all_results.json

{
    "epoch": 2.998430141287284,
    "eval_logits/chosen": -4.0,
    "eval_logits/rejected": -1.578125,
    "eval_logps/chosen": -1056.0,
    "eval_logps/rejected": -1248.0,
    "eval_loss": 0.812749981880188,
    "eval_rewards/accuracies": 0.7265625,
    "eval_rewards/chosen": -7.40625,
    "eval_rewards/margins": 2.21875,
    "eval_rewards/rejected": -9.625,
    "eval_runtime": 46.6963,
    "eval_samples": 2000,
    "eval_samples_per_second": 42.83,
    "eval_steps_per_second": 0.685,
    "total_flos": 0.0,
    "train_loss": 0.26550976372930185,
    "train_runtime": 11897.7673,
    "train_samples": 61134,
    "train_samples_per_second": 15.415,
    "train_steps_per_second": 0.241
}