Installation Precautions

Development Environment

The development environment is used to convert model files, develop and compile applications, and develop operators.

If you need to debug an application (for example, an image classification application) in the development environment in addition to developing and compiling the application, you need to configure Ascend AI devices (such as inference cards or AI accelerator modules) in the development environment.

The pure development environment requires only the development kit to be installed.

In the development + commissioning scenario, you also need to install the driver and firmware corresponding to the inference card or AI accelerator module.

Operating Environment

The operating environment is an environment for running applications developed by users. The operating environment must be configured with Ascend AI devices (such as inference cards or AI accelerator modules). Applications depend on the offline inference engine package, Toolbox, driver, and firmware to perform offline accelerated inference.

The operating environment includes the following two scenarios:

• Physical machine scenario: The inference card driver, firmware, offline inference engine package, and Toolbox are installed on a physical machine. Applications developed by users run on the physical machine.
• In the container scenario, you need to install Docker.
  1. The inference card driver, firmware, and Toolbox are installed on a host.
  2. The offline inference engine package, user-developed applications, and basic container images are encapsulated into a service inference image.
  3. In the operating environment, the image is used to start the containers. The applications then run as containers.

If the configured Ascend AI device is an Atlas 200 AI accelerator module (model 3000), the application operating environment supports only physical machines.