updated for 0.7.1

DeepSpeech-jni/org_mozilla_deepspeech_DeepSpeech.cpp

@@ -1,20 +1,16 @@
+#include <iostream>
 #include "org_mozilla_deepspeech_DeepSpeech.h"
 
 jint
 Java_org_mozilla_deepspeech_DeepSpeech_nCreateModel(JNIEnv *env, jclass, jstring modelPath,
-                                                    jlong nCep,
-                                                    jlong nContext, jstring alphabetConfigPath,
-                                                    jlong beamWidth,
                                                     jobject modelStatePtr) {
-    jboolean isModelPathCopy, isAlphaBetCopy;
+    jboolean isModelPathCopy;
     ModelState *ptr = nullptr;
     auto modelPathCStr = (char *) env->GetStringUTFChars(modelPath, &isModelPathCopy);
-    auto alphaBetCStr = (char *) env->GetStringUTFChars(alphabetConfigPath, &isAlphaBetCopy);
 
-    jint state = DS_CreateModel(modelPathCStr, static_cast<unsigned int>(nCep),
-                                static_cast<unsigned int>(nContext), alphaBetCStr,
-                                static_cast<unsigned int>(beamWidth),
-                                &ptr);
+// https://github.com/mozilla/DeepSpeech/commit/8c820817794d445746aefb1b5347b35bf5e0c621#diff-0317a0e76ece10e0dba742af310a2362
+    jint state = DS_CreateModel(modelPathCStr, &ptr);
+
     auto *bufferPtr = (jlong *) (env->GetDirectBufferAddress(modelStatePtr));
 
     bufferPtr[0] = reinterpret_cast<jlong>(ptr);
@@ -26,46 +22,62 @@ Java_org_mozilla_deepspeech_DeepSpeech_nCreateModel(JNIEnv *env, jclass, jstring
 }
 
 void Java_org_mozilla_deepspeech_DeepSpeech_destroyModel(JNIEnv *, jclass, jlong modelPtr) {
-    DS_DestroyModel(reinterpret_cast<ModelState *>(modelPtr));
+    DS_FreeModel(reinterpret_cast<ModelState *>(modelPtr));
+}
+
+jint
+Java_org_mozilla_deepspeech_DeepSpeech_getModelBeamWidth(JNIEnv *env, jclass, jlong modelStatePtr) {
+    return DS_GetModelBeamWidth((ModelState *) modelStatePtr);
+}
+
+jint
+Java_org_mozilla_deepspeech_DeepSpeech_setModelBeamWidth(JNIEnv *env, jclass, jlong modelStatePtr, jlong beamWidth) {
+    return DS_SetModelBeamWidth((ModelState *) modelStatePtr, static_cast<unsigned int>(beamWidth));
+}
+
+jint
+Java_org_mozilla_deepspeech_DeepSpeech_getModelSampleRate(JNIEnv *env, jclass, jlong modelStatePtr) {
+    return DS_GetModelSampleRate((ModelState *) modelStatePtr);
 }
 
 jint
 Java_org_mozilla_deepspeech_DeepSpeech_enableDecoderWithLM(JNIEnv *env, jclass, jlong modelStatePtr,
-                                                           jstring alphaBetConfigPath,
-                                                           jstring lmPath,
-                                                           jstring triePath, jfloat alpha,
-                                                           jfloat beta) {
-    jboolean isAlphabetStrCopy, isLmPathCopy, isTriePathCopy;
-    auto alphaBetConfigPathCStr = const_cast<char *>(env->GetStringUTFChars(alphaBetConfigPath,
-                                                                            &isAlphabetStrCopy));
-    auto lmPathCStr = const_cast<char *>(env->GetStringUTFChars(lmPath, &isLmPathCopy));
-    auto triePathCStr = const_cast<char *>(env->GetStringUTFChars(triePath, &isTriePathCopy));
-
-    jint status = DS_EnableDecoderWithLM((ModelState *) modelStatePtr, alphaBetConfigPathCStr,
-                                         lmPathCStr, triePathCStr,
-                                         alpha, beta);
-
-    if (isAlphabetStrCopy == JNI_TRUE) {
-        env->ReleaseStringUTFChars(alphaBetConfigPath, alphaBetConfigPathCStr);
-    }
+                                                           jstring scorerPath,
+                                                           jfloat alpha, jfloat beta) {
+    jboolean isLmPathCopy;
+    auto scorerPathCStr = const_cast<char *>(env->GetStringUTFChars(scorerPath, &isLmPathCopy));
+
+    // https://github.com/mozilla/DeepSpeech/pull/2681/files
+    // lm: models/lm.binary, trie: models/trie
+    // became scorer: models/kenlm.scorer
+    DS_EnableExternalScorer((ModelState *) modelStatePtr, scorerPathCStr);
+    jint status = DS_SetScorerAlphaBeta((ModelState *) modelStatePtr, alpha, beta);
+
     if (isLmPathCopy == JNI_TRUE) {
-        env->ReleaseStringUTFChars(lmPath, lmPathCStr);
-    }
-    if (isTriePathCopy == JNI_TRUE) {
-        env->ReleaseStringUTFChars(triePath, triePathCStr);
+        env->ReleaseStringUTFChars(scorerPath, scorerPathCStr);
     }
 
     return status;
 }
 
+jint
+Java_org_mozilla_deepspeech_DeepSpeech_setScorerAlphaBeta(JNIEnv *env, jclass, jlong modelStatePtr,
+                                                           jfloat alpha, jfloat beta) {
+    return DS_SetScorerAlphaBeta((ModelState *) modelStatePtr, alpha, beta);
+}
+
+jint
+Java_org_mozilla_deepspeech_DeepSpeech_disableExternalScorer(JNIEnv *env, jclass, jlong modelStatePtr) {
+    jint status = DS_DisableExternalScorer((ModelState *) modelStatePtr);
+    return status;
+}
+
 jstring
 Java_org_mozilla_deepspeech_DeepSpeech_nSpeechToText(JNIEnv *env, jclass, jlong modelStatePtr,
-                                                     jobject audioBuffer, jlong numSamples,
-                                                     jlong sampleRate) {
+                                                     jobject audioBuffer, jlong numSamples) {
     auto *array = (short *) (env->GetDirectBufferAddress(audioBuffer));
     char *cStr = DS_SpeechToText((ModelState *) modelStatePtr, array,
-                                 static_cast<unsigned int>(numSamples),
-                                 (unsigned int) sampleRate);
+                                 static_cast<unsigned int>(numSamples));
     if (cStr == nullptr) {
         return nullptr;
     }
@@ -76,12 +88,10 @@ Java_org_mozilla_deepspeech_DeepSpeech_nSpeechToText(JNIEnv *env, jclass, jlong
 
 jstring
 Java_org_mozilla_deepspeech_DeepSpeech_speechToTextUnsafe(JNIEnv *env, jclass, jlong modelStatePtr,
-                                                     jlong audioBuffer, jlong numSamples,
-                                                     jlong sampleRate) {
+                                                     jlong audioBuffer, jlong numSamples) {
     auto *array = (short *) (audioBuffer);
     char *cStr = DS_SpeechToText((ModelState *) modelStatePtr, array,
-                                 static_cast<unsigned int>(numSamples),
-                                 (unsigned int) sampleRate);
+                                 static_cast<unsigned int>(numSamples));
     if (cStr == nullptr) {
         return nullptr;
     }
@@ -95,36 +105,34 @@ Java_org_mozilla_deepspeech_DeepSpeech_nSpeechToTextWithMetadata(JNIEnv *env, jc
                                                                  jlong modelStatePtr,
                                                                  jobject audioBuffer,
                                                                  jlong bufferSize,
-                                                                 jlong sampleRate) {
+                                                                 jlong numResults) {
     auto *array = static_cast<short *>(env->GetDirectBufferAddress(audioBuffer));
     auto metaPtr = reinterpret_cast<jlong>(DS_SpeechToTextWithMetadata((ModelState *) modelStatePtr,
                                                                        array,
                                                                        static_cast<unsigned int>(bufferSize),
-                                                                       static_cast<unsigned int>(sampleRate)));
+                                                                       static_cast<unsigned int>(numResults)));
     return metaPtr;
 }
 jlong
 Java_org_mozilla_deepspeech_DeepSpeech_speechToTextWithMetadataUnsafe(JNIEnv *, jclass,
                                                                  jlong modelStatePtr,
                                                                  jlong audioBuffer,
                                                                  jlong bufferSize,
-                                                                 jlong sampleRate) {
+                                                                 jlong numResults) {
     auto *array = (short *)audioBuffer;
     auto metaPtr = reinterpret_cast<jlong>(DS_SpeechToTextWithMetadata((ModelState *) modelStatePtr,
                                                                        array,
                                                                        static_cast<unsigned int>(bufferSize),
-                                                                       static_cast<unsigned int>(sampleRate)));
+                                                                       static_cast<unsigned int>(numResults)));
     return metaPtr;
 }
 
 jint Java_org_mozilla_deepspeech_DeepSpeech_nSetupStream(JNIEnv *env, jclass, jlong modelStatePtr,
-                                                         jlong preAllocFrames, jlong sampleRate,
                                                          jobject streamPtr) {
     StreamingState *pStreamingState;
 
