Porting China's deep-learning textbook standard to PyTorch

What it does This repository re-implements the code from the Chinese edition of Dive into Deep Learning (动手学深度学习) in PyTorch, replacing the original MXNet implementations. It pairs Jupyter notebooks with markdown chapters covering fundamentals, CNNs, RNNs, optimization, computer vision, and NLP, all rendered through a docsify site.

The interesting bit The project acts as a framework translation layer: it preserves the original book’s chapter-by-chapter pedagogy but rewrites every exercise in PyTorch, which is useful for readers who prefer PyTorch’s imperative style over MXNet.

Key highlights

• Covers the standard curriculum: linear regression up through ResNet, LSTM, attention mechanisms, and seq2seq machine translation
• Most chapters include both from-scratch and PyTorch-native implementations
• Documentation is rendered with docsify for GitHub Pages, including LaTeX formula support
• Aimed at beginners with only basic Python, linear algebra, and probability
• Nearly 20,000 GitHub stars

Caveats

• Four computer-vision chapters are unfinished: SSD, FCN, and two Kaggle competition walkthroughs remain unchecked in the table of contents
• The text is in Chinese; the README explicitly points English readers to a separate port of the English edition

Verdict A solid entry point for Chinese-speaking students who want a structured deep-learning course in PyTorch rather than MXNet. Look elsewhere if you need the English text or complete coverage of every advanced CV topic.