https://github.com/muhammadhasyim/tps-torch

A PyTorch-based package for string methods and ML-inspired path sampling methods
https://github.com/muhammadhasyim/tps-torch

Last synced: 3 months ago
JSON representation

A PyTorch-based package for string methods and ML-inspired path sampling methods

• Host: GitHub
• URL: https://github.com/muhammadhasyim/tps-torch
• Owner: muhammadhasyim
• License: bsd-3-clause
• Created: 2020-11-03T04:25:41.000Z (about 4 years ago)
• Default Branch: main
• Last Pushed: 2022-03-12T01:50:11.000Z (over 2 years ago)
• Last Synced: 2024-07-04T01:00:51.266Z (4 months ago)
• Language: Python
• Size: 11.5 MB
• Stars: 2
• Watchers: 2
• Forks: 1
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# tps-torch

A PyTorch-based package implementing ML-based algorithms for computing committor functions and reactions. This package is used primarily in the following work:

- M. R. Hasyim, C. H. Batton, K. K. Mandadapu, "Supervised Learning and the Finite-Temperature String Method for Computing Committor Functions and Reaction Rates", [arXiv:2107.13522](https://arxiv.org/abs/2107.13522) (2021)

Before installing and building this project, it is important that you have CMake 2.8.12 above, any MPI implementation (OpenMPI or MPICH), and, of course, PyTorch (v 1.7.0 and above). 

At the moment, building and installation is just:

```console

mkdir build; cd build

cmake ../

make -j6

pip install .

```

Note that the code only supports runs on CPUs. 

Listed files and folders:

- `tpstorch`, contains C++ and Python source files

- `1dbrownian`, contains scripts to run the 1D Quartic Potential study

- `muller-brown`, contains scripts to build and run the 2D Muller-Brown potential study with the FTS method. 

- `muller-brown-ml`, contains scripts to build and run the 2D Muller-Brown potential study.