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 2.50, ONNX Runtime 1.27.1	Download
ONNX	w8a16	Universal	QAIRT 2.50, ONNX Runtime 1.27.1	Download
QNN_DLC	float	Universal	QAIRT 2.50	Download
QNN_DLC	w8a16	Universal	QAIRT 2.50	Download
TFLITE	float	Universal	QAIRT 2.50	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:

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

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
EfficientFormer	ONNX	float	Snapdragon® 8 Elite Gen 5 For Galaxy Mobile	0.572 ms	0 - 44 MB	NPU
EfficientFormer	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	0.647 ms	0 - 46 MB	NPU
EfficientFormer	ONNX	float	Snapdragon® X2 Elite	0.558 ms	2 - 2 MB	NPU
EfficientFormer	ONNX	float	Snapdragon® X Elite	1.2 ms	25 - 25 MB	NPU
EfficientFormer	ONNX	float	Snapdragon® 8 Gen 3 Mobile	0.835 ms	0 - 78 MB	NPU
EfficientFormer	ONNX	float	Snapdragon® 8 Gen 1 Mobile	4.996 ms	1 - 81 MB	NPU
EfficientFormer	ONNX	float	Qualcomm® Dragonwing™ IQ-8275	1.589 ms	1 - 5 MB	NPU
EfficientFormer	ONNX	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	1.166 ms	0 - 29 MB	NPU
EfficientFormer	ONNX	float	Qualcomm® QCS8450	4.996 ms	1 - 81 MB	NPU
EfficientFormer	ONNX	float	Qualcomm® Dragonwing™ IQ-9075	1.586 ms	1 - 4 MB	NPU
EfficientFormer	ONNX	float	Qualcomm® Dragonwing™ IQ-X7181	1.2 ms	25 - 25 MB	NPU
EfficientFormer	ONNX	float	Qualcomm® Dragonwing™ Q-8750	0.647 ms	0 - 46 MB	NPU
EfficientFormer	ONNX	w8a16	Snapdragon® 8 Elite Gen 5 For Galaxy Mobile	0.463 ms	0 - 62 MB	NPU
EfficientFormer	ONNX	w8a16	Snapdragon® 8 Elite For Galaxy Mobile	0.537 ms	0 - 62 MB	NPU
EfficientFormer	ONNX	w8a16	Snapdragon® X2 Elite	0.479 ms	1 - 1 MB	NPU
EfficientFormer	ONNX	w8a16	Snapdragon® X Elite	1.264 ms	13 - 13 MB	NPU
EfficientFormer	ONNX	w8a16	Snapdragon® 8 Gen 3 Mobile	0.82 ms	0 - 87 MB	NPU
EfficientFormer	ONNX	w8a16	Snapdragon® 8 Gen 1 Mobile	1.528 ms	0 - 92 MB	NPU
EfficientFormer	ONNX	w8a16	Qualcomm® Dragonwing™ QCS6490	2.986 ms	0 - 3 MB	NPU
EfficientFormer	ONNX	w8a16	Qualcomm® Dragonwing™ IQ-8275	1.228 ms	0 - 4 MB	NPU
EfficientFormer	ONNX	w8a16	Qualcomm® Dragonwing™ QCS8550 (Proxy)	1.219 ms	0 - 85 MB	NPU
EfficientFormer	ONNX	w8a16	Qualcomm® QCS8450	1.528 ms	0 - 92 MB	NPU
EfficientFormer	ONNX	w8a16	Qualcomm® Dragonwing™ IQ-9075	1.403 ms	0 - 3 MB	NPU
EfficientFormer	ONNX	w8a16	Qualcomm® Dragonwing™ IQ-X7181	1.264 ms	13 - 13 MB	NPU
EfficientFormer	ONNX	w8a16	Qualcomm® Dragonwing™ Q-6690	5.924 ms	0 - 71 MB	NPU
EfficientFormer	ONNX	w8a16	Qualcomm® Dragonwing™ Q-7790	1.357 ms	0 - 68 MB	NPU
EfficientFormer	ONNX	w8a16	Qualcomm® Dragonwing™ Q-8750	0.537 ms	0 - 62 MB	NPU
EfficientFormer	ONNX	w8a16	Snapdragon® 7 Gen 4 Mobile	1.357 ms	0 - 68 MB	NPU
EfficientFormer	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 For Galaxy Mobile	0.671 ms	1 - 45 MB	NPU
EfficientFormer	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	0.767 ms	0 - 41 MB	NPU
EfficientFormer	QNN_DLC	float	Snapdragon® X2 Elite	0.891 ms	1 - 1 MB	NPU
EfficientFormer	QNN_DLC	float	Snapdragon® X Elite	1.687 ms	1 - 1 MB	NPU
EfficientFormer	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	1.03 ms	0 - 78 MB	NPU
EfficientFormer	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	5.591 ms	0 - 80 MB	NPU
EfficientFormer	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-8275	1.824 ms	1 - 4 MB	NPU
