qwen4B_models/qwen_4B_d1_iter2_model/train_results.json

{
    "epoch": 2.0,
    "total_flos": 21111522983936.0,
    "train_loss": 0.8778598624117234,
    "train_runtime": 458.5184,
    "train_samples_per_second": 4.746,
    "train_steps_per_second": 0.297
}