Add support for Grammar/Tools + TGI-based specs in InferenceClient #2237
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
2 tasks
|
The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update. |
Wauplin
changed the title
Wauplin
changed the title
LysandreJik
approved these changes
Apr 23, 2024
|
Thanks for the review @LysandreJik! Agree with you about "I'm of the opinion of ship it and eventually fix if issues arise," 😄 |
Wauplin
added a commit
to huggingface/huggingface.js
that referenced
this pull request
Apr 30, 2024
From an idea mentioned by @OlivierDehaene and @drbh in #579 and [slack thread](https://huggingface.slack.com/archives/C05CFK1HM0T/p1711360022125399) (internal). This PR adds a script `inference-tgi-import.ts` to generate the `text-generation` and `chat_completion` specs from the auto-generated TGI [specifications](https://huggingface.github.io/text-generation-inference/). The goal is to keep in sync TGI improvements with @huggingface/tasks and therefore have a consistent "single source of truth". The converted specs are then compatible with our tooling to generate JS/Python code. This PR changes quite a lot of naming in the generated JS/Python types. Luckily I don't think they were yet used in the JS ecosystem so better to do that now than later. I also opened huggingface/huggingface_hub#2237 to include these changes in `huggingface_hub`. **TODO:** - [ ] fix lint errors. How to deal with a parsed json to avoid using `any`? - [ ] CI workflow to open a PR each time TGI is updated? => can be done in a future PR --------- Co-authored-by: SBrandeis <[email protected]>
This was referenced Jun 5, 2024
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Sorry for the huge PR 😬 It goes hand-in-hand with huggingface/huggingface.js#629 (and to a lesser extent huggingface/text-generation-inference#1798). Most of the changes to review are hopefully documentation / auto-generated stuff.
What's in this PR?
text_generationandchat_completionhave been updated based on TGI specs (see associated PR Generate specs from TGI openapi.json huggingface.js#629)text_generationtaskchat_completiontaskMODEL_KWARGS_NOT_USED_REGEX.text_generation, only send non-None parameters in payload.check_inference_input_params.pythat checks task parameters are correctly used and documented, to be consistent with the generated types. For now, raises an error when not consistent but doesn't provide an auto-fix (see TODOs in script).generate_async_inference_client.pyto generate KO package reference correctly.What to review
Mostly:
src/huggingface_hub/inference/_client.py=> lots of docs update (not interesting) + few tweaks in code (to review)src/huggingface_hub/inference/_common.py=> only small tweakstests/test_inference_client.py=> to check "it works"tests/test_inference_text_generation.py=> to check "it works"utils/check_inference_input_params.py=> new script to check input parametersThe rest is either auto-generated stuff or low-level scripts.
Generated docs:
What's not this PR?
=> will be handled in a follow-up PR