LSTM_Conv1d.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "KuMWoh4eXWWf",
   "metadata": {
    "id": "KuMWoh4eXWWf"
   },
   "source": [
    "### <font color='violet'>Import libraries</font>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "0529e0da",
   "metadata": {
    "id": "0529e0da"
   },
   "outputs": [],
   "source": [
    "\n",
    "import os\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import tensorflow as tf\n",
    "from tensorflow import keras\n",
    "from sklearn.preprocessing import OneHotEncoder\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "from tensorflow.keras import layers\n",
    "from sklearn.decomposition import PCA\n",
    "import matplotlib.pyplot as plt\n",
    "import matplotlib.image as mpimg\n",
    "import seaborn as sns\n",
    "import statistics\n",
    "from numpy import array\n",
    "from numpy import asarray\n",
    "from numpy import zeros\n",
    "from keras.preprocessing.text import Tokenizer\n",
    "from keras.preprocessing.sequence import pad_sequences\n",
    "from sklearn.preprocessing import LabelEncoder\n",
    "from tensorflow.keras import layers\n",
    "import tensorflow as tf\n",
    "from tensorflow import keras"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "TLrPmi0UXx_f",
   "metadata": {
    "id": "TLrPmi0UXx_f"
   },
   "source": [
    "### <font color='violet'>Load the preprocessed GloVe embeddings</font>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "zOBa6jO7QMXE",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "zOBa6jO7QMXE",
    "outputId": "1492374c-34ff-4e18-90f7-c87cba6f6709"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Mounted at /content/drive\n"
     ]
    }
   ],
   "source": [
    "from google.colab import drive\n",
    "drive.mount('/content/drive')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "6mInfaBlDR_g",
   "metadata": {
    "id": "6mInfaBlDR_g"
   },
   "outputs": [],
   "source": [
    "import pickle\n",
    "filename = 'drive/My Drive/Colab Notebooks/NLP_emotions_dataset/Untitled Folder 1/emotions_dataset/embeddings_index.sav'\n",
    "embeddings_index = pickle.load(open(filename, 'rb'))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "O2S6U03yX5eN",
   "metadata": {
    "id": "O2S6U03yX5eN"
   },
   "source": [
    "### <font color='violet'>Import datasets and preprocess the datasets</font>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "4710c931",
   "metadata": {
    "id": "4710c931"
   },
   "outputs": [],
   "source": [
    "\n",
    "max_length = 58\n",
    "f = open('drive/My Drive/Colab Notebooks/NLP_emotions_dataset/Untitled Folder 1/emotions_dataset/train.txt')\n",
    "g=open('drive/My Drive/Colab Notebooks/NLP_emotions_dataset/Untitled Folder 1/emotions_dataset/val.txt')\n",
    "h=open('drive/My Drive/Colab Notebooks/NLP_emotions_dataset/Untitled Folder 1/emotions_dataset/test.txt')\n",
    "\n",
    "\n",
    "lb = LabelEncoder()\n",
    "t = Tokenizer()\n",
    "#def tokenizer_transform(words):\n",
    "    #t.fit_on_texts(words)    \n",
    "words_train =[]\n",
    "labels=[]\n",
    "for line in f:\n",
    "    values =line.rsplit(\";\", maxsplit=1)\n",
    "    sentence = values[0].split()\n",
    "    label=values[1]#removesuffix('\\n')\n",
    "    word = sentence[:-1]\n",
    "    words_train.append(word)\n",
    "    labels.append(label)\n",
    "f.close()  \n",
    "\n",
    "t.fit_on_texts(words_train)\n",
    "encoded_docs_train = t.texts_to_sequences(words_train)\n",
    "padded_docs_train = pad_sequences(encoded_docs_train, maxlen=max_length, padding='post')\n",
    "\n",
    "\n",
    "labels_train=lb.fit(labels)\n",
    "labels_train=lb.transform(labels)\n",
    "\n",
    "\n",
    "vocab_size = len(t.word_index) + 1\n",
    "embedding_matrix_train = zeros((vocab_size, 100))\n",
    "for word, i in t.word_index.items():\n",
    "    embedding_vector = embeddings_index.get(word)\n",
    "    if embedding_vector is not None:\n",
    "        embedding_matrix_train[i] = embedding_vector\n",
    "\n",
    "def Preprocessing(file,embeddings_index):\n",
    "    words =[]\n",
    "    labels=[]\n",
    "    for line in file:\n",
    "        values =line.rsplit(\";\", maxsplit=1)\n",
    "        sentence = values[0].split()\n",
    "        label=values[1]#.removesuffix('\\n')\n",
    "        word = sentence[:-1]\n",
    "        words.append(word)\n",
    "        labels.append(label)\n",
    "    file.close()  \n",
    "    # integer encode the documents\n",
    "    encoded_docs = t.texts_to_sequences(words)\n",
    "    #print(encoded_docs)\n",
    "    # pad documents to a max length of8 words\n",
    "    \n",
    "    padded_docs = pad_sequences(encoded_docs, maxlen=max_length, padding='post')\n",
    "    \n",
    "    labels=lb.transform(labels)\n",
    "    \n",
    "    embedding_matrix = zeros((vocab_size, 100))\n",
    "    for word, i in t.word_index.items():\n",
    "        embedding_vector = embeddings_index.get(word)\n",
    "        if embedding_vector is not None:\n",
    "            embedding_matrix[i] = embedding_vector\n",
    "\n",
    "    \n",
    "    \n",
    "    return padded_docs, encoded_docs, embedding_matrix,words,labels\n",
    "\n",
    "#padded_docs_train, encoded_docs_train,embedding_matrix_train,words_train,labels_train=Preprocessing(f,embeddings_index)\n",
    "padded_docs_val, encoded_docs_val,embedding_matrix_val,words_val,labels_val=Preprocessing(g,embeddings_index)\n",
    "padded_docs_test, encoded_docs_test,embedding_matrix_test,words_test,labels_test=Preprocessing(h,embeddings_index)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "390a50c5",
   "metadata": {
    "id": "390a50c5"
   },
   "outputs": [],
   "source": [
    "# embedding_matrix\n",
    "#t.word_index.items()\n",
    "#words"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "7e7a60ff",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "7e7a60ff",
    "outputId": "16167dd7-ef35-40d2-c697-a4cfb53fdd2e"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'anger\\n', 'fear\\n', 'joy\\n', 'love\\n', 'sadness\\n', 'surprise\\n'}"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(padded_docs_train[3333])\n",
    "set(labels)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "qheaw4F0XEM3",
   "metadata": {
    "id": "qheaw4F0XEM3"
   },
   "source": [
    "### <font color='violet'>Train the model</font>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "a3lg-AWWeoeW",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "a3lg-AWWeoeW",
    "outputId": "5adf07f6-7a84-42c1-d1ee-9e315a43e162"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_3\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " embedding_3 (Embedding)     (None, 58, 100)           1418500   \n",
      "                                                                 \n",
      " dropout_4 (Dropout)         (None, 58, 100)           0         \n",
      "                                                                 \n",
      " lstm_3 (LSTM)               (None, 58, 60)            38640     \n",
      "                                                                 \n",
      " conv1d_4 (Conv1D)           (None, 56, 64)            11584     \n",
      "                                                                 \n",
      " max_pooling1d_4 (MaxPooling  (None, 28, 64)           0         \n",
      " 1D)                                                             \n",
      "                                                                 \n",
      " global_average_pooling1d_3   (None, 64)               0         \n",
      " (GlobalAveragePooling1D)                                        \n",
      "                                                                 \n",
      " dropout_5 (Dropout)         (None, 64)                0         \n",
      "                                                                 \n",
      " dense_12 (Dense)            (None, 60)                3900      \n",
      "                                                                 \n",
      " dropout_6 (Dropout)         (None, 60)                0         \n",
      "                                                                 \n",
      " dense_13 (Dense)            (None, 20)                1220      \n",
      "                                                                 \n",
      " dense_14 (Dense)            (None, 6)                 126       \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 1,473,970\n",
      "Trainable params: 55,470\n",
      "Non-trainable params: 1,418,500\n",
      "_________________________________________________________________\n",
      "None\n",
      "Epoch 1/70\n",
      "800/800 [==============================] - 31s 35ms/step - loss: 1.4801 - accuracy: 0.4225 - val_loss: 1.2918 - val_accuracy: 0.5115\n",
      "Epoch 2/70\n",
      "800/800 [==============================] - 29s 36ms/step - loss: 1.2577 - accuracy: 0.5251 - val_loss: 1.1119 - val_accuracy: 0.5795\n",
      "Epoch 3/70\n",
      "800/800 [==============================] - 28s 35ms/step - loss: 1.0979 - accuracy: 0.5829 - val_loss: 0.9377 - val_accuracy: 0.6580\n",
      "Epoch 4/70\n",
      "800/800 [==============================] - 28s 35ms/step - loss: 0.9248 - accuracy: 0.6608 - val_loss: 0.7837 - val_accuracy: 0.7095\n",
      "Epoch 5/70\n",
      "800/800 [==============================] - 28s 35ms/step - loss: 0.8020 - accuracy: 0.7044 - val_loss: 0.7282 - val_accuracy: 0.7350\n",
      "Epoch 6/70\n",
      "800/800 [==============================] - 27s 34ms/step - loss: 0.7213 - accuracy: 0.7297 - val_loss: 0.6615 - val_accuracy: 0.7530\n",
      "Epoch 7/70\n",
      "800/800 [==============================] - 27s 34ms/step - loss: 0.6520 - accuracy: 0.7567 - val_loss: 0.5953 - val_accuracy: 0.7780\n",
      "Epoch 8/70\n",
      "800/800 [==============================] - 27s 34ms/step - loss: 0.5931 - accuracy: 0.7791 - val_loss: 0.5511 - val_accuracy: 0.7950\n",
      "Epoch 9/70\n",
      "800/800 [==============================] - 27s 34ms/step - loss: 0.5746 - accuracy: 0.7857 - val_loss: 0.5664 - val_accuracy: 0.7865\n",
      "Epoch 10/70\n",
      "800/800 [==============================] - 27s 34ms/step - loss: 0.5323 - accuracy: 0.7987 - val_loss: 0.5192 - val_accuracy: 0.8115\n",
      "Epoch 11/70\n",
      "800/800 [==============================] - 27s 34ms/step - loss: 0.5118 - accuracy: 0.8091 - val_loss: 0.4953 - val_accuracy: 0.8125\n",
      "Epoch 12/70\n",
      "800/800 [==============================] - 27s 34ms/step - loss: 0.4887 - accuracy: 0.8139 - val_loss: 0.4776 - val_accuracy: 0.8195\n",
      "Epoch 13/70\n",
      "800/800 [==============================] - 31s 38ms/step - loss: 0.4721 - accuracy: 0.8196 - val_loss: 0.4948 - val_accuracy: 0.8145\n",
      "Epoch 14/70\n",
      "800/800 [==============================] - 27s 34ms/step - loss: 0.4537 - accuracy: 0.8271 - val_loss: 0.4673 - val_accuracy: 0.8245\n",
      "Epoch 15/70\n",
      "800/800 [==============================] - 27s 34ms/step - loss: 0.4408 - accuracy: 0.8307 - val_loss: 0.4623 - val_accuracy: 0.8220\n",
      "Epoch 16/70\n",
      "800/800 [==============================] - 28s 35ms/step - loss: 0.4262 - accuracy: 0.8378 - val_loss: 0.4673 - val_accuracy: 0.8235\n",
      "Epoch 17/70\n",
      "800/800 [==============================] - 27s 34ms/step - loss: 0.4178 - accuracy: 0.8414 - val_loss: 0.4771 - val_accuracy: 0.8190\n",
      "Epoch 18/70\n",
      "800/800 [==============================] - 27s 34ms/step - loss: 0.4067 - accuracy: 0.8453 - val_loss: 0.4335 - val_accuracy: 0.8320\n",
      "Epoch 19/70\n",
      "800/800 [==============================] - 27s 34ms/step - loss: 0.4010 - accuracy: 0.8460 - val_loss: 0.4376 - val_accuracy: 0.8285\n",
      "Epoch 20/70\n",
      "800/800 [==============================] - 27s 34ms/step - loss: 0.3987 - accuracy: 0.8446 - val_loss: 0.4374 - val_accuracy: 0.8310\n",
