Dataset · mikex86
Archive of 60 million StackOverflow posts in Markdown
mikex86/stackoverflow-posts
A large corpus of StackOverflow questions, answers, and code up to June 14, 2023, in Markdown text.
What's inside
The dataset includes around 60 million StackOverflow posts published up to June 14, 2023, with a total volume of about 35 GB (roughly 65 billion characters). English texts and code snippets are formatted in Markdown. Data is gathered from the official StackExchange dump in the Internet Archive.
What you can do with it
- Pre-training and fine-tuning (SFT) language models for code generation and programming answers
- Creating a knowledge base for developer assistance RAG systems
- Tasks
- Question answering, Text generation
- Languages
- Code, English
- Size
- 10M–100M rows
- Downloads this month
- 13.1K
- Likes
- 64
Models trained on it
How to load it
from datasets import load_dataset
ds = load_dataset("mikex86/stackoverflow-posts")