API Reference

Once the MLTK is installed into the Python environment, it may be imported into a python script using:

import mltk

Once the MLTK is imported, it’s various APIs may be accessed.

API Overview

The following provides a general overview of the MLTK Python API:

Name Description
Model Operations Modeling operations such as profiling and training
MLTK Model Provides the root object of a model specification
Tensorflow-Lite Model Enables reading/writing .tflite model flatbuffer
Tensorflow-Lite Micro Model Enables running .tflite models in the Tensorflow-Lite Micro interpreter
Keras Model The model object used by Tensorflow during model training
Data Preprocessing Dataset preprocessing utilities
Utilities Common utilities
Reference Models Pre-trained reference models
Reference Datasets Datasets used by reference models

Package Directory Structure

The MLTK Python package has the following structure:

Name Description
mltk The root of the MLTK package
mltk.core Core modeling utilities, see the Model Operations docs for more details
mltk.core.model Provides the root object of a model specification, more details in the MLTK Model docs
mltk.core.preprocess Data pre-processing utilities, see the Data Preprocessing docs for more info
mltk.core.tflite_model Enables reading/writing .tflite model flatbuffers, more details in the TfliteModel docs
mltk.core.tflite_model_parameters Enables read/writing custom parameters in a .tflite model flatbuffer
mltk.core.tflite_micro Enables running .tflite models in the Tensorflow-Lite Micro interpreter, more details in the Tensorflow-Lite Micro Wrapper docs
mltk.core.keras Helper scripts for the Keras API
mltk.utils Common utility scripts, more details in the utilities docs
mltk.cli MLTK Command-Line Interface (CLI) scripts
mltk.models Reference models, more details in the Reference models docs
mltk.datasets Reference datasets, more details in the Reference datasets docs