Public/koboldcpp
1
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2026-08-31 17:40:52 +02:00
Code Issues Packages Projects Releases Wiki Activity
Files
b84e210f0de90c7b657624f49cfc0a2eabaf4e37
koboldcpp/Makefile
T
Georgi Gerganov ec893798b7 llama : custom attention mask + parallel decoding + no context swaps (#3228) 
* tests : verify that RoPE is "additive"

* llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask)

* ggml : ggml_rope now takes a vector with positions instead of n_past

* metal : add rope_f16 kernel + optimize cpy kernels

* llama : unified KV cache + batch inference API

* llama : add new llama_decode() API that works with llama_batch

* llama : add cell_max heuristic for more efficient kv_cache

* llama : extend llama_kv_cache API

* llama : more robust cell_max heuristic + wip shift

* metal : disable concurrency optimization

* llama : add llama_kv_cache_shift_seq + no more context swaps

* llama : apply K-cache roping for Falcon and Baichuan

* speculative : fix KV cache management

* parallel : example for serving multiple users in parallel

* parallel : disable hot-plug to avoid cache fragmentation

* fixes : speculative KV cache + llama worst-case graph

* llama : extend batch API to select which logits to output

* llama : fix worst case graph build

* ggml-cuda : update rope implementation for parallel decoding (#3254)

* ggml-cuda : update rope implementation for parallel decoding

* better solution for p0 computation

* fix rope

* simpler rope implementation

---------

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>

* make : add parallel to build + fix static functions in llama.cpp

* simple : fix token counting

* parallel : various improvements

* llama : fix cell_max logic + rename functions

* parallel : try smaller batches when the KV cache is fragmented

* parallel : fix sequence termination criteria

* llama : silence errors KV cache errors

* parallel : remove new line from prompt

* parallel : process system prompt once + configurable paramters + llama API

* parallel : remove question with short answers

* parallel : count cache misses

* parallel : print misses on each request

* parallel : minor

* llama : fix n_kv to never become 0

* parallel : rename hot-plug to continuous-batching

* llama : improve llama_batch API + simplify parallel example

* simple : add parallel decoding support

* simple : improve comments + free batch

* ggml-cuda : add rope f16, restore performance with parallel decoding (#3272)

* ggml-cuda : add rope f16, restore performance

* offload KQ_mask with all models

* fix rope shift

---------

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>

* llama : disable MPI for now

ggml-ci

* train : make KQ_pos memory buffer permanent via dummy scale op

* ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275)

ggml-ci

* parallel : fix bug (extra BOS) + smaller token_prev array

* parallel : fix cases where the input prompts can overflow the batch

* parallel : add disabled experimental batch chunking in powers of two

* llama : llama.h formatting + comments

* simple : add README.md

* llama : fix kv cache heuristic when context is less than 32

* parallel : fix crash when `-n -1`

* llama : simplify returns if/else branches

* metal : use mm kernels for batch size > 2

* examples : utilize new llama_get_logits_ith()

* examples : add example for batched decoding

* examples : do not eval prompt 2 times (close #3348)

