Generic_GNN

A generic GNN framework to enable rapid research and development with DGL.

Description

Meant for general purpose graph learning tasks, specifically those which require large graphs and assays.

Code Style

Documentation follows Google's Python style guidelines, with TODOs and supporting comments scattered throughout on an as needed basis. Entry point for the entire codebase is in main.py. 3 pipelines are available to run, as described in main.py documentation. Configuration files are stored as json files in config directory.

Directory Tree

Stored in directory.txt (use Get-ChildItem | tree /F > foo.txt in PowerShell to create your own)

Getting Started

Conda Env Setup

First you'll want to create a new conda (or pip) env with Python 3.7

conda create -n env_name python=3.7 anaconda
source activate env_name

Before cloning into this repository:

git clone https://github.com/flawnson/Generic_GNN.git
OR
pip install git+https://github.com/flawnson/Generic_GNN.git

Then you can run setup.py

python setup.py

and install depedencies in the requirements.txt

pip install -r requirements.txt

Then you'll need to create an empty directory for model outputs (including saved models)

cd Generic_GNN && mkdir outputs

Finally you can run a demo version of the pipeline (default configs in configs directory)

python -c path/to/config/files/file.json -s path/to/schema/files/file.json

You can see the logged results using TensorBoard

tensorboard --logdir=logs/GAT_tuning/tune_model

None of the above will work without the correct data files, all of which are not publically available as of currently.

Docker Container Setup

Change directories into the one with Dockerfile and run (add tags as necessary):

docker build .

Copy the container ID and run the container by executing:

docker run -p port_number:8000 container_id

Be sure to stop (or kill if necessary) the container when not in use.

Acknowledgements

To my mentors and my friends whom have taught and inspired me all these years. Thanks to PyTorch Geometric for providing me with the background to use DGL, and DGL for providing me with a kick ass library.