Mobile-VIT: Optimized for Qualcomm Devices

MobileVit 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 Mobile-VIT 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 1.27.1	Download
ONNX	w8a16	Universal	QAIRT 2.45, ONNX Runtime 1.27.1	Download
QNN_DLC	float	Universal	QAIRT 2.45	Download
QNN_DLC	w8a16	Universal	QAIRT 2.45	Download
TFLITE	float	Universal	QAIRT 2.45	Download

For more device-specific assets and performance metrics, visit Mobile-VIT 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 Mobile-VIT 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: 5.57M
Model size (float): 21.4 MB
Model size (w8a16): 6.56 MB

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
Mobile-VIT	ONNX	float	Snapdragon® X2 Elite	2.272 ms	2 - 2 MB	NPU
Mobile-VIT	ONNX	float	Snapdragon® X Elite	4.569 ms	12 - 12 MB	NPU
Mobile-VIT	ONNX	float	Snapdragon® 8 Gen 3 Mobile	3.027 ms	0 - 93 MB	NPU
Mobile-VIT	ONNX	float	Snapdragon® 8 Gen 1 Mobile	6.59 ms	1 - 99 MB	NPU
Mobile-VIT	ONNX	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	4.355 ms	0 - 15 MB	NPU
Mobile-VIT	ONNX	float	Qualcomm® QCS8450	6.59 ms	1 - 99 MB	NPU
Mobile-VIT	ONNX	float	Qualcomm® Dragonwing™ IQ-9075	4.951 ms	1 - 4 MB	NPU
Mobile-VIT	ONNX	float	Qualcomm® Dragonwing™ IQ-X7181	4.569 ms	12 - 12 MB	NPU
Mobile-VIT	ONNX	float	Qualcomm® Dragonwing™ Q-8750	2.451 ms	0 - 70 MB	NPU
Mobile-VIT	ONNX	float	Snapdragon® 8 Elite Mobile	2.451 ms	0 - 70 MB	NPU
Mobile-VIT	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	2.053 ms	0 - 62 MB	NPU
Mobile-VIT	ONNX	w8a16	Snapdragon® X2 Elite	1.835 ms	1 - 1 MB	NPU
Mobile-VIT	ONNX	w8a16	Snapdragon® X Elite	4.391 ms	6 - 6 MB	NPU
Mobile-VIT	ONNX	w8a16	Snapdragon® 8 Gen 3 Mobile	2.837 ms	0 - 117 MB	NPU
Mobile-VIT	ONNX	w8a16	Snapdragon® 8 Gen 1 Mobile	4.799 ms	0 - 116 MB	NPU
Mobile-VIT	ONNX	w8a16	Qualcomm® Dragonwing™ QCS6490	13.266 ms	0 - 3 MB	NPU
Mobile-VIT	ONNX	w8a16	Qualcomm® Dragonwing™ QCS8550 (Proxy)	4.21 ms	0 - 138 MB	NPU
Mobile-VIT	ONNX	w8a16	Qualcomm® QCS8450	4.799 ms	0 - 116 MB	NPU
Mobile-VIT	ONNX	w8a16	Qualcomm® Dragonwing™ IQ-9075	6.175 ms	0 - 3 MB	NPU
Mobile-VIT	ONNX	w8a16	Qualcomm® Dragonwing™ IQ-X7181	4.391 ms	6 - 6 MB	NPU
Mobile-VIT	ONNX	w8a16	Qualcomm® Dragonwing™ Q-8750	2.069 ms	0 - 89 MB	NPU
Mobile-VIT	ONNX	w8a16	Snapdragon® 8 Elite Mobile	2.069 ms	0 - 89 MB	NPU
Mobile-VIT	ONNX	w8a16	Snapdragon® 8 Elite Gen 5 Mobile	1.739 ms	0 - 91 MB	NPU
Mobile-VIT	QNN_DLC	float	Snapdragon® X2 Elite	2.48 ms	1 - 1 MB	NPU
Mobile-VIT	QNN_DLC	float	Snapdragon® X Elite	4.625 ms	1 - 1 MB	NPU
Mobile-VIT	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	2.935 ms	0 - 86 MB	NPU
Mobile-VIT	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	6.464 ms	0 - 89 MB	NPU
Mobile-VIT	QNN_DLC	float	Qualcomm® Dragonwing™ QCS8275	11.157 ms	1 - 57 MB	NPU
Mobile-VIT	QNN_DLC	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	4.208 ms	1 - 99 MB	NPU
Mobile-VIT	QNN_DLC	float	Qualcomm® SA8775P	5.032 ms	1 - 66 MB	NPU
Mobile-VIT	QNN_DLC	float	Qualcomm® SA8650P	5.032 ms	1 - 66 MB	NPU
Mobile-VIT	QNN_DLC	float	Qualcomm® SA8255P	5.032 ms	1 - 66 MB	NPU
Mobile-VIT	QNN_DLC	float	Qualcomm® QCS8450	6.464 ms	0 - 89 MB	NPU
