Update to match required Tensorflow versions

[brendenconnors]

layers.Normalization() does not exist in Tensorflow 2.3.0, which is listed as a requirement.

And Tensorflow_probability 0.11.0 (listed) is stable against tf 2.3.0 shown here https://github.com/tensorflow/probability/releases?page=2

[fchollet]

We should target the latest releases whenever possible. Please try it with TF 2.0 and update the requirements accordingly.

[brendenconnors]

Ran with updated versions and ran correctly without the "experimental" reference. 😃

[fchollet]

Please also update the md and ipynb files.

[brendenconnors]

Arxiv link caused issue when creating md/ipynb files. Updated link in .py file and generated md/ipynb files.

[fchollet]

LGTM, thank you!

[fchollet]

There are some failing tests related to special characters:

UnicodeDecodeError: 'utf-8' codec can't decode byte 0x95 in position 704: invalid start byte
The command '/bin/sh -c python autogen.py make' returned a non-zero code: 1
make: *** [Makefile:2: container-test] Error 1

Did you use a non-utf8 character somewhere?

[brendenconnors]

Hmm no not that I am aware of. Could it be a Windows issue?

Seeing this Specials character as opposed to the dot locally in md file.
It looks like forward slashes are also being changed to backward slashes in that line? After removing the Specials character and rerunning, I get a further error because of bad escape \m. Probably user error, but seems like maybe windows paths are causing problems.

For this part of the displayed page

Not really sure when this would happen during generation

[brendenconnors]

Changed back slashes to forward slashes

[brendenconnors]

updated changes to image path and metadata name. 👍

[fchollet]

Thank you!