all_results.json

{
    "epoch": 4.0,
    "eval_logits/chosen": -2.1724095344543457,
    "eval_logits/rejected": -2.2160496711730957,
    "eval_logps/chosen": -102.96302795410156,
    "eval_logps/rejected": -101.06044006347656,
    "eval_loss": 0.6873850226402283,
    "eval_rewards/accuracies": 1.0,
    "eval_rewards/chosen": 0.08033924549818039,
    "eval_rewards/margins": 0.05054035410284996,
    "eval_rewards/rejected": 0.02979888767004013,
    "eval_runtime": 6.0716,
    "eval_samples": 30,
    "eval_samples_per_second": 4.941,
    "eval_steps_per_second": 0.165,
    "train_loss": 0.6922631859779358,
    "train_runtime": 883.8621,
    "train_samples": 626,
    "train_samples_per_second": 3.541,
    "train_steps_per_second": 0.006
}