Skip to content

Commit 5766f90

Browse files
JohannesGaesslerarthw
JohannesGaessler
authored and
arthw
committed
llama/ex: remove --logdir argument (ggml-org#10339)
1 parent 30beaa4 commit 5766f90

File tree

10 files changed

+0
-547
lines changed

10 files changed

+0
-547
lines changed

‎common/arg.cpp‎

Lines changed: 0 additions & 11 deletions
Original file line numberDiff line numberDiff line change
@@ -1939,17 +1939,6 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
19391939
params.simple_io = true;
19401940
}
19411941
).set_examples({LLAMA_EXAMPLE_MAIN, LLAMA_EXAMPLE_INFILL}));
1942-
add_opt(common_arg(
1943-
{"-ld", "--logdir"}, "LOGDIR",
1944-
"path under which to save YAML logs (no logging if unset)",
1945-
[](common_params & params, const std::string & value) {
1946-
params.logdir = value;
1947-
1948-
if (params.logdir.back() != DIRECTORY_SEPARATOR) {
1949-
params.logdir += DIRECTORY_SEPARATOR;
1950-
}
1951-
}
1952-
));
19531942
add_opt(common_arg(
19541943
{"--positive-file"}, "FNAME",
19551944
string_format("positive prompts file, one prompt per line (default: '%s')", params.cvector_positive_file.c_str()),

‎common/common.cpp‎

Lines changed: 0 additions & 210 deletions
Original file line numberDiff line numberDiff line change
@@ -1890,213 +1890,3 @@ common_control_vector_data common_control_vector_load(const std::vector<common_c
18901890
return result;
18911891
}
18921892

1893-
//
1894-
// YAML utils
1895-
//
1896-
1897-
void yaml_dump_vector_float(FILE * stream, const char * prop_name, const std::vector<float> & data) {
1898-
if (data.empty()) {
1899-
fprintf(stream, "%s:\n", prop_name);
1900-
return;
1901-
}
1902-
1903-
fprintf(stream, "%s: [", prop_name);
1904-
for (size_t i = 0; i < data.size() - 1; ++i) {
1905-
fprintf(stream, "%e, ", data[i]);
1906-
}
1907-
fprintf(stream, "%e]\n", data.back());
1908-
}
1909-
1910-
void yaml_dump_vector_int(FILE * stream, const char * prop_name, const std::vector<int> & data) {
1911-
if (data.empty()) {
1912-
fprintf(stream, "%s:\n", prop_name);
1913-
return;
1914-
}
1915-
1916-
fprintf(stream, "%s: [", prop_name);
1917-
for (size_t i = 0; i < data.size() - 1; ++i) {
1918-
fprintf(stream, "%d, ", data[i]);
1919-
}
1920-
fprintf(stream, "%d]\n", data.back());
1921-
}
1922-
1923-
void yaml_dump_string_multiline(FILE * stream, const char * prop_name, const char * data) {
1924-
std::string data_str(data == NULL ? "" : data);
1925-
1926-
if (data_str.empty()) {
1927-
fprintf(stream, "%s:\n", prop_name);
1928-
return;
1929-
}
1930-
1931-
size_t pos_start = 0;
1932-
size_t pos_found = 0;
1933-
1934-
if (std::isspace(data_str[0]) || std::isspace(data_str.back())) {
1935-
data_str = std::regex_replace(data_str, std::regex("\n"), "\\n");
1936-
data_str = std::regex_replace(data_str, std::regex("\""), "\\\"");
1937-
data_str = std::regex_replace(data_str, std::regex(R"(\\[^n"])"), R"(\$&)");
1938-
data_str = "\"" + data_str + "\"";
1939-
fprintf(stream, "%s: %s\n", prop_name, data_str.c_str());
1940-
return;
1941-
}
1942-
1943-
if (data_str.find('\n') == std::string::npos) {
1944-
fprintf(stream, "%s: %s\n", prop_name, data_str.c_str());
1945-
return;
1946-
}
1947-
1948-
fprintf(stream, "%s: |\n", prop_name);
1949-
while ((pos_found = data_str.find('\n', pos_start)) != std::string::npos) {
1950-
fprintf(stream, " %s\n", data_str.substr(pos_start, pos_found-pos_start).c_str());
1951-
pos_start = pos_found + 1;
1952-
}
1953-
}
1954-
1955-
void yaml_dump_non_result_info(FILE * stream, const common_params & params, const llama_context * lctx,
1956-
const std::string & timestamp, const std::vector<int> & prompt_tokens, const char * model_desc) {
1957-
ggml_cpu_init(); // some ARM features are detected at runtime
1958-
1959-
const auto & sparams = params.sparams;
1960-
1961-
fprintf(stream, "build_commit: %s\n", LLAMA_COMMIT);
1962-
fprintf(stream, "build_number: %d\n", LLAMA_BUILD_NUMBER);
1963-
fprintf(stream, "cpu_has_arm_fma: %s\n", ggml_cpu_has_arm_fma() ? "true" : "false");
1964-
fprintf(stream, "cpu_has_avx: %s\n", ggml_cpu_has_avx() ? "true" : "false");
1965-
fprintf(stream, "cpu_has_avx_vnni: %s\n", ggml_cpu_has_avx_vnni() ? "true" : "false");
1966-
fprintf(stream, "cpu_has_avx2: %s\n", ggml_cpu_has_avx2() ? "true" : "false");
1967-
fprintf(stream, "cpu_has_avx512: %s\n", ggml_cpu_has_avx512() ? "true" : "false");
1968-
fprintf(stream, "cpu_has_avx512_vbmi: %s\n", ggml_cpu_has_avx512_vbmi() ? "true" : "false");
1969-
fprintf(stream, "cpu_has_avx512_vnni: %s\n", ggml_cpu_has_avx512_vnni() ? "true" : "false");
1970-
fprintf(stream, "cpu_has_fma: %s\n", ggml_cpu_has_fma() ? "true" : "false");
1971-
fprintf(stream, "cpu_has_neon: %s\n", ggml_cpu_has_neon() ? "true" : "false");
1972-
fprintf(stream, "cpu_has_sve: %s\n", ggml_cpu_has_sve() ? "true" : "false");
1973-
fprintf(stream, "cpu_has_f16c: %s\n", ggml_cpu_has_f16c() ? "true" : "false");
1974-
fprintf(stream, "cpu_has_fp16_va: %s\n", ggml_cpu_has_fp16_va() ? "true" : "false");
1975-
fprintf(stream, "cpu_has_riscv_v: %s\n", ggml_cpu_has_riscv_v() ? "true" : "false");
1976-
fprintf(stream, "cpu_has_wasm_simd: %s\n", ggml_cpu_has_wasm_simd() ? "true" : "false");
1977-
fprintf(stream, "cpu_has_sse3: %s\n", ggml_cpu_has_sse3() ? "true" : "false");
1978-
fprintf(stream, "cpu_has_vsx: %s\n", ggml_cpu_has_vsx() ? "true" : "false");
1979-
fprintf(stream, "cpu_has_matmul_int8: %s\n", ggml_cpu_has_matmul_int8() ? "true" : "false");
1980-
1981-
#ifdef NDEBUG
1982-
fprintf(stream, "debug: false\n");
1983-
#else
1984-
fprintf(stream, "debug: true\n");
1985-
#endif // NDEBUG
1986-
1987-
fprintf(stream, "model_desc: %s\n", model_desc);
1988-
fprintf(stream, "n_vocab: %d # output size of the final layer, 32001 for some models\n", llama_n_vocab(llama_get_model(lctx)));
1989-
1990-
#ifdef __OPTIMIZE__
1991-
fprintf(stream, "optimize: true\n");
1992-
#else
1993-
fprintf(stream, "optimize: false\n");
1994-
#endif // __OPTIMIZE__
1995-
1996-
fprintf(stream, "time: %s\n", timestamp.c_str());
1997-
1998-
fprintf(stream, "\n");
1999-
fprintf(stream, "###############\n");
2000-
fprintf(stream, "# User Inputs #\n");
2001-
fprintf(stream, "###############\n");
2002-
fprintf(stream, "\n");
2003-
2004-
fprintf(stream, "alias: %s # default: unknown\n", params.model_alias.c_str());
2005-
fprintf(stream, "batch_size: %d # default: 512\n", params.n_batch);
2006-
fprintf(stream, "chunks: %d # default: -1 (unlimited)\n", params.n_chunks);
2007-
fprintf(stream, "color: %s # default: false\n", params.use_color ? "true" : "false");
2008-
fprintf(stream, "ctx_size: %d # default: 512\n", params.n_ctx);
2009-
fprintf(stream, "dry_allowed_length: %d # default: 2\n", sparams.dry_allowed_length);
2010-
fprintf(stream, "dry_base: %.2f # default: 1.75\n", sparams.dry_base);
2011-
fprintf(stream, "dry_multiplier: %.1f # default: 0.0\n", sparams.dry_multiplier);
2012-
fprintf(stream, "dry_penalty_last_n: %d # default: -1 (0 = disable, -1 = context size)\n", sparams.dry_penalty_last_n);
2013-
fprintf(stream, "escape: %s # default: false\n", params.escape ? "true" : "false");
2014-
fprintf(stream, "file: # never logged, see prompt instead. Can still be specified for input.\n");
2015-
fprintf(stream, "frequency_penalty: %f # default: 0.0 \n", sparams.penalty_freq);
2016-
yaml_dump_string_multiline(stream, "grammar", sparams.grammar.c_str());
2017-
fprintf(stream, "grammar-file: # never logged, see grammar instead. Can still be specified for input.\n");
2018-
fprintf(stream, "hellaswag: %s # default: false\n", params.hellaswag ? "true" : "false");
2019-
fprintf(stream, "hellaswag_tasks: %zu # default: 400\n", params.hellaswag_tasks);
2020-
fprintf(stream, "ignore_eos: %s # default: false\n", sparams.ignore_eos ? "true" : "false");
2021-
2022-
yaml_dump_string_multiline(stream, "in_prefix", params.input_prefix.c_str());
2023-
fprintf(stream, "in_prefix_bos: %s # default: false\n", params.input_prefix_bos ? "true" : "false");
2024-
yaml_dump_string_multiline(stream, "in_suffix", params.input_prefix.c_str());
2025-
fprintf(stream, "interactive: %s # default: false\n", params.interactive ? "true" : "false");
2026-
fprintf(stream, "interactive_first: %s # default: false\n", params.interactive_first ? "true" : "false");
2027-
fprintf(stream, "keep: %d # default: 0\n", params.n_keep);
2028-
fprintf(stream, "logdir: %s # default: unset (no logging)\n", params.logdir.c_str());
2029-
2030-
fprintf(stream, "logit_bias:\n");
2031-
for (const auto & logit_bias : sparams.logit_bias) {
2032-
fprintf(stream, " %d: %f", logit_bias.token, logit_bias.bias);
2033-
}
2034-
2035-
fprintf(stream, "lora:\n");
2036-
for (auto & la : params.lora_adapters) {
2037-
if (la.scale == 1.0f) {
2038-
fprintf(stream, " - %s\n", la.path.c_str());
2039-
}
2040-
}
2041-
fprintf(stream, "lora_scaled:\n");
2042-
for (auto & la : params.lora_adapters) {
2043-
if (la.scale != 1.0f) {
2044-
fprintf(stream, " - %s: %f\n", la.path.c_str(), la.scale);
2045-
}
2046-
}
2047-
fprintf(stream, "lora_init_without_apply: %s # default: false\n", params.lora_init_without_apply ? "true" : "false");
2048-
