modeltimesteps.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
    "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['fakenews-glove', 'ta-rumour-glove']\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "\n",
    "import os\n",
    "import time\n",
    "import datetime\n",
    "import gc\n",
    "import random\n",
    "import re\n",
    "import operator\n",
    "\n",
    "from tqdm import tqdm\n",
    "\n",
    "from sklearn.model_selection import StratifiedKFold,KFold\n",
    "from sklearn.metrics import f1_score,precision_score,recall_score\n",
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "\n",
    "\n",
    "from keras.layers import *\n",
    "from keras.models import Model\n",
    "\n",
    "from keras.preprocessing.text import Tokenizer,text_to_word_sequence\n",
    "from keras.preprocessing.sequence import pad_sequences\n",
    "print(os.listdir(\"../input/\"))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "_cell_guid": "79c7e3d0-c299-4dcb-8224-4455121ee9b0",
    "_uuid": "d629ff2d2480ee46fbb7e2d37f6b5fab8052498a"
   },
   "outputs": [],
   "source": [
    "rumour=np.load('../input/ta-rumour-glove/rumour.npy')\n",
    "fakenews=np.load('../input/fakenews-glove/fakenews.npy')\n",
    "# rumour"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(5331, 2)\n",
      "non-rumours\n",
      "(13, 300)\n"
     ]
    }
   ],
   "source": [
    "print(rumour.shape)\n",
    "print(rumour[4][1])\n",
    "print(rumour[1][0].shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(30000, 2)"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "fakenews.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5331/5331 [00:00<00:00, 31910.44it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "X shape :  (5331, 100, 300)\n",
      "Y shape :  (5331,)\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "timesteps=100\n",
    "embed_dim=300\n",
    "X_rumour=np.zeros((rumour.shape[0],timesteps,embed_dim))\n",
    "y_rumour=[]\n",
    "\n",
    "for i in tqdm(range(0,rumour.shape[0])):\n",
    "    num_of_sentences=rumour[i][0].shape[0]\n",
    "    if num_of_sentences>=timesteps:\n",
    "        X_rumour[i]=rumour[i][0][:timesteps,:]\n",
    "    else:\n",
    "        X_rumour[i,-num_of_sentences:]=rumour[i][0]\n",
    "    y_rumour.append(rumour[i][1])\n",
    "y_rumour=np.asarray(y_rumour)\n",
    "print(\"X shape : \",X_rumour.shape)\n",
    "print(\"Y shape : \",y_rumour.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 30000/30000 [00:00<00:00, 45686.82it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "X shape :  (30000, 100, 300)\n",
      "Y shape :  (30000,)\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "timesteps=100\n",
    "embed_dim=300\n",
    "X_fakenews=np.zeros((fakenews.shape[0],timesteps,embed_dim))\n",
    "y_fakenews=[]\n",
    "\n",
    "for i in tqdm(range(0,fakenews.shape[0])):\n",
    "    num_of_sentences=fakenews[i][0].shape[0]\n",
    "    if num_of_sentences>=timesteps:\n",
    "        X_fakenews[i]=fakenews[i][0][:timesteps,:]\n",
    "    else:\n",
    "        X_fakenews[i,-num_of_sentences:]=fakenews[i][0]\n",
    "    y_fakenews.append(fakenews[i][1])\n",
    "y_fakenews=np.asarray(y_fakenews)\n",
    "print(\"X shape : \",X_fakenews.shape)\n",
    "print(\"Y shape : \",y_fakenews.shape)   \n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array(['agree', 'agree', 'agree', ..., 'unrelated', 'unrelated',\n",
       "       'unrelated'], dtype='<U9')"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_fakenews"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['non-rumours' 'non-rumours' 'non-rumours' ... 'rumours' 'rumours'\n",
      " 'rumours']\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7f9ac0f452e8>"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "print(y_rumour)\n",
    "sns.countplot(y_rumour)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0\n",
      "non-rumours\n"
     ]
    }
   ],
   "source": [
    "rumour_label_dict={\"non-rumours\":0,\"rumours\":1}\n",
    "label_rumour_dict={0:\"non-rumours\",1:\"rumours\"}\n",
    "print(rumour_label_dict[\"non-rumours\"])\n",
    "print(label_rumour_dict[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "label_rumour=[]\n",
    "for i in range(0,len(y_rumour)):\n",
    "    if y_rumour[i]==\"non-rumours\":\n",
    "        label_rumour.append(0)\n",
    "    elif y_rumour[i]==\"rumours\":\n",
    "        label_rumour.append(1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(5331,)\n",
      "[0. 0. 0. ... 1. 1. 1.]\n"
     ]
    }
   ],
   "source": [
    "Y_rumour=np.zeros(len(label_rumour))\n",
    "for i in range(0,len(label_rumour)):\n",
    "    if label_rumour[i] == 0:\n",
    "        Y_rumour[i]=0\n",
    "    elif label_rumour[i] == 1:\n",
    "        Y_rumour[i]=1\n",
    "print(Y_rumour.shape)\n",
    "print(Y_rumour)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['agree' 'agree' 'agree' ... 'unrelated' 'unrelated' 'unrelated']\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7f9ac0a7ca90>"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "print(y_fakenews)\n",
    "sns.countplot(y_fakenews)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0\n",
      "unrelated\n"
     ]
    }
   ],
   "source": [
    "fakenews_label_dict={\"unrelated\":0,\"agree\":1,\"disagree\":2,\"discuss\":3}\n",
    "label_fakenews_dict={0:\"unrelated\",1:\"agree\",2:\"disagree\",3:\"discuss\"}\n",
    "print(fakenews_label_dict[\"unrelated\"])\n",
    "print(label_fakenews_dict[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "label_fakenews=[]\n",
    "for i in range(0,len(y_fakenews)):\n",
    "    label_fakenews.append(fakenews_label_dict[y_fakenews[i]])\n",
    "# label_fakenews"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(30000, 4)\n",
      "[[0. 1. 0. 0.]\n",
      " [0. 1. 0. 0.]\n",
      " [0. 1. 0. 0.]\n",
      " ...\n",
      " [1. 0. 0. 0.]\n",
      " [1. 0. 0. 0.]\n",
      " [1. 0. 0. 0.]]\n"
     ]
    }
   ],
   "source": [
    "Y_fakenews=np.zeros((len(label_fakenews),4))\n",
    "for i in range(0,len(label_fakenews)):\n",
    "    if label_fakenews[i] == 0:\n",
    "        Y_fakenews[i][0]=1\n",
    "    elif label_fakenews[i] == 1:\n",
    "        Y_fakenews[i][1]=1\n",
    "    elif label_fakenews[i] == 2:\n",
    "        Y_fakenews[i][2]=1\n",
    "    elif label_fakenews[i] == 3:\n",
    "        Y_fakenews[i][3]=1\n",
    "print(Y_fakenews.shape)\n",
    "print(Y_fakenews)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras import backend as K\n",
    "\n",
    "def f1(y_true, y_pred):\n",
    "    def recall(y_true, y_pred):\n",
    "        \"\"\"Recall metric.\n",
    "\n",
    "        Only computes a batch-wise average of recall.\n",
    "\n",
    "        Computes the recall, a metric for multi-label classification of\n",
    "        how many relevant items are selected.\n",
    "        \"\"\"\n",
    "        true_positives = K.sum(K.round(K.clip(y_true * y_pred, 0, 1)))\n",
    "        possible_positives = K.sum(K.round(K.clip(y_true, 0, 1)))\n",
    "        recall = true_positives / (possible_positives + K.epsilon())\n",
    "        return recall\n",
    "\n",
    "    def precision(y_true, y_pred):\n",
    "        \"\"\"Precision metric.\n",
    "\n",
    "        Only computes a batch-wise average of precision.\n",
    "\n",
    "        Computes the precision, a metric for multi-label classification of\n",
    "        how many selected items are relevant.\n",
    "        \"\"\"\n",
    "        true_positives = K.sum(K.round(K.clip(y_true * y_pred, 0, 1)))\n",
    "        predicted_positives = K.sum(K.round(K.clip(y_pred, 0, 1)))\n",
    "        precision = true_positives / (predicted_positives + K.epsilon())\n",
    "        return precision\n",
    "    precision = precision(y_true, y_pred)\n",
    "    recall = recall(y_true, y_pred)\n",
    "    return 2*((precision*recall)/(precision+recall+K.epsilon()))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From /opt/conda/lib/python3.6/site-packages/tensorflow/python/framework/op_def_library.py:263: colocate_with (from tensorflow.python.framework.ops) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Colocations handled automatically by placer.\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "input_1 (InputLayer)         (None, 100, 300)          0         \n",
      "_________________________________________________________________\n",
      "gru_1 (GRU)                  (None, 64)                70080     \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 32)                2080      \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 8)                 264       \n",
      "_________________________________________________________________\n",
      "dense_3 (Dense)              (None, 1)                 9         \n",
      "=================================================================\n",
      "Total params: 72,433\n",
      "Trainable params: 72,433\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "input_1 (InputLayer)         (None, 100, 300)          0         \n",
      "_________________________________________________________________\n",
      "gru_1 (GRU)                  (None, 64)                70080     \n",
      "_________________________________________________________________\n",
      "dense_4 (Dense)              (None, 32)                2080      \n",
      "_________________________________________________________________\n",
      "dense_5 (Dense)              (None, 16)                528       \n",
      "_________________________________________________________________\n",
      "dense_6 (Dense)              (None, 4)                 68        \n",
      "=================================================================\n",
      "Total params: 72,756\n",
      "Trainable params: 72,756\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "hidden_units=64\n",
    "inp=Input(shape=(100,300))\n",
    "h=GRU(units=64)(inp)\n",
    "\n",
    "o1=Dense(32,activation='relu')(h)\n",
    "o1=Dense(8,activation='relu')(o1)\n",
    "o1=Dense(1,activation='sigmoid')(o1)\n",
    "\n",
    "o2=Dense(32,activation='relu')(h)\n",
    "o2=Dense(16,activation='relu')(o2)\n",
    "o2=Dense(4,activation='softmax')(o2)\n",
    "\n",
    "rumour_model=Model(inputs=[inp],outputs=[o1])\n",
    "rumour_model.summary()\n",
    "\n",
    "fake_model=Model(inputs=[inp],outputs=[o2])\n",
    "fake_model.summary()\n",
    "\n",
    "rumour_model.compile(optimizer=\"Adagrad\",loss=\"binary_crossentropy\",metrics=[f1])\n",
    "fake_model.compile(optimizer=\"Adagrad\",loss=\"categorical_crossentropy\",metrics=[f1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "ag = Y_fakenews[Y_fakenews[:,1] == 1].shape[0]\n",
    "dg = ag + Y_fakenews[Y_fakenews[:,2] == 1].shape[0]\n",
    "ds = dg + Y_fakenews[Y_fakenews[:,3] == 1].shape[0]\n",
    "ur = ds + Y_fakenews[Y_fakenews[:,0] == 1].shape[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "3678\n",
      "840\n",
      "8909\n",
      "16573\n"
     ]
    }
   ],
   "source": [
    "print(ag)\n",
    "print(dg-ag)\n",
    "print(ds-dg)\n",
    "print(ur-ds)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "# temp = np.zeros((10000,60,300))\n",
    "# temp_y = np.zeros((10000,4))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "# temp[0:1500] = X_fakenews[0:1500]\n",
    "# temp[1500:1500+840] = X_fakenews[ag:dg]\n",
    "# temp[2340:5340] = X_fakenews[dg:dg+3000]\n",
    "# temp[5340:10000] = X_fakenews[ds:ds+4660]\n",
    "\n",
    "# temp_y[0:1500] = Y_fakenews[0:1500]\n",
    "# temp_y[1500:1500+840] = Y_fakenews[ag:dg]\n",
    "# temp_y[2340:5340] = Y_fakenews[dg:dg+3000]\n",
    "# temp_y[5340:10000] = Y_fakenews[ds:ds+4660]\n",
    "\n",
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "x_rumour_train,x_rumour_val,y_rumour_train,y_rumour_val=train_test_split(X_rumour,Y_rumour,test_size=0.15,random_state=42)\n",
    "x_fakenews_train,x_fakenews_val,y_fakenews_train,y_fakenews_val=train_test_split(X_fakenews[0:20000],Y_fakenews[0:20000],test_size=0.15,random_state=42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(17000, 4)\n",
      "(5594, 4)\n",
      "(3153, 4)\n",
      "(702, 4)\n",
      "(7551, 4)\n",
      "(3000, 4)\n",
      "(979, 4)\n",
      "(525, 4)\n",
      "(138, 4)\n",
      "(1358, 4)\n"
     ]
    }
   ],
   "source": [
    "print(y_fakenews_train.shape)\n",
    "print(y_fakenews_train[y_fakenews_train[:,0] == 1].shape)\n",
    "print(y_fakenews_train[y_fakenews_train[:,1] == 1].shape)\n",
    "print(y_fakenews_train[y_fakenews_train[:,2] == 1].shape)\n",
    "print(y_fakenews_train[y_fakenews_train[:,3] == 1].shape)\n",
    "print(y_fakenews_val.shape)\n",
    "print(y_fakenews_val[y_fakenews_val[:,0] == 1].shape)\n",
    "print(y_fakenews_val[y_fakenews_val[:,1] == 1].shape)\n",
    "print(y_fakenews_val[y_fakenews_val[:,2] == 1].shape)\n",
    "print(y_fakenews_val[y_fakenews_val[:,3] == 1].shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(17000, 4)\n",
      "(4531,)\n"
     ]
    }
   ],
   "source": [
    "print(y_fakenews_train.shape)\n",
    "print(y_rumour_train.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "index_rumour = np.array([i for i in range(y_rumour_train.shape[0])])\n",
    "index_fakenews = np.array([i for i in range(y_fakenews_train.shape[0])])\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "def sample(size,flag):\n",
    "    if flag == 1:\n",
    "        index = list(np.random.choice(index_rumour,size))\n",
    "        x_train = np.array([x_rumour_train[i] for i in index])\n",
    "        y_train = np.array([y_rumour_train[i] for i in index])\n",
    "    else:\n",
    "        #Size = size/4\n",
    "        index = list(np.random.choice(index_fakenews,size))\n",
    "        x_train = np.array([x_fakenews_train[i] for i in index])\n",
    "        y_train = np.array([y_fakenews_train[i] for i in index])\n",
    "    return x_train,y_train\n",
    "        \n",
    "    \n",
    "    \n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  0%|          | 0/2000 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From /opt/conda/lib/python3.6/site-packages/tensorflow/python/ops/math_ops.py:3066: to_int32 (from tensorflow.python.ops.math_ops) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Use tf.cast instead.\n",
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 6s 23ms/step - loss: 0.6523 - f1: 0.0250 - val_loss: 0.6355 - val_f1: 0.0000e+00\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 17ms/step - loss: 0.5985 - f1: 0.0000e+00 - val_loss: 0.6346 - val_f1: 0.0000e+00\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  0%|          | 1/2000 [00:10<6:05:54, 10.98s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 7s 14ms/step - loss: 1.2134 - f1: 0.0683 - val_loss: 1.1994 - val_f1: 0.4334\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 7s 13ms/step - loss: 1.1356 - f1: 0.1137 - val_loss: 1.1669 - val_f1: 0.3260\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  0%|          | 2/2000 [00:25<6:40:54, 12.04s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 7s 13ms/step - loss: 1.1438 - f1: 0.2242 - val_loss: 1.1624 - val_f1: 0.2618\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 7s 13ms/step - loss: 1.1186 - f1: 0.2258 - val_loss: 1.2553 - val_f1: 0.4554\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  0%|          | 3/2000 [00:38<6:52:31, 12.39s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 17ms/step - loss: 0.6032 - f1: 0.0000e+00 - val_loss: 0.6274 - val_f1: 0.0000e+00\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 17ms/step - loss: 0.5767 - f1: 0.0000e+00 - val_loss: 0.6729 - val_f1: 0.0000e+00\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  0%|          | 4/2000 [00:47<6:14:36, 11.26s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 7s 14ms/step - loss: 1.1964 - f1: 0.2211 - val_loss: 1.1542 - val_f1: 0.1162\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 7s 13ms/step - loss: 1.1616 - f1: 0.2066 - val_loss: 1.1412 - val_f1: 0.2080\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  0%|          | 5/2000 [01:00<6:37:50, 11.97s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 13ms/step - loss: 1.1146 - f1: 0.2746 - val_loss: 1.1384 - val_f1: 0.3968\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 13ms/step - loss: 1.0859 - f1: 0.3383 - val_loss: 1.1609 - val_f1: 0.4062\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  0%|          | 6/2000 [01:13<6:46:55, 12.24s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 13ms/step - loss: 1.1348 - f1: 0.2930 - val_loss: 1.1388 - val_f1: 0.1321\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 13ms/step - loss: 1.1101 - f1: 0.2392 - val_loss: 1.1226 - val_f1: 0.2307\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  0%|          | 7/2000 [01:26<6:53:29, 12.45s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.6015 - f1: 0.0000e+00 - val_loss: 0.6103 - val_f1: 0.0000e+00\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.5534 - f1: 0.0000e+00 - val_loss: 0.6446 - val_f1: 0.0000e+00\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  0%|          | 8/2000 [01:35<6:12:52, 11.23s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.6064 - f1: 0.0000e+00 - val_loss: 0.5991 - val_f1: 0.0252\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.5684 - f1: 0.1869 - val_loss: 0.6021 - val_f1: 0.0749\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  0%|          | 9/2000 [01:43<5:43:31, 10.35s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.5904 - f1: 0.1200 - val_loss: 0.5785 - val_f1: 0.3205\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.5582 - f1: 0.4083 - val_loss: 0.6048 - val_f1: 0.2214\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  0%|          | 10/2000 [01:51<5:23:24,  9.75s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 1.1549 - f1: 0.2406 - val_loss: 1.1262 - val_f1: 0.2112\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 1.1445 - f1: 0.2305 - val_loss: 1.1184 - val_f1: 0.2958\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  1%|          | 11/2000 [02:04<5:52:31, 10.63s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 7s 13ms/step - loss: 1.1158 - f1: 0.3568 - val_loss: 1.1126 - val_f1: 0.3561\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 7s 13ms/step - loss: 1.1016 - f1: 0.3745 - val_loss: 1.1099 - val_f1: 0.3750\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  1%|          | 12/2000 [02:18<6:22:28, 11.54s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.5902 - f1: 0.2895 - val_loss: 0.5747 - val_f1: 0.3765\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.5528 - f1: 0.4389 - val_loss: 0.5781 - val_f1: 0.5029\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  1%|          | 13/2000 [02:26<5:49:53, 10.57s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 1.1396 - f1: 0.3007 - val_loss: 1.1177 - val_f1: 0.2259\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 7s 13ms/step - loss: 1.1272 - f1: 0.2635 - val_loss: 1.1061 - val_f1: 0.3203\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  1%|          | 14/2000 [02:39<6:14:05, 11.30s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.5725 - f1: 0.5134 - val_loss: 0.5562 - val_f1: 0.4547\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.5402 - f1: 0.5932 - val_loss: 0.5517 - val_f1: 0.5365\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  1%|          | 15/2000 [02:47<5:43:44, 10.39s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 13ms/step - loss: 1.1188 - f1: 0.3125 - val_loss: 1.1040 - val_f1: 0.3091\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 7s 14ms/step - loss: 1.1007 - f1: 0.3103 - val_loss: 1.1018 - val_f1: 0.3151\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  1%|          | 16/2000 [03:01<6:14:01, 11.31s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.5638 - f1: 0.4610 - val_loss: 0.5491 - val_f1: 0.4991\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.5437 - f1: 0.5451 - val_loss: 0.5472 - val_f1: 0.4727\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  1%|          | 17/2000 [03:09<5:43:16, 10.39s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.5332 - f1: 0.4848 - val_loss: 0.5480 - val_f1: 0.4910\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.5106 - f1: 0.5342 - val_loss: 0.5608 - val_f1: 0.4256\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  1%|          | 18/2000 [03:17<5:21:42,  9.74s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.5838 - f1: 0.4941 - val_loss: 0.5512 - val_f1: 0.4801\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.5557 - f1: 0.5601 - val_loss: 0.5445 - val_f1: 0.5049\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  1%|          | 19/2000 [03:25<5:07:03,  9.30s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 7s 14ms/step - loss: 1.0986 - f1: 0.3687 - val_loss: 1.0992 - val_f1: 0.3420\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 7s 13ms/step - loss: 1.0755 - f1: 0.3737 - val_loss: 1.1009 - val_f1: 0.4145\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  1%|          | 20/2000 [03:39<5:51:04, 10.64s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 1.0453 - f1: 0.4686 - val_loss: 1.0944 - val_f1: 0.3807\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 1.0261 - f1: 0.4674 - val_loss: 1.0971 - val_f1: 0.3077\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  1%|          | 21/2000 [03:52<6:11:17, 11.26s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.5199 - f1: 0.4090 - val_loss: 0.5504 - val_f1: 0.5311\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4919 - f1: 0.5211 - val_loss: 0.5542 - val_f1: 0.4940\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  1%|          | 22/2000 [04:00<5:41:28, 10.36s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 1.1185 - f1: 0.3024 - val_loss: 1.0882 - val_f1: 0.3655\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 13ms/step - loss: 1.0985 - f1: 0.3210 - val_loss: 1.0876 - val_f1: 0.3862\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  1%|          | 23/2000 [04:13<6:05:36, 11.10s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.5629 - f1: 0.6186 - val_loss: 0.5474 - val_f1: 0.5754\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 17ms/step - loss: 0.5203 - f1: 0.6831 - val_loss: 0.5459 - val_f1: 0.5962\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  1%|          | 24/2000 [04:21<5:39:36, 10.31s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 7s 14ms/step - loss: 1.1046 - f1: 0.3581 - val_loss: 1.0934 - val_f1: 0.2957\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 7s 13ms/step - loss: 1.0827 - f1: 0.3729 - val_loss: 1.0972 - val_f1: 0.2745\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  1%|▏         | 25/2000 [04:35<6:14:16, 11.37s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.5369 - f1: 0.5636 - val_loss: 0.5339 - val_f1: 0.5577\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4997 - f1: 0.5875 - val_loss: 0.5357 - val_f1: 0.5451\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  1%|▏         | 26/2000 [04:44<5:43:20, 10.44s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 7s 13ms/step - loss: 1.1301 - f1: 0.3513 - val_loss: 1.0882 - val_f1: 0.3173\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 7s 14ms/step - loss: 1.1184 - f1: 0.3499 - val_loss: 1.0780 - val_f1: 0.3797\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  1%|▏         | 27/2000 [04:57<6:15:39, 11.42s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.5469 - f1: 0.5460 - val_loss: 0.5312 - val_f1: 0.5595\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.5424 - f1: 0.6339 - val_loss: 0.5325 - val_f1: 0.5692\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  1%|▏         | 28/2000 [05:06<5:44:05, 10.47s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.5631 - f1: 0.5704 - val_loss: 0.5386 - val_f1: 0.5446\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.5420 - f1: 0.5217 - val_loss: 0.5400 - val_f1: 0.5011\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  1%|▏         | 29/2000 [05:14<5:22:14,  9.81s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.5581 - f1: 0.5419 - val_loss: 0.5567 - val_f1: 0.4408\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.5292 - f1: 0.5583 - val_loss: 0.5310 - val_f1: 0.5286\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  2%|▏         | 30/2000 [05:22<5:06:56,  9.35s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 1.0779 - f1: 0.3517 - val_loss: 1.0788 - val_f1: 0.3940\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 13ms/step - loss: 1.0651 - f1: 0.3889 - val_loss: 1.0838 - val_f1: 0.2957\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  2%|▏         | 31/2000 [05:35<5:40:30, 10.38s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 1.0853 - f1: 0.3459 - val_loss: 1.0762 - val_f1: 0.3454\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 1.0726 - f1: 0.3592 - val_loss: 1.0793 - val_f1: 0.3228\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  2%|▏         | 32/2000 [05:47<6:02:54, 11.06s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 1.0882 - f1: 0.3338 - val_loss: 1.0759 - val_f1: 0.4434\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 1.0720 - f1: 0.3923 - val_loss: 1.0747 - val_f1: 0.3272\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  2%|▏         | 33/2000 [06:00<6:19:09, 11.57s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 1.1067 - f1: 0.2976 - val_loss: 1.0684 - val_f1: 0.3365\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 7s 14ms/step - loss: 1.0909 - f1: 0.3045 - val_loss: 1.0677 - val_f1: 0.3330\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  2%|▏         | 34/2000 [06:14<6:36:06, 12.09s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 7s 13ms/step - loss: 1.0983 - f1: 0.3612 - val_loss: 1.0644 - val_f1: 0.3709\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 1.0835 - f1: 0.3861 - val_loss: 1.0638 - val_f1: 0.3616\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  2%|▏         | 35/2000 [06:27<6:44:57, 12.36s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4967 - f1: 0.5222 - val_loss: 0.5352 - val_f1: 0.5063\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4845 - f1: 0.5041 - val_loss: 0.5268 - val_f1: 0.5567\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  2%|▏         | 36/2000 [06:35<6:04:28, 11.13s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 1.1055 - f1: 0.3512 - val_loss: 1.0602 - val_f1: 0.4234\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 1.0890 - f1: 0.3818 - val_loss: 1.0544 - val_f1: 0.4053\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  2%|▏         | 37/2000 [06:47<6:19:23, 11.60s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.5076 - f1: 0.5088 - val_loss: 0.5277 - val_f1: 0.5462\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4895 - f1: 0.5839 - val_loss: 0.5227 - val_f1: 0.5847\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  2%|▏         | 38/2000 [06:56<5:46:04, 10.58s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4549 - f1: 0.5788 - val_loss: 0.5343 - val_f1: 0.5235\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4430 - f1: 0.6028 - val_loss: 0.5419 - val_f1: 0.5009\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  2%|▏         | 39/2000 [07:04<5:22:46,  9.88s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 1.0522 - f1: 0.3659 - val_loss: 1.0625 - val_f1: 0.3402\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 1.0361 - f1: 0.3752 - val_loss: 1.0543 - val_f1: 0.3918\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  2%|▏         | 40/2000 [07:17<5:49:33, 10.70s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4810 - f1: 0.6224 - val_loss: 0.5211 - val_f1: 0.5710\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4654 - f1: 0.6465 - val_loss: 0.5366 - val_f1: 0.5181\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  2%|▏         | 41/2000 [07:25<5:24:53,  9.95s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 7s 13ms/step - loss: 1.0695 - f1: 0.3921 - val_loss: 1.0550 - val_f1: 0.3738\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 7s 14ms/step - loss: 1.0533 - f1: 0.3696 - val_loss: 1.0509 - val_f1: 0.4263\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  2%|▏         | 42/2000 [07:38<6:01:17, 11.07s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 1.0829 - f1: 0.3849 - val_loss: 1.0468 - val_f1: 0.4168\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 1.0581 - f1: 0.4250 - val_loss: 1.0471 - val_f1: 0.3782\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  2%|▏         | 43/2000 [07:51<6:16:29, 11.54s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.5071 - f1: 0.5719 - val_loss: 0.5208 - val_f1: 0.5841\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4784 - f1: 0.6205 - val_loss: 0.5186 - val_f1: 0.5942\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  2%|▏         | 44/2000 [07:59<5:44:20, 10.56s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 5s 18ms/step - loss: 0.4665 - f1: 0.6462 - val_loss: 0.5225 - val_f1: 0.5786\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4340 - f1: 0.6804 - val_loss: 0.5309 - val_f1: 0.6048\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  2%|▏         | 45/2000 [08:08<5:26:21, 10.02s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4841 - f1: 0.6937 - val_loss: 0.5181 - val_f1: 0.5843\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4578 - f1: 0.7272 - val_loss: 0.5192 - val_f1: 0.5717\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  2%|▏         | 46/2000 [08:16<5:08:16,  9.47s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4995 - f1: 0.6161 - val_loss: 0.5192 - val_f1: 0.5643\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4743 - f1: 0.6312 - val_loss: 0.5320 - val_f1: 0.5363\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  2%|▏         | 47/2000 [08:24<4:55:44,  9.09s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4883 - f1: 0.5518 - val_loss: 0.5165 - val_f1: 0.5671\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4640 - f1: 0.5889 - val_loss: 0.5435 - val_f1: 0.4965\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  2%|▏         | 48/2000 [08:33<4:46:41,  8.81s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.5362 - f1: 0.5483 - val_loss: 0.5140 - val_f1: 0.5823\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.5090 - f1: 0.6544 - val_loss: 0.5160 - val_f1: 0.5895\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  2%|▏         | 49/2000 [08:41<4:40:54,  8.64s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 1.0846 - f1: 0.3696 - val_loss: 1.0511 - val_f1: 0.3600\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 7s 14ms/step - loss: 1.0610 - f1: 0.3805 - val_loss: 1.0460 - val_f1: 0.4355\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  2%|▎         | 50/2000 [08:54<5:26:21, 10.04s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 18ms/step - loss: 0.4587 - f1: 0.5707 - val_loss: 0.5279 - val_f1: 0.5617\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4463 - f1: 0.6109 - val_loss: 0.5370 - val_f1: 0.5536\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  3%|▎         | 51/2000 [09:03<5:11:49,  9.60s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4963 - f1: 0.7045 - val_loss: 0.5185 - val_f1: 0.5934\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4721 - f1: 0.6976 - val_loss: 0.5354 - val_f1: 0.5692\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  3%|▎         | 52/2000 [09:11<4:58:42,  9.20s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 1.0113 - f1: 0.4506 - val_loss: 1.0446 - val_f1: 0.4372\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 1.0016 - f1: 0.4516 - val_loss: 1.0439 - val_f1: 0.4184\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  3%|▎         | 53/2000 [09:24<5:32:16, 10.24s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 1.0642 - f1: 0.3779 - val_loss: 1.0406 - val_f1: 0.4209\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 7s 14ms/step - loss: 1.0489 - f1: 0.3964 - val_loss: 1.1005 - val_f1: 0.2834\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  3%|▎         | 54/2000 [09:37<6:05:09, 11.26s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 7s 13ms/step - loss: 1.0552 - f1: 0.4016 - val_loss: 1.0415 - val_f1: 0.4401\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 1.0434 - f1: 0.4232 - val_loss: 1.0391 - val_f1: 0.4281\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  3%|▎         | 55/2000 [09:50<6:22:21, 11.79s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4873 - f1: 0.5807 - val_loss: 0.5164 - val_f1: 0.5921\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4635 - f1: 0.6278 - val_loss: 0.5217 - val_f1: 0.5642\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  3%|▎         | 56/2000 [09:59<5:46:45, 10.70s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 1.0289 - f1: 0.4215 - val_loss: 1.0398 - val_f1: 0.4096\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 7s 13ms/step - loss: 1.0159 - f1: 0.4178 - val_loss: 1.0388 - val_f1: 0.4547\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  3%|▎         | 57/2000 [10:11<6:08:00, 11.36s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 5s 18ms/step - loss: 0.5142 - f1: 0.6589 - val_loss: 0.5261 - val_f1: 0.6485\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 17ms/step - loss: 0.4986 - f1: 0.6356 - val_loss: 0.5209 - val_f1: 0.6515\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  3%|▎         | 58/2000 [10:20<5:44:16, 10.64s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 1.0198 - f1: 0.4084 - val_loss: 1.0462 - val_f1: 0.4130\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 1.0130 - f1: 0.4136 - val_loss: 1.0315 - val_f1: 0.4371\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  3%|▎         | 59/2000 [10:33<6:03:21, 11.23s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 1.0045 - f1: 0.4590 - val_loss: 1.0338 - val_f1: 0.4591\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 0.9895 - f1: 0.4689 - val_loss: 1.0364 - val_f1: 0.4623\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  3%|▎         | 60/2000 [10:46<6:17:04, 11.66s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.5345 - f1: 0.6000 - val_loss: 0.5156 - val_f1: 0.6012\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.5151 - f1: 0.6207 - val_loss: 0.5181 - val_f1: 0.5718\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  3%|▎         | 61/2000 [10:54<5:43:08, 10.62s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 1.0548 - f1: 0.3788 - val_loss: 1.0437 - val_f1: 0.3829\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 1.0379 - f1: 0.3826 - val_loss: 1.0302 - val_f1: 0.4325\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  3%|▎         | 62/2000 [11:07<6:02:35, 11.23s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.5144 - f1: 0.6147 - val_loss: 0.5127 - val_f1: 0.5746\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4820 - f1: 0.6134 - val_loss: 0.5167 - val_f1: 0.5634\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  3%|▎         | 63/2000 [11:15<5:32:35, 10.30s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4687 - f1: 0.5568 - val_loss: 0.5140 - val_f1: 0.5707\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4502 - f1: 0.5874 - val_loss: 0.5051 - val_f1: 0.6023\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  3%|▎         | 64/2000 [11:23<5:11:47,  9.66s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4846 - f1: 0.5772 - val_loss: 0.5081 - val_f1: 0.5813\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 17ms/step - loss: 0.4740 - f1: 0.5657 - val_loss: 0.5136 - val_f1: 0.5688\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  3%|▎         | 65/2000 [11:31<5:00:14,  9.31s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 5s 18ms/step - loss: 0.4160 - f1: 0.6292 - val_loss: 0.5185 - val_f1: 0.5712\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 17ms/step - loss: 0.3961 - f1: 0.6707 - val_loss: 0.5014 - val_f1: 0.6231\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  3%|▎         | 66/2000 [11:40<4:57:29,  9.23s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 1.0243 - f1: 0.4611 - val_loss: 1.0478 - val_f1: 0.4754\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 1.0162 - f1: 0.4748 - val_loss: 1.0339 - val_f1: 0.4586\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  3%|▎         | 67/2000 [11:53<5:30:37, 10.26s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 13ms/step - loss: 1.0050 - f1: 0.4374 - val_loss: 1.0458 - val_f1: 0.4777\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 1.0008 - f1: 0.4516 - val_loss: 1.0273 - val_f1: 0.4416\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  3%|▎         | 68/2000 [12:06<5:54:53, 11.02s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4347 - f1: 0.7285 - val_loss: 0.5347 - val_f1: 0.5415\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4295 - f1: 0.6676 - val_loss: 0.5062 - val_f1: 0.6032\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  3%|▎         | 69/2000 [12:14<5:27:37, 10.18s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 1.0414 - f1: 0.4325 - val_loss: 1.0285 - val_f1: 0.4472\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 1.0097 - f1: 0.4371 - val_loss: 1.0411 - val_f1: 0.4584\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  4%|▎         | 70/2000 [12:27<5:51:43, 10.93s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4863 - f1: 0.6503 - val_loss: 0.5075 - val_f1: 0.5973\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4487 - f1: 0.6896 - val_loss: 0.5075 - val_f1: 0.6074\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  4%|▎         | 71/2000 [12:35<5:26:36, 10.16s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 13ms/step - loss: 1.0599 - f1: 0.4447 - val_loss: 1.0247 - val_f1: 0.4353\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 13ms/step - loss: 1.0401 - f1: 0.4272 - val_loss: 1.0236 - val_f1: 0.4534\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  4%|▎         | 