eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -1.404586672782898,
    "eval_logits/rejected": -1.3435096740722656,
    "eval_logps/chosen": -788.7391357421875,
    "eval_logps/rejected": -854.9130859375,
    "eval_loss": 0.9015858173370361,
    "eval_rewards/accuracies": 0.656000018119812,
    "eval_rewards/chosen": -5.041457653045654,
    "eval_rewards/margins": 0.9218845367431641,
    "eval_rewards/margins_max": 4.525033473968506,
    "eval_rewards/margins_min": -2.841050386428833,
    "eval_rewards/margins_std": 2.5078797340393066,
    "eval_rewards/rejected": -5.96334171295166,
    "eval_runtime": 428.0105,
    "eval_samples": 2000,
    "eval_samples_per_second": 4.673,
    "eval_steps_per_second": 0.292
}