Pass number of graphs to dataloader

[dgm2]

Is it possible to pass as parameter the number of graphs in the datasets for splitting train, valid, test:
for example ZINC instead of 10.000, 1000, 1000 to 1,1,1 for local development?

Thanks!

[rusty1s]

You can simply slice the dataset after processing to achieve this, right?

train_dataset = train_dataset[:10]

[dgm2]

by doing ZINC[:1]
the slice object inside dataset is still set to the full dataset, do I need to slice each element inside slices?

[rusty1s]

Yes, slicing will just internally shrink the dataset (similar to how PyTorch slices Tensors). If you want to create a new dataset only containing the subset of elements, try ZINC[:1].copy().