README.md

metadata

language:
  - code
  - en
license: mit
task_categories:
  - text-generation
pretty_name: RepoExec-Instruct
viewer: true

Table of Contents

• Dataset Summary
• Dataset Structure
  • Data Instances
  • Data Fields
  • Data Splits
• Usage
• Additional Information
• • Other Resources
  • Licensing Information
  • Citation Information
  • Contributions

Dataset Description

• Repository: FSoft-AI4Code/RepoExec
• Paper: RepoExec: Evaluate Code Generation with a Repository-Level Executable Benchmark
• Contact: [email protected]
• Website: https://www.fpt-aicenter.com/ai-residency/

RepoExec: Evaluate Code Generation with a Repository-Level Executable Benchmark

Dataset Summary

This source contains the instruction-tuning dataset to fine-tune models in our work.

Dataset Structure

Data Instances

{
    "id": 0,
    "prompt": "import base64\nimport random\nimport unicodedata\nimport zlib\nfrom typing import Union\nfrom uuid import uuid4\nfrom ._regex import *\nfrom .errors import InvalidInputError\nfrom .validation import is_snake_case, is_full_string, is_camel_case, is_integer, is_string\n\nclass InvalidInputError(TypeError):\n    \"\"\"\n    Custom error raised when received object is not a string as expected.\n    \"\"\"\n\n    def __init__(self, input_data: Any):\n        \"\"\"\n        :param input_data: Any received object\n        \"\"\"\n        type_name = type(input_data).__name__\n        msg = 'Expected \"str\", received \"{}\"'.format(type_name)\n        super().__init__(msg)\n\ndef is_string(obj: Any) -> bool:\n    \"\"\"\n    Checks if an object is a string.\n\n    *Example:*\n\n    >>> is_string('foo') # returns true\n    >>> is_string(b'foo') # returns false\n\n    :param obj: Object to test.\n    :return: True if string, false otherwise.\n    \"\"\"\n    return isinstance(obj, str)\n\ndef reverse(input_string: str) -> str:\n    \"\"\"\n    Returns the string with its chars reversed.\n\n    *Example:*\n\n    >>> reverse('hello') # returns 'olleh'\n\n    :param input_string: String to revert.\n    :type input_string: str\n    :return: Reversed string.\n    \"\"\"\n",
    "docstring": 
}

Data Fields

Data fields for inline level:

• id (string): the unique id
• prompt (string): sequence to fine-tune LM
• docstring (string): docstring of the target function. If docstring is not None, instruction template is applied; otherwise raw format or small context is applied.

Data Splits

The instruction tuning dataset is not split and only contains data subset.

Usage

You can load this dataset using datasets library: pip install datasets

from datasets import load_dataset

# Load full dataset
dataset = load_dataset("Fsoft-AIC/RepoExec-Instruct")

Additional Information

Other Resources:

• Github: https://github.com/FSoft-AI4Code/RepoExec
• Webpage: https://fsoft-ai4code.github.io/repoexec
• Leaderboard: https://repoexec.github.io
• Paper: https://arxiv.org/html/2406.11927v1

Licensing Information

MIT License

Citation Information

@article{nam2024repoexec,
  title={RepoExec: Evaluate Code Generation with a Repository-Level Executable Benchmark},
  author={Hai, Nam Le and Manh, Dung Nguyen and Bui, Nghi DQ},
  journal={arXiv preprint arXiv:2406.11927v1},
  year={2024}
}

Contributions

This dataset is developed by FSOFT AI4Code team.