whisper : add GPU support via cuBLAS (#834)

* make : add WHISPER_CUBLAS

* make : fix CUBLAS build

* whisper : disable Flash Attention + adjust memory buffers

* whisper : remove old commented code

* readme : add cuBLAS instructions

* cmake : add WHISPER_CUBLAS option

* gitignore : ignore build-cublas
This commit is contained in:
Georgi Gerganov
2023-04-30 12:14:33 +03:00
committed by GitHub
parent 0ccd6746c9
commit 5fd1bdd7fc
10 changed files with 97 additions and 46 deletions

View File

@@ -1,3 +1,5 @@
default: main bench
ifndef UNAME_S
UNAME_S := $(shell uname -s)
endif
@@ -157,6 +159,18 @@ ifdef WHISPER_OPENBLAS
LDFLAGS += -lopenblas
endif
ifdef WHISPER_CUBLAS
CFLAGS += -DGGML_USE_CUBLAS -I/usr/local/cuda/include -I/opt/cuda/include -I$(CUDA_PATH)/targets/x86_64-linux/include
CXXFLAGS += -DGGML_USE_CUBLAS -I/usr/local/cuda/include -I/opt/cuda/include -I$(CUDA_PATH)/targets/x86_64-linux/include
LDFLAGS += -lcublas -lculibos -lcudart -lcublasLt -lpthread -ldl -lrt -L/usr/local/cuda/lib64 -L/opt/cuda/lib64 -L$(CUDA_PATH)/targets/x86_64-linux/lib
WHISPER_OBJ += ggml-cuda.o
NVCC = nvcc
NVCCFLAGS = --forward-unknown-to-host-compiler -arch=native
ggml-cuda.o: ggml-cuda.cu ggml-cuda.h
$(NVCC) $(NVCCFLAGS) $(CXXFLAGS) -Wno-pedantic -c $< -o $@
endif
ifdef WHISPER_GPROF
CFLAGS += -pg
CXXFLAGS += -pg
@@ -200,20 +214,18 @@ $(info I CC: $(CCV))
$(info I CXX: $(CXXV))
$(info )
default: main bench
#
# Build library
#
ggml.o: ggml.c ggml.h
$(CC) $(CFLAGS) -c ggml.c -o ggml.o
ggml.o: ggml.c ggml.h ggml-cuda.h
$(CC) $(CFLAGS) -c $< -o $@
whisper.o: whisper.cpp whisper.h ggml.h
$(CXX) $(CXXFLAGS) -c whisper.cpp -o whisper.o
whisper.o: whisper.cpp whisper.h ggml.h ggml-cuda.h
$(CXX) $(CXXFLAGS) -c $< -o $@
ifndef WHISPER_COREML
WHISPER_OBJ = whisper.o
WHISPER_OBJ += whisper.o
else
whisper-encoder.o: coreml/whisper-encoder.mm coreml/whisper-encoder.h
$(CXX) -O3 -I . -c coreml/whisper-encoder.mm -o whisper-encoder.o
@@ -221,7 +233,7 @@ whisper-encoder.o: coreml/whisper-encoder.mm coreml/whisper-encoder.h
whisper-encoder-impl.o: coreml/whisper-encoder-impl.m coreml/whisper-encoder-impl.h
$(CXX) -O3 -I . -fobjc-arc -c coreml/whisper-encoder-impl.m -o whisper-encoder-impl.o
WHISPER_OBJ = whisper.o whisper-encoder.o whisper-encoder-impl.o
WHISPER_OBJ += whisper.o whisper-encoder.o whisper-encoder-impl.o
endif
libwhisper.a: ggml.o $(WHISPER_OBJ)