The NLP projects which I programmed and shared them open source.
1.NER A token classification to predict each token to ( O : None, Loc : location, Per : person, Org : organization) label. The model accepts Persian and English and performed 95% f1-score. You can check the model on huggingface page
2.Summarization A summarization model (mT5) is trained on XLsum dataset for Persian and English. The model trained for about 16 hours and gave the below result:
ROUGE-1 -> 29.83 ROUGE-2 -> 17.4 ROUGE-L -> 21.95
Note that model will perform better with training in more steps/epochs.