cymetric is a Python package for learning of moduli-dependent Calabi-Yau metrics using neural networks implemented in TensorFlow.
NOTE development of the package continues here:
https://github.com/pythoncymetric/cymetric
This version will no longer be updated.
The current version is an alpha-release so not all features mentioned below are on the main branch yet. Features with an (*) will be released with the arXiv announcement of our paper.
- Point Generators for Complete Intersection Calabi-Yau manifolds and hypersurfaces from the Kreuzer-Skarke list (requires SageMath).
- A collection of custom TensorFlow models with different metric Ansätze.
- A mathematica API for the point generators and TensorFlow models(*).
- Custom models for the bundle metric(*).
- Documentation exists(*).
Create a new environment with
conda create -n cymetric python=3.9
Then install with pip directly from github
conda activate cymetric
pip install git+https://github.com/robin-schneider/cymetric.git
There are some tutorials
- In 1.PointGenerator.ipynb we explore the three different PointGenerators for codimension-1 CICY, general CICYs and CY in toric varieties on the Fermat Quintic.
- In 2.TensorFlow_models.ipynb we explore some of the TF custom models with the data generated in the first notebook.
- (*) There exists a Mathematica integration, which allows to call the PointGenerators and the TensorFlow models. Furthermore, there are arbitrary precision PointGenerators based on the wolfram language.
We welcome contributions to the project. Those can be bug reports or new features, that you have or want to be implemented. Please read more here.
There will soon be an accompanying paper.