all_results.json

{
    "epoch": 3.0,
    "eval_logits/chosen": -1.6946285963058472,
    "eval_logits/rejected": -1.6252007484436035,
    "eval_logps/chosen": -1460.7322998046875,
    "eval_logps/rejected": -1589.5469970703125,
    "eval_loss": 1.2165074348449707,
    "eval_rewards/accuracies": 0.6570000052452087,
    "eval_rewards/chosen": -11.761388778686523,
    "eval_rewards/margins": 1.5482909679412842,
    "eval_rewards/margins_max": 7.547528266906738,
    "eval_rewards/margins_min": -4.689306259155273,
    "eval_rewards/margins_std": 4.095370769500732,
    "eval_rewards/rejected": -13.309679985046387,
    "eval_runtime": 428.455,
    "eval_samples": 2000,
    "eval_samples_per_second": 4.668,
    "eval_steps_per_second": 0.292,
    "train_loss": 0.14573693349257882,
    "train_runtime": 13238.8899,
    "train_samples": 5678,
    "train_samples_per_second": 1.287,
    "train_steps_per_second": 0.08
}