ON THIS PAGE

Get Started
Cloud Resources
Mobile Resources
Edge Resources
Conference

Arm and PyTorch: Accelerating AI from Cloud to Edge

The collaboration between Arm and Meta empowers AI developers to build and deploy high-performance, energy-efficient generative AI (GenAI) applications across cloud, mobile, and edge using PyTorch and ExecuTorch.

Watch Video

Get Started

Get Started with PyTorch on Arm

Dive into the code repositories, model collections, and performance libraries to help you train, optimize, and deploy PyTorch on Arm today.

PyTorch on GitHub

Train and deploy AI models with PyTorch, optimized for high-performance on Arm from cloud to edge.

Access PyTorch Repo

ExecuTorch on GitHub

Build efficient AI applications for Arm-based edge and mobile devices using Meta’s lightweight runtime.

Access ExecuTorch Repo

Arm Kleidi Libraries

Arm Kleidi integrates with the most popular AI frameworks to enable optimum performance on Arm-based platforms out of the box.

Discover Arm Kleidi

Hugging Face Model Collections

Discover AI models you can deploy on Arm-based devices or cloud platforms, including those built with PyTorch and ExecuTorch.

Explore Model Collections

Cloud Resources

Resources for PyTorch on Server and Cloud

LEARNING PATHS
CODE-ALONGS & VIDEOS
BLOGS

Build and scale AI with PyTorch on Arm cloud

Visit our Developer Launchpad to learn how to deploy on Arm.

LLM chatbot with PyTorch on Arm servers

Quantize and run a Llama 3.1- based chatbot on Arm.

Run an NLP model from Hugging Face

Run PyTorch NLP models on Arm-based instances.

Train PyTorch models for digit classification

Train a feedforward neural network (FNN) and apply optimization.

Deploy a vision chatbot on Google Axion processors

Deploy a production-ready vision chatbot.

Measure and accelerate inference on Arm servers

Evaluate the performance of NLP and vision models.

Optimize MLOps with Arm-hosted GitHub runners

Automate training and testing steps in an ML workflow.

Use Keras Core with multiple backends

Use the Keras Core library with TensorFlow, PyTorch, and JAX.

Run Llama with PyTorch on Arm-based infrastructure

Create a browser-based LLM app that runs on Arm CPUs.

Accelerating LLMs on Arm-based Graviton processors

See how reducing memory footprint boosts performance.

GenAI with AWS Graviton processors | AWS OnAir

Learn when to use AWS Graviton for AI/ML workloads.

Accelerating DLRMv2 inference on Arm Neoverse

Accelerate recommendation model inference in the cloud.

AI performance boost with Arm Kleidi | PyTorch Conference 2024

How Kleidi enhances performance on AWS Graviton.

Optimizing CPU LLM inference | PyTorch Europe 2026

Build a high-performance CPU inference engine using vLLM.

Streamlining MLOps with GitHub Arm64 runners

Unlock new levels of efficiency and performance in your ML projects with GitHub on Arm.

Accelerated PyTorch inference with torch.compile

Use optimized torch.compile performance to improve PyTorch model inference.

Mobile Resources

Resources for Building Mobile AI Applications

LEARNING PATHS
CODE-ALONGS & VIDEOS
BLOGS

Fine-tuning neural graphics models with model gym

Train and deploy upscaling models for neural super sampling (NSS).

Build an Android chat app with ExecuTorch and Arm Kleidi

Build and run apps with Llama models using ExecuTorch on an Arm-based smartphone.

Build an Android chat app with Llama and Arm KleidiAI

Learn how to build and deploy a ChatAI application for Android mobile devices.

Edge AI acceleration demo | PyTorch Conference 2024

Learn how Arm KleidiAI, integrated with ExecuTorch, accelerates AI on mobile.

Faster GenAI at the edge with Arm and ExecuTorch

See how the Arm and ExecuTorch collaboration is improving edge AI performance.

Accelerate on-device inference with SME2 | PyTorch Europe 2026

Learn how ExecuTorch and Arm SME2 enable efficient CPU deployments.

ExecuTorch 1.0 is here with SME2 optimizations

With SME2 support via Arm KleidiAI, run ExecuTorch models natively on-device.

Arm neural technology in ExecuTorch 1.0

The ExecuTorch GA release includes backend support for Arm neural technology.

Bring GenAI to the masses with ExecuTorch and Arm KleidiAI

ExecuTorch 0.7 enables KleidiAI to automatically accelerate AI performance.

Interactive Image Segmentation Using SqueezeSAM

How SME2 enables up to 3.9x inference speedup for the Instagram Cutouts tool using Arm Kleidi, XNNPACK, ExecuTorch, and SqueezeSAM.

Automatic Speech Recognition with Conformer Models

Run high-quality speech and audio models at the edge using ExecuTorch and Arm Ethos-U85.

Edge Resources

Resources for Deploying Models at the Edge

LEARNING PATHS
CODE-ALONGS & VIDEOS
BLOGS

Run Llama 3 on a Raspberry Pi 5 using ExecuTorch

Learn techniques for running large language models (LLMs) in an embedded environment.

Introduction to TinyML on Arm using PyTorch and ExecuTorch

Set up and configure a TinyML development environment using Arm Corstone-320 FVP.

