cmake : fix GGML_USE_SYCL typo (#5555)

* server: enrich health endpoint with available slots, return 503 if not slots are available

* server: document new status no slot available in the README.md

CMakeLists.txt

@@ -526,7 +526,7 @@ if (LLAMA_SYCL)
 
     message(STATUS "SYCL found")
 
-    add_compile_definitions(GML_USE_SYCL)
+    add_compile_definitions(GGML_USE_SYCL)
 
     if (LLAMA_SYCL_F16)
         add_compile_definitions(GGML_SYCL_F16)

examples/server/README.md

@@ -39,6 +39,7 @@ see https://github.com/ggerganov/llama.cpp/issues/1437
 - `--mmproj MMPROJ_FILE`: Path to a multimodal projector file for LLaVA.
 - `--grp-attn-n`: Set the group attention factor to extend context size through self-extend(default: 1=disabled), used together with group attention width `--grp-attn-w`
 - `--grp-attn-w`: Set the group attention width to extend context size through self-extend(default: 512), used together with group attention factor `--grp-attn-n`
+- `-n, --n-predict`: Set the maximum tokens to predict (default: -1)
 
 ## Build
 
@@ -135,6 +136,7 @@ node index.js
   - `{"status": "loading model"}` if the model is still being loaded.
   - `{"status": "error"}` if the model failed to load.
   - `{"status": "ok"}` if the model is successfully loaded and the server is ready for further requests mentioned below.
+  - `{"status": "no slot available", "slots_idle": 0, "slots_processing": 32}` if no slot are currently available
 
 - **POST** `/completion`: Given a `prompt`, it returns the predicted completion.
 

examples/server/server.cpp

@@ -28,6 +28,7 @@
 #include <chrono>
 #include <condition_variable>
 #include <atomic>
+#include <signal.h>
 
 using json = nlohmann::json;
 
@@ -158,6 +159,7 @@ struct llama_client_slot
     int32_t n_decoded   = 0;
     int32_t n_remaining = -1;
     int32_t i_batch     = -1;
+    int32_t n_predict   = -1;
 
     int32_t num_prompt_tokens           = 0;
     int32_t num_prompt_tokens_processed = 0;
@@ -409,6 +411,7 @@ struct llama_server_context
 
             slot.id = i;
             slot.n_ctx = n_ctx_slot;
+            slot.n_predict = params.n_predict;
 
             LOG_TEE(" -> Slot %i - max context: %i\n", slot.id, n_ctx_slot);
 
@@ -545,6 +548,15 @@ struct llama_server_context
         slot->sparams.grammar           = json_value(data, "grammar",           default_sparams.grammar);
         slot->sparams.n_probs           = json_value(data, "n_probs",           default_sparams.n_probs);
 
+        if (slot->n_predict > 0 && slot->params.n_predict > slot->n_predict) {
+            // Might be better to reject the request with a 400 ?
+            LOG_WARNING("Max tokens to predict exceeds server configuration", {
+                {"params.n_predict", slot->params.n_predict},
+                {"slot.n_predict", slot->n_predict},
+            });
+            slot->params.n_predict = slot->n_predict;
+        }
+
         // infill
         if (data.count("input_prefix") != 0)
         {
@@ -1052,6 +1064,7 @@ struct llama_server_context
 
         return json {
             {"n_ctx",             slot.n_ctx},
+            {"n_predict",         slot.n_predict},
             {"model",             params.model_alias},
             {"seed",              slot.params.seed},
             {"temperature",       slot.sparams.temp},
@@ -1914,13 +1927,14 @@ static void server_print_usage(const char *argv0, const gpt_params &params,
     printf("  --mmproj MMPROJ_FILE      path to a multimodal projector file for LLaVA.\n");
     printf("  --log-disable             disables logging to a file.\n");
     printf("\n");
+    printf("  -n, --n-predict           maximum tokens to predict (default: %d)\n", params.n_predict);
     printf("  --override-kv KEY=TYPE:VALUE\n");
     printf("                            advanced option to override model metadata by key. may be specified multiple times.\n");
     printf("                            types: int, float, bool. example: --override-kv tokenizer.ggml.add_bos_token=bool:false\n");
     printf("  -gan N, --grp-attn-n N    set the group attention factor to extend context size through self-extend(default: 1=disabled), used together with group attention width `--grp-attn-w`");
     printf("  -gaw N, --grp-attn-w N    set the group attention width to extend context size through self-extend(default: 512), used together with group attention factor `--grp-attn-n`");
     printf("  --chat-template FORMAT_NAME");
-    printf("                            set chat template, possible valus is: llama2, chatml (default %s)", sparams.chat_template.c_str());
+    printf("                            set chat template, possible value is: llama2, chatml (default %s)", sparams.chat_template.c_str());
     printf("\n");
 }
 
