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

Merged three native APIs into one: get_server_info (#2152)

Browse Source
This commit is contained in:
Henry Hyeonmok Ko
2024-11-24 01:37:58 -08:00
committed by GitHub
parent 84a1698d67
commit dbe1729395
10 changed files with 81 additions and 126 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
test/srt/test_data_parallelism.py
View File

@@ -63,12 +63,13 @@ class TestDataParallelism(unittest.TestCase):
assert response.status_code == 200
def test_get_memory_pool_size(self):
response = requests.get(self.base_url + "/get_memory_pool_size")
# use `get_server_info` instead since `get_memory_pool_size` is merged into `get_server_info`
response = requests.get(self.base_url + "/get_server_info")
assert response.status_code == 200
time.sleep(5)
response = requests.get(self.base_url + "/get_memory_pool_size")
response = requests.get(self.base_url + "/get_server_info")
assert response.status_code == 200

15
test/srt/test_srt_endpoint.py
View File

@@ -154,9 +154,18 @@ class TestSRTEndpoint(unittest.TestCase):
self.assertEqual(res["meta_info"]["completion_tokens"], new_tokens)
self.assertEqual(len(res["meta_info"]["output_token_logprobs"]), new_tokens)
def test_get_memory_pool_size(self):
response = requests.post(self.base_url + "/get_memory_pool_size")
self.assertIsInstance(response.json(), int)
def test_get_server_info(self):
response = requests.get(self.base_url + "/get_server_info")
response_json = response.json()
max_total_num_tokens = response_json["max_total_num_tokens"]
self.assertIsInstance(max_total_num_tokens, int)
memory_pool_size = response_json["memory_pool_size"]
self.assertIsInstance(memory_pool_size, int)
attention_backend = response_json["attention_backend"]
self.assertIsInstance(attention_backend, str)
if __name__ == "__main__":