What's new

This site is a complete redesign of the current MXNet document site available at https://mxnet.apache.org. For readers:

1. Changed to a new sphinx theme based on the material design, which is the most modern sphinx theme we found. We customized it to fit MXNet and related projects such as GluonCV/GluonNLP/DiveIntoDeepLearning.

   

2. Separated front languages into different sites. So readers can select a language and then focus on it. For example, the currently available site is for Python only.

3. Re-organized contents into two folders: guide/ for tutorials and api/ for API references. Guidelines are further grouped into a) getting started, b) using mxnet to implement deep learning algorithms, and c) learning mxnet modules. APIs are grouped into ndarray, gluon, gluon-related, symbol, symbol-related and advanced.

4. All guidelines are jupyter notebooks, they are executed during building the website.

   

5. Gluon first. Both guidelines and API references are focusing on Gluon.

6. Enabled right TOC, which list the section titles in a local page, which is useful to navigate for long pages.

   

7. Every class and method has its own page to improve load time. Example We can browse all classes and methods on the right global TOC.

   

8. Added disqus at the bottom of almost every page, to make it easier to make comment or ask questions. We tried to embed the current mxnet forum as well, but it's less convenient as disqus.

   

9. Used card to list important topics.

   

We also simplified how to develop the website.

1. Removed almost all js hacks in the original site.

2. Simplified the installation pages, reduced the lines of codes by 10x.

3. Uses .md for tutorials and .rst for others. Tutorials will be converted into .ipynb during building. .rst has a rich set of directives, it makes writing sophisticated things easier. Such as adding a card:

   .. card::
      :title: GluonCV
      :link: https://gluon-cv.mxnet.io
   
      A deep learning toolkit for computer vision.