The Merlin data loader has a Padding operator that we can pass via the transforms parameter. To avoid code redundancy, I created a PR in T4Rec that removes custom padding from the T4Rec's MerlinDataLoader class.
In this PR, the default behavior of the Trainer is to set the data loader without padding. If the user wants to perform padding outside of the model, they can explicitly provide a data loader with a Padding transform.
Implementation Details :construction:
Removes the sparse_max logic from MerlinDataLoader
Extend MerlinDataLoader to accept the transforms parameter where the user can pass the Padding op.
Testing Details :mag:
Add a unit test to check that we can provide a data loader with Padding op to the Trainer class
Goals :soccer:
The Merlin data loader has a Padding operator that we can pass via the transforms parameter. To avoid code redundancy, I created a PR in T4Rec that removes custom padding from the T4Rec's
MerlinDataLoader
class. In this PR, the default behavior of the Trainer is to set the data loader without padding. If the user wants to perform padding outside of the model, they can explicitly provide a data loader with a Padding transform.Implementation Details :construction:
sparse_max
logic fromMerlinDataLoader
MerlinDataLoader
to accept thetransforms
parameter where the user can pass the Padding op.Testing Details :mag: