Class OnnxExporter

Namespace

AiDotNet.Onnx

Assembly

AiDotNet.dll

Exports AiDotNet models to the ONNX format.

public static class OnnxExporter

Inheritance

object

OnnxExporter

Inherited Members

object.Equals(object)

object.Equals(object, object)

object.GetHashCode()

object.GetType()

object.MemberwiseClone()

object.ReferenceEquals(object, object)

object.ToString()

Remarks

This exporter supports sequential models with common layer types:

• Dense/Linear layers (exported as MatMul + Add)
• Activation functions (ReLU, Sigmoid, Tanh, etc.)
• Dropout (exported as Identity in inference mode)

Limitations: This is a proof-of-concept implementation that works with specific, known model structures (sequential models with supported layers). For production use, consider using framework-native export tools.

For Beginners: Use this class to convert your trained AiDotNet models to ONNX format for deployment:

var model = // your trained model
OnnxExporter.Export(model, "model.onnx");

Methods

ExportToBytes<T, TInput, TOutput>(IFullModel<T, TInput, TOutput>, int[]?)

Exports a model to ONNX format and returns the bytes.

public static byte[] ExportToBytes<T, TInput, TOutput>(IFullModel<T, TInput, TOutput> model, int[]? inputShape = null)

Parameters

model IFullModel<T, TInput, TOutput>

The model to export.

inputShape int[]

Optional input shape.

Returns

byte[]

The ONNX model as a byte array.

Type Parameters

T

The numeric type of the model.

TInput

TOutput

Export<T, TInput, TOutput>(IFullModel<T, TInput, TOutput>, string, int[]?)

Exports a model to ONNX format.

public static void Export<T, TInput, TOutput>(IFullModel<T, TInput, TOutput> model, string outputPath, int[]? inputShape = null)

Parameters

model IFullModel<T, TInput, TOutput>

The model to export.

outputPath string

The path to write the ONNX file.

inputShape int[]

Optional input shape. If not provided, will try to infer from model.

Type Parameters

T

The numeric type of the model.

TInput

TOutput

Exceptions

ArgumentNullException

If model or outputPath is null.

NotSupportedException

If the model contains unsupported layer types.