https://github.com/ornl/ornl-hydragnn-graph-generative-models
Graph generative models using HydraGNN as neural network architecture
https://github.com/ornl/ornl-hydragnn-graph-generative-models
Last synced: 3 months ago
JSON representation
Graph generative models using HydraGNN as neural network architecture
- Host: GitHub
- URL: https://github.com/ornl/ornl-hydragnn-graph-generative-models
- Owner: ORNL
- License: bsd-3-clause
- Created: 2024-10-22T17:28:11.000Z (8 months ago)
- Default Branch: main
- Last Pushed: 2025-03-17T20:22:10.000Z (3 months ago)
- Last Synced: 2025-03-17T21:25:02.660Z (3 months ago)
- Language: Python
- Size: 290 KB
- Stars: 0
- Watchers: 7
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- Contributing: CONTRIBUTING.md
- License: LICENSE
- Citation: CITATION.cff
Awesome Lists containing this project
README
# Diffusion Models on Graphs with HydraGNN
This project builds on HydraGNN, leveraging its powerful GNN and ML utilities for training, testing, and model optimization.
## Features
* TBD
## Quick Start
Clone the repo:
```bash
git clone
cd
```
### Install Dependencies:
Make sure you have the HydraGNN environment set up:
```bash
pip install -r requirements.txt
```
### Run Training:
```bash
python
```
## How It Works
HydraGNN integration: We utilize the operational utilities from HydraGNN, such as model training, testing, and optimization, to simplify workflow.
Diffusion Process: Modeled on graph structures to simulate the propagation of information or features across the graph nodes. Perfect for dynamic systems!
Model Parallelization: Thanks to HydraGNN, training large models with multi-GPU support is integrated.
### ️Configuration
All model and training parameters can be easily set via our config.json file:
```json
model:
type: diffusion_gnn
layers: 5
hidden_dim: 128
train:
epochs: 100
batch_size: 32
learning_rate: 0.001
```
## Modules
`src/<>.py: `
### Performance
Our diffusion-enhanced GNNs show promising results in tasks such as:
### Contributing
We welcome contributions! If you're interested in extending the diffusion model or improving performance, feel free to submit a pull request or open an issue.