---
dataset_info:
  features:
  - name: prompt
    dtype: string
  - name: reference_code
    dtype: string
  - name: code_context
    dtype: string
  - name: problem_id
    dtype: int64
  - name: library_problem_id
    dtype: int64
  - name: library
    dtype:
      class_label:
        names:
          '0': Matplotlib
          '1': Numpy
          '2': Pandas
          '3': Pytorch
          '4': Scipy
          '5': Sklearn
          '6': Tensorflow
  - name: test_case_cnt
    dtype: int64
  - name: perturbation_type
    dtype:
      class_label:
        names:
          '0': Difficult-Rewrite
          '1': Origin
          '2': Semantic
          '3': Surface
  - name: perturbation_origin_id
    dtype: int64
  splits:
  - name: test
    num_bytes: 3136179
    num_examples: 1000
  download_size: 712717
  dataset_size: 3136179
configs:
- config_name: default
  data_files:
  - split: test
    path: data/test-*
license: cc-by-sa-4.0
language:
- code
task_categories:
- text2text-generation
tags:
- code-generation
arxiv: 2211.11501
---

This is a reupload of [DS-1000](https://huggingface.co/datasets/xlangai/DS-1000). The metadata dictionary has been extracted into columns and the categorical variables are now `ClassLabel` types, and the dataset is natively a parquet. The features are as follows:

|        Column        |                                                   Type                                                    |
|----------------------|-----------------------------------------------------------------------------------------------------------|
|problem_id            |`Value(dtype='int64', id=None)`                                                                            |
|prompt                |`Value(dtype='string', id=None)`                                                                           |
|reference_code        |`Value(dtype='string', id=None)`                                                                           |
|code_context          |`Value(dtype='string', id=None)`                                                                           |
|library_problem_id    |`Value(dtype='int64', id=None)`                                                                            |
|library               |`ClassLabel(names=['Matplotlib', 'Numpy', 'Pandas', 'Pytorch', 'Scipy', 'Sklearn', 'Tensorflow'], id=None)`|
|test_case_cnt         |`Value(dtype='int64', id=None)`                                                                            |
|perturbation_type     |`ClassLabel(names=['Difficult-Rewrite', 'Origin', 'Semantic', 'Surface'], id=None)`                        |
|perturbation_origin_id|`Value(dtype='int64', id=None)`                                                                            |

All credits go to the original authors below.

---

<h1 align="center"> DS-1000 in simplified format </h1>

See testing code and more information in the [DS-1000 repo](https://github.com/xlang-ai/DS-1000/).

Reformatting credits: Yuhang Lai, Sida Wang