Committed by
GitHub
Support not using external buffers for node-addon (#925)
正在显示
11 个修改的文件
包含
228 行增加
和
82 行删除
| @@ -18,7 +18,7 @@ fi | @@ -18,7 +18,7 @@ fi | ||
| 18 | SHERPA_ONNX_VERSION=$(grep "SHERPA_ONNX_VERSION" ./CMakeLists.txt | cut -d " " -f 2 | cut -d '"' -f 2) | 18 | SHERPA_ONNX_VERSION=$(grep "SHERPA_ONNX_VERSION" ./CMakeLists.txt | cut -d " " -f 2 | cut -d '"' -f 2) |
| 19 | echo "SHERPA_ONNX_VERSION $SHERPA_ONNX_VERSION" | 19 | echo "SHERPA_ONNX_VERSION $SHERPA_ONNX_VERSION" |
| 20 | 20 | ||
| 21 | -# SHERPA_ONNX_VERSION=1.0.25 | 21 | +# SHERPA_ONNX_VERSION=1.0.27 |
| 22 | 22 | ||
| 23 | if [ -z $owner ]; then | 23 | if [ -z $owner ]; then |
| 24 | owner=k2-fsa | 24 | owner=k2-fsa |
| @@ -55,7 +55,7 @@ jobs: | @@ -55,7 +55,7 @@ jobs: | ||
| 55 | 55 | ||
| 56 | SHERPA_ONNX_VERSION=$(grep "SHERPA_ONNX_VERSION" ./CMakeLists.txt | cut -d " " -f 2 | cut -d '"' -f 2) | 56 | SHERPA_ONNX_VERSION=$(grep "SHERPA_ONNX_VERSION" ./CMakeLists.txt | cut -d " " -f 2 | cut -d '"' -f 2) |
| 57 | echo "SHERPA_ONNX_VERSION $SHERPA_ONNX_VERSION" | 57 | echo "SHERPA_ONNX_VERSION $SHERPA_ONNX_VERSION" |
| 58 | - # SHERPA_ONNX_VERSION=1.0.25 | 58 | + # SHERPA_ONNX_VERSION=1.0.27 |
| 59 | 59 | ||
| 60 | src_dir=.github/scripts/node-addon | 60 | src_dir=.github/scripts/node-addon |
| 61 | sed -i.bak s/SHERPA_ONNX_VERSION/$SHERPA_ONNX_VERSION/g $src_dir/package.json | 61 | sed -i.bak s/SHERPA_ONNX_VERSION/$SHERPA_ONNX_VERSION/g $src_dir/package.json |
| @@ -24,7 +24,12 @@ const tts = createOfflineTts(); | @@ -24,7 +24,12 @@ const tts = createOfflineTts(); | ||
| 24 | const text = 'Alles hat ein Ende, nur die Wurst hat zwei.' | 24 | const text = 'Alles hat ein Ende, nur die Wurst hat zwei.' |
| 25 | 25 | ||
| 26 | let start = Date.now(); | 26 | let start = Date.now(); |
| 27 | -const audio = tts.generate({text: text, sid: 0, speed: 1.0}); | 27 | +const audio = tts.generate({ |
| 28 | + text: text, | ||
| 29 | + sid: 0, | ||
| 30 | + speed: 1.0, | ||
| 31 | + enableExternalBuffer: true, | ||
| 32 | +}); | ||
| 28 | let stop = Date.now(); | 33 | let stop = Date.now(); |
| 29 | const elapsed_seconds = (stop - start) / 1000; | 34 | const elapsed_seconds = (stop - start) / 1000; |
| 30 | const duration = audio.samples.length / audio.sampleRate; | 35 | const duration = audio.samples.length / audio.sampleRate; |
| @@ -99,7 +99,7 @@ function do_check() { | @@ -99,7 +99,7 @@ function do_check() { | ||
| 99 | ;; | 99 | ;; |
| 100 | 2) | 100 | 2) |
| 101 | echo "Check all files" | 101 | echo "Check all files" |
| 102 | - files=$(find $sherpa_onnx_dir/sherpa-onnx -name "*.h" -o -name "*.cc") | 102 | + files=$(find $sherpa_onnx_dir/sherpa-onnx/csrc $sherpa_onnx_dir/sherpa-onnx/python $sherpa_onnx_dir/scripts/node-addon-api/src $sherpa_onnx_dir/sherpa-onnx/jni $sherpa_onnx_dir/sherpa-onnx/c-api -name "*.h" -o -name "*.cc") |
| 103 | ;; | 103 | ;; |
| 104 | *) | 104 | *) |
| 105 | echo "Check last commit" | 105 | echo "Check last commit" |
| @@ -18,9 +18,9 @@ class SpeakerEmbeddingExtractor { | @@ -18,9 +18,9 @@ class SpeakerEmbeddingExtractor { | ||
| 18 | } | 18 | } |
| 19 | 19 | ||
| 20 | // return a float32 array | 20 | // return a float32 array |
| 21 | - compute(stream) { | 21 | + compute(stream, enableExternalBuffer = true) { |
| 22 | return addon.speakerEmbeddingExtractorComputeEmbedding( | 22 | return addon.speakerEmbeddingExtractorComputeEmbedding( |
| 23 | - this.handle, stream.handle); | 23 | + this.handle, stream.handle, enableExternalBuffer); |
| 24 | } | 24 | } |
| 25 | } | 25 | } |
| 26 | 26 |
| @@ -11,8 +11,9 @@ class CircularBuffer { | @@ -11,8 +11,9 @@ class CircularBuffer { | ||
| 11 | } | 11 | } |
| 12 | 12 | ||
| 13 | // return a float32 array | 13 | // return a float32 array |
| 14 | - get(startIndex, n) { | ||
| 15 | - return addon.circularBufferGet(this.handle, startIndex, n); | 14 | + get(startIndex, n, enableExternalBuffer = true) { |
| 15 | + return addon.circularBufferGet( | ||
| 16 | + this.handle, startIndex, n, enableExternalBuffer); | ||
| 16 | } | 17 | } |
| 17 | 18 | ||
| 18 | pop(n) { | 19 | pop(n) { |
| @@ -48,23 +49,23 @@ config = { | @@ -48,23 +49,23 @@ config = { | ||
| 48 | } | 49 | } |
| 49 | 50 | ||
| 50 | acceptWaveform(samples) { | 51 | acceptWaveform(samples) { |
| 51 | - addon.voiceActivityDetectorAcceptWaveform(this.handle, samples) | 52 | + addon.voiceActivityDetectorAcceptWaveform(this.handle, samples); |
| 52 | } | 53 | } |
| 53 | 54 | ||
| 54 | isEmpty() { | 55 | isEmpty() { |
| 55 | - return addon.voiceActivityDetectorIsEmpty(this.handle) | 56 | + return addon.voiceActivityDetectorIsEmpty(this.handle); |
| 56 | } | 57 | } |
| 57 | 58 | ||
| 58 | isDetected() { | 59 | isDetected() { |
| 59 | - return addon.voiceActivityDetectorIsDetected(this.handle) | 60 | + return addon.voiceActivityDetectorIsDetected(this.handle); |
| 60 | } | 61 | } |
| 61 | 62 | ||
| 62 | pop() { | 63 | pop() { |
| 63 | - addon.voiceActivityDetectorPop(this.handle) | 64 | + addon.voiceActivityDetectorPop(this.handle); |
| 64 | } | 65 | } |
| 65 | 66 | ||
| 66 | clear() { | 67 | clear() { |
| 67 | - addon.VoiceActivityDetectorClearWrapper(this.handle) | 68 | + addon.VoiceActivityDetectorClearWrapper(this.handle); |
| 68 | } | 69 | } |
| 69 | 70 | ||
| 70 | /* | 71 | /* |
| @@ -73,12 +74,12 @@ config = { | @@ -73,12 +74,12 @@ config = { | ||
| 73 | start: a int32 | 74 | start: a int32 |
| 74 | } | 75 | } |
| 75 | */ | 76 | */ |
| 76 | - front() { | ||
| 77 | - return addon.voiceActivityDetectorFront(this.handle) | 77 | + front(enableExternalBuffer = true) { |
| 78 | + return addon.voiceActivityDetectorFront(this.handle, enableExternalBuffer); | ||
| 78 | } | 79 | } |
| 79 | 80 | ||
| 80 | reset() { | 81 | reset() { |
| 81 | - return addon.VoiceActivityDetectorResetWrapper(this.handle) | 82 | + return addon.VoiceActivityDetectorResetWrapper(this.handle); |
| 82 | } | 83 | } |
| 83 | } | 84 | } |
| 84 | 85 |
| @@ -2,6 +2,7 @@ | @@ -2,6 +2,7 @@ | ||
| 2 | // | 2 | // |
| 3 | // Copyright (c) 2024 Xiaomi Corporation | 3 | // Copyright (c) 2024 Xiaomi Corporation |
| 4 | 4 | ||
| 5 | +#include <algorithm> | ||
| 5 | #include <sstream> | 6 | #include <sstream> |
| 6 | 7 | ||
| 7 | #include "macros.h" // NOLINT | 8 | #include "macros.h" // NOLINT |
| @@ -265,6 +266,13 @@ static Napi::Object OfflineTtsGenerateWrapper(const Napi::CallbackInfo &info) { | @@ -265,6 +266,13 @@ static Napi::Object OfflineTtsGenerateWrapper(const Napi::CallbackInfo &info) { | ||
| 265 | return {}; | 266 | return {}; |
| 266 | } | 267 | } |
| 267 | 268 | ||
| 269 | + bool enable_external_buffer = true; | ||
| 270 | + if (obj.Has("enableExternalBuffer") && | ||
| 271 | + obj.Get("enableExternalBuffer").IsBoolean()) { | ||
| 272 | + enable_external_buffer = | ||
| 273 | + obj.Get("enableExternalBuffer").As<Napi::Boolean>().Value(); | ||
| 274 | + } | ||
| 275 | + | ||
| 268 | Napi::String _text = obj.Get("text").As<Napi::String>(); | 276 | Napi::String _text = obj.Get("text").As<Napi::String>(); |
