added data generator note

Author: CaptainDario

CNN_kanji/jupyter/DaKanjiRecognizer.ipynb

@@ -37,7 +37,6 @@
     "from etldr.etl_data_reader import ETLDataReader\n",
     "from etldr.etl_character_groups import ETLCharacterGroups\n",
     "from etldr.etl_data_names import ETLDataNames\n",
-    "import DataGenerator\n",
     "\n",
     "#data handling\n",
     "import PIL\n",
@@ -496,10 +495,11 @@
    "source": [
     "## Defining a thread-safe Datagenerator\n",
     "\n",
-    "**caution: on windows multiprocesing in jupyter is difficult. For actual training use `training.py`** on windwos.\n",
+    "**caution: on windows multiprocesing in jupyter is difficult. For actual training use `training.py`** on windows.\n",
     "\n",
     "Because the data set is too large to fit into memory we will first generate a data generator to create batches on the fly. <br> \n",
-    "Instantiate a generator for training and testing."
+    "Instantiate a generator for training and testing. <br>\n",
+    "This generator uses multi processing. Multi processing does not work in jupyter notebooks (on windows) when defined in the notebook. Therefore the datagenerator is implemented in `DataGenerator.py`."
    ]
   },
   {
@@ -508,6 +508,8 @@
    "metadata": {},
    "outputs": [],
    "source": [
+    "import DataGenerator\n",
+    "\n",
     "train_generator = DataGenerator.DataGenerator(len(x_np), 2048, \n",
     "                                            80, \"training\", \n",
     "                                            x_shared, y_shared, \n",