datasets:
- JoaoJunior/python_java_dataset_APR
tags:
- APR
- AI
Introduction
This model, JoaoJunior/T5_APR_java_python_v4, is a fine-tuned version of the pre-trained CodeT5 model from Salesforce. The model is designed to understand and generate code, with a specific focus on bug fixing tasks in Python and Java languages.
Description
The CodeT5 model was introduced in the paper "CodeT5: Identifier-aware Unified Pre-trained Encoder-Decoder Models for Code Understanding and Generation". This model leverages the semantics conveyed from the developer-assigned identifiers in the code, allowing for effective code understanding and generation tasks.
JoaoJunior/T5_APR_java_python_v4 was trained on the python_java_dataset_APR dataset, which contains pairs of bugged and fixed code in Python and Java. This dataset was created using the coconut_java2006 and coconut_python2010 datasets from the CoCoNuT project as its base.
Objective
The primary objective of this model is to identify and fix bugs in Python and Java code. By fine-tuning the CodeT5 model on the python_java_dataset_APR dataset, this model aims to effectively learn the patterns and structures of these languages, enabling it to accurately detect and correct errors.
References
- CodeT5: Identifier-aware Unified Pre-trained Encoder-Decoder Models for Code Understanding and Generation by Yue Wang, Weishi Wang, Shafiq Joty, Steven C.H. Hoi
- python_java_dataset_APR: A dataset containing pairs of bugged and fixed code in Python and Java, created using the CoCoNuT project's coconut_java2006 and coconut_python2010 datasets
- CoCoNuT: Combining Context-Aware Neural Translation Models using Ensemble for Program Repair