赞
踩
默认分配的服务器,toolkit
的版本为20.1.rc1
,而msame模型推理工具的版本现在要求5.0.3.alpha001
及以上版本。虽然官方也有提供5.0.3.alpha001
前的版本,但我下载后编译会报错,所以还是升级一下CANN
开发环境,同时更新硬件驱动。
系统自带原驱动信息如下:
## 查询驱动信息
npu-smi info
## 返回如下结果
+------------------------------------------------------------------------------+
| npu-smi 20.1.0 Version: 20.1.0 |
+-------------------+-----------------+----------------------------------------+
| NPU Name | Health | Power(W) Temp(C) |
| Chip Device | Bus-Id | AICore(%) Memory-Usage(MB) |
+===================+=================+========================================+
| 0 310 | OK | 12.8 45 |
| 0 0 | 0000:00:0D.0 | 0 2703 / 8192 |
+===================+=================+========================================+
在华为官网下载支持CANN 5.1.RC1.alpha001
版本的驱动1.0.13.aplha
这里我下载的软件名称为:
A300-3010-npu_21.0.4_linux-x86_64.zip
可在服务器上执行如下命令进行下载并安装:
## 创建driver目录并进入driver目录
mkdir driver && cd driver
## 下载驱动安装包
wget --no-check-certificate https://obs-9be7.obs.cn-east-2.myhuaweicloud.com/turing/resourcecenter/Software/AtlasI/A300-3010%201.0.13/A300-3010-npu_21.0.4_linux-x86_64.zip
## 解压安装包到指定目录A310
unzip -d A310 A300-3010-npu_21.0.4_linux-x86_64.zip
## 进入A310目录
cd A310
## 再次执行解压命令(官方下载的默认解压有一个同名ZIP,需要二次解压)
## 解压出来后还有一个是固件,但云服务器无法更新固件所以更新固件
unzip A300-3010-npu_21.0.4_linux-x86_64.zip
## 安装驱动
./npu-driver_21.0.4_linux-x86_64.run --full
安装过程会得到如下提示:
Verifying archive integrity... 100% SHA256 checksums are OK. All good.
Uncompressing npu-driver-run-package 100%
[Driver] [2022-03-04 13:23:04] [INFO]Start time: 2022-03-04 13:23:04
[Driver] [2022-03-04 13:23:04] [INFO]LogFile: /var/log/ascend_seclog/ascend_install.log
[Driver] [2022-03-04 13:23:04] [INFO]OperationLogFile: /var/log/ascend_seclog/operation.log
[Driver] [2022-03-04 13:23:04] [INFO]base version is 20.1.0.
[Driver] [2022-03-04 13:23:04] [WARNING]Do not power off or restart the system during the installation/upgrade
[Driver] [2022-03-04 13:23:04] [INFO]set username and usergroup, HwHiAiUser:HwHiAiUser
[Driver] [2022-03-04 13:23:04] [INFO]Driver package has been installed on the path /usr/local/Ascend, the version is 20.1.0, and the version of this package is 21.0.4,do you want to continue? [y/n]
输入y
确认更新驱动,最终提示如下:
deleting install files...
remove install files successfully!
deleting installed folders...
remove install folders successfully!
[Driver] [2022-03-04 13:24:12] [INFO]driver install type: DKMS
[Driver] [2022-03-04 13:24:12] [INFO]upgradePercentage:10%
[Driver] [2022-03-04 13:24:16] [INFO]upgradePercentage:30%
[Driver] [2022-03-04 13:24:16] [INFO]upgradePercentage:40%
[Driver] [2022-03-04 13:25:09] [INFO]upgradePercentage:90%
[Driver] [2022-03-04 13:25:09] [INFO]upgradePercentage:100%
[Driver] [2022-03-04 13:25:10] [INFO]Driver package installed successfully! Reboot needed for installation/upgrade to take effect!
[Driver] [2022-03-04 13:25:10] [INFO]End time: 2022-03-04 13:25:10
提示重启生效,此时执行reboot
命令进行重启。
重启后执行npu-smi info
命令得到如下结果:
+--------------------------------------------------------------------------------------------+
| npu-smi 21.0.4 Version: 21.0.4 |
+-------------------+-----------------+------------------------------------------------------+
| NPU Name | Health | Power(W) Temp(C) Hugepages-Usage(page) |
| Chip Device | Bus-Id | AICore(%) Memory-Usage(MB) |
+===================+=================+======================================================+
| 13 310 | OK | 12.8 45 0 / 970 |
| 0 0 | 0000:00:0D.0 | 0 608 / 7765 |
+===================+=================+======================================================+
即可看到驱动版本已更新为21.0.4
。
在华为官网上下载CANN社区版,需要下载机器对应CPU架构的平台开发套件软件包,这里我下载的是x86_64架构的,对应软件名称为:
Ascend-cann-toolkit_5.1.RC1.alpha001_linux-x86_64.run
可在服务器上执行如下命令进行下载并安装:
推荐使用此方式下载,实测下载速度最高可达4.66MB/s,超过了我外网IP的固定带宽10 Mbit/s。
## 创建cann目录并进入cann目录
mkdir cann && cd cann
## 下载toolkit安装包
wget --no-check-certificate https://ascend-repo.obs.cn-east-2.myhuaweicloud.com/CANN/5.1.RC1.alpha001/Ascend-cann-toolkit_5.1.RC1.alpha001_linux-x86_64.run
## 添加可执行权限
chmod +x Ascend-cann-toolkit_5.1.RC1.alpha001_linux-x86_64.run
## 执行安装
./Ascend-cann-toolkit_5.1.RC1.alpha001_linux-x86_64.run --install
安装过程可得到如下提示:
Verifying archive integrity... 100% SHA256 checksums are OK. All good. Uncompressing ASCEND_RUN_PACKAGE 100% [Toolkit] [20220304-13:27:35] [INFO] mkdir /usr/local/Ascend/ascend-toolkit/5.1.RC1.alpha001/x86_64-linux [Toolkit] [20220304-13:27:35] [INFO] touch /var/log/ascend_seclog/ascend_toolkit_install.log [Toolkit] [20220304-13:27:35] [INFO] LogFile:/var/log/ascend_seclog/ascend_toolkit_install.log [Toolkit] [20220304-13:27:35] [INFO] install start [Toolkit] [20220304-13:27:35] [INFO] The install path is /usr/local/Ascend ! [Toolkit] [20220304-13:27:35] [INFO] install package CANN-runtime-5.1.RC1.alpha001-linux.x86_64.run start [Toolkit] [20220304-13:27:37] [INFO] CANN-runtime-5.1.RC1.alpha001-linux.x86_64.run --full --quiet --nox11 install success [Toolkit] [20220304-13:27:37] [INFO] install package CANN-compiler-5.1.RC1.alpha001-linux.x86_64.run start WARNING: You are using pip version 19.2.3, however version 22.0.3 is available. You should consider upgrading via the 'pip install --upgrade pip' command. WARNING: You are using pip version 19.2.3, however version 22.0.3 is available. You should consider upgrading via the 'pip install --upgrade pip' command. WARNING: You are using pip version 19.2.3, however version 22.0.3 is available. You should consider upgrading via the 'pip install --upgrade pip' command. WARNING: You are using pip version 19.2.3, however version 22.0.3 is available. You should consider upgrading via the 'pip install --upgrade pip' command. WARNING: You are using pip version 19.2.3, however version 22.0.3 is available. You should consider upgrading via the 'pip install --upgrade pip' command. WARNING: You are using pip version 19.2.3, however version 22.0.3 is available. You should consider upgrading via the 'pip install --upgrade pip' command. [Toolkit] [20220304-13:28:13] [INFO] CANN-compiler-5.1.RC1.alpha001-linux.x86_64.run --full --pylocal --quiet --nox11 install success [Toolkit] [20220304-13:28:13] [INFO] install package CANN-opp-5.1.RC1.alpha001-linux.x86_64.run start [Toolkit] [20220304-13:28:33] [INFO] CANN-opp-5.1.RC1.alpha001-linux.x86_64.run --full --quiet --nox11 install success [Toolkit] [20220304-13:28:33] [INFO] install package CANN-toolkit-5.1.RC1.alpha001-linux.x86_64.run start [Toolkit] [20220304-13:29:13] [INFO] CANN-toolkit-5.1.RC1.alpha001-linux.x86_64.run --full --pylocal --quiet --nox11 install success [Toolkit] [20220304-13:29:13] [INFO] install package Ascend-mindstudio-toolkit_3.0.3.B120_linux-x86_64.run start [Toolkit] [20220304-13:29:14] [INFO] Ascend-mindstudio-toolkit_3.0.3.B120_linux-x86_64.run --full --quiet --nox11 install success [Toolkit] [20220304-13:29:14] [INFO] install package CANN-aoe-5.1.RC1.alpha001-linux.x86_64.run start WARNING: You are using pip version 19.2.3, however version 22.0.3 is available. You should consider upgrading via the 'pip install --upgrade