all_results.json

{
    "epoch": 1.971563981042654,
    "eval_logits/chosen": 99.9845199584961,
    "eval_logits/rejected": 94.18653106689453,
    "eval_logps/chosen": -403.2098693847656,
    "eval_logps/rejected": -425.10888671875,
    "eval_loss": 0.9077228307723999,
    "eval_rewards/accuracies": 0.7291666865348816,
    "eval_rewards/chosen": -1.9770817756652832,
    "eval_rewards/margins": 1.1748138666152954,
    "eval_rewards/rejected": -3.151895523071289,
    "eval_runtime": 123.8059,
    "eval_samples": 750,
    "eval_samples_per_second": 6.058,
    "eval_steps_per_second": 0.194,
    "total_flos": 0.0,
    "train_loss": 0.9854127982488046,
    "train_runtime": 2188.3418,
    "train_samples": 6750,
    "train_samples_per_second": 6.169,
    "train_steps_per_second": 0.048
}