Practical overview
When it fits
Use the repository when you want a code-based starting point for graph classification or regression experiments and can work with its documented Python, PyTorch, CUDA and DGL setup.
Who it is for
- Researchers exploring graph representation learning.
- Practitioners comparing graph neural network variants on graph-level prediction tasks.
- Users preparing experiments on the documented OGB molecular datasets.
Documented uses
- Train gated-GCN, GCN or MLP models on ogbg-molhiv or ogbg-molpcba.
- Explore a combination of Transformer blocks, gated graph convolution and Laplacian-eigenvector positional encodings.
- Prepare and collate results for OGB leaderboard submissions.
Requirements & limits
The public GitHub repository is identified as MIT-licensed.
- The README documents OGB molecular training and submission workflows, not a hosted prediction interface.
- The setup instructions specify older Python, PyTorch and CUDA versions; the README separately records the environment used for testing.
- The project describes itself as a personal research project focused on graph representation learning.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Set up the documented training environment
Prepare the Python and graph-learning dependencies listed for the repository’s OGB molecular training workflow.
Before you start
- Conda
- Python 3.7 environment
- PyTorch 1.6 with CUDA 10.x
- DGL package for CUDA 10.x
- tqdm, scikit-learn, pandas, urllib3, tensorboard and OGB
Steps
Create and activate a Conda environment named new-env with Python 3.7.
Install PyTorch 1.6 with cudatoolkit 10.x from the PyTorch Conda channel.
Install the DGL CUDA 10.x package from the dglteam Conda channel.
Install the listed Conda dependencies and upgrade OGB with pip.
Expected outcome
The dependencies named in the README are installed in the new environment.
Guide limitations
- The README’s test note specifies Ubuntu 16.04, Python 3.6, PyTorch 1.6 and CUDA 10.1; it does not establish compatibility with other setups.
- This recipe covers environment setup only, not training execution.
Train a model on an OGB molecular dataset
Use the repository’s documented command pattern to train a selected GNN on an OGB molecular dataset.
Before you start
- The repository and its documented dependencies
- An OGB dataset choice of ogbg-molhiv or ogbg-molpcba
- A model choice of gated-gcn, gcn or mlp
Steps
Choose either ogbg-molhiv or ogbg-molpcba.
Choose gated-gcn, gcn or mlp.
Run main_mol.py with the selected values for --dataset and --gnn.
Expected outcome
The documented training script is invoked for the selected dataset and model.
Guide limitations
- The README does not specify expected metrics or training duration.
- The repository’s stated test environment is Ubuntu 16.04, Python 3.6, PyTorch 1.6 and CUDA 10.1.
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 18 October 2026
- Original website title
- chaitjo/gated-graph-transformers · Official homepage title · retained in its source language