Fix gemma3n with hybrid swa (#8240)

Signed-off-by: Xinyuan Tong <xinyuantong.cs@gmail.com>
Co-authored-by: Xinyuan Tong <xinyuantong.cs@gmail.com>
This commit is contained in:
Xinyuan Tong
2025-07-23 13:29:18 -07:00
committed by GitHub
parent 70251e935e
commit 38000a5f44
3 changed files with 28 additions and 3 deletions

View File

@@ -105,7 +105,7 @@ suites = {
TestFile("test_vision_chunked_prefill.py", 175),
TestFile("test_vlm_input_format.py", 300),
TestFile("test_vision_openai_server_a.py", 584),
TestFile("test_vision_openai_server_b.py", 556),
TestFile("test_vision_openai_server_b.py", 620),
TestFile("test_w8a8_quantization.py", 46),
TestFile("test_reasoning_parser.py", 5),
],

View File

@@ -151,6 +151,27 @@ class TestGemma3itServer(TestOpenAIVisionServer):
pass
class TestGemma3nServer(TestOpenAIVisionServer):
@classmethod
def setUpClass(cls):
cls.model = "google/gemma-3n-E2B-it"
cls.base_url = DEFAULT_URL_FOR_TEST
cls.api_key = "sk-123456"
cls.process = popen_launch_server(
cls.model,
cls.base_url,
timeout=DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
other_args=[
"--trust-remote-code",
"--mem-fraction-static",
"0.70",
"--cuda-graph-max-bs",
"1",
],
)
cls.base_url += "/v1"
class TestKimiVLServer(TestOpenAIVisionServer):
@classmethod
def setUpClass(cls):