NumPy 1.22, a Python library for scientific computing, is now available, which is focused on working with multidimensional arrays and matrices, and also provides a large collection of functions with the implementation of various algorithms related to the use of matrices. NumPy is one of the most popular libraries used for scientific computing. The project code is written in Python with C optimizations and distributed under the BSD license.
In the new version:
- Completed work on defining annotations for the main namespace.
- A preliminary version of the Array API has been proposed that conforms to the Python Array API standard and is implemented in a separate namespace. The new API is aimed at preparing standard functions for working with arrays, which can also be used in applications based on other libraries, such as CuPy and JAX.
- The DLPack backend has been implemented, providing support for the format of the same name for exchanging the contents of arrays (tensors) between different frameworks.
- Added a set of methods with the implementation of functions related to the concepts of quantile and percentile.
- Added a new custom memory manager (numpy-allocator).
- Work continued on optimizing features and platforms using SIMD vector instructions.
- Support for Python 3.7 has ended, it requires Python 3.8-3.10 to work.
Source: opennet.ru