| 269 | std::string text = _text.Utf8Value(); | 277 | std::string text = _text.Utf8Value(); |
| 270 | int32_t sid = obj.Get("sid").As<Napi::Number>().Int32Value(); | 278 | int32_t sid = obj.Get("sid").As<Napi::Number>().Int32Value(); |
| @@ -273,20 +281,37 @@ static Napi::Object OfflineTtsGenerateWrapper(const Napi::CallbackInfo &info) { | @@ -273,20 +281,37 @@ static Napi::Object OfflineTtsGenerateWrapper(const Napi::CallbackInfo &info) { | ||
| 273 | const SherpaOnnxGeneratedAudio *audio = | 281 | const SherpaOnnxGeneratedAudio *audio = |
| 274 | SherpaOnnxOfflineTtsGenerate(tts, text.c_str(), sid, speed); | 282 | SherpaOnnxOfflineTtsGenerate(tts, text.c_str(), sid, speed); |
| 275 | 283 | ||
| 276 | - Napi::ArrayBuffer arrayBuffer = Napi::ArrayBuffer::New( | ||
| 277 | - env, const_cast<float *>(audio->samples), sizeof(float) * audio->n, | ||
| 278 | - [](Napi::Env /*env*/, void * /*data*/, | ||
| 279 | - const SherpaOnnxGeneratedAudio *hint) { | ||
| 280 | - SherpaOnnxDestroyOfflineTtsGeneratedAudio(hint); | ||
| 281 | - }, | ||
| 282 | - audio); | ||
| 283 | - Napi::Float32Array float32Array = | ||
| 284 | - Napi::Float32Array::New(env, audio->n, arrayBuffer, 0); | ||
| 285 | - | ||
| 286 | - Napi::Object ans = Napi::Object::New(env); | ||
| 287 | - ans.Set(Napi::String::New(env, "samples"), float32Array); | ||
| 288 | - ans.Set(Napi::String::New(env, "sampleRate"), audio->sample_rate); | ||
| 289 | - return ans; | 284 | + if (enable_external_buffer) { |
| 285 | + Napi::ArrayBuffer arrayBuffer = Napi::ArrayBuffer::New( | ||
| 286 | + env, const_cast<float *>(audio->samples), sizeof(float) * audio->n, | ||
| 287 | + [](Napi::Env /*env*/, void * /*data*/, | ||
| 288 | + const SherpaOnnxGeneratedAudio *hint) { | ||
| 289 | + SherpaOnnxDestroyOfflineTtsGeneratedAudio(hint); | ||
| 290 | + }, | ||
| 291 | + audio); | ||
| 292 | + Napi::Float32Array float32Array = | ||
| 293 | + Napi::Float32Array::New(env, audio->n, arrayBuffer, 0); | ||
| 294 | + | ||
| 295 | + Napi::Object ans = Napi::Object::New(env); | ||
| 296 | + ans.Set(Napi::String::New(env, "samples"), float32Array); | ||
| 297 | + ans.Set(Napi::String::New(env, "sampleRate"), audio->sample_rate); | ||
| 298 | + return ans; | ||
| 299 | + } else { | ||
| 300 | + // don't use external buffer | ||
| 301 | + Napi::ArrayBuffer arrayBuffer = | ||
| 302 | + Napi::ArrayBuffer::New(env, sizeof(float) * audio->n); | ||
| 303 | + | ||
| 304 | + Napi::Float32Array float32Array = | ||
| 305 | + Napi::Float32Array::New(env, audio->n, arrayBuffer, 0); | ||
| 306 | + | ||
| 307 | + std::copy(audio->samples, audio->samples + audio->n, float32Array.Data()); | ||
| 308 | + | ||
| 309 | + Napi::Object ans = Napi::Object::New(env); | ||
| 310 | + ans.Set(Napi::String::New(env, "samples"), float32Array); | ||
| 311 | + ans.Set(Napi::String::New(env, "sampleRate"), audio->sample_rate); | ||
| 312 | + SherpaOnnxDestroyOfflineTtsGeneratedAudio(audio); | ||
| 313 | + return ans; | ||
| 314 | + } | ||
| 290 | } | 315 | } |
| 291 | 316 | ||
| 292 | void InitNonStreamingTts(Napi::Env env, Napi::Object exports) { | 317 | void InitNonStreamingTts(Napi::Env env, Napi::Object exports) { |
| 1 | // scripts/node-addon-api/src/speaker-identification.cc | 1 | // scripts/node-addon-api/src/speaker-identification.cc |
| 2 | // | 2 | // |
| 3 | // Copyright (c) 2024 Xiaomi Corporation | 3 | // Copyright (c) 2024 Xiaomi Corporation |
| 4 | +#include <algorithm> | ||
| 4 | #include <sstream> | 5 | #include <sstream> |
| 5 | 6 | ||
| 6 | #include "macros.h" // NOLINT | 7 | #include "macros.h" // NOLINT |
| @@ -175,9 +176,9 @@ static Napi::Boolean SpeakerEmbeddingExtractorIsReadyWrapper( | @@ -175,9 +176,9 @@ static Napi::Boolean SpeakerEmbeddingExtractorIsReadyWrapper( | ||
| 175 | static Napi::Float32Array SpeakerEmbeddingExtractorComputeEmbeddingWrapper( | 176 | static Napi::Float32Array SpeakerEmbeddingExtractorComputeEmbeddingWrapper( |
| 176 | const Napi::CallbackInfo &info) { | 177 | const Napi::CallbackInfo &info) { |
| 177 | Napi::Env env = info.Env(); | 178 | Napi::Env env = info.Env(); |
| 178 | - if (info.Length() != 2) { | 179 | + if (info.Length() != 2 && info.Length() != 3) { |
| 179 | std::ostringstream os; | 180 | std::ostringstream os; |
| 180 | - os << "Expect only 2 arguments. Given: " << info.Length(); | 181 | + os << "Expect only 2 or 3 arguments. Given: " << info.Length(); |
| 181 | 182 | ||
| 182 | Napi::TypeError::New(env, os.str()).ThrowAsJavaScriptException(); | 183 | Napi::TypeError::New(env, os.str()).ThrowAsJavaScriptException(); |
| 183 | 184 | ||
| @@ -199,6 +200,16 @@ static Napi::Float32Array SpeakerEmbeddingExtractorComputeEmbeddingWrapper( | @@ -199,6 +200,16 @@ static Napi::Float32Array SpeakerEmbeddingExtractorComputeEmbeddingWrapper( | ||
| 199 | return {}; | 200 | return {}; |
| 200 | } | 201 | } |
| 201 | 202 | ||
| 203 | + bool enable_external_buffer = true; | ||
| 204 | + if (info.Length() == 3) { | ||
| 205 | + if (info[2].IsBoolean()) { | ||
| 206 | + enable_external_buffer = info[2].As<Napi::Boolean>().Value(); | ||
| 207 | + } else { | ||
| 208 | + Napi::TypeError::New(env, "Argument 2 should be a boolean.") | ||
| 209 | + .ThrowAsJavaScriptException(); | ||
| 210 | + } | ||
| 211 | + } | ||
| 212 | + | ||
| 202 | SherpaOnnxSpeakerEmbeddingExtractor *extractor = | 213 | SherpaOnnxSpeakerEmbeddingExtractor *extractor = |
| 203 | info[0].As<Napi::External<SherpaOnnxSpeakerEmbeddingExtractor>>().Data(); | 214 | info[0].As<Napi::External<SherpaOnnxSpeakerEmbeddingExtractor>>().Data(); |
| 204 | 215 | ||
| @@ -210,14 +221,29 @@ static Napi::Float32Array SpeakerEmbeddingExtractorComputeEmbeddingWrapper( | @@ -210,14 +221,29 @@ static Napi::Float32Array SpeakerEmbeddingExtractorComputeEmbeddingWrapper( | ||
| 210 | 221 | ||
| 211 | int32_t dim = SherpaOnnxSpeakerEmbeddingExtractorDim(extractor); | 222 | int32_t dim = SherpaOnnxSpeakerEmbeddingExtractorDim(extractor); |
| 212 | 223 | ||
| 213 | - Napi::ArrayBuffer arrayBuffer = Napi::ArrayBuffer::New( | ||
| 214 | - env, const_cast<float *>(v), sizeof(float) * dim, | ||
| 215 | - [](Napi::Env /*env*/, void *data) { | ||
| 216 | - SherpaOnnxSpeakerEmbeddingExtractorDestroyEmbedding( | ||
| 217 | - reinterpret_cast<float *>(data)); | ||
| 218 | - }); | 224 | + if (enable_external_buffer) { |
| 225 | + Napi::ArrayBuffer arrayBuffer = Napi::ArrayBuffer::New( | ||
| 226 | + env, const_cast<float *>(v), sizeof(float) * dim, | ||
| 227 | + [](Napi::Env /*env*/, void *data) { | ||
| 228 | + SherpaOnnxSpeakerEmbeddingExtractorDestroyEmbedding( | ||
| 229 | + reinterpret_cast<float *>(data)); | ||
| 230 | + }); | ||
| 231 | + | ||
| 232 | + return Napi::Float32Array::New(env, dim, arrayBuffer, 0); | ||
| 233 | + } else { | ||
| 234 | + // don't use external buffer | ||
| 235 | + Napi::ArrayBuffer arrayBuffer = | ||
| 236 | + Napi::ArrayBuffer::New(env, sizeof(float) * dim); | ||
| 237 | + | ||
| 238 | + Napi::Float32Array float32Array = | ||
| 239 | + Napi::Float32Array::New(env, dim, arrayBuffer, 0); | ||
