EditPackFT: Code Editing Instructions for LLM Training

EditPackFT is a dataset for training large language models on instructional code editing tasks. It was created by the author 'nuprl' and last updated on February 29, 2024. The dataset is derived from CommitPackFT and provides formatted training windows for code transformation.

Use Cases

• Fine-tuning code LLMs for edit generation based on natural language instructions.
• Training models to understand code diffs based on the 'old_contents' and 'new_contents' fields.
• Studying patterns in code changes and their associated instructions.

Strengths

• Provides a structured format for training with '<before><instruction><after>' windows as mentioned in the description.
• Derived from CommitPackFT, suggesting a foundation in real-world code changes.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Row count is unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.

Provenance

Source

huggingface, author nuprl

Collection Method

Filtered from CommitPackFT.

Freshness

Last updated 2024-02-29 23:22:23; freshness should be verified.