eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -2.244168758392334,
    "eval_logits/rejected": -2.239292860031128,
    "eval_logps/chosen": -34.02364730834961,
    "eval_logps/rejected": -37.501853942871094,
    "eval_loss": 0.6989675760269165,
    "eval_rewards/accuracies": 0.49958473443984985,
    "eval_rewards/chosen": 0.009816422127187252,
    "eval_rewards/margins": -0.0034731123596429825,
    "eval_rewards/rejected": 0.01328953169286251,
    "eval_runtime": 145.7819,
    "eval_samples": 343,
    "eval_samples_per_second": 2.353,
    "eval_steps_per_second": 0.295
}