You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: llama_cpp/llama_chat_format.py
+6-17Lines changed: 6 additions & 17 deletions
Original file line number
Diff line number
Diff line change
@@ -4467,7 +4467,6 @@ def __init__(
4467
4467
self,
4468
4468
force_reasoning: bool=False,
4469
4469
add_vision_id: bool=True,
4470
-
image_min_tokens: int=-1,
4471
4470
**kwargs,
4472
4471
):
4473
4472
"""
@@ -4478,20 +4477,13 @@ def __init__(
4478
4477
- add_vision_id (bool):
4479
4478
- True (default): Count all the images. Recommended for multi-image.
4480
4479
- False: Doesn't count the images. Can save tokens with single-image.
4481
-
- image_min_tokens (int):
4482
-
It only takes effect when the value is greater than zero. the default value is -1 (i.e., using the default parameters in the model's preprocessor_config.json).
4483
-
Note: Qwen-VL models require at minimum 1024 image tokens to function correctly on bbox grounding tasks
0 commit comments