[Feature Gap] CUDA compared to other backends like XPU/CPU

[jiqing-feng]

Feature request

Hi @matthewdouglas .

For now, most features on XPU/CPU are aligned with CUDA, but still some gap btw them.
I can see from the legend that the Intel XPU only partially support 8-bit optimizer. I remembered the XPU has triton implementation for 8bit-optimizer. The partially support means we lack of some features on 8bit optimizer? Please let me know if we have any feature gap btw CUDA (including any new features). Thanks!

Motivation

Fill the feature gap btw CUDA and other backends

Your contribution

I will submit a PR to fix it.