blastwind/random_code_snippets

--- dataset_info: features: - name: lang dtype: string - name: seed dtype: string splits: - name: train num_bytes: 3114466 num_examples: 10000 download_size: 1629429 dataset_size: 3114466 configs: - config_name: default data_files: - split: train path: data/train-* --- This dataset contains 10000 random snippets of 5-15 lines parsed from [`bigcode/starcoderdata`](https://huggingface.co/datasets/bigcode/starcoderdata). Specifically, I consider 10 languages: Haskell, Python, cpp, java, typescript, shell, csharp, rust, php, and swift. And, I collect 1000 documents for each language, and then extract 5-15 random lines from the document to create this dataset. See MagiCoder and their [seed collection](https://github.com/ise-uiuc/magicoder/blob/main/experiments/collect_seed_documents.py#L35) process. In my usecase, I needed some inspiration documents for generating synthetic datasets.

原始信息汇总

数据集概述

数据集特征

• lang：数据类型为字符串
• seed：数据类型为字符串

数据集划分

• 训练集（train）：
  • 示例数量：10000
  • 数据大小：3114466字节

数据集大小

• 下载大小：1629429字节
• 数据集总大小：3114466字节

配置信息

• 配置名称：default
• 数据文件：
  • 划分：训练集
  • 路径：data/train-*