Add log for AVX2 and create log_utils

Author: CameronNg

controllers/llamaCPP.cc

@@ -3,14 +3,13 @@
 #include <fstream>
 #include <iostream>
 #include "log.h"
+#include "utils/nitro_utils.h"
+#include "utils/logging_utils.h"
 
 // External
 #include "common.h"
 #include "llama.h"
 
-#include "log.h"
-#include "utils/nitro_utils.h"
-
 using namespace inferences;
 using json = nlohmann::json;
 
@@ -189,7 +188,7 @@ void llamaCPP::InferenceImpl(
     inferences::ChatCompletionRequest&& completion,
     std::function<void(const HttpResponsePtr&)>& callback) {
   std::string formatted_output = pre_prompt;
-  int request_id = ++no_of_inference_requests;
+  int request_id = ++no_of_requests;
   LOG_INFO_REQUEST(request_id) << "Generating reponse for inference request";
 
   json data;
@@ -464,7 +463,7 @@ void llamaCPP::Embedding(
 void llamaCPP::EmbeddingImpl(
     std::shared_ptr<Json::Value> jsonBody,
     std::function<void(const HttpResponsePtr&)>& callback) {
-  int request_id = ++no_of_inference_requests;
+  int request_id = ++no_of_requests;
   LOG_INFO_REQUEST(request_id) << "Generating reponse for embedding request";
   // Queue embedding task
   auto state = create_inference_state(this);
@@ -559,6 +558,9 @@ void llamaCPP::LoadModel(
     return;
   }
 
+  if (!nitro_utils::isAVX2Supported() && ggml_cpu_has_avx2())
+    LOG_ERROR << "AVX2 is not supported by your processor, please download and replace the correct Nitro asset version";
+
   const auto& jsonBody = req->getJsonObject();
   if (!LoadModelImpl(jsonBody)) {
     // Error occurred during model loading

controllers/llamaCPP.h

@@ -27,10 +27,6 @@
 #define SERVER_VERBOSE 1
 #endif
 
-#define LOG_INFO_REQUEST(RID) LOG_INFO << "Request " << RID << ": " 
-#define LOG_WARN_REQUEST(RID) LOG_WARN << "Request " << RID << ": " 
-#define LOG_ERROR_REQUEST(RID) LOG_ERROR << "Request " << RID << ": " 
-
 using json = nlohmann::json;
 
 using namespace drogon;
@@ -89,7 +85,7 @@ class llamaCPP : public drogon::HttpController<llamaCPP>,
   std::string pre_prompt;
   int repeat_last_n;
   bool caching_enabled;
-  std::atomic<int> no_of_inference_requests = 0;
+  std::atomic<int> no_of_requests = 0;
   std::atomic<int> no_of_chats = 0;
   int clean_cache_threshold;
   std::string grammar_file_content;

utils/logging_utils.h

@@ -0,0 +1,5 @@
+#pragma once
+
+#define LOG_INFO_REQUEST(RID) LOG_INFO << "Request " << RID << ": " 
+#define LOG_WARN_REQUEST(RID) LOG_WARN << "Request " << RID << ": " 
+#define LOG_ERROR_REQUEST(RID) LOG_ERROR << "Request " << RID << ": " 

utils/nitro_utils.h

@@ -10,7 +10,13 @@
 #include <ostream>
 #include <regex>
 #include <vector>
-// Include platform-specific headers
+
+#if defined(__GNUC__) || defined(__clang__)
+#include <cpuid.h>
+#elif defined(_MSC_VER)
+#include <intrin.h>
+#endif
+
 #ifdef _WIN32
 #include <winsock2.h>
 #include <windows.h>
@@ -174,6 +180,25 @@ inline std::string generate_random_string(std::size_t length) {
   return random_string;
 }
 
+inline bool isAVX2Supported() {
+#if defined(__GNUC__) || defined(__clang__)
+    unsigned eax, ebx, ecx, edx;
+    if (__get_cpuid_max(0, nullptr) < 7) return false;
+
+    __get_cpuid_count(7, 0, &eax, &ebx, &ecx, &edx);
+    return (ebx & (1 << 5)) != 0;
+#elif defined(_MSC_VER)
+    int cpuInfo[4];
+    __cpuid(cpuInfo, 0);
+    int nIds = cpuInfo[0];
+    if (nIds >= 7) {
+        __cpuidex(cpuInfo, 7, 0);
+        return (cpuInfo[1] & (1 << 5)) != 0;
+    }
+    return false;
+#endif
+}
+
 inline void nitro_logo() {
   std::string rainbowColors[] = {
       "\033[93m", // Yellow