tencent-alchemy / Alchemy

Alchemy

Here is the repo for Tencent Alchemy Tools. We currently provide a PyG dataloader for Alchemy contest, as well as a PyG mpnn model.

HOWTO

How to use deep graph lib (dgl) dataloader for Alchemy

How to download Alchemy dataset

Simply run python3 train.py, the script will download and preprocess the dataset (dev set) automatically. You can download other dataset (valid set & test set) in a similar way or just do it manually.

How to run dgl based models

SchNet: expected MAE 0.065

python train.py --model sch --epochs 250

MGCN: expected MAE 0.050

python train.py --model mgcn --epochs 250

With Tesla V100, SchNet takes 80s/epoch and MGCN takes 110s/epoch.

Dependencies

• PyTorch 1.0+
• dgl 0.3+
• RDKit

Reference

• K.T. Schütt. P.-J. Kindermans, H. E. Sauceda, S. Chmiela, A. Tkatchenko, K.-R. Müller. SchNet: A continuous-filter convolutional neural network for modeling quantum interactions. Advances in Neural Information Processing Systems 30, pp. 992-1002 (2017) link
• C. Lu, Q. Liu, C. Wang, Z. Huang, P. Lin, L. He, Molecular Property Prediction: A Multilevel Quantum Interactions Modeling Perspective. The 33rd AAAI Conference on Artificial Intelligence (2019) link

How to use pytorch-geometric (PyG) dataloader for Alchemy

Environment setup

You need to install some packages before you can run the code. First you will need to install the Anaconda Distribution. Then you need to run the following commands to setup a anaconda environment and install the required libraries.

conda env create -f environment.yml

conda activate Alchemy

pip install -r requirements.txt

How to download Alchemy dataset

You may want to use the script at pyg/data-bin/download.sh. This script downloads dev.zip and valid.zip and extracts them at pyg/data-bin/raw.

You can also download manually from the homepage of Alchemy contest.

How to run PyG mpnn model

If you download Alchemy dataset and extract at pyg/data-bin/raw, you can simply run pyg/mpnn.py for training. After training, the example mpnn model will dump a target.csv file which is ready to submit to CodaLab for evaluation.

Citation

Please cite as:

@article{chen2019alchemy,
  title={Alchemy: A Quantum Chemistry Dataset for Benchmarking AI Models},
  author={Chen, Guangyong and Chen, Pengfei and Hsieh, Chang-Yu and Lee, Chee-Kong and Liao, Benben and Liao, Renjie and Liu, Weiwen and Qiu, Jiezhong and Sun, Qiming and Tang, Jie and Zemel, Richard and Zhang, Shengyu},
  journal={arXiv preprint arXiv:1906.09427},
  year={2019}
}