Add CUDA kernel support for 4-bit quantization with blocksize=32#1854

matthewdouglas merged 7 commits intobitsandbytes-foundation:mainbitsandbytes-foundation/bitsandbytes:mainfrom

Abdennacer-Badaoui:32-blocksize-supportAbdennacer-Badaoui/bitsandbytes:32-blocksize-supportCopy head branch name to clipboard