daruokta/t5gemma2-indonesia-vision-formatted

T5Gemma-2 Indonesian Vision & Multimodal Dataset A high-quality Indonesian language visual instruction tuning and preference alignment dataset, specifically formatted for multimodal sequence-to-sequence (Seq2Seq) models like T5-Gemma-2 Vision (e.g., google/t5gemma-2-270m-270m). This dataset provides embedded image features (datasets.Image) along with multi-turn conversations in Bahasa Indonesia, facilitating instruction-following and preference alignment training on… See the full description on the dataset page: https://huggingface.co/datasets/daruokta/t5gemma2-indonesia-vision-formatted.