uses tokenization

3 triples
GPTKB property

Random triples
Subject Object
gptkb:m_BERT Word Piece
gptkb:XLM-Ro_BERTa Byte Pair Encoding
gptkb:GPT-1 Byte Pair Encoding