Welcome to SpArSeMoD’s documentation!

SpArSeMOod is a NAS (Neural Architecture Search) framework. It is specially designed for developing and finding neural networks suitable for microcontroller deployment, taking into account model size, working memory, accuracy and latency. It is based on the paper SpArSe.

It is build over PyTorch, GPyTorch and Ax

This documentation describes the framework, how to get started and how to use it.

Contents:

• Installation Process
• Getting Started
  • Following cnn_cost_example
• Using SpArSeMoD
  • Main Call and Configuration file
  • Data Loading
  • Search Space
  • Network Builder
  • Morphisms
  • Inspecting results
• Extending Sparse
  • Changing the Model class

Indices and tables

• Index
• Module Index
• Search Page