models

models copied to clipboard

### Description Add an example of Wide&Deep ranking model in the [03-Exploring-different-models.ipynb](https://github.com/NVIDIA-Merlin/models/blob/main/examples/03-Exploring-different-models.ipynb) notebook.

gabrielspmoreira

XGBoost - Use DaskDMatrix for evals data to ensure metrics in logs match result of evaluate

9

### Goals :soccer: Ensure metrics in logs match the result of the `.evaluate` method. Example in this gist of the metrics difference: https://gist.github.com/oliverholworthy/db8dfca6e08d29fa6481f5f9d66c8049 Issue reported by @rnyak ### Implementation Details...

oliverholworthy

Update In Batch Negative Sampling Augmentation Layer for primary use with dataloader

6

Related to #596 and #626 ### Goals :soccer: - Improve speed of in-batch negative sampling when used as part of the dataloader - Simplify use of in-batch negative sampling when...

oliverholworthy

[Task] InputBlockV2 - Usage in ranking models high-level API

1

Change the following higher-level API ranking models to use the new `InputBlockV2`. This is part of RMP #479, which you can check for more context - [ ] `DCNModel` -...

gabrielspmoreira

Return correct type from data augmentation layer depending on the context

3

Related to https://github.com/NVIDIA-Merlin/models/issues/596 and https://github.com/NVIDIA-Merlin/models/issues/626 ### Goals :soccer: Enable the negative sampling (data augmentation) layer to be used both in the dataset (BatchedDataset.map) and as part of the model without...

oliverholworthy

Rename UniformNegativeSampling class

7

Addresses https://github.com/NVIDIA-Merlin/models/issues/650

rnyak

[BUG] Batch Size results in different prediction outputs.

1

### Bug description Predictions from different size data loader yields different results. Reproducing: 1. simplistic two tower model ``` model = mm.TwoTowerModel( schema, query_tower=mm.MLPBlock([64,64], no_activation_last_layer=True, ), item_tower=mm.MLPBlock([64,64], no_activation_last_layer=True, )) model.compile(...

angmc

Check for `Prediction` namedtuple in train/test steps

1

Based on by comments in #685 ### Goals :soccer: Check for `Prediction` namedtuple in train/test steps when unpacking data. ### Implementation Details :construction: Wrapping `unpack_x_y_sample_weight` to handle `Prediction` tuple as...

oliverholworthy

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...

marcromeyn

marcromeyn