cokeshao/HoliTom

[NeurIPS 2025] HoliTom: Holistic Token Merging for Fast Video Large Language Models

29
/ 100
Experimental

Processing long videos with large language models can be slow and expensive. This project helps reduce the computational resources needed by intelligently sifting out less important visual information from video data before and during its processing. It takes video inputs and delivers faster video analysis, making it easier for AI researchers and machine learning engineers to work with video LLMs efficiently.

No commits in the last 6 months.

Use this if you are a machine learning engineer or researcher developing or deploying video large language models and need to significantly speed up processing and reduce computational costs without sacrificing accuracy.

Not ideal if you are an end-user without a technical background in machine learning models, as this is a developer tool for optimizing existing video LLM infrastructure.

video-processing large-language-models machine-learning-optimization computational-efficiency AI-research
Stale 6m No Package No Dependents
Maintenance 2 / 25
Adoption 9 / 25
Maturity 15 / 25
Community 3 / 25

How are scores calculated?

Stars

72

Forks

1

Language

Python

License

Apache-2.0

Last pushed

Oct 10, 2025

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/transformers/cokeshao/HoliTom"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.