A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 14, 05:30 UTC
[Bugfix] Fix MoE BNB version (#22260)
Signed-off-by: Jee Jee Li <pandaleefree@gmail.com>
Jee Jee Li committed
Aug 6, 2025 at 10:56 UTC
8e6c7e873f1a2830ab096d69ee1812b323aef650
1 file changed
+4
-4
vllm/model_executor/layers/quantization/bitsandbytes.py
+4
-4
@@ -412,12 +412,12 @@ class BitsAndBytesMoEMethod(FusedMoEMethodBase):
412
def __init__(self, quant_config: BitsAndBytesConfig):
413
try:
414
import bitsandbytes
415
- if bitsandbytes.__version__ < "0.45.3":
415
+ if bitsandbytes.__version__ < "0.46.1":
416
raise ImportError("bitsandbytes version is wrong. Please "
417
- "install bitsandbytes>=0.45.3.")
417
+ "install bitsandbytes>=0.46.1.")
418
except ImportError as err:
419
- raise ImportError("Please install bitsandbytes>=0.45.3 via "
420
- "`pip install bitsandbytes>=0.45.3` to use "
419
+ raise ImportError("Please install bitsandbytes>=0.46.1 via "
420
+ "`pip install bitsandbytes>=0.46.1` to use "
421
"bitsandbytes quantizer.") from err
422
self.topk_indices_dtype = None
423
self.quant_config = quant_config