eval_results.json

{
    "epoch": 0.9990726429675425,
    "eval_log_odds_chosen": 9.089225769042969,
    "eval_log_odds_ratio": -0.14960747957229614,
    "eval_logits/chosen": -0.23287992179393768,
    "eval_logits/rejected": -0.2937111556529999,
    "eval_logps/chosen": -0.10372789204120636,
    "eval_logps/rejected": -7.177553653717041,
    "eval_loss": 0.2745336890220642,
    "eval_nll_loss": 0.10372789204120636,
    "eval_rewards/accuracies": 0.9399999976158142,
    "eval_rewards/chosen": -0.10372789204120636,
    "eval_rewards/margins": 7.073825836181641,
    "eval_rewards/rejected": -7.177553653717041,
    "eval_runtime": 7.3204,
    "eval_samples": 100,
    "eval_samples_per_second": 13.66,
    "eval_steps_per_second": 6.83
}