      "Epoch 21/70\n",
      "800/800 [==============================] - 28s 35ms/step - loss: 0.3790 - accuracy: 0.8511 - val_loss: 0.4367 - val_accuracy: 0.8300\n",
      "Epoch 22/70\n",
      "800/800 [==============================] - 27s 34ms/step - loss: 0.3735 - accuracy: 0.8561 - val_loss: 0.4263 - val_accuracy: 0.8295\n",
      "Epoch 23/70\n",
      "800/800 [==============================] - 27s 34ms/step - loss: 0.3662 - accuracy: 0.8559 - val_loss: 0.4379 - val_accuracy: 0.8320\n",
      "Epoch 24/70\n",
      "800/800 [==============================] - 28s 34ms/step - loss: 0.3591 - accuracy: 0.8595 - val_loss: 0.4313 - val_accuracy: 0.8290\n",
      "Epoch 25/70\n",
      "800/800 [==============================] - 27s 34ms/step - loss: 0.3572 - accuracy: 0.8604 - val_loss: 0.4157 - val_accuracy: 0.8390\n",
      "Epoch 26/70\n",
      "800/800 [==============================] - 27s 34ms/step - loss: 0.3510 - accuracy: 0.8602 - val_loss: 0.4556 - val_accuracy: 0.8330\n",
      "Epoch 27/70\n",
      "800/800 [==============================] - 27s 34ms/step - loss: 0.3440 - accuracy: 0.8647 - val_loss: 0.4352 - val_accuracy: 0.8345\n",
      "Epoch 28/70\n",
      "800/800 [==============================] - 27s 34ms/step - loss: 0.3418 - accuracy: 0.8658 - val_loss: 0.4280 - val_accuracy: 0.8390\n",
      "Epoch 29/70\n",
      "800/800 [==============================] - 27s 34ms/step - loss: 0.3325 - accuracy: 0.8689 - val_loss: 0.4346 - val_accuracy: 0.8305\n",
      "Epoch 30/70\n",
      "800/800 [==============================] - 27s 34ms/step - loss: 0.3311 - accuracy: 0.8716 - val_loss: 0.4164 - val_accuracy: 0.8385\n",
      "Epoch 31/70\n",
      "800/800 [==============================] - 27s 34ms/step - loss: 0.3265 - accuracy: 0.8711 - val_loss: 0.4241 - val_accuracy: 0.8380\n",
      "63/63 [==============================] - 2s 16ms/step - loss: 0.4157 - accuracy: 0.8390\n",
      "Accuracy: 83.899999\n"
     ]
    }
   ],
   "source": [
    "tf.random.set_seed(79)\n",
    "opt = keras.optimizers.Adam(learning_rate=0.0005, beta_1=0.8,beta_2=0.999)\n",
    "# define model\n",
    "model = keras.Sequential()\n",
    "#(14185, 100) shape dimension\n",
    "model.add(keras.layers.Embedding(vocab_size, 100, weights=[embedding_matrix_train], input_length=max_length, trainable=False))\n",
    "#model.add(keras.layers.Flatten())\n",
    "\n",
    "#model.add(tf.keras.layers.Conv1D(32, 2, activation='relu'))\n",
    "#model.add(tf.keras.layers.MaxPooling1D(pool_size=2))\n",
    "#model.add(tf.keras.layers.Conv1D(64, 2, activation='relu'))\n",
    "#model.add(tf.keras.layers.MaxPooling1D(pool_size=1))\n",
    "#model.add(tf.keras.layers.GlobalAveragePooling1D()) #put this if after the conv1d you have a dense layer\n",
    "#tf.expand_dims(embedding_matrix_train.shape,axis=1)\n",
    "model.add(keras.layers.Dropout(rate=0.1))\n",
    "model.add(tf.keras.layers.LSTM(60,input_shape=(embedding_matrix_train.shape[0],embedding_matrix_train.shape[1]),return_sequences=True))\n",
    "#put return_sequences=True if you put the \n",
    "#LSTM layer before Conv1d or before another LSTM layer\n",
    "#model.add(keras.layers.Flatten()) #insert this to pass from MaxPooling to Fully connected neural network without GlobalAveragePooling1d\n",
    "\n",
    "#model.add(tf.keras.layers.Conv1D(32, 6, activation='relu'))\n",
    "#model.add(tf.keras.layers.MaxPooling1D(pool_size=2))\n",
    "model.add(tf.keras.layers.Conv1D(64, 3, activation='relu'))\n",
    "model.add(tf.keras.layers.MaxPooling1D(pool_size=2))\n",
    "model.add(tf.keras.layers.GlobalAveragePooling1D())\n",
    "#tf.expand_dims(embedding_matrix_train.shape,axis=1)#np.expand_dims(embedding_matrix_train,axis=1)\n",
    "#model.add(keras.layers.Flatten())\n",
    "\n",
    "                     # keras.layers.Dense(3000,activation='elu'\n",
    "                  #,kernel_initializer='glorot_normal'\n",
    "                 #,kernel_regularizer=keras.regularizers.l2(0.1)\n",
