t5-en-de-translator

Model Card: T5 English ↔ German Translator

# ​ T5 English ↔ German Translator

This repository hosts a fine-tuned **T5 model** for **English ↔ German translation**. The model, training notebook, and interactive demo are maintained by [@chinesemusk](https://huggingface.co/chinesemusk).

---

##  Model Information

- **Architecture**: T5-small (Text-to-Text Transfer Transformer)  
- **Task**: English ↔ German Translation (seq2seq)  
- **Tokenizer**: SentencePiece (`spiece.model` + `tokenizer.json`)  
- **Training Code**: Available in this [Google Colab / GitHub notebook](https://github.com/Deon62/Eng-German-Translator-model/blob/main/translator.ipynb)  
- **Demo**: Interactive UI hosted via Gradio in my Hugging Face Space: [Kahnwald Translator Demo](https://huggingface.co/spaces/chinesemusk/Kahnwald)

---

##  Use the Model

Load and run translations with just a few lines:

```python
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM

model_id = "chinesemusk/t5-en-de-translator"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForSeq2SeqLM.from_pretrained(model_id)

text = "This is an example."
inputs = tokenizer(f"translate English to German: {text}", return_tensors="pt", truncation=True)
outputs = model.generate(**inputs, max_length=60)

print("EN:", text)
print("DE:", tokenizer.decode(outputs[0], skip_special_tokens=True))

---

Try It Live

Don't want to code? Try the model directly in your browser via this Gradio app:

Live Translator Demo

Enter text, select the direction (English → German or German → English), and get translations instantly.

---

Purpose & Limitations

• Purpose: Educational and prototyping usage—learn how translation fine-tuning works and test small-scale translation tasks.

• Limitations:

  • Fine-tuned on a small dataset slice — quality may vary on long or complex sentences.
  • Not designed for production-level accuracy or large-scale deployment.
  • Direction "German → English" works but may produce less accurate results since only lightly fine-tuned for that direction.

---

Acknowledgments

• Model built using Hugging Face transformers, datasets, and evaluate libraries.
• Huge thanks to the original T5 authors (Google Research).
• Demo powered by Gradio in a Hugging Face Space.

---

References

• Training Notebook: translator.ipynb on GitHub
• Gradio Demo Space: Kahnwald

---

---

###  Summary of Inclusions:

- Clear breakdown of model architecture and task.
- GitHub link to your code/notebook for transparency and reproducibility.
- Live demo link via Hugging Face Space for interactive testing.
- Usage snippet for quick adoption.
- Caveats and purpose for better user awareness.
- Proper acknowledgments and references.