Llama3.2-3B_Paper_Impact_dataset_SFT_1ep/train_results.json

{
    "epoch": 1.0,
    "total_flos": 1.8521361846213018e+17,
    "train_loss": 0.11334303547354306,
    "train_runtime": 974.5943,
    "train_samples_per_second": 4.38,
    "train_steps_per_second": 0.035
}