search

ergodicio / adept

Automatic-Differentiation-Enabled Plasma Transport in JAX
MIT License
27 stars 3 forks source link
automatic-differentiation boltzmann-equation computational-fluid-dynamics fluid-mechanics kinetic-modeling machine-learning plasma-physics

readme

ADEPT

Docs Tests

ADEPT

ADEPT is an Automatic Differentation Enabled Plasma Transport code.

Installation

User

pip install git+https://github.com/ergodicio/adept.git

Developer

Conda

  1. Install conda (we recommend mamba)
  2. mamba env create -f env.yaml or mamba env create -f env_gpu.yaml
  3. mamba activate adept

pip

  1. python3 -m venv venv
  2. source venv/bin/activate
  3. pip3 install -r requirements.txt

Docs

https://adept.readthedocs.io/en/latest/

Examples

https://github.com/ergodicio/adept-notebooks

There are other ways to use ADEPT, notably as part of a neural network training pipeline that leverages differentiable simulation. In reference [1], neural networks are trained to learn forcing functions that drive the system towards previously unseen behavior. In reference [2], neural networks are trained to help bridge the micro-macro physics gap in multiphysics simulations.

Usage

python3 run.py --cfg {config_path} without the .yaml extension

This runs the simulation defined in the config and stores the output to an mlflow server.

Unless you have separately deployed an mlflow server somewhere, it simply writes files using the mlflow specification to the current working directory.

To access and visualize the results, it is easiest to use the UI from the browser by typing mlflow ui in the command line from the same directory.

Contributing guide

The contributing guide is in development but for now, just make an issue / pull request and we can go from there :)

Citation

If you are using this package for your research, please cite the following

A. Joglekar and A. Thomas, “ADEPT - automatic differentiation enabled plasma transport,” 
ICML - SynS & ML Workshop (https://syns-ml.github.io/2023/contributions/), 2023

References

[1] A. S. Joglekar & A. G. R. Thomas. "Unsupervised discovery of nonlinear plasma physics using differentiable kinetic simulations." J. Plasma Phys. 88, 905880608 (2022).

[2] A. S. Joglekar and A. G. R. Thomas, “Machine learning of hidden variables in multiscale fluid simulation,” Mach. Learn.: Sci. Technol., vol. 4, no. 3, p. 035049, Sep. 2023, doi: 10.1088/2632-2153/acf81a.