https://github.com/cgarciae/einop

https://github.com/cgarciae/einop

Last synced: 7 days ago
JSON representation

• Host: GitHub
• URL: https://github.com/cgarciae/einop
• Owner: cgarciae
• License: mit
• Created: 2022-03-07T20:50:27.000Z (over 2 years ago)
• Default Branch: master
• Last Pushed: 2022-03-08T13:13:40.000Z (over 2 years ago)
• Last Synced: 2024-10-08T18:55:55.538Z (about 1 month ago)
• Language: Python
• Size: 48.8 KB
• Stars: 57
• Watchers: 4
• Forks: 2
• Open Issues: 1
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# Einop

_One op to rule them all_

Einop is a very thin wrapper around [einops](https://github.com/arogozhnikov/einops) that combines `rearrange`, `reduce`, and `repeat` into a single `einop` function. This library is a port of [arogozhnikov/einops#91](https://github.com/arogozhnikov/einops/pull/91) by [Miles Cranmer](https://github.com/MilesCranmer) into a separate library, if at some point that PR is merged use `einop` directly from einops instead.

## Installation

```

pip install einop

```

## Usage

```python

import numpy as np

from einop import einop

x = np.random.uniform(size=(10, 20))

y = einop(x, "height width -> batch width height", batch=32)

assert y.shape == (32, 20, 10)

```

#### Rearrange

```python

x = np.random.randn(100, 5, 3)

einop(x, 'i j k -> k i j').shape

>>> (3, 100, 5)

```

#### Reduction

```python

x = np.random.randn(100, 5, 3)

einop(x, 'i j k -> i j', reduction='sum').shape

>>> (100, 5)

```

#### Repeat

```python

x = np.random.randn(100, 5, 3)

einop(x, 'i j k -> i j k l', l=10).shape

>>> (100, 5, 3, 10)

```