Only captions of dataset without images

#2
by am-pak - opened

Hi,
Thank you for this great platform. I want to analyse the dataset used in Flava paper. I have used the following command to get the coco datast:
pmd = load_dataset("facebook/pmd",'coco', use_auth_token=True, streaming=True)

It successfully loads the dataset into my colab notebook as iterable object. But the issue is when I iterate through the train split of coco dataset, it takes very long time. I want to get only the text captions of the dataset without any image. Therefore I want to ask is there any way through which I can only get the text captions so that I can iterate through it more quickly. Thank you.

You can remove the columns that you don't want as specified in Hugging Face dataset documentation: https://huggingface.co/docs/datasets/process#remove

Sign up or log in to comment