categoricalNB - One Hot Encoding

[Tryd3x]

I've been working on implementing CategoricalNB for river and i would like to understand what's the best approach for encoding categorical values (Ordinal or One Hot or etc). Below I've considered the following data sample:

After performing one hot encoding, I have the following:

Currently, I have the following hierarchy to store the count of target class values c associated with each feature i with category t in a dictionary (according to sklearn's categoricalNB as follows:

{"feature" :
{"category" :
{"class": count } } }

My Question: Is the first level of hierarchy ('feature') necessary in order to store the count of each categorical value in the dictionary? or Should I just eliminate the first level and maintain a "category" -> "class" -> count hierarchy?