| 219 | 240 | ||
| 220 | - return Napi::Float32Array::New(env, dim, arrayBuffer, 0); | 241 | + std::copy(v, v + dim, float32Array.Data()); |
| 242 | + | ||
| 243 | + SherpaOnnxSpeakerEmbeddingExtractorDestroyEmbedding(v); | ||
| 244 | + | ||
| 245 | + return float32Array; | ||
| 246 | + } | ||
| 221 | } | 247 | } |
| 222 | 248 | ||
| 223 | static Napi::External<SherpaOnnxSpeakerEmbeddingManager> | 249 | static Napi::External<SherpaOnnxSpeakerEmbeddingManager> |
| @@ -2,6 +2,7 @@ | @@ -2,6 +2,7 @@ | ||
| 2 | // | 2 | // |
| 3 | // Copyright (c) 2024 Xiaomi Corporation | 3 | // Copyright (c) 2024 Xiaomi Corporation |
| 4 | 4 | ||
| 5 | +#include <algorithm> | ||
| 5 | #include <sstream> | 6 | #include <sstream> |
| 6 | 7 | ||
| 7 | #include "macros.h" // NOLINT | 8 | #include "macros.h" // NOLINT |
| @@ -75,9 +76,9 @@ static Napi::Float32Array CircularBufferGetWrapper( | @@ -75,9 +76,9 @@ static Napi::Float32Array CircularBufferGetWrapper( | ||
| 75 | const Napi::CallbackInfo &info) { | 76 | const Napi::CallbackInfo &info) { |
| 76 | Napi::Env env = info.Env(); | 77 | Napi::Env env = info.Env(); |
| 77 | 78 | ||
| 78 | - if (info.Length() != 3) { | 79 | + if (info.Length() != 3 && info.Length() != 4) { |
| 79 | std::ostringstream os; | 80 | std::ostringstream os; |
| 80 | - os << "Expect only 3 arguments. Given: " << info.Length(); | 81 | + os << "Expect only 3 or 4 arguments. Given: " << info.Length(); |
| 81 | 82 | ||
| 82 | Napi::TypeError::New(env, os.str()).ThrowAsJavaScriptException(); | 83 | Napi::TypeError::New(env, os.str()).ThrowAsJavaScriptException(); |
| 83 | 84 | ||
| @@ -108,21 +109,46 @@ static Napi::Float32Array CircularBufferGetWrapper( | @@ -108,21 +109,46 @@ static Napi::Float32Array CircularBufferGetWrapper( | ||
| 108 | return {}; | 109 | return {}; |
| 109 | } | 110 | } |
| 110 | 111 | ||
| 112 | + bool enable_external_buffer = true; | ||
| 113 | + if (info.Length() == 4) { | ||
| 114 | + if (info[3].IsBoolean()) { | ||
| 115 | + enable_external_buffer = info[3].As<Napi::Boolean>().Value(); | ||
| 116 | + } else { | ||
| 117 | + Napi::TypeError::New(env, "Argument 3 should be a boolean.") | ||
| 118 | + .ThrowAsJavaScriptException(); | ||
| 119 | + } | ||
| 120 | + } | ||
| 121 | + | ||
| 111 | int32_t start_index = info[1].As<Napi::Number>().Int32Value(); | 122 | int32_t start_index = info[1].As<Napi::Number>().Int32Value(); |
| 112 | int32_t n = info[2].As<Napi::Number>().Int32Value(); | 123 | int32_t n = info[2].As<Napi::Number>().Int32Value(); |
| 113 | 124 | ||
| 114 | const float *data = SherpaOnnxCircularBufferGet(buf, start_index, n); | 125 | const float *data = SherpaOnnxCircularBufferGet(buf, start_index, n); |
| 115 | 126 | ||
| 116 | - Napi::ArrayBuffer arrayBuffer = Napi::ArrayBuffer::New( | ||
| 117 | - env, const_cast<float *>(data), sizeof(float) * n, | ||
| 118 | - [](Napi::Env /*env*/, void *p) { | ||
| 119 | - SherpaOnnxCircularBufferFree(reinterpret_cast<const float *>(p)); | ||
| 120 | - }); | 127 | + if (enable_external_buffer) { |
| 128 | + Napi::ArrayBuffer arrayBuffer = Napi::ArrayBuffer::New( | ||
| 129 | + env, const_cast<float *>(data), sizeof(float) * n, | ||
| 130 | + [](Napi::Env /*env*/, void *p) { | ||
| 131 | + SherpaOnnxCircularBufferFree(reinterpret_cast<const float *>(p)); | ||
| 132 | + }); | ||
| 133 | + | ||
| 134 | + Napi::Float32Array float32Array = | ||
| 135 | + Napi::Float32Array::New(env, n, arrayBuffer, 0); | ||
| 121 | 136 | ||
| 122 | - Napi::Float32Array float32Array = | ||
