Hugging Face's logo

froggeric
/
Qwen-Fixed-Chat-Templates

MLX
jinja
chat-template
qwen
qwen3.5
qwen3.6
qwen3.8
lm-studio
llama.cpp
vllm
tool-calling
thinking
Model card Files
xet
 Community
101

This doesn't seem to work with this Qwen 3.6 model...

#67
by BTGbullseye - opened
Discussion
BTGbullseye

I'm using https://huggingface.co/DavidAU/Qwen3.6-27B-Fable-Fusion-711-Uncensored-Heretic-NM-DAU-NEO-MAX-MTP-GGUF for creative writing, and this template just causes the model to not load at all. I'm pretty new with LLMs, what might be causing this?
froggeric
 Owner

A chat template only formats prompt text during inference and cannot prevent a model from loading into memory. If the model fails to load at startup, it is usually due to one of two things:

  1. The path passed to --chat-template-file is incorrect or unreadable.
  2. The specific GGUF you downloaded is an MTP (Multi-Token Prediction) variant. MTP models require recent llama.cpp builds and specific draft speculation flags (--spec-type draft-mtp). If your build does not support MTP, load the standard non-MTP GGUF instead.

For creative writing where tool calling is not involved, loading the standard GGUF with --jinja --chat-template-file chat_template.jinja will work cleanly.

froggeric changed discussion status to closed

Sign up or log in to comment