[Fix][CPU] fix xeon ci failure by test_qwen35_flashinfer_fusion (#37338)

test/registered/unit/layers/moe/test_qwen35_flashinfer_fusion.py

@@ -1,5 +1,6 @@
 from dataclasses import dataclass
 from types import SimpleNamespace
+from unittest.mock import patch
 
 import pytest
 import torch
@@ -17,7 +18,7 @@ from sglang.srt.model_executor.forward_batch_info import ForwardMode
 from sglang.srt.models.qwen3_5_text import Qwen3_5ForCausalLM
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=20, suite="base-c-test-cpu")
+register_cpu_ci(est_time=20, suite="base-a-test-cpu")
 
 
 @dataclass(frozen=True)
@@ -59,15 +60,16 @@ def test_supported_forward_modes(forward_mode, expected):
     assert is_supported_forward_mode(forward_mode) is expected
 
 
-def test_framework_capacity_is_maximum_of_all_sources():
+@patch(
+    "sglang.srt.layers.moe.qwen35_flashinfer_fusion.cutedsl_moe_max_num_tokens",
+    return_value=8192,
+)
+def test_framework_capacity_is_maximum_of_all_sources(_cutedsl_moe_max_num_tokens):
     graph = SimpleNamespace(
         decode=SimpleNamespace(max_bs=512, bs=[1, 64, 256]),
         prefill=SimpleNamespace(max_bs=4096, bs=[1024, 2048, 4096]),
     )
-    server_args = SimpleNamespace(
-        cuda_graph_config=graph,
-        cutedsl_moe_max_num_tokens=lambda: 8192,
-    )
+    server_args = SimpleNamespace(cuda_graph_config=graph)
     runner = SimpleNamespace(server_args=server_args, max_running_requests=2048)
 
     assert resolve_max_m(runner) == 8192