func_def.mutable_attr()));

  for (int i = 0, e = function.getNumArguments(); i < e; ++i) {
    if (auto resource_arg_unique_id_attr =
            function.getArgAttrOfType<mlir::IntegerAttr>(
                i, kResourceArgUniqueIdAttr)) {
      (*func_def.mutable_resource_arg_unique_id())[i] =
          resource_arg_unique_id_attr.getInt();
    }
  }

namespace mlir {
namespace TF {

namespace {

// Creates ConstOp for int32_t value.
ConstOp createI32ConstOp(int32_t value, Location loc,
                         PatternRewriter* rewriter) {
  auto int_attr = IntegerAttr::get(rewriter->getIntegerType(32), value);
  return rewriter->create<ConstOp>(loc, int_attr);
}

// Creates ConstantOp for array of int32_t.
arith::ConstantOp createI32ConstantOp(llvm::ArrayRef<int32_t> values,

}

TypeAttr GetQuantizedTypeAttr(const Builder builder, const Type input_type,
                              const Attribute min, const Attribute max,
                              const int quant_dim, const IntegerAttr num_bits,
                              const BoolAttr narrow_range, const bool is_signed,
                              const bool legacy_float_scale,
                              const bool use_fake_quant_num_bits) {

// if symmetric quantization is required.
TypeAttr GetQuantizedTypeAttr(Builder builder, Type input_type, Attribute min,
                              Attribute max, int quant_dim,
                              IntegerAttr num_bits, BoolAttr narrow_range,
                              bool is_signed, bool legacy_float_scale = false,
                              bool use_fake_quant_num_bits = false);

    "whose size is at most " # n>;

// Ensures the given integer attribute has the given value.
class TFL_IntEqualsTo<int n> : AttrConstraint<
    CPred<"$_self.isa<IntegerAttr>() && "
            "$_self.cast<IntegerAttr>().getInt() == " # n>,
    "whose value is " # n>;

// Ensures the given LSTMKernelType attribute has the given value.
class TFL_LSTMKernelTypeEqualsTo<string value> : AttrConstraint<

  mlir::ElementsAttr layer_stats = mlir::DenseFPElementsAttr::get(
      tensorflow::GetTypeFromTFTensorShape({2}, b.getF32Type()),
      {min_maxs[0], min_maxs[1]});
  mlir::ElementsAttr axis_stats;
  mlir::IntegerAttr axis;
  if (mins.size() > 1) {
    llvm::SmallVector<int64_t, 4> axis_stats_shape{
        static_cast<int64_t>(mins.size()), 2};
    axis_stats = mlir::DenseFPElementsAttr::get(

    Value scalar_zero = CreateI32SplatConst(op.getLoc(), rewriter, {}, 0);
    Type shape_dtype = getElementTypeOrSelf(op.getElementShape().getType());
    Value num_elements = operands[1];
    IntegerAttr attr;
    if (matchPattern(num_elements, m_Constant(&attr))) {
      return CreateI32SplatConst(op.getLoc(), rewriter, {1}, attr.getInt());
    }
    if (auto const_op = num_elements.getDefiningOp<TF::ConstOp>()) {

         << " at index path: <value>";
      for (auto path_element : current_index_path_) {
        os << ".";
        if (auto integer = mlir::dyn_cast<mlir::IntegerAttr>(path_element)) {
          os << integer.getValue();
        } else {
          auto str = mlir::cast<mlir::StringAttr>(path_element);
          os << str.getValue();
        }
      }

  TF_DerivedOperandTypeAttr TI = TF_DerivedOperandTypeAttr<0>;

  let builders = [
    OpBuilder<(ins "Value":$indices, "Value":$depth, "Value":$on_value,
      "Value":$off_value, "IntegerAttr":$axis)>
  ];

  let hasVerifier = 1;
}

def TF_OneShotIteratorOp : TF_Op<"OneShotIterator", []> {
  let summary = [{
Makes a "one-shot" iterator that can be iterated only once.