mirror of
https://github.com/PaddlePaddle/FastDeploy.git
synced 2025-12-24 13:28:13 +08:00
8e0f4dfd0c
* add xpu ci * add case * add case * fix ci bug * Update Docker image tag to 'latest' in CI workflow * Fix set -e usage in run_xpu_ci_pytest.sh * add pd case * add case * Configure pip to use Tsinghua mirror for dependencies Set the global pip index URL to Tsinghua mirror. * fix ci bug * fix bug * fix bug --------- Co-authored-by: suijiaxin <suijiaxin@Suis-MacBook-Pro.local> Co-authored-by: root <root@gajl-bbc-onlinec-com-1511964.gajl.baidu.com> Co-authored-by: root <root@gajl-bbc-onlinec-com-1511972.gajl.baidu.com>
102 lines
2.8 KiB
Python
102 lines
2.8 KiB
Python
# Copyright (c) 2025 PaddlePaddle Authors. All Rights Reserved.
|
|
#
|
|
# Licensed under the Apache License, Version 2.0 (the "License");
|
|
# you may not use this file except in compliance with the License.
|
|
# You may obtain a copy of the License at
|
|
#
|
|
# http://www.apache.org/licenses/LICENSE-2.0
|
|
#
|
|
# Unless required by applicable law or agreed to in writing, software
|
|
# distributed under the License is distributed on an "AS IS" BASIS,
|
|
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
# See the License for the specific language governing permissions and
|
|
# limitations under the License.
|
|
|
|
"""
|
|
W4A8模式测试 - ERNIE-4.5-300B W4A8C8量化模型
|
|
|
|
测试配置:
|
|
- 模型: ERNIE-4.5-300B-A47B-W4A8C8-TP4-Paddle
|
|
- 量化: W4A8
|
|
- Tensor Parallel: 4
|
|
"""
|
|
|
|
|
|
import openai
|
|
import pytest
|
|
from conftest import get_model_path, get_port_num, print_logs_on_failure, start_server
|
|
|
|
|
|
def test_w4a8(xpu_env):
|
|
"""W4A8量化模式测试"""
|
|
|
|
print("\n============================开始W4A8测试!============================")
|
|
|
|
# 获取配置
|
|
port_num = get_port_num()
|
|
model_path = get_model_path()
|
|
|
|
# 构建服务器启动参数
|
|
server_args = [
|
|
"--model",
|
|
f"{model_path}/ERNIE-4.5-300B-A47B-W4A8C8-TP4-Paddle",
|
|
"--port",
|
|
str(port_num),
|
|
"--engine-worker-queue-port",
|
|
str(port_num + 1),
|
|
"--metrics-port",
|
|
str(port_num + 2),
|
|
"--cache-queue-port",
|
|
str(port_num + 47873),
|
|
"--tensor-parallel-size",
|
|
"4",
|
|
"--num-gpu-blocks-override",
|
|
"16384",
|
|
"--max-model-len",
|
|
"32768",
|
|
"--max-num-seqs",
|
|
"64",
|
|
"--quantization",
|
|
"W4A8",
|
|
]
|
|
|
|
# 启动服务器
|
|
if not start_server(server_args):
|
|
pytest.fail("W4A8模式服务启动失败")
|
|
|
|
# 执行测试
|
|
try:
|
|
ip = "0.0.0.0"
|
|
client = openai.Client(base_url=f"http://{ip}:{port_num}/v1", api_key="EMPTY_API_KEY")
|
|
|
|
# 非流式对话
|
|
response = client.chat.completions.create(
|
|
model="default",
|
|
messages=[
|
|
{"role": "user", "content": "你好,你是谁?"},
|
|
],
|
|
temperature=1,
|
|
top_p=0,
|
|
max_tokens=64,
|
|
stream=False,
|
|
)
|
|
|
|
print(f"\n模型回复: {response.choices[0].message.content}")
|
|
|
|
# 验证响应
|
|
assert any(
|
|
keyword in response.choices[0].message.content
|
|
for keyword in ["人工智能", "文心一言", "小度", "百度", "智能助手"]
|
|
), f"响应内容不符合预期: {response.choices[0].message.content}"
|
|
|
|
print("\nW4A8测试通过!")
|
|
|
|
except Exception as e:
|
|
print(f"\nW4A8测试失败: {str(e)}")
|
|
print_logs_on_failure()
|
|
pytest.fail(f"W4A8测试失败: {str(e)}")
|
|
|
|
|
|
if __name__ == "__main__":
|
|
pytest.main([__file__, "-v", "-s"])
|