Hugging Face's logo

qualcomm
/
Whisper-Base-En

Automatic Speech Recognition
PyTorch
LiteRT
foundation
android
Model card Files Community
1

Whisper-Base-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-Base-En found here.

This repository provides scripts to run Whisper-Base-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: base.en
    • Input resolution: 80x3000 (30 seconds audio)
    • Mean decoded sequence length: 112 tokens
    • Number of parameters (WhisperEncoderInf): 23.7M
    • Model size (WhisperEncoderInf) (float): 90.6 MB
    • Number of parameters (WhisperDecoderInf): 48.8M
    • Model size (WhisperDecoderInf) (float): 186 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 839.092 ms 37 - 60 MB GPU Whisper-Base-En.tflite
WhisperEncoderInf float QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) QNN_DLC 315.193 ms 1 - 1399 MB NPU Whisper-Base-En.dlc
WhisperEncoderInf float QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) TFLITE 277.305 ms 38 - 89 MB GPU Whisper-Base-En.tflite
WhisperEncoderInf float QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) TFLITE 200.519 ms 0 - 84 MB GPU Whisper-Base-En.tflite
WhisperEncoderInf float QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) QNN_DLC 205.831 ms 0 - 358 MB NPU Whisper-Base-En.dlc
WhisperEncoderInf float QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) TFLITE 353.451 ms 38 - 62 MB GPU Whisper-Base-En.tflite
WhisperEncoderInf float QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) QNN_DLC 204.363 ms 0 - 1398 MB NPU Whisper-Base-En.dlc
WhisperEncoderInf float SA7255P ADP Qualcomm® SA7255P TFLITE 839.092 ms 37 - 60 MB GPU Whisper-Base-En.tflite
WhisperEncoderInf float SA7255P ADP Qualcomm® SA7255P QNN_DLC 315.193 ms 1 - 1399 MB NPU Whisper-Base-En.dlc
WhisperEncoderInf float SA8255 (Proxy) Qualcomm® SA8255P (Proxy) TFLITE 221.305 ms 0 - 69 MB GPU Whisper-Base-En.tflite
WhisperEncoderInf float SA8255 (Proxy) Qualcomm® SA8255P (Proxy) QNN_DLC 167.203 ms 0 - 355 MB NPU Whisper-Base-En.dlc
WhisperEncoderInf float SA8295P ADP Qualcomm® SA8295P TFLITE 189.893 ms 38 - 68 MB GPU Whisper-Base-En.tflite
WhisperEncoderInf float SA8295P ADP Qualcomm® SA8295P QNN_DLC 217.911 ms 1 - 1392 MB NPU Whisper-Base-En.dlc
WhisperEncoderInf float SA8650 (Proxy) Qualcomm® SA8650P (Proxy) TFLITE 203.378 ms 0 - 69 MB GPU Whisper-Base-En.tflite
WhisperEncoderInf float SA8650 (Proxy) Qualcomm® SA8650P (Proxy) QNN_DLC 161.769 ms 0 - 357 MB NPU Whisper-Base-En.dlc
WhisperEncoderInf float SA8775P ADP Qualcomm® SA8775P TFLITE 353.451 ms 38 - 62 MB GPU Whisper-Base-En.tflite
WhisperEncoderInf float SA8775P ADP Qualcomm® SA8775P QNN_DLC 204.363 ms 0 - 1398 MB NPU Whisper-Base-En.dlc
WhisperEncoderInf float Samsung Galaxy S23 Snapdragon® 8 Gen 2 Mobile TFLITE 201.716 ms 0 - 68 MB GPU Whisper-Base-En.tflite
WhisperEncoderInf float Samsung Galaxy S23 Snapdragon® 8 Gen 2 Mobile QNN_DLC 218.463 ms 0 - 355 MB NPU Whisper-Base-En.dlc
WhisperEncoderInf float Samsung Galaxy S23 Snapdragon® 8 Gen 2 Mobile ONNX 259.434 ms 10 - 569 MB NPU Whisper-Base-En.onnx
WhisperEncoderInf float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile TFLITE 176.492 ms 37 - 83 MB GPU Whisper-Base-En.tflite
WhisperEncoderInf float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile QNN_DLC 142.136 ms 0 - 1370 MB NPU Whisper-Base-En.dlc
WhisperEncoderInf float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile ONNX 190.438 ms 92 - 1646 MB NPU Whisper-Base-En.onnx
WhisperEncoderInf float Snapdragon 8 Elite QRD Snapdragon® 8 Elite Mobile TFLITE 160.081 ms 39 - 67 MB GPU Whisper-Base-En.tflite
WhisperEncoderInf float Snapdragon 8 Elite QRD Snapdragon® 8 Elite Mobile QNN_DLC 115.265 ms 1 - 1377 MB NPU Whisper-Base-En.dlc
WhisperEncoderInf float Snapdragon 8 Elite QRD Snapdragon® 8 Elite Mobile ONNX 150.849 ms 90 - 1645 MB NPU Whisper-Base-En.onnx
WhisperEncoderInf float Snapdragon X Elite CRD Snapdragon® X Elite QNN_DLC 162.983 ms 136 - 136 MB NPU Whisper-Base-En.dlc
WhisperEncoderInf float Snapdragon X Elite CRD Snapdragon® X Elite ONNX 207.702 ms 133 - 133 MB NPU Whisper-Base-En.onnx
WhisperDecoderInf float QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) TFLITE 16.579 ms 5 - 137 MB NPU Whisper-Base-En.tflite
WhisperDecoderInf float QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) QNN_DLC 6.813 ms 11 - 80 MB NPU Whisper-Base-En.dlc
WhisperDecoderInf float QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) TFLITE 10.759 ms 6 - 136 MB NPU Whisper-Base-En.tflite
WhisperDecoderInf float QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) TFLITE 9.549 ms 5 - 39 MB NPU Whisper-Base-En.tflite
