This is the accompanied code of submission of Training a General Spiking Neural Network with Improved Efficiency and Minimum Latency:

• Note：due to the size of the model, please download the SNN.zip file and follow the instructions below to complete the training and verification :).

If you are using Anaconda, you might want to use virtual environment.

conda create -n SNN python=3.7 -y
conda activate SNN

This work are based on Pytorch 1.13.1, download instructions can be found on the official website.

You can install the required dependencies using pip/conda.

pip install -r requirements.txt

from models.layers_ecml_cnn_10 import Sequencer2DBlock, PatchEmbed, Downsample2D,SNN2D,PatchMerging

parser.add_argument('--sched', default='cosine', type=str, metavar='SCHEDULER',
                    help='LR scheduler (default: "step"')
parser.add_argument('--lr', type=float, default=0.05, metavar='LR',
                    help='learning rate (default: 0.05)')

Python train.py

To prove our research, we provide the pre-train file of two versions (CNN-SNN/LSTM-SNN) on the dataset CIFAR10 and CIFAR100. Please download the related datasets for validation and load the corresponding model.

from models.layers_ecml_cnn_10 import Sequencer2DBlock, PatchEmbed, Downsample2D,SNN2D,PatchMerging

'--dataset', default='torch/cifar10', help='dataset type cifar10 or cifar100'
'--load-model', default=None, help='Path of pre training weights'  #set 'output\CIFAR10\CNN\model_best.pth.tar' for this case.
'--num-classes', default=10, help='number of label classes'

The operation is consistent with CNN-SNN except for the path of the model file
set 'output\CIFAR10\CNN\cifar_10_not_best_model_best.pth.tar' in this case.

Python load_model.py

The training results are automatically output when training finishes and the .csv file can be found. The Validation results are presented in your own IDE console. Example as:

If you find our work helpful for your research, please consider citing our paper:

@inproceedings{ACMLSNN23,
author={Yao, Yunpeng and Wu, Man and Chen, Zheng and Zhang, Renyuan},
booktitle = {The 15th Asian Conference on Machine Learning}, 
title={Training a General Spiking Neural Network with Improved Efficiency and Minimum Latency}, 
year={2023},
series = {ACML '23}}