Hugging Face's logo

qualcomm
/
Whisper-Tiny-En

Automatic Speech Recognition
PyTorch
LiteRT
ONNX
foundation
android
Model card Files Community
2

Whisper-Tiny-En: Optimized for Mobile Deployment

Automatic speech recognition (ASR) model for English transcription as well as translation

OpenAI’s Whisper ASR (Automatic Speech Recognition) model is a state-of-the-art system designed for transcribing spoken language into written text. 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 mean decoded length specified below.

This model is an implementation of Whisper-Tiny-En found here.

This repository provides scripts to run Whisper-Tiny-En on Qualcomm® devices. More details on model performance across various devices, can be found here.

Model Details

  • Model Type: Model_use_case.speech_recognition
  • Model Stats:
    • Model checkpoint: tiny.en
    • Input resolution: 80x3000 (30 seconds audio)
    • Mean decoded sequence length: 112 tokens
    • Number of parameters (WhisperEncoder): 9.39M
    • Model size (WhisperEncoder): 35.9 MB
    • Number of parameters (WhisperDecoder): 28.2M
    • Model size (WhisperDecoder): 108 MB
Model Precision Device Chipset Target Runtime Inference Time (ms) Peak Memory Range (MB) Primary Compute Unit Target Model
WhisperEncoderInf float QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) TFLITE 399.974 ms 19 - 41 MB GPU Whisper-Tiny-En.tflite
WhisperEncoderInf float QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) QNN_DLC 106.541 ms 0 - 269 MB NPU Whisper-Tiny-En.dlc
WhisperEncoderInf float QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) TFLITE 128.571 ms 20 - 61 MB GPU Whisper-Tiny-En.tflite
WhisperEncoderInf float QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) QNN_DLC 114.494 ms 1 - 249 MB NPU Whisper-Tiny-En.dlc
WhisperEncoderInf float QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) TFLITE 100.558 ms 3 - 41 MB GPU Whisper-Tiny-En.tflite
WhisperEncoderInf float QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) QNN_DLC 51.037 ms 1 - 85 MB NPU Whisper-Tiny-En.dlc
WhisperEncoderInf float QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) TFLITE 168.84 ms 20 - 42 MB GPU Whisper-Tiny-En.tflite
WhisperEncoderInf float QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) QNN_DLC 54.084 ms 1 - 270 MB NPU Whisper-Tiny-En.dlc
WhisperEncoderInf float SA7255P ADP Qualcomm® SA7255P TFLITE 399.974 ms 19 - 41 MB GPU Whisper-Tiny-En.tflite
WhisperEncoderInf float SA7255P ADP Qualcomm® SA7255P QNN_DLC 106.541 ms 0 - 269 MB NPU Whisper-Tiny-En.dlc
WhisperEncoderInf float SA8255 (Proxy) Qualcomm® SA8255P (Proxy) TFLITE 100.247 ms 20 - 68 MB GPU Whisper-Tiny-En.tflite
WhisperEncoderInf float SA8255 (Proxy) Qualcomm® SA8255P (Proxy) QNN_DLC 53.614 ms 0 - 86 MB NPU Whisper-Tiny-En.dlc
WhisperEncoderInf float SA8295P ADP Qualcomm® SA8295P TFLITE 103.495 ms 20 - 50 MB GPU Whisper-Tiny-En.tflite
WhisperEncoderInf float SA8295P ADP Qualcomm® SA8295P QNN_DLC 81.281 ms 1 - 252 MB NPU Whisper-Tiny-En.dlc
WhisperEncoderInf float SA8650 (Proxy) Qualcomm® SA8650P (Proxy) TFLITE 102.579 ms 20 - 69 MB GPU Whisper-Tiny-En.tflite
WhisperEncoderInf float SA8650 (Proxy) Qualcomm® SA8650P (Proxy) QNN_DLC 52.042 ms 0 - 88 MB NPU Whisper-Tiny-En.dlc
WhisperEncoderInf float SA8775P ADP Qualcomm® SA8775P TFLITE 168.84 ms 20 - 42 MB GPU Whisper-Tiny-En.tflite
WhisperEncoderInf float SA8775P ADP Qualcomm® SA8775P QNN_DLC 54.084 ms 1 - 270 MB NPU Whisper-Tiny-En.dlc
WhisperEncoderInf float Samsung Galaxy S23 Snapdragon® 8 Gen 2 Mobile TFLITE 101.124 ms 3 - 46 MB GPU Whisper-Tiny-En.tflite
WhisperEncoderInf float Samsung Galaxy S23 Snapdragon® 8 Gen 2 Mobile QNN_DLC 53.969 ms 0 - 85 MB NPU Whisper-Tiny-En.dlc
WhisperEncoderInf float Samsung Galaxy S23 Snapdragon® 8 Gen 2 Mobile ONNX 62.752 ms 36 - 152 MB NPU Whisper-Tiny-En.onnx
WhisperEncoderInf float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile TFLITE 79.074 ms 18 - 54 MB GPU Whisper-Tiny-En.tflite
WhisperEncoderInf float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile QNN_DLC 43.62 ms 1 - 277 MB NPU Whisper-Tiny-En.dlc
WhisperEncoderInf float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile ONNX 51.705 ms 40 - 443 MB NPU Whisper-Tiny-En.onnx
