Spaces:
Configuration error
Configuration error
Commit
·
fff6695
1
Parent(s):
85af770
update model
Browse files- app-asr.js +1 -1
- sherpa-onnx-asr.js +469 -124
- sherpa-onnx-wasm-main-asr.js +0 -0
app-asr.js
CHANGED
|
@@ -45,7 +45,7 @@ Module.onRuntimeInitialized = function() {
|
|
| 45 |
|
| 46 |
startBtn.disabled = false;
|
| 47 |
|
| 48 |
-
recognizer =
|
| 49 |
console.log('recognizer is created!', recognizer);
|
| 50 |
};
|
| 51 |
|
|
|
|
| 45 |
|
| 46 |
startBtn.disabled = false;
|
| 47 |
|
| 48 |
+
recognizer = createOnlineRecognizer(Module);
|
| 49 |
console.log('recognizer is created!', recognizer);
|
| 50 |
};
|
| 51 |
|
sherpa-onnx-asr.js
CHANGED
|
@@ -1,161 +1,181 @@
|
|
| 1 |
-
function freeConfig(config) {
|
| 2 |
if ('buffer' in config) {
|
| 3 |
-
_free(config.buffer);
|
| 4 |
}
|
| 5 |
|
| 6 |
if ('config' in config) {
|
| 7 |
-
freeConfig(config.config)
|
| 8 |
}
|
| 9 |
|
| 10 |
if ('transducer' in config) {
|
| 11 |
-
freeConfig(config.transducer)
|
| 12 |
}
|
| 13 |
|
| 14 |
if ('paraformer' in config) {
|
| 15 |
-
freeConfig(config.paraformer)
|
| 16 |
}
|
| 17 |
|
| 18 |
if ('ctc' in config) {
|
| 19 |
-
freeConfig(config.ctc)
|
| 20 |
}
|
| 21 |
|
| 22 |
if ('feat' in config) {
|
| 23 |
-
freeConfig(config.feat)
|
| 24 |
}
|
| 25 |
|
| 26 |
if ('model' in config) {
|
| 27 |
-
freeConfig(config.model)
|
| 28 |
}
|
| 29 |
|
| 30 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 31 |
}
|
| 32 |
|
| 33 |
// The user should free the returned pointers
|
| 34 |
-
function initSherpaOnnxOnlineTransducerModelConfig(config) {
|
| 35 |
-
|
| 36 |
-
|
| 37 |
-
|
| 38 |
|
| 39 |
-
|
| 40 |
|
| 41 |
-
|
| 42 |
|
| 43 |
-
|
| 44 |
-
|
| 45 |
|
| 46 |
let offset = 0;
|
| 47 |
-
stringToUTF8(config.encoder, buffer + offset, encoderLen);
|
| 48 |
offset += encoderLen;
|
| 49 |
|
| 50 |
-
stringToUTF8(config.decoder, buffer + offset, decoderLen);
|
| 51 |
offset += decoderLen;
|
| 52 |
|
| 53 |
-
stringToUTF8(config.joiner, buffer + offset, joinerLen);
|
| 54 |
|
| 55 |
offset = 0;
|
| 56 |
-
setValue(ptr, buffer + offset, 'i8*');
|
| 57 |
offset += encoderLen;
|
| 58 |
|
| 59 |
-
setValue(ptr + 4, buffer + offset, 'i8*');
|
| 60 |
offset += decoderLen;
|
| 61 |
|
| 62 |
-
setValue(ptr + 8, buffer + offset, 'i8*');
|
| 63 |
|
| 64 |
return {
|
| 65 |
buffer: buffer, ptr: ptr, len: len,
|
| 66 |
}
|
| 67 |
}
|
| 68 |
|
| 69 |
-
function initSherpaOnnxOnlineParaformerModelConfig(config) {
|
| 70 |
-
|
| 71 |
-
|
| 72 |
|
| 73 |
-
|
| 74 |
-
|
| 75 |
|
| 76 |
-
|
| 77 |
-
|
| 78 |
|
| 79 |
let offset = 0;
|
| 80 |
-
stringToUTF8(config.encoder, buffer + offset, encoderLen);
|
| 81 |
offset += encoderLen;
|
| 82 |
|
| 83 |
-
stringToUTF8(config.decoder, buffer + offset, decoderLen);
|
| 84 |
|
| 85 |
offset = 0;
|
| 86 |
-
setValue(ptr, buffer + offset, 'i8*');
|
| 87 |
offset += encoderLen;
|
| 88 |
|
| 89 |
-
setValue(ptr + 4, buffer + offset, 'i8*');
|
| 90 |
|
| 91 |
return {
|
| 92 |
buffer: buffer, ptr: ptr, len: len,
|
| 93 |
}
|
| 94 |
}
|
| 95 |
|
| 96 |
-
function initSherpaOnnxOnlineZipformer2CtcModelConfig(config) {
|
| 97 |
-
|
| 98 |
-
|
| 99 |
|
| 100 |
-
|
| 101 |
-
|
| 102 |
|
| 103 |
-
stringToUTF8(config.model, buffer, n);
|
| 104 |
|
| 105 |
-
setValue(ptr, buffer, 'i8*');
|
| 106 |
|
| 107 |
return {
|
| 108 |
buffer: buffer, ptr: ptr, len: len,
|
| 109 |
}
|
| 110 |
}
|
| 111 |
|
| 112 |
-
function initSherpaOnnxOnlineModelConfig(config) {
|
| 113 |
-
|
| 114 |
-
|
| 115 |
-
|
|
|
|
|
|
|
|
|
|
| 116 |
|
| 117 |
-
|
| 118 |
-
|
| 119 |
|
| 120 |
let offset = 0;
|
| 121 |
-
_CopyHeap(transducer.ptr, transducer.len, ptr + offset);
|
| 122 |
offset += transducer.len;
|
| 123 |
|
| 124 |
-
_CopyHeap(paraformer.ptr, paraformer.len, ptr + offset);
|
| 125 |
offset += paraformer.len;
|
| 126 |
|
| 127 |
-
_CopyHeap(ctc.ptr, ctc.len, ptr + offset);
|
| 128 |
offset += ctc.len;
|
| 129 |
|
| 130 |
-
|
| 131 |
-
|
| 132 |
-
|
| 133 |
-
|
| 134 |
-
|
| 135 |
|
| 136 |
offset = 0;
|
| 137 |
-
stringToUTF8(config.tokens, buffer, tokensLen);
|
| 138 |
offset += tokensLen;
|
| 139 |
|
| 140 |
-
stringToUTF8(config.provider, buffer + offset, providerLen);
|
| 141 |
offset += providerLen;
|
| 142 |
|
| 143 |
-
stringToUTF8(config.modelType, buffer + offset, modelTypeLen);
|
| 144 |
|
| 145 |
offset = transducer.len + paraformer.len + ctc.len;
|
| 146 |
-
setValue(ptr + offset, buffer, 'i8*'); // tokens
|
| 147 |
offset += 4;
|
| 148 |
|
| 149 |
-
setValue(ptr + offset, config.numThreads, 'i32');
|
| 150 |
offset += 4;
|
| 151 |
|
| 152 |
-
setValue(ptr + offset, buffer + tokensLen, 'i8*'); // provider
|
| 153 |
offset += 4;
|
| 154 |
|
| 155 |
-
setValue(ptr + offset, config.debug, 'i32');
|
| 156 |
offset += 4;
|
| 157 |
|
| 158 |
-
setValue(
|
|
|
|
| 159 |
offset += 4;
|
| 160 |
|
| 161 |
return {
|
|
@@ -164,63 +184,63 @@ function initSherpaOnnxOnlineModelConfig(config) {
|
|
| 164 |
}
|
| 165 |
}
|
| 166 |
|
| 167 |
-
function initSherpaOnnxFeatureConfig(config) {
|
| 168 |
-
|
| 169 |
-
|
| 170 |
|
| 171 |
-
setValue(ptr, config.sampleRate, 'i32');
|
| 172 |
-
setValue(ptr + 4, config.featureDim, 'i32');
|
| 173 |
return {ptr: ptr, len: len};
|
| 174 |
}
|
| 175 |
|
| 176 |
-
function initSherpaOnnxOnlineRecognizerConfig(config) {
|
| 177 |
-
|
| 178 |
-
|
| 179 |
|
| 180 |
-
|
| 181 |
-
|
| 182 |
|
| 183 |
let offset = 0;
|
| 184 |
-
_CopyHeap(feat.ptr, feat.len, ptr + offset);
|
| 185 |
offset += feat.len;
|
| 186 |
|
| 187 |
-
_CopyHeap(model.ptr, model.len, ptr + offset);
|
| 188 |
offset += model.len;
|
| 189 |
|
| 190 |
-
|
| 191 |
-
|
| 192 |
-
|
| 193 |
-
|
| 194 |
|
| 195 |
offset = 0;
|
| 196 |
-
stringToUTF8(config.decodingMethod, buffer, decodingMethodLen);
|
| 197 |
offset += decodingMethodLen;
|
| 198 |
|
| 199 |
-
stringToUTF8(config.hotwordsFile, buffer + offset, hotwordsFileLen);
|
| 200 |
|
| 201 |
offset = feat.len + model.len;
|
| 202 |
-
setValue(ptr + offset, buffer, 'i8*'); // decoding method
|
| 203 |
offset += 4;
|
| 204 |
|
| 205 |
-
setValue(ptr + offset, config.maxActivePaths, 'i32');
|
| 206 |
offset += 4;
|
| 207 |
|
| 208 |
-
setValue(ptr + offset, config.enableEndpoint, 'i32');
|
| 209 |
offset += 4;
|
| 210 |
|
| 211 |
-
setValue(ptr + offset, config.rule1MinTrailingSilence, 'float');
|
| 212 |
offset += 4;
|
| 213 |
|
| 214 |
-
setValue(ptr + offset, config.rule2MinTrailingSilence, 'float');
|
| 215 |
offset += 4;
|
| 216 |
|
| 217 |
-
setValue(ptr + offset, config.rule3MinUtteranceLength, 'float');
|
| 218 |
offset += 4;
|
| 219 |
|
| 220 |
-
setValue(ptr + offset, buffer + decodingMethodLen, 'i8*');
|
| 221 |
offset += 4;
|
| 222 |
|
| 223 |
-
setValue(ptr + offset, config.hotwordsScore, 'float');
|
| 224 |
offset += 4;
|
| 225 |
|
| 226 |
return {
|
|
@@ -229,21 +249,21 @@ function initSherpaOnnxOnlineRecognizerConfig(config) {
|
|
| 229 |
}
|
| 230 |
|
| 231 |
|
| 232 |
-
function
|
| 233 |
-
|
| 234 |
encoder: '',
|
| 235 |
decoder: '',
|
| 236 |
joiner: '',
|
| 237 |
-
}
|
| 238 |
|
| 239 |
-
|
| 240 |
encoder: '',
|
| 241 |
decoder: '',
|
| 242 |
-
}
|
| 243 |
|
| 244 |
-
|
| 245 |
model: '',
|
| 246 |
-
}
|
| 247 |
|
| 248 |
let type = 0;
|
| 249 |
|
|
@@ -266,7 +286,7 @@ function createRecognizer() {
|
|
| 266 |
}
|
| 267 |
|
| 268 |
|
| 269 |
-
|
| 270 |
transducer: onlineTransducerModelConfig,
|
| 271 |
paraformer: onlineParaformerModelConfig,
|
| 272 |
zipformer2Ctc: onlineZipformer2CtcModelConfig,
|
|
@@ -275,12 +295,12 @@ function createRecognizer() {
|
|
| 275 |
provider: 'cpu',
|
| 276 |
debug: 1,
|
| 277 |
modelType: '',
|
| 278 |
-
}
|
| 279 |
|
| 280 |
-
|
| 281 |
sampleRate: 16000,
|
| 282 |
featureDim: 80,
|
| 283 |
-
}
|
| 284 |
|
| 285 |
let recognizerConfig = {
|
| 286 |
featConfig: featureConfig,
|
|
@@ -293,23 +313,336 @@ function createRecognizer() {
|
|
| 293 |
rule3MinUtteranceLength: 20,
|
| 294 |
hotwordsFile: '',
|
| 295 |
hotwordsScore: 1.5,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 296 |
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 297 |
|
| 298 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 299 |
}
|
| 300 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 301 |
class OnlineStream {
|
| 302 |
-
constructor(handle) {
|
| 303 |
this.handle = handle;
|
| 304 |
this.pointer = null; // buffer
|
| 305 |
this.n = 0; // buffer size
|
|
|
|
| 306 |
}
|
| 307 |
|
| 308 |
free() {
|
| 309 |
if (this.handle) {
|
| 310 |
-
_DestroyOnlineStream(this.handle);
|
| 311 |
this.handle = null;
|
| 312 |
-
_free(this.pointer);
|
| 313 |
this.pointer = null;
|
| 314 |
this.n = 0;
|
| 315 |
}
|
|
@@ -321,61 +654,73 @@ class OnlineStream {
|
|
| 321 |
*/
|
| 322 |
acceptWaveform(sampleRate, samples) {
|
| 323 |
if (this.n < samples.length) {
|
| 324 |
-
_free(this.pointer);
|
| 325 |
-
this.pointer =
|
|
|
|
| 326 |
this.n = samples.length
|
| 327 |
}
|
| 328 |
|
| 329 |
-
Module.HEAPF32.set(samples, this.pointer / samples.BYTES_PER_ELEMENT);
|
| 330 |
-
|
|
|
|
| 331 |
}
|
| 332 |
|
| 333 |
inputFinished() {
|
| 334 |
-
_InputFinished(this.handle);
|
| 335 |
}
|
| 336 |
};
|
| 337 |
|
| 338 |
class OnlineRecognizer {
|
| 339 |
-
constructor(configObj) {
|
| 340 |
-
|
| 341 |
-
|
|
|
|
| 342 |
|
| 343 |
-
freeConfig(config);
|
| 344 |
|
| 345 |
this.handle = handle;
|
|
|
|
| 346 |
}
|
| 347 |
|
| 348 |
free() {
|
| 349 |
-
_DestroyOnlineRecognizer(this.handle);
|
| 350 |
this.handle = 0
|
| 351 |
}
|
| 352 |
|
| 353 |
createStream() {
|
| 354 |
-
|
| 355 |
-
return new OnlineStream(handle);
|
| 356 |
}
|
| 357 |
|
| 358 |
isReady(stream) {
|
| 359 |
-
return _IsOnlineStreamReady(this.handle, stream.handle) == 1;
|
| 360 |
}
|
| 361 |
|
| 362 |
decode(stream) {
|
| 363 |
-
|
| 364 |
}
|
| 365 |
|
| 366 |
isEndpoint(stream) {
|
| 367 |
-
return _IsEndpoint(this.handle, stream.handle) == 1;
|
| 368 |
}
|
| 369 |
|
| 370 |
reset(stream) {
|
| 371 |
-
_Reset(this.handle, stream.handle);
|
| 372 |
}
|
| 373 |
|
| 374 |
getResult(stream) {
|
| 375 |
-
|
| 376 |
-
|
| 377 |
-
|
| 378 |
-
_DestroyOnlineRecognizerResult(r);
|
| 379 |
return text;
|
| 380 |
}
|
| 381 |
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
function freeConfig(config, Module) {
|
| 2 |
if ('buffer' in config) {
|
| 3 |
+
Module._free(config.buffer);
|
| 4 |
}
|
| 5 |
|
| 6 |
if ('config' in config) {
|
| 7 |
+
freeConfig(config.config, Module)
|
| 8 |
}
|
| 9 |
|
| 10 |
if ('transducer' in config) {
|
| 11 |
+
freeConfig(config.transducer, Module)
|
| 12 |
}
|
| 13 |
|
| 14 |
if ('paraformer' in config) {
|
| 15 |
+
freeConfig(config.paraformer, Module)
|
| 16 |
}
|
| 17 |
|
| 18 |
if ('ctc' in config) {
|
| 19 |
+
freeConfig(config.ctc, Module)
|
| 20 |
}
|
| 21 |
|
| 22 |
if ('feat' in config) {
|
| 23 |
+
freeConfig(config.feat, Module)
|
| 24 |
}
|
| 25 |
|
| 26 |
if ('model' in config) {
|
| 27 |
+
freeConfig(config.model, Module)
|
| 28 |
}
|
| 29 |
|
| 30 |
+
if ('nemoCtc' in config) {
|
| 31 |
+
freeConfig(config.nemoCtc, Module)
|
| 32 |
+
}
|
| 33 |
+
|
| 34 |
+
if ('whisper' in config) {
|
| 35 |
+
freeConfig(config.whisper, Module)
|
| 36 |
+
}
|
| 37 |
+
|
| 38 |
+
if ('tdnn' in config) {
|
| 39 |
+
freeConfig(config.tdnn, Module)
|
| 40 |
+
}
|
| 41 |
+
|
| 42 |
+
if ('lm' in config) {
|
| 43 |
+
freeConfig(config.lm, Module)
|
| 44 |
+
}
|
| 45 |
+
|
| 46 |
+
Module._free(config.ptr);
|
| 47 |
}
|
| 48 |
|
| 49 |
// The user should free the returned pointers
|
| 50 |
+
function initSherpaOnnxOnlineTransducerModelConfig(config, Module) {
|
| 51 |
+
const encoderLen = Module.lengthBytesUTF8(config.encoder) + 1;
|
| 52 |
+
const decoderLen = Module.lengthBytesUTF8(config.decoder) + 1;
|
| 53 |
+
const joinerLen = Module.lengthBytesUTF8(config.joiner) + 1;
|
| 54 |
|
| 55 |
+
const n = encoderLen + decoderLen + joinerLen;
|
| 56 |
|
| 57 |
+
const buffer = Module._malloc(n);
|
| 58 |
|
| 59 |
+
const len = 3 * 4; // 3 pointers
|
| 60 |
+
const ptr = Module._malloc(len);
|
| 61 |
|
| 62 |
let offset = 0;
|
| 63 |
+
Module.stringToUTF8(config.encoder, buffer + offset, encoderLen);
|
| 64 |
offset += encoderLen;
|
| 65 |
|
| 66 |
+
Module.stringToUTF8(config.decoder, buffer + offset, decoderLen);
|
| 67 |
offset += decoderLen;
|
| 68 |
|
| 69 |
+
Module.stringToUTF8(config.joiner, buffer + offset, joinerLen);
|
| 70 |
|
| 71 |
offset = 0;
|
| 72 |
+
Module.setValue(ptr, buffer + offset, 'i8*');
|
| 73 |
offset += encoderLen;
|
| 74 |
|
| 75 |
+
Module.setValue(ptr + 4, buffer + offset, 'i8*');
|
| 76 |
offset += decoderLen;
|
| 77 |
|
| 78 |
+
Module.setValue(ptr + 8, buffer + offset, 'i8*');
|
| 79 |
|
| 80 |
return {
|
| 81 |
buffer: buffer, ptr: ptr, len: len,
|
| 82 |
}
|
| 83 |
}
|
| 84 |
|
| 85 |
+
function initSherpaOnnxOnlineParaformerModelConfig(config, Module) {
|
| 86 |
+
const encoderLen = Module.lengthBytesUTF8(config.encoder) + 1;
|
| 87 |
+
const decoderLen = Module.lengthBytesUTF8(config.decoder) + 1;
|
| 88 |
|
| 89 |
+
const n = encoderLen + decoderLen;
|
| 90 |
+
const buffer = Module._malloc(n);
|
| 91 |
|
| 92 |
+
const len = 2 * 4; // 2 pointers
|
| 93 |
+
const ptr = Module._malloc(len);
|
| 94 |
|
| 95 |
let offset = 0;
|
| 96 |
+
Module.stringToUTF8(config.encoder, buffer + offset, encoderLen);
|
| 97 |
