added presplit-sentences to scripts

scripts/pretrain_bert.sh

@@ -10,6 +10,7 @@ python pretrain_bert.py \
     --tokenizer-model-type bert-large-uncased \
     --vocab-size 30522 \
     --train-data wikipedia \
+    --presplit-sentences \
     --loose-json \
     --text-key text \
     --split 1000,1,1 \

scripts/pretrain_bert_distributed.sh

@@ -17,6 +17,7 @@ python -m torch.distributed.launch $DISTRIBUTED_ARGS \
     --tokenizer-model-type bert-large-uncased \
     --vocab-size 30522 \
     --train-data wikipedia \
+    --presplit-sentences \
     --loose-json \
     --text-key text \
     --split 1000,1,1 \

scripts/pretrain_bert_sentencepiece.sh

@@ -10,6 +10,7 @@ python pretrain_bert.py \
     --tokenizer-path tokenizer.model \
     --vocab-size 30522 \
     --train-data wikipedia \
+    --presplit-sentences \
     --loose-json \
     --text-key text \
     --split 1000,1,1 \