(Mirror) cli : Use acquire/release semantics for stopping logic

common/arg.cpp

@@ -396,7 +396,7 @@ static bool parse_bool_value(const std::string & value) {
 // CLI argument parsing functions
 //
 
-bool common_params_handle_models(common_params & params, llama_example curr_ex) {
+bool common_params_handle_models(common_params & params, llama_example curr_ex, common_download_callback * callback) {
     const bool spec_type_draft_mtp = std::find(params.speculative.types.begin(),
                                          params.speculative.types.end(),
                                          COMMON_SPECULATIVE_TYPE_DRAFT_MTP) != params.speculative.types.end();
@@ -408,6 +408,10 @@ bool common_params_handle_models(common_params & params, llama_example curr_ex)
     opts.download_mtp    = spec_type_draft_mtp;
     opts.download_mmproj = !params.no_mmproj && params.mmproj.path.empty() && params.mmproj.url.empty();
 
+    if (callback) {
+        opts.callback = callback;
+    }
+
     // sub-models (draft, mmproj, vocoder) are explicitly specified by the user,
     // so we should not auto-discover mtp/mmproj siblings for them
     common_download_opts sub_opts = opts;
@@ -584,8 +588,11 @@ static bool common_params_parse_ex(int argc, char ** argv, common_params_context
         throw std::invalid_argument("error: --prompt-cache-all not supported in interactive mode yet\n");
     }
 
-    // export_graph_ops loads only metadata
-    const bool skip_model_download = ctx_arg.ex == LLAMA_EXAMPLE_EXPORT_GRAPH_OPS;
+    const bool skip_model_download =
+        // server will call common_params_handle_models() later, so we skip it here
+        ctx_arg.ex == LLAMA_EXAMPLE_SERVER ||
+        // export_graph_ops loads only metadata
+        ctx_arg.ex == LLAMA_EXAMPLE_EXPORT_GRAPH_OPS;
 
     if (!skip_model_download) {
         // handle model and download
@@ -594,7 +601,6 @@ static bool common_params_parse_ex(int argc, char ** argv, common_params_context
         // model is required (except for server)
         // TODO @ngxson : maybe show a list of available models in CLI in this case
         if (params.model.path.empty()
-                && ctx_arg.ex != LLAMA_EXAMPLE_SERVER
                 && !params.usage
                 && !params.completion) {
             throw std::invalid_argument("error: --model is required\n");

common/arg.h

@@ -1,6 +1,7 @@
 #pragma once
 
 #include "common.h"
+#include "download.h"
 
 #include <set>
 #include <map>
@@ -133,7 +134,10 @@ void common_params_add_preset_options(std::vector<common_arg> & args);
 // return true if the model is ready to use
 // throw an exception if there is an error that prevents the model from being used (e.g. network error, model not found, etc)
 // if params.skip_download is true, no downloads will be attempted. return false if the model is invalid or missing (e.g. ETag check failed)
-bool common_params_handle_models(common_params & params, llama_example curr_ex);
+bool common_params_handle_models(
+    common_params & params,
+    llama_example curr_ex,
+    common_download_callback * callback = nullptr);
 
 // initialize argument parser context - used by test-arg-parser and preset
 common_params_context common_params_parser_init(common_params & params, llama_example ex, void(*print_usage)(int, char **) = nullptr);

tools/cli/cli.cpp

@@ -12,17 +12,21 @@
 #include <array>
 #include <atomic>
 #include <algorithm>
+#include <csignal>
 #include <filesystem>
 #include <fstream>
 #include <thread>
-#include <signal.h>
 
 #if defined(_WIN32)
 #define WIN32_LEAN_AND_MEAN
 #ifndef NOMINMAX
 #   define NOMINMAX
 #endif
+#include <io.h>
+#include <stdio.h>
 #include <windows.h>
+#elif defined (__unix__) || (defined (__APPLE__) && defined (__MACH__))
+#include <unistd.h>
 #endif
 
 const char * LLAMA_ASCII_LOGO = R"(
@@ -37,19 +41,26 @@ const char * LLAMA_ASCII_LOGO = R"(
 
 static std::atomic<bool> g_is_interrupted = false;
 static bool should_stop() {
-    return g_is_interrupted.load();
+    return g_is_interrupted.load(std::memory_order_acquire);
+}
+static void reset_stop() {
+    g_is_interrupted.store(false, std::memory_order_release);
 }
 
 #if defined (__unix__) || (defined (__APPLE__) && defined (__MACH__)) || defined (_WIN32)
 static void signal_handler(int) {
-    if (g_is_interrupted.load()) {
+    const bool already_interrupted = g_is_interrupted.exchange(true, std::memory_order_acq_rel);
+    if (already_interrupted) {
         // second Ctrl+C - exit immediately
         // make sure to clear colors before exiting (not using LOG or console.cpp here to avoid deadlock)
-        fprintf(stdout, "\033[0m\n");
-        fflush(stdout);
-        std::exit(130);
+        static constexpr char color_reset[] = "\033[0m\n";
+#if defined(_WIN32)
+        _write(_fileno(stdout), color_reset, sizeof(color_reset) - 1);
+#else
+        [[maybe_unused]] ssize_t ret = write(STDOUT_FILENO, color_reset, sizeof(color_reset) - 1);
+#endif
+        _exit(128 + SIGINT);
     }
-    g_is_interrupted.store(true);
 }
 #endif
 
@@ -60,9 +71,6 @@ struct cli_context {
     task_params defaults;
     bool verbose_prompt;
 
-    // thread for showing "loading" animation
-    std::atomic<bool> loading_show;
-
     cli_context(const common_params & params) {
         defaults.sampling    = params.sampling;
         defaults.speculative = params.speculative;
@@ -150,10 +158,7 @@ struct cli_context {
         std::string curr_content;
         bool is_thinking = false;
 
-        while (result) {
-            if (should_stop()) {
-                break;
-            }
+        while (result && !should_stop()) {
             if (result->is_error()) {
                 json err_data = result->to_json();
                 if (err_data.contains("message")) {
@@ -195,7 +200,7 @@ struct cli_context {
             }
             result = rd.next(should_stop);
         }
-        g_is_interrupted.store(false);
+        reset_stop();
         // server_response_reader automatically cancels pending tasks upon destruction
         return curr_content;
     }
@@ -527,7 +532,7 @@ int llama_cli(int argc, char ** argv) {
         console::log("\n");
 
         if (should_stop()) {
-            g_is_interrupted.store(false);
+            reset_stop();
             break;
         }
 

tools/server/README-dev.md

@@ -204,9 +204,9 @@ Instead of building everything from the ground up (like what most AI agents will
 
 The flow for downloading a new model:
 - POST request comes in --> `post_router_models` --> validation
-- `server_models::download()` is called
-    - Sets up a new thread `inst.th` and runs the download inside
-- If a stop request comes in, set `stop_download` to `true`
+- A new `llama-server` subprocess will be spawned with special `SERVER_CHILD_MODE_DOWNLOAD`
+- Child process runs the download and report status back to router via stdin/out
+- If a stop request comes in, the router asks the child process to stop (same mechanism as running a model in child process)
 - Otherwise, upon completion, we call `load_models()` to refresh the list of models
 
 ### Notable Related PRs

tools/server/README.md

@@ -1230,8 +1230,6 @@ print(completion.choices[0].text)
 
 Given a ChatML-formatted json description in `messages`, it returns the predicted completion. Both synchronous and streaming mode are supported, so scripted and interactive applications work fine. While no strong claims of compatibility with OpenAI API spec is being made, in our experience it suffices to support many apps. Only models with a [supported chat template](https://github.com/ggml-org/llama.cpp/wiki/Templates-supported-by-llama_chat_apply_template) can be used optimally with this endpoint. By default, the ChatML template will be used.
 
-If model supports multimodal, you can input the media file via `image_url` content part. We support both base64 and remote URL as input. See OAI documentation for more.
-
 *Options:*
 
 See [OpenAI Chat Completions API documentation](https://platform.openai.com/docs/api-reference/chat). llama.cpp `/completion`-specific features such as `mirostat` are also supported.
@@ -1250,9 +1248,18 @@ The `response_format` parameter supports both plain JSON output (e.g. `{"type":
 
 `parallel_tool_calls` : Whether to enable parallel/multiple tool calls (only supported on some models, verification is based on jinja template).
 
-For multimodal input:
-- Content type `image_url` and `input_audio` are the same as OAI schema
-- Content type `input_video` is an extension from OAI schema. For now, it only accepts base64 input
+For multimodal input (typed content, `messages[i].content[j]`):
+- If `type == "image_url"`:
+    - `image_url.url` can be a remote URL, base64 (raw or URI-encoded via `data:image/...;base64`) or path to local file
+    - Accepts formats supported by `stb_image` (jpeg, png, tga, bmp, gif, ...)
+- If `type == "input_audio"`:
+    - Either `input_audio.data` or `input_audio.url` can be specified, can be a remote URL, raw base64 or path to local file
+    - Accepts formats supported by `miniaudio` (mp3, wav, flac)
+    - `input_audio.format` will be ignored, the file format will be determined automatically
+- If `type == "input_video"`:
+    - Either `input_video.data` or `input_video.url` can be specified, can be a remote URL, raw base64 or path to local file
+    - Accepts formats supported by `ffmpeg`
+- Note: for local file, make sure to set `--media-path`. File path must be prefixed by `file://`
 
 *Examples:*
 

tools/server/server-common.cpp

@@ -817,12 +817,21 @@ json oaicompat_completion_params_parse(const json & body) {
     return llama_params;
 }
 
-// media_path always end with '/', see arg.cpp
+// url can be
+// - http(s):// for remote files
+// - file:// for local files (only allowed if media_path is set)
+// - data: for base64 encoded data with uri scheme (e.g. data:image/png;base64,...)
+// - raw base64 encoded data
 static void handle_media(
         std::vector<raw_buffer> & out_files,
-        json & media_obj,
-        const std::string & media_path) {
-    std::string url = json_value(media_obj, "url", std::string());
+        const std::string & url,
+        const std::string & media_path,
+        bool accept_base64_uri) {
+    if (!media_path.empty()) {
+        // should already be enforced by arg.cpp, but checking just in case
+        GGML_ASSERT(media_path.back() == DIRECTORY_SEPARATOR);
+    }
+
     if (string_starts_with(url, "http")) {
         // download remote image
         // TODO @ngxson : maybe make these params configurable
@@ -858,20 +867,28 @@ static void handle_media(
         data.assign((std::istreambuf_iterator<char>(file)), std::istreambuf_iterator<char>());
         out_files.push_back(data);
 
-    } else {
+    } else if (accept_base64_uri && string_starts_with(url, "data:")) {
         // try to decode base64 image
         std::vector<std::string> parts = string_split<std::string>(url, /*separator*/ ',');
         if (parts.size() != 2) {
-            throw std::runtime_error("Invalid url value");
+            throw std::runtime_error("Invalid uri-encoded base64 value");
         } else if (!string_starts_with(parts[0], "data:image/")) {
-            throw std::runtime_error("Invalid url format: " + parts[0]);
+            throw std::runtime_error("Invalid uri format: " + parts[0]);
         } else if (!string_ends_with(parts[0], "base64")) {
-            throw std::runtime_error("url must be base64 encoded");
+            throw std::runtime_error("uri must be base64 encoded");
         } else {
             auto base64_data = parts[1];
             auto decoded_data = base64_decode(base64_data);
             out_files.push_back(decoded_data);
         }
+
+    } else {
+        // try as raw base64 string
+        auto decoded_data = base64_decode(url);
+        if (decoded_data.empty()) {
+            throw std::runtime_error("Invalid base64 value");
+        }
+        out_files.push_back(decoded_data);
     }
 }
 
@@ -957,14 +974,15 @@ json oaicompat_chat_params_parse(
         }
 
         for (auto & p : content) {
-            std::string type      = json_value(p, "type", std::string());
+            std::string type = json_value(p, "type", std::string());
             if (type == "image_url") {
                 if (!opt.allow_image) {
                     throw std::runtime_error("image input is not supported - hint: if this is unexpected, you may need to provide the mmproj");
                 }
 
                 json image_url = json_value(p, "image_url", json::object());
-                handle_media(out_files, image_url, opt.media_path);
+                std::string url = json_value(image_url, "url", std::string());
+                handle_media(out_files, url, opt.media_path, true);
 
                 p["type"] = "media_marker";
                 p["text"] = get_media_marker();
@@ -975,17 +993,11 @@ json oaicompat_chat_params_parse(
                     throw std::runtime_error("audio input is not supported - hint: if this is unexpected, you may need to provide the mmproj");
                 }
 
-                json input_audio   = json_value(p, "input_audio", json::object());
-                std::string data   = json_value(input_audio, "data", std::string());
-                std::string format = json_value(input_audio, "format", std::string());
-                // while we also support flac, we don't allow it here so we matches the OAI spec
-                if (format != "wav" && format != "mp3") {
-                    throw std::invalid_argument("input_audio.format must be either 'wav' or 'mp3'");
-                }
-                auto decoded_data = base64_decode(data); // expected to be base64 encoded
-                out_files.push_back(decoded_data);
-
-                // TODO: add audio_url support by reusing handle_media()
+                // note: don't need to validate "format", it's redundant
+                json input_audio = json_value(p, "input_audio", json::object());
+                std::string url  = json_value(input_audio, "data",
+                                        json_value(input_audio, "url", std::string()));
+                handle_media(out_files, url, opt.media_path, false);
 
                 p["type"] = "media_marker";
                 p["text"] = get_media_marker();
@@ -996,10 +1008,10 @@ json oaicompat_chat_params_parse(
                     throw std::runtime_error("video input is not supported - hint: if this is unexpected, you may need to provide the mmproj");
                 }
 
-                json input_video  = json_value(p, "input_video", json::object());
-                std::string data  = json_value(input_video, "data", std::string());
-                auto decoded_data = base64_decode(data); // expected to be base64 encoded
-                out_files.push_back(decoded_data);
+                json input_video = json_value(p, "input_video", json::object());
+                std::string url  = json_value(input_video, "data",
+                                        json_value(input_video, "url", std::string()));
+                handle_media(out_files, url, opt.media_path, false);
 
                 p["type"] = "media_marker";
                 p["text"] = get_media_marker();

tools/server/server-context.cpp

@@ -931,6 +931,8 @@ struct server_context_impl {
 
     bool sleeping = false;
 
+    int64_t t_last_load_progress_ms = 0;
+
     void destroy() {
         spec.reset();
         ctx_dft.reset();
@@ -1244,6 +1246,10 @@ struct server_context_impl {
         }
 
         if (has_mmproj) {
+            if (callback_state) {
+                callback_state(SERVER_STATE_LOADING, {{"stage", "mmproj_model"}});
+            }
+
             if (!is_resume) {
                 mtmd_helper_log_set(common_log_default_callback, nullptr);
             }

tools/server/server-context.h

@@ -53,14 +53,15 @@ struct server_context_meta {
 };
 
 enum server_state {
-    // SERVER_STATE_DOWNLOADING,
+    SERVER_STATE_DOWNLOADING,
     SERVER_STATE_LOADING,
     SERVER_STATE_READY,
     SERVER_STATE_SLEEPING,
 };
 
 static std::string server_state_to_str(server_state state) {
     switch (state) {
+        case SERVER_STATE_DOWNLOADING: return "downloading";
         case SERVER_STATE_LOADING:     return "loading";
         case SERVER_STATE_READY:       return "ready";
         case SERVER_STATE_SLEEPING:    return "sleeping";
@@ -69,6 +70,7 @@ static std::string server_state_to_str(server_state state) {
 }
 
 static server_state server_state_from_str(const std::string & str) {
+    if (str == "downloading") return SERVER_STATE_DOWNLOADING;
     if (str == "loading")     return SERVER_STATE_LOADING;
     if (str == "ready")       return SERVER_STATE_READY;
     if (str == "sleeping")    return SERVER_STATE_SLEEPING;