Mobile-Bert-Uncased-Google: Optimized for Qualcomm Devices

MOBILEBERT is a lightweight BERT model designed for efficient self-supervised learning of language representations. It can be used for masked language modeling and as a backbone for various NLP tasks.

This is based on the implementation of Mobile-Bert-Uncased-Google 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
QNN_DLC	float	Universal	QAIRT 2.50	Download
QNN_DLC	w8a16	Universal	QAIRT 2.50	Download
QNN_DLC	w8a16_mixed_fp16	Universal	QAIRT 2.50	Download

For more device-specific assets and performance metrics, visit Mobile-Bert-Uncased-Google 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 Mobile-Bert-Uncased-Google on GitHub for usage instructions.

Model Details

Model Type: Model_use_case.text_generation

Model Stats:

• Input resolution: 1x384
• Model checkpoint: mobile_bert_uncased_google
• Model size (float): 130 MB
• Number of parameters: 25.3M

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
Mobile-Bert-Uncased-Google	ONNX	float	Snapdragon® 8 Elite Gen 5 For Galaxy Mobile	4.353 ms	0 - 231 MB	NPU
Mobile-Bert-Uncased-Google	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	5.843 ms	0 - 230 MB	NPU
Mobile-Bert-Uncased-Google	ONNX	float	Snapdragon® X2 Elite	4.645 ms	1 - 1 MB	NPU
Mobile-Bert-Uncased-Google	ONNX	float	Snapdragon® X Elite	11.146 ms	81 - 81 MB	NPU
Mobile-Bert-Uncased-Google	ONNX	float	Snapdragon® 8 Gen 3 Mobile	7.839 ms	0 - 396 MB	NPU
Mobile-Bert-Uncased-Google	ONNX	float	Snapdragon® 8 Gen 1 Mobile	13.252 ms	0 - 376 MB	NPU
Mobile-Bert-Uncased-Google	ONNX	float	Qualcomm® Dragonwing™ IQ-8275	12.636 ms	0 - 4 MB	NPU
Mobile-Bert-Uncased-Google	ONNX	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	10.598 ms	0 - 83 MB	NPU
Mobile-Bert-Uncased-Google	ONNX	float	Qualcomm® QCS8450	13.252 ms	0 - 376 MB	NPU
Mobile-Bert-Uncased-Google	ONNX	float	Qualcomm® Dragonwing™ IQ-9075	13.034 ms	0 - 4 MB	NPU
Mobile-Bert-Uncased-Google	ONNX	float	Qualcomm® Dragonwing™ IQ-X7181	11.146 ms	81 - 81 MB	NPU
Mobile-Bert-Uncased-Google	ONNX	float	Qualcomm® Dragonwing™ Q-8750	5.843 ms	0 - 230 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 For Galaxy Mobile	4.554 ms	0 - 245 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	5.723 ms	0 - 241 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	float	Snapdragon® X2 Elite	5.194 ms	1 - 1 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	float	Snapdragon® X Elite	11.55 ms	1 - 1 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	7.922 ms	0 - 325 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	13.365 ms	0 - 305 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-8275	12.819 ms	0 - 3 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	10.796 ms	0 - 235 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	float	Qualcomm® QCS8450	13.365 ms	0 - 305 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-9075	12.869 ms	0 - 2 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-X7181	11.55 ms	1 - 1 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	float	Qualcomm® Dragonwing™ Q-8750	5.723 ms	0 - 241 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	float	Qualcomm® SA8295P	13.948 ms	0 - 200 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	w8a16	Snapdragon® 8 Elite Gen 5 For Galaxy Mobile	4.282 ms	0 - 316 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	w8a16	Snapdragon® 8 Elite For Galaxy Mobile	5.317 ms	0 - 313 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	w8a16	Snapdragon® X2 Elite	4.867 ms	1 - 1 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	w8a16	Snapdragon® X Elite	10.216 ms	1 - 1 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	w8a16	Snapdragon® 8 Gen 3 Mobile	7.292 ms	0 - 367 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	w8a16	Qualcomm® Dragonwing™ IQ-8275	9.049 ms	0 - 3 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	w8a16	Qualcomm® Dragonwing™ QCS8550 (Proxy)	9.513 ms	0 - 184 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	w8a16	Qualcomm® SA8775P	9.944 ms	0 - 295 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	w8a16	Qualcomm® SA8650P	9.944 ms	0 - 295 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	w8a16	Qualcomm® SA8255P	9.944 ms	0 - 295 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	w8a16	Qualcomm® Dragonwing™ IQ-9075	9.738 ms	2 - 4 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	w8a16	Qualcomm® Dragonwing™ IQ-X7181	10.216 ms	1 - 1 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	w8a16	Qualcomm® Dragonwing™ Q-8750	5.317 ms	0 - 313 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	w8a16_mixed_fp16	Snapdragon® 8 Elite Gen 5 For Galaxy Mobile	4.431 ms	0 - 316 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	w8a16_mixed_fp16	Snapdragon® 8 Elite For Galaxy Mobile	5.548 ms	0 - 312 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	w8a16_mixed_fp16	Snapdragon® X2 Elite	5.025 ms	0 - 0 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	w8a16_mixed_fp16	Snapdragon® X Elite	10.483 ms	1 - 1 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	w8a16_mixed_fp16	Snapdragon® 8 Gen 3 Mobile	7.453 ms	0 - 367 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	w8a16_mixed_fp16	Qualcomm® Dragonwing™ IQ-8275	9.588 ms	0 - 3 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	w8a16_mixed_fp16	Qualcomm® Dragonwing™ QCS8550 (Proxy)	9.691 ms	0 - 189 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	w8a16_mixed_fp16	Qualcomm® SA8775P	10.192 ms	0 - 295 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	w8a16_mixed_fp16	Qualcomm® SA8650P	10.192 ms	0 - 295 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	w8a16_mixed_fp16	Qualcomm® SA8255P	10.192 ms	0 - 295 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	w8a16_mixed_fp16	Qualcomm® Dragonwing™ IQ-9075	9.978 ms	0 - 2 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	w8a16_mixed_fp16	Qualcomm® Dragonwing™ IQ-X7181	10.483 ms	1 - 1 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	w8a16_mixed_fp16	Qualcomm® Dragonwing™ Q-8750	5.548 ms	0 - 312 MB	NPU

License

• The license for the original implementation of Mobile-Bert-Uncased-Google can be found here.

References

• MobileBERT: a Compact Task-Agnostic BERT for Resource-Limited Devices
• 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.