offset += encoderLen;
|
| 98 |
|
| 99 |
+
Module.stringToUTF8(config.decoder, buffer + offset, decoderLen);
|
| 100 |
|
| 101 |
offset = 0;
|
| 102 |
+
Module.setValue(ptr, buffer + offset, 'i8*');
|
| 103 |
offset += encoderLen;
|
| 104 |
|
| 105 |
+
Module.setValue(ptr + 4, buffer + offset, 'i8*');
|
| 106 |
|
| 107 |
return {
|
| 108 |
buffer: buffer, ptr: ptr, len: len,
|
| 109 |
}
|
| 110 |
}
|
| 111 |
|
| 112 |
+
function initSherpaOnnxOnlineZipformer2CtcModelConfig(config, Module) {
|
| 113 |
+
const n = Module.lengthBytesUTF8(config.model) + 1;
|
| 114 |
+
const buffer = Module._malloc(n);
|
| 115 |
|
| 116 |
+
const len = 1 * 4; // 1 pointer
|
| 117 |
+
const ptr = Module._malloc(len);
|
| 118 |
|
| 119 |
+
Module.stringToUTF8(config.model, buffer, n);
|
| 120 |
|
| 121 |
+
Module.setValue(ptr, buffer, 'i8*');
|
| 122 |
|
| 123 |
return {
|
| 124 |
buffer: buffer, ptr: ptr, len: len,
|
| 125 |
}
|
| 126 |
}
|
| 127 |
|
| 128 |
+
function initSherpaOnnxOnlineModelConfig(config, Module) {
|
| 129 |
+
const transducer =
|
| 130 |
+
initSherpaOnnxOnlineTransducerModelConfig(config.transducer, Module);
|
| 131 |
+
const paraformer =
|
| 132 |
+
initSherpaOnnxOnlineParaformerModelConfig(config.paraformer, Module);
|
| 133 |
+
const ctc = initSherpaOnnxOnlineZipformer2CtcModelConfig(
|
| 134 |
+
config.zipformer2Ctc, Module);
|
| 135 |
|
| 136 |
+
const len = transducer.len + paraformer.len + ctc.len + 5 * 4;
|
| 137 |
+
const ptr = Module._malloc(len);
|
| 138 |
|
| 139 |
let offset = 0;
|
| 140 |
+
Module._CopyHeap(transducer.ptr, transducer.len, ptr + offset);
|
| 141 |
offset += transducer.len;
|
| 142 |
|
| 143 |
+
Module._CopyHeap(paraformer.ptr, paraformer.len, ptr + offset);
|
| 144 |
offset += paraformer.len;
|
| 145 |
|
| 146 |
+
Module._CopyHeap(ctc.ptr, ctc.len, ptr + offset);
|
| 147 |
offset += ctc.len;
|
| 148 |
|
| 149 |
+
const tokensLen = Module.lengthBytesUTF8(config.tokens) + 1;
|
| 150 |
+
const providerLen = Module.lengthBytesUTF8(config.provider) + 1;
|
| 151 |
+
const modelTypeLen = Module.lengthBytesUTF8(config.modelType) + 1;
|
| 152 |
+
const bufferLen = tokensLen + providerLen + modelTypeLen;
|
| 153 |
+
const buffer = Module._malloc(bufferLen);
|
| 154 |
|
| 155 |
offset = 0;
|
| 156 |
+
Module.stringToUTF8(config.tokens, buffer, tokensLen);
|
| 157 |
offset += tokensLen;
|
| 158 |
|
| 159 |
+
Module.stringToUTF8(config.provider, buffer + offset, providerLen);
|
| 160 |
offset += providerLen;
|
| 161 |
|
| 162 |
+
Module.stringToUTF8(config.modelType, buffer + offset, modelTypeLen);
|
| 163 |
|
| 164 |
offset = transducer.len + paraformer.len + ctc.len;
|
| 165 |
+
Module.setValue(ptr + offset, buffer, 'i8*'); // tokens
|
| 166 |
offset += 4;
|
| 167 |
|
| 168 |
+
Module.setValue(ptr + offset, config.numThreads, 'i32');
|
| 169 |
offset += 4;
|
| 170 |
|
| 171 |
+
Module.setValue(ptr + offset, buffer + tokensLen, 'i8*'); // provider
|
| 172 |
offset += 4;
|
| 173 |
|
| 174 |
+
Module.setValue(ptr + offset, config.debug, 'i32');
|
| 175 |
offset += 4;
|
| 176 |
|
| 177 |
+
Module.setValue(
|
| 178 |
+
ptr + offset, buffer + tokensLen + providerLen, 'i8*'); // modelType
|
| 179 |
offset += 4;
|
| 180 |
|
| 181 |
return {
|
|
|
|
| 184 |
}
|
| 185 |
}
|
| 186 |
|
| 187 |
+
function initSherpaOnnxFeatureConfig(config, Module) {
|
| 188 |
+
const len = 2 * 4; // 2 pointers
|
| 189 |
+
const ptr = Module._malloc(len);
|
| 190 |
|
| 191 |
+
Module.setValue(ptr, config.sampleRate, 'i32');
|
| 192 |
+
Module.setValue(ptr + 4, config.featureDim, 'i32');
|
| 193 |
return {ptr: ptr, len: len};
|
| 194 |
}
|
| 195 |
|
| 196 |
+
function initSherpaOnnxOnlineRecognizerConfig(config, Module) {
|
| 197 |
+
const feat = initSherpaOnnxFeatureConfig(config.featConfig, Module);
|
| 198 |
+
const model = initSherpaOnnxOnlineModelConfig(config.modelConfig, Module);
|
| 199 |
|
| 200 |
+
const len = feat.len + model.len + 8 * 4;
|
| 201 |
+
const ptr = Module._malloc(len);
|
| 202 |
|
| 203 |
let offset = 0;
|
| 204 |
+
Module._CopyHeap(feat.ptr, feat.len, ptr + offset);
|
| 205 |
offset += feat.len;
|
| 206 |
|
| 207 |
+
Module._CopyHeap(model.ptr, model.len, ptr + offset);
|
| 208 |
offset += model.len;
|
| 209 |
|
| 210 |
+
const decodingMethodLen = Module.lengthBytesUTF8(config.decodingMethod) + 1;
|
| 211 |
+
const hotwordsFileLen = Module.lengthBytesUTF8(config.hotwordsFile) + 1;
|
| 212 |
+
