Sequencer2D: Optimized for Qualcomm Devices

Sequencer2D is a vision transformer model that can classify images from the Imagenet dataset.

This is based on the implementation of Sequencer2D 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 1.27.1	Download
QNN_DLC	float	Universal	QAIRT 2.45	Download
TFLITE	float	Universal	QAIRT 2.45	Download
TFLITE	w8a8	Universal	QAIRT 2.45	Download

For more device-specific assets and performance metrics, visit Sequencer2D 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 Sequencer2D on GitHub for usage instructions.

Model Details

Model Type: Model_use_case.image_classification

Model Stats:

• Model checkpoint: sequencer2d_s
• Input resolution: 224x224
• Number of parameters: 27.6M
• Model size (float): 106 MB
• Model size (w8a8): 69.1 MB

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
Sequencer2D	ONNX	float	Snapdragon® X2 Elite	10.777 ms	2 - 2 MB	NPU
Sequencer2D	ONNX	float	Snapdragon® X Elite	16.891 ms	60 - 60 MB	NPU
Sequencer2D	ONNX	float	Snapdragon® 8 Gen 3 Mobile	10.912 ms	1 - 960 MB	NPU
Sequencer2D	ONNX	float	Snapdragon® 8 Gen 1 Mobile	22.622 ms	1 - 661 MB	NPU
Sequencer2D	ONNX	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	16.048 ms	1 - 368 MB	NPU
Sequencer2D	ONNX	float	Qualcomm® QCS8450	22.622 ms	1 - 661 MB	NPU
Sequencer2D	ONNX	float	Qualcomm® Dragonwing™ IQ-9075	19.868 ms	1 - 4 MB	NPU
Sequencer2D	ONNX	float	Qualcomm® Dragonwing™ IQ-X7181	16.891 ms	60 - 60 MB	NPU
Sequencer2D	ONNX	float	Qualcomm® Dragonwing™ Q-8750	9.612 ms	1 - 742 MB	NPU
Sequencer2D	ONNX	float	Snapdragon® 8 Elite Mobile	9.612 ms	1 - 742 MB	NPU
Sequencer2D	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	8.05 ms	0 - 894 MB	NPU
Sequencer2D	QNN_DLC	float	Snapdragon® X2 Elite	11.786 ms	1 - 1 MB	NPU
Sequencer2D	QNN_DLC	float	Snapdragon® X Elite	21.344 ms	1 - 1 MB	NPU
Sequencer2D	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	14.16 ms	0 - 2205 MB	NPU
Sequencer2D	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	25.65 ms	0 - 825 MB	NPU
Sequencer2D	QNN_DLC	float	Qualcomm® Dragonwing™ QCS8275	51.674 ms	1 - 996 MB	NPU
Sequencer2D	QNN_DLC	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	20.18 ms	1 - 3 MB	NPU
Sequencer2D	QNN_DLC	float	Qualcomm® SA8775P	24.077 ms	1 - 908 MB	NPU
Sequencer2D	QNN_DLC	float	Qualcomm® SA8650P	24.077 ms	1 - 908 MB	NPU
Sequencer2D	QNN_DLC	float	Qualcomm® SA8255P	24.077 ms	1 - 908 MB	NPU
Sequencer2D	QNN_DLC	float	Qualcomm® QCS8450	25.65 ms	0 - 825 MB	NPU
Sequencer2D	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-9075	24.107 ms	1 - 3 MB	NPU
Sequencer2D	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-X7181	21.344 ms	1 - 1 MB	NPU
Sequencer2D	QNN_DLC	float	Qualcomm® Dragonwing™ Q-8750	12.101 ms	1 - 1045 MB	NPU
Sequencer2D	QNN_DLC	float	Qualcomm® SA7255P	51.674 ms	1 - 996 MB	NPU
Sequencer2D	QNN_DLC	float	Qualcomm® SA8295P	25.964 ms	1 - 665 MB	NPU
Sequencer2D	QNN_DLC	float	Snapdragon® 8 Elite Mobile	12.101 ms	1 - 1045 MB	NPU
Sequencer2D	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	9.197 ms	0 - 865 MB	NPU
Sequencer2D	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	11.921 ms	0 - 982 MB	NPU
Sequencer2D	TFLITE	float	Snapdragon® 8 Gen 1 Mobile	21.182 ms	0 - 732 MB	NPU
Sequencer2D	TFLITE	float	Qualcomm® Dragonwing™ QCS8275	37.479 ms	0 - 795 MB	NPU
Sequencer2D	TFLITE	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	17.463 ms	0 - 7 MB	NPU
Sequencer2D	TFLITE	float	Qualcomm® SA8775P	19.055 ms	0 - 795 MB	NPU
Sequencer2D	TFLITE	float	Qualcomm® SA8650P	19.055 ms	0 - 795 MB	NPU
Sequencer2D	TFLITE	float	Qualcomm® SA8255P	19.055 ms	0 - 795 MB	NPU
Sequencer2D	TFLITE	float	Qualcomm® QCS8450	21.182 ms	0 - 732 MB	NPU
Sequencer2D	TFLITE	float	Qualcomm® Dragonwing™ IQ-9075	18.493 ms	0 - 76 MB	NPU
Sequencer2D	TFLITE	float	Qualcomm® Dragonwing™ Q-8750	9.276 ms	0 - 785 MB	NPU
Sequencer2D	TFLITE	float	Qualcomm® SA7255P	37.479 ms	0 - 795 MB	NPU
Sequencer2D	TFLITE	float	Qualcomm® SA8295P	20.571 ms	0 - 484 MB	NPU
Sequencer2D	TFLITE	float	Snapdragon® 8 Elite Mobile	9.276 ms	0 - 785 MB	NPU
Sequencer2D	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	7.238 ms	0 - 761 MB	NPU
Sequencer2D	TFLITE	w8a8	Snapdragon® 8 Gen 3 Mobile	13.56 ms	14 - 965 MB	NPU
Sequencer2D	TFLITE	w8a8	Snapdragon® 8 Gen 1 Mobile	21.822 ms	0 - 816 MB	NPU
Sequencer2D	TFLITE	w8a8	Qualcomm® Dragonwing™ QCS8275	36.371 ms	0 - 862 MB	NPU
Sequencer2D	TFLITE	w8a8	Qualcomm® Dragonwing™ QCS8550 (Proxy)	19.0 ms	0 - 11 MB	NPU
Sequencer2D	TFLITE	w8a8	Qualcomm® SA8775P	20.421 ms	0 - 861 MB	NPU
Sequencer2D	TFLITE	w8a8	Qualcomm® SA8650P	20.421 ms	0 - 861 MB	NPU
Sequencer2D	TFLITE	w8a8	Qualcomm® SA8255P	20.421 ms	0 - 861 MB	NPU
Sequencer2D	TFLITE	w8a8	Qualcomm® QCS8450	21.822 ms	0 - 816 MB	NPU
Sequencer2D	TFLITE	w8a8	Qualcomm® Dragonwing™ IQ-9075	19.58 ms	0 - 64 MB	NPU
Sequencer2D	TFLITE	w8a8	Qualcomm® Dragonwing™ Q-8750	10.962 ms	0 - 792 MB	NPU
Sequencer2D	TFLITE	w8a8	Qualcomm® SA7255P	36.371 ms	0 - 862 MB	NPU
Sequencer2D	TFLITE	w8a8	Qualcomm® SA8295P	22.938 ms	0 - 616 MB	NPU
Sequencer2D	TFLITE	w8a8	Snapdragon® 8 Elite Mobile	10.962 ms	0 - 792 MB	NPU
Sequencer2D	TFLITE	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	9.073 ms	0 - 823 MB	NPU

License

• The license for the original implementation of Sequencer2D can be found here.

References

• Sequencer: Deep LSTM for Image Classification
• 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.