Mnist.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Mnist.ipynb",
      "provenance": [],
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/NeuroFan/LowVoltageDNNonFPGA/blob/main/Mnist.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ZddkOzI6ANFI"
      },
      "source": [
        ""
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "b2RWAmeP-Zyj",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "5de443c7-bec8-4af3-f6fd-789e66108dc3"
      },
      "source": [
        "! [ ! -z \"$COLAB_GPU\" ] && pip install torch scikit-learn==0.20.* skorch"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Requirement already satisfied: torch in /usr/local/lib/python3.7/dist-packages (1.9.0+cu102)\n",
            "Collecting scikit-learn==0.20.*\n",
            "  Downloading scikit_learn-0.20.4-cp37-cp37m-manylinux1_x86_64.whl (5.4 MB)\n",
            "\u001b[K     |████████████████████████████████| 5.4 MB 4.6 MB/s \n",
            "\u001b[?25hCollecting skorch\n",
            "  Downloading skorch-0.10.0-py3-none-any.whl (128 kB)\n",
            "\u001b[K     |████████████████████████████████| 128 kB 20.1 MB/s \n",
            "\u001b[?25hRequirement already satisfied: numpy>=1.8.2 in /usr/local/lib/python3.7/dist-packages (from scikit-learn==0.20.*) (1.19.5)\n",
            "Requirement already satisfied: scipy>=0.13.3 in /usr/local/lib/python3.7/dist-packages (from scikit-learn==0.20.*) (1.4.1)\n",
            "Requirement already satisfied: typing-extensions in /usr/local/lib/python3.7/dist-packages (from torch) (3.7.4.3)\n",
            "Requirement already satisfied: tabulate>=0.7.7 in /usr/local/lib/python3.7/dist-packages (from skorch) (0.8.9)\n",
            "Requirement already satisfied: tqdm>=4.14.0 in /usr/local/lib/python3.7/dist-packages (from skorch) (4.62.0)\n",
            "Installing collected packages: scikit-learn, skorch\n",
            "  Attempting uninstall: scikit-learn\n",
            "    Found existing installation: scikit-learn 0.22.2.post1\n",
            "    Uninstalling scikit-learn-0.22.2.post1:\n",
            "      Successfully uninstalled scikit-learn-0.22.2.post1\n",
            "Successfully installed scikit-learn-0.20.4 skorch-0.10.0\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "2vkLGGLBAcVO"
      },
      "source": [
        "from sklearn.datasets import fetch_openml\n",
        "from sklearn.model_selection import train_test_split\n",
        "import numpy as np\n",
        "import matplotlib.pyplot as plt\n",
        "import torch\n",
        "from torch import nn\n",
        "import torch.nn.functional as F\n",
        "from keras.models import Sequential\n",
        "from keras.layers import Dense\n",
        "from keras.models import model_from_json\n",
        "import numpy\n",
        "import os"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "YT81zncmAhKW",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "69be2cca-abcc-4989-b7fc-0234b016cb49"
      },
      "source": [
        "mnist = fetch_openml('mnist_784', cache=False)\n",
        "mnist.data.shape\n",
        "X = mnist.data.astype('float32')\n",
        "y = mnist.target.astype('int64')\n",
        "X /= 255.0\n",
        "X.min(), X.max()\n",
        "#X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.25, random_state=42)\n",
        " \n",
        "#assert(X_train.shape[0] + X_test.shape[0] == mnist.data.shape[0])\n",
        "#X_m=X\n",
        "#y_m=y"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(0.0, 1.0)"
            ]
          },
          "metadata": {},
          "execution_count": 3
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "8UHUCxRmDKVL"
      },
      "source": [
        ""
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "CjZUxB49slG5"
      },
      "source": [
        ""
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "XwKLCRBa9khL"
      },
      "source": [
        "i=-1\n",
        "T=4\n",
        "xt=np.zeros((280000,784), dtype=float)\n",
        "yt=np.zeros((280000),dtype=int)\n",
        "yt9=np.zeros((280000),dtype=int)\n",
        "from PIL import Image\n",
        "for temp in X:\n",
        "  i+=1;\n",
        "  #if i==2:\n",
        "    #break\n",
        "  img=temp.reshape(28, 28)\n",
        "  #plt.imshow(img)\n",
        "  j=0;\n",
        "  for j in range(T):\n",
        "      img=np.rot90(img)\n",
        "      T1=img.reshape(1,784)\n",
        "      xt[T*i+j,:]=T1[0]\n",
        "      #print(T*i,j)\n",
        "      yt[T*i+j]=y[i]*T+j\n",
        "      \n",
        "  \n",
        " \n",
        "#plt.imshow(img)\n",
        "#xt[0,:]=img.reshape(28*28).shape\n",
        "#X[0,:].shape"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 282
        },
        "id": "PHhmFnNA3xo6",
        "outputId": "ee9525cd-6c37-48d6-d28c-024d0b52bb07"
      },
      "source": [
        "i=20582\n",
        "plt.imshow(xt[i,:].reshape(28,28))\n",
        "print(yt[i])\n",
        "\n"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "26\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "F6Yjde2oEgUA"
      },
      "source": [
        "X=xt\n",
        "y=yt\n",
        "#T=1\n",
        "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)\n",
        "assert(X_train.shape[0] + X_test.shape[0] == mnist.data.shape[0]*T)\n"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "bIPFG6fJDSas",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 142
        },
        "outputId": "26ee7006-2121-4ba3-f586-e6d3255a80fc"
      },
      "source": [
        "def plot_example(X, y):\n",
        "    \"\"\"Plot the first 5 images and their labels in a row.\"\"\"\n",
        "    for i, (img, y) in enumerate(zip(X[:5].reshape(5, 28, 28), y[:5])):\n",
        "        plt.subplot(151 + i)\n",
        "        plt.imshow(img)\n",
        "        plt.xticks([])\n",
        "        plt.yticks([])\n",
        "        plt.title(y)\n",
        "plot_example(X_train, y_train)\n",
        "print(X.shape[0])\n",
        "(X[:5].shape)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "280000\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(5, 784)"
            ]
          },
          "metadata": {},
          "execution_count": 7
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 5 Axes>"
            ]
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "bz6LDCNi5Ygf",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "bb092446-33b1-47e1-c087-86bbcf78f68f"
      },
      "source": [
        "device = 'cuda' if torch.cuda.is_available() else 'cpu'\n",
        "mnist_dim = X.shape[1]\n",
        "hidden_dim = int(mnist_dim/8)\n",
        "output_dim = T*len(np.unique(mnist.target))\n",
        "mnist_dim, hidden_dim, output_dim\n",
        "class ClassifierModule(nn.Module):\n",
        "    def __init__(\n",
        "            self,\n",
        "            input_dim1=mnist_dim,\n",
        "            hidden_dim=hidden_dim,\n",
        "            output_dim=output_dim,\n",
        "            dropout=0,\n",
        "    ):\n",
        "        super(ClassifierModule, self).__init__()\n",
        "        self.dropout = nn.Dropout(dropout)\n",
        "\n",
        "        self.hidden = nn.Linear(input_dim1, 8)\n",
        "        self.hidden1 =nn.Linear(8, 16)\n",
        "        self.hidden2 =nn.Linear(16, 12)\n",
        "        self.hidden3 =nn.Linear(12, 8)\n",
        "        self.output4 = nn.Linear(8, output_dim)\n",
        "\n",
        "    def forward(self, X, **kwargs):\n",
        "        X = F.relu(self.hidden(X))\n",
        "        X = F.relu(self.hidden1(X))\n",
        "        X = F.relu(self.hidden2(X))\n",
        "        X = F.relu(self.hidden3(X))\n",
        "        \n",
        "\n",
        "        X = self.dropout(X)\n",
        "        if test_flag==True:\n",
        "                  X+=torch.randn(X.shape)\n",
        "                  #X=torch.float64(X)\n",
        "        X = F.softmax(self.output4(X), dim=-1)\n",
        "        return X\n",
        "from skorch import NeuralNetClassifier\n",
        "\n",
        "torch.manual_seed(0)\n",
        "\n",
        "net = NeuralNetClassifier(\n",
        "    ClassifierModule,\n",
        "    max_epochs=150,\n",
        "    lr=0.15,\n",
        "    device=device,\n",
        ")\n",
        "test_flag=False\n",
        "net.fit(np.float32(X_train), y_train);\n",
        "\n",
        "from sklearn.metrics import accuracy_score\n",
        "test_flag=False  # with noise in test time  \n",
        "y_pred = net.predict(np.float32(X_test))\n",
        "print(accuracy_score(y_test, y_pred))\n",
        "error_mask = y_pred != y_test\n",
        "plot_example(X_test[error_mask], y_pred[error_mask])\n"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "  epoch    train_loss    valid_acc    valid_loss     dur\n",
            "-------  ------------  -----------  ------------  ------\n",
            "      1        \u001b[36m2.6092\u001b[0m       \u001b[32m0.5208\u001b[0m        \u001b[35m1.5617\u001b[0m  4.0473\n",
            "      2        \u001b[36m1.3383\u001b[0m       \u001b[32m0.6513\u001b[0m        \u001b[35m1.1402\u001b[0m  3.8278\n",
            "      3        \u001b[36m1.0946\u001b[0m       \u001b[32m0.6917\u001b[0m        \u001b[35m1.0069\u001b[0m  3.8002\n",
            "      4        \u001b[36m0.9925\u001b[0m       \u001b[32m0.7162\u001b[0m        \u001b[35m0.9518\u001b[0m  3.7145\n",
            "      5        \u001b[36m0.9438\u001b[0m       \u001b[32m0.7218\u001b[0m        \u001b[35m0.9344\u001b[0m  3.7091\n",
            "      6        \u001b[36m0.9127\u001b[0m       \u001b[32m0.7392\u001b[0m        \u001b[35m0.8827\u001b[0m  3.8651\n",
            "      7        \u001b[36m0.8870\u001b[0m       0.7323        0.8963  3.7727\n",
            "      8        \u001b[36m0.8689\u001b[0m       \u001b[32m0.7513\u001b[0m        \u001b[35m0.8419\u001b[0m  3.8352\n",
            "      9        \u001b[36m0.8519\u001b[0m       0.7461        0.8665  3.8580\n",
            "     10        \u001b[36m0.8403\u001b[0m       \u001b[32m0.7528\u001b[0m        0.8430  3.7412\n",
            "     11        \u001b[36m0.8334\u001b[0m       0.7450        0.8749  3.7834\n",
            "     12        \u001b[36m0.8235\u001b[0m       \u001b[32m0.7652\u001b[0m        \u001b[35m0.8074\u001b[0m  3.7141\n",
            "     13        \u001b[36m0.8149\u001b[0m       0.7571        0.8254  3.8487\n",
            "     14        \u001b[36m0.8106\u001b[0m       \u001b[32m0.7726\u001b[0m        \u001b[35m0.7669\u001b[0m  3.7129\n",
            "     15        \u001b[36m0.8008\u001b[0m       0.7655        0.7867  3.8239\n",
            "     16        \u001b[36m0.7951\u001b[0m       0.7567        0.8210  3.8823\n",
            "     17        \u001b[36m0.7897\u001b[0m       0.7543        0.8329  3.8820\n",
            "     18        \u001b[36m0.7846\u001b[0m       \u001b[32m0.7767\u001b[0m        \u001b[35m0.7483\u001b[0m  3.8429\n",
            "     19        \u001b[36m0.7794\u001b[0m       \u001b[32m0.7772\u001b[0m        \u001b[35m0.7424\u001b[0m  3.8390\n",
            "     20        \u001b[36m0.7736\u001b[0m       0.7729        0.7487  3.8274\n",
            "     21        \u001b[36m0.7717\u001b[0m       0.7751        \u001b[35m0.7418\u001b[0m  3.7999\n",
            "     22        \u001b[36m0.7665\u001b[0m       \u001b[32m0.7794\u001b[0m        \u001b[35m0.7305\u001b[0m  3.7488\n",
            "     23        \u001b[36m0.7630\u001b[0m       \u001b[32m0.7806\u001b[0m        \u001b[35m0.7268\u001b[0m  3.7968\n",
            "     24        \u001b[36m0.7595\u001b[0m       \u001b[32m0.7862\u001b[0m        \u001b[35m0.7048\u001b[0m  3.8869\n",
            "     25        \u001b[36m0.7579\u001b[0m       \u001b[32m0.7867\u001b[0m        0.7050  3.7400\n",
            "     26        \u001b[36m0.7550\u001b[0m       \u001b[32m0.7878\u001b[0m        \u001b[35m0.7013\u001b[0m  3.7647\n",
            "     27        0.7552       0.7875        \u001b[35m0.7005\u001b[0m  3.6717\n",
            "     28        \u001b[36m0.7484\u001b[0m       0.7862        0.7014  3.7260\n",
            "     29        0.7512       0.7843        0.7119  3.7620\n",
            "     30        0.7497       0.7856        0.7066  3.7317\n",
            "     31        \u001b[36m0.7470\u001b[0m       0.7824        0.7214  3.8484\n",
            "     32        0.7473       \u001b[32m0.7880\u001b[0m        \u001b[35m0.6968\u001b[0m  3.7377\n",
            "     33        0.7491       0.7866        0.6973  3.7330\n",
            "     34        \u001b[36m0.7404\u001b[0m       \u001b[32m0.7934\u001b[0m        \u001b[35m0.6802\u001b[0m  3.7292\n",
            "     35        \u001b[36m0.7398\u001b[0m       0.7890        0.6915  3.7002\n",
            "     36        \u001b[36m0.7386\u001b[0m       0.7850        0.7009  3.7215\n",
            "     37        \u001b[36m0.7369\u001b[0m       0.7793        0.7181  3.7975\n",
            "     38        0.7380       0.7845        0.7040  3.7411\n",
            "     39        \u001b[36m0.7343\u001b[0m       0.7772        0.7298  3.7753\n",
            "     40        \u001b[36m0.7314\u001b[0m       0.7843        0.7054  3.8171\n",
            "     41        \u001b[36m0.7295\u001b[0m       0.7833        0.7083  3.7572\n",
            "     42        \u001b[36m0.7283\u001b[0m       0.7830        0.7044  3.8895\n",
            "     43        \u001b[36m0.7269\u001b[0m       0.7772        0.7327  3.8045\n",
            "     44        0.7293       0.7804        0.7133  3.8357\n",
            "     45        0.7301       0.7879        0.7050  3.7258\n",
            "     46        \u001b[36m0.7225\u001b[0m       0.7859        0.7065  3.9147\n",
            "     47        0.7281       0.7870        0.6993  3.7734\n",
            "     48        0.7228       0.7809        0.7132  3.8120\n",
            "     49        0.7250       0.7843        0.7081  3.7954\n",
            "     50        0.7263       0.7858        0.7061  3.8087\n",
            "     51        0.7243       0.7874        0.6961  3.7049\n",
            "     52        \u001b[36m0.7201\u001b[0m       0.7859        0.7042  3.7206\n",
            "     53        0.7212       0.7887        0.7075  3.7400\n",
            "     54        \u001b[36m0.7195\u001b[0m       0.7923        \u001b[35m0.6799\u001b[0m  3.7640\n",
            "     55        0.7213       0.7783        0.7352  3.7305\n",
            "     56        \u001b[36m0.7187\u001b[0m       0.7819        0.7148  3.8041\n",
            "     57        \u001b[36m0.7168\u001b[0m       0.7844        0.6999  3.7097\n",
            "     58        0.7175       0.7812        0.7199  3.7362\n",
            "     59        \u001b[36m0.7135\u001b[0m       0.7853        0.7016  3.7867\n",
            "     60        0.7148       0.7790        0.7271  3.8046\n",
            "     61        0.7143       0.7850        0.7125  3.8111\n",
            "     62        \u001b[36m0.7094\u001b[0m       0.7850        0.7086  3.7522\n",
            "     63        0.7146       0.7774        0.7362  3.8204\n",
            "     64        0.7112       0.7759        0.7288  3.7594\n",
            "     65        0.7103       0.7873        0.6875  3.7380\n",
            "     66        \u001b[36m0.7092\u001b[0m       0.7824        0.7086  3.7741\n",
            "     67        0.7127       0.7875        0.6989  3.7144\n",
            "     68        \u001b[36m0.7090\u001b[0m       0.7868        0.6981  3.7282\n",
            "     69        0.7097       0.7830        0.7060  3.7456\n",
            "     70        \u001b[36m0.7071\u001b[0m       0.7839        0.7075  3.7535\n",
            "     71        0.7083       0.7916        0.6816  3.8468\n",
            "     72        0.7099       0.7731        0.7390  3.7327\n",
            "     73        0.7078       0.7764        0.7171  3.7265\n",
            "     74        0.7081       0.7717        0.7372  3.8091\n",
            "     75        0.7087       0.7697        0.7454  3.7554\n",
            "     76        \u001b[36m0.7063\u001b[0m       0.7793        0.7136  3.7978\n",
            "     77        \u001b[36m0.7033\u001b[0m       0.7626        0.8027  3.7255\n",
            "     78        0.7035       0.7726        0.7393  3.7917\n",
            "     79        \u001b[36m0.7018\u001b[0m       0.7690        0.7593  3.7705\n",
            "     80        0.7055       0.7752        0.7261  3.7539\n",
            "     81        \u001b[36m0.6993\u001b[0m       0.7851        0.7091  3.7350\n",
            "     82        0.7018       0.7832        0.7050  3.7665\n",
            "     83        0.6995       0.7724        0.7349  3.7939\n",
