Qwen3-4B-Math-R1-CoT-SFT/all_results.json

{
    "effective_tokens_per_sec": 24290.389169769056,
    "epoch": 4.0,
    "total_flos": 4.3530238414695694e+19,
    "train_loss": 0.15123671168837785,
    "train_runtime": 82201.3053,
    "train_samples_per_second": 1.977,
    "train_steps_per_second": 0.031
}