mirror of
https://github.com/ggerganov/llama.cpp.git
synced 2025-01-24 18:39:19 +01:00
llama/ex: remove --logdir argument (#10339)
This commit is contained in:
parent
db4cfd5dbc
commit
4e54be0ec6
@ -1939,17 +1939,6 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
|
||||
params.simple_io = true;
|
||||
}
|
||||
).set_examples({LLAMA_EXAMPLE_MAIN, LLAMA_EXAMPLE_INFILL}));
|
||||
add_opt(common_arg(
|
||||
{"-ld", "--logdir"}, "LOGDIR",
|
||||
"path under which to save YAML logs (no logging if unset)",
|
||||
[](common_params & params, const std::string & value) {
|
||||
params.logdir = value;
|
||||
|
||||
if (params.logdir.back() != DIRECTORY_SEPARATOR) {
|
||||
params.logdir += DIRECTORY_SEPARATOR;
|
||||
}
|
||||
}
|
||||
));
|
||||
add_opt(common_arg(
|
||||
{"--positive-file"}, "FNAME",
|
||||
string_format("positive prompts file, one prompt per line (default: '%s')", params.cvector_positive_file.c_str()),
|
||||
|
@ -1890,213 +1890,3 @@ common_control_vector_data common_control_vector_load(const std::vector<common_c
|
||||
return result;
|
||||
}
|
||||
|
||||
//
|
||||
// YAML utils
|
||||
//
|
||||
|
||||
void yaml_dump_vector_float(FILE * stream, const char * prop_name, const std::vector<float> & data) {
|
||||
if (data.empty()) {
|
||||
fprintf(stream, "%s:\n", prop_name);
|
||||
return;
|
||||
}
|
||||
|
||||
fprintf(stream, "%s: [", prop_name);
|
||||
for (size_t i = 0; i < data.size() - 1; ++i) {
|
||||
fprintf(stream, "%e, ", data[i]);
|
||||
}
|
||||
fprintf(stream, "%e]\n", data.back());
|
||||
}
|
||||
|
||||
void yaml_dump_vector_int(FILE * stream, const char * prop_name, const std::vector<int> & data) {
|
||||
if (data.empty()) {
|
||||
fprintf(stream, "%s:\n", prop_name);
|
||||
return;
|
||||
}
|
||||
|
||||
fprintf(stream, "%s: [", prop_name);
|
||||
for (size_t i = 0; i < data.size() - 1; ++i) {
|
||||
fprintf(stream, "%d, ", data[i]);
|
||||
}
|
||||
fprintf(stream, "%d]\n", data.back());
|
||||
}
|
||||
|
||||
void yaml_dump_string_multiline(FILE * stream, const char * prop_name, const char * data) {
|
||||
std::string data_str(data == NULL ? "" : data);
|
||||
|
||||
if (data_str.empty()) {
|
||||
fprintf(stream, "%s:\n", prop_name);
|
||||
return;
|
||||
}
|
||||
|
||||
size_t pos_start = 0;
|
||||
size_t pos_found = 0;
|
||||
|
||||
if (std::isspace(data_str[0]) || std::isspace(data_str.back())) {
|
||||
data_str = std::regex_replace(data_str, std::regex("\n"), "\\n");
|
||||
data_str = std::regex_replace(data_str, std::regex("\""), "\\\"");
|
||||
data_str = std::regex_replace(data_str, std::regex(R"(\\[^n"])"), R"(\$&)");
|
||||
data_str = "\"" + data_str + "\"";
|
||||
fprintf(stream, "%s: %s\n", prop_name, data_str.c_str());
|
||||
return;
|
||||
}
|
||||
|
||||
if (data_str.find('\n') == std::string::npos) {
|
||||
fprintf(stream, "%s: %s\n", prop_name, data_str.c_str());
|
||||
return;
|
||||
}
|
||||
|
||||
fprintf(stream, "%s: |\n", prop_name);
|
||||
while ((pos_found = data_str.find('\n', pos_start)) != std::string::npos) {
|
||||
fprintf(stream, " %s\n", data_str.substr(pos_start, pos_found-pos_start).c_str());
|
||||
pos_start = pos_found + 1;
|
||||
}
|
||||
}
|
||||
|
||||
void yaml_dump_non_result_info(FILE * stream, const common_params & params, const llama_context * lctx,
|
||||
const std::string & timestamp, const std::vector<int> & prompt_tokens, const char * model_desc) {
|
||||
ggml_cpu_init(); // some ARM features are detected at runtime
|
||||
|
||||
const auto & sparams = params.sparams;
|
||||
|
||||
fprintf(stream, "build_commit: %s\n", LLAMA_COMMIT);
|
||||
fprintf(stream, "build_number: %d\n", LLAMA_BUILD_NUMBER);
|
||||
fprintf(stream, "cpu_has_arm_fma: %s\n", ggml_cpu_has_arm_fma() ? "true" : "false");
|
||||
fprintf(stream, "cpu_has_avx: %s\n", ggml_cpu_has_avx() ? "true" : "false");
|
||||
fprintf(stream, "cpu_has_avx_vnni: %s\n", ggml_cpu_has_avx_vnni() ? "true" : "false");
|
||||
fprintf(stream, "cpu_has_avx2: %s\n", ggml_cpu_has_avx2() ? "true" : "false");
|
||||
fprintf(stream, "cpu_has_avx512: %s\n", ggml_cpu_has_avx512() ? "true" : "false");
|
||||
fprintf(stream, "cpu_has_avx512_vbmi: %s\n", ggml_cpu_has_avx512_vbmi() ? "true" : "false");
|
||||
fprintf(stream, "cpu_has_avx512_vnni: %s\n", ggml_cpu_has_avx512_vnni() ? "true" : "false");
|
||||
fprintf(stream, "cpu_has_fma: %s\n", ggml_cpu_has_fma() ? "true" : "false");
|
||||
fprintf(stream, "cpu_has_neon: %s\n", ggml_cpu_has_neon() ? "true" : "false");
|
||||
fprintf(stream, "cpu_has_sve: %s\n", ggml_cpu_has_sve() ? "true" : "false");
|
||||
fprintf(stream, "cpu_has_f16c: %s\n", ggml_cpu_has_f16c() ? "true" : "false");
|
||||
fprintf(stream, "cpu_has_fp16_va: %s\n", ggml_cpu_has_fp16_va() ? "true" : "false");
|
||||
fprintf(stream, "cpu_has_riscv_v: %s\n", ggml_cpu_has_riscv_v() ? "true" : "false");
|
||||
fprintf(stream, "cpu_has_wasm_simd: %s\n", ggml_cpu_has_wasm_simd() ? "true" : "false");
|
||||
fprintf(stream, "cpu_has_sse3: %s\n", ggml_cpu_has_sse3() ? "true" : "false");
|
||||
fprintf(stream, "cpu_has_vsx: %s\n", ggml_cpu_has_vsx() ? "true" : "false");
|
||||
fprintf(stream, "cpu_has_matmul_int8: %s\n", ggml_cpu_has_matmul_int8() ? "true" : "false");
|
||||
|
||||
#ifdef NDEBUG
|
||||
fprintf(stream, "debug: false\n");
|
||||
#else
|
||||
fprintf(stream, "debug: true\n");
|
||||
#endif // NDEBUG
|
||||
|
||||
fprintf(stream, "model_desc: %s\n", model_desc);
|
||||
fprintf(stream, "n_vocab: %d # output size of the final layer, 32001 for some models\n", llama_n_vocab(llama_get_model(lctx)));
|
||||
|
||||
#ifdef __OPTIMIZE__
|
||||
fprintf(stream, "optimize: true\n");
|
||||
#else
|
||||
fprintf(stream, "optimize: false\n");
|
||||
#endif // __OPTIMIZE__
|
||||
|
||||
fprintf(stream, "time: %s\n", timestamp.c_str());
|
||||
|
||||
fprintf(stream, "\n");
|
||||
fprintf(stream, "###############\n");
|
||||
fprintf(stream, "# User Inputs #\n");
|
||||
fprintf(stream, "###############\n");
|
||||
fprintf(stream, "\n");
|
||||
|
||||
fprintf(stream, "alias: %s # default: unknown\n", params.model_alias.c_str());
|
||||
fprintf(stream, "batch_size: %d # default: 512\n", params.n_batch);
|
||||
fprintf(stream, "chunks: %d # default: -1 (unlimited)\n", params.n_chunks);
|
||||
fprintf(stream, "color: %s # default: false\n", params.use_color ? "true" : "false");
|
||||
fprintf(stream, "ctx_size: %d # default: 512\n", params.n_ctx);
|
||||
fprintf(stream, "dry_allowed_length: %d # default: 2\n", sparams.dry_allowed_length);
|
||||
fprintf(stream, "dry_base: %.2f # default: 1.75\n", sparams.dry_base);
|
||||
fprintf(stream, "dry_multiplier: %.1f # default: 0.0\n", sparams.dry_multiplier);
|
||||
fprintf(stream, "dry_penalty_last_n: %d # default: -1 (0 = disable, -1 = context size)\n", sparams.dry_penalty_last_n);
|
||||
fprintf(stream, "escape: %s # default: false\n", params.escape ? "true" : "false");
|
||||
fprintf(stream, "file: # never logged, see prompt instead. Can still be specified for input.\n");
|
||||
fprintf(stream, "frequency_penalty: %f # default: 0.0 \n", sparams.penalty_freq);
|
||||
yaml_dump_string_multiline(stream, "grammar", sparams.grammar.c_str());
|
||||
fprintf(stream, "grammar-file: # never logged, see grammar instead. Can still be specified for input.\n");
|
||||
fprintf(stream, "hellaswag: %s # default: false\n", params.hellaswag ? "true" : "false");
|
||||
fprintf(stream, "hellaswag_tasks: %zu # default: 400\n", params.hellaswag_tasks);
|
||||
fprintf(stream, "ignore_eos: %s # default: false\n", sparams.ignore_eos ? "true" : "false");
