Is there a way to use dynamic decoding rather than fix output size for the tflite model?

[SefaZeng]

The current tflite implementation needs to fix the length of the output sequence,e.g 250. Is there a way to use a really dynamic decodng method?

[gcervantes8]

Maybe? You'd mostly have to modify the opennmt/utils/decoding.py file to find out I believe. When I tried it with TensorFlow 2.3 it didn't work, but maybe with the newest TensorFlow it could work.