eval_results.json

{
    "epoch": 0.9982851866508377,
    "eval_logits/chosen": -3.2231318950653076,
    "eval_logits/rejected": -3.1088078022003174,
    "eval_logps/chosen": -623.1680297851562,
    "eval_logps/rejected": -685.9901733398438,
    "eval_loss": 0.6219658255577087,
    "eval_rewards/accuracies": 0.6350806355476379,
    "eval_rewards/chosen": -3.605348587036133,
    "eval_rewards/margins": 0.5176258683204651,
    "eval_rewards/rejected": -4.122974872589111,
    "eval_runtime": 146.9573,
    "eval_samples": 1976,
    "eval_samples_per_second": 13.446,
    "eval_steps_per_second": 0.844
}