sentencepiece
transformers==4.20.0
torch==1.11.0
datasets==2.3.2
tokenizers==0.12.1
numpy<2