mirrors/cosmopolitan
1
0
Fork 0
mirror of https://github.com/jart/cosmopolitan.git synced 2025-01-31 19:43:32 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions
cosmopolitan/third_party/ggml/ggml.mk

197 lines
6.2 KiB
Makefile
Raw Normal View History

Import llama.cpp https://github.com/ggerganov/llama.cpp 0b2da20538d01926b77ea237dd1c930c4d20b686 See third_party/ggml/README.cosmo for changes
2023-04-27 21:31:20 +00:00
#-*-mode:makefile-gmake;indent-tabs-mode:t;tab-width:8;coding:utf-8-*-┐
#───vi: set et ft=make ts=8 tw=8 fenc=utf-8 :vi───────────────────────┘
PKGS += THIRD_PARTY_GGML
################################################################################
# single file machine learning framework written in c
# make -j8 o//third_party/ggml/ggml.a
THIRD_PARTY_GGML_ARTIFACTS += THIRD_PARTY_GGML_A
THIRD_PARTY_GGML = $(THIRD_PARTY_GGML_A_DEPS) $(THIRD_PARTY_GGML_A)
THIRD_PARTY_GGML_A = o/$(MODE)/third_party/ggml/ggml.a
THIRD_PARTY_GGML_A_OBJS = $(THIRD_PARTY_GGML_A_SRCS:%.c=o/$(MODE)/%.o)
THIRD_PARTY_GGML_A_FILES = $(THIRD_PARTY_GGML_A_SRCS) $(THIRD_PARTY_GGML_A_HDRS)
THIRD_PARTY_GGML_A_CHECKS = $(THIRD_PARTY_GGML_A).pkg $(THIRD_PARTY_GGML_A_HDRS:%=o/$(MODE)/%.ok)
Add support for new GGJT v2 quantizers This change makes quantized models (e.g. q4_0) go 10% faster on Macs however doesn't offer much improvement for Intel PC hardware. This change syncs llama.cpp 699b1ad7fe6f7b9e41d3cb41e61a8cc3ea5fc6b5 which recently made a breaking change to nearly all its file formats without any migration. Since that'll break hundreds upon hundreds of models on websites like HuggingFace llama.com will support both file formats because llama.com will never ever break the GGJT file format
2023-05-13 15:08:32 +00:00
THIRD_PARTY_GGML_A_HDRS = \
third_party/ggml/fp16.h \
third_party/ggml/ggml.h \
third_party/ggml/ggjt.v1.q4_0.h \
third_party/ggml/ggjt.v1.q4_1.h \
third_party/ggml/ggjt.v1.q4_2.h \
third_party/ggml/ggjt.v1.q5_0.h \
third_party/ggml/ggjt.v1.q5_1.h \
third_party/ggml/ggjt.v1.q8_0.h \
third_party/ggml/ggjt.v1.q8_1.h \
third_party/ggml/fp16.internal.h \
third_party/ggml/ggjt.v1.internal.h
THIRD_PARTY_GGML_A_SRCS = \
third_party/ggml/fp16.c \
third_party/ggml/ggml.c \
third_party/ggml/ggjt.v1.c \
third_party/ggml/ggjt.v1.q4_0.c \
third_party/ggml/ggjt.v1.q4_1.c \
third_party/ggml/ggjt.v1.q4_2.c \
third_party/ggml/ggjt.v1.q5_0.c \
third_party/ggml/ggjt.v1.q5_1.c \
third_party/ggml/ggjt.v1.q8_0.c \
third_party/ggml/ggjt.v1.q8_1.c
Use Companion AI in llama.com by default