* server : clear the KV cache beyond n_past before llama_decode

* server : avoid context swaps by shifting the KV cache

---------

Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 19:04:36 +03:00

499 lines
18 KiB
Makefile
Raw Blame History

default: koboldcpp_default koboldcpp_failsafe koboldcpp_openblas koboldcpp_noavx2 koboldcpp_clblast koboldcpp_cublas koboldcpp_hipblas
tools: quantize_gpt2 quantize_gptj quantize_llama quantize_neox quantize_mpt
dev: koboldcpp_openblas
dev2: koboldcpp_clblast
ifndef UNAME_S
UNAME_S := $(shell uname -s)
endif
ifndef UNAME_P
UNAME_P := $(shell uname -p)
endif
ifndef UNAME_M
UNAME_M := $(shell uname -m)
endif
ifneq ($(shell grep -e "Arch Linux" -e "ID_LIKE=arch" /etc/os-release 2>/dev/null),)
ARCH_ADD = -lcblas
endif
# Mac OS + Arm can report x86_64
# ref: https://github.com/ggerganov/whisper.cpp/issues/66#issuecomment-1282546789
ifeq ($(UNAME_S),Darwin)
ifneq ($(UNAME_P),arm)
SYSCTL_M := $(shell sysctl -n hw.optional.arm64 2>/dev/null)
ifeq ($(SYSCTL_M),1)
# UNAME_P := arm
# UNAME_M := arm64
warn := $(warning Your arch is announced as x86_64, but it seems to actually be ARM64. Not fixing that can lead to bad performance. For more info see: https://github.com/ggerganov/whisper.cpp/issues/66\#issuecomment-1282546789)
endif
endif
endif
#
# Compile flags
#
# keep standard at C11 and C++11
CFLAGS = -I. -I./include -I./include/CL -I./otherarch -I./otherarch/tools -Ofast -DNDEBUG -std=c11 -fPIC -DLOG_DISABLE_LOGS -D_GNU_SOURCE
CXXFLAGS = -I. -I./common -I./include -I./include/CL -I./otherarch -I./otherarch/tools -Ofast -DNDEBUG -std=c++11 -fPIC -DLOG_DISABLE_LOGS -D_GNU_SOURCE
LDFLAGS =
ifndef LLAMA_NO_K_QUANTS
CFLAGS += -DGGML_USE_K_QUANTS
CXXFLAGS += -DGGML_USE_K_QUANTS
endif
# these are used on windows, to build some libraries with extra old device compatibility
SIMPLECFLAGS =
FULLCFLAGS =
NONECFLAGS =
OPENBLAS_FLAGS = -DGGML_USE_OPENBLAS -I/usr/local/include/openblas
CLBLAST_FLAGS = -DGGML_USE_CLBLAST
FAILSAFE_FLAGS = -DUSE_FAILSAFE
ifdef LLAMA_CUBLAS
CUBLAS_FLAGS = -DGGML_USE_CUBLAS
else
CUBLAS_FLAGS =
endif
CUBLASLD_FLAGS =
CUBLAS_OBJS =
#lets try enabling everything
CFLAGS += -pthread -s
CXXFLAGS += -pthread -s -Wno-multichar -Wno-write-strings
# OS specific
# TODO: support Windows
ifeq ($(UNAME_S),Linux)
CFLAGS += -pthread
CXXFLAGS += -pthread
endif
ifeq ($(UNAME_S),Darwin)
CFLAGS += -pthread
CXXFLAGS += -pthread
endif
ifeq ($(UNAME_S),FreeBSD)
CFLAGS += -pthread
CXXFLAGS += -pthread
endif
ifeq ($(UNAME_S),NetBSD)
CFLAGS += -pthread
CXXFLAGS += -pthread
endif
ifeq ($(UNAME_S),OpenBSD)
CFLAGS += -pthread
CXXFLAGS += -pthread
endif
ifeq ($(UNAME_S),Haiku)
CFLAGS += -pthread
CXXFLAGS += -pthread
endif
ifdef LLAMA_GPROF
CFLAGS += -pg
CXXFLAGS += -pg
endif
ifdef LLAMA_PERF
CFLAGS += -DGGML_PERF
CXXFLAGS += -DGGML_PERF
endif
# Architecture specific
# TODO: probably these flags need to be tweaked on some architectures
# feel free to update the Makefile for your architecture and send a pull request or issue
ifeq ($(UNAME_M),$(filter $(UNAME_M),x86_64 i686))
# Use all CPU extensions that are available:
# old library NEEDS mf16c to work. so we must build with it. new one doesnt
ifeq ($(OS),Windows_NT)
CFLAGS +=
NONECFLAGS +=
# -mno-sse3
SIMPLECFLAGS += -mavx -msse3
FULLCFLAGS += -mavx2 -msse3 -mfma -mf16c -mavx
else
# if not on windows, they are clearly building it themselves, so lets just use whatever is supported
CFLAGS += -march=native -mtune=native
endif
endif
ifneq ($(filter ppc64%,$(UNAME_M)),)
POWER9_M := $(shell grep "POWER9" /proc/cpuinfo)
ifneq (,$(findstring POWER9,$(POWER9_M)))
CFLAGS += -mcpu=power9
CXXFLAGS += -mcpu=power9
endif
# Require c++23's std::byteswap for big-endian support.
ifeq ($(UNAME_M),ppc64)
CXXFLAGS += -std=c++23 -DGGML_BIG_ENDIAN
endif
endif
ifndef LLAMA_NO_ACCELERATE
# Mac M1 - include Accelerate framework.
# `-framework Accelerate` works on Mac Intel as well, with negliable performance boost (as of the predict time).
ifeq ($(UNAME_S),Darwin)
CFLAGS += -DGGML_USE_ACCELERATE
LDFLAGS += -framework Accelerate
endif
endif
# it is recommended to use the CMAKE file to build for cublas if you can - will likely work better
ifdef LLAMA_CUBLAS
CUBLAS_FLAGS = -DGGML_USE_CUBLAS -I/usr/local/cuda/include -I/opt/cuda/include -I$(CUDA_PATH)/targets/x86_64-linux/include
CUBLASLD_FLAGS = -lcublas -lculibos -lcudart -lcublasLt -lpthread -ldl -lrt -L/usr/local/cuda/lib64 -L/opt/cuda/lib64 -L$(CUDA_PATH)/targets/x86_64-linux/lib
CUBLAS_OBJS = ggml-cuda.o ggml_v2-cuda.o ggml_v2-cuda-legacy.o
NVCC = nvcc
NVCCFLAGS = --forward-unknown-to-host-compiler -use_fast_math
ifdef CUDA_DOCKER_ARCH
NVCCFLAGS += -Wno-deprecated-gpu-targets -arch=$(CUDA_DOCKER_ARCH)
else
NVCCFLAGS += -arch=native
endif # CUDA_DOCKER_ARCH
ifdef LLAMA_CUDA_FORCE_DMMV
NVCCFLAGS += -DGGML_CUDA_FORCE_DMMV
endif # LLAMA_CUDA_FORCE_DMMV
ifdef LLAMA_CUDA_DMMV_X
NVCCFLAGS += -DGGML_CUDA_DMMV_X=$(LLAMA_CUDA_DMMV_X)
else
NVCCFLAGS += -DGGML_CUDA_DMMV_X=32
endif # LLAMA_CUDA_DMMV_X
ifdef LLAMA_CUDA_MMV_Y
NVCCFLAGS += -DGGML_CUDA_MMV_Y=$(LLAMA_CUDA_MMV_Y)
else ifdef LLAMA_CUDA_DMMV_Y
NVCCFLAGS += -DGGML_CUDA_MMV_Y=$(LLAMA_CUDA_DMMV_Y) # for backwards compatibility
else
NVCCFLAGS += -DGGML_CUDA_MMV_Y=1
endif # LLAMA_CUDA_MMV_Y
