all_results.json

{
    "epoch": 2.9158878504672896,
    "eval_logits/chosen": -2.5211000442504883,
    "eval_logits/rejected": -2.522141456604004,
    "eval_logps/chosen": -272.1190490722656,
    "eval_logps/rejected": -251.88446044921875,
    "eval_loss": 0.5770149230957031,
    "eval_rewards/accuracies": 0.78125,
    "eval_rewards/chosen": 1.9856371879577637,
    "eval_rewards/margins": 2.3065383434295654,
    "eval_rewards/rejected": -0.3209007680416107,
    "eval_runtime": 49.4101,
    "eval_samples_per_second": 15.381,
    "eval_steps_per_second": 0.243,
    "total_flos": 919378820333568.0,
    "train_loss": 0.3579892989916679,
    "train_runtime": 2691.602,
    "train_samples_per_second": 7.621,
    "train_steps_per_second": 0.029
}