What transform I should have in order to use LabelSample

Hello, I run into this issue while creating dataloader using torchio. I defined a sampler with LabelSampler, and it gives me this error:

RuntimeError: Empty probability map found: ...
Voxels with positive probability might be near the image border.
If you suspect that this is the case, try adding a padding transform

The transform I’m using is

training_transform = tio.Compose([
    tio.ToCanonical(),
    tio.Pad(int(patch_size[0]/2)),
    tio.CropOrPad(patch_size),
    tio.ZNormalization(masking_method=tio.ZNormalization.mean),
    tio.OneHot(3),
])

I means I have to pad after cropandpad?

Thanks

[fepegar]

In that case, you might want to lower the max_queue_length.

[fepegar]

[Please use syntax highlighting for code snippets]

Given the size of your volumes and patches, I suggest using a larger number of samples per volume, maybe 30 or 50. That will be less computationally expensive, as you will need to load and transform volumes less often.