ifdef LLAMA_CUDA_F16
NVCCFLAGS += -DGGML_CUDA_F16
endif # LLAMA_CUDA_F16
ifdef LLAMA_CUDA_DMMV_F16
NVCCFLAGS += -DGGML_CUDA_F16
endif # LLAMA_CUDA_DMMV_F16
ifdef LLAMA_CUDA_KQUANTS_ITER
NVCCFLAGS += -DK_QUANTS_PER_ITERATION=$(LLAMA_CUDA_KQUANTS_ITER)
else
NVCCFLAGS += -DK_QUANTS_PER_ITERATION=2
endif
ifdef LLAMA_CUDA_MMQ_Y
NVCCFLAGS += -DGGML_CUDA_MMQ_Y=$(LLAMA_CUDA_MMQ_Y)
else
NVCCFLAGS += -DGGML_CUDA_MMQ_Y=64
endif # LLAMA_CUDA_MMQ_Y
#ifdef LLAMA_CUDA_CUBLAS
# NVCCFLAGS += -DGGML_CUDA_CUBLAS
#endif # LLAMA_CUDA_CUBLAS
ifdef LLAMA_CUDA_CCBIN
NVCCFLAGS += -ccbin $(LLAMA_CUDA_CCBIN)
endif
ggml-cuda.o: ggml-cuda.cu ggml-cuda.h
$(NVCC) $(NVCCFLAGS) $(subst -Ofast,-O3,$(CXXFLAGS)) $(CUBLAS_FLAGS) $(CUBLAS_CXXFLAGS) -Wno-pedantic -c $< -o $@
ggml_v2-cuda.o: otherarch/ggml_v2-cuda.cu otherarch/ggml_v2-cuda.h
$(NVCC) $(NVCCFLAGS) $(subst -Ofast,-O3,$(CXXFLAGS)) $(CUBLAS_FLAGS) $(CUBLAS_CXXFLAGS) -Wno-pedantic -c $< -o $@
ggml_v2-cuda-legacy.o: otherarch/ggml_v2-cuda-legacy.cu otherarch/ggml_v2-cuda-legacy.h
$(NVCC) $(NVCCFLAGS) $(subst -Ofast,-O3,$(CXXFLAGS)) $(CUBLAS_FLAGS) $(CUBLAS_CXXFLAGS) -Wno-pedantic -c $< -o $@
endif # LLAMA_CUBLAS
ifdef LLAMA_HIPBLAS
ROCM_PATH ?= /opt/rocm
CC := $(ROCM_PATH)/llvm/bin/clang
CXX := $(ROCM_PATH)/llvm/bin/clang++
GPU_TARGETS ?= gfx803 gfx900 gfx906 gfx908 gfx90a gfx1030 gfx1100 $(shell $(ROCM_PATH)/llvm/bin/amdgpu-arch)
LLAMA_CUDA_DMMV_X ?= 32
LLAMA_CUDA_MMV_Y ?= 2
LLAMA_CUDA_KQUANTS_ITER ?= 2
HIPFLAGS += -DGGML_USE_HIPBLAS -DGGML_USE_CUBLAS $(shell $(ROCM_PATH)/bin/hipconfig -C)
ifdef LLAMA_CUDA_FORCE_DMMV
HIPFLAGS += -DGGML_CUDA_FORCE_DMMV
endif # LLAMA_CUDA_FORCE_DMMV
HIPLDFLAGS += -L$(ROCM_PATH)/lib -Wl,-rpath=$(ROCM_PATH)/lib -lhipblas -lamdhip64 -lrocblas
HIP_OBJS += ggml-cuda.o ggml_v2-cuda.o ggml_v2-cuda-legacy.o
ggml-cuda.o: HIPFLAGS += $(addprefix --offload-arch=,$(GPU_TARGETS)) \
-DGGML_CUDA_DMMV_X=$(LLAMA_CUDA_DMMV_X) \
-DGGML_CUDA_MMV_Y=$(LLAMA_CUDA_MMV_Y) \
-DK_QUANTS_PER_ITERATION=$(LLAMA_CUDA_KQUANTS_ITER)
ggml_v2-cuda.o: HIPFLAGS += $(addprefix --offload-arch=,$(GPU_TARGETS)) \
-DGGML_CUDA_DMMV_X=$(LLAMA_CUDA_DMMV_X) \
-DGGML_CUDA_MMV_Y=$(LLAMA_CUDA_MMV_Y) \
-DK_QUANTS_PER_ITERATION=$(LLAMA_CUDA_KQUANTS_ITER)
ggml_v2-cuda-legacy.o: HIPFLAGS += $(addprefix --offload-arch=,$(GPU_TARGETS)) \
-DGGML_CUDA_DMMV_X=$(LLAMA_CUDA_DMMV_X) \
-DGGML_CUDA_MMV_Y=$(LLAMA_CUDA_MMV_Y) \
-DK_QUANTS_PER_ITERATION=$(LLAMA_CUDA_KQUANTS_ITER)
ggml-cuda.o: ggml-cuda.cu ggml-cuda.h
$(CXX) $(CXXFLAGS) $(HIPFLAGS) -x hip -c -o $@ $<
ggml_v2-cuda.o: otherarch/ggml_v2-cuda.cu otherarch/ggml_v2-cuda.h
$(CXX) $(CXXFLAGS) $(HIPFLAGS) -x hip -c -o $@ $<