| 123 | - Napi::Float32Array::New(env, n, arrayBuffer, 0); | 137 | + return float32Array; |
| 138 | + } else { | ||
| 139 | + // don't use external buffer | ||
| 140 | + Napi::ArrayBuffer arrayBuffer = Napi::ArrayBuffer::New( | ||
| 141 | + env, const_cast<float *>(data), sizeof(float) * n); | ||
| 124 | 142 | ||
| 125 | - return float32Array; | 143 | + Napi::Float32Array float32Array = |
| 144 | + Napi::Float32Array::New(env, n, arrayBuffer, 0); | ||
| 145 | + | ||
| 146 | + std::copy(data, data + n, float32Array.Data()); | ||
| 147 | + | ||
| 148 | + SherpaOnnxCircularBufferFree(data); | ||
| 149 | + | ||
| 150 | + return float32Array; | ||
| 151 | + } | ||
| 126 | } | 152 | } |
| 127 | 153 | ||
| 128 | static void CircularBufferPopWrapper(const Napi::CallbackInfo &info) { | 154 | static void CircularBufferPopWrapper(const Napi::CallbackInfo &info) { |
| @@ -470,9 +496,9 @@ static Napi::Object VoiceActivityDetectorFrontWrapper( | @@ -470,9 +496,9 @@ static Napi::Object VoiceActivityDetectorFrontWrapper( | ||
| 470 | const Napi::CallbackInfo &info) { | 496 | const Napi::CallbackInfo &info) { |
| 471 | Napi::Env env = info.Env(); | 497 | Napi::Env env = info.Env(); |
| 472 | 498 | ||
| 473 | - if (info.Length() != 1) { | 499 | + if (info.Length() != 1 && info.Length() != 2) { |
| 474 | std::ostringstream os; | 500 | std::ostringstream os; |
| 475 | - os << "Expect only 1 argument. Given: " << info.Length(); | 501 | + os << "Expect only 1 or 2 arguments. Given: " << info.Length(); |
| 476 | 502 | ||
| 477 | Napi::TypeError::New(env, os.str()).ThrowAsJavaScriptException(); | 503 | Napi::TypeError::New(env, os.str()).ThrowAsJavaScriptException(); |
| 478 | 504 | ||
| @@ -486,28 +512,57 @@ static Napi::Object VoiceActivityDetectorFrontWrapper( | @@ -486,28 +512,57 @@ static Napi::Object VoiceActivityDetectorFrontWrapper( | ||
| 486 | return {}; | 512 | return {}; |
| 487 | } | 513 | } |
| 488 | 514 | ||
| 515 | + bool enable_external_buffer = true; | ||
| 516 | + if (info.Length() == 2) { | ||
| 517 | + if (info[1].IsBoolean()) { | ||
| 518 | + enable_external_buffer = info[1].As<Napi::Boolean>().Value(); | ||
| 519 | + } else { | ||
| 520 | + Napi::TypeError::New(env, "Argument 1 should be a boolean.") | ||
| 521 | + .ThrowAsJavaScriptException(); | ||
| 522 | + } | ||
| 523 | + } | ||
| 524 | + | ||
| 489 | SherpaOnnxVoiceActivityDetector *vad = | 525 | SherpaOnnxVoiceActivityDetector *vad = |
| 490 | info[0].As<Napi::External<SherpaOnnxVoiceActivityDetector>>().Data(); | 526 | info[0].As<Napi::External<SherpaOnnxVoiceActivityDetector>>().Data(); |
| 491 | 527 | ||
| 492 | const SherpaOnnxSpeechSegment *segment = | 528 | const SherpaOnnxSpeechSegment *segment = |
| 493 | SherpaOnnxVoiceActivityDetectorFront(vad); | 529 | SherpaOnnxVoiceActivityDetectorFront(vad); |
| 494 | 530 | ||
| 495 | - Napi::ArrayBuffer arrayBuffer = Napi::ArrayBuffer::New( | ||
| 496 | - env, const_cast<float *>(segment->samples), sizeof(float) * segment->n, | ||
| 497 | - [](Napi::Env /*env*/, void * /*data*/, | ||
| 498 | - const SherpaOnnxSpeechSegment *hint) { | ||
| 499 | - SherpaOnnxDestroySpeechSegment(hint); | ||
| 500 | - }, | ||
| 501 | - segment); | 531 | + if (enable_external_buffer) { |
| 532 | + Napi::ArrayBuffer arrayBuffer = Napi::ArrayBuffer::New( | ||
| 533 | + env, const_cast<float *>(segment->samples), sizeof(float) * segment->n, | ||
| 534 | + [](Napi::Env /*env*/, void * /*data*/, | ||
| 535 | + const SherpaOnnxSpeechSegment *hint) { | ||
