llama-7b-sft-DPO / trainer_state.json
AmberYifan's picture
Model save
5c51859 verified
{
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 1.0,
"eval_steps": 100,
"global_step": 1751,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.0,
"grad_norm": 2.70925584855601,
"learning_rate": 2.8409090909090907e-09,
"logits/chosen": -2.915217876434326,
"logits/rejected": -2.758937358856201,
"logps/chosen": -1027.0467529296875,
"logps/rejected": -801.54931640625,
"loss": 0.6931,
"rewards/accuracies": 0.0,
"rewards/chosen": 0.0,
"rewards/margins": 0.0,
"rewards/rejected": 0.0,
"step": 1
},
{
"epoch": 0.01,
"grad_norm": 2.305868983825744,
"learning_rate": 2.8409090909090908e-08,
"logits/chosen": -2.853972911834717,
"logits/rejected": -2.8207626342773438,
"logps/chosen": -688.71875,
"logps/rejected": -695.2182006835938,
"loss": 0.6932,
"rewards/accuracies": 0.4444444477558136,
"rewards/chosen": -8.214871922973543e-05,
"rewards/margins": 0.00015560892643406987,
"rewards/rejected": -0.00023775763111189008,
"step": 10
},
{
"epoch": 0.01,
"grad_norm": 2.40743980345329,
"learning_rate": 5.6818181818181815e-08,
"logits/chosen": -3.022716999053955,
"logits/rejected": -2.9431874752044678,
"logps/chosen": -762.6065063476562,
"logps/rejected": -704.7117919921875,
"loss": 0.6931,
"rewards/accuracies": 0.48124998807907104,
"rewards/chosen": -6.342886626953259e-05,
"rewards/margins": 0.00014013503096066415,
"rewards/rejected": -0.0002035639190580696,
"step": 20
},
{
"epoch": 0.02,
"grad_norm": 2.104625945949737,
"learning_rate": 8.522727272727271e-08,
"logits/chosen": -2.8718769550323486,
"logits/rejected": -2.782773494720459,
"logps/chosen": -806.1029052734375,
"logps/rejected": -715.8812255859375,
"loss": 0.6933,
"rewards/accuracies": 0.44999998807907104,
"rewards/chosen": 0.0003297650837339461,
"rewards/margins": -0.00031846793717704713,
"rewards/rejected": 0.0006482329918071628,
"step": 30
},
{
"epoch": 0.02,
"grad_norm": 2.55572189108705,
"learning_rate": 1.1363636363636363e-07,
"logits/chosen": -2.820253849029541,
"logits/rejected": -2.7977638244628906,
"logps/chosen": -628.6044921875,
"logps/rejected": -596.6134643554688,
"loss": 0.6929,
"rewards/accuracies": 0.5687500238418579,
"rewards/chosen": 0.0016179379308596253,
"rewards/margins": 0.0007343190954998136,
"rewards/rejected": 0.0008836188353598118,
"step": 40
},
{
"epoch": 0.03,
"grad_norm": 2.231257874310367,
"learning_rate": 1.4204545454545455e-07,
"logits/chosen": -2.891606569290161,
"logits/rejected": -2.8657639026641846,
"logps/chosen": -737.377197265625,
"logps/rejected": -629.8726806640625,
"loss": 0.693,
"rewards/accuracies": 0.518750011920929,
"rewards/chosen": 0.00467439740896225,
"rewards/margins": 0.0006132819689810276,
"rewards/rejected": 0.004061114974319935,
"step": 50
},
{
"epoch": 0.03,
"grad_norm": 2.205514048621008,
"learning_rate": 1.7045454545454543e-07,
"logits/chosen": -2.945676326751709,
"logits/rejected": -2.934192419052124,
"logps/chosen": -706.3095703125,
"logps/rejected": -644.1196899414062,
"loss": 0.6926,
"rewards/accuracies": 0.5,
"rewards/chosen": 0.008496693335473537,
"rewards/margins": 0.0008374860626645386,
"rewards/rejected": 0.007659205701202154,
"step": 60
},
{
"epoch": 0.04,
"grad_norm": 2.011345495557958,
"learning_rate": 1.9886363636363636e-07,
"logits/chosen": -2.988844394683838,
"logits/rejected": -2.8398101329803467,
"logps/chosen": -726.2923583984375,
"logps/rejected": -645.0869140625,
"loss": 0.6917,
"rewards/accuracies": 0.581250011920929,
"rewards/chosen": 0.01823153905570507,
"rewards/margins": 0.0032039093784987926,
"rewards/rejected": 0.01502762921154499,
"step": 70
},
{
"epoch": 0.05,
"grad_norm": 2.8344720832998354,
"learning_rate": 2.2727272727272726e-07,
"logits/chosen": -2.8927783966064453,
"logits/rejected": -2.8410775661468506,
"logps/chosen": -790.5596923828125,
"logps/rejected": -713.123046875,
"loss": 0.6912,
"rewards/accuracies": 0.5625,
"rewards/chosen": 0.03424953669309616,
"rewards/margins": 0.00334340101107955,
"rewards/rejected": 0.030906137079000473,
"step": 80
},
{
"epoch": 0.05,
"grad_norm": 2.005554229830089,
"learning_rate": 2.5568181818181816e-07,
"logits/chosen": -2.9193904399871826,
"logits/rejected": -2.8237431049346924,
"logps/chosen": -765.6085815429688,
"logps/rejected": -687.0435791015625,
"loss": 0.689,
"rewards/accuracies": 0.59375,
"rewards/chosen": 0.052040331065654755,
"rewards/margins": 0.007134866900742054,
"rewards/rejected": 0.044905465096235275,
"step": 90
},
{
"epoch": 0.06,
"grad_norm": 2.7536228949783403,
"learning_rate": 2.840909090909091e-07,
"logits/chosen": -2.7679712772369385,
"logits/rejected": -2.750290870666504,
"logps/chosen": -661.1817626953125,
"logps/rejected": -607.760986328125,
"loss": 0.6884,
"rewards/accuracies": 0.5375000238418579,
"rewards/chosen": 0.06972592324018478,
"rewards/margins": 0.008985200896859169,
"rewards/rejected": 0.06074073165655136,
"step": 100
},
{
"epoch": 0.06,
"eval_logits/chosen": -2.8753762245178223,
"eval_logits/rejected": -2.778611421585083,
"eval_logps/chosen": -714.9249877929688,
"eval_logps/rejected": -645.173095703125,
"eval_loss": 0.6886128187179565,
"eval_rewards/accuracies": 0.5647435784339905,
"eval_rewards/chosen": 0.08787301182746887,
"eval_rewards/margins": 0.010522950440645218,
"eval_rewards/rejected": 0.07735005766153336,
"eval_runtime": 837.7219,
"eval_samples_per_second": 14.863,
"eval_steps_per_second": 0.466,
"step": 100
},
{
"epoch": 0.06,
"grad_norm": 1.989207588161087,
"learning_rate": 3.1249999999999997e-07,
"logits/chosen": -2.879096508026123,
"logits/rejected": -2.6906533241271973,
"logps/chosen": -679.3154296875,
"logps/rejected": -712.8575439453125,
"loss": 0.688,
"rewards/accuracies": 0.5062500238418579,
"rewards/chosen": 0.0985824465751648,
"rewards/margins": 0.0015855018282309175,
"rewards/rejected": 0.09699694812297821,
"step": 110
},
{
"epoch": 0.07,
"grad_norm": 2.261619042072944,
"learning_rate": 3.4090909090909085e-07,
"logits/chosen": -2.7942328453063965,
"logits/rejected": -2.7030820846557617,
"logps/chosen": -665.7606201171875,
"logps/rejected": -582.5177612304688,
"loss": 0.6874,
"rewards/accuracies": 0.5562499761581421,
"rewards/chosen": 0.11793726682662964,
"rewards/margins": 0.015646493062376976,
"rewards/rejected": 0.10229077190160751,
"step": 120
},
{
"epoch": 0.07,
"grad_norm": 2.270843600939803,
"learning_rate": 3.693181818181818e-07,
"logits/chosen": -2.782252788543701,
"logits/rejected": -2.760941743850708,
"logps/chosen": -715.648193359375,
"logps/rejected": -647.896484375,
"loss": 0.6894,
"rewards/accuracies": 0.606249988079071,
"rewards/chosen": 0.14008425176143646,
"rewards/margins": 0.015175814740359783,
"rewards/rejected": 0.124908447265625,
"step": 130
},
{
"epoch": 0.08,
"grad_norm": 2.2952967047901565,
"learning_rate": 3.977272727272727e-07,
"logits/chosen": -2.753657817840576,
"logits/rejected": -2.718437910079956,
"logps/chosen": -732.1031494140625,
"logps/rejected": -675.71435546875,
"loss": 0.6835,
"rewards/accuracies": 0.543749988079071,
"rewards/chosen": 0.15980581939220428,
"rewards/margins": 0.014441452920436859,
"rewards/rejected": 0.14536437392234802,
"step": 140
},
{
"epoch": 0.09,
"grad_norm": 2.0745365771074096,
"learning_rate": 4.2613636363636364e-07,
"logits/chosen": -2.818753480911255,
"logits/rejected": -2.866856098175049,
"logps/chosen": -700.4889526367188,
"logps/rejected": -673.15234375,
"loss": 0.6822,
"rewards/accuracies": 0.543749988079071,
"rewards/chosen": 0.1768052875995636,
"rewards/margins": 0.01371350884437561,
"rewards/rejected": 0.163091778755188,
"step": 150
},
{
"epoch": 0.09,
"grad_norm": 2.3049728868710417,
"learning_rate": 4.545454545454545e-07,
"logits/chosen": -2.6952528953552246,
"logits/rejected": -2.6161999702453613,
"logps/chosen": -756.7824096679688,
"logps/rejected": -669.5311279296875,
"loss": 0.6842,
"rewards/accuracies": 0.5375000238418579,
"rewards/chosen": 0.22233958542346954,
"rewards/margins": 0.03233207389712334,
"rewards/rejected": 0.1900075227022171,
"step": 160
},
{
"epoch": 0.1,
"grad_norm": 2.503833153793772,
"learning_rate": 4.829545454545455e-07,
"logits/chosen": -2.7758500576019287,
"logits/rejected": -2.693150281906128,
"logps/chosen": -721.1317138671875,
"logps/rejected": -693.0599365234375,
"loss": 0.6912,
"rewards/accuracies": 0.5874999761581421,
"rewards/chosen": 0.2289244681596756,
"rewards/margins": 0.01609339565038681,
"rewards/rejected": 0.212831050157547,
"step": 170
},
{
"epoch": 0.1,
"grad_norm": 1.963645529752969,
"learning_rate": 4.999920426892062e-07,
"logits/chosen": -2.728468418121338,
"logits/rejected": -2.7938730716705322,
"logps/chosen": -697.4301147460938,
"logps/rejected": -639.5413818359375,
"loss": 0.6829,
"rewards/accuracies": 0.543749988079071,
"rewards/chosen": 0.22587458789348602,
"rewards/margins": 0.011290490627288818,
"rewards/rejected": 0.214584082365036,
"step": 180
},
{
"epoch": 0.11,
"grad_norm": 1.9105830549467013,
"learning_rate": 4.999025287600885e-07,
"logits/chosen": -2.569119453430176,
"logits/rejected": -2.600816011428833,
"logps/chosen": -764.7642211914062,
"logps/rejected": -736.1318359375,
"loss": 0.6825,
"rewards/accuracies": 0.5249999761581421,
"rewards/chosen": 0.24947170913219452,
"rewards/margins": 0.015052946284413338,
"rewards/rejected": 0.23441874980926514,
"step": 190
},
{
"epoch": 0.11,
"grad_norm": 2.1951075055673153,
"learning_rate": 4.997135899956001e-07,
"logits/chosen": -2.7701027393341064,
"logits/rejected": -2.6743502616882324,
"logps/chosen": -697.3702392578125,
"logps/rejected": -623.8272705078125,
"loss": 0.6769,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": 0.2433600127696991,
"rewards/margins": 0.034381039440631866,
"rewards/rejected": 0.20897898077964783,
"step": 200
},
{
"epoch": 0.11,
"eval_logits/chosen": -2.6971323490142822,
"eval_logits/rejected": -2.6093711853027344,
"eval_logps/chosen": -698.2555541992188,
"eval_logps/rejected": -630.9727783203125,
"eval_loss": 0.6808642745018005,
"eval_rewards/accuracies": 0.5746794939041138,
"eval_rewards/chosen": 0.25456854701042175,
"eval_rewards/margins": 0.03521544486284256,
"eval_rewards/rejected": 0.2193530946969986,
"eval_runtime": 833.7444,
"eval_samples_per_second": 14.934,
"eval_steps_per_second": 0.468,
"step": 200
},
{
"epoch": 0.12,
"grad_norm": 2.386315180732009,
"learning_rate": 4.994253015658708e-07,
"logits/chosen": -2.6340126991271973,
"logits/rejected": -2.6266329288482666,
"logps/chosen": -687.16845703125,
"logps/rejected": -614.4462280273438,
"loss": 0.682,
"rewards/accuracies": 0.5249999761581421,
"rewards/chosen": 0.24873919785022736,
"rewards/margins": 0.029811348766088486,
"rewards/rejected": 0.21892781555652618,
"step": 210
},
{
"epoch": 0.13,
"grad_norm": 2.2083799767077594,
"learning_rate": 4.990377781677378e-07,
"logits/chosen": -2.6552016735076904,
"logits/rejected": -2.60087251663208,
"logps/chosen": -786.2576904296875,
"logps/rejected": -686.0176391601562,
"loss": 0.6814,
"rewards/accuracies": 0.612500011920929,
"rewards/chosen": 0.29123130440711975,
"rewards/margins": 0.05502746254205704,
"rewards/rejected": 0.23620383441448212,
"step": 220
},
{
"epoch": 0.13,
"grad_norm": 2.096783018708116,
"learning_rate": 4.985511739791128e-07,
"logits/chosen": -2.7540948390960693,
"logits/rejected": -2.6966335773468018,
"logps/chosen": -676.9022216796875,
"logps/rejected": -581.7147827148438,
"loss": 0.6785,
"rewards/accuracies": 0.574999988079071,
"rewards/chosen": 0.25563615560531616,
"rewards/margins": 0.03196953982114792,
"rewards/rejected": 0.22366662323474884,
"step": 230
},
{
"epoch": 0.14,
"grad_norm": 2.084361296501474,
"learning_rate": 4.979656825976425e-07,
"logits/chosen": -2.655158281326294,
"logits/rejected": -2.5282509326934814,
"logps/chosen": -687.6748657226562,
"logps/rejected": -673.4171142578125,
"loss": 0.6809,
"rewards/accuracies": 0.550000011920929,
"rewards/chosen": 0.29197677969932556,
"rewards/margins": 0.03262047842144966,
"rewards/rejected": 0.2593563497066498,
"step": 240
},
{
"epoch": 0.14,
"grad_norm": 2.391265956326263,
"learning_rate": 4.972815369636832e-07,
"logits/chosen": -2.5233607292175293,
"logits/rejected": -2.435659170150757,
"logps/chosen": -794.3279418945312,
"logps/rejected": -715.5684204101562,
"loss": 0.6872,
"rewards/accuracies": 0.550000011920929,
"rewards/chosen": 0.3191401958465576,
"rewards/margins": 0.04753274470567703,
"rewards/rejected": 0.2716074585914612,
"step": 250
},
{
"epoch": 0.15,
"grad_norm": 2.149818689029419,
"learning_rate": 4.964990092676262e-07,
"logits/chosen": -2.6722426414489746,
"logits/rejected": -2.4782052040100098,
"logps/chosen": -764.1699829101562,
"logps/rejected": -645.8834228515625,
"loss": 0.6749,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": 0.31941699981689453,
"rewards/margins": 0.07869166880846024,
"rewards/rejected": 0.2407253533601761,
"step": 260
},
{
"epoch": 0.15,
"grad_norm": 2.035030934228571,
"learning_rate": 4.956184108416046e-07,
"logits/chosen": -2.5263333320617676,
"logits/rejected": -2.5625710487365723,
"logps/chosen": -683.1992797851562,
"logps/rejected": -625.9034423828125,
"loss": 0.6778,
"rewards/accuracies": 0.59375,
"rewards/chosen": 0.27234965562820435,
"rewards/margins": 0.041879888623952866,
"rewards/rejected": 0.23046977818012238,
"step": 270
},
{
"epoch": 0.16,
"grad_norm": 1.9569259926485463,
"learning_rate": 4.946400920356287e-07,
"logits/chosen": -2.7372500896453857,
"logits/rejected": -2.6020898818969727,
"logps/chosen": -719.0564575195312,
"logps/rejected": -637.1673583984375,
"loss": 0.6754,
"rewards/accuracies": 0.5562499761581421,
"rewards/chosen": 0.28345972299575806,
"rewards/margins": 0.04788089543581009,
"rewards/rejected": 0.23557886481285095,
"step": 280
},
{
"epoch": 0.17,
"grad_norm": 2.0726336037515845,
"learning_rate": 4.935644420781978e-07,
"logits/chosen": -2.479055881500244,
"logits/rejected": -2.4796347618103027,
"logps/chosen": -706.4686889648438,
"logps/rejected": -655.3255004882812,
"loss": 0.6822,
"rewards/accuracies": 0.59375,
"rewards/chosen": 0.3067619204521179,
"rewards/margins": 0.023757971823215485,
"rewards/rejected": 0.28300395607948303,
"step": 290
},
{
"epoch": 0.17,
"grad_norm": 2.7082840348298203,
"learning_rate": 4.923918889214436e-07,
"logits/chosen": -2.414355754852295,
"logits/rejected": -2.386721611022949,
"logps/chosen": -693.1402587890625,
"logps/rejected": -700.3583984375,
"loss": 0.6734,
"rewards/accuracies": 0.581250011920929,
"rewards/chosen": 0.2925127148628235,
"rewards/margins": 0.026889953762292862,
"rewards/rejected": 0.26562273502349854,
"step": 300
},
{
"epoch": 0.17,
"eval_logits/chosen": -2.6062464714050293,
"eval_logits/rejected": -2.522627353668213,
"eval_logps/chosen": -693.9142456054688,
"eval_logps/rejected": -628.194580078125,
"eval_loss": 0.6755158305168152,
"eval_rewards/accuracies": 0.5833333134651184,
"eval_rewards/chosen": 0.2979806959629059,
"eval_rewards/margins": 0.0508454330265522,
"eval_rewards/rejected": 0.24713526666164398,
"eval_runtime": 811.8177,
"eval_samples_per_second": 15.337,
"eval_steps_per_second": 0.48,
"step": 300
},
{
"epoch": 0.18,
"grad_norm": 2.075320347171257,
"learning_rate": 4.91122899070868e-07,
"logits/chosen": -2.5357015132904053,
"logits/rejected": -2.490417003631592,
"logps/chosen": -755.467041015625,
"logps/rejected": -663.8169555664062,
"loss": 0.6797,
"rewards/accuracies": 0.543749988079071,
"rewards/chosen": 0.3274896740913391,
"rewards/margins": 0.05784725025296211,
"rewards/rejected": 0.2696423828601837,
"step": 310
},
{
"epoch": 0.18,
"grad_norm": 2.145836596773884,
"learning_rate": 4.897579773997414e-07,
"logits/chosen": -2.5054690837860107,
"logits/rejected": -2.509059429168701,
"logps/chosen": -684.943359375,
"logps/rejected": -603.5328369140625,
"loss": 0.6739,
"rewards/accuracies": 0.5249999761581421,
"rewards/chosen": 0.30167508125305176,
"rewards/margins": 0.05937837436795235,
"rewards/rejected": 0.2422967255115509,
"step": 320
},
{
"epoch": 0.19,
"grad_norm": 2.048334012658428,
"learning_rate": 4.882976669482367e-07,
"logits/chosen": -2.792116165161133,
"logits/rejected": -2.664175271987915,
"logps/chosen": -693.2596435546875,
"logps/rejected": -672.1799926757812,
"loss": 0.6794,
"rewards/accuracies": 0.543749988079071,
"rewards/chosen": 0.3001101613044739,
"rewards/margins": 0.024096574634313583,
"rewards/rejected": 0.2760135531425476,
"step": 330
},
{
"epoch": 0.19,
"grad_norm": 2.008493005665859,
"learning_rate": 4.867425487073786e-07,
"logits/chosen": -2.3512167930603027,
"logits/rejected": -2.393057346343994,
"logps/chosen": -756.4876708984375,
"logps/rejected": -652.0723876953125,
"loss": 0.6783,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": 0.3165535628795624,
"rewards/margins": 0.07065500319004059,
"rewards/rejected": 0.2458985596895218,
"step": 340
},
{
"epoch": 0.2,
"grad_norm": 2.1734372333374883,
"learning_rate": 4.850932413878934e-07,
"logits/chosen": -2.340710401535034,
"logits/rejected": -2.298480272293091,
"logps/chosen": -745.8781127929688,
"logps/rejected": -674.5993041992188,
"loss": 0.6757,
"rewards/accuracies": 0.512499988079071,
"rewards/chosen": 0.2814175486564636,
"rewards/margins": 0.06212268024682999,
"rewards/rejected": 0.21929486095905304,
"step": 350
},
{
"epoch": 0.21,
"grad_norm": 2.4301611366343288,
"learning_rate": 4.833504011740522e-07,
"logits/chosen": -2.3556628227233887,
"logits/rejected": -2.399705410003662,
"logps/chosen": -717.0030517578125,
"logps/rejected": -651.791259765625,
"loss": 0.6731,
"rewards/accuracies": 0.5625,
"rewards/chosen": 0.2667848467826843,
"rewards/margins": 0.041503336280584335,
"rewards/rejected": 0.2252815216779709,
"step": 360
},
{
"epoch": 0.21,
"grad_norm": 2.1135061185396005,
"learning_rate": 4.815147214626056e-07,
"logits/chosen": -2.3883965015411377,
"logits/rejected": -2.427729845046997,
"logps/chosen": -803.894775390625,
"logps/rejected": -620.780517578125,
"loss": 0.6723,
"rewards/accuracies": 0.643750011920929,
"rewards/chosen": 0.3185453712940216,
"rewards/margins": 0.08323224633932114,
"rewards/rejected": 0.23531313240528107,
"step": 370
},
{
"epoch": 0.22,
"grad_norm": 2.3634608846601175,
"learning_rate": 4.795869325869116e-07,
"logits/chosen": -2.328643560409546,
"logits/rejected": -2.403995990753174,
"logps/chosen": -687.8956909179688,
"logps/rejected": -643.009765625,
"loss": 0.6724,
"rewards/accuracies": 0.518750011920929,
"rewards/chosen": 0.2998669743537903,
"rewards/margins": 0.047370560467243195,
"rewards/rejected": 0.2524964213371277,
"step": 380
},
{
"epoch": 0.22,
"grad_norm": 2.196700529213078,
"learning_rate": 4.775678015263708e-07,
"logits/chosen": -2.3591794967651367,
"logits/rejected": -2.330289125442505,
"logps/chosen": -672.6624755859375,
"logps/rejected": -570.5098266601562,
"loss": 0.6688,
"rewards/accuracies": 0.5874999761581421,
"rewards/chosen": 0.3420453667640686,
"rewards/margins": 0.07790260016918182,
"rewards/rejected": 0.264142781496048,
"step": 390
},
{
"epoch": 0.23,
"grad_norm": 2.1383346208020733,
"learning_rate": 4.7545813160127845e-07,
"logits/chosen": -2.499473810195923,
"logits/rejected": -2.3798794746398926,
"logps/chosen": -691.5140380859375,
"logps/rejected": -599.8192749023438,
"loss": 0.6684,
"rewards/accuracies": 0.637499988079071,
"rewards/chosen": 0.3622116148471832,
"rewards/margins": 0.1057838648557663,
"rewards/rejected": 0.25642773509025574,
"step": 400
},
{
"epoch": 0.23,
"eval_logits/chosen": -2.4781997203826904,
"eval_logits/rejected": -2.400707960128784,
"eval_logps/chosen": -688.9108276367188,
"eval_logps/rejected": -624.684814453125,
"eval_loss": 0.671257734298706,
"eval_rewards/accuracies": 0.5887820720672607,
"eval_rewards/chosen": 0.34801578521728516,
"eval_rewards/margins": 0.06578347086906433,
"eval_rewards/rejected": 0.2822323143482208,
"eval_runtime": 834.4654,
"eval_samples_per_second": 14.921,
"eval_steps_per_second": 0.467,
"step": 400
},
{
"epoch": 0.23,
"grad_norm": 2.1073018777983776,
"learning_rate": 4.732587621532214e-07,
"logits/chosen": -2.4502573013305664,
"logits/rejected": -2.4475185871124268,
"logps/chosen": -683.3175659179688,
"logps/rejected": -611.3568115234375,
"loss": 0.674,
"rewards/accuracies": 0.5562499761581421,
"rewards/chosen": 0.3536176085472107,
"rewards/margins": 0.057841986417770386,
"rewards/rejected": 0.29577556252479553,
"step": 410
},
{
"epoch": 0.24,
"grad_norm": 2.1659650914628155,
"learning_rate": 4.709705682111412e-07,
"logits/chosen": -2.4668691158294678,
"logits/rejected": -2.4713261127471924,
"logps/chosen": -733.1044921875,
"logps/rejected": -634.1087036132812,
"loss": 0.6715,
"rewards/accuracies": 0.637499988079071,
"rewards/chosen": 0.3784140944480896,
"rewards/margins": 0.07397519052028656,
"rewards/rejected": 0.30443888902664185,
"step": 420
},
{
"epoch": 0.25,
"grad_norm": 1.9726863761871838,
"learning_rate": 4.68594460143201e-07,
"logits/chosen": -2.469764232635498,
"logits/rejected": -2.364992618560791,
"logps/chosen": -707.831298828125,
"logps/rejected": -595.4861450195312,
"loss": 0.6661,
"rewards/accuracies": 0.637499988079071,
"rewards/chosen": 0.3741925060749054,
"rewards/margins": 0.128498837351799,
"rewards/rejected": 0.245693638920784,
"step": 430
},
{
"epoch": 0.25,
"grad_norm": 2.006300708272278,
"learning_rate": 4.661313832945903e-07,
"logits/chosen": -2.4452786445617676,
"logits/rejected": -2.4417500495910645,
"logps/chosen": -713.32763671875,
"logps/rejected": -662.4024047851562,
"loss": 0.6711,
"rewards/accuracies": 0.581250011920929,
"rewards/chosen": 0.35015571117401123,
"rewards/margins": 0.06933808326721191,
"rewards/rejected": 0.28081759810447693,
"step": 440
},
{
"epoch": 0.26,
"grad_norm": 2.6249198820895963,
"learning_rate": 4.635823176114162e-07,
"logits/chosen": -2.2743911743164062,
"logits/rejected": -2.248422384262085,
"logps/chosen": -747.7679443359375,
"logps/rejected": -706.8677978515625,
"loss": 0.6784,
"rewards/accuracies": 0.5375000238418579,
"rewards/chosen": 0.326291024684906,
"rewards/margins": 0.049748796969652176,
"rewards/rejected": 0.2765422463417053,
"step": 450
},
{
"epoch": 0.26,
"grad_norm": 2.2372785611262884,
"learning_rate": 4.6094827725082684e-07,
"logits/chosen": -2.3613054752349854,
"logits/rejected": -2.334916353225708,
"logps/chosen": -705.9910888671875,
"logps/rejected": -635.3782958984375,
"loss": 0.663,
"rewards/accuracies": 0.5874999761581421,
"rewards/chosen": 0.31965959072113037,
"rewards/margins": 0.08668084442615509,
"rewards/rejected": 0.2329787313938141,
"step": 460
},
{
"epoch": 0.27,
"grad_norm": 2.544144441997542,
"learning_rate": 4.582303101775248e-07,
"logits/chosen": -2.4770946502685547,
"logits/rejected": -2.3431713581085205,
"logps/chosen": -731.3635864257812,
"logps/rejected": -648.119873046875,
"loss": 0.6706,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": 0.3089877963066101,
"rewards/margins": 0.08075568079948425,
"rewards/rejected": 0.22823214530944824,
"step": 470
},
{
"epoch": 0.27,
"grad_norm": 2.269046581247883,
"learning_rate": 4.5542949774682956e-07,
"logits/chosen": -2.2857303619384766,
"logits/rejected": -2.31972074508667,
"logps/chosen": -721.4352416992188,
"logps/rejected": -664.8331909179688,
"loss": 0.6679,
"rewards/accuracies": 0.543749988079071,
"rewards/chosen": 0.34746861457824707,
"rewards/margins": 0.056263603270053864,
"rewards/rejected": 0.2912050187587738,
"step": 480
},
{
"epoch": 0.28,
"grad_norm": 2.5334126160826034,
"learning_rate": 4.5254695427445517e-07,
"logits/chosen": -2.4382026195526123,
"logits/rejected": -2.4473910331726074,
"logps/chosen": -746.6915283203125,
"logps/rejected": -620.1527099609375,
"loss": 0.666,
"rewards/accuracies": 0.59375,
"rewards/chosen": 0.37302321195602417,
"rewards/margins": 0.08707480877637863,
"rewards/rejected": 0.28594842553138733,
"step": 490
},
{
"epoch": 0.29,
"grad_norm": 2.319795440046946,
"learning_rate": 4.4958382659317536e-07,
"logits/chosen": -2.3574156761169434,
"logits/rejected": -2.31435489654541,
"logps/chosen": -725.4437866210938,
"logps/rejected": -665.6046142578125,
"loss": 0.6647,
"rewards/accuracies": 0.518750011920929,
"rewards/chosen": 0.3333023190498352,
"rewards/margins": 0.03737147897481918,
"rewards/rejected": 0.29593080282211304,
"step": 500
},
{
"epoch": 0.29,
"eval_logits/chosen": -2.374924659729004,
"eval_logits/rejected": -2.302616596221924,
"eval_logps/chosen": -688.75927734375,
"eval_logps/rejected": -625.8477172851562,
"eval_loss": 0.6670830249786377,
"eval_rewards/accuracies": 0.6048076748847961,
"eval_rewards/chosen": 0.349530965089798,
"eval_rewards/margins": 0.07892686128616333,
"eval_rewards/rejected": 0.27060407400131226,
"eval_runtime": 839.5836,
"eval_samples_per_second": 14.83,
"eval_steps_per_second": 0.465,
"step": 500
},
{
"epoch": 0.29,
"grad_norm": 2.36756980006423,
"learning_rate": 4.465412935965506e-07,
"logits/chosen": -2.3698456287384033,
"logits/rejected": -2.275355577468872,
"logps/chosen": -692.2882080078125,
"logps/rejected": -618.537109375,
"loss": 0.6787,
"rewards/accuracies": 0.5562499761581421,
"rewards/chosen": 0.33824020624160767,
"rewards/margins": 0.03582638129591942,
"rewards/rejected": 0.30241379141807556,
"step": 510
},
{
"epoch": 0.3,
"grad_norm": 2.0999015400485477,
"learning_rate": 4.434205657698999e-07,
"logits/chosen": -2.3519811630249023,
"logits/rejected": -2.349076747894287,
"logps/chosen": -606.4388427734375,
"logps/rejected": -607.9388427734375,
"loss": 0.6803,
"rewards/accuracies": 0.5249999761581421,
"rewards/chosen": 0.3264550268650055,
"rewards/margins": 0.020858457311987877,
"rewards/rejected": 0.3055965304374695,
"step": 520
},
{
"epoch": 0.3,
"grad_norm": 1.8949235983581547,
"learning_rate": 4.402228847087046e-07,
"logits/chosen": -2.265383720397949,
"logits/rejected": -2.20546555519104,
"logps/chosen": -687.4367065429688,
"logps/rejected": -624.896728515625,
"loss": 0.6653,
"rewards/accuracies": 0.581250011920929,
"rewards/chosen": 0.37758034467697144,
"rewards/margins": 0.0828850045800209,
"rewards/rejected": 0.2946953773498535,
"step": 530
},
{
"epoch": 0.31,
"grad_norm": 2.3607848761065453,
"learning_rate": 4.36949522624633e-07,
"logits/chosen": -2.2971131801605225,
"logits/rejected": -2.2627434730529785,
"logps/chosen": -667.89599609375,
"logps/rejected": -629.7573852539062,
"loss": 0.6669,
"rewards/accuracies": 0.5874999761581421,
"rewards/chosen": 0.3096492886543274,
"rewards/margins": 0.09841451048851013,
"rewards/rejected": 0.21123476326465607,
"step": 540
},
{
"epoch": 0.31,
"grad_norm": 2.3424108049054357,
"learning_rate": 4.33601781839386e-07,
"logits/chosen": -2.238157272338867,
"logits/rejected": -2.2228469848632812,
"logps/chosen": -761.5797119140625,
"logps/rejected": -710.4862670898438,
"loss": 0.6586,
"rewards/accuracies": 0.574999988079071,
"rewards/chosen": 0.3563387989997864,
"rewards/margins": 0.0723680704832077,
"rewards/rejected": 0.28397077322006226,
"step": 550
},
{
"epoch": 0.32,
"grad_norm": 2.0984681365446556,
"learning_rate": 4.301809942665625e-07,
"logits/chosen": -2.3568809032440186,
"logits/rejected": -2.238476037979126,
"logps/chosen": -666.4149780273438,
"logps/rejected": -600.7111206054688,
"loss": 0.669,
"rewards/accuracies": 0.612500011920929,
"rewards/chosen": 0.29119664430618286,
"rewards/margins": 0.07588323205709457,
"rewards/rejected": 0.2153133898973465,
"step": 560
},
{
"epoch": 0.33,
"grad_norm": 2.428757148167148,
"learning_rate": 4.2668852088175145e-07,
"logits/chosen": -2.1850478649139404,
"logits/rejected": -2.2067627906799316,
"logps/chosen": -727.734130859375,
"logps/rejected": -701.0474243164062,
"loss": 0.6748,
"rewards/accuracies": 0.5562499761581421,
"rewards/chosen": 0.28803592920303345,
"rewards/margins": 0.04421677812933922,
"rewards/rejected": 0.24381914734840393,
"step": 570
},
{
"epoch": 0.33,
"grad_norm": 3.3105303898170333,
"learning_rate": 4.231257511810618e-07,
"logits/chosen": -2.2972073554992676,
"logits/rejected": -2.2880618572235107,
"logps/chosen": -720.8897705078125,
"logps/rejected": -703.8472900390625,
"loss": 0.6734,
"rewards/accuracies": 0.5687500238418579,
"rewards/chosen": 0.29478007555007935,
"rewards/margins": 0.058770112693309784,
"rewards/rejected": 0.23600995540618896,
"step": 580
},
{
"epoch": 0.34,
"grad_norm": 2.414444271215555,
"learning_rate": 4.1949410262830523e-07,
"logits/chosen": -2.1948752403259277,
"logits/rejected": -2.104328155517578,
"logps/chosen": -677.14501953125,
"logps/rejected": -613.4793701171875,
"loss": 0.6555,
"rewards/accuracies": 0.706250011920929,
"rewards/chosen": 0.3152075409889221,
"rewards/margins": 0.1456335484981537,
"rewards/rejected": 0.16957402229309082,
"step": 590
},
{
"epoch": 0.34,
"grad_norm": 2.390588473535724,
"learning_rate": 4.1579502009105164e-07,
"logits/chosen": -2.2218821048736572,
"logits/rejected": -2.175379991531372,
"logps/chosen": -660.20458984375,
"logps/rejected": -598.1253051757812,
"loss": 0.6598,
"rewards/accuracies": 0.574999988079071,
"rewards/chosen": 0.28508657217025757,
"rewards/margins": 0.10129264742136002,
"rewards/rejected": 0.18379393219947815,
"step": 600
},
{
"epoch": 0.34,
"eval_logits/chosen": -2.2344727516174316,
"eval_logits/rejected": -2.1693804264068604,
"eval_logps/chosen": -690.60302734375,
"eval_logps/rejected": -628.6143188476562,
"eval_loss": 0.6635647416114807,
"eval_rewards/accuracies": 0.6057692170143127,
"eval_rewards/chosen": 0.3310932219028473,
"eval_rewards/margins": 0.08815512806177139,
"eval_rewards/rejected": 0.24293813109397888,
"eval_runtime": 836.7899,
"eval_samples_per_second": 14.879,
"eval_steps_per_second": 0.466,
"step": 600
},
{
"epoch": 0.35,
"grad_norm": 2.620103064645028,
"learning_rate": 4.120299752657827e-07,
"logits/chosen": -2.2362682819366455,
"logits/rejected": -2.1457152366638184,
"logps/chosen": -733.4589233398438,
"logps/rejected": -714.5718994140625,
"loss": 0.6743,
"rewards/accuracies": 0.581250011920929,
"rewards/chosen": 0.3162640631198883,
"rewards/margins": 0.06924857199192047,
"rewards/rejected": 0.24701550602912903,
"step": 610
},
{
"epoch": 0.35,
"grad_norm": 2.6435733608714713,
"learning_rate": 4.082004660923702e-07,
"logits/chosen": -2.16068696975708,
"logits/rejected": -2.0863850116729736,
"logps/chosen": -703.9278564453125,
"logps/rejected": -639.5438842773438,
"loss": 0.6648,
"rewards/accuracies": 0.612500011920929,
"rewards/chosen": 0.3491550087928772,
"rewards/margins": 0.07219501584768295,
"rewards/rejected": 0.2769600450992584,
"step": 620
},
{
"epoch": 0.36,
"grad_norm": 2.4580793288834273,
"learning_rate": 4.0430801615811437e-07,
"logits/chosen": -2.138211965560913,
"logits/rejected": -2.1236751079559326,
"logps/chosen": -681.3580322265625,
"logps/rejected": -662.194580078125,
"loss": 0.6677,
"rewards/accuracies": 0.53125,
"rewards/chosen": 0.3004602789878845,
"rewards/margins": 0.06852121651172638,
"rewards/rejected": 0.23193907737731934,
"step": 630
},
{
"epoch": 0.37,
"grad_norm": 3.1611029274796003,
"learning_rate": 4.00354174091578e-07,
"logits/chosen": -2.075774669647217,
"logits/rejected": -2.054658889770508,
"logps/chosen": -674.4803466796875,
"logps/rejected": -636.6451416015625,
"loss": 0.6658,
"rewards/accuracies": 0.6187499761581421,
"rewards/chosen": 0.34417495131492615,
"rewards/margins": 0.10877577215433121,
"rewards/rejected": 0.23539917171001434,
"step": 640
},
{
"epoch": 0.37,
"grad_norm": 2.955455771466899,
"learning_rate": 3.963405129464569e-07,
"logits/chosen": -2.1316115856170654,
"logits/rejected": -2.0590739250183105,
"logps/chosen": -729.1641845703125,
"logps/rejected": -655.9600219726562,
"loss": 0.6581,
"rewards/accuracies": 0.612500011920929,
"rewards/chosen": 0.31629639863967896,
"rewards/margins": 0.0853310376405716,
"rewards/rejected": 0.23096534609794617,
"step": 650
},
{
"epoch": 0.38,
"grad_norm": 3.20641486824199,
"learning_rate": 3.922686295757347e-07,
"logits/chosen": -2.1304194927215576,
"logits/rejected": -2.051994800567627,
"logps/chosen": -764.9317016601562,
"logps/rejected": -722.5567626953125,
"loss": 0.6564,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": 0.35463058948516846,
"rewards/margins": 0.09359289705753326,
"rewards/rejected": 0.2610377073287964,
"step": 660
},
{
"epoch": 0.38,
"grad_norm": 3.3096544804003214,
"learning_rate": 3.881401439963666e-07,
"logits/chosen": -2.057319164276123,
"logits/rejected": -2.015831232070923,
"logps/chosen": -588.3761596679688,
"logps/rejected": -567.8236083984375,
"loss": 0.6641,
"rewards/accuracies": 0.543749988079071,
"rewards/chosen": 0.2484780102968216,
"rewards/margins": 0.057387322187423706,
"rewards/rejected": 0.1910906732082367,
"step": 670
},
{
"epoch": 0.39,
"grad_norm": 2.8331676817387628,
"learning_rate": 3.839566987447491e-07,
"logits/chosen": -2.050976514816284,
"logits/rejected": -2.0163493156433105,
"logps/chosen": -749.3988037109375,
"logps/rejected": -670.8482055664062,
"loss": 0.6734,
"rewards/accuracies": 0.6187499761581421,
"rewards/chosen": 0.29123300313949585,
"rewards/margins": 0.083717942237854,
"rewards/rejected": 0.20751512050628662,
"step": 680
},
{
"epoch": 0.39,
"grad_norm": 2.6749133072665843,
"learning_rate": 3.7971995822322947e-07,
"logits/chosen": -2.019416332244873,
"logits/rejected": -2.0274858474731445,
"logps/chosen": -696.1351928710938,
"logps/rejected": -654.2576293945312,
"loss": 0.6615,
"rewards/accuracies": 0.5625,
"rewards/chosen": 0.31569501757621765,
"rewards/margins": 0.07416489720344543,
"rewards/rejected": 0.2415301501750946,
"step": 690
},
{
"epoch": 0.4,
"grad_norm": 2.457414583878444,
"learning_rate": 3.7543160803791545e-07,
"logits/chosen": -1.9960815906524658,
"logits/rejected": -1.9505565166473389,
"logps/chosen": -721.5948486328125,
"logps/rejected": -651.5316162109375,
"loss": 0.6598,
"rewards/accuracies": 0.550000011920929,
"rewards/chosen": 0.31404954195022583,
"rewards/margins": 0.10105651617050171,
"rewards/rejected": 0.21299299597740173,
"step": 700
},
{
"epoch": 0.4,
"eval_logits/chosen": -1.9781278371810913,
"eval_logits/rejected": -1.9252119064331055,
"eval_logps/chosen": -695.4718017578125,
"eval_logps/rejected": -634.3779296875,
"eval_loss": 0.6605694890022278,
"eval_rewards/accuracies": 0.6105769276618958,
"eval_rewards/chosen": 0.28240564465522766,
"eval_rewards/margins": 0.09710400551557541,
"eval_rewards/rejected": 0.18530163168907166,
"eval_runtime": 834.416,
"eval_samples_per_second": 14.922,
"eval_steps_per_second": 0.467,
"step": 700
},
{
"epoch": 0.41,
"grad_norm": 2.4659247559278366,
"learning_rate": 3.7109335432805e-07,
"logits/chosen": -1.9180772304534912,
"logits/rejected": -1.9245834350585938,
"logps/chosen": -758.48876953125,
"logps/rejected": -619.541748046875,
"loss": 0.6636,
"rewards/accuracies": 0.625,
"rewards/chosen": 0.3083231747150421,
"rewards/margins": 0.11108432710170746,
"rewards/rejected": 0.19723883271217346,
"step": 710
},
{
"epoch": 0.41,
"grad_norm": 2.620943282000028,
"learning_rate": 3.667069230872155e-07,
"logits/chosen": -1.853564977645874,
"logits/rejected": -1.833913803100586,
"logps/chosen": -760.102783203125,
"logps/rejected": -648.6535034179688,
"loss": 0.6558,
"rewards/accuracies": 0.612500011920929,
"rewards/chosen": 0.33468401432037354,
"rewards/margins": 0.09708525240421295,
"rewards/rejected": 0.23759880661964417,
"step": 720
},
{
"epoch": 0.42,
"grad_norm": 2.2579820514508264,
"learning_rate": 3.6227405947664003e-07,
"logits/chosen": -1.9256470203399658,
"logits/rejected": -1.885286569595337,
"logps/chosen": -790.8487548828125,
"logps/rejected": -700.6820678710938,
"loss": 0.6591,
"rewards/accuracies": 0.637499988079071,
"rewards/chosen": 0.38141968846321106,
"rewards/margins": 0.15458762645721436,
"rewards/rejected": 0.22683203220367432,
"step": 730
},
{
"epoch": 0.42,
"grad_norm": 2.536381576442856,
"learning_rate": 3.577965271308771e-07,
"logits/chosen": -1.9556684494018555,
"logits/rejected": -1.9257959127426147,
"logps/chosen": -735.0540161132812,
"logps/rejected": -654.5272216796875,
"loss": 0.6585,
"rewards/accuracies": 0.606249988079071,
"rewards/chosen": 0.3390718102455139,
"rewards/margins": 0.1305239498615265,
"rewards/rejected": 0.20854787528514862,
"step": 740
},
{
"epoch": 0.43,
"grad_norm": 2.3455105370717737,
"learning_rate": 3.5327610745613546e-07,
"logits/chosen": -1.9934170246124268,
"logits/rejected": -1.9228801727294922,
"logps/chosen": -700.4527587890625,
"logps/rejected": -634.6886596679688,
"loss": 0.657,
"rewards/accuracies": 0.606249988079071,
"rewards/chosen": 0.32780513167381287,
"rewards/margins": 0.12030024826526642,
"rewards/rejected": 0.20750489830970764,
"step": 750
},
{
"epoch": 0.43,
"grad_norm": 2.5681519746805845,
"learning_rate": 3.487145989215391e-07,
"logits/chosen": -1.9263842105865479,
"logits/rejected": -1.8778858184814453,
"logps/chosen": -745.1176147460938,
"logps/rejected": -652.140625,
"loss": 0.6579,
"rewards/accuracies": 0.643750011920929,
"rewards/chosen": 0.34303581714630127,
"rewards/margins": 0.13984538614749908,
"rewards/rejected": 0.203190416097641,
"step": 760
},
{
"epoch": 0.44,
"grad_norm": 2.8084666872092705,
"learning_rate": 3.4411381634359796e-07,
"logits/chosen": -1.9643045663833618,
"logits/rejected": -1.9424177408218384,
"logps/chosen": -738.8779907226562,
"logps/rejected": -653.993408203125,
"loss": 0.6513,
"rewards/accuracies": 0.53125,
"rewards/chosen": 0.3104199767112732,
"rewards/margins": 0.0740085169672966,
"rewards/rejected": 0.236411452293396,
"step": 770
},
{
"epoch": 0.45,
"grad_norm": 2.5437521131160805,
"learning_rate": 3.3947559016417554e-07,
"logits/chosen": -2.015496253967285,
"logits/rejected": -1.9664264917373657,
"logps/chosen": -730.8230590820312,
"logps/rejected": -708.7545776367188,
"loss": 0.6735,
"rewards/accuracies": 0.5562499761581421,
"rewards/chosen": 0.3310191333293915,
"rewards/margins": 0.07294587790966034,
"rewards/rejected": 0.2580733001232147,
"step": 780
},
{
"epoch": 0.45,
"grad_norm": 3.253152159173173,
"learning_rate": 3.3480176572223885e-07,
"logits/chosen": -2.022552728652954,
"logits/rejected": -1.94183349609375,
"logps/chosen": -637.3079223632812,
"logps/rejected": -579.6361083984375,
"loss": 0.6487,
"rewards/accuracies": 0.668749988079071,
"rewards/chosen": 0.3080602288246155,
"rewards/margins": 0.15390750765800476,
"rewards/rejected": 0.15415272116661072,
"step": 790
},
{
"epoch": 0.46,
"grad_norm": 2.4835060692710464,
"learning_rate": 3.300942025196824e-07,
"logits/chosen": -2.063943862915039,
"logits/rejected": -2.0479979515075684,
"logps/chosen": -713.1132202148438,
"logps/rejected": -639.5491943359375,
"loss": 0.6563,
"rewards/accuracies": 0.637499988079071,
"rewards/chosen": 0.389708936214447,
"rewards/margins": 0.15992382168769836,
"rewards/rejected": 0.22978505492210388,
"step": 800
},
{
"epoch": 0.46,
"eval_logits/chosen": -2.0598788261413574,
"eval_logits/rejected": -2.0029659271240234,
"eval_logps/chosen": -688.9520874023438,
"eval_logps/rejected": -629.170654296875,
"eval_loss": 0.658465564250946,
"eval_rewards/accuracies": 0.6070512533187866,
"eval_rewards/chosen": 0.3476024270057678,
"eval_rewards/margins": 0.11022823303937912,
"eval_rewards/rejected": 0.2373741716146469,
"eval_runtime": 832.2648,
"eval_samples_per_second": 14.96,
"eval_steps_per_second": 0.469,
"step": 800
},
{
"epoch": 0.46,
"grad_norm": 2.4337446470444024,
"learning_rate": 3.253547734815173e-07,
"logits/chosen": -2.0798535346984863,
"logits/rejected": -2.061521530151367,
"logps/chosen": -715.1067504882812,
"logps/rejected": -732.9197998046875,
"loss": 0.6594,
"rewards/accuracies": 0.6499999761581421,
"rewards/chosen": 0.4086131453514099,
"rewards/margins": 0.1309557557106018,
"rewards/rejected": 0.2776573598384857,
"step": 810
},
{
"epoch": 0.47,
"grad_norm": 2.2842891498898275,
"learning_rate": 3.2058536421071914e-07,
"logits/chosen": -2.098947048187256,
"logits/rejected": -2.087313652038574,
"logps/chosen": -768.4358520507812,
"logps/rejected": -626.5296630859375,
"loss": 0.6657,
"rewards/accuracies": 0.5687500238418579,
"rewards/chosen": 0.4160265028476715,
"rewards/margins": 0.11269855499267578,
"rewards/rejected": 0.3033279478549957,
"step": 820
},
{
"epoch": 0.47,
"grad_norm": 2.3495187388609238,
"learning_rate": 3.1578787223803294e-07,
"logits/chosen": -1.9842846393585205,
"logits/rejected": -1.9742408990859985,
"logps/chosen": -700.257568359375,
"logps/rejected": -584.7598876953125,
"loss": 0.662,
"rewards/accuracies": 0.5874999761581421,
"rewards/chosen": 0.3193957507610321,
"rewards/margins": 0.058227695524692535,
"rewards/rejected": 0.26116806268692017,
"step": 830
},
{
"epoch": 0.48,
"grad_norm": 2.268979187768203,
"learning_rate": 3.109642062670318e-07,
"logits/chosen": -2.0449118614196777,
"logits/rejected": -1.9595324993133545,
"logps/chosen": -633.672607421875,
"logps/rejected": -602.7111206054688,
"loss": 0.6523,
"rewards/accuracies": 0.606249988079071,
"rewards/chosen": 0.37880054116249084,
"rewards/margins": 0.11559071391820908,
"rewards/rejected": 0.26320984959602356,
"step": 840
},
{
"epoch": 0.49,
"grad_norm": 2.445431457272807,
"learning_rate": 3.0611628541473074e-07,
"logits/chosen": -2.0798251628875732,
"logits/rejected": -2.0806803703308105,
"logps/chosen": -776.8604125976562,
"logps/rejected": -754.062255859375,
"loss": 0.6676,
"rewards/accuracies": 0.59375,
"rewards/chosen": 0.4384000301361084,
"rewards/margins": 0.1085997074842453,
"rewards/rejected": 0.3298003077507019,
"step": 850
},
{
"epoch": 0.49,
"grad_norm": 2.4068587077954597,
"learning_rate": 3.012460384480576e-07,
"logits/chosen": -2.0535523891448975,
"logits/rejected": -2.003754138946533,
"logps/chosen": -708.5001220703125,
"logps/rejected": -663.521240234375,
"loss": 0.6466,
"rewards/accuracies": 0.6625000238418579,
"rewards/chosen": 0.4629546105861664,
"rewards/margins": 0.14931170642375946,
"rewards/rejected": 0.3136429190635681,
"step": 860
},
{
"epoch": 0.5,
"grad_norm": 2.540723241212297,
"learning_rate": 2.963554030164842e-07,
"logits/chosen": -1.893160104751587,
"logits/rejected": -1.9689306020736694,
"logps/chosen": -726.8738403320312,
"logps/rejected": -671.0678100585938,
"loss": 0.6633,
"rewards/accuracies": 0.5562499761581421,
"rewards/chosen": 0.4015054702758789,
"rewards/margins": 0.07470430433750153,
"rewards/rejected": 0.3268011808395386,
"step": 870
},
{
"epoch": 0.5,
"grad_norm": 3.0009975628742174,
"learning_rate": 2.9144632488112416e-07,
"logits/chosen": -2.020155429840088,
"logits/rejected": -2.016500473022461,
"logps/chosen": -721.1326904296875,
"logps/rejected": -660.15966796875,
"loss": 0.6551,
"rewards/accuracies": 0.6812499761581421,
"rewards/chosen": 0.38422372937202454,
"rewards/margins": 0.16717380285263062,
"rewards/rejected": 0.21704988181591034,
"step": 880
},
{
"epoch": 0.51,
"grad_norm": 2.651435689363449,
"learning_rate": 2.865207571406029e-07,
"logits/chosen": -2.0237433910369873,
"logits/rejected": -1.9571411609649658,
"logps/chosen": -692.2257690429688,
"logps/rejected": -613.42431640625,
"loss": 0.6684,
"rewards/accuracies": 0.59375,
"rewards/chosen": 0.34494125843048096,
"rewards/margins": 0.0645814761519432,
"rewards/rejected": 0.28035980463027954,
"step": 890
},
{
"epoch": 0.51,
"grad_norm": 2.5848228401368334,
"learning_rate": 2.815806594540089e-07,
"logits/chosen": -2.0166923999786377,
"logits/rejected": -2.052598476409912,
"logps/chosen": -720.5625,
"logps/rejected": -678.0027465820312,
"loss": 0.6636,
"rewards/accuracies": 0.581250011920929,
"rewards/chosen": 0.40049290657043457,
"rewards/margins": 0.06527121365070343,
"rewards/rejected": 0.33522164821624756,
"step": 900
},
{
"epoch": 0.51,
"eval_logits/chosen": -2.0439984798431396,
"eval_logits/rejected": -1.9872162342071533,
"eval_logps/chosen": -688.0209350585938,
"eval_logps/rejected": -628.6378784179688,
"eval_loss": 0.6572163701057434,
"eval_rewards/accuracies": 0.6118589639663696,
"eval_rewards/chosen": 0.3569147288799286,
"eval_rewards/margins": 0.11421282589435577,
"eval_rewards/rejected": 0.24270187318325043,
"eval_runtime": 811.7757,
"eval_samples_per_second": 15.338,
"eval_steps_per_second": 0.48,
"step": 900
},
{
"epoch": 0.52,
"grad_norm": 2.8783002475743547,
"learning_rate": 2.7662799726123454e-07,
"logits/chosen": -1.9933090209960938,
"logits/rejected": -1.9611231088638306,
"logps/chosen": -637.7457275390625,
"logps/rejected": -599.6551513671875,
"loss": 0.6639,
"rewards/accuracies": 0.543749988079071,
"rewards/chosen": 0.28943049907684326,
"rewards/margins": 0.0467146560549736,
"rewards/rejected": 0.24271583557128906,
"step": 910
},
{
"epoch": 0.53,
"grad_norm": 3.137998102462479,
"learning_rate": 2.7166474100101674e-07,
"logits/chosen": -1.9679996967315674,
"logits/rejected": -1.9321043491363525,
"logps/chosen": -665.1378173828125,
"logps/rejected": -610.6767578125,
"loss": 0.6592,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": 0.3013032078742981,
"rewards/margins": 0.11742081493139267,
"rewards/rejected": 0.18388235569000244,
"step": 920
},
{
"epoch": 0.53,
"grad_norm": 2.5841977136958727,
"learning_rate": 2.6669286532698964e-07,
"logits/chosen": -1.9860007762908936,
"logits/rejected": -1.9256454706192017,
"logps/chosen": -844.6925659179688,
"logps/rejected": -707.6446533203125,
"loss": 0.6465,
"rewards/accuracies": 0.581250011920929,
"rewards/chosen": 0.38923513889312744,
"rewards/margins": 0.14900648593902588,
"rewards/rejected": 0.24022865295410156,
"step": 930
},
{
"epoch": 0.54,
"grad_norm": 2.7091636904536887,
"learning_rate": 2.6171434832205953e-07,
"logits/chosen": -1.9356151819229126,
"logits/rejected": -1.9506133794784546,
"logps/chosen": -715.4945068359375,
"logps/rejected": -654.3660278320312,
"loss": 0.6441,
"rewards/accuracies": 0.5249999761581421,
"rewards/chosen": 0.35090070962905884,
"rewards/margins": 0.0896492600440979,
"rewards/rejected": 0.26125144958496094,
"step": 940
},
{
"epoch": 0.54,
"grad_norm": 2.724306235445107,
"learning_rate": 2.567311707114157e-07,
"logits/chosen": -2.0761585235595703,
"logits/rejected": -2.057460069656372,
"logps/chosen": -638.9573364257812,
"logps/rejected": -622.14697265625,
"loss": 0.6766,
"rewards/accuracies": 0.5562499761581421,
"rewards/chosen": 0.3293986916542053,
"rewards/margins": 0.06820010393857956,
"rewards/rejected": 0.26119858026504517,
"step": 950
},
{
"epoch": 0.55,
"grad_norm": 2.752955939468096,
"learning_rate": 2.5174531507449037e-07,
"logits/chosen": -2.070046901702881,
"logits/rejected": -1.9810707569122314,
"logps/chosen": -616.99169921875,
"logps/rejected": -543.088134765625,
"loss": 0.657,
"rewards/accuracies": 0.6499999761581421,
"rewards/chosen": 0.3693603575229645,
"rewards/margins": 0.14052534103393555,
"rewards/rejected": 0.22883501648902893,
"step": 960
},
{
"epoch": 0.55,
"grad_norm": 2.4658663047675575,
"learning_rate": 2.4675876505618045e-07,
"logits/chosen": -2.007403612136841,
"logits/rejected": -2.010920763015747,
"logps/chosen": -762.5975952148438,
"logps/rejected": -657.7164306640625,
"loss": 0.6634,
"rewards/accuracies": 0.6312500238418579,
"rewards/chosen": 0.3691919445991516,
"rewards/margins": 0.13456907868385315,
"rewards/rejected": 0.23462291061878204,
"step": 970
},
{
"epoch": 0.56,
"grad_norm": 2.8093493901482645,
"learning_rate": 2.4177350457764527e-07,
"logits/chosen": -1.9829814434051514,
"logits/rejected": -1.9325615167617798,
"logps/chosen": -760.7388305664062,
"logps/rejected": -746.89697265625,
"loss": 0.6639,
"rewards/accuracies": 0.6312500238418579,
"rewards/chosen": 0.3807603120803833,
"rewards/margins": 0.1183190792798996,
"rewards/rejected": 0.2624412178993225,
"step": 980
},
{
"epoch": 0.57,
"grad_norm": 3.2352313043894214,
"learning_rate": 2.367915170469956e-07,
"logits/chosen": -2.02712345123291,
"logits/rejected": -1.865962028503418,
"logps/chosen": -726.9395751953125,
"logps/rejected": -638.2818603515625,
"loss": 0.658,
"rewards/accuracies": 0.643750011920929,
"rewards/chosen": 0.3744068741798401,
"rewards/margins": 0.17671848833560944,
"rewards/rejected": 0.19768838584423065,
"step": 990
},
{
"epoch": 0.57,
"grad_norm": 3.0201719958390263,
"learning_rate": 2.3181478457018475e-07,
"logits/chosen": -2.0808539390563965,
"logits/rejected": -2.021824359893799,
"logps/chosen": -777.3084106445312,
"logps/rejected": -668.2219848632812,
"loss": 0.6436,
"rewards/accuracies": 0.612500011920929,
"rewards/chosen": 0.3560071587562561,
"rewards/margins": 0.17518427968025208,
"rewards/rejected": 0.18082287907600403,
"step": 1000
},
{
"epoch": 0.57,
"eval_logits/chosen": -2.018136501312256,
"eval_logits/rejected": -1.9617843627929688,
"eval_logps/chosen": -694.4999389648438,
"eval_logps/rejected": -635.5912475585938,
"eval_loss": 0.6557932496070862,
"eval_rewards/accuracies": 0.6096153855323792,
"eval_rewards/chosen": 0.29212433099746704,
"eval_rewards/margins": 0.11895613372325897,
"eval_rewards/rejected": 0.17316819727420807,
"eval_runtime": 831.0231,
"eval_samples_per_second": 14.983,
"eval_steps_per_second": 0.469,
"step": 1000
},
{
"epoch": 0.58,
"grad_norm": 2.5885677640784626,
"learning_rate": 2.2684528716241968e-07,
"logits/chosen": -1.9237916469573975,
"logits/rejected": -1.9125349521636963,
"logps/chosen": -777.046142578125,
"logps/rejected": -745.3409423828125,
"loss": 0.6548,
"rewards/accuracies": 0.550000011920929,
"rewards/chosen": 0.3333626985549927,
"rewards/margins": 0.11125592142343521,
"rewards/rejected": 0.22210678458213806,
"step": 1010
},
{
"epoch": 0.58,
"grad_norm": 2.5438960725196296,
"learning_rate": 2.218850019604028e-07,
"logits/chosen": -2.0038065910339355,
"logits/rejected": -1.925611138343811,
"logps/chosen": -680.6759033203125,
"logps/rejected": -662.7098388671875,
"loss": 0.6545,
"rewards/accuracies": 0.5625,
"rewards/chosen": 0.2969500422477722,
"rewards/margins": 0.14483226835727692,
"rewards/rejected": 0.1521177738904953,
"step": 1020
},
{
"epoch": 0.59,
"grad_norm": 2.8410942952474136,
"learning_rate": 2.1693590243571935e-07,
"logits/chosen": -2.0043246746063232,
"logits/rejected": -1.9441509246826172,
"logps/chosen": -804.6784057617188,
"logps/rejected": -714.288330078125,
"loss": 0.6581,
"rewards/accuracies": 0.550000011920929,
"rewards/chosen": 0.36427271366119385,
"rewards/margins": 0.061019230633974075,
"rewards/rejected": 0.3032534718513489,
"step": 1030
},
{
"epoch": 0.59,
"grad_norm": 2.611147125105862,
"learning_rate": 2.119999576096832e-07,
"logits/chosen": -2.049757480621338,
"logits/rejected": -1.9409106969833374,
"logps/chosen": -737.8204345703125,
"logps/rejected": -626.43701171875,
"loss": 0.6571,
"rewards/accuracies": 0.6499999761581421,
"rewards/chosen": 0.3496776223182678,
"rewards/margins": 0.1673753708600998,
"rewards/rejected": 0.18230223655700684,
"step": 1040
},
{
"epoch": 0.6,
"grad_norm": 2.289802791317009,
"learning_rate": 2.0707913126995244e-07,
"logits/chosen": -1.958674669265747,
"logits/rejected": -1.9054616689682007,
"logps/chosen": -794.5623779296875,
"logps/rejected": -620.9688720703125,
"loss": 0.6522,
"rewards/accuracies": 0.699999988079071,
"rewards/chosen": 0.39125892519950867,
"rewards/margins": 0.23107214272022247,
"rewards/rejected": 0.1601867973804474,
"step": 1050
},
{
"epoch": 0.61,
"grad_norm": 2.640295515988177,
"learning_rate": 2.021753811892281e-07,
"logits/chosen": -1.9902336597442627,
"logits/rejected": -2.0175299644470215,
"logps/chosen": -729.5810546875,
"logps/rejected": -599.620361328125,
"loss": 0.6502,
"rewards/accuracies": 0.6187499761581421,
"rewards/chosen": 0.37484925985336304,
"rewards/margins": 0.14430688321590424,
"rewards/rejected": 0.2305423766374588,
"step": 1060
},
{
"epoch": 0.61,
"grad_norm": 2.5023450766367965,
"learning_rate": 1.9729065834634528e-07,
"logits/chosen": -2.086310863494873,
"logits/rejected": -2.0709755420684814,
"logps/chosen": -746.2198486328125,
"logps/rejected": -680.5553588867188,
"loss": 0.6511,
"rewards/accuracies": 0.625,
"rewards/chosen": 0.39980658888816833,
"rewards/margins": 0.16294142603874207,
"rewards/rejected": 0.23686519265174866,
"step": 1070
},
{
"epoch": 0.62,
"grad_norm": 2.62481893045694,
"learning_rate": 1.924269061500678e-07,
"logits/chosen": -2.0319840908050537,
"logits/rejected": -2.07069730758667,
"logps/chosen": -641.7462768554688,
"logps/rejected": -621.5814819335938,
"loss": 0.6579,
"rewards/accuracies": 0.5062500238418579,
"rewards/chosen": 0.2810671627521515,
"rewards/margins": 0.05967780202627182,
"rewards/rejected": 0.22138938307762146,
"step": 1080
},
{
"epoch": 0.62,
"grad_norm": 2.405136558965165,
"learning_rate": 1.8758605966589434e-07,
"logits/chosen": -2.0415592193603516,
"logits/rejected": -1.9661998748779297,
"logps/chosen": -858.5528564453125,
"logps/rejected": -739.1558837890625,
"loss": 0.6549,
"rewards/accuracies": 0.6312500238418579,
"rewards/chosen": 0.3966771960258484,
"rewards/margins": 0.10974051058292389,
"rewards/rejected": 0.2869366705417633,
"step": 1090
},
{
"epoch": 0.63,
"grad_norm": 3.3922492394829504,
"learning_rate": 1.8277004484618357e-07,
"logits/chosen": -1.9904890060424805,
"logits/rejected": -1.9474655389785767,
"logps/chosen": -736.5589599609375,
"logps/rejected": -658.781005859375,
"loss": 0.6759,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": 0.3667674660682678,
"rewards/margins": 0.1037057489156723,
"rewards/rejected": 0.2630617022514343,
"step": 1100
},
{
"epoch": 0.63,
"eval_logits/chosen": -2.019803047180176,
"eval_logits/rejected": -1.9627211093902588,
"eval_logps/chosen": -689.3489379882812,
"eval_logps/rejected": -631.2626342773438,
"eval_loss": 0.6548019051551819,
"eval_rewards/accuracies": 0.6070512533187866,
"eval_rewards/chosen": 0.3436334729194641,
"eval_rewards/margins": 0.1271788775920868,
"eval_rewards/rejected": 0.21645459532737732,
"eval_runtime": 811.6476,
"eval_samples_per_second": 15.34,
"eval_steps_per_second": 0.481,
"step": 1100
},
{
"epoch": 0.63,
"grad_norm": 2.8285528920600167,
"learning_rate": 1.7798077776390603e-07,
"logits/chosen": -1.9690300226211548,
"logits/rejected": -1.923657774925232,
"logps/chosen": -721.0764770507812,
"logps/rejected": -610.729736328125,
"loss": 0.6657,
"rewards/accuracies": 0.606249988079071,
"rewards/chosen": 0.3101206123828888,
"rewards/margins": 0.13614089787006378,
"rewards/rejected": 0.17397968471050262,
"step": 1110
},
{
"epoch": 0.64,
"grad_norm": 3.086737649194682,
"learning_rate": 1.7322016385032536e-07,
"logits/chosen": -2.0291314125061035,
"logits/rejected": -1.9643595218658447,
"logps/chosen": -809.5503540039062,
"logps/rejected": -699.6201782226562,
"loss": 0.6512,
"rewards/accuracies": 0.6499999761581421,
"rewards/chosen": 0.4089950621128082,
"rewards/margins": 0.21505212783813477,
"rewards/rejected": 0.19394293427467346,
"step": 1120
},
{
"epoch": 0.65,
"grad_norm": 3.170259473129296,
"learning_rate": 1.6849009713691454e-07,
"logits/chosen": -1.9695608615875244,
"logits/rejected": -1.958207130432129,
"logps/chosen": -752.0731201171875,
"logps/rejected": -697.3651733398438,
"loss": 0.6638,
"rewards/accuracies": 0.5562499761581421,
"rewards/chosen": 0.29536861181259155,
"rewards/margins": 0.09291869401931763,
"rewards/rejected": 0.20244988799095154,
"step": 1130
},
{
"epoch": 0.65,
"grad_norm": 2.6277921924177963,
"learning_rate": 1.6379245950180666e-07,
"logits/chosen": -1.9219013452529907,
"logits/rejected": -1.9105087518692017,
"logps/chosen": -740.3037109375,
"logps/rejected": -609.6246948242188,
"loss": 0.633,
"rewards/accuracies": 0.625,
"rewards/chosen": 0.35257434844970703,
"rewards/margins": 0.16397470235824585,
"rewards/rejected": 0.18859967589378357,
"step": 1140
},
{
"epoch": 0.66,
"grad_norm": 2.5332783060270123,
"learning_rate": 1.591291199210813e-07,
"logits/chosen": -1.9232860803604126,
"logits/rejected": -1.9083904027938843,
"logps/chosen": -665.9305419921875,
"logps/rejected": -630.5094604492188,
"loss": 0.6625,
"rewards/accuracies": 0.5562499761581421,
"rewards/chosen": 0.28770583868026733,
"rewards/margins": 0.05961765721440315,
"rewards/rejected": 0.22808821499347687,
"step": 1150
},
{
"epoch": 0.66,
"grad_norm": 2.3275797683615105,
"learning_rate": 1.5450193372518438e-07,
"logits/chosen": -1.9562702178955078,
"logits/rejected": -1.8471969366073608,
"logps/chosen": -663.4168090820312,
"logps/rejected": -641.2056884765625,
"loss": 0.639,
"rewards/accuracies": 0.59375,
"rewards/chosen": 0.3817751407623291,
"rewards/margins": 0.15303686261177063,
"rewards/rejected": 0.22873827815055847,
"step": 1160
},
{
"epoch": 0.67,
"grad_norm": 2.782849328804318,
"learning_rate": 1.4991274186077628e-07,
"logits/chosen": -1.9461126327514648,
"logits/rejected": -1.8650624752044678,
"logps/chosen": -717.2433471679688,
"logps/rejected": -608.2086181640625,
"loss": 0.6518,
"rewards/accuracies": 0.59375,
"rewards/chosen": 0.38884252309799194,
"rewards/margins": 0.1517229527235031,
"rewards/rejected": 0.23711958527565002,
"step": 1170
},
{
"epoch": 0.67,
"grad_norm": 2.514896941566138,
"learning_rate": 1.4536337015830323e-07,
"logits/chosen": -1.964003562927246,
"logits/rejected": -1.9508689641952515,
"logps/chosen": -717.3407592773438,
"logps/rejected": -673.0404663085938,
"loss": 0.6594,
"rewards/accuracies": 0.5625,
"rewards/chosen": 0.3360191583633423,
"rewards/margins": 0.10252443701028824,
"rewards/rejected": 0.23349472880363464,
"step": 1180
},
{
"epoch": 0.68,
"grad_norm": 2.6031614387547606,
"learning_rate": 1.4085562860558253e-07,
"logits/chosen": -1.92083740234375,
"logits/rejected": -1.8824331760406494,
"logps/chosen": -742.9560546875,
"logps/rejected": -657.2242431640625,
"loss": 0.6493,
"rewards/accuracies": 0.637499988079071,
"rewards/chosen": 0.3182393014431,
"rewards/margins": 0.14292821288108826,
"rewards/rejected": 0.17531108856201172,
"step": 1190
},
{
"epoch": 0.69,
"grad_norm": 3.4953278158305303,
"learning_rate": 1.3639131062769075e-07,
"logits/chosen": -1.9272648096084595,
"logits/rejected": -1.9646079540252686,
"logps/chosen": -699.5802001953125,
"logps/rejected": -622.5048217773438,
"loss": 0.6679,
"rewards/accuracies": 0.625,
"rewards/chosen": 0.3422418236732483,
"rewards/margins": 0.12412543594837189,
"rewards/rejected": 0.2181164026260376,
"step": 1200
},
{
"epoch": 0.69,
"eval_logits/chosen": -1.959836721420288,
"eval_logits/rejected": -1.9057732820510864,
"eval_logps/chosen": -688.3819580078125,
"eval_logps/rejected": -630.7877807617188,
"eval_loss": 0.6542019844055176,
"eval_rewards/accuracies": 0.607692301273346,
"eval_rewards/chosen": 0.35330361127853394,
"eval_rewards/margins": 0.1320999264717102,
"eval_rewards/rejected": 0.22120368480682373,
"eval_runtime": 829.7168,
"eval_samples_per_second": 15.006,
"eval_steps_per_second": 0.47,
"step": 1200
},
{
"epoch": 0.69,
"grad_norm": 2.9526761996356634,
"learning_rate": 1.3197219237344155e-07,
"logits/chosen": -1.979193925857544,
"logits/rejected": -1.9904626607894897,
"logps/chosen": -770.7225341796875,
"logps/rejected": -613.1958618164062,
"loss": 0.6464,
"rewards/accuracies": 0.699999988079071,
"rewards/chosen": 0.4239157736301422,
"rewards/margins": 0.21235182881355286,
"rewards/rejected": 0.21156391501426697,
"step": 1210
},
{
"epoch": 0.7,
"grad_norm": 2.314858281629497,
"learning_rate": 1.2760003200873697e-07,
"logits/chosen": -1.9512065649032593,
"logits/rejected": -1.902094841003418,
"logps/chosen": -712.6738891601562,
"logps/rejected": -621.5015869140625,
"loss": 0.6594,
"rewards/accuracies": 0.6187499761581421,
"rewards/chosen": 0.3171978294849396,
"rewards/margins": 0.13169512152671814,
"rewards/rejected": 0.18550269305706024,
"step": 1220
},
{
"epoch": 0.7,
"grad_norm": 2.884626651322277,
"learning_rate": 1.2327656901707338e-07,
"logits/chosen": -1.8530919551849365,
"logits/rejected": -1.8408454656600952,
"logps/chosen": -772.9168701171875,
"logps/rejected": -740.7579345703125,
"loss": 0.6464,
"rewards/accuracies": 0.581250011920929,
"rewards/chosen": 0.3601977229118347,
"rewards/margins": 0.16855189204216003,
"rewards/rejected": 0.1916457712650299,
"step": 1230
},
{
"epoch": 0.71,
"grad_norm": 3.5188391499904563,
"learning_rate": 1.1900352350748024e-07,
"logits/chosen": -1.907289743423462,
"logits/rejected": -1.9461581707000732,
"logps/chosen": -801.3639526367188,
"logps/rejected": -783.6478271484375,
"loss": 0.6698,
"rewards/accuracies": 0.5687500238418579,
"rewards/chosen": 0.376972496509552,
"rewards/margins": 0.02767539583146572,
"rewards/rejected": 0.34929707646369934,
"step": 1240
},
{
"epoch": 0.71,
"grad_norm": 2.665457863564576,
"learning_rate": 1.147825955301668e-07,
"logits/chosen": -1.9209585189819336,
"logits/rejected": -1.8920555114746094,
"logps/chosen": -636.7376708984375,
"logps/rejected": -571.822021484375,
"loss": 0.6563,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": 0.3121941089630127,
"rewards/margins": 0.13192227482795715,
"rewards/rejected": 0.18027189373970032,
"step": 1250
},
{
"epoch": 0.72,
"grad_norm": 2.885847358162881,
"learning_rate": 1.1061546440014963e-07,
"logits/chosen": -1.9306738376617432,
"logits/rejected": -1.9074119329452515,
"logps/chosen": -690.8770141601562,
"logps/rejected": -606.33447265625,
"loss": 0.6675,
"rewards/accuracies": 0.53125,
"rewards/chosen": 0.32359814643859863,
"rewards/margins": 0.07375864684581757,
"rewards/rejected": 0.24983951449394226,
"step": 1260
},
{
"epoch": 0.73,
"grad_norm": 2.3702768828862864,
"learning_rate": 1.0650378802913007e-07,
"logits/chosen": -2.008939743041992,
"logits/rejected": -1.970020055770874,
"logps/chosen": -635.5045776367188,
"logps/rejected": -561.628173828125,
"loss": 0.652,
"rewards/accuracies": 0.65625,
"rewards/chosen": 0.31716689467430115,
"rewards/margins": 0.16777855157852173,
"rewards/rejected": 0.14938834309577942,
"step": 1270
},
{
"epoch": 0.73,
"grad_norm": 2.600006257784113,
"learning_rate": 1.0244920226588597e-07,
"logits/chosen": -1.9430663585662842,
"logits/rejected": -1.941568374633789,
"logps/chosen": -684.0612182617188,
"logps/rejected": -652.2069091796875,
"loss": 0.6688,
"rewards/accuracies": 0.543749988079071,
"rewards/chosen": 0.2591310143470764,
"rewards/margins": 0.06264869123697281,
"rewards/rejected": 0.1964823305606842,
"step": 1280
},
{
"epoch": 0.74,
"grad_norm": 2.598645248672983,
"learning_rate": 9.845332024544245e-08,
"logits/chosen": -1.965990662574768,
"logits/rejected": -1.998303771018982,
"logps/chosen": -738.8136596679688,
"logps/rejected": -617.0636596679688,
"loss": 0.6533,
"rewards/accuracies": 0.65625,
"rewards/chosen": 0.31749868392944336,
"rewards/margins": 0.15900316834449768,
"rewards/rejected": 0.1584954857826233,
"step": 1290
},
{
"epoch": 0.74,
"grad_norm": 2.600416268107498,
"learning_rate": 9.451773174727873e-08,
"logits/chosen": -1.9850645065307617,
"logits/rejected": -1.9690158367156982,
"logps/chosen": -758.544921875,
"logps/rejected": -651.073486328125,
"loss": 0.6358,
"rewards/accuracies": 0.6812499761581421,
"rewards/chosen": 0.35706740617752075,
"rewards/margins": 0.15773825347423553,
"rewards/rejected": 0.1993291676044464,
"step": 1300
},
{
"epoch": 0.74,
"eval_logits/chosen": -2.0015413761138916,
"eval_logits/rejected": -1.9446992874145508,
"eval_logps/chosen": -690.077880859375,
"eval_logps/rejected": -632.544921875,
"eval_loss": 0.6533253192901611,
"eval_rewards/accuracies": 0.6073718070983887,
"eval_rewards/chosen": 0.33634501695632935,
"eval_rewards/margins": 0.13271409273147583,
"eval_rewards/rejected": 0.2036309689283371,
"eval_runtime": 833.2952,
"eval_samples_per_second": 14.942,
"eval_steps_per_second": 0.468,
"step": 1300
},
{
"epoch": 0.75,
"grad_norm": 2.4248781446367054,
"learning_rate": 9.064400256282755e-08,
"logits/chosen": -1.9554182291030884,
"logits/rejected": -1.8983566761016846,
"logps/chosen": -636.9713134765625,
"logps/rejected": -603.3074951171875,
"loss": 0.6436,
"rewards/accuracies": 0.581250011920929,
"rewards/chosen": 0.26512694358825684,
"rewards/margins": 0.08946672827005386,
"rewards/rejected": 0.17566025257110596,
"step": 1310
},
{
"epoch": 0.75,
"grad_norm": 3.090428157345016,
"learning_rate": 8.68336738725177e-08,
"logits/chosen": -2.043494939804077,
"logits/rejected": -1.9082825183868408,
"logps/chosen": -733.5816650390625,
"logps/rejected": -669.7770385742188,
"loss": 0.667,
"rewards/accuracies": 0.5874999761581421,
"rewards/chosen": 0.37750545144081116,
"rewards/margins": 0.14476510882377625,
"rewards/rejected": 0.23274032771587372,
"step": 1320
},
{
"epoch": 0.76,
"grad_norm": 3.1532123855322376,
"learning_rate": 8.308826163260852e-08,
"logits/chosen": -1.9862855672836304,
"logits/rejected": -1.9692624807357788,
"logps/chosen": -690.51708984375,
"logps/rejected": -675.4782104492188,
"loss": 0.6531,
"rewards/accuracies": 0.643750011920929,
"rewards/chosen": 0.3360420763492584,
"rewards/margins": 0.157058984041214,
"rewards/rejected": 0.17898306250572205,
"step": 1330
},
{
"epoch": 0.77,
"grad_norm": 2.629746362223527,
"learning_rate": 7.940925597206052e-08,
"logits/chosen": -2.007011890411377,
"logits/rejected": -1.9624780416488647,
"logps/chosen": -647.2005615234375,
"logps/rejected": -650.79052734375,
"loss": 0.6582,
"rewards/accuracies": 0.612500011920929,
"rewards/chosen": 0.35882705450057983,
"rewards/margins": 0.11791452020406723,
"rewards/rejected": 0.24091258645057678,
"step": 1340
},
{
"epoch": 0.77,
"grad_norm": 2.7296472904697513,
"learning_rate": 7.579812059968014e-08,
"logits/chosen": -2.0613789558410645,
"logits/rejected": -2.0162453651428223,
"logps/chosen": -620.8883666992188,
"logps/rejected": -658.7117309570312,
"loss": 0.6554,
"rewards/accuracies": 0.606249988079071,
"rewards/chosen": 0.3870480954647064,
"rewards/margins": 0.126990407705307,
"rewards/rejected": 0.2600576877593994,
"step": 1350
},
{
"epoch": 0.78,
"grad_norm": 2.763125897839002,
"learning_rate": 7.225629222177715e-08,
"logits/chosen": -1.943429946899414,
"logits/rejected": -1.9874225854873657,
"logps/chosen": -738.5694580078125,
"logps/rejected": -762.2120361328125,
"loss": 0.649,
"rewards/accuracies": 0.543749988079071,
"rewards/chosen": 0.3255593180656433,
"rewards/margins": 0.08112556487321854,
"rewards/rejected": 0.24443373084068298,
"step": 1360
},
{
"epoch": 0.78,
"grad_norm": 2.5446108148638498,
"learning_rate": 6.878517997056457e-08,
"logits/chosen": -1.9231475591659546,
"logits/rejected": -1.9031331539154053,
"logps/chosen": -625.0635986328125,
"logps/rejected": -623.573486328125,
"loss": 0.6469,
"rewards/accuracies": 0.574999988079071,
"rewards/chosen": 0.32645702362060547,
"rewards/margins": 0.11162126064300537,
"rewards/rejected": 0.21483579277992249,
"step": 1370
},
{
"epoch": 0.79,
"grad_norm": 3.5073757146061677,
"learning_rate": 6.538616484352902e-08,
"logits/chosen": -1.910323143005371,
"logits/rejected": -1.9373595714569092,
"logps/chosen": -732.39697265625,
"logps/rejected": -635.9688110351562,
"loss": 0.666,
"rewards/accuracies": 0.59375,
"rewards/chosen": 0.3115167021751404,
"rewards/margins": 0.10787830501794815,
"rewards/rejected": 0.20363843441009521,
"step": 1380
},
{
"epoch": 0.79,
"grad_norm": 2.859757387740054,
"learning_rate": 6.206059915399495e-08,
"logits/chosen": -2.0179154872894287,
"logits/rejected": -1.9656012058258057,
"logps/chosen": -815.8009033203125,
"logps/rejected": -609.6783447265625,
"loss": 0.6424,
"rewards/accuracies": 0.581250011920929,
"rewards/chosen": 0.3753044605255127,
"rewards/margins": 0.12460635602474213,
"rewards/rejected": 0.25069814920425415,
"step": 1390
},
{
"epoch": 0.8,
"grad_norm": 2.8109306032090764,
"learning_rate": 5.88098059931004e-08,
"logits/chosen": -1.9886192083358765,
"logits/rejected": -1.8791803121566772,
"logps/chosen": -670.867919921875,
"logps/rejected": -652.6749877929688,
"loss": 0.6473,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": 0.31432944536209106,
"rewards/margins": 0.17133426666259766,
"rewards/rejected": 0.1429951936006546,
"step": 1400
},
{
"epoch": 0.8,
"eval_logits/chosen": -1.9620791673660278,
"eval_logits/rejected": -1.9071754217147827,
"eval_logps/chosen": -689.9299926757812,
"eval_logps/rejected": -632.6980590820312,
"eval_loss": 0.6527881622314453,
"eval_rewards/accuracies": 0.6080127954483032,
"eval_rewards/chosen": 0.337823748588562,
"eval_rewards/margins": 0.1357237845659256,
"eval_rewards/rejected": 0.2020999938249588,
"eval_runtime": 811.1011,
"eval_samples_per_second": 15.351,
"eval_steps_per_second": 0.481,
"step": 1400
},
{
"epoch": 0.81,
"grad_norm": 2.9339811794947375,
"learning_rate": 5.563507870339962e-08,
"logits/chosen": -1.9552795886993408,
"logits/rejected": -1.885912537574768,
"logps/chosen": -820.3865356445312,
"logps/rejected": -724.3788452148438,
"loss": 0.6536,
"rewards/accuracies": 0.6499999761581421,
"rewards/chosen": 0.40686970949172974,
"rewards/margins": 0.22154918313026428,
"rewards/rejected": 0.18532049655914307,
"step": 1410
},
{
"epoch": 0.81,
"grad_norm": 3.0504864373903873,
"learning_rate": 5.25376803643007e-08,
"logits/chosen": -2.018120050430298,
"logits/rejected": -1.973238229751587,
"logps/chosen": -761.1993408203125,
"logps/rejected": -684.3026733398438,
"loss": 0.6609,
"rewards/accuracies": 0.5687500238418579,
"rewards/chosen": 0.32909178733825684,
"rewards/margins": 0.12167789787054062,
"rewards/rejected": 0.20741388201713562,
"step": 1420
},
{
"epoch": 0.82,
"grad_norm": 3.300789834284616,
"learning_rate": 4.9518843289544007e-08,
"logits/chosen": -1.9833561182022095,
"logits/rejected": -1.9334217309951782,
"logps/chosen": -674.072265625,
"logps/rejected": -636.847412109375,
"loss": 0.666,
"rewards/accuracies": 0.625,
"rewards/chosen": 0.3223080039024353,
"rewards/margins": 0.14498993754386902,
"rewards/rejected": 0.17731806635856628,
"step": 1430
},
{
"epoch": 0.82,
"grad_norm": 2.6410110326405314,
"learning_rate": 4.6579768536920695e-08,
"logits/chosen": -2.024742603302002,
"logits/rejected": -1.9354236125946045,
"logps/chosen": -717.5321655273438,
"logps/rejected": -710.6961669921875,
"loss": 0.6625,
"rewards/accuracies": 0.612500011920929,
"rewards/chosen": 0.36842697858810425,
"rewards/margins": 0.11848391592502594,
"rewards/rejected": 0.2499430626630783,
"step": 1440
},
{
"epoch": 0.83,
"grad_norm": 3.2793314877461306,
"learning_rate": 4.372162543042623e-08,
"logits/chosen": -2.0009231567382812,
"logits/rejected": -1.942253828048706,
"logps/chosen": -727.9117431640625,
"logps/rejected": -638.6812744140625,
"loss": 0.6507,
"rewards/accuracies": 0.612500011920929,
"rewards/chosen": 0.40108004212379456,
"rewards/margins": 0.1539073884487152,
"rewards/rejected": 0.24717263877391815,
"step": 1450
},
{
"epoch": 0.83,
"grad_norm": 2.7921989940115157,
"learning_rate": 4.094555109503983e-08,
"logits/chosen": -1.9503885507583618,
"logits/rejected": -1.8943992853164673,
"logps/chosen": -737.1885375976562,
"logps/rejected": -671.646484375,
"loss": 0.6543,
"rewards/accuracies": 0.612500011920929,
"rewards/chosen": 0.3580954670906067,
"rewards/margins": 0.12970301508903503,
"rewards/rejected": 0.22839240729808807,
"step": 1460
},
{
"epoch": 0.84,
"grad_norm": 2.7639464701150445,
"learning_rate": 3.825265000431424e-08,
"logits/chosen": -1.8646436929702759,
"logits/rejected": -1.8763723373413086,
"logps/chosen": -699.7142333984375,
"logps/rejected": -632.5223388671875,
"loss": 0.6615,
"rewards/accuracies": 0.6499999761581421,
"rewards/chosen": 0.37401726841926575,
"rewards/margins": 0.1618153154850006,
"rewards/rejected": 0.21220192313194275,
"step": 1470
},
{
"epoch": 0.85,
"grad_norm": 2.995425337211788,
"learning_rate": 3.5643993540955844e-08,
"logits/chosen": -1.9095637798309326,
"logits/rejected": -1.8548505306243896,
"logps/chosen": -750.3823852539062,
"logps/rejected": -683.6522216796875,
"loss": 0.6587,
"rewards/accuracies": 0.581250011920929,
"rewards/chosen": 0.37580543756484985,
"rewards/margins": 0.0922856256365776,
"rewards/rejected": 0.28351980447769165,
"step": 1480
},
{
"epoch": 0.85,
"grad_norm": 2.6756247012610697,
"learning_rate": 3.312061957057061e-08,
"logits/chosen": -1.9792400598526,
"logits/rejected": -1.9158741235733032,
"logps/chosen": -642.9052124023438,
"logps/rejected": -682.2804565429688,
"loss": 0.6462,
"rewards/accuracies": 0.5874999761581421,
"rewards/chosen": 0.2730144262313843,
"rewards/margins": 0.13844364881515503,
"rewards/rejected": 0.13457079231739044,
"step": 1490
},
{
"epoch": 0.86,
"grad_norm": 2.5306381537983658,
"learning_rate": 3.0683532028744184e-08,
"logits/chosen": -1.9723201990127563,
"logits/rejected": -1.8639072179794312,
"logps/chosen": -649.2698364257812,
"logps/rejected": -595.0311889648438,
"loss": 0.6447,
"rewards/accuracies": 0.6312500238418579,
"rewards/chosen": 0.3062586486339569,
"rewards/margins": 0.15073780715465546,
"rewards/rejected": 0.15552084147930145,
"step": 1500
},
{
"epoch": 0.86,
"eval_logits/chosen": -1.9781183004379272,
"eval_logits/rejected": -1.9226415157318115,
"eval_logps/chosen": -691.5005493164062,
"eval_logps/rejected": -634.215576171875,
"eval_loss": 0.6525518298149109,
"eval_rewards/accuracies": 0.6080127954483032,
"eval_rewards/chosen": 0.3221181333065033,
"eval_rewards/margins": 0.13519282639026642,
"eval_rewards/rejected": 0.18692529201507568,
"eval_runtime": 833.3431,
"eval_samples_per_second": 14.941,
"eval_steps_per_second": 0.468,
"step": 1500
},
{
"epoch": 0.86,
"grad_norm": 2.4794478716281474,
"learning_rate": 2.8333700521621955e-08,
"logits/chosen": -1.9369399547576904,
"logits/rejected": -1.915985345840454,
"logps/chosen": -723.750732421875,
"logps/rejected": -571.1014404296875,
"loss": 0.6534,
"rewards/accuracies": 0.5874999761581421,
"rewards/chosen": 0.34681373834609985,
"rewards/margins": 0.15279236435890198,
"rewards/rejected": 0.1940213292837143,
"step": 1510
},
{
"epoch": 0.87,
"grad_norm": 2.9576634978157657,
"learning_rate": 2.6072059940146772e-08,
"logits/chosen": -2.0135951042175293,
"logits/rejected": -1.9804418087005615,
"logps/chosen": -681.3564453125,
"logps/rejected": -626.4763793945312,
"loss": 0.6443,
"rewards/accuracies": 0.5625,
"rewards/chosen": 0.3405224680900574,
"rewards/margins": 0.15179835259914398,
"rewards/rejected": 0.1887241005897522,
"step": 1520
},
{
"epoch": 0.87,
"grad_norm": 3.126710724613424,
"learning_rate": 2.3899510088108587e-08,
"logits/chosen": -1.9243135452270508,
"logits/rejected": -1.8939940929412842,
"logps/chosen": -654.2572021484375,
"logps/rejected": -603.611328125,
"loss": 0.648,
"rewards/accuracies": 0.606249988079071,
"rewards/chosen": 0.2771258056163788,
"rewards/margins": 0.08931798487901688,
"rewards/rejected": 0.1878078281879425,
"step": 1530
},
{
"epoch": 0.88,
"grad_norm": 3.3520779361226216,
"learning_rate": 2.1816915324153334e-08,
"logits/chosen": -2.021639347076416,
"logits/rejected": -1.9558565616607666,
"logps/chosen": -772.6861572265625,
"logps/rejected": -712.33544921875,
"loss": 0.6531,
"rewards/accuracies": 0.65625,
"rewards/chosen": 0.3145074248313904,
"rewards/margins": 0.17986731231212616,
"rewards/rejected": 0.13464009761810303,
"step": 1540
},
{
"epoch": 0.89,
"grad_norm": 2.8164037709731296,
"learning_rate": 1.9825104217894018e-08,
"logits/chosen": -1.9654910564422607,
"logits/rejected": -1.8914296627044678,
"logps/chosen": -652.0030517578125,
"logps/rejected": -626.3556518554688,
"loss": 0.6569,
"rewards/accuracies": 0.5874999761581421,
"rewards/chosen": 0.3254419267177582,
"rewards/margins": 0.10867112874984741,
"rewards/rejected": 0.21677079796791077,
"step": 1550
},
{
"epoch": 0.89,
"grad_norm": 2.6118898735650125,
"learning_rate": 1.7924869220260626e-08,
"logits/chosen": -1.9689853191375732,
"logits/rejected": -1.9209016561508179,
"logps/chosen": -692.149658203125,
"logps/rejected": -639.4564819335938,
"loss": 0.6507,
"rewards/accuracies": 0.581250011920929,
"rewards/chosen": 0.3543575406074524,
"rewards/margins": 0.1561117321252823,
"rewards/rejected": 0.1982457935810089,
"step": 1560
},
{
"epoch": 0.9,
"grad_norm": 2.704102602095005,
"learning_rate": 1.6116966348220046e-08,
"logits/chosen": -1.9350318908691406,
"logits/rejected": -1.9500646591186523,
"logps/chosen": -677.9044189453125,
"logps/rejected": -582.033447265625,
"loss": 0.6493,
"rewards/accuracies": 0.65625,
"rewards/chosen": 0.40810251235961914,
"rewards/margins": 0.22535303235054016,
"rewards/rejected": 0.18274952471256256,
"step": 1570
},
{
"epoch": 0.9,
"grad_norm": 2.0946434953536768,
"learning_rate": 1.4402114883991318e-08,
"logits/chosen": -2.010521411895752,
"logits/rejected": -1.9158645868301392,
"logps/chosen": -664.4788208007812,
"logps/rejected": -566.4072875976562,
"loss": 0.6531,
"rewards/accuracies": 0.5625,
"rewards/chosen": 0.2843988537788391,
"rewards/margins": 0.14584381878376007,
"rewards/rejected": 0.13855502009391785,
"step": 1580
},
{
"epoch": 0.91,
"grad_norm": 2.9882114113782885,
"learning_rate": 1.2780997088875866e-08,
"logits/chosen": -1.9896526336669922,
"logits/rejected": -1.9169244766235352,
"logps/chosen": -723.8212280273438,
"logps/rejected": -595.9112548828125,
"loss": 0.6595,
"rewards/accuracies": 0.625,
"rewards/chosen": 0.31389322876930237,
"rewards/margins": 0.14465804398059845,
"rewards/rejected": 0.16923515498638153,
"step": 1590
},
{
"epoch": 0.91,
"grad_norm": 2.597069789857251,
"learning_rate": 1.125425793181703e-08,
"logits/chosen": -2.011854410171509,
"logits/rejected": -1.9551817178726196,
"logps/chosen": -690.0392456054688,
"logps/rejected": -658.7679443359375,
"loss": 0.6546,
"rewards/accuracies": 0.6312500238418579,
"rewards/chosen": 0.3487124741077423,
"rewards/margins": 0.15366964042186737,
"rewards/rejected": 0.19504281878471375,
"step": 1600
},
{
"epoch": 0.91,
"eval_logits/chosen": -1.9683998823165894,
"eval_logits/rejected": -1.913440465927124,
"eval_logps/chosen": -690.6824340820312,
"eval_logps/rejected": -633.5018310546875,
"eval_loss": 0.6524997353553772,
"eval_rewards/accuracies": 0.6073718070983887,
"eval_rewards/chosen": 0.3302985429763794,
"eval_rewards/margins": 0.13623538613319397,
"eval_rewards/rejected": 0.19406315684318542,
"eval_runtime": 811.6062,
"eval_samples_per_second": 15.341,
"eval_steps_per_second": 0.481,
"step": 1600
},
{
"epoch": 0.92,
"grad_norm": 2.686772952442727,
"learning_rate": 9.822504832796036e-09,
"logits/chosen": -1.9362843036651611,
"logits/rejected": -1.8773858547210693,
"logps/chosen": -756.9738159179688,
"logps/rejected": -635.9459838867188,
"loss": 0.6479,
"rewards/accuracies": 0.6312500238418579,
"rewards/chosen": 0.3329898416996002,
"rewards/margins": 0.18678471446037292,
"rewards/rejected": 0.1462051421403885,
"step": 1610
},
{
"epoch": 0.93,
"grad_norm": 2.6447110973275665,
"learning_rate": 8.48630742116746e-09,
"logits/chosen": -1.9655792713165283,
"logits/rejected": -1.8795562982559204,
"logps/chosen": -771.013671875,
"logps/rejected": -664.7561645507812,
"loss": 0.6623,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": 0.33320263028144836,
"rewards/margins": 0.15516476333141327,
"rewards/rejected": 0.1780378520488739,
"step": 1620
},
{
"epoch": 0.93,
"grad_norm": 2.8021224333165144,
"learning_rate": 7.246197309029617e-09,
"logits/chosen": -1.9644882678985596,
"logits/rejected": -1.9270439147949219,
"logps/chosen": -812.21337890625,
"logps/rejected": -742.98095703125,
"loss": 0.6634,
"rewards/accuracies": 0.6312500238418579,
"rewards/chosen": 0.3916359543800354,
"rewards/margins": 0.15279091894626617,
"rewards/rejected": 0.2388450652360916,
"step": 1630
},
{
"epoch": 0.94,
"grad_norm": 2.7924440856518578,
"learning_rate": 6.102667879720164e-09,
"logits/chosen": -2.016839027404785,
"logits/rejected": -1.9747447967529297,
"logps/chosen": -696.322265625,
"logps/rejected": -629.0064697265625,
"loss": 0.6598,
"rewards/accuracies": 0.5562499761581421,
"rewards/chosen": 0.310738742351532,
"rewards/margins": 0.11902445554733276,
"rewards/rejected": 0.19171428680419922,
"step": 1640
},
{
"epoch": 0.94,
"grad_norm": 3.1884351048854946,
"learning_rate": 5.056174091521509e-09,
"logits/chosen": -2.0210442543029785,
"logits/rejected": -1.9764583110809326,
"logps/chosen": -786.7182006835938,
"logps/rejected": -743.8095703125,
"loss": 0.6557,
"rewards/accuracies": 0.625,
"rewards/chosen": 0.35495525598526,
"rewards/margins": 0.17099758982658386,
"rewards/rejected": 0.18395769596099854,
"step": 1650
},
{
"epoch": 0.95,
"grad_norm": 2.8161297313177394,
"learning_rate": 4.107132296653548e-09,
"logits/chosen": -1.9870693683624268,
"logits/rejected": -1.916751503944397,
"logps/chosen": -682.517822265625,
"logps/rejected": -588.7493896484375,
"loss": 0.6492,
"rewards/accuracies": 0.643750011920929,
"rewards/chosen": 0.37071341276168823,
"rewards/margins": 0.136337548494339,
"rewards/rejected": 0.23437583446502686,
"step": 1660
},
{
"epoch": 0.95,
"grad_norm": 2.6636522324368266,
"learning_rate": 3.255920075626084e-09,
"logits/chosen": -1.8954880237579346,
"logits/rejected": -1.9022176265716553,
"logps/chosen": -623.636962890625,
"logps/rejected": -596.3826293945312,
"loss": 0.6712,
"rewards/accuracies": 0.6187499761581421,
"rewards/chosen": 0.3262075185775757,
"rewards/margins": 0.11597733199596405,
"rewards/rejected": 0.21023023128509521,
"step": 1670
},
{
"epoch": 0.96,
"grad_norm": 2.7200883225079395,
"learning_rate": 2.5028760870168253e-09,
"logits/chosen": -1.9352178573608398,
"logits/rejected": -1.9500812292099,
"logps/chosen": -690.5252685546875,
"logps/rejected": -673.8720703125,
"loss": 0.6625,
"rewards/accuracies": 0.59375,
"rewards/chosen": 0.28286224603652954,
"rewards/margins": 0.084191232919693,
"rewards/rejected": 0.19867102801799774,
"step": 1680
},
{
"epoch": 0.97,
"grad_norm": 4.497441891242338,
"learning_rate": 1.8482999327343597e-09,
"logits/chosen": -2.021484613418579,
"logits/rejected": -1.9103187322616577,
"logps/chosen": -767.1332397460938,
"logps/rejected": -708.3283081054688,
"loss": 0.659,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": 0.40307435393333435,
"rewards/margins": 0.1727377474308014,
"rewards/rejected": 0.23033663630485535,
"step": 1690
},
{
"epoch": 0.97,
"grad_norm": 3.084782840541729,
"learning_rate": 1.2924520388204462e-09,
"logits/chosen": -1.9084007740020752,
"logits/rejected": -1.9263513088226318,
"logps/chosen": -614.9078369140625,
"logps/rejected": -623.568603515625,
"loss": 0.6725,
"rewards/accuracies": 0.512499988079071,
"rewards/chosen": 0.26621872186660767,
"rewards/margins": 0.04896958917379379,
"rewards/rejected": 0.21724911034107208,
"step": 1700
},
{
"epoch": 0.97,
"eval_logits/chosen": -1.9645206928253174,
"eval_logits/rejected": -1.9097665548324585,
"eval_logps/chosen": -690.5892333984375,
"eval_logps/rejected": -633.4114990234375,
"eval_loss": 0.6524692177772522,
"eval_rewards/accuracies": 0.6073718070983887,
"eval_rewards/chosen": 0.33123117685317993,
"eval_rewards/margins": 0.13626568019390106,
"eval_rewards/rejected": 0.19496554136276245,
"eval_runtime": 831.8192,
"eval_samples_per_second": 14.968,
"eval_steps_per_second": 0.469,
"step": 1700
},
{
"epoch": 0.98,
"grad_norm": 2.668345364222901,
"learning_rate": 8.355535518381751e-10,
"logits/chosen": -1.9214885234832764,
"logits/rejected": -1.947447419166565,
"logps/chosen": -736.5780029296875,
"logps/rejected": -594.7259521484375,
"loss": 0.6583,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": 0.33075276017189026,
"rewards/margins": 0.11735783517360687,
"rewards/rejected": 0.2133948802947998,
"step": 1710
},
{
"epoch": 0.98,
"grad_norm": 3.4262928984719894,
"learning_rate": 4.77786250887846e-10,
"logits/chosen": -1.8850606679916382,
"logits/rejected": -1.882394552230835,
"logps/chosen": -706.0784912109375,
"logps/rejected": -678.0963134765625,
"loss": 0.6542,
"rewards/accuracies": 0.59375,
"rewards/chosen": 0.3584180772304535,
"rewards/margins": 0.1582474410533905,
"rewards/rejected": 0.20017066597938538,
"step": 1720
},
{
"epoch": 0.99,
"grad_norm": 2.9771754271252857,
"learning_rate": 2.1929247528540418e-10,
"logits/chosen": -2.0322813987731934,
"logits/rejected": -1.9513972997665405,
"logps/chosen": -745.3729248046875,
"logps/rejected": -708.0712280273438,
"loss": 0.6572,
"rewards/accuracies": 0.6499999761581421,
"rewards/chosen": 0.36603301763534546,
"rewards/margins": 0.15421339869499207,
"rewards/rejected": 0.21181955933570862,
"step": 1730
},
{
"epoch": 0.99,
"grad_norm": 2.8018132635690183,
"learning_rate": 6.017506793212779e-11,
"logits/chosen": -1.9503310918807983,
"logits/rejected": -1.9407823085784912,
"logps/chosen": -713.3302612304688,
"logps/rejected": -686.0525512695312,
"loss": 0.6631,
"rewards/accuracies": 0.625,
"rewards/chosen": 0.31906238198280334,
"rewards/margins": 0.11456944793462753,
"rewards/rejected": 0.20449292659759521,
"step": 1740
},
{
"epoch": 1.0,
"grad_norm": 3.3199203561320076,
"learning_rate": 4.973343980252398e-13,
"logits/chosen": -1.9441578388214111,
"logits/rejected": -1.8932502269744873,
"logps/chosen": -728.4362182617188,
"logps/rejected": -602.8780517578125,
"loss": 0.6488,
"rewards/accuracies": 0.6875,
"rewards/chosen": 0.2982894778251648,
"rewards/margins": 0.17425718903541565,
"rewards/rejected": 0.12403228133916855,
"step": 1750
},
{
"epoch": 1.0,
"step": 1751,
"total_flos": 0.0,
"train_loss": 0.6642669952984608,
"train_runtime": 38056.9659,
"train_samples_per_second": 2.944,
"train_steps_per_second": 0.046
}
],
"logging_steps": 10,
"max_steps": 1751,
"num_input_tokens_seen": 0,
"num_train_epochs": 1,
"save_steps": 100,
"total_flos": 0.0,
"train_batch_size": 8,
"trial_name": null,
"trial_params": null
}