vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

common : reimplement logging (#9418)

Browse source 
https://github.com/ggerganov/llama.cpp/pull/9418
This commit is contained in:
Georgi Gerganov 2024-09-15 20:46:12 +03:00 • committed by GitHub
parent e6deac31f7
commit 6262d13e0b
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
54 changed files with 2092 additions and 2419 deletions
Download patch file Download diff file Expand all files Collapse all files

115
common/arg.cpp
View file

@ -1,15 +1,17 @@
#include "arg.h"
#include "log.h"
#include "sampling.h"
#include <algorithm>
#include <string>
#include <vector>
#include <set>
#include <climits>
#include <cstdarg>
#include <fstream>
#include <regex>
#include <cstdarg>
#include <climits>
#include <set>
#include <string>
#include <thread>
#include <vector>
#include "json-schema-to-grammar.h"
@ -383,20 +385,6 @@ gpt_params_context gpt_params_parser_init(gpt_params & params, llama_example ex,
exit(0);
}
));
add_opt(llama_arg(
{"-v", "--verbose"},
"print verbose information",
[](gpt_params & params) {
params.verbosity = 1;
}
));
add_opt(llama_arg(
{"--verbosity"}, "N",
format("set specific verbosity level (default: %d)", params.verbosity),
[](gpt_params & params, int value) {
params.verbosity = value;
}
));
add_opt(llama_arg(
{"--verbose-prompt"},
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,
[](gpt_params & params) {
params.use_color = true;
}
).set_examples({LLAMA_EXAMPLE_MAIN, LLAMA_EXAMPLE_INFILL}));
).set_examples({LLAMA_EXAMPLE_MAIN, LLAMA_EXAMPLE_INFILL, LLAMA_EXAMPLE_SPECULATIVE, LLAMA_EXAMPLE_LOOKUP}));
add_opt(llama_arg(
{"-t", "--threads"}, "N",
format("number of threads to use during generation (default: %d)", params.cpuparams.n_threads),
@ -876,7 +864,7 @@ gpt_params_context gpt_params_parser_init(gpt_params & params, llama_example ex,
params.input_prefix = value;
params.enable_chat_template = false;
}
).set_examples({LLAMA_EXAMPLE_MAIN}));
).set_examples({LLAMA_EXAMPLE_MAIN, LLAMA_EXAMPLE_INFILL}));
add_opt(llama_arg(
{"--in-suffix"}, "STRING",
"string to suffix after user inputs with (default: empty)",
@ -884,7 +872,7 @@ gpt_params_context gpt_params_parser_init(gpt_params & params, llama_example ex,
params.input_suffix = value;
params.enable_chat_template = false;
}
).set_examples({LLAMA_EXAMPLE_MAIN}));
).set_examples({LLAMA_EXAMPLE_MAIN, LLAMA_EXAMPLE_INFILL}));
add_opt(llama_arg(
{"--no-warmup"},
"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,
params.system_prompt = system_prompt;
}
).set_examples({LLAMA_EXAMPLE_SERVER}));
add_opt(llama_arg(
{"--log-format"}, "{text, json}",
"log output format: json or text (default: json)",
[](gpt_params & params, const std::string & value) {
if (value == "json") {
params.log_json = true;
} else if (value == "text") {
params.log_json = false;
} else {
throw std::invalid_argument("invalid value");
}
}
).set_examples({LLAMA_EXAMPLE_SERVER}));
add_opt(llama_arg(
{"--metrics"},
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,
else { std::invalid_argument("invalid value"); }
}
).set_examples({LLAMA_EXAMPLE_BENCH}));
#ifndef LOG_DISABLE_LOGS
// TODO: make this looks less weird
add_opt(llama_arg(
{"--log-test"},
"Log test",
[](gpt_params &) { log_param_single_parse("--log-test"); }
));
add_opt(llama_arg(
{"--log-disable"},
"Log disable",
[](gpt_params &) { log_param_single_parse("--log-disable"); }
));
add_opt(llama_arg(
{"--log-enable"},
"Log enable",
[](gpt_params &) { log_param_single_parse("--log-enable"); }
));
add_opt(llama_arg(
{"--log-new"},
"Log new",
[](gpt_params &) { log_param_single_parse("--log-new"); }
));
add_opt(llama_arg(
{"--log-append"},
"Log append",
[](gpt_params &) { log_param_single_parse("--log-append"); }
[](gpt_params &) {
gpt_log_pause(gpt_log_main());
}
));
add_opt(llama_arg(
{"--log-file"}, "FNAME",
"Log file",
[](gpt_params &, const std::string & value) { log_param_pair_parse(false, "--log-file", value); }
"Log to file",
[](gpt_params &, const std::string & value) {
gpt_log_set_file(gpt_log_main(), value.c_str());
}
));
#endif // LOG_DISABLE_LOGS
add_opt(llama_arg(
{"--log-colors"},
"Enable colored logging",
[](gpt_params &) {
gpt_log_set_colors(gpt_log_main(), true);
}
).set_env("LLAMA_LOG_COLORS"));
add_opt(llama_arg(
{"-v", "--verbose", "--log-verbose"},
"Set verbosity level to infinity (i.e. log all messages, useful for debugging)",
[](gpt_params & params) {
params.verbosity = INT_MAX;
gpt_log_set_verbosity_thold(INT_MAX);
}
));
add_opt(llama_arg(
{"-lv", "--verbosity", "--log-verbosity"}, "N",
"Set the verbosity threshold. Messages with a higher verbosity will be ignored.",
[](gpt_params & params, int value) {
params.verbosity = value;
gpt_log_set_verbosity_thold(value);
}
).set_env("LLAMA_LOG_VERBOSITY"));
add_opt(llama_arg(
{"--log-prefix"},
"Enable prefx in log messages",
[](gpt_params &) {
gpt_log_set_prefix(gpt_log_main(), true);
}
).set_env("LLAMA_LOG_PREFIX"));
add_opt(llama_arg(
{"--log-timestamps"},
"Enable timestamps in log messages",
[](gpt_params &) {
gpt_log_set_timestamps(gpt_log_main(), true);
}
).set_env("LLAMA_LOG_TIMESTAMPS"));
return ctx_arg;
}