all_results.json

{
    "epoch": 3.0,
    "eval_logits/chosen": -2.231706142425537,
    "eval_logits/rejected": -2.1925008296966553,
    "eval_logps/chosen": -259.9488830566406,
    "eval_logps/rejected": -234.73414611816406,
    "eval_loss": 0.5682210922241211,
    "eval_rewards/accuracies": 0.7958333492279053,
    "eval_rewards/chosen": 1.657994031906128,
    "eval_rewards/margins": 2.357875347137451,
    "eval_rewards/rejected": -0.6998813152313232,
    "eval_runtime": 126.4545,
    "eval_samples_per_second": 15.025,
    "eval_steps_per_second": 0.237,
    "total_flos": 2369906314051584.0,
    "train_loss": 0.3513777989208402,
    "train_runtime": 7410.5253,
    "train_samples_per_second": 6.921,
    "train_steps_per_second": 0.027
}