🤗 Transformers: the model-definition framework for state-of-the-art machine learning models in text, vision, audio, and multimodal models, for both inference and training.
Fix llama4 bnb mode (#44588)
* check float before using normal op Signed-off-by: jiqing-feng <[email protected]> * fix llama4 weight Signed-off-by: jiqing-feng <[email protected]> * add bnb quant skip module for llama4 Signed-off-by: jiqing-feng <[email protected]> * revert bnb integration Signed-off-by: jiqing-feng <[email protected]> * revert initialization.py Signed-off-by: jiqing-feng <[email protected]> * total revert init Signed-off-by: jiqing-feng <[email protected]> * fix _keep_in_fp32_modules Signed-off-by: jiqing-feng <[email protected]> * add _modules_to_not_quantize Signed-off-by: jiqing-feng <[email protected]> * fix modules_to_not_convert Signed-off-by: jiqing-feng <[email protected]> * update bnb quantize condition Signed-off-by: jiqing-feng <[email protected]> --------- Signed-off-by: jiqing-feng <[email protected]>
J
jiqing-feng committed
ce4a791c5277840c4c1d74eed03431b674869da5
Parent: cc4ef19
Committed by GitHub <[email protected]>
on 3/27/2026, 2:05:40 PM