You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Please state you issue using the following template and, most importantly, in English.
Describe the bug
If there are multi segments, the result of IVF_PQ IP is wrong.
There is a bug in merging task result.
Steps/Code to reproduce behavior
(0) create collection whose metric_type is IP
(1) inset data and product multi segments
(2) create IVF_PQ
(3) query the topK nearest items
The result is wrong.
Expected behavior
The result is correct.
Environment details
Milvus version (v0.10.1)
Screenshots
none
Additional context
from source review
The text was updated successfully, but these errors were encountered:
When the total data in a collection is large enough, ie. multiple gigabytes, Milvus will organize the data in multiple slices, or "segments". For instance, 1 GB of vector data per segment. Then, Milvus will build index for each of the data segment.
Thus, if you have a collection containing 10 GB of vector data, and 1GB per segment, we will have 10 segments.
That is where the "multi-segments" come from.
Please state you issue using the following template and, most importantly, in English.
Describe the bug
If there are multi segments, the result of IVF_PQ IP is wrong.
There is a bug in merging task result.
Steps/Code to reproduce behavior
(0) create collection whose metric_type is IP
(1) inset data and product multi segments
(2) create IVF_PQ
(3) query the topK nearest items
The result is wrong.
Expected behavior
The result is correct.
Environment details
Screenshots
none
Additional context
from source review
The text was updated successfully, but these errors were encountered: