all_results.json

{
    "epoch": 0.9974726200505476,
    "eval_kl": 0.04474467411637306,
    "eval_logps/chosen": -12.086200252242152,
    "eval_logps/rejected": -16.140498082129962,
    "eval_loss": 0.5002864599227905,
    "eval_rewards/chosen": 0.006133331311657824,
    "eval_rewards/margins": 0.003854976257181193,
    "eval_rewards/rejected": 0.0022783550544766312,
    "eval_runtime": 116.8015,
    "eval_samples_per_second": 4.281,
    "eval_steps_per_second": 1.07,
    "total_flos": 2.1537474524636774e+17,
    "train_loss": 0.5015917984214989,
    "train_runtime": 4835.0561,
    "train_samples_per_second": 1.964,
    "train_steps_per_second": 0.031
}