Immich•2y ago

Run Machine Learning container on remote machine

Hi. I was running immich on a raspberry pi 4 model B with raspbian arm64 via docker compose with ML on same machine with no problems but high CPU usage. I want to run the ML container on a more powerful remote machine. I have tailscale running on both rpi and remote machine. I have provided my trial setup and issue in comments. Please help me in solving the issue.

5 Replies

coding-doctorOP•2y ago

docker-compose file in remote machine:

services:
  immich-machine-learning:
    container_name: immich_machine_learning
    image: ghcr.io/immich-app/immich-machine-learning
    volumes:
      - ./immich_model-cache:/cache
    env_file:
      - .env
    ports:
      - 3003:3003
    restart: always

services:
  immich-machine-learning:
    container_name: immich_machine_learning
    image: ghcr.io/immich-app/immich-machine-learning
    volumes:
      - ./immich_model-cache:/cache
    env_file:
      - .env
    ports:
      - 3003:3003
    restart: always

I have removed the ml service in rpi compose. Added:

IMMICH_MACHINE_LEARNING_URL=http://<remote-server-ip>:3003

IMMICH_MACHINE_LEARNING_URL=http://<remote-server-ip>:3003

in .env of rpi I have tried with both tailscale ip and public ip after opening the firewall. But the ML remote container isn't producing any output or logs and the ML jobs on rpi are not being completed. How can I set it up correctly?

bo0tzz•2y ago

Have you made sure you can reach the remote machine from the rpi, just with ping/curl?

coding-doctorOP•2y ago

yes I can reach it

bo0tzz•2y ago

What's the result of curl http://<remote server>:3003/ping?

coding-doctorOP•2y ago

PING remote-ip (remote-ip) 56(84) bytes of data. 64 bytes from remote-ip: icmp_seq=1 ttl=50 time=233 ms 64 bytes from remote-ip: icmp_seq=2 ttl=50 time=232 ms 64 bytes from remote-ip: icmp_seq=3 ttl=50 time=232 ms ^X64 bytes from remote-ip: icmp_seq=4 ttl=50 time=232 ms ^C --- remote-ip ping statistics --- 4 packets transmitted, 4 received, 0% packet loss, time 3004ms rtt min/avg/max/mdev = 232.328/232.495/232.850/0.207 ms Issue resolved. It was an issue with correct version of the image to be pulled. Thanks @bo0tzz . Sorry for the trouble.

Gaming

Programming

Run Machine Learning container on remote machine

Did you find this page helpful?