fix import

Signed-off-by: Ming Yang <minos.future@gmail.com>

Author: minosfuture

vllm/model_executor/offloader/__init__.py

@@ -9,8 +9,8 @@
     get_offloader,
     set_offloader,
 )
-from vllm.model_executor.offloader.offloader_v2 import OffloaderV2
 from vllm.model_executor.offloader.uva import UVAOffloader
+from vllm.model_executor.offloader.v2 import OffloaderV2
 
 __all__ = [
     "BaseOffloader",

vllm/model_executor/offloader/base.py

@@ -106,8 +106,8 @@ def create_offloader(cache_config: "CacheConfig") -> BaseOffloader:
 
     Priority: V2 offloading if configured, else UVA, else noop.
     """
-    from vllm.model_executor.offloader.offloader_v2 import OffloaderV2
     from vllm.model_executor.offloader.uva import UVAOffloader
+    from vllm.model_executor.offloader.v2 import OffloaderV2
 
     if cache_config.offload_group_size > 0:
         # Use V2 offloading