[fix] Correctly pass mask in TransformerBlock.forward in transformer_layers.py #218
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
…layers.py The attention mask was not passed correctly to the `Attention` in `TransformerBlock.forward`. One problem it caused was that when passing two images in the image encoder, the attention would be done on all images at the same time, thus taking more resources and returning an incorrect result. This PR fixes this problem.
|
I observed the same bug, this PR fixes the bug |
kmk142789
approved these changes
Oct 12, 2025
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 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Outdated suggestions cannot be applied.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.
The attention mask was not passed correctly to the
AttentioninTransformerBlock.forward.One problem it caused was that when passing two images in the image encoder, the attention would be done on all images at the same time, thus taking more resources and returning an incorrect result.
This PR fixes this problem.
Tested by passing two images separately to the encoder, and checking that passing them as a batch gives "almost" the same thing.
Also tested on classic vision benchmarks where the results were very low (and inference very slow), now considerably better:
ImageNet KNN: