how to set validation split for dfdc? about icpr2020dfdc

[zeemezeeme]

ok i found out that dfdc-35-5-10 are division of data according to folders. so basically i have to download all 470 GB of data in order to run the code, is there any way i can run this for small dataset ?

[nicobonne]

You can easily modify the splitting code to suit your needs, like defining less folders for training and validation.

[zeemezeeme]

thank you for your swift reply , i have tried it .
i changed this part according to my needs

if split == 'train':
split_df = df[df['folder'].isin(range(35))]
elif split == 'val':
split_df = df[df['folder'].isin(range(35, 40))]
elif split == 'test':
split_df = df[df['folder'].isin(range(40, 50))]

but still having this error while working on colab:

[nicobonne]

You are probably trying to sample from an empty list, due to some path mismatch. Check your paths.