Fine-tune for Multi Label Classification

[bi1101]

Hi everyone, I came here through Predibase.

I'm not really well-versed in ML so I want to ask a few question on fine-tuning LLM if that's ok.

My use case is in essay evaluations where there are multiple labels (10 out of a total of 100 possible labels) assigned to an essay to produce a score.

These labels I think can be categorized into 4 main type (like Vocabulary, Grammar,...) which decreases the complexity of each API calls.

My question is, is it possible to fine tune an LLM to perform such a complex task, given I will eventually gather enough data?

And is this process any different from the docs for regular prompt-to-response fine tuning llm in Predibase docs?