use aesara and pymc4 instead of theano and pymc3

[AndrewAnnex]

…No clue if CI builds would work. todo need to update the github worker to allow this branch to run.

likely there are a lot of api breaking changes I did not address here, I did not attempt to run it locally yet. Also the CI environment for gempy is a bit confusing (still using travis?). Likely these scripts could be cleaned up a bit more to make testing easier

Description

swap out all usage of theano for aesara and pymc3 for pymc4.

Relates to
#701

Checklist

• My code follows the PEP 8 style guidelines.
• My code uses type hinting for function and method arguments and return values.
• My code contains descriptive and helpful docstrings
  which are formatted per the Google Python Style Guidelines.
• I have created tests which entirely cover my code.
• The test code either 1. demonstrates at least one valuable use case (e.g. integration tests)
  or 2. verifies that outputs are as expected for given inputs (e.g. unit tests).
• New and existing tests pass locally with my changes.

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[Japhiolite]

The checks were always failing at the installation step. Apparently with aesara 2.7.8 there'd been a change with setuptools (aesara-devs/aesara#1050). Out of curiosity, I temporary froze aesara to 2.7.7 just to see whether the checks pass. Seems...like there are all tests passing now 🎉
Will have a look at this setuptools thing to get aesara out of the ice-box.

[AndrewAnnex]

we should also do some benchmarks with a gpu to ensure things are still fast, can probably help with that

[AndrewAnnex]

@Japhiolite that bug in aesara was fixed a couple of weeks ago. I can perform some benchmarking tests if needed

[Japhiolite]

That would be great @AndrewAnnex

I've talked to @Leguark about merging, and we thought that we publish the switch to aesara as gempy 2.3

[Japhiolite]

Currently, the remaining failing tests are due to a (in the meantime fixed) deprecation in subsurface. Unfortunately I can't deploy a new subsurface version to pypi.

But I was wondering though....whether we just should move forward with this PR - as is - to Version 2.3 and work from there to mainly get the pandas fixes up to speed with the new major version...with pandas 2.0 apparently coming out within the next week.

Thoughts?

[AlexanderJuestel]

Agree with proceeding to get GemPy properly working again

[Japhiolite]

The newest arviz version ( v0.15.1 ) does not have a from_pymc3() method anymore...in fact, I don't see a from_pymc() method either....So I guess we'll have to see how to fix this as we go by.

[AndrewAnnex]

closing as per #796