The goal of this project is to reproduce the paper "Inductive Representation Learning on Large Graphs".
-
baselines/
Code for all the baselines. -
experiments/
Code reproducing all the experiments.fig2a/
Code reproducing the figure 2a of the original paper.fig2b/
Code reproducing the figure 2b of the original paper.fig3/
Code reproducing the figure 3 of the original paper.table1/
Code reproducing the table 1 of the original paper.
-
graphsage/
Code related to graphsage models and extensions.datasets/
Code related to datasets and transformations.models/
Code related to models.layers/
Code related to reusable pytorch modules.samplers/
Code related to samplers.trainers/
Custom abstractions to monitor training.
-
examples
/ Direct application of models on datasets. -
docs/
Auto-Generated & manual code documentation. -
scripts/
Contains bash scripts, this scripts might just be launchers for python scripts defined in the main package. Useful for running long experiments for example. -
data/
Auto-generated, contains original or intermediate synthetic data. -
examples/
All the examples, python scripts or notebooks, illustrating the usage of the package. -
graphsage/
Python package containing the main code for this project. -
results/
Auto-generated, For results, e.g. tables (csv files), and plots (images)
# Create the conda environment
conda env create -f env.yml
# Add the environment to your jupyter kernels
python -m ipykernel install --user --name graphsage
# Activate the environment
conda activate graphsage
To generate documentation, run
scripts/makedoc.sh
The documentation entrypoint will be generated at docs/_build/html/index.html