EfficientFormer	QNN_DLC	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	1.471 ms	1 - 2 MB	NPU
EfficientFormer	QNN_DLC	float	Qualcomm® SA8775P	2.023 ms	5 - 52 MB	NPU
EfficientFormer	QNN_DLC	float	Qualcomm® SA8650P	2.023 ms	5 - 52 MB	NPU
EfficientFormer	QNN_DLC	float	Qualcomm® SA8255P	2.023 ms	5 - 52 MB	NPU
EfficientFormer	QNN_DLC	float	Qualcomm® QCS8450	5.591 ms	0 - 80 MB	NPU
EfficientFormer	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-9075	1.89 ms	3 - 5 MB	NPU
EfficientFormer	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-X7181	1.687 ms	1 - 1 MB	NPU
EfficientFormer	QNN_DLC	float	Qualcomm® Dragonwing™ Q-8750	0.767 ms	0 - 41 MB	NPU
EfficientFormer	QNN_DLC	float	Qualcomm® SA7255P	4.915 ms	1 - 44 MB	NPU
EfficientFormer	QNN_DLC	float	Qualcomm® SA8295P	3.881 ms	0 - 42 MB	NPU
EfficientFormer	QNN_DLC	w8a16	Snapdragon® 8 Elite Gen 5 For Galaxy Mobile	0.537 ms	0 - 56 MB	NPU
EfficientFormer	QNN_DLC	w8a16	Snapdragon® 8 Elite For Galaxy Mobile	0.641 ms	0 - 54 MB	NPU
EfficientFormer	QNN_DLC	w8a16	Snapdragon® X2 Elite	0.752 ms	0 - 0 MB	NPU
EfficientFormer	QNN_DLC	w8a16	Snapdragon® X Elite	1.68 ms	0 - 0 MB	NPU
EfficientFormer	QNN_DLC	w8a16	Snapdragon® 8 Gen 3 Mobile	1.014 ms	0 - 75 MB	NPU
EfficientFormer	QNN_DLC	w8a16	Qualcomm® Dragonwing™ IQ-8275	1.431 ms	0 - 3 MB	NPU
EfficientFormer	QNN_DLC	w8a16	Qualcomm® Dragonwing™ QCS8550 (Proxy)	1.479 ms	0 - 2 MB	NPU
EfficientFormer	QNN_DLC	w8a16	Qualcomm® SA8775P	1.761 ms	0 - 62 MB	NPU
EfficientFormer	QNN_DLC	w8a16	Qualcomm® SA8650P	1.761 ms	0 - 62 MB	NPU
EfficientFormer	QNN_DLC	w8a16	Qualcomm® SA8255P	1.761 ms	0 - 62 MB	NPU
EfficientFormer	QNN_DLC	w8a16	Qualcomm® Dragonwing™ IQ-9075	1.62 ms	0 - 3 MB	NPU
EfficientFormer	QNN_DLC	w8a16	Qualcomm® Dragonwing™ IQ-X7181	1.68 ms	0 - 0 MB	NPU
EfficientFormer	QNN_DLC	w8a16	Qualcomm® Dragonwing™ Q-6690	6.734 ms	0 - 173 MB	NPU
EfficientFormer	QNN_DLC	w8a16	Qualcomm® Dragonwing™ Q-7790	1.6 ms	0 - 62 MB	NPU
EfficientFormer	QNN_DLC	w8a16	Qualcomm® Dragonwing™ Q-8750	0.641 ms	0 - 54 MB	NPU
EfficientFormer	QNN_DLC	w8a16	Snapdragon® 7 Gen 4 Mobile	1.6 ms	0 - 62 MB	NPU
EfficientFormer	TFLITE	float	Snapdragon® 8 Elite Gen 5 For Galaxy Mobile	0.661 ms	0 - 53 MB	NPU
EfficientFormer	TFLITE	float	Snapdragon® 8 Elite For Galaxy Mobile	0.763 ms	0 - 58 MB	NPU
EfficientFormer	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	1.012 ms	0 - 92 MB	NPU
EfficientFormer	TFLITE	float	Snapdragon® 8 Gen 1 Mobile	5.529 ms	0 - 97 MB	NPU
EfficientFormer	TFLITE	float	Qualcomm® Dragonwing™ IQ-8275	1.827 ms	0 - 28 MB	NPU
EfficientFormer	TFLITE	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	1.455 ms	0 - 5 MB	NPU
EfficientFormer	TFLITE	float	Qualcomm® SA8775P	2.046 ms	0 - 56 MB	NPU
EfficientFormer	TFLITE	float	Qualcomm® SA8650P	2.046 ms	0 - 56 MB	NPU
EfficientFormer	TFLITE	float	Qualcomm® SA8255P	2.046 ms	0 - 56 MB	NPU
EfficientFormer	TFLITE	float	Qualcomm® QCS8450	5.529 ms	0 - 97 MB	NPU
EfficientFormer	TFLITE	float	Qualcomm® Dragonwing™ IQ-9075	1.865 ms	0 - 27 MB	NPU
EfficientFormer	TFLITE	float	Qualcomm® Dragonwing™ Q-8750	0.763 ms	0 - 58 MB	NPU
EfficientFormer	TFLITE	float	Qualcomm® SA7255P	4.877 ms	0 - 55 MB	NPU
EfficientFormer	TFLITE	float	Qualcomm® SA8295P	3.87 ms	0 - 51 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.