[Task] Update Models API to update/remove deprecated blocks for Input, Retrieval and Output - Githubissues

NVIDIA-Merlin / models

Merlin Models is a collection of deep learning recommender system model reference implementations

https://nvidia-merlin.github.io/models/main/index.html

Apache License 2.0

262 stars 50 forks source link

[Task] Update Models API to update/remove deprecated blocks for Input, Retrieval and Output #1032

Open gabrielspmoreira opened 1 year ago

gabrielspmoreira commented 1 year ago

Description

Update Merling Models to remove update or deprecated blocks. The changes will cover updating:

API
Tests
Examples
Documentation

Additional context Here are the blocks that need to be updated or removed from the API

Input blocks

[ ] InputBlock -> InputBlockV2
[ ] EmbeddingFeatures -> Embedddings

Retrieval

[ ] RetrievalModel -> RetrievalModelV2
[ ] MatrixFactorizationModel -> MatrixFactorizationModelV2
[ ] TwoTowerModel -> TwoTowerModelV2
[ ] YoutubeDNNRetrievalModel -> YoutubeDNNRetrievalModelV2
[ ] InBatchSampler -> InBatchSamplerV2 Remove:
[ ] ItemRetrievalScorer
[ ] ItemRetrievalTask
[ ] NextItemPredictionTask

Output blocks

[ ] BinaryClassificationTask -> BinaryOutput
[ ] RegressionTask -> RegressionOutput
[ ] ParallelPredictionBlock -> OutputBlock

sararb commented 1 year ago

Thank you for the ticket @gabrielspmoreira. I just have a two additions to the list:

If we remove the ItemRetrievalScorer, we can also remove the PopularityBasedSampler: https://github.com/NVIDIA-Merlin/models/blob/5765c8ce2262f7f14fcb095f67b845235b8c16ac/merlin/models/tf/blocks/sampling/cross_batch.py#L26
We should also depreciate the logic of handling the PredictionTask instances in the base model class: ==> e.g. here