This repository provides the full LLAVA-LibMoE dataset, including LLaVA-665K and OneVision-1M2 image sources, organized into the following required directory tree:
libmoe/
└── data/
├── image_onevision/
├── coco/
│ └── train2017/
├── gqa/
│ └── images/
├── ocr_vqa/
│ └── images/
├── textvqa/
│ └── train_images/
└── vg/
├── VG_100K/
└──… See the full description on the dataset page:
https://huggingface.co/datasets/DavidNguyen/LLAVA-LibMoE.
