all_results.json

{
    "epoch": 3.0,
    "eval_logits/chosen": -2.8751940727233887,
    "eval_logits/rejected": -2.8561081886291504,
    "eval_logps/chosen": -253.02841186523438,
    "eval_logps/rejected": -269.08563232421875,
    "eval_loss": 0.6063873767852783,
    "eval_rewards/accuracies": 0.6825000047683716,
    "eval_rewards/chosen": -6.233180522918701,
    "eval_rewards/margins": 3.6988027095794678,
    "eval_rewards/rejected": -9.931983947753906,
    "eval_runtime": 168.7249,
    "eval_samples": 3156,
    "eval_samples_per_second": 18.705,
    "eval_steps_per_second": 0.296,
    "train_loss": 0.25373340110176973,
    "train_runtime": 3870.3099,
    "train_samples": 8239,
    "train_samples_per_second": 6.386,
    "train_steps_per_second": 0.399
}