Use AbstractBatchingProcessor for InferenceProcessor (#820)

* Use AbstractBatchingProcessor for InferenceProcessor Signed-off-by: Liyun Xiu <[email protected]> * Add chagnelog Signed-off-by: Liyun Xiu <[email protected]> --------- Signed-off-by: Liyun Xiu <[email protected]> (cherry picked from commit bf2fd5a)
opensearch-project · Jul 12, 2024 · fc9739b · fc9739b
1 parent ceee5a5
commit fc9739b
Show file tree

Hide file tree

Showing 17 changed files with 278 additions and 98 deletions.
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -16,6 +16,7 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 ### Features
 - Enable sorting and search_after features in Hybrid Search [#827](https://github.com/opensearch-project/neural-search/pull/827)
 ### Enhancements
+- InferenceProcessor inherits from AbstractBatchingProcessor to support sub batching in processor [#820](https://github.com/opensearch-project/neural-search/pull/820)
 - Adds dynamic knn query parameters efsearch and nprobes [#814](https://github.com/opensearch-project/neural-search/pull/814/)
 - Enable '.' for nested field in text embedding processor ([#811](https://github.com/opensearch-project/neural-search/pull/811))
 ### Bug Fixes

diff --git a/src/main/java/org/opensearch/neuralsearch/processor/InferenceProcessor.java b/src/main/java/org/opensearch/neuralsearch/processor/InferenceProcessor.java
@@ -29,7 +29,7 @@
 import org.opensearch.cluster.service.ClusterService;
 import org.opensearch.env.Environment;
 import org.opensearch.index.mapper.IndexFieldMapper;
-import org.opensearch.ingest.AbstractProcessor;
+import org.opensearch.ingest.AbstractBatchingProcessor;
 import org.opensearch.ingest.IngestDocument;
 import org.opensearch.ingest.IngestDocumentWrapper;
 import org.opensearch.neuralsearch.ml.MLCommonsClientAccessor;
@@ -46,7 +46,7 @@
  * and set the target fields according to the field name map.
  */
 @Log4j2
-public abstract class InferenceProcessor extends AbstractProcessor {
+public abstract class InferenceProcessor extends AbstractBatchingProcessor {
 
     public static final String MODEL_ID_FIELD = "model_id";
     public static final String FIELD_MAP_FIELD = "field_map";
@@ -69,6 +69,7 @@ public abstract class InferenceProcessor extends AbstractProcessor {
     public InferenceProcessor(
         String tag,
         String description,
+        int batchSize,
         String type,
         String listTypeNestedMapKey,
         String modelId,
@@ -77,7 +78,7 @@ public InferenceProcessor(
         Environment environment,
         ClusterService clusterService
     ) {
-        super(tag, description);
+        super(tag, description, batchSize);
         this.type = type;
         if (StringUtils.isBlank(modelId)) throw new IllegalArgumentException("model_id is null or empty, cannot process it");
         validateEmbeddingConfiguration(fieldMap);
@@ -144,7 +145,7 @@ public void execute(IngestDocument ingestDocument, BiConsumer<IngestDocument, Ex
     abstract void doBatchExecute(List<String> inferenceList, Consumer<List<?>> handler, Consumer<Exception> onException);
 
     @Override
-    public void batchExecute(List<IngestDocumentWrapper> ingestDocumentWrappers, Consumer<List<IngestDocumentWrapper>> handler) {
+    public void subBatchExecute(List<IngestDocumentWrapper> ingestDocumentWrappers, Consumer<List<IngestDocumentWrapper>> handler) {
         if (CollectionUtils.isEmpty(ingestDocumentWrappers)) {
             handler.accept(Collections.emptyList());
             return;

diff --git a/src/main/java/org/opensearch/neuralsearch/processor/SparseEncodingProcessor.java b/src/main/java/org/opensearch/neuralsearch/processor/SparseEncodingProcessor.java
@@ -31,13 +31,14 @@ public final class SparseEncodingProcessor extends InferenceProcessor {
     public SparseEncodingProcessor(
         String tag,
         String description,
+        int batchSize,
         String modelId,
         Map<String, Object> fieldMap,
         MLCommonsClientAccessor clientAccessor,
         Environment environment,
         ClusterService clusterService
     ) {
-        super(tag, description, TYPE, LIST_TYPE_NESTED_MAP_KEY, modelId, fieldMap, clientAccessor, environment, clusterService);
+        super(tag, description, batchSize, TYPE, LIST_TYPE_NESTED_MAP_KEY, modelId, fieldMap, clientAccessor, environment, clusterService);
     }
 
