Skip to content

ci: CLT test for vector quantization #3626

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.

Sign up for GitHub

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

Open
wants to merge 1 commit into
base: master
Choose a base branch
from
Open

ci: CLT test for vector quantization #3626

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
325 changes: 325 additions & 0 deletions test/clt-tests/vector-knn/vector-quantization.rec
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,325 @@
Test vector quantization functionality with 8bit, 1bit, and 1bitsimple quantization types. Tests basic functionality, edge cases, and error conditions.

––– comment –––
Start Manticore Search with Buddy
––– block: ../base/start-searchd-with-buddy –––
––– comment –––
Test 8bit quantization
––– input –––
mysql -h0 -P9306 -e "create table test_8bit ( title text, image_vector float_vector knn_type='hnsw' knn_dims='4' hnsw_similarity='l2' quantization='8bit');"
––– output –––
––– input –––
mysql -h0 -P9306 -e "insert into test_8bit values ( 1, 'test vector 1', (0.653448,0.192478,0.017971,0.339821) ), ( 2, 'test vector 2', (-0.148894,0.748278,0.091892,-0.095406) ), ( 3, 'test vector 3', (0.286569,-0.031816,0.066684,0.032926) )"
––– output –––
––– input –––
mysql -h0 -P9306 -e "select id, knn_dist() from test_8bit where knn ( image_vector, 5, (0.286569,-0.031816,0.066684,0.032926) );"
––– output –––
+------+------------+
| id | knn_dist() |
+------+------------+
| 3 | 0.000000 |
| 1 | 0.28146550 |
| 2 | 0.81527930 |
+------+------------+
––– comment –––
Test 1bit quantization
––– input –––
mysql -h0 -P9306 -e "create table test_1bit ( title text, image_vector float_vector knn_type='hnsw' knn_dims='4' hnsw_similarity='l2' quantization='1bit');"
––– output –––
––– input –––
mysql -h0 -P9306 -e "insert into test_1bit values ( 1, 'test vector 1', (0.653448,0.192478,0.017971,0.339821) ), ( 2, 'test vector 2', (-0.148894,0.748278,0.091892,-0.095406) ), ( 3, 'test vector 3', (0.286569,-0.031816,0.066684,0.032926) );"
––– output –––
––– input –––
mysql -h0 -P9306 -e "select id, knn_dist() from test_1bit where knn ( image_vector, 5, (0.286569,-0.031816,0.066684,0.032926) );"
––– output –––
+------+------------+
| id | knn_dist() |
+------+------------+
| 3 | 0.000000 |
| 1 | 0.28146550 |
| 2 | 0.81527930 |
+------+------------+
––– input –––
mysql -h0 -P9306 -e "flush ramchunk test_1bit; select id, knn_dist() from test_1bit where knn ( image_vector, 5, (0.286569,-0.031816,0.066684,0.032926) );"
––– output –––
+------+------------+
| id | knn_dist() |
+------+------------+
| 3 | #!/\d+\.\d+/!# |
| 1 | #!/\d+\.\d+/!# |
| 2 | #!/\d+\.\d+/!# |
+------+------------+
––– comment –––
Test 1bitsimple quantization
––– input –––
mysql -h0 -P9306 -e "create table test_1bitsimple ( title text, image_vector float_vector knn_type='hnsw' knn_dims='4' hnsw_similarity='l2' quantization='1bitsimple');"
––– output –––
––– input –––
mysql -h0 -P9306 -e "insert into test_1bitsimple values ( 1, 'test vector 1', (0.653448,0.192478,0.017971,0.339821) ), ( 2, 'test vector 2', (-0.148894,0.748278,0.091892,-0.095406) ), ( 3, 'test vector 3', (0.286569,-0.031816,0.066684,0.032926) );"
––– output –––
––– input –––
mysql -h0 -P9306 -e "select id, knn_dist() from test_1bitsimple where knn ( image_vector, 5, (0.286569,-0.031816,0.066684,0.032926) );"
