Skip to content
forked from evenmn/VMaChine

General variational Monte-Carlo solver written in C++

License

Notifications You must be signed in to change notification settings

mtrynoha/VMaChine

 
 

Repository files navigation

VMaChine

VMaChine is a general variational Monte Carlo (VMC) solver written in object-oriented C++. It was implemented with focus on quantum dot systems, and standard trial wave functions (Hermite functions) are implemented. However, we primary aim of examining trial wave functions where a less amount of physical intuition is required. Therefore, several trial wave functions based on neural networks and machine learning are also included, hence VMaChine. The code is largely parallelizable and is well-suited for large clusters.

Basic usage

After compiling the code, a configuration file input.in is run by

vmachine input.in

on a single thread or

mpirun -n 4 vmachine input.in

on 4 CPUs (might be any number).

What's new?

Update 2022-08-01

  • Added Xavier initialisation class
  • Now supporting weight initialisation from file

Update 2022-07-11

  • Added better comments
  • Splitted up system.cpp in several files to increase readability

Update 2021-01-01

  • An interaction class is added. The old style "interaction: true" is now superseded by "interactionStyle: coulomb".

Update 2020-12-28

  • The log (printed to the terminal by default) is now more readable both for a human eye and for the computer
  • Examples added

Update 2020-12-27

  • Fixed bug in CMakeLists.txt
  • Removed support for Armadillo, as it was found to be more confusing than useful

Prerequisites

To run the code without issues, C++14 is required. In addition, a few external packages are needed:

  • MPI
  • Eigen

MPI

MPI is used for parallel processing. On Linux, the package can be installed by the following commands

sudo apt-get install libopenmpi-dev
sudo apt-get install openmpi-bin

MPI is also avaliable on other platforms.

Eigen

Eigen is a C++ template library for linear algebra operations. See http://eigen.tuxfamily.org/ for installation details.

Install

VMaChine can be installed by cloning this repository. This is preferably done in the home directory:

cd
git clone https://github.com/evenmn/VMaChine.git

Then copy Eigen header files and blocker files to VMaChine:

cd ~/Download/eigen-3.3.9   # insert correct path here
cp -r Eigen ~/VMaChine/include/

Build

The code can be compiled by either CMake or QMake

CMake (recommended)

Build VMaChine in the usual CMake way:

mkdir build
cd build
cmake ..
make -j8

The executable is then found in ~/VMaChine/build folder, which is added to the bash resource path by

echo "export PATH=~/VMaChine/build:\$PATH" >> ~/.bashrc

QMake (QT-creator)

  1. Download QT-creator
  2. Configure the building file src/vmachine.pro

The project can then be run in QT-creator using ctrl + R.


Running simulation

The desired simulation parameters are specified in a configuration file. A typical configuration file can be found in input.in:

# system
numParticles: 2
numDimensions: 2
hamiltonian: harmonicOscillator
omega: 1.0
interactionStyle: coulomb

# wave function
waveFunctionElement: gaussian
waveFunctionElement: padeJastrow

# simulation
numIterations: 100
numSteps: 100000
learningRate: 0.1
stepLength: 0.1

The configuration file is simply run by

vmachine input.in

License

MIT

About

General variational Monte-Carlo solver written in C++

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • C++ 98.2%
  • C 1.2%
  • QMake 0.3%
  • CMake 0.1%
  • Python 0.1%
  • Haxe 0.1%