LightCompress
[EMNLP 2024 & AAAI 2026] A powerful toolkit for compressing large models including LLMs, VLMs, and video generative models.
- 755
- Python
- Apache-2.0
- 6 mo ago
[EMNLP 2024 & AAAI 2026] A powerful toolkit for compressing large models including LLMs, VLMs, and video generative models.
This is a collection of our research on efficient AI, covering hardware-aware NAS and model compression.
Official implementation of CVPR 2024 paper "vid-TLDR: Training Free Token merging for Light-weight Video Transformer".
Official Implementation (Pytorch) of the "Representation Shift: Unifying Token Compression with FlashAttention", ICCV 2025
Glance, then Select: Evidence-Tempered Visual Token Selection for Extremely Compressed Multimodal LLMs
Hard-coded attention sinks in vision encoders (AIMv2, SigLIP So400m, PaliGemma, LLaVA-OneVision): where they live, how they work, and how to move them