🕷️ Crawler Inspector

URL Lookup

Direct Parameter Lookup

Raw Queries and Responses

1. Shard Calculation

Query:

Response:

Calculated Shard: 57 (from laksa068)

2. Crawled Status Check

Query:

curl -X POST \
  'http://laksa057.int.ahrefs:8124/' \
  -H 'Content-Type: text/plain' \
  -H 'X-ClickHouse-Database: crawler3' \
  -H 'Authorization: Basic YXBpOg==' \
  -d 'SELECT getAhrefsURLFromUnparsed(src_unparsed) AS found_url, ifNull(toUnixTimestamp(download_stamp), 0) AS crawl_time, ifNull(toUnixTimestamp(props_url_first_seen), 0) AS first_indexed_time, download_http_code AS http_code, src_unparsed AS src_unparsed, src_root_hash AS src_root_hash, history_drop_reason AS history_drop_reason, meta_title AS meta_title, meta_descriptions AS meta_descriptions, attrs_boilerpipe_text AS attrs_boilerpipe_text, attrs_markdown AS attrs_markdown, attrs_readable_markdown AS attrs_readable_markdown, meta_canonical AS meta_canonical FROM crawler3.page_info_local FINAL PREWHERE (src_root_hash, src_unparsed) IN ((getAhrefsRootHashFromUnparsed(getAhrefsUnparsedNoserviceFromURL(\'https://onnxruntime.ai/docs/install/\')), getAhrefsUnparsedNoserviceFromURL(\'https://onnxruntime.ai/docs/install/\'))) FORMAT JSONEachRow'

Response:

{"found_url":"https:\/\/onnxruntime.ai\/docs\/install\/","crawl_time":1775410940,"first_indexed_time":1631810453,"http_code":200,"src_unparsed":"ai,onnxruntime!\/docs\/install\/ s443","src_root_hash":"16223958384227938257","history_drop_reason":null,"meta_title":"Install ONNX Runtime | onnxruntime","meta_descriptions":["Instructions to install ONNX Runtime on your target platform in your environment"],"attrs_boilerpipe_text":"See the\ninstallation matrix\nfor recommended instructions for desired combinations of target operating system, hardware, accelerator, and language.\nDetails on OS versions, compilers, language versions, dependent libraries, etc can be found under\nCompatibility\n.\nContents\nRequirements\nCUDA and CuDNN\nPython Installs\nInstall ONNX Runtime CPU\nInstall nightly\nInstall ONNX Runtime GPU (DirectML) - Sustained Engineering Mode\nInstall nightly\nInstall ONNX Runtime GPU (CUDA or TensorRT)\nCUDA 12.x\nNightly for CUDA 13.x\nNightly for CUDA 12.x\nCUDA 11.x\nInstall ONNX Runtime QNN\nInstall nightly\nC#\/C\/C++\/WinML Installs\nInstall ONNX Runtime\nInstall ONNX Runtime CPU\nInstall ONNX Runtime GPU (CUDA 12.x)\nInstall ONNX Runtime GPU (CUDA 11.8)\nDirectML (sustained engineering - use WinML for new projects)\nWinML (recommended for Windows)\nInstall on web and mobile\nJavaScript Installs\nInstall ONNX Runtime Web (browsers)\nInstall ONNX Runtime Node.js binding (Node.js)\nInstall ONNX Runtime for React Native\nInstall on iOS\nC\/C++\nObjective-C\nCustom build\nInstall on Android\nJava\/Kotlin\nC\/C++\nCustom build\nInstall for On-Device Training\nOffline Phase - Prepare for Training\nTraining Phase - On-Device Training\nLarge Model Training\nInference install table for all languages\nTraining install table for all languages\nRequirements\nAll builds require the English language package with\nen_US.UTF-8\nlocale. On Linux, install\nlanguage-pack-en package\nby running\nlocale-gen en_US.UTF-8\nand\nupdate-locale LANG=en_US.UTF-8\nWindows builds require\nVisual C++ 2019 runtime\n. The latest version is recommended.\nCUDA and CuDNN\nFor ONNX Runtime GPU package, it is required to install\nCUDA\nand\ncuDNN\n. Check\nCUDA execution provider requirements\nfor compatible version of CUDA and cuDNN.\nZlib is required by cuDNN 9.x for Linux only (zlib is statically linked into the cuDNN 9.x Windows dynamic libraries), or cuDNN 8.x for Linux and Windows. Follow the\ncuDNN 8.9 installation guide\nto install zlib in Linux or Windows.\nIn Windows, the path of CUDA\nbin\nand cuDNN\nbin\ndirectories must be added to the\nPATH\nenvironment variable.\nIn Linux, the path of CUDA\nlib64\nand cuDNN\nlib\ndirectories must be added to the\nLD_LIBRARY_PATH\nenvironment variable.\nFor\nonnxruntime-gpu\npackage, it is possible to work with PyTorch without the need for manual installations of CUDA or cuDNN. Refer to\nCompatibility with PyTorch\nfor more information.\nPython Installs\nInstall ONNX Runtime CPU\npip\ninstall\nonnxruntime\nInstall nightly\npip\ninstall\ncoloredlogs flatbuffers numpy packaging protobuf sympy\npip\ninstall\n--pre\n--index-url\nhttps:\/\/aiinfra.pkgs.visualstudio.com\/PublicPackages\/_packaging\/ORT-Nightly\/pypi\/simple\/ onnxruntime\nInstall ONNX Runtime GPU (DirectML) - Sustained Engineering Mode\nNote\n: DirectML is in sustained engineering. For new Windows projects, consider\nWinML\ninstead.\npip\ninstall\nonnxruntime-directml\nInstall nightly\npip\ninstall\ncoloredlogs flatbuffers numpy packaging protobuf sympy\npip\ninstall\n--pre\n--index-url\nhttps:\/\/aiinfra.pkgs.visualstudio.com\/PublicPackages\/_packaging\/ORT-Nightly\/pypi\/simple\/ onnxruntime-directml\nInstall ONNX Runtime GPU (CUDA or TensorRT)\nCUDA 12.x\nThe default CUDA version for\nonnxruntime-gpu in pypi\nis 12.x since 1.19.0.\npip\ninstall\nonnxruntime-gpu\nFor previous versions, you can download here:\n1.18.1\n,\n1.18.0\nNightly for CUDA 13.x\npip\ninstall\ncoloredlogs flatbuffers numpy packaging protobuf sympy\npip\ninstall\n--pre\n--index-url\nhttps:\/\/aiinfra.pkgs.visualstudio.com\/PublicPackages\/_packaging\/ort-cuda-13-nightly\/pypi\/simple\/ onnxruntime-gpu\nNightly for CUDA 12.x\npip\ninstall\ncoloredlogs flatbuffers numpy packaging protobuf sympy\npip\ninstall\n--pre\n--index-url\nhttps:\/\/aiinfra.pkgs.visualstudio.com\/PublicPackages\/_packaging\/ORT-Nightly\/pypi\/simple\/ onnxruntime-gpu\nCUDA 11.x\nFor Cuda 11.x, please use the following instructions to install from\nORT Azure Devops Feed\nfor 1.19.2 or later.\npip\ninstall\ncoloredlogs flatbuffers numpy packaging protobuf sympy\npip\ninstall\nonnxruntime-gpu\n--index-url\nhttps:\/\/aiinfra.pkgs.visualstudio.com\/PublicPackages\/_packaging\/onnxruntime-cuda-11\/pypi\/simple\/\nFor previous versions, you can download here:\n1.18.1\n,\n1.18.0\nInstall ONNX Runtime QNN\npip\ninstall\nonnxruntime-qnn\nInstall nightly\npip\ninstall\ncoloredlogs flatbuffers numpy packaging protobuf sympy\npip\ninstall\n--pre\n--index-url\nhttps:\/\/aiinfra.pkgs.visualstudio.com\/PublicPackages\/_packaging\/ORT-Nightly\/pypi\/simple\/ onnxruntime-qnn\nC#\/C\/C++\/WinML Installs\nInstall ONNX Runtime\nInstall ONNX Runtime CPU\n# CPU\ndotnet add package Microsoft.ML.OnnxRuntime\nInstall ONNX Runtime GPU (CUDA 12.x)\nThe default CUDA version for ORT is 12.x\n# GPU\ndotnet add package Microsoft.ML.OnnxRuntime.Gpu\nInstall ONNX Runtime GPU (CUDA 11.8)\nProject Setup\nEnsure you have installed the latest version of the Azure Artifacts keyring from the its\nGithub Repo\n.\nAdd a nuget.config file to your project in the same directory as your .csproj file.\n<?xml version=\"1.0\" encoding=\"utf-8\"?>\n<configuration>\n<packageSources>\n<clear\/>\n<add\nkey=\n\"onnxruntime-cuda-11\"\nvalue=\n\"https:\/\/aiinfra.pkgs.visualstudio.com\/PublicPackages\/_packaging\/onnxruntime-cuda-11\/nuget\/v3\/index.json\"\n\/>\n<\/packageSources>\n<\/configuration>\nRestore packages\nRestore packages (using the interactive flag, which allows dotnet to prompt you for credentials)\ndotnet add package Microsoft.ML.OnnxRuntime.Gpu\nNote: You don’t need –interactive every time. dotnet will prompt you to add –interactive if it needs updated credentials.\nDirectML (sustained engineering - use WinML for new projects)\ndotnet add package Microsoft.ML.OnnxRuntime.DirectML\nNote\n: DirectML is in sustained engineering. For new Windows projects, use WinML instead:\nWinML (recommended for Windows)\ndotnet add package Microsoft.AI.MachineLearning\nInstall on web and mobile\nThe pre-built packages have full support for all ONNX opsets and operators.\nIf the pre-built package is too large, you can create a\ncustom build\n. A custom build can include just the opsets and operators in your model\/s to reduce the size.\nJavaScript Installs\nInstall ONNX Runtime Web (browsers)\n# install latest release version\nnpm\ninstall\nonnxruntime-web\n# install nightly build dev version\nnpm\ninstall\nonnxruntime-web@dev\nInstall ONNX Runtime Node.js binding (Node.js)\n# install latest release version\nnpm\ninstall\nonnxruntime-node\nInstall ONNX Runtime for React Native\n# install latest release version\nnpm\ninstall\nonnxruntime-react-native\nInstall on iOS\nIn your CocoaPods\nPodfile\n, add the\nonnxruntime-c\nor\nonnxruntime-objc\npod, depending on which API you want to use.\nC\/C++\nuse_frameworks!\npod\n'onnxruntime-c'\nObjective-C\nuse_frameworks!\npod\n'onnxruntime-objc'\nRun\npod install\n.\nCustom build\nRefer to the instructions for creating a\ncustom iOS package\n.\nInstall on Android\nJava\/Kotlin\nIn your Android Studio Project, make the following changes to:\nbuild.gradle (Project):\nrepositories\n{\nmavenCentral\n()\n}\nbuild.gradle (Module):\ndependencies\n{\nimplementation\n'com.microsoft.onnxruntime:onnxruntime-android:latest.release'\n}\nC\/C++\nDownload the\nonnxruntime-android\nAAR hosted at MavenCentral, change the file extension from\n.aar\nto\n.zip\n, and unzip it. Include the header files from the\nheaders\nfolder, and the relevant\nlibonnxruntime.so\ndynamic library from the\njni\nfolder in your NDK project.\nCustom build\nRefer to the instructions for creating a\ncustom Android package\n.\nInstall for On-Device Training\nUnless stated otherwise, the installation instructions in this section refer to pre-built packages designed to perform on-device training.\nIf the pre-built training package supports your model but is too large, you can create a\ncustom training build\n.\nOffline Phase - Prepare for Training\npython\n-m\npip\ninstall\ncerberus flatbuffers h5py numpy>\n=\n1.16.6 onnx packaging protobuf sympy setuptools>\n=\n41.4.0\npip\ninstall\n-i\nhttps:\/\/aiinfra.pkgs.visualstudio.com\/PublicPackages\/_packaging\/ORT\/pypi\/simple\/ onnxruntime-training-cpu\nTraining Phase - On-Device Training\nDevice\nLanguage\nPackageName\nInstallation Instructions\nWindows\nC, C++, C#\nMicrosoft.ML.OnnxRuntime.Training\ndotnet add package Microsoft.ML.OnnxRuntime.Training\nLinux\nC, C++\nonnxruntime-training-linux*.tgz\nDownload the\n*.tgz\nfile from\nhere\n.\nExtract it.\nMove and include the header files in the\ninclude\ndirectory.\nMove the\nlibonnxruntime.so\ndynamic library to a desired path and include it.\nPython\nonnxruntime-training\npip install onnxruntime-training\nAndroid\nC, C++\nonnxruntime-training-android\nDownload the\nonnxruntime-training-android (full package)\nAAR hosted at Maven Central.\nChange the file extension from\n.aar\nto\n.zip\n, and unzip it.\nInclude the header files from the\nheaders\nfolder.\nInclude the relevant\nlibonnxruntime.so\ndynamic library from the\njni\nfolder in your NDK project.\nJava\/Kotlin\nonnxruntime-training-android\nIn your Android Studio Project, make the following changes to:\nbuild.gradle (Project):\nrepositories { mavenCentral() }\nbuild.gradle (Module):\ndependencies { implementation 'com.microsoft.onnxruntime:onnxruntime-training-android:latest.release' }\niOS\nC, C++\nCocoaPods: onnxruntime-training-c\nIn your CocoaPods\nPodfile\n, add the\nonnxruntime-training-c\npod:\nuse_frameworks!\npod 'onnxruntime-training-c'\nRun\npod install\n.\nObjective-C\nCocoaPods: onnxruntime-training-objc\nIn your CocoaPods\nPodfile\n, add the\nonnxruntime-training-objc\npod:\nuse_frameworks!\npod 'onnxruntime-training-objc'\nRun\npod install\n.\nWeb\nJavaScript, TypeScript\nonnxruntime-web\nnpm install onnxruntime-web\nUse either\nimport * as ort from 'onnxruntime-web\/training';\nor\nconst ort = require('onnxruntime-web\/training');\nLarge Model Training\npip\ninstall\ntorch-ort\npython\n-m\ntorch_ort.configure\nNote\n: This installs the default version of the\ntorch-ort\nand\nonnxruntime-training\npackages that are mapped to specific versions of the CUDA libraries. Refer to the install options in\nonnxruntime.ai\n.\nInference install table for all languages\nThe table below lists the build variants available as officially supported packages. Others can be\nbuilt from source\nfrom each\nrelease branch\n.\nIn addition to general\nrequirements\n, please note additional requirements and dependencies in the table below:\n \nOfficial build\nNightly build\nReqs\nPython\nIf using pip, run\npip install --upgrade pip\nprior to downloading.\n \n \n \nCPU:\nonnxruntime\nonnxruntime (nightly)\n \n \nGPU (CUDA\/TensorRT) for CUDA 12.x:\nonnxruntime-gpu\nonnxruntime-gpu (nightly)\nView\n \nGPU (DirectML)\nsustained engineering\n:\nonnxruntime-directml\nonnxruntime-directml (nightly)\nView\n \nOpenVINO:\nintel\/onnxruntime\n-\nIntel managed\n \nView\n \nTensorRT (Jetson):\nJetson Zoo\n-\nNVIDIA managed\n \n \n \nAzure (Cloud):\nonnxruntime-azure\n \n \nC#\/C\/C++\nCPU:\nMicrosoft.ML.OnnxRuntime\nonnxruntime (nightly)\n \n \nGPU (CUDA\/TensorRT):\nMicrosoft.ML.OnnxRuntime.Gpu\nonnxruntime (nightly)\nView\n \nGPU (DirectML)\nsustained engineering\n:\nMicrosoft.ML.OnnxRuntime.DirectML\nonnxruntime (nightly)\nView\nWinML\nrecommended for Windows\nMicrosoft.AI.MachineLearning\nonnxruntime (nightly)\nView\nJava\nCPU:\ncom.microsoft.onnxruntime:onnxruntime\n \nView\n \nGPU (CUDA\/TensorRT):\ncom.microsoft.onnxruntime:onnxruntime_gpu\n \nView\nAndroid\ncom.microsoft.onnxruntime:onnxruntime-android\n \nView\niOS (C\/C++)\nCocoaPods:\nonnxruntime-c\n \nView\nObjective-C\nCocoaPods:\nonnxruntime-objc\n \nView\nReact Native\nonnxruntime-react-native\n(latest)\nonnxruntime-react-native (dev)\nView\nNode.js\nonnxruntime-node\n(latest)\nonnxruntime-node (dev)\nView\nWeb\nonnxruntime-web\n(latest)\nonnxruntime-web (dev)\nView\nNote: Nightly builds created from the main branch are available for testing newer changes between official releases. Please use these at your own risk. We strongly advise against deploying these to production workloads as support is limited for nightly builds.\nTraining install table for all languages\nRefer to the getting started with\nOptimized Training\npage for more fine-grained installation instructions.","attrs_markdown":"[Skip to main content](https:\/\/onnxruntime.ai\/docs\/install\/#main-content)\n\n- [ONNX Runtime](https:\/\/onnxruntime.ai\/docs\/)\n- [Install ONNX Runtime](https:\/\/onnxruntime.ai\/docs\/install\/)\n- [Get Started](https:\/\/onnxruntime.ai\/docs\/get-started\/)\n  - [Python](https:\/\/onnxruntime.ai\/docs\/get-started\/with-python.html)\n  - [C++](https:\/\/onnxruntime.ai\/docs\/get-started\/with-cpp.html)\n  - [C](https:\/\/onnxruntime.ai\/docs\/get-started\/with-c.html)\n  - [C\\#](https:\/\/onnxruntime.ai\/docs\/get-started\/with-csharp.html)\n  - [Java](https:\/\/onnxruntime.ai\/docs\/get-started\/with-java.html)\n  - [JavaScript](https:\/\/onnxruntime.ai\/docs\/get-started\/with-javascript\/)\n    - [Web](https:\/\/onnxruntime.ai\/docs\/get-started\/with-javascript\/web.html)\n    - [Node.js binding](https:\/\/onnxruntime.ai\/docs\/get-started\/with-javascript\/node.html)\n    - [React Native](https:\/\/onnxruntime.ai\/docs\/get-started\/with-javascript\/react-native.html)\n  - [Objective-C](https:\/\/onnxruntime.ai\/docs\/get-started\/with-obj-c.html)\n  - [Julia, Ruby and Rust APIs](https:\/\/onnxruntime.ai\/docs\/get-started\/community-projects.html)\n  - [Windows](https:\/\/onnxruntime.ai\/docs\/get-started\/with-windows.html)\n  - [Mobile](https:\/\/onnxruntime.ai\/docs\/get-started\/with-mobile.html)\n  - [On-Device Training](https:\/\/onnxruntime.ai\/docs\/get-started\/training-on-device.html)\n  - [Large Model Training](https:\/\/onnxruntime.ai\/docs\/get-started\/training-pytorch.html)\n- [Tutorials](https:\/\/onnxruntime.ai\/docs\/tutorials\/)\n  - [API Basics](https:\/\/onnxruntime.ai\/docs\/tutorials\/api-basics.html)\n  - [Accelerate PyTorch](https:\/\/onnxruntime.ai\/docs\/tutorials\/accelerate-pytorch\/)\n    - [PyTorch Inference](https:\/\/onnxruntime.ai\/docs\/tutorials\/accelerate-pytorch\/pytorch.html)\n    - [Inference on multiple targets](https:\/\/onnxruntime.ai\/docs\/tutorials\/accelerate-pytorch\/resnet-inferencing.html)\n    - [Accelerate PyTorch Training](https:\/\/onnxruntime.ai\/docs\/tutorials\/accelerate-pytorch\/ort-training.html)\n  - [Accelerate TensorFlow](https:\/\/onnxruntime.ai\/docs\/tutorials\/tensorflow.html)\n  - [Accelerate Hugging Face](https:\/\/onnxruntime.ai\/docs\/tutorials\/huggingface.html)\n  - [Deploy on AzureML](https:\/\/onnxruntime.ai\/docs\/tutorials\/azureml.html)\n  - [Deploy on mobile](https:\/\/onnxruntime.ai\/docs\/tutorials\/mobile\/)\n    - [Object detection and pose estimation with YOLOv8](https:\/\/onnxruntime.ai\/docs\/tutorials\/mobile\/pose-detection.html)\n    - [Mobile image recognition on Android](https:\/\/onnxruntime.ai\/docs\/tutorials\/mobile\/deploy-android.html)\n    - [Improve image resolution on mobile](https:\/\/onnxruntime.ai\/docs\/tutorials\/mobile\/superres.html)\n    - [Mobile objection detection on iOS](https:\/\/onnxruntime.ai\/docs\/tutorials\/mobile\/deploy-ios.html)\n    - [ORT Mobile Model Export Helpers](https:\/\/onnxruntime.ai\/docs\/tutorials\/mobile\/helpers\/)\n  - [Web](https:\/\/onnxruntime.ai\/docs\/tutorials\/web\/)\n    - [Build a web app with ONNX Runtime](https:\/\/onnxruntime.ai\/docs\/tutorials\/web\/build-web-app.html)\n    - [The 'env' Flags and Session Options](https:\/\/onnxruntime.ai\/docs\/tutorials\/web\/env-flags-and-session-options.html)\n    - [Using WebGPU](https:\/\/onnxruntime.ai\/docs\/tutorials\/web\/ep-webgpu.html)\n    - [Using WebNN](https:\/\/onnxruntime.ai\/docs\/tutorials\/web\/ep-webnn.html)\n    - [Working with Large Models](https:\/\/onnxruntime.ai\/docs\/tutorials\/web\/large-models.html)\n    - [Performance Diagnosis](https:\/\/onnxruntime.ai\/docs\/tutorials\/web\/performance-diagnosis.html)\n    - [Deploying ONNX Runtime Web](https:\/\/onnxruntime.ai\/docs\/tutorials\/web\/deploy.html)\n    - [Troubleshooting](https:\/\/onnxruntime.ai\/docs\/tutorials\/web\/trouble-shooting.html)\n    - [Classify images with ONNX Runtime and Next.js](https:\/\/onnxruntime.ai\/docs\/tutorials\/web\/classify-images-nextjs-github-template.html)\n    - [Custom Excel Functions for BERT Tasks in JavaScript](https:\/\/onnxruntime.ai\/docs\/tutorials\/web\/excel-addin-bert-js.html)\n  - [Deploy on IoT and edge](https:\/\/onnxruntime.ai\/docs\/tutorials\/iot-edge\/)\n    - [IoT Deployment on Raspberry Pi](https:\/\/onnxruntime.ai\/docs\/tutorials\/iot-edge\/rasp-pi-cv.html)\n  - [Deploy traditional ML](https:\/\/onnxruntime.ai\/docs\/tutorials\/traditional-ml.html)\n  - [Inference with C\\#](https:\/\/onnxruntime.ai\/docs\/tutorials\/csharp\/)\n    - [Basic C\\# Tutorial](https:\/\/onnxruntime.ai\/docs\/tutorials\/csharp\/basic_csharp.html)\n    - [Inference BERT NLP with C\\#](https:\/\/onnxruntime.ai\/docs\/tutorials\/csharp\/bert-nlp-csharp-console-app.html)\n    - [Configure CUDA for GPU with C\\#](https:\/\/onnxruntime.ai\/docs\/tutorials\/csharp\/csharp-gpu.html)\n    - [Image recognition with ResNet50v2 in C\\#](https:\/\/onnxruntime.ai\/docs\/tutorials\/csharp\/resnet50_csharp.html)\n    - [Stable Diffusion with C\\#](https:\/\/onnxruntime.ai\/docs\/tutorials\/csharp\/stable-diffusion-csharp.html)\n    - [Object detection in C\\# using OpenVINO](https:\/\/onnxruntime.ai\/docs\/tutorials\/csharp\/yolov3_object_detection_csharp.html)\n    - [Object detection with Faster RCNN in C\\#](https:\/\/onnxruntime.ai\/docs\/tutorials\/csharp\/fasterrcnn_csharp.html)\n  - [On-Device Training](https:\/\/onnxruntime.ai\/docs\/tutorials\/on-device-training\/)\n    - [Building an Android Application](https:\/\/onnxruntime.ai\/docs\/tutorials\/on-device-training\/android-app.html)\n    - [Building an iOS Application](https:\/\/onnxruntime.ai\/docs\/tutorials\/on-device-training\/ios-app.html)\n- [API Docs](https:\/\/onnxruntime.ai\/docs\/api\/)\n- [Build ONNX Runtime](https:\/\/onnxruntime.ai\/docs\/build\/)\n  - [Build for inferencing](https:\/\/onnxruntime.ai\/docs\/build\/inferencing.html)\n  - [Build for training](https:\/\/onnxruntime.ai\/docs\/build\/training.html)\n  - [Build with different EPs](https:\/\/onnxruntime.ai\/docs\/build\/eps.html)\n  - [Build for web](https:\/\/onnxruntime.ai\/docs\/build\/web.html)\n  - [Build for Android](https:\/\/onnxruntime.ai\/docs\/build\/android.html)\n  - [Build for iOS](https:\/\/onnxruntime.ai\/docs\/build\/ios.html)\n  - [Custom build](https:\/\/onnxruntime.ai\/docs\/build\/custom.html)\n- [Execution Providers](https:\/\/onnxruntime.ai\/docs\/execution-providers\/)\n  - [NVIDIA - CUDA](https:\/\/onnxruntime.ai\/docs\/execution-providers\/CUDA-ExecutionProvider.html)\n  - [NVIDIA - TensorRT](https:\/\/onnxruntime.ai\/docs\/execution-providers\/TensorRT-ExecutionProvider.html)\n  - [NVIDIA - TensorRT RTX](https:\/\/onnxruntime.ai\/docs\/execution-providers\/TensorRTRTX-ExecutionProvider.html)\n  - [Intel - OpenVINO™](https:\/\/onnxruntime.ai\/docs\/execution-providers\/OpenVINO-ExecutionProvider.html)\n  - [Intel - oneDNN](https:\/\/onnxruntime.ai\/docs\/execution-providers\/oneDNN-ExecutionProvider.html)\n  - [Windows - DirectML](https:\/\/onnxruntime.ai\/docs\/execution-providers\/DirectML-ExecutionProvider.html)\n  - [Qualcomm - QNN](https:\/\/onnxruntime.ai\/docs\/execution-providers\/QNN-ExecutionProvider.html)\n  - [Android - NNAPI](https:\/\/onnxruntime.ai\/docs\/execution-providers\/NNAPI-ExecutionProvider.html)\n  - [Apple - CoreML](https:\/\/onnxruntime.ai\/docs\/execution-providers\/CoreML-ExecutionProvider.html)\n  - [XNNPACK](https:\/\/onnxruntime.ai\/docs\/execution-providers\/Xnnpack-ExecutionProvider.html)\n  - [AMD - ROCm](https:\/\/onnxruntime.ai\/docs\/execution-providers\/ROCm-ExecutionProvider.html)\n  - [AMD - MIGraphX](https:\/\/onnxruntime.ai\/docs\/execution-providers\/MIGraphX-ExecutionProvider.html)\n  - [AMD - Vitis AI](https:\/\/onnxruntime.ai\/docs\/execution-providers\/Vitis-AI-ExecutionProvider.html)\n  - [Cloud - Azure](https:\/\/onnxruntime.ai\/docs\/execution-providers\/Azure-ExecutionProvider.html)\n  - [Community-maintained](https:\/\/onnxruntime.ai\/docs\/execution-providers\/community-maintained\/)\n    - [Arm - ACL](https:\/\/onnxruntime.ai\/docs\/execution-providers\/community-maintained\/ACL-ExecutionProvider.html)\n    - [Arm - Arm NN](https:\/\/onnxruntime.ai\/docs\/execution-providers\/community-maintained\/ArmNN-ExecutionProvider.html)\n    - [Apache - TVM](https:\/\/onnxruntime.ai\/docs\/execution-providers\/community-maintained\/TVM-ExecutionProvider.html)\n    - [Rockchip - RKNPU](https:\/\/onnxruntime.ai\/docs\/execution-providers\/community-maintained\/RKNPU-ExecutionProvider.html)\n    - [Huawei - CANN](https:\/\/onnxruntime.ai\/docs\/execution-providers\/community-maintained\/CANN-ExecutionProvider.html)\n  - [Add a new provider](https:\/\/onnxruntime.ai\/docs\/execution-providers\/add-execution-provider.html)\n  - [EP Context Design](https:\/\/onnxruntime.ai\/docs\/execution-providers\/EP-Context-Design.html)\n  - [Plugin Execution Provider Libraries](https:\/\/onnxruntime.ai\/docs\/execution-providers\/plugin-ep-libraries\/)\n    - [Usage](https:\/\/onnxruntime.ai\/docs\/execution-providers\/plugin-ep-libraries\/usage.html)\n    - [Development](https:\/\/onnxruntime.ai\/docs\/execution-providers\/plugin-ep-libraries\/development.html)\n    - [Testing](https:\/\/onnxruntime.ai\/docs\/execution-providers\/plugin-ep-libraries\/testing.html)\n    - [Packaging](https:\/\/onnxruntime.ai\/docs\/execution-providers\/plugin-ep-libraries\/packaging.html)\n- [Generate API (Preview)](https:\/\/onnxruntime.ai\/docs\/genai\/)\n  - [Tutorials](https:\/\/onnxruntime.ai\/docs\/genai\/tutorials\/)\n    - [Phi-3.5 vision tutorial](https:\/\/onnxruntime.ai\/docs\/genai\/tutorials\/phi3-v.html)\n    - [Phi-3 tutorial](https:\/\/onnxruntime.ai\/docs\/genai\/tutorials\/phi3-python.html)\n    - [Phi-2 tutorial](https:\/\/onnxruntime.ai\/docs\/genai\/tutorials\/phi2-python.html)\n    - [Run with LoRA adapters](https:\/\/onnxruntime.ai\/docs\/genai\/tutorials\/finetune.html)\n    - [DeepSeek-R1-Distill tutorial](https:\/\/onnxruntime.ai\/docs\/genai\/tutorials\/deepseek-python.html)\n    - [Run on Snapdragon devices](https:\/\/onnxruntime.ai\/docs\/genai\/tutorials\/snapdragon.html)\n  - [API docs](https:\/\/onnxruntime.ai\/docs\/genai\/api\/)\n    - [Python API](https:\/\/onnxruntime.ai\/docs\/genai\/api\/python.html)\n    - [C\\# API](https:\/\/onnxruntime.ai\/docs\/genai\/api\/csharp.html)\n    - [C API](https:\/\/onnxruntime.ai\/docs\/genai\/api\/c.html)\n    - [C++ API](https:\/\/onnxruntime.ai\/docs\/genai\/api\/cpp.html)\n    - [Java API](https:\/\/onnxruntime.ai\/docs\/genai\/api\/java.html)\n  - [How to](https:\/\/onnxruntime.ai\/docs\/genai\/howto\/)\n    - [Install](https:\/\/onnxruntime.ai\/docs\/genai\/howto\/install.html)\n    - [Build from source](https:\/\/onnxruntime.ai\/docs\/genai\/howto\/build-from-source.html)\n    - [Build models](https:\/\/onnxruntime.ai\/docs\/genai\/howto\/build-model.html)\n    - [Build models for Snapdragon](https:\/\/onnxruntime.ai\/docs\/genai\/howto\/build-models-for-snapdragon.html)\n    - [Troubleshoot](https:\/\/onnxruntime.ai\/docs\/genai\/howto\/troubleshoot.html)\n    - [Migrate](https:\/\/onnxruntime.ai\/docs\/genai\/howto\/migrate.html)\n    - [Past present share buffer](https:\/\/onnxruntime.ai\/docs\/genai\/howto\/past-present-share-buffer.html)\n  - [Reference](https:\/\/onnxruntime.ai\/docs\/genai\/reference\/)\n    - [Config reference](https:\/\/onnxruntime.ai\/docs\/genai\/reference\/config.html)\n    - [Adapter file spec](https:\/\/onnxruntime.ai\/docs\/genai\/reference\/adapter.html)\n- [Extensions](https:\/\/onnxruntime.ai\/docs\/extensions\/)\n  - [Add Operators](https:\/\/onnxruntime.ai\/docs\/extensions\/add-op.html)\n  - [Build](https:\/\/onnxruntime.ai\/docs\/extensions\/build.html)\n- [Performance](https:\/\/onnxruntime.ai\/docs\/performance\/)\n  - [Tune performance](https:\/\/onnxruntime.ai\/docs\/performance\/tune-performance\/)\n    - [Profiling tools](https:\/\/onnxruntime.ai\/docs\/performance\/tune-performance\/profiling-tools.html)\n    - [Logging & Tracing](https:\/\/onnxruntime.ai\/docs\/performance\/tune-performance\/logging_tracing.html)\n    - [Memory consumption](https:\/\/onnxruntime.ai\/docs\/performance\/tune-performance\/memory.html)\n    - [Thread management](https:\/\/onnxruntime.ai\/docs\/performance\/tune-performance\/threading.html)\n    - [I\/O Binding](https:\/\/onnxruntime.ai\/docs\/performance\/tune-performance\/iobinding.html)\n    - [Troubleshooting](https:\/\/onnxruntime.ai\/docs\/performance\/tune-performance\/troubleshooting.html)\n  - [Model optimizations](https:\/\/onnxruntime.ai\/docs\/performance\/model-optimizations\/)\n    - [Quantize ONNX models](https:\/\/onnxruntime.ai\/docs\/performance\/model-optimizations\/quantization.html)\n    - [Float16 and mixed precision models](https:\/\/onnxruntime.ai\/docs\/performance\/model-optimizations\/float16.html)\n    - [Graph optimizations](https:\/\/onnxruntime.ai\/docs\/performance\/model-optimizations\/graph-optimizations.html)\n    - [ORT model format](https:\/\/onnxruntime.ai\/docs\/performance\/model-optimizations\/ort-format-models.html)\n    - [ORT model format runtime optimization](https:\/\/onnxruntime.ai\/docs\/performance\/model-optimizations\/ort-format-model-runtime-optimization.html)\n  - [Transformers optimizer](https:\/\/onnxruntime.ai\/docs\/performance\/transformers-optimization.html)\n  - [End to end optimization with Olive](https:\/\/onnxruntime.ai\/docs\/performance\/olive.html)\n  - [Device tensors](https:\/\/onnxruntime.ai\/docs\/performance\/device-tensor.html)\n- [Ecosystem](https:\/\/onnxruntime.ai\/docs\/ecosystem\/)\n  - [Azure Container for PyTorch (ACPT)](https:\/\/onnxruntime.ai\/docs\/ecosystem\/acpt.html)\n- [Reference](https:\/\/onnxruntime.ai\/docs\/reference\/)\n  - [Releases](https:\/\/onnxruntime.ai\/docs\/reference\/releases-servicing.html)\n  - [Compatibility](https:\/\/onnxruntime.ai\/docs\/reference\/compatibility.html)\n  - [Operators](https:\/\/onnxruntime.ai\/docs\/reference\/operators\/)\n    - [Operator kernels](https:\/\/onnxruntime.ai\/docs\/reference\/operators\/OperatorKernels.html)\n    - [Contrib operators](https:\/\/onnxruntime.ai\/docs\/reference\/operators\/ContribOperators.html)\n    - [Custom operators](https:\/\/onnxruntime.ai\/docs\/reference\/operators\/add-custom-op.html)\n    - [Reduced operator config file](https:\/\/onnxruntime.ai\/docs\/reference\/operators\/reduced-operator-config-file.html)\n  - [Architecture](https:\/\/onnxruntime.ai\/docs\/reference\/high-level-design.html)\n  - [Citing ONNX Runtime](https:\/\/onnxruntime.ai\/docs\/reference\/citing.html)\n- [Dependency Management in ONNX Runtime](https:\/\/onnxruntime.ai\/docs\/build\/dependencies.html)\n\n- [ONNX Runtime Docs on GitHub](https:\/\/github.com\/microsoft\/onnxruntime\/tree\/gh-pages)\n\nThis site uses [Just the Docs](https:\/\/github.com\/just-the-docs\/just-the-docs), a documentation theme for Jekyll.\n\nSearch onnxruntime\n\n- [ONNX Runtime](https:\/\/onnxruntime.ai\/)\n- [Install](https:\/\/onnxruntime.ai\/docs\/install\/)\n- [Get Started](https:\/\/onnxruntime.ai\/docs\/get-started\/)\n- [Tutorials](https:\/\/onnxruntime.ai\/docs\/tutorials\/)\n- [API Docs](https:\/\/onnxruntime.ai\/docs\/api\/)\n- [YouTube](https:\/\/www.youtube.com\/onnxruntime)\n- [GitHub](https:\/\/github.com\/microsoft\/onnxruntime)\n\n# Install ONNX Runtime\nSee the [installation matrix](https:\/\/onnxruntime.ai\/) for recommended instructions for desired combinations of target operating system, hardware, accelerator, and language.\n\nDetails on OS versions, compilers, language versions, dependent libraries, etc can be found under [Compatibility](https:\/\/onnxruntime.ai\/docs\/reference\/compatibility).\n\n## Contents\n- [Requirements](https:\/\/onnxruntime.ai\/docs\/install\/#requirements)\n  - [CUDA and CuDNN](https:\/\/onnxruntime.ai\/docs\/install\/#cuda-and-cudnn)\n- [Python Installs](https:\/\/onnxruntime.ai\/docs\/install\/#python-installs)\n  - [Install ONNX Runtime CPU](https:\/\/onnxruntime.ai\/docs\/install\/#install-onnx-runtime-cpu)\n    - [Install nightly](https:\/\/onnxruntime.ai\/docs\/install\/#install-nightly)\n  - [Install ONNX Runtime GPU (DirectML) - Sustained Engineering Mode](https:\/\/onnxruntime.ai\/docs\/install\/#install-onnx-runtime-gpu-directml---sustained-engineering-mode)\n    - [Install nightly](https:\/\/onnxruntime.ai\/docs\/install\/#install-nightly-1)\n  - [Install ONNX Runtime GPU (CUDA or TensorRT)](https:\/\/onnxruntime.ai\/docs\/install\/#install-onnx-runtime-gpu-cuda-or-tensorrt)\n    - [CUDA 12.x](https:\/\/onnxruntime.ai\/docs\/install\/#cuda-12x)\n    - [Nightly for CUDA 13.x](https:\/\/onnxruntime.ai\/docs\/install\/#nightly-for-cuda-13x)\n    - [Nightly for CUDA 12.x](https:\/\/onnxruntime.ai\/docs\/install\/#nightly-for-cuda-12x)\n    - [CUDA 11.x](https:\/\/onnxruntime.ai\/docs\/install\/#cuda-11x)\n  - [Install ONNX Runtime QNN](https:\/\/onnxruntime.ai\/docs\/install\/#install-onnx-runtime-qnn)\n    - [Install nightly](https:\/\/onnxruntime.ai\/docs\/install\/#install-nightly-2)\n- [C\\#\/C\/C++\/WinML Installs](https:\/\/onnxruntime.ai\/docs\/install\/#cccwinml-installs)\n  - [Install ONNX Runtime](https:\/\/onnxruntime.ai\/docs\/install\/#install-onnx-runtime-1)\n    - [Install ONNX Runtime CPU](https:\/\/onnxruntime.ai\/docs\/install\/#install-onnx-runtime-cpu-1)\n    - [Install ONNX Runtime GPU (CUDA 12.x)](https:\/\/onnxruntime.ai\/docs\/install\/#install-onnx-runtime-gpu-cuda-12x)\n    - [Install ONNX Runtime GPU (CUDA 11.8)](https:\/\/onnxruntime.ai\/docs\/install\/#install-onnx-runtime-gpu-cuda-118)\n    - [DirectML (sustained engineering - use WinML for new projects)](https:\/\/onnxruntime.ai\/docs\/install\/#directml-sustained-engineering---use-winml-for-new-projects)\n    - [WinML (recommended for Windows)](https:\/\/onnxruntime.ai\/docs\/install\/#winml-recommended-for-windows)\n- [Install on web and mobile](https:\/\/onnxruntime.ai\/docs\/install\/#install-on-web-and-mobile)\n  - [JavaScript Installs](https:\/\/onnxruntime.ai\/docs\/install\/#javascript-installs)\n    - [Install ONNX Runtime Web (browsers)](https:\/\/onnxruntime.ai\/docs\/install\/#install-onnx-runtime-web-browsers)\n    - [Install ONNX Runtime Node.js binding (Node.js)](https:\/\/onnxruntime.ai\/docs\/install\/#install-onnx-runtime-nodejs-binding-nodejs)\n    - [Install ONNX Runtime for React Native](https:\/\/onnxruntime.ai\/docs\/install\/#install-onnx-runtime-for-react-native)\n  - [Install on iOS](https:\/\/onnxruntime.ai\/docs\/install\/#install-on-ios)\n    - [C\/C++](https:\/\/onnxruntime.ai\/docs\/install\/#cc)\n    - [Objective-C](https:\/\/onnxruntime.ai\/docs\/install\/#objective-c)\n    - [Custom build](https:\/\/onnxruntime.ai\/docs\/install\/#custom-build)\n  - [Install on Android](https:\/\/onnxruntime.ai\/docs\/install\/#install-on-android)\n    - [Java\/Kotlin](https:\/\/onnxruntime.ai\/docs\/install\/#javakotlin)\n    - [C\/C++](https:\/\/onnxruntime.ai\/docs\/install\/#cc-1)\n    - [Custom build](https:\/\/onnxruntime.ai\/docs\/install\/#custom-build-1)\n- [Install for On-Device Training](https:\/\/onnxruntime.ai\/docs\/install\/#install-for-on-device-training)\n  - [Offline Phase - Prepare for Training](https:\/\/onnxruntime.ai\/docs\/install\/#offline-phase---prepare-for-training)\n  - [Training Phase - On-Device Training](https:\/\/onnxruntime.ai\/docs\/install\/#training-phase---on-device-training)\n- [Large Model Training](https:\/\/onnxruntime.ai\/docs\/install\/#large-model-training)\n- [Inference install table for all languages](https:\/\/onnxruntime.ai\/docs\/install\/#inference-install-table-for-all-languages)\n- [Training install table for all languages](https:\/\/onnxruntime.ai\/docs\/install\/#training-install-table-for-all-languages)\n## Requirements\n- All builds require the English language package with `en_US.UTF-8` locale. On Linux, install [language-pack-en package](https:\/\/packages.ubuntu.com\/search?keywords=language-pack-en) by running `locale-gen en_US.UTF-8` and `update-locale LANG=en_US.UTF-8`\n- Windows builds require [Visual C++ 2019 runtime](https:\/\/learn.microsoft.com\/en-us\/cpp\/windows\/latest-supported-vc-redist?view=msvc-170#latest-microsoft-visual-c-redistributable-version). The latest version is recommended.\n### CUDA and CuDNN\nFor ONNX Runtime GPU package, it is required to install [CUDA](https:\/\/developer.nvidia.com\/cuda-toolkit) and [cuDNN](https:\/\/developer.nvidia.com\/cudnn). Check [CUDA execution provider requirements](https:\/\/onnxruntime.ai\/docs\/execution-providers\/CUDA-ExecutionProvider.html#requirements) for compatible version of CUDA and cuDNN.\n\n- Zlib is required by cuDNN 9.x for Linux only (zlib is statically linked into the cuDNN 9.x Windows dynamic libraries), or cuDNN 8.x for Linux and Windows. Follow the [cuDNN 8.9 installation guide](https:\/\/docs.nvidia.com\/deeplearning\/cudnn\/archives\/cudnn-890\/install-guide\/index.html) to install zlib in Linux or Windows.\n- In Windows, the path of CUDA `bin` and cuDNN `bin` directories must be added to the `PATH` environment variable.\n- In Linux, the path of CUDA `lib64` and cuDNN `lib` directories must be added to the `LD_LIBRARY_PATH` environment variable.\n\nFor `onnxruntime-gpu` package, it is possible to work with PyTorch without the need for manual installations of CUDA or cuDNN. Refer to [Compatibility with PyTorch](https:\/\/onnxruntime.ai\/docs\/execution-providers\/CUDA-ExecutionProvider.html#compatibility-with-pytorch) for more information.\n\n## Python Installs\n### Install ONNX Runtime CPU\n```\npip install onnxruntime\n```\n#### Install nightly\n```\npip install coloredlogs flatbuffers numpy packaging protobuf sympy\npip install --pre --index-url https:\/\/aiinfra.pkgs.visualstudio.com\/PublicPackages\/_packaging\/ORT-Nightly\/pypi\/simple\/ onnxruntime\n```\n### Install ONNX Runtime GPU (DirectML) - Sustained Engineering Mode\n**Note**: DirectML is in sustained engineering. For new Windows projects, consider [WinML](https:\/\/onnxruntime.ai\/docs\/install\/#winml-recommended-for-windows) instead.\n```\npip install onnxruntime-directml\n```\n#### Install nightly\n```\npip install coloredlogs flatbuffers numpy packaging protobuf sympy\npip install --pre --index-url https:\/\/aiinfra.pkgs.visualstudio.com\/PublicPackages\/_packaging\/ORT-Nightly\/pypi\/simple\/ onnxruntime-directml\n```\n### Install ONNX Runtime GPU (CUDA or TensorRT)\n#### CUDA 12.x\nThe default CUDA version for [onnxruntime-gpu in pypi](https:\/\/pypi.org\/project\/onnxruntime-gpu) is 12.x since 1.19.0.