Commit 386c0a76 authored by gaoqiong's avatar gaoqiong
Browse files

根据ubuntu编译做一些源码调整

parent 1884fd60
...@@ -55,13 +55,18 @@ docker run -it --name baichuan --shm-size=1024G -v /opt/hyhal:/opt/hyhal --devic ...@@ -55,13 +55,18 @@ docker run -it --name baichuan --shm-size=1024G -v /opt/hyhal:/opt/hyhal --devic
2、要是非光源提供镜像,配置环境:(若安装过慢,可以添加源:pip3 install xxx -i https://pypi.tuna.tsinghua.edu.cn/simple/) 2、要是非光源提供镜像,配置环境:(若安装过慢,可以添加源:pip3 install xxx -i https://pypi.tuna.tsinghua.edu.cn/simple/)
``` ```
#centos7.6环境
pip3 install -r requirements.txt pip3 install -r requirements.txt
pip3 install urllib3==1.24 pip3 install urllib3==1.24
yum install rapidjson yum install rapidjson
# gcc版本需要>=9 安装高版本gcc 要是必须使用gcc7,可以下载对应的gcc7的分支
yum install -y centos-release-scl #ubuntu20.04环境
yum install -y devtoolset-9 export NCCL_LIB_DIR=/opt/dtk/cuda/lib64
scl enable devtoolset-9 bash pip3 install -r requirements.txt
pip3 install urllib3==1.24
#apt-get 换源,添加清华源
sudo apt-get update
apt-get install rapidjson-dev
# 执行nccl环境变量 # 执行nccl环境变量
export NCCL_LAUNCH_MODE=GROUP export NCCL_LAUNCH_MODE=GROUP
...@@ -156,8 +161,8 @@ lmdeploy convert ${model_name} ${model_path} --model_format awq --group-size ${g ...@@ -156,8 +161,8 @@ lmdeploy convert ${model_name} ${model_path} --model_format awq --group-size ${g
#step1:模型运行 #step1:模型运行
lmdeploy chat turbomind ./workspace --tp ${tp} lmdeploy chat turbomind ./workspace --tp ${tp}
``` ```
注意事项: 注意事项:
1.该版本暂时仅支持tp=1 单卡量化推理; 1.该版本暂时仅支持tp=1 单卡量化推理,仅支持卡型KM-AI,暂不支持K100/Z100/Z100L
2.该版本量化推理功能仅支持先通过convert模型转换为turbomind格式,然后进行推理运行,暂时不知道hf模型直接量化推理; 2.该版本量化推理功能仅支持先通过convert模型转换为turbomind格式,然后进行推理运行,暂时不知道hf模型直接量化推理;
3.该版本暂时不支持通过数据集进行量化功能,需要在别处获取量化模型; 3.该版本暂时不支持通过数据集进行量化功能,需要在别处获取量化模型;
......
pybind11 pybind11
setuptools setuptools
wheel
\ No newline at end of file
...@@ -60,12 +60,12 @@ def get_version_add(sha: Optional[str] = None) -> str: ...@@ -60,12 +60,12 @@ def get_version_add(sha: Optional[str] = None) -> str:
version += ".dtk" + rocm_version version += ".dtk" + rocm_version
# torch version # torch version
version += ".torch" + torch.__version__[:4] version += ".torch" + torch.__version__[:5]
lines=[] lines=[]
with open(add_version_path, 'r',encoding='utf-8') as file: with open(add_version_path, 'r',encoding='utf-8') as file:
lines = file.readlines() lines = file.readlines()
lines[2] = "__dcu_version__ = '0.2.6+{}'\n".format(version) lines[2] = "__dcu_version__ = '0.2.6+das1.1.{}'\n".format(version)
with open(add_version_path, encoding="utf-8",mode="w") as file: with open(add_version_path, encoding="utf-8",mode="w") as file:
file.writelines(lines) file.writelines(lines)
file.close() file.close()
...@@ -212,6 +212,7 @@ if __name__ == '__main__': ...@@ -212,6 +212,7 @@ if __name__ == '__main__':
author='OpenMMLab', author='OpenMMLab',
author_email='openmmlab@gmail.com', author_email='openmmlab@gmail.com',
packages=find_packages(exclude=()), packages=find_packages(exclude=()),
package_data={ package_data={
'lmdeploy': lmdeploy_package_data, 'lmdeploy': lmdeploy_package_data,
}, },
...@@ -235,4 +236,4 @@ if __name__ == '__main__': ...@@ -235,4 +236,4 @@ if __name__ == '__main__':
'Intended Audience :: Science/Research', 'Intended Audience :: Science/Research',
], ],
entry_points={'console_scripts': ['lmdeploy = lmdeploy.cli:run']}, entry_points={'console_scripts': ['lmdeploy = lmdeploy.cli:run']},
) )
\ No newline at end of file
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment