Edit model card

PersianEase

This model is fine-tuned to generate informal text from formal text based on the input provided. It has been fine-tuned on [Mohavere Dataset] (Takalli vahideh, Kalantari, Fateme, Shamsfard, Mehrnoush, Developing an Informal-Formal Persian Corpus, 2022.) using the pretrained model persian-t5-formality-transfer.

Evaluation Metrics

Metric Basic Model Base Persian T5 Previous Semester Model Our Model
BLEU-1 0.269 0.256 0.397 0.664
BLEU-2 0.137 0.171 0.299 0.539
BLEU-3 0.084 0.121 0.231 0.444
BLEU-4 0.054 0.086 0.177 0.364
Bert-Score Precision 0.581 0.583 0.665 0.826
Bert-Score Recall 0.629 0.614 0.659 0.820
Bert-Score F1 Score 0.603 0.595 0.658 0.822
ROUGE-1 F1 Score 0.259 - - 0.701
ROUGE-2 F1 Score 0.061 - - 0.475
ROUGE-l F1 Score 0.250 - - 0.675

Usage


from transformers import (T5ForConditionalGeneration, AutoTokenizer, pipeline)
import torch

model = T5ForConditionalGeneration.from_pretrained('parsi-ai-nlpclass/PersianEase')
tokenizer = AutoTokenizer.from_pretrained('parsi-ai-nlpclass/PersianEase')

pipe = pipeline(task='text2text-generation', model=model, tokenizer=tokenizer)
def test_model(text):
  device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')  
  model.to(device) 

  inputs = tokenizer.encode("formal: " + text, return_tensors='pt', max_length=128, truncation=True, padding='max_length')
  inputs = inputs.to(device) 

  outputs = model.generate(inputs, max_length=128, num_beams=4, temperature=0.7)
  print("Output:", tokenizer.decode(outputs[0], skip_special_tokens=True))  

text = "   من فقط می‌خواستم بگویم که چقدر قدردان همه چیزهایی هستم که برای من انجام داده ای."
print("Original:", text)
test_model(text)

# output: من فقط میخوام بگم که چقدر قدردان همه کاریم که برای من انجام دادی. دوستی تو برای من یه هدیه بزرگه و من همیشه از داشتن یه دوست مثل تو خوشحالم.

text = "   آرزویش است او را یک رستوران ببرم."
print("Original:", text)
test_model(text)

# output: آرزوشه یه رستوران ببرمش

Downloads last month
3
Safetensors
Model size
248M params
Tensor type
F32
·
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.