Skip to main content

Mojo function

col_sum_acc

col_sum_acc(mut col_accum: TileTensor[DType.float32, address_space=AddressSpace.LOCAL], src: TileTensor[DType.float32, address_space=AddressSpace.LOCAL], src_accum: TileTensor[DType.float32, address_space=AddressSpace.LOCAL])

Running-norm for online softmax: col_accum[j] = src_accum[j] + sum(src[*, j, *]).