[Frontend] Re-enable running MaxSim on GPU (#38620)

Signed-off-by: wang.yuqi <yuqi.wang@daocloud.io>

tests/entrypoints/pooling/scoring/test_late_interaction_online.py

@@ -26,13 +26,18 @@ TEXTS_2 = [
 ]
 
 
-@pytest.fixture(scope="module")
-def server():
+@pytest.fixture(scope="module", params=[True, False])
+def server(request):
     args = [
         "--max-model-len",
         str(MAX_MODEL_LEN),
     ]
 
+    # Test run pooling score MaxSim on worker side (GPU)
+    # aka flash-late-interaction
+    if not request.param:
+        args += ["--no-enable-flash-late-interaction"]
+
     with RemoteOpenAIServer(MODEL_NAME, args) as remote_server:
         yield remote_server