zephyr-dpo-qlora-uf-oursuf6k-5e-7 / trainer_state.json
just1nseo's picture
Model save
f7ec731 verified
{
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 1.0,
"eval_steps": 100,
"global_step": 684,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.0,
"grad_norm": 1.972968290218204,
"learning_rate": 7.246376811594203e-09,
"logits/chosen": -2.8746490478515625,
"logits/rejected": -2.840811252593994,
"logps/chosen": -227.73272705078125,
"logps/rejected": -174.71890258789062,
"loss": 0.6931,
"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
},
{
"epoch": 0.01,
"grad_norm": 1.8679339461073734,
"learning_rate": 7.246376811594203e-08,
"logits/chosen": -2.769803047180176,
"logits/rejected": -2.7261266708374023,
"logps/chosen": -244.6114959716797,
"logps/rejected": -240.41116333007812,
"loss": 0.6931,
"rewards/accuracies": 0.3611111044883728,
"rewards/chosen": 0.00021203850337769836,
"rewards/margins": 0.00021335652854759246,
"rewards/margins_max": 0.0021964015904814005,
"rewards/margins_min": -0.0016354921972379088,
"rewards/margins_std": 0.0016972733428701758,
"rewards/rejected": -1.318060753874306e-06,
"step": 10
},
{
"epoch": 0.03,
"grad_norm": 2.1095097251535906,
"learning_rate": 1.4492753623188405e-07,
"logits/chosen": -2.8978943824768066,
"logits/rejected": -2.859205961227417,
"logps/chosen": -317.23260498046875,
"logps/rejected": -239.8469696044922,
"loss": 0.6931,
"rewards/accuracies": 0.5375000238418579,
"rewards/chosen": 0.00018418056424707174,
"rewards/margins": 9.037666313815862e-05,
"rewards/margins_max": 0.003943216986954212,
"rewards/margins_min": -0.003490231465548277,
"rewards/margins_std": 0.0033192276023328304,
"rewards/rejected": 9.380385745316744e-05,
"step": 20
},
{
"epoch": 0.04,
"grad_norm": 2.08467778375872,
"learning_rate": 2.1739130434782607e-07,
"logits/chosen": -2.811413288116455,
"logits/rejected": -2.7905545234680176,
"logps/chosen": -264.98712158203125,
"logps/rejected": -223.62734985351562,
"loss": 0.693,
"rewards/accuracies": 0.512499988079071,
"rewards/chosen": 0.00012606059317477047,
"rewards/margins": -7.177007319114637e-06,
"rewards/margins_max": 0.0028896895237267017,
"rewards/margins_min": -0.0032867384143173695,
"rewards/margins_std": 0.0027837478555738926,
"rewards/rejected": 0.000133237597765401,
"step": 30
},
{
"epoch": 0.06,
"grad_norm": 1.940398970016681,
"learning_rate": 2.898550724637681e-07,
"logits/chosen": -2.8652420043945312,
"logits/rejected": -2.83544659614563,
"logps/chosen": -305.147216796875,
"logps/rejected": -267.1144714355469,
"loss": 0.6928,
"rewards/accuracies": 0.5625,
"rewards/chosen": 0.0004788968653883785,
"rewards/margins": 0.0008589730714447796,
"rewards/margins_max": 0.0041023739613592625,
"rewards/margins_min": -0.002046389738097787,
"rewards/margins_std": 0.0028562676161527634,
"rewards/rejected": -0.0003800761769525707,
"step": 40
},
{
"epoch": 0.07,
"grad_norm": 1.6929010917265717,
"learning_rate": 3.6231884057971015e-07,
"logits/chosen": -2.8695337772369385,
"logits/rejected": -2.837745189666748,
"logps/chosen": -260.9767150878906,
"logps/rejected": -235.2560577392578,
"loss": 0.6924,
"rewards/accuracies": 0.637499988079071,
"rewards/chosen": 0.0011892963666468859,
"rewards/margins": 0.0010751936351880431,
"rewards/margins_max": 0.004216945730149746,
"rewards/margins_min": -0.002206298988312483,
"rewards/margins_std": 0.0029160729609429836,
"rewards/rejected": 0.00011410261504352093,
"step": 50
},
{
"epoch": 0.09,
"grad_norm": 1.6771532679917263,
"learning_rate": 4.3478260869565214e-07,
"logits/chosen": -2.8912997245788574,
"logits/rejected": -2.8439412117004395,
"logps/chosen": -281.3103942871094,
"logps/rejected": -248.8180389404297,
"loss": 0.6919,
"rewards/accuracies": 0.75,
"rewards/chosen": 0.0023077281657606363,
"rewards/margins": 0.0024695510510355234,
"rewards/margins_max": 0.006759033538401127,
"rewards/margins_min": -0.0015218419721350074,
"rewards/margins_std": 0.00368516705930233,
"rewards/rejected": -0.0001618233509361744,
"step": 60
},
{
"epoch": 0.1,
"grad_norm": 1.9158671065853772,
"learning_rate": 4.999967381905813e-07,
"logits/chosen": -2.7889466285705566,
"logits/rejected": -2.7540149688720703,
"logps/chosen": -290.43804931640625,
"logps/rejected": -212.5797882080078,
"loss": 0.691,
"rewards/accuracies": 0.762499988079071,
"rewards/chosen": 0.003653948428109288,
"rewards/margins": 0.004532678984105587,
"rewards/margins_max": 0.0123423608019948,
"rewards/margins_min": -0.001225766958668828,
"rewards/margins_std": 0.0061333803460001945,
"rewards/rejected": -0.000878730439580977,
"step": 70
},
{
"epoch": 0.12,
"grad_norm": 1.71294323830613,
"learning_rate": 4.996054240392509e-07,
"logits/chosen": -2.777132749557495,
"logits/rejected": -2.7576847076416016,
"logps/chosen": -265.20281982421875,
"logps/rejected": -243.9671173095703,
"loss": 0.6907,
"rewards/accuracies": 0.737500011920929,
"rewards/chosen": 0.003907615318894386,
"rewards/margins": 0.003999439999461174,
"rewards/margins_max": 0.012583956122398376,
"rewards/margins_min": -0.0027496658731251955,
"rewards/margins_std": 0.00690504303202033,
"rewards/rejected": -9.18240548344329e-05,
"step": 80
},
{
"epoch": 0.13,
"grad_norm": 2.4508716900973826,
"learning_rate": 4.985629178361649e-07,
"logits/chosen": -2.889789581298828,
"logits/rejected": -2.8571457862854004,
"logps/chosen": -273.2809753417969,
"logps/rejected": -246.13528442382812,
"loss": 0.6894,
"rewards/accuracies": 0.737500011920929,
"rewards/chosen": 0.0063544102013111115,
"rewards/margins": 0.007208968047052622,
"rewards/margins_max": 0.018225526437163353,
"rewards/margins_min": -0.0013279046397656202,
"rewards/margins_std": 0.008873926475644112,
"rewards/rejected": -0.0008545577293261886,
"step": 90
},
{
"epoch": 0.15,
"grad_norm": 1.8386412495416258,
"learning_rate": 4.968719393609756e-07,
"logits/chosen": -2.8424429893493652,
"logits/rejected": -2.7764132022857666,
"logps/chosen": -337.29986572265625,
"logps/rejected": -243.8036651611328,
"loss": 0.6876,
"rewards/accuracies": 0.8374999761581421,
"rewards/chosen": 0.010191970504820347,
"rewards/margins": 0.012033768929541111,
"rewards/margins_max": 0.029753312468528748,
"rewards/margins_min": -0.0020332676358520985,
"rewards/margins_std": 0.014395820908248425,
"rewards/rejected": -0.0018417991232126951,
"step": 100
},
{
"epoch": 0.15,
"eval_logits/chosen": -2.8052334785461426,
"eval_logits/rejected": -2.7664124965667725,
"eval_logps/chosen": -284.0943603515625,
"eval_logps/rejected": -258.4696960449219,
"eval_loss": 0.691686749458313,
"eval_rewards/accuracies": 0.6079999804496765,
"eval_rewards/chosen": 0.004990490153431892,
"eval_rewards/margins": 0.0038980983663350344,
"eval_rewards/margins_max": 0.023511478677392006,
"eval_rewards/margins_min": -0.012509307824075222,
"eval_rewards/margins_std": 0.011862216517329216,
"eval_rewards/rejected": 0.0010923919035121799,
"eval_runtime": 444.4734,
"eval_samples_per_second": 4.5,
"eval_steps_per_second": 0.281,
"step": 100
},
{
"epoch": 0.16,
"grad_norm": 1.9556267546868065,
"learning_rate": 4.945369001834514e-07,
"logits/chosen": -2.8761301040649414,
"logits/rejected": -2.8216347694396973,
"logps/chosen": -287.6670837402344,
"logps/rejected": -223.6521453857422,
"loss": 0.6881,
"rewards/accuracies": 0.7749999761581421,
"rewards/chosen": 0.009855161421000957,
"rewards/margins": 0.010875609703361988,
"rewards/margins_max": 0.027963850647211075,
"rewards/margins_min": -0.00028645730344578624,
"rewards/margins_std": 0.012777927331626415,
"rewards/rejected": -0.0010204474674537778,
"step": 110
},
{
"epoch": 0.18,
"grad_norm": 1.9579674466888082,
"learning_rate": 4.915638921541951e-07,
"logits/chosen": -2.8273448944091797,
"logits/rejected": -2.8115439414978027,
"logps/chosen": -257.2424011230469,
"logps/rejected": -237.67904663085938,
"loss": 0.6866,
"rewards/accuracies": 0.7875000238418579,
"rewards/chosen": 0.00945218000560999,
"rewards/margins": 0.012427609413862228,
"rewards/margins_max": 0.033436521887779236,
"rewards/margins_min": -0.0018920926377177238,
"rewards/margins_std": 0.015944166108965874,
"rewards/rejected": -0.0029754305724054575,
"step": 120
},
{
"epoch": 0.19,
"grad_norm": 1.5966946942385376,
"learning_rate": 4.879606715117018e-07,
"logits/chosen": -2.878997325897217,
"logits/rejected": -2.8253443241119385,
"logps/chosen": -275.1449890136719,
"logps/rejected": -235.9576873779297,
"loss": 0.6859,
"rewards/accuracies": 0.75,
"rewards/chosen": 0.00955723412334919,
"rewards/margins": 0.012582411989569664,
"rewards/margins_max": 0.03178320452570915,
"rewards/margins_min": -0.004044829867780209,
"rewards/margins_std": 0.0158962644636631,
"rewards/rejected": -0.003025178564712405,
"step": 130
},
{
"epoch": 0.2,
"grad_norm": 1.612892528929156,
"learning_rate": 4.837366386472174e-07,
"logits/chosen": -2.8901007175445557,
"logits/rejected": -2.827146053314209,
"logps/chosen": -297.2471618652344,
"logps/rejected": -254.73318481445312,
"loss": 0.686,
"rewards/accuracies": 0.7875000238418579,
"rewards/chosen": 0.015631090849637985,
"rewards/margins": 0.017191508784890175,
"rewards/margins_max": 0.04147082567214966,
"rewards/margins_min": -0.0013556934427469969,
"rewards/margins_std": 0.019727854058146477,
"rewards/rejected": -0.0015604153741151094,
"step": 140
},
{
"epoch": 0.22,
"grad_norm": 1.7836254409467616,
"learning_rate": 4.789028135801918e-07,
"logits/chosen": -2.847933769226074,
"logits/rejected": -2.8365511894226074,
"logps/chosen": -297.95208740234375,
"logps/rejected": -272.38201904296875,
"loss": 0.684,
"rewards/accuracies": 0.762499988079071,
"rewards/chosen": 0.016203617677092552,
"rewards/margins": 0.018767360597848892,
"rewards/margins_max": 0.05341927334666252,
"rewards/margins_min": -0.011627629399299622,
"rewards/margins_std": 0.029866989701986313,
"rewards/rejected": -0.0025637417566031218,
"step": 150
},
{
"epoch": 0.23,
"grad_norm": 8.25250622221843,
"learning_rate": 4.7347180720830627e-07,
"logits/chosen": -2.867987632751465,
"logits/rejected": -2.7911014556884766,
"logps/chosen": -300.35626220703125,
"logps/rejected": -261.74908447265625,
"loss": 0.6847,
"rewards/accuracies": 0.8125,
"rewards/chosen": 0.016810361295938492,
"rewards/margins": 0.01963501051068306,
"rewards/margins_max": 0.05201994627714157,
"rewards/margins_min": -0.0029959846287965775,
"rewards/margins_std": 0.024577533826231956,
"rewards/rejected": -0.002824649680405855,
"step": 160
},
{
"epoch": 0.25,
"grad_norm": 1.8565831100637469,
"learning_rate": 4.6745778840708107e-07,
"logits/chosen": -2.7958624362945557,
"logits/rejected": -2.783989429473877,
"logps/chosen": -295.93438720703125,
"logps/rejected": -227.43661499023438,
"loss": 0.6779,
"rewards/accuracies": 0.824999988079071,
"rewards/chosen": 0.024501120671629906,
"rewards/margins": 0.02991095557808876,
"rewards/margins_max": 0.06729420274496078,
"rewards/margins_min": -0.003415555926039815,
"rewards/margins_std": 0.03072194755077362,
"rewards/rejected": -0.005409830249845982,
"step": 170
},
{
"epoch": 0.26,
"grad_norm": 1.8914594463949175,
"learning_rate": 4.6087644706489703e-07,
"logits/chosen": -2.8270645141601562,
"logits/rejected": -2.787436008453369,
"logps/chosen": -294.5668640136719,
"logps/rejected": -255.82962036132812,
"loss": 0.6816,
"rewards/accuracies": 0.7749999761581421,
"rewards/chosen": 0.01939496397972107,
"rewards/margins": 0.02427856996655464,
"rewards/margins_max": 0.059351809322834015,
"rewards/margins_min": -0.008079716935753822,
"rewards/margins_std": 0.03051997348666191,
"rewards/rejected": -0.004883607849478722,
"step": 180
},
{
"epoch": 0.28,
"grad_norm": 1.724680638811993,
"learning_rate": 4.537449531498687e-07,
"logits/chosen": -2.767610788345337,
"logits/rejected": -2.719372272491455,
"logps/chosen": -281.36297607421875,
"logps/rejected": -229.79306030273438,
"loss": 0.6809,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": 0.02440481074154377,
"rewards/margins": 0.026266787201166153,
"rewards/margins_max": 0.06888638436794281,
"rewards/margins_min": -0.009829925373196602,
"rewards/margins_std": 0.034937743097543716,
"rewards/rejected": -0.0018619761103764176,
"step": 190
},
{
"epoch": 0.29,
"grad_norm": 1.587562818110368,
"learning_rate": 4.4608191191535736e-07,
"logits/chosen": -2.869741201400757,
"logits/rejected": -2.8258328437805176,
"logps/chosen": -275.5715637207031,
"logps/rejected": -244.32763671875,
"loss": 0.6792,
"rewards/accuracies": 0.75,
"rewards/chosen": 0.02022332139313221,
"rewards/margins": 0.026959722861647606,
"rewards/margins_max": 0.07024272531270981,
"rewards/margins_min": -0.013307643122971058,
"rewards/margins_std": 0.037044934928417206,
"rewards/rejected": -0.006736403796821833,
"step": 200
},
{
"epoch": 0.29,
"eval_logits/chosen": -2.7954251766204834,
"eval_logits/rejected": -2.7569642066955566,
"eval_logps/chosen": -283.2934875488281,
"eval_logps/rejected": -258.5861511230469,
"eval_loss": 0.6882554888725281,
"eval_rewards/accuracies": 0.6370000243186951,
"eval_rewards/chosen": 0.012999121099710464,
"eval_rewards/margins": 0.013071166351437569,
"eval_rewards/margins_max": 0.07187327742576599,
"eval_rewards/margins_min": -0.03631452098488808,
"eval_rewards/margins_std": 0.035920411348342896,
"eval_rewards/rejected": -7.204585563158616e-05,
"eval_runtime": 453.7395,
"eval_samples_per_second": 4.408,
"eval_steps_per_second": 0.275,
"step": 200
},
{
"epoch": 0.31,
"grad_norm": 1.7443006213577599,
"learning_rate": 4.379073153609896e-07,
"logits/chosen": -2.8446342945098877,
"logits/rejected": -2.8066139221191406,
"logps/chosen": -298.4417724609375,
"logps/rejected": -260.84881591796875,
"loss": 0.6787,
"rewards/accuracies": 0.737500011920929,
"rewards/chosen": 0.028581559658050537,
"rewards/margins": 0.03288479894399643,
"rewards/margins_max": 0.08662258088588715,
"rewards/margins_min": -0.013487100601196289,
"rewards/margins_std": 0.04499204084277153,
"rewards/rejected": -0.0043032425455749035,
"step": 210
},
{
"epoch": 0.32,
"grad_norm": 1.7530104764635095,
"learning_rate": 4.292424900758128e-07,
"logits/chosen": -2.8034873008728027,
"logits/rejected": -2.775111675262451,
"logps/chosen": -279.15478515625,
"logps/rejected": -223.8057403564453,
"loss": 0.6778,
"rewards/accuracies": 0.875,
"rewards/chosen": 0.028650784865021706,
"rewards/margins": 0.03896017372608185,
"rewards/margins_max": 0.08817549049854279,
"rewards/margins_min": -0.0010216787923127413,
"rewards/margins_std": 0.04009110480546951,
"rewards/rejected": -0.010309383273124695,
"step": 220
},
{
"epoch": 0.34,
"grad_norm": 1.6934833415236739,
"learning_rate": 4.201100415996597e-07,
"logits/chosen": -2.7763991355895996,
"logits/rejected": -2.7445781230926514,
"logps/chosen": -263.20074462890625,
"logps/rejected": -261.6719970703125,
"loss": 0.6785,
"rewards/accuracies": 0.75,
"rewards/chosen": 0.030025389045476913,
"rewards/margins": 0.031550828367471695,
"rewards/margins_max": 0.09253410995006561,
"rewards/margins_min": -0.01012241281569004,
"rewards/margins_std": 0.04607797786593437,
"rewards/rejected": -0.001525437692180276,
"step": 230
},
{
"epoch": 0.35,
"grad_norm": 1.9254815601584274,
"learning_rate": 4.1053379544787557e-07,
"logits/chosen": -2.8438620567321777,
"logits/rejected": -2.8013641834259033,
"logps/chosen": -286.17901611328125,
"logps/rejected": -249.131591796875,
"loss": 0.6777,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": 0.03425859659910202,
"rewards/margins": 0.04053955897688866,
"rewards/margins_max": 0.09026241302490234,
"rewards/margins_min": -0.004998114425688982,
"rewards/margins_std": 0.043195050209760666,
"rewards/rejected": -0.006280961446464062,
"step": 240
},
{
"epoch": 0.37,
"grad_norm": 1.773759451391351,
"learning_rate": 4.0053873495326964e-07,
"logits/chosen": -2.802912950515747,
"logits/rejected": -2.7872378826141357,
"logps/chosen": -243.8607177734375,
"logps/rejected": -229.35678100585938,
"loss": 0.674,
"rewards/accuracies": 0.8125,
"rewards/chosen": 0.026193741708993912,
"rewards/margins": 0.03452175855636597,
"rewards/margins_max": 0.08425874263048172,
"rewards/margins_min": -0.007185367401689291,
"rewards/margins_std": 0.04130570963025093,
"rewards/rejected": -0.00832801777869463,
"step": 250
},
{
"epoch": 0.38,
"grad_norm": 1.89562327580133,
"learning_rate": 3.9015093608745143e-07,
"logits/chosen": -2.8584208488464355,
