LeViT: Optimized for Qualcomm Devices

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

This is based on the implementation of LeViT 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 2.42, ONNX Runtime 1.24.3	Download
ONNX	w8a16	Universal	QAIRT 2.42, ONNX Runtime 1.24.3	Download
ONNX	w8a16_mixed_int16	Universal	QAIRT 2.42, ONNX Runtime 1.24.3	Download
QNN_DLC	float	Universal	QAIRT 2.45	Download
QNN_DLC	w8a16	Universal	QAIRT 2.45	Download
QNN_DLC	w8a16_mixed_int16	Universal	QAIRT 2.45	Download
TFLITE	float	Universal	QAIRT 2.45	Download

For more device-specific assets and performance metrics, visit LeViT 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 LeViT on GitHub for usage instructions.

Model Details

Model Type: Model_use_case.image_classification

Model Stats:

• Model checkpoint: LeViT-128S
• Input resolution: 224x224
• Number of parameters: 7.82M
• Model size (float): 29.9 MB
• Model size (w8a16): 8.83 MB

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
LeViT	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	0.653 ms	1 - 76 MB	NPU
LeViT	ONNX	float	Snapdragon® X2 Elite	0.71 ms	16 - 16 MB	NPU
LeViT	ONNX	float	Snapdragon® X Elite	1.479 ms	16 - 16 MB	NPU
LeViT	ONNX	float	Snapdragon® 8 Gen 3 Mobile	0.864 ms	0 - 93 MB	NPU
LeViT	ONNX	float	Qualcomm® QCS8550 (Proxy)	1.254 ms	0 - 22 MB	NPU
LeViT	ONNX	float	Qualcomm® QCS9075	1.617 ms	1 - 3 MB	NPU
LeViT	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	0.706 ms	0 - 67 MB	NPU
LeViT	ONNX	w8a16	Snapdragon® 8 Elite Gen 5 Mobile	0.467 ms	0 - 62 MB	NPU
LeViT	ONNX	w8a16	Snapdragon® X2 Elite	0.507 ms	8 - 8 MB	NPU
LeViT	ONNX	w8a16	Snapdragon® X Elite	1.253 ms	8 - 8 MB	NPU
LeViT	ONNX	w8a16	Snapdragon® 8 Gen 3 Mobile	0.719 ms	0 - 80 MB	NPU
LeViT	ONNX	w8a16	Qualcomm® QCS6490	43.779 ms	11 - 16 MB	CPU
LeViT	ONNX	w8a16	Qualcomm® QCS8550 (Proxy)	1.054 ms	0 - 13 MB	NPU
LeViT	ONNX	w8a16	Qualcomm® QCS9075	1.275 ms	0 - 3 MB	NPU
LeViT	ONNX	w8a16	Qualcomm® QCM6690	18.551 ms	11 - 21 MB	CPU
LeViT	ONNX	w8a16	Snapdragon® 8 Elite For Galaxy Mobile	0.529 ms	0 - 60 MB	NPU
LeViT	ONNX	w8a16	Snapdragon® 7 Gen 4 Mobile	16.116 ms	19 - 29 MB	CPU
LeViT	ONNX	w8a16_mixed_int16	Snapdragon® 8 Elite Gen 5 Mobile	0.485 ms	0 - 61 MB	NPU
LeViT	ONNX	w8a16_mixed_int16	Snapdragon® X2 Elite	0.506 ms	10 - 10 MB	NPU
LeViT	ONNX	w8a16_mixed_int16	Snapdragon® X Elite	1.28 ms	10 - 10 MB	NPU
LeViT	ONNX	w8a16_mixed_int16	Snapdragon® 8 Gen 3 Mobile	0.738 ms	0 - 87 MB	NPU
LeViT	ONNX	w8a16_mixed_int16	Qualcomm® QCS6490	43.894 ms	15 - 20 MB	CPU
LeViT	ONNX	w8a16_mixed_int16	Qualcomm® QCS8550 (Proxy)	1.062 ms	0 - 110 MB	NPU
LeViT	ONNX	w8a16_mixed_int16	Qualcomm® QCS9075	1.312 ms	0 - 3 MB	NPU
LeViT	ONNX	w8a16_mixed_int16	Qualcomm® QCM6690	18.563 ms	12 - 23 MB	CPU
LeViT	ONNX	w8a16_mixed_int16	Snapdragon® 8 Elite For Galaxy Mobile	0.549 ms	0 - 59 MB	NPU
LeViT	ONNX	w8a16_mixed_int16	Snapdragon® 7 Gen 4 Mobile	15.119 ms	19 - 28 MB	CPU
