eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": 99.08063507080078,
    "eval_logits/rejected": 99.07344818115234,
    "eval_logps/chosen": -32.34668731689453,
    "eval_logps/rejected": -35.868736267089844,
    "eval_loss": 1.664738655090332,
    "eval_rewards/accuracies": 0.510797381401062,
    "eval_rewards/chosen": 0.08684198558330536,
    "eval_rewards/margins": -0.0011658848961815238,
    "eval_rewards/rejected": 0.08800788223743439,
    "eval_runtime": 103.8246,
    "eval_samples": 343,
    "eval_samples_per_second": 3.304,
    "eval_steps_per_second": 0.414
}