ggml_v2-cuda-legacy.o: otherarch/ggml_v2-cuda-legacy.cu otherarch/ggml_v2-cuda-legacy.h
$(CXX) $(CXXFLAGS) $(HIPFLAGS) -x hip -c -o $@ $<
endif # LLAMA_HIPBLAS
ifdef LLAMA_METAL
CFLAGS += -DGGML_USE_METAL -DGGML_METAL_NDEBUG
CXXFLAGS += -DGGML_USE_METAL
LDFLAGS += -framework Foundation -framework Metal -framework MetalKit -framework MetalPerformanceShaders
OBJS += ggml-metal.o
ggml-metal.o: ggml-metal.m ggml-metal.h
$(CC) $(CFLAGS) -c $< -o $@
endif # LLAMA_METAL
ifneq ($(filter aarch64%,$(UNAME_M)),)
# Apple M1, M2, etc.
# Raspberry Pi 3, 4, Zero 2 (64-bit)
CFLAGS +=
CXXFLAGS +=
endif
ifneq ($(filter armv6%,$(UNAME_M)),)
# Raspberry Pi 1, Zero
CFLAGS += -mfpu=neon-fp-armv8 -mfp16-format=ieee -mno-unaligned-access
endif
ifneq ($(filter armv7%,$(UNAME_M)),)
# Raspberry Pi 2
CFLAGS += -mfpu=neon-fp-armv8 -mfp16-format=ieee -mno-unaligned-access -funsafe-math-optimizations
endif
ifneq ($(filter armv8%,$(UNAME_M)),)
# Raspberry Pi 3, 4, Zero 2 (32-bit)
CFLAGS += -mfp16-format=ieee -mno-unaligned-access
endif
CCV := $(shell $(CC) --version | head -n 1)
CXXV := $(shell $(CXX) --version | head -n 1)
DEFAULT_BUILD =
FAILSAFE_BUILD =
OPENBLAS_BUILD =
NOAVX2_BUILD =
CLBLAST_BUILD =
CUBLAS_BUILD =
HIPBLAS_BUILD =
ifeq ($(OS),Windows_NT)
DEFAULT_BUILD = $(CXX) $(CXXFLAGS) $^ -shared -o $@.dll $(LDFLAGS)
FAILSAFE_BUILD = $(CXX) $(CXXFLAGS) $^ -shared -o $@.dll $(LDFLAGS)
OPENBLAS_BUILD = $(CXX) $(CXXFLAGS) $^ lib/libopenblas.lib -shared -o $@.dll $(LDFLAGS)
NOAVX2_BUILD = $(CXX) $(CXXFLAGS) $^ -shared -o $@.dll $(LDFLAGS)
CLBLAST_BUILD = $(CXX) $(CXXFLAGS) $^ lib/OpenCL.lib lib/clblast.lib -shared -o $@.dll $(LDFLAGS)
ifdef LLAMA_CUBLAS
CUBLAS_BUILD = $(CXX) $(CXXFLAGS) $(CUBLAS_FLAGS) $^ -shared -o $@.dll $(CUBLASLD_FLAGS) $(LDFLAGS)
endif
ifdef LLAMA_HIPBLAS
HIPBLAS_BUILD = $(CXX) $(CXXFLAGS) $(HIPFLAGS) $^ -shared -o $@.dll $(HIPLDFLAGS) $(LDFLAGS)
endif
else
DEFAULT_BUILD = $(CXX) $(CXXFLAGS) $^ -shared -o $@.so $(LDFLAGS)
ifdef LLAMA_OPENBLAS
OPENBLAS_BUILD = $(CXX) $(CXXFLAGS) $^ $(ARCH_ADD) -lopenblas -shared -o $@.so $(LDFLAGS)
endif
ifdef LLAMA_CLBLAST
ifeq ($(UNAME_S),Darwin)
CLBLAST_BUILD = $(CXX) $(CXXFLAGS) $^ -lclblast -framework OpenCL $(ARCH_ADD) -lopenblas -shared -o $@.so $(LDFLAGS)
else
CLBLAST_BUILD = $(CXX) $(CXXFLAGS) $^ -lclblast -lOpenCL $(ARCH_ADD) -lopenblas -shared -o $@.so $(LDFLAGS)
endif
endif
ifdef LLAMA_CUBLAS
CUBLAS_BUILD = $(CXX) $(CXXFLAGS) $(CUBLAS_FLAGS) $^ -shared -o $@.so $(CUBLASLD_FLAGS) $(LDFLAGS)
endif
ifdef LLAMA_HIPBLAS
HIPBLAS_BUILD = $(CXX) $(CXXFLAGS) $(HIPFLAGS) $^ -shared -o $@.so $(HIPLDFLAGS) $(LDFLAGS)
