GPUNet: Optimized for Qualcomm Devices

GPUNet 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 is based on the implementation of GPUNet found here. This repository contains pre-exported model files optimized for Qualcomm® devices. You can use the Qualcomm® AI Hub Models library to export with custom configurations. More details on model performance across various devices, can be found here.

Qualcomm AI Hub Models uses Qualcomm AI Hub Workbench to compile, profile, and evaluate this model. Sign up to run these models on a hosted Qualcomm® device.

Getting Started

There are two ways to deploy this model on your device:

Option 1: Download Pre-Exported Models

Below are pre-exported model assets ready for deployment.

Runtime	Precision	Chipset	SDK Versions	Download
ONNX	float	Universal	QAIRT 2.45, ONNX Runtime 1.25.0	Download
ONNX	w8a16	Universal	QAIRT 2.45, ONNX Runtime 1.25.0	Download
ONNX	w8a8	Universal	QAIRT 2.45, ONNX Runtime 1.25.0	Download
QNN_DLC	float	Universal	QAIRT 2.45	Download
QNN_DLC	w8a16	Universal	QAIRT 2.45	Download
QNN_DLC	w8a8	Universal	QAIRT 2.45	Download
TFLITE	float	Universal	QAIRT 2.45	Download
TFLITE	w8a8	Universal	QAIRT 2.45	Download

For more device-specific assets and performance metrics, visit GPUNet on Qualcomm® AI Hub.

Option 2: Export with Custom Configurations

Use the Qualcomm® AI Hub Models Python library to compile and export the model with your own:

Custom weights (e.g., fine-tuned checkpoints)
Custom input shapes
Target device and runtime configurations

This option is ideal if you need to customize the model beyond the default configuration provided here.

See our repository for GPUNet on GitHub for usage instructions.

Model Details

Model Type: Model_use_case.image_classification

Model Stats:

Model checkpoint: Imagenet
Input resolution: 224x224
Number of parameters: 10.49M
Model size (float): 45.28MB
Model size (w8a8): 21.3MB

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
GPUNet	ONNX	float	Snapdragon® X2 Elite	0.477 ms	180 - 180 MB	NPU
GPUNet	ONNX	float	Snapdragon® X Elite	0.987 ms	181 - 181 MB	NPU
GPUNet	ONNX	float	Snapdragon® 8 Gen 3 Mobile	0.769 ms	0 - 51 MB	NPU
GPUNet	ONNX	float	Snapdragon® 8 Gen 1 Mobile	1.721 ms	1 - 53 MB	NPU
GPUNet	ONNX	float	Qualcomm® QCS8550 (Proxy)	1.01 ms	1 - 4 MB	NPU
GPUNet	ONNX	float	Qualcomm® QCS8450	1.721 ms	1 - 53 MB	NPU
GPUNet	ONNX	float	Snapdragon® 8 Elite Mobile	0.598 ms	0 - 34 MB	NPU
GPUNet	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	0.513 ms	1 - 31 MB	NPU
GPUNet	ONNX	float	Qualcomm® QCS9075	1.357 ms	1 - 46 MB	NPU
GPUNet	ONNX	float	Qualcomm® QCS8750	0.598 ms	0 - 34 MB	NPU
GPUNet	ONNX	float	Qualcomm® QCS7181	0.987 ms	181 - 181 MB	NPU
GPUNet	ONNX	w8a16	Snapdragon® X2 Elite	0.381 ms	181 - 181 MB	NPU
GPUNet	ONNX	w8a16	Snapdragon® X Elite	0.832 ms	181 - 181 MB	NPU
GPUNet	ONNX	w8a16	Snapdragon® 8 Gen 3 Mobile	0.576 ms	0 - 53 MB	NPU
GPUNet	ONNX	w8a16	Snapdragon® 8 Gen 1 Mobile	1.12 ms	0 - 59 MB	NPU
GPUNet	ONNX	w8a16	Qualcomm® QCS6490	2.79 ms	0 - 45 MB	NPU
GPUNet	ONNX	w8a16	Qualcomm® QCS8550 (Proxy)	0.824 ms	0 - 20 MB	NPU
GPUNet	ONNX	w8a16	Qualcomm® QCS8450	1.12 ms	0 - 59 MB	NPU
GPUNet	ONNX	w8a16	Snapdragon® 8 Elite Gen 5 Mobile	0.364 ms	0 - 44 MB	NPU
GPUNet	ONNX	w8a16	Snapdragon® 8 Elite Mobile	0.435 ms	0 - 52 MB	NPU
GPUNet	ONNX	w8a16	Qualcomm® QCS9075	0.985 ms	0 - 45 MB	NPU
GPUNet	ONNX	w8a16	Snapdragon® 7 Gen 4 Mobile	1.056 ms	0 - 43 MB	NPU
GPUNet	ONNX	w8a16	Qualcomm® QCM6690	5.765 ms	0 - 172 MB	NPU
GPUNet	ONNX	w8a16	Qualcomm® QCS7790	1.056 ms	0 - 43 MB	NPU
GPUNet	ONNX	w8a16	Qualcomm® QCS8750	0.435 ms	0 - 52 MB	NPU
GPUNet	ONNX	w8a16	Qualcomm® QCS7181	0.832 ms	181 - 181 MB	NPU
GPUNet	ONNX	w8a8	Snapdragon® X2 Elite	0.305 ms	180 - 180 MB	NPU
GPUNet	ONNX	w8a8	Snapdragon® X Elite	0.598 ms	176 - 176 MB	NPU
GPUNet	ONNX	w8a8	Snapdragon® 8 Gen 3 Mobile	0.467 ms	0 - 56 MB	NPU
GPUNet	ONNX	w8a8	Snapdragon® 8 Gen 1 Mobile	0.904 ms	0 - 56 MB	NPU
GPUNet	ONNX	w8a8	Qualcomm® QCS6490	1.915 ms	0 - 45 MB	NPU
GPUNet	ONNX	w8a8	Qualcomm® QCS8550 (Proxy)	0.623 ms	0 - 4 MB	NPU
GPUNet	ONNX	w8a8	Qualcomm® QCS8450	0.904 ms	0 - 56 MB	NPU
GPUNet	ONNX	w8a8	Qualcomm® QCS9075	0.733 ms	0 - 45 MB	NPU
GPUNet	ONNX	w8a8	Qualcomm® QCM6690	3.276 ms	0 - 44 MB	NPU
GPUNet	ONNX	w8a8	Snapdragon® 7 Gen 4 Mobile	0.76 ms	0 - 40 MB	NPU
GPUNet	ONNX	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	0.36 ms	0 - 45 MB	NPU
GPUNet	ONNX	w8a8	Snapdragon® 8 Elite Mobile	0.398 ms	0 - 43 MB	NPU
GPUNet	ONNX	w8a8	Qualcomm® QCS7790	0.76 ms	0 - 40 MB	NPU
GPUNet	ONNX	w8a8	Qualcomm® QCS8750	0.398 ms	0 - 43 MB	NPU
GPUNet	ONNX	w8a8	Qualcomm® QCS7181	0.598 ms	176 - 176 MB	NPU
GPUNet	QNN_DLC	float	Snapdragon® X2 Elite	0.647 ms	1 - 1 MB	NPU
GPUNet	QNN_DLC	float	Snapdragon® X Elite	1.372 ms	1 - 1 MB	NPU
