DNN/DNN_TELUGU.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "DNN_TELUGU.ipynb",
      "provenance": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "J9DtPbakB5hO",
        "colab_type": "text"
      },
      "source": [
        "**DEEP NEURAL NETWORK FOR TELUGU DIGITS RECOGNITION**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "PxNfWeHQtGQs",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 122
        },
        "outputId": "5c9e56af-1237-49f5-e3b1-5186c0cce89a"
      },
      "source": [
        "from google.colab import drive\n",
        "drive.mount('/content/gdrive')"
      ],
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Go to this URL in a browser: https://accounts.google.com/o/oauth2/auth?client_id=947318989803-6bn6qk8qdgf4n4g3pfee6491hc0brc4i.apps.googleusercontent.com&redirect_uri=urn%3aietf%3awg%3aoauth%3a2.0%3aoob&response_type=code&scope=email%20https%3a%2f%2fwww.googleapis.com%2fauth%2fdocs.test%20https%3a%2f%2fwww.googleapis.com%2fauth%2fdrive%20https%3a%2f%2fwww.googleapis.com%2fauth%2fdrive.photos.readonly%20https%3a%2f%2fwww.googleapis.com%2fauth%2fpeopleapi.readonly\n",
            "\n",
            "Enter your authorization code:\n",
            "··········\n",
            "Mounted at /content/gdrive\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "CoOg7Hn_CfVe",
        "colab_type": "text"
      },
      "source": [
        "**IMPORTS AND LIBRARIES**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "2STM-JTKz6n5",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "import os\n",
        "import numpy as np\n",
        "import pandas as pd\n",
        "from matplotlib import pyplot as plt\n",
        "from tqdm import tqdm\n",
        "import glob\n",
        "import librosa\n",
        "import datetime as dt\n",
        "import torch\n",
        "import torch.nn as nn\n",
        "import torch.nn.functional as F\n",
        "import torch.optim as optim\n",
        "from sklearn.metrics import confusion_matrix"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "pkR6e9O4Cnh4",
        "colab_type": "text"
      },
      "source": [
        "**DATA PRE-PROCESSING**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "p_DFU5kkz6fq",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "bc711c33-cb69-42b7-841c-e51ff79b1053"
      },
      "source": [
        "os.chdir(\"/content/gdrive/My Drive/SRU Project/Final Submission/Telugu/telugu digits\")\n",
        "d = {'Rendu' : 2,'nalugu' : 4,'enimidi': 8, 'aidu' : 5, 'tommidi' : 9, 'padi' : 0, 'edu' : 7, 'Okati' : 1, 'Mudu' : 3,'aaru' : 6}\n",
        "\n",
        "data2 = []\n",
        "for i in d:\n",
        "  c = 1\n",
        "  for file in glob.glob(i + \"/\" + \"*.wav\"):\n",
        "    print(file)\n",
        "    print(c)\n",
        "    c =c + 1\n",
        "    y, sr = librosa.load(file)\n",
        "    mfccs = librosa.feature.mfcc(y = y)\n",
        "    data2.append([mfccs , int(d[i])])\n",
        "\n",
        "  print(\"change in digit\")\n",
        "\n",
        "\n",
        "data2 = np.array(data2)\n",
        "np.random.shuffle(data2)\n",
        "\n",
        "\n",
        "shp = []\n",
        "for i in range(len(data2)):\n",
        "  print(data2[i][0].shape)\n",
        "  shp.append(data2[i][0].shape[1])\n",
        "\n",
        "z = max(shp)\n",
        "\n",
        "np.save(\"mfcc_telegu\" , data2)"
      ],
      "execution_count": 3,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Rendu/2a.wav\n",
            "1\n",
            "Rendu/2e.wav\n",
            "2\n",
            "Rendu/2d.wav\n",
            "3\n",
            "Rendu/2c.wav\n",
            "4\n",
            "Rendu/2b.wav\n",
            "5\n",
            "Rendu/2f.wav\n",
            "6\n",
            "change in digit\n",
            "nalugu/4a.wav\n",
            "1\n",
            "nalugu/4f.wav\n",
            "2\n",
            "nalugu/4e.wav\n",
            "3\n",
            "nalugu/4d.wav\n",
            "4\n",
            "nalugu/4c.wav\n",
            "5\n",
            "nalugu/4b.wav\n",
            "6\n",
            "change in digit\n",
            "enimidi/8b.wav\n",
            "1\n",
            "enimidi/8a.wav\n",
            "2\n",
            "enimidi/8f.wav\n",
            "3\n",
            "enimidi/8e.wav\n",
            "4\n",
            "enimidi/8d.wav\n",
            "5\n",
            "enimidi/8c.wav\n",
            "6\n",
            "change in digit\n",
            "aidu/5a.wav\n",
            "1\n",
            "aidu/5b.wav\n",
            "2\n",
            "aidu/5g.wav\n",
            "3\n",
            "aidu/5f.wav\n",
            "4\n",
            "aidu/5e.wav\n",
            "5\n",
            "aidu/5d.wav\n",
            "6\n",
            "aidu/5c.wav\n",
            "7\n",
            "change in digit\n",
            "tommidi/9b.wav\n",
            "1\n",
            "tommidi/9a.wav\n",
            "2\n",
            "tommidi/9f.wav\n",
            "3\n",
            "tommidi/9e.wav\n",
            "4\n",
            "tommidi/9d.wav\n",
            "5\n",
            "tommidi/9c.wav\n",
            "6\n",
            "change in digit\n",
            "padi/10b.wav\n",
            "1\n",
            "padi/10a.wav\n",
            "2\n",
            "padi/10f.wav\n",
            "3\n",
            "padi/10e.wav\n",
            "4\n",
            "padi/10d.wav\n",
            "5\n",
            "padi/10c.wav\n",
            "6\n",
            "change in digit\n",
            "edu/7a.wav\n",
            "1\n",
            "edu/7b.wav\n",
            "2\n",
            "edu/7f.wav\n",
            "3\n",
            "edu/7e.wav\n",
            "4\n",
            "edu/7d.wav\n",
            "5\n",
            "edu/7c.wav\n",
            "6\n",
            "change in digit\n",
            "Okati/1a.wav\n",
            "1\n",
            "Okati/1b.wav\n",
            "2\n",
            "Okati/1f.wav\n",
            "3\n",
            "Okati/1e.wav\n",
            "4\n",
            "Okati/1d.wav\n",
            "5\n",
            "Okati/1c.wav\n",
            "6\n",
            "change in digit\n",
            "Mudu/3a.wav\n",
            "1\n",
            "Mudu/3f.wav\n",
            "2\n",
            "Mudu/3e.wav\n",
            "3\n",
            "Mudu/3d.wav\n",
            "4\n",
            "Mudu/3c.wav\n",
            "5\n",
            "Mudu/3b.wav\n",
            "6\n",
            "change in digit\n",
            "aaru/6b.wav\n",
            "1\n",
            "aaru/6a.wav\n",
            "2\n",
            "aaru/6f.wav\n",
            "3\n",
            "aaru/6e.wav\n",
            "4\n",
            "aaru/6d.wav\n",
            "5\n",
            "aaru/6c.wav\n",
            "6\n",
            "change in digit\n",
            "(20, 66)\n",
            "(20, 66)\n",
            "(20, 34)\n",
            "(20, 59)\n",
            "(20, 66)\n",
            "(20, 33)\n",
            "(20, 71)\n",
            "(20, 69)\n",
            "(20, 67)\n",
            "(20, 53)\n",
            "(20, 36)\n",
            "(20, 59)\n",
            "(20, 67)\n",
            "(20, 29)\n",
            "(20, 58)\n",
            "(20, 58)\n",
            "(20, 50)\n",
            "(20, 50)\n",
            "(20, 59)\n",
            "(20, 59)\n",
            "(20, 53)\n",
            "(20, 27)\n",
            "(20, 50)\n",
            "(20, 69)\n",
            "(20, 71)\n",
            "(20, 75)\n",
            "(20, 71)\n",
            "(20, 58)\n",
            "(20, 38)\n",
            "(20, 67)\n",
            "(20, 66)\n",
            "(20, 75)\n",
            "(20, 53)\n",
            "(20, 50)\n",
            "(20, 48)\n",
            "(20, 48)\n",
            "(20, 75)\n",
            "(20, 71)\n",
            "(20, 48)\n",
            "(20, 25)\n",
            "(20, 58)\n",
            "(20, 53)\n",
            "(20, 75)\n",
            "(20, 58)\n",
            "(20, 75)\n",
            "(20, 50)\n",
            "(20, 50)\n",
            "(20, 67)\n",
            "(20, 53)\n",
            "(20, 67)\n",
            "(20, 69)\n",
            "(20, 66)\n",
            "(20, 71)\n",
            "(20, 24)\n",
            "(20, 30)\n",
            "(20, 48)\n",
            "(20, 35)\n",
            "(20, 69)\n",
            "(20, 69)\n",
            "(20, 59)\n",
            "(20, 48)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "6panr8Vo2Rq2",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "x = []\n",
        "y = []\n",
        "for i in range(len(data2)):\n",
        "  val = data2[i][0].shape[1]\n",
        "  if(val != z):\n",
        "    zero = np.zeros((20 , z - val))\n",
        "    new_data = np.append(data2[i][0] , zero , axis = 1)\n",
        "    x.append(new_data)\n",
        "  else:\n",
        "    x.append(data2[i][0])\n",
        "  y.append(data2[i][1])\n",
        "\n",
        "x = np.array(x)\n",
        "y = np.array(y)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "UTED0eD527zP",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 85
        },
        "outputId": "29ba6c94-a2e8-493a-9baa-a2490047e022"
      },
      "source": [
        "train_x = x[:50]\n",
        "train_y = y[:50]\n",
        "test_x = x[45:]\n",
        "test_y = y[45:]\n",
        "\n",
        "print(\"Training Data Dimensions = \" + str(train_x.shape))\n",
        "print(\"Training Labels Dimensions = \" + str(train_y.shape))\n",
        "print(\"Test Data Dimensions = \" + str(test_x.shape))\n",
        "print(\"Test Labels Dimensions = \" + str(test_y.shape))"
      ],
      "execution_count": 5,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Training Data Dimensions = (50, 20, 75)\n",
            "Training Labels Dimensions = (50,)\n",
            "Test Data Dimensions = (16, 20, 75)\n",
            "Test Labels Dimensions = (16,)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ZlhMkLPWCyMO",
        "colab_type": "text"
      },
      "source": [
        "**LSTM MODEL**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "yoY48u9U49Jk",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "a2276c34-9bac-485b-d531-f3972076dbe4"
      },
      "source": [
        "class LSTM(nn.Module):\n",
        "\n",
        "\n",
        "  def __init__(self , inp, hidden, batch, out, layers):\n",
        "\n",
        "    super(LSTM, self).__init__()\n",
        "\n",
        "\n",
        "    self.inp = inp\n",
        "    self.hidden = hidden\n",
        "    self.batch = batch\n",
        "    self.out = out\n",
        "    self.layers = layers\n",
        "\n",
        "\n",
        "    self.rnn = nn.LSTM(self.inp, self.hidden, self.layers, dropout = 0.1)\n",
        "    self.fc = nn.Sequential(nn.Linear(self.hidden, self.out))\n",
        "\n",
        "\n",
        "  \n",
        "  def forward(self, x):\n",
        "\n",
        "    output , hide = self.rnn(x)\n",
        "    lstm_out = self.fc(output[-1])\n",
        "    prediction = F.softmax(lstm_out, dim = 1)\n",
        "\n",
        "    return prediction\n",
        "\n",
        "train_y_ohe = []\n",
        "for i  in range(len(train_y)):\n",
        "  k = np.eye(10)[int(train_y[i])]\n",
        "  train_y_ohe.append(k)\n",
        "\n",
        "test_y_ohe = []\n",
        "for i  in range(len(test_y)):\n",
        "  k = np.eye(10)[int(test_y[i])]\n",
        "  test_y_ohe.append(k)\n",
        "\n",
        "train_y_ohe = np.array(train_y_ohe)\n",
        "test_y_ohe = np.array(test_y_ohe)\n",
        "print(test_y_ohe.shape)\n",
        "print(train_y_ohe.shape)\n",
        "\n",
        "batch = 5\n",
        "epochs = 20\n",
        "loss = []\n",
        "\n",
        "\n",
        "lstm = LSTM(inp = 75, hidden = 10, batch = batch, out = 10, layers=2)\n",
        "#loss_function = nn.CrossEntropyLoss()\n",
        "loss_function = nn.MSELoss()\n",
        "optimizer = optim.Adam(lstm.parameters(), lr=0.01)\n",
        "\n",
        "for ep in range(epochs):\n",
        "\n",
        "\n",
        "  for i in range(0, len(train_x), batch):\n",
        "\n",
        "\n",
        "    batch_x = torch.Tensor(train_x[i : i + batch]).permute(1,0,2)\n",
        "    #batch_y = torch.Tensor(train_y[i : i + batch]).type(torch.LongTensor)\n",
        "    batch_y = torch.Tensor(train_y_ohe[i : i + batch])#.type(torch.LongTensor)\n",
        "\n",
        "\n",
        "    lstm.zero_grad()\n",
        "    pred = lstm(batch_x)\n",
        "    los = loss_function(pred , batch_y)\n",
        "    los.backward()\n",
        "    optimizer.step()\n",
        "\n",
        "\n",
        "  print('EPOCH = ' + str(ep))\n",
        "  print(los)\n",
        "  loss.append(los)\n",
        "\n",
        "\n",
        "correct = 0\n",
        "total = 0\n",
        "predict = []\n",
        "real = []\n",
        "\n",
        "\n",
        "with torch.no_grad():\n",
        "\n",
        "\n",
        "  #real_class = torch.Tensor(test_y).type(torch.LongTensor)\n",
        "  real_class = torch.Tensor(train_y_ohe)#.type(torch.LongTensor)\n",
        "  scores = lstm(torch.Tensor(train_x).permute(1,0,2))\n",
        "\n",
        "print(real_class)\n",
        "print(scores)\n",
        "\n",
        "\n",
        "array = []\n",
        "for i in range(len(scores)):\n",
        "  k = scores[i]\n",
        "  m = torch.max(k)\n",
        "  for j in range(len(k)):\n",
        "    if(k[j] == m):\n",
        "      array.append(j)\n",
        "      break\n",
        "\n",
        "for i in range(len(scores)):\n",
        "  if(array[i] == int(train_y[i])):\n",
        "    correct = correct + 1\n",
        "\n",
        "print(\"Accuracy = \" + str(correct*100/len(scores)))"
      ],
      "execution_count": 6,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(16, 10)\n",
            "(50, 10)\n",
            "EPOCH = 0\n",
            "tensor(0.0878, grad_fn=<MseLossBackward>)\n",
            "EPOCH = 1\n",
            "tensor(0.0826, grad_fn=<MseLossBackward>)\n",
            "EPOCH = 2\n",
            "tensor(0.0750, grad_fn=<MseLossBackward>)\n",
            "EPOCH = 3\n",
            "tensor(0.0624, grad_fn=<MseLossBackward>)\n",
            "EPOCH = 4\n",
            "tensor(0.0525, grad_fn=<MseLossBackward>)\n",
            "EPOCH = 5\n",
            "tensor(0.0495, grad_fn=<MseLossBackward>)\n",
            "EPOCH = 6\n",
            "tensor(0.0355, grad_fn=<MseLossBackward>)\n",
            "EPOCH = 7\n",
            "tensor(0.0268, grad_fn=<MseLossBackward>)\n",
            "EPOCH = 8\n",
            "tensor(0.0100, grad_fn=<MseLossBackward>)\n",
            "EPOCH = 9\n",
            "tensor(0.0084, grad_fn=<MseLossBackward>)\n",
            "EPOCH = 10\n",
            "tensor(0.0152, grad_fn=<MseLossBackward>)\n",
            "EPOCH = 11\n",
            "tensor(0.0063, grad_fn=<MseLossBackward>)\n",
            "EPOCH = 12\n",
            "tensor(0.0063, grad_fn=<MseLossBackward>)\n",
            "EPOCH = 13\n",
            "tensor(0.0046, grad_fn=<MseLossBackward>)\n",
            "EPOCH = 14\n",
            "tensor(0.0029, grad_fn=<MseLossBackward>)\n",
            "EPOCH = 15\n",
            "tensor(0.0039, grad_fn=<MseLossBackward>)\n",
            "EPOCH = 16\n",
            "tensor(0.0058, grad_fn=<MseLossBackward>)\n",
            "EPOCH = 17\n",
            "tensor(0.0019, grad_fn=<MseLossBackward>)\n",
            "EPOCH = 18\n",
            "tensor(0.0382, grad_fn=<MseLossBackward>)\n",
            "EPOCH = 19\n",
            "tensor(0.0145, grad_fn=<MseLossBackward>)\n",
            "tensor([[0., 0., 1., 0., 0., 0., 0., 0., 0., 0.],\n",
            "        [0., 0., 1., 0., 0., 0., 0., 0., 0., 0.],\n",
            "        [0., 0., 0., 0., 0., 0., 0., 0., 0., 1.],\n",
            "        [0., 0., 0., 0., 0., 0., 0., 0., 1., 0.],\n",
            "        [0., 0., 1., 0., 0., 0., 0., 0., 0., 0.],\n",
            "        [0., 0., 1., 0., 0., 0., 0., 0., 0., 0.],\n",
            "        [0., 0., 0., 0., 0., 0., 0., 1., 0., 0.],\n",
            "        [0., 1., 0., 0., 0., 0., 0., 0., 0., 0.],\n",
            "        [0., 0., 0., 0., 0., 0., 0., 0., 0., 1.],\n",
            "        [0., 0., 0., 0., 0., 0., 1., 0., 0., 0.],\n",
            "        [0., 0., 0., 0., 0., 0., 0., 1., 0., 0.],\n",
            "        [0., 0., 0., 0., 0., 0., 0., 0., 1., 0.],\n",
            "        [0., 0., 0., 0., 0., 0., 0., 0., 0., 1.],\n",
            "        [0., 0., 0., 0., 1., 0., 0., 0., 0., 0.],\n",
            "        [0., 0., 0., 0., 1., 0., 0., 0., 0., 0.],\n",
            "        [0., 0., 0., 0., 1., 0., 0., 0., 0., 0.],\n",
            "        [0., 0., 0., 0., 0., 1., 0., 0., 0., 0.],\n",
            "        [0., 0., 0., 0., 0., 1., 0., 0., 0., 0.],\n",
            "        [0., 0., 0., 0., 0., 0., 0., 0., 1., 0.],\n",
            "        [0., 0., 0., 0., 0., 0., 0., 0., 1., 0.],\n",
            "        [0., 0., 0., 0., 0., 0., 1., 0., 0., 0.],\n",
            "        [0., 0., 0., 0., 0., 0., 1., 0., 0., 0.],\n",
            "        [0., 0., 0., 0., 0., 1., 0., 0., 0., 0.],\n",
            "        [0., 1., 0., 0., 0., 0., 0., 0., 0., 0.],\n",
            "        [0., 0., 0., 0., 0., 0., 0., 1., 0., 0.],\n",
            "        [0., 0., 0., 1., 0., 0., 0., 0., 0., 0.],\n",
            "        [0., 0., 0., 0., 0., 0., 0., 1., 0., 0.],\n",
            "        [0., 0., 0., 0., 1., 0., 0., 0., 0., 0.],\n",
            "        [0., 0., 0., 1., 0., 0., 0., 0., 0., 0.],\n",
            "        [0., 0., 0., 0., 0., 0., 0., 0., 0., 1.],\n",
            "        [0., 0., 1., 0., 0., 0., 0., 0., 0., 0.],\n",
            "        [0., 0., 0., 1., 0., 0., 0., 0., 0., 0.],\n",
            "        [0., 0., 0., 0., 0., 0., 1., 0., 0., 0.],\n",
            "        [0., 0., 0., 0., 0., 1., 0., 0., 0., 0.],\n",
            "        [1., 0., 0., 0., 0., 0., 0., 0., 0., 0.],\n",
            "        [1., 0., 0., 0., 0., 0., 0., 0., 0., 0.],\n",
            "        [0., 0., 0., 1., 0., 0., 0., 0., 0., 0.],\n",
            "        [0., 0., 0., 0., 0., 0., 0., 1., 0., 0.],\n",
            "        [1., 0., 0., 0., 0., 0., 0., 0., 0., 0.],\n",
            "        [0., 0., 0., 0., 0., 1., 0., 0., 0., 0.],\n",
            "        [0., 0., 0., 0., 1., 0., 0., 0., 0., 0.],\n",
            "        [0., 0., 0., 0., 0., 0., 1., 0., 0., 0.],\n",
            "        [0., 0., 0., 1., 0., 0., 0., 0., 0., 0.],\n",
