Llama-3.2-3B-Instruct/ReasoningBench/MetaMath/train_results.json

{
    "epoch": 1.0,
    "total_flos": 8414511274868736.0,
    "train_loss": 0.19171363378167153,
    "train_runtime": 843.4917,
    "train_samples_per_second": 5.928,
    "train_steps_per_second": 5.928
}