Move model loading back to main.cpp

Signed-off-by: Thiago Padilha <[email protected]>

Author: tarruda

llama.cpp

@@ -829,41 +829,12 @@ void sigint_handler(int signo) {
 }
 #endif
 
-const char * llama_print_system_info(void) {
-    static std::string s;
-
-    s  = "";
-    s += "AVX = "       + std::to_string(ggml_cpu_has_avx())       + " | ";
-    s += "AVX2 = "      + std::to_string(ggml_cpu_has_avx2())      + " | ";
-    s += "AVX512 = "    + std::to_string(ggml_cpu_has_avx512())    + " | ";
-    s += "FMA = "       + std::to_string(ggml_cpu_has_fma())       + " | ";
-    s += "NEON = "      + std::to_string(ggml_cpu_has_neon())      + " | ";
-    s += "ARM_FMA = "   + std::to_string(ggml_cpu_has_arm_fma())   + " | ";
-    s += "F16C = "      + std::to_string(ggml_cpu_has_f16c())      + " | ";
-    s += "FP16_VA = "   + std::to_string(ggml_cpu_has_fp16_va())   + " | ";
-    s += "WASM_SIMD = " + std::to_string(ggml_cpu_has_wasm_simd()) + " | ";
-    s += "BLAS = "      + std::to_string(ggml_cpu_has_blas())      + " | ";
-    s += "SSE3 = "      + std::to_string(ggml_cpu_has_sse3())      + " | ";
-    s += "VSX = "       + std::to_string(ggml_cpu_has_vsx())       + " | ";
-
-    return s.c_str();
-}
-
-int llama_main(int argc, char ** argv) {
-    ggml_time_init();
-    const int64_t t_main_start_us = ggml_time_us();
-
-    gpt_params params;
-    params.model = "models/llama-7B/ggml-model.bin";
-
-    if (gpt_params_parse(argc, argv, params) == false) {
-        return 1;
-    }
-
-    if (params.n_ctx > 2048) {
-        fprintf(stderr, "%s: warning: model does not support context sizes greater than 2048 tokens (%d specified);"
-                "expect poor results\n", __func__, params.n_ctx);
-    }
+int llama_main(
+    gpt_params params,
+    llama_vocab vocab,
+    llama_model model,
+    int64_t t_load_us,
+    int64_t t_main_start_us) {
 
     if (params.seed < 0) {
         params.seed = time(NULL);
@@ -879,30 +850,6 @@ int llama_main(int argc, char ** argv) {
 //    params.prompt = R"(// this function checks if the number n is prime
 //bool is_prime(int n) {)";
 
-    int64_t t_load_us = 0;
-
-    llama_vocab vocab;
-    llama_model model;
-
-    // load the model
-    {
-        const ggml_type memory_type = params.memory_f16 ? GGML_TYPE_F16 : GGML_TYPE_F32;
-        const int64_t t_start_us = ggml_time_us();
-        if (!llama_model_load(params.model, model, vocab, params.n_ctx, params.n_parts, memory_type)) {
-            fprintf(stderr, "%s: failed to load model from '%s'\n", __func__, params.model.c_str());
-            return 1;
-        }
-
-        t_load_us = ggml_time_us() - t_start_us;
-    }
-
-    // print system information
-    {
-        fprintf(stderr, "\n");
-        fprintf(stderr, "system_info: n_threads = %d / %d | %s\n",
-                params.n_threads, std::thread::hardware_concurrency(), llama_print_system_info());
-    }
-
     std::vector<float> logits;
 
     // determine the required inference memory per token:

llama.h

@@ -6,6 +6,7 @@
 #include <string>
 
 #include "ggml.h"
+#include "utils.h"
 
 
 // default hparams (LLaMA 7B)
@@ -58,4 +59,10 @@ struct llama_model {
     std::map<std::string, struct ggml_tensor *> tensors;
 };
 
-int llama_main(int argc, char ** argv);
+int llama_main(
+    gpt_params params,
+    llama_vocab vocab,
+    llama_model model,
+    int64_t t_load_us,
+    int64_t t_main_start_us);
+bool llama_model_load(const std::string & fname, llama_model & model, llama_vocab & vocab, int n_ctx, int n_parts, ggml_type memory_type);

main.cpp

@@ -1,5 +1,67 @@
+#include "ggml.h"
+#include "utils.h"
 #include "llama.h"
 
+const char * llama_print_system_info(void) {
+    static std::string s;
+
+    s  = "";
+    s += "AVX = "       + std::to_string(ggml_cpu_has_avx())       + " | ";
+    s += "AVX2 = "      + std::to_string(ggml_cpu_has_avx2())      + " | ";
+    s += "AVX512 = "    + std::to_string(ggml_cpu_has_avx512())    + " | ";
+    s += "FMA = "       + std::to_string(ggml_cpu_has_fma())       + " | ";
+    s += "NEON = "      + std::to_string(ggml_cpu_has_neon())      + " | ";
+    s += "ARM_FMA = "   + std::to_string(ggml_cpu_has_arm_fma())   + " | ";
+    s += "F16C = "      + std::to_string(ggml_cpu_has_f16c())      + " | ";
+    s += "FP16_VA = "   + std::to_string(ggml_cpu_has_fp16_va())   + " | ";
+    s += "WASM_SIMD = " + std::to_string(ggml_cpu_has_wasm_simd()) + " | ";
+    s += "BLAS = "      + std::to_string(ggml_cpu_has_blas())      + " | ";
+    s += "SSE3 = "      + std::to_string(ggml_cpu_has_sse3())      + " | ";
+    s += "VSX = "       + std::to_string(ggml_cpu_has_vsx())       + " | ";
+
+    return s.c_str();
+}
+
 int main(int argc, char ** argv) {
-    return llama_main(argc, argv);
+
+    ggml_time_init();
+    const int64_t t_main_start_us = ggml_time_us();
+
+    gpt_params params;
+    params.model = "models/llama-7B/ggml-model.bin";
+
+    if (gpt_params_parse(argc, argv, params) == false) {
+        return 1;
+    }
+
+    if (params.n_ctx > 2048) {
+        fprintf(stderr, "%s: warning: model does not support context sizes greater than 2048 tokens (%d specified);"
+                "expect poor results\n", __func__, params.n_ctx);
+    }
+
+    int64_t t_load_us = 0;
+
+    llama_vocab vocab;
+    llama_model model;
+
+    // load the model
+    {
+        const ggml_type memory_type = params.memory_f16 ? GGML_TYPE_F16 : GGML_TYPE_F32;
+        const int64_t t_start_us = ggml_time_us();
+        if (!llama_model_load(params.model, model, vocab, params.n_ctx, params.n_parts, memory_type)) {
+            fprintf(stderr, "%s: failed to load model from '%s'\n", __func__, params.model.c_str());
+            return 1;
+        }
+
+        t_load_us = ggml_time_us() - t_start_us;
+    }
+
+    // print system information
+    {
+        fprintf(stderr, "\n");
+        fprintf(stderr, "system_info: n_threads = %d / %d | %s\n",
+                params.n_threads, std::thread::hardware_concurrency(), llama_print_system_info());
+    }
+
+    return llama_main(params, vocab, model, t_main_start_us, t_load_us);
 }