--- title: Introducing TorchNeuron: Native PyTorch backend for AWS... description: AWS and the PyTorch team announced TorchNeuron, an open source native PyTorch backend for AWS Trainium chips, launching as a public beta on Oct. 20 under... canonical: https://daily.dev/posts/introducing-torchneuron-native-pytorch-backend-for-aws-trainium-kdnpfsytl twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Introducing TorchNeuron: Native PyTorch backend for AWS Trainium | daily.dev og:description: AWS and the PyTorch team announced TorchNeuron, an open source native PyTorch backend for AWS Trainium chips, launching as a public beta on Oct. 20 under... og:url: https://daily.dev/posts/introducing-torchneuron-native-pytorch-backend-for-aws-trainium-kdnpfsytl og:image: https://api.daily.dev/og/posts/KDNpfsytL.png og:image:alt: Introducing TorchNeuron: Native PyTorch backend for AWS Trainium og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Introducing TorchNeuron: Native PyTorch backend for AWS Trainium **[PyTorch](https://daily.dev/sources/pytorch)** · 16 min read · 0 upvotes · 0 comments ## Summary AWS and the PyTorch team announced TorchNeuron, an open source native PyTorch backend for AWS Trainium chips, launching as a public beta on Oct. 20 under Apache 2.0. It registers Trainium through PyTorch's PrivateUse1 extension point so existing PyTorch code, libraries (Hugging Face Transformers v5, TorchTitan, vLLM), and distributed APIs (DDP, FSDP2, tensor/pipeline parallelism) run on Trainium with minimal changes. It supports eager execution (including a new adaptive eager mode), torch.compile for production, and the Neuron Kernel Interface (NKI) for custom kernels, plus Neuron Explorer for profiling. Customers including Cartesia, Deepgram, and Hugging Face shared early results, such as Deepgram reporting about 60% more audio seconds processed per dollar versus their GPU baseline. Testing covered 100k Hugging Face models and PyTorch's upstream test suites, with remaining gaps in operator coverage and some explicitly unsupported features (sparse tensors, dynamic shapes, etc.) to close in future releases. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### What is TorchNeuron and how do I run PyTorch code on AWS Trainium? TorchNeuron is an open source native PyTorch backend for AWS Trainium, released under Apache 2.0 as a public beta starting October 20, 2026. It registers Trainium through PyTorch's PrivateUse1 extension point, so existing code works by simply calling .to('neuron') or setting device_map='neuron', with no rewrites needed for torch.nn, torch.distributed, or torch.profiler. _Teams weighing Trainium against GPUs for training follow backend changes like this on daily.dev._ ### Does TorchNeuron support torch.compile and custom kernels on Trainium? Yes, TorchNeuron supports both eager execution (with an experimental adaptive eager mode for lower launch overhead) and torch.compile for production performance, plus the Neuron Kernel Interface (NKI) for writing custom operators. torch.compile uses TorchDynamo and AOTAutograd, lowering through Torch-MLIR and StableHLO into a Neuron Executable File Format (NEFF) that runs on Trainium2 and Trainium3. _Engineers optimizing inference pipelines track execution-mode tradeoffs like these on daily.dev._ ### What performance benefit did Deepgram see from moving to AWS Trainium with PyTorch? Deepgram reported that Trainium processes about 60% more input audio seconds per dollar compared to their existing GPU baselines, based on benchmarks run after a multi-month engagement with the AWS Generative AI Innovation Center to port their model to PyTorch Native on Trainium. This helps them deliver better models to customers at lower prices. _Developers comparing hardware costs for inference workloads follow real benchmark reports like this on daily.dev._ ## Similar posts on daily.dev - [An exclusive tour of Amazon’s Trainium lab, the chip that’s won over Anthropic, OpenAI, even Apple](https://daily.dev/posts/an-exclusive-tour-of-amazon-s-trainium-lab-the-chip-that-s-won-over-anthropic-openai-even-apple-lkzx6swqj) · TechCrunch · 2 upvotes · 0 comments - [Run cost-effective AI workloads on OpenShift with AWS Neuron