[feature] enable NPU CI (#7935)

Co-authored-by: Even Zhou <14368888+iforgetmyname@users.noreply.github.com>

test/srt/test_ascend_attention_backend.py

@@ -20,22 +20,10 @@ from sglang.test.test_utils import (
     run_bench_offline_throughput,
 )
 
+DEFAULT_MODEL_NAME_FOR_TEST = "Qwen/Qwen2.5-7B-Instruct"
+
 
 class TestAscendAttnBackend(CustomTestCase):
-    def test_latency(self):
-        output_throughput = run_bench_offline_throughput(
-            DEFAULT_MODEL_NAME_FOR_TEST,
-            [
-                "--attention-backend",
-                "ascend",
-            ],
-        )
-
-        print(f"{output_throughput=}")
-
-        if is_in_ci():
-            self.assertGreater(output_throughput, 18)
-
     def test_gsm8k(self):
         model = DEFAULT_MODEL_NAME_FOR_TEST
         base_url = DEFAULT_URL_FOR_TEST