EngineX-Iluvatar/enginex-mr_series-sherpa-onnx
Archived
16
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity

Add Java API for non-streaming ASR (#807)

Browse Source
This commit is contained in:
Fangjun Kuang
2024-04-24 21:03:26 +08:00
committed by GitHub
parent c3a2e8a67c
commit 83cd533f67
24 changed files with 834 additions and 27 deletions
Download Patch File Download Diff File Expand all files Collapse all files

117
sherpa-onnx/java-api/src/com/k2fsa/sherpa/onnx/OfflineModelConfig.java Normal file
View File

@@ -0,0 +1,117 @@
// Copyright 2024 Xiaomi Corporation
package com.k2fsa.sherpa.onnx;
public class OfflineModelConfig {
private final OfflineTransducerModelConfig transducer;
private final OfflineParaformerModelConfig paraformer;
private final OfflineWhisperModelConfig whisper;
private final String tokens;
private final int numThreads;
private final boolean debug;
private final String provider;
private final String modelType;
private OfflineModelConfig(Builder builder) {
this.transducer = builder.transducer;
this.paraformer = builder.paraformer;
this.whisper = builder.whisper;
this.tokens = builder.tokens;
this.numThreads = builder.numThreads;
this.debug = builder.debug;
this.provider = builder.provider;
this.modelType = builder.modelType;
}
public static Builder builder() {
return new Builder();
}
public OfflineParaformerModelConfig getParaformer() {
return paraformer;
}
public OfflineTransducerModelConfig getTransducer() {
return transducer;
}
public OfflineWhisperModelConfig getZipformer2Ctc() {
return whisper;
}
public String getTokens() {
return tokens;
}
public int getNumThreads() {
return numThreads;
}
public boolean getDebug() {
return debug;
}
public String getProvider() {
return provider;
}
public String getModelType() {
return modelType;
}
public static class Builder {
private OfflineParaformerModelConfig paraformer = OfflineParaformerModelConfig.builder().build();
private OfflineTransducerModelConfig transducer = OfflineTransducerModelConfig.builder().build();
private OfflineWhisperModelConfig whisper = OfflineWhisperModelConfig.builder().build();
private String tokens = "";
private int numThreads = 1;
private boolean debug = true;
private String provider = "cpu";
private String modelType = "";
public OfflineModelConfig build() {
return new OfflineModelConfig(this);
}
public Builder setTransducer(OfflineTransducerModelConfig transducer) {
this.transducer = transducer;
return this;
}
public Builder setParaformer(OfflineParaformerModelConfig paraformer) {
this.paraformer = paraformer;
return this;
}
public Builder setWhisper(OfflineWhisperModelConfig whisper) {
this.whisper = whisper;
return this;
}
public Builder setTokens(String tokens) {
this.tokens = tokens;
return this;
}
public Builder setNumThreads(int numThreads) {
this.numThreads = numThreads;
return this;
}
public Builder setDebug(boolean debug) {
this.debug = debug;
return this;
}
public Builder setProvider(String provider) {
this.provider = provider;
return this;
}
public Builder setModelType(String modelType) {
this.modelType = modelType;
return this;
}
}
}

32
sherpa-onnx/java-api/src/com/k2fsa/sherpa/onnx/OfflineParaformerModelConfig.java Normal file
View File

@@ -0,0 +1,32 @@
// Copyright 2024 Xiaomi Corporation
package com.k2fsa.sherpa.onnx;
public class OfflineParaformerModelConfig {
private final String model;
private OfflineParaformerModelConfig(Builder builder) {
this.model = builder.model;
}
public static Builder builder() {
return new Builder();
}
public String getModel() {
return model;
}
public static class Builder {
private String model = "";
public OfflineParaformerModelConfig build() {
return new OfflineParaformerModelConfig(this);
}
public Builder setModel(String model) {
this.model = model;
return this;
}
}
}

55
sherpa-onnx/java-api/src/com/k2fsa/sherpa/onnx/OfflineRecognizer.java Normal file
View File

@@ -0,0 +1,55 @@
// Copyright 2024 Xiaomi Corporation
package com.k2fsa.sherpa.onnx;
public class OfflineRecognizer {
static {
System.loadLibrary("sherpa-onnx-jni");
}
private long ptr = 0; // this is the asr engine ptrss
public OfflineRecognizer(OfflineRecognizerConfig config) {
ptr = newFromFile(config);
}
public void decode(OfflineStream s) {
decode(ptr, s.getPtr());
}
public OfflineStream createStream() {
long p = createStream(ptr);
return new OfflineStream(p);
}
@Override
protected void finalize() throws Throwable {
release();
}
// You'd better call it manually if it is not used anymore
public void release() {
if (this.ptr == 0) {
return;
}
delete(this.ptr);
this.ptr = 0;
}
public OfflineRecognizerResult getResult(OfflineStream s) {
Object[] arr = getResult(s.getPtr());
String text = (String) arr[0];
String[] tokens = (String[]) arr[1];
float[] timestamps = (float[]) arr[2];
return new OfflineRecognizerResult(text, tokens, timestamps);
}
private native void delete(long ptr);
private native long newFromFile(OfflineRecognizerConfig config);
private native long createStream(long ptr);
private native void decode(long ptr, long streamPtr);
private native Object[] getResult(long streamPtr);
}

71
sherpa-onnx/java-api/src/com/k2fsa/sherpa/onnx/OfflineRecognizerConfig.java Normal file
View File

@@ -0,0 +1,71 @@
// Copyright 2024 Xiaomi Corporation
package com.k2fsa.sherpa.onnx;
public class OfflineRecognizerConfig {
private final FeatureConfig featConfig;
private final OfflineModelConfig modelConfig;
private final String decodingMethod;
private final int maxActivePaths;
private final String hotwordsFile;
private final float hotwordsScore;
private OfflineRecognizerConfig(Builder builder) {
this.featConfig = builder.featConfig;
this.modelConfig = builder.modelConfig;
this.decodingMethod = builder.decodingMethod;
this.maxActivePaths = builder.maxActivePaths;
this.hotwordsFile = builder.hotwordsFile;
this.hotwordsScore = builder.hotwordsScore;
}
public static Builder builder() {
return new Builder();
}
public OfflineModelConfig getModelConfig() {
return modelConfig;
}
public static class Builder {
private FeatureConfig featConfig = FeatureConfig.builder().build();
private OfflineModelConfig modelConfig = OfflineModelConfig.builder().build();
private String decodingMethod = "greedy_search";
private int maxActivePaths = 4;
private String hotwordsFile = "";
private float hotwordsScore = 1.5f;
public OfflineRecognizerConfig build() {
return new OfflineRecognizerConfig(this);
}
public Builder setFeatureConfig(FeatureConfig featConfig) {
this.featConfig = featConfig;
return this;
}
public Builder setOfflineModelConfig(OfflineModelConfig modelConfig) {
this.modelConfig = modelConfig;
return this;
}
public Builder setDecodingMethod(String decodingMethod) {
this.decodingMethod = decodingMethod;
return this;
}
public Builder setMaxActivePaths(int maxActivePaths) {
this.maxActivePaths = maxActivePaths;
return this;
}
public Builder setHotwordsFile(String hotwordsFile) {
this.hotwordsFile = hotwordsFile;
return this;
}
public Builder setHotwordsScore(float hotwordsScore) {
this.hotwordsScore = hotwordsScore;
return this;
}
}
}

26
sherpa-onnx/java-api/src/com/k2fsa/sherpa/onnx/OfflineRecognizerResult.java Normal file
View File

@@ -0,0 +1,26 @@
// Copyright 2024 Xiaomi Corporation
package com.k2fsa.sherpa.onnx;
public class OfflineRecognizerResult {
private final String text;
private final String[] tokens;
private final float[] timestamps;
public OfflineRecognizerResult(String text, String[] tokens, float[] timestamps) {
this.text = text;
this.tokens = tokens;
this.timestamps = timestamps;
}
public String getText() {
return text;
}
public String[] getTokens() {
return tokens;
}
public float[] getTimestamps() {
return timestamps;
}
}

49
sherpa-onnx/java-api/src/com/k2fsa/sherpa/onnx/OfflineStream.java Normal file
View File

@@ -0,0 +1,49 @@
// Copyright 2024 Xiaomi Corporation
package com.k2fsa.sherpa.onnx;
public class OfflineStream {
static {
System.loadLibrary("sherpa-onnx-jni");
}
private long ptr = 0;
public OfflineStream() {
this.ptr = 0;
}
public OfflineStream(long ptr) {
this.ptr = ptr;
}
public long getPtr() {
return ptr;
}
public void setPtr(long ptr) {
this.ptr = ptr;
}
public void acceptWaveform(float[] samples, int sampleRate) {
acceptWaveform(this.ptr, samples, sampleRate);
}
public void release() {
// stream object must be release after used
if (this.ptr == 0) {
return;
}
delete(this.ptr);
this.ptr = 0;
}
@Override
protected void finalize() throws Throwable {
release();
super.finalize();
}
private native void acceptWaveform(long ptr, float[] samples, int sampleRate);
private native void delete(long ptr);
}

55
sherpa-onnx/java-api/src/com/k2fsa/sherpa/onnx/OfflineTransducerModelConfig.java Normal file
View File

@@ -0,0 +1,55 @@
// Copyright 2024 Xiaomi Corporation
package com.k2fsa.sherpa.onnx;
public class OfflineTransducerModelConfig {
private final String encoder;
private final String decoder;
private final String joiner;
private OfflineTransducerModelConfig(Builder builder) {
this.encoder = builder.encoder;
this.decoder = builder.decoder;
this.joiner = builder.joiner;
}
public static Builder builder() {
return new Builder();
}
public String getEncoder() {
return encoder;
}
public String getDecoder() {
return decoder;
}
public String getJoiner() {
return joiner;
}
public static class Builder {
private String encoder = "";
private String decoder = "";
private String joiner = "";
public OfflineTransducerModelConfig build() {
return new OfflineTransducerModelConfig(this);
}
public Builder setEncoder(String encoder) {
this.encoder = encoder;
return this;
}
public Builder setDecoder(String decoder) {
this.decoder = decoder;
return this;
}
public Builder setJoiner(String joiner) {
this.joiner = joiner;
return this;
}
}
}

80
sherpa-onnx/java-api/src/com/k2fsa/sherpa/onnx/OfflineWhisperModelConfig.java Normal file
View File

@@ -0,0 +1,80 @@
// Copyright 2024 Xiaomi Corporation
package com.k2fsa.sherpa.onnx;
public class OfflineWhisperModelConfig {
private final String encoder;
private final String decoder;
private final String language;
private final String task;
private final int tailPaddings;
private OfflineWhisperModelConfig(Builder builder) {
this.encoder = builder.encoder;
this.decoder = builder.decoder;
this.language = builder.language;
this.task = builder.task;
this.tailPaddings = builder.tailPaddings;
}
public static Builder builder() {
return new Builder();
}
public String getEncoder() {
return encoder;
}
public String getDecoder() {
return decoder;
}
public String getLanguage() {
return language;
}
public String getTask() {
return task;
}
public int getTailPaddings() {
return tailPaddings;
}
public static class Builder {
private String encoder = "";
private String decoder = "";
private String language = "en"; // used only with multilingual models
private String task = "transcribe"; // used only with multilingual models
private int tailPaddings = 1000; // number of frames to pad
public OfflineWhisperModelConfig build() {
return new OfflineWhisperModelConfig(this);
}
public Builder setEncoder(String encoder) {
this.encoder = encoder;
return this;
}
public Builder setDecoder(String decoder) {
this.decoder = decoder;
return this;
}
public Builder setLanguage(String language) {
this.language = language;
return this;
}
public Builder setTask(String task) {
this.task = task;
return this;
}
public Builder setTailPaddings(int tailPaddings) {
this.tailPaddings = tailPaddings;
return this;
}
}
}

1
sherpa-onnx/java-api/src/com/k2fsa/sherpa/onnx/OnlineModelConfig.java
View File

@@ -12,6 +12,7 @@ public class OnlineModelConfig {
private final boolean debug;
private final String provider;
private final String modelType;
private OnlineModelConfig(Builder builder) {
this.transducer = builder.transducer;
this.paraformer = builder.paraformer;

15
sherpa-onnx/java-api/src/com/k2fsa/sherpa/onnx/OnlineRecognizer.java
View File

@@ -15,19 +15,6 @@ public class OnlineRecognizer {
ptr = newFromFile(config);
}
/*
public static float[] readWavFile(String fileName) {
// read data from the filename
Object[] wavdata = readWave(fileName);
Object data = wavdata[0]; // data[0] is float data, data[1] sample rate
float[] floatData = (float[]) data;
return floatData;
}
*/
public void decode(OnlineStream s) {
decode(ptr, s.getPtr());
}
@@ -55,7 +42,7 @@ public class OnlineRecognizer {
release();
}
// recognizer release, you'd better call it manually if not use anymore
// You'd better call it manually if it is not used anymore
public void release() {
if (this.ptr == 0) {
return;

1
sherpa-onnx/java-api/src/com/k2fsa/sherpa/onnx/OnlineRecognizerConfig.java
View File

@@ -12,6 +12,7 @@ public class OnlineRecognizerConfig {
private final int maxActivePaths;
private final String hotwordsFile;
private final float hotwordsScore;
private OnlineRecognizerConfig(Builder builder) {
this.featConfig = builder.featConfig;
this.modelConfig = builder.modelConfig;