Add notebook smoke tests, QOL changes, part 2

[matthewcarbone]

Same as #38 but I actually remembered to pull from upstream first:

Just playing around/exploring the code a bit. While I was I figured I'd make a few QOL changes to the testing system:

• Trying something new by adding a smoke test to the notebook. Figured it could be useful considering the notebooks are hosted on Colab and should always work. It's a fun hack that you can use: ipython -c "%run simpleGP.ipynb", for example.
• Added osx-latest to the runs-on in the testing jobs. Since the project is open source your CI hours are free so it doesn't matter that it's a more expensive OS to test on.
• Added python version 3.11 to the list of tested versions.
• Added .flake8 and pyproject.toml files for running black (with line length 127).
• Ran black on gpax.models.gp as a test (feel free to tell me you don't like this and I'll revert, but stylistically I think it's much easier to read!).

(there will no doubt be bugs which I will fix 👍)

[matthewcarbone]

@ziatdinovmax looks like these jobs just failed with no error messages... Doesn't even look like they ran. 😁
Any idea what's going on?

[matthewcarbone]

@ziatdinovmax Pretty sure I found the bug. I just did something really silly in the workflow files.

[codecov-commenter]

Codecov Report

Merging #39 (b5348d0) into main (6d5bc4c) will decrease coverage by 0.04%.
The diff coverage is 95.23%.

❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

@@            Coverage Diff             @@
##             main      #39      +/-   ##
==========================================
- Coverage   95.60%   95.57%   -0.04%     
==========================================
  Files          39       39              
  Lines        3072     3072              
==========================================
- Hits         2937     2936       -1     
- Misses        135      136       +1     

Flag	Coverage Δ
unittests	95.57% <95.23%> (-0.04%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files	Coverage Δ
gpax/models/gp.py	93.08% <95.23%> (ø)

... and 1 file with indirect coverage changes

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[ziatdinovmax]

Looks like the notebook tests are failing.
@matthewcarbone can you please look into it?

[matthewcarbone]

@ziatdinovmax Yeah I think it's because I missed two notebook-specific dependencies. That's fixed now. Question is whether or not this will work anyway.

I do think notebooks should have smoke tests, but if this doesn't work I'll revert the smoke tests pieces and experiment elsewhere 😁. Once I get it right I can just open another PR.

[ziatdinovmax]

@matthewcarbone For some reason, I can't run the osx workflows. Given that JAX support on the Apple M series is still limited and buggy, I'm not sure how helpful they are anyway. Maybe you can remove them for now?

[matthewcarbone]

@ziatdinovmax yeah sure I can remove them. OSX is more expensive than Linux resource-wise, though it's odd it's been 13 hours...

Will remove them but definitely consider putting them back in the future. Note I don't believe that OSX latest == M1. Stay tuned.

[matthewcarbone]

@ziatdinovmax start of the new FY so I'll probably start working on this a bit more! How strict are you regarding the code coverage differences? I didn't actually change any code, I just put some new lines in for clarity 😁

[ziatdinovmax]

I wouldn't worry about the slight decrease in the code coverage due to basic formatting.