machinelearning icon indicating copy to clipboard operation
machinelearning copied to clipboard

Suggestion: Update Model Builder to support non-CUDA GPU's using OneAPI

Open tauheedul opened this issue 3 years ago • 3 comments

The Model Builder extension currently supports GPU acceleration using CUDA Graphics card.

Please add support for GPU acceleration for non-CUDA graphics cards e.g. AMD Radeon and Intel Arc series. It is possible to migrate existing CUDA code to OneAPI to support the Intel series of graphics cards

e.g. https://codeplay.com/solutions/oneapi/for-cuda/

https://www.intel.com/content/www/us/en/develop/documentation/oneapi-programming-guide/top/software-development-process/migrating-code-to-dpcpp/migrating-from-cuda-to-dpcpp.html

tauheedul avatar Oct 27 '22 11:10 tauheedul

Thanks for the feedback!

@LittleLittleCloud and @JakeRadMSFT Is this possible for Model Builder now? Or is this a framework suggestion?

beccamc avatar Oct 27 '22 15:10 beccamc

@beccamc this is a framework suggestion. Let's transfer the issue there. No action items on MB / CLI side unless this becomes available in the framework.

luisquintanilla avatar Oct 27 '22 16:10 luisquintanilla

In fact, being able to use only an NVidia GPU is now a major hindrance to development, as there are competitive cards from other brands to do this job.

marcos-henrique-tech avatar Jun 14 '24 03:06 marcos-henrique-tech