Standardizing on a single N-dimensional array API for Python |
![]() |
Ralf Gommers, Director @ Quansight Labs
Numerical computing and deep learning libraries for Python all offer array (or tensor) data structures and associated compute functionality with similar APIs. There are many subtle differences however, making it hard for users to migrate from one library to another, or for library authors to write code that supports multiple array libraries. The Consortium for Python Data API Standards (https://data-apis.org/) just released a first version of its array API standard - which aims to address these issues - for community review. In this talk, we will start with an overview of the full API surface, and then focus on some of the key choices made, such as dtype casting rules and which mutability features are supported. Finally we will look at how this new API aligns with the current MXNet API. |