[Question] Benchmark baseline for Qwen-Image (Original vs. TensorRT)

[Jiuchaoyao]

Hi there,

First of all, thank you for this great repository! I have successfully deployed and run the entire pipeline locally, and the functionality works perfectly.

I am writing to ask if you could share some baseline performance data so I can verify if my local inference results are within the expected range. Specifically, I am interested in comparing the Original Qwen-Image vs. the TensorRT-accelerated version regarding:

1。 Inference Speed
2. GPU Memory Usage
Knowing the hardware environment (e.g., A100, 4090) you tested on would also be very helpful.

Thanks again for your work!