common : gpt_params_parse do not print usage

common/common.cpp

@@ -265,19 +265,20 @@ bool gpt_params_parse_ex(int argc, char ** argv, gpt_params & params) {
 }
 
 bool gpt_params_parse(int argc, char ** argv, gpt_params & params) {
-    bool result = true;
+    const auto params_org = params; // the example can modify the default params
+
     try {
-        if (!gpt_params_parse_ex(argc, argv, params)) {
+        if (!gpt_params_parse_ex(argc, argv, params) || params.usage) {
-            gpt_params_print_usage(argc, argv, gpt_params());
+            params = params_org;
-            exit(0);
+            params.usage = true;
+            return false;
         }
-    }
+    } catch (const std::invalid_argument & ex) {
-    catch (const std::invalid_argument & ex) {
         fprintf(stderr, "%s\n", ex.what());
-        gpt_params_print_usage(argc, argv, gpt_params());
+        return false;
-        exit(1);
     }
-    return result;
+
+    return true;
 }
 
 bool gpt_params_find_arg(int argc, char ** argv, const std::string & arg, gpt_params & params, int & i, bool & invalid_param) {
@@ -1242,9 +1243,9 @@ bool gpt_params_find_arg(int argc, char ** argv, const std::string & arg, gpt_pa
         }
         return true;
     }
-    if (arg == "-h" || arg == "--help") {
+    if (arg == "-h" || arg == "--help" || arg == "--usage"  ) {
-        gpt_params_print_usage(argc, argv, gpt_params());
+        params.usage = true;
-        exit(0);
+        return true;
     }
     if (arg == "--version") {
         fprintf(stderr, "version: %d (%s)\n", LLAMA_BUILD_NUMBER, LLAMA_COMMIT);
@@ -1363,7 +1364,7 @@ void gpt_params_print_usage(int /*argc*/, char ** argv, const gpt_params & param
     printf("usage: %s [options]\n", argv[0]);
     printf("\n");
     printf("options:\n");
-    printf("  -h, --help            show this help message and exit\n");
+    printf("  -h, --help, --usage   print usage and exit\n");
     printf("  --version             show version and build info\n");
     printf("  -i, --interactive     run in interactive mode\n");
     printf("  --special             special tokens output enabled\n");

common/common.h

@@ -142,6 +142,7 @@ struct gpt_params {
 
     bool   kl_divergence   = false; // compute KL divergence
 
+    bool usage             = false; // print usage
     bool random_prompt     = false; // do not randomize prompt if none provided
     bool use_color         = false; // use color to distinguish generations and inputs
     bool interactive       = false; // interactive mode

examples/embedding/embedding.cpp

@@ -63,6 +63,7 @@ int main(int argc, char ** argv) {
     gpt_params params;
 
     if (!gpt_params_parse(argc, argv, params)) {
+        gpt_params_print_usage(argc, argv, params);
         return 1;
     }
 

examples/eval-callback/eval-callback.cpp

@@ -140,11 +140,12 @@ static bool run(llama_context * ctx, const gpt_params & params) {
 }
 
 int main(int argc, char ** argv) {
-
     callback_data cb_data;
 
     gpt_params params;
+
     if (!gpt_params_parse(argc, argv, params)) {
+        gpt_params_print_usage(argc, argv, params);
         return 1;
     }
 

examples/gritlm/gritlm.cpp

@@ -153,7 +153,9 @@ static std::string gritlm_instruction(const std::string & instruction) {
 
 int main(int argc, char * argv[]) {
     gpt_params params;
+
     if (!gpt_params_parse(argc, argv, params)) {
+        gpt_params_print_usage(argc, argv, params);
         return 1;
     }
 

examples/imatrix/imatrix.cpp

@@ -533,7 +533,6 @@ static bool compute_imatrix(llama_context * ctx, const gpt_params & params, bool
 }
 
 int main(int argc, char ** argv) {
-
     StatParams sparams;
     std::string prev_result_file;
     std::string combine_files;
@@ -581,7 +580,9 @@ int main(int argc, char ** argv) {
 
     gpt_params params;
     params.n_batch = 512;
-    if (!gpt_params_parse(args.size(), args.data(), params)) {
+
+    if (!gpt_params_parse(argc, argv, params)) {
+        gpt_params_print_usage(argc, argv, params);
         return 1;
     }
 

examples/infill/infill.cpp

@@ -107,6 +107,7 @@ int main(int argc, char ** argv) {
     g_params = &params;
 
     if (!gpt_params_parse(argc, argv, params)) {
+        gpt_params_print_usage(argc, argv, params);
         return 1;
     }
 

examples/llava/llava-cli.cpp

@@ -112,9 +112,12 @@ struct llava_context {
     struct llama_model * model = NULL;
 };
 
-static void show_additional_info(int /*argc*/, char ** argv) {
+static void print_usage(int argc, char ** argv, const gpt_params & params) {
-    LOG_TEE("\n example usage: %s -m <llava-v1.5-7b/ggml-model-q5_k.gguf> --mmproj <llava-v1.5-7b/mmproj-model-f16.gguf> --image <path/to/an/image.jpg> --image <path/to/another/image.jpg> [--temp 0.1] [-p \"describe the image in detail.\"]\n", argv[0]);
+    gpt_params_print_usage(argc, argv, params);
-    LOG_TEE("  note: a lower temperature value like 0.1 is recommended for better quality.\n");
+
+    LOG_TEE("\n example usage:\n");
+    LOG_TEE("\n     %s -m <llava-v1.5-7b/ggml-model-q5_k.gguf> --mmproj <llava-v1.5-7b/mmproj-model-f16.gguf> --image <path/to/an/image.jpg> --image <path/to/another/image.jpg> [--temp 0.1] [-p \"describe the image in detail.\"]\n", argv[0]);
+    LOG_TEE("\n note: a lower temperature value like 0.1 is recommended for better quality.\n");
 }
 
 static struct llava_image_embed * load_image(llava_context * ctx_llava, gpt_params * params, const std::string & fname) {
@@ -278,7 +281,7 @@ int main(int argc, char ** argv) {
     gpt_params params;
 
     if (!gpt_params_parse(argc, argv, params)) {
-        show_additional_info(argc, argv);
+        print_usage(argc, argv, params);
         return 1;
     }
 
@@ -290,8 +293,7 @@ int main(int argc, char ** argv) {
 #endif // LOG_DISABLE_LOGS
 
     if (params.mmproj.empty() || (params.image.empty() && !prompt_contains_image(params.prompt))) {
-        gpt_params_print_usage(argc, argv, params);
+        print_usage(argc, argv, {});
-        show_additional_info(argc, argv);
         return 1;
     }
     auto model = llava_init(&params);

examples/lookahead/lookahead.cpp

@@ -37,7 +37,8 @@ struct ngram_container {
 int main(int argc, char ** argv) {
     gpt_params params;
 
-    if (gpt_params_parse(argc, argv, params) == false) {
+    if (!gpt_params_parse(argc, argv, params)) {
+        gpt_params_print_usage(argc, argv, params);
         return 1;
     }
 

examples/lookup/lookup-create.cpp

@@ -14,8 +14,10 @@ int main(int argc, char ** argv){
     gpt_params params;
 
     if (!gpt_params_parse(argc, argv, params)) {
+        gpt_params_print_usage(argc, argv, params);
         return 1;
     }
+
     // init llama.cpp
     llama_backend_init();
     llama_numa_init(params.numa);

examples/lookup/lookup-stats.cpp

@@ -16,6 +16,7 @@ int main(int argc, char ** argv){
     gpt_params params;
 
     if (!gpt_params_parse(argc, argv, params)) {
+        gpt_params_print_usage(argc, argv, params);
         return 1;
     }
 

examples/lookup/lookup.cpp

@@ -15,6 +15,7 @@ int main(int argc, char ** argv){
     gpt_params params;
 
     if (!gpt_params_parse(argc, argv, params)) {
+        gpt_params_print_usage(argc, argv, params);
         return 1;
     }
 

examples/main/main.cpp

@@ -122,8 +122,10 @@ int main(int argc, char ** argv) {
     g_params = &params;
 
     if (!gpt_params_parse(argc, argv, params)) {
+        gpt_params_print_usage(argc, argv, params);
         return 1;
     }
+
     llama_sampling_params & sparams = params.sparams;
 
 #ifndef LOG_DISABLE_LOGS

examples/parallel/parallel.cpp

@@ -100,7 +100,8 @@ int main(int argc, char ** argv) {
 
     gpt_params params;
 
-    if (gpt_params_parse(argc, argv, params) == false) {
+    if (!gpt_params_parse(argc, argv, params)) {
+        gpt_params_print_usage(argc, argv, params);
         return 1;
     }
 

examples/perplexity/perplexity.cpp

@@ -1965,6 +1965,7 @@ int main(int argc, char ** argv) {
     gpt_params params;
 
     if (!gpt_params_parse(argc, argv, params)) {
+        gpt_params_print_usage(argc, argv, params);
         return 1;
     }
 

examples/retrieval/retrieval.cpp

@@ -27,7 +27,7 @@ static void retrieval_params_parse(int argc, char ** argv, gpt_params & gpt_para
     while (i < argc) {
         arg = argv[i];
         bool invalid_gpt_param = false;
-        if(gpt_params_find_arg(argc, argv, argv[i], gpt_params, i, invalid_gpt_param)) {
+        if (gpt_params_find_arg(argc, argv, argv[i], gpt_params, i, invalid_gpt_param)) {
             if (invalid_gpt_param) {
                 fprintf(stderr, "error: invalid argument: %s\n", arg.c_str());
                 retrieval_params_print_usage(argc, argv, gpt_params, retrieval_params);

examples/save-load-state/save-load-state.cpp

@@ -11,6 +11,7 @@ int main(int argc, char ** argv) {
     params.prompt = "The quick brown fox";
 
     if (!gpt_params_parse(argc, argv, params)) {
+        gpt_params_print_usage(argc, argv, params);
         return 1;
     }
 

examples/speculative/speculative.cpp

@@ -27,7 +27,8 @@ struct seq_draft {
 int main(int argc, char ** argv) {
     gpt_params params;
 
-    if (gpt_params_parse(argc, argv, params) == false) {
+    if (!gpt_params_parse(argc, argv, params)) {
+        gpt_params_print_usage(argc, argv, params);
         return 1;
     }