README.md

---
language:
- en
multilinguality:
- monolingual
size_categories:
- 1M<n<10M
task_categories:
- feature-extraction
- sentence-similarity
pretty_name: CodeSearchNet
tags:
- sentence-transformers
dataset_info:
  config_name: pair
  features:
  - name: comment
    dtype: string
  - name: code
    dtype: string
  splits:
  - name: train
    num_bytes: 1016891681
    num_examples: 1375067
  download_size: 491861063
  dataset_size: 1016891681
configs:
- config_name: pair
  data_files:
  - split: train
    path: pair/train-*
---

# Dataset Card for CodeSearchNet

This dataset is a collection of comment-code pairs of various programming languages. See [code_search_net](https://huggingface.co/datasets/code_search_net) for additional information.
This dataset can be used directly with Sentence Transformers to train embedding models.

## Dataset Subsets

### `pair` subset

* Columns: "comment", "code"
* Column types: `str`, `str`
* Examples:
    ```python
    {
      'comment': 'Computes the new parent id for the node being moved.\n\n@return int',
      'code': "protected function parentId()\n\t{\n\t\tswitch ( $this->position )\n\t\t{\n\t\t\tcase 'root':\n\t\t\t\treturn null;\n\n\t\t\tcase 'child':\n\t\t\t\treturn $this->target->getKey();\n\n\t\t\tdefault:\n\t\t\t\treturn $this->target->getParentId();\n\t\t}\n\t}",
    }
    ```
* Collection strategy: Reading the Code Search Net dataset from [embedding-training-data](https://huggingface.co/datasets/sentence-transformers/embedding-training-data).
* Deduplified: No