[Frontend] Skip stop in reasoning content

[chaunceyjiang]

Purpose

When the stop="" parameter is set, it causes the reasoning phase to stop when encountering a stop token, which interrupts the process and prevents the user from seeing any content.

Test Plan

from openai import OpenAI

# Modify OpenAI's API key and API base to use vLLM's API server.
openai_api_key = "EMPTY"
openai_api_base = "http://localhost:8000/v1"

client = OpenAI(
    api_key=openai_api_key,
    base_url=openai_api_base,
)

models = client.models.list()
model = models.data[0].id

# Round 1
messages = [{"role": "user", "content": "9.11 and 9.8, which is greater?"}]
response = client.chat.completions.create(
    model=model,
    messages=messages,
    stop="9.8",
)

reasoning_content = response.choices[0].message.reasoning_content
content = response.choices[0].message.content

print("reasoning_content for Round 1:", reasoning_content)
print("-" * 80)
print("content for Round 1:", content)

Test Result

reasoning_content for Round 1: 
Okay, so I need to figure out whether 9.11 is greater than 9.8 or not. Let me start by recalling how to compare decimal numbers. I think the general rule is to look at the digits from left to right and compare them one by one. 

First, both numbers are in the ones place. Let me write them down to visualize better:

9.11 and 9.8
...
...
...
So, in conclusion, 9.8 is greater than 9.11.

--------------------------------------------------------------------------------
content for Round 1: 

To determine which number is greater between **9.11** and **

---

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.
• (Optional) Release notes update. If your change is user facing, please update the release notes draft in the Google Doc.

[chaunceyjiang]

/cc @mgoin @njhill @aarnphm PTAL.

[mergify]

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

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

[mergify]

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

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

[chaunceyjiang]

/cc @njhill PTAL.

[gaocegege]

I think this only addresses part of the performance problem though. For the other part, I think changes would be needed on the reasoning parser side. There should be a way of doing incremental checking, rather than searching the entire prompt + output tokens for every generated token (while reasoning)...

@njhill I agree with your point. I’ll look into this issue and then submit a new PR. Currently, is_reasoning_end is used in multiple places.

I’ll explore how to perform incremental checks.

Could you please create an issue to keep track? It is a potential problem since day 1, I am also interested in it.

[mergify]

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

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

[chaunceyjiang]

I think this only addresses part of the performance problem though. For the other part, I think changes would be needed on the reasoning parser side. There should be a way of doing incremental checking, rather than searching the entire prompt + output tokens for every generated token (while reasoning)...

Hi, @gaocegege @njhill PTAL. #25735

[mergify]

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

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

[bbartels]

@chaunceyjiang is there something still missing from having this merged?

[mergify]

Hi @chaunceyjiang, the pre-commit checks have failed. Please run:

uv pip install pre-commit
pre-commit install
pre-commit run --all-files

Then, commit the changes and push to your branch.

For future commits, pre-commit will run automatically on changed files before each commit.

[github-actions]

This pull request has been automatically marked as stale because it has not had any activity within 90 days. It will be automatically closed if no further activity occurs within 30 days. Leave a comment if you feel this pull request should remain open. Thank you!