| 536 | + SherpaOnnxDestroySpeechSegment(hint); | ||
| 537 | + }, | ||
| 538 | + segment); | ||
| 539 | + | ||
| 540 | + Napi::Float32Array float32Array = | ||
| 541 | + Napi::Float32Array::New(env, segment->n, arrayBuffer, 0); | ||
| 502 | 542 | ||
| 503 | - Napi::Float32Array float32Array = | ||
| 504 | - Napi::Float32Array::New(env, segment->n, arrayBuffer, 0); | 543 | + Napi::Object obj = Napi::Object::New(env); |
| 544 | + obj.Set(Napi::String::New(env, "start"), segment->start); | ||
| 545 | + obj.Set(Napi::String::New(env, "samples"), float32Array); | ||
| 505 | 546 | ||
| 506 | - Napi::Object obj = Napi::Object::New(env); | ||
| 507 | - obj.Set(Napi::String::New(env, "start"), segment->start); | ||
| 508 | - obj.Set(Napi::String::New(env, "samples"), float32Array); | 547 | + return obj; |
| 548 | + } else { | ||
| 549 | + Napi::ArrayBuffer arrayBuffer = | ||
| 550 | + Napi::ArrayBuffer::New(env, sizeof(float) * segment->n); | ||
| 509 | 551 | ||
| 510 | - return obj; | 552 | + Napi::Float32Array float32Array = |
| 553 | + Napi::Float32Array::New(env, segment->n, arrayBuffer, 0); | ||
| 554 | + | ||
| 555 | + std::copy(segment->samples, segment->samples + segment->n, | ||
| 556 | + float32Array.Data()); | ||
| 557 | + | ||
| 558 | + Napi::Object obj = Napi::Object::New(env); | ||
| 559 | + obj.Set(Napi::String::New(env, "start"), segment->start); | ||
| 560 | + obj.Set(Napi::String::New(env, "samples"), float32Array); | ||
| 561 | + | ||
| 562 | + SherpaOnnxDestroySpeechSegment(segment); | ||
| 563 | + | ||
| 564 | + return obj; | ||
| 565 | + } | ||
| 511 | } | 566 | } |
| 512 | 567 | ||
| 513 | static void VoiceActivityDetectorResetWrapper(const Napi::CallbackInfo &info) { | 568 | static void VoiceActivityDetectorResetWrapper(const Napi::CallbackInfo &info) { |
| @@ -2,6 +2,7 @@ | @@ -2,6 +2,7 @@ | ||
| 2 | // | 2 | // |
| 3 | // Copyright (c) 2024 Xiaomi Corporation | 3 | // Copyright (c) 2024 Xiaomi Corporation |
| 4 | 4 | ||
| 5 | +#include <algorithm> | ||
| 5 | #include <sstream> | 6 | #include <sstream> |
| 6 | 7 | ||
| 7 | #include "napi.h" // NOLINT | 8 | #include "napi.h" // NOLINT |
| @@ -9,16 +10,17 @@ | @@ -9,16 +10,17 @@ | ||
| 9 | 10 | ||
| 10 | static Napi::Object ReadWaveWrapper(const Napi::CallbackInfo &info) { | 11 | static Napi::Object ReadWaveWrapper(const Napi::CallbackInfo &info) { |
| 11 | Napi::Env env = info.Env(); | 12 | Napi::Env env = info.Env(); |
| 12 | - if (info.Length() != 1) { | 13 | + if (info.Length() > 2) { |
| 13 | std::ostringstream os; | 14 | std::ostringstream os; |
| 14 | - os << "Expect only 1 argument. Given: " << info.Length(); | 15 | + os << "Expect only 2 arguments. Given: " << info.Length(); |
| 15 | 16 | ||
| 16 | Napi::TypeError::New(env, os.str()).ThrowAsJavaScriptException(); | 17 | Napi::TypeError::New(env, os.str()).ThrowAsJavaScriptException(); |
| 17 | 18 | ||
| 18 | return {}; | 19 | return {}; |
| 19 | } | 20 | } |
| 21 | + | ||
| 20 | if (!info[0].IsString()) { | 22 | if (!info[0].IsString()) { |
| 21 | - Napi::TypeError::New(env, "Argument should be a string") | 23 | + Napi::TypeError::New(env, "Argument 0 should be a string") |
| 22 | .ThrowAsJavaScriptException(); | 24 | .ThrowAsJavaScriptException(); |
| 23 | 25 | ||
| 24 | return {}; | 26 | return {}; |
| @@ -26,6 +28,18 @@ static Napi::Object ReadWaveWrapper(const Napi::CallbackInfo &info) { | @@ -26,6 +28,18 @@ static Napi::Object ReadWaveWrapper(const Napi::CallbackInfo &info) { | ||
| 26 | 28 | ||
| 27 | std::string filename = info[0].As<Napi::String>().Utf8Value(); | 29 | std::string filename = info[0].As<Napi::String>().Utf8Value(); |
| 28 | 30 | ||
| 31 | + bool enable_external_buffer = true; | ||
| 32 | + if (info.Length() == 2) { | ||
| 33 | + if (info[1].IsBoolean()) { | ||
| 34 | + enable_external_buffer = info[1].As<Napi::Boolean>().Value(); | ||
| 35 | + } else { | ||
| 36 | + Napi::TypeError::New(env, "Argument 1 should be a boolean") | ||
| 37 | + .ThrowAsJavaScriptException(); | ||
| 38 | + | ||
| 39 | + return {}; | ||
| 40 | + } | ||
| 41 | + } | ||
| 42 | + | ||
| 29 | const SherpaOnnxWave *wave = SherpaOnnxReadWave(filename.c_str()); | 43 | const SherpaOnnxWave *wave = SherpaOnnxReadWave(filename.c_str()); |
| 30 | if (!wave) { | 44 | if (!wave) { |
| 31 | std::ostringstream os; | 45 | std::ostringstream os; |
| @@ -35,20 +49,40 @@ static Napi::Object ReadWaveWrapper(const Napi::CallbackInfo &info) { | @@ -35,20 +49,40 @@ static Napi::Object ReadWaveWrapper(const Napi::CallbackInfo &info) { | ||
| 35 | return {}; | 49 | return {}; |
| 36 | } | 50 | } |
| 37 | 51 | ||
| 38 | - Napi::ArrayBuffer arrayBuffer = Napi::ArrayBuffer::New( | ||
| 39 | - env, const_cast<float *>(wave->samples), | ||
| 40 | - sizeof(float) * wave->num_samples, | ||
| 41 | - [](Napi::Env /*env*/, void * /*data*/, const SherpaOnnxWave *hint) { | ||
| 42 | - SherpaOnnxFreeWave(hint); | ||
| 43 | - }, | ||
| 44 | - wave); | ||
| 45 | - Napi::Float32Array float32Array = | ||
| 46 | - Napi::Float32Array::New(env, wave->num_samples, arrayBuffer, 0); | ||
| 47 | - | ||
| 48 | - Napi::Object obj = Napi::Object::New(env); | ||
| 49 | - obj.Set(Napi::String::New(env, "samples"), float32Array); | ||
| 50 | - obj.Set(Napi::String::New(env, "sampleRate"), wave->sample_rate); | ||
| 51 | - return obj; | 52 | + if (enable_external_buffer) { |
| 53 | + Napi::ArrayBuffer arrayBuffer = Napi::ArrayBuffer::New( | ||
| 54 | + env, const_cast<float *>(wave->samples), | ||
| 55 | + sizeof(float) * wave->num_samples, | ||
| 56 | + [](Napi::Env /*env*/, void * /*data*/, const SherpaOnnxWave *hint) { | ||
| 57 | + SherpaOnnxFreeWave(hint); | ||
| 58 | + }, | ||
| 59 | + wave); | ||
| 60 | + Napi::Float32Array float32Array = | ||
| 61 | + Napi::Float32Array::New(env, wave->num_samples, arrayBuffer, 0); | ||
| 62 | + | ||
| 63 | + Napi::Object obj = Napi::Object::New(env); | ||
| 64 | + obj.Set(Napi::String::New(env, "samples"), float32Array); | ||
| 65 | + obj.Set(Napi::String::New(env, "sampleRate"), wave->sample_rate); | ||
| 66 | + return obj; | ||
| 67 | + } else { | ||
| 68 | + // don't use external buffer | ||
| 69 | + Napi::ArrayBuffer arrayBuffer = | ||
| 70 | + Napi::ArrayBuffer::New(env, sizeof(float) * wave->num_samples); | ||
| 71 | + | ||
| 72 | + Napi::Float32Array float32Array = | ||
| 73 | + Napi::Float32Array::New(env, wave->num_samples, arrayBuffer, 0); | ||
| 74 | + | ||
| 75 | + std::copy(wave->samples, wave->samples + wave->num_samples, | ||
| 76 | + float32Array.Data()); | ||
| 77 | + | ||
| 78 | + Napi::Object obj = Napi::Object::New(env); | ||
| 79 | + obj.Set(Napi::String::New(env, "samples"), float32Array); | ||
| 80 | + obj.Set(Napi::String::New(env, "sampleRate"), wave->sample_rate); | ||
| 81 | + | ||
| 82 | + SherpaOnnxFreeWave(wave); | ||
| 83 | + | ||
| 84 | + return obj; | ||
| 85 | + } | ||
| 52 | } | 86 | } |
| 53 | 87 | ||
| 54 | void InitWaveReader(Napi::Env env, Napi::Object exports) { | 88 | void InitWaveReader(Napi::Env env, Napi::Object exports) { |
-
请 注册 或 登录 后发表评论