llama3.2-1b-gsm8k-full/all_results.json

{
    "epoch": 3.0,
    "total_flos": 246796222464.0,
    "train_loss": 0.3023774076490572,
    "train_runtime": 943.8917,
    "train_samples_per_second": 3.458,
    "train_steps_per_second": 1.729
}