Note Diffs only, no full files

Note Contains full new and old file

Note Megadiff: A Dataset of 600k Java Source Code Changes Categorized by Diff Size -- https://arxiv.org/pdf/2108.04631 Refined version of "ASSERT-KTH/megadiff" where each line has the old buggy function and the corrected one Only code, no commit messages

Note Patches, no diffs or full files. Only taken from high quality files

Note Has new file, old file and diff

Note Has full old and new files Filtered bigcode/commitpack for high quality commit messages

Note Diff and commit message only Filtered version of the JetBrains-Research/commit-chronicle

Note Diffs with meta data

Note Add a prefix question to the commit message as an instruction Data taken from bigcode/commitpackft

Note 4 years old but looks good quality Diff and commit message

Note 6 splits containing input output pairs where the input is the code with a bug and the output is the correction RepairLLaMA: Efficient Representations and Fine-Tuned Adapters for Program Repair -- https://arxiv.org/abs/2312.15698