Qwen2.5-3B-Instruct-RG-Math/train_results.json

{
    "total_flos": 0.0,
    "train_loss": 0.15968478212249465,
    "train_runtime": 15013.7279,
    "train_samples_per_second": 0.639,
    "train_steps_per_second": 0.04
}