Swin-Tiny: Optimized for Qualcomm Devices

SwinTiny 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-Tiny 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
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-Tiny 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-Tiny 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: 28.8M
• Model size (float): 110 MB
• Model size (w8a16): 29.9 MB

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
Swin-Tiny	ONNX	float	Snapdragon® X2 Elite	3.973 ms	212 - 212 MB	NPU
Swin-Tiny	ONNX	float	Snapdragon® X Elite	9.602 ms	148 - 148 MB	NPU
Swin-Tiny	ONNX	float	Snapdragon® 8 Gen 3 Mobile	5.968 ms	1 - 242 MB	NPU
Swin-Tiny	ONNX	float	Snapdragon® 8 Gen 1 Mobile	14.193 ms	1 - 235 MB	NPU
Swin-Tiny	ONNX	float	Qualcomm® QCS8550 (Proxy)	9.105 ms	0 - 67 MB	NPU
Swin-Tiny	ONNX	float	Qualcomm® QCS8450	14.193 ms	1 - 235 MB	NPU
Swin-Tiny	ONNX	float	Snapdragon® 8 Elite Mobile	4.584 ms	0 - 435 MB	NPU
Swin-Tiny	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	3.767 ms	1 - 429 MB	NPU
Swin-Tiny	ONNX	float	Qualcomm® QCS9075	10.99 ms	1 - 46 MB	NPU
Swin-Tiny	ONNX	float	Qualcomm® QCS8750	4.584 ms	0 - 435 MB	NPU
Swin-Tiny	ONNX	float	Qualcomm® QCS7181	9.602 ms	148 - 148 MB	NPU
Swin-Tiny	ONNX	w8a16	Snapdragon® X2 Elite	3.471 ms	212 - 212 MB	NPU
Swin-Tiny	ONNX	w8a16	Snapdragon® X Elite	8.23 ms	149 - 149 MB	NPU
Swin-Tiny	ONNX	w8a16	Snapdragon® 8 Gen 3 Mobile	5.164 ms	0 - 323 MB	NPU
Swin-Tiny	ONNX	w8a16	Qualcomm® QCS8550 (Proxy)	7.82 ms	0 - 36 MB	NPU
Swin-Tiny	ONNX	w8a16	Qualcomm® QCS9075	9.363 ms	0 - 45 MB	NPU
Swin-Tiny	ONNX	w8a16	Snapdragon® 8 Elite Gen 5 Mobile	3.263 ms	0 - 271 MB	NPU
Swin-Tiny	ONNX	w8a16	Snapdragon® 8 Elite Mobile	3.979 ms	0 - 267 MB	NPU
Swin-Tiny	ONNX	w8a16	Snapdragon® 7 Gen 4 Mobile	8.254 ms	0 - 234 MB	NPU
Swin-Tiny	ONNX	w8a16	Qualcomm® QCM6690	35.976 ms	0 - 383 MB	NPU
Swin-Tiny	ONNX	w8a16	Qualcomm® QCS7790	8.254 ms	0 - 234 MB	NPU
Swin-Tiny	ONNX	w8a16	Qualcomm® QCS8750	3.979 ms	0 - 267 MB	NPU
Swin-Tiny	ONNX	w8a16	Qualcomm® QCS7181	8.23 ms	149 - 149 MB	NPU
Swin-Tiny	QNN_DLC	float	Snapdragon® X2 Elite	4.259 ms	1 - 1 MB	NPU
Swin-Tiny	QNN_DLC	float	Snapdragon® X Elite	9.885 ms	1 - 1 MB	NPU
Swin-Tiny	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	5.97 ms	0 - 245 MB	NPU
Swin-Tiny	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	14.538 ms	0 - 242 MB	NPU
Swin-Tiny	QNN_DLC	float	Qualcomm® QCS8275	21.241 ms	1 - 176 MB	NPU
Swin-Tiny	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	9.049 ms	1 - 3 MB	NPU
Swin-Tiny	QNN_DLC	float	Qualcomm® QCS8450	14.538 ms	0 - 242 MB	NPU
Swin-Tiny	QNN_DLC	float	Snapdragon® 8 Elite Mobile	4.529 ms	0 - 414 MB	NPU
Swin-Tiny	QNN_DLC	float	Qualcomm® SA8295P	13.797 ms	0 - 433 MB	NPU
Swin-Tiny	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	3.762 ms	1 - 414 MB	NPU
Swin-Tiny	QNN_DLC	float	Qualcomm® SA7255P	21.241 ms	1 - 176 MB	NPU
Swin-Tiny	QNN_DLC	float	Qualcomm® QCS9075	11.446 ms	1 - 3 MB	NPU
Swin-Tiny	QNN_DLC	float	Qualcomm® QCS8750	4.529 ms	0 - 414 MB	NPU
Swin-Tiny	QNN_DLC	float	Qualcomm® QCS7181	9.885 ms	1 - 1 MB	NPU
Swin-Tiny	QNN_DLC	w8a16	Snapdragon® X2 Elite	4.429 ms	0 - 0 MB	NPU
Swin-Tiny	QNN_DLC	w8a16	Snapdragon® X Elite	10.91 ms	0 - 0 MB	NPU
Swin-Tiny	QNN_DLC	w8a16	Snapdragon® 8 Gen 3 Mobile	6.59 ms	0 - 506 MB	NPU
Swin-Tiny	QNN_DLC	w8a16	Qualcomm® QCS8275	17.761 ms	0 - 436 MB	NPU
Swin-Tiny	QNN_DLC	w8a16	Qualcomm® QCS8550 (Proxy)	10.192 ms	0 - 31 MB	NPU
Swin-Tiny	QNN_DLC	w8a16	Qualcomm® QCS9075	11.509 ms	0 - 2 MB	NPU
Swin-Tiny	QNN_DLC	w8a16	Snapdragon® 8 Elite Gen 5 Mobile	3.885 ms	0 - 438 MB	NPU
Swin-Tiny	QNN_DLC	w8a16	Snapdragon® 8 Elite Mobile	4.859 ms	0 - 428 MB	NPU
Swin-Tiny	QNN_DLC	w8a16	Snapdragon® 7 Gen 4 Mobile	10.563 ms	0 - 436 MB	NPU
Swin-Tiny	QNN_DLC	w8a16	Qualcomm® QCM6690	43.878 ms	0 - 424 MB	NPU
Swin-Tiny	QNN_DLC	w8a16	Qualcomm® SA7255P	17.761 ms	0 - 436 MB	NPU
Swin-Tiny	QNN_DLC	w8a16	Qualcomm® QCS7790	10.563 ms	0 - 436 MB	NPU
Swin-Tiny	QNN_DLC	w8a16	Qualcomm® QCS8750	4.859 ms	0 - 428 MB	NPU
Swin-Tiny	QNN_DLC	w8a16	Qualcomm® QCS7181	10.91 ms	0 - 0 MB	NPU
Swin-Tiny	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	6.024 ms	0 - 272 MB	NPU
Swin-Tiny	TFLITE	float	Snapdragon® 8 Gen 1 Mobile	14.301 ms	0 - 258 MB	NPU
Swin-Tiny	TFLITE	float	Qualcomm® QCS8275	21.21 ms	0 - 201 MB	NPU
Swin-Tiny	TFLITE	float	Qualcomm® QCS8550 (Proxy)	9.261 ms	0 - 3 MB	NPU
Swin-Tiny	TFLITE	float	Qualcomm® SA8775P	38.271 ms	0 - 30 MB	GPU
Swin-Tiny	TFLITE	float	Qualcomm® SA8650P	38.271 ms	0 - 30 MB	GPU
Swin-Tiny	TFLITE	float	Qualcomm® SA8255P	38.271 ms	0 - 30 MB	GPU
Swin-Tiny	TFLITE	float	Qualcomm® QCS8450	14.301 ms	0 - 258 MB	NPU
Swin-Tiny	TFLITE	float	Snapdragon® 8 Elite Mobile	4.639 ms	0 - 196 MB	NPU
Swin-Tiny	TFLITE	float	Qualcomm® SA8295P	13.905 ms	0 - 198 MB	NPU
Swin-Tiny	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	3.83 ms	0 - 214 MB	NPU
Swin-Tiny	TFLITE	float	Qualcomm® SA7255P	21.21 ms	0 - 201 MB	NPU
Swin-Tiny	TFLITE	float	Qualcomm® QCS9075	11.387 ms	0 - 60 MB	NPU
Swin-Tiny	TFLITE	float	Qualcomm® QCS8750	4.639 ms	0 - 196 MB	NPU

License

• The license for the original implementation of Swin-Tiny 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.