renormalizer 0.0.11

No project description provided

Renormalizer is a python package based on tensor network states for electron-phonon quantum dynamics.

Installation

pip install renormalizer

For users who are not familiar with python, step-by-step installation guide can be found in the document.

Documentation

Primitive documentation could be found here.

Notice

Renormalizer relies on linear algebra libraries such as OpenBLAS or MKL for matrix operations. These libraries by default parallel over as many CPU cores as possible. However, we have found empirically that the calculations carried out in Renormalizer has very poor parallelism efficiency over 4 cores (see this paper). Thus, we highly recommend limiting the number of parallel CPU cores to 4 for large scale calculations and 1 for small scale tests. To do this, set the environment variable before running Python

export RENO_NUM_THREADS=1

which sets all environment variables for underlying linear algebra libraries, such as MKL_NUM_THREADS.

After importing NumPy or Renormalizer, setting the environment variables will have no effect.

In fact, limiting the cores was once the default behavior of Renormalizer. It is later changed in this PR because Renormalizer is sometimes imported as a utility package.