what is in the pretrain_data for mobileVLM2? @er-muyue

[xiyuanhao]

It is really great job and I want to reproduce your work。 Can you help me about this issue?
└── pretrain_data
├── share-captioner_coco_lcs_sam_1246k_1107.json
├── blip_laion_cc_sbu_558k.json
├── images
├── coco
│ └── train2017
├── llava
│ └── llava_pretrain
└── sam
└── images
where can I get the llava_pretrain , and this images? can you give me the link ? Thank you very much.

[huyiming2018]

You can refer to the original LLaVA paper to obtain the corresponding training data.

[daichi1207]

you can get all the data following this https://github.com/Meituan-AutoML/MobileVLM?tab=readme-ov-file#2%EF%B8%8F%E2%83%A3-prepare-data
llava_pretrain can be obtained from https://huggingface.co/datasets/liuhaotian/LLaVA-Pretrain/blob/main/images.zip