all_results.json

{
    "epoch": 0.9988571428571429,
    "eval_logits/chosen": 0.5811245441436768,
    "eval_logits/rejected": 1.8620274066925049,
    "eval_logps/chosen": -361.8296203613281,
    "eval_logps/rejected": -406.50030517578125,
    "eval_loss": 0.5201585292816162,
    "eval_rewards/accuracies": 0.7715517282485962,
    "eval_rewards/chosen": -0.8622565269470215,
    "eval_rewards/margins": 1.0119850635528564,
    "eval_rewards/rejected": -1.8742417097091675,
    "eval_runtime": 91.4795,
    "eval_samples": 1831,
    "eval_samples_per_second": 20.015,
    "eval_steps_per_second": 0.317,
    "total_flos": 0.0,
    "train_loss": 0.5594796479703087,
    "train_runtime": 11252.5956,
    "train_samples": 55999,
    "train_samples_per_second": 4.977,
    "train_steps_per_second": 0.039
}