all_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": 1.0391736030578613,
    "eval_logits/rejected": 1.7624521255493164,
    "eval_logps/chosen": -388.8264465332031,
    "eval_logps/rejected": -460.791259765625,
    "eval_loss": 0.5025966167449951,
    "eval_rewards/accuracies": 0.761904776096344,
    "eval_rewards/chosen": -1.0471669435501099,
    "eval_rewards/margins": 0.9429601430892944,
    "eval_rewards/rejected": -1.9901270866394043,
    "eval_runtime": 213.579,
    "eval_samples": 2000,
    "eval_samples_per_second": 9.364,
    "eval_steps_per_second": 0.295,
    "train_loss": 0.540120158305458,
    "train_runtime": 18392.9814,
    "train_samples": 61135,
    "train_samples_per_second": 3.324,
    "train_steps_per_second": 0.026
}