VIT: Optimized for Qualcomm Devices

VIT 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 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 2.42, ONNX Runtime 1.24.1	Download
ONNX	w8a16	Universal	QAIRT 2.42, ONNX Runtime 1.24.1	Download
ONNX	w8a8	Universal	QAIRT 2.42, ONNX Runtime 1.24.1	Download
ONNX	w8a8_mixed_int16	Universal	QAIRT 2.42, ONNX Runtime 1.24.1	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 2.17.0	Download
TFLITE	w8a8	Universal	QAIRT 2.43, TFLite 2.17.0	Download

For more device-specific assets and performance metrics, visit 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 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: 86.6M
Model size (float): 330 MB
Model size (w8a16): 86.2 MB
Model size (w8a8): 83.2 MB

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
VIT	ONNX	float	Snapdragon® X2 Elite	3.862 ms	170 - 170 MB	NPU
VIT	ONNX	float	Snapdragon® X Elite	11.125 ms	170 - 170 MB	NPU
VIT	ONNX	float	Snapdragon® 8 Gen 3 Mobile	7.176 ms	0 - 379 MB	NPU
VIT	ONNX	float	Qualcomm® QCS8550 (Proxy)	10.443 ms	0 - 198 MB	NPU
VIT	ONNX	float	Qualcomm® QCS9075	14.381 ms	1 - 4 MB	NPU
VIT	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	5.0 ms	0 - 347 MB	NPU
VIT	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	3.655 ms	1 - 356 MB	NPU
VIT	ONNX	w8a16	Snapdragon® X2 Elite	3.889 ms	86 - 86 MB	NPU
VIT	ONNX	w8a16	Snapdragon® X Elite	11.271 ms	86 - 86 MB	NPU
VIT	ONNX	w8a16	Snapdragon® 8 Gen 3 Mobile	7.346 ms	0 - 375 MB	NPU
VIT	ONNX	w8a16	Qualcomm® QCS6490	1121.601 ms	43 - 59 MB	CPU
VIT	ONNX	w8a16	Qualcomm® QCS8550 (Proxy)	10.685 ms	0 - 5 MB	NPU
VIT	ONNX	w8a16	Qualcomm® QCS9075	13.06 ms	0 - 3 MB	NPU
VIT	ONNX	w8a16	Qualcomm® QCM6690	619.324 ms	68 - 83 MB	CPU
VIT	ONNX	w8a16	Snapdragon® 8 Elite For Galaxy Mobile	5.263 ms	0 - 295 MB	NPU
VIT	ONNX	w8a16	Snapdragon® 7 Gen 4 Mobile	595.167 ms	85 - 103 MB	CPU
VIT	ONNX	w8a16	Snapdragon® 8 Elite Gen 5 Mobile	3.798 ms	0 - 303 MB	NPU
VIT	ONNX	w8a8	Snapdragon® X2 Elite	5.096 ms	85 - 85 MB	NPU
VIT	ONNX	w8a8	Snapdragon® X Elite	13.639 ms	85 - 85 MB	NPU
VIT	ONNX	w8a8	Snapdragon® 8 Gen 3 Mobile	8.828 ms	0 - 464 MB	NPU
VIT	ONNX	w8a8	Qualcomm® QCS6490	312.895 ms	20 - 60 MB	CPU
VIT	ONNX	w8a8	Qualcomm® QCS8550 (Proxy)	12.938 ms	0 - 107 MB	NPU
VIT	ONNX	w8a8	Qualcomm® QCS9075	13.258 ms	0 - 3 MB	NPU
VIT	ONNX	w8a8	Qualcomm® QCM6690	134.641 ms	13 - 31 MB	CPU
VIT	ONNX	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	7.25 ms	0 - 321 MB	NPU
VIT	ONNX	w8a8	Snapdragon® 7 Gen 4 Mobile	129.747 ms	12 - 33 MB	CPU
VIT	ONNX	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	4.584 ms	0 - 350 MB	NPU
VIT	ONNX	w8a8_mixed_int16	Snapdragon® X2 Elite	55.106 ms	79 - 79 MB	NPU
VIT	ONNX	w8a8_mixed_int16	Snapdragon® X Elite	203.253 ms	79 - 79 MB	NPU
VIT	ONNX	w8a8_mixed_int16	Snapdragon® 8 Gen 3 Mobile	82.072 ms	69 - 445 MB	NPU
VIT	ONNX	w8a8_mixed_int16	Qualcomm® QCS6490	720.944 ms	96 - 124 MB	CPU
VIT	ONNX	w8a8_mixed_int16	Qualcomm® QCS8550 (Proxy)	105.062 ms	10 - 95 MB	NPU
