3 ms·
Most people don’t care what hardware it is running on as long as it can do fast matrix multiplications. Right now CUDA is the lowest friction way to do that.
by janalsncm 2y ago
Most people don’t care what hardware it is running on as long as it can do fast matrix multiplications. Right now CUDA is the lowest friction way to do that.