Closed safakkbilici closed 2 years ago
Currently, we only support displaying images. We will release a new version in this week, which supports showing text, or other user-customized information panels.
const feder = new Feder({
source: 'faiss',
filePath: ivfflatIndexFilePath,
viewParams: {
mediaType: 'img',
mediaCallback: () => null,
},
});
Set mediaCallback
to () => null
, and hover the node you are interested in, Feder will show the rowID and the clusterID.
If you want to know which vectors are in a specific cluster, you can try it.
// get the rowIds of vectors that belong to cluster-123.
feder.core.indexMeta.listIds[123] // "list" means "cluster"
thanks for the quick reply, i'm looking forward to the new release! is there any pythonic version of feder.core.indexMeta.listIds[123]
code snippet. I think this is for JS.
This is the internal data of federjs
. federpy
only supports generating visualizations and cannot access it directly.
Maybe you can try to use faiss to get it.
I want to visualize my text documents and so i did. However i didn't find any way to get docids in a specific cluster. For example, when I search over a random vector with feder, library gives me a bunch of clusters. Is it possible to obtain which vectors are stored in a specified cluster?
Actually what i exactly wanted and asked is that, in your example you are able to see images with fine search (distance) (first attached file). I want to show small texts instead of images (second attached file).