vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

build : rename flag GGML_CUDA_USE_GRAPHS -> GGML_CUDA_GRAPHS

Browse source 
for consistency

ggml-ci
This commit is contained in:
Georgi Gerganov 2024-09-16 09:00:43 +03:00
parent 2ac8a91fbe
commit f80e679696
No known key found for this signature in database
GPG key ID: 449E073F9DC10735
4 changed files with 7 additions and 7 deletions
Download patch file Download diff file Expand all files Collapse all files

4
CMakeLists.txt
View file

@ -85,8 +85,8 @@ if (NOT DEFINED GGML_LLAMAFILE)
set(GGML_LLAMAFILE_DEFAULT ON) set(GGML_LLAMAFILE_DEFAULT ON)
endif() endif()
if (NOT DEFINED GGML_CUDA_USE_GRAPHS) if (NOT DEFINED GGML_CUDA_GRAPHS)
set(GGML_CUDA_USE_GRAPHS_DEFAULT ON) set(GGML_CUDA_GRAPHS_DEFAULT ON)
endif() endif()
# transition helpers # transition helpers

2
Makefile
View file

@ -619,7 +619,7 @@ ifdef GGML_CUDA
CUDA_PATH ?= /usr/local/cuda CUDA_PATH ?= /usr/local/cuda
endif endif
MK_CPPFLAGS += -DGGML_USE_CUDA -I$(CUDA_PATH)/include -I$(CUDA_PATH)/targets/$(UNAME_M)-linux/include -DGGML_CUDA_USE_GRAPHS MK_CPPFLAGS += -DGGML_USE_CUDA -DGGML_CUDA_USE_GRAPHS -I$(CUDA_PATH)/include -I$(CUDA_PATH)/targets/$(UNAME_M)-linux/include
MK_LDFLAGS += -lcuda -lcublas -lculibos -lcudart -lcublasLt -lpthread -ldl -lrt -L$(CUDA_PATH)/lib64 -L/usr/lib64 -L$(CUDA_PATH)/targets/$(UNAME_M)-linux/lib -L$(CUDA_PATH)/lib64/stubs -L/usr/lib/wsl/lib MK_LDFLAGS += -lcuda -lcublas -lculibos -lcudart -lcublasLt -lpthread -ldl -lrt -L$(CUDA_PATH)/lib64 -L/usr/lib64 -L$(CUDA_PATH)/targets/$(UNAME_M)-linux/lib -L$(CUDA_PATH)/lib64/stubs -L/usr/lib/wsl/lib
MK_NVCCFLAGS += -use_fast_math MK_NVCCFLAGS += -use_fast_math
endif # GGML_MUSA endif # GGML_MUSA

6
ggml/CMakeLists.txt
View file

@ -61,8 +61,8 @@ if (NOT GGML_LLAMAFILE_DEFAULT)
set(GGML_LLAMAFILE_DEFAULT OFF) set(GGML_LLAMAFILE_DEFAULT OFF)
endif() endif()
if (NOT GGML_CUDA_USE_GRAPHS_DEFAULT) if (NOT GGML_CUDA_GRAPHS_DEFAULT)
set(GGML_CUDA_USE_GRAPHS_DEFAULT OFF) set(GGML_CUDA_GRAPHS_DEFAULT OFF)
endif() endif()
# general # general
@ -136,7 +136,7 @@ set (GGML_CUDA_PEER_MAX_BATCH_SIZE "128" CACHE STRING
option(GGML_CUDA_NO_PEER_COPY "ggml: do not use peer to peer copies" OFF) option(GGML_CUDA_NO_PEER_COPY "ggml: do not use peer to peer copies" OFF)
option(GGML_CUDA_NO_VMM "ggml: do not try to use CUDA VMM" OFF) option(GGML_CUDA_NO_VMM "ggml: do not try to use CUDA VMM" OFF)
option(GGML_CUDA_FA_ALL_QUANTS "ggml: compile all quants for FlashAttention" OFF) option(GGML_CUDA_FA_ALL_QUANTS "ggml: compile all quants for FlashAttention" OFF)
option(GGML_CUDA_USE_GRAPHS "ggml: use CUDA graphs (llama.cpp only)" ${GGML_CUDA_USE_GRAPHS_DEFAULT}) option(GGML_CUDA_GRAPHS "ggml: use CUDA graphs (llama.cpp only)" ${GGML_CUDA_GRAPHS_DEFAULT})
option(GGML_HIPBLAS "ggml: use hipBLAS" OFF) option(GGML_HIPBLAS "ggml: use hipBLAS" OFF)
option(GGML_HIP_UMA "ggml: use HIP unified memory architecture" OFF) option(GGML_HIP_UMA "ggml: use HIP unified memory architecture" OFF)

2
ggml/src/CMakeLists.txt
View file

@ -329,7 +329,7 @@ if (GGML_CUDA)
add_compile_definitions(K_QUANTS_PER_ITERATION=${GGML_CUDA_KQUANTS_ITER}) add_compile_definitions(K_QUANTS_PER_ITERATION=${GGML_CUDA_KQUANTS_ITER})
add_compile_definitions(GGML_CUDA_PEER_MAX_BATCH_SIZE=${GGML_CUDA_PEER_MAX_BATCH_SIZE}) add_compile_definitions(GGML_CUDA_PEER_MAX_BATCH_SIZE=${GGML_CUDA_PEER_MAX_BATCH_SIZE})
if (GGML_CUDA_USE_GRAPHS) if (GGML_CUDA_GRAPHS)
add_compile_definitions(GGML_CUDA_USE_GRAPHS) add_compile_definitions(GGML_CUDA_USE_GRAPHS)
endif() endif()