[8.17][ML] Set max allocations to 32 in default configs (#115518) (#115564)

* [ML] Set max allocations to 32 in default configs (#115518)

* fix test

---------

Co-authored-by: Elastic Machine <[email protected]>

Author: davidkyle

x-pack/plugin/inference/qa/inference-service-tests/src/javaRestTest/java/org/elasticsearch/xpack/inference/DefaultEndPointsIT.java

@@ -64,7 +64,7 @@ private static void assertDefaultElserConfig(Map<String, Object> modelConfig) {
         assertThat(
             modelConfig.toString(),
             adaptiveAllocations,
-            Matchers.is(Map.of("enabled", true, "min_number_of_allocations", 0, "max_number_of_allocations", 8))
+            Matchers.is(Map.of("enabled", true, "min_number_of_allocations", 0, "max_number_of_allocations", 32))
         );
     }
 
@@ -99,7 +99,7 @@ private static void assertDefaultE5Config(Map<String, Object> modelConfig) {
         assertThat(
             modelConfig.toString(),
             adaptiveAllocations,
-            Matchers.is(Map.of("enabled", true, "min_number_of_allocations", 0, "max_number_of_allocations", 8))
+            Matchers.is(Map.of("enabled", true, "min_number_of_allocations", 0, "max_number_of_allocations", 32))
         );
     }
 }

x-pack/plugin/inference/src/main/java/org/elasticsearch/xpack/inference/services/elasticsearch/ElasticsearchInternalService.java

@@ -870,7 +870,7 @@ private List<Model> defaultConfigs(boolean useLinuxOptimizedModel) {
                 null,
                 1,
                 useLinuxOptimizedModel ? ELSER_V2_MODEL_LINUX_X86 : ELSER_V2_MODEL,
-                new AdaptiveAllocationsSettings(Boolean.TRUE, 0, 8)
+                new AdaptiveAllocationsSettings(Boolean.TRUE, 0, 32)
             ),
             ElserMlNodeTaskSettings.DEFAULT,
             null // default chunking settings
@@ -883,7 +883,7 @@ private List<Model> defaultConfigs(boolean useLinuxOptimizedModel) {
                 null,
                 1,
                 useLinuxOptimizedModel ? MULTILINGUAL_E5_SMALL_MODEL_ID_LINUX_X86 : MULTILINGUAL_E5_SMALL_MODEL_ID,
-                new AdaptiveAllocationsSettings(Boolean.TRUE, 0, 8)
+                new AdaptiveAllocationsSettings(Boolean.TRUE, 0, 32)
             ),
             null // default chunking settings
         );