Conversation
Overview of the DeepSpeed Committers Responsibility.
Adding Zhipeng Wang to the DeepSpeed TSC Committers list.
Adding Zhipeng Wang to the DeepSpeed TSC Committers.
Adding Zhipeng Wang to the DeepSpeed TSC Committers.
|
|
||
| ### Reviewing Pull Requests, Issues and Proposals (RFCs) in a timely manner, and help maintain the necessary code quality for all contributions. | ||
|
|
||
| Committers are responsible for maintaining the quality of the codebase and ensuring that contributions meet the project's standards. |
There was a problem hiding this comment.
One open is PR progress tracking. Should committers responsible for PR progress or we leave it to contributor? If the PR itself have potential but original submitter no longer have time to work on it, what would be the right action to do? We might need a next level PR guideline document on these details.
There was a problem hiding this comment.
@delock, I agree that we need more clarity here. To reduce committer burden, I think committers should engage with PRs when the contributor requests a review (or come out of Draft mode). So, there will be two cases of abandoned PRs.
- Draft PRs - I think we can configure automatic closing after X number of days of no activity
- Under review - Committer who is tracking progress can decide when to close.
If the PR is of great interest, then it is likely that someone else will pick it up. If under review, the involved committer could choose to complete it.
There was a problem hiding this comment.
We could also learn from other repos like HF transformers and Pytorch on their process.
There was a problem hiding this comment.
@delock @sfc-gh-truwase I agree with both of you that we need a better mechanism to handle incoming PRs. Here are my thoughts: 1. We can introduce the "two-stage" review process (similar to the AI conference paper reviews). If the PR quality below certain threshold or is not promising/interesting to the community, we can introduce "desk-reject" to close it out before reviewing it further; 2. I agree with Tunji that for the interesting PRs, if no actions are taken after X numbers of days, we can close it and convert it into a RFC; The committer who choose to complete it should work on it, or he/she can invite other contributors to complete it.
There was a problem hiding this comment.
I don't think there can be guidelines about how contributor committers do things wrt whether they bring the PR to the finish line or not if the original PR creator is gone.
If one or more persons are interested in completing the PR they will.
There was a problem hiding this comment.
If nobody wants to continue engaging it's probably fine to auto-close the PR after 1 month of no activity.
There was a problem hiding this comment.
I will incorporate the suggestions and modify the PR shortly.
5 participants
Overview of the DeepSpeed Committers Responsibility.