---
base_model: unsloth/gemma-3-4b-it-unsloth-bnb-4bit
tags:
- text-generation-inference
- transformers
- unsloth
- gemma3
- dutch
- fine-tuned
license: apache-2.0
language:
- nl
---

# 🇳🇱 DUTCHGPT — Fine-Tuned Gemma 3 4B Model

- **Developed by:** [aacudad](https://github.com/aacudad) & [AkramChakrouni](https://github.com/AkramChakrouni)
- **License:** apache-2.0  
- **Base model:** [unsloth/gemma-3-4b-it-unsloth-bnb-4bit](https://huggingface.co/unsloth/gemma-3-4b-it-unsloth-bnb-4bit)

---

## 📚 Training Datasets

This model was fine-tuned on a diverse mix of high-quality **Dutch datasets** to support instruction following, legal reasoning, summarization, and conversational tasks:

- 🧾 [BramVanroy/ultrachat_200k_dutch](https://huggingface.co/datasets/BramVanroy/ultrachat_200k_dutch) — Dutch UltraChat conversational dataset  
- 🌐 [aacudad/8K_DUTCH_NEMOTRON_TRANSLATION](https://huggingface.co/datasets/aacudad/8K_DUTCH_NEMOTRON_TRANSLATION) — 8K translated instruction-style samples from NVIDIA's Nemotron dataset  
- ⚖️ [aacudad/5K_DUTCH_LEGAL_SUMMARY](https://huggingface.co/datasets/aacudad/5K_DUTCH_LEGAL_SUMMARY) — Legal cases with Gemini-generated summaries  
- 💬 [aacudad/86k_DUTCH_conversational](https://huggingface.co/datasets/aacudad/86k_DUTCH_conversational) — Instructional and chat-style Dutch conversations

---

## 🚀 Powered by Unsloth

This **Gemma 3 model** was trained **2x faster** using [Unsloth](https://github.com/unslothai/unsloth) along with Hugging Face’s TRL library.

[<img src="https://raw.githubusercontent.com/unslothai/unsloth/main/images/unsloth%20made%20with%20love.png" width="200"/>](https://github.com/unslothai/unsloth)

---

## 🔗 Resources

- 🧠 **GitHub (DutchGPT):** [github.com/aacudad/DutchGPT](https://github.com/aacudad/DutchGPT)  
- 🤗 **Hugging Face Profile:** [huggingface.co/aacudad](https://huggingface.co/aacudad)

Feel free to use the model in your own projects. If you do, I’d really appreciate a reference or tag — thanks! 🙌