I would advise to check first the train_val dataset codes first as I can see dataset mismatch.
Next your fit_label vectorizer error mentions you haven’t mentions the MAX_LEN as per the instructions. Check how you have vectorize or instantiate the textvectorization class.
I suspect more of variables mix up issue, you are suppose to use the arguments provided in the grade function preprocess data.
Go through this comment, see if it helps you debug your codes.