Skip to content

NWAVE SDK Documentation

The NWAVE SDK helps you prepare data, perform inference & train spiking neural networks, model hardware non-idealities for the H1 chip, and evaluate results. This documentation is organized to guide from setup to experiments.

How the Documentation Is Organized

  • Environment Setup: Install the SDK and validate your environment before you start.
  • Introduction to H1 v1: Learn the first Neuronova's hardware and its constraints.
  • Introduction to H1 v2: Second Neuronova's hardware exposing characteristics and main difference from first version of H1.
  • Data Preparation: Supported input formats and the preprocessing pipeline of both chips.
  • Losses: Network-level and hardware-aware loss functions.
  • Metrics: Evaluation metrics and analysis tools.
  • Utilities: Plotting and helper utilities.
  • Tutorials: End-to-end workflows for common training and deployment scenarios.
  • Reference: API-style details for layers and surrogate functions.
  • FAQ: Quick answers to common questions.
  • Releases & Downloads: Version-specific docs and downloadable release bundles.

What to Read First

  1. Environment Setup → Procedure To Install The Environment
  2. Introduction to H1 v1 → Chip Overview
  3. Introduction to H1 v2 → Chip Overview
  4. Data Preparation → Data Formats
  5. Tutorials → Start with Tutorial 1: H1v1 Model Overview
  6. Reference → Layers, Surrogates and Schedulers
  7. (Advanced) Hyper-parameter optimization for production runs → This section has complex concept that may be useful for production training on big datasets and where metrics are fundamental