OpenAI-Clip: Optimized for Qualcomm Devices

Contrastive Language-Image Pre-Training (CLIP) uses a ViT like transformer to get visual features and a causal language model to get the text features. Both the text and visual features can then be used for a variety of zero-shot learning tasks.

This is based on the implementation of OpenAI-Clip 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 1.30.0	Download
ONNX	w8a16	Universal	QAIRT 2.50, ONNX Runtime 1.30.0	Download
QNN_DLC	float	Universal	QAIRT 2.50	Download
QNN_DLC	w8a16	Universal	QAIRT 2.50	Download
TFLITE	float	Universal	QAIRT 2.50	Download

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

Model Details

Model Type: Model_use_case.image_classification

Model Stats:

• Image input resolution: 224x224
• Model checkpoint: ViT-B/16
• Model size (float): 571 MB
• Number of parameters: 150M
• Text context length: 77

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
OpenAI-Clip	ONNX	float	Snapdragon® 8 Elite Gen 5 For Galaxy Mobile	10.414 ms	1 - 509 MB	NPU
OpenAI-Clip	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	14.432 ms	1 - 505 MB	NPU
OpenAI-Clip	ONNX	float	Snapdragon® X2 Elite	11.102 ms	2 - 2 MB	NPU
OpenAI-Clip	ONNX	float	Snapdragon® X Elite	27.428 ms	293 - 293 MB	NPU
OpenAI-Clip	ONNX	float	Snapdragon® 8 Gen 3 Mobile	18.48 ms	1 - 713 MB	NPU
OpenAI-Clip	ONNX	float	Snapdragon® 8 Gen 1 Mobile	33.254 ms	1 - 700 MB	NPU
OpenAI-Clip	ONNX	float	Qualcomm® Dragonwing™ IQ-8275	31.356 ms	0 - 5 MB	NPU
OpenAI-Clip	ONNX	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	26.34 ms	0 - 323 MB	NPU
OpenAI-Clip	ONNX	float	Qualcomm® QCS8450	33.254 ms	1 - 700 MB	NPU
OpenAI-Clip	ONNX	float	Qualcomm® Dragonwing™ IQ-9075	31.178 ms	0 - 4 MB	NPU
OpenAI-Clip	ONNX	float	Qualcomm® Dragonwing™ IQ-X7181	27.428 ms	293 - 293 MB	NPU
OpenAI-Clip	ONNX	float	Qualcomm® Dragonwing™ Q-8750	14.432 ms	1 - 505 MB	NPU
OpenAI-Clip	ONNX	w8a16	Snapdragon® 8 Elite Gen 5 For Galaxy Mobile	10.523 ms	0 - 627 MB	NPU
OpenAI-Clip	ONNX	w8a16	Snapdragon® 8 Elite For Galaxy Mobile	13.977 ms	0 - 587 MB	NPU
OpenAI-Clip	ONNX	w8a16	Snapdragon® X2 Elite	11.0 ms	1 - 1 MB	NPU
OpenAI-Clip	ONNX	w8a16	Snapdragon® X Elite	24.436 ms	172 - 172 MB	NPU
OpenAI-Clip	ONNX	w8a16	Snapdragon® 8 Gen 3 Mobile	16.901 ms	0 - 699 MB	NPU
OpenAI-Clip	ONNX	w8a16	Snapdragon® 8 Gen 1 Mobile	32.249 ms	0 - 703 MB	NPU
OpenAI-Clip	ONNX	w8a16	Qualcomm® Dragonwing™ QCS6490	113.314 ms	1 - 4 MB	NPU
OpenAI-Clip	ONNX	w8a16	Qualcomm® Dragonwing™ IQ-8275	22.465 ms	0 - 5 MB	NPU
OpenAI-Clip	ONNX	w8a16	Qualcomm® Dragonwing™ QCS8550 (Proxy)	23.735 ms	0 - 195 MB	NPU
OpenAI-Clip	ONNX	w8a16	Qualcomm® QCS8450	32.249 ms	0 - 703 MB	NPU
OpenAI-Clip	ONNX	w8a16	Qualcomm® Dragonwing™ IQ-9075	24.034 ms	0 - 4 MB	NPU
OpenAI-Clip	ONNX	w8a16	Qualcomm® Dragonwing™ IQ-X7181	24.436 ms	172 - 172 MB	NPU
OpenAI-Clip	ONNX	w8a16	Qualcomm® Dragonwing™ Q-6690	270.271 ms	0 - 713 MB	NPU
OpenAI-Clip	ONNX	w8a16	Qualcomm® Dragonwing™ Q-7790	41.106 ms	1 - 630 MB	NPU
OpenAI-Clip	ONNX	w8a16	Qualcomm® Dragonwing™ Q-8750	13.977 ms	0 - 587 MB	NPU
OpenAI-Clip	ONNX	w8a16	Snapdragon® 7 Gen 4 Mobile	41.106 ms	1 - 630 MB	NPU
OpenAI-Clip	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 For Galaxy Mobile	13.197 ms	0 - 490 MB	NPU
OpenAI-Clip	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	16.632 ms	1 - 483 MB	NPU
OpenAI-Clip	QNN_DLC	float	Snapdragon® X2 Elite	14.147 ms	1 - 1 MB	NPU
OpenAI-Clip	QNN_DLC	float	Snapdragon® X Elite	33.039 ms	1 - 1 MB	NPU
OpenAI-Clip	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	21.362 ms	0 - 690 MB	NPU
OpenAI-Clip	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	37.709 ms	1 - 815 MB	NPU
OpenAI-Clip	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-8275	35.747 ms	1 - 4 MB	NPU
OpenAI-Clip	QNN_DLC	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	31.548 ms	1 - 2 MB	NPU
OpenAI-Clip	QNN_DLC	float	Qualcomm® QCS8450	37.709 ms	1 - 815 MB	NPU
OpenAI-Clip	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-9075	36.446 ms	1 - 3 MB	NPU
OpenAI-Clip	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-X7181	33.039 ms	1 - 1 MB	NPU
OpenAI-Clip	QNN_DLC	float	Qualcomm® Dragonwing™ Q-8750	16.632 ms	1 - 483 MB	NPU
OpenAI-Clip	QNN_DLC	float	Qualcomm® SA8295P	37.564 ms	1 - 590 MB	NPU
OpenAI-Clip	QNN_DLC	w8a16	Snapdragon® 8 Elite Gen 5 For Galaxy Mobile	11.493 ms	0 - 629 MB	NPU
OpenAI-Clip	QNN_DLC	w8a16	Snapdragon® 8 Elite For Galaxy Mobile	15.278 ms	0 - 591 MB	NPU
OpenAI-Clip	QNN_DLC	w8a16	Snapdragon® X2 Elite	12.18 ms	0 - 0 MB	NPU
OpenAI-Clip	QNN_DLC	w8a16	Snapdragon® X Elite	28.168 ms	0 - 0 MB	NPU
OpenAI-Clip	QNN_DLC	w8a16	Snapdragon® 8 Gen 3 Mobile	18.835 ms	0 - 703 MB	NPU
OpenAI-Clip	QNN_DLC	w8a16	Qualcomm® Dragonwing™ IQ-8275	24.98 ms	0 - 3 MB	NPU
OpenAI-Clip	QNN_DLC	w8a16	Qualcomm® Dragonwing™ QCS8550 (Proxy)	27.128 ms	0 - 3 MB	NPU
OpenAI-Clip	QNN_DLC	w8a16	Qualcomm® SA8650P	26.675 ms	0 - 567 MB	NPU
OpenAI-Clip	QNN_DLC	w8a16	Qualcomm® SA8255P	26.675 ms	0 - 567 MB	NPU
OpenAI-Clip	QNN_DLC	w8a16	Qualcomm® Dragonwing™ IQ-9075	26.92 ms	0 - 3 MB	NPU
OpenAI-Clip	QNN_DLC	w8a16	Qualcomm® Dragonwing™ IQ-X7181	28.168 ms	0 - 0 MB	NPU
OpenAI-Clip	QNN_DLC	w8a16	Qualcomm® Dragonwing™ Q-6690	400.308 ms	1 - 693 MB	NPU
OpenAI-Clip	QNN_DLC	w8a16	Qualcomm® Dragonwing™ Q-7790	37.034 ms	0 - 611 MB	NPU
OpenAI-Clip	QNN_DLC	w8a16	Qualcomm® Dragonwing™ Q-8750	15.278 ms	0 - 591 MB	NPU
OpenAI-Clip	QNN_DLC	w8a16	Snapdragon® 7 Gen 4 Mobile	37.034 ms	0 - 611 MB	NPU
OpenAI-Clip	TFLITE	float	Snapdragon® 8 Elite Gen 5 For Galaxy Mobile	12.81 ms	0 - 497 MB	NPU
OpenAI-Clip	TFLITE	float	Snapdragon® 8 Elite For Galaxy Mobile	17.062 ms	0 - 489 MB	NPU
OpenAI-Clip	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	21.145 ms	0 - 689 MB	NPU
OpenAI-Clip	TFLITE	float	Snapdragon® 8 Gen 1 Mobile	36.358 ms	0 - 677 MB	NPU
OpenAI-Clip	TFLITE	float	Qualcomm® Dragonwing™ IQ-8275	35.003 ms	0 - 297 MB	NPU
OpenAI-Clip	TFLITE	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	30.909 ms	0 - 4 MB	NPU
OpenAI-Clip	TFLITE	float	Qualcomm® SA8775P	520.632 ms	11 - 35 MB	CPU
OpenAI-Clip	TFLITE	float	Qualcomm® SA8650P	520.632 ms	11 - 35 MB	CPU
OpenAI-Clip	TFLITE	float	Qualcomm® SA8255P	520.632 ms	11 - 35 MB	CPU
OpenAI-Clip	TFLITE	float	Qualcomm® QCS8450	36.358 ms	0 - 677 MB	NPU
OpenAI-Clip	TFLITE	float	Qualcomm® Dragonwing™ IQ-9075	35.45 ms	0 - 297 MB	NPU
OpenAI-Clip	TFLITE	float	Qualcomm® Dragonwing™ Q-8750	17.062 ms	0 - 489 MB	NPU
OpenAI-Clip	TFLITE	float	Qualcomm® SA7255P	750.392 ms	11 - 35 MB	CPU
OpenAI-Clip	TFLITE	float	Qualcomm® SA8295P	37.239 ms	0 - 440 MB	NPU

License

• The license for the original implementation of OpenAI-Clip can be found here.

References

• Learning Transferable Visual Models From Natural Language Supervision
• 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.