|
||||
|
||||
yaml_dump_string_multiline(stream, "in_prefix", params.input_prefix.c_str());
|
||||
fprintf(stream, "in_prefix_bos: %s # default: false\n", params.input_prefix_bos ? "true" : "false");
|
||||
yaml_dump_string_multiline(stream, "in_suffix", params.input_prefix.c_str());
|
||||
fprintf(stream, "interactive: %s # default: false\n", params.interactive ? "true" : "false");
|
||||
fprintf(stream, "interactive_first: %s # default: false\n", params.interactive_first ? "true" : "false");
|
||||
fprintf(stream, "keep: %d # default: 0\n", params.n_keep);
|
||||
fprintf(stream, "logdir: %s # default: unset (no logging)\n", params.logdir.c_str());
|
||||
|
||||
fprintf(stream, "logit_bias:\n");
|
||||
for (const auto & logit_bias : sparams.logit_bias) {
|
||||
fprintf(stream, " %d: %f", logit_bias.token, logit_bias.bias);
|
||||
}
|
||||
|
||||
fprintf(stream, "lora:\n");
|
||||
for (auto & la : params.lora_adapters) {
|
||||
if (la.scale == 1.0f) {
|
||||
fprintf(stream, " - %s\n", la.path.c_str());
|
||||
}
|
||||
}
|
||||
fprintf(stream, "lora_scaled:\n");
|
||||
for (auto & la : params.lora_adapters) {
|
||||
if (la.scale != 1.0f) {
|
||||
fprintf(stream, " - %s: %f\n", la.path.c_str(), la.scale);
|
||||
}
|
||||
}
|
||||
fprintf(stream, "lora_init_without_apply: %s # default: false\n", params.lora_init_without_apply ? "true" : "false");
|
||||
fprintf(stream, "main_gpu: %d # default: 0\n", params.main_gpu);
|
||||
fprintf(stream, "min_keep: %d # default: 0 (disabled)\n", sparams.min_keep);
|
||||
fprintf(stream, "mirostat: %d # default: 0 (disabled)\n", sparams.mirostat);
|
||||
fprintf(stream, "mirostat_ent: %f # default: 5.0\n", sparams.mirostat_tau);
|
||||
fprintf(stream, "mirostat_lr: %f # default: 0.1\n", sparams.mirostat_eta);
|
||||
fprintf(stream, "mlock: %s # default: false\n", params.use_mlock ? "true" : "false");
|
||||
fprintf(stream, "model: %s # default: %s\n", params.model.c_str(), DEFAULT_MODEL_PATH);
|
||||
fprintf(stream, "model_draft: %s # default:\n", params.model_draft.c_str());
|
||||
fprintf(stream, "multiline_input: %s # default: false\n", params.multiline_input ? "true" : "false");
|
||||
fprintf(stream, "n_gpu_layers: %d # default: -1\n", params.n_gpu_layers);
|
||||
fprintf(stream, "n_predict: %d # default: -1 (unlimited)\n", params.n_predict);
|
||||
fprintf(stream, "n_probs: %d # only used by server binary, default: 0\n", sparams.n_probs);
|
||||
fprintf(stream, "no_mmap: %s # default: false\n", !params.use_mmap ? "true" : "false");
|
||||
fprintf(stream, "penalize_nl: %s # default: false\n", sparams.penalize_nl ? "true" : "false");
|
||||
fprintf(stream, "ppl_output_type: %d # default: 0\n", params.ppl_output_type);
|
||||
fprintf(stream, "ppl_stride: %d # default: 0\n", params.ppl_stride);
|
||||
fprintf(stream, "presence_penalty: %f # default: 0.0\n", sparams.penalty_present);
|
||||
yaml_dump_string_multiline(stream, "prompt", params.prompt.c_str());
|
||||
fprintf(stream, "prompt_cache: %s\n", params.path_prompt_cache.c_str());
|
||||
fprintf(stream, "prompt_cache_all: %s # default: false\n", params.prompt_cache_all ? "true" : "false");
|
||||
fprintf(stream, "prompt_cache_ro: %s # default: false\n", params.prompt_cache_ro ? "true" : "false");
|
||||
yaml_dump_vector_int(stream, "prompt_tokens", prompt_tokens);
|
||||
fprintf(stream, "repeat_penalty: %f # default: 1.1\n", sparams.penalty_repeat);
|
||||
|
||||
fprintf(stream, "reverse_prompt:\n");
|
||||
for (std::string ap : params.antiprompt) {
|
||||
size_t pos = 0;
|
||||
while ((pos = ap.find('\n', pos)) != std::string::npos) {
|
||||
ap.replace(pos, 1, "\\n");
|
||||
pos += 1;
|
||||
}
|
||||
|
||||
fprintf(stream, " - %s\n", ap.c_str());
|
||||
}
|
||||
|
||||
