Skip to content

Dataset · bigcode

The Stack v2 code dataset with IDs for training

bigcode/the-stack-v2-train-full-ids

Cleaned code across 600+ programming languages, grouped by repository, for training LLMs.

What's inside

The Stack v2 train-full-ids version based on the deduplicated The Stack v2, further filtered by heuristics. The data covers 600+ programming languages and is grouped by repository. In the metadata, the size is specified as 10-100 million examples; this is a code corpus for training generative models.

What you can do with it

  • Pretrain code LLMs
  • Fine-tune code auto-completion models
  • Research code quality across languages
Tasks
Text generation
Languages
Code
Size
10M–100M rows
Downloads this month
1.2K
Likes
73
Access
on request on Hugging Face

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("bigcode/the-stack-v2-train-full-ids")