all_results.json

{
    "epoch": 3.0,
    "eval_logits/chosen": -1.726296067237854,
    "eval_logits/rejected": -1.7025630474090576,
    "eval_logps/chosen": -203.75299072265625,
    "eval_logps/rejected": -244.15496826171875,
    "eval_loss": 0.6307257413864136,
    "eval_rewards/accuracies": 0.6433550119400024,
    "eval_rewards/chosen": -1.4504109621047974,
    "eval_rewards/margins": 0.3593376576900482,
    "eval_rewards/rejected": -1.809748649597168,
    "eval_runtime": 384.3784,
    "eval_samples": 4304,
    "eval_samples_per_second": 11.197,
    "eval_steps_per_second": 1.4,
    "total_flos": 0.0,
    "train_loss": 0.5724969553024556,
    "train_runtime": 86264.9537,
    "train_samples": 92858,
    "train_samples_per_second": 3.229,
    "train_steps_per_second": 0.202
}