BM1684-based AI Edge Server - Model Conversion, Large Model Integrated Machine
#人工智能#服务器#运维
Introduction
We operate in an SoC mode, meaning we perform model compilation, quantization, and program cross-compilation on an x86 host using tpu-nntc and libsophon. For deployment, the compiled programs are copied to the SoC platform (1684 development board / SE microserver / SM module) for execution.
Note: All the following operations are performed on an Ubuntu 20.04 system. Of course, Ubuntu 18 and 22 are also acceptable, as we primarily use the official Docker environment for configuration.
Preparation
Install Docker
First, install Docker.
# Update libraries sudo apt-get updatesudo apt-gefat upgrade# Install Docker sudo apt-get install docker.io# Execute docker commands without root privileges # Create docker user group; if it already exists, an error will be reported, which can be ignored.sudo groupadd docker # Add current user to the docker groupsudo gpasswd -