Giter VIP home page Giter VIP logo

Comments (3)

spficklin avatar spficklin commented on August 29, 2024

Oh, I just found a way to see what process are running on the GPUs. In fact all 4 are running on the first GPU with none running on the others 3:

  [Processes
       Process ID                  : 80379
           Type                    : C
           Name                    : /data/ficklin/software/KINC-3.1.1/kinc-cli
           Used GPU Memory         : 210 MiB
       Process ID                  : 80380
           Type                    : C
           Name                    : /data/ficklin/software/KINC-3.1.1/kinc-cli
           Used GPU Memory         : 210 MiB
       Process ID                  : 80381
           Type                    : C
           Name                    : /data/ficklin/software/KINC-3.1.1/kinc-cli
           Used GPU Memory         : 210 MiB
       Process ID                  : 80382
           Type                    : C
           Name                    : /data/ficklin/software/KINC-3.1.1/kinc-cli
           Used GPU Memory         : 210 MiB](url)

from kinc.

spficklin avatar spficklin commented on August 29, 2024

@bentsherman explained by email that current ACE/KINC defaults to always using the first GPU. And future version of ACE/KINC will balance out the load among the visible GPUs. So that makes sense now.

But, it seems SLURM will give me the GPU assigned to my job with the SLURM_JOB_GPUS environment
variable. KINC apparently sees all of the GPUs regardless how many I told it I wanted to use. So, I think even though future ACE/KINC will better divide up the workers by the visible GPUs, I will still need to provide the indexes to the GPUs that SLURM tells me I am allowed to run on.

Unfortunately, the format of the SLURM_JOB_GPUS is not defined in any documentation that I've looked at, so I'm not sure if it's list separated by commas or what...

from kinc.

bentsherman avatar bentsherman commented on August 29, 2024

I think you can also use the CUDA_VISIBLE_DEVICES environment variable to control which GPUs can be seen by a node. More information here:

https://devblogs.nvidia.com/cuda-pro-tip-control-gpu-visibility-cuda_visible_devices/

from kinc.

Related Issues (20)

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.