3 ms·
Can hardware vendors even put their differences aside to build such a thing? We can't even build a unified open raster graphics API, and now you're asking for m
by smoldesu 3y ago
Can hardware vendors even put their differences aside to build such a thing? We can't even build a unified open raster graphics API, and now you're asking for machine learning acceleration in that vein?
- OkayPhysicist 3y agoMachine learning would probably be the simpler API. If you can speak Linear Algebra, you're most of the way there.
- smoldesu 3y agoYou're right, and it's why projects like the ONNX runtime exist to unify vendor-specific AI accelerators. Covering the basics isn't too hard. What GP seems to be asking for is an open CUDA replacement, which is kinda like asking someone to fund a Free and Open Source cruise ship to compete with Carnival for you. You'll get somewhere with some effort, luck and good old human intuition, but Nvidia can outspend you 10:1 unless you have funding leverage from FAANG.