TensorFlow tutorials from Session to GAN, circa 2017

What it does

This repo is a curated set of Python scripts teaching TensorFlow 1.x concepts in a strict progression: variables and placeholders, then regression and classification, then CNNs, RNNs, autoencoders, DQN, GANs, and transfer learning. Each script pairs with a Chinese video and text lesson on the author’s site, making it a full course in classic deep-learning archetypes. It is essentially a syllabus frozen around 2017.

The interesting bit

The author openly treats this as a period piece: the README immediately redirects visitors to a TF2+ repo and a PyTorch equivalent, yet leaves this collection intact because the underlying math and architectures—CNNs, LSTMs, GANs—haven’t changed, only the API has. That honesty is rare in tutorial land.

Key highlights

• Covers the full stack from Session and Placeholder basics up to Conditional GANs and transfer learning
• Every method has a matching Chinese video and text tutorial at 莫烦 Python
• Includes now-vintage but still-relevant advanced topics: DQN reinforcement learning, autoencoders, and batch normalization
• Author maintains parallel PyTorch and TF2+ successor repos for modern refugees

Caveats

• Built for TensorFlow 1.x (Session, Placeholder, Variable); the author explicitly recommends his TF2+ repo for modern usage
• Several later topics—dropout, batch normalization, distributed training—are listed as WIP and may be less polished than the core sections

Verdict

Grab these scripts if you are a Chinese-speaking beginner who wants to see how the sausage was made in 2017, or if you are maintaining legacy TF1 code and need a Rosetta Stone. Skip it if you want production-ready TF2 or PyTorch patterns today.