Luka Govedič

This would be a very desired feature; There are a lot of times when we want to use a `Trace` with `DebugLevel`. Current best option is to manually implement that....

@bigPYJ1151 @WoosukKwon I did not want to slow down the merging of this PR but I was wondering if there's a plan to decouple the CPU backend via an additional...

Oh, I see. What words are allowed? Can I say `"failed with"`?

@apuaaChen while @tlrmchlsmth ended up using a custom visitor that loads both a row and a scalar from the `float*` argument, I ran into this use-case when the scalar is...

I guess could you let me know if you plan to merge it, or if there's any cleanup you want me to do before we merge. I also have a...

Perfect, thank you!!

And please don't hesitate to ask for any changes or improved comments, and feel free to make edits yourself if there are any style/formatting issues.

@apuaaChen were you able to get the PR run on the internal CI?

@calvin0327 @youkaichao I'll take a detailed look once I'm done reproing an urgent DP/TP bug, but I think this PR looks slightly better than mine, so we should likely do...

Is there a reason you removed the calls to `memory_profiling`? I liked that utility but this is fine as well