Vowpal Wabbit

Vowpal Wabbit is an open-source machine learning library with a focus on online learning, contextual bandits, and reinforcement learning.

Installation

dotnet add package VowpalWabbit

Platform Support

Runtime Package	Platform
`VowpalWabbit.runtime.win-x64`	Windows x64
`VowpalWabbit.runtime.linux-x64`	Linux x64
`VowpalWabbit.runtime.osx-x64`	macOS x64
`VowpalWabbit.runtime.osx-arm64`	macOS Apple Silicon

All runtime packages are pulled in automatically when you install the main VowpalWabbit package.

Quick Start

Basic Regression

csharp

using VW;

using (var vw = new VowpalWabbit("--quiet"))
{
    vw.Learn("1 | feature1 feature2");
    vw.Learn("0 | feature3 feature4");

    var prediction = vw.Predict("| feature1 feature2");
}

Contextual Bandits (CB ADF)

csharp

using VW;

using (var vw = new VowpalWabbit("--cb_explore_adf --quiet"))
{
    var example = @"shared | user_age:25
        | action1 sport
        | action2 politics
        | action3 music";

    vw.Learn(example);
    var prediction = vw.Predict(example);
}

Save and Load Models

csharp

using (var vw = new VowpalWabbit("--quiet"))
{
    // ... train ...
    vw.SaveModel("model.vw");
}

using (var vw = new VowpalWabbit("--quiet -i model.vw"))
{
    var prediction = vw.Predict("| feature1 feature2");
}

Documentation

License

BSD-3-Clause