Skip to main content

T1C Ecosystem Introduction

T1C-IR is a neuromorphic intermediate representation designed for Type 1 Compute hardware. The ecosystem consists of four packages that work together to bridge ML frameworks with neuromorphic deployment.

The T1C-SDK​

For most users, the recommended entry point is the T1C-SDK:

pip install t1c-sdk

This installs all ecosystem packages and provides a unified CLI. See the SDK Documentation for details.

Package Overview​

PackagePurposeDependencies
t1c-sdkSDK meta-package and CLIt1c-ir, t1c-bridge, t1c-viz
t1c.ir (t1c-ir)Core IR primitives and HDF5 serializationnumpy, h5py, rustworkx
t1c.bridge (t1c-bridge)PyTorch export/import bridget1c-ir, torch, snntorch
t1c.viz (t1c-viz)Interactive graph and spike visualizationt1c-ir, pillow

Design Philosophy​

The architecture mirrors the ONNX ecosystem:

PyTorch Model → bridge.to_ir() → T1C-IR Graph → ir.write() → .t1c file
                                                                     ↓
Hardware Runtime ← ir.read() ← .t1c file ← bridge.ir_to_torch() ←

(This assumes from t1c import ir, bridge.)

This separation enables:

  • ML engineers export models without understanding hardware details
  • Hardware teams consume standardized IR without touching PyTorch
  • Visualization works independently of both

Key Features​

  • 22+ primitives covering SNNs, ANNs, and hybrid architectures
  • Hybrid ANN-SNN support with activations (ReLU, Sigmoid, GELU, Softmax), normalization (BatchNorm, LayerNorm), and dropout
  • NIR-compliant LIF neuron dynamics for interoperability
  • RustworkX-powered graph algorithms for 100x faster validation and cycle detection
  • HDF5 serialization for efficient storage and portability
  • snnTorch integration via thin wrapper functions
  • Netron-style visualization with clickable nodes and parameter inspection
  • Spike/event visualization for neuromorphic datasets (23-30M events/sec)

Use Cases​

T1C-IR is designed for:

  • Model deployment to Type 1 Compute neuromorphic hardware
  • Hybrid ANN-SNN architectures with encoder-SNN-decoder patterns
  • Model exchange between ML and hardware teams
  • Architecture visualization for debugging and documentation
  • Event-based data visualization for neuromorphic datasets
  • Round-trip verification ensuring export/import fidelity

Project Repositories​

Repositories (Git clone names) and their installed package names:

  • t1c-sdk: SDK meta-package and CLI (installs as t1c-sdk)
  • t1cir: Core IR package (installs as t1c-ir; use from t1c import ir)
  • t1ctorch: PyTorch bridge (installs as t1c-bridge; use from t1c import bridge)
  • t1cviz: Visualization tools (installs as t1c-viz; use from t1c import viz)

Next Steps​