Swin-Small: Optimized for Qualcomm Devices

SwinSmall 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 Swin-Small 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 1.25.0	Download
ONNX	w8a16	Universal	QAIRT 2.42, ONNX Runtime 1.25.0	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 Swin-Small 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 Swin-Small 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: 50.4M
• Model size (float): 193 MB
• Model size (w8a16): 52.5 MB

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
Swin-Small	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	6.327 ms	1 - 636 MB	NPU
Swin-Small	ONNX	float	Snapdragon® X2 Elite	6.806 ms	211 - 211 MB	NPU
Swin-Small	ONNX	float	Snapdragon® X Elite	15.983 ms	148 - 148 MB	NPU
Swin-Small	ONNX	float	Snapdragon® 8 Gen 3 Mobile	10.165 ms	1 - 1029 MB	NPU
Swin-Small	ONNX	float	Qualcomm® QCS8550 (Proxy)	15.302 ms	0 - 123 MB	NPU
Swin-Small	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	7.676 ms	1 - 625 MB	NPU
Swin-Small	ONNX	float	Qualcomm® QCS9075	19.672 ms	1 - 46 MB	NPU
Swin-Small	ONNX	float	Qualcomm® QCS8750	7.676 ms	1 - 625 MB	NPU
Swin-Small	ONNX	float	Qualcomm® QCS7181	15.983 ms	148 - 148 MB	NPU
Swin-Small	ONNX	w8a16	Snapdragon® 8 Elite Gen 5 Mobile	5.884 ms	0 - 377 MB	NPU
Swin-Small	ONNX	w8a16	Snapdragon® X2 Elite	6.326 ms	212 - 212 MB	NPU
Swin-Small	ONNX	w8a16	Snapdragon® X Elite	14.42 ms	181 - 181 MB	NPU
Swin-Small	ONNX	w8a16	Snapdragon® 8 Gen 3 Mobile	9.24 ms	3 - 545 MB	NPU
Swin-Small	ONNX	w8a16	Qualcomm® QCS6490	761.587 ms	98 - 111 MB	CPU
Swin-Small	ONNX	w8a16	Qualcomm® QCS8550 (Proxy)	13.91 ms	0 - 71 MB	NPU
Swin-Small	ONNX	w8a16	Qualcomm® QCM6690	371.408 ms	76 - 95 MB	CPU
Swin-Small	ONNX	w8a16	Snapdragon® 7 Gen 4 Mobile	349.419 ms	91 - 112 MB	CPU
Swin-Small	ONNX	w8a16	Snapdragon® 8 Elite For Galaxy Mobile	7.286 ms	0 - 448 MB	NPU
Swin-Small	ONNX	w8a16	Qualcomm® QCS9075	16.996 ms	0 - 45 MB	NPU
Swin-Small	ONNX	w8a16	Qualcomm® QCS7790	349.419 ms	91 - 112 MB	CPU
Swin-Small	ONNX	w8a16	Qualcomm® QCS8750	7.286 ms	0 - 448 MB	NPU
Swin-Small	ONNX	w8a16	Qualcomm® QCS7181	14.42 ms	181 - 181 MB	NPU
Swin-Small	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	6.219 ms	1 - 596 MB	NPU
Swin-Small	QNN_DLC	float	Snapdragon® X2 Elite	7.071 ms	1 - 1 MB	NPU
Swin-Small	QNN_DLC	float	Snapdragon® X Elite	16.362 ms	1 - 1 MB	NPU
Swin-Small	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	9.927 ms	0 - 404 MB	NPU
Swin-Small	QNN_DLC	float	Qualcomm® QCS8275	37.906 ms	1 - 276 MB	NPU
Swin-Small	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	15.337 ms	1 - 299 MB	NPU
Swin-Small	QNN_DLC	float	Qualcomm® SA8775P	17.396 ms	1 - 593 MB	NPU
Swin-Small	QNN_DLC	float	Qualcomm® SA8650P	17.396 ms	1 - 593 MB	NPU
Swin-Small	QNN_DLC	float	Qualcomm® SA8255P	17.396 ms	1 - 593 MB	NPU
Swin-Small	QNN_DLC	float	Qualcomm® QCS8450 (Proxy)	23.99 ms	0 - 397 MB	NPU
Swin-Small	QNN_DLC	float	Qualcomm® SA7255P	37.906 ms	1 - 276 MB	NPU
Swin-Small	QNN_DLC	float	Qualcomm® SA8295P	22.637 ms	1 - 269 MB	NPU
Swin-Small	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	7.594 ms	1 - 587 MB	NPU
Swin-Small	QNN_DLC	float	Qualcomm® QCS9075	19.704 ms	1 - 3 MB	NPU
Swin-Small	QNN_DLC	float	Qualcomm® QCS8750	7.594 ms	1 - 587 MB	NPU
Swin-Small	QNN_DLC	float	Qualcomm® QCS7181	16.362 ms	1 - 1 MB	NPU
Swin-Small	QNN_DLC	w8a16	Snapdragon® 8 Elite Gen 5 Mobile	6.427 ms	0 - 635 MB	NPU
Swin-Small	QNN_DLC	w8a16	Snapdragon® X2 Elite	7.176 ms	0 - 0 MB	NPU
Swin-Small	QNN_DLC	w8a16	Snapdragon® X Elite	17.26 ms	0 - 0 MB	NPU
Swin-Small	QNN_DLC	w8a16	Snapdragon® 8 Gen 3 Mobile	10.768 ms	0 - 865 MB	NPU
Swin-Small	QNN_DLC	w8a16	Qualcomm® QCS8275	29.078 ms	0 - 635 MB	NPU
Swin-Small	QNN_DLC	w8a16	Qualcomm® QCS8550 (Proxy)	16.212 ms	0 - 2 MB	NPU
Swin-Small	QNN_DLC	w8a16	Qualcomm® SA8775P	16.714 ms	0 - 639 MB	NPU
Swin-Small	QNN_DLC	w8a16	Qualcomm® SA8650P	16.714 ms	0 - 639 MB	NPU
Swin-Small	QNN_DLC	w8a16	Qualcomm® SA8255P	16.714 ms	0 - 639 MB	NPU
Swin-Small	QNN_DLC	w8a16	Qualcomm® QCM6690	70.847 ms	0 - 626 MB	NPU
Swin-Small	QNN_DLC	w8a16	Qualcomm® SA7255P	29.078 ms	0 - 635 MB	NPU
Swin-Small	QNN_DLC	w8a16	Snapdragon® 7 Gen 4 Mobile	16.99 ms	0 - 655 MB	NPU
Swin-Small	QNN_DLC	w8a16	Snapdragon® 8 Elite For Galaxy Mobile	8.115 ms	0 - 623 MB	NPU
Swin-Small	QNN_DLC	w8a16	Qualcomm® QCS9075	18.94 ms	0 - 2 MB	NPU
Swin-Small	QNN_DLC	w8a16	Qualcomm® QCS7790	16.99 ms	0 - 655 MB	NPU
Swin-Small	QNN_DLC	w8a16	Qualcomm® QCS8750	8.115 ms	0 - 623 MB	NPU
Swin-Small	QNN_DLC	w8a16	Qualcomm® QCS7181	17.26 ms	0 - 0 MB	NPU
Swin-Small	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	6.549 ms	0 - 320 MB	NPU
Swin-Small	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	10.147 ms	0 - 435 MB	NPU
Swin-Small	TFLITE	float	Qualcomm® QCS8275	37.41 ms	0 - 306 MB	NPU
Swin-Small	TFLITE	float	Qualcomm® QCS8550 (Proxy)	15.478 ms	0 - 4 MB	NPU
Swin-Small	TFLITE	float	Qualcomm® SA8775P	17.462 ms	0 - 305 MB	NPU
Swin-Small	TFLITE	float	Qualcomm® SA8650P	17.462 ms	0 - 305 MB	NPU
Swin-Small	TFLITE	float	Qualcomm® SA8255P	17.462 ms	0 - 305 MB	NPU
Swin-Small	TFLITE	float	Qualcomm® QCS8450 (Proxy)	23.411 ms	0 - 411 MB	NPU
Swin-Small	TFLITE	float	Qualcomm® SA7255P	37.41 ms	0 - 306 MB	NPU
Swin-Small	TFLITE	float	Qualcomm® SA8295P	22.965 ms	0 - 297 MB	NPU
Swin-Small	TFLITE	float	Snapdragon® 8 Elite For Galaxy Mobile	7.824 ms	0 - 304 MB	NPU
Swin-Small	TFLITE	float	Qualcomm® QCS9075	20.027 ms	0 - 104 MB	NPU
Swin-Small	TFLITE	float	Qualcomm® QCS8750	7.824 ms	0 - 304 MB	NPU

License

• The license for the original implementation of Swin-Small can be found here.

References

• Swin Transformer: Hierarchical Vision Transformer using Shifted Windows
• Source Model Implementation

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.