CUDA 8.0 Setup
TensorFlow 1.15 works with CUDA 10.0 and cuDNN 7.4
TensorFlow 1.4 works with CUDA 8.0 and cuDNN 6.
If you have tensorflow 2.xx , Cuda > 11.0
Windows 10 OS
CUDA Toolkits
cuda 8.0 downloading https://developer.nvidia.com/cuda-80-ga2-download-archive
Set environment
- Go to Start and Search “environment variables”
- click the Environment Variables button
- Click on the Path system variable and select edit
- add the following paths
- `C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v8.0\bin`
- `C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v8.0\libnvvp`
- `C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v8.0\extras\CUPTI\libx64`
cudnn 6.0 downloading https://developer.nvidia.com/rdp/cudnn-archive
-
Looking for Download cuDNN v6.0 (April 27, 2017), for CUDA 8.0
-
Extract the files: Once you’ve downloaded the cuDNN zip file, extract it.
-
Copy the files:
- Copy the contents of the extracted
cuda
folder to your CUDA installation directory. For example, if you have installed CUDA in the default directory (C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v8.0
), copy the following:- Copy
bin/cudnn*.dll
toC:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v8.0\bin
. - Copy
include/cudnn*.h
toC:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v8.0\include
. - Copy
lib/x64/cudnn*.lib
toC:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v8.0\lib\x64
- Copy
- Copy the contents of the extracted
Ubuntu 16.04
Please note: Ubuntu 22.04 is not comptabile with CUDA 8.0
Reference https://gist.github.com/Mahedi-61/7b1f737e328fe19f4063f08a009c82c4