minke.yu/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 12 Packages Projects Releases Wiki Activity

fix: restore VLM nightly regression coverage (#34662)

Browse Source
This commit is contained in:
Mick
2026-08-12 21:47:16 -07:00
committed by GitHub
parent 1286a50bb9
commit 69bf601e3c
4 changed files with 49 additions and 74 deletions
Download Patch File Download Diff File Expand all files Collapse all files
test/registered/unit/server_args/test_server_args.py
+7 -9
View File
@@ -262,19 +262,17 @@ class TestMultimodalFeatureTransport(CustomTestCase):
self.assertFalse(envs.SGLANG_USE_CUDA_IPC_TRANSPORT.get())
@patch("sglang.srt.server_args.is_cuda", return_value=True)
def test_default_transport_is_cuda_ipc_for_multimodal_model(self, _mock_is_cuda):
def test_default_transport_is_cpu_for_multimodal_model(self, _mock_is_cuda):
server_args = ServerArgs(model_path="dummy")
self._set_model_type(server_args, is_multimodal=True)
with patch.dict(os.environ, {}, clear=False):
envs.SGLANG_USE_CUDA_IPC_TRANSPORT.clear()
with self.assertLogs(server_args_module.logger, level="INFO") as logs:
with self.assertNoLogs(server_args_module.logger, level="INFO"):
server_args._handle_multimodal_feature_transport()
self.assertEqual(server_args.mm_feature_transport, "cuda_ipc")
self.assertTrue(envs.SGLANG_USE_CUDA_IPC_TRANSPORT.get())
self.assertIn("auto-resolved to cuda_ipc", "\n".join(logs.output))
self.assertEqual(server_args.mm_feature_transport, "cpu")
self.assertFalse(envs.SGLANG_USE_CUDA_IPC_TRANSPORT.get())
@patch("sglang.srt.server_args.os.path.exists", return_value=True)
@patch("sglang.srt.server_args.is_mnnvl_fabric_device", return_value=True)
@@ -362,7 +360,7 @@ class TestMultimodalFeatureTransport(CustomTestCase):
self.assertFalse(envs.SGLANG_USE_CUDA_IPC_TRANSPORT.get())
@patch("sglang.srt.server_args.is_cuda", return_value=True)
def test_default_transport_is_cuda_ipc_for_language_only_model(self, _mock_is_cuda):
def test_default_transport_is_cpu_for_language_only_model(self, _mock_is_cuda):
server_args = ServerArgs(model_path="dummy", language_only=True)
self._set_model_type(server_args, is_multimodal=True)
@@ -370,8 +368,8 @@ class TestMultimodalFeatureTransport(CustomTestCase):
envs.SGLANG_USE_CUDA_IPC_TRANSPORT.clear()
server_args._handle_multimodal_feature_transport()
self.assertEqual(server_args.mm_feature_transport, "cuda_ipc")
self.assertTrue(envs.SGLANG_USE_CUDA_IPC_TRANSPORT.get())
self.assertEqual(server_args.mm_feature_transport, "cpu")
self.assertFalse(envs.SGLANG_USE_CUDA_IPC_TRANSPORT.get())
@patch("sglang.srt.server_args.is_cuda", return_value=False)
def test_cuda_ipc_rejects_non_nvidia_platforms(self, _mock_is_cuda):