endif
ifndef LLAMA_OPENBLAS
ifndef LLAMA_CLBLAST
ifndef LLAMA_CUBLAS
ifndef LLAMA_HIPBLAS
OPENBLAS_BUILD = @echo 'Your OS $(OS) does not appear to be Windows. For faster speeds, install and link a BLAS library. Set LLAMA_OPENBLAS=1 to compile with OpenBLAS support or LLAMA_CLBLAST=1 to compile with ClBlast support. This is just a reminder, not an error.'
endif
endif
endif
endif
endif
#
# Print build information
#
$(info I llama.cpp build info: )
$(info I UNAME_S: $(UNAME_S))
$(info I UNAME_P: $(UNAME_P))
$(info I UNAME_M: $(UNAME_M))
$(info I CFLAGS: $(CFLAGS))
$(info I CXXFLAGS: $(CXXFLAGS))
$(info I LDFLAGS: $(LDFLAGS))
$(info I CC: $(CCV))
$(info I CXX: $(CXXV))
$(info )
#
# Build library
#
ggml.o: ggml.c ggml.h ggml-cuda.h k_quants.h
$(CC) $(CFLAGS) $(FULLCFLAGS) -c $< -o $@
ggml_openblas.o: ggml.c ggml.h ggml-cuda.h k_quants.h
$(CC) $(CFLAGS) $(FULLCFLAGS) $(OPENBLAS_FLAGS) -c $< -o $@
ggml_failsafe.o: ggml.c ggml.h ggml-cuda.h k_quants.h
$(CC) $(CFLAGS) $(NONECFLAGS) -c $< -o $@
ggml_noavx2.o: ggml.c ggml.h ggml-cuda.h k_quants.h
$(CC) $(CFLAGS) $(SIMPLECFLAGS) -c $< -o $@
ggml_clblast.o: ggml.c ggml.h ggml-cuda.h k_quants.h
$(CC) $(CFLAGS) $(FULLCFLAGS) $(CLBLAST_FLAGS) -c $< -o $@
ggml_cublas.o: ggml.c ggml.h ggml-cuda.h k_quants.h
$(CC) $(CFLAGS) $(FULLCFLAGS) $(CUBLAS_FLAGS) $(HIPFLAGS) -c $< -o $@
#quants K
ifndef LLAMA_NO_K_QUANTS
k_quants.o: k_quants.c k_quants.h ggml.h ggml-cuda.h
$(CC) $(CFLAGS) $(FULLCFLAGS) -c $< -o $@
k_quants_noavx2.o: k_quants.c k_quants.h ggml.h ggml-cuda.h
$(CC) $(CFLAGS) $(SIMPLECFLAGS) -c $< -o $@
k_quants_failsafe.o: k_quants.c k_quants.h ggml.h ggml-cuda.h
$(CC) $(CFLAGS) $(NONECFLAGS) -c $< -o $@
else
k_quants.o:
k_quants_noavx2.o:
k_quants_failsafe.o:
endif # LLAMA_NO_K_QUANTS
#there's no intrinsics or special gpu ops used here, so we can have a universal object
ggml-alloc.o: ggml-alloc.c ggml.h ggml-alloc.h
$(CC) $(CFLAGS) -c $< -o $@
#version 2 libs
ggml_v2.o: otherarch/ggml_v2.c otherarch/ggml_v2.h
$(CC) $(CFLAGS) $(FULLCFLAGS) -c $< -o $@
ggml_v2_openblas.o: otherarch/ggml_v2.c otherarch/ggml_v2.h
$(CC) $(CFLAGS) $(FULLCFLAGS) $(OPENBLAS_FLAGS) -c $< -o $@
ggml_v2_failsafe.o: otherarch/ggml_v2.c otherarch/ggml_v2.h
$(CC) $(CFLAGS) $(NONECFLAGS) -c $< -o $@
ggml_v2_noavx2.o: otherarch/ggml_v2.c otherarch/ggml_v2.h
$(CC) $(CFLAGS) $(SIMPLECFLAGS) -c $< -o $@
ggml_v2_clblast.o: otherarch/ggml_v2.c otherarch/ggml_v2.h
$(CC) $(CFLAGS) $(FULLCFLAGS) $(CLBLAST_FLAGS) -c $< -o $@
ggml_v2_cublas.o: otherarch/ggml_v2.c otherarch/ggml_v2.h
$(CC) $(CFLAGS) $(FULLCFLAGS) $(CUBLAS_FLAGS) $(HIPFLAGS) -c $< -o $@
#extreme old version compat
