{ "cells": [ { "cell_type": "markdown", "metadata": { "id": "AlaMdWF26k2H" }, "source": [ "# LSTM과 GRU 셀" ] }, { "cell_type": "markdown", "metadata": { "id": "o_Csb7v96k2M" }, "source": [ "\n", " \n", "
\n", " 구글 코랩에서 실행하기\n", "
" ] }, { "cell_type": "code", "execution_count": 1, "metadata": { "id": "fJIx0dHbIAmO" }, "outputs": [], "source": [ "# 실행마다 동일한 결과를 얻기 위해 케라스에 랜덤 시드를 사용하고 텐서플로 연산을 결정적으로 만듭니다.\n", "import tensorflow as tf\n", "\n", "tf.keras.utils.set_random_seed(42)" ] }, { "cell_type": "markdown", "metadata": { "id": "KWi9yl0q6k2N" }, "source": [ "## LSTM 신경망 훈련하기" ] }, { "cell_type": "code", "execution_count": 2, "metadata": { "id": "XL7x0NCx6k2N" }, "outputs": [], "source": [ "from tensorflow.keras.datasets import imdb\n", "from sklearn.model_selection import train_test_split\n", "\n", "(train_input, train_target), (test_input, test_target) = imdb.load_data(\n", " num_words=500)\n", "\n", "train_input, val_input, train_target, val_target = train_test_split(\n", " train_input, train_target, test_size=0.2, random_state=42)" ] }, { "cell_type": "code", "execution_count": 3, "metadata": { "id": "5bjJ14QJ6k2O" }, "outputs": [], "source": [ "from tensorflow.keras.preprocessing.sequence import pad_sequences\n", "\n", "train_seq = pad_sequences(train_input, maxlen=100)\n", "val_seq = pad_sequences(val_input, maxlen=100)" ] }, { "cell_type": "code", "execution_count": 4, "metadata": { "colab": { "base_uri": "https://localhost:8080/", "height": 286 }, "id": "uMlTD4zv6k2O", "outputId": "c03b1784-d740-453d-8d70-2c716231fb48" }, "outputs": [ { "output_type": "stream", "name": "stderr", "text": [ "/usr/local/lib/python3.10/dist-packages/keras/src/layers/core/embedding.py:93: UserWarning: Do not pass an `input_shape`/`input_dim` argument to a layer. When using Sequential models, prefer using an `Input(shape)` object as the first layer in the model instead.\n", " super().__init__(**kwargs)\n" ] }, { "output_type": "display_data", "data": { "text/plain": [ "\u001b[1mModel: \"sequential\"\u001b[0m\n" ], "text/html": [ "
Model: \"sequential\"\n",
              "
\n" ] }, "metadata": {} }, { "output_type": "display_data", "data": { "text/plain": [ "┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━┓\n", "┃\u001b[1m \u001b[0m\u001b[1mLayer (type) \u001b[0m\u001b[1m \u001b[0m┃\u001b[1m \u001b[0m\u001b[1mOutput Shape \u001b[0m\u001b[1m \u001b[0m┃\u001b[1m \u001b[0m\u001b[1m Param #\u001b[0m\u001b[1m \u001b[0m┃\n", "┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━┩\n", "│ embedding (\u001b[38;5;33mEmbedding\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m100\u001b[0m, \u001b[38;5;34m16\u001b[0m) │ \u001b[38;5;34m8,000\u001b[0m │\n", "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n", "│ lstm (\u001b[38;5;33mLSTM\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m8\u001b[0m) │ \u001b[38;5;34m800\u001b[0m │\n", "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n", "│ dense (\u001b[38;5;33mDense\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m1\u001b[0m) │ \u001b[38;5;34m9\u001b[0m │\n", "└──────────────────────────────────────┴─────────────────────────────┴─────────────────┘\n" ], "text/html": [ "
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━┓\n",
              "┃ Layer (type)                         ┃ Output Shape                ┃         Param # ┃\n",
              "┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━┩\n",
              "│ embedding (Embedding)                │ (None, 100, 16)             │           8,000 │\n",
              "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n",
              "│ lstm (LSTM)                          │ (None, 8)                   │             800 │\n",
              "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n",
              "│ dense (Dense)                        │ (None, 1)                   │               9 │\n",
              "└──────────────────────────────────────┴─────────────────────────────┴─────────────────┘\n",
              "
\n" ] }, "metadata": {} }, { "output_type": "display_data", "data": { "text/plain": [ "\u001b[1m Total params: \u001b[0m\u001b[38;5;34m8,809\u001b[0m (34.41 KB)\n" ], "text/html": [ "
 Total params: 8,809 (34.41 KB)\n",
              "
\n" ] }, "metadata": {} }, { "output_type": "display_data", "data": { "text/plain": [ "\u001b[1m Trainable params: \u001b[0m\u001b[38;5;34m8,809\u001b[0m (34.41 KB)\n" ], "text/html": [ "
 Trainable params: 8,809 (34.41 KB)\n",
              "
\n" ] }, "metadata": {} }, { "output_type": "display_data", "data": { "text/plain": [ "\u001b[1m Non-trainable params: \u001b[0m\u001b[38;5;34m0\u001b[0m (0.00 B)\n" ], "text/html": [ "
 Non-trainable params: 0 (0.00 B)\n",
              "
\n" ] }, "metadata": {} } ], "source": [ "from tensorflow import keras\n", "\n", "model = keras.Sequential()\n", "\n", "model.add(keras.layers.Embedding(500, 16, input_shape=(100,)))\n", "model.add(keras.layers.LSTM(8))\n", "model.add(keras.layers.Dense(1, activation='sigmoid'))\n", "\n", "model.summary()" ] }, { "cell_type": "code", "execution_count": 5, "metadata": { "colab": { "base_uri": "https://localhost:8080/" }, "id": "nCmqOkjI6k2P", "outputId": "ec304aa1-bb25-415b-fcfe-795c6c9424db" }, "outputs": [ { "output_type": "stream", "name": "stdout", "text": [ "Epoch 1/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m9s\u001b[0m 8ms/step - accuracy: 0.5110 - loss: 0.6929 - val_accuracy: 0.5864 - val_loss: 0.6911\n", "Epoch 2/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m5s\u001b[0m 7ms/step - accuracy: 0.5784 - loss: 0.6907 - val_accuracy: 0.6258 - val_loss: 0.6874\n", "Epoch 3/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 8ms/step - accuracy: 0.6217 - loss: 0.6858 - val_accuracy: 0.6562 - val_loss: 0.6765\n", "Epoch 4/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 9ms/step - accuracy: 0.6681 - loss: 0.6682 - val_accuracy: 0.7096 - val_loss: 0.6177\n", "Epoch 5/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 7ms/step - accuracy: 0.7180 - loss: 0.6006 - val_accuracy: 0.7248 - val_loss: 0.5760\n", "Epoch 6/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.7282 - loss: 0.5690 - val_accuracy: 0.7404 - val_loss: 0.5555\n", "Epoch 7/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 8ms/step - accuracy: 0.7481 - loss: 0.5465 - val_accuracy: 0.7450 - val_loss: 0.5383\n", "Epoch 8/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 10ms/step - accuracy: 0.7600 - loss: 0.5274 - val_accuracy: 0.7588 - val_loss: 0.5216\n", "Epoch 9/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.7688 - loss: 0.5110 - val_accuracy: 0.7698 - val_loss: 0.5076\n", "Epoch 10/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 7ms/step - accuracy: 0.7777 - loss: 0.4972 - val_accuracy: 0.7746 - val_loss: 0.4959\n", "Epoch 11/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.7831 - loss: 0.4853 - val_accuracy: 0.7802 - val_loss: 0.4862\n", "Epoch 12/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 8ms/step - accuracy: 0.7885 - loss: 0.4750 - val_accuracy: 0.7850 - val_loss: 0.4781\n", "Epoch 13/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 10ms/step - accuracy: 0.7936 - loss: 0.4662 - val_accuracy: 0.7874 - val_loss: 0.4713\n", "Epoch 14/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.7961 - loss: 0.4585 - val_accuracy: 0.7876 - val_loss: 0.4659\n", "Epoch 15/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.8009 - loss: 0.4519 - val_accuracy: 0.7904 - val_loss: 0.4614\n", "Epoch 16/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.8036 - loss: 0.4461 - val_accuracy: 0.7920 - val_loss: 0.4577\n", "Epoch 17/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.8054 - loss: 0.4411 - val_accuracy: 0.7900 - val_loss: 0.4548\n", "Epoch 18/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 11ms/step - accuracy: 0.8083 - loss: 0.4369 - val_accuracy: 0.7898 - val_loss: 0.4524\n", "Epoch 19/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.8104 - loss: 0.4332 - val_accuracy: 0.7908 - val_loss: 0.4504\n", "Epoch 20/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.8117 - loss: 0.4300 - val_accuracy: 0.7904 - val_loss: 0.4488\n", "Epoch 21/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 10ms/step - accuracy: 0.8128 - loss: 0.4273 - val_accuracy: 0.7902 - val_loss: 0.4475\n", "Epoch 22/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m5s\u001b[0m 10ms/step - accuracy: 0.8135 - loss: 0.4249 - val_accuracy: 0.7888 - val_loss: 0.4463\n", "Epoch 23/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.8139 - loss: 0.4228 - val_accuracy: 0.7894 - val_loss: 0.4453\n", "Epoch 24/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.8139 - loss: 0.4210 - val_accuracy: 0.7886 - val_loss: 0.4444\n", "Epoch 25/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.8144 - loss: 0.4194 - val_accuracy: 0.7884 - val_loss: 0.4436\n", "Epoch 26/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 7ms/step - accuracy: 0.8152 - loss: 0.4179 - val_accuracy: 0.7902 - val_loss: 0.4429\n", "Epoch 27/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 11ms/step - accuracy: 0.8148 - loss: 0.4166 - val_accuracy: 0.7914 - val_loss: 0.4422\n", "Epoch 28/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.8140 - loss: 0.4154 - val_accuracy: 0.7914 - val_loss: 0.4416\n", "Epoch 29/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 7ms/step - accuracy: 0.8145 - loss: 0.4144 - val_accuracy: 0.7932 - val_loss: 0.4410\n", "Epoch 30/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.8139 - loss: 0.4134 - val_accuracy: 0.7940 - val_loss: 0.4405\n", "Epoch 31/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.8141 - loss: 0.4125 - val_accuracy: 0.7936 - val_loss: 0.4400\n", "Epoch 32/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 11ms/step - accuracy: 0.8137 - loss: 0.4117 - val_accuracy: 0.7940 - val_loss: 0.4395\n", "Epoch 33/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 7ms/step - accuracy: 0.8138 - loss: 0.4109 - val_accuracy: 0.7936 - val_loss: 0.4391\n", "Epoch 34/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.8141 - loss: 0.4102 - val_accuracy: 0.7946 - val_loss: 0.4387\n", "Epoch 35/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.8147 - loss: 0.4095 - val_accuracy: 0.7944 - val_loss: 0.4384\n", "Epoch 36/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 9ms/step - accuracy: 0.8151 - loss: 0.4089 - val_accuracy: 0.7946 - val_loss: 0.4381\n", "Epoch 37/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 9ms/step - accuracy: 0.8153 - loss: 0.4083 - val_accuracy: 0.7934 - val_loss: 0.4379\n", "Epoch 38/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m5s\u001b[0m 7ms/step - accuracy: 0.8170 - loss: 0.4077 - val_accuracy: 0.7940 - val_loss: 0.4376\n", "Epoch 39/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.8170 - loss: 0.4072 - val_accuracy: 0.7932 - val_loss: 0.4375\n", "Epoch 40/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 9ms/step - accuracy: 0.8162 - loss: 0.4067 - val_accuracy: 0.7938 - val_loss: 0.4373\n", "Epoch 41/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 10ms/step - accuracy: 0.8163 - loss: 0.4061 - val_accuracy: 0.7940 - val_loss: 0.4372\n", "Epoch 42/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 7ms/step - accuracy: 0.8160 - loss: 0.4057 - val_accuracy: 0.7940 - val_loss: 0.4370\n", "Epoch 43/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.8167 - loss: 0.4052 - val_accuracy: 0.7936 - val_loss: 0.4370\n", "Epoch 44/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 8ms/step - accuracy: 0.8166 - loss: 0.4047 - val_accuracy: 0.7930 - val_loss: 0.4369\n", "Epoch 45/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 10ms/step - accuracy: 0.8167 - loss: 0.4043 - val_accuracy: 0.7934 - val_loss: 0.4368\n", "Epoch 46/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.8169 - loss: 0.4038 - val_accuracy: 0.7934 - val_loss: 0.4368\n", "Epoch 47/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.8173 - loss: 0.4034 - val_accuracy: 0.7938 - val_loss: 0.4368\n", "Epoch 48/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 7ms/step - accuracy: 0.8173 - loss: 0.4030 - val_accuracy: 0.7940 - val_loss: 0.4368\n", "Epoch 49/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 8ms/step - accuracy: 0.8171 - loss: 0.4025 - val_accuracy: 0.7944 - val_loss: 0.4368\n", "Epoch 50/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 11ms/step - accuracy: 0.8166 - loss: 0.4021 - val_accuracy: 0.7944 - val_loss: 0.4368\n", "Epoch 51/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 7ms/step - accuracy: 0.8164 - loss: 0.4017 - val_accuracy: 0.7942 - val_loss: 0.4368\n", "Epoch 52/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.8167 - loss: 0.4013 - val_accuracy: 0.7944 - val_loss: 0.4368\n" ] } ], "source": [ "rmsprop = keras.optimizers.RMSprop(learning_rate=1e-4)\n", "model.compile(optimizer=rmsprop, loss='binary_crossentropy',\n", " metrics=['accuracy'])\n", "\n", "checkpoint_cb = keras.callbacks.ModelCheckpoint('best-lstm-model.keras',\n", " save_best_only=True)\n", "early_stopping_cb = keras.callbacks.EarlyStopping(patience=3,\n", " restore_best_weights=True)\n", "\n", "history = model.fit(train_seq, train_target, epochs=100, batch_size=64,\n", " validation_data=(val_seq, val_target),\n", " callbacks=[checkpoint_cb, early_stopping_cb])" ] }, { "cell_type": "code", "execution_count": 6, "metadata": { "colab": { "base_uri": "https://localhost:8080/", "height": 449 }, "id": "Xl1bliNr6k2Q", "outputId": "19d7bfc2-fe79-41b7-abd6-3d177cd8e1d4" }, "outputs": [ { "output_type": "display_data", "data": { "text/plain": [ "