    "                # ),\n",
    "#keras.layers.BatchNormalization(),\n",
    "model.add(keras.layers.Dropout(rate=0.1))\n",
    "model.add(keras.layers.Dense(60,activation='elu'\n",
    "                  ,kernel_initializer='GlorotNormal'\n",
    "               #  ,kernel_regularizer=keras.regularizers.l2(0.02)\n",
    "                 ))\n",
    "#model.add(keras.layers.Dropout(rate=0.1))\n",
    "#keras.layers.BatchNormalization(),\n",
    "#model.add(keras.layers.Dense(40,activation='elu'\n",
    "           #       ,kernel_initializer='GlorotNormal'\n",
    "               #  ,kernel_regularizer=keras.regularizers.l2(0.02)\n",
    "               #  ))\n",
    "model.add(keras.layers.Dropout(rate=0.1))\n",
    "#keras.layers.BatchNormalization(),\n",
    "model.add(keras.layers.Dense(20,activation='elu',kernel_initializer='GlorotNormal'\n",
    "#  ,kernel_regularizer=keras.regularizers.l2(0.02)\n",
    "                ))\n",
    "model.add(keras.layers.Dense(6,activation='sigmoid'))\n",
    "# compile the model\n",
    "opt = keras.optimizers.Adam(learning_rate=0.00038, beta_1=0.8,beta_2=0.999)\n",
    "model.compile(optimizer=opt, loss='sparse_categorical_crossentropy', metrics=['accuracy'])\n",
    "# summarize the model\n",
    "print(model.summary())\n",
    "# fit the model\n",
    "early_stopping_cb=keras.callbacks.EarlyStopping(patience=6,restore_best_weights=True)\n",
    "history=model.fit(padded_docs_train, labels_train, epochs=70,validation_data=(padded_docs_val, labels_val),callbacks=[early_stopping_cb],batch_size=20)\n",
    "# evaluate the model\n",
    "loss, accuracy = model.evaluate(padded_docs_val, labels_val)\n",
    "print('Accuracy: %f' % (accuracy*100))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "iuMr1qvjYGwl",
   "metadata": {
    "id": "iuMr1qvjYGwl"
   },
   "source": [
    "### <font color='violet'> History plot </font>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "eQmr4TAvOWcH",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 324
    },
    "id": "eQmr4TAvOWcH",
    "outputId": "ffc1cdce-0b44-48c2-e15b-c0e7826569b2"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x360 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "pd.DataFrame(history.history).plot(figsize=(8, 5))\n",
    "plt.grid(True)\n",
    "#plt.gca().set_xlim(0,33)\n",
    "plt.gca().set_ylim(0,1)\n",
    "model.save('Conv1d_LSTM_58_3.h5')\n",
    "#Accuracy: 0.839"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "wx9sRqqxYKtN",
   "metadata": {
    "id": "wx9sRqqxYKtN"
   },
   "source": [
    "### <font color='violet'>More details about the model</font>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "Gv5T_YL0OW5G",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "Gv5T_YL0OW5G",
    "outputId": "b3b91976-e41c-497d-d43f-5d6b1177d4e5"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_3\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " embedding_3 (Embedding)     (None, 58, 100)           1418500   \n",
      "                                                                 \n",
      " dropout_4 (Dropout)         (None, 58, 100)           0         \n",
      "                                                                 \n",
      " lstm_3 (LSTM)               (None, 58, 60)            38640     \n",
      "                                                                 \n",
      " conv1d_4 (Conv1D)           (None, 56, 64)            11584     \n",
      "                                                                 \n",
      " max_pooling1d_4 (MaxPooling  (None, 28, 64)           0         \n",
      " 1D)                                                             \n",
      "                                                                 \n",
      " global_average_pooling1d_3   (None, 64)               0         \n",
      " (GlobalAveragePooling1D)                                        \n",
      "                                                                 \n",
      " dropout_5 (Dropout)         (None, 64)                0         \n",
      "                                                                 \n",
      " dense_12 (Dense)            (None, 60)                3900      \n",
      "                                                                 \n",
      " dropout_6 (Dropout)         (None, 60)                0         \n",
      "                                                                 \n",
      " dense_13 (Dense)            (None, 20)                1220      \n",
      "                                                                 \n",
      " dense_14 (Dense)            (None, 6)                 126       \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 1,473,970\n",
      "Trainable params: 55,470\n",
      "Non-trainable params: 1,418,500\n",
      "_________________________________________________________________\n",
      "0 <keras.layers.embeddings.Embedding object at 0x7f801d28c790>\n",
      "   no activation attribute\n",
      "1 <keras.layers.core.dropout.Dropout object at 0x7f801d28c550>\n",
      "   no activation attribute\n",
      "2 <keras.layers.recurrent_v2.LSTM object at 0x7f801d29bf10>\n",
      "     <function tanh at 0x7f802a8ed290>\n",
      "3 <keras.layers.convolutional.Conv1D object at 0x7f801d2a98d0>\n",
      "     <function relu at 0x7f802a8e8ef0>\n",
      "4 <keras.layers.pooling.MaxPooling1D object at 0x7f801d2b7e90>\n",
      "   no activation attribute\n",
      "5 <keras.layers.pooling.GlobalAveragePooling1D object at 0x7f801d112a90>\n",
      "   no activation attribute\n",
      "6 <keras.layers.core.dropout.Dropout object at 0x7f801d1fd590>\n",
      "   no activation attribute\n",
      "7 <keras.layers.core.dense.Dense object at 0x7f801d2a5090>\n",
      "     <function elu at 0x7f802a8e8680>\n",
      "8 <keras.layers.core.dropout.Dropout object at 0x7f801d296490>\n",
      "   no activation attribute\n",
      "9 <keras.layers.core.dense.Dense object at 0x7f801dd05850>\n",
      "     <function elu at 0x7f802a8e8680>\n",
      "10 <keras.layers.core.dense.Dense object at 0x7f801d126f90>\n",
      "     <function sigmoid at 0x7f802a8ed440>\n",
      "0   <keras.layers.embeddings.Embedding object at 0x7f801d28c790>: \n",
      "{'name': 'embedding_3', 'trainable': False, 'batch_input_shape': (None, 58), 'dtype': 'float32', 'input_dim': 14185, 'output_dim': 100, 'embeddings_initializer': {'class_name': 'RandomUniform', 'config': {'minval': -0.05, 'maxval': 0.05, 'seed': None}}, 'embeddings_regularizer': None, 'activity_regularizer': None, 'embeddings_constraint': None, 'mask_zero': False, 'input_length': 58} \n",
      "\n",
      "1   <keras.layers.core.dropout.Dropout object at 0x7f801d28c550>: \n",
      "{'name': 'dropout_4', 'trainable': True, 'dtype': 'float32', 'rate': 0.1, 'noise_shape': None, 'seed': None} \n",
      "\n",
      "2   <keras.layers.recurrent_v2.LSTM object at 0x7f801d29bf10>: \n",
      "{'name': 'lstm_3', 'trainable': True, 'batch_input_shape': (None, 14185, 100), 'dtype': 'float32', 'return_sequences': True, 'return_state': False, 'go_backwards': False, 'stateful': False, 'unroll': False, 'time_major': False, 'units': 60, 'activation': 'tanh', 'recurrent_activation': 'sigmoid', 'use_bias': True, 'kernel_initializer': {'class_name': 'GlorotUniform', 'config': {'seed': None}}, 'recurrent_initializer': {'class_name': 'Orthogonal', 'config': {'gain': 1.0, 'seed': None}}, 'bias_initializer': {'class_name': 'Zeros', 'config': {}}, 'unit_forget_bias': True, 'kernel_regularizer': None, 'recurrent_regularizer': None, 'bias_regularizer': None, 'activity_regularizer': None, 'kernel_constraint': None, 'recurrent_constraint': None, 'bias_constraint': None, 'dropout': 0.0, 'recurrent_dropout': 0.0, 'implementation': 2} \n",
      "\n",
      "3   <keras.layers.convolutional.Conv1D object at 0x7f801d2a98d0>: \n",
