nikola/llama_cpp
1
0
Fork 0
Code Issues Pull Requests Actions 17 Packages Projects Releases Wiki Activity

chat: trim messages sent to StepFun parser (fixes long reasoning loops) (#25238)

Browse Source 
* chat: trim messages sent to StepFun parser (fixes long reasoning loops)

* add regression test; remove duplicate template

* chat: trim StepFun content parts before rendering

The StepFun trim workaround ran on the already-rendered messages, where
typed content parts have been concatenated into a single string, so the
per-part whitespace could no longer be reached. Move the trim ahead of
rendering and apply it to content_parts text as well as the string
content and reasoning_content. Adds a content-parts regression test.

Co-Authored-By: Piotr Wilkin <ilintar@gmail.com>
Assisted-By: Claude Fable 5 <noreply@anthropic.com>

---------

Co-authored-by: tarruda <tpadilha84@gmail.com>
This commit is contained in:
Piotr Wilkin (ilintar)
2026-07-03 23:12:11 +02:00
committed by GitHub
co-authored by Piotr Wilkin tarruda
parent d4cff114c0
commit 2d973636e2
4 changed files with 80 additions and 82 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tests/test-chat-auto-parser.cpp
-1
View File
@@ -1887,7 +1887,6 @@ static void test_role_markers_all_templates(testing & t) {
{ "Qwen-Qwen3-0.6B.jinja", "<|im_start|>user", "<|im_start|>assistant" },
{ "Qwen-QwQ-32B.jinja", "<|im_start|>user", "<|im_start|>assistant" },
{ "StepFun3.5-Flash.jinja", "<|im_start|>user", "<|im_start|>assistant" },
{ "stepfun-ai-Step-3.5-Flash.jinja", "<|im_start|>user", "<|im_start|>assistant" },
// DeepSeek family
{ "deepseek-ai-DeepSeek-R1-Distill-Llama-8B.jinja", "<｜User｜>", "<｜Assistant｜>" },