Fix: num_logits_to_keep in composite models
#33168
Merged
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
|
The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update. |
5 tasks
LysandreJik
approved these changes
Sep 2, 2024
|
Would it be possible to add a test to ensure no regression on the issue #33164 (comment) ? |
|
Oh, also cc @gante for generation :) |
|
@LysandreJik yes, I am working on enabling generation tests on all VLMs. That should be 2-3 PRs away, after we fix some issues after the latest refactoring on VLM processing code We've overlooked many issues on VLM side due to the absence of fast generation tests and slow tests are usually not ran if modeling code is unchanged. |
BernardZach
pushed a commit
to BernardZach/transformers
that referenced
this pull request
Dec 5, 2024
* fix * paligemma
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Outdated suggestions cannot be applied.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.
What does this PR do?
Fixes issue from #33164 (comment).
We cannot assume that the model will have
num_logits_to_keepin its forward, because an LM can be used as backbone in composite models. So the current PR setsnum_logits_to_keepto None by default and adds it as a kwarg only when it is explicitly set to an integer by usersPlus, updated VLMs accordingly with
num_logits_to_keep. Other VLMs like BLIP don't need this arg because they delegate the whole generation process to the LM.cc @merveenoyan