Operator](https://daily.dev/posts/run-cost-effective-ai-workloads-on-openshift-with-aws-neuron-operator-gdgyktewi) · Red Hat Developer · 0 upvotes · 0 comments - [AWS Neuron announces support for Dynamic Resource Allocation with Amazon EKS](https://daily.dev/posts/aws-neuron-announces-support-for-dynamic-resource-allocation-with-amazon-eks-jfdhdf0iz) · AWS · 1 upvotes · 0 comments - [Uber joins Amazon’s Trainium roster with AWS expansion deal](https://daily.dev/posts/uber-joins-amazon-s-trainium-roster-with-aws-expansion-deal-vtnhjpvfx) · The Next Web · 1 upvotes · 0 comments --- Tags: [#machine-learning](https://daily.dev/tags/machine-learning), [#pytorch](https://daily.dev/tags/pytorch) [View this post on daily.dev](https://daily.dev/posts/introducing-torchneuron-native-pytorch-backend-for-aws-trainium-kdnpfsytl) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Introducing TorchNeuron: Native PyTorch backend for AWS Trainium","url":"https://daily.dev/posts/introducing-torchneuron-native-pytorch-backend-for-aws-trainium-kdnpfsytl","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/introducing-torchneuron-native-pytorch-backend-for-aws-trainium-kdnpfsytl"},"datePublished":"2026-10-19T12:50:01.071Z","dateModified":"2026-10-21T16:23:35.480Z","description":"AWS and the PyTorch team announced TorchNeuron, an open source native PyTorch backend for AWS Trainium chips, launching as a public beta on Oct. 20 under...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/6434ed11a23ede557cb8fba41d45acab?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/6434ed11a23ede557cb8fba41d45acab?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"PyTorch","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"PyTorch","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/5fdab3d75a834a97a52d433d9c7e5ff9","url":"https://daily.dev/sources/pytorch"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/introducing-torchneuron-native-pytorch-backend-for-aws-trainium-kdnpfsytl","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"machine-learning,pytorch","timeRequired":"PT16M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"PyTorch","item":"https://daily.dev/sources/pytorch"},{"@type":"ListItem","position":3,"name":"Introducing TorchNeuron: Native PyTorch backend for AWS Trainium"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/introducing-torchneuron-native-pytorch-backend-for-aws-trainium-kdnpfsytl#faq","mainEntity":[{"@type":"Question","name":"What is TorchNeuron and how do I run PyTorch code on AWS Trainium?","acceptedAnswer":{"@type":"Answer","text":"TorchNeuron is an open source native PyTorch backend for AWS Trainium, released under Apache 2.0 as a public beta starting October 20, 2026. It registers Trainium through PyTorch's PrivateUse1 extension point, so existing code works by simply calling .to('neuron') or setting device_map='neuron', with no rewrites needed for torch.nn, torch.distributed, or torch.profiler. Teams weighing Trainium against GPUs for training follow backend changes like this on daily.dev."}},{"@type":"Question","name":"Does TorchNeuron support torch.compile and custom kernels on Trainium?","acceptedAnswer":{"@type":"Answer","text":"Yes, TorchNeuron supports both eager execution (with an experimental adaptive eager mode for lower launch overhead) and torch.compile for production performance, plus the Neuron Kernel Interface (NKI) for writing custom operators. torch.compile uses TorchDynamo and AOTAutograd, lowering through Torch-MLIR and StableHLO into a Neuron Executable File Format (NEFF) that runs on Trainium2 and Trainium3. Engineers optimizing inference pipelines track execution-mode tradeoffs like these on daily.dev."}},{"@type":"Question","name":"What performance benefit did Deepgram see from moving to AWS Trainium with PyTorch?","acceptedAnswer":{"@type":"Answer","text":"Deepgram reported that Trainium processes about 60% more input audio seconds per dollar compared to their existing GPU baselines, based on benchmarks run after a multi-month engagement with the AWS Generative AI Innovation Center to port their model to PyTorch Native on Trainium. This helps them deliver better models to customers at lower prices. Developers comparing hardware costs for inference workloads follow real benchmark reports like this on daily.dev."}}]} ```