Fix bug in prepare_inputs_for_generation function: in Qwen2-VL (#36037)#36038
Open
JamesHujy wants to merge 2 commits intohuggingface:mainfrom
Open
Fix bug in prepare_inputs_for_generation function: in Qwen2-VL (#36037)#36038JamesHujy wants to merge 2 commits intohuggingface:mainfrom
JamesHujy wants to merge 2 commits intohuggingface:mainfrom
Conversation
Member
|
cc @gante (and sorry!) |
gante
approved these changes
Feb 14, 2025
Contributor
gante
left a comment
gante
left a comment
There was a problem hiding this comment.
LGTM, thank you for fixing 🙏
Contributor
|
@JamesHujy after sorting the conflicts, it seems like the fix is already on |
shethaadit
approved these changes
Feb 25, 2025
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Fix the problem in Qwen2-VL. A similar problem is mentioned in PR #34095.
When using synced_gpus, after one peer finishes generating, the cache position in the generation process continues to increase. This leads to the input IDs going out of bounds, resulting in errors. The issue specifically occurs in the following line of code:
modeling_qwen2_vl.py#L1739.
The root cause seems to be the difference in the implementation of the prepare_inputs_for_generation function compared to the default implementation found here:
utils.py#L388.