PyTorch: trying to create a joint dataset with different transforms results in both datasets having same transform
Summary The core issue stems from a subtle naming conflict and shared mutable state in the user’s PyTorch data pipeline. Specifically, the variable names train_set and val_set are first assigned to the unmasked dataset subsets. However, later in the script, the variables train_set and val_set are accidentally reused when creating the masked dataset subsets. This … Read more