" ], "image/png": "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\n" }, "metadata": {} } ], "source": [ "import matplotlib.pyplot as plt\n", "\n", "plt.plot(history.history['loss'])\n", "plt.plot(history.history['val_loss'])\n", "plt.xlabel('epoch')\n", "plt.ylabel('loss')\n", "plt.legend(['train', 'val'])\n", "plt.show()" ] }, { "cell_type": "markdown", "metadata": { "id": "ob79h7Kc6k2Q" }, "source": [ "## 순환 층에 드롭아웃 적용하기" ] }, { "cell_type": "code", "execution_count": 7, "metadata": { "id": "Fr0DUR-Y6k2R" }, "outputs": [], "source": [ "model2 = keras.Sequential()\n", "\n", "model2.add(keras.layers.Embedding(500, 16, input_shape=(100,)))\n", "model2.add(keras.layers.LSTM(8, dropout=0.3))\n", "model2.add(keras.layers.Dense(1, activation='sigmoid'))" ] }, { "cell_type": "code", "execution_count": 8, "metadata": { "colab": { "base_uri": "https://localhost:8080/" }, "id": "Ap8XxUO46k2R", "outputId": "d8ef0c12-37a6-456e-ce5a-8c746d370392" }, "outputs": [ { "output_type": "stream", "name": "stdout", "text": [ "Epoch 1/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m6s\u001b[0m 12ms/step - accuracy: 0.5145 - loss: 0.6930 - val_accuracy: 0.5546 - val_loss: 0.6923\n", "Epoch 2/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 8ms/step - accuracy: 0.5619 - loss: 0.6920 - val_accuracy: 0.5872 - val_loss: 0.6909\n", "Epoch 3/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.6026 - loss: 0.6902 - val_accuracy: 0.6164 - val_loss: 0.6879\n", "Epoch 4/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.6257 - loss: 0.6865 - val_accuracy: 0.6406 - val_loss: 0.6809\n", "Epoch 5/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 8ms/step - accuracy: 0.6458 - loss: 0.6773 - val_accuracy: 0.6486 - val_loss: 0.6513\n", "Epoch 6/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 11ms/step - accuracy: 0.6848 - loss: 0.6394 - val_accuracy: 0.7208 - val_loss: 0.6041\n", "Epoch 7/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 8ms/step - accuracy: 0.7198 - loss: 0.5997 - val_accuracy: 0.7310 - val_loss: 0.5860\n", "Epoch 8/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 8ms/step - accuracy: 0.7326 - loss: 0.5804 - val_accuracy: 0.7364 - val_loss: 0.5722\n", "Epoch 9/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 8ms/step - accuracy: 0.7458 - loss: 0.5618 - val_accuracy: 0.7530 - val_loss: 0.5525\n", "Epoch 10/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 10ms/step - accuracy: 0.7549 - loss: 0.5479 - val_accuracy: 0.7516 - val_loss: 0.5480\n", "Epoch 11/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m5s\u001b[0m 8ms/step - accuracy: 0.7634 - loss: 0.5337 - val_accuracy: 0.7580 - val_loss: 0.5328\n", "Epoch 12/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.7698 - loss: 0.5203 - val_accuracy: 0.7562 - val_loss: 0.5290\n", "Epoch 13/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.7742 - loss: 0.5099 - val_accuracy: 0.7674 - val_loss: 0.5149\n", "Epoch 14/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 10ms/step - accuracy: 0.7796 - loss: 0.5000 - val_accuracy: 0.7638 - val_loss: 0.5109\n", "Epoch 15/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 9ms/step - accuracy: 0.7813 - loss: 0.4916 - val_accuracy: 0.7728 - val_loss: 0.4965\n", "Epoch 16/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 8ms/step - accuracy: 0.7863 - loss: 0.4829 - val_accuracy: 0.7736 - val_loss: 0.4912\n", "Epoch 17/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 8ms/step - accuracy: 0.7900 - loss: 0.4736 - val_accuracy: 0.7786 - val_loss: 0.4847\n", "Epoch 18/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 8ms/step - accuracy: 0.7908 - loss: 0.4681 - val_accuracy: 0.7790 - val_loss: 0.4792\n", "Epoch 19/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 10ms/step - accuracy: 0.7957 - loss: 0.4608 - val_accuracy: 0.7774 - val_loss: 0.4762\n", "Epoch 20/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 9ms/step - accuracy: 0.7987 - loss: 0.4565 - val_accuracy: 0.7802 - val_loss: 0.4736\n", "Epoch 21/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m5s\u001b[0m 8ms/step - accuracy: 0.7961 - loss: 0.4506 - val_accuracy: 0.7836 - val_loss: 0.4689\n", "Epoch 22/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 8ms/step - accuracy: 0.7972 - loss: 0.4478 - val_accuracy: 0.7882 - val_loss: 0.4632\n", "Epoch 23/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 10ms/step - accuracy: 0.8023 - loss: 0.4423 - val_accuracy: 0.7852 - val_loss: 0.4641\n", "Epoch 24/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 9ms/step - accuracy: 0.8029 - loss: 0.4391 - val_accuracy: 0.7856 - val_loss: 0.4571\n", "Epoch 25/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m5s\u001b[0m 8ms/step - accuracy: 0.8027 - loss: 0.4375 - val_accuracy: 0.7858 - val_loss: 0.4577\n", "Epoch 26/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.8056 - loss: 0.4343 - val_accuracy: 0.7858 - val_loss: 0.4592\n", "Epoch 27/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 10ms/step - accuracy: 0.8060 - loss: 0.4317 - val_accuracy: 0.7866 - val_loss: 0.4538\n", "Epoch 28/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 8ms/step - accuracy: 0.8093 - loss: 0.4307 - val_accuracy: 0.7920 - val_loss: 0.4526\n", "Epoch 29/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 8ms/step - accuracy: 0.8097 - loss: 0.4274 - val_accuracy: 0.7894 - val_loss: 0.4539\n", "Epoch 30/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 8ms/step - accuracy: 0.8052 - loss: 0.4292 - val_accuracy: 0.7904 - val_loss: 0.4473\n", "Epoch 31/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 9ms/step - accuracy: 0.8128 - loss: 0.4233 - val_accuracy: 0.7908 - val_loss: 0.4487\n", "Epoch 32/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 10ms/step - accuracy: 0.8098 - loss: 0.4228 - val_accuracy: 0.7922 - val_loss: 0.4489\n", "Epoch 33/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 8ms/step - accuracy: 0.8109 - loss: 0.4198 - val_accuracy: 0.7902 - val_loss: 0.4500\n" ] } ], "source": [ "rmsprop = keras.optimizers.RMSprop(learning_rate=1e-4)\n", "model2.compile(optimizer=rmsprop, loss='binary_crossentropy',\n", " metrics=['accuracy'])\n", "\n", "checkpoint_cb = keras.callbacks.ModelCheckpoint('best-dropout-model.keras',\n", " save_best_only=True)\n", "early_stopping_cb = keras.callbacks.EarlyStopping(patience=3,\n", " restore_best_weights=True)\n", "\n", "history = model2.fit(train_seq, train_target, epochs=100, batch_size=64,\n", " validation_data=(val_seq, val_target),\n", " callbacks=[checkpoint_cb, early_stopping_cb])" ] }, { "cell_type": "code", "execution_count": 9, "metadata": { "colab": { "base_uri": "https://localhost:8080/", "height": 449 }, "id": "Qzwsp3pl6k2R", "outputId": "16a3ba30-8ed9-4038-9a1b-23466e52c0b2" }, "outputs": [ { "output_type": "display_data", "data": { "text/plain": [ "
" ], "image/png": "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\n" }, "metadata": {} } ], "source": [ "plt.plot(history.history['loss'])\n", "plt.plot(history.history['val_loss'])\n", "plt.xlabel('epoch')\n", "plt.ylabel('loss')\n", "plt.legend(['train', 'val'])\n", "plt.show()" ] }, { "cell_type": "markdown", "metadata": { "id": "DKtQByM96k2S" }, "source": [ "## 2개의 층을 연결하기" ] }, { "cell_type": "code", "execution_count": 10, "metadata": { "colab": { "base_uri": "https://localhost:8080/", "height": 262 }, "id": "YkQboCpD6k2S", "outputId": "b4e92187-0d5a-4772-88e3-8dd97bccfa34" }, "outputs": [ { "output_type": "display_data", "data": { "text/plain": [ "\u001b[1mModel: \"sequential_2\"\u001b[0m\n" ], "text/html": [ "
Model: \"sequential_2\"\n",
              "
\n" ] }, "metadata": {} }, { "output_type": "display_data", "data": { "text/plain": [ "┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━┓\n", "┃\u001b[1m \u001b[0m\u001b[1mLayer (type) \u001b[0m\u001b[1m \u001b[0m┃\u001b[1m \u001b[0m\u001b[1mOutput Shape \u001b[0m\u001b[1m \u001b[0m┃\u001b[1m \u001b[0m\u001b[1m Param #\u001b[0m\u001b[1m \u001b[0m┃\n", "┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━┩\n", "│ embedding_2 (\u001b[38;5;33mEmbedding\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m100\u001b[0m, \u001b[38;5;34m16\u001b[0m) │ \u001b[38;5;34m8,000\u001b[0m │\n", "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n", "│ lstm_2 (\u001b[38;5;33mLSTM\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m100\u001b[0m, \u001b[38;5;34m8\u001b[0m) │ \u001b[38;5;34m800\u001b[0m │\n", "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n", "│ lstm_3 (\u001b[38;5;33mLSTM\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m8\u001b[0m) │ \u001b[38;5;34m544\u001b[0m │\n", "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n", "│ dense_2 (\u001b[38;5;33mDense\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m1\u001b[0m) │ \u001b[38;5;34m9\u001b[0m │\n", "└──────────────────────────────────────┴─────────────────────────────┴─────────────────┘\n" ], "text/html": [ "
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━┓\n",
              "┃ Layer (type)                         ┃ Output Shape                ┃         Param # ┃\n",
              "┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━┩\n",
              "│ embedding_2 (Embedding)              │ (None, 100, 16)             │           8,000 │\n",
              "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n",
              "│ lstm_2 (LSTM)                        │ (None, 100, 8)              │             800 │\n",
              "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n",
              "│ lstm_3 (LSTM)                        │ (None, 8)                   │             544 │\n",
              "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n",
              "│ dense_2 (Dense)                      │ (None, 1)                   │               9 │\n",
              "└──────────────────────────────────────┴─────────────────────────────┴─────────────────┘\n",
              "
\n" ] }, "metadata": {} }, { "output_type": "display_data", "data": { "text/plain": [ "\u001b[1m Total params: \u001b[0m\u001b[38;5;34m9,353\u001b[0m (36.54 KB)\n" ], "text/html": [ "
 Total params: 9,353 (36.54 KB)\n",
              "
\n" ] }, "metadata": {} }, { "output_type": "display_data", "data": { "text/plain": [ "\u001b[1m Trainable params: \u001b[0m\u001b[38;5;34m9,353\u001b[0m (36.54 KB)\n" ], "text/html": [ "
 Trainable params: 9,353 (36.54 KB)\n",
              "
\n" ] }, "metadata": {} }, { "output_type": "display_data", "data": { "text/plain": [ "\u001b[1m Non-trainable params: \u001b[0m\u001b[38;5;34m0\u001b[0m (0.00 B)\n" ], "text/html": [ "
 Non-trainable params: 0 (0.00 B)\n",
              "
\n" ] }, "metadata": {} } ], "source": [ "model3 = keras.Sequential()\n", "\n", "model3.add(keras.layers.Embedding(500, 16, input_shape=(100,)))\n", "model3.add(keras.layers.LSTM(8, dropout=0.3, return_sequences=True))\n", "model3.add(keras.layers.LSTM(8, dropout=0.3))\n", "model3.add(keras.layers.Dense(1, activation='sigmoid'))\n", "\n", "model3.summary()" ] }, { "cell_type": "code", "execution_count": 11, "metadata": { "colab": { "base_uri": "https://localhost:8080/" }, "id": "hDk_ffAo6k2S", "outputId": "22d03e20-af67-4b94-ea47-7f95c052411f" }, "outputs": [ { "output_type": "stream", "name": "stdout", "text": [ "Epoch 1/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m6s\u001b[0m 13ms/step - accuracy: 0.5235 - loss: 0.6927 - val_accuracy: 0.5738 - val_loss: 0.6904\n", "Epoch 2/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m5s\u001b[0m 12ms/step - accuracy: 0.5854 - loss: 0.6888 - val_accuracy: 0.6384 - val_loss: 0.6765\n", "Epoch 3/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m5s\u001b[0m 11ms/step - accuracy: 0.6495 - loss: 0.6604 - val_accuracy: 0.6932 - val_loss: 0.5969\n", "Epoch 4/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m6s\u001b[0m 15ms/step - accuracy: 0.7134 - loss: 0.5717 - val_accuracy: 0.7300 - val_loss: 0.5459\n", "Epoch 5/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 11ms/step - accuracy: 0.7405 - loss: 0.5346 - val_accuracy: 0.7478 - val_loss: 0.5202\n", "Epoch 6/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 11ms/step - accuracy: 0.7571 - loss: 0.5090 - val_accuracy: 0.7558 - val_loss: 0.5120\n", "Epoch 7/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 14ms/step - accuracy: 0.7712 - loss: 0.4930 - val_accuracy: 0.7612 - val_loss: 0.5026\n", "Epoch 8/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 11ms/step - accuracy: 0.7732 - loss: 0.4813 - val_accuracy: 0.7716 - val_loss: 0.4902\n", "Epoch 9/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 11ms/step - accuracy: 0.7836 - loss: 0.4727 - val_accuracy: 0.7662 - val_loss: 0.4928\n", "Epoch 10/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m6s\u001b[0m 15ms/step - accuracy: 0.7806 - loss: 0.4711 - val_accuracy: 0.7742 - val_loss: 0.4794\n", "Epoch 11/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 11ms/step - accuracy: 0.7831 - loss: 0.4691 - val_accuracy: 0.7776 - val_loss: 0.4795\n", "Epoch 12/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 11ms/step - accuracy: 0.7909 - loss: 0.4590 - val_accuracy: 0.7828 - val_loss: 0.4705\n", "Epoch 13/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 12ms/step - accuracy: 0.7906 - loss: 0.4552 - val_accuracy: 0.7762 - val_loss: 0.4754\n", "Epoch 14/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 14ms/step - accuracy: 0.7940 - loss: 0.4542 - val_accuracy: 0.7844 - val_loss: 0.4670\n", "Epoch 15/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 11ms/step - accuracy: 0.7940 - loss: 0.4510 - val_accuracy: 0.7806 - val_loss: 0.4688\n", "Epoch 16/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 11ms/step - accuracy: 0.7992 - loss: 0.4484 - val_accuracy: 0.7820 - val_loss: 0.4728\n", "Epoch 17/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m6s\u001b[0m 13ms/step - accuracy: 0.7982 - loss: 0.4452 - val_accuracy: 0.7812 - val_loss: 0.4698\n" ] } ], "source": [ "rmsprop = keras.optimizers.RMSprop(learning_rate=1e-4)\n", "model3.compile(optimizer=rmsprop, loss='binary_crossentropy',\n", " metrics=['accuracy'])\n", "\n", "checkpoint_cb = keras.callbacks.ModelCheckpoint('best-2rnn-model.keras',\n", " save_best_only=True)\n", "early_stopping_cb = keras.callbacks.EarlyStopping(patience=3,\n", " restore_best_weights=True)\n", "\n", "history = model3.fit(train_seq, train_target, epochs=100, batch_size=64,\n", " validation_data=(val_seq, val_target),\n", " callbacks=[checkpoint_cb, early_stopping_cb])" ] }, { "cell_type": "code", "execution_count": 12, "metadata": { "colab": { "base_uri": "https://localhost:8080/", "height": 449 }, "id": "1paBqczf6k2S", "outputId": "6b4eddf2-09f2-4a05-dece-e9a1e0f62732" }, "outputs": [ { "output_type": "display_data", "data": { "text/plain": [ "
" ], "image/png": "iVBORw0KGgoAAAANSUhEUgAAAkAAAAGwCAYAAABB4NqyAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjcuMSwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/bCgiHAAAACXBIWXMAAA9hAAAPYQGoP6dpAABiFElEQVR4nO3dd3hUVf7H8ffMJJn0hCSkQUhC7y0UA9hxwYJdUVEUV9110VVZG7s/cC0rrquujRV1dUFd2yIqimJBQXoVBaRDQihJSCCdtJn5/XGTgZhCyiSTZD6v55mHO3fOvfO94Cafvefcc0wOh8OBiIiIiAcxu7sAERERkZamACQiIiIeRwFIREREPI4CkIiIiHgcBSARERHxOApAIiIi4nEUgERERMTjeLm7gNbIbrdz+PBhgoKCMJlM7i5HRERE6sHhcJCfn09sbCxmc933eBSAanD48GHi4uLcXYaIiIg0QlpaGp07d66zjQJQDYKCggDjLzA4ONjN1YiIiEh95OXlERcX5/w9XhcFoBpUdnsFBwcrAImIiLQx9Rm+0ioGQc+ePZuEhAR8fX0ZOXIk69atq7XtOeecg8lkqva6+OKLnW0cDgczZ84kJiYGPz8/xo4dy+7du1viUkRERKQNcHsA+uCDD5g2bRqPPPIImzZtYtCgQYwbN47MzMwa2y9YsIAjR444X1u3bsVisXDNNdc42zz99NO8+OKLzJkzh7Vr1xIQEMC4ceMoLi5uqcsSERGRVszk7tXgR44cyfDhw3n55ZcB4wmsuLg47r77bh5++OHTHv/8888zc+ZMjhw5QkBAAA6Hg9jYWP70pz9x//33A5Cbm0tUVBRz587luuuuq3aOkpISSkpKnO8r+xBzc3PVBSYiItJG5OXlERISUq/f324dA1RaWsrGjRuZPn26c5/ZbGbs2LGsXr26Xud44403uO666wgICABg//79pKenM3bsWGebkJAQRo4cyerVq2sMQLNmzeLRRx9t4tWIiIicnt1up7S01N1ltEne3t5YLBaXnMutASgrKwubzUZUVFSV/VFRUezYseO0x69bt46tW7fyxhtvOPelp6c7z/Hrc1Z+9mvTp09n2rRpzveVd4BERERcqbS0lP3792O3291dSpsVGhpKdHR0k+fpa9NPgb3xxhsMGDCAESNGNOk8VqsVq9XqoqpERESqczgcHDlyBIvFQlxc3Gkn6pOqHA4HRUVFzjHCMTExTTqfWwNQREQEFouFjIyMKvszMjKIjo6u89jCwkLef/99HnvssSr7K4/LyMio8peTkZHB4MGDXVO4iIhIA5WXl1NUVERsbCz+/v7uLqdN8vPzAyAzM5PIyMgmdYe5NX76+PiQlJTEkiVLnPvsdjtLliwhOTm5zmP/97//UVJSwo033lhlf2JiItHR0VXOmZeXx9q1a097ThERkeZis9kA43efNF5leCwrK2vSedzeBTZt2jRuvvlmhg0bxogRI3j++ecpLCxkypQpAEyePJlOnToxa9asKse98cYbXH755YSHh1fZbzKZuPfee3niiSfo0aMHiYmJzJgxg9jYWC6//PKWuiwREZEaaY3JpnHV35/bA9DEiRM5evQoM2fOJD09ncGDB7N48WLnIOYDBw5U6yfduXMnK1as4Ouvv67xnA8++CCFhYXccccd5OTkMGbMGBYvXoyvr2+zX4+IiIi0fm6fB6g1asg8AiIiIvVRXFzM/v37SUxM1P8hb4K6/h4b8vtbQ9BFRESkRSQkJPD888+7uwygFXSBeZKcolIOHj9Bv9hg9QGLiEibcM455zB48GCXBJf169c7Jy52NwWgFrR4azoPL9hCdLAv5/WJZGyfSEZ1i8DX2zWzWoqIiLQ0h8OBzWbDy+v0kaJjx44tUFH9qAusBWUXlhLqbSM9r5h31x7g1rkbGPzY19w2bz3vrj1Aeq4WaxUR8RQOh4Oi0nK3vOo7/PeWW25h2bJlvPDCC5hMJkwmE3PnzsVkMvHll1+SlJSE1WplxYoV7N27l8suu4yoqCgCAwMZPnw43377bZXz/boLzGQy8e9//5srrrgCf39/evTowcKFC13511wr3QFqQVN7F/GH9fexfeBDvHcimSU7MjmcW8y32zP5drsxs2X/TsGc1zuKsX0i6R8bgtmsrjIRkfboRJmNvjO/cst3//LYOPx9Th8BXnjhBXbt2kX//v2dEw9v27YNgIcffphnnnmGrl270qFDB9LS0rjooov429/+htVq5a233mLChAns3LmTLl261Podjz76KE8//TT/+Mc/eOmll5g0aRKpqamEhYW55mJroQDUkta8gqkoi75rHuDx3pfw2F3PsaPAj+92ZPLt9gw2p+Ww9VAeWw/l8eKS3XQMsnJer0jO7xPJmB4R9fqPVURExFVCQkLw8fHB39/fudJC5Vqdjz32GBdccIGzbVhYGIMGDXK+f/zxx/n4449ZuHAhd911V63fccstt3D99dcD8OSTT/Liiy+ybt06xo8f3xyX5KTfqC3p0pcgvBssfQp2fI7pwGr6XPwcfc69nKnndieroISlO4+yZHsGP+w6ytH8Ej7YkMYHG9Lw8TIzqls45/eO5Lw+UXQK9XP31YiISBP4eVv45bFxbvvupho2bFiV9wUFBfz1r39l0aJFHDlyhPLyck6cOMGBAwfqPM/AgQOd2wEBAQQHBzvX+2pOCkAtyeIFZ90PPcfBx7+HjK3wv5th+1Vw0TNEBIZxdVJnrk7qTEm5jXX7j7FkeyZLdmSQduwES3ceZenOo8z4dBu9o4M4v08k5/eJYlDnUCzqKhMRaVNMJlObvrP/66e57r//fr755hueeeYZunfvjp+fH1dffTWlpaV1nsfb27vKe5PJhN1ud3m9v9Z2/+bbsugBcPv38MPTsPw52PoRpKyACS9CL+OWn9XLwpk9OnJmj448MqEvezILWLIjkyXbM9iYepwd6fnsSM9n9vd7CQ/w4dzekZzf2+gqC/L1Pk0BIiIi9ePj4+Ncx6wuK1eu5JZbbuGKK64AjDtCKSkpzVxd4ykAuYuXD5z3f9DrQuNuUNYueG8iDL4Rxj8JviHOpiaTiR5RQfSICuL3Z3fjeGEpS3dlsmR7Jst2HSW7sJT5Gw8yf+NBvC0mzugaznm9IxnbJ4q4MK04LCIijZeQkMDatWtJSUkhMDCw1rszPXr0YMGCBUyYMAGTycSMGTNa5E5OY+kxeHfrlAS/+wGS7wJMsPkd+Nco2Pt9rYd0CPDhiiGdefmGoWyacQHv3j6S28YkkhgRQJnNwfLdWTz62S+c+fT3XPDcMr7f2fx9qSIi0j7df//9WCwW+vbtS8eOHWsd0/Pcc8/RoUMHRo0axYQJExg3bhxDhw5t4WrrT2uB1cBta4GlroZP7oTj+433w34LFzwG1sB6n2Lf0QLnU2XrU45jszsI9fdm9cPn4+ejCRdFRNxFa4G5htYCa4/ik+HOlTD8duP9hjdgzmhIXVXvU3TtGMhtZ3bl/TuS2fR/FxAX5kdOURmfbj7UTEWLiIi0PQpArY1PAFz8DEz+FII7w/EU+M9F8NVfoOxEg04V4u/NzckJAMxdlVLvmT9FRETaOwWg1qrrOfCHVTDkRsABq1+GV8+CgxsbdJprhsXh72NhR3o+q/dlN0upIiIibY0CUGvmGwKXzYYbPoTAKONJsTfGwpLHoLykXqcI8fPmqqGdAZi7MqUZixUREWk7FIDagp7j4A9rYMA14LDD8mfh9fMgfUu9Dr95VDwA32zPIO1YUXNWKiIi0iYoALUV/mFw1b/h2rfAP9yYRfq1c2HZP8BWXueh3SODOLNHBA4HvLU6pWXqFRERacUUgNqavpfBH9ZC70vAXgbfP2F0i2XuqPOwW0cnAvD++jQKS+oOTCIiIu2dAlBbFNgRJr4DV75ujBM6/KMxQHrli2Cvebrys3t2JCHcn/zichb8qEfiRUTEsykAtVUmEwy81hgb1P0CsJXANzOMR+az91ZrbjabuHlUAgBzV+7XI/EiItIiEhISeP75591dRjUKQG1dcCxM+p+xkKpPIKStgTljYN3r8Ks1WK5O6kyg1Yu9RwtZvjvLTQWLiIi4nwJQe2AyQdLNcOcqSDgTyorgi/vh7csg5+SaLUG+3lydVPFI/KoUNxUrIiLifgpA7UmHeJi8EC78B3j5wf4fjIVVd33lbHLLqARMJvhuRyb7swrdWKyIiLR2r732GrGxsdVWdb/sssu49dZb2bt3L5dddhlRUVEEBgYyfPhwvv32WzdV2zAKQO2N2Qwj7zDWFOs8Akrz4cuHnB8nRARwbq9IAObpLpCIiPs4HFBa6J5XPceBXnPNNWRnZ/P999879x07dozFixczadIkCgoKuOiii1iyZAk//vgj48ePZ8KECbWuGN+aeLm7AGkm4d3gpgXwVLyxunzOAQjtAhh3gb7bkcn8jQf50296EuTr7eZiRUQ8UFkRPBnrnu/+82Fj7cnT6NChAxdeeCHvvvsu559/PgDz588nIiKCc889F7PZzKBBg5ztH3/8cT7++GMWLlzIXXfd1Wzlu4LuALVn1iDolGRs7//BufvMHhF0jwykoKSc+RsPuqk4ERFpCyZNmsRHH31ESYmxBNN///tfrrvuOsxmMwUFBdx///306dOH0NBQAgMD2b59u+4ASSuQeBYcXGcEoCE3AmAyGY/Ez/hkK/NWpXBzcgJms8nNhYqIeBhvf+NOjLu+u54mTJiAw+Fg0aJFDB8+nOXLl/PPf/4TgPvvv59vvvmGZ555hu7du+Pn58fVV19NaWlpc1XuMgpA7V3iWbD8GSMAORzGE2PAVUM78fTiHaRkF7Fs11HO7R3p5kJFRDyMyVSvbih38/X15corr+S///0ve/bsoVevXgwdOhSAlStXcsstt3DFFVcAUFBQQEpKihurrT91gbV3cSPBYoX8I5C127nb38eL64bHAfDmyv3uqk5ERNqASZMmsWjRIt58800mTZrk3N+jRw8WLFjA5s2b+emnn7jhhhuqPTHWWikAtXfevtBlpLG9f1mVjyYnJ2A2wfLdWezJzHdDcSIi0hacd955hIWFsXPnTm644Qbn/ueee44OHTowatQoJkyYwLhx45x3h1o7dYF5gsSzjC6w/T/AiNudu+PC/BnbJ4qvf8lg7qoUnrh8gBuLFBGR1spsNnP4cPXxSgkJCXz33XdV9k2dOrXK+9baJaY7QJ4g8Wzjz5Tl1ZbHuGV0AgAfbTxE7omyFi5MRETEPRSAPEHsUPAJghPHIWNLlY+Su4bTKyqIE2U2/rchzU0FioiItCwFIE9g8YL4Ucb2KfMBgfFI/JSKu0BzV6Vgs2uVeBERaf8UgDxF4lnGn78KQACXDe5EqL83B4+fYMn2jBYuTEREpOUpAHmKygCUugpsVcf6+PlYuH6EsUzGf1amtHBhIiKexVHPdbikZq76+1MA8hRR/cEvDEoL4NCmah/feEY8FrOJ1fuy2ZGe54YCRUTaN4vFAtAmZkluzYqKigDw9m7aOpZ6DN5TmM2QeCb88qnRDVY5N1CFTqF+jOsXxRdb0pm3KoVZVw50U6EiIu2Tl5cX/v7+HD16FG9vb8xm3YNoCIfDQVFREZmZmYSGhjoDZWMpAHmSxLMqAtAyOPuBah9PGZ3IF1vSWbDpEA+O602HAB83FCki0j6ZTCZiYmLYv38/qamp7i6nzQoNDSU6OrrJ51EA8iSV8wGlrYOyE+DtV+XjYfEd6BcbzLbDeby/Po07z+nmhiJFRNovHx8fevTooW6wRvL29m7ynZ9KCkCeJLw7BMUY64KlrYOuZ1f52HgkPpH7//cTb69O4fYzE/Gy6BatiIgrmc1mfH193V2Gx9NvN09iMp28C/SrdcEqXTIwhvAAHw7nFvP1L3okXkRE2icFIE9Tx3xAAL7eFm4YaTwSP1ePxIuISDulAORpEs80/jy0CYprftz9xjPi8TKbWJdyjK2HcluwOBERkZahAORpQrtAh0Rw2ODA6hqbRAX7ctGAGMBYHkNERKS9UQDyRJWDn/fVPA4IcK4PtnDzYbIKSlqgKBERkZajAOSJTjMOCGBIlw4Migul1GbnvbUHWqgwERGRlqEA5IkSKsYBZWyBwuxam00ZlQDA22tSKbPZW6AwERGRlqEA5IkCIyGyr7GdsrzWZhcNiKFjkJXM/BK+2HKkhYoTERFpfgpAnsrZDVb7OCAfLzM3jowHNBhaRETaFwUgT+WcELH2cUAAN4zsgo/FzI8HcticltP8dYmIiLQABSBPFT8KTGbI3gO5h2pt1jHIyiWDKh6JX7m/paoTERFpVgpAnsovFGIGG9t1jAMCmDIqEYBFW46QmVfcvHWJiIi0AAUgT1aPx+EBBnQOYVh8B8psDt7RI/EiItIOKAB5slMnRHQ46mx6S8XEiO+uTaWk3NbMhYmIiDQvBSBPFncGmL0h7yAc21dn03H9ookO9iWroJTPf9Ij8SIi0ra5PQDNnj2bhIQEfH19GTlyJOvWrauzfU5ODlOnTiUmJgar1UrPnj354osvnJ//9a9/xWQyVXn17t27uS+jbfLxh7gRxvZpusG8LWZuSj75SLzjNHeMREREWjO3BqAPPviAadOm8cgjj7Bp0yYGDRrEuHHjyMzMrLF9aWkpF1xwASkpKcyfP5+dO3fy+uuv06lTpyrt+vXrx5EjR5yvFStWtMTltE31HAcEcP2ILli9zGw5lMumA8ebuTAREZHm49YA9Nxzz3H77bczZcoU+vbty5w5c/D39+fNN9+ssf2bb77JsWPH+OSTTxg9ejQJCQmcffbZDBo0qEo7Ly8voqOjna+IiIiWuJy26dT5gOx1L3cRFuDD5YONsPnmypRmLkxERKT5uC0AlZaWsnHjRsaOHXuyGLOZsWPHsnr16hqPWbhwIcnJyUydOpWoqCj69+/Pk08+ic1WdVDu7t27iY2NpWvXrkyaNIkDB+p+cqmkpIS8vLwqL4/RKQm8/aEoC45uP23zmyvWB1u8NZ0juSeauTgREZHm4bYAlJWVhc1mIyoqqsr+qKgo0tPTazxm3759zJ8/H5vNxhdffMGMGTN49tlneeKJJ5xtRo4cydy5c1m8eDGvvPIK+/fv58wzzyQ/P7/WWmbNmkVISIjzFRcX55qLbAu8fKBLsrFdj26wvrHBjEwMw2Z38Pbq1GYuTkREpHm4fRB0Q9jtdiIjI3nttddISkpi4sSJ/OUvf2HOnDnONhdeeCHXXHMNAwcOZNy4cXzxxRfk5OTw4Ycf1nre6dOnk5ub63ylpaW1xOW0Hg0YBwQwZbQxMeJ76w5QXKZH4kVEpO3xctcXR0REYLFYyMjIqLI/IyOD6OjoGo+JiYnB29sbi8Xi3NenTx/S09MpLS3Fx8en2jGhoaH07NmTPXv21FqL1WrFarU28kragcoAlLICbOVgqfs/i7F9IukU6sehnBMs3HyYa4d70B0zERFpF9x2B8jHx4ekpCSWLFni3Ge321myZAnJyck1HjN69Gj27NmD/ZTBurt27SImJqbG8ANQUFDA3r17iYmJce0FtCcxg8A3BEry4MhPp23uZTEzueKR+DdX7tcj8SIi0ua4tQts2rRpvP7668ybN4/t27dz5513UlhYyJQpUwCYPHky06dPd7a/8847OXbsGPfccw+7du1i0aJFPPnkk0ydOtXZ5v7772fZsmWkpKSwatUqrrjiCiwWC9dff32LX1+bYbZAwpnG9v5l9TrkuuFd8PO2sCM9n7X7jzVjcSIiIq7nti4wgIkTJ3L06FFmzpxJeno6gwcPZvHixc6B0QcOHMBsPpnR4uLi+Oqrr7jvvvsYOHAgnTp14p577uGhhx5ytjl48CDXX3892dnZdOzYkTFjxrBmzRo6duzY4tfXpiSeBTs+N8YBnTnttM1D/L25Ymgn3l17gP+s3M8ZXcNboEgRERHXMDnUf1FNXl4eISEh5ObmEhwc7O5yWkbmdvjXGeDlBw+ngtfpx0Ttzsjngn/+gNkEyx44l7gw/xYoVEREpGYN+f3dpp4Ck2bUsTcEREL5CTi4vl6H9IgKYkz3COwOeGeNHokXEZG2QwFIDCZTgx+HB7ilYmLE99YdoKi0vBkKExERcT0FIDmpEQHovN6RxIf7k1dczsc/HmqmwkRERFxLAUhOqgxAB9dDaWG9DjGbTUxOTgBg7kqtEi8iIm2DApCc1CEBQrqAvRxSa16PrSbXDOtMgI+F3ZkFrNyT3Xz1iYiIuIgCkJxkMkHXym6w+s0HBBDs683VSZ0B+M/K/c1RmYiIiEspAElViWcbfzZgHBDA5IrB0N/tzCQlq37dZyIiIu6iACRVVc4IfeQnOHG83od16xjIOb064nDAW1olXkREWjkFIKkqOAYiegIOY3HUBqh8JP5/G9Mos9nrbiwiIuJGCkBSXSO7wc7q0ZEO/t7kF5fzU1qO6+sSERFxEQUgqa4R8wGB8Uj8qG4RAKzYk+XqqkRERFxGAUiqSxgDmODoDsjPaNCho7sbAWiVHocXEZFWTAFIqvMPg+gBxnbK8gYdOqYiAG06cJzCEi2NISIirZMCkNSsa8U4oH1LG3RYl3B/4sL8KLc7WLf/mOvrEhERcQEFIKlZIwdCw8m7QBoHJCIirZUCkNSsyxlg9oKcVDie0qBDKwdCr1QAEhGRVkoBSGpmDYJOScb2/oaNAxrVLRyAHen5HM0vcXVlIiIiTaYAJLVLbPi6YADhgVb6xgQDsGqv7gKJiEjrowAktTt1HJDD0aBDx/RQN5iIiLReCkBSu87DwcsXCjIga1eDDq2cD2jF7iwcDQxPIiIizU0BSGrn7QtxI43tBj4NNjyhAz4WM4dzi0nJLmqG4kRERBpPAUjqVjkOqIHzAfn7eDGkSyigx+FFRKT1UQCSunU9x/gzZQXYbQ06dIxzWQwFIBERaV0UgKRuMYPBJwiKcyB9S4MOHV0xEHrV3mxsdo0DEhGR1kMBSOpm8YKE0cZ2A8cBDewUQpDVi9wTZWw7nNsMxYmIiDSOApCcnnM+oIYFIC+LmTMqJkXUOCAREWlNFIDk9CrnA0pdBeWlDTq0chyQ5gMSEZHWRAFITi+yL/iHQ1khHN7UoENHdzfuAK1POU5xWcMGUYuIiDQXBSA5PbMZEs40thvYDdatYyBRwVZKy+1sTD3eDMWJiIg0nAKQ1E8jxwGZTKaTs0KrG0xERFoJBSCpn8pxQGlrobRhMztrHJCIiLQ2CkBSP+HdILgT2EqNENQAlXeAthzKJaeoYYOoRUREmoMCkNSPydTobrCoYF96RAbicMDqvdnNUJyIiEjDKABJ/TUyAMHJu0Ar96obTERE3E8BSOqvMgAd3gTFDZvZ2RmA9ugOkIiIuJ8CkNRfSGcI6wYOuzEpYgOM7BqGxWxif1YhB483bBC1iIiIqykAScM0shss2NebQZ1DAFilu0AiIuJmCkDSME0YBzRG8wGJiEgroQAkDVMZgDK2QsHRBh1aOQ5o1d4sHA6HqysTERGpNwUgaZiACIjqb2ynLG/QoUO6dMDP20JWQSk7M/KboTgREZH6UQCShmtkN5iPl5kRiWEArNitbjAREXEfBSBpOBeMA9KyGCIi4k4KQNJw8aPAZIZjeyH3YIMOrRwHtHb/MUrL7c1RnYiIyGkpAEnD+YZA7BBju4F3gXpHBxEe4ENRqY3NaTmur01ERKQeFICkcSpXh29gADKbTSR3CwfUDSYiIu6jACSNc+o4oAY+0q5xQCIi4m4KQNI4cSPB4gN5h+DYvgYdWjkO6Me0HPKLy5qjOhERkTopAEnj+PhD5xHG9r6lDTo0Lsyf+HB/bHYH6/Yfc31tIiIip6EAJI3XtXHjgODkXSAtiyEiIu6gACSNVzkOKGU52Bv2SLvGAYmIiDspAEnjxQ4F7wAoyobMXxp0aHLXcEwm2JVRQGZ+cTMVKCIiUjMFIGk8Lx+ITza29y9r0KEdAnzoFxsMwKo92a6uTEREpE4KQNI0TVgWQ+OARETEXRSApGkqJ0RMWQm28gYdeuo4IEcD5xISERFpCgUgaZroAeAbCqX5cGRzgw4dnhCGj5eZI7nF7MsqbJbyREREaqIAJE1jtkDCGGO7geOAfL0tDIvvAMAqdYOJiEgLUgCSpqvsBtvXsAAEGgckIiLuoQAkTVc5IWLaWihr2CPtlQFo1d5sbHaNAxIRkZbh9gA0e/ZsEhIS8PX1ZeTIkaxbt67O9jk5OUydOpWYmBisVis9e/bkiy++aNI5pYkiekJgFJQXw8H1DTp0QKcQgny9yC8uZ8uh3GYqUEREpCq3BqAPPviAadOm8cgjj7Bp0yYGDRrEuHHjyMzMrLF9aWkpF1xwASkpKcyfP5+dO3fy+uuv06lTp0afU1zAZGr04/AWs4lR3cIBzQotIiItx60B6LnnnuP2229nypQp9O3blzlz5uDv78+bb75ZY/s333yTY8eO8cknnzB69GgSEhI4++yzGTRoUKPPCVBSUkJeXl6VlzSQMwA1fBxQ5ePwK3YrAImISMtwWwAqLS1l48aNjB079mQxZjNjx45l9erVNR6zcOFCkpOTmTp1KlFRUfTv358nn3wSm83W6HMCzJo1i5CQEOcrLi7ORVfpQSoHQh/aCCX5DTq0chzQxtTjnCi1uboyERGRatwWgLKysrDZbERFRVXZHxUVRXp6eo3H7Nu3j/nz52Oz2fjiiy+YMWMGzz77LE888USjzwkwffp0cnNzna+0tLQmXp0H6hAPofFgL4cDaxp0aGJEADEhvpTa7GxIPdZMBYqIiJzk9kHQDWG324mMjOS1114jKSmJiRMn8pe//IU5c+Y06bxWq5Xg4OAqL2mERnaDmUwmPQ4vIiItym0BKCIiAovFQkZGRpX9GRkZREdH13hMTEwMPXv2xGKxOPf16dOH9PR0SktLG3VOcaEmzAd06rIYIiIizc1tAcjHx4ekpCSWLFni3Ge321myZAnJyck1HjN69Gj27NmD3W537tu1axcxMTH4+Pg06pziQolnASZI/xmy9zbo0FHdjSfBth3O43hhaTMUJyIicpJbu8CmTZvG66+/zrx589i+fTt33nknhYWFTJkyBYDJkyczffp0Z/s777yTY8eOcc8997Br1y4WLVrEk08+ydSpU+t9TmlGQVHQvWIA+sa5DTo0MsiXXlFBOBywel+262sTERE5hZc7v3zixIkcPXqUmTNnkp6ezuDBg1m8eLFzEPOBAwcwm09mtLi4OL766ivuu+8+Bg4cSKdOnbjnnnt46KGH6n1OaWbDpsCeb2Dzf+G8/wMva70PHd09gp0Z+azYk8VFA2KasUgREfF0JofDofUHfiUvL4+QkBByc3M1ILqhbOXw/ADIPwxXvQEDrq73oUu2Z/DbeRuID/dn2QPnNmORIiLSHjXk93ebegpM2gCLFwy9ydje8J8GHTqyazgWs4nU7CLSjhU1Q3EiIiIGBSBxvaGTwWSG1BVwdFe9Dwu0ejEkLhTQ02AiItK8FIDE9UI6Q4/fGNsNHAyt+YBERKQlKABJ8xh2q/HnT+9CWXG9DxvTwwhAq/dmY7dreJqIiDQPBSBpHt3HQkgcnDgOv3xa78MGx4US4GMhu7CUHekNW1NMRESkvhSApHmYLcZYIICN9R8M7W0xMyIxDNA4IBERaT4KQNJ8htwEJgscWA2Z2+t9mMYBiYhIc1MAkuYTHAO9LjS2G/BIfOU4oHX7j1Fabj9NaxERkYZTAJLmlVSxBMlP70Np/eb26RUVRESgDyfKbPx44HgzFiciIp5KAUiaV7fzILQLlOTCto/rdYjJZHJ2g2kckIiINAcFIGleZjMk3WJsN2Aw9OhuGgckIiLNRwFImt/gG8HsBQfXQ/rWeh0yumIc0E8Hc8krLmvO6kRExAMpAEnzC4qC3hcb2/W8C9Qp1I/EiABsdgdr9x1rxuJERMQTKQBJy6gcDP3zh1BaWK9DRncPBzQOSEREXE8BSFpG4tnQIRFK8mDrR/U6ZIwGQouISDNRAJKWcepg6A1v1uuQ5K4RmEywO7OAjLz6rycmIiJyOgpA0nIGTwKzNxz+EQ5vPm3zEH9vBnQKAXQXSEREXEsBSFpOYEfoe6mxXc/B0FoWQ0REmoMCkLSsysHQW+ZDyelXez91HJDD4WjOykRExIMoAEnLShgD4T2gtAC2/O+0zZPiO2D1MpORV8Leo/V7ekxEROR0FICkZZlMpwyG/g+c5q6Or7eF4QlhgMYBiYiI6ygAScsbfANYrJD+MxzedNrmGgckIiKupgAkLc8/DPpeZmzX45H4ygkR1+zNptxmb87KRETEQygAiXsMqxgMvXUBFOfW2bRfbAghft7kl5Tz86G624qIiNSHApC4R5dk6NgbyoqM5THqYDGbGNWtYlmM3eoGExGRpmtUAJo3bx6LFi1yvn/wwQcJDQ1l1KhRpKamuqw4acdMppOPxNdjMHTlOKCVexWARESk6RoVgJ588kn8/PwAWL16NbNnz+bpp58mIiKC++67z6UFSjs2aCJ4+ULmNji4vs6mlfMBbUrNoai0vCWqExGRdqxRASgtLY3u3bsD8Mknn3DVVVdxxx13MGvWLJYvX+7SAqUd8+sA/a40tjfUPTN0fLg/nUL9KLXZWZ9yvAWKExGR9qxRASgwMJDs7GwAvv76ay644AIAfH19OXHihOuqk/avcjD0tgVwovZgYzKZnE+DaT4gERFpqkYFoAsuuIDbbruN2267jV27dnHRRRcBsG3bNhISElxZn7R3nYdDZD8oL4afPqizqXM+IA2EFhGRJmpUAJo9ezbJyckcPXqUjz76iPBw4/+Zb9y4keuvv96lBUo7ZzKdvAu04c06B0OP6mYEoF+O5HGssLQlqhMRkXbK5NAKk9Xk5eUREhJCbm4uwcHB7i6n/SvOhWcrHomf8iXEj6q16fjnf2BHej4v3zCESwbGtmCRIiLS2jXk93ej7gAtXryYFStWON/Pnj2bwYMHc8MNN3D8uAaoSgP5hkD/q4zt0wyGPnV1eBERkcZqVAB64IEHyMvLA2DLli386U9/4qKLLmL//v1MmzbNpQWKh6jsBvvlUyg6Vmuz0T20LpiIiDRdowLQ/v376du3LwAfffQRl1xyCU8++SSzZ8/myy+/dGmB4iFih0L0QLCVwOZ3a202IiEML7OJtGMnOJBd1IIFiohIe9KoAOTj40NRkfHL59tvv+U3v/kNAGFhYc47QyINcupg6I1zax0MHWD1YmiXDoDuAomISOM1KgCNGTOGadOm8fjjj7Nu3TouvvhiAHbt2kXnzp1dWqB4kAHXgE8gZO+GlBW1NtOyGCIi0lSNCkAvv/wyXl5ezJ8/n1deeYVOnToB8OWXXzJ+/HiXFigexBpkhCCAjbUPhh7Tw5h2YdWeLOx2PcQoIiINp8fga6DH4N3oyE/w6llg9oZp2yGwY7UmZTY7Qx77hoKScj6/ewz9O4W4oVAREWltGvL726uxX2Kz2fjkk0/Yvn07AP369ePSSy/FYrE09pQiEDPIGBB9eBNs/i+MubdaE2+LmTO6hvHt9kxW7slSABIRkQZrVBfYnj176NOnD5MnT2bBggUsWLCAG2+8kX79+rF3715X1yie5tTB0HZ7jU0qZ4XWQGgREWmMRgWgP/7xj3Tr1o20tDQ2bdrEpk2bOHDgAImJifzxj390dY3iafpfBdZgOL4f9i+rscmYivmA1qcco6Tc1pLViYhIO9CoALRs2TKefvppwsLCnPvCw8N56qmnWLas5l9YIvXmEwADrzW2axkM3SMykI5BVorL7GxKzWm52kREpF1oVACyWq3k5+dX219QUICPj0+TixIhqaIbbMciyM+o9rHJZNKyGCIi0miNCkCXXHIJd9xxB2vXrsXhcOBwOFizZg2///3vufTSS11do3ii6P7QeQTYy2HzOzU2qZwPSOOARESkoRoVgF588UW6detGcnIyvr6++Pr6MmrUKLp3787zzz/v4hLFY51mMPTo7sZ8QD8fzCG7oKQFCxMRkbauUQEoNDSUTz/9lF27djF//nzmz5/Prl27+PjjjwkNDXVxieKx+l1hrBSfcwD2flft45gQPwZ2DsHugGe+3umGAkVEpK2q9zxAp1vl/fvvv3duP/fcc42vSKSStx8Muh7WzjEGQ/cYW63JjEv6cs2c1by/Po1rh8UxpGKdMBERkbrUOwD9+OOP9WpnMpkaXYxINUlTjAC080vIOwLBMVU+Hp4QxlVDO/PRpoPM+HQrn04dg8Ws/wZFRKRu9Q5Ap97hEWkxkb2hSzIcWA0/vg1nP1itycMX9ubrX9LZeiiPd9cd4KYz4t1QqIiItCWNGgMk0qIqH4nfOA/s1Sc97Bhk5YFxvQD4x+IdZGlAtIiInIYCkLR+fS8Dvw6QdxD2fFtjk0kj4+kXG0xecTl//3JHCxcoIiJtjQKQtH7evjB4krG9oeaZoS1mE49f3h+A/208yIaUYy1VnYiItEEKQNI2JN1i/Ln7K8g9WGOToV06MHFYHAAzPt1Gua3mhVRFREQUgKRtiOgBCWeCww6b3qq12YPjexHi5832I3m8vSa1BQsUEZG2RAFI2o7Ku0Cb3gJbeY1NwgOtPDjeGBD93Ne7yMwvbqHiRESkLVEAkrajzwTwD4f8I0ZXWC2uG96FgZ1DyC8p56kvNCBaRESqaxUBaPbs2SQkJODr68vIkSNZt25drW3nzp2LyWSq8vL19a3S5pZbbqnWZvz48c19GdLcvKynHQwNFQOiL+uPyQQLfjzE2n3ZLVSgiIi0FW4PQB988AHTpk3jkUceYdOmTQwaNIhx48aRmZlZ6zHBwcEcOXLE+UpNrT7WY/z48VXavPfee815GdJSKrvB9nwLx2sf4zMoLpTrR3QBYOan2yjTgGgRETmF2wPQc889x+23386UKVPo27cvc+bMwd/fnzfffLPWY0wmE9HR0c5XVFRUtTZWq7VKmw4dtEZUuxDeDbqeAzjqHAwN8OC4XnTw92ZnRj7zVqW0RHUiItJGuDUAlZaWsnHjRsaOPbnIpdlsZuzYsaxevbrW4woKCoiPjycuLo7LLruMbdu2VWuzdOlSIiMj6dWrF3feeSfZ2bV3g5SUlJCXl1flJa1Y5czQP74NtrJam4X6+/Dwhb0B+Oc3u8jI04BoERExuDUAZWVlYbPZqt3BiYqKIj09vcZjevXqxZtvvsmnn37KO++8g91uZ9SoURw8eHJumPHjx/PWW2+xZMkS/v73v7Ns2TIuvPBCbLbqyygAzJo1i5CQEOcrLi7OdRcprtf7YgiIhIIM2PlFnU2vSYpjSJdQCkttPLFoewsVKCIirZ3bu8AaKjk5mcmTJzN48GDOPvtsFixYQMeOHXn11Vedba677jouvfRSBgwYwOWXX87nn3/O+vXrWbp0aY3nnD59Orm5uc5XWlpaC12NNIrFG4bcaGzXMRgawFwxINpsgs9+OsyqPVktUKCIiLR2bg1AERERWCwWMjIyquzPyMggOjq6Xufw9vZmyJAh7Nmzp9Y2Xbt2JSIiotY2VquV4ODgKi9p5ZJuBkyw73s4tq/Opv07hThXiJ/x6VZKyzUgWkTE07k1APn4+JCUlMSSJUuc++x2O0uWLCE5Oble57DZbGzZsoWYmJha2xw8eJDs7Ow620gb0yEBup1nbG+cd9rm037Ti4hAH/YeLeTNlfubtzYREWn13N4FNm3aNF5//XXmzZvH9u3bufPOOyksLGTKFGOg6+TJk5k+fbqz/WOPPcbXX3/Nvn372LRpEzfeeCOpqancdtttgDFA+oEHHmDNmjWkpKSwZMkSLrvsMrp37864cePcco3STIbdavz54ztQWlRn0xA/b6Zf2AeAF77dzeGcE81dnYiItGJuD0ATJ07kmWeeYebMmQwePJjNmzezePFi58DoAwcOcOTIEWf748ePc/vtt9OnTx8uuugi8vLyWLVqFX379gXAYrHw888/c+mll9KzZ09++9vfkpSUxPLly7FarW65RmkmPcdDcGcoyoLP/ggOR53NrxzaieEJHThRZuOJRb+0UJEiItIamRyO0/zW8EB5eXmEhISQm5ur8UCtXcoKmHcpOGzwm7/BqLvqbL79SB6XvLQCm93BW7eO4KyeHVuoUBERaW4N+f3t9jtAIk2SMAbGzzK2v5kBe7+vs3mfmGBuTk4A4JGF2ygpr3lqBBERad8UgKTtG3GHsUaYww7zp8DxlDqb33tBDzoGWdmfVci/l2tAtIiIJ1IAkrbPZIKLn4PYoXDiOLw/CUoLa20e7OvN/11sDIh+6bvdpB2rewC1iIi0PwpA0j54+8LEdyCgI2RshU/vqnNQ9KWDYjmjaxjFZXYe/1wDokVEPI0CkLQfIZ3g2rfA7AXbFsDKF2ptajKZeOyy/niZTXz9Swbf78hswUJFRMTdFICkfYkfBeOfMraXPAp7vq21ac+oIG4dkwgYA6KLyzQgWkTEUygASfsz/DYYclPFoOhb61wq44/n9yA62JcDx4qYs2xvCxYpIiLupAAk7Y/JBBc/C52GQXGuMSi6pKDGpoFWL/7vEmNA9L+W7uVAtgZEi4h4AgUgaZ+8rDDxbQiMgsxf4NM/1Doo+uIBMYzpHkFpuZ2/frYNzQ0qItL+KQBJ+xUcC9e+DWZv+OVTWPFcjc1MJhOPXtYPb4uJ73Zk8u12DYgWEWnvFICkfesyEi76h7G95HHY/U2Nzbp1DOT2M7sC8NeF2zhRqgHRIiLtmQKQtH/DpkDSLYAD5v8Wsmse7HzXed2JDfHlUM4J/rV0T4uWKCIiLUsBSDzDhU9D3EgoyYX3b4CS/GpN/H28mDmhLwCvLtvH/qzaZ5MWEZG2TQFIPIOX1ZgkMSgGju6Aj38Pdnu1ZuP6RXN2z46U2uzM/HSrBkSLiLRTCkDiOYKijUHRFh/Y8Tksf7ZaE5PJxF8v7YePxczy3Vl8tS3dDYWKiEhzUwASzxI3HC56xtj+/m+wc3G1JokRAfz+bGNA9GOf/UJRaXlLVigiIi1AAUg8T9LNMOy3gAMW3A5Zu6s1ufOc7nTu4Mfh3GJe+k4DokVE2hsFIPFM45+CLslQkmcMii7Oq/Kxn4+Fv07oB8C/l+9jT2bNM0mLiEjbpAAknsnLp2JQdCxk7YKPf1dtUPTYvlGc3zuSMpuDRxZqQLSISHuiACSeKzASrnsHLFbY+QX88HS1Jn+9tB9WLzMr92Tz+c9H3FCkiIg0BwUg8WydkuCSfxrbS2fBjkVVPo4L8+cP53QH4IlFv1BQogHRIiLtgQKQyJBJMOJ3xvaC38HRnVU+/t3ZXYkP9ycjr4QXvt3lhgJFRMTVFIBEAMb9DeLHQGl+xaDoXOdHvt4W/nqpMSD6zZUp7EyvPou0iIi0LQpAIgAWb7hmLgR3huw98NHtVQZFn9srknH9orDZHczQDNEiIm2eApBIpcCOxqBoL1/Y/RUsfbLKxzMu6Yuvt5l1+4/x+vJ9bipSRERcQQFI5FSxQ2DCC8b2D/+AXxY6P+rcwZ/pF/YB4MkvdvDFFj0VJiLSVikAifzaoOvgjD8Y2x//HjK3Oz+anBzPzcnxANz7wWY2ph5zR4UiItJECkAiNbngcUg4E8oKjUHRJ44DxmKpMyf0Y2yfSErL7dz+1kZSsgrdXKyIiDSUApBITSxexqDokC5wbB98dBvYbcZHZhMvXj+EgZ1DOFZYyi3/WcexwlL31isiIg2iACRSm4CIikHRfrDnW/juCedH/j5e/PvmYXQK9SMlu4g73tpAcZnNjcWKiEhDKACJ1CVmEFz6krG94jnY9rHzo8ggX+ZOGU6QrxcbUo/zp//9hN2ux+NFRNoCBSCR0xl4DSTfZWx/8gfI2Ob8qEdUEK/elIS3xcSin4/w9692uKlIERFpCAUgkfoY+yh0PQfKiuC962Hvd86JEkd1i+DvVw0E4NVl+3hnTaobCxURkfpQABKpD4sXXP0fCI2HnFR4+wp4aQgsfw7yM7hyaGemXdATgJmfbuX7HZluLlhEROqiACRSX/5hMOULGH47WIPheAoseRT+2Rc+uIm749O4ZmgsdgdMfXcTWw/lnvaUIiLiHiaHFjWqJi8vj5CQEHJzcwkODnZ3OdIalRbCtk9g41w4uM652xEaz4f283gmczimoCg+njqaTqF+bitTRMSTNOT3twJQDRSApEEythlB6KcPoMS461OOha9tSSwPupjpd99JsJ/VvTWKiHgABaAmUgCSRiktgl8+McJQ2lrn7kxLNGFn3oZX0mQIinJbeSIi7Z0CUBMpAEmTZWwj+4fX8N76IcGmIgAcZi9MvS6EpFug63lg1hA8ERFXUgBqIgUgcZVlW1NZ+N6/uM7yHcPNu05+ENoFht4MQ26EoGj3FSgi0o4oADWRApC40jtrUvm/T7bSw3SQV3r/TPcjn0FxxRNiJgv0uhCSpkC3c8FscW+xIiJtmAJQEykAiavN+nI7ry7bh7fFxNuTB3JG8QpjrNCB1ScbhXSBoZONu0LBMW6rVUSkrVIAaiIFIHE1u93B3e//yKKfjxDk68WCO0fRIyoIMrfDxnnw03tQnGM0dt4VugW6nae7QiIi9aQA1EQKQNIcists3PjvtWxIPU6nUD8+njqKyCBf48OyE/DLwoq7QqtOHlR5VyjpFgjs6I6yRUTaDAWgJlIAkuZyrLCUq15Zxf6sQgZ0CuGD352Bv49X1UZHdxpBaPO7J+8KefnCoOth1N0Q3q2lyxYRaRMUgJpIAUiaU0pWIVe+sopjhaWM7RPJqzcNw2I2VW9YVgzbF8LaOXBoY8VOE/S+GEbfA3EjWrRuEZHWriG/vzURiUgLS4gI4PXJw/DxMvPt9kwe/WwbNf7/EG9fGHgt3LYEpnwJPS8EHLDjc3jjAnhjHOxY5FyVXkRE6k8BSMQNkuI78PzEwZhM8NbqVN5Ysb/2xiYTxI+CG96HqetgyE1g8YG0NfD+DTB7hNFlVlbcYvWLiLR16gKrgbrApKW8/sM+/vbFdkwm+NcNQ7lwQD0ff89Ph7WvwoY3Ts4pFBAJI++AYb81Vq4XEfEwGgPURApA0lIcDgczP93G22tSsXqZeff2M0iK71D/E5Tkw6a3YPW/IO+gsc87AIbeBGf8ATrEN0/hIiKtkAJQEykASUsqt9n53dsbWbIjk7AAHz7+wyjiwwMadhJbGWz7GFa+CBlbjH0mC/S7HEb9EWIHu7psEZFWRwGoiRSApKUVlpQz8bXVbD2UR2JEAAvuHEWHAJ+Gn8jhgH3fG0Fo3/cn9yeeBaPuge7nG2OKRETaIQWgJlIAEnfIzCvmin+t4lDOCYYndODt347E17sJs0Af+RlWvQRbPwKHzdgX2c+YS6j/VeDViIAlItKKKQA1kQKQuMuujHyuemUV+cXlXDIwhhevG4K5pjmCGiInDda8UvGkWKGxLygWzrjTmGHaV/+Ni0j7oADURApA4k6r9mQx+c11lNsd/P7sbjx8YW/XnPjEcdjwpvH0WEGGsc8abISgM+6E4FjXfI+IiJsoADWRApC42/yNB7n/fz8B8Lcr+jNppAuf5iovgZ8/MLrHsnYZ+8zeMOAao3ssqq/rvktEpAVpJmiRNu7qpM7cO7YHADM/3caHG9Jqni26MbysxgKrf1gL178PXUaBvQx+ehdeSYZ3rj5l6Q0RkfapVQSg2bNnk5CQgK+vLyNHjmTdunW1tp07dy4mk6nKy9fXt0obh8PBzJkziYmJwc/Pj7Fjx7J79+7mvgwRl7rn/B5cndQZm93Bg/N/5u73fiS3qMx1X2A2Q68L4dYvjeU2+lwKmGDPN/D6+fDZPVB0zHXfJyLSirg9AH3wwQdMmzaNRx55hE2bNjFo0CDGjRtHZmZmrccEBwdz5MgR5ys1NbXK508//TQvvvgic+bMYe3atQQEBDBu3DiKi7VUgLQdJpOJv181kAfG9cJiNvH5z0e48IUfWLsv2/Vf1nkYTHwb7t4IAycCDmPQ9EtDYcN/tN6YiLQ7bh8DNHLkSIYPH87LL78MgN1uJy4ujrvvvpuHH364Wvu5c+dy7733kpOTU+P5HA4HsbGx/OlPf+L+++8HIDc3l6ioKObOnct111132po0Bkham81pOdzz/o+kZhdhNsEfzunOPWN74G1ppv8Pk7ISvngAMrcZ72OHwsXPQqehzfN9IiIu0GbGAJWWlrJx40bGjh3r3Gc2mxk7diyrV6+u9biCggLi4+OJi4vjsssuY9u2bc7P9u/fT3p6epVzhoSEMHLkyFrPWVJSQl5eXpWXSGsyOC6URX88k6uTOmN3wMvf7+HqOatJzS5sni9MGA2/+wHGPwU+QXB4E7x+Hnx+n7rFRKRdcGsAysrKwmazERUVVWV/VFQU6enpNR7Tq1cv3nzzTT799FPeeecd7HY7o0aN4uBBYx2kyuMacs5Zs2YREhLifMXFxTX10kRcLtDqxTPXDOLlG4YQ5OvFT2k5XPTCcv7nygHSp7J4GY/H373hZLfYhjfhpSTYOE/dYiLSprl9DFBDJScnM3nyZAYPHszZZ5/NggUL6NixI6+++mqjzzl9+nRyc3Odr7S0NBdWLOJalwyMZfG9ZzEiIYzCUhsPzP+Zu1w9QPpUQdFw5WtwyxcQ2RdOHIPP/ghvXACHf2ye7xQRaWZuDUARERFYLBYyMjKq7M/IyCA6Orpe5/D29mbIkCHs2bMHwHlcQ85ptVoJDg6u8hJpzTqF+vHeHWc4B0gvas4B0pUqu8XGPWl0ix3aAK+dC59PU7eYiLQ5bg1APj4+JCUlsWTJEuc+u93OkiVLSE5Ortc5bDYbW7ZsISYmBoDExESio6OrnDMvL4+1a9fW+5wibYHFbGLqud356M5RxIf7czi3mOtfX8MzX+2kzNZM3VMWb0ieanSLDbgGo1vsDXh5GGx6S91iItJmuL0LbNq0abz++uvMmzeP7du3c+edd1JYWMiUKVMAmDx5MtOnT3e2f+yxx/j666/Zt28fmzZt4sYbbyQ1NZXbbrsNMB4dvvfee3niiSdYuHAhW7ZsYfLkycTGxnL55Ze74xJFmlWLD5AGo1vsqn/DzZ9Dx95QlA0L74Y3fwOHNzff94qIuIiXuwuYOHEiR48eZebMmaSnpzN48GAWL17sHMR84MABzOaTOe348ePcfvvtpKen06FDB5KSkli1ahV9+56cvv/BBx+ksLCQO+64g5ycHMaMGcPixYurTZgo0l5UDpA+p1dHpi/Y4hwg/ddL+3F1UmdMpiYuqFqbxDPh9yuM9cWWzoKD6+H1c2HYb+G8v4Bfh+b5XhGRJnL7PECtkeYBkrbsUM4J7vtgM+v2G+NyLh4Yw5OXDyDE37t5vzjvCHz9f7B1vvHePwIueBQG3WDMOi0i0sy0GGoTKQBJW2ezO5izbC/PfbMLm91BbIgv/5w4mJFdw5v/y/f/AIvuh6ydxvvOI+DiZyBmUPN/t4h4NAWgJlIAkvbi1BmkTSb4wznduHdsz+abQbqSrQzWvALL/g6lBWAyw/Db4Ny/gF9o83736djtYDIZLxFpVxSAmkgBSNqTgpJy/rpwG/M3GpOFDooL5YWJg0mICGj+L887XNEt9pHx3j8CfvM4DLyu+brFbGWQdwhyDkBOmvFnbsWfOQeMz0I6w4X/gJ6/aZ4aRMQtFICaSAFI2qPPfz7M9AVbyC8uJ8DH0vwDpE+1b5mxtlhlt1jcGUa3WPSAhp+rrBhyD0JOakWw+VXIyT8Cjno+jj/gWmO5j4AW6BoUkWanANRECkDSXrltgDRAeSmsfQWW/h3KCo1usRF3wDnTq3aLlRScEmxqCDkFGbV+hZPFatzlCe0CoXEQ0uXkdlA0rH8D1vzLCEr+4XDh09D/KnWLibRxCkBNpAAk7ZlbB0gD5B4yusW2LTDeB3SEuJEng86Jeswq7R1QEWziTgab0C4VQScOAiJP38V2cCMsvAsyfzHe9xwPFz8HIZ2adn0i4jYKQE2kACSewG0DpCvtW1rRLbar+mfWkF8Fm7iqIcc/zDV3a8pLYcU/4Yd/gL3MWOLjgkchaYoe3RdpgxSAmkgBSDyFWwdIgxFAts6Hkvyqd3N8Q1rm+ytlbjdmsj643ngfPxoufQnCu7VsHSLSJApATaQAJJ7m858P8+cFW8irGCD9u7O7cc2wzsSE+Lm7tJZjt8G612DJY1BWBF6+xvik5LvA4vZJ80WkHhSAmkgBSDzRrwdIm01wds+OTBwex/l9olqua8zdjqfAZ/fCvu+N9zGD4NKXIWagO6sSkXpQAGoiBSDxVDa7g4U/HeK9dWnOIAQQEejDVUM7c+3wOLp1DHRjhS3E4YDN78JX06E4F0wWGHMvnPUgeGtNQZHWSgGoiRSARGDf0QI+3HCQ+RsPklVQ4tw/PKED1w6L4+KBMfj7tPOuofwM+PIB+OVT4314D2NsUHyye+sSkRopADWRApDISWU2O9/vyOSD9Wl8vzMTe8VPjECrFxMGxXLd8DgGdg5pmQkV3WX7Z7DoTyfnIBp+O4x9BKxB7q1LRKpQAGoiBSCRmqXnFvPRpoN8uCGN1Owi5/7e0UFMHB7H5YM70SHAx40VNqMTx+HrGfDj28b74M4w4XnocYFbyxKRkxSAmkgBSKRudruDNfuz+XB9Gl9sTae03Fh6wsdiZlz/aCYOi2NUt3DM5nZ4V2jfUlj4R2OWaoCBE2HcLC2nIdIKKAA1kQKQSP3lFpXxacXA6e1H8pz7O3fw49phce3zcfrSQvj+yVOW04iAi56GfldqOQ0RN1IAaiIFIJGGczgcbD2UxwcbDvDpj4fJLykHjMfpz+rZkeuGx3Fe7yh8vNrR4/QHN8Cnd8HR7cb7XhfBxc9CcKx76xLxUApATaQAJNI0J0ptfLn1CO+vr/o4fXiAD1cldebaYXF0j2wnj9P/ejkNazBc8BgMvVnLaYi0MAWgJlIAEnGd2h6nHxbfgYnD29Hj9JnbjbtBhzYY7xPOhAkvaDkNkRakANRECkAirlf5OP2HG9L4bkfVx+kvGxzL5OQEekW38cfK7TZY+yp89/jJ5TTO/TOcMVXLaYi0AAWgJlIAEmleGXnFzN9Y/XH6M7qGcXNyAhf0jcKrLS+9cTwFPrvHeGIMjEVefUMrPnSA86euw5h1unJHTdvOH9E1bf+6bcXH3n7Q/yoYdisEdnT11Ym0WgpATaQAJNIy7HYHa/Zl89bqVL7+Jd15VygmxJcbz4hn4vA4IgKt7i2ysX69nIY7WHxgwDUw8vday0w8ggJQEykAibS8QzkneHdtKu+tS+NYYSlgzCt0ycAYJo9KYHBcqHsLbKyiY3B4k3F3xvmEvKnicfmKHZXbzkfoa9r+VdtajwOy98LaOXBo48k64sfAGb83nlQzW1x/nSKtgAJQEykAibhPcZmNL7YcYd6qFH46ePLOyaDOIUxOTuDigTH4eusXeL2krYe1rxhrmdmNaQkI7QIjfgdDbgS/ULeWJ+JqCkBNpAAk0jpsTsvhrVUpfP7zEUptxmzTYQE+XDc8jhvPiCc2tJ1NsNhccg/B+n/DxrlwomJaAu8AGHyD0T0W0d2t5Ym4igJQEykAibQuWQUlfLA+jXfWpHIktxgwJlj8Td9oJo+KJ7lrePtejNVVyk7Azx/CmldOTt4I0OM3RhDqdl7rn8m67ARk/gKhCVp+RKpRAGoiBSCR1qncZufb7RnMW5XK6n3Zzv09IgOZPCqBK4d0IsCqx81Py+GA/cuMILTrK5yPj3XsDSN/BwOvAx9/t5boVHQM0tZC6io4sAYO/2hMOGnxgb6XwbDfQpczWn9wkxahANRECkAird+ujHzeWp3Cgk2HKCq1ARBk9eKqpM5MTo6na8d2MtN0c8veC+tegx/fgdICY59vKCTdAiNuh5DOLVtPzgEj6FQGnlPvVFWyhkDJKU/WRfY1HvkfOBF89TPbkykANZECkEjbkXuijI82HuTtNanszyp07j+rZ0duTo7nnF6RWNrjqvSuVpwLP/4X1r1qzGMEYLJAnwlwxh8gboTr77LY7UbAqQw7B9ZA3sHq7SJ6Gnd5uiQbrw4Jxp2gDW/ClvlQfsJo5x0AA681wpAe+/dICkBNpAAk0vbY7Q6W78nirVUpfLcz0zlnYFyYHzedEc+1w+II9fdxb5Ftgd0GuxYb3WMpy0/ujx0CI++EfleAVyP/HstLjOBSGXjS1lSfI8nsBTGDToadLmdAQETt5zyRAz+9DxvegKxdJ/d3Hm50j/W7Arx9G1dve1ZWDJnb4MhPcHgzFGQYTwiGdTOWbwnrCqHxbW4GcwWgJlIAEmnbDmQX8c7aVD5Yn0buiTIArF5mLh/cicmj4ukXG+LmCtuI9K3GfEI/fwi2inXcAqNg+G2QNOX0s0yfyIGD608GnkMbT56nkncAxA2HLqOMsNN5GPgENLxWhwNSVhhBaPtnJx/79+sAgycZd4U8dV220iLI2Hoy7Bz5ybjzVvl3VBuzlxGCwrtVDUbh3SAkrlXOJ6UA1EQKQCLtw4lSGwt/OsTcValsP5Ln3N8jMpBuHQOJj/AnMTyAhIgAEsIDiAq26mmymhRmwcb/wLp/Q0G6sc9iNWaZPuP3ED3A2Jd3+JTurNWQsY1T1v0wBHSsencneqDr7zLkZ8CPb8HGeZCbdnJ/13Nh+G+h54Vt7s5GvZXkQ/qWqmEnayc47NXb+oVB7GCIGWyM9co5AMf2QvY+OLbvZNdiTSw+Rlfkr4NRWDcI7gRm9yxlowDURApAIu2Lw+FgQ+px5q1KYfHWdMrtNf/Y8/U2kxBuhKHKcBQfHkBihMIRAOWlxqSKa1+pOst0pyQoPGr8Av21sK4n7+7EjzLet9Tfo90Gu782xgrt/gZnGAuKhaSbYehkCI5tmVqaQ3GuEXAqX4c3Q/YeqoVOgIDIk2EnZpDxCulc+7+F3Q75RyoC0d5TgtFeOLa/+p28U3n5QofE6sEovBsExTTrv78CUBMpAIm0X0fzS9h6OJeUrELjlV1ESnYhB4+fwFZLMIK6w1FkkBWzpw20TlsPa/5lBCKH8RQeJrNxR6dLMsQnQ9wZEBTl3jorHU8xJoLc9DYUZRn7TBbodaFxVyjxHLfdtaiXomMVQWfzybBzfH/NbYM7nQw5lYEnOMZ1tdhtkHeoejDK3mv8PdvLaj/W298IRWFdjWkMBlzturpQAGoyBSARz1Nms3Pw+AlSso1glJpdxP6swgaFo/hwfxIiAjwrHOUeMu6ydIg3Bh5bg9xdUd3KS4wxQuvfgAOrTu4P6wbDphjjhfzD3FObw2FMRVB41AgVR36sCDs/QW4Nd9fAGLjsDDtDjKffAiNbtu5T2cqNbsdfB6Nje+F46smwDHDm/XD+DJd+vQJQEykAicipagtHqdmFpNUjHMWHBdAl3J8uYSdfcWH+dO7gp3XN3CnjF6N77Kf3oTTf2GexQv8rjSfIOg9reneN3QZF2UaoKcg0xlMVHoXCzIo/s6rur2vcTVjX6nd23BXWGsNWZnSTVgaizsONv2MXUgBqIgUgEamvMpudQ8dPsD+7kNSKLrX6hiOA6GBfZyDqEuZPl3A/5/uOgRp31CJKCmDL/4wnyNK3nNwfPcAIQgOuAespE2uWFv4qvFQGmqzqQacomxrH5NTF298Yn3Nq2IkeoMVr60EBqIkUgETEFSrDUUpFGEo7VkRqdiEHjhnbBSV1P4bs622ucsfo13eQdPfIxRwOY3D3+jdg2wIoN9adwycIOvasCDVZUFZY93mqMYF/uPEEXECE0UVVuR1Qud3RmFYgoGPjpgEQQAGoyRSARKS5ORwOjheVceBYEQeOFZF2rIgD2UXO90dyT3Cam0dEBlmrBaTKcUgRgdaWuZD2qugYbH7X6CI7trf6516+FeElomp4Ceh4cn9l0PELa7+P3bcyCkBNpAAkIu5WWm7ncM4JDhwrIrWGgHS6u0cDO4dw0YAYLh4QQ1xYK1nYtC2y240Zq4uyjWDjvEsTqAVYWyEFoCZSABKR1szhcJBzyt0j5x2kitehnBOc+pN9QKeTYahLuMKQtF8KQE2kACQibdnR/BK+2pbOF1uOsGZfdpWutH6xwc4wlBChsSbSvigANZECkIi0F1kFRhj6cks6q/dlV3kqrW9MMBcPjOGiATEkKgxJO6AA1EQKQCLSHmUXlPD1Lxl8seUIq/ZWDUO9o4O4eEAMFw2MoVvHwDrOItJ6KQA1kQKQiLR3xwpL+XpbOl9sTWfVnqwq66P1jg7iogHGnaHukQpD0nYoADWRApCIeJLjhaV880sGi7YcYeWvwlDPqEDnmKEeUa18mQvxeApATaQAJCKeKqeolK9/yeDLLUdYsSeLMtvJXxE9IgOdd4Z6RgVqlmppdRSAmkgBSEQEcovK+Ga7MWZo+e6jVcJQt44BzjFDvaKCFIakVVAAaiIFIBGRqnJPlPHtLxl8ufUIP+zKotRmd37WNSKAAZ1DiA/zJz48gPhw48+IQB8FI2lRCkBNpAAkIlK7vOIylmzPYNHP6fyw62iVMHSqAB8LXcIDSAj3p0u4PwnhAUZIigggJtgXs1nhSFxLAaiJFIBEROonv7iMVXuz2Z9VSGq2sdhranYRh3Orzkb9az4WM3FhfifvGFUEo/gwfzp38MfHy9xyFyHtRkN+f2t1NhERabQgX2/G9Yuutr+k3MbB4yecgejUcJR2vIhSm529RwvZe7T6yupmE3Tq4Ed8WGV32ilda2EB+PlYWuLSpJ3THaAa6A6QiEjzsdkdzoVeU7ILOZBt/FkZlE6U2eo8PjLISpcwf+IqXx38nNvRwb5Y1LXmsdQF1kQKQCIi7uFwODiaX0LqsSJSsgorQlIRB7ILSckuIvdEWZ3He1tMdAo1AlHnDv7EhfkZYamDEZA6+HtrYHY7pi4wERFpk0wmE5HBvkQG+zI8Iaza5zlFpc5utLRjxl2kg8eLSDtWxKGcE5TZHKRkG6GpJgE+llPuHBkBqTIcxYX54e+jX4ueQv/SIiLSZoT6+xDq78OguNBqn9nsDtLzikk7ZgSitOMnTtkuIiOvhMJSGzvS89mRnl/j+SMCfSruHJ3sWusS5k9MiBHKAq36tdleqAusBuoCExFpf4rLbByqGHt08JSAdKAiJOUVl5/2HP4+FiKDrEQG+dIx2OrcjgyyEhl8cjtUXW1uoS4wERGRX/H1ttCtY2Ctq93nnigjraJLzQhFJyq62opIzy2msNRGUamtzi62Sj4WMx2DrHQMslYLR5HBVjoG+hIZbCU8wAcvix75d4dWEYBmz57NP/7xD9LT0xk0aBAvvfQSI0aMOO1x77//Ptdffz2XXXYZn3zyiXP/Lbfcwrx586q0HTduHIsXL3Z16SIi0k6E+HkT0imE/p1Cavy8sKSczPwSMvOKjT/zS8jML+Zo3sntzPwScorKKLXZOZRzgkM5J+r8TrMJwgJODUlGUIoN9aNHVCA9I4MI8fdujsv1eG4PQB988AHTpk1jzpw5jBw5kueff55x48axc+dOIiMjaz0uJSWF+++/nzPPPLPGz8ePH89//vMf53ur1ery2kVExHMEWL1ItHqRGBFQZ7uSchtHKwNSXglHC0o4+qvQlJlXQlZBCXYHZBUY278cqfl8kUFWekYF0SMqkB6RQfSMCqRHVBAhfgpGTeH2MUAjR45k+PDhvPzyywDY7Xbi4uK4++67efjhh2s8xmazcdZZZ3HrrbeyfPlycnJyqt0B+vW+htAYIBERaW42u4PswoqQdEowyswv4cCxInZn5HM4t7jW46OCrfSINIJRzygjGHWP9Oxg1GbGAJWWlrJx40amT5/u3Gc2mxk7diyrV6+u9bjHHnuMyMhIfvvb37J8+fIa2yxdupTIyEg6dOjAeeedxxNPPEF4eHiNbUtKSigpKXG+z8vLa+QViYiI1I/FbKoYF+Rba5v84jL2ZBawO6OAXRn57MosYE9FMMrIKyEjr4QVe7KqHBMVXHHHyBmOFIxq4tYAlJWVhc1mIyoqqsr+qKgoduzYUeMxK1as4I033mDz5s21nnf8+PFceeWVJCYmsnfvXv785z9z4YUXsnr1aiyW6lOoz5o1i0cffbRJ1yIiIuJqQb7eDOnSgSFdOlTZn19cxu7MAnZn5BvhqGL7yCnBaPnuqsEoOti3Wjdaj6hAgn09Mxi5fQxQQ+Tn53PTTTfx+uuvExERUWu76667zrk9YMAABg4cSLdu3Vi6dCnnn39+tfbTp09n2rRpzvd5eXnExcW5tngREREXCfL1ZmiXDgz9VTDKc94xymdXRoEzJB3JLSY9z3jVFIwSIwII8fMmyNeLQF8vgny9Cfb1Mt5bjf0nX8Z7P29Lm37U360BKCIiAovFQkZGRpX9GRkZREdXX1xv7969pKSkMGHCBOc+u90OgJeXFzt37qRbt27VjuvatSsRERHs2bOnxgBktVo1SFpERNq84DqC0e6MAvZkGsFoV8Wdo8pQlJ5X+1ij2ljMJgKtp4Qia9WQFOhb02cVIcvqRViADwFunFjSrQHIx8eHpKQklixZwuWXXw4YgWbJkiXcdddd1dr37t2bLVu2VNn3f//3f+Tn5/PCCy/Uetfm4MGDZGdnExMT4/JrEBERae2Cfb1Jiu9AUnzNwSjtWBH5JeXkF5eRX2z8WVBcXrFdTl5xGQUl5c7P7A5jEHfuibKK9dnqfty/JreNSeT/LunroitsOLd3gU2bNo2bb76ZYcOGMWLECJ5//nkKCwuZMmUKAJMnT6ZTp07MmjULX19f+vfvX+X40NBQAOf+goICHn30Ua666iqio6PZu3cvDz74IN27d2fcuHEtem0iIiKtWW3BqC4Oh4OiUltFICojryIkFRRXDVD5pwSm/OLyKgEqr7icIDePPXJ7AJo4cSJHjx5l5syZpKenM3jwYBYvXuwcGH3gwAHM5vrPkmmxWPj555+ZN28eOTk5xMbG8pvf/IbHH39c3VwiIiJNZDKZCLB6EWD1Iiq49ifYTsdud+9KXG6fB6g10jxAIiIibU9Dfn9rARIRERHxOApAIiIi4nEUgERERMTjKACJiIiIx1EAEhEREY+jACQiIiIeRwFIREREPI4CkIiIiHgcBSARERHxOApAIiIi4nEUgERERMTjKACJiIiIx1EAEhEREY/j5e4CWiOHwwEYq8qKiIhI21D5e7vy93hdFIBqkJ+fD0BcXJybKxEREZGGys/PJyQkpM42Jkd9YpKHsdvtHD58mKCgIEwmk0vPnZeXR1xcHGlpaQQHB7v03K2Brq/ta+/XqOtr+9r7Ner6Gs/hcJCfn09sbCxmc92jfHQHqAZms5nOnTs363cEBwe3y/+wK+n62r72fo26vravvV+jrq9xTnfnp5IGQYuIiIjHUQASERERj6MA1MKsViuPPPIIVqvV3aU0C11f29fer1HX1/a192vU9bUMDYIWERERj6M7QCIiIuJxFIBERETE4ygAiYiIiMdRABIRERGPowDUgmbPnk1CQgK+vr6MHDmSdevWubskl5k1axbDhw8nKCiIyMhILr/8cnbu3OnusprNU089hclk4t5773V3KS5z6NAhbrzxRsLDw/Hz82PAgAFs2LDB3WW5jM1mY8aMGSQmJuLn50e3bt14/PHH67VmUGv0ww8/MGHCBGJjYzGZTHzyySdVPnc4HMycOZOYmBj8/PwYO3Ysu3fvdk+xjVTXNZaVlfHQQw8xYMAAAgICiI2NZfLkyRw+fNh9BTfQ6f4NT/X73/8ek8nE888/32L1NVV9rm/79u1ceumlhISEEBAQwPDhwzlw4ECL1KcA1EI++OADpk2bxiOPPMKmTZsYNGgQ48aNIzMz092lucSyZcuYOnUqa9as4ZtvvqGsrIzf/OY3FBYWurs0l1u/fj2vvvoqAwcOdHcpLnP8+HFGjx6Nt7c3X375Jb/88gvPPvssHTp0cHdpLvP3v/+dV155hZdffpnt27fz97//naeffpqXXnrJ3aU1SmFhIYMGDWL27Nk1fv7000/z4osvMmfOHNauXUtAQADjxo2juLi4hSttvLqusaioiE2bNjFjxgw2bdrEggUL2LlzJ5deeqkbKm2c0/0bVvr4449Zs2YNsbGxLVSZa5zu+vbu3cuYMWPo3bs3S5cu5eeff2bGjBn4+vq2TIEOaREjRoxwTJ061fneZrM5YmNjHbNmzXJjVc0nMzPTATiWLVvm7lJcKj8/39GjRw/HN9984zj77LMd99xzj7tLcomHHnrIMWbMGHeX0awuvvhix6233lpl35VXXumYNGmSmypyHcDx8ccfO9/b7XZHdHS04x//+IdzX05OjsNqtTree+89N1TYdL++xpqsW7fOAThSU1NbpigXqu36Dh486OjUqZNj69atjvj4eMc///nPFq/NFWq6vokTJzpuvPFG9xTkcDh0B6gFlJaWsnHjRsaOHevcZzabGTt2LKtXr3ZjZc0nNzcXgLCwMDdX4lpTp07l4osvrvJv2R4sXLiQYcOGcc011xAZGcmQIUN4/fXX3V2WS40aNYolS5awa9cuAH766SdWrFjBhRde6ObKXG///v2kp6dX+e80JCSEkSNHttufOWD83DGZTISGhrq7FJew2+3cdNNNPPDAA/Tr18/d5biU3W5n0aJF9OzZk3HjxhEZGcnIkSPr7AZ0NQWgFpCVlYXNZiMqKqrK/qioKNLT091UVfOx2+3ce++9jB49mv79+7u7HJd5//332bRpE7NmzXJ3KS63b98+XnnlFXr06MFXX33FnXfeyR//+EfmzZvn7tJc5uGHH+a6666jd+/eeHt7M2TIEO69914mTZrk7tJcrvLniqf8zAEoLi7moYce4vrrr283C4j+/e9/x8vLiz/+8Y/uLsXlMjMzKSgo4KmnnmL8+PF8/fXXXHHFFVx55ZUsW7asRWrQavDiclOnTmXr1q2sWLHC3aW4TFpaGvfccw/ffPNNy/VPtyC73c6wYcN48sknARgyZAhbt25lzpw53HzzzW6uzjU+/PBD/vvf//Luu+/Sr18/Nm/ezL333ktsbGy7uUZPVVZWxrXXXovD4eCVV15xdzkusXHjRl544QU2bdqEyWRydzkuZ7fbAbjsssu47777ABg8eDCrVq1izpw5nH322c1eg+4AtYCIiAgsFgsZGRlV9mdkZBAdHe2mqprHXXfdxeeff873339P586d3V2Oy2zcuJHMzEyGDh2Kl5cXXl5eLFu2jBdffBEvLy9sNpu7S2ySmJgY+vbtW2Vfnz59WuxpjJbwwAMPOO8CDRgwgJtuuon77ruvXd7Rq/y54gk/cyrDT2pqKt988027ufuzfPlyMjMz6dKli/NnTmpqKn/6059ISEhwd3lNFhERgZeXl1t/7igAtQAfHx+SkpJYsmSJc5/dbmfJkiUkJye7sTLXcTgc3HXXXXz88cd89913JCYmursklzr//PPZsmULmzdvdr6GDRvGpEmT2Lx5MxaLxd0lNsno0aOrTVuwa9cu4uPj3VSR6xUVFWE2V/2RZ7FYnP9PtD1JTEwkOjq6ys+cvLw81q5d225+5sDJ8LN7926+/fZbwsPD3V2Sy9x00038/PPPVX7mxMbG8sADD/DVV1+5u7wm8/HxYfjw4W79uaMusBYybdo0br75ZoYNG8aIESN4/vnnKSwsZMqUKe4uzSWmTp3Ku+++y6effkpQUJBznEFISAh+fn5urq7pgoKCqo1nCggIIDw8vF2Mc7rvvvsYNWoUTz75JNdeey3r1q3jtdde47XXXnN3aS4zYcIE/va3v9GlSxf69evHjz/+yHPPPcett97q7tIapaCggD179jjf79+/n82bNxMWFkaXLl249957eeKJJ+jRoweJiYnMmDGD2NhYLr/8cvcV3UB1XWNMTAxXX301mzZt4vPPP8dmszl/7oSFheHj4+OusuvtdP+Gvw503t7eREdH06tXr5YutVFOd30PPPAAEydO5KyzzuLcc89l8eLFfPbZZyxdurRlCnTb82ce6KWXXnJ06dLF4ePj4xgxYoRjzZo17i7JZYAaX//5z3/cXVqzaU+PwTscDsdnn33m6N+/v8NqtTp69+7teO2119xdkkvl5eU57rnnHkeXLl0cvr6+jq5duzr+8pe/OEpKStxdWqN8//33Nf5v7uabb3Y4HMaj8DNmzHBERUU5rFar4/zzz3fs3LnTvUU3UF3XuH///lp/7nz//ffuLr1eTvdv+Gtt7TH4+lzfG2+84ejevbvD19fXMWjQIMcnn3zSYvWZHI42Og2qiIiISCNpDJCIiIh4HAUgERER8TgKQCIiIuJxFIBERETE4ygAiYiIiMdRABIRERGPowAkIiIiHkcBSERERDyOApCISD0sXboUk8lETk6Ou0sRERdQABIRERGPowAkIiIiHkcBSETaBLvdzqxZs0hMTMTPz49BgwYxf/584GT31KJFixg4cCC+vr6cccYZbN26tco5PvroI/r164fVaiUhIYFnn322yuclJSU89NBDxMXFYbVa6d69O2+88UaVNhs3bmTYsGH4+/szatQodu7c2bwXLiLNQgFIRNqEWbNm8dZbbzFnzhy2bdvGfffdx4033siyZcucbR544AGeffZZ1q9fT8eOHZkwYQJlZWWAEVyuvfZarrvuOrZs2cJf//pXZsyYwdy5c53HT548mffee48XX3yR7du38+qrrxIYGFiljr/85S88++yzbNiwAS8vL2699dYWuX4RcS2tBi8irV5JSQlhYWF8++23JCcnO/ffdtttFBUVcccdd3Duuefy/vvvM3HiRACOHTtG586dmTt3Ltdeey2TJk3i6NGjfP31187jH3zwQRYtWsS2bdvYtWsXvXr14ptvvmHs2LHVali6dCnnnnsu3377Leeffz4AX3zxBRdffDEnTpzA19e3mf8WRMSVdAdIRFq9PXv2UFRUxAUXXEBgYKDz9dZbb7F3715nu1PDUVhYGL169WL79u0AbN++ndGjR1c57+jRo9m9ezc2m43NmzdjsVg4++yz66xl4MCBzu2YmBgAMjMzm3yNItKyvNxdgIjI6RQUFACwaNEiOnXqVOUzq9VaJQQ1lp+fX73aeXt7O7dNJhNgjE8SkbZFd4BEpNXr27cvVquVAwcO0L179yqvuLg4Z7s1a9Y4t48fP86uXbvo06cPAH369GHlypVVzrty5Up69uyJxWJhwIAB2O32KmOKRKT90h0gEWn1goKCuP/++7nvvvuw2+2MGTOG3NxcVq5cSXBwMPHx8QA89thjhIeHExUVxV/+8hciIiK4/PLLAfjTn/7E8OHDefzxx5k4cSKrV6/m5Zdf5l//+hcACQkJ3Hzzzdx66628+OKLDBo0iNTUVDIzM7n22mvddeki0kwUgESkTXj88cfp2LEjs2bNYt++fYSGhjJ06FD+/Oc/O7ugnnrqKe655x52797N4MGD+eyzz/Dx8QFg6NChfPjhh8ycOZPHH3+cmJgYHnvsMW655Rbnd7zyyiv8+c9/5g9/+APZ2dl06dKFP//5z+64XBFpZnoKTETavMontI4fP05oaKi7yxGRNkBjgERERMTjKACJiIiIx1EXmIiIiHgc3QESERERj6MAJCIiIh5HAUhEREQ8jgKQiIiIeBwFIBEREfE4CkAiIiLicRSARERExOMoAImIiIjH+X960zjS1K6uRAAAAABJRU5ErkJggg==\n" }, "metadata": {} } ], "source": [ "plt.plot(history.history['loss'])\n", "plt.plot(history.history['val_loss'])\n", "plt.xlabel('epoch')\n", "plt.ylabel('loss')\n", "plt.legend(['train', 'val'])\n", "plt.show()" ] }, { "cell_type": "markdown", "metadata": { "id": "Jisn0P9h6k2T" }, "source": [ "## GRU 신경망 훈련하기" ] }, { "cell_type": "code", "execution_count": 13, "metadata": { "colab": { "base_uri": "https://localhost:8080/", "height": 230 }, "id": "dwYF578d6k2T", "outputId": "ecc04a1b-9c37-46b0-ce69-da653ab97a21" }, "outputs": [ { "output_type": "display_data", "data": { "text/plain": [ "\u001b[1mModel: \"sequential_3\"\u001b[0m\n" ], "text/html": [ "
Model: \"sequential_3\"\n",
              "
\n" ] }, "metadata": {} }, { "output_type": "display_data", "data": { "text/plain": [ "┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━┓\n", "┃\u001b[1m \u001b[0m\u001b[1mLayer (type) \u001b[0m\u001b[1m \u001b[0m┃\u001b[1m \u001b[0m\u001b[1mOutput Shape \u001b[0m\u001b[1m \u001b[0m┃\u001b[1m \u001b[0m\u001b[1m Param #\u001b[0m\u001b[1m \u001b[0m┃\n", "┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━┩\n", "│ embedding_3 (\u001b[38;5;33mEmbedding\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m100\u001b[0m, \u001b[38;5;34m16\u001b[0m) │ \u001b[38;5;34m8,000\u001b[0m │\n", "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n", "│ gru (\u001b[38;5;33mGRU\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m8\u001b[0m) │ \u001b[38;5;34m624\u001b[0m │\n", "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n", "│ dense_3 (\u001b[38;5;33mDense\u001b[0m) │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m1\u001b[0m) │ \u001b[38;5;34m9\u001b[0m │\n", "└──────────────────────────────────────┴─────────────────────────────┴─────────────────┘\n" ], "text/html": [ "
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━┓\n",
              "┃ Layer (type)                         ┃ Output Shape                ┃         Param # ┃\n",
              "┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━┩\n",
              "│ embedding_3 (Embedding)              │ (None, 100, 16)             │           8,000 │\n",
              "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n",
              "│ gru (GRU)                            │ (None, 8)                   │             624 │\n",
              "├──────────────────────────────────────┼─────────────────────────────┼─────────────────┤\n",
              "│ dense_3 (Dense)                      │ (None, 1)                   │               9 │\n",
              "└──────────────────────────────────────┴─────────────────────────────┴─────────────────┘\n",
              "
\n" ] }, "metadata": {} }, { "output_type": "display_data", "data": { "text/plain": [ "\u001b[1m Total params: \u001b[0m\u001b[38;5;34m8,633\u001b[0m (33.72 KB)\n" ], "text/html": [ "
 Total params: 8,633 (33.72 KB)\n",
              "
\n" ] }, "metadata": {} }, { "output_type": "display_data", "data": { "text/plain": [ "\u001b[1m Trainable params: \u001b[0m\u001b[38;5;34m8,633\u001b[0m (33.72 KB)\n" ], "text/html": [ "
 Trainable params: 8,633 (33.72 KB)\n",
              "
\n" ] }, "metadata": {} }, { "output_type": "display_data", "data": { "text/plain": [ "\u001b[1m Non-trainable params: \u001b[0m\u001b[38;5;34m0\u001b[0m (0.00 B)\n" ], "text/html": [ "
 Non-trainable params: 0 (0.00 B)\n",
              "
\n" ] }, "metadata": {} } ], "source": [ "model4 = keras.Sequential()\n", "\n", "model4.add(keras.layers.Embedding(500, 16, input_shape=(100,)))\n", "model4.add(keras.layers.GRU(8))\n", "model4.add(keras.layers.Dense(1, activation='sigmoid'))\n", "\n", "model4.summary()" ] }, { "cell_type": "code", "execution_count": 14, "metadata": { "colab": { "base_uri": "https://localhost:8080/" }, "id": "j2SfOZ3v6k2U", "outputId": "deb2b2fc-f0bd-4509-d0d3-9b7d77271b3e" }, "outputs": [ { "output_type": "stream", "name": "stdout", "text": [ "Epoch 1/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 8ms/step - accuracy: 0.5052 - loss: 0.6931 - val_accuracy: 0.5206 - val_loss: 0.6928\n", "Epoch 2/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m6s\u001b[0m 10ms/step - accuracy: 0.5532 - loss: 0.6925 - val_accuracy: 0.5616 - val_loss: 0.6921\n", "Epoch 3/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.5831 - loss: 0.6917 - val_accuracy: 0.5714 - val_loss: 0.6911\n", "Epoch 4/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 7ms/step - accuracy: 0.5901 - loss: 0.6903 - val_accuracy: 0.5844 - val_loss: 0.6893\n", "Epoch 5/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.6033 - loss: 0.6882 - val_accuracy: 0.5938 - val_loss: 0.6866\n", "Epoch 6/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 8ms/step - accuracy: 0.6136 - loss: 0.6849 - val_accuracy: 0.6036 - val_loss: 0.6822\n", "Epoch 7/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 10ms/step - accuracy: 0.6197 - loss: 0.6796 - val_accuracy: 0.6120 - val_loss: 0.6752\n", "Epoch 8/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 7ms/step - accuracy: 0.6300 - loss: 0.6713 - val_accuracy: 0.6294 - val_loss: 0.6640\n", "Epoch 9/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.6418 - loss: 0.6578 - val_accuracy: 0.6488 - val_loss: 0.6453\n", "Epoch 10/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.6644 - loss: 0.6351 - val_accuracy: 0.6844 - val_loss: 0.6111\n", "Epoch 11/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 11ms/step - accuracy: 0.6996 - loss: 0.5917 - val_accuracy: 0.7326 - val_loss: 0.5492\n", "Epoch 12/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.7397 - loss: 0.5360 - val_accuracy: 0.7484 - val_loss: 0.5242\n", "Epoch 13/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.7523 - loss: 0.5158 - val_accuracy: 0.7634 - val_loss: 0.5107\n", "Epoch 14/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.7617 - loss: 0.5014 - val_accuracy: 0.7724 - val_loss: 0.5008\n", "Epoch 15/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 7ms/step - accuracy: 0.7712 - loss: 0.4883 - val_accuracy: 0.7750 - val_loss: 0.4949\n", "Epoch 16/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 10ms/step - accuracy: 0.7816 - loss: 0.4780 - val_accuracy: 0.7762 - val_loss: 0.4916\n", "Epoch 17/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 8ms/step - accuracy: 0.7873 - loss: 0.4701 - val_accuracy: 0.7766 - val_loss: 0.4870\n", "Epoch 18/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.7921 - loss: 0.4630 - val_accuracy: 0.7782 - val_loss: 0.4798\n", "Epoch 19/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.7938 - loss: 0.4567 - val_accuracy: 0.7816 - val_loss: 0.4736\n", "Epoch 20/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.7963 - loss: 0.4512 - val_accuracy: 0.7846 - val_loss: 0.4702\n", "Epoch 21/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 10ms/step - accuracy: 0.7992 - loss: 0.4466 - val_accuracy: 0.7868 - val_loss: 0.4677\n", "Epoch 22/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 8ms/step - accuracy: 0.8031 - loss: 0.4426 - val_accuracy: 0.7878 - val_loss: 0.4655\n", "Epoch 23/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.8064 - loss: 0.4392 - val_accuracy: 0.7868 - val_loss: 0.4636\n", "Epoch 24/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.8089 - loss: 0.4364 - val_accuracy: 0.7880 - val_loss: 0.4621\n", "Epoch 25/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.8102 - loss: 0.4341 - val_accuracy: 0.7888 - val_loss: 0.4608\n", "Epoch 26/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 10ms/step - accuracy: 0.8112 - loss: 0.4320 - val_accuracy: 0.7894 - val_loss: 0.4597\n", "Epoch 27/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 7ms/step - accuracy: 0.8121 - loss: 0.4303 - val_accuracy: 0.7878 - val_loss: 0.4587\n", "Epoch 28/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.8134 - loss: 0.4288 - val_accuracy: 0.7882 - val_loss: 0.4578\n", "Epoch 29/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.8133 - loss: 0.4274 - val_accuracy: 0.7890 - val_loss: 0.4570\n", "Epoch 30/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 10ms/step - accuracy: 0.8134 - loss: 0.4262 - val_accuracy: 0.7908 - val_loss: 0.4562\n", "Epoch 31/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 7ms/step - accuracy: 0.8124 - loss: 0.4250 - val_accuracy: 0.7892 - val_loss: 0.4554\n", "Epoch 32/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.8128 - loss: 0.4240 - val_accuracy: 0.7900 - val_loss: 0.4547\n", "Epoch 33/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 7ms/step - accuracy: 0.8132 - loss: 0.4230 - val_accuracy: 0.7906 - val_loss: 0.4540\n", "Epoch 34/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 9ms/step - accuracy: 0.8131 - loss: 0.4220 - val_accuracy: 0.7908 - val_loss: 0.4533\n", "Epoch 35/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 9ms/step - accuracy: 0.8144 - loss: 0.4211 - val_accuracy: 0.7894 - val_loss: 0.4527\n", "Epoch 36/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.8145 - loss: 0.4203 - val_accuracy: 0.7898 - val_loss: 0.4521\n", "Epoch 37/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 7ms/step - accuracy: 0.8139 - loss: 0.4194 - val_accuracy: 0.7904 - val_loss: 0.4515\n", "Epoch 38/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 7ms/step - accuracy: 0.8147 - loss: 0.4187 - val_accuracy: 0.7898 - val_loss: 0.4509\n", "Epoch 39/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 10ms/step - accuracy: 0.8152 - loss: 0.4179 - val_accuracy: 0.7894 - val_loss: 0.4504\n", "Epoch 40/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 7ms/step - accuracy: 0.8154 - loss: 0.4172 - val_accuracy: 0.7884 - val_loss: 0.4499\n", "Epoch 41/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.8158 - loss: 0.4165 - val_accuracy: 0.7884 - val_loss: 0.4495\n", "Epoch 42/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 7ms/step - accuracy: 0.8158 - loss: 0.4159 - val_accuracy: 0.7880 - val_loss: 0.4490\n", "Epoch 43/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 10ms/step - accuracy: 0.8160 - loss: 0.4153 - val_accuracy: 0.7882 - val_loss: 0.4486\n", "Epoch 44/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 9ms/step - accuracy: 0.8159 - loss: 0.4147 - val_accuracy: 0.7868 - val_loss: 0.4481\n", "Epoch 45/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.8166 - loss: 0.4141 - val_accuracy: 0.7872 - val_loss: 0.4477\n", "Epoch 46/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.8166 - loss: 0.4136 - val_accuracy: 0.7864 - val_loss: 0.4473\n", "Epoch 47/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.8169 - loss: 0.4131 - val_accuracy: 0.7866 - val_loss: 0.4469\n", "Epoch 48/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 9ms/step - accuracy: 0.8170 - loss: 0.4126 - val_accuracy: 0.7870 - val_loss: 0.4465\n", "Epoch 49/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 9ms/step - accuracy: 0.8172 - loss: 0.4121 - val_accuracy: 0.7870 - val_loss: 0.4461\n", "Epoch 50/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.8171 - loss: 0.4116 - val_accuracy: 0.7870 - val_loss: 0.4457\n", "Epoch 51/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 7ms/step - accuracy: 0.8175 - loss: 0.4112 - val_accuracy: 0.7872 - val_loss: 0.4453\n", "Epoch 52/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.8175 - loss: 0.4107 - val_accuracy: 0.7876 - val_loss: 0.4449\n", "Epoch 53/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 8ms/step - accuracy: 0.8179 - loss: 0.4103 - val_accuracy: 0.7878 - val_loss: 0.4445\n", "Epoch 54/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m5s\u001b[0m 7ms/step - accuracy: 0.8179 - loss: 0.4099 - val_accuracy: 0.7884 - val_loss: 0.4442\n", "Epoch 55/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 7ms/step - accuracy: 0.8177 - loss: 0.4094 - val_accuracy: 0.7894 - val_loss: 0.4438\n", "Epoch 56/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.8181 - loss: 0.4090 - val_accuracy: 0.7898 - val_loss: 0.4434\n", "Epoch 57/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 9ms/step - accuracy: 0.8185 - loss: 0.4086 - val_accuracy: 0.7906 - val_loss: 0.4431\n", "Epoch 58/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m5s\u001b[0m 8ms/step - accuracy: 0.8183 - loss: 0.4082 - val_accuracy: 0.7906 - val_loss: 0.4427\n", "Epoch 59/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.8186 - loss: 0.4078 - val_accuracy: 0.7912 - val_loss: 0.4424\n", "Epoch 60/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 8ms/step - accuracy: 0.8184 - loss: 0.4075 - val_accuracy: 0.7912 - val_loss: 0.4420\n", "Epoch 61/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 9ms/step - accuracy: 0.8179 - loss: 0.4071 - val_accuracy: 0.7920 - val_loss: 0.4417\n", "Epoch 62/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m5s\u001b[0m 7ms/step - accuracy: 0.8181 - loss: 0.4067 - val_accuracy: 0.7924 - val_loss: 0.4413\n", "Epoch 63/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.8181 - loss: 0.4063 - val_accuracy: 0.7922 - val_loss: 0.4410\n", "Epoch 64/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.8186 - loss: 0.4060 - val_accuracy: 0.7920 - val_loss: 0.4407\n", "Epoch 65/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 9ms/step - accuracy: 0.8190 - loss: 0.4056 - val_accuracy: 0.7922 - val_loss: 0.4404\n", "Epoch 66/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m5s\u001b[0m 8ms/step - accuracy: 0.8190 - loss: 0.4052 - val_accuracy: 0.7922 - val_loss: 0.4400\n", "Epoch 67/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.8193 - loss: 0.4048 - val_accuracy: 0.7926 - val_loss: 0.4397\n", "Epoch 68/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 8ms/step - accuracy: 0.8195 - loss: 0.4045 - val_accuracy: 0.7930 - val_loss: 0.4394\n", "Epoch 69/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 10ms/step - accuracy: 0.8202 - loss: 0.4041 - val_accuracy: 0.7932 - val_loss: 0.4391\n", "Epoch 70/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 9ms/step - accuracy: 0.8204 - loss: 0.4037 - val_accuracy: 0.7938 - val_loss: 0.4388\n", "Epoch 71/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m5s\u001b[0m 8ms/step - accuracy: 0.8205 - loss: 0.4034 - val_accuracy: 0.7932 - val_loss: 0.4385\n", "Epoch 72/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.8207 - loss: 0.4030 - val_accuracy: 0.7936 - val_loss: 0.4382\n", "Epoch 73/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 9ms/step - accuracy: 0.8207 - loss: 0.4027 - val_accuracy: 0.7940 - val_loss: 0.4380\n", "Epoch 74/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 10ms/step - accuracy: 0.8206 - loss: 0.4023 - val_accuracy: 0.7942 - val_loss: 0.4377\n", "Epoch 75/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 7ms/step - accuracy: 0.8204 - loss: 0.4019 - val_accuracy: 0.7948 - val_loss: 0.4374\n", "Epoch 76/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.8208 - loss: 0.4016 - val_accuracy: 0.7944 - val_loss: 0.4371\n", "Epoch 77/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.8206 - loss: 0.4012 - val_accuracy: 0.7948 - val_loss: 0.4369\n", "Epoch 78/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 11ms/step - accuracy: 0.8207 - loss: 0.4009 - val_accuracy: 0.7950 - val_loss: 0.4366\n", "Epoch 79/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.8206 - loss: 0.4005 - val_accuracy: 0.7946 - val_loss: 0.4363\n", "Epoch 80/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.8208 - loss: 0.4001 - val_accuracy: 0.7952 - val_loss: 0.4361\n", "Epoch 81/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 7ms/step - accuracy: 0.8212 - loss: 0.3998 - val_accuracy: 0.7950 - val_loss: 0.4358\n", "Epoch 82/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.8213 - loss: 0.3994 - val_accuracy: 0.7954 - val_loss: 0.4356\n", "Epoch 83/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 11ms/step - accuracy: 0.8215 - loss: 0.3990 - val_accuracy: 0.7960 - val_loss: 0.4353\n", "Epoch 84/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.8218 - loss: 0.3987 - val_accuracy: 0.7960 - val_loss: 0.4351\n", "Epoch 85/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 8ms/step - accuracy: 0.8223 - loss: 0.3983 - val_accuracy: 0.7958 - val_loss: 0.4348\n", "Epoch 86/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.8222 - loss: 0.3980 - val_accuracy: 0.7962 - val_loss: 0.4346\n", "Epoch 87/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.8221 - loss: 0.3976 - val_accuracy: 0.7962 - val_loss: 0.4344\n", "Epoch 88/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 11ms/step - accuracy: 0.8225 - loss: 0.3972 - val_accuracy: 0.7960 - val_loss: 0.4341\n", "Epoch 89/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 7ms/step - accuracy: 0.8224 - loss: 0.3969 - val_accuracy: 0.7964 - val_loss: 0.4339\n", "Epoch 90/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.8225 - loss: 0.3965 - val_accuracy: 0.7964 - val_loss: 0.4337\n", "Epoch 91/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 7ms/step - accuracy: 0.8226 - loss: 0.3962 - val_accuracy: 0.7968 - val_loss: 0.4335\n", "Epoch 92/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 10ms/step - accuracy: 0.8226 - loss: 0.3958 - val_accuracy: 0.7970 - val_loss: 0.4333\n", "Epoch 93/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 7ms/step - accuracy: 0.8230 - loss: 0.3954 - val_accuracy: 0.7970 - val_loss: 0.4331\n", "Epoch 94/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.8229 - loss: 0.3951 - val_accuracy: 0.7968 - val_loss: 0.4329\n", "Epoch 95/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.8231 - loss: 0.3947 - val_accuracy: 0.7970 - val_loss: 0.4327\n", "Epoch 96/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 9ms/step - accuracy: 0.8232 - loss: 0.3944 - val_accuracy: 0.7966 - val_loss: 0.4325\n", "Epoch 97/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 9ms/step - accuracy: 0.8236 - loss: 0.3940 - val_accuracy: 0.7962 - val_loss: 0.4323\n", "Epoch 98/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m5s\u001b[0m 7ms/step - accuracy: 0.8236 - loss: 0.3936 - val_accuracy: 0.7970 - val_loss: 0.4321\n", "Epoch 99/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 7ms/step - accuracy: 0.8238 - loss: 0.3933 - val_accuracy: 0.7972 - val_loss: 0.4319\n", "Epoch 100/100\n", "\u001b[1m313/313\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 9ms/step - accuracy: 0.8237 - loss: 0.3929 - val_accuracy: 0.7978 - val_loss: 0.4317\n" ] } ], "source": [ "rmsprop = keras.optimizers.RMSprop(learning_rate=1e-4)\n", "model4.compile(optimizer=rmsprop, loss='binary_crossentropy',\n", " metrics=['accuracy'])\n", "\n", "checkpoint_cb = keras.callbacks.ModelCheckpoint('best-gru-model.keras',\n", " save_best_only=True)\n", "early_stopping_cb = keras.callbacks.EarlyStopping(patience=3,\n", " restore_best_weights=True)\n", "\n", "history = model4.fit(train_seq, train_target, epochs=100, batch_size=64,\n", " validation_data=(val_seq, val_target),\n", " callbacks=[checkpoint_cb, early_stopping_cb])" ] }, { "cell_type": "code", "execution_count": 15, "metadata": { "colab": { "base_uri": "https://localhost:8080/", "height": 449 }, "id": "3yUJM4nA6k2U", "outputId": "96b05c6c-9adb-4c9f-afcf-aae0abd0c2c0" }, "outputs": [ { "output_type": "display_data", "data": { "text/plain": [ "
" ], "image/png": "iVBORw0KGgoAAAANSUhEUgAAAkAAAAGwCAYAAABB4NqyAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjcuMSwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/bCgiHAAAACXBIWXMAAA9hAAAPYQGoP6dpAABYoUlEQVR4nO3deXxU1f3/8dfMJDPZN7KHQNhkkU1ZIkJr1SiopeCKFIui1W8ttipaK7ZiXSou1SLKT5SvFPxqK61brSgucakggoIoIvsWtuzLZF9m7u+PmUwSSSBkm2Tm/Xw87mNm7ty585n7UPP2nHPPMRmGYSAiIiLiR8zeLkBERESkqykAiYiIiN9RABIRERG/owAkIiIifkcBSERERPyOApCIiIj4HQUgERER8TsB3i6gO3I6nRw9epTw8HBMJpO3yxEREZFWMAyD0tJSkpOTMZtP3MajANSMo0ePkpqa6u0yREREpA0OHTpE7969T3iMAlAzwsPDAdcFjIiI8HI1IiIi0hp2u53U1FTP3/ETUQBqRn23V0REhAKQiIhID9Oa4SsaBC0iIiJ+RwFIRERE/E63CEBLliwhLS2NoKAg0tPT2bhxY4vH/uQnP8FkMh23XXLJJZ5jDMNgwYIFJCUlERwcTEZGBrt37+6KnyIiIiI9gNfHAK1atYp58+axdOlS0tPTWbRoEZMnT2bnzp3Ex8cfd/zrr79OTU2N53VBQQGjRo3iyiuv9Ox77LHHWLx4MStXrqRfv37ce++9TJ48me+//56goKAu+V0iIiLNcTqdTf6OSesFBgZisVg65FwmwzCMDjlTG6WnpzNu3DieeeYZwPUPRmpqKr/5zW+4++67T/r5RYsWsWDBAo4dO0ZoaCiGYZCcnMwdd9zBnXfeCUBJSQkJCQmsWLGCq6+++qTntNvtREZGUlJSokHQIiLSYWpqati/fz9Op9PbpfRYUVFRJCYmNjvQ+VT+fnu1BaimpoZNmzYxf/58zz6z2UxGRgbr169v1TleeOEFrr76akJDQwHYv38/2dnZZGRkeI6JjIwkPT2d9evXNxuAqqurqa6u9ry22+1t/UkiIiLNMgyDY8eOYbFYSE1NPelEfdKUYRhUVFSQm5sLQFJSUrvO59UAlJ+fj8PhICEhocn+hIQEduzYcdLPb9y4ke+++44XXnjBsy87O9tzjh+es/69H1q4cCH333//qZYvIiLSanV1dVRUVJCcnExISIi3y+mRgoODAcjNzSU+Pr5d3WE9On6+8MILjBgxgvHjx7frPPPnz6ekpMSzHTp0qIMqFBERcXE4HABYrVYvV9Kz1YfH2tradp3HqwEoNjYWi8VCTk5Ok/05OTkkJiae8LPl5eW88sor3HDDDU3213/uVM5ps9k8kx5q8kMREelMWmOyfTrq+nk1AFmtVsaMGUNmZqZnn9PpJDMzkwkTJpzws//617+orq7mmmuuabK/X79+JCYmNjmn3W5nw4YNJz2niIiI+Aev3wY/b948rr32WsaOHcv48eNZtGgR5eXlzJkzB4DZs2eTkpLCwoULm3zuhRdeYPr06fTq1avJfpPJxG233cZDDz3EoEGDPLfBJycnM3369K76WSIiItKNeT0AzZgxg7y8PBYsWEB2djajR49mzZo1nkHMWVlZx42U37lzJ2vXruX9999v9px33XUX5eXl3HTTTRQXFzNp0iTWrFmjOYBERES8KC0tjdtuu43bbrvN26V4fx6g7qiz5gEqKCmlpNqgT68wAiw9evy5iIicoqqqKvbv30+/fv161P+Q/+QnP2H06NEsWrSo3efKy8sjNDS0XXfBneg69ph5gPzNzveXMfq7R9hJEscC+1IaPgBn7CDC+oxm8JCR9I0N1eA4ERHpUQzDwOFwEBBw8kgRFxfXBRW1jpohulCwfR8hpmpONx0go+5TLi1azuW75zM58yICnhnFWw9cyv8ueYSV729gV06pt8sVEZFOZBgGFTV1Xtla2/lz3XXX8emnn/LUU0951t5csWIFJpOJd999lzFjxmCz2Vi7di179+5l2rRpJCQkEBYWxrhx4/jwww+bnC8tLa1JS5LJZOJ///d/ufTSSwkJCWHQoEG89dZbHXmZW6QWoC50xnWLcBbeStHBrdgPfYcjdyfBxbuJr9xLb1M+vY2PIe9jyFvIF2uH8ueI6SSOu4ypZ/QhPqLnNJeKiMjJVdY6GLbgPa989/cPTCbEevII8NRTT7Fr1y6GDx/OAw88AMC2bdsAuPvuu/nLX/5C//79iY6O5tChQ1x88cX8+c9/xmaz8eKLLzJ16lR27txJnz59WvyO+++/n8cee4zHH3+cp59+mlmzZnHw4EFiYmI65se2QAGoK1kCMMcNolfcIHqNvaxhf005NfvXUbj1QwIO/peY0h2cZd7OWWXbOfLR86z44AIO9r2CX100nhG9I71Xv4iI+JXIyEisVishISGeufTqV2p44IEHuOCCCzzHxsTEMGrUKM/rBx98kDfeeIO33nqLW265pcXvuO6665g5cyYADz/8MIsXL2bjxo1MmTKlM36ShwJQd2ANxTr4QhIHX+h6XXKEyvXLMG1aQUptAXcFvELF4Td49NmreXn09dwxZShx4Tbv1iwiIu0SHGjh+wcme+2722vs2LFNXpeVlfGnP/2J1atXc+zYMerq6qisrCQrK+uE5xk5cqTneWhoKBEREZ71vjqTAlB3FJlC8JQ/wfl3w7bXqV63hJC877g/cCWffbuJmd/9mivPO4vrJqZhC2j/P8QiItL1TCZTq7qhuqv6Rcjr3XnnnXzwwQf85S9/YeDAgQQHB3PFFVdQU1NzwvMEBgY2eW0ymXA6nR1e7w9pEHR3FhgEo3+O7ebP4KLHcVqC+JHlO17nDna99zxX/L/PyS+rPvl5RERE2shqtXrWMTuRdevWcd1113HppZcyYsQIEhMTOXDgQOcX2EYKQD2B2QzpN2G+eR1GyjgiTJU8YV3KdXmPcPWzazlUWOHtCkVExEelpaWxYcMGDhw4QH5+foutM4MGDeL1119ny5YtfPPNN/z85z/vkpactlIA6kliB2K6fg2cfx+GOYDLLWu5rmQJVzy7jp3Zum1eREQ63p133onFYmHYsGHExcW1OKbnySefJDo6mrPPPpupU6cyefJkzjzzzC6utvU0E3QzOmsm6A713WsYr96ACYPFddN5IXAWy68by5i+nXvboIiItE1PnQm6u+momaDVAtRTDb8c00+fBOC3AW9yec1bzPrfDXydVeTlwkRERLo/BaCebOz1cN69ACwI/D8udnzCvH9+Q0VNnZcLExER6d4UgHq6H90BE1wTTD1mfZ64gq949N0dXi5KRESke1MA6ulMJrjwIRhxJQE4uT9wBS+t38fa3fnerkxERKTbUgDyBSYTXPQYBEUx1HyImZaP+N2r31BSWevtykRERLolBSBfERID5/0RgN8FvkpFST73v7XNy0WJiIh0TwpAvmTMHIgfRiSl3B7wGq9/fYQ13x3zdlUiIiLdjgKQL7EEwJRHAJgd8CGnmQ5x77+3UV138inMRURE/IkCkK/pfw4MnYoZB38Oepm80ipWf6tWIBER8Y60tDQWLVrk7TKOowDkiy58CCw2xhnfcqH5K/627gCa8FtERKSBApAvik6Ds38DwD2B/+C7I0Vszir2akkiIiLdiQKQr5p0O1jDSTNlM9a0ixWfH/B2RSIi0sM8//zzJCcnH7eq+7Rp07j++uvZu3cv06ZNIyEhgbCwMMaNG8eHH37opWpPjQKQr7KFwbBpAFxq+Yx3tx4ju6TKy0WJiIiHYUBNuXe2Vg6LuPLKKykoKODjjz/27CssLGTNmjXMmjWLsrIyLr74YjIzM/n666+ZMmUKU6dObXHF+O4kwNsFSCcaeRVseYlpgRu5v/JaXt5wkDsuHOztqkREBKC2Ah5O9s5333MUrKEnPSw6OpqLLrqIv//975x//vkAvPrqq8TGxnLuuediNpsZNWqU5/gHH3yQN954g7feeotbbrml08rvCGoB8mVpkyA8mVCjnJ+Yt/D3DVlU1eqWeBERab1Zs2bx2muvUV1dDcDLL7/M1VdfjdlspqysjDvvvJOhQ4cSFRVFWFgY27dvVwuQeJnZAiOvhHVPMTNoPe+Vj+ftb49xxZje3q5MREQCQ1wtMd767laaOnUqhmGwevVqxo0bx2effcZf//pXAO68804++OAD/vKXvzBw4ECCg4O54oorqKmp6azKO4wCkK8bOQPWPcWPjE1EUsbf1u3n8jNTMJlM3q5MRMS/mUyt6obytqCgIC677DJefvll9uzZw+DBgznzzDMBWLduHddddx2XXnopAGVlZRw4cMCL1baeusB8XcLpkDAci1HHtMCNbDtq56uDRd6uSkREepBZs2axevVqli9fzqxZszz7Bw0axOuvv86WLVv45ptv+PnPf37cHWPdlQKQPxg5A4DrwjcC8J9vvNTkKiIiPdJ5551HTEwMO3fu5Oc//7ln/5NPPkl0dDRnn302U6dOZfLkyZ7Woe5OXWD+YMQV8MEC+ld8S29TLuv2dP8mVxER6T7MZjNHjx7/P89paWl89NFHTfbNnTu3yevu2iWmFiB/EJEM/X4MwKWWdezNK9ecQCIi4tcUgPzFqKsBmGFbDxis25Pv3XpERES8SAHIXwz5KQQE09txmJGmfQpAIiLi1xSA/EVQBAy+CIALLV+xdk++VogXERG/pQDkT/qfA8AY8x5yS6vZk1vm5YJERPyP/uezfTrq+ikA+ZPe4wAYbdmPGSdr1Q0mItJlLBYLQI+YJbk7q6ioACAwMLBd59Ft8P4kbghYwwiuKWOg6Qjr9iQyZ2I/b1clIuIXAgICCAkJIS8vj8DAQMxmtUGcCsMwqKioIDc3l6ioKE+gbCsFIH9itkDyGXDgM0ab9/DOvn7UOZwEWPQvoYhIZzOZTCQlJbF//34OHjzo7XJ6rKioKBITE9t9HgUgf9N7LBz4jLMC9/HPqjq+OVzCmL7R3q5KRMQvWK1WBg0apG6wNgoMDGx3y089BSB/kzIWgHTrfqiCdXvyFYBERLqQ2WwmKCjI22X4PfV9+JvergCUXLOfUCo1EFpERPySApC/CU+EyFRMGIw07+PrrCLKq+u8XZWIiEiXUgDyR+5WoHNCDlLrMNh4oNDLBYmIiHQtBSB/5B4H9KOQAwB8rm4wERHxMwpA/sg9IeKA6h2Awdo9Bd6tR0REpIspAPmjpJFgDiCoOp8U8tl+zI69qtbbVYmIiHQZrwegJUuWkJaWRlBQEOnp6WzcuPGExxcXFzN37lySkpKw2WycdtppvPPOO573//SnP2EymZpsQ4YM6eyf0bMEBkPCcADOCXVNxrUvr9ybFYmIiHQprwagVatWMW/ePO677z42b97MqFGjmDx5Mrm5uc0eX1NTwwUXXMCBAwd49dVX2blzJ8uWLSMlJaXJcaeffjrHjh3zbGvXru2Kn9OzuLvBJgbtB2CvFkYVERE/4tWJEJ988kluvPFG5syZA8DSpUtZvXo1y5cv5+677z7u+OXLl1NYWMjnn3/uWQQtLS3tuOMCAgJOaZrs6upqqqurPa/tdvsp/pIeqPdY+HIZw43dAOzLVwASERH/4bUWoJqaGjZt2kRGRkZDMWYzGRkZrF+/vtnPvPXWW0yYMIG5c+eSkJDA8OHDefjhh3E4HE2O2717N8nJyfTv359Zs2aRlZV1wloWLlxIZGSkZ0tNTW3/D+zu3C1AKVW7CKROXWAiIuJXvBaA8vPzcTgcJCQkNNmfkJBAdnZ2s5/Zt28fr776Kg6Hg3feeYd7772XJ554goceeshzTHp6OitWrGDNmjU8++yz7N+/nx/96EeUlpa2WMv8+fMpKSnxbIcOHeqYH9mdxfSH4GgCnDUMNR1kb55agERExH/0qLXAnE4n8fHxPP/881gsFsaMGcORI0d4/PHHue+++wC46KKLPMePHDmS9PR0+vbtyz//+U9uuOGGZs9rs9mw2Wxd8hu6DZPJNR/Qng8Ybd7DKwWDcDgNLGaTtysTERHpdF5rAYqNjcVisZCTk9Nkf05OTovjd5KSkjjttNOarAQ7dOhQsrOzW1xZNyoqitNOO409e/Z0XPG+wj0j9BjLXmrqnBwpqvRyQSIiIl3DawHIarUyZswYMjMzPfucTieZmZlMmDCh2c9MnDiRPXv24HQ6Pft27dpFUlISVqu12c+UlZWxd+9ekpKSOvYH+AJ3ABobsBdA3WAiIuI3vHob/Lx581i2bBkrV65k+/bt3HzzzZSXl3vuCps9ezbz58/3HH/zzTdTWFjIrbfeyq5du1i9ejUPP/wwc+fO9Rxz55138umnn3LgwAE+//xzLr30UiwWCzNnzuzy39ftpYxxPTiPEUaFApCIiPgNr44BmjFjBnl5eSxYsIDs7GxGjx7NmjVrPAOjs7KyMJsbMlpqairvvfcet99+OyNHjiQlJYVbb72V3//+955jDh8+zMyZMykoKCAuLo5JkybxxRdfEBcX1+W/r9sLjnZtlUWkmPLZl687wURExD+YDMMwvF1Ed2O324mMjKSkpISIiAhvl9O5lk6C7K3MqfkdFX3PZ9X/NN/9KCIi0t2dyt9vry+FIV4W2QdALUAiIuJXFID8XWRvwBWA8kqrtSiqiIj4BQUgfxflmvW6X2ARoEVRRUTEPygA+Tt3C1BaYCEA+3QnmIiI+AEFIH8X6WoBSjTyAbUAiYiIf1AA8nfuABRRm08AdVoVXkRE/IICkL8LjQOLFTNOEk1F7M1VC5CIiPg+BSB/ZzZ7xgElk8/+gnIcTk0NJSIivk0BSDwBqE9AITV1To4Wa1FUERHxbQpA4hkHNDykBIA9uhNMRER8nAKQeALQAFsxoDvBRETE9ykAScNs0OYCQHMBiYiI71MAEs9s0LF1OQDsVQASEREfpwAkni6w0KpswFAXmIiI+DwFIIGIFAAsdRVEUUZuaTWlWhRVRER8mAKQQGAQhMYDcHqYHdBAaBER8W0KQOLiHgg9Otw1/kdLYoiIiC9TABIXdwAaElwMqAVIRER8mwKQuET1ASDVUgjA4SLNBi0iIr5LAUhc3C1AvepyASiqqPFmNSIiIp1KAUhc3LfCR1QfA6CoXAFIRER8lwKQuLhbgEIqXQGoUC1AIiLiwwK8XYB0E+4WoMDKPGzUUFSufzRERMR3qQVIXEJiIDAEgERTIWXVdVTXObxclIiISOdQABIXk8nTDZbqXhS1uEKzQYuIiG9SAJIG7m6wQbYiAAo1EFpERHyUApA0cLcApQW6ApBuhRcREV+lACQN3C1AfdxdYEXl6gITERHfpAAkDaJcASiRfEC3wouIiO9SAJIG7i6wOId7NmiNARIRER+lACQN3AEoqjYXE04NghYREZ+lACQNIlIAEwFGDb0o1SBoERHxWQpA0sASCOFJACSb8tUCJCIiPksBSJpyD4ROMeWrBUhERHyWApA05R4HlGzK123wIiLisxSApCl3AOqtLjAREfFhCkDSVFgCADGmUiprHVTWaEFUERHxPQpA0lRQJACRpgpAy2GIiIhvUgCSpoKiAOhlcQUgdYOJiIgvUgCSpoKjAIgylQNQXKGB0CIi4nsUgKQpdwtQBK4ApPXARETEFykASVPuMUAhRjlgaD0wERHxSQpA0pS7CyzQqCWIGo0BEhERn6QAJE1Zw8BkASCSct0FJiIiPkkBSJoymTytQJGmcrUAiYiIT/J6AFqyZAlpaWkEBQWRnp7Oxo0bT3h8cXExc+fOJSkpCZvNxmmnncY777zTrnPKD7jHAUWoBUhERHyUVwPQqlWrmDdvHvfddx+bN29m1KhRTJ48mdzc3GaPr6mp4YILLuDAgQO8+uqr7Ny5k2XLlpGSktLmc0oz3HeCuVqAdBu8iIj4Hq8GoCeffJIbb7yROXPmMGzYMJYuXUpISAjLly9v9vjly5dTWFjIm2++ycSJE0lLS+Occ85h1KhRbT4nQHV1NXa7vcnm1+q7wCjXXWAiIuKTvBaAampq2LRpExkZGQ3FmM1kZGSwfv36Zj/z1ltvMWHCBObOnUtCQgLDhw/n4YcfxuFwtPmcAAsXLiQyMtKzpaamdtCv7KHq5wIyVVBYUYNhGN6tR0REpIN5LQDl5+fjcDhISEhosj8hIYHs7OxmP7Nv3z5effVVHA4H77zzDvfeey9PPPEEDz30UJvPCTB//nxKSko826FDh9r563q4+vXAKKemzkmFFkQVEREfE+DtAk6F0+kkPj6e559/HovFwpgxYzhy5AiPP/449913X5vPa7PZsNlsHVhpD+fuAouxVIDDtR5YqK1H/aMiIiJyQl77qxYbG4vFYiEnJ6fJ/pycHBITE5v9TFJSEoGBgVgsFs++oUOHkp2dTU1NTZvOKc1wd4HFBVRCjWtF+NSYEO/WJCIi0oG81gVmtVoZM2YMmZmZnn1Op5PMzEwmTJjQ7GcmTpzInj17cDqdnn27du0iKSkJq9XapnNKM9wtQNGWSkArwouIiO/x6l1g8+bNY9myZaxcuZLt27dz8803U15ezpw5cwCYPXs28+fP9xx/8803U1hYyK233squXbtYvXo1Dz/8MHPnzm31OaUV3GOAorUivIiI+CivDuyYMWMGeXl5LFiwgOzsbEaPHs2aNWs8g5izsrIwmxsyWmpqKu+99x633347I0eOJCUlhVtvvZXf//73rT6ntMIPV4RXC5CIiPgYk6F7nI9jt9uJjIykpKSEiIgIb5fT9Y5+Dc//hJLAOEaVPsVvzhvIHRcO9nZVIiIiJ3Qqf7+9vhSGdEPuFqBgRxmgFiAREfE9CkByPPcYIKuzkgDqtB6YiIj4HAUgOZ47AIFrMkS1AImIiK9RAJLjmS1gc88GbSqnSAuiioiIj1EAkuYFuwJQBK71wERERHyJApA0L6hxC5AWRBUREd+iACTNc98JFkk5dU6D0uo679YjIiLSgRSApHnu5TBiAyoAKNJAaBER8SEKQNI8dxdYgrUa0FxAIiLiWxSApHmNV4QHzQUkIiI+RQFImufuAoux1HeB6VZ4ERHxHQpA0jx3C1C0yR2A1AIkIiI+RAFImucOQOEmrQgvIiK+RwFImufuAgtzuhZEVQuQiIj4EgUgaZ67BSjEWQqoBUhERHyLApA0z90CZKtzBSANghYREV+iACTNc88DFFBbhgmn1gMTERGfogAkzXN3gZkwCKdCM0GLiIhPUQCS5gVYITAEcC+IWlGD06kFUUVExDcoAEnLGi2I6jTAXqVxQCIi4hsUgKRl7nFAie71wArUDSYiIj5CAUha5r4TLCVIC6KKiIhvUQCSlrm7wBKtVQAUlFV7sRgREZGOowAkLXO3AMUHulaEzytTC5CIiPgGBSBpmXsMUK8AtQCJiIhvUQCSlnlWhHctiFqgFiAREfERCkDSMncXWIQ7AOWrBUhERHyEApC0zN0CVL8ivFqARETEVygAScvcY4CC3QFILUAiIuIrFICkZe4uMGutHVAAEhER36EAJC1zd4EF1LgCkL2qjpo6pxcLEhER6RgKQNIydwuQqbqEAPc/KQXlagUSEZGeTwFIWuYeA2Ry1tE71NXyo4HQIiLiCxSApGWBIWAOBKBPiGsleI0DEhERX6AAJC0zmTzdYL3dC6LmqwVIRER8gAKQnJh7IHRykCv4aDkMERHxBQpAcmLucUAJ7gVRC8rVAiQiIj2fApCcmLsLLDbAFYDyS9UCJCIiPZ8CkJyYuwssxlwBQL5agERExAcoAMmJ1S+IWh+A1AIkIiI+QAFITsw9Bii8fkFUTYQoIiI+QAFITszdBRbSaEV4wzC8WJCIiEj7KQDJibm7wGx1rvXA6pwGJZW1XixIRESk/RSA5MTcLUCWajvhQQGAJkMUEZGeTwFITsw9BoiqYuLCbIAmQxQRkZ6vWwSgJUuWkJaWRlBQEOnp6WzcuLHFY1esWIHJZGqyBQUFNTnmuuuuO+6YKVOmdPbP8E3uLjAqi+kVZgXUAiQiIj1fgLcLWLVqFfPmzWPp0qWkp6ezaNEiJk+ezM6dO4mPj2/2MxEREezcudPz2mQyHXfMlClT+Nvf/uZ5bbPZOr54f+DuAqOqmF4J7hYg3QkmIiI9XJtagFauXMnq1as9r++66y6ioqI4++yzOXjw4Cmd68knn+TGG29kzpw5DBs2jKVLlxISEsLy5ctb/IzJZCIxMdGzJSQkHHeMzWZrckx0dPQp1SVu9S1AdVUkhrru/lILkIiI9HRtCkAPP/wwwcHBAKxfv54lS5bw2GOPERsby+23397q89TU1LBp0yYyMjIaCjKbycjIYP369S1+rqysjL59+5Kamsq0adPYtm3bccd88sknxMfHM3jwYG6++WYKCgpaPF91dTV2u73JJm7WcMDVwpZkcwWffI0BEhGRHq5NAejQoUMMHDgQgDfffJPLL7+cm266iYULF/LZZ5+1+jz5+fk4HI7jWnASEhLIzs5u9jODBw9m+fLl/Pvf/+all17C6XRy9tlnc/jwYc8xU6ZM4cUXXyQzM5NHH32UTz/9lIsuugiHw9HsORcuXEhkZKRnS01NbfVv8Hlms6cVKDmwHNAgaBER6fnaFIDCwsI8LSrvv/8+F1xwAQBBQUFUVlZ2XHXNmDBhArNnz2b06NGcc845vP7668TFxfHcc895jrn66qv52c9+xogRI5g+fTpvv/02X375JZ988kmz55w/fz4lJSWe7dChQ536G3qc6H4A9Ha4QmaBusBERKSHa9Mg6AsuuIBf/vKXnHHGGezatYuLL74YgG3btpGWltbq88TGxmKxWMjJyWmyPycnh8TExFadIzAwkDPOOIM9e/a0eEz//v2JjY1lz549nH/++ce9b7PZNEj6ROKGwNHNxFUfBOLVBSYiIj1em1qAlixZwoQJE8jLy+O1116jV69eAGzatImZM2e2+jxWq5UxY8aQmZnp2ed0OsnMzGTChAmtOofD4WDr1q0kJSW1eMzhw4cpKCg44TFyAnGDAYgs3weoBUhERHq+NrUARUVF8cwzzxy3//777z/lc82bN49rr72WsWPHMn78eBYtWkR5eTlz5swBYPbs2aSkpLBw4UIAHnjgAc466ywGDhxIcXExjz/+OAcPHuSXv/wl4Bogff/993P55ZeTmJjI3r17ueuuuxg4cCCTJ09uy88VdwAKKXa1spVW11FV6yAo0OLNqkRERNqsTQFozZo1hIWFMWnSJMDVIrRs2TKGDRvGkiVLTumW8xkzZpCXl8eCBQvIzs5m9OjRrFmzxjMwOisrC7O5oaGqqKiIG2+8kezsbKKjoxkzZgyff/45w4YNA8BisfDtt9+ycuVKiouLSU5O5sILL+TBBx9UN1dbuQOQuXAPQRaockBBeQ0pUcFeLkxERKRtTEYblvYeMWIEjz76KBdffDFbt25l3LhxzJs3j48//pghQ4Y0mYCwJ7Lb7URGRlJSUkJERIS3y/E+pwMeToa6Kq60LuFLezRv3TKRkb2jvF2ZiIiIx6n8/W5TC9D+/fs9LS6vvfYaP/3pT3n44YfZvHmzZ0C0+BCzBXoNgpytjLBl8yXRGggtIiI9WpsGQVutVioqKgD48MMPufDCCwGIiYnRJIK+yt0NNthyFNBs0CIi0rO1qQVo0qRJzJs3j4kTJ7Jx40ZWrVoFwK5du+jdu3eHFijdhDsA9Tc0F5CIiPR8bWoBeuaZZwgICODVV1/l2WefJSUlBYB3331Xq677KncASqlzTRKpLjAREenJ2tQC1KdPH95+++3j9v/1r39td0HSTcUNASC26gBgaDkMERHp0doUgMA1AeGbb77J9u3bATj99NP52c9+hsWiuWF8Ukx/MAdgdVSQRCEF5XHerkhERKTN2hSA9uzZw8UXX8yRI0cYPNjVNbJw4UJSU1NZvXo1AwYM6NAipRuwBELMAMjfySDzYXJL07xdkYiISJu1aQzQb3/7WwYMGMChQ4fYvHkzmzdvJisri379+vHb3/62o2uU7iLuNAAGmY5QUK5B0CIi0nO1qQXo008/5YsvviAmJsazr1evXjzyyCNMnDixw4qTbiZuCGz/DwNMRygsr8HpNDCbTd6uSkRE5JS1qQXIZrNRWlp63P6ysjKsVmu7i5Juyj0QepD5CA6nQXFlrZcLEhERaZs2BaCf/vSn3HTTTWzYsAHDMDAMgy+++IJf/epX/OxnP+voGqW7iHV1gZ1mPoLuBBMRkZ6sTQFo8eLFDBgwgAkTJhAUFERQUBBnn302AwcOZNGiRR1conQbsYMAE5GUE4udPAUgERHpodo0BigqKop///vf7Nmzx3Mb/NChQxk4cGCHFifdTGAwRKdB0X4GmQ9rNmgREemxWh2A5s2bd8L3P/74Y8/zJ598su0VSfcWNxiK9jPAdFRdYCIi0mO1OgB9/fXXrTrOZNJdQT4tbjDsWsMg02Hy1AIkIiI9VKsDUOMWHvFjsa6JLweZjrCjXC1AIiLSM7VpELT4Mfet8APNR8m1KwCJiEjPpAAkp8Y9G3S8qZiDR45gGIaXCxIRETl1CkByamzhOMOTAYgo28/hokovFyQiInLqFIDklJnjG2aE3nSwyMvViIiInDoFIDl1noHQhxWARESkR1IAklPnbgE63XSQrxSARESkB1IAklOX9iMAxpp3cjQ7m9IqLYoqIiI9iwKQnLpeA6DXIAJNDiaavuXrrGJvVyQiInJKFICkbQZPAeB8y2Z1g4mISI+jACRtc5orAJ1r3sLmA3leLkZEROTUKABJ26SehcMaSYypDOPQl9Q5nN6uSEREpNUUgKRtLAGYT7sAgInOTezILvVyQSIiIq2nACRtZhp8EQDnmb/mqwOFXq5GRESk9RSApO0GnIcTC0PMh9i3Z7u3qxEREWk1BSBpu5AYSuPHABB5KNPLxYiIiLSeApC0S/DwSwAYV72BI8VaGFVERHoGBSBpF+vQiwFIN29ny+5DXq5GRESkdRSApH1iB1Fg643NVIf9+w+8XY2IiEirKABJ+5hMlKSeD0CvIx95uRgREZHWUQCSdosaPRWAM6s3UlZV4+VqRERETk4BSNotZsg5lBFCrMnOrq90N5iIiHR/CkDSfgFWdkaf43q+aaV3axEREWkFBSDpEMbY6wE4vehDjPICL1cjIiJyYgpA0iGGjzuPbUYaNmrJ+Wy5t8sRERE5IQUg6RBB1gC+ir3M9XzLSnBqdXgREem+FICkw4SMmYHdCCaq6hDs/8Tb5YiIiLRIAUg6zDnD03jN8WMAqj9/3svViIiItEwBSDpMfEQQm2KnAxC49z0oOeLdgkRERFqgACQdavCIcXzhHIoZJ2zWLfEiItI9dYsAtGTJEtLS0ggKCiI9PZ2NGze2eOyKFSswmUxNtqCgoCbHGIbBggULSEpKIjg4mIyMDHbv3t3ZP0OA84cm8H91FwBgbFoJjlovVyQiInI8rwegVatWMW/ePO677z42b97MqFGjmDx5Mrm5uS1+JiIigmPHjnm2gwcPNnn/scceY/HixSxdupQNGzYQGhrK5MmTqaqq6uyf4/eGJoWzNWwieUYkprJs2LHa2yWJiIgcx+sB6Mknn+TGG29kzpw5DBs2jKVLlxISEsLy5S3PJWMymUhMTPRsCQkJnvcMw2DRokX88Y9/ZNq0aYwcOZIXX3yRo0eP8uabb3bBL/JvJpOJc4b15hXHua4dnz8NTod3ixIREfkBrwagmpoaNm3aREZGhmef2WwmIyOD9evXt/i5srIy+vbtS2pqKtOmTWPbtm2e9/bv3092dnaTc0ZGRpKent7iOaurq7Hb7U02abvzh8bzUl0G5QTBka/gi//n7ZJERESa8GoAys/Px+FwNGnBAUhISCA7O7vZzwwePJjly5fz73//m5deegmn08nZZ5/N4cOHATyfO5VzLly4kMjISM+Wmpra3p/m187q34tSaxwP1P7CtSPzAcj53rtFiYiINOL1LrBTNWHCBGbPns3o0aM555xzeP3114mLi+O5555r8znnz59PSUmJZzt06FAHVux/ggIt/HhQHKscP2Fv9CRw1MAbN0FdjbdLExERAbwcgGJjY7FYLOTk5DTZn5OTQ2JiYqvOERgYyBlnnMGePXsAPJ87lXPabDYiIiKabNI+5w+NB0wscP4PBMdA9lb49BFvlyUiIgJ4OQBZrVbGjBlDZmamZ5/T6SQzM5MJEya06hwOh4OtW7eSlJQEQL9+/UhMTGxyTrvdzoYNG1p9Tmm/84cmYDGbWJdjIe8nj7p2rv0rHGp5igMREZGu4vUusHnz5rFs2TJWrlzJ9u3bufnmmykvL2fOnDkAzJ49m/nz53uOf+CBB3j//ffZt28fmzdv5pprruHgwYP88pe/BFx3Id1222089NBDvPXWW2zdupXZs2eTnJzM9OnTvfET/VJMqJUJ/XsB8FrVGBg5AwwnvPE/UKVB5iIi4l0B3i5gxowZ5OXlsWDBArKzsxk9ejRr1qzxDGLOysrCbG7IaUVFRdx4441kZ2cTHR3NmDFj+Pzzzxk2bJjnmLvuuovy8nJuuukmiouLmTRpEmvWrDluwkTpXBeNSGTtnnze2XqMX93wGBxYC4X7YPlkuPrvENPP2yWKiIifMhmGYXi7iO7GbrcTGRlJSUmJxgO1Q15pNekPf4jTgM/uOpfUqp3w9xlQlgPB0XDlCuj/E2+XKSIiPuJU/n57vQtMfFdcuI3x/WIAWPNdNiSfATd9AiljoLII/u8y+GIpKIOLiEgXUwCSTnXxCNfg9He+O+baEZEM170DI68GwwFrfg9v364QJCIiXUoBSDrVlNMTMZng66xijhZXunYGBsGlS+HCP4PJDJv+Bh//2buFioiIX1EAkk4VHxHEuL6ubrB3v2s0E7fJBGffAlOfcr3+7+OwaUXXFygiIn5JAUg63UUjXBNQvrv12PFvnjkbfnyX6/nb82D3B11YmYiI+CsFIOl0U4a7AtBXB4vILqk6/oBz74FRP3eNCfrntXD06y6uUERE/I0CkHS6pMhgzuwTBcB725pZkNZkcnWF9f8J1JbDy1dBcVaX1igiIv5FAUi6hOdusOa6wQACrHDV/0HCcCjPhXfv7sLqRETE3ygASZeo7wbbeKCQvNLq5g8KioArloPJAjtXw/7/dmGFIiLiTxSApEv0jg5hVGoUhgFvfXO05QPjBsPY613P37sHnI6uKVBERPyKApB0mSvOTAHg7xsOcsIVWH4yH2yRkL0VvvlHF1UnIiL+RAFIusz0M1IItVrYm1fO+n0FLR8Y2gvO+Z3reeYDUF3WNQWKiIjfUACSLhMeFMj0M1ytQC9/cZK7vMbfBNH9XAunrnuqC6oTERF/ogAkXeqas/oCrtvhc+3NzAlUL8AGFzzgev7501ByuAuqExERf6EAJF1qaFIEY/pGU+c0eOXLQyc5eCr0ORvqKl1dYSIiIh1EAUi63C/crUD/2JhFncPZ8oEmE0x2L5L67SrY/1kXVCciIv5AAUi63EUjEokJtXKspIqPduSe+OCUM2HMHNfz//wWais7v0AREfF5CkDS5WwBFq4c2xuAlza0YsmLC+6H8CQo3AefLOzk6kRExB8oAIlXzBrfF5MJ/rsrj4MF5Sc+OCgSLnnS9fzzZ+Dolk6vT0REfJsCkHhFn14hnHNaHAB/b00r0JCL4fRLXSvGv3ULOGo7uUIREfFlCkDiNdekuwZDv/LlIexVrQg0Fz0GQVGuGaI/f7pzixMREZ+mACRec+6QeAbFh1FSWcuy/+47+QfC4mGKewzQJ49A3q7OLVBERHyWApB4jcVs4o4LBwPwwtr9La8S39iomTDgPHBUw4pL4NCXnVyliIj4IgUg8arJpycwKjWKihoHSz7ec/IPmEww7f9BwnAoz4UVF8M3r3R+oSIi4lMUgMSrTCYTv5/sagV6ecNBDhVWnPxDEUlw/Xsw5KfgqIE3/gc+WABORydXKyIivkIBSLzu7IGxTBoYS63DYNGHu1v3IVsYXPV/8GP3qvHrnoJ/XA32Y51XqIiI+AwFIOkWfuduBXrj68Psyilt3YfMZjjvj3D5CxAQBLvfhyXjYeMytQaJiMgJKQBJtzAqNYqLhifiNOAv7+08tQ+PuAJ+mQkpY6DaDu/cCS9cCNnfdU6xIiLS4ykASbdxx4WnYTbB+9/nsDmr6NQ+nDgcbvgALv4LWMPhyFfw3I/h3buhvKBzChYRkR5LAUi6jYHx4VwxxrVG2AP/+R6n0zi1E5gtMP5GuGUjDP2Za9boDc/C4tHw38eh5iRLboiIiN9QAJJu5c4LBxNmC2DLoWJe3Xy4bSeJSIYZ/we/eAMSR7q6xT56CBafCV8th7pWzDckIiI+TQFIupX4iCBuPX8QAI++u4OSynas+TXgPLjpU7jsfyGqD5Rlw9u3w1OjYN1iqLJ3UNUiItLTKABJt3PdxDQGxodRUF7Dog/budyF2Qwjr4RbvoIpj0J4EpQegw/uhb8Ohw//BPajHVK3iIj0HApA0u0EWsz8aerpALy4/iA7sjugpSbABmf9Cm79BqYtgdjToLoE1v4V/no6vHwlfP9vqKtp/3eJiEi3pwAk3dKkQbFcNDwRh9Pgvn9vwzBOcUB0SwJscMY18OsNcPXfoc/ZYDhdcwj9czY8OcR151jWF+B0dsx3iohIt2MyOuwvi++w2+1ERkZSUlJCRESEt8vxW4eLKsh48lOqap0snnkGPxuV3DlflL8HtrwEW/7hGidULzQehlwMQ6ZCvx9DgLVzvl9ERDrEqfz9VgBqhgJQ97E4czdPfrCLhAgb799+DpHBgZ33ZY462PMhfPcq7HrPdfdYvcAQ6DsRBpwL/c+F+KGuhVlFRKTbUABqJwWg7qOq1sGURf/lQEEFl52ZwpNXje6aL66rgQP/he1vw853oCyn6fthCa5A1Pds1xY31DXgWkREvEYBqJ0UgLqXTQcLuXLpepwGLL1mDFOGJ3ZtAU4n5H4P+z6GvR/Dwc+hrrLpMUFR0Ocs6D0Weo93LcthC+vaOkVE/JwCUDspAHU/j67ZwbOf7KVXqJX3bv8xsWE27xVTWwWHv4Ss9a4wdGgj1P5glmmTGeJPh5QzXWEo5UxXK5ElwDs1i4j4AQWgdlIA6n6q6xxMe2YdO7JLuXBYAs/9Ygym7jIGx1EL2d+6glD9Zm9mFuuAYEga5dqSR7seYwcrFImIdBAFoHZSAOqevj9qZ9qStdQ6DJ64chSXu9cN65bsR12tREc2w5FNcHQL1JQef1xAEMQPg8QRDVvC6WAL7/KSRUR6OgWgdlIA6r6WfLyHx9/bSbgtgPdu/zHJUcHeLql1nE4o2O0KQse2wLFv4Ni3zYcigOg0SBjuCkcJp7u26H5qLRIROQEFoHZSAOq+6hxOrnxuPV9nFTOmbzT/uPEsrAE99O4rpxMK90HOVshutJUea/54iw3iTnOFovihrjFFcYMhqq/uQBMRQQGo3RSAurcD+eVMfWYtpVV1/OKsvjw4fbi3S+pY5QWQuw1ytkHOd67H3B3H33lWLyAYYgdB3BDXEh9xp7nGFsX01+SNIuJXelwAWrJkCY8//jjZ2dmMGjWKp59+mvHjx5/0c6+88gozZ85k2rRpvPnmm5791113HStXrmxy7OTJk1mzZk2r6lEA6v4+2pHDDSu/wjDgsStGctXYVG+X1LmcTig+ALnbXbfk526HvJ2QvwscLaxfZrJAdF/oNQh6DYTYgRAzAHoNgPBktRqJiM/pUQFo1apVzJ49m6VLl5Kens6iRYv417/+xc6dO4mPj2/xcwcOHGDSpEn079+fmJiY4wJQTk4Of/vb3zz7bDYb0dHRrapJAahnqJ8l2hpg5l//M4FRqVHeLqnrOeqg+KArEOXvhPzd7mC0u+XxReAafB3dzxWGotPcWz+I6QeRqWo5EpEeqUcFoPT0dMaNG8czzzwDgNPpJDU1ld/85jfcfffdzX7G4XDw4x//mOuvv57PPvuM4uLi4wLQD/edCgWgnsHpNPiflzbxwfc5JEUG8Z/fTPLu/EDdiWG4xhIV7HGFoYK9rkHYBXtdgclZ1/JnTWZXC1FUH1cLUlQf1xaZClGpENFbAUlEuqVT+fvt1VtKampq2LRpE/Pnz/fsM5vNZGRksH79+hY/98ADDxAfH88NN9zAZ5991uwxn3zyCfHx8URHR3Peeefx0EMP0atXr2aPra6uprq62vPabrc3e5x0L2aziSevGsW0JevYl1fO3Jc38+IN47EFWLxdmveZTBCR7Nr6/bjpe446KMlyDcAu2AdFB6Bov+uxcL9rrJH9sGvL+ry5k0N4IkT2btgiekNkCkS4t9A4dbGJSLfm1QCUn5+Pw+EgISGhyf6EhAR27NjR7GfWrl3LCy+8wJYtW1o875QpU7jsssvo168fe/fu5Z577uGiiy5i/fr1WCzH/3FcuHAh999/f7t+i3hHeFAgz/9iLNOXrGPD/kJuX7WFp2eeicXcTSZJ7I4sAa4B0jH9YeAP3jMMKM+DooOulqKiA1CcBSWHoPiQ67GuytW6VHrMNddRc8yBEJHkDkTJEF7/3P0YnuQKUZZOXNxWROQEetSkIqWlpfziF79g2bJlxMbGtnjc1Vdf7Xk+YsQIRo4cyYABA/jkk084//zzjzt+/vz5zJs3z/PabreTmurjg2p9yMD4MJZeM4brV3zJO1uzCbN9yyOXjcSsEHTqTCYIi3dtqeOOf98woDzf1YJUcgRKDru3Q67JH+1HoDQbnLWu4FScdaIvc7UURSS5A1FSQzCKSHY9hidBcIxak0Skw3k1AMXGxmKxWMjJabrSdk5ODomJxy94uXfvXg4cOMDUqVM9+5xOJwABAQHs3LmTAQMGHPe5/v37Exsby549e5oNQDabDZtNY0d6skmDYlk88wx+/fIm/vnVYcJsgdz706HdZ7kMX2EyQVica0sZ0/wxjlpXCLIfcYeio43C0TGwu1uPnLVQnuvajn3T8neaA91hKLEhFIUluMOS+zEsEUJiXPWJiLSCVwOQ1WplzJgxZGZmMn36dMAVaDIzM7nllluOO37IkCFs3bq1yb4//vGPlJaW8tRTT7XYanP48GEKCgpISkrq8N8g3ceU4Yk8dsUo7vzXNyxft5/I4EBuzRjk7bL8jyXQNVg66gStqE4nVBRA6VF3WHI/1r8uPeZ6LM9zBaUSd/fbiZgD3cEowRWIWnoMjdOM2iLi/S6wefPmce211zJ27FjGjx/PokWLKC8vZ86cOQDMnj2blJQUFi5cSFBQEMOHN530LioqCsCzv6ysjPvvv5/LL7+cxMRE9u7dy1133cXAgQOZPHlyl/426XpXjOlNWVUtf/rP9/z1w10EBZr5n3OObxUULzObG1qSkka1fFxdDZTluLb6UFR6DEpzoCzb/TobKvJdQal+8PYJubvewhNcgSks0dXlF+5+rH8dlgC2sA792SLSfXg9AM2YMYO8vDwWLFhAdnY2o0ePZs2aNZ6B0VlZWZhPof/fYrHw7bffsnLlSoqLi0lOTubCCy/kwQcfVDeXn7huYj9Kq+p44oNdLHx3BwXlNdw9ZYjGBPVEAdaTtyaBKyiV57qCUekxdzjKOf6xPBcMZ0PXG1tPfN7A0IYw5HlMcIe3BAiNbxgzFaD/voj0JF6fB6g70jxAPZ9hGDz/330sfNd1N+GlZ6Tw2BUjCbRoMK1fczpcg7jLsqEs19V6VN/CVJbTEJJKc6C2/NTOHRTZNBCFxrtamsLiGvaHxrk2a0jn/D4RP9dj5gES6Swmk4n/OWcAsWE27nrtW974+ggF5TU8O+tMQm36x95vmS3ugdMJJz+2uqxROMp1b+7WpLI8V1Aqy3Ptc9ZCVYlrK9h98nNbwxrCUFg8hMY2vG7yPA6Co111i0iHUgtQM9QC5Fs+3pnLr1/aTGWtg1G9I3nuF2NJjAzydlniKwwDKotcA7bL3F1r9YGpPNfd4pTb8L6j+uTnbMxkdk0F4AlHsRBS/9jrB69jXXfDKTCJn+pRS2F0RwpAvufrrCKuX/ElRRW1xIZZWTzzDM4e0PJcUiKdwjCgutQVhjyBKc8VksrdrUrlBQ3vVxa24UtMrhAU4g5IITENYcmzz72//rk1VFMIiE9QAGonBSDfdLCgnF+9tJntx+yYTXDHhYO5+ZwBGhwt3ZejznWHW3l+w6PnuTs4VRQ0PK8qbtv3WGzNB6P658ExEBLd6HmMqxtPoUm6GQWgdlIA8l1VtQ7uffM7/rXJdav0+UPiefKq0USGaEkG8QGOOlerUX1IqihoFJLyXe9VFLhamerfd9S07bvMga4gVB+IgqNd23H7Yhrtj9bdctKpFIDaSQHI9636Mot7/72NmjoniRFBPDh9OBcMa8XAWBFfYhhQU+4KQhUFUFHYEJIqChv2Vxa6X7v3neo4psYCQxpCUXBUQ3A6bmv0XlCUuumkVRSA2kkByD98d6SE3/zja/bnu253vmRkEn+aejpx4fo/VJEWGQbUVjQKS+7HyiKoKGq0r6jR/kJX95zhbPv3mgMbQlFQlOt5k8foZvZFuaYnCAxRePITCkDtpADkP6pqHSz6cDfLPtuHw2kQGRzIHy8ZyuVn9tbYIJGO5HRCtd0djOrDUXGj142Ckme/+xhnbfu+uz48BUW6glFQZKPXjfa19DrA2r7vly6jANROCkD+57sjJfz+tW/ZdtQOwMjekdx90RDdKSbibfUtTpVFDcGocUiqf97So+Fofw0BwY0CUYTr0RbRdJ8twh2cIhq9535uDXMt/yKdTgGonRSA/FOdw8n/rt3P05m7Ka9x/Ufz3MFx/P6iIQxJ1D8HIj1O/RgnTyhyT1bZ3OuqEte+anvD/mp7BxVicoeiiGYew5vus7n3ed4Lb9ivlqiTUgBqJwUg/5ZXWs3TH+3m7xuyqHMamEwwfXQKN/9kAKclhHu7PBHpKo46qCltFJTqN3tDQGocmjzhyd7w3FnXcfVYbI1CUXhDWLKFHb/fGtaw39roGKv70Ucny1QAaicFIAHYn1/O4+/t4J2t2Z59GUMTuPknAxjTN9qLlYlIj2AYUFvpDkN21ySY1Y0CUnVpo+eNj2n8Xumpr0vXGgHBrlBkbRSe6kOTNbQhNFlDG46pf25172/8vqV7TCWiANROCkDS2LeHi3n2k72s2ZZN/b8t4/vFMOfsNDKGJWiBVRHpXI46qClrFI4aP2/0uqasITw13ldld7VkVZe1f0B5SyzWRgEptGELrH8e0vBeoPt58hmQOq5Dy1AAaicFIGnO3rwynv90H69/fZhah+tfm/hwG1eP78PM8akkRQZ7uUIRkZOoq3YFofpAVFN2/GvPvrLm99VUuB/L2zcn1MTb4IL7O+yngQJQuykAyYlkl1Txf18cYNWXh8gvc82iazbBeUPiufSM3pw/NJ6gQN/sXxcRacJR6wpCNeVNg1JtxQ/2u0NT4/1Dp8KIKzq0HAWgdlIAktaoqXPy3rZsXvriIBv2NyxaGW4L4OIRSVx6Zgrj02I0n5CISBdRAGonBSA5VXtyS3l98xHe/PoIR0uqPPvjwm1MPj2BKacnkd4/RuOFREQ6kQJQOykASVs5nQYbDxTyxuYjvPPdMUqrGm6BjQoJ5PwhCZw3JJ5Jg2KJDO4ed02IiPgKBaB2UgCSjlBT5+Tzvfms+S6b97/PobC8YdVti9nE2L7RnDsknh8NimVoYoS6ykRE2kkBqJ0UgKSj1TmcfHmgiI925PDRjlz25jWd1yMm1MqEAb2YNDCWiQNiSY0JxqTFG0VETokCUDspAElnyyqo4JNduXy8I5cN+wupqGm6XlFSZBDj+8Uwvl8M6f1iGBAXpkAkInISCkDtpAAkXammzsk3h4tZuzufz/fms+VQsWeeoXrRIYGc2SeaM/tGM6ZvNKN6RxFs1a32IiKNKQC1kwKQeFNljYOvs4rYsL+QjfsL2ZxVRHWds8kxFrOJwQnhjEqN4ozUKEalRjEwPgyLxhGJiB9TAGonBSDpTmrqnGw7WsLmrGI2Hyziq4OF5NiPn301ONDCsOQIRqREMjwlkhEpkQyICyVAt96LiJ9QAGonBSDpzgzD4FhJFd8eLmbLoRK+OVTMt4eLKf/BOCIAW4CZIYnhDEuOZFhyBMOSIhiSGE6oLcALlYuIdC4FoHZSAJKexuE02J9fxtYjJWw9bOe7IyV8d7TkuMHV9fr2CmFIYjhDEl2B6LTEcPrGhKi1SER6NAWgdlIAEl/gdBocLKxg29ESvj9qZ9tROzuy7c12nwFYA8wMjAtjcGI4gxLCGBQfzqD4MFJjQjS2SER6BAWgdlIAEl9WWF7DjmN2tmeXsuOYnV05pezKKaOytvnWIluAmf5xYQyIC2VgfBgD4lxbv9hQ3YkmIt2KAlA7KQCJv3E6DQ4VVbAju5Q9uWXsdoeivXllx92B1lhKVDD940LpHxtKv9hQ+sWF0T82lOSoYLUaiUiXUwBqJwUgEReH0+BQYQV788rYk1vW6LGcksraFj9ntZhJjQmmX2wofXuFktYrhL69QunbK4TkqGAtCisinUIBqJ0UgEROzDAMCstr2Jdfzr68MvbllbM/37UdLKyg5gStRhaziZSoYPrEhNCnV4jrMSaE1GjXY2SIFokVkbY5lb/fuhdWRE6ZyWSiV5iNXmE2xqXFNHnP4TQ4WlzJgYJyDhRUkOV+PJBfTlZhBdV1TrIKK8gqrIA9x587PCiA1OgQekcHkxoTQmp0ML2jQ+gdE0xKVDDhQQpIItJ+agFqhlqARDqH02mQW1rNwQJXS1FWQQWHiio4VFhBVmEl+WXN36HWWGRwIClRwaREuwJRSlQwyVHBJEcFkRIVTGyYDbPGH4n4JbUAiUi3ZDabSIwMIjEyiPT+vY57v6KmjiNFle5QVMlh9+ORYtfzoopaSipd2/fH7M1+R6DFRFJkMEmRrkCUFBVEUqQrICVFBpMcGUxEcIAWlxXxcwpAItJthFgDGJQQzqCE8GbfL6+u43BRJUeLKzlcXMmRIlc4OlJUwbGSKnLsVdQ6jIYutha/x0JiZBDJkcGuQBbhCmVJ7nCWGBFETKhVIUnEhykAiUiPEWoLYHBiOIMTmw9ItQ4nOfYqjpVUcbS4kqPFrsdjJa7nx0oqKaqopaLGwb68cvbllbf4XdYAMwkRNhIjgkiIaAhJCRENISku3EZQoOZCEumJFIBExGcEWsyuAdPRIS0eU1Xr4FiJKwwdK64i215Fdv1rdytSflkNNXVODhVWcqiw8oTfGR0SSEJEEPERQSSE20iICCIhMoj4+ucRNmLDbLr1X6SbUQASEb8SFGhxTdoYG9riMdV1DnLt1eTYGwKS63k1OSWufTn2KqrrnBRV1FJUUcuO7NIWz2cyQa9QK3HhrkAUH24jPjyIePfzuHBXYIqPsGELUIuSSFdQABIR+QFbgMV1C35Myy1JhmFgr6xzBSR7Fbn2KnJL3aGpxPW8fl+d0yC/rIb8shq2Hzvxd0cGB3rCULw7GMW5t/hwV7dbfISNcJsGcou0hwKQiEgbmEwmIkMCiQwJbHFMErhu/S+qqCHHXk1uaRW57kfP69Jqcu3V5JVWU+Nweu5y251bdsLvtwWY3aHo+IAUF9awLzbMhjVA3W8iP6QAJCLSiczmhkkjh9HyvCSGYVBcUUteWbUnJHnCUVk1ee7XefZqSqvrqK5zcrioksNFJx6jBBAVEtgkFMWF2YhtJijFhFq1hpv4DQUgEZFuwGQyER1qJTrUymktTANQr7LGQX6ZKyTllVaTW1pNvvvR87rM9bzO6QpWxRUnb1Uym6BXmCsMuUKRtUmLUmyjx6jgQE04KT2aApCISA8TbD35GCVwdb+VVNa6W5AaQtEPQ1JeaTWFFTU4DTyvTzZWKcBsopc7IMWGNWye4NQoLEUqLEk3pAAkIuKjzObWtyrVOZwUlNc0CUr5ZT987eqOK66opc5pkGOvJsd+8uVLWgpLjVuYYtWyJF1MAUhERAiwmN3zFgWd9NiaOieFPwhLjVuZXJvr/ZLKUwtLFrOJXqFWV0Bq1Jrkem1tEqA0ZknaQwFIREROiTXA7FnT7WRq6pwUlDcKR6U1zYalfHfLksO9YG5uaTWcpBvObIKYUGuTFqXYRq1J9a/jwl1hSZNRSmPdIgAtWbKExx9/nOzsbEaNGsXTTz/N+PHjT/q5V155hZkzZzJt2jTefPNNz37DMLjvvvtYtmwZxcXFTJw4kWeffZZBgwZ14q8QEZEfsgaY3YvTBp/02PqWpfqutvzShnBUv9V3zRW5xyzVz68ELU9EWS86JJDYMBu9wqzHjVnyhCj3a01I6fu8HoBWrVrFvHnzWLp0Kenp6SxatIjJkyezc+dO4uPjW/zcgQMHuPPOO/nRj3503HuPPfYYixcvZuXKlfTr1497772XyZMn8/333xMUdPL/YxERka53Ki1LdQ53N5y7BamgcWuSu0uuPjwVltfgcBqeWbt35568lvCggCZdb71Cj++Gi3OHqVCb1/+UShuYDMMwvFlAeno648aN45lnngHA6XSSmprKb37zG+6+++5mP+NwOPjxj3/M9ddfz2effUZxcbGnBcgwDJKTk7njjju48847ASgpKSEhIYEVK1Zw9dVXH3e+6upqqqsb+qbtdjupqamUlJQQEdHyvB0iItL91U9G2bg1Ka+0moLyxmHJ1T1XUF5NrePU/iwGB1qOG5/UpFUpzOrplosI0gzenclutxMZGdmqv99eja01NTVs2rSJ+fPne/aZzWYyMjJYv359i5974IEHiI+P54YbbuCzzz5r8t7+/fvJzs4mIyPDsy8yMpL09HTWr1/fbABauHAh999/fwf8IhER6W4aT0Y5mBPfDVe/xEleo263/PqwVFZNXmnTLrmqWieVtY5WLZwLYLWYG3XBNYxZcq0V1zRARYdYdUdcJ/JqAMrPz8fhcJCQkNBkf0JCAjt27Gj2M2vXruWFF15gy5Ytzb6fnZ3tOccPz1n/3g/Nnz+fefPmeV7XtwCJiIh/abzEycD4sBMeaxgG5TUOT/db43BU0GTskqulqbS6jhqHk2MlVRwrqTppLRaziegQq2e6gNgwV1D64SDv+nFNGuR9anpUx2VpaSm/+MUvWLZsGbGxsR12XpvNhs1m67DziYiI7zOZTITZAgizBdC3V+hJj6+qdXi63X44XcAP9xe574irf70j++SDvKNCAptMIRDXKDDVP9aPawq2apC3VwNQbGwsFouFnJycJvtzcnJITEw87vi9e/dy4MABpk6d6tnndDoBCAgIYOfOnZ7P5eTkkJSU1OSco0eP7oRfISIicnJBgRZSooJJiTr5HXG1joY74pob5J3fKDAVuAd51y95sjev/KTnD7Va3MueWN0tSDbiwurDkmt/L/dA74hg3xy35NUAZLVaGTNmDJmZmUyfPh1wBZrMzExuueWW444fMmQIW7dubbLvj3/8I6WlpTz11FOkpqYSGBhIYmIimZmZnsBjt9vZsGEDN998c2f/JBERkXYLPIWJKZ1Og+LK2uO63ArKXQO788uaBqbqOiflNQ7KCyvIKqxoRS0mVyj64d1woQ13xdW/HxNiJaCHdMV5vQts3rx5XHvttYwdO5bx48ezaNEiysvLmTNnDgCzZ88mJSWFhQsXEhQUxPDhw5t8PioqCqDJ/ttuu42HHnqIQYMGeW6DT05O9oQsERERX2E2m4gJtRLTiiVPDMOgrLru+Fal48YuNYxbqnUYZNuryLaffNySyYRn3JIrFDUe4H38vqBA73XFeT0AzZgxg7y8PBYsWEB2djajR49mzZo1nkHMWVlZmM2nlibvuusuysvLuemmmyguLmbSpEmsWbNGcwCJiIhfM5lMhAcFEh4USL/Yto9bKvhBaCood3XFGQYUltdQWF4DlJ3w3HMmpnHf1NM76JedOq/PA9Qdnco8AiIiIoJ7ssmaJi1J9TN3F5Q1TCVQUOaawPLW8wcx99yBHVpDj5kHSERERHyDxWzy3JZ/MoZh4HB6t/1FAUhERES6lMlkIsDi3TvLesZQbREREZEOpAAkIiIifkcBSERERPyOApCIiIj4HQUgERER8TsKQCIiIuJ3FIBERETE7ygAiYiIiN9RABIRERG/owAkIiIifkcBSERERPyOApCIiIj4HQUgERER8TtaDb4ZhmEAYLfbvVyJiIiItFb93+36v+MnogDUjNLSUgBSU1O9XImIiIicqtLSUiIjI094jMloTUzyM06nk6NHjxIeHo7JZOrQc9vtdlJTUzl06BAREREdem5pSte66+hadx1d666ja911OupaG4ZBaWkpycnJmM0nHuWjFqBmmM1mevfu3anfERERoX+huoiuddfRte46utZdR9e663TEtT5Zy089DYIWERERv6MAJCIiIn5HAaiL2Ww27rvvPmw2m7dL8Xm61l1H17rr6Fp3HV3rruONa61B0CIiIuJ31AIkIiIifkcBSERERPyOApCIiIj4HQUgERER8TsKQF1oyZIlpKWlERQURHp6Ohs3bvR2ST3ewoULGTduHOHh4cTHxzN9+nR27tzZ5Jiqqirmzp1Lr169CAsL4/LLLycnJ8dLFfuORx55BJPJxG233ebZp2vdcY4cOcI111xDr169CA4OZsSIEXz11Vee9w3DYMGCBSQlJREcHExGRga7d+/2YsU9k8Ph4N5776Vfv34EBwczYMAAHnzwwSZrSelat81///tfpk6dSnJyMiaTiTfffLPJ+625roWFhcyaNYuIiAiioqK44YYbKCsr65D6FIC6yKpVq5g3bx733XcfmzdvZtSoUUyePJnc3Fxvl9ajffrpp8ydO5cvvviCDz74gNraWi688ELKy8s9x9x+++385z//4V//+heffvopR48e5bLLLvNi1T3fl19+yXPPPcfIkSOb7Ne17hhFRUVMnDiRwMBA3n33Xb7//nueeOIJoqOjPcc89thjLF68mKVLl7JhwwZCQ0OZPHkyVVVVXqy853n00Ud59tlneeaZZ9i+fTuPPvoojz32GE8//bTnGF3rtikvL2fUqFEsWbKk2fdbc11nzZrFtm3b+OCDD3j77bf573//y0033dQxBRrSJcaPH2/MnTvX89rhcBjJycnGwoULvViV78nNzTUA49NPPzUMwzCKi4uNwMBA41//+pfnmO3btxuAsX79em+V2aOVlpYagwYNMj744APjnHPOMW699VbDMHStO9Lvf/97Y9KkSS2+73Q6jcTEROPxxx/37CsuLjZsNpvxj3/8oytK9BmXXHKJcf311zfZd9lllxmzZs0yDEPXuqMAxhtvvOF53Zrr+v333xuA8eWXX3qOeffddw2TyWQcOXKk3TWpBagL1NTUsGnTJjIyMjz7zGYzGRkZrF+/3ouV+Z6SkhIAYmJiANi0aRO1tbVNrv2QIUPo06ePrn0bzZ07l0suuaTJNQVd64701ltvMXbsWK688kri4+M544wzWLZsmef9/fv3k52d3eRaR0ZGkp6ermt9is4++2wyMzPZtWsXAN988w1r167loosuAnStO0trruv69euJiopi7NixnmMyMjIwm81s2LCh3TVoMdQukJ+fj8PhICEhocn+hIQEduzY4aWqfI/T6eS2225j4sSJDB8+HIDs7GysVitRUVFNjk1ISCA7O9sLVfZsr7zyCps3b+bLL7887j1d646zb98+nn32WebNm8c999zDl19+yW9/+1usVivXXnut53o2998UXetTc/fdd2O32xkyZAgWiwWHw8Gf//xnZs2aBaBr3Ulac12zs7OJj49v8n5AQAAxMTEdcu0VgMRnzJ07l++++461a9d6uxSfdOjQIW699VY++OADgoKCvF2OT3M6nYwdO5aHH34YgDPOOIPvvvuOpUuXcu2113q5Ot/yz3/+k5dffpm///3vnH766WzZsoXbbruN5ORkXWsfpy6wLhAbG4vFYjnubpicnBwSExO9VJVvueWWW3j77bf5+OOP6d27t2d/YmIiNTU1FBcXNzle1/7Ubdq0idzcXM4880wCAgIICAjg008/ZfHixQQEBJCQkKBr3UGSkpIYNmxYk31Dhw4lKysLwHM99d+U9vvd737H3XffzdVXX82IESP4xS9+we23387ChQsBXevO0prrmpiYeNyNQnV1dRQWFnbItVcA6gJWq5UxY8aQmZnp2ed0OsnMzGTChAlerKznMwyDW265hTfeeIOPPvqIfv36NXl/zJgxBAYGNrn2O3fuJCsrS9f+FJ1//vls3bqVLVu2eLaxY8cya9Ysz3Nd644xceLE46Zz2LVrF3379gWgX79+JCYmNrnWdrudDRs26FqfooqKCszmpn8KLRYLTqcT0LXuLK25rhMmTKC4uJhNmzZ5jvnoo49wOp2kp6e3v4h2D6OWVnnllVcMm81mrFixwvj++++Nm266yYiKijKys7O9XVqPdvPNNxuRkZHGJ598Yhw7dsyzVVRUeI751a9+ZfTp08f46KOPjK+++sqYMGGCMWHCBC9W7Tsa3wVmGLrWHWXjxo1GQECA8ec//9nYvXu38fLLLxshISHGSy+95DnmkUceMaKioox///vfxrfffmtMmzbN6Nevn1FZWenFynuea6+91khJSTHefvttY//+/cbrr79uxMbGGnfddZfnGF3rtiktLTW+/vpr4+uvvzYA48knnzS+/vpr4+DBg4ZhtO66TpkyxTjjjDOMDRs2GGvXrjUGDRpkzJw5s0PqUwDqQk8//bTRp08fw2q1GuPHjze++OILb5fU4wHNbn/72988x1RWVhq//vWvjejoaCMkJMS49NJLjWPHjnmvaB/ywwCka91x/vOf/xjDhw83bDabMWTIEOP5559v8r7T6TTuvfdeIyEhwbDZbMb5559v7Ny500vV9lx2u9249dZbjT59+hhBQUFG//79jT/84Q9GdXW15xhd67b5+OOPm/3v87XXXmsYRuuua0FBgTFz5kwjLCzMiIiIMObMmWOUlpZ2SH0mw2g03aWIiIiIH9AYIBEREfE7CkAiIiLidxSARERExO8oAImIiIjfUQASERERv6MAJCIiIn5HAUhERET8jgKQiIiI+B0FIBGRVvjkk08wmUzHLfYqIj2TApCIiIj4HQUgERER8TsKQCLSIzidThYuXEi/fv0IDg5m1KhRvPrqq0BD99Tq1asZOXIkQUFBnHXWWXz33XdNzvHaa69x+umnY7PZSEtL44knnmjyfnV1Nb///e9JTU3FZrMxcOBAXnjhhSbHbNq0ibFjxxISEsLZZ5/Nzp07O/eHi0inUAASkR5h4cKFvPjiiyxdupRt27Zx++23c8011/Dpp596jvnd737HE088wZdffklcXBxTp06ltrYWcAWXq666iquvvpqtW7fypz/9iXvvvZcVK1Z4Pj979mz+8Y9/sHjxYrZv385zzz1HWFhYkzr+8Ic/8MQTT/DVV18REBDA9ddf3yW/X0Q6llaDF5Fur7q6mpiYGD788EMmTJjg2f/LX/6SiooKbrrpJs4991xeeeUVZsyYAUBhYSG9e/dmxYoVXHXVVcyaNYu8vDzef/99z+fvuusuVq9ezbZt29i1axeDBw/mgw8+ICMj47gaPvnkE84991w+/PBDzj//fADeeecdLrnkEiorKwkKCurkqyAiHUktQCLS7e3Zs4eKigouuOACwsLCPNuLL77I3r17Pcc1DkcxMTEMHjyY7du3A7B9+3YmTpzY5LwTJ05k9+7dOBwOtmzZgsVi4ZxzzjlhLSNHjvQ8T0pKAiA3N7fdv1FEulaAtwsQETmZsrIyAFavXk1KSkqT92w2W5MQ1FbBwcGtOi4wMNDz3GQyAa7xSSLSs6gFSES6vWHDhmGz2cjKymLgwIFNttTUVM9xX3zxhed5UVERu3btYujQoQAMHTqUdevWNTnvunXrOO2007BYLIwYMQKn09lkTJGI+C61AIlItxceHs6dd97J7bffjtPpZNKkSZSUlLBu3ToiIiLo27cvAA888AC9evUiISGBP/zhD8TGxjJ9+nQA7rjjDsaNG8eDDz7IjBkzWL9+Pc888wz/7//9PwDS0tK49tpruf7661m8eDGjRo3i4MGD5ObmctVVV3nrp4tIJ1EAEpEe4cEHHyQuLo6FCxeyb98+oqKiOPPMM7nnnns8XVCPPPIIt956K7t372b06NH85z//wWq1AnDmmWfyz3/+kwULFvDggw+SlJTEAw88wHXXXef5jmeffZZ77rmHX//61xQUFNCnTx/uueceb/xcEelkugtMRHq8+ju0ioqKiIqK8nY5ItIDaAyQiIiI+B0FIBEREfE76gITERERv6MWIBEREfE7CkAiIiLidxSARERExO8oAImIiIjfUQASERERv6MAJCIiIn5HAUhERET8jgKQiIiI+J3/D9bNStpNqnPVAAAAAElFTkSuQmCC\n" }, "metadata": {} } ], "source": [ "plt.plot(history.history['loss'])\n", "plt.plot(history.history['val_loss'])\n", "plt.xlabel('epoch')\n", "plt.ylabel('loss')\n", "plt.legend(['train', 'val'])\n", "plt.show()" ] }, { "cell_type": "markdown", "metadata": { "id": "Ja3-116b6k2V" }, "source": [ "## 마무리" ] }, { "cell_type": "code", "execution_count": 16, "metadata": { "colab": { "base_uri": "https://localhost:8080/" }, "id": "-5WtFRN-6k2V", "outputId": "8c31a7f9-276a-49d7-fe75-9d78efcafc87", "tags": [] }, "outputs": [ { "output_type": "stream", "name": "stdout", "text": [ "\u001b[1m782/782\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m5s\u001b[0m 5ms/step - accuracy: 0.7866 - loss: 0.4654\n" ] }, { "output_type": "execute_result", "data": { "text/plain": [ "[0.4673371911048889, 0.7847599983215332]" ] }, "metadata": {}, "execution_count": 16 } ], "source": [ "test_seq = pad_sequences(test_input, maxlen=100)\n", "\n", "rnn_model = keras.models.load_model('best-2rnn-model.keras')\n", "\n", "rnn_model.evaluate(test_seq, test_target)" ] } ], "metadata": { "accelerator": "GPU", "colab": { "name": "9-3.ipynb", "provenance": [] }, "kernelspec": { "display_name": "default:Python", "language": "python", "name": "conda-env-default-py" }, "language_info": { "codemirror_mode": { "name": "ipython", "version": 3 }, "file_extension": ".py", "mimetype": "text/x-python", "name": "python", "nbconvert_exporter": "python", "pygments_lexer": "ipython3", "version": "3.9.10" } }, "nbformat": 4, "nbformat_minor": 0 }