     @Override

diff --git a/src/main/java/org/opensearch/neuralsearch/processor/TextEmbeddingProcessor.java b/src/main/java/org/opensearch/neuralsearch/processor/TextEmbeddingProcessor.java
@@ -30,13 +30,14 @@ public final class TextEmbeddingProcessor extends InferenceProcessor {
     public TextEmbeddingProcessor(
         String tag,
         String description,
+        int batchSize,
         String modelId,
         Map<String, Object> fieldMap,
         MLCommonsClientAccessor clientAccessor,
         Environment environment,
         ClusterService clusterService
     ) {
-        super(tag, description, TYPE, LIST_TYPE_NESTED_MAP_KEY, modelId, fieldMap, clientAccessor, environment, clusterService);
+        super(tag, description, batchSize, TYPE, LIST_TYPE_NESTED_MAP_KEY, modelId, fieldMap, clientAccessor, environment, clusterService);
     }
 
     @Override

diff --git a/...in/java/org/opensearch/neuralsearch/processor/factory/SparseEncodingProcessorFactory.java b/...in/java/org/opensearch/neuralsearch/processor/factory/SparseEncodingProcessorFactory.java
@@ -14,7 +14,7 @@
 
 import org.opensearch.cluster.service.ClusterService;
 import org.opensearch.env.Environment;
-import org.opensearch.ingest.Processor;
+import org.opensearch.ingest.AbstractBatchingProcessor;
 import org.opensearch.neuralsearch.ml.MLCommonsClientAccessor;
 import org.opensearch.neuralsearch.processor.SparseEncodingProcessor;
 
@@ -24,27 +24,23 @@
  * Factory for sparse encoding ingest processor for ingestion pipeline. Instantiates processor based on user provided input.
  */
 @Log4j2
-public class SparseEncodingProcessorFactory implements Processor.Factory {
+public class SparseEncodingProcessorFactory extends AbstractBatchingProcessor.Factory {
     private final MLCommonsClientAccessor clientAccessor;
     private final Environment environment;
     private final ClusterService clusterService;
 
     public SparseEncodingProcessorFactory(MLCommonsClientAccessor clientAccessor, Environment environment, ClusterService clusterService) {
+        super(TYPE);
         this.clientAccessor = clientAccessor;
         this.environment = environment;
         this.clusterService = clusterService;
     }
 
     @Override
-    public SparseEncodingProcessor create(
-        Map<String, Processor.Factory> registry,
-        String processorTag,
-        String description,
-        Map<String, Object> config
-    ) throws Exception {
-        String modelId = readStringProperty(TYPE, processorTag, config, MODEL_ID_FIELD);
-        Map<String, Object> fieldMap = readMap(TYPE, processorTag, config, FIELD_MAP_FIELD);
-
-        return new SparseEncodingProcessor(processorTag, description, modelId, fieldMap, clientAccessor, environment, clusterService);
+    protected AbstractBatchingProcessor newProcessor(String tag, String description, int batchSize, Map<String, Object> config) {
+        String modelId = readStringProperty(TYPE, tag, config, MODEL_ID_FIELD);
+        Map<String, Object> fieldMap = readMap(TYPE, tag, config, FIELD_MAP_FIELD);
+
+        return new SparseEncodingProcessor(tag, description, batchSize, modelId, fieldMap, clientAccessor, environment, clusterService);
     }
 }
diff --git a/...ain/java/org/opensearch/neuralsearch/processor/factory/TextEmbeddingProcessorFactory.java b/...ain/java/org/opensearch/neuralsearch/processor/factory/TextEmbeddingProcessorFactory.java
@@ -14,14 +14,14 @@
 
 import org.opensearch.cluster.service.ClusterService;
 import org.opensearch.env.Environment;
-import org.opensearch.ingest.Processor;
+import org.opensearch.ingest.AbstractBatchingProcessor;
 import org.opensearch.neuralsearch.ml.MLCommonsClientAccessor;
 import org.opensearch.neuralsearch.processor.TextEmbeddingProcessor;
 
