Remove breaking torchrun config for single-node runs#292
Merged
hemildesai merged 2 commits intoNVIDIA-NeMo:mainfrom Jul 25, 2025
Merged
Remove breaking torchrun config for single-node runs#292hemildesai merged 2 commits intoNVIDIA-NeMo:mainfrom
hemildesai merged 2 commits intoNVIDIA-NeMo:mainfrom
Conversation
Signed-off-by: Roee Landesman <roeeland@cisco.com>
ri-roee
force-pushed
the
fix-single-node-launcher
branch
from
c95855d to
1f4dfb3
Compare
Contributor
|
Did you test this on multi-node? This workaround was added for supporting torchrun on multi-node K8s as dynamic rendezvous was running into connection errors. |
Contributor
Author
|
Shoot you're right! This means that |
Contributor
|
Hey folks! 👋 SkyPilot recently fixed dynamic rdvz discovery for multi-node in this PR: skypilot-org/skypilot#5960. So using static rdvz should no longer be required. The fix is available in SkyPilot 0.10. |
Contributor
Author
|
Amazing, thank you @romilbhardwaj! Recently upgraded #297, so re-opening PR |
hemildesai
approved these changes
Jul 25, 2025
zoeyz101
pushed a commit
to zoeyz101/NeMo-Run
that referenced
this pull request
Nov 12, 2025
* remove breaking torchrun config for single-node runs Signed-off-by: Roee Landesman <roeeland@cisco.com> * fix lint Signed-off-by: Roee Landesman <roeeland@cisco.com> --------- Signed-off-by: Roee Landesman <roeeland@cisco.com> Signed-off-by: Zoey Zhang <zozhang@nvidia.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
I suspect there were upstream changes between now and when this code block was original written because it actually breaks single-node torchrun runs today. I tested this by overwritting this config change locally and my
launcher=torchrun(on a single-node in k8s) started working. The error I was seeing originally was[default7]:[I715 22:47:11.296780083 socket.cpp:872] [c10d] No socket on (localhost, 0) is listening yet, will retry.for context, which led me to believe that rdzv backend was misconfigured, which finally led me to the overwritting ofc10d.