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