08139985054731369,
"step": 610
},
{
"epoch": 0.48848292295472595,
"grad_norm": 119.18313598632812,
"learning_rate": 2.9348619317054494e-06,
"log_odds_chosen": 0.6101234555244446,
"log_odds_ratio": -0.4588192403316498,
"logits/chosen": 344.44158935546875,
"logits/rejected": 381.56640625,
"logps/chosen": -0.940362274646759,
"logps/rejected": -1.3556063175201416,
"loss": 1.7092,
"nll_loss": 1.7841463088989258,
"rewards/accuracies": 1.0,
"rewards/chosen": -0.04701811820268631,
"rewards/margins": 0.02076220139861107,
"rewards/rejected": -0.06778031587600708,
"step": 615
},
{
"epoch": 0.4924543288324067,
"grad_norm": 41.579444885253906,
"learning_rate": 2.9014569628543577e-06,
"log_odds_chosen": 0.6708475947380066,
"log_odds_ratio": -0.4850080907344818,
"logits/chosen": 271.7330017089844,
"logits/rejected": 390.313232421875,
"logps/chosen": -1.4829562902450562,
"logps/rejected": -1.998425841331482,
"loss": 1.5568,
"nll_loss": 1.7184091806411743,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.07414782792329788,
"rewards/margins": 0.02577347680926323,
"rewards/rejected": -0.09992130100727081,
"step": 620
},
{
"epoch": 0.4964257347100874,
"grad_norm": 42.15742111206055,
"learning_rate": 2.867978253647416e-06,
"log_odds_chosen": 0.20527370274066925,
"log_odds_ratio": -0.798372209072113,
"logits/chosen": 467.48321533203125,
"logits/rejected": 273.2369079589844,
"logps/chosen": -1.0584286451339722,
"logps/rejected": -1.0820119380950928,
"loss": 1.6456,
"nll_loss": 1.5411746501922607,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.05292143300175667,
"rewards/margins": 0.0011791624128818512,
"rewards/rejected": -0.05410059541463852,
"step": 625
},
{
"epoch": 0.5003971405877681,
"grad_norm": 27.90037727355957,
"learning_rate": 2.8344319535157174e-06,
"log_odds_chosen": 0.42426902055740356,
"log_odds_ratio": -0.5486973524093628,
"logits/chosen": 346.8106994628906,
"logits/rejected": 325.4391174316406,
"logps/chosen": -1.075537085533142,
"logps/rejected": -1.3731486797332764,
"loss": 1.4489,
"nll_loss": 1.245110034942627,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.053776852786540985,
"rewards/margins": 0.014880577102303505,
"rewards/rejected": -0.06865743547677994,
"step": 630
},
{
"epoch": 0.5043685464654488,
"grad_norm": 73.80306243896484,
"learning_rate": 2.800824224305584e-06,
"log_odds_chosen": 0.22638097405433655,
"log_odds_ratio": -0.6381832361221313,
"logits/chosen": 359.15911865234375,
"logits/rejected": 273.5746154785156,
"logps/chosen": -1.2116836309432983,
"logps/rejected": -1.3346078395843506,
"loss": 1.6291,
"nll_loss": 1.5670521259307861,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.060584187507629395,
"rewards/margins": 0.006146208383142948,
"rewards/rejected": -0.06673039495944977,
"step": 635
},
{
"epoch": 0.5083399523431295,
"grad_norm": 49.89323043823242,
"learning_rate": 2.76716123914674e-06,
"log_odds_chosen": 0.07012102752923965,
"log_odds_ratio": -0.7102433443069458,
"logits/chosen": 303.50506591796875,
"logits/rejected": 324.7231750488281,
"logps/chosen": -1.0607610940933228,
"logps/rejected": -1.0976377725601196,
"loss": 1.6129,
"nll_loss": 1.4195703268051147,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.05303805321455002,
"rewards/margins": 0.0018438354600220919,
"rewards/rejected": -0.05488189309835434,
"step": 640
},
{
"epoch": 0.5123113582208102,
"grad_norm": 46.98134994506836,
"learning_rate": 2.7334491813184276e-06,
"log_odds_chosen": 0.28301459550857544,
"log_odds_ratio": -0.5637701749801636,
"logits/chosen": 274.5257568359375,
"logits/rejected": 413.9043884277344,
"logps/chosen": -0.8029153943061829,
"logps/rejected": -0.9688779711723328,
"loss": 1.6859,
"nll_loss": 1.5517845153808594,
"rewards/accuracies": 1.0,
"rewards/chosen": -0.04014577344059944,
"rewards/margins": 0.008298131637275219,
"rewards/rejected": -0.04844390228390694,
"step": 645
},
{
"epoch": 0.5162827640984908,
"grad_norm": 26.37677764892578,
"learning_rate": 2.6996942431136466e-06,
"log_odds_chosen": 0.342237263917923,
"log_odds_ratio": -0.613301694393158,
"logits/chosen": 273.4066162109375,
"logits/rejected": 396.9400939941406,
"logps/chosen": -0.9960271716117859,
"logps/rejected": -1.2160321474075317,
"loss": 1.3995,
"nll_loss": 1.4034605026245117,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.049801357090473175,
"rewards/margins": 0.011000247672200203,
"rewards/rejected": -0.06080160662531853,
"step": 650
},
{
"epoch": 0.5202541699761716,
"grad_norm": 46.33094024658203,
"learning_rate": 2.6659026247017418e-06,
"log_odds_chosen": 0.3311120867729187,
"log_odds_ratio": -0.7188401222229004,
"logits/chosen": 384.7994079589844,
"logits/rejected": 300.3538818359375,
"logps/chosen": -1.2231776714324951,
"logps/rejected": -1.4374616146087646,
"loss": 1.5018,
"nll_loss": 1.7496258020401,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.06115889549255371,
"rewards/margins": 0.01071419008076191,
"rewards/rejected": -0.07187308371067047,
"step": 655
},
{
"epoch": 0.5242255758538522,
"grad_norm": 70.27367401123047,
"learning_rate": 2.6320805329895495e-06,
"log_odds_chosen": 0.4950195848941803,
"log_odds_ratio": -0.6416879296302795,
"logits/chosen": 338.72930908203125,
"logits/rejected": 325.9527282714844,
"logps/chosen": -1.1382685899734497,
"logps/rejected": -1.4173352718353271,
"loss": 1.6332,
"nll_loss": 1.4613453149795532,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.056913428008556366,
"rewards/margins": 0.01395334117114544,
"rewards/rejected": -0.07086677849292755,
"step": 660
},
{
"epoch": 0.528196981731533,
"grad_norm": 32.2733039855957,
"learning_rate": 2.5982341804812946e-06,
"log_odds_chosen": 0.3531644642353058,
"log_odds_ratio": -0.5361741781234741,
"logits/chosen": 295.8695068359375,
"logits/rejected": 391.4307556152344,
"logps/chosen": -1.4157756567001343,
"logps/rejected": -1.6907964944839478,
"loss": 1.6299,
"nll_loss": 1.7296804189682007,
"rewards/accuracies": 1.0,
"rewards/chosen": -0.07078878581523895,
"rewards/margins": 0.01375104021281004,
"rewards/rejected": -0.08453982323408127,
"step": 665
},
{
"epoch": 0.5321683876092137,
"grad_norm": 42.08566665649414,
"learning_rate": 2.5643697841374722e-06,
"log_odds_chosen": 0.050149548798799515,
"log_odds_ratio": -0.7359659075737,
"logits/chosen": 320.44049072265625,
"logits/rejected": 269.22320556640625,
"logps/chosen": -1.551636815071106,
"logps/rejected": -1.6407535076141357,
"loss": 1.8105,
"nll_loss": 1.6291691064834595,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.07758183777332306,
"rewards/margins": 0.004455844406038523,
"rewards/rejected": -0.08203768730163574,
"step": 670
},
{
"epoch": 0.5361397934868943,
"grad_norm": 52.635318756103516,
"learning_rate": 2.5304935642329e-06,
"log_odds_chosen": 0.18347088992595673,
"log_odds_ratio": -0.621238112449646,
"logits/chosen": 344.8230895996094,
"logits/rejected": 346.5325927734375,
"logps/chosen": -1.3301855325698853,
"logps/rejected": -1.4845483303070068,
"loss": 1.7328,
"nll_loss": 1.6084213256835938,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.06650927662849426,
"rewards/margins": 0.007718136068433523,
"rewards/rejected": -0.07422741502523422,
"step": 675
},
{
"epoch": 0.5401111993645751,
"grad_norm": 43.74726486206055,
"learning_rate": 2.4966117432141726e-06,
"log_odds_chosen": 0.24497541785240173,
"log_odds_ratio": -0.6605706810951233,
"logits/chosen": 298.27667236328125,
"logits/rejected": 422.50897216796875,
"logps/chosen": -0.9108420610427856,
"logps/rejected": -1.1715190410614014,
"loss": 1.6037,
"nll_loss": 1.2029683589935303,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.04554210603237152,
"rewards/margins": 0.013033849187195301,
"rewards/rejected": -0.05857595056295395,
"step": 680
},
{
"epoch": 0.5440826052422557,
"grad_norm": 64.52481079101562,
"learning_rate": 2.4627305445567048e-06,
"log_odds_chosen": 0.09129991382360458,
"log_odds_ratio": -0.6612669229507446,
"logits/chosen": 390.5311279296875,
"logits/rejected": 266.16888427734375,
"logps/chosen": -1.044427752494812,
"logps/rejected": -1.137475848197937,
"loss": 1.4905,
"nll_loss": 1.4350093603134155,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.0522213876247406,
"rewards/margins": 0.0046524060890078545,
"rewards/rejected": -0.05687378719449043,
"step": 685
},
{
"epoch": 0.5480540111199365,
"grad_norm": 50.33750534057617,
"learning_rate": 2.428856191621596e-06,
"log_odds_chosen": -0.4379616677761078,
"log_odds_ratio": -0.9384675025939941,
"logits/chosen": 295.41632080078125,
"logits/rejected": 335.3627624511719,
"logps/chosen": -1.5291237831115723,
"logps/rejected": -1.2289823293685913,
"loss": 1.5525,
"nll_loss": 1.6753852367401123,
"rewards/accuracies": 0.0,
"rewards/chosen": -0.07645618915557861,
"rewards/margins": -0.015007075853645802,
"rewards/rejected": -0.061449117958545685,
"step": 690
},
{
"epoch": 0.5520254169976172,
"grad_norm": 62.857566833496094,
"learning_rate": 2.3949949065125107e-06,
"log_odds_chosen": -0.1392946094274521,
"log_odds_ratio": -0.8439092636108398,
"logits/chosen": 298.41705322265625,
