eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -2.2261412143707275,
    "eval_logits/rejected": -2.2213146686553955,
    "eval_logps/chosen": -34.100101470947266,
    "eval_logps/rejected": -37.61717224121094,
    "eval_loss": 0.6907125115394592,
    "eval_rewards/accuracies": 0.5132890343666077,
    "eval_rewards/chosen": -0.013110256753861904,
    "eval_rewards/margins": 0.007000547368079424,
    "eval_rewards/rejected": -0.020110804587602615,
    "eval_runtime": 145.4751,
    "eval_samples": 343,
    "eval_samples_per_second": 2.358,
    "eval_steps_per_second": 0.296
}