[Relax] Handle dynamic arguments in legalization of nn.attention #16592
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
slyubomirsky
approved these changes
Feb 22, 2024
Contributor
slyubomirsky
left a comment
slyubomirsky
left a comment
There was a problem hiding this comment.
Good catch and it's a nice improvement. We should really have "always think about the dynamic case" as a motto.
Contributor
Author
I like the motto, as it's the most user-friendly option at a high level. Since specialization can go from dynamic to static, but going from static to dynamic would be ambiguous, I really like "dynamic until specialized" as the default behavior. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 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Outdated suggestions cannot be applied.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.
Prior to this commit, when using
causal_mask="BottomRight"inR.nn.attention, the legalization would assume that the query and key/value sequence lengths were static integers. This commit updates the legalization to allow dynamic shapes.