Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Add stardist 0.8.5 recipe to containers #551

Open
wants to merge 2 commits into
base: master
Choose a base branch
from
Open

Add stardist 0.8.5 recipe to containers #551

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
676e791
Added stardist 0.8.5 recipe.
FloWuenne Nov 29, 2023
1a5a762
Added missing checklist items.
FloWuenne Nov 29, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
29 changes: 29 additions & 0 deletions stardist/0.8.5/Dockerfile
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,29 @@
FROM tensorflow/tensorflow:2.11.0-gpu-jupyter

LABEL base_image="tensorflow:2.11.0-gpu-jupyter"
LABEL version="1"
LABEL maintainer="[email protected]"
LABEL software="stardist"
LABEL software.version="0.8.5"
LABEL about.summary="Object Detection with Star-convex Shapes."
LABEL about.home="https://github.com/stardist/stardist"
LABEL about.license="BSD-3-Clause"
LABEL about.license_file="https://github.com/stardist/stardist/blob/master/LICENSE.txt"

MAINTAINER Florian Wuennemann <[email protected]>

ARG DEBIAN_FRONTEND="noninteractive"
ARG STARDIST_VERSION="0.8.5"
ARG NVIDIA_DRIVER_VERSION=470
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I feel like this will be an issue? According to the stardist github, the NVIDIA_DRIVER_VERSION should be adjusted to the runtime environnement (depending on the GPU you have, I guess).

No sure how the container will behave if the runtime env does not match

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks for the review @mboudet !

What would you suggest how to handle this issue?

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

No quite sure, to be honest. It would be possible to install the correct lib at runtime I guess, but it would be quite dirty, and would probably not work with singularity.

I'm not sure how does it work with the conda package? Would it be possible to directly install the package in the dockerfile (like here), or does it require external librairies ?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I don't think I understand what you mean whether it would be possible to directly install the package in the dockerfile? Which package are you referring to? I don't see anything related to nvidia drivers in the recipe you linked unforunately? 🤷🏼

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Well, you use the 'NVIDIA_DRIVER_VERSION' variable to install a specific version of the libnvidia-compute in the next step. (Line 24). According the the stardist github, the libnvidia-compute version should match the GPU driver version.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

So as is custom these days, I asked ChatGPT for help. The suggestion from it was to specify --build-arg when building the docker container at runtime to adjust the ENV variable NVIDIA_DRIVER_VERSION to the one available:
docker build --build-arg NVIDIA_DRIVER_VERSION=<desired_version> -t <your_image_name> .

I am not sure there is a good way to automatically detect this version number. I also don't really know whether we could build a container for cpu completely without NVIDIA drivers...

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yeah, this would work, but that mean that the user would need to build the image themselve (so, not much point in having the image on biocontainers). Also, building an image usually require root/sudo privileges, which are not usually available in computing clusters.

My question regarding the conda package stands though. I'm not sure if a mamba/conda install in the container would work. This might need input from the software's developpers.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I will ask in their repo if they can advise regarding the mamba / conda install. What would be the advantage over mamba/conda install over the pip install that we are currently doing?

Totally agree on the container build comment. One of the main goals here is to be able to build a singularity container from the docker container and if we can't do that, you are right, pretty useless to put the docker container on biocontainers 🙃

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I asked them in stardist/stardist#259


RUN apt-get update && apt-get install -y --no-install-recommends \
ocl-icd-dev \
ocl-icd-opencl-dev \
opencl-headers \
clinfo \
libnvidia-compute-${NVIDIA_DRIVER_VERSION} \
&& apt-get clean \
&& rm -rf /var/lib/apt/lists/*

RUN python3 -m pip install --upgrade pip
RUN pip install stardist==$STARDIST_VERSION gputools edt