all_results.json

{
    "epoch": 1.9986120749479528,
    "eval_logits/chosen": 1.8379676342010498,
    "eval_logits/rejected": 4.30747127532959,
    "eval_logps/chosen": -870.468017578125,
    "eval_logps/rejected": -1735.4202880859375,
    "eval_loss": 0.24658846855163574,
    "eval_rewards/accuracies": 0.896616518497467,
    "eval_rewards/chosen": -4.8588385581970215,
    "eval_rewards/margins": 8.39414119720459,
    "eval_rewards/rejected": -13.252979278564453,
    "eval_runtime": 384.8738,
    "eval_samples": 8491,
    "eval_samples_per_second": 22.062,
    "eval_steps_per_second": 0.346,
    "total_flos": 0.0,
    "train_loss": 0.25082930790053476,
    "train_runtime": 42170.4729,
    "train_samples": 184443,
    "train_samples_per_second": 8.747,
    "train_steps_per_second": 0.034
}