examples : move gpt_init() after parsing the cli args

ggerganov · ggerganov · commit 2afe0a0c7dec · 2024-09-13T11:28:20.000+03:00
diff --git a/examples/batched-bench/batched-bench.cpp b/examples/batched-bench/batched-bench.cpp
@@ -15,14 +15,14 @@ static void print_usage(int, char ** argv) {
 }
 
 int main(int argc, char ** argv) {
-    gpt_init();
-
     gpt_params params;
 
     if (!gpt_params_parse(argc, argv, params, LLAMA_EXAMPLE_BENCH, print_usage)) {
         return 1;
     }
 
+    gpt_init();
+
     int is_pp_shared = params.is_pp_shared;
 
     std::vector<int> n_pp = params.n_pp;
diff --git a/examples/batched/batched.cpp b/examples/batched/batched.cpp
@@ -15,8 +15,6 @@ static void print_usage(int, char ** argv) {
 }
 
 int main(int argc, char ** argv) {
-    gpt_log_init();
-
     gpt_params params;
 
     params.prompt = "Hello my name is";
@@ -26,6 +24,7 @@ int main(int argc, char ** argv) {
         return 1;
     }
 
+    gpt_init();
 
     // number of parallel batches
     int n_parallel = params.n_parallel;
diff --git a/examples/embedding/embedding.cpp b/examples/embedding/embedding.cpp
@@ -79,14 +79,14 @@ static void batch_decode(llama_context * ctx, llama_batch & batch, float * outpu
 }
 
 int main(int argc, char ** argv) {
-    gpt_log_init();
-
     gpt_params params;
 
     if (!gpt_params_parse(argc, argv, params, LLAMA_EXAMPLE_EMBEDDING)) {
         return 1;
     }
 
+    gpt_init();
+
     params.embedding = true;
     // For non-causal models, batch size must be equal to ubatch size
     params.n_ubatch = params.n_batch;
diff --git a/examples/eval-callback/eval-callback.cpp b/examples/eval-callback/eval-callback.cpp
@@ -140,8 +140,6 @@ static bool run(llama_context * ctx, const gpt_params & params) {
 }
 
 int main(int argc, char ** argv) {
-    gpt_init();
-
     callback_data cb_data;
 
     gpt_params params;
@@ -150,6 +148,8 @@ int main(int argc, char ** argv) {
         return 1;
     }
 
+    gpt_init();
+
     llama_backend_init();
     llama_numa_init(params.numa);
 
diff --git a/examples/gritlm/gritlm.cpp b/examples/gritlm/gritlm.cpp
@@ -158,6 +158,8 @@ int main(int argc, char * argv[]) {
         return 1;
     }
 
+    gpt_init();
+
     llama_model_params mparams = llama_model_params_from_gpt_params(params);
     llama_context_params cparams = llama_context_params_from_gpt_params(params);
 
diff --git a/examples/imatrix/imatrix.cpp b/examples/imatrix/imatrix.cpp
@@ -573,8 +573,6 @@ static bool compute_imatrix(llama_context * ctx, const gpt_params & params) {
 }
 
 int main(int argc, char ** argv) {
-    gpt_init();
-
     gpt_params params;
 
     params.n_ctx = 512;
@@ -585,6 +583,8 @@ int main(int argc, char ** argv) {
         return 1;
     }
 
+    gpt_init();
+
     params.n_batch = std::min(params.n_batch, params.n_ctx);
 
     g_collector.set_params(params);
diff --git a/examples/infill/infill.cpp b/examples/infill/infill.cpp
@@ -104,15 +104,15 @@ static void sigint_handler(int signo) {
 #endif
 
 int main(int argc, char ** argv) {
-    gpt_init();
-
     gpt_params params;
     g_params = &params;
 
     if (!gpt_params_parse(argc, argv, params, LLAMA_EXAMPLE_INFILL)) {
         return 1;
     }
 
+    gpt_init();
+
     auto & sparams = params.sparams;
 
     console::init(params.simple_io, params.use_color);
diff --git a/examples/llava/llava-cli.cpp b/examples/llava/llava-cli.cpp
@@ -270,8 +270,6 @@ static void llava_free(struct llava_context * ctx_llava) {
 }
 
 int main(int argc, char ** argv) {
-    gpt_init();
-
     ggml_time_init();
 
     gpt_params params;
@@ -280,6 +278,8 @@ int main(int argc, char ** argv) {
         return 1;
     }
 
+    gpt_init();
+
     if (params.mmproj.empty() || (params.image.empty() && !prompt_contains_image(params.prompt))) {
         print_usage(argc, argv);
         return 1;
diff --git a/examples/llava/minicpmv-cli.cpp b/examples/llava/minicpmv-cli.cpp
@@ -248,8 +248,6 @@ static const char * llama_loop(struct llava_context * ctx_llava,struct gpt_sampl
 }
 
 int main(int argc, char ** argv) {
-    gpt_init();
-
     ggml_time_init();
 
     gpt_params params;
@@ -258,6 +256,8 @@ int main(int argc, char ** argv) {
         return 1;
     }
 
+    gpt_init();
+
     if (params.mmproj.empty() || (params.image.empty())) {
         show_additional_info(argc, argv);
         return 1;
diff --git a/examples/lookahead/lookahead.cpp b/examples/lookahead/lookahead.cpp
@@ -37,14 +37,14 @@ struct ngram_container {
 };
 
 int main(int argc, char ** argv) {
-    gpt_init();
-
     gpt_params params;
 
     if (!gpt_params_parse(argc, argv, params, LLAMA_EXAMPLE_COMMON)) {
         return 1;
     }
 
+    gpt_init();
+
     const int W = 15; // lookahead window
     const int N = 5;  // n-gram size
     const int G = 15; // max verification n-grams
diff --git a/examples/lookup/lookup-stats.cpp b/examples/lookup/lookup-stats.cpp
@@ -13,14 +13,14 @@
 #include <vector>
 
 int main(int argc, char ** argv){
-    gpt_init();
-
     gpt_params params;
 
     if (!gpt_params_parse(argc, argv, params, LLAMA_EXAMPLE_LOOKUP)) {
         return 1;
     }
 
+    gpt_init();
+
     const int n_draft = params.n_draft;
 
     // init llama.cpp
diff --git a/examples/lookup/lookup.cpp b/examples/lookup/lookup.cpp
@@ -13,14 +13,14 @@
 #include <vector>
 
 int main(int argc, char ** argv){
-    gpt_init();
-
     gpt_params params;
 
     if (!gpt_params_parse(argc, argv, params, LLAMA_EXAMPLE_LOOKUP)) {
         return 1;
     }
 
+    gpt_init();
+
     // max. number of additional tokens to draft if match is found
     const int n_draft = params.n_draft;
 
diff --git a/examples/main/main.cpp b/examples/main/main.cpp
@@ -131,14 +131,14 @@ static std::string chat_add_and_format(struct llama_model * model, std::vector<l
 }
 
 int main(int argc, char ** argv) {
-    gpt_init();
-
     gpt_params params;
     g_params = &params;
     if (!gpt_params_parse(argc, argv, params, LLAMA_EXAMPLE_MAIN, print_usage)) {
         return 1;
     }
 
+    gpt_init();
+
     auto & sparams = params.sparams;
 
     // save choice to use color for later
diff --git a/examples/parallel/parallel.cpp b/examples/parallel/parallel.cpp
@@ -103,14 +103,14 @@ static std::vector<std::string> split_string(const std::string& input, char deli
 int main(int argc, char ** argv) {
     srand(1234);
 
-    gpt_init();
-
     gpt_params params;
 
     if (!gpt_params_parse(argc, argv, params, LLAMA_EXAMPLE_PARALLEL)) {
         return 1;
     }
 
+    gpt_init();
+
     // number of simultaneous "clients" to simulate
     const int32_t n_clients = params.n_parallel;
 
diff --git a/examples/passkey/passkey.cpp b/examples/passkey/passkey.cpp
@@ -15,8 +15,6 @@ static void print_usage(int, char ** argv) {
 }
 
 int main(int argc, char ** argv) {
-    gpt_init();
-
     gpt_params params;
 
     params.n_junk = 250;
@@ -27,6 +25,8 @@ int main(int argc, char ** argv) {
         return 1;
     }
 
+    gpt_init();
+
     int n_junk = params.n_junk;
     int n_keep = params.n_keep;
     int n_grp  = params.grp_attn_n;
diff --git a/examples/perplexity/perplexity.cpp b/examples/perplexity/perplexity.cpp
@@ -1957,8 +1957,6 @@ static void kl_divergence(llama_context * ctx, const gpt_params & params) {
 }
 
 int main(int argc, char ** argv) {
-    gpt_init();
-
     gpt_params params;
 
     params.n_ctx = 512;
@@ -1968,6 +1966,8 @@ int main(int argc, char ** argv) {
         return 1;
     }
 
+    gpt_init();
+
     const int32_t n_ctx = params.n_ctx;
 
     if (n_ctx <= 0) {
diff --git a/examples/retrieval/retrieval.cpp b/examples/retrieval/retrieval.cpp
@@ -112,14 +112,14 @@ static void batch_decode(llama_context * ctx, llama_batch & batch, float * outpu
 }
 
 int main(int argc, char ** argv) {
-    gpt_init();
-
     gpt_params params;
 
     if (!gpt_params_parse(argc, argv, params, LLAMA_EXAMPLE_RETRIEVAL, print_usage)) {
         return 1;
     }
 
+    gpt_init();
+
     // For BERT models, batch size must be equal to ubatch size
     params.n_ubatch = params.n_batch;
     params.embedding = true;
diff --git a/examples/server/server.cpp b/examples/server/server.cpp
@@ -2316,15 +2316,15 @@ inline void signal_handler(int signal) {
 }
 
 int main(int argc, char ** argv) {
-    gpt_init();
-
     // own arguments required by this example
     gpt_params params;
 
     if (!gpt_params_parse(argc, argv, params, LLAMA_EXAMPLE_SERVER)) {
         return 1;
     }
 
+    gpt_init();
+
     const bool verbose = params.verbosity > 0;
 
     // struct that contains llama context and inference
diff --git a/examples/simple/simple.cpp b/examples/simple/simple.cpp
@@ -12,8 +12,6 @@ static void print_usage(int, char ** argv) {
 }
 
 int main(int argc, char ** argv) {
-    gpt_init();
-
     gpt_params params;
 
     params.prompt = "Hello my name is";
@@ -23,6 +21,8 @@ int main(int argc, char ** argv) {
         return 1;
     }
 
+    gpt_init();
+
     // total length of the sequence including the prompt
     const int n_predict = params.n_predict;
 
diff --git a/examples/speculative/speculative.cpp b/examples/speculative/speculative.cpp
@@ -30,14 +30,14 @@ struct seq_draft {
 };
 
 int main(int argc, char ** argv) {
-    gpt_init();
-
     gpt_params params;
 
     if (!gpt_params_parse(argc, argv, params, LLAMA_EXAMPLE_SPECULATIVE)) {
         return 1;
     }
 
+    gpt_init();
+
     if (params.model_draft.empty()) {
         LOG_ERR("%s: --model-draft is required\n", __func__);
         return 1;

Original file line number	Diff line number	Diff line change
`@@ -15,14 +15,14 @@ static void print_usage(int, char ** argv) {`
`15`	`15`	`}`
`16`	`16`
`17`	`17`	`int main(int argc, char ** argv) {`
`18`		`- gpt_init();`
`19`		`-`
`20`	`18`	`gpt_params params;`
`21`	`19`
`22`	`20`	`if (!gpt_params_parse(argc, argv, params, LLAMA_EXAMPLE_BENCH, print_usage)) {`
`23`	`21`	`return 1;`
`24`	`22`	`}`
`25`	`23`
	`24`	`+ gpt_init();`
	`25`	`+`
`26`	`26`	`int is_pp_shared = params.is_pp_shared;`
`27`	`27`
`28`	`28`	`std::vector<int> n_pp = params.n_pp;`
Original file line number	Diff line number	Diff line change
`@@ -15,8 +15,6 @@ static void print_usage(int, char ** argv) {`
`15`	`15`	`}`
`16`	`16`
`17`	`17`	`int main(int argc, char ** argv) {`
`18`		`- gpt_log_init();`
`19`		`-`
`20`	`18`	`gpt_params params;`
`21`	`19`
`22`	`20`	`params.prompt = "Hello my name is";`
`@@ -26,6 +24,7 @@ int main(int argc, char ** argv) {`
`26`	`24`	`return 1;`
`27`	`25`	`}`
`28`	`26`
	`27`	`+ gpt_init();`
`29`	`28`
`30`	`29`	`// number of parallel batches`
`31`	`30`	`int n_parallel = params.n_parallel;`
Original file line number	Diff line number	Diff line change
`@@ -79,14 +79,14 @@ static void batch_decode(llama_context * ctx, llama_batch & batch, float * outpu`
`79`	`79`	`}`
`80`	`80`
`81`	`81`	`int main(int argc, char ** argv) {`
`82`		`- gpt_log_init();`
`83`		`-`
`84`	`82`	`gpt_params params;`
`85`	`83`
`86`	`84`	`if (!gpt_params_parse(argc, argv, params, LLAMA_EXAMPLE_EMBEDDING)) {`
`87`	`85`	`return 1;`
`88`	`86`	`}`
`89`	`87`
	`88`	`+ gpt_init();`
	`89`	`+`
`90`	`90`	`params.embedding = true;`
`91`	`91`	`// For non-causal models, batch size must be equal to ubatch size`
`92`	`92`	`params.n_ubatch = params.n_batch;`
Original file line number	Diff line number	Diff line change
`@@ -140,8 +140,6 @@ static bool run(llama_context * ctx, const gpt_params & params) {`
`140`	`140`	`}`
`141`	`141`
`142`	`142`	`int main(int argc, char ** argv) {`
`143`		`- gpt_init();`
`144`		`-`
`145`	`143`	`callback_data cb_data;`
`146`	`144`
`147`	`145`	`gpt_params params;`
`@@ -150,6 +148,8 @@ int main(int argc, char ** argv) {`
`150`	`148`	`return 1;`
`151`	`149`	`}`
`152`	`150`
	`151`	`+ gpt_init();`
	`152`	`+`
`153`	`153`	`llama_backend_init();`
`154`	`154`	`llama_numa_init(params.numa);`
`155`	`155`
Original file line number	Diff line number	Diff line change
`@@ -158,6 +158,8 @@ int main(int argc, char * argv[]) {`
`158`	`158`	`return 1;`
`159`	`159`	`}`
`160`	`160`
	`161`	`+ gpt_init();`
	`162`	`+`
`161`	`163`	`llama_model_params mparams = llama_model_params_from_gpt_params(params);`
`162`	`164`	`llama_context_params cparams = llama_context_params_from_gpt_params(params);`
`163`	`165`
Original file line number	Diff line number	Diff line change
`@@ -573,8 +573,6 @@ static bool compute_imatrix(llama_context * ctx, const gpt_params & params) {`
`573`	`573`	`}`
`574`	`574`
`575`	`575`	`int main(int argc, char ** argv) {`