2023-04-29 07:48:14 +00:00
THIRD_PARTY_GGML_A_DIRECTDEPS = \
LIBC_CALLS \
LIBC_INTRIN \
LIBC_MEM \
LIBC_NEXGEN32E \
LIBC_RUNTIME \
LIBC_STDIO \
LIBC_THREAD \
LIBC_STR \
LIBC_STUBS \
LIBC_SYSV \
Get llama.com building as an aarch64 native binary
2023-05-09 19:14:57 +00:00
LIBC_TINYMATH \
THIRD_PARTY_COMPILER_RT
Import llama.cpp https://github.com/ggerganov/llama.cpp 0b2da20538d01926b77ea237dd1c930c4d20b686 See third_party/ggml/README.cosmo for changes
2023-04-27 21:31:20 +00:00
Use Companion AI in llama.com by default
2023-04-29 07:48:14 +00:00
THIRD_PARTY_GGML_A_DEPS := \
Import llama.cpp https://github.com/ggerganov/llama.cpp 0b2da20538d01926b77ea237dd1c930c4d20b686 See third_party/ggml/README.cosmo for changes
2023-04-27 21:31:20 +00:00
$(call uniq,$(foreach x,$(THIRD_PARTY_GGML_A_DIRECTDEPS),$($(x))))
Use Companion AI in llama.com by default
2023-04-29 07:48:14 +00:00
$(THIRD_PARTY_GGML_A): \
third_party/ggml/ \
$(THIRD_PARTY_GGML_A).pkg \
Import llama.cpp https://github.com/ggerganov/llama.cpp 0b2da20538d01926b77ea237dd1c930c4d20b686 See third_party/ggml/README.cosmo for changes
2023-04-27 21:31:20 +00:00
$(THIRD_PARTY_GGML_A_OBJS)
Use Companion AI in llama.com by default
2023-04-29 07:48:14 +00:00
$(THIRD_PARTY_GGML_A).pkg: \
$(THIRD_PARTY_GGML_A_OBJS) \
Import llama.cpp https://github.com/ggerganov/llama.cpp 0b2da20538d01926b77ea237dd1c930c4d20b686 See third_party/ggml/README.cosmo for changes
2023-04-27 21:31:20 +00:00
$(foreach x,$(THIRD_PARTY_GGML_A_DIRECTDEPS),$($(x)_A).pkg)
Use Companion AI in llama.com by default
2023-04-29 07:48:14 +00:00
$(THIRD_PARTY_GGML_A_OBJS): private \
OVERRIDE_CFLAGS += \
-O3 \
-ffunction-sections \
Get LIBC_MEM and LIBC_STDIO building with aarch64
2023-05-09 15:08:56 +00:00
-fdata-sections
ifeq ($(ARCH), x86_64)
$(THIRD_PARTY_GGML_A_OBJS): private \
OVERRIDE_CFLAGS += \
Use Companion AI in llama.com by default
2023-04-29 07:48:14 +00:00
-msse3 \
-mavx \
-mavx2 \
-mf16c \
Import llama.cpp https://github.com/ggerganov/llama.cpp 0b2da20538d01926b77ea237dd1c930c4d20b686 See third_party/ggml/README.cosmo for changes
2023-04-27 21:31:20 +00:00
-mfma
Get LIBC_MEM and LIBC_STDIO building with aarch64
2023-05-09 15:08:56 +00:00
endif
Import llama.cpp https://github.com/ggerganov/llama.cpp 0b2da20538d01926b77ea237dd1c930c4d20b686 See third_party/ggml/README.cosmo for changes
2023-04-27 21:31:20 +00:00
Remove division from matrix multiplication This change reduces llama.com CPU cycles systemically by 2.5% according to the Linux Kernel `perf stat -Bddd` utility.
