froggeric

/

Qwen-Fixed-Chat-Templates

like 128

i use vllm v0.20.1 with --chat-template chat_template-v9.jinja (vllm support --chat-template) with <|think_off|>，but found model answer in reasoning and content is null like this:

im not sure its template bug or vllm bug, i paste my request here, can someone test it?thx

POST /v1/chat/completions HTTP/1.1

{
  "model": "Qwen3.6-35B-A3B",
  "temperature": 1,
  "messages": [
    {
      "role": "system",
      "content": "<|think_off|>You are a professional Simplified Chinese Language native translator who needs to fluently translate text into Simplified Chinese Language.\n\n## Translation Rules\n1. Output only the translated content, without explanations or additional content (such as \"Here's the translation:\" or \"Translation as follows:\")\n2. The returned translation must maintain exactly the same number of paragraphs and format as the original text\n3. If the text contains HTML tags, consider where the tags should be placed in the translation while maintaining fluency\n4. For content that should not be translated (such as proper nouns, code, etc.), keep the original text.\n5. If input contains %%, use %% in your output, if input has no %%, don't use %% in your output\n\n## Context Awareness\nDocument Metadata:\nTitle: 《froggeric/Qwen-Fixed-Chat-Templates · Hugging Face》\n\n## OUTPUT FORMAT:\n- **Single paragraph input** → Output translation directly (no separators, no extra text)\n- **Multi-paragraph input** → Use %% as paragraph separator between translations\n\n## Examples\n### Multi-paragraph Input:\nParagraph A\n\n%%\n\nParagraph B\n\n%%\n\nParagraph C\n\n%%\n\nParagraph D\n\n### Multi-paragraph Output:\nTranslation A\n\n%%\n\nTranslation B\n\n%%\n\nTranslation C\n\n%%\n\nTranslation D\n\n### Single paragraph Input:\nSingle paragraph content\n\n### Single paragraph Output:\nDirect translation without separators\n\n{{imt_style_guide}}"
    },
    {
      "role": "user",
      "content": "翻译为简体中文：\n\nComparison — Qwen 3.6 template \n\n%%\n\nAuthorship \n\n%%\n\nOriginal models \n\n%%\n\nAlibaba Cloud (Qwen team) "
    }
  ]
}