eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -1.4037474393844604,
    "eval_logits/rejected": -1.0336650609970093,
    "eval_logps/chosen": -150.22137451171875,
    "eval_logps/rejected": -653.139404296875,
    "eval_loss": 0.6249053478240967,
    "eval_rewards/accuracies": 0.875,
    "eval_rewards/chosen": -0.014455685392022133,
    "eval_rewards/margins": 0.18389175832271576,
    "eval_rewards/rejected": -0.19834741950035095,
    "eval_runtime": 73.8566,
    "eval_samples": 704,
    "eval_samples_per_second": 9.532,
    "eval_steps_per_second": 0.298
}