Spaces:
Running
Running
File size: 3,137 Bytes
6747ba1 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 |
Metadata-Version: 2.1
Name: mhg-gnn
Version: 0.0
Summary: Package for mhg-gnn
Author: team
License: TBD
Classifier: Programming Language :: Python :: 3
Classifier: Programming Language :: Python :: 3.9
Description-Content-Type: text/markdown
Requires-Dist: networkx>=2.8
Requires-Dist: numpy<2.0.0,>=1.23.5
Requires-Dist: pandas>=1.5.3
Requires-Dist: rdkit-pypi<2023.9.6,>=2022.9.4
Requires-Dist: torch>=2.0.0
Requires-Dist: torchinfo>=1.8.0
Requires-Dist: torch-geometric>=2.3.1
# mhg-gnn
This repository provides PyTorch source code assosiated with our publication, "MHG-GNN: Combination of Molecular Hypergraph Grammar with Graph Neural Network"
**Paper:** [Arxiv Link](https://arxiv.org/pdf/2309.16374)
For more information contact: [email protected]
## Introduction
We present MHG-GNN, an autoencoder architecture
that has an encoder based on GNN and a decoder based on a sequential model with MHG.
Since the encoder is a GNN variant, MHG-GNN can accept any molecule as input, and
demonstrate high predictive performance on molecular graph data.
In addition, the decoder inherits the theoretical guarantee of MHG on always generating a structurally valid molecule as output.
## Table of Contents
1. [Getting Started](#getting-started)
1. [Pretrained Models and Training Logs](#pretrained-models-and-training-logs)
2. [Replicating Conda Environment](#replicating-conda-environment)
2. [Feature Extraction](#feature-extraction)
## Getting Started
**This code and environment have been tested on Intel E5-2667 CPUs at 3.30GHz and NVIDIA A100 Tensor Core GPUs.**
### Pretrained Models and Training Logs
We provide checkpoints of the MHG-GNN model pre-trained on a dataset of ~1.34M molecules curated from PubChem. (later) For model weights: [HuggingFace Link]()
Add the MHG-GNN `pre-trained weights.pt` to the `models/` directory according to your needs.
### Replacicating Conda Environment
Follow these steps to replicate our Conda environment and install the necessary libraries:
```
conda create --name mhg-gnn-env python=3.8.18
conda activate mhg-gnn-env
```
#### Install Packages with Conda
```
conda install -c conda-forge networkx=2.8
conda install numpy=1.23.5
# conda install -c conda-forge rdkit=2022.9.4
conda install pytorch=2.0.0 torchvision torchaudio -c pytorch
conda install -c conda-forge torchinfo=1.8.0
conda install pyg -c pyg
```
#### Install Packages with pip
```
pip install rdkit torch-nl==0.3 torch-scatter torch-sparse
```
## Feature Extraction
The example notebook [mhg-gnn_encoder_decoder_example.ipynb](notebooks/mhg-gnn_encoder_decoder_example.ipynb) contains code to load checkpoint files and use the pre-trained model for encoder and decoder tasks.
To load mhg-gnn, you can simply use:
```python
import torch
import load
model = load.load()
```
To encode SMILES into embeddings, you can use:
```python
with torch.no_grad():
repr = model.encode(["CCO", "O=C=O", "OC(=O)c1ccccc1C(=O)O"])
```
For decoder, you can use the function, so you can return from embeddings to SMILES strings:
```python
orig = model.decode(repr)
```
|