Open ar4s-gh opened 6 months ago
Thank you for the feature request to create an ML-powered pre-commit hook. However, this issue currently lacks some key details needed for our AI coding assistant Sweep to implement it:
The exact code files and locations that need to be modified should be specified. Sweep is great at making targeted code changes but needs more context on where the changes are needed.
The linked research PDF provides useful background but Sweep is not able to access and interpret information from external files like this. Please summarize the key points from the paper directly in the issue description.
The issue mentions some command-line operations and changes to the git workflow that are outside of Sweep's current capabilities. Those requirements should be removed or noted as being handled separately.
Overall this is a fairly complex and ambitious feature. It would be helpful to break it down further into smaller, incremental coding tasks that Sweep can handle. Let's start with something bite-sized.
Please update the issue description with those details and re-submit it. We appreciate you taking the time to report this and look forward to working with you to get it implemented!
[!TIP] To recreate the pull request, edit the issue title or description.
This is an automated message generated by Sweep AI.
I'm a bit of a novice, for each of the suggested points to expand on, could I have you provide exact deliverables.
You can provide mock examples of what each should contain, down to the leaf
Based on the Delta-tuning explained here, we will design an git-based incremental fine-tuning MVP. The pre-commit should work with git(hub)flow and stacked-git workflows. A high level is: A modelfile will live in .git that will adjust weights based on commit history and code changes within a repo. The flow