xuning / sherpaonnx

转到一个项目
Toggle navigation Toggle navigation pinning
Fangjun Kuang
Committed by GitHub
fa2af5dc 1 parent 035a82df

Add TTS demo for C# API (#557)

内嵌 并排对比
正在显示 13 个修改的文件 包含 500 行增加 和 9 行删除
@@ -55,3 +55,19 @@ jobs: @@ -55,3 +55,19 @@ jobs:
55 ./run-zipformer.sh 55 ./run-zipformer.sh
56 ./run-whisper.sh 56 ./run-whisper.sh
57 ./run-tdnn-yesno.sh 57 ./run-tdnn-yesno.sh
  58 +
  59 + cd ../offline-tts
  60 + ./run-aishell3.sh
  61 + ./run-piper.sh
  62 + ls -lh
  63 +
  64 + cd ../..
  65 +
  66 + mkdir tts
  67 +
  68 + cp dotnet-examples/offline-tts/*.wav ./tts
  69 +
  70 + - uses: actions/upload-artifact@v3
  71 + with:
  72 + name: dot-net-tts-generated-test-files-${{ matrix.os }}
  73 + path: tts
@@ -131,6 +131,7 @@ jobs: @@ -131,6 +131,7 @@ jobs:
131 - name: Copy files 131 - name: Copy files
132 shell: bash 132 shell: bash
133 run: | 133 run: |
  134 + cp -v scripts/dotnet/examples/offline-tts.csproj dotnet-examples/offline-tts/
134 cp -v scripts/dotnet/examples/offline-decode-files.csproj dotnet-examples/offline-decode-files/ 135 cp -v scripts/dotnet/examples/offline-decode-files.csproj dotnet-examples/offline-decode-files/
135 cp -v scripts/dotnet/examples/online-decode-files.csproj dotnet-examples/online-decode-files/ 136 cp -v scripts/dotnet/examples/online-decode-files.csproj dotnet-examples/online-decode-files/
136 cp -v scripts/dotnet/examples/speech-recognition-from-microphone.csproj dotnet-examples/speech-recognition-from-microphone/ 137 cp -v scripts/dotnet/examples/speech-recognition-from-microphone.csproj dotnet-examples/speech-recognition-from-microphone/
@@ -153,3 +154,19 @@ jobs: @@ -153,3 +154,19 @@ jobs:
153 ./run-zipformer.sh 154 ./run-zipformer.sh
154 ./run-whisper.sh 155 ./run-whisper.sh
155 ./run-tdnn-yesno.sh 156 ./run-tdnn-yesno.sh
  157 +
  158 + cd ../offline-tts
  159 + ./run-aishell3.sh
  160 + ./run-piper.sh
  161 + ls -lh
  162 +
  163 + cd ../..
  164 +
  165 + mkdir tts
  166 +
  167 + cp dotnet-examples/offline-tts/*.wav ./tts
  168 +
  169 + - uses: actions/upload-artifact@v3
  170 + with:
  171 + name: dot-net-tts-generated-test-files-${{ matrix.os }}
  172 + path: tts
1 cmake_minimum_required(VERSION 3.13 FATAL_ERROR) 1 cmake_minimum_required(VERSION 3.13 FATAL_ERROR)
2 project(sherpa-onnx) 2 project(sherpa-onnx)
3 3
4 -set(SHERPA_ONNX_VERSION "1.9.9") 4 +set(SHERPA_ONNX_VERSION "1.9.10")
5 5
6 # Disable warning about 6 # Disable warning about
7 # 7 #
  1 +# How to create a new project in this folder
  2 +
  3 +```bash
  4 +mkdir offline-tts
  5 +cd offline-tts
  6 +dotnet new console
  7 +cd ..
