llama.cpp
Support jinja extra template kwargs (Qwen3 enable_thinking feature), from command line and from client
#13196
Merged
Merged
Support jinja extra template kwargs (Qwen3 enable_thinking feature), from command line and from client #13196
CISC
merged
16 commits into
ggml-org:master
from
matteoserva:enable_thinking
matteoserva
changed the title 
ochafik
commented
on 2025-05-16
ochafik
approved these changes
on 2025-06-03
initial commit for handling extra template kwargs
2950f624
enable_thinking and assistant prefill cannot be enabled at the same time
46064b49
can set chat_template_kwargs in command line
91681d45
added doc
a92e790b
fixed formatting
abda1aed
add support for extra context in generic template init
570018b1
coding standard: common/chat.cpp
8c8b2903
coding standard: common/chat.cpp
56b3a691
Apply suggestions from code review
fe6e44ad
fix merge conflict
67789ef0
chat.cpp: simplify calls to apply to ensure systematic propagation of…
9a938632
normalize environment variable name
74f6060c
simplify code
cdc3cbe0
prefill cannot be used with thinking models
226e37d8
compatibility with the new reasoning-budget parameter
4e1c329d
exxocism
commented
on 2025-06-08
fix prefill for non thinking models
a056e536
exxocism
commented
on 2025-06-14
Login to write a write a comment.
Login via GitHub