Remove DataParallel.scale_loss & apply_collective_grads#27603
Merged
chenwhql merged 4 commits intoPaddlePaddle:developfrom Sep 29, 2020
Merged
Conversation
|
Thanks for your contribution! |
chenwhql
requested review from
MRXLT and
danleifeng
and removed request for
MRXLT
phlrain
previously approved these changes
Sep 28, 2020
chenwhql
dismissed stale reviews from saxon-zh, phlrain, and danleifeng
via
0ea3483
phlrain
approved these changes
Sep 29, 2020
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
6 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
PR types
New features
PR changes
APIs
Describe
In order to simplify the writing of dynamic parallel training, remove
DataParallel.scale_loss & apply_collective_grads.DataParallel.scale_losss->loss.backward()DataParallel.apply_collective_grads->optimizer.step/minimizeTODO:
scale_losssandapply_collective_gradsinto C++ for better preformance (in the future, now only for API stability of 2.0-RC)