Fix long CPU->GPU synchronization during Gradient clipping #3318
Conversation
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
Summary: 1. Recent changes from D79128843 introduced sync point `clipping.py` which was seen in trace 2. It was creating CPU tensors which were being moved **synchronously** to cuda devices consequently causing long wait times in training with `CudaStreamSychronization` exhibiting in trace. 3. This caused QPS degradation in CTX FM model which I was actively working on optimizing and also it cause QPS degradation in most model enabling Optimizer Gradient clipping in their yaml config. 4. This fix helps bump qps by around 5% while keep NE unimpacted. Differential Revision: D80959986
meta-cla
bot
added
the
CLA Signed
|
This pull request was exported from Phabricator. Differential Revision: D80959986 |
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Outdated suggestions cannot be applied.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:
clipping.pywhich was seen in traceCudaStreamSychronizationexhibiting in trace.Differential Revision: D80959986