Details
-
Bug
-
Status: Open (View Workflow)
-
Major
-
Resolution: Unresolved
-
11.8.5
-
None
-
Distributor ID: Ubuntu
Description: Ubuntu 22.04.5 LTS
Release: 22.04
Codename: jammy
Description
The number of rows returned with below query is not always the same
SELECT run_errorString_embeddings.run_id, VEC_DISTANCE_EUCLIDEAN(embedding, VEC_FromText(@myvec)) as distance FROM run_errorString_embeddings ORDER BY distance limit 1000; |
running this query multiple times gives us the following results:
495 rows in set (0.219 sec)
|
1000 rows in set (0.509 sec)
|
479 rows in set (0.252 sec)
|
501 rows in set (0.263 sec)
|
1000 rows in set (0.476 sec)
|
481 rows in set (0.244 sec)
|
the table format is created like below:
CREATE TABLE `run_errorString_embeddings` ( |
`run_id` INT(10) UNSIGNED NOT NULL, |
`attributes` LONGTEXT NOT NULL COLLATE 'utf8mb4_bin', |
`embedding` VECTOR(1024) NOT NULL, |
`vector_added` TINYINT(1) NOT NULL DEFAULT '0', |
PRIMARY KEY (`run_id`) USING BTREE, |
VECTOR INDEX `vect_embeddings` (`embedding`), |
CONSTRAINT `attributes` CHECK (json_valid(`attributes`)) |
)
|
COLLATE='utf8mb4_bin' |
ENGINE=InnoDB
|
;
|