README.md

---
license: apache-2.0
language:
  - sv
library_name: transformers
base_model: Qwen/Qwen2.5-7B-Instruct
tags:
  - gguf
  - swedish
  - driving-license
  - körkort
  - fine-tuned
  - qwen2
pipeline_tag: text-generation
---

# Korkortlarare - Swedish Driving License Theory Assistant

A fine-tuned Qwen2.5-7B-Instruct model specialized for Swedish driving license theory (körkortsteori).

## Model Details

- **Base Model:** [Qwen2.5-7B-Instruct](https://huggingface.co/Qwen/Qwen2.5-7B-Instruct)
- **Fine-tuning Method:** LoRA with Unsloth
- **Training Data:** 4,080 Swedish driving license Q&A pairs
- **Format:** GGUF F16 (full precision)
- **Size:** 15.2 GB
- **Language:** Swedish (sv)

## Intended Use

This model is designed to help users study for the Swedish driving license theory test (körkortsteori). It can answer questions about:

- Traffic rules (Trafikregler)
- Road signs (Vägmärken)
- Safety (Säkerhet)
- Vehicle and environment (Fordon och Miljö)
- Parking and special situations
- Laws and regulations
- Practical driving

## Usage with Ollama

```bash
ollama run SlipsKnuten/korkortlarare
```

Or download and create locally:

```bash
ollama create korkortlarare -f Modelfile
```

## Usage with llama.cpp

```bash
./main -m korkortlarare-f16.gguf -p "Vad innebär högerregeln?"
```

## Training

The model was fine-tuned using:
- **Framework:** Unsloth + TRL
- **Method:** SFT (Supervised Fine-Tuning) with LoRA
- **Epochs:** 5
- **Dataset:** Custom Swedish driving license Q&A dataset

## Limitations

- The model is specifically trained for Swedish driving license theory and may not perform well on other topics
- Always verify important information with official sources (Trafikverket)
- This is an educational tool and should not replace official study materials

## License

Apache 2.0