Committed Train: Code Diff to Commit Message Pairs for Model Fine-Tuning

Committed Train is a dataset of filtered diff and Conventional Commit message pairs derived from CommitChronicle. It was built by the Committed project for fine-tuning small models to write commit messages from a code diff. The dataset was last updated on June 4, 2026, by the author marzoukbaig14.

Use Cases

• Fine-tuning language models to generate commit messages based on code diffs.
• Training models to follow the Conventional Commits specification for structured commit messages.
• Researching automated code documentation and summarization techniques.

Strengths

• Pairs are specifically filtered for Conventional Commits format, providing a structured training target.
• Derived from the CommitChronicle source, suggesting a foundation in real-world commit history.

Limitations

• Row count and dataset size are unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• The 'reasoning_trace' field is reserved for a future version and is currently null.

Provenance

Source

Derived from CommitChronicle.

Collection Method

Filtered diff and Conventional Commit message pairs.

Freshness

Last updated 2026-06-04 17:45:23.