Electra-Bert-Base-Discrim-Google: Optimized for Qualcomm Devices

ELECTRABERT is a lightweight BERT model designed for efficient self-supervised learning of language representations. It can be used for identify unnatural or artificially modified text and as a backbone for various NLP tasks.

This is based on the implementation of Electra-Bert-Base-Discrim-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 2.37, ONNX Runtime 1.23.0	Download
QNN_DLC	float	Universal	QAIRT 2.42	Download
TFLITE	float	Universal	QAIRT 2.42, TFLite 2.17.0	Download

For more device-specific assets and performance metrics, visit Electra-Bert-Base-Discrim-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 Electra-Bert-Base-Discrim-Google on GitHub for usage instructions.

Model Details

Model Type: Model_use_case.text_generation

Model Stats:

• Model checkpoint: google/electra-base-discriminator
• Input resolution: 1x384
• Number of parameters: 109M
• Model size (float): 417 MB

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
Electra-Bert-Base-Discrim-Google	ONNX	float	Snapdragon® X Elite	24.086 ms	221 - 221 MB	NPU
Electra-Bert-Base-Discrim-Google	ONNX	float	Snapdragon® 8 Gen 3 Mobile	16.672 ms	0 - 532 MB	NPU
Electra-Bert-Base-Discrim-Google	ONNX	float	Qualcomm® QCS8550 (Proxy)	23.954 ms	0 - 227 MB	NPU
Electra-Bert-Base-Discrim-Google	ONNX	float	Qualcomm® QCS9075	29.184 ms	0 - 3 MB	NPU
Electra-Bert-Base-Discrim-Google	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	11.399 ms	0 - 498 MB	NPU
Electra-Bert-Base-Discrim-Google	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	8.259 ms	0 - 508 MB	NPU
Electra-Bert-Base-Discrim-Google	QNN_DLC	float	Snapdragon® X Elite	18.013 ms	0 - 0 MB	NPU
Electra-Bert-Base-Discrim-Google	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	11.887 ms	0 - 496 MB	NPU
Electra-Bert-Base-Discrim-Google	QNN_DLC	float	Qualcomm® QCS8275 (Proxy)	66.442 ms	0 - 419 MB	NPU
Electra-Bert-Base-Discrim-Google	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	17.212 ms	0 - 2 MB	NPU
Electra-Bert-Base-Discrim-Google	QNN_DLC	float	Qualcomm® SA8775P	98.946 ms	0 - 419 MB	NPU
Electra-Bert-Base-Discrim-Google	QNN_DLC	float	Qualcomm® QCS9075	21.677 ms	0 - 2 MB	NPU
Electra-Bert-Base-Discrim-Google	QNN_DLC	float	Qualcomm® QCS8450 (Proxy)	24.804 ms	0 - 450 MB	NPU
Electra-Bert-Base-Discrim-Google	QNN_DLC	float	Qualcomm® SA7255P	66.442 ms	0 - 419 MB	NPU
Electra-Bert-Base-Discrim-Google	QNN_DLC	float	Qualcomm® SA8295P	26.614 ms	0 - 373 MB	NPU
Electra-Bert-Base-Discrim-Google	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	8.06 ms	0 - 445 MB	NPU
Electra-Bert-Base-Discrim-Google	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	5.841 ms	0 - 449 MB	NPU
Electra-Bert-Base-Discrim-Google	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	11.971 ms	0 - 505 MB	NPU
Electra-Bert-Base-Discrim-Google	TFLITE	float	Qualcomm® QCS8275 (Proxy)	66.445 ms	0 - 427 MB	NPU
Electra-Bert-Base-Discrim-Google	TFLITE	float	Qualcomm® QCS8550 (Proxy)	17.246 ms	0 - 3 MB	NPU
Electra-Bert-Base-Discrim-Google	TFLITE	float	Qualcomm® SA8775P	21.325 ms	0 - 425 MB	NPU
Electra-Bert-Base-Discrim-Google	TFLITE	float	Qualcomm® QCS9075	22.353 ms	0 - 214 MB	NPU
Electra-Bert-Base-Discrim-Google	TFLITE	float	Qualcomm® QCS8450 (Proxy)	24.99 ms	0 - 453 MB	NPU
Electra-Bert-Base-Discrim-Google	TFLITE	float	Qualcomm® SA7255P	66.445 ms	0 - 427 MB	NPU
Electra-Bert-Base-Discrim-Google	TFLITE	float	Qualcomm® SA8295P	26.588 ms	0 - 379 MB	NPU
Electra-Bert-Base-Discrim-Google	TFLITE	float	Snapdragon® 8 Elite For Galaxy Mobile	8.091 ms	0 - 454 MB	NPU
Electra-Bert-Base-Discrim-Google	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	5.816 ms	0 - 460 MB	NPU

License

• The license for the original implementation of Electra-Bert-Base-Discrim-Google can be found here.

References

• ELECTRA: Pre-training Text Encoders as Discriminators Rather Than Generators
• 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.