[Frontend][Core] Move logits processor construction to engine #7666
Conversation
Signed-off-by: Joe Runde <[email protected]>
|
👋 Hi! Thank you for contributing to the vLLM project. Once the PR is approved and ready to go, please make sure to run full CI as it is required to merge (or just use auto-merge). To run full CI, you can do one of these:
🚀 |
There was a problem hiding this comment.
Thanks @joerunde! I haven't had a chance to look in detail yet but have a couple of initial comments:
- I agree we should check that these changes don't result in any worse overhead from LP construction due to different thread blocking behaviour for example
- Not sure whether you looked at Support for guided decoding for offline LLM #6878 which was added recently, but I guess it would make sense to reconcile those changes as part of this. If we're adding new params into SamplingParams then we should probably deprecate/remove the new LLM
guided_options_requestarg which was added there
| class LogitsProcessorParams: | ||
| """Parameters required to construct one or more logits processors""" | ||
| logit_bias: Optional[Union[Dict[int, float], Dict[str, float]]] | ||
| allowed_token_ids: Optional[List[int]] | ||
| guided_decoding_request: Optional[ # type: ignore | ||
| 'GuidedDecodingRequest'] # noqa |
There was a problem hiding this comment.
I think these should probably just be top-level sampling params, the fact that these get turned into logits processors is a bit of an implementation detail.
Ah shoot, yeah I just saw those in the conflicts, not sure how those didn't end up here when I started this :( I could go either way on adding these to sampling params or using a |
|
closing for #8252 |
Followup for #6797
This PR moves the logits processor construction out of the frontend api server, instead passing through the relevant request parameters so that the
LLMEnginecan construct our standard set of logits processors. This does not remove thelogits_processorsparameter fromSamplingParamsto maintain compatibility for other use cases besides the OpenAI frontend where users may be passing in their own logits processors.The goal here is to
I still need to do some profiling and look at all the tests. I may have to end up putting back an async method for constructing these logits processors that can run in a separate thread in the engine, I think as-is compiling a hefty FSM might block up the gpu event loop.
PR Checklist (Click to Expand)
Thank you for your contribution to vLLM! Before submitting the pull request, please ensure the PR meets the following criteria. This helps vLLM maintain the code quality and improve the efficiency of the review process.
PR Title and Classification
Only specific types of PRs will be reviewed. The PR title is prefixed appropriately to indicate the type of change. Please use one of the following:
[Bugfix]for bug fixes.[CI/Build]for build or continuous integration improvements.[Doc]for documentation fixes and improvements.[Model]for adding a new model or improving an existing model. Model name should appear in the title.[Frontend]For changes on the vLLM frontend (e.g., OpenAI API server,LLMclass, etc.)[Kernel]for changes affecting CUDA kernels or other compute kernels.[Core]for changes in the core vLLM logic (e.g.,LLMEngine,AsyncLLMEngine,Scheduler, etc.)[Hardware][Vendor]for hardware-specific changes. Vendor name should appear in the prefix (e.g.,[Hardware][AMD]).[Misc]for PRs that do not fit the above categories. Please use this sparingly.Note: If the PR spans more than one category, please include all relevant prefixes.
Code Quality
The PR need to meet the following code quality standards:
format.shto format your code.docs/source/if the PR modifies the user-facing behaviors of vLLM. It helps vLLM user understand and utilize the new features or changes.Notes for Large Changes
Please keep the changes as concise as possible. For major architectural changes (>500 LOC excluding kernel/data/config/test), we would expect a GitHub issue (RFC) discussing the technical design and justification. Otherwise, we will tag it with
rfc-requiredand might not go through the PR.What to Expect for the Reviews
The goal of the vLLM team is to be a transparent reviewing machine. We would like to make the review process transparent and efficient and make sure no contributor feel confused or frustrated. However, the vLLM team is small, so we need to prioritize some PRs over others. Here is what you can expect from the review process:
action-requiredlabel on the PR if there are changes required. The contributor should address the comments and ping the reviewer to re-review the PR.Thank You
Finally, thank you for taking the time to read these guidelines and for your interest in contributing to vLLM. Your contributions make vLLM a great tool for everyone!