"logits/rejected": 304.42584228515625,
"logps/chosen": -1.1663509607315063,
"logps/rejected": -1.0636799335479736,
"loss": 1.3371,
"nll_loss": 1.299024224281311,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.05831754207611084,
"rewards/margins": -0.005133545026183128,
"rewards/rejected": -0.05318400263786316,
"step": 695
},
{
"epoch": 0.5559968228752978,
"grad_norm": 36.2435417175293,
"learning_rate": 2.3611529089327893e-06,
"log_odds_chosen": 0.14748263359069824,
"log_odds_ratio": -0.6694773435592651,
"logits/chosen": 278.55145263671875,
"logits/rejected": 415.44378662109375,
"logps/chosen": -1.3171262741088867,
"logps/rejected": -1.4007153511047363,
"loss": 1.4775,
"nll_loss": 1.3410961627960205,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.06585632264614105,
"rewards/margins": 0.004179453942924738,
"rewards/rejected": -0.07003577053546906,
"step": 700
},
{
"epoch": 0.5599682287529786,
"grad_norm": 47.188453674316406,
"learning_rate": 2.327336415043006e-06,
"log_odds_chosen": 0.15575894713401794,
"log_odds_ratio": -0.6959660649299622,
"logits/chosen": 335.40582275390625,
"logits/rejected": 465.32843017578125,
"logps/chosen": -1.123983383178711,
"logps/rejected": -1.2077070474624634,
"loss": 1.5045,
"nll_loss": 1.2318366765975952,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.056199170649051666,
"rewards/margins": 0.004186179488897324,
"rewards/rejected": -0.06038535386323929,
"step": 705
},
{
"epoch": 0.5639396346306592,
"grad_norm": 33.867916107177734,
"learning_rate": 2.2935516363191695e-06,
"log_odds_chosen": -0.3543465733528137,
"log_odds_ratio": -0.9505079388618469,
"logits/chosen": 280.46343994140625,
"logits/rejected": 292.0199279785156,
"logps/chosen": -1.207983374595642,
"logps/rejected": -0.9264053106307983,
"loss": 1.5108,
"nll_loss": 1.4482814073562622,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.060399167239665985,
"rewards/margins": -0.014078897424042225,
"rewards/rejected": -0.046320270746946335,
"step": 710
},
{
"epoch": 0.56791104050834,
"grad_norm": 71.63556671142578,
"learning_rate": 2.259804778411786e-06,
"log_odds_chosen": 0.571071982383728,
"log_odds_ratio": -0.4987887740135193,
"logits/chosen": 361.73345947265625,
"logits/rejected": 394.4234313964844,
"logps/chosen": -0.8434446454048157,
"logps/rejected": -1.139594554901123,
"loss": 1.5099,
"nll_loss": 1.2623536586761475,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.042172230780124664,
"rewards/margins": 0.014807499945163727,
"rewards/rejected": -0.05697972699999809,
"step": 715
},
{
"epoch": 0.5718824463860207,
"grad_norm": 31.155038833618164,
"learning_rate": 2.2261020400059986e-06,
"log_odds_chosen": 0.15770220756530762,
"log_odds_ratio": -0.6557624340057373,
"logits/chosen": 311.5298156738281,
"logits/rejected": 288.30810546875,
"logps/chosen": -1.1949565410614014,
"logps/rejected": -1.3707664012908936,
"loss": 1.4682,
"nll_loss": 1.4724897146224976,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.05974782258272171,
"rewards/margins": 0.008790492080152035,
"rewards/rejected": -0.06853832304477692,
"step": 720
},
{
"epoch": 0.5758538522637013,
"grad_norm": 32.814144134521484,
"learning_rate": 2.1924496116829996e-06,
"log_odds_chosen": 0.4697895050048828,
"log_odds_ratio": -0.6313791871070862,
"logits/chosen": 327.9530334472656,
"logits/rejected": 266.0186462402344,
"logps/chosen": -1.1744751930236816,
"logps/rejected": -1.497837781906128,
"loss": 1.4959,
"nll_loss": 1.3718366622924805,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.05872376635670662,
"rewards/margins": 0.016168128699064255,
"rewards/rejected": -0.07489189505577087,
"step": 725
},
{
"epoch": 0.5798252581413821,
"grad_norm": 48.00249099731445,
"learning_rate": 2.158853674782928e-06,
"log_odds_chosen": -0.15365850925445557,
"log_odds_ratio": -1.1835048198699951,
"logits/chosen": 280.81500244140625,
"logits/rejected": 461.069580078125,
"logps/chosen": -1.4112962484359741,
"logps/rejected": -1.6498725414276123,
"loss": 1.4424,
"nll_loss": 1.5177103281021118,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.07056482136249542,
"rewards/margins": 0.011928820051252842,
"rewards/rejected": -0.08249364048242569,
"step": 730
},
{
"epoch": 0.5837966640190627,
"grad_norm": 38.61474609375,
"learning_rate": 2.1253204002694777e-06,
"log_odds_chosen": 0.62263023853302,
"log_odds_ratio": -0.541793704032898,
"logits/chosen": 304.44573974609375,
"logits/rejected": 331.6595458984375,
"logps/chosen": -1.1065720319747925,
"logps/rejected": -1.4321861267089844,
"loss": 1.5616,
"nll_loss": 1.639901876449585,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.055328596383333206,
"rewards/margins": 0.01628071442246437,
"rewards/rejected": -0.07160931080579758,
"step": 735
},
{
"epoch": 0.5877680698967435,
"grad_norm": 39.873226165771484,
"learning_rate": 2.091855947596401e-06,
"log_odds_chosen": -0.07302029430866241,
"log_odds_ratio": -0.8576405644416809,
"logits/chosen": 356.81842041015625,
"logits/rejected": 376.3813781738281,
"logps/chosen": -1.1168615818023682,
"logps/rejected": -1.0692174434661865,
"loss": 1.4753,
"nll_loss": 1.271024465560913,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.05584307760000229,
"rewards/margins": -0.0023822046350687742,
"rewards/rejected": -0.053460873663425446,
"step": 740
},
{
"epoch": 0.5917394757744241,
"grad_norm": 54.21368408203125,
"learning_rate": 2.058466463576124e-06,
"log_odds_chosen": -0.6579837203025818,
"log_odds_ratio": -1.344585657119751,
"logits/chosen": 289.74786376953125,
"logits/rejected": 320.3457946777344,
"logps/chosen": -1.8980462551116943,
"logps/rejected": -1.3352105617523193,
"loss": 1.8089,
"nll_loss": 1.8561862707138062,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.09490232169628143,
"rewards/margins": -0.02814178541302681,
"rewards/rejected": -0.06676053255796432,
"step": 745
},
{
"epoch": 0.5957108816521048,
"grad_norm": 33.83228302001953,
"learning_rate": 2.0251580812506938e-06,
"log_odds_chosen": -0.6078636050224304,
"log_odds_ratio": -1.1107518672943115,
"logits/chosen": 339.91595458984375,
"logits/rejected": 320.0901794433594,
"logps/chosen": -1.4174226522445679,
"logps/rejected": -1.0588816404342651,
"loss": 1.4017,
"nll_loss": 1.365027904510498,
"rewards/accuracies": 0.20000000298023224,
"rewards/chosen": -0.07087112963199615,
"rewards/margins": -0.017927050590515137,
"rewards/rejected": -0.052944086492061615,
"step": 750
},
{
"epoch": 0.5996822875297856,
"grad_norm": 37.01515579223633,
"learning_rate": 1.9919369187652483e-06,
"log_odds_chosen": -0.07303062826395035,
"log_odds_ratio": -0.7524539232254028,
"logits/chosen": 389.5413513183594,
"logits/rejected": 321.3714294433594,
"logps/chosen": -1.183814287185669,
"logps/rejected": -1.1148512363433838,
"loss": 1.619,
"nll_loss": 1.339825987815857,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.05919071286916733,
"rewards/margins": -0.003448158036917448,
"rewards/rejected": -0.05574256181716919,
"step": 755
},
{
"epoch": 0.6036536934074662,
"grad_norm": 42.6160888671875,
"learning_rate": 1.9588090782442257e-06,
"log_odds_chosen": -0.02819465473294258,
"log_odds_ratio": -0.7264882922172546,
"logits/chosen": 313.9991760253906,
"logits/rejected": 334.552978515625,
"logps/chosen": -1.269689917564392,
"logps/rejected": -1.2629872560501099,
"loss": 1.5967,
"nll_loss": 1.4182217121124268,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.06348450481891632,
"rewards/margins": -0.0003351382911205292,
"rewards/rejected": -0.0631493628025055,
"step": 760
},
{
"epoch": 0.607625099285147,
"grad_norm": 45.021705627441406,
"learning_rate": 1.9257806446705116e-06,
"log_odds_chosen": -0.48236551880836487,
"log_odds_ratio": -1.207824468612671,
"logits/chosen": 385.98260498046875,
"logits/rejected": 320.9479675292969,
"logps/chosen": -1.317742109298706,
"logps/rejected": -0.8753318786621094,
"loss": 1.6117,
"nll_loss": 1.4493136405944824,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.06588710099458694,
"rewards/margins": -0.022120505571365356,
"rewards/rejected": -0.04376659542322159,
"step": 765
},
{
"epoch": 0.6115965051628276,
"grad_norm": 45.63967514038086,
"learning_rate": 1.8928576847677404e-06,
"log_odds_chosen": 0.15900571644306183,
"log_odds_ratio": -0.71577388048172,
"logits/chosen": 349.0858459472656,
"logits/rejected": 374.12872314453125,
"logps/chosen": -0.9837914705276489,
"logps/rejected": -0.9930256009101868,
"loss": 1.8407,
"nll_loss": 1.8363087177276611,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.049189578741788864,
"rewards/margins": 0.0004617050290107727,
"rewards/rejected": -0.04965128004550934,
"step": 770
},
{
"epoch": 0.6155679110405083,
"grad_norm": 49.38869094848633,
"learning_rate": 1.8600462458859492e-06,
"log_odds_chosen": 0.5004615187644958,
"log_odds_ratio": -0.5722527503967285,
"logits/chosen": 338.10382080078125,
"logits/rejected": 324.82861328125,
"logps/chosen": -0.9794312715530396,
"logps/rejected": -1.2764912843704224,
"loss": 1.6962,
"nll_loss": 1.4810049533843994,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.04897156357765198,
"rewards/margins": 0.014853003434836864,
"rewards/rejected": -0.06382457166910172,
"step": 775
},
{
"epoch": 0.6195393169181891,
"grad_norm": 55.28865432739258,
"learning_rate": 1.8273523548907867e-06,
"log_odds_chosen": 0.8983039855957031,
"log_odds_ratio": -0.7501333951950073,
"logits/chosen": 323.56842041015625,
"logits/rejected": 315.9183654785156,
"logps/chosen": -1.0870964527130127,
"logps/rejected": -1.8820436000823975,
"loss": 1.7622,
"nll_loss": 1.5614144802093506,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.054354824125766754,
"rewards/margins": 0.03974735736846924,
"rewards/rejected": -0.09410218894481659,
"step": 780
},
{
"epoch": 0.6235107227958697,
"grad_norm": 38.31996154785156,
"learning_rate": 1.7947820170564897e-06,
"log_odds_chosen": 1.0626842975616455,
"log_odds_ratio": -0.44282132387161255,
"logits/chosen": 351.11236572265625,
"logits/rejected": 300.69366455078125,
"logps/chosen": -1.2781195640563965,
"logps/rejected": -2.0489819049835205,
"loss": 1.4668,
"nll_loss": 1.7934401035308838,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.0639059767127037,
"rewards/margins": 0.03854311257600784,
"rewards/rejected": -0.10244909673929214,
"step": 785
},
{
"epoch": 0.6274821286735505,
"grad_norm": 49.6168212890625,
"learning_rate": 1.7623412149628216e-06,
"log_odds_chosen": -0.2784636616706848,
"log_odds_ratio": -0.9712227582931519,
"logits/chosen": 294.8133239746094,
"logits/rejected": 370.743896484375,
"logps/chosen": -1.2229773998260498,
"logps/rejected": -1.0270025730133057,
"loss": 1.6546,
"nll_loss": 1.4282915592193604,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.06114886328577995,
"rewards/margins": -0.009798737242817879,
"rewards/rejected": -0.051350127905607224,
"step": 790
},
{
"epoch": 0.6314535345512311,
"grad_norm": 28.367610931396484,
"learning_rate": 1.7300359073961834e-06,
"log_odds_chosen": 0.42548590898513794,
"log_odds_ratio": -0.5335083603858948,
"logits/chosen": 337.03436279296875,
"logits/rejected": 383.9117736816406,
"logps/chosen": -1.062105655670166,
"logps/rejected": -1.415236473083496,
"loss": 1.3773,
"nll_loss": 1.3322608470916748,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.05310528352856636,
"rewards/margins": 0.017656544223427773,
"rewards/rejected": -0.07076182961463928,
"step": 795
},
{
"epoch": 0.6354249404289118,
"grad_norm": 98.34535217285156,
"learning_rate": 1.6978720282550897e-06,
"log_odds_chosen": 0.07821528613567352,
"log_odds_ratio": -0.665870726108551,
"logits/chosen": 305.55938720703125,
"logits/rejected": 395.2098388671875,
"logps/chosen": -0.9890663027763367,
"logps/rejected": -1.0237197875976562,
"loss": 1.6046,
"nll_loss": 1.4329578876495361,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.049453310668468475,
"rewards/margins": 0.0017326741944998503,
"rewards/rejected": -0.051185984164476395,
"step": 800
},
{
"epoch": 0.6393963463065926,
"grad_norm": 34.472469329833984,
"learning_rate": 1.6658554854602222e-06,
"log_odds_chosen": 0.6897698640823364,
"log_odds_ratio": -0.41707152128219604,
"logits/chosen": 309.0857849121094,
"logits/rejected": 351.38153076171875,
"logps/chosen": -1.0435580015182495,
"logps/rejected": -1.5444471836090088,
"loss": 1.5348,
"nll_loss": 1.5355165004730225,
"rewards/accuracies": 1.0,
"rewards/chosen": -0.052177898585796356,
"rewards/margins": 0.025044452399015427,
"rewards/rejected": -0.07722235471010208,
"step": 805
},
{
"epoch": 0.6433677521842732,
"grad_norm": 48.47563171386719,
"learning_rate": 1.6339921598692476e-06,
"log_odds_chosen": 0.11802919209003448,
"log_odds_ratio": -0.7510842084884644,
"logits/chosen": 293.94256591796875,
"logits/rejected": 311.3507995605469,
"logps/chosen": -1.730385184288025,
"logps/rejected": -1.8177807331085205,
"loss": 1.7041,
"nll_loss": 1.842095136642456,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.086519256234169,
"rewards/margins": 0.0043697720393538475,
"rewards/rejected": -0.09088902175426483,
"step": 810
},
{
"epoch": 0.647339158061954,
"grad_norm": 31.41144371032715,
"learning_rate": 1.6022879041966188e-06,
"log_odds_chosen": 0.3533809185028076,
"log_odds_ratio": -0.6280057430267334,
"logits/chosen": 335.3319396972656,
"logits/rejected": 264.1844787597656,
"logps/chosen": -1.0379221439361572,
"logps/rejected": -1.2540075778961182,
"loss": 1.3573,
"nll_loss": 1.4223954677581787,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.0518961064517498,
"rewards/margins": 0.01080426573753357,
"rewards/rejected": -0.06270037591457367,
"step": 815
},
{
"epoch": 0.6513105639396346,
"grad_norm": 44.874664306640625,
"learning_rate": 1.5707485419385293e-06,
"log_odds_chosen": -0.14202973246574402,
"log_odds_ratio": -0.8004404306411743,
"logits/chosen": 360.9006042480469,
"logits/rejected": 264.46337890625,
"logps/chosen": -1.2262499332427979,
"logps/rejected": -1.134313941001892,
"loss": 1.496,
"nll_loss": 1.3937432765960693,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.06131250411272049,
"rewards/margins": -0.00459679588675499,
"rewards/rejected": -0.0567157045006752,
"step": 820
},
{
"epoch": 0.6552819698173153,
"grad_norm": 36.3102912902832,
"learning_rate": 1.539379866303245e-06,
"log_odds_chosen": 0.1242959052324295,
"log_odds_ratio": -0.6936241984367371,
"logits/chosen": 281.64874267578125,
"logits/rejected": 311.4866943359375,
"logps/chosen": -1.1951546669006348,
"logps/rejected": -1.2739152908325195,
"loss": 1.5673,
"nll_loss": 1.4124051332473755,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.05975773185491562,
"rewards/margins": 0.003938031382858753,
"rewards/rejected": -0.0636957660317421,
"step": 825
},
{
"epoch": 0.659253375694996,
"grad_norm": 58.44258499145508,
"learning_rate": 1.508187639147001e-06,
"log_odds_chosen": 0.12791205942630768,
"log_odds_ratio": -0.6473852396011353,
"logits/chosen": 344.5693664550781,
"logits/rejected": 367.2503662109375,
"logps/chosen": -1.0457253456115723,
"logps/rejected": -1.142913579940796,
"loss": 1.5898,
"nll_loss": 1.3428993225097656,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.05228627473115921,
"rewards/margins": 0.004859411157667637,
"rewards/rejected": -0.05714568495750427,
"step": 830
},
{
"epoch": 0.6632247815726767,
"grad_norm": 33.92708206176758,
"learning_rate": 1.4771775899156487e-06,
"log_odds_chosen": 0.5014594793319702,
"log_odds_ratio": -0.5115988254547119,
"logits/chosen": 318.1674499511719,
"logits/rejected": 471.2735290527344,
"logps/chosen": -1.0235928297042847,
"logps/rejected": -1.36759614944458,
"loss": 1.4523,
"nll_loss": 1.3124425411224365,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.05117964744567871,
"rewards/margins": 0.017200157046318054,
"rewards/rejected": -0.06837980449199677,
"step": 835
},
{
"epoch": 0.6671961874503575,
"grad_norm": 27.84844207763672,
"learning_rate": 1.4463554145922603e-06,
"log_odds_chosen": -0.31632497906684875,
"log_odds_ratio": -0.9493728876113892,
"logits/chosen": 424.86517333984375,
"logits/rejected": 282.5466613769531,
"logps/chosen": -1.1070573329925537,
"logps/rejected": -0.9589263796806335,
"loss": 1.4065,
"nll_loss": 1.1693588495254517,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.055352866649627686,
"rewards/margins": -0.007406541611999273,
"rewards/rejected": -0.047946326434612274,
"step": 840
},
{
"epoch": 0.6711675933280381,
"grad_norm": 51.31157302856445,
"learning_rate": 1.4157267746508834e-06,
"log_odds_chosen": 0.07127873599529266,
"log_odds_ratio": -0.7540755867958069,
"logits/chosen": 357.64971923828125,
"logits/rejected": 300.25811767578125,
"logps/chosen": -1.264819860458374,
"logps/rejected": -1.2909610271453857,
"loss": 1.2753,
"nll_loss": 1.4475698471069336,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.06324099004268646,
"rewards/margins": 0.0013070597779005766,
"rewards/rejected": -0.0645480528473854,
"step": 845
},
{
"epoch": 0.6751389992057188,
"grad_norm": 47.41500473022461,
"learning_rate": 1.385297296016631e-06,
"log_odds_chosen": 0.2415420562028885,
"log_odds_ratio": -0.5941085815429688,
"logits/chosen": 419.50408935546875,
"logits/rejected": 281.9603576660156,
"logps/chosen": -0.820796012878418,
"logps/rejected": -0.9756487011909485,
"loss": 1.5865,
"nll_loss": 1.0082799196243286,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.041039805859327316,
"rewards/margins": 0.007742627058178186,
"rewards/rejected": -0.048782430589199066,
"step": 850
},
{
"epoch": 0.6791104050833995,
"grad_norm": 42.014591217041016,
"learning_rate": 1.3550725680322973e-06,
"log_odds_chosen": 1.7263425588607788,
"log_odds_ratio": -0.2393535077571869,
"logits/chosen": 317.6227111816406,
"logits/rejected": 302.24859619140625,
"logps/chosen": -0.7767224311828613,
"logps/rejected": -2.0884933471679688,
"loss": 1.5027,
"nll_loss": 1.3212028741836548,
"rewards/accuracies": 1.0,
"rewards/chosen": -0.038836125284433365,
"rewards/margins": 0.06558854132890701,
"rewards/rejected": -0.10442467033863068,
"step": 855
},
{
"epoch": 0.6830818109610802,
"grad_norm": 32.96746063232422,
"learning_rate": 1.3250581424317012e-06,
"log_odds_chosen": -0.05438203737139702,
"log_odds_ratio": -0.8108514547348022,
"logits/chosen": 328.84979248046875,
"logits/rejected": 378.38507080078125,
"logps/chosen": -1.1635057926177979,
"logps/rejected": -1.1345611810684204,
"loss": 1.456,
"nll_loss": 1.441853642463684,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.05817528814077377,
"rewards/margins": -0.0014472283655777574,
"rewards/rejected": -0.0567280575633049,
"step": 860
},
{
"epoch": 0.687053216838761,
"grad_norm": 25.131879806518555,
"learning_rate": 1.295259532319927e-06,
"log_odds_chosen": 1.491008996963501,
"log_odds_ratio": -0.33983761072158813,
"logits/chosen": 305.7498474121094,
"logits/rejected": 395.27630615234375,
"logps/chosen": -0.9085075259208679,
"logps/rejected": -1.8494819402694702,
"loss": 1.5229,
"nll_loss": 1.9765069484710693,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.045425377786159515,
"rewards/margins": 0.04704872891306877,
"rewards/rejected": -0.09247410297393799,
"step": 865
},
{
"epoch": 0.6910246227164416,
"grad_norm": 37.967525482177734,
"learning_rate": 1.265682211160678e-06,
"log_odds_chosen": -0.6112550497055054,
"log_odds_ratio": -1.088639497756958,
"logits/chosen": 310.4002685546875,
"logits/rejected": 406.2005615234375,
"logps/chosen": -1.192030906677246,
"logps/rejected": -0.812456488609314,
"loss": 1.6589,
"nll_loss": 1.2748486995697021,
"rewards/accuracies": 0.20000000298023224,
"rewards/chosen": -0.059601545333862305,
"rewards/margins": -0.01897871494293213,
"rewards/rejected": -0.040622830390930176,
"step": 870
},
{
"epoch": 0.6949960285941224,
"grad_norm": 75.36614227294922,
"learning_rate": 1.2363316117708912e-06,
"log_odds_chosen": 0.7237299680709839,
"log_odds_ratio": -0.6293952465057373,
"logits/chosen": 310.7578430175781,
"logits/rejected": 359.1610412597656,
"logps/chosen": -1.019719123840332,
"logps/rejected": -1.6588159799575806,
"loss": 1.5328,
"nll_loss": 1.2692029476165771,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.05098595470190048,
"rewards/margins": 0.031954843550920486,
"rewards/rejected": -0.08294080197811127,
"step": 875
},
{
"epoch": 0.698967434471803,
"grad_norm": 36.7714958190918,
"learning_rate": 1.2072131253228292e-06,
"log_odds_chosen": -0.17302027344703674,
"log_odds_ratio": -0.8415447473526001,
"logits/chosen": 327.3581237792969,
"logits/rejected": 300.59027099609375,
"logps/chosen": -1.4463417530059814,
"logps/rejected": -1.3052947521209717,
"loss": 1.5306,
"nll_loss": 1.9659268856048584,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.07231709361076355,
"rewards/margins": -0.007052358239889145,
"rewards/rejected": -0.0652647316455841,
"step": 880
},
{
"epoch": 0.7029388403494837,
"grad_norm": 39.60774612426758,
"learning_rate": 1.1783321003538262e-06,
"log_odds_chosen": 0.14880752563476562,
"log_odds_ratio": -0.6801349520683289,
"logits/chosen": 326.66217041015625,
"logits/rejected": 288.07977294921875,
"logps/chosen": -1.3033047914505005,
"logps/rejected": -1.410351037979126,
"loss": 1.5044,
"nll_loss": 1.411853551864624,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.06516523659229279,
"rewards/margins": 0.005352319683879614,
"rewards/rejected": -0.07051756232976913,
"step": 885
},
{
"epoch": 0.7069102462271644,
"grad_norm": 49.855010986328125,
"learning_rate": 1.1496938417838466e-06,
"log_odds_chosen": -0.6302076578140259,
"log_odds_ratio": -1.108147382736206,
"logits/chosen": 303.8426208496094,
"logits/rejected": 299.96282958984375,
"logps/chosen": -1.1776927709579468,
"logps/rejected": -0.8100347518920898,
"loss": 1.6152,
"nll_loss": 1.3265931606292725,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.0588846430182457,
"rewards/margins": -0.018382901325821877,
"rewards/rejected": -0.04050173982977867,
"step": 890
},
{
"epoch": 0.7108816521048451,
"grad_norm": 65.90828704833984,
"learning_rate": 1.1213036099410799e-06,
"log_odds_chosen": 0.3805133104324341,
"log_odds_ratio": -0.5482473373413086,
"logits/chosen": 334.39703369140625,
"logits/rejected": 280.8984680175781,
"logps/chosen": -1.1065565347671509,
"logps/rejected": -1.3260886669158936,
"loss": 1.5504,
"nll_loss": 1.5361745357513428,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.055327825248241425,
"rewards/margins": 0.010976609773933887,
"rewards/rejected": -0.06630443036556244,
"step": 895
},
{
"epoch": 0.7148530579825259,
"grad_norm": 44.0485725402832,
"learning_rate": 1.0931666195957053e-06,
"log_odds_chosen": 0.4383140504360199,
"log_odds_ratio": -0.5253010392189026,
"logits/chosen": 337.86090087890625,
"logits/rejected": 394.5022888183594,
"logps/chosen": -0.7901648283004761,
"logps/rejected": -1.034003496170044,
"loss": 1.4815,
"nll_loss": 1.1732326745986938,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.03950824216008186,
"rewards/margins": 0.012191934511065483,
"rewards/rejected": -0.0517001748085022,
"step": 900
},
{
"epoch": 0.7188244638602065,
"grad_norm": 40.06319046020508,
"learning_rate": 1.0652880390020398e-06,
"log_odds_chosen": 2.0690839290618896,
"log_odds_ratio": -0.2347683608531952,
"logits/chosen": 423.53961181640625,
"logits/rejected": 276.1210021972656,
"logps/chosen": -0.6150510311126709,
"logps/rejected": -1.5576661825180054,
"loss": 1.4353,
"nll_loss": 1.4714704751968384,
"rewards/accuracies": 1.0,
"rewards/chosen": -0.030752548947930336,
"rewards/margins": 0.0471307598054409,
"rewards/rejected": -0.07788331806659698,
"step": 905
},
{
"epoch": 0.7227958697378872,
"grad_norm": 29.090539932250977,
"learning_rate": 1.0376729889492178e-06,
"log_odds_chosen": -0.0568159744143486,
"log_odds_ratio": -0.7396942377090454,
"logits/chosen": 287.9375915527344,
"logits/rejected": 300.72698974609375,
"logps/chosen": -1.0032362937927246,
"logps/rejected": -0.9725478887557983,
"loss": 1.5766,
"nll_loss": 1.5186008214950562,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.05016181990504265,
"rewards/margins": -0.0015344202984124422,
"rewards/rejected": -0.04862739518284798,
"step": 910
},
{
"epoch": 0.7267672756155679,
"grad_norm": 43.51217269897461,
"learning_rate": 1.0103265418205984e-06,
"log_odds_chosen": 0.6717264652252197,
"log_odds_ratio": -0.5441080927848816,
"logits/chosen": 299.4734802246094,
"logits/rejected": 346.14617919921875,
"logps/chosen": -0.7290914058685303,
"logps/rejected": -1.1217130422592163,
"loss": 1.5629,
"nll_loss": 1.4440972805023193,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.03645457327365875,
"rewards/margins": 0.01963108405470848,
"rewards/rejected": -0.05608565732836723,
"step": 915
},
{
"epoch": 0.7307386814932486,
"grad_norm": 43.0380859375,
"learning_rate": 9.832537206620594e-07,
"log_odds_chosen": 0.04898405075073242,
"log_odds_ratio": -0.8131389617919922,
"logits/chosen": 331.40570068359375,
"logits/rejected": 315.16229248046875,
"logps/chosen": -0.9992687106132507,
"logps/rejected": -1.0054936408996582,
"loss": 1.5654,
"nll_loss": 1.4301942586898804,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.049963437020778656,
"rewards/margins": 0.00031124576344154775,
"rewards/rejected": -0.05027468129992485,
"step": 920
},
{
"epoch": 0.7347100873709294,
"grad_norm": 37.88023376464844,
"learning_rate": 9.564594982593559e-07,
"log_odds_chosen": 0.1898471564054489,
"log_odds_ratio": -0.6842805743217468,
"logits/chosen": 346.6091613769531,
"logits/rejected": 315.18408203125,
"logps/chosen": -0.9871004819869995,
"logps/rejected": -1.0915769338607788,
"loss": 1.439,
"nll_loss": 1.487571358680725,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.049355026334524155,
"rewards/margins": 0.005223819054663181,
"rewards/rejected": -0.05457884818315506,
"step": 925
},
{
"epoch": 0.73868149324861,
"grad_norm": 54.20380783081055,
"learning_rate": 9.299487962247089e-07,
"log_odds_chosen": 0.022501707077026367,
"log_odds_ratio": -0.8337327837944031,
"logits/chosen": 302.9130554199219,
"logits/rejected": 279.71685791015625,
"logps/chosen": -1.2005774974822998,
"logps/rejected": -1.1395219564437866,
"loss": 1.6788,
"nll_loss": 1.584804654121399,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.06002888083457947,
"rewards/margins": -0.0030527785420417786,
"rewards/rejected": -0.05697610229253769,
"step": 930
},
{
"epoch": 0.7426528991262907,
"grad_norm": 47.39323425292969,
"learning_rate": 9.037264840927945e-07,
"log_odds_chosen": -0.10585136711597443,
"log_odds_ratio": -0.8327314257621765,
"logits/chosen": 284.0058898925781,
"logits/rejected": 345.33587646484375,
"logps/chosen": -1.1005109548568726,
"logps/rejected": -1.0587420463562012,
"loss": 1.3425,
"nll_loss": 1.6860910654067993,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.05502554774284363,
"rewards/margins": -0.002088439418002963,
"rewards/rejected": -0.052937109023332596,
"step": 935
},
{
"epoch": 0.7466243050039714,
"grad_norm": 38.21221923828125,
"learning_rate": 8.777973784263016e-07,
"log_odds_chosen": -0.43362635374069214,
"log_odds_ratio": -0.9653336405754089,
"logits/chosen": 285.54193115234375,
"logits/rejected": 329.3501892089844,
"logps/chosen": -1.3442434072494507,
"logps/rejected": -1.126430869102478,
"loss": 1.4812,
"nll_loss": 1.4815846681594849,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.06721217930316925,
"rewards/margins": -0.01089063473045826,
"rewards/rejected": -0.05632154271006584,
"step": 940
},
{
"epoch": 0.7505957108816521,
"grad_norm": 33.12229919433594,
"learning_rate": 8.521662419312124e-07,
"log_odds_chosen": 0.15825173258781433,
"log_odds_ratio": -0.7438204884529114,
"logits/chosen": 337.7537841796875,
"logits/rejected": 271.5417175292969,
"logps/chosen": -1.0434350967407227,
"logps/rejected": -1.243912696838379,
"loss": 1.5446,
"nll_loss": 1.3849962949752808,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.05217175558209419,
"rewards/margins": 0.010023881681263447,
"rewards/rejected": -0.062195636332035065,
"step": 945
},
{
"epoch": 0.7545671167593329,
"grad_norm": 48.55085372924805,
"learning_rate": 8.268377825819821e-07,
"log_odds_chosen": -0.014403104782104492,
"log_odds_ratio": -0.7055137157440186,
"logits/chosen": 371.1424865722656,
"logits/rejected": 286.8011779785156,
"logps/chosen": -1.4028490781784058,
"logps/rejected": -1.39651620388031,
"loss": 1.5356,
"nll_loss": 1.620958924293518,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.07014245539903641,
"rewards/margins": -0.00031664298148825765,
"rewards/rejected": -0.06982581317424774,
"step": 950
},
{
"epoch": 0.7585385226370135,
"grad_norm": 36.20191192626953,
"learning_rate": 8.018166527567672e-07,
"log_odds_chosen": 1.0636457204818726,
"log_odds_ratio": -0.43774813413619995,
"logits/chosen": 259.2046813964844,
"logits/rejected": 375.35198974609375,
"logps/chosen": -0.868695080280304,
"logps/rejected": -1.445894479751587,
"loss": 1.4218,
"nll_loss": 1.3990987539291382,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.0434347540140152,
"rewards/margins": 0.02885996736586094,
"rewards/rejected": -0.07229472696781158,
"step": 955
},
{
"epoch": 0.7625099285146942,
"grad_norm": 29.8554744720459,
"learning_rate": 7.771074483828747e-07,
"log_odds_chosen": -0.15444841980934143,
"log_odds_ratio": -0.8071505427360535,
"logits/chosen": 290.59674072265625,
"logits/rejected": 347.6033630371094,
"logps/chosen": -1.40049147605896,
"logps/rejected": -1.3093515634536743,
"loss": 1.3432,
"nll_loss": 1.450141191482544,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.07002457231283188,
"rewards/margins": -0.0045569948852062225,
"rewards/rejected": -0.06546757370233536,
"step": 960
},
{
"epoch": 0.7664813343923749,
"grad_norm": 39.03826904296875,
"learning_rate": 7.52714708092565e-07,
"log_odds_chosen": 0.04803264141082764,
"log_odds_ratio": -0.7985066175460815,
"logits/chosen": 321.512939453125,
"logits/rejected": 325.80853271484375,
"logps/chosen": -1.1335227489471436,
"logps/rejected": -1.0179483890533447,
"loss": 1.5082,
"nll_loss": 1.2471481561660767,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.05667613074183464,
"rewards/margins": -0.005778718274086714,
"rewards/rejected": -0.050897419452667236,
"step": 965
},
{
"epoch": 0.7704527402700556,
"grad_norm": 43.7759895324707,
"learning_rate": 7.286429123893931e-07,
"log_odds_chosen": 0.35435453057289124,
"log_odds_ratio": -0.6048498153686523,
"logits/chosen": 308.8065490722656,
"logits/rejected": 375.87982177734375,
"logps/chosen": -1.6246349811553955,
"logps/rejected": -1.9498169422149658,
"loss": 1.6459,
"nll_loss": 1.7503254413604736,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.0812317505478859,
"rewards/margins": 0.016259105876088142,
"rewards/rejected": -0.09749085456132889,
"step": 970
},
{
"epoch": 0.7744241461477362,
"grad_norm": 54.47321319580078,
"learning_rate": 7.048964828252188e-07,
"log_odds_chosen": 0.9730485081672668,
"log_odds_ratio": -0.47064799070358276,
"logits/chosen": 364.9896545410156,
"logits/rejected": 361.5113830566406,
"logps/chosen": -0.9137029647827148,
"logps/rejected": -1.395819067955017,
"loss": 1.3761,
"nll_loss": 1.2809137105941772,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.04568514600396156,
"rewards/margins": 0.02410580962896347,
"rewards/rejected": -0.06979095190763474,
"step": 975
},
{
"epoch": 0.778395552025417,
"grad_norm": 46.42416000366211,
"learning_rate": 6.814797811880525e-07,
"log_odds_chosen": 0.18214160203933716,
"log_odds_ratio": -0.714030921459198,
"logits/chosen": 386.58782958984375,
"logits/rejected": 275.394775390625,
"logps/chosen": -1.3727108240127563,
"logps/rejected": -1.474959373474121,
"loss": 1.4134,
"nll_loss": 1.827099084854126,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.06863553822040558,
"rewards/margins": 0.005112423561513424,
"rewards/rejected": -0.07374797016382217,
"step": 980
},
{
"epoch": 0.7823669579030977,
"grad_norm": 53.367584228515625,
"learning_rate": 6.583971087008654e-07,
"log_odds_chosen": 0.1975199282169342,
"log_odds_ratio": -0.6967960596084595,
"logits/chosen": 292.4072265625,
"logits/rejected": 433.8272399902344,
"logps/chosen": -0.9060415029525757,
"logps/rejected": -0.9691941142082214,
"loss": 1.3796,
"nll_loss": 1.407875657081604,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.04530208185315132,
"rewards/margins": 0.0031576238106936216,
"rewards/rejected": -0.04845970869064331,
"step": 985
},
{
"epoch": 0.7863383637807784,
"grad_norm": 50.85255813598633,
"learning_rate": 6.356527052315403e-07,
"log_odds_chosen": -0.2651984691619873,
"log_odds_ratio": -0.8623536825180054,
"logits/chosen": 372.86358642578125,
"logits/rejected": 285.08172607421875,
"logps/chosen": -1.2994807958602905,
"logps/rejected": -1.1402140855789185,
"loss": 1.5812,
"nll_loss": 1.4798548221588135,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.06497403979301453,
"rewards/margins": -0.007963338866829872,
"rewards/rejected": -0.0570107102394104,
"step": 990
},
{
"epoch": 0.7903097696584591,
"grad_norm": 79.17274475097656,
"learning_rate": 6.132507485140843e-07,
"log_odds_chosen": 1.5265482664108276,
"log_odds_ratio": -0.2821381688117981,
"logits/chosen": 388.9686279296875,
"logits/rejected": 307.1571350097656,
"logps/chosen": -0.7675724625587463,
"logps/rejected": -1.6737315654754639,
"loss": 1.4555,
"nll_loss": 1.2786924839019775,
"rewards/accuracies": 1.0,
"rewards/chosen": -0.03837861865758896,
"rewards/margins": 0.045307956635951996,
"rewards/rejected": -0.08368657529354095,
"step": 995
},
{
"epoch": 0.7942811755361397,
"grad_norm": 41.524253845214844,
"learning_rate": 5.911953533812506e-07,
"log_odds_chosen": -0.3479143977165222,
"log_odds_ratio": -0.9029040336608887,
"logits/chosen": 338.65325927734375,
"logits/rejected": 374.2637939453125,
"logps/chosen": -0.9410643577575684,
"logps/rejected": -0.7491464614868164,
"loss": 1.655,
"nll_loss": 1.3791284561157227,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.047053221613168716,
"rewards/margins": -0.009595893323421478,
"rewards/rejected": -0.03745732456445694,
"step": 1000
},
{
"epoch": 0.7982525814138205,
"grad_norm": 47.42415237426758,
"learning_rate": 5.694905710087217e-07,
"log_odds_chosen": 0.20326891541481018,
"log_odds_ratio": -0.6220360398292542,
"logits/chosen": 397.6610107421875,
"logits/rejected": 313.22088623046875,
"logps/chosen": -0.8882730603218079,
"logps/rejected": -1.0516657829284668,
"loss": 1.6773,
"nll_loss": 1.243209719657898,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.04441365599632263,
"rewards/margins": 0.008169631473720074,
"rewards/rejected": -0.05258328840136528,
"step": 1005
},
{
"epoch": 0.8022239872915012,
"grad_norm": 45.351253509521484,
"learning_rate": 5.481403881709815e-07,
"log_odds_chosen": 0.5789100527763367,
"log_odds_ratio": -0.886620044708252,
"logits/chosen": 345.7013854980469,
"logits/rejected": 279.64801025390625,
"logps/chosen": -0.9864595532417297,
"logps/rejected": -1.3232471942901611,
"loss": 1.6636,
"nll_loss": 1.7329130172729492,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.04932297766208649,
"rewards/margins": 0.01683938130736351,
"rewards/rejected": -0.0661623626947403,
"step": 1010
},
{
"epoch": 0.8061953931691819,
"grad_norm": 45.76054382324219,
"learning_rate": 5.271487265090163e-07,
"log_odds_chosen": 0.1340581774711609,
"log_odds_ratio": -0.6442204117774963,
"logits/chosen": 253.32131958007812,
"logits/rejected": 293.978759765625,
"logps/chosen": -0.9866166114807129,
"logps/rejected": -1.0805059671401978,
"loss": 1.3506,
"nll_loss": 1.1058270931243896,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.04933083802461624,
"rewards/margins": 0.004694463685154915,
"rewards/rejected": -0.05402529984712601,
"step": 1015
},
{
"epoch": 0.8101667990468626,
"grad_norm": 30.294597625732422,
"learning_rate": 5.06519441809982e-07,
"log_odds_chosen": 0.6967889666557312,
"log_odds_ratio": -0.5373490452766418,
"logits/chosen": 318.8917541503906,
"logits/rejected": 270.747802734375,
"logps/chosen": -1.074588418006897,
"logps/rejected": -1.4418222904205322,
"loss": 1.4542,
"nll_loss": 1.5221776962280273,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.05372941493988037,
"rewards/margins": 0.01836169883608818,
"rewards/rejected": -0.07209111005067825,
"step": 1020
},
{
"epoch": 0.8141382049245433,
"grad_norm": 53.206661224365234,
"learning_rate": 4.862563232989643e-07,
"log_odds_chosen": 0.20271643996238708,
"log_odds_ratio": -0.6173663139343262,
"logits/chosen": 485.3902893066406,
"logits/rejected": 313.65185546875,
"logps/chosen": -1.0637105703353882,
"logps/rejected": -1.2300751209259033,
"loss": 1.4923,
"nll_loss": 1.1044560670852661,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.05318553373217583,
"rewards/margins": 0.008318223990499973,
"rewards/rejected": -0.061503756791353226,
"step": 1025
},
{
"epoch": 0.818109610802224,
"grad_norm": 78.89067077636719,
"learning_rate": 4.663630929429674e-07,
"log_odds_chosen": 0.3912231922149658,
"log_odds_ratio": -0.5344873666763306,
"logits/chosen": 278.98651123046875,
"logits/rejected": 279.85418701171875,
"logps/chosen": -0.7328051924705505,
"logps/rejected": -0.9687238931655884,
"loss": 1.497,
"nll_loss": 1.046706199645996,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.036640264093875885,
"rewards/margins": 0.011795936152338982,
"rewards/rejected": -0.048436202108860016,
"step": 1030
},
{
"epoch": 0.8220810166799047,
"grad_norm": 51.44734573364258,
"learning_rate": 4.46843404767259e-07,
"log_odds_chosen": 0.3833610415458679,
"log_odds_ratio": -0.8149446249008179,
"logits/chosen": 314.2817077636719,
"logits/rejected": 307.8243103027344,
"logps/chosen": -1.357617735862732,
"logps/rejected": -1.7892076969146729,
"loss": 1.459,
"nll_loss": 1.4600521326065063,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.06788089126348495,
"rewards/margins": 0.021579492837190628,
"rewards/rejected": -0.08946038782596588,
"step": 1035
},
{
"epoch": 0.8260524225575854,
"grad_norm": 39.02664566040039,
"learning_rate": 4.2770084418418736e-07,
"log_odds_chosen": 0.27395057678222656,
"log_odds_ratio": -0.7102149128913879,
"logits/chosen": 414.571533203125,
"logits/rejected": 287.80877685546875,
"logps/chosen": -0.8411673307418823,
"logps/rejected": -0.9671209454536438,
"loss": 1.5205,
"nll_loss": 1.5072977542877197,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.042058371007442474,
"rewards/margins": 0.006297673098742962,
"rewards/rejected": -0.04835604503750801,
"step": 1040
},
{
"epoch": 0.8300238284352661,
"grad_norm": 44.10149383544922,
"learning_rate": 4.089389273346084e-07,
"log_odds_chosen": 1.4467592239379883,
"log_odds_ratio": -0.36090224981307983,
"logits/chosen": 336.5260009765625,
"logits/rejected": 268.7255859375,
"logps/chosen": -0.579176664352417,
"logps/rejected": -1.433186650276184,
"loss": 1.4451,
"nll_loss": 2.0459094047546387,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.028958836570382118,
"rewards/margins": 0.042700495570898056,
"rewards/rejected": -0.07165933400392532,
"step": 1045
},
{
"epoch": 0.8339952343129468,
"grad_norm": 29.071208953857422,
"learning_rate": 3.9056110044203594e-07,
"log_odds_chosen": 0.3594241142272949,
"log_odds_ratio": -0.6073828935623169,
"logits/chosen": 340.73431396484375,
"logits/rejected": 293.41363525390625,
"logps/chosen": -1.039194107055664,
"logps/rejected": -1.395516037940979,
"loss": 1.3767,
"nll_loss": 1.1890077590942383,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.05195971205830574,
"rewards/margins": 0.017816094681620598,
"rewards/rejected": -0.06977580487728119,
"step": 1050
},
{
"epoch": 0.8379666401906275,
"grad_norm": 50.16731643676758,
"learning_rate": 3.72570739179631e-07,
"log_odds_chosen": 0.49487370252609253,
"log_odds_ratio": -0.49297910928726196,
"logits/chosen": 280.0902404785156,
"logits/rejected": 468.6788024902344,
"logps/chosen": -1.0811553001403809,
"logps/rejected": -1.4258912801742554,
"loss": 1.453,
"nll_loss": 1.3579655885696411,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.0540577657520771,
"rewards/margins": 0.017236804589629173,
"rewards/rejected": -0.07129456847906113,
"step": 1055
},
{
"epoch": 0.8419380460683081,
"grad_norm": 33.406272888183594,
"learning_rate": 3.5497114805015223e-07,
"log_odds_chosen": 0.14397627115249634,
"log_odds_ratio": -0.6715080142021179,
"logits/chosen": 373.4464416503906,
"logits/rejected": 325.99163818359375,
"logps/chosen": -0.9457298517227173,
"logps/rejected": -1.0310219526290894,
"loss": 1.4311,
"nll_loss": 1.5625425577163696,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.047286491841077805,
"rewards/margins": 0.004264607094228268,
"rewards/rejected": -0.051551103591918945,
"step": 1060
},
{
"epoch": 0.8459094519459889,
"grad_norm": 50.40607833862305,
"learning_rate": 3.377655597789789e-07,
"log_odds_chosen": -0.3970710337162018,
"log_odds_ratio": -1.0087158679962158,
"logits/chosen": 305.99798583984375,
"logits/rejected": 319.93890380859375,
"logps/chosen": -1.3028199672698975,
"logps/rejected": -1.107723593711853,
"loss": 1.5254,
"nll_loss": 1.708296537399292,
"rewards/accuracies": 0.20000000298023224,
"rewards/chosen": -0.06514099985361099,
"rewards/margins": -0.009754816070199013,
"rewards/rejected": -0.05538617819547653,
"step": 1065
},
{
"epoch": 0.8498808578236696,
"grad_norm": 32.866085052490234,
"learning_rate": 3.209571347203197e-07,
"log_odds_chosen": -0.3324670195579529,
"log_odds_ratio": -0.9437648057937622,
"logits/chosen": 288.600830078125,
"logits/rejected": 395.1521911621094,
"logps/chosen": -0.9841065406799316,
"logps/rejected": -0.9132230877876282,
"loss": 1.2591,
"nll_loss": 1.0712697505950928,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.04920532926917076,
"rewards/margins": -0.0035441755317151546,
"rewards/rejected": -0.04566115140914917,
"step": 1070
},
{
"epoch": 0.8538522637013503,
"grad_norm": 63.01152038574219,
"learning_rate": 3.0454896027671073e-07,
"log_odds_chosen": -0.15130704641342163,
"log_odds_ratio": -0.8452315330505371,
"logits/chosen": 408.84552001953125,
"logits/rejected": 319.49493408203125,
"logps/chosen": -1.28704035282135,
"logps/rejected": -1.2764626741409302,
"loss": 1.6339,
"nll_loss": 1.682464361190796,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.06435202062129974,
"rewards/margins": -0.000528886157553643,
"rewards/rejected": -0.06382313370704651,
"step": 1075
},
{
"epoch": 0.857823669579031,
"grad_norm": 32.351932525634766,
"learning_rate": 2.885440503319145e-07,
"log_odds_chosen": -0.4667375087738037,
"log_odds_ratio": -1.0818705558776855,
"logits/chosen": 459.02410888671875,
"logits/rejected": 257.15716552734375,
"logps/chosen": -1.1935930252075195,
"logps/rejected": -0.7796539068222046,
"loss": 1.548,
"nll_loss": 1.7794479131698608,
"rewards/accuracies": 0.20000000298023224,
"rewards/chosen": -0.05967964977025986,
"rewards/margins": -0.020696954801678658,
"rewards/rejected": -0.03898269310593605,
"step": 1080
},
{
"epoch": 0.8617950754567116,
"grad_norm": 36.15253829956055,
"learning_rate": 2.7294534469732794e-07,
"log_odds_chosen": 1.6522690057754517,
"log_odds_ratio": -0.4097142219543457,
"logits/chosen": 344.11962890625,
"logits/rejected": 319.623046875,
"logps/chosen": -0.894513726234436,
"logps/rejected": -2.088923692703247,
"loss": 1.4534,
"nll_loss": 1.7392299175262451,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.0447256863117218,
"rewards/margins": 0.05972049757838249,
"rewards/rejected": -0.10444619506597519,
"step": 1085
},
{
"epoch": 0.8657664813343924,
"grad_norm": 47.798858642578125,
"learning_rate": 2.5775570857199144e-07,
"log_odds_chosen": -0.20097847282886505,
"log_odds_ratio": -0.8294545412063599,
"logits/chosen": 281.59942626953125,
"logits/rejected": 324.57403564453125,
"logps/chosen": -1.1493308544158936,
"logps/rejected": -0.9629890322685242,
"loss": 1.3747,
"nll_loss": 1.3210439682006836,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.0574665442109108,
"rewards/margins": -0.009317094460129738,
"rewards/rejected": -0.04814944788813591,
"step": 1090
},
{
"epoch": 0.8697378872120731,
"grad_norm": 29.671043395996094,
"learning_rate": 2.4297793201630113e-07,
"log_odds_chosen": 0.020139653235673904,
"log_odds_ratio": -0.7046443819999695,
"logits/chosen": 300.97998046875,
"logits/rejected": 333.9520568847656,
"logps/chosen": -0.972399115562439,
"logps/rejected": -0.9247520565986633,
"loss": 1.4489,
"nll_loss": 1.3582823276519775,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.04861995577812195,
"rewards/margins": -0.0023823559749871492,
"rewards/rejected": -0.04623759910464287,
"step": 1095
},
{
"epoch": 0.8737092930897538,
"grad_norm": 30.31666374206543,
"learning_rate": 2.286147294395283e-07,
"log_odds_chosen": 0.29012542963027954,
"log_odds_ratio": -0.7615915536880493,
"logits/chosen": 315.9916687011719,
"logits/rejected": 422.93017578125,
"logps/chosen": -1.0184037685394287,
"logps/rejected": -1.0094739198684692,
"loss": 1.5729,
"nll_loss": 1.4798800945281982,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.05092019587755203,
"rewards/margins": -0.00044649915071204305,
"rewards/rejected": -0.05047369748353958,
"step": 1100
},
{
"epoch": 0.8776806989674345,
"grad_norm": 32.19295120239258,
"learning_rate": 2.1466873910123058e-07,
"log_odds_chosen": -0.07446761429309845,
"log_odds_ratio": -0.8099315762519836,
"logits/chosen": 334.15740966796875,
"logits/rejected": 331.76031494140625,
