Working with GPUs

Working with GPUs

Ubuntu 18.04

1. Download Anaconda

Go to https://www.anaconda.com/download/ and download Anaconda for your operating system.

2. Install Anaconda

Under Linux, that is:

chmod +x ~/Downloads/Anaconda3-5.2.0-Linux-x86_64.sh
~/Downloads/Anaconda3-5.2.0-Linux-x86_64.sh

During installation, select to have anaconda added to your path. That will add something like the following to your ~/.bashrc file:

export PATH="/home/dblank/anaconda3/bin:$PATH"

You'll either want to log out of your terminal and log back in, or source ~/.bashrc to get your updated path.

3. Update Anaconda:

conda update conda
conda update anaconda
conda update python
conda update --all

4. Create an environment

conda create --name tf-gpu

5. Activate the environment:

source activate tf-gpu

6. Install TensorFlow with support for GPUs:

conda install tensorflow-gpu

7. Install the ConX dependencies that are provided via conda:

conda install keras ipywidgets Pillow h5py tqdm requests pydot matplotlib

8. Install other, non-conda-based packages:

pip install svgwrite sklearn cairosvg conx

9. Optional, but useful: install other, system-wide packages:

sudo apt install cython

To deactivate this environment:

source deactivate 

Then, next time, all you need to do is:

source activate tf-gpu