LeViT	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	0.73 ms	1 - 59 MB	NPU
LeViT	QNN_DLC	float	Snapdragon® X2 Elite	0.978 ms	1 - 1 MB	NPU
LeViT	QNN_DLC	float	Snapdragon® X Elite	1.791 ms	1 - 1 MB	NPU
LeViT	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	1.064 ms	0 - 79 MB	NPU
LeViT	QNN_DLC	float	Qualcomm® QCS8275 (Proxy)	3.814 ms	1 - 55 MB	NPU
LeViT	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	1.565 ms	1 - 2 MB	NPU
LeViT	QNN_DLC	float	Qualcomm® QCS9075	1.863 ms	1 - 3 MB	NPU
LeViT	QNN_DLC	float	Qualcomm® QCS8450 (Proxy)	2.351 ms	0 - 77 MB	NPU
LeViT	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	0.836 ms	0 - 61 MB	NPU
LeViT	QNN_DLC	w8a16	Snapdragon® 8 Elite Gen 5 Mobile	0.634 ms	0 - 44 MB	NPU
LeViT	QNN_DLC	w8a16	Snapdragon® X2 Elite	0.851 ms	0 - 0 MB	NPU
LeViT	QNN_DLC	w8a16	Snapdragon® X Elite	1.633 ms	0 - 0 MB	NPU
LeViT	QNN_DLC	w8a16	Snapdragon® 8 Gen 3 Mobile	0.991 ms	0 - 60 MB	NPU
LeViT	QNN_DLC	w8a16	Qualcomm® QCS8275 (Proxy)	2.955 ms	0 - 39 MB	NPU
LeViT	QNN_DLC	w8a16	Qualcomm® QCS8550 (Proxy)	1.419 ms	0 - 2 MB	NPU
LeViT	QNN_DLC	w8a16	Qualcomm® QCS9075	1.711 ms	0 - 2 MB	NPU
LeViT	QNN_DLC	w8a16	Qualcomm® QCM6690	5.769 ms	0 - 165 MB	NPU
LeViT	QNN_DLC	w8a16	Snapdragon® 8 Elite For Galaxy Mobile	0.721 ms	0 - 42 MB	NPU
LeViT	QNN_DLC	w8a16	Snapdragon® 7 Gen 4 Mobile	1.469 ms	0 - 39 MB	NPU
LeViT	QNN_DLC	w8a16_mixed_int16	Snapdragon® 8 Elite Gen 5 Mobile	0.654 ms	0 - 44 MB	NPU
LeViT	QNN_DLC	w8a16_mixed_int16	Snapdragon® X2 Elite	0.848 ms	0 - 0 MB	NPU
LeViT	QNN_DLC	w8a16_mixed_int16	Snapdragon® X Elite	1.638 ms	0 - 0 MB	NPU
LeViT	QNN_DLC	w8a16_mixed_int16	Snapdragon® 8 Gen 3 Mobile	1.01 ms	0 - 60 MB	NPU
LeViT	QNN_DLC	w8a16_mixed_int16	Qualcomm® QCS8275 (Proxy)	2.979 ms	0 - 38 MB	NPU
LeViT	QNN_DLC	w8a16_mixed_int16	Qualcomm® QCS8550 (Proxy)	1.439 ms	0 - 38 MB	NPU
LeViT	QNN_DLC	w8a16_mixed_int16	Qualcomm® QCS9075	1.723 ms	0 - 2 MB	NPU
LeViT	QNN_DLC	w8a16_mixed_int16	Qualcomm® QCM6690	6.127 ms	0 - 164 MB	NPU
LeViT	QNN_DLC	w8a16_mixed_int16	Snapdragon® 8 Elite For Galaxy Mobile	0.736 ms	0 - 37 MB	NPU
LeViT	QNN_DLC	w8a16_mixed_int16	Snapdragon® 7 Gen 4 Mobile	1.48 ms	0 - 39 MB	NPU
LeViT	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	0.659 ms	0 - 68 MB	NPU
LeViT	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	1.038 ms	0 - 91 MB	NPU
LeViT	TFLITE	float	Qualcomm® QCS8275 (Proxy)	4.012 ms	0 - 62 MB	NPU
LeViT	TFLITE	float	Qualcomm® QCS8550 (Proxy)	1.516 ms	0 - 2 MB	NPU
LeViT	TFLITE	float	Qualcomm® QCS9075	1.875 ms	0 - 19 MB	NPU
LeViT	TFLITE	float	Qualcomm® QCS8450 (Proxy)	2.378 ms	0 - 79 MB	NPU
LeViT	TFLITE	float	Snapdragon® 8 Elite For Galaxy Mobile	0.785 ms	0 - 61 MB	NPU

License

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

References

• LeViT: a Vision Transformer in ConvNet's Clothing for Faster Inference
• 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.