Original implementation of the paper Adafold: Adapting folding trajectories of cloths via feedback-loop manipulation.

We encourage installing the repo with a conda environment.

conda create -n adafold python=3.9
git clone [email protected]:albiLo17/Adafold.git
cd Adafold
conda activate adafold

We use python3.9 and cuda 12.1 for our experiments.You can use the following commands to install the required cuda version and torch dependencies.

conda install cuda -c nvidia/label/cuda-12.1.0
pip install torch==2.2.0 torchvision --index-url https://download.pytorch.org/whl/cu121

For the installation of the torch_geometric dependencies, for more information refer to the official installation guide.

pip install torch_geometric
pip install pyg_lib torch_scatter torch_sparse torch_cluster torch_spline_conv -f https://data.pyg.org/whl/torch-2.2.0+cu121.html

For the remaining pip dependencies, you can install them using :

pip install -e .

To run the data collection for the folding task, run:

python data_collection.py

Refer to Adafold/dataset_collection/dataset_args.py for the available arguments for the data collection.

To train the dynamics model on the collected dataset, run:

python train_dynamics.py

Refer to Adafold/args/arguments.py for the available arguments for the data collection.

An example of how to plan with the learned model can be seen by running:

python planning.py

Please refer to the paper for more implementation details.