gemma-7b-borpo-low-quality-v2 / trainer_state.json
c-alfano's picture
Model save
4190f79 verified
{
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 3.0,
"eval_steps": 500,
"global_step": 504,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.02976190476190476,
"grad_norm": 1776.2471923828125,
"learning_rate": 2.5000000000000004e-07,
"log_odds_chosen": -0.3503778874874115,
"log_odds_ratio": -1.0639952421188354,
"logits/chosen": 125.49534606933594,
"logits/rejected": 180.3563232421875,
"logps/chosen": -15.494331359863281,
"logps/rejected": -15.143954277038574,
"loss": 14.9629,
"nll_loss": 15.042287826538086,
"rewards/accuracies": 0.25,
"rewards/chosen": -0.7747165560722351,
"rewards/margins": -0.01751876249909401,
"rewards/rejected": -0.7571978569030762,
"step": 5
},
{
"epoch": 0.05952380952380952,
"grad_norm": 1193.9168701171875,
"learning_rate": 5.000000000000001e-07,
"log_odds_chosen": 0.06253216415643692,
"log_odds_ratio": -0.8492839932441711,
"logits/chosen": 210.31631469726562,
"logits/rejected": 245.1850128173828,
"logps/chosen": -12.538459777832031,
"logps/rejected": -12.600992202758789,
"loss": 12.6111,
"nll_loss": 12.388693809509277,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.6269229650497437,
"rewards/margins": 0.0031266347505152225,
"rewards/rejected": -0.6300495862960815,
"step": 10
},
{
"epoch": 0.08928571428571429,
"grad_norm": 723.246826171875,
"learning_rate": 7.5e-07,
"log_odds_chosen": -0.26854413747787476,
"log_odds_ratio": -0.9210551977157593,
"logits/chosen": 300.58953857421875,
"logits/rejected": 289.635009765625,
"logps/chosen": -8.324029922485352,
"logps/rejected": -8.055707931518555,
"loss": 8.2823,
"nll_loss": 8.378087997436523,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.41620150208473206,
"rewards/margins": -0.013416108675301075,
"rewards/rejected": -0.40278539061546326,
"step": 15
},
{
"epoch": 0.11904761904761904,
"grad_norm": 216.30667114257812,
"learning_rate": 1.0000000000000002e-06,
"log_odds_chosen": -0.09513016790151596,
"log_odds_ratio": -0.9796999096870422,
"logits/chosen": 255.70156860351562,
"logits/rejected": 235.23721313476562,
"logps/chosen": -5.404101848602295,
"logps/rejected": -5.310047149658203,
"loss": 5.4458,
"nll_loss": 5.430812835693359,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": -0.2702050805091858,
"rewards/margins": -0.004702714271843433,
"rewards/rejected": -0.265502393245697,
"step": 20
},
{
"epoch": 0.1488095238095238,
"grad_norm": 164.33297729492188,
"learning_rate": 1.25e-06,
"log_odds_chosen": 0.25551921129226685,
"log_odds_ratio": -0.6588834524154663,
"logits/chosen": 274.9960632324219,
"logits/rejected": 302.8502502441406,
"logps/chosen": -3.1612606048583984,
"logps/rejected": -3.390745162963867,
"loss": 3.5011,
"nll_loss": 3.502755641937256,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.15806300938129425,
"rewards/margins": 0.01147423591464758,
"rewards/rejected": -0.1695372611284256,
"step": 25
},
{
"epoch": 0.17857142857142858,
"grad_norm": 78.93386840820312,
"learning_rate": 1.5e-06,
"log_odds_chosen": 0.24210360646247864,
"log_odds_ratio": -0.6352943181991577,
"logits/chosen": 290.36334228515625,
"logits/rejected": 385.3616638183594,
"logps/chosen": -2.418250799179077,
"logps/rejected": -2.630479097366333,
"loss": 2.5553,
"nll_loss": 2.509606122970581,
"rewards/accuracies": 0.699999988079071,
"rewards/chosen": -0.12091253697872162,
"rewards/margins": 0.01061142049729824,
"rewards/rejected": -0.1315239518880844,
"step": 30
},
{
"epoch": 0.20833333333333334,
"grad_norm": 42.19651412963867,
"learning_rate": 1.75e-06,
"log_odds_chosen": 0.4038185477256775,
"log_odds_ratio": -0.5650765299797058,
"logits/chosen": 365.90753173828125,
"logits/rejected": 389.90338134765625,
"logps/chosen": -1.5057640075683594,
"logps/rejected": -1.8460853099822998,
"loss": 2.129,
"nll_loss": 1.8491789102554321,
"rewards/accuracies": 0.75,
"rewards/chosen": -0.07528821378946304,
"rewards/margins": 0.01701606810092926,
"rewards/rejected": -0.0923042744398117,
"step": 35
},
{
"epoch": 0.23809523809523808,
"grad_norm": 35.26750183105469,
"learning_rate": 2.0000000000000003e-06,
"log_odds_chosen": 0.15938052535057068,
"log_odds_ratio": -0.6910892724990845,
"logits/chosen": 331.43438720703125,
"logits/rejected": 372.8791198730469,
"logps/chosen": -1.6345170736312866,
"logps/rejected": -1.7539236545562744,
"loss": 1.935,
"nll_loss": 1.968824028968811,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.08172585070133209,
"rewards/margins": 0.005970318801701069,
"rewards/rejected": -0.08769617974758148,
"step": 40
},
{
"epoch": 0.26785714285714285,
"grad_norm": 87.45353698730469,
"learning_rate": 2.25e-06,
"log_odds_chosen": -0.15498922765254974,
"log_odds_ratio": -0.8517929911613464,
"logits/chosen": 386.3527526855469,
"logits/rejected": 371.41851806640625,
"logps/chosen": -1.4574557542800903,
"logps/rejected": -1.3248956203460693,
"loss": 1.8765,
"nll_loss": 1.7655121088027954,
"rewards/accuracies": 0.44999998807907104,
"rewards/chosen": -0.07287278026342392,
"rewards/margins": -0.006628001574426889,
"rewards/rejected": -0.06624479591846466,
"step": 45
},
{
"epoch": 0.2976190476190476,
"grad_norm": 41.48765563964844,
"learning_rate": 2.5e-06,
"log_odds_chosen": 0.043729472905397415,
"log_odds_ratio": -0.7361106872558594,
"logits/chosen": 437.00823974609375,
"logits/rejected": 443.38232421875,
"logps/chosen": -1.658603310585022,
"logps/rejected": -1.6873445510864258,
"loss": 1.8933,
"nll_loss": 2.108198642730713,
"rewards/accuracies": 0.550000011920929,
"rewards/chosen": -0.08293016999959946,
"rewards/margins": 0.0014370663557201624,
"rewards/rejected": -0.08436723798513412,
"step": 50
},
{
"epoch": 0.3273809523809524,
"grad_norm": 51.39496994018555,
"learning_rate": 2.7500000000000004e-06,
"log_odds_chosen": 0.10381323099136353,
"log_odds_ratio": -0.7126073241233826,
"logits/chosen": 365.14544677734375,
"logits/rejected": 364.1676330566406,
"logps/chosen": -1.4783101081848145,
"logps/rejected": -1.550569772720337,
"loss": 1.8734,
"nll_loss": 1.8976672887802124,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.073915496468544,
"rewards/margins": 0.003612985834479332,
"rewards/rejected": -0.07752849161624908,
"step": 55
},
{
"epoch": 0.35714285714285715,
"grad_norm": 47.07830810546875,
"learning_rate": 3e-06,
"log_odds_chosen": 0.20662184059619904,
"log_odds_ratio": -0.6513880491256714,
"logits/chosen": 386.7112731933594,
"logits/rejected": 356.2147216796875,
"logps/chosen": -1.1965901851654053,
"logps/rejected": -1.3431804180145264,
"loss": 1.6989,
"nll_loss": 1.9430646896362305,
"rewards/accuracies": 0.550000011920929,
"rewards/chosen": -0.05982951074838638,
"rewards/margins": 0.007329514715820551,
"rewards/rejected": -0.0671590268611908,
"step": 60
},
{
"epoch": 0.3869047619047619,
"grad_norm": 116.96641540527344,
"learning_rate": 3.2500000000000002e-06,
"log_odds_chosen": 0.029183167964220047,
"log_odds_ratio": -0.7434911727905273,
"logits/chosen": 385.2690734863281,
"logits/rejected": 389.1334228515625,
"logps/chosen": -1.3286149501800537,
"logps/rejected": -1.3333098888397217,
"loss": 1.636,
"nll_loss": 1.6027374267578125,
"rewards/accuracies": 0.6499999761581421,
"rewards/chosen": -0.06643076241016388,
"rewards/margins": 0.00023474842600990087,
"rewards/rejected": -0.06666550040245056,
"step": 65
},
{
"epoch": 0.4166666666666667,
"grad_norm": 40.99021530151367,
"learning_rate": 3.5e-06,
"log_odds_chosen": 0.045442335307598114,
"log_odds_ratio": -0.7358086705207825,
"logits/chosen": 395.22906494140625,
"logits/rejected": 390.57305908203125,
"logps/chosen": -1.7554155588150024,
"logps/rejected": -1.765885591506958,
"loss": 1.6688,
"nll_loss": 1.9699609279632568,
"rewards/accuracies": 0.550000011920929,
"rewards/chosen": -0.08777078241109848,
"rewards/margins": 0.0005234999698586762,
"rewards/rejected": -0.08829428255558014,
"step": 70
},
{
"epoch": 0.44642857142857145,
"grad_norm": 67.57577514648438,
"learning_rate": 3.7500000000000005e-06,
"log_odds_chosen": 0.24833233654499054,
"log_odds_ratio": -0.6203548908233643,
"logits/chosen": 386.412109375,
"logits/rejected": 376.1787414550781,
"logps/chosen": -1.1873780488967896,
"logps/rejected": -1.3402230739593506,
"loss": 1.5454,
"nll_loss": 1.4503333568572998,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.059368908405303955,
"rewards/margins": 0.0076422421261668205,
"rewards/rejected": -0.06701115518808365,
"step": 75
},
{
"epoch": 0.47619047619047616,
"grad_norm": 29.741676330566406,
"learning_rate": 4.000000000000001e-06,
"log_odds_chosen": 0.29179519414901733,
"log_odds_ratio": -0.6332293748855591,
"logits/chosen": 389.09295654296875,
"logits/rejected": 405.5395202636719,
"logps/chosen": -1.2367178201675415,
"logps/rejected": -1.3924906253814697,
"loss": 1.5966,
"nll_loss": 1.5792012214660645,
"rewards/accuracies": 0.550000011920929,
"rewards/chosen": -0.061835892498493195,
"rewards/margins": 0.007788646034896374,
"rewards/rejected": -0.06962453573942184,
"step": 80
},
{
"epoch": 0.5059523809523809,
"grad_norm": 53.26546096801758,
"learning_rate": 4.25e-06,
"log_odds_chosen": -0.2370494306087494,
"log_odds_ratio": -0.9606464505195618,
"logits/chosen": 405.4431457519531,
"logits/rejected": 386.801025390625,
"logps/chosen": -1.5545951128005981,
"logps/rejected": -1.2866960763931274,
"loss": 1.6757,
"nll_loss": 1.9027442932128906,
"rewards/accuracies": 0.5,
"rewards/chosen": -0.07772975414991379,
"rewards/margins": -0.013394953683018684,
"rewards/rejected": -0.06433480232954025,
"step": 85
},
{
"epoch": 0.5357142857142857,
"grad_norm": 28.4445743560791,
"learning_rate": 4.5e-06,
"log_odds_chosen": 0.7336848378181458,
"log_odds_ratio": -0.5354053378105164,
"logits/chosen": 416.51580810546875,
"logits/rejected": 425.767578125,
"logps/chosen": -1.2461140155792236,
"logps/rejected": -1.842694878578186,
"loss": 1.4576,
"nll_loss": 1.5203516483306885,
"rewards/accuracies": 0.699999988079071,
"rewards/chosen": -0.06230570003390312,
"rewards/margins": 0.02982904389500618,
"rewards/rejected": -0.0921347513794899,
"step": 90
},
{
"epoch": 0.5654761904761905,
"grad_norm": 46.642581939697266,
"learning_rate": 4.75e-06,
"log_odds_chosen": 0.25576168298721313,
"log_odds_ratio": -0.6427351832389832,
"logits/chosen": 347.84832763671875,
"logits/rejected": 380.0849609375,
"logps/chosen": -1.119447946548462,
"logps/rejected": -1.234407663345337,
"loss": 1.523,
"nll_loss": 1.489450454711914,
"rewards/accuracies": 0.550000011920929,
"rewards/chosen": -0.05597240477800369,
"rewards/margins": 0.005747987423092127,
"rewards/rejected": -0.06172039359807968,
"step": 95
},
{
"epoch": 0.5952380952380952,
"grad_norm": 43.77986526489258,
"learning_rate": 5e-06,
"log_odds_chosen": 0.12196314334869385,
"log_odds_ratio": -0.7031902074813843,
"logits/chosen": 417.86639404296875,
"logits/rejected": 436.10858154296875,
"logps/chosen": -1.2873280048370361,
"logps/rejected": -1.3734104633331299,
"loss": 1.5418,
"nll_loss": 1.5198986530303955,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.0643664002418518,
"rewards/margins": 0.004304117523133755,
"rewards/rejected": -0.06867052614688873,
"step": 100
},
{
"epoch": 0.625,
"grad_norm": 32.244384765625,
"learning_rate": 4.8795003647426654e-06,
"log_odds_chosen": 0.27784407138824463,
"log_odds_ratio": -0.6329095959663391,
"logits/chosen": 412.528564453125,
"logits/rejected": 394.52130126953125,
"logps/chosen": -1.0501761436462402,
"logps/rejected": -1.182625412940979,
"loss": 1.5085,
"nll_loss": 1.6539300680160522,
"rewards/accuracies": 0.699999988079071,
"rewards/chosen": -0.05250881239771843,
"rewards/margins": 0.006622466258704662,
"rewards/rejected": -0.059131283313035965,
"step": 105
},
{
"epoch": 0.6547619047619048,
"grad_norm": 69.666259765625,
"learning_rate": 4.767312946227961e-06,
"log_odds_chosen": 0.30888259410858154,
"log_odds_ratio": -0.5859929919242859,
"logits/chosen": 375.71612548828125,
"logits/rejected": 356.3163146972656,
"logps/chosen": -1.0360513925552368,
"logps/rejected": -1.2427626848220825,
"loss": 1.5381,
"nll_loss": 1.5933644771575928,
"rewards/accuracies": 0.699999988079071,
"rewards/chosen": -0.05180257558822632,
"rewards/margins": 0.01033556554466486,
"rewards/rejected": -0.06213812902569771,
"step": 110
},
{
"epoch": 0.6845238095238095,
"grad_norm": 34.408199310302734,
"learning_rate": 4.662524041201569e-06,
"log_odds_chosen": 0.2710966467857361,
"log_odds_ratio": -0.6335155367851257,
"logits/chosen": 389.82244873046875,
"logits/rejected": 437.4444274902344,
"logps/chosen": -0.9292726516723633,
"logps/rejected": -1.0246301889419556,
"loss": 1.4878,
"nll_loss": 1.3399275541305542,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.04646363481879234,
"rewards/margins": 0.0047678761184215546,
"rewards/rejected": -0.0512315109372139,
"step": 115
},
{
"epoch": 0.7142857142857143,
"grad_norm": 41.90970230102539,
"learning_rate": 4.564354645876385e-06,
"log_odds_chosen": 0.2518194615840912,
"log_odds_ratio": -0.6617739796638489,
"logits/chosen": 393.5774841308594,
"logits/rejected": 395.57147216796875,
"logps/chosen": -1.0596468448638916,
"logps/rejected": -1.1526801586151123,
"loss": 1.5378,
"nll_loss": 1.5159982442855835,
"rewards/accuracies": 0.550000011920929,
"rewards/chosen": -0.05298234894871712,
"rewards/margins": 0.004651663359254599,
"rewards/rejected": -0.05763401836156845,
"step": 120
},
{
"epoch": 0.7440476190476191,
"grad_norm": 30.15009880065918,
"learning_rate": 4.47213595499958e-06,
"log_odds_chosen": -0.14496295154094696,
"log_odds_ratio": -0.9012987017631531,
"logits/chosen": 378.92242431640625,
"logits/rejected": 388.03662109375,
"logps/chosen": -1.3249809741973877,
"logps/rejected": -1.1592341661453247,
"loss": 1.4803,
"nll_loss": 1.3684661388397217,
"rewards/accuracies": 0.44999998807907104,
"rewards/chosen": -0.0662490501999855,
"rewards/margins": -0.00828734040260315,
"rewards/rejected": -0.05796170234680176,
"step": 125
},
{
"epoch": 0.7738095238095238,
"grad_norm": 24.6285457611084,
"learning_rate": 4.385290096535147e-06,
"log_odds_chosen": 0.3137187361717224,
"log_odds_ratio": -0.6695318818092346,
"logits/chosen": 401.2060546875,
"logits/rejected": 372.0254821777344,
"logps/chosen": -1.065645456314087,
"logps/rejected": -1.1312780380249023,
"loss": 1.5122,
"nll_loss": 1.4193568229675293,
"rewards/accuracies": 0.5,
"rewards/chosen": -0.053282272070646286,
"rewards/margins": 0.003281626384705305,
"rewards/rejected": -0.05656389519572258,
"step": 130
},
{
"epoch": 0.8035714285714286,
"grad_norm": 22.474136352539062,
"learning_rate": 4.303314829119352e-06,
"log_odds_chosen": 0.08262600004673004,
"log_odds_ratio": -0.7314961552619934,
"logits/chosen": 373.8929138183594,
"logits/rejected": 405.0035705566406,
"logps/chosen": -1.1277202367782593,
"logps/rejected": -1.2345101833343506,
"loss": 1.5189,
"nll_loss": 1.4631812572479248,
"rewards/accuracies": 0.550000011920929,
"rewards/chosen": -0.056386012583971024,
"rewards/margins": 0.005339500494301319,
"rewards/rejected": -0.06172550842165947,
"step": 135
},
{
"epoch": 0.8333333333333334,
"grad_norm": 23.9536190032959,
"learning_rate": 4.2257712736425835e-06,
"log_odds_chosen": -0.12259285151958466,
"log_odds_ratio": -0.7952130436897278,
"logits/chosen": 394.89117431640625,
"logits/rejected": 361.0680847167969,
"logps/chosen": -1.0518170595169067,
"logps/rejected": -0.986822247505188,
"loss": 1.5033,
"nll_loss": 1.4649903774261475,
"rewards/accuracies": 0.44999998807907104,
"rewards/chosen": -0.052590854465961456,
"rewards/margins": -0.003249742556363344,
"rewards/rejected": -0.0493411123752594,
"step": 140
},
{
"epoch": 0.8630952380952381,
"grad_norm": 22.85877227783203,
"learning_rate": 4.1522739926869985e-06,
"log_odds_chosen": -0.07754195481538773,
"log_odds_ratio": -0.7574380040168762,
"logits/chosen": 365.57147216796875,
"logits/rejected": 380.41046142578125,
"logps/chosen": -1.0989421606063843,
"logps/rejected": -1.0417240858078003,
"loss": 1.4892,
"nll_loss": 1.3651247024536133,
"rewards/accuracies": 0.550000011920929,
"rewards/chosen": -0.05494711548089981,
"rewards/margins": -0.0028609037399291992,
"rewards/rejected": -0.052086204290390015,
"step": 145
},
{
"epoch": 0.8928571428571429,
"grad_norm": 28.741422653198242,
"learning_rate": 4.082482904638631e-06,
"log_odds_chosen": 0.4838029742240906,
"log_odds_ratio": -0.5155819654464722,
"logits/chosen": 390.33367919921875,
"logits/rejected": 395.64117431640625,
"logps/chosen": -1.0308912992477417,
"logps/rejected": -1.3601925373077393,
"loss": 1.4181,
"nll_loss": 1.4082247018814087,
"rewards/accuracies": 0.75,
"rewards/chosen": -0.051544565707445145,
"rewards/margins": 0.016465062275528908,
"rewards/rejected": -0.0680096298456192,
"step": 150
},
{
"epoch": 0.9226190476190477,
"grad_norm": 29.028520584106445,
"learning_rate": 4.016096644512495e-06,
"log_odds_chosen": 0.07582991570234299,
"log_odds_ratio": -0.7352056503295898,
"logits/chosen": 384.6861572265625,
"logits/rejected": 382.9585876464844,
"logps/chosen": -1.1874030828475952,
"logps/rejected": -1.2361079454421997,
"loss": 1.4165,
"nll_loss": 1.365952730178833,
"rewards/accuracies": 0.550000011920929,
"rewards/chosen": -0.05937015265226364,
"rewards/margins": 0.0024352427572011948,
"rewards/rejected": -0.06180540472269058,
"step": 155
},
{
"epoch": 0.9523809523809523,
"grad_norm": 47.11198043823242,
"learning_rate": 3.952847075210474e-06,
"log_odds_chosen": 0.17520497739315033,
"log_odds_ratio": -0.691135048866272,
"logits/chosen": 389.0434265136719,
"logits/rejected": 431.94989013671875,
"logps/chosen": -0.9387935400009155,
"logps/rejected": -1.0681307315826416,
"loss": 1.3893,
"nll_loss": 1.2998160123825073,
"rewards/accuracies": 0.44999998807907104,
"rewards/chosen": -0.046939678490161896,
"rewards/margins": 0.006466855760663748,
"rewards/rejected": -0.05340652912855148,
"step": 160
},
{
"epoch": 0.9821428571428571,
"grad_norm": 16.736385345458984,
"learning_rate": 3.892494720807615e-06,
"log_odds_chosen": -0.01084871869534254,
"log_odds_ratio": -0.759717583656311,
"logits/chosen": 427.4103088378906,
"logits/rejected": 431.9371032714844,
"logps/chosen": -1.1341661214828491,
"logps/rejected": -1.1380488872528076,
"loss": 1.4218,
"nll_loss": 1.4594279527664185,
"rewards/accuracies": 0.5,
"rewards/chosen": -0.05670831352472305,
"rewards/margins": 0.00019413381232880056,
"rewards/rejected": -0.0569024458527565,
"step": 165
},
{
"epoch": 1.0,
"eval_log_odds_chosen": 0.23235350847244263,
"eval_log_odds_ratio": -0.6711738109588623,
"eval_logits/chosen": 326.8028564453125,
"eval_logits/rejected": 273.7525634765625,
"eval_logps/chosen": -1.007071614265442,
"eval_logps/rejected": -1.159067988395691,
"eval_loss": 1.4487849473953247,
"eval_nll_loss": 1.455283284187317,
"eval_rewards/accuracies": 0.5571428537368774,
"eval_rewards/chosen": -0.050353582948446274,
"eval_rewards/margins": 0.007599818520247936,
"eval_rewards/rejected": -0.057953398674726486,
"eval_runtime": 201.1698,
"eval_samples_per_second": 2.749,
"eval_steps_per_second": 0.348,
"step": 168
},
{
"epoch": 1.0119047619047619,
"grad_norm": 19.479097366333008,
"learning_rate": 3.834824944236852e-06,
"log_odds_chosen": 0.4537879526615143,
"log_odds_ratio": -0.5628765225410461,
"logits/chosen": 367.5221252441406,
"logits/rejected": 402.8450622558594,
"logps/chosen": -0.841064453125,
"logps/rejected": -1.091392159461975,
"loss": 1.2782,
"nll_loss": 1.1047321557998657,
"rewards/accuracies": 0.699999988079071,
"rewards/chosen": -0.0420532263815403,
"rewards/margins": 0.012516376562416553,
"rewards/rejected": -0.054569609463214874,
"step": 170
},
{
"epoch": 1.0416666666666667,
"grad_norm": 18.193614959716797,
"learning_rate": 3.7796447300922724e-06,
"log_odds_chosen": 0.6258941292762756,
"log_odds_ratio": -0.5433114171028137,
"logits/chosen": 404.79443359375,
"logits/rejected": 432.61151123046875,
"logps/chosen": -0.7016412615776062,
"logps/rejected": -1.0624569654464722,
"loss": 1.05,
"nll_loss": 1.2699940204620361,
"rewards/accuracies": 0.8500000238418579,
"rewards/chosen": -0.03508206456899643,
"rewards/margins": 0.018040789291262627,
"rewards/rejected": -0.053122855722904205,
"step": 175
},
{
"epoch": 1.0714285714285714,
"grad_norm": 15.39194393157959,
"learning_rate": 3.72677996249965e-06,
"log_odds_chosen": 0.7378727197647095,
"log_odds_ratio": -0.47085660696029663,
"logits/chosen": 343.8995056152344,
"logits/rejected": 338.2045593261719,
"logps/chosen": -0.8047459721565247,
"logps/rejected": -1.2078239917755127,
"loss": 1.1048,
"nll_loss": 1.2225919961929321,
"rewards/accuracies": 0.699999988079071,
"rewards/chosen": -0.04023730009794235,
"rewards/margins": 0.020153898745775223,
"rewards/rejected": -0.060391198843717575,
"step": 180
},
{
"epoch": 1.1011904761904763,
"grad_norm": 20.72310447692871,
"learning_rate": 3.6760731104690393e-06,
"log_odds_chosen": 1.557734727859497,
"log_odds_ratio": -0.26492685079574585,
"logits/chosen": 435.18865966796875,
"logits/rejected": 362.96868896484375,
"logps/chosen": -0.6113921403884888,
"logps/rejected": -1.4927384853363037,
"loss": 0.9894,
"nll_loss": 1.086829423904419,
"rewards/accuracies": 0.949999988079071,
"rewards/chosen": -0.03056960739195347,
"rewards/margins": 0.044067323207855225,
"rewards/rejected": -0.07463693618774414,
"step": 185
},
{
"epoch": 1.130952380952381,
"grad_norm": 16.229610443115234,
"learning_rate": 3.6273812505500587e-06,
"log_odds_chosen": 0.5972079038619995,
"log_odds_ratio": -0.5253661870956421,
"logits/chosen": 386.9327087402344,
"logits/rejected": 405.801513671875,
"logps/chosen": -0.7007042169570923,
"logps/rejected": -1.028989553451538,
"loss": 1.1004,
"nll_loss": 1.0136523246765137,
"rewards/accuracies": 0.6499999761581421,
"rewards/chosen": -0.03503521531820297,
"rewards/margins": 0.01641426980495453,
"rewards/rejected": -0.0514494851231575,
"step": 190
},
{
"epoch": 1.1607142857142858,
"grad_norm": 21.483884811401367,
"learning_rate": 3.5805743701971648e-06,
"log_odds_chosen": 0.9149399995803833,
"log_odds_ratio": -0.4483817219734192,
"logits/chosen": 371.6880187988281,
"logits/rejected": 403.46356201171875,
"logps/chosen": -0.7223442196846008,
"logps/rejected": -1.1754904985427856,
"loss": 1.0787,
"nll_loss": 1.0912402868270874,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.03611721470952034,
"rewards/margins": 0.022657308727502823,
"rewards/rejected": -0.05877452343702316,
"step": 195
},
{
"epoch": 1.1904761904761905,
"grad_norm": 19.23330307006836,
"learning_rate": 3.5355339059327378e-06,
"log_odds_chosen": 1.305513620376587,
"log_odds_ratio": -0.32568031549453735,
"logits/chosen": 431.67156982421875,
"logits/rejected": 391.12017822265625,
"logps/chosen": -0.4990530014038086,
"logps/rejected": -1.1961959600448608,
"loss": 1.0519,
"nll_loss": 0.9312452077865601,
"rewards/accuracies": 0.8999999761581421,
"rewards/chosen": -0.02495265007019043,
"rewards/margins": 0.03485715016722679,
"rewards/rejected": -0.05980980396270752,
"step": 200
},
{
"epoch": 1.2202380952380953,
"grad_norm": 18.404104232788086,
"learning_rate": 3.4921514788478916e-06,
"log_odds_chosen": 1.2872587442398071,
"log_odds_ratio": -0.3010835349559784,
"logits/chosen": 335.2537841796875,
"logits/rejected": 387.68621826171875,
"logps/chosen": -0.6490974426269531,
"logps/rejected": -1.4184123277664185,
"loss": 1.0519,
"nll_loss": 1.0444796085357666,
"rewards/accuracies": 0.949999988079071,
"rewards/chosen": -0.032454878091812134,
"rewards/margins": 0.03846573829650879,
"rewards/rejected": -0.07092060893774033,
"step": 205
},
{
"epoch": 1.25,
"grad_norm": 13.957087516784668,
"learning_rate": 3.450327796711771e-06,
"log_odds_chosen": 1.1957916021347046,
"log_odds_ratio": -0.36826610565185547,
"logits/chosen": 365.1176452636719,
"logits/rejected": 363.3384094238281,
"logps/chosen": -0.6256347298622131,
"logps/rejected": -1.242398738861084,
"loss": 0.9696,
"nll_loss": 1.0164039134979248,
"rewards/accuracies": 0.8500000238418579,
"rewards/chosen": -0.031281743198633194,
"rewards/margins": 0.03083820268511772,
"rewards/rejected": -0.06211994215846062,
"step": 210
},
{
"epoch": 1.2797619047619047,
"grad_norm": 25.830501556396484,
"learning_rate": 3.409971697352368e-06,
"log_odds_chosen": 1.1722948551177979,
"log_odds_ratio": -0.3577140271663666,
"logits/chosen": 410.3714904785156,
"logits/rejected": 416.93621826171875,
"logps/chosen": -0.7465067505836487,
"logps/rejected": -1.4228246212005615,
"loss": 1.0589,
"nll_loss": 1.059061050415039,
"rewards/accuracies": 0.949999988079071,
"rewards/chosen": -0.037325333803892136,
"rewards/margins": 0.0338158942759037,
"rewards/rejected": -0.07114122807979584,
"step": 215
},
{
"epoch": 1.3095238095238095,
"grad_norm": 15.7648344039917,
"learning_rate": 3.3709993123162106e-06,
"log_odds_chosen": 0.9515246152877808,
"log_odds_ratio": -0.4343441128730774,
"logits/chosen": 352.2547607421875,
"logits/rejected": 370.08245849609375,
"logps/chosen": -0.6594604253768921,
"logps/rejected": -1.1094262599945068,
"loss": 1.0138,
"nll_loss": 0.9946004748344421,
"rewards/accuracies": 0.8500000238418579,
"rewards/chosen": -0.032973017543554306,
"rewards/margins": 0.022498302161693573,
"rewards/rejected": -0.05547132343053818,
"step": 220
},
{
"epoch": 1.3392857142857144,
"grad_norm": 21.596298217773438,
"learning_rate": 3.3333333333333333e-06,
"log_odds_chosen": 0.4564495086669922,
"log_odds_ratio": -0.5201828479766846,
"logits/chosen": 387.99542236328125,
"logits/rejected": 365.8666076660156,
"logps/chosen": -1.059605360031128,
"logps/rejected": -1.3452465534210205,
"loss": 1.0353,
"nll_loss": 1.2666301727294922,
"rewards/accuracies": 0.8500000238418579,
"rewards/chosen": -0.05298026651144028,
"rewards/margins": 0.014282059855759144,
"rewards/rejected": -0.06726232916116714,
"step": 225
},
{
"epoch": 1.369047619047619,
"grad_norm": 16.963186264038086,
"learning_rate": 3.296902366978936e-06,
"log_odds_chosen": 1.108178973197937,
"log_odds_ratio": -0.3512019217014313,
"logits/chosen": 355.7499084472656,
"logits/rejected": 366.8659362792969,
"logps/chosen": -0.6812053322792053,
"logps/rejected": -1.3409600257873535,
"loss": 1.0135,
"nll_loss": 0.873543381690979,
"rewards/accuracies": 0.8500000238418579,
"rewards/chosen": -0.034060269594192505,
"rewards/margins": 0.03298773616552353,
"rewards/rejected": -0.06704800575971603,
"step": 230
},
{
"epoch": 1.3988095238095237,
"grad_norm": 22.207305908203125,
"learning_rate": 3.2616403652672114e-06,
"log_odds_chosen": 1.4102222919464111,
"log_odds_ratio": -0.3304731547832489,
"logits/chosen": 382.0793151855469,
"logits/rejected": 389.06683349609375,
"logps/chosen": -0.5401080846786499,
"logps/rejected": -1.397993803024292,
"loss": 1.0737,
"nll_loss": 0.9223471879959106,
"rewards/accuracies": 0.8999999761581421,
"rewards/chosen": -0.027005404233932495,
"rewards/margins": 0.04289429262280464,
"rewards/rejected": -0.06989969313144684,
"step": 235
},
{
"epoch": 1.4285714285714286,
"grad_norm": 16.992250442504883,
"learning_rate": 3.2274861218395142e-06,
"log_odds_chosen": 0.807520866394043,
"log_odds_ratio": -0.4530153274536133,
"logits/chosen": 414.99261474609375,
"logits/rejected": 423.21392822265625,
"logps/chosen": -0.7907418012619019,
"logps/rejected": -1.2757242918014526,
"loss": 1.0285,
"nll_loss": 1.0256245136260986,
"rewards/accuracies": 0.8500000238418579,
"rewards/chosen": -0.03953709080815315,
"rewards/margins": 0.0242491252720356,
"rewards/rejected": -0.06378621608018875,
"step": 240
},
{
"epoch": 1.4583333333333333,
"grad_norm": 17.553632736206055,
"learning_rate": 3.1943828249997e-06,
"log_odds_chosen": 0.7604994177818298,
"log_odds_ratio": -0.42883071303367615,
"logits/chosen": 386.42657470703125,
"logits/rejected": 363.54931640625,
"logps/chosen": -0.6057090163230896,
"logps/rejected": -1.0264109373092651,
"loss": 1.0502,
"nll_loss": 1.1268291473388672,
"rewards/accuracies": 0.8999999761581421,
"rewards/chosen": -0.03028545156121254,
"rewards/margins": 0.021035097539424896,
"rewards/rejected": -0.05132054537534714,
"step": 245
},
{
"epoch": 1.4880952380952381,
"grad_norm": 22.123960494995117,
"learning_rate": 3.1622776601683796e-06,
"log_odds_chosen": 1.0444424152374268,
"log_odds_ratio": -0.38615161180496216,
"logits/chosen": 369.87274169921875,
"logits/rejected": 389.29583740234375,
"logps/chosen": -0.7160728573799133,
"logps/rejected": -1.2788223028182983,
"loss": 0.9816,
"nll_loss": 0.8135589361190796,
"rewards/accuracies": 0.8999999761581421,
"rewards/chosen": -0.035803645849227905,
"rewards/margins": 0.02813747525215149,
"rewards/rejected": -0.0639411136507988,
"step": 250
},
{
"epoch": 1.5178571428571428,
"grad_norm": 26.341325759887695,
"learning_rate": 3.131121455425748e-06,
"log_odds_chosen": 1.0527693033218384,
"log_odds_ratio": -0.35840553045272827,
"logits/chosen": 404.608154296875,
"logits/rejected": 410.81854248046875,
"logps/chosen": -0.6503961682319641,
"logps/rejected": -1.250218152999878,
"loss": 1.0367,
"nll_loss": 0.9888055920600891,
"rewards/accuracies": 0.8999999761581421,
"rewards/chosen": -0.032519809901714325,
"rewards/margins": 0.029991086572408676,
"rewards/rejected": -0.0625109076499939,
"step": 255
},
{
"epoch": 1.5476190476190477,
"grad_norm": 25.059885025024414,
"learning_rate": 3.1008683647302113e-06,
"log_odds_chosen": 1.0121057033538818,
"log_odds_ratio": -0.4098740220069885,
"logits/chosen": 375.2867126464844,
"logits/rejected": 426.01507568359375,
"logps/chosen": -0.8014397621154785,
"logps/rejected": -1.4673702716827393,
"loss": 1.0023,
"nll_loss": 1.092798113822937,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.0400719977915287,
"rewards/margins": 0.03329651802778244,
"rewards/rejected": -0.07336851954460144,
"step": 260
},
{
"epoch": 1.5773809523809523,
"grad_norm": 17.973594665527344,
"learning_rate": 3.0714755841697565e-06,
"log_odds_chosen": 0.830208957195282,
"log_odds_ratio": -0.5377876162528992,
"logits/chosen": 421.629150390625,
"logits/rejected": 414.95703125,
"logps/chosen": -0.8159686923027039,
"logps/rejected": -1.314146876335144,
"loss": 1.0967,
"nll_loss": 1.1257387399673462,
"rewards/accuracies": 0.6499999761581421,
"rewards/chosen": -0.040798433125019073,
"rewards/margins": 0.024908915162086487,
"rewards/rejected": -0.06570734083652496,
"step": 265
},
{
"epoch": 1.6071428571428572,
"grad_norm": 17.566476821899414,
"learning_rate": 3.0429030972509227e-06,
"log_odds_chosen": 0.7149374485015869,
"log_odds_ratio": -0.45594334602355957,
"logits/chosen": 354.78692626953125,
"logits/rejected": 364.1693115234375,
"logps/chosen": -0.836654543876648,
"logps/rejected": -1.2243211269378662,
"loss": 1.0429,
"nll_loss": 1.1530247926712036,
"rewards/accuracies": 0.949999988079071,
"rewards/chosen": -0.041832733899354935,
"rewards/margins": 0.019383331760764122,
"rewards/rejected": -0.06121605634689331,
"step": 270
},
{
"epoch": 1.6369047619047619,
"grad_norm": 16.325292587280273,
"learning_rate": 3.0151134457776365e-06,
"log_odds_chosen": 0.7199637293815613,
"log_odds_ratio": -0.434516578912735,
"logits/chosen": 322.65130615234375,
"logits/rejected": 338.8601989746094,
"logps/chosen": -0.5899932384490967,
"logps/rejected": -0.9299219250679016,
"loss": 1.076,
"nll_loss": 0.9019113779067993,
"rewards/accuracies": 0.8500000238418579,
"rewards/chosen": -0.029499661177396774,
"rewards/margins": 0.016996433958411217,
"rewards/rejected": -0.04649610072374344,
"step": 275
},
{
"epoch": 1.6666666666666665,
"grad_norm": 14.822625160217285,
"learning_rate": 2.988071523335984e-06,
"log_odds_chosen": 0.9259305000305176,
"log_odds_ratio": -0.5428971648216248,
"logits/chosen": 395.42230224609375,
"logits/rejected": 360.65234375,
"logps/chosen": -0.5504172444343567,
"logps/rejected": -1.1162601709365845,
"loss": 1.0232,
"nll_loss": 0.8841646909713745,
"rewards/accuracies": 0.8999999761581421,
"rewards/chosen": -0.027520865201950073,
"rewards/margins": 0.02829214558005333,
"rewards/rejected": -0.0558130145072937,
"step": 280
},
{
"epoch": 1.6964285714285714,
"grad_norm": 13.360857009887695,
"learning_rate": 2.961744388795462e-06,
"log_odds_chosen": 0.7002909183502197,
"log_odds_ratio": -0.46074408292770386,
"logits/chosen": 344.73712158203125,
"logits/rejected": 373.6070556640625,
"logps/chosen": -0.6551269292831421,
"logps/rejected": -1.0137001276016235,
"loss": 0.9926,
"nll_loss": 0.9025999307632446,
"rewards/accuracies": 0.8500000238418579,
"rewards/chosen": -0.032756343483924866,
"rewards/margins": 0.017928656190633774,
"rewards/rejected": -0.05068499967455864,
"step": 285
},
{
"epoch": 1.7261904761904763,
"grad_norm": 12.45747184753418,
"learning_rate": 2.9361010975735177e-06,
"log_odds_chosen": 1.1189637184143066,
"log_odds_ratio": -0.3354729413986206,
"logits/chosen": 367.6058349609375,
"logits/rejected": 436.7281188964844,
"logps/chosen": -0.7302691340446472,
"logps/rejected": -1.3499362468719482,
"loss": 1.0165,
"nll_loss": 0.9672770500183105,
"rewards/accuracies": 0.8999999761581421,
"rewards/chosen": -0.03651345893740654,
"rewards/margins": 0.03098336234688759,
"rewards/rejected": -0.06749682128429413,
"step": 290
},
{
"epoch": 1.755952380952381,
"grad_norm": 20.8387508392334,
"learning_rate": 2.9111125486979104e-06,
"log_odds_chosen": 0.8264445066452026,
"log_odds_ratio": -0.45267271995544434,
"logits/chosen": 380.7887268066406,
"logits/rejected": 392.38018798828125,
"logps/chosen": -0.7333472967147827,
"logps/rejected": -1.1829124689102173,
"loss": 1.0473,
"nll_loss": 1.2072994709014893,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.036667369306087494,
"rewards/margins": 0.0224782545119524,
"rewards/rejected": -0.059145621955394745,
"step": 295
},
{
"epoch": 1.7857142857142856,
"grad_norm": 14.306224822998047,
"learning_rate": 2.8867513459481293e-06,
"log_odds_chosen": 1.0914583206176758,
"log_odds_ratio": -0.3539409041404724,
"logits/chosen": 376.03033447265625,
"logits/rejected": 378.3951416015625,
"logps/chosen": -0.58739173412323,
"logps/rejected": -1.1421259641647339,
"loss": 0.975,
"nll_loss": 0.8565952181816101,
"rewards/accuracies": 0.8500000238418579,
"rewards/chosen": -0.02936958707869053,
"rewards/margins": 0.027736714109778404,
"rewards/rejected": -0.05710630491375923,
"step": 300
},
{
"epoch": 1.8154761904761905,
"grad_norm": 15.119916915893555,
"learning_rate": 2.862991671569341e-06,
"log_odds_chosen": 0.3856947720050812,
"log_odds_ratio": -0.5676103830337524,
"logits/chosen": 432.51239013671875,
"logits/rejected": 413.9574279785156,
"logps/chosen": -1.001999855041504,
"logps/rejected": -1.1941574811935425,
"loss": 1.026,
"nll_loss": 1.176692247390747,
"rewards/accuracies": 0.75,
"rewards/chosen": -0.050099991261959076,
"rewards/margins": 0.009607886895537376,
"rewards/rejected": -0.0597078800201416,
"step": 305
},
{
"epoch": 1.8452380952380953,
"grad_norm": 13.858372688293457,
"learning_rate": 2.839809171235324e-06,
"log_odds_chosen": 0.9851268529891968,
"log_odds_ratio": -0.46969637274742126,
"logits/chosen": 347.86224365234375,
"logits/rejected": 354.42236328125,
"logps/chosen": -0.7486366629600525,
"logps/rejected": -1.323521375656128,
"loss": 1.0778,
"nll_loss": 1.036675214767456,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.037431832402944565,
"rewards/margins": 0.028744244948029518,
"rewards/rejected": -0.06617607176303864,
"step": 310
},
{
"epoch": 1.875,
"grad_norm": 17.435771942138672,
"learning_rate": 2.817180849095055e-06,
"log_odds_chosen": 0.7195825576782227,
"log_odds_ratio": -0.5751198530197144,
"logits/chosen": 372.2769470214844,
"logits/rejected": 380.11322021484375,
"logps/chosen": -0.9793826937675476,
"logps/rejected": -1.4906439781188965,
"loss": 1.0867,
"nll_loss": 1.2049812078475952,
"rewards/accuracies": 0.6499999761581421,
"rewards/chosen": -0.04896913468837738,
"rewards/margins": 0.025563066825270653,
"rewards/rejected": -0.07453219592571259,
"step": 315
},
{
"epoch": 1.9047619047619047,
"grad_norm": 13.72044563293457,
"learning_rate": 2.7950849718747376e-06,
"log_odds_chosen": 1.036712884902954,
"log_odds_ratio": -0.36927738785743713,
"logits/chosen": 385.5801086425781,
"logits/rejected": 423.3285217285156,
"logps/chosen": -0.7053465247154236,
"logps/rejected": -1.304882526397705,
"loss": 0.9866,
"nll_loss": 0.9042797088623047,
"rewards/accuracies": 0.949999988079071,
"rewards/chosen": -0.03526733070611954,
"rewards/margins": 0.029976800084114075,
"rewards/rejected": -0.06524413079023361,
"step": 320
},
{
"epoch": 1.9345238095238095,
"grad_norm": 17.074472427368164,
"learning_rate": 2.773500981126146e-06,
"log_odds_chosen": 1.224524736404419,
"log_odds_ratio": -0.35412582755088806,
"logits/chosen": 397.96649169921875,
"logits/rejected": 423.78094482421875,
"logps/chosen": -0.6596122980117798,
"logps/rejected": -1.2944166660308838,
"loss": 0.9893,
"nll_loss": 0.9460498690605164,
"rewards/accuracies": 0.8500000238418579,
"rewards/chosen": -0.03298061341047287,
"rewards/margins": 0.0317402258515358,
"rewards/rejected": -0.06472083181142807,
"step": 325
},
{
"epoch": 1.9642857142857144,
"grad_norm": 23.940942764282227,
"learning_rate": 2.752409412815902e-06,
"log_odds_chosen": 1.0460882186889648,
"log_odds_ratio": -0.416635125875473,
"logits/chosen": 391.8972473144531,
"logits/rejected": 401.4590148925781,
"logps/chosen": -0.7760294675827026,
"logps/rejected": -1.4733283519744873,
"loss": 1.0094,
"nll_loss": 0.9179704785346985,
"rewards/accuracies": 0.8500000238418579,
"rewards/chosen": -0.038801468908786774,
"rewards/margins": 0.034864943474531174,
"rewards/rejected": -0.07366641610860825,
"step": 330
},
{
"epoch": 1.994047619047619,
"grad_norm": 18.142581939697266,
"learning_rate": 2.7317918235407652e-06,
"log_odds_chosen": 0.5594000816345215,
"log_odds_ratio": -0.544743537902832,
"logits/chosen": 377.28729248046875,
"logits/rejected": 394.92413330078125,
"logps/chosen": -0.9491284489631653,
"logps/rejected": -1.2867456674575806,
"loss": 1.0804,
"nll_loss": 1.2649915218353271,
"rewards/accuracies": 0.699999988079071,
"rewards/chosen": -0.047456420958042145,
"rewards/margins": 0.016880858689546585,
"rewards/rejected": -0.06433728337287903,
"step": 335
},
{
"epoch": 2.0,
"eval_log_odds_chosen": 0.21522486209869385,
"eval_log_odds_ratio": -0.6897445917129517,
"eval_logits/chosen": 330.50665283203125,
"eval_logits/rejected": 278.2473449707031,
"eval_logps/chosen": -1.022003173828125,
"eval_logps/rejected": -1.1829901933670044,
"eval_loss": 1.4225263595581055,
"eval_nll_loss": 1.4082757234573364,
"eval_rewards/accuracies": 0.5142857432365417,
"eval_rewards/chosen": -0.05110016465187073,
"eval_rewards/margins": 0.008049344643950462,
"eval_rewards/rejected": -0.05914951115846634,
"eval_runtime": 201.3887,
"eval_samples_per_second": 2.746,
"eval_steps_per_second": 0.348,
"step": 336
},
{
"epoch": 2.0238095238095237,
"grad_norm": 25.43450927734375,
"learning_rate": 2.711630722733202e-06,
"log_odds_chosen": 1.8818715810775757,
"log_odds_ratio": -0.2647668421268463,
"logits/chosen": 383.9913330078125,
"logits/rejected": 380.02801513671875,
"logps/chosen": -0.47935551404953003,
"logps/rejected": -1.454525351524353,
"loss": 0.6307,
"nll_loss": 0.6181924343109131,
"rewards/accuracies": 0.949999988079071,
"rewards/chosen": -0.023967772722244263,
"rewards/margins": 0.04875849559903145,
"rewards/rejected": -0.07272626459598541,
"step": 340
},
{
"epoch": 2.0535714285714284,
"grad_norm": 16.127397537231445,
"learning_rate": 2.691909510290828e-06,
"log_odds_chosen": 2.698786497116089,
"log_odds_ratio": -0.14153851568698883,
"logits/chosen": 388.04534912109375,
"logits/rejected": 355.0823669433594,
"logps/chosen": -0.29280781745910645,
"logps/rejected": -1.475531816482544,
"loss": 0.5566,
"nll_loss": 0.6942065954208374,
"rewards/accuracies": 0.8999999761581421,
"rewards/chosen": -0.014640390872955322,
"rewards/margins": 0.059136200696229935,
"rewards/rejected": -0.07377658784389496,
"step": 345
},
{
"epoch": 2.0833333333333335,
"grad_norm": 13.852057456970215,
"learning_rate": 2.6726124191242444e-06,
"log_odds_chosen": 2.8243956565856934,
"log_odds_ratio": -0.09990663826465607,
"logits/chosen": 354.74615478515625,
"logits/rejected": 387.60528564453125,
"logps/chosen": -0.37845298647880554,
"logps/rejected": -2.111022472381592,
"loss": 0.5438,
"nll_loss": 0.5637595057487488,
"rewards/accuracies": 1.0,
"rewards/chosen": -0.018922649323940277,
"rewards/margins": 0.08662847429513931,
"rewards/rejected": -0.10555113852024078,
"step": 350
},
{
"epoch": 2.113095238095238,
"grad_norm": 16.1456241607666,
"learning_rate": 2.6537244621713765e-06,
"log_odds_chosen": 2.1223108768463135,
"log_odds_ratio": -0.14685922861099243,
"logits/chosen": 359.0018005371094,
"logits/rejected": 385.0552062988281,
"logps/chosen": -0.425741970539093,
"logps/rejected": -1.6288859844207764,
"loss": 0.5366,
"nll_loss": 0.5897840261459351,
"rewards/accuracies": 1.0,
"rewards/chosen": -0.0212871003895998,
"rewards/margins": 0.060157209634780884,
"rewards/rejected": -0.08144429326057434,
"step": 355
},
{
"epoch": 2.142857142857143,
"grad_norm": 11.441978454589844,
"learning_rate": 2.6352313834736496e-06,
"log_odds_chosen": 2.8245463371276855,
"log_odds_ratio": -0.12314938008785248,
"logits/chosen": 369.9977722167969,
"logits/rejected": 407.8392333984375,
"logps/chosen": -0.3003098964691162,
"logps/rejected": -1.5362733602523804,
"loss": 0.5329,
"nll_loss": 0.5958473086357117,
"rewards/accuracies": 1.0,
"rewards/chosen": -0.015015493147075176,
"rewards/margins": 0.06179817393422127,
"rewards/rejected": -0.07681366801261902,
"step": 360
},
{
"epoch": 2.1726190476190474,
"grad_norm": 12.86784839630127,
"learning_rate": 2.6171196129510684e-06,
"log_odds_chosen": 2.503121852874756,
"log_odds_ratio": -0.1363641321659088,
"logits/chosen": 334.23468017578125,
"logits/rejected": 311.822509765625,
"logps/chosen": -0.3001948595046997,
"logps/rejected": -1.4599236249923706,
"loss": 0.5156,
"nll_loss": 0.5041288137435913,
"rewards/accuracies": 1.0,
"rewards/chosen": -0.0150097431614995,
"rewards/margins": 0.05798644572496414,
"rewards/rejected": -0.07299618422985077,
"step": 365
},
{
"epoch": 2.2023809523809526,
"grad_norm": 13.525810241699219,
"learning_rate": 2.599376224550182e-06,
"log_odds_chosen": 1.8203538656234741,
"log_odds_ratio": -0.19545204937458038,
"logits/chosen": 313.6293029785156,
"logits/rejected": 326.6126403808594,
"logps/chosen": -0.4296380877494812,
"logps/rejected": -1.4019193649291992,
"loss": 0.5492,
"nll_loss": 0.6865583062171936,
"rewards/accuracies": 0.949999988079071,
"rewards/chosen": -0.02148190513253212,
"rewards/margins": 0.04861406981945038,
"rewards/rejected": -0.0700959712266922,
"step": 370
},
{
"epoch": 2.232142857142857,
"grad_norm": 14.674184799194336,
"learning_rate": 2.5819888974716113e-06,
"log_odds_chosen": 1.8546769618988037,
"log_odds_ratio": -0.22947852313518524,
"logits/chosen": 377.0060119628906,
"logits/rejected": 399.48590087890625,
"logps/chosen": -0.3884517252445221,
"logps/rejected": -1.292641520500183,
"loss": 0.5785,
"nll_loss": 0.5873233079910278,
"rewards/accuracies": 0.8999999761581421,
"rewards/chosen": -0.019422587007284164,
"rewards/margins": 0.045209504663944244,
"rewards/rejected": -0.06463208049535751,
"step": 375
},
{
"epoch": 2.261904761904762,
"grad_norm": 11.674169540405273,
"learning_rate": 2.564945880212886e-06,
"log_odds_chosen": 2.2116830348968506,
"log_odds_ratio": -0.12257333844900131,
"logits/chosen": 367.1352233886719,
"logits/rejected": 336.3498840332031,
"logps/chosen": -0.3219259977340698,
