ShenaoZ's picture
Model save
842684f verified
{
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 0.9968652037617555,
"eval_steps": 500,
"global_step": 159,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.006269592476489028,
"grad_norm": 18.55070569719312,
"learning_rate": 3.125e-08,
"logits/chosen": -2.554591178894043,
"logits/rejected": -2.434877395629883,
"logps/chosen": -323.11175537109375,
"logps/pi_response": -94.49095153808594,
"logps/ref_response": -94.49095153808594,
"logps/rejected": -208.82200622558594,
"loss": 0.6931,
"rewards/accuracies": 0.0,
"rewards/chosen": 0.0,
"rewards/margins": 0.0,
"rewards/rejected": 0.0,
"step": 1
},
{
"epoch": 0.06269592476489028,
"grad_norm": 17.53145291009543,
"learning_rate": 3.1249999999999997e-07,
"logits/chosen": -2.4522640705108643,
"logits/rejected": -2.350480079650879,
"logps/chosen": -193.23435974121094,
"logps/pi_response": -86.61455535888672,
"logps/ref_response": -86.81534576416016,
"logps/rejected": -259.4031066894531,
"loss": 0.6916,
"rewards/accuracies": 0.5763888955116272,
"rewards/chosen": -0.0016433527925983071,
"rewards/margins": 0.0050659943372011185,
"rewards/rejected": -0.0067093465477228165,
"step": 10
},
{
"epoch": 0.12539184952978055,
"grad_norm": 23.589101524345324,
"learning_rate": 4.990353313429303e-07,
"logits/chosen": -2.4247384071350098,
"logits/rejected": -2.4102516174316406,
"logps/chosen": -214.6630096435547,
"logps/pi_response": -80.90936279296875,
"logps/ref_response": -82.33551788330078,
"logps/rejected": -277.02093505859375,
"loss": 0.6494,
"rewards/accuracies": 0.768750011920929,
"rewards/chosen": -0.09652389585971832,
"rewards/margins": 0.16998200118541718,
"rewards/rejected": -0.2665058970451355,
"step": 20
},
{
"epoch": 0.18808777429467086,
"grad_norm": 34.67188819309952,
"learning_rate": 4.882681251368548e-07,
"logits/chosen": -2.363760471343994,
"logits/rejected": -2.2931735515594482,
"logps/chosen": -265.7935485839844,
"logps/pi_response": -118.18830871582031,
"logps/ref_response": -89.96348571777344,
"logps/rejected": -357.9932556152344,
"loss": 0.5909,
"rewards/accuracies": 0.6812499761581421,
"rewards/chosen": -0.6331563591957092,
"rewards/margins": 0.4989984929561615,
"rewards/rejected": -1.132154941558838,
"step": 30
},
{
"epoch": 0.2507836990595611,
"grad_norm": 24.39087061903397,
"learning_rate": 4.6604720940421207e-07,
"logits/chosen": -2.3351082801818848,
"logits/rejected": -2.240893602371216,
"logps/chosen": -282.7699890136719,
"logps/pi_response": -137.54034423828125,
"logps/ref_response": -89.04374694824219,
"logps/rejected": -359.68048095703125,
"loss": 0.5472,
"rewards/accuracies": 0.6812499761581421,
"rewards/chosen": -0.6459968686103821,
"rewards/margins": 0.5398008227348328,
"rewards/rejected": -1.1857976913452148,
"step": 40
},
{
"epoch": 0.31347962382445144,
"grad_norm": 26.791698094673826,
"learning_rate": 4.3344075855595097e-07,
"logits/chosen": -2.3063526153564453,
"logits/rejected": -2.2795004844665527,
"logps/chosen": -300.38531494140625,
"logps/pi_response": -154.4958953857422,
"logps/ref_response": -83.67357635498047,
"logps/rejected": -391.9969177246094,
"loss": 0.519,
"rewards/accuracies": 0.7250000238418579,
"rewards/chosen": -0.7796777486801147,
"rewards/margins": 0.6542943120002747,
"rewards/rejected": -1.4339721202850342,
"step": 50
},
{
"epoch": 0.3761755485893417,
"grad_norm": 38.46368190958448,
"learning_rate": 3.920161866827889e-07,
"logits/chosen": -2.1805338859558105,
"logits/rejected": -2.1266684532165527,
"logps/chosen": -285.82476806640625,
"logps/pi_response": -186.3391876220703,
"logps/ref_response": -86.60485076904297,
"logps/rejected": -381.95941162109375,
"loss": 0.5073,
"rewards/accuracies": 0.768750011920929,
"rewards/chosen": -0.9567915201187134,
"rewards/margins": 0.6549071073532104,
"rewards/rejected": -1.6116985082626343,
"step": 60
},
{
"epoch": 0.438871473354232,
"grad_norm": 33.61675096777802,
"learning_rate": 3.4376480090239047e-07,
"logits/chosen": -2.225067138671875,
"logits/rejected": -2.197097063064575,
"logps/chosen": -311.00189208984375,
"logps/pi_response": -195.21932983398438,
"logps/ref_response": -87.1160659790039,
"logps/rejected": -443.4674377441406,
"loss": 0.4642,
"rewards/accuracies": 0.7875000238418579,
"rewards/chosen": -1.0373988151550293,
"rewards/margins": 0.9388235211372375,
"rewards/rejected": -1.976222276687622,
"step": 70
},
{
"epoch": 0.5015673981191222,
