Added DLRM notebook and generated python script

[kharshith-k]

Added DLRM.ipynb and generated dlrm.py

[abheesht17]

Thanks for the PR! Left some comments, let me know what you think :).

Here are some notes for posterity (we don't need to implement these now in this PR):

• We should, in the future, also think about the specific case of how how DLRM implements distributed training. The embeddings are sharded across devices, whereas the MLP layers are DDP'd.
• The paper mentions that they use torch.nn.EmbeddingBag instead of torch.nn.Embedding. Apparently, torch.nn.EmbeddingBag is more efficient than torch.nn.Embedding followed by some aggregation op. We should think about this at some later point (maybe, keras_rs.embeddings.DistributedEmbedding solves this). Anyway, we can worry about this later!

[abheesht17]

@kharshith-k - let me know when this is ready for another round of review. Thanks!

[kharshith-k]

@kharshith-k - let me know when this is ready for another round of review. Thanks!
@abheesht17 - The changes are ready for review. Please let me know if any more changes would be required. Thanks!

[abheesht17]

Thanks! Great work, this looks good to me, overall. Just one major comment on using dense features (it isn't DLRM without the two blocks for dense and categorical features)

[kharshith-k]

Hi @abheesht17, The changes are ready for review. I've used raw_user_age and timestamp for continuous features. Please let me know if any more changes would be required. Thanks!