Skip to content

[ML] Default endpoint allocations are now configurable #127783

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.

Sign up for GitHub

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

Closed
wants to merge 3 commits into from
Closed

[ML] Default endpoint allocations are now configurable #127783

Show file tree
Hide file tree
Changes from 2 commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
32517b6
[ML] Default endpoint allocations are now configurable
prwhelan May 6, 2025
4202d49
Update docs/changelog/127783.yaml
prwhelan May 6, 2025
972e309
Merge branch 'main' into ml/settings
prwhelan May 19, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
5 changes: 5 additions & 0 deletions docs/changelog/127783.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,5 @@
pr: 127783
summary: Default endpoint allocations are now configurable
area: Machine Learning
type: enhancement
issues: []
19 changes: 19 additions & 0 deletions ...ava/org/elasticsearch/xpack/core/ml/inference/assignment/AdaptiveAllocationsSettings.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -11,6 +11,7 @@
import org.elasticsearch.common.io.stream.StreamInput;
import org.elasticsearch.common.io.stream.StreamOutput;
import org.elasticsearch.common.io.stream.Writeable;
import org.elasticsearch.common.settings.Setting;
import org.elasticsearch.xcontent.ObjectParser;
import org.elasticsearch.xcontent.ParseField;
import org.elasticsearch.xcontent.ToXContentObject;
Expand All @@ -22,6 +23,24 @@

public class AdaptiveAllocationsSettings implements ToXContentObject, Writeable {

public static final Setting<Integer> DEFAULT_MIN_ALLOCATIONS = Setting.intSetting(
"xpack.ml.models.default_allocations.min",
0,
0,
32,
Setting.Property.Dynamic,
Copy link
Contributor

@jonathan-buttner jonathan-buttner May 7, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think dynamic will only be useful if we're going to allow changes without having to reboot. I think we typically add a addSettingsUpdateConsumer call so we can listen for the changes for example:

https://github.com/elastic/elasticsearch/blob/main/x-pack/plugin/inference/src/main/java/org/elasticsearch/xpack/inference/external/http/HttpSettings.java#L48

Setting.Property.NodeScope
);

public static final Setting<Integer> DEFAULT_MAX_ALLOCATIONS = Setting.intSetting(
"xpack.ml.models.default_allocations.max",
32,
1,
32,
Setting.Property.Dynamic,
Setting.Property.NodeScope
);

public static final AdaptiveAllocationsSettings RESET_PLACEHOLDER = new AdaptiveAllocationsSettings(false, -1, -1);

public static final ParseField ENABLED = new ParseField("enabled");
Expand Down
12 changes: 10 additions & 2 deletions .../elasticsearch/xpack/inference/services/elasticsearch/ElasticRerankerServiceSettings.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -9,19 +9,27 @@

import org.elasticsearch.common.ValidationException;
import org.elasticsearch.common.io.stream.StreamInput;
import org.elasticsearch.common.settings.Settings;
import org.elasticsearch.xpack.core.ml.inference.assignment.AdaptiveAllocationsSettings;

import java.io.IOException;
import java.util.Map;

import static org.elasticsearch.xpack.core.ml.inference.assignment.AdaptiveAllocationsSettings.DEFAULT_MAX_ALLOCATIONS;
import static org.elasticsearch.xpack.core.ml.inference.assignment.AdaptiveAllocationsSettings.DEFAULT_MIN_ALLOCATIONS;
import static org.elasticsearch.xpack.inference.services.elasticsearch.ElasticsearchInternalService.RERANKER_ID;

public class ElasticRerankerServiceSettings extends ElasticsearchInternalServiceSettings {

public static final String NAME = "elastic_reranker_service_settings";

public static ElasticRerankerServiceSettings defaultEndpointSettings() {
return new ElasticRerankerServiceSettings(null, 1, RERANKER_ID, new AdaptiveAllocationsSettings(Boolean.TRUE, 0, 32));
public static ElasticRerankerServiceSettings defaultEndpointSettings(Settings settings) {
return new ElasticRerankerServiceSettings(
null,
1,
RERANKER_ID,
new AdaptiveAllocationsSettings(Boolean.TRUE, DEFAULT_MIN_ALLOCATIONS.get(settings), DEFAULT_MAX_ALLOCATIONS.get(settings))
);
}

public ElasticRerankerServiceSettings(ElasticsearchInternalServiceSettings other) {
Expand Down
6 changes: 3 additions & 3 deletions ...rg/elasticsearch/xpack/inference/services/elasticsearch/ElasticsearchInternalService.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -923,22 +923,22 @@ private List<Model> defaultConfigs(boolean useLinuxOptimizedModel) {
DEFAULT_ELSER_ID,
TaskType.SPARSE_EMBEDDING,
NAME,
ElserInternalServiceSettings.defaultEndpointSettings(useLinuxOptimizedModel),
ElserInternalServiceSettings.defaultEndpointSettings(useLinuxOptimizedModel, settings),
ElserMlNodeTaskSettings.DEFAULT,
ChunkingSettingsBuilder.DEFAULT_SETTINGS
);
var defaultE5 = new MultilingualE5SmallModel(
DEFAULT_E5_ID,
TaskType.TEXT_EMBEDDING,
NAME,
MultilingualE5SmallInternalServiceSettings.defaultEndpointSettings(useLinuxOptimizedModel),
MultilingualE5SmallInternalServiceSettings.defaultEndpointSettings(useLinuxOptimizedModel, settings),
ChunkingSettingsBuilder.DEFAULT_SETTINGS
);
var defaultRerank = new ElasticRerankerModel(
DEFAULT_RERANK_ID,
TaskType.RERANK,
NAME,
ElasticRerankerServiceSettings.defaultEndpointSettings(),
ElasticRerankerServiceSettings.defaultEndpointSettings(settings),
RerankTaskSettings.DEFAULT_SETTINGS
);
return List.of(defaultElser, defaultE5, defaultRerank);
Expand Down
7 changes: 5 additions & 2 deletions ...rg/elasticsearch/xpack/inference/services/elasticsearch/ElserInternalServiceSettings.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -10,11 +10,14 @@
import org.elasticsearch.TransportVersion;
import org.elasticsearch.TransportVersions;
import org.elasticsearch.common.io.stream.StreamInput;
import org.elasticsearch.common.settings.Settings;
import org.elasticsearch.inference.MinimalServiceSettings;
import org.elasticsearch.xpack.core.ml.inference.assignment.AdaptiveAllocationsSettings;

import java.io.IOException;

import static org.elasticsearch.xpack.core.ml.inference.assignment.AdaptiveAllocationsSettings.DEFAULT_MAX_ALLOCATIONS;
import static org.elasticsearch.xpack.core.ml.inference.assignment.AdaptiveAllocationsSettings.DEFAULT_MIN_ALLOCATIONS;
import static org.elasticsearch.xpack.inference.services.elasticsearch.ElserModels.ELSER_V2_MODEL;
import static org.elasticsearch.xpack.inference.services.elasticsearch.ElserModels.ELSER_V2_MODEL_LINUX_X86;

Expand All @@ -26,12 +29,12 @@ public static MinimalServiceSettings minimalServiceSettings() {
return MinimalServiceSettings.sparseEmbedding(ElasticsearchInternalService.NAME);
}

public static ElserInternalServiceSettings defaultEndpointSettings(boolean useLinuxOptimizedModel) {
public static ElserInternalServiceSettings defaultEndpointSettings(boolean useLinuxOptimizedModel, Settings settings) {
return new ElserInternalServiceSettings(
null,
1,
useLinuxOptimizedModel ? ELSER_V2_MODEL_LINUX_X86 : ELSER_V2_MODEL,
new AdaptiveAllocationsSettings(Boolean.TRUE, 0, 32)
new AdaptiveAllocationsSettings(Boolean.TRUE, DEFAULT_MIN_ALLOCATIONS.get(settings), DEFAULT_MAX_ALLOCATIONS.get(settings))
);
}

Expand Down
7 changes: 5 additions & 2 deletions ...ch/xpack/inference/services/elasticsearch/MultilingualE5SmallInternalServiceSettings.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -9,6 +9,7 @@

import org.elasticsearch.common.ValidationException;
import org.elasticsearch.common.io.stream.StreamInput;
import org.elasticsearch.common.settings.Settings;
import org.elasticsearch.index.mapper.vectors.DenseVectorFieldMapper;
import org.elasticsearch.inference.MinimalServiceSettings;
import org.elasticsearch.inference.SimilarityMeasure;
Expand All @@ -18,6 +19,8 @@
import java.util.Arrays;
import java.util.Map;

import static org.elasticsearch.xpack.core.ml.inference.assignment.AdaptiveAllocationsSettings.DEFAULT_MAX_ALLOCATIONS;
import static org.elasticsearch.xpack.core.ml.inference.assignment.AdaptiveAllocationsSettings.DEFAULT_MIN_ALLOCATIONS;
import static org.elasticsearch.xpack.inference.services.elasticsearch.ElasticsearchInternalService.MULTILINGUAL_E5_SMALL_MODEL_ID;
import static org.elasticsearch.xpack.inference.services.elasticsearch.ElasticsearchInternalService.MULTILINGUAL_E5_SMALL_MODEL_ID_LINUX_X86;

Expand All @@ -37,12 +40,12 @@ public static MinimalServiceSettings minimalServiceSettings() {
);
}

public static MultilingualE5SmallInternalServiceSettings defaultEndpointSettings(boolean useLinuxOptimizedModel) {
public static MultilingualE5SmallInternalServiceSettings defaultEndpointSettings(boolean useLinuxOptimizedModel, Settings settings) {
return new MultilingualE5SmallInternalServiceSettings(
null,
1,
useLinuxOptimizedModel ? MULTILINGUAL_E5_SMALL_MODEL_ID_LINUX_X86 : MULTILINGUAL_E5_SMALL_MODEL_ID,
new AdaptiveAllocationsSettings(Boolean.TRUE, 0, 32)
new AdaptiveAllocationsSettings(Boolean.TRUE, DEFAULT_MIN_ALLOCATIONS.get(settings), DEFAULT_MAX_ALLOCATIONS.get(settings))
);
}

Expand Down
Loading