[FEAT] EVA: ensure deterministic behavior of SVD on multi gpu setups #2225
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. |
|
@BenjaminBossan thanks for the suggestions, I integrated them! |
BenjaminBossan
approved these changes
Nov 21, 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Outdated suggestions cannot be applied.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.
Deterministic behavior of SVD computations on layer inputs is important on multi gpu setups to ensure resulting EVA state dicts are identical. Currently deterministic behaviour of SVD is ensured by just setting torch.manual_seed() inside
_get_eva_state_dictwhich is not ideal because it can interfere with the seed set by the user. I therefore moved this functionality inside theIncrementalPCAclass.I also added a new argument for
get_eva_state_dict:gather_distributed_inputswhich isFalseby default and defines if inputs should be concatenated for the SVD computation for distributed runs.