Tweets fine-tuning

How can I fine-tuning Arabic tweets with my custom datasets? no one of the available models run with unsupervised Arabic tweets text classification