"logps/rejected": -1.350303292274475,
"loss": 0.5297,
"nll_loss": 0.5648508071899414,
"rewards/accuracies": 1.0,
"rewards/chosen": -0.01609629951417446,
"rewards/margins": 0.051418863236904144,
"rewards/rejected": -0.06751517206430435,
"step": 380
},
{
"epoch": 2.2916666666666665,
"grad_norm": 13.154436111450195,
"learning_rate": 2.5482359571881276e-06,
"log_odds_chosen": 2.5731985569000244,
"log_odds_ratio": -0.1179199069738388,
"logits/chosen": 340.62567138671875,
"logits/rejected": 364.66888427734375,
"logps/chosen": -0.27915579080581665,
"logps/rejected": -1.4784090518951416,
"loss": 0.5102,
"nll_loss": 0.49411922693252563,
"rewards/accuracies": 0.949999988079071,
"rewards/chosen": -0.013957786373794079,
"rewards/margins": 0.059962667524814606,
"rewards/rejected": -0.07392045855522156,
"step": 385
},
{
"epoch": 2.3214285714285716,
"grad_norm": 11.973630905151367,
"learning_rate": 2.5318484177091667e-06,
"log_odds_chosen": 2.2964727878570557,
"log_odds_ratio": -0.13063772022724152,
"logits/chosen": 359.48345947265625,
"logits/rejected": 400.7521667480469,
"logps/chosen": -0.3307945430278778,
"logps/rejected": -1.4601058959960938,
"loss": 0.5685,
"nll_loss": 0.5173304677009583,
"rewards/accuracies": 1.0,
"rewards/chosen": -0.01653972640633583,
"rewards/margins": 0.056465573608875275,
"rewards/rejected": -0.0730053037405014,
"step": 390
},
{
"epoch": 2.3511904761904763,
"grad_norm": 10.832448959350586,
"learning_rate": 2.515773027133138e-06,
"log_odds_chosen": 2.685337543487549,
"log_odds_ratio": -0.10510516166687012,
"logits/chosen": 360.8955383300781,
"logits/rejected": 359.0787048339844,
"logps/chosen": -0.2764904201030731,
"logps/rejected": -1.409620761871338,
"loss": 0.511,
"nll_loss": 0.500057578086853,
"rewards/accuracies": 1.0,
"rewards/chosen": -0.013824522495269775,
"rewards/margins": 0.05665650963783264,
"rewards/rejected": -0.07048103958368301,
"step": 395
},
{
"epoch": 2.380952380952381,
"grad_norm": 12.639362335205078,
"learning_rate": 2.5e-06,
"log_odds_chosen": 2.6170766353607178,
"log_odds_ratio": -0.1315409243106842,
"logits/chosen": 342.3817138671875,
"logits/rejected": 396.52410888671875,
"logps/chosen": -0.328637957572937,
"logps/rejected": -1.6732807159423828,
"loss": 0.527,
"nll_loss": 0.5181549787521362,
"rewards/accuracies": 0.949999988079071,
"rewards/chosen": -0.01643189787864685,
"rewards/margins": 0.067232146859169,
"rewards/rejected": -0.08366404473781586,
"step": 400
},
{
"epoch": 2.4107142857142856,
"grad_norm": 14.651439666748047,
"learning_rate": 2.484519974999767e-06,
"log_odds_chosen": 2.2201273441314697,
"log_odds_ratio": -0.16681890189647675,
"logits/chosen": 424.72918701171875,
"logits/rejected": 371.58966064453125,
"logps/chosen": -0.30684176087379456,
"logps/rejected": -1.3665101528167725,
"loss": 0.5749,
"nll_loss": 0.5082268714904785,
"rewards/accuracies": 1.0,
"rewards/chosen": -0.015342088416218758,
"rewards/margins": 0.05298342555761337,
"rewards/rejected": -0.06832550466060638,
"step": 405
},
{
"epoch": 2.4404761904761907,
"grad_norm": 11.325146675109863,
"learning_rate": 2.4693239916239746e-06,
"log_odds_chosen": 2.499159336090088,
"log_odds_ratio": -0.21561208367347717,
"logits/chosen": 356.08258056640625,
"logits/rejected": 390.337890625,
"logps/chosen": -0.372717022895813,
"logps/rejected": -1.7423985004425049,
"loss": 0.5384,
"nll_loss": 0.5256937742233276,
"rewards/accuracies": 0.949999988079071,
"rewards/chosen": -0.01863585226237774,
"rewards/margins": 0.06848406046628952,
"rewards/rejected": -0.087119922041893,
"step": 410
},
{
"epoch": 2.4702380952380953,
"grad_norm": 12.71360969543457,
"learning_rate": 2.4544034683690802e-06,
"log_odds_chosen": 2.8460097312927246,
"log_odds_ratio": -0.09599516540765762,
"logits/chosen": 386.35302734375,
"logits/rejected": 372.95367431640625,
"logps/chosen": -0.26167118549346924,
"logps/rejected": -1.7513694763183594,
"loss": 0.551,
"nll_loss": 0.4728317856788635,
"rewards/accuracies": 1.0,
"rewards/chosen": -0.013083559460937977,
"rewards/margins": 0.0744849145412445,
"rewards/rejected": -0.08756847679615021,
"step": 415
},
{
"epoch": 2.5,
"grad_norm": 19.943899154663086,
"learning_rate": 2.4397501823713327e-06,
"log_odds_chosen": 2.2042346000671387,
"log_odds_ratio": -0.17198964953422546,
"logits/chosen": 370.604736328125,
"logits/rejected": 320.15875244140625,
"logps/chosen": -0.30728116631507874,
"logps/rejected": -1.3875327110290527,
"loss": 0.5247,
"nll_loss": 0.6666563153266907,
"rewards/accuracies": 0.949999988079071,
"rewards/chosen": -0.015364060178399086,
"rewards/margins": 0.05401257798075676,
"rewards/rejected": -0.0693766325712204,
"step": 420
},
{
"epoch": 2.5297619047619047,
"grad_norm": 12.133928298950195,
"learning_rate": 2.4253562503633297e-06,
"log_odds_chosen": 2.90407133102417,
"log_odds_ratio": -0.06831072270870209,
"logits/chosen": 342.95733642578125,
"logits/rejected": 367.10894775390625,
"logps/chosen": -0.3182448446750641,
"logps/rejected": -1.8485863208770752,
"loss": 0.507,
"nll_loss": 0.6053184270858765,
"rewards/accuracies": 1.0,
"rewards/chosen": -0.015912240371108055,
"rewards/margins": 0.07651706039905548,
"rewards/rejected": -0.09242931008338928,
"step": 425
},
{
"epoch": 2.5595238095238093,
"grad_norm": 12.26125717163086,
"learning_rate": 2.411214110852061e-06,
"log_odds_chosen": 2.55293607711792,
"log_odds_ratio": -0.10798110067844391,
"logits/chosen": 391.63848876953125,
"logits/rejected": 404.0960388183594,
"logps/chosen": -0.27310800552368164,
"logps/rejected": -1.4932405948638916,
"loss": 0.5084,
"nll_loss": 0.4452734887599945,
"rewards/accuracies": 1.0,
"rewards/chosen": -0.013655401766300201,
"rewards/margins": 0.06100662797689438,
"rewards/rejected": -0.07466202974319458,
"step": 430
},
{
"epoch": 2.5892857142857144,
"grad_norm": 13.624907493591309,
"learning_rate": 2.3973165074269213e-06,
"log_odds_chosen": 2.3622078895568848,
"log_odds_ratio": -0.2018093764781952,
"logits/chosen": 356.6548767089844,
"logits/rejected": 360.92474365234375,
"logps/chosen": -0.36742302775382996,
"logps/rejected": -1.6279608011245728,
"loss": 0.5587,
"nll_loss": 0.5465327501296997,
"rewards/accuracies": 0.949999988079071,
"rewards/chosen": -0.018371151760220528,
"rewards/margins": 0.06302689015865326,
"rewards/rejected": -0.08139804750680923,
"step": 435
},
{
"epoch": 2.619047619047619,
"grad_norm": 13.956061363220215,
"learning_rate": 2.3836564731139807e-06,
"log_odds_chosen": 2.7665021419525146,
"log_odds_ratio": -0.10336035490036011,
"logits/chosen": 352.8879089355469,
"logits/rejected": 368.2629699707031,
"logps/chosen": -0.2728215754032135,
"logps/rejected": -1.5103857517242432,
"loss": 0.5462,
"nll_loss": 0.46932536363601685,
"rewards/accuracies": 1.0,
"rewards/chosen": -0.01364107709378004,
"rewards/margins": 0.06187821552157402,
"rewards/rejected": -0.07551928609609604,
"step": 440
},
{
"epoch": 2.6488095238095237,
"grad_norm": 11.78285026550293,
"learning_rate": 2.3702273156998867e-06,
"log_odds_chosen": 2.3770580291748047,
"log_odds_ratio": -0.116610586643219,
"logits/chosen": 337.458740234375,
"logits/rejected": 374.8058776855469,
"logps/chosen": -0.3450239896774292,
"logps/rejected": -1.6122970581054688,
"loss": 0.5447,
"nll_loss": 0.5509533882141113,
