May 08, 2019 17:00:00

Public beta version of Google's machine learning machine 'Cloud TPU Pod' is available

Google is a machine 'TPU Pod' which was developed in order to speed up the machine learning, is a super computer that was in a cloud-based '

Cloud TPU Pod is'. At Google's developer event ' Google I / O 2019, ' Google announced that it will start offering beta versions of ' Cloud TPU v2 Pod ' and ' Cloud TPU v3 Pod ', and machine learning researchers and engineers will be faster It appeals that it will be possible to train machine learning.

Google's scalable supercomputers for machine learning, Cloud TPU Pods, now public available in beta | Google Cloud Blog
https://cloud.google.com/blog/products/ai-machine-learning/googles-scalable-supercomputers-for-machine-learning-cloud-tpu-pods-are-now-publicly-available-in-beta

In machine learning, a large amount of data is used to train a model, which requires a large amount of arithmetic processing, and training takes a long time even using a powerful machine. In order to solve this problem, Google has developed its own custom silicon chip ' TPU ' for machine learning, and Google Cloud Platform (GPC) provides Cloud TPU, a service for renting TPU systems on the cloud. You are

More than 1000 such TPU chips connected in a data center network is TPU Pod. Cloud TPU v2 Pod and Cloud TPU v3 Pod are cloud-based computers using second-generation TPU and third-generation TPU, respectively.

Cloud TPU v3 Pod using 3rd generation processor 'TPU 3.0 ' is liquid-cooled to achieve high performance. With Cloud TPU v3 Pod, training of ' ResNet-50 ' using ImageNet , which is a data set of Stanford University, is completed in just 2 minutes . While some custom silicon chips can perform only a single function, TPU can be fully programmed, so Cloud TPU Pod can be used to train a wide range of machine learning models.

Cloud TPU Pod can also be used in a small section called 'slice'. Google recommends that the machine learning team develop the first model on an individual Cloud TPU, and later use a larger Cloud TPU Pod slice when scaling up the training.

If you want to apply to Google for Cloud TPU Pod and Slice, you need to contact Google Cloud sales representative from the inquiry form .