Model deployment regarding flashocc

[layemiao]

Thank you for your open-source work. The link you provided previously for the TensorRT (C++) example is now invalid. I tried to run the Demo using the following link, and repeated the experiments in multiple environments. https://github.com/hituwangli/TRT_Flashocc The code, data, and ONNX model remained consistent. However, there is a significant accuracy difference between PyTorch and TensorRT, and they are almost unmatched. Could you please reactivate that link? https://github.com/drilistbox/TRT_FlashOcc Regarding the model deployment environment, such as CUDA, cuDNN, and TensorRT versions, do you have any recommended configurations? I am looking forward to your reply. Thank you again.

[layemiao]

@drilistbox @Yzichen