# 🧠 MarianMT-Text-Translation-AI-Model-"en-de"

A sequence-to-sequence translation model fine-tuned on English–German sentence pairs. This model translates English text into German and is built using the Hugging Face MarianMTModel. It’s suitable for general-purpose translation, language learning, and formal or semi-formal communication across English and German.

---

## ✨ Model Highlights

- 📌 Base Model: Helsinki-NLP/opus-mt-en-de
- 📚 Fine-tuned on a cleaned and tokenized parallel English-German dataset
- 🌍 Direction: English → German
- 🔧 Framework: Hugging Face Transformers + PyTorch

---

## 🧠 Intended Uses

- ✅ Translating English content (emails, documentation, support text) into German
- ✅ Use in educational platforms for learning German
- ✅ Supporting cross-lingual customer service, product documentation, or semi-formal communications

---

## 🚫 Limitations

- ❌ Not optimized for informal, idiomatic, or slang expressions
- ❌ Not ideal for legal, medical, or sensitive content translation
- 📏 Sentences longer than 128 tokens are truncated
- ⚠️ Domain-specific accuracy may vary (e.g., legal, technical)

---

## 🏋️‍♂️ Training Details

| Attribute          | Value                            |
|--------------------|----------------------------------|
| Base Model         | `Helsinki-NLP/opus-mt-en-de`     |
| Dataset            |  WMT14 English-German            |
| Task Type          | Translation                      |
| Max Token Length   | 128                              |
| Epochs             | 3                                |
| Batch Size         | 16                               |
| Optimizer          | AdamW                            |
| Loss Function      | CrossEntropyLoss                 |
| Framework          | PyTorch + Transformers           |
| Hardware           | CUDA-enabled GPU                 |

---

## 📊 Evaluation Metrics

| Metric     | Score   |
|------------|---------|
| BLEU Score | 30.42   |

---

## 🔎 Output Details

- Input: English text string 
- Output: Translated German text string

---

## 🚀 Usage

```python
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
import torch

model_name = "AventIQ-AI/Ai-Translate-Model-Eng-German"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForSeq2SeqLM.from_pretrained(model_name)
model.eval()

def translate(text):
    device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
    model.to(device)
    inputs = tokenizer(text, return_tensors="pt", padding=True, truncation=True).to(device)
    outputs = model.generate(**inputs)
    return tokenizer.decode(outputs[0], skip_special_tokens=True)

# Example
print(translate("How are you doing today?"))

```
---

## 📁 Repository Structure
```
finetuned-model/
├── config.json               ✅ Model architecture & config
├── pytorch_model.bin         ✅ Model weights
├── tokenizer_config.json     ✅ Tokenizer settings
├── tokenizer.json            ✅ Tokenizer vocabulary (JSON format)
├── source.spm                ✅ SentencePiece model for source language
├── target.spm                ✅ SentencePiece model for target language
├── special_tokens_map.json   ✅ Special tokens mapping
├── generation_config.json    ✅ (Optional) Generation defaults
├── README.md                 ✅ Model card

```

## 🤝 Contributing
Contributions are welcome! Feel free to open an issue or pull request to improve the model, training scripts, or documentation.