 /**
  * Factory for text embedding ingest processor for ingestion pipeline. Instantiates processor based on user provided input.
  */
-public class TextEmbeddingProcessorFactory implements Processor.Factory {
+public final class TextEmbeddingProcessorFactory extends AbstractBatchingProcessor.Factory {
 
     private final MLCommonsClientAccessor clientAccessor;
 
@@ -34,20 +34,16 @@ public TextEmbeddingProcessorFactory(
         final Environment environment,
         final ClusterService clusterService
     ) {
+        super(TYPE);
         this.clientAccessor = clientAccessor;
         this.environment = environment;
         this.clusterService = clusterService;
     }
 
     @Override
-    public TextEmbeddingProcessor create(
-        final Map<String, Processor.Factory> registry,
-        final String processorTag,
-        final String description,
-        final Map<String, Object> config
-    ) throws Exception {
-        String modelId = readStringProperty(TYPE, processorTag, config, MODEL_ID_FIELD);
-        Map<String, Object> filedMap = readMap(TYPE, processorTag, config, FIELD_MAP_FIELD);
-        return new TextEmbeddingProcessor(processorTag, description, modelId, filedMap, clientAccessor, environment, clusterService);
+    protected AbstractBatchingProcessor newProcessor(String tag, String description, int batchSize, Map<String, Object> config) {
+        String modelId = readStringProperty(TYPE, tag, config, MODEL_ID_FIELD);
+        Map<String, Object> filedMap = readMap(TYPE, tag, config, FIELD_MAP_FIELD);
+        return new TextEmbeddingProcessor(tag, description, batchSize, modelId, filedMap, clientAccessor, environment, clusterService);
     }
 }
diff --git a/...ava/org/opensearch/neuralsearch/processor/factory/TextImageEmbeddingProcessorFactory.java b/...ava/org/opensearch/neuralsearch/processor/factory/TextImageEmbeddingProcessorFactory.java
@@ -6,7 +6,6 @@
 
 import static org.opensearch.ingest.ConfigurationUtils.readMap;
 import static org.opensearch.ingest.ConfigurationUtils.readStringProperty;
-import static org.opensearch.neuralsearch.processor.TextEmbeddingProcessor.Factory;
 import static org.opensearch.neuralsearch.processor.TextImageEmbeddingProcessor.EMBEDDING_FIELD;
 import static org.opensearch.neuralsearch.processor.TextImageEmbeddingProcessor.FIELD_MAP_FIELD;
 import static org.opensearch.neuralsearch.processor.TextImageEmbeddingProcessor.MODEL_ID_FIELD;
@@ -16,6 +15,7 @@
 
 import org.opensearch.cluster.service.ClusterService;
 import org.opensearch.env.Environment;
+import org.opensearch.ingest.Processor;
 import org.opensearch.neuralsearch.ml.MLCommonsClientAccessor;
 import org.opensearch.neuralsearch.processor.TextImageEmbeddingProcessor;
 
@@ -25,31 +25,18 @@
  * Factory for text_image embedding ingest processor for ingestion pipeline. Instantiates processor based on user provided input.
  */
 @AllArgsConstructor
-public class TextImageEmbeddingProcessorFactory implements Factory {
+public class TextImageEmbeddingProcessorFactory implements Processor.Factory {
 
     private final MLCommonsClientAccessor clientAccessor;
     private final Environment environment;
     private final ClusterService clusterService;
 
     @Override
-    public TextImageEmbeddingProcessor create(
-        final Map<String, Factory> registry,
-        final String processorTag,
-        final String description,
-        final Map<String, Object> config
-    ) throws Exception {
-        String modelId = readStringProperty(TYPE, processorTag, config, MODEL_ID_FIELD);
-        String embedding = readStringProperty(TYPE, processorTag, config, EMBEDDING_FIELD);
-        Map<String, String> filedMap = readMap(TYPE, processorTag, config, FIELD_MAP_FIELD);
-        return new TextImageEmbeddingProcessor(
-            processorTag,
-            description,
-            modelId,
-            embedding,
-            filedMap,
-            clientAccessor,
-            environment,
-            clusterService
-        );
+    public Processor create(Map<String, Processor.Factory> processorFactories, String tag, String description, Map<String, Object> config)
+        throws Exception {
+        String modelId = readStringProperty(TYPE, tag, config, MODEL_ID_FIELD);
+        String embedding = readStringProperty(TYPE, tag, config, EMBEDDING_FIELD);
+        Map<String, String> filedMap = readMap(TYPE, tag, config, FIELD_MAP_FIELD);
+        return new TextImageEmbeddingProcessor(tag, description, modelId, embedding, filedMap, clientAccessor, environment, clusterService);
     }
 }
diff --git a/src/test/java/org/opensearch/neuralsearch/processor/InferenceProcessorTests.java b/src/test/java/org/opensearch/neuralsearch/processor/InferenceProcessorTests.java
@@ -4,6 +4,7 @@
  */
 package org.opensearch.neuralsearch.processor;
 
