facebookresearch / MetaCLIP

ICLR2024 Spotlight: curation/training code, metadata, distribution and pre-trained models for MetaCLIP; CVPR 2024: MoDE: CLIP Data Experts via Clustering

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

KMeans is not implemented in the original kmeans_pytorch

rezoo opened this issue · comments

Currently I am trying to reproduce MoDE in my own dataset.
Although it worked successfully up to feature extraction, problems arose with kmeans clustering.
Your repository describes that mode/prep_hrchy.py imports KMeans class from kmeans_pytorch, however, the original kmeans_pytorch does not support such a class.
I think that this library may be a modified one, but is this implementation published anywhere?

Thank you for your question. Regarding the KMeans installation, please refer the following command. The corresponding ReadMe will be updated shortly.

git clone https://github.com/kernelmachine/balanced-kmeans.git
cd balanced-kmeans
pip3 install -e .

More details of KMeans usage can be found here.