2023-05-11 04:19:54 +00:00
o/opt/third_party/ggml/ggml.o: private \
OVERRIDE_CFLAGS += \
-x-no-pg
ifeq ($(ARCH), x86_64)
o/rel/third_party/ggml/ggml.o \
o/opt/third_party/ggml/ggml.o: private \
OVERRIDE_CFLAGS += \
-fschedule-insns2 \
-mred-zone
endif
Import llama.cpp https://github.com/ggerganov/llama.cpp 0b2da20538d01926b77ea237dd1c930c4d20b686 See third_party/ggml/README.cosmo for changes
2023-04-27 21:31:20 +00:00
################################################################################
# command for running inference on large language models
# make -j8 o//third_party/ggml/llama.com
THIRD_PARTY_GGML_ARTIFACTS += THIRD_PARTY_GGML_LLAMA
THIRD_PARTY_GGML_LLAMA = o/$(MODE)/third_party/ggml/llama.com
THIRD_PARTY_GGML_LLAMA_OBJS = $(THIRD_PARTY_GGML_LLAMA_SRCS:%.cc=o/$(MODE)/%.o)
THIRD_PARTY_GGML_LLAMA_FILES := $(THIRD_PARTY_GGML_LLAMA_SRCS) $(THIRD_PARTY_GGML_LLAMA_HDRS)
THIRD_PARTY_GGML_LLAMA_CHECKS = $(THIRD_PARTY_GGML_LLAMA).pkg $(THIRD_PARTY_GGML_LLAMA_HDRS:%=o/$(MODE)/%.okk)
Perform some code cleanup
2023-05-15 23:32:10 +00:00
THIRD_PARTY_GGML_LLAMA_HDRS = \
third_party/ggml/common.cc \
third_party/ggml/llama.h \
third_party/ggml/llama_util.h \
third_party/ggml/common.h
THIRD_PARTY_GGML_LLAMA_SRCS = \
third_party/ggml/main.cc \
third_party/ggml/llama.cc \
third_party/ggml/common.cc \
third_party/ggml/quantize.cc \
third_party/ggml/perplexity.cc
Use Companion AI in llama.com by default
2023-04-29 07:48:14 +00:00
THIRD_PARTY_GGML_LLAMA_DIRECTDEPS = \
LIBC_CALLS \
LIBC_FMT \
LIBC_INTRIN \
LIBC_MEM \
LIBC_NEXGEN32E \
LIBC_RUNTIME \
LIBC_STDIO \
LIBC_LOG \
LIBC_STR \
LIBC_STUBS \
LIBC_SYSV \
LIBC_THREAD \
LIBC_TINYMATH \
LIBC_ZIPOS \
THIRD_PARTY_GGML \
Import llama.cpp https://github.com/ggerganov/llama.cpp 0b2da20538d01926b77ea237dd1c930c4d20b686 See third_party/ggml/README.cosmo for changes
2023-04-27 21:31:20 +00:00
THIRD_PARTY_LIBCXX
Use Companion AI in llama.com by default
2023-04-29 07:48:14 +00:00
THIRD_PARTY_GGML_LLAMA_DEPS := \
Import llama.cpp https://github.com/ggerganov/llama.cpp 0b2da20538d01926b77ea237dd1c930c4d20b686 See third_party/ggml/README.cosmo for changes
2023-04-27 21:31:20 +00:00
$(call uniq,$(foreach x,$(THIRD_PARTY_GGML_LLAMA_DIRECTDEPS),$($(x))))
Use Companion AI in llama.com by default
2023-04-29 07:48:14 +00:00
$(THIRD_PARTY_GGML_LLAMA).dbg: \
$(THIRD_PARTY_GGML_LLAMA).pkg \
$(THIRD_PARTY_GGML_LLAMA_DEPS) \
o/$(MODE)/third_party/ggml/companionai.txt.zip.o \
o/$(MODE)/third_party/ggml/common.o \
o/$(MODE)/third_party/ggml/llama.o \
o/$(MODE)/third_party/ggml/main.o \
$(CRT) \
Import llama.cpp https://github.com/ggerganov/llama.cpp 0b2da20538d01926b77ea237dd1c930c4d20b686 See third_party/ggml/README.cosmo for changes
