GPU image

[ekourlit]

Hi, I would like to request a new image to appear at the UChicago JupyterLab for usage on GPU machines. I think the specification file below would work for me. Thanks in advance!

channels:
- conda-forge
- defaults
dependencies:
- coffea
- cupy
- fsspec-xrootd
- matplotlib
- mplhep
- numba
- numpy
- pandas
- scipy
- pip:
  - awkward==2.6.6

[ekourlit]

Is there any chance we can get this image into the system? BTW, Awkward has been updated to 2.6.8

[matthewfeickert]

I think @fengpinghu can comment best on this now.

[fengpinghu]

The following would work to get the coda-toolkit added to this image. But it would increase the image size significantly. I guess we have a few options:
1)add cuda-toolkit to analysisbase-dask-uc
2)created a separate image(analysisbase-dask-uc-gpu) with cuda-toolkit added
3)add coffea to one of the ML-platform images that Ilija maintains.
Thoughts?
@ivukotic

+RUN yum install -y dnf-plugins-core
+RUN dnf install -y https://dl.fedoraproject.org/pub/epel/epel-release-latest-9.noarch.rpm
+RUN dnf config-manager --enable epel
+#install cuda-toolkit
+RUN  dnf config-manager --add-repo https://developer.download.nvidia.com/compute/cuda/repos/rhel9/x86_64/cuda-rhel9.repo
+RUN  dnf clean expire-cache
+RUN  dnf install -y cuda-toolkit
+RUN  dnf install -y nvidia-gds

[ivukotic]

I think we can do both 2 and 3. Adding cuda-toolkit to an already very big and complex image would be asking for trouble… My 2cents, Ilija From: Fengping Hu ***@***.***> Sent: Thursday, October 17, 2024 21:01 To: usatlas/analysisbase-dask-uc ***@***.***> Cc: Ilija Vukotic ***@***.***>; Mention ***@***.***> Subject: Re: [usatlas/analysisbase-dask-uc] GPU image (Issue #10) The following would work to get the coda-toolkit added to this image. But it would increase the image size significantly. I guess we have a few options: 1)add cuda-toolkit to analysisbase-dask-uc 2)created a separate image(analysisbase-dask-uc-gpu) with cuda-toolkit added 3)add coffea to one of the ML-platform images that Ilija maintains. Thoughts? @ivukotic<https://urldefense.com/v3/__https:/github.com/ivukotic__;!!BpyFHLRN4TMTrA!_wdLtkI7SGUsgnBSLeP1u_NIsp9Tyj-QC1vvZqt6bFu8eULYQTidNV-OGonyGXnFG6gMsmFxUDynQdRPlS0EUhRbLp9i$> +RUN yum install -y dnf-plugins-core +RUN dnf install -y https://dl.fedoraproject.org/pub/epel/epel-release-latest-9.noarch.rpm<https://urldefense.com/v3/__https:/dl.fedoraproject.org/pub/epel/epel-release-latest-9.noarch.rpm__;!!BpyFHLRN4TMTrA!_wdLtkI7SGUsgnBSLeP1u_NIsp9Tyj-QC1vvZqt6bFu8eULYQTidNV-OGonyGXnFG6gMsmFxUDynQdRPlS0EUgivMQOZ$> +RUN dnf config-manager --enable epel +#install cuda-toolkit +RUN dnf config-manager --add-repo https://developer.download.nvidia.com/compute/cuda/repos/rhel9/x86_64/cuda-rhel9.repo<https://urldefense.com/v3/__https:/developer.download.nvidia.com/compute/cuda/repos/rhel9/x86_64/cuda-rhel9.repo__;!!BpyFHLRN4TMTrA!_wdLtkI7SGUsgnBSLeP1u_NIsp9Tyj-QC1vvZqt6bFu8eULYQTidNV-OGonyGXnFG6gMsmFxUDynQdRPlS0EUt5Tsqn3$> +RUN dnf clean expire-cache +RUN dnf install -y cuda-toolkit +RUN dnf install -y nvidia-gds — Reply to this email directly, view it on GitHub<https://urldefense.com/v3/__https:/github.com/usatlas/analysisbase-dask-uc/issues/10*issuecomment-2420315737__;Iw!!BpyFHLRN4TMTrA!_wdLtkI7SGUsgnBSLeP1u_NIsp9Tyj-QC1vvZqt6bFu8eULYQTidNV-OGonyGXnFG6gMsmFxUDynQdRPlS0EUmEszO2q$>, or unsubscribe<https://urldefense.com/v3/__https:/github.com/notifications/unsubscribe-auth/AANICGCAVJ6QSRCBJTHDYXLZ4ACNXAVCNFSM6AAAAABLFFRKXKVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDIMRQGMYTKNZTG4__;!!BpyFHLRN4TMTrA!_wdLtkI7SGUsgnBSLeP1u_NIsp9Tyj-QC1vvZqt6bFu8eULYQTidNV-OGonyGXnFG6gMsmFxUDynQdRPlS0EUtyTz9cq$>. You are receiving this because you were mentioned.Message ID: ***@***.******@***.***>>

[ekourlit]

At the moment this is independent of AnalysisBase. So for me the option 3) would work and might just be the simplest?