Fix tensor parallelism with SGMV to use true rank of the LoRA after splitting #324
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
4 tasks
jeffreyftang
approved these changes
Mar 13, 2024
| lora_b_list = [pad_rank(w, dim=0, world_size=self.world_size) for w in lora_b_list] | ||
|
|
||
| if lora_b_list: | ||
| if lora_a_list: |
Contributor
There was a problem hiding this comment.
Curious what the significance of swapping from lora_b_list to lora_a_list here is.
Contributor
Author
There was a problem hiding this comment.
It is the same, actually.
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Outdated suggestions cannot be applied.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 #308.
Previous implementation assumed the rank set in the config was correct and we only need to scale by process group size. However, this may not always be the case depending on how the weights get split for tensor parallelism (column vs row parallel). As such, we should ignore the config and instead rely on the true rank dimension of the tensor independent of the process group size.