approx_nash_again_iter_3 / trainer_state.json
YYYYYYibo's picture
Model save
37ab888 verified
{
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 0.9992429977289932,
"eval_steps": 500,
"global_step": 165,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.01,
"eta": 0.0010000000474974513,
"grad_norm": 19.93140330088016,
"learning_rate": 2.941176470588235e-08,
"logits/chosen": -2.397038221359253,
"logits/rejected": -2.213353395462036,
"logps/chosen": -180.87660217285156,
"logps/pi_response": -160.3468780517578,
"logps/ref_response": -160.3468780517578,
"logps/rejected": -188.15975952148438,
"loss": 0.6931,
"rewards/accuracies": 0.0,
"rewards/chosen": 0.0,
"rewards/margins": 0.0,
"rewards/rejected": 0.0,
"step": 1
},
{
"epoch": 0.06,
"eta": 0.0010000000474974513,
"grad_norm": 19.32986244730549,
"learning_rate": 2.941176470588235e-07,
"logits/chosen": -2.334710121154785,
"logits/rejected": -2.268401861190796,
"logps/chosen": -204.74749755859375,
"logps/pi_response": -172.67669677734375,
"logps/ref_response": -171.70980834960938,
"logps/rejected": -210.5995330810547,
"loss": 0.6931,
"rewards/accuracies": 0.4930555522441864,
"rewards/chosen": -0.009038920514285564,
"rewards/margins": 0.0021236613392829895,
"rewards/rejected": -0.011162581853568554,
"step": 10
},
{
"epoch": 0.12,
"eta": 0.0010000000474974513,
"grad_norm": 38.14550506704795,
"learning_rate": 4.994932636402031e-07,
"logits/chosen": -2.1406025886535645,
"logits/rejected": -2.2008185386657715,
"logps/chosen": -219.163330078125,
"logps/pi_response": -189.20216369628906,
"logps/ref_response": -168.5287628173828,
"logps/rejected": -222.94271850585938,
"loss": 0.6888,
"rewards/accuracies": 0.550000011920929,
"rewards/chosen": -0.19716469943523407,
"rewards/margins": 0.0258896853774786,
"rewards/rejected": -0.22305437922477722,
"step": 20
},
{
"epoch": 0.18,
"eta": 0.0010000000474974513,
"grad_norm": 17.675340945525093,
"learning_rate": 4.905416503522123e-07,
"logits/chosen": -2.074061870574951,
"logits/rejected": -2.080552101135254,
"logps/chosen": -198.05816650390625,
"logps/pi_response": -164.6136932373047,
"logps/ref_response": -160.591552734375,
"logps/rejected": -201.87464904785156,
"loss": 0.6934,
"rewards/accuracies": 0.515625,
"rewards/chosen": -0.06703463941812515,
"rewards/margins": 0.023267237469553947,
"rewards/rejected": -0.09030187875032425,
"step": 30
},
{
"epoch": 0.24,
"eta": 0.0010000000474974513,
"grad_norm": 20.77937618350303,
"learning_rate": 4.707922373336523e-07,
"logits/chosen": -2.1749043464660645,
"logits/rejected": -2.1247994899749756,
"logps/chosen": -209.81790161132812,
"logps/pi_response": -183.0597381591797,
"logps/ref_response": -177.06118774414062,
"logps/rejected": -218.8212890625,
"loss": 0.6946,
"rewards/accuracies": 0.53125,
"rewards/chosen": -0.03284044936299324,
"rewards/margins": 0.03580809757113457,
"rewards/rejected": -0.06864854693412781,
"step": 40
},
{
"epoch": 0.3,
"eta": 0.0010000000474974513,
"grad_norm": 21.485794770779655,
"learning_rate": 4.4113156629677313e-07,
"logits/chosen": -2.161895990371704,
"logits/rejected": -2.047234058380127,
"logps/chosen": -253.85107421875,
"logps/pi_response": -220.5885467529297,
"logps/ref_response": -183.68719482421875,
"logps/rejected": -255.03494262695312,
"loss": 0.6961,