ggml_v1.o: otherarch/ggml_v1.c otherarch/ggml_v1.h
$(CC) $(CFLAGS) $(FULLCFLAGS) -c $< -o $@
ggml_v1_failsafe.o: otherarch/ggml_v1.c otherarch/ggml_v1.h
$(CC) $(CFLAGS) $(NONECFLAGS) -c $< -o $@
#opencl
ggml-opencl.o: ggml-opencl.cpp ggml-opencl.h
$(CXX) $(CXXFLAGS) $(CLBLAST_FLAGS) -c $< -o $@
ggml_v2-opencl.o: otherarch/ggml_v2-opencl.cpp otherarch/ggml_v2-opencl.h
$(CXX) $(CXXFLAGS) $(CLBLAST_FLAGS) -c $< -o $@
ggml_v2-opencl-legacy.o: otherarch/ggml_v2-opencl-legacy.c otherarch/ggml_v2-opencl-legacy.h
$(CC) $(CFLAGS) -c $< -o $@
# intermediate objects
llama.o: llama.cpp ggml.h ggml-alloc.h ggml-cuda.h ggml-metal.h llama.h otherarch/llama-util.h
$(CXX) $(CXXFLAGS) -c $< -o $@
common.o: common/common.cpp common/common.h common/log.h
$(CXX) $(CXXFLAGS) -c $< -o $@
console.o: common/console.cpp common/console.h
$(CXX) $(CXXFLAGS) -c $< -o $@
grammar-parser.o: common/grammar-parser.cpp common/grammar-parser.h
$(CXX) $(CXXFLAGS) -c $< -o $@
expose.o: expose.cpp expose.h
$(CXX) $(CXXFLAGS) -c $< -o $@
# idiotic "for easier compilation"
GPTTYPE_ADAPTER = gpttype_adapter.cpp otherarch/llama_v2.cpp otherarch/llama_v3.cpp llama.cpp otherarch/utils.cpp otherarch/gptj_v1.cpp otherarch/gptj_v2.cpp otherarch/gptj_v3.cpp otherarch/gpt2_v1.cpp otherarch/gpt2_v2.cpp otherarch/gpt2_v3.cpp otherarch/rwkv_v2.cpp otherarch/rwkv_v3.cpp otherarch/neox_v2.cpp otherarch/neox_v3.cpp otherarch/mpt_v3.cpp ggml.h ggml-cuda.h llama.h otherarch/llama-util.h
gpttype_adapter_failsafe.o: $(GPTTYPE_ADAPTER)
$(CXX) $(CXXFLAGS) $(FAILSAFE_FLAGS) -c $< -o $@
gpttype_adapter.o: $(GPTTYPE_ADAPTER)
$(CXX) $(CXXFLAGS) -c $< -o $@
gpttype_adapter_clblast.o: $(GPTTYPE_ADAPTER)
$(CXX) $(CXXFLAGS) $(CLBLAST_FLAGS) -c $< -o $@
gpttype_adapter_cublas.o: $(GPTTYPE_ADAPTER)
$(CXX) $(CXXFLAGS) $(CUBLAS_FLAGS) $(HIPFLAGS) -c $< -o $@
clean:
rm -vf *.o main quantize_llama quantize_gpt2 quantize_gptj quantize_neox quantize_mpt quantize-stats perplexity embedding benchmark-matmult save-load-state gguf gguf.exe main.exe quantize_llama.exe quantize_gptj.exe quantize_gpt2.exe quantize_neox.exe quantize_mpt.exe koboldcpp_default.dll koboldcpp_openblas.dll koboldcpp_failsafe.dll koboldcpp_noavx2.dll koboldcpp_clblast.dll koboldcpp_cublas.dll koboldcpp_hipblas.dll koboldcpp_default.so koboldcpp_openblas.so koboldcpp_failsafe.so koboldcpp_noavx2.so koboldcpp_clblast.so koboldcpp_cublas.so koboldcpp_hipblas.so
main: examples/main/main.cpp build-info.h ggml.o k_quants.o ggml-alloc.o llama.o common.o console.o grammar-parser.o $(OBJS)
$(CXX) $(CXXFLAGS) $(filter-out %.h,$^) -o $@ $(LDFLAGS)
@echo
