[tokenizer] Return if exceed max token length (#2957)

Author: frankfliu

extensions/tokenizers/src/main/java/ai/djl/huggingface/tokenizers/Encoding.java

@@ -27,6 +27,7 @@ public class Encoding {
     private long[] specialTokenMask;
     private CharSpan[] charTokenSpans;
     private Encoding[] overflowing;
+    private boolean exceedMaxLength;
 
     protected Encoding(
             long[] ids,
@@ -36,6 +37,7 @@ protected Encoding(
             long[] attentionMask,
             long[] specialTokenMask,
             CharSpan[] charTokenSpans,
+            boolean exceedMaxLength,
             Encoding[] overflowing) {
         this.ids = ids;
         this.typeIds = typeIds;
@@ -44,6 +46,7 @@ protected Encoding(
         this.attentionMask = attentionMask;
         this.specialTokenMask = specialTokenMask;
         this.charTokenSpans = charTokenSpans;
+        this.exceedMaxLength = exceedMaxLength;
         this.overflowing = overflowing;
     }
 
@@ -127,6 +130,15 @@ public CharSpan[] getCharTokenSpans() {
         return charTokenSpans;
     }
 
+    /**
+     * Returns if tokens exceed max length.
+     *
+     * @return {@code true} if tokens exceed max length
+     */
+    public boolean exceedMaxLength() {
+        return exceedMaxLength;
+    }
+
     /**
      * Returns an array of overflowing encodings.
      *

extensions/tokenizers/src/main/java/ai/djl/huggingface/tokenizers/HuggingFaceTokenizer.java

@@ -530,10 +530,10 @@ private Encoding toEncoding(long encoding, boolean withOverflowingTokens) {
         long[] specialTokenMask = TokenizersLibrary.LIB.getSpecialTokenMask(encoding);
         CharSpan[] charSpans = TokenizersLibrary.LIB.getTokenCharSpans(encoding);
 
+        long[] overflowingHandles = TokenizersLibrary.LIB.getOverflowing(encoding);
+        boolean exceedMaxLength = overflowingHandles.length > 0;
         Encoding[] overflowing;
         if (withOverflowingTokens) {
-            long[] overflowingHandles = TokenizersLibrary.LIB.getOverflowing(encoding);
-
             overflowing = new Encoding[overflowingHandles.length];
             for (int i = 0; i < overflowingHandles.length; ++i) {
                 overflowing[i] = toEncoding(overflowingHandles[i], true);
@@ -551,6 +551,7 @@ private Encoding toEncoding(long encoding, boolean withOverflowingTokens) {
                 attentionMask,
                 specialTokenMask,
                 charSpans,
+                exceedMaxLength,
                 overflowing);
     }
 

extensions/tokenizers/src/test/java/ai/djl/huggingface/tokenizers/HuggingFaceTokenizerTest.java

@@ -330,7 +330,9 @@ public void testTruncationStride() throws IOException {
                         .build()) {
             String text = "Hello there my friend I am happy to see you";
             String textPair = "How are you my friend";
-            Encoding[] overflowing = tokenizer.encode(text, textPair).getOverflowing();
+            Encoding encoding = tokenizer.encode(text, textPair);
+            Assert.assertTrue(encoding.exceedMaxLength());
+            Encoding[] overflowing = encoding.getOverflowing();
 
             int expectedNumberOfOverflowEncodings = 7;
             Assert.assertEquals(overflowing.length, expectedNumberOfOverflowEncodings);