Skip to content

Dataset · jordangong

Materialized Source Code from The Stack v2

jordangong/the-stack-v2-smollm3

1 to 10 billion source code snippets across 16 programming languages for training code models.

What's inside

Contains downloaded and decoded source code from the bigcode/the-stack-v2 repository (exact commit and Software Heritage source specified). Features 1 to 10 billion records split by configurations for 16 languages and formats, ranging from Python, C++, and Rust to SQL, Jupyter Notebook, and Markdown. Tables include columns for code text and download status.

What you can do with it

  • Pre-training and fine-tuning language models for code generation
  • Training autocompletion and IDE assistants for specific programming languages
  • Evaluating the parsing and filtering quality of large code corpora
Tasks
Text generation
Languages
Code
Size
1B–10B rows
Downloads this month
84.1K
Likes
8

How to load it

from datasets import load_dataset

ds = load_dataset("jordangong/the-stack-v2-smollm3")