Distribute and run LLMs with a single file.
C++ cross-platformggufllama-cpplocal-ailocal-inferencelocal-llmopen-source-aisingle-file-executablespeech-to-text
respect user NO_COLOR preference in localscore
CJ Pais81da07b34c7a
diff --git a/localscore/localscore.cpp b/localscore/localscore.cppindex 891ceb61..036d5250 100644--- a/localscore/localscore.cpp+++ b/localscore/localscore.cpp@@ -34,6 +34,7 @@#include "cmd.h"#include "benchmark.h"#include "printer.h"+#include "utils.h"#include "localscore.h"@@ -338,7 +339,7 @@ static LocalScoreResultsSummary getResultsSummary(Json data) {static void displayResults(LocalScoreResultsSummary results_summary, bool plaintext) {- printf("\n\033[1;35m");+ printf("\n%s", utils::color_str("\033[1;35m"));if (!plaintext) {ascii_display::print_logo();printf("\n");@@ -346,12 +347,33 @@ static void displayResults(LocalScoreResultsSummary results_summary, bool plaint} else {printf("LocalScore: \t\t %d", (int)results_summary.performance_score);}- printf("\033[0m\n");-- printf("\033[32mToken Generation: \t \033[1;32m%.2f\033[0m \033[3;32mtok/s\033[0m\n", results_summary.avg_gen_tps);- printf("\033[36mPrompt Processing: \t \033[1;36m%.2f\033[0m \033[3;36mtok/s\033[0m\n", results_summary.avg_prompt_tps);- printf("\033[33mTime to First Token:\t \033[1;33m%.2f\033[0m \033[3;33mms\033[0m\n", results_summary.avg_ttft_ms);- printf("\033[0m");+ printf("%s\n", utils::color_str("\033[0m"));++ printf("%sToken Generation: \t %s%.2f%s %stok/s%s\n",+ utils::color_str("\033[32m"),+ utils::color_str("\033[1;32m"),+ results_summary.avg_gen_tps,+ utils::color_str("\033[0m"),+ utils::color_str("\033[3;32m"),+ utils::color_str("\033[0m"));++ printf("%sPrompt Processing: \t %s%.2f%s %stok/s%s\n",+ utils::color_str("\033[36m"),+ utils::color_str("\033[1;36m"),+ results_summary.avg_prompt_tps,+ utils::color_str("\033[0m"),+ utils::color_str("\033[3;36m"),+ utils::color_str("\033[0m"));++ printf("%sTime to First Token:\t %s%.2f%s %sms%s\n",+ utils::color_str("\033[33m"),+ utils::color_str("\033[1;33m"),+ results_summary.avg_ttft_ms,+ utils::color_str("\033[0m"),+ utils::color_str("\033[3;33m"),+ utils::color_str("\033[0m"));++ printf("%s", utils::color_str("\033[0m"));}struct SystemData {diff --git a/localscore/system.cpp b/localscore/system.cppindex 2b0e040d..f23393a5 100644--- a/localscore/system.cpp+++ b/localscore/system.cpp@@ -158,12 +158,12 @@ void get_runtime_info(RuntimeInfo* info) {strncpy(info->llamafile_version, LLAMAFILE_VERSION_STRING, MAX_STRING_LENGTH - 1);strncpy(info->llama_commit, LLAMA_COMMIT, MAX_STRING_LENGTH - 1);- fprintf(stderr, "\033[0;35m\n"); // Sets purple color- utils::print_centered(stderr, 70, '=', "\033[1mLocalScore Runtime Information\033[0;35m");+ fprintf(stderr, "%s\n", utils::color_str("\033[0;35m")); // Sets purple color+ utils::print_centered(stderr, 70, '=', "%sLocalScore Runtime Information%s", utils::color_str("\033[1m"), utils::color_str("\033[0;35m"));fprintf(stderr, "\n");- fprintf(stderr, "%-20s \033[1m%s\033[22m\n", "llamafile version:", info->llamafile_version);+ fprintf(stderr, "%-20s %s%s%s\n", "llamafile version:", utils::color_str("\033[1m"), info->llamafile_version, utils::color_str("\033[22m"));fprintf(stderr, "%-20s %s\n", "llama.cpp commit:", info->llama_commit);- fprintf(stderr, "\n======================================================================\n\n\033[0m");+ fprintf(stderr, "\n======================================================================\n\n%s", utils::color_str("\033[0m"));}double get_mem_gb() {@@ -194,7 +194,7 @@ void get_sys_info(SystemInfo* info) {info->ram_gb = get_mem_gb();- utils::print_centered(stderr, 70, '=', "\033[1mSystem Information\033[0m");+ utils::print_centered(stderr, 70, '=', "%sSystem Information%s", utils::color_str("\033[1m"), utils::color_str("\033[0m"));fprintf(stderr, "\n");fprintf(stderr, "%-20s %s\n", "Kernel