Scalable Ecosystem, Driver, and Analyzer for Complex Chemistry Simulations (SEDACS) enables massively parallel atomistic simulations that can seamlessly integrate with a diverse range of available and emerging quantum chemistry codes at different levels of theory.
Supporting ab initio, semiempirical quantum mechanics (SEQM),and coarse-grained flexible charge equilibration (ChEQ) models, this is a unified framework to simulate and analyze the MD of complex chemical systems and materials.
SEDACS also enables the anlysis of trajectories using novel graph-based ML schemes and quantum-response information to capture and visualize hidden, non-local quantum features that cannot be seen from the geometry alone.
Finally, SEDACS provides advanced mixed-precision electronic structure solver library that uses AI-hardware accelerators.
Our target customer is a Computational Chemist domain expert working on complex materials systems or developing new quantum capabilities that can easily be deployed at scale.
We hence provide transparent implementations that closely follow a “white-board” physics and mathematics presentation. The threshold to understand and work with this codebase for a domain expert is purposely kept low.
conda create -n sedacs pytorch::pytorch torchvision torchaudio openmpi mpi4py scipy jupyter nb_conda_kernels python=3.10 -c pytorchconda create -n sedacs pytorch torchvision torchaudio pytorch-cuda=11.8 openmpi mpi4py scipy jupyter nb_conda_kernels python=3.10 -c pytorch -c nvidiaThe current codebase has the following folder structure:
.
├── docs
├── driver
├── gpu
├── latte
├── mods
├── proxya
└── test
Proxya code as explained in the proposal. This proxy code should perform up to a full SCF optimization of the density matrix. It is written in three different languages: Python, Fortran, and C.
This is an implementation of the GPU/AI-solver library.
This is a code that generates "Latte" Hamiltonians from input coordinates
files (xyz or pdb) and constructs the density matrix.
Auxiliary Python modules. These modules will be used as building blocks to develop SEDACS.
Scripts to exercise the code.
This program is open source under the BSD-3 License.