einops-0.5.0.tar.gz

<!-- this link magically rendered as video, unfortunately not in docs --> <!-- <a href='http://arogozhnikov.github.io/images/einops/einops_video.mp4' > <div align="center"> <img src="http://arogozhnikov.github.io/images/einops/einops_video.gif" alt="einops package examples" /> <br> <small><a href='http://arogozhnikov.github.io/images/einops/einops_video.mp4'>This video in high quality (mp4)</a></small> <br><br> </div> </a> --> https://user-images.githubusercontent.com/6318811/177030658-66f0eb5d-e136-44d8-99c9-86ae298ead5b.mp4 # einops [](https://github.com/arogozhnikov/einops/actions/workflows/run_tests.yml) [](https://badge.fury.io/py/einops) [](https://einops.rocks/)  Flexible and powerful tensor operations for readable and reliable code. Supports numpy, pytorch, tensorflow, jax, and [others](#supported-frameworks). ## Recent updates: - einsum is now a part of einops - [Einops paper](https://openreview.net/pdf?id=oapKSVM2bcj) is accepted for oral presentation at ICLR 2022 (yes, it worth reading) - flax and oneflow backend added - torch.jit.script is supported for pytorch layers - powerful EinMix added to einops. [Einmix tutorial notebook](https://github.com/arogozhnikov/einops/blob/master/docs/3-einmix-layer.ipynb) <!--<div align="center"> <img src="http://arogozhnikov.github.io/images/einops/einops_logo_350x350.png" alt="einops package logo" width="250" height="250" /> <br><br> </div> --> ## Tweets > In case you need convincing arguments for setting aside time to learn about einsum and einops... [Tim Rockt채schel, FAIR](https://twitter.com/_rockt/status/1230818967205425152) > Writing better code with PyTorch and einops �윉� [Andrej Karpathy, AI at Tesla](https://twitter.com/karpathy/status/1290826075916779520) > Slowly but surely, einops is seeping in to every nook and cranny of my code. If you find yourself shuffling around bazillion dimensional tensors, this might change your life [Nasim Rahaman, MILA (Montreal)](https://twitter.com/nasim_rahaman/status/1216022614755463169) [More testimonials](https://einops.rocks/pages/testimonials/) ## Recordings of talk at ICLR 2022 <a href='https://iclr.cc/virtual/2022/oral/6603'> <img width="922" alt="Screen Shot 2022-07-03 at 1 00 15 AM" src="https://user-images.githubusercontent.com/6318811/177030789-89d349bf-ef75-4af5-a71f-609896d1c8d9.png"> </a> Watch [a 15-minute talk](https://iclr.cc/virtual/2022/oral/6603) that focuses on main problems of standard tensor manipulation methods, and how einops improves this process. ## Contents - [Installation](#Installation) - [Documentation](https://einops.rocks/) - [Tutorial](#Tutorials) - [API micro-reference](#API) - [Why using einops](#Why-using-einops-notation) - [Supported frameworks](#Supported-frameworks) - [Contributing](#Contributing) - [Repository](https://github.com/arogozhnikov/einops) and [discussions](https://github.com/arogozhnikov/einops/discussions) ## Installation <a name="Installation"></a> Plain and simple: ```bash pip install einops ``` <!-- `einops` has no mandatory dependencies (code examples also require jupyter, pillow + backends). To obtain the latest github version ```bash pip install https://github.com/arogozhnikov/einops/archive/master.zip ``` --> ## Tutorials <a name="Tutorials"></a> Tutorials are the most convenient way to see `einops` in action - part 1: [einops fundamentals](https://github.com/arogozhnikov/einops/blob/master/docs/1-einops-basics.ipynb) - part 2: [einops for deep learning](https://github.com/arogozhnikov/einops/blob/master/docs/2-einops-for-deep-learning.ipynb) - part 3: [improve pytorch code with einops](https://arogozhnikov.github.io/einops/pytorch-examples.html) ## API <a name="API"></a> `einops` has a minimalistic yet powerful API. Three