EngineX-Hygon/sglang
5
0
Fork 0
Code Issues Pull Requests Actions 7 Projects Releases Wiki Activity

[Feature]Support ragged prefill in flashinfer mla backend (#3967)

Browse Source 
Co-authored-by: Yineng Zhang <me@zhyncs.com>
Co-authored-by: pankajroark <pankajroark@users.noreply.github.com>
This commit is contained in:
Baizhou Zhang
2025-02-28 18:13:56 -08:00
committed by GitHub
parent f3b99f73b3
commit 90a4b7d98a
9 changed files with 308 additions and 407 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
test/srt/run_suite.py
View File

@@ -23,6 +23,7 @@ suites = {
"test_gguf.py",
"test_input_embeddings.py",
"test_mla.py",
"test_mla_flashinfer.py",
"test_mla_fp8.py",
"test_json_constrained.py",
"test_large_max_new_tokens.py",

104
test/srt/test_mla_flashinfer.py Normal file
View File

@@ -0,0 +1,104 @@
import unittest
from types import SimpleNamespace
import torch
from sglang.srt.utils import kill_process_tree
from sglang.test.few_shot_gsm8k import run_eval as run_eval_few_shot_gsm8k
from sglang.test.run_eval import run_eval
from sglang.test.test_utils import (
DEFAULT_MLA_MODEL_NAME_FOR_TEST,
DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
DEFAULT_URL_FOR_TEST,
popen_launch_server,
)
class TestFlashinferMLA(unittest.TestCase):
@classmethod
def setUpClass(cls):
cls.model = "sgl-project/sglang-ci-dsv3-test"
cls.base_url = DEFAULT_URL_FOR_TEST
other_args = ["--trust-remote-code"]
if torch.cuda.is_available() and torch.version.cuda:
other_args.extend(
[
"--enable-torch-compile",
"--cuda-graph-max-bs",
"2",
"--enable-flashinfer-mla",
]
)
cls.process = popen_launch_server(
cls.model,
cls.base_url,
timeout=DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
other_args=other_args,
)
@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_few_shot_gsm8k(args)
print(metrics)
self.assertGreater(metrics["accuracy"], 0.62)
class TestFlashinferMLANoRagged(unittest.TestCase):
@classmethod
def setUpClass(cls):
cls.model = "sgl-project/sglang-ci-dsv3-test"
cls.base_url = DEFAULT_URL_FOR_TEST
other_args = ["--trust-remote-code"]
if torch.cuda.is_available() and torch.version.cuda:
other_args.extend(
[
"--enable-torch-compile",
"--disable-cuda-graph",
"--cuda-graph-max-bs",
"2",
"--enable-flashinfer-mla",
"--flashinfer-mla-disable-ragged",
]
)
cls.process = popen_launch_server(
cls.model,
cls.base_url,
timeout=DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
other_args=other_args,
)
@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_few_shot_gsm8k(args)
print(metrics)
self.assertGreater(metrics["accuracy"], 0.62)
if __name__ == "__main__":
unittest.main()