111 lines
2.9 KiB
Python
111 lines
2.9 KiB
Python
"""
|
|
Usage:
|
|
python -m unittest test_moe_eval_accuracy_large.TestMoEEvalAccuracyLarge.test_mmlu
|
|
"""
|
|
|
|
import os
|
|
import unittest
|
|
from types import SimpleNamespace
|
|
|
|
from sglang.srt.utils import kill_process_tree
|
|
from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
|
|
from sglang.test.run_eval import run_eval
|
|
from sglang.test.test_utils import (
|
|
DEFAULT_MOE_MODEL_NAME_FOR_TEST,
|
|
DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
|
|
DEFAULT_URL_FOR_TEST,
|
|
CustomTestCase,
|
|
is_in_amd_ci,
|
|
is_in_ci,
|
|
popen_launch_server,
|
|
write_github_step_summary,
|
|
)
|
|
|
|
register_cuda_ci(est_time=500, suite="stage-b-test-large-2-gpu")
|
|
register_amd_ci(est_time=500, suite="stage-b-test-large-2-gpu-amd")
|
|
|
|
|
|
class TestMoEEvalAccuracyLarge(CustomTestCase):
|
|
@classmethod
|
|
def setUpClass(cls):
|
|
cls.model = DEFAULT_MOE_MODEL_NAME_FOR_TEST
|
|
cls.base_url = DEFAULT_URL_FOR_TEST
|
|
|
|
# Disable AITER for AMD CI to ensure consistent results
|
|
env = None
|
|
if is_in_amd_ci():
|
|
env = os.environ.copy()
|
|
env["SGLANG_USE_AITER"] = "0"
|
|
env["SGLANG_USE_AITER_AR"] = "0"
|
|
env["HF_HUB_ENABLE_HF_TRANSFER"] = "0"
|
|
|
|
cls.process = popen_launch_server(
|
|
cls.model,
|
|
cls.base_url,
|
|
timeout=DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
|
|
other_args=[
|
|
"--log-level-http",
|
|
"warning",
|
|
"--tp",
|
|
"2",
|
|
],
|
|
env=env,
|
|
)
|
|
|
|
@classmethod
|
|
def tearDownClass(cls):
|
|
kill_process_tree(cls.process.pid)
|
|
|
|
def test_mmlu(self):
|
|
args = SimpleNamespace(
|
|
base_url=self.base_url,
|
|
model=self.model,
|
|
eval_name="mmlu",
|
|
num_examples=5000,
|
|
num_threads=1024,
|
|
)
|
|
|
|
metrics = run_eval(args)
|
|
self.assertGreater(metrics["score"], 0.62)
|
|
|
|
if is_in_ci():
|
|
write_github_step_summary(f"### test_mmlu\n" f'{metrics["score"]=:.4f}\n')
|
|
|
|
def test_human_eval(self):
|
|
args = SimpleNamespace(
|
|
base_url=self.base_url,
|
|
model=self.model,
|
|
eval_name="humaneval",
|
|
num_examples=None,
|
|
num_threads=1024,
|
|
)
|
|
|
|
metrics = run_eval(args)
|
|
self.assertGreater(metrics["score"], 0.40)
|
|
|
|
if is_in_ci():
|
|
write_github_step_summary(
|
|
f"### test_human_eval\n" f'{metrics["score"]=:.4f}\n'
|
|
)
|
|
|
|
def test_mgsm_en(self):
|
|
args = SimpleNamespace(
|
|
base_url=self.base_url,
|
|
model=self.model,
|
|
eval_name="mgsm_en",
|
|
num_examples=None,
|
|
num_threads=1024,
|
|
)
|
|
|
|
metrics = run_eval(args)
|
|
self.assertGreater(metrics["score"], 0.61)
|
|
|
|
if is_in_ci():
|
|
write_github_step_summary(
|
|
f"### test_mgsm_en\n" f'{metrics["score"]=:.4f}\n'
|
|
)
|
|
|
|
|
|
if __name__ == "__main__":
|
|
unittest.main()
|