Debian Codegraph @

docs/installation.md

raw
1 # Installation
2
3 `autoray` is available on both [pypi](https://pypi.org/project/autoray/) and
4 [conda-forge](https://anaconda.org/conda-forge/autoray). While `autoray` is
5 pure python and has no direct dependencies itself, the recommended distribution
6 would be [mambaforge](https://github.com/conda-forge/miniforge#mambaforge)
7 for installing the various backend array libraries and their dependencies.
8
9 **Installing with `pip`:**
10 ```bash
11 pip install autoray
12 ```
13
14 **Installing with `conda`:**
15 ```bash
16 conda install -c conda-forge autoray
17 ```
18
19 **Installing with `mambaforge`:**
20 ```bash
21 mamba install autoray
22 ```
23 ```{hint}
24 Mamba is a faster version of `conda`, and the -forge distritbution comes
25 pre-configured with only the `conda-forge` channel, which further simplifies
26 and speeds up installing dependencies.
27 ```
28
29 **Installing the latest version directly from github:**
30
31 If you want to checkout the latest version of features and fixes, you can
32 install directly from the github repository:
33 ```bash
34 pip install -U git+https://github.com/jcmgray/autoray.git
35 ```
36
37 **Installing a local, editable development version:**
38
39 If you want to make changes to the source code and test them out, you can
40 install a local editable version of the package:
41 ```bash
42 git clone https://github.com/jcmgray/autoray.git
43 pip install --no-deps -U -e autoray/
44 ```
45
46 ```{note}
47 **No-install version:**
48 The entirety of the automatic dispatch mechanism is contained in the single
49 file `autoray.py`, which you could simply copy into your project if you don't
50 want add a dependency.
51 ```
52
53 ## Optional plotting requirements
54
55 The [`autoray.lazy.draw`](autoray.lazy.draw) visualizations variously require:
56
57 * [`matplotlib`](https://matplotlib.org/)
58 * [`networkx`](https://networkx.org/) - for computational graph drawing
59 * [`pygraphviz`](https://pygraphviz.github.io/) - optional, for better and
60 faster graph layouts than `networkx`.