Skip to content

FasterDecoding/SnapKV

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

28 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

SnapKV 📷

We introduce an innovative and out-of-box KV cache compression method, SnapKV.

Requirements

Currently tested with transformers==4.37.0, need to check if it is compatible with higher version.

transformers>=4.36
flash-attn==2.4.0

Installation

git clone [email protected]:FasterDecoding/SnapKV.git
cd SnapKV
pip install -e .

Quick Start

Use SnapKV-optimized Models

For example:

from snapkv.monkeypatch.monkeypatch import replace_mistral
replace_mistral() # Use monkey patches enable SnapKV

Check the example notebook.

Customize Your SnapKV-optimized Models

SnapKV can be easily integrated with other models.

You can follow the comment marked with [SnapKV] in existing models to construct your own models. (Currently we support Llama family/ Mistral/ Mixtral)

The detailed algorithm of SnapKV is in snapkv_utils.py

Partial Results

Comprehensive Experiment Results on LongBench Pressure Test Result on Needle-in-a-Haystack

TODO

  • Add observation experiments for reduplication.
  • Add LongBench for reduplication.
  • Explore the prompt phase compression.

Citation

If you feel this project is helpful, please consider cite our report 😊

@article{li2024snapkv,
  title={SnapKV: LLM Knows What You are Looking for Before Generation},
  author={Li, Yuhong and Huang, Yingbing and Yang, Bowen and Venkitesh, Bharat and Locatelli, Acyr and Ye, Hanchen and Cai, Tianle and Lewis, Patrick and Chen, Deming},
  journal={arXiv preprint arXiv:2404.14469},
  year={2024}
}