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.

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

Paper work to do over break #83

Open
10 of 16 tasks
vsoch opened this issue Dec 22, 2024 · 0 comments
Open
10 of 16 tasks

Paper work to do over break #83

vsoch opened this issue Dec 22, 2024 · 0 comments
Assignees

Comments

@vsoch
Copy link
Member

vsoch commented Dec 22, 2024

  • Go through current paper and un-generalize (I left pieces general that didn't need pointing fingers)
  • Write about issues with CUDA_VISIBLE_DEVICES
  • Colors for different environments need to be consistent across plots for environments a731e64
  • Finish adding each application, including description in methods (paper / docs citation), and results (and plots)
  • Redo all plots with white and black axes (I decided the gray won't print well and is distracting)
  • Think about what we want for single-node benchmarks (which we haven't solved)
  • Figure out how to make institution latex wider so prettier looking
  • Talk about GPU error correction setting as a possible explanation for stream on GPU (discussion)
  • Add mixbench to paper, since we need results for error correction (note that I added a mention of it just in the context of stream, I don't think we can use the full results)

I will add more items as I hit them.

Discussion items (for after break):

  • We have a lot of content - discussion of costs / contention of AI might be out of scope here.
  • I don't think we should add multi-gpu-models - it didn't run anywhere except for lassen and compute engine, and there were a lot of error messages
  • Should we add cores to the table of environments? (see node characteristics)
  • Which apps/plots to include, and for which do we want speedup (I'm currently doing miniFE for speedups)
  • I'm currently writing main parameters into application under methods. Is that sufficient?
  • Do we still want to mention all apps (e.g., kripke / quicksilver) even if we don't include results / data?
  • Look at stream docs and decide if we are using the right FOM / calculation, beyond reporting raw values.
@vsoch vsoch self-assigned this Dec 22, 2024
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

1 participant