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

Added prints for stopping sequences, made makefile 1% friendlier to arch linux users

Browse source
This commit is contained in:
Concedo 2023-04-16 20:43:17 +08:00
parent 525184930d
commit 6548d3b3fb
4 changed files with 15 additions and 6 deletions
Download patch file Download diff file Expand all files Collapse all files

17
Makefile
View file

@ -10,6 +10,10 @@ ifndef UNAME_M
UNAME_M := $(shell uname -m)
endif
ifndef ARCH_LINUX
ARCH_LINUX := $(shell grep "Arch Linux" /etc/os-release 2>/dev/null)
endif
CCV := $(shell $(CC) --version | head -n 1)
CXXV := $(shell $(CXX) --version | head -n 1)
@ -40,7 +44,7 @@ BONUSCFLAGS1 =
BONUSCFLAGS2 =
#lets try enabling everything
CFLAGS += -pthread -s
CFLAGS += -pthread -s
CXXFLAGS += -pthread -s -Wno-multichar
# OS specific
@ -48,6 +52,9 @@ CXXFLAGS += -pthread -s -Wno-multichar
ifeq ($(UNAME_S),Linux)
CFLAGS += -pthread
CXXFLAGS += -pthread
ifdef ARCH_LINUX
LDFLAGS += -lcblas
endif
endif
ifeq ($(UNAME_S),Darwin)
CFLAGS += -pthread
@ -75,11 +82,11 @@ endif
# feel free to update the Makefile for your architecture and send a pull request or issue
ifeq ($(UNAME_M),$(filter $(UNAME_M),x86_64 i686))
# Use all CPU extensions that are available:
CFLAGS += -mavx
CFLAGS += -mavx
# old library NEEDS mf16c to work. so we must build with it. new one doesnt
ifeq ($(OS),Windows_NT)
BONUSCFLAGS1 += -mf16c
BONUSCFLAGS2 += -mavx2 -msse3 -mfma
BONUSCFLAGS1 += -mf16c
BONUSCFLAGS2 += -mavx2 -msse3 -mfma
else
# if not on windows, they are clearly building it themselves, so lets just use whatever is supported
CFLAGS += -march=native -mtune=native
@ -135,7 +142,7 @@ endif
OPENBLAS_BUILD =
CLBLAST_BUILD =
NOAVX2_BUILD =
NOAVX2_BUILD =
OPENBLAS_NOAVX2_BUILD =
ifeq ($(OS),Windows_NT)

1
gpttype_adapter.cpp
View file

@ -350,6 +350,7 @@ generation_outputs gpttype_generate(const generation_inputs inputs, generation_o
if (concat_output.find(matched) != std::string::npos)
{
remaining_tokens = 0;
printf("\n(Stop sequence triggered)");
break;
}
}

2
koboldcpp.py
View file

@ -435,7 +435,7 @@ def main(args):
RunServerMultiThreaded(args.host, args.port, embedded_kailite)
if __name__ == '__main__':
print("Welcome to KoboldCpp - Version 1.8") # just update version manually
print("Welcome to KoboldCpp - Version 1.9") # just update version manually
parser = argparse.ArgumentParser(description='Kobold llama.cpp server')
modelgroup = parser.add_mutually_exclusive_group() #we want to be backwards compatible with the unnamed positional args
modelgroup.add_argument("--model", help="Model file to load", nargs="?")

1
llama_adapter.cpp
View file

@ -246,6 +246,7 @@ generation_outputs llama_generate(const generation_inputs inputs, generation_out
if (concat_output.find(matched) != std::string::npos)
{
remaining_tokens = 0;
printf("\n(Stop sequence triggered)");
break;
}
}