torch-m3gnet¶
🚧This repository is heavily under construction!🚧
This is an unofficial PyTorch implementation of M3GNet. The authors’ reimplementation of MEGNet and M3GNet with PyTorch/DGL will be available here.
Documentation https://lan496.github.io/torch-m3gnet/
Limitations¶
Only support structures with PBC
Only support potential prediction
No support:
state
,MaterialGraph.phi
Installation¶
See docs/installation.md.
Preparing benchmark datasets¶
Development¶
Docuement
sphinx-autobuild --host 0.0.0.0 --port 8000 docs docs_build
GPU container
docker run --gpus '"device=1"' -it -v $(pwd):/app -p 6006:6006 m3gnet
# tensorboard
tensorboard --logdir results/ --host 0.0.0.0 --port 6006
References¶
MatDeepLearn https://github.com/vxfung/MatDeepLearn
change_transfer_nnp https://github.com/pfnet-research/charge_transfer_nnp
nn-template https://github.com/grok-ai/nn-template