zephyr-dpop-qlora-gpt4-5e-7 / trainer_state.json
just1nseo's picture
Model save
8bb065b verified
raw
history blame
30.1 kB
{
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 1.0,
"eval_steps": 100,
"global_step": 355,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"dpo_losses": 0.6931471824645996,
"epoch": 0.0,
"grad_norm": 2.1563079383916617,
"learning_rate": 1.3888888888888887e-08,
"logits/chosen": -2.804708957672119,
"logits/rejected": -2.8150453567504883,
"logps/chosen": -217.97438049316406,
"logps/rejected": -216.58865356445312,
"loss": 0.6931,
"positive_losses": 0.0,
"rewards/accuracies": 0.0,
"rewards/chosen": 0.0,
"rewards/margins": 0.0,
"rewards/margins_max": 0.0,
"rewards/margins_min": 0.0,
"rewards/margins_std": 0.0,
"rewards/rejected": 0.0,
"step": 1
},
{
"dpo_losses": 0.6929957270622253,
"epoch": 0.03,
"grad_norm": 29.018601338328015,
"learning_rate": 1.3888888888888888e-07,
"logits/chosen": -2.884633779525757,
"logits/rejected": -2.7994985580444336,
"logps/chosen": -366.7369689941406,
"logps/rejected": -275.4845275878906,
"loss": 0.7009,
"positive_losses": 0.08276741951704025,
"rewards/accuracies": 0.4027777910232544,
"rewards/chosen": -2.229649544460699e-05,
"rewards/margins": 0.00030543722095899284,
"rewards/margins_max": 0.003608349245041609,
"rewards/margins_min": -0.0025843861512839794,
"rewards/margins_std": 0.002678285352885723,
"rewards/rejected": -0.00032773372367955744,
"step": 10
},
{
"dpo_losses": 0.6929633021354675,
"epoch": 0.06,
"grad_norm": 11.724263026129163,
"learning_rate": 2.7777777777777776e-07,
"logits/chosen": -2.7479777336120605,
"logits/rejected": -2.6943278312683105,
"logps/chosen": -329.0453186035156,
"logps/rejected": -216.6789093017578,
"loss": 0.6981,
"positive_losses": 0.05849536508321762,
"rewards/accuracies": 0.5625,
"rewards/chosen": 0.0010186245199292898,
"rewards/margins": 0.00036991172237321734,
"rewards/margins_max": 0.003288195002824068,
"rewards/margins_min": -0.002696308773010969,
"rewards/margins_std": 0.0026607036124914885,
"rewards/rejected": 0.0006487127975560725,
"step": 20
},
{
"dpo_losses": 0.6923447251319885,
"epoch": 0.08,
"grad_norm": 2.313455126668315,
"learning_rate": 4.1666666666666667e-07,
"logits/chosen": -2.835400342941284,
"logits/rejected": -2.7531635761260986,
"logps/chosen": -328.7791442871094,
"logps/rejected": -232.63723754882812,
"loss": 0.6929,
"positive_losses": 0.0025131225120276213,
"rewards/accuracies": 0.6499999761581421,
"rewards/chosen": 0.005060128401964903,
"rewards/margins": 0.0016096157487481833,
"rewards/margins_max": 0.006041892804205418,
"rewards/margins_min": -0.0021330646704882383,
"rewards/margins_std": 0.003625147510319948,
"rewards/rejected": 0.003450512420386076,
"step": 30
},
{
"dpo_losses": 0.6913660764694214,
"epoch": 0.11,
"grad_norm": 1.9684885850555545,
"learning_rate": 4.998060489154965e-07,
"logits/chosen": -2.813957929611206,
"logits/rejected": -2.7604265213012695,
"logps/chosen": -284.87249755859375,
"logps/rejected": -227.00289916992188,
"loss": 0.6915,
"positive_losses": 0.0,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": 0.009648630395531654,
"rewards/margins": 0.003571807174012065,
"rewards/margins_max": 0.008924934081733227,
"rewards/margins_min": -0.001162365428172052,
"rewards/margins_std": 0.004516431130468845,
"rewards/rejected": 0.006076823454350233,
"step": 40
},
{
"dpo_losses": 0.6890588998794556,
"epoch": 0.14,
"grad_norm": 2.3506499385956174,
"learning_rate": 4.976275538042932e-07,
"logits/chosen": -2.8134660720825195,
"logits/rejected": -2.73073148727417,
"logps/chosen": -316.21136474609375,
"logps/rejected": -233.4212646484375,
"loss": 0.689,
"positive_losses": 0.0,
"rewards/accuracies": 0.8374999761581421,
"rewards/chosen": 0.015740128234028816,
"rewards/margins": 0.008210490457713604,
"rewards/margins_max": 0.017676424235105515,
"rewards/margins_min": 0.0003599710180424154,
"rewards/margins_std": 0.008204594254493713,
"rewards/rejected": 0.00752963637933135,
"step": 50
},
{
"dpo_losses": 0.6860449910163879,
"epoch": 0.17,
"grad_norm": 2.3197802822711284,
"learning_rate": 4.930493069997119e-07,
"logits/chosen": -2.7519192695617676,
"logits/rejected": -2.7033615112304688,
"logps/chosen": -342.56121826171875,
"logps/rejected": -263.17974853515625,
"loss": 0.6864,
"positive_losses": 0.0,
"rewards/accuracies": 0.8999999761581421,
"rewards/chosen": 0.02219192497432232,
"rewards/margins": 0.01428920030593872,
"rewards/margins_max": 0.02703510783612728,
"rewards/margins_min": 0.0032699487637728453,
"rewards/margins_std": 0.010757345706224442,
"rewards/rejected": 0.007902727462351322,
"step": 60
},
{
"dpo_losses": 0.6826337575912476,
"epoch": 0.2,
"grad_norm": 1.81612559785793,
"learning_rate": 4.861156761634013e-07,
"logits/chosen": -2.8005824089050293,
"logits/rejected": -2.7131030559539795,
"logps/chosen": -359.32147216796875,
"logps/rejected": -235.8487091064453,
"loss": 0.6831,
"positive_losses": 0.0,
"rewards/accuracies": 0.949999988079071,
"rewards/chosen": 0.030173394829034805,
"rewards/margins": 0.021212318912148476,
"rewards/margins_max": 0.04077567160129547,
"rewards/margins_min": 0.003378701163455844,
"rewards/margins_std": 0.01716596446931362,
"rewards/rejected": 0.008961074985563755,
"step": 70
},
{
"dpo_losses": 0.6795681715011597,
"epoch": 0.23,
"grad_norm": 2.0311863815908806,
"learning_rate": 4.768938549177392e-07,
"logits/chosen": -2.8440804481506348,
"logits/rejected": -2.7800514698028564,
"logps/chosen": -328.42352294921875,
"logps/rejected": -286.81768798828125,
"loss": 0.6796,
"positive_losses": 0.0,
"rewards/accuracies": 0.949999988079071,
"rewards/chosen": 0.0342494361102581,
"rewards/margins": 0.027495408430695534,
"rewards/margins_max": 0.05417575687170029,
"rewards/margins_min": 0.004369550384581089,
"rewards/margins_std": 0.022978123277425766,
"rewards/rejected": 0.006754028145223856,
"step": 80
},
{
"dpo_losses": 0.6761659383773804,
"epoch": 0.25,
"grad_norm": 2.441277633330898,
"learning_rate": 4.654732116743193e-07,
"logits/chosen": -2.784029722213745,
"logits/rejected": -2.700582981109619,
"logps/chosen": -334.9620056152344,
"logps/rejected": -198.4322967529297,
"loss": 0.6754,
"positive_losses": 0.0,
"rewards/accuracies": 0.9375,
"rewards/chosen": 0.03997080400586128,
"rewards/margins": 0.034528493881225586,
"rewards/margins_max": 0.07119405269622803,
"rewards/margins_min": 0.006210760213434696,