"rewards/accuracies": 0.5249999761581421,
"rewards/chosen": -0.3808112144470215,
"rewards/margins": 0.016062479466199875,
"rewards/rejected": -0.3968736529350281,
"step": 50
},
{
"epoch": 0.36,
"eta": 0.0010000000474974513,
"grad_norm": 20.861550744842482,
"learning_rate": 4.0289109058972283e-07,
"logits/chosen": -2.0308499336242676,
"logits/rejected": -2.0642848014831543,
"logps/chosen": -249.21365356445312,
"logps/pi_response": -218.06405639648438,
"logps/ref_response": -171.6521453857422,
"logps/rejected": -256.102294921875,
"loss": 0.6918,
"rewards/accuracies": 0.5249999761581421,
"rewards/chosen": -0.45084279775619507,
"rewards/margins": 0.007402978837490082,
"rewards/rejected": -0.45824581384658813,
"step": 60
},
{
"epoch": 0.42,
"eta": 0.0010000000474974513,
"grad_norm": 28.00866705635057,
"learning_rate": 3.577874068920446e-07,
"logits/chosen": -1.9535369873046875,
"logits/rejected": -1.9607006311416626,
"logps/chosen": -277.5329284667969,
"logps/pi_response": -250.90200805664062,
"logps/ref_response": -169.15725708007812,
"logps/rejected": -291.10015869140625,
"loss": 0.6856,
"rewards/accuracies": 0.550000011920929,
"rewards/chosen": -0.7873853445053101,
"rewards/margins": 0.044238921254873276,
"rewards/rejected": -0.8316243290901184,
"step": 70
},
{
"epoch": 0.48,
"eta": 0.0010000000474974513,
"grad_norm": 30.93214776345345,
"learning_rate": 3.078451980100854e-07,
"logits/chosen": -1.9619076251983643,
"logits/rejected": -1.8502511978149414,
"logps/chosen": -281.6351318359375,
"logps/pi_response": -248.10684204101562,
"logps/ref_response": -168.71682739257812,
"logps/rejected": -294.75604248046875,
"loss": 0.6782,
"rewards/accuracies": 0.581250011920929,
"rewards/chosen": -0.7786952257156372,
"rewards/margins": 0.08077356964349747,
"rewards/rejected": -0.8594688177108765,
"step": 80
},
{
"epoch": 0.55,
"eta": 0.0010000000474974513,
"grad_norm": 25.352348701139025,
"learning_rate": 2.553063458334059e-07,
"logits/chosen": -1.952636957168579,
"logits/rejected": -2.0428998470306396,
"logps/chosen": -253.1057586669922,
"logps/pi_response": -229.44277954101562,
"logps/ref_response": -163.27528381347656,
"logps/rejected": -273.68780517578125,
"loss": 0.6755,
"rewards/accuracies": 0.546875,
"rewards/chosen": -0.6530182361602783,
"rewards/margins": 0.06835106015205383,
"rewards/rejected": -0.7213693261146545,
"step": 90
},
{
"epoch": 0.61,
"eta": 0.0010000000474974513,
"grad_norm": 30.010292143430778,
"learning_rate": 2.0252929432814287e-07,
"logits/chosen": -2.0059685707092285,
"logits/rejected": -1.9757779836654663,
"logps/chosen": -277.2453308105469,
"logps/pi_response": -250.5970458984375,
"logps/ref_response": -177.19424438476562,
"logps/rejected": -285.8126525878906,
"loss": 0.6861,
"rewards/accuracies": 0.5562499761581421,
"rewards/chosen": -0.7217534780502319,
"rewards/margins": 0.028179144486784935,
"rewards/rejected": -0.7499326467514038,
"step": 100
},
{
"epoch": 0.67,
"eta": 0.0010000000474974513,
"grad_norm": 24.160814982479724,
"learning_rate": 1.5188318011445906e-07,
"logits/chosen": -1.880448579788208,
"logits/rejected": -1.9354912042617798,
"logps/chosen": -272.9020080566406,
"logps/pi_response": -249.7644805908203,
"logps/ref_response": -170.73162841796875,
"logps/rejected": -296.87628173828125,
"loss": 0.6738,
