all_results.json

{
    "epoch": 2.994495412844037,
    "eval_logits/chosen": -0.21071076393127441,
    "eval_logits/rejected": -1.2749640941619873,
    "eval_logps/chosen": -291.2645263671875,
    "eval_logps/rejected": -212.66262817382812,
    "eval_loss": 0.06446235626935959,
    "eval_rewards/accuracies": 1.0,
    "eval_rewards/chosen": 1.6396679878234863,
    "eval_rewards/margins": 4.35139274597168,
    "eval_rewards/rejected": -2.7117245197296143,
    "eval_runtime": 8.9352,
    "eval_samples_per_second": 5.036,
    "eval_steps_per_second": 5.036,
    "total_flos": 7.837376281021809e+17,
    "train_loss": 0.2111055671474805,
    "train_runtime": 8097.1834,
    "train_samples_per_second": 1.614,
    "train_steps_per_second": 0.05
}