Tensor2Tensor - TensorFlow at O’Reilly AI Conference, San Francisco '18

Explore the powerful Tensor2Tensor library in this 39-minute conference talk from the O'Reilly AI Conference in San Francisco. Dive into the world of deep learning models and datasets, learning how to create state-of-the-art models for various machine learning applications such as translation, parsing, and image captioning. Discover how Tensor2Tensor accelerates the exploration of new ideas in the field. Gain insights into machine translation, speech recognition, and image transformation using Transformer models. Understand the benefits of MultiGPU and Cloud TPU Pod for training. Learn about tuning hyperparameters, importing data, and working with problem classes and models. Explore the research subdirectory and get a glimpse of future developments in mesh tensor flow and image generation. By the end of this talk, you'll have a comprehensive understanding of Tensor2Tensor's capabilities and how to leverage them in your machine learning projects.

Introduction
Motivation
Machine Translation
Transformer
TensorFlow
Tensor2Tensor
Training
Open Source
Collaboration
Papers
How does it work
Translation
Speech Recognition
Transformer Models
Image Transformer
MultiGPU
Cloud TPU Pod
Tuning hyper parameters
Mesh tensor flow
Importing data
Problem class
Models
Subclasses
Research subdirectory
Looking Ahead
Mesh
Image Generation
Slice Back
Build Every Tensor
Model Layout
Model Transformer
GitHub