[Public release 26/04] Introducing Mega MoE, FP4 Indexer and other features/fixes (#304)

* Merge with private repo

* Update README

* Update README

* Update README

* Add PyTorch requirements

* Fix sync scopes for MQA logits (#256)

* Update README
This commit is contained in:
Chenggang Zhao
2026-04-17 09:45:14 +08:00
committed by GitHub
parent d30fc36c8f
commit 7f2a703ed5
109 changed files with 12101 additions and 3219 deletions

View File

@@ -3,8 +3,7 @@ cmake_minimum_required(VERSION 3.10)
project(deep_gemm LANGUAGES CXX CUDA)
set(CMAKE_VERBOSE_MAKEFILE ON)
set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -O3 -fPIC -Wno-psabi")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -O3 -fPIC -Wno-psabi")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -O3 -fPIC -Wno-psabi -Wno-deprecated-declarations")
set(CUDA_SEPARABLE_COMPILATION ON)
list(APPEND CUDA_NVCC_FLAGS "-DENABLE_FAST_DEBUG")
list(APPEND CUDA_NVCC_FLAGS "-O3")
@@ -18,11 +17,11 @@ find_package(CUDAToolkit REQUIRED)
find_package(pybind11 REQUIRED)
find_package(Torch REQUIRED)
set(CMAKE_CXX_STANDARD 17)
set(CMAKE_CUDA_STANDARD 17)
set(CMAKE_CXX_STANDARD 20)
set(CMAKE_CUDA_STANDARD 20)
include_directories(deep_gemm/include third-party/cutlass/include third-party/cutlass/tools/util/include third-party/fmt/include)
include_directories(${CUDA_TOOLKIT_ROOT_DIR}/targets/x86_64-linux/include ${TORCH_INCLUDE_DIRS} ${PYTHON_INCLUDE_DIRS})
include_directories(${CUDA_TOOLKIT_ROOT_DIR}/include/cccl ${TORCH_INCLUDE_DIRS} ${PYTHON_INCLUDE_DIRS})
link_directories(${TORCH_INSTALL_PREFIX}/lib ${CUDA_TOOLKIT_ROOT_DIR}/lib64 ${CUDA_TOOLKIT_ROOT_DIR}/lib64/stubs)
# The main Python API entrance