maxjcohen/transformer
A PyTorch implementation of Transformer neural networks adapted for time series forecasting and building consumption prediction.
Not currently ranked — collecting fresh signals.
star history
This repository implements the Transformer architecture (originally from NLP) applied to time series prediction. It adapts the model by replacing embeddings with linear layers, using modified positional encodings that better match day/night patterns, and applying windowed attention to focus on short-term patterns. The use case is modeling a numerical simulator for building energy consumption prediction.
Frequently asked
- What is maxjcohen/transformer?
- A PyTorch implementation of Transformer neural networks adapted for time series forecasting and building consumption prediction.
- Is transformer open source?
- Yes — maxjcohen/transformer is open source, released under the GPL-3.0 license.
- What language is transformer written in?
- maxjcohen/transformer is primarily written in Jupyter Notebook.
- How popular is transformer?
- maxjcohen/transformer has 905 stars on GitHub.
- Where can I find transformer?
- maxjcohen/transformer is on GitHub at https://github.com/maxjcohen/transformer.