tf.NET 0.2.3.1

.NET Standard 2.0 .NET Framework 4.5

dotnet add package tf.NET --version 0.2.3.1

NuGet\Install-Package tf.NET -Version 0.2.3.1

This command is intended to be used within the Package Manager Console in Visual Studio, as it uses the NuGet module's version of Install-Package.

<PackageReference Include="tf.NET" Version="0.2.3.1" />

For projects that support PackageReference, copy this XML node into the project file to reference the package.

paket add tf.NET --version 0.2.3.1

The NuGet Team does not provide support for this client. Please contact its maintainers for support.

#r "nuget: tf.NET, 0.2.3.1"

#r directive can be used in F# Interactive and Polyglot Notebooks. Copy this into the interactive tool or source code of the script to reference the package.

// Install tf.NET as a Cake Addin
#addin nuget:?package=tf.NET&version=0.2.3.1

// Install tf.NET as a Cake Tool
#tool nuget:?package=tf.NET&version=0.2.3.1

The NuGet Team does not provide support for this client. Please contact its maintainers for support.

tf.NET

About

.NET bindings for the TensorFlow 2 C API. These are low-level bindings suitable for creating higher-level idiomatic interfaces to TensorFlow 2 in .NET languages like C# and F#.

Implementation

C# bindings autogenerated from the TensorFlow C header files using CppSharp.
C# op definitions autogenerated from the TensorFlow 2 runtime op definitions. The generator is based on the one created by Miguel de Icaza for TensorFlowSharp.
Tensor object implementation that uses zero-copy reinterpretation of existing .NET multi-dimensional arrays as tensor data.

Differences

The main differences between this and TensorFlow.NET and TensorSharp and other .NET TensorFlow projects are:

tf.NET supports TensorFlow 2 only. TensorFlow.NET and other libraries currently only support TF 1.
tf.NET does not try to implement any higher-level APIs for ML or things like that. I made this because I wanted a low-level direct binding to the TensorFlow 2 API that I could use to create my own .NET scientific computing language in F#. Looking at the code for TensorFlow.NET, it seems to me that it mixes a lot of higher-level C# code and the bindings together. tf.NET has implementations of essential memory-backed objects like buffers and tensors but usage of these implementations is purely optional.

Installation

Available from NuGet

PM> Install-Package tf.NET

Note that you will still need the TensorFlow 2 native runtime library (i.e. libtensorflow) for your OS. There doesn't seem to be an official Google libtensorflow package for version 2 as yet. For Windows you can build it yourself or check out this package which installs a CPU-only TF 2 native library dependency into your .NET project. For Linux you can download nightly builds of libtensorflow from the TensorFlow Github page.

For usage examples see the unit tests.

Product	Versions
.NET	net5.0 net5.0-windows net6.0 net6.0-android net6.0-ios net6.0-maccatalyst net6.0-macos net6.0-tvos net6.0-windows net7.0 net7.0-android net7.0-ios net7.0-maccatalyst net7.0-macos net7.0-tvos net7.0-windows
.NET Core	netcoreapp2.0 netcoreapp2.1 netcoreapp2.2 netcoreapp3.0 netcoreapp3.1
.NET Standard	netstandard2.0 netstandard2.1
.NET Framework	net45 net451 net452 net46 net461 net462 net463 net47 net471 net472 net48 net481
MonoAndroid	monoandroid
MonoMac	monomac
MonoTouch	monotouch
Tizen	tizen40 tizen60
Xamarin.iOS	xamarinios
Xamarin.Mac	xamarinmac
Xamarin.TVOS	xamarintvos
Xamarin.WatchOS	xamarinwatchos