ByteDance Quantum Chemistry: ByteQC

ByteQC is a high-performance, GPU-accelerated quantum chemistry package designed for large-scale quantum chemistry simulations. It currently supports a range of methods, including mean-field calculations for both open and periodic boundary conditions, MP2 simulations, CCSD, and CCSD(T) calculations. All of these methods are optimized to support multiple GPUs, enabling efficient parallelization. Additionally, ByteQC includes an integrated functionality for systematically improvable embedding (SIE), which allows for scalable simulations of complex systems. The package also exports several useful tools for the development of GPU-based quantum chemistry applications.

Exteranl dependencies

This package incorporates parts of its code adapted from several external open-source projects:

• https://github.com/bytedance/gpu4pyscf
• https://github.com/hongzhouye/pyscf/tree/rsdf_direct
• https://github.com/pyscf/pyscf
• https://github.com/sunqm/libcint
• https://github.com/cupy/cupy
• https://github.com/BoothGroup/Vayesta
• https://github.com/gkclab/libdmet_preview

Installation

Requirement:

• cupy (master branch in github)
• nvmath-python
• h5py
• mpi4py
• numpy
• scipy
• pyscf = 2.5.0

Build dependencied:

• libcutensor (installation from conda is also supported)
• libcublas

Build the package by run command python byteqc/setup.py.

Package structure

ByteQC
├── cucc          # CCSD and CCSD(T) solver
├── cump2         # MP2 solver
├── cupbc         # mean-field for PBC
├── cuobc         # mean-field for OBC
├── lib           # GPU utilities
├── embyte        # systematically improvable embedding
├── __init__.py
├── LICENSE
├── README.md
└── setup.py      # installation script

The lib subpackage is exported directly when importing ByteQC. The usage of each module is described in the README.md file in the corresponding directory.

Citations

@Misc{Huang2024,
  author    = {Huang, Zigeng and Guo, Zhen and Cao, Changsu and Pham, Hung Q. and Wen, Xuelan and Booth, George H. and Chen, Ji and Lv, Dingshun},
  title     = {Advancing surface chemistry with large-scale ab-initio quantum many-body simulations},
  year      = {2024},
  doi       = {10.48550/arXiv.2412.18553},
  keywords  = {Materials Science (cond-mat.mtrl-sci), Chemical Physics (physics.chem-ph), FOS: Physical sciences, FOS: Physical sciences},
  publisher = {arXiv},
}