all_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -1.9487414360046387,
    "eval_logits/rejected": -1.7909051179885864,
    "eval_logps/chosen": -291.80194091796875,
    "eval_logps/rejected": -278.0226135253906,
    "eval_loss": 0.05501076579093933,
    "eval_rewards/accuracies": 0.38749998807907104,
    "eval_rewards/chosen": -0.0251043438911438,
    "eval_rewards/margins": -0.0020215215627104044,
    "eval_rewards/rejected": -0.023082822561264038,
    "eval_runtime": 728.1274,
    "eval_samples": 2000,
    "eval_samples_per_second": 2.747,
    "eval_steps_per_second": 1.373,
    "train_loss": 0.3462253446340561,
    "train_runtime": 13267.124,
    "train_samples": 61135,
    "train_samples_per_second": 0.377,
    "train_steps_per_second": 0.094
}