Keras inputs/outputs. Use `GradientTape` on the actual Tensors passed to
        the already-constructed model instead.
    *   Code that requires very tricky shape manipulation via converted op
        layers in order to work, where the Keras symbolic shape inference proves
        insufficient.
    *   Code that tries manually walking a `tf.keras.Model` layer by layer and

// A list of TF_ShapeAndType elements..
struct TF_ShapeAndTypeList {
  int num_items;
  TF_ShapeAndType* items;
};
typedef struct TF_ShapeAndTypeList TF_ShapeAndTypeList;

// API for manipulating TF_ShapeAndTypeList objects.
//
TF_CAPI_EXPORT extern TF_ShapeAndTypeList* TF_NewShapeAndTypeList(
    int num_shapes);
TF_CAPI_EXPORT extern void TF_ShapeAndTypeListSetShape(