ryo-neko/example_docker_gpuenv

Usage of the docker environment with Cuda

For server users in our Lab.

Last update: 1st Mar. 2024.

Preparation

1. Create work directory at "."
2. Search your UID using id -u command.
3. In the docker-compose.yml, Change USERNAME, USERID(=UID), GROUPNAME, and GROUPID following your name and your ID.

      args:
        USERNAME: "test"
        USERID: "1000"
        GROUPNAME: "test" # it's ok [same as USERNAME]
        GROUPID: "1000" # it's ok [same as USERID]

4. In the docker-compose.uml, change "user".

        user: 1000:1000 # UID

5. Search YOUR GPU's UUID using nvidia-smi -L command.

→ nvidia-smi -L
GPU 0: NVIDIA A100 80GB PCIe (UUID: GPU-aaaaaaaaaaaaa)
  MIG 3g.40gb     Device  0: (UUID: MIG-bbbbbbbbbbbbb)
  MIG 1g.20gb     Device  1: (UUID: MIG-ccccccccccccc)
  MIG 1g.10gb     Device  2: (UUID: MIG-ddddddddddddd)

6. In the docker-compose.yml, fill NVIDIA_VISIBLE_DEVICES.

      - NVIDIA_VISIBLE_DEVICES={FILL YOUR UUID}

7. build and up using docker compose

docker compose build # build the image.
docker compose up -d # start the container.
docker compose attach dlenv # attach the container (/bin/bash)