If you want to use the new custom op, run the script to build the example.
./build_custom_op.sh
Now we can train and export the model with this script.
./main.py
If we run the serving service like this, we will get the error about missing custom op.
simple_tensorflow_serving --model_base_path="./model/"
Now we can load the custom op while starting the serving service with this command.
simple_tensorflow_serving --model_base_path="./model/" --custom_op_paths="./"