log : cleanup, comments, build flags

ggml-ci

Makefile

@@ -149,6 +149,14 @@ GGML_NO_METAL := 1
 DEPRECATE_WARNING := 1
 endif
 
+ifdef LLAMA_DISABLE_LOGS
+REMOVE_WARNING := 1
+endif
+
+ifdef LLAMA_SERVER_VERBOSE
+REMOVE_WARNING := 1
+endif
+
 ifndef UNAME_S
 UNAME_S := $(shell uname -s)
 endif
@@ -352,19 +360,11 @@ ifdef LLAMA_SANITIZE_UNDEFINED
 	MK_LDFLAGS  += -fsanitize=undefined -g
 endif
 
-ifdef LLAMA_SERVER_VERBOSE
-	MK_CPPFLAGS += -DSERVER_VERBOSE=$(LLAMA_SERVER_VERBOSE)
-endif
-
 ifdef LLAMA_SERVER_SSL
 	MK_CPPFLAGS += -DCPPHTTPLIB_OPENSSL_SUPPORT
 	MK_LDFLAGS += -lssl -lcrypto
 endif
 
-ifdef LLAMA_DISABLE_LOGS
-	MK_CPPFLAGS += -DLOG_DISABLE_LOGS
-endif # LLAMA_DISABLE_LOGS
-
 # warnings
 WARN_FLAGS = \
 	-Wall \
@@ -1029,6 +1029,14 @@ $(info   - LLAMA_NO_CCACHE)
 $(info )
 endif
 
+ifdef REMOVE_WARNING
+$(info !!! REMOVAL WARNING !!!)
+$(info The following LLAMA_ options have been removed and are no longer supported)
+$(info   - LLAMA_DISABLE_LOGS   (https://github.com/ggerganov/llama.cpp/pull/9418))
+$(info   - LLAMA_SERVER_VERBOSE (https://github.com/ggerganov/llama.cpp/pull/9418))
+$(info )
+endif
+
 #
 # Build libraries
 #

common/common.h

@@ -341,7 +341,8 @@ struct gpt_params {
     bool batched_bench_output_jsonl = false;
 };
 
-// call once at the start of a program using common
+// call once at the start of a program if it uses libcommon
+// initializes the logging system and prints info about the build
 void gpt_init();
 
 std::string gpt_params_get_system_info(const gpt_params & params);

common/log.cpp

@@ -69,8 +69,8 @@ struct gpt_log_entry {
     void print(FILE * file = nullptr) const {
         FILE * fcur = file;
         if (!fcur) {
-            // stderr displays DBG messages only when the verbosity is high
+            // stderr displays DBG messages only when their verbosity level is not higher than the threshold
-            // these messages can still be logged to a file
+            // these messages will still be logged to a file
             if (level == GGML_LOG_LEVEL_DEBUG && gpt_log_verbosity_thold < LOG_DEFAULT_DEBUG) {
                 return;
             }

common/log.h

@@ -1,6 +1,6 @@
 #pragma once
 
-#include "ggml.h"
+#include "ggml.h" // for ggml_log_level
 
 #ifndef __GNUC__
 #    define LOG_ATTRIBUTE_FORMAT(...)
@@ -48,7 +48,10 @@ void gpt_log_add(struct gpt_log * log, enum ggml_log_level level, const char * f
 //   0.00.090.578 I llm_load_tensors: offloading 32 repeating layers to GPU
 //   0.00.090.579 I llm_load_tensors: offloading non-repeating layers to GPU
 //
-// I - info, W - warning, E - error, D - debug
+// I - info    (stdout, V = 0)
+// W - warning (stderr, V = 0)
+// E - error   (stderr, V = 0)
+// D - debug   (stderr, V = LOG_DEFAULT_DEBUG)
 //
 
 void gpt_log_set_file      (struct gpt_log * log, const char * file);       // not thread-safe
@@ -57,13 +60,13 @@ void gpt_log_set_prefix    (struct gpt_log * log,       bool   prefix);     // w
 void gpt_log_set_timestamps(struct gpt_log * log,       bool   timestamps); // whether to output timestamps in the prefix
 
 // helper macros for logging
-// use these to avoid computing log arguments if the verbosity is lower than the threshold
+// use these to avoid computing log arguments if the verbosity of the log is higher than the threshold
 //
 // for example:
 //
 //   LOG_DBG("this is a debug message: %d\n", expensive_function());
 //
-// this will avoid calling expensive_function() if the verbosity is lower than LOG_DEFAULT_DEBUG
+// this will avoid calling expensive_function() if LOG_DEFAULT_DEBUG > gpt_log_verbosity_thold
 //
 
 #define LOG_TMPL(level, verbosity, ...) \

examples/llava/clip.cpp

@@ -42,7 +42,7 @@
 #define LOG_INF(...) do { fprintf(stdout, __VA_ARGS__); } while (0)
 #define LOG_WRN(...) do { fprintf(stderr, __VA_ARGS__); } while (0)
 #define LOG_ERR(...) do { fprintf(stderr, __VA_ARGS__); } while (0)
-#define LOG_DBG(...) do { fprintf(stdout, __VA_ARGS__); } while (0)
+#define LOG_DBG(...) do { fprintf(stderr, __VA_ARGS__); } while (0)
 
 //#define CLIP_DEBUG_FUNCTIONS
 

examples/server/CMakeLists.txt

@@ -1,5 +1,5 @@
 set(TARGET llama-server)
-option(LLAMA_SERVER_VERBOSE "Build verbose logging option for Server" ON)
+
 option(LLAMA_SERVER_SSL "Build SSL support for the server" OFF)
 
 include_directories(${CMAKE_CURRENT_SOURCE_DIR} ${CMAKE_CURRENT_BINARY_DIR})
@@ -45,9 +45,6 @@ endforeach()
 
 add_executable(${TARGET} ${TARGET_SRCS})
 install(TARGETS ${TARGET} RUNTIME)
-target_compile_definitions(${TARGET} PRIVATE
-    SERVER_VERBOSE=$<BOOL:${LLAMA_SERVER_VERBOSE}>
-)
 
 target_link_libraries(${TARGET} PRIVATE common ${CMAKE_THREAD_LIBS_INIT})
 

examples/server/server.cpp

@@ -2077,13 +2077,6 @@ struct server_context {
                         slot_npast++;
                     }
 
-                    //LOG_VERBOSE("prompt processing progress", {
-                    //    {"id_slot",  slot.id},
-                    //    {"n_past",   slot.n_past},
-                    //    {"n_ctx",    n_ctx},
-                    //    {"n_tokens", batch.n_tokens},
-                    //    {"progress", (float) slot.n_prompt_tokens_processed / slot.n_prompt_tokens},
-                    //});
                     SLT_INF(slot, "prompt processing progress, n_past = %d, n_tokens = %d, progress = %f\n", slot.n_past, batch.n_tokens, (float) slot.n_prompt_tokens_processed / slot.n_prompt_tokens);
 
                     // entire prompt has been processed
@@ -2098,12 +2091,6 @@ struct server_context {
                         slot.n_decoded = 0;
                         slot.i_batch   = batch.n_tokens - 1;
 
-                        //LOG_VERBOSE("prompt done", {
-                        //    {"id_slot",  slot.id},
-                        //    {"n_past",   slot.n_past},
-                        //    {"n_ctx",    n_ctx},
-                        //    {"n_tokens", batch.n_tokens},
-                        //});
                         SLT_INF(slot, "prompt done, n_past = %d, n_tokens = %d\n", slot.n_past, batch.n_tokens);
                     }
                 }