intgemm icon indicating copy to clipboard operation
intgemm copied to clipboard

Published 20 hours ago • verified publisher icon kpu

Multiple quantization weights

Open kpu opened this issue 4 years ago • 0 comments

Allow parts of matrices to have different quantization multipliers: https://github.com/marian-nmt/marian-dev/blob/master/src/tensors/cpu/fbgemm/packed_gemm.cpp#L368

kpu avatar May 08 '20 13:05 kpu