const bufferLen = decodingMethodLen + hotwordsFileLen;
|
| 213 |
+
const buffer = Module._malloc(bufferLen);
|
| 214 |
|
| 215 |
offset = 0;
|
| 216 |
+
Module.stringToUTF8(config.decodingMethod, buffer, decodingMethodLen);
|
| 217 |
offset += decodingMethodLen;
|
| 218 |
|
| 219 |
+
Module.stringToUTF8(config.hotwordsFile, buffer + offset, hotwordsFileLen);
|
| 220 |
|
| 221 |
offset = feat.len + model.len;
|
| 222 |
+
Module.setValue(ptr + offset, buffer, 'i8*'); // decoding method
|
| 223 |
offset += 4;
|
| 224 |
|
| 225 |
+
Module.setValue(ptr + offset, config.maxActivePaths, 'i32');
|
| 226 |
offset += 4;
|
| 227 |
|
| 228 |
+
Module.setValue(ptr + offset, config.enableEndpoint, 'i32');
|
| 229 |
offset += 4;
|
| 230 |
|
| 231 |
+
Module.setValue(ptr + offset, config.rule1MinTrailingSilence, 'float');
|
| 232 |
offset += 4;
|
| 233 |
|
| 234 |
+
Module.setValue(ptr + offset, config.rule2MinTrailingSilence, 'float');
|
| 235 |
offset += 4;
|
| 236 |
|
| 237 |
+
Module.setValue(ptr + offset, config.rule3MinUtteranceLength, 'float');
|
| 238 |
offset += 4;
|
| 239 |
|
| 240 |
+
Module.setValue(ptr + offset, buffer + decodingMethodLen, 'i8*');
|
| 241 |
offset += 4;
|
| 242 |
|
| 243 |
+
Module.setValue(ptr + offset, config.hotwordsScore, 'float');
|
| 244 |
offset += 4;
|
| 245 |
|
| 246 |
return {
|
|
|
|
| 249 |
}
|
| 250 |
|
| 251 |
|
| 252 |
+
function createOnlineRecognizer(Module, myConfig) {
|
| 253 |
+
const onlineTransducerModelConfig = {
|
| 254 |
encoder: '',
|
| 255 |
decoder: '',
|
| 256 |
joiner: '',
|
| 257 |
+
};
|
| 258 |
|
| 259 |
+
const onlineParaformerModelConfig = {
|
| 260 |
encoder: '',
|
| 261 |
decoder: '',
|
| 262 |
+
};
|
| 263 |
|
| 264 |
+
const onlineZipformer2CtcModelConfig = {
|
| 265 |
model: '',
|
| 266 |
+
};
|
| 267 |
|
| 268 |
let type = 0;
|
| 269 |
|
|
|
|
| 286 |
}
|
| 287 |
|
| 288 |
|
| 289 |
+
const onlineModelConfig = {
|
| 290 |
transducer: onlineTransducerModelConfig,
|
| 291 |
paraformer: onlineParaformerModelConfig,
|
| 292 |
zipformer2Ctc: onlineZipformer2CtcModelConfig,
|
|
|
|
| 295 |
provider: 'cpu',
|
| 296 |
debug: 1,
|
| 297 |
modelType: '',
|
| 298 |
+
};
|
| 299 |
|
| 300 |
+
const featureConfig = {
|
| 301 |
sampleRate: 16000,
|
| 302 |
featureDim: 80,
|
| 303 |
+
};
|
| 304 |
|
| 305 |
let recognizerConfig = {
|
| 306 |
featConfig: featureConfig,
|
|
|
|
| 313 |
rule3MinUtteranceLength: 20,
|
| 314 |
hotwordsFile: '',
|
| 315 |
hotwordsScore: 1.5,
|
| 316 |
+
};
|
| 317 |
+
if (myConfig) {
|
| 318 |
+
recognizerConfig = myConfig;
|
| 319 |
+
}
|
| 320 |
+
|
| 321 |
+
return new OnlineRecognizer(recognizerConfig, Module);
|
| 322 |
+
}
|
| 323 |
+
|
| 324 |
+
function initSherpaOnnxOfflineTransducerModelConfig(config, Module) {
|
| 325 |
+
const encoderLen = Module.lengthBytesUTF8(config.encoder) + 1;
|
| 326 |
+
const decoderLen = Module.lengthBytesUTF8(config.decoder) + 1;
|
| 327 |
+
const joinerLen = Module.lengthBytesUTF8(config.joiner) + 1;
|
| 328 |
+
|
| 329 |
+
const n = encoderLen + decoderLen + joinerLen;
|
| 330 |
+
|
| 331 |
+
const buffer = Module._malloc(n);
|
| 332 |
+
|
| 333 |
+
const len = 3 * 4; // 3 pointers
|
| 334 |
+
const ptr = Module._malloc(len);
|
| 335 |
+
|
| 336 |
+
let offset = 0;
|
| 337 |
+
Module.stringToUTF8(config.encoder, buffer + offset, encoderLen);
|
| 338 |
+
offset += encoderLen;
|
| 339 |
+
|
| 340 |
+
Module.stringToUTF8(config.decoder, buffer + offset, decoderLen);
|
| 341 |
+
offset += decoderLen;
|
| 342 |
+
|
| 343 |
+
Module.stringToUTF8(config.joiner, buffer + offset, joinerLen);
|
| 344 |
+
|
| 345 |
+
offset = 0;
|
| 346 |
+
Module.setValue(ptr, buffer + offset, 'i8*');
|
| 347 |
+
offset += encoderLen;
|
| 348 |
+
|
| 349 |
+
Module.setValue(ptr + 4, buffer + offset, 'i8*');
|
| 350 |
+
offset += decoderLen;
|
| 351 |
+
|
| 352 |
+
Module.setValue(ptr + 8, buffer + offset, 'i8*');
|
| 353 |
+
|
| 354 |
+
return {
|
| 355 |
+
buffer: buffer, ptr: ptr, len: len,
|
| 356 |
+
}
|
| 357 |
+
}
|
| 358 |
+
|
| 359 |
+
function initSherpaOnnxOfflineParaformerModelConfig(config, Module) {
|
| 360 |
+
const n = Module.lengthBytesUTF8(config.model) + 1;
|
| 361 |
+
|
| 362 |
+
const buffer = Module._malloc(n);
|
| 363 |
+
|
| 364 |
+
const len = 1 * 4; // 1 pointer
|
| 365 |
+
const ptr = Module._malloc(len);
|
| 366 |
+
|
| 367 |
+
Module.stringToUTF8(config.model, buffer, n);
|
| 368 |
+
|
| 369 |
+
Module.setValue(ptr, buffer, 'i8*');
|
| 370 |
+
|
| 371 |
+
return {
|
| 372 |
+
buffer: buffer, ptr: ptr, len: len,
|
| 373 |
+
}
|
| 374 |
+
}
|
| 375 |
+
|
| 376 |
+
function initSherpaOnnxOfflineNemoEncDecCtcModelConfig(config, Module) {
|
| 377 |
+
const n = Module.lengthBytesUTF8(config.model) + 1;
|
| 378 |
+
|
| 379 |
+
const buffer = Module._malloc(n);
|
| 380 |
+
|
| 381 |
+
const len = 1 * 4; // 1 pointer
|
| 382 |
+
const ptr = Module._malloc(len);
|
| 383 |
+
|
| 384 |
+
Module.stringToUTF8(config.model, buffer, n);
|
| 385 |
+
|
| 386 |
+
Module.setValue(ptr, buffer, 'i8*');
|
| 387 |
+
|
| 388 |
+
return {
|
| 389 |
+
buffer: buffer, ptr: ptr, len: len,
|
| 390 |
+
}
|
| 391 |
+
}
|
| 392 |
+
|
| 393 |
+
function initSherpaOnnxOfflineWhisperModelConfig(config, Module) {
|
| 394 |
+
const encoderLen = Module.lengthBytesUTF8(config.encoder) + 1;
|
| 395 |
+
const decoderLen = Module.lengthBytesUTF8(config.decoder) + 1;
|
| 396 |
+
|
| 397 |
+
const n = encoderLen + decoderLen;
|
| 398 |
+
const buffer = Module._malloc(n);
|
| 399 |
+
|
| 400 |
+
const len = 2 * 4; // 2 pointers
|
| 401 |
+
const ptr = Module._malloc(len);
|
| 402 |
+
|
| 403 |
+
let offset = 0;
|
| 404 |
+
Module.stringToUTF8(config.encoder, buffer + offset, encoderLen);
|
| 405 |
+
offset += encoderLen;
|
| 406 |
+
|
| 407 |
+
Module.stringToUTF8(config.decoder, buffer + offset, decoderLen);
|
| 408 |
+
|
| 409 |
+
offset = 0;
|
| 410 |
+
Module.setValue(ptr, buffer + offset, 'i8*');
|
| 411 |
+
offset += encoderLen;
|
| 412 |
+
|
| 413 |
+
Module.setValue(ptr + 4, buffer + offset, 'i8*');
|
| 414 |
+
|
| 415 |
+
return {
|
| 416 |
+
buffer: buffer, ptr: ptr, len: len,
|
| 417 |
+
}
|
| 418 |
+
}
|
| 419 |
+
|
| 420 |
+
function initSherpaOnnxOfflineTdnnModelConfig(config, Module) {
|
| 421 |
+
const n = Module.lengthBytesUTF8(config.model) + 1;
|
| 422 |
+
const buffer = Module._malloc(n);
|
| 423 |
+
|
| 424 |
+
const len = 1 * 4; // 1 pointer
|
| 425 |
+
const ptr = Module._malloc(len);
|
| 426 |
+
|
| 427 |
+
Module.stringToUTF8(config.model, buffer, n);
|
| 428 |
+
|
| 429 |
+
Module.setValue(ptr, buffer, 'i8*');
|
| 430 |
+
|
| 431 |
+
return {
|
| 432 |
+
buffer: buffer, ptr: ptr, len: len,
|
| 433 |
+
}
|
| 434 |
+
}
|
| 435 |
+
|
| 436 |
+
function initSherpaOnnxOfflineLMConfig(config, Module) {
|
| 437 |
+
const n = Module.lengthBytesUTF8(config.model) + 1;
|
| 438 |
+
const buffer = Module._malloc(n);
|
| 439 |
+
|
| 440 |
+
const len = 2 * 4;
|
| 441 |
+
const ptr = Module._malloc(len);
|
| 442 |
+
|
| 443 |
+
Module.stringToUTF8(config.model, buffer, n);
|
| 444 |
+
Module.setValue(ptr, buffer, 'i8*');
|
| 445 |
+
Module.setValue(ptr + 4, config.scale, 'float');
|
| 446 |
+
|
| 447 |
+
return {
|
| 448 |
+
buffer: buffer, ptr: ptr, len: len,
|
| 449 |
+
}
|
| 450 |
+
}
|
| 451 |
+
|
| 452 |
+
function initSherpaOnnxOfflineModelConfig(config, Module) {
|
| 453 |
+
const transducer =
|
| 454 |
+
initSherpaOnnxOfflineTransducerModelConfig(config.transducer, Module);
|
| 455 |
+
const paraformer =
|
| 456 |
+
initSherpaOnnxOfflineParaformerModelConfig(config.paraformer, Module);
|
| 457 |
+
const nemoCtc =
|
| 458 |
+
initSherpaOnnxOfflineNemoEncDecCtcModelConfig(config.nemoCtc, Module);
|
| 459 |
+
const whisper =
|
| 460 |
+
initSherpaOnnxOfflineWhisperModelConfig(config.whisper, Module);
|
| 461 |
+
const tdnn = initSherpaOnnxOfflineTdnnModelConfig(config.tdnn, Module);
|
| 462 |
+
|
| 463 |
+
const len = transducer.len + paraformer.len + nemoCtc.len + whisper.len +
|
| 464 |
+
tdnn.len + 5 * 4;
|
| 465 |
+
const ptr = Module._malloc(len);
|
| 466 |
+
|
| 467 |
+
let offset = 0;
|
| 468 |
+
Module._CopyHeap(transducer.ptr, transducer.len, ptr + offset);
|
| 469 |
+
offset += transducer.len;
|
| 470 |
+
|
| 471 |
+
Module._CopyHeap(paraformer.ptr, paraformer.len, ptr + offset);
|
| 472 |
+
offset += paraformer.len;
|
| 473 |
+
|
| 474 |
+
Module._CopyHeap(nemoCtc.ptr, nemoCtc.len, ptr + offset);
|
| 475 |
+
offset += nemoCtc.len;
|
| 476 |
+
|
| 477 |
+
Module._CopyHeap(whisper.ptr, whisper.len, ptr + offset);
|
| 478 |
+
offset += whisper.len;
|
| 479 |
+
|
| 480 |
+
Module._CopyHeap(tdnn.ptr, tdnn.len, ptr + offset);
|
| 481 |
+
offset += tdnn.len;
|
| 482 |
+
|
| 483 |
+
const tokensLen = Module.lengthBytesUTF8(config.tokens) + 1;
|
| 484 |
+
const providerLen = Module.lengthBytesUTF8(config.provider) + 1;
|
