tomaarsen/attention_sinks
Extend existing LLMs way beyond the original training length with constant memory usage, without retraining
PythonApache-2.0
Stargazers
- aflah02Indraprastha Institute of Information Technology Delhi
- alexanderfreymarkets.sh GmbH
- alexbalandi
- benbenbenbenbenben
- cceydaKakaoStyle
- chainyo@owkin
- chuan298
- creatorrr@julep-ai
- dsindexhttps://github.com/kakaobrain
- fly51flyPRIS
- fursoviaEx-Human
- gagan3012
- ganesh3
- GeneZC
- Guangxuan-XiaoMIT
- JeffCarpenterCanada
- jordanparker6
- josaumJAI
- kgrozdanovski@cognirum
- LaxmanSinghTomarJiffyShirts.com
- learning-chip
- LysandreJikHugging Face
- manigithub-lab
- Mortadha-abderrhim
- MuhtashamTU Munich
- omarmhaimdatDatarise
- oplatekhttps://ufal.mff.cuni.cz/
- Oran-Ac@soarsmu | @RUCAIBOX
- SandalotsVolcanak
- sparverius
- tomaarsenHugging Face
- trannhatquyHUST
- webtaskforce3FosterAI
- Xu-KaiNational University of Singapore
- YangWang92
- younesbelkada@huggingface