eval_results.json

{
    "epoch": 0.9994944388270981,
    "eval_chosen_logps": -94.18656158447266,
    "eval_chosen_rewards": 0.024786589667201042,
    "eval_log_diff_policy": 2.3620760440826416,
    "eval_logits": -2.1694083213806152,
    "eval_logp_accuracy": 0.5450223684310913,
    "eval_loss": 8.375486373901367,
    "eval_objective": 8.485334396362305,
    "eval_rejected_logps": -96.54862976074219,
    "eval_rejected_rewards": -0.07420787215232849,
    "eval_reward_accuracy": 0.6261186003684998,
    "eval_runtime": 1121.9363,
    "eval_samples": 10722,
    "eval_samples_per_second": 9.557,
    "eval_steps_per_second": 1.062
}