The following bitsandbytes quantization config was used during training:

load_in_8bit: True 

Framework versions

PEFT 0.4.0.dev0