Embed man page into --help flag of each program

See #166

Makefile

@@ -22,7 +22,7 @@ o/$(MODE)/: o/$(MODE)/llama.cpp o/$(MODE)/llamafile
 # for installing to `make PREFIX=/usr/local`
 .PHONY: install
 install:	llamafile/zipalign.1					\
-		llama.cpp/main/main.1					\
+		llamafile/llamafile.1					\
 		llama.cpp/quantize/quantize.1				\
 		llama.cpp/perplexity/perplexity.1			\
 		llama.cpp/llava/llava-quantize.1			\
@@ -40,7 +40,7 @@ install:	llamafile/zipalign.1					\
 	$(INSTALL) o/$(MODE)/llama.cpp/llava/llava-quantize $(PREFIX)/bin/llava-quantize
 	mkdir -p $(PREFIX)/share/man/man1
 	$(INSTALL) -m 0644 llamafile/zipalign.1 $(PREFIX)/share/man/man1/zipalign.1
-	$(INSTALL) -m 0644 llama.cpp/main/main.1 $(PREFIX)/share/man/man1/llamafile.1
+	$(INSTALL) -m 0644 llamafile/llamafile.1 $(PREFIX)/share/man/man1/llamafile.1
 	$(INSTALL) -m 0644 llama.cpp/quantize/quantize.1 $(PREFIX)/share/man/man1/llamafile-quantize.1
 	$(INSTALL) -m 0644 llama.cpp/perplexity/perplexity.1 $(PREFIX)/share/man/man1/llamafile-perplexity.1
 	$(INSTALL) -m 0644 llama.cpp/llava/llava-quantize.1 $(PREFIX)/share/man/man1/llava-quantize.1

build/rules.mk

@@ -23,6 +23,11 @@ o/$(MODE)/%.o: %.cpp $(COSMOCC)
 o/$(MODE)/%: o/$(MODE)/%.o
 	$(LINK.o) $^ $(LOADLIBES) $(LDLIBS) -o $@
 
+.PRECIOUS: %.1.asc
+%.1.asc: %.1
+	-man $< >$@.tmp && mv -f $@.tmp $@
+	@rm -f $@.tmp
+
 o/$(MODE)/%.zip.o: % $(COSMOCC)
 	@mkdir -p $(dir $@)/.aarch64
 	$(ZIPOBJ) $(ZIPOBJ_FLAGS) -a x86_64 -o $@ $<

llama.cpp/llava/BUILD.mk

@@ -10,15 +10,16 @@ LLAMA_CPP_LLAVA_OBJS = $(LLAMA_CPP_LLAVA_SRCS:%.cpp=o/$(MODE)/%.o)
 
 .PHONY: tool/args/args.h
 
-o/$(MODE)/llama.cpp/llava/llava.a:				\
+o/$(MODE)/llama.cpp/llava/llava.a:					\
 		$(LLAMA_CPP_LLAVA_OBJS)
 
-o/$(MODE)/llama.cpp/llava/llava-quantize:			\
-		o/$(MODE)/llama.cpp/llava/llava-quantize.o	\
-		o/$(MODE)/llama.cpp/llava/llava.a		\
+o/$(MODE)/llama.cpp/llava/llava-quantize:				\
+		o/$(MODE)/llama.cpp/llava/llava-quantize.o		\
+		o/$(MODE)/llama.cpp/llava/llava-quantize.1.asc.zip.o	\
+		o/$(MODE)/llama.cpp/llava/llava.a			\
 		o/$(MODE)/llama.cpp/llama.cpp.a
 
 .PHONY: o/$(MODE)/llama.cpp/llava
-o/$(MODE)/llama.cpp/llava:					\
-		o/$(MODE)/llama.cpp/llava/llava.a		\
+o/$(MODE)/llama.cpp/llava:						\
+		o/$(MODE)/llama.cpp/llava/llava.a			\
 		o/$(MODE)/llama.cpp/llava/llava-quantize

llama.cpp/llava/llava-quantize.1

@@ -1,6 +1,6 @@
 .Dd December 5, 2023
-.Dt llava-quantize 1
-.Os
+.Dt LLAVA-QUANTIZE 1
+.Os Llamafile Manual
 .Sh NAME
 .Nm llava-quantize
 .Nd CLIP model quantizer

llama.cpp/llava/llava-quantize.1.asc

@@ -0,0 +1,46 @@
+LLAVA-QUANTIZE(1)           General Commands Manual          LLAVA-QUANTIZE(1)
+
+NNAAMMEE
+     llllaavvaa--qquuaannttiizzee - CLIP model quantizer
+
+SSYYNNOOPPSSIISS
+     llllaavvaa--qquuaannttiizzee [options...] _I_N_P_U_T _O_U_T_P_U_T _F_O_R_M_A_T
+
+DDEESSCCRRIIPPTTIIOONN
+     llllaavvaa--qquuaannttiizzee makes LLaVA mmproj files smaller.
+
+AARRGGUUMMEENNTTSS
+     The following positional arguments are accepted:
+
+     _I_N_P_U_T   Is the input file, which should be a CLIP model in the GGUF
+             format using float16 values.
+
+     _O_U_T_P_U_T  Is the output file, which will be a CLIP model in the GGUF format
+             using the desired number type.
+
+     _F_O_R_M_A_T  Is the desired quantization format, which may be the integer id
+             of a supported quantization type. See the quantization types
+             section below for acceptable formats.
+
+OOPPTTIIOONNSS
+     The following options are accepted:
+
+     --hh, ----hheellpp
+             Show help message and exit.
+
+     ----vveerrssiioonn
+             Print llamafile version.
+
+QQUUAANNTTIIZZAATTIIOONN TTYYPPEESS
+     The following quantization types are available:
+
+     --   2 is Q4_0
+     --   3 is Q4_1
+     --   6 is Q5_0
+     --   7 is Q5_1
+     --   8 is Q8_0
+
+SSEEEE AALLSSOO
+     llamafile(1)
+
+Llamafile Manual               December 5, 2023               Llamafile Manual

llama.cpp/llava/llava-quantize.cpp

@@ -8,23 +8,27 @@
 #include <string.h>
 
 int main(int argc, char *argv[]) {
-    if (argc == 2 && !strcmp(argv[1], "--version")) {
-        printf("llamafile v" LLAMAFILE_VERSION_STRING " llava-quantize\n");
-        exit(0);
+
+    if (llamafile_has(argv, "--version")) {
+        puts("llava-quantize v" LLAMAFILE_VERSION_STRING);
+        return 0;
+    }
+
+    if (llamafile_has(argv, "-h") ||
+        llamafile_has(argv, "-help") ||
+        llamafile_has(argv, "--help")) {
+        llamafile_help("/zip/llama.cpp/llava/llava-quantize.1.asc");
+        __builtin_unreachable();
     }
+
     llamafile_init();
     llamafile_check_cpu();
+
     if (argc != 4) {
-        fprintf(stderr,
-                "Usage: %s INPUT OUTPUT FORMAT\n"
-                "  - 2 is Q4_0\n"
-                "  - 3 is Q4_1\n"
-                "  - 6 is Q5_0\n"
-                "  - 7 is Q5_1\n"
-                "  - 8 is Q8_0\n",
-                argv[0]);
+        fprintf(stderr, "%s: missing argument\n", argv[0]);
         return 1;
     }
+
     if (!clip_model_quantize(argv[1], argv[2], atoi(argv[3]))) {
         exit(1);
     }

llama.cpp/main/BUILD.mk

@@ -13,13 +13,9 @@ o/$(MODE)/llama.cpp/main/main:					\
 		o/$(MODE)/llama.cpp/server/server.a		\
 		o/$(MODE)/llama.cpp/llava/llava.a		\
 		o/$(MODE)/llama.cpp/llama.cpp.a			\
-		o/$(MODE)/llama.cpp/main/main.1.asc.zip.o	\
+		o/$(MODE)/llamafile/llamafile.1.asc.zip.o	\
 		$(LLAMA_CPP_SERVER_ASSETS:%=o/$(MODE)/%.zip.o)
 
-llama.cpp/main/main.1.asc: llama.cpp/main/main.1
-	-man $< >$@.tmp && mv -f $@.tmp $@
-	@rm -f $@.tmp
-
 .PHONY: o/$(MODE)/llama.cpp/main
 o/$(MODE)/llama.cpp/main:					\
 		o/$(MODE)/llama.cpp/main/main

llama.cpp/main/main.cpp

@@ -99,47 +99,41 @@ static void llama_log_callback_logTee(ggml_log_level level, const char * text, v
     LOG_TEE("%s", text);
 }
 
-static bool has_argument(int argc, char ** argv, const char * arg) {
-    for (int i = 1; i < argc; ++i) {
-        if (!strcmp(argv[i], arg)) {
-            return true;
-        }
-    }
-    return false;
-}
-
 int main(int argc, char ** argv) {
-    llamafile_init();
-    llamafile_check_cpu();
-    ShowCrashReports();
-    LoadZipArgs(&argc, &argv);
 
-    if (has_argument(argc, argv, "--version")) {
-        printf("llamafile v" LLAMAFILE_VERSION_STRING "\n");
+    if (llamafile_has(argv, "--version")) {
+        puts("llamafile v" LLAMAFILE_VERSION_STRING);
         return 0;
     }
 
-    if (has_argument(argc, argv, "--help")) {
-        llamafile_help("/zip/llama.cpp/main/main.1.asc");
+    if (llamafile_has(argv, "-h") ||
+        llamafile_has(argv, "-help") ||
+        llamafile_has(argv, "--help")) {
+        llamafile_help("/zip/llamafile/llamafile.1.asc");
         __builtin_unreachable();
     }
 
+    llamafile_init();
+    llamafile_check_cpu();
+    ShowCrashReports();
+    LoadZipArgs(&argc, &argv);
+
     if (!IsXnuSilicon() &&
-        (!has_argument(argc, argv, "-ngl") &&
-         !has_argument(argc, argv, "--gpu-layers") &&
-         !has_argument(argc, argv, "--n-gpu-layers"))) {
+        (!llamafile_has(argv, "-ngl") &&
+         !llamafile_has(argv, "--gpu-layers") &&
+         !llamafile_has(argv, "--n-gpu-layers"))) {
         FLAG_gpu = LLAMAFILE_GPU_DISABLE;
     }
 
-    if (!has_argument(argc, argv, "--cli") &&
-        (has_argument(argc, argv, "--server") ||
-         (!has_argument(argc, argv, "-p") &&
-          !has_argument(argc, argv, "-f") &&
-          !has_argument(argc, argv, "--random-prompt")))) {
+    if (!llamafile_has(argv, "--cli") &&
+        (llamafile_has(argv, "--server") ||
+         (!llamafile_has(argv, "-p") &&
+          !llamafile_has(argv, "-f") &&
+          !llamafile_has(argv, "--random-prompt")))) {
         return server_cli(argc, argv);
     }
 
-    if (has_argument(argc, argv, "--image")) {
+    if (llamafile_has(argv, "--image")) {
         return llava_cli(argc, argv);
     }
 

llama.cpp/perplexity/BUILD.mk

@@ -9,9 +9,10 @@ LLAMA_CPP_PERPLEXITY_SRCS = $(filter %.cpp,$(LLAMA_CPP_PERPLEXITY_FILES))
 LLAMA_CPP_PERPLEXITY_OBJS = $(LLAMA_CPP_PERPLEXITY_SRCS:%.cpp=o/$(MODE)/%.o)
 
 .PHONY: o/$(MODE)/llama.cpp/perplexity
-o/$(MODE)/llama.cpp/perplexity:					\
+o/$(MODE)/llama.cpp/perplexity:						\
 		o/$(MODE)/llama.cpp/perplexity/perplexity
 
-o/$(MODE)/llama.cpp/perplexity/perplexity:			\
-		o/$(MODE)/llama.cpp/perplexity/perplexity.o	\
+o/$(MODE)/llama.cpp/perplexity/perplexity:				\
+		o/$(MODE)/llama.cpp/perplexity/perplexity.o		\
+		o/$(MODE)/llama.cpp/perplexity/perplexity.1.asc.zip.o	\
 		o/$(MODE)/llama.cpp/llama.cpp.a

llama.cpp/perplexity/perplexity.1

@@ -1,6 +1,6 @@
 .Dd December 5, 2023
-.Dt llamafile-perplexity 1
-.Os
+.Dt LLAMAFILE-PERPLEXITY 1
+.Os Llamafile Manual
 .Sh NAME
 .Nm llamafile-perplexity
 .Nd LLM benchmarking tool

llama.cpp/perplexity/perplexity.1.asc

@@ -0,0 +1,35 @@
+LLAMAFILE-PERPLEXITY(1)     General Commands Manual    LLAMAFILE-PERPLEXITY(1)
+
+NNAAMMEE
+     llllaammaaffiillee--ppeerrpplleexxiittyy - LLM benchmarking tool
+
+SSYYNNOOPPSSIISS
+     llllaammaaffiillee--ppeerrpplleexxiittyy [flags...]
+
+DDEESSCCRRIIPPTTIIOONN
+     llllaammaaffiillee--ppeerrpplleexxiittyy can be used to gauge the quality of an LLM
+     implementation.
+
+OOPPTTIIOONNSS
+     The following options are available:
+
+     --hh, ----hheellpp
+             Show help message and exit.
+
+     --mm _F_N_A_M_E, ----mmooddeell _F_N_A_M_E
+             Model path (default: models/7B/ggml-model-f16.gguf)
+
+     --ff _F_N_A_M_E, ----ffiillee _F_N_A_M_E
+             Raw data input file.
+
+     --tt _N, ----tthhrreeaaddss _N
+             Number of threads to use during generation (default: nproc/2)
+
+     --ss _S_E_E_D, ----sseeeedd _S_E_E_D
+             Random Number Generator (RNG) seed (default: -1, use random seed
+             for < 0)
+
+SSEEEE AALLSSOO
+     llamafile(1)
+
+Llamafile Manual               December 5, 2023               Llamafile Manual

llama.cpp/perplexity/perplexity.cpp

@@ -680,11 +680,23 @@ static void hellaswag_score(llama_context * ctx, const gpt_params & params) {
 }
 
 int main(int argc, char ** argv) {
-    gpt_params params;
+
+    if (llamafile_has(argv, "--version")) {
+        puts("llamafile-perplexity v" LLAMAFILE_VERSION_STRING);
+        return 0;
+    }
+
+    if (llamafile_has(argv, "-h") ||
+        llamafile_has(argv, "-help") ||
+        llamafile_has(argv, "--help")) {
+        llamafile_help("/zip/llama.cpp/perplexity/perplexity.1.asc");
+        __builtin_unreachable();
+    }
 
     llamafile_init();
     llamafile_check_cpu();
 
+    gpt_params params;
     params.n_batch = 512;
     if (!gpt_params_parse(argc, argv, params)) {
         return 1;

llama.cpp/quantize/BUILD.mk

@@ -9,9 +9,10 @@ LLAMA_CPP_QUANTIZE_SRCS = $(filter %.cpp,$(LLAMA_CPP_QUANTIZE_FILES))
 LLAMA_CPP_QUANTIZE_OBJS = $(LLAMA_CPP_QUANTIZE_SRCS:%.cpp=o/$(MODE)/%.o)
 
 .PHONY: o/$(MODE)/llama.cpp/quantize
-o/$(MODE)/llama.cpp/quantize:					\
+o/$(MODE)/llama.cpp/quantize:						\
 		o/$(MODE)/llama.cpp/quantize/quantize
 
-o/$(MODE)/llama.cpp/quantize/quantize:				\
-		o/$(MODE)/llama.cpp/quantize/quantize.o		\
+o/$(MODE)/llama.cpp/quantize/quantize:					\
+		o/$(MODE)/llama.cpp/quantize/quantize.o			\
+		o/$(MODE)/llama.cpp/quantize/quantize.1.asc.zip.o	\
 		o/$(MODE)/llama.cpp/llama.cpp.a

llama.cpp/quantize/quantize.1

@@ -1,6 +1,6 @@
 .Dd December 5, 2023
-.Dt llamafile-quantize 1
-.Os
+.Dt LLAMAFILE-QUANTIZE 1
+.Os Llamafile Manual
 .Sh NAME
 .Nm llamafile-quantize
 .Nd large language model quantizer