fakezeta

{# =========================
   Merged Qwen Multimodal Chat Template from
   - https://github.com/allanchan339/vLLM-Qwen3-3.5-3.6-chat-template-fix
   - https://huggingface.co/froggeric/Qwen-Fixed-Chat-Templates
   Features:
   - Long, strict tool system prompt (from allanchan339)
   - developer role supported (from froggeric)
   - <|think_on|> / <|think_off|> toggles (from froggeric)
   - Historical reasoning HIDDEN by default (from allanchan339)
   - String-form tool arguments parsed as JSON (from allanchan339)

fakezeta

  server/llama-server.exe 
    --port 9001 
    -ngl 40
    -t 6 
    -c 10000 
    -fa 
    --slots 
    --no-warmup  
    --jinja 
    --reasoning-format deepseek