all_results.json

{
    "epoch": 2.999098751126561,
    "eval_kl/chosen": 0.00028143005329184234,
    "eval_kl/rejected": 0.0005092654610052705,
    "eval_logits/chosen": -1.40625,
    "eval_logits/rejected": -1.375,
    "eval_logps/chosen": -322.0,
    "eval_logps/rejected": -298.0,
    "eval_loss": 0.9176015853881836,
    "eval_nll_loss": 1.0546875,
    "eval_ref_probs/chosen": 0.0038271306548267603,
    "eval_ref_probs/rejected": 0.0034569373819977045,
    "eval_rejected_term/max": 0.0005092654610052705,
    "eval_rejected_term/min": 0.0005092654610052705,
    "eval_rewards/accuracies": 0.5720000267028809,
    "eval_rewards/chosen": -0.049072265625,
    "eval_rewards/margins": 0.10546875,
    "eval_rewards/rejected": -0.154296875,
    "eval_runtime": 25.2526,
    "eval_samples_per_second": 39.6,
    "eval_steps_per_second": 9.9
}