Make sure you have a local GPU to run vllm with, for now.
pyenv local 3.12.4
poetry env use $(pyenv which python)
poetry install --with dev --no-root
poetry shell
sudo -E bash dev/bootstrap.sh
poetry env use 3.12
poetry install --no-root
poetry shell
docker network create chutes
docker compose up -d
docker compose -f docker-compose-gpu.yml up -d vllm
docker compose -f docker-compose-gpu.yml up -d graval
If you do NOT want to verify GPUs/don't have GPU locally, be sure to set the SKIP_GPU_VERIFICATION
env variable to true
in the api service.