"rewards/margins_std": 0.030152523890137672,
"rewards/rejected": 0.00544231291860342,
"step": 90
},
{
"dpo_losses": 0.6722386479377747,
"epoch": 0.28,
"grad_norm": 2.6661097024380993,
"learning_rate": 4.519644235671752e-07,
"logits/chosen": -2.858572244644165,
"logits/rejected": -2.7656137943267822,
"logps/chosen": -341.40325927734375,
"logps/rejected": -263.1445007324219,
"loss": 0.6709,
"positive_losses": 0.0,
"rewards/accuracies": 0.9125000238418579,
"rewards/chosen": 0.04941926524043083,
"rewards/margins": 0.042630504816770554,
"rewards/margins_max": 0.08383534103631973,
"rewards/margins_min": 0.00657398859038949,
"rewards/margins_std": 0.03647482395172119,
"rewards/rejected": 0.00678876182064414,
"step": 100
},
{
"epoch": 0.28,
"eval_dpo_losses": 0.6899906396865845,
"eval_logits/chosen": -2.799083709716797,
"eval_logits/rejected": -2.7607357501983643,
"eval_logps/chosen": -283.0979919433594,
"eval_logps/rejected": -257.7293395996094,
"eval_loss": 0.7044196128845215,
"eval_positive_losses": 0.13342785835266113,
"eval_rewards/accuracies": 0.5830000042915344,
"eval_rewards/chosen": 0.014954007230699062,
"eval_rewards/margins": 0.0064581576734781265,
"eval_rewards/margins_max": 0.04341236874461174,
"eval_rewards/margins_min": -0.024124523624777794,
"eval_rewards/margins_std": 0.022434763610363007,
"eval_rewards/rejected": 0.008495849557220936,
"eval_runtime": 428.6351,
"eval_samples_per_second": 4.666,
"eval_steps_per_second": 0.292,
"step": 100
},
{
"dpo_losses": 0.6648027300834656,
"epoch": 0.31,
"grad_norm": 2.4903000527490997,
"learning_rate": 4.364984038837727e-07,
"logits/chosen": -2.8712093830108643,
"logits/rejected": -2.7597901821136475,
"logps/chosen": -384.63873291015625,
"logps/rejected": -286.42755126953125,
"loss": 0.6644,
"positive_losses": 0.0,
"rewards/accuracies": 0.9375,
"rewards/chosen": 0.06723800301551819,
"rewards/margins": 0.05809164047241211,
"rewards/margins_max": 0.11163143068552017,
"rewards/margins_min": 0.012343182228505611,
"rewards/margins_std": 0.04587184637784958,
"rewards/rejected": 0.009146357886493206,
"step": 110
},
{
"dpo_losses": 0.6599195599555969,
"epoch": 0.34,
"grad_norm": 2.0272270859977986,
"learning_rate": 4.1922503338800447e-07,
"logits/chosen": -2.8636832237243652,
"logits/rejected": -2.788536548614502,
"logps/chosen": -386.91357421875,
"logps/rejected": -265.55084228515625,
"loss": 0.6624,
"positive_losses": 0.0009745836141519248,
"rewards/accuracies": 0.949999988079071,
"rewards/chosen": 0.07831867039203644,
"rewards/margins": 0.0684511810541153,
"rewards/margins_max": 0.13835951685905457,
"rewards/margins_min": 0.014530710875988007,
"rewards/margins_std": 0.057229261845350266,
"rewards/rejected": 0.00986748468130827,
"step": 120
},
{
"dpo_losses": 0.6510322093963623,
"epoch": 0.37,
"grad_norm": 1.9824231979290536,
"learning_rate": 4.003117078299021e-07,
"logits/chosen": -2.8207156658172607,
"logits/rejected": -2.7434334754943848,
"logps/chosen": -395.66802978515625,
"logps/rejected": -300.08343505859375,
"loss": 0.6531,
"positive_losses": 0.0,
"rewards/accuracies": 0.9750000238418579,
"rewards/chosen": 0.09558561444282532,
"rewards/margins": 0.08716461062431335,
