zephyr-7b-dpo-iter3/train_results.json

{
    "epoch": 2.0,
    "total_flos": 0.0,
    "train_loss": 0.4729907219554042,
    "train_runtime": 6750.8442,
    "train_samples": 40756,
    "train_samples_per_second": 12.074,
    "train_steps_per_second": 0.377
}