Add C# API for Moonshine models. (#1483)

* Also, return timestamps for non-streaming ASR.
k2-fsa · Oct 27, 2024 · 3622104 · 3622104
1 parent cdd8e1b
commit 3622104
Show file tree

Hide file tree

Showing 6 changed files with 143 additions and 8 deletions.
diff --git a/.github/scripts/test-dot-net.sh b/.github/scripts/test-dot-net.sh
@@ -9,6 +9,9 @@ rm -fv *.wav
 rm -rfv sherpa-onnx-pyannote-*
 
 cd ../offline-decode-files
+./run-moonshine.sh
+rm -rf sherpa-onnx-*
+
 ./run-sense-voice-ctc.sh
 rm -rf sherpa-onnx-*
 

diff --git a/dotnet-examples/offline-decode-files/Program.cs b/dotnet-examples/offline-decode-files/Program.cs
@@ -17,7 +17,7 @@ class Options
   {
 
     [Option("sample-rate", Required = false, Default = 16000, HelpText = "Sample rate of the data used to train the model")]
-    public int SampleRate { get; set; }  = 16000;
+    public int SampleRate { get; set; } = 16000;
 
     [Option("feat-dim", Required = false, Default = 80, HelpText = "Dimension of the features used to train the model")]
     public int FeatureDim { get; set; } = 80;
@@ -31,7 +31,7 @@ class Options
     [Option(Required = false, Default = "", HelpText = "Path to transducer decoder.onnx. Used only for transducer models")]
     public string Decoder { get; set; } = "";
 
-    [Option(Required = false,  Default = "",HelpText = "Path to transducer joiner.onnx. Used only for transducer models")]
+    [Option(Required = false, Default = "", HelpText = "Path to transducer joiner.onnx. Used only for transducer models")]
     public string Joiner { get; set; } = "";
 
     [Option("model-type", Required = false, Default = "", HelpText = "model type")]
@@ -44,10 +44,22 @@ class Options
     public string WhisperDecoder { get; set; } = "";
 
     [Option("whisper-language", Required = false, Default = "", HelpText = "Language of the input file. Can be empty")]
-    public string WhisperLanguage{ get; set; } = "";
+    public string WhisperLanguage { get; set; } = "";
 
     [Option("whisper-task", Required = false, Default = "transcribe", HelpText = "transcribe or translate")]
-    public string WhisperTask{ get; set; } = "transcribe";
+    public string WhisperTask { get; set; } = "transcribe";
+
+    [Option("moonshine-preprocessor", Required = false, Default = "", HelpText = "Path to preprocess.onnx. Used only for Moonshine models")]
+    public string MoonshinePreprocessor { get; set; } = "";
+
+    [Option("moonshine-encoder", Required = false, Default = "", HelpText = "Path to encode.onnx. Used only for Moonshine models")]
+    public string MoonshineEncoder { get; set; } = "";
+
+    [Option("moonshine-uncached-decoder", Required = false, Default = "", HelpText = "Path to uncached_decode.onnx. Used only for Moonshine models")]
+    public string MoonshineUncachedDecoder { get; set; } = "";
+
+    [Option("moonshine-cached-decoder", Required = false, Default = "", HelpText = "Path to cached_decode.onnx. Used only for Moonshine models")]
+    public string MoonshineCachedDecoder { get; set; } = "";
 
     [Option("tdnn-model", Required = false, Default = "", HelpText = "Path to tdnn yesno model")]
     public string TdnnModel { get; set; } = "";
@@ -90,7 +102,7 @@ class Options
     public float HotwordsScore { get; set; } = 1.5F;
 
     [Option("files", Required = true, HelpText = "Audio files for decoding")]
-    public IEnumerable<string> Files { get; set; } = new string[] {};
+    public IEnumerable<string> Files { get; set; } = new string[] { };
   }
 
   static void Main(string[] args)
@@ -236,6 +248,13 @@ private static void Run(Options options)
       config.ModelConfig.SenseVoice.Model = options.SenseVoiceModel;
       config.ModelConfig.SenseVoice.UseInverseTextNormalization = options.SenseVoiceUseItn;
     }
+    else if (!String.IsNullOrEmpty(options.MoonshinePreprocessor))
+    {
+      config.ModelConfig.Moonshine.Preprocessor = options.MoonshinePreprocessor;
+      config.ModelConfig.Moonshine.Encoder = options.MoonshineEncoder;
+      config.ModelConfig.Moonshine.UncachedDecoder = options.MoonshineUncachedDecoder;
+      config.ModelConfig.Moonshine.CachedDecoder = options.MoonshineCachedDecoder;
+    }
     else
     {
       Console.WriteLine("Please provide a model");
@@ -273,10 +292,21 @@ private static void Run(Options options)
     // display results
     for (int i = 0; i != files.Length; ++i)
     {
-      var text = streams[i].Result.Text;
+      var r = streams[i].Result;
       Console.WriteLine("--------------------");
       Console.WriteLine(files[i]);
-      Console.WriteLine(text);
+      Console.WriteLine("Text: {0}", r.Text);
+      Console.WriteLine("Tokens: [{0}]", string.Join(", ", r.Tokens));
+      if (r.Timestamps != null && r.Timestamps.Length > 0) {
+        Console.Write("Timestamps: [");
+        var sep = "";
+        for (int k = 0; k != r.Timestamps.Length; ++k)
+        {
+          Console.Write("{0}{1}", sep, r.Timestamps[k].ToString("0.00"));
+          sep = ", ";
+        }
+        Console.WriteLine("]");
+      }
     }
     Console.WriteLine("--------------------");
   }

