understanding-search / maze-transformer

This repo is built to facilitate the training and analysis of autoregressive transformers on maze-solving tasks.
24 stars 6 forks source link

Checks Coverage GitHub commit activity GitHub closed pull requests GitHub closed issues code size, bytes

Maze Transformer

This repo is built to facilitate the training and analysis of autoregressive transformers on maze-solving tasks.

Example of a baseline solver navigating through the maze Plotting the generated path and attention over the maze of a (poorly trained) model An example maze generated via randomized depth-first search and percolation

Visualization of attention over the input sequence of tokens

Installation

pip install git+ssh://git@github.com/understanding-search/maze-transformer.git

Note: if you want to install the library in colab, follow the steps in this Colab notebook.

You can run all tests via

make test

(this will take a few minutes)

Usage

Most of the functionality is demonstrated in the ipython notebooks in the notebooks/ folder.

Development

Prerequisites

Testing & Static analysis

make help will print all available commands.