72/2000 [12:48<5:53:22, 11.00s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 17ms/step - loss: 0.4129 - f1: 0.7456 - val_loss: 0.5311 - val_f1: 0.5760\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 5s 18ms/step - loss: 0.4057 - f1: 0.7342 - val_loss: 0.5087 - val_f1: 0.6045\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  4%|▎         | 73/2000 [12:57<5:34:23, 10.41s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 7s 13ms/step - loss: 0.9959 - f1: 0.4619 - val_loss: 1.0205 - val_f1: 0.4435\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 7s 14ms/step - loss: 0.9781 - f1: 0.4516 - val_loss: 1.0253 - val_f1: 0.4769\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  4%|▎         | 74/2000 [13:11<6:04:46, 11.36s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 1.0590 - f1: 0.4053 - val_loss: 1.0269 - val_f1: 0.4763\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 1.0346 - f1: 0.4372 - val_loss: 1.0209 - val_f1: 0.4434\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  4%|▍         | 75/2000 [13:23<6:17:41, 11.77s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 13ms/step - loss: 1.0517 - f1: 0.3782 - val_loss: 1.0224 - val_f1: 0.4484\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 13ms/step - loss: 1.0416 - f1: 0.3981 - val_loss: 1.0168 - val_f1: 0.4347\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  4%|▍         | 76/2000 [13:36<6:27:53, 12.10s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 13ms/step - loss: 0.9931 - f1: 0.4771 - val_loss: 1.0117 - val_f1: 0.4476\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 0.9751 - f1: 0.4827 - val_loss: 1.0111 - val_f1: 0.4627\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  4%|▍         | 77/2000 [13:49<6:34:29, 12.31s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 13ms/step - loss: 1.0350 - f1: 0.4216 - val_loss: 1.0187 - val_f1: 0.4199\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 13ms/step - loss: 1.0209 - f1: 0.4121 - val_loss: 1.0145 - val_f1: 0.4339\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  4%|▍         | 78/2000 [14:02<6:39:26, 12.47s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 1.0146 - f1: 0.4167 - val_loss: 1.0211 - val_f1: 0.4873\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 7s 14ms/step - loss: 0.9950 - f1: 0.4246 - val_loss: 1.0090 - val_f1: 0.4615\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  4%|▍         | 79/2000 [14:15<6:47:38, 12.73s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 5s 18ms/step - loss: 0.4533 - f1: 0.7177 - val_loss: 0.5121 - val_f1: 0.5876\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4265 - f1: 0.7278 - val_loss: 0.5139 - val_f1: 0.6311\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  4%|▍         | 80/2000 [14:24<6:10:02, 11.56s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4643 - f1: 0.6857 - val_loss: 0.5049 - val_f1: 0.6260\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4371 - f1: 0.6637 - val_loss: 0.5055 - val_f1: 0.6356\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  4%|▍         | 81/2000 [14:32<5:38:13, 10.57s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4679 - f1: 0.5999 - val_loss: 0.5226 - val_f1: 0.5746\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4419 - f1: 0.6667 - val_loss: 0.5234 - val_f1: 0.5691\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  4%|▍         | 82/2000 [14:41<5:16:51,  9.91s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 7s 14ms/step - loss: 0.9977 - f1: 0.4661 - val_loss: 1.0489 - val_f1: 0.4001\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 7s 13ms/step - loss: 0.9890 - f1: 0.4682 - val_loss: 1.0098 - val_f1: 0.4698\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  4%|▍         | 83/2000 [14:55<5:56:59, 11.17s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.5084 - f1: 0.6097 - val_loss: 0.5087 - val_f1: 0.6231\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4983 - f1: 0.6267 - val_loss: 0.5117 - val_f1: 0.6441\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  4%|▍         | 84/2000 [15:03<5:28:57, 10.30s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 0.9526 - f1: 0.5268 - val_loss: 1.0086 - val_f1: 0.4710\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 13ms/step - loss: 0.9338 - f1: 0.5270 - val_loss: 1.0092 - val_f1: 0.4594\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  4%|▍         | 85/2000 [15:16<5:52:47, 11.05s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 13ms/step - loss: 0.9969 - f1: 0.4752 - val_loss: 1.0088 - val_f1: 0.4307\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 13ms/step - loss: 0.9772 - f1: 0.4693 - val_loss: 1.0046 - val_f1: 0.4582\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  4%|▍         | 86/2000 [15:29<6:10:21, 11.61s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 7s 14ms/step - loss: 0.9834 - f1: 0.4679 - val_loss: 1.0321 - val_f1: 0.4068\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 7s 13ms/step - loss: 0.9681 - f1: 0.4927 - val_loss: 0.9994 - val_f1: 0.4757\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  4%|▍         | 87/2000 [15:43<6:31:04, 12.27s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 13ms/step - loss: 1.0424 - f1: 0.4568 - val_loss: 0.9999 - val_f1: 0.4656\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 13ms/step - loss: 1.0251 - f1: 0.4708 - val_loss: 1.0020 - val_f1: 0.4705\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  4%|▍         | 88/2000 [15:55<6:36:49, 12.45s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4386 - f1: 0.6428 - val_loss: 0.5219 - val_f1: 0.5919\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4216 - f1: 0.6635 - val_loss: 0.5166 - val_f1: 0.5934\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  4%|▍         | 89/2000 [16:04<5:57:49, 11.23s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4516 - f1: 0.7226 - val_loss: 0.5194 - val_f1: 0.5788\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 17ms/step - loss: 0.4415 - f1: 0.7619 - val_loss: 0.5199 - val_f1: 0.5781\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  4%|▍         | 90/2000 [16:12<5:31:11, 10.40s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4641 - f1: 0.6557 - val_loss: 0.5118 - val_f1: 0.6410\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4395 - f1: 0.6917 - val_loss: 0.5127 - val_f1: 0.6241\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  5%|▍         | 91/2000 [16:21<5:11:54,  9.80s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 13ms/step - loss: 1.0099 - f1: 0.4664 - val_loss: 1.0010 - val_f1: 0.4506\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 13ms/step - loss: 0.9987 - f1: 0.4605 - val_loss: 1.0012 - val_f1: 0.4457\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  5%|▍         | 92/2000 [16:34<5:41:36, 10.74s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 7s 13ms/step - loss: 1.0086 - f1: 0.4367 - val_loss: 0.9993 - val_f1: 0.4591\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 13ms/step - loss: 0.9888 - f1: 0.4667 - val_loss: 1.0020 - val_f1: 0.4497\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  5%|▍         | 93/2000 [16:47<6:02:44, 11.41s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 7s 13ms/step - loss: 1.0036 - f1: 0.4462 - val_loss: 1.0050 - val_f1: 0.4540\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 7s 14ms/step - loss: 0.9934 - f1: 0.4535 - val_loss: 1.0018 - val_f1: 0.4670\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  5%|▍         | 94/2000 [17:00<6:24:44, 12.11s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 13ms/step - loss: 1.0301 - f1: 0.4171 - val_loss: 0.9978 - val_f1: 0.4802\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 13ms/step - loss: 1.0115 - f1: 0.4135 - val_loss: 1.0106 - val_f1: 0.4243\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  5%|▍         | 95/2000 [17:13<6:31:45, 12.34s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 13ms/step - loss: 1.0213 - f1: 0.4203 - val_loss: 0.9985 - val_f1: 0.4580\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 1.0133 - f1: 0.4379 - val_loss: 0.9969 - val_f1: 0.4847\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  5%|▍         | 96/2000 [17:26<6:36:26, 12.49s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4863 - f1: 0.6873 - val_loss: 0.5053 - val_f1: 0.6322\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 17ms/step - loss: 0.4798 - f1: 0.6297 - val_loss: 0.5078 - val_f1: 0.6009\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  5%|▍         | 97/2000 [17:35<5:58:28, 11.30s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 7s 13ms/step - loss: 1.0165 - f1: 0.4747 - val_loss: 0.9902 - val_f1: 0.4913\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 7s 13ms/step - loss: 1.0036 - f1: 0.4687 - val_loss: 0.9884 - val_f1: 0.4920\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  5%|▍         | 98/2000 [17:48<6:16:16, 11.87s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.5142 - f1: 0.6284 - val_loss: 0.5105 - val_f1: 0.6347\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4884 - f1: 0.6910 - val_loss: 0.5118 - val_f1: 0.6254\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  5%|▍         | 99/2000 [17:56<5:42:21, 10.81s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 0.9788 - f1: 0.4559 - val_loss: 1.0006 - val_f1: 0.5056\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 0.9605 - f1: 0.4824 - val_loss: 0.9914 - val_f1: 0.4971\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  5%|▌         | 100/2000 [18:09<6:01:02, 11.40s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 5s 18ms/step - loss: 0.4942 - f1: 0.6151 - val_loss: 0.5101 - val_f1: 0.5931\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 5s 18ms/step - loss: 0.4507 - f1: 0.7068 - val_loss: 0.5185 - val_f1: 0.5764\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  5%|▌         | 101/2000 [18:18<5:40:27, 10.76s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 7s 13ms/step - loss: 0.9884 - f1: 0.4427 - val_loss: 0.9888 - val_f1: 0.4590\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 13ms/step - loss: 0.9762 - f1: 0.4454 - val_loss: 0.9865 - val_f1: 0.4851\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  5%|▌         | 102/2000 [18:31<6:03:02, 11.48s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 13ms/step - loss: 1.0106 - f1: 0.4725 - val_loss: 0.9937 - val_f1: 0.4930\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 7s 13ms/step - loss: 0.9963 - f1: 0.4918 - val_loss: 0.9992 - val_f1: 0.4944\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  5%|▌         | 103/2000 [18:44<6:17:08, 11.93s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4478 - f1: 0.5949 - val_loss: 0.5011 - val_f1: 0.6312\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4361 - f1: 0.6045 - val_loss: 0.5044 - val_f1: 0.6452\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  5%|▌         | 104/2000 [18:53<5:42:59, 10.85s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4657 - f1: 0.6046 - val_loss: 0.5231 - val_f1: 0.5772\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4450 - f1: 0.6304 - val_loss: 0.5367 - val_f1: 0.5714\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  5%|▌         | 105/2000 [19:01<5:18:51, 10.10s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 7s 13ms/step - loss: 1.0132 - f1: 0.4553 - val_loss: 1.0016 - val_f1: 0.4397\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 7s 13ms/step - loss: 0.9998 - f1: 0.4677 - val_loss: 0.9882 - val_f1: 0.4778\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  5%|▌         | 106/2000 [19:14<5:49:31, 11.07s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 17ms/step - loss: 0.3960 - f1: 0.5638 - val_loss: 0.5111 - val_f1: 0.5801\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.3754 - f1: 0.6765 - val_loss: 0.5272 - val_f1: 0.5785\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  5%|▌         | 107/2000 [19:23<5:24:52, 10.30s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4137 - f1: 0.7059 - val_loss: 0.5102 - val_f1: 0.5855\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.3982 - f1: 0.6971 - val_loss: 0.5116 - val_f1: 0.5854\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  5%|▌         | 108/2000 [19:31<5:06:17,  9.71s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 5s 18ms/step - loss: 0.4478 - f1: 0.6518 - val_loss: 0.5041 - val_f1: 0.6169\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 5s 18ms/step - loss: 0.4278 - f1: 0.7167 - val_loss: 0.5267 - val_f1: 0.5767\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  5%|▌         | 109/2000 [19:40<5:01:12,  9.56s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 13ms/step - loss: 