diff --git a/dotnet-examples/offline-decode-files/run-moonshine.sh b/dotnet-examples/offline-decode-files/run-moonshine.sh
@@ -0,0 +1,18 @@
+#!/usr/bin/env bash
+
+set -ex
+
+if [ ! -f ./sherpa-onnx-moonshine-tiny-en-int8/tokens.txt ]; then
+  curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-moonshine-tiny-en-int8.tar.bz2
+  tar xvf sherpa-onnx-moonshine-tiny-en-int8.tar.bz2
+  rm sherpa-onnx-moonshine-tiny-en-int8.tar.bz2
+fi
+
+dotnet run \
+  --num-threads=2 \
+  --moonshine-preprocessor=./sherpa-onnx-moonshine-tiny-en-int8/preprocess.onnx \
+  --moonshine-encoder=./sherpa-onnx-moonshine-tiny-en-int8/encode.int8.onnx \
+  --moonshine-uncached-decoder=./sherpa-onnx-moonshine-tiny-en-int8/uncached_decode.int8.onnx \
+  --moonshine-cached-decoder=./sherpa-onnx-moonshine-tiny-en-int8/cached_decode.int8.onnx \
+  --tokens=./sherpa-onnx-moonshine-tiny-en-int8/tokens.txt \
+  --files ./sherpa-onnx-moonshine-tiny-en-int8/test_wavs/0.wav
diff --git a/scripts/dotnet/OfflineModelConfig.cs b/scripts/dotnet/OfflineModelConfig.cs
@@ -24,6 +24,7 @@ public OfflineModelConfig()
             BpeVocab = "";
             TeleSpeechCtc = "";
             SenseVoice = new OfflineSenseVoiceModelConfig();
+            Moonshine = new OfflineMoonshineModelConfig();
         }
         public OfflineTransducerModelConfig Transducer;
         public OfflineParaformerModelConfig Paraformer;
@@ -54,5 +55,6 @@ public OfflineModelConfig()
         public string TeleSpeechCtc;
 
         public OfflineSenseVoiceModelConfig SenseVoice;
+        public OfflineMoonshineModelConfig Moonshine;
     }
 }
diff --git a/scripts/dotnet/OfflineMoonshineModelConfig.cs b/scripts/dotnet/OfflineMoonshineModelConfig.cs
@@ -0,0 +1,29 @@
+/// Copyright (c)  2024  Xiaomi Corporation (authors: Fangjun Kuang)
+
+using System.Runtime.InteropServices;
+
+namespace SherpaOnnx
+{
+    [StructLayout(LayoutKind.Sequential)]
+    public struct OfflineMoonshineModelConfig
+    {
+        public OfflineMoonshineModelConfig()
+        {
+            Preprocessor = "";
+            Encoder = "";
+            UncachedDecoder = "";
+            CachedDecoder = "";
+        }
+        [MarshalAs(UnmanagedType.LPStr)]
+        public string Preprocessor;
+
+        [MarshalAs(UnmanagedType.LPStr)]
+        public string Encoder;
+
+        [MarshalAs(UnmanagedType.LPStr)]
+        public string UncachedDecoder;
+
+        [MarshalAs(UnmanagedType.LPStr)]
+        public string CachedDecoder;
+    }
+}
diff --git a/scripts/dotnet/OfflineRecognizerResult.cs b/scripts/dotnet/OfflineRecognizerResult.cs
@@ -31,17 +31,70 @@ public OfflineRecognizerResult(IntPtr handle)
             byte[] stringBuffer = new byte[length];
             Marshal.Copy(impl.Text, stringBuffer, 0, length);
             _text = Encoding.UTF8.GetString(stringBuffer);
+
+            _tokens = new String[impl.Count];
+
+            unsafe
+            {
+                byte* buf = (byte*)impl.Tokens;
+                for (int i = 0; i < impl.Count; i++)
+                {
+                    length = 0;
+                    byte* start = buf;
+                    while (*buf != 0)
+                    {
+                        ++buf;
+                        length += 1;
+                    }
+                    ++buf;
+
+                    stringBuffer = new byte[length];
+                    fixed (byte* pTarget = stringBuffer)
+                    {
+                        for (int k = 0; k < length; k++)
+                        {
+                            pTarget[k] = start[k];
+                        }
+                    }
+
+                    _tokens[i] = Encoding.UTF8.GetString(stringBuffer);
+                }
+            }
+
+            unsafe
+            {
+              if (impl.Timestamps != IntPtr.Zero)
+              {
+                float *t = (float*)impl.Timestamps;
+                _timestamps = new float[impl.Count];
+                fixed (float* f = _timestamps)
+                {
+                  for (int k = 0; k < impl.Count; k++)
+                  {
+                    f[k] = t[k];
+                  }
+                }
+              }
+            }
+
         }
 
         [StructLayout(LayoutKind.Sequential)]
         struct Impl
         {
             public IntPtr Text;
+            public IntPtr Timestamps;
+            public int Count;
+            public IntPtr Tokens;
         }
 
         private String _text;
         public String Text => _text;
-    }
 
+        private String[] _tokens;
+        public String[] Tokens => _tokens;
 
+        private float[] _timestamps;
+        public float[] Timestamps => _timestamps;
+    }
 }