I’m trying to load a local custom image captioning dataset that I have filtered from a much larger one but I’m having issues where it seems that the Image Folder doesn’t seem to be picking up the captions/labels for the images or the csv file that they are contained on
Hi! What’s the structure of the Train_Images and Test_Images folders and the contents of the metadata file(s)? You can obtain the folder structure using the tree command.
Thanks for responding. The structure is as shown in this image. Also how would I use the tree command? Thanks.
The metadata files must be named metadata.csv as explained here.
Can you replace data_dir= with data_files= in the last load_dataset call and try again?
Hi, thanks for responding. I fixed the issue yesterday once I changed my metadata from a csv to a jsonl file it just worked