What it does

This repo is a collection of TensorFlow exercises organized by topic—constants, variables, math operations, neural networks, seq2seq, audio processing—each paired with its NumPy equivalent. The author assumes you already know NumPy and want to map that knowledge onto TensorFlow’s API surface.

The interesting bit

The pedagogical bet is that analogy beats abstraction: instead of explaining TensorFlow’s computational graph model from first principles, it shows you the NumPy way and the TensorFlow way side by side. For someone who can already vectorize in their sleep, this skips the “what is a tensor?” throat-clearing.

Key highlights

• Covers ~15 topic areas, most marked complete (Strings, Images, and Neural Network Part 3 are still WIP)
• Explicitly targets the “NumPy mental model → TensorFlow” translation layer
• Includes practical territory often skipped in intros: reading data, sparse tensors, audio processing, seq2seq
• Last major update appears to be circa 2017 (TensorFlow 1.x era, based on the README’s own timestamp)

Caveats

• The “2017” framing and graph/session terminology strongly suggest this predates TensorFlow 2.x’s eager-by-default redesign; some patterns may be outdated
• Several sections remain unfinished (WIP) with no indication of when or if they will be completed
• No automated tests or CI visible; it’s a personal notebook-style repo

Verdict

Worth a look if you’re maintaining legacy TensorFlow 1.x code or prefer learning by pattern-matching against familiar APIs. Skip it if you need modern TF2/Keras guidance or want production-ready examples—this is a phrasebook, not a framework.