What model would best fit a structured text generator?


I apologize if this question seems misguided.

I was wondering what model would be good for generating structured/tabular data.
I have a dataset, where each row includes data attributes from two tables. If the attributes from the two tables are alike, the classification is considered a “Match”.
I’ve tried to use the GPT-2 model and fine-tune it on structured data regarding entity alignment. However, it’s rather unpredictable when it comes to its generation. I feel it is “overkill” to use GPT-2, and have been eyeing this as a “seq2seq” task, rather than “text-generation”.

I was wondering if anybody had a better suggestion for a model to use?

Some pointers would be lovely!