transformer_engine: Control backward FP8 metadata synchronization (v1.6 onwards) #379
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
IvanYashchuk
approved these changes
May 10, 2024
kshitij12345
changed the title
kshitij12345
requested review from
mruberry,
lantiga,
robieta,
t-vi and
carmocca
as code owners
Co-authored-by: Carlos Mocholí <carlossmocholi@gmail.com>
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Outdated suggestions cannot be applied.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.
Fixes: #368
Failure in above issue happens because we don't always correctly reorder the inputs to the
te_linearin distributed setting.This PR takes a different approach and instead inserts a sync operation in the backward trace which makes us agnostic to the previous constraint.
Details of the approach can be found at :
NOTE: Backward FP8 metadata syncintransformer_engineex.py.NOTE: Once v1.6 is released, I am planning to drop the support for older versions and clean-up the code.