all_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -3.2432265281677246,
    "eval_logits/rejected": -3.091603994369507,
    "eval_logps/chosen": -750.3968505859375,
    "eval_logps/rejected": -1521.8148193359375,
    "eval_loss": 0.2945583164691925,
    "eval_rewards/accuracies": 0.8792613744735718,
    "eval_rewards/chosen": -2.2046568393707275,
    "eval_rewards/margins": 3.7011642456054688,
    "eval_rewards/rejected": -5.905820846557617,
    "eval_runtime": 293.4477,
    "eval_samples": 2804,
    "eval_samples_per_second": 9.555,
    "eval_steps_per_second": 0.3,
    "train_loss": 0.04516718044779659,
    "train_runtime": 721.6494,
    "train_samples": 9808,
    "train_samples_per_second": 13.591,
    "train_steps_per_second": 0.212
}