---
license: mit
task_categories:
- text-generation
language:
- de
- en
tags:
- gdscript
- godot
- code
- game-development
pretty_name: Godot LORA Dataset
size_categories:
- n<1K
---

# Godot LORA Dataset

GDScript training dataset for fine-tuning code models on Godot engine development.

## Dataset Info

- **Total samples:** 476
- **Train split:** 428
- **Validation split:** 48
- **Format:** JSONL (instruction, input, output)
- **Language:** GDScript (Godot 4.x)
- **Languages:** German instructions, GDScript code

## Sources

- godotengine/godot-demo-projects
- GDQuest/godot-open-rpg
- GDQuest/godot-3d-dodge-the-creeps
- bitbrain/beehave (behavior trees)
- limboai/limboai (AI for Godot)

## Sample Format

```json
{
  "instruction": "Erstelle einen Player-Controller in Godot 4",
  "input": "",
  "output": "extends CharacterBody3D\n\nconst SPEED = 5.0\n..."
}
```

## Usage

```python
from datasets import load_dataset

dataset = load_dataset("matzejo/godot-lora-dataset")
```

## Training

Use with Unsloth/PEFT for LORA fine-tuning:

```python
from unsloth import FastLanguageModel

model, tokenizer = FastLanguageModel.from_pretrained(
    "Qwen/Qwen2.5-Coder-7B-Instruct",
    load_in_4bit=True,
)

model = FastLanguageModel.get_peft_model(
    model,
    r=16,
    target_modules=["q_proj", "k_proj", "v_proj", "o_proj"],
)
```

## License

MIT License. Individual licenses from source repositories may apply to specific code samples.