trainsum 0.0.1

A Python package for dealing with quantics tensor trains

Welcome to trainsum, a Python package designed for working with quantics tensor trains. The development was done by the ZAQC-team at the Fraunhofer Institute for Graphical Data Analysis (IGD). trainsum is licensed under EUPL 1.2 (similar to GPL).

The main features are:

• easy definition of N-dimensional tensor trains

• quantization of dimensions independent of their size

• einsum-operations equivalent to NumPy’s einsum function

• generic backends for NumPy, Torch and CuPy

• tensorized solver for eigenvalue equations and linear equation systems

Installation

You can install trainsum using pip:

pip install trainsum

The dependencies are:

• numpy

• array_api_compat

• opt_einsum

• hdf5

Documentation

The documentation for trainsum can be found at https://trainsum.readthedocs.io.