Inference Service Base Classes

AbstractInferenceService Class

The AbstractInferenceService class implements the basic infrastructure required for serving inference requests. It has the following member variables

Name	Type	Description	Developer Guidelines
model	Dictionary	Stores the model for the service	Developer should implement the load_model method to load the model from the file path provided as a parameter

The class has the following methods:

Name	Parameters	Description	Developer Guidelines
load	None	Loads the model from the model versioning system	This method should be called from the inference service entry point. Developers must implement the get_crdentials method to fetch the credentials for the versioning system and the load_model method to load the model from the file system, after they have been fetched from the model versioning system
get_credentials	None	Gets user credentials for the model versioning system	Should return a dictionary withe elements “uid” (User ID), “pwd” (password) and “env” (environment from which to fetch the model)
load_model	path of model files (string)	Loads model from the path provided	Should set the model variable to the loaded model
transform_input	request (pandas DataFrame)	Transforms the request if required	should return a pandas DataFrame with the transformed data
predict	request (pandas DataFrame)	makes a prediction using the loaded model
transform_output	response	Transforms the output response if required