[nit]: Fix import for the lmcache integration (#27600)

Signed-off-by: Samuel Shen <slshen@uchicago.edu>
Co-authored-by: Samuel Shen <slshen@uchicago.edu>

vllm/distributed/kv_transfer/kv_connector/v1/lmcache_connector.py

@@ -13,9 +13,6 @@ from vllm.distributed.kv_transfer.kv_connector.v1.base import (
     KVConnectorMetadata,
     KVConnectorRole,
 )
-from vllm.distributed.kv_transfer.kv_connector.v1.lmcache_integration import (
-    vllm_v1_adapter as _adapter,
-)
 from vllm.logger import init_logger
 from vllm.v1.core.sched.output import SchedulerOutput
 
@@ -37,6 +34,11 @@ class LMCacheConnectorV1(KVConnectorBase_V1):
         )
         if use_native:
             logger.info("Initializing native LMCache connector")
+            # lazy import
+            from vllm.distributed.kv_transfer.kv_connector.v1 import lmcache_integration
+
+            _adapter = lmcache_integration.vllm_v1_adapter
+
             cls = _adapter.LMCacheConnectorV1Impl
         else:
             logger.info("Initializing latest dev LMCache connector")

vllm/distributed/kv_transfer/kv_connector/v1/lmcache_integration/__init__.py

@@ -1,2 +1,7 @@
 # SPDX-License-Identifier: Apache-2.0
 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+
+
+from . import vllm_v1_adapter
+
+__all__ = ["vllm_v1_adapter"]

vllm/distributed/kv_transfer/kv_connector/v1/lmcache_integration/utils.py

@@ -131,7 +131,7 @@ def create_lmcache_metadata(
     # First Party
     from lmcache.config import LMCacheEngineMetadata
 
-    from vllm.utils import get_kv_cache_torch_dtype
+    from vllm.utils.torch_utils import get_kv_cache_torch_dtype
 
     config = lmcache_get_or_create_config()
     # Support both vllm_config object and individual config parameters