Beit: Optimized for Qualcomm Devices

Beit 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 Beit 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 Beit 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 Beit 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: 92.0M
• Model size (float): 351 MB

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
Beit	ONNX	float	Snapdragon® X2 Elite	7.435 ms	212 - 212 MB	NPU
Beit	ONNX	float	Snapdragon® X Elite	15.243 ms	183 - 183 MB	NPU
Beit	ONNX	float	Snapdragon® 8 Gen 3 Mobile	10.457 ms	1 - 439 MB	NPU
Beit	ONNX	float	Snapdragon® 8 Gen 1 Mobile	17.929 ms	1 - 409 MB	NPU
Beit	ONNX	float	Qualcomm® QCS8550 (Proxy)	14.844 ms	0 - 195 MB	NPU
Beit	ONNX	float	Qualcomm® QCS8450	17.929 ms	1 - 409 MB	NPU
Beit	ONNX	float	Snapdragon® 8 Elite Mobile	8.521 ms	1 - 305 MB	NPU
Beit	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	7.153 ms	1 - 302 MB	NPU
Beit	ONNX	float	Qualcomm® QCS9075	18.251 ms	1 - 46 MB	NPU
Beit	ONNX	float	Qualcomm® QCS8750	8.521 ms	1 - 305 MB	NPU
Beit	ONNX	float	Qualcomm® QCS7181	15.243 ms	183 - 183 MB	NPU
Beit	ONNX	w8a16	Snapdragon® X2 Elite	2.667 ms	212 - 212 MB	NPU
Beit	ONNX	w8a16	Snapdragon® X Elite	6.899 ms	149 - 149 MB	NPU
Beit	ONNX	w8a16	Snapdragon® 8 Gen 3 Mobile	4.568 ms	0 - 413 MB	NPU
Beit	ONNX	w8a16	Qualcomm® QCS8550 (Proxy)	6.687 ms	0 - 99 MB	NPU
Beit	ONNX	w8a16	Qualcomm® QCS9075	6.861 ms	0 - 46 MB	NPU
Beit	ONNX	w8a16	Snapdragon® 8 Elite Gen 5 Mobile	2.374 ms	0 - 254 MB	NPU
Beit	ONNX	w8a16	Snapdragon® 8 Elite Mobile	3.61 ms	0 - 361 MB	NPU
Beit	ONNX	w8a16	Snapdragon® 7 Gen 4 Mobile	9.088 ms	0 - 407 MB	NPU
Beit	ONNX	w8a16	Qualcomm® QCM6690	112.753 ms	0 - 438 MB	NPU
Beit	ONNX	w8a16	Qualcomm® QCS7790	9.088 ms	0 - 407 MB	NPU
Beit	ONNX	w8a16	Qualcomm® QCS8750	3.61 ms	0 - 361 MB	NPU
Beit	ONNX	w8a16	Qualcomm® QCS7181	6.899 ms	149 - 149 MB	NPU
Beit	QNN_DLC	float	Snapdragon® X2 Elite	7.858 ms	1 - 1 MB	NPU
Beit	QNN_DLC	float	Snapdragon® X Elite	15.192 ms	1 - 1 MB	NPU
Beit	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	10.472 ms	0 - 394 MB	NPU
Beit	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	17.511 ms	0 - 383 MB	NPU
Beit	QNN_DLC	float	Qualcomm® QCS8275	47.994 ms	1 - 295 MB	NPU
Beit	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	14.427 ms	1 - 3 MB	NPU
Beit	QNN_DLC	float	Qualcomm® QCS8450	17.511 ms	0 - 383 MB	NPU
Beit	QNN_DLC	float	Snapdragon® 8 Elite Mobile	8.54 ms	1 - 300 MB	NPU
Beit	QNN_DLC	float	Qualcomm® SA8295P	15.324 ms	1 - 289 MB	NPU
Beit	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	7.175 ms	1 - 301 MB	NPU
Beit	QNN_DLC	float	Qualcomm® SA7255P	47.994 ms	1 - 295 MB	NPU
Beit	QNN_DLC	float	Qualcomm® QCS9075	18.266 ms	1 - 3 MB	NPU
Beit	QNN_DLC	float	Qualcomm® QCS8750	8.54 ms	1 - 300 MB	NPU
Beit	QNN_DLC	float	Qualcomm® QCS7181	15.192 ms	1 - 1 MB	NPU
Beit	QNN_DLC	w8a16	Snapdragon® X2 Elite	3.197 ms	0 - 0 MB	NPU
Beit	QNN_DLC	w8a16	Snapdragon® X Elite	7.314 ms	0 - 0 MB	NPU
Beit	QNN_DLC	w8a16	Snapdragon® 8 Gen 3 Mobile	4.652 ms	0 - 411 MB	NPU
Beit	QNN_DLC	w8a16	Qualcomm® QCS8275	15.107 ms	0 - 352 MB	NPU
Beit	QNN_DLC	w8a16	Qualcomm® QCS8550 (Proxy)	6.801 ms	0 - 2 MB	NPU
Beit	QNN_DLC	w8a16	Qualcomm® QCS9075	7.785 ms	2 - 4 MB	NPU
Beit	QNN_DLC	w8a16	Snapdragon® 8 Elite Gen 5 Mobile	2.542 ms	0 - 245 MB	NPU
Beit	QNN_DLC	w8a16	Snapdragon® 8 Elite Mobile	3.554 ms	0 - 350 MB	NPU
Beit	QNN_DLC	w8a16	Snapdragon® 7 Gen 4 Mobile	8.659 ms	0 - 395 MB	NPU
Beit	QNN_DLC	w8a16	Qualcomm® QCM6690	65.174 ms	0 - 428 MB	NPU
Beit	QNN_DLC	w8a16	Qualcomm® SA7255P	15.107 ms	0 - 352 MB	NPU
Beit	QNN_DLC	w8a16	Qualcomm® QCS7790	8.659 ms	0 - 395 MB	NPU
Beit	QNN_DLC	w8a16	Qualcomm® QCS8750	3.554 ms	0 - 350 MB	NPU
Beit	QNN_DLC	w8a16	Qualcomm® QCS7181	7.314 ms	0 - 0 MB	NPU
Beit	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	10.502 ms	0 - 405 MB	NPU
Beit	TFLITE	float	Snapdragon® 8 Gen 1 Mobile	17.568 ms	0 - 372 MB	NPU
Beit	TFLITE	float	Qualcomm® QCS8275	48.075 ms	0 - 301 MB	NPU
Beit	TFLITE	float	Qualcomm® QCS8550 (Proxy)	14.098 ms	0 - 539 MB	NPU
Beit	TFLITE	float	Qualcomm® SA8775P	130.183 ms	0 - 25 MB	GPU
Beit	TFLITE	float	Qualcomm® SA8650P	130.183 ms	0 - 25 MB	GPU
Beit	TFLITE	float	Qualcomm® SA8255P	130.183 ms	0 - 25 MB	GPU
Beit	TFLITE	float	Qualcomm® QCS8450	17.568 ms	0 - 372 MB	NPU
Beit	TFLITE	float	Snapdragon® 8 Elite Mobile	8.495 ms	0 - 311 MB	NPU
Beit	TFLITE	float	Qualcomm® SA8295P	15.315 ms	0 - 293 MB	NPU
Beit	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	7.181 ms	0 - 312 MB	NPU
Beit	TFLITE	float	Qualcomm® SA7255P	48.075 ms	0 - 301 MB	NPU
Beit	TFLITE	float	Qualcomm® QCS9075	18.638 ms	0 - 186 MB	NPU
Beit	TFLITE	float	Qualcomm® QCS8750	8.495 ms	0 - 311 MB	NPU

License

• The license for the original implementation of Beit can be found here.

References

• BEIT: BERT Pre-Training of Image Transformers
• 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.