--- library_name: pytorch license: other tags: - foundation - android pipeline_tag: automatic-speech-recognition --- ![](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/whisper_small_quantized/web-assets/model_demo.png) # Whisper-Small-Quantized: Optimized for Mobile Deployment ## Transformer-based automatic speech recognition (ASR) model for multilingual transcription and translation available on HuggingFace 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 model is an implementation of Whisper-Small-Quantized found [here](https://github.com/huggingface/transformers/tree/v4.42.3/src/transformers/models/whisper). This repository provides scripts to run Whisper-Small-Quantized on Qualcomm® devices. More details on model performance across various devices, can be found [here](https://aihub.qualcomm.com/models/whisper_small_quantized). ### 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 | Model | Precision | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit | Target Model |---|---|---|---|---|---|---|---|---| | WhisperSmallEncoderQuantizable | w8a16 | Dragonwing RB3 Gen 2 Vision Kit | Qualcomm® QCS6490 | PRECOMPILED_QNN_ONNX | 537.889 ms | 30 - 33 MB | NPU | Use Export Script | | WhisperSmallEncoderQuantizable | w8a16 | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN_CONTEXT_BINARY | 458.786 ms | 1 - 10 MB | NPU | Use Export Script | | WhisperSmallEncoderQuantizable | w8a16 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_CONTEXT_BINARY | 353.084 ms | 1 - 3 MB | NPU | Use Export Script | | WhisperSmallEncoderQuantizable | w8a16 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | PRECOMPILED_QNN_ONNX | 62.712 ms | 0 - 113 MB | NPU | Use Export Script | | WhisperSmallEncoderQuantizable | w8a16 | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_CONTEXT_BINARY | 267.301 ms | 1 - 10 MB | NPU | Use Export Script | | WhisperSmallEncoderQuantizable | w8a16 | SA7255P ADP | Qualcomm® SA7255P | QNN_CONTEXT_BINARY | 458.786 ms | 1 - 10 MB | NPU | Use Export Script | | WhisperSmallEncoderQuantizable | w8a16 | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | QNN_CONTEXT_BINARY | 335.324 ms | 1 - 3 MB | NPU | Use Export Script | | WhisperSmallEncoderQuantizable | w8a16 | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | QNN_CONTEXT_BINARY | 330.105 ms | 1 - 2 MB | NPU | Use Export Script | | WhisperSmallEncoderQuantizable | w8a16 | SA8775P ADP | Qualcomm® SA8775P | QNN_CONTEXT_BINARY | 267.301 ms | 1 - 10 MB | NPU | Use Export Script | | WhisperSmallEncoderQuantizable | w8a16 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_CONTEXT_BINARY | 250.011 ms | 1 - 19 MB | NPU | Use Export Script | | WhisperSmallEncoderQuantizable | w8a16 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | PRECOMPILED_QNN_ONNX | 44.505 ms | 64 - 82 MB | NPU | Use Export Script | | WhisperSmallEncoderQuantizable | w8a16 | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | QNN_CONTEXT_BINARY | 202.875 ms | 1 - 14 MB | NPU | Use Export Script | | WhisperSmallEncoderQuantizable | w8a16 | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | PRECOMPILED_QNN_ONNX | 35.201 ms | 63 - 78 MB | NPU | Use Export Script | | WhisperSmallEncoderQuantizable | w8a16 | Snapdragon 7 Gen 4 QRD | Snapdragon® 7 Gen 4 Mobile | QNN_CONTEXT_BINARY | 526.452 ms | 3 - 17 MB | NPU | Use Export Script | | WhisperSmallEncoderQuantizable | w8a16 | Snapdragon 7 Gen 4 QRD | Snapdragon® 7 Gen 4 Mobile | PRECOMPILED_QNN_ONNX | 183.764 ms | 63 - 77 MB | NPU | Use Export Script | | WhisperSmallEncoderQuantizable | w8a16 | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | QNN_CONTEXT_BINARY | 179.19 ms | 1 - 11 MB | NPU | Use Export Script | | WhisperSmallEncoderQuantizable | w8a16 | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | PRECOMPILED_QNN_ONNX | 29.961 ms | 62 - 73 MB | NPU | Use Export Script | | WhisperSmallEncoderQuantizable | w8a16 | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_CONTEXT_BINARY | 243.159 ms | 0 - 0 MB | NPU | Use Export Script | | WhisperSmallEncoderQuantizable | w8a16 | Snapdragon X Elite CRD | Snapdragon® X Elite | PRECOMPILED_QNN_ONNX | 61.68 ms | 107 - 107 MB | NPU | Use Export Script | | WhisperSmallDecoderQuantizable | w8a16 | Dragonwing RB3 Gen 2 Vision Kit | Qualcomm® QCS6490 | PRECOMPILED_QNN_ONNX | 32.659 ms | 27 - 59 MB | NPU | Use Export Script | | WhisperSmallDecoderQuantizable | w8a16 | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN_CONTEXT_BINARY | 13.485 ms | 29 - 39 MB | NPU | Use Export Script | | WhisperSmallDecoderQuantizable | w8a16 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_CONTEXT_BINARY | 8.416 ms | 27 - 29 MB | NPU | Use Export Script | | WhisperSmallDecoderQuantizable | w8a16 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | PRECOMPILED_QNN_ONNX | 8.737 ms | 0 - 193 MB | NPU | Use Export Script | | WhisperSmallDecoderQuantizable | w8a16 | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_CONTEXT_BINARY | 9.562 ms | 27 - 37 MB | NPU | Use Export Script | | WhisperSmallDecoderQuantizable | w8a16 | SA7255P ADP | Qualcomm® SA7255P | QNN_CONTEXT_BINARY | 13.485 ms | 29 - 39 MB | NPU | Use Export Script | | WhisperSmallDecoderQuantizable | w8a16 | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | QNN_CONTEXT_BINARY | 8.495 ms | 27 - 29 MB | NPU | Use Export Script | | WhisperSmallDecoderQuantizable | w8a16 | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | QNN_CONTEXT_BINARY | 8.279 ms | 34 - 37 MB | NPU | Use Export Script | | WhisperSmallDecoderQuantizable | w8a16 | SA8775P ADP | Qualcomm® SA8775P | QNN_CONTEXT_BINARY | 9.562 ms | 27 - 37 MB | NPU | Use Export Script | | WhisperSmallDecoderQuantizable | w8a16 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_CONTEXT_BINARY | 6.409 ms | 30 - 49 MB | NPU | Use Export Script | | WhisperSmallDecoderQuantizable | w8a16 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | PRECOMPILED_QNN_ONNX | 6.785 ms | 38 - 57 MB | NPU | Use Export Script | | WhisperSmallDecoderQuantizable | w8a16 | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | QNN_CONTEXT_BINARY | 4.763 ms | 21 - 38 MB | NPU | Use Export Script | | WhisperSmallDecoderQuantizable | w8a16 | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | PRECOMPILED_QNN_ONNX | 5.095 ms | 26 - 39 MB | NPU | Use Export Script | | WhisperSmallDecoderQuantizable | w8a16 | Snapdragon 7 Gen 4 QRD | Snapdragon® 7 Gen 4 Mobile | QNN_CONTEXT_BINARY | 11.037 ms | 30 - 45 MB | NPU | Use Export Script | | WhisperSmallDecoderQuantizable | w8a16 | Snapdragon 7 Gen 4 QRD | Snapdragon® 7 Gen 4 Mobile | PRECOMPILED_QNN_ONNX | 11.94 ms | 30 - 43 MB | NPU | Use Export Script | | WhisperSmallDecoderQuantizable | w8a16 | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | QNN_CONTEXT_BINARY | 4.012 ms | 30 - 42 MB | NPU | Use Export Script | | WhisperSmallDecoderQuantizable | w8a16 | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | PRECOMPILED_QNN_ONNX | 4.368 ms | 37 - 48 MB | NPU | Use Export Script | | WhisperSmallDecoderQuantizable | w8a16 | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_CONTEXT_BINARY | 7.725 ms | 30 - 30 MB | NPU | Use Export Script | | WhisperSmallDecoderQuantizable | w8a16 | Snapdragon X Elite CRD | Snapdragon® X Elite | PRECOMPILED_QNN_ONNX | 7.841 ms | 186 - 186 MB | NPU | Use Export Script | ## Installation Install the package via pip: ```bash # NOTE: 3.10 <= PYTHON_VERSION < 3.14 is supported. pip install "qai-hub-models[whisper-small-quantized]" ``` ## Configure Qualcomm® AI Hub Workbench to run this model on a cloud-hosted device Sign-in to [Qualcomm® AI Hub Workbench](https://workbench.aihub.qualcomm.com/) with your Qualcomm® ID. Once signed in navigate to `Account -> Settings -> API Token`. With this API token, you can configure your client to run models on the cloud hosted devices. ```bash qai-hub configure --api_token API_TOKEN ``` Navigate to [docs](https://workbench.aihub.qualcomm.com/docs/) for more information. ## Demo off target The package contains a simple end-to-end demo that downloads pre-trained weights and runs this model on a sample input. ```bash python -m qai_hub_models.models.whisper_small_quantized.demo ``` The above demo runs a reference implementation of pre-processing, model inference, and post processing. **NOTE**: If you want running in a Jupyter Notebook or Google Colab like environment, please add the following to your cell (instead of the above). ``` %run -m qai_hub_models.models.whisper_small_quantized.demo ``` ### Run model on a cloud-hosted device In addition to the demo, you can also run the model on a cloud-hosted Qualcomm® device. This script does the following: * Performance check on-device on a cloud-hosted device * Downloads compiled assets that can be deployed on-device for Android. * Accuracy check between PyTorch and on-device outputs. ```bash python -m qai_hub_models.models.whisper_small_quantized.export ``` ## Deploying compiled model to Android The models can be deployed using multiple runtimes: - TensorFlow Lite (`.tflite` export): [This tutorial](https://www.tensorflow.org/lite/android/quickstart) provides a guide to deploy the .tflite model in an Android application. - QNN (`.so` export ): This [sample app](https://docs.qualcomm.com/bundle/publicresource/topics/80-63442-50/sample_app.html) provides instructions on how to use the `.so` shared library in an Android application. ## View on Qualcomm® AI Hub Get more details on Whisper-Small-Quantized's performance across various devices [here](https://aihub.qualcomm.com/models/whisper_small_quantized). Explore all available models on [Qualcomm® AI Hub](https://aihub.qualcomm.com/) ## License * The license for the original implementation of Whisper-Small-Quantized can be found [here](https://github.com/huggingface/transformers/blob/v4.42.3/LICENSE). * The license for the compiled assets for on-device deployment can be found [here](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/Qualcomm+AI+Hub+Proprietary+License.pdf) ## References * [Robust Speech Recognition via Large-Scale Weak Supervision](https://cdn.openai.com/papers/whisper.pdf) * [Source Model Implementation](https://github.com/huggingface/transformers/tree/v4.42.3/src/transformers/models/whisper) ## Community * Join [our AI Hub Slack community](https://aihub.qualcomm.com/community/slack) to collaborate, post questions and learn more about on-device AI. * For questions or feedback please [reach out to us](mailto:ai-hub-support@qti.qualcomm.com).