sparsity-pattern
Generate different types of sparsity pattern for sparse matrices.
Installation
The sparsity-pattern git repo is available as PyPi package
pip install sparsity-pattern
pip install git+ssh://git@github.com/ulf1/sparsity-pattern.git
Usage
The block-diagonal pattern for tensorflow
import sparsity_pattern
import tensorflow as tf
n_rows, n_cols = 10, 12
idx = sparsity_pattern.get('block', min(n_rows, n_cols), block_sizes=[3, 1, 2])
mat = tf.sparse.SparseTensor(
dense_shape=(n_rows, n_cols),
indices=tf.convert_to_tensor(idx, dtype=tf.int64),
values=range(1, len(idx)+1))
print(tf.sparse.to_dense(mat))The circle pattern for pytorch
import sparsity_pattern
import torch
n_rows, n_cols = 5, 7
idx = sparsity_pattern.get('circle', min(n_rows, n_cols), offsets=[1, 2])
mat = torch.sparse_coo_tensor(
indices=torch.tensor(idx).transpose(0, 1),
values=range(1, len(idx)+1),
size=[n_rows, n_cols])
print(mat.to_dense())The triu pattern for scipy
import sparsity_pattern
import scipy.sparse
import numpy as np
n, k = 4, -1
idx = sparsity_pattern.get('triu', n, k)
idx_rows, idx_cols = np.array(idx)[:, 0], np.array(idx)[:, 1]
mat = scipy.sparse.lil_matrix((n, n), dtype=np.int64)
mat[idx_rows, idx_cols] = range(1, len(idx)+1)
print(mat.todense())Check the examples folder for more notebooks.
Appendix
Install a virtual environment
python3 -m venv .venv
source .venv/bin/activate
pip3 install --upgrade pip
pip3 install -r requirements-dev.txt
pip3 install -r requirements-demo.txt
(If your git repo is stored in a folder with whitespaces, then don't use the subfolder .venv. Use an absolute path without whitespaces.)
Python commands
- Jupyter for the examples:
jupyter lab - Check syntax:
flake8 --ignore=F401 --exclude=$(grep -v '^#' .gitignore | xargs | sed -e 's/ /,/g') - Run Unit Tests:
pytest
Publish
pandoc README.md --from markdown --to rst -s -o README.rst
python setup.py sdist
twine upload -r pypi dist/*Clean up
find . -type f -name "*.pyc" | xargs rm
find . -type d -name "__pycache__" | xargs rm -r
rm -r .pytest_cache
rm -r .venv
Support
Please open an issue for support.
License and citation
This software is licensed under Apache License 2.0 and archived on Zenodo. If you would like to cite the software, please use this DOI: 10.5281/zenodo.4357290.
Contributing
Please contribute using Github Flow. Create a branch, add commits, and open a pull request.
