Since onnxruntime updated, I was forced to update tensorrt and cuDNN to get rembg working again.
Unfortunately, it's incredibly slow (1:30) now and I am not able to get it faster again, using a RTX 3080 10GB.
During processing, the terminal is spammed by the following:
[W:onnxruntime:Default, tensorrt_execution_provider.h:63 onnxruntime::TensorrtLogger::log] [2023-03-23 15:28:50 WARNING] CUDA lazy loading is not enabled. Enabling it can significantly reduce device memory usage and speed up TensorRT initialization. See "Lazy Loading" section of CUDA documentation https://docs.nvidia.com/cuda/cuda-c-programming-guide/index.html#lazy-loading
Since onnxruntime updated, I was forced to update tensorrt and cuDNN to get rembg working again.
Unfortunately, it's incredibly slow (1:30) now and I am not able to get it faster again, using a RTX 3080 10GB.
During processing, the terminal is spammed by the following: [W:onnxruntime:Default, tensorrt_execution_provider.h:63 onnxruntime::TensorrtLogger::log] [2023-03-23 15:28:50 WARNING] CUDA lazy loading is not enabled. Enabling it can significantly reduce device memory usage and speed up TensorRT initialization. See "Lazy Loading" section of CUDA documentation https://docs.nvidia.com/cuda/cuda-c-programming-guide/index.html#lazy-loading
So, I've tried to set the env variable following https://github.com/AUTOMATIC1111/stable-diffusion-webui/discussions/8459 to my web-user.bat.
Unfortunately, neither set CUDA_MODULE_LOADING="LAZY" nor set CUDA_MODULE_LOADING=LAZY
Seem to set the variable.
I'd appreciate help or a information about which version combination of CUDA, cuDNN and tensorrt would get me back to the previous speed.
My current versions: CUDA 12.1.0 cuDNN 8.8.1.3 tensorrt 8.6.0.12
Thanks in advance!