torch<2.0.0, >=1.12.1 packaging==23.1 colossalai==0.3.2 autoflake==2.2.1 black==23.9.1 transformers tensorboard==2.14.0 six==1.16.0 datasets ninja==1.11.1 flash-attn>=2.0.0,<=2.0.5 tqdm sentencepiece==0.1.99 protobuf<=3.20.0
