Dataset · JetBrains-Research
10.7M GitHub commits for description generation
JetBrains-Research/commit-chronicle
A large code change history corpus across 20 programming languages for generating and supplementing commit messages.
What's inside
The dataset from JetBrains Research contains 10.7 million commits and their metadata, collected from 11.9 thousand GitHub repositories across 20 programming languages (descriptions in English). The data is presented in text and tabular formats without strict diff structure filtering. The sample was prepared for a research paper at the ASE 2023 conference.
What you can do with it
- Training and fine-tuning LLMs for automatic commit message generation
- Building autocomplete systems for code change descriptions based on repository history
- Benchmarking source code summarization models
- Tasks
- Text generation, Summarization
- Languages
- Code, English
- Size
- 10M–100M rows
- Downloads this month
- 3.9K
- Likes
- 14
Models trained on it
- mradermacher/smart-commit-0.5b-v0.3-i1-GGUF 3.3K
- mradermacher/smart-commit-1.5b-v0.3-i1-GGUF 3.3K
- mradermacher/smart-commit-0.5b-v0.3-GGUF 999
- mradermacher/smart-commit-1.5b-v0.3-GGUF 911
- LucasBRT/pionus-1.5b-v0.3Text generation 784
- LucasBRT/pionus-0.5b-v0.3Text generation 765
- mradermacher/pionus-0.5b-v0.3-GGUF 653
- mradermacher/pionus-1.5b-v0.3-GGUF 593
How to load it
from datasets import load_dataset
ds = load_dataset("JetBrains-Research/commit-chronicle")