vrr/koboldcpp
2
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2026-04-26 10:41:25 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 1

add support for llguidance (#1728)

Browse source 
* add llguidance

remove tab indentation for makefile if statements - these are dangerous
fix broken tool compilation commands
add USE_LLGUIDANCE env var to enable llguidance for faster structured
output generation
add llguidance as an optional submodule

* rm submodule
This commit is contained in:
tsite 2025-09-11 01:46:03 -07:00 • committed by GitHub
parent f7fa283bb6
commit 27c443f01e
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
2 changed files with 22 additions and 10 deletions
Download patch file Download diff file Expand all files Collapse all files

0
.gitmodules vendored Normal file
View file

32
Makefile
View file

@ -57,6 +57,7 @@ CXXFLAGS += -fsanitize=undefined -fsanitize-undefined-trap-on-error
endif
CFLAGS += -I. -Iggml/include -Iggml/src -Iggml/src/ggml-cpu -Iinclude -Isrc -I./common -I./vendor -I./vendor/stb -I./include -I./include/CL -I./otherarch -I./otherarch/tools -I./otherarch/sdcpp -I./otherarch/ttscpp/include -I./otherarch/ttscpp/src -I./otherarch/sdcpp/thirdparty -I./include/vulkan -O3 -fno-finite-math-only -std=c11 -fPIC -DLOG_DISABLE_LOGS -D_GNU_SOURCE -DGGML_USE_CPU -DGGML_USE_CPU_REPACK
CXXFLAGS += -I. -Iggml/include -Iggml/src -Iggml/src/ggml-cpu -Iinclude -Isrc -I./common -I./vendor -I./vendor/stb -I./include -I./include/CL -I./otherarch -I./otherarch/tools -I./otherarch/sdcpp -I./otherarch/ttscpp/include -I./otherarch/ttscpp/src -I./otherarch/sdcpp/thirdparty -I./include/vulkan -O3 -fno-finite-math-only -std=c++17 -fPIC -DLOG_DISABLE_LOGS -D_GNU_SOURCE -DGGML_USE_CPU -DGGML_USE_CPU_REPACK
ifndef KCPP_DEBUG
CFLAGS += -DNDEBUG -s
CXXFLAGS += -DNDEBUG -s
@ -74,6 +75,17 @@ CFLAGS += -pthread -Wno-deprecated -Wno-deprecated-declarations -Wno-unused-va
CXXFLAGS += -pthread -Wno-multichar -Wno-write-strings -Wno-deprecated -Wno-deprecated-declarations -Wno-unused-variable
LDFLAGS =
ifdef USE_LLGUIDANCE
CFLAGS += -DLLAMA_USE_LLGUIDANCE
CXXFLAGS += -DLLAMA_USE_LLGUIDANCE
LDFLAGS += -Lllguidance/target/release -lllguidance
OBJS_FULL += llguidance.o
llguidance.o: common/llguidance.cpp
cd llguidance && cargo build --release
$(CXX) $(CXXFLAGS) -Illguidance/target/release -c $< -o $@
endif
FASTCFLAGS = $(subst -O3,-Ofast,$(CFLAGS))
FASTCXXFLAGS = $(subst -O3,-Ofast,$(CXXFLAGS))
@ -300,10 +312,10 @@ endif # LLAMA_HIPBLAS
ifdef LLAMA_METAL
CFLAGS += -DGGML_USE_METAL -DGGML_METAL_NDEBUG -DSD_USE_METAL
CXXFLAGS += -DGGML_USE_METAL -DSD_USE_METAL
LDFLAGS += -framework Foundation -framework Metal -framework MetalKit -framework MetalPerformanceShaders
OBJS += ggml-metal.o
CFLAGS += -DGGML_USE_METAL -DGGML_METAL_NDEBUG -DSD_USE_METAL
CXXFLAGS += -DGGML_USE_METAL -DSD_USE_METAL
LDFLAGS += -framework Foundation -framework Metal -framework MetalKit -framework MetalPerformanceShaders
OBJS += ggml-metal.o
ggml-metal.o: ggml/src/ggml-metal/ggml-metal.m ggml/src/ggml-metal/ggml-metal-impl.h ggml/include/ggml-metal.h
@echo "== Preparing merged Metal file =="
@ -908,17 +920,17 @@ koboldcpp_vulkan_noavx2:
endif
# tools
quantize_gguf: tools/quantize/quantize.cpp ggml.o ggml-cpu.o ggml-ops.o ggml-vec.o ggml-binops.o ggml-unops.o llama.o llavaclip_default.o llava.o ggml-backend_default.o ggml-backend-reg_default.o ggml-repack.o $(OBJS_FULL)
quantize_gguf: tools/quantize/quantize.cpp ggml.o ggml-cpu.o ggml-ops.o ggml-vec.o ggml-binops.o ggml-unops.o llama.o llavaclip_default.o llava.o ggml-backend_default.o ggml-backend-reg_default.o ggml-repack.o $(OBJS_FULL) $(OBJS)
$(CXX) $(CXXFLAGS) $^ -o $@ $(LDFLAGS)
quantize_gptj: otherarch/tools/gptj_quantize.cpp otherarch/tools/common-ggml.cpp ggml_v3.o ggml.o ggml-cpu.o ggml-ops.o ggml-vec.o ggml-binops.o ggml-unops.o llama.o llavaclip_default.o llava.o ggml-backend_default.o ggml-backend-reg_default.o ggml-repack.o $(OBJS_FULL)
quantize_gptj: otherarch/tools/gptj_quantize.cpp otherarch/tools/common-ggml.cpp ggml_v3.o ggml.o ggml-cpu.o ggml-ops.o ggml-vec.o ggml-binops.o ggml-unops.o llama.o llavaclip_default.o llava.o ggml-backend_default.o ggml-backend-reg_default.o ggml-repack.o $(OBJS_FULL) $(OBJS)
$(CXX) $(CXXFLAGS) $^ -o $@ $(LDFLAGS)
quantize_gpt2: otherarch/tools/gpt2_quantize.cpp otherarch/tools/common-ggml.cpp ggml_v3.o ggml.o ggml-cpu.o ggml-ops.o ggml-vec.o ggml-binops.o ggml-unops.o llama.o llavaclip_default.o llava.o ggml-backend_default.o ggml-backend-reg_default.o ggml-repack.o $(OBJS_FULL)
quantize_gpt2: otherarch/tools/gpt2_quantize.cpp otherarch/tools/common-ggml.cpp ggml_v3.o ggml.o ggml-cpu.o ggml-ops.o ggml-vec.o ggml-binops.o ggml-unops.o llama.o llavaclip_default.o llava.o ggml-backend_default.o ggml-backend-reg_default.o ggml-repack.o $(OBJS_FULL) $(OBJS)
$(CXX) $(CXXFLAGS) $^ -o $@ $(LDFLAGS)
quantize_neox: otherarch/tools/neox_quantize.cpp otherarch/tools/common-ggml.cpp ggml_v3.o ggml.o ggml-cpu.o ggml-ops.o ggml-vec.o ggml-binops.o ggml-unops.o llama.o llavaclip_default.o llava.o ggml-backend_default.o ggml-backend-reg_default.o ggml-repack.o $(OBJS_FULL)
quantize_neox: otherarch/tools/neox_quantize.cpp otherarch/tools/common-ggml.cpp ggml_v3.o ggml.o ggml-cpu.o ggml-ops.o ggml-vec.o ggml-binops.o ggml-unops.o llama.o llavaclip_default.o llava.o ggml-backend_default.o ggml-backend-reg_default.o ggml-repack.o $(OBJS_FULL) $(OBJS)
$(CXX) $(CXXFLAGS) $^ -o $@ $(LDFLAGS)
quantize_mpt: otherarch/tools/mpt_quantize.cpp otherarch/tools/common-ggml.cpp ggml_v3.o ggml.o ggml-cpu.o ggml-ops.o ggml-vec.o ggml-binops.o ggml-unops.o llama.o llavaclip_default.o llava.o ggml-backend_default.o ggml-backend-reg_default.o ggml-repack.o $(OBJS_FULL)
quantize_mpt: otherarch/tools/mpt_quantize.cpp otherarch/tools/common-ggml.cpp ggml_v3.o ggml.o ggml-cpu.o ggml-ops.o ggml-vec.o ggml-binops.o ggml-unops.o llama.o llavaclip_default.o llava.o ggml-backend_default.o ggml-backend-reg_default.o ggml-repack.o $(OBJS_FULL) $(OBJS)
$(CXX) $(CXXFLAGS) $^ -o $@ $(LDFLAGS)
quantize_clip: tools/mtmd/clip.cpp tools/mtmd/clip.h tools/quantclip.cpp ggml_v3.o ggml.o ggml-cpu.o ggml-ops.o ggml-vec.o ggml-binops.o ggml-unops.o llama.o ggml-backend_default.o ggml-backend-reg_default.o ggml-repack.o $(OBJS_FULL)
quantize_clip: tools/mtmd/clip.cpp tools/quantclip.cpp ggml_v3.o ggml.o ggml-cpu.o ggml-ops.o ggml-vec.o ggml-binops.o ggml-unops.o llama.o ggml-backend_default.o ggml-backend-reg_default.o ggml-repack.o $(OBJS_FULL) $(OBJS)
$(CXX) $(CXXFLAGS) $^ -o $@ $(LDFLAGS)
#window simple clinfo