all_results.json

{
    "epoch": 2.0,
    "eval_logits/chosen": -2.3005340099334717,
    "eval_logits/rejected": -2.3817031383514404,
    "eval_logps/chosen": -316.1898498535156,
    "eval_logps/rejected": -322.1933898925781,
    "eval_loss": 0.4346597194671631,
    "eval_rewards/accuracies": 0.7658227682113647,
    "eval_rewards/chosen": -0.9460535049438477,
    "eval_rewards/margins": 1.8284220695495605,
    "eval_rewards/rejected": -2.7744758129119873,
    "eval_runtime": 117.6177,
    "eval_samples": 2500,
    "eval_samples_per_second": 21.255,
    "eval_steps_per_second": 0.672,
    "train_loss": 0.5184940074794384,
    "train_runtime": 19743.6623,
    "train_samples": 73494,
    "train_samples_per_second": 7.445,
    "train_steps_per_second": 0.058
}