add annotation

Author: NicholasTao

tests/e2e/multicard/test_torchair_graph_mode.py

@@ -223,6 +223,7 @@ def _qwen_torchair_test_fixture(
 def test_e2e_qwen3_moe_with_torchair():
 
     def stubbed_get_state(ep_size, with_prefill, is_deepseek_v3_r1):
+        # force ep_size=16 to use MC2
         return _get_fused_moe_state(16, with_prefill, is_deepseek_v3_r1)
 
     with patch('vllm_ascend.ascend_forward_context._get_fused_moe_state',