-
Notifications
You must be signed in to change notification settings - Fork 1.2k
GroupVarInt Encoding Implementation for HNSW Graphs #14932
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
Merged
Merged
Changes from 3 commits
Commits
Show all changes
10 commits
Select commit
Hold shift + click to select a range
723dcf6
initial commit
aylonsk 0857c05
Merge branch 'refs/heads/main' into groupvarint_hnsw
aylonsk db86abf
added backwards compatibility implementation and testing
aylonsk ea9612b
Merge branch 'refs/heads/main' into groupvarint_hnsw
aylonsk 4fc0f24
Condensed testing to existing classes, removed top-level abstraction …
aylonsk c6fc06d
made version logic consistent
aylonsk ad58abf
fixed constructors, small details and nits
aylonsk a40b5d4
Merge branch 'main' into groupvarint_hnsw
aylonsk a47dd20
small changes to Lucene99HnswVectorsWriter constructors
aylonsk 02d7a35
corrected indentation on docstring (./gradlew tidy)
aylonsk File filter
Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
There are no files selected for viewing
233 changes: 233 additions & 0 deletions
233
...decs/src/test/org/apache/lucene/backward_codecs/lucene99/Lucene99HnswVectorsFormatV0.java
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,233 @@ | ||
| /* | ||
| * Licensed to the Apache Software Foundation (ASF) under one or more | ||
| * contributor license agreements. See the NOTICE file distributed with | ||
| * this work for additional information regarding copyright ownership. | ||
| * The ASF licenses this file to You under the Apache License, Version 2.0 | ||
| * (the "License"); you may not use this file except in compliance with | ||
| * the License. You may obtain a copy of the License at | ||
| * | ||
| * http://www.apache.org/licenses/LICENSE-2.0 | ||
| * | ||
| * Unless required by applicable law or agreed to in writing, software | ||
| * distributed under the License is distributed on an "AS IS" BASIS, | ||
| * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. | ||
| * See the License for the specific language governing permissions and | ||
| * limitations under the License. | ||
| */ | ||
|
|
||
| package org.apache.lucene.backward_codecs.lucene99; | ||
|
|
||
| import java.io.IOException; | ||
| import java.util.concurrent.ExecutorService; | ||
| import org.apache.lucene.codecs.KnnVectorsFormat; | ||
| import org.apache.lucene.codecs.KnnVectorsReader; | ||
| import org.apache.lucene.codecs.KnnVectorsWriter; | ||
| import org.apache.lucene.codecs.hnsw.FlatVectorScorerUtil; | ||
| import org.apache.lucene.codecs.hnsw.FlatVectorsFormat; | ||
| import org.apache.lucene.codecs.lucene99.Lucene99FlatVectorsFormat; | ||
| import org.apache.lucene.codecs.lucene99.Lucene99HnswVectorsReader; | ||
| import org.apache.lucene.index.MergePolicy; | ||
| import org.apache.lucene.index.MergeScheduler; | ||
| import org.apache.lucene.index.SegmentReadState; | ||
| import org.apache.lucene.index.SegmentWriteState; | ||
| import org.apache.lucene.search.TaskExecutor; | ||
| import org.apache.lucene.util.hnsw.HnswGraph; | ||
| import org.apache.lucene.util.hnsw.HnswGraphBuilder; | ||
|
|
||
| /** | ||
| * Lucene 9.9 vector format, which encodes numeric vector values into an associated graph connecting | ||
| * the documents having values. The graph is used to power HNSW search. The format consists of two | ||
| * files, and requires a {@link FlatVectorsFormat} to store the actual vectors: | ||
| * | ||
| * <h2>.vex (vector index)</h2> | ||
| * | ||
| * <p>Stores graphs connecting the documents for each field organized as a list of nodes' neighbours | ||
| * as following: | ||
| * | ||
| * <ul> | ||
| * <li>For each level: | ||
| * <ul> | ||
| * <li>For each node: | ||
| * <ul> | ||
| * <li><b>[vint]</b> the number of neighbor nodes | ||
| * <li><b>array[vint]</b> the delta encoded neighbor ordinals | ||
| * </ul> | ||
| * </ul> | ||
| * <li>After all levels are encoded, memory offsets for each node's neighbor nodes are appended to | ||
| * the end of the file. The offsets are encoded by {@link | ||
| * org.apache.lucene.util.packed.DirectMonotonicWriter}. | ||
| * </ul> | ||
| * | ||
| * <h2>.vem (vector metadata) file</h2> | ||
| * | ||
| * <p>For each field: | ||
| * | ||
| * <ul> | ||
| * <li><b>[int32]</b> field number | ||
| * <li><b>[int32]</b> vector similarity function ordinal | ||
| * <li><b>[vlong]</b> offset to this field's index in the .vex file | ||
| * <li><b>[vlong]</b> length of this field's index data, in bytes | ||
| * <li><b>[vint]</b> dimension of this field's vectors | ||
| * <li><b>[int]</b> the number of documents having values for this field | ||
| * <li><b>[vint]</b> the maximum number of connections (neighbours) that each node can have | ||
| * <li><b>[vint]</b> number of levels in the graph | ||
| * <li>Graph nodes by level. For each level | ||
| * <ul> | ||
| * <li><b>[vint]</b> the number of nodes on this level | ||
| * <li><b>array[vint]</b> for levels greater than 0 list of nodes on this level, stored as | ||
| * the level 0th delta encoded nodes' ordinals. | ||
| * </ul> | ||
| * </ul> | ||
| * | ||
| * @lucene.experimental | ||
| */ | ||
| public final class Lucene99HnswVectorsFormatV0 extends KnnVectorsFormat { | ||
|
|
||
| static final String META_CODEC_NAME = "Lucene99HnswVectorsFormatMeta"; | ||
| static final String VECTOR_INDEX_CODEC_NAME = "Lucene99HnswVectorsFormatIndex"; | ||
| static final String META_EXTENSION = "vem"; | ||
| static final String VECTOR_INDEX_EXTENSION = "vex"; | ||
|
|
||
| public static final int VERSION_START = 0; | ||
| public static final int VERSION_GROUPVARINT = 1; | ||
| public static final int VERSION_CURRENT = VERSION_GROUPVARINT; | ||
|
|
||
| /** | ||
| * A maximum configurable maximum max conn. | ||
| * | ||
| * <p>NOTE: We eagerly populate `float[MAX_CONN*2]` and `int[MAX_CONN*2]`, so exceptionally large | ||
| * numbers here will use an inordinate amount of heap | ||
| */ | ||
| public static final int MAXIMUM_MAX_CONN = 512; | ||
|
|
||
| /** Default number of maximum connections per node */ | ||
| public static final int DEFAULT_MAX_CONN = HnswGraphBuilder.DEFAULT_MAX_CONN; | ||
|
|
||
| /** | ||
| * The maximum size of the queue to maintain while searching during graph construction. This | ||
| * maximum value preserves the ratio of the `DEFAULT_BEAM_WIDTH`/`DEFAULT_MAX_CONN` (i.e. `6.25 * | ||
| * 16 = 3200`). | ||
| */ | ||
| public static final int MAXIMUM_BEAM_WIDTH = 3200; | ||
|
|
||
| /** | ||
| * Default number of the size of the queue maintained while searching during a graph construction. | ||
| */ | ||
| public static final int DEFAULT_BEAM_WIDTH = HnswGraphBuilder.DEFAULT_BEAM_WIDTH; | ||
|
|
||
| /** Default to use single thread merge */ | ||
| public static final int DEFAULT_NUM_MERGE_WORKER = 1; | ||
|
|
||
| static final int DIRECT_MONOTONIC_BLOCK_SHIFT = 16; | ||
|
|
||
| /** | ||
| * Controls how many of the nearest neighbor candidates are connected to the new node. Defaults to | ||
| * {@link org.apache.lucene.codecs.lucene99.Lucene99HnswVectorsFormat#DEFAULT_MAX_CONN}. See | ||
| * {@link HnswGraph} for more details. | ||
| */ | ||
| private final int maxConn; | ||
|
|
||
| /** | ||
| * The number of candidate neighbors to track while searching the graph for each newly inserted | ||
| * node. Defaults to {@link | ||
| * org.apache.lucene.codecs.lucene99.Lucene99HnswVectorsFormat#DEFAULT_BEAM_WIDTH}. See {@link | ||
| * HnswGraph} for details. | ||
| */ | ||
| private final int beamWidth; | ||
|
|
||
| /** The format for storing, reading, and merging vectors on disk. */ | ||
| private static final FlatVectorsFormat flatVectorsFormat = | ||
| new Lucene99FlatVectorsFormat(FlatVectorScorerUtil.getLucene99FlatVectorsScorer()); | ||
|
|
||
| private final int numMergeWorkers; | ||
| private final TaskExecutor mergeExec; | ||
|
|
||
| /** Constructs a format using default graph construction parameters */ | ||
| public Lucene99HnswVectorsFormatV0() { | ||
| this(DEFAULT_MAX_CONN, DEFAULT_BEAM_WIDTH, DEFAULT_NUM_MERGE_WORKER, null); | ||
| } | ||
|
|
||
| /** | ||
| * Constructs a format using the given graph construction parameters. | ||
| * | ||
| * @param maxConn the maximum number of connections to a node in the HNSW graph | ||
| * @param beamWidth the size of the queue maintained during graph construction. | ||
| */ | ||
| public Lucene99HnswVectorsFormatV0(int maxConn, int beamWidth) { | ||
| this(maxConn, beamWidth, DEFAULT_NUM_MERGE_WORKER, null); | ||
