English | 中文

SpikingJelly is an open-source deep learning framework for Spiking Neural Network (SNN) based on PyTorch.

The documentation of SpikingJelly is written in both English and Chinese: https://spikingjelly.readthedocs.io

• Installation
• Build SNN In An Unprecedented Simple Way
• Fast And Handy ANN-SNN Conversion
• CUDA-Enhanced Neuron
• Device Supports
• Neuromorphic Datasets Supports
• Tutorials
• Citation
• Frequently Asked Questions
• About

Note that SpikingJelly is based on PyTorch. Please make sure that you have installed PyTorch before you install SpikingJelly.

Install the last stable version (0.0.0.0.4) from PyPI:

pip install spikingjelly

Note that the CUDA extensions are not included in the PyPI package. If you want to use the CUDA extensions, please install from the source codes:

From GitHub:

git clone https://github.com/fangwei123456/spikingjelly.git
cd spikingjelly
git checkout 0.0.0.0.4  # switch to the last stable version if you do not want to use the master version
python setup.py install

From OpenI：

git clone https://git.openi.org.cn/OpenI/spikingjelly.git
cd spikingjelly
git checkout 0.0.0.0.4  # switch to the last stable version if you do not want to use the master version
python setup.py install

When install from the source codes, SpikingJelly will detect whether CUDA is installed. If not, the CUDA extensions will also not be compiled.

SpikingJelly is user-friendly. Building SNN with SpikingJelly is as simple as building ANN in PyTorch:

class Net(nn.Module):
    def __init__(self, tau=100.0, v_threshold=1.0, v_reset=0.0):
        super().__init__()
        # Network structure, a simple two-layer fully connected network, each layer is followed by LIF neurons
        self.fc = nn.Sequential(
            nn.Flatten(),
            nn.Linear(28 * 28, 14 * 14, bias=False),
            neuron.LIFNode(tau=tau, v_threshold=v_threshold, v_reset=v_reset),
            nn.Linear(14 * 14, 10, bias=False),
            neuron.LIFNode(tau=tau, v_threshold=v_threshold, v_reset=v_reset)
        )

    def forward(self, x):
        return self.fc(x)

This simple network with a Poisson encoder can achieve 92% accuracy on MNIST test dataset. Read the tutorial of clock driven for more details. You can also run this code in Python terminal for training on classifying MNIST:

>>> import spikingjelly.clock_driven.examples.lif_fc_mnist as lif_fc_mnist
>>> lif_fc_mnist.main()

Read spikingjelly.clock_driven.examples to explore more advanced networks!

SpikingJelly implements a relatively general ANN-SNN Conversion interface. Users can realize the conversion through PyTorch or ONNX packages. What's more, users can customize the conversion module to add to the conversion.

class ANN(nn.Module):
    def __init__(self):
        super().__init__()
        self.network = nn.Sequential(
            nn.Conv2d(1, 32, 3, 1),
            nn.BatchNorm2d(32, eps=1e-3),
            nn.ReLU(),
            nn.AvgPool2d(2, 2),

            nn.Conv2d(32, 32, 3, 1),
            nn.BatchNorm2d(32, eps=1e-3),
            nn.ReLU(),
            nn.AvgPool2d(2, 2),

            nn.Conv2d(32, 32, 3, 1),
            nn.BatchNorm2d(32, eps=1e-3),
            nn.ReLU(),
            nn.AvgPool2d(2, 2),

            nn.Flatten(),
            nn.Linear(32, 10),
            nn.ReLU()
        )

    def forward(self,x):
        x = self.network(x)
        return x

This simple network with analog encoding can achieve 98.51% accuracy after converiosn on MNIST test dataset. Read the tutorial of ann2snn for more details. You can also run this code in Python terminal for training on classifying MNIST using converted model:

>>> import spikingjelly.clock_driven.ann2snn.examples.cnn_mnist as cnn_mnist
>>> cnn_mnist.main()

SpikingJelly provides two versions of spiking neurons: user-friendly PyTorch version and high-speed CUDA version. The followed figure compares execution time of different LIF neurons:

• Nvidia GPU
• CPU

As simple as using PyTorch.

>>> net = nn.Sequential(nn.Flatten(), nn.Linear(28 * 28, 10, bias=False), neuron.LIFNode(tau=tau))
>>> net = net.to(device) # Can be CPU or CUDA devices

SpikingJelly includes the following neuromorphic datasets:

Users can use both the origin events data and frames data integrated by SpikingJelly:

from spikingjelly.datasets.dvs128_gesture import DVS128Gesture
root_dir = 'D:/datasets/DVS128Gesture'
event_set = DVS128Gesture(root_dir, train=True, data_type='event')
frame_set = DVS128Gesture(root_dir, train=True, data_type='frame', frames_number=20, split_by='number')

More datasets will be included in the future.

If some datasets' download link are not available for some users, the users can download from the OpenI mirror:

https://git.openi.org.cn/OpenI/spikingjelly/datasets?type=0

All datasets saved in the OpenI mirror are allowable by their licence or authors' agreement.

SpikingJelly provides elaborate tutorials. Here are some of tutorials:

Figure	Tutorial
	Neurons
	Encoder
	Use single-layer fully connected SNN to identify MNIST
	Use convolutional SNN to identify Fashion-MNIST
	ANN2SNN
	Reinforcement Learning: Deep Q Learning
	Propagation Pattern
	Neuromorphic Datasets Processing
	Classify DVS128 Gesture

If you use SpikingJelly in your work, please cite it as follows:

@misc{SpikingJelly,
	title = {SpikingJelly},
	author = {Fang, Wei and Chen, Yanqi and Ding, Jianhao and Chen, Ding and Yu, Zhaofei and Zhou, Huihui and Tian, Yonghong and other contributors},
	year = {2020},
	publisher = {GitHub},
	journal = {GitHub repository},
	howpublished = {\url{https://github.com/fangwei123456/spikingjelly}},
}

"_C_..." modules in SpikingJelly are C/CUDA extensions, e.g., "_C_neuron" is the compiled C/CUDA module. Note that the CUDA extensions are not included in the PyPI package. If you need CUDA extensions, you can install from the source codes.

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

The list of developers can be found here.

Any contributions to SpikingJelly is welcome!