"grad_norm": 32.26496956211775,
"learning_rate": 2.910060778827554e-07,
"logits/chosen": -2.2063775062561035,
"logits/rejected": -2.1627447605133057,
"logps/chosen": -366.4414367675781,
"logps/pi_response": -222.93441772460938,
"logps/ref_response": -94.23348236083984,
"logps/rejected": -445.9207458496094,
"loss": 0.4839,
"rewards/accuracies": 0.768750011920929,
"rewards/chosen": -1.322712779045105,
"rewards/margins": 0.9108338356018066,
"rewards/rejected": -2.233546495437622,
"step": 80
},
{
"epoch": 0.5642633228840125,
"grad_norm": 26.726350157115625,
"learning_rate": 2.3627616503391812e-07,
"logits/chosen": -2.0913848876953125,
"logits/rejected": -2.097029209136963,
"logps/chosen": -338.84674072265625,
"logps/pi_response": -222.7672119140625,
"logps/ref_response": -87.508056640625,
"logps/rejected": -487.66766357421875,
"loss": 0.4588,
"rewards/accuracies": 0.78125,
"rewards/chosen": -1.2909362316131592,
"rewards/margins": 0.9323557019233704,
"rewards/rejected": -2.2232918739318848,
"step": 90
},
{
"epoch": 0.6269592476489029,
"grad_norm": 36.897937868099504,
"learning_rate": 1.8220596619089573e-07,
"logits/chosen": -2.0451536178588867,
"logits/rejected": -2.017054319381714,
"logps/chosen": -363.85626220703125,
"logps/pi_response": -238.09249877929688,
"logps/ref_response": -86.1615219116211,
"logps/rejected": -467.4317321777344,
"loss": 0.4936,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -1.4615881443023682,
"rewards/margins": 0.8330486416816711,
"rewards/rejected": -2.2946369647979736,
"step": 100
},
{
"epoch": 0.6896551724137931,
"grad_norm": 36.50147029957021,
"learning_rate": 1.3139467229135998e-07,
"logits/chosen": -1.8684625625610352,
"logits/rejected": -1.8239994049072266,
"logps/chosen": -318.9664001464844,
"logps/pi_response": -232.65066528320312,
"logps/ref_response": -80.07207489013672,
"logps/rejected": -486.52655029296875,
"loss": 0.4637,
"rewards/accuracies": 0.75,
"rewards/chosen": -1.443070650100708,
"rewards/margins": 1.0364656448364258,
"rewards/rejected": -2.479536294937134,
"step": 110
},
{
"epoch": 0.7523510971786834,
"grad_norm": 32.80757403172816,
"learning_rate": 8.628481651367875e-08,
"logits/chosen": -1.8891398906707764,
"logits/rejected": -1.7928073406219482,
"logps/chosen": -345.98785400390625,
"logps/pi_response": -232.44729614257812,
"logps/ref_response": -78.10755920410156,
"logps/rejected": -474.845703125,
"loss": 0.4495,
"rewards/accuracies": 0.8062499761581421,
"rewards/chosen": -1.4001835584640503,
"rewards/margins": 1.0585042238235474,
"rewards/rejected": -2.4586877822875977,
"step": 120
},
{
"epoch": 0.8150470219435737,
"grad_norm": 31.6821642347826,
"learning_rate": 4.904486005914027e-08,
"logits/chosen": -1.808752417564392,
"logits/rejected": -1.694044828414917,
"logps/chosen": -374.7220764160156,
"logps/pi_response": -225.9062957763672,
"logps/ref_response": -80.09992218017578,
"logps/rejected": -484.637451171875,
"loss": 0.4598,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -1.4728864431381226,
"rewards/margins": 0.967255711555481,
"rewards/rejected": -2.4401423931121826,
"step": 130
},
{
"epoch": 0.877742946708464,
"grad_norm": 33.378981761704864,
"learning_rate": 2.1464952759020856e-08,
"logits/chosen": -1.7705777883529663,
"logits/rejected": -1.6775932312011719,
"logps/chosen": -368.3900146484375,
"logps/pi_response": -242.3651580810547,
"logps/ref_response": -80.24095916748047,
"logps/rejected": -515.4364624023438,
"loss": 0.4255,
"rewards/accuracies": 0.824999988079071,
"rewards/chosen": -1.5394630432128906,
"rewards/margins": 1.2102196216583252,
"rewards/rejected": -2.7496824264526367,
"step": 140
},
{
"epoch": 0.9404388714733543,
"grad_norm": 32.52480870828172,
"learning_rate": 4.8708793644441086e-09,
"logits/chosen": -1.8091413974761963,
"logits/rejected": -1.7029527425765991,
"logps/chosen": -412.6189880371094,
"logps/pi_response": -255.9366455078125,
"logps/ref_response": -90.39550018310547,
"logps/rejected": -488.56060791015625,
"loss": 0.437,
"rewards/accuracies": 0.75,
"rewards/chosen": -1.6722618341445923,
"rewards/margins": 0.8418585062026978,
"rewards/rejected": -2.514120101928711,
"step": 150
},
{
"epoch": 0.9968652037617555,
"step": 159,
"total_flos": 0.0,
"train_loss": 0.5087880788359253,
"train_runtime": 3966.6039,
"train_samples_per_second": 5.137,
"train_steps_per_second": 0.04
}
],
"logging_steps": 10,
"max_steps": 159,
"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
}