Skip to content

Commit e7b2e5a

Browse files
ggerganovarthw
ggerganov
authored and
arthw
committed
common : reimplement logging (ggml-org#9418)
ggml-org#9418
1 parent c1fd2b7 commit e7b2e5a

Some content is hidden

Large Commits have some content hidden by default. Use the searchbox below for content that may be hidden.

54 files changed

+2074
-2401
lines changed

‎.github/workflows/build.yml

Lines changed: 3 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -23,6 +23,9 @@ env:
2323
BRANCH_NAME: ${{ github.head_ref || github.ref_name }}
2424
GGML_NLOOP: 3
2525
GGML_N_THREADS: 1
26+
LLAMA_LOG_COLORS: 1
27+
LLAMA_LOG_PREFIX: 1
28+
LLAMA_LOG_TIMESTAMPS: 1
2629

2730
jobs:
2831
macOS-latest-cmake-arm64:

‎.github/workflows/server.yml

Lines changed: 6 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -20,6 +20,12 @@ on:
2020
types: [opened, synchronize, reopened]
2121
paths: ['.github/workflows/server.yml', '**/CMakeLists.txt', '**/Makefile', '**/*.h', '**/*.hpp', '**/*.c', '**/*.cpp', '**/*.cu', '**/*.swift', '**/*.m', 'examples/server/**.*']
2222

23+
env:
24+
LLAMA_LOG_COLORS: 1
25+
LLAMA_LOG_PREFIX: 1
26+
LLAMA_LOG_TIMESTAMPS: 1
27+
LLAMA_LOG_VERBOSITY: 10
28+
2329
concurrency:
2430
group: ${{ github.workflow }}-${{ github.ref }}-${{ github.head_ref || github.run_id }}
2531
cancel-in-progress: true

‎Makefile

Lines changed: 29 additions & 9 deletions
Original file line numberDiff line numberDiff line change
@@ -54,6 +54,7 @@ TEST_TARGETS = \
5454
tests/test-grammar-parser \
5555
tests/test-json-schema-to-grammar \
5656
tests/test-llama-grammar \
57+
tests/test-log \
5758
tests/test-model-load-cancel \
5859
tests/test-opt \
5960
tests/test-quantize-fns \
@@ -148,6 +149,14 @@ GGML_NO_METAL := 1
148149
DEPRECATE_WARNING := 1
149150
endif
150151

152+
ifdef LLAMA_DISABLE_LOGS
153+
REMOVE_WARNING := 1
154+
endif
155+
156+
ifdef LLAMA_SERVER_VERBOSE
157+
REMOVE_WARNING := 1
158+
endif
159+
151160
ifndef UNAME_S
152161
UNAME_S := $(shell uname -s)
153162
endif
@@ -351,19 +360,11 @@ ifdef LLAMA_SANITIZE_UNDEFINED
351360
MK_LDFLAGS += -fsanitize=undefined -g
352361
endif
353362

354-
ifdef LLAMA_SERVER_VERBOSE
355-
MK_CPPFLAGS += -DSERVER_VERBOSE=$(LLAMA_SERVER_VERBOSE)
356-
endif
357-
358363
ifdef LLAMA_SERVER_SSL
359364
MK_CPPFLAGS += -DCPPHTTPLIB_OPENSSL_SUPPORT
360365
MK_LDFLAGS += -lssl -lcrypto
361366
endif
362367

363-
ifdef LLAMA_DISABLE_LOGS
364-
MK_CPPFLAGS += -DLOG_DISABLE_LOGS
365-
endif # LLAMA_DISABLE_LOGS
366-
367368
# warnings
368369
WARN_FLAGS = \
369370
-Wall \
@@ -930,6 +931,7 @@ OBJ_LLAMA = \
930931
OBJ_COMMON = \
931932
common/common.o \
932933
common/arg.o \
934+
common/log.o \
933935
common/console.o \
934936
common/ngram-cache.o \
935937
common/sampling.o \
@@ -1026,6 +1028,14 @@ $(info - LLAMA_NO_CCACHE)
10261028
$(info )
10271029
endif
10281030

1031+
ifdef REMOVE_WARNING
1032+
$(info !!! REMOVAL WARNING !!!)
1033+
$(info The following LLAMA_ options have been removed and are no longer supported)
1034+
$(info - LLAMA_DISABLE_LOGS (https://github.com/ggerganov/llama.cpp/pull/9418))
1035+
$(info - LLAMA_SERVER_VERBOSE (https://github.com/ggerganov/llama.cpp/pull/9418))
1036+
$(info )
1037+
endif
1038+
10291039
#
10301040
# Build libraries
10311041
#
@@ -1167,6 +1177,11 @@ common/arg.o: \
11671177
common/arg.h
11681178
$(CXX) $(CXXFLAGS) -c $< -o $@
11691179

1180+
common/log.o: \
1181+
common/log.cpp \
1182+
common/log.h
1183+
$(CXX) $(CXXFLAGS) -c $< -o $@
1184+
11701185
common/sampling.o: \
11711186
common/sampling.cpp \
11721187
common/sampling.h \
@@ -1345,7 +1360,7 @@ llama-cvector-generator: examples/cvector-generator/cvector-generator.cpp \
13451360
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
13461361

13471362
llama-convert-llama2c-to-ggml: examples/convert-llama2c-to-ggml/convert-llama2c-to-ggml.cpp \
1348-
$(OBJ_GGML) $(OBJ_LLAMA)
1363+
$(OBJ_ALL)
13491364
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
13501365
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
13511366

@@ -1527,6 +1542,11 @@ tests/test-llama-grammar: tests/test-llama-grammar.cpp \
15271542
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
15281543
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
15291544

1545+
tests/test-log: tests/test-log.cpp \
1546+
$(OBJ_ALL)
1547+
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
1548+
$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
1549+
15301550
tests/test-grammar-parser: tests/test-grammar-parser.cpp \
15311551
$(OBJ_ALL)
15321552
$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)

‎ci/run.sh

Lines changed: 3 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -737,6 +737,9 @@ function gg_sum_embd_bge_small {
737737

738738
## main
739739

740+
export LLAMA_LOG_PREFIX=1
741+
export LLAMA_LOG_TIMESTAMPS=1
742+
740743
if [ -z ${GG_BUILD_LOW_PERF} ]; then
741744
# Create symlink: ./llama.cpp/models-mnt -> $MNT/models/models-mnt
742745
rm -rf ${SRC}/models-mnt

‎common/CMakeLists.txt

Lines changed: 12 additions & 10 deletions
Original file line numberDiff line numberDiff line change
@@ -51,21 +51,23 @@ endif()
5151
set(TARGET common)
5252

5353
add_library(${TARGET} STATIC
54+
arg.cpp
55+
arg.h
5456
base64.hpp
55-
common.h
5657
common.cpp
57-
arg.h
58-
arg.cpp
59-
sampling.h
60-
sampling.cpp
61-
console.h
58+
common.h
6259
console.cpp
63-
json.hpp
60+
console.h
6461
json-schema-to-grammar.cpp
65-
train.h
66-
train.cpp
67-
ngram-cache.h
62+
json.hpp
63+
log.cpp
64+
log.h
6865
ngram-cache.cpp
66+
ngram-cache.h
67+
sampling.cpp
68+
sampling.h
69+
train.cpp
70+
train.h
6971
)
7072

7173
if (BUILD_SHARED_LIBS)

‎common/arg.cpp

Lines changed: 51 additions & 58 deletions
Original file line numberDiff line numberDiff line change
@@ -1,15 +1,17 @@
11
#include "arg.h"
22

3+
#include "log.h"
34
#include "sampling.h"
45

56
#include <algorithm>
6-
#include <string>
7-
#include <vector>
8-
#include <set>
7+
#include <climits>
8+
#include <cstdarg>
99
#include <fstream>
1010
#include <regex>
11-
#include <cstdarg>
12-
#include <climits>
11+
#include <set>
12+
#include <string>
13+
#include <thread>
14+
#include <vector>
1315

1416
#include "json-schema-to-grammar.h"
1517

@@ -383,20 +385,6 @@ gpt_params_context gpt_params_parser_init(gpt_params & params, llama_example ex,
383385
exit(0);
384386
}
385387
));
386-
add_opt(llama_arg(
387-
{"-v", "--verbose"},
388-
"print verbose information",
389-
[](gpt_params & params) {
390-
params.verbosity = 1;
391-
}
392-
));
393-
add_opt(llama_arg(
394-
{"--verbosity"}, "N",
395-
format("set specific verbosity level (default: %d)", params.verbosity),
396-
[](gpt_params & params, int value) {
397-
params.verbosity = value;
398-
}
399-
));
400388
add_opt(llama_arg(
401389
{"--verbose-prompt"},
402390
format("print a verbose prompt before generation (default: %s)", params.verbose_prompt ? "true" : "false"),
@@ -417,7 +405,7 @@ gpt_params_context gpt_params_parser_init(gpt_params & params, llama_example ex,
417405
[](gpt_params & params) {
418406
params.use_color = true;
419407
}
420-
).set_examples({LLAMA_EXAMPLE_MAIN, LLAMA_EXAMPLE_INFILL}));
408+
).set_examples({LLAMA_EXAMPLE_MAIN, LLAMA_EXAMPLE_INFILL, LLAMA_EXAMPLE_SPECULATIVE, LLAMA_EXAMPLE_LOOKUP}));
421409
add_opt(llama_arg(
422410
{"-t", "--threads"}, "N",
423411
format("number of threads to use during generation (default: %d)", params.cpuparams.n_threads),
@@ -876,15 +864,15 @@ gpt_params_context gpt_params_parser_init(gpt_params & params, llama_example ex,
876864
params.input_prefix = value;
877865
params.enable_chat_template = false;
878866
}
879-
).set_examples({LLAMA_EXAMPLE_MAIN}));
867+
).set_examples({LLAMA_EXAMPLE_MAIN, LLAMA_EXAMPLE_INFILL}));
880868
add_opt(llama_arg(
881869
{"--in-suffix"}, "STRING",
882870
"string to suffix after user inputs with (default: empty)",
883871
[](gpt_params & params, const std::string & value) {
884872
params.input_suffix = value;
885873
params.enable_chat_template = false;
886874
}
887-
).set_examples({LLAMA_EXAMPLE_MAIN}));
875+
).set_examples({LLAMA_EXAMPLE_MAIN, LLAMA_EXAMPLE_INFILL}));
888876
add_opt(llama_arg(
889877
{"--no-warmup"},
890878
"skip warming up the model with an empty run",
@@ -1824,19 +1812,6 @@ gpt_params_context gpt_params_parser_init(gpt_params & params, llama_example ex,
18241812
params.system_prompt = system_prompt;
18251813
}
18261814
).set_examples({LLAMA_EXAMPLE_SERVER}));
1827-
add_opt(llama_arg(
1828-
{"--log-format"}, "{text, json}",
1829-
"log output format: json or text (default: json)",
1830-
[](gpt_params & params, const std::string & value) {
1831-
if (value == "json") {
1832-
params.log_json = true;
1833-
} else if (value == "text") {
1834-
params.log_json = false;
1835-
} else {
1836-
throw std::invalid_argument("invalid value");
1837-
}
1838-
}
1839-
).set_examples({LLAMA_EXAMPLE_SERVER}));
18401815
add_opt(llama_arg(
18411816
{"--metrics"},
18421817
format("enable prometheus compatible metrics endpoint (default: %s)", params.endpoint_metrics ? "enabled" : "disabled"),
@@ -1956,39 +1931,57 @@ gpt_params_context gpt_params_parser_init(gpt_params & params, llama_example ex,
19561931
else { std::invalid_argument("invalid value"); }
19571932
}
19581933
).set_examples({LLAMA_EXAMPLE_BENCH}));
1959-
#ifndef LOG_DISABLE_LOGS
1960-
// TODO: make this looks less weird
1961-
add_opt(llama_arg(
1962-
{"--log-test"},
1963-
"Log test",
1964-
[](gpt_params &) { log_param_single_parse("--log-test"); }
1965-
));
19661934
add_opt(llama_arg(
19671935
{"--log-disable"},
19681936
"Log disable",
1969-
[](gpt_params &) { log_param_single_parse("--log-disable"); }
1937+
[](gpt_params &) {
1938+
gpt_log_pause(gpt_log_main());
1939+
}
19701940
));
19711941
add_opt(llama_arg(
1972-
{"--log-enable"},
1973-
"Log enable",
1974-
[](gpt_params &) { log_param_single_parse("--log-enable"); }
1942+
{"--log-file"}, "FNAME",
1943+
"Log to file",
1944+
[](gpt_params &, const std::string & value) {
1945+
gpt_log_set_file(gpt_log_main(), value.c_str());
1946+
}
19751947
));
19761948
add_opt(llama_arg(
1977-
{"--log-new"},
1978-
"Log new",
1979-
[](gpt_params &) { log_param_single_parse("--log-new"); }
1980-
));
1949+
{"--log-colors"},
1950+
"Enable colored logging",
1951+
[](gpt_params &) {
1952+
gpt_log_set_colors(gpt_log_main(), true);
1953+
}
1954+
).set_env("LLAMA_LOG_COLORS"));
19811955
add_opt(llama_arg(
1982-
{"--log-append"},
1983-
"Log append",
1984-
[](gpt_params &) { log_param_single_parse("--log-append"); }
1956+
{"-v", "--verbose", "--log-verbose"},
1957+
"Set verbosity level to infinity (i.e. log all messages, useful for debugging)",
1958+
[](gpt_params & params) {
1959+
params.verbosity = INT_MAX;
1960+
gpt_log_set_verbosity_thold(INT_MAX);
1961+
}
19851962
));
19861963
add_opt(llama_arg(
1987-
{"--log-file"}, "FNAME",
1988-
"Log file",
1989-
[](gpt_params &, const std::string & value) { log_param_pair_parse(false, "--log-file", value); }
1990-
));
1991-
#endif // LOG_DISABLE_LOGS
1964+
{"-lv", "--verbosity", "--log-verbosity"}, "N",
1965+
"Set the verbosity threshold. Messages with a higher verbosity will be ignored.",
1966+
[](gpt_params & params, int value) {
1967+
params.verbosity = value;
1968+
gpt_log_set_verbosity_thold(value);
1969+
}
1970+
).set_env("LLAMA_LOG_VERBOSITY"));
1971+
add_opt(llama_arg(
1972+
{"--log-prefix"},
1973+
"Enable prefx in log messages",
1974+
[](gpt_params &) {
1975+
gpt_log_set_prefix(gpt_log_main(), true);
1976+
}
1977+
).set_env("LLAMA_LOG_PREFIX"));
1978+
add_opt(llama_arg(
1979+
{"--log-timestamps"},
1980+
"Enable timestamps in log messages",
1981+
[](gpt_params &) {
1982+
gpt_log_set_timestamps(gpt_log_main(), true);
1983+
}
1984+
).set_env("LLAMA_LOG_TIMESTAMPS"));
19921985

19931986
return ctx_arg;
19941987
}

0 commit comments

Comments
 (0)