WhisperDecoderInf float QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) QNN_DLC 4.071 ms 20 - 47 MB NPU Whisper-Base-En.dlc
WhisperDecoderInf float QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) TFLITE 10.754 ms 0 - 132 MB NPU Whisper-Base-En.tflite
WhisperDecoderInf float QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) QNN_DLC 4.864 ms 19 - 83 MB NPU Whisper-Base-En.dlc
WhisperDecoderInf float SA7255P ADP Qualcomm® SA7255P TFLITE 16.579 ms 5 - 137 MB NPU Whisper-Base-En.tflite
WhisperDecoderInf float SA7255P ADP Qualcomm® SA7255P QNN_DLC 6.813 ms 11 - 80 MB NPU Whisper-Base-En.dlc
WhisperDecoderInf float SA8255 (Proxy) Qualcomm® SA8255P (Proxy) TFLITE 9.651 ms 5 - 41 MB NPU Whisper-Base-En.tflite
WhisperDecoderInf float SA8255 (Proxy) Qualcomm® SA8255P (Proxy) QNN_DLC 4.071 ms 20 - 46 MB NPU Whisper-Base-En.dlc
WhisperDecoderInf float SA8295P ADP Qualcomm® SA8295P TFLITE 11.602 ms 5 - 128 MB NPU Whisper-Base-En.tflite
WhisperDecoderInf float SA8295P ADP Qualcomm® SA8295P QNN_DLC 5.223 ms 11 - 70 MB NPU Whisper-Base-En.dlc
WhisperDecoderInf float SA8650 (Proxy) Qualcomm® SA8650P (Proxy) TFLITE 9.717 ms 5 - 34 MB NPU Whisper-Base-En.tflite
WhisperDecoderInf float SA8650 (Proxy) Qualcomm® SA8650P (Proxy) QNN_DLC 4.036 ms 20 - 45 MB NPU Whisper-Base-En.dlc
WhisperDecoderInf float SA8775P ADP Qualcomm® SA8775P TFLITE 10.754 ms 0 - 132 MB NPU Whisper-Base-En.tflite
WhisperDecoderInf float SA8775P ADP Qualcomm® SA8775P QNN_DLC 4.864 ms 19 - 83 MB NPU Whisper-Base-En.dlc
WhisperDecoderInf float Samsung Galaxy S23 Snapdragon® 8 Gen 2 Mobile TFLITE 9.691 ms 5 - 35 MB NPU Whisper-Base-En.tflite
WhisperDecoderInf float Samsung Galaxy S23 Snapdragon® 8 Gen 2 Mobile QNN_DLC 4.062 ms 20 - 43 MB NPU Whisper-Base-En.dlc
WhisperDecoderInf float Samsung Galaxy S23 Snapdragon® 8 Gen 2 Mobile ONNX 10.009 ms 0 - 142 MB NPU Whisper-Base-En.onnx
WhisperDecoderInf float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile TFLITE 7.411 ms 5 - 148 MB NPU Whisper-Base-En.tflite
WhisperDecoderInf float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile QNN_DLC 3.149 ms 16 - 91 MB NPU Whisper-Base-En.dlc
WhisperDecoderInf float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile ONNX 8.306 ms 56 - 183 MB NPU Whisper-Base-En.onnx
WhisperDecoderInf float Snapdragon 8 Elite QRD Snapdragon® 8 Elite Mobile TFLITE 7.378 ms 5 - 134 MB NPU Whisper-Base-En.tflite
WhisperDecoderInf float Snapdragon 8 Elite QRD Snapdragon® 8 Elite Mobile QNN_DLC 2.704 ms 19 - 90 MB NPU Whisper-Base-En.dlc
WhisperDecoderInf float Snapdragon 8 Elite QRD Snapdragon® 8 Elite Mobile ONNX 7.631 ms 56 - 166 MB NPU Whisper-Base-En.onnx
WhisperDecoderInf float Snapdragon X Elite CRD Snapdragon® X Elite QNN_DLC 3.752 ms 231 - 231 MB NPU Whisper-Base-En.dlc
WhisperDecoderInf float Snapdragon X Elite CRD Snapdragon® X Elite ONNX 8.905 ms 107 - 107 MB NPU Whisper-Base-En.onnx

Installation

Install the package via pip:

pip install "qai-hub-models[whisper-base-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_base_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_base_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_base_en.export

Profiling Results
------------------------------------------------------------
WhisperEncoderInf
Device                          : cs_8275 (ANDROID 14)                  
Runtime                         : TFLITE                                
Estimated inference time (ms)   : 839.1                                 
Estimated peak memory usage (MB): [37, 60]                              
Total # Ops                     : 419                                   
Compute Unit(s)                 : npu (0 ops) gpu (408 ops) cpu (11 ops)

------------------------------------------------------------
WhisperDecoderInf
Device                          : cs_8275 (ANDROID 14)                 
Runtime                         : TFLITE                               
Estimated inference time (ms)   : 16.6                                 
Estimated peak memory usage (MB): [5, 137]                             
Total # Ops                     : 983                                  
Compute Unit(s)                 : npu (983 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_base_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-Base-En's performance across various devices here. Explore all available models on Qualcomm® AI Hub

License

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

References

Community

Downloads last month
202
Inference Providers NEW
Automatic Speech Recognition
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support