VIT	ONNX	w8a8_mixed_int16	Qualcomm® QCS9075	131.1 ms	68 - 71 MB	NPU
VIT	ONNX	w8a8_mixed_int16	Qualcomm® QCM6690	389.169 ms	44 - 63 MB	CPU
VIT	ONNX	w8a8_mixed_int16	Snapdragon® 8 Elite For Galaxy Mobile	73.051 ms	68 - 338 MB	NPU
VIT	ONNX	w8a8_mixed_int16	Snapdragon® 7 Gen 4 Mobile	372.237 ms	92 - 114 MB	CPU
VIT	ONNX	w8a8_mixed_int16	Snapdragon® 8 Elite Gen 5 Mobile	62.988 ms	0 - 268 MB	NPU
VIT	QNN_DLC	float	Snapdragon® X2 Elite	4.494 ms	1 - 1 MB	NPU
VIT	QNN_DLC	float	Snapdragon® X Elite	11.849 ms	1 - 1 MB	NPU
VIT	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	7.714 ms	1 - 371 MB	NPU
VIT	QNN_DLC	float	Qualcomm® QCS8275 (Proxy)	40.621 ms	1 - 340 MB	NPU
VIT	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	11.137 ms	1 - 3 MB	NPU
VIT	QNN_DLC	float	Qualcomm® SA8775P	13.894 ms	1 - 339 MB	NPU
VIT	QNN_DLC	float	Qualcomm® QCS9075	15.323 ms	1 - 3 MB	NPU
VIT	QNN_DLC	float	Qualcomm® QCS8450 (Proxy)	19.175 ms	0 - 351 MB	NPU
VIT	QNN_DLC	float	Qualcomm® SA7255P	40.621 ms	1 - 340 MB	NPU
VIT	QNN_DLC	float	Qualcomm® SA8295P	17.18 ms	1 - 333 MB	NPU
VIT	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	5.303 ms	1 - 339 MB	NPU
VIT	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	4.048 ms	0 - 345 MB	NPU
VIT	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	5.872 ms	0 - 319 MB	NPU
VIT	TFLITE	float	Qualcomm® QCS8275 (Proxy)	35.901 ms	0 - 288 MB	NPU
VIT	TFLITE	float	Qualcomm® QCS8550 (Proxy)	8.009 ms	0 - 3 MB	NPU
VIT	TFLITE	float	Qualcomm® SA8775P	11.166 ms	0 - 288 MB	NPU
VIT	TFLITE	float	Qualcomm® QCS9075	11.738 ms	0 - 174 MB	NPU
VIT	TFLITE	float	Qualcomm® QCS8450 (Proxy)	13.889 ms	0 - 291 MB	NPU
VIT	TFLITE	float	Qualcomm® SA7255P	35.901 ms	0 - 288 MB	NPU
VIT	TFLITE	float	Qualcomm® SA8295P	13.347 ms	0 - 261 MB	NPU
VIT	TFLITE	float	Snapdragon® 8 Elite For Galaxy Mobile	3.933 ms	0 - 288 MB	NPU
VIT	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	3.092 ms	0 - 279 MB	NPU
VIT	TFLITE	w8a8	Snapdragon® 8 Gen 3 Mobile	4.722 ms	0 - 185 MB	NPU
VIT	TFLITE	w8a8	Qualcomm® QCS6490	66.292 ms	1 - 99 MB	NPU
VIT	TFLITE	w8a8	Qualcomm® QCS8275 (Proxy)	14.352 ms	0 - 85 MB	NPU
VIT	TFLITE	w8a8	Qualcomm® QCS8550 (Proxy)	6.76 ms	0 - 3 MB	NPU
VIT	TFLITE	w8a8	Qualcomm® SA8775P	7.083 ms	0 - 86 MB	NPU
VIT	TFLITE	w8a8	Qualcomm® QCS9075	7.561 ms	0 - 89 MB	NPU
VIT	TFLITE	w8a8	Qualcomm® QCM6690	97.501 ms	2 - 185 MB	NPU
VIT	TFLITE	w8a8	Qualcomm® QCS8450 (Proxy)	8.757 ms	0 - 180 MB	NPU
VIT	TFLITE	w8a8	Qualcomm® SA7255P	14.352 ms	0 - 85 MB	NPU
VIT	TFLITE	w8a8	Qualcomm® SA8295P	9.687 ms	0 - 89 MB	NPU
VIT	TFLITE	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	3.365 ms	0 - 86 MB	NPU
VIT	TFLITE	w8a8	Snapdragon® 7 Gen 4 Mobile	20.163 ms	2 - 75 MB	NPU
VIT	TFLITE	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	2.29 ms	0 - 90 MB	NPU

License

The license for the original implementation of 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/VIT

An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale

Paper • 2010.11929 • Published Oct 22, 2020 • 15