* If you work for a company that wants to allow you to contribute your work, then you'll need to sign a [corporate CLA](https://code.google.com/legal/corporate-cla-v1.0.html).

  tensorflow::mutex execution_mutex_;
  ExecutionState execution_state_ TF_GUARDED_BY(execution_mutex_) =
      ExecutionState::kIdle;
  // Tells the worker thread that there is new work.
  tensorflow::condition_variable start_execute_;
  // The worker thread notifies that work has finished.
  tensorflow::condition_variable finished_execute_;
  // Notifies a StartExecute that the previous Join has finished.
  tensorflow::condition_variable finished_join_;

//  - target: dy in dx/dy
//  - source: dx in dx/dy
//  - tensor: one of the many inputs or outputs of an operation
//
// Below here we do the gradient algorithm. It works as follows:
//
// First we filter the tape to just the subset of operations we want to
// differentiate. In the process of doing so we count how many times each Tensor

    forward_op_->outputs.push_back(retvals[i]);
  }
  // TODO(b/166669239): This is needed to support AttrBuilder::Get for string
  // attributes. Number type attrs and DataType attrs work fine without this.
  // Consider getting rid of this and making the behavior between number types
  // and string consistent.
  forward_op_->attrs.BuildNodeDef();
  std::unique_ptr<GradientFunction> gradient_fn;