BloodAxe/Catalyst-Inria-Segmentation-Example

tiles training mode

I8dNLo opened this issue · 1 comments

In inria/dataset at "tiles" training mode the last usage of the returned sampler is
train_sampler = None
but

torch/utils/data/sampler

expects it to be a positive integer.

ValueError: num_samples should be a positive integer value, but got num_samples=0

This should be fixed since 6e1d47a