String classification, how to encode character-by-character and train?

Question

I am trying to build a classifier to classify some files into 150 categories based on the name of those files. Here are some examples of file names in my dataset (~700k file