qwen2.5-3b-sft-10/train_results.json

{
    "epoch": 10.0,
    "total_flos": 737402935050240.0,
    "train_loss": 0.556043832511017,
    "train_runtime": 12515.2297,
    "train_samples": 164653,
    "train_samples_per_second": 19.797,
    "train_steps_per_second": 0.31
}