Mobile-VIT	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-9075	4.771 ms	1 - 3 MB	NPU
Mobile-VIT	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-X7181	4.625 ms	1 - 1 MB	NPU
Mobile-VIT	QNN_DLC	float	Qualcomm® Dragonwing™ Q-8750	2.388 ms	1 - 59 MB	NPU
Mobile-VIT	QNN_DLC	float	Qualcomm® SA7255P	11.157 ms	1 - 57 MB	NPU
Mobile-VIT	QNN_DLC	float	Qualcomm® SA8295P	7.355 ms	1 - 61 MB	NPU
Mobile-VIT	QNN_DLC	float	Snapdragon® 8 Elite Mobile	2.388 ms	1 - 59 MB	NPU
Mobile-VIT	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	1.99 ms	1 - 61 MB	NPU
Mobile-VIT	QNN_DLC	w8a16	Snapdragon® X2 Elite	2.184 ms	0 - 0 MB	NPU
Mobile-VIT	QNN_DLC	w8a16	Snapdragon® X Elite	4.865 ms	0 - 0 MB	NPU
Mobile-VIT	QNN_DLC	w8a16	Snapdragon® 8 Gen 3 Mobile	2.962 ms	0 - 94 MB	NPU
Mobile-VIT	QNN_DLC	w8a16	Qualcomm® Dragonwing™ QCS8275	8.02 ms	0 - 69 MB	NPU
Mobile-VIT	QNN_DLC	w8a16	Qualcomm® Dragonwing™ QCS8550 (Proxy)	4.484 ms	0 - 2 MB	NPU
Mobile-VIT	QNN_DLC	w8a16	Qualcomm® SA8775P	4.87 ms	0 - 72 MB	NPU
Mobile-VIT	QNN_DLC	w8a16	Qualcomm® SA8650P	4.87 ms	0 - 72 MB	NPU
Mobile-VIT	QNN_DLC	w8a16	Qualcomm® SA8255P	4.87 ms	0 - 72 MB	NPU
Mobile-VIT	QNN_DLC	w8a16	Qualcomm® Dragonwing™ IQ-9075	4.696 ms	0 - 2 MB	NPU
Mobile-VIT	QNN_DLC	w8a16	Qualcomm® Dragonwing™ IQ-X7181	4.865 ms	0 - 0 MB	NPU
Mobile-VIT	QNN_DLC	w8a16	Qualcomm® Dragonwing™ Q-6690	17.742 ms	0 - 223 MB	NPU
Mobile-VIT	QNN_DLC	w8a16	Qualcomm® Dragonwing™ Q-7790	4.88 ms	0 - 214 MB	NPU
Mobile-VIT	QNN_DLC	w8a16	Qualcomm® Dragonwing™ Q-8750	2.187 ms	0 - 68 MB	NPU
Mobile-VIT	QNN_DLC	w8a16	Qualcomm® SA7255P	8.02 ms	0 - 69 MB	NPU
Mobile-VIT	QNN_DLC	w8a16	Snapdragon® 8 Elite Mobile	2.187 ms	0 - 68 MB	NPU
Mobile-VIT	QNN_DLC	w8a16	Snapdragon® 8 Elite Gen 5 Mobile	1.811 ms	0 - 77 MB	NPU
Mobile-VIT	QNN_DLC	w8a16	Snapdragon® 7 Gen 4 Mobile	4.88 ms	0 - 214 MB	NPU
Mobile-VIT	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	2.945 ms	0 - 101 MB	NPU
Mobile-VIT	TFLITE	float	Snapdragon® 8 Gen 1 Mobile	6.439 ms	0 - 102 MB	NPU
Mobile-VIT	TFLITE	float	Qualcomm® Dragonwing™ QCS8275	11.212 ms	0 - 78 MB	NPU
Mobile-VIT	TFLITE	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	4.218 ms	0 - 40 MB	NPU
Mobile-VIT	TFLITE	float	Qualcomm® SA8775P	5.043 ms	0 - 80 MB	NPU
Mobile-VIT	TFLITE	float	Qualcomm® SA8650P	5.043 ms	0 - 80 MB	NPU
Mobile-VIT	TFLITE	float	Qualcomm® SA8255P	5.043 ms	0 - 80 MB	NPU
Mobile-VIT	TFLITE	float	Qualcomm® QCS8450	6.439 ms	0 - 102 MB	NPU
Mobile-VIT	TFLITE	float	Qualcomm® Dragonwing™ IQ-9075	4.783 ms	0 - 15 MB	NPU
Mobile-VIT	TFLITE	float	Qualcomm® Dragonwing™ Q-8750	2.388 ms	0 - 76 MB	NPU
Mobile-VIT	TFLITE	float	Qualcomm® SA7255P	11.212 ms	0 - 78 MB	NPU
Mobile-VIT	TFLITE	float	Qualcomm® SA8295P	7.336 ms	0 - 64 MB	NPU
Mobile-VIT	TFLITE	float	Snapdragon® 8 Elite Mobile	2.388 ms	0 - 76 MB	NPU
Mobile-VIT	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	2.013 ms	0 - 78 MB	NPU

License

The license for the original implementation of Mobile-VIT 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/Mobile-VIT

MobileViT: Light-weight, General-purpose, and Mobile-friendly Vision Transformer

Paper • 2110.02178 • Published Oct 5, 2021 • 3