all_results.json

{
    "epoch": 3.0,
    "eval_logits/chosen": -2.0068085193634033,
    "eval_logits/rejected": -1.9822146892547607,
    "eval_logps/chosen": -345.02728271484375,
    "eval_logps/rejected": -404.28973388671875,
    "eval_loss": 0.7099329233169556,
    "eval_rewards/accuracies": 0.6319702863693237,
    "eval_rewards/chosen": -2.8601176738739014,
    "eval_rewards/margins": 0.5552833080291748,
    "eval_rewards/rejected": -3.4154014587402344,
    "eval_runtime": 360.569,
    "eval_samples": 4304,
    "eval_samples_per_second": 11.937,
    "eval_steps_per_second": 1.492,
    "total_flos": 0.0,
    "train_loss": 0.49131362667692796,
    "train_runtime": 84120.8122,
    "train_samples": 92858,
    "train_samples_per_second": 3.312,
    "train_steps_per_second": 0.207
}