"logits/rejected": -2.7942659854888916,
"logps/chosen": -283.54974365234375,
"logps/rejected": -242.9059295654297,
"loss": 0.6717,
"rewards/accuracies": 0.824999988079071,
"rewards/chosen": 0.03558926284313202,
"rewards/margins": 0.0444478802382946,
"rewards/margins_max": 0.11205202341079712,
"rewards/margins_min": -0.004177084192633629,
"rewards/margins_std": 0.05269969254732132,
"rewards/rejected": -0.008858618326485157,
"step": 260
},
{
"epoch": 0.39,
"grad_norm": 2.30599597059098,
"learning_rate": 3.79397499431599e-07,
"logits/chosen": -2.8217787742614746,
"logits/rejected": -2.7673909664154053,
"logps/chosen": -291.3817138671875,
"logps/rejected": -263.5201721191406,
"loss": 0.672,
"rewards/accuracies": 0.75,
"rewards/chosen": 0.04040498286485672,
"rewards/margins": 0.04550846666097641,
"rewards/margins_max": 0.10735081136226654,
"rewards/margins_min": -0.009063487872481346,
"rewards/margins_std": 0.051175691187381744,
"rewards/rejected": -0.00510348379611969,
"step": 270
},
{
"epoch": 0.41,
"grad_norm": 1.6888761913438506,
"learning_rate": 3.6830647947413694e-07,
"logits/chosen": -2.88932728767395,
"logits/rejected": -2.8418757915496826,
"logps/chosen": -274.30218505859375,
"logps/rejected": -248.6231231689453,
"loss": 0.671,
"rewards/accuracies": 0.762499988079071,
"rewards/chosen": 0.03000582382082939,
"rewards/margins": 0.04024919122457504,
"rewards/margins_max": 0.10416732728481293,
"rewards/margins_min": -0.011732708662748337,
"rewards/margins_std": 0.05244187265634537,
"rewards/rejected": -0.0102433692663908,
"step": 280
},
{
"epoch": 0.42,
"grad_norm": 1.6649865105801362,
"learning_rate": 3.5690681141977837e-07,
"logits/chosen": -2.822051525115967,
"logits/rejected": -2.7766671180725098,
"logps/chosen": -253.5765380859375,
"logps/rejected": -206.1942901611328,
"loss": 0.6729,
"rewards/accuracies": 0.7749999761581421,
"rewards/chosen": 0.031104344874620438,
"rewards/margins": 0.04246490076184273,
"rewards/margins_max": 0.12114210426807404,
"rewards/margins_min": -0.018173199146986008,
"rewards/margins_std": 0.06279204040765762,
"rewards/rejected": -0.011360556818544865,
"step": 290
},
{
"epoch": 0.44,
"grad_norm": 2.1922548116517397,
"learning_rate": 3.4522823570088067e-07,
"logits/chosen": -2.821171998977661,
"logits/rejected": -2.808584451675415,
"logps/chosen": -259.92913818359375,
"logps/rejected": -254.540283203125,
"loss": 0.6697,
"rewards/accuracies": 0.7250000238418579,
"rewards/chosen": 0.03274675831198692,
"rewards/margins": 0.047333501279354095,
"rewards/margins_max": 0.13525982201099396,
"rewards/margins_min": -0.01479897927492857,
"rewards/margins_std": 0.06737245619297028,
"rewards/rejected": -0.01458674855530262,
"step": 300
},
{
"epoch": 0.44,
"eval_logits/chosen": -2.786345958709717,
"eval_logits/rejected": -2.7488794326782227,
"eval_logps/chosen": -282.5645751953125,
"eval_logps/rejected": -258.747314453125,
"eval_loss": 0.6848979592323303,
"eval_rewards/accuracies": 0.6169999837875366,
"eval_rewards/chosen": 0.020288635045289993,
"eval_rewards/margins": 0.02197239361703396,
"eval_rewards/margins_max": 0.1184224858880043,
"eval_rewards/margins_min": -0.059722770005464554,
"eval_rewards/margins_std": 0.05925743281841278,
"eval_rewards/rejected": -0.0016837569419294596,
"eval_runtime": 440.5364,
"eval_samples_per_second": 4.54,
"eval_steps_per_second": 0.284,
"step": 300
},
{
"epoch": 0.45,
"grad_norm": 1.884851957725623,
"learning_rate": 3.3330122038805277e-07,
"logits/chosen": -2.8432490825653076,
"logits/rejected": -2.795261859893799,
"logps/chosen": -272.4532470703125,
"logps/rejected": -227.98348999023438,
"loss": 0.6672,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": 0.038232967257499695,
"rewards/margins": 0.0511365607380867,
"rewards/margins_max": 0.1235017329454422,
"rewards/margins_min": -0.007081184536218643,
"rewards/margins_std": 0.05993686243891716,
"rewards/rejected": -0.012903591617941856,
"step": 310
},
{
"epoch": 0.47,
"grad_norm": 1.732519789963906,
"learning_rate": 3.2115688170243734e-07,
"logits/chosen": -2.831395387649536,
"logits/rejected": -2.8028111457824707,
"logps/chosen": -270.9764709472656,
"logps/rejected": -252.1440887451172,
"loss": 0.6681,
"rewards/accuracies": 0.824999988079071,
"rewards/chosen": 0.037659503519535065,
"rewards/margins": 0.05491337180137634,
"rewards/margins_max": 0.132755845785141,
"rewards/margins_min": -0.010344445705413818,
"rewards/margins_std": 0.06564854830503464,
"rewards/rejected": -0.01725386641919613,
"step": 320
},
{
"epoch": 0.48,
"grad_norm": 1.5242926911564552,
"learning_rate": 3.088269028370435e-07,
"logits/chosen": -2.8538591861724854,
"logits/rejected": -2.8225364685058594,
"logps/chosen": -295.54388427734375,
"logps/rejected": -250.438232421875,
"loss": 0.663,
"rewards/accuracies": 0.862500011920929,
"rewards/chosen": 0.04337853938341141,
"rewards/margins": 0.06583191454410553,
"rewards/margins_max": 0.15647490322589874,
"rewards/margins_min": -0.0003548143431544304,
"rewards/margins_std": 0.07130275666713715,
"rewards/rejected": -0.022453375160694122,
"step": 330
},
{
"epoch": 0.5,
"grad_norm": 1.6819074997159351,
"learning_rate": 2.9634345129891294e-07,
"logits/chosen": -2.80195689201355,
"logits/rejected": -2.765192985534668,
"logps/chosen": -288.1705627441406,
"logps/rejected": -258.40667724609375,
"loss": 0.6647,
"rewards/accuracies": 0.7875000238418579,
"rewards/chosen": 0.04105386883020401,
"rewards/margins": 0.06338892132043839,
