SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2409.10994 · 2024

Less is More: A Simple yet Effective Token Reduction Method for Efficient Multi-modal LLMs

arXiv · PDF · Open in the Atlas

Code that ran

We lifted 8 functions out of this paper's own repositories and ran 8 of them in a sandbox. "Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results.

RepositoryRoleRan
freedomintelligence/trim canonical 8 of 8
FunctionStatusWhere it lives
collate_fn Ran freedomintelligence/trim/llava/eval/model_vqa_loader.py
code served (permissive licence) · get_code("20e4f665698a3d18")
divide_to_patches Ran freedomintelligence/trim/llava/mm_utils.py
code served (permissive licence) · get_code("7e03b180fa317c9a")
get_chunk Ran freedomintelligence/trim/llava/eval/model_vqa.py
code served (permissive licence) · get_code("42a46570620cd9fa")
is_none Ran freedomintelligence/trim/llava/eval/model_vqa_mmbench.py
code served (permissive licence) · get_code("bae18947b56f2be1")
load_image Ran freedomintelligence/trim/predict.py
code served (permissive licence) · get_code("9b3c1cb391672ccb")
resize_and_pad_image Ran freedomintelligence/trim/llava/mm_utils.py
code served (permissive licence) · get_code("468eedeba67f1b00")
select_best_resolution Ran freedomintelligence/trim/llava/mm_utils.py
code served (permissive licence) · get_code("3999ff487573f32c")
split_list Ran freedomintelligence/trim/llava/eval/model_vqa.py
code served (permissive licence) · get_code("076c252c52cbb161")

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

The rapid advancement of Multimodal Large Language Models (MLLMs) has led to remarkable performances across various domains. However, this progress is accompanied by a substantial surge in the resource consumption of these models. We address this pressing issue by introducing a new approach, Token Reduction using CLIP Metric (TRIM), aimed at improving the efficiency of MLLMs without sacrificing their performance. Inspired by human attention patterns in Visual Question Answering (VQA) tasks, TRIM presents a fresh perspective on the selection and reduction of image tokens. The TRIM method has been extensively tested across 12 datasets, and the results demonstrate a significant reduction in computational overhead while maintaining a consistent level of performance. This research marks a critical stride in efficient MLLM development, promoting greater accessibility and sustainability of high-performing models.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2409.10994")
get_code_for_paper("2409.10994")
have("2409.10994")

Connect an agent — have() is free.