Stop String Plumbing #24
Conversation
|
👋 Hi! Thank you for contributing to the vLLM project. 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 do one of these:
🚀 |
| await self.input_socket.send_multipart(( | ||
| self.encoder.encode(request_type), | ||
| self.encoder.encode(request), | ||
| ), |
There was a problem hiding this comment.
@robertgshaw2-neuralmagic - @njhill and I spoke about using a single socket so the poll doesn't become a bottleneck. as part of it, we send both the "Request Type" and "Request Data" - the "Request Type" is used to determine what appropriate decoder type in the LLMEngineCore. What do you think ?
There was a problem hiding this comment.
I like this design
There was a problem hiding this comment.
Thanks @varun-sundar-rabindranath!
We also need to wire in the abort for request cancellation here. That could either be made async or per my comment there we could just put it into a queue which we drain at the same time as the per-step aborts are done.
vllm/v1/engine/__init__.py
Outdated
| class EngineCoreRequestType(enum.Enum): | ||
| AddRequest = enum.auto() | ||
| AbortRequest = enum.auto() |
There was a problem hiding this comment.
I don't think we need a class for this, we can just have e.g.
ENGINE_CORE_ADD_REQ = b'\00'
ENGINE_CORE_ABORT_REQ = b'\01'No need for msgspec decoder etc.
There was a problem hiding this comment.
This is a good idea. I have made the change but left the enum class encapsulation as-is for better code readability.
vllm/v1/engine/llm_engine.py
Outdated
| self.engine_core.abort_requests( | ||
| EngineCoreAbortRequest( | ||
| request_ids=request_id, | ||
| request_status=RequestStatus.FINISHED_STOPPED)) |
There was a problem hiding this comment.
the status here should be FINISHED_ABORTED I think
There was a problem hiding this comment.
We are overloading what it means to "abort" - I believe,
- if some sequence is aborted due to a stop_string detection, It should have "FINISHED_STOPPED"
- if some sequence is aborted due to client termination, it should have "FINISHED_ABORTED".
I thinkfinish_requestwould be a better name.
There was a problem hiding this comment.
sounds good to me.. I think this method on llm_engine is meant for the latter case though .. stopped is more "internal"
There was a problem hiding this comment.
I see what you are saying - let me change it to FINISHED_ABORTED in this PR and try to maybe add a separate function for detokenization stop strings in another PR.
varun-sundar-rabindranath
force-pushed
the
varun/9826-stop-strings
branch
from
fc507d5 to
78af701
Compare
varun-sundar-rabindranath
force-pushed
the
varun/9826-stop-strings
branch
from
78af701 to
35580c6
Compare
|
nice work! |
Plumb stop-string stop sequence from Detokenizer to LLMEngineCore
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.Adding or changing kernels
Each custom kernel needs a schema and one or more implementations to be registered with PyTorch.
Tensorsrequire meta-functions. Meta-functions should be implemented and registered in python so that dynamic dims can be handled automatically. See above documents for a description of meta-functions.torch.libary.opcheck()to test the function registration and meta-function for any registered ops. Seetests/kernelsfor examples.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!