![]() |
VOOZH | about |
dotnet add package Microsoft.DeepDev.TokenizerLib --version 1.3.3
NuGet\Install-Package Microsoft.DeepDev.TokenizerLib -Version 1.3.3
<PackageReference Include="Microsoft.DeepDev.TokenizerLib" Version="1.3.3" />
<PackageVersion Include="Microsoft.DeepDev.TokenizerLib" Version="1.3.3" />Directory.Packages.props
<PackageReference Include="Microsoft.DeepDev.TokenizerLib" />Project file
paket add Microsoft.DeepDev.TokenizerLib --version 1.3.3
#r "nuget: Microsoft.DeepDev.TokenizerLib, 1.3.3"
#:package Microsoft.DeepDev.TokenizerLib@1.3.3
#addin nuget:?package=Microsoft.DeepDev.TokenizerLib&version=1.3.3Install as a Cake Addin
#tool nuget:?package=Microsoft.DeepDev.TokenizerLib&version=1.3.3Install as a Cake Tool
This repo contains C# and Typescript implementation of byte pair encoding(BPE) tokenizer for OpenAI LLMs, it's based on open sourced rust implementation in the OpenAI tiktoken. Both implementation are valuable to run prompt tokenization in .NET and Nodejs environment before feeding prompt into a LLM.
The TokenizerLib is built in .NET Standard 2.0, which can be consumed in projects on any version of .NET later than .NET Core 2.0 or .NET Framework 4.6.1.
You can download and install the nuget package of TokenizerLib here.
Example C# code to use TokenizerLib in your code:
using System.Collections.Generic;
using Microsoft.DeepDev;
var IM_START = "<|im_start|>";
var IM_END = "<|im_end|>";
var specialTokens = new Dictionary<string, int>{
{ IM_START, 100264},
{ IM_END, 100265},
};
var tokenizer = await TokenizerBuilder.CreateByModelNameAsync("gpt-4", specialTokens);
var text = "<|im_start|>Hello World<|im_end|>";
var encoded = tokenizer.Encode(text, new HashSet<string>(specialTokens.Keys));
Console.WriteLine(encoded.Count);
var decoded = tokenizer.Decode(encoded.ToArray());
Console.WriteLine(decoded);
In production setting, you should pre-download the BPE rank file and call TokenizerBuilder.CreateTokenizer API to avoid downloading the BPE rank file on the fly.
You can find the model to encoder and encoder to BPE rank file link mapping in: TokenizerBuilder.cs.
PerfBenchmark result based on :
BenchmarkDotNet=v0.13.3, OS=Windows 11 (10.0.22621.1702)
Intel Core i7-1065G7 CPU 1.30GHz, 1 CPU, 8 logical and 4 physical cores
.NET SDK=7.0.300-preview.23179.2
[Host] : .NET 6.0.16 (6.0.1623.17311), X64 RyuJIT AVX2
DefaultJob : .NET 6.0.16 (6.0.1623.17311), X64 RyuJIT AVX2
| Method | Mean | Error | StdDev |
|------- |--------:|---------:|---------:|
| Encode | 2.414 s | 0.0303 s | 0.0253 s |
Please follow .
We welcome contributions. Please follow .
This project may contain trademarks or logos for projects, products, or services. Authorized use of Microsoft trademarks or logos is subject to and must follow Microsoft's Trademark & Brand Guidelines. Use of Microsoft trademarks or logos in modified versions of this project must not cause confusion or imply Microsoft sponsorship. Any use of third-party trademarks or logos are subject to those third-party's policies.
| Product | Versions Compatible and additional computed target framework versions. |
|---|---|
| .NET | net5.0 net5.0 was computed. net5.0-windows net5.0-windows was computed. net6.0 net6.0 was computed. net6.0-android net6.0-android was computed. net6.0-ios net6.0-ios was computed. net6.0-maccatalyst net6.0-maccatalyst was computed. net6.0-macos net6.0-macos was computed. net6.0-tvos net6.0-tvos was computed. net6.0-windows net6.0-windows was computed. net7.0 net7.0 was computed. net7.0-android net7.0-android was computed. net7.0-ios net7.0-ios was computed. net7.0-maccatalyst net7.0-maccatalyst was computed. net7.0-macos net7.0-macos was computed. net7.0-tvos net7.0-tvos was computed. net7.0-windows net7.0-windows was computed. net8.0 net8.0 was computed. net8.0-android net8.0-android was computed. net8.0-browser net8.0-browser was computed. net8.0-ios net8.0-ios was computed. net8.0-maccatalyst net8.0-maccatalyst was computed. net8.0-macos net8.0-macos was computed. net8.0-tvos net8.0-tvos was computed. net8.0-windows net8.0-windows was computed. net9.0 net9.0 was computed. net9.0-android net9.0-android was computed. net9.0-browser net9.0-browser was computed. net9.0-ios net9.0-ios was computed. net9.0-maccatalyst net9.0-maccatalyst was computed. net9.0-macos net9.0-macos was computed. net9.0-tvos net9.0-tvos was computed. net9.0-windows net9.0-windows was computed. net10.0 net10.0 was computed. net10.0-android net10.0-android was computed. net10.0-browser net10.0-browser was computed. net10.0-ios net10.0-ios was computed. net10.0-maccatalyst net10.0-maccatalyst was computed. net10.0-macos net10.0-macos was computed. net10.0-tvos net10.0-tvos was computed. net10.0-windows net10.0-windows was computed. |
| .NET Core | netcoreapp2.0 netcoreapp2.0 was computed. netcoreapp2.1 netcoreapp2.1 was computed. netcoreapp2.2 netcoreapp2.2 was computed. netcoreapp3.0 netcoreapp3.0 was computed. netcoreapp3.1 netcoreapp3.1 was computed. |
| .NET Standard | netstandard2.0 netstandard2.0 is compatible. netstandard2.1 netstandard2.1 was computed. |
| .NET Framework | net461 net461 was computed. net462 net462 was computed. net463 net463 was computed. net47 net47 was computed. net471 net471 was computed. net472 net472 was computed. net48 net48 was computed. net481 net481 was computed. |
| MonoAndroid | monoandroid monoandroid was computed. |
| MonoMac | monomac monomac was computed. |
| MonoTouch | monotouch monotouch was computed. |
| Tizen | tizen40 tizen40 was computed. tizen60 tizen60 was computed. |
| Xamarin.iOS | xamarinios xamarinios was computed. |
| Xamarin.Mac | xamarinmac xamarinmac was computed. |
| Xamarin.TVOS | xamarintvos xamarintvos was computed. |
| Xamarin.WatchOS | xamarinwatchos xamarinwatchos was computed. |
Showing the top 5 NuGet packages that depend on Microsoft.DeepDev.TokenizerLib:
| Package | Downloads |
|---|---|
|
FoundationaLLM.Common
FoundationaLLM.Common is a .NET library that the FoundationaLLM.Client.Core and FoundationaLLM.Client.Management client libraries share as a common dependency. |
|
|
Microsoft.DotNet.Interactive.AIUtilities
Utilities for AI workload in .NET Interactive and Polyglot Notebooks |
|
|
LangChain.NET
LangChain.NET provides the ability to build applications with LLMs through composability |
|
|
Cnblogs.DashScope.Core
Provide pure api access to DashScope without extra references. Cnblogs.DashScope.Sdk should be used for general purpose. |
|
|
ContextFlow
Package Description |
Showing the top 5 popular GitHub repositories that depend on Microsoft.DeepDev.TokenizerLib:
| Repository | Stars |
|---|---|
|
microsoft/WhatTheHack
A collection of challenge based hack-a-thons including student guide, coach guide, lecture presentations, sample/instructional code and templates. Please visit the What The Hack website at: https://aka.ms/wth
|
|
|
axzxs2001/Asp.NetCoreExperiment
原来所有项目都移动到**OleVersion**目录下进行保留。新的案例装以.net 5.0为主,一部分对以前案例进行升级,一部分将以前的工作经验总结出来,以供大家参考!
|
|
|
dmitry-brazhenko/SharpToken
SharpToken is a C# library for tokenizing natural language text. It's based on the tiktoken Python library and designed to be fast and accurate.
|
|
|
Azure/Vector-Search-AI-Assistant
Microsoft Official Build Modern AI Apps reference solutions and content. Demonstrate how to build Copilot applications that incorporate Hero Azure Services including Azure OpenAI Service, Azure Container Apps (or AKS) and Azure Cosmos DB for NoSQL with Vector Search.
|
|
|
cnblogs/dashscope-sdk
An unofficial DashScope( SDK for .NET maintained by Cnblogs.
|