+import lombok.Getter;
 import org.junit.Before;
 import org.mockito.ArgumentCaptor;
 import org.mockito.MockitoAnnotations;
@@ -15,6 +16,7 @@
 import org.opensearch.ingest.IngestDocumentWrapper;
 import org.opensearch.neuralsearch.ml.MLCommonsClientAccessor;
 
+import java.util.ArrayList;
 import java.util.Arrays;
 import java.util.Collections;
 import java.util.List;
@@ -42,6 +44,7 @@ public class InferenceProcessorTests extends InferenceProcessorTestCase {
     private static final String DESCRIPTION = "description";
     private static final String MAP_KEY = "map_key";
     private static final String MODEL_ID = "model_id";
+    private static final int BATCH_SIZE = 10;
     private static final Map<String, Object> FIELD_MAP = Map.of("key1", "embedding_key1", "key2", "embedding_key2");
 
     @Before
@@ -54,7 +57,7 @@ public void setup() {
     }
 
     public void test_batchExecute_emptyInput() {
-        TestInferenceProcessor processor = new TestInferenceProcessor(createMockVectorResult(), null);
+        TestInferenceProcessor processor = new TestInferenceProcessor(createMockVectorResult(), BATCH_SIZE, null);
         Consumer resultHandler = mock(Consumer.class);
         processor.batchExecute(Collections.emptyList(), resultHandler);
         ArgumentCaptor<List<IngestDocumentWrapper>> captor = ArgumentCaptor.forClass(List.class);
@@ -65,7 +68,7 @@ public void test_batchExecute_emptyInput() {
 
     public void test_batchExecute_allFailedValidation() {
         final int docCount = 2;
-        TestInferenceProcessor processor = new TestInferenceProcessor(createMockVectorResult(), null);
+        TestInferenceProcessor processor = new TestInferenceProcessor(createMockVectorResult(), BATCH_SIZE, null);
         List<IngestDocumentWrapper> wrapperList = createIngestDocumentWrappers(docCount);
         wrapperList.get(0).getIngestDocument().setFieldValue("key1", Arrays.asList("", "value1"));
         wrapperList.get(1).getIngestDocument().setFieldValue("key1", Arrays.asList("", "value1"));
@@ -83,7 +86,7 @@ public void test_batchExecute_allFailedValidation() {
 
