Infatoshi/batmobile
High-performance CUDA kernels for equivariant graph neural networks (MACE, NequIP, Allegro). 10-20x faster than e3nn.
This project dramatically speeds up the core calculations for complex simulations in molecular dynamics and materials science. It takes in structural data and other relevant inputs for molecular systems, and outputs significantly faster computations for models like MACE, NequIP, and Allegro. Researchers and engineers working on detailed atomic and molecular simulations will find this highly beneficial.
Use this if you are running molecular dynamics or materials science simulations with equivariant graph neural networks and need to significantly reduce computation time for models operating at L_max=3.
Not ideal if your work does not involve molecular dynamics or materials science, or if your graph neural network models do not use the specific equivariant architectures targeted.
Stars
43
Forks
6
Language
Cuda
License
MIT
Category
Last pushed
Jan 17, 2026
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/ml-frameworks/Infatoshi/batmobile"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Higher-rated alternatives
apache/tvm
Open Machine Learning Compiler Framework
uxlfoundation/oneDNN
oneAPI Deep Neural Network Library (oneDNN)
Tencent/ncnn
ncnn is a high-performance neural network inference framework optimized for the mobile platform
OpenMined/TenSEAL
A library for doing homomorphic encryption operations on tensors
iree-org/iree-turbine
IREE's PyTorch Frontend, based on Torch Dynamo.