Add JavaScript API (WebAssembly) for FireRedAsr model. (#1874)

Author: csukuangfj

new-release.sh

@@ -14,6 +14,7 @@ find dart-api-examples -name *.yaml -type f -exec sed -i.bak 's/1\.10\.43/1\.10\
 find flutter-examples -name *.yaml -type f -exec sed -i.bak 's/1\.10\.43/1\.10\.44/g' {} \;
 find flutter -name *.podspec -type f -exec sed -i.bak 's/1\.10\.43/1\.10\.44/g' {} \;
 find nodejs-addon-examples -name package.json -type f -exec sed -i.bak 's/1\.10\.43/1\.10\.44/g' {} \;
+find nodejs-examples -name package.json -type f -exec sed -i.bak 's/1\.10\.43/1\.10\.44/g' {} \;
 
 find harmony-os -name "README.md" -type f -exec sed -i.bak 's/1\.10\.43/1\.10\.44/g' {} \;
 find harmony-os -name oh-package.json5 -type f -exec sed -i.bak 's/1\.10\.43/1\.10\.44/g' {} \;

nodejs-examples/README.md

@@ -216,6 +216,21 @@ tar xvf sherpa-onnx-whisper-tiny.en.tar.bz2
 node ./test-offline-whisper.js
 ```
 
+## ./test-offline-fire-red-asr.js
+
+[./test-offline-fire-red-asr.js](./test-offline-fire-red-asr.js) demonstrates
+how to decode a file with a FireRedAsr AED model.
+
+You can use the following command to run it:
+
+```bash
+wget https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-fire-red-asr-large-zh_en-2025-02-16.tar.bz2
+tar xvf sherpa-onnx-fire-red-asr-large-zh_en-2025-02-16.tar.bz2
+rm sherpa-onnx-fire-red-asr-large-zh_en-2025-02-16.tar.bz2
+
+node ./test-offline-fire-red-asr.js
+```
+
 ## ./test-offline-moonshine.js
 
 [./test-offline-moonshine.js](./test-offline-moonshine.js) demonstrates

nodejs-examples/package.json

@@ -1,7 +1,7 @@
 {
   "dependencies": {
     "naudiodon2": "^2.4.0",
-    "sherpa-onnx": "*",
+    "sherpa-onnx": "^1.10.44",
     "wav": "^1.0.2"
   }
 }

nodejs-examples/test-offline-fire-red-asr.js

@@ -0,0 +1,37 @@
+// Copyright (c)  2025  Xiaomi Corporation (authors: Fangjun Kuang)
+//
+const sherpa_onnx = require('sherpa-onnx');
+
+function createOfflineRecognizer() {
+  let modelConfig = {
+    fireRedAsr: {
+      encoder:
+          './sherpa-onnx-fire-red-asr-large-zh_en-2025-02-16/encoder.int8.onnx',
+      decoder:
+          './sherpa-onnx-fire-red-asr-large-zh_en-2025-02-16/decoder.int8.onnx',
+    },
+    tokens: './sherpa-onnx-fire-red-asr-large-zh_en-2025-02-16/tokens.txt',
+    debug: 1,
+  };
+
+  let config = {
+    modelConfig: modelConfig,
+  };
+
+  return sherpa_onnx.createOfflineRecognizer(config);
+}
+
+recognizer = createOfflineRecognizer();
+stream = recognizer.createStream();
+
+const waveFilename =
+    './sherpa-onnx-fire-red-asr-large-zh_en-2025-02-16/test_wavs/0.wav';
+const wave = sherpa_onnx.readWave(waveFilename);
+stream.acceptWaveform(wave.sampleRate, wave.samples);
+
+recognizer.decode(stream);
+const text = recognizer.getResult(stream).text;
+console.log(text);
+
+stream.free();
+recognizer.free();

wasm/asr/sherpa-onnx-asr.js

@@ -35,6 +35,10 @@ function freeConfig(config, Module) {
     freeConfig(config.whisper, Module)
   }
 
+  if ('fireRedAsr' in config) {
+    freeConfig(config.fireRedAsr, Module)
+  }
+
   if ('moonshine' in config) {
     freeConfig(config.moonshine, Module)
   }
@@ -651,6 +655,35 @@ function initSherpaOnnxOfflineMoonshineModelConfig(config, Module) {
   }
 }
 
+function initSherpaOnnxOfflineFireRedAsrModelConfig(config, Module) {
+  const encoderLen = Module.lengthBytesUTF8(config.encoder || '') + 1;
+  const decoderLen = Module.lengthBytesUTF8(config.decoder || '') + 1;
+
+  const n = encoderLen + decoderLen;
+  const buffer = Module._malloc(n);
+
+  const len = 2 * 4;  // 2 pointers
+  const ptr = Module._malloc(len);
+
+  let offset = 0;
+  Module.stringToUTF8(config.encoder || '', buffer + offset, encoderLen);
+  offset += encoderLen;
+
+  Module.stringToUTF8(config.decoder || '', buffer + offset, decoderLen);
+  offset += decoderLen;
+
+  offset = 0;
+  Module.setValue(ptr, buffer + offset, 'i8*');
+  offset += encoderLen;
+
+  Module.setValue(ptr + 4, buffer + offset, 'i8*');
+  offset += decoderLen;
+
+  return {
+    buffer: buffer, ptr: ptr, len: len,
+  }
+}
+
 function initSherpaOnnxOfflineTdnnModelConfig(config, Module) {
   const n = Module.lengthBytesUTF8(config.model || '') + 1;
   const buffer = Module._malloc(n);
@@ -755,6 +788,13 @@ function initSherpaOnnxOfflineModelConfig(config, Module) {
     };
   }
 
+  if (!('fireRedAsr' in config)) {
+    config.fireRedAsr = {
+      encoder: '',
+      decoder: '',
+    };
+  }
+
   if (!('tdnn' in config)) {
     config.tdnn = {
       model: '',
@@ -789,8 +829,11 @@ function initSherpaOnnxOfflineModelConfig(config, Module) {
   const moonshine =
       initSherpaOnnxOfflineMoonshineModelConfig(config.moonshine, Module);
 
+  const fireRedAsr =
+      initSherpaOnnxOfflineFireRedAsrModelConfig(config.fireRedAsr, Module);
+
   const len = transducer.len + paraformer.len + nemoCtc.len + whisper.len +
-      tdnn.len + 8 * 4 + senseVoice.len + moonshine.len;
+      tdnn.len + 8 * 4 + senseVoice.len + moonshine.len + fireRedAsr.len;
 
   const ptr = Module._malloc(len);
 
@@ -884,11 +927,15 @@ function initSherpaOnnxOfflineModelConfig(config, Module) {
   offset += senseVoice.len;
 
   Module._CopyHeap(moonshine.ptr, moonshine.len, ptr + offset);
+  offset += moonshine.len;
+
+  Module._CopyHeap(fireRedAsr.ptr, fireRedAsr.len, ptr + offset);
+  offset += fireRedAsr.len;
 
   return {
     buffer: buffer, ptr: ptr, len: len, transducer: transducer,
         paraformer: paraformer, nemoCtc: nemoCtc, whisper: whisper, tdnn: tdnn,
-        senseVoice: senseVoice, moonshine: moonshine,
+        senseVoice: senseVoice, moonshine: moonshine, fireRedAsr: fireRedAsr
   }
 }
 

wasm/nodejs/sherpa-onnx-wasm-nodejs.cc

@@ -15,6 +15,7 @@ static_assert(sizeof(SherpaOnnxOfflineParaformerModelConfig) == 4, "");
 
 static_assert(sizeof(SherpaOnnxOfflineNemoEncDecCtcModelConfig) == 4, "");
 static_assert(sizeof(SherpaOnnxOfflineWhisperModelConfig) == 5 * 4, "");
+static_assert(sizeof(SherpaOnnxOfflineFireRedAsrModelConfig) == 2 * 4, "");
 static_assert(sizeof(SherpaOnnxOfflineMoonshineModelConfig) == 4 * 4, "");
 static_assert(sizeof(SherpaOnnxOfflineTdnnModelConfig) == 4, "");
 static_assert(sizeof(SherpaOnnxOfflineSenseVoiceModelConfig) == 3 * 4, "");
@@ -27,7 +28,9 @@ static_assert(sizeof(SherpaOnnxOfflineModelConfig) ==
                       sizeof(SherpaOnnxOfflineWhisperModelConfig) +
                       sizeof(SherpaOnnxOfflineTdnnModelConfig) + 8 * 4 +
                       sizeof(SherpaOnnxOfflineSenseVoiceModelConfig) +
-                      sizeof(SherpaOnnxOfflineMoonshineModelConfig),
+                      sizeof(SherpaOnnxOfflineMoonshineModelConfig) +
+                      sizeof(SherpaOnnxOfflineFireRedAsrModelConfig),
+
               "");
 static_assert(sizeof(SherpaOnnxFeatureConfig) == 2 * 4, "");
 static_assert(sizeof(SherpaOnnxOfflineRecognizerConfig) ==
@@ -69,6 +72,7 @@ void PrintOfflineRecognizerConfig(SherpaOnnxOfflineRecognizerConfig *config) {
   auto tdnn = &model_config->tdnn;
   auto sense_voice = &model_config->sense_voice;
   auto moonshine = &model_config->moonshine;
+  auto fire_red_asr = &model_config->fire_red_asr;
 
   fprintf(stdout, "----------offline transducer model config----------\n");
   fprintf(stdout, "encoder: %s\n", transducer->encoder);
@@ -102,6 +106,10 @@ void PrintOfflineRecognizerConfig(SherpaOnnxOfflineRecognizerConfig *config) {
   fprintf(stdout, "uncached_decoder: %s\n", moonshine->uncached_decoder);
   fprintf(stdout, "cached_decoder: %s\n", moonshine->cached_decoder);
 
+  fprintf(stdout, "----------offline FireRedAsr model config----------\n");
+  fprintf(stdout, "encoder: %s\n", fire_red_asr->encoder);
+  fprintf(stdout, "decoder: %s\n", fire_red_asr->decoder);
+
   fprintf(stdout, "tokens: %s\n", model_config->tokens);
   fprintf(stdout, "num_threads: %d\n", model_config->num_threads);
   fprintf(stdout, "provider: %s\n", model_config->provider);