# Requirements for model training
torch
transformers
datasets
peft
bitsandbytes
trl==0.8.6
accelerate
scipy
hf_transfer
rich