Why TensorFlow eager execution matters
summary
Demonstrates how TensorFlow Eager Execution (with TF.Keras) replaces the static graph workarounds by letting code run step‑by‑step so GANs and VAEs are written as ordinary Python loops, using gradient tape for per‑step gradients and custom models for reusable encoder/decoder or down/up‑sampling blocks. Shows a GAN trained on Google QuickDraw penguins that looks poor after one epoch but produces convincing penguins after ~80 epochs, enables printf debugging of tensors and integrates with TensorFlow Probability.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 1:23What is deep learning
- 6:33Custom models
- 8:07Execution
- 12:40Modular code
- 15:45Interleaving logic
- 17:43Documentation
- 18:12Tensorflow Probability
- 19:01Questions
related talks
24mA Journey through Generative AI's Frontier Python Libraries
41mLearning TensorFlow
41mTensorFlow ever higher
47mSwifty TensorFlow
32mGetting Started with TensorFlow 2.0
44mTensorflow, Tensorflow, What is it that you see?
42mHigh Performance Distributed Tensorflow
36mHigh-Performance Distributed Tensorflow