all_results.json

{
    "epoch": 2.0,
    "eval_logits/chosen": 0.7325530052185059,
    "eval_logits/rejected": 0.7531598806381226,
    "eval_logps/chosen": -94.8434829711914,
    "eval_logps/rejected": -98.55415344238281,
    "eval_loss": 0.687246561050415,
    "eval_rewards/accuracies": 0.33399999141693115,
    "eval_rewards/chosen": -0.03116113506257534,
    "eval_rewards/margins": 0.012410260736942291,
    "eval_rewards/rejected": -0.04357139766216278,
    "eval_runtime": 272.5724,
    "eval_samples": 2000,
    "eval_samples_per_second": 7.337,
    "eval_steps_per_second": 0.459,
    "train_loss": 0.3994182998029441,
    "train_runtime": 8026.3391,
    "train_samples": 30567,
    "train_samples_per_second": 7.617,
    "train_steps_per_second": 0.119
}