WhisperEncoderInf float Snapdragon 8 Elite QRD Snapdragon® 8 Elite Mobile TFLITE 67.215 ms 20 - 46 MB GPU Whisper-Tiny-En.tflite
WhisperEncoderInf float Snapdragon 8 Elite QRD Snapdragon® 8 Elite Mobile QNN_DLC 34.396 ms 1 - 265 MB NPU Whisper-Tiny-En.dlc
WhisperEncoderInf float Snapdragon 8 Elite QRD Snapdragon® 8 Elite Mobile ONNX 43.803 ms 39 - 406 MB NPU Whisper-Tiny-En.onnx
WhisperEncoderInf float Snapdragon X Elite CRD Snapdragon® X Elite QNN_DLC 51.039 ms 31 - 31 MB NPU Whisper-Tiny-En.dlc
WhisperEncoderInf float Snapdragon X Elite CRD Snapdragon® X Elite ONNX 66.012 ms 66 - 66 MB NPU Whisper-Tiny-En.onnx
WhisperDecoderInf float QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) TFLITE 6.417 ms 3 - 96 MB NPU Whisper-Tiny-En.tflite
WhisperDecoderInf float QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) QNN_DLC 5.893 ms 17 - 90 MB NPU Whisper-Tiny-En.dlc
WhisperDecoderInf float QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) TFLITE 4.175 ms 3 - 93 MB NPU Whisper-Tiny-En.tflite
WhisperDecoderInf float QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) QNN_DLC 5.992 ms 3 - 77 MB NPU Whisper-Tiny-En.dlc
WhisperDecoderInf float QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) TFLITE 3.647 ms 3 - 50 MB NPU Whisper-Tiny-En.tflite
WhisperDecoderInf float QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) QNN_DLC 3.447 ms 19 - 41 MB NPU Whisper-Tiny-En.dlc
WhisperDecoderInf float QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) TFLITE 4.497 ms 3 - 97 MB NPU Whisper-Tiny-En.tflite
WhisperDecoderInf float QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) QNN_DLC 4.362 ms 19 - 93 MB NPU Whisper-Tiny-En.dlc
WhisperDecoderInf float SA7255P ADP Qualcomm® SA7255P TFLITE 6.417 ms 3 - 96 MB NPU Whisper-Tiny-En.tflite
WhisperDecoderInf float SA7255P ADP Qualcomm® SA7255P QNN_DLC 5.893 ms 17 - 90 MB NPU Whisper-Tiny-En.dlc
WhisperDecoderInf float SA8255 (Proxy) Qualcomm® SA8255P (Proxy) TFLITE 3.601 ms 3 - 50 MB NPU Whisper-Tiny-En.tflite
WhisperDecoderInf float SA8255 (Proxy) Qualcomm® SA8255P (Proxy) QNN_DLC 3.457 ms 20 - 42 MB NPU Whisper-Tiny-En.dlc
WhisperDecoderInf float SA8295P ADP Qualcomm® SA8295P TFLITE 4.747 ms 3 - 88 MB NPU Whisper-Tiny-En.tflite
WhisperDecoderInf float SA8295P ADP Qualcomm® SA8295P QNN_DLC 4.41 ms 3 - 70 MB NPU Whisper-Tiny-En.dlc
WhisperDecoderInf float SA8650 (Proxy) Qualcomm® SA8650P (Proxy) TFLITE 3.627 ms 3 - 55 MB NPU Whisper-Tiny-En.tflite
WhisperDecoderInf float SA8650 (Proxy) Qualcomm® SA8650P (Proxy) QNN_DLC 3.498 ms 20 - 43 MB NPU Whisper-Tiny-En.dlc
WhisperDecoderInf float SA8775P ADP Qualcomm® SA8775P TFLITE 4.497 ms 3 - 97 MB NPU Whisper-Tiny-En.tflite
WhisperDecoderInf float SA8775P ADP Qualcomm® SA8775P QNN_DLC 4.362 ms 19 - 93 MB NPU Whisper-Tiny-En.dlc
WhisperDecoderInf float Samsung Galaxy S23 Snapdragon® 8 Gen 2 Mobile TFLITE 3.645 ms 3 - 43 MB NPU Whisper-Tiny-En.tflite
WhisperDecoderInf float Samsung Galaxy S23 Snapdragon® 8 Gen 2 Mobile QNN_DLC 3.423 ms 20 - 43 MB NPU Whisper-Tiny-En.dlc
WhisperDecoderInf float Samsung Galaxy S23 Snapdragon® 8 Gen 2 Mobile ONNX 4.102 ms 4 - 176 MB NPU Whisper-Tiny-En.onnx
WhisperDecoderInf float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile TFLITE 2.762 ms 0 - 103 MB NPU Whisper-Tiny-En.tflite
WhisperDecoderInf float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile QNN_DLC 2.861 ms 0 - 83 MB NPU Whisper-Tiny-En.dlc
WhisperDecoderInf float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile ONNX 3.372 ms 26 - 110 MB NPU Whisper-Tiny-En.onnx
WhisperDecoderInf float Snapdragon 8 Elite QRD Snapdragon® 8 Elite Mobile TFLITE 2.568 ms 0 - 90 MB NPU Whisper-Tiny-En.tflite
WhisperDecoderInf float Snapdragon 8 Elite QRD Snapdragon® 8 Elite Mobile QNN_DLC 2.383 ms 6 - 84 MB NPU Whisper-Tiny-En.dlc
WhisperDecoderInf float Snapdragon 8 Elite QRD Snapdragon® 8 Elite Mobile ONNX 2.606 ms 0 - 190 MB NPU Whisper-Tiny-En.onnx
WhisperDecoderInf float Snapdragon X Elite CRD Snapdragon® X Elite QNN_DLC 3.309 ms 200 - 200 MB NPU Whisper-Tiny-En.dlc
WhisperDecoderInf float Snapdragon X Elite CRD Snapdragon® X Elite ONNX 3.318 ms 73 - 73 MB NPU Whisper-Tiny-En.onnx

