- Install the required packages by running the following command:
pip install -r requirements.txt
- Download the SMILES dataset from Kaggle ZINC 250k. Change the file extension to
.smi
and remove the header row. - Place the prepared
.smi
file in thedatasets
folder.
- Use
preprocess.ipynb
and run the notebook to preprocess the.smi
file and obtain the tokenization mapping table.
We have provided a token file for this project, so you can skip the Load Training Data
section and proceed with running the rest of the code.
Note: Due to the nature of Variational Autoencoders (VAE), there might be instances where new compounds are not generated (sampling problem). If this happens, please run the code multiple times to obtain a valid compound.
This project is licensed under the Apache License 2.0 - see the LICENSE file for details.