AWS First Steps
LanguageEnglish
New feature

AWS Announces TorchNeuron: Native PyTorch Support for AWS Trainium (Public Beta)

AWS has launched the public beta of TorchNeuron, an open source native PyTorch backend for AWS Trainium. It allows AI scientists and researchers to build and experiment with models natively on Trainium using PyTorch.

Today, AWS launched the public beta of TorchNeuron, an open-source native PyTorch backend for AWS Trainium. TorchNeuron lets AI scientists build and experiment models natively on Trainium using PyTorch. Developers can use eager execution, torch.compile, and standard PyTorch APIs while maintaining full PyTorch compatibility. It integrates with the Neuron Kernel Interface (NKI) for custom operators and performance optimizations on Trainium hardware. TorchNeuron is available open-source on GitHub for Amazon EC2 Trainium2 and Trainium3 instances.

Why it matters

AI developers and researchers can now use PyTorch natively on Trainium chips, making model development and experimentation easier.

Read the original AWS announcement

Related articles and terms

The Layered Architecture of AWS AI/ML Services - Flexibility Through the Three Tiers of SageMaker, Bedrock, and API ServicesThis article organizes AWS AI/ML services into three layers - SageMaker (full control), Bedrock (managed generative AI), and Rekognition/Comprehend/etc. (API-based) - and explains AWS's flexibility through comparisons with GCP Vertex AI and Azure OpenAI Service, including custom silicon integration.AWS Graviton and Custom Silicon Strategy - How In-House Chips Are Reshaping Cloud EconomicsExplore how AWS's Arm-based Graviton processors and AI-focused custom silicon (Inferentia and Trainium) are transforming cloud cost structures and performance, compared with Azure and GCP approaches.AWS Fault Injection SimulatorA service that safely runs chaos engineering experiments to verify system resilienceGPU-Based Machine Learning Training with AWS Batch - Cost-Efficient Large-Scale TrainingRun GPU training with your existing Docker containers, and cut costs by up to 90% using Spot Instances and checkpointing. Includes guidance on when to choose Batch over SageMaker.Amazon Elastic Inference[Discontinued] Service that attached fractional GPU accelerators to EC2 and SageMaker for inference. Inferentia (Inf1 / Inf2), GPU instances, and SageMaker AI inference are the replacements