llama-3-8b-base-hh-harmless-sft-4xh100/train_results.json

{
    "epoch": 0.9927710843373494,
    "total_flos": 7.598970625641677e+16,
    "train_loss": 2.118080861360124,
    "train_runtime": 272.0741,
    "train_samples": 42537,
    "train_samples_per_second": 48.781,
    "train_steps_per_second": 0.379
}