build settings

.gitignore

@@ -34,6 +34,7 @@ models/*
 /embedding
 /benchmark-matmult
 /vdot
+/server
 /Pipfile
 /libllama.so
 

Makefile

@@ -3,6 +3,8 @@ BUILD_TARGETS = main quantize quantize-stats perplexity embedding vdot
 
 ifdef LLAMA_BUILD_SERVER
 	BUILD_TARGETS += server
+	LLAMA_SERVER_VERBOSE ?= 1
+server: CXXFLAGS += -DSERVER_VERBOSE=$(LLAMA_SERVER_VERBOSE)
 endif
 
 default: $(BUILD_TARGETS)

examples/server/CMakeLists.txt

@@ -1,4 +1,5 @@
 set(TARGET server)
+option(LLAMA_SERVER_VERBOSE "Build verbose logging option for Server" ON)
 include_directories(${CMAKE_CURRENT_SOURCE_DIR})
 add_executable(${TARGET} server.cpp json.hpp httplib.h)
 target_compile_definitions(${TARGET} PRIVATE
@@ -8,6 +9,8 @@ target_compile_definitions(${TARGET} PRIVATE
     $<$<CONFIG:Debug>:
         CPPHTTPLIB_NO_EXCEPTIONS=1
     >
+
+    SERVER_VERBOSE=$<BOOL:${LLAMA_SERVER_VERBOSE}>
 )
 target_link_libraries(${TARGET} PRIVATE common llama ${CMAKE_THREAD_LIBS_INIT})
 target_compile_features(${TARGET} PRIVATE cxx_std_11)

examples/server/server.cpp

@@ -5,6 +5,10 @@
 #include "httplib.h"
 #include "json.hpp"
 
+#ifndef SERVER_VERBOSE
+#define SERVER_VERBOSE 1
+#endif
+
 using namespace httplib;
 using json = nlohmann::json;
 
@@ -553,7 +557,7 @@ void server_params_parse(int argc, char ** argv, server_params & sparams,
             params.lora_base = argv[i];
         } else if (arg == "-v" || arg == "--verbose") {
             sparams.verbose = true;
-#ifndef SERVER_VERBOSE
+#if SERVER_VERBOSE != 1
             LOG_WARNING("server.cpp is not built with verbose logging.", {});
 #endif
         } else if (arg == "--mlock") {
@@ -743,7 +747,7 @@ int main(int argc, char ** argv) {
 
     server_params_parse(argc, argv, sparams, params);
 
-#ifdef SERVER_VERBOSE
+#if SERVER_VERBOSE == 1
     server_verbose = sparams.verbose;
 #endif