Qwen
/

Text Generation
Transformers
Safetensors
qwen3_moe
conversational

chat_template issue

#40
by Alex-Chan - opened

Hi, thank you for your excellent work.
When I'm using Dynamic Cache in iterative generation, Qwen3 chat template will cause a bug in the transformers code, which is because it delete contents between <think> and </think>. More information can be found in https://github.com/huggingface/transformers/issues/38727.
I wonder why chat template defualtly delete thinking content, and should it raise confusion between codes? Would it be better if not deleting?
Besides, I'm from ali-inc too huhuh.

Sign up or log in to comment