PyTorch* Optimizations from Intel

Speed Up AI from Research to Production Deployment

  

Maximize PyTorch Performance on Intel® Hardware

PyTorch* is an AI and machine learning framework popular for both research and production usage. This open source library is often used for deep learning applications whose compute-intensive training and inference test the limits of available hardware resources.

Intel contributes optimizations and features directly to open source PyTorch. Starting with PyTorch 2.5, Intel GPU (XPU) support is natively available in stock PyTorch.

With Intel hardware and stock PyTorch, you can:

  • Take advantage of Intel CPU and GPU optimizations contributed upstream by Intel engineers.
  • Use native torch.xpu APIs to run workloads on Intel Data Center GPU Max Series, Intel Arc GPUs, and Intel Core Ultra processors with Arc graphics.
  • Use torch.compile, AMP (FP16/BF16), and eager mode—all supported natively.

Intel works closely with the open source PyTorch project to contribute optimizations for Intel hardware directly upstream. These optimizations are part of the end-to-end suite of Intel® AI and machine learning development tools and resources.

Prerequisites

Install the Intel GPU driver and, if building PyTorch from source, Intel® Deep Learning Essentials.

PyTorch Prerequisites for Intel GPUs

Get PyTorch on Intel GPU

Get Started with PyTorch on Intel GPUs


Features

Open Source PyTorch Powered by Optimizations from Intel

  • Get the best PyTorch training and inference performance on Intel CPU or GPU hardware through open source contributions from Intel.
  • Take advantage of Intel® Deep Learning Boost, Intel® Advanced Vector Extensions (Intel® AVX-512), and Intel® Advanced Matrix Extensions (Intel® AMX) instruction set features to parallelize and accelerate PyTorch workloads.
  • Perform distributed training with oneAPI Collective Communications Library (oneCCL) bindings for PyTorch.

Native Intel GPU Support in PyTorch

  • Use torch.xpu APIs directly in stock PyTorch.
  • AMP (FP32, BF16, FP16), eager mode, and torch.compile are all supported natively.
  • Install via pip install torch --index-url https://download.pytorch.org/whl/xpu.

Optimized Deployment with OpenVINO™ Toolkit

  • Import your PyTorch model into OpenVINO Runtime to compress model size and increase inference speed.
  • Instantly target Intel CPUs, GPUs (integrated or discrete), or NPUs.
  • Deploy with OpenVINO model server for optimized inference in microservice applications, container-based, or cloud environments. Scale using the same architecture API as KServe for inference execution and inference service provided in gRPC* or REST.

Access the latest AI benchmark performance data for PyTorch and OpenVINO toolkit when running on data center products from Intel. Performance Data.

Demonstrations

PyTorch and DINOv2 for Multi-label Plant Species Classification

Perform transfer learning with self-supervised vision transformers (DINOv2) for multi-label plant species classification and a dataset of 1.4 million images that use PyTorch Lightning*.

Watch

Use PyTorch for Monocular Depth Estimation

Learn how to use a model based on Hugging Face Transformers to produce a clipped image with background clutter removed, ultimately creating a depth estimate from a single image.

Watch

Case Studies

L&T Technology Services Enhances Chest Radiology Outcomes

Chest-rAI* is a deep learning algorithm developed by L&T Technology Services (LTTS) to detect and isolate abnormalities in chest X-ray imagery. LTTS adopted the AI Tools and OpenVINO toolkit, reducing inference time by 46% and reducing their product development time from eight weeks to two weeks.

Learn More

HippoScreen Improves AI Performance by 2.4x

The Taiwan-based neurotechnology startup used tools and frameworks in the Intel® oneAPI Base Toolkit and AI Tools to improve the efficiency and training times of deep learning models used in its Brain Waves AI system.

Learn More

 

 

Specifications

Processors:

  • Intel® Xeon® processor
  • Intel® Core™ processor
  • Intel Core Ultra Processors with Intel® Arc™ graphics
  • Intel Arc GPUs
  • Intel® Data Center GPU Max Series

Operating systems:

  • Linux*
  • Windows* (Intel GPU XPU support available from PyTorch 2.5+)

Languages:

  • Python
  • C++

 

Deploy PyTorch models to a variety of devices and operating systems with OpenVINO™ toolkit.

 

Get Help

Your success is our success. Access these support resources when you need assistance.