lora is merged to model in inference, but the model weight is unchanged

Hi@kohya-ss, sorry to interrupt you, I am very confused that in `flux_minimal_inference.py` how lora weight is merged into the original unet. The `def generate_image` does not include lora weight, whereas the output image is indeed different from the original dev output. I am a beginner to flux model and lora, could you please give me some hint? many thanks !!!!

Start with flux_minimal_inference.py and trace how the LoRA weight is loaded or merged in relation to generate_image. Compare that path with the unchanged model weight described in the issue; done means explaining why the output differs and whether the reported behavior is expected.

Written by the indexing model from the issue text.