Skip to content
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

[Backport 2.x] Integrates FAISS iterative builds with NativeEngines990KnnVectorsForm… #1992

Merged

Conversation

shatejas
Copy link
Collaborator

Original Approved PR #1950

…at (opensearch-project#1950)

* Iterative Vector Insertion (opensearch-project#1840)

* Rebased with new version of k-NN

Signed-off-by: Andrew Klepchick <[email protected]>

* Optimized faiss insertion

Signed-off-by: Andrew Klepchick <[email protected]>

* Optimized threadCount logic

Signed-off-by: Andrew Klepchick <[email protected]>

* Removed IDEA files

Signed-off-by: Andrew Klepchick <[email protected]>

* Removed unnecessary cmake file

Signed-off-by: Andrew Klepchick <[email protected]>

* Added comments to new functions

Signed-off-by: Andrew Klepchick <[email protected]>

* Removed createIndex and fixed test cases that use it

Signed-off-by: Andrew Klepchick <[email protected]>

* Removed unused code

Signed-off-by: Andrew Klepchick <[email protected]>

* Explained zero initialization for vector transfer

Signed-off-by: Andrew Klepchick <[email protected]>

* Added locale

Signed-off-by: Andrew Klepchick <[email protected]>

* Spotless Apply

Signed-off-by: Andrew Klepchick <[email protected]>

* Account for zero documents in finished batch

Signed-off-by: Andrew Klepchick <[email protected]>

* Changed where we check for zero docs

Signed-off-by: Andrew Klepchick <[email protected]>

* Changed tip for return

Signed-off-by: Andrew Klepchick <[email protected]>

* Use unique pointers to make sure resources are released on exception

Signed-off-by: Andrew Klepchick <[email protected]>

* Moved createIndex to testUtils

Signed-off-by: Andrew Klepchick <[email protected]>

* Fixed memory management so that the underlying index is not deleted after initialized

Signed-off-by: Andrew Klepchick <[email protected]>

* Created new KNNIndexBuilder graph to make index building more modular

Signed-off-by: Andrew Klepchick <[email protected]>

* Streamlined logic in KNNIndexBuilder.

Signed-off-by: Andrew Klepchick <[email protected]>

* Cleaned up unnecessary code in KNN80DocValuesConsumer

Signed-off-by: Andrew Klepchick <[email protected]>

* Fixed memory management process

Signed-off-by: Andrew Klepchick <[email protected]>

* Added note about index initialization in faiss_index_service

Signed-off-by: Andrew Klepchick <[email protected]>

* Accounted for case where the exception happens after the indexWriter is released.

Signed-off-by: Andrew Klepchick <[email protected]>

* Delete jni/src/.idea/modules.xml

Signed-off-by: Andrew Klepchick <[email protected]>

* Delete jni/src/.idea/vcs.xml

Signed-off-by: Andrew Klepchick <[email protected]>

* Delete jni/src/.idea/workspace.xml

Signed-off-by: Andrew Klepchick <[email protected]>

* Spotless apply and free iterative index on exception

Signed-off-by: Andrew Klepchick <[email protected]>

* Undid hack for checking first document metrics

Signed-off-by: Andrew Klepchick <[email protected]>

* Removed print statements

Signed-off-by: Andrew Klepchick <[email protected]>

* Free Vector Transfer on batch ingestion

Signed-off-by: Andrew Klepchick <[email protected]>

* Undid free

Signed-off-by: Andrew Klepchick <[email protected]>

* Fixed check for transfer ready

Signed-off-by: Andrew Klepchick <[email protected]>

* Don't crash when zero vectors inserted?

Signed-off-by: Andrew Klepchick <[email protected]>

* Reverted to old insertion process?

Signed-off-by: Andrew Klepchick <[email protected]>

* Spotless apply

Signed-off-by: Andrew Klepchick <[email protected]>

* Added back createOutput

Signed-off-by: Andrew Klepchick <[email protected]>

* Removed prior createOutput

Signed-off-by: Andrew Klepchick <[email protected]>

* Test remaking vectorTransfer

Signed-off-by: Andrew Klepchick <[email protected]>

* Test restructuring of insertion

Signed-off-by: Andrew Klepchick <[email protected]>

* Fixed case where vector address is immediately discarded

Signed-off-by: Andrew Klepchick <[email protected]>

* Spotless apply

Signed-off-by: Andrew Klepchick <[email protected]>

* Split Index Builder into multiple classes

Signed-off-by: Andrew Klepchick <[email protected]>

* Fixed descriptions of functions in faiss_index_service

Signed-off-by: Andrew Klepchick <[email protected]>

* Added back copyright files

Signed-off-by: Andrew Klepchick <[email protected]>

* Removed unused builder names

Signed-off-by: Andrew Klepchick <[email protected]>

* Modified tests to work with new insertion methods

Signed-off-by: Andrew Klepchick <[email protected]>

* Track index insertions

Signed-off-by: Andrew Klepchick <[email protected]>

* Tracked insertions for binary indices

Signed-off-by: Andrew Klepchick <[email protected]>

* Added back insertIds

Signed-off-by: Andrew Klepchick <[email protected]>

* Added check for freeVectorData to see if it works with an already deleted address

Signed-off-by: Andrew Klepchick <[email protected]>

* Cleaned up logs and comments in KNNIndexBuilder

Signed-off-by: Andrew Klepchick <[email protected]>

* Restructured the logic for KNNIndexBuilder

Signed-off-by: Andrew Klepchick <[email protected]>

* Changed package name of KNNIndexBuilder

Signed-off-by: Andrew Klepchick <[email protected]>

* Changed all package names and deleted unnecessary headers

Signed-off-by: Andrew Klepchick <[email protected]>

* Fixed for loop

Signed-off-by: Andrew Klepchick <[email protected]>

* Removed createIndex methods for faiss index service

Signed-off-by: Andrew Klepchick <[email protected]>

* Fixed package to fit naming conventions

Signed-off-by: Andrew Klepchick <[email protected]>

* Changed name of index builder

Signed-off-by: Andrew Klepchick <[email protected]>

* Spotless apply

Signed-off-by: Andrew Klepchick <[email protected]>

* Added comments to NativeIndexBuilder and restructured

Signed-off-by: Andrew Klepchick <[email protected]>

* Added deletion for memoryAddress

Signed-off-by: Andrew Klepchick <[email protected]>

* Spotless apply

Signed-off-by: Andrew Klepchick <[email protected]>

* Changed naming of classes to Writer and changed package name to fit conventions

Signed-off-by: Andrew Klepchick <[email protected]>

* Changed NativeIndexInfo and NativeVectorInfo to follow builder pattern

Signed-off-by: Andrew Klepchick <[email protected]>

* Added feature to changelog

Signed-off-by: Andrew Klepchick <[email protected]>

* Added class descriptions to each NativeIndexWriter

Signed-off-by: Andrew Klepchick <[email protected]>

* Changed name to getBytesPerVector

Signed-off-by: Andrew Klepchick <[email protected]>

* Added == false instead of ! for readability

Signed-off-by: Andrew Klepchick <[email protected]>

* Fixed changelog

Signed-off-by: Andrew Klepchick <[email protected]>

* Fixed naming in docvaluesconsumer

Signed-off-by: Andrew Klepchick <[email protected]>

* SpotlessApply

Signed-off-by: Andrew Klepchick <[email protected]>

* Made it so that we don't reuse testValues and removed a foot gun

Signed-off-by: Andrew Klepchick <[email protected]>

* Removed another foot gun in getIndexInfo

Signed-off-by: Andrew Klepchick <[email protected]>

* Fixed javadoc

Signed-off-by: Andrew Klepchick <[email protected]>

* Added deletion on exception cases

Signed-off-by: Andrew Klepchick <[email protected]>

* Removed unnecessary delete (NativeIndexWriter will handle deletion of vectors on exception)

Signed-off-by: Andrew Klepchick <[email protected]>

* Added correct logger and getWriter method to NativeIndexWriter

Signed-off-by: Andrew Klepchick <[email protected]>

* Ensured memory safety on JNI layer so that Java doesn't have to wrap everything in a try catch loop.

