Welcome to sinabs-dynapcnn’s documentation!

This sinabs plugin enables porting sinabs models to chips and dev-kits with DYNAP-CNN technology.

To get started take a look at the

• The documentation below

• cnn-to-dynapcnn tutorial.

• examples

Table of contents

• Installation
  • Requirements
  • Installing sinabs-dynapcnn
• The basics
  • TLDR;
  • Model conversion to DYNAP-CNN core structure
  • Layer conversion
  • Parameter quantization
  • Device selection
  • List of devices currently recognized by samna
  • Placement of layers on device cores
  • Porting model to device
  • Sending and receiving spikes
  • Monitoring layer activity
• Advanced
  • Under the hood
  • The config object
  • Memory constraints
  • Attributes of interest
  • Conversion between raster and spike streams
  • Samna: The interface library to the chip
• Dangers
  • Biases are a NO-NO
  • Event driven vs time-step based evaluation
• Visualizing DynapCNN input and output events
  • Introduction
  • Available plots in samnagui
  • Useful samna nodes for visualizing and Just-In-Time (JIT) compiled nodes.
  • Dynapcnn Visualizer
• Tutorials
  • Overview of DYNAP-CNN and SPECK devlopment kit
  • Role of sinabs-dynapcnn Library
  • Chip Resources
  • Known dev-kit Bugs
  • FAQs
  • From a CNN model to a DYNAP-CNN DevKit
  • Visualize DVS Input
  • Power Monitoring
  • Using Readout Layer
  • Spike Count Visualization
• API
  • sinabs.backend.dynapcnn
• Developer’s guide
  • How to add support for a new chip
  • Contact us
• Release History
  • version 0.3

Indices and tables

• Index

• Module Index

• Search Page