Use master weights for bfloat16 FusedAdam when master_weights=True#1731
Open
cbcase wants to merge 1 commit intoNVIDIA:masterfrom
Open
Use master weights for bfloat16 FusedAdam when master_weights=True#1731cbcase wants to merge 1 commit intoNVIDIA:masterfrom
cbcase wants to merge 1 commit intoNVIDIA:masterfrom
Conversation
Contributor
Author
|
Ping @minitu, looks like you added this support originally -- could you take a look? Thanks |
Contributor
|
LGTM, we only looked at adding master weights for FP16 AMP at the time of the original PR. |
crcrpar
reviewed
Oct 18, 2023
Collaborator
crcrpar
left a comment
crcrpar
left a comment
There was a problem hiding this comment.
looks good but could you add a test case of bfloat16 model with fp32 weights to
apex/tests/L0/run_optimizers/test_adam.py
Line 151 in 19cc873
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
As mentioned in #1728, the FusedAdam optimizer ignores
master_weights=Truefor bfloat16 parameters. This PR fixes that oversight. I have confirmed that the behavior now matches a "by hand" implementation of master weights (hand-copying) along with vanilla torch.optim.AdamW on the fp32 copy.