Help Center/ ModelArts/ Drawer/ Creating a Model/ Runtime Dependency
Updated on 2024-12-31 GMT+08:00
Share

Runtime Dependency

Description

Dependency of a model on the environment. Model developers need to provide the dependency name, installation method, version, and constraints. If the dependencies parameter is configured in the configuration file of a model, the configured installation dependencies are automatically listed in Runtime Dependency after this model is selected as the meta model.

Dependency packages can be installed only in pip mode. The default pip source is pypi.org. Check whether the package of the target version is available in pypi.org and check the package installation restrictions.

The wheel package on premises can be used for installation. (The wheel package must be stored in the same directory as the model file).

FAQs

How Do I Edit dependencies in the Model Configuration File?

dependencies in the model configuration file specifies the dependencies required for configuring the model inference code. This parameter requires the package name, installation method, and version constraints. Comply with the following rules when editing dependencies:

  • The installation packages must be installed in sequence. For example, before installing mmcv-full, install Cython, pytest-runner, and pytest. In the configuration file, Cython, pytest-runner, and pytest are ahead of mmcv-full.
  • If the dependency package cannot be installed online, use the wheel package on premises for installation. (The wheel package must be stored in the same directory as the model file).
  • dependencies in the model configuration file supports multiple dependency structure arrays in list format.

For details about the sample code, see How to Edit the Dependency Parameter in the Model Configuration File During Model Import.

What Can I Do if the pip Installation Package Fails to Be Downloaded?

The following shows causes of such a failure:

  • The package is not available in the pip source. The default pip source is pypi.org. Check whether the package of the target version is available in pypi.org and check the package installation restrictions.
  • The downloaded package does not match the architecture in the base image. For example, an x86 package is downloaded for Arm, or a Python 3 package is downloaded for Python 2. Check whether the installation restrictions and dependencies of the package meet the requirements. For details about the runtime environment of a base image, see Available Inference Base Images.
  • The dependencies are not configured in a correct sequence. Edit dependencies in the model configuration file according to the specifications.
Parent Topic: Creating a Model