ci: add moe test (#1053)

test/srt/test_serving_throughput.py

@@ -5,11 +5,14 @@ from types import SimpleNamespace
 from sglang.bench_serving import run_benchmark
 from sglang.srt.server_args import ServerArgs
 from sglang.srt.utils import kill_child_process
-from sglang.test.test_utils import DEFAULT_MODEL_NAME_FOR_TEST, popen_launch_server
+from sglang.test.test_utils import (
+    DEFAULT_MODEL_NAME_FOR_TEST,
+    DEFAULT_URL_FOR_E2E_TEST,
+    popen_launch_server,
+)
 
 
 class TestServingThroughput(unittest.TestCase):
-
     def run_test(self, disable_radix_cache, disable_flashinfer, chunked_prefill_size):
         # Launch the server
         other_args = []
@@ -20,7 +23,7 @@ class TestServingThroughput(unittest.TestCase):
         other_args.extend(["--chunked-prefill-size", str(chunked_prefill_size)])
 
         model = DEFAULT_MODEL_NAME_FOR_TEST
-        base_url = "http://127.0.0.1:9157"
+        base_url = DEFAULT_URL_FOR_E2E_TEST
         process = popen_launch_server(
             model, base_url, timeout=300, other_args=other_args
         )