from transformers import AutoModelForSeq2SeqLM, AutoTokenizer

model_name = "LuigiJoseph/Helsinki-ar-en-ft"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForSeq2SeqLM.from_pretrained(model_name)

text = "مرحبا بالعالم, كيف حالك؟"
inputs = tokenizer.encode(text, return_tensors="pt")
outputs = model.generate(inputs)
translation = tokenizer.decode(outputs[0], skip_special_tokens=True)

print(translation)  # English translations

Evaluation Metrics

Source language: Arabic ( ar )
Target language: English ( en )
Fine-tuned on: Custom dataset for e-commerce translations
Base model: Helsinki-NLP/opus-mt-ar-en

The performance of the base model and the fine-tuned model was measured using the BLEU metric:

Model	BLEU Score
Base Model (Helsinki-NLP/opus-mt-ar-en)	42.85
Fine-Tuned Model (this one)	80.79