[V0 deprecation] Guided decoding

[rzabarazesh]

Essential Elements of an Effective PR Description Checklist

• The purpose of the PR, such as "Fix some issue (link existing issues this PR will resolve)".
• The test plan, such as providing test command.
• The test results, such as pasting the results comparison before and after, or e2e results
• (Optional) The necessary documentation update, such as updating supported_models.md and examples for a new model.

Purpose

Remove v0 guided decoding code.

Test Plan

Test Result

(Optional) Documentation Update

[gemini-code-assist]

Code Review

This pull request effectively removes the v0 guided decoding functionality, which simplifies the codebase significantly. The changes mostly consist of deleting old files and removing references to the deprecated v0 guided decoding API, which has been done correctly across the board.

However, I've identified a critical inconsistency regarding the outlines guided decoding backend. While its implementation has been removed, it's still listed as a valid backend in the configuration, leading to contradictory behavior in the tests. My review includes a detailed comment and a suggestion to resolve this ambiguity.

[github-actions]

👋 Hi! Thank you for contributing to the vLLM project.

💬 Join our developer Slack at https://slack.vllm.ai to discuss your PR in #pr-reviews, coordinate on features in #feat- channels, or join special interest groups in #sig- channels.

Just a reminder: PRs would not trigger full CI run by default. Instead, it would only run fastcheck CI which starts running only a small and essential subset of CI tests to quickly catch errors. You can run other CI tests on top of those by going to your fastcheck build on Buildkite UI (linked in the PR checks section) and unblock them. If you do not have permission to unblock, ping simon-mo or khluu to add you in our Buildkite org.

Once the PR is approved and ready to go, your PR reviewer(s) can run CI to test the changes comprehensively before merging.

To run CI, PR reviewers can either: Add ready label to the PR or enable auto-merge.

🚀

[mergify]

This pull request has merge conflicts that must be resolved before it can be
merged. Please rebase the PR, @rzabarazesh.

https://docs.github.com/en/pull-requests/collaborating-with-pull-requests/working-with-forks/syncing-a-fork

[rzabarazesh]

Doesn't this overlap with #20928? cc @aarnphm

Unfortunately I didn't see #20928 as this task was unclaimed. This PR should cover everything in #20928 as well as some missing integration points (llm_engine.py, etc) and broken tests.

[mergify]

This pull request has merge conflicts that must be resolved before it can be
merged. Please rebase the PR, @rzabarazesh.

https://docs.github.com/en/pull-requests/collaborating-with-pull-requests/working-with-forks/syncing-a-fork

[hmellor]

I've pushed a few tweaks and removed ready until we feel it's ready for merging

[hmellor]

Too mant relevant tests have been deleted/disabled because of erroneous edits made to LLMEngine. These bad edits removed support for logit_bias and bad_words in V0 and deleted the corresponding tests even if they were still relevant for V1.

[rzabarazesh]

Too mant relevant tests have been deleted/disabled because of erroneous edits made to LLMEngine. These bad edits removed support for logit_bias and bad_words in V0 and deleted the corresponding tests even if they were still relevant for V1.

Thank you for updating the PR @hmellor. I thought we wanted to remove logit_bias and bad_words in v0 as well.
Do you know why test_guided_generate is failing now? Should we skip it for V0?

[hmellor]

I suppose ultimately we want to remove all of V0, but removing logit_bias and bad_words in this PR meant touching more tests to disable them in V0 (hence why I started to revert those changes).

@WoosukKwon should we also be removing logit_bias and bad_words in this PR?

[hmellor]

Do you know why test_guided_generate is failing now? Should we skip it for V0?

I didn't spend much time looking into those ones. I added them back so that they could be run on V1. We could change the buildkite pipeline to only run this file in V1?

[WoosukKwon]

should we also be removing logit_bias and bad_words in this PR?

I'm ok with either, as long as these features are supported and tested in V1. V0 is officially deprecated anyways.