Neural Monte Carlo

The main part of this repository contains the python package neumc. It contains all the necessary ingredients to build, train and sample neural network models for lattice field theories (LTF). It implements two-dimensional phi^4 scalar field, U(1) gauge and Schwinger (U(1) gauge with Wilson fermions) LTF.

The rest of the repository contains the auxiliary files like scripts and notebooks that show typical usages of the package. Once the installation, described in the next section, is complete, you can start with notebooks in the notebooks directory or one of the scripts in scripts directory.

For example you can run the script that trains the phi^4 model by running

python ./scripts/phi4.py 

This will start training the phi^4 scalar field model on a 8x8 lattice with m^2 == 1.25 and lambda == 0.0 (free field) using the REINFORCE estimator. The training will be on 'cuda' device if available. More information can be found in the scripts/README.md file. The whole script should take less than a minute on a GPU and several minutes on a CPU.

The rest of the scripts is described in the scripts/README.md file.

Installation

To use Python modules from this repository, you should install them as the package neumc. You will need Python>=3.11 and pip3>= 21.3. So please upgrade your python installation beforehand. As always it is best to create a separate virtual Python environment for the project. Probably the easiest way is to use venv, but you may use conda/mamba if you are familiar with it. Let's assume that you will call this environment neumc-env.

Linux/Unix

In your terminal, go to the directory where you want to store this environment and type

python -m venv neumc-env

to create an environment and then activate it by running

. ./neumc-env/bin/activate

Now you can start installing required packages. Go to the root directory of this repository and type

pip install -e neumc/

followed by

pip install -r neumc/requirements.txt

to install required packages (at the moment only numpy, scipy). The -e option installs the package in so-called developer mode. This allows you to make changes to the code inside package that will be immediately visible without reloading the package.

PyTorch

You will also need to install PyTorch. I am not giving any instructions for that, as it depends on your operating system and hardware. You can get instruction directly from their site.

If you would like to use notebooks provided in the repository, you will also need to install jupyterlab, jupytext and matplotlib. You can do that by running

pip install notebook jupyterlab jupytext matplotlib

Of course, you can use some other environment to view and run notebooks like e.g. Visual Studio Code or PyCharm, but you will need jupytext to transform notebooks that are stored in this repository as R Markdown (.Rmd) files. That can be done via the command line like this

jupytext some_notebook.Rmd --to notebook 

That will create a some_notebook.ipynb file. In jupyter lab this happens automatically when you have the jupytext.toml in you directory (it is provided in this repository). Then it is enough to click on the file name and choose Open With>Notebook option from the menu.

After finishing work you can deactivate the environment by running

deactivate

If something goes very wrong, you can always delete the whole environment by running

rm -rf  neumc-env

Windows

On Windows you essentially follow up same steps in command prompt or PowerShell, but you activate the environment by running the activate script

neumc-env\Scripts\activate