Add Java and Kotlin API for NeMo Canary models (#2359)
Add support for the NeMo Canary model in both Java and Kotlin APIs, wiring it through JNI and updating examples and CI. - Introduce OfflineCanaryModelConfig in Kotlin and Java with builder patterns - Extend OfflineRecognizer to accept and apply the new canary config via setConfig - Update JNI binding (GetOfflineConfig) and getOfflineModelConfig mapping (type 32), plus examples and CI workflows
This commit is contained in:
@@ -0,0 +1,80 @@
|
||||
// Copyright 2025 Xiaomi Corporation
|
||||
|
||||
package com.k2fsa.sherpa.onnx;
|
||||
|
||||
public class OfflineCanaryModelConfig {
|
||||
private final String encoder;
|
||||
private final String decoder;
|
||||
private final String srcLang;
|
||||
private final String tgtLang;
|
||||
private final boolean usePnc;
|
||||
|
||||
private OfflineCanaryModelConfig(Builder builder) {
|
||||
this.encoder = builder.encoder;
|
||||
this.decoder = builder.decoder;
|
||||
this.srcLang = builder.srcLang;
|
||||
this.tgtLang = builder.tgtLang;
|
||||
this.usePnc = builder.usePnc;
|
||||
}
|
||||
|
||||
public static Builder builder() {
|
||||
return new Builder();
|
||||
}
|
||||
|
||||
public String getEncoder() {
|
||||
return encoder;
|
||||
}
|
||||
|
||||
public String getDecoder() {
|
||||
return decoder;
|
||||
}
|
||||
|
||||
public String getSrcLang() {
|
||||
return srcLang;
|
||||
}
|
||||
|
||||
public String getTgtLang() {
|
||||
return tgtLang;
|
||||
}
|
||||
|
||||
public boolean isUsePnc() {
|
||||
return usePnc;
|
||||
}
|
||||
|
||||
public static class Builder {
|
||||
private String encoder = "";
|
||||
private String decoder = "";
|
||||
private String srcLang = "en";
|
||||
private String tgtLang = "en";
|
||||
private boolean usePnc = true;
|
||||
|
||||
public OfflineCanaryModelConfig build() {
|
||||
return new OfflineCanaryModelConfig(this);
|
||||
}
|
||||
|
||||
public Builder setEncoder(String encoder) {
|
||||
this.encoder = encoder;
|
||||
return this;
|
||||
}
|
||||
|
||||
public Builder setDecoder(String decoder) {
|
||||
this.decoder = decoder;
|
||||
return this;
|
||||
}
|
||||
|
||||
public Builder setSrcLang(String srcLang) {
|
||||
this.srcLang = srcLang;
|
||||
return this;
|
||||
}
|
||||
|
||||
public Builder setTgtLang(String tgtLang) {
|
||||
this.tgtLang = tgtLang;
|
||||
return this;
|
||||
}
|
||||
|
||||
public Builder setUsePnc(boolean usePnc) {
|
||||
this.usePnc = usePnc;
|
||||
return this;
|
||||
}
|
||||
}
|
||||
}
|
||||
@@ -12,6 +12,7 @@ public class OfflineModelConfig {
|
||||
private final OfflineSenseVoiceModelConfig senseVoice;
|
||||
private final OfflineDolphinModelConfig dolphin;
|
||||
private final OfflineZipformerCtcModelConfig zipformerCtc;
|
||||
private final OfflineCanaryModelConfig canary;
|
||||
private final String teleSpeech;
|
||||
private final String tokens;
|
||||
private final int numThreads;
|
||||
@@ -30,6 +31,7 @@ public class OfflineModelConfig {
|
||||
this.moonshine = builder.moonshine;
|
||||
this.nemo = builder.nemo;
|
||||
this.zipformerCtc = builder.zipformerCtc;
|
||||
this.canary = builder.canary;
|
||||
this.senseVoice = builder.senseVoice;
|
||||
this.dolphin = builder.dolphin;
|
||||
this.teleSpeech = builder.teleSpeech;
|
||||
@@ -78,6 +80,10 @@ public class OfflineModelConfig {
|
||||
return zipformerCtc;
|
||||
}
|
||||
|
||||
public OfflineCanaryModelConfig getCanary() {
|
||||
return canary;
|
||||
}
|
||||
|
||||
public String getTokens() {
|
||||
return tokens;
|
||||
}
|
||||
@@ -120,6 +126,7 @@ public class OfflineModelConfig {
|
||||
private OfflineSenseVoiceModelConfig senseVoice = OfflineSenseVoiceModelConfig.builder().build();
|
||||
private OfflineDolphinModelConfig dolphin = OfflineDolphinModelConfig.builder().build();
|
||||
private OfflineZipformerCtcModelConfig zipformerCtc = OfflineZipformerCtcModelConfig.builder().build();
|
||||
private OfflineCanaryModelConfig canary = OfflineCanaryModelConfig.builder().build();
|
||||
private String teleSpeech = "";
|
||||
private String tokens = "";
|
||||
private int numThreads = 1;
|
||||
@@ -158,6 +165,11 @@ public class OfflineModelConfig {
|
||||
return this;
|
||||
}
|
||||
|
||||
public Builder setCanary(OfflineCanaryModelConfig canary) {
|
||||
this.canary = canary;
|
||||
return this;
|
||||
}
|
||||
|
||||
public Builder setTeleSpeech(String teleSpeech) {
|
||||
this.teleSpeech = teleSpeech;
|
||||
return this;
|
||||
|
||||
@@ -4,10 +4,22 @@ package com.k2fsa.sherpa.onnx;
|
||||
|
||||
public class OfflineRecognizer {
|
||||
private long ptr = 0;
|
||||
private final OfflineRecognizerConfig config;
|
||||
|
||||
public OfflineRecognizer(OfflineRecognizerConfig config) {
|
||||
LibraryLoader.maybeLoad();
|
||||
ptr = newFromFile(config);
|
||||
|
||||
this.config = config;
|
||||
}
|
||||
|
||||
public void setConfig(OfflineRecognizerConfig config) {
|
||||
setConfig(ptr, config);
|
||||
// we don't update this.config
|
||||
}
|
||||
|
||||
public OfflineRecognizerConfig getConfig() {
|
||||
return config;
|
||||
}
|
||||
|
||||
public void decode(OfflineStream s) {
|
||||
@@ -60,6 +72,8 @@ public class OfflineRecognizer {
|
||||
|
||||
private native void decode(long ptr, long streamPtr);
|
||||
|
||||
private native void setConfig(long ptr, OfflineRecognizerConfig config);
|
||||
|
||||
private native void decodeStreams(long ptr, long[] streamPtrs);
|
||||
|
||||
private native Object[] getResult(long streamPtr);
|
||||
|
||||
Reference in New Issue
Block a user