October 2019
Intermediate to advanced
520 pages
13h 5m
English
In this section, we will explain what TPUs are. They are Google's custom-developed, application-specific integrated circuits (ASICs), which are used to speed up ML workloads.
They enhance the performance of linear algebra computation, which is used heavily in ML applications. TPUs facilitate a reduction in model training time from weeks to hours. In the following photograph, we can see a rack of servers using TPUs, which was presented at the Google Next conference:

As indicated by Google, the following factors make your model suitable for being trained with TPUs:
Read now
Unlock full access