eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -0.4925585687160492,
    "eval_logits/rejected": -0.15963555872440338,
    "eval_logps/chosen": -334.72607421875,
    "eval_logps/rejected": -671.7160034179688,
    "eval_loss": 0.6914218664169312,
    "eval_rewards/accuracies": 0.6104797720909119,
    "eval_rewards/chosen": -0.001606568112038076,
    "eval_rewards/margins": 0.003300695214420557,
    "eval_rewards/rejected": -0.004907263442873955,
    "eval_runtime": 630.223,
    "eval_samples": 6317,
    "eval_samples_per_second": 10.023,
    "eval_steps_per_second": 0.314
}