asr-models-qnn-binary

k2-fsa/sherpa-onnxasr-models-qnn-binaryDec 9, 2025by csukuangfj

AI Summary

This release provides binary files for QNN (Qualcomm Neural Processing Unit) SDK 2.40.0.251030, specifically targeting SenseVoice ASR models for Android devices like the Xiaomi 17 Pro. It includes necessary .so libraries and documentation for deployment.

Key Highlights

Context binaries generated using QNN SDK 2.40.0.251030.
Download links for required .so libraries are provided.
Documentation for Android usage via Kotlin API.
Support for SenseVoice model on Xiaomi 17 Pro.

New Features

SenseVoice ASR support on Qualcomm QNN (Android).

Full Release Notes

Context binary files in this page are generated using qnn sdk 2.40.0.251030

You can download the requires `.so` lib files from qnn sdk 2.40.0.251030 from 
https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models-qnn/qnn-libs-2.40.0.251030.tar.bz2

Please refer to
  - https://k2-fsa.github.io/sherpa/onnx/qnn/run-executables-on-your-phone-binary.html
  - https://k2-fsa.github.io/sherpa/onnx/qnn/models.html
for usages.

See
https://github.com/k2-fsa/sherpa-onnx/blob/master/sherpa-onnx/kotlin-api/OfflineRecognizer.kt#L1119
for how to use within Android.
```kotlin
        9022 -> {
            // for my Xiaomi 17 Pro
            val modelDir = "sherpa-onnx-qnn-SM8850-binary-10-seconds-sense-voice-zh-en-ja-ko-yue-2024-07-17-int8"
            return OfflineModelConfig(
                provider = "qnn",
                senseVoice = OfflineSenseVoiceModelConfig(
                    qnnConfig = QnnConfig(
                        // Please copy libQnnHtp.so and libQnnSystem.so to jniLibs/arm64-v8a by yourself
                        backendLib = "libQnnHtp.so",
                        systemLib = "libQnnSystem.so",
                        contextBinary = "$modelDir/model.bin",
                    ),
                ),
                tokens = "$modelDir/tokens.txt",
                debug = true,
            )
        }
```