const int input_rank = input_type.getRank();

    // Create a 1D I32 tensor for representing the dimension permutation.
    auto permuation_tensor_type =
        RankedTensorType::get({input_rank}, rewriter.getIntegerType(32));
    llvm::SmallVector<Attribute, 4> permute;
    permute.reserve(input_rank);
    // First create an identity permutation tensor.

  // begin = [1, 2, 1], size = [1, 3, 2]
  Input input = op.input(0);
  Input begin = op.input(1);
  // input_rank = 3
  auto input_rank = Rank(scope, input);
  // slice_size = [1, 3, 2]
  auto slice_size = Shape(scope, op.output(0));
  // padding_shape = [3, 1]
  auto padding_shape = Stack(scope, {input_rank, 1});
  // before_padding = [[1]
  //                   [2]
  //                   [1]]

  ShapedType weight_type = mlir::cast<ShapedType>(weight.getType());
  const int32_t input_rank = input_type.getRank();
  const int32_t weight_rank = weight_type.getRank();
  const int32_t broadcasted_rank = std::max(input_rank, weight_rank);

  const int32_t num_matmul_dim = 2;
  const int32_t num_input_batch_dim = input_rank - num_matmul_dim;
  const int32_t num_weight_batch_dim = weight_rank - num_matmul_dim;