fix gradient allreduce #215
Merged
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
This was referenced Jun 15, 2025
facebook-github-bot
added
the
CLA Signed
tushar00jain
force-pushed
the
pr215
branch
4 times, most recently
from
2546bbe to
7aba771
Compare
Merged
H-Huang
approved these changes
Jun 16, 2025
| else: | ||
| p.grad = self._grads[name] | ||
|
|
||
| del self._grads[name] |
Summary: - fix setting `_local_tensor` of a dtensor directly - fix allreduce bucketized to not use `parameter.grad` - simplify some code Test Plan: - added a test to validate the gradient are saved and set correctly - the previous test in `local_sgd_test` fails because allreduce is not performed on `param.grad` - updated the test to first set the grads, then load the grads to make sure they reflect the allreduce result
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Outdated suggestions cannot be applied.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.
Summary:
_local_tensorof a dtensor directlyparameter.gradTest Plan:
local_sgd_testfails because allreduce is not performed onparam.gradStack created with Sapling. Best reviewed with ReviewStack.