Molecular AutoEncoder in PyTorch

Aug 29, 2021 1 min read

MolEncoder

Install

$ git clone https://github.com/cxhernandez/molencoder.git && cd molencoder
$ python setup.py install

Download Dataset

$ molencoder download --dataset chembl22

Train

$ molencoder train --dataset data/chembl22.h5

Add --cuda flag to enable CUDA. Add --cont to continue training a model from a checkpoint file.

Pre-Trained Model

A pre-trained reference model is available in the ref/ directory. Currently, it performs with ~98% accuracy on the validation set after 100 epochs of training. However, if you succeed at training a better model, feel free to submit a pull request!

TODO

[x] Implement encoder
[x] Implement decoder
[x] Add download command
[x] Add train command
[ ] Add encode command
[ ] Add decode command
[x] Add pre-trained model

Shoutouts

Original paper by Gómez-Bombarelli, et al.
keras-molecules by Max Hodak
DeepChem
PyTorch

GitHub

https://github.com/cxhernandez/molencoder

PyTorch Deep Learning

John was the first writer to have joined pythonawesome.com. He has since then inculcated very effective writing and reviewing culture at pythonawesome which rivals have found impossible to imitate.