Installation

Florence2-Sharp is distributed on NuGet as Florence2.

dotnet add package Florence2

Or, with the PackageReference syntax in your .csproj:

<ItemGroup>
  <PackageReference Include="Florence2" Version="*" />
</ItemGroup>

Target frameworks

What gets installed

• Florence2.dll — the managed library.
• The CPU ONNX Runtime NuGet package, which ships native binaries for Windows, Linux, and macOS.
• Tokeniser resources embedded in the assembly.

The model files themselves are not included in the package — they are downloaded on first run (see Quick Start).

GPU acceleration (optional)

For CUDA, DirectML, or CoreML acceleration, replace the bundled CPU runtime with the matching ONNX Runtime package:

dotnet add package Microsoft.ML.OnnxRuntime.Gpu       # CUDA (Windows, Linux)
dotnet add package Microsoft.ML.OnnxRuntime.DirectML  # Windows DirectML

You don't need to change any Florence2-Sharp code — the runtime picks the best available execution provider automatically.

Verify the install

using Florence2;

Console.WriteLine($"Florence2 task count: {Enum.GetValues<TaskTypes>().Length}");

If that compiles and prints 15, the package is ready. Continue with the Quick Start.