[Misc] Upgrade BNB version (#15183)

Author: jeejeelee

Diff for: Dockerfile

@@ -286,7 +286,7 @@ RUN --mount=type=cache,target=/root/.cache/uv \
     if [ "$TARGETPLATFORM" = "linux/arm64" ]; then \
         uv pip install accelerate hf_transfer 'modelscope!=1.15.0' 'bitsandbytes>=0.42.0' 'timm==0.9.10' boto3 runai-model-streamer runai-model-streamer[s3]; \
     else \
-        uv pip install accelerate hf_transfer 'modelscope!=1.15.0' 'bitsandbytes>=0.45.0' 'timm==0.9.10' boto3 runai-model-streamer runai-model-streamer[s3]; \
+        uv pip install accelerate hf_transfer 'modelscope!=1.15.0' 'bitsandbytes>=0.45.3' 'timm==0.9.10' boto3 runai-model-streamer runai-model-streamer[s3]; \
     fi
 
 ENV VLLM_USAGE_SOURCE production-docker-image

Diff for: docs/source/features/quantization/bnb.md

@@ -9,7 +9,7 @@ Compared to other quantization methods, BitsAndBytes eliminates the need for cal
 Below are the steps to utilize BitsAndBytes with vLLM.
 
 ```console
-pip install bitsandbytes>=0.45.0
+pip install bitsandbytes>=0.45.3
 ```
 
 vLLM reads the model's config file and supports both in-flight quantization and pre-quantized checkpoint.

Diff for: vllm/model_executor/layers/quantization/bitsandbytes.py

@@ -155,12 +155,12 @@ class BitsAndBytesLinearMethod(LinearMethodBase):
     def __init__(self, quant_config: BitsAndBytesConfig):
         try:
             import bitsandbytes
-            if bitsandbytes.__version__ < "0.45.0":
+            if bitsandbytes.__version__ < "0.45.3":
                 raise ImportError("bitsandbytes version is wrong. Please "
-                                  "install bitsandbytes>=0.45.0.")
+                                  "install bitsandbytes>=0.45.3.")
         except ImportError as err:
-            raise ImportError("Please install bitsandbytes>=0.45.0 via "
-                              "`pip install bitsandbytes>=0.45.0` to use "
+            raise ImportError("Please install bitsandbytes>=0.45.3 via "
+                              "`pip install bitsandbytes>=0.45.3` to use "
                               "bitsandbytes quantizer.") from err
 
         self.quant_config = quant_config

Diff for: vllm/model_executor/model_loader/loader.py

@@ -862,12 +862,12 @@ def _get_quantized_weights_iterator(
         try:
             import bitsandbytes
 
-            if bitsandbytes.__version__ < "0.45.0":
+            if bitsandbytes.__version__ < "0.45.3":
                 raise ImportError("bitsandbytes version is wrong. Please "
-                                  "install bitsandbytes>=0.45.0.")
+                                  "install bitsandbytes>=0.45.3.")
         except ImportError as err:
-            raise ImportError("Please install bitsandbytes>=0.45.0 via "
-                              "`pip install bitsandbytes>=0.45.0` to use "
+            raise ImportError("Please install bitsandbytes>=0.45.3 via "
+                              "`pip install bitsandbytes>=0.45.3` to use "
                               "bitsandbytes quantizer.") from err
 
         hf_weights_files, use_safetensors = self._prepare_weights(