[Core] Add Ascend Quant Config to main branch#33

Closed

Angazenn wants to merge 28 commits intovllm-project:mainfrom Angazenn:main

+436-1

Commits on Feb 7, 2025

add ascend quantize
angazenn
committed
add license
angazenn
committed
fix quantization import bugs
angazenn
committed
support skipping unquantized layers
angazenn
committed
fix ci
angazenn
committed
remove unnecessary params
angazenn
committed
fix import errors
angazenn
committed
avoid import check
angazenn
committed
add ascend quantization ut
angazenn
committed
fix quant description bugs
angazenn
committed
fix ut bugs
angazenn
committed
move quantizer initialization to linear method
angazenn
committed
fix bugs
angazenn
committed

Commits on Feb 8, 2025

narrow down the codes that try-catch structure covers when importing quantizer
angazenn
committed
move packages imported to the head
angazenn
committed
fix import bugs
angazenn
committed
fix import bugs
angazenn
committed
remove anti-outlier conditions from vllm_ascend
angazenn
committed
remove unnecessary spaces
angazenn
committed
add example quantization codes to ut
angazenn
committed

Commits on Feb 10, 2025

fix bugs
angazenn
committed
move importation of AscendQuantConfig
angazenn
committed

Commits on Feb 11, 2025

adapt to new config.json
angazenn
committed
Merge remote-tracking branch 'upstream/main'
angazenn
committed
move import of AsecndQuantConfig to pre_register
angazenn
committed
clean code
angazenn
committed

Commits on Feb 21, 2025

Merge remote-tracking branch 'upstream/main'
angazenn
committed
add int8 cache dtype when using attention quantization
angazenn
committed