Fix: multigpu training #33271
Merged
Fix: multigpu training #33271
Conversation
This file contains 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
|
The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update. |
SunMarc
approved these changes
Sep 3, 2024
There was a problem hiding this comment.
Thanks for the fix @zucchini-nlp ! This is the right attribute to check !
amyeroberts
approved these changes
Sep 3, 2024
There was a problem hiding this comment.
LGTM - thanks for digging into this and fixing!
I wonder whether this means we might end up with different states in the trainer workflow for multigpu vs single gpu models i.e. we can't assume correctly whether we're on the device or not.
A more robust fix in the future would be to have _move_model_to_device to be compatible with the multigpu environment
BernardZach
pushed a commit
to BernardZach/transformers
that referenced
this pull request
Dec 5, 2024
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.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.
What does this PR do?
Addresses issues in LLaVA-VL/LLaVA-NeXT#79 (comment) and huggingface/trl#1785 (comment).
I could trace the issue to the line where we move the model to
args.devicewhich apparently moved the model from multigpu to a single device. That caused device mismatch errors where the kwargs are moved to device fromhf-device-mapwhile the weights are all inargs.device. I am not 100% sure this is the right fix, as I am not a Trainer master so pls let me know if there's a better way to check thatI guess @SunMarc for multigpu related issues