Meta-Llama-3-8B-Instruct/ReasoningBench/MetaMath/train_results.json

{
    "epoch": 1.0,
    "total_flos": 2.239561404560179e+16,
    "train_loss": 0.3289372156143188,
    "train_runtime": 837.4572,
    "train_samples_per_second": 5.97,
    "train_steps_per_second": 5.97
}