\n```\npip install onnxruntime-gpu\n```\nFor previous versions, you can download here: [1\\.18.1](https:\/\/aiinfra.visualstudio.com\/PublicPackages\/_artifacts\/feed\/onnxruntime-cuda-12\/PyPI\/onnxruntime-gpu\/overview\/1.18.1), [1\\.18.0](https:\/\/aiinfra.visualstudio.com\/PublicPackages\/_artifacts\/feed\/onnxruntime-cuda-12\/PyPI\/onnxruntime-gpu\/overview\/1.18.0)\n\n#### Nightly for CUDA 13.x\n```\npip install coloredlogs flatbuffers numpy packaging protobuf sympy\npip install --pre --index-url https:\/\/aiinfra.pkgs.visualstudio.com\/PublicPackages\/_packaging\/ort-cuda-13-nightly\/pypi\/simple\/ onnxruntime-gpu\n```\n#### Nightly for CUDA 12.x\n```\npip install coloredlogs flatbuffers numpy packaging protobuf sympy\npip install --pre --index-url https:\/\/aiinfra.pkgs.visualstudio.com\/PublicPackages\/_packaging\/ORT-Nightly\/pypi\/simple\/ onnxruntime-gpu\n```\n#### CUDA 11.x\nFor Cuda 11.x, please use the following instructions to install from [ORT Azure Devops Feed](https:\/\/aiinfra.visualstudio.com\/PublicPackages\/_artifacts\/feed\/onnxruntime-cuda-11\/PyPI\/onnxruntime-gpu\/overview) for 1.19.2 or later.\n```\npip install coloredlogs flatbuffers numpy packaging protobuf sympy\npip install onnxruntime-gpu --index-url https:\/\/aiinfra.pkgs.visualstudio.com\/PublicPackages\/_packaging\/onnxruntime-cuda-11\/pypi\/simple\/\n```\nFor previous versions, you can download here: [1\\.18.1](https:\/\/pypi.org\/project\/onnxruntime-gpu\/1.18.1\/), [1\\.18.0](https:\/\/pypi.org\/project\/onnxruntime-gpu\/1.18.0\/)\n\n### Install ONNX Runtime QNN\n```\npip install onnxruntime-qnn\n```\n#### Install nightly\n```\npip install coloredlogs flatbuffers numpy packaging protobuf sympy\npip install --pre --index-url https:\/\/aiinfra.pkgs.visualstudio.com\/PublicPackages\/_packaging\/ORT-Nightly\/pypi\/simple\/ onnxruntime-qnn\n```\n## C\\#\/C\/C++\/WinML Installs\n### Install ONNX Runtime\n#### Install ONNX Runtime CPU\n```\n# CPU\ndotnet add package Microsoft.ML.OnnxRuntime\n```\n#### Install ONNX Runtime GPU (CUDA 12.x)\nThe default CUDA version for ORT is 12.x\n```\n# GPU\ndotnet add package Microsoft.ML.OnnxRuntime.Gpu\n```\n#### Install ONNX Runtime GPU (CUDA 11.8)\n1. Project Setup\n\nEnsure you have installed the latest version of the Azure Artifacts keyring from the its [Github Repo](https:\/\/github.com\/microsoft\/artifacts-credprovider#azure-artifacts-credential-provider).   \n Add a nuget.config file to your project in the same directory as your .csproj file.\n```\n<?xml version=\"1.0\" encoding=\"utf-8\"?>\n<configuration>\n    <packageSources>\n        <clear\/>\n        <add key=\"onnxruntime-cuda-11\"\n             value=\"https:\/\/aiinfra.pkgs.visualstudio.com\/PublicPackages\/_packaging\/onnxruntime-cuda-11\/nuget\/v3\/index.json\"\/>\n    <\/packageSources>\n<\/configuration>\n```\n1. Restore packages\n\nRestore packages (using the interactive flag, which allows dotnet to prompt you for credentials)\n```\ndotnet add package Microsoft.ML.OnnxRuntime.Gpu\n```\nNote: You don’t need –interactive every time. dotnet will prompt you to add –interactive if it needs updated credentials.\n\n#### DirectML (sustained engineering - use WinML for new projects)\n```\ndotnet add package Microsoft.ML.OnnxRuntime.DirectML\n```\n**Note**: DirectML is in sustained engineering. For new Windows projects, use WinML instead:\n\n#### WinML (recommended for Windows)\n```\ndotnet add package Microsoft.AI.MachineLearning\n```\n## Install on web and mobile\nThe pre-built packages have full support for all ONNX opsets and operators.\n\nIf the pre-built package is too large, you can create a [custom build](https:\/\/onnxruntime.ai\/docs\/build\/custom.html). A custom build can include just the opsets and operators in your model\/s to reduce the size.\n\n### JavaScript Installs\n#### Install ONNX Runtime Web (browsers)\n```\n# install latest release version\nnpm install onnxruntime-web\n\n# install nightly build dev version\nnpm install onnxruntime-web@dev\n```\n#### Install ONNX Runtime Node.js binding (Node.js)\n```\n# install latest release version\nnpm install onnxruntime-node\n```\n#### Install ONNX Runtime for React Native\n```\n# install latest release version\nnpm install onnxruntime-react-native\n```\n### Install on iOS\nIn your CocoaPods `Podfile`, add the `onnxruntime-c` or `onnxruntime-objc` pod, depending on which API you want to use.\n\n#### C\/C++\n```\n  use_frameworks!\n\n  pod 'onnxruntime-c'\n```\n#### Objective-C\n```\n  use_frameworks!\n\n  pod 'onnxruntime-objc'\n```\nRun `pod install`.\n\n#### Custom build\nRefer to the instructions for creating a [custom iOS package](https:\/\/onnxruntime.ai\/docs\/build\/custom.html#ios).\n\n### Install on Android\n#### Java\/Kotlin\nIn your Android Studio Project, make the following changes to:\n\n1. build.gradle (Project):\n   ```\n    repositories {\n     mavenCentral()\n }\n   ```\n2. build.gradle (Module):\n   ```\n    dependencies {\n     implementation 'com.microsoft.onnxruntime:onnxruntime-android:latest.release'\n }\n   ```\n#### C\/C++\nDownload the [onnxruntime-android](https:\/\/mvnrepository.com\/artifact\/com.microsoft.onnxruntime\/onnxruntime-android) AAR hosted at MavenCentral, change the file extension from `.aar` to `.zip`, and unzip it. Include the header files from the `headers` folder, and the relevant `libonnxruntime.so` dynamic library from the `jni` folder in your NDK project.\n\n#### Custom build\nRefer to the instructions for creating a [custom Android package](https:\/\/onnxruntime.ai\/docs\/build\/custom.html#android).\n\n## Install for On-Device Training\nUnless stated otherwise, the installation instructions in this section refer to pre-built packages designed to perform on-device training.\n\nIf the pre-built training package supports your model but is too large, you can create a [custom training build](https:\/\/onnxruntime.ai\/docs\/build\/custom.html).\n\n### Offline Phase - Prepare for Training\n```\npython -m pip install cerberus flatbuffers h5py numpy>=1.16.6 onnx packaging protobuf sympy setuptools>=41.4.0\npip install -i https:\/\/aiinfra.pkgs.visualstudio.com\/PublicPackages\/_packaging\/ORT\/pypi\/simple\/ onnxruntime-training-cpu\n```\n### Training Phase - On-Device Training\n| Device | Language | PackageName | Installation Instructions |\n|---|---|---|---|\n| Windows | C, C++, C\\# | [Microsoft.ML.OnnxRuntime.Training](https:\/\/www.nuget.org\/packages\/Microsoft.ML.OnnxRuntime) | `dotnet add package Microsoft.ML.OnnxRuntime.Training` |\n| Linux | C, C++ | [onnxruntime-training-linux\\*.tgz](https:\/\/github.com\/microsoft\/onnxruntime\/releases) | Download the `*.tgz` file from [here](https:\/\/github.com\/microsoft\/onnxruntime\/releases). Extract it. Move and include the header files in the `include` directory. Move the `libonnxruntime.so` dynamic library to a desired path and include it. |\n|   | Python | [onnxruntime-training](https:\/\/pypi.org\/project\/onnxruntime-training\/) | `pip install onnxruntime-training` |\n| Android | C, C++ | [onnxruntime-training-android](https:\/\/mvnrepository.com\/artifact\/com.microsoft.onnxruntime\/onnxruntime-training-android) | Download the [onnxruntime-training-android (full package)](https:\/\/mvnrepository.com\/artifact\/com.microsoft.onnxruntime\/onnxruntime-android) AAR hosted at Maven Central. Change the file extension from `.aar` to `.zip`, and unzip it. Include the header files from the `headers` folder. Include the relevant `libonnxruntime.so` dynamic library from the `jni` folder in your NDK project. |\n|   | Java\/Kotlin | [onnxruntime-training-android](https:\/\/mvnrepository.com\/artifact\/com.microsoft.onnxruntime\/onnxruntime-android) | In your Android Studio Project, make the following changes to: build.gradle (Project): ` repositories { mavenCentral() }` build.gradle (Module): ` dependencies { implementation 'com.microsoft.onnxruntime:onnxruntime-training-android:latest.release' }` |\n| iOS | C, C++ | **CocoaPods: onnxruntime-training-c** | In your CocoaPods `Podfile`, add the `onnxruntime-training-c` pod: Run `pod install`. |\n|   | Objective-C | **CocoaPods: onnxruntime-training-objc** | In your CocoaPods `Podfile`, add the `onnxruntime-training-objc` pod: Run `pod install`. |\n| Web | JavaScript, TypeScript | onnxruntime-web | Use either `import * as ort from 'onnxruntime-web\/training';` or `const ort = require('onnxruntime-web\/training');` |\n## Large Model Training\n```\npip install torch-ort\npython -m torch_ort.configure\n```\n**Note**: This installs the default version of the `torch-ort` and `onnxruntime-training` packages that are mapped to specific versions of the CUDA libraries. Refer to the install options in [onnxruntime.ai](https:\/\/onnxruntime.ai\/).\n\n## Inference install table for all languages\nThe table below lists the build variants available as officially supported packages. Others can be [built from source](https:\/\/onnxruntime.ai\/docs\/build\/inferencing) from each [release branch](https:\/\/github.com\/microsoft\/onnxruntime\/tags).\n\nIn addition to general [requirements](https:\/\/onnxruntime.ai\/docs\/install\/#requirements), please note additional requirements and dependencies in the table below:\n\n|   | Official build | Nightly build | Reqs |\n|---|---|---|---|\n| Python | If using pip, run `pip install --upgrade pip` prior to downloading. |   |   |\n|   | CPU: [**onnxruntime**](https:\/\/pypi.org\/project\/onnxruntime) | [onnxruntime (nightly)](https:\/\/aiinfra.visualstudio.com\/PublicPackages\/_artifacts\/feed\/ORT-Nightly\/PyPI\/onnxruntime\/overview) |   |\n|   | GPU (CUDA\/TensorRT) for CUDA 12.x: [**onnxruntime-gpu**](https:\/\/pypi.org\/project\/onnxruntime-gpu) | [onnxruntime-gpu (nightly)](https:\/\/aiinfra.visualstudio.com\/PublicPackages\/_artifacts\/feed\/ORT-Nightly\/PyPI\/onnxruntime-gpu\/overview\/) | [View](https:\/\/onnxruntime.ai\/docs\/execution-providers\/CUDA-ExecutionProvider.html#requirements) |\n|   | GPU (DirectML) **sustained engineering**: [**onnxruntime-directml**](https:\/\/pypi.org\/project\/onnxruntime-directml\/) | [onnxruntime-directml (nightly)](https:\/\/aiinfra.visualstudio.com\/PublicPackages\/_artifacts\/feed\/ORT-Nightly\/PyPI\/onnxruntime-directml\/overview\/) | [View](https:\/\/onnxruntime.ai\/docs\/execution-providers\/DirectML-ExecutionProvider.html#requirements) |\n|   | OpenVINO: [**intel\/onnxruntime**](https:\/\/github.com\/intel\/onnxruntime\/releases\/latest) - *Intel managed* |   | [View](https:\/\/onnxruntime.ai\/docs\/build\/eps.html#openvino) |\n|   | TensorRT (Jetson): [**Jetson Zoo**](https:\/\/elinux.org\/Jetson_Zoo#ONNX_Runtime) - *NVIDIA managed* |   |   |\n|   | Azure (Cloud): [**onnxruntime-azure**](https:\/\/pypi.org\/project\/onnxruntime-azure\/) |   |   |\n| C\\#\/C\/C++ | CPU: [**Microsoft.ML.OnnxRuntime**](https:\/\/www.nuget.org\/packages\/Microsoft.ML.OnnxRuntime) | [onnxruntime (nightly)](https:\/\/aiinfra.visualstudio.com\/PublicPackages\/_packaging?_a=feed&feed=ORT-Nightly) |   |\n|   | GPU (CUDA\/TensorRT): [**Microsoft.ML.OnnxRuntime.Gpu**](https:\/\/www.nuget.org\/packages\/Microsoft.ML.OnnxRuntime.gpu) | [onnxruntime (nightly)](https:\/\/aiinfra.visualstudio.com\/PublicPackages\/_packaging?_a=feed&feed=ORT-Nightly) | [View](https:\/\/onnxruntime.ai\/docs\/execution-providers\/CUDA-ExecutionProvider) |\n|   | GPU (DirectML) **sustained engineering**: [**Microsoft.ML.OnnxRuntime.DirectML**](https:\/\/www.nuget.org\/packages\/Microsoft.ML.OnnxRuntime.DirectML) | [onnxruntime (nightly)](https:\/\/aiinfra.visualstudio.com\/PublicPackages\/_artifacts\/feed\/ORT-Nightly\/PyPI\/ort-nightly-directml\/overview) | [View](https:\/\/onnxruntime.ai\/docs\/execution-providers\/DirectML-ExecutionProvider) |\n| WinML **recommended for Windows** | [**Microsoft.AI.MachineLearning**](https:\/\/www.nuget.org\/packages\/Microsoft.AI.MachineLearning) | [onnxruntime (nightly)](https:\/\/aiinfra.visualstudio.com\/PublicPackages\/_artifacts\/feed\/ORT-Nightly\/NuGet\/Microsoft.AI.MachineLearning\/overview) | [View](https:\/\/docs.microsoft.com\/en-us\/windows\/ai\/windows-ml\/port-app-to-nuget#prerequisites) |\n| Java | CPU: [**com.microsoft.onnxruntime:onnxruntime**](https:\/\/search.maven.org\/artifact\/com.microsoft.onnxruntime\/onnxruntime) |   | [View](https:\/\/onnxruntime.ai\/docs\/api\/java) |\n|   | GPU (CUDA\/TensorRT): [**com.microsoft.onnxruntime:onnxruntime\\_gpu**](https:\/\/search.maven.org\/artifact\/com.microsoft.onnxruntime\/onnxruntime_gpu) |   | [View](https:\/\/onnxruntime.ai\/docs\/api\/java) |\n| Android | [**com.microsoft.onnxruntime:onnxruntime-android**](https:\/\/search.maven.org\/artifact\/com.microsoft.onnxruntime\/onnxruntime-android) |   | [View](https:\/\/onnxruntime.ai\/docs\/install\/#install-on-android) |\n| iOS (C\/C++) | CocoaPods: **onnxruntime-c** |   | [View](https:\/\/onnxruntime.ai\/docs\/install\/#install-on-ios) |\n| Objective-C | CocoaPods: **onnxruntime-objc** |   | [View](https:\/\/onnxruntime.ai\/docs\/install\/#install-on-ios) |\n| React Native | [**onnxruntime-react-native** (latest)](https:\/\/www.npmjs.com\/package\/onnxruntime-react-native) | [onnxruntime-react-native (dev)](https:\/\/www.npmjs.com\/package\/onnxruntime-react-native?activeTab=versions) | [View](https:\/\/onnxruntime.ai\/docs\/api\/js) |\n| Node.js | [**onnxruntime-node** (latest)](https:\/\/www.npmjs.com\/package\/onnxruntime-node) | [onnxruntime-node (dev)](https:\/\/www.npmjs.com\/package\/onnxruntime-node?activeTab=versions) | [View](https:\/\/onnxruntime.ai\/docs\/api\/js) |\n| Web | [**onnxruntime-web** (latest)](https:\/\/www.npmjs.com\/package\/onnxruntime-web) | [onnxruntime-web (dev)](https:\/\/www.npmjs.com\/package\/onnxruntime-web?activeTab=versions) | [View](https:\/\/onnxruntime.ai\/docs\/api\/js) |\n\n*Note: Nightly builds created from the main branch are available for testing newer changes between official releases. Please use these at your own risk. We strongly advise against deploying these to production workloads as support is limited for nightly builds.*\n\n## Training install table for all languages\nRefer to the getting started with [Optimized Training](https:\/\/onnxruntime.ai\/getting-started) page for more fine-grained installation instructions.\n***\nFor documentation questions, please [file an issue](https:\/\/github.com\/microsoft\/onnxruntime\/issues\/new?assignees=&labels=documentation&projects=&template=02-documentation.yml&title=%5BDocumentation%5D+).\n\n[Edit this page on GitHub](https:\/\/github.com\/microsoft\/onnxruntime\/tree\/gh-pages\/docs\/install\/index.md)\n\nThis site uses [Just the Docs](https:\/\/github.com\/just-the-docs\/just-the-docs), a documentation theme for Jekyll.","attrs_readable_markdown":"See the [installation matrix](https:\/\/onnxruntime.ai\/) for recommended instructions for desired combinations of target operating system, hardware, accelerator, and language.\n\nDetails on OS versions, compilers, language versions, dependent libraries, etc can be found under [Compatibility](https:\/\/onnxruntime.ai\/docs\/reference\/compatibility).\n\n## Contents\n- [Requirements](https:\/\/onnxruntime.ai\/docs\/install\/#requirements)\n  - [CUDA and CuDNN](https:\/\/onnxruntime.ai\/docs\/install\/#cuda-and-cudnn)\n- [Python Installs](https:\/\/onnxruntime.ai\/docs\/install\/#python-installs)\n  - [Install ONNX Runtime CPU](https:\/\/onnxruntime.ai\/docs\/install\/#install-onnx-runtime-cpu)\n    - [Install nightly](https:\/\/onnxruntime.ai\/docs\/install\/#install-nightly)\n  - [Install ONNX Runtime GPU (DirectML) - Sustained Engineering Mode](https:\/\/onnxruntime.ai\/docs\/install\/#install-onnx-runtime-gpu-directml---sustained-engineering-mode)\n    - [Install nightly](https:\/\/onnxruntime.ai\/docs\/install\/#install-nightly-1)\n  - [Install ONNX Runtime GPU (CUDA or TensorRT)](https:\/\/onnxruntime.ai\/docs\/install\/#install-onnx-runtime-gpu-cuda-or-tensorrt)\n    - [CUDA 12.x](https:\/\/onnxruntime.ai\/docs\/install\/#cuda-12x)\n    - [Nightly for CUDA 13.x](https:\/\/onnxruntime.ai\/docs\/install\/#nightly-for-cuda-13x)\n    - [Nightly for CUDA 12.x](https:\/\/onnxruntime.ai\/docs\/install\/#nightly-for-cuda-12x)\n    - [CUDA 11.x](https:\/\/onnxruntime.ai\/docs\/install\/#cuda-11x)\n  - [Install ONNX Runtime QNN](https:\/\/onnxruntime.ai\/docs\/install\/#install-onnx-runtime-qnn)\n    - [Install nightly](https:\/\/onnxruntime.ai\/docs\/install\/#install-nightly-2)\n- [C\\#\/C\/C++\/WinML Installs](https:\/\/onnxruntime.ai\/docs\/install\/#cccwinml-installs)\n  - [Install ONNX Runtime](https:\/\/onnxruntime.ai\/docs\/install\/#install-onnx-runtime-1)\n    - [Install ONNX Runtime CPU](https:\/\/onnxruntime.ai\/docs\/install\/#install-onnx-runtime-cpu-1)\n    - [Install ONNX Runtime GPU (CUDA 12.x)](https:\/\/onnxruntime.ai\/docs\/install\/#install-onnx-runtime-gpu-cuda-12x)\n    - [Install ONNX Runtime GPU (CUDA 11.8)](https:\/\/onnxruntime.ai\/docs\/install\/#install-onnx-runtime-gpu-cuda-118)\n    - [DirectML (sustained engineering - use WinML for new projects)](https:\/\/onnxruntime.ai\/docs\/install\/#directml-sustained-engineering---use-winml-for-new-projects)\n    - [WinML (recommended for Windows)](https:\/\/onnxruntime.ai\/docs\/install\/#winml-recommended-for-windows)\n- [Install on web and mobile](https:\/\/onnxruntime.ai\/docs\/install\/#install-on-web-and-mobile)\n  - [JavaScript Installs](https:\/\/onnxruntime.ai\/docs\/install\/#javascript-installs)\n    - [Install ONNX Runtime Web (browsers)](https:\/\/onnxruntime.ai\/docs\/install\/#install-onnx-runtime-web-browsers)\n    - [Install ONNX Runtime Node.js binding (Node.js)](https:\/\/onnxruntime.ai\/docs\/install\/#install-onnx-runtime-nodejs-binding-nodejs)\n    - [Install ONNX Runtime for React Native](https:\/\/onnxruntime.ai\/docs\/install\/#install-onnx-runtime-for-react-native)\n  - [Install on iOS](https:\/\/onnxruntime.ai\/docs\/install\/#install-on-ios)\n    - [C\/C++](https:\/\/onnxruntime.ai\/docs\/install\/#cc)\n    - [Objective-C](https:\/\/onnxruntime.ai\/docs\/install\/#objective-c)\n    - [Custom build](https:\/\/onnxruntime.ai\/docs\/install\/#custom-build)\n  - [Install on Android](https:\/\/onnxruntime.ai\/docs\/install\/#install-on-android)\n    - [Java\/Kotlin](https:\/\/onnxruntime.ai\/docs\/install\/#javakotlin)\n    - [C\/C++](https:\/\/onnxruntime.ai\/docs\/install\/#cc-1)\n    - [Custom build](https:\/\/onnxruntime.ai\/docs\/install\/#custom-build-1)\n- [Install for On-Device Training](https:\/\/onnxruntime.ai\/docs\/install\/#install-for-on-device-training)\n  - [Offline Phase - Prepare for Training](https:\/\/onnxruntime.ai\/docs\/install\/#offline-phase---prepare-for-training)\n  - [Training Phase - On-Device Training](https:\/\/onnxruntime.ai\/docs\/install\/#training-phase---on-device-training)\n- [Large Model Training](https:\/\/onnxruntime.ai\/docs\/install\/#large-model-training)\n- [Inference install table for all languages](https:\/\/onnxruntime.ai\/docs\/install\/#inference-install-table-for-all-languages)\n- [Training install table for all languages](https:\/\/onnxruntime.ai\/docs\/install\/#training-install-table-for-all-languages)\n## Requirements\n- All builds require the English language package with `en_US.UTF-8` locale. On Linux, install [language-pack-en package](https:\/\/packages.ubuntu.com\/search?keywords=language-pack-en) by running `locale-gen en_US.UTF-8` and `update-locale LANG=en_US.UTF-8`\n- Windows builds require [Visual C++ 2019 runtime](https:\/\/learn.microsoft.com\/en-us\/cpp\/windows\/latest-supported-vc-redist?view=msvc-170#latest-microsoft-visual-c-redistributable-version). The latest version is recommended.