all_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": 0.19371525943279266,
    "eval_logits/rejected": 0.24551750719547272,
    "eval_logps/chosen": -0.5973814725875854,
    "eval_logps/rejected": -0.7979374527931213,
    "eval_loss": 1.46084463596344,
    "eval_rewards/accuracies": 0.6290322542190552,
    "eval_rewards/chosen": -1.4934533834457397,
    "eval_rewards/margins": 0.5013901591300964,
    "eval_rewards/rejected": -1.9948437213897705,
    "eval_runtime": 100.7391,
    "eval_samples": 1961,
    "eval_samples_per_second": 19.466,
    "eval_steps_per_second": 0.308,
    "total_flos": 0.0,
    "train_loss": 1.5168422256779468,
    "train_runtime": 9435.3453,
    "train_samples": 59876,
    "train_samples_per_second": 6.346,
    "train_steps_per_second": 0.025
}