TF_TriState dependency_optimization;
  TF_TriState auto_parallel;
  TF_TriState memory_optimization;
  TF_TriState scoped_allocator_optimization;
} TP_OptimizerConfigs;

#define TP_OPTIMIZER_CONFIGS_STRUCT_SIZE \
  TF_OFFSET_OF_END(TP_OptimizerConfigs, scoped_allocator_optimization)

// Struct for Optimizer. Plugin authors must provide an optimize function.
// Creation and deletion functions are optional.

// efficiency of forward-mode autodiff. Instead, language bindings should call
// DeleteGradient as soon as a Tensor which was `Watch`ed or was an output
// Tensor passed to Accumulate goes out of scope.
//
// Not thread-safe.
template <typename Gradient, typename BackwardFunction, typename TapeTensor>
class ForwardAccumulator {
 public:
  // Does not take ownership of `vspace`, which must outlive the