Home
Generated with Grok Build: Grok 4.5 · xAI Imagine (/imagine)
Pure-Rust implementation of the Neuromorphic Intermediate Representation (NIR) — the standard interchange format for spiking neural networks.
| Area | Capability |
|---|---|
| Graph model |
NirGraph — insertion-ordered nodes, directed edges, optional nesting (no nesting-depth stack growth) |
| Nodes | Closed NirNode enum, wire-exact type strings (CubaLIF, Conv2d, …) |
File I/O (hdf5 feature) |
io::read / write with decoded-allocation budgets, atomic writes, opt-in version policy + read budgets |
serde (serde feature) |
Debug/export Serialize / Deserialize for graph types |
| Validation |
validate_structure (heap-allocated, MAX_NESTING_DEPTH = 1024); opt-in validate_parameters (conv/pool invariants) |
| Testing | Vendored Python-produced fixtures, Hugging Face SNN→NIR corpus, proptest + cargo-fuzz |
| CI | Multi-OS matrix (Linux/macOS/Windows), cargo package artifact checks, public-API semver gate |
Mirrored from Devin DeepWiki (generated 2026-09-19, commit
b4dc7835, branchmain) by Devin (Cognition) againstmain@b4dc783. Source: Devin wiki
| Section | What's inside |
|---|---|
| 1. Overview | NIR serves as a framework-agnostic graph format for Spiking Neural Networks (SNNs)—acting as an equivalent to… |
| 2. Core Graph Model | The nir-rs crate provides a typed, in-memory representation of the Neuromorphic Intermediate Representation (N… |
| 3. HDF5 I/O Layer | The io module provides the implementation for reading and writing .nir files, the official HDF5-based intercha… |
| 4. Testing and Quality Assurance | The nir-rs project employs a multi-layered testing strategy designed to ensure wire-format fidelity with the P… |
| 5. CI/CD and Infrastructure | This section provides a high-level overview of the automation, containerization, quality enforcement, and pack… |
| 6. Compatibility Reference | The purpose of this page is to document the fixture-backed compatibility philosophy of nir-rs, the supported N… |
| 7. Glossary | This page provides definitions for the domain-specific terminology, architectural concepts, and data structure… |
Full index — 27 pages
- 1. Overview
- 1.1. Getting Started
- 1.2. Project Conventions and Versioning
- 2. Core Graph Model
- 2.1. NirGraph: Graph Container and Validation
- 2.2. NirNode: Neuromorphic Primitives
- 2.3. Tensor and Type System
- 2.4. Structural and Parameter Validation
- 2.5. Error Handling
- 3. HDF5 I/O Layer
- 3.1. Public I/O API: read and write
- 3.2. Wire Format and HDF5 Layout
- 3.3. Read Pipeline: Decoding and Field Synthesis
- 3.4. Write Pipeline: Staging, Compression, and Atomic Rename
- 3.5. Version Policy and Resource Budgets
- 4. Testing and Quality Assurance
- 4.1. Property-Based and Unit Tests
- 4.2. Fixture-Based Interoperability Tests
- 4.2.1. Hugging Face Fixtures and Attribution
- 4.3. Error-Path and Security Tests
- 4.4. Fuzz Harnesses
- 5. CI/CD and Infrastructure
- 5.1. CI Workflow
- 5.2. Packaging, Semver Gate, and Docker
- 5.3. Coverage, Static Analysis, and Quality Bar
- 6. Compatibility Reference
- 7. Glossary
Last updated: 2026-09-18
Updated by: Devin (Cognition)
Package tip reference: b4dc783 (main)