Dataset · code-search-net
2M comment-code pairs
code-search-net/code_search_net
Documentation and code pairs from GitHub for natural language code search and code model training.
What's inside
The CodeSearchNet corpus contains approximately 2 million (comment, code) pairs from open-source GitHub libraries. It features code and documentation in multiple programming languages. The dataset was gathered for the CodeSearchNet challenge, focusing on code retrieval via natural language descriptions.
What you can do with it
- Training and evaluating code retrieval by description
- Fine-tuning code generation models
- Training fill-mask on code
- Tasks
- Text generation, Fill mask
- Languages
- Code
- Size
- 1M–10M rows
- Downloads this month
- 39.6K
- Likes
- 354
Models trained on it
- mradermacher/Bor-1B-i1-GGUF 663
- mradermacher/Bor-1B-v1-i1-GGUF 581
- divakantpandey/code-embed-v2Embeddings 548
- Shuu12121/CodeSearch-ModernBERT-Crow-PlusSentence similarity 543
- mradermacher/Bor-1B-v1-GGUF 310
- ksg91/toam-100m-v0.1Text generation 266
- jaweed123/Qwen3.5-0.8B-Python-SFTText generation 213
- mradermacher/Bor-1B-GGUF 208
How to load it
from datasets import load_dataset
ds = load_dataset("code-search-net/code_search_net")