My cheat-sheet:
!sudo apt-get install -y build-essential cmake
!pip install wandb
!pip install -U peft git+https://github.com/huggingface/trl accelerate optimum
!pip install numpy==1.26
!pip install tf-keras
!git clone https://github.com/bitsandbytes-foundation/bitsandbytes.git
!cd bitsandbytes/ && cmake -DCOMPUTE_BACKEND=cuda -S . && make -j && pip install --prefix=/home/ubuntu/.local/lib/python3.10/site-packages/bitsandbytes/ -e .
After this restart kernel.
Note numpy, manual bits and bytes.
After this training runs Ok!
Just posting here for you guys to be able to google it out.