`576`		`- gpt_init();`
`577`		`-`
`578`	`576`	`gpt_params params;`
`579`	`577`
`580`	`578`	`params.n_ctx = 512;`
`@@ -585,6 +583,8 @@ int main(int argc, char ** argv) {`
`585`	`583`	`return 1;`
`586`	`584`	`}`
`587`	`585`
	`586`	`+ gpt_init();`
	`587`	`+`
`588`	`588`	`params.n_batch = std::min(params.n_batch, params.n_ctx);`
`589`	`589`
`590`	`590`	`g_collector.set_params(params);`
Original file line number	Diff line number	Diff line change
`@@ -270,8 +270,6 @@ static void llava_free(struct llava_context * ctx_llava) {`
`270`	`270`	`}`
`271`	`271`
`272`	`272`	`int main(int argc, char ** argv) {`
`273`		`- gpt_init();`
`274`		`-`
`275`	`273`	`ggml_time_init();`
`276`	`274`
`277`	`275`	`gpt_params params;`
`@@ -280,6 +278,8 @@ int main(int argc, char ** argv) {`
`280`	`278`	`return 1;`
`281`	`279`	`}`
`282`	`280`
	`281`	`+ gpt_init();`
	`282`	`+`
`283`	`283`	`if (params.mmproj.empty() \|\| (params.image.empty() && !prompt_contains_image(params.prompt))) {`
`284`	`284`	`print_usage(argc, argv);`
`285`	`285`	`return 1;`
Original file line number	Diff line number	Diff line change
`@@ -248,8 +248,6 @@ static const char * llama_loop(struct llava_context * ctx_llava,struct gpt_sampl`
`248`	`248`	`}`
`249`	`249`
`250`	`250`	`int main(int argc, char ** argv) {`
`251`		`- gpt_init();`
`252`		`-`
`253`	`251`	`ggml_time_init();`
`254`	`252`
`255`	`253`	`gpt_params params;`
`@@ -258,6 +256,8 @@ int main(int argc, char ** argv) {`
`258`	`256`	`return 1;`
`259`	`257`	`}`
`260`	`258`
	`259`	`+ gpt_init();`
	`260`	`+`
`261`	`261`	`if (params.mmproj.empty() \|\| (params.image.empty())) {`
`262`	`262`	`show_additional_info(argc, argv);`
`263`	`263`	`return 1;`
Original file line number	Diff line number	Diff line change
`@@ -131,14 +131,14 @@ static std::string chat_add_and_format(struct llama_model * model, std::vector<l`
`131`	`131`	`}`
`132`	`132`
`133`	`133`	`int main(int argc, char ** argv) {`
`134`		`- gpt_init();`
`135`		`-`
`136`	`134`	`gpt_params params;`
`137`	`135`	`g_params = &params;`
`138`	`136`	`if (!gpt_params_parse(argc, argv, params, LLAMA_EXAMPLE_MAIN, print_usage)) {`
`139`	`137`	`return 1;`
`140`	`138`	`}`
`141`	`139`
	`140`	`+ gpt_init();`
	`141`	`+`
`142`	`142`	`auto & sparams = params.sparams;`
`143`	`143`
`144`	`144`	`// save choice to use color for later`
Original file line number	Diff line number	Diff line change
`@@ -1957,8 +1957,6 @@ static void kl_divergence(llama_context * ctx, const gpt_params & params) {`
`1957`	`1957`	`}`
`1958`	`1958`
`1959`	`1959`	`int main(int argc, char ** argv) {`
`1960`		`- gpt_init();`
`1961`		`-`
`1962`	`1960`	`gpt_params params;`
`1963`	`1961`
`1964`	`1962`	`params.n_ctx = 512;`
`@@ -1968,6 +1966,8 @@ int main(int argc, char ** argv) {`
`1968`	`1966`	`return 1;`
`1969`	`1967`	`}`
`1970`	`1968`
	`1969`	`+ gpt_init();`
	`1970`	`+`
`1971`	`1971`	`const int32_t n_ctx = params.n_ctx;`
`1972`	`1972`
`1973`	`1973`	`if (n_ctx <= 0) {`