Hugging Face's logo

qualcomm
/
Whisper-Small-Quantized

Automatic Speech Recognition
PyTorch
foundation
android
Model card Files
xet
 Community

Whisper-Small-Quantized: Optimized for Qualcomm Devices

We have applied w8a16 quantization to significantly enhance performance and efficiency. HuggingFace Whisper-Small ASR (Automatic Speech Recognition) model is a state-of-the-art system designed for transcribing spoken language into written text. This model is based on the transformer architecture and has been optimized for edge inference by replacing Multi-Head Attention (MHA) with Single-Head Attention (SHA) and linear layers with convolutional (conv) layers. It exhibits robust performance in realistic, noisy environments, making it highly reliable for real-world applications. Specifically, it excels in long-form transcription, capable of accurately transcribing audio clips up to 30 seconds long. Time to the first token is the encoder's latency, while time to each additional token is decoder's latency, where we assume a max decoded length specified below.

This is based on the implementation of Whisper-Small-Quantized found here. This repository contains pre-exported model files optimized for Qualcomm® devices. You can use the Qualcomm® AI Hub Models library to export with custom configurations. More details on model performance across various devices, can be found here.

Qualcomm AI Hub Models uses Qualcomm AI Hub Workbench to compile, profile, and evaluate this model. Sign up to run these models on a hosted Qualcomm® device.

Deploying Whisper-Small-Quantized on-device

This model is compatible with the Qualcomm Voice AI SDK. Download the SDK from the Qualcomm Package Manager to deploy this model on-device.

Getting Started

There are two ways to deploy this model on your device:

Option 1: Download Pre-Exported Models

Below are pre-exported model assets ready for deployment.

Runtime Precision Chipset SDK Versions Download
PRECOMPILED_QNN_ONNX w8a16 Snapdragon® X2 Elite QAIRT 2.45, ONNX Runtime 1.25.0 Download
PRECOMPILED_QNN_ONNX w8a16 Snapdragon® X Elite QAIRT 2.45, ONNX Runtime 1.25.0 Download
PRECOMPILED_QNN_ONNX w8a16 Snapdragon® 8 Gen 3 Mobile QAIRT 2.45, ONNX Runtime 1.25.0 Download
PRECOMPILED_QNN_ONNX w8a16 Snapdragon® 8 Gen 1 Mobile QAIRT 2.45, ONNX Runtime 1.25.0 Download
PRECOMPILED_QNN_ONNX w8a16 Qualcomm® QCS8550 (Proxy) QAIRT 2.45, ONNX Runtime 1.25.0 Download
PRECOMPILED_QNN_ONNX w8a16 Snapdragon® 8 Elite Mobile QAIRT 2.45, ONNX Runtime 1.25.0 Download
PRECOMPILED_QNN_ONNX w8a16 Snapdragon® 7 Gen 4 Mobile QAIRT 2.45, ONNX Runtime 1.25.0 Download
PRECOMPILED_QNN_ONNX w8a16 Snapdragon® 8 Elite Gen 5 Mobile QAIRT 2.45, ONNX Runtime 1.25.0 Download
PRECOMPILED_QNN_ONNX w8a16 Qualcomm® QCM6690 QAIRT 2.45, ONNX Runtime 1.25.0 Download
PRECOMPILED_QNN_ONNX w8a16 Qualcomm® QCS9075 QAIRT 2.45, ONNX Runtime 1.25.0 Download
QNN_CONTEXT_BINARY w8a16 Snapdragon® X2 Elite QAIRT 2.45 Download
QNN_CONTEXT_BINARY w8a16 Snapdragon® X Elite QAIRT 2.45 Download
QNN_CONTEXT_BINARY w8a16 Snapdragon® 8 Gen 3 Mobile QAIRT 2.45 Download
QNN_CONTEXT_BINARY w8a16 Qualcomm® QCS8550 (Proxy) QAIRT 2.45 Download
QNN_CONTEXT_BINARY w8a16 Qualcomm® SA8775P QAIRT 2.45 Download
QNN_CONTEXT_BINARY w8a16 Snapdragon® 8 Elite Mobile QAIRT 2.45 Download
QNN_CONTEXT_BINARY w8a16 Snapdragon® 7 Gen 4 Mobile QAIRT 2.45 Download
QNN_CONTEXT_BINARY w8a16 Snapdragon® 8 Elite Gen 5 Mobile QAIRT 2.45 Download
QNN_CONTEXT_BINARY w8a16 Qualcomm® SA7255P QAIRT 2.45 Download
QNN_CONTEXT_BINARY w8a16 Qualcomm® QCM6690 QAIRT 2.45 Download
QNN_CONTEXT_BINARY w8a16 Qualcomm® QCS9075 QAIRT 2.45 Download
VOICE_AI w8a16 Snapdragon® X2 Elite QAIRT 2.45 Download
VOICE_AI w8a16 Snapdragon® X Elite QAIRT 2.45 Download
VOICE_AI w8a16 Snapdragon® 8 Gen 3 Mobile QAIRT 2.45 Download
VOICE_AI w8a16 Qualcomm® QCS8550 (Proxy) QAIRT 2.45 Download
VOICE_AI w8a16 Qualcomm® SA8775P QAIRT 2.45 Download
VOICE_AI w8a16 Snapdragon® 8 Elite Mobile QAIRT 2.45 Download
VOICE_AI w8a16 Snapdragon® 7 Gen 4 Mobile QAIRT 2.45 Download
VOICE_AI w8a16 Snapdragon® 8 Elite Gen 5 Mobile QAIRT 2.45 Download
VOICE_AI w8a16 Qualcomm® SA7255P QAIRT 2.45 Download
VOICE_AI w8a16 Qualcomm® QCM6690 QAIRT 2.45 Download
VOICE_AI w8a16 Qualcomm® QCS9075 QAIRT 2.45 Download

For more device-specific assets and performance metrics, visit Whisper-Small-Quantized on Qualcomm® AI Hub.

Option 2: Export with Custom Configurations

Use the Qualcomm® AI Hub Models Python library to compile and export the model with your own:

  • Custom weights (e.g., fine-tuned checkpoints)
  • Custom input shapes
  • Target device and runtime configurations

This option is ideal if you need to customize the model beyond the default configuration provided here.

See our repository for Whisper-Small-Quantized on GitHub for usage instructions.

Model Details

Model Type: Model_use_case.speech_recognition

Model Stats:

  • Model checkpoint: openai/whisper-small
  • Input resolution: 80x3000 (30 seconds audio)
  • Max decoded sequence length: 200 tokens

Performance Summary

Model Runtime Precision Chipset Inference Time (ms) Peak Memory Range (MB) Primary Compute Unit
decoder PRECOMPILED_QNN_ONNX w8a16 Snapdragon® X2 Elite 3.926 ms 33 - 33 MB NPU
decoder PRECOMPILED_QNN_ONNX w8a16 Snapdragon® X Elite 7.416 ms 187 - 187 MB NPU
decoder PRECOMPILED_QNN_ONNX w8a16 Snapdragon® 8 Gen 3 Mobile 6.162 ms 26 - 34 MB NPU
decoder PRECOMPILED_QNN_ONNX w8a16 Snapdragon® 8 Gen 1 Mobile 10.099 ms 37 - 52 MB NPU
decoder PRECOMPILED_QNN_ONNX w8a16 Qualcomm® QCS8550 (Proxy) 7.886 ms 28 - 29 MB NPU
decoder PRECOMPILED_QNN_ONNX w8a16 Qualcomm® QCS8450 10.099 ms 37 - 52 MB NPU
decoder PRECOMPILED_QNN_ONNX w8a16 Qualcomm® QCM6690 32.377 ms 28 - 37 MB NPU
decoder PRECOMPILED_QNN_ONNX w8a16 Qualcomm® QCS9075 8.86 ms 25 - 58 MB NPU
decoder PRECOMPILED_QNN_ONNX w8a16 Snapdragon® 7 Gen 4 Mobile 10.575 ms 28 - 34 MB NPU
decoder PRECOMPILED_QNN_ONNX w8a16 Snapdragon® 8 Elite Mobile 4.689 ms 28 - 40 MB NPU
decoder PRECOMPILED_QNN_ONNX w8a16 Snapdragon® 8 Elite Gen 5 Mobile 3.985 ms 23 - 31 MB NPU
decoder PRECOMPILED_QNN_ONNX w8a16 Qualcomm® QCS7790 10.575 ms 28 - 34 MB NPU
decoder PRECOMPILED_QNN_ONNX w8a16 Qualcomm® QCS8750 4.689 ms 28 - 40 MB NPU
decoder PRECOMPILED_QNN_ONNX w8a16 Qualcomm® QCS7181 7.416 ms 187 - 187 MB NPU
decoder QNN_CONTEXT_BINARY w8a16 Snapdragon® X2 Elite 4.169 ms 30 - 30 MB NPU
decoder QNN_CONTEXT_BINARY w8a16 Snapdragon® X Elite 7.271 ms 30 - 30 MB NPU
decoder QNN_CONTEXT_BINARY w8a16 Snapdragon® 8 Gen 3 Mobile 5.981 ms 30 - 39 MB NPU
decoder QNN_CONTEXT_BINARY w8a16 Qualcomm® QCS8275 12.513 ms 30 - 39 MB NPU
decoder QNN_CONTEXT_BINARY w8a16 Qualcomm® QCS8550 (Proxy) 7.856 ms 30 - 32 MB NPU
decoder QNN_CONTEXT_BINARY w8a16 Qualcomm® QCM6690 30.591 ms 30 - 37 MB NPU
