Lightrun Answers was designed to reduce the constant googling that comes with debugging 3rd party libraries. It collects links to all the places you might be looking at while hunting down a tough bug.
And, if you’re still stuck at the end, we’re happy to hop on a call to see how we can help out.
MoleculeNet PyTorch Port
See original GitHub issuePyTorch makes it much easier than Tensorflow to define dynamic models. Some cheminformatics deep models (notably http://pubs.acs.org/doi/abs/10.1021/ci400187y) are very dynamic and may be easier to implement in PyTorch than in DeepChem.
I’d like us to start building up a suite of PyTorch models. The install for PyTorch looks very simple (conda install pytorch torchvision cuda80 -c soumith for the Cuda 8.0 version). We can use MoleculeNet to benchmark the new models. The goal would be to replicate the MoleculeNet model implementations in PyTorch so new contributors can use either Tensorflow or PyTorch as desired.
I believe @miaecle has taken a crack at this already. If anyone else is interested in PyTorch porting, please chime in on this thread and we can coordinate.
Issue Analytics
- State:
- Created 7 years ago
- Reactions:4
- Comments:6 (4 by maintainers)
Top Results From Across the Web
Top Related Medium Post
Top Related StackOverflow Question
Troubleshoot Live Code
Top Related Reddit Thread
Top Related Hackernoon Post
Top Related Tweet
Top Related Dev.to Post
Top Related Hashnode Post
hey guys, it’s cool that you’re thinking of a pytorch port. what kind of help could i offer?
@soumith Hi, thanks for participating! PyTorch is a real wonderful package, it has been a pleasure to play around with it. We are currently still on the very first step of utilizing the model(and planning to use it to build up some more fancy models), will let you know if there is any concerns or requests!