From 27547a10e6ad36760650feedf4043f086922c0d0 Mon Sep 17 00:00:00 2001
From: Shanshan Shen <467638484@qq.com>
Date: Thu, 6 Nov 2025 20:30:40 +0800
Subject: [PATCH] [MM][Bugfix] Add MoE verification for multi-modal models
 (#3897) (#4027)

### What this PR does / why we need it?

Fix #3891.

The empty of `moe_comm_method` in the above issue is due to the wrong
check for MoE models. To be specific, the method `is_moe_model` only
checks whether a text-only model is a MoE model, without considering
multi-modal models, e.g., `VL` and `Omni`.

Check the config dict recursively to find if it has a key contains
"expert", without checking the model architecture.

It is worth noting that, we can't verify a model by if it contains
`FusedMoE` module because `is_moe_model` is called somewhere before the
model loading, e.g., it's called when updating the ACLGraph config in
platform initialization.

- vLLM version: v0.11.0
- vLLM main:
https://github.com/vllm-project/vllm/commit/83f478bb19489b41e9d208b47b4bb5a95ac171ac

---------

Signed-off-by: shen-shanshan <467638484@qq.com>
---
 vllm_ascend/utils.py | 16 +++++++++++++---
 1 file changed, 13 insertions(+), 3 deletions(-)

diff --git a/vllm_ascend/utils.py b/vllm_ascend/utils.py
index 27020de..f65aaa0 100644
--- a/vllm_ascend/utils.py
+++ b/vllm_ascend/utils.py
@@ -624,14 +624,24 @@ def shared_expert_dp_enabled() -> bool:
 
 
 def is_moe_model(vllm_config: VllmConfig):
+    """Checks if the model is a MoE model by config"""
     global _IS_MOE_MODEL
     if _IS_MOE_MODEL is None:
-        config = vllm_config.model_config.hf_config
-        _IS_MOE_MODEL = any('experts' in key.lower()
-                            for key in config.to_dict())
+        model_configs = vllm_config.model_config.hf_config.to_dict()
+        _IS_MOE_MODEL = _is_contain_expert(model_configs)
     return _IS_MOE_MODEL
 
 
+def _is_contain_expert(config: Any):
+    if isinstance(config, dict):
+        for k, v in config.items():
+            if "expert" in str(k):
+                return True
+            if _is_contain_expert(v):
+                return True
+    return False
+
+
 def weak_ref_tensor(tensor: Any) -> Any:
     """
     Create a weak reference to a tensor.