[ML] Consistently apply the default truncation option for the BERT tokenizer (#80339)

The default is Truncate.First

docs/reference/ml/ml-shared.asciidoc

@@ -897,7 +897,7 @@ end::inference-config-classification-num-top-classes[]
 
 tag::inference-config-classification-num-top-feature-importance-values[]
 Specifies the maximum number of
-{ml-docs}/ml-feature-importance.html[{feat-imp}] values per document. Defaults 
+{ml-docs}/ml-feature-importance.html[{feat-imp}] values per document. Defaults
 to 0 which means no {feat-imp} calculation occurs.
 end::inference-config-classification-num-top-feature-importance-values[]
 
@@ -908,7 +908,7 @@ end::inference-config-classification-top-classes-results-field[]
 
 tag::inference-config-classification-prediction-field-type[]
 Specifies the type of the predicted field to write.
-Valid values are: `string`, `number`, `boolean`. When `boolean` is provided 
+Valid values are: `string`, `number`, `boolean`. When `boolean` is provided
 `1.0` is transformed to `true` and `0.0` to `false`.
 end::inference-config-classification-prediction-field-type[]
 
@@ -921,13 +921,13 @@ BERT-style tokenization is to be performed with the enclosed settings.
 end::inference-config-nlp-tokenization-bert[]
 
 tag::inference-config-nlp-tokenization-bert-do-lower-case[]
-Specifies if the tokenization lower case the text sequence when building the 
+Specifies if the tokenization lower case the text sequence when building the
 tokens.
 end::inference-config-nlp-tokenization-bert-do-lower-case[]
 
 tag::inference-config-nlp-tokenization-bert-truncate[]
 Indicates how tokens are truncated when they exceed `max_sequence_length`.
-The default value is `none`.
+The default value is `first`.
 +
 --
 * `none`: No truncation occurs; the inference request receives an error.
@@ -936,7 +936,7 @@ The default value is `none`.
 					 that sequence is truncated.
 --
 
-NOTE: For `zero_shot_classification`, the hypothesis sequence is always the second 
+NOTE: For `zero_shot_classification`, the hypothesis sequence is always the second
 sequence. Therefore, do not use `second` in this case.
 
 end::inference-config-nlp-tokenization-bert-truncate[]
@@ -951,21 +951,21 @@ Tokenize with special tokens. The tokens typically included in BERT-style tokeni
 end::inference-config-nlp-tokenization-bert-with-special-tokens[]
 
 tag::inference-config-nlp-tokenization-bert-max-sequence-length[]
-Specifies the maximum number of tokens allowed to be output by the tokenizer. 
+Specifies the maximum number of tokens allowed to be output by the tokenizer.
 The default for BERT-style tokenization is `512`.
 end::inference-config-nlp-tokenization-bert-max-sequence-length[]
 
 tag::inference-config-nlp-vocabulary[]
-The configuration for retreiving the vocabulary of the model. The vocabulary is 
-then used at inference time. This information is usually provided automatically 
+The configuration for retreiving the vocabulary of the model. The vocabulary is
+then used at inference time. This information is usually provided automatically
 by storing vocabulary in a known, internally managed index.
 end::inference-config-nlp-vocabulary[]
 
 tag::inference-config-nlp-fill-mask[]
-Configuration for a fill_mask natural language processing (NLP) task. The 
-fill_mask task works with models optimized for a fill mask action. For example, 
-for BERT models, the following text may be provided: "The capital of France is 
-[MASK].". The response indicates the value most likely to replace `[MASK]`. In 
+Configuration for a fill_mask natural language processing (NLP) task. The
+fill_mask task works with models optimized for a fill mask action. For example,
+for BERT models, the following text may be provided: "The capital of France is
+[MASK].". The response indicates the value most likely to replace `[MASK]`. In
 this instance, the most probable token is `paris`.
 end::inference-config-nlp-fill-mask[]
 
@@ -993,7 +993,7 @@ end::inference-config-text-classification[]
 tag::inference-config-text-embedding[]
 Text embedding takes an input sequence and transforms it into a vector of
 numbers. These embeddings capture not simply tokens, but semantic meanings and
-context. These embeddings can be used in a <<dense-vector,dense vector>> field 
+context. These embeddings can be used in a <<dense-vector,dense vector>> field
 for powerful insights.
 end::inference-config-text-embedding[]
 
@@ -1021,7 +1021,7 @@ it is possible to adjust the labels to classify. This makes this type of model
 and task exceptionally flexible.
 +
 --
-If consistently classifying the same labels, it may be better to use a 
+If consistently classifying the same labels, it may be better to use a
 fine-tuned text classification model.
 --
 end::inference-config-zero-shot-classification[]

x-pack/plugin/core/src/main/java/org/elasticsearch/xpack/core/ml/inference/trainedmodel/Tokenization.java

@@ -47,7 +47,7 @@ public abstract class Tokenization implements NamedXContentObject, NamedWriteabl
     private static final int DEFAULT_MAX_SEQUENCE_LENGTH = 512;
     private static final boolean DEFAULT_DO_LOWER_CASE = false;
     private static final boolean DEFAULT_WITH_SPECIAL_TOKENS = true;
-    private static final Truncate DEFAULT_TRUNCATION = Truncate.NONE;
+    private static final Truncate DEFAULT_TRUNCATION = Truncate.FIRST;
 
     static <T extends Tokenization> void declareCommonFields(ConstructingObjectParser<T, ?> parser) {
         parser.declareBoolean(ConstructingObjectParser.optionalConstructorArg(), DO_LOWER_CASE);
@@ -57,7 +57,7 @@ public abstract class Tokenization implements NamedXContentObject, NamedWriteabl
     }
 
     public static BertTokenization createDefault() {
-        return new BertTokenization(null, null, null, Truncate.FIRST);
+        return new BertTokenization(null, null, null, Tokenization.DEFAULT_TRUNCATION);
     }
 
     protected final boolean doLowerCase;