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 2.50, 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 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:

Input resolution: 224x224
Model checkpoint: Imagenet
Model size (float): 243 MB
Number of parameters: 63.7M

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
EfficientViT-l2-cls	ONNX	float	Snapdragon® 8 Elite Gen 5 For Galaxy Mobile	5.736 ms	2 - 254 MB	NPU
EfficientViT-l2-cls	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	7.67 ms	2 - 247 MB	NPU
EfficientViT-l2-cls	ONNX	float	Snapdragon® X2 Elite	6.813 ms	2 - 2 MB	NPU
EfficientViT-l2-cls	ONNX	float	Snapdragon® X Elite	14.459 ms	131 - 131 MB	NPU
EfficientViT-l2-cls	ONNX	float	Snapdragon® 8 Gen 3 Mobile	9.92 ms	2 - 378 MB	NPU
EfficientViT-l2-cls	ONNX	float	Snapdragon® 8 Gen 1 Mobile	33.95 ms	2 - 354 MB	NPU
EfficientViT-l2-cls	ONNX	float	Qualcomm® Dragonwing™ IQ-8275	18.572 ms	2 - 7 MB	NPU
EfficientViT-l2-cls	ONNX	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	14.071 ms	0 - 161 MB	NPU
EfficientViT-l2-cls	ONNX	float	Qualcomm® QCS8450	33.95 ms	2 - 354 MB	NPU
EfficientViT-l2-cls	ONNX	float	Qualcomm® Dragonwing™ IQ-9075	18.438 ms	1 - 6 MB	NPU
EfficientViT-l2-cls	ONNX	float	Qualcomm® Dragonwing™ IQ-X7181	14.459 ms	131 - 131 MB	NPU
EfficientViT-l2-cls	ONNX	float	Qualcomm® Dragonwing™ Q-8750	7.67 ms	2 - 247 MB	NPU
EfficientViT-l2-cls	ONNX	w8a16	Snapdragon® 8 Elite Gen 5 For Galaxy Mobile	4.41 ms	0 - 217 MB	NPU
EfficientViT-l2-cls	ONNX	w8a16	Snapdragon® 8 Elite For Galaxy Mobile	5.548 ms	1 - 217 MB	NPU
EfficientViT-l2-cls	ONNX	w8a16	Snapdragon® X2 Elite	5.044 ms	2 - 2 MB	NPU
EfficientViT-l2-cls	ONNX	w8a16	Snapdragon® X Elite	12.109 ms	68 - 68 MB	NPU
EfficientViT-l2-cls	ONNX	w8a16	Snapdragon® 8 Gen 3 Mobile	7.513 ms	1 - 335 MB	NPU
EfficientViT-l2-cls	ONNX	w8a16	Snapdragon® 8 Gen 1 Mobile	14.403 ms	1 - 339 MB	NPU
EfficientViT-l2-cls	ONNX	w8a16	Qualcomm® Dragonwing™ QCS6490	37.676 ms	0 - 4 MB	NPU
EfficientViT-l2-cls	ONNX	w8a16	Qualcomm® Dragonwing™ IQ-8275	10.239 ms	1 - 6 MB	NPU
EfficientViT-l2-cls	ONNX	w8a16	Qualcomm® Dragonwing™ QCS8550 (Proxy)	11.6 ms	0 - 83 MB	NPU
EfficientViT-l2-cls	ONNX	w8a16	Qualcomm® QCS8450	14.403 ms	1 - 339 MB	NPU
EfficientViT-l2-cls	ONNX	w8a16	Qualcomm® Dragonwing™ IQ-9075	11.767 ms	0 - 4 MB	NPU
EfficientViT-l2-cls	ONNX	w8a16	Qualcomm® Dragonwing™ IQ-X7181	12.109 ms	68 - 68 MB	NPU
EfficientViT-l2-cls	ONNX	w8a16	Qualcomm® Dragonwing™ Q-6690	103.917 ms	0 - 424 MB	NPU
EfficientViT-l2-cls	ONNX	w8a16	Qualcomm® Dragonwing™ Q-7790	14.078 ms	1 - 406 MB	NPU
EfficientViT-l2-cls	ONNX	w8a16	Qualcomm® Dragonwing™ Q-8750	5.548 ms	1 - 217 MB	NPU
EfficientViT-l2-cls	ONNX	w8a16	Snapdragon® 7 Gen 4 Mobile	14.078 ms	1 - 406 MB	NPU
EfficientViT-l2-cls	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 For Galaxy Mobile	6.268 ms	2 - 247 MB	NPU
EfficientViT-l2-cls	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	8.04 ms	2 - 241 MB	NPU
EfficientViT-l2-cls	QNN_DLC	float	Snapdragon® X2 Elite	7.616 ms	2 - 2 MB	NPU
EfficientViT-l2-cls	QNN_DLC	float	Snapdragon® X Elite	16.27 ms	2 - 2 MB	NPU
EfficientViT-l2-cls	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	10.701 ms	0 - 369 MB	NPU
EfficientViT-l2-cls	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	36.359 ms	0 - 346 MB	NPU
EfficientViT-l2-cls	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-8275	19.37 ms	2 - 6 MB	NPU
EfficientViT-l2-cls	QNN_DLC	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	15.542 ms	2 - 4 MB	NPU
