Login
- Dokumenttyp:
- Forschungsdaten
- Verantwortlich:
- Zügner, Daniel
- Autorinnen / Autoren:
- Zügner, Daniel (1); Kirschstein, Tobias (1); Catasta, Michele (2); Leskovec, Jure (2); Günnemann, Stephan (1)
- Institutionszugehörigkeit:
- 1. Technical University of Munich (TUM)
2. Computer Science Department, Stanford University - Herausgeber:
- TUM
- Titel:
- Code Transformer: Pretrained Models and Preprocessed Data
- Enddatum der Datenerzeugung:
- 31.03.2021
- Fachgebiet:
- DAT Datenverarbeitung, Informatik
- Quellen der Daten:
- Experimente und Beobachtungen / experiments and observations
- Andere Quellen der Daten:
- Pre-trained models
- Datentyp:
- mehrdimensionale Visualisierungen oder Modelle / models
- Methode der Datenerhebung:
- This repository contains the preprocessed data and pretrained model from the ICLR 2021 paper Language-Agnostic Representation Learning of Source Code from Structure and Context. The preprocessed data is from the CodeSearchNet Challenge ( https://github.com/github/CodeSearchNet/) as well as the code2seq paper (Java-small, Java-medium, Java-large; https://github.com/tech-srl/code2seq/). We first use GitHub semantic to construct abstract syntax trees (ASTs) for each code snippet. We then compute d... »
- Beschreibung:
- This repository contains pretrained models as well as preprocessed data from the ICLR 2021 paper Language-Agnostic Representation Learning of Source Code from Structure and Context. See https://www.in.tum.de/daml/code-transformer/ for additional information such as the code, paper, and poster. The preprocessed data is based on the CodeSearchNet Challenge ( https://github.com/github/CodeSearchNet/) as well as the code2seq paper (Java-small, Java-medium, Java-large; https://github.com/tech-srl/code2seq/) datasets.
See the original datasets for their licenses and terms of use. - Links:
Additional Information: https://openreview.net/forum?id=Xh5eMZVONGF
- Schlagworte:
- machine learning; deep learning; transformer models; source code; representation learning
- Technische Hinweise:
- View and download (40 GB total, 15 Files)
The data server also offers downloads with FTP
The data server also offers downloads with rsync (password m1647000):
rsync rsync:// m1647000@dataserv.ub.tum.de/ m1647000/ - Sprache:
- en
- Rechte:
- by, http://creativecommons.org/licenses/by/4.0
BibTeX