"rewards/margins_max": 0.1566300094127655,
"rewards/margins_min": -0.008608223870396614,
"rewards/margins_std": 0.0741976946592331,
"rewards/rejected": -0.022335056215524673,
"step": 340
},
{
"epoch": 0.51,
"grad_norm": 1.9298838156481428,
"learning_rate": 2.8373909498776744e-07,
"logits/chosen": -2.8505892753601074,
"logits/rejected": -2.8334896564483643,
"logps/chosen": -280.2142028808594,
"logps/rejected": -279.97857666015625,
"loss": 0.6686,
"rewards/accuracies": 0.8374999761581421,
"rewards/chosen": 0.03649730607867241,
"rewards/margins": 0.05252464860677719,
"rewards/margins_max": 0.12718316912651062,
"rewards/margins_min": -0.00450880965217948,
"rewards/margins_std": 0.061397988349199295,
"rewards/rejected": -0.016027342528104782,
"step": 350
},
{
"epoch": 0.53,
"grad_norm": 1.9549755637356359,
"learning_rate": 2.710467172300768e-07,
"logits/chosen": -2.756687879562378,
"logits/rejected": -2.71925687789917,
"logps/chosen": -318.5474548339844,
"logps/rejected": -252.8844757080078,
"loss": 0.6597,
"rewards/accuracies": 0.8374999761581421,
"rewards/chosen": 0.05029547959566116,
"rewards/margins": 0.06931338459253311,
"rewards/margins_max": 0.15269415080547333,
"rewards/margins_min": 0.0053079272620379925,
"rewards/margins_std": 0.06585155427455902,
"rewards/rejected": -0.01901790127158165,
"step": 360
},
{
"epoch": 0.54,
"grad_norm": 2.265857763501263,
"learning_rate": 2.582994309902146e-07,
"logits/chosen": -2.7980430126190186,
"logits/rejected": -2.7501606941223145,
"logps/chosen": -305.3988342285156,
"logps/rejected": -258.4122314453125,
"loss": 0.6649,
"rewards/accuracies": 0.7749999761581421,
"rewards/chosen": 0.04739305004477501,
"rewards/margins": 0.06349317729473114,
"rewards/margins_max": 0.17864832282066345,
"rewards/margins_min": -0.03638879954814911,
"rewards/margins_std": 0.09656090289354324,
"rewards/rejected": -0.01610013097524643,
"step": 370
},
{
"epoch": 0.56,
"grad_norm": 2.2730703155677743,
"learning_rate": 2.455304924825151e-07,
"logits/chosen": -2.813772678375244,
"logits/rejected": -2.8081743717193604,
"logps/chosen": -276.4326477050781,
"logps/rejected": -260.1085510253906,
"loss": 0.6546,
"rewards/accuracies": 0.8125,
"rewards/chosen": 0.044434770941734314,
"rewards/margins": 0.07333989441394806,
"rewards/margins_max": 0.16991741955280304,
"rewards/margins_min": 0.007076957728713751,
"rewards/margins_std": 0.07423131167888641,
"rewards/rejected": -0.028905129060149193,
"step": 380
},
{
"epoch": 0.57,
"grad_norm": 1.8322064736841432,
"learning_rate": 2.3277321440960732e-07,
"logits/chosen": -2.811725616455078,
"logits/rejected": -2.7685177326202393,
"logps/chosen": -268.05914306640625,
"logps/rejected": -266.083740234375,
"loss": 0.6614,
"rewards/accuracies": 0.7875000238418579,
"rewards/chosen": 0.04107924923300743,
"rewards/margins": 0.06845887005329132,
"rewards/margins_max": 0.18714329600334167,
"rewards/margins_min": -0.010029973462224007,
"rewards/margins_std": 0.08763924241065979,
"rewards/rejected": -0.02737962268292904,
"step": 390
},
{
"epoch": 0.58,
"grad_norm": 2.128975347395003,
"learning_rate": 2.2006087905337698e-07,
"logits/chosen": -2.8421759605407715,
"logits/rejected": -2.816889762878418,
"logps/chosen": -239.36984252929688,
"logps/rejected": -236.58364868164062,
"loss": 0.6571,
"rewards/accuracies": 0.862500011920929,
"rewards/chosen": 0.04006009176373482,
"rewards/margins": 0.07172206044197083,
"rewards/margins_max": 0.17035157978534698,
"rewards/margins_min": -0.0009761411929503083,
"rewards/margins_std": 0.07894248515367508,
"rewards/rejected": -0.03166196495294571,
"step": 400
},
{
"epoch": 0.58,
"eval_logits/chosen": -2.7727160453796387,
"eval_logits/rejected": -2.735231399536133,
"eval_logps/chosen": -282.5451965332031,
"eval_logps/rejected": -259.56536865234375,
"eval_loss": 0.6818826198577881,
"eval_rewards/accuracies": 0.6330000162124634,
"eval_rewards/chosen": 0.02048237808048725,
"eval_rewards/margins": 0.03034677356481552,
"eval_rewards/margins_max": 0.15978190302848816,
"eval_rewards/margins_min": -0.08066722005605698,
"eval_rewards/margins_std": 0.08006121963262558,
"eval_rewards/rejected": -0.00986439362168312,
"eval_runtime": 428.8143,
"eval_samples_per_second": 4.664,
"eval_steps_per_second": 0.292,
"step": 400
},
{
"epoch": 0.6,
"grad_norm": 1.464979648644022,
"learning_rate": 2.0742665144529372e-07,
"logits/chosen": -2.8245086669921875,
"logits/rejected": -2.7860279083251953,
"logps/chosen": -308.59234619140625,
"logps/rejected": -286.37725830078125,
"loss": 0.6545,
"rewards/accuracies": 0.7749999761581421,
"rewards/chosen": 0.05577210336923599,
"rewards/margins": 0.08082688599824905,
"rewards/margins_max": 0.21429844200611115,
"rewards/margins_min": -0.014085543341934681,
"rewards/margins_std": 0.10368019342422485,
"rewards/rejected": -0.02505478635430336,
"step": 410
},
{
"epoch": 0.61,
"grad_norm": 2.2565152886733024,
"learning_rate": 1.9490349284263033e-07,
"logits/chosen": -2.7606394290924072,
"logits/rejected": -2.7328662872314453,
"logps/chosen": -257.64727783203125,
"logps/rejected": -247.62216186523438,
"loss": 0.6651,
"rewards/accuracies": 0.7875000238418579,
"rewards/chosen": 0.034510333091020584,
"rewards/margins": 0.05616886541247368,
"rewards/margins_max": 0.16494083404541016,
"rewards/margins_min": -0.04768489673733711,
"rewards/margins_std": 0.09831468015909195,
"rewards/rejected": -0.021658534184098244,
"step": 420
},
{
"epoch": 0.63,