––– output –––
+------+------------+
| id | knn_dist() |
+------+------------+
| 3 | 0.000000 |
| 1 | 0.28146550 |
| 2 | 0.81527930 |
+------+------------+
––– input –––
mysql -h0 -P9306 -e "flush ramchunk test_1bitsimple; select id, knn_dist() from test_1bitsimple where knn ( image_vector, 5, (0.286569,-0.031816,0.066684,0.032926) );"
––– output –––
+------+------------+
| id | knn_dist() |
+------+------------+
| 3 | -4.000000 |
| 1 | -2.000000 |
| 2 | 2.000000 |
+------+------------+
––– comment –––
Test quantization with different similarity metrics
––– input –––
mysql -h0 -P9306 -e "create table test_cosine_8bit ( title text, image_vector float_vector knn_type='hnsw' knn_dims='4' hnsw_similarity='cosine' quantization='8bit');"
––– output –––
––– input –––
mysql -h0 -P9306 -e "insert into test_cosine_8bit values ( 1, 'test vector 1', (1.0,0.0,0.0,0.0) ), ( 2, 'test vector 2', (0.0,1.0,0.0,0.0) ), ( 3, 'test vector 3', (0.707,0.707,0.0,0.0) );"
––– output –––
––– input –––
mysql -h0 -P9306 -e "select id, knn_dist() from test_cosine_8bit where knn ( image_vector, 5, (1.0,0.0,0.0,0.0) );"
––– output –––
+------+------------+
| id | knn_dist() |
+------+------------+
| 1 | 0.000000 |
| 3 | 0.29289317 |
| 2 | 1.000000 |
+------+------------+
––– input –––
mysql -h0 -P9306 -e "flush ramchunk test_cosine_8bit; select id, knn_dist() from test_cosine_8bit where knn ( image_vector, 5, (1.0,0.0,0.0,0.0) );"
––– output –––
+------+-------------+
| id | knn_dist() |
+------+-------------+
| 1 | -0.00000012 |
| 3 | 0.29411757 |
| 2 | 1.000000 |
+------+-------------+
––– comment –––
Test edge cases - zero vectors
––– input –––
mysql -h0 -P9306 -e "create table test_zero_vectors ( title text, image_vector float_vector knn_type='hnsw' knn_dims='4' hnsw_similarity='l2' quantization='8bit');"
––– output –––
––– input –––
mysql -h0 -P9306 -e "insert into test_zero_vectors values ( 1, 'zero vector', (0.0,0.0,0.0,0.0) ), ( 2, 'small vector', (0.001,0.001,0.001,0.001) ), ( 3, 'normal vector', (1.0,1.0,1.0,1.0) );"
––– output –––
––– input –––
mysql -h0 -P9306 -e "select id, knn_dist() from test_zero_vectors where knn ( image_vector, 5, (0.0,0.0,0.0,0.0) );"
––– output –––
+------+------------+
| id | knn_dist() |
+------+------------+
| 1 | 0.000000 |
| 2 | 0.00000400 |
| 3 | 4.000000 |
+------+------------+
––– input –––
mysql -h0 -P9306 -e "flush ramchunk test_zero_vectors; select id, knn_dist() from test_zero_vectors where knn ( image_vector, 5, (0.0,0.0,0.0,0.0) );"
––– output –––
+------+------------+
| id | knn_dist() |
+------+------------+
| 1 | 0.000000 |
| 2 | 0.000000 |
| 3 | 0.25196466 |
+------+------------+
––– comment –––
Test edge cases - extreme values
––– input –––
mysql -h0 -P9306 -e "create table test_extreme_values ( title text, image_vector float_vector knn_type='hnsw' knn_dims='4' hnsw_similarity='l2' quantization='1bit');"
––– output –––
––– input –––
mysql -h0 -P9306 -e "insert into test_extreme_values values ( 1, 'large positive', (1000.0,1000.0,1000.0,1000.0) ), ( 2, 'large negative', (-1000.0,-1000.0,-1000.0,-1000.0) ), ( 3, 'mixed extreme', (1000.0,-1000.0,1000.0,-1000.0) );"
––– output –––
––– input –––
mysql -h0 -P9306 -e "select id, knn_dist() from test_extreme_values where knn ( image_vector, 5, (500.0,500.0,500.0,500.0) );"
––– output –––
+------+----------------+
| id | knn_dist() |
+------+----------------+
| 1 | 1000000.000000 |
| 3 | 5000000.000000 |
| 2 | 9000000.000000 |
+------+----------------+
––– comment –––
Test error cases - invalid quantization type
––– input –––
mysql -h0 -P9306 -e "create table test_invalid_quantization ( title text, image_vector float_vector knn_type='hnsw' knn_dims='4' hnsw_similarity='l2' quantization='invalid');"
––– output –––
ERROR 1064 (42000) at line 1: P03: Unknown quantization 'invalid' near ''invalid')'
––– comment –––
Test quantization with different vector dimensions
––– input –––
mysql -h0 -P9306 -e "create table test_high_dim ( title text, image_vector float_vector knn_type='hnsw' knn_dims='8' hnsw_similarity='l2' quantization='8bit');"
––– output –––
––– input –––
mysql -h0 -P9306 -e "insert into test_high_dim values ( 1, 'high dim vector', (0.1,0.2,0.3,0.4,0.5,0.6,0.7,0.8) );"
––– output –––
––– input –––
mysql -h0 -P9306 -e "select count(*) from test_high_dim;"
––– output –––
+----------+
| count(*) |
+----------+
| 1 |
+----------+
––– comment –––
Test quantization with single dimension
––– input –––
mysql -h0 -P9306 -e "create table test_single_dim ( title text, image_vector float_vector knn_type='hnsw' knn_dims='1' hnsw_similarity='l2' quantization='1bit');"
––– output –––
––– input –––
mysql -h0 -P9306 -e "insert into test_single_dim values ( 1, 'single dim 1', (0.5) ), ( 2, 'single dim 2', (-0.5) ), ( 3, 'single dim 3', (1.0) );"
––– output –––
––– input –––
mysql -h0 -P9306 -e "select id, knn_dist() from test_single_dim where knn ( image_vector, 5, (0.0) );"
––– output –––
+------+------------+
| id | knn_dist() |
+------+------------+
| 1 | 0.250000 |
| 2 | 0.250000 |
| 3 | 1.000000 |
+------+------------+
––– comment –––
Test quantization with IP (inner product) similarity
––– input –––
mysql -h0 -P9306 -e "create table test_ip_quantization ( title text, image_vector float_vector knn_type='hnsw' knn_dims='4' hnsw_similarity='ip' quantization='8bit');"
––– output –––
––– input –––
mysql -h0 -P9306 -e "insert into test_ip_quantization values ( 1, 'ip vector 1', (1.0,0.0,0.0,0.0) ), ( 2, 'ip vector 2', (0.0,1.0,0.0,0.0) ), ( 3, 'ip vector 3', (1.0,1.0,0.0,0.0) );"
––– output –––
––– input –––
mysql -h0 -P9306 -e "select id, knn_dist() from test_ip_quantization where knn ( image_vector, 5, (1.0,1.0,0.0,0.0) );"
––– output –––
+------+------------+
| id | knn_dist() |
+------+------------+
| 3 | -1.000000 |
| 1 | 0.000000 |
| 2 | 0.000000 |
+------+------------+
––– input –––
mysql -h0 -P9306 -e "flush ramchunk test_ip_quantization; select id, knn_dist() from test_ip_quantization where knn ( image_vector, 5, (1.0,1.0,0.0,0.0) );"
––– output –––
+------+-------------+
| id | knn_dist() |
+------+-------------+
| 3 | -1.00000024 |
| 1 | -0.00000012 |
| 2 | -0.00000012 |
+------+-------------+
––– comment –––
Verify table structure shows quantization parameter
––– input –––
mysql -h0 -P9306 -e "describe test_8bit;"
––– output –––
+--------------+--------------+----------------+
| Field | Type | Properties |
+--------------+--------------+----------------+
| id | bigint | |
| title | text | indexed stored |
| image_vector | float_vector | knn |
+--------------+--------------+----------------+
––– comment –––
Test quantization with rescoring and oversampling
––– input –––
mysql -h0 -P9306 -e "create table test_rescoring ( title text, image_vector float_vector knn_type='hnsw' knn_dims='4' hnsw_similarity='l2' quantization='1bit');"
––– output –––
––– input –––
mysql -h0 -P9306 -e "insert into test_rescoring values ( 1, 'vector 1', (0.1,0.2,0.3,0.4) ), ( 2, 'vector 2', (0.5,0.6,0.7,0.8) ), ( 3, 'vector 3', (0.9,1.0,1.1,1.2) ), ( 4, 'vector 4', (1.3,1.4,1.5,1.6) ), ( 5, 'vector 5', (1.7,1.8,1.9,2.0) );"
––– output –––
––– input –––
mysql -h0 -P9306 -e "select id, knn_dist() from test_rescoring where knn ( image_vector, 3, (0.5,0.6,0.7,0.8), {rescore=1} );"
––– output –––
+------+------------+
| id | knn_dist() |
+------+------------+
| 2 | 0.000000 |
| 3 | 0.640000 |
| 1 | 0.64000005 |
| 4 | 2.560000 |
| 5 | 5.75999975 |
+------+------------+
––– input –––
mysql -h0 -P9306 -e "flush ramchunk test_rescoring; select id, knn_dist() from test_rescoring where knn ( image_vector, 3, (0.5,0.6,0.7,0.8), {rescore=1} );"
––– output –––
+------+------------+
| id | knn_dist() |
+------+------------+
| 2 | 0.000000 |
| 3 | 0.640000 |
| 1 | 0.64000005 |
+------+------------+
––– comment –––
Test quantization with different ef parameters
––– input –––
mysql -h0 -P9306 -e "select id, knn_dist() from test_rescoring where knn ( image_vector, 3, (0.5,0.6,0.7,0.8), {ef=100} );"
––– output –––
+------+------------+
| id | knn_dist() |
+------+------------+
| 2 | 0.00000024 |
| 3 | 0.64000005 |
| 1 | 0.64020139 |
+------+------------+
––– comment –––
Test quantization memory usage comparison
––– input –––
mysql -h0 -P9306 -e "create table test_no_quantization ( title text, image_vector float_vector knn_type='hnsw' knn_dims='4' hnsw_similarity='l2');"
––– output –––
––– input –––
mysql -h0 -P9306 -e "insert into test_no_quantization values ( 1, 'no quant vector', (0.1,0.2,0.3,0.4) );"
––– output –––
––– input –––
mysql -h0 -P9306 -e "select id, knn_dist() from test_no_quantization where knn ( image_vector, 5, (0.1,0.2,0.3,0.4) );"
––– output –––
+------+------------+
| id | knn_dist() |
+------+------------+
| 1 | #!/[0-9]+\.[0-9]+/!# |
+------+------------+
––– input –––
mysql -h0 -P9306 -e "flush ramchunk test_no_quantization; select id, knn_dist() from test_no_quantization where knn ( image_vector, 5, (0.1,0.2,0.3,0.4) );"
––– output –––
+------+------------+
| id | knn_dist() |
+------+------------+
| 1 | 0.000000 |
+------+------------+
––– comment –––
Test quantization with normalized vectors
––– input –––
mysql -h0 -P9306 -e "create table test_normalized ( title text, image_vector float_vector knn_type='hnsw' knn_dims='4' hnsw_similarity='cosine' quantization='8bit');"
––– output –––
––– input –––
mysql -h0 -P9306 -e "insert into test_normalized values ( 1, 'normalized vector', (0.5,0.5,0.5,0.5) ), ( 2, 'unit vector', (1.0,0.0,0.0,0.0) );"
––– output –––
––– input –––
mysql -h0 -P9306 -e "select id, knn_dist() from test_normalized where knn ( image_vector, 5, (0.707,0.707,0.0,0.0) );"
––– output –––
+------+------------+
| id | knn_dist() |
+------+------------+
| 1 | 0.29289317 |
| 2 | 0.29289317 |
+------+------------+
––– input –––
mysql -h0 -P9306 -e "flush ramchunk test_normalized; select id, knn_dist() from test_normalized where knn ( image_vector, 5, (0.707,0.707,0.0,0.0) );"
––– output –––
+------+------------+
| id | knn_dist() |
+------+------------+
| 1 | 0.29134941 |
| 2 | 0.29411757 |
+------+------------+
Loading