pip' command. [Toolkit] [20220304-13:29:16] [INFO] CANN-aoe-5.1.RC1.alpha001-linux.x86_64.run --full --quiet --nox11 install success [Toolkit] [20220304-13:29:16] [INFO] install package Ascend-test-ops_5.0.3_linux.run start [Toolkit] [20220304-13:29:17] [INFO] Ascend-test-ops_5.0.3_linux.run --full --quiet --nox11 install success [Toolkit] [20220304-13:29:17] [INFO] install package Ascend-pyACL_5.0.3_linux-x86_64.run start [Toolkit] [20220304-13:29:17] [INFO] Ascend-pyACL_5.0.3_linux-x86_64.run --full --quiet --nox11 install success [Toolkit] [20220304-13:29:17] [INFO] /etc/Ascend/ascend_cann_install.info generate success [Toolkit] [20220304-13:29:17] [INFO] Please make sure that: PATH includes : /usr/local/Ascend/ascend-toolkit/latest/bin: /usr/local/Ascend/ascend-toolkit/latest/compiler/ccec_compiler/bin: LD_LIBRARY_PATH includes : /usr/local/Ascend/ascend-toolkit/latest/lib64: /usr/local/Ascend/ascend-toolkit/latest/compiler/lib64/plugin/opskernel: /usr/local/Ascend/ascend-toolkit/latest/compiler/lib64/plugin/nnengine: PYTHONPATH includes : /usr/local/Ascend/ascend-toolkit/latest/python/site-packages: /usr/local/Ascend/ascend-toolkit/latest/opp/op_impl/built-in/ai_core/tbe: ASCEND_AICPU_PATH includes : /usr/local/Ascend/ascend-toolkit/latest: ASCEND_OPP_PATH includes : /usr/local/Ascend/ascend-toolkit/latest/opp: TOOLCHAIN_HOME includes : /usr/local/Ascend/ascend-toolkit/latest/toolkit: ASCEND_HOME_PATH includes : /usr/local/Ascend/ascend-toolkit/latest: [Toolkit] [20220304-13:29:17] [INFO] If your service is started using the shell script, you can call the /usr/local/Ascend/ascend-toolkit/set_env.sh script to configure environment variables. Note that this script can not be executed mannually. [Toolkit] [20220304-13:29:17] [INFO] Ascend-cann-toolkit_5.1.RC1.alpha001_linux-x86_64 install success,The install path is /usr/local/Ascend !
## 下载msame源码
git clone https://gitee.com/ascend/tools.git
## 切换到代码目录
cd tools/msame
## 添加可执行权限
chmod +x build.sh
## 开始编译工具
./build.sh g++
提示编译成功
-- The C compiler identification is GNU 7.5.0 -- The CXX compiler identification is GNU 7.5.0 -- Check for working C compiler: /usr/bin/cc -- Check for working C compiler: /usr/bin/cc -- works -- Detecting C compiler ABI info -- Detecting C compiler ABI info - done -- Detecting C compile features -- Detecting C compile features - done -- Check for working CXX compiler: /usr/bin/g++ -- Check for working CXX compiler: /usr/bin/g++ -- works -- Detecting CXX compiler ABI info -- Detecting CXX compiler ABI info - done -- Detecting CXX compile features -- Detecting CXX compile features - done -- env INC_PATH: /usr/local/Ascend/ascend-toolkit/latest -- env LIB_PATH: /usr/local/Ascend/ascend-toolkit/latest/acllib/lib64/stub -- Configuring done -- Generating done -- Build files have been written to: /root/cann/tools/msame/build/intermediates/host Scanning dependencies of target main [ 20%] Building CXX object CMakeFiles/main.dir/utils.cpp.o [ 40%] Building CXX object CMakeFiles/main.dir/model_process.cpp.o [ 60%] Building CXX object CMakeFiles/main.dir/sample_process.cpp.o [ 80%] Building CXX object CMakeFiles/main.dir/main.cpp.o [100%] Linking CXX executable /root/cann/tools/msame/out/main [100%] Built target main
## 创建googlenet目录并进入googlenet目录
mkdir googlenet && cd googlenet
wget https://modelzoo-train-atc.obs.cn-north-4.myhuaweicloud.com/003_Atc_Models/AE/ATC%20Model/classification/googlenet.om
../out/msame --model ./googlenet.om --output ./output --outfmt BIN --loop 1 [INFO] acl init success [INFO] open device 0 success [INFO] create context success [INFO] create stream success [INFO] get run mode success [INFO] load model ./googlenet.om success [INFO] create model description success [INFO] get input dynamic gear count success [INFO] create model output success ./output/202234_12_58_29_218043 [INFO] model execute success Inference time: 1.963ms [INFO] get max dynamic batch size success [INFO] output data success Inference average time: 1.963000 ms [INFO] destroy model input success [INFO] unload model success, model Id is 1 [INFO] Execute sample success [INFO] end to destroy stream [INFO] end to destroy context [INFO] end to reset device is 0 [INFO] end to finalize acl
上述输出结果提示成功,无报错!
Copyright © 2003-2013 www.wpsshop.cn 版权所有,并保留所有权利。