  8 +dotnet sln ./sherpa-onnx.sln add ./offline-tts
  9 +```
  1 +// Copyright (c) 2024 Xiaomi Corporation
  2 +//
  3 +// This file shows how to use a non-streaming TTS model for text-to-speech
  4 +// Please refer to
  5 +// https://k2-fsa.github.io/sherpa/onnx/pretrained_models/index.html
  6 +// and
  7 +// https://github.com/k2-fsa/sherpa-onnx/releases/tag/tts-models
  8 +// to download pre-trained models
  9 +using CommandLine.Text;
  10 +using CommandLine;
  11 +using SherpaOnnx;
  12 +using System.Collections.Generic;
  13 +using System;
  14 +
  15 +class OfflineTtsDemo
  16 +{
  17 + class Options
  18 + {
  19 +
  20 + [Option("tts-rule-fsts", Required = false, Default = "", HelpText = "path to rule.fst")]
  21 + public string RuleFsts { get; set; }
  22 +
  23 + [Option("vits-data-dir", Required = false, Default = "", HelpText = "Path to the directory containing dict for espeak-ng.")]
  24 + public string DataDir { get; set; }
  25 +
  26 + [Option("vits-length-scale", Required = false, Default = 1, HelpText = "speech speed. Larger->Slower; Smaller->faster")]
  27 + public float LengthScale { get; set; }
  28 +
  29 + [Option("vits-noise-scale", Required = false, Default = 0.667f, HelpText = "noise_scale for VITS models")]
  30 + public float NoiseScale { get; set; }
  31 +
  32 + [Option("vits-noise-scale-w", Required = false, Default = 0.8f, HelpText = "noise_scale_w for VITS models")]
  33 + public float NoiseScaleW { get; set; }
  34 +
  35 + [Option("vits-lexicon", Required = false, Default = "", HelpText = "Path to lexicon.txt")]
  36 + public string Lexicon { get; set; }
  37 +
  38 + [Option("vits-tokens", Required = false, Default = "", HelpText = "Path to tokens.txt")]
  39 + public string Tokens { get; set; }
  40 +
  41 + [Option("tts-max-num-sentences", Required = false, Default = 1, HelpText = "Maximum number of sentences that we process at a time.")]
  42 + public int MaxNumSentences { get; set; }
  43 +
  44 + [Option(Required = false, Default = 0, HelpText = "1 to show debug messages.")]
  45 + public int Debug { get; set; }
  46 +
  47 + [Option("vits-model", Required = true, HelpText = "Path to VITS model")]
  48 + public string Model { get; set; }
  49 +
  50 + [Option("sid", Required = false, Default = 0, HelpText = "Speaker ID")]
  51 + public int SpeakerId { get; set; }
  52 +
  53 + [Option("text", Required = true, HelpText = "Text to synthesize")]
  54 + public string Text { get; set; }
  55 +
  56 + [Option("output-filename", Required = true, Default = "./generated.wav", HelpText = "Path to save the generated audio")]
  57 + public string OutputFilename { get; set; }
  58 + }
  59 +
  60 + static void Main(string[] args)
  61 + {
  62 + var parser = new CommandLine.Parser(with => with.HelpWriter = null);
  63 + var parserResult = parser.ParseArguments<Options>(args);
  64 +
  65 + parserResult
  66 + .WithParsed<Options>(options => Run(options))
  67 + .WithNotParsed(errs => DisplayHelp(parserResult, errs));
  68 + }
  69 +
  70 + private static void DisplayHelp<T>(ParserResult<T> result, IEnumerable<Error> errs)
  71 + {
  72 + string usage = @"
  73 +# vits-aishell3
  74 +
  75 +wget -qq https://github.com/k2-fsa/sherpa-onnx/releases/download/tts-models/vits-zh-aishell3.tar.bz2
  76 +tar xf vits-zh-aishell3.tar.bz2
  77 +
  78 +dotnet run \
  79 + --vits-model=./vits-zh-aishell3/vits-aishell3.onnx \
  80 + --vits-tokens=./vits-zh-aishell3/tokens.txt \
  81 + --vits-lexicon=./vits-zh-aishell3/lexicon.txt \
  82 + --tts-rule-fsts=./vits-zh-aishell3/rule.fst \
  83 + --sid=66 \
  84 + --debug=1 \
  85 + --output-filename=./aishell3-66.wav \
  86 + --text=这是一个语音合成测试
  87 +
  88 +# Piper models
  89 +
  90 +wget -qq https://github.com/k2-fsa/sherpa-onnx/releases/download/tts-models/vits-piper-en_US-amy-low.tar.bz2
  91 +tar xf vits-piper-en_US-amy-low.tar.bz2
  92 +
  93 +dotnet run \
  94 + --vits-model=./vits-piper-en_US-amy-low/en_US-amy-low.onnx \
  95 + --vits-tokens=./vits-piper-en_US-amy-low/tokens.txt \
  96 + --vits-data-dir=./vits-piper-en_US-amy-low/espeak-ng-data \
  97 + --debug=1 \
  98 + --output-filename=./amy.wav \
  99 + --text='This is a text to speech application in dotnet with Next Generation Kaldi'
  100 +
  101 +Please refer to
  102 +https://k2-fsa.github.io/sherpa/onnx/tts/pretrained_models/index.html
  103 +to download more models.
  104 +";
  105 +
  106 + var helpText = HelpText.AutoBuild(result, h =>
  107 + {
  108 + h.AdditionalNewLineAfterOption = false;
  109 + h.Heading = usage;
  110 + h.Copyright = "Copyright (c) 2024 Xiaomi Corporation";
  111 + return HelpText.DefaultParsingErrorsHandler(result, h);
  112 + }, e => e);
  113 + Console.WriteLine(helpText);
  114 + }
  115 +
  116 + private static void Run(Options options)
  117 + {
  118 + OfflineTtsConfig config = new OfflineTtsConfig();
  119 + config.Model.Vits.Model = options.Model;
  120 + config.Model.Vits.Lexicon = options.Lexicon;
  121 + config.Model.Vits.Tokens = options.Tokens;
  122 + config.Model.Vits.DataDir = options.DataDir;
  123 + config.Model.Vits.NoiseScale = options.NoiseScale;
  124 + config.Model.Vits.NoiseScaleW = options.NoiseScaleW;
  125 + config.Model.Vits.LengthScale = options.LengthScale;
  126 + config.Model.NumThreads = 1;
  127 + config.Model.Debug = options.Debug;
  128 + config.Model.Provider = "cpu";
  129 + config.RuleFsts = options.RuleFsts;
  130 + config.MaxNumSentences = options.MaxNumSentences;
  131 +
  132 + OfflineTts tts = new OfflineTts(config);
  133 + float speed = 1.0f / options.LengthScale;
  134 + int sid = options.SpeakerId;
  135 + OfflineTtsGeneratedAudio audio = tts.Generate(options.Text, speed, sid);
  136 + bool ok = audio.SaveToWaveFile(options.OutputFilename);
  137 +
  138 + if (ok)
  139 + {
  140 + Console.WriteLine($"Wrote to {options.OutputFilename} succeeded!");
  141 + }
  142 + else
  143 + {
  144 + Console.WriteLine($"Failed to write {options.OutputFilename}");
  145 + }
  146 + }
  147 +}
  1 +<Project Sdk="Microsoft.NET.Sdk">
  2 +
  3 + <PropertyGroup>
  4 + <OutputType>Exe</OutputType>
  5 + <TargetFramework>net6.0</TargetFramework>
  6 + <RootNamespace>offline_tts</RootNamespace>
  7 + <ImplicitUsings>enable</ImplicitUsings>
  8 + <Nullable>enable</Nullable>
  9 + </PropertyGroup>
  10 +
  11 + <ItemGroup>
  12 + <PackageReference Include="CommandLineParser" Version="2.9.1" />
  13 + <PackageReference Include="org.k2fsa.sherpa.onnx" Version="*" />
  14 + </ItemGroup>
  15 +
  16 +</Project>
  1 +#!/usr/bin/env bash
  2 +
  3 +if [ ! -f ./vits-zh-aishell3/vits-aishell3.onnx ]; then
  4 + wget -qq https://github.com/k2-fsa/sherpa-onnx/releases/download/tts-models/vits-zh-aishell3.tar.bz2
  5 + tar xf vits-zh-aishell3.tar.bz2
  6 + rm vits-zh-aishell3.tar.bz2
  7 +fi
  8 +
  9 +dotnet run \
  10 + --vits-model=./vits-zh-aishell3/vits-aishell3.onnx \
  11 + --vits-tokens=./vits-zh-aishell3/tokens.txt \
  12 + --vits-lexicon=./vits-zh-aishell3/lexicon.txt \
  13 + --tts-rule-fsts=./vits-zh-aishell3/rule.fst \
  14 + --sid=66 \
  15 + --debug=1 \
  16 + --output-filename=./aishell3-66.wav \
