Skip to content

Latest commit

 

History

History
82 lines (51 loc) · 3.97 KB

CONTRIBUTING.md

File metadata and controls

82 lines (51 loc) · 3.97 KB
Raw

Contributing to DIG

Thank you very much for your interest in contributing to Dive into Graphs (DIG)!

Table of Contents

  1. Overview
  2. Environment and Setup
  3. Issues and Bug Reporting
  4. Adding New Algorithms
  5. [Adding Datasets]
  6. [Adding Evaluation Metrics]
  7. [Pull Requests]

Overview

Before contributing, it is helpful to have an understanding of the general structure of DIG. Namely, DIG is not one single graph model; instead, it is a collection of algorithms across different topics. The objective of DIG is to enable researchers to benchmark their work and implement new ideas.

Structurally, DIG is divided into four topics: Graph Generation, Self-supervised Learning on Graphs, Explainability of Graph Neural Networks, and Deep Learning on 3D Graphs.

Each topic has its own sets of standards and evaluation metrics, but a uniform format is used. Specifically, every directory under the /dig sources root contains several implemented algorithms, a directory of datasets used by those methods, and a directory with utilities or metrics to augment and evaluate the model. By principle, DIG strives to use the uniform tools whenever possible. So, for example, it is preferable that a contribution to Graph Generation is evaluated against the metrics in that topic.

Environment and Setup

DIG uses Git as its version control system. These instructions will help with synchronization of your local environment and this repository.

To facilitate this, fork a local copy of DIG by clicking "Fork" in the top right of the screen at this URL.

Then, clone your fork:

$ git clone https://github.com/[YOUR_USERNAME]/DIG.git

Next, add the official DIG repository as a remote:

$ git remote add upstream https://github.com/divelab/DIG.git

It is recommended that a separate branch is used during development:

$ git branch [BRANCH_NAME]
$ git checkout [BRANCH_NAME]

Finally, once the contributions are ready to be pushed back to the original repository, execute the following:

git checkout master
git pull upstream master
git push origin master  # This updates your fork of the repository
git branch -D [BRANCH_NAME]  # This deletes the branch you made during development

Note about requirements:

Requirements are managed at the model level; however, there are several standards across DIG:

  • Python = 3.7
  • CUDA Toolkit = 10.1

Issues and Bug Reporting

We use the GitHub issues tracker to manage any issues, questions, and reports. Please use the label feature to indicate what topic your issue concerns.

Adding New Algorithms

We welcome contributions of new algorithms that are implemented in PyTorch. To contribute, follow these steps:

  1. Create a new directory under the relevant topic. For example, DIG/dig/ggraph/[YOUR_ALGORITHM]. [For sslgraph, add your method to a new python file under DIG/dig/sslgraph/contrastive/model/ and add functions to DIG/dig/sslgraph/contrastive/views_fn/ if necessary.]

  2. Add your algorithm to this directory. Take care to craft a README that is compatible with the DIG standards. An example of a strong README is available here. [For sslgraph, creating a jupyter notebook with your example is greatly recommended.]

  3. Follow the steps provided in the Environment & Setup section to push your contributions to your fork. Then, navigate to your fork on GitHub, and create a pull request. The pull request will be reviewed by a member familiar with the parent topic of the contributed algorithm.

  4. Organize data within the datasets directory of the parent topic, and add any additional metrics to the metrics or utils of the parent topic.

  5. Take note of any comments made to the pull request, and work with the reviewer to implement any changes.

  6. That's it. Thank you for your contributions!