docs: Add interactive plots with Plotly

[akihironitta]

📚 Documentation

It would be nice to have interactive plots instead of the current screenshots of losses and metrics on tensorboard in our docs.

We can employ sphinx-plotly-directive like Optuna, a hyperparameter optimization framework, does in their docs.

Example

Source

.. plotly::

   >>> import plotly.express as px
   >>> px.scatter(x=[0, 1, 2, 3, 4], y=[0, 1, 4, 9, 16])

Output

Alternative

Suggested by @Borda http://nipy.org/nipy/devel/guidelines/sphinx_helpers.html#inserting-matplotlib-plots

[Borda]

see an alternative with basic matplotlib: http://nipy.org/nipy/devel/guidelines/sphinx_helpers.html#inserting-matplotlib-plots