How do I access data from only one group?

[krishnap25]

Hello, Thanks for the fantastic library!

I have two questions:

1. Is there any way I can get a per-group dataloader in wilds? This will help with, for instance, training a separate model for each group of data.
2. Can I change the split of data for each dataset? My application requires 50% of the data for each group/domain for testing.

Thanks!

[kohpangwei]

Hello! Thanks for your interest in using WILDS.

1. Not right now, sorry. (If you're interested in writing a PR, we'd be happy to review it!) One thing to note: some datasets have many groups, so initializing a separate dataloader for each group might be a problem. (We tried something like this when developing WILDS and ran into performance issues.)

2. You're welcome to change the data split for your own research, though that means that any performance numbers won't be comparable to others who are using the standard datasets. To change the split, feel free to modify the existing dataset classes. (For example, you could write a function that takes a dataset, looks at split_array, and reassigns accordingly.)

[krishnap25]

Hey @kohpangwei, thanks for your prompt response!

1. While I do not have the cycles to work on a PR now, I will keep you posted if I get around to doing it.

2. Got it, thanks!

[kohpangwei]

Sounds good, thanks!