[CI] Remove some unneeded CP tests (#33763)
This commit is contained in:
@@ -1,63 +0,0 @@
|
|||||||
import unittest
|
|
||||||
|
|
||||||
from sglang.srt.utils import kill_process_tree
|
|
||||||
from sglang.test.ci.ci_register import register_cuda_ci
|
|
||||||
from sglang.test.kits.eval_accuracy_kit import GSM8KMixin
|
|
||||||
from sglang.test.test_utils import (
|
|
||||||
DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
|
|
||||||
DEFAULT_URL_FOR_TEST,
|
|
||||||
CustomTestCase,
|
|
||||||
popen_launch_server,
|
|
||||||
)
|
|
||||||
|
|
||||||
register_cuda_ci(est_time=400, stage="extra-b", runner_config="4-gpu-b200")
|
|
||||||
|
|
||||||
GLM52_NVFP4_MODEL_PATH = "nvidia/GLM-5.2-NVFP4"
|
|
||||||
|
|
||||||
|
|
||||||
class TestGLM52CPInterleave(GSM8KMixin, CustomTestCase):
|
|
||||||
gsm8k_accuracy_thres = 0.935
|
|
||||||
gsm8k_num_examples = 500
|
|
||||||
gsm8k_num_threads = 32
|
|
||||||
gsm8k_num_shots = 20
|
|
||||||
gsm8k_accept_length_thres = 3
|
|
||||||
|
|
||||||
@classmethod
|
|
||||||
def setUpClass(cls):
|
|
||||||
cls.model = GLM52_NVFP4_MODEL_PATH
|
|
||||||
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=[
|
|
||||||
"--tp",
|
|
||||||
"4",
|
|
||||||
"--attn-cp-size",
|
|
||||||
"4",
|
|
||||||
"--enable-prefill-cp",
|
|
||||||
"--cp-strategy",
|
|
||||||
"interleave",
|
|
||||||
"--speculative-algorithm",
|
|
||||||
"EAGLE",
|
|
||||||
"--speculative-num-steps",
|
|
||||||
"3",
|
|
||||||
"--speculative-eagle-topk",
|
|
||||||
"1",
|
|
||||||
"--speculative-num-draft-tokens",
|
|
||||||
"4",
|
|
||||||
"--mem-frac",
|
|
||||||
"0.85",
|
|
||||||
"--model-loader-extra-config",
|
|
||||||
'{"enable_multithread_load": true, "num_threads": 64}',
|
|
||||||
],
|
|
||||||
)
|
|
||||||
|
|
||||||
@classmethod
|
|
||||||
def tearDownClass(cls):
|
|
||||||
if hasattr(cls, "process") and cls.process:
|
|
||||||
kill_process_tree(cls.process.pid)
|
|
||||||
|
|
||||||
|
|
||||||
if __name__ == "__main__":
|
|
||||||
unittest.main()
|
|
||||||
@@ -1,136 +0,0 @@
|
|||||||
import unittest
|
|
||||||
from types import SimpleNamespace
|
|
||||||
|
|
||||||
from sglang.test.ci.ci_register import register_cuda_ci
|
|
||||||
from sglang.test.run_eval import run_eval
|
|
||||||
from sglang.test.test_utils import (
|
|
||||||
DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
|
|
||||||
DEFAULT_URL_FOR_TEST,
|
|
||||||
CustomTestCase,
|
|
||||||
kill_process_tree,
|
|
||||||
popen_launch_server,
|
|
||||||
)
|
|
||||||
|
|
||||||
register_cuda_ci(est_time=260, stage="extra-b", runner_config="4-gpu-h100")
|
|
||||||
|
|
||||||
GQA_MODEL_PATH = "Qwen/Qwen3-30B-A3B-FP8"
|
|
||||||
|
|
||||||
GSM8K_BASELINE_ACCURACY = 0.93
|
|
||||||
|
|
||||||
|
|
||||||
class TestGQACP2TP2EP2(CustomTestCase):
|
|
||||||
@classmethod
|
|
||||||
def setUpClass(cls):
|
|
||||||
cls.model = GQA_MODEL_PATH
|
|
||||||
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=[
|
|
||||||
"--tp-size",
|
|
||||||
"4",
|
|
||||||
"--moe-dp-size",
|
|
||||||
"2",
|
|
||||||
"--ep-size",
|
|
||||||
"2",
|
|
||||||
"--attn-cp-size",
|
|
||||||
"2",
|
|
||||||
"--enable-prefill-context-parallel",
|
|
||||||
"--cuda-graph-max-bs-decode",
|
|
||||||
"32",
|
|
||||||
"--max-running-requests",
|
|
||||||
"32",
|
|
||||||
"--trust-remote-code",
|
|
||||||
"--disable-piecewise-cuda-graph",
|
|
||||||
"--model-loader-extra-config",
|
|
||||||
'{"enable_multithread_load": true, "num_threads": 64}',
|
|
||||||
],
|
|
||||||
env={"SGLANG_ENABLE_CP_V2": "0"},
|
|
||||||
)
|
|
||||||
|
|
||||||
@classmethod
|
|
||||||
def tearDownClass(cls):
|
|
||||||
if hasattr(cls, "process") and cls.process:
|
|
||||||
kill_process_tree(cls.process.pid)
|
|
||||||
|
|
||||||
def test_gsm8k(self):
|
|
||||||
args = SimpleNamespace(
|
|
||||||
model=self.model,
|
|
||||||
eval_name="gsm8k",
|
|
||||||
num_shots=5,
|
|
||||||
num_examples=200,
|
|
||||||
max_tokens=16000,
|
|
||||||
num_threads=128,
|
|
||||||
repeat=1,
|
|
||||||
temperature=0.6,
|
|
||||||
top_p=0.95,
|
|
||||||
top_k=20,
|
|
||||||
base_url=self.base_url,
|
|
||||||
host="http://127.0.0.1",
|
|
||||||
port=int(self.base_url.split(":")[-1]),
|
|
||||||
)
|
|
||||||
metrics = run_eval(args)
|
|
||||||
print(f"{metrics=}")
|
|
||||||
self.assertGreaterEqual(metrics["score"], GSM8K_BASELINE_ACCURACY)
|
|
||||||
|
|
||||||
|
|
||||||
class TestGQACPTP2CP2EP4(CustomTestCase):
|
|
||||||
@classmethod
|
|
||||||
def setUpClass(cls):
|
|
||||||
cls.model = GQA_MODEL_PATH
|
|
||||||
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=[
|
|
||||||
"--tp-size",
|
|
||||||
"4",
|
|
||||||
"--moe-dp-size",
|
|
||||||
"1",
|
|
||||||
"--ep-size",
|
|
||||||
"4",
|
|
||||||
"--attn-cp-size",
|
|
||||||
"2",
|
|
||||||
"--enable-prefill-context-parallel",
|
|
||||||
"--cuda-graph-max-bs-decode",
|
|
||||||
"32",
|
|
||||||
"--max-running-requests",
|
|
||||||
"32",
|
|
||||||
"--trust-remote-code",
|
|
||||||
"--disable-piecewise-cuda-graph",
|
|
||||||
"--model-loader-extra-config",
|
|
||||||
'{"enable_multithread_load": true, "num_threads": 64}',
|
|
||||||
],
|
|
||||||
env={"SGLANG_ENABLE_CP_V2": "0"},
|
|
||||||
)
|
|
||||||
|
|
||||||
@classmethod
|
|
||||||
def tearDownClass(cls):
|
|
||||||
if hasattr(cls, "process") and cls.process:
|
|
||||||
kill_process_tree(cls.process.pid)
|
|
||||||
|
|
||||||
def test_gsm8k(self):
|
|
||||||
args = SimpleNamespace(
|
|
||||||
model=self.model,
|
|
||||||
eval_name="gsm8k",
|
|
||||||
num_shots=5,
|
|
||||||
num_examples=200,
|
|
||||||
max_tokens=16000,
|
|
||||||
num_threads=128,
|
|
||||||
repeat=1,
|
|
||||||
temperature=0.6,
|
|
||||||
top_p=0.95,
|
|
||||||
top_k=20,
|
|
||||||
base_url=self.base_url,
|
|
||||||
host="http://127.0.0.1",
|
|
||||||
port=int(self.base_url.split(":")[-1]),
|
|
||||||
)
|
|
||||||
metrics = run_eval(args)
|
|
||||||
print(f"{metrics=}")
|
|
||||||
self.assertGreaterEqual(metrics["score"], GSM8K_BASELINE_ACCURACY)
|
|
||||||
|
|
||||||
|
|
||||||
if __name__ == "__main__":
|
|
||||||
unittest.main()
|
|
||||||
@@ -1,80 +0,0 @@
|
|||||||
import unittest
|
|
||||||
from types import SimpleNamespace
|
|
||||||
|
|
||||||
from sglang.test.ci.ci_register import register_cuda_ci
|
|
||||||
from sglang.test.run_eval import run_eval
|
|
||||||
from sglang.test.test_utils import (
|
|
||||||
DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
|
|
||||||
DEFAULT_URL_FOR_TEST,
|
|
||||||
CustomTestCase,
|
|
||||||
kill_process_tree,
|
|
||||||
popen_launch_server,
|
|
||||||
)
|
|
||||||
|
|
||||||
register_cuda_ci(est_time=500, suite="nightly-8-gpu-b200", nightly=True)
|
|
||||||
|
|
||||||
MIMO_V2_MODEL_PATH = "XiaomiMiMo/MiMo-V2.5"
|
|
||||||
GSM8K_BASELINE_ACCURACY = 0.93
|
|
||||||
|
|
||||||
|
|
||||||
class TestMiMoV2ContextParallel(CustomTestCase):
|
|
||||||
@classmethod
|
|
||||||
def setUpClass(cls):
|
|
||||||
cls.model = MIMO_V2_MODEL_PATH
|
|
||||||
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=[
|
|
||||||
"--trust-remote-code",
|
|
||||||
"--language-only",
|
|
||||||
"--tp",
|
|
||||||
"8",
|
|
||||||
"--attn-cp-size",
|
|
||||||
"2",
|
|
||||||
"--attention-backend",
|
|
||||||
"fa4",
|
|
||||||
"--enable-prefill-cp",
|
|
||||||
"--cp-strategy",
|
|
||||||
"zigzag",
|
|
||||||
"--moe-runner-backend",
|
|
||||||
"flashinfer_trtllm",
|
|
||||||
"--moe-dense-tp-size",
|
|
||||||
"1",
|
|
||||||
"--mem-fraction-static",
|
|
||||||
"0.8",
|
|
||||||
"--chunked-prefill-size",
|
|
||||||
"8192",
|
|
||||||
],
|
|
||||||
)
|
|
||||||
|
|
||||||
@classmethod
|
|
||||||
def tearDownClass(cls):
|
|
||||||
if hasattr(cls, "process") and cls.process:
|
|
||||||
kill_process_tree(cls.process.pid)
|
|
||||||
|
|
||||||
def test_gsm8k(self):
|
|
||||||
metrics = run_eval(
|
|
||||||
SimpleNamespace(
|
|
||||||
model=self.model,
|
|
||||||
eval_name="gsm8k",
|
|
||||||
api="chat",
|
|
||||||
num_shots=5,
|
|
||||||
num_examples=200,
|
|
||||||
max_tokens=4096,
|
|
||||||
num_threads=8,
|
|
||||||
repeat=1,
|
|
||||||
temperature=0.0,
|
|
||||||
top_p=1.0,
|
|
||||||
base_url=self.base_url,
|
|
||||||
host="http://127.0.0.1",
|
|
||||||
port=int(self.base_url.split(":")[-1]),
|
|
||||||
)
|
|
||||||
)
|
|
||||||
print(f"{metrics=}")
|
|
||||||
self.assertGreaterEqual(metrics["score"], GSM8K_BASELINE_ACCURACY)
|
|
||||||
|
|
||||||
|
|
||||||
if __name__ == "__main__":
|
|
||||||
unittest.main()
|
|
||||||
Reference in New Issue
Block a user