File size: 3,170 Bytes
998df90 b39b124 998df90 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 | ---
license: mit
tags:
- molecular-property-prediction
- graph-neural-network
- chemistry
- pytorch
datasets:
- qm9
- spice
- pfas
metrics:
- mse
- mae
pipeline_tag: graph-ml
library_name: moml
---
# djmgnn-base
## Model Description
This is a base **DJMGNN** (Dense Jump Multi-Graph Neural Network) model for molecular property prediction. The model is designed to predict various molecular properties from graph representations of molecules.
### Architecture
- **Model Type**: Dense Jump Multi-Graph Neural Network (DJMGNN)
- **Framework**: PyTorch
- **Library**: MoML (Molecular Machine Learning)
- **Task**: Molecular Property Prediction
### Model Architecture Details
- **Hidden Dimensions**: 128
- **Number of Blocks**: 3
- **Layers per Block**: 6
- **Input Node Dimensions**: 11
- **Input Edge Dimensions**: 0
- **Node Output Dimensions**: 3
- **Graph Output Dimensions**: 19
- **Energy Output Dimensions**: 1
- **Jumping Knowledge Mode**: cat
- **Dropout Rate**: 0.2
- **Uses Supernode**: True
- **Uses RBF Features**: True
- **RBF K**: 32
## Training Details
### Datasets
The model was trained on the following datasets:
- **QM9**: Quantum mechanical properties of small molecules
- **SPICE**: Molecular dynamics data with forces and energies
- **PFAS**: Per- and polyfluoroalkyl substances dataset
### Training Configuration
```yaml
batch_size: 32
early_stopping: true
epochs: 100
learning_rate: 0.001
optimizer: Adam
patience: 10
validation_split: 0.2
```
## Usage
### Loading the Model
```python
import torch
from moml.models.mgnn.djmgnn import DJMGNN
# Load the model
model = DJMGNN(
in_node_dim=11,
in_edge_dim=0,
hidden_dim=128,
n_blocks=3,
layers_per_block=6,
node_output_dims=3,
graph_output_dims=19,
energy_output_dims=1,
jk_mode="cat",
dropout=0.2,
use_supernode=true,
use_rbf=true,
rbf_K=32
)
# Load the checkpoint
checkpoint = torch.load("path/to/pytorch_model.pt", map_location="cpu")
model.load_state_dict(checkpoint["model_state_dict"])
model.eval()
```
### Making Predictions
```python
# Assuming you have a molecular graph 'data' (torch_geometric.data.Data object)
with torch.no_grad():
output = model(
x=data.x,
edge_index=data.edge_index,
edge_attr=data.edge_attr,
batch=data.batch
)
# Extract predictions
node_predictions = output["node_pred"] # Node-level predictions
graph_predictions = output["graph_pred"] # Graph-level predictions
energy_predictions = output["energy_pred"] # Energy predictions
```
## Model Performance
This is the base DJMGNN model trained on QM9, SPICE, and PFAS datasets.
## Citation
If you use this model in your research, please cite:
```bibtex
@misc{djmgnn_model,
title={DJMGNN: Dense Jump Multi-Graph Neural Network for Molecular Property Prediction},
author={Your Name},
year={2024},
url={https://github.com/SAKETH11111/MoML-CA}
}
```
## License
This model is released under the MIT License.
## Contact
For questions or issues, please contact sakethbaddam10@gmail.com or open an issue in the [GitHub repository](https://github.com/SAKETH11111/MoML-CA).
|