Introduction
This document provides the KAIL installation guide, adaptation guide, operator description, and sample code to help users quickly get familiar with KAIL.
The Kunpeng Artificial Intelligence Library (KAIL) is a high-performance AI operator library optimized by Huawei for the Kunpeng platform. It consists of a deep neural network library and an extension library that contains the softmax and random_choice operators.
AI Library Overview
KAIL provides high-performance AI operators optimized for the Kunpeng platform, which are implemented by C, C++, and assembly languages. Table 1 describes the composition of KAIL.
No. |
Library |
Description |
Application |
|---|---|---|---|
1 |
KAIL_DNN |
A deep neural network library that contains AI operators optimized based on the microarchitecture of the Kunpeng processor and software optimization methods. It can be integrated into open source oneDNN as an operator library plugin. |
Suitable for various machine learning applications, including image classification, object detection, and speech recognition. It can be integrated with various deep learning frameworks, such as TensorFlow and PyTorch. |
2 |
KAIL_DNN_EXT |
A deep neural network extension library that contains the softmax and random_choice operators. They are encapsulated as Python interfaces. |
Suitable for scenarios including multi-classification and random selection. It can be integrated with various deep learning frameworks, such as TensorFlow and PyTorch. |
KAIL is available only for Kunpeng processors, where:
KAIL_DNN applies to new Kunpeng 920 processor models and supports SVE instructions (256-bit width and 512-bit width) and SME instructions (512-bit width).
KAIL_DNN_EXT applies to Kunpeng 920 processors and supports NEON instructions (128-bit width).
To achieve the optimal performance, KAIL interfaces do not verify all input parameters. The validity of input parameters is ensured by the service that calls the interfaces.
Application Scenarios
KAIL is mainly used in the following scenarios:
- Deep learning acceleration: image classification, object detection and segmentation, natural language processing, recommendation systems
- HPC: meteorology, life sciences, manufacturing, education and scientific research
- Big data: machine learning algorithms