Tensorflow AVX instructions error when using ARM build

[camlloyd]

Using the new 2.7.8-arm image:

(base) cameron@Camerons-MacBook-Pro ~ % nextflow run labsyspharm/mcmicro --in /Users/cameron/exemplar-002

 N E X T F L O W   ~  version 24.10.4

Launching `https://github.com/labsyspharm/mcmicro` [pedantic_keller] DSL2 - revision: 950169628b [master]

executor >  local (4)
[-        ] staging:phenoimager2mc         -
[-        ] illumination                   -
[-        ] registration:ashlar            -
[-        ] background:backsub             -
[-        ] dearray:coreograph             -
[-        ] segmentation:roadie:runTask    -
[ad/0bc4cf] seg…ntation:worker (unmicst-2) | 0 of 4
[-        ] segmentation:s3seg             -
[-        ] quantification:mcquant         -
[-        ] downstream:worker              -
[-        ] viz:autominerva                -
ERROR ~ Error executing process > 'segmentation:worker (unmicst-1)'

Caused by:
  Process `segmentation:worker (unmicst-1)` terminated with an error exit status (134)


Command executed:

  python /app/unmicstWrapper.py --stackOutput --outputPath .  4.ome.tif

Command exit status:
executor >  local (4)
[-        ] staging:phenoimager2mc         -
[-        ] illumination                   -
[-        ] registration:ashlar            -
[-        ] background:backsub             -
[-        ] dearray:coreograph             -
[-        ] segmentation:roadie:runTask    -
[ad/0bc4cf] seg…ntation:worker (unmicst-2) | 2 of 4, failed: 2
[-        ] segmentation:s3seg             -
[-        ] quantification:mcquant         -
[-        ] downstream:worker              -
[-        ] viz:autominerva                -
ERROR ~ Error executing process > 'segmentation:worker (unmicst-1)'

Caused by:
  Process `segmentation:worker (unmicst-1)` terminated with an error exit status (134)


Command executed:

  python /app/unmicstWrapper.py --stackOutput --outputPath .  4.ome.tif

Command exit status:
  134

Command output:
  (empty)

Command error:
  The TensorFlow library was compiled to use AVX instructions, but these aren't available on your machine.
  .command.sh: line 2:     7 Aborted                 python /app/unmicstWrapper.py --stackOutput --outputPath . 4.ome.tif

Work dir:
  /Users/cameron/work/b4/fd42b2aee8e651fffd39d4a9f8eefd

Container:
  labsyspharm/unmicst:2.7.8-arm

Tip: you can replicate the issue by changing to the process work dir and entering the command `bash .command.run`

 -- Check '.nextflow.log' file for details

(base) cameron@Camerons-MacBook-Pro ~ % 

[camlloyd]

Quickly, changing as little as possible, I can build a cpu- and arm- only proof-of-concept that produces the following output:

├── probability-maps
│   └── unmicst
│       ├── 1-pmap.tif
│       ├── 2-pmap.tif
│       ├── 3-pmap.tif
│       └── 4-pmap.tif

Sure, it runs slow, but it runs!

Dockerfile:

FROM python:3.10-slim
RUN apt-get update && apt-get install -y curl && rm -rf /var/lib/apt/lists/*
RUN pip install tensorflow==2.14.0 numpy==1.26.4 scikit-image matplotlib tifffile==2025.1.10 pynvml czifile==2019.7.2 nd2reader==3.2.3 imagecodecs

COPY . /app
RUN curl -f -o /app/models/nucleiDAPILAMIN/model.ckpt.data-00000-of-00001 https://mcmicro.s3.amazonaws.com/models/unmicst2/model.ckpt.data-00000-of-00001
RUN curl -f -o /app/models/nucleiDAPI1-5/model.ckpt.data-00000-of-00001 https://mcmicro.s3.amazonaws.com/models/unmicst1-5/model.ckpt.data-00000-of-00001

Also needed to pin a few dependencies due to breaking changes (added numpy until I find the correct version of scikit-image).

[ArtemSokolov]

Thanks for checking these.

I suspect we may need to build FROM an ARM-based Tensorflow image. There are some v2.7 builds available at https://hub.docker.com/r/armswdev/tensorflow-arm-neoverse/tags