Type:", info->kernel_type);fprintf(stderr, "%-20s %s\n", "Kernel Release:", info->kernel_release);@@ -232,11 +232,11 @@ void get_accelerator_info(AcceleratorInfo* info, cmd_params * params) {}if (i == params->main_gpu) {- fprintf(stderr, "\033[0;32m"); // Sets green color- utils::print_centered(stderr, 70, '=', "\033[1mActive GPU (GPU %d) Information\033[0;32m", i);+ fprintf(stderr, "%s", utils::color_str("\033[0;32m")); // Sets green color+ utils::print_centered(stderr, 70, '=', "%sActive GPU (GPU %d) Information%s", utils::color_str("\033[1m"), i, utils::color_str("\033[0;32m"));fprintf(stderr, "\n");} else {- fprintf(stderr, "\033[0;90m"); // Sets gray color+ fprintf(stderr, "%s", utils::color_str("\033[0;90m")); // Sets gray colorutils::print_centered(stderr, 70, '=', "GPU %d Information", i);fprintf(stderr, "\n");}@@ -245,7 +245,7 @@ void get_accelerator_info(AcceleratorInfo* info, cmd_params * params) {fprintf(stderr, "%-26s %.1f GiB\n", "VRAM:", rounded_memory_gb);fprintf(stderr, "%-26s %d\n", "Streaming Multiprocessors:", props.multiProcessorCount);fprintf(stderr, "%-26s %.1f\n", "CUDA Capability:", atof(props.compute));- fprintf(stderr, "\n======================================================================\n\n\033[0m");+ fprintf(stderr, "\n======================================================================\n\n%s", utils::color_str("\033[0m"));}}@@ -276,14 +276,14 @@ void get_accelerator_info(AcceleratorInfo* info, cmd_params * params) {strncpy(info->manufacturer, "Apple", MAX_STRING_LENGTH - 1);- fprintf(stderr, "\033[0;32m===== GPU information =====\n\n");+ fprintf(stderr, "%s===== GPU information =====\n\n", utils::color_str("\033[0;32m"));fprintf(stderr, "%-26s %s\n", "GPU Name:", props.name);fprintf(stderr, "%-26s %.1f GiB\n", "VRAM:", info->total_memory_gb);fprintf(stderr, "%-26s %d\n", "Core Count:", props.core_count);fprintf(stderr, "%-26s %d\n", "Metal Version:", props.metal_version);fprintf(stderr, "%-26s %d\n", "GPU Family:", props.gpu_family);fprintf(stderr, "%-26s %d\n", "Common GPU Family:", props.gpu_family_common);- fprintf(stderr, "\n======================================================================\n\n\033[0m");+ fprintf(stderr, "\n======================================================================\n\n%s", utils::color_str("\033[0m"));}} else {#ifdef __x86_64__@@ -306,7 +306,7 @@ void list_available_accelerators() {fprintf(stderr, "Apple Metal\n");} else if (llamafile_has_cuda()) {int count = ggml_backend_cuda_get_device_count();- fprintf(stderr, "\n\033[0;32m==================== Available GPUs ====================\n\n");+ fprintf(stderr, "\n%s==================== Available GPUs ====================\n\n", utils::color_str("\033[0;32m"));for (int i = 0; i < count; i++) {struct ggml_cuda_device_properties props;ggml_backend_cuda_get_device_properties(i, &props);@@ -318,7 +318,7 @@ void list_available_accelerators() {} else {fprintf(stderr, "No Accelerator support available\n");}- fprintf(stderr, "\n======================================================================\n\033[0m");+ fprintf(stderr, "\n======================================================================\n%s", utils::color_str("\033[0m"));}void get_model_info(ModelInfo *info, llama_model *model) {diff --git a/localscore/utils.h b/localscore/utils.hindex f28b3de6..2d365777 100644--- a/localscore/utils.h+++ b/localscore/utils.h@@ -11,6 +11,7 @@#include <memory>#include <array>#include <cstdio>+#include <cstring>#include <stdexcept>#include <string.h>#include <stdarg.h>@@ -185,6 +186,20 @@ namespace utils {return round(value * multiplier) / multiplier;}+ inline bool should_use_color() {+ const char* no_color = getenv("NO_COLOR");+ if (no_color != NULL && no_color[0] != '\0') {+ return false;+ }+ return true;+ }++ inline const char* color_str(const char* color_code) {+ if (should_use_color()) {+ return color_code;+ }+ return "";+ }} // namespace utils