all_results.json

{
    "epoch": 2.999098751126561,
    "eval_kl/chosen": 0.0006424170569516718,
    "eval_kl/rejected": 0.00019059143960475922,
    "eval_logits/chosen": -1.2265625,
    "eval_logits/rejected": -1.1875,
    "eval_logps/chosen": -324.0,
    "eval_logps/rejected": -298.0,
    "eval_loss": 0.8567168116569519,
    "eval_nll_loss": 1.0546875,
    "eval_ref_probs/chosen": 0.0038271306548267603,
    "eval_ref_probs/rejected": 0.0034569373819977045,
    "eval_rejected_term/max": 0.00019059143960475922,
    "eval_rejected_term/min": 0.00019059143960475922,
    "eval_rewards/accuracies": 0.36800000071525574,
    "eval_rewards/chosen": -0.294921875,
    "eval_rewards/margins": -0.142578125,
    "eval_rewards/rejected": -0.1533203125,
    "eval_runtime": 25.3543,
    "eval_samples_per_second": 39.441,
    "eval_steps_per_second": 9.86
}