qwen3-8b-base-sft-hh-helpful-8xh200/train_results.json

{
    "epoch": 1.0,
    "total_flos": 5.022030486516531e+16,
    "train_loss": 1.7195094338169805,
    "train_runtime": 501.5523,
    "train_samples": 43598,
    "train_samples_per_second": 34.284,
    "train_steps_per_second": 0.269
}