math-think-s1-v2-qwen3-4b/train_results.json

{
    "epoch": 2.0,
    "total_flos": 1161363418251264.0,
    "train_loss": 0.6327840509308371,
    "train_runtime": 63341.305,
    "train_samples_per_second": 0.634,
    "train_steps_per_second": 0.01
}