Why
The documentation describes a how-to, to use QuAC on the user's dataset. Very early on, we give them the option of downloading an example dataset and classifier checkpoint and following along. For somebody who just wants to try things, however, they may not be able to or want to spend the compute necessary to train the full thing.
We should, therefore, give the users at least one fully fledged example to play with, including trained models and generated images + masks.
How
Use disc_b or fictus as the example dataset.
Use zenodo or figshare to store the full setup.
Why
The documentation describes a how-to, to use QuAC on the user's dataset. Very early on, we give them the option of downloading an example dataset and classifier checkpoint and following along. For somebody who just wants to try things, however, they may not be able to or want to spend the compute necessary to train the full thing.
We should, therefore, give the users at least one fully fledged example to play with, including trained models and generated images + masks.
How
Use
disc_borfictusas the example dataset.Use
zenodoorfigshareto store the full setup.