https://github.com/epronovost/eincheck

Tensor shape checks inspired by einstein notation
https://github.com/epronovost/eincheck

deep-learning jax numpy pytorch tensor tensorflow

Last synced: 2 days ago
JSON representation

Tensor shape checks inspired by einstein notation

• Host: GitHub
• URL: https://github.com/epronovost/eincheck
• Owner: EPronovost
• License: mit
• Created: 2022-10-29T02:38:07.000Z (almost 2 years ago)
• Default Branch: main
• Last Pushed: 2024-09-12T04:32:30.000Z (15 days ago)
• Last Synced: 2024-09-24T01:31:10.737Z (3 days ago)
• Topics: deep-learning, jax, numpy, pytorch, tensor, tensorflow
• Language: Python
• Homepage:
• Size: 203 KB
• Stars: 3
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# eincheck

[![CI](https://github.com/epronovost/eincheck/actions/workflows/pr.yaml/badge.svg)](https://github.com/epronovost/eincheck/actions/workflows/pr.yaml)

[![Documentation Status](https://readthedocs.org/projects/eincheck/badge/?version=main)](https://eincheck.readthedocs.io/en/main/?badge=main)

[![PyPI version](https://badge.fury.io/py/eincheck.svg)](https://badge.fury.io/py/eincheck)

Tensor shape checks inspired by einstein notation

## Overview

This library has three main functions:

* `check_shapes` takes tuples of `(Tensor, shape)` and checks that all the Tensors match the shapes

```

check_shapes((x, "i 3"), (y, "i 3"))

```

* `check_func` is a function decorator to check the input and output shapes of a function

```

@check_func("*i x, *i y -> *i (x + y)")

def concat(a, b):

    return np.concatenate([a, b], -1)

```

* `check_data` is a class decorator to check the fields of a data class

```

@check_data(start="i 2", end="i 2")

class LineSegment2D(NamedTuple):

    start: torch.Tensor

    end: torch.Tensor

```

For more info, [read the docs!](https://eincheck.readthedocs.io/en/main)