all_results.json

{
    "epoch": 5.0,
    "eval_logits/chosen": -1.844208002090454,
    "eval_logits/rejected": -1.731345772743225,
    "eval_logps/chosen": -468.6473388671875,
    "eval_logps/rejected": -486.3179626464844,
    "eval_loss": 0.573898196220398,
    "eval_rewards/accuracies": 0.716269850730896,
    "eval_rewards/chosen": -1.1928762197494507,
    "eval_rewards/margins": 0.5913423895835876,
    "eval_rewards/rejected": -1.7842185497283936,
    "eval_runtime": 226.0135,
    "eval_samples": 2000,
    "eval_samples_per_second": 8.849,
    "eval_steps_per_second": 0.279,
    "train_loss": 0.15748632995245967,
    "train_runtime": 23969.9997,
    "train_samples": 61135,
    "train_samples_per_second": 12.752,
    "train_steps_per_second": 0.199
}