I had tried to install oneAPI on windows,then install a conda,followed instrutions on:Welcome to Intel® Extension for PyTorch* Documentation!
But it report errors like:Error loading "C:\ProgramData\anaconda3\Lib\site-packages\intel_extension_for_pytorch\bin\intel-ext-pt-gpu.dll" or one of its dependencies.
Just like what happened when I tried to install it on wsl,on ubuntu 22.04,if I directly install AI tool kit and source it,it returns segmentation fault then teminated the python complier after I used to('xpu') methor.And if I tried to install with commands above for linux,it reports missing on a gpu related .so package.
In ipex document,oneAPI ver 2023.2.0 is requested,but it automatically downloads 2024.0 when I tried to get through.
What should I do to train my model on my windows system?Not directML,but ipex because even 12G Gmem is too small for dml consumptions.I need a way for a speed and compacity at lease meets GTX 1080,better be able to accelerate by integrated graphics.If oneAPI ver 2023.2.0 is basical but no support with 2024.0,is there any way to get a installer?
Main sys:windows 10 22H2 19045.3693
python tried but faild:3.11.5,intelpython 3.9.
NUC X15 laptop kit,added in 32*2 4800mhz mem,1T KIOXIA RC20 SSD.
Hi 子棋 彭, we can see that you have created a duplicate thread here: https://community.intel.com/t5/oneAPI-Registration-Download/Is-Intel-extention-of-torch-died/m-p/1549997#M6577, with a similar concern. We will close this thread as a duplicate and we will continue our support on that thread.
If you have any further query, please post a new question as this thread will no longer be monitored by Intel®.