gradio>=3.0
transformers>=4.30
torch>=1.13
sentencepiece>=0.1.97
# optional but recommended for better model sharding/loading on some envs:
accelerate>=0.20