Raise error when using KV sharing with KV connector

Signed-off-by: Yong Hoon Shin <[email protected]>

Author: sarckk

vllm/attention/layer.py

@@ -58,12 +58,19 @@ def __init__(
         """
         super().__init__()
 
-        if not envs.VLLM_USE_V1:
-            assert kv_sharing_target_layer_name is None, NotImplementedError(
-                "KV sharing is not supported in V0.")
-
-        # Verify target layer is valid for cross-layer KV sharing
         if kv_sharing_target_layer_name is not None:
+            if not envs.VLLM_USE_V1:
+                raise NotImplementedError(
+                    "Cross-layer KV sharing is not supported in V0.")
+
+            if has_kv_transfer_group():
+                # TODO(sarckk): look into gaps for making cross-layer
+                # KV sharing work with KV connectors
+                raise NotImplementedError(
+                    "Cross-layer KV sharing is not yet compatible "
+                    "with KV connector.")
+
+            # Verify target layer is valid for cross-layer KV sharing
             from vllm.model_executor.models.utils import extract_layer_index
             current_layer_idx = extract_layer_index(prefix)
             target_layer_idx = extract_layer_index(