operations provided ([einops tutorial](https://github.com/arogozhnikov/einops/blob/master/docs/) shows those cover stacking, reshape, transposition, squeeze/unsqueeze, repeat, tile, concatenate, view and numerous reductions) ```python from einops import rearrange, reduce, repeat # rearrange elements according to the pattern output_tensor = rearrange(input_tensor, 't b c -> b c t') # combine rearrangement and reduction output_tensor = reduce(input_tensor, 'b c (h h2) (w w2) -> b h w c', 'mean', h2=2, w2=2) # copy along a new axis output_tensor = repeat(input_tensor, 'h w -> h w c', c=3) ``` And two corresponding layers (`einops` keeps a separate version for each framework) with the same API. ```python from einops.layers.chainer import Rearrange, Reduce from einops.layers.gluon import Rearrange, Reduce from einops.layers.keras import Rearrange, Reduce from einops.layers.torch import Rearrange, Reduce from einops.layers.tensorflow import Rearrange, Reduce ``` Layers behave similarly to operations and have the same parameters (with the exception of the first argument, which is passed during call) ```python layer = Rearrange(pattern, **axes_lengths) layer = Reduce(pattern, reduction, **axes_lengths) # apply created layer to a tensor / variable x = layer(x) ``` Example of using layers within a model: ```python # example given for pytorch, but code in other frameworks is almost identical from torch.nn import Sequential, Conv2d, MaxPool2d, Linear, ReLU from einops.layers.torch import Rearrange model = Sequential( Conv2d(3, 6, kernel_size=5), MaxPool2d(kernel_size=2), Conv2d(6, 16, kernel_size=5), MaxPool2d(kernel_size=2), # flattening Rearrange('b c h w -> b (c h w)'), Linear(16*5*5, 120), ReLU(), Linear(120, 10), ) ``` <!--- Additionally two auxiliary functions provided ```python from einops import asnumpy, parse_shape # einops.asnumpy converts tensors of imperative frameworks to numpy numpy_tensor = asnumpy(input_tensor) # einops.parse_shape gives a shape of axes of interest parse_shape(input_tensor, 'batch _ h w') # e.g {'batch': 64, 'h': 128, 'w': 160} ``` --> ## Naming <a name="Naming"></a> `einops` stands for Einstein-Inspired Notation for operations (though "Einstein operations" is more attractive and easier to remember). Notation was loosely inspired by Einstein summation (in particular by `numpy.einsum` operation). ## Why use `einops` notation?! <a name="Why-using-einops-notation"></a> ### Semantic information (being verbose in expectations) ```python y = x.view(x.shape[0], -1) y = rearrange(x, 'b c h w -> b (c h w)') ``` While these two lines are doing the same job in *some* context, the second one provides information about the input and output. In other words, `einops` focuses on interface: *what is the input and output*, not *how* the output is computed. The next operation looks similar: ```python y = rearrange(x, 'time c h w -> time (c h w)') ``` but it gives the reader a hint: this is not an independent batch of images we are processing, but rather a sequence (video). Semantic information makes the code easier to read and maintain. ### Convenient checks Reconsider the same example: ```python y = x.view(x.shape[0], -1) # x: (batch, 256, 19, 19) y = rearrange(x, 'b c h w -> b (c h w)') ``` The second line checks that the input has four dimensions, but you can also specify particular dimensions. That's opposed to just writing comments about shapes since [comments don't work and don't prevent mistakes](https://medium.freecodecamp.org/code-comments-the-good-the-bad-and-the-ugly-be9cc65fbf83) as we know ```python y = x.view(x.shape[0], -1) # x: (batch, 256, 19, 19) y = rearrange(x, 'b c h w -> b (c h w)', c=256, h=19, w=19) ``` ### Result is strictly determined Below we have at least two ways to define