This is the simplest implementation of the paper "Gradient-based learning applied to document recognition" in PyTorch.
Note that this repo's implementation uses MaxPooling and ReLU instead of AvgPooling and Sigmoid activation. 
Python3
PyTorch >= 0.4.0
torchvision >= 0.1.8
$git clone https://github.com/ChawDoe/LeNet-5-MNIST-PyTorch.git $cd LeNet5-MNIST-PyTorch $python3 train.py model will now run on GPU if available
This repo includes the mnist dataset.
Average precision on test set: 99%