Share tensors between comp replay and comms replay #194
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
Summary: Share tensors between comp replay and comms replay When run full replay in et_replay, compute replay and comms replay manage the tensor allocation separately. So some tensors are double allocated, this leads to the full replay of Llama4 70B out of memory. This DIFF is to fix it by allocating tensors in comp replay and passes them to comms replay. Reviewed By: sanrise Differential Revision: D67353163
shengfukevin
requested review from
kingchc,
louisfeng,
sunghlin,
shengbao-zheng and
briancoutinho
as code owners
facebook-github-bot
added
the
CLA Signed
|
This pull request was exported from Phabricator. Differential Revision: D67353163 |
|
This pull request has been merged in c5f8d06. |
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Outdated suggestions cannot be applied.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:
Share tensors between comp replay and comms replay
When run full replay in et_replay, compute replay and comms replay manage the tensor allocation separately. So some tensors are double allocated, this leads to the full replay of Llama4 70B out of memory.
This DIFF is to fix it by allocating tensors in comp replay and passes them to comms replay.
Reviewed By: sanrise
Differential Revision: D67353163