"logps/chosen": -1.026379942893982,
"logps/rejected": -0.9773386120796204,
"loss": 1.4675,
"nll_loss": 1.4952067136764526,
"rewards/accuracies": 0.20000000298023224,
"rewards/chosen": -0.05131899565458298,
"rewards/margins": -0.002452067332342267,
"rewards/rejected": -0.048866935074329376,
"step": 1105
},
{
"epoch": 0.8816521048451151,
"grad_norm": 34.077449798583984,
"learning_rate": 2.0114252262665086e-07,
"log_odds_chosen": 0.30518868565559387,
"log_odds_ratio": -0.6247283220291138,
"logits/chosen": 313.799560546875,
"logits/rejected": 278.57293701171875,
"logps/chosen": -1.2211542129516602,
"logps/rejected": -1.4464585781097412,
"loss": 1.4043,
"nll_loss": 1.2219561338424683,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.06105770915746689,
"rewards/margins": 0.011265222914516926,
"rewards/rejected": -0.07232292741537094,
"step": 1110
},
{
"epoch": 0.8856235107227959,
"grad_norm": 33.00147247314453,
"learning_rate": 1.880385645361951e-07,
"log_odds_chosen": 2.0590970516204834,
"log_odds_ratio": -0.4269631803035736,
"logits/chosen": 304.67572021484375,
"logits/rejected": 437.95751953125,
"logps/chosen": -0.8812467455863953,
"logps/rejected": -2.7326838970184326,
"loss": 1.5588,
"nll_loss": 1.2233951091766357,
"rewards/accuracies": 1.0,
"rewards/chosen": -0.04406233876943588,
"rewards/margins": 0.09257186949253082,
"rewards/rejected": -0.1366342008113861,
"step": 1115
},
{
"epoch": 0.8895949166004765,
"grad_norm": 29.0517578125,
"learning_rate": 1.7535927178906898e-07,
"log_odds_chosen": 0.6221317648887634,
"log_odds_ratio": -0.5823326110839844,
"logits/chosen": 288.1285705566406,
"logits/rejected": 357.75701904296875,
"logps/chosen": -1.0755956172943115,
"logps/rejected": -1.5605504512786865,
"loss": 1.4892,
"nll_loss": 1.3548178672790527,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.053779780864715576,
"rewards/margins": 0.024247746914625168,
"rewards/rejected": -0.07802753150463104,
"step": 1120
},
{
"epoch": 0.8935663224781573,
"grad_norm": 37.735111236572266,
"learning_rate": 1.6310697334116583e-07,
"log_odds_chosen": -0.2765834629535675,
"log_odds_ratio": -0.8877069354057312,
"logits/chosen": 357.4658203125,
"logits/rejected": 302.4728088378906,
"logps/chosen": -1.121751070022583,
"logps/rejected": -0.9259787797927856,
"loss": 1.4662,
"nll_loss": 1.56728196144104,
"rewards/accuracies": 0.20000000298023224,
"rewards/chosen": -0.05608755350112915,
"rewards/margins": -0.009788615629076958,
"rewards/rejected": -0.04629894345998764,
"step": 1125
},
{
"epoch": 0.897537728355838,
"grad_norm": 39.21086120605469,
"learning_rate": 1.512839197172758e-07,
"log_odds_chosen": 0.1607791781425476,
"log_odds_ratio": -0.7086135745048523,
"logits/chosen": 431.327880859375,
"logits/rejected": 309.34210205078125,
"logps/chosen": -1.1171777248382568,
"logps/rejected": -1.3307039737701416,
"loss": 1.3869,
"nll_loss": 1.7646287679672241,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.05585888773202896,
"rewards/margins": 0.010676311329007149,
"rewards/rejected": -0.06653519719839096,
"step": 1130
},
{
"epoch": 0.9015091342335186,
"grad_norm": 49.552589416503906,
"learning_rate": 1.398922825977092e-07,
"log_odds_chosen": 1.1862103939056396,
"log_odds_ratio": -0.36970359086990356,
"logits/chosen": 287.95318603515625,
"logits/rejected": 344.77764892578125,
"logps/chosen": -1.1711792945861816,
"logps/rejected": -2.0627903938293457,
"loss": 1.4555,
"nll_loss": 1.4977771043777466,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.05855896323919296,
"rewards/margins": 0.044580571353435516,
"rewards/rejected": -0.10313953459262848,
"step": 1135
},
{
"epoch": 0.9054805401111994,
"grad_norm": 34.04672622680664,
"learning_rate": 1.2893415441939588e-07,
"log_odds_chosen": 0.042095281183719635,
"log_odds_ratio": -0.7100853323936462,
"logits/chosen": 319.3915710449219,
"logits/rejected": 372.4937438964844,
"logps/chosen": -0.8466545343399048,
"logps/rejected": -0.8484467267990112,
"loss": 1.3861,
"nll_loss": 1.0329768657684326,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.0423327274620533,
"rewards/margins": 8.96111159818247e-05,
"rewards/rejected": -0.0424223393201828,
"step": 1140
},
{
"epoch": 0.90945194598888,
"grad_norm": 43.70923614501953,
"learning_rate": 1.1841154799154376e-07,
"log_odds_chosen": 0.09307994693517685,
"log_odds_ratio": -0.8356366157531738,
"logits/chosen": 383.5526428222656,
"logits/rejected": 280.0104064941406,
"logps/chosen": -1.0310574769973755,
"logps/rejected": -1.0444084405899048,
"loss": 1.4938,
"nll_loss": 1.3081772327423096,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.051552869379520416,
"rewards/margins": 0.0006675481563434005,
"rewards/rejected": -0.0522204227745533,
"step": 1145
},
{
"epoch": 0.9134233518665608,
"grad_norm": 42.19453048706055,
"learning_rate": 1.083263961259215e-07,
"log_odds_chosen": 0.5610286593437195,
"log_odds_ratio": -0.5318921804428101,
"logits/chosen": 313.7818298339844,
"logits/rejected": 281.55303955078125,
"logps/chosen": -0.8616586923599243,
"logps/rejected": -1.13971745967865,
"loss": 1.3678,
"nll_loss": 1.5508089065551758,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.043082933872938156,
"rewards/margins": 0.01390293799340725,
"rewards/rejected": -0.056985873728990555,
"step": 1150
},
{
"epoch": 0.9173947577442415,
"grad_norm": 62.5867805480957,
"learning_rate": 9.86805512818359e-08,
"log_odds_chosen": 0.5819055438041687,
"log_odds_ratio": -0.5563138723373413,
"logits/chosen": 308.6764221191406,
"logits/rejected": 280.77093505859375,
"logps/chosen": -0.7943710684776306,
"logps/rejected": -1.1860836744308472,
"loss": 1.5596,
"nll_loss": 1.2872906923294067,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.03971855714917183,
"rewards/margins": 0.019585633650422096,
"rewards/rejected": -0.059304188936948776,
"step": 1155
},
{
"epoch": 0.9213661636219221,
"grad_norm": 39.823516845703125,
"learning_rate": 8.947578522587097e-08,
"log_odds_chosen": 0.5100874304771423,
"log_odds_ratio": -0.7220025062561035,
"logits/chosen": 421.84722900390625,
"logits/rejected": 306.4421081542969,
"logps/chosen": -0.7023354768753052,
"logps/rejected": -0.9484280347824097,
"loss": 1.2855,
"nll_loss": 0.8443538546562195,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.0351167730987072,
"rewards/margins": 0.012304631993174553,
"rewards/rejected": -0.0474214032292366,
"step": 1160
},
{
"epoch": 0.9253375694996029,
"grad_norm": 34.763492584228516,
"learning_rate": 8.071378870644381e-08,
"log_odds_chosen": 0.068526491522789,
"log_odds_ratio": -0.7172547578811646,
"logits/chosen": 306.42303466796875,
"logits/rejected": 306.2086486816406,
"logps/chosen": -1.1717783212661743,
"logps/rejected": -1.2519972324371338,
"loss": 1.4611,
"nll_loss": 1.6088817119598389,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.05858892202377319,
"rewards/margins": 0.004010946489870548,
"rewards/rejected": -0.06259986758232117,
"step": 1165
},
{
"epoch": 0.9293089753772835,
"grad_norm": 57.37542724609375,
"learning_rate": 7.239617114324499e-08,
"log_odds_chosen": -0.01497584581375122,
"log_odds_ratio": -0.7752578854560852,
"logits/chosen": 278.559326171875,
"logits/rejected": 374.15869140625,
"logps/chosen": -1.2931535243988037,
"logps/rejected": -1.2771342992782593,
"loss": 1.5557,
"nll_loss": 1.4230643510818481,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.06465767323970795,
"rewards/margins": -0.0008009634912014008,
"rewards/rejected": -0.06385671347379684,
"step": 1170
},
{
"epoch": 0.9332803812549643,
"grad_norm": 54.0869026184082,
"learning_rate": 6.452446033161946e-08,
"log_odds_chosen": -0.19031484425067902,
"log_odds_ratio": -0.8042638897895813,
"logits/chosen": 299.294921875,
"logits/rejected": 398.8330078125,
"logps/chosen": -1.0967943668365479,
"logps/rejected": -1.014966607093811,
"loss": 1.5468,
"nll_loss": 1.4041001796722412,
"rewards/accuracies": 0.20000000298023224,
"rewards/chosen": -0.05483972281217575,
"rewards/margins": -0.0040913899429142475,
"rewards/rejected": -0.05074832960963249,
"step": 1175
},
{
"epoch": 0.937251787132645,
"grad_norm": 30.728069305419922,
"learning_rate": 5.7100102161937176e-08,
"log_odds_chosen": 0.08359535038471222,
"log_odds_ratio": -0.6820527911186218,
"logits/chosen": 354.07806396484375,
"logits/rejected": 264.95947265625,
"logps/chosen": -0.9352580904960632,
"logps/rejected": -0.9271550178527832,
"loss": 1.3888,
"nll_loss": 1.2889841794967651,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.04676290601491928,
"rewards/margins": -0.0004051584692206234,
"rewards/rejected": -0.04635775089263916,
"step": 1180
},
{
"epoch": 0.9412231930103256,
"grad_norm": 30.694597244262695,
"learning_rate": 5.012446035400881e-08,
"log_odds_chosen": -0.18491533398628235,
"log_odds_ratio": -0.8581596612930298,
"logits/chosen": 431.51483154296875,
"logits/rejected": 256.1930236816406,
"logps/chosen": -1.3761582374572754,
"logps/rejected": -1.2526451349258423,
"loss": 1.5465,
"nll_loss": 1.7214624881744385,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.06880791485309601,
