A light-weight Transformer model for Kaggle House Prices Regression Competition

A simple Pytorch deep learning model for predicting the house price. Lightweight Transformer model is tested for accuracy. The Transformer architecture is utilized to capture pair-wise affinity of all the features.

This is my first project in PyTorch. The aim of the project is to perform a simple multivariate regression using Transformer model.

Get packages by using conda or pip.

1. PyTorch=1.8.0
2. numpy=1.19.2
3. matplotlib=3.3.4
4. pandas=1.1.5
5. seaborn=0.11.2

Once finished, you can upload your prediction.csv to the kaggle website where you can compare your score with other users.

You need to have more than 4GB GPU memory to train the model with default settings, or you need to change batchsize or the network sturctures.