The 2023.2 Intel® AI Analytics Toolkit (AI Kit) is now available for download. This release delivers new benefits for AI practitioners, performance engineers, and framework builders, including greater performance, enhanced productivity, and more portability. Highlights include:
Greater Performance
Speed up deep learning with new PyTorch* and TensorFlow* capabilities. Intel® Extension for PyTorch* now features PyTorch 2.0 compatibility, and adds experimental support for Intel® Arc™ A-Series Graphics cards. Meanwhile, Intel® Extension for TensorFlow* adds new CPU optimizations that streamline execution, memory allocation, and task scheduling.
Accelerate data preprocessing on CPU and GPU with pandas 2.0 support in the latest Intel® Distribution of Modin*, now combining the pandas 2.0 faster memory-efficient operations with the scaling benefits of parallel and distributed computing that are core to Modin.
Faster classic machine learning. Intel® Extension for Scikit-learn* now features CPU optimizations for extremely random trees and Intel® oneAPI Data Analytics Library (oneDAL) distributed algorithms. And Intel® Optimization for XGBoost* now supports Intel® Data Center GPU Max Series.
Enhanced Productivity
New model compression automation with Intel® Neural Compressor. The 2023.2 release adds new productivity features: “Neural Insights” provides analysis reports and visualization to help tune quantization; “Neural Solution” offers model optimization-as-a-service to automate accuracy-aware tuning across multiple nodes. The release also adds built-in validation functions for popular large language models (LLMs) and enhanced compatibility with mainstream AI frameworks, including PyTorch, TensorFlow, and ONNX Runtime*.
Boost prediction accuracy with new support for missing values when using the daal4py “Model Builder” feature to convert popular gradient boosting models to use optimized algorithmic building blocks in oneDAL for training and inference.
Streamlined package installation using the new Intel AI Tools web-based tools selector (beta), which can install individual components, popular bundles, or the entire release through multiple popular package managers.
More Portability
Expanded hardware choice using Intel® oneAPI Deep Neural Network Library (oneDNN). It adds new cross-platform support for ARM*, NVIDIA*, and AMD* processors. And it builds on its performance optimizations for Intel CPU and GPU. Finally, it adds developer productivity features such as simpler debugging and diagnostics, an experimental Graph compiler backend, and more.
Enhanced scaling efficiency for the cross-platform Intel® oneAPI Collective Communications Library (oneCCL), featuring new support for the Intel® Data Streaming Accelerator in 4th Gen Intel® Xeon® Scalable Processors.
To learn more about and download the software mentioned in this post, visit the AI Kit overview.
Technical marketing manager for Intel AI/ML product and solutions. Previous to Intel, I spent 7.5 years at MathWorks in technical marketing for the HDL product line, and 20 years at Cadence Design Systems in various technical and marketing roles for synthesis, simulation, and other verification technologies.