transformers
torch
gradio
sentencepiece==0.1.95
datasets
seqeval
safetensors