all_results.json

{
    "epoch": 3.0,
    "eval_logits/chosen": -2.353081703186035,
    "eval_logits/rejected": -2.308103084564209,
    "eval_logps/chosen": -299.4560546875,
    "eval_logps/rejected": -340.154052734375,
    "eval_loss": 0.7496059536933899,
    "eval_rewards/accuracies": 0.78125,
    "eval_rewards/chosen": -4.522095203399658,
    "eval_rewards/margins": 3.7963125705718994,
    "eval_rewards/rejected": -8.318408012390137,
    "eval_runtime": 48.0152,
    "eval_samples": 1000,
    "eval_samples_per_second": 20.827,
    "eval_steps_per_second": 0.333,
    "train_loss": 0.2172969928600547,
    "train_runtime": 23865.9828,
    "train_samples": 61966,
    "train_samples_per_second": 7.789,
    "train_steps_per_second": 0.243
}