Franck Charras

I cross posted the issues with intel extension for pytorch repo at https://github.com/intel/intel-extension-for-pytorch/issues/368 . Environment informations: ``` Collecting environment information... PyTorch version: 1.13.0a0+git49444c3 PyTorch CXX11 ABI: Yes IPEX version: 1.13.120+gitb243ae3...

Wouldn't it be `(threads per compute unit) * (number of compute units)`, where the former is indeed `8` as the optimization guide shows, but the latter is `number of compute...

I can understand that there are several sources for getting the runtime and that some of those are more reliable or up to date than other, my concern is that...

Maybe you could use github actions to set up an automated build & push to pypi when new tags are issued ? The runners available with github workflow look strong...

Well `numpy.from_dlpack` can't take dlpacks from cuda tensors apparently. I raise the question of wether this should be working or not at the [dlpack repo](https://github.com/dmlc/dlpack/issues/132).

In last commits, I try adding a more exhaustive `_average` function and remove `_weighted_sum`. With all the added input validation and type inspection, the numpy runtime might be considerably slower,...

After the latest commit, I think here are the remaining todos / discussions : - [ ] let a comment explaining the reason for the `xp.reshape` call (and figure out...

The PR is mainly blocked on this particular point of discussion that does not manage to settle, could we make it progress independently of that ? the array api dispatch...

@adrinjalali that said, I don't agree with your last points, in regard of @betatim [previous proposals](https://github.com/scikit-learn/scikit-learn/pull/27904#issuecomment-1959622801). I agree that we can't commit to returning "same results" but wouldn't say that...

Thanks everyone for continuing this PR, I now caught up with latest diff and I'm also a happy bunny. I want to mention 2 differences I think I've spotted between...