You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Using the OSCD datamodule with a segmentation task results in the error:
RuntimeError: Input type (torch.cuda.LongTensor) and weight type (torch.cuda.FloatTensor) should be the same
It is necessary to cast the image to a float to resolve this (using x = batch["image"].float()), which I verified using a custom task. However I'm pretty sure this is not the intention to do this
This relates to #985. We currently don't have a good way to test this. OSCD isn't really intended for use with SemanticSegmentationTask, we need a new trainer for change detection. I would be fine with a PR that simply casts the image to float32.
Description
Using the OSCD datamodule with a segmentation task results in the error:
It is necessary to cast the image to a float to resolve this (using
x = batch["image"].float()
), which I verified using a custom task. However I'm pretty sure this is not the intention to do thisSteps to reproduce
I note that the data is type int64:
Version
0.5.0
The text was updated successfully, but these errors were encountered: