search

lmb-freiburg / Unet-Segmentation

The U-Net Segmentation plugin for Fiji (ImageJ)
https://lmb.informatik.uni-freiburg.de/resources/opensource/unet
GNU General Public License v3.0
87 stars 25 forks source link

CUDA 11 support #80

Open BolekZapiec opened 3 years ago

BolekZapiec commented 3 years ago

Hi, I recently upgraded my GPU from the RTX 2080ti to a RTX 3090 and now realize that the UNET plugin only specifies up to CUDA 10.x which is no longer supported by the 3090. I was wondering if you were planning on releasing a CUDA 11 compatible version of the backend?

Many thanks!

lienkamp commented 3 years ago

Hi Thorsten,

We love the plugin and have been training like crazy. It's really awesome, and the plugin makes is so much more accessible.

Now, we also struggle to get the plugin running on an RTX 3090, using CUDA 11.2 and cuDNN 8.1.1.

When running the "cmake" command we end up with an error message indicating a problem with the detection of the newer cudNN version: -- CUDA detected: 11.2 -- Found cuDNN: ver. ??? found (include: /usr/local/cuda/include, library: /usr/local/cuda/lib64/libcudnn.so) CMake Error at cmake/Cuda.cmake:227 (message): cuDNN version >3 is required. Call Stack (most recent call first): cmake/Cuda.cmake:255 (detect_cuDNN) cmake/Dependencies.cmake:85 (include) CMakeLists.txt:49 (include) Any help/hints are highly appreciated!

Borease commented 3 years ago

Hi Thorsten,

We got similar issue as above mentioned. Our workstation is built on Ubuntu 20.04 which only support CUDA 11+, we love the plugin and eagerly waiting for any future release to support CUDA 11. Thank you!

ThorstenFalk commented 2 years ago

I am not officially maintaining the package any more since leaving university. I'll try to find some time during Christmas holidays to look into it. Brief answer: caffe hasn't been actively developed for years, thus a simple update is not possible. I will check whether I can manually build with slight adaptations.