pub fn reduce_kernel_virtual<In: Numeric, InSize: Size, Out: Numeric, OutSize: Size, Acc: Numeric>(
input: &VirtualTensor<In, InSize>,
output: &mut VirtualTensor<Out, OutSize, ReadWrite>,
reduce_axis: usize,
out_vec_axis: usize,
blueprint: ReduceBlueprint,
config: ReduceOperationConfig,
)