Datasetpublic

Kalso42/WorldModelForMaze

WorldModelForMaze Code, datasets, and trained checkpoints for studying world-model representations in maze navigation, based on a modified NanoGPT. Contents *.py — training, testing, probing, and visualization scripts (see readme.md). model/ — architectures: transformer, transformer-rope, transformer-nextlat, mamba, mamba2, gated-deltanet, gru. data/maze/100/ — tokenized maze datasets for Tasks A/C/E/H/I (RWs paths, 100 nodes). out/ — final (10000-iter)… See the full description on the dataset page: https://huggingface.co/datasets/Kalso42/WorldModelForMaze.

sourceHugging Facemitupdated 3mo agoView on Hugging Face

0likes644downloads

rootREADME.md

README.md38 linesDownload Raw Back to root

1---2license: mit3task_categories:4  - other5tags:6  - maze7  - world-model8  - sequence-modeling9  - transformer10  - mamba11pretty_name: World Model For Maze12---13 14# WorldModelForMaze15 16Code, datasets, and trained checkpoints for studying world-model representations in maze navigation, based on a modified NanoGPT.17 18## Contents19 20- `*.py` — training, testing, probing, and visualization scripts (see `readme.md`).21- `model/` — architectures: transformer, transformer-rope, transformer-nextlat, mamba, mamba2, gated-deltanet, gru.22- `data/maze/100/` — tokenized maze datasets for Tasks A/C/E/H/I (RWs paths, 100 nodes).23- `out/` — final (10000-iter) checkpoints, `maze_kdetour` results, and plots.24 25> Intermediate (non-10000-iter) checkpoints (`out2/`) are not included here.26 27## Usage28 29See `readme.md` for full instructions on data generation, training, testing, probing, and the k-step detour analysis.30 31Quick start:32 33```bash34git clone https://huggingface.co/datasets/Kalso42/WorldModelForMaze35cd WorldModelForMaze36python train_maze.py --tasks C1 --path_type RWs --num_train_dataset 10M --config 6_6_38437```38