Added Support for guided decoding in offline interface #4130
Conversation
kevinbu233
changed the title
Co-authored-by: Jack Boylan <70636379+jackboyla@users.noreply.github.com>
| self.guided_regex is not None, | ||
| self.guided_choice is not None, | ||
| self.guided_grammar is not None, | ||
| self.guided_json_object is not None, |
There was a problem hiding this comment.
Issue
One typical usage pattern in OpenAI is to request tool use and json response format. This is because OpenAI does not support real guided decoding as vLLM does through Outlines and lm-format-enforcer. OpenAI can still generate invalid JSON or a JSON not respecting the specified schema.
In vLLM this is not necessary. At the same time, not allowing this pattern in vLLM means people will have to treat calls to OpenAI and vLLM differently. So this is not good.
Currently, if someone passes a named function in a tool call (the only supported tool call in vLLM) together with the response format requesting a json answer, the named function call will take precedence; the response format option will be ignored. Since the json schema in the tools will necessarily specify a json output, this ends up behaving the way users would expect.
|
This is awesome and exactly what I hope to have. I will take a look at it tmrw! |
| ) | ||
|
|
||
|
|
||
| def get_guided_decoding_logits_processor( |
There was a problem hiding this comment.
For the offline use case it makes sense to have a non-async way to get the processor. But orthogonal to that is the issue of sharing of stateful logits processors. In PR #5329 both concerns are addressed because it adds a factory with an async and a sync builder method.
|
Closing as superseded by #6878. |
Added support for guided decoding (JSON, regex, choice) on offline interface that is raised in issue #3536 .
This is a continuation and build off the work of #2815
BEFORE SUBMITTING, PLEASE READ THE CHECKLIST BELOW AND FILL IN THE DESCRIPTION ABOVE
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!