eval_results.json

{
    "epoch": 0.9984168865435357,
    "eval_logits/chosen": -2.766282558441162,
    "eval_logits/rejected": -2.6962404251098633,
    "eval_logps/chosen": -612.2565307617188,
    "eval_logps/rejected": -809.2787475585938,
    "eval_loss": 0.3973737061023712,
    "eval_rewards/accuracies": 0.8286290168762207,
    "eval_rewards/chosen": -3.4936444759368896,
    "eval_rewards/margins": 1.8601694107055664,
    "eval_rewards/rejected": -5.353814125061035,
    "eval_runtime": 316.1088,
    "eval_samples": 1976,
    "eval_samples_per_second": 6.251,
    "eval_steps_per_second": 0.392
}