all_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": 0.39978858828544617,
    "eval_logits/rejected": 1.363898754119873,
    "eval_logps/chosen": -552.1046142578125,
    "eval_logps/rejected": -778.7791137695312,
    "eval_loss": 0.4203870892524719,
    "eval_rewards/accuracies": 0.8019999861717224,
    "eval_rewards/chosen": -1.9644380807876587,
    "eval_rewards/margins": 1.633329153060913,
    "eval_rewards/rejected": -3.5977675914764404,
    "eval_runtime": 1376.4504,
    "eval_samples": 2000,
    "eval_samples_per_second": 1.453,
    "eval_steps_per_second": 0.363,
    "train_loss": 0.43769783746998087,
    "train_runtime": 91189.3347,
    "train_samples": 61135,
    "train_samples_per_second": 0.67,
    "train_steps_per_second": 0.042
}