fprintf(stream, "main_gpu: %d # default: 0\n", params.main_gpu);
2049-
fprintf(stream, "min_keep: %d # default: 0 (disabled)\n", sparams.min_keep);
2050-
fprintf(stream, "mirostat: %d # default: 0 (disabled)\n", sparams.mirostat);
2051-
fprintf(stream, "mirostat_ent: %f # default: 5.0\n", sparams.mirostat_tau);
2052-
fprintf(stream, "mirostat_lr: %f # default: 0.1\n", sparams.mirostat_eta);
2053-
fprintf(stream, "mlock: %s # default: false\n", params.use_mlock ? "true" : "false");
2054-
fprintf(stream, "model: %s # default: %s\n", params.model.c_str(), DEFAULT_MODEL_PATH);
2055-
fprintf(stream, "model_draft: %s # default:\n", params.model_draft.c_str());
2056-
fprintf(stream, "multiline_input: %s # default: false\n", params.multiline_input ? "true" : "false");
2057-
fprintf(stream, "n_gpu_layers: %d # default: -1\n", params.n_gpu_layers);
2058-
fprintf(stream, "n_predict: %d # default: -1 (unlimited)\n", params.n_predict);
2059-
fprintf(stream, "n_probs: %d # only used by server binary, default: 0\n", sparams.n_probs);
2060-
fprintf(stream, "no_mmap: %s # default: false\n", !params.use_mmap ? "true" : "false");
2061-
fprintf(stream, "penalize_nl: %s # default: false\n", sparams.penalize_nl ? "true" : "false");
2062-
fprintf(stream, "ppl_output_type: %d # default: 0\n", params.ppl_output_type);
2063-
fprintf(stream, "ppl_stride: %d # default: 0\n", params.ppl_stride);
2064-
fprintf(stream, "presence_penalty: %f # default: 0.0\n", sparams.penalty_present);
2065-
yaml_dump_string_multiline(stream, "prompt", params.prompt.c_str());
2066-
fprintf(stream, "prompt_cache: %s\n", params.path_prompt_cache.c_str());
2067-
fprintf(stream, "prompt_cache_all: %s # default: false\n", params.prompt_cache_all ? "true" : "false");
2068-
fprintf(stream, "prompt_cache_ro: %s # default: false\n", params.prompt_cache_ro ? "true" : "false");
2069-
yaml_dump_vector_int(stream, "prompt_tokens", prompt_tokens);
2070-
fprintf(stream, "repeat_penalty: %f # default: 1.1\n", sparams.penalty_repeat);
2071-
2072-
fprintf(stream, "reverse_prompt:\n");
2073-
for (std::string ap : params.antiprompt) {
2074-
size_t pos = 0;
2075-
while ((pos = ap.find('\n', pos)) != std::string::npos) {
2076-
ap.replace(pos, 1, "\\n");
2077-
pos += 1;
2078-
}
2079-
2080-
fprintf(stream, " - %s\n", ap.c_str());
2081-
}
2082-
2083-
fprintf(stream, "rope_freq_base: %f # default: 10000.0\n", params.rope_freq_base);
2084-
fprintf(stream, "rope_freq_scale: %f # default: 1.0\n", params.rope_freq_scale);
2085-
fprintf(stream, "simple_io: %s # default: false\n", params.simple_io ? "true" : "false");
2086-
fprintf(stream, "cont_batching: %s # default: false\n", params.cont_batching ? "true" : "false");
2087-
fprintf(stream, "flash_attn: %s # default: false\n", params.flash_attn ? "true" : "false");
2088-
fprintf(stream, "temp: %f # default: 0.8\n", sparams.temp);
2089-
2090-
const std::vector<float> tensor_split_vector(params.tensor_split, params.tensor_split + llama_max_devices());
2091-
yaml_dump_vector_float(stream, "tensor_split", tensor_split_vector);
2092-
2093-
fprintf(stream, "threads: %d # default: %u\n", params.cpuparams.n_threads, std::thread::hardware_concurrency());
2094-
fprintf(stream, "top_k: %d # default: 40\n", sparams.top_k);
2095-
fprintf(stream, "top_p: %f # default: 0.95\n", sparams.top_p);
2096-
fprintf(stream, "min_p: %f # default: 0.0\n", sparams.min_p);
2097-
fprintf(stream, "xtc_probability: %f # default: 0.0\n", sparams.xtc_probability);
2098-
fprintf(stream, "xtc_threshold: %f # default: 0.1\n", sparams.xtc_threshold);
2099-
fprintf(stream, "typ_p: %f # default: 1.0\n", sparams.typ_p);
2100-
fprintf(stream, "verbose_prompt: %s # default: false\n", params.verbose_prompt ? "true" : "false");
2101-
fprintf(stream, "display_prompt: %s # default: true\n", params.display_prompt ? "true" : "false");
2102-
}

‎common/common.h‎

Lines changed: 0 additions & 13 deletions
Original file line numberDiff line numberDiff line change
@@ -209,7 +209,6 @@ struct common_params {
209209
std::string path_prompt_cache = ""; // path to file for saving/loading prompt eval state // NOLINT
210210
std::string input_prefix = ""; // string to prefix user inputs with // NOLINT
211211
std::string input_suffix = ""; // string to suffix user inputs with // NOLINT
212-
std::string logdir = ""; // directory in which to save YAML log files // NOLINT
213212
std::string lookup_cache_static = ""; // path of static ngram cache file for lookup decoding // NOLINT
214213
std::string lookup_cache_dynamic = ""; // path of dynamic ngram cache file for lookup decoding // NOLINT
215214
std::string logits_file = ""; // file for saving *all* logits // NOLINT
@@ -584,15 +583,3 @@ common_control_vector_data common_control_vector_load(const std::vector<common_c
584583
static const char * const LLM_KV_SPLIT_NO = "split.no";
585584
static const char * const LLM_KV_SPLIT_COUNT = "split.count";
586585
static const char * const LLM_KV_SPLIT_TENSORS_COUNT = "split.tensors.count";
587-
588-
//
589-
// YAML utils
590-
//
591-
592-
void yaml_dump_vector_float (FILE * stream, const char * prop_name, const std::vector<float> & data);
593-
void yaml_dump_vector_int (FILE * stream, const char * prop_name, const std::vector<int> & data);
594-
void yaml_dump_string_multiline(FILE * stream, const char * prop_name, const char * data);
595-
596-
void yaml_dump_non_result_info(
597-
FILE * stream, const common_params & params, const llama_context * lctx,
598-
const std::string & timestamp, const std::vector<int> & prompt_tokens, const char * model_desc);

‎examples/infill/infill.cpp‎

Lines changed: 0 additions & 46 deletions
Original file line numberDiff line numberDiff line change
@@ -43,50 +43,6 @@ static std::vector<llama_token> * g_output_tokens;
4343

4444
static bool is_interacting = false;
4545

46-
static void write_logfile(
47-
const llama_context * ctx, const common_params & params, const llama_model * model,
48-
const std::vector<llama_token> & input_tokens, const std::string & output,
49-
const std::vector<llama_token> & output_tokens
50-
) {
51-
if (params.logdir.empty()) {
52-
return;
53-
}
54-
55-
const std::string timestamp = string_get_sortable_timestamp();
56-
57-
const bool success = fs_create_directory_with_parents(params.logdir);
58-
if (!success) {
59-
LOG_ERR("%s: warning: failed to create logdir %s, cannot write logfile\n",
60-
__func__, params.logdir.c_str());
61-
return;
62-
}
63-
64-
const std::string logfile_path = params.logdir + timestamp + ".yml";
65-
FILE * logfile = fopen(logfile_path.c_str(), "w");
66-
67-
if (logfile == NULL) {
68-
LOG_ERR("%s: failed to open logfile %s\n", __func__, logfile_path.c_str());
69-
return;
70-
}
71-
72-
fprintf(logfile, "binary: infill\n");
73-
char model_desc[128];
74-
llama_model_desc(model, model_desc, sizeof(model_desc));
75-
yaml_dump_non_result_info(logfile, params, ctx, timestamp, input_tokens, model_desc);
76-
77-
fprintf(logfile, "\n");
78-
fprintf(logfile, "######################\n");
79-
fprintf(logfile, "# Generation Results #\n");
80-
fprintf(logfile, "######################\n");
81-
fprintf(logfile, "\n");
82-
83-
yaml_dump_string_multiline(logfile, "output", output.c_str());
84-
yaml_dump_vector_int(logfile, "output_tokens", output_tokens);
85-
86-
llama_perf_dump_yaml(logfile, ctx);
87-
fclose(logfile);
88-
}
89-
9046
#if defined (__unix__) || (defined (__APPLE__) && defined (__MACH__)) || defined (_WIN32)
9147
static void sigint_handler(int signo) {
9248
if (signo == SIGINT) {
@@ -96,7 +52,6 @@ static void sigint_handler(int signo) {
9652
console::cleanup();
9753
LOG("\n");
9854
common_perf_print(*g_ctx, *g_smpl);
99-
write_logfile(*g_ctx, *g_params, *g_model, *g_input_tokens, g_output_ss->str(), *g_output_tokens);
10055

10156
// make sure all logs are flushed
10257
LOG("Interrupted by user\n");
@@ -625,7 +580,6 @@ int main(int argc, char ** argv) {
625580

626581
LOG("\n");
627582
common_perf_print(ctx, smpl);
628-
write_logfile(ctx, params, model, input_tokens, output_ss.str(), output_tokens);
629583

630584
llama_free(ctx);
631585
llama_free_model(model);

‎examples/main/main.cpp‎

Lines changed: 0 additions & 45 deletions
Original file line numberDiff line numberDiff line change
@@ -62,49 +62,6 @@ static bool file_is_empty(const std::string & path) {
6262
return f.tellg() == 0;
6363
}
6464

65-
static void write_logfile(
66-
const llama_context * ctx, const common_params & params, const llama_model * model,
67-
const std::vector<llama_token> & input_tokens, const std::string & output,
68-
const std::vector<llama_token> & output_tokens
69-
) {
70-
if (params.logdir.empty()) {
71-
return;
72-
}
73-
74-
const std::string timestamp = string_get_sortable_timestamp();
75-
76-
const bool success = fs_create_directory_with_parents(params.logdir);
77-
if (!success) {
78-
LOG_ERR("%s: failed to create logdir %s, cannot write logfile\n", __func__, params.logdir.c_str());
79-
return;
80-
}
81-
82-
const std::string logfile_path = params.logdir + timestamp + ".yml";
83-
FILE * logfile = fopen(logfile_path.c_str(), "w");
84-
85-
if (logfile == NULL) {
86-
LOG_ERR("%s: failed to open logfile %s\n", __func__, logfile_path.c_str());
87-
return;
88-
}
89-
90-
fprintf(logfile, "binary: main\n");
91-
char model_desc[128];
92-
llama_model_desc(model, model_desc, sizeof(model_desc));
93-
yaml_dump_non_result_info(logfile, params, ctx, timestamp, input_tokens, model_desc);
94-
95-
fprintf(logfile, "\n");
96-
fprintf(logfile, "######################\n");
97-
fprintf(logfile, "# Generation Results #\n");
98-
fprintf(logfile, "######################\n");
99-
fprintf(logfile, "\n");
100-
101-
yaml_dump_string_multiline(logfile, "output", output.c_str());
102-
yaml_dump_vector_int(logfile, "output_tokens", output_tokens);
103-
104-
llama_perf_dump_yaml(logfile, ctx);
105-
fclose(logfile);
106-
}
107-
10865
#if defined (__unix__) || (defined (__APPLE__) && defined (__MACH__)) || defined (_WIN32)
10966
static void sigint_handler(int signo) {
11067
if (signo == SIGINT) {
@@ -115,7 +72,6 @@ static void sigint_handler(int signo) {
11572
console::cleanup();
11673
LOG("\n");
11774
common_perf_print(*g_ctx, *g_smpl);
118-
write_logfile(*g_ctx, *g_params, *g_model, *g_input_tokens, g_output_ss->str(), *g_output_tokens);
11975

12076
// make sure all logs are flushed
12177
LOG("Interrupted by user\n");
@@ -926,7 +882,6 @@ int main(int argc, char ** argv) {
926882

927883
LOG("\n\n");
928884
common_perf_print(ctx, smpl);
929-
write_logfile(ctx, params, model, input_tokens, output_ss.str(), output_tokens);
930885

931886
common_sampler_free(smpl);
932887

0 commit comments

Comments
 (0)