distributed init updates (addition of non-meta-tensor/rank0-broadcast path) #674
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
… path) Summary: add non-meta tensor initialization path to distributed init. in this path, the model is materialized by all processes on the cpu (duplicated—resulting in proportionally increased cpu memory requirements) instead of just by rank 0. afterwards, all modules/buffers/params are iterated over and sharded as previously; however, instead of rank0 first broadcasting its tensor data to all other meta-tensor ranks, all ranks now simply copy their own cpu-init'd data onto their corresponding gpu and continue. Reviewed By: xiecong Differential Revision: D57022318
facebook-github-bot
added
the
CLA Signed
|
This pull request was exported from Phabricator. Differential Revision: D57022318 |
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Outdated suggestions cannot be applied.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:
add non-meta tensor initialization path to distributed init.
in this path, the model is materialized by all processes on the cpu (duplicated—resulting in proportionally increased cpu memory requirements) instead of just by rank 0. afterwards, all modules/buffers/params are iterated over and sharded as previously; however, instead of rank0 first broadcasting its tensor data to all other meta-tensor ranks, all ranks now simply copy their own cpu-init'd data onto their corresponding gpu and continue.
Reviewed By: xiecong
Differential Revision: D57022318