---
language:
- ar
- en
tags:
- translation
- darija
- algerian
- maghrebi
- nllb
base_model: facebook/nllb-200-1.3B
---

# 🇩🇿🇲🇦🇹🇳 Maghrebi Darija to English Translator (V3 Ultimate)

This model is a state-of-the-art fine-tuned version of facebook/nllb-200-1.3B, specifically designed to translate Maghrebi Arabic Dialects (Algerian, Moroccan, and Tunisian Darija) into English.

## 🚀 Model Details
- Base Model: NLLB-200-1.3B (Meta)
- Training Data: Over 120,000 highly curated parallel sentences, including daily conversations, social media slang, and deep cultural proverbs.
- Hardware Used: Dual T4 GPUs (Kaggle)
- Developer: Zakaria CH

## 🧠 Why this model?
Standard translation models often fail at Maghrebi dialects, usually defaulting to Modern Standard Arabic (MSA) literal translations. This model was explicitly trained to understand the *context* and *slang* of the Maghreb region (e.g., translating "قتلو" contextually as "I told him" instead of the literal "I killed him").

## 💻 How to use in Python

`python
from transformers import AutoModelForSeq2SeqLM, AutoTokenizer

model_name = "zakaCH/Algerian-Darija-Translator"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForSeq2SeqLM.from_pretrained(model_name)

text = "واش يا خويا كاش جديد في الحومة؟"
tokenizer.src_lang = "arb_Arab"
inputs = tokenizer(text, return_tensors="pt")

outputs = model.generate(**inputs, forced_bos_token_id=tokenizer.convert_tokens_to_ids("eng_Latn"), max_length=128)
print(tokenizer.batch_decode(outputs, skip_special_tokens=True)[0])
# Output: "What's up bro, any news in the neighborhood?"