OnnxStack.Core 0.8.0

OnnxStack.Core - Onnx Services for .NET Applications

OnnxStack.Core is a library that provides higher-level ONNX services for use in .NET applications. It offers extensive support for features such as dependency injection, .NET configuration implementations, ASP.NET Core integration, and IHostedService support.

You can configure a model set for runtime, offloading individual models to different devices to make better use of resources or run on lower-end hardware. The first use-case is StableDiffusion; however, it will be expanded, and other model sets, such as object detection and classification, will be added.

Getting Started

OnnxStack.Core can be found via the nuget package manager, download and install it.

PM> Install-Package OnnxStack.Core

.NET Core Registration

You can easily integrate OnnxStack.Core into your application services layer. This registration process sets up the necessary services and loads the appsettings.json configuration.

Example: Registering OnnxStack

builder.Services.AddOnnxStack();

Configuration example

The appsettings.json is the easiest option for configuring model sets. Below is an example of clip tokenizer.

{
	"Logging": {
		"LogLevel": {
			"Default": "Information",
			"Microsoft.AspNetCore": "Warning"
		}
	},
	"AllowedHosts": "*",

	"OnnxStackConfig": {
		"Name": "Clip Tokenizer",
		"TokenizerLimit": 77,
		"ModelConfigurations": [{
			"Type": "Tokenizer",
			"DeviceId": 0,
			"ExecutionProvider": "Cpu",
			"OnnxModelPath": "D:\\Repositories\\stable-diffusion-v1-5\\cliptokenizer.onnx"
		}]
	}
}

Basic C# Example

// From DI
IOnnxModelService _onnxModelService;


// Tokenizer model Example
var text = "Text To Tokenize";
var inputTensor = new DenseTensor<string>(new string[] { text }, new int[] { 1 });
var inputString = new List<NamedOnnxValue>
{
	NamedOnnxValue.CreateFromTensor("string_input", inputTensor)
};

// Create an InferenceSession from the Onnx clip tokenizer.
// Run session and send the input data in to get inference output. 
using (var tokens = _onnxModelService.RunInference(OnnxModelType.Tokenizer, inputString))
{
	var resultTensor = tokens.ToArray();
}

Basic C# Example (No DI)

// Create Configuration
var onnxStackConfig = new OnnxStackConfig
{
    Name = "OnnxStack",
    TokenizerLimit = 77,
    ModelConfigurations = new List<OnnxModelSessionConfig>
    {
        new OnnxModelSessionConfig
        {
            DeviceId = 0,
            ExecutionProvider = ExecutionProvider.DirectML,

            Type = OnnxModelType.Tokenizer,
            OnnxModelPath = "clip_tokenizer.onnx",
        }
    }
};

// Create Service
var onnxModelService = new OnnxModelService(onnxStackConfig);


// Tokenizer model Example
var text = "Text To Tokenize";
var inputTensor = new DenseTensor<string>(new string[] { text }, new int[] { 1 });
var inputString = new List<NamedOnnxValue>
{
    NamedOnnxValue.CreateFromTensor("string_input", inputTensor)
};

// Create an InferenceSession from the Onnx clip tokenizer.
// Run session and send the input data in to get inference output. 
using (var tokens = onnxModelService.RunInference(OnnxModelType.Tokenizer, inputString))
{
    var resultTensor = tokens.ToArray();
}