all_results.json

{
    "epoch": 0.9982631930527722,
    "eval_logits/chosen": -1.3516209125518799,
    "eval_logits/rejected": -1.361035943031311,
    "eval_logps/chosen": -1.7249912023544312,
    "eval_logps/rejected": -2.307142972946167,
    "eval_loss": 1.355675220489502,
    "eval_rewards/accuracies": 0.8455284833908081,
    "eval_rewards/chosen": -17.24991226196289,
    "eval_rewards/margins": 5.821517467498779,
    "eval_rewards/rejected": -23.071430206298828,
    "eval_runtime": 87.7742,
    "eval_samples": 1961,
    "eval_samples_per_second": 22.341,
    "eval_steps_per_second": 1.401,
    "total_flos": 0.0,
    "train_loss": 1.9719815800480975,
    "train_runtime": 11459.6716,
    "train_samples": 59876,
    "train_samples_per_second": 5.225,
    "train_steps_per_second": 0.041
}