by wyem on 6/28/23, 3:46 AM with 1 comments
by brucethemoose2 on 6/28/23, 4:04 AM
There are Vulkan runtimes, but still, this is a big deal.
I wonder how performant it will be? I am thinking the dedicated ML frameworks will be better for big models (not to speak of the difficulty of ports), but smaller models will benefit from the lower overhead integration.