@echo '==== Run ./main -h for help. ===='
@echo
gguf: examples/gguf/gguf.cpp build-info.h ggml.o llama.o $(OBJS)
$(CXX) $(CXXFLAGS) $(filter-out %.h,$^) -o $@ $(LDFLAGS)
#generated libraries
koboldcpp_default: ggml.o ggml_v2.o ggml_v1.o expose.o common.o gpttype_adapter.o k_quants.o ggml-alloc.o grammar-parser.o $(OBJS)
$(DEFAULT_BUILD)
ifdef OPENBLAS_BUILD
koboldcpp_openblas: ggml_openblas.o ggml_v2_openblas.o ggml_v1.o expose.o common.o gpttype_adapter.o k_quants.o ggml-alloc.o grammar-parser.o $(OBJS)
$(OPENBLAS_BUILD)
else
koboldcpp_openblas:
endif
ifdef FAILSAFE_BUILD
koboldcpp_failsafe: ggml_failsafe.o ggml_v2_failsafe.o ggml_v1_failsafe.o expose.o common.o gpttype_adapter_failsafe.o k_quants_failsafe.o ggml-alloc.o grammar-parser.o $(OBJS)
$(FAILSAFE_BUILD)
else
koboldcpp_failsafe:
endif
ifdef NOAVX2_BUILD
koboldcpp_noavx2: ggml_noavx2.o ggml_v2_noavx2.o ggml_v1_failsafe.o expose.o common.o gpttype_adapter_failsafe.o k_quants_noavx2.o ggml-alloc.o grammar-parser.o $(OBJS)
$(NOAVX2_BUILD)
else
koboldcpp_noavx2:
endif
ifdef CLBLAST_BUILD
koboldcpp_clblast: ggml_clblast.o ggml_v2_clblast.o ggml_v1.o expose.o common.o gpttype_adapter_clblast.o ggml-opencl.o ggml_v2-opencl.o ggml_v2-opencl-legacy.o k_quants.o ggml-alloc.o grammar-parser.o $(OBJS)
$(CLBLAST_BUILD)
else
koboldcpp_clblast:
endif
ifdef CUBLAS_BUILD
koboldcpp_cublas: ggml_cublas.o ggml_v2_cublas.o ggml_v1.o expose.o common.o gpttype_adapter_cublas.o k_quants.o ggml-alloc.o grammar-parser.o $(CUBLAS_OBJS) $(OBJS)
$(CUBLAS_BUILD)
else
koboldcpp_cublas:
endif
ifdef HIPBLAS_BUILD
koboldcpp_hipblas: ggml_cublas.o ggml_v2_cublas.o ggml_v1.o expose.o common.o gpttype_adapter_cublas.o k_quants.o ggml-alloc.o grammar-parser.o $(HIP_OBJS) $(OBJS)
$(HIPBLAS_BUILD)
else
koboldcpp_hipblas:
endif
# tools
quantize_llama: examples/quantize/quantize.cpp ggml.o llama.o k_quants.o ggml-alloc.o
$(CXX) $(CXXFLAGS) $^ -o $@ $(LDFLAGS)
quantize_gptj: ggml.o llama.o k_quants.o ggml-alloc.o otherarch/tools/gptj_quantize.cpp otherarch/tools/common-ggml.cpp
$(CXX) $(CXXFLAGS) $^ -o $@ $(LDFLAGS)
quantize_gpt2: ggml.o llama.o k_quants.o ggml-alloc.o otherarch/tools/gpt2_quantize.cpp otherarch/tools/common-ggml.cpp
$(CXX) $(CXXFLAGS) $^ -o $@ $(LDFLAGS)
quantize_neox: ggml.o llama.o k_quants.o ggml-alloc.o otherarch/tools/neox_quantize.cpp otherarch/tools/common-ggml.cpp
$(CXX) $(CXXFLAGS) $^ -o $@ $(LDFLAGS)
quantize_mpt: ggml.o llama.o k_quants.o ggml-alloc.o otherarch/tools/mpt_quantize.cpp otherarch/tools/common-ggml.cpp
$(CXX) $(CXXFLAGS) $^ -o $@ $(LDFLAGS)
build-info.h:
$(DONOTHING)
Reference in New Issue View Git Blame Copy Permalink