Files
sglang/test/registered/backends/test_flashinfer_trtllm_gen_moe_backend.py

111 lines
3.1 KiB
Python

import os
import unittest
from types import SimpleNamespace
from sglang.srt.utils import kill_process_tree
from sglang.test.ci.ci_register import register_cuda_ci
from sglang.test.few_shot_gsm8k import run_eval
from sglang.test.test_utils import (
DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
DEFAULT_URL_FOR_TEST,
CustomTestCase,
popen_launch_server,
)
register_cuda_ci(est_time=300, suite="nightly-4-gpu-b200", nightly=True)
class TestFlashinferTrtllmGenMoeBackendFP8(CustomTestCase):
@classmethod
def setUpClass(cls):
cls.model = "Qwen/Qwen3-Next-80B-A3B-Instruct-FP8"
cls.base_url = DEFAULT_URL_FOR_TEST
cls.process = popen_launch_server(
cls.model,
cls.base_url,
timeout=DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
env={**os.environ, "SGLANG_ENABLE_JIT_DEEPGEMM": "False"},
other_args=[
"--attention-backend",
"triton",
"--moe-runner-backend",
"flashinfer_trtllm",
"--tp-size",
"4",
"--ep-size",
"4",
"--mem-fraction-static",
"0.7",
"--mamba-ssm-dtype",
"bfloat16",
],
)
@classmethod
def tearDownClass(cls):
kill_process_tree(cls.process.pid)
def test_gsm8k(self):
args = SimpleNamespace(
num_shots=5,
data_path=None,
num_questions=200,
max_new_tokens=512,
parallel=128,
host="http://127.0.0.1",
port=int(self.base_url.split(":")[-1]),
)
metrics = run_eval(args)
print(f"{metrics=}")
self.assertGreater(metrics["accuracy"], 0.93)
class TestFlashinferTrtllmGenMoeBackendBF16(CustomTestCase):
@classmethod
def setUpClass(cls):
cls.model = "Qwen/Qwen3-Next-80B-A3B-Instruct"
cls.base_url = DEFAULT_URL_FOR_TEST
cls.process = popen_launch_server(
cls.model,
cls.base_url,
timeout=DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
other_args=[
"--attention-backend",
"triton",
"--moe-runner-backend",
"flashinfer_trtllm",
"--cuda-graph-max-bs",
"512",
"--tp-size",
"4",
"--ep-size",
"4",
"--mem-fraction-static",
"0.7",
"--mamba-ssm-dtype",
"bfloat16",
],
)
@classmethod
def tearDownClass(cls):
kill_process_tree(cls.process.pid)
def test_gsm8k(self):
args = SimpleNamespace(
num_shots=5,
data_path=None,
num_questions=200,
max_new_tokens=512,
parallel=128,
host="http://127.0.0.1",
port=int(self.base_url.split(":")[-1]),
)
metrics = run_eval(args)
print(f"{metrics=}")
self.assertGreater(metrics["accuracy"], 0.93)
if __name__ == "__main__":
unittest.main()