ac453b253f
Co-authored-by: Alison Shao <alisonshao@mac.lan> Co-authored-by: Alison Shao <alisonshao@Mac.attlocal.net>
75 lines
2.4 KiB
Python
75 lines
2.4 KiB
Python
import unittest
|
|
|
|
from sglang.test.accuracy_test_runner import AccuracyTestParams
|
|
from sglang.test.ci.ci_register import register_cuda_ci
|
|
from sglang.test.performance_test_runner import PerformanceTestParams
|
|
from sglang.test.run_combined_tests import run_combined_tests
|
|
from sglang.test.test_utils import ModelLaunchSettings
|
|
|
|
# Runs on both H200 and B200 via nightly-8-gpu-common suite
|
|
register_cuda_ci(est_time=1800, suite="nightly-8-gpu-common", nightly=True)
|
|
|
|
QWEN35_MODEL_PATH = "Qwen/Qwen3.5-397B-A17B"
|
|
|
|
|
|
class TestQwen35(unittest.TestCase):
|
|
"""Unified test class for Qwen3.5-397B-A17B performance and accuracy.
|
|
|
|
Qwen3.5 is a 397B MoE VLM with 17B active params.
|
|
Features hybrid reasoning, tool calling, and multimodal capabilities.
|
|
Runs BOTH:
|
|
- Performance test (using NightlyBenchmarkRunner)
|
|
- Accuracy test (using run_eval with gsm8k)
|
|
"""
|
|
|
|
def test_qwen35(self):
|
|
"""Run performance and accuracy for Qwen3.5-397B-A17B."""
|
|
base_args = [
|
|
"--trust-remote-code",
|
|
"--reasoning-parser=qwen3",
|
|
"--tool-call-parser=qwen3_coder",
|
|
"--mem-fraction-static=0.8",
|
|
]
|
|
mtp_args = [
|
|
"--speculative-algorithm=EAGLE",
|
|
"--speculative-num-steps=3",
|
|
"--speculative-eagle-topk=1",
|
|
"--speculative-num-draft-tokens=4",
|
|
"--mamba-scheduler-strategy=extra_buffer",
|
|
]
|
|
|
|
variants = [
|
|
ModelLaunchSettings(
|
|
QWEN35_MODEL_PATH,
|
|
tp_size=8,
|
|
extra_args=base_args,
|
|
variant="TP8",
|
|
),
|
|
ModelLaunchSettings(
|
|
QWEN35_MODEL_PATH,
|
|
tp_size=8,
|
|
extra_args=base_args + mtp_args,
|
|
variant="TP8+MTP",
|
|
env={"SGLANG_ENABLE_SPEC_V2": "1"},
|
|
),
|
|
]
|
|
|
|
run_combined_tests(
|
|
models=variants,
|
|
test_name="Qwen3.5-397B-A17B",
|
|
accuracy_params=AccuracyTestParams(
|
|
dataset="gsm8k",
|
|
baseline_accuracy=0.95,
|
|
thinking_mode="qwen3",
|
|
max_tokens=8192,
|
|
num_examples=200,
|
|
),
|
|
performance_params=PerformanceTestParams(
|
|
profile_dir="performance_profiles_qwen35",
|
|
),
|
|
)
|
|
|
|
|
|
if __name__ == "__main__":
|
|
unittest.main()
|