-
-
Notifications
You must be signed in to change notification settings - Fork 510
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Wrong column added when alter table with additional vector column #2230
Comments
could you provide the stream of the create table or alter table statements that recreates the issue here locally? |
hi! create table create TABLE rt10 ( then subsequently alter table rt10 add column vector_zh float_vector knn_type='hnsw' knn_dims='768' hnsw_similarity='cosine'; then result of show create table rt10 CREATE TABLE rt10 ( if I create the index directly with 2 vector columns of different dimensions: create TABLE rt20 ( it then shows correctly when show create table rt20 CREATE TABLE rt20 ( is this what you need? |
MRE
Expected non-zero dimensions:
|
fixed missed KNN options for adding attribute at alter table at 69ca250 You need to update daemon package from the dev branch to get issue fixed. |
Great!
…On Tue, 18 Jun 2024 at 8:56 PM, Stas ***@***.***> wrote:
fixed missed KNN options for adding attribute at alter table at 69ca250
<69ca250>
You need to update daemon package from the dev branch to get issue fixed.
—
Reply to this email directly, view it on GitHub
<#2230 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/ABYQTYKA7UZSYNDPYRIMTMLZIAU63AVCNFSM6AAAAABII2FPRCVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDCNZWGAZTMNRXGQ>
.
You are receiving this because you authored the thread.Message ID:
***@***.***>
|
Bug Description:
in my current index, I already have a realtime index with the following vector column:
vector float_vector knn_type='hnsw' knn_dims='1024' hnsw_similarity='cosine'
This is fine if I am using one text embedding model to handle English but we also have chinese content. the English text-embedding model cannot handle chinese phrases therefore I tried on another chinese specific text embedding model. however, the English model generates 1024 dimensions of vectors which is good for the above column, the chinese model that I chose generates only 768 dimensions.
when I tried to use the same 1024 vector column for chinese text-embedding vector, I can't seems to insert or update/replace the record to refresh the chinese vector.
then I went on to add a column as below:
alter table myindex add column vector_zh float_vector knn_type='hnsw' knn_dims='768' hnsw_similarity='cosine';
column Is added, but result from show create table myindex seems weird:
vector float_vector knn_type='hnsw' knn_dims='1024' hnsw_similarity='COSINE',
vector_zh float_vector knn_type='hnsw' knn_dims='0' hnsw_similarity='L2'
note that knn_dims is 0 and hnsw_similarity=L2
which is not what I asked for.
if I create a new index, the vector columns correctly created:
vector float_vector knn_type='hnsw' knn_dims='1024' hnsw_similarity='COSINE',
vector_zh float_vector knn_type='hnsw' knn_dims='768' hnsw_similarity='COSINE'
Affected versions: 6.2.13 and 6.3.0
Manticore Search Version:
6.3.0
Operating System Version:
Ubuntu 22.04 server
Have you tried the latest development version?
Internal Checklist:
To be completed by the assignee. Check off tasks that have been completed or are not applicable.
The text was updated successfully, but these errors were encountered: