Frigate app: Fails to restart after config save

Apps and Virtualization
22

While waiting on a fix, I switched to a custom YAML app, which solved the issue – and a few other things only seemed to work this way per the docs.

Context: I’m on an AMD GPU. Initially I wasn’t using it for detection, but the docs recommend the stable-rocm image instead of the default. After switching, I also discovered the community app doesn’t map /dev/kfd, and without it hardware acceleration for detection won’t work.

In the end, I deployed via a custom Compose file:

services:
  frigate:
    container_name: frigate
    privileged: false
    restart: unless-stopped
    stop_grace_period: 30s
    image: ghcr.io/blakeblackshear/frigate:stable-rocm
    shm_size: "512mb"
    devices:
      - /dev/dri
      - /dev/kfd
    volumes:
      - /etc/localtime:/etc/localtime:ro
      - /mnt/SSDs/Applications/frigate/config:/config
      - /mnt/HDDs/Applications/frigate/media:/media/frigate
      - type: tmpfs
        target: /tmp/cache
        tmpfs:
          size: 1000000000
    ports:
      - "30193:8971"
      - "8554:8554"
      - "8555:8555/tcp"
      - "8555:8555/udp"
    environment:
      FRIGATE_RTSP_PASSWORD: "password"
      HSA_OVERRIDE_GFX_VERSION: "11.0.0"
x-portals:
  - host: 0.0.0.0
    name: Web UI
    path: /
    port: 30193
    scheme: https

I also added a few config tweaks:

detectors:
  onnx:
    type: onnx

model:
  model_type: yolo-generic
  width: 320
  height: 320
  input_tensor: nchw
  input_dtype: float
  path: /config/model_cache/yolov9-m.onnx
  labelmap_path: /labelmap/coco-80.txt

Finally, I set the app’s icon using the tips in How to change icon of custom app? - #31 by houndtt.

With these changes, the app now runs smoothly, and config saves/restarts are clean.

1 Like