PySDK basics: Three types of model zoos

Guides PySDK Basics
1

A model zoo is a collection of AI models. PySDK supports three types of model zoos.

  1. Hosted zoo: The models are stored in the host. When loading of a model from a hosted zoo is requested, the machine running the AI model downloads the model from host to a local folder and runs inferences on the model. This type of model zoo works with all three types of inference: hosted inference, AI server inference, and local inference.
  2. Local zoo: This option works only for AI server inference. In this case, the models are served from a local folder which is specified when the AI server is started on the machine.
  3. Single file zoo: This option works only for local inference. In this case, the model zoo consists of a single model. The path to the model configuration JSON file is used to specify the model zoo.
Inference type Hosted zoo Local zoo Single file zoo
Hosted inference :white_check_mark: :x: :x:
AI server inference :white_check_mark: :white_check_mark: :x:
Local inference :white_check_mark: :x: :white_check_mark: