--- 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_en/web-assets/model_demo.png) # Whisper-Small-En: Optimized for Mobile Deployment ## Transformer-based automatic speech recognition (ASR) model for multilingual transcription and translation available on HuggingFace 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-En found [here](https://github.com/huggingface/transformers/tree/v4.42.3/src/transformers/models/whisper). This repository provides scripts to run Whisper-Small-En on Qualcomm® devices. More details on model performance across various devices, can be found [here](https://aihub.qualcomm.com/models/whisper_small_en). ### 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 - Number of parameters (HfWhisperEncoder): 102M - Model size (HfWhisperEncoder) (float): 391 MB - Number of parameters (HfWhisperDecoder): 139M - Model size (HfWhisperDecoder) (float): 533 MB | Model | Precision | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit | Target Model |---|---|---|---|---|---|---|---|---| | HfWhisperEncoder | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE | 3333.928 ms | 109 - 159 MB | GPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) | | HfWhisperEncoder | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN_DLC | 424.425 ms | 0 - 443 MB | NPU | [Whisper-Small-En.dlc](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.dlc) | | HfWhisperEncoder | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 1115.67 ms | 18 - 223 MB | GPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) | | HfWhisperEncoder | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_DLC | 135.465 ms | 1 - 28 MB | NPU | [Whisper-Small-En.dlc](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.dlc) | | HfWhisperEncoder | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 1416.508 ms | 106 - 155 MB | GPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) | | HfWhisperEncoder | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_DLC | 154.567 ms | 0 - 442 MB | NPU | [Whisper-Small-En.dlc](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.dlc) | | HfWhisperEncoder | float | SA7255P ADP | Qualcomm® SA7255P | TFLITE | 3333.928 ms | 109 - 159 MB | GPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) | | HfWhisperEncoder | float | SA7255P ADP | Qualcomm® SA7255P | QNN_DLC | 424.425 ms | 0 - 443 MB | NPU | [Whisper-Small-En.dlc](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.dlc) | | HfWhisperEncoder | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | TFLITE | 882.114 ms | 18 - 186 MB | GPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) | | HfWhisperEncoder | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | QNN_DLC | 137.649 ms | 0 - 33 MB | NPU | [Whisper-Small-En.dlc](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.dlc) | | HfWhisperEncoder | float | SA8295P ADP | Qualcomm® SA8295P | TFLITE | 798.622 ms | 109 - 158 MB | GPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) | | HfWhisperEncoder | float | SA8295P ADP | Qualcomm® SA8295P | QNN_DLC | 241.302 ms | 0 - 440 MB | NPU | [Whisper-Small-En.dlc](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.dlc) | | HfWhisperEncoder | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | TFLITE | 768.865 ms | 18 - 222 MB | GPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) | | HfWhisperEncoder | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | QNN_DLC | 138.695 ms | 0 - 32 MB | NPU | [Whisper-Small-En.dlc](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.dlc) | | HfWhisperEncoder | float | SA8775P ADP | Qualcomm® SA8775P | TFLITE | 1416.508 ms | 106 - 155 MB | GPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) | | HfWhisperEncoder | float | SA8775P ADP | Qualcomm® SA8775P | QNN_DLC | 154.567 ms | 0 - 442 MB | NPU | [Whisper-Small-En.dlc](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.dlc) | | HfWhisperEncoder | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | TFLITE | 775.503 ms | 110 - 130 MB | GPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) | | HfWhisperEncoder | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | QNN_DLC | 137.037 ms | 0 - 30 MB | NPU | [Whisper-Small-En.dlc](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.dlc) | | HfWhisperEncoder | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 569.222 ms | 111 - 300 MB | GPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) | | HfWhisperEncoder | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_DLC | 106.478 ms | 1 - 449 MB | NPU | [Whisper-Small-En.dlc](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.dlc) | | HfWhisperEncoder | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | TFLITE | 622.531 ms | 110 - 156 MB | GPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) | | HfWhisperEncoder | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | QNN_DLC | 87.001 ms | 0 - 426 MB | NPU | [Whisper-Small-En.dlc](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.dlc) | | HfWhisperEncoder | float | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC | 130.938 ms | 158 - 158 MB | NPU | [Whisper-Small-En.dlc](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.dlc) | | HfWhisperDecoder | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE | 28.37 ms | 14 - 506 MB | NPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) | | HfWhisperDecoder | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN_DLC | 18.043 ms | 38 - 335 MB | NPU | [Whisper-Small-En.dlc](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.dlc) | | HfWhisperDecoder | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 19.146 ms | 14 - 59 MB | NPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) | | HfWhisperDecoder | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_DLC | 12.048 ms | 60 - 74 MB | NPU | [Whisper-Small-En.dlc](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.dlc) | | HfWhisperDecoder | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 21.142 ms | 14 - 505 MB | NPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) | | HfWhisperDecoder | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_DLC | 13.304 ms | 55 - 346 MB | NPU | [Whisper-Small-En.dlc](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.dlc) | | HfWhisperDecoder | float | SA7255P ADP | Qualcomm® SA7255P | TFLITE | 28.37 ms | 14 - 506 MB | NPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) | | HfWhisperDecoder | float | SA7255P ADP | Qualcomm® SA7255P | QNN_DLC | 18.043 ms | 38 - 335 MB | NPU | [Whisper-Small-En.dlc](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.dlc) | | HfWhisperDecoder | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | TFLITE | 19.112 ms | 14 - 40 MB | NPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) | | HfWhisperDecoder | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | QNN_DLC | 11.743 ms | 57 - 83 MB | NPU | [Whisper-Small-En.dlc](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.dlc) | | HfWhisperDecoder | float | SA8295P ADP | Qualcomm® SA8295P | TFLITE | 20.478 ms | 14 - 429 MB | NPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) | | HfWhisperDecoder | float | SA8295P ADP | Qualcomm® SA8295P | QNN_DLC | 14.571 ms | 47 - 328 MB | NPU | [Whisper-Small-En.dlc](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.dlc) | | HfWhisperDecoder | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | TFLITE | 19.092 ms | 0 - 57 MB | NPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) | | HfWhisperDecoder | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | QNN_DLC | 11.972 ms | 43 - 67 MB | NPU | [Whisper-Small-En.dlc](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.dlc) | | HfWhisperDecoder | float | SA8775P ADP | Qualcomm® SA8775P | TFLITE | 21.142 ms | 14 - 505 MB | NPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) | | HfWhisperDecoder | float | SA8775P ADP | Qualcomm® SA8775P | QNN_DLC | 13.304 ms | 55 - 346 MB | NPU | [Whisper-Small-En.dlc](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.dlc) | | HfWhisperDecoder | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | TFLITE | 18.961 ms | 14 - 66 MB | NPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) | | HfWhisperDecoder | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | QNN_DLC | 11.942 ms | 60 - 85 MB | NPU | [Whisper-Small-En.dlc](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.dlc) | | HfWhisperDecoder | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 15.165 ms | 14 - 668 MB | NPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) | | HfWhisperDecoder | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_DLC | 9.579 ms | 0 - 308 MB | NPU | [Whisper-Small-En.dlc](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.dlc) | | HfWhisperDecoder | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | TFLITE | 13.094 ms | 14 - 459 MB | NPU | [Whisper-Small-En.tflite](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.tflite) | | HfWhisperDecoder | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | QNN_DLC | 8.145 ms | 58 - 169 MB | NPU | [Whisper-Small-En.dlc](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.dlc) | | HfWhisperDecoder | float | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC | 10.743 ms | 392 - 392 MB | NPU | [Whisper-Small-En.dlc](https://huggingface.co/qualcomm/Whisper-Small-En/blob/main/Whisper-Small-En.dlc) | ## Installation Install the package via pip: ```bash pip install "qai-hub-models[whisper-small-en]" ``` ## Configure Qualcomm® AI Hub to run this model on a cloud-hosted device Sign-in to [Qualcomm® AI Hub](https://app.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://app.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_en.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_en.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_en.export ``` ## How does this work? This [export script](https://aihub.qualcomm.com/models/whisper_small_en/qai_hub_models/models/Whisper-Small-En/export.py) leverages [Qualcomm® AI Hub](https://aihub.qualcomm.com/) to optimize, validate, and deploy this model on-device. Lets go through each step below in detail: Step 1: **Compile model for on-device deployment** To compile a PyTorch model for on-device deployment, we first trace the model in memory using the `jit.trace` and then call the `submit_compile_job` API. ```python import torch import qai_hub as hub from qai_hub_models.models.whisper_small_en import Model # Load the model torch_model = Model.from_pretrained() # Device device = hub.Device("Samsung Galaxy S24") # Trace model input_shape = torch_model.get_input_spec() sample_inputs = torch_model.sample_inputs() pt_model = torch.jit.trace(torch_model, [torch.tensor(data[0]) for _, data in sample_inputs.items()]) # Compile model on a specific device compile_job = hub.submit_compile_job( model=pt_model, device=device, input_specs=torch_model.get_input_spec(), ) # Get target model to run on-device target_model = compile_job.get_target_model() ``` Step 2: **Performance profiling on cloud-hosted device** After compiling models from step 1. Models can be profiled model on-device using the `target_model`. Note that this scripts runs the model on a device automatically provisioned in the cloud. Once the job is submitted, you can navigate to a provided job URL to view a variety of on-device performance metrics. ```python profile_job = hub.submit_profile_job( model=target_model, device=device, ) ``` Step 3: **Verify on-device accuracy** To verify the accuracy of the model on-device, you can run on-device inference on sample input data on the same cloud hosted device. ```python input_data = torch_model.sample_inputs() inference_job = hub.submit_inference_job( model=target_model, device=device, inputs=input_data, ) on_device_output = inference_job.download_output_data() ``` With the output of the model, you can compute like PSNR, relative errors or spot check the output with expected output. **Note**: This on-device profiling and inference requires access to Qualcomm® AI Hub. [Sign up for access](https://myaccount.qualcomm.com/signup). ## 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-En's performance across various devices [here](https://aihub.qualcomm.com/models/whisper_small_en). Explore all available models on [Qualcomm® AI Hub](https://aihub.qualcomm.com/) ## License * The license for the original implementation of Whisper-Small-En 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).