streamline the fine-tuning process for multimodal models: PaliGemma 2, Florence-2, and Qwen2.5-VL - View it on GitHub