            "        [0., 0., 0., 0., 1., 0., 0., 0., 0., 0.],\n",
            "        [0., 0., 0., 1., 0., 0., 0., 0., 0., 0.],\n",
            "        [0., 0., 0., 0., 0., 1., 0., 0., 0., 0.],\n",
            "        [0., 0., 0., 0., 0., 1., 0., 0., 0., 0.],\n",
            "        [0., 0., 0., 0., 0., 0., 0., 0., 0., 1.],\n",
            "        [0., 0., 0., 0., 0., 0., 1., 0., 0., 0.],\n",
            "        [0., 0., 0., 0., 0., 0., 0., 0., 0., 1.]])\n",
            "tensor([[6.6425e-02, 4.2958e-02, 5.4672e-01, 1.5070e-03, 6.7244e-02, 1.7308e-02,\n",
            "         2.3933e-01, 2.0866e-03, 1.5370e-02, 1.0463e-03],\n",
            "        [7.1445e-02, 4.7440e-02, 4.8655e-01, 1.6837e-03, 6.3341e-02, 1.5650e-02,\n",
            "         2.9663e-01, 2.5520e-03, 1.3794e-02, 9.1418e-04],\n",
            "        [3.8375e-03, 5.4091e-03, 9.8500e-03, 2.2581e-01, 2.6192e-02, 1.2677e-01,\n",
            "         1.2562e-02, 1.8248e-02, 1.3128e-02, 5.5820e-01],\n",
            "        [1.6476e-01, 3.0723e-01, 1.0670e-01, 1.6011e-02, 9.1761e-02, 1.7478e-02,\n",
            "         1.3291e-02, 2.9144e-03, 2.7535e-01, 4.5027e-03],\n",
            "        [9.0092e-02, 2.1435e-02, 6.8393e-01, 1.1693e-03, 2.7492e-02, 2.6322e-02,\n",
            "         1.1872e-01, 3.8825e-03, 2.4728e-02, 2.2297e-03],\n",
            "        [3.9034e-02, 3.7329e-02, 3.3349e-01, 2.0777e-02, 2.1542e-01, 8.8972e-02,\n",
            "         1.7387e-01, 1.1302e-02, 4.0531e-02, 3.9272e-02],\n",
            "        [1.8591e-02, 9.7320e-04, 7.3435e-04, 7.6283e-02, 3.7971e-04, 2.1687e-03,\n",
            "         8.0579e-03, 8.7026e-01, 5.3439e-04, 2.2013e-02],\n",
            "        [1.2005e-01, 3.4951e-01, 5.2204e-02, 5.2765e-02, 4.0813e-02, 6.2194e-02,\n",
            "         7.0294e-03, 1.2597e-02, 2.9361e-01, 9.2176e-03],\n",
            "        [1.3993e-02, 2.9922e-03, 2.2419e-03, 3.1681e-01, 1.8029e-03, 3.0226e-02,\n",
            "         8.5395e-03, 3.4801e-01, 6.6125e-03, 2.6878e-01],\n",
            "        [6.5145e-03, 8.6147e-04, 9.1680e-03, 2.6360e-02, 1.1845e-02, 9.8988e-04,\n",
            "         9.2222e-01, 1.8912e-02, 1.9083e-04, 2.9359e-03],\n",
            "        [1.6751e-02, 7.8966e-04, 5.0339e-04, 6.0237e-02, 4.4245e-04, 8.4813e-04,\n",
            "         7.6682e-03, 9.0085e-01, 3.6807e-04, 1.1540e-02],\n",
            "        [5.2661e-02, 1.1115e-01, 4.9030e-02, 8.6237e-03, 2.1518e-02, 2.9570e-02,\n",
            "         1.6168e-03, 1.0643e-03, 7.1805e-01, 6.7099e-03],\n",
            "        [1.1451e-03, 4.6307e-04, 3.0674e-03, 5.1881e-02, 1.4357e-02, 1.7325e-02,\n",
            "         8.0556e-03, 2.1623e-02, 2.1242e-03, 8.7996e-01],\n",
            "        [1.1869e-02, 3.5970e-02, 7.5353e-02, 5.8435e-02, 5.4863e-01, 4.2361e-02,\n",
            "         1.3174e-01, 1.5976e-02, 2.7699e-02, 5.1971e-02],\n",
            "        [7.6991e-03, 1.8970e-02, 2.2387e-02, 4.0294e-02, 6.8938e-01, 1.7496e-03,\n",
            "         1.8704e-01, 6.6778e-03, 7.9516e-03, 1.7854e-02],\n",
            "        [2.3605e-02, 1.0014e-01, 2.2142e-02, 1.3675e-01, 5.4029e-01, 7.7314e-03,\n",
            "         9.1518e-02, 2.5904e-02, 3.6150e-02, 1.5769e-02],\n",
            "        [1.5201e-02, 4.0377e-02, 6.8492e-02, 7.6389e-03, 8.6149e-03, 4.1581e-01,\n",
            "         8.3113e-04, 2.9325e-03, 4.1490e-01, 2.5202e-02],\n",
            "        [1.0311e-02, 2.0415e-02, 2.1050e-01, 1.3157e-03, 1.8172e-02, 2.7058e-01,\n",
            "         1.4174e-03, 8.8762e-04, 4.4148e-01, 2.4912e-02],\n",
            "        [3.7707e-02, 9.1682e-02, 5.7307e-02, 6.5651e-03, 2.2155e-02, 3.6794e-02,\n",
            "         1.5006e-03, 6.6787e-04, 7.3896e-01, 6.6672e-03],\n",
            "        [1.9498e-02, 7.7296e-02, 1.0230e-01, 7.6176e-03, 3.7636e-02, 1.5327e-01,\n",
            "         2.2104e-03, 1.3763e-03, 5.7865e-01, 2.0142e-02],\n",
            "        [1.4372e-02, 1.9702e-03, 9.7139e-03, 6.5148e-02, 1.5617e-02, 1.6734e-03,\n",
            "         8.1585e-01, 6.9043e-02, 3.8402e-04, 6.2324e-03],\n",
            "        [1.0659e-02, 2.4381e-03, 2.4803e-02, 2.9380e-02, 2.4526e-02, 3.7433e-03,\n",
            "         8.6125e-01, 3.7055e-02, 7.9595e-04, 5.3487e-03],\n",
            "        [3.0347e-03, 4.4346e-03, 4.0969e-02, 4.0892e-03, 2.4746e-03, 8.5610e-01,\n",
            "         5.4770e-04, 1.6308e-03, 4.6798e-02, 3.9923e-02],\n",
            "        [1.5315e-01, 4.1973e-01, 5.9701e-02, 5.3940e-02, 1.0480e-01, 2.3585e-02,\n",
            "         2.4562e-02, 1.1426e-02, 1.4249e-01, 6.6208e-03],\n",
