Slide 1

Slide 1 text

@Vikram_Tiwari @gdgvancouver Slides by: @nitya #TFDevSummit #gdgcv TensorFlow Dev Summit 2018 (Recap)

Slide 2

Slide 2 text

What is TensorFlow? https://www.tensorflow.org TensorFlow™ is an open source software library for high performance numerical computation. Its flexible architecture allows easy deployment of computation across a variety of platforms (CPUs, GPUs, TPUs), and from desktops to clusters of servers to mobile and embedded systems

Slide 3

Slide 3 text

1-day Developer Summit held in Mountain View on March 30, 2018. Brings together ML/AI researchers, developers, practitioners & hobbyists. Signature event for TensorFlow news & updates from Google Brain (research) & TensorFlow (dev) teams What is the TensorFlow DevSummit? https://www.tensorflow.org/dev-summit /

Slide 4

Slide 4 text

● Keynote (summary of announcements)) ● Tf.data (input pipelining) ● Eager Execution (imperative vs. graphs) ● TensorFlow.js (training & inference in browser) ● TensorFlow Lite (for mobile & embedded devices) ● TensorFlow Hub (package & share models) ● Debugging TensorFlow (TensorFlow Board UI) ● Project Magenta (ML for art & music) Were the talks recorded? YouTube Playlist http://bit.ly/tfds18-playlist

Slide 5

Slide 5 text

Highlights in a nutshell Read the Medium BlogPost Make it easier to use (Eager Execution, TensorFlow Hub, debugging UI) More languages (JavaScript, Swift) and platforms (TF Lite) New app domains (astronomy, agriculture, air traffic control., music & art) New community resources (blog, youtube, mail lists, SIGs) New ways to learn (Machine Learning Crash Course, Colab Notebooks)

Slide 6

Slide 6 text

https://www.tensorflow.org/dev-summit / Keynote New application domains, New features, resources, tools Performance improvements “Help People’s Lives”

Slide 7

Slide 7 text

tf.data Fast (extract in parallel, fuse, prefetch), Flexible (generators, file formats), Easy-to-use (standard recipes, helper fns) Input Pipelines https://www.tensorflow.org/dev-summit /

Slide 8

Slide 8 text

https://www.tensorflow.org/dev-summit / Eager Execution Imperative vs. graph Dynamic models Faster iterations Profiling & Debugging “Pythonic” familiarity

Slide 9

Slide 9 text

https://www.tensorflow.org/dev-summit / tf.hub Shoulder of Giants Modules Composable, reusable and retrainable

Slide 10

Slide 10 text

https://www.tensorflow.org/dev-summit / ML in browser DeepLearn.js => tf.js Sensors & interactive Training & inferences Node.js coming soon

Slide 11

Slide 11 text

https://www.tensorflow.org/dev-summit / Magenta Generative Models Research

Slide 12

Slide 12 text

Resources TensorFlow Blog http://blog.tensorflow.org/ TensorFlow YouTube http://youtube.com/tensorflow TensorFlow Community: https://www.tensorflow.org/community/ TensorFlow Hub https://www.tensorflow.org/hub/ TensorFlow DevSummit: https://www.tensorflow.org/dev-summit/ These Slides: http://bit.ly/2018-tfdevsummit-recap

Slide 13

Slide 13 text

Key Takeaway Core focus is on democratizing usage & improving productivity in ML/AI with better tools, more languages & new resources for developers & researchers.

Slide 14

Slide 14 text

- ML Study Jam - Google IO Extended @gdgvancouver #TFDevSummit #gdgcv What’s Next?