// Match Add
    mlir::TFL::AddOp add_op =
        dyn_cast_or_null<TFL::AddOp>(strided_slice_op.getEnd().getDefiningOp());
    mlir::TFL::SubOp sub_op =
        dyn_cast_or_null<TFL::SubOp>(strided_slice_op.getEnd().getDefiningOp());
    if (!(add_op || sub_op)) {
      return failure();
    }

    // Check that add rhs is constant.
    DenseElementsAttr added_value;

    }
  }

  // Slice exactly the first shape dimension:
  //   begin = [0] end = [1], strides = [1]
  auto begin = get_const_int(slice_op.getBegin(), /*expected_rank=*/1);
  auto end = get_const_int(slice_op.getEnd(), /*expected_rank=*/1);
  auto strides = get_const_int(slice_op.getStrides(), /*expected_rank=*/1);
  if (!begin.has_value() || !end.has_value() || !strides.has_value() ||
      *begin != 0 || *end != 1 || *strides != 1)

  if (offset && !IsOfRankedFloatTensorType(offset, 1))
    return op.emitOpError("requires offset to be a 1D float tensor");

  auto mean = GetRankedTensorTypeForOperand(op.getMean());
  if (mean && !IsOfRankedFloatTensorType(mean, 1))
    return op.emitOpError("requires mean to be a 1D float tensor");

  auto variance = GetRankedTensorTypeForOperand(op.getVariance());

    if (begin_type.getRank() != 1) return failure();
    if (begin_type.getDimSize(0) > num_input_dims) return failure();
  }

  if (auto end_type = op.getEnd().getType().dyn_cast<RankedTensorType>()) {
    if (end_type.getRank() != 1) return failure();
    if (end_type.getDimSize(0) > num_input_dims) return failure();
  }

  if (auto strides_type =