Application to process jobs submitted from BIIGLE on a GPU.
All GPU modules should use Python 3 for computation, so run all scripts with python3. See the requirements.txt for the available Python 3 packages. In addition to those, TensorFlow is provided.
Head over to the distribution repository to configure and build your production setup.
To develop the BIIGLE GPU server on your local machine you may use Docker containers, too. This way you don't need to install any of the requirements like Python or special PHP extensions and keep your development environment clearly separated from your regular OS.
Set up the project in the biigle-gpus directory (using Composer):
composer create-project biigle/gpus:dev-master --repository='{"type":"vcs","url":"[email protected]:biigle/gpus.git"}' --keep-vcs --ignore-platform-reqs --prefer-source biigle-gpus
Note the --ignore-platform-reqs flag to keep Composer from complaining about missing requirements. These requirements will be met by the Docker containers.
Now you have to configure GITHUB_OAUTH_TOKEN in the .env file with a personal access token of GitHub (with authorization for the repo scope).
This allows you to build the Docker images and start the application with docker-compose up. The first time may take a while. The BIIGLE GPU server application is now running at http://localhost:8001. Stop the containers with docker-compose stop. Destroy them (and the development database) with docker-compose down.
To build and publish the Docker images, run ./build.sh. The script asks you if you want to publish the new images to the GitHub package registry. Before you can do this, you have to configure Docker to use the package registry and you obviously need package write permissions in this repository.