EfficientViT-l2-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-l2-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-l2-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-l2-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: 63.7M
• Model size (float): 243 MB

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
EfficientViT-l2-cls	ONNX	float	Snapdragon® X2 Elite	3.21 ms	212 - 212 MB	NPU
EfficientViT-l2-cls	ONNX	float	Snapdragon® X Elite	6.616 ms	148 - 148 MB	NPU
EfficientViT-l2-cls	ONNX	float	Snapdragon® 8 Gen 3 Mobile	4.463 ms	0 - 212 MB	NPU
EfficientViT-l2-cls	ONNX	float	Snapdragon® 8 Gen 1 Mobile	13.655 ms	1 - 185 MB	NPU
EfficientViT-l2-cls	ONNX	float	Qualcomm® QCS8550 (Proxy)	6.488 ms	0 - 163 MB	NPU
EfficientViT-l2-cls	ONNX	float	Qualcomm® QCS8450	13.655 ms	1 - 185 MB	NPU
EfficientViT-l2-cls	ONNX	float	Snapdragon® 8 Elite Mobile	3.622 ms	0 - 117 MB	NPU
EfficientViT-l2-cls	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	3.007 ms	1 - 112 MB	NPU
EfficientViT-l2-cls	ONNX	float	Qualcomm® QCS9075	7.902 ms	1 - 46 MB	NPU
EfficientViT-l2-cls	ONNX	float	Qualcomm® QCS8750	3.622 ms	0 - 117 MB	NPU
EfficientViT-l2-cls	ONNX	float	Qualcomm® QCS7181	6.616 ms	148 - 148 MB	NPU
EfficientViT-l2-cls	QNN_DLC	float	Snapdragon® X2 Elite	3.938 ms	1 - 1 MB	NPU
EfficientViT-l2-cls	QNN_DLC	float	Snapdragon® X Elite	8.11 ms	1 - 1 MB	NPU
EfficientViT-l2-cls	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	5.235 ms	0 - 222 MB	NPU
EfficientViT-l2-cls	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	14.904 ms	0 - 190 MB	NPU
EfficientViT-l2-cls	QNN_DLC	float	Qualcomm® QCS8275	24.362 ms	1 - 108 MB	NPU
EfficientViT-l2-cls	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	7.774 ms	1 - 2 MB	NPU
EfficientViT-l2-cls	QNN_DLC	float	Qualcomm® QCS8450	14.904 ms	0 - 190 MB	NPU
EfficientViT-l2-cls	QNN_DLC	float	Snapdragon® 8 Elite Mobile	3.965 ms	0 - 117 MB	NPU
EfficientViT-l2-cls	QNN_DLC	float	Qualcomm® SA8295P	14.71 ms	1 - 97 MB	NPU
EfficientViT-l2-cls	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	3.213 ms	1 - 115 MB	NPU
EfficientViT-l2-cls	QNN_DLC	float	Qualcomm® SA7255P	24.362 ms	1 - 108 MB	NPU
EfficientViT-l2-cls	QNN_DLC	float	Qualcomm® QCS9075	8.564 ms	1 - 3 MB	NPU
EfficientViT-l2-cls	QNN_DLC	float	Qualcomm® QCS8750	3.965 ms	0 - 117 MB	NPU
EfficientViT-l2-cls	QNN_DLC	float	Qualcomm® QCS7181	8.11 ms	1 - 1 MB	NPU
EfficientViT-l2-cls	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	5.205 ms	0 - 296 MB	NPU
EfficientViT-l2-cls	TFLITE	float	Snapdragon® 8 Gen 1 Mobile	14.86 ms	0 - 277 MB	NPU
EfficientViT-l2-cls	TFLITE	float	Qualcomm® QCS8275	24.326 ms	0 - 179 MB	NPU
EfficientViT-l2-cls	TFLITE	float	Qualcomm® QCS8550 (Proxy)	7.825 ms	0 - 3 MB	NPU
EfficientViT-l2-cls	TFLITE	float	Qualcomm® SA8775P	40.994 ms	0 - 30 MB	GPU
EfficientViT-l2-cls	TFLITE	float	Qualcomm® SA8650P	40.994 ms	0 - 30 MB	GPU
EfficientViT-l2-cls	TFLITE	float	Qualcomm® SA8255P	40.994 ms	0 - 30 MB	GPU
EfficientViT-l2-cls	TFLITE	float	Qualcomm® QCS8450	14.86 ms	0 - 277 MB	NPU
EfficientViT-l2-cls	TFLITE	float	Snapdragon® 8 Elite Mobile	3.96 ms	0 - 183 MB	NPU
EfficientViT-l2-cls	TFLITE	float	Qualcomm® SA8295P	14.653 ms	0 - 168 MB	NPU
EfficientViT-l2-cls	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	3.215 ms	0 - 187 MB	NPU
EfficientViT-l2-cls	TFLITE	float	Qualcomm® SA7255P	24.326 ms	0 - 179 MB	NPU
EfficientViT-l2-cls	TFLITE	float	Qualcomm® QCS9075	8.48 ms	0 - 134 MB	NPU
EfficientViT-l2-cls	TFLITE	float	Qualcomm® QCS8750	3.96 ms	0 - 183 MB	NPU

License

• The license for the original implementation of EfficientViT-l2-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.