biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[NVIDIA] Support nvfp4 quantization (#12784)

Browse Source
This commit is contained in:
Kaixi Hou
2025-02-12 19:51:51 -08:00
committed by GitHub
parent 9f9704dca6
commit 4fc5c23bb6
12 changed files with 688 additions and 13 deletions
Download Patch File Download Diff File Expand all files Collapse all files

18
cmake/utils.cmake
View File

@@ -257,9 +257,9 @@ endmacro()
# where `<=` is the version comparison operator.
# In other words, for each version in `TGT_CUDA_ARCHS` find the highest version
# in `SRC_CUDA_ARCHS` that is less or equal to the version in `TGT_CUDA_ARCHS`.
# We have special handling for 9.0a, if 9.0a is in `SRC_CUDA_ARCHS` and 9.0 is
# in `TGT_CUDA_ARCHS` then we should remove 9.0a from `SRC_CUDA_ARCHS` and add
# 9.0a to the result (and remove 9.0 from TGT_CUDA_ARCHS).
# We have special handling for x.0a, if x.0a is in `SRC_CUDA_ARCHS` and x.0 is
# in `TGT_CUDA_ARCHS` then we should remove x.0a from `SRC_CUDA_ARCHS` and add
# x.0a to the result (and remove x.0 from TGT_CUDA_ARCHS).
# The result is stored in `OUT_CUDA_ARCHS`.
#
# Example:
@@ -272,8 +272,8 @@ function(cuda_archs_loose_intersection OUT_CUDA_ARCHS SRC_CUDA_ARCHS TGT_CUDA_AR
list(REMOVE_DUPLICATES SRC_CUDA_ARCHS)
set(TGT_CUDA_ARCHS_ ${TGT_CUDA_ARCHS})
# if 9.0a is in SRC_CUDA_ARCHS and 9.0 is in CUDA_ARCHS then we should
# remove 9.0a from SRC_CUDA_ARCHS and add 9.0a to _CUDA_ARCHS
# if x.0a is in SRC_CUDA_ARCHS and x.0 is in CUDA_ARCHS then we should
# remove x.0a from SRC_CUDA_ARCHS and add x.0a to _CUDA_ARCHS
set(_CUDA_ARCHS)
if ("9.0a" IN_LIST SRC_CUDA_ARCHS)
list(REMOVE_ITEM SRC_CUDA_ARCHS "9.0a")
@@ -283,6 +283,14 @@ function(cuda_archs_loose_intersection OUT_CUDA_ARCHS SRC_CUDA_ARCHS TGT_CUDA_AR
endif()
endif()
if ("10.0a" IN_LIST SRC_CUDA_ARCHS)
list(REMOVE_ITEM SRC_CUDA_ARCHS "10.0a")
if ("10.0" IN_LIST TGT_CUDA_ARCHS)
list(REMOVE_ITEM TGT_CUDA_ARCHS_ "10.0")
set(_CUDA_ARCHS "10.0a")
endif()
endif()
list(SORT SRC_CUDA_ARCHS COMPARE NATURAL ORDER ASCENDING)
# for each ARCH in TGT_CUDA_ARCHS find the highest arch in SRC_CUDA_ARCHS that