HuggingFace-WavLM-Base-Plus: Optimized for Mobile Deployment

Real-time Speech processing

HuggingFaceWavLMBasePlus is a real time speech processing backbone based on Microsoft's WavLM model.

This model is an implementation of HuggingFace-WavLM-Base-Plus found here.

This repository provides scripts to run HuggingFace-WavLM-Base-Plus on Qualcomm® devices. More details on model performance across various devices, can be found here.

Model Details

Model Type: Speech recognition
Model Stats:
- Model checkpoint: wavlm-libri-clean-100h-base-plus
- Input resolution: 1x320000
- Number of parameters: 95.1M
- Model size: 363 MB

Model	Device	Chipset	Target Runtime	Inference Time (ms)	Peak Memory Range (MB)	Precision	Primary Compute Unit	Target Model
HuggingFace-WavLM-Base-Plus	Samsung Galaxy S23	Snapdragon® 8 Gen 2	TFLITE	164.023 ms	0 - 50 MB	FP16	NPU	HuggingFace-WavLM-Base-Plus.tflite
HuggingFace-WavLM-Base-Plus	Samsung Galaxy S23	Snapdragon® 8 Gen 2	ONNX	181.343 ms	3 - 442 MB	FP16	NPU	HuggingFace-WavLM-Base-Plus.onnx
HuggingFace-WavLM-Base-Plus	Samsung Galaxy S24	Snapdragon® 8 Gen 3	TFLITE	120.214 ms	1 - 193 MB	FP16	NPU	HuggingFace-WavLM-Base-Plus.tflite
HuggingFace-WavLM-Base-Plus	Samsung Galaxy S24	Snapdragon® 8 Gen 3	ONNX	131.235 ms	2 - 728 MB	FP16	NPU	HuggingFace-WavLM-Base-Plus.onnx
HuggingFace-WavLM-Base-Plus	Snapdragon 8 Elite QRD	Snapdragon® 8 Elite	TFLITE	117.174 ms	1 - 206 MB	FP16	NPU	HuggingFace-WavLM-Base-Plus.tflite
HuggingFace-WavLM-Base-Plus	Snapdragon 8 Elite QRD	Snapdragon® 8 Elite	ONNX	114.035 ms	2 - 749 MB	FP16	NPU	HuggingFace-WavLM-Base-Plus.onnx
HuggingFace-WavLM-Base-Plus	SA7255P ADP	SA7255P	TFLITE	2268.84 ms	0 - 202 MB	FP16	NPU	HuggingFace-WavLM-Base-Plus.tflite
HuggingFace-WavLM-Base-Plus	SA8255 (Proxy)	SA8255P Proxy	TFLITE	164.093 ms	2 - 51 MB	FP16	NPU	HuggingFace-WavLM-Base-Plus.tflite
HuggingFace-WavLM-Base-Plus	SA8295P ADP	SA8295P	TFLITE	270.626 ms	1 - 196 MB	FP16	NPU	HuggingFace-WavLM-Base-Plus.tflite
HuggingFace-WavLM-Base-Plus	SA8650 (Proxy)	SA8650P Proxy	TFLITE	164.721 ms	0 - 66 MB	FP16	NPU	HuggingFace-WavLM-Base-Plus.tflite
HuggingFace-WavLM-Base-Plus	SA8775P ADP	SA8775P	TFLITE	225.045 ms	1 - 203 MB	FP16	NPU	HuggingFace-WavLM-Base-Plus.tflite
HuggingFace-WavLM-Base-Plus	QCS8275 (Proxy)	QCS8275 Proxy	TFLITE	2268.84 ms	0 - 202 MB	FP16	NPU	HuggingFace-WavLM-Base-Plus.tflite
HuggingFace-WavLM-Base-Plus	QCS8550 (Proxy)	QCS8550 Proxy	TFLITE	163.494 ms	0 - 52 MB	FP16	NPU	HuggingFace-WavLM-Base-Plus.tflite
HuggingFace-WavLM-Base-Plus	QCS9075 (Proxy)	QCS9075 Proxy	TFLITE	225.045 ms	1 - 203 MB	FP16	NPU	HuggingFace-WavLM-Base-Plus.tflite
HuggingFace-WavLM-Base-Plus	QCS8450 (Proxy)	QCS8450 Proxy	TFLITE	247.182 ms	1 - 191 MB	FP16	NPU	HuggingFace-WavLM-Base-Plus.tflite
HuggingFace-WavLM-Base-Plus	Snapdragon X Elite CRD	Snapdragon® X Elite	ONNX	185.751 ms	207 - 207 MB	FP16	NPU	HuggingFace-WavLM-Base-Plus.onnx

Installation

Install the package via pip:

pip install "qai-hub-models[huggingface-wavlm-base-plus]"

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.huggingface_wavlm_base_plus.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.huggingface_wavlm_base_plus.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.huggingface_wavlm_base_plus.export

Profiling Results
------------------------------------------------------------
HuggingFace-WavLM-Base-Plus
Device                          : Samsung Galaxy S23 (13)
Runtime                         : TFLITE                 
Estimated inference time (ms)   : 164.0                  
Estimated peak memory usage (MB): [0, 50]                
Total # Ops                     : 871                    
Compute Unit(s)                 : NPU (871 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.huggingface_wavlm_base_plus 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 HuggingFace-WavLM-Base-Plus's performance across various devices here. Explore all available models on Qualcomm® AI Hub

License

The license for the original implementation of HuggingFace-WavLM-Base-Plus can be found here.
The license for the compiled assets for on-device deployment can be found here

References

Community

Join our AI Hub Slack community to collaborate, post questions and learn more about on-device AI.
For questions or feedback please reach out to us.