What	Links and Status
OnnxSharp
dotnet-onnx

OnnxSharp library and dotnet-onnx tool

ONNX format parsing and manipulation in C# and with command line .NET tool.

Quick Guide

Install latest version of .NET:

• PowerShell (Windows): https://dot.net/v1/dotnet-install.ps1
• Bash (Linux/macOS): https://dot.net/v1/dotnet-install.sh

Code

What	How
Install	dotnet add PROJECT.csproj package OnnxSharp
Parse	var model = ModelProto.Parser.ParseFromFile("mnist-8.onnx");
Info	var info = model.Graph.Info();
Clean	model.Graph.Clean();
SetDim	model.Graph.SetDim();
Write	model.WriteToFile("mnist-8-clean-dynamic.onnx");

Tool

What	How
Install	dotnet tool install dotnet-onnx -g
Info	dotnet onnx info mnist-8.onnx
Info	dotnet onnx info mnist-8.onnx
Clean	dotnet onnx clean mnist-8.onnx mnist-8-clean.onnx
SetDim	dotnet onnx setdim mnist-8.onnx mnist-8-setdim.onnx

Source Code

Base functionality is based on:

.\protoc.exe .\onnx.proto3 --csharp_out=OnnxSharp

Everything else written in beautiful C# 9.0 as extensions to this.

Example Code

using Onnx;

// Examples see https://github.com/onnx/models
var onnxInputFilePath = @"mnist-8.onnx";

var model = ModelProto.Parser.ParseFromFile(onnxInputFilePath);

var graph = model.Graph;
// Clean graph e.g. remove initializers from inputs that may prevent constant folding
graph.Clean();
// Set dimension in graph to enable dynamic batch size during inference
graph.SetDim(dimIndex: 0, DimParamOrValue.New("N"));
// Get summarized info about the graph
var info = graph.Info();

System.Console.WriteLine(info);

model.WriteToFile(@"mnist-8-clean-dynamic-batch-size.onnx");