✔️ Linting Passed

All linting checks passed. Your pull request is in excellent shape! ☀️

_{Generated for commit:92afcee. Link to the linter CI:here}

+1 for referencing RAPIDS in that section, but the disclaimer for the scikit-learn-intelex item should be extracted as a new paragraph at the beginning of the "Model throughput" section because it would apply for both projects.

I'm not sure about this one. Should it not instead be the cuml project?

I'm not sure about this one. Should it not instead be the cuml project?

Indeed. It is cuML. But cuML is a part of RAPIDS. Not a separate project.

Indeed. It is cuML. But cuML is a part of RAPIDS. Not a separate project.

And intelex is a part of intel, that doesn't mean we refer to intel as a related project 😅

From the perspective of our users, and us, cuML is a separate project, separate repo, separatepypi package. It's as if you reference PyData packages or include numpy/scipy/scikit-learn all as the same "project".

And intelex is a part of intel, that doesn't mean we refer to intel as a related project 😅

Sorry, the project is literally calledRAPIDS, it's the preferred way of the maintainers. When were we discussing the word "Nvidia"? My comment said:

Indeed. It is cuML. But cuML is a part of RAPIDS. Not a separate project.

I am happy for you to drop "Nvidia". (Even though I think it makes sense to keep "Nvidia" in it, because it's strictly only for Nvidia chips.)

And intelex is a part of intel, that doesn't mean we refer to intel as a related project 😅

No, you do. In thedocs you literally callintelex "Intel(R) Extension for scikit-learn". (as shown below)

I'd be okay with includingcuML as the main item, but mentioning it's a part of the RAPIDS ecosystem.

Why not refer to "RAPIDS cuML" to be maximally informative?

I agree with@adrinjalali that we should be specific about which sub-package of the RAPIDS ecosystem is relevant to accelerate scikit-learn on nvidia hardware.

👍 to the suggestion of referring directly to cuml. Otherwise you need to explain to people that they don't need to install all of RAPIDS but only the cuml part. Of course you could install of RAPIDS to get cuml, but it is not needed. IMHO it is a bit like telling people to install anaconda (the distribution) in order to use scikit-learn. (I work for NVIDIA as part of the RAPIDS team - jupp RAPIDS is an overloaded term :D)