            "        [1.5550e-02, 8.1392e-04, 3.2310e-03, 2.8321e-02, 1.7214e-03, 2.3257e-03,\n",
            "         5.2141e-02, 8.7633e-01, 2.8861e-04, 1.9282e-02],\n",
            "        [6.6824e-03, 7.2193e-03, 1.0250e-03, 8.0394e-01, 4.9500e-03, 1.3627e-02,\n",
            "         2.2144e-02, 9.3335e-02, 2.8437e-03, 4.4236e-02],\n",
            "        [7.7636e-03, 3.4276e-04, 1.1264e-03, 1.6467e-02, 6.0649e-04, 1.5079e-03,\n",
            "         1.3425e-02, 9.4393e-01, 2.1090e-04, 1.4620e-02],\n",
            "        [2.2097e-03, 1.8124e-02, 1.9796e-02, 1.1569e-02, 8.8869e-01, 5.4526e-03,\n",
            "         3.0112e-02, 2.7246e-03, 8.0459e-03, 1.3273e-02],\n",
            "        [1.3651e-02, 1.4123e-02, 2.1538e-03, 8.0967e-01, 9.7565e-03, 1.4951e-02,\n",
            "         2.2101e-02, 4.6491e-02, 8.3435e-03, 5.8761e-02],\n",
            "        [7.2056e-04, 3.6900e-04, 6.9853e-04, 6.2429e-02, 3.2945e-03, 1.9538e-02,\n",
            "         8.2856e-04, 4.0357e-02, 2.1790e-03, 8.6959e-01],\n",
            "        [6.9506e-02, 2.3841e-02, 7.1097e-01, 8.1415e-04, 3.2271e-02, 2.1526e-02,\n",
            "         1.1189e-01, 2.1212e-03, 2.5730e-02, 1.3297e-03],\n",
            "        [1.5483e-02, 9.8544e-03, 3.9771e-03, 5.6825e-01, 1.4703e-02, 1.3673e-02,\n",
            "         1.2150e-01, 2.0520e-01, 2.3054e-03, 4.5055e-02],\n",
            "        [8.9921e-03, 1.6613e-03, 1.6412e-02, 9.4354e-03, 2.2884e-02, 5.8142e-04,\n",
            "         9.2097e-01, 1.6759e-02, 4.2578e-04, 1.8807e-03],\n",
            "        [2.5335e-03, 6.0967e-03, 4.0020e-02, 8.9171e-03, 5.4478e-03, 8.6759e-01,\n",
            "         1.4042e-03, 1.2791e-03, 2.4201e-02, 4.2507e-02],\n",
            "        [7.2094e-01, 9.1936e-03, 2.6383e-02, 2.6427e-02, 1.5052e-03, 1.7312e-03,\n",
            "         4.9502e-02, 1.4727e-01, 7.0499e-03, 1.0002e-02],\n",
            "        [4.9060e-01, 1.0555e-01, 7.5040e-02, 4.6224e-02, 3.5334e-02, 3.0192e-03,\n",
            "         1.8796e-01, 2.7267e-02, 2.5938e-02, 3.0580e-03],\n",
            "        [1.2080e-02, 1.2211e-02, 4.9014e-03, 6.7438e-01, 1.6732e-02, 3.4576e-02,\n",
            "         4.5892e-02, 6.7435e-02, 7.3207e-03, 1.2447e-01],\n",
            "        [1.8901e-02, 9.2615e-04, 7.7054e-04, 6.8989e-02, 4.3573e-04, 1.7450e-03,\n",
            "         9.6547e-03, 8.7946e-01, 4.6366e-04, 1.8656e-02],\n",
            "        [6.1631e-01, 8.4749e-02, 9.6940e-02, 2.2437e-02, 1.9242e-02, 4.0701e-03,\n",
            "         7.1475e-02, 3.0177e-02, 5.0853e-02, 3.7442e-03],\n",
            "        [2.5260e-03, 8.9533e-03, 5.6468e-02, 3.5414e-02, 4.0517e-02, 6.8888e-01,\n",
            "         7.3984e-03, 1.3113e-03, 3.2259e-02, 1.2627e-01],\n",
            "        [3.0729e-03, 2.8397e-02, 1.8042e-02, 8.9098e-03, 8.9594e-01, 4.1345e-03,\n",
            "         1.6531e-02, 1.8373e-03, 1.6288e-02, 6.8463e-03],\n",
            "        [1.0457e-02, 1.1602e-03, 1.2514e-02, 2.9295e-02, 1.1831e-02, 1.2586e-03,\n",
            "         9.0396e-01, 2.5755e-02, 3.0564e-04, 3.4634e-03],\n",
            "        [1.2146e-02, 6.0315e-03, 8.7397e-04, 6.0593e-01, 1.4656e-03, 2.0785e-02,\n",
            "         5.3908e-03, 2.4351e-01, 6.2794e-03, 9.7582e-02],\n",
            "        [1.1146e-02, 3.7066e-02, 5.2051e-02, 1.4035e-02, 6.7221e-01, 4.7696e-03,\n",
            "         1.8594e-01, 6.2362e-03, 1.2282e-02, 4.2705e-03],\n",
            "        [5.1627e-03, 8.1547e-03, 2.1757e-03, 7.3899e-01, 1.9544e-02, 2.6612e-02,\n",
            "         1.2967e-02, 3.0214e-02, 7.2708e-03, 1.4891e-01],\n",
            "        [8.4876e-03, 1.5501e-02, 1.7352e-01, 1.6443e-03, 1.3876e-02, 4.3621e-01,\n",
            "         1.0074e-03, 1.2392e-03, 3.1141e-01, 3.7105e-02],\n",
            "        [1.0316e-02, 1.0472e-02, 1.4328e-01, 4.1061e-03, 8.3672e-03, 7.1776e-01,\n",
            "         2.3677e-03, 3.6631e-03, 6.1905e-02, 3.7754e-02],\n",
            "        [7.5157e-04, 2.9131e-04, 1.7985e-03, 3.7152e-02, 7.0368e-03, 1.6838e-02,\n",
            "         3.3240e-03, 2.5332e-02, 1.6842e-03, 9.0579e-01],\n",
            "        [9.0821e-03, 2.1808e-03, 1.2696e-02, 8.9878e-02, 1.9137e-02, 3.5962e-03,\n",
            "         8.3385e-01, 2.2157e-02, 6.2248e-04, 6.8050e-03],\n",
            "        [3.2795e-03, 4.0597e-04, 6.6985e-03, 3.5240e-02, 1.9689e-02, 1.0633e-02,\n",
            "         1.7279e-02, 7.6956e-02, 1.5227e-03, 8.2829e-01]])\n",
            "Accuracy = 94.0\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "6ES2pf1UC5VC",
        "colab_type": "text"
      },
      "source": [
