I'm going through several classic papers of deep learning that brought nothing less than a breakthrough in the field, also opened many opportuinities and introduced some really great ideas. Papers on VGG Net, GoogLeNet, ResNet, GAN etc. To understand them deeper and to get a real taste, I'm implementing those models and here you can find those implementations. I'm using Pytorch, cause I found it more Pythonic and gives you more freedom (though I've very limited experience with TensorFlow).