-    jint status = DS_SetupStream((ModelState *) modelStatePtr,
-                                 static_cast<unsigned int>(preAllocFrames),
-                                 static_cast<unsigned int>(sampleRate), &pStreamingState);
+    jint status = DS_CreateStream((ModelState *) modelStatePtr, &pStreamingState);
+
     auto p = (StreamingState **) env->GetDirectBufferAddress(streamPtr);
     *p = pStreamingState;
     return status;
@@ -155,20 +163,34 @@ jstring Java_org_mozilla_deepspeech_DeepSpeech_finishStream(JNIEnv *env, jclass,
 }
 
 jlong
-Java_org_mozilla_deepspeech_DeepSpeech_finishStreamWithMetadata(JNIEnv *, jclass, jlong streamPtr) {
-    return reinterpret_cast<jlong>(DS_FinishStreamWithMetadata((StreamingState *) streamPtr));
+Java_org_mozilla_deepspeech_DeepSpeech_finishStreamWithMetadata(JNIEnv *, jclass, jlong streamPtr, jlong numResults) {
+    return reinterpret_cast<jlong>(DS_FinishStreamWithMetadata((StreamingState *) streamPtr, numResults));
 }
 
 void Java_org_mozilla_deepspeech_DeepSpeech_discardStream(JNIEnv *, jclass, jlong streamPtr) {
-    DS_DiscardStream((StreamingState *) streamPtr);
+    DS_FreeStream((StreamingState *) streamPtr);
 }
 
 void Java_org_mozilla_deepspeech_DeepSpeech_freeMetadata(JNIEnv *, jclass, jlong metaPtr) {
     DS_FreeMetadata((Metadata *) metaPtr);
 }
 
-void Java_org_mozilla_deepspeech_DeepSpeech_printVersions(JNIEnv *, jclass) {
-    DS_PrintVersions();
+jstring Java_org_mozilla_deepspeech_DeepSpeech_getVersion(JNIEnv *env, jclass) {
+    char *cString = DS_Version();
+    size_t cStrLen = strlen(cString);
+    jstring str = env->NewString(reinterpret_cast<const jchar *>(cString),
+                                     static_cast<jsize>(cStrLen));
+    DS_FreeString(cString);
+    return str;
+}
+
+jstring Java_org_mozilla_deepspeech_DeepSpeech_errorCodeToErrorMessage(JNIEnv *env, jclass, jlong errorCode) {
+    char *cString = DS_ErrorCodeToErrorMessage(errorCode);
+    size_t cStrLen = strlen(cString);
+    jstring str = env->NewString(reinterpret_cast<const jchar *>(cString),
+                                     static_cast<jsize>(cStrLen));
+    DS_FreeString(cString);
+    return str;
 }
 
 jint Java_org_mozilla_deepspeech_DeepSpeech_nGetConfiguration(JNIEnv *, jclass) {
@@ -181,4 +203,4 @@ jint Java_org_mozilla_deepspeech_DeepSpeech_nGetConfiguration(JNIEnv *, jclass)
     return BuildConfiguration::INVALID; // This should never be returned
 #endif
 #endif
-}
+}

DeepSpeech-jni/org_mozilla_deepspeech_DeepSpeech.h

@@ -16,19 +16,35 @@ enum BuildConfiguration {
 };
 
 JNIEXPORT jint JNICALL Java_org_mozilla_deepspeech_DeepSpeech_nCreateModel
-        (JNIEnv *, jclass, jstring, jlong, jlong, jstring, jlong, jobject);
+        (JNIEnv *, jclass, jstring, jobject);
 
 JNIEXPORT void JNICALL Java_org_mozilla_deepspeech_DeepSpeech_destroyModel
         (JNIEnv *, jclass, jlong);
 
+
+JNIEXPORT jint JNICALL Java_org_mozilla_deepspeech_DeepSpeech_getModelBeamWidth
+        (JNIEnv *, jclass, jlong);
+
+JNIEXPORT jint JNICALL Java_org_mozilla_deepspeech_DeepSpeech_setModelBeamWidth
+        (JNIEnv *, jclass, jlong, jlong);
+
+JNIEXPORT jint JNICALL Java_org_mozilla_deepspeech_DeepSpeech_getModelSampleRate
+        (JNIEnv *, jclass, jlong);
+
 JNIEXPORT jint JNICALL Java_org_mozilla_deepspeech_DeepSpeech_enableDecoderWithLM
-        (JNIEnv *, jclass, jlong, jstring, jstring, jstring, jfloat, jfloat);
+        (JNIEnv *, jclass, jlong, jstring, jfloat, jfloat);
+
+JNIEXPORT jint JNICALL Java_org_mozilla_deepspeech_DeepSpeech_setScorerAlphaBeta
+        (JNIEnv *env, jclass, jlong, jfloat, jfloat);
+
+JNIEXPORT jint JNICALL Java_org_mozilla_deepspeech_DeepSpeech_disableExternalScorer
+        (JNIEnv *env, jclass, jlong);
 
 JNIEXPORT jstring JNICALL Java_org_mozilla_deepspeech_DeepSpeech_nSpeechToText
-        (JNIEnv *, jclass, jlong, jobject, jlong, jlong);
+        (JNIEnv *, jclass, jlong, jobject, jlong);
 
 JNIEXPORT jstring JNICALL Java_org_mozilla_deepspeech_DeepSpeech_speechToTextUnsafe
-        (JNIEnv *, jclass, jlong, jlong, jlong, jlong);
+        (JNIEnv *, jclass, jlong, jlong, jlong);
 
 JNIEXPORT jlong JNICALL Java_org_mozilla_deepspeech_DeepSpeech_nSpeechToTextWithMetadata
         (JNIEnv *, jclass, jlong, jobject, jlong, jlong);
@@ -37,7 +53,7 @@ JNIEXPORT jlong JNICALL Java_org_mozilla_deepspeech_DeepSpeech_speechToTextWithM
         (JNIEnv *, jclass, jlong, jlong, jlong, jlong);
 
 JNIEXPORT jint JNICALL Java_org_mozilla_deepspeech_DeepSpeech_nSetupStream
-        (JNIEnv *, jclass, jlong, jlong, jlong, jobject);
+        (JNIEnv *, jclass, jlong, jobject);
 
 JNIEXPORT void JNICALL Java_org_mozilla_deepspeech_DeepSpeech_nFeedAudioContent
         (JNIEnv *, jclass, jlong, jobject, jlong);
@@ -57,9 +73,12 @@ JNIEXPORT void JNICALL Java_org_mozilla_deepspeech_DeepSpeech_discardStream
 JNIEXPORT void JNICALL Java_org_mozilla_deepspeech_DeepSpeech_freeMetadata
         (JNIEnv *, jclass, jlong);
 
-JNIEXPORT void JNICALL Java_org_mozilla_deepspeech_DeepSpeech_printVersions
+JNIEXPORT jstring JNICALL Java_org_mozilla_deepspeech_DeepSpeech_getVersion
         (JNIEnv *, jclass);
 
+JNIEXPORT jstring JNICALL Java_org_mozilla_deepspeech_DeepSpeech_errorCodeToErrorMessage
+        (JNIEnv *, jclass, jlong);
+
 JNIEXPORT jint JNICALL Java_org_mozilla_deepspeech_DeepSpeech_nGetConfiguration
         (JNIEnv *, jclass);
 

Dockerfile

@@ -0,0 +1,6 @@
+FROM java:8
+
+COPY . /idear
+WORKDIR /idear
+RUN ./gradlew buildPlugin
+

Makefile

@@ -0,0 +1,13 @@
+default:
+	@docker run --rm -v $(PWD):/idear -w /idear openasr/idear ./gradlew buildPlugin
+
+test:
+	@docker run --rm -v $(PWD):/idear -w /idear openasr/idear ./gradlew test
+
+docker:
+	@rm -rf build out
+	@docker build -t openasr/idear .
+
+push:
+	@docker push openasr/idear:latest
+

libdeepspeech/build.gradle

@@ -1,9 +1,11 @@
 plugins {
     id 'java'
+    id 'distribution'
+    id 'maven-publish'
 }
 
 group 'org.mozilla.deepspeech'
-version '1.0'
+version '1.0.4-SNAPSHOT'
 
 sourceCompatibility = 1.6
 
@@ -14,4 +16,50 @@ repositories {
 dependencies {
     testCompile group: 'junit', name: 'junit', version: '4.12'
     compile("org.jetbrains:annotations:16.0.1") // Jetbrains annotations library
-}
+}
+
+//task linuxJniZip(type: Zip) {
+//    archiveName("deepspeech-jni.zip")
+//    destinationDir(file("$buildDir/dist"))
+////    archiveFileName = "deepspeech-jni.zip"
+////    destinationDirectory = file("$buildDir/target")
+//    from "../DeepSpeech-jni/libs/linux"
+//}
+
+//artifacts {
+//    linuxJniZip
+//}
+
+distributions {
+    linux {
+        contents {
+            from '../DeepSpeech-jni/libs/linux'
+        }
+    }
+    windows {
+        contents {
+            from '../DeepSpeech-jni/libs/windows'
+        }
+    }
+}
+
+publishing {
+  publications {
+      libdeepspeech(MavenPublication) {
+          from components.java
+          artifact linuxDistZip
+          artifact windowsDistZip
+      }
+  }
+  repositories {
+    maven {
+      name "GitHubPackages"
+      url "https://maven.pkg.github.com/" + System.getenv('GITHUB_REPOSITORY')
+      credentials {
+        username System.getenv('GITHUB_ACTOR')
+        password System.getenv('GITHUB_TOKEN')
+      }
+    }
+  }
+}
+