| 485 |
+
const modelTypeLen = Module.lengthBytesUTF8(config.modelType) + 1;
|
| 486 |
+
const bufferLen = tokensLen + providerLen + modelTypeLen;
|
| 487 |
+
const buffer = Module._malloc(bufferLen);
|
| 488 |
+
|
| 489 |
+
offset = 0;
|
| 490 |
+
Module.stringToUTF8(config.tokens, buffer, tokensLen);
|
| 491 |
+
offset += tokensLen;
|
| 492 |
+
|
| 493 |
+
Module.stringToUTF8(config.provider, buffer + offset, providerLen);
|
| 494 |
+
offset += providerLen;
|
| 495 |
+
|
| 496 |
+
Module.stringToUTF8(config.modelType, buffer + offset, modelTypeLen);
|
| 497 |
+
|
| 498 |
+
offset =
|
| 499 |
+
transducer.len + paraformer.len + nemoCtc.len + whisper.len + tdnn.len;
|
| 500 |
+
Module.setValue(ptr + offset, buffer, 'i8*'); // tokens
|
| 501 |
+
offset += 4;
|
| 502 |
+
|
| 503 |
+
Module.setValue(ptr + offset, config.numThreads, 'i32');
|
| 504 |
+
offset += 4;
|
| 505 |
+
|
| 506 |
+
Module.setValue(ptr + offset, config.debug, 'i32');
|
| 507 |
+
offset += 4;
|
| 508 |
+
|
| 509 |
+
Module.setValue(ptr + offset, buffer + tokensLen, 'i8*'); // provider
|
| 510 |
+
offset += 4;
|
| 511 |
+
|
| 512 |
+
Module.setValue(
|
| 513 |
+
ptr + offset, buffer + tokensLen + providerLen, 'i8*'); // modelType
|
| 514 |
+
offset += 4;
|
| 515 |
+
|
| 516 |
+
return {
|
| 517 |
+
buffer: buffer, ptr: ptr, len: len, transducer: transducer,
|
| 518 |
+
paraformer: paraformer, nemoCtc: nemoCtc, whisper: whisper, tdnn: tdnn
|
| 519 |
}
|
| 520 |
+
}
|
| 521 |
+
|
| 522 |
+
function initSherpaOnnxOfflineRecognizerConfig(config, Module) {
|
| 523 |
+
const feat = initSherpaOnnxFeatureConfig(config.featConfig, Module);
|
| 524 |
+
const model = initSherpaOnnxOfflineModelConfig(config.modelConfig, Module);
|
| 525 |
+
const lm = initSherpaOnnxOfflineLMConfig(config.lmConfig, Module);
|
| 526 |
+
|
| 527 |
+
const len = feat.len + model.len + lm.len + 4 * 4;
|
| 528 |
+
const ptr = Module._malloc(len);
|
| 529 |
+
|
| 530 |
+
let offset = 0;
|
| 531 |
+
Module._CopyHeap(feat.ptr, feat.len, ptr + offset);
|
| 532 |
+
offset += feat.len;
|
| 533 |
+
|
| 534 |
+
Module._CopyHeap(model.ptr, model.len, ptr + offset);
|
| 535 |
+
offset += model.len;
|
| 536 |
+
|
| 537 |
+
Module._CopyHeap(lm.ptr, lm.len, ptr + offset);
|
| 538 |
+
offset += lm.len;
|
| 539 |
+
|
| 540 |
+
const decodingMethodLen = Module.lengthBytesUTF8(config.decodingMethod) + 1;
|
| 541 |
+
const hotwordsFileLen = Module.lengthBytesUTF8(config.hotwordsFile) + 1;
|
| 542 |
+
const bufferLen = decodingMethodLen + hotwordsFileLen;
|
| 543 |
+
const buffer = Module._malloc(bufferLen);
|
| 544 |
+
|
| 545 |
+
offset = 0;
|
| 546 |
+
Module.stringToUTF8(config.decodingMethod, buffer, decodingMethodLen);
|
| 547 |
+
offset += decodingMethodLen;
|
| 548 |
+
|
| 549 |
+
Module.stringToUTF8(config.hotwordsFile, buffer + offset, hotwordsFileLen);
|
| 550 |
+
|
| 551 |
+
offset = feat.len + model.len + lm.len;
|
| 552 |
+
|
| 553 |
+
Module.setValue(ptr + offset, buffer, 'i8*'); // decoding method
|
| 554 |
+
offset += 4;
|
| 555 |
+
|
| 556 |
+
Module.setValue(ptr + offset, config.maxActivePaths, 'i32');
|
| 557 |
+
offset += 4;
|
| 558 |
|
| 559 |
+
Module.setValue(ptr + offset, buffer + decodingMethodLen, 'i8*');
|
| 560 |
+
offset += 4;
|
| 561 |
+
|
| 562 |
+
Module.setValue(ptr + offset, config.hotwordsScore, 'float');
|
| 563 |
+
offset += 4;
|
| 564 |
+
|
| 565 |
+
return {
|
| 566 |
+
buffer: buffer, ptr: ptr, len: len, feat: feat, model: model, lm: lm
|
| 567 |
+
}
|
| 568 |
}
|
| 569 |
|
| 570 |
+
class OfflineStream {
|
| 571 |
+
constructor(handle, Module) {
|
| 572 |
+
this.handle = handle;
|
| 573 |
+
this.Module = Module;
|
| 574 |
+
}
|
| 575 |
+
|
| 576 |
+
free() {
|
| 577 |
+
if (this.handle) {
|
| 578 |
+
this.Module._DestroyOfflineStream(this.handle);
|
| 579 |
+
this.handle = null;
|
| 580 |
+
}
|
| 581 |
+
}
|
| 582 |
+
|
| 583 |
+
/**
|
| 584 |
+
* @param sampleRate {Number}
|
| 585 |
+
* @param samples {Float32Array} Containing samples in the range [-1, 1]
|
| 586 |
+
*/
|
| 587 |
+
acceptWaveform(sampleRate, samples) {
|
| 588 |
+
const pointer =
|
| 589 |
+
this.Module._malloc(samples.length * samples.BYTES_PER_ELEMENT);
|
| 590 |
+
this.Module.HEAPF32.set(samples, pointer / samples.BYTES_PER_ELEMENT);
|
| 591 |
+
this.Module._AcceptWaveformOffline(
|
| 592 |
+
this.handle, sampleRate, pointer, samples.length);
|
| 593 |
+
this.Module._free(pointer);
|
| 594 |
+
}
|
| 595 |
+
};
|
| 596 |
+
|
| 597 |
+
class OfflineRecognizer {
|
| 598 |
+
constructor(configObj, Module) {
|
| 599 |
+
this.config = configObj;
|
| 600 |
+
const config = initSherpaOnnxOfflineRecognizerConfig(configObj, Module);
|
| 601 |
+
const handle = Module._CreateOfflineRecognizer(config.ptr);
|
| 602 |
+
freeConfig(config, Module);
|
| 603 |
+
|
| 604 |
+
this.handle = handle;
|
| 605 |
+
this.Module = Module;
|
| 606 |
+
}
|
| 607 |
+
|
| 608 |
+
free() {
|
| 609 |
+
this.Module._DestroyOfflineRecognizer(this.handle);
|
| 610 |
+
this.handle = 0
|
| 611 |
+
}
|
| 612 |
+
|
| 613 |
+
createStream() {
|
| 614 |
+
const handle = this.Module._CreateOfflineStream(this.handle);
|
| 615 |
+
return new OfflineStream(handle, this.Module);
|
| 616 |
+
}
|
| 617 |
+
|
| 618 |
+
decode(stream) {
|
| 619 |
+
this.Module._DecodeOfflineStream(this.handle, stream.handle);
|
| 620 |
+
}
|
| 621 |
+
|
| 622 |
+
getResult(stream) {
|
| 623 |
+
const r = this.Module._GetOfflineStreamResult(stream.handle);
|
| 624 |
+
|
| 625 |
+
const textPtr = this.Module.getValue(r, 'i8*');
|
| 626 |
+
const text = this.Module.UTF8ToString(textPtr);
|
| 627 |
+
|
| 628 |
+
this.Module._DestroyOfflineRecognizerResult(r);
|
| 629 |
+
return text;
|
| 630 |
+
}
|
| 631 |
+
};
|
| 632 |
+
|
| 633 |
class OnlineStream {
|
| 634 |
+
constructor(handle, Module) {
|
| 635 |
this.handle = handle;
|
| 636 |
this.pointer = null; // buffer
|
| 637 |
this.n = 0; // buffer size
|
| 638 |
+
this.Module = Module;
|
| 639 |
}
|
| 640 |
|
| 641 |
free() {
|
| 642 |
if (this.handle) {
|
| 643 |
+
this.Module._DestroyOnlineStream(this.handle);
|
| 644 |
this.handle = null;
|
| 645 |
+
this.Module._free(this.pointer);
|
| 646 |
this.pointer = null;
|
| 647 |
this.n = 0;
|
| 648 |
}
|
|
|
|
| 654 |
*/
|
| 655 |
acceptWaveform(sampleRate, samples) {
|
| 656 |
if (this.n < samples.length) {
|
| 657 |
+
this.Module._free(this.pointer);
|
| 658 |
+
this.pointer =
|
| 659 |
+
this.Module._malloc(samples.length * samples.BYTES_PER_ELEMENT);
|
| 660 |
this.n = samples.length
|
| 661 |
}
|
| 662 |
|
| 663 |
+
this.Module.HEAPF32.set(samples, this.pointer / samples.BYTES_PER_ELEMENT);
|
| 664 |
+
this.Module._AcceptWaveform(
|
| 665 |
+
this.handle, sampleRate, this.pointer, samples.length);
|
| 666 |
}
|
| 667 |
|
| 668 |
inputFinished() {
|
| 669 |
+
this.Module._InputFinished(this.handle);
|
| 670 |
}
|
| 671 |
};
|
| 672 |
|
| 673 |
class OnlineRecognizer {
|
| 674 |
+
constructor(configObj, Module) {
|
| 675 |
+
this.config = configObj;
|
| 676 |
+
const config = initSherpaOnnxOnlineRecognizerConfig(configObj, Module)
|
| 677 |
+
const handle = Module._CreateOnlineRecognizer(config.ptr);
|
| 678 |
|
| 679 |
+
freeConfig(config, Module);
|
| 680 |
|
| 681 |
this.handle = handle;
|
| 682 |
+
this.Module = Module;
|
| 683 |
}
|
| 684 |
|
| 685 |
free() {
|
| 686 |
+
this.Module._DestroyOnlineRecognizer(this.handle);
|
| 687 |
this.handle = 0
|
| 688 |
}
|
| 689 |
|
| 690 |
createStream() {
|
| 691 |
+
const handle = this.Module._CreateOnlineStream(this.handle);
|
| 692 |
+
return new OnlineStream(handle, this.Module);
|
| 693 |
}
|
| 694 |
|
| 695 |
isReady(stream) {
|
| 696 |
+
return this.Module._IsOnlineStreamReady(this.handle, stream.handle) == 1;
|
| 697 |
}
|
| 698 |
|
| 699 |
decode(stream) {
|
| 700 |
+
this.Module._DecodeOnlineStream(this.handle, stream.handle);
|
| 701 |
}
|
| 702 |
|
| 703 |
isEndpoint(stream) {
|
| 704 |
+
return this.Module._IsEndpoint(this.handle, stream.handle) == 1;
|
| 705 |
}
|
| 706 |
|
| 707 |
reset(stream) {
|
| 708 |
+
this.Module._Reset(this.handle, stream.handle);
|
| 709 |
}
|
| 710 |
|
| 711 |
getResult(stream) {
|
| 712 |
+
const r = this.Module._GetOnlineStreamResult(this.handle, stream.handle);
|
| 713 |
+
const textPtr = this.Module.getValue(r, 'i8*');
|
| 714 |
+
const text = this.Module.UTF8ToString(textPtr);
|
| 715 |
+
this.Module._DestroyOnlineRecognizerResult(r);
|
| 716 |
return text;
|
| 717 |
}
|
| 718 |
}
|
| 719 |
+
|
| 720 |
+
if (typeof process == 'object' && typeof process.versions == 'object' &&
|
| 721 |
+
typeof process.versions.node == 'string') {
|
| 722 |
+
module.exports = {
|
| 723 |
+
createOnlineRecognizer,
|
| 724 |
+
OfflineRecognizer,
|
| 725 |
+
};
|
| 726 |
+
}
|
sherpa-onnx-wasm-main-asr.js
CHANGED
|
The diff for this file is too large to render.
See raw diff
|
|
|