improve the threshold and ports in tests (#1215)

This commit is contained in:
Mingyi
2024-08-25 19:02:08 -07:00
committed by GitHub
parent d3efcb3930
commit 158e8f1e2d
18 changed files with 122 additions and 86 deletions

View File

@@ -23,18 +23,14 @@ from sglang.utils import get_exception_traceback
DEFAULT_MODEL_NAME_FOR_TEST = "meta-llama/Meta-Llama-3.1-8B-Instruct"
DEFAULT_MOE_MODEL_NAME_FOR_TEST = "mistralai/Mixtral-8x7B-Instruct-v0.1"
DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH = 600
if os.getenv("SGLANG_IS_IN_CI", "false") == "true":
DEFAULT_PORT_FOR_SRT_TEST_RUNNER = 5157
DEFAULT_URL_FOR_MOE_TEST = "http://127.0.0.1:6157"
DEFAULT_URL_FOR_ACCURACY_TEST = "http://127.0.0.1:7157"
DEFAULT_URL_FOR_UNIT_TEST = "http://127.0.0.1:8157"
DEFAULT_URL_FOR_E2E_TEST = "http://127.0.0.1:9157"
DEFAULT_URL_FOR_TEST = "http://127.0.0.1:6157"
else:
DEFAULT_URL_FOR_MOE_TEST = "http://127.0.0.1:1157"
DEFAULT_URL_FOR_ACCURACY_TEST = "http://127.0.0.1:1257"
DEFAULT_URL_FOR_UNIT_TEST = "http://127.0.0.1:1357"
DEFAULT_URL_FOR_E2E_TEST = "http://127.0.0.1:1457"
DEFAULT_PORT_FOR_SRT_TEST_RUNNER = 1157
DEFAULT_URL_FOR_TEST = "http://127.0.0.1:2157"
def call_generate_lightllm(prompt, temperature, max_tokens, stop=None, url=None):