all_results.json

{
    "epoch": 2.999098751126561,
    "eval_kl/both": 0.012443912215530872,
    "eval_kl/chosen": 0.006274462211877108,
    "eval_kl/rejected": 0.0061685326509177685,
    "eval_logits/chosen": -1.8515625,
    "eval_logits/rejected": -1.859375,
    "eval_logps/chosen": -328.0,
    "eval_logps/rejected": -310.0,
    "eval_loss": 1.6035839319229126,
    "eval_nll_loss": 1.0703125,
    "eval_ref_probs/chosen": 0.0038271306548267603,
    "eval_ref_probs/rejected": 0.0034569373819977045,
    "eval_rejected_term/max": 0.0061685326509177685,
    "eval_rejected_term/min": 0.0061685326509177685,
    "eval_rewards/accuracies": 0.7400000095367432,
    "eval_rewards/chosen": -0.578125,
    "eval_rewards/margins": 0.734375,
    "eval_rewards/rejected": -1.3125,
    "eval_runtime": 25.2611,
    "eval_samples_per_second": 39.587,
    "eval_steps_per_second": 9.897
}