TensorFlow-BoW-LSTM-Simple-Model.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/anaconda3/lib/python3.6/site-packages/h5py/__init__.py:36: FutureWarning: Conversion of the second argument of issubdtype from `float` to `np.floating` is deprecated. In future, it will be treated as `np.float64 == np.dtype(float).type`.\n",
      "  from ._conv import register_converters as _register_converters\n"
     ]
    }
   ],
   "source": [
    "%matplotlib inline\n",
    "import matplotlib\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import pandas as pd \n",
    "import re\n",
    "import collections \n",
    "import tensorflow as tf\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.feature_extraction.text import CountVectorizer\n",
    "from sklearn.feature_extraction.text import TfidfTransformer\n",
    "from sklearn.linear_model import LogisticRegression\n",
    "from sklearn.metrics import log_loss"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Data gathering and data cleaning"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "data = pd.read_csv('/Users/mohsenkiskani/.kaggle/competitions/cmps242-spring18-hw5/train.csv')\n",
    "test_data = pd.read_csv('/Users/mohsenkiskani/.kaggle/competitions/cmps242-spring18-hw5/test.csv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "def string_from_list(A):\n",
    "    if len(A)>0:\n",
    "        B = (\" \".join(A)).lower()\n",
    "        return re.sub('[^a-zA-z0-9\\s]','',B)\n",
    "    else:\n",
    "        return None "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "data['hashtags'] =  data['tweet'].apply((lambda x: re.findall(r\"#(\\w+)\", x)))\n",
    "test_data['hashtags'] =  test_data['tweet'].apply((lambda x: re.findall(r\"#(\\w+)\", x)))\n",
    "\n",
    "trump_tags = [\"littlemarco\", \"makeamericagreatagain\", \"trump2016\", \"crookedhillary\", \"votetrumpsc\", \n",
    "              \"votetrump\", \"trumppence16\", \"trumppence2016\", \"americafirst\", \"votetrumpma\", \"neverhillary\",\n",
    "              \"trump4president\", \"lightweightsenatormarcorubio\", \"trumpinstagram\", \"americagreatagain\", \n",
    "              \"trump4vets\", \"votetrumpmi\", \"votetrumphi\", \"votetrumpnh\", \"caucusfortrump\", \"trumpforpresident\",\n",
    "              \"makeyoutubegreatagain\", \"votetrumpms\", \"votetrumpil\", \"votetrumpny\", \"votetrumpks\", \"trumppence2016\",\n",
    "             ]\n",
    "\n",
    "trump_list   = []\n",
    "clinton_list = []\n",
    "\n",
    "for i in range(test_data.shape[0]):\n",
    "    test_data['hashtags'][i] = [x.lower() for x in test_data['hashtags'][i]]\n",
    "    \n",
    "    for item in trump_tags:\n",
    "        if item in test_data['hashtags'][i]:\n",
    "            trump_list.append(i)\n",
    "            break\n",
    "        \n",
    "    if \"lovetrumpshate\" in test_data['hashtags'][i] or \"nevertrump\" in test_data['hashtags'][i]:\n",
    "        clinton_list.append(i)\n",
    "        \n",
    "trump_list   = set(trump_list)\n",
    "clinton_list = set(clinton_list)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "all_hashtags_trump   = []\n",
    "all_hashtags_clinton = []\n",
    "\n",
    "for i in range(data.shape[0]):\n",
    "    data['hashtags'][i] = [x.lower() for x in data['hashtags'][i]]\n",
    "    if data['handle'][i]=='realDonaldTrump':\n",
    "        all_hashtags_trump.extend(data['hashtags'][i])\n",
    "    else:\n",
    "        all_hashtags_clinton.extend(data['hashtags'][i])\n",
    "\n",
    "all_hashtags_trump_set = set(all_hashtags_trump)\n",
    "all_hashtags_clinton_set = set(all_hashtags_clinton)\n",
    "#print(all_hashtags_trump_set)\n",
    "#print(all_hashtags_clinton_set)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "data['handle']    = data['handle'].replace({'realDonaldTrump':1, 'HillaryClinton':0})\n",
    "\n",
    "data['tweet']     = data['tweet'].apply((lambda x: re.sub(r\"http\\S+\", 'urlurlurlurl', x)))\n",
    "data['tweet']     = data['tweet'].apply(lambda x: x.lower())\n",
    "data['tweet']     = data['tweet'].apply((lambda x: re.sub('[^a-zA-z0-9\\s]','',x)))\n",
    "data['tweet']     = data['tweet'].apply((lambda x: re.sub(' +',' ',x)))\n",
    "data['tweet']     = data['tweet'].apply((lambda x: re.sub(r'(\\n+)(?=[A-Z])', r' ', x)))\n",
    "\n",
    "test_data['tweet']     = test_data['tweet'].apply((lambda x: re.sub(r\"http\\S+\", 'urlurlurlurl', x)))\n",
    "test_data['tweet']     = test_data['tweet'].apply(lambda x: x.lower())\n",
    "test_data['tweet']     = test_data['tweet'].apply((lambda x: re.sub('[^a-zA-z0-9\\s]','',x)))\n",
    "test_data['tweet']     = test_data['tweet'].apply((lambda x: re.sub(' +',' ',x)))\n",
    "test_data['tweet']     = test_data['tweet'].apply((lambda x: re.sub(r'(\\n+)(?=[A-Z])', r' ', x)))\n",
    "\n",
    "\n",
    "all_tweets = pd.concat([data['tweet'], test_data['tweet']])\n",
    "\n",
    "X       = data['tweet']\n",
    "Y       = np.array(data['handle'].tolist())\n",
    "X_test  = test_data['tweet']\n",
    "\n",
    "X_train_bow_pd, X_val_bow_pd, Y_train_bow_pd, Y_val_bow_pd = train_test_split(X, Y,  test_size=0.1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Bag of word's model, feature selection"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "count_vec         = CountVectorizer()\n",
    "tfidf_transformer = TfidfTransformer()\n",
    "\n",
    "X_train_counts    = count_vec.fit_transform(X_train_bow_pd)\n",
    "X_train_tfidf     = tfidf_transformer.fit_transform(X_train_counts)\n",
    "\n",
    "X_val_counts      = count_vec.transform(X_val_bow_pd)\n",
    "X_val_tfidf       = tfidf_transformer.transform(X_val_counts)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "hyperVec                  = [1e6,1e5,1e4,1e3,1e2,1e1,1e0,1e-1,1e-2]\n",
    "loss_dic_bow              = {}\n",
    "\n",
    "for C in hyperVec:\n",
    "    lf_tmp = LogisticRegression(C=C)\n",
    "    lf_tmp.fit(X_train_tfidf, Y_train_bow_pd)\n",
    "    y_val_predicted_tmp = lf_tmp.predict(X_val_tfidf)\n",
    "    loss_dic_bow[C] = log_loss(y_pred = y_val_predicted_tmp, y_true =  Y_val_bow_pd)\n",
    "\n",
    "    \n",
    "C_optimal = min(loss_dic_bow, key=loss_dic_bow.get)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_counts           = count_vec.fit_transform(X)\n",
    "X_tfidf            = tfidf_transformer.fit_transform(X_counts)\n",
    "\n",
    "X_test_counts      = count_vec.transform(X_test)\n",
    "X_test_tfidf       = tfidf_transformer.transform(X_test_counts)\n",
    "\n",
    "lf = LogisticRegression(C = C_optimal)\n",
    "lf.fit(X_tfidf, Y)\n",
    "Y_bow_predicted_test = lf.predict(X_test_tfidf)\n",
    "\n",
    "#print(len(Y_bow_predicted_test))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_tfidf_np      = X_tfidf.toarray()\n",
    "X_test_tfidf_np = X_test_tfidf.toarray()\n",
    "X_train_tfidf, X_val_tfidf, Y_train_tfidf, Y_val_tfidf = train_test_split(X_tfidf_np,\\\n",
    "                                                                          Y, test_size = 0.1, random_state = 42)\n",
    "bow_dim_size    = X_tfidf.shape[1]\n",
    "#print(X_test_tfidf_np.shape)\n",
    "#print(X_tfidf_np.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_all = []\n",
    "for i in range(X.shape[0]):\n",
    "    X_all.append(X[i].split())\n",
    "    \n",
    "for i in range(X_test.shape[0]):\n",
    "    X_all.append(X_test[i].split())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "tweets         = all_tweets.tolist()\n",
    "words          = \" \".join(tweets)\n",
    "words          = \" \".join(words.split(\"\\n\")) \n",
    "words          = \" \".join(words.split(\"\\t\"))\n",
    "words          = \" \".join(words.split(\"\\xa0\")).split()\n",
    "whole_text     = \" \".join(words)\n",
    "\n",
    "count          = collections.Counter(words)\n",
    "count_clean    = [(item, count[item]) for item in count if count[item]>1]\n",
    "vocab_size     = len(count_clean)\n",
    "num_words      = vocab_size+2 \n",
    "\n",
    "dic = {}\n",
    "\n",
    "for i in range(vocab_size):\n",
    "    dic[count_clean[i][0]] = (i+1, count_clean[i][1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "def string_to_token_list(A):\n",
    "    B      = A.strip().split(\" \")\n",
    "    output = []\n",
    "    for item in B:\n",
    "        if item in dic:\n",
    "            output.append(dic[item][0])\n",
    "        else:\n",
    "            output.append(vocab_size + 1)\n",
    "    return output"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_tokens       = X.apply(lambda x: string_to_token_list(x))\n",
    "X_test_tokens  = X_test.apply(lambda x: string_to_token_list(x)) \n",
    "\n",
    "num_tokens     = [len(tokens) for tokens in X_tokens]\n",
    "num_tokens     = np.array(num_tokens)\n",
    "max_tokens     = int(np.max(num_tokens))\n",
    "\n",
    "X_pad          = []\n",
    "X_test_pad     = []\n",
    "\n",
    "for item in X_tokens:\n",
    "    n = len(item)\n",
    "    X_pad.append([0]*(max_tokens-n) + item)\n",
    "\n",
    "for item in X_test_tokens:\n",
    "    n = len(item)\n",
    "    X_test_pad.append([0]*(max_tokens-n) + item)\n",
    "    \n",
    "X_pad        = np.array(X_pad)\n",
    "X_test_pad   = np.array(X_test_pad)\n",
    "X_all_pad    = np.concatenate((X_pad , X_test_pad), axis=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train_pad, X_val_pad, Y_train_pad, Y_val_pad = train_test_split(X_pad, Y, test_size = 0.1, random_state = 42)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Hyper parameter selection"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 146,
   "metadata": {},
   "outputs": [],
   "source": [
    "lr                       = 1e-5\n",
    "\n",
    "lstm_dropout_prob        = 0.5\n",
    "lstm_forget_bias         = 1.0\n",
    "reg_const                = 1.0\n",
    "reg_const_bow            = 1.0\n",
    "\n",
    "embedding_size           = 64\n",
    "batch_size               = 128\n",
    "lstm_out                 = 30\n",
    "\n",
    "bow_hidden_layer_1_size  = 128\n",
    "bow_hidden_layer_2_size  = 2\n",
    "\n",
    "hidden_layer_1_size      = 32#lstm_out + bow_hidden_layer_2_size\n",
    "hidden_layer_2_size      = 32#lstm_out + bow_hidden_layer_2_size"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Word embedding initialization "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 147,
   "metadata": {},
   "outputs": [],
   "source": [
    "from gensim.models import Word2Vec\n",
    "\n",
    "model = Word2Vec(X_all, iter=100, min_count=2, size=embedding_size, workers=4)\n",
    "#words = list(model.wv.vocab)\n",
    "#print(words)\n",
    "#print(model.wv.similar_by_word('makeamericagreatagain'))\n",
    "#model.wv['trump'] "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 148,
   "metadata": {},
   "outputs": [],
   "source": [
    "embedding_matrix = np.zeros((1,embedding_size), np.float32)\n",
    "\n",
    "for i in range(vocab_size):\n",
    "    word             = count_clean[i][0]\n",
    "    word_vec         = model.wv[word]\n",
    "    embedding_matrix = np.vstack((embedding_matrix, word_vec))\n",
    "\n",
    "embedding_matrix = np.vstack((embedding_matrix, 5* np.ones((1,embedding_size), np.float32)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Utility functions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 149,
   "metadata": {},
   "outputs": [],
   "source": [
    "def evaluate_test_data_set(filename, X_test_pad, X_test_tfidf_np, sess):\n",
    "        test_size   = X_test_pad.shape[0]\n",
    "        test_remain = batch_size - (test_size % batch_size) \n",
    "        X_test_pad      = np.concatenate((X_test_pad, np.zeros((test_remain, max_tokens), np.float32)), axis=0)\n",
    "        X_test_tfidf_np = np.concatenate((X_test_tfidf_np, np.zeros((test_remain, bow_dim_size), np.float32)), axis=0)\n",
    "        m = test_size // batch_size \n",
    "        pred_test_vals = np.empty((0, 2), np.float32)\n",
    "\n",
    "        for i in range(m+1):\n",
    "            input_test_batch      = X_test_pad[i * batch_size : (i+1) * batch_size, :]\n",
    "            input_X_test_tfidf    = X_test_tfidf_np[i * batch_size : (i+1) * batch_size, :]\n",
    "            test_preds_list  = sess.run([predictions], feed_dict = {X_bow : input_X_test_tfidf,\n",
    "                                                                    X     : input_test_batch})\n",
    "            pred_test_batch  = np.asarray(test_preds_list).reshape(batch_size, 2)\n",
    "            pred_test_vals   = np.concatenate((pred_test_vals, pred_test_batch), axis=0)\n",
    "            \n",
    "        with open(filename,\"w+\") as outputfile:\n",
    "            outputfile.write(\"id,realDonaldTrump,HillaryClinton\\n\")\n",
    "            for j in range(test_size):\n",
    "                if j in trump_list:\n",
    "                    hillary = 0 \n",
    "                    donald  = 1 \n",
    "                elif j in clinton_list:\n",
    "                    hillary = 1 \n",
    "                    donald  = 0\n",
    "                else:\n",
    "                    hillary = pred_test_vals[j][1]\n",
    "                    donald  = pred_test_vals[j][0]\n",
    "                outputfile.write(str(j)+\",\"+str(donald)+\",\"+str(hillary)+\"\\n\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 150,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_batch(X, Y, X_tfidf, size):\n",
    "    rdm         = np.random.choice(X.shape[0], size , replace = False)\n",
    "    y1          = Y[rdm].reshape((size,1))\n",
    "    y2          = (y1+1)%2\n",
    "    Y_out       = np.concatenate((y1, y2), axis=1)\n",
    "    X_out       = X[rdm,:]\n",
    "    X_tfidf_out = X_tfidf[rdm,:] \n",
    "    return X_tfidf_out, X_out, Y_out"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Look at \n",
    "https://github.com/bernhard2202/twitter-sentiment-analysis/blob/master/model/lstm.py\n",
    "\n",
    "For a similar implementation"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Computational graph"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 151,
   "metadata": {},
   "outputs": [],
   "source": [
    "with tf.device(\"/cpu:0\"):\n",
    "    \n",
    "    tf.reset_default_graph()\n",
    "    tf.set_random_seed(0)\n",
    "\n",
    "    X_bow = tf.placeholder(tf.float32, shape=[batch_size, bow_dim_size])\n",
    "    X     = tf.placeholder(tf.int32, shape=[batch_size, max_tokens])\n",
    "    Y     = tf.placeholder(tf.float32, shape=[batch_size,2])\n",
    "        \n",
    "    embeddings     = tf.Variable(embedding_matrix)    \n",
    "    embed          = tf.nn.embedding_lookup(embeddings, X)\n",
    "    embed          = tf.unstack(embed, max_tokens, 1)\n",
    "    \n",
    "    lstm_cell                 = tf.contrib.rnn.BasicLSTMCell(lstm_out, forget_bias = lstm_forget_bias)\n",
    "    lstm_cell                 = tf.contrib.rnn.DropoutWrapper(lstm_cell, output_keep_prob=lstm_dropout_prob)\n",
    "    lstm_output , lstm_state  = tf.contrib.rnn.static_rnn(lstm_cell, inputs=embed , dtype=tf.float32)\n",
    "    \n",
    "    outputs                   = tf.stack(lstm_output)\n",
    "    outputs                   = tf.gather(outputs, max_tokens-1, axis=0)\n",
    "    outputs                   = tf.reshape(outputs, [batch_size, lstm_out])\n",
    "    lstm_final_output         = tf.reshape(outputs, [batch_size , lstm_out])\n",
    "    \n",
    "    bow_hidden_layer_1_output = tf.layers.dense(X_bow, bow_hidden_layer_1_size, \n",
    "                                               kernel_regularizer = tf.contrib.layers.l2_regularizer(reg_const_bow))\n",
    "    bow_hidden_layer_2_output = tf.layers.dense(bow_hidden_layer_1_output, bow_hidden_layer_2_size, \n",
    "                                               kernel_regularizer = tf.contrib.layers.l2_regularizer(reg_const_bow))\n",
    "    \n",
    "    bow_concat_lstm           = tf.concat( [lstm_final_output, X_bow], axis=1)\n",
    "    \n",
    "    hidden_layer_1_output     = tf.layers.dense(bow_concat_lstm, hidden_layer_1_size,\n",
    "                                                kernel_regularizer = tf.contrib.layers.l2_regularizer(reg_const) )\n",
    "    hidden_layer_2_output     = tf.layers.dense(hidden_layer_1_output, hidden_layer_2_size,\n",
    "                                                kernel_regularizer = tf.contrib.layers.l2_regularizer(reg_const))\n",
    "    \n",
    "    out_weight      = tf.Variable(tf.random_normal([hidden_layer_2_size, 2]))\n",
    "    out_bias        = tf.Variable(tf.random_normal([2]))\n",
    "    \n",
    "    scores          = tf.nn.xw_plus_b(hidden_layer_2_output, out_weight,out_bias)\n",
    "    predictions     = tf.nn.softmax(scores)\n",
    "    \n",
    "    losses          = tf.nn.softmax_cross_entropy_with_logits_v2(logits = scores,\n",
    "                                                              labels = Y)\n",
    "    loss            = tf.reduce_mean(losses)\n",
    "    \n",
    "    correct_pred    = tf.equal(tf.argmax(predictions, 1), tf.argmax(Y, 1))\n",
    "    accuracy        = tf.reduce_mean(tf.cast(correct_pred, tf.float32))\n",
    "    optimizer       = tf.train.AdamOptimizer(lr).minimize(loss)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Running the computational graph"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 154,
   "metadata": {
    "slideshow": {
     "slide_type": "-"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 0, Batch loss =  1.019,  training accuray =  0.484,  val loss =  0.898,  val accuray =  0.562, \n",
      "Epoch 10, Batch loss =  0.870,  training accuray =  0.570,  val loss =  1.041,  val accuray =  0.477, \n",
      "Epoch 20, Batch loss =  0.962,  training accuray =  0.516,  val loss =  1.016,  val accuray =  0.492, \n",
      "Epoch 30, Batch loss =  0.872,  training accuray =  0.586,  val loss =  1.001,  val accuray =  0.492, \n",
      "Epoch 40, Batch loss =  0.960,  training accuray =  0.508,  val loss =  0.838,  val accuray =  0.578, \n",
      "Epoch 50, Batch loss =  1.003,  training accuray =  0.461,  val loss =  0.943,  val accuray =  0.516, \n",
      "Epoch 60, Batch loss =  1.031,  training accuray =  0.453,  val loss =  0.943,  val accuray =  0.508, \n",
      "Epoch 70, Batch loss =  0.845,  training accuray =  0.570,  val loss =  0.988,  val accuray =  0.477, \n",
      "Epoch 80, Batch loss =  0.903,  training accuray =  0.523,  val loss =  0.927,  val accuray =  0.508, \n",
      "Epoch 90, Batch loss =  1.000,  training accuray =  0.438,  val loss =  0.888,  val accuray =  0.523, \n",
      "Epoch 100, Batch loss =  0.889,  training accuray =  0.508,  val loss =  0.902,  val accuray =  0.516, \n",
      "Epoch 110, Batch loss =  0.946,  training accuray =  0.469,  val loss =  0.866,  val accuray =  0.523, \n",
      "Epoch 120, Batch loss =  0.927,  training accuray =  0.477,  val loss =  0.903,  val accuray =  0.500, \n",
      "Epoch 130, Batch loss =  0.941,  training accuray =  0.477,  val loss =  0.861,  val accuray =  0.531, \n",
      "Epoch 140, Batch loss =  0.927,  training accuray =  0.477,  val loss =  0.874,  val accuray =  0.508, \n",
      "Epoch 150, Batch loss =  1.003,  training accuray =  0.414,  val loss =  0.877,  val accuray =  0.516, \n",
      "Epoch 160, Batch loss =  0.799,  training accuray =  0.562,  val loss =  0.904,  val accuray =  0.477, \n",
      "Epoch 170, Batch loss =  0.883,  training accuray =  0.484,  val loss =  0.805,  val accuray =  0.547, \n",
      "Epoch 180, Batch loss =  0.898,  training accuray =  0.477,  val loss =  0.926,  val accuray =  0.461, \n",
      "Epoch 190, Batch loss =  0.911,  training accuray =  0.445,  val loss =  0.812,  val accuray =  0.531, \n",
      "Epoch 200, Batch loss =  0.798,  training accuray =  0.523,  val loss =  0.791,  val accuray =  0.539, \n",
      "Epoch 210, Batch loss =  0.765,  training accuray =  0.547,  val loss =  0.769,  val accuray =  0.547, \n",
      "Epoch 220, Batch loss =  0.827,  training accuray =  0.508,  val loss =  0.879,  val accuray =  0.484, \n",
      "Epoch 230, Batch loss =  0.850,  training accuray =  0.477,  val loss =  0.812,  val accuray =  0.516, \n",
      "Epoch 240, Batch loss =  0.844,  training accuray =  0.477,  val loss =  0.829,  val accuray =  0.516, \n",
