Refactor the JNI interface to make it more modular and maintainable (#802)

This commit is contained in:
Fangjun Kuang
2024-04-24 09:48:42 +08:00
committed by GitHub
parent dc5af04830
commit 9b67a476e6
116 changed files with 3502 additions and 3316 deletions

View File

@@ -1,188 +0,0 @@
package com.k2fsa.sherpa.onnx
import android.content.res.AssetManager
const val TAG = "sherpa-onnx"
data class OfflineZipformerAudioTaggingModelConfig(
var model: String = "",
)
data class AudioTaggingModelConfig(
var zipformer: OfflineZipformerAudioTaggingModelConfig = OfflineZipformerAudioTaggingModelConfig(),
var ced: String = "",
var numThreads: Int = 1,
var debug: Boolean = false,
var provider: String = "cpu",
)
data class AudioTaggingConfig(
var model: AudioTaggingModelConfig,
var labels: String,
var topK: Int = 5,
)
data class AudioEvent(
val name: String,
val index: Int,
val prob: Float,
)
class AudioTagging(
assetManager: AssetManager? = null,
config: AudioTaggingConfig,
) {
private var ptr: Long
init {
ptr = if (assetManager != null) {
newFromAsset(assetManager, config)
} else {
newFromFile(config)
}
}
protected fun finalize() {
if (ptr != 0L) {
delete(ptr)
ptr = 0
}
}
fun release() = finalize()
fun createStream(): OfflineStream {
val p = createStream(ptr)
return OfflineStream(p)
}
@Suppress("UNCHECKED_CAST")
fun compute(stream: OfflineStream, topK: Int = -1): ArrayList<AudioEvent> {
val events: Array<Any> = compute(ptr, stream.ptr, topK)
val ans = ArrayList<AudioEvent>()
for (e in events) {
val p: Array<Any> = e as Array<Any>
ans.add(
AudioEvent(
name = p[0] as String,
index = p[1] as Int,
prob = p[2] as Float,
)
)
}
return ans
}
private external fun newFromAsset(
assetManager: AssetManager,
config: AudioTaggingConfig,
): Long
private external fun newFromFile(
config: AudioTaggingConfig,
): Long
private external fun delete(ptr: Long)
private external fun createStream(ptr: Long): Long
private external fun compute(ptr: Long, streamPtr: Long, topK: Int): Array<Any>
companion object {
init {
System.loadLibrary("sherpa-onnx-jni")
}
}
}
// please refer to
// https://github.com/k2-fsa/sherpa-onnx/releases/tag/audio-tagging-models
// to download more models
//
// See also
// https://k2-fsa.github.io/sherpa/onnx/audio-tagging/
fun getAudioTaggingConfig(type: Int, numThreads: Int = 1): AudioTaggingConfig? {
when (type) {
0 -> {
val modelDir = "sherpa-onnx-zipformer-small-audio-tagging-2024-04-15"
return AudioTaggingConfig(
model = AudioTaggingModelConfig(
zipformer = OfflineZipformerAudioTaggingModelConfig(model = "$modelDir/model.int8.onnx"),
numThreads = numThreads,
debug = true,
),
labels = "$modelDir/class_labels_indices.csv",
topK = 3,
)
}
1 -> {
val modelDir = "sherpa-onnx-zipformer-audio-tagging-2024-04-09"
return AudioTaggingConfig(
model = AudioTaggingModelConfig(
zipformer = OfflineZipformerAudioTaggingModelConfig(model = "$modelDir/model.int8.onnx"),
numThreads = numThreads,
debug = true,
),
labels = "$modelDir/class_labels_indices.csv",
topK = 3,
)
}
2 -> {
val modelDir = "sherpa-onnx-ced-tiny-audio-tagging-2024-04-19"
return AudioTaggingConfig(
model = AudioTaggingModelConfig(
ced = "$modelDir/model.int8.onnx",
numThreads = numThreads,
debug = true,
),
labels = "$modelDir/class_labels_indices.csv",
topK = 3,
)
}
3 -> {
val modelDir = "sherpa-onnx-ced-mini-audio-tagging-2024-04-19"
return AudioTaggingConfig(
model = AudioTaggingModelConfig(
ced = "$modelDir/model.int8.onnx",
numThreads = numThreads,
debug = true,
),
labels = "$modelDir/class_labels_indices.csv",
topK = 3,
)
}
4 -> {
val modelDir = "sherpa-onnx-ced-small-audio-tagging-2024-04-19"
return AudioTaggingConfig(
model = AudioTaggingModelConfig(
ced = "$modelDir/model.int8.onnx",
numThreads = numThreads,
debug = true,
),
labels = "$modelDir/class_labels_indices.csv",
topK = 3,
)
}
5 -> {
val modelDir = "sherpa-onnx-ced-base-audio-tagging-2024-04-19"
return AudioTaggingConfig(
model = AudioTaggingModelConfig(
ced = "$modelDir/model.int8.onnx",
numThreads = numThreads,
debug = true,
),
labels = "$modelDir/class_labels_indices.csv",
topK = 3,
)
}
}
return null
}

