Revert "split serving and monitoring, and merging monitoring and evaluation s…"

[axsaucedo]

Reverts #536

@zhimin-z As discussed let's please review larger changes especially on splits of sections. Reverting as serving and monitoring is closer to serving than evaluation as the latter is only benchmarks. We can open a thread to explore this further.

[zhimin-z]

"the latter is only benchmarks" this is not true... @axsaucedo
Industry-strength evaluation is a collection of evaluation harnesses for ML & LLM models. Evaluation harness, as far as I understand, are tools that are designed specifically to execute benchmarks, do not necessarily be benchmarks or contain benchmarks.

[zhimin-z]

As discussed in our last meeting, you agree to split monitoring and serving, right?
I also removed all benchmarks and datasets in this old pr: #511

Thus, basically, there are no pure benchmarks in this list anymore since this proposal was rejected in our last meeting, check here

I follow very closely what you expect and what we discussed in the previous meetings. @axsaucedo

[axsaucedo]

Industry-strength evaluation is a collection of evaluation harnesses for ML & LLM models.

Ok yes, that's a more accurate wording, that's what I mentioned when I was referring to benchmarks, but certainly benchmark framework / evaluation harness is more accurate.

I follow very closely what you expect and what we discussed in the previous meetings.

Much appreciated, it seems it was just a slight miscommunication. I indeed remember mentioning that I have been keen on finding a way to split the serving and monitoring sections, but mentioned it as an example of being quite hard to split as I hadn't seen enough growth from monitoring tools. I can see how that could have come across as a thumbs up, but yeah I meant more as something to explore.

Ok let's dive into this further in the document as discussion, as indeed I would be interested to explore splitting but I'm not yet sure how it could be merged with another section (as opposed to serving), but would be open to discuss