@@ -2511,6 +2525,9 @@ static void append_to_generated_text_from_generated_token_probs(llama_server_con
     }
 }
 
+std::function<void(int)> shutdown_handler;
+inline void signal_handler(int signal) { shutdown_handler(signal); }
+
 int main(int argc, char **argv)
 {
 #if SERVER_VERBOSE != 1
@@ -2561,8 +2578,35 @@ int main(int argc, char **argv)
         server_state current_state = state.load();
         switch(current_state) {
             case SERVER_STATE_READY:
-                res.set_content(R"({"status": "ok"})", "application/json");
-                res.status = 200; // HTTP OK
+                if (llama.all_slots_are_idle) {
+                    res.set_content(R"({"status": "ok"})", "application/json");
+                    res.status = 200; // HTTP OK
+                } else {
+                    int available_slots = 0;
+                    int processing_slots = 0;
+                    for (llama_client_slot & slot : llama.slots) {
+                        if (slot.available()) {
+                            available_slots++;
+                        } else {
+                            processing_slots++;
+                        }
+                    }
+                    if (available_slots > 0) {
+                        json health = {
+                                {"status",           "ok"},
+                                {"slots_idle",       available_slots},
+                                {"slots_processing", processing_slots}};
+                        res.set_content(health.dump(), "application/json");
+                        res.status = 200; // HTTP OK
+                    } else {
+                        json health = {
+                                {"status",           "no slot available"},
+                                {"slots_idle",       available_slots},
+                                {"slots_processing", processing_slots}};
+                        res.set_content(health.dump(), "application/json");
+                        res.status = 503; // HTTP Service Unavailable
+                    }
+                }
                 break;
             case SERVER_STATE_LOADING_MODEL:
                 res.set_content(R"({"status": "loading model"})", "application/json");
@@ -3128,8 +3172,25 @@ int main(int argc, char **argv)
         std::placeholders::_2,
         std::placeholders::_3
     ));
-    llama.queue_tasks.start_loop();
 
+    shutdown_handler = [&](int) {
+        llama.queue_tasks.terminate();
+    };
+
+#if defined (__unix__) || (defined (__APPLE__) && defined (__MACH__))
+    struct sigaction sigint_action;
+    sigint_action.sa_handler = signal_handler;
+    sigemptyset (&sigint_action.sa_mask);
+    sigint_action.sa_flags = 0;
+    sigaction(SIGINT, &sigint_action, NULL);
+#elif defined (_WIN32)
+    auto console_ctrl_handler = +[](DWORD ctrl_type) -> BOOL {
+        return (ctrl_type == CTRL_C_EVENT) ? (signal_handler(SIGINT), true) : false;
+    };
+    SetConsoleCtrlHandler(reinterpret_cast<PHANDLER_ROUTINE>(console_ctrl_handler), true);
+#endif
+    llama.queue_tasks.start_loop();
+    svr.stop();
     t.join();
 
     llama_backend_free();

examples/server/utils.hpp

@@ -220,6 +220,7 @@ inline std::string format_chatml(std::vector<json> messages)
 struct llama_server_queue {
     int id = 0;
     std::mutex mutex_tasks;
+    bool running;
     // queues
     std::vector<task_server> queue_tasks;
     std::vector<task_server> queue_tasks_deferred;
@@ -278,9 +279,18 @@ struct llama_server_queue {
         queue_tasks_deferred.clear();
     }
 
-    // Start the main loop. This call is blocking
-    [[noreturn]]
+    // end the start_loop routine
+    void terminate() {
+        {
+            std::unique_lock<std::mutex> lock(mutex_tasks);
+            running = false;
+        }
+        condition_tasks.notify_all();
+    }
+
+    // Start the main loop.
     void start_loop() {
+        running = true;
         while (true) {
             // new task arrived
             LOG_VERBOSE("have new task", {});
@@ -324,8 +334,12 @@ struct llama_server_queue {
             {
                 std::unique_lock<std::mutex> lock(mutex_tasks);
                 if (queue_tasks.empty()) {
+                    if (!running) {
+                        LOG_VERBOSE("ending start_loop", {});
+                        return;
+                    }
                     condition_tasks.wait(lock, [&]{
-                        return !queue_tasks.empty();
+                        return (!queue_tasks.empty() || !running);
                     });
                 }
             }