0.9357 - f1: 0.5276 - val_loss: 0.9882 - val_f1: 0.4808\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 7s 13ms/step - loss: 0.9305 - f1: 0.5145 - val_loss: 0.9884 - val_f1: 0.4883\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  6%|▌         | 110/2000 [19:54<5:35:57, 10.67s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 7s 14ms/step - loss: 1.0177 - f1: 0.4500 - val_loss: 1.0005 - val_f1: 0.4647\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 13ms/step - loss: 1.0056 - f1: 0.4434 - val_loss: 0.9924 - val_f1: 0.4823\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  6%|▌         | 111/2000 [20:07<6:05:13, 11.60s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.3953 - f1: 0.6980 - val_loss: 0.5206 - val_f1: 0.5898\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.3871 - f1: 0.7177 - val_loss: 0.5092 - val_f1: 0.6480\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  6%|▌         | 112/2000 [20:16<5:33:29, 10.60s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 13ms/step - loss: 1.0108 - f1: 0.4802 - val_loss: 0.9853 - val_f1: 0.4860\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 13ms/step - loss: 0.9888 - f1: 0.4570 - val_loss: 0.9950 - val_f1: 0.4957\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  6%|▌         | 113/2000 [20:29<5:55:06, 11.29s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 13ms/step - loss: 0.9789 - f1: 0.4955 - val_loss: 0.9856 - val_f1: 0.4971\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 13ms/step - loss: 0.9628 - f1: 0.4896 - val_loss: 0.9886 - val_f1: 0.4969\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  6%|▌         | 114/2000 [20:41<6:10:02, 11.77s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 0.9516 - f1: 0.4827 - val_loss: 0.9856 - val_f1: 0.5056\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 7s 13ms/step - loss: 0.9398 - f1: 0.4932 - val_loss: 0.9837 - val_f1: 0.4944\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  6%|▌         | 115/2000 [20:54<6:21:22, 12.14s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 7s 14ms/step - loss: 0.9518 - f1: 0.5427 - val_loss: 0.9861 - val_f1: 0.5096\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 13ms/step - loss: 0.9398 - f1: 0.5321 - val_loss: 0.9910 - val_f1: 0.5068\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  6%|▌         | 116/2000 [21:08<6:33:49, 12.54s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 1.0087 - f1: 0.4678 - val_loss: 0.9821 - val_f1: 0.4941\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 13ms/step - loss: 0.9884 - f1: 0.4739 - val_loss: 0.9918 - val_f1: 0.4758\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  6%|▌         | 117/2000 [21:21<6:35:47, 12.61s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 13ms/step - loss: 0.9873 - f1: 0.4822 - val_loss: 0.9862 - val_f1: 0.4979\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 13ms/step - loss: 0.9665 - f1: 0.5086 - val_loss: 0.9812 - val_f1: 0.4936\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  6%|▌         | 118/2000 [21:34<6:38:03, 12.69s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 7s 13ms/step - loss: 0.9628 - f1: 0.4932 - val_loss: 0.9843 - val_f1: 0.5051\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 7s 13ms/step - loss: 0.9492 - f1: 0.5290 - val_loss: 0.9851 - val_f1: 0.4886\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  6%|▌         | 119/2000 [21:47<6:41:07, 12.80s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 13ms/step - loss: 0.9871 - f1: 0.4979 - val_loss: 0.9879 - val_f1: 0.5050\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 7s 13ms/step - loss: 0.9581 - f1: 0.5207 - val_loss: 0.9834 - val_f1: 0.5046\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  6%|▌         | 120/2000 [22:00<6:43:10, 12.87s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 0.9956 - f1: 0.5230 - val_loss: 0.9904 - val_f1: 0.4731\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 13ms/step - loss: 0.9858 - f1: 0.5139 - val_loss: 0.9834 - val_f1: 0.4792\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  6%|▌         | 121/2000 [22:12<6:41:49, 12.83s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 5s 18ms/step - loss: 0.4419 - f1: 0.7318 - val_loss: 0.5160 - val_f1: 0.6537\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 5s 18ms/step - loss: 0.4165 - f1: 0.7553 - val_loss: 0.5183 - val_f1: 0.5860\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  6%|▌         | 122/2000 [22:22<6:07:19, 11.74s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 7s 13ms/step - loss: 0.9435 - f1: 0.5222 - val_loss: 0.9758 - val_f1: 0.5067\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 13ms/step - loss: 0.9330 - f1: 0.5133 - val_loss: 0.9852 - val_f1: 0.5071\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  6%|▌         | 123/2000 [22:35<6:18:48, 12.11s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4381 - f1: 0.6897 - val_loss: 0.5198 - val_f1: 0.5934\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.3981 - f1: 0.7348 - val_loss: 0.5099 - val_f1: 0.6408\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  6%|▌         | 124/2000 [22:43<5:42:24, 10.95s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 0.9982 - f1: 0.4892 - val_loss: 0.9750 - val_f1: 0.4926\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 13ms/step - loss: 0.9820 - f1: 0.4928 - val_loss: 0.9738 - val_f1: 0.4885\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  6%|▋         | 125/2000 [22:56<5:59:08, 11.49s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 1.0035 - f1: 0.4306 - val_loss: 0.9768 - val_f1: 0.4837\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 0.9935 - f1: 0.4465 - val_loss: 0.9746 - val_f1: 0.4874\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  6%|▋         | 126/2000 [23:08<6:10:00, 11.85s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 13ms/step - loss: 0.9324 - f1: 0.5504 - val_loss: 0.9729 - val_f1: 0.5017\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 7s 14ms/step - loss: 0.9157 - f1: 0.5584 - val_loss: 0.9779 - val_f1: 0.5128\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  6%|▋         | 127/2000 [23:22<6:24:11, 12.31s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 7s 13ms/step - loss: 0.9782 - f1: 0.4803 - val_loss: 0.9777 - val_f1: 0.4855\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 7s 13ms/step - loss: 0.9617 - f1: 0.4837 - val_loss: 0.9724 - val_f1: 0.5014\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  6%|▋         | 128/2000 [23:35<6:31:38, 12.55s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 7s 14ms/step - loss: 0.9595 - f1: 0.5012 - val_loss: 0.9715 - val_f1: 0.5076\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 13ms/step - loss: 0.9382 - f1: 0.5077 - val_loss: 0.9763 - val_f1: 0.5189\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  6%|▋         | 129/2000 [23:48<6:40:41, 12.85s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 0.9460 - f1: 0.5158 - val_loss: 0.9800 - val_f1: 0.5082\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 0.9341 - f1: 0.5358 - val_loss: 0.9752 - val_f1: 0.5052\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  6%|▋         | 130/2000 [24:01<6:39:50, 12.83s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 0.9060 - f1: 0.5583 - val_loss: 0.9716 - val_f1: 0.5064\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 0.8855 - f1: 0.5833 - val_loss: 0.9729 - val_f1: 0.5119\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  7%|▋         | 131/2000 [24:14<6:38:41, 12.80s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 13ms/step - loss: 0.9534 - f1: 0.5053 - val_loss: 0.9729 - val_f1: 0.5159\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 7s 13ms/step - loss: 0.9314 - f1: 0.5223 - val_loss: 0.9751 - val_f1: 0.4874\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  7%|▋         | 132/2000 [24:27<6:40:03, 12.85s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 0.9322 - f1: 0.5180 - val_loss: 0.9946 - val_f1: 0.5154\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 13ms/step - loss: 0.9221 - f1: 0.5492 - val_loss: 0.9676 - val_f1: 0.5160\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  7%|▋         | 133/2000 [24:40<6:39:24, 12.84s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4635 - f1: 0.6378 - val_loss: 0.5199 - val_f1: 0.5988\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4260 - f1: 0.6962 - val_loss: 0.5188 - val_f1: 0.5884\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  7%|▋         | 134/2000 [24:48<5:56:33, 11.47s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4971 - f1: 0.6328 - val_loss: 0.5165 - val_f1: 0.5903\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 17ms/step - loss: 0.4795 - f1: 0.6629 - val_loss: 0.5217 - val_f1: 0.5958\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  7%|▋         | 135/2000 [24:56<5:29:06, 10.59s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 7s 14ms/step - loss: 1.0112 - f1: 0.4823 - val_loss: 0.9737 - val_f1: 0.4950\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 7s 14ms/step - loss: 0.9900 - f1: 0.4930 - val_loss: 0.9678 - val_f1: 0.4990\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  7%|▋         | 136/2000 [25:11<6:06:05, 11.78s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 13ms/step - loss: 0.9711 - f1: 0.4951 - val_loss: 0.9680 - val_f1: 0.5099\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 0.9560 - f1: 0.5178 - val_loss: 0.9712 - val_f1: 0.4865\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  7%|▋         | 137/2000 [25:24<6:15:30, 12.09s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4262 - f1: 0.7126 - val_loss: 0.5281 - val_f1: 0.5894\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4120 - f1: 0.6980 - val_loss: 0.5205 - val_f1: 0.5902\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  7%|▋         | 138/2000 [25:32<5:39:41, 10.95s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4305 - f1: 0.7568 - val_loss: 0.5198 - val_f1: 0.6355\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4039 - f1: 0.7662 - val_loss: 0.5243 - val_f1: 0.6348\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  7%|▋         | 139/2000 [25:40<5:14:18, 10.13s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4300 - f1: 0.6920 - val_loss: 0.5289 - val_f1: 0.6099\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4237 - f1: 0.6233 - val_loss: 0.5155 - val_f1: 0.6408\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  7%|▋         | 140/2000 [25:49<4:56:29,  9.56s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4748 - f1: 0.6709 - val_loss: 0.5100 - val_f1: 0.6421\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4481 - f1: 0.6779 - val_loss: 0.5093 - val_f1: 0.6457\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  7%|▋         | 141/2000 [25:57<4:43:53,  9.16s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4876 - f1: 0.6392 - val_loss: 0.5103 - val_f1: 0.6339\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4650 - f1: 0.6437 - val_loss: 0.5125 - val_f1: 0.6273\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  7%|▋         | 142/2000 [26:05<4:35:02,  8.88s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4618 - f1: 0.6569 - val_loss: 0.5010 - val_f1: 0.6155\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4237 - f1: 0.7044 - val_loss: 0.4999 - val_f1: 0.6154\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  7%|▋         | 143/2000 [26:13<4:28:48,  8.69s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 7s 14ms/step - loss: 0.9823 - f1: 0.4935 - val_loss: 0.9761 - val_f1: 0.4919\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 7s 13ms/step - loss: 0.9551 - f1: 0.5117 - val_loss: 0.9685 - val_f1: 0.5093\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  7%|▋         | 144/2000 [26:27<5:14:58, 10.18s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4139 - f1: 0.6954 - val_loss: 0.5292 - val_f1: 0.6497\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4023 - f1: 0.7334 - val_loss: 0.5037 - val_f1: 0.6391\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  7%|▋         | 145/2000 [26:35<4:56:44,  9.60s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 0.9565 - f1: 0.5062 - val_loss: 0.9699 - val_f1: 0.4926\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 0.9446 - f1: 0.5366 - val_loss: 0.9727 - val_f1: 0.4873\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  7%|▋         | 146/2000 [26:48<5:25:17, 10.53s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 0.9832 - f1: 0.4585 - val_loss: 0.9672 - val_f1: 0.5060\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 0.9654 - f1: 0.4679 - val_loss: 0.9915 - val_f1: 0.4648\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  7%|▋         | 147/2000 [27:01<5:46:03, 11.21s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4451 - f1: 0.6609 - val_loss: 0.5021 - val_f1: 0.6237\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4220 - f1: 0.6485 - val_loss: 0.5019 - val_f1: 0.6287\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  7%|▋         | 148/2000 [27:09<5:18:34, 10.32s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.3774 - f1: 0.7029 - val_loss: 0.5163 - val_f1: 0.5920\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.3488 - f1: 0.7233 - val_loss: 0.5045 - val_f1: 0.6125\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  7%|▋         | 149/2000 [27:17<4:58:58,  9.69s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.3911 - f1: 0.7330 - val_loss: 0.5114 - val_f1: 0.6007\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.3842 - f1: 0.7065 - val_loss: 0.5108 - val_f1: 0.5969\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  8%|▊         | 150/2000 [27:25<4:45:26,  9.26s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 0.9834 - f1: 0.5037 - val_loss: 0.9714 - val_f1: 0.5052\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 7s 13ms/step - loss: 0.9667 - f1: 0.5140 - val_loss: 0.9628 - val_f1: 0.5083\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  8%|▊         | 151/2000 [27:38<5:20:47, 10.41s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 7s 14ms/step - loss: 0.9812 - f1: 0.5270 - val_loss: 0.9629 - val_f1: 0.5067\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 0.9628 - f1: 0.5298 - val_loss: 0.9617 - val_f1: 0.5194\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  8%|▊         | 152/2000 [27:52<5:47:35, 11.29s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4296 - f1: 0.6792 - val_loss: 0.5033 - val_f1: 0.6346\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4109 - f1: 0.6568 - val_loss: 0.5073 - val_f1: 0.6368\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  8%|▊         | 153/2000 [28:00<5:19:59, 10.40s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.3667 - f1: 0.7991 - val_loss: 0.5122 - val_f1: 0.6345\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.3440 - f1: 0.8006 - val_loss: 0.5185 - val_f1: 0.6283\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  8%|▊         | 154/2000 [28:08<5:00:13,  9.76s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4190 - f1: 0.7476 - val_loss: 0.5144 - val_f1: 0.6340\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.3907 - f1: 0.7718 - val_loss: 0.5150 - val_f1: 0.6343\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  8%|▊         | 155/2000 [28:17<4:45:31,  9.29s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 7s 13ms/step - loss: 0.9357 - f1: 0.5020 - val_loss: 0.9608 - val_f1: 0.5239\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 7s 13ms/step - loss: 0.9159 - f1: 0.5227 - val_loss: 0.9649 - val_f1: 0.5341\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  8%|▊         | 156/2000 [28:30<5:25:02, 10.58s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.3878 - f1: 0.7511 - val_loss: 0.5212 - val_f1: 0.6621\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.3748 - f1: 0.7839 - val_loss: 0.5171 - val_f1: 0.6366\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  8%|▊         | 157/2000 [28:38<5:03:50,  9.89s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 0.9572 - f1: 0.4995 - val_loss: 0.9619 - val_f1: 0.5110\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 13ms/step - loss: 0.9431 - f1: 0.4866 - val_loss: 0.9607 - val_f1: 0.5130\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  8%|▊         | 158/2000 [28:51<5:30:32, 10.77s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4357 - f1: 0.6387 - val_loss: 0.5199 - val_f1: 0.5904\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 5s 18ms/step - loss: 0.4100 - f1: 0.6232 - val_loss: 0.5195 - val_f1: 0.5875\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  8%|▊         | 159/2000 [29:00<5:11:55, 10.17s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 5s 18ms/step - loss: 0.4658 - f1: 0.6856 - val_loss: 0.5108 - val_f1: 0.6206\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4506 - f1: 0.7054 - val_loss: 0.5104 - val_f1: 0.6392\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  8%|▊         | 160/2000 [29:09<4:58:43,  9.74s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.3803 - f1: 0.7069 - val_loss: 0.5114 - val_f1: 0.6388\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.3589 - f1: 0.7399 - val_loss: 0.5440 - val_f1: 0.5886\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  8%|▊         | 161/2000 [29:17<4:44:28,  9.28s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 0.9745 - f1: 0.5273 - val_loss: 0.9600 - val_f1: 0.5102\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 7s 13ms/step - loss: 0.9567 - f1: 0.5297 - val_loss: 0.9597 - val_f1: 0.5090\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  8%|▊         | 162/2000 [29:30<5:17:58, 10.38s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 13ms/step - loss: 0.9746 - f1: 0.5007 - val_loss: 0.9624 - val_f1: 0.5215\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 13ms/step - loss: 0.9604 - f1: 0.5148 - val_loss: 0.9618 - val_f1: 0.5169\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  8%|▊         | 163/2000 [29:43<5:40:22, 11.12s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 0.9115 - f1: 0.5442 - val_loss: 0.9745 - val_f1: 0.5250\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 0.8957 - f1: 0.5483 - val_loss: 0.9656 - val_f1: 0.5181\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  8%|▊         | 164/2000 [29:56<5:55:11, 11.61s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4016 - f1: 0.7621 - val_loss: 0.5358 - val_f1: 0.6496\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.3804 - f1: 0.7824 - val_loss: 0.5445 - val_f1: 0.6042\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  8%|▊         | 165/2000 [30:04<5:24:38, 10.62s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4161 - f1: 0.6786 - val_loss: 0.5297 - val_f1: 0.6161\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.3906 - f1: 0.7188 - val_loss: 0.5207 - val_f1: 0.6662\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  8%|▊         | 166/2000 [30:12<5:03:21,  9.92s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 7s 14ms/step - loss: 0.9937 - f1: 0.4815 - val_loss: 0.9676 - val_f1: 0.5103\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 7s 14ms/step - loss: 0.9755 - f1: 0.5041 - val_loss: 0.9703 - val_f1: 0.5104\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  8%|▊         | 167/2000 [30:27<5:44:08, 11.26s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 0.9807 - f1: 0.4898 - val_loss: 0.9769 - val_f1: 0.5231\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 0.9620 - f1: 0.5086 - val_loss: 0.9631 - val_f1: 0.5141\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  8%|▊         | 168/2000 [30:39<5:57:29, 11.71s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4433 - f1: 0.7156 - val_loss: 0.5135 - val_f1: 0.6589\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4243 - f1: 0.7437 - val_loss: 0.5153 - val_f1: 0.6292\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  8%|▊         | 169/2000 [30:48<5:25:39, 10.67s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 0.9575 - f1: 0.4919 - val_loss: 0.9608 - val_f1: 0.5186\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 0.9307 - f1: 0.5263 - val_loss: 0.9613 - val_f1: 0.5296\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  8%|▊         | 170/2000 [31:00<5:44:42, 11.30s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4755 - f1: 0.6749 - val_loss: 0.5124 - val_f1: 0.6479\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4536 - f1: 0.6584 - val_loss: 0.5155 - val_f1: 0.6426\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  9%|▊         | 171/2000 [31:08<5:16:11, 10.37s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 0.9276 - f1: 0.5300 - val_loss: 0.9663 - val_f1: 0.5146\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 6s 12ms/step - loss: 0.9061 - f1: 0.5362 - val_loss: 0.9584 - val_f1: 0.5311\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  9%|▊         | 172/2000 [31:21<5:38:02, 11.10s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4356 - f1: 0.6592 - val_loss: 0.5061 - val_f1: 0.6357\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.3914 - f1: 0.7051 - val_loss: 0.5142 - val_f1: 0.5949\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  9%|▊         | 173/2000 [31:30<5:11:52, 10.24s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 512 samples, validate on 3000 samples\n",
      "Epoch 1/2\n",
      "512/512 [==============================] - 6s 13ms/step - loss: 0.9342 - f1: 0.5510 - val_loss: 0.9645 - val_f1: 0.5269\n",
      "Epoch 2/2\n",
      "512/512 [==============================] - 7s 14ms/step - loss: 0.9188 - f1: 0.5532 - val_loss: 0.9641 - val_f1: 0.5272\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  9%|▊         | 174/2000 [31:43<5:41:10, 11.21s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 17ms/step - loss: 0.4047 - f1: 0.6785 - val_loss: 0.5077 - val_f1: 0.6608\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.3894 - f1: 0.7014 - val_loss: 0.5075 - val_f1: 0.6274\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  9%|▉         | 175/2000 [31:52<5:16:54, 10.42s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4238 - f1: 0.6577 - val_loss: 0.5021 - val_f1: 0.6287\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.4011 - f1: 0.6538 - val_loss: 0.5097 - val_f1: 0.6154\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  9%|▉         | 176/2000 [32:00<4:57:15,  9.78s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.3518 - f1: 0.7106 - val_loss: 0.5648 - val_f1: 0.5583\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.3235 - f1: 0.7673 - val_loss: 0.5235 - val_f1: 0.5946\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  9%|▉         | 177/2000 [32:08<4:42:53,  9.31s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 256 samples, validate on 800 samples\n",
      "Epoch 1/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.3687 - f1: 0.7596 - val_loss: 0.5361 - val_f1: 0.5721\n",
      "Epoch 2/2\n",
      "256/256 [==============================] - 4s 16ms/step - loss: 0.3538 - f1: 0.7384 - val_loss: 0.5383 - val_f1: 0.5746\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  9%|▉         | 178/2000 [32:16<4:32:53,  8.99s/it]"
     ]
    }
   ],
   "source": [
    "j = 0\n",
    "k = 0\n",
    "for i in tqdm(range(2000)):\n",
    "    value=np.random.binomial(1,0.5)\n",
    "    if value==1:\n",
    "        #x,y = sample(64,value)\n",
    "        if(j == 16):\n",
    "            j = 0\n",
    "        x = x_rumour_train[j*256:(j*256)+256]\n",
    "        y = y_rumour_train[j*256:(j*256)+256]\n",
    "        j+=1\n",
    "        rumour_model.fit(x,y,batch_size=16,epochs=2,validation_data=(x_rumour_val,y_rumour_val ))\n",
    "    else:\n",
    "        #x,y = sample(128,value)\n",
    "        if(k == 32):\n",
    "            k = 0\n",
    "        x = x_fakenews_train[k*512:(k*512)+512]\n",
    "        y = y_fakenews_train[k*512:(k*512)+512]\n",
    "        k+=1\n",
    "        fake_model.fit(x,y,batch_size=32,epochs=2,validation_data=(x_fakenews_val,y_fakenews_val))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1999\n"
     ]
    }
   ],
   "source": [
    "print(i)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "#rumour_model.fit(x_rumour_train,y_rumour_train,batch_size=32,epochs=3,validation_data=(x_rumour_val,y_rumour_val ))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "#fake_model.fit(x_fakenews_train,y_fakenews_train,batch_size=32,epochs=8,validation_data=(x_fakenews_val,y_fakenews_val))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(800,)\n",
      "(800,)\n"
     ]
    }
   ],
   "source": [
    "y_rumour_predictions = rumour_model.predict(x_rumour_val)\n",
    "# print(y_rumour_predictions.shape)\n",
    "for i in range(0,y_rumour_predictions.shape[0]):\n",
    "    if y_rumour_predictions[i]>0.5:\n",
    "        y_rumour_predictions[i]=1\n",
    "    else :\n",
    "        y_rumour_predictions[i]=0\n",
    "# np.squeeze(y_rumour_predictions)\n",
    "y_rumour_predictions = y_rumour_predictions.squeeze()\n",
    "print(y_rumour_predictions.shape)\n",
    "# y_rumour_val= y_rumour_val.reshape((y_rumour_predictions.shape[0],1))\n",
    "print(y_rumour_val.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "No of correctly predicted points: 592\n",
      "Accuracy: 74.0\n"
     ]
    }
   ],
   "source": [
    "accuracy=0\n",
