Article Directory
The previous docker version of the tool felt cumbersome to use. This time there was a new version to install the toolchain locally under the virtual machine.
System: ubuntu 18.04
floating point conversion tool version: 1.1.17
Environment configuration details
- 1. The sh command under Ubuntu uses dash by default, and the release package script needs to be executed with bash. You can use the sudo dpkg-reconfigure dash command to select NO
- 2. The system default python is switched to python3 (python3.6 should be needed here, students who are not version 3.6 should install it by themselves)
sudo update-alternatives --install /usr/bin/python python /usr/bin/python2 100
sudo update-alternatives --install /usr/bin/python python /usr/bin/python3 150
- 3. Install and update the pip version, replace the source of pip, which will be used for later installation
pip config set global.index-url https://mirrors.aliyun.com/pypi/simple/
- 4. LINARO cross compilation tool system variable problem
If it is not configured correctly, the following error will appear:
Refer to the picture, the path is your own installation directory:
Installation process
- 1. Download the toolkit from the official website and unzip it
- 2. Run the environment detection script, cmake may be installed, just follow the prompt steps
cd install
bash install_ubuntu_deps.sh
If the environment is configured correctly:
- 3. Install the hb_mapper tool
bash install.sh
A successful operation will require restarting the terminal, indicating that it is correct.
- 4. Install promoter related tools
My virtual machine has no gpu, so the installed cpu version
sh install_horizon_mxnet_cpu.sh
Conclusion
- The tool chain is installed here, and then a model conversion, development process, etc. will be done later.
reference
https://developer.horizon.ai/forum/id=5f5f19e8cc8b1e59c8582b0a