Docker Image
How to Start
To facilitate the use of the HAT algorithm toolkit, we provide a command to start the docker environment with one key to simplify the procedure of environment setup.
After getting the release package, you can go to the path where the package is located, and then complete the basic docker environment installation, add docker group users, and pull docker images according to the Docker Container Deployment , and then use the following command to prepare the environment for running the HAT example.
The version number {version} here is only an example, please replace it with the real version number of the image you have obtained.
where openexplorer/ai_toolchain_ubuntu_20_j6_gpu:{version} is the image name and -v is used to mount the local path to the docker path.
Get Execution Script
After successfully starting the docker environment, get the compressed file from the release package and extract the configs and tools folders.
Once you have all the environments, you can follow the training tutorial to train a fixed-point model using HAT step by step.