在hugging face上下载的liuhaotian--LLaVA-CC3M-Pretrain-595K数据集没有cc3m.csv文件

[lsuae]

您好，我根据tvl_llama下的readme的指示下载CCM3数据集，发现下载到的数据集中并没有.csv文件，这个文件是需要自己写吗？还是说我理解错了，恳请指教，谢谢！

[Max-Fu]

Hi,

Thanks for the question. The data server of our lab is down, so I can't upload the csv file. I remember I downloaded from somewhere else, please check the following two links for references:

1. https://huggingface.co/spaces/flax-community/dalle-mini/commit/75b01a0a3a29bb2eb6962f5f2fdf160e5c784647
2. https://github.com/rom1504/img2dataset/blob/main/dataset_examples/cc3m.md

I think it is also fine to create a two-column csv file from the json that is available on:
https://huggingface.co/datasets/liuhaotian/LLaVA-CC3M-Pretrain-595K?row=0

Let me know if it works! I will adjust the readme accordingly!