all_results.json

{
    "epoch": 5.0,
    "eval_logits/chosen": -1.985146403312683,
    "eval_logits/rejected": -1.8253288269042969,
    "eval_logps/chosen": -287.1579895019531,
    "eval_logps/rejected": -271.7846374511719,
    "eval_loss": 0.09761956334114075,
    "eval_rewards/accuracies": 0.3440000116825104,
    "eval_rewards/chosen": -0.20460382103919983,
    "eval_rewards/margins": -0.03615570068359375,
    "eval_rewards/rejected": -0.16844810545444489,
    "eval_runtime": 700.8137,
    "eval_samples": 2000,
    "eval_samples_per_second": 2.854,
    "eval_steps_per_second": 1.427,
    "train_loss": 0.3515237546205521,
    "train_runtime": 12848.6235,
    "train_samples": 61135,
    "train_samples_per_second": 0.389,
    "train_steps_per_second": 0.097
}