ConvNext-Tiny-w8a16-Quantized: Optimized for Mobile Deployment

Imagenet classifier and general purpose backbone

ConvNextTiny is a machine learning model that can classify images from the Imagenet dataset. It can also be used as a backbone in building more complex models for specific use cases.

This model is an implementation of ConvNext-Tiny-w8a16-Quantized found here.

This repository provides scripts to run ConvNext-Tiny-w8a16-Quantized on Qualcomm® devices. More details on model performance across various devices, can be found here.

Model Details

Model Type: Image classification
Model Stats:
- Model checkpoint: Imagenet
- Input resolution: 224x224
- Number of parameters: 28.6M
- Model size: 28 MB
- Precision: w8a16 (8-bit weights, 16-bit activations)

Model	Device	Chipset	Target Runtime	Inference Time (ms)	Peak Memory Range (MB)	Precision	Primary Compute Unit	Target Model
ConvNext-Tiny-w8a16-Quantized	Samsung Galaxy S23	Snapdragon® 8 Gen 2	QNN	3.429 ms	0 - 136 MB	INT8	NPU	ConvNext-Tiny-w8a16-Quantized.so
ConvNext-Tiny-w8a16-Quantized	Samsung Galaxy S24	Snapdragon® 8 Gen 3	QNN	2.473 ms	0 - 37 MB	INT8	NPU	ConvNext-Tiny-w8a16-Quantized.so
ConvNext-Tiny-w8a16-Quantized	Snapdragon 8 Elite QRD	Snapdragon® 8 Elite	QNN	2.431 ms	0 - 36 MB	INT8	NPU	Use Export Script
ConvNext-Tiny-w8a16-Quantized	RB3 Gen 2 (Proxy)	QCS6490 Proxy	QNN	13.088 ms	0 - 7 MB	INT8	NPU	Use Export Script
ConvNext-Tiny-w8a16-Quantized	QCS8550 (Proxy)	QCS8550 Proxy	QNN	3.096 ms	0 - 2 MB	INT8	NPU	Use Export Script
ConvNext-Tiny-w8a16-Quantized	SA7255P ADP	SA7255P	QNN	26.821 ms	0 - 10 MB	INT8	NPU	Use Export Script
ConvNext-Tiny-w8a16-Quantized	SA8255 (Proxy)	SA8255P Proxy	QNN	3.095 ms	0 - 1 MB	INT8	NPU	Use Export Script
ConvNext-Tiny-w8a16-Quantized	SA8295P ADP	SA8295P	QNN	4.658 ms	0 - 6 MB	INT8	NPU	Use Export Script
ConvNext-Tiny-w8a16-Quantized	SA8650 (Proxy)	SA8650P Proxy	QNN	3.108 ms	0 - 2 MB	INT8	NPU	Use Export Script
ConvNext-Tiny-w8a16-Quantized	SA8775P ADP	SA8775P	QNN	4.446 ms	0 - 6 MB	INT8	NPU	Use Export Script
ConvNext-Tiny-w8a16-Quantized	QCS8450 (Proxy)	QCS8450 Proxy	QNN	4.258 ms	0 - 38 MB	INT8	NPU	Use Export Script
ConvNext-Tiny-w8a16-Quantized	Snapdragon X Elite CRD	Snapdragon® X Elite	QNN	3.38 ms	0 - 0 MB	INT8	NPU	Use Export Script

Installation

This model can be installed as a Python package via pip.

pip install "qai-hub-models[convnext_tiny_w8a16_quantized]"

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.convnext_tiny_w8a16_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.convnext_tiny_w8a16_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.

python -m qai_hub_models.models.convnext_tiny_w8a16_quantized.export

Profiling Results
------------------------------------------------------------
ConvNext-Tiny-w8a16-Quantized
Device                          : Samsung Galaxy S23 (13)
Runtime                         : QNN                    
Estimated inference time (ms)   : 3.4                    
Estimated peak memory usage (MB): [0, 136]               
Total # Ops                     : 215                    
Compute Unit(s)                 : NPU (215 ops)

Run demo on a cloud-hosted device

You can also run the demo on-device.

python -m qai_hub_models.models.convnext_tiny_w8a16_quantized.demo --on-device

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.convnext_tiny_w8a16_quantized.demo -- --on-device

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 ConvNext-Tiny-w8a16-Quantized's performance across various devices here. Explore all available models on Qualcomm® AI Hub

License

The license for the original implementation of ConvNext-Tiny-w8a16-Quantized 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.