View File

@@ -0,0 +1 @@
../../../../../../../../../../../../sherpa-onnx/kotlin-api/AudioTagging.kt

View File

@@ -46,7 +46,6 @@ import androidx.compose.ui.unit.dp
import androidx.compose.ui.unit.sp
import androidx.core.app.ActivityCompat
import com.k2fsa.sherpa.onnx.AudioEvent
import com.k2fsa.sherpa.onnx.Tagger
import kotlin.concurrent.thread

View File

@@ -13,13 +13,14 @@ import androidx.compose.material3.Surface
import androidx.compose.runtime.Composable
import androidx.compose.ui.Modifier
import androidx.core.app.ActivityCompat
import com.k2fsa.sherpa.onnx.Tagger
import com.k2fsa.sherpa.onnx.audio.tagging.ui.theme.SherpaOnnxAudioTaggingTheme
const val TAG = "sherpa-onnx"
private const val REQUEST_RECORD_AUDIO_PERMISSION = 200
// adb emu avd hostmicon
// to enable mic inside the emulator
class MainActivity : ComponentActivity() {
private val permissions: Array<String> = arrayOf(Manifest.permission.RECORD_AUDIO)
override fun onCreate(savedInstanceState: Bundle?) {

View File

@@ -1,24 +0,0 @@
package com.k2fsa.sherpa.onnx
class OfflineStream(var ptr: Long) {
fun acceptWaveform(samples: FloatArray, sampleRate: Int) =
acceptWaveform(ptr, samples, sampleRate)
protected fun finalize() {
if (ptr != 0L) {
delete(ptr)
ptr = 0
}
}
fun release() = finalize()
private external fun acceptWaveform(ptr: Long, samples: FloatArray, sampleRate: Int)
private external fun delete(ptr: Long)
companion object {
init {
System.loadLibrary("sherpa-onnx-jni")
}
}
}

View File

@@ -0,0 +1 @@
../../../../../../../../../../../../sherpa-onnx/kotlin-api/OfflineStream.kt

View File

@@ -1,7 +1,9 @@
package com.k2fsa.sherpa.onnx
package com.k2fsa.sherpa.onnx.audio.tagging
import android.content.res.AssetManager
import android.util.Log
import com.k2fsa.sherpa.onnx.AudioTagging
import com.k2fsa.sherpa.onnx.getAudioTaggingConfig
object Tagger {
@@ -17,7 +19,7 @@ object Tagger {
return
}
Log.i(TAG, "Initializing audio tagger")
Log.i("sherpa-onnx", "Initializing audio tagger")
val config = getAudioTaggingConfig(type = 0, numThreads = numThreads)!!
_tagger = AudioTagging(assetManager, config)
}