mcbal / afem

Implementation of approximate free-energy minimization in PyTorch

Repository from Github https://github.commcbal/afem

artificial-intelligence deep-learning transformers attention-mechanism partition-functions free-energy statistical-physics spin-models pytorch steepest-descent saddle-point ising-spin-models neural-networks boltzmann-machines hopfield-networks

Approximate Free-Energy Minimization

Experimental implementation of approximate free-energy minimization in PyTorch.

Blog post: Transformers from Spin Models: Approximate Free Energy Minimization

Setup

Install package in editable mode:

$ pip install -e .

Run (very slow) tests with:

$ python -m unittest

About

Implementation of approximate free-energy minimization in PyTorch

artificial-intelligence deep-learning transformers attention-mechanism partition-functions free-energy statistical-physics spin-models pytorch steepest-descent saddle-point ising-spin-models neural-networks boltzmann-machines hopfield-networks

MIT License

Languages

Language:Python 86.1%Language:Jupyter Notebook 13.9%