I do not undestand, how the deployment of a ML-model works in the reality.
A given dataset needs to be mostly time pre-processed (for example One Hot Encoding). After will be a model cretead and stored.
As next will be the model with Streamlit deployed. I mean the option when the new data could be uploaded and a "Label" column will be generated - predicted values.
Here is the point. The new data are not being pre-processed (for example One Hot Encoding) and they are in the form as they were before modelling. The model deployment is mainly used from the customers and they do not know "about pre-processing".
Have I something misunderstood or could someone bring more light into this topics?