Code

Quantized KV Cache#30483

Code

Merged

zucchini-nlp merged 21 commits intohuggingface:mainhuggingface/transformers:mainfrom zucchini-nlp:quantzucchini-nlp/transformers:quantCopy head branch name to clipboardMay 23, 2024

+652-28

Commits

Commits on May 10, 2024

• clean-up

  zucchini-nlp

  committed
  16731be

• Update src/transformers/cache_utils.py

  Show description for cf00de6
  

  zucchini-nlp

  and

  ArthurZucker

  authored
  cf00de6

• Update src/transformers/cache_utils.py

  Show description for 6de0d8a
  

  zucchini-nlp

  and

  ArthurZucker

  authored
  6de0d8a

• Update src/transformers/cache_utils.py

  Show description for 5a87cbb
  

  zucchini-nlp

  and

  ArthurZucker

  authored
  5a87cbb

• fixup

  zucchini-nlp

  committed
  bfe0804

Commits on May 13, 2024

• Update tests/quantization/quanto_integration/test_quanto.py

  Show description for 62abc33
  

  zucchini-nlp

  and

  younesbelkada

  authored
  62abc33

• Update src/transformers/generation/configuration_utils.py

  Show description for 519682c
  

  zucchini-nlp

  and

  ArthurZucker

  authored
  519682c

• more suggestions

  zucchini-nlp

  committed
  6f19cee

• mapping if torch available

  zucchini-nlp

  committed
  2b8e042

Commits on May 14, 2024

• Merge branch 'main' into quant

  zucchini-nlp

  authored
  4c65b5b

• run tests & add 'support_quantized' flag

  zucchini-nlp

  committed
  1d9cf15

• Merge branch 'main' into quant

  zucchini-nlp

  authored
  a58aa9d

• fix jamba test

  zucchini-nlp

  committed
  d181339

• revert, will be fixed by another PR

  zucchini-nlp

  committed
  e658c9f

Commits on May 18, 2024

• Merge branch 'main' into quant

  zucchini-nlp

  authored
  9397dda

• codestyle

  zucchini-nlp

  committed
  045e406

Commits on May 20, 2024

• HQQ and versatile cache classes

  zucchini-nlp

  committed
  3193b43

Commits on May 21, 2024

• final update

  zucchini-nlp

  committed
  126ce84

Commits on May 23, 2024

• Merge "main"

  zucchini-nlp

  committed
  0d1df5f

• typo

  zucchini-nlp

  committed
  89413d3

• make tests happy

  zucchini-nlp

  committed
  9d0f668