    "for i in range(0,y_rumour_val.shape[0]):\n",
    "    if(y_rumour_val[i]==y_rumour_predictions[i]):\n",
    "        accuracy+=1  \n",
    "print(\"No of correctly predicted points:\",accuracy)\n",
    "print(\"Accuracy:\", (accuracy*100)/y_rumour_val.shape[0]) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(3000, 4)\n",
      "[[4.73333485e-02 6.44735701e-04 6.62013481e-04 9.51359987e-01]\n",
      " [9.00628328e-01 8.64438564e-02 3.65091313e-04 1.25627685e-02]\n",
      " [4.02883410e-01 3.04137897e-02 4.21832763e-02 5.24519503e-01]\n",
      " ...\n",
      " [1.36195078e-01 4.48247930e-03 3.89518100e-03 8.55427265e-01]\n",
      " [2.35466771e-02 9.20963958e-02 1.77098215e-02 8.66647065e-01]\n",
      " [9.88073409e-01 1.71927700e-03 8.11068894e-05 1.01262201e-02]]\n",
      "(3000,)\n",
      "[3 0 3 ... 3 3 0]\n",
      "(3000,)\n",
      "[3 1 3 ... 3 3 0]\n"
     ]
    }
   ],
   "source": [
    "y_fakenews_predictions = fake_model.predict(x_fakenews_val)\n",
    "print(y_fakenews_predictions.shape)\n",
    "print(y_fakenews_predictions)\n",
    "y_fakenews_predictions=np.argmax(y_fakenews_predictions,axis=1)\n",
    "print(y_fakenews_predictions.shape)\n",
    "print(y_fakenews_predictions)\n",
    "y_fakenews_actual=np.argmax(y_fakenews_val,axis=1)\n",
    "print(y_fakenews_actual.shape)\n",
    "print(y_fakenews_actual)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "No of correctly predicted points: 2076\n",
      "Accuracy: 69.2\n"
     ]
    }
   ],
   "source": [
    "accuracy=0\n",
    "for i in range(0,y_fakenews_actual.shape[0]):\n",
    "    if(y_fakenews_actual[i]==y_fakenews_predictions[i]):\n",
    "        accuracy+=1  \n",
    "print(\"No of correctly predicted points:\",accuracy)\n",
    "print(\"Accuracy:\", (accuracy*100)/y_fakenews_actual.shape[0]) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [],
   "source": [
    "def confusionMatrix(a,b):\n",
    "    #T_class = list(map(int,y_adm_val >= t ))\n",
    "    #out = predict(X_adm_val,Theta,t,Mean,Std)\n",
    "    cm = confusion_matrix(a, b)\n",
    "    ax= plt.subplot()\n",
    "    sns.heatmap(cm, annot=True, ax = ax); #annot=True to annotate cells\n",
    "\n",
    "    # labels, title and ticks\n",
    "    ax.set_xlabel('Predicted labels');ax.set_ylabel('True labels'); \n",
    "    ax.set_title('Confusion Matrix'); \n",
    "    #ax.xaxis.set_ticklabels(['1', '0']); ax.yaxis.set_ticklabels(['1', '0'])\n",
    "    print(\"Micro-F1-score\",f1_score(a,b, average = 'micro'))\n",
    "    print(\"Macro-F1-score\",f1_score(a,b, average = 'macro'))\n",
    "#     print(\"precision\",precision_score(T_class,out, average = 'micro'))\n",
    "#     print(\"Accuracy\",accuracy_score(T_class,out))\n",
    "#     print(\"Recall\",recall_score(T_class,out, average = 'micro'))\n",
    "\n",
    "    \n",
    "    \n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [],
   "source": [
    "import seaborn as sns\n",
    "from sklearn.metrics import confusion_matrix"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Micro-F1-score 0.74\n",
      "Macro-F1-score 0.7191447417279351\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAW4AAAEWCAYAAABG030jAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4zLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvnQurowAAIABJREFUeJzt3XecFdX5x/HPl6VKF5CqsWGP2IItxq6gKIRYUGMlEgua2EuMir0bUTRgsHcDdhRL9GcXFEVFCWBDigIiIGBhd5/fH3cgy7Ls3mX37t1Zvm9f89p7z8w958y6PHv2mTNnFBGYmVl61Mt3B8zMrHIcuM3MUsaB28wsZRy4zcxSxoHbzCxlHLjNzFLGgduqTFITSU9Jmi/p0SrUc4Sk56uzb/kg6VlJR+e7H1Z3OXCvRiQdLuldSQslzUwCzG+roeqDgPZAm4g4eFUriYj7I2KfaujPciTtJikkPVaqvFtS/kqW9Vws6b6KjouInhFx9yp216xCDtyrCUmnA/8AriATZNcBbgV6V0P1vwImRURhNdSVK7OBHSW1KVF2NDCpuhpQhv9NWc75h2w1IKklcAlwckSMjIhFEbEkIp6KiLOSYxpJ+oekGcn2D0mNkn27SZom6QxJs5LR+rHJvkHAhcChyUi+f+mRqaR1k5Ft/eT9MZI+l/SDpC8kHVGi/PUSn9tJ0tgkBTNW0k4l9r0i6VJJbyT1PC+pbTnfhl+Ax4F+yecLgEOB+0t9r26S9LWkBZLek7RLUt4DOL/EeY4v0Y/LJb0BLAbWT8r+lOy/TdKIEvVfLeklScr6f6BZKQ7cq4cdgcbAY+Uc8zdgB2AroBvQHbigxP4OQEugM9AfGCKpdURcRGYU/3BENIuI4eV1RFJTYDDQMyKaAzsBH5Rx3JrAM8mxbYAbgGdKjZgPB44F1gIaAmeW1zZwD3BU8npf4GNgRqljxpL5HqwJPAA8KqlxRDxX6jy7lfjMkcAAoDnwVan6zgB+nfxS2oXM9+7o8FoTVgUO3KuHNsCcClIZRwCXRMSsiJgNDCITkJZakuxfEhGjgIXAxqvYn2JgC0lNImJmREwo45j9gckRcW9EFEbEg8BE4IASx9wZEZMi4kfgETIBd6Ui4k1gTUkbkwng95RxzH0R8V3S5vVAIyo+z7siYkLymSWl6ltM5vt4A3AfcEpETKugPrNyOXCvHr4D2i5NVaxEJ5YfLX6VlC2ro1TgXww0q2xHImIRmRTFCcBMSc9I2iSL/iztU+cS779Zhf7cCwwEdqeMv0AknSnp0yQ9M4/MXxnlpWAAvi5vZ0S8A3wOiMwvGLMqceBePbwF/Az0KeeYGWQuMi61DiumEbK1CFijxPsOJXdGxOiI2BvoSGYUfXsW/Vnap+mr2Kel7gVOAkYlo+FlklTG2cAhQOuIaAXMJxNwAVaW3ig37SHpZDIj9xlJ/WZV4sC9GoiI+WQuIA6R1EfSGpIaSOop6ZrksAeBCyS1Sy7yXUjmT/tV8QHwO0nrJBdGz1u6Q1J7Sb2TXPfPZFIuxWXUMQrYKJnCWF/SocBmwNOr2CcAIuILYFcyOf3SmgOFZGag1Jd0IdCixP5vgXUrM3NE0kbAZcAfyaRMzpZUbkrHrCIO3KuJJF97OpkLjrPJ/Hk/kMxMC8gEl3eBD4GPgHFJ2aq09QLwcFLXeywfbOsl/ZgBzCUTRE8so47vgF5kLu59R2ak2isi5qxKn0rV/XpElPXXxGjgOTJTBL8CfmL5NMjSm4u+kzSuonaS1NR9wNURMT4iJpOZmXLv0hk7ZqtCvrhtZpYuHnGbmaWMA7eZWco4cJuZpYwDt5lZypR3Q0ZeLZnzua+a2go223SVFx+0Omzy7PeqvPZLZWJOg7br53WtGY+4zcxSptaOuM3MalRxUb57kDUHbjMzgKLavJz88hy4zcyAiLJWXqidHLjNzACKHbjNzNLFI24zs5TxxUkzs5TxiNvMLF3Cs0rMzFLGFyfNzFLGqRIzs5TxxUkzs5TxiNvMLGV8cdLMLGV8cdLMLF0inOM2M0sX57jNzFLGqRIzs5TxiNvMLGWKluS7B1lz4DYzA6dKzMxSx6kSM7OU8YjbzCxlHLjNzNIlfHHSzCxlnOM2M0uZFKVK6uW7A2ZmtUIUZ7+VQ1JjSWMkjZc0QdKgpFySLpc0SdKnkk4tUT5Y0hRJH0rapqKuesRtZgbVOeL+GdgjIhZKagC8LulZYFNgbWCTiCiWtFZyfE+ga7JtD9yWfF0pB24zM6i2HHdEBLAwedsg2QI4ETg8ItNQRMxKjukN3JN87m1JrSR1jIiZK2vDqRIzM4DCwqw3SQMkvVtiG1CyKkkFkj4AZgEvRMQ7wAbAocnxz0rqmhzeGfi6xMenJWUr5RG3mRlUasQdEcOAYeXsLwK2ktQKeEzSFkAj4KeI2E5SX+AOYJdV6apH3GZmkMlxZ7tlKSLmAS8DPciMpEcmux4DtkxeTyeT+16qS1K2Ug7cZmZQnbNK2iUjbSQ1AfYGJgKPA7snh+0KTEpePwkclcwu2QGYX15+G5wqMTPLqL5ZJR2BuyUVkBkcPxIRT0t6Hbhf0mlkLl7+KTl+FLAfMAVYDBxbUQMO3GZmUJ2zSj4Eti6jfB6wfxnlAZxcmTYcuM3MIDNjJCUcuM3MACLy3YOsOXCbmUGq1ipx4DYzAwduM7PU8bKuZmYpU1SU7x5kzYHbzAycKjEzSx0HbjOzlHGO28wsXaLY87jNzNLFqRIzs5TxrBIzs5TxiNvMLGUcuK2oqIhD+5/KWu3acuu1g5bb98svv3DepdfzyX8n06plC6675Dw6d2xfpfamzfiGsy66innzF7DZxl256sIzadCgAXc/NJIRTz1HQUEBa7ZqyaXnn0anDlVry1bNlTddyO5778J3c+ay/+8OXWH/gX/oyfGnHI0kFi1cxEVnX8nECZOr1GbDhg24ZsglbNFtU+bNnc9fjj+X6V/PZOddt+fMv59CgwYNWLJkCVdffBNvvz62Sm2lXooWmfITcHLkvkefYP111ylz38inn6dF82Y8+8gdHHloH2649Y6s6338mRcYMvy+FcpvvC1T17OP3EGL5s0Y8fRoADbtugEPDx/MY/fcxt67/5brh2TfllWvkQ89xXH9Tlnp/q+nTueI3sfTa9dDGXLDv7js+guyrrvz2h257/GhK5QfdEQfFsxbwF7d+3DnP+/nrAtPBeD7ufP48xF/pdeuh3L2wIu49tZLKn9CdU0OHl2WKzkL3JI2kXSOpMHJdo6kTXPVXm3yzazZvPrmGP5wwL5l7v/Pa2/Re7+9ANhnt114570PiAiKioq47pZ/cWj/U/n9USfyyOOjsmovInjnvfHss1vmuaO999uL/7z6FgDdt+1Gk8aNAei2+SZ8O3tOVU/PVtHYt95n/vfzV7r//bEfsmD+DwB88O5HtO+01rJ9Bx7Uk3+PvpsnX36AS687n3r1svunu1fPXRn58NMAPPfUS+y4S3cAPvnov8z6NvOzMHniZzRu3IiGDRus0nnVGcWR/ZZnOQncks4BHgIEjEk2AQ9KOjcXbdYmV980lNNP6o9U9rd31uzv6LBWWwDq1y+gWdM1mDd/ASOfHk3zZk15ePhgHv7XTfz7yeeYNuObCtubN38BzZs1pX79AgDat2vLrNnfrXDcyKeeZ5cdtqvCmVlNOfiIPrz60psAbNB1Xfbvsw/99u/PgbsfTlFRMQce1DOretp3aMc3078FMum7hQsW0nrNVssd0+OAPZnw4UR++WVJ9Z5E2hQVZb/lWa5y3P2BzSNiuZ8ESTcAE4CryvqQpAHAAIBbr7+MPx11WI66lzuvvPEOa7ZuxeabdGXMuA8r9dk3x4xj0mdf8vzLrwOwcNEivvp6Os2arkH/U88DYP4PP7BkSeGyEfWVF55JuzZrVlj3U6P/w4SJk7hryDWVPCOradvvvB0HH9Gbfr36A7Dj77qzebdNGfnCPQA0atyI7+bMBWDIXdex9q860aBBAzp26cCTLz8AwN3DHmTEg09V2NaGG6/PWX8/lWMPqdSTs+qkqAUpkGzlKnAXA52Ar0qVd0z2lSkihgHDAJbM+Tz/f4+sgvc//IRXXn+b194ay8+/LGHRosWcM+garr7o7GXHrNWuDd/MmkOHtdpRWFjEwkWLadWyBRFw/mknsvP2265Q74i7hwCZHPf0b77l5P5/XLYvIvhh4SIKC4uoX7+Ab2fPYa12bZbtf2vs+wy7+yHuGnINDRs2zOHZW1VtvNmGXHHj3+nf7xTmJWkVSTz28NNcf9ktKxx/8jFnApkc99U3X8wf+/x5uf3ffjObDp3b883MWRQUFNCsRTO+nzsPgA4d1+LWu6/jrIEXMvXLaTk+sxSoBSmQbOUqx/1X4CVJz0oalmzPAS8Bf8lRm7XCaScey0uP38fzI+7m2kHn0n3bbssFbYDdf7sDT4x6EYDnX3mN7bfthiR23n4bHn7sGZYkz777cuo0Fv/4U4VtSqL7Nlvy/CuvAfDEqBfZY5cdAfh00hQGXTOYW66+iDatW5VXjeVZx84dGHLXdZx58t/58vOpy8rfenUMPQ7YkzXbtgagZasWdOrSIas6X3ru/+h7aC8gkxJZOnOkeYtmDHvgJq679GbGjRlfzWeSUlGc/ZZnORlxR8RzkjYCugOdk+LpwNiIyH+CKA9uuf0eNt9kI3bfZQf69tqX8y69lp6HHEfLFs25dlAm7f+HA3owfeYsDjn2FCKC1q1aMviqC7Oq/7QTj+Osi67i5mH3sOlGG9C31z4AXD9kOIt//InTL7gCgI7t23HLNRfn5BytfDcOvZzuO29H6zVb8dr4Udx0zVAa1M/8E3zw7hEMPPN4WrVuyaBrMj8PhYVF9N37SKZM+oIbr7yVux4dglSPwsJCBp1zFTOmVXz949H7n+C6Wy/lxTGPM+/7+Zw24HwAjvzTofxqvbUZeObxDDzzeACOOfhk5s75PkdnnwIpGnErauncxbSmSiy3Ntv04Hx3wWqhybPfU1XrWHRhv6xjTtNLHqpye1XhG3DMzKBWpECy5cBtZgapSpU4cJuZ4emAZmbp4xG3mVnKOHCbmaVMLbiVPVsO3GZm+JmTZmbp48BtZpYyKZpV4gcpmJlBta3HLamxpDGSxkuaIGlQUr6epHckTZH0sKSGSXmj5P2UZP+6FXXVgdvMDKrzQQo/A3tERDdgK6CHpB2Aq4EbI2JD4Hsyy1+TfP0+Kb8xOa5cDtxmZkAUFWe9lVtPxsLkbYNkC2AP4N9J+d1An+R17+Q9yf49JZW7FooDt5kZVOujyyQVSPoAmAW8AHwGzIuIwuSQafxv5dTOwNcAyf75QBvK4cBtZkZmOmC2m6QBkt4tsQ1Yrq6IoojYCuhCZnnrTaqzr55VYmYGlZoOWPJpXRUcN0/Sy8COQCtJ9ZNRdRcyzygg+bo2ME1SfaAlsOJDY0vwiNvMDDIPVcx2K4ekdpJaJa+bAHsDnwIvAwclhx0NPJG8fjJ5T7L/P1HBgxI84jYzA6Kw2uZxdwTullRAZnD8SEQ8LekT4CFJlwHvA8OT44cD90qaAswF+lXUgAO3mRlUOJLOVkR8CGxdRvnnZPLdpct/Air1aCcHbjMzvFaJmVn6pOeOdwduMzPwiNvMLH084jYzS5dl9zSmgAO3mRkQKRpxV3gDjqS+kponr8+V9IikrXLfNTOzGlRNN+DUhGzunLw4In6QtBOwH3A/8M/cdsvMrGZFcfZbvmUTuJc+QbMXMDQingAa5a5LZmY1L02BO5sc90xJQ4AewHbJUxu8xomZ1SlRVO4S2LVKNgH4EOD/gP0j4nugLXBuTntlZlbD6sSIW1KLEm+fK1G2EHgjx/0yM6tRUZyeEXd5qZIJZB63U/Jslr4PYJ0c9svMrEbVhpF0tlYauCNi7ZrsiJlZPkWkZ8Sd1UVGSf0knZ+87iJp29x2y8ysZqUpx53NDTi3ALsDRyZFi/E8bjOrY4qLlPWWb9lMB9wpIraR9D5ARMxNpgSamdUZdeXi5FJLJNUjc0ESSW2oFTd9mplVnzQF7mxy3EOAEUA7SYOA14Grc9orM7MaFpH9lm8Vjrgj4h5J7wF7JUUHR8THue2WmVnNStOIO9tlXQuAJWTSJb7d3czqnDo1HVDS34AHgU5AF+ABSeflumNmZjWpqEhZb/mWzYj7KGDriFgMIOly4H3gylx2zMysJqVpxJ3V6oCljquflJmZ1Rl1Isct6UYyOe25wARJo5P3+wBja6Z7ZmY1ozbMFslWeSPupTNHJgDPlCh/O3fdMTPLjzox4o6I4TXZETOzfCoqTs+EuQpz3JI2AC4HNgMaLy2PiI1y2C8zsxqVplRJNr9i7gLuJLMOd0/gEeDhHPbJzKzGFYey3vItm8C9RkSMBoiIzyLiAjIB3MyszohQ1lu+ZTMd8OdkkanPJJ0ATAea57ZbZmY1K02pkmwC92lAU+BUMrnulsBxuewUQJNOu+S6CUuhczrtmu8uWB1VG1Ig2cpmkal3kpc/8L+HKZiZ1Sl1YlaJpMdI1uAuS0T0zUmPzMzyoLoyJZLWBu4B2ifVDouIm0rsPwO4DmgXEXMkCbgJ2I/ME8aOiYhx5bVR3oj7lir238wsNaoxVVIInBER4yQ1B96T9EJEfJIE9X2AqSWO7wl0TbbtgduSrytV3g04L1W192ZmaVFds0UiYibJek4R8YOkT4HOwCfAjcDZwBMlPtIbuCciAnhbUitJHZN6ypSepI6ZWQ4VV2KTNEDSuyW2AWXVKWldYGvgHUm9gekRMb7UYZ2Br0u8n5aUrVS2D1IwM6vTguxH3BExDBhW3jGSmpF57ONfyaRPzieTJqmyrAO3pEYR8XN1NGpmVtsUVuN0QEkNyATt+yNipKRfA+sB4zPXIukCjJPUncy9MWuX+HiXpGylsnkCTndJHwGTk/fdJN28KidjZlZbBcp6K08yS2Q48GlE3AAQER9FxFoRsW5ErEsmHbJNRHwDPAkcpYwdgPnl5bchuxz3YKAX8F3SgfHA7ll8zswsNSqT467AzmTuedlD0gfJtl85x48CPgemALcDJ1XUQDapknoR8VUyvF+qKIvPmZmlRmVy3OXWE/E6lF9ZMupe+jqAkyvTRjaB++skDxOSCoBTgEmVacTMrLbLYiRda2QTuE8kky5ZB/gWeDEpMzOrM4qqacRdE7JZq2QW0K8G+mJmljcpenJZVk/AuZ0ybuOPiDInnJuZpVFxXRpxk0mNLNUY+D3L3+VjZpZ6KVqOO6tUyXKPKZN0L/B6znpkZpYHde3iZGnrkVmu0MyszihWHUqVSPqe//0VUQ+YC5yby06ZmdW0NN2cUm7gTm7d7Mb/7psvTiaLm5nVKWmaVVLuLe9JkB4VEUXJ5qBtZnVSMcp6y7ds1ir5QNLWOe+JmVkeRSW2fCvvmZP1I6KQzCLgYyV9Biwicw9+RMQ2NdRHM7OcS1OqpLwc9xhgG+DAGuqLmVne1JXpgAKIiM9qqC9mZnlTVEdG3O0knb6ynUsXCDczqwvqyoi7AGhGBevKmpnVBXUlcM+MiEtqrCdmZnlUjY+czLkKc9xmZquDujLi3rPGemFmlmd14pb3iJhbkx0xM8unujKP28xstVFXUiVmZqsNB24zs5SpDWuQZMuB28wM57jNzFKnTswqMTNbnRSnKFniwG1mhi9OmpmlTnrG2w7cZmaAR9xmZqlTqPSMuR24zcxwqsTMLHXSlCrJ5invZmZ1XjGR9VYRSXdImiXp4xJlW0l6W9IHkt6V1D0pl6TBkqZI+lBShQ9id+A2MyOTKsl2y8JdQI9SZdcAgyJiK+DC5D1AT6Brsg0AbquocgduMzMyqZJst4pExKtA6aWxA2iRvG4JzEhe9wbuiYy3gVaSOpZXv3PcZmZAUSUuT0oaQGZ0vNSwiBhWwcf+CoyWdB2ZQfNOSXln4OsSx01LymaurCIHbjMzKndxMgnSFQXq0k4ETouIEZIOAYYDe1WyDsCpEjMzAKIS/62io4GRyetHge7J6+nA2iWO65KUrZQDt5kZ1ZvjXokZwK7J6z2AycnrJ4GjktklOwDzI2KlaRJwqqTW+cupx3PccYcREXz88UT6/+l0dtpxO66++u80bNiAceM+4vgBZ1BUlKZFKA2g7zUD2HiPrVn03QIG73vOCvsbt2jKH64dwJrrtKfw5yWMOHsosyZNq1KbBQ3rc9ANJ9J5i/VYPG8hDw0czLxpc9jgt1uw7zmHUdCggKIlRTx3xf18/tYnVWor7apzdUBJDwK7AW0lTQMuAo4HbpJUH/iJ/+XIRwH7AVOAxcCxFdXvEXct0qlTBwaefBzb77AfW229JwUFBRzWrw93DP8HR/zxJLbaek+mTp3GUUcenO+u2ioY9+9Xufvoq1e6f7eTezPzk6+4uee5PHrGbfS66Kis627VpS39H7pghfLtDtmNn+Yv4obdTueN4c+y77mHAbD4+x+4t/+13NzjXP59xm0cfONJlT+hOqY6pwNGxGER0TEiGkREl4gYHhGvR8S2EdEtIraPiPeSYyMiTo6IDSLi1xHxbkX1O3DXMvXr16dJk8YUFBSwRpMmLFr8I7/88guTJ38OwIsvvkrf3++X517aqvhyzEQWz1+40v1rde3MZ29OAGDOZzNo1aUdTdtmZo9167MzJz5+KQNHXUHvK/qjetk9rmXTfbZj3IjXAJgw6h022GkLAGZO+IofZs0DYNakadRv3JCChqv3H+CFRNZbvjlw1yIzZnzDDTf+ky8+G8O0qe8zf8ECHn30SerXr8+222wJQN+++9Nl7U557qnlwsxPp7J5j98A0KXbBrTq3JaWHdrQboNObNlrR4YedDG37Hc+UVRMtz6/zarOFu1bM3/GdwAUFxXz0w+LWaN18+WO2bxnd2Z8/CVFvxRW7wmlTA1cnKw2Nf4rVtKxEXHnSvYtmxupgpbUq9e0RvuWb61ateTAA/Zlw412YN68BTz80FAOP7wvR/zxJK6/7mIaNWrICy++SlFRmlZVsGy9etuT7H/RUQwcdQXfTPyamRO+JIqL2WDnLej06/U46clLAajfqCELv1sAwBFDT6P12u0oaFCflp3aMnDUFQC8eedoxj36fxW2uVbXzux77mHcdeSVuTuxlEjTv6p8/G00CCgzcJecG1m/Yef8/1qrYXvuuQtffDmVOXMyN1w99viz7LjDdjzwwEh226MvAHvv9Tu6dl0/n920HPl54Y+MPGvosvdnvn4Tc6fO4le/2Zj3R7zK89c8vMJn7v/zjUAmx/2H605geL/Lltu/4NvvadmpDQu+mUu9gno0br4Gi7//AYAWHdbkiKGn8+/Tb2Pu1Fk5PLN0qA0j6WzlJFWSLJRS1vYR0D4XbdYFX0+dzvbbb0OTJo0B2GP33zJx4mTatWsDQMOGDTnrzJMZNuzefHbTcqRxizUoaFAAwHb9dufLdyby88If+eyNCWzec3uatsnku5u0bEqrzm2zqvPTF95jmz/sAsDm+23P50kOvXGLNTjqzrMYffVDTH1vUg7OJn1qYDpgtcnViLs9sC/wfalyAW/mqM3UGzP2fUaOfIaxY0ZTWFjIBx9M4PZ/3c+lg85mv/33ol69egwdeg8vv/JGvrtqq+CQwQNZf4dNWaN1c85+62ZeunHEskA95v6XaLdhZw667gQiglmTpzHy7NsBmD1lOi9e/wjH3nsuUj2KCot46sI7mTd9ToVtvvfIKxx0w0mc/soN/DhvEQ+dcjMAOxy1D21+1Z49/vJ79vjL7wG488irWJSkYFZHRZGeEbciB52VNBy4MyJeL2PfAxFxeEV1rI6pEqvYOZ12rfggW+1c/uUD2U2zKcfhv/p91jHnga8eq3J7VZGTEXdE9C9nX4VB28yspqUpx716T9w0M0vUhtx1thy4zcyo3lvec82B28wMp0rMzFInTbNKHLjNzHCqxMwsdXxx0swsZZzjNjNLGadKzMxSJhd3keeKA7eZGVDkEbeZWbo4VWJmljJOlZiZpYxH3GZmKePpgGZmKeNb3s3MUsapEjOzlHHgNjNLGc8qMTNLGY+4zcxSxrNKzMxSpijSs7CrA7eZGc5xm5mljnPcZmYpk6Ycd718d8DMrDYojsh6q4ikOyTNkvRxibJrJU2U9KGkxyS1KrHvPElTJP1X0r4V1e/AbWZGZsSd7X9ZuAvoUarsBWCLiNgSmAScByBpM6AfsHnymVslFZRXuQO3mRmZWSXZbhWJiFeBuaXKno+IwuTt20CX5HVv4KGI+DkivgCmAN3Lq9+B28yMyqVKJA2Q9G6JbUAlmzsOeDZ53Rn4usS+aUnZSvnipJkZlbs4GRHDgGGr0o6kvwGFwP2r8nlw4DYzA8jqomNVSToG6AXsGf+bOD4dWLvEYV2SspVyqsTMjGq/OLkCST2As4EDI2JxiV1PAv0kNZK0HtAVGFNeXR5xm5kBRVFUbXVJehDYDWgraRpwEZlZJI2AFyQBvB0RJ0TEBEmPAJ+QSaGcHFF+Zxy4zcyo3lveI+KwMoqHl3P85cDl2dbvwG1mhm95NzNLHS8yZWaWMjUxq6S6OHCbmZGuRaYcuM3M8IMUzMxSxzluM7OUcY7bzCxlPOI2M0sZz+M2M0sZj7jNzFLGs0rMzFLGFyfNzFLGqRIzs5TxnZNmZinjEbeZWcqkKcetNP2WWV1JGpA8nNRsGf9crL78zMl0GJDvDlit5J+L1ZQDt5lZyjhwm5mljAN3OjiPaWXxz8VqyhcnzcxSxiNuM7OUceA2M0sZB+5aTlIPSf+VNEXSufnuj+WfpDskzZL0cb77YvnhwF2LSSoAhgA9gc2AwyRtlt9eWS1wF9Aj352w/HHgrt26A1Mi4vOI+AV4COid5z5ZnkXEq8DcfPfD8seBu3brDHxd4v20pMzMVmMO3GZmKePAXbtNB9Yu8b5LUmZmqzEH7tptLNBV0nqSGgL9gCfz3CczyzMH7losIgqBgcBo4FPgkYiYkN9eWb5JehB4C9hY0jRJ/fPdJ6tZvuXdzCxlPOI2M0sZB24zs5Rx4DYzSxkHbjOzlHHgNjNLGQduW4GkIkkfSPpY0qOS1qhCXbtJejp5fWB5KxxKaiU0/515AAADQ0lEQVTppFVo42JJZ2ZbXuqYuyQdVIm21vWqfJZvDtxWlh8jYquI2AL4BTih5E5lVPpnJyKejIiryjmkFVDpwG22unHgtoq8BmyYjDT/K+ke4GNgbUn7SHpL0rhkZN4Mlq0hPlHSOKDv0ookHSPpluR1e0mPSRqfbDsBVwEbJKP9a5PjzpI0VtKHkgaVqOtvkiZJeh3YuKKTkHR8Us94SSNK/RWxl6R3k/p6JccXSLq2RNt/LqPOzSWNSfr7oaSulf/2mlWeA7etlKT6ZNYC/ygp6grcGhGbA4uAC4C9ImIb4F3gdEmNgduBA4BtgQ4rqX4w8H8R0Q3YBpgAnAt8loz2z5K0T9Jmd2ArYFtJv5O0LZnb/7cC9gN+k8XpjIyI3yTtfQqUvNtw3aSN/YF/JufQH5gfEb9J6j9e0nql6jwBuCkitgK2I7N6o1nO1c93B6xWaiLpg+T1a8BwoBPwVUS8nZTvQObhDm9IAmhI5jbsTYAvImIygKT7gAFltLEHcBRARBQB8yW1LnXMPsn2fvK+GZlA3hx4LCIWJ21ks37LFpIuI5OOaUZmGYGlHomIYmCypM+Tc9gH2LJE/rtl0vakEp97C/ibpC5kfjFMzqIfZlXmwG1l+TEZRS6TBOdFJYuAFyLisFLHLfe5KhJwZUQMLdXGX1ehrruAPhExXtIxwG4l9pVe9yGStk+JiJIBHknrLjso4gFJ75AZqY+S9OeI+M8q9M2sUpwqsVX1NrCzpA0BJDWVtBEwEVhX0gbJcYet5PMvAScmny2Q1BL4gcxoeqnRwHElcuedJa0FvAr0kdREUnMyaZmKNAdmSmoAHFFq38GS6iV9Xh/4b9L2icnxSNpIUtOSH5K0PvB5RAwGngC2zKIfZlXmEbetkoiYnYxcH5TUKCm+ICImSRoAPCNpMZlUS/MyqvgLMCxZ2a4IODEi3pL0RjLd7tkkz70p8FYy4l8I/DEixkl6GBgPzCKz/G1F/g68A8xOvpbs01RgDNACOCEifpL0LzK573HKND4b6FOqzkOAIyUtAb4BrsiiH2ZV5tUBzcxSxqkSM7OUceA2M0sZB24zs5Rx4DYzSxkHbjOzlHHgNjNLGQduM7OU+X8I5H4b3OdUIAAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "confusionMatrix(y_rumour_val, y_rumour_predictions)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Micro-F1-score 0.692\n",
      "Macro-F1-score 0.5791031006648775\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "confusionMatrix(y_fakenews_actual, y_fakenews_predictions)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "F1 macro score rumour model 0.7191447417279351\n",
      "F1 micro score rumour model 0.74\n",
      "F1 macro score fakenews model 0.5791031006648775\n",
      "F1 micro score fakenews model 0.692\n"
     ]
    }
   ],
   "source": [
    "print(\"F1 macro score rumour model\",f1_score(y_rumour_val, y_rumour_predictions, average='macro'))\n",
    "print(\"F1 micro score rumour model\",f1_score(y_rumour_val, y_rumour_predictions, average='micro'))\n",
    "\n",
    "print(\"F1 macro score fakenews model\",f1_score(y_fakenews_actual, y_fakenews_predictions, average='macro'))\n",
    "print(\"F1 micro score fakenews model\",f1_score(y_fakenews_actual, y_fakenews_predictions, average='micro'))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "input_2 (InputLayer)         (None, 100, 300)          0         \n",
      "_________________________________________________________________\n",
      "gru_2 (GRU)                  (None, 64)                70080     \n",
      "_________________________________________________________________\n",
      "dense_7 (Dense)              (None, 32)                2080      \n",
      "_________________________________________________________________\n",
      "dense_8 (Dense)              (None, 8)                 264       \n",
      "_________________________________________________________________\n",
      "dense_9 (Dense)              (None, 1)                 9         \n",
      "=================================================================\n",
      "Total params: 72,433\n",
      "Trainable params: 72,433\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "input_2 (InputLayer)         (None, 100, 300)          0         \n",
      "_________________________________________________________________\n",
      "gru_3 (GRU)                  (None, 64)                70080     \n",
      "_________________________________________________________________\n",
      "dense_10 (Dense)             (None, 32)                2080      \n",
      "_________________________________________________________________\n",
      "dense_11 (Dense)             (None, 16)                528       \n",
      "_________________________________________________________________\n",
      "dense_12 (Dense)             (None, 4)                 68        \n",
      "=================================================================\n",
      "Total params: 72,756\n",
      "Trainable params: 72,756\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "hidden_units1=64\n",
    "inp1=Input(shape=(100,300))\n",
    "h1=GRU(units=64)(inp1)\n",
    "h2=GRU(units=64)(inp1)\n",
    "\n",
    "o3=Dense(32,activation='relu')(h1)\n",
    "o3=Dense(8,activation='relu')(o3)\n",
    "o3=Dense(1,activation='sigmoid')(o3)\n",
    "\n",
    "o4=Dense(32,activation='relu')(h2)\n",
    "o4=Dense(16,activation='relu')(o4)\n",
    "o4=Dense(4,activation='softmax')(o4)\n",
    "\n",
    "rumour_model1=Model(inputs=[inp1],outputs=[o3])\n",
    "rumour_model1.summary()\n",
    "\n",
    "fake_model1=Model(inputs=[inp1],outputs=[o4])\n",
    "fake_model1.summary()\n",
    "\n",
    "rumour_model1.compile(optimizer=\"Adagrad\",loss=\"binary_crossentropy\",metrics=[f1])\n",
    "fake_model1.compile(optimizer=\"Adagrad\",loss=\"categorical_crossentropy\",metrics=[f1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 4531 samples, validate on 800 samples\n",
      "Epoch 1/3\n",
      "4531/4531 [==============================] - 20s 4ms/step - loss: 0.5932 - f1: 0.1662 - val_loss: 0.6088 - val_f1: 0.3080\n",
      "Epoch 2/3\n",
      "4531/4531 [==============================] - 18s 4ms/step - loss: 0.5326 - f1: 0.5594 - val_loss: 0.5286 - val_f1: 0.5775\n",
      "Epoch 3/3\n",
      "4531/4531 [==============================] - 18s 4ms/step - loss: 0.5123 - f1: 0.6074 - val_loss: 0.5405 - val_f1: 0.5324\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.callbacks.History at 0x7f9ac0376860>"
      ]
     },
     "execution_count": 41,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "rumour_model1.fit(x_rumour_train,y_rumour_train,batch_size=32,epochs=3,validation_data=(x_rumour_val,y_rumour_val ))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 17000 samples, validate on 3000 samples\n",
      "Epoch 1/8\n",
      "17000/17000 [==============================] - 70s 4ms/step - loss: 1.1107 - f1: 0.3146 - val_loss: 1.0934 - val_f1: 0.4434\n",
      "Epoch 2/8\n",
      "17000/17000 [==============================] - 69s 4ms/step - loss: 1.0493 - f1: 0.4072 - val_loss: 1.0351 - val_f1: 0.4733\n",
      "Epoch 3/8\n",
      "17000/17000 [==============================] - 68s 4ms/step - loss: 1.0197 - f1: 0.4466 - val_loss: 1.0061 - val_f1: 0.4443\n",
      "Epoch 4/8\n",
      "17000/17000 [==============================] - 70s 4ms/step - loss: 0.9903 - f1: 0.4796 - val_loss: 0.9835 - val_f1: 0.4895\n",
      "Epoch 5/8\n",
      "17000/17000 [==============================] - 68s 4ms/step - loss: 0.9653 - f1: 0.5125 - val_loss: 0.9639 - val_f1: 0.5105\n",
      "Epoch 6/8\n",
      "17000/17000 [==============================] - 68s 4ms/step - loss: 0.9433 - f1: 0.5309 - val_loss: 0.9631 - val_f1: 0.5368\n",
      "Epoch 7/8\n",
      "17000/17000 [==============================] - 69s 4ms/step - loss: 0.9267 - f1: 0.5475 - val_loss: 0.9543 - val_f1: 0.5532\n",
      "Epoch 8/8\n",
      "17000/17000 [==============================] - 69s 4ms/step - loss: 0.9100 - f1: 0.5603 - val_loss: 0.9353 - val_f1: 0.5569\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.callbacks.History at 0x7f9722fbdf60>"
      ]
     },
     "execution_count": 42,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "fake_model1.fit(x_fakenews_train,y_fakenews_train,batch_size=32,epochs=8,validation_data=(x_fakenews_val,y_fakenews_val))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(800,)\n",
      "(800,)\n"
     ]
    }
   ],
   "source": [
    "y_rumour_predictions1 = rumour_model1.predict(x_rumour_val)\n",
    "# print(y_rumour_predictions.shape)\n",
    "for i in range(0,y_rumour_predictions1.shape[0]):\n",
    "    if y_rumour_predictions1[i]>0.5:\n",
    "        y_rumour_predictions1[i]=1\n",
    "    else :\n",
    "        y_rumour_predictions1[i]=0\n",
    "# np.squeeze(y_rumour_predictions)\n",
    "y_rumour_predictions1 = y_rumour_predictions1.squeeze()\n",
    "print(y_rumour_predictions1.shape)\n",
    "# y_rumour_val= y_rumour_val.reshape((y_rumour_predictions.shape[0],1))\n",
    "print(y_rumour_val.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "No of correctly predicted points: 594\n",
      "Accuracy: 74.25\n"
     ]
    }
   ],
   "source": [
    "accuracy=0\n",
    "for i in range(0,y_rumour_val.shape[0]):\n",
    "    if(y_rumour_val[i]==y_rumour_predictions1[i]):\n",
    "        accuracy+=1  \n",
    "print(\"No of correctly predicted points:\",accuracy)\n",
    "print(\"Accuracy:\", (accuracy*100)/y_rumour_val.shape[0]) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(3000, 4)\n",
      "[[0.09430002 0.02941806 0.00711679 0.8691651 ]\n",
      " [0.34704942 0.28350863 0.05545031 0.31399164]\n",
      " [0.23562951 0.05481791 0.03514959 0.67440295]\n",
      " ...\n",
      " [0.27069992 0.04140913 0.01594665 0.6719443 ]\n",
      " [0.15602566 0.0721337  0.01593032 0.7559103 ]\n",
      " [0.41761976 0.01490598 0.00218489 0.5652894 ]]\n",
      "(3000,)\n",
      "[3 0 3 ... 3 3 3]\n",
      "(3000,)\n",
      "[3 1 3 ... 3 3 0]\n"
     ]
    }
   ],
   "source": [
    "y_fakenews_predictions1 = fake_model1.predict(x_fakenews_val)\n",
    "print(y_fakenews_predictions1.shape)\n",
    "print(y_fakenews_predictions1)\n",
    "y_fakenews_predictions1=np.argmax(y_fakenews_predictions1,axis=1)\n",
    "print(y_fakenews_predictions1.shape)\n",
    "print(y_fakenews_predictions1)\n",
    "y_fakenews_actual=np.argmax(y_fakenews_val,axis=1)\n",
    "print(y_fakenews_actual.shape)\n",
    "print(y_fakenews_actual)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "No of correctly predicted points: 1770\n",
      "Accuracy: 59.0\n"
     ]
    }
   ],
   "source": [
    "accuracy=0\n",
    "for i in range(0,y_fakenews_actual.shape[0]):\n",
    "    if(y_fakenews_actual[i]==y_fakenews_predictions1[i]):\n",
    "        accuracy+=1  \n",
    "print(\"No of correctly predicted points:\",accuracy)\n",
    "print(\"Accuracy:\", (accuracy*100)/y_fakenews_actual.shape[0]) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Micro-F1-score 0.7425\n",
      "Macro-F1-score 0.6840878419825789\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "confusionMatrix(y_rumour_val, y_rumour_predictions1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Micro-F1-score 0.59\n",
      "Macro-F1-score 0.4683010896359845\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "confusionMatrix(y_fakenews_actual, y_fakenews_predictions1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "F1 macro score rumour model 0.6840878419825789\n",
      "F1 micro score rumour model 0.7425\n",
      "F1 macro score fakenews model 0.4683010896359845\n",
      "F1 micro score fakenews model 0.59\n"
     ]
    }
   ],
   "source": [
    "print(\"F1 macro score rumour model\",f1_score(y_rumour_val, y_rumour_predictions1, average='macro'))\n",
    "print(\"F1 micro score rumour model\",f1_score(y_rumour_val, y_rumour_predictions1, average='micro'))\n",
    "\n",
    "print(\"F1 macro score fakenews model\",f1_score(y_fakenews_actual, y_fakenews_predictions1, average='macro'))\n",
    "print(\"F1 micro score fakenews model\",f1_score(y_fakenews_actual, y_fakenews_predictions1, average='micro'))\n"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}