README.md

metadata

license: unknown
task_categories:
  - text-generation
language:
  - code
pretty_name: Enriched Commit Diff Fine-tuning Dataset

Enriched Commit Diff Fine-tuning Dataset

Generated from jedisct1/rust on 2026-06-07T17:21:21.169834+00:00.

Each kept source commit produces four supervised fine-tuning variants:

• message_to_diff: original commit message -> original diff
• diff_to_message: original diff -> original commit message
• minimized_message_to_diff: concise/minimized commit message -> original diff
• message_to_shuffled_diff: original commit message -> original diff with file blocks shuffled

Output files are line-safe JSONL shards. By default they are gzip-compressed (.jsonl.gz) and rotated at roughly the configured uncompressed shard size.

Rows include both plain prompt/completion fields and chat messages fields:

• instruction
• input
• output
• prompt
• completion
• messages

Build stats

Language prompt: Rust
File filter: rust

Split	Source rows seen	Source rows kept	Enriched rows	Skipped by file filter
train	356184	356183	1424732	1

Total source rows kept: 356183
Total enriched rows: 1424732

Regeneration

python build_enriched_rust_dataset.py   --source-dataset jedisct1/rust   --output-dir enriched-rust-finetune-dataset

To upload after regeneration:

python build_enriched_rust_dataset.py   --source-dataset jedisct1/rust   --output-dir enriched-rust-finetune-dataset   --push-to YOUR_USERNAME/YOUR_DATASET_NAME