TensorFlow Tutorials That Bring Their Own Professor

What it does A collection of 23 Jupyter notebooks covering deep learning fundamentals and TensorFlow specifics, from simple linear models to reinforcement learning and machine translation. Each tutorial isolates a single topic and pairs the notebook with a narrated YouTube video. The author targets absolute beginners and documents the source code extensively.

The interesting bit The project straddles two eras of TensorFlow: about half the notebooks have been updated for TensorFlow 2 (some still relying on v1 compatibility mode), while the rest remain locked to the TF1 API. That split turns the repo into an accidental museum of the framework’s evolution—complete with obsolete APIs like Pretty Tensor and the Estimator API still on display.

Key highlights

• One topic per notebook: CNNs, transfer learning, adversarial examples, DeepDream, style transfer, NLP, time-series prediction, and more
• Every tutorial has a matching YouTube video in a single playlist
• Notebooks run in Google Colab without local installation
• Chinese translation available; author actively invites more translations and community video dubs
• Source code is split into reusable helper files shared across multiple tutorials

Caveats

• Roughly half the tutorials still require TensorFlow 1, which the author explicitly states would take too much effort to port
• Some TF2 tutorials still run in “v.1 compatibility mode,” so the code may not feel idiomatic to modern Keras-native developers
• The installation instructions mention Python 3.5/3.6 and warn that Python 2.7 throws errors, suggesting parts of the repo are aging

Verdict Newcomers who learn best by watching someone walk through code will find this invaluable. If you already speak fluent Keras and expect pure TF2 idioms, you will spend a lot of time reading legacy API surface area.