Nan Wang's Highlights on 'The “gradient” argument in Pytorch’s “backward” function — explained by examples' | Glasp