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 2.45, ONNX Runtime 1.27.1	Download
ONNX	w8a16	Universal	QAIRT 2.45, ONNX Runtime 1.27.1	Download
QNN_DLC	float	Universal	QAIRT 2.45	Download
QNN_DLC	w8a16	Universal	QAIRT 2.45	Download
TFLITE	float	Universal	QAIRT 2.45	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® X2 Elite	15.006 ms	2 - 2 MB	NPU
OpenAI-Clip	ONNX	float	Snapdragon® X Elite	33.203 ms	295 - 295 MB	NPU
OpenAI-Clip	ONNX	float	Snapdragon® 8 Gen 3 Mobile	21.524 ms	0 - 1185 MB	NPU
OpenAI-Clip	ONNX	float	Snapdragon® 8 Gen 1 Mobile	47.606 ms	1 - 894 MB	NPU
OpenAI-Clip	ONNX	float	Qualcomm® Dragonwing™ IQ-8275	36.0 ms	0 - 4 MB	NPU
OpenAI-Clip	ONNX	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	31.887 ms	0 - 323 MB	NPU
OpenAI-Clip	ONNX	float	Qualcomm® QCS8450	47.606 ms	1 - 894 MB	NPU
OpenAI-Clip	ONNX	float	Qualcomm® Dragonwing™ IQ-9075	36.741 ms	0 - 4 MB	NPU
OpenAI-Clip	ONNX	float	Qualcomm® Dragonwing™ IQ-X7181	33.203 ms	295 - 295 MB	NPU
OpenAI-Clip	ONNX	float	Qualcomm® Dragonwing™ Q-8750	16.93 ms	1 - 957 MB	NPU
OpenAI-Clip	ONNX	float	Snapdragon® 8 Elite Mobile	16.93 ms	1 - 957 MB	NPU
OpenAI-Clip	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	13.451 ms	1 - 806 MB	NPU
OpenAI-Clip	ONNX	w8a16	Snapdragon® X2 Elite	13.417 ms	1 - 1 MB	NPU
OpenAI-Clip	ONNX	w8a16	Snapdragon® X Elite	27.039 ms	173 - 173 MB	NPU
OpenAI-Clip	ONNX	w8a16	Snapdragon® 8 Gen 3 Mobile	18.746 ms	0 - 753 MB	NPU
OpenAI-Clip	ONNX	w8a16	Snapdragon® 8 Gen 1 Mobile	39.685 ms	1 - 759 MB	NPU
OpenAI-Clip	ONNX	w8a16	Qualcomm® Dragonwing™ QCS6490	145.115 ms	1 - 4 MB	NPU
OpenAI-Clip	ONNX	w8a16	Qualcomm® Dragonwing™ IQ-8275	24.74 ms	0 - 4 MB	NPU
OpenAI-Clip	ONNX	w8a16	Qualcomm® Dragonwing™ QCS8550 (Proxy)	26.287 ms	0 - 4 MB	NPU
OpenAI-Clip	ONNX	w8a16	Qualcomm® QCS8450	39.685 ms	1 - 759 MB	NPU
OpenAI-Clip	ONNX	w8a16	Qualcomm® Dragonwing™ IQ-9075	25.895 ms	0 - 3 MB	NPU
OpenAI-Clip	ONNX	w8a16	Qualcomm® Dragonwing™ IQ-X7181	27.039 ms	173 - 173 MB	NPU
OpenAI-Clip	ONNX	w8a16	Qualcomm® Dragonwing™ Q-6690	267.969 ms	1 - 666 MB	NPU
OpenAI-Clip	ONNX	w8a16	Qualcomm® Dragonwing™ Q-7790	37.402 ms	1 - 670 MB	NPU
OpenAI-Clip	ONNX	w8a16	Qualcomm® Dragonwing™ Q-8750	14.826 ms	0 - 640 MB	NPU
OpenAI-Clip	ONNX	w8a16	Snapdragon® 8 Elite Mobile	14.826 ms	0 - 640 MB	NPU
OpenAI-Clip	ONNX	w8a16	Snapdragon® 8 Elite Gen 5 Mobile	12.91 ms	0 - 684 MB	NPU
OpenAI-Clip	ONNX	w8a16	Snapdragon® 7 Gen 4 Mobile	37.402 ms	1 - 670 MB	NPU
OpenAI-Clip	QNN_DLC	float	Snapdragon® X2 Elite	14.729 ms	1 - 1 MB	NPU
OpenAI-Clip	QNN_DLC	float	Snapdragon® X Elite	33.367 ms	1 - 1 MB	NPU
OpenAI-Clip	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	21.882 ms	1 - 927 MB	NPU
OpenAI-Clip	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	46.662 ms	0 - 855 MB	NPU
OpenAI-Clip	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-8275	36.472 ms	1 - 4 MB	NPU
OpenAI-Clip	QNN_DLC	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	31.561 ms	1 - 496 MB	NPU
OpenAI-Clip	QNN_DLC	float	Qualcomm® SA8775P	36.675 ms	1 - 700 MB	NPU
OpenAI-Clip	QNN_DLC	float	Qualcomm® SA8650P	36.675 ms	1 - 700 MB	NPU
OpenAI-Clip	QNN_DLC	float	Qualcomm® SA8255P	36.675 ms	1 - 700 MB	NPU
OpenAI-Clip	QNN_DLC	float	Qualcomm® QCS8450	46.662 ms	0 - 855 MB	NPU
OpenAI-Clip	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-9075	37.141 ms	1 - 3 MB	NPU
OpenAI-Clip	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-X7181	33.367 ms	1 - 1 MB	NPU
OpenAI-Clip	QNN_DLC	float	Qualcomm® Dragonwing™ Q-8750	17.263 ms	0 - 694 MB	NPU
OpenAI-Clip	QNN_DLC	float	Qualcomm® SA7255P	99.039 ms	1 - 704 MB	NPU
OpenAI-Clip	QNN_DLC	float	Qualcomm® SA8295P	38.463 ms	1 - 798 MB	NPU
OpenAI-Clip	QNN_DLC	float	Snapdragon® 8 Elite Mobile	17.263 ms	0 - 694 MB	NPU
OpenAI-Clip	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	13.35 ms	0 - 703 MB	NPU
OpenAI-Clip	QNN_DLC	w8a16	Snapdragon® X2 Elite	14.418 ms	0 - 0 MB	NPU
OpenAI-Clip	QNN_DLC	w8a16	Snapdragon® X Elite	33.348 ms	0 - 0 MB	NPU
OpenAI-Clip	QNN_DLC	w8a16	Snapdragon® 8 Gen 3 Mobile	22.824 ms	0 - 747 MB	NPU
OpenAI-Clip	QNN_DLC	w8a16	Qualcomm® Dragonwing™ IQ-8275	30.108 ms	0 - 3 MB	NPU
OpenAI-Clip	QNN_DLC	w8a16	Qualcomm® Dragonwing™ QCS8550 (Proxy)	32.881 ms	0 - 665 MB	NPU
OpenAI-Clip	QNN_DLC	w8a16	Qualcomm® SA8775P	32.453 ms	0 - 603 MB	NPU
OpenAI-Clip	QNN_DLC	w8a16	Qualcomm® SA8650P	32.453 ms	0 - 603 MB	NPU
OpenAI-Clip	QNN_DLC	w8a16	Qualcomm® SA8255P	32.453 ms	0 - 603 MB	NPU
OpenAI-Clip	QNN_DLC	w8a16	Qualcomm® Dragonwing™ IQ-9075	32.143 ms	0 - 2 MB	NPU
OpenAI-Clip	QNN_DLC	w8a16	Qualcomm® Dragonwing™ IQ-X7181	33.348 ms	0 - 0 MB	NPU
OpenAI-Clip	QNN_DLC	w8a16	Qualcomm® Dragonwing™ Q-6690	446.478 ms	0 - 687 MB	NPU
OpenAI-Clip	QNN_DLC	w8a16	Qualcomm® Dragonwing™ Q-7790	41.615 ms	0 - 657 MB	NPU
OpenAI-Clip	QNN_DLC	w8a16	Qualcomm® Dragonwing™ Q-8750	18.359 ms	0 - 633 MB	NPU
OpenAI-Clip	QNN_DLC	w8a16	Qualcomm® SA7255P	67.742 ms	1 - 604 MB	NPU
OpenAI-Clip	QNN_DLC	w8a16	Snapdragon® 8 Elite Mobile	18.359 ms	0 - 633 MB	NPU
OpenAI-Clip	QNN_DLC	w8a16	Snapdragon® 8 Elite Gen 5 Mobile	14.479 ms	0 - 677 MB	NPU
OpenAI-Clip	QNN_DLC	w8a16	Snapdragon® 7 Gen 4 Mobile	41.615 ms	0 - 657 MB	NPU
OpenAI-Clip	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	21.756 ms	0 - 898 MB	NPU
OpenAI-Clip	TFLITE	float	Snapdragon® 8 Gen 1 Mobile	40.291 ms	0 - 906 MB	NPU
OpenAI-Clip	TFLITE	float	Qualcomm® Dragonwing™ IQ-8275	35.022 ms	0 - 297 MB	NPU
OpenAI-Clip	TFLITE	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	31.532 ms	0 - 4 MB	NPU
OpenAI-Clip	TFLITE	float	Qualcomm® SA8775P	35.976 ms	0 - 670 MB	NPU
OpenAI-Clip	TFLITE	float	Qualcomm® SA8650P	35.976 ms	0 - 670 MB	NPU
OpenAI-Clip	TFLITE	float	Qualcomm® SA8255P	35.976 ms	0 - 670 MB	NPU
OpenAI-Clip	TFLITE	float	Qualcomm® QCS8450	40.291 ms	0 - 906 MB	NPU
OpenAI-Clip	TFLITE	float	Qualcomm® Dragonwing™ IQ-9075	36.589 ms	0 - 296 MB	NPU
OpenAI-Clip	TFLITE	float	Qualcomm® Dragonwing™ Q-8750	17.215 ms	0 - 693 MB	NPU
OpenAI-Clip	TFLITE	float	Qualcomm® SA7255P	95.568 ms	0 - 665 MB	NPU
OpenAI-Clip	TFLITE	float	Qualcomm® SA8295P	38.34 ms	0 - 621 MB	NPU
OpenAI-Clip	TFLITE	float	Snapdragon® 8 Elite Mobile	17.215 ms	0 - 693 MB	NPU
OpenAI-Clip	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	13.028 ms	0 - 537 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.