        "**TRAINING RESULTS**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "LH1xJmub5wnc",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 832
        },
        "outputId": "2c016443-b070-4135-9052-27e356c94e22"
      },
      "source": [
        "print(\"Total number of test samples = \" + str(len(scores)))\n",
        "print(\" \")\n",
        "print(\"Correct predictions = \" + str(correct))\n",
        "print(\" \")\n",
        "print(\"Accuracy = \" + str(((correct/len(scores))*100)) + \"%\") \n",
        "print(\" \")\n",
        "\n",
        "\n",
        "cfm = confusion_matrix(train_y, array)\n",
        "print(cfm)\n",
        "print(\" \")\n",
        "plt.imshow(cfm)\n",
        "plt.show()\n",
        "\n",
        "\n",
        "for i in range(len(loss)):\n",
        "  loss[i] = float(loss[i])\n",
        "\n",
        "\n",
        "plt.plot(loss)\n",
        "plt.xlabel('Epochs')\n",
        "plt.ylabel('Loss')\n",
        "plt.title('Training convergence plot')\n",
        "plt.show()"
      ],
      "execution_count": 7,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Total number of test samples = 50\n",
            " \n",
            "Correct predictions = 47\n",
            " \n",
            "Accuracy = 94.0%\n",
            " \n",
            "[[3 0 0 0 0 0 0 0 0 0]\n",
            " [0 2 0 0 0 0 0 0 0 0]\n",
            " [0 0 5 0 0 0 0 0 0 0]\n",
            " [0 0 0 6 0 0 0 0 0 0]\n",
            " [0 0 0 0 6 0 0 0 0 0]\n",
            " [0 0 0 0 0 6 0 0 1 0]\n",
            " [0 0 0 0 0 0 6 0 0 0]\n",
            " [0 0 0 0 0 0 0 5 0 0]\n",
            " [0 1 0 0 0 0 0 0 3 0]\n",
            " [0 0 0 0 0 0 0 1 0 5]]\n",
            " \n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAPUAAAD4CAYAAAA0L6C7AAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjEsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+j8jraAAAKGElEQVR4nO3d34udB53H8fdnZ9LUpIu2uDdNyjaguERRWwZJLXjRCuuPYG92oUIFvcnNqm0pSPXGf6C4eiFCqHpjsRexF1KKdVn1QtgGp2lRk1gIrTZpK0bothqwSfS7FzNCNunkPDk5T5+ZL+8XFDLnnD79UPLuc+bJmaepKiT18Q9TD5C0WEYtNWPUUjNGLTVj1FIzy2McdOm6nbV8ww0LP+72k2cWfkxpK/oLZzhbb+TNnhsl6uUbbuDGB+5b+HHfdf9TCz+mtBUdrv/e8DnffkvNGLXUjFFLzRi11IxRS80YtdTMoKiTfCzJc0lOJHlw7FGS5jcz6iRLwDeBjwN7gU8n2Tv2MEnzGXKm/hBwoqqer6qzwKPAXePOkjSvIVHvAk5e8PWp9cf+nyQHkqwmWf3rn/04pzSVhV0oq6qDVbVSVStL1+1c1GElXaEhUb8E3HTB17vXH5O0CQ2J+hfAu5PsSXINcDfww3FnSZrXzJ/SqqrzST4PPAksAd+pqqOjL5M0l0E/ellVTwBPjLxF0gL4iTKpGaOWmjFqqRmjlpoxaqmZUW48uP3kmVFuEvjHA7ct/JgA7zz4P6McV5qCZ2qpGaOWmjFqqRmjlpoxaqkZo5aaMWqpGaOWmjFqqRmjlpoxaqkZo5aaMWqpGaOWmjFqqRmjlpoxaqkZo5aaMWqpGaOWmjFqqZlR7iY6lrHu+rn/6KsLP+bj771+4ceUhvBMLTVj1FIzRi01Y9RSM0YtNWPUUjNGLTUzM+okNyX5aZJjSY4mufetGCZpPkM+fHIeeKCqjiT5R+DpJP9VVcdG3iZpDjPP1FX1SlUdWf/1n4DjwK6xh0mazxV9TDTJzcAtwOE3ee4AcADgWnYsYJqkeQy+UJbkOuAHwH1V9frFz1fVwapaqaqVbWxf5EZJV2BQ1Em2sRb0I1X12LiTJF2NIVe/A3wbOF5VXxt/kqSrMeRMfTvwGeCOJM+u//WJkXdJmtPMC2VV9XMgb8EWSQvgJ8qkZoxaasaopWaMWmpmS914cCxj3CTwyZefXfgxAf71xg+Oclz14ZlaasaopWaMWmrGqKVmjFpqxqilZoxaasaopWaMWmrGqKVmjFpqxqilZoxaasaopWaMWmrGqKVmjFpqxqilZoxaasaopWaMWmrGu4mOZKy7fnqXUs3imVpqxqilZoxaasaopWaMWmrGqKVmjFpqZnDUSZaSPJPk8TEHSbo6V3Kmvhc4PtYQSYsxKOoku4FPAg+PO0fS1Rp6pv468CXgbxu9IMmBJKtJVs/xxkLGSbpyM6NOsh/4Q1U9fbnXVdXBqlqpqpVtbF/YQElXZsiZ+nbgU0l+CzwK3JHke6OukjS3mVFX1ZerandV3QzcDfykqu4ZfZmkufjn1FIzV/Tz1FX1M+BnoyyRtBCeqaVmjFpqxqilZoxaasaopWa8m+gW411KR7Tv/eMc96lfjnPcDXimlpoxaqkZo5aaMWqpGaOWmjFqqRmjlpoxaqkZo5aaMWqpGaOWmjFqqRmjlpoxaqkZo5aaMWqpGaOWmjFqqRmjlpoxaqkZo5aa8W6iArxLKfCW3/VzLJ6ppWaMWmrGqKVmjFpqxqilZoxaasaopWYGRZ3kHUkOJflNkuNJbht7mKT5DP3wyTeAH1XVvyW5Btgx4iZJV2Fm1EneDnwE+CxAVZ0Fzo47S9K8hrz93gOcBr6b5JkkDyfZefGLkhxIsppk9RxvLHyopGGGRL0M3Ap8q6puAc4AD178oqo6WFUrVbWyje0LnilpqCFRnwJOVdXh9a8PsRa5pE1oZtRV9XvgZJL3rD90J3Bs1FWS5jb06vcXgEfWr3w/D3xuvEmSrsagqKvqWWBl5C2SFsBPlEnNGLXUjFFLzRi11IxRS814N1GNaqy7lO4/+urCj/n4e69f+DGn4JlaasaopWaMWmrGqKVmjFpqxqilZoxaasaopWaMWmrGqKVmjFpqxqilZoxaasaopWaMWmrGqKVmjFpqxqilZoxaasaopWa21o0H971/nOM+9ctxjqvRjHGTwBP/uW/hxwR41/1PjXLcjXimlpoxaqkZo5aaMWqpGaOWmjFqqRmjlpoZFHWS+5McTfLrJN9Pcu3YwyTNZ2bUSXYBXwRWqup9wBJw99jDJM1n6NvvZeBtSZaBHcDL402SdDVmRl1VLwEPAS8CrwCvVdWPL35dkgNJVpOsnuONxS+VNMiQt9/XA3cBe4AbgZ1J7rn4dVV1sKpWqmplG9sXv1TSIEPefn8UeKGqTlfVOeAx4MPjzpI0ryFRvwjsS7IjSYA7gePjzpI0ryHfUx8GDgFHgF+t/z0HR94laU6Dfp66qr4KfHXkLZIWwE+USc0YtdSMUUvNGLXUjFFLzWytu4l610/93Qh3lh3rrp/7j7668GM+9+/nN3zOM7XUjFFLzRi11IxRS80YtdSMUUvNGLXUjFFLzRi11IxRS80YtdSMUUvNGLXUjFFLzRi11IxRS80YtdSMUUvNGLXUjFFLzRi11EyqavEHTU4Dvxvw0ncCf1z4gPFspb1baStsrb2bYes/V9U/vdkTo0Q9VJLVqlqZbMAV2kp7t9JW2Fp7N/tW335LzRi11MzUUW+1/3n9Vtq7lbbC1tq7qbdO+j21pMWb+kwtacGMWmpmsqiTfCzJc0lOJHlwqh2zJLkpyU+THEtyNMm9U28aIslSkmeSPD71lstJ8o4kh5L8JsnxJLdNvelykty//vvg10m+n+TaqTddbJKokywB3wQ+DuwFPp1k7xRbBjgPPFBVe4F9wH9s4q0Xuhc4PvWIAb4B/Kiq/gX4AJt4c5JdwBeBlap6H7AE3D3tqktNdab+EHCiqp6vqrPAo8BdE225rKp6paqOrP/6T6z9pts17arLS7Ib+CTw8NRbLifJ24GPAN8GqKqzVfW/066aaRl4W5JlYAfw8sR7LjFV1LuAkxd8fYpNHgpAkpuBW4DD0y6Z6evAl4C/TT1khj3AaeC7698qPJxk59SjNlJVLwEPAS8CrwCvVdWPp111KS+UDZTkOuAHwH1V9frUezaSZD/wh6p6euotAywDtwLfqqpbgDPAZr6+cj1r7yj3ADcCO5PcM+2qS00V9UvATRd8vXv9sU0pyTbWgn6kqh6bes8MtwOfSvJb1r6tuSPJ96adtKFTwKmq+vs7n0OsRb5ZfRR4oapOV9U54DHgwxNvusRUUf8CeHeSPUmuYe1iww8n2nJZScLa93zHq+prU++Zpaq+XFW7q+pm1v69/qSqNt3ZBKCqfg+cTPKe9YfuBI5NOGmWF4F9SXas/764k014YW95in9oVZ1P8nngSdauIH6nqo5OsWWA24HPAL9K8uz6Y1+pqicm3NTJF4BH1v/j/jzwuYn3bKiqDic5BBxh7U9FnmETfmTUj4lKzXihTGrGqKVmjFpqxqilZoxaasaopWaMWmrm/wAqpC2CjajaFwAAAABJRU5ErkJggg==\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "k0bovHVVC9ut",
        "colab_type": "text"
      },
      "source": [
        "**MODEL ARCHITECTURE**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "NfE2JF9L52Jh",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 119
        },
        "outputId": "23c35cf6-6590-4b1a-ba26-62dd0eb47d8f"
      },
      "source": [
        "lstm"
      ],
      "execution_count": 8,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "LSTM(\n",
              "  (rnn): LSTM(75, 10, num_layers=2, dropout=0.1)\n",
              "  (fc): Sequential(\n",
              "    (0): Linear(in_features=10, out_features=10, bias=True)\n",
              "  )\n",
              ")"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 8
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "5v6p1gUQC_aj",
        "colab_type": "text"
      },
      "source": [
        "**TESTING RESULTS**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "9I4KWvzH56eo",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 520
        },
        "outputId": "4d3de928-2f6a-4b05-91df-f54848806d73"
      },
      "source": [
        "correct = 0\n",
        "total = 0\n",
        "predict = []\n",
        "real = []\n",
        "\n",
        "\n",
        "with torch.no_grad():\n",
        "\n",
        "\n",
        "  #real_class = torch.Tensor(test_y).type(torch.LongTensor)\n",
        "  real_class = torch.Tensor(test_y_ohe)#.type(torch.LongTensor)\n",
        "  scores = lstm(torch.Tensor(test_x).permute(1,0,2))\n",
        "\n",
        "\n",
        "\n",
        "array = []\n",
        "for i in range(len(scores)):\n",
        "  k = scores[i]\n",
        "  m = torch.max(k)\n",
        "  for j in range(len(k)):\n",
        "    if(k[j] == m):\n",
        "      array.append(j)\n",
        "      break\n",
        "\n",
        "for i in range(len(scores)):\n",
        "  if(array[i] == int(test_y[i])):\n",
        "    correct = correct + 1\n",
        "\n",
        "print(\"Total number of test samples = \" + str(len(scores)))\n",
        "print(\" \")\n",
        "print(\"Correct predictions = \" + str(correct))\n",
        "print(\" \")\n",
        "print(\"Accuracy = \" + str((correct)/len(scores)*100) + \"%\") \n",
        "print(\" \")\n",
        "\n",
        "\n",
        "cfm = confusion_matrix(test_y, array)\n",
        "print(cfm)\n",
        "print(\" \")\n",
        "plt.imshow(cfm)\n",
        "plt.show()\n"
      ],
      "execution_count": 9,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Total number of test samples = 16\n",
            " \n",
            "Correct predictions = 11\n",
            " \n",
            "Accuracy = 68.75%\n",
            " \n",
            "[[2 0 0 0 1 0 0 0]\n",
            " [0 2 1 0 0 0 1 0]\n",
            " [0 0 1 0 0 0 0 0]\n",
            " [0 0 0 1 0 0 1 0]\n",
            " [0 0 0 0 1 0 0 0]\n",
            " [0 0 0 0 0 1 0 0]\n",
            " [0 0 0 1 0 0 1 0]\n",
            " [0 0 0 0 0 0 0 2]]\n",
            " \n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Lzgo3HVJDF5w",
        "colab_type": "text"
      },
      "source": [
        "**SAVING MODEL WEIGHTS**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "8nNA5kJB9-0F",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 71
        },
        "outputId": "700843ca-6b5b-44ea-bdf7-5766cf8f7c0c"
      },
      "source": [
        "torch.save(lstm , \"telugu_lstm_weights.pt\")"
      ],
      "execution_count": 10,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/torch/serialization.py:402: UserWarning: Couldn't retrieve source code for container of type LSTM. It won't be checked for correctness upon loading.\n",
            "  \"type \" + obj.__name__ + \". It won't be checked \"\n"
          ],
          "name": "stderr"
        }
      ]
    }
  ]
}