How does the model process process categorical variable embeddings (Lesson 3/4)

In Lessons 3 + 4, we cover how to generate embeddings for categorical variables for days of the week, text data, etc…

I understand why + how to construct these embeddings, but not how the model ingests all the different embedding matrices of different sizes? Does it process each variable separately or combine them into one huge matrix?

(Note: did my best to search through the previous topics for an answer, but please let me know if I missed an existing discussion!)