decoder QNN_CONTEXT_BINARY w8a16 Qualcomm® QCS9075 8.713 ms 25 - 60 MB NPU
decoder QNN_CONTEXT_BINARY w8a16 Snapdragon® 7 Gen 4 Mobile 10.4 ms 30 - 36 MB NPU
decoder QNN_CONTEXT_BINARY w8a16 Snapdragon® 8 Elite Mobile 4.645 ms 28 - 42 MB NPU
decoder QNN_CONTEXT_BINARY w8a16 Snapdragon® 8 Elite Gen 5 Mobile 3.938 ms 24 - 33 MB NPU
decoder QNN_CONTEXT_BINARY w8a16 Qualcomm® SA7255P 12.513 ms 30 - 39 MB NPU
decoder QNN_CONTEXT_BINARY w8a16 Qualcomm® QCS7790 10.4 ms 30 - 36 MB NPU
decoder QNN_CONTEXT_BINARY w8a16 Qualcomm® QCS8750 4.645 ms 28 - 42 MB NPU
decoder QNN_CONTEXT_BINARY w8a16 Qualcomm® QCS7181 7.271 ms 30 - 30 MB NPU
decoder VOICE_AI w8a16 Snapdragon® X2 Elite 4.316 ms 30 - 30 MB NPU
decoder VOICE_AI w8a16 Snapdragon® X Elite 7.277 ms 30 - 30 MB NPU
decoder VOICE_AI w8a16 Snapdragon® 8 Gen 3 Mobile 5.934 ms 30 - 37 MB NPU
decoder VOICE_AI w8a16 Qualcomm® QCS8275 12.725 ms 10 - 17 MB NPU
decoder VOICE_AI w8a16 Qualcomm® QCS8550 (Proxy) 7.926 ms 30 - 32 MB NPU
decoder VOICE_AI w8a16 Qualcomm® QCM6690 29.962 ms 30 - 37 MB NPU
decoder VOICE_AI w8a16 Qualcomm® QCS9075 8.662 ms 25 - 60 MB NPU
decoder VOICE_AI w8a16 Snapdragon® 7 Gen 4 Mobile 10.383 ms 30 - 37 MB NPU
decoder VOICE_AI w8a16 Snapdragon® 8 Elite Mobile 4.662 ms 30 - 43 MB NPU
decoder VOICE_AI w8a16 Snapdragon® 8 Elite Gen 5 Mobile 3.918 ms 24 - 33 MB NPU
decoder VOICE_AI w8a16 Qualcomm® SA7255P 12.725 ms 10 - 17 MB NPU
decoder VOICE_AI w8a16 Qualcomm® QCS7790 10.383 ms 30 - 37 MB NPU
decoder VOICE_AI w8a16 Qualcomm® QCS8750 4.662 ms 30 - 43 MB NPU
decoder VOICE_AI w8a16 Qualcomm® QCS7181 7.277 ms 30 - 30 MB NPU
encoder PRECOMPILED_QNN_ONNX w8a16 Snapdragon® X2 Elite 157.34 ms 66 - 66 MB NPU
encoder PRECOMPILED_QNN_ONNX w8a16 Snapdragon® X Elite 301.972 ms 127 - 127 MB NPU
encoder PRECOMPILED_QNN_ONNX w8a16 Snapdragon® 8 Gen 3 Mobile 271.303 ms 58 - 65 MB NPU
encoder PRECOMPILED_QNN_ONNX w8a16 Snapdragon® 8 Gen 1 Mobile 306.977 ms 64 - 78 MB NPU
encoder PRECOMPILED_QNN_ONNX w8a16 Qualcomm® QCS8550 (Proxy) 372.687 ms 52 - 53 MB NPU
encoder PRECOMPILED_QNN_ONNX w8a16 Qualcomm® QCS8450 306.977 ms 64 - 78 MB NPU
encoder PRECOMPILED_QNN_ONNX w8a16 Qualcomm® QCM6690 4492.943 ms 2 - 13 MB NPU
encoder PRECOMPILED_QNN_ONNX w8a16 Qualcomm® QCS9075 298.137 ms 64 - 67 MB NPU
encoder PRECOMPILED_QNN_ONNX w8a16 Snapdragon® 7 Gen 4 Mobile 477.335 ms 48 - 58 MB NPU
encoder PRECOMPILED_QNN_ONNX w8a16 Snapdragon® 8 Elite Mobile 224.256 ms 64 - 75 MB NPU
encoder PRECOMPILED_QNN_ONNX w8a16 Snapdragon® 8 Elite Gen 5 Mobile 174.077 ms 64 - 78 MB NPU
encoder PRECOMPILED_QNN_ONNX w8a16 Qualcomm® QCS7790 477.335 ms 48 - 58 MB NPU
encoder PRECOMPILED_QNN_ONNX w8a16 Qualcomm® QCS8750 224.256 ms 64 - 75 MB NPU
encoder PRECOMPILED_QNN_ONNX w8a16 Qualcomm® QCS7181 301.972 ms 127 - 127 MB NPU
encoder QNN_CONTEXT_BINARY w8a16 Snapdragon® X2 Elite 157.161 ms 0 - 0 MB NPU
encoder QNN_CONTEXT_BINARY w8a16 Snapdragon® X Elite 302.31 ms 0 - 0 MB NPU
encoder QNN_CONTEXT_BINARY w8a16 Snapdragon® 8 Gen 3 Mobile 270.388 ms 3 - 10 MB NPU
encoder QNN_CONTEXT_BINARY w8a16 Qualcomm® QCS8275 518.157 ms 1 - 8 MB NPU
encoder QNN_CONTEXT_BINARY w8a16 Qualcomm® QCS8550 (Proxy) 376.895 ms 1 - 3 MB NPU
encoder QNN_CONTEXT_BINARY w8a16 Qualcomm® QCM6690 4449.38 ms 1 - 13 MB NPU
encoder QNN_CONTEXT_BINARY w8a16 Qualcomm® QCS9075 297.144 ms 0 - 29 MB NPU
encoder QNN_CONTEXT_BINARY w8a16 Snapdragon® 7 Gen 4 Mobile 490.096 ms 1 - 7 MB NPU
encoder QNN_CONTEXT_BINARY w8a16 Snapdragon® 8 Elite Mobile 222.448 ms 1 - 10 MB NPU
encoder QNN_CONTEXT_BINARY w8a16 Snapdragon® 8 Elite Gen 5 Mobile 180.871 ms 1 - 13 MB NPU
encoder QNN_CONTEXT_BINARY w8a16 Qualcomm® SA7255P 518.157 ms 1 - 8 MB NPU
encoder QNN_CONTEXT_BINARY w8a16 Qualcomm® QCS7790 490.096 ms 1 - 7 MB NPU
encoder QNN_CONTEXT_BINARY w8a16 Qualcomm® QCS8750 222.448 ms 1 - 10 MB NPU
encoder QNN_CONTEXT_BINARY w8a16 Qualcomm® QCS7181 302.31 ms 0 - 0 MB NPU
encoder VOICE_AI w8a16 Snapdragon® X2 Elite 156.726 ms 0 - 0 MB NPU
encoder VOICE_AI w8a16 Snapdragon® X Elite 305.709 ms 0 - 0 MB NPU
encoder VOICE_AI w8a16 Snapdragon® 8 Gen 3 Mobile 268.687 ms 1 - 8 MB NPU
encoder VOICE_AI w8a16 Qualcomm® QCS8275 516.993 ms 1 - 9 MB NPU
encoder VOICE_AI w8a16 Qualcomm® QCS8550 (Proxy) 364.857 ms 1 - 3 MB NPU
encoder VOICE_AI w8a16 Qualcomm® QCM6690 4431.688 ms 1 - 13 MB NPU
encoder VOICE_AI w8a16 Qualcomm® QCS9075 297.037 ms 0 - 29 MB NPU
encoder VOICE_AI w8a16 Snapdragon® 7 Gen 4 Mobile 485.403 ms 1 - 7 MB NPU
encoder VOICE_AI w8a16 Snapdragon® 8 Elite Mobile 228.111 ms 1 - 10 MB NPU
encoder VOICE_AI w8a16 Snapdragon® 8 Elite Gen 5 Mobile 174.642 ms 1 - 11 MB NPU
encoder VOICE_AI w8a16 Qualcomm® SA7255P 516.993 ms 1 - 9 MB NPU
encoder VOICE_AI w8a16 Qualcomm® QCS7790 485.403 ms 1 - 7 MB NPU
encoder VOICE_AI w8a16 Qualcomm® QCS8750 228.111 ms 1 - 10 MB NPU
encoder VOICE_AI w8a16 Qualcomm® QCS7181 305.709 ms 0 - 0 MB NPU

License

  • The license for the original implementation of Whisper-Small-Quantized can be found here.

References

Community

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
Automatic Speech Recognition
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support