Custom Dataset and DataLoader: list object is not callable

[ahmedramly]

Hi,
I have a dataset that is sorted as (subjects * trials) where each trial is a graph. I wrote a Dataset class to load the data to the DataLoader where the training is on n subjects and the test is on other subjects. the problem is when I load the data in the DataLoader and start training I get this error

the code I used for creating the dataset is this

class EEDataset(Dataset):
  def __init__(self, root, folders, transform=None, pre_transform=None):
    self.root = root
    self.folders = folders
    super(EEDataset, self).__init__(root, folders, transform, pre_transform)

  @property
  def raw_file_names(self):
      return []

  @property
  def processed_file_names(self):
    fnames = []
    for folder in self.folders:
      files = os.listdir(os.path.join(self.root, folder))
      nfiles = [os.path.join(folder, f) for f in files]
      fnames.extend(nfiles)
    return fnames

  def len(self):
      return len(self.processed_file_names)

  def get(self, idx):
    subj = torch.load(os.path.join(self.root, self.processed_file_names[idx]))
    return subj

and for loading the data

from torch_geometric.data import DataLoader

trainset = EEDataset(root_dir,  folders=Dirs[:20])  #Dirs is a list of dir names 'subject ids' 
testset = EEDataset(root_dir,  folders=Dirs[30:40])
train_loader = DataLoader(trainset, batch_size=64, shuffle=True)
test_loader = DataLoader(testset, batch_size=64, shuffle=True)

What is possibly wrong with the code that caused this issue?

[wsad1]

I think the problem is down to this line super(EEDataset, self).__init__(root, folders, transform, pre_transform). The signature of __init__ in the dataset class is

def __init__(self, root: Optional[str] = None,transform: Optional[Callable] = None,pre_transform:Optional[Callable] = None,pre_filter: Optional[Callable] = None):

The second argument is transform, in your case the second argument is folder that's the reason you get an error when self.transform gets executed. Simply removing folder from the super call should fix the issue.

[ahmedramly]

Thanks a lot @wsad1
It solved my problem