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 2.45, ONNX Runtime 1.27.1	Download
QNN_DLC	float	Universal	QAIRT 2.45	Download
TFLITE	float	Universal	QAIRT 2.45	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® X2 Elite	12.229 ms	1 - 1 MB	NPU
Mobile-Bert-Uncased-Google	ONNX	float	Snapdragon® X Elite	23.162 ms	80 - 80 MB	NPU
Mobile-Bert-Uncased-Google	ONNX	float	Snapdragon® 8 Gen 3 Mobile	15.489 ms	0 - 366 MB	NPU
Mobile-Bert-Uncased-Google	ONNX	float	Snapdragon® 8 Gen 1 Mobile	30.419 ms	0 - 484 MB	NPU
Mobile-Bert-Uncased-Google	ONNX	float	Qualcomm® Dragonwing™ IQ-8275	22.813 ms	0 - 4 MB	NPU
Mobile-Bert-Uncased-Google	ONNX	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	22.044 ms	0 - 83 MB	NPU
Mobile-Bert-Uncased-Google	ONNX	float	Qualcomm® QCS8450	30.419 ms	0 - 484 MB	NPU
Mobile-Bert-Uncased-Google	ONNX	float	Qualcomm® Dragonwing™ IQ-9075	25.935 ms	0 - 3 MB	NPU
Mobile-Bert-Uncased-Google	ONNX	float	Qualcomm® Dragonwing™ IQ-X7181	23.162 ms	80 - 80 MB	NPU
Mobile-Bert-Uncased-Google	ONNX	float	Qualcomm® Dragonwing™ Q-8750	13.409 ms	0 - 204 MB	NPU
Mobile-Bert-Uncased-Google	ONNX	float	Snapdragon® 8 Elite Mobile	13.409 ms	0 - 204 MB	NPU
Mobile-Bert-Uncased-Google	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	11.761 ms	0 - 204 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	float	Snapdragon® X2 Elite	12.628 ms	1 - 1 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	float	Snapdragon® X Elite	23.569 ms	1 - 1 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	15.736 ms	0 - 305 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	29.562 ms	0 - 423 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-8275	22.971 ms	0 - 2 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-8275	53.614 ms	0 - 193 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	22.345 ms	0 - 2 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	float	Qualcomm® SA8775P	25.761 ms	0 - 193 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	float	Qualcomm® SA8650P	25.761 ms	0 - 193 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	float	Qualcomm® SA8255P	25.761 ms	0 - 193 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	float	Qualcomm® QCS8450	29.562 ms	0 - 423 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-9075	25.975 ms	0 - 2 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-X7181	23.569 ms	1 - 1 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	float	Qualcomm® Dragonwing™ Q-8750	13.214 ms	0 - 219 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	float	Qualcomm® SA7255P	53.614 ms	0 - 193 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	float	Qualcomm® SA8295P	27.551 ms	0 - 309 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	float	Snapdragon® 8 Elite Mobile	13.214 ms	0 - 219 MB	NPU
Mobile-Bert-Uncased-Google	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	11.921 ms	0 - 224 MB	NPU
Mobile-Bert-Uncased-Google	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	15.073 ms	0 - 313 MB	NPU
Mobile-Bert-Uncased-Google	TFLITE	float	Snapdragon® 8 Gen 1 Mobile	29.967 ms	0 - 416 MB	NPU
Mobile-Bert-Uncased-Google	TFLITE	float	Qualcomm® Dragonwing™ IQ-8275	22.102 ms	0 - 83 MB	NPU
Mobile-Bert-Uncased-Google	TFLITE	float	Qualcomm® Dragonwing™ IQ-8275	52.051 ms	0 - 198 MB	NPU
Mobile-Bert-Uncased-Google	TFLITE	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	21.53 ms	0 - 3 MB	NPU
Mobile-Bert-Uncased-Google	TFLITE	float	Qualcomm® SA8775P	24.863 ms	0 - 198 MB	NPU
Mobile-Bert-Uncased-Google	TFLITE	float	Qualcomm® SA8650P	24.863 ms	0 - 198 MB	NPU
Mobile-Bert-Uncased-Google	TFLITE	float	Qualcomm® SA8255P	24.863 ms	0 - 198 MB	NPU
Mobile-Bert-Uncased-Google	TFLITE	float	Qualcomm® QCS8450	29.967 ms	0 - 416 MB	NPU
Mobile-Bert-Uncased-Google	TFLITE	float	Qualcomm® Dragonwing™ IQ-9075	25.326 ms	0 - 82 MB	NPU
Mobile-Bert-Uncased-Google	TFLITE	float	Qualcomm® Dragonwing™ Q-8750	12.767 ms	0 - 232 MB	NPU
Mobile-Bert-Uncased-Google	TFLITE	float	Qualcomm® SA7255P	52.051 ms	0 - 198 MB	NPU
Mobile-Bert-Uncased-Google	TFLITE	float	Qualcomm® SA8295P	27.232 ms	0 - 297 MB	NPU
Mobile-Bert-Uncased-Google	TFLITE	float	Snapdragon® 8 Elite Mobile	12.767 ms	0 - 232 MB	NPU
Mobile-Bert-Uncased-Google	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	11.513 ms	0 - 230 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.