[BUG] machine learning exists with code 132

[tarik02]

The bug

machine-learning container always exits with code 132

The OS that Immich Server is running on

Official Helm chart within k3s (kubernetes) inside systemd-nspawn container inside qemu VM (8.0, machine=virt-8.0, cpu=host) on Raspberry PI 4

Version of Immich Server

v1.54.1

Version of Immich Mobile App

v1.54.0 build.94

Platform with the issue

• [X] Server
• [ ] Web
• [ ] Mobile

Your docker-compose.yml content

Not used (official helm chart is used instead)

Your .env content

env:
  REDIS_HOSTNAME: <hidden>
  REDIS_PORT: '6379'
  DB_HOSTNAME: <hidden>
  DB_USERNAME: "{{ .Values.postgresql.global.postgresql.auth.username }}"
  DB_DATABASE_NAME: "{{ .Values.postgresql.global.postgresql.auth.database }}"
  # -- You should provide your own secret outside of this helm-chart and use `postgresql.global.postgresql.auth.existingSecret` to provide credentials to the postgresql instance
  DB_PASSWORD: "{{ .Values.postgresql.global.postgresql.auth.password }}"
  TYPESENSE_ENABLED: "{{ .Values.typesense.enabled }}"
  TYPESENSE_API_KEY: "{{ .Values.typesense.env.TYPESENSE_API_KEY }}"
  TYPESENSE_HOST: '{{ printf "%s-typesense" .Release.Name }}'
  IMMICH_WEB_URL: '{{ printf "http://%s-web:3000" .Release.Name }}'
  IMMICH_SERVER_URL: '{{ printf "http://%s-server:3001" .Release.Name }}'
  IMMICH_MACHINE_LEARNING_URL: '{{ printf "http://%s-machine-learning:3003" .Release.Name }}'

Reproduction steps

1. Build QEMU 8.0 (was using 5.2 from upstream before) on Raspbian.
2. Install any gnu/linux OS within VM.
3. Create systemd-nspawn container.
4. Install k3s inside container.
5. Apply official immich helm chart.

Additional information

No response

[bo0tzz]

This is potentially related to https://github.com/pytorch/pytorch/issues/97226