hkristof03

How to match embedding dimensions for different cardinalities of user and item sets?

1

Thanks for the developers and for the community for the library and the excellent questions and answers. [According to this source, ](https://developers.googleblog.com/2017/11/introducing-tensorflow-feature-columns.html?fbclid=IwAR23SR4lbfmyllJ2c6VWT77kclmhBirYwlExCYw93NYqnac0VR5ia_Cu8AE) it is usually a good rule of thumb...

Random datetimes extracted from nowhere

1

Hi! I found your this library on stackoverflow: https://stackoverflow.com/questions/53759224/how-to-write-a-regex-to-validate-date-format-of-type-day-month-dd-yyyy Although it successfully extracted the timestamps that appear in the following log, it produces artificial ones as well: s = 'Jan...

Totally bad data

4

I just came across this library and performed some initial queries but what I see is a bit far from reality. ``` aapl = yf.Ticker("AAPL") bs = aapl.balance_sheet print(f"{bs.transpose()['Common Stock'][::-1][-1]:,}")...

Ambiguous DCN layer

3

I would like to ask for some help regarding stacked Cross layers. From [this tutorial](https://www.tensorflow.org/recommenders/examples/dcn), the calling of a simple DCN layer looks like this: ``` if use_cross_layer: self._cross_layer =...

[Question] How to use Attention to aggregate history?

2

@OmarMAmin mentioned [Exploring Heterogeneous Metadata for Video Recommendation with Two-tower Model](https://assets.amazon.science/1e/e6/4d7f8a2741a4a3b148e20a953946/exploring-heterogeneous-metadata-for-video-recommendation-with-two-tower-model.pdf) paper in [this discussion](https://github.com/tensorflow/recommenders/issues/633). I read through the paper multiple times and a few resources about existing Attention mechanisms,...

Sample weights for both Retrieval and Ranking

3

I'd like to ask about using the `'sample_weight'` argument for both the Retrieval and the Ranking Tasks' call methods. [Retrieval docs](https://www.tensorflow.org/recommenders/api_docs/python/tfrs/tasks/Retrieval) [Ranking docs](https://www.tensorflow.org/recommenders/api_docs/python/tfrs/tasks/Ranking) 1. For both models, does it make...

[QST] How to get candidate features after using pre-trained embedding tables?

3

# ❓ Questions & Help ## Details I am following the [tutorial here](https://github.com/NVIDIA-Merlin/models/blob/main/examples/usecases/entertainment-with-pretrained-embeddings.ipynb)to include pre-computed embeddings when I train a Two Tower Retrieval model. Specifically, I am using this method...

[BUG] ModelCheckpoint Callback results in an error due to unsupported overwrite=True parameter

7

### Bug description The[ callback ModelCheckpoint calls](https://github.com/tensorflow/tensorflow/blob/d0228dc874ed4f19f750e5df7e8ac61827fa7332/tensorflow/python/keras/callbacks.py#L1434) the model's save method, which is overwritten by the [Merlin BaseModel implementation](https://github.com/NVIDIA-Merlin/models/blob/eb1e54196a64a70950b2a7e7744d2150e052d53e/merlin/models/tf/models/base.py#L1687). The callback passes the `overwrite=True` parameter to the save method, which...

**Describe the bug** The default method that computes the embedding dimensions for categorical features based on their cardinalities often leads to embedding collapse, when the subsequent Two Tower model recommends...

### Bug description The default embedding dimensions for categorical features computed by NVTabular often causes embedding collapse. I detailed this issue [here](https://github.com/NVIDIA-Merlin/NVTabular/issues/1890). When you modify the embedding dimensions directly in...