SpikingJelly

中文 | English

Contents

• Why SpikingJelly
• Installation
• Quick Start
• Core Capabilities
  • Backend Performance
  • Large-Scale SNN Systems
  • Datasets
  • Interchange and Deployment
• Project Status and Version Notes
• Acknowledgement
• Contributing
• Citation

Why SpikingJelly

SpikingJelly is a PyTorch-native framework for spiking neural networks (SNNs), with support for large-scale SNN training and inference.

• Beginner-friendly API
• ANN2SNN conversion
• Event-based datasets
• Acceleration backends: torch, cupy, triton
• Memory-efficient training, distributed execution, precision control
• Hardware deployment and framework exchange

Installation

SpikingJelly is built on PyTorch. Install PyTorch, torchvision, and torchaudio first.

• Python >=3.11
• PyTorch >=2.6.0 (tested with 2.7.1)

Install the latest stable release 0.0.0.0.15:

pip install spikingjelly

Install the latest development version from source:

git clone https://github.com/fangwei123456/spikingjelly.git
cd spikingjelly
pip install .

Optional dependencies:

Feature	Install
CuPy backend	pip install cupy-cuda12x or pip install cupy-cuda11x
Triton backend	pip install triton==3.3.1
NIR exchange	pip install nir nirtorch
Lightning integration	pip install lightning jsonargparse[signatures]

Version note: SpikingJelly uses a 0.0.0.0.X scheme where odd X tracks development versions and even X tracks stable releases.

Quick Start

Define an SNN in the same way that you would define any PyTorch model:

from torch import nn
from spikingjelly.activation_based import layer, neuron, surrogate

net = nn.Sequential(
    layer.Flatten(),
    layer.Linear(28 * 28, 10, bias=False),
    neuron.LIFNode(tau=2.0, surrogate_function=surrogate.ATan())
)

Next steps:

• Learn about spiking neurons
• Train on event-based datasets
• Convert ANN to SNN

Core Capabilities

Area	What SpikingJelly provides
SNN modeling	Activation-based SNN components: spiking neurons, surrogate gradients, stateful and stateless modules. Predefined SNN models.
Training workflows	PyTorch-native training flows, online-learning utilities, and ANN2SNN conversion
Performance	torch, cupy, and triton backends, FlexSN for customized neuron kernels, and mixed-precision training utilities (e.g., fp8)
Scaling	Memory-efficient training, and distributed training
Datasets	Neuromorphic datasets, and data preprocessing pipelines
Analysis	FLOPs / SynOps / memory-access profiling, and inference energy estimation
Interchange and deployment	NIR, Lava, and Lynxi-oriented exchange interfaces for neuromorphic workflows

Backend Performance

Spiking neuron models run on torch, cupy, or triton backends. The backend is set at neuron creation and can be changed later. All backends are compatible with torch.compile.

Below: execution time comparison for multi-step LIF neurons on torch vs cupy. Triton is covered in the backend tutorials.

Large-Scale SNN Systems

For large-scale SNN systems, SpikingJelly provides:

• Memory-efficient training with spike compression (memopt)
• Experimental distributed execution for multi-GPU workloads
• Precision policy tools for large-scale training and inference
• Spiking transformer components

Datasets

SpikingJelly includes the following event-based and neuromorphic datasets:

• ASL-DVS
• Bullying10K
• CIFAR10-DVS
• DVS-Lip
• DVS128 Gesture
• ES-ImageNet
• HARDVS
• N-Caltech101
• N-MNIST
• Nav Gesture
• SHD
• SSC
• Speech Commands

Each dataset supports raw event access and frame representations. See the neuromorphic datasets tutorial for the full workflow.

Interchange and Deployment

Export SpikingJelly models to neuromorphic hardware or other frameworks:

• NIR exchange
• Lava exchange
• Lynxi deployment

Project Status and Version Notes

Development / release policy:

Odd version numbers track the development branch on GitHub / OpenI. Even version numbers are stable releases published to PyPI.

Compatibility, migration, and older docs

• From 0.0.0.0.14, modules including clock_driven and event_driven were renamed. See Migrate From Old Versions.
• The default documentation points to the latest development version.
• If you rely on an older release, check bugs.md and switch to the matching documentation version.

Historical documentation:

• zero
• 0.0.0.0.4
• 0.0.0.0.6
• 0.0.0.0.8
• 0.0.0.0.10
• 0.0.0.0.12
• 0.0.0.0.14
• latest

Acknowledgement

Maintainers

Current maintainers (since July 2024):

• Yifan Huang
• Peng Xue

Previous core maintainers (before July 2024):

• Wei Fang
• Yanqi Chen
• Jianhao Ding
• Ding Chen
• Liwei Huang

Contributors

The full contributor list is on the contributors page.

Institutes

The main institutions behind SpikingJelly are Multimedia Learning Group, Institute of Digital Media (NELVT), Peking University and Peng Cheng Laboratory.

Community and related links

• Documentation
• Contributing Guide
• Issues and development discussion
• OpenI mirror
• Community Jupyter tutorials in Chinese

Contributing

We welcome issues, pull requests, documentation improvements, and translations.

• Read the Contributing Guide.
• Check issues for ongoing work.
• API docs are not fully bilingual yet; translation contributions are especially welcome.

Citation

Publications using SpikingJelly are listed on the documentation page. The source of truth in this repository is publications.json.

If you use SpikingJelly in your work, please cite:

@article{
doi:10.1126/sciadv.adi1480,
author = {Wei Fang and Yanqi Chen and Jianhao Ding and Zhaofei Yu and Timothee Masquelier and Ding Chen and Liwei Huang and Huihui Zhou and Guoqi Li and Yonghong Tian},
title = {SpikingJelly: An open-source machine learning infrastructure platform for spike-based intelligence},
journal = {Science Advances},
volume = {9},
number = {40},
pages = {eadi1480},
year = {2023},
doi = {10.1126/sciadv.adi1480},
url = {https://www.science.org/doi/abs/10.1126/sciadv.adi1480},
eprint = {https://www.science.org/doi/pdf/10.1126/sciadv.adi1480}
}