robertgshaw2-neuralmagic/mistral-self-rag

Training mistral on self-rag task

sudo docker build -t llmf .

git clone https://github.com/mosaicml/llm-foundry.git

sudo docker run --rm --gpus all --shm-size 100gb --runtime nvidia -v $PWD/data:/data -v $PWD/experiments:/experiments -v $PWD/llm-foundry:/llm-foundry -e HF_HOME="/data" -it llmf

pip install -e ./llm-foundry

If model not downloaded, download it:

python3 experiments/edit_and_save_model.py

Launch

composer experiments/train.py experiments/yamls/run0.yaml