How to properly modify attribute of Batch objects?

[ekosman]

Assuming I have already constructed a Batch object, that could contain either Data or HeteroData - what is the proper way to modify its attributes? modifying them directly won't work because the objects returned from Batch.to_data_list() won't recognize all attributes. For example:

b = Batch.from_data_list([HeteroData()])
b["a"].x = torch.randn(3,3)
data = b.to_data_list()[0]
contains_attrs = "x" in data["a"]
print(f"Does the data contains any nodes? {contains_attrs }")

Edit: I just noticed that all attributes must be contained within _slice_dict - which is created only if from_data_list was used to create the batch. Is there a pre implemented way to manually add items to this dictionary?

[rusty1s]

The only way currently would be to modify _slice_dict by yourself for this. Otherwise, this is not possible as we don't know how to split x in this case :(