New vector_rescore parameter as a quantized index type option #124581
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
elasticsearchmachine added needs:triage 
benwtrent added >enhancement :Search Relevance/Vectors elasticsearchmachine added the Team:Search Relevance 
| Pinging @elastic/es-search-relevance (Team:Search Relevance) |
| Hi @benwtrent, I've created a changelog YAML for you. |
benwtrent added auto-backport elasticsearchmachine added the Team:Search Relevance …wtrent/elasticsearch into feature/add-rescore-to-index-options
Comment on lines 2245 to 2250
| Float oversample = indexOptions instanceof QuantizedIndexOptions quantizedIndexOptions | ||
| ? quantizedIndexOptions.rescoreVector != null ? quantizedIndexOptions.rescoreVector.oversample() : null | ||
| : null; | ||
| if (queryOversample != null) { | ||
| oversample = queryOversample; | ||
| } |
There was a problem hiding this comment.
Nit - I find hard to follow multiple ternary operators
Suggested change
| Float oversample = indexOptions instanceof QuantizedIndexOptions quantizedIndexOptions | |
| ? quantizedIndexOptions.rescoreVector != null ? quantizedIndexOptions.rescoreVector.oversample() : null | |
| : null; | |
| if (queryOversample != null) { | |
| oversample = queryOversample; | |
| } | |
| Float oversample = null; | |
| if (queryOversample != null) { | |
| oversample = queryOversample; | |
| } else if (queryOversample indexOptions instanceof QuantizedIndexOptions quantizedIndexOptions) { | |
| oversample = quantizedIndexOptions.rescoreVector != null ? quantizedIndexOptions.rescoreVector.oversample() : null | |
| } |
Co-authored-by: Carlos Delgado <6339205+carlosdelest@users.noreply.github.com>
carlosdelest approved these changes Mar 13, 2025
| ? quantizedIndexOptions.rescoreVector != null ? quantizedIndexOptions.rescoreVector.oversample() : null | ||
| : null; | ||
| Float oversample = null; | ||
| if (indexOptions instanceof QuantizedIndexOptions quantizedIndexOptions && quantizedIndexOptions.rescoreVector != null) { |
There was a problem hiding this comment.
Nit - Why go through this if branch if queryOversample is not null? It will be overriden in the next if.
benwtrent removed auto-backport 
| @carlosdelest I am restricting by index version. It came to mind that its possible to rolling upgrade, have a mixed cluster, then update a value to use the rescore parameter, but on an old index version. This would be bad as that index would no longer be readable on the old data nodes. |
Good call. Will keep in mind for future mapping options changes ✍️ |
…wtrent/elasticsearch into feature/add-rescore-to-index-options
benwtrent added the auto-merge-without-approval jfreden pushed a commit to jfreden/elasticsearch that referenced this pull request Mar 13, 2025
…tic#124581) This adds a new parameter to the quantized index mapping that allows default oversampling and rescoring to occur. This doesn't adjust any of the defaults. It allows it to be configured. When the user provides `rescore_vector: {oversample: <number>}` in the query it will overwrite it. For example, here is how to use it with bbq: ``` PUT rescored_bbq { "mappings": { "properties": { "vector": { "type": "dense_vector", "index_options": { "type": "bbq_hnsw", "rescore_vector": {"oversample": 3.0} } } } } } ``` Then, when querying, it will auto oversample the `k` by `3x` and rerank with the raw vectors. ``` POST _search { "knn": { "query_vector": [...], "field": "vector" } } ```
This was referenced Mar 18, 2025
This was referenced Apr 17, 2025
Closed
jimczi added a commit to jimczi/elasticsearch that referenced this pull request Apr 30, 2025
This PR is a partial backport of elastic#127285 that fixes the validation of the inference id when mappings are restored or dynamically updated. This change doesn't include defaulting semantic text dense vector to BBQ since it requires elastic#124581 to be backported first.
jimczi added a commit to jimczi/elasticsearch that referenced this pull request Apr 30, 2025
This PR is a partial backport of elastic#127285 that fixes the validation of the inference id when mappings are restored or dynamically updated. This change doesn't include defaulting semantic text dense vector to BBQ since it requires elastic#124581 to be backported first.
💚 All backports created successfully
Questions ?Please refer to the Backport tool documentation |
benwtrent added a commit to benwtrent/elasticsearch that referenced this pull request May 2, 2025
benwtrent added a commit that referenced this pull request May 2, 2025
benwtrent added a commit to benwtrent/elasticsearch that referenced this pull request May 2, 2025
benwtrent added a commit that referenced this pull request May 2, 2025
#124581) (#127644) * New `vector_rescore` parameter as a quantized index type option (#124581) This adds a new parameter to the quantized index mapping that allows default oversampling and rescoring to occur. This doesn't adjust any of the defaults. It allows it to be configured. When the user provides `rescore_vector: {oversample: <number>}` in the query it will overwrite it. For example, here is how to use it with bbq: ``` PUT rescored_bbq { "mappings": { "properties": { "vector": { "type": "dense_vector", "index_options": { "type": "bbq_hnsw", "rescore_vector": {"oversample": 3.0} } } } } } ``` Then, when querying, it will auto oversample the `k` by `3x` and rerank with the raw vectors. ``` POST _search { "knn": { "query_vector": [...], "field": "vector" } } ``` (cherry picked from commit b2c1c4e) * Adds new BWC version for 8.19 backport of (#124581) (#127647) * [CI] Auto commit changes from spotless --------- Co-authored-by: elasticsearchmachine <infra-root+elasticsearchmachine@elastic.co>
jfreden pushed a commit to jfreden/elasticsearch that referenced this pull request May 12, 2025
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.
This adds a new parameter to the quantized index mapping that allows default oversampling and rescoring to occur.
This doesn't adjust any of the defaults. It allows it to be configured. When the user provides
rescore_vector: {oversample: <number>}in the query it will overwrite it.For example, here is how to use it with bbq:
Then, when querying, it will auto oversample the
kby3xand rerank with the raw vectors.