Signed-off-by: Andrew Klepchick <[email protected]>

* Refactored NativeIndexWriter and added comments to FaissService

Signed-off-by: Andrew Klepchick <[email protected]>

* Removed free in the JNIExport since index will always be freed in writeIndex.

Signed-off-by: Andrew Klepchick <[email protected]>

* Changed getVectorTransfer back to accept VectorDataType

Signed-off-by: Andrew Klepchick <[email protected]>

* Reverted free since not guaranteed to be IDMap.

Signed-off-by: Andrew Klepchick <[email protected]>

* Added all processes in addKNNBinaryField to NativeIndexWriter.createKNNIndex

Signed-off-by: Andrew Klepchick <[email protected]>

* Fixed javadoc

Signed-off-by: Andrew Klepchick <[email protected]>

* Applied spotless

Signed-off-by: Andrew Klepchick <[email protected]>

* Added back writeFooter

Signed-off-by: Andrew Klepchick <[email protected]>

* Removed threadCount fron writeIndex

Signed-off-by: Andrew Klepchick <[email protected]>

* Removed redundancies in KNN80DocValuesConsumer

Signed-off-by: Andrew Klepchick <[email protected]>

* Removed serializationMode

Signed-off-by: Andrew Klepchick <[email protected]>

* Fixed changelog

Signed-off-by: Andrew Klepchick <[email protected]>

* Fixed changelog

Signed-off-by: Andrew Klepchick <[email protected]>

* Removed double free test as we don't have to worry about that anymore

Signed-off-by: Andrew Klepchick <[email protected]>

* Accounted for HNSWSQ in index service

Signed-off-by: Andrew Klepchick <[email protected]>

* Removed delete in catch

Signed-off-by: Andrew Klepchick <[email protected]>

* Fixed faiss tests to work with writeIndex

Signed-off-by: Andrew Klepchick <[email protected]>

---------

Signed-off-by: Andrew Klepchick <[email protected]>

* Index Initialization Alloc Method (opensearch-project#1933)

* Added methods for allocating memory before inserting vectors to a faiss index

Signed-off-by: Andrew Klepchick <[email protected]>

* Fixed logic that gets type of index

Signed-off-by: Andrew Klepchick <[email protected]>

* Removed print statement

Signed-off-by: Andrew Klepchick <[email protected]>

* Removed unnecessary iostream

Signed-off-by: Andrew Klepchick <[email protected]>

* Removed flat index

Signed-off-by: Andrew Klepchick <[email protected]>

* Fixed flat index case

Signed-off-by: Andrew Klepchick <[email protected]>

* Fixed naming

Signed-off-by: Andrew Klepchick <[email protected]>

* Properly allocate HNSWSQ storage

Signed-off-by: Andrew Klepchick <[email protected]>

* Removed print statements

Signed-off-by: Andrew Klepchick <[email protected]>

* Fixed changelog

Signed-off-by: Andrew Klepchick <[email protected]>

* Removed unnecessary lib

Signed-off-by: Andrew Klepchick <[email protected]>

* Made alloc adaptive to different code sizes

Signed-off-by: Andrew Klepchick <[email protected]>

---------

Signed-off-by: Andrew Klepchick <[email protected]>

* Integrates FAISS iterative builds with NativeEngines990KnnVectorsFormat

Changes include reusing the same vector buffer in the JNI layer

Signed-off-by: Tejas Shah <[email protected]>

---------

Signed-off-by: Andrew Klepchick <[email protected]>
Signed-off-by: Tejas Shah <[email protected]>
Co-authored-by: Andrew Klepchick <[email protected]>
(cherry picked from commit fd59b9a)
jmazanec15
jmazanec15 previously approved these changes Aug 20, 2024
navneet1v
navneet1v previously approved these changes Aug 22, 2024
@navneet1v navneet1v merged commit 020ab1f into opensearch-project:2.x Aug 22, 2024
90 of 91 checks passed
@shatejas shatejas deleted the backport/backport-1950-to-2.x branch August 29, 2024 00:54
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants