Skip to content

Dataset · google

Code and documentation for description generation

google/code_x_glue_ct_code_to_text

Code-to-text documentation pairs from CodeSearchNet for training description generation.

What's inside

The CodeXGLUE code-to-text dataset contains pairs of code snippets and text descriptions (documentation). It is gathered from CodeSearchNet with filtering: examples with unparsable ASTs, descriptions shorter than 3 or longer than 256 tokens, and special tokens are removed. The languages are code and English, with a size ranging from 1 to 10 million examples. The format is not specified in the card.

What you can do with it

  • Fine-tune a model to generate docstrings from code
  • Evaluate code-to-text on the CodeXGLUE benchmark
  • Train an encoder-decoder for function descriptions
Tasks
Translation
Languages
Code, English
Size
1M–10M rows
Downloads this month
5.2K
Likes
80

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("google/code_x_glue_ct_code_to_text")