README.md

---
license: mit
task_categories:
- reinforcement-learning
tags:
- imitation_learning
- gymnasium
- agents
pretty_name: Labyrinth 4x4 Dataset
size_categories:
- 1K<n<10K
---

# Labyrinth 5x5 Dataset

This dataset is for the gymnasium-base environment [Labyrinth-v0](https://github.com/NathanGavenski/maze-gym).
It contains `obs`, `actions`, `rewards`, `episodeed_starts` and `info` for each entry based on `100` train, `100` validation and `100` test different structures.

## Dataset Details

### Dataset Description

The dataset consists of the shortest route for each Labyrinth structure over `100` different structures for `3` different splits (`train`, `validation` and `test`).
Each entry cosists of:

```{python}
obs (str): path to the image of the observation.
action (int): the action for that observation (0-4).
reward (float): the reward for that observation.
episode_starts (bool): whether that observation was the initial timestep for that path.
info (str): the custom setting language to load that structure into the environment if needed.
```

The dataset has `4` different splits (`train`, `validation`, `test`, and `random`).
The `train`, `validation` and `test` are `100` different structures for training, validating and testing, respectively.
The `random` split is for imitation learning and offline inverse reinforcement learning methods that require random samples to start (i.e., BCO, IUPE, SAIL).

### Usage
Feel free to download all files from this dataset and use it as you please.
If you are interested in using our PyTorch Dataset implementation, feel free to check the [IL Datasets](https://github.com/NathanGavenski/IL-Datasets/blob/main/src/imitation_datasets/dataset/dataset.py) project.
There, we implement a base Dataset that downloads this dataset and all other datasets directly from HuggingFace.
The BaselineDataset also allows for more control over the splits and how many episodes you want to use (in cases where the `100` episodes aren't necessary).

## Citation

```
```