This repository has been archived by the owner on Nov 17, 2023. It is now read-only.
[BUGFIX][v1.x] Fix a bug for 1bit quantizer #20007
Merged
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Outdated suggestions cannot be applied.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.
Description
Fix a bug for encoding parameters in gradient_compression.cc, which may lead to inconsistent threshold on worker and server when using 1bit quantizer (if the threshod is set to a non-zero value).
releated pr: #17952
Checklist
Essentials
Changes
Comments