please avoid deleting CM automations and artifacts from this repository that are used in other projects

[gfursin]

Dear all,

We need to find a better way to coordinate changes in this repository.

Yesterday's removal of many "unused" artifacts broke other MLCommons projects.

@arjunsuresh - I managed to restore the functionality of other projects by reverting your 630 PR that deleted many artifacts that were not created by you and that are required by other projects. Please be careful with such changes.

If you are not sure about some artifacts in this repository and want to delete them, please check the author of that artifact/automation and then create an issue in this repository and tag that author for a discussion - that will help prevent breaking other projects.

Also, please restore these files in the mlperf-inference branch that you are maintaining - it also breaks dependencies in other projects.

I remind that I designed CK and CM for collaborative and reproducible development, i.e. we can enhance functionality but we need to keep all artifacts for backwards compatibility and reproducibility. If a given CM automation or artifact requires complete redesign, we create a copy with new tags/UID and gradually change dependencies in other automations to the new one but we do not delete artifacts from others.

Thank you very much for understanding!

[gfursin]

We may convert it into a discussion if needed.

[arjunsuresh]

Hi @gfursin The usage of the deleted folders are unaware to us as they are not documented anywhere and no tests were also added for those. We are currently working on significant changes in the CM script automation as we are writing an arxiv paper and documentation explaining the usage of CM scripts. For now we'll be using only the mlperf-inference branch and eventually will move that to a separate repository to avoid conflicts with the generic CM script usage. We won't be touching the main branch of this repository again.

[gfursin]

Hi @arjunsuresh . May I ask you who made this decision about not using the main branch and where it is documented?

[arjunsuresh]

Hi @gfursin usage of "mlperf-inference" branch and not the main branch is internal to our MLPerf related automation work. Our interface to the external users is the MLPerf inference documentation

Do you have any concern here?

[gfursin]

Hi @gfursin usage of "mlperf-inference" branch and not the main branch is internal to our MLPerf related automation work. Our interface to the external users is the MLPerf inference documentation

Do you have any concern here?

Yes, several MLCommons members have serious concerns - checking with MLCommons ...

[gfursin]

I have restored files required by other projects.