- Author: Lina Faik
- Creation date: June 2023
- Last update: June 2023
This repository contains the code and the notebook used to train a DLRM model using PyTorch library, torchrec. It was developed as an experimentation project to support the explanation blog posts around the topic.
The model and code are explained in more detail in the following article:
You can find all my technical blog posts here.
notebook.ipynb # central code where a DLRM model is trained and evaluated using synthetical data
src
├── batch.py # general functions used to build batch from raw data
├── model.py # model related class and functions
The notebook is based on synthetic data. It can easily be adapted to any other data set with categorical and continuous variables and a binary target to predict.
The code uses a GPU and relies on the following libraries:
torch
torchrec==0.4.0
plotly==5.13.1
sklearn-pandas==2.2.0
To run experiments, you need to run the notebook notebook.ipynb.
The associated code is in the src directory.