all_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -2.2439703941345215,
    "eval_logits/rejected": -2.2390873432159424,
    "eval_logps/chosen": -34.037574768066406,
    "eval_logps/rejected": -37.50224685668945,
    "eval_loss": 0.7015302777290344,
    "eval_rewards/accuracies": 0.4925249218940735,
    "eval_rewards/chosen": -0.002115294337272644,
    "eval_rewards/margins": -0.012174182571470737,
    "eval_rewards/rejected": 0.010058889165520668,
    "eval_runtime": 145.8307,
    "eval_samples": 343,
    "eval_samples_per_second": 2.352,
    "eval_steps_per_second": 0.295,
    "train_loss": 0.6898305428492558,
    "train_runtime": 3256.6227,
    "train_samples": 3079,
    "train_samples_per_second": 0.945,
    "train_steps_per_second": 0.118
}