From bfbf8def19867d7176384f356dc8f470c9d956dc Mon Sep 17 00:00:00 2001
From: Kangyan Zhou <zky314343421@gmail.com>
Date: Fri, 21 Nov 2025 16:17:09 -0800
Subject: [PATCH] Fix B200 Nightly tests and move one manual test back to unit
 test to prevent the same issue

---
 python/sglang/srt/layers/moe/fused_moe_triton/layer.py | 4 +++-
 test/srt/run_suite.py                                  | 1 +
 test/{manual => srt}/test_triton_fused_moe.py          | 2 +-
 3 files changed, 5 insertions(+), 2 deletions(-)
 rename test/{manual => srt}/test_triton_fused_moe.py (98%)

diff --git a/python/sglang/srt/layers/moe/fused_moe_triton/layer.py b/python/sglang/srt/layers/moe/fused_moe_triton/layer.py
index 47a9da781b7e..3b3b8e7b716b 100644
--- a/python/sglang/srt/layers/moe/fused_moe_triton/layer.py
+++ b/python/sglang/srt/layers/moe/fused_moe_triton/layer.py
@@ -1034,7 +1034,9 @@ def forward(self, hidden_states: torch.Tensor, topk_output: TopKOutput):
         final_hidden_states = self.quant_method.apply_with_router_logits(
             layer=self,
             dispatch_output=StandardDispatchOutput(
-                hidden_states=hidden_states, topk_output=topk_output
+                hidden_states=hidden_states,
+                hidden_states_scale=None,
+                topk_output=topk_output,
             ),
         )
 
diff --git a/test/srt/run_suite.py b/test/srt/run_suite.py
index 48510c959eef..863192c00495 100644
--- a/test/srt/run_suite.py
+++ b/test/srt/run_suite.py
@@ -116,6 +116,7 @@
         TestFile("test_swa_unittest.py", 1),
         TestFile("test_torch_compile.py", 76),
         TestFile("test_torch_compile_moe.py", 210),
+        TestFile("test_triton_fused_moe.py", 80),
         TestFile("test_torch_native_attention_backend.py", 123),
         TestFile("test_torchao.py", 70),
         TestFile("test_triton_attention_kernels.py", 4),
diff --git a/test/manual/test_triton_fused_moe.py b/test/srt/test_triton_fused_moe.py
similarity index 98%
rename from test/manual/test_triton_fused_moe.py
rename to test/srt/test_triton_fused_moe.py
index d989494c95d0..07e774a9b0cc 100644
--- a/test/manual/test_triton_fused_moe.py
+++ b/test/srt/test_triton_fused_moe.py
@@ -115,7 +115,7 @@ def _test_case(self, m, n, k, e, topk, dtype):
         quant_info = TritonKernelsQuantInfo(w13_weight=w1_tri, w2_weight=w2_tri)
 
         dispatch_output = StandardDispatchOutput(
-            hidden_states=a, topk_output=triton_topk_output
+            hidden_states=a, hidden_states_scale=None, topk_output=triton_topk_output
         )
 
         torch_per_expert = self.torch_naive_moe(