Replies: 3 comments
-
please use Clustering::iteration_stats |
Beta Was this translation helpful? Give feedback.
-
I use I read the changelog and find the iteration_stats is added in 1.6.3, while my Faiss version is 1.5.1. Is this the reason that I cannot use it? My environment does not support conda installation, so I can only use faiss 1.5.1 instead of faiss-cpu or faiss-gpu. Could you pls tell me how this metric is computed? I read faiss/Clustering.cpp and guess it is kind of like accumulative square error between each sample and its centroid?
|
Beta Was this translation helpful? Give feedback.
-
Sorry that I misclose the issue, pls check my first reply. |
Beta Was this translation helpful? Give feedback.
-
Summary
Hello,
I have tried to use faiss on clustering problems with faiss.Clustering, and I want to measure the result by some metric. I notice that faiss.Kmeans has an attribute .obj in a notebook file of this repo, but I fail to get it with faiss.Clustering.
I am wondering if faiss.Clustering has this property or some other metrics? Or, could you please tell me how kmeans.obj is computed?
Thank you
Platform
OS:
Faiss version:
Installed from:
Faiss compilation options:
Running on:
Interface:
Beta Was this translation helpful? Give feedback.
All reactions