  17 + --text="这是一个语音合成测试, 写于公元 2024 年 1 月 28 号, 23点27分，星期天。"
  1 +#!/usr/bin/env bash
  2 +
  3 +if [ ! -f ./vits-piper-en_US-amy-low/en_US-amy-low.onnx ]; then
  4 + wget -qq https://github.com/k2-fsa/sherpa-onnx/releases/download/tts-models/vits-piper-en_US-amy-low.tar.bz2
  5 + tar xf vits-piper-en_US-amy-low.tar.bz2
  6 + rm vits-piper-en_US-amy-low.tar.bz2
  7 +fi
  8 +
  9 +dotnet run \
  10 + --vits-model=./vits-piper-en_US-amy-low/en_US-amy-low.onnx \
  11 + --vits-tokens=./vits-piper-en_US-amy-low/tokens.txt \
  12 + --vits-data-dir=./vits-piper-en_US-amy-low/espeak-ng-data \
  13 + --debug=1 \
  14 + --output-filename=./amy.wav \
  15 + --text="This is a text to speech application in dotnet with Next Generation Kaldi"
  16 +
@@ -9,6 +9,8 @@ Project("{FAE04EC0-301F-11D3-BF4B-00C04F79EFBC}") = "offline-decode-files", "off @@ -9,6 +9,8 @@ Project("{FAE04EC0-301F-11D3-BF4B-00C04F79EFBC}") = "offline-decode-files", "off
9 EndProject 9 EndProject
10 Project("{FAE04EC0-301F-11D3-BF4B-00C04F79EFBC}") = "speech-recognition-from-microphone", "speech-recognition-from-microphone\speech-recognition-from-microphone.csproj", "{FE4EA1FF-062A-46B3-B78D-C828FED7B82E}" 10 Project("{FAE04EC0-301F-11D3-BF4B-00C04F79EFBC}") = "speech-recognition-from-microphone", "speech-recognition-from-microphone\speech-recognition-from-microphone.csproj", "{FE4EA1FF-062A-46B3-B78D-C828FED7B82E}"
11 EndProject 11 EndProject
  12 +Project("{FAE04EC0-301F-11D3-BF4B-00C04F79EFBC}") = "offline-tts", "offline-tts\offline-tts.csproj", "{72196886-7143-4043-96E2-BCACEC6C79EB}"
  13 +EndProject
12 Global 14 Global
13 GlobalSection(SolutionConfigurationPlatforms) = preSolution 15 GlobalSection(SolutionConfigurationPlatforms) = preSolution
14 Debug|Any CPU = Debug|Any CPU 16 Debug|Any CPU = Debug|Any CPU
@@ -30,5 +32,9 @@ Global @@ -30,5 +32,9 @@ Global
30 {FE4EA1FF-062A-46B3-B78D-C828FED7B82E}.Debug|Any CPU.Build.0 = Debug|Any CPU 32 {FE4EA1FF-062A-46B3-B78D-C828FED7B82E}.Debug|Any CPU.Build.0 = Debug|Any CPU
31 {FE4EA1FF-062A-46B3-B78D-C828FED7B82E}.Release|Any CPU.ActiveCfg = Release|Any CPU 33 {FE4EA1FF-062A-46B3-B78D-C828FED7B82E}.Release|Any CPU.ActiveCfg = Release|Any CPU
32 {FE4EA1FF-062A-46B3-B78D-C828FED7B82E}.Release|Any CPU.Build.0 = Release|Any CPU 34 {FE4EA1FF-062A-46B3-B78D-C828FED7B82E}.Release|Any CPU.Build.0 = Release|Any CPU
  35 + {72196886-7143-4043-96E2-BCACEC6C79EB}.Debug|Any CPU.ActiveCfg = Debug|Any CPU
  36 + {72196886-7143-4043-96E2-BCACEC6C79EB}.Debug|Any CPU.Build.0 = Debug|Any CPU
  37 + {72196886-7143-4043-96E2-BCACEC6C79EB}.Release|Any CPU.ActiveCfg = Release|Any CPU
  38 + {72196886-7143-4043-96E2-BCACEC6C79EB}.Release|Any CPU.Build.0 = Release|Any CPU
33 EndGlobalSection 39 EndGlobalSection
34 EndGlobal 40 EndGlobal
  1 +<Project Sdk="Microsoft.NET.Sdk">
