Reduce buffer copying by using one device to reduce and distribute #957
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
Alex-Vasile
force-pushed
the
all_reduce_change
branch
from
c91965d to
f79044a
Compare
|
Should we hold off on merging this until the sharding accuracy is stable again? |
Alex-Vasile
force-pushed
the
all_reduce_change
branch
2 times, most recently
from
e4ca7a4 to
9bceebf
Compare
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Outdated suggestions cannot be applied.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.
For tensor parallelization (TP) > 1, this reduces buffer copies from (TP-1)*TP copies down to (TP-1)*2. Setup work for future improvements. On its own, it only results in minor speedups since all devices have to wait for the reduction result.