Thank you Bhogayatachandrakant for your interest in my article. Regarding your questions:

(1) Not sure what you define long text. Gpt2 models usually have 1024 max sequence lenght. So any text of size 1024 tokens ids or smaller will fit in the model.

(2) Yes.

(3) No. You will need to train your own GPT2 model on Gujarati language. I would also google if there is a GPT2 model out there already trained on Gujarati.