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
You're correct that TransformsWrapper should be set up to apply transformations for training. However, this code specified only for testing. I am using the MNIST dataset which includes separate train and test sets, and then I concatenate these sets for splitting them into new train, validation, and test sets, which required for testing of different code parts (see more in tests/). So, in this context, the exact mode of TransformsWrapper is not critical in MNISTDataModule class, as the transformations applied to the initial MNIST sets are the same (see more in configs/datamodule/mnist.yaml) due to further splitting.
I realize that this might not be clear from the code. To improve clarity, I added a comment explaining why the same transformations are used, and changed the mode name to "train". I hope it will help avoid any misunderstanding.
The main logic of DataModule in this pipeline could be seen in src/datamodules/datamodules.py. MNISTDataModule is used just for testing.
Thank you for your contribution to improving the quality and clarity of this project. If you have any further questions or suggestions, feel free to share!
yet-another-lightning-hydra-template/src/datamodules/mnist_datamodule.py
Line 34 in 936e99f
The text was updated successfully, but these errors were encountered: