Multi vector support for Faiss HNSW - approximate search only #1371

heemin32 · 2024-01-03T21:58:26Z

Description

Apply the parentId filter to the Faiss HNSW search method. This ensures that documents are deduplicated based on their parentId, and the method returns k results for documents with nested fields.

Note: This is only for approximate search only. Exact search case will be handled in a follow up PR.

Issues Resolved

#1065

Check List

New functionality includes testing.
- All tests pass
New functionality has been documented.
- New functionality has javadoc added
Commits are signed as per the DCO using --signoff

By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.
For more information on following Developer Certificate of Origin and signing off your commits, please check here.

navneet1v · 2024-01-03T22:02:13Z

How we are handling the exact search use case here? I can only see that we have made changes for ANN search. But we also do exact search based on various conditions.

heemin32 · 2024-01-03T22:06:05Z

How we are handling the exact search use case here?

Update: Missed to handle the exact search case. Will update the PR.

Hmm. I thought that exact search has no issue with multi vector because it searches every data anyway. Let me verify it again though.

src/main/java/org/opensearch/knn/index/query/KNNWeight.java

src/test/java/org/opensearch/knn/jni/JNIServiceTests.java

src/test/java/org/opensearch/knn/index/NestedSearchIT.java

jmazanec15 · 2024-01-04T18:11:27Z

@heemin32 Can you look into failing CI?

heemin32 · 2024-01-04T18:16:56Z

@heemin32 Can you look into failing CI?

Not related with my change. Will retry.

jmazanec15 · 2024-01-04T21:47:06Z

@heemin32 actually, this looks like it might be related: https://github.com/opensearch-project/k-NN/actions/runs/7414868454/job/20176854501?pr=1371

heemin32 · 2024-01-04T22:07:25Z

@heemin32 actually, this looks like it might be related: https://github.com/opensearch-project/k-NN/actions/runs/7414868454/job/20176854501?pr=1371

Oops. Unused code checked in. Removed it.

jmazanec15

Approve, you may want to rebase branch with main to avoid flaky tests

codecov · 2024-01-04T22:25:16Z

Codecov Report

Attention: 1 lines in your changes are missing coverage. Please review.

Comparison is base (e64f762) 85.12% compared to head (4c05bd6) 85.18%.

Files	Patch %	Lines
...java/org/opensearch/knn/index/query/KNNWeight.java	93.33%	0 Missing and 1 partial ⚠️

Additional details and impacted files

@@                    Coverage Diff                     @@
##             feature/multi-vector    #1371      +/-   ##
==========================================================
+ Coverage                   85.12%   85.18%   +0.05%     
- Complexity                   1258     1261       +3     
==========================================================
  Files                         163      163              
  Lines                        5110     5115       +5     
  Branches                      479      479              
==========================================================
+ Hits                         4350     4357       +7     
+ Misses                        555      554       -1     
+ Partials                      205      204       -1

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

src/main/java/org/opensearch/knn/index/query/KNNQuery.java

src/test/java/org/opensearch/knn/jni/JNIServiceTests.java

src/main/java/org/opensearch/knn/index/query/KNNWeight.java

navneet1v

Try rebasing the feature branch and your forked repo with main branch as there are some conflicting changes being added to main branch.

jni/include/knn_extension/faiss/utils/Heap.h

jni/src/knn_extension/faiss/utils/BitSet.cpp

src/main/java/org/opensearch/knn/index/query/KNNWeight.java

src/test/java/org/opensearch/knn/index/NestedSearchIT.java

heemin32 · 2024-01-05T16:56:26Z

Try rebasing the feature branch and your forked repo with main branch as there are some conflicting changes being added to main branch.

Will do

src/test/java/org/opensearch/knn/index/NestedSearchIT.java

Apply the parentId filter to the Faiss HNSW search method. This ensures that documents are deduplicated based on their parentId, and the method returns k results for documents with nested fields. Signed-off-by: Heemin Kim <[email protected]>

* Add patch to support multi vector in faiss (#1358) Signed-off-by: Heemin Kim <[email protected]> * Initialize id_map as null (#1363) Signed-off-by: Heemin Kim <[email protected]> * Add support of multi vector in jni (#1364) Signed-off-by: Heemin Kim <[email protected]> * Multi vector support for Faiss HNSW (#1371) Apply the parentId filter to the Faiss HNSW search method. This ensures that documents are deduplicated based on their parentId, and the method returns k results for documents with nested fields. Signed-off-by: Heemin Kim <[email protected]> * Add data generation script for nested field (#1388) Signed-off-by: Heemin Kim <[email protected]> * Add perf test for nested field (#1394) Signed-off-by: Heemin Kim <[email protected]> --------- Signed-off-by: Heemin Kim <[email protected]>

Apply the parentId filter to the Faiss HNSW search method. This ensures that documents are deduplicated based on their parentId, and the method returns k results for documents with nested fields. Signed-off-by: Heemin Kim <[email protected]>

* Add patch to support multi vector in faiss (opensearch-project#1358) Signed-off-by: Heemin Kim <[email protected]> * Initialize id_map as null (opensearch-project#1363) Signed-off-by: Heemin Kim <[email protected]> * Add support of multi vector in jni (opensearch-project#1364) Signed-off-by: Heemin Kim <[email protected]> * Multi vector support for Faiss HNSW (opensearch-project#1371) Apply the parentId filter to the Faiss HNSW search method. This ensures that documents are deduplicated based on their parentId, and the method returns k results for documents with nested fields. Signed-off-by: Heemin Kim <[email protected]> * Add data generation script for nested field (opensearch-project#1388) Signed-off-by: Heemin Kim <[email protected]> * Add perf test for nested field (opensearch-project#1394) Signed-off-by: Heemin Kim <[email protected]> --------- Signed-off-by: Heemin Kim <[email protected]> (cherry picked from commit 709b448)

* Add patch to support multi vector in faiss (#1358) Signed-off-by: Heemin Kim <[email protected]> * Initialize id_map as null (#1363) Signed-off-by: Heemin Kim <[email protected]> * Add support of multi vector in jni (#1364) Signed-off-by: Heemin Kim <[email protected]> * Multi vector support for Faiss HNSW (#1371) Apply the parentId filter to the Faiss HNSW search method. This ensures that documents are deduplicated based on their parentId, and the method returns k results for documents with nested fields. Signed-off-by: Heemin Kim <[email protected]> * Add data generation script for nested field (#1388) Signed-off-by: Heemin Kim <[email protected]> * Add perf test for nested field (#1394) Signed-off-by: Heemin Kim <[email protected]> --------- Signed-off-by: Heemin Kim <[email protected]> (cherry picked from commit 709b448)

heemin32 added the skip-changelog label Jan 3, 2024

heemin32 requested review from navneet1v, VijayanB, vamshin, jmazanec15, naveentatikonda, junqiu-lei, martin-gaievski and ryanbogan as code owners January 3, 2024 21:58

heemin32 changed the title ~~Multi vector support for Faiss HNSW~~ Multi vector support for Faiss HNSW - approximate search only Jan 3, 2024

heemin32 force-pushed the parent-filter branch from a8e496d to f2b02af Compare January 3, 2024 23:57

jmazanec15 reviewed Jan 4, 2024

View reviewed changes

heemin32 force-pushed the parent-filter branch from f2b02af to 009f5be Compare January 4, 2024 19:23

heemin32 requested a review from jmazanec15 January 4, 2024 19:25

heemin32 force-pushed the parent-filter branch from 009f5be to 5c8b2af Compare January 4, 2024 21:10

jmazanec15 approved these changes Jan 4, 2024

View reviewed changes

jmazanec15 self-requested a review January 4, 2024 21:47

heemin32 force-pushed the parent-filter branch from 5c8b2af to af172f4 Compare January 4, 2024 22:06

jmazanec15 approved these changes Jan 4, 2024

View reviewed changes

heemin32 force-pushed the parent-filter branch 2 times, most recently from 7474efc to 4d2e517 Compare January 4, 2024 23:19

martin-gaievski reviewed Jan 5, 2024

View reviewed changes

navneet1v reviewed Jan 5, 2024

View reviewed changes

src/main/java/org/opensearch/knn/index/query/KNNWeight.java Outdated Show resolved Hide resolved

navneet1v reviewed Jan 5, 2024

View reviewed changes

src/main/java/org/opensearch/knn/index/query/KNNWeight.java Outdated Show resolved Hide resolved

heemin32 force-pushed the parent-filter branch from 4d2e517 to f1ae907 Compare January 5, 2024 05:11

navneet1v reviewed Jan 5, 2024

View reviewed changes

heemin32 force-pushed the parent-filter branch from f1ae907 to 1ac3f6d Compare January 5, 2024 17:31

martin-gaievski approved these changes Jan 5, 2024

View reviewed changes

heemin32 force-pushed the parent-filter branch from 1ac3f6d to e2f2cb5 Compare January 5, 2024 18:05

heemin32 force-pushed the feature/multi-vector branch from df6d1fa to e64f762 Compare January 5, 2024 18:10

heemin32 added skip-changelog and removed skip-changelog labels Jan 5, 2024

heemin32 force-pushed the parent-filter branch 2 times, most recently from 803e5b2 to d1ea68a Compare January 5, 2024 18:29

navneet1v reviewed Jan 5, 2024

View reviewed changes

src/test/java/org/opensearch/knn/index/NestedSearchIT.java Outdated Show resolved Hide resolved

heemin32 force-pushed the parent-filter branch from d1ea68a to e156363 Compare January 5, 2024 19:15

heemin32 requested a review from navneet1v January 5, 2024 19:19

heemin32 force-pushed the parent-filter branch from e156363 to 5f4cebe Compare January 5, 2024 19:23

Multi vector support for Faiss HNSW

4c05bd6

Apply the parentId filter to the Faiss HNSW search method. This ensures that documents are deduplicated based on their parentId, and the method returns k results for documents with nested fields. Signed-off-by: Heemin Kim <[email protected]>

heemin32 force-pushed the parent-filter branch from 5f4cebe to 4c05bd6 Compare January 5, 2024 19:32

navneet1v approved these changes Jan 5, 2024

View reviewed changes

heemin32 merged commit 0abed23 into opensearch-project:feature/multi-vector Jan 5, 2024
48 checks passed

heemin32 deleted the parent-filter branch January 5, 2024 19:58

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Multi vector support for Faiss HNSW - approximate search only #1371

Multi vector support for Faiss HNSW - approximate search only #1371

heemin32 commented Jan 3, 2024 •

edited

Loading

navneet1v commented Jan 3, 2024 •

edited

Loading

heemin32 commented Jan 3, 2024 •

edited

Loading

jmazanec15 commented Jan 4, 2024

heemin32 commented Jan 4, 2024

jmazanec15 commented Jan 4, 2024

heemin32 commented Jan 4, 2024

jmazanec15 left a comment

codecov bot commented Jan 4, 2024 •

edited

Loading

navneet1v left a comment

heemin32 commented Jan 5, 2024 •

edited

Loading

Multi vector support for Faiss HNSW - approximate search only #1371

Multi vector support for Faiss HNSW - approximate search only #1371

Conversation

heemin32 commented Jan 3, 2024 • edited Loading

Description

Issues Resolved

Check List

navneet1v commented Jan 3, 2024 • edited Loading

heemin32 commented Jan 3, 2024 • edited Loading

jmazanec15 commented Jan 4, 2024

heemin32 commented Jan 4, 2024

jmazanec15 commented Jan 4, 2024

heemin32 commented Jan 4, 2024

jmazanec15 left a comment

Choose a reason for hiding this comment

codecov bot commented Jan 4, 2024 • edited Loading

Codecov Report

navneet1v left a comment

Choose a reason for hiding this comment

heemin32 commented Jan 5, 2024 • edited Loading

heemin32 commented Jan 3, 2024 •

edited

Loading

navneet1v commented Jan 3, 2024 •

edited

Loading

heemin32 commented Jan 3, 2024 •

edited

Loading

codecov bot commented Jan 4, 2024 •

edited

Loading

heemin32 commented Jan 5, 2024 •

edited

Loading