Transformers_HF/keras_hf_RoBERTa.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "virtual-optimum",
   "metadata": {
    "executionInfo": {
     "elapsed": 686,
     "status": "ok",
     "timestamp": 1632978616491,
     "user": {
      "displayName": "Prabath Hetti Mudiyanselage",
      "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s64",
      "userId": "08605282177904379825"
     },
     "user_tz": -600
    },
    "id": "virtual-optimum"
   },
   "outputs": [],
   "source": [
    "# GPU Reset if needed\n",
    "\n",
    "from numba import cuda \n",
    "device = cuda.get_current_device()\n",
    "device.reset()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "4c19ba3c",
   "metadata": {
    "executionInfo": {
     "elapsed": 829,
     "status": "ok",
     "timestamp": 1632978617898,
     "user": {
      "displayName": "Prabath Hetti Mudiyanselage",
      "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s64",
      "userId": "08605282177904379825"
     },
     "user_tz": -600
    },
    "id": "4c19ba3c"
   },
   "outputs": [],
   "source": [
    "# Disable CUDA for TF 2.x\n",
    "\n",
    "# import os\n",
    "# os.environ[\"CUDA_VISIBLE_DEVICES\"]=\"-1\" "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "iA9Ipq-4zuPT",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 6,
     "status": "ok",
     "timestamp": 1632978617899,
     "user": {
      "displayName": "Prabath Hetti Mudiyanselage",
      "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s64",
      "userId": "08605282177904379825"
     },
     "user_tz": -600
    },
    "id": "iA9Ipq-4zuPT",
    "outputId": "622e58cd-9251-4c7b-bd63-33103b40de7b"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "D:\\NLP_Last\\files\\UPLOAD\\Transformers_HF\n"
     ]
    }
   ],
   "source": [
    "# Mount Google Drive if running on Google Colab\n",
    "\n",
    "# from google.colab import drive\n",
    "import os\n",
    "\n",
    "# drive.mount('/content/drive')\n",
    "# os.chdir('/content/drive/My Drive/Colab Notebooks/hf/')\n",
    "print(os.getcwd())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "MI6QgDjY2g7o",
   "metadata": {
    "executionInfo": {
     "elapsed": 5,
     "status": "ok",
     "timestamp": 1632978617900,
     "user": {
      "displayName": "Prabath Hetti Mudiyanselage",
      "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s64",
      "userId": "08605282177904379825"
     },
     "user_tz": -600
    },
    "id": "MI6QgDjY2g7o"
   },
   "outputs": [],
   "source": [
    "# List current directory\n",
    "\n",
    "#!dir"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "wWbrBVZ7CoPF",
   "metadata": {
    "executionInfo": {
     "elapsed": 4,
     "status": "ok",
     "timestamp": 1632978617900,
     "user": {
      "displayName": "Prabath Hetti Mudiyanselage",
      "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s64",
      "userId": "08605282177904379825"
     },
     "user_tz": -600
    },
    "id": "wWbrBVZ7CoPF"
   },
   "outputs": [],
   "source": [
    "# Remove current logs\n",
    "\n",
    "#!rm -rf logs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "distinct-breath",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 1918,
     "status": "ok",
     "timestamp": 1632978619814,
     "user": {
      "displayName": "Prabath Hetti Mudiyanselage",
      "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s64",
      "userId": "08605282177904379825"
     },
     "user_tz": -600
    },
    "id": "distinct-breath",
    "outputId": "54231406-bc2c-4652-d5cd-5ffd31969919"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Tensorflow precision mode : float32\n",
      "TensorFlow version : 2.9.0-dev20220102\n",
      "Numpy version : 1.22.0\n",
      "Pandas version : 1.3.4\n"
     ]
    }
   ],
   "source": [
    "# Import Tensorflow and related libraries\n",
    "\n",
    "import tensorflow\n",
    "\n",
    "print(\"Tensorflow precision mode : \" + tensorflow.keras.backend.floatx())\n",
    "\n",
    "# # Set Tensorflow precision mode to 'float16' to speedup with newer GPU\n",
    "# tensorflow.keras.backend.set_floatx('float32')\n",
    "\n",
    "# print(\"Tensorflow precision mode (after change) : \" + tensorflow.keras.backend.floatx())\n",
    "\n",
    "from tensorflow.keras.models import Model\n",
    "from tensorflow.keras.layers import Dense\n",
    "from tensorflow.keras.layers import Flatten\n",
    "from tensorflow.keras.layers import Dropout\n",
    "\n",
    "from tensorflow.keras.layers import Input\n",
    "\n",
    "from tensorflow.keras.utils import plot_model\n",
    "\n",
    "from sklearn.preprocessing import LabelEncoder\n",
    "\n",
    "print(\"TensorFlow version : \" + tensorflow.__version__)\n",
    "\n",
    "import numpy as np\n",
    "print(\"Numpy version : \" + np.version.version)\n",
    "\n",
    "import pandas as pd\n",
    "print(\"Pandas version : \" + pd.__version__)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "circular-angle",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 10,
     "status": "ok",
     "timestamp": 1632978619814,
     "user": {
      "displayName": "Prabath Hetti Mudiyanselage",
      "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s64",
      "userId": "08605282177904379825"
     },
     "user_tz": -600
    },
    "id": "circular-angle",
    "outputId": "2a47cece-f4dd-40cf-f26e-d15bfe89815b"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Name: /physical_device:GPU:0   Type: GPU\n"
     ]
    }
   ],
   "source": [
    "# Display available GPUs\n",
    "\n",
    "physical_devices = tensorflow.config.list_physical_devices(('GPU'))\n",
    "for gpu in physical_devices:\n",
    "    print(\"Name:\", gpu.name, \"  Type:\", gpu.device_type)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "618b9a2b",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 8,
     "status": "ok",
     "timestamp": 1632978619814,
     "user": {
      "displayName": "Prabath Hetti Mudiyanselage",
      "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s64",
      "userId": "08605282177904379825"
     },
     "user_tz": -600
    },
    "id": "618b9a2b",
    "outputId": "e372b3da-c98b-4689-e225-3ffefcba075c"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fri Feb 25 08:50:24 2022       \n",
      "+-----------------------------------------------------------------------------+\n",
      "| NVIDIA-SMI 496.13       Driver Version: 496.13       CUDA Version: 11.5     |\n",
      "|-------------------------------+----------------------+----------------------+\n",
      "| GPU  Name            TCC/WDDM | Bus-Id        Disp.A | Volatile Uncorr. ECC |\n",
      "| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |\n",
      "|                               |                      |               MIG M. |\n",
      "|===============================+======================+======================|\n",
      "|   0  NVIDIA GeForce ... WDDM  | 00000000:0D:00.0  On |                  N/A |\n",
      "|  0%   51C    P2    42W / 170W |   2169MiB / 12288MiB |      0%      Default |\n",
      "|                               |                      |                  N/A |\n",
      "+-------------------------------+----------------------+----------------------+\n",
      "                                                                               \n",
      "+-----------------------------------------------------------------------------+\n",
      "| Processes:                                                                  |\n",
      "|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |\n",
      "|        ID   ID                                                   Usage      |\n",
      "|=============================================================================|\n",
      "|    0   N/A  N/A      1604    C+G   Insufficient Permissions        N/A      |\n",
      "|    0   N/A  N/A      2476      C   ...onda3\\envs\\tf2\\python.exe    N/A      |\n",
      "|    0   N/A  N/A      5204    C+G   C:\\Windows\\explorer.exe         N/A      |\n",
      "|    0   N/A  N/A      5748      C   ...onda3\\envs\\tf2\\python.exe    N/A      |\n",
      "|    0   N/A  N/A      6308    C+G   ...2txyewy\\TextInputHost.exe    N/A      |\n",
      "|    0   N/A  N/A      7460    C+G   ...y\\ShellExperienceHost.exe    N/A      |\n",
      "|    0   N/A  N/A      7488    C+G   ...5n1h2txyewy\\SearchApp.exe    N/A      |\n",
      "|    0   N/A  N/A      7616    C+G   Insufficient Permissions        N/A      |\n",
      "|    0   N/A  N/A      9452    C+G   ...zilla Firefox\\firefox.exe    N/A      |\n",
      "|    0   N/A  N/A     13036    C+G   ... Download Manager\\fdm.exe    N/A      |\n",
      "|    0   N/A  N/A     13108    C+G   ...zilla Firefox\\firefox.exe    N/A      |\n",
      "|    0   N/A  N/A     14216    C+G   ...5.0.3.0\\GoogleDriveFS.exe    N/A      |\n",
      "|    0   N/A  N/A     15260    C+G   ...enTabletDriver.UX.Wpf.exe    N/A      |\n",
      "|    0   N/A  N/A     15884    C+G   ...in7x64\\steamwebhelper.exe    N/A      |\n",
      "+-----------------------------------------------------------------------------+\n"
     ]
    }
   ],
   "source": [
    "# Show GPU status\n",
    "\n",
    "!nvidia-smi"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "0f583e47",
   "metadata": {
    "executionInfo": {
     "elapsed": 5,
     "status": "ok",
     "timestamp": 1632978619815,
     "user": {
      "displayName": "Prabath Hetti Mudiyanselage",
      "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s64",
      "userId": "08605282177904379825"
     },
     "user_tz": -600
    },
    "id": "0f583e47"
   },
   "outputs": [],
   "source": [
    "# Fix for \"CUDA_ERROR_ILLEGAL_ADDRESS\"\n",
    "\n",
    "# Ref : https://github.com/tensorflow/tensorflow/issues/40814#issuecomment-663838196\n",
    "\n",
    "os.environ['TF_CUDNN_DETERMINISTIC']='1'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "preceding-travel",
   "metadata": {
    "executionInfo": {
     "elapsed": 5,
     "status": "ok",
     "timestamp": 1632978619815,
     "user": {
      "displayName": "Prabath Hetti Mudiyanselage",
      "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s64",
      "userId": "08605282177904379825"
     },
     "user_tz": -600
    },
    "id": "preceding-travel"
   },
   "outputs": [],
   "source": [
    "# Enable Memory Growth\n",
    "\n",
    "try:\n",
    "    if physical_devices != []:\n",
    "        tensorflow.config.experimental.set_memory_growth(physical_devices[0], True)\n",
    "        physical_devices[0]\n",
    "except:\n",
    "    print(\"error\")      "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "primary-bryan",
   "metadata": {
    "executionInfo": {
     "elapsed": 4,
     "status": "ok",
     "timestamp": 1632978619815,
     "user": {
      "displayName": "Prabath Hetti Mudiyanselage",
      "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s64",
      "userId": "08605282177904379825"
     },
     "user_tz": -600
    },
    "id": "primary-bryan"
   },
   "outputs": [],
   "source": [
    "# Global Parameters\n",
    "\n",
    "author_list = ['Charles Dickens','Jane Austen','Sir Arthur Conan Doyle','George Eliot','Jules Verne']\n",
    "seed_number = 18\n",
    "dataset_file  = \"dataset.csv\"\n",
    "validation_dataset_file  = \"validation_dataset.csv\"\n",
    "\n",
    "# For 3 GB VRAM : \n",
    "# batch_size = 2 when seq_length = 128, otherwise \"out of memory\"\n",
    "# Typical single epoch runtime = 20 mins\n",
    "\n",
    "batch_size = 64 # number of batch size\n",
    "epochs = 5 # number of maximum epochs to run\n",
    "seq_length = 100#300 # BERT Model sequence length\n",
    "dropout_value = 0.1 # Dropout layer value for Model\n",
    "init_lr = 3e-5 # Initial learning rate"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "e41a7202",
   "metadata": {
    "executionInfo": {
     "elapsed": 5,
     "status": "ok",
     "timestamp": 1632978619816,
     "user": {
      "displayName": "Prabath Hetti Mudiyanselage",
      "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s64",
      "userId": "08605282177904379825"
     },
     "user_tz": -600
    },
    "id": "e41a7202"
   },
   "outputs": [],
   "source": [
    "# Reading validation data (from seperate books)\n",
    "\n",
    "validation_data = pd.read_csv(validation_dataset_file, usecols = ['author','text'])\n",
    "X_val_2 = validation_data.text\n",
    "y_val_2 = validation_data.author"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "attractive-oriental",
   "metadata": {
    "executionInfo": {
     "elapsed": 4,
     "status": "ok",
     "timestamp": 1632978619816,
     "user": {
      "displayName": "Prabath Hetti Mudiyanselage",
      "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s64",
      "userId": "08605282177904379825"
     },
     "user_tz": -600
    },
    "id": "attractive-oriental"
   },
   "outputs": [],
   "source": [
    "# Reading training/test/validation data\n",
    "\n",
    "data = pd.read_csv(dataset_file, usecols = ['author','text'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "dried-aerospace",
   "metadata": {
    "executionInfo": {
     "elapsed": 12,
     "status": "ok",
     "timestamp": 1632978620779,
     "user": {
      "displayName": "Prabath Hetti Mudiyanselage",
      "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s64",
      "userId": "08605282177904379825"
     },
     "user_tz": -600
    },
    "id": "dried-aerospace"
   },
   "outputs": [],
   "source": [
    "#data.sample(5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "signal-football",
   "metadata": {
    "executionInfo": {
     "elapsed": 11,
     "status": "ok",
     "timestamp": 1632978620779,
     "user": {
      "displayName": "Prabath Hetti Mudiyanselage",
      "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s64",
      "userId": "08605282177904379825"
     },
     "user_tz": -600
    },
    "id": "signal-football"
   },
   "outputs": [],
   "source": [
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "# Train, Test, Validation Spit ratios 60%, 20%, 20% \n",
    "\n",
    "X_train, X_val, y_train, y_val = train_test_split(data[\"text\"], data[\"author\"], test_size=0.2, random_state=2000)\n",
    "X_train, X_test, y_train, y_test = train_test_split(X_train, y_train, test_size=0.25, random_state=seed_number) # 0.25 x 0.8 = 0.2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "a34aaa20",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 11,
     "status": "ok",
     "timestamp": 1632978620779,
     "user": {
      "displayName": "Prabath Hetti Mudiyanselage",
      "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s64",
      "userId": "08605282177904379825"
     },
     "user_tz": -600
    },
    "id": "a34aaa20",
    "outputId": "0b1ac93d-84a8-4493-a0be-6c2477e902bf"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of training data: 15000\n",
      "Number of test data: 5000\n",
      "Number of validation data (from same books): 5000\n",
      "Number of validation data (from seperate books): 4000\n",
      "Number of max characters in training dataset: 370\n"
     ]
    }
   ],
   "source": [
    "print(\"Number of training data:\",len(X_train))\n",
    "print(\"Number of test data:\",len(X_test))\n",
    "print(\"Number of validation data (from same books):\",len(X_val))\n",
    "print(\"Number of validation data (from seperate books):\",len(X_val_2))\n",
    "print(\"Number of max characters in training dataset:\",max(X_train.apply(len)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "strong-ukraine",
   "metadata": {
    "executionInfo": {
     "elapsed": 8,
     "status": "ok",
     "timestamp": 1632978620780,
     "user": {
      "displayName": "Prabath Hetti Mudiyanselage",
      "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s64",
      "userId": "08605282177904379825"
     },
     "user_tz": -600
    },
    "id": "strong-ukraine"
   },
   "outputs": [],
   "source": [
    "# Encoding label text to integeres\n",
    "\n",
    "label_tokenizer = LabelEncoder()\n",
    "label_tokenizer.fit(author_list)\n",
    "label_index = label_tokenizer.classes_\n",
    "\n",
    "y_train_sequences = np.array(label_tokenizer.transform(y_train))\n",
    "y_test_sequences = np.array(label_tokenizer.transform(y_test))\n",
    "y_val_sequences = np.array(label_tokenizer.transform(y_val))\n",
    "y_val_sequences_2 = np.array(label_tokenizer.transform(y_val_2))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "informed-precipitation",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 7,
     "status": "ok",
     "timestamp": 1632978620780,
     "user": {
      "displayName": "Prabath Hetti Mudiyanselage",
      "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s64",
      "userId": "08605282177904379825"
     },
     "user_tz": -600
    },
    "id": "informed-precipitation",
    "outputId": "d49da39e-4588-43ee-dc4c-9ea9bbec9252"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array(['Charles Dickens', 'George Eliot', 'Jane Austen', 'Jules Verne',\n",
       "       'Sir Arthur Conan Doyle'], dtype='<U22')"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "label_index"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "numeric-produce",
   "metadata": {
    "id": "numeric-produce"
   },
   "source": [
    "### RoBERTa Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "6c978cf2",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 4241,
     "status": "ok",
     "timestamp": 1632978625017,
     "user": {
      "displayName": "Prabath Hetti Mudiyanselage",
      "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s64",
      "userId": "08605282177904379825"
     },
     "user_tz": -600
    },
    "id": "6c978cf2",
    "outputId": "f13b2fef-098d-4643-884a-1ce3b3aa2add"
   },
   "outputs": [],
   "source": [
    "# !pip install transformers==4.6.0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "5029e0df",
   "metadata": {
    "executionInfo": {
     "elapsed": 14,
     "status": "ok",
     "timestamp": 1632978625025,
     "user": {
      "displayName": "Prabath Hetti Mudiyanselage",
      "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s64",
      "userId": "08605282177904379825"
     },
     "user_tz": -600
    },
    "id": "5029e0df"
   },
   "outputs": [],
   "source": [
    "# Load Huggingface transformers\n",
    "from transformers import RobertaConfig, TFRobertaModel, RobertaTokenizerFast"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "fb14f7a9",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 7491,
     "status": "ok",
     "timestamp": 1632978632502,
     "user": {
      "displayName": "Prabath Hetti Mudiyanselage",
      "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s64",
      "userId": "08605282177904379825"
     },
     "user_tz": -600
    },
    "id": "fb14f7a9",
    "outputId": "cc74b50b-16d1-4eb6-921b-c260005d6d75"
   },
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "9028b1fd554d4626852c8c4a99af5f8a",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Downloading:   0%|          | 0.00/899k [00:00<?, ?B/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "79bd1b814a284f63be7ef6a99ee46ec3",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Downloading:   0%|          | 0.00/456k [00:00<?, ?B/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "5681c76060e54bbea428d7669120b604",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Downloading:   0%|          | 0.00/1.36M [00:00<?, ?B/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "e49540b804a14167936b62016a8ffe09",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Downloading:   0%|          | 0.00/481 [00:00<?, ?B/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "9187bbc9a00c47b8b57e124cf5a33d0c",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Downloading:   0%|          | 0.00/657M [00:00<?, ?B/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Some layers from the model checkpoint at roberta-base were not used when initializing TFRobertaModel: ['lm_head']\n",
      "- This IS expected if you are initializing TFRobertaModel from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).\n",
      "- This IS NOT expected if you are initializing TFRobertaModel from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).\n",
      "All the layers of TFRobertaModel were initialized from the model checkpoint at roberta-base.\n",
      "If your task is similar to the task the model of the checkpoint was trained on, you can already use TFRobertaModel for predictions without further training.\n"
     ]
    }
   ],
   "source": [
    "#######################################\n",
    "### --------- Setup  ---------- ###\n",
    "\n",
    "# Ref: https://huggingface.co/transformers/model_doc/roberta.html\n",
    "\n",
    "\n",
    "# Name of the  model to use\n",
    "model_name = 'roberta-base'\n",
    "\n",
    "# Max length of tokens\n",
    "max_length = seq_length #100\n",
    "\n",
    "# Load transformers config and set output_hidden_states to False\n",
    "# config = BertConfig.from_pretrained(model_name)\n",
    "# config.output_hidden_states = False\n",
    "\n",
    "# Load BERT tokenizer\n",
    "tokenizer = RobertaTokenizerFast.from_pretrained(model_name)#BertTokenizerFast.from_pretrained(pretrained_model_name_or_path = model_name, config = config)\n",
    "\n",
    "# Load the Transformers BERT model\n",
    "transformer_model = TFRobertaModel.from_pretrained(model_name) #TFBertModel.from_pretrained(model_name, config = config)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "d2676225",
   "metadata": {
    "executionInfo": {
     "elapsed": 5392,
     "status": "ok",
     "timestamp": 1632978637892,
     "user": {
      "displayName": "Prabath Hetti Mudiyanselage",
      "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s64",
      "userId": "08605282177904379825"
     },
     "user_tz": -600
    },
    "id": "d2676225"
   },
   "outputs": [],
   "source": [
    "#######################################\n",
    "### ------- Build the model ------- ###\n",
    "\n",
    "# TF Keras documentation: https://www.tensorflow.org/api_docs/python/tf/keras/Model\n",
    "\n",
    "# Load the MainLayer\n",
    "bert = transformer_model.layers[0]\n",
    "\n",
    "# Build your model input\n",
    "input_ids = Input(shape=(max_length,), name='input_ids', dtype='int32')\n",
    "\n",
    "inputs = {'input_ids': input_ids}\n",
    "\n",
    "# Load the Transformers BERT model as a layer in a Keras model\n",
    "bert_model = bert(inputs)[1]\n",
    "dropout = Dropout(dropout_value, name='pooled_output')#Dropout(config.hidden_dropout_prob, name='pooled_output')\n",
    "pooled_output = dropout(bert_model, training=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "93802723",
   "metadata": {
    "executionInfo": {
     "elapsed": 4,
     "status": "ok",
     "timestamp": 1632978637893,
     "user": {
      "displayName": "Prabath Hetti Mudiyanselage",
      "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s64",
      "userId": "08605282177904379825"
     },
     "user_tz": -600
    },
    "id": "93802723"
   },
   "outputs": [],
   "source": [
    "def X_tokenizer(X_original):\n",
    "    X_original = tokenizer(\n",
    "    text=X_original.to_list(),\n",
    "    add_special_tokens=True,\n",
    "    max_length=max_length,\n",
    "    truncation=True,\n",
    "    padding='max_length', \n",
    "   # pad_to_max_length=\"right\",\n",
    "    return_tensors='tf',\n",
    "    return_token_type_ids = False,\n",
    "    return_attention_mask = False,\n",
    "    verbose = True)\n",
    "\n",
    "    X_original={'input_ids': X_original['input_ids']}\n",
    "    return X_original\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "9be0308f",
   "metadata": {
    "executionInfo": {
     "elapsed": 3836,
     "status": "ok",
     "timestamp": 1632978641727,
     "user": {
      "displayName": "Prabath Hetti Mudiyanselage",
      "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s64",
      "userId": "08605282177904379825"
     },
     "user_tz": -600
    },
    "id": "9be0308f"
   },
   "outputs": [],
   "source": [
    "X_train = X_tokenizer(X_train)\n",
    "X_test = X_tokenizer(X_test)\n",
    "X_val = X_tokenizer(X_val)\n",
    "X_val_2 = X_tokenizer(X_val_2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "acting-midwest",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 11,
     "status": "ok",
     "timestamp": 1632978641731,
     "user": {
      "displayName": "Prabath Hetti Mudiyanselage",
      "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s64",
      "userId": "08605282177904379825"
     },
     "user_tz": -600
    },
    "id": "acting-midwest",
    "outputId": "92844f4f-0bd8-47bf-989b-d5d9bc2eab54"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"nlp_roberta_18\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " input_ids (InputLayer)      [(None, 100)]             0         \n",
      "                                                                 \n",
      " roberta (TFRobertaMainLayer  TFBaseModelOutputWithPoo  124645632\n",
      " )                           ling(last_hidden_state=(            \n",
      "                             None, 100, 768),                    \n",
      "                              pooler_output=(None, 76            \n",
      "                             8),                                 \n",
      "                              hidden_states=None, att            \n",
      "                             entions=None)                       \n",
      "                                                                 \n",
      " pooled_output (Dropout)     (None, 768)               0         \n",
      "                                                                 \n",
      " dense (Dense)               (None, 5)                 3845      \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 124,649,477\n",
      "Trainable params: 124,649,477\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# define new model\n",
    "\n",
    "model_output = tensorflow.keras.layers.Dense(5, activation='softmax')(pooled_output)\n",
    "\n",
    "model = tensorflow.keras.Model(inputs, model_output)\n",
    "\n",
    "model._name = \"nlp_roberta_\" + str(seed_number) \n",
    "model.summary()\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "d206d079",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 369
    },
    "executionInfo": {
     "elapsed": 10,
     "status": "ok",
     "timestamp": 1632978641732,
     "user": {
      "displayName": "Prabath Hetti Mudiyanselage",
      "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s64",
      "userId": "08605282177904379825"
     },
     "user_tz": -600
    },
    "id": "d206d079",
    "outputId": "6028a783-4508-450c-e6b2-69098cd139fb"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Save the model as a png file\n",
    "plot_model(model, to_file=model._name +'.png', show_layer_names=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "stuck-dispatch",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 9,
     "status": "ok",
     "timestamp": 1632978641732,
     "user": {
      "displayName": "Prabath Hetti Mudiyanselage",
      "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s64",
      "userId": "08605282177904379825"
     },
     "user_tz": -600
    },
    "id": "stuck-dispatch",
    "outputId": "66685b6f-23c3-4d9b-c1f9-bf182d753d63"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0 input_ids True\n",
      "1 roberta True\n",
      "2 pooled_output True\n",
      "3 dense True\n"
     ]
    }
   ],
   "source": [
    "# Show the trainable status of the layers\n",
    "\n",
    "for i, layer in enumerate(model.layers):\n",
    "    print(i, layer.name, layer.trainable)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "13a91642",
   "metadata": {
    "executionInfo": {
     "elapsed": 7,
     "status": "ok",
     "timestamp": 1632978641732,
     "user": {
      "displayName": "Prabath Hetti Mudiyanselage",
      "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s64",
      "userId": "08605282177904379825"
     },
     "user_tz": -600
    },
    "id": "13a91642"
   },
   "outputs": [],
   "source": [
    "# Define Optimizer\n",
    "\n",
    "# steps_per_epoch = int(len(X_train)/batch_size) #tensorflow.data.experimental.cardinality(data_train).numpy()\n",
    "# num_train_steps = steps_per_epoch * epochs\n",
    "# num_warmup_steps = int(0.1*num_train_steps)\n",
    "# print(steps_per_epoch,num_train_steps,num_warmup_steps)\n",
    "\n",
    "# optimizer = optimization.create_optimizer(init_lr=init_lr,\n",
    "#                                           num_train_steps=num_train_steps,\n",
    "#                                           num_warmup_steps=num_warmup_steps,\n",
    "#                                           optimizer_type='adamw')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "c45b8d76",
   "metadata": {
    "executionInfo": {
     "elapsed": 7,
     "status": "ok",
     "timestamp": 1632978641732,
     "user": {
      "displayName": "Prabath Hetti Mudiyanselage",
      "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s64",
      "userId": "08605282177904379825"
     },
     "user_tz": -600
    },
    "id": "c45b8d76"
   },
   "outputs": [],
   "source": [
    "from tensorflow.keras.optimizers import Adam\n",
    "\n",
    "optimizer = Adam(\n",
    "    learning_rate=5e-05,\n",
    "    epsilon=1e-08,\n",
    "    decay=0.01,\n",
    "    clipnorm=1.0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "id": "manufactured-niagara",
   "metadata": {
    "executionInfo": {
     "elapsed": 6,
     "status": "ok",
     "timestamp": 1632978641732,
     "user": {
      "displayName": "Prabath Hetti Mudiyanselage",
      "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s64",
      "userId": "08605282177904379825"
     },
     "user_tz": -600
    },
    "id": "manufactured-niagara"
   },
   "outputs": [],
   "source": [
    "# Compile the model\n",
    "\n",
    "model.compile(\n",
    "    #optimizer=tensorflow.keras.optimizers.Adam(learning_rate=3e-5),\n",
    "    #optimizer=tensorflow.keras.optimizers.RMSprop(learning_rate=0.0001),\n",
    "    optimizer=optimizer,\n",
    "    loss=\"sparse_categorical_crossentropy\",\n",
    "    metrics=[\"accuracy\"],\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "id": "passing-chambers",
   "metadata": {
    "executionInfo": {
     "elapsed": 611,
     "status": "ok",
     "timestamp": 1632978642337,
     "user": {
      "displayName": "Prabath Hetti Mudiyanselage",
      "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s64",
      "userId": "08605282177904379825"
     },
     "user_tz": -600
    },
    "id": "passing-chambers"
   },
   "outputs": [],
   "source": [
    "# Tensorboard parameters\n",
    "\n",
    "import datetime\n",
    "log_dir = \"logs/fit/\" + datetime.datetime.now().strftime(\"%Y%m%d-%H%M%S\")\n",
    "tensorboard_callback = tensorflow.keras.callbacks.TensorBoard(log_dir=log_dir, histogram_freq=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "id": "diagnostic-hydrogen",
   "metadata": {
    "executionInfo": {
     "elapsed": 4,
     "status": "ok",
     "timestamp": 1632978642337,
     "user": {
      "displayName": "Prabath Hetti Mudiyanselage",
      "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s64",
      "userId": "08605282177904379825"
     },
     "user_tz": -600
    },
    "id": "diagnostic-hydrogen"
   },
   "outputs": [],
   "source": [
    "# Use best weights for the model\n",
    "\n",
    "early_stopping_monitor = tensorflow.keras.callbacks.EarlyStopping(\n",
    "    monitor='val_accuracy',\n",
    "    min_delta=0,\n",
    "    patience=50,\n",
    "    verbose=0,\n",
    "    mode='max',\n",
    "    baseline=None,\n",
    "    restore_best_weights=True\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "id": "unlike-ballot",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 1872494,
     "status": "ok",
     "timestamp": 1632980514827,
     "user": {
      "displayName": "Prabath Hetti Mudiyanselage",
      "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s64",
      "userId": "08605282177904379825"
     },
     "user_tz": -600
    },
    "id": "unlike-ballot",
    "outputId": "4647cd96-664c-4421-f587-e0faf73c3db7"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/5\n",
      "235/235 [==============================] - 197s 793ms/step - loss: 0.8020 - accuracy: 0.6827 - val_loss: 0.4250 - val_accuracy: 0.8436\n",
      "Epoch 2/5\n",
      "235/235 [==============================] - 188s 799ms/step - loss: 0.3293 - accuracy: 0.8775 - val_loss: 0.3567 - val_accuracy: 0.8730\n",
      "Epoch 3/5\n",
      "235/235 [==============================] - 186s 791ms/step - loss: 0.2093 - accuracy: 0.9264 - val_loss: 0.3592 - val_accuracy: 0.8822\n",
      "Epoch 4/5\n",
      "235/235 [==============================] - 185s 787ms/step - loss: 0.1442 - accuracy: 0.9495 - val_loss: 0.3504 - val_accuracy: 0.8892\n",
      "Epoch 5/5\n",
      "235/235 [==============================] - 184s 783ms/step - loss: 0.1050 - accuracy: 0.9630 - val_loss: 0.3584 - val_accuracy: 0.8966\n"
     ]
    }
   ],
   "source": [
    "# Training the model\n",
    "# callbacks are not used : early_stopping_monitor (to speed up), tensorboard_callback (has a bug with BERT)\n",
    "\n",
    "history = model.fit(X_train, y_train_sequences, epochs=epochs, batch_size = batch_size, shuffle = True, callbacks=[], validation_data=(X_test , y_test_sequences ),)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "id": "f6784583",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 441
    },
    "executionInfo": {
     "elapsed": 1020,
     "status": "ok",
     "timestamp": 1632980515835,
     "user": {
      "displayName": "Prabath Hetti Mudiyanselage",
      "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s64",
      "userId": "08605282177904379825"
     },
     "user_tz": -600
    },
    "id": "f6784583",
    "outputId": "d2de76d7-92c9-4714-e796-cd5135731726"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "dict_keys(['loss', 'accuracy', 'val_loss', 'val_accuracy'])\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x26327489a00>"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x432 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot the accuracy and loss over time \n",
    "# Reusing sample code from : https://www.tensorflow.org/text/tutorials/classify_text_with_bert\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "history_dict = history.history\n",
    "print(history_dict.keys())\n",
    "\n",
    "acc = history_dict['accuracy']\n",
    "val_acc = history_dict['val_accuracy']\n",
    "loss = history_dict['loss']\n",
    "val_loss = history_dict['val_loss']\n",
    "\n",
    "#print(loss)\n",
    "\n",
    "epochs_number = range(1, (epochs) + 1)\n",
    "fig = plt.figure(figsize=(10, 6))\n",
    "fig.tight_layout()\n",
    "\n",
    "plt.subplot(2, 1, 1)\n",
    "# \"bo\" is for \"blue dot\"\n",
    "plt.plot(epochs_number, loss, 'r', label='Training loss')\n",
    "# b is for \"solid blue line\"\n",
    "plt.plot(epochs_number, val_loss, 'b', label='Validation loss')\n",
    "plt.title('Training and validation loss')\n",
    "# plt.xlabel('Epochs')\n",
    "plt.ylabel('Loss')\n",
    "plt.legend()\n",
    "\n",
    "plt.subplot(2, 1, 2)\n",
    "plt.plot(epochs_number, acc, 'r', label='Training acc')\n",
    "plt.plot(epochs_number, val_acc, 'b', label='Validation acc')\n",
    "plt.title('Training and validation accuracy')\n",
    "plt.xlabel('Epochs')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.legend(loc='lower right')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "id": "apart-christian",
   "metadata": {
    "executionInfo": {
     "elapsed": 16,
     "status": "ok",
     "timestamp": 1632980515836,
     "user": {
      "displayName": "Prabath Hetti Mudiyanselage",
      "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s64",
      "userId": "08605282177904379825"
     },
     "user_tz": -600
    },
    "id": "apart-christian"
   },
   "outputs": [],
   "source": [
    "# Show tensorboard\n",
    "\n",
    "# %load_ext tensorboard\n",
    "# %tensorboard --logdir logs/fit\n",
    "#%reload_ext tensorboard # just in case tensorboard didn't refresh"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "id": "5gMTwri775Ex",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 16,
     "status": "ok",
     "timestamp": 1632980515837,
     "user": {
      "displayName": "Prabath Hetti Mudiyanselage",
      "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s64",
      "userId": "08605282177904379825"
     },
     "user_tz": -600
    },
    "id": "5gMTwri775Ex",
    "outputId": "b3ebdc50-65e0-4585-bad9-ff01a6edad20"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'input_ids': <tf.Tensor: shape=(4000, 100), dtype=int32, numpy=\n",
       " array([[    0,   118,    56, ...,     1,     1,     1],\n",
       "        [    0, 14595,    52, ...,     1,     1,     1],\n",
       "        [    0,   463,    77, ...,     1,     1,     1],\n",
       "        ...,\n",
       "        [    0,  6025,  2528, ...,     1,     1,     1],\n",
       "        [    0,   102, 18969, ...,     1,     1,     1],\n",
       "        [    0,  7325,   220, ...,     1,     1,     1]])>}"
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_val_2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "id": "axoHnl6t-Nzp",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 11,
     "status": "ok",
     "timestamp": 1632980515837,
     "user": {
      "displayName": "Prabath Hetti Mudiyanselage",
      "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s64",
      "userId": "08605282177904379825"
     },
     "user_tz": -600
    },
    "id": "axoHnl6t-Nzp",
    "outputId": "dd39e825-3d00-4097-a044-41736341bc39"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'input_ids': <tf.Tensor: shape=(5000, 100), dtype=int32, numpy=\n",
       " array([[    0, 44290,    30, ...,     1,     1,     1],\n",
       "        [    0,  5317,   225, ...,     1,     1,     1],\n",
       "        [    0,   627, 31150, ...,     1,     1,     1],\n",
       "        ...,\n",
       "        [    0, 10339, 16963, ...,     1,     1,     1],\n",
       "        [    0,  8494, 16637, ...,     1,     1,     1],\n",
       "        [    0, 42065,   278, ...,     1,     1,     1]])>}"
      ]
     },
     "execution_count": 37,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "id": "jxu631dY7zqL",
   "metadata": {
    "executionInfo": {
     "elapsed": 9,
     "status": "ok",
     "timestamp": 1632980515838,
     "user": {
      "displayName": "Prabath Hetti Mudiyanselage",
      "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s64",
      "userId": "08605282177904379825"
     },
     "user_tz": -600
    },
    "id": "jxu631dY7zqL"
   },
   "outputs": [],
   "source": [
    "#model.predict([X_test])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "id": "c0aeef60",
   "metadata": {
    "executionInfo": {
     "elapsed": 8,
     "status": "ok",
     "timestamp": 1632980515838,
     "user": {
      "displayName": "Prabath Hetti Mudiyanselage",
      "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s64",
      "userId": "08605282177904379825"
     },
     "user_tz": -600
    },
    "id": "c0aeef60"
   },
   "outputs": [],
   "source": [
    "# Function to generate accuracy report\n",
    "\n",
    "from sklearn.metrics import confusion_matrix\n",
    "import matplotlib.pyplot as plt\n",
    "from sklearn.metrics import ConfusionMatrixDisplay\n",
    "from sklearn.metrics import accuracy_score, f1_score\n",
    "\n",
    "def accuracy_report(X_val_seq,y_val_seq):\n",
    "    # Calculate preditions for the validation dataset\n",
    "\n",
    "    y_pred = []  # store predicted labels\n",
    "    y_true = []  # store true labels\n",
    "\n",
    "    y_pred = model.predict([X_val_seq])\n",
    "    y_pred = np.argmax(y_pred, axis=1)\n",
    "    #len(y_pred)\n",
    "\n",
    "    y_true= y_val_seq\n",
    "    #len(y_true)\n",
    "\n",
    "    # Convert the true and predicted labels into arrays\n",
    "    true_categories = label_tokenizer.inverse_transform(y_true)\n",
    "    predicted_categories = label_tokenizer.inverse_transform(y_pred)\n",
    "    # print(predicted_categories)\n",
    "\n",
    "    # Generate confusion matrix\n",
    "\n",
    "    cnf_matrix = confusion_matrix(true_categories,predicted_categories,  normalize='pred')\n",
    "\n",
    "    # Plot confusion matrix\n",
    "    disp = ConfusionMatrixDisplay(confusion_matrix=cnf_matrix,display_labels=label_index)\n",
    "    disp = disp.plot(cmap='Greens', xticks_rotation=45)\n",
    "    plt.show()\n",
    "\n",
    "    print(\"Accuracy for validation split: \" ,accuracy_score(y_true, y_pred))\n",
    "    print( \"F1_score: \" + str(f1_score(y_true, y_pred, average='micro')))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "id": "environmental-transcription",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 395
    },
    "executionInfo": {
     "elapsed": 62071,
     "status": "ok",
     "timestamp": 1632980577901,
     "user": {
      "displayName": "Prabath Hetti Mudiyanselage",
      "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s64",
      "userId": "08605282177904379825"
     },
     "user_tz": -600
    },
    "id": "environmental-transcription",
    "outputId": "c20b3512-e99e-4536-dba1-ef2d51641e38"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy for validation split:  0.8948\n",
      "F1_score: 0.8948\n"
     ]
    }
   ],
   "source": [
    "# accuracy_report for same books\n",
    "\n",
    "accuracy_report(X_val,y_val_sequences)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "id": "fundamental-raising",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 395
    },
    "executionInfo": {
     "elapsed": 46091,
     "status": "ok",
     "timestamp": 1632980623988,
     "user": {
      "displayName": "Prabath Hetti Mudiyanselage",
      "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s64",
      "userId": "08605282177904379825"
     },
     "user_tz": -600
    },
    "id": "fundamental-raising",
    "outputId": "3244d914-230f-4982-da20-ac649d6670d2"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy for validation split:  0.674\n",
      "F1_score: 0.674\n"
     ]
    }
   ],
   "source": [
    "# accuracy_report for seperate books\n",
    "\n",
    "accuracy_report(X_val_2,y_val_sequences_2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "id": "CPLiRurgZed_",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 18,
     "status": "ok",
     "timestamp": 1632980623989,
     "user": {
      "displayName": "Prabath Hetti Mudiyanselage",
      "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s64",
      "userId": "08605282177904379825"
     },
     "user_tz": -600
    },
    "id": "CPLiRurgZed_",
    "outputId": "f3668867-dee7-4ea8-b120-ebda7398f048"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"nlp_roberta_18\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " input_ids (InputLayer)      [(None, 100)]             0         \n",
      "                                                                 \n",
      " roberta (TFRobertaMainLayer  TFBaseModelOutputWithPoo  124645632\n",
      " )                           ling(last_hidden_state=(            \n",
      "                             None, 100, 768),                    \n",
      "                              pooler_output=(None, 76            \n",
      "                             8),                                 \n",
      "                              hidden_states=None, att            \n",
      "                             entions=None)                       \n",
      "                                                                 \n",
      " pooled_output (Dropout)     (None, 768)               0         \n",
      "                                                                 \n",
      " dense (Dense)               (None, 5)                 3845      \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 124,649,477\n",
      "Trainable params: 124,649,477\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "id": "electoral-arrangement",
   "metadata": {
    "executionInfo": {
     "elapsed": 14,
     "status": "ok",
     "timestamp": 1632980623990,
     "user": {
      "displayName": "Prabath Hetti Mudiyanselage",
      "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s64",
      "userId": "08605282177904379825"
     },
     "user_tz": -600
    },
    "id": "electoral-arrangement"
   },
   "outputs": [],
   "source": [
    "# Saving final model for future use\n",
    "\n",
    "#model.save(\"bert_5.h5\", save_format='h5')\n",
    "#tensorflow.keras.models.save_model(model,model._name)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "id": "mk7Mp05XdxxU",
   "metadata": {
    "executionInfo": {
     "elapsed": 13,
     "status": "ok",
     "timestamp": 1632980623990,
     "user": {
      "displayName": "Prabath Hetti Mudiyanselage",
      "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s64",
      "userId": "08605282177904379825"
     },
     "user_tz": -600
    },
    "id": "mk7Mp05XdxxU"
   },
   "outputs": [],
   "source": [
    "#!tar czf bert_epoch5_adamw_compress.tar.gz bert_epoch5_adamw/"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "id": "VqWmI63N5AIy",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 11,
     "status": "ok",
     "timestamp": 1632980623990,
     "user": {
      "displayName": "Prabath Hetti Mudiyanselage",
      "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s64",
      "userId": "08605282177904379825"
     },
     "user_tz": -600
    },
    "id": "VqWmI63N5AIy",
    "outputId": "651e66f5-02b7-491d-eccc-dce27d0cfd1f"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fri Feb 25 09:07:57 2022       \n",
      "+-----------------------------------------------------------------------------+\n",
      "| NVIDIA-SMI 496.13       Driver Version: 496.13       CUDA Version: 11.5     |\n",
      "|-------------------------------+----------------------+----------------------+\n",
      "| GPU  Name            TCC/WDDM | Bus-Id        Disp.A | Volatile Uncorr. ECC |\n",
      "| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |\n",
      "|                               |                      |               MIG M. |\n",
      "|===============================+======================+======================|\n",
      "|   0  NVIDIA GeForce ... WDDM  | 00000000:0D:00.0  On |                  N/A |\n",
      "| 80%   67C    P2   134W / 170W |  12059MiB / 12288MiB |     26%      Default |\n",
      "|                               |                      |                  N/A |\n",
      "+-------------------------------+----------------------+----------------------+\n",
      "                                                                               \n",
      "+-----------------------------------------------------------------------------+\n",
      "| Processes:                                                                  |\n",
      "|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |\n",
      "|        ID   ID                                                   Usage      |\n",
      "|=============================================================================|\n",
      "|    0   N/A  N/A      1604    C+G   Insufficient Permissions        N/A      |\n",
      "|    0   N/A  N/A      2476      C   ...onda3\\envs\\tf2\\python.exe    N/A      |\n",
      "|    0   N/A  N/A      5204    C+G   C:\\Windows\\explorer.exe         N/A      |\n",
      "|    0   N/A  N/A      5748      C   ...onda3\\envs\\tf2\\python.exe    N/A      |\n",
      "|    0   N/A  N/A      6308    C+G   ...2txyewy\\TextInputHost.exe    N/A      |\n",
      "|    0   N/A  N/A      7460    C+G   ...y\\ShellExperienceHost.exe    N/A      |\n",
      "|    0   N/A  N/A      7488    C+G   ...5n1h2txyewy\\SearchApp.exe    N/A      |\n",
      "|    0   N/A  N/A      7616    C+G   Insufficient Permissions        N/A      |\n",
