README.md

---
base_model: unsloth/meta-llama-3.1-8b-unsloth-bnb-4bit
tags:
- text-generation-inference
- transformers
- unsloth
- llama
- education
license: apache-2.0
language:
- en
---

# 🦙 Uploaded Finetuned Model – Llama 3.1 (8B) by Matteo Angeloni

- **Developed by:** [matteoangeloni](https://huggingface.co/matteoangeloni)  
- **License:** apache-2.0  
- **Base model:** [unsloth/meta-llama-3.1-8b-unsloth-bnb-4bit](https://huggingface.co/unsloth/meta-llama-3.1-8b-unsloth-bnb-4bit)  
- **Libraries used:** [Unsloth](https://github.com/unslothai/unsloth), Hugging Face TRL  

This model is my **first finetuned Llama model**, built for **educational and legal-domain text generation**.  
Training was accelerated with **Unsloth** (2x faster fine-tuning) and integrated with Hugging Face tools.

---

## 📚 Training Data

The model was trained on:

- **Dataset:** [louisbrulenaudet/code-education](https://huggingface.co/datasets/louisbrulenaudet/code-education)  
  → educational dataset for code-related instructions.

---

## 🎯 Intended Use

- Experimentation with **educational text generation**  
- Testing **instruction-following capabilities** in code/education-related contexts  
- Benchmarking performance of Unsloth-accelerated LLaMA models  

⚠️ **Not suitable for production**. This is an **experimental finetune**.

---

## 🚀 Example Usage

```python
from transformers import AutoModelForCausalLM, AutoTokenizer

model_name = "matteoangeloni/llama3-8b-edu"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name)

prompt = "Summarize the main points of the Italian privacy law."
inputs = tokenizer(prompt, return_tensors="pt")
outputs = model.generate(**inputs, max_new_tokens=200)

print(tokenizer.decode(outputs[0], skip_special_tokens=True))