using tensorflow::NodeDef;
  using tensorflow::OpRegistrationData;
  using tensorflow::Tensor;
  using tensorflow::shape_inference::DimensionHandle;
  using tensorflow::shape_inference::InferenceContext;
  using tensorflow::shape_inference::ShapeAndType;
  using tensorflow::shape_inference::ShapeHandle;

  const int num_inputs = input_shapes->num_items;
  NodeDef node_def;

}

namespace {

// Helper method that creates a shape handle for a shape described by dims.
tensorflow::shape_inference::ShapeHandle ShapeHandleFromDims(
    tensorflow::shape_inference::InferenceContext* ic, int num_dims,
    const int64_t* dims) {
  if (num_dims != -1) {
    std::vector<tensorflow::shape_inference::DimensionHandle> dim_vec;
    dim_vec.reserve(num_dims);
    for (int i = 0; i < num_dims; ++i) {

import org.jetbrains.kotlin.analysis.api.symbols.pointers.KtSymbolPointer
import org.jetbrains.kotlin.analysis.api.types.KtType
import org.jetbrains.kotlin.descriptors.FunctionDescriptor
import org.jetbrains.kotlin.resolve.calls.inference.returnTypeOrNothing
import org.jetbrains.kotlin.resolve.descriptorUtil.isExtension

internal class KtFe10DescAnonymousFunctionSymbol(
    override val descriptor: FunctionDescriptor,

namespace tensorflow {
namespace tracing {
namespace graph {

class GraphContext;
class GraphOperation;
class GraphTensor;

auto& kUnknownDim = shape_inference::InferenceContext::kUnknownDim;
auto& kUnknownRank = shape_inference::InferenceContext::kUnknownRank;

// GraphTensor wraps a `TF_Output`, i.e. a pointer to TF_Operation and the index
// into the list of outputs for the operation.

import org.jetbrains.kotlin.descriptors.PropertyGetterDescriptor
import org.jetbrains.kotlin.descriptors.hasBody
import org.jetbrains.kotlin.name.CallableId
import org.jetbrains.kotlin.resolve.calls.inference.returnTypeOrNothing

internal class KtFe10DescPropertyGetterSymbol(
    override val descriptor: PropertyGetterDescriptor,
    override val analysisContext: Fe10AnalysisContext

import org.jetbrains.kotlin.analysis.api.types.KtType
import org.jetbrains.kotlin.name.CallableId
import org.jetbrains.kotlin.name.Name
import org.jetbrains.kotlin.resolve.calls.inference.returnTypeOrNothing
import org.jetbrains.kotlin.resolve.descriptorUtil.isExtension
import org.jetbrains.kotlin.resolve.sam.SamConstructorDescriptor
import org.jetbrains.kotlin.resolve.sam.SamTypeAliasConstructorDescriptor

inspected and debugged and it is intended to be used during the development
phase.

As part of the differences that make Eager mode easier to debug, the [shape
inference
functions](https://www.tensorflow.org/guide/create_op#define_the_op_interface)
are skipped, and any checks implemented inside the shape inference code are not
executed.

The security impact of skipping those checks should be low, since the attack

import org.jetbrains.kotlin.name.CallableId
import org.jetbrains.kotlin.name.Name
import org.jetbrains.kotlin.resolve.DescriptorUtils
import org.jetbrains.kotlin.resolve.calls.inference.returnTypeOrNothing
import org.jetbrains.kotlin.resolve.calls.tasks.isDynamic
import org.jetbrains.kotlin.resolve.descriptorUtil.isExtension

internal class KtFe10DescFunctionSymbol private constructor(

            }
        else firFunctionSymbol.valueParameters[valueParameterReference.parameterIndex]
    )

    // Util function to avoid hard coding names of the classes. Type inference will do a better job figuring out the best type to cast to.
    // This visitor isn't type-safe anyway
    private inline fun <reified T> ConeContractDescriptionElement.accept() =

import org.jetbrains.kotlin.name.StandardClassIds
import org.jetbrains.kotlin.renderer.DescriptorRenderer
import org.jetbrains.kotlin.renderer.render
import org.jetbrains.kotlin.resolve.calls.inference.CapturedType
import org.jetbrains.kotlin.resolve.descriptorUtil.annotationClass
import org.jetbrains.kotlin.types.*
import org.jetbrains.kotlin.types.checker.NewCapturedType