skip to main content
Nov 2018 software

Implementation, data and a trained model for the code2vec paper

Description

This artifact contains data, trained models and code for the paper "code2vec: Learning Distributed Representations of Code". Updated instructions may be found in: https://github.com/tech-srl/code2vec


Assets

Read Me (popl19main-p32-p-3290353-artifact-accepted-readme.txt)
Original Artifact (popl19main-p32-p-3290353-artifact-accepted-archive.zip)

Instructions

General Installation

Software Dependencies:

TensorFlow, Python 3.5

Hardware Dependencies:

A GPU is recommended only for training a model from scratch. Using our provided trained model does not require a GPU.

General Installation:

See README.

Other Instructions:

The data for training a model from scratch is under the directory: data/java14m/ The trained model is under: models/java14_model/ For more information, see README.

Experimental Installation

Installation:

See README.

Parameterization:

See README.

Work Flows:

See README.

Evaluation:

See README.


Data Documentation

A dataset of about 13M Java methods with their names, taken from top-starred GitHub repositories; A code2vec model that was trained on this dataset; code from https://github.com/tech-srl/code2vec


Provenance

https://github.com/tech-srl/code2vec


Comments