models

models copied to clipboard

[BUG] The evaluation metrics of YouTubeDNN model is worse after the refactory of model.compile()

2

### Bug description After the refactory moving loss and metrics to model.compile(), the loss and eval metrics are worse for a YouTubeDNN retrieval models. For the LastFM dataset for example...

gabrielspmoreira

[FEA] Export embeddings from youtube-dnn

1

# 🚀 Feature request Make it possible to export query-/item-embeddings from the youtube-dnn retrieval model similar to MF & Two-Tower. Right now the `YoutubeDNNRetrievalModel` returns a `Model` object instead of...

marcromeyn

### Description Implement a mapping function to apply on top of the data loader in order to extract targets from the input sequence. The goal is to define different strategies...

sararb

[RMP] Expand supported datasets in Merlin Models (to support the tutorial?)

5

## Problem: A clear description of the problem and how it impacts the customer experience of Merlin users. Why is this important? Why/Should we prioritize this work? ## Goal: -...

gabrielspmoreira

[BUG] get_aliccp() failed while processing Ali CCP dataset

4

### Bug description I'm using ```get_aliccp()``` to process Ali CCP dataset following the [multi-stage recsys example](https://github.com/NVIDIA-Merlin/Merlin/tree/main/examples/Building-and-deploying-multi-stage-RecSys) but met a problem: ``` Reading common features...: 730600it [09:54, 1229.12it/s] Processing data...: 100000it...

future-xy

### Description Providing an example using the dataset from RecSys2022 challenge and teaching concepts of Merlin Models. Concepts: - NextItem Prediction - Sequential input - ...

bschifferer

## Problem: Customers need to be able to train models that exceed the memory of a single GPU. Recommender Systems models can grow to over hundreds of millions of users...

benfred

Add preprocessing and schema for H&M dataset

4

The preprocessing workflow we have been used for retrieval models research is available [here](https://github.com/NVIDIA-Merlin/research/tree/main/retrieval_exp/notebooks/preproc/handm)

benfred

Currently we calculate the metrics for both the items in the dataset as well as the sampled negatives. We shouldn’t calculate it for the sampled negatives.

marcromeyn

[TASK] Masking shouldn't be used for test data

2

Masking class: sequences are shifted to keep the last item as a label during evaluation (Training=False) but we should disable masking and keep the sequence as it is for the...

sararb