  2 +
  3 + <PropertyGroup>
  4 + <OutputType>Exe</OutputType>
  5 + <TargetFramework>net6.0</TargetFramework>
  6 + <RootNamespace>offline_tts</RootNamespace>
  7 + <ImplicitUsings>enable</ImplicitUsings>
  8 + <Nullable>enable</Nullable>
  9 + </PropertyGroup>
  10 +
  11 + <PropertyGroup>
  12 + <RestoreSources>/tmp/packages;$(RestoreSources);https://api.nuget.org/v3/index.json</RestoreSources>
  13 + </PropertyGroup>
  14 +
  15 + <ItemGroup>
  16 + <PackageReference Include="CommandLineParser" Version="2.9.1" />
  17 + <PackageReference Include="org.k2fsa.sherpa.onnx" Version="*" />
  18 + </ItemGroup>
  19 +
  20 +</Project>
@@ -11,6 +11,214 @@ namespace SherpaOnnx @@ -11,6 +11,214 @@ namespace SherpaOnnx
11 { 11 {
12 12
13 [StructLayout(LayoutKind.Sequential)] 13 [StructLayout(LayoutKind.Sequential)]
  14 + public struct OfflineTtsVitsModelConfig
  15 + {
  16 + public OfflineTtsVitsModelConfig()
  17 + {
  18 + Model = "";
  19 + Lexicon = "";
  20 + Tokens = "";
  21 + DataDir = "";
  22 +
  23 + NoiseScale = 0.667F;
  24 + NoiseScaleW = 0.8F;
  25 + LengthScale = 1.0F;
  26 + }
  27 + [MarshalAs(UnmanagedType.LPStr)]
  28 + public string Model;
  29 +
  30 + [MarshalAs(UnmanagedType.LPStr)]
  31 + public string Lexicon;
  32 +
  33 + [MarshalAs(UnmanagedType.LPStr)]
  34 + public string Tokens;
  35 +
  36 + [MarshalAs(UnmanagedType.LPStr)]
  37 + public string DataDir;
  38 +
  39 + public float NoiseScale;
  40 + public float NoiseScaleW;
  41 + public float LengthScale;
  42 + }
  43 +
  44 + [StructLayout(LayoutKind.Sequential)]
  45 + public struct OfflineTtsModelConfig
  46 + {
  47 + public OfflineTtsModelConfig()
  48 + {
  49 + Vits = new OfflineTtsVitsModelConfig();
  50 + NumThreads = 1;
  51 + Debug = 0;
  52 + Provider = "cpu";
  53 + }
  54 +
  55 + public OfflineTtsVitsModelConfig Vits;
  56 + public int NumThreads;
  57 + public int Debug;
  58 + [MarshalAs(UnmanagedType.LPStr)]
  59 + public string Provider;
  60 + }
  61 +
  62 + public struct OfflineTtsConfig
  63 + {
  64 + public OfflineTtsConfig()
  65 + {
  66 + Model = new OfflineTtsModelConfig();
  67 + RuleFsts = "";
  68 + MaxNumSentences = 1;
  69 + }
  70 + public OfflineTtsModelConfig Model;
  71 +
  72 + [MarshalAs(UnmanagedType.LPStr)]
  73 + public string RuleFsts;
  74 +
  75 + public int MaxNumSentences;
  76 + }
  77 +
  78 + public class OfflineTtsGeneratedAudio
  79 + {
  80 + public OfflineTtsGeneratedAudio(IntPtr p)
  81 + {
  82 + _handle = new HandleRef(this, p);
  83 + }
  84 +
  85 + public bool SaveToWaveFile(String filename)
  86 + {
  87 + Impl impl = (Impl)Marshal.PtrToStructure(Handle, typeof(Impl));
  88 + int status = SherpaOnnxWriteWave(impl.Samples, impl.NumSamples, impl.SampleRate, filename);
