Skip to content

chat : force \n</think> on reasoning budget end for qwen3-coder - #28869

Merged
pwilkin merged 1 commit into
ggml-org:masterfrom
aldehir:qwen3-think-newline
Sep 16, 2026
Merged

pwilkin merged 1 commit into
ggml-org:masterfrom
aldehir:qwen3-think-newline

Conversation

@aldehir

@aldehir aldehir commented Sep 13, 2026

Copy link
Copy Markdown
Contributor

Overview

Force \n</think> on reasoning budget end for qwen3-coder templates. When forcing a message, it currently appends </think> which doesn't match the template rendering on subsequent requests, preventing prompt cache reuse. This forces the \n to be included.

ref: #23478 (comment)

Requirements

@aldehir
aldehir requested a review from a team as a code owner September 13, 2026 22:46
@aldehir
aldehir requested a review from ggerganov September 13, 2026 23:37
@aldehir aldehir added the merge ready A maintainer can use this label to indicate that they consider the changes final and ready to merge. label Sep 15, 2026
@pwilkin
pwilkin merged commit 0bec16e into ggml-org:master Sep 16, 2026
25 of 27 checks passed
quimmedes pushed a commit to quimmedes/cafe-llama.cpp that referenced this pull request Sep 16, 2026
zsogitbe pushed a commit to zsogitbe/llama.cpp that referenced this pull request Sep 17, 2026
frostyautumnleaf pushed a commit to frostyautumnleaf/llama.cpp that referenced this pull request Oct 5, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

merge ready A maintainer can use this label to indicate that they consider the changes final and ready to merge.

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants