fixed crash bug in unigram model training

google · Jan 28, 2024 · 0fe7add · 0fe7add
1 parent 41c4b7f
commit 0fe7add
Showing 1 changed file with 1 addition and 1 deletion.
diff --git a/src/unigram_model_trainer.cc b/src/unigram_model_trainer.cc
@@ -229,7 +229,7 @@ TrainerModel::SentencePieces Trainer::MakeSeedSentencePiecesInternal() {
   for (node_int_type i = 0; i < node_num; ++i) {
     const node_int_type offset = SA[L[i]];
     const node_int_type len = D[i];
-    if (len <= 1) {
+    if (len <= 1 || offset >= array.size() || offset + len >= array.size()) {
       continue;
     }
     const char32 *begin = &array[offset];