      "{'name': 'conv1d_4', 'trainable': True, 'dtype': 'float32', 'filters': 64, 'kernel_size': (3,), 'strides': (1,), 'padding': 'valid', 'data_format': 'channels_last', 'dilation_rate': (1,), 'groups': 1, 'activation': 'relu', 'use_bias': True, 'kernel_initializer': {'class_name': 'GlorotUniform', 'config': {'seed': None}}, 'bias_initializer': {'class_name': 'Zeros', 'config': {}}, 'kernel_regularizer': None, 'bias_regularizer': None, 'activity_regularizer': None, 'kernel_constraint': None, 'bias_constraint': None} \n",
      "\n",
      "4   <keras.layers.pooling.MaxPooling1D object at 0x7f801d2b7e90>: \n",
      "{'name': 'max_pooling1d_4', 'trainable': True, 'dtype': 'float32', 'strides': (2,), 'pool_size': (2,), 'padding': 'valid', 'data_format': 'channels_last'} \n",
      "\n",
      "5   <keras.layers.pooling.GlobalAveragePooling1D object at 0x7f801d112a90>: \n",
      "{'name': 'global_average_pooling1d_3', 'trainable': True, 'dtype': 'float32', 'data_format': 'channels_last', 'keepdims': False} \n",
      "\n",
      "6   <keras.layers.core.dropout.Dropout object at 0x7f801d1fd590>: \n",
      "{'name': 'dropout_5', 'trainable': True, 'dtype': 'float32', 'rate': 0.1, 'noise_shape': None, 'seed': None} \n",
      "\n",
      "7   <keras.layers.core.dense.Dense object at 0x7f801d2a5090>: \n",
      "{'name': 'dense_12', 'trainable': True, 'dtype': 'float32', 'units': 60, 'activation': 'elu', 'use_bias': True, 'kernel_initializer': {'class_name': 'GlorotNormal', 'config': {'seed': None}}, 'bias_initializer': {'class_name': 'Zeros', 'config': {}}, 'kernel_regularizer': None, 'bias_regularizer': None, 'activity_regularizer': None, 'kernel_constraint': None, 'bias_constraint': None} \n",
      "\n",
      "8   <keras.layers.core.dropout.Dropout object at 0x7f801d296490>: \n",
      "{'name': 'dropout_6', 'trainable': True, 'dtype': 'float32', 'rate': 0.1, 'noise_shape': None, 'seed': None} \n",
      "\n",
      "9   <keras.layers.core.dense.Dense object at 0x7f801dd05850>: \n",
      "{'name': 'dense_13', 'trainable': True, 'dtype': 'float32', 'units': 20, 'activation': 'elu', 'use_bias': True, 'kernel_initializer': {'class_name': 'GlorotNormal', 'config': {'seed': None}}, 'bias_initializer': {'class_name': 'Zeros', 'config': {}}, 'kernel_regularizer': None, 'bias_regularizer': None, 'activity_regularizer': None, 'kernel_constraint': None, 'bias_constraint': None} \n",
      "\n",
      "10   <keras.layers.core.dense.Dense object at 0x7f801d126f90>: \n",
      "{'name': 'dense_14', 'trainable': True, 'dtype': 'float32', 'units': 6, 'activation': 'sigmoid', 'use_bias': True, 'kernel_initializer': {'class_name': 'GlorotUniform', 'config': {'seed': None}}, 'bias_initializer': {'class_name': 'Zeros', 'config': {}}, 'kernel_regularizer': None, 'bias_regularizer': None, 'activity_regularizer': None, 'kernel_constraint': None, 'bias_constraint': None} \n",
      "\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "{'amsgrad': False,\n",
       " 'beta_1': 0.8,\n",
       " 'beta_2': 0.999,\n",
       " 'decay': 0.0,\n",
       " 'epsilon': 1e-07,\n",
       " 'learning_rate': 0.00038,\n",
       " 'name': 'Adam'}"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model=keras.models.load_model('Conv1d_LSTM_58_3.h5') \n",
    "model.summary()\n",
    "\n",
    "#to visualize activation functions\n",
    "for i, layer in enumerate (model.layers):\n",
    "    print (i, layer)\n",
    "    try:\n",
    "        print (\"    \",layer.activation)\n",
    "    except AttributeError:\n",
    "        print('   no activation attribute')\n",
    "#specific info about each layer\n",
    "for i in range(len(model.layers)):\n",
    "    print(f'{i}   {model.layers[i]}: \\n{model.layers[i].get_config()} \\n')\n",
    "#info about optimizers\n",
    "model.optimizer.get_config()        "
   ]
  }
 ],
 "metadata": {
  "colab": {
   "collapsed_sections": [],
   "name": "LSTM_Conv1d.ipynb",
   "provenance": []
  },
  "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": "3.9.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}