fprintf(stream, "rope_freq_base: %f # default: 10000.0\n", params.rope_freq_base);
|
||||
fprintf(stream, "rope_freq_scale: %f # default: 1.0\n", params.rope_freq_scale);
|
||||
fprintf(stream, "simple_io: %s # default: false\n", params.simple_io ? "true" : "false");
|
||||
fprintf(stream, "cont_batching: %s # default: false\n", params.cont_batching ? "true" : "false");
|
||||
fprintf(stream, "flash_attn: %s # default: false\n", params.flash_attn ? "true" : "false");
|
||||
fprintf(stream, "temp: %f # default: 0.8\n", sparams.temp);
|
||||
|
||||
const std::vector<float> tensor_split_vector(params.tensor_split, params.tensor_split + llama_max_devices());
|
||||
yaml_dump_vector_float(stream, "tensor_split", tensor_split_vector);
|
||||
|
||||
fprintf(stream, "threads: %d # default: %u\n", params.cpuparams.n_threads, std::thread::hardware_concurrency());
|
||||
fprintf(stream, "top_k: %d # default: 40\n", sparams.top_k);
|
||||
fprintf(stream, "top_p: %f # default: 0.95\n", sparams.top_p);
|
||||
fprintf(stream, "min_p: %f # default: 0.0\n", sparams.min_p);
|
||||
fprintf(stream, "xtc_probability: %f # default: 0.0\n", sparams.xtc_probability);
|
||||
fprintf(stream, "xtc_threshold: %f # default: 0.1\n", sparams.xtc_threshold);
|
||||
fprintf(stream, "typ_p: %f # default: 1.0\n", sparams.typ_p);
|
||||
fprintf(stream, "verbose_prompt: %s # default: false\n", params.verbose_prompt ? "true" : "false");
|
||||
fprintf(stream, "display_prompt: %s # default: true\n", params.display_prompt ? "true" : "false");
|
||||
}
|
||||
|
@ -209,7 +209,6 @@ struct common_params {
|
||||
std::string path_prompt_cache = ""; // path to file for saving/loading prompt eval state // NOLINT
|
||||
std::string input_prefix = ""; // string to prefix user inputs with // NOLINT
|
||||
std::string input_suffix = ""; // string to suffix user inputs with // NOLINT
|
||||
std::string logdir = ""; // directory in which to save YAML log files // NOLINT
|
||||
std::string lookup_cache_static = ""; // path of static ngram cache file for lookup decoding // NOLINT
|
||||
std::string lookup_cache_dynamic = ""; // path of dynamic ngram cache file for lookup decoding // NOLINT
|
||||
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
|
||||
static const char * const LLM_KV_SPLIT_NO = "split.no";
|
||||
static const char * const LLM_KV_SPLIT_COUNT = "split.count";
|
||||
static const char * const LLM_KV_SPLIT_TENSORS_COUNT = "split.tensors.count";
|
||||
|
||||
//
|
||||
// YAML utils
|
||||
//
|
||||
|
||||
void yaml_dump_vector_float (FILE * stream, const char * prop_name, const std::vector<float> & data);
|
||||
void yaml_dump_vector_int (FILE * stream, const char * prop_name, const std::vector<int> & data);
|
||||
void yaml_dump_string_multiline(FILE * stream, const char * prop_name, const char * data);
|
||||
|
||||
void yaml_dump_non_result_info(
|
||||
FILE * stream, const common_params & params, const llama_context * lctx,
|
||||
const std::string & timestamp, const std::vector<int> & prompt_tokens, const char * model_desc);
|
||||
|
@ -43,50 +43,6 @@ static std::vector<llama_token> * g_output_tokens;
|
||||
|
||||
static bool is_interacting = false;
|
||||
|
||||
static void write_logfile(
|
||||
const llama_context * ctx, const common_params & params, const llama_model * model,
|
||||
const std::vector<llama_token> & input_tokens, const std::string & output,
|
||||
const std::vector<llama_token> & output_tokens
|
||||
) {
|
||||
if (params.logdir.empty()) {
|
||||
return;
|
||||
}
|
||||
|
||||
const std::string timestamp = string_get_sortable_timestamp();
|
||||
|
||||
const bool success = fs_create_directory_with_parents(params.logdir);
|
||||
if (!success) {
|
||||
LOG_ERR("%s: warning: failed to create logdir %s, cannot write logfile\n",
|
||||
__func__, params.logdir.c_str());
|
||||
return;
|
||||
}
|
||||
|
||||
const std::string logfile_path = params.logdir + timestamp + ".yml";
|
||||
FILE * logfile = fopen(logfile_path.c_str(), "w");
|
||||
|
||||
if (logfile == NULL) {
|
||||
LOG_ERR("%s: failed to open logfile %s\n", __func__, logfile_path.c_str());
|
||||
return;
|
||||
}
|
||||
|
||||
fprintf(logfile, "binary: infill\n");
|
||||
char model_desc[128];
|
||||
llama_model_desc(model, model_desc, sizeof(model_desc));
|
||||
yaml_dump_non_result_info(logfile, params, ctx, timestamp, input_tokens, model_desc);
|
||||
|
||||
fprintf(logfile, "\n");
|
||||
fprintf(logfile, "######################\n");
|
||||
fprintf(logfile, "# Generation Results #\n");
|
||||
fprintf(logfile, "######################\n");
|
||||
fprintf(logfile, "\n");
|
||||
|
||||
yaml_dump_string_multiline(logfile, "output", output.c_str());
|
||||
yaml_dump_vector_int(logfile, "output_tokens", output_tokens);
|
||||
|
||||
llama_perf_dump_yaml(logfile, ctx);
|
||||
fclose(logfile);
|
||||
}
|
||||
|
||||
#if defined (__unix__) || (defined (__APPLE__) && defined (__MACH__)) || defined (_WIN32)
|
||||
static void sigint_handler(int signo) {
|
||||
if (signo == SIGINT) {
|
||||
@ -96,7 +52,6 @@ static void sigint_handler(int signo) {
|
||||
console::cleanup();
|
||||
LOG("\n");
|
||||
common_perf_print(*g_ctx, *g_smpl);
|
||||
write_logfile(*g_ctx, *g_params, *g_model, *g_input_tokens, g_output_ss->str(), *g_output_tokens);
|
||||
|
||||
// make sure all logs are flushed
|
||||
LOG("Interrupted by user\n");
|
||||
@ -625,7 +580,6 @@ int main(int argc, char ** argv) {
|
||||
|
||||
LOG("\n");
|
||||
common_perf_print(ctx, smpl);
|
||||
write_logfile(ctx, params, model, input_tokens, output_ss.str(), output_tokens);
|
||||
|
||||
llama_free(ctx);
|
||||
llama_free_model(model);
|
||||
|
@ -62,49 +62,6 @@ static bool file_is_empty(const std::string & path) {
|
||||
return f.tellg() == 0;
|
||||
}
|
||||
|
||||
static void write_logfile(
|
||||
const llama_context * ctx, const common_params & params, const llama_model * model,
|
||||
const std::vector<llama_token> & input_tokens, const std::string & output,
|
||||
const std::vector<llama_token> & output_tokens
|
||||
) {
|
||||
if (params.logdir.empty()) {
|
||||
return;
|
||||
}
|
||||
|
||||
const std::string timestamp = string_get_sortable_timestamp();
|
||||
|
||||
const bool success = fs_create_directory_with_parents(params.logdir);
|
||||
if (!success) {
|
||||
LOG_ERR("%s: failed to create logdir %s, cannot write logfile\n", __func__, params.logdir.c_str());
|
||||
return;
|
||||
}
|
||||
|
||||
const std::string logfile_path = params.logdir + timestamp + ".yml";
|
||||
FILE * logfile = fopen(logfile_path.c_str(), "w");
|
||||
|
||||
if (logfile == NULL) {
|
||||
LOG_ERR("%s: failed to open logfile %s\n", __func__, logfile_path.c_str());
|
||||
return;
|
||||
}
|
||||
|
||||
fprintf(logfile, "binary: main\n");
|
||||
char model_desc[128];
|
||||
llama_model_desc(model, model_desc, sizeof(model_desc));
|
||||
yaml_dump_non_result_info(logfile, params, ctx, timestamp, input_tokens, model_desc);
|
||||
|
||||
fprintf(logfile, "\n");
|
||||
fprintf(logfile, "######################\n");
|
||||
fprintf(logfile, "# Generation Results #\n");
|
||||
fprintf(logfile, "######################\n");
|
||||
fprintf(logfile, "\n");
|
||||
|
||||
yaml_dump_string_multiline(logfile, "output", output.c_str());
|
||||
yaml_dump_vector_int(logfile, "output_tokens", output_tokens);
|
||||
|
||||
llama_perf_dump_yaml(logfile, ctx);
|
||||
fclose(logfile);
|
||||
}
|
||||
|
||||
#if defined (__unix__) || (defined (__APPLE__) && defined (__MACH__)) || defined (_WIN32)
|
||||
static void sigint_handler(int signo) {
|
||||
if (signo == SIGINT) {
|
||||
@ -115,7 +72,6 @@ static void sigint_handler(int signo) {
|
||||
console::cleanup();
|
||||
LOG("\n");
|
||||
common_perf_print(*g_ctx, *g_smpl);
|
||||
write_logfile(*g_ctx, *g_params, *g_model, *g_input_tokens, g_output_ss->str(), *g_output_tokens);
|
||||
|
||||
// make sure all logs are flushed
|
||||
LOG("Interrupted by user\n");
|
||||
@ -926,7 +882,6 @@ int main(int argc, char ** argv) {
|
||||
|
||||
LOG("\n\n");
|
||||
common_perf_print(ctx, smpl);
|
||||
write_logfile(ctx, params, model, input_tokens, output_ss.str(), output_tokens);
|
||||
|
||||
common_sampler_free(smpl);
|
||||
|
||||
|
@ -34,55 +34,6 @@ struct results_log_softmax {
|
||||
float prob;
|
||||
};
|
||||
|
||||
static void write_logfile(
|
||||
const llama_context * ctx, const common_params & params, const llama_model * model,
|
||||
const struct results_perplexity & results
|
||||
) {
|
||||
if (params.logdir.empty()) {
|
||||
return;
|
||||
}
|
||||
|
||||
if (params.hellaswag) {
|
||||
LOG_WRN("%s: logging results is not implemented for HellaSwag. No files will be written.\n", __func__);
|
||||
return;
|
||||
}
|
||||
|
||||
const std::string timestamp = string_get_sortable_timestamp();
|
||||
|
||||
const bool success = fs_create_directory_with_parents(params.logdir);
|
||||
if (!success) {
|
||||
LOG_WRN("%s: failed to create logdir %s, cannot write logfile\n",
|
||||
__func__, params.logdir.c_str());
|
||||
return;
|
||||
}
|
||||
|
||||
const std::string logfile_path = params.logdir + timestamp + ".yml";
|
||||
FILE * logfile = fopen(logfile_path.c_str(), "w");
|
||||
|
||||
if (logfile == NULL) {
|
||||
LOG_ERR("%s: failed to open logfile %s\n", __func__, logfile_path.c_str());
|
||||
return;
|
||||
}
|
||||
|
||||
fprintf(logfile, "binary: main\n");
|
||||
char model_desc[128];
|
||||
llama_model_desc(model, model_desc, sizeof(model_desc));
|
||||
yaml_dump_non_result_info(logfile, params, ctx, timestamp, results.tokens, model_desc);
|
||||
|
||||
fprintf(logfile, "\n");
|
||||
fprintf(logfile, "######################\n");
|
||||
fprintf(logfile, "# Perplexity Results #\n");
|
||||
fprintf(logfile, "######################\n");
|
||||
fprintf(logfile, "\n");
|
||||
|
||||
yaml_dump_vector_float(logfile, "logits", results.logits);
|
||||
fprintf(logfile, "ppl_value: %f\n", results.ppl_value);
|
||||
yaml_dump_vector_float(logfile, "probs", results.probs);
|
||||
|
||||
llama_perf_dump_yaml(logfile, ctx);
|
||||
fclose(logfile);
|
||||
}
|
||||
|
||||
static std::vector<float> softmax(const std::vector<float>& logits) {
|
||||
std::vector<float> probs(logits.size());
|
||||
float max_logit = logits[0];
|
||||
@ -2072,8 +2023,6 @@ int main(int argc, char ** argv) {
|
||||
LOG("\n");
|
||||
llama_perf_context_print(ctx);
|
||||
|
||||
write_logfile(ctx, params, model, results);
|
||||
|
||||
llama_free(ctx);
|
||||
llama_free_model(model);
|
||||
|
||||
|
@ -85,7 +85,6 @@ The project is under active development, and we are [looking for feedback and co
|
||||
| `-hfr, --hf-repo REPO` | Hugging Face model repository (default: unused)<br/>(env: LLAMA_ARG_HF_REPO) |
|
||||
| `-hff, --hf-file FILE` | Hugging Face model file (default: unused)<br/>(env: LLAMA_ARG_HF_FILE) |
|
||||
| `-hft, --hf-token TOKEN` | Hugging Face access token (default: value from HF_TOKEN environment variable)<br/>(env: HF_TOKEN) |
|
||||
| `-ld, --logdir LOGDIR` | path under which to save YAML logs (no logging if unset) |
|
||||
| `--log-disable` | Log disable |
|
||||
| `--log-file FNAME` | Log to file |
|
||||
| `--log-colors` | Enable colored logging<br/>(env: LLAMA_LOG_COLORS) |
|
||||
|
@ -1244,8 +1244,6 @@ extern "C" {
|
||||
LLAMA_API void llama_perf_sampler_print(const struct llama_sampler * chain);
|
||||
LLAMA_API void llama_perf_sampler_reset( struct llama_sampler * chain);
|
||||
|
||||
LLAMA_API void llama_perf_dump_yaml(FILE * stream, const struct llama_context * ctx);
|
||||
|
||||
#ifdef __cplusplus
|
||||
}
|
||||
#endif
|
||||
|
@ -1,146 +0,0 @@
|
||||
#!/usr/bin/env python3
|
||||
|
||||
import logging
|
||||
import argparse
|
||||
import os
|
||||
import subprocess
|
||||
import sys
|
||||
|
||||
import yaml
|
||||
|
||||
logger = logging.getLogger("run-with-preset")
|
||||
|
||||
CLI_ARGS_LLAMA_CLI_PERPLEXITY = [
|
||||
"batch-size", "cfg-negative-prompt", "cfg-scale", "chunks", "color", "ctx-size", "escape",
|
||||
"export", "file", "frequency-penalty", "grammar", "grammar-file", "hellaswag",
|
||||
"hellaswag-tasks", "ignore-eos", "in-prefix", "in-prefix-bos", "in-suffix",
|
||||
"interactive", "interactive-first", "keep", "logdir", "logit-bias", "lora", "lora-base",
|
||||
"low-vram", "main-gpu", "mirostat", "mirostat-ent", "mirostat-lr", "mlock",
|
||||
"model", "multiline-input", "n-gpu-layers", "n-predict", "no-mmap", "no-mul-mat-q",
|
||||
"np-penalize-nl", "numa", "ppl-output-type", "ppl-stride", "presence-penalty", "prompt",
|
||||
"prompt-cache", "prompt-cache-all", "prompt-cache-ro", "repeat-last-n",
|
||||
"repeat-penalty", "reverse-prompt", "rope-freq-base", "rope-freq-scale", "rope-scale", "seed",
|
||||
"simple-io", "tensor-split", "threads", "temp", "top-k", "top-p", "typical",
|
||||
"verbose-prompt"
|
||||
]
|
||||
|
||||
CLI_ARGS_LLAMA_BENCH = [
|
||||
"batch-size", "low-vram", "model", "mul-mat-q", "n-gen", "n-gpu-layers",
|
||||
"n-prompt", "output", "repetitions", "tensor-split", "threads", "verbose"
|
||||
]
|
||||
|
||||
CLI_ARGS_LLAMA_SERVER = [
|
||||
"alias", "batch-size", "ctx-size", "embedding", "host", "lora", "lora-base",
|
||||
"low-vram", "main-gpu", "mlock", "model", "n-gpu-layers", "n-probs", "no-mmap", "no-mul-mat-q",
|
||||
"numa", "path", "port", "rope-freq-base", "timeout", "rope-freq-scale", "tensor-split",
|
||||
"threads", "verbose"
|
||||
]
|
||||
|
||||
description = """Run llama.cpp binaries with presets from YAML file(s).
|
||||
To specify which binary should be run, specify the "binary" property (llama-cli, llama-perplexity, llama-bench, and llama-server are supported).
|
||||
To get a preset file template, run a llama.cpp binary with the "--logdir" CLI argument.
|
||||
|
||||
Formatting considerations:
|
||||
- The YAML property names are the same as the CLI argument names of the corresponding binary.
|
||||
- Properties must use the long name of their corresponding llama.cpp CLI arguments.
|
||||
- Like the llama.cpp binaries the property names do not differentiate between hyphens and underscores.
|
||||
- Flags must be defined as "<PROPERTY_NAME>: true" to be effective.
|
||||
- To define the logit_bias property, the expected format is "<TOKEN_ID>: <BIAS>" in the "logit_bias" namespace.
|
||||
- To define multiple "reverse_prompt" properties simultaneously the expected format is a list of strings.
|
||||
- To define a tensor split, pass a list of floats.
|
||||
"""
|
||||
usage = "run-with-preset.py [-h] [yaml_files ...] [--<ARG_NAME> <ARG_VALUE> ...]"
|
||||
epilog = (" --<ARG_NAME> specify additional CLI ars to be passed to the binary (override all preset files). "
|
||||
"Unknown args will be ignored.")
|
||||
|
||||
parser = argparse.ArgumentParser(
|
||||
description=description, usage=usage, epilog=epilog, formatter_class=argparse.RawTextHelpFormatter)
|
||||
parser.add_argument("-bin", "--binary", help="The binary to run.")
|
||||
parser.add_argument("yaml_files", nargs="*",
|
||||
help="Arbitrary number of YAML files from which to read preset values. "
|
||||
"If two files specify the same values the later one will be used.")
|
||||
parser.add_argument("--verbose", action="store_true", help="increase output verbosity")
|
||||
|
||||
known_args, unknown_args = parser.parse_known_args()
|
||||
|
||||
if not known_args.yaml_files and not unknown_args:
|
||||
parser.print_help()
|
||||
sys.exit(0)
|
||||
|
||||
logging.basicConfig(level=logging.DEBUG if known_args.verbose else logging.INFO)
|
||||
|
||||
props = dict()
|
||||
|
||||
for yaml_file in known_args.yaml_files:
|
||||
with open(yaml_file, "r") as f:
|
||||
props.update(yaml.load(f, yaml.SafeLoader))
|
||||
|
||||
props = {prop.replace("_", "-"): val for prop, val in props.items()}
|
||||
|
||||
binary = props.pop("binary", "llama-cli")
|
||||
if known_args.binary:
|
||||
binary = known_args.binary
|
||||
|
||||
if os.path.exists(f"./{binary}"):
|
||||
binary = f"./{binary}"
|
||||
|
||||
if binary.lower().endswith("llama-cli") or binary.lower().endswith("llama-perplexity"):
|
||||
cli_args = CLI_ARGS_LLAMA_CLI_PERPLEXITY
|
||||
elif binary.lower().endswith("llama-bench"):
|
||||
cli_args = CLI_ARGS_LLAMA_BENCH
|
||||
elif binary.lower().endswith("llama-server"):
|
||||
cli_args = CLI_ARGS_LLAMA_SERVER
|
||||
else:
|
||||
logger.error(f"Unknown binary: {binary}")
|
||||
sys.exit(1)
|
||||
|
||||
command_list = [binary]
|
||||
|
||||
for cli_arg in cli_args:
|
||||
value = props.pop(cli_arg, None)
|
||||
|
||||
if not value or value == -1:
|
||||
continue
|
||||
|
||||
if cli_arg == "logit-bias":
|
||||
for token, bias in value.items():
|
||||
command_list.append("--logit-bias")
|
||||
command_list.append(f"{token}{bias:+}")
|
||||
continue
|
||||
|
||||
if cli_arg == "reverse-prompt" and not isinstance(value, str):
|
||||
for rp in value:
|
||||
command_list.append("--reverse-prompt")
|
||||
command_list.append(str(rp))
|
||||
continue
|
||||
|
||||
command_list.append(f"--{cli_arg}")
|
||||
|
||||
if cli_arg == "tensor-split":
|
||||
command_list.append(",".join([str(v) for v in value]))
|
||||
continue
|
||||
|
||||
value = str(value)
|
||||
|
||||
if value != "True":
|
||||
command_list.append(str(value))
|
||||
|
||||
num_unused = len(props)
|
||||
if num_unused > 10:
|
||||
logger.info(f"The preset file contained a total of {num_unused} unused properties.")
|
||||
elif num_unused > 0:
|
||||
logger.info("The preset file contained the following unused properties:")
|
||||
for prop, value in props.items():
|
||||
logger.info(f" {prop}: {value}")
|
||||
|
||||
command_list += unknown_args
|
||||
|
||||
sp = subprocess.Popen(command_list)
|
||||
|
||||
while sp.returncode is None:
|
||||
try:
|
||||
sp.wait()
|
||||
except KeyboardInterrupt:
|
||||
pass
|
||||
|
||||
sys.exit(sp.returncode)
|
@ -22075,28 +22075,6 @@ void llama_perf_context_reset(struct llama_context * ctx) {
|
||||
ctx->t_p_eval_us = ctx->n_p_eval = 0;
|
||||
}
|
||||
|
||||
void llama_perf_dump_yaml(FILE * stream, const llama_context * ctx) {
|
||||
fprintf(stream, "\n");
|
||||
fprintf(stream, "###########\n");
|
||||
fprintf(stream, "# Timings #\n");
|
||||
fprintf(stream, "###########\n");
|
||||
fprintf(stream, "\n");
|
||||
|
||||
fprintf(stream, "mst_eval: %.2f # ms / token during generation\n",
|
||||
1.0e-3 * ctx->t_eval_us / ctx->n_eval);
|
||||
fprintf(stream, "mst_p_eval: %.2f # ms / token during prompt processing\n",
|
||||
1.0e-3 * ctx->t_p_eval_us / ctx->n_p_eval);
|
||||
fprintf(stream, "n_eval: %d # number of tokens generated (excluding the first one)\n", ctx->n_eval);
|
||||
fprintf(stream, "n_p_eval: %d # number of tokens processed in batches at the beginning\n", ctx->n_p_eval);
|
||||
fprintf(stream, "t_eval_us: %" PRId64 " # total microseconds spent generating tokens\n", ctx->t_eval_us);
|
||||
fprintf(stream, "t_load_us: %" PRId64 " # total microseconds spent loading the model\n", ctx->t_load_us);
|
||||
fprintf(stream, "t_p_eval_us: %" PRId64 " # total microseconds spent prompt processing\n", ctx->t_p_eval_us);
|
||||
fprintf(stream, "ts_eval: %.2f # tokens / second during generation\n",
|
||||
1.0e6 * ctx->n_eval / ctx->t_eval_us);
|
||||
fprintf(stream, "ts_p_eval: %.2f # tokens / second during prompt processing\n",
|
||||
1.0e6 * ctx->n_p_eval / ctx->t_p_eval_us);
|
||||
}
|
||||
|
||||
// For internal test use
|
||||
const std::vector<std::pair<std::string, struct ggml_tensor *>> & llama_internal_get_tensor_map(
|
||||
struct llama_context * ctx
|
||||
|
Loading…
Reference in New Issue
Block a user