auto shape = mlir::cast<ShapedType>(value.getType());
  if (shape.getRank() != 1) {
    SmallVector<int64_t> new_shape;
    new_shape.push_back(shape.getNumElements());
    value = builder.create<TF::ReshapeOp>(
        loc, value, Create1DConstValue(builder, loc, new_shape));
  }
  return ConstantFoldOpIfPossible(value.getDefiningOp()).front();
}

    // Get the new shape.
    SmallVector<int64_t, 4> new_shape;
    for (int i = 0; i < 4 - rank; ++i) {
      new_shape.push_back(1);
    }
    for (auto size : input_type.getShape()) {
      new_shape.push_back(size);
    }

    auto reshape_op =
        InsertReshapeOp(concat_op.getLoc(), input, input_type.getElementType(),
                        new_shape, &rewriter);
    reshape_ops.push_back(reshape_op.getResult());

          slice_sizes));

  // Collapses dimensions by reshaping.
  SmallVector<int64_t> new_shape(operand_rank - collapsed_dims.size());
  for (int64_t i = 0, j = 0; i < operand_rank; i++) {
    if (!collapsed_dims.contains(i)) {
      new_shape[j++] = operand_shape[i];
    }
  }
  if (!new_shape.empty()) new_shape[0] = -1;
  return builder.create<TF::ReshapeOp>(
      loc, output.getType(), slice_op,

TF::ReshapeOp GetReshapeOpForConv2DFilter(ArrayRef<int64_t> new_shape,
                                          Value input, OpBuilder* builder) {
  auto reshape_result_type =
      RankedTensorType::get(new_shape, getElementTypeOrSelf(input));
  auto reshape_type = RankedTensorType::get(
      {static_cast<int64_t>(new_shape.size())}, builder->getIntegerType(64));

    TensorType new_shape = RankedTensorType::get(
        {cur_shape[0], cur_shape[1], 1, cur_shape[2] * cur_shape[3]},
        attr.getElementType());

    // Inserts a reshape op.
    auto shape_spec_type =
        RankedTensorType::get({cur_rank}, rewriter.getIntegerType(64));
    auto new_shape_const_attr =
        DenseElementsAttr::get(shape_spec_type, new_shape.getShape());

  }
  int rank = shaped_type.getRank();
  if (rank < 2) {
    return nullptr;
  }
  SmallVector<int64_t> new_shape(shaped_type.getShape().begin(),
                                 shaped_type.getShape().end());
  std::swap(new_shape[rank - 1], new_shape[rank - 2]);
  return shaped_type.clone(new_shape);
}

// Returns a ShapedType for a permutation and the shape of input after

        op->getLoc(), permute_attr.getType(), permute_attr);
    llvm::SmallVector<int64_t, 2> new_shape{type.getShape()[1],
                                            type.getShape()[0]};
    auto output = rewriter.create<TFL::TransposeOp>(
        op->getLoc(), RankedTensorType::get(new_shape, type.getElementType()),
        input, permute);
    return {success(), output};
  };

                                        llvm::ArrayRef<int32_t> perm) {
  llvm::SmallVector<int64_t> new_shape(shape.size());
  for (const auto &perm_element : enumerate(perm)) {
    new_shape[perm_element.index()] = shape[perm_element.value()];
  }
  return new_shape;
}

// Determine if op commutes with transposes. Requires a strict
// definition of Elementwise, all i/o shapes and types must be same-rank

      return {};
    }
  } else if ((new_shape.size() == shape.size() + 1) && new_shape.front() == 1) {
    // Handle the [A, B, C] -> [1, A, B, C] reshape case.
    if (!(std::equal(shape.begin(), shape.end(), new_shape.begin() + 1) &&
          quant_dim == new_shape.size() - 1)) {
      return {};
    }
  } else {
    return {};
  }

  if (!type.hasRank()) return type;

  auto shape = type.getShape();
  SmallVector<int64_t, 4> new_shape(shape.size());

  for (int i = 0; i < permutation.size(); ++i) {
    int64_t index = permutation[i];
    assert(index < shape.size());
    new_shape[index] = shape[i];
  }

  return type.clone(new_shape);
}

// Move Transpose operations that permute `op` operands after the `op`.