            "     84        0.7008       0.7825        0.7088  3.7104\n",
            "     85        \u001b[36m0.6961\u001b[0m       0.7846        0.7043  3.7532\n",
            "     86        0.6963       0.7794        0.7222  3.7540\n",
            "     87        0.6961       0.7797        0.7051  3.7842\n",
            "     88        \u001b[36m0.6951\u001b[0m       0.7808        0.7149  3.7480\n",
            "     89        \u001b[36m0.6926\u001b[0m       0.7814        0.7081  3.7927\n",
            "     90        0.6940       0.7796        0.7127  3.7516\n",
            "     91        \u001b[36m0.6904\u001b[0m       0.7797        0.7035  3.8763\n",
            "     92        \u001b[36m0.6896\u001b[0m       0.7883        0.6864  3.7744\n",
            "     93        \u001b[36m0.6890\u001b[0m       0.7748        0.7349  3.7375\n",
            "     94        \u001b[36m0.6872\u001b[0m       0.7767        0.7226  3.9407\n",
            "     95        0.6877       0.7785        0.7155  3.8055\n",
            "     96        \u001b[36m0.6864\u001b[0m       0.7742        0.7272  3.8862\n",
            "     97        \u001b[36m0.6860\u001b[0m       0.7772        0.7117  3.8381\n",
            "     98        0.6872       0.7799        0.7100  3.7576\n",
            "     99        \u001b[36m0.6822\u001b[0m       0.7782        0.7172  3.7134\n",
            "    100        \u001b[36m0.6813\u001b[0m       0.7813        0.7023  3.7231\n",
            "    101        \u001b[36m0.6781\u001b[0m       0.7786        0.7075  3.7690\n",
            "    102        \u001b[36m0.6771\u001b[0m       0.7737        0.7226  3.8717\n",
            "    103        \u001b[36m0.6753\u001b[0m       0.7812        0.7053  3.7468\n",
            "    104        \u001b[36m0.6731\u001b[0m       0.7779        0.7133  3.7187\n",
            "    105        \u001b[36m0.6702\u001b[0m       0.7772        0.7184  3.7345\n",
            "    106        \u001b[36m0.6684\u001b[0m       0.7846        0.6952  3.7300\n",
            "    107        0.6708       0.7825        0.7000  3.7729\n",
            "    108        \u001b[36m0.6681\u001b[0m       0.7803        0.6992  3.8352\n",
            "    109        0.6683       0.7738        0.7263  3.7925\n",
            "    110        \u001b[36m0.6680\u001b[0m       0.7764        0.7186  3.6989\n",
            "    111        \u001b[36m0.6644\u001b[0m       0.7820        0.7016  3.7422\n",
            "    112        \u001b[36m0.6624\u001b[0m       0.7829        0.6994  3.7292\n",
            "    113        \u001b[36m0.6620\u001b[0m       0.7711        0.7375  3.7438\n",
            "    114        \u001b[36m0.6578\u001b[0m       0.7884        \u001b[35m0.6769\u001b[0m  3.7292\n",
            "    115        0.6583       0.7861        0.6864  3.7388\n",
            "    116        0.6582       \u001b[32m0.7981\u001b[0m        \u001b[35m0.6616\u001b[0m  3.8362\n",
            "    117        \u001b[36m0.6560\u001b[0m       0.7980        \u001b[35m0.6559\u001b[0m  3.8217\n",
            "    118        \u001b[36m0.6537\u001b[0m       0.7875        0.6979  3.7913\n",
            "    119        \u001b[36m0.6523\u001b[0m       0.7954        0.6602  3.8328\n",
            "    120        \u001b[36m0.6513\u001b[0m       0.7913        0.6668  3.7636\n",
