onnx

name	onnx
description	Use when running pre-trained ONNX models for inference in .NET with ONNX Runtime. Covers session management, tensor inputs/outputs, execution providers (CPU/GPU/DirectML), model optimization, and integration with ASP.NET Core. USE FOR: running pre-trained ONNX models in .NET, image classification inference, NLP model inference, cross-framework model deployment (PyTorch/TensorFlow to .NET), GPU-accelerated inference with CUDA or DirectML DO NOT USE FOR: training ML models from scratch (use mlnet or Python), calling cloud-hosted LLMs (use microsoft-extensions-ai or azure-ai-inference), building agent workflows (use agent-framework), custom ML pipelines with feature engineering (use mlnet)
license	MIT
metadata	{"displayName":"ONNX Runtime","author":"Tyler-R-Kendrick","version":"1.0.0"}
compatibility	claude, copilot, cursor
references	[{"title":"ONNX Runtime Documentation","url":"https://onnxruntime.ai"},{"title":"ONNX Runtime GitHub Repository","url":"https://github.com/microsoft/onnxruntime"},{"title":"Microsoft.ML.OnnxRuntime NuGet Package","url":"https://www.nuget.org/packages/Microsoft.ML.OnnxRuntime"}]

ONNX Runtime

Overview

ONNX Runtime is a high-performance inference engine for running models in the Open Neural Network Exchange (ONNX) format. It enables .NET applications to run models trained in PyTorch, TensorFlow, scikit-learn, or any framework that exports to ONNX. The runtime supports multiple execution providers (CPU, CUDA, DirectML, TensorRT) for hardware-accelerated inference with minimal code changes.

NuGet Packages

dotnet add package Microsoft.ML.OnnxRuntime           # CPU inference
dotnet add package Microsoft.ML.OnnxRuntime.Gpu       # CUDA GPU inference
dotnet add package Microsoft.ML.OnnxRuntime.DirectML  # DirectML (Windows GPU)

Basic Inference

using Microsoft.ML.OnnxRuntime;
using Microsoft.ML.OnnxRuntime.Tensors;

using var session = new InferenceSession("model.onnx");

// Inspect model inputs and outputs
foreach (var input in session.InputMetadata)
{
    Console.WriteLine($"Input: {input.Key}, Shape: [{string.Join(",", input.Value.Dimensions)}], Type: {input.Value.ElementType}");
}
foreach (var output in session.OutputMetadata)
{
    Console.WriteLine($"Output: {output.Key}, Shape: [{string.Join(",", output.Value.Dimensions)}], Type: {output.Value.ElementType}");
}

// Create input tensor
var inputData = new float[] { 1.0f, 2.0f, 3.0f, 4.0f };
var inputTensor =  DenseTensor<>(inputData, [] { ,  });

 inputs =  List<NamedOnnxValue>
{
    NamedOnnxValue.CreateFromTensor(, inputTensor)
};


 IDisposableReadOnlyCollection<DisposableNamedOnnxValue> results = session.Run(inputs);
 output = results.First().AsTensor<>();

Console.WriteLine();
Console.WriteLine();