Dataset split

❓ Questions & Help

Hi

So i’m currently trying to combine two very similar datasets. I have them at two distinct locations and my goal would be to load both into the same dataset, using one as the training data and the other one as the test and val data. Now when using an InMemoryDataset, i thought about using masks for that. But as far as i have found in the datasets provided, masks are used to use certain parts of graphs for training and other parts for testing (am i right, or am i misunderstanding them?)

Would there be a way, to declare certain graphs as part of the training data, and others as part of the test data?

Sorry if it’s a simple question, but i haven’t worked with masks before.

[rusty1s]

Just split them into train and test set and load them separaretly, e.g., like here.

[SaschaStenger]

Yes, for whatever reason it crashes when i try to load it this way. But I just made a workaround. I initialize the encode that i’m using on both datasets beforehand and then pass it to both datasets as an argument, when creating them, so that both are using the same encoder instance.
It’s possible, that i have made a mistake, when coding the above section somewhere, but i cant find it and so i’ll just be moving on to this solution.