[WIP] LAMB optimizer #1460
Open
[WIP] LAMB optimizer #1460
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
|
LGTM |
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Outdated suggestions cannot be applied.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.
[DO NOT MERGE]
This is a WIP on implementing LAMB optimizer from BERT. It apparently allows to scale training on huge batches. There are some ambiguities : different algorithms between v1 and v2/v3 of the paper, some blurry definitions and no official implementation yet (a few ones are out there but differ on a few points), no clear learning_rate schedule in the paper despite detailed experiments, etc.
Also, there might be some significant tuning to do in order to find appropriate values for our tasks.
I open this PR for future work, when we'll have more elements.
The current version here is based on https://github.com/cybertronai/pytorch-lamb, which itself is based on
torch.optimizers.Adam.