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 1.24.1	Download
ONNX	w8a16	Universal	QAIRT 2.42, ONNX Runtime 1.24.1	Download
ONNX	w8a16_mixed_int16	Universal	QAIRT 2.42, ONNX Runtime 1.24.1	Download
QNN_DLC	float	Universal	QAIRT 2.43	Download
QNN_DLC	w8a16	Universal	QAIRT 2.43	Download
QNN_DLC	w8a16_mixed_int16	Universal	QAIRT 2.43	Download
TFLITE	float	Universal	QAIRT 2.43, TFLite 2.17.0	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® X Elite	1.462 ms	16 - 16 MB	NPU
LeViT	ONNX	float	Snapdragon® 8 Gen 3 Mobile	0.877 ms	0 - 101 MB	NPU
LeViT	ONNX	float	Qualcomm® QCS8550 (Proxy)	1.252 ms	0 - 22 MB	NPU
LeViT	ONNX	float	Qualcomm® QCS9075	1.65 ms	1 - 3 MB	NPU
LeViT	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	0.704 ms	0 - 67 MB	NPU
LeViT	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	0.653 ms	1 - 76 MB	NPU
LeViT	ONNX	float	Snapdragon® X2 Elite	0.69 ms	16 - 16 MB	NPU
LeViT	QNN_DLC	float	Snapdragon® X Elite	1.824 ms	1 - 1 MB	NPU
LeViT	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	1.084 ms	0 - 83 MB	NPU
LeViT	QNN_DLC	float	Qualcomm® QCS8275 (Proxy)	3.833 ms	1 - 58 MB	NPU
LeViT	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	1.595 ms	1 - 2 MB	NPU
LeViT	QNN_DLC	float	Qualcomm® QCS9075	1.882 ms	3 - 5 MB	NPU
LeViT	QNN_DLC	float	Qualcomm® QCS8450 (Proxy)	2.374 ms	0 - 80 MB	NPU
LeViT	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	0.847 ms	0 - 62 MB	NPU
LeViT	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	0.75 ms	1 - 62 MB	NPU
LeViT	QNN_DLC	float	Snapdragon® X2 Elite	1.009 ms	1 - 1 MB	NPU
LeViT	QNN_DLC	w8a16	Snapdragon® X Elite	1.673 ms	0 - 0 MB	NPU
LeViT	QNN_DLC	w8a16	Snapdragon® 8 Gen 3 Mobile	1.016 ms	0 - 61 MB	NPU
LeViT	QNN_DLC	w8a16	Qualcomm® QCS8275 (Proxy)	2.98 ms	0 - 40 MB	NPU
LeViT	QNN_DLC	w8a16	Qualcomm® QCS8550 (Proxy)	1.452 ms	0 - 26 MB	NPU
LeViT	QNN_DLC	w8a16	Qualcomm® QCS9075	1.727 ms	0 - 2 MB	NPU
LeViT	QNN_DLC	w8a16	Qualcomm® QCM6690	5.836 ms	0 - 164 MB	NPU
LeViT	QNN_DLC	w8a16	Snapdragon® 8 Elite For Galaxy Mobile	0.747 ms	0 - 41 MB	NPU
LeViT	QNN_DLC	w8a16	Snapdragon® 7 Gen 4 Mobile	1.47 ms	0 - 39 MB	NPU
LeViT	QNN_DLC	w8a16	Snapdragon® 8 Elite Gen 5 Mobile	0.639 ms	0 - 42 MB	NPU
LeViT	QNN_DLC	w8a16	Snapdragon® X2 Elite	0.865 ms	0 - 0 MB	NPU
LeViT	QNN_DLC	w8a16_mixed_int16	Snapdragon® X Elite	1.7 ms	0 - 0 MB	NPU
LeViT	QNN_DLC	w8a16_mixed_int16	Snapdragon® 8 Gen 3 Mobile	1.046 ms	0 - 63 MB	NPU
LeViT	QNN_DLC	w8a16_mixed_int16	Qualcomm® QCS8275 (Proxy)	3.027 ms	0 - 40 MB	NPU
LeViT	QNN_DLC	w8a16_mixed_int16	Qualcomm® QCS8550 (Proxy)	1.485 ms	0 - 2 MB	NPU
LeViT	QNN_DLC	w8a16_mixed_int16	Qualcomm® QCS9075	1.738 ms	0 - 2 MB	NPU
LeViT	QNN_DLC	w8a16_mixed_int16	Qualcomm® QCM6690	6.101 ms	0 - 166 MB	NPU
LeViT	QNN_DLC	w8a16_mixed_int16	Snapdragon® 8 Elite For Galaxy Mobile	0.753 ms	0 - 42 MB	NPU
LeViT	QNN_DLC	w8a16_mixed_int16	Snapdragon® 7 Gen 4 Mobile	1.507 ms	0 - 39 MB	NPU
LeViT	QNN_DLC	w8a16_mixed_int16	Snapdragon® 8 Elite Gen 5 Mobile	0.649 ms	0 - 41 MB	NPU
LeViT	QNN_DLC	w8a16_mixed_int16	Snapdragon® X2 Elite	0.903 ms	0 - 0 MB	NPU
LeViT	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	1.058 ms	0 - 96 MB	NPU
LeViT	TFLITE	float	Qualcomm® QCS8275 (Proxy)	4.059 ms	0 - 66 MB	NPU
LeViT	TFLITE	float	Qualcomm® QCS8550 (Proxy)	1.556 ms	0 - 2 MB	NPU
LeViT	TFLITE	float	Qualcomm® QCS9075	1.869 ms	0 - 19 MB	NPU
LeViT	TFLITE	float	Qualcomm® QCS8450 (Proxy)	2.356 ms	0 - 83 MB	NPU
LeViT	TFLITE	float	Snapdragon® 8 Elite For Galaxy Mobile	0.807 ms	0 - 66 MB	NPU
LeViT	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	0.679 ms	0 - 72 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.