A Chinese developer’s reproduction of the entire TF 2.0 syllabus

What it does This repo collects a Chinese developer’s personal reproduction of Google’s official TensorFlow 2.0 tutorials, rendered in Mandarin and packaged as Jupyter notebooks. It covers Keras APIs, Eager execution, AutoGraph, and standard deep-learning tasks—image classification, text classification, regression, structured data—plus installation guides for Ubuntu and Windows. Each topic links to a matching Zhihu article for extended reading.

The interesting bit With 7,787 stars, the project’s popularity suggests it filled a gap that official documentation did not: a learner-paced, fully Chinese rewrite of Google’s material, seasoned with the author’s own adaptations of Keras-style network building. It is curated pedagogy, not a new framework.

Key highlights

• Keras-centric introductions: quickstart, functional API, custom layers, model saving, Variables, and Eager mode
• Practical deep-learning coverage: image and text classification, overfitting, structured data, regression, and model persistence
• Architecture tour: MLP, CNNs (plus variants), text CNN, LSTM/GRU, autoencoders, word embeddings, DCGAN, and Boosted Trees via Estimator
• Platform-specific installation guides for Ubuntu and Windows
• Content is mirrored as Zhihu column articles with code maintained in the repository

Caveats

• The README explicitly targets TensorFlow 2.0, and it is unclear whether the notebooks have been updated for subsequent 2.x releases
• The author notes that updates are irregular (“不定期更新”)
• The material is largely a reproduction and reorganization of official Google tutorials rather than original tooling

Verdict A solid starting point if you read Chinese and want TensorFlow 2.0 concepts explained in your own language through notebooks. If you are already comfortable with English, the official docs cover the same ground more directly.