mann-pytorch

The mann-pytorch project is a PyTorch implementation of the Mode-Adaptive Neural Networks (MANN) architecture, originally proposed in H. Zhang, S. Starke, T. Komura, and J. Saito, “Mode-adaptive neural networks for quadruped motion control,” ACM Trans. Graph., vol. 37, no. 4, pp. 1–11, 2018.

---

⚠️ REPOSITORY UNDER DEVELOPMENT ⚠️
We cannot guarantee stable API

---

💾 Installation on Ubuntu/Debian

Install python3, pip and venv via:

sudo apt-get install python3-pip python3-venv

Clone and install the repo in a virtual environment:

python3 -m venv mann-pytorch-env
source mann-pytorch-env/bin/activate
git clone https://github.com/ami-iit/mann-pytorch.git
cd mann-pytorch
pip install -r requirements.txt
pip install .

🚀 Usage

Training

You can execute a sample training script by:

cd mann-pytorch/scripts
python3 training.py

The training data will be periodically stored in a dedicated mann-pytorch/models/storage_<training_start_time> folder. You can also monitor the training progress by:

cd mann-pytorch/models/storage_<training_start_time>
python3 -m tensorboard.main --logdir=logs

Testing

You can execute a sample testing script by:

cd mann-pytorch/scripts
python3 testing.py

The average loss of the learned model on the testing dataset will be printed. Moreover, you will be able to inspect the learned model performances by comparing the ground truth and the predicted output on each instance of the testing dataset.

⚙️ Contributing

mann-pytorch is an open-source project, for which contributions are welcome. Open an issue with your feature request. Then, you can also proceed with a Pull-Request!