Test BBQ recall with different bit sizes #129729
Draft
Conversation
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
…e-quantization-bits # Conflicts: # qa/vector/src/main/java/org/elasticsearch/test/knn/CmdLineArgs.java # server/src/main/java/org/elasticsearch/index/mapper/vectors/DenseVectorFieldMapper.java
benwtrent
reviewed
Jul 1, 2025
| this.byteBuffer = ByteBuffer.allocate(dimension); | ||
| this.binaryValue = byteBuffer.array(); | ||
| this.binaryQuantizer = quantizer; | ||
| this.discretizedDimensions = BQVectorUtils.discretize(dimension, 64); |
There was a problem hiding this comment.
Doesn't seem needed, you are just storing the raw bytes.
benwtrent
reviewed
Jul 1, 2025
Comment on lines
736
to
744
| OffHeapBinarizedQueryVectorValues(IndexInput data, int dimension, int size) { | ||
| this.slice = data; | ||
| this.dimension = dimension; | ||
| this.size = size; | ||
| // 4x the quantized binary dimensions | ||
| int binaryDimensions = (BQVectorUtils.discretize(dimension, 64) / 8) * BQSpaceUtils.B_QUERY; | ||
| this.byteBuffer = ByteBuffer.allocate(binaryDimensions); | ||
| this.binaryValue = byteBuffer.array(); | ||
| // + 1 for the quantized sum |
There was a problem hiding this comment.
I am not sure if this is used, but its incorrect as you aren't packing the bits, binaryDimensions is just dimensions.
…e option combinations
…ntization-bits' into non-issue/bbq-multiple-quantization-bits
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
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Outdated suggestions cannot be applied.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.
Related to #129421
Creates a new
ES910BinaryQuantizedVectorsFormatthat accepts variable bit sizes for indexing and querying.KnnIndexTester and related classes have been modified to allow testing with different bit sizes. Testing is possible via the following configuration file:
{ "doc_vectors" : "/path_to_vectors", "query_vectors" : "/path_to_queries", "num_docs" : 100000, "num_queries" : 1000, "index_type" : "flat", "over_sampling_factor" : 1, "search_threads" : 40, "index_threads" : 40, "reindex" : true, "force_merge" : true, "vector_space" : "cosine", "quantize_bits" : 1, "vector_encoding" : "float32", "dimensions" : 768, "use_new_flat_vectors_format": true, "quantize_query_bits" : 4 }Where:
quantize_bitsdetermine the quantization for vectors at index timequantize_query_bitsdetermine the quantization for queriesuse_new_flat_vectors_format: true to allow testing the variable bit sizes, false to fall back to the previous vector format (useful to compare quantize_bits:1 and quantize_query_bits: 4 to check the baseline)