Question about example workflow 02_Deploying_a_Neural_Machine_Translation_model(deep learning)

Thank you so much for this precious example.
As I see “02_Deploying_a_Neural_Machine_Translation_model(deep learning)” example, I found out that “Read decoder with tf loop” was not prepared from example workflow “01_Training_a_Neural_Machine_Translation_Model”.

Would you please guide me the way to prepare “Read decoder with tf loop”?

Thank you so much!

Hi @JSLim

nice to meet you!
The workflow you are talking about should be this one on the KNIME Hub: https://hub.knime.com/knime/spaces/Examples/latest/04_Analytics/14_Deep_Learning/02_Keras/12_Machine_Translation/02_Deploying_a_Neural_Machine_Translation_Model

@Kathrin can you help here?

(You can also find companion Forum threads to each of our hub workflows, those are linked at the bottom of the hub pages. The respective one would be here: Neural Machine Translation from English to German: Deployment Workflow)

Hi @JSLim,

welcome to the KNIME Forum!

Unfortunately I don’t have the code, which adds the loop to the TensorFlow model. It has been added by one of our developers for faster execution. Thank you for pointing out that this part is not in the workflow. I will update the workflow so it uses a KNIME Loop instead.

The component NMT Predictor in the training workflow shows you how this loop can be implemented using KNIME nodes (using a recursive loop).

Best
Kathrin