all_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -2.596123456954956,
    "eval_logits/rejected": -2.594705104827881,
    "eval_logps/chosen": -524.208984375,
    "eval_logps/rejected": -555.1446533203125,
    "eval_loss": 0.7283747792243958,
    "eval_rewards/accuracies": 0.5703125,
    "eval_rewards/chosen": -2.3185923099517822,
    "eval_rewards/margins": 0.11669808626174927,
    "eval_rewards/rejected": -2.435290575027466,
    "eval_runtime": 128.6637,
    "eval_samples": 1000,
    "eval_samples_per_second": 7.772,
    "eval_steps_per_second": 0.249,
    "train_loss": 0.5676067341596652,
    "train_runtime": 5367.57,
    "train_samples": 19996,
    "train_samples_per_second": 3.725,
    "train_steps_per_second": 0.058
}