[ICCV'25] The official code implementation of paper "Combining Similarity and Importance for Video Token Reduction on Large Visual Language Models"

efficient-deep-learning llm lvlm video
1 Open Issue Need Help Last updated: Aug 2, 2025

Open Issues Need Help

View All on GitHub

AI Summary: The user is encountering an AttributeError ('Qwen2VLForConditionalGeneration' object has no attribute 'rope_deltas') when running the example script for the Qwen2-VL model with FrameFusion. The task is to debug this error, likely stemming from an incompatibility or missing attribute in the custom Qwen2-VL model adapter within the FrameFusion codebase. This involves examining the `modeling_qwen2_vl.py` file, specifically line 206, and potentially modifying the model adapter to correctly handle the 'rope_deltas' attribute or remove the dependency if unnecessary.

Complexity: 4/5
help wanted

[ICCV'25] The official code implementation of paper "Combining Similarity and Importance for Video Token Reduction on Large Visual Language Models"

Python
#efficient-deep-learning#llm#lvlm#video