EfficientFormer: Optimized for Qualcomm Devices

EfficientFormer is a vision transformer model that can classify images from the Imagenet dataset.

This is based on the implementation of EfficientFormer 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 EfficientFormer 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 EfficientFormer on GitHub for usage instructions.

Model Details

Model Type: Model_use_case.image_classification

Model Stats:

• Model checkpoint: efficientformer_l1_300d
• Input resolution: 224x224
• Number of parameters: 12.3M
• Model size (float): 46.9 MB
• Model size (w8a16): 12.2 MB

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
EfficientFormer	ONNX	float	Snapdragon® X2 Elite	0.639 ms	212 - 212 MB	NPU
EfficientFormer	ONNX	float	Snapdragon® X Elite	1.359 ms	181 - 181 MB	NPU
EfficientFormer	ONNX	float	Snapdragon® 8 Gen 3 Mobile	0.915 ms	0 - 70 MB	NPU
EfficientFormer	ONNX	float	Snapdragon® 8 Gen 1 Mobile	5.103 ms	1 - 74 MB	NPU
EfficientFormer	ONNX	float	Qualcomm® QCS8550 (Proxy)	1.319 ms	1 - 44 MB	NPU
EfficientFormer	ONNX	float	Qualcomm® QCS8450	5.103 ms	1 - 74 MB	NPU
EfficientFormer	ONNX	float	Snapdragon® 8 Elite Mobile	0.703 ms	0 - 45 MB	NPU
EfficientFormer	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	0.609 ms	1 - 40 MB	NPU
EfficientFormer	ONNX	float	Qualcomm® QCS9075	1.867 ms	1 - 46 MB	NPU
EfficientFormer	ONNX	float	Qualcomm® QCS8750	0.703 ms	0 - 45 MB	NPU
EfficientFormer	ONNX	float	Qualcomm® QCS7181	1.359 ms	181 - 181 MB	NPU
EfficientFormer	ONNX	w8a16	Snapdragon® X2 Elite	0.563 ms	212 - 212 MB	NPU
EfficientFormer	ONNX	w8a16	Snapdragon® X Elite	1.365 ms	181 - 181 MB	NPU
EfficientFormer	ONNX	w8a16	Snapdragon® 8 Gen 3 Mobile	0.895 ms	0 - 82 MB	NPU
EfficientFormer	ONNX	w8a16	Snapdragon® 8 Gen 1 Mobile	2.245 ms	0 - 76 MB	NPU
EfficientFormer	ONNX	w8a16	Qualcomm® QCS6490	4.292 ms	1 - 46 MB	NPU
EfficientFormer	ONNX	w8a16	Qualcomm® QCS8550 (Proxy)	1.36 ms	0 - 4 MB	NPU
EfficientFormer	ONNX	w8a16	Qualcomm® QCS8450	2.245 ms	0 - 76 MB	NPU
EfficientFormer	ONNX	w8a16	Qualcomm® QCS9075	1.461 ms	0 - 45 MB	NPU
EfficientFormer	ONNX	w8a16	Snapdragon® 8 Elite Gen 5 Mobile	0.529 ms	0 - 66 MB	NPU
EfficientFormer	ONNX	w8a16	Snapdragon® 8 Elite Mobile	0.602 ms	0 - 63 MB	NPU
EfficientFormer	ONNX	w8a16	Snapdragon® 7 Gen 4 Mobile	1.473 ms	0 - 65 MB	NPU
EfficientFormer	ONNX	w8a16	Qualcomm® QCM6690	6.213 ms	0 - 182 MB	NPU
EfficientFormer	ONNX	w8a16	Qualcomm® QCS7790	1.473 ms	0 - 65 MB	NPU
EfficientFormer	ONNX	w8a16	Qualcomm® QCS8750	0.602 ms	0 - 63 MB	NPU
EfficientFormer	ONNX	w8a16	Qualcomm® QCS7181	1.365 ms	181 - 181 MB	NPU
EfficientFormer	QNN_DLC	float	Snapdragon® X2 Elite	0.918 ms	1 - 1 MB	NPU
EfficientFormer	QNN_DLC	float	Snapdragon® X Elite	1.697 ms	1 - 1 MB	NPU
EfficientFormer	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	1.034 ms	0 - 75 MB	NPU
EfficientFormer	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	5.588 ms	0 - 79 MB	NPU
EfficientFormer	QNN_DLC	float	Qualcomm® QCS8275	4.889 ms	1 - 40 MB	NPU
