[Example] Training offloading service example

- This app uses ml service APIs
- You must run the receiver first with receiver.conf for the sender to work

Signed-off-by: hyunil park <[email protected]>

Tizen.platform/training_offloading_service/gen_training_offloading_service_rpm.sh

@@ -0,0 +1,8 @@
+#!/usr/bin/env bash
+
+git init
+git add *
+git commit -m 'Initial commit'
+#gbs build -A armv7l --include-all --clean
+gbs build  --include-all -A armv7l -P public_9.0_arm
+rm -rf .git

Tizen.platform/training_offloading_service/main.c

@@ -0,0 +1,284 @@
+/**
+ * @file  main.c
+ * @date  2 May 2024
+ * @brief  Training offloading service app.
+ * @see  https://github.com/nnstreamer/nnstreamer
+ * @author  <[email protected]>
+ * @bug  No known bugs.
+ */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <glib.h>
+#include <ml-api-service.h>
+
+#define MAX_STRING_LEN	2048
+
+enum {
+  CURRENT_STATUS_MAINMENU,
+  CURRENT_STATUS_INPUT_FILENAME,
+  CURRENT_STATUS_INPUT_RW_PATH,
+};
+
+int g_menu_state = CURRENT_STATUS_MAINMENU;
+GMainLoop *loop;
+static gchar g_config_path[MAX_STRING_LEN];
+static gchar g_path[MAX_STRING_LEN];
+ml_service_h service_h;
+#define ML_SERVICE_EVENT_REPLY 4
+
+/**
+ * @brief Callback function for reply test.
+ */
+static void
+_receive_trained_model_cb (ml_service_event_e event, ml_information_h event_data, void *user_data)
+{
+  switch ((int) event) {
+    case ML_SERVICE_EVENT_REPLY:
+      {
+        g_warning ("Get ML_SERVICE_EVENT_REPLY and received trained model");
+        break;
+      }
+    default:
+      break;
+  }
+}
+
+/**
+ * @brief Callback function for sink node.
+ */
+static void
+_sink_register_cb (ml_service_event_e event, ml_information_h event_data, void *user_data)
+{
+  ml_tensors_data_h data = NULL;
+  double *output = NULL;
+  size_t data_size = 0;
+  int status, i;
+  double result_data[4];
+  char *output_node_name = NULL;
+
+  switch (event) {
+    case ML_SERVICE_EVENT_NEW_DATA:
+      g_return_if_fail (event_data != NULL);
+
+      status = ml_information_get (event_data, "name", (void **) &output_node_name);
+      if (status != ML_ERROR_NONE)
+        return;
+      if (!output_node_name)
+        return;
+
+      status = ml_information_get (event_data, "data", &data);
+      if (status != ML_ERROR_NONE)
+        return;
+
+      status = ml_tensors_data_get_tensor_data (data, 0, (void **) &output, &data_size);
+      if (status != ML_ERROR_NONE)
+        return;
+      break;
+    default:
+      break;
+  }
+
+  if (output) {
+    for (i = 0; i < 4; i++)
+      result_data[i] = output[i];
+
+    g_print ("name:%s >> [training_loss: %f, training_accuracy: %f, validation_loss: %f, validation_accuracy: %f]",
+        output_node_name, result_data[0], result_data[1], result_data[2],
+        result_data[3]);
+  }
+}
+
+/**
+ * @brief menu interface function
+ */
+static void 
+input_filepath (gchar *src, gchar *path)
+{
+  gint len = 0;
+  gsize size = 0;
+  g_return_if_fail (path != NULL);
+
+  len = strlen(path);
+  if (len < 0 || len > MAX_STRING_LEN -1)
+    return;
+  size = g_strlcpy(src, path, MAX_STRING_LEN);
+  if (len != size)
+    return;
+}
+
+/**
+ * @brief menu interface function
+ */
+void
+main_menu ()
+{
+  g_print("\n");
+	g_print("============================================================================\n");
+	g_print("  Training Offloading Services Test (press q to quit) \n");
+	g_print("----------------------------------------------------------------------------\n");
+  g_print("a. Create \n");
+  g_print("b. Set R/W path \n");
+  g_print("c. Set callback for receiver \n");
+  g_print("d. Set callback for sender \n");
+  g_print("e. Start \n");
+  g_print("f. Stop \n");
+  g_print("g. End \n");
+  g_print("q. Quit \n");
+	g_print("============================================================================\n");  
+}
+
+/**
+ * @brief menu interface function
+ */
+static void
+quit_program ()
+{
+  g_main_loop_quit (loop);
+}
+
+/**
+ * @brief menu interface function
+ */
+void
+reset_menu_status (void)
+{
+	g_menu_state = CURRENT_STATUS_MAINMENU;
+}
+
+/**
+ * @brief menu interface function
+ */
+static void
+display_menu ()
+{
+  if (g_menu_state == CURRENT_STATUS_MAINMENU) {
+    main_menu();
+  } else if (g_menu_state == CURRENT_STATUS_INPUT_FILENAME) {
+    g_print("*** input config file path.\n");
+  } else if (g_menu_state == CURRENT_STATUS_INPUT_RW_PATH) {
+    g_print("*** input path for reading or writing. e.g. /opt \n");
+  } else {
+    g_print("*** Unknown status. \n");
+  }
+  g_print(" >>> ");
+}
+
+/**
+ * @brief interpret main menu
+ */
+static void
+interpret_main_menu (char *cmd)
+{
+  int ret = ML_ERROR_NONE;
+  if (!g_strcmp0 (cmd, "a")) {
+    g_menu_state = CURRENT_STATUS_INPUT_FILENAME;
+  } else if (!g_strcmp0 (cmd, "b")) {
+    g_menu_state = CURRENT_STATUS_INPUT_RW_PATH;
+  } else if (!g_strcmp0 (cmd, "c")) {
+    ret = ml_service_set_event_cb (service_h, _sink_register_cb, NULL);
+  } else if (!g_strcmp0 (cmd, "d")) {
+    ret = ml_service_set_event_cb (service_h, _receive_trained_model_cb, NULL);
+  } else if (!g_strcmp0 (cmd, "e")) {
+    ret = ml_service_start (service_h);
+  } else if (!g_strcmp0 (cmd, "f")) {
+    ret = ml_service_stop (service_h);
+  } else if (!g_strcmp0 (cmd, "g")) {
+    ret = ml_service_destroy (service_h);
+  } else if (!g_strcmp0 (cmd, "q")) {
+    quit_program();
+    return;
+  }
+
+  if (ret == ML_ERROR_NONE)
+      g_print("*** Success *** \n");
+  else
+      g_print("*** Failed *** \n");
+}
+
+/**
+ * @brief timeout for menu display
+ */
+gboolean
+timeout_menu_display (void *data)
+{
+	display_menu ();
+	return FALSE;
+}
+
+/**
+ * @brief interpret function
+ */
+static void
+interpret (gchar *cmd)
+{
+  switch (g_menu_state) {
+    case CURRENT_STATUS_MAINMENU:
+      interpret_main_menu (cmd);
+      break;
+    case CURRENT_STATUS_INPUT_FILENAME:
+      input_filepath (g_config_path, cmd);
+      reset_menu_status ();
+      g_print("config path=%s\n", g_config_path);
+      if (ML_ERROR_NONE == ml_service_new (g_config_path, &service_h))
+        g_print("Success to create service(handle=%p)\n", service_h);
+      else
+        g_print("Failed to create service \n");
+      break;
+    case CURRENT_STATUS_INPUT_RW_PATH:
+      input_filepath (g_path, cmd);
+      reset_menu_status ();
+      if (ML_ERROR_NONE == ml_service_set_information (service_h, "path", g_path))
+        g_print ("Success to set path \n");
+      else
+        g_print ("Failed to set path \n");
+      break;
+    default:
+      break;
+  }
+  g_timeout_add (100, timeout_menu_display, 0);
+}
+
+/**
+ * @brief input function
+ */
+gboolean
+input (GIOChannel *channel)
+{
+	gchar buf[MAX_STRING_LEN];
+	ssize_t cnt;
+
+	memset (buf, 0, MAX_STRING_LEN);
+	cnt = read (0, (void *)buf, MAX_STRING_LEN);
+	if (cnt == 0) return TRUE;
+	buf[cnt - 1] = 0;
+
+	interpret (buf);
+
+	return TRUE;
+}
+
+/**
+ * @brief Main function.
+ */
+int
+main (int argc, char **argv)
+{
+  GIOChannel *stdin_channel;
+
+
+  stdin_channel = g_io_channel_unix_new (0);
+	g_io_channel_set_flags (stdin_channel, G_IO_FLAG_NONBLOCK, NULL);
+	g_io_add_watch (stdin_channel, G_IO_IN, (GIOFunc)input, NULL);
+
+  loop = g_main_loop_new (NULL, FALSE);
+
+  g_print("running\n");
+  display_menu ();
+  g_main_loop_run (loop);
+  g_print("exit training offloading services\n");
+  g_main_loop_unref (loop);
+
+  return 0;
+}

Tizen.platform/training_offloading_service/meson.build

@@ -0,0 +1,30 @@
+project('nnstreamer-example', 'c', 'cpp',
+  version: '0.1.0',
+  license: ['LGPL-2.1'],
+  meson_version: '>=0.50.0',
+  default_options: [
+    'werror=true',
+    'warning_level=1',
+    'c_std=gnu89',
+    'cpp_std=c++11'
+  ]
+)
+
+# Set install path
+nnst_exam_prefix = get_option('prefix')
+nnst_exam_bindir = join_paths(nnst_exam_prefix, get_option('bindir'))
+examples_install_dir = nnst_exam_bindir
+
+# Dependencies
+glib_dep = dependency('glib-2.0')
+nns_edge_dep = dependency('nnstreamer-edge', required: false)
+nns_capi_common_dep = dependency('capi-ml-common', required: false)
+ml_service_dep = dependency('capi-ml-service', required: false)
+
+executable('training_offloadinge_service',
+  'main.c',
+  dependencies: [glib_dep, nns_edge_dep, nns_capi_common_dep, ml_service_dep],
+  install: true,
+  install_dir: examples_install_dir
+)
+

Tizen.platform/training_offloading_service/packaging/training_offloading_service.manifest

@@ -0,0 +1,5 @@
+<manifest>
+ <request>
+    <domain name="_"/>
+ </request>
+</manifest>

Tizen.platform/training_offloading_service/packaging/training_offloading_service.spec

@@ -0,0 +1,40 @@
+%define		nnstexampledir	/usr/lib/nnstreamer/bin
+
+Name:		training_offloading_service
+Summary:	training_offloadingi_service app
+Version:	1.0.0
+Release:	0
+Group:		Machine Learning/ML Framework
+Packager:	Hyunil Park <[email protected]>
+License:	LGPL-2.1
+Source0:	%{name}-%{version}.tar.gz
+Source1001:	%{name}.manifest
+
+Requires:	nnstreamer
+Requires:	nnstreamer-edge
+BuildRequires:	meson
+BuildRequires:	pkgconfig(glib-2.0)
+BuildRequires:	pkgconfig(capi-ml-service)
+
+
+%description
+training offloading service sample app with Tizen IoT platform
+
+%prep
+%setup -q
+cp %{SOURCE1001} .
+
+%build
+mkdir -p build
+
+meson --buildtype=plain --prefix=%{_prefix} --libdir=%{_libdir} --bindir=%{nnstexampledir} --includedir=%{_includedir} build
+ninja -C build %{?_smp_mflags}
+
+%install
+DESTDIR=%{buildroot} ninja -C build install
+
+%files
+%manifest %{name}.manifest
+%defattr(-,root,root,-)
+%{nnstexampledir}/*
+