Documentation for handle_deletion

Functionality

The handle_deletion function is responsible for managing the deletion of a deployed embedding model from the Triton Inference Server. This function performs several key tasks, including validating the deletion request, checking for the model iteration, and ensuring that the model utilizes a supported plugin. Additionally, it prevents concurrent deletion attempts by implementing a file lock. The function also handles the cleanup process by deleting the query model directory and the items model directory, provided that the latter is not shared.

Parameters

• task_id: This parameter represents the ID of the model deletion task that needs to be processed.

Usage

• Purpose: The function is designed to safely remove a deployed embedding model while executing necessary validations, managing file locks, and performing the cleanup of model files.

Example

To use the handle_deletion function, one would call it as follows:

handle_deletion("task123")