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 2.42, ONNX Runtime 1.24.3	Download
ONNX	w8a16	Universal	QAIRT 2.42, ONNX Runtime 1.24.3	Download
ONNX	w8a8	Universal	QAIRT 2.42, ONNX Runtime 1.24.3	Download
QNN_DLC	float	Universal	QAIRT 2.43	Download
QNN_DLC	w8a16	Universal	QAIRT 2.43	Download
QNN_DLC	w8a8	Universal	QAIRT 2.43	Download
TFLITE	float	Universal	QAIRT 2.43, TFLite 2.19.1	Download
TFLITE	w8a8	Universal	QAIRT 2.43, TFLite 2.19.1	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® 8 Elite Gen 5 Mobile	0.506 ms	1 - 36 MB	NPU
GPUNet	ONNX	float	Snapdragon® X2 Elite	0.469 ms	24 - 24 MB	NPU
GPUNet	ONNX	float	Snapdragon® X Elite	1.131 ms	24 - 24 MB	NPU
GPUNet	ONNX	float	Snapdragon® 8 Gen 3 Mobile	0.792 ms	0 - 58 MB	NPU
GPUNet	ONNX	float	Qualcomm® QCS8550 (Proxy)	1.045 ms	1 - 2 MB	NPU
GPUNet	ONNX	float	Qualcomm® QCS9075	1.354 ms	1 - 3 MB	NPU
GPUNet	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	0.622 ms	0 - 36 MB	NPU
GPUNet	ONNX	w8a16	Snapdragon® 8 Elite Gen 5 Mobile	0.358 ms	0 - 48 MB	NPU
GPUNet	ONNX	w8a16	Snapdragon® X2 Elite	0.386 ms	12 - 12 MB	NPU
GPUNet	ONNX	w8a16	Snapdragon® X Elite	0.986 ms	12 - 12 MB	NPU
GPUNet	ONNX	w8a16	Snapdragon® 8 Gen 3 Mobile	0.597 ms	0 - 62 MB	NPU
GPUNet	ONNX	w8a16	Qualcomm® QCS6490	98.939 ms	23 - 32 MB	CPU
GPUNet	ONNX	w8a16	Qualcomm® QCS8550 (Proxy)	0.828 ms	0 - 16 MB	NPU
GPUNet	ONNX	w8a16	Qualcomm® QCS9075	0.978 ms	0 - 3 MB	NPU
GPUNet	ONNX	w8a16	Qualcomm® QCM6690	49.574 ms	30 - 37 MB	CPU
GPUNet	ONNX	w8a16	Snapdragon® 8 Elite For Galaxy Mobile	0.433 ms	0 - 50 MB	NPU
GPUNet	ONNX	w8a16	Snapdragon® 7 Gen 4 Mobile	38.247 ms	29 - 37 MB	CPU
GPUNet	ONNX	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	0.358 ms	0 - 47 MB	NPU
GPUNet	ONNX	w8a8	Snapdragon® X2 Elite	0.308 ms	12 - 12 MB	NPU
GPUNet	ONNX	w8a8	Snapdragon® X Elite	0.737 ms	12 - 12 MB	NPU
GPUNet	ONNX	w8a8	Snapdragon® 8 Gen 3 Mobile	0.473 ms	0 - 67 MB	NPU
GPUNet	ONNX	w8a8	Qualcomm® QCS6490	16.535 ms	1 - 16 MB	CPU
GPUNet	ONNX	w8a8	Qualcomm® QCS8550 (Proxy)	0.616 ms	0 - 15 MB	NPU
GPUNet	ONNX	w8a8	Qualcomm® QCS9075	0.729 ms	0 - 3 MB	NPU
GPUNet	ONNX	w8a8	Qualcomm® QCM6690	10.103 ms	7 - 14 MB	CPU
GPUNet	ONNX	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	0.391 ms	0 - 46 MB	NPU
GPUNet	ONNX	w8a8	Snapdragon® 7 Gen 4 Mobile	7.718 ms	7 - 15 MB	CPU
GPUNet	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	0.567 ms	1 - 37 MB	NPU
GPUNet	QNN_DLC	float	Snapdragon® X2 Elite	0.661 ms	1 - 1 MB	NPU
GPUNet	QNN_DLC	float	Snapdragon® X Elite	1.43 ms	1 - 1 MB	NPU
GPUNet	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	0.926 ms	1 - 64 MB	NPU
GPUNet	QNN_DLC	float	Qualcomm® QCS8275 (Proxy)	4.647 ms	1 - 33 MB	NPU
GPUNet	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	1.29 ms	1 - 110 MB	NPU
GPUNet	QNN_DLC	float	Qualcomm® SA8775P	1.767 ms	1 - 36 MB	NPU
GPUNet	QNN_DLC	float	Qualcomm® QCS9075	1.577 ms	3 - 5 MB	NPU
GPUNet	QNN_DLC	float	Qualcomm® QCS8450 (Proxy)	2.412 ms	0 - 64 MB	NPU
GPUNet	QNN_DLC	float	Qualcomm® SA7255P	4.647 ms	1 - 33 MB	NPU
GPUNet	QNN_DLC	float	Qualcomm® SA8295P	2.3 ms	1 - 34 MB	NPU
GPUNet	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	0.707 ms	1 - 33 MB	NPU
GPUNet	QNN_DLC	w8a16	Snapdragon® 8 Elite Gen 5 Mobile	0.442 ms	0 - 45 MB	NPU
GPUNet	QNN_DLC	w8a16	Snapdragon® X2 Elite	0.571 ms	0 - 0 MB	NPU
GPUNet	QNN_DLC	w8a16	Snapdragon® X Elite	1.252 ms	0 - 0 MB	NPU
GPUNet	QNN_DLC	w8a16	Snapdragon® 8 Gen 3 Mobile	0.754 ms	0 - 58 MB	NPU
GPUNet	QNN_DLC	w8a16	Qualcomm® QCS6490	3.232 ms	2 - 4 MB	NPU
GPUNet	QNN_DLC	w8a16	Qualcomm® QCS8275 (Proxy)	2.451 ms	0 - 42 MB	NPU
GPUNet	QNN_DLC	w8a16	Qualcomm® QCS8550 (Proxy)	1.073 ms	0 - 2 MB	NPU
GPUNet	QNN_DLC	w8a16	Qualcomm® SA8775P	1.292 ms	0 - 44 MB	NPU
GPUNet	QNN_DLC	w8a16	Qualcomm® QCS9075	1.228 ms	2 - 4 MB	NPU
GPUNet	QNN_DLC	w8a16	Qualcomm® QCM6690	6.556 ms	0 - 159 MB	NPU
GPUNet	QNN_DLC	w8a16	Qualcomm® QCS8450 (Proxy)	1.44 ms	0 - 61 MB	NPU
GPUNet	QNN_DLC	w8a16	Qualcomm® SA7255P	2.451 ms	0 - 42 MB	NPU
GPUNet	QNN_DLC	w8a16	Qualcomm® SA8295P	1.715 ms	0 - 39 MB	NPU
GPUNet	QNN_DLC	w8a16	Snapdragon® 8 Elite For Galaxy Mobile	0.521 ms	0 - 45 MB	NPU
GPUNet	QNN_DLC	w8a16	Snapdragon® 7 Gen 4 Mobile	1.281 ms	0 - 42 MB	NPU
GPUNet	QNN_DLC	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	0.297 ms	0 - 43 MB	NPU
GPUNet	QNN_DLC	w8a8	Snapdragon® X2 Elite	0.356 ms	0 - 0 MB	NPU
GPUNet	QNN_DLC	w8a8	Snapdragon® X Elite	0.741 ms	0 - 0 MB	NPU
GPUNet	QNN_DLC	w8a8	Snapdragon® 8 Gen 3 Mobile	0.461 ms	0 - 56 MB	NPU
GPUNet	QNN_DLC	w8a8	Qualcomm® QCS6490	2.01 ms	2 - 4 MB	NPU
GPUNet	QNN_DLC	w8a8	Qualcomm® QCS8275 (Proxy)	1.398 ms	0 - 38 MB	NPU