     public void test_batchExecute_partialFailedValidation() {
         final int docCount = 2;
-        TestInferenceProcessor processor = new TestInferenceProcessor(createMockVectorResult(), null);
+        TestInferenceProcessor processor = new TestInferenceProcessor(createMockVectorResult(), BATCH_SIZE, null);
         List<IngestDocumentWrapper> wrapperList = createIngestDocumentWrappers(docCount);
         wrapperList.get(0).getIngestDocument().setFieldValue("key1", Arrays.asList("", "value1"));
         wrapperList.get(1).getIngestDocument().setFieldValue("key1", Arrays.asList("value3", "value4"));
@@ -105,7 +108,7 @@ public void test_batchExecute_partialFailedValidation() {
     public void test_batchExecute_happyCase() {
         final int docCount = 2;
         List<List<Float>> inferenceResults = createMockVectorWithLength(6);
-        TestInferenceProcessor processor = new TestInferenceProcessor(inferenceResults, null);
+        TestInferenceProcessor processor = new TestInferenceProcessor(inferenceResults, BATCH_SIZE, null);
         List<IngestDocumentWrapper> wrapperList = createIngestDocumentWrappers(docCount);
         wrapperList.get(0).getIngestDocument().setFieldValue("key1", Arrays.asList("value1", "value2"));
         wrapperList.get(1).getIngestDocument().setFieldValue("key1", Arrays.asList("value3", "value4"));
@@ -126,7 +129,7 @@ public void test_batchExecute_happyCase() {
     public void test_batchExecute_sort() {
         final int docCount = 2;
         List<List<Float>> inferenceResults = createMockVectorWithLength(100);
-        TestInferenceProcessor processor = new TestInferenceProcessor(inferenceResults, null);
+        TestInferenceProcessor processor = new TestInferenceProcessor(inferenceResults, BATCH_SIZE, null);
         List<IngestDocumentWrapper> wrapperList = createIngestDocumentWrappers(docCount);
         wrapperList.get(0).getIngestDocument().setFieldValue("key1", Arrays.asList("aaaaa", "bbb"));
         wrapperList.get(1).getIngestDocument().setFieldValue("key1", Arrays.asList("cc", "ddd"));
@@ -158,7 +161,7 @@ public void test_batchExecute_sort() {
     public void test_doBatchExecute_exception() {
         final int docCount = 2;
         List<List<Float>> inferenceResults = createMockVectorWithLength(6);
-        TestInferenceProcessor processor = new TestInferenceProcessor(inferenceResults, new RuntimeException());
+        TestInferenceProcessor processor = new TestInferenceProcessor(inferenceResults, BATCH_SIZE, new RuntimeException());
         List<IngestDocumentWrapper> wrapperList = createIngestDocumentWrappers(docCount);
         wrapperList.get(0).getIngestDocument().setFieldValue("key1", Arrays.asList("value1", "value2"));
         wrapperList.get(1).getIngestDocument().setFieldValue("key1", Arrays.asList("value3", "value4"));
@@ -174,12 +177,36 @@ public void test_doBatchExecute_exception() {
         verify(clientAccessor).inferenceSentences(anyString(), anyList(), any());
     }
 
+    public void test_batchExecute_subBatches() {
+        final int docCount = 5;
+        List<List<Float>> inferenceResults = createMockVectorWithLength(6);
+        TestInferenceProcessor processor = new TestInferenceProcessor(inferenceResults, 2, null);
+        List<IngestDocumentWrapper> wrapperList = createIngestDocumentWrappers(docCount);
+        for (int i = 0; i < docCount; ++i) {
+            wrapperList.get(i).getIngestDocument().setFieldValue("key1", Collections.singletonList("value" + i));
+        }
+        List<IngestDocumentWrapper> allResults = new ArrayList<>();
+        processor.batchExecute(wrapperList, allResults::addAll);
+        for (int i = 0; i < docCount; ++i) {
+            assertEquals(allResults.get(i).getIngestDocument(), wrapperList.get(i).getIngestDocument());
+            assertEquals(allResults.get(i).getSlot(), wrapperList.get(i).getSlot());
+            assertEquals(allResults.get(i).getException(), wrapperList.get(i).getException());
+        }
+        assertEquals(3, processor.getAllInferenceInputs().size());
+        assertEquals(List.of("value0", "value1"), processor.getAllInferenceInputs().get(0));
+        assertEquals(List.of("value2", "value3"), processor.getAllInferenceInputs().get(1));
+        assertEquals(List.of("value4"), processor.getAllInferenceInputs().get(2));
+    }
+
     private class TestInferenceProcessor extends InferenceProcessor {
         List<?> vectors;
         Exception exception;
 
-        public TestInferenceProcessor(List<?> vectors, Exception exception) {
-            super(TAG, DESCRIPTION, TYPE, MAP_KEY, MODEL_ID, FIELD_MAP, clientAccessor, environment, clusterService);
+        @Getter
+        List<List<String>> allInferenceInputs = new ArrayList<>();
+
+        public TestInferenceProcessor(List<?> vectors, int batchSize, Exception exception) {
+            super(TAG, DESCRIPTION, batchSize, TYPE, MAP_KEY, MODEL_ID, FIELD_MAP, clientAccessor, environment, clusterService);
             this.vectors = vectors;
             this.exception = exception;
         }
@@ -196,6 +223,7 @@ public void doExecute(
         void doBatchExecute(List<String> inferenceList, Consumer<List<?>> handler, Consumer<Exception> onException) {
             // use to verify if doBatchExecute is called from InferenceProcessor
             clientAccessor.inferenceSentences(MODEL_ID, inferenceList, ActionListener.wrap(results -> {}, ex -> {}));
+            allInferenceInputs.add(inferenceList);
             if (this.exception != null) {
                 onException.accept(this.exception);
             } else {