quant::UniformQuantizedPerAxisType ResetAxisAndBroadcast(
    const ArrayRef<int64_t> shape,
    const quant::UniformQuantizedPerAxisType qtype, const Type target,
    const int quant_dim) {
  const auto shaped = dyn_cast<RankedTensorType>(target);
  if (!shaped) return {};
  const ArrayRef<int64_t> new_shape = shaped.getShape();

  SmallVector<double, 4> scales(qtype.getScales().begin(),
                                qtype.getScales().end());

            /*scale=*/1.0 / output_inverse_scale_value,
            output_zero_point_value);

    Value conv_output_value = op.getResult();
    auto output_uniform_quantized_tensor_type = RankedTensorType::getChecked(
        rewriter.getUnknownLoc(),
        /*shape=*/
        mlir::cast<TensorType>(conv_output_value.getType()).getShape(),
        output_uniform_quantized_type);

    builder.setInsertionPoint(op);
    if (!const_true)
      const_true = builder.create<TF::ConstOp>(
          op.getLoc(),
          DenseIntElementsAttr::get(
              RankedTensorType::get(/*shape=*/{}, builder.getI1Type()), true));

    op.getIsInitialized().replaceAllUsesWith(const_true);
    op.erase();
  }
}

// Performs store-load forwarding. This effectively removes

  CPred<"$_op.getOperand(" # n # ").getType().isa<RankedTensorType>()">,
  CPred<"$_op.getOperand(" # n # ").getType().cast<ShapedType>().getRank() == "
    # dim>]>;

// True if operand n is ranked and has a rank > dim.
class TFL_OperandIsRankedAndHasDimPred<int n, int dim> : And<[
  CPred<"$_op.getOperand(" # n # ").getType().isa<RankedTensorType>()">,

constexpr bool kDoAssertions = true;

void AddAssertion(OpBuilder& builder, Location& loc, Value cond,
                  const std::string& message) {
  if (!kDoAssertions) return;
  auto shape_type =
      RankedTensorType::get({1}, builder.getType<TF::StringType>());
  auto msg = builder.create<TF::ConstOp>(
      loc, DenseStringElementsAttr::get(shape_type,
                                        llvm::ArrayRef<StringRef>{message}));

    OpBuilder<(ins "StringAttr":$container, "StringAttr":$shared_name,
      "BoolAttr":$use_node_name_sharing, "TypeAttr":$key_dtype, "TypeAttr":$value_dtype),
    [{
      build($_builder, $_state,
      mlir::RankedTensorType::get({},
      $_builder.getType<mlir::TF::ResourceType>()),
      container, shared_name, use_node_name_sharing, key_dtype, value_dtype);
    }]>];
}

def TF_IFFTOp : TF_Op<"IFFT", [Pure]> {