[ENH] Improving docs, examples, and tutorials of pytorch-forecasting

[julian-fong]

relates to #1718

Implements the proposed changes in above issue.

[fkiraly]

question, since this is not a draft - is this complete or still in progress? The changes currently are only a new table of contents, which is nice but not the entire scope of the issue.

[julian-fong]

It's still in progress at the moment! Do drafts also trigger the read the docs build? I'm primarily keeping it in this state so that I can see how it looks like after changes

[fkiraly]

yes, CI is triggered both by "ready" and "draft" PR. The "draft" state is mainly cosmetic and for signposting that it is not ready.

[julian-fong]

yes, CI is triggered both by "ready" and "draft" PR. The "draft" state is mainly cosmetic and for signposting that it is not ready.

Thanks! If you have anymore suggestions, feel free to post it here in or in the issue thread :). I'm not sure if we are planning to refactor the TimeSeriesDataSet class, but I feel like storing some pretrained dataset objects somewhere in the repo would make it easy to load and would make the tutorial much more simple.

[fkiraly]

I'm not sure if we are planning to refactor the TimeSeriesDataSet class,

Yes, we probably will, see here: #1736
Comments appreciated, especially on usage!

What did you have in mind for the examples?