Skip to content

tud-ccc/learning-compiler-graphs

Repository files navigation

Requirements

This artifact is designed to be pretty self-contained and produces all its requirements. In principle it only needs a working docker installation. See https://docs.docker.com/install/

Quick Result Overview

It is possible to see the Jupyter notebook without actually running anything. For this, you need a working Jupyter installation (see https://jupyter.org/install).


Note that the docker container builds Jupyter, so you don't need a local installation if you run the standard method.


For this, just execute:

jupyter notebook artifact.ipynb

Running the artifact

This artifact is prepared with a docker container, which will fetch all dependencies, build and prepare the artifact such that it can be executed from the Jupyter notebook artifact.ipynb. Make sure you have docker installed and execute the runscript to build and run everything:

./run.sh

After fetching and compiling everything (warning: this will take several hours), you should see the Jupyter notebook running. It will be forwarded to port 8888. You should have a link with a token in the terminal output, like:

https://127.0.0.1:8888/notebooks/artifact.ipynb?token=<some-long-hash-here>

This link should work on your local browser to see the Jupyter notebook.

Re-training the models

Since training time would take several weeks on a commodity desktop machine, we've included trained models to reproduce our results. If you want to re-train, just run rm -fr results to remove the trained models.