EngineX-Hygon/sglang
5
0
Fork 0
Code Issues Pull Requests Actions 7 Projects Releases Wiki Activity

Feat: Support audio in Phi4-mm model (#8048)

Browse Source
This commit is contained in:
Binyao Jiang
2025-07-18 21:03:53 -07:00
committed by GitHub
parent d918ab7985
commit b7e951a6db
11 changed files with 3333 additions and 54 deletions
Download Patch File Download Diff File Expand all files Collapse all files

22
test/srt/test_vision_openai_server_common.py
View File

@@ -47,6 +47,12 @@ class TestOpenAIVisionServer(CustomTestCase):
def tearDownClass(cls):
kill_process_tree(cls.process.pid)
def get_audio_request_kwargs(self):
return self.get_request_kwargs()
def get_vision_request_kwargs(self):
return self.get_request_kwargs()
def get_request_kwargs(self):
return {}
@@ -71,7 +77,7 @@ class TestOpenAIVisionServer(CustomTestCase):
},
],
temperature=0,
**(self.get_request_kwargs()),
**(self.get_vision_request_kwargs()),
)
assert response.choices[0].message.role == "assistant"
@@ -134,7 +140,7 @@ class TestOpenAIVisionServer(CustomTestCase):
},
],
temperature=0,
**(self.get_request_kwargs()),
**(self.get_vision_request_kwargs()),
)
assert response.choices[0].message.role == "assistant"
@@ -177,7 +183,7 @@ class TestOpenAIVisionServer(CustomTestCase):
},
],
temperature=0,
**(self.get_request_kwargs()),
**(self.get_vision_request_kwargs()),
)
assert response.choices[0].message.role == "assistant"
@@ -333,7 +339,7 @@ class TestOpenAIVisionServer(CustomTestCase):
temperature=0,
max_tokens=1024,
stream=False,
**(self.get_request_kwargs()),
**(self.get_vision_request_kwargs()),
)
video_response = response.choices[0].message.content
@@ -376,7 +382,7 @@ class TestOpenAIVisionServer(CustomTestCase):
+ r"""\}"""
)
extra_kwargs = self.get_request_kwargs()
extra_kwargs = self.get_vision_request_kwargs()
extra_kwargs.setdefault("extra_body", {})["regex"] = regex
response = client.chat.completions.create(
@@ -443,7 +449,7 @@ class TestOpenAIVisionServer(CustomTestCase):
{"role": "user", "content": content},
],
temperature=0,
**(self.get_request_kwargs()),
**(self.get_vision_request_kwargs()),
)
assert response.choices[0].message.role == "assistant"
@@ -486,7 +492,7 @@ class TestOpenAIVisionServer(CustomTestCase):
temperature=0,
max_tokens=128,
stream=False,
**(self.get_request_kwargs()),
**(self.get_audio_request_kwargs()),
)
audio_response = response.choices[0].message.content
@@ -500,7 +506,7 @@ class TestOpenAIVisionServer(CustomTestCase):
self.assertIsNotNone(audio_response)
self.assertGreater(len(audio_response), 0)
return audio_response
return audio_response.lower()
def _test_audio_speech_completion(self):
# a fragment of Trump's speech