      "Epoch 250, Batch loss =  0.854,  training accuray =  0.469,  val loss =  0.754,  val accuray =  0.562, \n",
      "Epoch 260, Batch loss =  0.838,  training accuray =  0.484,  val loss =  0.837,  val accuray =  0.477, \n",
      "Epoch 270, Batch loss =  0.795,  training accuray =  0.516,  val loss =  0.854,  val accuray =  0.461, \n",
      "Epoch 280, Batch loss =  0.780,  training accuray =  0.555,  val loss =  0.781,  val accuray =  0.539, \n",
      "Epoch 290, Batch loss =  0.703,  training accuray =  0.570,  val loss =  0.787,  val accuray =  0.484, \n",
      "Epoch 300, Batch loss =  0.811,  training accuray =  0.477,  val loss =  0.806,  val accuray =  0.508, \n",
      "Epoch 310, Batch loss =  0.754,  training accuray =  0.539,  val loss =  0.714,  val accuray =  0.578, \n",
      "Epoch 320, Batch loss =  0.785,  training accuray =  0.523,  val loss =  0.780,  val accuray =  0.547, \n",
      "Epoch 330, Batch loss =  0.726,  training accuray =  0.531,  val loss =  0.734,  val accuray =  0.547, \n",
      "Epoch 340, Batch loss =  0.813,  training accuray =  0.477,  val loss =  0.678,  val accuray =  0.586, \n",
      "Epoch 350, Batch loss =  0.753,  training accuray =  0.547,  val loss =  0.750,  val accuray =  0.516, \n",
      "Epoch 360, Batch loss =  0.718,  training accuray =  0.523,  val loss =  0.766,  val accuray =  0.500, \n",
      "Epoch 370, Batch loss =  0.763,  training accuray =  0.531,  val loss =  0.784,  val accuray =  0.484, \n",
      "Epoch 380, Batch loss =  0.765,  training accuray =  0.477,  val loss =  0.702,  val accuray =  0.570, \n",
      "Epoch 390, Batch loss =  0.755,  training accuray =  0.508,  val loss =  0.738,  val accuray =  0.500, \n",
      "Epoch 400, Batch loss =  0.669,  training accuray =  0.578,  val loss =  0.709,  val accuray =  0.523, \n",
      "Epoch 410, Batch loss =  0.745,  training accuray =  0.492,  val loss =  0.725,  val accuray =  0.508, \n",
      "Epoch 420, Batch loss =  0.767,  training accuray =  0.445,  val loss =  0.725,  val accuray =  0.531, \n",
      "Epoch 430, Batch loss =  0.725,  training accuray =  0.500,  val loss =  0.703,  val accuray =  0.531, \n",
      "Epoch 440, Batch loss =  0.696,  training accuray =  0.562,  val loss =  0.787,  val accuray =  0.445, \n",
      "Epoch 450, Batch loss =  0.689,  training accuray =  0.547,  val loss =  0.696,  val accuray =  0.516, \n",
      "Epoch 460, Batch loss =  0.706,  training accuray =  0.492,  val loss =  0.735,  val accuray =  0.477, \n",
      "Epoch 470, Batch loss =  0.675,  training accuray =  0.586,  val loss =  0.659,  val accuray =  0.625, \n",
      "Epoch 480, Batch loss =  0.728,  training accuray =  0.492,  val loss =  0.684,  val accuray =  0.562, \n",
      "Epoch 490, Batch loss =  0.654,  training accuray =  0.570,  val loss =  0.674,  val accuray =  0.555, \n",
      "Epoch 500, Batch loss =  0.633,  training accuray =  0.617,  val loss =  0.707,  val accuray =  0.531, \n",
      "Epoch 510, Batch loss =  0.724,  training accuray =  0.477,  val loss =  0.710,  val accuray =  0.508, \n",
      "Epoch 520, Batch loss =  0.653,  training accuray =  0.594,  val loss =  0.711,  val accuray =  0.523, \n",
      "Epoch 530, Batch loss =  0.595,  training accuray =  0.656,  val loss =  0.686,  val accuray =  0.586, \n",
      "Epoch 540, Batch loss =  0.634,  training accuray =  0.602,  val loss =  0.662,  val accuray =  0.555, \n",
      "Epoch 550, Batch loss =  0.626,  training accuray =  0.609,  val loss =  0.704,  val accuray =  0.516, \n",
      "Epoch 560, Batch loss =  0.685,  training accuray =  0.547,  val loss =  0.708,  val accuray =  0.531, \n",
      "Epoch 570, Batch loss =  0.679,  training accuray =  0.555,  val loss =  0.670,  val accuray =  0.594, \n",
      "Epoch 580, Batch loss =  0.642,  training accuray =  0.625,  val loss =  0.666,  val accuray =  0.562, \n",
      "Epoch 590, Batch loss =  0.672,  training accuray =  0.562,  val loss =  0.711,  val accuray =  0.477, \n",
      "Epoch 600, Batch loss =  0.675,  training accuray =  0.547,  val loss =  0.629,  val accuray =  0.633, \n",
      "Epoch 610, Batch loss =  0.665,  training accuray =  0.547,  val loss =  0.668,  val accuray =  0.547, \n",
      "Epoch 620, Batch loss =  0.626,  training accuray =  0.594,  val loss =  0.612,  val accuray =  0.648, \n",
      "Epoch 630, Batch loss =  0.672,  training accuray =  0.555,  val loss =  0.674,  val accuray =  0.531, \n",
      "Epoch 640, Batch loss =  0.610,  training accuray =  0.633,  val loss =  0.671,  val accuray =  0.570, \n",
      "Epoch 650, Batch loss =  0.682,  training accuray =  0.570,  val loss =  0.630,  val accuray =  0.602, \n",
      "Epoch 660, Batch loss =  0.611,  training accuray =  0.625,  val loss =  0.650,  val accuray =  0.594, \n",
      "Epoch 670, Batch loss =  0.640,  training accuray =  0.539,  val loss =  0.618,  val accuray =  0.656, \n",
      "Epoch 680, Batch loss =  0.617,  training accuray =  0.656,  val loss =  0.641,  val accuray =  0.578, \n",
      "Epoch 690, Batch loss =  0.622,  training accuray =  0.594,  val loss =  0.653,  val accuray =  0.562, \n",
      "Epoch 700, Batch loss =  0.617,  training accuray =  0.609,  val loss =  0.607,  val accuray =  0.680, \n",
      "Epoch 710, Batch loss =  0.562,  training accuray =  0.719,  val loss =  0.624,  val accuray =  0.656, \n",
      "Epoch 720, Batch loss =  0.587,  training accuray =  0.664,  val loss =  0.647,  val accuray =  0.562, \n",
      "Epoch 730, Batch loss =  0.641,  training accuray =  0.586,  val loss =  0.579,  val accuray =  0.688, \n",
      "Epoch 740, Batch loss =  0.607,  training accuray =  0.672,  val loss =  0.620,  val accuray =  0.656, \n",
      "Epoch 750, Batch loss =  0.596,  training accuray =  0.664,  val loss =  0.591,  val accuray =  0.680, \n",
      "Epoch 760, Batch loss =  0.570,  training accuray =  0.695,  val loss =  0.592,  val accuray =  0.672, \n",
      "Epoch 770, Batch loss =  0.547,  training accuray =  0.719,  val loss =  0.583,  val accuray =  0.703, \n",
      "Epoch 780, Batch loss =  0.581,  training accuray =  0.711,  val loss =  0.615,  val accuray =  0.633, \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 790, Batch loss =  0.557,  training accuray =  0.750,  val loss =  0.592,  val accuray =  0.648, \n",
      "Epoch 800, Batch loss =  0.574,  training accuray =  0.695,  val loss =  0.631,  val accuray =  0.609, \n",
      "Epoch 810, Batch loss =  0.570,  training accuray =  0.719,  val loss =  0.585,  val accuray =  0.680, \n",
      "Epoch 820, Batch loss =  0.515,  training accuray =  0.766,  val loss =  0.575,  val accuray =  0.703, \n",
      "Epoch 830, Batch loss =  0.567,  training accuray =  0.727,  val loss =  0.574,  val accuray =  0.719, \n",
      "Epoch 840, Batch loss =  0.587,  training accuray =  0.641,  val loss =  0.582,  val accuray =  0.664, \n",
      "Epoch 850, Batch loss =  0.552,  training accuray =  0.703,  val loss =  0.547,  val accuray =  0.734, \n",
      "Epoch 860, Batch loss =  0.542,  training accuray =  0.750,  val loss =  0.584,  val accuray =  0.688, \n",
      "Epoch 870, Batch loss =  0.547,  training accuray =  0.758,  val loss =  0.561,  val accuray =  0.711, \n",
      "Epoch 880, Batch loss =  0.509,  training accuray =  0.766,  val loss =  0.592,  val accuray =  0.688, \n",
      "Epoch 890, Batch loss =  0.598,  training accuray =  0.711,  val loss =  0.562,  val accuray =  0.742, \n",
      "Epoch 900, Batch loss =  0.546,  training accuray =  0.789,  val loss =  0.549,  val accuray =  0.742, \n",
      "Epoch 910, Batch loss =  0.549,  training accuray =  0.719,  val loss =  0.536,  val accuray =  0.727, \n",
      "Epoch 920, Batch loss =  0.587,  training accuray =  0.641,  val loss =  0.566,  val accuray =  0.727, \n",
      "Epoch 930, Batch loss =  0.538,  training accuray =  0.758,  val loss =  0.547,  val accuray =  0.766, \n",
      "Epoch 940, Batch loss =  0.575,  training accuray =  0.711,  val loss =  0.528,  val accuray =  0.750, \n",
      "Epoch 950, Batch loss =  0.488,  training accuray =  0.812,  val loss =  0.581,  val accuray =  0.703, \n",
      "Epoch 960, Batch loss =  0.530,  training accuray =  0.773,  val loss =  0.568,  val accuray =  0.688, \n",
      "Epoch 970, Batch loss =  0.555,  training accuray =  0.750,  val loss =  0.507,  val accuray =  0.805, \n",
      "Epoch 980, Batch loss =  0.493,  training accuray =  0.820,  val loss =  0.535,  val accuray =  0.805, \n",
      "Epoch 990, Batch loss =  0.504,  training accuray =  0.812,  val loss =  0.520,  val accuray =  0.766, \n",
      "Epoch 1000, Batch loss =  0.451,  training accuray =  0.852,  val loss =  0.530,  val accuray =  0.773, \n",
      "Epoch 1010, Batch loss =  0.503,  training accuray =  0.805,  val loss =  0.542,  val accuray =  0.781, \n",
      "Epoch 1020, Batch loss =  0.480,  training accuray =  0.812,  val loss =  0.539,  val accuray =  0.758, \n",
      "Epoch 1030, Batch loss =  0.503,  training accuray =  0.805,  val loss =  0.546,  val accuray =  0.734, \n",
      "Epoch 1040, Batch loss =  0.488,  training accuray =  0.805,  val loss =  0.523,  val accuray =  0.773, \n",
      "Epoch 1050, Batch loss =  0.490,  training accuray =  0.828,  val loss =  0.513,  val accuray =  0.805, \n",
      "Epoch 1060, Batch loss =  0.524,  training accuray =  0.828,  val loss =  0.541,  val accuray =  0.758, \n",
      "Epoch 1070, Batch loss =  0.495,  training accuray =  0.828,  val loss =  0.530,  val accuray =  0.789, \n",
      "Epoch 1080, Batch loss =  0.473,  training accuray =  0.859,  val loss =  0.459,  val accuray =  0.852, \n",
      "Epoch 1090, Batch loss =  0.486,  training accuray =  0.812,  val loss =  0.531,  val accuray =  0.742, \n",
      "Epoch 1100, Batch loss =  0.457,  training accuray =  0.805,  val loss =  0.504,  val accuray =  0.773, \n",
      "Epoch 1110, Batch loss =  0.458,  training accuray =  0.812,  val loss =  0.501,  val accuray =  0.844, \n",
      "Epoch 1120, Batch loss =  0.478,  training accuray =  0.797,  val loss =  0.513,  val accuray =  0.820, \n",
      "Epoch 1130, Batch loss =  0.478,  training accuray =  0.828,  val loss =  0.548,  val accuray =  0.742, \n",
      "Epoch 1140, Batch loss =  0.481,  training accuray =  0.836,  val loss =  0.542,  val accuray =  0.719, \n",
      "Epoch 1150, Batch loss =  0.485,  training accuray =  0.820,  val loss =  0.473,  val accuray =  0.844, \n",
      "Epoch 1160, Batch loss =  0.469,  training accuray =  0.828,  val loss =  0.486,  val accuray =  0.812, \n",
      "Epoch 1170, Batch loss =  0.465,  training accuray =  0.812,  val loss =  0.517,  val accuray =  0.797, \n",
      "Epoch 1180, Batch loss =  0.464,  training accuray =  0.797,  val loss =  0.500,  val accuray =  0.789, \n",
      "Epoch 1190, Batch loss =  0.473,  training accuray =  0.828,  val loss =  0.465,  val accuray =  0.844, \n",
      "Epoch 1200, Batch loss =  0.438,  training accuray =  0.859,  val loss =  0.513,  val accuray =  0.750, \n",
      "Epoch 1210, Batch loss =  0.444,  training accuray =  0.867,  val loss =  0.502,  val accuray =  0.812, \n",
      "Epoch 1220, Batch loss =  0.437,  training accuray =  0.867,  val loss =  0.478,  val accuray =  0.805, \n",
      "Epoch 1230, Batch loss =  0.459,  training accuray =  0.852,  val loss =  0.458,  val accuray =  0.828, \n",
      "Epoch 1240, Batch loss =  0.435,  training accuray =  0.852,  val loss =  0.467,  val accuray =  0.820, \n",
      "Epoch 1250, Batch loss =  0.472,  training accuray =  0.820,  val loss =  0.494,  val accuray =  0.797, \n",
      "Epoch 1260, Batch loss =  0.496,  training accuray =  0.828,  val loss =  0.500,  val accuray =  0.781, \n",
      "Epoch 1270, Batch loss =  0.444,  training accuray =  0.844,  val loss =  0.518,  val accuray =  0.758, \n",
      "Epoch 1280, Batch loss =  0.427,  training accuray =  0.867,  val loss =  0.485,  val accuray =  0.766, \n",
      "Epoch 1290, Batch loss =  0.433,  training accuray =  0.883,  val loss =  0.527,  val accuray =  0.742, \n",
      "Epoch 1300, Batch loss =  0.451,  training accuray =  0.852,  val loss =  0.500,  val accuray =  0.805, \n",
      "Epoch 1310, Batch loss =  0.430,  training accuray =  0.883,  val loss =  0.476,  val accuray =  0.781, \n",
      "Epoch 1320, Batch loss =  0.420,  training accuray =  0.883,  val loss =  0.533,  val accuray =  0.727, \n",
      "Epoch 1330, Batch loss =  0.435,  training accuray =  0.844,  val loss =  0.449,  val accuray =  0.875, \n",
      "Epoch 1340, Batch loss =  0.441,  training accuray =  0.805,  val loss =  0.465,  val accuray =  0.844, \n",
      "Epoch 1350, Batch loss =  0.427,  training accuray =  0.875,  val loss =  0.480,  val accuray =  0.797, \n",
      "Epoch 1360, Batch loss =  0.423,  training accuray =  0.867,  val loss =  0.476,  val accuray =  0.781, \n",
      "Epoch 1370, Batch loss =  0.418,  training accuray =  0.836,  val loss =  0.522,  val accuray =  0.758, \n",
      "Epoch 1380, Batch loss =  0.366,  training accuray =  0.930,  val loss =  0.467,  val accuray =  0.828, \n",
      "Epoch 1390, Batch loss =  0.428,  training accuray =  0.820,  val loss =  0.494,  val accuray =  0.828, \n",
      "Epoch 1400, Batch loss =  0.424,  training accuray =  0.883,  val loss =  0.522,  val accuray =  0.766, \n",
      "Epoch 1410, Batch loss =  0.429,  training accuray =  0.836,  val loss =  0.470,  val accuray =  0.805, \n",
      "Epoch 1420, Batch loss =  0.425,  training accuray =  0.852,  val loss =  0.480,  val accuray =  0.758, \n",
      "Epoch 1430, Batch loss =  0.437,  training accuray =  0.812,  val loss =  0.478,  val accuray =  0.820, \n",
      "Epoch 1440, Batch loss =  0.428,  training accuray =  0.852,  val loss =  0.462,  val accuray =  0.820, \n",
      "Epoch 1450, Batch loss =  0.449,  training accuray =  0.789,  val loss =  0.443,  val accuray =  0.828, \n",
      "Epoch 1460, Batch loss =  0.424,  training accuray =  0.867,  val loss =  0.462,  val accuray =  0.781, \n",
      "Epoch 1470, Batch loss =  0.399,  training accuray =  0.859,  val loss =  0.467,  val accuray =  0.750, \n",
      "Epoch 1480, Batch loss =  0.379,  training accuray =  0.891,  val loss =  0.440,  val accuray =  0.836, \n",
      "Epoch 1490, Batch loss =  0.372,  training accuray =  0.898,  val loss =  0.481,  val accuray =  0.758, \n",
      "Epoch 1500, Batch loss =  0.430,  training accuray =  0.828,  val loss =  0.405,  val accuray =  0.820, \n",
      "Epoch 1510, Batch loss =  0.383,  training accuray =  0.875,  val loss =  0.462,  val accuray =  0.797, \n",
      "Epoch 1520, Batch loss =  0.383,  training accuray =  0.898,  val loss =  0.487,  val accuray =  0.789, \n",
      "Epoch 1530, Batch loss =  0.389,  training accuray =  0.867,  val loss =  0.444,  val accuray =  0.859, \n",
      "Epoch 1540, Batch loss =  0.344,  training accuray =  0.930,  val loss =  0.463,  val accuray =  0.797, \n",
      "Epoch 1550, Batch loss =  0.385,  training accuray =  0.867,  val loss =  0.447,  val accuray =  0.836, \n",
      "Epoch 1560, Batch loss =  0.394,  training accuray =  0.836,  val loss =  0.474,  val accuray =  0.789, \n",
      "Epoch 1570, Batch loss =  0.362,  training accuray =  0.883,  val loss =  0.415,  val accuray =  0.852, \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1580, Batch loss =  0.384,  training accuray =  0.875,  val loss =  0.433,  val accuray =  0.797, \n",
      "Epoch 1590, Batch loss =  0.376,  training accuray =  0.898,  val loss =  0.432,  val accuray =  0.836, \n",
      "Epoch 1600, Batch loss =  0.408,  training accuray =  0.844,  val loss =  0.477,  val accuray =  0.789, \n",
      "Epoch 1610, Batch loss =  0.383,  training accuray =  0.859,  val loss =  0.368,  val accuray =  0.898, \n",
      "Epoch 1620, Batch loss =  0.383,  training accuray =  0.852,  val loss =  0.451,  val accuray =  0.797, \n",
      "Epoch 1630, Batch loss =  0.394,  training accuray =  0.867,  val loss =  0.442,  val accuray =  0.844, \n",
      "Epoch 1640, Batch loss =  0.389,  training accuray =  0.867,  val loss =  0.458,  val accuray =  0.805, \n",
      "Epoch 1650, Batch loss =  0.326,  training accuray =  0.891,  val loss =  0.458,  val accuray =  0.781, \n",
      "Epoch 1660, Batch loss =  0.391,  training accuray =  0.875,  val loss =  0.400,  val accuray =  0.820, \n",
      "Epoch 1670, Batch loss =  0.320,  training accuray =  0.938,  val loss =  0.442,  val accuray =  0.781, \n",
      "Epoch 1680, Batch loss =  0.338,  training accuray =  0.883,  val loss =  0.434,  val accuray =  0.805, \n",
      "Epoch 1690, Batch loss =  0.368,  training accuray =  0.891,  val loss =  0.397,  val accuray =  0.844, \n",
      "Epoch 1700, Batch loss =  0.392,  training accuray =  0.859,  val loss =  0.373,  val accuray =  0.875, \n",
      "Epoch 1710, Batch loss =  0.393,  training accuray =  0.852,  val loss =  0.405,  val accuray =  0.836, \n",
      "Epoch 1720, Batch loss =  0.348,  training accuray =  0.898,  val loss =  0.486,  val accuray =  0.781, \n",
      "Epoch 1730, Batch loss =  0.363,  training accuray =  0.898,  val loss =  0.424,  val accuray =  0.820, \n",
      "Epoch 1740, Batch loss =  0.334,  training accuray =  0.891,  val loss =  0.432,  val accuray =  0.836, \n",
      "Epoch 1750, Batch loss =  0.353,  training accuray =  0.859,  val loss =  0.382,  val accuray =  0.867, \n",
      "Epoch 1760, Batch loss =  0.360,  training accuray =  0.844,  val loss =  0.409,  val accuray =  0.844, \n",
      "Epoch 1770, Batch loss =  0.374,  training accuray =  0.828,  val loss =  0.430,  val accuray =  0.789, \n",
      "Epoch 1780, Batch loss =  0.351,  training accuray =  0.867,  val loss =  0.389,  val accuray =  0.867, \n",
      "Epoch 1790, Batch loss =  0.346,  training accuray =  0.867,  val loss =  0.427,  val accuray =  0.805, \n",
      "Epoch 1800, Batch loss =  0.330,  training accuray =  0.914,  val loss =  0.455,  val accuray =  0.805, \n",
      "Epoch 1810, Batch loss =  0.296,  training accuray =  0.953,  val loss =  0.372,  val accuray =  0.867, \n",
      "Epoch 1820, Batch loss =  0.350,  training accuray =  0.875,  val loss =  0.388,  val accuray =  0.859, \n",
      "Epoch 1830, Batch loss =  0.338,  training accuray =  0.938,  val loss =  0.446,  val accuray =  0.773, \n",
      "Epoch 1840, Batch loss =  0.279,  training accuray =  0.938,  val loss =  0.404,  val accuray =  0.828, \n",
      "Epoch 1850, Batch loss =  0.352,  training accuray =  0.859,  val loss =  0.370,  val accuray =  0.898, \n",
      "Epoch 1860, Batch loss =  0.335,  training accuray =  0.898,  val loss =  0.443,  val accuray =  0.797, \n",
      "Epoch 1870, Batch loss =  0.283,  training accuray =  0.938,  val loss =  0.430,  val accuray =  0.781, \n",
      "Epoch 1880, Batch loss =  0.287,  training accuray =  0.938,  val loss =  0.403,  val accuray =  0.867, \n",
      "Epoch 1890, Batch loss =  0.305,  training accuray =  0.906,  val loss =  0.411,  val accuray =  0.844, \n",
      "Epoch 1900, Batch loss =  0.366,  training accuray =  0.875,  val loss =  0.395,  val accuray =  0.844, \n",
      "Epoch 1910, Batch loss =  0.337,  training accuray =  0.906,  val loss =  0.419,  val accuray =  0.797, \n",
      "Epoch 1920, Batch loss =  0.361,  training accuray =  0.875,  val loss =  0.443,  val accuray =  0.812, \n",
      "Epoch 1930, Batch loss =  0.322,  training accuray =  0.875,  val loss =  0.396,  val accuray =  0.867, \n",
      "Epoch 1940, Batch loss =  0.359,  training accuray =  0.867,  val loss =  0.329,  val accuray =  0.891, \n",
      "Epoch 1950, Batch loss =  0.348,  training accuray =  0.859,  val loss =  0.358,  val accuray =  0.883, \n",
      "Epoch 1960, Batch loss =  0.356,  training accuray =  0.883,  val loss =  0.387,  val accuray =  0.875, \n",
      "Epoch 1970, Batch loss =  0.335,  training accuray =  0.883,  val loss =  0.408,  val accuray =  0.797, \n",
      "Epoch 1980, Batch loss =  0.325,  training accuray =  0.891,  val loss =  0.363,  val accuray =  0.867, \n",
      "Epoch 1990, Batch loss =  0.314,  training accuray =  0.914,  val loss =  0.345,  val accuray =  0.914, \n",
      "Epoch 2000, Batch loss =  0.307,  training accuray =  0.922,  val loss =  0.398,  val accuray =  0.820, \n",
      "Epoch 2010, Batch loss =  0.305,  training accuray =  0.922,  val loss =  0.386,  val accuray =  0.852, \n",
      "Epoch 2020, Batch loss =  0.328,  training accuray =  0.883,  val loss =  0.414,  val accuray =  0.820, \n",
      "Epoch 2030, Batch loss =  0.281,  training accuray =  0.898,  val loss =  0.324,  val accuray =  0.867, \n",
      "Epoch 2040, Batch loss =  0.311,  training accuray =  0.945,  val loss =  0.375,  val accuray =  0.852, \n",
      "Epoch 2050, Batch loss =  0.288,  training accuray =  0.891,  val loss =  0.410,  val accuray =  0.820, \n",
      "Epoch 2060, Batch loss =  0.336,  training accuray =  0.875,  val loss =  0.385,  val accuray =  0.875, \n",
      "Epoch 2070, Batch loss =  0.323,  training accuray =  0.906,  val loss =  0.388,  val accuray =  0.844, \n",
      "Epoch 2080, Batch loss =  0.344,  training accuray =  0.891,  val loss =  0.383,  val accuray =  0.828, \n",
      "Epoch 2090, Batch loss =  0.305,  training accuray =  0.875,  val loss =  0.352,  val accuray =  0.898, \n",
      "Epoch 2100, Batch loss =  0.309,  training accuray =  0.891,  val loss =  0.350,  val accuray =  0.867, \n",
      "Epoch 2110, Batch loss =  0.318,  training accuray =  0.914,  val loss =  0.371,  val accuray =  0.836, \n",
      "Epoch 2120, Batch loss =  0.292,  training accuray =  0.898,  val loss =  0.354,  val accuray =  0.867, \n",
      "Epoch 2130, Batch loss =  0.339,  training accuray =  0.891,  val loss =  0.309,  val accuray =  0.914, \n",
      "Epoch 2140, Batch loss =  0.296,  training accuray =  0.891,  val loss =  0.447,  val accuray =  0.797, \n",
      "Epoch 2150, Batch loss =  0.360,  training accuray =  0.875,  val loss =  0.448,  val accuray =  0.812, \n",
      "Epoch 2160, Batch loss =  0.320,  training accuray =  0.883,  val loss =  0.385,  val accuray =  0.836, \n",
      "Epoch 2170, Batch loss =  0.332,  training accuray =  0.875,  val loss =  0.371,  val accuray =  0.836, \n",
      "Epoch 2180, Batch loss =  0.303,  training accuray =  0.891,  val loss =  0.353,  val accuray =  0.844, \n",
      "Epoch 2190, Batch loss =  0.299,  training accuray =  0.891,  val loss =  0.366,  val accuray =  0.836, \n",
      "Epoch 2200, Batch loss =  0.328,  training accuray =  0.828,  val loss =  0.399,  val accuray =  0.844, \n",
      "Epoch 2210, Batch loss =  0.289,  training accuray =  0.898,  val loss =  0.443,  val accuray =  0.758, \n",
      "Epoch 2220, Batch loss =  0.263,  training accuray =  0.922,  val loss =  0.370,  val accuray =  0.867, \n",
      "Epoch 2230, Batch loss =  0.298,  training accuray =  0.898,  val loss =  0.329,  val accuray =  0.891, \n",
      "Epoch 2240, Batch loss =  0.289,  training accuray =  0.906,  val loss =  0.399,  val accuray =  0.844, \n",
      "Epoch 2250, Batch loss =  0.303,  training accuray =  0.914,  val loss =  0.373,  val accuray =  0.867, \n",
      "Epoch 2260, Batch loss =  0.297,  training accuray =  0.906,  val loss =  0.377,  val accuray =  0.844, \n",
      "Epoch 2270, Batch loss =  0.311,  training accuray =  0.883,  val loss =  0.351,  val accuray =  0.867, \n",
      "Epoch 2280, Batch loss =  0.277,  training accuray =  0.930,  val loss =  0.301,  val accuray =  0.922, \n",
      "Epoch 2290, Batch loss =  0.228,  training accuray =  0.969,  val loss =  0.352,  val accuray =  0.844, \n",
      "Epoch 2300, Batch loss =  0.275,  training accuray =  0.922,  val loss =  0.369,  val accuray =  0.875, \n",
      "Epoch 2310, Batch loss =  0.270,  training accuray =  0.922,  val loss =  0.373,  val accuray =  0.867, \n",
      "Epoch 2320, Batch loss =  0.273,  training accuray =  0.922,  val loss =  0.352,  val accuray =  0.852, \n",
      "Epoch 2330, Batch loss =  0.308,  training accuray =  0.898,  val loss =  0.353,  val accuray =  0.859, \n",
      "Epoch 2340, Batch loss =  0.218,  training accuray =  0.945,  val loss =  0.362,  val accuray =  0.875, \n",
      "Epoch 2350, Batch loss =  0.277,  training accuray =  0.891,  val loss =  0.371,  val accuray =  0.844, \n",
      "Epoch 2360, Batch loss =  0.253,  training accuray =  0.945,  val loss =  0.404,  val accuray =  0.828, \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 2370, Batch loss =  0.225,  training accuray =  0.961,  val loss =  0.311,  val accuray =  0.930, \n",
      "Epoch 2380, Batch loss =  0.249,  training accuray =  0.953,  val loss =  0.339,  val accuray =  0.844, \n",
      "Epoch 2390, Batch loss =  0.296,  training accuray =  0.891,  val loss =  0.328,  val accuray =  0.891, \n",
      "Epoch 2400, Batch loss =  0.291,  training accuray =  0.898,  val loss =  0.351,  val accuray =  0.852, \n",
      "Epoch 2410, Batch loss =  0.267,  training accuray =  0.938,  val loss =  0.308,  val accuray =  0.883, \n",
      "Epoch 2420, Batch loss =  0.274,  training accuray =  0.922,  val loss =  0.304,  val accuray =  0.898, \n",
      "Epoch 2430, Batch loss =  0.268,  training accuray =  0.914,  val loss =  0.377,  val accuray =  0.812, \n",
      "Epoch 2440, Batch loss =  0.253,  training accuray =  0.906,  val loss =  0.359,  val accuray =  0.875, \n",
      "Epoch 2450, Batch loss =  0.290,  training accuray =  0.906,  val loss =  0.315,  val accuray =  0.875, \n",
      "Epoch 2460, Batch loss =  0.239,  training accuray =  0.938,  val loss =  0.306,  val accuray =  0.898, \n",
      "Epoch 2470, Batch loss =  0.247,  training accuray =  0.930,  val loss =  0.324,  val accuray =  0.859, \n",
      "Epoch 2480, Batch loss =  0.283,  training accuray =  0.891,  val loss =  0.302,  val accuray =  0.883, \n",
      "Epoch 2490, Batch loss =  0.261,  training accuray =  0.891,  val loss =  0.370,  val accuray =  0.852, \n",
      "Epoch 2500, Batch loss =  0.291,  training accuray =  0.906,  val loss =  0.307,  val accuray =  0.875, \n",
      "Epoch 2510, Batch loss =  0.240,  training accuray =  0.938,  val loss =  0.313,  val accuray =  0.875, \n",
      "Epoch 2520, Batch loss =  0.279,  training accuray =  0.898,  val loss =  0.343,  val accuray =  0.867, \n",
      "Epoch 2530, Batch loss =  0.315,  training accuray =  0.891,  val loss =  0.329,  val accuray =  0.883, \n",
      "Epoch 2540, Batch loss =  0.238,  training accuray =  0.938,  val loss =  0.315,  val accuray =  0.867, \n",
      "Epoch 2550, Batch loss =  0.239,  training accuray =  0.945,  val loss =  0.409,  val accuray =  0.820, \n",
      "Epoch 2560, Batch loss =  0.287,  training accuray =  0.898,  val loss =  0.286,  val accuray =  0.867, \n",
      "Epoch 2570, Batch loss =  0.294,  training accuray =  0.875,  val loss =  0.303,  val accuray =  0.867, \n",
      "Epoch 2580, Batch loss =  0.248,  training accuray =  0.922,  val loss =  0.291,  val accuray =  0.891, \n",
      "Epoch 2590, Batch loss =  0.227,  training accuray =  0.953,  val loss =  0.331,  val accuray =  0.852, \n",
      "Epoch 2600, Batch loss =  0.231,  training accuray =  0.922,  val loss =  0.295,  val accuray =  0.883, \n",
      "Epoch 2610, Batch loss =  0.247,  training accuray =  0.922,  val loss =  0.339,  val accuray =  0.852, \n",
      "Epoch 2620, Batch loss =  0.259,  training accuray =  0.914,  val loss =  0.341,  val accuray =  0.867, \n",
      "Epoch 2630, Batch loss =  0.219,  training accuray =  0.922,  val loss =  0.385,  val accuray =  0.812, \n",
      "Epoch 2640, Batch loss =  0.254,  training accuray =  0.914,  val loss =  0.377,  val accuray =  0.844, \n",
      "Epoch 2650, Batch loss =  0.253,  training accuray =  0.945,  val loss =  0.373,  val accuray =  0.805, \n",
      "Epoch 2660, Batch loss =  0.287,  training accuray =  0.891,  val loss =  0.318,  val accuray =  0.898, \n",
      "Epoch 2670, Batch loss =  0.232,  training accuray =  0.930,  val loss =  0.431,  val accuray =  0.750, \n",
      "Epoch 2680, Batch loss =  0.233,  training accuray =  0.930,  val loss =  0.292,  val accuray =  0.867, \n",
      "Epoch 2690, Batch loss =  0.274,  training accuray =  0.883,  val loss =  0.334,  val accuray =  0.852, \n",
      "Epoch 2700, Batch loss =  0.297,  training accuray =  0.883,  val loss =  0.311,  val accuray =  0.867, \n",
      "Epoch 2710, Batch loss =  0.260,  training accuray =  0.922,  val loss =  0.271,  val accuray =  0.930, \n",
      "Epoch 2720, Batch loss =  0.223,  training accuray =  0.930,  val loss =  0.253,  val accuray =  0.914, \n",
      "Epoch 2730, Batch loss =  0.292,  training accuray =  0.883,  val loss =  0.319,  val accuray =  0.867, \n",
      "Epoch 2740, Batch loss =  0.254,  training accuray =  0.914,  val loss =  0.294,  val accuray =  0.883, \n",
      "Epoch 2750, Batch loss =  0.254,  training accuray =  0.859,  val loss =  0.366,  val accuray =  0.828, \n",
      "Epoch 2760, Batch loss =  0.261,  training accuray =  0.922,  val loss =  0.313,  val accuray =  0.859, \n",
      "Epoch 2770, Batch loss =  0.230,  training accuray =  0.938,  val loss =  0.293,  val accuray =  0.914, \n",
      "Epoch 2780, Batch loss =  0.269,  training accuray =  0.883,  val loss =  0.290,  val accuray =  0.891, \n",
      "Epoch 2790, Batch loss =  0.205,  training accuray =  0.938,  val loss =  0.303,  val accuray =  0.906, \n",
      "Epoch 2800, Batch loss =  0.242,  training accuray =  0.930,  val loss =  0.353,  val accuray =  0.836, \n",
      "Epoch 2810, Batch loss =  0.201,  training accuray =  0.953,  val loss =  0.286,  val accuray =  0.867, \n",
      "Epoch 2820, Batch loss =  0.230,  training accuray =  0.906,  val loss =  0.294,  val accuray =  0.891, \n",
      "Epoch 2830, Batch loss =  0.173,  training accuray =  0.977,  val loss =  0.342,  val accuray =  0.859, \n",
      "Epoch 2840, Batch loss =  0.223,  training accuray =  0.930,  val loss =  0.288,  val accuray =  0.867, \n",
      "Epoch 2850, Batch loss =  0.231,  training accuray =  0.945,  val loss =  0.329,  val accuray =  0.898, \n",
      "Epoch 2860, Batch loss =  0.216,  training accuray =  0.938,  val loss =  0.338,  val accuray =  0.875, \n",
      "Epoch 2870, Batch loss =  0.193,  training accuray =  0.977,  val loss =  0.305,  val accuray =  0.898, \n",
      "Epoch 2880, Batch loss =  0.195,  training accuray =  0.938,  val loss =  0.317,  val accuray =  0.883, \n",
      "Epoch 2890, Batch loss =  0.189,  training accuray =  0.953,  val loss =  0.325,  val accuray =  0.898, \n",
      "Epoch 2900, Batch loss =  0.217,  training accuray =  0.938,  val loss =  0.286,  val accuray =  0.883, \n",
      "Epoch 2910, Batch loss =  0.260,  training accuray =  0.898,  val loss =  0.375,  val accuray =  0.828, \n",
      "Epoch 2920, Batch loss =  0.209,  training accuray =  0.969,  val loss =  0.297,  val accuray =  0.891, \n",
      "Epoch 2930, Batch loss =  0.169,  training accuray =  0.969,  val loss =  0.283,  val accuray =  0.906, \n",
      "Epoch 2940, Batch loss =  0.228,  training accuray =  0.938,  val loss =  0.324,  val accuray =  0.867, \n",
      "Epoch 2950, Batch loss =  0.239,  training accuray =  0.930,  val loss =  0.284,  val accuray =  0.883, \n",
      "Epoch 2960, Batch loss =  0.257,  training accuray =  0.898,  val loss =  0.314,  val accuray =  0.891, \n",
      "Epoch 2970, Batch loss =  0.281,  training accuray =  0.898,  val loss =  0.285,  val accuray =  0.883, \n",
      "Epoch 2980, Batch loss =  0.255,  training accuray =  0.906,  val loss =  0.258,  val accuray =  0.906, \n",
      "Epoch 2990, Batch loss =  0.250,  training accuray =  0.914,  val loss =  0.304,  val accuray =  0.891, \n",
      "Epoch 3000, Batch loss =  0.208,  training accuray =  0.953,  val loss =  0.311,  val accuray =  0.859, \n",
      "Epoch 3010, Batch loss =  0.229,  training accuray =  0.922,  val loss =  0.299,  val accuray =  0.867, \n",
      "Epoch 3020, Batch loss =  0.225,  training accuray =  0.930,  val loss =  0.322,  val accuray =  0.859, \n",
      "Epoch 3030, Batch loss =  0.198,  training accuray =  0.945,  val loss =  0.284,  val accuray =  0.883, \n",
      "Epoch 3040, Batch loss =  0.206,  training accuray =  0.938,  val loss =  0.294,  val accuray =  0.875, \n",
      "Epoch 3050, Batch loss =  0.182,  training accuray =  0.953,  val loss =  0.329,  val accuray =  0.883, \n",
      "Epoch 3060, Batch loss =  0.251,  training accuray =  0.898,  val loss =  0.297,  val accuray =  0.852, \n",
      "Epoch 3070, Batch loss =  0.235,  training accuray =  0.930,  val loss =  0.270,  val accuray =  0.906, \n",
      "Epoch 3080, Batch loss =  0.237,  training accuray =  0.914,  val loss =  0.282,  val accuray =  0.883, \n",
      "Epoch 3090, Batch loss =  0.225,  training accuray =  0.953,  val loss =  0.281,  val accuray =  0.883, \n",
      "Epoch 3100, Batch loss =  0.242,  training accuray =  0.914,  val loss =  0.331,  val accuray =  0.852, \n",
      "Epoch 3110, Batch loss =  0.203,  training accuray =  0.930,  val loss =  0.310,  val accuray =  0.875, \n",
      "Epoch 3120, Batch loss =  0.161,  training accuray =  0.953,  val loss =  0.336,  val accuray =  0.867, \n",
      "Epoch 3130, Batch loss =  0.223,  training accuray =  0.922,  val loss =  0.258,  val accuray =  0.891, \n",
      "Epoch 3140, Batch loss =  0.191,  training accuray =  0.953,  val loss =  0.262,  val accuray =  0.922, \n",
      "Epoch 3150, Batch loss =  0.261,  training accuray =  0.922,  val loss =  0.315,  val accuray =  0.828, \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 3160, Batch loss =  0.221,  training accuray =  0.938,  val loss =  0.242,  val accuray =  0.922, \n",
      "Epoch 3170, Batch loss =  0.191,  training accuray =  0.953,  val loss =  0.328,  val accuray =  0.867, \n",
      "Epoch 3180, Batch loss =  0.218,  training accuray =  0.945,  val loss =  0.289,  val accuray =  0.891, \n",
      "Epoch 3190, Batch loss =  0.195,  training accuray =  0.930,  val loss =  0.326,  val accuray =  0.867, \n",
      "Epoch 3200, Batch loss =  0.225,  training accuray =  0.922,  val loss =  0.306,  val accuray =  0.875, \n",
      "Epoch 3210, Batch loss =  0.248,  training accuray =  0.922,  val loss =  0.260,  val accuray =  0.883, \n",
      "Epoch 3220, Batch loss =  0.168,  training accuray =  0.969,  val loss =  0.340,  val accuray =  0.820, \n",
      "Epoch 3230, Batch loss =  0.247,  training accuray =  0.922,  val loss =  0.368,  val accuray =  0.820, \n",
      "Epoch 3240, Batch loss =  0.231,  training accuray =  0.914,  val loss =  0.294,  val accuray =  0.891, \n",
      "Epoch 3250, Batch loss =  0.229,  training accuray =  0.914,  val loss =  0.278,  val accuray =  0.891, \n",
      "Epoch 3260, Batch loss =  0.211,  training accuray =  0.914,  val loss =  0.260,  val accuray =  0.938, \n",
      "Epoch 3270, Batch loss =  0.201,  training accuray =  0.953,  val loss =  0.305,  val accuray =  0.891, \n",
      "Epoch 3280, Batch loss =  0.195,  training accuray =  0.969,  val loss =  0.242,  val accuray =  0.914, \n",
      "Epoch 3290, Batch loss =  0.208,  training accuray =  0.938,  val loss =  0.277,  val accuray =  0.906, \n",
      "Epoch 3300, Batch loss =  0.202,  training accuray =  0.930,  val loss =  0.267,  val accuray =  0.875, \n",
      "Epoch 3310, Batch loss =  0.223,  training accuray =  0.938,  val loss =  0.299,  val accuray =  0.883, \n",
      "Epoch 3320, Batch loss =  0.190,  training accuray =  0.930,  val loss =  0.259,  val accuray =  0.906, \n",
      "Epoch 3330, Batch loss =  0.233,  training accuray =  0.938,  val loss =  0.301,  val accuray =  0.906, \n",
      "Epoch 3340, Batch loss =  0.249,  training accuray =  0.922,  val loss =  0.337,  val accuray =  0.859, \n",
      "Epoch 3350, Batch loss =  0.187,  training accuray =  0.961,  val loss =  0.273,  val accuray =  0.883, \n",
      "Epoch 3360, Batch loss =  0.203,  training accuray =  0.922,  val loss =  0.260,  val accuray =  0.922, \n",
      "Epoch 3370, Batch loss =  0.208,  training accuray =  0.938,  val loss =  0.285,  val accuray =  0.875, \n",
      "Epoch 3380, Batch loss =  0.243,  training accuray =  0.914,  val loss =  0.258,  val accuray =  0.906, \n",
      "Epoch 3390, Batch loss =  0.203,  training accuray =  0.938,  val loss =  0.280,  val accuray =  0.883, \n",
      "Epoch 3400, Batch loss =  0.186,  training accuray =  0.945,  val loss =  0.271,  val accuray =  0.906, \n",
      "Epoch 3410, Batch loss =  0.216,  training accuray =  0.945,  val loss =  0.280,  val accuray =  0.875, \n",
      "Epoch 3420, Batch loss =  0.203,  training accuray =  0.953,  val loss =  0.316,  val accuray =  0.867, \n",
      "Epoch 3430, Batch loss =  0.248,  training accuray =  0.906,  val loss =  0.254,  val accuray =  0.914, \n",
      "Epoch 3440, Batch loss =  0.179,  training accuray =  0.969,  val loss =  0.253,  val accuray =  0.914, \n",
      "Epoch 3450, Batch loss =  0.209,  training accuray =  0.961,  val loss =  0.258,  val accuray =  0.914, \n",
      "Epoch 3460, Batch loss =  0.182,  training accuray =  0.953,  val loss =  0.274,  val accuray =  0.891, \n",
      "Epoch 3470, Batch loss =  0.190,  training accuray =  0.953,  val loss =  0.259,  val accuray =  0.930, \n",
      "Epoch 3480, Batch loss =  0.182,  training accuray =  0.938,  val loss =  0.259,  val accuray =  0.891, \n",
      "Epoch 3490, Batch loss =  0.205,  training accuray =  0.930,  val loss =  0.306,  val accuray =  0.867, \n",
      "Epoch 3500, Batch loss =  0.185,  training accuray =  0.961,  val loss =  0.266,  val accuray =  0.914, \n",
      "Epoch 3510, Batch loss =  0.223,  training accuray =  0.930,  val loss =  0.242,  val accuray =  0.922, \n",
      "Epoch 3520, Batch loss =  0.123,  training accuray =  0.984,  val loss =  0.324,  val accuray =  0.867, \n",
      "Epoch 3530, Batch loss =  0.168,  training accuray =  0.977,  val loss =  0.355,  val accuray =  0.852, \n",
      "Epoch 3540, Batch loss =  0.214,  training accuray =  0.922,  val loss =  0.342,  val accuray =  0.844, \n",
      "Epoch 3550, Batch loss =  0.162,  training accuray =  0.984,  val loss =  0.286,  val accuray =  0.906, \n",
      "Epoch 3560, Batch loss =  0.192,  training accuray =  0.938,  val loss =  0.301,  val accuray =  0.859, \n",
      "Epoch 3570, Batch loss =  0.180,  training accuray =  0.945,  val loss =  0.270,  val accuray =  0.914, \n",
      "Epoch 3580, Batch loss =  0.173,  training accuray =  0.961,  val loss =  0.282,  val accuray =  0.875, \n",
      "Epoch 3590, Batch loss =  0.125,  training accuray =  0.992,  val loss =  0.285,  val accuray =  0.867, \n",
      "Epoch 3600, Batch loss =  0.195,  training accuray =  0.945,  val loss =  0.366,  val accuray =  0.828, \n",
      "Epoch 3610, Batch loss =  0.164,  training accuray =  0.961,  val loss =  0.297,  val accuray =  0.898, \n",
      "Epoch 3620, Batch loss =  0.183,  training accuray =  0.953,  val loss =  0.282,  val accuray =  0.867, \n",
      "Epoch 3630, Batch loss =  0.172,  training accuray =  0.961,  val loss =  0.282,  val accuray =  0.883, \n",
      "Epoch 3640, Batch loss =  0.153,  training accuray =  0.977,  val loss =  0.291,  val accuray =  0.898, \n",
      "Epoch 3650, Batch loss =  0.170,  training accuray =  0.945,  val loss =  0.239,  val accuray =  0.930, \n",
      "Epoch 3660, Batch loss =  0.200,  training accuray =  0.945,  val loss =  0.273,  val accuray =  0.914, \n",
      "Epoch 3670, Batch loss =  0.166,  training accuray =  0.953,  val loss =  0.239,  val accuray =  0.922, \n",
      "Epoch 3680, Batch loss =  0.155,  training accuray =  0.977,  val loss =  0.300,  val accuray =  0.875, \n",
      "Epoch 3690, Batch loss =  0.165,  training accuray =  0.953,  val loss =  0.274,  val accuray =  0.906, \n",
      "Epoch 3700, Batch loss =  0.182,  training accuray =  0.945,  val loss =  0.237,  val accuray =  0.938, \n",
      "Epoch 3710, Batch loss =  0.198,  training accuray =  0.945,  val loss =  0.286,  val accuray =  0.906, \n",
      "Epoch 3720, Batch loss =  0.203,  training accuray =  0.922,  val loss =  0.274,  val accuray =  0.906, \n",
      "Epoch 3730, Batch loss =  0.173,  training accuray =  0.953,  val loss =  0.257,  val accuray =  0.906, \n",
      "Epoch 3740, Batch loss =  0.183,  training accuray =  0.930,  val loss =  0.336,  val accuray =  0.859, \n",
      "Epoch 3750, Batch loss =  0.170,  training accuray =  0.984,  val loss =  0.335,  val accuray =  0.859, \n",
      "Epoch 3760, Batch loss =  0.148,  training accuray =  0.977,  val loss =  0.349,  val accuray =  0.820, \n",
      "Epoch 3770, Batch loss =  0.141,  training accuray =  0.969,  val loss =  0.271,  val accuray =  0.914, \n",
      "Epoch 3780, Batch loss =  0.155,  training accuray =  0.984,  val loss =  0.229,  val accuray =  0.922, \n",
      "Epoch 3790, Batch loss =  0.170,  training accuray =  0.961,  val loss =  0.239,  val accuray =  0.930, \n",
      "Epoch 3800, Batch loss =  0.162,  training accuray =  0.977,  val loss =  0.317,  val accuray =  0.867, \n",
      "Epoch 3810, Batch loss =  0.182,  training accuray =  0.945,  val loss =  0.227,  val accuray =  0.938, \n",
      "Epoch 3820, Batch loss =  0.180,  training accuray =  0.938,  val loss =  0.276,  val accuray =  0.922, \n",
      "Epoch 3830, Batch loss =  0.270,  training accuray =  0.883,  val loss =  0.292,  val accuray =  0.875, \n",
      "Epoch 3840, Batch loss =  0.157,  training accuray =  0.953,  val loss =  0.299,  val accuray =  0.875, \n",
      "Epoch 3850, Batch loss =  0.162,  training accuray =  0.961,  val loss =  0.274,  val accuray =  0.898, \n",
      "Epoch 3860, Batch loss =  0.151,  training accuray =  0.945,  val loss =  0.241,  val accuray =  0.906, \n",
      "Epoch 3870, Batch loss =  0.217,  training accuray =  0.938,  val loss =  0.286,  val accuray =  0.875, \n",
      "Epoch 3880, Batch loss =  0.179,  training accuray =  0.953,  val loss =  0.327,  val accuray =  0.883, \n",
      "Epoch 3890, Batch loss =  0.167,  training accuray =  0.961,  val loss =  0.271,  val accuray =  0.891, \n",
      "Epoch 3900, Batch loss =  0.194,  training accuray =  0.945,  val loss =  0.244,  val accuray =  0.922, \n",
      "Epoch 3910, Batch loss =  0.152,  training accuray =  0.977,  val loss =  0.232,  val accuray =  0.898, \n",
      "Epoch 3920, Batch loss =  0.192,  training accuray =  0.961,  val loss =  0.188,  val accuray =  0.945, \n",
      "Epoch 3930, Batch loss =  0.188,  training accuray =  0.953,  val loss =  0.278,  val accuray =  0.898, \n",
      "Epoch 3940, Batch loss =  0.158,  training accuray =  0.953,  val loss =  0.217,  val accuray =  0.922, \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 3950, Batch loss =  0.161,  training accuray =  0.953,  val loss =  0.276,  val accuray =  0.906, \n",
      "Epoch 3960, Batch loss =  0.170,  training accuray =  0.945,  val loss =  0.235,  val accuray =  0.922, \n",
      "Epoch 3970, Batch loss =  0.157,  training accuray =  0.984,  val loss =  0.333,  val accuray =  0.859, \n",
      "Epoch 3980, Batch loss =  0.147,  training accuray =  0.977,  val loss =  0.271,  val accuray =  0.883, \n",
      "Epoch 3990, Batch loss =  0.177,  training accuray =  0.953,  val loss =  0.288,  val accuray =  0.891, \n",
      "Epoch 4000, Batch loss =  0.146,  training accuray =  0.969,  val loss =  0.280,  val accuray =  0.898, \n",
      "Epoch 4010, Batch loss =  0.153,  training accuray =  0.961,  val loss =  0.291,  val accuray =  0.922, \n",
      "Epoch 4020, Batch loss =  0.215,  training accuray =  0.930,  val loss =  0.264,  val accuray =  0.906, \n",
      "Epoch 4030, Batch loss =  0.174,  training accuray =  0.945,  val loss =  0.282,  val accuray =  0.914, \n",
      "Epoch 4040, Batch loss =  0.177,  training accuray =  0.953,  val loss =  0.206,  val accuray =  0.945, \n",
      "Epoch 4050, Batch loss =  0.133,  training accuray =  0.977,  val loss =  0.243,  val accuray =  0.914, \n",
      "Epoch 4060, Batch loss =  0.138,  training accuray =  0.969,  val loss =  0.255,  val accuray =  0.891, \n",
      "Epoch 4070, Batch loss =  0.164,  training accuray =  0.969,  val loss =  0.329,  val accuray =  0.859, \n",
      "Epoch 4080, Batch loss =  0.123,  training accuray =  0.977,  val loss =  0.248,  val accuray =  0.883, \n",
      "Epoch 4090, Batch loss =  0.158,  training accuray =  0.945,  val loss =  0.296,  val accuray =  0.883, \n",
      "Epoch 4100, Batch loss =  0.143,  training accuray =  0.977,  val loss =  0.221,  val accuray =  0.914, \n",
      "Epoch 4110, Batch loss =  0.176,  training accuray =  0.961,  val loss =  0.287,  val accuray =  0.898, \n",
      "Epoch 4120, Batch loss =  0.171,  training accuray =  0.961,  val loss =  0.262,  val accuray =  0.891, \n",
      "Epoch 4130, Batch loss =  0.139,  training accuray =  0.977,  val loss =  0.279,  val accuray =  0.883, \n",
      "Epoch 4140, Batch loss =  0.165,  training accuray =  0.945,  val loss =  0.208,  val accuray =  0.945, \n",
      "Epoch 4150, Batch loss =  0.120,  training accuray =  0.984,  val loss =  0.235,  val accuray =  0.898, \n",
      "Epoch 4160, Batch loss =  0.177,  training accuray =  0.961,  val loss =  0.293,  val accuray =  0.844, \n",
      "Epoch 4170, Batch loss =  0.141,  training accuray =  0.969,  val loss =  0.275,  val accuray =  0.859, \n",
      "Epoch 4180, Batch loss =  0.143,  training accuray =  0.977,  val loss =  0.272,  val accuray =  0.930, \n",
      "Epoch 4190, Batch loss =  0.169,  training accuray =  0.938,  val loss =  0.305,  val accuray =  0.906, \n",
      "Epoch 4200, Batch loss =  0.145,  training accuray =  0.961,  val loss =  0.280,  val accuray =  0.898, \n",
      "Epoch 4210, Batch loss =  0.159,  training accuray =  0.969,  val loss =  0.256,  val accuray =  0.891, \n",
      "Epoch 4220, Batch loss =  0.143,  training accuray =  0.969,  val loss =  0.219,  val accuray =  0.938, \n",
      "Epoch 4230, Batch loss =  0.131,  training accuray =  0.977,  val loss =  0.255,  val accuray =  0.914, \n",
      "Epoch 4240, Batch loss =  0.133,  training accuray =  0.969,  val loss =  0.311,  val accuray =  0.852, \n",
      "Epoch 4250, Batch loss =  0.122,  training accuray =  0.977,  val loss =  0.240,  val accuray =  0.883, \n",
      "Epoch 4260, Batch loss =  0.182,  training accuray =  0.945,  val loss =  0.229,  val accuray =  0.914, \n",
      "Epoch 4270, Batch loss =  0.159,  training accuray =  0.977,  val loss =  0.276,  val accuray =  0.898, \n",
      "Epoch 4280, Batch loss =  0.201,  training accuray =  0.938,  val loss =  0.270,  val accuray =  0.898, \n",
      "Epoch 4290, Batch loss =  0.153,  training accuray =  0.977,  val loss =  0.249,  val accuray =  0.922, \n",
      "Epoch 4300, Batch loss =  0.104,  training accuray =  0.992,  val loss =  0.257,  val accuray =  0.906, \n",
      "Epoch 4310, Batch loss =  0.155,  training accuray =  0.938,  val loss =  0.280,  val accuray =  0.906, \n",
      "Epoch 4320, Batch loss =  0.126,  training accuray =  0.984,  val loss =  0.315,  val accuray =  0.867, \n",
      "Epoch 4330, Batch loss =  0.113,  training accuray =  0.977,  val loss =  0.280,  val accuray =  0.891, \n",
      "Epoch 4340, Batch loss =  0.156,  training accuray =  0.961,  val loss =  0.265,  val accuray =  0.914, \n",
      "Epoch 4350, Batch loss =  0.139,  training accuray =  0.969,  val loss =  0.225,  val accuray =  0.898, \n",
      "Epoch 4360, Batch loss =  0.132,  training accuray =  0.961,  val loss =  0.276,  val accuray =  0.859, \n",
      "Epoch 4370, Batch loss =  0.154,  training accuray =  0.961,  val loss =  0.212,  val accuray =  0.930, \n",
      "Epoch 4380, Batch loss =  0.132,  training accuray =  0.977,  val loss =  0.281,  val accuray =  0.906, \n",
      "Epoch 4390, Batch loss =  0.193,  training accuray =  0.930,  val loss =  0.241,  val accuray =  0.906, \n",
      "Epoch 4400, Batch loss =  0.127,  training accuray =  0.977,  val loss =  0.250,  val accuray =  0.906, \n",
      "Epoch 4410, Batch loss =  0.138,  training accuray =  0.977,  val loss =  0.257,  val accuray =  0.906, \n",
      "Epoch 4420, Batch loss =  0.157,  training accuray =  0.961,  val loss =  0.249,  val accuray =  0.945, \n",
      "Epoch 4430, Batch loss =  0.168,  training accuray =  0.969,  val loss =  0.192,  val accuray =  0.945, \n",
      "Epoch 4440, Batch loss =  0.160,  training accuray =  0.961,  val loss =  0.213,  val accuray =  0.945, \n",
      "Epoch 4450, Batch loss =  0.142,  training accuray =  0.961,  val loss =  0.260,  val accuray =  0.906, \n",
      "Epoch 4460, Batch loss =  0.167,  training accuray =  0.945,  val loss =  0.256,  val accuray =  0.922, \n",
      "Epoch 4470, Batch loss =  0.149,  training accuray =  0.953,  val loss =  0.222,  val accuray =  0.922, \n",
      "Epoch 4480, Batch loss =  0.112,  training accuray =  0.984,  val loss =  0.185,  val accuray =  0.961, \n",
      "Epoch 4490, Batch loss =  0.165,  training accuray =  0.938,  val loss =  0.309,  val accuray =  0.914, \n",
      "Epoch 4500, Batch loss =  0.159,  training accuray =  0.953,  val loss =  0.252,  val accuray =  0.906, \n",
      "Epoch 4510, Batch loss =  0.127,  training accuray =  0.969,  val loss =  0.229,  val accuray =  0.922, \n",
      "Epoch 4520, Batch loss =  0.164,  training accuray =  0.953,  val loss =  0.222,  val accuray =  0.883, \n",
      "Epoch 4530, Batch loss =  0.144,  training accuray =  0.961,  val loss =  0.272,  val accuray =  0.906, \n",
      "Epoch 4540, Batch loss =  0.110,  training accuray =  0.984,  val loss =  0.273,  val accuray =  0.906, \n",
      "Epoch 4550, Batch loss =  0.111,  training accuray =  0.977,  val loss =  0.282,  val accuray =  0.875, \n",
      "Epoch 4560, Batch loss =  0.126,  training accuray =  0.953,  val loss =  0.222,  val accuray =  0.922, \n",
      "Epoch 4570, Batch loss =  0.133,  training accuray =  0.961,  val loss =  0.314,  val accuray =  0.859, \n",
      "Epoch 4580, Batch loss =  0.114,  training accuray =  0.961,  val loss =  0.289,  val accuray =  0.891, \n",
      "Epoch 4590, Batch loss =  0.157,  training accuray =  0.961,  val loss =  0.232,  val accuray =  0.914, \n",
      "Epoch 4600, Batch loss =  0.129,  training accuray =  0.977,  val loss =  0.263,  val accuray =  0.891, \n",
      "Epoch 4610, Batch loss =  0.145,  training accuray =  0.969,  val loss =  0.248,  val accuray =  0.898, \n",
      "Epoch 4620, Batch loss =  0.109,  training accuray =  0.977,  val loss =  0.327,  val accuray =  0.859, \n",
      "Epoch 4630, Batch loss =  0.122,  training accuray =  0.984,  val loss =  0.244,  val accuray =  0.883, \n",
      "Epoch 4640, Batch loss =  0.126,  training accuray =  0.969,  val loss =  0.210,  val accuray =  0.898, \n",
      "Epoch 4650, Batch loss =  0.104,  training accuray =  0.969,  val loss =  0.246,  val accuray =  0.914, \n",
      "Epoch 4660, Batch loss =  0.138,  training accuray =  0.977,  val loss =  0.289,  val accuray =  0.906, \n",
      "Epoch 4670, Batch loss =  0.159,  training accuray =  0.961,  val loss =  0.179,  val accuray =  0.945, \n",
      "Epoch 4680, Batch loss =  0.156,  training accuray =  0.945,  val loss =  0.267,  val accuray =  0.906, \n",
      "Epoch 4690, Batch loss =  0.136,  training accuray =  0.961,  val loss =  0.231,  val accuray =  0.930, \n",
      "Epoch 4700, Batch loss =  0.076,  training accuray =  0.992,  val loss =  0.284,  val accuray =  0.859, \n",
      "Epoch 4710, Batch loss =  0.115,  training accuray =  0.969,  val loss =  0.209,  val accuray =  0.938, \n",
      "Epoch 4720, Batch loss =  0.143,  training accuray =  0.969,  val loss =  0.266,  val accuray =  0.906, \n",
      "Epoch 4730, Batch loss =  0.119,  training accuray =  0.969,  val loss =  0.223,  val accuray =  0.930, \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 4740, Batch loss =  0.136,  training accuray =  0.969,  val loss =  0.197,  val accuray =  0.914, \n",
      "Epoch 4750, Batch loss =  0.175,  training accuray =  0.953,  val loss =  0.188,  val accuray =  0.945, \n",
      "Epoch 4760, Batch loss =  0.119,  training accuray =  0.977,  val loss =  0.275,  val accuray =  0.898, \n",
      "Epoch 4770, Batch loss =  0.144,  training accuray =  0.977,  val loss =  0.248,  val accuray =  0.875, \n",
      "Epoch 4780, Batch loss =  0.122,  training accuray =  0.969,  val loss =  0.186,  val accuray =  0.945, \n",
      "Epoch 4790, Batch loss =  0.102,  training accuray =  0.977,  val loss =  0.320,  val accuray =  0.883, \n",
      "Epoch 4800, Batch loss =  0.123,  training accuray =  0.953,  val loss =  0.266,  val accuray =  0.898, \n",
      "Epoch 4810, Batch loss =  0.104,  training accuray =  0.984,  val loss =  0.239,  val accuray =  0.914, \n",
      "Epoch 4820, Batch loss =  0.170,  training accuray =  0.945,  val loss =  0.247,  val accuray =  0.914, \n",
      "Epoch 4830, Batch loss =  0.096,  training accuray =  0.992,  val loss =  0.251,  val accuray =  0.898, \n",
      "Epoch 4840, Batch loss =  0.109,  training accuray =  0.984,  val loss =  0.205,  val accuray =  0.945, \n",
      "Epoch 4850, Batch loss =  0.147,  training accuray =  0.953,  val loss =  0.237,  val accuray =  0.906, \n",
      "Epoch 4860, Batch loss =  0.099,  training accuray =  0.977,  val loss =  0.221,  val accuray =  0.930, \n",
      "Epoch 4870, Batch loss =  0.132,  training accuray =  0.961,  val loss =  0.271,  val accuray =  0.891, \n",
      "Epoch 4880, Batch loss =  0.132,  training accuray =  0.969,  val loss =  0.225,  val accuray =  0.938, \n",
      "Epoch 4890, Batch loss =  0.105,  training accuray =  0.977,  val loss =  0.261,  val accuray =  0.891, \n",
      "Epoch 4900, Batch loss =  0.118,  training accuray =  0.984,  val loss =  0.240,  val accuray =  0.922, \n",
      "Epoch 4910, Batch loss =  0.111,  training accuray =  0.984,  val loss =  0.230,  val accuray =  0.938, \n",
      "Epoch 4920, Batch loss =  0.138,  training accuray =  0.961,  val loss =  0.204,  val accuray =  0.938, \n",
      "Epoch 4930, Batch loss =  0.115,  training accuray =  0.984,  val loss =  0.239,  val accuray =  0.922, \n",
      "Epoch 4940, Batch loss =  0.170,  training accuray =  0.938,  val loss =  0.213,  val accuray =  0.930, \n",
      "Epoch 4950, Batch loss =  0.110,  training accuray =  0.969,  val loss =  0.174,  val accuray =  0.945, \n",
      "Epoch 4960, Batch loss =  0.125,  training accuray =  0.977,  val loss =  0.243,  val accuray =  0.867, \n",
      "Epoch 4970, Batch loss =  0.125,  training accuray =  0.969,  val loss =  0.235,  val accuray =  0.914, \n",
      "Epoch 4980, Batch loss =  0.124,  training accuray =  0.977,  val loss =  0.282,  val accuray =  0.898, \n",
      "Epoch 4990, Batch loss =  0.127,  training accuray =  0.969,  val loss =  0.256,  val accuray =  0.898, \n",
      "Epoch 5000, Batch loss =  0.115,  training accuray =  0.992,  val loss =  0.291,  val accuray =  0.891, \n",
      "Epoch 5010, Batch loss =  0.119,  training accuray =  0.977,  val loss =  0.272,  val accuray =  0.898, \n",
      "Epoch 5020, Batch loss =  0.106,  training accuray =  0.984,  val loss =  0.280,  val accuray =  0.883, \n",
      "Epoch 5030, Batch loss =  0.121,  training accuray =  0.977,  val loss =  0.207,  val accuray =  0.953, \n",
      "Epoch 5040, Batch loss =  0.118,  training accuray =  0.977,  val loss =  0.234,  val accuray =  0.906, \n",
      "Epoch 5050, Batch loss =  0.101,  training accuray =  0.984,  val loss =  0.249,  val accuray =  0.914, \n",
      "Epoch 5060, Batch loss =  0.104,  training accuray =  0.969,  val loss =  0.186,  val accuray =  0.945, \n",
      "Epoch 5070, Batch loss =  0.106,  training accuray =  0.977,  val loss =  0.228,  val accuray =  0.898, \n",
      "Epoch 5080, Batch loss =  0.135,  training accuray =  0.969,  val loss =  0.171,  val accuray =  0.984, \n",
      "Epoch 5090, Batch loss =  0.106,  training accuray =  0.984,  val loss =  0.220,  val accuray =  0.938, \n",
      "Epoch 5100, Batch loss =  0.118,  training accuray =  0.969,  val loss =  0.239,  val accuray =  0.930, \n",
      "Epoch 5110, Batch loss =  0.116,  training accuray =  0.977,  val loss =  0.255,  val accuray =  0.898, \n",
      "Epoch 5120, Batch loss =  0.120,  training accuray =  0.969,  val loss =  0.219,  val accuray =  0.938, \n",
      "Epoch 5130, Batch loss =  0.132,  training accuray =  0.977,  val loss =  0.184,  val accuray =  0.922, \n",
      "Epoch 5140, Batch loss =  0.087,  training accuray =  0.984,  val loss =  0.209,  val accuray =  0.906, \n",
      "Epoch 5150, Batch loss =  0.115,  training accuray =  0.977,  val loss =  0.250,  val accuray =  0.930, \n",
      "Epoch 5160, Batch loss =  0.112,  training accuray =  0.969,  val loss =  0.225,  val accuray =  0.938, \n",
      "Epoch 5170, Batch loss =  0.103,  training accuray =  0.984,  val loss =  0.191,  val accuray =  0.945, \n",
      "Epoch 5180, Batch loss =  0.082,  training accuray =  0.984,  val loss =  0.199,  val accuray =  0.938, \n",
      "Epoch 5190, Batch loss =  0.107,  training accuray =  0.992,  val loss =  0.255,  val accuray =  0.891, \n",
      "Epoch 5200, Batch loss =  0.107,  training accuray =  0.969,  val loss =  0.217,  val accuray =  0.938, \n",
      "Epoch 5210, Batch loss =  0.082,  training accuray =  0.984,  val loss =  0.192,  val accuray =  0.930, \n",
      "Epoch 5220, Batch loss =  0.113,  training accuray =  0.977,  val loss =  0.280,  val accuray =  0.859, \n",
      "Epoch 5230, Batch loss =  0.121,  training accuray =  0.953,  val loss =  0.213,  val accuray =  0.930, \n",
      "Epoch 5240, Batch loss =  0.094,  training accuray =  0.984,  val loss =  0.242,  val accuray =  0.906, \n",
      "Epoch 5250, Batch loss =  0.116,  training accuray =  0.984,  val loss =  0.223,  val accuray =  0.914, \n",
      "Epoch 5260, Batch loss =  0.084,  training accuray =  0.992,  val loss =  0.293,  val accuray =  0.906, \n",
      "Epoch 5270, Batch loss =  0.072,  training accuray =  1.000,  val loss =  0.273,  val accuray =  0.914, \n",
      "Epoch 5280, Batch loss =  0.129,  training accuray =  0.961,  val loss =  0.251,  val accuray =  0.938, \n",
      "Epoch 5290, Batch loss =  0.105,  training accuray =  0.961,  val loss =  0.269,  val accuray =  0.891, \n",
      "Epoch 5300, Batch loss =  0.109,  training accuray =  0.969,  val loss =  0.189,  val accuray =  0.938, \n",
      "Epoch 5310, Batch loss =  0.140,  training accuray =  0.945,  val loss =  0.256,  val accuray =  0.922, \n",
      "Epoch 5320, Batch loss =  0.095,  training accuray =  0.984,  val loss =  0.235,  val accuray =  0.906, \n",
      "Epoch 5330, Batch loss =  0.085,  training accuray =  1.000,  val loss =  0.220,  val accuray =  0.922, \n",
      "Epoch 5340, Batch loss =  0.099,  training accuray =  0.992,  val loss =  0.248,  val accuray =  0.906, \n",
      "Epoch 5350, Batch loss =  0.072,  training accuray =  0.992,  val loss =  0.238,  val accuray =  0.938, \n",
      "Epoch 5360, Batch loss =  0.104,  training accuray =  0.969,  val loss =  0.225,  val accuray =  0.914, \n",
      "Epoch 5370, Batch loss =  0.077,  training accuray =  1.000,  val loss =  0.272,  val accuray =  0.898, \n",
      "Epoch 5380, Batch loss =  0.130,  training accuray =  0.969,  val loss =  0.182,  val accuray =  0.930, \n",
      "Epoch 5390, Batch loss =  0.073,  training accuray =  1.000,  val loss =  0.240,  val accuray =  0.906, \n",
      "Epoch 5400, Batch loss =  0.085,  training accuray =  0.992,  val loss =  0.207,  val accuray =  0.930, \n",
      "Epoch 5410, Batch loss =  0.097,  training accuray =  0.977,  val loss =  0.186,  val accuray =  0.945, \n",
      "Epoch 5420, Batch loss =  0.094,  training accuray =  0.992,  val loss =  0.268,  val accuray =  0.891, \n",
      "Epoch 5430, Batch loss =  0.115,  training accuray =  0.961,  val loss =  0.202,  val accuray =  0.945, \n",
      "Epoch 5440, Batch loss =  0.125,  training accuray =  0.992,  val loss =  0.180,  val accuray =  0.938, \n",
      "Epoch 5450, Batch loss =  0.101,  training accuray =  0.977,  val loss =  0.246,  val accuray =  0.914, \n",
      "Epoch 5460, Batch loss =  0.117,  training accuray =  0.961,  val loss =  0.288,  val accuray =  0.875, \n",
      "Epoch 5470, Batch loss =  0.124,  training accuray =  0.961,  val loss =  0.218,  val accuray =  0.922, \n",
      "Epoch 5480, Batch loss =  0.101,  training accuray =  0.977,  val loss =  0.199,  val accuray =  0.938, \n",
      "Epoch 5490, Batch loss =  0.095,  training accuray =  0.977,  val loss =  0.249,  val accuray =  0.906, \n",
      "Epoch 5500, Batch loss =  0.105,  training accuray =  0.977,  val loss =  0.264,  val accuray =  0.867, \n",
      "Epoch 5510, Batch loss =  0.138,  training accuray =  0.969,  val loss =  0.215,  val accuray =  0.930, \n",
      "Epoch 5520, Batch loss =  0.137,  training accuray =  0.961,  val loss =  0.264,  val accuray =  0.898, \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 5530, Batch loss =  0.096,  training accuray =  0.984,  val loss =  0.196,  val accuray =  0.930, \n",
      "Epoch 5540, Batch loss =  0.096,  training accuray =  1.000,  val loss =  0.216,  val accuray =  0.953, \n",
      "Epoch 5550, Batch loss =  0.098,  training accuray =  0.977,  val loss =  0.270,  val accuray =  0.906, \n",
      "Epoch 5560, Batch loss =  0.084,  training accuray =  0.969,  val loss =  0.208,  val accuray =  0.922, \n",
      "Epoch 5570, Batch loss =  0.129,  training accuray =  0.961,  val loss =  0.195,  val accuray =  0.930, \n",
      "Epoch 5580, Batch loss =  0.102,  training accuray =  0.984,  val loss =  0.245,  val accuray =  0.883, \n",
      "Epoch 5590, Batch loss =  0.094,  training accuray =  0.984,  val loss =  0.220,  val accuray =  0.922, \n",
      "Epoch 5600, Batch loss =  0.090,  training accuray =  0.984,  val loss =  0.234,  val accuray =  0.906, \n",
      "Epoch 5610, Batch loss =  0.105,  training accuray =  0.969,  val loss =  0.262,  val accuray =  0.906, \n",
      "Epoch 5620, Batch loss =  0.073,  training accuray =  0.992,  val loss =  0.223,  val accuray =  0.930, \n",
      "Epoch 5630, Batch loss =  0.102,  training accuray =  0.969,  val loss =  0.271,  val accuray =  0.883, \n",
      "Epoch 5640, Batch loss =  0.113,  training accuray =  0.969,  val loss =  0.281,  val accuray =  0.891, \n",
      "Epoch 5650, Batch loss =  0.092,  training accuray =  0.992,  val loss =  0.210,  val accuray =  0.914, \n",
      "Epoch 5660, Batch loss =  0.082,  training accuray =  0.977,  val loss =  0.208,  val accuray =  0.945, \n",
      "Epoch 5670, Batch loss =  0.098,  training accuray =  0.977,  val loss =  0.174,  val accuray =  0.922, \n",
      "Epoch 5680, Batch loss =  0.115,  training accuray =  0.961,  val loss =  0.254,  val accuray =  0.891, \n",
      "Epoch 5690, Batch loss =  0.114,  training accuray =  0.977,  val loss =  0.224,  val accuray =  0.898, \n",
      "Epoch 5700, Batch loss =  0.091,  training accuray =  0.992,  val loss =  0.308,  val accuray =  0.883, \n",
      "Epoch 5710, Batch loss =  0.083,  training accuray =  0.984,  val loss =  0.230,  val accuray =  0.938, \n",
      "Epoch 5720, Batch loss =  0.059,  training accuray =  0.992,  val loss =  0.274,  val accuray =  0.867, \n",
      "Epoch 5730, Batch loss =  0.065,  training accuray =  0.984,  val loss =  0.216,  val accuray =  0.922, \n",
      "Epoch 5740, Batch loss =  0.077,  training accuray =  0.992,  val loss =  0.214,  val accuray =  0.922, \n",
      "Epoch 5750, Batch loss =  0.088,  training accuray =  0.992,  val loss =  0.217,  val accuray =  0.914, \n",
      "Epoch 5760, Batch loss =  0.093,  training accuray =  0.984,  val loss =  0.232,  val accuray =  0.906, \n",
      "Epoch 5770, Batch loss =  0.091,  training accuray =  0.977,  val loss =  0.202,  val accuray =  0.945, \n",
      "Epoch 5780, Batch loss =  0.096,  training accuray =  0.977,  val loss =  0.283,  val accuray =  0.898, \n",
      "Epoch 5790, Batch loss =  0.070,  training accuray =  0.992,  val loss =  0.165,  val accuray =  0.953, \n",
      "Epoch 5800, Batch loss =  0.130,  training accuray =  0.969,  val loss =  0.202,  val accuray =  0.945, \n",
      "Epoch 5810, Batch loss =  0.085,  training accuray =  0.992,  val loss =  0.284,  val accuray =  0.883, \n",
      "Epoch 5820, Batch loss =  0.090,  training accuray =  0.977,  val loss =  0.207,  val accuray =  0.914, \n",
      "Epoch 5830, Batch loss =  0.094,  training accuray =  0.992,  val loss =  0.205,  val accuray =  0.945, \n",
      "Epoch 5840, Batch loss =  0.098,  training accuray =  0.984,  val loss =  0.278,  val accuray =  0.875, \n",
      "Epoch 5850, Batch loss =  0.077,  training accuray =  0.984,  val loss =  0.204,  val accuray =  0.922, \n",
      "Epoch 5860, Batch loss =  0.128,  training accuray =  0.977,  val loss =  0.216,  val accuray =  0.914, \n",
      "Epoch 5870, Batch loss =  0.099,  training accuray =  0.977,  val loss =  0.179,  val accuray =  0.938, \n",
      "Epoch 5880, Batch loss =  0.086,  training accuray =  0.969,  val loss =  0.234,  val accuray =  0.906, \n",
      "Epoch 5890, Batch loss =  0.095,  training accuray =  0.992,  val loss =  0.175,  val accuray =  0.953, \n",
      "Epoch 5900, Batch loss =  0.068,  training accuray =  0.992,  val loss =  0.233,  val accuray =  0.922, \n",
      "Epoch 5910, Batch loss =  0.071,  training accuray =  0.992,  val loss =  0.201,  val accuray =  0.906, \n",
      "Epoch 5920, Batch loss =  0.066,  training accuray =  0.992,  val loss =  0.184,  val accuray =  0.938, \n",
      "Epoch 5930, Batch loss =  0.085,  training accuray =  0.977,  val loss =  0.238,  val accuray =  0.906, \n",
      "Epoch 5940, Batch loss =  0.073,  training accuray =  1.000,  val loss =  0.206,  val accuray =  0.938, \n",
      "Epoch 5950, Batch loss =  0.062,  training accuray =  0.992,  val loss =  0.197,  val accuray =  0.930, \n",
      "Epoch 5960, Batch loss =  0.076,  training accuray =  0.977,  val loss =  0.189,  val accuray =  0.945, \n",
      "Epoch 5970, Batch loss =  0.074,  training accuray =  0.984,  val loss =  0.221,  val accuray =  0.898, \n",
      "Epoch 5980, Batch loss =  0.097,  training accuray =  0.961,  val loss =  0.236,  val accuray =  0.906, \n",
      "Epoch 5990, Batch loss =  0.086,  training accuray =  0.984,  val loss =  0.230,  val accuray =  0.922, \n",
      "Epoch 6000, Batch loss =  0.061,  training accuray =  0.992,  val loss =  0.178,  val accuray =  0.945, \n",
      "Epoch 6010, Batch loss =  0.084,  training accuray =  0.992,  val loss =  0.232,  val accuray =  0.906, \n",
      "Epoch 6020, Batch loss =  0.066,  training accuray =  0.992,  val loss =  0.192,  val accuray =  0.938, \n",
      "Epoch 6030, Batch loss =  0.098,  training accuray =  0.977,  val loss =  0.258,  val accuray =  0.914, \n",
      "Epoch 6040, Batch loss =  0.084,  training accuray =  0.984,  val loss =  0.200,  val accuray =  0.930, \n",
      "Epoch 6050, Batch loss =  0.084,  training accuray =  0.977,  val loss =  0.178,  val accuray =  0.953, \n",
      "Epoch 6060, Batch loss =  0.080,  training accuray =  0.977,  val loss =  0.262,  val accuray =  0.891, \n",
      "Epoch 6070, Batch loss =  0.081,  training accuray =  0.977,  val loss =  0.243,  val accuray =  0.883, \n",
      "Epoch 6080, Batch loss =  0.072,  training accuray =  0.992,  val loss =  0.215,  val accuray =  0.906, \n",
      "Epoch 6090, Batch loss =  0.132,  training accuray =  0.938,  val loss =  0.195,  val accuray =  0.930, \n",
      "Epoch 6100, Batch loss =  0.063,  training accuray =  1.000,  val loss =  0.256,  val accuray =  0.898, \n",
      "Epoch 6110, Batch loss =  0.083,  training accuray =  0.961,  val loss =  0.214,  val accuray =  0.906, \n",
      "Epoch 6120, Batch loss =  0.067,  training accuray =  1.000,  val loss =  0.220,  val accuray =  0.938, \n",
      "Epoch 6130, Batch loss =  0.105,  training accuray =  0.977,  val loss =  0.232,  val accuray =  0.922, \n",
      "Epoch 6140, Batch loss =  0.078,  training accuray =  0.992,  val loss =  0.225,  val accuray =  0.930, \n",
      "Epoch 6150, Batch loss =  0.075,  training accuray =  0.992,  val loss =  0.241,  val accuray =  0.906, \n",
      "Epoch 6160, Batch loss =  0.088,  training accuray =  0.984,  val loss =  0.185,  val accuray =  0.945, \n",
      "Epoch 6170, Batch loss =  0.088,  training accuray =  0.984,  val loss =  0.255,  val accuray =  0.883, \n",
      "Epoch 6180, Batch loss =  0.076,  training accuray =  0.992,  val loss =  0.215,  val accuray =  0.914, \n",
      "Epoch 6190, Batch loss =  0.075,  training accuray =  0.984,  val loss =  0.166,  val accuray =  0.953, \n",
      "Epoch 6200, Batch loss =  0.091,  training accuray =  0.992,  val loss =  0.217,  val accuray =  0.914, \n",
      "Epoch 6210, Batch loss =  0.083,  training accuray =  0.992,  val loss =  0.187,  val accuray =  0.930, \n",
      "Epoch 6220, Batch loss =  0.092,  training accuray =  0.992,  val loss =  0.232,  val accuray =  0.938, \n",
      "Epoch 6230, Batch loss =  0.082,  training accuray =  0.992,  val loss =  0.153,  val accuray =  0.930, \n",
      "Epoch 6240, Batch loss =  0.056,  training accuray =  0.984,  val loss =  0.217,  val accuray =  0.914, \n",
      "Epoch 6250, Batch loss =  0.062,  training accuray =  0.992,  val loss =  0.231,  val accuray =  0.906, \n",
      "Epoch 6260, Batch loss =  0.070,  training accuray =  0.984,  val loss =  0.229,  val accuray =  0.914, \n",
      "Epoch 6270, Batch loss =  0.063,  training accuray =  0.992,  val loss =  0.205,  val accuray =  0.930, \n",
      "Epoch 6280, Batch loss =  0.097,  training accuray =  0.969,  val loss =  0.219,  val accuray =  0.906, \n",
      "Epoch 6290, Batch loss =  0.054,  training accuray =  1.000,  val loss =  0.241,  val accuray =  0.922, \n",
      "Epoch 6300, Batch loss =  0.080,  training accuray =  0.984,  val loss =  0.185,  val accuray =  0.922, \n",
      "Epoch 6310, Batch loss =  0.061,  training accuray =  0.992,  val loss =  0.163,  val accuray =  0.961, \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 6320, Batch loss =  0.065,  training accuray =  0.992,  val loss =  0.216,  val accuray =  0.930, \n",
      "Epoch 6330, Batch loss =  0.071,  training accuray =  0.984,  val loss =  0.216,  val accuray =  0.922, \n",
      "Epoch 6340, Batch loss =  0.073,  training accuray =  0.992,  val loss =  0.143,  val accuray =  0.961, \n",
      "Epoch 6350, Batch loss =  0.070,  training accuray =  0.992,  val loss =  0.196,  val accuray =  0.922, \n",
      "Epoch 6360, Batch loss =  0.094,  training accuray =  0.977,  val loss =  0.220,  val accuray =  0.922, \n",
      "Epoch 6370, Batch loss =  0.058,  training accuray =  0.992,  val loss =  0.230,  val accuray =  0.938, \n",
      "Epoch 6380, Batch loss =  0.075,  training accuray =  0.992,  val loss =  0.201,  val accuray =  0.930, \n",
      "Epoch 6390, Batch loss =  0.073,  training accuray =  0.984,  val loss =  0.175,  val accuray =  0.945, \n",
      "Epoch 6400, Batch loss =  0.106,  training accuray =  0.984,  val loss =  0.213,  val accuray =  0.914, \n",
      "Epoch 6410, Batch loss =  0.079,  training accuray =  0.977,  val loss =  0.193,  val accuray =  0.922, \n",
      "Epoch 6420, Batch loss =  0.067,  training accuray =  0.992,  val loss =  0.166,  val accuray =  0.945, \n",
      "Epoch 6430, Batch loss =  0.072,  training accuray =  0.992,  val loss =  0.169,  val accuray =  0.945, \n",
      "Epoch 6440, Batch loss =  0.082,  training accuray =  0.969,  val loss =  0.195,  val accuray =  0.930, \n",
      "Epoch 6450, Batch loss =  0.078,  training accuray =  0.984,  val loss =  0.215,  val accuray =  0.914, \n",
      "Epoch 6460, Batch loss =  0.076,  training accuray =  0.984,  val loss =  0.175,  val accuray =  0.945, \n",
      "Epoch 6470, Batch loss =  0.074,  training accuray =  0.984,  val loss =  0.237,  val accuray =  0.906, \n",
      "Epoch 6480, Batch loss =  0.053,  training accuray =  1.000,  val loss =  0.262,  val accuray =  0.898, \n",
      "Epoch 6490, Batch loss =  0.050,  training accuray =  1.000,  val loss =  0.249,  val accuray =  0.891, \n",
      "Epoch 6500, Batch loss =  0.091,  training accuray =  0.984,  val loss =  0.202,  val accuray =  0.914, \n",
      "Epoch 6510, Batch loss =  0.056,  training accuray =  1.000,  val loss =  0.223,  val accuray =  0.922, \n",
      "Epoch 6520, Batch loss =  0.089,  training accuray =  0.984,  val loss =  0.242,  val accuray =  0.906, \n",
      "Epoch 6530, Batch loss =  0.066,  training accuray =  0.992,  val loss =  0.236,  val accuray =  0.906, \n",
      "Epoch 6540, Batch loss =  0.053,  training accuray =  1.000,  val loss =  0.226,  val accuray =  0.891, \n",
      "Epoch 6550, Batch loss =  0.056,  training accuray =  0.992,  val loss =  0.198,  val accuray =  0.906, \n",
      "Epoch 6560, Batch loss =  0.063,  training accuray =  0.984,  val loss =  0.216,  val accuray =  0.938, \n",
      "Epoch 6570, Batch loss =  0.073,  training accuray =  0.984,  val loss =  0.240,  val accuray =  0.898, \n",
      "Epoch 6580, Batch loss =  0.082,  training accuray =  0.984,  val loss =  0.199,  val accuray =  0.930, \n",
      "Epoch 6590, Batch loss =  0.093,  training accuray =  0.977,  val loss =  0.190,  val accuray =  0.945, \n",
      "Epoch 6600, Batch loss =  0.074,  training accuray =  0.984,  val loss =  0.153,  val accuray =  0.938, \n",
      "Epoch 6610, Batch loss =  0.056,  training accuray =  0.992,  val loss =  0.193,  val accuray =  0.945, \n",
      "Epoch 6620, Batch loss =  0.084,  training accuray =  0.977,  val loss =  0.181,  val accuray =  0.922, \n",
      "Epoch 6630, Batch loss =  0.075,  training accuray =  0.992,  val loss =  0.206,  val accuray =  0.906, \n",
      "Epoch 6640, Batch loss =  0.069,  training accuray =  0.984,  val loss =  0.169,  val accuray =  0.945, \n",
      "Epoch 6650, Batch loss =  0.081,  training accuray =  0.977,  val loss =  0.222,  val accuray =  0.938, \n",
      "Epoch 6660, Batch loss =  0.062,  training accuray =  1.000,  val loss =  0.162,  val accuray =  0.938, \n",
      "Epoch 6670, Batch loss =  0.050,  training accuray =  1.000,  val loss =  0.245,  val accuray =  0.883, \n",
      "Epoch 6680, Batch loss =  0.084,  training accuray =  0.977,  val loss =  0.244,  val accuray =  0.914, \n",
      "Epoch 6690, Batch loss =  0.093,  training accuray =  0.992,  val loss =  0.169,  val accuray =  0.945, \n",
      "Epoch 6700, Batch loss =  0.072,  training accuray =  0.992,  val loss =  0.282,  val accuray =  0.898, \n",
      "Epoch 6710, Batch loss =  0.054,  training accuray =  0.984,  val loss =  0.161,  val accuray =  0.945, \n",
      "Epoch 6720, Batch loss =  0.060,  training accuray =  0.992,  val loss =  0.231,  val accuray =  0.930, \n",
      "Epoch 6730, Batch loss =  0.098,  training accuray =  0.977,  val loss =  0.163,  val accuray =  0.953, \n",
      "Epoch 6740, Batch loss =  0.077,  training accuray =  0.984,  val loss =  0.172,  val accuray =  0.930, \n",
      "Epoch 6750, Batch loss =  0.053,  training accuray =  0.992,  val loss =  0.224,  val accuray =  0.906, \n",
      "Epoch 6760, Batch loss =  0.054,  training accuray =  1.000,  val loss =  0.205,  val accuray =  0.906, \n",
      "Epoch 6770, Batch loss =  0.069,  training accuray =  0.984,  val loss =  0.220,  val accuray =  0.883, \n",
      "Epoch 6780, Batch loss =  0.041,  training accuray =  1.000,  val loss =  0.253,  val accuray =  0.898, \n",
      "Epoch 6790, Batch loss =  0.066,  training accuray =  1.000,  val loss =  0.205,  val accuray =  0.953, \n",
      "Epoch 6800, Batch loss =  0.052,  training accuray =  0.992,  val loss =  0.191,  val accuray =  0.922, \n",
      "Epoch 6810, Batch loss =  0.049,  training accuray =  1.000,  val loss =  0.194,  val accuray =  0.938, \n",
      "Epoch 6820, Batch loss =  0.061,  training accuray =  1.000,  val loss =  0.190,  val accuray =  0.938, \n",
      "Epoch 6830, Batch loss =  0.049,  training accuray =  1.000,  val loss =  0.335,  val accuray =  0.836, \n",
      "Epoch 6840, Batch loss =  0.047,  training accuray =  1.000,  val loss =  0.228,  val accuray =  0.914, \n",
      "Epoch 6850, Batch loss =  0.049,  training accuray =  1.000,  val loss =  0.169,  val accuray =  0.930, \n",
      "Epoch 6860, Batch loss =  0.076,  training accuray =  0.977,  val loss =  0.164,  val accuray =  0.945, \n",
      "Epoch 6870, Batch loss =  0.056,  training accuray =  1.000,  val loss =  0.179,  val accuray =  0.930, \n",
      "Epoch 6880, Batch loss =  0.037,  training accuray =  1.000,  val loss =  0.236,  val accuray =  0.906, \n",
      "Epoch 6890, Batch loss =  0.103,  training accuray =  0.969,  val loss =  0.222,  val accuray =  0.914, \n",
      "Epoch 6900, Batch loss =  0.068,  training accuray =  0.984,  val loss =  0.199,  val accuray =  0.922, \n",
      "Epoch 6910, Batch loss =  0.071,  training accuray =  1.000,  val loss =  0.214,  val accuray =  0.914, \n",
      "Epoch 6920, Batch loss =  0.048,  training accuray =  1.000,  val loss =  0.216,  val accuray =  0.945, \n",
      "Epoch 6930, Batch loss =  0.086,  training accuray =  0.977,  val loss =  0.181,  val accuray =  0.953, \n",
      "Epoch 6940, Batch loss =  0.051,  training accuray =  0.992,  val loss =  0.220,  val accuray =  0.906, \n",
      "Epoch 6950, Batch loss =  0.062,  training accuray =  1.000,  val loss =  0.202,  val accuray =  0.938, \n",
      "Epoch 6960, Batch loss =  0.057,  training accuray =  0.992,  val loss =  0.169,  val accuray =  0.953, \n",
      "Epoch 6970, Batch loss =  0.060,  training accuray =  0.992,  val loss =  0.180,  val accuray =  0.938, \n",
      "Epoch 6980, Batch loss =  0.057,  training accuray =  1.000,  val loss =  0.189,  val accuray =  0.930, \n",
      "Epoch 6990, Batch loss =  0.063,  training accuray =  0.992,  val loss =  0.248,  val accuray =  0.906, \n",
      "Epoch 7000, Batch loss =  0.049,  training accuray =  1.000,  val loss =  0.209,  val accuray =  0.922, \n",
      "Epoch 7010, Batch loss =  0.070,  training accuray =  0.977,  val loss =  0.188,  val accuray =  0.945, \n",
      "Epoch 7020, Batch loss =  0.058,  training accuray =  1.000,  val loss =  0.204,  val accuray =  0.922, \n",
      "Epoch 7030, Batch loss =  0.074,  training accuray =  0.977,  val loss =  0.158,  val accuray =  0.969, \n",
      "Epoch 7040, Batch loss =  0.046,  training accuray =  1.000,  val loss =  0.219,  val accuray =  0.914, \n",
      "Epoch 7050, Batch loss =  0.084,  training accuray =  0.984,  val loss =  0.241,  val accuray =  0.922, \n",
      "Epoch 7060, Batch loss =  0.041,  training accuray =  0.992,  val loss =  0.212,  val accuray =  0.898, \n",
      "Epoch 7070, Batch loss =  0.065,  training accuray =  0.977,  val loss =  0.213,  val accuray =  0.922, \n",
      "Epoch 7080, Batch loss =  0.040,  training accuray =  1.000,  val loss =  0.173,  val accuray =  0.930, \n",
      "Epoch 7090, Batch loss =  0.058,  training accuray =  0.992,  val loss =  0.183,  val accuray =  0.953, \n",
      "Epoch 7100, Batch loss =  0.053,  training accuray =  0.992,  val loss =  0.205,  val accuray =  0.922, \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 7110, Batch loss =  0.048,  training accuray =  1.000,  val loss =  0.206,  val accuray =  0.914, \n",
      "Epoch 7120, Batch loss =  0.074,  training accuray =  0.984,  val loss =  0.158,  val accuray =  0.930, \n",
      "Epoch 7130, Batch loss =  0.063,  training accuray =  0.992,  val loss =  0.263,  val accuray =  0.898, \n",
      "Epoch 7140, Batch loss =  0.054,  training accuray =  1.000,  val loss =  0.239,  val accuray =  0.891, \n",
      "Epoch 7150, Batch loss =  0.059,  training accuray =  0.984,  val loss =  0.216,  val accuray =  0.938, \n",
      "Epoch 7160, Batch loss =  0.057,  training accuray =  0.992,  val loss =  0.167,  val accuray =  0.945, \n",
      "Epoch 7170, Batch loss =  0.041,  training accuray =  1.000,  val loss =  0.139,  val accuray =  0.953, \n",
      "Epoch 7180, Batch loss =  0.051,  training accuray =  1.000,  val loss =  0.173,  val accuray =  0.945, \n",
      "Epoch 7190, Batch loss =  0.068,  training accuray =  0.992,  val loss =  0.239,  val accuray =  0.922, \n",
      "Epoch 7200, Batch loss =  0.057,  training accuray =  1.000,  val loss =  0.188,  val accuray =  0.938, \n",
      "Epoch 7210, Batch loss =  0.057,  training accuray =  0.992,  val loss =  0.234,  val accuray =  0.914, \n",
      "Epoch 7220, Batch loss =  0.058,  training accuray =  0.977,  val loss =  0.199,  val accuray =  0.922, \n",
      "Epoch 7230, Batch loss =  0.051,  training accuray =  0.992,  val loss =  0.158,  val accuray =  0.953, \n",
      "Epoch 7240, Batch loss =  0.067,  training accuray =  0.992,  val loss =  0.203,  val accuray =  0.906, \n",
      "Epoch 7250, Batch loss =  0.051,  training accuray =  0.992,  val loss =  0.184,  val accuray =  0.938, \n",
      "Epoch 7260, Batch loss =  0.051,  training accuray =  0.992,  val loss =  0.198,  val accuray =  0.922, \n",
      "Epoch 7270, Batch loss =  0.048,  training accuray =  1.000,  val loss =  0.263,  val accuray =  0.891, \n",
      "Epoch 7280, Batch loss =  0.088,  training accuray =  0.977,  val loss =  0.234,  val accuray =  0.898, \n",
