refac: made some attributes of VAEAlgorithmConfig optional

[federico-carrara]

Description

• What: Made all attributes, except model, Optional in VAEAlgortihmConfig.
• Why: At inference time the only the model attribute is needed. This allows to simplify model initialization for evaluation.
• How: Simply marked attributes as Optional, added None default if needed.

---

Please ensure your PR meets the following requirements:

• Code builds and passes tests locally, including doctests
• New tests have been added (for bug fixes/features)
• Pre-commit passes
• PR to the documentation exists (for bug fixes / features)

[jdeschamps]

That means that we would need to check whether they have been assigned for training.

Would it make sense to simply give them a default value?

[federico-carrara]

I think I got your point, but let me be sure. The problem you are pointing out is that in this way we could instantiate an Algorithm config that will raise a RuntimrError during training since some attributes are not defined, right?

If this is the case, yeah good point. I guess a solution is to change the default values of the pydantic configs, e.g.,

loss: Optional[LVAELossConfig] = LVAELossConfig()

What do you think?

[federico-carrara]

Or we could also skip this PR. However, it is really ugly to load a loss config or an optimizer config which are just useless at inference time...

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 87.80%. Comparing base (b29fc6c) to head (4f9e5bd).

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #261   +/-   ##
=======================================
  Coverage   87.80%   87.80%           
=======================================
  Files         137      137           
  Lines        4971     4971           
=======================================
  Hits         4365     4365           
  Misses        606      606           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

---

🚨 Try these New Features:

• Flaky Tests Detection - Detect and resolve failed and flaky tests

[jdeschamps]

Sorry I lost track of this.

I do prefer having an unused parameter class than guarding against it being None in the training calls.

[federico-carrara]

I completely see your point. Let's skip this PR then, not important at all!
If you agree I will close it :)