infinityofspace/python_codestyles-random-1k

---
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
- split: test
path: data/test-*
dataset_info:
features:
- name: code
dtype: string
- name: code_codestyle
dtype: int64
- name: style_context
dtype: string
- name: style_context_codestyle
dtype: int64
- name: label
dtype: int64
splits:
- name: train
num_bytes: 3604934957
num_examples: 308000
- name: test
num_bytes: 645620388
num_examples: 56400
download_size: 671035436
dataset_size: 4250555345
license: mit
tags:
- python
- code-style
- random
size_categories:
- 100K<n<1M
---
# Dataset Card for "python_codestyles-random-1k"

This dataset contains negative and positive examples with python code of compliance with a code style. A positive
example represents compliance with the code style (label is 1). Each example is composed of two components, the first
component consists of a code that either conforms to the code style or violates it and the second component
corresponding to an example code that already conforms to a code style. In total, the dataset contains `1.000` completely
different code styles. The code styles differ in at least one codestyle rule, which is called a `random` codestyle
dataset variant. The dataset consists of a training and test group, with none of the code styles overlapping between
groups. In addition, both groups contain completely different underlying codes.

The examples contain source code from the following repositories:

| repository | tag or commit |
|:-----------------------------------------------------------------------:|:----------------------------------------:|
| [TheAlgorithms/Python](https://github.com/TheAlgorithms/Python) | f614ed72170011d2d439f7901e1c8daa7deac8c4 |
| [huggingface/transformers](https://github.com/huggingface/transformers) | v4.31.0 |
| [huggingface/datasets](https://github.com/huggingface/datasets) | 2.13.1 |
| [huggingface/diffusers](https://github.com/huggingface/diffusers) | v0.18.2 |
| [huggingface/accelerate](https://github.com/huggingface/accelerate) | v0.21.0 |

You can find the corresponding code styles of the examples in the file [additional_data.json](additional_data.json).
The code styles in the file are split by training and test group and the index corresponds to the class for the
columns `code_codestyle` and `style_context_codestyle` in the dataset.

There are 364.400 samples in total and 182.200 positive and 182.200 negative samples.

原始信息汇总

数据集概述

数据集配置

• 默认配置：
  • 训练集：路径为 data/train-*
  • 测试集：路径为 data/test-*

数据集信息

• 特征：

  • code：类型为 string
  • code_codestyle：类型为 int64
  • style_context：类型为 string
  • style_context_codestyle：类型为 int64
  • label：类型为 int64

• 数据分割：

  • 训练集：
    • 字节数：3604934957
    • 样本数：308000
  • 测试集：
    • 字节数：645620388
    • 样本数：56400

• 数据集大小：

  • 下载大小：671035436
  • 实际大小：4250555345

• 许可证：MIT

• 标签：

  • python
  • code-style
  • random

• 大小类别：

  • 100K<n<1M

数据集描述

• 该数据集包含符合和违反代码风格的Python代码的正负样本。
• 正样本表示符合代码风格（标签为1）。
• 每个样本由两部分组成：
  • 第一部分是符合或违反代码风格的代码。
  • 第二部分是已经符合代码风格的示例代码。
• 数据集包含1000种完全不同的代码风格。
• 代码风格至少在一个代码风格规则上有所不同，称为random代码风格数据集变体。
• 数据集分为训练集和测试集，两组之间没有代码风格重叠。
• 训练集和测试集包含完全不同的底层代码。
• 数据集总共包含364400个样本，其中182200个正样本和182200个负样本。