EfficientViT-b2-cls: Optimized for Qualcomm Devices

EfficientViT 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 EfficientViT-b2-cls 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
QNN_DLC	float	Universal	QAIRT 2.45	Download
TFLITE	float	Universal	QAIRT 2.45	Download

For more device-specific assets and performance metrics, visit EfficientViT-b2-cls 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 EfficientViT-b2-cls 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: 24.3M
• Model size (float): 92.9 MB

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
EfficientViT-b2-cls	ONNX	float	Snapdragon® X2 Elite	2.487 ms	2 - 2 MB	NPU
EfficientViT-b2-cls	ONNX	float	Snapdragon® X Elite	5.031 ms	49 - 49 MB	NPU
EfficientViT-b2-cls	ONNX	float	Snapdragon® 8 Gen 3 Mobile	3.202 ms	0 - 141 MB	NPU
EfficientViT-b2-cls	ONNX	float	Snapdragon® 8 Gen 1 Mobile	6.362 ms	1 - 141 MB	NPU
EfficientViT-b2-cls	ONNX	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	4.838 ms	1 - 41 MB	NPU
EfficientViT-b2-cls	ONNX	float	Qualcomm® QCS8450	6.362 ms	1 - 141 MB	NPU
EfficientViT-b2-cls	ONNX	float	Qualcomm® Dragonwing™ IQ-9075	5.486 ms	1 - 4 MB	NPU
EfficientViT-b2-cls	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	2.408 ms	0 - 69 MB	NPU
EfficientViT-b2-cls	ONNX	float	Snapdragon® 8 Elite Mobile	2.655 ms	0 - 68 MB	NPU
EfficientViT-b2-cls	ONNX	float	Qualcomm® Dragonwing™ Q-8750	2.655 ms	0 - 68 MB	NPU
EfficientViT-b2-cls	ONNX	float	Qualcomm® Dragonwing™ IQ-X7181	5.031 ms	49 - 49 MB	NPU
EfficientViT-b2-cls	QNN_DLC	float	Snapdragon® X2 Elite	2.947 ms	1 - 1 MB	NPU
EfficientViT-b2-cls	QNN_DLC	float	Snapdragon® X Elite	6.232 ms	1 - 1 MB	NPU
EfficientViT-b2-cls	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	3.715 ms	0 - 141 MB	NPU
EfficientViT-b2-cls	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	7.229 ms	0 - 144 MB	NPU
EfficientViT-b2-cls	QNN_DLC	float	Qualcomm® QCS8275	12.792 ms	1 - 67 MB	NPU
EfficientViT-b2-cls	QNN_DLC	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	5.449 ms	1 - 2 MB	NPU
EfficientViT-b2-cls	QNN_DLC	float	Qualcomm® SA8775P	6.749 ms	1 - 69 MB	NPU
EfficientViT-b2-cls	QNN_DLC	float	Qualcomm® SA8650P	6.749 ms	1 - 69 MB	NPU
EfficientViT-b2-cls	QNN_DLC	float	Qualcomm® SA8255P	6.749 ms	1 - 69 MB	NPU
EfficientViT-b2-cls	QNN_DLC	float	Qualcomm® QCS8450	7.229 ms	0 - 144 MB	NPU
EfficientViT-b2-cls	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-9075	6.078 ms	1 - 3 MB	NPU
EfficientViT-b2-cls	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	2.321 ms	0 - 72 MB	NPU
EfficientViT-b2-cls	QNN_DLC	float	Qualcomm® SA7255P	12.792 ms	1 - 67 MB	NPU
EfficientViT-b2-cls	QNN_DLC	float	Snapdragon® 8 Elite Mobile	2.769 ms	0 - 69 MB	NPU
EfficientViT-b2-cls	QNN_DLC	float	Qualcomm® SA8295P	7.372 ms	1 - 73 MB	NPU
EfficientViT-b2-cls	QNN_DLC	float	Qualcomm® Dragonwing™ Q-8750	2.769 ms	0 - 69 MB	NPU
EfficientViT-b2-cls	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-X7181	6.232 ms	1 - 1 MB	NPU
EfficientViT-b2-cls	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	3.718 ms	0 - 177 MB	NPU
EfficientViT-b2-cls	TFLITE	float	Snapdragon® 8 Gen 1 Mobile	7.222 ms	0 - 186 MB	NPU
EfficientViT-b2-cls	TFLITE	float	Qualcomm® QCS8275	12.826 ms	0 - 110 MB	NPU
EfficientViT-b2-cls	TFLITE	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	5.375 ms	0 - 3 MB	NPU
EfficientViT-b2-cls	TFLITE	float	Qualcomm® SA8775P	6.712 ms	0 - 111 MB	NPU
EfficientViT-b2-cls	TFLITE	float	Qualcomm® SA8650P	6.712 ms	0 - 111 MB	NPU
EfficientViT-b2-cls	TFLITE	float	Qualcomm® SA8255P	6.712 ms	0 - 111 MB	NPU
EfficientViT-b2-cls	TFLITE	float	Qualcomm® QCS8450	7.222 ms	0 - 186 MB	NPU
EfficientViT-b2-cls	TFLITE	float	Qualcomm® Dragonwing™ IQ-9075	6.085 ms	0 - 52 MB	NPU
EfficientViT-b2-cls	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	2.331 ms	0 - 119 MB	NPU
EfficientViT-b2-cls	TFLITE	float	Qualcomm® SA7255P	12.826 ms	0 - 110 MB	NPU
EfficientViT-b2-cls	TFLITE	float	Snapdragon® 8 Elite Mobile	2.757 ms	0 - 114 MB	NPU
EfficientViT-b2-cls	TFLITE	float	Qualcomm® SA8295P	7.418 ms	0 - 114 MB	NPU
EfficientViT-b2-cls	TFLITE	float	Qualcomm® Dragonwing™ Q-8750	2.757 ms	0 - 114 MB	NPU

License

• The license for the original implementation of EfficientViT-b2-cls can be found here.

References

• EfficientViT: Multi-Scale Linear Attention for High-Resolution Dense Prediction
• 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.