slices from collate_fn returns unexpected result

I am trying to create a custom dataset. My dataset.data is Data(edge_index=[2, 93003], edge_to_id=[1], test_mask=[93003], train_mask=[93003], val_mask=[93003], vertice_to_id=[1], y=[93003]). The number in square brace is shape

However, dataset.slice is {'edge_index': tensor([0, 93003]), 'y': tensor([0, 93003]), 'train_mask': tensor([0, 93003]), 'val_mask': tensor([0, 93003]), 'test_mask': tensor([0, 93003])}. The number in ([...]) is actual value. Therefore, I cannot get an item with index larger than 2, while I have 93003 datapoint.

I have seen that slices typically has a longer length. What did I do wrong? My code of creating the dataset is

edge_index = torch.LongTensor(edge_index)
data = Data(edge_index=edge_index.t().contiguous())
data.y = torch.tensor(edge_type)
data.train_mask = torch.cat((torch.ones(n_train, dtype=torch.bool), torch.zeros(n_val+n_test, dtype=torch.bool)))
data.val_mask = torch.cat((torch.zeros(n_train, dtype=torch.bool), torch.ones(n_val, dtype=torch.bool), torch.zeros(n_test, dtype=torch.bool)))
data.test_mask = torch.cat((torch.zeros(n_train+n_val, dtype=torch.bool), torch.ones(n_test, dtype=torch.bool)))
collated_data, slices = self.collate([data])
torch.save((collated_data, slices), *self.processed_paths)

[minhtriet]

It is done

[minhtriet]

In addition, it has been noted that the WN18 andFB15k datasets suffer from test set leakage, due to inverse relations from the training set being present in the test set – however, the extent of this issue has so far not been quantified.

… but enough for them to create a new dataset. Source