"rewards/accuracies": 0.612500011920929,
"rewards/chosen": -0.7538261413574219,
"rewards/margins": 0.10054464638233185,
"rewards/rejected": -0.8543707132339478,
"step": 110
},
{
"epoch": 0.73,
"eta": 0.0010000000474974513,
"grad_norm": 21.79427862950612,
"learning_rate": 1.0564148305586295e-07,
"logits/chosen": -1.8592274188995361,
"logits/rejected": -1.8245065212249756,
"logps/chosen": -286.49853515625,
"logps/pi_response": -259.8140869140625,
"logps/ref_response": -175.83364868164062,
"logps/rejected": -298.8387756347656,
"loss": 0.6773,
"rewards/accuracies": 0.546875,
"rewards/chosen": -0.8258917927742004,
"rewards/margins": 0.05054790526628494,
"rewards/rejected": -0.8764396905899048,
"step": 120
},
{
"epoch": 0.79,
"eta": 0.0010000000474974513,
"grad_norm": 19.644675719528685,
"learning_rate": 6.587997083462196e-08,
"logits/chosen": -2.0082201957702637,
"logits/rejected": -1.852614164352417,
"logps/chosen": -262.72113037109375,
"logps/pi_response": -234.92135620117188,
"logps/ref_response": -170.3947296142578,
"logps/rejected": -271.8548278808594,
"loss": 0.6733,
"rewards/accuracies": 0.546875,
"rewards/chosen": -0.6122658848762512,
"rewards/margins": 0.0654686689376831,
"rewards/rejected": -0.6777344942092896,
"step": 130
},
{
"epoch": 0.85,
"eta": 0.0010000000474974513,
"grad_norm": 23.231586788018557,
"learning_rate": 3.438351873250492e-08,
"logits/chosen": -2.007563829421997,
"logits/rejected": -1.9652036428451538,
"logps/chosen": -268.14642333984375,
"logps/pi_response": -241.33349609375,
"logps/ref_response": -177.7176971435547,
"logps/rejected": -281.2812805175781,
"loss": 0.6679,
"rewards/accuracies": 0.5687500238418579,
"rewards/chosen": -0.6138588190078735,
"rewards/margins": 0.07413015514612198,
"rewards/rejected": -0.6879889369010925,
"step": 140
},
{
"epoch": 0.91,
"eta": 0.0010000000474974513,
"grad_norm": 22.182591656592486,
"learning_rate": 1.256598743236703e-08,
"logits/chosen": -1.9013135433197021,
"logits/rejected": -1.9633811712265015,
"logps/chosen": -258.21563720703125,
"logps/pi_response": -229.17837524414062,
"logps/ref_response": -161.02003479003906,
"logps/rejected": -270.0240478515625,
"loss": 0.6754,
"rewards/accuracies": 0.5718749761581421,
"rewards/chosen": -0.6737670302391052,
"rewards/margins": 0.05831047147512436,
"rewards/rejected": -0.7320775985717773,
"step": 150
},
{
"epoch": 0.97,
"eta": 0.0010000000474974513,
"grad_norm": 28.927536436184628,
"learning_rate": 1.406755487774386e-09,
"logits/chosen": -1.9876502752304077,
"logits/rejected": -1.8038742542266846,
"logps/chosen": -269.11968994140625,
"logps/pi_response": -239.73095703125,
"logps/ref_response": -171.18202209472656,
"logps/rejected": -278.70245361328125,
"loss": 0.6711,
"rewards/accuracies": 0.643750011920929,
"rewards/chosen": -0.6629125475883484,
"rewards/margins": 0.0806988924741745,
"rewards/rejected": -0.7436113953590393,
"step": 160
},
{
"epoch": 1.0,
"step": 165,
"total_flos": 0.0,
"train_loss": 0.6822745196747058,
"train_runtime": 33626.4283,
"train_samples_per_second": 0.629,
"train_steps_per_second": 0.005
}
],
"logging_steps": 10,
"max_steps": 165,
"num_input_tokens_seen": 0,
"num_train_epochs": 1,
"save_steps": 200,
"total_flos": 0.0,
"train_batch_size": 1,
"trial_name": null,
"trial_params": null
}