Qwen2.5-7B-Instruct/ReasoningBench/MetaMath/train_results.json

{
    "epoch": 1.0,
    "total_flos": 2.3466648993521664e+16,
    "train_loss": 0.3158116644740105,
    "train_runtime": 807.301,
    "train_samples_per_second": 6.193,
    "train_steps_per_second": 6.193
}