All Gather Once (FSDP Zero-one) #1865
Draft
Conversation
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
gobbleturk
reviewed
Jun 23, 2025
MaxText/configs/base.yml
Outdated
| @@ -186,6 +186,9 @@ pipeline_delay_activation_forwarding: False # This delays the activation forward | |||
| # the communication and compute in each iteration are now independent. However this comes at the cost of doubling the pipeline bubble, | |||
| # and you must set the number of microbatches to at least 2 * num_stages (the minimum 2 * num_stages is set by default with this delay). | |||
|
|
|||
| model_fsdp_ag_once: False # This controls whether the Zero-1 optimization is active. | |||
There was a problem hiding this comment.
how does this interplay with pipeline_fsdp_ag_once? If you set the model one to true, what happens when we set pipeline_fsdp_ag_once to true/false?
gobbleturk
reviewed
Jun 23, 2025
MaxText/layers/models.py
Outdated
| The goal of this optimization is to reduce communication overhead. In the standard | ||
| FSDP implementation, an all-gather operation on the model weights is performed twice | ||
| for each microbatch (once for the forward pass, once for the backward pass). |
There was a problem hiding this comment.
microbatch -> gradient accumulation microbatch
gobbleturk
reviewed
Jun 23, 2025
gobbleturk
reviewed
Jun 23, 2025
gobbleturk
reviewed
Jun 23, 2025
gobbleturk
reviewed
Jun 23, 2025
|
Can we also add a unit test (integration test) with this feature on in train_tests in your PR description you can link the profiles before and after (ideally with pdb=1 to make the feature more obvious) |
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Outdated suggestions cannot be applied.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.
Description
Start with a short description of what the PR does and how this is a change from
the past.
The rest of the description includes relevant details and context, examples:
If the change fixes a bug or a Github issue, please include a link, e.g.,:
FIXES: b/123456
FIXES: #123456
Notice 1: Once all tests pass, the "pull ready" label will automatically be assigned.
This label is used for administrative purposes. Please do not add it manually.
Notice 2: For external contributions, our settings currently require an approval from a MaxText maintainer to trigger CI tests.
Tests
Please describe how you tested this change, and include any instructions and/or
commands to reproduce.
Checklist
Before submitting this PR, please make sure (put X in square brackets):