| } | ||
|
|
||
| /** | ||
| * Constructs a format using the given graph construction parameters and scalar quantization. | ||
| * | ||
| * @param maxConn the maximum number of connections to a node in the HNSW graph | ||
| * @param beamWidth the size of the queue maintained during graph construction. | ||
| * @param numMergeWorkers number of workers (threads) that will be used when doing merge. If | ||
| * larger than 1, a non-null {@link ExecutorService} must be passed as mergeExec | ||
| * @param mergeExec the {@link ExecutorService} that will be used by ALL vector writers that are | ||
| * generated by this format to do the merge. If null, the configured {@link | ||
| * MergeScheduler#getIntraMergeExecutor(MergePolicy.OneMerge)} is used. | ||
| */ | ||
| public Lucene99HnswVectorsFormatV0( | ||
| int maxConn, int beamWidth, int numMergeWorkers, ExecutorService mergeExec) { | ||
| super("Lucene99HnswVectorsFormat"); | ||
| if (maxConn <= 0 || maxConn > MAXIMUM_MAX_CONN) { | ||
| throw new IllegalArgumentException( | ||
| "maxConn must be positive and less than or equal to " | ||
| + MAXIMUM_MAX_CONN | ||
| + "; maxConn=" | ||
| + maxConn); | ||
| } | ||
| if (beamWidth <= 0 || beamWidth > MAXIMUM_BEAM_WIDTH) { | ||
| throw new IllegalArgumentException( | ||
| "beamWidth must be positive and less than or equal to " | ||
| + MAXIMUM_BEAM_WIDTH | ||
| + "; beamWidth=" | ||
| + beamWidth); | ||
| } | ||
| this.maxConn = maxConn; | ||
| this.beamWidth = beamWidth; | ||
| if (numMergeWorkers == 1 && mergeExec != null) { | ||
| throw new IllegalArgumentException( | ||
| "No executor service is needed as we'll use single thread to merge"); | ||
| } | ||
| this.numMergeWorkers = numMergeWorkers; | ||
| if (mergeExec != null) { | ||
| this.mergeExec = new TaskExecutor(mergeExec); | ||
| } else { | ||
| this.mergeExec = null; | ||
| } | ||
| } | ||
|
|
||
| @Override | ||
| public KnnVectorsWriter fieldsWriter(SegmentWriteState state) throws IOException { | ||
| return new Lucene99HnswVectorsWriterV0( | ||
| state, | ||
| maxConn, | ||
| beamWidth, | ||
| flatVectorsFormat.fieldsWriter(state), | ||
| numMergeWorkers, | ||
| mergeExec); | ||
| } | ||
|
|
||
| @Override | ||
| public KnnVectorsReader fieldsReader(SegmentReadState state) throws IOException { | ||
| return new Lucene99HnswVectorsReader(state, flatVectorsFormat.fieldsReader(state)); | ||
| } | ||
|
|
||
| @Override | ||
| public int getMaxDimensions(String fieldName) { | ||
| return 4096; | ||
| } | ||
|
|
||
| @Override | ||
| public String toString() { | ||
| return "Lucene99HnswVectorsFormat(name=Lucene99HnswVectorsFormat, maxConn=" | ||
| + maxConn | ||
| + ", beamWidth=" | ||
| + beamWidth | ||
| + ", flatVectorFormat=" | ||
| + flatVectorsFormat | ||
| + ")"; | ||
| } | ||
| } | ||
Oops, something went wrong.
Oops, something went wrong.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
So I think we are introducing these backward_codecs in order to be able to write graphs in the old VInt (v0) format in order to test that we are able to read them back with the existing codec?
Given that, I think any additional classes we add here could live in the tests rather than in backward_codecs, since we won't need these for reading old indexes (which is what backward_codecs are for).
Having said that, I wonder if we could add a test-only constructor to the format that would enable it to continue writing the old format?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Why don't we have a package private constructor that allows setting the version for the writer? That way tests can write with the old version if necessary?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Ah, I think that is what you mean by "test only ctor".
I agree, a new package private ctor is likely best and easiest
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I added a test-only constructor to the format class, eliminating the need for the "V0" testing files. However, this also required me to make a similar type of constructor in the writer class. If these changes feel excessive, @kaivalnp showed me a way to keep the version logic out of the format class and specify the writer version separately when initializing.