Skip to content

Running the CPU profile on ARM devices breaks if the reranker is enabled. #50

@EnjoyBacon7

Description

@EnjoyBacon7

As mentioned in the documentation, the reranker is unavailable on arm devices when running the cpu profile:

[!INFO]
Reranking is not available on ARM architectures when using the CPU profile. This is particularly relevant for Apple Silicon devices. The reranking service should be disabled in such cases.

This still lets the user enable the reranker manually, leading to undefined behavior. Additionally, the infinity container for the reranker is still created, though not used.

Related compose snippet:

  reranker-cpu:
    <<: *reranker_template
    platform: linux/amd64
    image: michaelf34/infinity:latest-cpu
    build:
      context: https://github.com/michaelfeil/infinity.git#main:libs/infinity_emb
      dockerfile: Dockerfile.cpu_auto
    deploy: {}
    command: >
      v2
      --engine torch
      --model-id ${RERANKER_MODEL:-Alibaba-NLP/gte-multilingual-reranker-base}
      --port ${RERANKER_PORT:-7997}
    profiles:
      - 'cpu'

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions