llama.cpp/Makefile

1055 lines
39 KiB
Makefile
Raw Normal View History

# Define the default target now so that it is always the first target
BUILD_TARGETS = \
main quantize quantize-stats perplexity imatrix embedding vdot q8dot train-text-from-scratch convert-llama2c-to-ggml \
2024-06-04 18:23:05 +00:00
simple batched batched-bench save-load-state server gguf gguf-split eval-callback llama-bench libllava.a llava-cli baby-llama \
retrieval speculative infill tokenize benchmark-matmult parallel finetune export-lora lookahead lookup passkey gritlm tests/test-c.o
# Binaries only useful for tests
TEST_TARGETS = \
llama : fix BPE pre-tokenization (#6920) * merged the changes from deepseeker models to main branch * Moved regex patterns to unicode.cpp and updated unicode.h * Moved header files * Resolved issues * added and refactored unicode_regex_split and related functions * Updated/merged the deepseek coder pr * Refactored code * Adding unicode regex mappings * Adding unicode regex function * Added needed functionality, testing remains * Fixed issues * Fixed issue with gpt2 regex custom preprocessor * unicode : fix? unicode_wstring_to_utf8 * lint : fix whitespaces * tests : add tokenizer tests for numbers * unicode : remove redundant headers * tests : remove and rename tokenizer test scripts * tests : add sample usage * gguf-py : reader prints warnings on duplicate keys * llama : towards llama3 tokenization support (wip) * unicode : shot in the dark to fix tests on Windows * unicode : first try custom implementations * convert : add "tokenizer.ggml.pre" GGUF KV (wip) * llama : use new pre-tokenizer type * convert : fix pre-tokenizer type writing * lint : fix * make : add test-tokenizer-0-llama-v3 * wip * models : add llama v3 vocab file * llama : adapt punctuation regex + add llama 3 regex * minor * unicode : set bomb * unicode : set bomb * unicode : always use std::wregex * unicode : support \p{N}, \p{L} and \p{P} natively * unicode : try fix windows * unicode : category support via std::regex * unicode : clean-up * unicode : simplify * convert : add convert-hf-to-gguf-update.py ggml-ci * lint : update * convert : add falcon ggml-ci * unicode : normalize signatures * lint : fix * lint : fix * convert : remove unused functions * convert : add comments * convert : exercise contractions ggml-ci * lint : fix * cmake : refactor test targets * tests : refactor vocab tests ggml-ci * tests : add more vocabs and tests ggml-ci * unicode : cleanup * scripts : ignore new update script in check-requirements.sh * models : add phi-3, mpt, gpt-2, starcoder * tests : disable obsolete ggml-ci * tests : use faster bpe test ggml-ci * llama : more prominent warning for old BPE models * tests : disable test-tokenizer-1-bpe due to slowness ggml-ci --------- Co-authored-by: Jaggzh <jaggz.h@gmail.com> Co-authored-by: Kazim Abrar Mahi <kazimabrarmahi135@gmail.com>
2024-04-29 13:58:41 +00:00
tests/test-autorelease \
tests/test-backend-ops \
tests/test-double-float \
tests/test-grad0 \
tests/test-grammar-integration \
tests/test-grammar-parser \
tests/test-json-schema-to-grammar \
tests/test-llama-grammar \
tests/test-model-load-cancel \
tests/test-opt \
tests/test-quantize-fns \
tests/test-quantize-perf \
tests/test-rope \
tests/test-sampling \
tests/test-tokenizer-0 \
tests/test-tokenizer-1-bpe \
tests/test-tokenizer-1-spm
# Code coverage output files
COV_TARGETS = *.gcno tests/*.gcno *.gcda tests/*.gcda *.gcov tests/*.gcov lcov-report gcovr-report
ifndef UNAME_S
UNAME_S := $(shell uname -s)
endif
ifndef UNAME_P
UNAME_P := $(shell uname -p)
endif
ifndef UNAME_M
UNAME_M := $(shell uname -m)
endif
# In GNU make default CXX is g++ instead of c++. Let's fix that so that users
# of non-gcc compilers don't have to provide g++ alias or wrapper.
DEFCC := cc
DEFCXX := c++
ifeq ($(origin CC),default)
CC := $(DEFCC)
endif
ifeq ($(origin CXX),default)
CXX := $(DEFCXX)
endif
# Mac OS + Arm can report x86_64
# ref: https://github.com/ggerganov/whisper.cpp/issues/66#issuecomment-1282546789
ifeq ($(UNAME_S),Darwin)
ifndef LLAMA_NO_METAL
LLAMA_METAL := 1
endif
LLAMA_NO_OPENMP := 1
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
ifdef LLAMA_RPC
BUILD_TARGETS += rpc-server
endif
default: $(BUILD_TARGETS)
2023-09-07 14:15:01 +00:00
test: $(TEST_TARGETS)
@failures=0; \
for test_target in $(TEST_TARGETS); do \
llama : fix BPE pre-tokenization (#6920) * merged the changes from deepseeker models to main branch * Moved regex patterns to unicode.cpp and updated unicode.h * Moved header files * Resolved issues * added and refactored unicode_regex_split and related functions * Updated/merged the deepseek coder pr * Refactored code * Adding unicode regex mappings * Adding unicode regex function * Added needed functionality, testing remains * Fixed issues * Fixed issue with gpt2 regex custom preprocessor * unicode : fix? unicode_wstring_to_utf8 * lint : fix whitespaces * tests : add tokenizer tests for numbers * unicode : remove redundant headers * tests : remove and rename tokenizer test scripts * tests : add sample usage * gguf-py : reader prints warnings on duplicate keys * llama : towards llama3 tokenization support (wip) * unicode : shot in the dark to fix tests on Windows * unicode : first try custom implementations * convert : add "tokenizer.ggml.pre" GGUF KV (wip) * llama : use new pre-tokenizer type * convert : fix pre-tokenizer type writing * lint : fix * make : add test-tokenizer-0-llama-v3 * wip * models : add llama v3 vocab file * llama : adapt punctuation regex + add llama 3 regex * minor * unicode : set bomb * unicode : set bomb * unicode : always use std::wregex * unicode : support \p{N}, \p{L} and \p{P} natively * unicode : try fix windows * unicode : category support via std::regex * unicode : clean-up * unicode : simplify * convert : add convert-hf-to-gguf-update.py ggml-ci * lint : update * convert : add falcon ggml-ci * unicode : normalize signatures * lint : fix * lint : fix * convert : remove unused functions * convert : add comments * convert : exercise contractions ggml-ci * lint : fix * cmake : refactor test targets * tests : refactor vocab tests ggml-ci * tests : add more vocabs and tests ggml-ci * unicode : cleanup * scripts : ignore new update script in check-requirements.sh * models : add phi-3, mpt, gpt-2, starcoder * tests : disable obsolete ggml-ci * tests : use faster bpe test ggml-ci * llama : more prominent warning for old BPE models * tests : disable test-tokenizer-1-bpe due to slowness ggml-ci --------- Co-authored-by: Jaggzh <jaggz.h@gmail.com> Co-authored-by: Kazim Abrar Mahi <kazimabrarmahi135@gmail.com>
2024-04-29 13:58:41 +00:00
if [ "$$test_target" = "tests/test-tokenizer-0" ]; then \
./$$test_target $(CURDIR)/models/ggml-vocab-llama-spm.gguf; \
./$$test_target $(CURDIR)/models/ggml-vocab-llama-bpe.gguf; \
./$$test_target $(CURDIR)/models/ggml-vocab-phi-3.gguf; \
./$$test_target $(CURDIR)/models/ggml-vocab-falcon.gguf; \
llama : fix BPE pre-tokenization (#6920) * merged the changes from deepseeker models to main branch * Moved regex patterns to unicode.cpp and updated unicode.h * Moved header files * Resolved issues * added and refactored unicode_regex_split and related functions * Updated/merged the deepseek coder pr * Refactored code * Adding unicode regex mappings * Adding unicode regex function * Added needed functionality, testing remains * Fixed issues * Fixed issue with gpt2 regex custom preprocessor * unicode : fix? unicode_wstring_to_utf8 * lint : fix whitespaces * tests : add tokenizer tests for numbers * unicode : remove redundant headers * tests : remove and rename tokenizer test scripts * tests : add sample usage * gguf-py : reader prints warnings on duplicate keys * llama : towards llama3 tokenization support (wip) * unicode : shot in the dark to fix tests on Windows * unicode : first try custom implementations * convert : add "tokenizer.ggml.pre" GGUF KV (wip) * llama : use new pre-tokenizer type * convert : fix pre-tokenizer type writing * lint : fix * make : add test-tokenizer-0-llama-v3 * wip * models : add llama v3 vocab file * llama : adapt punctuation regex + add llama 3 regex * minor * unicode : set bomb * unicode : set bomb * unicode : always use std::wregex * unicode : support \p{N}, \p{L} and \p{P} natively * unicode : try fix windows * unicode : category support via std::regex * unicode : clean-up * unicode : simplify * convert : add convert-hf-to-gguf-update.py ggml-ci * lint : update * convert : add falcon ggml-ci * unicode : normalize signatures * lint : fix * lint : fix * convert : remove unused functions * convert : add comments * convert : exercise contractions ggml-ci * lint : fix * cmake : refactor test targets * tests : refactor vocab tests ggml-ci * tests : add more vocabs and tests ggml-ci * unicode : cleanup * scripts : ignore new update script in check-requirements.sh * models : add phi-3, mpt, gpt-2, starcoder * tests : disable obsolete ggml-ci * tests : use faster bpe test ggml-ci * llama : more prominent warning for old BPE models * tests : disable test-tokenizer-1-bpe due to slowness ggml-ci --------- Co-authored-by: Jaggzh <jaggz.h@gmail.com> Co-authored-by: Kazim Abrar Mahi <kazimabrarmahi135@gmail.com>
2024-04-29 13:58:41 +00:00
./$$test_target $(CURDIR)/models/ggml-vocab-bert-bge.gguf; \
./$$test_target $(CURDIR)/models/ggml-vocab-starcoder.gguf; \
./$$test_target $(CURDIR)/models/ggml-vocab-gpt-2.gguf; \
./$$test_target $(CURDIR)/models/ggml-vocab-refact.gguf; \
llama : fix BPE pre-tokenization (#6920) * merged the changes from deepseeker models to main branch * Moved regex patterns to unicode.cpp and updated unicode.h * Moved header files * Resolved issues * added and refactored unicode_regex_split and related functions * Updated/merged the deepseek coder pr * Refactored code * Adding unicode regex mappings * Adding unicode regex function * Added needed functionality, testing remains * Fixed issues * Fixed issue with gpt2 regex custom preprocessor * unicode : fix? unicode_wstring_to_utf8 * lint : fix whitespaces * tests : add tokenizer tests for numbers * unicode : remove redundant headers * tests : remove and rename tokenizer test scripts * tests : add sample usage * gguf-py : reader prints warnings on duplicate keys * llama : towards llama3 tokenization support (wip) * unicode : shot in the dark to fix tests on Windows * unicode : first try custom implementations * convert : add "tokenizer.ggml.pre" GGUF KV (wip) * llama : use new pre-tokenizer type * convert : fix pre-tokenizer type writing * lint : fix * make : add test-tokenizer-0-llama-v3 * wip * models : add llama v3 vocab file * llama : adapt punctuation regex + add llama 3 regex * minor * unicode : set bomb * unicode : set bomb * unicode : always use std::wregex * unicode : support \p{N}, \p{L} and \p{P} natively * unicode : try fix windows * unicode : category support via std::regex * unicode : clean-up * unicode : simplify * convert : add convert-hf-to-gguf-update.py ggml-ci * lint : update * convert : add falcon ggml-ci * unicode : normalize signatures * lint : fix * lint : fix * convert : remove unused functions * convert : add comments * convert : exercise contractions ggml-ci * lint : fix * cmake : refactor test targets * tests : refactor vocab tests ggml-ci * tests : add more vocabs and tests ggml-ci * unicode : cleanup * scripts : ignore new update script in check-requirements.sh * models : add phi-3, mpt, gpt-2, starcoder * tests : disable obsolete ggml-ci * tests : use faster bpe test ggml-ci * llama : more prominent warning for old BPE models * tests : disable test-tokenizer-1-bpe due to slowness ggml-ci --------- Co-authored-by: Jaggzh <jaggz.h@gmail.com> Co-authored-by: Kazim Abrar Mahi <kazimabrarmahi135@gmail.com>
2024-04-29 13:58:41 +00:00
elif [ "$$test_target" = "tests/test-tokenizer-1-spm" ]; then \
continue; \
elif [ "$$test_target" = "tests/test-tokenizer-1-bpe" ]; then \
continue; \
else \
2023-09-07 14:15:01 +00:00
echo "Running test $$test_target..."; \
./$$test_target; \
fi; \
2023-09-07 14:15:01 +00:00
if [ $$? -ne 0 ]; then \
printf 'Test %s FAILED!\n\n' $$test_target; \
2023-09-07 14:15:01 +00:00
failures=$$(( failures + 1 )); \
else \
printf 'Test %s passed.\n\n' $$test_target; \
fi; \
done; \
if [ $$failures -gt 0 ]; then \
printf '\n%s tests failed.\n' $$failures; \
exit 1; \
fi
@echo 'All tests passed.'
all: $(BUILD_TARGETS) $(TEST_TARGETS)
coverage: ## Run code coverage
gcov -pb tests/*.cpp
lcov-report: coverage ## Generate lcov report
mkdir -p lcov-report
lcov --capture --directory . --output-file lcov-report/coverage.info
genhtml lcov-report/coverage.info --output-directory lcov-report
gcovr-report: coverage ## Generate gcovr report
mkdir -p gcovr-report
gcovr --root . --html --html-details --output gcovr-report/coverage.html
ifdef RISCV_CROSS_COMPILE
CC := riscv64-unknown-linux-gnu-gcc
CXX := riscv64-unknown-linux-gnu-g++
endif
2023-03-10 18:40:58 +00:00
#
# Compile flags
#
# keep standard at C11 and C++11
2024-02-19 11:41:51 +00:00
MK_CPPFLAGS = -I. -Icommon
MK_CFLAGS = -std=c11 -fPIC
MK_CXXFLAGS = -std=c++11 -fPIC
MK_NVCCFLAGS = -std=c++11
ggml : add SOTA 2,3,4,5,6 bit k-quantizations (#1684) * Starting to add k-quantization to ggml I think it is better to have quantization separate from ggml. For now just adding the k-quants there, but it would be better to also factor out the existing ggml quantizations. * Adding Q3_K and Q8_K (de)-quantization * Q3_K now working on CUDA and AVX2/scalar CUDA is not ideal - ~50% slower than Q4_0 for single token prediction, about the same in batch mode (perplexity). CPU single token is ~55 ms (on Ryzen 7950X). * Some improvement for Q3_K on CUDA It is now ~22.5 ms/token on my GPU, so ~30% slower than Q4_0. * Some more CUDA optimizations for Q3_K Single token is now 20.5 ms/token (~20% slower than Q4_0). Perplexity is on par with Q4_0. * Adding Q4_K - scalar, AVX2, CUDA Performance is the same or perhaps very slightly better than Q4_0 on the CPU. On the GPU, single token prediction is ~10% better than Q4_0, batch mode (perplexity is about the same). * Adding Q6_K - scalar, AVX2, CUDA Performance is ~40% lower compared to Q4_K on the CPU. This is to be expected, considering that we are memory bound on the CPU and the 6-bit model is ~44% larger than the 4-bit. On the GPU, single token prediction is ~6% lower than Q4_0, batch mode (perplexity) is even closer (but still slower). * Adding Q5_K - scalar, AVX2, CUDA Performance is ~20% lower compared to Q4_K on the CPU. This is to be expected, considering that we are memory bound on the CPU and the 5-bit model is ~22% larger than the 4-bit. On the GPU, single token prediction is about the same as Q4_0 for both, single token and batch prediction. * Per convention, all QX_K quantizations use Q5_K for output.weight * Adding quantization mixes * Quantization mixes: didn't quite get what I wanted in the last commit * Q4_K dot product for ARM_NEON * Q6_K dot product for ARM_NEON * Q5_K dot product for ARM_NEON * Adding Q3_K dot for ARM_NEON It is 22% slower than Q4_K, despite the smaller model size. On x86_64, where we are memory bound, the Q3_K model is quite a bit faster than Q4_K. * A very slightly faster ARM_NEON Q3_K dot * Adding Q2_K - just CUDA for now Token prediction is pretty good - about 15.5 ms on a RTX 4080. Perplexity is about the same as Q4_K. * Adding scalar and AVX2 Q2_K dot * Adding ARM_NEON Q2_K dot About the same performance as Q4_K. * A slightly faster ARM_NEON Q2_K dot Single token prediction is now ~36 ms on M2 Max. The code is much simpler too. * Fixed bug in Q2_K CUDA dot product kernel Stranegly enough, for the few prompts I tried with the 7B model the responses looked perfectly reasonable. Only realized something is not quite right when I tried the larger models and started getting nonse back. In any case, Q2_K single token evaluation time on an RTX 4080 in a Ryzen7950X box iusing CUDA and model fully loaded on the GPU are ~15.5 ms for 7B, ~25.4 ms for 13B, and ~55.8 ms for 30B. The max number of layers that fit in VRAM for The 65B is 32. With that, we get ~330 ms per token, which is not that much faster than just running on the CPU (~470 ms per token). * Don't print zeros/NaNs when no count histogram has been collected * A 10% faster CUDA vector dot kernel for Q3_K Q3_K is now running at ~18.5 ms / token on CUDA, so the gap to Q4_0 is only 10%. It seems memory acccess pattern is more important for performance than the amount of computation the kernel does. * A slightly daster Q4_K AVX2 dot product For perplexity, where we are less memory bound, time per pass drops by ~5%. Barely measurable difference for single token prediction. * A slightly faster ARM_NEON A4_K dot product * Minor * Fix quantization error test We cannot possibly be expecting rmse < 0.002 for 2- and 3-bit quantization variants. * Fix docker build I have been sloppy with vector reinterpret casts on ARM_NEON. It seems clang is very forgiving in that regard. * Added forgotten ggml.o dependence on k_quants.h to the Makefile * Had unintentionally committed the Makefile with -Ofast enabled * ggml : rename k_quants -> ggml-quants-k, use lowercase in code --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-06-05 19:56:18 +00:00
# -Ofast tends to produce faster code, but may not be available for some compilers.
k-quants : support for super-block size of 64 (#2001) * k_quants: WIP super-blocks with 64 weights * k_quants: WIP super-blocks with 64 weights Q6_K scalar and AVX2 works * k_quants: WIP super-blocks with 64 weights Q4_K scalar and AVX2 works * k_quants: WIP super-blocks with 64 weights Q2_K scalar and AVX2 works. Q2_K is way too slow (it is actually slower than the scalar implementation) * k_quants: WIP super-blocks with 64 weights Q3_K scalar and AVX2 works. * k_quants: WIP super-blocks with 64 weights Q5_K scalar and AVX2 works, and with that all k_quants are done on AVX2 and scalar * k_quants: WIP super-blocks with 64 weights Q6_K working on CUDA. Cannot make it run quite as gast as with super-blocks with 256 weigths: 8% slower on 4080, 20% slower on the 1660 (but there we fit 1 less layer on the GPU because pf the larger model size), so some fraction of these 20% is due to that, * k_quants: WIP super-blocks with 64 weights Q4_K working on CUDA. ~10% slower on GTX-1660, 16% slower on 4080. * k_quants: WIP super-blocks with 64 weights Q2_K working on CUDA. ~3% slower on GTX-1660, 10% slower on 4080. * k_quants: WIP super-blocks with 64 weights Q3_K working on CUDA. * k_quants: WIP super-blocks with 64 weights Q5_K working on CUDA, and with this CUDA is done. * k_quants: WIP super-blocks with 64 weights Q6_K working on ARM_NEON * k_quants: WIP super-blocks with 64 weights Q4_K working on ARM_NEON, but quite a bit slower than 256 weights * k_quants: WIP super-blocks with 64 weights Q2_K working on ARM_NEON, but quite a bit slower than 256 weights * k_quants: WIP super-blocks with 64 weights Q3_K working on ARM_NEON, but quite a bit slower than 256 weights. * k_quants: WIP super-blocks with 64 weights Q5_K working on ARM_NEON, but quite a bit slower than 256 weights. With that, we have full support for ARM_NEON, although performance is not quite there. * k_quants: WIP super-blocks with 64 weights Slightly more efficient Q3_K and Q5_K * k_quants: WIP super-blocks with 64 weights Another small improvement for Q3_K and Q5_K on ARM_NEON * k_quants: WIP super-blocks with 64 weights Yet another speedup for Q5_K on ARM_NEON. We are now within 10% of the QK_K = 256 version. * k_quants: WIP super-blocks with 64 weights * We are able to pass preprocessor macros to the Metal compiler * Q6_K works and is actually slightly more efficient than the QK_K = 256 version (25.2 ms vs 25.8 ms) * k_quants: WIP super-blocks with 64 weights Q4_K works on Metal and is actually slightly faster than QK_K = 256 (21.95 ms vs 24.0 ms). * k_quants: WIP super-blocks with 64 weights Q2_K works on Metal and is very slightly faster than QK_K = 256 (23.8 ms vs 24.2 ms). * k_quants: WIP super-blocks with 64 weights Q3_K works on Metal and is slightly faster than QK_K = 256 (26.6 ms vs 28.3 ms). * k_quants: WIP super-blocks with 64 weights Q5_K works on Metal and is slightly faster than QK_K = 256 (23.7 ms vs 26.3 ms). * k_quants: call them _K, not _k, also on Metal * k_quants: correctly define QK_K in llama.cpp * Fixed bug in q4_K quantization added with the 64-block addition * Simplify via lambda * k_quants: swicth Q3_K to 4-bit scales when QK_K = 64 Otherwise there isn't much benefit from this quantization type. There is some very slight loss in accuracy, but we reduce size by ~7%. E.g., for OpenLLaMA-3B, Q3_K_S perplexity is 8.6131 with 8-bit scales and 8.6352 with 4-bit, while file size decreases from 1.53G to 1.44G. * k_quants: switch Q4_K to 4-bit scales when QK_K = 64 Here the loss in accuracy is greater than for Q3_K, but the Q4_K points still move further to the left on the perplexity vs size curve. * k_quants: forgot to add the Metal changes in last commit * k_quants: change Q5_K to be type 0 when QK_K = 64 Still needs AVX2 implementation * k_quants: AVX2 implementation for new 64-weight Q5_K * k_quants: 10% faster ARM_NEON Q5_K dot product * k_quants: fixed issue caused by merging with master --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2023-06-26 16:43:07 +00:00
ifdef LLAMA_FAST
MK_CFLAGS += -Ofast
HOST_CXXFLAGS += -Ofast
ifndef LLAMA_DEBUG
MK_NVCCFLAGS += -O3
endif # LLAMA_DEBUG
k-quants : support for super-block size of 64 (#2001) * k_quants: WIP super-blocks with 64 weights * k_quants: WIP super-blocks with 64 weights Q6_K scalar and AVX2 works * k_quants: WIP super-blocks with 64 weights Q4_K scalar and AVX2 works * k_quants: WIP super-blocks with 64 weights Q2_K scalar and AVX2 works. Q2_K is way too slow (it is actually slower than the scalar implementation) * k_quants: WIP super-blocks with 64 weights Q3_K scalar and AVX2 works. * k_quants: WIP super-blocks with 64 weights Q5_K scalar and AVX2 works, and with that all k_quants are done on AVX2 and scalar * k_quants: WIP super-blocks with 64 weights Q6_K working on CUDA. Cannot make it run quite as gast as with super-blocks with 256 weigths: 8% slower on 4080, 20% slower on the 1660 (but there we fit 1 less layer on the GPU because pf the larger model size), so some fraction of these 20% is due to that, * k_quants: WIP super-blocks with 64 weights Q4_K working on CUDA. ~10% slower on GTX-1660, 16% slower on 4080. * k_quants: WIP super-blocks with 64 weights Q2_K working on CUDA. ~3% slower on GTX-1660, 10% slower on 4080. * k_quants: WIP super-blocks with 64 weights Q3_K working on CUDA. * k_quants: WIP super-blocks with 64 weights Q5_K working on CUDA, and with this CUDA is done. * k_quants: WIP super-blocks with 64 weights Q6_K working on ARM_NEON * k_quants: WIP super-blocks with 64 weights Q4_K working on ARM_NEON, but quite a bit slower than 256 weights * k_quants: WIP super-blocks with 64 weights Q2_K working on ARM_NEON, but quite a bit slower than 256 weights * k_quants: WIP super-blocks with 64 weights Q3_K working on ARM_NEON, but quite a bit slower than 256 weights. * k_quants: WIP super-blocks with 64 weights Q5_K working on ARM_NEON, but quite a bit slower than 256 weights. With that, we have full support for ARM_NEON, although performance is not quite there. * k_quants: WIP super-blocks with 64 weights Slightly more efficient Q3_K and Q5_K * k_quants: WIP super-blocks with 64 weights Another small improvement for Q3_K and Q5_K on ARM_NEON * k_quants: WIP super-blocks with 64 weights Yet another speedup for Q5_K on ARM_NEON. We are now within 10% of the QK_K = 256 version. * k_quants: WIP super-blocks with 64 weights * We are able to pass preprocessor macros to the Metal compiler * Q6_K works and is actually slightly more efficient than the QK_K = 256 version (25.2 ms vs 25.8 ms) * k_quants: WIP super-blocks with 64 weights Q4_K works on Metal and is actually slightly faster than QK_K = 256 (21.95 ms vs 24.0 ms). * k_quants: WIP super-blocks with 64 weights Q2_K works on Metal and is very slightly faster than QK_K = 256 (23.8 ms vs 24.2 ms). * k_quants: WIP super-blocks with 64 weights Q3_K works on Metal and is slightly faster than QK_K = 256 (26.6 ms vs 28.3 ms). * k_quants: WIP super-blocks with 64 weights Q5_K works on Metal and is slightly faster than QK_K = 256 (23.7 ms vs 26.3 ms). * k_quants: call them _K, not _k, also on Metal * k_quants: correctly define QK_K in llama.cpp * Fixed bug in q4_K quantization added with the 64-block addition * Simplify via lambda * k_quants: swicth Q3_K to 4-bit scales when QK_K = 64 Otherwise there isn't much benefit from this quantization type. There is some very slight loss in accuracy, but we reduce size by ~7%. E.g., for OpenLLaMA-3B, Q3_K_S perplexity is 8.6131 with 8-bit scales and 8.6352 with 4-bit, while file size decreases from 1.53G to 1.44G. * k_quants: switch Q4_K to 4-bit scales when QK_K = 64 Here the loss in accuracy is greater than for Q3_K, but the Q4_K points still move further to the left on the perplexity vs size curve. * k_quants: forgot to add the Metal changes in last commit * k_quants: change Q5_K to be type 0 when QK_K = 64 Still needs AVX2 implementation * k_quants: AVX2 implementation for new 64-weight Q5_K * k_quants: 10% faster ARM_NEON Q5_K dot product * k_quants: fixed issue caused by merging with master --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2023-06-26 16:43:07 +00:00
else
MK_CFLAGS += -O3
MK_CXXFLAGS += -O3
ifndef LLAMA_DEBUG
MK_NVCCFLAGS += -O3
endif # LLAMA_DEBUG
endif # LLAMA_FAST
2023-03-10 18:40:58 +00:00
ifndef LLAMA_NO_CCACHE
CCACHE := $(shell which ccache)
ifdef CCACHE
export CCACHE_SLOPPINESS = time_macros
$(info I ccache found, compilation results will be cached. Disable with LLAMA_NO_CCACHE.)
CC := $(CCACHE) $(CC)
CXX := $(CCACHE) $(CXX)
else
$(info I ccache not found. Consider installing it for faster compilation.)
endif # CCACHE
endif # LLAMA_NO_CCACHE
# clock_gettime came in POSIX.1b (1993)
# CLOCK_MONOTONIC came in POSIX.1-2001 / SUSv3 as optional
# posix_memalign came in POSIX.1-2001 / SUSv3
# M_PI is an XSI extension since POSIX.1-2001 / SUSv3, came in XPG1 (1985)
MK_CPPFLAGS += -D_XOPEN_SOURCE=600
# Somehow in OpenBSD whenever POSIX conformance is specified
# some string functions rely on locale_t availability,
# which was introduced in POSIX.1-2008, forcing us to go higher
ifeq ($(UNAME_S),OpenBSD)
MK_CPPFLAGS += -U_XOPEN_SOURCE -D_XOPEN_SOURCE=700
endif
# Data types, macros and functions related to controlling CPU affinity and
# some memory allocation are available on Linux through GNU extensions in libc
ifeq ($(UNAME_S),Linux)
MK_CPPFLAGS += -D_GNU_SOURCE
endif
# RLIMIT_MEMLOCK came in BSD, is not specified in POSIX.1,
# and on macOS its availability depends on enabling Darwin extensions
# similarly on DragonFly, enabling BSD extensions is necessary
ifeq ($(UNAME_S),Darwin)
MK_CPPFLAGS += -D_DARWIN_C_SOURCE
endif
ifeq ($(UNAME_S),DragonFly)
MK_CPPFLAGS += -D__BSD_VISIBLE
endif
# alloca is a non-standard interface that is not visible on BSDs when
# POSIX conformance is specified, but not all of them provide a clean way
# to enable it in such cases
ifeq ($(UNAME_S),FreeBSD)
MK_CPPFLAGS += -D__BSD_VISIBLE
endif
ifeq ($(UNAME_S),NetBSD)
MK_CPPFLAGS += -D_NETBSD_SOURCE
endif
ifeq ($(UNAME_S),OpenBSD)
MK_CPPFLAGS += -D_BSD_SOURCE
endif
ifdef LLAMA_SCHED_MAX_COPIES
MK_CPPFLAGS += -DGGML_SCHED_MAX_COPIES=$(LLAMA_SCHED_MAX_COPIES)
endif
2023-05-28 19:01:02 +00:00
ifdef LLAMA_DEBUG
MK_CFLAGS += -O0 -g
MK_CXXFLAGS += -O0 -g
MK_LDFLAGS += -g
MK_NVCCFLAGS += -O0 -g
ifeq ($(UNAME_S),Linux)
MK_CPPFLAGS += -D_GLIBCXX_ASSERTIONS
endif
2023-05-28 19:01:02 +00:00
else
MK_CPPFLAGS += -DNDEBUG
endif
ifdef LLAMA_SANITIZE_THREAD
MK_CFLAGS += -fsanitize=thread -g
MK_CXXFLAGS += -fsanitize=thread -g
MK_LDFLAGS += -fsanitize=thread -g
endif
ifdef LLAMA_SANITIZE_ADDRESS
MK_CFLAGS += -fsanitize=address -fno-omit-frame-pointer -g
MK_CXXFLAGS += -fsanitize=address -fno-omit-frame-pointer -g
MK_LDFLAGS += -fsanitize=address -fno-omit-frame-pointer -g
endif
ifdef LLAMA_SANITIZE_UNDEFINED
MK_CFLAGS += -fsanitize=undefined -g
MK_CXXFLAGS += -fsanitize=undefined -g
MK_LDFLAGS += -fsanitize=undefined -g
endif
ifdef LLAMA_SERVER_VERBOSE
MK_CPPFLAGS += -DSERVER_VERBOSE=$(LLAMA_SERVER_VERBOSE)
endif
ifdef LLAMA_SERVER_SSL
MK_CPPFLAGS += -DCPPHTTPLIB_OPENSSL_SUPPORT
MK_LDFLAGS += -lssl -lcrypto
endif
ifdef LLAMA_CODE_COVERAGE
MK_CXXFLAGS += -fprofile-arcs -ftest-coverage -dumpbase ''
endif
ifdef LLAMA_DISABLE_LOGS
MK_CPPFLAGS += -DLOG_DISABLE_LOGS
endif # LLAMA_DISABLE_LOGS
# warnings
WARN_FLAGS = -Wall -Wextra -Wpedantic -Wcast-qual -Wno-unused-function
MK_CFLAGS += $(WARN_FLAGS) -Wshadow -Wstrict-prototypes -Wpointer-arith -Wmissing-prototypes -Werror=implicit-int \
-Werror=implicit-function-declaration
MK_CXXFLAGS += $(WARN_FLAGS) -Wmissing-declarations -Wmissing-noreturn
ifeq ($(LLAMA_FATAL_WARNINGS),1)
MK_CFLAGS += -Werror
MK_CXXFLAGS += -Werror
endif
# this version of Apple ld64 is buggy
ifneq '' '$(findstring dyld-1015.7,$(shell $(CC) $(LDFLAGS) -Wl,-v 2>&1))'
MK_CPPFLAGS += -DHAVE_BUGGY_APPLE_LINKER
endif
2023-03-10 18:40:58 +00:00
# OS specific
# TODO: support Windows
ifneq '' '$(filter $(UNAME_S),Linux Darwin FreeBSD NetBSD OpenBSD Haiku)'
MK_CFLAGS += -pthread
MK_CXXFLAGS += -pthread
2023-03-10 18:40:58 +00:00
endif
# detect Windows
ifneq ($(findstring _NT,$(UNAME_S)),)
_WIN32 := 1
endif
# library name prefix
ifneq ($(_WIN32),1)
LIB_PRE := lib
endif
# Dynamic Shared Object extension
ifneq ($(_WIN32),1)
DSO_EXT := .so
else
DSO_EXT := .dll
endif
# Windows Sockets 2 (Winsock) for network-capable apps
ifeq ($(_WIN32),1)
LWINSOCK2 := -lws2_32
endif
2023-05-13 14:25:09 +00:00
ifdef LLAMA_GPROF
MK_CFLAGS += -pg
MK_CXXFLAGS += -pg
2023-05-13 14:25:09 +00:00
endif
ifdef LLAMA_PERF
MK_CPPFLAGS += -DGGML_PERF
2023-05-13 14:25:09 +00:00
endif
2023-03-10 18:40:58 +00:00
# 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
ifndef RISCV
ifeq ($(UNAME_M),$(filter $(UNAME_M),x86_64 i686 amd64))
# Use all CPU extensions that are available:
MK_CFLAGS += -march=native -mtune=native
HOST_CXXFLAGS += -march=native -mtune=native
# Usage AVX-only
#MK_CFLAGS += -mfma -mf16c -mavx
#MK_CXXFLAGS += -mfma -mf16c -mavx
# Usage SSSE3-only (Not is SSE3!)
#MK_CFLAGS += -mssse3
#MK_CXXFLAGS += -mssse3
2023-03-10 18:40:58 +00:00
endif
llama : Metal inference (#1642) * mtl : export the LLaMA computation graph * ci : disable temporary * mtl : adapt the MNIST example as starter * mtl : no need for mtl-export tool, add cli arg for main instead * mtl : export just a small part of the graph for now to make it easier * mtl : move MSL code into separate file for easy editing * mtl : initial get_rows_q4_0 kernel * mtl : confirmed get_rows_q4_0 is working correctly * mtl : add rms_norm kernel + confirm working * mtl : add mul kernel + confirm working * mtl : initial mul_mat Q4 kernel (wrong results) * mtl : mul_mat fixes (still wrong) * mtl : another mul_mat Q4 (still does not work) * mtl : working mul_mat q4 * ggml : fix handling of "view" ops in ggml_graph_import() * mtl : add rope kernel * mtl : add reshape and transpose handling * ggml : store offset as opt arg for ggml_view_xd() operators * mtl : add cpy kernel + handle view ops * mtl : confirm f16 x f32 attention mul mat * mtl : add scale kernel * mtl : add diag_mask_inf kernel * mtl : fix soft_max kernel * ggml : update ggml_nbytes() to handle non-contiguous tensors * mtl : verify V tensor contents * mtl : add f32 -> f32 cpy kernel * mtl : add silu kernel * mtl : add non-broadcast mul kernel * mtl : full GPU inference of the computation graph * mtl : optimize rms_norm and soft_max kernels * mtl : add f16 mat x f32 vec multiplication kernel * mtl : fix bug in f16 x f32 mul mat + speed-up computation * mtl : faster mul_mat_q4_0_f32 kernel * mtl : fix kernel signature + roll inner loop * mtl : more threads for rms_norm + better timing * mtl : remove printfs from inner loop * mtl : simplify implementation * mtl : add save/load vocab to ggml file * mtl : plug Metal inference into llama.cpp (very quick-n-dirty) * mtl : make it work with main example Lots of hacks but at least now it generates text * mtl : preparing for merge * mtl : clean-up ggml mtl interface + suport scratch / inplace * mtl : remove temp / debug code * metal : final refactoring and simplification * Revert "ci : disable temporary" This reverts commit 98c267fc77fe811082f672538fc91bcfc9072d63. * metal : add comments * metal : clean-up stuff, fix typos * readme : add Metal instructions * readme : add example for main
2023-06-04 20:34:30 +00:00
ifneq '' '$(findstring mingw,$(shell $(CC) -dumpmachine))'
# The stack is only 16-byte aligned on Windows, so don't let gcc emit aligned moves.
# https://gcc.gnu.org/bugzilla/show_bug.cgi?id=54412
# https://github.com/ggerganov/llama.cpp/issues/2922
MK_CFLAGS += -Xassembler -muse-unaligned-vector-move
MK_CXXFLAGS += -Xassembler -muse-unaligned-vector-move
# Target Windows 8 for PrefetchVirtualMemory
MK_CPPFLAGS += -D_WIN32_WINNT=0x602
endif
ifneq ($(filter aarch64%,$(UNAME_M)),)
# Apple M1, M2, etc.
# Raspberry Pi 3, 4, Zero 2 (64-bit)
# Nvidia Jetson
MK_CFLAGS += -mcpu=native
MK_CXXFLAGS += -mcpu=native
JETSON_RELEASE_INFO = $(shell jetson_release)
ifdef JETSON_RELEASE_INFO
ifneq ($(filter TX2%,$(JETSON_RELEASE_INFO)),)
JETSON_EOL_MODULE_DETECT = 1
CC = aarch64-unknown-linux-gnu-gcc
cxx = aarch64-unknown-linux-gnu-g++
endif
endif
endif
ifneq ($(filter armv6%,$(UNAME_M)),)
# Raspberry Pi 1, Zero
MK_CFLAGS += -mfpu=neon-fp-armv8 -mfp16-format=ieee -mno-unaligned-access
MK_CXXFLAGS += -mfpu=neon-fp-armv8 -mfp16-format=ieee -mno-unaligned-access
endif
ifneq ($(filter armv7%,$(UNAME_M)),)
# Raspberry Pi 2
MK_CFLAGS += -mfpu=neon-fp-armv8 -mfp16-format=ieee -mno-unaligned-access -funsafe-math-optimizations
MK_CXXFLAGS += -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)
MK_CFLAGS += -mfp16-format=ieee -mno-unaligned-access
MK_CXXFLAGS += -mfp16-format=ieee -mno-unaligned-access
endif
2023-03-10 18:40:58 +00:00
ifneq ($(filter ppc64%,$(UNAME_M)),)
POWER9_M := $(shell grep "POWER9" /proc/cpuinfo)
ifneq (,$(findstring POWER9,$(POWER9_M)))
MK_CFLAGS += -mcpu=power9
MK_CXXFLAGS += -mcpu=power9
2023-03-10 18:40:58 +00:00
endif
endif
llama : Metal inference (#1642) * mtl : export the LLaMA computation graph * ci : disable temporary * mtl : adapt the MNIST example as starter * mtl : no need for mtl-export tool, add cli arg for main instead * mtl : export just a small part of the graph for now to make it easier * mtl : move MSL code into separate file for easy editing * mtl : initial get_rows_q4_0 kernel * mtl : confirmed get_rows_q4_0 is working correctly * mtl : add rms_norm kernel + confirm working * mtl : add mul kernel + confirm working * mtl : initial mul_mat Q4 kernel (wrong results) * mtl : mul_mat fixes (still wrong) * mtl : another mul_mat Q4 (still does not work) * mtl : working mul_mat q4 * ggml : fix handling of "view" ops in ggml_graph_import() * mtl : add rope kernel * mtl : add reshape and transpose handling * ggml : store offset as opt arg for ggml_view_xd() operators * mtl : add cpy kernel + handle view ops * mtl : confirm f16 x f32 attention mul mat * mtl : add scale kernel * mtl : add diag_mask_inf kernel * mtl : fix soft_max kernel * ggml : update ggml_nbytes() to handle non-contiguous tensors * mtl : verify V tensor contents * mtl : add f32 -> f32 cpy kernel * mtl : add silu kernel * mtl : add non-broadcast mul kernel * mtl : full GPU inference of the computation graph * mtl : optimize rms_norm and soft_max kernels * mtl : add f16 mat x f32 vec multiplication kernel * mtl : fix bug in f16 x f32 mul mat + speed-up computation * mtl : faster mul_mat_q4_0_f32 kernel * mtl : fix kernel signature + roll inner loop * mtl : more threads for rms_norm + better timing * mtl : remove printfs from inner loop * mtl : simplify implementation * mtl : add save/load vocab to ggml file * mtl : plug Metal inference into llama.cpp (very quick-n-dirty) * mtl : make it work with main example Lots of hacks but at least now it generates text * mtl : preparing for merge * mtl : clean-up ggml mtl interface + suport scratch / inplace * mtl : remove temp / debug code * metal : final refactoring and simplification * Revert "ci : disable temporary" This reverts commit 98c267fc77fe811082f672538fc91bcfc9072d63. * metal : add comments * metal : clean-up stuff, fix typos * readme : add Metal instructions * readme : add example for main
2023-06-04 20:34:30 +00:00
ifneq ($(filter ppc64le%,$(UNAME_M)),)
MK_CFLAGS += -mcpu=powerpc64le
MK_CXXFLAGS += -mcpu=powerpc64le
CUDA_POWER_ARCH = 1
endif
ifneq ($(filter loongarch64%,$(UNAME_M)),)
MK_CFLAGS += -mlasx
MK_CXXFLAGS += -mlasx
endif
else
MK_CFLAGS += -march=rv64gcv -mabi=lp64d
MK_CXXFLAGS += -march=rv64gcv -mabi=lp64d
endif
2023-03-11 10:26:16 +00:00
ifndef LLAMA_NO_ACCELERATE
# Mac OS - include Accelerate framework.
# `-framework Accelerate` works both with Apple Silicon and Mac Intel
2023-03-10 18:40:58 +00:00
ifeq ($(UNAME_S),Darwin)
MK_CPPFLAGS += -DGGML_USE_ACCELERATE
MK_CPPFLAGS += -DACCELERATE_NEW_LAPACK
MK_CPPFLAGS += -DACCELERATE_LAPACK_ILP64
MK_LDFLAGS += -framework Accelerate
2023-03-10 18:40:58 +00:00
endif
llama : Metal inference (#1642) * mtl : export the LLaMA computation graph * ci : disable temporary * mtl : adapt the MNIST example as starter * mtl : no need for mtl-export tool, add cli arg for main instead * mtl : export just a small part of the graph for now to make it easier * mtl : move MSL code into separate file for easy editing * mtl : initial get_rows_q4_0 kernel * mtl : confirmed get_rows_q4_0 is working correctly * mtl : add rms_norm kernel + confirm working * mtl : add mul kernel + confirm working * mtl : initial mul_mat Q4 kernel (wrong results) * mtl : mul_mat fixes (still wrong) * mtl : another mul_mat Q4 (still does not work) * mtl : working mul_mat q4 * ggml : fix handling of "view" ops in ggml_graph_import() * mtl : add rope kernel * mtl : add reshape and transpose handling * ggml : store offset as opt arg for ggml_view_xd() operators * mtl : add cpy kernel + handle view ops * mtl : confirm f16 x f32 attention mul mat * mtl : add scale kernel * mtl : add diag_mask_inf kernel * mtl : fix soft_max kernel * ggml : update ggml_nbytes() to handle non-contiguous tensors * mtl : verify V tensor contents * mtl : add f32 -> f32 cpy kernel * mtl : add silu kernel * mtl : add non-broadcast mul kernel * mtl : full GPU inference of the computation graph * mtl : optimize rms_norm and soft_max kernels * mtl : add f16 mat x f32 vec multiplication kernel * mtl : fix bug in f16 x f32 mul mat + speed-up computation * mtl : faster mul_mat_q4_0_f32 kernel * mtl : fix kernel signature + roll inner loop * mtl : more threads for rms_norm + better timing * mtl : remove printfs from inner loop * mtl : simplify implementation * mtl : add save/load vocab to ggml file * mtl : plug Metal inference into llama.cpp (very quick-n-dirty) * mtl : make it work with main example Lots of hacks but at least now it generates text * mtl : preparing for merge * mtl : clean-up ggml mtl interface + suport scratch / inplace * mtl : remove temp / debug code * metal : final refactoring and simplification * Revert "ci : disable temporary" This reverts commit 98c267fc77fe811082f672538fc91bcfc9072d63. * metal : add comments * metal : clean-up stuff, fix typos * readme : add Metal instructions * readme : add example for main
2023-06-04 20:34:30 +00:00
endif # LLAMA_NO_ACCELERATE
ifndef LLAMA_NO_OPENMP
MK_CPPFLAGS += -DGGML_USE_OPENMP
MK_CFLAGS += -fopenmp
MK_CXXFLAGS += -fopenmp
endif # LLAMA_NO_OPENMP
2023-03-11 10:26:16 +00:00
ifdef LLAMA_OPENBLAS
MK_CPPFLAGS += -DGGML_USE_OPENBLAS $(shell pkg-config --cflags-only-I openblas)
MK_CFLAGS += $(shell pkg-config --cflags-only-other openblas)
MK_LDFLAGS += $(shell pkg-config --libs openblas)
llama : Metal inference (#1642) * mtl : export the LLaMA computation graph * ci : disable temporary * mtl : adapt the MNIST example as starter * mtl : no need for mtl-export tool, add cli arg for main instead * mtl : export just a small part of the graph for now to make it easier * mtl : move MSL code into separate file for easy editing * mtl : initial get_rows_q4_0 kernel * mtl : confirmed get_rows_q4_0 is working correctly * mtl : add rms_norm kernel + confirm working * mtl : add mul kernel + confirm working * mtl : initial mul_mat Q4 kernel (wrong results) * mtl : mul_mat fixes (still wrong) * mtl : another mul_mat Q4 (still does not work) * mtl : working mul_mat q4 * ggml : fix handling of "view" ops in ggml_graph_import() * mtl : add rope kernel * mtl : add reshape and transpose handling * ggml : store offset as opt arg for ggml_view_xd() operators * mtl : add cpy kernel + handle view ops * mtl : confirm f16 x f32 attention mul mat * mtl : add scale kernel * mtl : add diag_mask_inf kernel * mtl : fix soft_max kernel * ggml : update ggml_nbytes() to handle non-contiguous tensors * mtl : verify V tensor contents * mtl : add f32 -> f32 cpy kernel * mtl : add silu kernel * mtl : add non-broadcast mul kernel * mtl : full GPU inference of the computation graph * mtl : optimize rms_norm and soft_max kernels * mtl : add f16 mat x f32 vec multiplication kernel * mtl : fix bug in f16 x f32 mul mat + speed-up computation * mtl : faster mul_mat_q4_0_f32 kernel * mtl : fix kernel signature + roll inner loop * mtl : more threads for rms_norm + better timing * mtl : remove printfs from inner loop * mtl : simplify implementation * mtl : add save/load vocab to ggml file * mtl : plug Metal inference into llama.cpp (very quick-n-dirty) * mtl : make it work with main example Lots of hacks but at least now it generates text * mtl : preparing for merge * mtl : clean-up ggml mtl interface + suport scratch / inplace * mtl : remove temp / debug code * metal : final refactoring and simplification * Revert "ci : disable temporary" This reverts commit 98c267fc77fe811082f672538fc91bcfc9072d63. * metal : add comments * metal : clean-up stuff, fix typos * readme : add Metal instructions * readme : add example for main
2023-06-04 20:34:30 +00:00
endif # LLAMA_OPENBLAS
ifndef LLAMA_NO_LLAMAFILE
MK_CPPFLAGS += -DGGML_USE_LLAMAFILE
OBJS += sgemm.o
endif
ifdef LLAMA_BLIS
MK_CPPFLAGS += -DGGML_USE_OPENBLAS -I/usr/local/include/blis -I/usr/include/blis
MK_LDFLAGS += -lblis -L/usr/local/lib
llama : Metal inference (#1642) * mtl : export the LLaMA computation graph * ci : disable temporary * mtl : adapt the MNIST example as starter * mtl : no need for mtl-export tool, add cli arg for main instead * mtl : export just a small part of the graph for now to make it easier * mtl : move MSL code into separate file for easy editing * mtl : initial get_rows_q4_0 kernel * mtl : confirmed get_rows_q4_0 is working correctly * mtl : add rms_norm kernel + confirm working * mtl : add mul kernel + confirm working * mtl : initial mul_mat Q4 kernel (wrong results) * mtl : mul_mat fixes (still wrong) * mtl : another mul_mat Q4 (still does not work) * mtl : working mul_mat q4 * ggml : fix handling of "view" ops in ggml_graph_import() * mtl : add rope kernel * mtl : add reshape and transpose handling * ggml : store offset as opt arg for ggml_view_xd() operators * mtl : add cpy kernel + handle view ops * mtl : confirm f16 x f32 attention mul mat * mtl : add scale kernel * mtl : add diag_mask_inf kernel * mtl : fix soft_max kernel * ggml : update ggml_nbytes() to handle non-contiguous tensors * mtl : verify V tensor contents * mtl : add f32 -> f32 cpy kernel * mtl : add silu kernel * mtl : add non-broadcast mul kernel * mtl : full GPU inference of the computation graph * mtl : optimize rms_norm and soft_max kernels * mtl : add f16 mat x f32 vec multiplication kernel * mtl : fix bug in f16 x f32 mul mat + speed-up computation * mtl : faster mul_mat_q4_0_f32 kernel * mtl : fix kernel signature + roll inner loop * mtl : more threads for rms_norm + better timing * mtl : remove printfs from inner loop * mtl : simplify implementation * mtl : add save/load vocab to ggml file * mtl : plug Metal inference into llama.cpp (very quick-n-dirty) * mtl : make it work with main example Lots of hacks but at least now it generates text * mtl : preparing for merge * mtl : clean-up ggml mtl interface + suport scratch / inplace * mtl : remove temp / debug code * metal : final refactoring and simplification * Revert "ci : disable temporary" This reverts commit 98c267fc77fe811082f672538fc91bcfc9072d63. * metal : add comments * metal : clean-up stuff, fix typos * readme : add Metal instructions * readme : add example for main
2023-06-04 20:34:30 +00:00
endif # LLAMA_BLIS
ifdef LLAMA_RPC
MK_CPPFLAGS += -DGGML_USE_RPC
OBJS += ggml-rpc.o
endif # LLAMA_RPC
2023-04-19 09:22:45 +00:00
ifdef LLAMA_CUBLAS
# LLAMA_CUBLAS is deprecated and will be removed in the future
LLAMA_CUDA := 1
endif
OBJS_CUDA_TEMP_INST = $(patsubst %.cu,%.o,$(wildcard ggml-cuda/template-instances/fattn-wmma*.cu))
ifdef LLAMA_CUDA_FA_ALL_QUANTS
OBJS_CUDA_TEMP_INST += $(patsubst %.cu,%.o,$(wildcard ggml-cuda/template-instances/fattn-vec*.cu))
else
OBJS_CUDA_TEMP_INST += $(patsubst %.cu,%.o,$(wildcard ggml-cuda/template-instances/fattn-vec*q4_0-q4_0.cu))
OBJS_CUDA_TEMP_INST += $(patsubst %.cu,%.o,$(wildcard ggml-cuda/template-instances/fattn-vec*q8_0-q8_0.cu))
OBJS_CUDA_TEMP_INST += $(patsubst %.cu,%.o,$(wildcard ggml-cuda/template-instances/fattn-vec*f16-f16.cu))
endif # LLAMA_CUDA_FA_ALL_QUANTS
ifdef LLAMA_CUDA
ifneq ('', '$(wildcard /opt/cuda)')
CUDA_PATH ?= /opt/cuda
else
CUDA_PATH ?= /usr/local/cuda
endif
Introduction of CUDA Graphs to LLama.cpp (#6766) * DRAFT: Introduction of CUDA Graphs to LLama.cpp * FIx issues raised in comments * Tidied to now only use CUDA runtime (not mixed with driver calls) * disable for multi-gpu and batch size > 1 * Disable CUDA graphs for old GPU arch and with env var * added missing CUDA_CHECKs * Addressed comments * further addressed comments * limit to GGML_ALLOW_CUDA_GRAPHS defined in llama.cpp cmake * Added more comprehensive graph node checking * With mechanism to fall back if graph capture fails * Revert "With mechanism to fall back if graph capture fails" This reverts commit eb9f15fb6fcb81384f732c4601a5b25c016a5143. * Fall back if graph capture fails and address other comments * - renamed GGML_ALLOW_CUDA_GRAPHS to GGML_CUDA_USE_GRAPHS - rename env variable to disable CUDA graphs to GGML_CUDA_DISABLE_GRAPHS - updated Makefile build to enable CUDA graphs - removed graph capture failure checking in ggml_cuda_error using a global variable to track this is not thread safe, but I am also not safistied with checking an error by string if this is necessary to workaround some issues with graph capture with eg. cuBLAS, we can pass the ggml_backend_cuda_context to the error checking macro and store the result in the context - fixed several resource leaks - fixed issue with zero node graphs - changed fixed size arrays to vectors - removed the count of number of evaluations before start capturing, and instead changed the capture mode to relaxed - removed the check for multiple devices so that it is still possible to use a single device, instead checks for split buffers to disable cuda graphs with -sm row - changed the op for checking batch size to GGML_OP_ADD, should be more reliable than GGML_OP_SOFT_MAX - code style fixes - things to look into - VRAM usage of the cudaGraphExec_t, if it is significant we may need to make it optional - possibility of using cudaStreamBeginCaptureToGraph to keep track of which ggml graph nodes correspond to which cuda graph nodes * fix build without cuda graphs * remove outdated comment * replace minimum cc value with a constant --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-05-08 20:55:49 +00:00
MK_CPPFLAGS += -DGGML_USE_CUDA -I$(CUDA_PATH)/include -I$(CUDA_PATH)/targets/$(UNAME_M)-linux/include -DGGML_CUDA_USE_GRAPHS
MK_LDFLAGS += -lcuda -lcublas -lculibos -lcudart -lcublasLt -lpthread -ldl -lrt -L$(CUDA_PATH)/lib64 -L/usr/lib64 -L$(CUDA_PATH)/targets/$(UNAME_M)-linux/lib -L/usr/lib/wsl/lib
OBJS += ggml-cuda.o
OBJS += $(patsubst %.cu,%.o,$(wildcard ggml-cuda/*.cu))
OBJS += $(OBJS_CUDA_TEMP_INST)
MK_NVCCFLAGS += -use_fast_math
ifdef LLAMA_FATAL_WARNINGS
MK_NVCCFLAGS += -Werror all-warnings
endif # LLAMA_FATAL_WARNINGS
ifndef JETSON_EOL_MODULE_DETECT
MK_NVCCFLAGS += --forward-unknown-to-host-compiler
endif # JETSON_EOL_MODULE_DETECT
llama : add Mixtral support (#4406) * convert : support Mixtral as LLAMA arch * convert : fix n_ff typo * llama : model loading * ggml : sync latest ggml_mul_mat_id * llama : update graph to support MoE * llama : fix cur -> cur_expert * llama : first working version * llama : fix expert weighting in the FFN * ggml : ggml_get_rows support 2D indexing [n_tokens, n_experts] (cpu only) * ggml : add n_as argument to ggml_mul_mat_id * ggml : fix ggml_get_rows to take into account ne02 / ne11 * metal : add more general support for ggml_get_rows + tests * llama : add basic support for offloading moe with CUDA * metal : add/mul/div use general kernel when src1 not cont * metal : reduce the kernel launches for ggml_mul_mat_id * ggml : get_rows : support non-contiguos tensors with gaps, generalize up to 3D * ggml : update get_rows f16 and q * cuda : support non-contiguous src1 in get_rows * llama : offload missing ffn_moe_silu * metal : fix ggml_get_rows to work with non-cont src1 * metal : add indirect mat-vec kernels for all quantization types * llama : do not quantize expert gating tensors * llama : add n_expert and n_expert_used to hparams + change quants * test-backend-ops : add moe test * cuda : fix get_rows when ncols is odd * convert : determine n_ctx correctly * metal : fix ggml_mul_mat_id for F32 * test-backend-ops : make experts more evenly probable (test_moe) * test-backend-ops : cleanup, add moe test for batches * test-backend-ops : add cpy from f32 -> all types test * test-backend-ops : fix dequantize block offset * llama : fix hard-coded number of experts * test-backend-ops : simplify and disable slow tests to avoid CI timeout * test-backend-ops : disable MOE test with thread sanitizer * cuda : fix mul_mat_id with multi gpu * convert : use 1e6 rope_freq_base for mixtral * convert : fix style * convert : support safetensors format * gguf-py : bump version * metal : add cpy f16 -> f32 kernel * metal : fix binary ops for ne10 % 4 != 0 * test-backend-ops : add one more sum_rows test * ggml : do not use BLAS with ggml_mul_mat_id * convert-hf : support for mixtral-instruct (#4428) * convert : typo fix, add additional hyperparameters, use LLaMA arch for Mixtral-instruct * convert : use sentencepiece tokenizer for Mixtral-instruct * convert : make flake8 happy * metal : fix soft_max kernels ref: https://github.com/ggerganov/ggml/pull/621/commits/1914017863d2f9ab8ecc0281cc2a56d683668b92 * metal : limit kernels to not use more than the allowed threads --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Radek Pilar <github@mrkva.eu>
2023-12-13 12:04:25 +00:00
ifdef LLAMA_DEBUG
MK_NVCCFLAGS += -lineinfo
endif # LLAMA_DEBUG
ifdef LLAMA_CUDA_DEBUG
MK_NVCCFLAGS += --device-debug
endif # LLAMA_CUDA_DEBUG
ifdef LLAMA_CUDA_NVCC
NVCC = $(CCACHE) $(LLAMA_CUDA_NVCC)
else
NVCC = $(CCACHE) nvcc
endif #LLAMA_CUDA_NVCC
ifdef CUDA_DOCKER_ARCH
MK_NVCCFLAGS += -Wno-deprecated-gpu-targets -arch=$(CUDA_DOCKER_ARCH)
else ifndef CUDA_POWER_ARCH
MK_NVCCFLAGS += -arch=native
endif # CUDA_DOCKER_ARCH
ifdef LLAMA_CUDA_FORCE_DMMV
MK_NVCCFLAGS += -DGGML_CUDA_FORCE_DMMV
endif # LLAMA_CUDA_FORCE_DMMV
ifdef LLAMA_CUDA_FORCE_MMQ
MK_NVCCFLAGS += -DGGML_CUDA_FORCE_MMQ
endif # LLAMA_CUDA_FORCE_MMQ
ifdef LLAMA_CUDA_DMMV_X
MK_NVCCFLAGS += -DGGML_CUDA_DMMV_X=$(LLAMA_CUDA_DMMV_X)
else
MK_NVCCFLAGS += -DGGML_CUDA_DMMV_X=32
endif # LLAMA_CUDA_DMMV_X
ifdef LLAMA_CUDA_MMV_Y
MK_NVCCFLAGS += -DGGML_CUDA_MMV_Y=$(LLAMA_CUDA_MMV_Y)
else ifdef LLAMA_CUDA_DMMV_Y
MK_NVCCFLAGS += -DGGML_CUDA_MMV_Y=$(LLAMA_CUDA_DMMV_Y) # for backwards compatibility
else
MK_NVCCFLAGS += -DGGML_CUDA_MMV_Y=1
endif # LLAMA_CUDA_MMV_Y
ifdef LLAMA_CUDA_F16
MK_NVCCFLAGS += -DGGML_CUDA_F16
endif # LLAMA_CUDA_F16
ifdef LLAMA_CUDA_DMMV_F16
MK_NVCCFLAGS += -DGGML_CUDA_F16
endif # LLAMA_CUDA_DMMV_F16
ifdef LLAMA_CUDA_KQUANTS_ITER
MK_NVCCFLAGS += -DK_QUANTS_PER_ITERATION=$(LLAMA_CUDA_KQUANTS_ITER)
else
MK_NVCCFLAGS += -DK_QUANTS_PER_ITERATION=2
endif
ifdef LLAMA_CUDA_PEER_MAX_BATCH_SIZE
MK_NVCCFLAGS += -DGGML_CUDA_PEER_MAX_BATCH_SIZE=$(LLAMA_CUDA_PEER_MAX_BATCH_SIZE)
else
MK_NVCCFLAGS += -DGGML_CUDA_PEER_MAX_BATCH_SIZE=128
endif # LLAMA_CUDA_PEER_MAX_BATCH_SIZE
ifdef LLAMA_CUDA_NO_PEER_COPY
MK_NVCCFLAGS += -DGGML_CUDA_NO_PEER_COPY
endif # LLAMA_CUDA_NO_PEER_COPY
ifdef LLAMA_CUDA_CCBIN
MK_NVCCFLAGS += -ccbin $(LLAMA_CUDA_CCBIN)
endif # LLAMA_CUDA_CCBIN
ifdef LLAMA_CUDA_FA_ALL_QUANTS
MK_NVCCFLAGS += -DGGML_CUDA_FA_ALL_QUANTS
endif # LLAMA_CUDA_FA_ALL_QUANTS
ifdef JETSON_EOL_MODULE_DETECT
define NVCC_COMPILE
$(NVCC) -I. -Icommon -D_XOPEN_SOURCE=600 -D_GNU_SOURCE -DNDEBUG -DGGML_USE_CUDA -I/usr/local/cuda/include -I/opt/cuda/include -I/usr/local/cuda/targets/aarch64-linux/include -std=c++11 -O3 $(NVCCFLAGS) $(CPPFLAGS) -Xcompiler "$(CUDA_CXXFLAGS)" -c $< -o $@
endef # NVCC_COMPILE
else
define NVCC_COMPILE
$(NVCC) $(NVCCFLAGS) $(CPPFLAGS) -Xcompiler "$(CUDA_CXXFLAGS)" -c $< -o $@
endef # NVCC_COMPILE
endif # JETSON_EOL_MODULE_DETECT
ggml-cuda/%.o: ggml-cuda/%.cu ggml.h ggml-common.h ggml-cuda/common.cuh
$(NVCC_COMPILE)
ggml-cuda.o: ggml-cuda.cu ggml-cuda.h ggml.h ggml-backend.h ggml-backend-impl.h ggml-common.h $(wildcard ggml-cuda/*.cuh)
$(NVCC_COMPILE)
endif # LLAMA_CUDA
llama : Metal inference (#1642) * mtl : export the LLaMA computation graph * ci : disable temporary * mtl : adapt the MNIST example as starter * mtl : no need for mtl-export tool, add cli arg for main instead * mtl : export just a small part of the graph for now to make it easier * mtl : move MSL code into separate file for easy editing * mtl : initial get_rows_q4_0 kernel * mtl : confirmed get_rows_q4_0 is working correctly * mtl : add rms_norm kernel + confirm working * mtl : add mul kernel + confirm working * mtl : initial mul_mat Q4 kernel (wrong results) * mtl : mul_mat fixes (still wrong) * mtl : another mul_mat Q4 (still does not work) * mtl : working mul_mat q4 * ggml : fix handling of "view" ops in ggml_graph_import() * mtl : add rope kernel * mtl : add reshape and transpose handling * ggml : store offset as opt arg for ggml_view_xd() operators * mtl : add cpy kernel + handle view ops * mtl : confirm f16 x f32 attention mul mat * mtl : add scale kernel * mtl : add diag_mask_inf kernel * mtl : fix soft_max kernel * ggml : update ggml_nbytes() to handle non-contiguous tensors * mtl : verify V tensor contents * mtl : add f32 -> f32 cpy kernel * mtl : add silu kernel * mtl : add non-broadcast mul kernel * mtl : full GPU inference of the computation graph * mtl : optimize rms_norm and soft_max kernels * mtl : add f16 mat x f32 vec multiplication kernel * mtl : fix bug in f16 x f32 mul mat + speed-up computation * mtl : faster mul_mat_q4_0_f32 kernel * mtl : fix kernel signature + roll inner loop * mtl : more threads for rms_norm + better timing * mtl : remove printfs from inner loop * mtl : simplify implementation * mtl : add save/load vocab to ggml file * mtl : plug Metal inference into llama.cpp (very quick-n-dirty) * mtl : make it work with main example Lots of hacks but at least now it generates text * mtl : preparing for merge * mtl : clean-up ggml mtl interface + suport scratch / inplace * mtl : remove temp / debug code * metal : final refactoring and simplification * Revert "ci : disable temporary" This reverts commit 98c267fc77fe811082f672538fc91bcfc9072d63. * metal : add comments * metal : clean-up stuff, fix typos * readme : add Metal instructions * readme : add example for main
2023-06-04 20:34:30 +00:00
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 17:03:59 +00:00
ifdef LLAMA_VULKAN
MK_CPPFLAGS += -DGGML_USE_VULKAN
MK_LDFLAGS += -lvulkan
OBJS += ggml-vulkan.o
ifdef LLAMA_VULKAN_CHECK_RESULTS
MK_CPPFLAGS += -DGGML_VULKAN_CHECK_RESULTS
endif
ifdef LLAMA_VULKAN_DEBUG
MK_CPPFLAGS += -DGGML_VULKAN_DEBUG
endif
ifdef LLAMA_VULKAN_VALIDATE
MK_CPPFLAGS += -DGGML_VULKAN_VALIDATE
endif
ifdef LLAMA_VULKAN_RUN_TESTS
MK_CPPFLAGS += -DGGML_VULKAN_RUN_TESTS
endif
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 17:03:59 +00:00
ggml-vulkan.o: ggml-vulkan.cpp ggml-vulkan.h
$(CXX) $(CXXFLAGS) -c $< -o $@
endif # LLAMA_VULKAN
ifdef LLAMA_HIPBLAS
ifeq ($(wildcard /opt/rocm),)
ROCM_PATH ?= /usr
ROCm: use native CMake HIP support (#5966) Supercedes #4024 and #4813. CMake's native HIP support has become the recommended way to add HIP code into a project (see [here](https://rocm.docs.amd.com/en/docs-6.0.0/conceptual/cmake-packages.html#using-hip-in-cmake)). This PR makes the following changes: 1. The environment variable `HIPCXX` or CMake option `CMAKE_HIP_COMPILER` should be used to specify the HIP compiler. Notably this shouldn't be `hipcc`, but ROCm's clang, which usually resides in `$ROCM_PATH/llvm/bin/clang`. Previously this was control by `CMAKE_C_COMPILER` and `CMAKE_CXX_COMPILER`. Note that since native CMake HIP support is not yet available on Windows, on Windows we fall back to the old behavior. 2. CMake option `CMAKE_HIP_ARCHITECTURES` is used to control the GPU architectures to build for. Previously this was controled by `GPU_TARGETS`. 3. Updated the Nix recipe to account for these new changes. 4. The GPU targets to build against in the Nix recipe is now consistent with the supported GPU targets in nixpkgs. 5. Added CI checks for HIP on both Linux and Windows. On Linux, we test both the new and old behavior. The most important part about this PR is the separation of the HIP compiler and the C/C++ compiler. This allows users to choose a different C/C++ compiler if desired, compared to the current situation where when building for ROCm support, everything must be compiled with ROCm's clang. ~~Makefile is unchanged. Please let me know if we want to be consistent on variables' naming because Makefile still uses `GPU_TARGETS` to control architectures to build for, but I feel like setting `CMAKE_HIP_ARCHITECTURES` is a bit awkward when you're calling `make`.~~ Makefile used `GPU_TARGETS` but the README says to use `AMDGPU_TARGETS`. For consistency with CMake, all usage of `GPU_TARGETS` in Makefile has been updated to `AMDGPU_TARGETS`. Thanks to the suggestion of @jin-eld, to maintain backwards compatibility (and not break too many downstream users' builds), if `CMAKE_CXX_COMPILER` ends with `hipcc`, then we still compile using the original behavior and emit a warning that recommends switching to the new HIP support. Similarly, if `AMDGPU_TARGETS` is set but `CMAKE_HIP_ARCHITECTURES` is not, then we forward `AMDGPU_TARGETS` to `CMAKE_HIP_ARCHITECTURES` to ease the transition to the new HIP support. Signed-off-by: Gavin Zhao <git@gzgz.dev>
2024-05-17 15:03:03 +00:00
AMDGPU_TARGETS ?= $(shell $(shell which amdgpu-arch))
else
ROCM_PATH ?= /opt/rocm
ROCm: use native CMake HIP support (#5966) Supercedes #4024 and #4813. CMake's native HIP support has become the recommended way to add HIP code into a project (see [here](https://rocm.docs.amd.com/en/docs-6.0.0/conceptual/cmake-packages.html#using-hip-in-cmake)). This PR makes the following changes: 1. The environment variable `HIPCXX` or CMake option `CMAKE_HIP_COMPILER` should be used to specify the HIP compiler. Notably this shouldn't be `hipcc`, but ROCm's clang, which usually resides in `$ROCM_PATH/llvm/bin/clang`. Previously this was control by `CMAKE_C_COMPILER` and `CMAKE_CXX_COMPILER`. Note that since native CMake HIP support is not yet available on Windows, on Windows we fall back to the old behavior. 2. CMake option `CMAKE_HIP_ARCHITECTURES` is used to control the GPU architectures to build for. Previously this was controled by `GPU_TARGETS`. 3. Updated the Nix recipe to account for these new changes. 4. The GPU targets to build against in the Nix recipe is now consistent with the supported GPU targets in nixpkgs. 5. Added CI checks for HIP on both Linux and Windows. On Linux, we test both the new and old behavior. The most important part about this PR is the separation of the HIP compiler and the C/C++ compiler. This allows users to choose a different C/C++ compiler if desired, compared to the current situation where when building for ROCm support, everything must be compiled with ROCm's clang. ~~Makefile is unchanged. Please let me know if we want to be consistent on variables' naming because Makefile still uses `GPU_TARGETS` to control architectures to build for, but I feel like setting `CMAKE_HIP_ARCHITECTURES` is a bit awkward when you're calling `make`.~~ Makefile used `GPU_TARGETS` but the README says to use `AMDGPU_TARGETS`. For consistency with CMake, all usage of `GPU_TARGETS` in Makefile has been updated to `AMDGPU_TARGETS`. Thanks to the suggestion of @jin-eld, to maintain backwards compatibility (and not break too many downstream users' builds), if `CMAKE_CXX_COMPILER` ends with `hipcc`, then we still compile using the original behavior and emit a warning that recommends switching to the new HIP support. Similarly, if `AMDGPU_TARGETS` is set but `CMAKE_HIP_ARCHITECTURES` is not, then we forward `AMDGPU_TARGETS` to `CMAKE_HIP_ARCHITECTURES` to ease the transition to the new HIP support. Signed-off-by: Gavin Zhao <git@gzgz.dev>
2024-05-17 15:03:03 +00:00
AMDGPU_TARGETS ?= $(shell $(ROCM_PATH)/llvm/bin/amdgpu-arch)
endif
HIPCC ?= $(CCACHE) $(ROCM_PATH)/bin/hipcc
LLAMA_CUDA_DMMV_X ?= 32
LLAMA_CUDA_MMV_Y ?= 1
LLAMA_CUDA_KQUANTS_ITER ?= 2
MK_CPPFLAGS += -DGGML_USE_HIPBLAS -DGGML_USE_CUDA
ifdef LLAMA_HIP_UMA
MK_CPPFLAGS += -DGGML_HIP_UMA
endif # LLAMA_HIP_UMA
MK_LDFLAGS += -L$(ROCM_PATH)/lib -Wl,-rpath=$(ROCM_PATH)/lib
2024-05-31 14:00:29 +00:00
MK_LDFLAGS += -L$(ROCM_PATH)/lib64 -Wl,-rpath=$(ROCM_PATH)/lib64
MK_LDFLAGS += -lhipblas -lamdhip64 -lrocblas
ROCm: use native CMake HIP support (#5966) Supercedes #4024 and #4813. CMake's native HIP support has become the recommended way to add HIP code into a project (see [here](https://rocm.docs.amd.com/en/docs-6.0.0/conceptual/cmake-packages.html#using-hip-in-cmake)). This PR makes the following changes: 1. The environment variable `HIPCXX` or CMake option `CMAKE_HIP_COMPILER` should be used to specify the HIP compiler. Notably this shouldn't be `hipcc`, but ROCm's clang, which usually resides in `$ROCM_PATH/llvm/bin/clang`. Previously this was control by `CMAKE_C_COMPILER` and `CMAKE_CXX_COMPILER`. Note that since native CMake HIP support is not yet available on Windows, on Windows we fall back to the old behavior. 2. CMake option `CMAKE_HIP_ARCHITECTURES` is used to control the GPU architectures to build for. Previously this was controled by `GPU_TARGETS`. 3. Updated the Nix recipe to account for these new changes. 4. The GPU targets to build against in the Nix recipe is now consistent with the supported GPU targets in nixpkgs. 5. Added CI checks for HIP on both Linux and Windows. On Linux, we test both the new and old behavior. The most important part about this PR is the separation of the HIP compiler and the C/C++ compiler. This allows users to choose a different C/C++ compiler if desired, compared to the current situation where when building for ROCm support, everything must be compiled with ROCm's clang. ~~Makefile is unchanged. Please let me know if we want to be consistent on variables' naming because Makefile still uses `GPU_TARGETS` to control architectures to build for, but I feel like setting `CMAKE_HIP_ARCHITECTURES` is a bit awkward when you're calling `make`.~~ Makefile used `GPU_TARGETS` but the README says to use `AMDGPU_TARGETS`. For consistency with CMake, all usage of `GPU_TARGETS` in Makefile has been updated to `AMDGPU_TARGETS`. Thanks to the suggestion of @jin-eld, to maintain backwards compatibility (and not break too many downstream users' builds), if `CMAKE_CXX_COMPILER` ends with `hipcc`, then we still compile using the original behavior and emit a warning that recommends switching to the new HIP support. Similarly, if `AMDGPU_TARGETS` is set but `CMAKE_HIP_ARCHITECTURES` is not, then we forward `AMDGPU_TARGETS` to `CMAKE_HIP_ARCHITECTURES` to ease the transition to the new HIP support. Signed-off-by: Gavin Zhao <git@gzgz.dev>
2024-05-17 15:03:03 +00:00
HIPFLAGS += $(addprefix --offload-arch=,$(AMDGPU_TARGETS))
HIPFLAGS += -DGGML_CUDA_DMMV_X=$(LLAMA_CUDA_DMMV_X)
HIPFLAGS += -DGGML_CUDA_MMV_Y=$(LLAMA_CUDA_MMV_Y)
HIPFLAGS += -DK_QUANTS_PER_ITERATION=$(LLAMA_CUDA_KQUANTS_ITER)
ifdef LLAMA_CUDA_FORCE_DMMV
HIPFLAGS += -DGGML_CUDA_FORCE_DMMV
endif # LLAMA_CUDA_FORCE_DMMV
ifdef LLAMA_CUDA_NO_PEER_COPY
HIPFLAGS += -DGGML_CUDA_NO_PEER_COPY
endif # LLAMA_CUDA_NO_PEER_COPY
OBJS += ggml-cuda.o
OBJS += $(patsubst %.cu,%.o,$(wildcard ggml-cuda/*.cu))
OBJS += $(OBJS_CUDA_TEMP_INST)
2024-03-27 13:02:49 +00:00
ggml-cuda.o: ggml-cuda.cu ggml-cuda.h ggml.h ggml-backend.h ggml-backend-impl.h ggml-common.h $(wildcard ggml-cuda/*.cuh)
$(HIPCC) $(CXXFLAGS) $(HIPFLAGS) -x hip -c -o $@ $<
ggml-cuda/%.o: ggml-cuda/%.cu ggml.h ggml-common.h ggml-cuda/common.cuh
$(HIPCC) $(CXXFLAGS) $(HIPFLAGS) -x hip -c -o $@ $<
endif # LLAMA_HIPBLAS
llama : Metal inference (#1642) * mtl : export the LLaMA computation graph * ci : disable temporary * mtl : adapt the MNIST example as starter * mtl : no need for mtl-export tool, add cli arg for main instead * mtl : export just a small part of the graph for now to make it easier * mtl : move MSL code into separate file for easy editing * mtl : initial get_rows_q4_0 kernel * mtl : confirmed get_rows_q4_0 is working correctly * mtl : add rms_norm kernel + confirm working * mtl : add mul kernel + confirm working * mtl : initial mul_mat Q4 kernel (wrong results) * mtl : mul_mat fixes (still wrong) * mtl : another mul_mat Q4 (still does not work) * mtl : working mul_mat q4 * ggml : fix handling of "view" ops in ggml_graph_import() * mtl : add rope kernel * mtl : add reshape and transpose handling * ggml : store offset as opt arg for ggml_view_xd() operators * mtl : add cpy kernel + handle view ops * mtl : confirm f16 x f32 attention mul mat * mtl : add scale kernel * mtl : add diag_mask_inf kernel * mtl : fix soft_max kernel * ggml : update ggml_nbytes() to handle non-contiguous tensors * mtl : verify V tensor contents * mtl : add f32 -> f32 cpy kernel * mtl : add silu kernel * mtl : add non-broadcast mul kernel * mtl : full GPU inference of the computation graph * mtl : optimize rms_norm and soft_max kernels * mtl : add f16 mat x f32 vec multiplication kernel * mtl : fix bug in f16 x f32 mul mat + speed-up computation * mtl : faster mul_mat_q4_0_f32 kernel * mtl : fix kernel signature + roll inner loop * mtl : more threads for rms_norm + better timing * mtl : remove printfs from inner loop * mtl : simplify implementation * mtl : add save/load vocab to ggml file * mtl : plug Metal inference into llama.cpp (very quick-n-dirty) * mtl : make it work with main example Lots of hacks but at least now it generates text * mtl : preparing for merge * mtl : clean-up ggml mtl interface + suport scratch / inplace * mtl : remove temp / debug code * metal : final refactoring and simplification * Revert "ci : disable temporary" This reverts commit 98c267fc77fe811082f672538fc91bcfc9072d63. * metal : add comments * metal : clean-up stuff, fix typos * readme : add Metal instructions * readme : add example for main
2023-06-04 20:34:30 +00:00
ifdef LLAMA_METAL
MK_CPPFLAGS += -DGGML_USE_METAL
MK_LDFLAGS += -framework Foundation -framework Metal -framework MetalKit
OBJS += ggml-metal.o
ifdef LLAMA_METAL_NDEBUG
MK_CPPFLAGS += -DGGML_METAL_NDEBUG
endif
ifdef LLAMA_METAL_EMBED_LIBRARY
MK_CPPFLAGS += -DGGML_METAL_EMBED_LIBRARY
OBJS += ggml-metal-embed.o
endif
llama : Metal inference (#1642) * mtl : export the LLaMA computation graph * ci : disable temporary * mtl : adapt the MNIST example as starter * mtl : no need for mtl-export tool, add cli arg for main instead * mtl : export just a small part of the graph for now to make it easier * mtl : move MSL code into separate file for easy editing * mtl : initial get_rows_q4_0 kernel * mtl : confirmed get_rows_q4_0 is working correctly * mtl : add rms_norm kernel + confirm working * mtl : add mul kernel + confirm working * mtl : initial mul_mat Q4 kernel (wrong results) * mtl : mul_mat fixes (still wrong) * mtl : another mul_mat Q4 (still does not work) * mtl : working mul_mat q4 * ggml : fix handling of "view" ops in ggml_graph_import() * mtl : add rope kernel * mtl : add reshape and transpose handling * ggml : store offset as opt arg for ggml_view_xd() operators * mtl : add cpy kernel + handle view ops * mtl : confirm f16 x f32 attention mul mat * mtl : add scale kernel * mtl : add diag_mask_inf kernel * mtl : fix soft_max kernel * ggml : update ggml_nbytes() to handle non-contiguous tensors * mtl : verify V tensor contents * mtl : add f32 -> f32 cpy kernel * mtl : add silu kernel * mtl : add non-broadcast mul kernel * mtl : full GPU inference of the computation graph * mtl : optimize rms_norm and soft_max kernels * mtl : add f16 mat x f32 vec multiplication kernel * mtl : fix bug in f16 x f32 mul mat + speed-up computation * mtl : faster mul_mat_q4_0_f32 kernel * mtl : fix kernel signature + roll inner loop * mtl : more threads for rms_norm + better timing * mtl : remove printfs from inner loop * mtl : simplify implementation * mtl : add save/load vocab to ggml file * mtl : plug Metal inference into llama.cpp (very quick-n-dirty) * mtl : make it work with main example Lots of hacks but at least now it generates text * mtl : preparing for merge * mtl : clean-up ggml mtl interface + suport scratch / inplace * mtl : remove temp / debug code * metal : final refactoring and simplification * Revert "ci : disable temporary" This reverts commit 98c267fc77fe811082f672538fc91bcfc9072d63. * metal : add comments * metal : clean-up stuff, fix typos * readme : add Metal instructions * readme : add example for main
2023-06-04 20:34:30 +00:00
endif # LLAMA_METAL
ifdef LLAMA_METAL
2024-03-15 09:36:50 +00:00
ggml-metal.o: ggml-metal.m ggml-metal.h ggml.h
$(CC) $(CFLAGS) -c $< -o $@
ifdef LLAMA_METAL_EMBED_LIBRARY
ggml-metal-embed.o: ggml-metal.metal ggml-common.h
@echo "Embedding Metal library"
@sed -e '/#include "ggml-common.h"/r ggml-common.h' -e '/#include "ggml-common.h"/d' < ggml-metal.metal > ggml-metal-embed.metal
$(eval TEMP_ASSEMBLY=$(shell mktemp))
@echo ".section __DATA, __ggml_metallib" > $(TEMP_ASSEMBLY)
@echo ".globl _ggml_metallib_start" >> $(TEMP_ASSEMBLY)
@echo "_ggml_metallib_start:" >> $(TEMP_ASSEMBLY)
@echo ".incbin \"ggml-metal-embed.metal\"" >> $(TEMP_ASSEMBLY)
@echo ".globl _ggml_metallib_end" >> $(TEMP_ASSEMBLY)
@echo "_ggml_metallib_end:" >> $(TEMP_ASSEMBLY)
@$(AS) $(TEMP_ASSEMBLY) -o $@
@rm -f ${TEMP_ASSEMBLY}
endif
endif # LLAMA_METAL
OBJS += ggml-alloc.o ggml-backend.o ggml-quants.o unicode.o unicode-data.o
COMMON_H_DEPS = common/common.h common/sampling.h common/log.h llama.h
COMMON_DEPS = common.o sampling.o grammar-parser.o build-info.o json-schema-to-grammar.o
ifndef LLAMA_NO_LLAMAFILE
sgemm.o: sgemm.cpp sgemm.h ggml.h
$(CXX) $(CXXFLAGS) -c $< -o $@
endif
ifdef LLAMA_RPC
ggml-rpc.o: ggml-rpc.cpp ggml-rpc.h
$(CXX) $(CXXFLAGS) -c $< -o $@
rpc-server.o: examples/rpc/rpc-server.cpp ggml-rpc.h
$(CXX) $(CXXFLAGS) -c $< -o $@
rpc-server: rpc-server.o ggml.o llama.o $(COMMON_DEPS) $(OBJS)
$(CXX) $(CXXFLAGS) $^ -o $@ $(LDFLAGS)
endif # LLAMA_RPC
GF_CC := $(CC)
include scripts/get-flags.mk
# combine build flags with cmdline overrides
override CPPFLAGS := $(MK_CPPFLAGS) $(CPPFLAGS)
override CFLAGS := $(CPPFLAGS) $(MK_CFLAGS) $(GF_CFLAGS) $(CFLAGS)
BASE_CXXFLAGS := $(MK_CXXFLAGS) $(CXXFLAGS)
override CXXFLAGS := $(BASE_CXXFLAGS) $(HOST_CXXFLAGS) $(GF_CXXFLAGS) $(CPPFLAGS)
override NVCCFLAGS := $(MK_NVCCFLAGS) $(NVCCFLAGS)
override LDFLAGS := $(MK_LDFLAGS) $(LDFLAGS)
# identify CUDA host compiler
ifdef LLAMA_CUDA
GF_CC := $(NVCC) $(NVCCFLAGS) 2>/dev/null .c -Xcompiler
include scripts/get-flags.mk
CUDA_CXXFLAGS := $(BASE_CXXFLAGS) $(GF_CXXFLAGS) -Wno-pedantic
endif
ifdef LLAMA_CURL
override CXXFLAGS := $(CXXFLAGS) -DLLAMA_USE_CURL
override LDFLAGS := $(LDFLAGS) -lcurl
endif
2023-03-10 18:40:58 +00:00
#
# 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 NVCCFLAGS: $(NVCCFLAGS))
$(info I LDFLAGS: $(LDFLAGS))
2024-02-03 19:15:13 +00:00
$(info I CC: $(shell $(CC) --version | head -n 1))
$(info I CXX: $(shell $(CXX) --version | head -n 1))
ifdef LLAMA_CUDA
2024-02-03 19:15:13 +00:00
$(info I NVCC: $(shell $(NVCC) --version | tail -n 1))
CUDA_VERSION := $(shell $(NVCC) --version | grep -oP 'release (\K[0-9]+\.[0-9])')
ifeq ($(shell awk -v "v=$(CUDA_VERSION)" 'BEGIN { print (v < 11.7) }'),1)
ifndef CUDA_DOCKER_ARCH
ifndef CUDA_POWER_ARCH
$(error I ERROR: For CUDA versions < 11.7 a target CUDA architecture must be explicitly provided via environment variable CUDA_DOCKER_ARCH, e.g. by running "export CUDA_DOCKER_ARCH=compute_XX" on Unix-like systems, where XX is the minimum compute capability that the code needs to run on. A list with compute capabilities can be found here: https://developer.nvidia.com/cuda-gpus )
endif # CUDA_POWER_ARCH
endif # CUDA_DOCKER_ARCH
endif # eq ($(shell echo "$(CUDA_VERSION) < 11.7" | bc),1)
endif # LLAMA_CUDA
2023-03-10 18:40:58 +00:00
$(info )
ifdef LLAMA_CUBLAS
$(info !!!!)
$(info LLAMA_CUBLAS is deprecated and will be removed in the future. Use LLAMA_CUDA instead.)
$(info !!!!)
$(info )
endif
2023-03-10 18:40:58 +00:00
#
# Build library
#
ggml.o: ggml.c ggml.h ggml-cuda.h
$(CC) $(CFLAGS) -c $< -o $@
2023-03-10 18:40:58 +00:00
ggml-alloc.o: ggml-alloc.c ggml.h ggml-alloc.h
$(CC) $(CFLAGS) -c $< -o $@
ggml-backend.o: ggml-backend.c ggml.h ggml-backend.h
$(CC) $(CFLAGS) -c $< -o $@
ggml-quants.o: ggml-quants.c ggml.h ggml-quants.h ggml-common.h
$(CC) $(CFLAGS) -c $< -o $@
unicode.o: unicode.cpp unicode.h
$(CXX) $(CXXFLAGS) -c $< -o $@
unicode-data.o: unicode-data.cpp unicode-data.h
$(CXX) $(CXXFLAGS) -c $< -o $@
llama.o: llama.cpp unicode.h ggml.h ggml-alloc.h ggml-backend.h ggml-cuda.h ggml-metal.h llama.h
$(CXX) $(CXXFLAGS) -c $< -o $@
common.o: common/common.cpp $(COMMON_H_DEPS)
$(CXX) $(CXXFLAGS) -c $< -o $@
sampling.o: common/sampling.cpp $(COMMON_H_DEPS)
$(CXX) $(CXXFLAGS) -c $< -o $@
2023-03-10 18:40:58 +00:00
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
console.o: common/console.cpp common/console.h
$(CXX) $(CXXFLAGS) -c $< -o $@
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
grammar-parser.o: common/grammar-parser.cpp common/grammar-parser.h
$(CXX) $(CXXFLAGS) -c $< -o $@
json-schema-to-grammar improvements (+ added to server) (#5978) * json: fix arrays (disallow `[,1]`) * json: support tuple types (`[number, string]`) * json: support additionalProperties (`{[k: string]: [string,number][]}`) * json: support required / optional properties * json: add support for pattern * json: resolve $ref (and support https schema urls) * json: fix $ref resolution * join: support union types (mostly for nullable types I think) * json: support allOf + nested anyOf * json: support any (`{}` or `{type: object}`) * json: fix merge * json: temp fix for escapes * json: spaces in output and unrestricted output spaces * json: add typings * json:fix typo * Create ts-type-to-grammar.sh * json: fix _format_literal (json.dumps already escapes quotes) * json: merge lit sequences and handle negatives {"type": "string", "pattern": "^({\"question\": \"[^\"]+\", \"response\": \"[^\"]+\"}\\n)+$"} * json: handle pattern repetitions * Update json-schema-to-grammar.mjs * Create regex-to-grammar.py * json: extract repeated regexp patterns to subrule * Update json-schema-to-grammar.py * Update json-schema-to-grammar.py * Update json-schema-to-grammar.py * json: handle schema from pydantic Optional fields * Update json-schema-to-grammar.py * Update json-schema-to-grammar.py * Update ts-type-to-grammar.sh * Update ts-type-to-grammar.sh * json: simplify nullable fields handling * json: accept duplicate identical rules * json: revert space to 1 at most * json: reuse regexp pattern subrules * json: handle uuid string format * json: fix literal escapes * json: add --allow-fetch * json: simplify range escapes * json: support negative ranges in patterns * Delete commit.txt * json: custom regex parser, adds dot support & JS-portable * json: rm trailing spaces * Update json-schema-to-grammar.mjs * json: updated server & chat `( cd examples/server && ./deps.sh )` * json: port fixes from mjs to python * Update ts-type-to-grammar.sh * json: support prefixItems alongside array items * json: add date format + fix uuid * json: add date, time, date-time formats * json: preserve order of props from TS defs * json: port schema converter to C++, wire in ./server * json: nits * Update json-schema-to-grammar.cpp * Update json-schema-to-grammar.cpp * Update json-schema-to-grammar.cpp * json: fix mjs implementation + align outputs * Update json-schema-to-grammar.mjs.hpp * json: test C++, JS & Python versions * json: nits + regen deps * json: cleanup test * json: revert from c++17 to 11 * json: nit fixes * json: dirty include for test * json: fix zig build * json: pass static command to std::system in tests (fixed temp files) * json: fix top-level $refs * json: don't use c++20 designated initializers * nit * json: basic support for reserved names `{number:{number:{root:number}}}` * Revamp test cmake to allow args (WORKING_DIRECTORY needed for JSON test) * json: re-ran server deps.sh * json: simplify test * json: support mix of additional props & required/optional * json: add tests for some expected failures * json: fix type=const in c++, add failure expectations for non-str const&enum * json: test (& simplify output of) empty schema * json: check parsing in test + fix value & string refs * json: add server tests for OAI JSON response_format * json: test/fix top-level anyOf * json: improve grammar parsing failures * json: test/fix additional props corner cases * json: fix string patterns (was missing quotes) * json: ws nit * json: fix json handling in server when there's no response_format * json: catch schema conversion errors in server * json: don't complain about unknown format type in server if unset * json: cleaner build of test * json: create examples/json-schema-pydantic-example.py * json: fix date pattern * json: move json.hpp & json-schema-to-grammar.{cpp,h} to common * json: indent 4 spaces * json: fix naming of top-level c++ function (+ drop unused one) * json: avoid using namespace std * json: fix zig build * Update server.feature * json: iostream -> fprintf * json: space before & refs for consistency * json: nits
2024-03-21 11:50:43 +00:00
json-schema-to-grammar.o: common/json-schema-to-grammar.cpp common/json-schema-to-grammar.h
$(CXX) $(CXXFLAGS) -c $< -o $@
train : finetune LORA (#2632) * fix track_max_mem in forward_batch_wo_cache_flash_attn_train * remove unnecessary Adam(W) optimizer tensors. reduces optimizer memory overhead from 7*modelsize to 2*modelsize. additionally allows to optimize models with more than 2^31 parameters by replacing int with int64_t. bumps training checkpoint file version, but old checkpoints can still be read. new version with less tensors is saved. * add gradient clipping to AdamW * Fix reset of unused g->nodes and g->grads to NULL * implement gradient checkpointing for training reduces memory overhead from O(n_layer) to O(sqrt(n_layer)) as explained in readme of https://github.com/cybertronai/gradient-checkpointing * remove unused compute buffer 3 * add and use function ggml_build_backward_expand to avoid stack overflows with large maximum number of nodes GGML_API void ggml_build_backward_expand(struct ggml_context * ctx, struct ggml_cgraph * gf, struct ggml_cgraph * gb, bool keep); * change AdamW decay parameter to work like the torch AdamW decay parameter It is now relative to Adam learning rate `alpha*sched`. Before that it was relative to `sched` only. `alpha` being the maximum learning rate and `sched` being a scaling parameter in [0..1] * change default AdamW weight decay parameter used in training to 0.1 as used in nanoGPT * change default AdamW weight decay parameter defined in ggml to 0.0, making Adam default instead of AdamW btw: the default weight decay parameter for torch.optim.AdamW is 0.01 * bug fixes for cross entropy loss ggml_cross_entropy_loss: sums where not correctly added in workload of each thread ggml_cross_entropy_loss_back: simplify backward process, reducing numerical issues guard usage of exp f16 lookup in cross entropy by #define GGML_CROSS_ENTROPY_EXP_FP16 cross entropy loss is only used once during training, but it is quite sensitive to numerical errors introduced by exp-f16-lookup. so exp-f16-lookup for cross entropy loss is disabled by default, trading better gradients for very slightly worse runtime performance. * fix test-grad0 for cross_entropy_loss the second argument to cross_entropy_loss must sum up to 1 for each row * fix test-grad0 for soft_max dont use only sum as aggregation, because sum of softmax is always 1 -> finite differences should not work instead use sum(log(soft_max()*(1-eps)+eps)); use eps to avoid log(0) * improve finite differences of test-grad0 by using double instead of float * change cross_entropy_loss to output average over all rows this helps keeping the loss and gradients in a sane range * improve gradient checkpointing sqrt(n_layers) is only the best checkpoint step when mem size of checkpoints and mem size of layers are equal. since layers require more memory than the single-tensor-checkpoint we use, the optimal values are compute different: ``` given: n, u, v objective: minimize(a*u+b*v) where a*b=n, a>0, b>0 b=n/a minimize(a*u+v*n/a) diff(a*u+v*n/a, a) = u - (v*n/a)/a diff(a*u+v*n/a, a) == 0 u - (v*n/a)/a == 0 u == v*n/(a*a) u*a*a = v*n a*a = v*n/u a = sqrt(n*v/u) ``` this change results in more checkpoints, requiring less layers to store between checkpoints, overall improving memory usage. * disable gradient checkpointing debug output * llama : fix rope usage in train-text-from-scratch after ChatGLM change * add more training parameters: --enable-restart N Only for Adam optimizer. Enable restarts of cos-decay --disable-restart N Only for Adam optimizer. Disable restarts of cos-decay --opt-past N Number of optimization iterations to track for delta convergence test. Disabled when zero. --opt-delta N Maximum delta for delta convergence test. Disabled when <= zero. --opt-max-no-improvement N Maximum number of optimization iterations with no improvement. Disabled when <= zero. --adam-epsf N AdamW epsilon for convergence test. Disabled when <= zero. --adam-min-alpha N Adam minimum learning rate alpha, usually 0.1 * alpha * replace memcpy with reshape operation so that the graph is not cut at the input this makes it possible to store other values into the input tensor and then simply recompute the graph without rebuilding it * remove unused function argument from get_example_targets_batch * measure and print total training time * add optimization callback to ggml_opt_resume_g this callback is called before each iteration with custom data and pointer to learning schedule parameter (only used in Adam(W)). can be used for dynamic learning schedule and setting input data for batches before each iteration * use optimization callback in training allows dynamic learning schedule and different batch data for each iteration without relying on low n_iter and high n_examples parameters reduces runtime by avoiding restart of optimization function and improves training convergence by providing a different batch for each iteration * add minimum number of tensor dimensions to apply weight decay (default 2) this allows to not apply weight decay to bias parameters * rename training parameter cos-decay-alpha to cos-decay-min and clarify that adam-min-alpha also applies to warmup * fix increase of model.train_samples and model.train_tokens now that each optimizer iteration gets its own batch we need to multiply by number of opt iterations * change sampling parameters for prediction after training to defaults of common.h and clarify what is context for prediction and what are generated tokens * tighten abs error bounds for cross_entropy_loss in test-grad0 * add conditional compilation of using F16 exp in flash attention uncomment `// #define GGML_FLASH_ATTN_EXP_FP16` to enable usage of f16 exp in flash attention * tighten abs error bounds for flash_attn in test-grad0 * tighten abs error bounds for sqrt in test-grad0 * remove out-commented vectorized code of opt_adam the vectorized code might be bit faster for low number of parameters, but it had a big memory usage overhead * ggml : update ggml_rms_norm_back with configurable eps * llama training : fix ggml_rms_norm_back calls to pass configurable eps * remove trailing whitespace * add train function using automatic gradient checkpointing backward pass and allocator * in train function replace add_inplace by regular add because using add_inplace seems to result in different gradients * don't use allocate hash_map on context because the context has no_alloc=True when using memory allocator resulting in NULL data pointers * correctly clone reshape and permute operations by also cloning tensor->nb values * fix variable name and add missing type cast * terminate recursive tensor cloning when reaching tensor without src tensors * correctly clone view tensors by setting data pointers without this the checkpointing would only work when being used together with memory allocator * fix variable names * swap arguments to commutative ops to be the same as in `forward_batch_wo_cache_flash_attn` * add input tensors as checkpoints so that recursive tensor cloning of gradient checkpointing terminates on input tensors * fix variable name and add missing boolean negation * make sure some tensors are not reallocated by inserting new temporary nodes depending on them: output and parameter gradient tensors need to be available at the end of the graph execution parameter gradient tensors also need to be available before the graph execution because they are set to zero before each optimizer iteration checkpoint tensors are allocated all together to reduce memory allocator fragmentation afterwards, in addition to the temporary nodes, we also need to reset the temporary leafs * fix ASSERT to work with zero layers * add training options whether to use allocator and/or unified training function * integrate unified training function which may use memory allocator the unified training function also supports arguments whether to use flash attention and/or gradient checkpointing * format name of cloned tensors with " (clone)" suffix * set names for tensors in unified train function for easier debugging * allocate graph on context using ggml_new_graph * remove handwritten training functions * remove unused training parameters "use_scratch" and "use_unified" * remove trailing whitespace * remove unused train params: mem_compute1_gb & mem_compute2_gb mem_compute_gb is used for compute when automatic memory allocator is not enabled, otherwise it can be very small to only hold the tensor definitions mem_compute0_gb is used for automatic memory allocator (as long as measurement of max required size is not implemented) * remove unused forward_batch function * add debug asserts in ggml_allocr_alloc to some common pitfalls when using this function directly * only use ggml_allocr_alloc when tensor has NULL data and is no view * fix test when to create temporary backward graph temporary backward graph is only necessary when using checkpointing * fix memory "leak" in optimizers each iteration a new cplan with new memory for work data was allocated. now cplan creation only happens at the start of optimization, with each iteration reusing the cplan and its work data. * reverse order of for loop in ggml_build_backward_expand to save memory when using gradient checkpointing and allocator with this loop order gradient checkpointing with allocator on 16 layer model saves 13% memory; 2 layer memory it saves 2% memory. the computation results are the same * add API functions to access llama model tensors * add stub example for finetuning, based on train-text-from-scratch * move and remove code * add API functions to access remaining model parameters: mult, head and rot * first draft for LORA finetune training * remove const model and layer arguments in API functions for accessing model tensors * bug fixes to make finetune compile automatic allocator does not work yet * add debug prints for training memory improvements * fix names of lora tensors * avoid stack overflow resulting from big ggml_cgraph replace stack allocation and ggml_build_forward by ggml_new_graph in combination with ggml_build_forward_expand * replace llama API functions to get model tensors by one function to get model tensor by name LLAMA_API struct ggml_tensor * llama_get_model_tensor(struct llama_model * model, const char * name); * remove unused call to not existing llama_get_layer_from_model * implement ggml_compute_forward_out_prod_q_f32 * remove trailing whitespace * add lora finetune support on quantized base model tensors * add ggml_add_cast API function this function works like ggml_add, but accepts a data type for the resulting tensor. only supported for quantized src0 input. * use ggml_add_cast in finetuning lora-applied weights will now have data type F32, which improves gradients when finetuning quantized base models * bug fix: actually use result type passed to ggml_add_cast * make sure base model tensors data cannot be used in viewable operations memory allocator would try to make lora application inplace on base model tensors. since those are memory mapped this will result in memory access violations * fix bug in ggml_out_prod which resulted in wrong n_dims of result tensors * avoid keeping in memory ALL of the gradients The problem here stems from ggml_graph_reset. This function is called in the optimization function, before each graph computation, to reset the gradients to zero. This required a unique memory slot for each gradient: allocating memory from a previosly freed memory location might lead to non-zero input gradients. During ggml_compute_backward the gradients are build stepwise by adding or substracting new values, starting from a OP_NONE tensor which needs to contain zero-values. This requires the graph reset. To avoid this I now remember in ggml_build_backward_expand the original OP_NONE gradient tensors in a hash table, which is passed to ggml_compute_backward. There instead of using add (or sub or similar) I test whether the existing gradient to be changed is a zero-valued-tensor by looking up its existence in the hash table. When it is such a zero-tensor it will not be modified, but replaced by the value to be added, otherwise the regular add (not inplace, allocator will take care of this) will be used. This way none of those zero-tensor values will be necessary in the final backward graph and more importantly they won't need a unique memory slot, just to make them zero. * remove trailing whitespace * remove debug prints and function to compute tensor data hash * improve optimization iteration prints * adjust maximal values to support finetuning 3B models * change default finetune params lora_r and lora_alpha to match the n_rank parameters of 4 * bug fix: make sure finetune input gradient is allocated at begin and kept until end * remove unnecessary src tensor from ggml_get_rows_back we don't need data of src[2] for computation, only to setup the correct output shape. remove dependency on src[2], so that allocator can work more freely. the computational graph is still completely determined, because the output shape is naturally included. this is similar to how ggml_reshape does it. * remove unnecessary src tensor from ggml_repeat & ggml_repeat_back we don't need data of src[1] for computation, only to setup the correct output shape. remove dependency on src[1], so that allocator can work more freely. the computational graph is still completely determined, because the output shape is naturally included * resolve todo allocator will only make it inplace when they are of the same type * mixing multiple LORA adapters is now possible pass more than one '--lora FNAME' argument to apply more than one LORA. use '--lora-scaled FNAME S' when you want to specify a user-defined scale for an adapter. * add option to save finetune output every N iterations * also save latest finetune output with ITERATION="LATEST" and print where files are saved saving with LATEST makes it easier to resume training from the latest checkpoint the string "LATEST" can be configured with command line option "--fn-latest STR" * update checkpoint train stats before saving via "--save-every" * add command line option `--rank-wo N` for rank of wo tensor * update finetune README * fix dump_non_result_info_yaml to output multiple lora adapters * bug fix: replace GGML_TYPE_SIZE[t] by ggml_type_size(t) * replace llama_n_mult by llama_n_ff * finetune bug fixes to compile with merged in code from master * remove prediction related code to reduce duplicated code with main use main instead * reduce large memory overhead in train-text-from-scratch all gradients had to be pinned so that graph_reset works correctly. this is no longer necessary with the changes to ggml_compute_backward introduced in this PR. * add comment explaining why finetune checkpoints are allocated in one block * make default value of float member a float literal * handle rms_norm and rope parameters the same as in train-text-from-scratch * remove unused code * remove vocab related code as it is unnecessary * add LLM_KV_TRAINING_TYPE to train-text-from-scratch checkpoints so that they can be differentiated from lora finetune checkpoints * add gguf constants and load/save functions from train-text-from-scratch * add load & save lora finetune checkpoints via gguf * add python script to convert old finetune checkpoint files to gguf * remove old checkpoint save & load code * remove code to print data checksums which was used to verify correctness of new gguf code * omit tokenization when training is disabled, only save llama lora adapter training can be disabled by passing '-n 0' to finetune * remove trailing whitespace * update README.md * implement ggml_compute_forward_repeat_f16 * avoid stack overflow of large cgraphs in test-grad0 * add ggml API functions ggml_unravel_index, ggml_get_i32_nd and its analogs for set and for f32 ggml_get_i32_1d, ggml_set_i32_1d, ggml_get_f32_1d, ggml_set_f32_1d now support non-contiguous tensors. in case of non-contiguous tensor, the 1d index is unraveled into a multi index using ggml_unravel_index to be passed to '_nd' function equivalent. this fixes a bug in test-grad0 which happens due to ggml_build_backward not building purely contiguous tensors anymore * increase test-grad0 context mem size to accommodate for bigger cgraph * add sanity check to ggml_compute_backward, asserting the correct shape of gradients * fix ggml_acc_or_set to return tensor of correct shape * remove unused 'inplace' argument from ggml_compute_backward function inplace operations to add gradients are no longer created by ggml_compute_backward use allocator to automatically make inplace operations * add missing argument 'int i0' to ggml_get_i32_nd & ggml_set_i32_nd header declarations * fix error message in ggml_allocr_alloc to display actual max_avail * fix check_gradient ggml_build_backward_expand was previously replaced by ggml_build_backward, but the assignment of forward graph to backward graph missing * use tensor->view_src instead of ggml_is_view and get_view_source * move gradient checkpointing code into ggml, new API function: // build gradient checkpointing backward graph gb for gf using provided checkpoints // gb_tmp will contain original backward graph with rewritten backward process nodes, // but without the second forward pass nodes. GGML_API void ggml_build_backward_gradient_checkpointing( struct ggml_context * ctx, struct ggml_cgraph * gf, struct ggml_cgraph * gb, struct ggml_cgraph * gb_tmp, struct ggml_tensor * * checkpoints, int n_checkpoints); * replace custom data getters and setters by ggml functions * train-text-from-scratch can train (full finetune) gguf models just pass the gguf model via `--checkpoint-in FN`. after this, to continue training, pass the generated checkpoint instead of the original gguf model. tested with smaller models, bigger models may exceed available memory. use (LORA) finetune for those. * remove trailing whitespace * add option to save train-text-from-scratch output every N iterations * update README.md * fix warnings * fix warnings * remove finetune option to disable allocator the allocator should always be used. by making sure that it is always used it gets easier to implement automatic memory requirements computation * add tensor checkpoints only when gradient checkpointing is enabled * initialize opt ggml context if none was provided * add ggml-alloc API function 'ggml_allocr_max_size' to get max size of alloc GGML_API size_t ggml_allocr_max_size(struct ggml_allocr * alloc); * finetune: automatically allocate all memory and changes to command line options remove '--n_examples N' parameter, as it no longer makes sense to call optimization process multiple times in a loop. add '--only_write_lora' command line option: will skip tokenization and training, to only write a llama.cpp comptabile LORA adapter. remove memory buffer related command line options. improve iteration console output. * add finetune to Makefile * update README.md * print time per iteration and estimate remaining time * increase measured alloc size by tensor_alignment ggml_allocr_reset will reduce the given size by up to tensor_alignment-1 * fix README.md * add some more allocator debug prints * bug fix, probably solves the 'ggml_allocr_alloc: not enough space in the buffer' issue * revert last commit "bug fix, probably solves the 'ggml_allocr_alloc: not enough space in the buffer' issue" "alloc was freeing an externally allocated tensor, because it calculated the end of allocator memory as alloc->data + alloc->max_size instead of alloc->data + alloc->size." This is intentional to reduce the risk of freeing external tensors when measuring. Unless max_size is not properly calculated, I don't see why this is an issue. * remove unnecessary "0x" before "%p" output * move measurement memory segment to upper region of the address space * update README.md * fix printf format warnings * add missing gguf_free in load_checkpoint_lora_file * load default rms_norm and rope parameters from base model * add gradient accumulation specify number accumulation steps with '--grad-acc N'. this will simulate a bigger batch size of grad_acc*batch. * fix tracking of train_samples and train_tokens * build : fix compile warnings * ggml : fix L-BFGS linesearch loop * improve finetune time measurement fix printf warnings on system where int64_t is (long int). change time datatypes to double because values get big with long training times. exclude file saving from time measurement. converge faster to actual time per iteration by removing very small first duration before first iteration was performed. fix bug in output of total training time, the reported value was 1000 times to small. * specify default lora rank with '--lora-r N' '--lora-r N' will specify default rank for all tensors '--rank-wq N', etc. will override this default rank for specific tensor types. * fix gradient accumulation bug where the same batch was used for each microstep * fix gradient accumulation bug where the same batch was used for each microstep * support grouped-query-attention in ggml_flash_attn and ggml_flash_attn_back k and v can now be repeated in q along ne[2] in forward pass just use modulo to compute k and v indices, like ik2 = iq2 % nek2. in backard pass this won't work as easy, because multiple threads will compete to accumulate to the same k->grad[:,ik1,ik2,ik3] and v->grad[:,iv1,iv2,iv3]. so we change the parallelization over q rows to be over k rows. this ensures non-overlapping (ik2,ik3) across threads. in each thread we then iterate over the number of repetitions of k/v in q to compute iq2 as iq2 = ik2 + irep*nek2. since ne2 is not the same for q,k and v we also change how the gradients are concatenated into the result tensor. additionally the offsets of gradq, gradk and gradv in the result tensor are now memory aligned. we also simplify the compute_backward part of flash_attn to use ggml_reshape instead of switching over the number of dimensions. this needs a small change to ggml_reshape, removing the assertion of second argument to be contiguous. since only the shape (ne) of the second reshape argument is of relevance, its memory layout (nb) is irrelevant -> it can very well be non-contiguous. change test-grad0 to also test for repeated k/v in q. this changes the rng and now results in small gradient differences in softmax. these solely come from using f16 exp table lookup in forward softmax: when temporarily changing softmax to use actual exp function, the reported gradient differences go away. gradient differences coming solely from f16 table lookup are acceptable. added a note to explain this. * add llama API functions to get grouped-query-attention n_head parameter 'n_head_kv'. * fix finetune to support grouped-query-attention (using flash-attention) note: ggml changes to ggml_out_prod are necessary to support grouped-query-attention without flash-attention. * support broadcastable a in out_prod(a, b) and backward pass of broadcasting mul_mat(a, b) * test broadcasting mul_mat backward pass * decouple random number generator of each operation test when changing one test the rng of others tests is not influenced anymore * add comment briefly describing what ggml_repeat_back does * simplify broadcasting mul_mat backward using ggml_repeat_back * add cgraph evaluation order member and corresponding enum type this controls in which order ggml_build_forward visits source nodes. by default the nodes are visited left to right, i.e. src[0] first. in some cases it is beneficial for ggml-alloc to visit in a different order. two possible orders are supported: left-to-right (src[0] first) and right-to-left (src[0] last). * measure max compute size for each cgraph eval order and use best order this can bring huge memory savings: e.g. codellama-34b with n_ctx=64, n_batch=1 goes from 92927.8mb down to 4627.6 MB * remove unused command line options * add sample start patterns and options to force new or by default resume last shuffling * update shuffle rng state on reshuffle * exclude known zero values from computations in flash_attn_f32 & flash_attn_back_f32 * remove probably unnecessary exception type flags from stringstream * pass correct max number of tokens to llama_tokenize * account for possible leading whitespace that will be added by tokenizer e.g. '\t' will be tokenized by llama spm tokenizer to [29871, 12] * use unrolled vec_mad in out_prod y is vec_mad result vec. x is vec_mad input vec. v is vec_mad input scalar. ggml_vec_mad_f32_unroll will internally loop over x and v with same y. GGML_VEC_MAD_UNROLL is by default defined to 32. This value is empirical optimized using performance test runs of out-prod in openllama-3b finetune with 256 context length and batch size 1. It gives 23% performance boost for out_prod. Full measurements of out-prod runtime in ms: unroll_xv unroll_yv 1 67014.643 87826.469 2 77117.552 89077.656 4 72091.311 109121.657 8 61077.543 88678.334 16 56914.67 79514.947 24 59024.595 84350.254 28 55952.446 83368.73 32 51476.658 85177.745 36 55973.792 84659.92 40 55139.616 93844.738 48 60736.392 93330.267 64 99856.878 116994.99 Second column is when unrollying yv instead of xv * set lora_alpha to value of lora_r if it is not set via command line otherwise only changing lora_r will change scaling of lora adapter used in prediction * reshuffle original sample order instead of the previous shuffled order otherwise resumed reshuffle will not result in same sample order * block tiling for out-prod inspired by mul-mat block sizes are empirically optimized roughly doubles the flops of out-prod * exclude some more known zero values from computations in flash_attn_f32 & flash_attn_back_f32 * add static keywords * remove outcommented old code * update train-text-from-scratch with tokenization, sample selection and shuffling from finetune * remove lbfgs related train parameters * move common train functions into common/train.[h|cpp] * move train state into struct train_state * move train data saving code into callback to unify code of opt_callback train_params are still different in finetune and train-text-from-scratch, so it can't yet be moved to train.h|cpp * move common train params into common/train * move common opt_callback into common/train * fix consume_common_train_arg * save and load head_count_kv in lora checkpoints * increase train_samples by used_samples instead of number of batches on batch can contain more than one sample when option "fill_with_next_samples" is used * fix usage of llama_tokenize * remove static from process_escape since we need it exposed in header * fix code formating of long function declarations * fix condition in load_train_state_gguf * use die("msg") instead of replace GGML_ASSERT(!"msg") or throw std::runtime_error("msg") * fix saving and loading of training type * remove terminating '\0' from tokenization (llama_tokenize is now passed the string length instead of relying on terminating '\0') * fix compile warnings * fix compile warnings * use new/delete for train_state instead of malloc/free using malloc may result in seg faults when trying to assign string fields * assert that sample_count > 0, avoiding division by zero * fix frand to return value in interval [0,1) * add train option "--sample-random-offsets" Use samples beginning at random offsets. The offset is only applied to the first sample in each batch context window. Together with "--fill-with-next-samples" this may help for training endless text generation. For example given a dataset containing samples "abcd", "ABCD", "0123". With context size of 8 and options "--fill-with-next-samples", "--no-separate-with-eos", "--no-separate-with-bos", the context windows of batches could only be filled with "abcdABCD", "ABCDabcd", "0123abcd", etc. With "--sample-random-offsets" it can also be filled with "23abcdAB", "bcd0123A", etc. * deduplicate code into function * remove n_rot hparam, as it must always be hparam.n_embd_head() * align code * assert correct base model tensor shapes * move some params from lora hparams into model hparams and load model params from gguf this equalizes the model definition in finetune and text-from-scratch and removes the need for additional llama api functions to get model parameters * remove now unnecessary llama API functions to get model params that where added by this PR * train-text-from-scratch: automatically allocate model tensors, remove option '--mem-model N' * train-text-from-scratch: automatically allocate opt context * train-text-from-scratch: automatically allocate input tensors * train-text-from-scratch: automatically allocate compute memory * remove unused options and equalize train-text-from-scratch with finetune * initialize opt->loss_after with zero * add export-lora program * remove trailing whitespace * add export-lora build in Makefile * remove unused struct tensor_info from export-lora * add export-lora build dependency to llama because it depends on common, which depends on llama * update finetune README.md * cancel optimization when specified number of epochs is completed * improve handling of export-lora arguments print errors and warnings when files could not be read or created * Fix export-lora.cpp "not enough space in the context's memory pool" (#1) * Fix export-lora.cpp "not enough space in the context's memory pool" Without this patch, export-lora would sometimes error with "not enough space in the context's memory pool (needed 656784, available 656800)". * increase required context size by 5*GGML_MEM_ALIGN instead of plain 16 --------- Co-authored-by: xaedes <xaedes@gmail.com> * improve handling of not yet supported tensor types --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: meatbag-18a <145869052+meatbag-18a@users.noreply.github.com>
2023-09-28 18:40:11 +00:00
train.o: common/train.cpp common/train.h
$(CXX) $(CXXFLAGS) -c $< -o $@
ngram-cache.o: common/ngram-cache.cpp common/ngram-cache.h
$(CXX) $(CXXFLAGS) -c $< -o $@
libllama.so: llama.o ggml.o $(OBJS)
$(CXX) $(CXXFLAGS) -shared -fPIC -o $@ $^ $(LDFLAGS)
libllama.a: llama.o ggml.o $(OBJS) $(COMMON_DEPS)
ar rcs libllama.a llama.o ggml.o $(OBJS) $(COMMON_DEPS)
2023-03-10 18:40:58 +00:00
clean:
rm -vrf *.o tests/*.o *.so *.a *.dll benchmark-matmult lookup-create lookup-merge lookup-stats common/build-info.cpp *.dot $(COV_TARGETS) $(BUILD_TARGETS) $(TEST_TARGETS)
rm -vrf ggml-cuda/*.o
rm -vrf ggml-cuda/template-instances/*.o
find examples pocs -type f -name "*.o" -delete
2023-03-10 18:40:58 +00:00
#
# Examples
#
# $< is the first prerequisite, i.e. the source file.
# Explicitly compile this to an object file so that it can be cached with ccache.
# The source file is then filtered out from $^ (the list of all prerequisites) and the object file is added instead.
# Helper function that replaces .c, .cpp, and .cu file endings with .o:
GET_OBJ_FILE = $(patsubst %.c,%.o,$(patsubst %.cpp,%.o,$(patsubst %.cu,%.o,$(1))))
main: examples/main/main.cpp ggml.o llama.o $(COMMON_DEPS) console.o grammar-parser.o $(OBJS)
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
@echo
@echo '==== Run ./main -h for help. ===='
@echo
2023-03-10 18:40:58 +00:00
infill: examples/infill/infill.cpp ggml.o llama.o $(COMMON_DEPS) console.o grammar-parser.o $(OBJS)
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
simple: examples/simple/simple.cpp ggml.o llama.o $(COMMON_DEPS) $(OBJS)
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
tokenize: examples/tokenize/tokenize.cpp ggml.o llama.o $(COMMON_DEPS) $(OBJS)
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
batched: examples/batched/batched.cpp ggml.o llama.o $(COMMON_DEPS) $(OBJS)
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
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 16:04:36 +00:00
batched-bench: examples/batched-bench/batched-bench.cpp build-info.o ggml.o llama.o $(COMMON_DEPS) $(OBJS)
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
quantize: examples/quantize/quantize.cpp ggml.o llama.o $(COMMON_DEPS) $(OBJS)
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
quantize-stats: examples/quantize-stats/quantize-stats.cpp build-info.o ggml.o llama.o $(OBJS)
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
perplexity: examples/perplexity/perplexity.cpp ggml.o llama.o $(COMMON_DEPS) $(OBJS)
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
imatrix: examples/imatrix/imatrix.cpp ggml.o llama.o $(COMMON_DEPS) $(OBJS)
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
2023-03-10 18:40:58 +00:00
embedding: examples/embedding/embedding.cpp ggml.o llama.o $(COMMON_DEPS) $(OBJS)
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
gritlm: examples/gritlm/gritlm.cpp ggml.o llama.o $(COMMON_DEPS) $(OBJS)
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
save-load-state: examples/save-load-state/save-load-state.cpp ggml.o llama.o $(COMMON_DEPS) $(OBJS)
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
server: examples/server/server.cpp examples/server/utils.hpp examples/server/httplib.h common/json.hpp examples/server/colorthemes.css.hpp examples/server/style.css.hpp examples/server/theme-beeninorder.css.hpp examples/server/theme-ketivah.css.hpp examples/server/theme-mangotango.css.hpp examples/server/theme-playground.css.hpp examples/server/theme-polarnight.css.hpp examples/server/theme-snowstorm.css.hpp examples/server/index.html.hpp examples/server/index-new.html.hpp examples/server/index.js.hpp examples/server/completion.js.hpp examples/server/system-prompts.js.hpp examples/server/prompt-formats.js.hpp examples/server/json-schema-to-grammar.mjs.hpp common/stb_image.h ggml.o llama.o $(COMMON_DEPS) grammar-parser.o $(OBJS)
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
2024-03-07 09:41:53 +00:00
$(CXX) $(CXXFLAGS) $(filter-out %.h %.hpp $<,$^) -Iexamples/server $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS) $(LWINSOCK2)
# Portable equivalent of `cd examples/server/public && xxd -i $(notdir $<) ../$(notdir $<).hpp`:
examples/server/%.hpp: examples/server/public/% Makefile
@( export NAME=$(subst .,_,$(subst -,_,$(notdir $<))) && \
echo "unsigned char $${NAME}[] = {" && \
cat $< | od -v -t x1 -An | sed -E 's/([0-9a-fA-F]+)/0x\1, /g' && \
echo "};" && \
echo "unsigned int $${NAME}_len = $(shell cat $< | wc -c );" \
) > $@
2023-12-21 21:07:58 +00:00
gguf: examples/gguf/gguf.cpp ggml.o $(OBJS)
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
gguf-split: examples/gguf-split/gguf-split.cpp ggml.o llama.o $(COMMON_DEPS) $(OBJS)
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
eval-callback: examples/eval-callback/eval-callback.cpp ggml.o llama.o $(COMMON_DEPS) $(OBJS)
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
train-text-from-scratch: examples/train-text-from-scratch/train-text-from-scratch.cpp ggml.o llama.o $(COMMON_DEPS) train.o $(OBJS)
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
convert-llama2c-to-ggml: examples/convert-llama2c-to-ggml/convert-llama2c-to-ggml.cpp ggml.o llama.o $(OBJS)
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
llama-bench: examples/llama-bench/llama-bench.cpp ggml.o llama.o $(COMMON_DEPS) $(OBJS)
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
libllava.a: examples/llava/llava.cpp examples/llava/llava.h examples/llava/clip.cpp examples/llava/clip.h common/stb_image.h common/base64.hpp ggml.o llama.o $(COMMON_DEPS) $(OBJS)
$(CXX) $(CXXFLAGS) -static -fPIC -c $< -o $@ -Wno-cast-qual
llava-cli: examples/llava/llava-cli.cpp examples/llava/clip.h examples/llava/clip.cpp examples/llava/llava.h examples/llava/llava.cpp ggml.o llama.o $(COMMON_DEPS) $(OBJS)
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) -c examples/llava/clip.cpp -o $(call GET_OBJ_FILE, examples/llava/clip.cpp) -Wno-cast-qual
$(CXX) $(CXXFLAGS) -c examples/llava/llava.cpp -o $(call GET_OBJ_FILE, examples/llava/llava.cpp)
$(CXX) $(CXXFLAGS) $(filter-out %.h $< examples/llava/clip.cpp examples/llava/llava.cpp,$^) $(call GET_OBJ_FILE, $<) $(call GET_OBJ_FILE, examples/llava/clip.cpp) $(call GET_OBJ_FILE, examples/llava/llava.cpp) -o $@ $(LDFLAGS)
baby-llama: examples/baby-llama/baby-llama.cpp ggml.o llama.o $(COMMON_DEPS) train.o $(OBJS)
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
finetune: examples/finetune/finetune.cpp ggml.o llama.o $(COMMON_DEPS) train.o $(OBJS)
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
train : finetune LORA (#2632) * fix track_max_mem in forward_batch_wo_cache_flash_attn_train * remove unnecessary Adam(W) optimizer tensors. reduces optimizer memory overhead from 7*modelsize to 2*modelsize. additionally allows to optimize models with more than 2^31 parameters by replacing int with int64_t. bumps training checkpoint file version, but old checkpoints can still be read. new version with less tensors is saved. * add gradient clipping to AdamW * Fix reset of unused g->nodes and g->grads to NULL * implement gradient checkpointing for training reduces memory overhead from O(n_layer) to O(sqrt(n_layer)) as explained in readme of https://github.com/cybertronai/gradient-checkpointing * remove unused compute buffer 3 * add and use function ggml_build_backward_expand to avoid stack overflows with large maximum number of nodes GGML_API void ggml_build_backward_expand(struct ggml_context * ctx, struct ggml_cgraph * gf, struct ggml_cgraph * gb, bool keep); * change AdamW decay parameter to work like the torch AdamW decay parameter It is now relative to Adam learning rate `alpha*sched`. Before that it was relative to `sched` only. `alpha` being the maximum learning rate and `sched` being a scaling parameter in [0..1] * change default AdamW weight decay parameter used in training to 0.1 as used in nanoGPT * change default AdamW weight decay parameter defined in ggml to 0.0, making Adam default instead of AdamW btw: the default weight decay parameter for torch.optim.AdamW is 0.01 * bug fixes for cross entropy loss ggml_cross_entropy_loss: sums where not correctly added in workload of each thread ggml_cross_entropy_loss_back: simplify backward process, reducing numerical issues guard usage of exp f16 lookup in cross entropy by #define GGML_CROSS_ENTROPY_EXP_FP16 cross entropy loss is only used once during training, but it is quite sensitive to numerical errors introduced by exp-f16-lookup. so exp-f16-lookup for cross entropy loss is disabled by default, trading better gradients for very slightly worse runtime performance. * fix test-grad0 for cross_entropy_loss the second argument to cross_entropy_loss must sum up to 1 for each row * fix test-grad0 for soft_max dont use only sum as aggregation, because sum of softmax is always 1 -> finite differences should not work instead use sum(log(soft_max()*(1-eps)+eps)); use eps to avoid log(0) * improve finite differences of test-grad0 by using double instead of float * change cross_entropy_loss to output average over all rows this helps keeping the loss and gradients in a sane range * improve gradient checkpointing sqrt(n_layers) is only the best checkpoint step when mem size of checkpoints and mem size of layers are equal. since layers require more memory than the single-tensor-checkpoint we use, the optimal values are compute different: ``` given: n, u, v objective: minimize(a*u+b*v) where a*b=n, a>0, b>0 b=n/a minimize(a*u+v*n/a) diff(a*u+v*n/a, a) = u - (v*n/a)/a diff(a*u+v*n/a, a) == 0 u - (v*n/a)/a == 0 u == v*n/(a*a) u*a*a = v*n a*a = v*n/u a = sqrt(n*v/u) ``` this change results in more checkpoints, requiring less layers to store between checkpoints, overall improving memory usage. * disable gradient checkpointing debug output * llama : fix rope usage in train-text-from-scratch after ChatGLM change * add more training parameters: --enable-restart N Only for Adam optimizer. Enable restarts of cos-decay --disable-restart N Only for Adam optimizer. Disable restarts of cos-decay --opt-past N Number of optimization iterations to track for delta convergence test. Disabled when zero. --opt-delta N Maximum delta for delta convergence test. Disabled when <= zero. --opt-max-no-improvement N Maximum number of optimization iterations with no improvement. Disabled when <= zero. --adam-epsf N AdamW epsilon for convergence test. Disabled when <= zero. --adam-min-alpha N Adam minimum learning rate alpha, usually 0.1 * alpha * replace memcpy with reshape operation so that the graph is not cut at the input this makes it possible to store other values into the input tensor and then simply recompute the graph without rebuilding it * remove unused function argument from get_example_targets_batch * measure and print total training time * add optimization callback to ggml_opt_resume_g this callback is called before each iteration with custom data and pointer to learning schedule parameter (only used in Adam(W)). can be used for dynamic learning schedule and setting input data for batches before each iteration * use optimization callback in training allows dynamic learning schedule and different batch data for each iteration without relying on low n_iter and high n_examples parameters reduces runtime by avoiding restart of optimization function and improves training convergence by providing a different batch for each iteration * add minimum number of tensor dimensions to apply weight decay (default 2) this allows to not apply weight decay to bias parameters * rename training parameter cos-decay-alpha to cos-decay-min and clarify that adam-min-alpha also applies to warmup * fix increase of model.train_samples and model.train_tokens now that each optimizer iteration gets its own batch we need to multiply by number of opt iterations * change sampling parameters for prediction after training to defaults of common.h and clarify what is context for prediction and what are generated tokens * tighten abs error bounds for cross_entropy_loss in test-grad0 * add conditional compilation of using F16 exp in flash attention uncomment `// #define GGML_FLASH_ATTN_EXP_FP16` to enable usage of f16 exp in flash attention * tighten abs error bounds for flash_attn in test-grad0 * tighten abs error bounds for sqrt in test-grad0 * remove out-commented vectorized code of opt_adam the vectorized code might be bit faster for low number of parameters, but it had a big memory usage overhead * ggml : update ggml_rms_norm_back with configurable eps * llama training : fix ggml_rms_norm_back calls to pass configurable eps * remove trailing whitespace * add train function using automatic gradient checkpointing backward pass and allocator * in train function replace add_inplace by regular add because using add_inplace seems to result in different gradients * don't use allocate hash_map on context because the context has no_alloc=True when using memory allocator resulting in NULL data pointers * correctly clone reshape and permute operations by also cloning tensor->nb values * fix variable name and add missing type cast * terminate recursive tensor cloning when reaching tensor without src tensors * correctly clone view tensors by setting data pointers without this the checkpointing would only work when being used together with memory allocator * fix variable names * swap arguments to commutative ops to be the same as in `forward_batch_wo_cache_flash_attn` * add input tensors as checkpoints so that recursive tensor cloning of gradient checkpointing terminates on input tensors * fix variable name and add missing boolean negation * make sure some tensors are not reallocated by inserting new temporary nodes depending on them: output and parameter gradient tensors need to be available at the end of the graph execution parameter gradient tensors also need to be available before the graph execution because they are set to zero before each optimizer iteration checkpoint tensors are allocated all together to reduce memory allocator fragmentation afterwards, in addition to the temporary nodes, we also need to reset the temporary leafs * fix ASSERT to work with zero layers * add training options whether to use allocator and/or unified training function * integrate unified training function which may use memory allocator the unified training function also supports arguments whether to use flash attention and/or gradient checkpointing * format name of cloned tensors with " (clone)" suffix * set names for tensors in unified train function for easier debugging * allocate graph on context using ggml_new_graph * remove handwritten training functions * remove unused training parameters "use_scratch" and "use_unified" * remove trailing whitespace * remove unused train params: mem_compute1_gb & mem_compute2_gb mem_compute_gb is used for compute when automatic memory allocator is not enabled, otherwise it can be very small to only hold the tensor definitions mem_compute0_gb is used for automatic memory allocator (as long as measurement of max required size is not implemented) * remove unused forward_batch function * add debug asserts in ggml_allocr_alloc to some common pitfalls when using this function directly * only use ggml_allocr_alloc when tensor has NULL data and is no view * fix test when to create temporary backward graph temporary backward graph is only necessary when using checkpointing * fix memory "leak" in optimizers each iteration a new cplan with new memory for work data was allocated. now cplan creation only happens at the start of optimization, with each iteration reusing the cplan and its work data. * reverse order of for loop in ggml_build_backward_expand to save memory when using gradient checkpointing and allocator with this loop order gradient checkpointing with allocator on 16 layer model saves 13% memory; 2 layer memory it saves 2% memory. the computation results are the same * add API functions to access llama model tensors * add stub example for finetuning, based on train-text-from-scratch * move and remove code * add API functions to access remaining model parameters: mult, head and rot * first draft for LORA finetune training * remove const model and layer arguments in API functions for accessing model tensors * bug fixes to make finetune compile automatic allocator does not work yet * add debug prints for training memory improvements * fix names of lora tensors * avoid stack overflow resulting from big ggml_cgraph replace stack allocation and ggml_build_forward by ggml_new_graph in combination with ggml_build_forward_expand * replace llama API functions to get model tensors by one function to get model tensor by name LLAMA_API struct ggml_tensor * llama_get_model_tensor(struct llama_model * model, const char * name); * remove unused call to not existing llama_get_layer_from_model * implement ggml_compute_forward_out_prod_q_f32 * remove trailing whitespace * add lora finetune support on quantized base model tensors * add ggml_add_cast API function this function works like ggml_add, but accepts a data type for the resulting tensor. only supported for quantized src0 input. * use ggml_add_cast in finetuning lora-applied weights will now have data type F32, which improves gradients when finetuning quantized base models * bug fix: actually use result type passed to ggml_add_cast * make sure base model tensors data cannot be used in viewable operations memory allocator would try to make lora application inplace on base model tensors. since those are memory mapped this will result in memory access violations * fix bug in ggml_out_prod which resulted in wrong n_dims of result tensors * avoid keeping in memory ALL of the gradients The problem here stems from ggml_graph_reset. This function is called in the optimization function, before each graph computation, to reset the gradients to zero. This required a unique memory slot for each gradient: allocating memory from a previosly freed memory location might lead to non-zero input gradients. During ggml_compute_backward the gradients are build stepwise by adding or substracting new values, starting from a OP_NONE tensor which needs to contain zero-values. This requires the graph reset. To avoid this I now remember in ggml_build_backward_expand the original OP_NONE gradient tensors in a hash table, which is passed to ggml_compute_backward. There instead of using add (or sub or similar) I test whether the existing gradient to be changed is a zero-valued-tensor by looking up its existence in the hash table. When it is such a zero-tensor it will not be modified, but replaced by the value to be added, otherwise the regular add (not inplace, allocator will take care of this) will be used. This way none of those zero-tensor values will be necessary in the final backward graph and more importantly they won't need a unique memory slot, just to make them zero. * remove trailing whitespace * remove debug prints and function to compute tensor data hash * improve optimization iteration prints * adjust maximal values to support finetuning 3B models * change default finetune params lora_r and lora_alpha to match the n_rank parameters of 4 * bug fix: make sure finetune input gradient is allocated at begin and kept until end * remove unnecessary src tensor from ggml_get_rows_back we don't need data of src[2] for computation, only to setup the correct output shape. remove dependency on src[2], so that allocator can work more freely. the computational graph is still completely determined, because the output shape is naturally included. this is similar to how ggml_reshape does it. * remove unnecessary src tensor from ggml_repeat & ggml_repeat_back we don't need data of src[1] for computation, only to setup the correct output shape. remove dependency on src[1], so that allocator can work more freely. the computational graph is still completely determined, because the output shape is naturally included * resolve todo allocator will only make it inplace when they are of the same type * mixing multiple LORA adapters is now possible pass more than one '--lora FNAME' argument to apply more than one LORA. use '--lora-scaled FNAME S' when you want to specify a user-defined scale for an adapter. * add option to save finetune output every N iterations * also save latest finetune output with ITERATION="LATEST" and print where files are saved saving with LATEST makes it easier to resume training from the latest checkpoint the string "LATEST" can be configured with command line option "--fn-latest STR" * update checkpoint train stats before saving via "--save-every" * add command line option `--rank-wo N` for rank of wo tensor * update finetune README * fix dump_non_result_info_yaml to output multiple lora adapters * bug fix: replace GGML_TYPE_SIZE[t] by ggml_type_size(t) * replace llama_n_mult by llama_n_ff * finetune bug fixes to compile with merged in code from master * remove prediction related code to reduce duplicated code with main use main instead * reduce large memory overhead in train-text-from-scratch all gradients had to be pinned so that graph_reset works correctly. this is no longer necessary with the changes to ggml_compute_backward introduced in this PR. * add comment explaining why finetune checkpoints are allocated in one block * make default value of float member a float literal * handle rms_norm and rope parameters the same as in train-text-from-scratch * remove unused code * remove vocab related code as it is unnecessary * add LLM_KV_TRAINING_TYPE to train-text-from-scratch checkpoints so that they can be differentiated from lora finetune checkpoints * add gguf constants and load/save functions from train-text-from-scratch * add load & save lora finetune checkpoints via gguf * add python script to convert old finetune checkpoint files to gguf * remove old checkpoint save & load code * remove code to print data checksums which was used to verify correctness of new gguf code * omit tokenization when training is disabled, only save llama lora adapter training can be disabled by passing '-n 0' to finetune * remove trailing whitespace * update README.md * implement ggml_compute_forward_repeat_f16 * avoid stack overflow of large cgraphs in test-grad0 * add ggml API functions ggml_unravel_index, ggml_get_i32_nd and its analogs for set and for f32 ggml_get_i32_1d, ggml_set_i32_1d, ggml_get_f32_1d, ggml_set_f32_1d now support non-contiguous tensors. in case of non-contiguous tensor, the 1d index is unraveled into a multi index using ggml_unravel_index to be passed to '_nd' function equivalent. this fixes a bug in test-grad0 which happens due to ggml_build_backward not building purely contiguous tensors anymore * increase test-grad0 context mem size to accommodate for bigger cgraph * add sanity check to ggml_compute_backward, asserting the correct shape of gradients * fix ggml_acc_or_set to return tensor of correct shape * remove unused 'inplace' argument from ggml_compute_backward function inplace operations to add gradients are no longer created by ggml_compute_backward use allocator to automatically make inplace operations * add missing argument 'int i0' to ggml_get_i32_nd & ggml_set_i32_nd header declarations * fix error message in ggml_allocr_alloc to display actual max_avail * fix check_gradient ggml_build_backward_expand was previously replaced by ggml_build_backward, but the assignment of forward graph to backward graph missing * use tensor->view_src instead of ggml_is_view and get_view_source * move gradient checkpointing code into ggml, new API function: // build gradient checkpointing backward graph gb for gf using provided checkpoints // gb_tmp will contain original backward graph with rewritten backward process nodes, // but without the second forward pass nodes. GGML_API void ggml_build_backward_gradient_checkpointing( struct ggml_context * ctx, struct ggml_cgraph * gf, struct ggml_cgraph * gb, struct ggml_cgraph * gb_tmp, struct ggml_tensor * * checkpoints, int n_checkpoints); * replace custom data getters and setters by ggml functions * train-text-from-scratch can train (full finetune) gguf models just pass the gguf model via `--checkpoint-in FN`. after this, to continue training, pass the generated checkpoint instead of the original gguf model. tested with smaller models, bigger models may exceed available memory. use (LORA) finetune for those. * remove trailing whitespace * add option to save train-text-from-scratch output every N iterations * update README.md * fix warnings * fix warnings * remove finetune option to disable allocator the allocator should always be used. by making sure that it is always used it gets easier to implement automatic memory requirements computation * add tensor checkpoints only when gradient checkpointing is enabled * initialize opt ggml context if none was provided * add ggml-alloc API function 'ggml_allocr_max_size' to get max size of alloc GGML_API size_t ggml_allocr_max_size(struct ggml_allocr * alloc); * finetune: automatically allocate all memory and changes to command line options remove '--n_examples N' parameter, as it no longer makes sense to call optimization process multiple times in a loop. add '--only_write_lora' command line option: will skip tokenization and training, to only write a llama.cpp comptabile LORA adapter. remove memory buffer related command line options. improve iteration console output. * add finetune to Makefile * update README.md * print time per iteration and estimate remaining time * increase measured alloc size by tensor_alignment ggml_allocr_reset will reduce the given size by up to tensor_alignment-1 * fix README.md * add some more allocator debug prints * bug fix, probably solves the 'ggml_allocr_alloc: not enough space in the buffer' issue * revert last commit "bug fix, probably solves the 'ggml_allocr_alloc: not enough space in the buffer' issue" "alloc was freeing an externally allocated tensor, because it calculated the end of allocator memory as alloc->data + alloc->max_size instead of alloc->data + alloc->size." This is intentional to reduce the risk of freeing external tensors when measuring. Unless max_size is not properly calculated, I don't see why this is an issue. * remove unnecessary "0x" before "%p" output * move measurement memory segment to upper region of the address space * update README.md * fix printf format warnings * add missing gguf_free in load_checkpoint_lora_file * load default rms_norm and rope parameters from base model * add gradient accumulation specify number accumulation steps with '--grad-acc N'. this will simulate a bigger batch size of grad_acc*batch. * fix tracking of train_samples and train_tokens * build : fix compile warnings * ggml : fix L-BFGS linesearch loop * improve finetune time measurement fix printf warnings on system where int64_t is (long int). change time datatypes to double because values get big with long training times. exclude file saving from time measurement. converge faster to actual time per iteration by removing very small first duration before first iteration was performed. fix bug in output of total training time, the reported value was 1000 times to small. * specify default lora rank with '--lora-r N' '--lora-r N' will specify default rank for all tensors '--rank-wq N', etc. will override this default rank for specific tensor types. * fix gradient accumulation bug where the same batch was used for each microstep * fix gradient accumulation bug where the same batch was used for each microstep * support grouped-query-attention in ggml_flash_attn and ggml_flash_attn_back k and v can now be repeated in q along ne[2] in forward pass just use modulo to compute k and v indices, like ik2 = iq2 % nek2. in backard pass this won't work as easy, because multiple threads will compete to accumulate to the same k->grad[:,ik1,ik2,ik3] and v->grad[:,iv1,iv2,iv3]. so we change the parallelization over q rows to be over k rows. this ensures non-overlapping (ik2,ik3) across threads. in each thread we then iterate over the number of repetitions of k/v in q to compute iq2 as iq2 = ik2 + irep*nek2. since ne2 is not the same for q,k and v we also change how the gradients are concatenated into the result tensor. additionally the offsets of gradq, gradk and gradv in the result tensor are now memory aligned. we also simplify the compute_backward part of flash_attn to use ggml_reshape instead of switching over the number of dimensions. this needs a small change to ggml_reshape, removing the assertion of second argument to be contiguous. since only the shape (ne) of the second reshape argument is of relevance, its memory layout (nb) is irrelevant -> it can very well be non-contiguous. change test-grad0 to also test for repeated k/v in q. this changes the rng and now results in small gradient differences in softmax. these solely come from using f16 exp table lookup in forward softmax: when temporarily changing softmax to use actual exp function, the reported gradient differences go away. gradient differences coming solely from f16 table lookup are acceptable. added a note to explain this. * add llama API functions to get grouped-query-attention n_head parameter 'n_head_kv'. * fix finetune to support grouped-query-attention (using flash-attention) note: ggml changes to ggml_out_prod are necessary to support grouped-query-attention without flash-attention. * support broadcastable a in out_prod(a, b) and backward pass of broadcasting mul_mat(a, b) * test broadcasting mul_mat backward pass * decouple random number generator of each operation test when changing one test the rng of others tests is not influenced anymore * add comment briefly describing what ggml_repeat_back does * simplify broadcasting mul_mat backward using ggml_repeat_back * add cgraph evaluation order member and corresponding enum type this controls in which order ggml_build_forward visits source nodes. by default the nodes are visited left to right, i.e. src[0] first. in some cases it is beneficial for ggml-alloc to visit in a different order. two possible orders are supported: left-to-right (src[0] first) and right-to-left (src[0] last). * measure max compute size for each cgraph eval order and use best order this can bring huge memory savings: e.g. codellama-34b with n_ctx=64, n_batch=1 goes from 92927.8mb down to 4627.6 MB * remove unused command line options * add sample start patterns and options to force new or by default resume last shuffling * update shuffle rng state on reshuffle * exclude known zero values from computations in flash_attn_f32 & flash_attn_back_f32 * remove probably unnecessary exception type flags from stringstream * pass correct max number of tokens to llama_tokenize * account for possible leading whitespace that will be added by tokenizer e.g. '\t' will be tokenized by llama spm tokenizer to [29871, 12] * use unrolled vec_mad in out_prod y is vec_mad result vec. x is vec_mad input vec. v is vec_mad input scalar. ggml_vec_mad_f32_unroll will internally loop over x and v with same y. GGML_VEC_MAD_UNROLL is by default defined to 32. This value is empirical optimized using performance test runs of out-prod in openllama-3b finetune with 256 context length and batch size 1. It gives 23% performance boost for out_prod. Full measurements of out-prod runtime in ms: unroll_xv unroll_yv 1 67014.643 87826.469 2 77117.552 89077.656 4 72091.311 109121.657 8 61077.543 88678.334 16 56914.67 79514.947 24 59024.595 84350.254 28 55952.446 83368.73 32 51476.658 85177.745 36 55973.792 84659.92 40 55139.616 93844.738 48 60736.392 93330.267 64 99856.878 116994.99 Second column is when unrollying yv instead of xv * set lora_alpha to value of lora_r if it is not set via command line otherwise only changing lora_r will change scaling of lora adapter used in prediction * reshuffle original sample order instead of the previous shuffled order otherwise resumed reshuffle will not result in same sample order * block tiling for out-prod inspired by mul-mat block sizes are empirically optimized roughly doubles the flops of out-prod * exclude some more known zero values from computations in flash_attn_f32 & flash_attn_back_f32 * add static keywords * remove outcommented old code * update train-text-from-scratch with tokenization, sample selection and shuffling from finetune * remove lbfgs related train parameters * move common train functions into common/train.[h|cpp] * move train state into struct train_state * move train data saving code into callback to unify code of opt_callback train_params are still different in finetune and train-text-from-scratch, so it can't yet be moved to train.h|cpp * move common train params into common/train * move common opt_callback into common/train * fix consume_common_train_arg * save and load head_count_kv in lora checkpoints * increase train_samples by used_samples instead of number of batches on batch can contain more than one sample when option "fill_with_next_samples" is used * fix usage of llama_tokenize * remove static from process_escape since we need it exposed in header * fix code formating of long function declarations * fix condition in load_train_state_gguf * use die("msg") instead of replace GGML_ASSERT(!"msg") or throw std::runtime_error("msg") * fix saving and loading of training type * remove terminating '\0' from tokenization (llama_tokenize is now passed the string length instead of relying on terminating '\0') * fix compile warnings * fix compile warnings * use new/delete for train_state instead of malloc/free using malloc may result in seg faults when trying to assign string fields * assert that sample_count > 0, avoiding division by zero * fix frand to return value in interval [0,1) * add train option "--sample-random-offsets" Use samples beginning at random offsets. The offset is only applied to the first sample in each batch context window. Together with "--fill-with-next-samples" this may help for training endless text generation. For example given a dataset containing samples "abcd", "ABCD", "0123". With context size of 8 and options "--fill-with-next-samples", "--no-separate-with-eos", "--no-separate-with-bos", the context windows of batches could only be filled with "abcdABCD", "ABCDabcd", "0123abcd", etc. With "--sample-random-offsets" it can also be filled with "23abcdAB", "bcd0123A", etc. * deduplicate code into function * remove n_rot hparam, as it must always be hparam.n_embd_head() * align code * assert correct base model tensor shapes * move some params from lora hparams into model hparams and load model params from gguf this equalizes the model definition in finetune and text-from-scratch and removes the need for additional llama api functions to get model parameters * remove now unnecessary llama API functions to get model params that where added by this PR * train-text-from-scratch: automatically allocate model tensors, remove option '--mem-model N' * train-text-from-scratch: automatically allocate opt context * train-text-from-scratch: automatically allocate input tensors * train-text-from-scratch: automatically allocate compute memory * remove unused options and equalize train-text-from-scratch with finetune * initialize opt->loss_after with zero * add export-lora program * remove trailing whitespace * add export-lora build in Makefile * remove unused struct tensor_info from export-lora * add export-lora build dependency to llama because it depends on common, which depends on llama * update finetune README.md * cancel optimization when specified number of epochs is completed * improve handling of export-lora arguments print errors and warnings when files could not be read or created * Fix export-lora.cpp "not enough space in the context's memory pool" (#1) * Fix export-lora.cpp "not enough space in the context's memory pool" Without this patch, export-lora would sometimes error with "not enough space in the context's memory pool (needed 656784, available 656800)". * increase required context size by 5*GGML_MEM_ALIGN instead of plain 16 --------- Co-authored-by: xaedes <xaedes@gmail.com> * improve handling of not yet supported tensor types --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: meatbag-18a <145869052+meatbag-18a@users.noreply.github.com>
2023-09-28 18:40:11 +00:00
export-lora: examples/export-lora/export-lora.cpp ggml.o common/common.h $(OBJS)
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
train : finetune LORA (#2632) * fix track_max_mem in forward_batch_wo_cache_flash_attn_train * remove unnecessary Adam(W) optimizer tensors. reduces optimizer memory overhead from 7*modelsize to 2*modelsize. additionally allows to optimize models with more than 2^31 parameters by replacing int with int64_t. bumps training checkpoint file version, but old checkpoints can still be read. new version with less tensors is saved. * add gradient clipping to AdamW * Fix reset of unused g->nodes and g->grads to NULL * implement gradient checkpointing for training reduces memory overhead from O(n_layer) to O(sqrt(n_layer)) as explained in readme of https://github.com/cybertronai/gradient-checkpointing * remove unused compute buffer 3 * add and use function ggml_build_backward_expand to avoid stack overflows with large maximum number of nodes GGML_API void ggml_build_backward_expand(struct ggml_context * ctx, struct ggml_cgraph * gf, struct ggml_cgraph * gb, bool keep); * change AdamW decay parameter to work like the torch AdamW decay parameter It is now relative to Adam learning rate `alpha*sched`. Before that it was relative to `sched` only. `alpha` being the maximum learning rate and `sched` being a scaling parameter in [0..1] * change default AdamW weight decay parameter used in training to 0.1 as used in nanoGPT * change default AdamW weight decay parameter defined in ggml to 0.0, making Adam default instead of AdamW btw: the default weight decay parameter for torch.optim.AdamW is 0.01 * bug fixes for cross entropy loss ggml_cross_entropy_loss: sums where not correctly added in workload of each thread ggml_cross_entropy_loss_back: simplify backward process, reducing numerical issues guard usage of exp f16 lookup in cross entropy by #define GGML_CROSS_ENTROPY_EXP_FP16 cross entropy loss is only used once during training, but it is quite sensitive to numerical errors introduced by exp-f16-lookup. so exp-f16-lookup for cross entropy loss is disabled by default, trading better gradients for very slightly worse runtime performance. * fix test-grad0 for cross_entropy_loss the second argument to cross_entropy_loss must sum up to 1 for each row * fix test-grad0 for soft_max dont use only sum as aggregation, because sum of softmax is always 1 -> finite differences should not work instead use sum(log(soft_max()*(1-eps)+eps)); use eps to avoid log(0) * improve finite differences of test-grad0 by using double instead of float * change cross_entropy_loss to output average over all rows this helps keeping the loss and gradients in a sane range * improve gradient checkpointing sqrt(n_layers) is only the best checkpoint step when mem size of checkpoints and mem size of layers are equal. since layers require more memory than the single-tensor-checkpoint we use, the optimal values are compute different: ``` given: n, u, v objective: minimize(a*u+b*v) where a*b=n, a>0, b>0 b=n/a minimize(a*u+v*n/a) diff(a*u+v*n/a, a) = u - (v*n/a)/a diff(a*u+v*n/a, a) == 0 u - (v*n/a)/a == 0 u == v*n/(a*a) u*a*a = v*n a*a = v*n/u a = sqrt(n*v/u) ``` this change results in more checkpoints, requiring less layers to store between checkpoints, overall improving memory usage. * disable gradient checkpointing debug output * llama : fix rope usage in train-text-from-scratch after ChatGLM change * add more training parameters: --enable-restart N Only for Adam optimizer. Enable restarts of cos-decay --disable-restart N Only for Adam optimizer. Disable restarts of cos-decay --opt-past N Number of optimization iterations to track for delta convergence test. Disabled when zero. --opt-delta N Maximum delta for delta convergence test. Disabled when <= zero. --opt-max-no-improvement N Maximum number of optimization iterations with no improvement. Disabled when <= zero. --adam-epsf N AdamW epsilon for convergence test. Disabled when <= zero. --adam-min-alpha N Adam minimum learning rate alpha, usually 0.1 * alpha * replace memcpy with reshape operation so that the graph is not cut at the input this makes it possible to store other values into the input tensor and then simply recompute the graph without rebuilding it * remove unused function argument from get_example_targets_batch * measure and print total training time * add optimization callback to ggml_opt_resume_g this callback is called before each iteration with custom data and pointer to learning schedule parameter (only used in Adam(W)). can be used for dynamic learning schedule and setting input data for batches before each iteration * use optimization callback in training allows dynamic learning schedule and different batch data for each iteration without relying on low n_iter and high n_examples parameters reduces runtime by avoiding restart of optimization function and improves training convergence by providing a different batch for each iteration * add minimum number of tensor dimensions to apply weight decay (default 2) this allows to not apply weight decay to bias parameters * rename training parameter cos-decay-alpha to cos-decay-min and clarify that adam-min-alpha also applies to warmup * fix increase of model.train_samples and model.train_tokens now that each optimizer iteration gets its own batch we need to multiply by number of opt iterations * change sampling parameters for prediction after training to defaults of common.h and clarify what is context for prediction and what are generated tokens * tighten abs error bounds for cross_entropy_loss in test-grad0 * add conditional compilation of using F16 exp in flash attention uncomment `// #define GGML_FLASH_ATTN_EXP_FP16` to enable usage of f16 exp in flash attention * tighten abs error bounds for flash_attn in test-grad0 * tighten abs error bounds for sqrt in test-grad0 * remove out-commented vectorized code of opt_adam the vectorized code might be bit faster for low number of parameters, but it had a big memory usage overhead * ggml : update ggml_rms_norm_back with configurable eps * llama training : fix ggml_rms_norm_back calls to pass configurable eps * remove trailing whitespace * add train function using automatic gradient checkpointing backward pass and allocator * in train function replace add_inplace by regular add because using add_inplace seems to result in different gradients * don't use allocate hash_map on context because the context has no_alloc=True when using memory allocator resulting in NULL data pointers * correctly clone reshape and permute operations by also cloning tensor->nb values * fix variable name and add missing type cast * terminate recursive tensor cloning when reaching tensor without src tensors * correctly clone view tensors by setting data pointers without this the checkpointing would only work when being used together with memory allocator * fix variable names * swap arguments to commutative ops to be the same as in `forward_batch_wo_cache_flash_attn` * add input tensors as checkpoints so that recursive tensor cloning of gradient checkpointing terminates on input tensors * fix variable name and add missing boolean negation * make sure some tensors are not reallocated by inserting new temporary nodes depending on them: output and parameter gradient tensors need to be available at the end of the graph execution parameter gradient tensors also need to be available before the graph execution because they are set to zero before each optimizer iteration checkpoint tensors are allocated all together to reduce memory allocator fragmentation afterwards, in addition to the temporary nodes, we also need to reset the temporary leafs * fix ASSERT to work with zero layers * add training options whether to use allocator and/or unified training function * integrate unified training function which may use memory allocator the unified training function also supports arguments whether to use flash attention and/or gradient checkpointing * format name of cloned tensors with " (clone)" suffix * set names for tensors in unified train function for easier debugging * allocate graph on context using ggml_new_graph * remove handwritten training functions * remove unused training parameters "use_scratch" and "use_unified" * remove trailing whitespace * remove unused train params: mem_compute1_gb & mem_compute2_gb mem_compute_gb is used for compute when automatic memory allocator is not enabled, otherwise it can be very small to only hold the tensor definitions mem_compute0_gb is used for automatic memory allocator (as long as measurement of max required size is not implemented) * remove unused forward_batch function * add debug asserts in ggml_allocr_alloc to some common pitfalls when using this function directly * only use ggml_allocr_alloc when tensor has NULL data and is no view * fix test when to create temporary backward graph temporary backward graph is only necessary when using checkpointing * fix memory "leak" in optimizers each iteration a new cplan with new memory for work data was allocated. now cplan creation only happens at the start of optimization, with each iteration reusing the cplan and its work data. * reverse order of for loop in ggml_build_backward_expand to save memory when using gradient checkpointing and allocator with this loop order gradient checkpointing with allocator on 16 layer model saves 13% memory; 2 layer memory it saves 2% memory. the computation results are the same * add API functions to access llama model tensors * add stub example for finetuning, based on train-text-from-scratch * move and remove code * add API functions to access remaining model parameters: mult, head and rot * first draft for LORA finetune training * remove const model and layer arguments in API functions for accessing model tensors * bug fixes to make finetune compile automatic allocator does not work yet * add debug prints for training memory improvements * fix names of lora tensors * avoid stack overflow resulting from big ggml_cgraph replace stack allocation and ggml_build_forward by ggml_new_graph in combination with ggml_build_forward_expand * replace llama API functions to get model tensors by one function to get model tensor by name LLAMA_API struct ggml_tensor * llama_get_model_tensor(struct llama_model * model, const char * name); * remove unused call to not existing llama_get_layer_from_model * implement ggml_compute_forward_out_prod_q_f32 * remove trailing whitespace * add lora finetune support on quantized base model tensors * add ggml_add_cast API function this function works like ggml_add, but accepts a data type for the resulting tensor. only supported for quantized src0 input. * use ggml_add_cast in finetuning lora-applied weights will now have data type F32, which improves gradients when finetuning quantized base models * bug fix: actually use result type passed to ggml_add_cast * make sure base model tensors data cannot be used in viewable operations memory allocator would try to make lora application inplace on base model tensors. since those are memory mapped this will result in memory access violations * fix bug in ggml_out_prod which resulted in wrong n_dims of result tensors * avoid keeping in memory ALL of the gradients The problem here stems from ggml_graph_reset. This function is called in the optimization function, before each graph computation, to reset the gradients to zero. This required a unique memory slot for each gradient: allocating memory from a previosly freed memory location might lead to non-zero input gradients. During ggml_compute_backward the gradients are build stepwise by adding or substracting new values, starting from a OP_NONE tensor which needs to contain zero-values. This requires the graph reset. To avoid this I now remember in ggml_build_backward_expand the original OP_NONE gradient tensors in a hash table, which is passed to ggml_compute_backward. There instead of using add (or sub or similar) I test whether the existing gradient to be changed is a zero-valued-tensor by looking up its existence in the hash table. When it is such a zero-tensor it will not be modified, but replaced by the value to be added, otherwise the regular add (not inplace, allocator will take care of this) will be used. This way none of those zero-tensor values will be necessary in the final backward graph and more importantly they won't need a unique memory slot, just to make them zero. * remove trailing whitespace * remove debug prints and function to compute tensor data hash * improve optimization iteration prints * adjust maximal values to support finetuning 3B models * change default finetune params lora_r and lora_alpha to match the n_rank parameters of 4 * bug fix: make sure finetune input gradient is allocated at begin and kept until end * remove unnecessary src tensor from ggml_get_rows_back we don't need data of src[2] for computation, only to setup the correct output shape. remove dependency on src[2], so that allocator can work more freely. the computational graph is still completely determined, because the output shape is naturally included. this is similar to how ggml_reshape does it. * remove unnecessary src tensor from ggml_repeat & ggml_repeat_back we don't need data of src[1] for computation, only to setup the correct output shape. remove dependency on src[1], so that allocator can work more freely. the computational graph is still completely determined, because the output shape is naturally included * resolve todo allocator will only make it inplace when they are of the same type * mixing multiple LORA adapters is now possible pass more than one '--lora FNAME' argument to apply more than one LORA. use '--lora-scaled FNAME S' when you want to specify a user-defined scale for an adapter. * add option to save finetune output every N iterations * also save latest finetune output with ITERATION="LATEST" and print where files are saved saving with LATEST makes it easier to resume training from the latest checkpoint the string "LATEST" can be configured with command line option "--fn-latest STR" * update checkpoint train stats before saving via "--save-every" * add command line option `--rank-wo N` for rank of wo tensor * update finetune README * fix dump_non_result_info_yaml to output multiple lora adapters * bug fix: replace GGML_TYPE_SIZE[t] by ggml_type_size(t) * replace llama_n_mult by llama_n_ff * finetune bug fixes to compile with merged in code from master * remove prediction related code to reduce duplicated code with main use main instead * reduce large memory overhead in train-text-from-scratch all gradients had to be pinned so that graph_reset works correctly. this is no longer necessary with the changes to ggml_compute_backward introduced in this PR. * add comment explaining why finetune checkpoints are allocated in one block * make default value of float member a float literal * handle rms_norm and rope parameters the same as in train-text-from-scratch * remove unused code * remove vocab related code as it is unnecessary * add LLM_KV_TRAINING_TYPE to train-text-from-scratch checkpoints so that they can be differentiated from lora finetune checkpoints * add gguf constants and load/save functions from train-text-from-scratch * add load & save lora finetune checkpoints via gguf * add python script to convert old finetune checkpoint files to gguf * remove old checkpoint save & load code * remove code to print data checksums which was used to verify correctness of new gguf code * omit tokenization when training is disabled, only save llama lora adapter training can be disabled by passing '-n 0' to finetune * remove trailing whitespace * update README.md * implement ggml_compute_forward_repeat_f16 * avoid stack overflow of large cgraphs in test-grad0 * add ggml API functions ggml_unravel_index, ggml_get_i32_nd and its analogs for set and for f32 ggml_get_i32_1d, ggml_set_i32_1d, ggml_get_f32_1d, ggml_set_f32_1d now support non-contiguous tensors. in case of non-contiguous tensor, the 1d index is unraveled into a multi index using ggml_unravel_index to be passed to '_nd' function equivalent. this fixes a bug in test-grad0 which happens due to ggml_build_backward not building purely contiguous tensors anymore * increase test-grad0 context mem size to accommodate for bigger cgraph * add sanity check to ggml_compute_backward, asserting the correct shape of gradients * fix ggml_acc_or_set to return tensor of correct shape * remove unused 'inplace' argument from ggml_compute_backward function inplace operations to add gradients are no longer created by ggml_compute_backward use allocator to automatically make inplace operations * add missing argument 'int i0' to ggml_get_i32_nd & ggml_set_i32_nd header declarations * fix error message in ggml_allocr_alloc to display actual max_avail * fix check_gradient ggml_build_backward_expand was previously replaced by ggml_build_backward, but the assignment of forward graph to backward graph missing * use tensor->view_src instead of ggml_is_view and get_view_source * move gradient checkpointing code into ggml, new API function: // build gradient checkpointing backward graph gb for gf using provided checkpoints // gb_tmp will contain original backward graph with rewritten backward process nodes, // but without the second forward pass nodes. GGML_API void ggml_build_backward_gradient_checkpointing( struct ggml_context * ctx, struct ggml_cgraph * gf, struct ggml_cgraph * gb, struct ggml_cgraph * gb_tmp, struct ggml_tensor * * checkpoints, int n_checkpoints); * replace custom data getters and setters by ggml functions * train-text-from-scratch can train (full finetune) gguf models just pass the gguf model via `--checkpoint-in FN`. after this, to continue training, pass the generated checkpoint instead of the original gguf model. tested with smaller models, bigger models may exceed available memory. use (LORA) finetune for those. * remove trailing whitespace * add option to save train-text-from-scratch output every N iterations * update README.md * fix warnings * fix warnings * remove finetune option to disable allocator the allocator should always be used. by making sure that it is always used it gets easier to implement automatic memory requirements computation * add tensor checkpoints only when gradient checkpointing is enabled * initialize opt ggml context if none was provided * add ggml-alloc API function 'ggml_allocr_max_size' to get max size of alloc GGML_API size_t ggml_allocr_max_size(struct ggml_allocr * alloc); * finetune: automatically allocate all memory and changes to command line options remove '--n_examples N' parameter, as it no longer makes sense to call optimization process multiple times in a loop. add '--only_write_lora' command line option: will skip tokenization and training, to only write a llama.cpp comptabile LORA adapter. remove memory buffer related command line options. improve iteration console output. * add finetune to Makefile * update README.md * print time per iteration and estimate remaining time * increase measured alloc size by tensor_alignment ggml_allocr_reset will reduce the given size by up to tensor_alignment-1 * fix README.md * add some more allocator debug prints * bug fix, probably solves the 'ggml_allocr_alloc: not enough space in the buffer' issue * revert last commit "bug fix, probably solves the 'ggml_allocr_alloc: not enough space in the buffer' issue" "alloc was freeing an externally allocated tensor, because it calculated the end of allocator memory as alloc->data + alloc->max_size instead of alloc->data + alloc->size." This is intentional to reduce the risk of freeing external tensors when measuring. Unless max_size is not properly calculated, I don't see why this is an issue. * remove unnecessary "0x" before "%p" output * move measurement memory segment to upper region of the address space * update README.md * fix printf format warnings * add missing gguf_free in load_checkpoint_lora_file * load default rms_norm and rope parameters from base model * add gradient accumulation specify number accumulation steps with '--grad-acc N'. this will simulate a bigger batch size of grad_acc*batch. * fix tracking of train_samples and train_tokens * build : fix compile warnings * ggml : fix L-BFGS linesearch loop * improve finetune time measurement fix printf warnings on system where int64_t is (long int). change time datatypes to double because values get big with long training times. exclude file saving from time measurement. converge faster to actual time per iteration by removing very small first duration before first iteration was performed. fix bug in output of total training time, the reported value was 1000 times to small. * specify default lora rank with '--lora-r N' '--lora-r N' will specify default rank for all tensors '--rank-wq N', etc. will override this default rank for specific tensor types. * fix gradient accumulation bug where the same batch was used for each microstep * fix gradient accumulation bug where the same batch was used for each microstep * support grouped-query-attention in ggml_flash_attn and ggml_flash_attn_back k and v can now be repeated in q along ne[2] in forward pass just use modulo to compute k and v indices, like ik2 = iq2 % nek2. in backard pass this won't work as easy, because multiple threads will compete to accumulate to the same k->grad[:,ik1,ik2,ik3] and v->grad[:,iv1,iv2,iv3]. so we change the parallelization over q rows to be over k rows. this ensures non-overlapping (ik2,ik3) across threads. in each thread we then iterate over the number of repetitions of k/v in q to compute iq2 as iq2 = ik2 + irep*nek2. since ne2 is not the same for q,k and v we also change how the gradients are concatenated into the result tensor. additionally the offsets of gradq, gradk and gradv in the result tensor are now memory aligned. we also simplify the compute_backward part of flash_attn to use ggml_reshape instead of switching over the number of dimensions. this needs a small change to ggml_reshape, removing the assertion of second argument to be contiguous. since only the shape (ne) of the second reshape argument is of relevance, its memory layout (nb) is irrelevant -> it can very well be non-contiguous. change test-grad0 to also test for repeated k/v in q. this changes the rng and now results in small gradient differences in softmax. these solely come from using f16 exp table lookup in forward softmax: when temporarily changing softmax to use actual exp function, the reported gradient differences go away. gradient differences coming solely from f16 table lookup are acceptable. added a note to explain this. * add llama API functions to get grouped-query-attention n_head parameter 'n_head_kv'. * fix finetune to support grouped-query-attention (using flash-attention) note: ggml changes to ggml_out_prod are necessary to support grouped-query-attention without flash-attention. * support broadcastable a in out_prod(a, b) and backward pass of broadcasting mul_mat(a, b) * test broadcasting mul_mat backward pass * decouple random number generator of each operation test when changing one test the rng of others tests is not influenced anymore * add comment briefly describing what ggml_repeat_back does * simplify broadcasting mul_mat backward using ggml_repeat_back * add cgraph evaluation order member and corresponding enum type this controls in which order ggml_build_forward visits source nodes. by default the nodes are visited left to right, i.e. src[0] first. in some cases it is beneficial for ggml-alloc to visit in a different order. two possible orders are supported: left-to-right (src[0] first) and right-to-left (src[0] last). * measure max compute size for each cgraph eval order and use best order this can bring huge memory savings: e.g. codellama-34b with n_ctx=64, n_batch=1 goes from 92927.8mb down to 4627.6 MB * remove unused command line options * add sample start patterns and options to force new or by default resume last shuffling * update shuffle rng state on reshuffle * exclude known zero values from computations in flash_attn_f32 & flash_attn_back_f32 * remove probably unnecessary exception type flags from stringstream * pass correct max number of tokens to llama_tokenize * account for possible leading whitespace that will be added by tokenizer e.g. '\t' will be tokenized by llama spm tokenizer to [29871, 12] * use unrolled vec_mad in out_prod y is vec_mad result vec. x is vec_mad input vec. v is vec_mad input scalar. ggml_vec_mad_f32_unroll will internally loop over x and v with same y. GGML_VEC_MAD_UNROLL is by default defined to 32. This value is empirical optimized using performance test runs of out-prod in openllama-3b finetune with 256 context length and batch size 1. It gives 23% performance boost for out_prod. Full measurements of out-prod runtime in ms: unroll_xv unroll_yv 1 67014.643 87826.469 2 77117.552 89077.656 4 72091.311 109121.657 8 61077.543 88678.334 16 56914.67 79514.947 24 59024.595 84350.254 28 55952.446 83368.73 32 51476.658 85177.745 36 55973.792 84659.92 40 55139.616 93844.738 48 60736.392 93330.267 64 99856.878 116994.99 Second column is when unrollying yv instead of xv * set lora_alpha to value of lora_r if it is not set via command line otherwise only changing lora_r will change scaling of lora adapter used in prediction * reshuffle original sample order instead of the previous shuffled order otherwise resumed reshuffle will not result in same sample order * block tiling for out-prod inspired by mul-mat block sizes are empirically optimized roughly doubles the flops of out-prod * exclude some more known zero values from computations in flash_attn_f32 & flash_attn_back_f32 * add static keywords * remove outcommented old code * update train-text-from-scratch with tokenization, sample selection and shuffling from finetune * remove lbfgs related train parameters * move common train functions into common/train.[h|cpp] * move train state into struct train_state * move train data saving code into callback to unify code of opt_callback train_params are still different in finetune and train-text-from-scratch, so it can't yet be moved to train.h|cpp * move common train params into common/train * move common opt_callback into common/train * fix consume_common_train_arg * save and load head_count_kv in lora checkpoints * increase train_samples by used_samples instead of number of batches on batch can contain more than one sample when option "fill_with_next_samples" is used * fix usage of llama_tokenize * remove static from process_escape since we need it exposed in header * fix code formating of long function declarations * fix condition in load_train_state_gguf * use die("msg") instead of replace GGML_ASSERT(!"msg") or throw std::runtime_error("msg") * fix saving and loading of training type * remove terminating '\0' from tokenization (llama_tokenize is now passed the string length instead of relying on terminating '\0') * fix compile warnings * fix compile warnings * use new/delete for train_state instead of malloc/free using malloc may result in seg faults when trying to assign string fields * assert that sample_count > 0, avoiding division by zero * fix frand to return value in interval [0,1) * add train option "--sample-random-offsets" Use samples beginning at random offsets. The offset is only applied to the first sample in each batch context window. Together with "--fill-with-next-samples" this may help for training endless text generation. For example given a dataset containing samples "abcd", "ABCD", "0123". With context size of 8 and options "--fill-with-next-samples", "--no-separate-with-eos", "--no-separate-with-bos", the context windows of batches could only be filled with "abcdABCD", "ABCDabcd", "0123abcd", etc. With "--sample-random-offsets" it can also be filled with "23abcdAB", "bcd0123A", etc. * deduplicate code into function * remove n_rot hparam, as it must always be hparam.n_embd_head() * align code * assert correct base model tensor shapes * move some params from lora hparams into model hparams and load model params from gguf this equalizes the model definition in finetune and text-from-scratch and removes the need for additional llama api functions to get model parameters * remove now unnecessary llama API functions to get model params that where added by this PR * train-text-from-scratch: automatically allocate model tensors, remove option '--mem-model N' * train-text-from-scratch: automatically allocate opt context * train-text-from-scratch: automatically allocate input tensors * train-text-from-scratch: automatically allocate compute memory * remove unused options and equalize train-text-from-scratch with finetune * initialize opt->loss_after with zero * add export-lora program * remove trailing whitespace * add export-lora build in Makefile * remove unused struct tensor_info from export-lora * add export-lora build dependency to llama because it depends on common, which depends on llama * update finetune README.md * cancel optimization when specified number of epochs is completed * improve handling of export-lora arguments print errors and warnings when files could not be read or created * Fix export-lora.cpp "not enough space in the context's memory pool" (#1) * Fix export-lora.cpp "not enough space in the context's memory pool" Without this patch, export-lora would sometimes error with "not enough space in the context's memory pool (needed 656784, available 656800)". * increase required context size by 5*GGML_MEM_ALIGN instead of plain 16 --------- Co-authored-by: xaedes <xaedes@gmail.com> * improve handling of not yet supported tensor types --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: meatbag-18a <145869052+meatbag-18a@users.noreply.github.com>
2023-09-28 18:40:11 +00:00
retrieval: examples/retrieval/retrieval.cpp ggml.o llama.o $(COMMON_DEPS) $(OBJS)
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
speculative: examples/speculative/speculative.cpp ggml.o llama.o $(COMMON_DEPS) grammar-parser.o $(OBJS)
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
2023-09-04 10:39:57 +00:00
parallel: examples/parallel/parallel.cpp ggml.o llama.o $(COMMON_DEPS) $(OBJS)
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
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 16:04:36 +00:00
lookahead: examples/lookahead/lookahead.cpp ggml.o llama.o $(COMMON_DEPS) $(OBJS)
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
lookup: examples/lookup/lookup.cpp ggml.o llama.o ngram-cache.o $(COMMON_DEPS) $(OBJS)
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
$(CXX) $(CXXFLAGS) -c examples/lookup/lookup-create.cpp -o $(call GET_OBJ_FILE, examples/lookup/lookup-create.cpp)
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, examples/lookup/lookup-create.cpp) -o lookup-create $(LDFLAGS)
$(CXX) $(CXXFLAGS) -c examples/lookup/lookup-merge.cpp -o $(call GET_OBJ_FILE, examples/lookup/lookup-merge.cpp)
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, examples/lookup/lookup-merge.cpp) -o lookup-merge $(LDFLAGS)
$(CXX) $(CXXFLAGS) -c examples/lookup/lookup-stats.cpp -o $(call GET_OBJ_FILE, examples/lookup/lookup-stats.cpp)
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, examples/lookup/lookup-stats.cpp) -o lookup-stats $(LDFLAGS)
passkey: examples/passkey/passkey.cpp ggml.o llama.o $(COMMON_DEPS) $(OBJS)
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
gbnf-validator: examples/gbnf-validator/gbnf-validator.cpp ggml.o llama.o $(COMMON_DEPS) grammar-parser.o $(OBJS)
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
ifeq ($(UNAME_S),Darwin)
swift: examples/batched.swift
(cd examples/batched.swift; make build)
endif
common/build-info.cpp: $(wildcard .git/index) scripts/build-info.sh
@sh scripts/build-info.sh "$(CC)" > $@.tmp
@if ! cmp -s $@.tmp $@; then \
mv $@.tmp $@; \
else \
rm $@.tmp; \
fi
2023-04-13 14:03:57 +00:00
build-info.o: common/build-info.cpp
$(CXX) $(CXXFLAGS) -c $(filter-out %.h,$^) -o $@
2023-03-10 18:40:58 +00:00
#
# Tests
#
tests: $(TEST_TARGETS)
benchmark-matmult: examples/benchmark/benchmark-matmult.cpp build-info.o ggml.o $(OBJS)
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
run-benchmark-matmult: benchmark-matmult
./$@
2023-04-13 14:03:57 +00:00
.PHONY: run-benchmark-matmult swift
vdot: pocs/vdot/vdot.cpp ggml.o $(OBJS)
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) $(filter-out $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
q8dot: pocs/vdot/q8dot.cpp ggml.o $(OBJS)
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) $(filter-out $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
tests/test-llama-grammar: tests/test-llama-grammar.cpp ggml.o grammar-parser.o $(OBJS)
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
tests/test-grammar-parser: tests/test-grammar-parser.cpp ggml.o llama.o grammar-parser.o $(OBJS)
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
tests/test-grammar-integration: tests/test-grammar-integration.cpp ggml.o llama.o grammar-parser.o $(OBJS)
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
tests/test-double-float: tests/test-double-float.cpp ggml.o $(OBJS)
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
json-schema-to-grammar improvements (+ added to server) (#5978) * json: fix arrays (disallow `[,1]`) * json: support tuple types (`[number, string]`) * json: support additionalProperties (`{[k: string]: [string,number][]}`) * json: support required / optional properties * json: add support for pattern * json: resolve $ref (and support https schema urls) * json: fix $ref resolution * join: support union types (mostly for nullable types I think) * json: support allOf + nested anyOf * json: support any (`{}` or `{type: object}`) * json: fix merge * json: temp fix for escapes * json: spaces in output and unrestricted output spaces * json: add typings * json:fix typo * Create ts-type-to-grammar.sh * json: fix _format_literal (json.dumps already escapes quotes) * json: merge lit sequences and handle negatives {"type": "string", "pattern": "^({\"question\": \"[^\"]+\", \"response\": \"[^\"]+\"}\\n)+$"} * json: handle pattern repetitions * Update json-schema-to-grammar.mjs * Create regex-to-grammar.py * json: extract repeated regexp patterns to subrule * Update json-schema-to-grammar.py * Update json-schema-to-grammar.py * Update json-schema-to-grammar.py * json: handle schema from pydantic Optional fields * Update json-schema-to-grammar.py * Update json-schema-to-grammar.py * Update ts-type-to-grammar.sh * Update ts-type-to-grammar.sh * json: simplify nullable fields handling * json: accept duplicate identical rules * json: revert space to 1 at most * json: reuse regexp pattern subrules * json: handle uuid string format * json: fix literal escapes * json: add --allow-fetch * json: simplify range escapes * json: support negative ranges in patterns * Delete commit.txt * json: custom regex parser, adds dot support & JS-portable * json: rm trailing spaces * Update json-schema-to-grammar.mjs * json: updated server & chat `( cd examples/server && ./deps.sh )` * json: port fixes from mjs to python * Update ts-type-to-grammar.sh * json: support prefixItems alongside array items * json: add date format + fix uuid * json: add date, time, date-time formats * json: preserve order of props from TS defs * json: port schema converter to C++, wire in ./server * json: nits * Update json-schema-to-grammar.cpp * Update json-schema-to-grammar.cpp * Update json-schema-to-grammar.cpp * json: fix mjs implementation + align outputs * Update json-schema-to-grammar.mjs.hpp * json: test C++, JS & Python versions * json: nits + regen deps * json: cleanup test * json: revert from c++17 to 11 * json: nit fixes * json: dirty include for test * json: fix zig build * json: pass static command to std::system in tests (fixed temp files) * json: fix top-level $refs * json: don't use c++20 designated initializers * nit * json: basic support for reserved names `{number:{number:{root:number}}}` * Revamp test cmake to allow args (WORKING_DIRECTORY needed for JSON test) * json: re-ran server deps.sh * json: simplify test * json: support mix of additional props & required/optional * json: add tests for some expected failures * json: fix type=const in c++, add failure expectations for non-str const&enum * json: test (& simplify output of) empty schema * json: check parsing in test + fix value & string refs * json: add server tests for OAI JSON response_format * json: test/fix top-level anyOf * json: improve grammar parsing failures * json: test/fix additional props corner cases * json: fix string patterns (was missing quotes) * json: ws nit * json: fix json handling in server when there's no response_format * json: catch schema conversion errors in server * json: don't complain about unknown format type in server if unset * json: cleaner build of test * json: create examples/json-schema-pydantic-example.py * json: fix date pattern * json: move json.hpp & json-schema-to-grammar.{cpp,h} to common * json: indent 4 spaces * json: fix naming of top-level c++ function (+ drop unused one) * json: avoid using namespace std * json: fix zig build * Update server.feature * json: iostream -> fprintf * json: space before & refs for consistency * json: nits
2024-03-21 11:50:43 +00:00
tests/test-json-schema-to-grammar: tests/test-json-schema-to-grammar.cpp json-schema-to-grammar.o ggml.o llama.o grammar-parser.o $(OBJS)
$(CXX) $(CXXFLAGS) -Iexamples/server -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
tests/test-grad0: tests/test-grad0.cpp ggml.o $(OBJS)
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
tests/test-opt: tests/test-opt.cpp ggml.o $(OBJS)
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
tests/test-quantize-fns: tests/test-quantize-fns.cpp ggml.o $(OBJS)
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
tests/test-quantize-perf: tests/test-quantize-perf.cpp ggml.o $(OBJS)
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
tests/test-sampling: tests/test-sampling.cpp ggml.o llama.o $(OBJS)
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
llama : fix BPE pre-tokenization (#6920) * merged the changes from deepseeker models to main branch * Moved regex patterns to unicode.cpp and updated unicode.h * Moved header files * Resolved issues * added and refactored unicode_regex_split and related functions * Updated/merged the deepseek coder pr * Refactored code * Adding unicode regex mappings * Adding unicode regex function * Added needed functionality, testing remains * Fixed issues * Fixed issue with gpt2 regex custom preprocessor * unicode : fix? unicode_wstring_to_utf8 * lint : fix whitespaces * tests : add tokenizer tests for numbers * unicode : remove redundant headers * tests : remove and rename tokenizer test scripts * tests : add sample usage * gguf-py : reader prints warnings on duplicate keys * llama : towards llama3 tokenization support (wip) * unicode : shot in the dark to fix tests on Windows * unicode : first try custom implementations * convert : add "tokenizer.ggml.pre" GGUF KV (wip) * llama : use new pre-tokenizer type * convert : fix pre-tokenizer type writing * lint : fix * make : add test-tokenizer-0-llama-v3 * wip * models : add llama v3 vocab file * llama : adapt punctuation regex + add llama 3 regex * minor * unicode : set bomb * unicode : set bomb * unicode : always use std::wregex * unicode : support \p{N}, \p{L} and \p{P} natively * unicode : try fix windows * unicode : category support via std::regex * unicode : clean-up * unicode : simplify * convert : add convert-hf-to-gguf-update.py ggml-ci * lint : update * convert : add falcon ggml-ci * unicode : normalize signatures * lint : fix * lint : fix * convert : remove unused functions * convert : add comments * convert : exercise contractions ggml-ci * lint : fix * cmake : refactor test targets * tests : refactor vocab tests ggml-ci * tests : add more vocabs and tests ggml-ci * unicode : cleanup * scripts : ignore new update script in check-requirements.sh * models : add phi-3, mpt, gpt-2, starcoder * tests : disable obsolete ggml-ci * tests : use faster bpe test ggml-ci * llama : more prominent warning for old BPE models * tests : disable test-tokenizer-1-bpe due to slowness ggml-ci --------- Co-authored-by: Jaggzh <jaggz.h@gmail.com> Co-authored-by: Kazim Abrar Mahi <kazimabrarmahi135@gmail.com>
2024-04-29 13:58:41 +00:00
tests/test-tokenizer-0: tests/test-tokenizer-0.cpp ggml.o llama.o $(COMMON_DEPS) console.o $(OBJS)
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
tests/test-tokenizer-1-bpe: tests/test-tokenizer-1-bpe.cpp ggml.o llama.o $(COMMON_DEPS) console.o $(OBJS)
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
llama : fix BPE pre-tokenization (#6920) * merged the changes from deepseeker models to main branch * Moved regex patterns to unicode.cpp and updated unicode.h * Moved header files * Resolved issues * added and refactored unicode_regex_split and related functions * Updated/merged the deepseek coder pr * Refactored code * Adding unicode regex mappings * Adding unicode regex function * Added needed functionality, testing remains * Fixed issues * Fixed issue with gpt2 regex custom preprocessor * unicode : fix? unicode_wstring_to_utf8 * lint : fix whitespaces * tests : add tokenizer tests for numbers * unicode : remove redundant headers * tests : remove and rename tokenizer test scripts * tests : add sample usage * gguf-py : reader prints warnings on duplicate keys * llama : towards llama3 tokenization support (wip) * unicode : shot in the dark to fix tests on Windows * unicode : first try custom implementations * convert : add "tokenizer.ggml.pre" GGUF KV (wip) * llama : use new pre-tokenizer type * convert : fix pre-tokenizer type writing * lint : fix * make : add test-tokenizer-0-llama-v3 * wip * models : add llama v3 vocab file * llama : adapt punctuation regex + add llama 3 regex * minor * unicode : set bomb * unicode : set bomb * unicode : always use std::wregex * unicode : support \p{N}, \p{L} and \p{P} natively * unicode : try fix windows * unicode : category support via std::regex * unicode : clean-up * unicode : simplify * convert : add convert-hf-to-gguf-update.py ggml-ci * lint : update * convert : add falcon ggml-ci * unicode : normalize signatures * lint : fix * lint : fix * convert : remove unused functions * convert : add comments * convert : exercise contractions ggml-ci * lint : fix * cmake : refactor test targets * tests : refactor vocab tests ggml-ci * tests : add more vocabs and tests ggml-ci * unicode : cleanup * scripts : ignore new update script in check-requirements.sh * models : add phi-3, mpt, gpt-2, starcoder * tests : disable obsolete ggml-ci * tests : use faster bpe test ggml-ci * llama : more prominent warning for old BPE models * tests : disable test-tokenizer-1-bpe due to slowness ggml-ci --------- Co-authored-by: Jaggzh <jaggz.h@gmail.com> Co-authored-by: Kazim Abrar Mahi <kazimabrarmahi135@gmail.com>
2024-04-29 13:58:41 +00:00
tests/test-tokenizer-1-spm: tests/test-tokenizer-1-spm.cpp ggml.o llama.o $(COMMON_DEPS) console.o $(OBJS)
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
tests/test-rope: tests/test-rope.cpp ggml.o $(OBJS)
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
tests/test-c.o: tests/test-c.c llama.h
$(CC) $(CFLAGS) -c $(filter-out %.h,$^) -o $@
tests/test-backend-ops: tests/test-backend-ops.cpp ggml.o $(OBJS)
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
tests/test-model-load-cancel: tests/test-model-load-cancel.cpp ggml.o llama.o tests/get-model.cpp $(COMMON_DEPS) $(OBJS)
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
tests/test-autorelease: tests/test-autorelease.cpp ggml.o llama.o tests/get-model.cpp $(COMMON_DEPS) $(OBJS)
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
tests/test-chat-template: tests/test-chat-template.cpp ggml.o llama.o $(COMMON_DEPS) $(OBJS)
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)