An overview of Cloud TPU resources.
Services that can access TPUs
Helps you decide which Cloud TPU service (Compute Engine, Google Kubernetes Engine, or AI Platform) best suits your needs.
TPU types and zones
Details the types of Cloud TPU and their zones of availability.
Storage options for Cloud TPU
Details the types of storage that can be used with Cloud TPU.
Internal IP address ranges
Setting up internal IP addresses for Cloud TPU nodes.
High-level performance tips for running TensorFlow programs on Cloud TPU.
Training on TPU Pods
An overview of training on Cloud TPU Pods.
Using TPU Estimator API
The differences between training a TensorFlow model with the standard TensorFlow Estimator and with the TPUEstimator API.
Processing large images with TPUs
Using spatial partitioning with Cloud TPUs and TPUEstimator for training with very large images and video.
An overview of Cloud TPU hardware and software architecture.
Using the bfloat16 format to store activations and gradients in memory.
Advanced Guide to Inception v3
An advanced view of the guide to running Inception v3 on Cloud TPU.
Except as otherwise noted, the content of this page is licensed under the Creative Commons Attribution 4.0 License, and code samples are licensed under the Apache 2.0 License. For details, see the Google Developers Site Policies. Java is a registered trademark of Oracle and/or its affiliates.