// so not passing it during conversion. This design might change if needed.
  ArrayRef<double> scales = qtype.getScales();
  ArrayRef<int64_t> zero_points = qtype.getZeroPoints();
  const int num_channels = scales.size();
  TensorType scale_type = RankedTensorType::get(
      {static_cast<int64_t>(num_channels)}, rewriter.getF32Type());

          new_qtype = mlir::quant::UniformQuantizedPerAxisType::get(
              per_axis.getFlags(), per_axis.getStorageType(),
              per_axis.getExpressedType(), per_axis.getScales(),
              per_axis.getZeroPoints(), per_axis.getQuantizedDimension(),
              per_axis.getStorageTypeMin() - 1, per_axis.getStorageTypeMax());
        } else if (auto per_tensor =