feat: Adds support for zero-shot document classification (#7669)

deepset-ai / haystack

AI orchestration framework to build customizable, production-ready LLM applications. Connect components (models, vector DBs, file converters) to pipelines or agents that can interact with your data. With advanced retrieval methods, it's best suited for building RAG, question answering, semantic search or conversational agent chatbots.

https://haystack.deepset.ai

Apache License 2.0

17.72k stars 1.92k forks source link

feat: Adds support for zero-shot document classification (#7669) #8193

Closed jpatra72 closed 2 months ago

jpatra72 commented 3 months ago

Related Issues

fixes #7669

Proposed Changes:

Added support for zero-shot document classification aligned with 1.x version. Additionally, introduced support for multi-label classification.

How did you test it?

unit tests
integration tests

Notes for the reviewer

did not implement a .batch_predict feature as found in v1.x (since I couldn't identify a similar implementation across v2.x).
added the classification_field arg to the TransformersZeroShotDocumentClassifier class, consistent with v1.x version

Checklist

I have read the contributors guidelines and the code of conduct
I have updated the related issue with new insights and changes
I added unit tests and updated the docstrings
I've used one of the conventional commit types for my PR title: fix:, feat:, build:, chore:, ci:, docs:, style:, refactor:, perf:, test:.
I documented my code
I ran pre-commit hooks and fixed any issue

CLAassistant commented 3 months ago

All committers have signed the CLA.

shadeMe commented 3 months ago

Thanks for the PR! Could you fix the lints/tests?

jpatra72 commented 3 months ago

@shadeMe , no problem! I think the issue was that the licence info was missing in the scripts. I have fixed it now.

coveralls commented 3 months ago

Pull Request Test Coverage Report for Build 10779480371

Details

0 of 0 changed or added relevant lines in 0 files are covered.
No unchanged relevant lines lost coverage.
Overall coverage increased (+0.02%) to 90.33%

Totals
Change from base Build 10774951533:	0.02%
Covered Lines:	7146
Relevant Lines:	7911

💛 - Coveralls

jpatra72 commented 3 months ago

@shadeMe, the pr now passes all the git checks!

dfokina commented 2 months ago

Hi! I will review the docstrings for this today 👀

jpatra72 commented 2 months ago

Hi @dfokina , your suggestions seem very valid. I have incorporated all of them (Y)

jpatra72 commented 2 months ago

@shadeMe , any idea why the unit test fails here - check results ? The fail test refers to the following line - Link I did not run into this issue when I run the pytest on my macos.

shadeMe commented 2 months ago

@shadeMe , any idea why the unit test fails here - check results ? The fail test refers to the following line - Link I did not run into this issue when I run the pytest on my macos.

Maybe there's a version mismatch b'ween the libraries used in the CI and your local env? Have you tried running the failing hatch commands locally? The failures seen above are in the lints, not the tests.

jpatra72 commented 2 months ago

@shadeMe , I was referring to the failures in the following test - Link. This was before i made some mods to the doc strings based on dfokina's feedback. I have already taken care of linting issues locally. :)

shadeMe commented 2 months ago

Hmm, looking at the failure, seems like the test is making some assumptions about the presence of the HF token in the execution environment? You might want to try removing any HF-related env vars and retest.