"rewards/margins_max": 0.1589736044406891,
"rewards/margins_min": 0.021759631112217903,
"rewards/margins_std": 0.06302757561206818,
"rewards/rejected": 0.008421008475124836,
"step": 130
},
{
"dpo_losses": 0.6501553654670715,
"epoch": 0.39,
"grad_norm": 1.7119399665574373,
"learning_rate": 3.799417157181075e-07,
"logits/chosen": -2.795180082321167,
"logits/rejected": -2.7392399311065674,
"logps/chosen": -363.2869567871094,
"logps/rejected": -269.78350830078125,
"loss": 0.6548,
"positive_losses": 0.0016069173580035567,
"rewards/accuracies": 0.9624999761581421,
"rewards/chosen": 0.09409914910793304,
"rewards/margins": 0.08933846652507782,
"rewards/margins_max": 0.172917902469635,
"rewards/margins_min": 0.012987839989364147,
"rewards/margins_std": 0.07409818470478058,
"rewards/rejected": 0.004760689102113247,
"step": 140
},
{
"dpo_losses": 0.6526910662651062,
"epoch": 0.42,
"grad_norm": 2.0809425241432886,
"learning_rate": 3.583124620760659e-07,
"logits/chosen": -2.8274710178375244,
"logits/rejected": -2.7291388511657715,
"logps/chosen": -313.82757568359375,
"logps/rejected": -213.179443359375,
"loss": 0.6516,
"positive_losses": 0.0,
"rewards/accuracies": 0.9624999761581421,
"rewards/chosen": 0.09023541212081909,
"rewards/margins": 0.08422856032848358,
"rewards/margins_max": 0.17488643527030945,
"rewards/margins_min": 0.01652991771697998,
"rewards/margins_std": 0.07335906475782394,
"rewards/rejected": 0.006006857845932245,
"step": 150
},
{
"dpo_losses": 0.6424547433853149,
"epoch": 0.45,
"grad_norm": 1.7536600530591888,
"learning_rate": 3.356335553954679e-07,
"logits/chosen": -2.7450592517852783,
"logits/rejected": -2.685469627380371,
"logps/chosen": -334.85797119140625,
"logps/rejected": -234.3365020751953,
"loss": 0.6447,
"positive_losses": 0.005527782253921032,
"rewards/accuracies": 0.9750000238418579,
"rewards/chosen": 0.10652638971805573,
"rewards/margins": 0.10596431791782379,
"rewards/margins_max": 0.20612521469593048,
"rewards/margins_min": 0.020822737365961075,
"rewards/margins_std": 0.08536245673894882,
"rewards/rejected": 0.0005620642914436758,
"step": 160
},
{
"dpo_losses": 0.6378639936447144,
"epoch": 0.48,
"grad_norm": 6.472916445339736,
"learning_rate": 3.121247763262235e-07,
"logits/chosen": -2.82389497756958,
"logits/rejected": -2.7423415184020996,
"logps/chosen": -363.29315185546875,
"logps/rejected": -295.8577575683594,
"loss": 0.6445,
"positive_losses": 0.00014963149442337453,
"rewards/accuracies": 0.8999999761581421,
"rewards/chosen": 0.12112192809581757,
"rewards/margins": 0.11697308719158173,
"rewards/margins_max": 0.24103912711143494,
"rewards/margins_min": 0.013107486069202423,
"rewards/margins_std": 0.10543470084667206,
"rewards/rejected": 0.00414883391931653,
"step": 170
},
{
"dpo_losses": 0.6460212469100952,
"epoch": 0.51,
"grad_norm": 2.067078450412541,
"learning_rate": 2.880139477883347e-07,
"logits/chosen": -2.793970823287964,
"logits/rejected": -2.705854892730713,
"logps/chosen": -337.114501953125,
"logps/rejected": -292.83624267578125,
"loss": 0.6403,
"positive_losses": 0.0,
"rewards/accuracies": 0.9125000238418579,
"rewards/chosen": 0.10858919471502304,
"rewards/margins": 0.098782978951931,
"rewards/margins_max": 0.19024236500263214,
"rewards/margins_min": 0.0153144970536232,
"rewards/margins_std": 0.07897777855396271,
"rewards/rejected": 0.009806220419704914,
"step": 180
},
{
"dpo_losses": 0.634928286075592,
"epoch": 0.54,
"grad_norm": 2.5683292981841586,
"learning_rate": 2.635347271463544e-07,
"logits/chosen": -2.793365001678467,
"logits/rejected": -2.6589195728302,
"logps/chosen": -346.8443908691406,
"logps/rejected": -237.68563842773438,
"loss": 0.6365,
"positive_losses": 0.0,
"rewards/accuracies": 0.987500011920929,
"rewards/chosen": 0.12526604533195496,
"rewards/margins": 0.12293746322393417,
"rewards/margins_max": 0.2294788807630539,
"rewards/margins_min": 0.023495811969041824,
"rewards/margins_std": 0.09407813102006912,
"rewards/rejected": 0.002328577684238553,
"step": 190
},
{
"dpo_losses": 0.6293594837188721,
"epoch": 0.56,
"grad_norm": 6.199913442684288,
"learning_rate": 2.3892434184240534e-07,
"logits/chosen": -2.862529993057251,
"logits/rejected": -2.7567837238311768,
"logps/chosen": -384.57025146484375,
"logps/rejected": -265.1017761230469,
"loss": 0.6375,
"positive_losses": 0.026654411107301712,
"rewards/accuracies": 0.9750000238418579,
"rewards/chosen": 0.14424578845500946,
"rewards/margins": 0.13493841886520386,
"rewards/margins_max": 0.2611827552318573,
"rewards/margins_min": 0.03425807133316994,
"rewards/margins_std": 0.10205696523189545,
"rewards/rejected": 0.009307395666837692,
"step": 200
},
{
"epoch": 0.56,
"eval_dpo_losses": 0.6862263083457947,
"eval_logits/chosen": -2.7717671394348145,
"eval_logits/rejected": -2.7334673404693604,
"eval_logps/chosen": -282.3363037109375,
"eval_logps/rejected": -257.78778076171875,
"eval_loss": 0.7538697123527527,
"eval_positive_losses": 0.6397833228111267,
"eval_rewards/accuracies": 0.5879999995231628,
"eval_rewards/chosen": 0.022570965811610222,
"eval_rewards/margins": 0.0146592753008008,
"eval_rewards/margins_max": 0.10217105597257614,
"eval_rewards/margins_min": -0.05958281084895134,
"eval_rewards/margins_std": 0.0534750372171402,
"eval_rewards/rejected": 0.007911693304777145,
"eval_runtime": 427.9807,
"eval_samples_per_second": 4.673,
"eval_steps_per_second": 0.292,
"step": 200
},
{
"dpo_losses": 0.6285057067871094,
"epoch": 0.59,
"grad_norm": 2.0332452218189423,
"learning_rate": 2.1442129043167873e-07,
"logits/chosen": -2.756385326385498,
"logits/rejected": -2.685586929321289,
"logps/chosen": -341.33837890625,
"logps/rejected": -256.91766357421875,
"loss": 0.6311,
"positive_losses": 0.0025708198081701994,
"rewards/accuracies": 0.9750000238418579,
"rewards/chosen": 0.13986453413963318,
"rewards/margins": 0.13733358681201935,
"rewards/margins_max": 0.27369070053100586,
"rewards/margins_min": 0.027926146984100342,
"rewards/margins_std": 0.11671511083841324,
"rewards/rejected": 0.0025309394113719463,
"step": 210
},
{
"dpo_losses": 0.621319055557251,
"epoch": 0.62,
"grad_norm": 2.027345866825123,
"learning_rate": 1.9026303129961048e-07,
"logits/chosen": -2.856189727783203,
"logits/rejected": -2.73331880569458,
"logps/chosen": -391.1002197265625,
"logps/rejected": -274.0777893066406,
"loss": 0.627,
"positive_losses": 0.0,
"rewards/accuracies": 0.9375,
"rewards/chosen": 0.15911123156547546,
"rewards/margins": 0.1526704728603363,
"rewards/margins_max": 0.2787471413612366,
"rewards/margins_min": 0.029059389606118202,
"rewards/margins_std": 0.11583348363637924,
"rewards/rejected": 0.006440767087042332,
"step": 220
},
{
"dpo_losses": 0.6225648522377014,
"epoch": 0.65,
"grad_norm": 2.1898711296381363,
"learning_rate": 1.6668368145931396e-07,
"logits/chosen": -2.8812718391418457,
"logits/rejected": -2.7513253688812256,
"logps/chosen": -387.0869140625,
"logps/rejected": -262.6962585449219,
"loss": 0.6199,
"positive_losses": 0.0,
"rewards/accuracies": 0.9750000238418579,
"rewards/chosen": 0.1572650671005249,
"rewards/margins": 0.15020257234573364,
"rewards/margins_max": 0.27498555183410645,
"rewards/margins_min": 0.032824836671352386,
"rewards/margins_std": 0.11001060903072357,
"rewards/rejected": 0.007062485907226801,
"step": 230
},
{
"dpo_losses": 0.636019766330719,
"epoch": 0.68,
"grad_norm": 1.8191230391360291,
"learning_rate": 1.4391174773015834e-07,
"logits/chosen": -2.807938814163208,
"logits/rejected": -2.716116428375244,
"logps/chosen": -329.6391296386719,
"logps/rejected": -283.71624755859375,
"loss": 0.6338,
"positive_losses": 0.016074705868959427,
"rewards/accuracies": 0.949999988079071,
"rewards/chosen": 0.13542965054512024,
"rewards/margins": 0.12056276947259903,
"rewards/margins_max": 0.23423099517822266,
"rewards/margins_min": 0.021957222372293472,
"rewards/margins_std": 0.0950225442647934,
"rewards/rejected": 0.01486687920987606,
"step": 240
},
{
"dpo_losses": 0.6204310655593872,
"epoch": 0.7,
"grad_norm": 1.9142563702239555,
"learning_rate": 1.2216791228457775e-07,
"logits/chosen": -2.8029227256774902,
"logits/rejected": -2.6856589317321777,
"logps/chosen": -348.2162780761719,
"logps/rejected": -252.9592742919922,
"loss": 0.6225,
"positive_losses": 0.0,
"rewards/accuracies": 0.987500011920929,
"rewards/chosen": 0.15677711367607117,
"rewards/margins": 0.15525582432746887,
"rewards/margins_max": 0.30013835430145264,
"rewards/margins_min": 0.039689190685749054,
"rewards/margins_std": 0.11695368587970734,
"rewards/rejected": 0.0015213176375254989,
"step": 250
},
{
"dpo_losses": 0.6270495653152466,
"epoch": 0.73,
"grad_norm": 5.620542158536291,
"learning_rate": 1.0166289402331391e-07,
"logits/chosen": -2.8551700115203857,
"logits/rejected": -2.744173765182495,
"logps/chosen": -341.2341003417969,
"logps/rejected": -258.6510925292969,
"loss": 0.6278,
"positive_losses": 0.016833018511533737,
"rewards/accuracies": 0.987500011920929,
"rewards/chosen": 0.15211349725723267,
"rewards/margins": 0.14102734625339508,
"rewards/margins_max": 0.29794004559516907,
"rewards/margins_min": 0.023187965154647827,
"rewards/margins_std": 0.12820349633693695,
"rewards/rejected": 0.011086158454418182,
"step": 260
},
{
"dpo_losses": 0.6170064210891724,
"epoch": 0.76,
"grad_norm": 2.1039206735574796,
"learning_rate": 8.259540650444734e-08,
"logits/chosen": -2.807176351547241,
"logits/rejected": -2.7167766094207764,
"logps/chosen": -361.86981201171875,
"logps/rejected": -262.83673095703125,
"loss": 0.6263,
"positive_losses": 0.006254339125007391,
"rewards/accuracies": 0.949999988079071,
"rewards/chosen": 0.16030336916446686,
"rewards/margins": 0.16262248158454895,
"rewards/margins_max": 0.2933579683303833,
"rewards/margins_min": 0.03731368109583855,
"rewards/margins_std": 0.11614950001239777,
"rewards/rejected": -0.002319112652912736,
"step": 270
},
{
"dpo_losses": 0.6323626041412354,
"epoch": 0.79,
"grad_norm": 3.6286773445196094,
"learning_rate": 6.515023221586721e-08,
"logits/chosen": -2.756715774536133,
"logits/rejected": -2.709307909011841,
"logps/chosen": -316.50714111328125,
"logps/rejected": -272.3856506347656,
"loss": 0.6292,
"positive_losses": 0.029051829129457474,
"rewards/accuracies": 0.9125000238418579,
"rewards/chosen": 0.1414763480424881,
"rewards/margins": 0.12991073727607727,
"rewards/margins_max": 0.2901129424571991,
"rewards/margins_min": 0.026102622970938683,
"rewards/margins_std": 0.12171909958124161,
"rewards/rejected": 0.01156560517847538,
"step": 280
},
{
"dpo_losses": 0.6301154494285583,
"epoch": 0.82,
"grad_norm": 2.1273936086157175,
"learning_rate": 4.949643185335287e-08,
"logits/chosen": -2.7689366340637207,
"logits/rejected": -2.6886589527130127,
"logps/chosen": -326.63336181640625,
"logps/rejected": -265.2763671875,
"loss": 0.6315,
"positive_losses": 0.0,
"rewards/accuracies": 0.949999988079071,
"rewards/chosen": 0.14358708262443542,
"rewards/margins": 0.134388267993927,
"rewards/margins_max": 0.2748602032661438,
"rewards/margins_min": 0.020886730402708054,
"rewards/margins_std": 0.11660180240869522,
"rewards/rejected": 0.009198823012411594,
"step": 290
},
{
"dpo_losses": 0.6222444772720337,
"epoch": 0.85,
"grad_norm": 3.916824188852417,
"learning_rate": 3.578570595810274e-08,
"logits/chosen": -2.813533306121826,
"logits/rejected": -2.7393226623535156,
"logps/chosen": -347.9246520996094,
"logps/rejected": -288.87091064453125,
"loss": 0.6203,
"positive_losses": 0.004223608877509832,
"rewards/accuracies": 0.9125000238418579,
"rewards/chosen": 0.15709543228149414,
"rewards/margins": 0.15124014019966125,
"rewards/margins_max": 0.2926793098449707,
"rewards/margins_min": 0.03863045573234558,
"rewards/margins_std": 0.11517532914876938,
"rewards/rejected": 0.005855279043316841,
"step": 300
},
{
"epoch": 0.85,
"eval_dpo_losses": 0.6847304105758667,
"eval_logits/chosen": -2.769089937210083,
"eval_logits/rejected": -2.7314584255218506,
"eval_logps/chosen": -282.0690612792969,
"eval_logps/rejected": -257.8544921875,
"eval_loss": 0.7727298140525818,
"eval_positive_losses": 0.8306741118431091,
"eval_rewards/accuracies": 0.5920000076293945,
"eval_rewards/chosen": 0.02524358406662941,
"eval_rewards/margins": 0.017999326810240746,
"eval_rewards/margins_max": 0.12274405360221863,
"eval_rewards/margins_min": -0.07009908556938171,
"eval_rewards/margins_std": 0.06381046026945114,
"eval_rewards/rejected": 0.007244254928082228,
"eval_runtime": 428.2098,
"eval_samples_per_second": 4.671,
"eval_steps_per_second": 0.292,
"step": 300
},
{
"dpo_losses": 0.6221127510070801,
"epoch": 0.87,
"grad_norm": 1.8941985477237735,
"learning_rate": 2.415092479103503e-08,
"logits/chosen": -2.8478198051452637,
"logits/rejected": -2.7160096168518066,
"logps/chosen": -340.7911071777344,
"logps/rejected": -214.91964721679688,
"loss": 0.6237,
"positive_losses": 0.0,
"rewards/accuracies": 0.9750000238418579,
"rewards/chosen": 0.154912531375885,
"rewards/margins": 0.15173590183258057,
"rewards/margins_max": 0.29650163650512695,
"rewards/margins_min": 0.04189513251185417,
"rewards/margins_std": 0.12011243402957916,
"rewards/rejected": 0.0031766362953931093,
"step": 310
},
{
"dpo_losses": 0.6243400573730469,
"epoch": 0.9,
"grad_norm": 1.80973934530786,
"learning_rate": 1.4704840690808656e-08,
"logits/chosen": -2.8038718700408936,
"logits/rejected": -2.719196319580078,
"logps/chosen": -334.8002014160156,
"logps/rejected": -260.89532470703125,
"loss": 0.6212,
"positive_losses": 0.0,
"rewards/accuracies": 0.9375,
"rewards/chosen": 0.15841911733150482,
"rewards/margins": 0.1470610648393631,
"rewards/margins_max": 0.30652207136154175,
"rewards/margins_min": 0.019394170492887497,
"rewards/margins_std": 0.13127803802490234,
"rewards/rejected": 0.011358057148754597,
"step": 320
},
{
"dpo_losses": 0.6155394315719604,
"epoch": 0.93,
"grad_norm": 1.9914736197963372,
"learning_rate": 7.538995394063995e-09,
"logits/chosen": -2.8728480339050293,
"logits/rejected": -2.768115282058716,
"logps/chosen": -383.182373046875,
"logps/rejected": -267.9715881347656,
"loss": 0.6235,
"positive_losses": 0.017542457208037376,
"rewards/accuracies": 0.9375,
"rewards/chosen": 0.17095208168029785,
"rewards/margins": 0.16700315475463867,
"rewards/margins_max": 0.31099098920822144,
"rewards/margins_min": 0.040303539484739304,
"rewards/margins_std": 0.12561064958572388,
"rewards/rejected": 0.003948924597352743,
"step": 330
},
{
"dpo_losses": 0.6262079477310181,
"epoch": 0.96,
"grad_norm": 2.0098982490102606,
"learning_rate": 2.7228329070159705e-09,
"logits/chosen": -2.7688167095184326,
"logits/rejected": -2.681577205657959,
"logps/chosen": -329.7587890625,
"logps/rejected": -251.1042022705078,
"loss": 0.6217,
"positive_losses": 0.0,
"rewards/accuracies": 0.925000011920929,
"rewards/chosen": 0.1526114046573639,
"rewards/margins": 0.14340537786483765,
"rewards/margins_max": 0.28913384675979614,
"rewards/margins_min": 0.027879271656274796,
"rewards/margins_std": 0.1190972700715065,
"rewards/rejected": 0.009206023998558521,
"step": 340
},
{
"dpo_losses": 0.6229046583175659,
"epoch": 0.99,
"grad_norm": 3.4990163510778087,
"learning_rate": 3.0302652553296226e-10,
"logits/chosen": -2.761152744293213,
"logits/rejected": -2.6877024173736572,
"logps/chosen": -344.25860595703125,
"logps/rejected": -286.8030090332031,
"loss": 0.6221,
"positive_losses": 0.0,
"rewards/accuracies": 0.9624999761581421,
"rewards/chosen": 0.15005084872245789,
"rewards/margins": 0.1498393714427948,
"rewards/margins_max": 0.2942337691783905,
"rewards/margins_min": 0.04408826678991318,
"rewards/margins_std": 0.11335112154483795,
"rewards/rejected": 0.00021145511709619313,
"step": 350
},
{
"epoch": 1.0,
"step": 355,
"total_flos": 0.0,
"train_loss": 0.6494339942932129,
"train_runtime": 4013.1211,
"train_samples_per_second": 1.415,
"train_steps_per_second": 0.088
}
],
"logging_steps": 10,
"max_steps": 355,
"num_input_tokens_seen": 0,
"num_train_epochs": 1,
"save_steps": 100,
"total_flos": 0.0,
"train_batch_size": 4,
"trial_name": null,
"trial_params": null
}