"rewards/accuracies": 1.0,
"rewards/chosen": -0.01725119911134243,
"rewards/margins": 0.06336364895105362,
"rewards/rejected": -0.0806148499250412,
"step": 445
},
{
"epoch": 2.678571428571429,
"grad_norm": 14.717430114746094,
"learning_rate": 2.357022603955159e-06,
"log_odds_chosen": 2.9325308799743652,
"log_odds_ratio": -0.08914806693792343,
"logits/chosen": 375.6985778808594,
"logits/rejected": 367.75787353515625,
"logps/chosen": -0.35901179909706116,
"logps/rejected": -2.006959915161133,
"loss": 0.5752,
"nll_loss": 0.49780726432800293,
"rewards/accuracies": 1.0,
"rewards/chosen": -0.01795058883726597,
"rewards/margins": 0.08239741623401642,
"rewards/rejected": -0.10034799575805664,
"step": 450
},
{
"epoch": 2.7083333333333335,
"grad_norm": 18.606168746948242,
"learning_rate": 2.3440361546924774e-06,
"log_odds_chosen": 2.5973100662231445,
"log_odds_ratio": -0.09789351373910904,
"logits/chosen": 371.6011962890625,
"logits/rejected": 356.3915710449219,
"logps/chosen": -0.27857938408851624,
"logps/rejected": -1.5046110153198242,
"loss": 0.5909,
"nll_loss": 0.5146271586418152,
"rewards/accuracies": 1.0,
"rewards/chosen": -0.013928967528045177,
"rewards/margins": 0.061301589012145996,
"rewards/rejected": -0.07523055374622345,
"step": 455
},
{
"epoch": 2.738095238095238,
"grad_norm": 14.296780586242676,
"learning_rate": 2.3312620206007847e-06,
"log_odds_chosen": 2.4838976860046387,
"log_odds_ratio": -0.13478006422519684,
"logits/chosen": 412.72332763671875,
"logits/rejected": 433.41070556640625,
"logps/chosen": -0.3284691870212555,
"logps/rejected": -1.5628798007965088,
"loss": 0.5718,
"nll_loss": 0.5497859716415405,
"rewards/accuracies": 1.0,
"rewards/chosen": -0.016423460096120834,
"rewards/margins": 0.061720531433820724,
"rewards/rejected": -0.07814399152994156,
"step": 460
},
{
"epoch": 2.767857142857143,
"grad_norm": 12.61571216583252,
"learning_rate": 2.3186944788008413e-06,
"log_odds_chosen": 2.7270541191101074,
"log_odds_ratio": -0.11491680145263672,
"logits/chosen": 401.2771301269531,
"logits/rejected": 386.7737121582031,
"logps/chosen": -0.25855037569999695,
"logps/rejected": -1.5313479900360107,
"loss": 0.5771,
"nll_loss": 0.6066937446594238,
"rewards/accuracies": 1.0,
"rewards/chosen": -0.012927519157528877,
"rewards/margins": 0.06363988667726517,
"rewards/rejected": -0.0765674039721489,
"step": 465
},
{
"epoch": 2.7976190476190474,
"grad_norm": 11.338821411132812,
"learning_rate": 2.3063280200722128e-06,
"log_odds_chosen": 2.1205978393554688,
"log_odds_ratio": -0.20822449028491974,
"logits/chosen": 357.4671630859375,
"logits/rejected": 335.1904296875,
"logps/chosen": -0.36784037947654724,
"logps/rejected": -1.465673804283142,
"loss": 0.5175,
"nll_loss": 0.5160819888114929,
"rewards/accuracies": 0.8999999761581421,
"rewards/chosen": -0.018392018973827362,
"rewards/margins": 0.0548916831612587,
"rewards/rejected": -0.07328370213508606,
"step": 470
},
{
"epoch": 2.8273809523809526,
"grad_norm": 11.929247856140137,
"learning_rate": 2.2941573387056174e-06,
"log_odds_chosen": 2.6770148277282715,
"log_odds_ratio": -0.0899493470788002,
"logits/chosen": 325.32476806640625,
"logits/rejected": 420.31280517578125,
"logps/chosen": -0.3526560664176941,
"logps/rejected": -1.848586082458496,
"loss": 0.5393,
"nll_loss": 0.48061132431030273,
"rewards/accuracies": 1.0,
"rewards/chosen": -0.017632806673645973,
"rewards/margins": 0.07479649782180786,
"rewards/rejected": -0.09242929518222809,
"step": 475
},
{
"epoch": 2.857142857142857,
"grad_norm": 13.174921035766602,
"learning_rate": 2.2821773229381924e-06,
"log_odds_chosen": 2.5945560932159424,
"log_odds_ratio": -0.09449335187673569,
"logits/chosen": 369.5697326660156,
"logits/rejected": 390.010009765625,
"logps/chosen": -0.26720184087753296,
"logps/rejected": -1.5164011716842651,
"loss": 0.4922,
"nll_loss": 0.5011499524116516,
"rewards/accuracies": 1.0,
"rewards/chosen": -0.013360092416405678,
"rewards/margins": 0.06245996803045273,
"rewards/rejected": -0.07582006603479385,
"step": 480
},
{
"epoch": 2.886904761904762,
"grad_norm": 17.229890823364258,
"learning_rate": 2.270383045932499e-06,
"log_odds_chosen": 2.310241222381592,
"log_odds_ratio": -0.1323787122964859,
"logits/chosen": 352.09368896484375,
"logits/rejected": 390.89129638671875,
"logps/chosen": -0.41382861137390137,
"logps/rejected": -1.641650915145874,
"loss": 0.5059,
"nll_loss": 0.5348228216171265,
"rewards/accuracies": 1.0,
"rewards/chosen": -0.020691432058811188,
"rewards/margins": 0.06139112263917923,
"rewards/rejected": -0.08208255469799042,
"step": 485
},
{
"epoch": 2.9166666666666665,
"grad_norm": 11.300827026367188,
"learning_rate": 2.2587697572631284e-06,
"log_odds_chosen": 1.7844845056533813,
"log_odds_ratio": -0.32421550154685974,
"logits/chosen": 398.19256591796875,
"logits/rejected": 312.98248291015625,
"logps/chosen": -0.5920530557632446,
"logps/rejected": -1.449924349784851,
"loss": 0.6026,
"nll_loss": 0.5389891862869263,
"rewards/accuracies": 0.8999999761581421,
"rewards/chosen": -0.0296026524156332,
"rewards/margins": 0.04289356991648674,
"rewards/rejected": -0.07249622046947479,
"step": 490
},
{
"epoch": 2.946428571428571,
"grad_norm": 13.277922630310059,
"learning_rate": 2.2473328748774737e-06,
"log_odds_chosen": 2.5535902976989746,
"log_odds_ratio": -0.12692758440971375,
"logits/chosen": 344.2357177734375,
"logits/rejected": 398.01348876953125,
"logps/chosen": -0.27592435479164124,
"logps/rejected": -1.5041601657867432,
"loss": 0.5166,
"nll_loss": 0.4667043685913086,
"rewards/accuracies": 1.0,
"rewards/chosen": -0.013796217739582062,
"rewards/margins": 0.06141179800033569,
"rewards/rejected": -0.07520802319049835,
"step": 495
},
{
"epoch": 2.9761904761904763,
"grad_norm": 10.682828903198242,
"learning_rate": 2.23606797749979e-06,
"log_odds_chosen": 2.8015060424804688,
"log_odds_ratio": -0.12307514250278473,
"logits/chosen": 391.4792785644531,
"logits/rejected": 362.94403076171875,
"logps/chosen": -0.2695181965827942,
"logps/rejected": -1.6667454242706299,
"loss": 0.5651,
"nll_loss": 0.40066272020339966,
"rewards/accuracies": 1.0,
"rewards/chosen": -0.01347590796649456,
"rewards/margins": 0.06986136734485626,
"rewards/rejected": -0.08333728462457657,
"step": 500
},
{
"epoch": 3.0,
"eval_log_odds_chosen": 0.3315908908843994,
"eval_log_odds_ratio": -0.6744564771652222,
"eval_logits/chosen": 304.2648620605469,
"eval_logits/rejected": 249.09341430664062,
"eval_logps/chosen": -1.1561349630355835,
"eval_logps/rejected": -1.3795461654663086,
"eval_loss": 1.6016675233840942,
"eval_nll_loss": 1.5642986297607422,
"eval_rewards/accuracies": 0.5714285969734192,
"eval_rewards/chosen": -0.057806748896837234,
"eval_rewards/margins": 0.011170565150678158,
"eval_rewards/rejected": -0.06897731870412827,
"eval_runtime": 201.3868,
"eval_samples_per_second": 2.746,
"eval_steps_per_second": 0.348,
"step": 504
},
{
"epoch": 3.0,
"step": 504,
"total_flos": 0.0,
"train_loss": 1.4336541661667446,
"train_runtime": 14833.1355,
"train_samples_per_second": 1.085,
"train_steps_per_second": 0.034
}
],
"logging_steps": 5,
"max_steps": 504,
"num_input_tokens_seen": 0,
"num_train_epochs": 3,
"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": 2,
"trial_name": null,
"trial_params": null
}