EfficientFormer	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	1.511 ms	1 - 2 MB	NPU
EfficientFormer	QNN_DLC	float	Qualcomm® QCS8450	5.588 ms	0 - 79 MB	NPU
EfficientFormer	QNN_DLC	float	Snapdragon® 8 Elite Mobile	0.772 ms	0 - 40 MB	NPU
EfficientFormer	QNN_DLC	float	Qualcomm® SA8295P	3.995 ms	0 - 39 MB	NPU
EfficientFormer	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	0.652 ms	1 - 44 MB	NPU
EfficientFormer	QNN_DLC	float	Qualcomm® SA7255P	4.889 ms	1 - 40 MB	NPU
EfficientFormer	QNN_DLC	float	Qualcomm® QCS9075	1.954 ms	1 - 3 MB	NPU
EfficientFormer	QNN_DLC	float	Qualcomm® QCS8750	0.772 ms	0 - 40 MB	NPU
EfficientFormer	QNN_DLC	float	Qualcomm® QCS7181	1.697 ms	1 - 1 MB	NPU
EfficientFormer	QNN_DLC	w8a16	Snapdragon® X2 Elite	0.852 ms	0 - 0 MB	NPU
EfficientFormer	QNN_DLC	w8a16	Snapdragon® X Elite	1.786 ms	0 - 0 MB	NPU
EfficientFormer	QNN_DLC	w8a16	Snapdragon® 8 Gen 3 Mobile	1.076 ms	0 - 80 MB	NPU
EfficientFormer	QNN_DLC	w8a16	Qualcomm® QCS8275	3.232 ms	0 - 58 MB	NPU
EfficientFormer	QNN_DLC	w8a16	Qualcomm® QCS8550 (Proxy)	1.564 ms	0 - 2 MB	NPU
EfficientFormer	QNN_DLC	w8a16	Qualcomm® QCS9075	1.766 ms	0 - 2 MB	NPU
EfficientFormer	QNN_DLC	w8a16	Snapdragon® 8 Elite Gen 5 Mobile	0.614 ms	0 - 66 MB	NPU
EfficientFormer	QNN_DLC	w8a16	Snapdragon® 8 Elite Mobile	0.71 ms	0 - 55 MB	NPU
EfficientFormer	QNN_DLC	w8a16	Snapdragon® 7 Gen 4 Mobile	1.713 ms	0 - 62 MB	NPU
EfficientFormer	QNN_DLC	w8a16	Qualcomm® QCM6690	6.995 ms	0 - 178 MB	NPU
EfficientFormer	QNN_DLC	w8a16	Qualcomm® SA7255P	3.232 ms	0 - 58 MB	NPU
EfficientFormer	QNN_DLC	w8a16	Qualcomm® QCS7790	1.713 ms	0 - 62 MB	NPU
EfficientFormer	QNN_DLC	w8a16	Qualcomm® QCS8750	0.71 ms	0 - 55 MB	NPU
EfficientFormer	QNN_DLC	w8a16	Qualcomm® QCS7181	1.786 ms	0 - 0 MB	NPU
EfficientFormer	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	1.033 ms	0 - 96 MB	NPU
EfficientFormer	TFLITE	float	Snapdragon® 8 Gen 1 Mobile	5.546 ms	0 - 97 MB	NPU
EfficientFormer	TFLITE	float	Qualcomm® QCS8275	4.887 ms	0 - 52 MB	NPU
EfficientFormer	TFLITE	float	Qualcomm® QCS8550 (Proxy)	1.514 ms	0 - 16 MB	NPU
EfficientFormer	TFLITE	float	Qualcomm® SA8775P	8.986 ms	0 - 29 MB	GPU
EfficientFormer	TFLITE	float	Qualcomm® SA8650P	8.986 ms	0 - 29 MB	GPU
EfficientFormer	TFLITE	float	Qualcomm® SA8255P	8.986 ms	0 - 29 MB	GPU
EfficientFormer	TFLITE	float	Qualcomm® QCS8450	5.546 ms	0 - 97 MB	NPU
EfficientFormer	TFLITE	float	Snapdragon® 8 Elite Mobile	0.764 ms	0 - 49 MB	NPU
EfficientFormer	TFLITE	float	Qualcomm® SA8295P	3.995 ms	0 - 48 MB	NPU
EfficientFormer	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	0.653 ms	0 - 53 MB	NPU
EfficientFormer	TFLITE	float	Qualcomm® SA7255P	4.887 ms	0 - 52 MB	NPU
EfficientFormer	TFLITE	float	Qualcomm® QCS9075	1.95 ms	0 - 27 MB	NPU
EfficientFormer	TFLITE	float	Qualcomm® QCS8750	0.764 ms	0 - 49 MB	NPU

License

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

References

• Rethinking Vision Transformers for MobileNet Size and Speed
• 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.