llama.cpp thinking mode support

#10
by bpool - opened

How to enable & disable thinking mode in llama.cpp?

I can't seem to find any relevant documentation.

You can try to brute force it by adding this to your jinja template, but all it will do is output its thought process as a response and do nothing else.
{%- if not enable_thinking is defined %}
{%- set enable_thinking = true %}
{%- endif %}

Sign up or log in to comment