[yaml] add RunInference support with VertexAI

[Polber]

This PR adds support for calling RunInference with preliminary support for VertexAI through the use of the VertexAIModelHandlerJSON ModelHandler.

In order to implement more robust validation for the transform, the SafeLineLoader was refactored out of yaml_transform.py into a separate yaml_utils.py file to avoid import cycles.

---

Thank you for your contribution! Follow this checklist to help us incorporate your contribution quickly and easily:

• Mention the appropriate issue in your description (for example: addresses #123), if applicable. This will automatically add a link to the pull request in the issue. If you would like the issue to automatically close on merging the pull request, comment fixes #<ISSUE NUMBER> instead.
• Update CHANGES.md with noteworthy changes.
• If this contribution is large, please file an Apache Individual Contributor License Agreement.

See the Contributor Guide for more tips on how to make review process smoother.

To check the build health, please visit https://github.com/apache/beam/blob/master/.test-infra/BUILD_STATUS.md

GitHub Actions Tests Status (on master branch)

See CI.md for more information about GitHub Actions CI or the workflows README to see a list of phrases to trigger workflows.

[Polber]

R: @robertwb

[Polber]

R: @damccorm

[github-actions]

Stopping reviewer notifications for this pull request: review requested by someone other than the bot, ceding control. If you'd like to restart, comment assign set of reviewers

[robertwb]

Thanks, this'll be great!

[robertwb]

This won't actually get called until runtime, right? Perhaps we should call this default_preprocess_fn() and have it return an lambda (or raise an error). And then change postprocess_fn for symmetry.

[Polber]

Great suggestion, I ended up making the preprocess parameter required on the VertexAI handler, so the error is redundant in this case... but I think moving forward it could be useful, and the refactor in general is cleaner.

[robertwb]

Could we ask the handler for the type of inference? (Presumably the example type is that of the input as well.) Some handlers may not be able to provide this, but some can.

[damccorm]

We don't generally today, but we could for a very limited set of handlers. The ones I can think of with predictable output types are hugging face pipelines and vLLM; the rest are all dependent on the model.

I think this is probably worth doing when we can, it probably requires some slight modification to the PredictionResult type, though, and might be worth scoping into a follow on PR.

[damccorm]

We don't generally today, but we could for a very limited set of handlers. The ones I can think of with predictable output types are hugging face pipelines and vLLM; the rest are all dependent on the model.

I think this is probably worth doing when we can, it probably requires some slight modification to the PredictionResult type, though, and might be worth scoping into a follow on PR.

[codecov]

Codecov Report

Attention: Patch coverage is 55.39568% with 62 lines in your changes missing coverage. Please review.

Project coverage is 57.42%. Comparing base (d7502fa) to head (e8bc920).
Report is 52 commits behind head on master.

Files with missing lines	Patch %	Lines
sdks/python/apache_beam/yaml/yaml_ml.py	39.21%	62 Missing ⚠️

Additional details and impacted files

@@             Coverage Diff              @@
##             master   #33406      +/-   ##
============================================
+ Coverage     57.38%   57.42%   +0.03%     
  Complexity     1475     1475              
============================================
  Files           973      980       +7     
  Lines        154978   155311     +333     
  Branches       1076     1076              
============================================
+ Hits          88939    89187     +248     
- Misses        63829    63908      +79     
- Partials       2210     2216       +6     

Flag	Coverage Δ
python	81.22% <55.39%> (-0.05%)	⬇️

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

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

[damccorm]

This LGTM, just had one more suggestion - looks like linting is also broken

[Polber]

@damccorm Failing tests now appear to be unrelated

[damccorm]

Thanks!