      "|    0   N/A  N/A      9452    C+G   ...zilla Firefox\\firefox.exe    N/A      |\n",
      "|    0   N/A  N/A     13036    C+G   ... Download Manager\\fdm.exe    N/A      |\n",
      "|    0   N/A  N/A     13108    C+G   ...zilla Firefox\\firefox.exe    N/A      |\n",
      "|    0   N/A  N/A     14216    C+G   ...5.0.3.0\\GoogleDriveFS.exe    N/A      |\n",
      "|    0   N/A  N/A     15260    C+G   ...enTabletDriver.UX.Wpf.exe    N/A      |\n",
      "|    0   N/A  N/A     15884    C+G   ...in7x64\\steamwebhelper.exe    N/A      |\n",
      "+-----------------------------------------------------------------------------+\n"
     ]
    }
   ],
   "source": [
    "!nvidia-smi"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "id": "nZzulGEJnDSz",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 33282,
     "status": "ok",
     "timestamp": 1632980657265,
     "user": {
      "displayName": "Prabath Hetti Mudiyanselage",
      "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s64",
      "userId": "08605282177904379825"
     },
     "user_tz": -600
    },
    "id": "nZzulGEJnDSz",
    "outputId": "df2d4ce3-2e5c-42cd-adfe-4b25d65dacd0"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:absl:Found untraced functions such as encoder_layer_call_fn, encoder_layer_call_and_return_conditional_losses, pooler_layer_call_fn, pooler_layer_call_and_return_conditional_losses, embeddings_layer_call_fn while saving (showing 5 of 422). These functions will not be directly callable after loading.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: nlp_roberta_18\\assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: nlp_roberta_18\\assets\n"
     ]
    }
   ],
   "source": [
    "model.save(model._name, include_optimizer=False)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "id": "JOr9fAJytC2a",
   "metadata": {
    "executionInfo": {
     "elapsed": 12,
     "status": "ok",
     "timestamp": 1632980657265,
     "user": {
      "displayName": "Prabath Hetti Mudiyanselage",
      "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s64",
      "userId": "08605282177904379825"
     },
     "user_tz": -600
    },
    "id": "JOr9fAJytC2a"
   },
   "outputs": [],
   "source": [
    "#!tar czf bert_no_optimizer.tar.gz bert_no_optimizer/"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "id": "X-QjGeq3qDFd",
   "metadata": {
    "executionInfo": {
     "elapsed": 10,
     "status": "ok",
     "timestamp": 1632980657265,
     "user": {
      "displayName": "Prabath Hetti Mudiyanselage",
      "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s64",
      "userId": "08605282177904379825"
     },
     "user_tz": -600
    },
    "id": "X-QjGeq3qDFd"
   },
   "outputs": [],
   "source": [
    "# Reloading model\n",
    "\n",
    "#reloaded_model = tensorflow.saved_model.load(\"bert_no_optimizer\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "id": "qYa60UyQ1QA1",
   "metadata": {
    "executionInfo": {
     "elapsed": 11,
     "status": "ok",
     "timestamp": 1632980657266,
     "user": {
      "displayName": "Prabath Hetti Mudiyanselage",
      "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s64",
      "userId": "08605282177904379825"
     },
     "user_tz": -600
    },
    "id": "qYa60UyQ1QA1"
   },
   "outputs": [],
   "source": [
    "# Comparing results of model in memory and reloaded model\n",
    "\n",
    "#text_string = X_val.iloc[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "id": "ET2pOYH82gR7",
   "metadata": {
    "executionInfo": {
     "elapsed": 11,
     "status": "ok",
     "timestamp": 1632980657266,
     "user": {
      "displayName": "Prabath Hetti Mudiyanselage",
      "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s64",
      "userId": "08605282177904379825"
     },
     "user_tz": -600
    },
    "id": "ET2pOYH82gR7"
   },
   "outputs": [],
   "source": [
    "# Correct result\n",
    "\n",
    "#y_val.iloc[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "id": "C_aCtz1YwjOW",
   "metadata": {
    "executionInfo": {
     "elapsed": 10,
     "status": "ok",
     "timestamp": 1632980657266,
     "user": {
      "displayName": "Prabath Hetti Mudiyanselage",
      "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s64",
      "userId": "08605282177904379825"
     },
     "user_tz": -600
    },
    "id": "C_aCtz1YwjOW"
   },
   "outputs": [],
   "source": [
    "#model_result = model(tensorflow.constant([text_string]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "id": "WS-j-7At1NfE",
   "metadata": {
    "executionInfo": {
     "elapsed": 11,
     "status": "ok",
     "timestamp": 1632980657267,
     "user": {
      "displayName": "Prabath Hetti Mudiyanselage",
      "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s64",
      "userId": "08605282177904379825"
     },
     "user_tz": -600
    },
    "id": "WS-j-7At1NfE"
   },
   "outputs": [],
   "source": [
    "#reloaded_model_result = reloaded_model(tensorflow.constant([text_string]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "id": "0FRxvPjP1ttd",
   "metadata": {
    "executionInfo": {
     "elapsed": 10,
     "status": "ok",
     "timestamp": 1632980657267,
     "user": {
      "displayName": "Prabath Hetti Mudiyanselage",
      "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s64",
      "userId": "08605282177904379825"
     },
     "user_tz": -600
    },
    "id": "0FRxvPjP1ttd"
   },
   "outputs": [],
   "source": [
    "# Result from model in memory\n",
    "\n",
    "#label_index[np.argmax(model_result, axis=1)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "id": "wpirJtoY1uA8",
   "metadata": {
    "executionInfo": {
     "elapsed": 10,
     "status": "ok",
     "timestamp": 1632980657267,
     "user": {
      "displayName": "Prabath Hetti Mudiyanselage",
      "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s64",
      "userId": "08605282177904379825"
     },
     "user_tz": -600
    },
    "id": "wpirJtoY1uA8"
   },
   "outputs": [],
   "source": [
    "# Result from reloaded model\n",
    "\n",
    "#label_index[np.argmax(reloaded_model_result, axis=1)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "54On8BXI-Fwx",
   "metadata": {
    "executionInfo": {
     "elapsed": 10,
     "status": "ok",
     "timestamp": 1632980657267,
     "user": {
      "displayName": "Prabath Hetti Mudiyanselage",
      "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s64",
      "userId": "08605282177904379825"
     },
     "user_tz": -600
    },
    "id": "54On8BXI-Fwx"
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "accelerator": "GPU",
  "colab": {
   "collapsed_sections": [],
   "name": "keras_hf_RoBERTa.ipynb",
   "provenance": []
  },
  "kernelspec": {
   "display_name": "Python 3.9 (Tensorflow 2.6)",
   "language": "python",
   "name": "tf2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.9.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}