When I looked into Tensorflow Lite, I found that custom operations will be exported as a "flex op" not as a native. I don't understand what is "flex op" and what is "native". Thank you!
The code related my questions is in https://github.com/tensorflow/tensorflow/blob/master/tensorflow/lite/toco/tflite/export.cc#L368
See the documentation here: Using TensorFlow Lite with select TensorFlow ops
It's an experimental feature to support TensorFlow ops which are not supported by TensorFlow Lite, by linking and running TensorFlow kernels.
"Flex" is the internal code name for the "Using TensorFlow Lite with select TensorFlow ops" feature.
