Hi All!
In short, would I expect to be able to install and run pytorch for the intel xpu relatively
easily and would I expect such a version of pytorch not to conflict with my “regular”
cuda version?
I have a laptop that has an “ultra 9 185h” processor that supports “arc graphics,” as
well as an rtx 3000 laptop gpu.
This documentation, “Getting Started on Intel GPU,” suggests that I should be able
to run pytorch on the intel “xpu.”
Am I right that there is not a pre-built distribution that supports cuda and xpu tensors
simultaneously? (As an aside, would it be currently practical build such a version, or
does the current structure of pytorch not support using both cuda and xpu tensors in
the same pytorch instance?)
If I were to install a separate version of pytorch for the intel xpu, I would presumably
do so in a separate conda environment. (I routinely run different versions of pytorch
in different conda environments without issue.)
If I were to install cuda and xpu versions of pytorch in separate conda environments,
would it be likely that they would not conflict with one another?
Thanks for any pointers about this.
K. Frank