torch
fairscale
fire
sentencepiece
tqdm
matplotlib
seaborn
datasets==2.13.1
evaluate
transformers
wandb
sacrebleu
nltk
bert_score
rouge_score