GPUNet	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	0.917 ms	0 - 56 MB	NPU
GPUNet	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	2.419 ms	0 - 63 MB	NPU
GPUNet	QNN_DLC	float	Qualcomm® QCS8275	4.69 ms	1 - 31 MB	NPU
GPUNet	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	1.269 ms	1 - 2 MB	NPU
GPUNet	QNN_DLC	float	Qualcomm® QCS8450	2.419 ms	0 - 63 MB	NPU
GPUNet	QNN_DLC	float	Snapdragon® 8 Elite Mobile	0.705 ms	0 - 35 MB	NPU
GPUNet	QNN_DLC	float	Qualcomm® SA7255P	4.69 ms	1 - 31 MB	NPU
GPUNet	QNN_DLC	float	Qualcomm® SA8295P	2.275 ms	1 - 32 MB	NPU
GPUNet	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	0.566 ms	1 - 38 MB	NPU
GPUNet	QNN_DLC	float	Qualcomm® QCS9075	1.574 ms	1 - 3 MB	NPU
GPUNet	QNN_DLC	float	Qualcomm® QCS8750	0.705 ms	0 - 35 MB	NPU
GPUNet	QNN_DLC	float	Qualcomm® QCS7181	1.372 ms	1 - 1 MB	NPU
GPUNet	QNN_DLC	w8a16	Snapdragon® X2 Elite	0.591 ms	0 - 0 MB	NPU
GPUNet	QNN_DLC	w8a16	Snapdragon® X Elite	1.212 ms	0 - 0 MB	NPU
GPUNet	QNN_DLC	w8a16	Snapdragon® 8 Gen 3 Mobile	0.767 ms	0 - 60 MB	NPU
GPUNet	QNN_DLC	w8a16	Snapdragon® 8 Gen 1 Mobile	1.456 ms	0 - 64 MB	NPU
GPUNet	QNN_DLC	w8a16	Qualcomm® QCS6490	3.232 ms	0 - 2 MB	NPU
GPUNet	QNN_DLC	w8a16	Qualcomm® QCS8275	2.515 ms	0 - 43 MB	NPU
GPUNet	QNN_DLC	w8a16	Qualcomm® QCS8550 (Proxy)	1.069 ms	0 - 8 MB	NPU
GPUNet	QNN_DLC	w8a16	Qualcomm® QCS8450	1.456 ms	0 - 64 MB	NPU
GPUNet	QNN_DLC	w8a16	Snapdragon® 8 Elite Gen 5 Mobile	0.431 ms	0 - 43 MB	NPU
GPUNet	QNN_DLC	w8a16	Snapdragon® 8 Elite Mobile	0.522 ms	0 - 43 MB	NPU
GPUNet	QNN_DLC	w8a16	Qualcomm® QCS9075	1.243 ms	0 - 2 MB	NPU
GPUNet	QNN_DLC	w8a16	Snapdragon® 7 Gen 4 Mobile	1.282 ms	0 - 44 MB	NPU
GPUNet	QNN_DLC	w8a16	Qualcomm® QCM6690	6.577 ms	0 - 165 MB	NPU
GPUNet	QNN_DLC	w8a16	Qualcomm® SA8295P	1.663 ms	18 - 58 MB	NPU
GPUNet	QNN_DLC	w8a16	Qualcomm® SA7255P	2.515 ms	0 - 43 MB	NPU
GPUNet	QNN_DLC	w8a16	Qualcomm® QCS7790	1.282 ms	0 - 44 MB	NPU
GPUNet	QNN_DLC	w8a16	Qualcomm® QCS8750	0.522 ms	0 - 43 MB	NPU
GPUNet	QNN_DLC	w8a16	Qualcomm® QCS7181	1.212 ms	0 - 0 MB	NPU
GPUNet	QNN_DLC	w8a8	Snapdragon® X2 Elite	0.344 ms	0 - 0 MB	NPU
GPUNet	QNN_DLC	w8a8	Snapdragon® X Elite	0.701 ms	0 - 0 MB	NPU
GPUNet	QNN_DLC	w8a8	Snapdragon® 8 Gen 3 Mobile	0.466 ms	0 - 57 MB	NPU
GPUNet	QNN_DLC	w8a8	Snapdragon® 8 Gen 1 Mobile	0.853 ms	0 - 63 MB	NPU
GPUNet	QNN_DLC	w8a8	Qualcomm® QCS6490	2.043 ms	0 - 2 MB	NPU
GPUNet	QNN_DLC	w8a8	Qualcomm® QCS8275	1.419 ms	0 - 40 MB	NPU
GPUNet	QNN_DLC	w8a8	Qualcomm® QCS8550 (Proxy)	0.613 ms	0 - 2 MB	NPU
GPUNet	QNN_DLC	w8a8	Qualcomm® QCS8450	0.853 ms	0 - 63 MB	NPU
GPUNet	QNN_DLC	w8a8	Qualcomm® QCS9075	0.694 ms	0 - 2 MB	NPU
GPUNet	QNN_DLC	w8a8	Qualcomm® SA7255P	1.419 ms	0 - 40 MB	NPU
GPUNet	QNN_DLC	w8a8	Qualcomm® QCM6690	3.559 ms	0 - 44 MB	NPU
GPUNet	QNN_DLC	w8a8	Snapdragon® 7 Gen 4 Mobile	0.789 ms	0 - 41 MB	NPU
GPUNet	QNN_DLC	w8a8	Qualcomm® SA8295P	1.104 ms	0 - 38 MB	NPU
GPUNet	QNN_DLC	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	0.292 ms	0 - 42 MB	NPU
GPUNet	QNN_DLC	w8a8	Snapdragon® 8 Elite Mobile	0.346 ms	0 - 38 MB	NPU
GPUNet	QNN_DLC	w8a8	Qualcomm® QCS7790	0.789 ms	0 - 41 MB	NPU
GPUNet	QNN_DLC	w8a8	Qualcomm® QCS8750	0.346 ms	0 - 38 MB	NPU
GPUNet	QNN_DLC	w8a8	Qualcomm® QCS7181	0.701 ms	0 - 0 MB	NPU
GPUNet	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	0.906 ms	0 - 79 MB	NPU
GPUNet	TFLITE	float	Snapdragon® 8 Gen 1 Mobile	2.394 ms	0 - 80 MB	NPU
GPUNet	TFLITE	float	Qualcomm® QCS8275	4.755 ms	0 - 43 MB	NPU
GPUNet	TFLITE	float	Qualcomm® QCS8550 (Proxy)	1.266 ms	0 - 13 MB	NPU
GPUNet	TFLITE	float	Qualcomm® SA8775P	8.412 ms	0 - 28 MB	GPU
GPUNet	TFLITE	float	Qualcomm® SA8650P	8.412 ms	0 - 28 MB	GPU
GPUNet	TFLITE	float	Qualcomm® SA8255P	8.412 ms	0 - 28 MB	GPU
GPUNet	TFLITE	float	Qualcomm® QCS8450	2.394 ms	0 - 80 MB	NPU
GPUNet	TFLITE	float	Snapdragon® 8 Elite Mobile	0.698 ms	0 - 43 MB	NPU
GPUNet	TFLITE	float	Qualcomm® SA7255P	4.755 ms	0 - 43 MB	NPU
GPUNet	TFLITE	float	Qualcomm® SA8295P	2.284 ms	0 - 44 MB	NPU
GPUNet	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	0.557 ms	0 - 47 MB	NPU
GPUNet	TFLITE	float	Qualcomm® QCS9075	1.584 ms	0 - 27 MB	NPU
GPUNet	TFLITE	float	Qualcomm® QCS8750	0.698 ms	0 - 43 MB	NPU
GPUNet	TFLITE	w8a8	Snapdragon® 8 Gen 3 Mobile	0.342 ms	0 - 55 MB	NPU
GPUNet	TFLITE	w8a8	Snapdragon® 8 Gen 1 Mobile	0.683 ms	0 - 62 MB	NPU
GPUNet	TFLITE	w8a8	Qualcomm® QCS6490	1.604 ms	0 - 15 MB	NPU
GPUNet	TFLITE	w8a8	Qualcomm® QCS8275	1.115 ms	0 - 40 MB	NPU
GPUNet	TFLITE	w8a8	Qualcomm® QCS8550 (Proxy)	0.424 ms	0 - 2 MB	NPU
GPUNet	TFLITE	w8a8	Qualcomm® SA8775P	8.702 ms	1 - 29 MB	GPU
GPUNet	TFLITE	w8a8	Qualcomm® SA8650P	8.702 ms	1 - 29 MB	GPU
GPUNet	TFLITE	w8a8	Qualcomm® SA8255P	8.702 ms	1 - 29 MB	GPU
GPUNet	TFLITE	w8a8	Qualcomm® QCS8450	0.683 ms	0 - 62 MB	NPU
GPUNet	TFLITE	w8a8	Qualcomm® QCS9075	0.521 ms	0 - 14 MB	NPU
GPUNet	TFLITE	w8a8	Qualcomm® SA7255P	1.115 ms	0 - 40 MB	NPU
GPUNet	TFLITE	w8a8	Qualcomm® QCM6690	2.984 ms	0 - 41 MB	NPU
GPUNet	TFLITE	w8a8	Snapdragon® 7 Gen 4 Mobile	0.618 ms	0 - 39 MB	NPU
GPUNet	TFLITE	w8a8	Qualcomm® SA8295P	0.908 ms	0 - 36 MB	NPU
GPUNet	TFLITE	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	0.24 ms	0 - 41 MB	NPU
GPUNet	TFLITE	w8a8	Snapdragon® 8 Elite Mobile	0.278 ms	0 - 37 MB	NPU
GPUNet	TFLITE	w8a8	Qualcomm® QCS7790	0.618 ms	0 - 39 MB	NPU
GPUNet	TFLITE	w8a8	Qualcomm® QCS8750	0.278 ms	0 - 37 MB	NPU

License

The license for the original implementation of GPUNet 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.

Downloads last month: -; Downloads are not tracked for this model. How to track

Paper for qualcomm/GPUNet

GPUNet: Searching the Deployable Convolution Neural Networks for GPUs

Paper • 2205.00841 • Published Apr 26, 2022