CUDA installation
Tensorflow 1.4 support CUDA 8.0 and CUDNN 6.0.
The latest version is CUDA 9.1 and CUDNN 7.0.
To avoid compatibility issue, we use lower version of CUDA.
Steps:
1) Goto: https://developer.nvidia.com/cuda-toolkit-archive
2) Select CUDA Toolkit 8.0 GA2
3) Download the files, and follow the instructions for installation.
- `sudo dpkg -i cuda-repo-ubuntu1604-8-0-local-ga2_8.0.61-1_amd64.deb`
- `sudo apt-get update`
- `sudo apt-get install cuda
- environment setup:
export PATH=/usr/local/cuda-8.0/bin${PATH:+:${PATH}}
export LD_LIBRARY_PATH=/usr/local/cuda-8.0/lib64${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}
- testing CUDA:
Copy the samples in /usr/local/cuda-8.0/samples to ~/CUDA-8_samples
makefile
go to one of the folder: ~/NVIDIA_CUDA-8_Samples/5_Simulations/nbody
make
./nbody
5) Install cuDNN 6.0 for CUDA 8.0 (deep learning library)
https://developer.nvidia.com/rdp/cudnn-download (need to register and login)
PREREQUISITES CUDA 7.5 or higher version and a GPU of compute capability 3.0 or higher are required.
For Library:
Extract the cuDNN archive to a directory of your choice, referred to below as <installpath>. Then follow the platform-specific instructions as follows. cd <installpath>/lib export LD_LIBRARY_PATH=`pwd`:$LD_LIBRARY_PATH Add <installpath> to your build and link process by adding -I<installpath>/include to your compile line and -L<installpath>/lib -lcudnn to your link line.
For Deb:
SUPPORTED PLATFORMS Ubuntu 14.04, Ubuntu 16.04, POWER8 Then follow the platform-specific instructions as follows. 1. Install Runtime library sudo dpkg -i $(runtime library deb) 2. Install developer library sudo dpkg -i $(developer library deb) 3. Install code samples and user guide sudo dpkg -i $(document library deb)