  89 + return status == 1;
  90 + }
  91 +
  92 + ~OfflineTtsGeneratedAudio()
  93 + {
  94 + Cleanup();
  95 + }
  96 +
  97 + public void Dispose()
  98 + {
  99 + Cleanup();
  100 + // Prevent the object from being placed on the
  101 + // finalization queue
  102 + System.GC.SuppressFinalize(this);
  103 + }
  104 +
  105 + private void Cleanup()
  106 + {
  107 + SherpaOnnxDestroyOfflineTtsGeneratedAudio(Handle);
  108 +
  109 + // Don't permit the handle to be used again.
  110 + _handle = new HandleRef(this, IntPtr.Zero);
  111 + }
  112 +
  113 + [StructLayout(LayoutKind.Sequential)]
  114 + struct Impl
  115 + {
  116 + public IntPtr Samples;
  117 + public int NumSamples;
  118 + public int SampleRate;
  119 + }
  120 +
  121 + private HandleRef _handle;
  122 + public IntPtr Handle => _handle.Handle;
  123 +
  124 + public int NumSamples
  125 + {
  126 + get
  127 + {
  128 + Impl impl = (Impl)Marshal.PtrToStructure(Handle, typeof(Impl));
  129 + return impl.NumSamples;
  130 + }
  131 + }
  132 +
  133 + public int SampleRate
  134 + {
  135 + get
  136 + {
  137 + Impl impl = (Impl)Marshal.PtrToStructure(Handle, typeof(Impl));
  138 + return impl.SampleRate;
  139 + }
  140 + }
  141 +
  142 + public float[] Samples
  143 + {
  144 + get
  145 + {
  146 + Impl impl = (Impl)Marshal.PtrToStructure(Handle, typeof(Impl));
  147 +
  148 + float[] samples = new float[impl.NumSamples];
  149 + Marshal.Copy(impl.Samples, samples, 0, impl.NumSamples);
  150 + return samples;
  151 + }
  152 + }
  153 +
  154 + [DllImport(Dll.Filename)]
  155 + private static extern void SherpaOnnxDestroyOfflineTtsGeneratedAudio(IntPtr handle);
  156 +
  157 + [DllImport(Dll.Filename)]
  158 + private static extern int SherpaOnnxWriteWave(IntPtr samples, int n, int sample_rate, [MarshalAs(UnmanagedType.LPStr)] string filename);
  159 + }
  160 +
  161 + public class OfflineTts : IDisposable
  162 + {
  163 + public OfflineTts(OfflineTtsConfig config)
  164 + {
  165 + IntPtr h = SherpaOnnxCreateOfflineTts(ref config);
  166 + _handle = new HandleRef(this, h);
  167 + }
  168 +
  169 + public OfflineTtsGeneratedAudio Generate(String text, float speed, int speakerId)
  170 + {
  171 + IntPtr p = SherpaOnnxOfflineTtsGenerate(_handle.Handle, text, speakerId, speed);
  172 + return new OfflineTtsGeneratedAudio(p);
  173 + }
  174 +
  175 + public void Dispose()
  176 + {
  177 + Cleanup();
  178 + // Prevent the object from being placed on the
  179 + // finalization queue
  180 + System.GC.SuppressFinalize(this);
  181 + }
  182 +
  183 + ~OfflineTts()
  184 + {
  185 + Cleanup();
  186 + }
  187 +
  188 + private void Cleanup()
  189 + {
  190 + SherpaOnnxDestroyOfflineTts(_handle.Handle);
  191 +
  192 + // Don't permit the handle to be used again.
  193 + _handle = new HandleRef(this, IntPtr.Zero);
  194 + }
  195 +
  196 + private HandleRef _handle;
  197 +
  198 + public int SampleRate
  199 + {
  200 + get
  201 + {
  202 + return SherpaOnnxOfflineTtsSampleRate(_handle.Handle);
  203 + }
  204 + }
  205 +
  206 + [DllImport(Dll.Filename)]
  207 + private static extern IntPtr SherpaOnnxCreateOfflineTts(ref OfflineTtsConfig config);
  208 +
  209 + [DllImport(Dll.Filename)]
  210 + private static extern void SherpaOnnxDestroyOfflineTts(IntPtr handle);
  211 +
  212 + [DllImport(Dll.Filename)]
  213 + private static extern int SherpaOnnxOfflineTtsSampleRate(IntPtr handle);
  214 +
  215 + [DllImport(Dll.Filename)]
  216 + private static extern IntPtr SherpaOnnxOfflineTtsGenerate(IntPtr handle, [MarshalAs(UnmanagedType.LPStr)] string text, int sid, float speed);
  217 + }
  218 +
  219 +
  220 +
  221 + [StructLayout(LayoutKind.Sequential)]
14 public struct OfflineTransducerModelConfig 222 public struct OfflineTransducerModelConfig
15 { 223 {
16 public OfflineTransducerModelConfig() 224 public OfflineTransducerModelConfig()
@@ -15,12 +15,23 @@ pushd /tmp @@ -15,12 +15,23 @@ pushd /tmp
15 15
16 mkdir -p linux macos windows 16 mkdir -p linux macos windows
17 17
  18 +# You can pre-download the required wheels to /tmp
  19 +src_dir=/tmp
  20 +
  21 +linux_wheel=$src_dir/sherpa_onnx-${SHERPA_ONNX_VERSION}-cp38-cp38-manylinux_2_17_x86_64.manylinux2014_x86_64.whl
  22 +macos_wheel=$src_dir/sherpa_onnx-${SHERPA_ONNX_VERSION}-cp38-cp38-macosx_10_14_x86_64.whl
  23 +windows_wheel=$src_dir/sherpa_onnx-${SHERPA_ONNX_VERSION}-cp38-cp38-win_amd64.whl
  24 +
18 if [ ! -f /tmp/linux/libsherpa-onnx-core.so ]; then 25 if [ ! -f /tmp/linux/libsherpa-onnx-core.so ]; then
19 echo "---linux x86_64---" 26 echo "---linux x86_64---"
20 cd linux 27 cd linux
21 - mkdir wheel 28 + mkdir -p wheel
22 cd wheel 29 cd wheel
23 - curl -OL https://huggingface.co/csukuangfj/sherpa-onnx-wheels/resolve/main/sherpa_onnx-${SHERPA_ONNX_VERSION}-cp38-cp38-manylinux_2_17_x86_64.manylinux2014_x86_64.whl 30 + if [ -f $linux_wheel ]; then
  31 + cp -v $linux_wheel .
  32 + else
  33 + curl -OL https://huggingface.co/csukuangfj/sherpa-onnx-wheels/resolve/main/sherpa_onnx-${SHERPA_ONNX_VERSION}-cp38-cp38-manylinux_2_17_x86_64.manylinux2014_x86_64.whl
  34 + fi
24 unzip sherpa_onnx-${SHERPA_ONNX_VERSION}-cp38-cp38-manylinux_2_17_x86_64.manylinux2014_x86_64.whl 35 unzip sherpa_onnx-${SHERPA_ONNX_VERSION}-cp38-cp38-manylinux_2_17_x86_64.manylinux2014_x86_64.whl
25 cp -v sherpa_onnx/lib/*.so* ../ 36 cp -v sherpa_onnx/lib/*.so* ../
26 cd .. 37 cd ..
