Method handle

tensorflow/stream_executor/cuda/cuda_dnn.cc:182–182

Returns cuDNN handle. To be passed directly to cuDNN APIs, don't keep a copy.

  // Returns cuDNN handle. To be passed directly to cuDNN APIs, don't keep
  // a copy.
  cudnnHandle_t handle() const { return handle_; }

 private:
  gpu::ScopedActivateExecutorContext context_;