Llama-3.2-3B_hh_harmful/train_results.json

{
    "total_flos": 1.939658958385971e+16,
    "train_loss": 2.081411361694336,
    "train_runtime": 20.4328,
    "train_samples_per_second": 108.845,
    "train_steps_per_second": 1.713
}