all_results.json

{
    "epoch": 0.995910949568378,
    "eval_logits/chosen": -6.920379161834717,
    "eval_logits/rejected": -6.706053733825684,
    "eval_logps/chosen": -0.6168988347053528,
    "eval_logps/rejected": -0.7138991951942444,
    "eval_loss": 4.313390731811523,
    "eval_rewards/accuracies": 0.6024590134620667,
    "eval_rewards/chosen": -6.168987274169922,
    "eval_rewards/margins": 0.9700047373771667,
    "eval_rewards/rejected": -7.138991832733154,
    "eval_runtime": 65.9401,
    "eval_samples": 1941,
    "eval_samples_per_second": 29.436,
    "eval_steps_per_second": 1.85,
    "total_flos": 0.0,
    "train_loss": 4.287031420826041,
    "train_runtime": 1653.4486,
    "train_samples": 17605,
    "train_samples_per_second": 10.647,
    "train_steps_per_second": 0.083
}