Published January 23, 2023 | Version 1.0
Conference paper Open

PYEVOLVE: Automating Frequent Code Changes in Python ML Systems

Authors/Creators

  • 1. University of Colorado Boulder

Description

This is to make the artifacts described in the paper 

We release one tool
* [PyEvolve](https://github.com/maldil/PyEvolve) - a tool that automatically transplants code change patterns to Python software systems. The tool is available in the "./Tool" folder. 

PyEvolve's dependency RulePhaser is available in Tool/RulePhaser

 

We made two types of large datasets available.

* More than 40,000 code transformation trials were used to evaluate PyEvolve. This was achieved using a dataset of change patterns collected from real repositories. The dataset is publically available. This is available in the file "./Data/CrossValidationICSE2023.xlsx" 
* PyEvolve atomatically transplanted code change patterns by submitting 40 pull requests to open source repositories. We made the list of the pull requests public. This is available in the file "./Data/Pull_Requests.xlsx" 

 

Instructions that need to execute PyEvolve in VirtualBox image is available in ICSE2023_PyEvolve_Artifacts

Files

PyEvolve.zip

Files (59.0 MB)

Name Size
md5:e700e86637554abdd9b239fda0b87f72
59.0 MB Preview Download