Install nvidia driver:
add-apt-repository ppa:graphics-drivers/ppa
apt-get update
apt-get install nvidia-<version number>
Verify installation by executing nvidia-settings -q NvidiaDriverVersion
.
Load CUDA from https://developer.nvidia.com/cuda-downloads
.
Run CUDA installer. Select default values for questions except for nvidia driver (select no).
sudo sh cuda_8.0.61_375.26_linux.run
A warning should occur, ignore it. Fix it by adding following lines into bash.rc (or other bash config):
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}}
Take care, the cuda version is assumed to be 8.0. Otherwise change the exports accordingly.
Optional: Install pip and python-dev packages: sudo apt-get install python-pip python-dev
.
Optional: Upgrade pip: pip install --upgrade pip
.
Install the TensorFlow gpu version: sudo pip install tensorflow-gpu
.
Download cuDNN from https://developer.nvidia.com/cudnn
.
Install cuDNN by copying it's contents into the CUDA folder:
sudo tar -xvf cudnn-8.0-linux-x64-v5.1.tgz -C /usr/local
In case of correct installation, importing tensorflow
in Python should result in following output:
I tensorflow/stream_executor/dso_loader.cc:135] successfully opened CUDA library libcublas.so.8.0 locally
I tensorflow/stream_executor/dso_loader.cc:135] successfully opened CUDA library libcudnn.so.5 locally
I tensorflow/stream_executor/dso_loader.cc:135] successfully opened CUDA library libcufft.so.8.0 locally
I tensorflow/stream_executor/dso_loader.cc:135] successfully opened CUDA library libcuda.so.1 locally
I tensorflow/stream_executor/dso_loader.cc:135] successfully opened CUDA library libcurand.so.8.0 locally