DGDataset indexing for training

[amundra15]

Could you please explain how the datasets are getting samples?

Suppose I am trying to run exp-id 60 with the GTA and DG datasets. Here I see that initially the source dataset is sampled and then source2. Is that correct?

Can you also confirm the expected sizes of these two datasets?

[amundra15]

And could it be possible that source2 is getting completely ignored while building the Dataset here?