all_results.json

{
    "epoch": 2.9906542056074765,
    "eval_logits/chosen": -2.534839391708374,
    "eval_logits/rejected": -2.5236949920654297,
    "eval_logps/chosen": -242.10446166992188,
    "eval_logps/rejected": -217.54747009277344,
    "eval_loss": 0.5131691098213196,
    "eval_rewards/accuracies": 0.8090277910232544,
    "eval_rewards/chosen": 1.2483490705490112,
    "eval_rewards/margins": 2.6867623329162598,
    "eval_rewards/rejected": -1.438413143157959,
    "eval_runtime": 152.0829,
    "eval_samples_per_second": 14.992,
    "eval_steps_per_second": 0.237,
    "total_flos": 2829829665718272.0,
    "train_loss": 0.3309280425310135,
    "train_runtime": 8801.0779,
    "train_samples_per_second": 6.993,
    "train_steps_per_second": 0.027
}