Pytorch implementation of standard metrics for clustering. Tested on PyTorch = 1.10.1 / cuda11.3_cudnn8_0

So far only the Silhouette score is implemented. Code for the Silhouette score was developed in NumPy by Alexandre Abraham: https://gist.github.com/AlexandreAbraham/5544803

1. Open a terminal, navigate to the folder where you want to put the repository and clone it:

git clone https://github.com/maxschelski/pytorch-cluster-metrics.git

2. Navigate into the folder of the repository (pytorch-cluster-metrics):

cd pytorch-cluster-metrics

3. Install torchclustermetrics locally using pip:

pip install -e .

For any questions feel free to contact me via E-Mail to [email protected].

from torchclustermetrics import silhouette

score = silhouette.score(X, labels)