ResNeXt101: Optimized for Qualcomm Devices

ResNeXt101 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 ResNeXt101 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
ONNX	w8a8	Universal	QAIRT 2.45, ONNX Runtime 1.25.0	Download
QNN_DLC	float	Universal	QAIRT 2.45	Download
QNN_DLC	w8a8	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 ResNeXt101 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 ResNeXt101 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: 88.7M
• Model size (float): 338 MB
• Model size (w8a8): 87.3 MB

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
ResNeXt101	ONNX	float	Snapdragon® X2 Elite	3.06 ms	180 - 180 MB	NPU
ResNeXt101	ONNX	float	Snapdragon® X Elite	6.459 ms	172 - 172 MB	NPU
ResNeXt101	ONNX	float	Snapdragon® 8 Gen 3 Mobile	4.494 ms	1 - 358 MB	NPU
ResNeXt101	ONNX	float	Snapdragon® 8 Gen 1 Mobile	9.561 ms	1 - 302 MB	NPU
ResNeXt101	ONNX	float	Qualcomm® QCS8550 (Proxy)	6.548 ms	0 - 196 MB	NPU
ResNeXt101	ONNX	float	Qualcomm® QCS8450	9.561 ms	1 - 302 MB	NPU
ResNeXt101	ONNX	float	Snapdragon® 8 Elite Mobile	3.908 ms	0 - 182 MB	NPU
ResNeXt101	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	3.053 ms	1 - 182 MB	NPU
ResNeXt101	ONNX	float	Qualcomm® QCS9075	9.773 ms	1 - 46 MB	NPU
ResNeXt101	ONNX	float	Qualcomm® QCS8750	3.908 ms	0 - 182 MB	NPU
ResNeXt101	ONNX	float	Qualcomm® QCS7181	6.459 ms	172 - 172 MB	NPU
ResNeXt101	ONNX	w8a8	Snapdragon® X2 Elite	1.389 ms	181 - 181 MB	NPU
ResNeXt101	ONNX	w8a8	Snapdragon® X Elite	2.92 ms	149 - 149 MB	NPU
ResNeXt101	ONNX	w8a8	Snapdragon® 8 Gen 3 Mobile	2.153 ms	0 - 255 MB	NPU
ResNeXt101	ONNX	w8a8	Snapdragon® 8 Gen 1 Mobile	3.991 ms	0 - 256 MB	NPU
ResNeXt101	ONNX	w8a8	Qualcomm® QCS6490	8.975 ms	0 - 45 MB	NPU
ResNeXt101	ONNX	w8a8	Qualcomm® QCS8550 (Proxy)	2.886 ms	0 - 100 MB	NPU
ResNeXt101	ONNX	w8a8	Qualcomm® QCS8450	3.991 ms	0 - 256 MB	NPU
ResNeXt101	ONNX	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	1.592 ms	0 - 218 MB	NPU
ResNeXt101	ONNX	w8a8	Snapdragon® 7 Gen 4 Mobile	4.07 ms	0 - 245 MB	NPU
ResNeXt101	ONNX	w8a8	Qualcomm® QCM6690	33.534 ms	0 - 376 MB	NPU
ResNeXt101	ONNX	w8a8	Qualcomm® QCS9075	3.121 ms	0 - 45 MB	NPU
ResNeXt101	ONNX	w8a8	Snapdragon® 8 Elite Mobile	1.898 ms	0 - 216 MB	NPU
ResNeXt101	ONNX	w8a8	Qualcomm® QCS7790	4.07 ms	0 - 245 MB	NPU
ResNeXt101	ONNX	w8a8	Qualcomm® QCS8750	1.898 ms	0 - 216 MB	NPU
ResNeXt101	ONNX	w8a8	Qualcomm® QCS7181	2.92 ms	149 - 149 MB	NPU
ResNeXt101	QNN_DLC	float	Snapdragon® X2 Elite	3.654 ms	1 - 1 MB	NPU
ResNeXt101	QNN_DLC	float	Snapdragon® X Elite	6.939 ms	1 - 1 MB	NPU
ResNeXt101	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	4.682 ms	0 - 363 MB	NPU
ResNeXt101	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	9.91 ms	0 - 311 MB	NPU
ResNeXt101	QNN_DLC	float	Qualcomm® QCS8275	35.988 ms	1 - 180 MB	NPU
ResNeXt101	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	6.717 ms	1 - 3 MB	NPU
ResNeXt101	QNN_DLC	float	Qualcomm® QCS8450	9.91 ms	0 - 311 MB	NPU
ResNeXt101	QNN_DLC	float	Snapdragon® 8 Elite Mobile	3.97 ms	1 - 182 MB	NPU
ResNeXt101	QNN_DLC	float	Qualcomm® SA8295P	10.928 ms	1 - 134 MB	NPU
ResNeXt101	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	3.145 ms	1 - 193 MB	NPU
ResNeXt101	QNN_DLC	float	Qualcomm® SA7255P	35.988 ms	1 - 180 MB	NPU
ResNeXt101	QNN_DLC	float	Qualcomm® QCS9075	10.008 ms	1 - 3 MB	NPU
ResNeXt101	QNN_DLC	float	Qualcomm® QCS8750	3.97 ms	1 - 182 MB	NPU
ResNeXt101	QNN_DLC	float	Qualcomm® QCS7181	6.939 ms	1 - 1 MB	NPU
ResNeXt101	QNN_DLC	w8a8	Snapdragon® X2 Elite	1.624 ms	0 - 0 MB	NPU
ResNeXt101	QNN_DLC	w8a8	Snapdragon® X Elite	3.095 ms	0 - 0 MB	NPU
ResNeXt101	QNN_DLC	w8a8	Snapdragon® 8 Gen 3 Mobile	2.103 ms	0 - 258 MB	NPU
ResNeXt101	QNN_DLC	w8a8	Snapdragon® 8 Gen 1 Mobile	3.959 ms	0 - 260 MB	NPU
ResNeXt101	QNN_DLC	w8a8	Qualcomm® QCS6490	9.27 ms	0 - 2 MB	NPU
ResNeXt101	QNN_DLC	w8a8	Qualcomm® QCS8275	6.655 ms	0 - 214 MB	NPU
ResNeXt101	QNN_DLC	w8a8	Qualcomm® QCS8550 (Proxy)	2.929 ms	0 - 2 MB	NPU
ResNeXt101	QNN_DLC	w8a8	Qualcomm® QCS8450	3.959 ms	0 - 260 MB	NPU
ResNeXt101	QNN_DLC	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	1.511 ms	0 - 219 MB	NPU
ResNeXt101	QNN_DLC	w8a8	Snapdragon® 7 Gen 4 Mobile	4.048 ms	0 - 244 MB	NPU
ResNeXt101	QNN_DLC	w8a8	Qualcomm® QCM6690	34.474 ms	0 - 372 MB	NPU
ResNeXt101	QNN_DLC	w8a8	Qualcomm® QCS9075	3.119 ms	0 - 2 MB	NPU
ResNeXt101	QNN_DLC	w8a8	Qualcomm® SA7255P	6.655 ms	0 - 214 MB	NPU
ResNeXt101	QNN_DLC	w8a8	Snapdragon® 8 Elite Mobile	1.804 ms	0 - 211 MB	NPU
ResNeXt101	QNN_DLC	w8a8	Qualcomm® SA8295P	4.278 ms	0 - 217 MB	NPU
ResNeXt101	QNN_DLC	w8a8	Qualcomm® QCS7790	4.048 ms	0 - 244 MB	NPU
ResNeXt101	QNN_DLC	w8a8	Qualcomm® QCS8750	1.804 ms	0 - 211 MB	NPU
ResNeXt101	QNN_DLC	w8a8	Qualcomm® QCS7181	3.095 ms	0 - 0 MB	NPU
ResNeXt101	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	4.613 ms	0 - 532 MB	NPU
ResNeXt101	TFLITE	float	Snapdragon® 8 Gen 1 Mobile	9.863 ms	0 - 476 MB	NPU
ResNeXt101	TFLITE	float	Qualcomm® QCS8275	35.979 ms	0 - 358 MB	NPU
ResNeXt101	TFLITE	float	Qualcomm® QCS8550 (Proxy)	6.761 ms	0 - 3 MB	NPU
ResNeXt101	TFLITE	float	Qualcomm® SA8775P	66.442 ms	0 - 21 MB	GPU
ResNeXt101	TFLITE	float	Qualcomm® SA8650P	66.442 ms	0 - 21 MB	GPU
ResNeXt101	TFLITE	float	Qualcomm® SA8255P	66.442 ms	0 - 21 MB	GPU
ResNeXt101	TFLITE	float	Qualcomm® QCS8450	9.863 ms	0 - 476 MB	NPU
ResNeXt101	TFLITE	float	Snapdragon® 8 Elite Mobile	3.944 ms	0 - 359 MB	NPU
ResNeXt101	TFLITE	float	Qualcomm® SA8295P	11.011 ms	0 - 305 MB	NPU
ResNeXt101	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	3.119 ms	0 - 368 MB	NPU
ResNeXt101	TFLITE	float	Qualcomm® SA7255P	35.979 ms	0 - 358 MB	NPU
ResNeXt101	TFLITE	float	Qualcomm® QCS9075	10.084 ms	0 - 174 MB	NPU
ResNeXt101	TFLITE	float	Qualcomm® QCS8750	3.944 ms	0 - 359 MB	NPU
ResNeXt101	TFLITE	w8a8	Snapdragon® 8 Gen 3 Mobile	1.975 ms	0 - 250 MB	NPU
ResNeXt101	TFLITE	w8a8	Snapdragon® 8 Gen 1 Mobile	3.793 ms	0 - 254 MB	NPU
ResNeXt101	TFLITE	w8a8	Qualcomm® QCS6490	8.902 ms	0 - 89 MB	NPU
ResNeXt101	TFLITE	w8a8	Qualcomm® QCS8275	6.356 ms	0 - 213 MB	NPU
ResNeXt101	TFLITE	w8a8	Qualcomm® QCS8550 (Proxy)	2.636 ms	0 - 2 MB	NPU
ResNeXt101	TFLITE	w8a8	Qualcomm® SA8775P	67.149 ms	1 - 25 MB	GPU
ResNeXt101	TFLITE	w8a8	Qualcomm® SA8650P	67.149 ms	1 - 25 MB	GPU
ResNeXt101	TFLITE	w8a8	Qualcomm® SA8255P	67.149 ms	1 - 25 MB	GPU
ResNeXt101	TFLITE	w8a8	Qualcomm® QCS8450	3.793 ms	0 - 254 MB	NPU
ResNeXt101	TFLITE	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	1.465 ms	0 - 214 MB	NPU
ResNeXt101	TFLITE	w8a8	Snapdragon® 7 Gen 4 Mobile	3.986 ms	0 - 243 MB	NPU
ResNeXt101	TFLITE	w8a8	Qualcomm® QCM6690	29.594 ms	0 - 374 MB	NPU
ResNeXt101	TFLITE	w8a8	Qualcomm® QCS9075	2.924 ms	0 - 89 MB	NPU
ResNeXt101	TFLITE	w8a8	Qualcomm® SA7255P	6.356 ms	0 - 213 MB	NPU
ResNeXt101	TFLITE	w8a8	Snapdragon® 8 Elite Mobile	1.734 ms	0 - 212 MB	NPU
ResNeXt101	TFLITE	w8a8	Qualcomm® SA8295P	4.107 ms	0 - 215 MB	NPU
ResNeXt101	TFLITE	w8a8	Qualcomm® QCS7790	3.986 ms	0 - 243 MB	NPU
ResNeXt101	TFLITE	w8a8	Qualcomm® QCS8750	1.734 ms	0 - 212 MB	NPU

License

• The license for the original implementation of ResNeXt101 can be found here.

References

• Aggregated Residual Transformations for Deep Neural Networks
• 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.