Migrate documentation from Sphinx to Mkdocs about pytorch-widedeep HOT 2 CLOSED

• automate.py script goes through whole library and gathers funbctions and classes
  puts them into mkgendocs.yml
• gendocs util tranforms mkgendocs.yml into .md files in docs_mk
• mkdocs build tranforms md files in docs_mk into documentation in /site
• [optional] mkdocs serve will create http server to check the documentation
• [optional] running of automate script can be part of prehook, but this add additional overhead so I think it's enough to keep the procedure only in readme.md or somewhere else

notes:

• mkdocs does not like classes without init(), so I added init(self):pass where it was missing
• automate script uses improtlib which does not like relative imports so I changed relative import "from .wdtypes import *" at couple of places to absolute path

procedure to generate documentation (all in root dir):

pip install -r mkdocs_requirements.txt
python automate.py
gendocs --config mkgendocs.yml
mkdocs build
# if you want to run webserver
# mkdocs serve

from pytorch-widedeep.

• jupyternotebook are now displayed in documentation - each notebook has to start with level 1 heading, I adjusted couple of notebooks that started with level 2
• mkgendocs copies everything from templates dir to sources_dir (sources_dir is recreated each time), if we want to have examples/notebooks/*.ipynb in the documentation we have to copy them to templates ourselves - fixed it at the end of automate.py
• markdown "image links" in readme.md were not displayed correctly, I changed them to HTML - fixed
• I changed mknotebooks plugin to mkdocs-jupyter as new one includes scrollbar if printout is wider than width of webpage(eg pandas with many columns)
• I copied webpage(documentation) related files (pictures, favicon, etc) to docs_mk/templates, again as contents of this directory is copied to docs_mk/sources when gendocs is run

summary
new documentation looks awesome! @jrzaurin let me know when you have time and I can give you 15min tour how it works, afterwards I create pullrequest to merge the changes

from pytorch-widedeep.