PyTorch and ExecuTorch rock-paper-scissors

Deploy TinyML models on Arm-based edge devices using PyTorch and ExecuTorch.

Visualize Arm Ethos-U NPU performance on Arm FVPs

Deploy TinyML models and visualize execution using the FVP graphical interface.

Bring PyTorch models to Arm Cortex-M | Arm AI tech talk

Develop a deep learning model with PyTorch and deploy it on Cortex-M-based systems.

Run ExecuTorch apps with silicon acceleration | PyTorch Europe 2026

How to build an end-to-end speech recognition app using ExecuTorch.

Edge-native AI agents with MCP | PyTorch Europe 2026

MCP, ExecuTorch, and Arm hardware unlock production-ready agentic AI at the edge.

How Arm Ethos-U and ExecuTorch 1.0 powers edge AI

ExecuTorch 1.0 brings PyTorch performance and portability to the most efficient Arm devices.

Deploying PyTorch models on Arm edge devices

Tutorial for deploying PyTorch on devices like Raspberry Pi or NVIDIA Jetson Nano.

Get started with PyTorch, ExecuTorch, and Ethos-U85

Check out the code and tools for seamlessly deploying AI models on Arm solutions.

Conference attendees chatting during a networking session.

Arm + ExecuTorch: Edge AI Challenge

Join Arm and Meta for a two-day, hands-on hackathon and turn your edge AI idea into a working prototype on Arm-powered hardware, with support from technical experts throughout the event.

Conference

Arm at PyTorch Conference 2026

Meet Arm engineers at booth P3 on October 20 – 21 in San Jose. Explore technical sessions, live demos, and research posters on PyTorch, ExecuTorch. Learn to build and optimize AI on Arm from cloud to edge.

Technical Sessions

Hardware-Aware AI: Building Agentic Systems from Cloud To Edge With Pytorch, Executorch, And Arm Device Connect

Kavya Sri Chennoju, Staff AI Engineer, Arm

Learn how PyTorch, ExecuTorch, vLLM, and Arm Device Connect can work together to connect AI applications with devices across cloud and edge.

View Session Details

From Prompt to Physical Action: A Live Hardware-Aware AI Demo with PyTorch, ExecuTorch, and Arm Device Connect

Kavya Sri Chennoju, Staff AI Engineer, Arm

See a live workflow where an AI application discovers devices, uses edge models and sensor data, and coordinates real hardware.

View Session Details

“What Do I Do Now?” Bridging the ExecuTorch Adoption Gap Through Developer Learning

Matt Cossins, Developer Ecosystem Manager, Arm

Join a discussion on practical ExecuTorch learning and adoption, from model preparation and optimization to deploying on-device applications.

View Session Details

From PyTorch to the Edge: Agentic Synthesis of Inference Runtimes for Heterogeneous Hardware

Thomas Cottenier, Senior Principal Engineer, Arm

Explore an agentic approach to generating target-specific inference runtimes from PyTorch models for heterogeneous edge hardware.

View Session Details

Poster Presentations

Accelerating Whisper Inference on Arm CPUs through PyTorch Ecosystem Optimizations

View Poster Details

Beyond the Learner: The CPU’s Role in Scaling PyTorch Multi-Agent Reinforcement Learning

View Poster Details

Bridging PyTorch Export to Edge GPU Backends: Enabling High-Performance grid_sample with ExecuTorch

View Poster Details

Bringing SmolLM2-135M to Resource-Constrained Arm Edge Devices with ExecuTorch

View Poster Details

From Model Integrity to Runtime Integrity: Trusted MCP Agents on Arm Edge Devices

View Poster Details

From PT2E to Arm/TOSA: Dynamic W8A8 Linear Quantization in ExecuTorch

View Poster Details

High-Performance CPU Inference in PyTorch: TorchInductor, oneDNN, and vLLM Ecosystem Integration

View Poster Details

Importance Aware Attention for Faster PyTorch Inference on Arm CPUs

View Poster Details

Improving PyTorch Efficiency on Arm with SVE-Accelerated Memory Primitives

View Poster Details

Join the Arm Developer Program

Build AI at scale on Arm with a personalized dashboard, early access tools, on-demand labs, courses and certifications, badging, and direct access to Arm experts.

Explore the Program

PyTorch News

Redefining Edge AI with ExecuTorch 1.0 GA Release

Seamlessly deploy PyTorch models across billions of Arm-based edge devices, unlocking faster on-device AI.

Read Blog

Key Enhancements in the PyTorch 2.9 Release

Learn about the key enhancements Arm has contributed to the latest PyTorch release.

Read Blog

PyTorch Growth: 2024 Year in Review

PyTorch has become the dominant open source deep learning framework for AI and ML with a 63% adoption rate.

Read Blog

ExecuTorch Beta: On-Device AI

Deploy LLMs at the edge with ExecuTorch, optimized for Arm-based devices.

Read Blog

GenAI at the Edge on Arm with ExecuTorch Beta

Integrate new quantized models into your applications for faster AI at the edge.

Read Blog

Arm Joins the PyTorch Foundation

PyTorch Foundation welcomes Arm as a premier member and part of the PyTorch Foundation Governing Board.

Read Blog