"grad_norm": 2.3425120835618545,
"learning_rate": 1.8252407473630605e-07,
"logits/chosen": -2.8338799476623535,
"logits/rejected": -2.780625820159912,
"logps/chosen": -296.00457763671875,
"logps/rejected": -234.919189453125,
"loss": 0.6593,
"rewards/accuracies": 0.824999988079071,
"rewards/chosen": 0.05103176832199097,
"rewards/margins": 0.07940290123224258,
"rewards/margins_max": 0.18009448051452637,
"rewards/margins_min": -0.008073708042502403,
"rewards/margins_std": 0.08305726200342178,
"rewards/rejected": -0.028371136635541916,
"step": 430
},
{
"epoch": 0.64,
"grad_norm": 1.9987304949927014,
"learning_rate": 1.7032069361469764e-07,
"logits/chosen": -2.7400341033935547,
"logits/rejected": -2.7252862453460693,
"logps/chosen": -228.4139404296875,
"logps/rejected": -274.11822509765625,
"loss": 0.6614,
"rewards/accuracies": 0.7875000238418579,
"rewards/chosen": 0.04276793450117111,
"rewards/margins": 0.0704236626625061,
"rewards/margins_max": 0.18232765793800354,
"rewards/margins_min": -0.0089184595271945,
"rewards/margins_std": 0.08663706481456757,
"rewards/rejected": -0.02765573561191559,
"step": 440
},
{
"epoch": 0.66,
"grad_norm": 1.5785697838984911,
"learning_rate": 1.58325186705788e-07,
"logits/chosen": -2.8510193824768066,
"logits/rejected": -2.829686403274536,
"logps/chosen": -269.9481506347656,
"logps/rejected": -256.58563232421875,
"loss": 0.6643,
"rewards/accuracies": 0.7250000238418579,
"rewards/chosen": 0.028526362031698227,
"rewards/margins": 0.06000509113073349,
"rewards/margins_max": 0.16429737210273743,
"rewards/margins_min": -0.04647505283355713,
"rewards/margins_std": 0.09461401402950287,
"rewards/rejected": -0.03147872909903526,
"step": 450
},
{
"epoch": 0.67,
"grad_norm": 1.737884201823339,
"learning_rate": 1.4656884891747395e-07,
"logits/chosen": -2.7824554443359375,
"logits/rejected": -2.7381398677825928,
"logps/chosen": -283.9850158691406,
"logps/rejected": -239.4892120361328,
"loss": 0.6607,
"rewards/accuracies": 0.7749999761581421,
"rewards/chosen": 0.037972342222929,
"rewards/margins": 0.07290490716695786,
"rewards/margins_max": 0.1753791719675064,
"rewards/margins_min": -0.014705635607242584,
"rewards/margins_std": 0.0863322764635086,
"rewards/rejected": -0.03493257611989975,
"step": 460
},
{
"epoch": 0.69,
"grad_norm": 1.8923794511223242,
"learning_rate": 1.3508235119272466e-07,
"logits/chosen": -2.8365092277526855,
"logits/rejected": -2.7932674884796143,
"logps/chosen": -324.05743408203125,
"logps/rejected": -294.9861755371094,
"loss": 0.6479,
"rewards/accuracies": 0.75,
"rewards/chosen": 0.05230081081390381,
"rewards/margins": 0.09341531991958618,
"rewards/margins_max": 0.25887981057167053,
"rewards/margins_min": -0.017766449600458145,
"rewards/margins_std": 0.12934701144695282,
"rewards/rejected": -0.04111451655626297,
"step": 470
},
{
"epoch": 0.7,
"grad_norm": 1.5554031948375129,
"learning_rate": 1.2389566049259336e-07,
"logits/chosen": -2.7813267707824707,
"logits/rejected": -2.7599194049835205,
"logps/chosen": -239.9017333984375,
"logps/rejected": -236.3795928955078,
"loss": 0.6554,
"rewards/accuracies": 0.762499988079071,
"rewards/chosen": 0.04591267183423042,
"rewards/margins": 0.08855243027210236,
"rewards/margins_max": 0.20911240577697754,
"rewards/margins_min": -0.007775710429996252,
"rewards/margins_std": 0.09931908547878265,
"rewards/rejected": -0.042639754712581635,
"step": 480
},
{
"epoch": 0.72,
"grad_norm": 2.494715637025419,
"learning_rate": 1.1303796161583762e-07,
"logits/chosen": -2.889634847640991,
"logits/rejected": -2.8090715408325195,
"logps/chosen": -317.6177062988281,
"logps/rejected": -262.18414306640625,
"loss": 0.6536,
"rewards/accuracies": 0.7875000238418579,
"rewards/chosen": 0.06907807290554047,
"rewards/margins": 0.10660415887832642,
"rewards/margins_max": 0.251437246799469,
"rewards/margins_min": -0.01369224488735199,
"rewards/margins_std": 0.11950767040252686,
"rewards/rejected": -0.03752607852220535,
"step": 490
},
{
"epoch": 0.73,
"grad_norm": 1.9249901065324473,
"learning_rate": 1.0253758105911167e-07,
"logits/chosen": -2.8541712760925293,
"logits/rejected": -2.8100945949554443,
"logps/chosen": -331.4628601074219,
"logps/rejected": -291.27984619140625,
"loss": 0.6508,
"rewards/accuracies": 0.862500011920929,
"rewards/chosen": 0.07663814723491669,
"rewards/margins": 0.11299224942922592,
"rewards/margins_max": 0.2441258728504181,
"rewards/margins_min": 0.010978538542985916,
"rewards/margins_std": 0.10471439361572266,
"rewards/rejected": -0.036354102194309235,
"step": 500
},
{
"epoch": 0.73,
"eval_logits/chosen": -2.768547296524048,
"eval_logits/rejected": -2.7314090728759766,
"eval_logps/chosen": -283.3109436035156,
"eval_logps/rejected": -260.8111267089844,
"eval_loss": 0.6802051067352295,
"eval_rewards/accuracies": 0.6269999742507935,
"eval_rewards/chosen": 0.012824743054807186,
"eval_rewards/margins": 0.035146910697221756,
"eval_rewards/margins_max": 0.1843804121017456,
"eval_rewards/margins_min": -0.09391897916793823,
"eval_rewards/margins_std": 0.09255214780569077,
"eval_rewards/rejected": -0.022322168573737144,
"eval_runtime": 449.4803,
"eval_samples_per_second": 4.45,
"eval_steps_per_second": 0.278,
"step": 500
},
{
"epoch": 0.75,
"grad_norm": 2.267072438870165,
"learning_rate": 9.242191311637049e-08,
"logits/chosen": -2.799065113067627,
"logits/rejected": -2.7682583332061768,
"logps/chosen": -279.16363525390625,
"logps/rejected": -241.4091796875,
"loss": 0.6585,
"rewards/accuracies": 0.8374999761581421,
"rewards/chosen": 0.043598730117082596,
"rewards/margins": 0.08321089297533035,
"rewards/margins_max": 0.2153284102678299,
"rewards/margins_min": -0.01286692637950182,
"rewards/margins_std": 0.10618630796670914,
"rewards/rejected": -0.03961215913295746,
"step": 510
},
{
"epoch": 0.76,
"grad_norm": 1.9414414953336778,
"learning_rate": 8.271734841028552e-08,
"logits/chosen": -2.8196640014648438,
"logits/rejected": -2.747915029525757,
"logps/chosen": -337.2265625,
"logps/rejected": -270.56146240234375,
"loss": 0.6494,
"rewards/accuracies": 0.8374999761581421,
"rewards/chosen": 0.054617054760456085,
"rewards/margins": 0.09917386621236801,
"rewards/margins_max": 0.24081799387931824,
"rewards/margins_min": -0.0010756913106888533,
"rewards/margins_std": 0.11212246119976044,
"rewards/rejected": -0.04455682262778282,
"step": 520
},
{
"epoch": 0.77,
"grad_norm": 1.952363376892767,
"learning_rate": 7.344920504212243e-08,
"logits/chosen": -2.8049850463867188,
"logits/rejected": -2.7458760738372803,
"logps/chosen": -263.7733154296875,
"logps/rejected": -222.8297882080078,
"loss": 0.6563,
"rewards/accuracies": 0.7749999761581421,
"rewards/chosen": 0.02564847469329834,
"rewards/margins": 0.06961268186569214,
"rewards/margins_max": 0.18673083186149597,
"rewards/margins_min": -0.030092215165495872,
"rewards/margins_std": 0.09981914609670639,
"rewards/rejected": -0.0439642071723938,
"step": 530
},
{
"epoch": 0.79,
"grad_norm": 1.7758645410300213,
"learning_rate": 6.46416625397067e-08,
"logits/chosen": -2.801713466644287,
"logits/rejected": -2.745816230773926,
"logps/chosen": -321.74554443359375,
"logps/rejected": -301.52325439453125,
"loss": 0.6518,
"rewards/accuracies": 0.8500000238418579,
"rewards/chosen": 0.055145345628261566,
"rewards/margins": 0.0942390188574791,
"rewards/margins_max": 0.2147323191165924,
"rewards/margins_min": -0.01778433658182621,
"rewards/margins_std": 0.10189126431941986,
"rewards/rejected": -0.03909367322921753,
"step": 540
},
{
"epoch": 0.8,
"grad_norm": 1.6203543941779939,
"learning_rate": 5.6317698775795344e-08,
"logits/chosen": -2.8398869037628174,
"logits/rejected": -2.8026833534240723,
"logps/chosen": -280.53826904296875,
"logps/rejected": -275.3709411621094,
"loss": 0.6538,
"rewards/accuracies": 0.75,
"rewards/chosen": 0.03587827831506729,
"rewards/margins": 0.06856991350650787,
"rewards/margins_max": 0.1956954300403595,
"rewards/margins_min": -0.03170696645975113,
"rewards/margins_std": 0.10253816843032837,
"rewards/rejected": -0.03269163519144058,
"step": 550
},
{
"epoch": 0.82,
"grad_norm": 1.8985615756457714,
"learning_rate": 4.849903002143113e-08,
"logits/chosen": -2.875913143157959,
"logits/rejected": -2.8281137943267822,
"logps/chosen": -329.9501037597656,
"logps/rejected": -278.819580078125,
"loss": 0.6436,
"rewards/accuracies": 0.8125,
"rewards/chosen": 0.06364526599645615,
"rewards/margins": 0.11146412044763565,
"rewards/margins_max": 0.2525936961174011,
"rewards/margins_min": -0.006486054509878159,
"rewards/margins_std": 0.11787240207195282,
"rewards/rejected": -0.047818850725889206,
"step": 560
},
{
"epoch": 0.83,
"grad_norm": 1.7787852048841497,
"learning_rate": 4.1206054290670535e-08,
"logits/chosen": -2.827500104904175,
"logits/rejected": -2.7943129539489746,
"logps/chosen": -267.7633056640625,
"logps/rejected": -264.48822021484375,
"loss": 0.6537,
"rewards/accuracies": 0.862500011920929,
"rewards/chosen": 0.05329934507608414,
"rewards/margins": 0.09477965533733368,
"rewards/margins_max": 0.20895680785179138,
"rewards/margins_min": 0.0013521288055926561,
"rewards/margins_std": 0.0940418690443039,
"rewards/rejected": -0.041480325162410736,
"step": 570
},
{
"epoch": 0.85,
"grad_norm": 2.0202098422067625,
"learning_rate": 3.44577981244944e-08,
"logits/chosen": -2.7897348403930664,
"logits/rejected": -2.777766704559326,
"logps/chosen": -260.3923645019531,
"logps/rejected": -262.2550964355469,
"loss": 0.6591,
"rewards/accuracies": 0.699999988079071,
"rewards/chosen": 0.034461263567209244,
"rewards/margins": 0.07005412131547928,
"rewards/margins_max": 0.2330094575881958,
"rewards/margins_min": -0.05284310132265091,
"rewards/margins_std": 0.1283414661884308,
"rewards/rejected": -0.03559286147356033,
"step": 580
},
{
"epoch": 0.86,
"grad_norm": 1.9903127787741515,
"learning_rate": 2.8271866952734814e-08,
"logits/chosen": -2.8326079845428467,
"logits/rejected": -2.8090529441833496,
"logps/chosen": -304.954345703125,
"logps/rejected": -290.8445129394531,
"loss": 0.6574,
"rewards/accuracies": 0.7124999761581421,
"rewards/chosen": 0.037466924637556076,
"rewards/margins": 0.07924413681030273,
"rewards/margins_max": 0.22350183129310608,
"rewards/margins_min": -0.030017787590622902,
"rewards/margins_std": 0.11654887348413467,
"rewards/rejected": -0.04177721589803696,
"step": 590
},
{
"epoch": 0.88,
"grad_norm": 2.150802586702874,
"learning_rate": 2.2664399163518782e-08,
"logits/chosen": -2.7674005031585693,
"logits/rejected": -2.7506704330444336,
"logps/chosen": -290.6200256347656,
"logps/rejected": -253.0946044921875,
"loss": 0.6444,
"rewards/accuracies": 0.8125,
"rewards/chosen": 0.06507185846567154,
"rewards/margins": 0.10532490164041519,
"rewards/margins_max": 0.24639275670051575,
"rewards/margins_min": -0.00248835701495409,
"rewards/margins_std": 0.11410228908061981,
"rewards/rejected": -0.040253035724163055,
"step": 600
},
{
"epoch": 0.88,
"eval_logits/chosen": -2.7678329944610596,
"eval_logits/rejected": -2.7307965755462646,
"eval_logps/chosen": -283.465576171875,
"eval_logps/rejected": -261.1460266113281,
"eval_loss": 0.6795856356620789,
"eval_rewards/accuracies": 0.6230000257492065,
"eval_rewards/chosen": 0.011278249323368073,
"eval_rewards/margins": 0.036949291825294495,
"eval_rewards/margins_max": 0.1936866044998169,
"eval_rewards/margins_min": -0.09831613302230835,
"eval_rewards/margins_std": 0.0972999855875969,
"eval_rewards/rejected": -0.025671038776636124,
"eval_runtime": 428.6504,
"eval_samples_per_second": 4.666,
"eval_steps_per_second": 0.292,
"step": 600
},
{
"epoch": 0.89,
"grad_norm": 1.4413450481375123,
"learning_rate": 1.7650024000056414e-08,
"logits/chosen": -2.750030517578125,
"logits/rejected": -2.7483627796173096,
"logps/chosen": -243.78244018554688,
"logps/rejected": -249.83358764648438,
"loss": 0.6525,
"rewards/accuracies": 0.75,
"rewards/chosen": 0.04149278253316879,
"rewards/margins": 0.08110538125038147,
"rewards/margins_max": 0.20331616699695587,
"rewards/margins_min": -0.0176301971077919,
"rewards/margins_std": 0.09766165912151337,
"rewards/rejected": -0.03961259126663208,
"step": 610
},
{
"epoch": 0.91,
"grad_norm": 1.9882715371125113,
"learning_rate": 1.3241823394615437e-08,
"logits/chosen": -2.872929811477661,
"logits/rejected": -2.835972309112549,
"logps/chosen": -306.00555419921875,
"logps/rejected": -293.34014892578125,
"loss": 0.6513,
"rewards/accuracies": 0.8374999761581421,
"rewards/chosen": 0.05589998513460159,
"rewards/margins": 0.10602303594350815,
"rewards/margins_max": 0.22401413321495056,
"rewards/margins_min": -0.0072759948670864105,
"rewards/margins_std": 0.10588717460632324,
"rewards/rejected": -0.05012305825948715,
"step": 620
},
{
"epoch": 0.92,
"grad_norm": 2.1198299570085473,
"learning_rate": 9.451297839253913e-09,
"logits/chosen": -2.749361276626587,
"logits/rejected": -2.7258260250091553,
"logps/chosen": -277.8694763183594,
"logps/rejected": -285.40814208984375,
"loss": 0.6532,
"rewards/accuracies": 0.824999988079071,
"rewards/chosen": 0.05345848202705383,
"rewards/margins": 0.10288135707378387,
"rewards/margins_max": 0.24511775374412537,
"rewards/margins_min": -0.012477993965148926,
"rewards/margins_std": 0.11347142606973648,
"rewards/rejected": -0.04942287132143974,
"step": 630
},
{
"epoch": 0.94,
"grad_norm": 1.5752496491228836,
"learning_rate": 6.288336382349463e-09,
"logits/chosen": -2.767456293106079,
"logits/rejected": -2.73591685295105,
"logps/chosen": -247.0557098388672,
"logps/rejected": -214.10733032226562,
"loss": 0.6441,
"rewards/accuracies": 0.8500000238418579,
"rewards/chosen": 0.04774732142686844,
"rewards/margins": 0.09429731965065002,
"rewards/margins_max": 0.22121095657348633,
"rewards/margins_min": 0.002260456560179591,
"rewards/margins_std": 0.10257305949926376,
"rewards/rejected": -0.046550001949071884,
"step": 640
},
{
"epoch": 0.95,
"grad_norm": 2.9091264169866813,
"learning_rate": 3.7611908292010665e-09,
"logits/chosen": -2.8238768577575684,
"logits/rejected": -2.7892489433288574,
"logps/chosen": -310.2026672363281,
"logps/rejected": -265.02740478515625,
"loss": 0.6485,
"rewards/accuracies": 0.75,
"rewards/chosen": 0.030774693936109543,
"rewards/margins": 0.08213461190462112,
"rewards/margins_max": 0.2235296070575714,
"rewards/margins_min": -0.03506668284535408,
"rewards/margins_std": 0.11517021805047989,
"rewards/rejected": -0.05135990306735039,
"step": 650
},
{
"epoch": 0.96,
"grad_norm": 2.099534541726007,
"learning_rate": 1.8764542140112527e-09,
"logits/chosen": -2.767094135284424,
"logits/rejected": -2.772709369659424,
"logps/chosen": -222.73178100585938,
"logps/rejected": -231.007080078125,
"loss": 0.6568,
"rewards/accuracies": 0.7250000238418579,
"rewards/chosen": 0.018592417240142822,
"rewards/margins": 0.06605438143014908,
"rewards/margins_max": 0.18254120647907257,
"rewards/margins_min": -0.025726070627570152,
"rewards/margins_std": 0.09429889917373657,
"rewards/rejected": -0.047461964190006256,
"step": 660
},
{
"epoch": 0.98,
"grad_norm": 2.1873875180344093,
"learning_rate": 6.390435994127752e-10,
"logits/chosen": -2.793241024017334,
"logits/rejected": -2.760741710662842,
"logps/chosen": -281.3489074707031,
"logps/rejected": -301.64892578125,
"loss": 0.6538,
"rewards/accuracies": 0.762499988079071,
"rewards/chosen": 0.04053753241896629,
"rewards/margins": 0.08871294558048248,
"rewards/margins_max": 0.2136339396238327,
"rewards/margins_min": -0.008021386340260506,
"rewards/margins_std": 0.1009400486946106,
"rewards/rejected": -0.04817541316151619,
"step": 670
},
{
"epoch": 0.99,
"grad_norm": 2.9786376127683343,
"learning_rate": 5.2187248413465555e-11,
"logits/chosen": -2.7850821018218994,
"logits/rejected": -2.7602639198303223,
"logps/chosen": -298.0012512207031,
"logps/rejected": -265.384521484375,
"loss": 0.6519,
"rewards/accuracies": 0.824999988079071,
"rewards/chosen": 0.05158674716949463,
"rewards/margins": 0.10068739950656891,
"rewards/margins_max": 0.258645623922348,
"rewards/margins_min": -0.020790638402104378,
"rewards/margins_std": 0.1269642412662506,
"rewards/rejected": -0.04910064488649368,
"step": 680
},
{
"epoch": 1.0,
"step": 684,
"total_flos": 0.0,
"train_loss": 0.6679948979651021,
"train_runtime": 9131.6092,
"train_samples_per_second": 1.198,
"train_steps_per_second": 0.075
}
],
"logging_steps": 10,
"max_steps": 684,
"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
}