昇腾CANN模型转换

发表于 2024-08-09 更新于 2024-08-19

一、开发环境配置

选择Ubuntu20.04系统

换源

1	sudo vim /etc/apt/sources.list

使用国内镜像源

deb http://mirrors.aliyun.com/ubuntu/ bionic main restricted universe multiverse
deb-src http://mirrors.aliyun.com/ubuntu/ bionic main restricted universe multiverse
 
deb http://mirrors.aliyun.com/ubuntu/ bionic-security main restricted universe multiverse
deb-src http://mirrors.aliyun.com/ubuntu/ bionic-security main restricted universe multiverse
 
deb http://mirrors.aliyun.com/ubuntu/ bionic-updates main restricted universe multiverse
deb-src http://mirrors.aliyun.com/ubuntu/ bionic-updates main restricted universe multiverse
 
deb http://mirrors.aliyun.com/ubuntu/ bionic-proposed main restricted universe multiverse
deb-src http://mirrors.aliyun.com/ubuntu/ bionic-proposed main restricted universe multiverse
 
deb http://mirrors.aliyun.com/ubuntu/ bionic-backports main restricted universe multiverse
deb-src http://mirrors.aliyun.com/ubuntu/ bionic-backports main restricted universe multiverse

安装环境依赖

1
2
3

sudo apt update && sudo apt upgrade
sudo apt-get install -y gcc g++ make cmake zlib1g zlib1g-dev openssl libsqlite3-dev libssl-dev libffi-dev 
sudo apt-get install -y unzip pciutils net-tools libblas-dev gfortran libblas3 libopenblas-dev libncursesw5-dev

python环境配置
安装miniconda

mkdir -p ~/miniconda3
wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh -O ~/miniconda3/miniconda.sh
bash ~/miniconda3/miniconda.sh -b -u -p ~/miniconda3
rm -rf ~/miniconda3/miniconda.sh
~/miniconda3/bin/conda init zsh

创建cann的conda环境

1 2	conda create -n cann python=3.9 conda activate cann

pip、conda换源,此处省略，自行查阅相关资料
安装基础包

1
2
3

pip3 install --upgrade pip
pip3 install attrs numpy==1.17.2 decorator sympy cffi pyyaml pathlib2 psutil protobuf scipy requests
pip install onnx

二、安装CANN toolkit

下载toolkit开发工具
官方最新：华为社区版资源下载
个人：Ascend-cann-toolkit_8.0.RC3.alpha001_linux-x86_64.run 提取码: abel
个人：Ascend-cann-toolkit_8.0.RC3.alpha001_linux-aarch64.run 提取码: abel

安装开发工具

1 2	chmod +x Ascend-cann-toolkit_8.0.RC3.alpha001_linux-x86_64.run ./Ascend-cann-toolkit_8.0.RC3.alpha001_linux-x86_64.run --install

配置交叉编译环境
对于Atlas 200 AI加速模块（RC场景）和Atlas 500 小站(运行环境aarch64架构)来说，当开发环境是一台X86 PC进行环境搭建时，需要在开发环境中安装交叉编译工具
1
sudo apt install g++-aarch64-linux-gnu

添加依赖库
出现报错

1	libascend_hal.so: cannot open shared object file:No such……

进入到Ascend目录下输入

1	find . -name libascend_hal.so

把libascend_hal.so复制到/usr/local/lib/下

1	sudo cp ~/Ascend/ascend-toolkit/8.0.RC3.alpha001/x86_64-linux/devlib/linux/x86_64/libascend_hal.so /usr/local/lib/

进入“/etc/ld.so.conf”,并将“/usr/local/lib”添加至文件最后一行

1	/usr/local/lib/

更新共享库缓存

1 2	ldconfig sudo ldconfig

添加环境变量
在”~/.zshrc”中加入”source /home/abelxiaoxing/Ascend/ascend-toolkit/set_env.sh”
1
lvim ~/.zshrc
1
source /home/abelxiaoxing/Ascend/ascend-toolkit/set_env.sh
重新载入zsh配置
1
source ~/.zshrc

三、onnx到om模型转换

查看模型输入名称，此处用’convnextv2_nano.onnx’来示范

import onnx

model = onnx.load('convnextv2_nano.onnx')
for input in model.graph.input:
    print(input.name)

此处输出”input”，因此于–input_shape中，填入的模型名称为”input”

atc模型转换

1	atc --model=./convnextv2_nano.onnx --framework=5 --output=./model --soc_version=Ascend310B1 --input_shape="input:1,3,224,224"

一、开发环境配置

二、安装CANN toolkit

三、onnx到om模型转换

参考文献