From 93e9db5eb89d51e6818d43062d0296612bf53061 Mon Sep 17 00:00:00 2001
From: zijiec <zijie.chen@amd.com>
Date: Tue, 11 Aug 2026 21:50:05 -0400
Subject: [PATCH] [Fix][Qwen]: fused shared-expert detection PP-safe protection
 (#34447)

---
 python/sglang/srt/models/qwen3_5.py           | 12 ++--
 .../models/test_qwen3_5_pipeline_parallel.py  | 69 +++++++++++++++++++
 2 files changed, 75 insertions(+), 6 deletions(-)
 create mode 100644 test/registered/unit/models/test_qwen3_5_pipeline_parallel.py

diff --git a/python/sglang/srt/models/qwen3_5.py b/python/sglang/srt/models/qwen3_5.py
index 1a15ffc67..40d1810f9 100644
--- a/python/sglang/srt/models/qwen3_5.py
+++ b/python/sglang/srt/models/qwen3_5.py
@@ -1961,13 +1961,13 @@ class Qwen3_5MoeForConditionalGeneration(Qwen3VLForConditionalGeneration):
         return module_name.startswith("model.layers.")
 
     def _get_num_fused_shared_experts(self):
-        if not (
-            hasattr(self.model, "layers")
-            and len(self.model.layers) > 0
-            and hasattr(self.model.layers[0].mlp, "num_fused_shared_experts")
-        ):
+        if not hasattr(self.model, "layers"):
             return 0
-        return self.model.layers[0].mlp.num_fused_shared_experts
+        for layer_id in range(self.model.start_layer, self.model.end_layer):
+            mlp = getattr(self.model.layers[layer_id], "mlp", None)
+            if hasattr(mlp, "num_fused_shared_experts"):
+                return mlp.num_fused_shared_experts
+        return 0
 
     def get_embed_and_head(self):
         embed = self.model.embed_tokens.weight if self.pp_group.is_first_rank else None
diff --git a/test/registered/unit/models/test_qwen3_5_pipeline_parallel.py b/test/registered/unit/models/test_qwen3_5_pipeline_parallel.py
new file mode 100644
index 000000000..f239893a3
--- /dev/null
+++ b/test/registered/unit/models/test_qwen3_5_pipeline_parallel.py
@@ -0,0 +1,69 @@
+import unittest
+from types import SimpleNamespace
+
+from sglang.srt.layers.utils import PPMissingLayer
+from sglang.srt.models.qwen3_5 import Qwen3_5MoeForConditionalGeneration
+from sglang.test.ci.ci_register import register_cpu_ci
+from sglang.test.test_utils import CustomTestCase
+
+register_cpu_ci(est_time=2, suite="base-a-test-cpu")
+
+
+class TestQwen3_5PipelineParallel(CustomTestCase):
+    @staticmethod
+    def _get_num_fused_shared_experts(layers, start_layer, end_layer):
+        model = SimpleNamespace(
+            model=SimpleNamespace(
+                layers=layers,
+                start_layer=start_layer,
+                end_layer=end_layer,
+            )
+        )
+        return Qwen3_5MoeForConditionalGeneration._get_num_fused_shared_experts(model)
+
+    def test_get_num_fused_shared_experts_returns_zero_without_layers(self):
+        model = SimpleNamespace(model=SimpleNamespace())
+
+        num_fused_shared_experts = (
+            Qwen3_5MoeForConditionalGeneration._get_num_fused_shared_experts(model)
+        )
+
+        self.assertEqual(num_fused_shared_experts, 0)
+
+    def test_get_num_fused_shared_experts_uses_local_pp_layers(self):
+        layers = [
+            PPMissingLayer(),
+            PPMissingLayer(),
+            SimpleNamespace(
+                mlp=SimpleNamespace(num_fused_shared_experts=1),
+            ),
+            SimpleNamespace(
+                mlp=SimpleNamespace(num_fused_shared_experts=1),
+            ),
+        ]
+
+        num_fused_shared_experts = self._get_num_fused_shared_experts(
+            layers,
+            start_layer=2,
+            end_layer=4,
+        )
+
+        self.assertEqual(num_fused_shared_experts, 1)
+
+    def test_get_num_fused_shared_experts_returns_zero_without_local_fusion(self):
+        layers = [
+            PPMissingLayer(),
+            SimpleNamespace(mlp=SimpleNamespace()),
+        ]
+
+        num_fused_shared_experts = self._get_num_fused_shared_experts(
+            layers,
+            start_layer=1,
+            end_layer=2,
+        )
+
+        self.assertEqual(num_fused_shared_experts, 0)
+
+
+if __name__ == "__main__":
+    unittest.main()