Overview
GPU4PySCF is an NVIDIA GPU acceleration plugin for PySCF that moves PySCF objects to GPU computation via .to_gpu(). It supports molecular and periodic Hartree-Fock and DFT, analytical derivatives, excited states, solvent models, properties, geometry optimization and non-adiabatic dynamics.
The project shares the pyscf organization with PySCF and is released as separate pip packages (gpu4pyscf-cuda11x/cuda12x/cuda13x), requiring CuPy and cuTENSOR, under Apache-2.0. GPU acceleration is not artificial intelligence; this resource is categorized as a computational chemistry tool.
Key Features
- NVIDIA GPU acceleration plugin for PySCF, moving computation via .to_gpu().
- Supports molecular/periodic HF and DFT, analytical derivatives, excited states and geometry optimization.
- Released as separate pip packages requiring CuPy and cuTENSOR.
Use Cases
- Accelerate PySCF HF/DFT and geometry optimization on NVIDIA GPUs.
- Use GPUs to handle larger molecular or periodic systems.
How to Use
Install gpu4pyscf-cudaXXx and cutensor matching your CUDA version, then call .to_gpu() on PySCF objects. See the GitHub README.
Sources and verification
This entry describes capabilities documented by the linked project sources. Catalog edits are not evidence of a successful run, compatibility test or scientific performance evaluation.
Read the editorial and verification policy