Added CONTRIBUTING.md

[vfdev-5]

Related #2651

Description:

• Added a draft for CONTRIBUTING.md

This PR is not closing the issue #2651 as there are 2 sections to be defined later.

EDIT: There is already a PR submitted by the bot : #1564
I was not aware before.

[codecov]

Codecov Report

Merging #2663 (8fb19e4) into master (ce34258) will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff           @@
##           master    #2663   +/-   ##
=======================================
  Coverage   72.75%   72.75%           
=======================================
  Files          99       99           
  Lines        8979     8979           
  Branches     1431     1431           
=======================================
  Hits         6533     6533           
  Misses       1999     1999           
  Partials      447      447           

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update ce34258...8fb19e4. Read the comment docs.

[datumbox]

Any chance we could merge this to master? It's a good source of info for new comers as it describes setting up the dev environment. We could continue improving it post merge.

[jgbradley1]

This is just a suggestion. Should new datasets support at a minimum, the same constructor arguments (when appropriate)?

• root
• split
• target_type
• target_transform
• transform
• download - for large datasets that are unreasonable to download, should always raise a RuntimeError when download=True

[datumbox]

@vfdev-5 @fmassa I really think we should merge this, even if there are a few things left to TBD. This doc contains useful information that will allow new contributors to learn how to build the library from source. What is the bare minimum we need to do to merge this?

[fmassa]

Thanks a lot for working on this @vfdev-5 !

Let's get this merged, and at the same time can you create an issue (or keep the existing one open) so that we track completing the points about datasets / models?

[fmassa]

@jgbradley1 that's a good suggestion, although I'm not sure it can be generically implemented everywhere. FYI there was some discussion already in the past about dataset standardization in #1080

[vfdev-5]

@fmassa let's keep existing issue open and I add a message about that.