Installation

Install the package via pip:

pip install "qai-hub-models[whisper-tiny-en]"

Configure Qualcomm® AI Hub to run this model on a cloud-hosted device

Sign-in to Qualcomm® AI Hub 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.

qai-hub configure --api_token API_TOKEN

Navigate to 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.

python -m qai_hub_models.models.whisper_tiny_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_tiny_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.
python -m qai_hub_models.models.whisper_tiny_en.export

Profiling Results
------------------------------------------------------------
WhisperEncoderInf
Device                          : cs_8275 (ANDROID 14)                  
Runtime                         : TFLITE                                
Estimated inference time (ms)   : 400.0                                 
Estimated peak memory usage (MB): [19, 41]                              
Total # Ops                     : 271                                   
Compute Unit(s)                 : npu (0 ops) gpu (260 ops) cpu (11 ops)

------------------------------------------------------------
WhisperDecoderInf
Device                          : cs_8275 (ANDROID 14)                 
Runtime                         : TFLITE                               
Estimated inference time (ms)   : 6.4                                  
Estimated peak memory usage (MB): [3, 96]                              
Total # Ops                     : 557                                  
Compute Unit(s)                 : npu (557 ops) gpu (0 ops) cpu (0 ops)

How does this work?

This export script leverages Qualcomm® AI Hub 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.

import torch

import qai_hub as hub
from qai_hub_models.models.whisper_tiny_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.

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.

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.

Deploying compiled model to Android

The models can be deployed using multiple runtimes:

  • TensorFlow Lite (.tflite export): This tutorial provides a guide to deploy the .tflite model in an Android application.

  • QNN (.so export ): This sample app provides instructions on how to use the .so shared library in an Android application.

View on Qualcomm® AI Hub

Get more details on Whisper-Tiny-En's performance across various devices here. Explore all available models on Qualcomm® AI Hub

License

  • The license for the original implementation of Whisper-Tiny-En can be found here.
  • The license for the compiled assets for on-device deployment can be found here

References

Community

Downloads last month
250
Inference Providers NEW
Automatic Speech Recognition
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support