Repository navigation
Conversation
AdnanElAssadi
changed the title
Samoed
approved these changes
Oct 25, 2026
Samoed
left a comment
Samoed
left a comment
Member
There was a problem hiding this comment.
I updated implementation to not add extra parameters to create_dataloader
Contributor
Author
looks good, we can merge? |
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Rerankers crash on the video QA tasks (EgoSchemaVideoCentricQA, OmniVideoBenchVideoCentricQA) with
ArrowInvalid: offset overflow while concatenating arrays.To rerank, mteb pairs each question with each of its answer options, so every question is repeated once per option. The questions are then prepared for the model, and that step copies each repeated row, including the question's video. On EgoSchema that is 2036 copies of the videos, more than Arrow allows in one column.
This prepares each question once and repeats it afterwards, so nothing is copied. It adds an
indicesargument tocreate_dataloaderfor that. The corpus side is unchanged.Qwen3-VL-Reranker-2B:
ms-marco-MiniLM-L6-v2 on AskUbuntuDupQuestions also gives identical scores (0.55516).