Description:
I’m working on a sequence-to-sequence model using TensorFlow and SentencePiece to dynamically replace placeholders like {NAMESPACE} in user inputs. My training data includes examples with both the placeholder and actual values. During inference, I want the model to replace {NAMESPACE} with the user-specified value without hardcoding it. How can I ensure the model learns this pattern effectively and handles dynamic replacements during inference?

Input : list pods in namespace abc
Actual Output : oc get pods -n def

Expected output : oc get pods -n abc

I know I can implement NER .. just wanted to know if it can be done through model training
I don’t want to use pre trained models.