EfficientViT-l2-cls	QNN_DLC	float	Qualcomm® QCS8450	36.359 ms	0 - 346 MB	NPU
EfficientViT-l2-cls	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-9075	19.349 ms	2 - 5 MB	NPU
EfficientViT-l2-cls	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-X7181	16.27 ms	2 - 2 MB	NPU
EfficientViT-l2-cls	QNN_DLC	float	Qualcomm® Dragonwing™ Q-8750	8.04 ms	2 - 241 MB	NPU
EfficientViT-l2-cls	QNN_DLC	float	Qualcomm® SA7255P	56.562 ms	2 - 241 MB	NPU
EfficientViT-l2-cls	QNN_DLC	float	Qualcomm® SA8295P	32.88 ms	2 - 241 MB	NPU
EfficientViT-l2-cls	QNN_DLC	w8a16	Snapdragon® 8 Elite Gen 5 For Galaxy Mobile	5.192 ms	1 - 209 MB	NPU
EfficientViT-l2-cls	QNN_DLC	w8a16	Snapdragon® 8 Elite For Galaxy Mobile	6.601 ms	1 - 206 MB	NPU
EfficientViT-l2-cls	QNN_DLC	w8a16	Snapdragon® X2 Elite	6.117 ms	1 - 1 MB	NPU
EfficientViT-l2-cls	QNN_DLC	w8a16	Snapdragon® X Elite	15.076 ms	1 - 1 MB	NPU
EfficientViT-l2-cls	QNN_DLC	w8a16	Snapdragon® 8 Gen 3 Mobile	9.289 ms	0 - 326 MB	NPU
EfficientViT-l2-cls	QNN_DLC	w8a16	Qualcomm® Dragonwing™ IQ-8275	12.616 ms	1 - 5 MB	NPU
EfficientViT-l2-cls	QNN_DLC	w8a16	Qualcomm® Dragonwing™ QCS8550 (Proxy)	14.206 ms	1 - 4 MB	NPU
EfficientViT-l2-cls	QNN_DLC	w8a16	Qualcomm® SA8650P	14.996 ms	1 - 255 MB	NPU
EfficientViT-l2-cls	QNN_DLC	w8a16	Qualcomm® SA8255P	14.996 ms	1 - 255 MB	NPU
EfficientViT-l2-cls	QNN_DLC	w8a16	Qualcomm® Dragonwing™ IQ-9075	14.891 ms	3 - 6 MB	NPU
EfficientViT-l2-cls	QNN_DLC	w8a16	Qualcomm® Dragonwing™ IQ-X7181	15.076 ms	1 - 1 MB	NPU
EfficientViT-l2-cls	QNN_DLC	w8a16	Qualcomm® Dragonwing™ Q-6690	105.55 ms	1 - 414 MB	NPU
EfficientViT-l2-cls	QNN_DLC	w8a16	Qualcomm® Dragonwing™ Q-7790	16.853 ms	1 - 392 MB	NPU
EfficientViT-l2-cls	QNN_DLC	w8a16	Qualcomm® Dragonwing™ Q-8750	6.601 ms	1 - 206 MB	NPU
EfficientViT-l2-cls	QNN_DLC	w8a16	Snapdragon® 7 Gen 4 Mobile	16.853 ms	1 - 392 MB	NPU
EfficientViT-l2-cls	TFLITE	float	Snapdragon® 8 Elite Gen 5 For Galaxy Mobile	6.178 ms	0 - 325 MB	NPU
EfficientViT-l2-cls	TFLITE	float	Snapdragon® 8 Elite For Galaxy Mobile	7.795 ms	0 - 313 MB	NPU
EfficientViT-l2-cls	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	10.014 ms	0 - 455 MB	NPU
EfficientViT-l2-cls	TFLITE	float	Snapdragon® 8 Gen 1 Mobile	36.163 ms	0 - 449 MB	NPU
EfficientViT-l2-cls	TFLITE	float	Qualcomm® Dragonwing™ IQ-8275	18.93 ms	0 - 136 MB	NPU
EfficientViT-l2-cls	TFLITE	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	14.376 ms	0 - 3 MB	NPU
EfficientViT-l2-cls	TFLITE	float	Qualcomm® SA8775P	79.943 ms	0 - 32 MB	GPU
EfficientViT-l2-cls	TFLITE	float	Qualcomm® SA8650P	79.943 ms	0 - 32 MB	GPU
EfficientViT-l2-cls	TFLITE	float	Qualcomm® SA8255P	79.943 ms	0 - 32 MB	GPU
EfficientViT-l2-cls	TFLITE	float	Qualcomm® QCS8450	36.163 ms	0 - 449 MB	NPU
EfficientViT-l2-cls	TFLITE	float	Qualcomm® Dragonwing™ IQ-9075	18.467 ms	0 - 136 MB	NPU
EfficientViT-l2-cls	TFLITE	float	Qualcomm® Dragonwing™ Q-8750	7.795 ms	0 - 313 MB	NPU
EfficientViT-l2-cls	TFLITE	float	Qualcomm® SA7255P	204.442 ms	0 - 32 MB	GPU
EfficientViT-l2-cls	TFLITE	float	Qualcomm® SA8295P	32.298 ms	0 - 307 MB	NPU

License

The license for the original implementation of EfficientViT-l2-cls can be found here.

References

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.

Downloads last month: -; Downloads are not tracked for this model. How to track

Paper for qualcomm/EfficientViT-l2-cls

EfficientViT: Lightweight Multi-Scale Attention for On-Device Semantic Segmentation

Paper • 2205.14756 • Published May 29, 2022 • 1