all_results.json

{
    "epoch": 0.9934640522875817,
    "eval_logits/chosen": -0.3643321692943573,
    "eval_logits/rejected": -0.4401712417602539,
    "eval_logps/chosen": -394.849853515625,
    "eval_logps/rejected": -451.100830078125,
    "eval_loss": 0.6792144775390625,
    "eval_rewards/accuracies": 0.75,
    "eval_rewards/chosen": -0.15057459473609924,
    "eval_rewards/margins": 0.030382413417100906,
    "eval_rewards/rejected": -0.18095700442790985,
    "eval_runtime": 28.9938,
    "eval_samples": 20,
    "eval_samples_per_second": 0.69,
    "eval_steps_per_second": 0.172,
    "total_flos": 0.0,
    "train_loss": 0.6808264882940995,
    "train_runtime": 1624.4693,
    "train_samples": 612,
    "train_samples_per_second": 0.377,
    "train_steps_per_second": 0.047
}