packed_params, %a):
        %w_unpacked : Tensor, %b : Tensor? = quantized::linear_unpack_fp16(%packed_params)
        %r = aten::linear(%a, %w_unpacked, %b)
        return (%r)