Photonic device dataset for multimodal fine-tuning, including image-text pairs and text-completion pairs.