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
- 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
- Navigate into the folder of the repository (pytorch-cluster-metrics):
cd pytorch-cluster-metrics
- Install torchclustermetrics locally using pip:
pip install -e .
For any questions feel free to contact me via E-Mail to max.schelski@googlemail.com.
from torchclustermetrics import silhouette
score = silhouette.score(X, labels)