@@ -36,9 +47,13 @@ fi @@ -36,9 +47,13 @@ fi
36 if [ ! -f /tmp/macos/libsherpa-onnx-core.dylib ]; then 47 if [ ! -f /tmp/macos/libsherpa-onnx-core.dylib ]; then
37 echo "---macOS x86_64---" 48 echo "---macOS x86_64---"
38 cd macos 49 cd macos
39 - mkdir wheel 50 + mkdir -p wheel
40 cd wheel 51 cd wheel
41 - curl -OL https://huggingface.co/csukuangfj/sherpa-onnx-wheels/resolve/main/sherpa_onnx-${SHERPA_ONNX_VERSION}-cp38-cp38-macosx_10_14_x86_64.whl 52 + if [ -f $macos_wheel ]; then
  53 + cp -v $macos_wheel .
  54 + else
  55 + curl -OL https://huggingface.co/csukuangfj/sherpa-onnx-wheels/resolve/main/sherpa_onnx-${SHERPA_ONNX_VERSION}-cp38-cp38-macosx_10_14_x86_64.whl
  56 + fi
42 unzip sherpa_onnx-${SHERPA_ONNX_VERSION}-cp38-cp38-macosx_10_14_x86_64.whl 57 unzip sherpa_onnx-${SHERPA_ONNX_VERSION}-cp38-cp38-macosx_10_14_x86_64.whl
43 cp -v sherpa_onnx/lib/*.dylib ../ 58 cp -v sherpa_onnx/lib/*.dylib ../
44 59
@@ -57,9 +72,13 @@ fi @@ -57,9 +72,13 @@ fi
57 if [ ! -f /tmp/windows/libsherpa-onnx-core.dll ]; then 72 if [ ! -f /tmp/windows/libsherpa-onnx-core.dll ]; then
58 echo "---windows x64---" 73 echo "---windows x64---"
59 cd windows 74 cd windows
60 - mkdir wheel 75 + mkdir -p wheel
61 cd wheel 76 cd wheel
62 - curl -OL https://huggingface.co/csukuangfj/sherpa-onnx-wheels/resolve/main/sherpa_onnx-${SHERPA_ONNX_VERSION}-cp38-cp38-win_amd64.whl 77 + if [ -f $windows_wheel ]; then
  78 + cp -v $windows_wheel .
  79 + else
  80 + curl -OL https://huggingface.co/csukuangfj/sherpa-onnx-wheels/resolve/main/sherpa_onnx-${SHERPA_ONNX_VERSION}-cp38-cp38-win_amd64.whl
  81 + fi
63 unzip sherpa_onnx-${SHERPA_ONNX_VERSION}-cp38-cp38-win_amd64.whl 82 unzip sherpa_onnx-${SHERPA_ONNX_VERSION}-cp38-cp38-win_amd64.whl
64 cp -v sherpa_onnx-${SHERPA_ONNX_VERSION}.data/data/bin/*.dll ../ 83 cp -v sherpa_onnx-${SHERPA_ONNX_VERSION}.data/data/bin/*.dll ../
65 cp -v sherpa_onnx-${SHERPA_ONNX_VERSION}.data/data/bin/*.lib ../ 84 cp -v sherpa_onnx-${SHERPA_ONNX_VERSION}.data/data/bin/*.lib ../
@@ -101,5 +120,5 @@ popd @@ -101,5 +120,5 @@ popd
101 120
102 ls -lh packages 121 ls -lh packages
103 122
104 -mkdir /tmp/packages 123 +mkdir -p /tmp/packages
105 cp -v packages/*.nupkg /tmp/packages 124 cp -v packages/*.nupkg /tmp/packages
@@ -18,7 +18,7 @@ @@ -18,7 +18,7 @@
18 18
19 namespace sherpa_onnx { 19 namespace sherpa_onnx {
20 20
21 -class TransducerKeywordResult; 21 +struct TransducerKeywordResult;
22 class OnlineStream { 22 class OnlineStream {
23 public: 23 public:
24 explicit OnlineStream(const FeatureExtractorConfig &config = {}, 24 explicit OnlineStream(const FeatureExtractorConfig &config = {},