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
this is just a tiny issue that I came across while adding a custom dataset, that's not in ImageFolder format.
In solo/methods/base.py L553 the validation_step says the following:
"""Validation step for pytorch lightning. It does all the shared operations, such as
forwarding a batch of images, computing logits and computing metrics.
Args:
batch (List[torch.Tensor]):a batch of data in the format of [img_indexes, X, Y].
batch_idx (int): index of the batch.
Returns:
Dict[str, Any]: dict with the batch_size (used for averaging), the classification loss
and accuracies.
"""
However the batch is only split to X, Y according to L568: X, targets = batch
As I said, it's a tiny thing, but not sure why or where does the library modify the dataset/dataloader to have a different return value depending on train/val split.
Sorry for the rookie question.
Thanks,
Csabi
The text was updated successfully, but these errors were encountered:
Hey! I think we just messed up the docstring and forgot to update it a long time ago. I would say that for sure we have other inconsistencies in the documentation that we plan on progressively updating. I'll try to open a small PR to fix this one.
Hi,
this is just a tiny issue that I came across while adding a custom dataset, that's not in ImageFolder format.
In
solo/methods/base.py L553
thevalidation_step
says the following:However the batch is only split to X, Y according to L568:
X, targets = batch
As I said, it's a tiny thing, but not sure why or where does the library modify the dataset/dataloader to have a different return value depending on train/val split.
Sorry for the rookie question.
Thanks,
Csabi
The text was updated successfully, but these errors were encountered: