moe router tp removed #1091
Closed
moe router tp removed #1091
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
|
@megha95 thanks for the contribution! June |
Merged
|
@megha95 close this since it has already been merged in this push |
4 tasks
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Outdated suggestions cannot be applied.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.
As the title suggests, this PR removes TP (tensor parallelism) for MoE router. Duplicating router across GPUs removes an allreduce for each MoE layer. This small change leads to 4-18% speedup in decoding for Mixtral-8x7B-v0.1 (4% at bs=1, 10-18% at batch size 2-16). Measured on 2xA100-80GB.