2023-04-27 21:31:20 +00:00
$(APE_NO_MODIFY_SELF)
@$(APELINK)
Perform some code cleanup
2023-05-15 23:32:10 +00:00
o/$(MODE)/third_party/ggml/quantize.com.dbg: \
$(THIRD_PARTY_GGML_LLAMA).pkg \
$(THIRD_PARTY_GGML_LLAMA_DEPS) \
o/$(MODE)/third_party/ggml/common.o \
o/$(MODE)/third_party/ggml/llama.o \
o/$(MODE)/third_party/ggml/quantize.o \
$(CRT) \
$(APE_NO_MODIFY_SELF)
@$(APELINK)
o/$(MODE)/third_party/ggml/perplexity.com.dbg: \
$(THIRD_PARTY_GGML_LLAMA).pkg \
$(THIRD_PARTY_GGML_LLAMA_DEPS) \
o/$(MODE)/third_party/ggml/common.o \
o/$(MODE)/third_party/ggml/llama.o \
o/$(MODE)/third_party/ggml/perplexity.o \
$(CRT) \
$(APE_NO_MODIFY_SELF)
@$(APELINK)
Use Companion AI in llama.com by default
2023-04-29 07:48:14 +00:00
$(THIRD_PARTY_GGML_LLAMA).pkg: \
$(THIRD_PARTY_GGML_LLAMA_OBJS) \
Import llama.cpp https://github.com/ggerganov/llama.cpp 0b2da20538d01926b77ea237dd1c930c4d20b686 See third_party/ggml/README.cosmo for changes
2023-04-27 21:31:20 +00:00
$(foreach x,$(THIRD_PARTY_GGML_LLAMA_DIRECTDEPS),$($(x)_A).pkg)
Use Companion AI in llama.com by default
2023-04-29 07:48:14 +00:00
o/$(MODE)/third_party/ggml/companionai.txt.zip.o: private \
ZIPOBJ_FLAGS += \
-B
Import llama.cpp https://github.com/ggerganov/llama.cpp 0b2da20538d01926b77ea237dd1c930c4d20b686 See third_party/ggml/README.cosmo for changes
2023-04-27 21:31:20 +00:00
################################################################################
Perform some code cleanup
2023-05-15 23:32:10 +00:00
THIRD_PARTY_GGML_COMS = \
$(THIRD_PARTY_GGML_LLAMA) \
o/$(MODE)/third_party/ggml/perplexity.com
Import llama.cpp https://github.com/ggerganov/llama.cpp 0b2da20538d01926b77ea237dd1c930c4d20b686 See third_party/ggml/README.cosmo for changes
2023-04-27 21:31:20 +00:00
THIRD_PARTY_GGML_BINS = $(THIRD_PARTY_GGML_COMS) $(THIRD_PARTY_GGML_COMS:%=%.dbg)
THIRD_PARTY_GGML_LIBS = $(foreach x,$(THIRD_PARTY_GGML_ARTIFACTS),$($(x)))
THIRD_PARTY_GGML_SRCS = $(foreach x,$(THIRD_PARTY_GGML_ARTIFACTS),$($(x)_SRCS))
THIRD_PARTY_GGML_HDRS = $(foreach x,$(THIRD_PARTY_GGML_ARTIFACTS),$($(x)_HDRS))
THIRD_PARTY_GGML_OBJS = $(foreach x,$(THIRD_PARTY_GGML_ARTIFACTS),$($(x)_OBJS))
THIRD_PARTY_GGML_CHECKS = $(foreach x,$(THIRD_PARTY_GGML_ARTIFACTS),$($(x)_CHECKS))
$(THIRD_PARTY_GGML_OBJS): third_party/ggml/ggml.mk
.PHONY: o/$(MODE)/third_party/ggml
Use Companion AI in llama.com by default
2023-04-29 07:48:14 +00:00
o/$(MODE)/third_party/ggml: \
$(THIRD_PARTY_GGML_BINS) \
Import llama.cpp https://github.com/ggerganov/llama.cpp 0b2da20538d01926b77ea237dd1c930c4d20b686 See third_party/ggml/README.cosmo for changes
2023-04-27 21:31:20 +00:00
$(THIRD_PARTY_GGML_CHECKS)
Reference in a new issue Copy permalink