      "Epoch 7290, Batch loss =  0.048,  training accuray =  1.000,  val loss =  0.177,  val accuray =  0.922, \n",
      "Epoch 7300, Batch loss =  0.054,  training accuray =  0.992,  val loss =  0.205,  val accuray =  0.938, \n",
      "Epoch 7310, Batch loss =  0.048,  training accuray =  0.992,  val loss =  0.246,  val accuray =  0.898, \n",
      "Epoch 7320, Batch loss =  0.060,  training accuray =  0.992,  val loss =  0.201,  val accuray =  0.922, \n",
      "Epoch 7330, Batch loss =  0.065,  training accuray =  0.992,  val loss =  0.224,  val accuray =  0.922, \n",
      "Epoch 7340, Batch loss =  0.040,  training accuray =  1.000,  val loss =  0.209,  val accuray =  0.938, \n",
      "Epoch 7350, Batch loss =  0.060,  training accuray =  0.984,  val loss =  0.173,  val accuray =  0.930, \n",
      "Epoch 7360, Batch loss =  0.051,  training accuray =  0.992,  val loss =  0.233,  val accuray =  0.906, \n",
      "Epoch 7370, Batch loss =  0.080,  training accuray =  0.977,  val loss =  0.199,  val accuray =  0.914, \n",
      "Epoch 7380, Batch loss =  0.070,  training accuray =  1.000,  val loss =  0.208,  val accuray =  0.922, \n",
      "Epoch 7390, Batch loss =  0.048,  training accuray =  1.000,  val loss =  0.195,  val accuray =  0.945, \n",
      "Epoch 7400, Batch loss =  0.062,  training accuray =  0.984,  val loss =  0.195,  val accuray =  0.938, \n",
      "Epoch 7410, Batch loss =  0.049,  training accuray =  1.000,  val loss =  0.193,  val accuray =  0.953, \n",
      "Epoch 7420, Batch loss =  0.037,  training accuray =  1.000,  val loss =  0.143,  val accuray =  0.953, \n",
      "Epoch 7430, Batch loss =  0.048,  training accuray =  1.000,  val loss =  0.181,  val accuray =  0.930, \n",
      "Epoch 7440, Batch loss =  0.043,  training accuray =  0.992,  val loss =  0.201,  val accuray =  0.945, \n",
      "Epoch 7450, Batch loss =  0.072,  training accuray =  0.984,  val loss =  0.188,  val accuray =  0.930, \n",
      "Epoch 7460, Batch loss =  0.052,  training accuray =  1.000,  val loss =  0.230,  val accuray =  0.898, \n",
      "Epoch 7470, Batch loss =  0.042,  training accuray =  1.000,  val loss =  0.213,  val accuray =  0.914, \n",
      "Epoch 7480, Batch loss =  0.057,  training accuray =  0.992,  val loss =  0.188,  val accuray =  0.938, \n",
      "Epoch 7490, Batch loss =  0.083,  training accuray =  0.977,  val loss =  0.208,  val accuray =  0.922, \n",
      "Epoch 7500, Batch loss =  0.045,  training accuray =  1.000,  val loss =  0.205,  val accuray =  0.922, \n",
      "Epoch 7510, Batch loss =  0.073,  training accuray =  0.992,  val loss =  0.207,  val accuray =  0.953, \n",
      "Epoch 7520, Batch loss =  0.052,  training accuray =  1.000,  val loss =  0.195,  val accuray =  0.922, \n",
      "Epoch 7530, Batch loss =  0.047,  training accuray =  0.992,  val loss =  0.152,  val accuray =  0.961, \n",
      "Epoch 7540, Batch loss =  0.044,  training accuray =  1.000,  val loss =  0.178,  val accuray =  0.938, \n",
      "Epoch 7550, Batch loss =  0.037,  training accuray =  1.000,  val loss =  0.239,  val accuray =  0.922, \n",
      "Epoch 7560, Batch loss =  0.061,  training accuray =  0.984,  val loss =  0.187,  val accuray =  0.930, \n",
      "Epoch 7570, Batch loss =  0.040,  training accuray =  1.000,  val loss =  0.206,  val accuray =  0.906, \n",
      "Epoch 7580, Batch loss =  0.039,  training accuray =  1.000,  val loss =  0.168,  val accuray =  0.938, \n",
      "Epoch 7590, Batch loss =  0.052,  training accuray =  0.992,  val loss =  0.200,  val accuray =  0.938, \n",
      "Epoch 7600, Batch loss =  0.072,  training accuray =  0.984,  val loss =  0.220,  val accuray =  0.914, \n",
      "Epoch 7610, Batch loss =  0.063,  training accuray =  0.992,  val loss =  0.182,  val accuray =  0.906, \n",
      "Epoch 7620, Batch loss =  0.042,  training accuray =  1.000,  val loss =  0.160,  val accuray =  0.953, \n",
      "Epoch 7630, Batch loss =  0.039,  training accuray =  1.000,  val loss =  0.157,  val accuray =  0.961, \n",
      "Epoch 7640, Batch loss =  0.061,  training accuray =  0.992,  val loss =  0.201,  val accuray =  0.922, \n",
      "Epoch 7650, Batch loss =  0.052,  training accuray =  0.984,  val loss =  0.223,  val accuray =  0.898, \n",
      "Epoch 7660, Batch loss =  0.073,  training accuray =  0.961,  val loss =  0.204,  val accuray =  0.938, \n",
      "Epoch 7670, Batch loss =  0.059,  training accuray =  0.992,  val loss =  0.178,  val accuray =  0.922, \n",
      "Epoch 7680, Batch loss =  0.037,  training accuray =  1.000,  val loss =  0.274,  val accuray =  0.906, \n",
      "Epoch 7690, Batch loss =  0.044,  training accuray =  1.000,  val loss =  0.194,  val accuray =  0.953, \n",
      "Epoch 7700, Batch loss =  0.051,  training accuray =  0.992,  val loss =  0.129,  val accuray =  0.961, \n",
      "Epoch 7710, Batch loss =  0.044,  training accuray =  1.000,  val loss =  0.173,  val accuray =  0.961, \n",
      "Epoch 7720, Batch loss =  0.055,  training accuray =  0.984,  val loss =  0.238,  val accuray =  0.898, \n",
      "Epoch 7730, Batch loss =  0.068,  training accuray =  0.984,  val loss =  0.168,  val accuray =  0.945, \n",
      "Epoch 7740, Batch loss =  0.055,  training accuray =  0.992,  val loss =  0.192,  val accuray =  0.945, \n",
      "Epoch 7750, Batch loss =  0.041,  training accuray =  0.992,  val loss =  0.179,  val accuray =  0.953, \n",
      "Epoch 7760, Batch loss =  0.060,  training accuray =  0.992,  val loss =  0.206,  val accuray =  0.930, \n",
      "Epoch 7770, Batch loss =  0.071,  training accuray =  0.984,  val loss =  0.184,  val accuray =  0.930, \n",
      "Epoch 7780, Batch loss =  0.038,  training accuray =  0.992,  val loss =  0.173,  val accuray =  0.945, \n",
      "Epoch 7790, Batch loss =  0.051,  training accuray =  0.992,  val loss =  0.174,  val accuray =  0.922, \n",
      "Epoch 7800, Batch loss =  0.041,  training accuray =  1.000,  val loss =  0.197,  val accuray =  0.922, \n",
      "Epoch 7810, Batch loss =  0.051,  training accuray =  0.992,  val loss =  0.204,  val accuray =  0.938, \n",
      "Epoch 7820, Batch loss =  0.039,  training accuray =  1.000,  val loss =  0.236,  val accuray =  0.914, \n",
      "Epoch 7830, Batch loss =  0.050,  training accuray =  0.992,  val loss =  0.221,  val accuray =  0.914, \n",
      "Epoch 7840, Batch loss =  0.042,  training accuray =  1.000,  val loss =  0.149,  val accuray =  0.961, \n",
      "Epoch 7850, Batch loss =  0.043,  training accuray =  1.000,  val loss =  0.221,  val accuray =  0.930, \n",
      "Epoch 7860, Batch loss =  0.057,  training accuray =  0.977,  val loss =  0.186,  val accuray =  0.914, \n",
      "Epoch 7870, Batch loss =  0.029,  training accuray =  1.000,  val loss =  0.216,  val accuray =  0.922, \n",
      "Epoch 7880, Batch loss =  0.067,  training accuray =  0.992,  val loss =  0.240,  val accuray =  0.906, \n",
      "Epoch 7890, Batch loss =  0.052,  training accuray =  0.992,  val loss =  0.204,  val accuray =  0.945, \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 7900, Batch loss =  0.039,  training accuray =  0.992,  val loss =  0.206,  val accuray =  0.914, \n",
      "Epoch 7910, Batch loss =  0.049,  training accuray =  0.984,  val loss =  0.231,  val accuray =  0.906, \n",
      "Epoch 7920, Batch loss =  0.040,  training accuray =  1.000,  val loss =  0.158,  val accuray =  0.945, \n",
      "Epoch 7930, Batch loss =  0.050,  training accuray =  1.000,  val loss =  0.178,  val accuray =  0.930, \n",
      "Epoch 7940, Batch loss =  0.036,  training accuray =  1.000,  val loss =  0.247,  val accuray =  0.906, \n",
      "Epoch 7950, Batch loss =  0.046,  training accuray =  0.992,  val loss =  0.174,  val accuray =  0.930, \n",
      "Epoch 7960, Batch loss =  0.041,  training accuray =  0.984,  val loss =  0.209,  val accuray =  0.930, \n",
      "Epoch 7970, Batch loss =  0.035,  training accuray =  1.000,  val loss =  0.318,  val accuray =  0.867, \n",
      "Epoch 7980, Batch loss =  0.037,  training accuray =  1.000,  val loss =  0.223,  val accuray =  0.914, \n",
      "Epoch 7990, Batch loss =  0.050,  training accuray =  0.984,  val loss =  0.186,  val accuray =  0.922, \n",
      "Epoch 8000, Batch loss =  0.044,  training accuray =  0.984,  val loss =  0.175,  val accuray =  0.938, \n",
      "Epoch 8010, Batch loss =  0.035,  training accuray =  0.992,  val loss =  0.184,  val accuray =  0.914, \n",
      "Epoch 8020, Batch loss =  0.042,  training accuray =  0.992,  val loss =  0.172,  val accuray =  0.938, \n",
      "Epoch 8030, Batch loss =  0.033,  training accuray =  1.000,  val loss =  0.194,  val accuray =  0.930, \n",
      "Epoch 8040, Batch loss =  0.050,  training accuray =  0.992,  val loss =  0.252,  val accuray =  0.914, \n",
      "Epoch 8050, Batch loss =  0.037,  training accuray =  1.000,  val loss =  0.261,  val accuray =  0.891, \n",
      "Epoch 8060, Batch loss =  0.030,  training accuray =  1.000,  val loss =  0.156,  val accuray =  0.930, \n",
      "Epoch 8070, Batch loss =  0.050,  training accuray =  0.992,  val loss =  0.163,  val accuray =  0.930, \n",
      "Epoch 8080, Batch loss =  0.031,  training accuray =  1.000,  val loss =  0.152,  val accuray =  0.945, \n",
      "Epoch 8090, Batch loss =  0.040,  training accuray =  1.000,  val loss =  0.222,  val accuray =  0.930, \n",
      "Epoch 8100, Batch loss =  0.044,  training accuray =  0.992,  val loss =  0.232,  val accuray =  0.898, \n",
      "Epoch 8110, Batch loss =  0.029,  training accuray =  1.000,  val loss =  0.189,  val accuray =  0.938, \n",
      "Epoch 8120, Batch loss =  0.046,  training accuray =  0.992,  val loss =  0.215,  val accuray =  0.914, \n",
      "Epoch 8130, Batch loss =  0.042,  training accuray =  0.984,  val loss =  0.210,  val accuray =  0.930, \n",
      "Epoch 8140, Batch loss =  0.028,  training accuray =  1.000,  val loss =  0.207,  val accuray =  0.922, \n",
      "Epoch 8150, Batch loss =  0.059,  training accuray =  0.984,  val loss =  0.165,  val accuray =  0.945, \n",
      "Epoch 8160, Batch loss =  0.036,  training accuray =  1.000,  val loss =  0.219,  val accuray =  0.938, \n",
      "Epoch 8170, Batch loss =  0.048,  training accuray =  0.992,  val loss =  0.134,  val accuray =  0.945, \n",
      "Epoch 8180, Batch loss =  0.048,  training accuray =  1.000,  val loss =  0.150,  val accuray =  0.953, \n",
      "Epoch 8190, Batch loss =  0.038,  training accuray =  1.000,  val loss =  0.210,  val accuray =  0.930, \n",
      "Epoch 8200, Batch loss =  0.034,  training accuray =  1.000,  val loss =  0.153,  val accuray =  0.945, \n",
      "Epoch 8210, Batch loss =  0.037,  training accuray =  1.000,  val loss =  0.175,  val accuray =  0.938, \n",
      "Epoch 8220, Batch loss =  0.033,  training accuray =  1.000,  val loss =  0.178,  val accuray =  0.938, \n",
      "Epoch 8230, Batch loss =  0.047,  training accuray =  0.984,  val loss =  0.189,  val accuray =  0.914, \n",
      "Epoch 8240, Batch loss =  0.049,  training accuray =  0.992,  val loss =  0.126,  val accuray =  0.977, \n",
      "Epoch 8250, Batch loss =  0.038,  training accuray =  1.000,  val loss =  0.182,  val accuray =  0.930, \n",
      "Epoch 8260, Batch loss =  0.038,  training accuray =  1.000,  val loss =  0.154,  val accuray =  0.938, \n",
      "Epoch 8270, Batch loss =  0.050,  training accuray =  0.984,  val loss =  0.193,  val accuray =  0.906, \n",
      "Epoch 8280, Batch loss =  0.050,  training accuray =  0.977,  val loss =  0.167,  val accuray =  0.914, \n",
      "Epoch 8290, Batch loss =  0.055,  training accuray =  0.984,  val loss =  0.256,  val accuray =  0.875, \n",
      "Epoch 8300, Batch loss =  0.032,  training accuray =  1.000,  val loss =  0.206,  val accuray =  0.938, \n",
      "Epoch 8310, Batch loss =  0.040,  training accuray =  0.992,  val loss =  0.272,  val accuray =  0.898, \n",
      "Epoch 8320, Batch loss =  0.046,  training accuray =  0.992,  val loss =  0.311,  val accuray =  0.875, \n",
      "Epoch 8330, Batch loss =  0.043,  training accuray =  0.992,  val loss =  0.133,  val accuray =  0.953, \n",
      "Epoch 8340, Batch loss =  0.031,  training accuray =  1.000,  val loss =  0.193,  val accuray =  0.914, \n",
      "Epoch 8350, Batch loss =  0.034,  training accuray =  1.000,  val loss =  0.230,  val accuray =  0.906, \n",
      "Epoch 8360, Batch loss =  0.058,  training accuray =  0.984,  val loss =  0.169,  val accuray =  0.930, \n",
      "Epoch 8370, Batch loss =  0.035,  training accuray =  1.000,  val loss =  0.243,  val accuray =  0.906, \n",
      "Epoch 8380, Batch loss =  0.037,  training accuray =  0.992,  val loss =  0.158,  val accuray =  0.945, \n",
      "Epoch 8390, Batch loss =  0.041,  training accuray =  0.977,  val loss =  0.267,  val accuray =  0.922, \n",
      "Epoch 8400, Batch loss =  0.041,  training accuray =  1.000,  val loss =  0.167,  val accuray =  0.953, \n",
      "Epoch 8410, Batch loss =  0.030,  training accuray =  1.000,  val loss =  0.165,  val accuray =  0.945, \n",
      "Epoch 8420, Batch loss =  0.038,  training accuray =  1.000,  val loss =  0.164,  val accuray =  0.945, \n",
      "Epoch 8430, Batch loss =  0.029,  training accuray =  1.000,  val loss =  0.165,  val accuray =  0.930, \n",
      "Epoch 8440, Batch loss =  0.045,  training accuray =  0.992,  val loss =  0.220,  val accuray =  0.914, \n",
      "Epoch 8450, Batch loss =  0.041,  training accuray =  0.992,  val loss =  0.131,  val accuray =  0.961, \n",
      "Epoch 8460, Batch loss =  0.063,  training accuray =  0.977,  val loss =  0.225,  val accuray =  0.906, \n",
      "Epoch 8470, Batch loss =  0.029,  training accuray =  1.000,  val loss =  0.155,  val accuray =  0.945, \n",
      "Epoch 8480, Batch loss =  0.040,  training accuray =  0.992,  val loss =  0.184,  val accuray =  0.945, \n",
      "Epoch 8490, Batch loss =  0.046,  training accuray =  0.992,  val loss =  0.237,  val accuray =  0.922, \n",
      "Epoch 8500, Batch loss =  0.047,  training accuray =  0.984,  val loss =  0.228,  val accuray =  0.906, \n",
      "Epoch 8510, Batch loss =  0.046,  training accuray =  0.992,  val loss =  0.106,  val accuray =  0.961, \n",
      "Epoch 8520, Batch loss =  0.038,  training accuray =  0.992,  val loss =  0.220,  val accuray =  0.945, \n",
      "Epoch 8530, Batch loss =  0.024,  training accuray =  1.000,  val loss =  0.279,  val accuray =  0.891, \n",
      "Epoch 8540, Batch loss =  0.042,  training accuray =  1.000,  val loss =  0.189,  val accuray =  0.906, \n",
      "Epoch 8550, Batch loss =  0.031,  training accuray =  1.000,  val loss =  0.171,  val accuray =  0.953, \n",
      "Epoch 8560, Batch loss =  0.026,  training accuray =  1.000,  val loss =  0.235,  val accuray =  0.930, \n",
      "Epoch 8570, Batch loss =  0.034,  training accuray =  1.000,  val loss =  0.239,  val accuray =  0.914, \n",
      "Epoch 8580, Batch loss =  0.050,  training accuray =  0.992,  val loss =  0.098,  val accuray =  0.961, \n",
      "Epoch 8590, Batch loss =  0.041,  training accuray =  1.000,  val loss =  0.234,  val accuray =  0.922, \n",
      "Epoch 8600, Batch loss =  0.070,  training accuray =  0.977,  val loss =  0.238,  val accuray =  0.922, \n",
      "Epoch 8610, Batch loss =  0.055,  training accuray =  0.984,  val loss =  0.170,  val accuray =  0.938, \n",
      "Epoch 8620, Batch loss =  0.026,  training accuray =  1.000,  val loss =  0.174,  val accuray =  0.945, \n",
      "Epoch 8630, Batch loss =  0.059,  training accuray =  0.984,  val loss =  0.198,  val accuray =  0.938, \n",
      "Epoch 8640, Batch loss =  0.018,  training accuray =  1.000,  val loss =  0.242,  val accuray =  0.906, \n",
      "Epoch 8650, Batch loss =  0.041,  training accuray =  0.984,  val loss =  0.117,  val accuray =  0.953, \n",
      "Epoch 8660, Batch loss =  0.041,  training accuray =  0.992,  val loss =  0.209,  val accuray =  0.914, \n",
      "Epoch 8670, Batch loss =  0.025,  training accuray =  1.000,  val loss =  0.226,  val accuray =  0.906, \n",
      "Epoch 8680, Batch loss =  0.034,  training accuray =  1.000,  val loss =  0.205,  val accuray =  0.914, \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 8690, Batch loss =  0.044,  training accuray =  0.992,  val loss =  0.282,  val accuray =  0.922, \n",
      "Epoch 8700, Batch loss =  0.033,  training accuray =  1.000,  val loss =  0.187,  val accuray =  0.906, \n",
      "Epoch 8710, Batch loss =  0.027,  training accuray =  1.000,  val loss =  0.237,  val accuray =  0.906, \n",
      "Epoch 8720, Batch loss =  0.033,  training accuray =  1.000,  val loss =  0.235,  val accuray =  0.898, \n",
      "Epoch 8730, Batch loss =  0.046,  training accuray =  0.984,  val loss =  0.264,  val accuray =  0.883, \n",
      "Epoch 8740, Batch loss =  0.026,  training accuray =  1.000,  val loss =  0.204,  val accuray =  0.922, \n",
      "Epoch 8750, Batch loss =  0.036,  training accuray =  1.000,  val loss =  0.167,  val accuray =  0.930, \n",
      "Epoch 8760, Batch loss =  0.029,  training accuray =  1.000,  val loss =  0.294,  val accuray =  0.898, \n",
      "Epoch 8770, Batch loss =  0.036,  training accuray =  1.000,  val loss =  0.214,  val accuray =  0.914, \n",
      "Epoch 8780, Batch loss =  0.032,  training accuray =  0.992,  val loss =  0.192,  val accuray =  0.922, \n",
      "Epoch 8790, Batch loss =  0.030,  training accuray =  1.000,  val loss =  0.161,  val accuray =  0.953, \n",
      "Epoch 8800, Batch loss =  0.029,  training accuray =  1.000,  val loss =  0.177,  val accuray =  0.938, \n",
      "Epoch 8810, Batch loss =  0.032,  training accuray =  1.000,  val loss =  0.251,  val accuray =  0.898, \n",
      "Epoch 8820, Batch loss =  0.025,  training accuray =  1.000,  val loss =  0.204,  val accuray =  0.930, \n",
      "Epoch 8830, Batch loss =  0.030,  training accuray =  1.000,  val loss =  0.130,  val accuray =  0.953, \n",
      "Epoch 8840, Batch loss =  0.033,  training accuray =  0.992,  val loss =  0.156,  val accuray =  0.930, \n",
      "Epoch 8850, Batch loss =  0.049,  training accuray =  0.977,  val loss =  0.217,  val accuray =  0.906, \n",
      "Epoch 8860, Batch loss =  0.034,  training accuray =  1.000,  val loss =  0.169,  val accuray =  0.922, \n",
      "Epoch 8870, Batch loss =  0.036,  training accuray =  0.992,  val loss =  0.193,  val accuray =  0.914, \n",
      "Epoch 8880, Batch loss =  0.040,  training accuray =  0.992,  val loss =  0.206,  val accuray =  0.914, \n",
      "Epoch 8890, Batch loss =  0.037,  training accuray =  0.992,  val loss =  0.243,  val accuray =  0.898, \n",
      "Epoch 8900, Batch loss =  0.034,  training accuray =  0.992,  val loss =  0.203,  val accuray =  0.938, \n",
      "Epoch 8910, Batch loss =  0.047,  training accuray =  0.984,  val loss =  0.208,  val accuray =  0.922, \n",
      "Epoch 8920, Batch loss =  0.029,  training accuray =  1.000,  val loss =  0.233,  val accuray =  0.898, \n",
      "Epoch 8930, Batch loss =  0.025,  training accuray =  1.000,  val loss =  0.213,  val accuray =  0.922, \n",
      "Epoch 8940, Batch loss =  0.044,  training accuray =  0.992,  val loss =  0.177,  val accuray =  0.922, \n",
      "Epoch 8950, Batch loss =  0.030,  training accuray =  1.000,  val loss =  0.152,  val accuray =  0.914, \n",
      "Epoch 8960, Batch loss =  0.019,  training accuray =  1.000,  val loss =  0.198,  val accuray =  0.938, \n",
      "Epoch 8970, Batch loss =  0.032,  training accuray =  0.992,  val loss =  0.175,  val accuray =  0.930, \n",
      "Epoch 8980, Batch loss =  0.021,  training accuray =  1.000,  val loss =  0.240,  val accuray =  0.914, \n",
      "Epoch 8990, Batch loss =  0.034,  training accuray =  1.000,  val loss =  0.232,  val accuray =  0.898, \n",
      "Epoch 9000, Batch loss =  0.025,  training accuray =  1.000,  val loss =  0.194,  val accuray =  0.930, \n",
      "Epoch 9010, Batch loss =  0.033,  training accuray =  0.992,  val loss =  0.159,  val accuray =  0.938, \n",
      "Epoch 9020, Batch loss =  0.020,  training accuray =  1.000,  val loss =  0.199,  val accuray =  0.914, \n",
      "Epoch 9030, Batch loss =  0.026,  training accuray =  1.000,  val loss =  0.250,  val accuray =  0.906, \n",
      "Epoch 9040, Batch loss =  0.027,  training accuray =  0.992,  val loss =  0.191,  val accuray =  0.922, \n",
      "Epoch 9050, Batch loss =  0.043,  training accuray =  0.992,  val loss =  0.142,  val accuray =  0.953, \n",
      "Epoch 9060, Batch loss =  0.017,  training accuray =  1.000,  val loss =  0.157,  val accuray =  0.969, \n",
      "Epoch 9070, Batch loss =  0.026,  training accuray =  1.000,  val loss =  0.178,  val accuray =  0.922, \n",
      "Epoch 9080, Batch loss =  0.031,  training accuray =  0.992,  val loss =  0.211,  val accuray =  0.930, \n",
      "Epoch 9090, Batch loss =  0.027,  training accuray =  1.000,  val loss =  0.205,  val accuray =  0.945, \n",
      "Epoch 9100, Batch loss =  0.027,  training accuray =  1.000,  val loss =  0.216,  val accuray =  0.930, \n",
      "Epoch 9110, Batch loss =  0.028,  training accuray =  1.000,  val loss =  0.165,  val accuray =  0.945, \n",
      "Epoch 9120, Batch loss =  0.018,  training accuray =  1.000,  val loss =  0.175,  val accuray =  0.938, \n",
      "Epoch 9130, Batch loss =  0.028,  training accuray =  1.000,  val loss =  0.141,  val accuray =  0.961, \n",
      "Epoch 9140, Batch loss =  0.016,  training accuray =  1.000,  val loss =  0.227,  val accuray =  0.922, \n",
      "Epoch 9150, Batch loss =  0.076,  training accuray =  0.977,  val loss =  0.192,  val accuray =  0.914, \n",
      "Epoch 9160, Batch loss =  0.022,  training accuray =  1.000,  val loss =  0.232,  val accuray =  0.922, \n",
      "Epoch 9170, Batch loss =  0.041,  training accuray =  0.984,  val loss =  0.254,  val accuray =  0.898, \n",
      "Epoch 9180, Batch loss =  0.025,  training accuray =  1.000,  val loss =  0.225,  val accuray =  0.914, \n",
      "Epoch 9190, Batch loss =  0.028,  training accuray =  1.000,  val loss =  0.183,  val accuray =  0.945, \n",
      "Epoch 9200, Batch loss =  0.054,  training accuray =  0.992,  val loss =  0.245,  val accuray =  0.914, \n",
      "Epoch 9210, Batch loss =  0.021,  training accuray =  0.992,  val loss =  0.170,  val accuray =  0.922, \n",
      "Epoch 9220, Batch loss =  0.029,  training accuray =  1.000,  val loss =  0.152,  val accuray =  0.938, \n",
      "Epoch 9230, Batch loss =  0.020,  training accuray =  1.000,  val loss =  0.155,  val accuray =  0.938, \n",
      "Epoch 9240, Batch loss =  0.022,  training accuray =  1.000,  val loss =  0.236,  val accuray =  0.875, \n",
      "Epoch 9250, Batch loss =  0.025,  training accuray =  1.000,  val loss =  0.178,  val accuray =  0.945, \n",
      "Epoch 9260, Batch loss =  0.030,  training accuray =  0.992,  val loss =  0.193,  val accuray =  0.930, \n",
      "Epoch 9270, Batch loss =  0.024,  training accuray =  1.000,  val loss =  0.173,  val accuray =  0.938, \n",
      "Epoch 9280, Batch loss =  0.023,  training accuray =  1.000,  val loss =  0.151,  val accuray =  0.953, \n",
      "Epoch 9290, Batch loss =  0.025,  training accuray =  1.000,  val loss =  0.248,  val accuray =  0.891, \n",
      "Epoch 9300, Batch loss =  0.020,  training accuray =  1.000,  val loss =  0.223,  val accuray =  0.898, \n",
      "Epoch 9310, Batch loss =  0.023,  training accuray =  1.000,  val loss =  0.161,  val accuray =  0.945, \n",
      "Epoch 9320, Batch loss =  0.031,  training accuray =  0.992,  val loss =  0.218,  val accuray =  0.930, \n",
      "Epoch 9330, Batch loss =  0.041,  training accuray =  0.992,  val loss =  0.194,  val accuray =  0.938, \n",
      "Epoch 9340, Batch loss =  0.028,  training accuray =  1.000,  val loss =  0.231,  val accuray =  0.922, \n",
      "Epoch 9350, Batch loss =  0.025,  training accuray =  0.992,  val loss =  0.192,  val accuray =  0.961, \n",
      "Epoch 9360, Batch loss =  0.018,  training accuray =  1.000,  val loss =  0.229,  val accuray =  0.930, \n",
      "Epoch 9370, Batch loss =  0.026,  training accuray =  1.000,  val loss =  0.167,  val accuray =  0.938, \n",
      "Epoch 9380, Batch loss =  0.029,  training accuray =  0.992,  val loss =  0.292,  val accuray =  0.883, \n",
      "Epoch 9390, Batch loss =  0.025,  training accuray =  1.000,  val loss =  0.160,  val accuray =  0.914, \n",
      "Epoch 9400, Batch loss =  0.024,  training accuray =  1.000,  val loss =  0.120,  val accuray =  0.961, \n",
      "Epoch 9410, Batch loss =  0.025,  training accuray =  1.000,  val loss =  0.200,  val accuray =  0.914, \n",
      "Epoch 9420, Batch loss =  0.031,  training accuray =  0.992,  val loss =  0.132,  val accuray =  0.945, \n",
      "Epoch 9430, Batch loss =  0.028,  training accuray =  0.992,  val loss =  0.202,  val accuray =  0.922, \n",
      "Epoch 9440, Batch loss =  0.031,  training accuray =  0.992,  val loss =  0.188,  val accuray =  0.953, \n",
      "Epoch 9450, Batch loss =  0.025,  training accuray =  1.000,  val loss =  0.265,  val accuray =  0.891, \n",
      "Epoch 9460, Batch loss =  0.032,  training accuray =  1.000,  val loss =  0.230,  val accuray =  0.930, \n",
      "Epoch 9470, Batch loss =  0.029,  training accuray =  1.000,  val loss =  0.176,  val accuray =  0.945, \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 9480, Batch loss =  0.028,  training accuray =  1.000,  val loss =  0.238,  val accuray =  0.906, \n",
      "Epoch 9490, Batch loss =  0.021,  training accuray =  1.000,  val loss =  0.212,  val accuray =  0.906, \n",
      "Epoch 9500, Batch loss =  0.036,  training accuray =  0.992,  val loss =  0.166,  val accuray =  0.938, \n",
      "Epoch 9510, Batch loss =  0.032,  training accuray =  0.992,  val loss =  0.228,  val accuray =  0.898, \n",
      "Epoch 9520, Batch loss =  0.055,  training accuray =  0.992,  val loss =  0.146,  val accuray =  0.945, \n",
      "Epoch 9530, Batch loss =  0.017,  training accuray =  1.000,  val loss =  0.178,  val accuray =  0.938, \n",
      "Epoch 9540, Batch loss =  0.041,  training accuray =  0.992,  val loss =  0.264,  val accuray =  0.906, \n",
      "Epoch 9550, Batch loss =  0.014,  training accuray =  1.000,  val loss =  0.250,  val accuray =  0.906, \n",
      "Epoch 9560, Batch loss =  0.020,  training accuray =  1.000,  val loss =  0.172,  val accuray =  0.938, \n",
      "Epoch 9570, Batch loss =  0.016,  training accuray =  1.000,  val loss =  0.181,  val accuray =  0.930, \n",
      "Epoch 9580, Batch loss =  0.043,  training accuray =  0.992,  val loss =  0.209,  val accuray =  0.922, \n",
      "Epoch 9590, Batch loss =  0.030,  training accuray =  0.992,  val loss =  0.191,  val accuray =  0.930, \n",
      "Epoch 9600, Batch loss =  0.030,  training accuray =  1.000,  val loss =  0.272,  val accuray =  0.906, \n",
      "Epoch 9610, Batch loss =  0.025,  training accuray =  1.000,  val loss =  0.181,  val accuray =  0.922, \n",
      "Epoch 9620, Batch loss =  0.026,  training accuray =  1.000,  val loss =  0.227,  val accuray =  0.914, \n",
      "Epoch 9630, Batch loss =  0.032,  training accuray =  1.000,  val loss =  0.187,  val accuray =  0.930, \n",
      "Epoch 9640, Batch loss =  0.024,  training accuray =  1.000,  val loss =  0.226,  val accuray =  0.914, \n",
      "Epoch 9650, Batch loss =  0.016,  training accuray =  1.000,  val loss =  0.200,  val accuray =  0.938, \n",
      "Epoch 9660, Batch loss =  0.049,  training accuray =  0.984,  val loss =  0.127,  val accuray =  0.961, \n",
      "Epoch 9670, Batch loss =  0.025,  training accuray =  1.000,  val loss =  0.230,  val accuray =  0.930, \n",
      "Epoch 9680, Batch loss =  0.031,  training accuray =  0.992,  val loss =  0.141,  val accuray =  0.961, \n",
      "Epoch 9690, Batch loss =  0.021,  training accuray =  1.000,  val loss =  0.122,  val accuray =  0.953, \n",
      "Epoch 9700, Batch loss =  0.023,  training accuray =  1.000,  val loss =  0.204,  val accuray =  0.922, \n",
      "Epoch 9710, Batch loss =  0.027,  training accuray =  0.992,  val loss =  0.186,  val accuray =  0.938, \n",
      "Epoch 9720, Batch loss =  0.015,  training accuray =  1.000,  val loss =  0.210,  val accuray =  0.922, \n",
      "Epoch 9730, Batch loss =  0.016,  training accuray =  1.000,  val loss =  0.277,  val accuray =  0.891, \n",
      "Epoch 9740, Batch loss =  0.027,  training accuray =  1.000,  val loss =  0.117,  val accuray =  0.961, \n",
      "Epoch 9750, Batch loss =  0.020,  training accuray =  1.000,  val loss =  0.157,  val accuray =  0.945, \n",
      "Epoch 9760, Batch loss =  0.018,  training accuray =  1.000,  val loss =  0.246,  val accuray =  0.922, \n",
      "Epoch 9770, Batch loss =  0.018,  training accuray =  1.000,  val loss =  0.210,  val accuray =  0.906, \n",
      "Epoch 9780, Batch loss =  0.025,  training accuray =  1.000,  val loss =  0.215,  val accuray =  0.914, \n",
      "Epoch 9790, Batch loss =  0.015,  training accuray =  1.000,  val loss =  0.202,  val accuray =  0.922, \n",
      "Epoch 9800, Batch loss =  0.038,  training accuray =  0.992,  val loss =  0.201,  val accuray =  0.938, \n",
      "Epoch 9810, Batch loss =  0.022,  training accuray =  1.000,  val loss =  0.185,  val accuray =  0.938, \n",
      "Epoch 9820, Batch loss =  0.019,  training accuray =  1.000,  val loss =  0.318,  val accuray =  0.883, \n",
      "Epoch 9830, Batch loss =  0.019,  training accuray =  0.992,  val loss =  0.205,  val accuray =  0.938, \n",
      "Epoch 9840, Batch loss =  0.020,  training accuray =  1.000,  val loss =  0.184,  val accuray =  0.945, \n",
      "Epoch 9850, Batch loss =  0.028,  training accuray =  0.992,  val loss =  0.236,  val accuray =  0.891, \n",
      "Epoch 9860, Batch loss =  0.019,  training accuray =  1.000,  val loss =  0.224,  val accuray =  0.930, \n",
      "Epoch 9870, Batch loss =  0.015,  training accuray =  1.000,  val loss =  0.154,  val accuray =  0.953, \n",
      "Epoch 9880, Batch loss =  0.027,  training accuray =  0.992,  val loss =  0.162,  val accuray =  0.930, \n",
      "Epoch 9890, Batch loss =  0.025,  training accuray =  1.000,  val loss =  0.263,  val accuray =  0.891, \n",
      "Epoch 9900, Batch loss =  0.021,  training accuray =  1.000,  val loss =  0.146,  val accuray =  0.953, \n",
      "Epoch 9910, Batch loss =  0.028,  training accuray =  1.000,  val loss =  0.223,  val accuray =  0.914, \n",
      "Epoch 9920, Batch loss =  0.019,  training accuray =  0.992,  val loss =  0.223,  val accuray =  0.898, \n",
      "Epoch 9930, Batch loss =  0.025,  training accuray =  0.992,  val loss =  0.200,  val accuray =  0.938, \n",
      "Epoch 9940, Batch loss =  0.017,  training accuray =  1.000,  val loss =  0.083,  val accuray =  0.984, \n",
      "Epoch 9950, Batch loss =  0.044,  training accuray =  0.984,  val loss =  0.179,  val accuray =  0.930, \n",
      "Epoch 9960, Batch loss =  0.023,  training accuray =  0.992,  val loss =  0.262,  val accuray =  0.898, \n",
      "Epoch 9970, Batch loss =  0.029,  training accuray =  1.000,  val loss =  0.146,  val accuray =  0.953, \n",
      "Epoch 9980, Batch loss =  0.024,  training accuray =  0.992,  val loss =  0.150,  val accuray =  0.930, \n",
      "Epoch 9990, Batch loss =  0.025,  training accuray =  1.000,  val loss =  0.115,  val accuray =  0.977, \n"
     ]
    }
   ],
   "source": [
    "epochs           = 10000\n",
    "display_epoch    = 10\n",
    "\n",
    "import time\n",
    "timestr = time.strftime(\"%Y%m%d-%H%M%S\")\n",
    "allstr = \"batchSize_\" + str(batch_size) + \"_embeddingSize_\" + str(embedding_size) + \\\n",
    "         \"_lstmOut_\" + str(lstm_out) + \"_bowSize2_\" + str(bow_hidden_layer_2_size) + \\\n",
    "         \"_hidden1_\" + str(hidden_layer_1_size) + \"_hidden2_\" + str(hidden_layer_2_size) + \"_time_\"+ timestr\n",
    "filename = \"/Users/mohsenkiskani/Downloads/submissions/structred/lstm_bow_\"+str(allstr)+\".csv\"\n",
    "\n",
    "loss_val_vec     = []\n",
    "accur_val_vec    = []\n",
    "loss_train_vec   = []\n",
    "accur_train_vec  = []\n",
    "epoch_vec        = []\n",
    "\n",
    "with tf.Session() as sess:\n",
    "        sess.run(tf.global_variables_initializer())\n",
    "\n",
    "        for epoch in range(epochs):\n",
    "            batch_bow_vals, batch_inputs, batch_labels = get_batch(X_train_pad, \n",
    "                                                                   Y_train_pad, \n",
    "                                                                   X_train_tfidf,\n",
    "                                                                   batch_size)\n",
    "            loss_train, accur_train, _ = sess.run([loss, accuracy, optimizer],\n",
    "                                                  feed_dict = {X_bow : batch_bow_vals,\n",
    "                                                               X     : batch_inputs,\n",
    "                                                               Y     : batch_labels})\n",
    "            \n",
    "            batch_bow_vals, batch_inputs, batch_labels = get_batch(X_val_pad, \n",
    "                                                                   Y_val_pad, \n",
    "                                                                   X_val_tfidf,\n",
    "                                                                   batch_size)\n",
    "            \n",
    "            loss_val, accur_val = sess.run([loss, accuracy],\n",
    "                                           feed_dict = {X_bow : batch_bow_vals,\n",
    "                                                        X     : batch_inputs,\n",
    "                                                        Y     : batch_labels})\n",
    "                \n",
    "            loss_train_vec.append(loss_train)\n",
    "            accur_train_vec.append(accur_train)\n",
    "            loss_val_vec.append(loss_val)\n",
    "            accur_val_vec.append(accur_val)\n",
    "            epoch_vec.append(epoch)\n",
    "            \n",
    "            if epoch % display_epoch == 0:\n",
    "                print(\"Epoch \" +str(epoch) + \", Batch loss = \", \n",
    "                      \"%.3f, \" % loss_train , \"training accuray = \",  \"%.3f, \" % accur_train, \"val loss = \",\n",
    "                      \"%.3f, \" % loss_val , \"val accuray = \",  \"%.3f, \" % accur_val)\n",
    "        \n",
    "        evaluate_test_data_set(filename, X_test_pad, X_test_tfidf_np, sess)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Plotting the results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 155,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1c62d13b70>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1c653aed68>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "%matplotlib inline\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "plotting_parameter = 300\n",
    "\n",
    "fig1, (ay1, ay2) = plt.subplots(2, 1)\n",
    "ay1.plot(epoch_vec[plotting_parameter:], loss_train_vec[plotting_parameter:])\n",
    "ay2.plot(epoch_vec[plotting_parameter:], loss_val_vec[plotting_parameter:])\n",
    "\n",
    "fig2, (ay3, ay4) = plt.subplots(2, 1)\n",
    "ay3.plot(epoch_vec[plotting_parameter:], accur_train_vec[plotting_parameter:])\n",
    "ay4.plot(epoch_vec[plotting_parameter:], accur_val_vec[plotting_parameter:])\n",
    "\n",
    "ay1.set_title('Training batch loss vs epoch')\n",
    "ay1.set_xlabel('Epoch')\n",
    "ay1.set_ylabel('Training batch loss')\n",
    "\n",
    "ay2.set_title('Validation batch loss vs epoch')\n",
    "ay2.set_xlabel('Epoch')\n",
    "ay2.set_ylabel('Validation batch loss')\n",
    "\n",
    "ay3.set_title('Training batch accuracy vs epoch')\n",
    "ay3.set_xlabel('Epoch')\n",
    "ay3.set_ylabel('Train accuracy')\n",
    "\n",
    "ay4.set_title('Validation batch accuracy vs epoch')\n",
    "ay4.set_xlabel('Epoch')\n",
    "ay4.set_ylabel('Validation accuracy')\n",
    "\n",
    "fig1.set_size_inches(18.5, 10.5)\n",
    "fig1.savefig(\"/Users/mohsenkiskani/Downloads/submissions/structred/loss_\"+timestr+\".png\", dpi=100)\n",
    "\n",
    "fig2.set_size_inches(18.5, 10.5)\n",
    "fig2.savefig(\"/Users/mohsenkiskani/Downloads/submissions/structred/accuracy_\"+timestr+\".png\", dpi=100)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'kellyannepolls kellyanne you were fantastic on meetthepress today keep going i will win for the people make america great again'"
      ]
     },
     "execution_count": 53,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "test_data['tweet'][5]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 156,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[0.89821243,\n",
       " 0.85832083,\n",
       " 0.967569,\n",
       " 1.0272363,\n",
       " 0.9696244,\n",
       " 1.0445185,\n",
       " 1.0075109,\n",
       " 1.0253726,\n",
       " 0.99645746,\n",
       " 1.0089962,\n",
       " 1.0409911,\n",
       " 1.0509422,\n",
       " 0.9268191,\n",
       " 0.93204427,\n",
       " 1.0407597,\n",
       " 0.9849317,\n",
       " 0.9393016,\n",
       " 0.9150517,\n",
       " 1.0766394,\n",
       " 0.8977665,\n",
       " 1.0161712,\n",
       " 0.8196515,\n",
       " 0.9891337,\n",
       " 0.9778171,\n",
       " 0.9340654,\n",
       " 1.085321,\n",
       " 1.0524616,\n",
       " 0.996031,\n",
       " 0.9385838,\n",
       " 0.9664972,\n",
       " 1.0013951,\n",
       " 0.9237012,\n",
       " 0.9763791,\n",
       " 0.9838004,\n",
       " 0.96621156,\n",
       " 1.0280716,\n",
       " 1.079411,\n",
       " 0.88553774,\n",
       " 1.0584745,\n",
       " 0.9378973,\n",
       " 0.83817416,\n",
       " 1.1112394,\n",
       " 0.98311317,\n",
       " 0.8893344,\n",
       " 0.94265556,\n",
       " 0.9555483,\n",
       " 0.9812451,\n",
       " 0.9281052,\n",
       " 0.91257393,\n",
       " 0.95072204,\n",
       " 0.9433082,\n",
       " 0.9201872,\n",
       " 0.9316915,\n",
       " 0.9242649,\n",
       " 0.91812384,\n",
       " 0.9388125,\n",
       " 0.90751994,\n",
       " 0.96497905,\n",
       " 0.84622335,\n",
       " 0.86785847,\n",
       " 0.94315195,\n",
       " 1.0161517,\n",
       " 1.0226009,\n",
       " 0.86603624,\n",
       " 0.9329372,\n",
       " 0.8744401,\n",
       " 0.90020645,\n",
       " 0.830003,\n",
       " 0.9207852,\n",
       " 0.947881,\n",
       " 0.98842347,\n",
       " 0.95132494,\n",
       " 0.94301766,\n",
       " 0.9795476,\n",
       " 0.89751446,\n",
       " 1.0179832,\n",
       " 0.8825786,\n",
       " 0.86630875,\n",
       " 0.93034875,\n",
       " 0.84062135,\n",
       " 0.9272141,\n",
       " 0.9747637,\n",
       " 0.95819294,\n",
       " 0.9396049,\n",
       " 0.8695518,\n",
       " 0.94210166,\n",
       " 0.9741209,\n",
       " 0.9243814,\n",
       " 0.8235057,\n",
       " 0.96642697,\n",
       " 0.8883334,\n",
       " 0.904743,\n",
       " 0.9334764,\n",
       " 0.9504887,\n",
       " 1.0377295,\n",
       " 1.0134778,\n",
       " 0.98900896,\n",
       " 0.99270344,\n",
       " 0.8564359,\n",
       " 0.9622657,\n",
       " 0.9024029,\n",
       " 0.9513303,\n",
       " 0.9557351,\n",
       " 0.94458115,\n",
       " 0.9081423,\n",
       " 0.9600439,\n",
       " 0.7709401,\n",
       " 0.93839526,\n",
       " 0.83101535,\n",
       " 0.95932496,\n",
       " 0.8655735,\n",
       " 0.8141078,\n",
       " 0.8905761,\n",
       " 0.8663773,\n",
       " 1.0378492,\n",
       " 0.94280076,\n",
       " 0.93502927,\n",
       " 0.89796525,\n",
       " 0.90712816,\n",
       " 0.9503732,\n",
       " 0.90333533,\n",
       " 0.90965265,\n",
       " 0.8403228,\n",
       " 0.8681443,\n",
       " 0.868999,\n",
       " 0.84000015,\n",
       " 0.8393717,\n",
       " 0.9074854,\n",
       " 0.9397199,\n",
       " 0.82566655,\n",
       " 0.86115444,\n",
       " 0.74039304,\n",
       " 0.94244003,\n",
       " 0.95773876,\n",
       " 0.84363675,\n",
       " 0.8919831,\n",
       " 0.8588883,\n",
       " 0.8448039,\n",
       " 0.8188257,\n",
       " 0.93812406,\n",
       " 0.87384033,\n",
       " 0.9010638,\n",
       " 0.9027385,\n",
       " 0.9169159,\n",
       " 0.86084294,\n",
       " 0.7425935,\n",
       " 0.8516073,\n",
       " 0.90159565,\n",
       " 0.9058862,\n",
       " 0.87439287,\n",
       " 0.8771848,\n",
       " 0.8774707,\n",
       " 0.8961099,\n",
       " 0.896126,\n",
       " 0.8658047,\n",
       " 0.84893143,\n",
       " 0.92012334,\n",
       " 0.88078624,\n",
       " 0.87155145,\n",
       " 0.77769053,\n",
       " 0.9043523,\n",
       " 0.86992025,\n",
       " 0.88570166,\n",
       " 0.8612176,\n",
       " 0.77695626,\n",
       " 0.8399923,\n",
       " 0.9168811,\n",
       " 0.8913237,\n",
       " 0.81319535,\n",
       " 0.8547987,\n",
       " 0.8046651,\n",
       " 0.94187677,\n",
       " 0.79652333,\n",
       " 0.8398687,\n",
       " 0.790511,\n",
       " 0.80179536,\n",
       " 0.8138347,\n",
       " 0.88505065,\n",
       " 0.87469184,\n",
       " 0.82723176,\n",
       " 0.92638916,\n",
       " 0.8205172,\n",
       " 0.7621038,\n",
       " 0.85196674,\n",
       " 0.80080694,\n",
       " 0.8263533,\n",
       " 0.814676,\n",
       " 0.8575784,\n",
       " 0.89459217,\n",
       " 0.86792177,\n",
       " 0.8116634,\n",
       " 0.81633466,\n",
       " 0.7494805,\n",
       " 0.82418984,\n",
       " 0.85186243,\n",
       " 0.7708552,\n",
       " 0.86406314,\n",
       " 0.76551247,\n",
       " 0.88722944,\n",
       " 0.86996675,\n",
       " 0.79119265,\n",
       " 0.8474277,\n",
       " 0.79923385,\n",
       " 0.8878013,\n",
       " 0.82106316,\n",
       " 0.815221,\n",
       " 0.8718258,\n",
       " 0.8795309,\n",
       " 0.8254424,\n",
       " 0.8206466,\n",
       " 0.7694839,\n",
       " 0.84582895,\n",
       " 0.7698753,\n",
       " 0.841967,\n",
       " 0.81843954,\n",
       " 0.82618684,\n",
       " 0.7777567,\n",
       " 0.849234,\n",
       " 0.8602288,\n",
       " 0.78425145,\n",
       " 0.8790425,\n",
       " 0.89043903,\n",
       " 0.7530522,\n",
       " 0.8527262,\n",
       " 0.7313622,\n",
       " 0.8709003,\n",
       " 0.83188814,\n",
       " 0.79929864,\n",
       " 0.7881516,\n",
       " 0.8615348,\n",
       " 0.8120098,\n",
       " 0.85670227,\n",
       " 0.773281,\n",
       " 0.78797245,\n",
       " 0.7896366,\n",
       " 0.80060947,\n",
       " 0.72924685,\n",
       " 0.8283478,\n",
       " 0.76646024,\n",
       " 0.85534036,\n",
       " 0.8292638,\n",
       " 0.8963269,\n",
       " 0.8345226,\n",
       " 0.8318985,\n",
       " 0.80464613,\n",
       " 0.72256535,\n",
       " 0.80118185,\n",
       " 0.79312027,\n",
       " 0.8067199,\n",
       " 0.82117337,\n",
       " 0.75357133,\n",
       " 0.8118148,\n",
       " 0.7982203,\n",
       " 0.7929261,\n",
       " 0.7883422,\n",
       " 0.8224498,\n",
       " 0.83238447,\n",
       " 0.7502435,\n",
       " 0.7534903,\n",
       " 0.83808506,\n",
       " 0.83738756,\n",
       " 0.7623192,\n",
       " 0.7719124,\n",
       " 0.79014057,\n",
       " 0.7370973,\n",
       " 0.7711003,\n",
       " 0.8293199,\n",
       " 0.8380872,\n",
       " 0.75362873,\n",
       " 0.8474868,\n",
       " 0.8538864,\n",
       " 0.81574774,\n",
       " 0.7543555,\n",
       " 0.7480302,\n",
       " 0.80395,\n",
       " 0.76395065,\n",
       " 0.76994306,\n",
       " 0.7412557,\n",
       " 0.7882339,\n",
       " 0.81599295,\n",
       " 0.7807076,\n",
       " 0.68593323,\n",
       " 0.8283912,\n",
       " 0.7099951,\n",
       " 0.8021803,\n",
       " 0.77661103,\n",
       " 0.78647506,\n",
       " 0.7942347,\n",
       " 0.7844867,\n",
       " 0.8080335,\n",
       " 0.78702205,\n",
       " 0.7983848,\n",
       " 0.78240883,\n",
       " 0.8253603,\n",
       " 0.7918451,\n",
       " 0.7844975,\n",
       " 0.8240132,\n",
       " 0.7169112,\n",
       " 0.74521303,\n",
       " 0.7014509,\n",
       " 0.80580163,\n",
       " 0.7924547,\n",
       " 0.7296467,\n",
       " 0.78230625,\n",
       " 0.805863,\n",
       " 0.82662034,\n",
       " 0.71190345,\n",
       " 0.70815307,\n",
       " 0.7468208,\n",
       " 0.8457039,\n",
       " 0.7135993,\n",
       " 0.78427243,\n",
       " 0.8220713,\n",
       " 0.7727828,\n",
       " 0.76233506,\n",
       " 0.8019433,\n",
       " 0.7256209,\n",
       " 0.7264408,\n",
       " 0.70493364,\n",
       " 0.7795992,\n",
       " 0.7797924,\n",
       " 0.7312496,\n",
       " 0.75255275,\n",
       " 0.7841079,\n",
       " 0.7265321,\n",
       " 0.7348685,\n",
       " 0.7888098,\n",
       " 0.7453581,\n",
       " 0.77688956,\n",
       " 0.73513633,\n",
       " 0.7339103,\n",
       " 0.66730595,\n",
       " 0.7362472,\n",
       " 0.7389959,\n",
       " 0.7568603,\n",
       " 0.79359007,\n",
       " 0.7445438,\n",
       " 0.7463137,\n",
       " 0.79570574,\n",
       " 0.8006176,\n",
       " 0.67809534,\n",
       " 0.7264746,\n",
       " 0.78302217,\n",
       " 0.78252816,\n",
       " 0.78351116,\n",
       " 0.7895013,\n",
       " 0.76660407,\n",
       " 0.74412245,\n",
       " 0.75645006,\n",
       " 0.77318776,\n",
       " 0.75021404,\n",
       " 0.7113441,\n",
       " 0.708586,\n",
       " 0.75163877,\n",
       " 0.75859916,\n",
       " 0.77657247,\n",
       " 0.7810736,\n",
       " 0.68892264,\n",
       " 0.746956,\n",
       " 0.7363491,\n",
       " 0.7662526,\n",
       " 0.72246706,\n",
       " 0.70727885,\n",
       " 0.68015194,\n",
       " 0.74711776,\n",
       " 0.7521655,\n",
       " 0.701475,\n",
       " 0.83618927,\n",
       " 0.7594243,\n",
       " 0.7464582,\n",
       " 0.7841916,\n",
       " 0.7508923,\n",
       " 0.71847355,\n",
       " 0.67094284,\n",
       " 0.73102313,\n",
       " 0.72369325,\n",
       " 0.74344575,\n",
       " 0.7541201,\n",
       " 0.7150341,\n",
       " 0.6834309,\n",
       " 0.70180327,\n",
       " 0.77028346,\n",
       " 0.73871756,\n",
       " 0.7622352,\n",
       " 0.739247,\n",
       " 0.7016166,\n",
       " 0.72241503,\n",
       " 0.7706257,\n",
       " 0.73802006,\n",
       " 0.7112092,\n",
       " 0.738379,\n",
       " 0.75220513,\n",
       " 0.6786652,\n",
       " 0.7367083,\n",
       " 0.7812217,\n",
       " 0.6633923,\n",
       " 0.72319835,\n",
       " 0.6768712,\n",
       " 0.72056675,\n",
       " 0.6147715,\n",
       " 0.7091341,\n",
       " 0.7578559,\n",
       " 0.71777165,\n",
       " 0.6874329,\n",
       " 0.7330987,\n",
       " 0.7218384,\n",
       " 0.7368691,\n",
       " 0.7324085,\n",
       " 0.74650294,\n",
       " 0.7388367,\n",
       " 0.72546905,\n",
       " 0.69628453,\n",
       " 0.7530206,\n",
       " 0.77540946,\n",
       " 0.6777218,\n",
       " 0.72060597,\n",
       " 0.65025616,\n",
       " 0.7297852,\n",
       " 0.7395959,\n",
       " 0.752375,\n",
       " 0.725243,\n",
       " 0.75602853,\n",
       " 0.72480714,\n",
       " 0.7604641,\n",
       " 0.7021176,\n",
       " 0.6805196,\n",
       " 0.7433125,\n",
       " 0.7255553,\n",
       " 0.7225344,\n",
       " 0.68347216,\n",
       " 0.70298135,\n",
       " 0.76106465,\n",
       " 0.7059547,\n",
       " 0.683098,\n",
       " 0.64310014,\n",
       " 0.69392633,\n",
       " 0.69976604,\n",
       " 0.71458256,\n",
       " 0.68450886,\n",
       " 0.6584922,\n",
       " 0.7873199,\n",
       " 0.73416436,\n",
       " 0.70388377,\n",
       " 0.6272776,\n",
       " 0.7488146,\n",
       " 0.71524656,\n",
       " 0.66840595,\n",
       " 0.6759395,\n",
       " 0.68182063,\n",
       " 0.6597556,\n",
       " 0.69576955,\n",
       " 0.680401,\n",
       " 0.6596805,\n",
       " 0.70935345,\n",
       " 0.7642203,\n",
       " 0.70818305,\n",
       " 0.672909,\n",
       " 0.7132175,\n",
       " 0.6897455,\n",
       " 0.7097896,\n",
       " 0.7353106,\n",
       " 0.66050166,\n",
       " 0.6832876,\n",
       " 0.69609755,\n",
       " 0.6992676,\n",
       " 0.74761635,\n",
       " 0.7245989,\n",
       " 0.67560935,\n",
       " 0.70838267,\n",
       " 0.67487293,\n",
       " 0.65936035,\n",
       " 0.6863451,\n",
       " 0.6813632,\n",
       " 0.7324086,\n",
       " 0.6833483,\n",
       " 0.68074393,\n",
       " 0.7085129,\n",
       " 0.6976886,\n",
       " 0.7122992,\n",
       " 0.73090297,\n",
       " 0.6841806,\n",
       " 0.6691271,\n",
       " 0.65166247,\n",
       " 0.7101129,\n",
       " 0.6830439,\n",
       " 0.6709037,\n",
       " 0.7144092,\n",
       " 0.70015204,\n",
       " 0.72264886,\n",
       " 0.72696674,\n",
       " 0.6738516,\n",
       " 0.69587,\n",
       " 0.6996765,\n",
       " 0.738683,\n",
       " 0.6680864,\n",
       " 0.68606615,\n",
       " 0.6979571,\n",
       " 0.70253974,\n",
       " 0.7092405,\n",
       " 0.6686784,\n",
       " 0.7072183,\n",
       " 0.6904533,\n",
       " 0.68654275,\n",
       " 0.6723224,\n",
       " 0.6799298,\n",
       " 0.66493297,\n",
       " 0.64054114,\n",
       " 0.6962967,\n",
       " 0.6657629,\n",
       " 0.70997757,\n",
       " 0.7099631,\n",
       " 0.6856167,\n",
       " 0.66527736,\n",
       " 0.66504014,\n",
       " 0.6566026,\n",
       " 0.6375448,\n",
       " 0.65356433,\n",
       " 0.6803187,\n",
       " 0.7038826,\n",
       " 0.65877324,\n",
       " 0.7107332,\n",
       " 0.6859967,\n",
       " 0.72926295,\n",
       " 0.70298284,\n",
       " 0.69627035,\n",
       " 0.72277635,\n",
       " 0.69307566,\n",
       " 0.6956855,\n",
       " 0.6607703,\n",
       " 0.6854023,\n",
       " 0.685863,\n",
       " 0.66363597,\n",
       " 0.72615105,\n",
       " 0.7034724,\n",
       " 0.64828604,\n",
       " 0.70651233,\n",
       " 0.6898358,\n",
       " 0.6614989,\n",
       " 0.69694924,\n",
       " 0.6860908,\n",
       " 0.66189027,\n",
       " 0.6876374,\n",
       " 0.70367324,\n",
       " 0.6798407,\n",
       " 0.6731928,\n",
       " 0.7475494,\n",
       " 0.65446603,\n",
       " 0.64137363,\n",
       " 0.70530784,\n",
       " 0.6230328,\n",
       " 0.7043046,\n",
       " 0.7207713,\n",
       " 0.7137089,\n",
       " 0.70291525,\n",
       " 0.6796056,\n",
       " 0.6339651,\n",
       " 0.6421159,\n",
       " 0.66147673,\n",
       " 0.7251011,\n",
       " 0.6629526,\n",
       " 0.70793056,\n",
       " 0.6048691,\n",
       " 0.6363005,\n",
       " 0.67919415,\n",
       " 0.7096528,\n",
       " 0.62762177,\n",
       " 0.6182374,\n",
       " 0.6775838,\n",
       " 0.6182085,\n",
       " 0.6782043,\n",
       " 0.6700297,\n",
       " 0.6738249,\n",
       " 0.67932445,\n",
       " 0.6233339,\n",
       " 0.63244224,\n",
       " 0.7027116,\n",
       " 0.6122137,\n",
       " 0.6599533,\n",
       " 0.6902081,\n",
       " 0.66403997,\n",
       " 0.6663609,\n",
       " 0.60648537,\n",
       " 0.68189234,\n",
       " 0.6357635,\n",
       " 0.6799469,\n",
       " 0.67191225,\n",
       " 0.6954394,\n",
       " 0.63821816,\n",
       " 0.5788188,\n",
       " 0.6410897,\n",
       " 0.71102,\n",
       " 0.6733182,\n",
       " 0.64386594,\n",
       " 0.66233873,\n",
       " 0.6431771,\n",
       " 0.5962447,\n",
       " 0.64121836,\n",
       " 0.6328832,\n",
       " 0.6718802,\n",
       " 0.63991463,\n",
       " 0.6288387,\n",
       " 0.66806304,\n",
       " 0.6616337,\n",
       " 0.7063463,\n",
       " 0.65946233,\n",
       " 0.6568704,\n",
       " 0.6410686,\n",
       " 0.62501764,\n",
       " 0.59460294,\n",
       " 0.676664,\n",
       " 0.6676227,\n",
       " 0.69185823,\n",
       " 0.65691245,\n",
       " 0.6276652,\n",
       " 0.6517056,\n",
       " 0.64595807,\n",
       " 0.63561106,\n",
       " 0.65314597,\n",
       " 0.6477401,\n",
       " 0.6385168,\n",
       " 0.6116257,\n",
       " 0.7040861,\n",
       " 0.61309695,\n",
       " 0.6499467,\n",
       " 0.6331825,\n",
       " 0.6411644,\n",
       " 0.6402433,\n",
       " 0.6601018,\n",
       " 0.6312703,\n",
       " 0.66259706,\n",
       " 0.6737344,\n",
       " 0.679221,\n",
       " 0.6710788,\n",
       " 0.6197356,\n",
       " 0.64072967,\n",
       " 0.60913193,\n",
       " 0.6829848,\n",
       " 0.65908384,\n",
       " 0.69484603,\n",
       " 0.63182235,\n",
       " 0.6706188,\n",
       " 0.6672894,\n",
       " 0.65737027,\n",
       " 0.6399713,\n",
       " 0.643759,\n",
       " 0.65577996,\n",
       " 0.66071326,\n",
       " 0.63019717,\n",
       " 0.6999431,\n",
       " 0.63328934,\n",
       " 0.6302676,\n",
       " 0.6064582,\n",
       " 0.675341,\n",
       " 0.67006606,\n",
       " 0.59331,\n",
       " 0.64402854,\n",
       " 0.6415259,\n",
       " 0.6717983,\n",
       " 0.608222,\n",
       " 0.607728,\n",
       " 0.64982086,\n",
       " 0.6250571,\n",
       " 0.6111983,\n",
       " 0.6047924,\n",
       " 0.623262,\n",
       " 0.61474645,\n",
       " 0.655102,\n",
       " 0.6100596,\n",
       " 0.61509246,\n",
       " 0.6704414,\n",
       " 0.6178236,\n",
       " 0.62699103,\n",
       " 0.6501869,\n",
       " 0.63382703,\n",
       " 0.6544076,\n",
       " 0.62717825,\n",
       " 0.5703063,\n",
       " 0.62187666,\n",
       " 0.6347672,\n",
       " 0.644097,\n",
       " 0.64119,\n",
       " 0.6342933,\n",
       " 0.65282834,\n",
       " 0.64791375,\n",
       " 0.59961396,\n",
       " 0.6281224,\n",
       " 0.6503864,\n",
       " 0.5742403,\n",
       " 0.57510066,\n",
       " 0.6199403,\n",
       " 0.6529306,\n",
       " 0.62847936,\n",
       " 0.62847733,\n",
       " 0.6039808,\n",
       " 0.64580613,\n",
       " 0.6599021,\n",
       " 0.61875033,\n",
       " 0.6442957,\n",
       " 0.6222846,\n",
       " 0.65089166,\n",
       " 0.6074192,\n",
       " 0.6058692,\n",
       " 0.6368191,\n",
       " 0.61300653,\n",
       " 0.63125896,\n",
       " 0.6224141,\n",
       " 0.608272,\n",
       " 0.60436696,\n",
       " 0.60963964,\n",
       " 0.615533,\n",
       " 0.6238665,\n",
       " 0.6268909,\n",
       " 0.64223236,\n",
       " 0.6023531,\n",
       " 0.64186716,\n",
       " 0.6485458,\n",
       " 0.58265156,\n",
       " 0.62572885,\n",
       " 0.6569003,\n",
       " 0.6106112,\n",
       " 0.64691365,\n",
       " 0.60958725,\n",
       " 0.66320753,\n",
       " 0.6166732,\n",
       " 0.590767,\n",
       " 0.57527184,\n",
       " 0.61216676,\n",
       " 0.57436913,\n",
       " 0.6364205,\n",
       " 0.6178471,\n",
       " 0.578709,\n",
       " 0.6338783,\n",
       " 0.608114,\n",
       " 0.6258444,\n",
       " 0.6118808,\n",
       " 0.59912694,\n",
       " 0.641369,\n",
       " 0.62932,\n",
       " 0.6117339,\n",
       " 0.6015667,\n",
       " 0.62033564,\n",
       " 0.6201978,\n",
       " 0.63008356,\n",
       " 0.6560751,\n",
       " 0.5880186,\n",
       " 0.62962574,\n",
       " 0.62344086,\n",
       " 0.5941686,\n",
       " 0.6069073,\n",
       " 0.6531284,\n",
       " 0.59109056,\n",
       " 0.6130407,\n",
       " 0.6045146,\n",
       " 0.60767496,\n",
       " 0.5509547,\n",
       " 0.63022375,\n",
       " 0.61607146,\n",
       " 0.60624814,\n",
       " 0.62363845,\n",
       " 0.5832456,\n",
       " 0.5920778,\n",
       " 0.5978969,\n",
       " 0.6121981,\n",
       " 0.66311705,\n",
       " 0.617405,\n",
       " 0.6346986,\n",
       " 0.6342221,\n",
       " 0.59557384,\n",
       " 0.59736836,\n",
       " 0.5721047,\n",
       " 0.58254164,\n",
       " 0.5971682,\n",
       " 0.5677976,\n",
       " 0.6141303,\n",
       " 0.6049481,\n",
       " 0.61333144,\n",
       " 0.6217978,\n",
       " 0.63013387,\n",
       " 0.57908016,\n",
       " 0.55670524,\n",
       " 0.6146185,\n",
       " 0.61264575,\n",
       " 0.5919787,\n",
       " 0.57500845,\n",
       " 0.57251394,\n",
       " 0.62447834,\n",
       " 0.6209264,\n",
       " 0.6145299,\n",
       " 0.6117736,\n",
       " 0.5960772,\n",
       " 0.5917507,\n",
       " 0.6056448,\n",
       " 0.6024214,\n",
       " 0.59561217,\n",
       " 0.5712807,\n",
       " 0.58298147,\n",
       " 0.59874564,\n",
       " 0.5654972,\n",
       " 0.57707304,\n",
       " 0.62413096,\n",
       " 0.6310878,\n",
       " 0.5721212,\n",
       " 0.6156844,\n",
       " 0.5656717,\n",
       " 0.6529214,\n",
       " 0.531739,\n",
       " 0.6045329,\n",
       " 0.60775816,\n",
       " 0.56740934,\n",
       " 0.58199936,\n",
       " 0.58531725,\n",
       " 0.5870731,\n",
       " 0.59043676,\n",
       " 0.56943154,\n",
       " 0.5684296,\n",
       " 0.61817926,\n",
       " 0.59159195,\n",
       " 0.6085997,\n",
       " 0.6254742,\n",
       " 0.64895046,\n",
       " 0.57478416,\n",
       " 0.55925643,\n",
       " 0.57722414,\n",
       " 0.54501384,\n",
       " 0.617965,\n",
       " 0.5638292,\n",
       " 0.5881349,\n",
       " 0.6171851,\n",
       " 0.56821597,\n",
       " 0.5749701,\n",
       " 0.5740379,\n",
       " 0.585696,\n",
       " 0.5759352,\n",
       " 0.5836304,\n",
       " 0.58316904,\n",
       " 0.5792879,\n",
       " 0.57782733,\n",
       " 0.59258795,\n",
       " 0.58293056,\n",
       " 0.6127636,\n",
       " 0.58241045,\n",
       " 0.5897398,\n",
       " 0.533157,\n",
       " 0.5769533,\n",
       " 0.5654464,\n",
       " 0.58932906,\n",
       " 0.5717645,\n",
       " 0.60221356,\n",
       " 0.55411667,\n",
       " 0.5442171,\n",
       " 0.547006,\n",
       " 0.5558543,\n",
       " 0.59267104,\n",
       " 0.6314485,\n",
       " 0.61241364,\n",
       " 0.55310726,\n",
       " 0.5696618,\n",
       " 0.58538735,\n",
       " 0.6288893,\n",
       " 0.559286,\n",
       " 0.58392674,\n",
       " 0.5542231,\n",
       " 0.5638149,\n",
       " 0.5732347,\n",
       " 0.65186197,\n",
       " 0.5792265,\n",
       " 0.607239,\n",
       " 0.5970536,\n",
       " 0.56160474,\n",
       " 0.6138692,\n",
       " 0.56129825,\n",
       " 0.5979272,\n",
       " 0.57831705,\n",
       " 0.5862368,\n",
       " 0.58144605,\n",
       " 0.5634861,\n",
       " 0.59463775,\n",
       " 0.53345025,\n",
       " 0.569681,\n",
       " 0.5943918,\n",
       " 0.5919598,\n",
       " 0.57837504,\n",
       " 0.5866701,\n",
       " 0.58434904,\n",
       " 0.57331705,\n",
       " 0.58334565,\n",
       " 0.5412966,\n",
       " 0.59462255,\n",
       " 0.5821363,\n",
       " 0.579322,\n",
       " 0.56172854,\n",
       " 0.56638795,\n",
       " 0.5706394,\n",
       " 0.53812236,\n",
       " 0.56466985,\n",
       " 0.5488336,\n",
       " 0.5998,\n",
       " 0.5795649,\n",
       " 0.545029,\n",
       " 0.55858946,\n",
       " 0.548765,\n",
       " 0.55426824,\n",
       " 0.55697423,\n",
       " 0.5452593,\n",
       " 0.5733266,\n",
       " 0.5314634,\n",
       " 0.5507034,\n",
       " 0.5607968,\n",
       " 0.5464129,\n",
       " 0.5888603,\n",
       " 0.5360422,\n",
       " 0.58839166,\n",
       " 0.5371002,\n",
       " 0.5482539,\n",
       " 0.5765852,\n",
       " 0.59269476,\n",
       " 0.5935638,\n",
       " 0.5414504,\n",
       " 0.53435713,\n",
       " 0.54956925,\n",
       " 0.5659154,\n",
       " 0.5737396,\n",
       " 0.5531458,\n",
       " 0.6104747,\n",
       " 0.5565983,\n",
       " 0.5827235,\n",
       " 0.5275856,\n",
       " 0.5598418,\n",
       " 0.5835726,\n",
       " 0.5766543,\n",
       " 0.54663366,\n",
       " 0.54529226,\n",
       " 0.6164675,\n",
       " 0.55693436,\n",
       " 0.57461214,\n",
       " 0.55240417,\n",
       " 0.5537334,\n",
       " 0.5264449,\n",
       " 0.5550831,\n",
       " 0.5662645,\n",
       " 0.52756906,\n",
       " 0.5893189,\n",
       " 0.5472802,\n",
       " 0.56670296,\n",
       " 0.5467144,\n",
       " 0.54930896,\n",
       " 0.55956656,\n",
       " 0.5599304,\n",
       " 0.6064353,\n",
       " 0.5398005,\n",
       " 0.5808153,\n",
       " 0.55048746,\n",
       " 0.55216515,\n",
       " 0.53458166,\n",
       " 0.5252447,\n",
       " 0.5397391,\n",
       " 0.5624631,\n",
       " 0.5666756,\n",
       " 0.5344042,\n",
       " 0.5707776,\n",
       " 0.5680146,\n",
       " 0.5792857,\n",
       " 0.5417013,\n",
       " 0.54783934,\n",
       " 0.5340024,\n",
       " 0.56592774,\n",
       " 0.57506585,\n",
       " 0.50113016,\n",
       " 0.55373895,\n",
       " 0.52907455,\n",
       " 0.50678694,\n",
       " 0.6148273,\n",
       " 0.52590543,\n",
       " 0.5426067,\n",
       " 0.5381917,\n",
       " 0.5540221,\n",
       " 0.5229553,\n",
       " 0.48216003,\n",
       " 0.56547916,\n",
       " 0.5528487,\n",
       " 0.53522605,\n",
       " 0.5684202,\n",
       " 0.50613654,\n",
       " 0.54246974,\n",
       " 0.55319035,\n",
       " 0.5520916,\n",
       " 0.5308556,\n",
       " 0.55296326,\n",
       " 0.4938717,\n",
       " 0.5268759,\n",
       " 0.5203129,\n",
       " 0.5202305,\n",
       " 0.53284365,\n",
       " 0.5092518,\n",
       " 0.5589188,\n",
       " 0.5470788,\n",
       " 0.5329519,\n",
       " 0.5066443,\n",
       " 0.57331735,\n",
       " 0.5493945,\n",
       " ...]"
      ]
     },
     "execution_count": 156,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "loss_val_vec"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "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": 2
}