eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -2.7795233726501465,
    "eval_logits/rejected": -2.7412893772125244,
    "eval_logps/chosen": -280.4747619628906,
    "eval_logps/rejected": -256.65057373046875,
    "eval_loss": 0.6853376626968384,
    "eval_rewards/accuracies": 0.5879999995231628,
    "eval_rewards/chosen": 0.041186705231666565,
    "eval_rewards/margins": 0.02190316654741764,
    "eval_rewards/margins_max": 0.12900841236114502,
    "eval_rewards/margins_min": -0.07138193398714066,
    "eval_rewards/margins_std": 0.06693326681852341,
    "eval_rewards/rejected": 0.019283540546894073,
    "eval_runtime": 428.8609,
    "eval_samples": 2000,
    "eval_samples_per_second": 4.664,
    "eval_steps_per_second": 0.291
}