search

gasteigerjo / dimenet

DimeNet and DimeNet++ models, as proposed in "Directional Message Passing for Molecular Graphs" (ICLR 2020) and "Fast and Uncertainty-Aware Directional Message Passing for Non-Equilibrium Molecules" (NeurIPS-W 2020)
https://www.daml.in.tum.de/dimenet
Other
286 stars 60 forks source link

[Notice] PyTorch version #4

Closed akirasosa closed 4 years ago

akirasosa commented 4 years ago

Hi

I would like to share that I have ported this repo to PyTorch version. Though it's during training still, I have seen quite good result. https://github.com/akirasosa/pytorch-dimenet

Feel free to close this post, as this is just a notice.

Thanks for great paper, Akira

gasteigerjo commented 4 years ago

Nice, well done! I can spot some differences in the weight initialization, optimizer and training schedule. I'm curious, please do let me know how it goes! :)

Also, we couldn't get half-precision to work properly, which seemed reasonable considering that the achieved accuracy of DimeNet is better than 16-bit machine precision for some targets. But maybe something else was the issue? If you can get this to work, I'd be very interested!