Tensorflow Hub - DevFest Nairobi 2018

Tensorflow Hub is being introduced by Google with a focus on discovery, and consumption of reusable parts of ML models. This presentation will give a subtle introduction to the library, and the reusable modules one can leverage presently

Transcript

2. Tensorflow Hub Frontier for Reusable ML Modules Chris Barsolai Organizer, Nairobi AI @chrisbarso GDG DevFest Nairobi 2018
7. https://toolbox.google.com/datasetsearch
8. TensorFlow Hub Frontier for Reusable ML Modules Chris Barsolai @chrisbarso
9. Make machines intelligent. Improve people’s lives.
10. • Make solar energy affordable • Provide energy from fusion • Develop carbon sequestration methods • Manage the nitrogen cycle • Provide access to clean water • Restore & improve urban infrastructure • Advance health informatics • Engineer better medicines • Reverse-engineer the brain • Prevent nuclear terror • Secure cyberspace • Enhance virtual reality • Advance personalized learning • Engineer the tools for scientific discovery www.engineeringchallenges.org/challenges.aspx U.S. NAE Grand Engineering Challenges for 21st Century
11. Deep Learning for Image-Based Cassava Disease Detection
12. Half screen photo slide if text is necessary Using Tensorflow to keep farmers happy and cows healthy Source: https://www.blog.google/technology/ai/using-tensorflow-keep-farmers-happy-and-cows-healthy/
13. AutoML: Automated machine learning (“learning to learn”)
14. Current: Solution = ML expertise + data + computation
15. Current: Solution = ML expertise + data + computation Can we turn this into: Solution = data + 100X computation ???
16. Neural Architecture Search to find a model Controller: proposes ML models Train & evaluate models 20K times Iterate to find the most accurate model
18. Inception-ResNet-v2 computational cost Accuracy (precision @1) accuracy AutoML outperforms handcrafted models Learning Transferable Architectures for Scalable Image Recognition, Zoph et al. 2017, https://arxiv.org/abs/1707.07012
19. Inception-ResNet-v2 Years of effort by top ML researchers in the world computational cost Accuracy (precision @1) accuracy Learning Transferable Architectures for Scalable Image Recognition, Zoph et al. 2017, https://arxiv.org/abs/1707.07012 AutoML outperforms handcrafted models
20. Learning Transferable Architectures for Scalable Image Recognition, Zoph et al. 2017, https://arxiv.org/abs/1707.07012 computational cost Accuracy (precision @1) accuracy AutoML outperforms handcrafted models
21. cloud.google.com/automl
22. TensorFlow Hub The Next Frontier in AI Collaboration Chris Barsolai @chrisbarso
23. “Doing ML in production is hard.” -Everyone who has ever tried
24. Because, in addition to the actual ML... ML Code
25. ...you have to worry about so much more. Configuration Data Collection Data Verification Feature Extraction Process Management Tools Analysis Tools Machine Resource Management Serving Infrastructure Monitoring Source: Sculley et al.: Hidden Technical Debt in Machine Learning Systems ML Code
26. Typical ML Pipeline batch processing During training “online” processing During serving data request
27. 2005, no source control 2010, source control & C.I. … but not for ML 2018, great tools and practices … some still to be defined Story time ML Software
28. Repositories: Shared Code Repositories
29. TensorFlow Hub Tensorflow Hub: Shared ML
30. Data Algorithm Expertise Compute Ingredients of ML
31. Tensorflow Hub: Shared ML Module TensorFlow Hub Model
32. A repository of pre-trained model components, packaged for one-line reuse Easiest way to share and use best approaches for your task tensorflow.org/hub Module
33. Modules contain pretrained weights and graphs
34. Modules are composable, reusable, retrainable
35. Image Retraining
36. Model structure
38. # Download and use NASNet feature vector module. module = hub.Module( "https://tfhub.dev/google/imagenet/nasnet_large/feature_vector/1") features = module(my_images) logits = tf.layers.dense(features, NUM_CLASSES) probabilities = tf.nn.softmax(logits) 62,000+ GPU hours!
39. # Download and use NASNet feature vector module. module = hub.Module( "https://tfhub.dev/google/imagenet/nasnet_large/feature_vector/1", trainable=True, tags={“train”}) features = module(my_images) logits = tf.layers.dense(features, NUM_CLASSES) probabilities = tf.nn.softmax(logits)
40. • NASNet-A mobile, large • PNASNet-5 large • MobileNet V2 (all sizes) • Inception V1, V2, V3 • Inception-ResNet V2 • ResNet V1 & V2 50/101/152 • MobileNet V1 (all sizes) Available image modules
41. Text Classification
42. Sentence embeddings “The quick brown fox”
43. Sentence embeddings “DevFest Nairobi is super awesome so far”
44. Available text modules • Neural network language model (en, jp, de, es) • Word2vec trained on Wikipedia • ELMo (Embeddings from Language Models)
45. From researchers to you
46. Universal Sentence Encoder embed = hub.Module( "https://tfhub.dev/google/universal-sentence-encoder/1") messages = [ “The food was great, and USIU rocks”, “I didn’t understand a thing today” ….. “Where is jollof rice?” ] print session.run(embeddings) Check out the semantic similarity colab at https://alpha.tfhub.dev/google/universal-sentence-encoder/2
47. # Use pre-trained universal sentence encoder to build text vector column. review = hub.text_embedding_column( "review", "https://tfhub.dev/google/universal-sentence-encoder/1") features = { "review": np.array(["an arugula masterpiece", "inedible shoe leather", ...]) } labels = np.array([[1], [0], ...]) input_fn = tf.estimator.input.numpy_input_fn(features, labels, shuffle=True) estimator = tf.estimator.DNNClassifier(hidden_units, [review]) estimator.train(input_fn, max_steps=100)
48. # Use pre-trained universal sentence encoder to build text vector column. review = hub.text_embedding_column( "review", "https://tfhub.dev/google/universal-sentence-encoder/1", trainable=True) features = { "review": np.array(["an arugula masterpiece", "inedible shoe leather", ...]) } labels = np.array([[1], [0], ...]) input_fn = tf.estimator.input.numpy_input_fn(features, labels, shuffle=True) estimator = tf.estimator.DNNClassifier(hidden_units, [review]) estimator.train(input_fn, max_steps=100)
49. # Use pre-trained universal sentence encoder to build text vector column. review = hub.text_embedding_column( "review", "https://tfhub.dev/google/universal-sentence-encoder/1") features = { "review": np.array(["an arugula masterpiece", "inedible shoe leather", ...]) } labels = np.array([[1], [0], ...]) input_fn = tf.estimator.input.numpy_input_fn(features, labels, shuffle=True) estimator = tf.estimator.DNNClassifier(hidden_units, [review]) estimator.train(input_fn, max_steps=100)
51. More modules • Progressive GAN • Google Landmarks Deep Local Features (DELF) • Audio, video, and more coming...
52. New Web Experience
53. Try it! tensorflow.org/hub Build notebooks Share modules #tfhub
54. Blog blog.tensorflow.org
56. One URL to get involved tensorflow.org/community
58. https://meetup.com/NairobiAI To Join: