The Hitchhikers Guide to Tensorflow
summary
Starts from tensors and backprop but focuses on practical pipelines: train on GPUs, reuse pretrained nets, and serve models in production. Demonstrates Inception v3 → 2048‑D features and using them for image similarity and transfer learning rather than full retraining, cites the large ImageNet corpus (ImageNet ~14M images) as the teacher, shows Keras simplifying model construction and training (Keras one‑line fit), explains TensorFlow Serving (and GPU/NVIDIA variants) with Kubernetes deployment and mentions eager execution is useful for debugging but not yet production ready.
AI-written summary · about
chapters (22)
- 0:00Intro
- 1:48Machine Translation
- 2:33Voice Basic
- 4:02Slides
- 4:29Draw Easy
- 4:49Hardcore
- 5:26Simple
- 5:58Read the code
- 7:13Pain points
- 7:58Intuition
- 9:38Formulas
- 13:00Optimization
- 16:58Interface
- 18:45Speaker Picker
- 21:11Cost Function
- 22:03Model
- 27:07How to use Tensorflow
- 30:02Tensorflow Libraries
- 32:10Saving Results
- 34:39Modular Approach
- 39:40Mobile Support
- 42:18Conclusion
related talks
38mThe Hitchhiker's Guide to Tensorflow: from Tensors to running models in production
44mTensorflow, Tensorflow, What is it that you see?
55mTensorflow: Learning Machines
41mTensorFlow ever higher
41mLearning TensorFlow
43mDeep learning beyond the learning
01h 18mPutting Tensorflow Models into Production
07mWTT: What the Tensor?