[BugFix] Add async communication check for capturing mode (#8149)

### What this PR does / why we need it?
Introduce a check to not using asynchronous communication under
`enable_dsa_cp_with_layer_shard` branch on capturing mode. This change
prevents potential stream and event issues when operating in
graph/capturing mode, ensuring safer communication practices.

### Does this PR introduce _any_ user-facing change?
No.

### How was this patch tested?
E2E test with dsv32 + FC1 + FULL_DECODE_ONLY +
kv_transfer_config(kv_both)

---------

Signed-off-by: chenchuw886 <chenchuw@huawei.com>
Co-authored-by: chenchuw886 <chenchuw@huawei.com>

vllm_ascend/platform.py

@@ -233,6 +233,20 @@ class NPUPlatform(Platform):
     def set_device(cls, device: torch.device):
         torch.npu.set_device(device)
 
+    @classmethod
+    def _validate_layer_sharding_config(cls, vllm_config: VllmConfig) -> None:
+        additional_config = vllm_config.additional_config or {}
+        layer_sharding = additional_config.get("layer_sharding") or []
+        if not layer_sharding:
+            return
+
+        kv_transfer_config = vllm_config.kv_transfer_config
+        if kv_transfer_config is not None and kv_transfer_config.kv_role != "kv_producer":
+            raise ValueError(
+                "additional_config.layer_sharding is only supported on P nodes "
+                "(kv_role='kv_producer') when KV transfer is enabled."
+            )
+
     @classmethod
     def check_and_update_config(cls, vllm_config: VllmConfig) -> None:
         from vllm_ascend.quantization.utils import maybe_auto_detect_quantization
@@ -240,6 +254,8 @@ class NPUPlatform(Platform):
         if vllm_config.model_config is not None:
             maybe_auto_detect_quantization(vllm_config)
 
+        cls._validate_layer_sharding_config(vllm_config)
+
         # initialize ascend config from vllm additional_config
         cls._fix_incompatible_config(vllm_config)
 

vllm_ascend/utils.py

@@ -1238,7 +1238,9 @@ def enable_dsa_cp_with_layer_shard() -> bool:
     vllm_config = get_current_vllm_config()
     # because the broadcast in layer sharding needs to be overlapped with a heavy compute stream to be
     # effectively hidden, it is enabled only during the prefill stage.
-    is_prefill_instance = vllm_config.kv_transfer_config is not None and vllm_config.kv_transfer_config.is_kv_producer
+    is_prefill_instance = (
+        vllm_config.kv_transfer_config is not None and vllm_config.kv_transfer_config.kv_role == "kv_producer"
+    )
     return is_prefill_instance