I suggest you use the pytorch or tensorflow framework with the library used, namely huggingface.simply you fine-tune the model after it has been pre-trained. My suggestion is that you use the GPT 2 model with the precision size that you use.