torch
transformers
numpy
pandas
tokenizers
sentencepiece
tqdm
datasets
scikit-learn
altair<5