File size: 3,019 Bytes
a402b9b | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 | import unittest
import openai
from sglang.srt.utils import kill_process_tree
from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
from sglang.test.test_utils import (
DEFAULT_SMALL_MODEL_NAME_FOR_TEST,
DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
DEFAULT_URL_FOR_TEST,
CustomTestCase,
popen_launch_server,
)
register_cuda_ci(est_time=38, suite="stage-b-test-large-1-gpu")
register_amd_ci(est_time=31, suite="stage-b-test-small-1-gpu-amd")
class TestRequestLengthValidation(CustomTestCase):
@classmethod
def setUpClass(cls):
cls.base_url = DEFAULT_URL_FOR_TEST
cls.api_key = "sk-123456"
# Start server with auto truncate disabled
cls.process = popen_launch_server(
DEFAULT_SMALL_MODEL_NAME_FOR_TEST,
cls.base_url,
timeout=DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
api_key=cls.api_key,
other_args=("--max-total-tokens", "1000", "--context-length", "1000"),
)
@classmethod
def tearDownClass(cls):
kill_process_tree(cls.process.pid)
def test_input_length_longer_than_context_length(self):
client = openai.Client(api_key=self.api_key, base_url=f"{self.base_url}/v1")
long_text = "hello " * 1200 # Will tokenize to more than context length
with self.assertRaises(openai.BadRequestError) as cm:
client.chat.completions.create(
model=DEFAULT_SMALL_MODEL_NAME_FOR_TEST,
messages=[
{"role": "user", "content": long_text},
],
temperature=0,
)
self.assertIn("is longer than the model's context length", str(cm.exception))
def test_input_length_longer_than_maximum_allowed_length(self):
client = openai.Client(api_key=self.api_key, base_url=f"{self.base_url}/v1")
long_text = "hello " * 999 # the maximum allowed length is 994 tokens
with self.assertRaises(openai.BadRequestError) as cm:
client.chat.completions.create(
model=DEFAULT_SMALL_MODEL_NAME_FOR_TEST,
messages=[
{"role": "user", "content": long_text},
],
temperature=0,
)
self.assertIn("is longer than the model's context length", str(cm.exception))
def test_max_tokens_validation(self):
client = openai.Client(api_key=self.api_key, base_url=f"{self.base_url}/v1")
long_text = "hello "
with self.assertRaises(openai.BadRequestError) as cm:
client.chat.completions.create(
model=DEFAULT_SMALL_MODEL_NAME_FOR_TEST,
messages=[
{"role": "user", "content": long_text},
],
temperature=0,
max_tokens=1200,
)
self.assertIn(
"max_completion_tokens is too large",
str(cm.exception),
)
if __name__ == "__main__":
unittest.main()
|