eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -2.8198559284210205,
    "eval_logits/rejected": -2.817356824874878,
    "eval_logps/chosen": -31.425559997558594,
    "eval_logps/rejected": -35.012386322021484,
    "eval_loss": 0.6725602746009827,
    "eval_rewards/accuracies": 0.5627076625823975,
    "eval_rewards/chosen": -0.08586728572845459,
    "eval_rewards/margins": 0.1026720479130745,
    "eval_rewards/rejected": -0.1885393261909485,
    "eval_runtime": 112.8393,
    "eval_samples": 343,
    "eval_samples_per_second": 3.04,
    "eval_steps_per_second": 0.381
}