"rewards/margins": -0.006175657268613577,
"rewards/rejected": -0.062632255256176,
"step": 1185
},
{
"epoch": 0.9451945988880064,
"grad_norm": 52.927608489990234,
"learning_rate": 4.359881620659534e-08,
"log_odds_chosen": 0.6782919764518738,
"log_odds_ratio": -0.6247768998146057,
"logits/chosen": 436.1754455566406,
"logits/rejected": 285.18365478515625,
"logps/chosen": -0.985084056854248,
"logps/rejected": -1.5693267583847046,
"loss": 1.4816,
"nll_loss": 1.398045301437378,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.04925420135259628,
"rewards/margins": 0.029212135821580887,
"rewards/rejected": -0.07846634089946747,
"step": 1190
},
{
"epoch": 0.949166004765687,
"grad_norm": 48.66614532470703,
"learning_rate": 3.7524368362057415e-08,
"log_odds_chosen": -0.1631685197353363,
"log_odds_ratio": -0.8233796954154968,
"logits/chosen": 418.45928955078125,
"logits/rejected": 318.4512939453125,
"logps/chosen": -0.8457492589950562,
"logps/rejected": -0.7526192665100098,
"loss": 1.4522,
"nll_loss": 1.3443362712860107,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.04228746145963669,
"rewards/margins": -0.004656502045691013,
"rewards/rejected": -0.03763096407055855,
"step": 1195
},
{
"epoch": 0.9531374106433678,
"grad_norm": 29.535236358642578,
"learning_rate": 3.1902232586185635e-08,
"log_odds_chosen": 0.14746162295341492,
"log_odds_ratio": -0.7185263633728027,
"logits/chosen": 346.12823486328125,
"logits/rejected": 365.977783203125,
"logps/chosen": -0.9943090677261353,
"logps/rejected": -0.9647199511528015,
"loss": 1.5476,
"nll_loss": 1.3550399541854858,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.04971545934677124,
"rewards/margins": -0.001479454687796533,
"rewards/rejected": -0.048236001282930374,
"step": 1200
},
{
"epoch": 0.9571088165210484,
"grad_norm": 46.18953323364258,
"learning_rate": 2.673344156325558e-08,
"log_odds_chosen": 0.25677961111068726,
"log_odds_ratio": -0.6748029589653015,
"logits/chosen": 321.4141845703125,
"logits/rejected": 390.1539611816406,
"logps/chosen": -1.306235432624817,
"logps/rejected": -1.5235410928726196,
"loss": 1.4511,
"nll_loss": 1.348838448524475,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.06531177461147308,
"rewards/margins": 0.010865284129977226,
"rewards/rejected": -0.07617706060409546,
"step": 1205
},
{
"epoch": 0.9610802223987291,
"grad_norm": 48.14284133911133,
"learning_rate": 2.2018944706341528e-08,
"log_odds_chosen": 0.25043779611587524,
"log_odds_ratio": -0.6027860045433044,
"logits/chosen": 383.79913330078125,
"logits/rejected": 336.7133483886719,
"logps/chosen": -1.0423628091812134,
"logps/rejected": -1.1916873455047607,
"loss": 1.4619,
"nll_loss": 1.6256072521209717,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.05211814120411873,
"rewards/margins": 0.007466231472790241,
"rewards/rejected": -0.059584371745586395,
"step": 1210
},
{
"epoch": 0.9650516282764099,
"grad_norm": 32.60062026977539,
"learning_rate": 1.775960798292731e-08,
"log_odds_chosen": -0.04052457585930824,
"log_odds_ratio": -0.8082249760627747,
"logits/chosen": 331.26397705078125,
"logits/rejected": 329.55096435546875,
"logps/chosen": -1.347312092781067,
"logps/rejected": -1.3537461757659912,
"loss": 1.5079,
"nll_loss": 1.5286957025527954,
"rewards/accuracies": 0.20000000298023224,
"rewards/chosen": -0.0673656016588211,
"rewards/margins": 0.00032170413760468364,
"rewards/rejected": -0.06768731772899628,
"step": 1215
},
{
"epoch": 0.9690230341540905,
"grad_norm": 43.27484893798828,
"learning_rate": 1.3956213755842718e-08,
"log_odds_chosen": -0.26221469044685364,
"log_odds_ratio": -0.9925212860107422,
"logits/chosen": 304.05120849609375,
"logits/rejected": 393.05517578125,
"logps/chosen": -1.015187382698059,
"logps/rejected": -0.9859923124313354,
"loss": 1.3977,
"nll_loss": 1.1624058485031128,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.050759367644786835,
"rewards/margins": -0.0014597497647628188,
"rewards/rejected": -0.04929962009191513,
"step": 1220
},
{
"epoch": 0.9729944400317713,
"grad_norm": 43.98046875,
"learning_rate": 1.0609460639559033e-08,
"log_odds_chosen": 0.6014237999916077,
"log_odds_ratio": -0.5421421527862549,
"logits/chosen": 339.303955078125,
"logits/rejected": 391.6441345214844,
"logps/chosen": -1.0844001770019531,
"logps/rejected": -1.4398688077926636,
"loss": 1.3258,
"nll_loss": 1.4467300176620483,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.054220009595155716,
"rewards/margins": 0.017773432657122612,
"rewards/rejected": -0.07199344038963318,
"step": 1225
},
{
"epoch": 0.9769658459094519,
"grad_norm": 34.32419967651367,
"learning_rate": 7.719963371865259e-09,
"log_odds_chosen": 0.7910041809082031,
"log_odds_ratio": -0.4584100842475891,
"logits/chosen": 316.9669494628906,
"logits/rejected": 363.3733825683594,
"logps/chosen": -0.8123126029968262,
"logps/rejected": -1.2943851947784424,
"loss": 1.3404,
"nll_loss": 1.313084363937378,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.04061562940478325,
"rewards/margins": 0.02410362847149372,
"rewards/rejected": -0.06471925973892212,
"step": 1230
},
{
"epoch": 0.9809372517871326,
"grad_norm": 42.33149337768555,
"learning_rate": 5.288252700952068e-09,
"log_odds_chosen": 0.8038623929023743,
"log_odds_ratio": -0.43806830048561096,
"logits/chosen": 310.6473083496094,
"logits/rejected": 314.3547668457031,
"logps/chosen": -0.8453804850578308,
"logps/rejected": -1.4276442527770996,
"loss": 1.4819,
"nll_loss": 1.6422332525253296,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.0422690249979496,
"rewards/margins": 0.02911318838596344,
"rewards/rejected": -0.07138221710920334,
"step": 1235
},
{
"epoch": 0.9849086576648134,
"grad_norm": 45.44175720214844,
"learning_rate": 3.3147752879236773e-09,
"log_odds_chosen": 0.02175927720963955,
"log_odds_ratio": -0.8097355961799622,
"logits/chosen": 277.007080078125,
"logits/rejected": 415.473388671875,
"logps/chosen": -1.0552597045898438,
"logps/rejected": -1.2169201374053955,
"loss": 1.5311,
"nll_loss": 1.5463203191757202,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.05276298522949219,
"rewards/margins": 0.008083020336925983,
"rewards/rejected": -0.060846008360385895,
"step": 1240
},
{
"epoch": 0.988880063542494,
"grad_norm": 62.612850189208984,
"learning_rate": 1.7998936247534681e-09,
"log_odds_chosen": 0.2898419499397278,
"log_odds_ratio": -0.5893855094909668,
"logits/chosen": 399.6653747558594,
"logits/rejected": 274.2057800292969,
"logps/chosen": -1.0792999267578125,
"logps/rejected": -1.2781049013137817,
"loss": 1.4564,
"nll_loss": 1.5806140899658203,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.053964994847774506,
"rewards/margins": 0.009940249845385551,
"rewards/rejected": -0.0639052465558052,
"step": 1245
},
{
"epoch": 0.9928514694201748,
"grad_norm": 51.08999252319336,
"learning_rate": 7.438859677008636e-10,
"log_odds_chosen": 0.09502691775560379,
"log_odds_ratio": -0.7220104336738586,
"logits/chosen": 325.39263916015625,
"logits/rejected": 361.7767639160156,
"logps/chosen": -1.1207284927368164,
"logps/rejected": -1.173607587814331,
"loss": 1.3621,
"nll_loss": 1.241156816482544,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.05603642389178276,
"rewards/margins": 0.0026439554058015347,
"rewards/rejected": -0.05868038535118103,
"step": 1250
},
{
"epoch": 0.9968228752978554,
"grad_norm": 44.36956787109375,
"learning_rate": 1.4694628620137708e-10,
"log_odds_chosen": -0.33857935667037964,
"log_odds_ratio": -0.9029590487480164,
"logits/chosen": 293.49420166015625,
"logits/rejected": 431.3218688964844,
"logps/chosen": -0.9575881958007812,
"logps/rejected": -0.7752519845962524,
"loss": 1.3423,
"nll_loss": 1.1781036853790283,
"rewards/accuracies": 0.20000000298023224,
"rewards/chosen": -0.0478794127702713,
"rewards/margins": -0.009116815403103828,
"rewards/rejected": -0.038762595504522324,
"step": 1255
},
{
"epoch": 1.0,
"eval_log_odds_chosen": 0.23976314067840576,
"eval_log_odds_ratio": -0.6928443908691406,
"eval_logits/chosen": 340.5321350097656,
"eval_logits/rejected": 312.9670104980469,
"eval_logps/chosen": -1.0267834663391113,
"eval_logps/rejected": -1.1786600351333618,
"eval_loss": 1.455647587776184,
"eval_nll_loss": 1.4095592498779297,
"eval_rewards/accuracies": 0.5107913613319397,
"eval_rewards/chosen": -0.051339175552129745,
"eval_rewards/margins": 0.0075938161462545395,
"eval_rewards/rejected": -0.058932989835739136,
"eval_runtime": 91.5981,
"eval_samples_per_second": 6.037,
"eval_steps_per_second": 1.517,
"step": 1259
},
{
"epoch": 1.0,
"step": 1259,
"total_flos": 0.0,
"train_loss": 1.8019611810861456,
"train_runtime": 4470.8327,
"train_samples_per_second": 1.126,
"train_steps_per_second": 0.282
}
],
"logging_steps": 5,
"max_steps": 1259,
"num_input_tokens_seen": 0,
"num_train_epochs": 1,
"save_steps": 500,
"stateful_callbacks": {
"TrainerControl": {
"args": {
"should_epoch_stop": false,
"should_evaluate": false,
"should_log": false,
"should_save": false,
"should_training_stop": false
},
"attributes": {}
}
},
"total_flos": 0.0,
"train_batch_size": 1,
"trial_name": null,
"trial_params": null
}