Getting Started with TensorFlow

Transcript

1. TensorFlow Tutorial Rebecca Murphy rebecca.murphy@ocado.com @rebecca_roisin TensorFlow Meetup Monday 21st March 2016
2. Talk Overview • TensorFlow overview • Programming Model • Mechanics of TensorFlow • Installation • Model Definition • Fitting • Checkpointing • TensorBoard visualisations • Why TensorFlow?
3. TensorFlow: Overview
4. What Is TensorFlow? • Google’s 2nd generation deep learning library • Simple API (Python, C++) for: • Describing Machine Learning models • Implementing Machine Learning algorithms
5. What Can We Do With TensorFlow? • Regression models • Neural networks • Deep learning: • Distributed representations • Convolutional Networks • Recurrent Neural Networks • LSTM Neural Networks
6. TensorFlow: Programming Model
7. TensorFlow: What Is a Tensor? • Tensor: n-dimensional array • Scalar: 0D Tensor • Vector: 1D Tensor • Matrix: 2D Tensor • Typed: • Int, double, complex, string
8. Tensor Flows • Tensor Flow computations: stateful dataflow graphs • Deep learning model = Directed graph • Node: (mathematical) operation • Edge: • Control dependencies • Data flow • Describe graph -> initialize -> execute (parts of ) graph
9. TensorFlow: Mechanics
10. Installing TensorFlow • Python API • Python 2.7 • Python 3.3+ • Setup instructions • pip install: • pip install --upgrade https://storage.googleapis. com/tensorflow/mac/tensorflow-0.7.1-cp27-none-any.whl • Docker: • docker run -it b.gcr.io/tensorflow/tensorflow
11. Mechanics of Learning • Define model • Load data • Feed data • Make predictions • Evaluate • Visualise
12. Example Code • Try-tf github repositories • Associated blogpost • Jason Baldridge @jasonbaldridge
13. Let’s get Started
14. Defining the Model
15. Model Definition: Key Features (1) • Tensor shapes are pre-defined: • Tensors support mathematical manipulation • Operations are nodes in the model graph
16. Model Definition: Key Features (2) • Built-in functions for common Deep Learning operations: • See Neural Network API for more • Gradient descent optimisation: • Variables store current state of model
17. Training the Model: Loading Data (1) • Load data into variables • Need to write custom functions to parse data
18. Training the Model: Loading Data (2)
19. Training the Model: Sessions • Model graph describes computations • Computations evaluated within a session: • Places graph onto CPU / GPU • Supplies methods to evaluate graph operations
20. The Feed Dict: Training the Model • Predefined placeholder tensors • Feed-dict supplies batch of data
21. Training the Model: Evaluation • Pre-defined evaluation nodes compare predicted and true labels: • Evaluate accuracy function within a session:
22. Checkpoints: Saving Models • Saver class allows model state to be stored and reloaded • Use checkpoints to periodically save the state of the model
23. • Saver class allows model state to be stored and reloaded • Restore a previously trained model Checkpoints: Loading Saved Models
24. Flags: Controlling Training • tf.app.flags: set command-line arguments • Wraps python gflags • tf.app.run() parses flags before calling main()
25. TensorBoard: Visualising Learning
26. TensorBoard: Basics • TensorFlow visualisation tool • View • Graph models • Training behaviour • Simple modifications to model code • Browser-based tool
27. TensorBoard: Annotations
28. TensorBoard: Scopes
29. TensorBoard: Saving Output • Set up summary and writer objects • Periodically run evaluation and store output: • tensorboard --logdir=try_tf_logs/
30. TensorBoard: Model Visualisation (1)
31. TensorBoard: Model Visualisation (2)
32. TensorBoard: Training Visualisation (1)
33. TensorBoard: Training Visualisation (2)
34. TensorFlow: Where Next?
35. Why Use TensorFlow: Great Examples • TensorFlow Tutorials • Handwriting generation from @hardmaru • Next letter prediction from @karpathy
36. Why Use TensorFlow: Active Community
37. TensorFlow: Future Developments • Improved memory usage in gradient calculations • JIT Compilation • Improved node execution scheduling • Support for parallelisation across many machines • Support for more languages (Java, Lua, Go, R …) • Source: TensorFlow Whitepaper
38. Thank You!
39. Questions?