[ci] fix broken tests (#10956)

Signed-off-by: youkaichao <youkaichao@gmail.com>

vllm/worker/model_runner.py

@@ -1782,6 +1782,9 @@ class ModelRunner(GPUModelRunnerBase[ModelInputForGPUWithSamplingMetadata]):
             kv_caches: vLLM's paged memory
         """
 
+        if self.vllm_config.kv_transfer_config is None:
+            return False
+
         prefill_meta = model_input.attn_metadata.prefill_metadata
 
         # check if the current run is profiling
@@ -1789,9 +1792,6 @@ class ModelRunner(GPUModelRunnerBase[ModelInputForGPUWithSamplingMetadata]):
         # check if the current run is prefill
         is_prefill_run = prefill_meta is not None
 
-        if self.vllm_config.kv_transfer_config is None:
-            return False
-
         return self.vllm_config.kv_transfer_config.is_kv_consumer and (
             not is_profile_run) and is_prefill_run
 
@@ -1807,6 +1807,9 @@ class ModelRunner(GPUModelRunnerBase[ModelInputForGPUWithSamplingMetadata]):
             kv_caches: vLLM's paged memory
         """
 
+        if self.vllm_config.kv_transfer_config is None:
+            return False
+
         prefill_meta = model_input.attn_metadata.prefill_metadata
 
         # check if the current run is profiling
@@ -1814,9 +1817,6 @@ class ModelRunner(GPUModelRunnerBase[ModelInputForGPUWithSamplingMetadata]):
         # check if the current run is prefill
         is_prefill_run = prefill_meta is not None
 
-        if self.vllm_config.kv_transfer_config is None:
-            return False
-
         return self.vllm_config.kv_transfer_config.is_kv_producer and (
             not is_profile_run) and is_prefill_run