all_results.json

{
    "epoch": 2.0,
    "eval_logits/chosen": -2.5584018230438232,
    "eval_logits/rejected": -2.5515997409820557,
    "eval_logps/chosen": -108.46733856201172,
    "eval_logps/rejected": -123.28102111816406,
    "eval_loss": 0.6549462676048279,
    "eval_rewards/accuracies": 0.6194238066673279,
    "eval_rewards/chosen": -0.49755439162254333,
    "eval_rewards/margins": 0.10345453023910522,
    "eval_rewards/rejected": -0.6010088920593262,
    "eval_runtime": 359.6814,
    "eval_samples": 4304,
    "eval_samples_per_second": 11.966,
    "eval_steps_per_second": 1.496,
    "total_flos": 0.0,
    "train_loss": 0.6539983197297005,
    "train_runtime": 91180.3592,
    "train_samples": 92858,
    "train_samples_per_second": 2.037,
    "train_steps_per_second": 0.127
}