Conversation
|
|
||
| curl -LsSf https://astral.sh/uv/install.sh | sh | ||
|
|
||
| mkdir aghilan-workspace |
Collaborator
There was a problem hiding this comment.
perhaps user-workspace?
| mkdir aghilan-workspace | ||
|
|
||
| cd aghilan-workspace | ||
| echo "Cloning torchtitan" |
Collaborator
There was a problem hiding this comment.
perhaps "Cloning kingong (torchtitan fork)" or something
Contributor
Author
There was a problem hiding this comment.
fyi this pr isn't ready for review, I meant to make it draft
Comment on lines
+11
to
+17
| # Configure git credentials if GITHUB_TOKEN is available | ||
| if [ -n "${GITHUB_TOKEN:-}" ]; then | ||
| echo "Configuring git credentials..." | ||
| git config --global credential.helper store | ||
| echo "https://aghilann:${GITHUB_TOKEN}@github.com" > ~/.git-credentials | ||
| chmod 600 ~/.git-credentials | ||
| git config --global url."https://aghilann:${GITHUB_TOKEN}@github.com/".insteadOf "https://github.com/" |
Collaborator
There was a problem hiding this comment.
does this work for any user?
Contributor
Author
There was a problem hiding this comment.
Yes it does, actually realized you don't actually need this
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
2 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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
When you attempt to load the model in from HF, it triggers a brief GPU memory spike which will OOM on a 4 Node Setup. This is the problematic line. I did not have time to run a profiler on this but I would try to debug this locally by loading in a debug HF model of DSV3 with 2 layers and run the profiler on it.
It will work on a 8 Node Setup but I then noticed I need to bump the expert parallelism to 64, once I did that in begin to train but at every epoch it was producing NaN for loss.