PyPI page
Home page
Author:
Simon Jegou, Maximilian Jeblick, Alessio Devoto, David Austin
Summary:
Efficiently compress the KV cache of any pretrained transformer
Latest version:
0.5.4
Required dependencies:
accelerate
|
cachetools
|
datasets
|
fire
|
numpy
|
pandas
|
requests
|
torch
|
transformers
Optional dependencies:
bert-score
|
flash-attn
|
fuzzywuzzy
|
jieba
|
nltk
|
rouge
|
scipy
|
tqdm
Downloads last day:
123
Downloads last week:
532
Downloads last month:
2,100