I am creating a system that creates playlist through text description and have decided to use LLM by fine-tuning it

I have data consisting of playlist name tracks in the playlist each with it’s name music features eg (speechiness , acousticness etc)

How can i continue with my research?

What kind of LLM model can I use?

Are there any references I can take?

Any recommendations or suggestions is welcomed thank you!

I know basics of LLM but never have fine tuned one.
Additionally i have looked at project implementing LLM but my requirement of train it with music data for playlist generation didn’t seem to match any. Below is what my data looks like.

enter image description here