Marc Romeyn
Marc Romeyn
## Problem: There is a considerable amount of tech-debt in the code that’s responsible for negative sampling in retrieval. The biggest issue is that it relies heavily on the model-context,...
Currently blocked by #52.
Port the [TabularLayerNorm ](https://github.com/NVIDIA-Merlin/Transformers4Rec/blob/538fc54bb8f2e3dc79224e497bebee15b00e4ab7/transformers4rec/torch/tabular/transformations.py#L90)from PyTorch to TF
Currently blocked by #51.
This ticket includes the 3 basic aggregators we support for non-sequential data: - [ ] ConcatFeatures - [ ] StackFeatures - [ ] ElementwiseSum
Fixes # (issue) ### Goals :soccer: ### Implementation Details :construction: ### Testing Details :mag:
### Goals :soccer: To prepare for migrating to our new dataloaders package, this PR renames `BatchedDataset` to `Loader` since that will be the new name. This PR also exposes the...
Fixes # (issue) ### Goals :soccer: ### Implementation Details :construction: ### Testing Details :mag:
We would like to add capabilities to a `ParallelBlock` to select sub-graphs. This can be useful for instance to select the `item-id` embedding-table from a `InputBlock`. Another instance where this...
There’s a need to introduce a new concept: `Encoder` which is a block that encodes features into some representation. This class can be used for prediction but can’t be trained/evaluated...