GPUNet	QNN_DLC	w8a8	Qualcomm® QCS8550 (Proxy)	0.603 ms	0 - 2 MB	NPU
GPUNet	QNN_DLC	w8a8	Qualcomm® SA8775P	0.819 ms	0 - 42 MB	NPU
GPUNet	QNN_DLC	w8a8	Qualcomm® QCS9075	0.699 ms	2 - 4 MB	NPU
GPUNet	QNN_DLC	w8a8	Qualcomm® QCM6690	3.444 ms	0 - 41 MB	NPU
GPUNet	QNN_DLC	w8a8	Qualcomm® QCS8450 (Proxy)	0.857 ms	0 - 58 MB	NPU
GPUNet	QNN_DLC	w8a8	Qualcomm® SA7255P	1.398 ms	0 - 38 MB	NPU
GPUNet	QNN_DLC	w8a8	Qualcomm® SA8295P	1.112 ms	0 - 36 MB	NPU
GPUNet	QNN_DLC	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	0.338 ms	0 - 37 MB	NPU
GPUNet	QNN_DLC	w8a8	Snapdragon® 7 Gen 4 Mobile	0.791 ms	0 - 38 MB	NPU
GPUNet	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	0.566 ms	0 - 66 MB	NPU
GPUNet	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	0.893 ms	0 - 93 MB	NPU
GPUNet	TFLITE	float	Qualcomm® QCS8275 (Proxy)	4.616 ms	0 - 62 MB	NPU
GPUNet	TFLITE	float	Qualcomm® QCS8550 (Proxy)	1.287 ms	0 - 3 MB	NPU
GPUNet	TFLITE	float	Qualcomm® SA8775P	1.763 ms	0 - 65 MB	NPU
GPUNet	TFLITE	float	Qualcomm® QCS9075	1.582 ms	0 - 27 MB	NPU
GPUNet	TFLITE	float	Qualcomm® QCS8450 (Proxy)	2.415 ms	0 - 94 MB	NPU
GPUNet	TFLITE	float	Qualcomm® SA7255P	4.616 ms	0 - 62 MB	NPU
GPUNet	TFLITE	float	Qualcomm® SA8295P	2.281 ms	0 - 61 MB	NPU
GPUNet	TFLITE	float	Snapdragon® 8 Elite For Galaxy Mobile	0.703 ms	0 - 61 MB	NPU
GPUNet	TFLITE	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	0.232 ms	0 - 45 MB	NPU
GPUNet	TFLITE	w8a8	Snapdragon® 8 Gen 3 Mobile	0.345 ms	0 - 56 MB	NPU
GPUNet	TFLITE	w8a8	Qualcomm® QCS6490	1.669 ms	0 - 16 MB	NPU
GPUNet	TFLITE	w8a8	Qualcomm® QCS8275 (Proxy)	1.108 ms	0 - 39 MB	NPU
GPUNet	TFLITE	w8a8	Qualcomm® QCS8550 (Proxy)	0.429 ms	0 - 7 MB	NPU
GPUNet	TFLITE	w8a8	Qualcomm® SA8775P	0.633 ms	0 - 42 MB	NPU
GPUNet	TFLITE	w8a8	Qualcomm® QCS9075	0.52 ms	0 - 14 MB	NPU
GPUNet	TFLITE	w8a8	Qualcomm® QCM6690	3.026 ms	0 - 40 MB	NPU
GPUNet	TFLITE	w8a8	Qualcomm® QCS8450 (Proxy)	0.698 ms	0 - 57 MB	NPU
GPUNet	TFLITE	w8a8	Qualcomm® SA7255P	1.108 ms	0 - 39 MB	NPU
GPUNet	TFLITE	w8a8	Qualcomm® SA8295P	0.894 ms	0 - 37 MB	NPU
GPUNet	TFLITE	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	0.266 ms	0 - 37 MB	NPU
GPUNet	TFLITE	w8a8	Snapdragon® 7 Gen 4 Mobile	0.619 ms	0 - 38 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