\n### CUDA and CuDNN\nFor ONNX Runtime GPU package, it is required to install [CUDA](https:\/\/developer.nvidia.com\/cuda-toolkit) and [cuDNN](https:\/\/developer.nvidia.com\/cudnn). Check [CUDA execution provider requirements](https:\/\/onnxruntime.ai\/docs\/execution-providers\/CUDA-ExecutionProvider.html#requirements) for compatible version of CUDA and cuDNN.\n\n- Zlib is required by cuDNN 9.x for Linux only (zlib is statically linked into the cuDNN 9.x Windows dynamic libraries), or cuDNN 8.x for Linux and Windows. Follow the [cuDNN 8.9 installation guide](https:\/\/docs.nvidia.com\/deeplearning\/cudnn\/archives\/cudnn-890\/install-guide\/index.html) to install zlib in Linux or Windows.\n- In Windows, the path of CUDA `bin` and cuDNN `bin` directories must be added to the `PATH` environment variable.\n- In Linux, the path of CUDA `lib64` and cuDNN `lib` directories must be added to the `LD_LIBRARY_PATH` environment variable.\n\nFor `onnxruntime-gpu` package, it is possible to work with PyTorch without the need for manual installations of CUDA or cuDNN. Refer to [Compatibility with PyTorch](https:\/\/onnxruntime.ai\/docs\/execution-providers\/CUDA-ExecutionProvider.html#compatibility-with-pytorch) for more information.\n\n## Python Installs\n### Install ONNX Runtime CPU\n```\npip install onnxruntime\n```\n#### Install nightly\n```\npip install coloredlogs flatbuffers numpy packaging protobuf sympy\npip install --pre --index-url https:\/\/aiinfra.pkgs.visualstudio.com\/PublicPackages\/_packaging\/ORT-Nightly\/pypi\/simple\/ onnxruntime\n```\n### Install ONNX Runtime GPU (DirectML) - Sustained Engineering Mode\n**Note**: DirectML is in sustained engineering. For new Windows projects, consider [WinML](https:\/\/onnxruntime.ai\/docs\/install\/#winml-recommended-for-windows) instead.\n```\npip install onnxruntime-directml\n```\n#### Install nightly\n```\npip install coloredlogs flatbuffers numpy packaging protobuf sympy\npip install --pre --index-url https:\/\/aiinfra.pkgs.visualstudio.com\/PublicPackages\/_packaging\/ORT-Nightly\/pypi\/simple\/ onnxruntime-directml\n```\n### Install ONNX Runtime GPU (CUDA or TensorRT)\n#### CUDA 12.x\nThe default CUDA version for [onnxruntime-gpu in pypi](https:\/\/pypi.org\/project\/onnxruntime-gpu) is 12.x since 1.19.0.\n```\npip install onnxruntime-gpu\n```\nFor previous versions, you can download here: [1\\.18.1](https:\/\/aiinfra.visualstudio.com\/PublicPackages\/_artifacts\/feed\/onnxruntime-cuda-12\/PyPI\/onnxruntime-gpu\/overview\/1.18.1), [1\\.18.0](https:\/\/aiinfra.visualstudio.com\/PublicPackages\/_artifacts\/feed\/onnxruntime-cuda-12\/PyPI\/onnxruntime-gpu\/overview\/1.18.0)\n\n#### Nightly for CUDA 13.x\n```\npip install coloredlogs flatbuffers numpy packaging protobuf sympy\npip install --pre --index-url https:\/\/aiinfra.pkgs.visualstudio.com\/PublicPackages\/_packaging\/ort-cuda-13-nightly\/pypi\/simple\/ onnxruntime-gpu\n```\n#### Nightly for CUDA 12.x\n```\npip install coloredlogs flatbuffers numpy packaging protobuf sympy\npip install --pre --index-url https:\/\/aiinfra.pkgs.visualstudio.com\/PublicPackages\/_packaging\/ORT-Nightly\/pypi\/simple\/ onnxruntime-gpu\n```\n#### CUDA 11.x\nFor Cuda 11.x, please use the following instructions to install from [ORT Azure Devops Feed](https:\/\/aiinfra.visualstudio.com\/PublicPackages\/_artifacts\/feed\/onnxruntime-cuda-11\/PyPI\/onnxruntime-gpu\/overview) for 1.19.2 or later.\n```\npip install coloredlogs flatbuffers numpy packaging protobuf sympy\npip install onnxruntime-gpu --index-url https:\/\/aiinfra.pkgs.visualstudio.com\/PublicPackages\/_packaging\/onnxruntime-cuda-11\/pypi\/simple\/\n```\nFor previous versions, you can download here: [1\\.18.1](https:\/\/pypi.org\/project\/onnxruntime-gpu\/1.18.1\/), [1\\.18.0](https:\/\/pypi.org\/project\/onnxruntime-gpu\/1.18.0\/)\n\n### Install ONNX Runtime QNN\n```\npip install onnxruntime-qnn\n```\n#### Install nightly\n```\npip install coloredlogs flatbuffers numpy packaging protobuf sympy\npip install --pre --index-url https:\/\/aiinfra.pkgs.visualstudio.com\/PublicPackages\/_packaging\/ORT-Nightly\/pypi\/simple\/ onnxruntime-qnn\n```\n## C\\#\/C\/C++\/WinML Installs\n### Install ONNX Runtime\n#### Install ONNX Runtime CPU\n```\n# CPU\ndotnet add package Microsoft.ML.OnnxRuntime\n```\n#### Install ONNX Runtime GPU (CUDA 12.x)\nThe default CUDA version for ORT is 12.x\n```\n# GPU\ndotnet add package Microsoft.ML.OnnxRuntime.Gpu\n```\n#### Install ONNX Runtime GPU (CUDA 11.8)\n1. Project Setup\n\nEnsure you have installed the latest version of the Azure Artifacts keyring from the its [Github Repo](https:\/\/github.com\/microsoft\/artifacts-credprovider#azure-artifacts-credential-provider).   \n Add a nuget.config file to your project in the same directory as your .csproj file.\n```\n<?xml version=\"1.0\" encoding=\"utf-8\"?>\n<configuration>\n    <packageSources>\n        <clear\/>\n        <add key=\"onnxruntime-cuda-11\"\n             value=\"https:\/\/aiinfra.pkgs.visualstudio.com\/PublicPackages\/_packaging\/onnxruntime-cuda-11\/nuget\/v3\/index.json\"\/>\n    <\/packageSources>\n<\/configuration>\n```\n1. Restore packages\n\nRestore packages (using the interactive flag, which allows dotnet to prompt you for credentials)\n```\ndotnet add package Microsoft.ML.OnnxRuntime.Gpu\n```\nNote: You don’t need –interactive every time. dotnet will prompt you to add –interactive if it needs updated credentials.\n\n#### DirectML (sustained engineering - use WinML for new projects)\n```\ndotnet add package Microsoft.ML.OnnxRuntime.DirectML\n```\n**Note**: DirectML is in sustained engineering. For new Windows projects, use WinML instead:\n\n#### WinML (recommended for Windows)\n```\ndotnet add package Microsoft.AI.MachineLearning\n```\n## Install on web and mobile\nThe pre-built packages have full support for all ONNX opsets and operators.\n\nIf the pre-built package is too large, you can create a [custom build](https:\/\/onnxruntime.ai\/docs\/build\/custom.html). A custom build can include just the opsets and operators in your model\/s to reduce the size.\n\n### JavaScript Installs\n#### Install ONNX Runtime Web (browsers)\n```\n# install latest release version\nnpm install onnxruntime-web\n\n# install nightly build dev version\nnpm install onnxruntime-web@dev\n```\n#### Install ONNX Runtime Node.js binding (Node.js)\n```\n# install latest release version\nnpm install onnxruntime-node\n```\n#### Install ONNX Runtime for React Native\n```\n# install latest release version\nnpm install onnxruntime-react-native\n```\n### Install on iOS\nIn your CocoaPods `Podfile`, add the `onnxruntime-c` or `onnxruntime-objc` pod, depending on which API you want to use.\n\n#### C\/C++\n```\n  use_frameworks!\n\n  pod 'onnxruntime-c'\n```\n#### Objective-C\n```\n  use_frameworks!\n\n  pod 'onnxruntime-objc'\n```\nRun `pod install`.\n\n#### Custom build\nRefer to the instructions for creating a [custom iOS package](https:\/\/onnxruntime.ai\/docs\/build\/custom.html#ios).\n\n### Install on Android\n#### Java\/Kotlin\nIn your Android Studio Project, make the following changes to:\n\n1. build.gradle (Project):\n   ```\n    repositories {\n     mavenCentral()\n }\n   ```\n2. build.gradle (Module):\n   ```\n    dependencies {\n     implementation 'com.microsoft.onnxruntime:onnxruntime-android:latest.release'\n }\n   ```\n#### C\/C++\nDownload the [onnxruntime-android](https:\/\/mvnrepository.com\/artifact\/com.microsoft.onnxruntime\/onnxruntime-android) AAR hosted at MavenCentral, change the file extension from `.aar` to `.zip`, and unzip it. Include the header files from the `headers` folder, and the relevant `libonnxruntime.so` dynamic library from the `jni` folder in your NDK project.\n\n#### Custom build\nRefer to the instructions for creating a [custom Android package](https:\/\/onnxruntime.ai\/docs\/build\/custom.html#android).\n\n## Install for On-Device Training\nUnless stated otherwise, the installation instructions in this section refer to pre-built packages designed to perform on-device training.\n\nIf the pre-built training package supports your model but is too large, you can create a [custom training build](https:\/\/onnxruntime.ai\/docs\/build\/custom.html).\n\n### Offline Phase - Prepare for Training\n```\npython -m pip install cerberus flatbuffers h5py numpy>=1.16.6 onnx packaging protobuf sympy setuptools>=41.4.0\npip install -i https:\/\/aiinfra.pkgs.visualstudio.com\/PublicPackages\/_packaging\/ORT\/pypi\/simple\/ onnxruntime-training-cpu\n```\n### Training Phase - On-Device Training\n| Device | Language | PackageName | Installation Instructions |\n|---|---|---|---|\n| Windows | C, C++, C\\# | [Microsoft.ML.OnnxRuntime.Training](https:\/\/www.nuget.org\/packages\/Microsoft.ML.OnnxRuntime) | `dotnet add package Microsoft.ML.OnnxRuntime.Training` |\n| Linux | C, C++ | [onnxruntime-training-linux\\*.tgz](https:\/\/github.com\/microsoft\/onnxruntime\/releases) | Download the `*.tgz` file from [here](https:\/\/github.com\/microsoft\/onnxruntime\/releases). Extract it. Move and include the header files in the `include` directory. Move the `libonnxruntime.so` dynamic library to a desired path and include it. |\n|   | Python | [onnxruntime-training](https:\/\/pypi.org\/project\/onnxruntime-training\/) | `pip install onnxruntime-training` |\n| Android | C, C++ | [onnxruntime-training-android](https:\/\/mvnrepository.com\/artifact\/com.microsoft.onnxruntime\/onnxruntime-training-android) | Download the [onnxruntime-training-android (full package)](https:\/\/mvnrepository.com\/artifact\/com.microsoft.onnxruntime\/onnxruntime-android) AAR hosted at Maven Central. Change the file extension from `.aar` to `.zip`, and unzip it. Include the header files from the `headers` folder. Include the relevant `libonnxruntime.so` dynamic library from the `jni` folder in your NDK project. |\n|   | Java\/Kotlin | [onnxruntime-training-android](https:\/\/mvnrepository.com\/artifact\/com.microsoft.onnxruntime\/onnxruntime-android) | In your Android Studio Project, make the following changes to: build.gradle (Project): ` repositories { mavenCentral() }` build.gradle (Module): ` dependencies { implementation 'com.microsoft.onnxruntime:onnxruntime-training-android:latest.release' }` |\n| iOS | C, C++ | **CocoaPods: onnxruntime-training-c** | In your CocoaPods `Podfile`, add the `onnxruntime-training-c` pod: Run `pod install`. |\n|   | Objective-C | **CocoaPods: onnxruntime-training-objc** | In your CocoaPods `Podfile`, add the `onnxruntime-training-objc` pod: Run `pod install`. |\n| Web | JavaScript, TypeScript | onnxruntime-web | Use either `import * as ort from 'onnxruntime-web\/training';` or `const ort = require('onnxruntime-web\/training');` |\n## Large Model Training\n```\npip install torch-ort\npython -m torch_ort.configure\n```\n**Note**: This installs the default version of the `torch-ort` and `onnxruntime-training` packages that are mapped to specific versions of the CUDA libraries. Refer to the install options in [onnxruntime.ai](https:\/\/onnxruntime.ai\/).\n\n## Inference install table for all languages\nThe table below lists the build variants available as officially supported packages. Others can be [built from source](https:\/\/onnxruntime.ai\/docs\/build\/inferencing) from each [release branch](https:\/\/github.com\/microsoft\/onnxruntime\/tags).\n\nIn addition to general [requirements](https:\/\/onnxruntime.ai\/docs\/install\/#requirements), please note additional requirements and dependencies in the table below:\n\n|   | Official build | Nightly build | Reqs |\n|---|---|---|---|\n| Python | If using pip, run `pip install --upgrade pip` prior to downloading. |   |   |\n|   | CPU: [**onnxruntime**](https:\/\/pypi.org\/project\/onnxruntime) | [onnxruntime (nightly)](https:\/\/aiinfra.visualstudio.com\/PublicPackages\/_artifacts\/feed\/ORT-Nightly\/PyPI\/onnxruntime\/overview) |   |\n|   | GPU (CUDA\/TensorRT) for CUDA 12.x: [**onnxruntime-gpu**](https:\/\/pypi.org\/project\/onnxruntime-gpu) | [onnxruntime-gpu (nightly)](https:\/\/aiinfra.visualstudio.com\/PublicPackages\/_artifacts\/feed\/ORT-Nightly\/PyPI\/onnxruntime-gpu\/overview\/) | [View](https:\/\/onnxruntime.ai\/docs\/execution-providers\/CUDA-ExecutionProvider.html#requirements) |\n|   | GPU (DirectML) **sustained engineering**: [**onnxruntime-directml**](https:\/\/pypi.org\/project\/onnxruntime-directml\/) | [onnxruntime-directml (nightly)](https:\/\/aiinfra.visualstudio.com\/PublicPackages\/_artifacts\/feed\/ORT-Nightly\/PyPI\/onnxruntime-directml\/overview\/) | [View](https:\/\/onnxruntime.ai\/docs\/execution-providers\/DirectML-ExecutionProvider.html#requirements) |\n|   | OpenVINO: [**intel\/onnxruntime**](https:\/\/github.com\/intel\/onnxruntime\/releases\/latest) - *Intel managed* |   | [View](https:\/\/onnxruntime.ai\/docs\/build\/eps.html#openvino) |\n|   | TensorRT (Jetson): [**Jetson Zoo**](https:\/\/elinux.org\/Jetson_Zoo#ONNX_Runtime) - *NVIDIA managed* |   |   |\n|   | Azure (Cloud): [**onnxruntime-azure**](https:\/\/pypi.org\/project\/onnxruntime-azure\/) |   |   |\n| C\\#\/C\/C++ | CPU: [**Microsoft.ML.OnnxRuntime**](https:\/\/www.nuget.org\/packages\/Microsoft.ML.OnnxRuntime) | [onnxruntime (nightly)](https:\/\/aiinfra.visualstudio.com\/PublicPackages\/_packaging?_a=feed&feed=ORT-Nightly) |   |\n|   | GPU (CUDA\/TensorRT): [**Microsoft.ML.OnnxRuntime.Gpu**](https:\/\/www.nuget.org\/packages\/Microsoft.ML.OnnxRuntime.gpu) | [onnxruntime (nightly)](https:\/\/aiinfra.visualstudio.com\/PublicPackages\/_packaging?_a=feed&feed=ORT-Nightly) | [View](https:\/\/onnxruntime.ai\/docs\/execution-providers\/CUDA-ExecutionProvider) |\n|   | GPU (DirectML) **sustained engineering**: [**Microsoft.ML.OnnxRuntime.DirectML**](https:\/\/www.nuget.org\/packages\/Microsoft.ML.OnnxRuntime.DirectML) | [onnxruntime (nightly)](https:\/\/aiinfra.visualstudio.com\/PublicPackages\/_artifacts\/feed\/ORT-Nightly\/PyPI\/ort-nightly-directml\/overview) | [View](https:\/\/onnxruntime.ai\/docs\/execution-providers\/DirectML-ExecutionProvider) |\n| WinML **recommended for Windows** | [**Microsoft.AI.MachineLearning**](https:\/\/www.nuget.org\/packages\/Microsoft.AI.MachineLearning) | [onnxruntime (nightly)](https:\/\/aiinfra.visualstudio.com\/PublicPackages\/_artifacts\/feed\/ORT-Nightly\/NuGet\/Microsoft.AI.MachineLearning\/overview) | [View](https:\/\/docs.microsoft.com\/en-us\/windows\/ai\/windows-ml\/port-app-to-nuget#prerequisites) |\n| Java | CPU: [**com.microsoft.onnxruntime:onnxruntime**](https:\/\/search.maven.org\/artifact\/com.microsoft.onnxruntime\/onnxruntime) |   | [View](https:\/\/onnxruntime.ai\/docs\/api\/java) |\n|   | GPU (CUDA\/TensorRT): [**com.microsoft.onnxruntime:onnxruntime\\_gpu**](https:\/\/search.maven.org\/artifact\/com.microsoft.onnxruntime\/onnxruntime_gpu) |   | [View](https:\/\/onnxruntime.ai\/docs\/api\/java) |\n| Android | [**com.microsoft.onnxruntime:onnxruntime-android**](https:\/\/search.maven.org\/artifact\/com.microsoft.onnxruntime\/onnxruntime-android) |   | [View](https:\/\/onnxruntime.ai\/docs\/install\/#install-on-android) |\n| iOS (C\/C++) | CocoaPods: **onnxruntime-c** |   | [View](https:\/\/onnxruntime.ai\/docs\/install\/#install-on-ios) |\n| Objective-C | CocoaPods: **onnxruntime-objc** |   | [View](https:\/\/onnxruntime.ai\/docs\/install\/#install-on-ios) |\n| React Native | [**onnxruntime-react-native** (latest)](https:\/\/www.npmjs.com\/package\/onnxruntime-react-native) | [onnxruntime-react-native (dev)](https:\/\/www.npmjs.com\/package\/onnxruntime-react-native?activeTab=versions) | [View](https:\/\/onnxruntime.ai\/docs\/api\/js) |\n| Node.js | [**onnxruntime-node** (latest)](https:\/\/www.npmjs.com\/package\/onnxruntime-node) | [onnxruntime-node (dev)](https:\/\/www.npmjs.com\/package\/onnxruntime-node?activeTab=versions) | [View](https:\/\/onnxruntime.ai\/docs\/api\/js) |\n| Web | [**onnxruntime-web** (latest)](https:\/\/www.npmjs.com\/package\/onnxruntime-web) | [onnxruntime-web (dev)](https:\/\/www.npmjs.com\/package\/onnxruntime-web?activeTab=versions) | [View](https:\/\/onnxruntime.ai\/docs\/api\/js) |\n\n*Note: Nightly builds created from the main branch are available for testing newer changes between official releases. Please use these at your own risk. We strongly advise against deploying these to production workloads as support is limited for nightly builds.*\n\n## Training install table for all languages\nRefer to the getting started with [Optimized Training](https:\/\/onnxruntime.ai\/getting-started) page for more fine-grained installation instructions.","meta_canonical":null}

3. Robots.txt Check

Query:

Response:

4. Spam/Ban Check

Query:

Response:

5. Seen Status Check

ℹ️ Skipped - page is already crawled

📄

INDEXABLE

✅

CRAWLED

1 day ago

🤖

ROBOTS ALLOWED

Page Info Filters

Filter	Status	Condition	Details
HTTP status	PASS	`download_http_code = 200`	HTTP 200
Age cutoff	PASS	`download_stamp > now() - 6 MONTH`	0 months ago
History drop	PASS	`isNull(history_drop_reason)`	No drop reason
Spam/ban	PASS	`fh_dont_index != 1 AND ml_spam_score = 0`	ml_spam_score=0
Canonical	PASS	`meta_canonical IS NULL OR = '' OR = src_unparsed`	Not set

Page Details

Property	Value
URL	https://onnxruntime.ai/docs/install/
Last Crawled	2026-04-05 17:42:20 (1 day ago)
First Indexed	2021-09-16 16:40:53 (4 years ago)
HTTP Status Code	200
Meta Title	Install ONNX Runtime \| onnxruntime
Meta Description	Instructions to install ONNX Runtime on your target platform in your environment
Meta Canonical	null
Boilerpipe Text	See the installation matrix for recommended instructions for desired combinations of target operating system, hardware, accelerator, and language. Details on OS versions, compilers, language versions, dependent libraries, etc can be found under Compatibility . Contents Requirements CUDA and CuDNN Python Installs Install ONNX Runtime CPU Install nightly Install ONNX Runtime GPU (DirectML) - Sustained Engineering Mode Install nightly Install ONNX Runtime GPU (CUDA or TensorRT) CUDA 12.x Nightly for CUDA 13.x Nightly for CUDA 12.x CUDA 11.x Install ONNX Runtime QNN Install nightly C#/C/C++/WinML Installs Install ONNX Runtime Install ONNX Runtime CPU Install ONNX Runtime GPU (CUDA 12.x) Install ONNX Runtime GPU (CUDA 11.8) DirectML (sustained engineering - use WinML for new projects) WinML (recommended for Windows) Install on web and mobile JavaScript Installs Install ONNX Runtime Web (browsers) Install ONNX Runtime Node.js binding (Node.js) Install ONNX Runtime for React Native Install on iOS C/C++ Objective-C Custom build Install on Android Java/Kotlin C/C++ Custom build Install for On-Device Training Offline Phase - Prepare for Training Training Phase - On-Device Training Large Model Training Inference install table for all languages Training install table for all languages Requirements All builds require the English language package with en_US.UTF-8 locale. On Linux, install language-pack-en package by running locale-gen en_US.UTF-8 and update-locale LANG=en_US.UTF-8 Windows builds require Visual C++ 2019 runtime . The latest version is recommended. CUDA and CuDNN For ONNX Runtime GPU package, it is required to install CUDA and cuDNN . Check CUDA execution provider requirements for compatible version of CUDA and cuDNN. Zlib is required by cuDNN 9.x for Linux only (zlib is statically linked into the cuDNN 9.x Windows dynamic libraries), or cuDNN 8.x for Linux and Windows. Follow the cuDNN 8.9 installation guide to install zlib in Linux or Windows. In Windows, the path of CUDA bin and cuDNN bin directories must be added to the PATH environment variable. In Linux, the path of CUDA lib64 and cuDNN lib directories must be added to the LD_LIBRARY_PATH environment variable. For onnxruntime-gpu package, it is possible to work with PyTorch without the need for manual installations of CUDA or cuDNN. Refer to Compatibility with PyTorch for more information. Python Installs Install ONNX Runtime CPU pip install onnxruntime Install nightly pip install coloredlogs flatbuffers numpy packaging protobuf sympy pip install --pre --index-url https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/ORT-Nightly/pypi/simple/ onnxruntime Install ONNX Runtime GPU (DirectML) - Sustained Engineering Mode Note : DirectML is in sustained engineering. For new Windows projects, consider WinML instead. pip install onnxruntime-directml Install nightly pip install coloredlogs flatbuffers numpy packaging protobuf sympy pip install --pre --index-url https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/ORT-Nightly/pypi/simple/ onnxruntime-directml Install ONNX Runtime GPU (CUDA or TensorRT) CUDA 12.x The default CUDA version for onnxruntime-gpu in pypi is 12.x since 1.19.0. pip install onnxruntime-gpu For previous versions, you can download here: 1.18.1 , 1.18.0 Nightly for CUDA 13.x pip install coloredlogs flatbuffers numpy packaging protobuf sympy pip install --pre --index-url https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/ort-cuda-13-nightly/pypi/simple/ onnxruntime-gpu Nightly for CUDA 12.x pip install coloredlogs flatbuffers numpy packaging protobuf sympy pip install --pre --index-url https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/ORT-Nightly/pypi/simple/ onnxruntime-gpu CUDA 11.x For Cuda 11.x, please use the following instructions to install from ORT Azure Devops Feed for 1.19.2 or later. pip install coloredlogs flatbuffers numpy packaging protobuf sympy pip install onnxruntime-gpu --index-url https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/onnxruntime-cuda-11/pypi/simple/ For previous versions, you can download here: 1.18.1 , 1.18.0 Install ONNX Runtime QNN pip install onnxruntime-qnn Install nightly pip install coloredlogs flatbuffers numpy packaging protobuf sympy pip install --pre --index-url https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/ORT-Nightly/pypi/simple/ onnxruntime-qnn C#/C/C++/WinML Installs Install ONNX Runtime Install ONNX Runtime CPU # CPU dotnet add package Microsoft.ML.OnnxRuntime Install ONNX Runtime GPU (CUDA 12.x) The default CUDA version for ORT is 12.x # GPU dotnet add package Microsoft.ML.OnnxRuntime.Gpu Install ONNX Runtime GPU (CUDA 11.8) Project Setup Ensure you have installed the latest version of the Azure Artifacts keyring from the its Github Repo . Add a nuget.config file to your project in the same directory as your .csproj file. <?xml version="1.0" encoding="utf-8"?> <configuration> <packageSources> <clear/> <add key= "onnxruntime-cuda-11" value= "https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/onnxruntime-cuda-11/nuget/v3/index.json" /> </packageSources> </configuration> Restore packages Restore packages (using the interactive flag, which allows dotnet to prompt you for credentials) dotnet add package Microsoft.ML.OnnxRuntime.Gpu Note: You don’t need –interactive every time. dotnet will prompt you to add –interactive if it needs updated credentials. DirectML (sustained engineering - use WinML for new projects) dotnet add package Microsoft.ML.OnnxRuntime.DirectML Note : DirectML is in sustained engineering. For new Windows projects, use WinML instead: WinML (recommended for Windows) dotnet add package Microsoft.AI.MachineLearning Install on web and mobile The pre-built packages have full support for all ONNX opsets and operators. If the pre-built package is too large, you can create a custom build . A custom build can include just the opsets and operators in your model/s to reduce the size. JavaScript Installs Install ONNX Runtime Web (browsers) # install latest release version npm install onnxruntime-web # install nightly build dev version npm install onnxruntime-web@dev Install ONNX Runtime Node.js binding (Node.js) # install latest release version npm install onnxruntime-node Install ONNX Runtime for React Native # install latest release version npm install onnxruntime-react-native Install on iOS In your CocoaPods Podfile , add the onnxruntime-c or onnxruntime-objc pod, depending on which API you want to use. C/C++ use_frameworks! pod 'onnxruntime-c' Objective-C use_frameworks! pod 'onnxruntime-objc' Run pod install . Custom build Refer to the instructions for creating a custom iOS package . Install on Android Java/Kotlin In your Android Studio Project, make the following changes to: build.gradle (Project): repositories { mavenCentral () } build.gradle (Module): dependencies { implementation 'com.microsoft.onnxruntime:onnxruntime-android:latest.release' } C/C++ Download the onnxruntime-android AAR hosted at MavenCentral, change the file extension from .aar to .zip , and unzip it. Include the header files from the headers folder, and the relevant libonnxruntime.so dynamic library from the jni folder in your NDK project. Custom build Refer to the instructions for creating a custom Android package . Install for On-Device Training Unless stated otherwise, the installation instructions in this section refer to pre-built packages designed to perform on-device training. If the pre-built training package supports your model but is too large, you can create a custom training build . Offline Phase - Prepare for Training python -m pip install cerberus flatbuffers h5py numpy> = 1.16.6 onnx packaging protobuf sympy setuptools> = 41.4.0 pip install -i https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/ORT/pypi/simple/ onnxruntime-training-cpu Training Phase - On-Device Training Device Language PackageName Installation Instructions Windows C, C++, C# Microsoft.ML.OnnxRuntime.Training dotnet add package Microsoft.ML.OnnxRuntime.Training Linux C, C++ onnxruntime-training-linux.tgz Download the .tgz file from here . Extract it. Move and include the header files in the include directory. Move the libonnxruntime.so dynamic library to a desired path and include it. Python onnxruntime-training pip install onnxruntime-training Android C, C++ onnxruntime-training-android Download the onnxruntime-training-android (full package) AAR hosted at Maven Central. Change the file extension from .aar to .zip , and unzip it. Include the header files from the headers folder. Include the relevant libonnxruntime.so dynamic library from the jni folder in your NDK project. Java/Kotlin onnxruntime-training-android In your Android Studio Project, make the following changes to: build.gradle (Project): repositories { mavenCentral() } build.gradle (Module): dependencies { implementation 'com.microsoft.onnxruntime:onnxruntime-training-android:latest.release' } iOS C, C++ CocoaPods: onnxruntime-training-c In your CocoaPods Podfile , add the onnxruntime-training-c pod: use_frameworks! pod 'onnxruntime-training-c' Run pod install . Objective-C CocoaPods: onnxruntime-training-objc In your CocoaPods Podfile , add the onnxruntime-training-objc pod: use_frameworks! pod 'onnxruntime-training-objc' Run pod install . Web JavaScript, TypeScript onnxruntime-web npm install onnxruntime-web Use either import * as ort from 'onnxruntime-web/training'; or const ort = require('onnxruntime-web/training'); Large Model Training pip install torch-ort python -m torch_ort.configure Note : This installs the default version of the torch-ort and onnxruntime-training packages that are mapped to specific versions of the CUDA libraries. Refer to the install options in onnxruntime.ai . Inference install table for all languages The table below lists the build variants available as officially supported packages. Others can be built from source from each release branch . In addition to general requirements , please note additional requirements and dependencies in the table below: Official build Nightly build Reqs Python If using pip, run pip install --upgrade pip prior to downloading. CPU: onnxruntime onnxruntime (nightly) GPU (CUDA/TensorRT) for CUDA 12.x: onnxruntime-gpu onnxruntime-gpu (nightly) View GPU (DirectML) sustained engineering : onnxruntime-directml onnxruntime-directml (nightly) View OpenVINO: intel/onnxruntime - Intel managed View TensorRT (Jetson): Jetson Zoo - NVIDIA managed Azure (Cloud): onnxruntime-azure C#/C/C++ CPU: Microsoft.ML.OnnxRuntime onnxruntime (nightly) GPU (CUDA/TensorRT): Microsoft.ML.OnnxRuntime.Gpu onnxruntime (nightly) View GPU (DirectML) sustained engineering : Microsoft.ML.OnnxRuntime.DirectML onnxruntime (nightly) View WinML recommended for Windows Microsoft.AI.MachineLearning onnxruntime (nightly) View Java CPU: com.microsoft.onnxruntime:onnxruntime View GPU (CUDA/TensorRT): com.microsoft.onnxruntime:onnxruntime_gpu View Android com.microsoft.onnxruntime:onnxruntime-android View iOS (C/C++) CocoaPods: onnxruntime-c View Objective-C CocoaPods: onnxruntime-objc View React Native onnxruntime-react-native (latest) onnxruntime-react-native (dev) View Node.js onnxruntime-node (latest) onnxruntime-node (dev) View Web onnxruntime-web (latest) onnxruntime-web (dev) View Note: Nightly builds created from the main branch are available for testing newer changes between official releases. Please use these at your own risk. We strongly advise against deploying these to production workloads as support is limited for nightly builds. Training install table for all languages Refer to the getting started with Optimized Training page for more fine-grained installation instructions.
Markdown	[Skip to main content](https://onnxruntime.ai/docs/install/#main-content) - [ONNX Runtime](https://onnxruntime.ai/docs/) - [Install ONNX Runtime](https://onnxruntime.ai/docs/install/) - [Get Started](https://onnxruntime.ai/docs/get-started/) - [Python](https://onnxruntime.ai/docs/get-started/with-python.html) - [C++](https://onnxruntime.ai/docs/get-started/with-cpp.html) - [C](https://onnxruntime.ai/docs/get-started/with-c.html) - [C\#](https://onnxruntime.ai/docs/get-started/with-csharp.html) - [Java](https://onnxruntime.ai/docs/get-started/with-java.html) - [JavaScript](https://onnxruntime.ai/docs/get-started/with-javascript/) - [Web](https://onnxruntime.ai/docs/get-started/with-javascript/web.html) - [Node.js binding](https://onnxruntime.ai/docs/get-started/with-javascript/node.html) - [React Native](https://onnxruntime.ai/docs/get-started/with-javascript/react-native.html) - [Objective-C](https://onnxruntime.ai/docs/get-started/with-obj-c.html) - [Julia, Ruby and Rust APIs](https://onnxruntime.ai/docs/get-started/community-projects.html) - [Windows](https://onnxruntime.ai/docs/get-started/with-windows.html) - [Mobile](https://onnxruntime.ai/docs/get-started/with-mobile.html) - [On-Device Training](https://onnxruntime.ai/docs/get-started/training-on-device.html) - [Large Model Training](https://onnxruntime.ai/docs/get-started/training-pytorch.html) - [Tutorials](https://onnxruntime.ai/docs/tutorials/) - [API Basics](https://onnxruntime.ai/docs/tutorials/api-basics.html) - [Accelerate PyTorch](https://onnxruntime.ai/docs/tutorials/accelerate-pytorch/) - [PyTorch Inference](https://onnxruntime.ai/docs/tutorials/accelerate-pytorch/pytorch.html) - [Inference on multiple targets](https://onnxruntime.ai/docs/tutorials/accelerate-pytorch/resnet-inferencing.html) - [Accelerate PyTorch Training](https://onnxruntime.ai/docs/tutorials/accelerate-pytorch/ort-training.html) - [Accelerate TensorFlow](https://onnxruntime.ai/docs/tutorials/tensorflow.html) - [Accelerate Hugging Face](https://onnxruntime.ai/docs/tutorials/huggingface.html) - [Deploy on AzureML](https://onnxruntime.ai/docs/tutorials/azureml.html) - [Deploy on mobile](https://onnxruntime.ai/docs/tutorials/mobile/) - [Object detection and pose estimation with YOLOv8](https://onnxruntime.ai/docs/tutorials/mobile/pose-detection.html) - [Mobile image recognition on Android](https://onnxruntime.ai/docs/tutorials/mobile/deploy-android.html) - [Improve image resolution on mobile](https://onnxruntime.ai/docs/tutorials/mobile/superres.html) - [Mobile objection detection on iOS](https://onnxruntime.ai/docs/tutorials/mobile/deploy-ios.html) - [ORT Mobile Model Export Helpers](https://onnxruntime.ai/docs/tutorials/mobile/helpers/) - [Web](https://onnxruntime.ai/docs/tutorials/web/) - [Build a web app with ONNX Runtime](https://onnxruntime.ai/docs/tutorials/web/build-web-app.html) - [The 'env' Flags and Session Options](https://onnxruntime.ai/docs/tutorials/web/env-flags-and-session-options.html) - [Using WebGPU](https://onnxruntime.ai/docs/tutorials/web/ep-webgpu.html) - [Using WebNN](https://onnxruntime.ai/docs/tutorials/web/ep-webnn.html) - [Working with Large Models](https://onnxruntime.ai/docs/tutorials/web/large-models.html) - [Performance Diagnosis](https://onnxruntime.ai/docs/tutorials/web/performance-diagnosis.html) - [Deploying ONNX Runtime Web](https://onnxruntime.ai/docs/tutorials/web/deploy.html) - [Troubleshooting](https://onnxruntime.ai/docs/tutorials/web/trouble-shooting.html) - [Classify images with ONNX Runtime and Next.js](https://onnxruntime.ai/docs/tutorials/web/classify-images-nextjs-github-template.html) - [Custom Excel Functions for BERT Tasks in JavaScript](https://onnxruntime.ai/docs/tutorials/web/excel-addin-bert-js.html) - [Deploy on IoT and edge](https://onnxruntime.ai/docs/tutorials/iot-edge/) - [IoT Deployment on Raspberry Pi](https://onnxruntime.ai/docs/tutorials/iot-edge/rasp-pi-cv.html) - [Deploy traditional ML](https://onnxruntime.ai/docs/tutorials/traditional-ml.html) - [Inference with C\#](https://onnxruntime.ai/docs/tutorials/csharp/) - [Basic C\# Tutorial](https://onnxruntime.ai/docs/tutorials/csharp/basic_csharp.html) - [Inference BERT NLP with C\#](https://onnxruntime.ai/docs/tutorials/csharp/bert-nlp-csharp-console-app.html) - [Configure CUDA for GPU with C\#](https://onnxruntime.ai/docs/tutorials/csharp/csharp-gpu.html) - [Image recognition with ResNet50v2 in C\#](https://onnxruntime.ai/docs/tutorials/csharp/resnet50_csharp.html) - [Stable Diffusion with C\#](https://onnxruntime.ai/docs/tutorials/csharp/stable-diffusion-csharp.html) - [Object detection in C\# using OpenVINO](https://onnxruntime.ai/docs/tutorials/csharp/yolov3_object_detection_csharp.html) - [Object detection with Faster RCNN in C\#](https://onnxruntime.ai/docs/tutorials/csharp/fasterrcnn_csharp.html) - [On-Device Training](https://onnxruntime.ai/docs/tutorials/on-device-training/) - [Building an Android Application](https://onnxruntime.ai/docs/tutorials/on-device-training/android-app.html) - [Building an iOS Application](https://onnxruntime.ai/docs/tutorials/on-device-training/ios-app.html) - [API Docs](https://onnxruntime.ai/docs/api/) - [Build ONNX Runtime](https://onnxruntime.ai/docs/build/) - [Build for inferencing](https://onnxruntime.ai/docs/build/inferencing.html) - [Build for training](https://onnxruntime.ai/docs/build/training.html) - [Build with different EPs](https://onnxruntime.ai/docs/build/eps.html) - [Build for web](https://onnxruntime.ai/docs/build/web.html) - [Build for Android](https://onnxruntime.ai/docs/build/android.html) - [Build for iOS](https://onnxruntime.ai/docs/build/ios.html) - [Custom build](https://onnxruntime.ai/docs/build/custom.html) - [Execution Providers](https://onnxruntime.ai/docs/execution-providers/) - [NVIDIA - CUDA](https://onnxruntime.ai/docs/execution-providers/CUDA-ExecutionProvider.html) - [NVIDIA - TensorRT](https://onnxruntime.ai/docs/execution-providers/TensorRT-ExecutionProvider.html) - [NVIDIA - TensorRT RTX](https://onnxruntime.ai/docs/execution-providers/TensorRTRTX-ExecutionProvider.html) - [Intel - OpenVINO™](https://onnxruntime.ai/docs/execution-providers/OpenVINO-ExecutionProvider.html) - [Intel - oneDNN](https://onnxruntime.ai/docs/execution-providers/oneDNN-ExecutionProvider.html) - [Windows - DirectML](https://onnxruntime.ai/docs/execution-providers/DirectML-ExecutionProvider.html) - [Qualcomm - QNN](https://onnxruntime.ai/docs/execution-providers/QNN-ExecutionProvider.html) - [Android - NNAPI](https://onnxruntime.ai/docs/execution-providers/NNAPI-ExecutionProvider.html) - [Apple - CoreML](https://onnxruntime.ai/docs/execution-providers/CoreML-ExecutionProvider.html) - [XNNPACK](https://onnxruntime.ai/docs/execution-providers/Xnnpack-ExecutionProvider.html) - [AMD - ROCm](https://onnxruntime.ai/docs/execution-providers/ROCm-ExecutionProvider.html) - [AMD - MIGraphX](https://onnxruntime.ai/docs/execution-providers/MIGraphX-ExecutionProvider.html) - [AMD - Vitis AI](https://onnxruntime.ai/docs/execution-providers/Vitis-AI-ExecutionProvider.html) - [Cloud - Azure](https://onnxruntime.ai/docs/execution-providers/Azure-ExecutionProvider.html) - [Community-maintained](https://onnxruntime.ai/docs/execution-providers/community-maintained/) - [Arm - ACL](https://onnxruntime.ai/docs/execution-providers/community-maintained/ACL-ExecutionProvider.html) - [Arm - Arm NN](https://onnxruntime.ai/docs/execution-providers/community-maintained/ArmNN-ExecutionProvider.html) - [Apache - TVM](https://onnxruntime.ai/docs/execution-providers/community-maintained/TVM-ExecutionProvider.html) - [Rockchip - RKNPU](https://onnxruntime.ai/docs/execution-providers/community-maintained/RKNPU-ExecutionProvider.html) - [Huawei - CANN](https://onnxruntime.ai/docs/execution-providers/community-maintained/CANN-ExecutionProvider.html) - [Add a new provider](https://onnxruntime.ai/docs/execution-providers/add-execution-provider.html) - [EP Context Design](https://onnxruntime.ai/docs/execution-providers/EP-Context-Design.html) - [Plugin Execution Provider Libraries](https://onnxruntime.ai/docs/execution-providers/plugin-ep-libraries/) - [Usage](https://onnxruntime.ai/docs/execution-providers/plugin-ep-libraries/usage.html) - [Development](https://onnxruntime.ai/docs/execution-providers/plugin-ep-libraries/development.html) - [Testing](https://onnxruntime.ai/docs/execution-providers/plugin-ep-libraries/testing.html) - [Packaging](https://onnxruntime.ai/docs/execution-providers/plugin-ep-libraries/packaging.html) - [Generate API (Preview)](https://onnxruntime.ai/docs/genai/) - [Tutorials](https://onnxruntime.ai/docs/genai/tutorials/) - [Phi-3.5 vision tutorial](https://onnxruntime.ai/docs/genai/tutorials/phi3-v.html) - [Phi-3 tutorial](https://onnxruntime.ai/docs/genai/tutorials/phi3-python.html) - [Phi-2 tutorial](https://onnxruntime.ai/docs/genai/tutorials/phi2-python.html) - [Run with LoRA adapters](https://onnxruntime.ai/docs/genai/tutorials/finetune.html) - [DeepSeek-R1-Distill tutorial](https://onnxruntime.ai/docs/genai/tutorials/deepseek-python.html) - [Run on Snapdragon devices](https://onnxruntime.ai/docs/genai/tutorials/snapdragon.html) - [API docs](https://onnxruntime.ai/docs/genai/api/) - [Python API](https://onnxruntime.ai/docs/genai/api/python.html) - [C\# API](https://onnxruntime.ai/docs/genai/api/csharp.html) - [C API](https://onnxruntime.ai/docs/genai/api/c.html) - [C++ API](https://onnxruntime.ai/docs/genai/api/cpp.html) - [Java API](https://onnxruntime.ai/docs/genai/api/java.html) - [How to](https://onnxruntime.ai/docs/genai/howto/) - [Install](https://onnxruntime.ai/docs/genai/howto/install.html) - [Build from source](https://onnxruntime.ai/docs/genai/howto/build-from-source.html) - [Build models](https://onnxruntime.ai/docs/genai/howto/build-model.html) - [Build models for Snapdragon](https://onnxruntime.ai/docs/genai/howto/build-models-for-snapdragon.html) - [Troubleshoot](https://onnxruntime.ai/docs/genai/howto/troubleshoot.html) - [Migrate](https://onnxruntime.ai/docs/genai/howto/migrate.html) - [Past present share buffer](https://onnxruntime.ai/docs/genai/howto/past-present-share-buffer.html) - [Reference](https://onnxruntime.ai/docs/genai/reference/) - [Config reference](https://onnxruntime.ai/docs/genai/reference/config.html) - [Adapter file spec](https://onnxruntime.ai/docs/genai/reference/adapter.html) - [Extensions](https://onnxruntime.ai/docs/extensions/) - [Add Operators](https://onnxruntime.ai/docs/extensions/add-op.html) - [Build](https://onnxruntime.ai/docs/extensions/build.html) - [Performance](https://onnxruntime.ai/docs/performance/) - [Tune performance](https://onnxruntime.ai/docs/performance/tune-performance/) - [Profiling tools](https://onnxruntime.ai/docs/performance/tune-performance/profiling-tools.html) - [Logging & Tracing](https://onnxruntime.ai/docs/performance/tune-performance/logging_tracing.html) - [Memory consumption](https://onnxruntime.ai/docs/performance/tune-performance/memory.html) - [Thread management](https://onnxruntime.ai/docs/performance/tune-performance/threading.html) - [I/O Binding](https://onnxruntime.ai/docs/performance/tune-performance/iobinding.html) - [Troubleshooting](https://onnxruntime.ai/docs/performance/tune-performance/troubleshooting.html) - [Model optimizations](https://onnxruntime.ai/docs/performance/model-optimizations/) - [Quantize ONNX models](https://onnxruntime.ai/docs/performance/model-optimizations/quantization.html) - [Float16 and mixed precision models](https://onnxruntime.ai/docs/performance/model-optimizations/float16.html) - [Graph optimizations](https://onnxruntime.ai/docs/performance/model-optimizations/graph-optimizations.html) - [ORT model format](https://onnxruntime.ai/docs/performance/model-optimizations/ort-format-models.html) - [ORT model format runtime optimization](https://onnxruntime.ai/docs/performance/model-optimizations/ort-format-model-runtime-optimization.html) - [Transformers optimizer](https://onnxruntime.ai/docs/performance/transformers-optimization.html) - [End to end optimization with Olive](https://onnxruntime.ai/docs/performance/olive.html) - [Device tensors](https://onnxruntime.ai/docs/performance/device-tensor.html) - [Ecosystem](https://onnxruntime.ai/docs/ecosystem/) - [Azure Container for PyTorch (ACPT)](https://onnxruntime.ai/docs/ecosystem/acpt.html) - [Reference](https://onnxruntime.ai/docs/reference/) - [Releases](https://onnxruntime.ai/docs/reference/releases-servicing.html) - [Compatibility](https://onnxruntime.ai/docs/reference/compatibility.html) - [Operators](https://onnxruntime.ai/docs/reference/operators/) - [Operator kernels](https://onnxruntime.ai/docs/reference/operators/OperatorKernels.html) - [Contrib operators](https://onnxruntime.ai/docs/reference/operators/ContribOperators.html) - [Custom operators](https://onnxruntime.ai/docs/reference/operators/add-custom-op.html) - [Reduced operator config file](https://onnxruntime.ai/docs/reference/operators/reduced-operator-config-file.html) - [Architecture](https://onnxruntime.ai/docs/reference/high-level-design.html) - [Citing ONNX Runtime](https://onnxruntime.ai/docs/reference/citing.html) - [Dependency Management in ONNX Runtime](https://onnxruntime.ai/docs/build/dependencies.html) - [ONNX Runtime Docs on GitHub](https://github.com/microsoft/onnxruntime/tree/gh-pages) This site uses [Just the Docs](https://github.com/just-the-docs/just-the-docs), a documentation theme for Jekyll. Search onnxruntime - [ONNX Runtime](https://onnxruntime.ai/) - [Install](https://onnxruntime.ai/docs/install/) - [Get Started](https://onnxruntime.ai/docs/get-started/) - [Tutorials](https://onnxruntime.ai/docs/tutorials/) - [API Docs](https://onnxruntime.ai/docs/api/) - [YouTube](https://www.youtube.com/onnxruntime) - [GitHub](https://github.com/microsoft/onnxruntime) # Install ONNX Runtime See the [installation matrix](https://onnxruntime.ai/) for recommended instructions for desired combinations of target operating system, hardware, accelerator, and language. Details on OS versions, compilers, language versions, dependent libraries, etc can be found under [Compatibility](https://onnxruntime.ai/docs/reference/compatibility). ## Contents - [Requirements](https://onnxruntime.ai/docs/install/#requirements) - [CUDA and CuDNN](https://onnxruntime.ai/docs/install/#cuda-and-cudnn) - [Python Installs](https://onnxruntime.ai/docs/install/#python-installs) - [Install ONNX Runtime CPU](https://onnxruntime.ai/docs/install/#install-onnx-runtime-cpu) - [Install nightly](https://onnxruntime.ai/docs/install/#install-nightly) - [Install ONNX Runtime GPU (DirectML) - Sustained Engineering Mode](https://onnxruntime.ai/docs/install/#install-onnx-runtime-gpu-directml---sustained-engineering-mode) - [Install nightly](https://onnxruntime.ai/docs/install/#install-nightly-1) - [Install ONNX Runtime GPU (CUDA or TensorRT)](https://onnxruntime.ai/docs/install/#install-onnx-runtime-gpu-cuda-or-tensorrt) - [CUDA 12.x](https://onnxruntime.ai/docs/install/#cuda-12x) - [Nightly for CUDA 13.x](https://onnxruntime.ai/docs/install/#nightly-for-cuda-13x) - [Nightly for CUDA 12.x](https://onnxruntime.ai/docs/install/#nightly-for-cuda-12x) - [CUDA 11.x](https://onnxruntime.ai/docs/install/#cuda-11x) - [Install ONNX Runtime QNN](https://onnxruntime.ai/docs/install/#install-onnx-runtime-qnn) - [Install nightly](https://onnxruntime.ai/docs/install/#install-nightly-2) - [C\#/C/C++/WinML Installs](https://onnxruntime.ai/docs/install/#cccwinml-installs) - [Install ONNX Runtime](https://onnxruntime.ai/docs/install/#install-onnx-runtime-1) - [Install ONNX Runtime CPU](https://onnxruntime.ai/docs/install/#install-onnx-runtime-cpu-1) - [Install ONNX Runtime GPU (CUDA 12.x)](https://onnxruntime.ai/docs/install/#install-onnx-runtime-gpu-cuda-12x) - [Install ONNX Runtime GPU (CUDA 11.8)](https://onnxruntime.ai/docs/install/#install-onnx-runtime-gpu-cuda-118) - [DirectML (sustained engineering - use WinML for new projects)](https://onnxruntime.ai/docs/install/#directml-sustained-engineering---use-winml-for-new-projects) - [WinML (recommended for Windows)](https://onnxruntime.ai/docs/install/#winml-recommended-for-windows) - [Install on web and mobile](https://onnxruntime.ai/docs/install/#install-on-web-and-mobile) - [JavaScript Installs](https://onnxruntime.ai/docs/install/#javascript-installs) - [Install ONNX Runtime Web (browsers)](https://onnxruntime.ai/docs/install/#install-onnx-runtime-web-browsers) - [Install ONNX Runtime Node.js binding (Node.js)](https://onnxruntime.ai/docs/install/#install-onnx-runtime-nodejs-binding-nodejs) - [Install ONNX Runtime for React Native](https://onnxruntime.ai/docs/install/#install-onnx-runtime-for-react-native) - [Install on iOS](https://onnxruntime.ai/docs/install/#install-on-ios) - [C/C++](https://onnxruntime.ai/docs/install/#cc) - [Objective-C](https://onnxruntime.ai/docs/install/#objective-c) - [Custom build](https://onnxruntime.ai/docs/install/#custom-build) - [Install on Android](https://onnxruntime.ai/docs/install/#install-on-android) - [Java/Kotlin](https://onnxruntime.ai/docs/install/#javakotlin) - [C/C++](https://onnxruntime.ai/docs/install/#cc-1) - [Custom build](https://onnxruntime.ai/docs/install/#custom-build-1) - [Install for On-Device Training](https://onnxruntime.ai/docs/install/#install-for-on-device-training) - [Offline Phase - Prepare for Training](https://onnxruntime.ai/docs/install/#offline-phase---prepare-for-training) - [Training Phase - On-Device Training](https://onnxruntime.ai/docs/install/#training-phase---on-device-training) - [Large Model Training](https://onnxruntime.ai/docs/install/#large-model-training) - [Inference install table for all languages](https://onnxruntime.ai/docs/install/#inference-install-table-for-all-languages) - [Training install table for all languages](https://onnxruntime.ai/docs/install/#training-install-table-for-all-languages) ## Requirements - All builds require the English language package with `en_US.UTF-8` locale. On Linux, install [language-pack-en package](https://packages.ubuntu.com/search?keywords=language-pack-en) by running `locale-gen en_US.UTF-8` and `update-locale LANG=en_US.UTF-8` - Windows builds require [Visual C++ 2019 runtime](https://learn.microsoft.com/en-us/cpp/windows/latest-supported-vc-redist?view=msvc-170#latest-microsoft-visual-c-redistributable-version). The latest version is recommended. ### CUDA and CuDNN For ONNX Runtime GPU package, it is required to install [CUDA](https://developer.nvidia.com/cuda-toolkit) and [cuDNN](https://developer.nvidia.com/cudnn). Check [CUDA execution provider requirements](https://onnxruntime.ai/docs/execution-providers/CUDA-ExecutionProvider.html#requirements) for compatible version of CUDA and cuDNN. - Zlib is required by cuDNN 9.x for Linux only (zlib is statically linked into the cuDNN 9.x Windows dynamic libraries), or cuDNN 8.x for Linux and Windows. Follow the [cuDNN 8.9 installation guide](https://docs.nvidia.com/deeplearning/cudnn/archives/cudnn-890/install-guide/index.html) to install zlib in Linux or Windows. - In Windows, the path of CUDA `bin` and cuDNN `bin` directories must be added to the `PATH` environment variable. - In Linux, the path of CUDA `lib64` and cuDNN `lib` directories must be added to the `LD_LIBRARY_PATH` environment variable. For `onnxruntime-gpu` package, it is possible to work with PyTorch without the need for manual installations of CUDA or cuDNN. Refer to [Compatibility with PyTorch](https://onnxruntime.ai/docs/execution-providers/CUDA-ExecutionProvider.html#compatibility-with-pytorch) for more information. ## Python Installs ### Install ONNX Runtime CPU ``` pip install onnxruntime ``` #### Install nightly ``` pip install coloredlogs flatbuffers numpy packaging protobuf sympy pip install --pre --index-url https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/ORT-Nightly/pypi/simple/ onnxruntime ``` ### Install ONNX Runtime GPU (DirectML) - Sustained Engineering Mode Note: DirectML is in sustained engineering. For new Windows projects, consider [WinML](https://onnxruntime.ai/docs/install/#winml-recommended-for-windows) instead. ``` pip install onnxruntime-directml ``` #### Install nightly ``` pip install coloredlogs flatbuffers numpy packaging protobuf sympy pip install --pre --index-url https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/ORT-Nightly/pypi/simple/ onnxruntime-directml ``` ### Install ONNX Runtime GPU (CUDA or TensorRT) #### CUDA 12.x The default CUDA version for [onnxruntime-gpu in pypi](https://pypi.org/project/onnxruntime-gpu) is 12.x since 1.19.0. ``` pip install onnxruntime-gpu ``` For previous versions, you can download here: [1\.18.1](https://aiinfra.visualstudio.com/PublicPackages/_artifacts/feed/onnxruntime-cuda-12/PyPI/onnxruntime-gpu/overview/1.18.1), [1\.18.0](https://aiinfra.visualstudio.com/PublicPackages/_artifacts/feed/onnxruntime-cuda-12/PyPI/onnxruntime-gpu/overview/1.18.0) #### Nightly for CUDA 13.x ``` pip install coloredlogs flatbuffers numpy packaging protobuf sympy pip install --pre --index-url https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/ort-cuda-13-nightly/pypi/simple/ onnxruntime-gpu ``` #### Nightly for CUDA 12.x ``` pip install coloredlogs flatbuffers numpy packaging protobuf sympy pip install --pre --index-url https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/ORT-Nightly/pypi/simple/ onnxruntime-gpu ``` #### CUDA 11.x For Cuda 11.x, please use the following instructions to install from [ORT Azure Devops Feed](https://aiinfra.visualstudio.com/PublicPackages/_artifacts/feed/onnxruntime-cuda-11/PyPI/onnxruntime-gpu/overview) for 1.19.2 or later. ``` pip install coloredlogs flatbuffers numpy packaging protobuf sympy pip install onnxruntime-gpu --index-url https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/onnxruntime-cuda-11/pypi/simple/ ``` For previous versions, you can download here: [1\.18.1](https://pypi.org/project/onnxruntime-gpu/1.18.1/), [1\.18.0](https://pypi.org/project/onnxruntime-gpu/1.18.0/) ### Install ONNX Runtime QNN ``` pip install onnxruntime-qnn ``` #### Install nightly ``` pip install coloredlogs flatbuffers numpy packaging protobuf sympy pip install --pre --index-url https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/ORT-Nightly/pypi/simple/ onnxruntime-qnn ``` ## C\#/C/C++/WinML Installs ### Install ONNX Runtime #### Install ONNX Runtime CPU ``` # CPU dotnet add package Microsoft.ML.OnnxRuntime ``` #### Install ONNX Runtime GPU (CUDA 12.x) The default CUDA version for ORT is 12.x ``` # GPU dotnet add package Microsoft.ML.OnnxRuntime.Gpu ``` #### Install ONNX Runtime GPU (CUDA 11.8) 1. Project Setup Ensure you have installed the latest version of the Azure Artifacts keyring from the its [Github Repo](https://github.com/microsoft/artifacts-credprovider#azure-artifacts-credential-provider). Add a nuget.config file to your project in the same directory as your .csproj file. ``` <?xml version="1.0" encoding="utf-8"?> <configuration> <packageSources> <clear/> <add key="onnxruntime-cuda-11" value="https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/onnxruntime-cuda-11/nuget/v3/index.json"/> </packageSources> </configuration> ``` 1. Restore packages Restore packages (using the interactive flag, which allows dotnet to prompt you for credentials) ``` dotnet add package Microsoft.ML.OnnxRuntime.Gpu ``` Note: You don’t need –interactive every time. dotnet will prompt you to add –interactive if it needs updated credentials. #### DirectML (sustained engineering - use WinML for new projects) ``` dotnet add package Microsoft.ML.OnnxRuntime.DirectML ``` Note: DirectML is in sustained engineering. For new Windows projects, use WinML instead: #### WinML (recommended for Windows) ``` dotnet add package Microsoft.AI.MachineLearning ``` ## Install on web and mobile The pre-built packages have full support for all ONNX opsets and operators. If the pre-built package is too large, you can create a [custom build](https://onnxruntime.ai/docs/build/custom.html). A custom build can include just the opsets and operators in your model/s to reduce the size. ### JavaScript Installs #### Install ONNX Runtime Web (browsers) ``` # install latest release version npm install onnxruntime-web # install nightly build dev version npm install onnxruntime-web@dev ``` #### Install ONNX Runtime Node.js binding (Node.js) ``` # install latest release version npm install onnxruntime-node ``` #### Install ONNX Runtime for React Native ``` # install latest release version npm install onnxruntime-react-native ``` ### Install on iOS In your CocoaPods `Podfile`, add the `onnxruntime-c` or `onnxruntime-objc` pod, depending on which API you want to use. #### C/C++ ``` use_frameworks! pod 'onnxruntime-c' ``` #### Objective-C ``` use_frameworks! pod 'onnxruntime-objc' ``` Run `pod install`. #### Custom build Refer to the instructions for creating a [custom iOS package](https://onnxruntime.ai/docs/build/custom.html#ios). ### Install on Android #### Java/Kotlin In your Android Studio Project, make the following changes to: 1. build.gradle (Project): ``` repositories { mavenCentral() } ``` 2. build.gradle (Module): ``` dependencies { implementation 'com.microsoft.onnxruntime:onnxruntime-android:latest.release' } ``` #### C/C++ Download the [onnxruntime-android](https://mvnrepository.com/artifact/com.microsoft.onnxruntime/onnxruntime-android) AAR hosted at MavenCentral, change the file extension from `.aar` to `.zip`, and unzip it. Include the header files from the `headers` folder, and the relevant `libonnxruntime.so` dynamic library from the `jni` folder in your NDK project. #### Custom build Refer to the instructions for creating a [custom Android package](https://onnxruntime.ai/docs/build/custom.html#android). ## Install for On-Device Training Unless stated otherwise, the installation instructions in this section refer to pre-built packages designed to perform on-device training. If the pre-built training package supports your model but is too large, you can create a [custom training build](https://onnxruntime.ai/docs/build/custom.html). ### Offline Phase - Prepare for Training ``` python -m pip install cerberus flatbuffers h5py numpy>=1.16.6 onnx packaging protobuf sympy setuptools>=41.4.0 pip install -i https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/ORT/pypi/simple/ onnxruntime-training-cpu ``` ### Training Phase - On-Device Training \| Device \| Language \| PackageName \| Installation Instructions \| \|---\|---\|---\|---\| \| Windows \| C, C++, C\# \| [Microsoft.ML.OnnxRuntime.Training](https://www.nuget.org/packages/Microsoft.ML.OnnxRuntime) \| `dotnet add package Microsoft.ML.OnnxRuntime.Training` \| \| Linux \| C, C++ \| [onnxruntime-training-linux\.tgz](https://github.com/microsoft/onnxruntime/releases) \| Download the `.tgz` file from [here](https://github.com/microsoft/onnxruntime/releases). Extract it. Move and include the header files in the `include` directory. Move the `libonnxruntime.so` dynamic library to a desired path and include it. \| \| \| Python \| [onnxruntime-training](https://pypi.org/project/onnxruntime-training/) \| `pip install onnxruntime-training` \| \| Android \| C, C++ \| [onnxruntime-training-android](https://mvnrepository.com/artifact/com.microsoft.onnxruntime/onnxruntime-training-android) \| Download the [onnxruntime-training-android (full package)](https://mvnrepository.com/artifact/com.microsoft.onnxruntime/onnxruntime-android) AAR hosted at Maven Central. Change the file extension from `.aar` to `.zip`, and unzip it. Include the header files from the `headers` folder. Include the relevant `libonnxruntime.so` dynamic library from the `jni` folder in your NDK project. \| \| \| Java/Kotlin \| [onnxruntime-training-android](https://mvnrepository.com/artifact/com.microsoft.onnxruntime/onnxruntime-android) \| In your Android Studio Project, make the following changes to: build.gradle (Project): ` repositories { mavenCentral() }` build.gradle (Module): ` dependencies { implementation 'com.microsoft.onnxruntime:onnxruntime-training-android:latest.release' }` \| \| iOS \| C, C++ \| CocoaPods: onnxruntime-training-c \| In your CocoaPods `Podfile`, add the `onnxruntime-training-c` pod: Run `pod install`. \| \| \| Objective-C \| CocoaPods: onnxruntime-training-objc \| In your CocoaPods `Podfile`, add the `onnxruntime-training-objc` pod: Run `pod install`. \| \| Web \| JavaScript, TypeScript \| onnxruntime-web \| Use either `import * as ort from 'onnxruntime-web/training';` or `const ort = require('onnxruntime-web/training');` \| ## Large Model Training ``` pip install torch-ort python -m torch_ort.configure ``` Note: This installs the default version of the `torch-ort` and `onnxruntime-training` packages that are mapped to specific versions of the CUDA libraries. Refer to the install options in [onnxruntime.ai](https://onnxruntime.ai/). ## Inference install table for all languages The table below lists the build variants available as officially supported packages. Others can be [built from source](https://onnxruntime.ai/docs/build/inferencing) from each [release branch](https://github.com/microsoft/onnxruntime/tags). In addition to general [requirements](https://onnxruntime.ai/docs/install/#requirements), please note additional requirements and dependencies in the table below: \| \| Official build \| Nightly build \| Reqs \| \|---\|---\|---\|---\| \| Python \| If using pip, run `pip install --upgrade pip` prior to downloading. \| \| \| \| \| CPU: [onnxruntime](https://pypi.org/project/onnxruntime) \| [onnxruntime (nightly)](https://aiinfra.visualstudio.com/PublicPackages/_artifacts/feed/ORT-Nightly/PyPI/onnxruntime/overview) \| \| \| \| GPU (CUDA/TensorRT) for CUDA 12.x: [onnxruntime-gpu](https://pypi.org/project/onnxruntime-gpu) \| [onnxruntime-gpu (nightly)](https://aiinfra.visualstudio.com/PublicPackages/_artifacts/feed/ORT-Nightly/PyPI/onnxruntime-gpu/overview/) \| [View](https://onnxruntime.ai/docs/execution-providers/CUDA-ExecutionProvider.html#requirements) \| \| \| GPU (DirectML) sustained engineering: [onnxruntime-directml](https://pypi.org/project/onnxruntime-directml/) \| [onnxruntime-directml (nightly)](https://aiinfra.visualstudio.com/PublicPackages/_artifacts/feed/ORT-Nightly/PyPI/onnxruntime-directml/overview/) \| [View](https://onnxruntime.ai/docs/execution-providers/DirectML-ExecutionProvider.html#requirements) \| \| \| OpenVINO: [intel/onnxruntime](https://github.com/intel/onnxruntime/releases/latest) - Intel managed \| \| [View](https://onnxruntime.ai/docs/build/eps.html#openvino) \| \| \| TensorRT (Jetson): [Jetson Zoo](https://elinux.org/Jetson_Zoo#ONNX_Runtime) - NVIDIA managed \| \| \| \| \| Azure (Cloud): [onnxruntime-azure](https://pypi.org/project/onnxruntime-azure/) \| \| \| \| C\#/C/C++ \| CPU: [Microsoft.ML.OnnxRuntime](https://www.nuget.org/packages/Microsoft.ML.OnnxRuntime) \| [onnxruntime (nightly)](https://aiinfra.visualstudio.com/PublicPackages/_packaging?_a=feed&feed=ORT-Nightly) \| \| \| \| GPU (CUDA/TensorRT): [Microsoft.ML.OnnxRuntime.Gpu](https://www.nuget.org/packages/Microsoft.ML.OnnxRuntime.gpu) \| [onnxruntime (nightly)](https://aiinfra.visualstudio.com/PublicPackages/_packaging?_a=feed&feed=ORT-Nightly) \| [View](https://onnxruntime.ai/docs/execution-providers/CUDA-ExecutionProvider) \| \| \| GPU (DirectML) sustained engineering: [Microsoft.ML.OnnxRuntime.DirectML](https://www.nuget.org/packages/Microsoft.ML.OnnxRuntime.DirectML) \| [onnxruntime (nightly)](https://aiinfra.visualstudio.com/PublicPackages/_artifacts/feed/ORT-Nightly/PyPI/ort-nightly-directml/overview) \| [View](https://onnxruntime.ai/docs/execution-providers/DirectML-ExecutionProvider) \| \| WinML recommended for Windows \| [Microsoft.AI.MachineLearning](https://www.nuget.org/packages/Microsoft.AI.MachineLearning) \| [onnxruntime (nightly)](https://aiinfra.visualstudio.com/PublicPackages/_artifacts/feed/ORT-Nightly/NuGet/Microsoft.AI.MachineLearning/overview) \| [View](https://docs.microsoft.com/en-us/windows/ai/windows-ml/port-app-to-nuget#prerequisites) \| \| Java \| CPU: [com.microsoft.onnxruntime:onnxruntime](https://search.maven.org/artifact/com.microsoft.onnxruntime/onnxruntime) \| \| [View](https://onnxruntime.ai/docs/api/java) \| \| \| GPU (CUDA/TensorRT): [com.microsoft.onnxruntime:onnxruntime\_gpu](https://search.maven.org/artifact/com.microsoft.onnxruntime/onnxruntime_gpu) \| \| [View](https://onnxruntime.ai/docs/api/java) \| \| Android \| [com.microsoft.onnxruntime:onnxruntime-android](https://search.maven.org/artifact/com.microsoft.onnxruntime/onnxruntime-android) \| \| [View](https://onnxruntime.ai/docs/install/#install-on-android) \| \| iOS (C/C++) \| CocoaPods: onnxruntime-c \| \| [View](https://onnxruntime.ai/docs/install/#install-on-ios) \| \| Objective-C \| CocoaPods: onnxruntime-objc \| \| [View](https://onnxruntime.ai/docs/install/#install-on-ios) \| \| React Native \| [onnxruntime-react-native (latest)](https://www.npmjs.com/package/onnxruntime-react-native) \| [onnxruntime-react-native (dev)](https://www.npmjs.com/package/onnxruntime-react-native?activeTab=versions) \| [View](https://onnxruntime.ai/docs/api/js) \| \| Node.js \| [onnxruntime-node (latest)](https://www.npmjs.com/package/onnxruntime-node) \| [onnxruntime-node (dev)](https://www.npmjs.com/package/onnxruntime-node?activeTab=versions) \| [View](https://onnxruntime.ai/docs/api/js) \| \| Web \| [onnxruntime-web (latest)](https://www.npmjs.com/package/onnxruntime-web) \| [onnxruntime-web (dev)](https://www.npmjs.com/package/onnxruntime-web?activeTab=versions) \| [View](https://onnxruntime.ai/docs/api/js) \| Note: Nightly builds created from the main branch are available for testing newer changes between official releases. Please use these at your own risk. We strongly advise against deploying these to production workloads as support is limited for nightly builds. ## Training install table for all languages Refer to the getting started with [Optimized Training](https://onnxruntime.ai/getting-started) page for more fine-grained installation instructions. *** For documentation questions, please [file an issue](https://github.com/microsoft/onnxruntime/issues/new?assignees=&labels=documentation&projects=&template=02-documentation.yml&title=%5BDocumentation%5D+). [Edit this page on GitHub](https://github.com/microsoft/onnxruntime/tree/gh-pages/docs/install/index.md) This site uses [Just the Docs](https://github.com/just-the-docs/just-the-docs), a documentation theme for Jekyll.
Readable Markdown	See the [installation matrix](https://onnxruntime.ai/) for recommended instructions for desired combinations of target operating system, hardware, accelerator, and language. Details on OS versions, compilers, language versions, dependent libraries, etc can be found under [Compatibility](https://onnxruntime.ai/docs/reference/compatibility). ## Contents - [Requirements](https://onnxruntime.ai/docs/install/#requirements) - [CUDA and CuDNN](https://onnxruntime.ai/docs/install/#cuda-and-cudnn) - [Python Installs](https://onnxruntime.ai/docs/install/#python-installs) - [Install ONNX Runtime CPU](https://onnxruntime.ai/docs/install/#install-onnx-runtime-cpu) - [Install nightly](https://onnxruntime.ai/docs/install/#install-nightly) - [Install ONNX Runtime GPU (DirectML) - Sustained Engineering Mode](https://onnxruntime.ai/docs/install/#install-onnx-runtime-gpu-directml---sustained-engineering-mode) - [Install nightly](https://onnxruntime.ai/docs/install/#install-nightly-1) - [Install ONNX Runtime GPU (CUDA or TensorRT)](https://onnxruntime.ai/docs/install/#install-onnx-runtime-gpu-cuda-or-tensorrt) - [CUDA 12.x](https://onnxruntime.ai/docs/install/#cuda-12x) - [Nightly for CUDA 13.x](https://onnxruntime.ai/docs/install/#nightly-for-cuda-13x) - [Nightly for CUDA 12.x](https://onnxruntime.ai/docs/install/#nightly-for-cuda-12x) - [CUDA 11.x](https://onnxruntime.ai/docs/install/#cuda-11x) - [Install ONNX Runtime QNN](https://onnxruntime.ai/docs/install/#install-onnx-runtime-qnn) - [Install nightly](https://onnxruntime.ai/docs/install/#install-nightly-2) - [C\#/C/C++/WinML Installs](https://onnxruntime.ai/docs/install/#cccwinml-installs) - [Install ONNX Runtime](https://onnxruntime.ai/docs/install/#install-onnx-runtime-1) - [Install ONNX Runtime CPU](https://onnxruntime.ai/docs/install/#install-onnx-runtime-cpu-1) - [Install ONNX Runtime GPU (CUDA 12.x)](https://onnxruntime.ai/docs/install/#install-onnx-runtime-gpu-cuda-12x) - [Install ONNX Runtime GPU (CUDA 11.8)](https://onnxruntime.ai/docs/install/#install-onnx-runtime-gpu-cuda-118) - [DirectML (sustained engineering - use WinML for new projects)](https://onnxruntime.ai/docs/install/#directml-sustained-engineering---use-winml-for-new-projects) - [WinML (recommended for Windows)](https://onnxruntime.ai/docs/install/#winml-recommended-for-windows) - [Install on web and mobile](https://onnxruntime.ai/docs/install/#install-on-web-and-mobile) - [JavaScript Installs](https://onnxruntime.ai/docs/install/#javascript-installs) - [Install ONNX Runtime Web (browsers)](https://onnxruntime.ai/docs/install/#install-onnx-runtime-web-browsers) - [Install ONNX Runtime Node.js binding (Node.js)](https://onnxruntime.ai/docs/install/#install-onnx-runtime-nodejs-binding-nodejs) - [Install ONNX Runtime for React Native](https://onnxruntime.ai/docs/install/#install-onnx-runtime-for-react-native) - [Install on iOS](https://onnxruntime.ai/docs/install/#install-on-ios) - [C/C++](https://onnxruntime.ai/docs/install/#cc) - [Objective-C](https://onnxruntime.ai/docs/install/#objective-c) - [Custom build](https://onnxruntime.ai/docs/install/#custom-build) - [Install on Android](https://onnxruntime.ai/docs/install/#install-on-android) - [Java/Kotlin](https://onnxruntime.ai/docs/install/#javakotlin) - [C/C++](https://onnxruntime.ai/docs/install/#cc-1) - [Custom build](https://onnxruntime.ai/docs/install/#custom-build-1) - [Install for On-Device Training](https://onnxruntime.ai/docs/install/#install-for-on-device-training) - [Offline Phase - Prepare for Training](https://onnxruntime.ai/docs/install/#offline-phase---prepare-for-training) - [Training Phase - On-Device Training](https://onnxruntime.ai/docs/install/#training-phase---on-device-training) - [Large Model Training](https://onnxruntime.ai/docs/install/#large-model-training) - [Inference install table for all languages](https://onnxruntime.ai/docs/install/#inference-install-table-for-all-languages) - [Training install table for all languages](https://onnxruntime.ai/docs/install/#training-install-table-for-all-languages) ## Requirements - All builds require the English language package with `en_US.UTF-8` locale. On Linux, install [language-pack-en package](https://packages.ubuntu.com/search?keywords=language-pack-en) by running `locale-gen en_US.UTF-8` and `update-locale LANG=en_US.UTF-8` - Windows builds require [Visual C++ 2019 runtime](https://learn.microsoft.com/en-us/cpp/windows/latest-supported-vc-redist?view=msvc-170#latest-microsoft-visual-c-redistributable-version). The latest version is recommended. ### CUDA and CuDNN For ONNX Runtime GPU package, it is required to install [CUDA](https://developer.nvidia.com/cuda-toolkit) and [cuDNN](https://developer.nvidia.com/cudnn). Check [CUDA execution provider requirements](https://onnxruntime.ai/docs/execution-providers/CUDA-ExecutionProvider.html#requirements) for compatible version of CUDA and cuDNN. - Zlib is required by cuDNN 9.x for Linux only (zlib is statically linked into the cuDNN 9.x Windows dynamic libraries), or cuDNN 8.x for Linux and Windows. Follow the [cuDNN 8.9 installation guide](https://docs.nvidia.com/deeplearning/cudnn/archives/cudnn-890/install-guide/index.html) to install zlib in Linux or Windows. - In Windows, the path of CUDA `bin` and cuDNN `bin` directories must be added to the `PATH` environment variable. - In Linux, the path of CUDA `lib64` and cuDNN `lib` directories must be added to the `LD_LIBRARY_PATH` environment variable. For `onnxruntime-gpu` package, it is possible to work with PyTorch without the need for manual installations of CUDA or cuDNN. Refer to [Compatibility with PyTorch](https://onnxruntime.ai/docs/execution-providers/CUDA-ExecutionProvider.html#compatibility-with-pytorch) for more information. ## Python Installs ### Install ONNX Runtime CPU ``` pip install onnxruntime ``` #### Install nightly ``` pip install coloredlogs flatbuffers numpy packaging protobuf sympy pip install --pre --index-url https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/ORT-Nightly/pypi/simple/ onnxruntime ``` ### Install ONNX Runtime GPU (DirectML) - Sustained Engineering Mode Note: DirectML is in sustained engineering. For new Windows projects, consider [WinML](https://onnxruntime.ai/docs/install/#winml-recommended-for-windows) instead. ``` pip install onnxruntime-directml ``` #### Install nightly ``` pip install coloredlogs flatbuffers numpy packaging protobuf sympy pip install --pre --index-url https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/ORT-Nightly/pypi/simple/ onnxruntime-directml ``` ### Install ONNX Runtime GPU (CUDA or TensorRT) #### CUDA 12.x The default CUDA version for [onnxruntime-gpu in pypi](https://pypi.org/project/onnxruntime-gpu) is 12.x since 1.19.0. ``` pip install onnxruntime-gpu ``` For previous versions, you can download here: [1\.18.1](https://aiinfra.visualstudio.com/PublicPackages/_artifacts/feed/onnxruntime-cuda-12/PyPI/onnxruntime-gpu/overview/1.18.1), [1\.18.0](https://aiinfra.visualstudio.com/PublicPackages/_artifacts/feed/onnxruntime-cuda-12/PyPI/onnxruntime-gpu/overview/1.18.0) #### Nightly for CUDA 13.x ``` pip install coloredlogs flatbuffers numpy packaging protobuf sympy pip install --pre --index-url https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/ort-cuda-13-nightly/pypi/simple/ onnxruntime-gpu ``` #### Nightly for CUDA 12.x ``` pip install coloredlogs flatbuffers numpy packaging protobuf sympy pip install --pre --index-url https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/ORT-Nightly/pypi/simple/ onnxruntime-gpu ``` #### CUDA 11.x For Cuda 11.x, please use the following instructions to install from [ORT Azure Devops Feed](https://aiinfra.visualstudio.com/PublicPackages/_artifacts/feed/onnxruntime-cuda-11/PyPI/onnxruntime-gpu/overview) for 1.19.2 or later. ``` pip install coloredlogs flatbuffers numpy packaging protobuf sympy pip install onnxruntime-gpu --index-url https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/onnxruntime-cuda-11/pypi/simple/ ``` For previous versions, you can download here: [1\.18.1](https://pypi.org/project/onnxruntime-gpu/1.18.1/), [1\.18.0](https://pypi.org/project/onnxruntime-gpu/1.18.0/) ### Install ONNX Runtime QNN ``` pip install onnxruntime-qnn ``` #### Install nightly ``` pip install coloredlogs flatbuffers numpy packaging protobuf sympy pip install --pre --index-url https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/ORT-Nightly/pypi/simple/ onnxruntime-qnn ``` ## C\#/C/C++/WinML Installs ### Install ONNX Runtime #### Install ONNX Runtime CPU ``` # CPU dotnet add package Microsoft.ML.OnnxRuntime ``` #### Install ONNX Runtime GPU (CUDA 12.x) The default CUDA version for ORT is 12.x ``` # GPU dotnet add package Microsoft.ML.OnnxRuntime.Gpu ``` #### Install ONNX Runtime GPU (CUDA 11.8) 1. Project Setup Ensure you have installed the latest version of the Azure Artifacts keyring from the its [Github Repo](https://github.com/microsoft/artifacts-credprovider#azure-artifacts-credential-provider). Add a nuget.config file to your project in the same directory as your .csproj file. ``` <?xml version="1.0" encoding="utf-8"?> <configuration> <packageSources> <clear/> <add key="onnxruntime-cuda-11" value="https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/onnxruntime-cuda-11/nuget/v3/index.json"/> </packageSources> </configuration> ``` 1. Restore packages Restore packages (using the interactive flag, which allows dotnet to prompt you for credentials) ``` dotnet add package Microsoft.ML.OnnxRuntime.Gpu ``` Note: You don’t need –interactive every time. dotnet will prompt you to add –interactive if it needs updated credentials. #### DirectML (sustained engineering - use WinML for new projects) ``` dotnet add package Microsoft.ML.OnnxRuntime.DirectML ``` Note: DirectML is in sustained engineering. For new Windows projects, use WinML instead: #### WinML (recommended for Windows) ``` dotnet add package Microsoft.AI.MachineLearning ``` ## Install on web and mobile The pre-built packages have full support for all ONNX opsets and operators. If the pre-built package is too large, you can create a [custom build](https://onnxruntime.ai/docs/build/custom.html). A custom build can include just the opsets and operators in your model/s to reduce the size. ### JavaScript Installs #### Install ONNX Runtime Web (browsers) ``` # install latest release version npm install onnxruntime-web # install nightly build dev version npm install onnxruntime-web@dev ``` #### Install ONNX Runtime Node.js binding (Node.js) ``` # install latest release version npm install onnxruntime-node ``` #### Install ONNX Runtime for React Native ``` # install latest release version npm install onnxruntime-react-native ``` ### Install on iOS In your CocoaPods `Podfile`, add the `onnxruntime-c` or `onnxruntime-objc` pod, depending on which API you want to use. #### C/C++ ``` use_frameworks! pod 'onnxruntime-c' ``` #### Objective-C ``` use_frameworks! pod 'onnxruntime-objc' ``` Run `pod install`. #### Custom build Refer to the instructions for creating a [custom iOS package](https://onnxruntime.ai/docs/build/custom.html#ios). ### Install on Android #### Java/Kotlin In your Android Studio Project, make the following changes to: 1. build.gradle (Project): ``` repositories { mavenCentral() } ``` 2. build.gradle (Module): ``` dependencies { implementation 'com.microsoft.onnxruntime:onnxruntime-android:latest.release' } ``` #### C/C++ Download the [onnxruntime-android](https://mvnrepository.com/artifact/com.microsoft.onnxruntime/onnxruntime-android) AAR hosted at MavenCentral, change the file extension from `.aar` to `.zip`, and unzip it. Include the header files from the `headers` folder, and the relevant `libonnxruntime.so` dynamic library from the `jni` folder in your NDK project. #### Custom build Refer to the instructions for creating a [custom Android package](https://onnxruntime.ai/docs/build/custom.html#android). ## Install for On-Device Training Unless stated otherwise, the installation instructions in this section refer to pre-built packages designed to perform on-device training. If the pre-built training package supports your model but is too large, you can create a [custom training build](https://onnxruntime.ai/docs/build/custom.html). ### Offline Phase - Prepare for Training ``` python -m pip install cerberus flatbuffers h5py numpy>=1.16.6 onnx packaging protobuf sympy setuptools>=41.4.0 pip install -i https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/ORT/pypi/simple/ onnxruntime-training-cpu ``` ### Training Phase - On-Device Training \| Device \| Language \| PackageName \| Installation Instructions \| \|---\|---\|---\|---\| \| Windows \| C, C++, C\# \| [Microsoft.ML.OnnxRuntime.Training](https://www.nuget.org/packages/Microsoft.ML.OnnxRuntime) \| `dotnet add package Microsoft.ML.OnnxRuntime.Training` \| \| Linux \| C, C++ \| [onnxruntime-training-linux\.tgz](https://github.com/microsoft/onnxruntime/releases) \| Download the `.tgz` file from [here](https://github.com/microsoft/onnxruntime/releases). Extract it. Move and include the header files in the `include` directory. Move the `libonnxruntime.so` dynamic library to a desired path and include it. \| \| \| Python \| [onnxruntime-training](https://pypi.org/project/onnxruntime-training/) \| `pip install onnxruntime-training` \| \| Android \| C, C++ \| [onnxruntime-training-android](https://mvnrepository.com/artifact/com.microsoft.onnxruntime/onnxruntime-training-android) \| Download the [onnxruntime-training-android (full package)](https://mvnrepository.com/artifact/com.microsoft.onnxruntime/onnxruntime-android) AAR hosted at Maven Central. Change the file extension from `.aar` to `.zip`, and unzip it. Include the header files from the `headers` folder. Include the relevant `libonnxruntime.so` dynamic library from the `jni` folder in your NDK project. \| \| \| Java/Kotlin \| [onnxruntime-training-android](https://mvnrepository.com/artifact/com.microsoft.onnxruntime/onnxruntime-android) \| In your Android Studio Project, make the following changes to: build.gradle (Project): ` repositories { mavenCentral() }` build.gradle (Module): ` dependencies { implementation 'com.microsoft.onnxruntime:onnxruntime-training-android:latest.release' }` \| \| iOS \| C, C++ \| CocoaPods: onnxruntime-training-c \| In your CocoaPods `Podfile`, add the `onnxruntime-training-c` pod: Run `pod install`. \| \| \| Objective-C \| CocoaPods: onnxruntime-training-objc \| In your CocoaPods `Podfile`, add the `onnxruntime-training-objc` pod: Run `pod install`. \| \| Web \| JavaScript, TypeScript \| onnxruntime-web \| Use either `import * as ort from 'onnxruntime-web/training';` or `const ort = require('onnxruntime-web/training');` \| ## Large Model Training ``` pip install torch-ort python -m torch_ort.configure ``` Note: This installs the default version of the `torch-ort` and `onnxruntime-training` packages that are mapped to specific versions of the CUDA libraries. Refer to the install options in [onnxruntime.ai](https://onnxruntime.ai/). ## Inference install table for all languages The table below lists the build variants available as officially supported packages. Others can be [built from source](https://onnxruntime.ai/docs/build/inferencing) from each [release branch](https://github.com/microsoft/onnxruntime/tags). In addition to general [requirements](https://onnxruntime.ai/docs/install/#requirements), please note additional requirements and dependencies in the table below: \| \| Official build \| Nightly build \| Reqs \| \|---\|---\|---\|---\| \| Python \| If using pip, run `pip install --upgrade pip` prior to downloading. \| \| \| \| \| CPU: [onnxruntime](https://pypi.org/project/onnxruntime) \| [onnxruntime (nightly)](https://aiinfra.visualstudio.com/PublicPackages/_artifacts/feed/ORT-Nightly/PyPI/onnxruntime/overview) \| \| \| \| GPU (CUDA/TensorRT) for CUDA 12.x: [onnxruntime-gpu](https://pypi.org/project/onnxruntime-gpu) \| [onnxruntime-gpu (nightly)](https://aiinfra.visualstudio.com/PublicPackages/_artifacts/feed/ORT-Nightly/PyPI/onnxruntime-gpu/overview/) \| [View](https://onnxruntime.ai/docs/execution-providers/CUDA-ExecutionProvider.html#requirements) \| \| \| GPU (DirectML) sustained engineering: [onnxruntime-directml](https://pypi.org/project/onnxruntime-directml/) \| [onnxruntime-directml (nightly)](https://aiinfra.visualstudio.com/PublicPackages/_artifacts/feed/ORT-Nightly/PyPI/onnxruntime-directml/overview/) \| [View](https://onnxruntime.ai/docs/execution-providers/DirectML-ExecutionProvider.html#requirements) \| \| \| OpenVINO: [intel/onnxruntime](https://github.com/intel/onnxruntime/releases/latest) - Intel managed \| \| [View](https://onnxruntime.ai/docs/build/eps.html#openvino) \| \| \| TensorRT (Jetson): [Jetson Zoo](https://elinux.org/Jetson_Zoo#ONNX_Runtime) - NVIDIA managed \| \| \| \| \| Azure (Cloud): [onnxruntime-azure](https://pypi.org/project/onnxruntime-azure/) \| \| \| \| C\#/C/C++ \| CPU: [Microsoft.ML.OnnxRuntime](https://www.nuget.org/packages/Microsoft.ML.OnnxRuntime) \| [onnxruntime (nightly)](https://aiinfra.visualstudio.com/PublicPackages/_packaging?_a=feed&feed=ORT-Nightly) \| \| \| \| GPU (CUDA/TensorRT): [Microsoft.ML.OnnxRuntime.Gpu](https://www.nuget.org/packages/Microsoft.ML.OnnxRuntime.gpu) \| [onnxruntime (nightly)](https://aiinfra.visualstudio.com/PublicPackages/_packaging?_a=feed&feed=ORT-Nightly) \| [View](https://onnxruntime.ai/docs/execution-providers/CUDA-ExecutionProvider) \| \| \| GPU (DirectML) sustained engineering: [Microsoft.ML.OnnxRuntime.DirectML](https://www.nuget.org/packages/Microsoft.ML.OnnxRuntime.DirectML) \| [onnxruntime (nightly)](https://aiinfra.visualstudio.com/PublicPackages/_artifacts/feed/ORT-Nightly/PyPI/ort-nightly-directml/overview) \| [View](https://onnxruntime.ai/docs/execution-providers/DirectML-ExecutionProvider) \| \| WinML recommended for Windows \| [Microsoft.AI.MachineLearning](https://www.nuget.org/packages/Microsoft.AI.MachineLearning) \| [onnxruntime (nightly)](https://aiinfra.visualstudio.com/PublicPackages/_artifacts/feed/ORT-Nightly/NuGet/Microsoft.AI.MachineLearning/overview) \| [View](https://docs.microsoft.com/en-us/windows/ai/windows-ml/port-app-to-nuget#prerequisites) \| \| Java \| CPU: [com.microsoft.onnxruntime:onnxruntime](https://search.maven.org/artifact/com.microsoft.onnxruntime/onnxruntime) \| \| [View](https://onnxruntime.ai/docs/api/java) \| \| \| GPU (CUDA/TensorRT): [com.microsoft.onnxruntime:onnxruntime\_gpu](https://search.maven.org/artifact/com.microsoft.onnxruntime/onnxruntime_gpu) \| \| [View](https://onnxruntime.ai/docs/api/java) \| \| Android \| [com.microsoft.onnxruntime:onnxruntime-android](https://search.maven.org/artifact/com.microsoft.onnxruntime/onnxruntime-android) \| \| [View](https://onnxruntime.ai/docs/install/#install-on-android) \| \| iOS (C/C++) \| CocoaPods: onnxruntime-c \| \| [View](https://onnxruntime.ai/docs/install/#install-on-ios) \| \| Objective-C \| CocoaPods: onnxruntime-objc \| \| [View](https://onnxruntime.ai/docs/install/#install-on-ios) \| \| React Native \| [onnxruntime-react-native (latest)](https://www.npmjs.com/package/onnxruntime-react-native) \| [onnxruntime-react-native (dev)](https://www.npmjs.com/package/onnxruntime-react-native?activeTab=versions) \| [View](https://onnxruntime.ai/docs/api/js) \| \| Node.js \| [onnxruntime-node (latest)](https://www.npmjs.com/package/onnxruntime-node) \| [onnxruntime-node (dev)](https://www.npmjs.com/package/onnxruntime-node?activeTab=versions) \| [View](https://onnxruntime.ai/docs/api/js) \| \| Web \| [onnxruntime-web (latest)](https://www.npmjs.com/package/onnxruntime-web) \| [onnxruntime-web (dev)](https://www.npmjs.com/package/onnxruntime-web?activeTab=versions) \| [View](https://onnxruntime.ai/docs/api/js) \| Note: Nightly builds created from the main branch are available for testing newer changes between official releases. Please use these at your own risk. We strongly advise against deploying these to production workloads as support is limited for nightly builds. ## Training install table for all languages Refer to the getting started with [Optimized Training](https://onnxruntime.ai/getting-started) page for more fine-grained installation instructions.
Shard	57 (laksa)
Root Hash	16223958384227938257
Unparsed URL	ai,onnxruntime!/docs/install/ s443