all_results.json

{
    "epoch": 3.997808219178082,
    "eval_logits/chosen": -2.860849618911743,
    "eval_logits/rejected": -2.3769683837890625,
    "eval_logps/chosen": -706.0082397460938,
    "eval_logps/rejected": -596.4307861328125,
    "eval_loss": 0.40534183382987976,
    "eval_rewards/accuracies": 0.7142857313156128,
    "eval_rewards/chosen": 2.9905052185058594,
    "eval_rewards/margins": 4.6127777099609375,
    "eval_rewards/rejected": -1.6222723722457886,
    "eval_runtime": 14.2958,
    "eval_samples_per_second": 7.695,
    "eval_steps_per_second": 0.979,
    "total_flos": 519662108934144.0,
    "train_loss": 0.17924007512469045,
    "train_runtime": 53882.8337,
    "train_samples_per_second": 3.251,
    "train_steps_per_second": 0.102
}