import datasets
ds = datasets.load_dataset("atasoglu/flickr8k-dataset", data_dir="data")
print(ds)

DatasetDict({
    train: Dataset({
        features: ['image_id', 'image_path', 'captions'],
        num_rows: 6000
    })
    test: Dataset({
        features: ['image_id', 'image_path', 'captions'],
        num_rows: 1000
    })
    validation: Dataset({
        features: ['image_id', 'image_path', 'captions'],
        num_rows: 1000
    })
})