At present,the difference version of the CUDA could be installed at same time. CuDNN,NCCL,TensorRT was installed to such /usr/lib/x86_64-linux-gnu linux runtime library directory through debian package Installation Method. Could to be modified so it be installed into corresponding CUDA directory? Just a suggestion.
Related topics
Topic | Replies | Views | Activity | |
---|---|---|---|---|
Installing specific version of CUDNN on Ubuntu 20.04 with Package Manager | 5 | 7859 | February 28, 2023 | |
multiple cuDNN installation possible? | 2 | 4488 | November 20, 2017 | |
Cudnn issues with nvidia-cuda-toolkit on ubuntu 20 | 6 | 3175 | August 1, 2020 | |
cuda directory | 6 | 6324 | February 6, 2019 | |
CUDNN installation path by Jetpack | 1 | 1096 | September 21, 2020 | |
Download cuDNN issues | 5 | 1557 | October 26, 2020 | |
Why Installing cuDNN From Package Manager Doesn't Required Nvidia Developer Account | 2 | 4189 | October 12, 2021 | |
Issues while verifying the cuDNN installation | 0 | 816 | September 2, 2019 | |
Windows CUDNN installation instructions are incompatible with Tensorflow (at least version 2.10) | 1 | 751 | January 25, 2023 | |
Issue: CUDA and cuDNN installation on Ubuntu 22.04 | 1 | 4185 | November 16, 2023 |