            "    121        0.6583       \u001b[32m0.8014\u001b[0m        \u001b[35m0.6431\u001b[0m  3.7429\n",
            "    122        0.6518       \u001b[32m0.8039\u001b[0m        \u001b[35m0.6387\u001b[0m  3.7840\n",
            "    123        \u001b[36m0.6485\u001b[0m       0.8021        0.6404  3.7090\n",
            "    124        0.6506       0.7983        0.6521  3.6681\n",
            "    125        \u001b[36m0.6483\u001b[0m       0.7995        0.6520  3.7071\n",
            "    126        0.6499       \u001b[32m0.8049\u001b[0m        \u001b[35m0.6329\u001b[0m  3.7087\n",
            "    127        \u001b[36m0.6432\u001b[0m       \u001b[32m0.8079\u001b[0m        \u001b[35m0.6246\u001b[0m  3.7768\n",
            "    128        0.6471       0.8070        0.6351  3.7109\n",
            "    129        \u001b[36m0.6430\u001b[0m       0.8064        0.6285  3.8015\n",
            "    130        \u001b[36m0.6410\u001b[0m       0.8078        0.6261  3.7392\n",
            "    131        0.6425       0.8068        0.6284  3.8424\n",
            "    132        \u001b[36m0.6388\u001b[0m       0.8041        0.6340  3.7208\n",
            "    133        \u001b[36m0.6372\u001b[0m       0.8070        0.6357  3.7793\n",
            "    134        \u001b[36m0.6345\u001b[0m       0.8038        0.6365  3.8399\n",
            "    135        \u001b[36m0.6327\u001b[0m       0.8028        0.6453  3.7293\n",
            "    136        0.6341       \u001b[32m0.8079\u001b[0m        0.6348  3.7090\n",
            "    137        \u001b[36m0.6311\u001b[0m       0.7965        0.6856  3.7477\n",
            "    138        0.6315       0.8006        0.6535  3.7417\n",
            "    139        0.6346       0.8044        0.6464  3.7085\n",
            "    140        \u001b[36m0.6298\u001b[0m       0.7814        0.7793  3.7548\n",
            "    141        0.6310       0.8041        0.6490  3.8325\n",
            "    142        \u001b[36m0.6280\u001b[0m       0.8074        0.6343  3.8417\n",
            "    143        \u001b[36m0.6266\u001b[0m       0.8041        0.6490  3.7276\n",
            "    144        0.6288       0.8054        0.6414  3.7416\n",
            "    145        0.6280       0.8031        0.6527  3.8324\n",
            "    146        0.6273       0.8064        0.6380  3.7795\n",
            "    147        \u001b[36m0.6242\u001b[0m       0.7906        0.7247  3.9216\n",
            "    148        0.6344       0.8012        0.6558  3.8306\n",
            "    149        0.6246       0.8037        0.6500  3.7930\n",
            "    150        0.6250       \u001b[32m0.8082\u001b[0m        0.6292  3.7163\n",
            "0.8076607142857143\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAWAAAABbCAYAAABNq1+WAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAgAElEQVR4nO29d5Rc13Wn+52bqupWrurqnNHIAAGYBDPFIIpKFpWsQMv2OMiyPdaMHGbG4/ccNLae/TSeYL83kix7LMuWR5IleiRKlqxAEhSDGEEARGgADTQ658q5bt175o9qgAAFkIDY6GoA91urF9aqul296+Dc391nn733EVJKXFxcXFxWH6XZBri4uLhcq7gC7OLi4tIkXAF2cXFxaRKuALu4uLg0CVeAXVxcXJqEK8AuLi4uTcIVYBcXF5cmsSYEWAjhEUL8jRBiXAiRF0LsF0K89az3PyyEOCGEKAghviOE6GymvauFEOKjQogXhBBVIcTnL3DNHwghpBDi3lU2b9W5iHnyfiHE8PJ7R4QQ72qmvavFheaJEMIQQjwohBhbniN3Nc/K1eXV7h0hhCmE+LQQYkkIkRVCPN4kM9eGAAMaMAncCYSB3wO+IoToX540fwK8E4gBp4AvNcnO1WYG+ATwufO9KYRYB7wPmF1No5rIq82TLuAfgN8CQsC/B74ohGhtlrGryKvNkyeBnwHmVtWi5vNqY/JXNLRk8/K/v7mKdp2D1qw/fDZSyiLw8bNe+mchxCngeuAW4KtSysMAQog/BqaFEOuklCdX3dhVREr5vwGEEDcA3ee55FPA7wCfXk27msVrzJMpICOl/Jfl974lhCgC64CFVTV0lbnQPJFS1oA/X37Pbo51zeFCYyKE2ATcD3RLKXPLL+9dfQsbrBUP+ByEEG3ABuDw6ZfOfnv5322ratQaQwjxPqAqpfx2s21pFq+YJy8Aw0KI+4UQ6nL4oQq81EwbXdYcNwLjwH9aDkEcFEK8t1nGrAkP+GyEEDrwv4C/k1IeFUJ8B/iyEOIvgRHgDwAJmE00s6kIIYI0wjJvarYtzeKV82T5tb8Hvgh4gRrwvmWv2cXlNN00nLd/AjpprLC/JYQ4IqUcXm1j1pQHLIRQgC/QuHk+CiClfBj4QxoDNrb8k6ex5LxW+TjwBSnlWJPtaArnmyfLm5D/GbgLMGjEif+nEGJnk8x0WZuUAQv4hJSyJqX8AbAHuK8ZxqwZARZCCOBvgDbgvVJK6/R7UspPSSnXSynbaAixBhxqjqVrgjcC/1YIMSeEmAN6aGxG/U6T7brsvMo82Qk8LqV8QUrpSCmfB54FrvrsEJdL4nwhqaa1hFwzAgx8hsau5DuklOXTLwohvEKIbaJBL40dzL+QUqabZehqIYTQhBBeQAXU5bHQaAjwNhqis5PGju+v0NiUu9o57zwBngfuOO3xCiF2AXdwDcSAX2WenE7d8y5faiy/Jy74YVcJrzImjwMTwO8uX3MbcDfw3aYYKqVs+g/QR+MpVAEKZ/18CIjQuImKNFJp/hRQm23zKo3Lx5fH5eyfj5/nujHg3mbb28x5svz+R4ETNEJUo8BvN9vmZs+T5bnxyvf6m21zk8dkK/D0sqYcAd7dLDvFskEuLi4uLqvMWgpBuLi4uFxTuALs4uLi0iRcAXZxcXFpEq4Au7i4uDQJV4BdXFxcmsQllSIbwiO9+C+XLWuCCkVqsnrReZLXwpgA5EkvSSkTF3OtOybn51oYF/f+OT8XmiuXJMBe/Nwk3rhyVq1BnpWPXNL118KYADwsHxy/2GvdMTk/18K4uPfP+bnQXHFDEC4uLi5NwhVgFxcXlyax5tpRuri4rBBCIAwDoaqgqpzTAkJv3PpOoYis1cCtiG0KrgC7uFxtKCpqKIAIh8hd30klrJDvF9i+hshKRUJnBU2zaf1fPgKPjzSE2Ko12fBrD1eAXdYWQvyIxyalRFp1kI7rqb0GQtMQHg8iGsFuCZHtV6m0SLxb0sTMRvM4XbV5V8d+4mqBT/7gAYIBP6JWuzoE+Oz5o2mgnBVltZdPZXKcRsMe20Y6EpzmndbkCrDLmkFtiUM8SnF9jPRGDduAul9izgoS+4poySL2yKmm3jBrFiFQW1qo7Ooj362TfVOJRDTPPYlhWvQCm70z+JUqAAoO6/Q0NoLsFhsj30P8eR8MjzT5S7wOloVX7e6kuLWNXK9G5pYquqdOwKxQtXSKC36UkoJ3UUEvQvRYDSNdRR2fR2ZzODVr1eeWK8AuawYR8FNLBMgOahR3lfGZNbpDeUZPtuFb8uFXFdQxDVlzPeFzEAKh6RAOkO3XyQ3Bf73hq/yEZ4GE6kFDpU5DWJTlfXddBCg4FTytJfI9QSLHfc38Bq8boaoIw8COB8kO6GQ32/zJTV9j0Fhgo15n3nb4fOoWxkpx9k71UMp4UCs6pqkSTAcQto3IF1Z9brkC7LJmKG5pY/ZWDXVTjn+z+UmCSoWIWuKlRA+P9G3kxEScjq5d+BYtvIenkKUydj5/bYuxEGj9vSRv6yDXpzB03yhvDs2wyVikJAWfSm9krBJnz+R6qlWNnpYMbb48/6HzOwzpIIREqiCF4Irs0r7s+Tq7tzJ7m59KQmIMZQlrNv//qbsp13QyaT8UdAInVdQaeFXwqFBqh0KvwswdrQinlfanJYHxIsqpGexkalXMdwXYZc1Q6NAwtmZ5R/8hfi3y8nL4Ht8cH4o8x2cTd/DP8zfhnzZonwo2BKNQBHmNhiSEAKFgt4RIXifwDGX57MCDtKomBalwylJ4LLmB0WQcngvjL8DYoMmpeI2TiQRD+gKKIrGvSOVdRigITaPY48O+OUdvJMdN8TGO5Do4cKwXLaURPw7ejEPw8AJISaU3QjWmM38TOIkatw6NEtHL7Eldj5B+QksBSKVX5cG+NgVYUVH8JkooSPG6LqQm8CxVUWp1RN0BW6JkC8haDSeTRVarzbbYZQUodQh+bcNT7PCdWzSkC4WYYvO2yAFSb/JzaKmDU22t+KcTtD5Ywc5km2Rxc9H6e8nc0E56k8rtbzhIhyfLQ4WNHC138NALu9DSGuGTEMpLApMlABzVpFzzsFgPAgt4dIucTyI15Yr0gLWeTmr9LaQ3KNzSPcbzcz089MjtGDnomaqjFWt4lsqIcg0WkgD4ahZenwcjG6YaNfjhbRvRWip4b0yTvRGSW7oJTHYReymHcnISp1y5bBuUa1KAhSJQ/CZOPERyu07dC6ExFb3koNQlog6+OQ0lX0EpV7BdAb4qqMYdfjUyCoBz1uu6UNGFyu3eCrf3PMoLrSp/4H0XoyPttP2LH65RAbbaIyxer+DdlOaTXd9hzlb5L7P38cx4PwNfdfBOL+GMTpxxUBS/n2DHNhxdJWubKCj49DoZr0RqV6L8gh0LkVnnpdxd557oMD8YHWLDg3OQzjbCCFKeM5cASDeOkzSGwRcKUQ1tJT9o8sG3PcMHwnv5sP9DjJ1qxSgECM+ZiHr9GhFgRUXxeVFaW0je2kGxQ8F/1wKt/gJT2TAlW8VxBLatUE2ZqMUAib0tBMcr6Cdnqc/NN/sbuKwC7WqJ+9qG2aPYTH6gn8BML9GnpnBSaZxS6ZqJCUtDwQrZyJKHj4y+h9F0DPlklOiCxDs2B9lCI31vGaGqFFtVyh0OLVoeB4f5VAhzWkHL15p3NPDrwDF1KnEBhsOpagI7ryPyRZxS+aLmgVOtEj9Ywj/v4QulN/I/E3dBsI4asEhuM6nEBmh5IQf7Dl8W+9eUAAtVRQkGsDoiLNwIZk+Wv9ryD2zW9XOuc3DYV1U4abXye773UomYtBdi4ArwNUGLqvLu0H6GPPP89/s8TE/GCY5G0SwLWa0i6/XX/pCrAFtXUIIW9YrG4WcHCUwI2v/uIE4+z3mj4rpGJQF0VEhoOWwkdtpDYNpByZXP/ztrnLpPpRqVKIbNeDmOUlRxMlmcSuWifl9WqyjPH8ZUVQZ+GEIE/Zz4xQ682zJUtjik16t4UwH8+y6P/WtKgJWAn9pQB9khHwNbp7kpPkZMqQM6JVnDkZJFR2JJhaACm4xZdl93ksNt7cyqURKRXRiji9Qnp5r9VV4ToRsIY/nB8opkcadSdYsOXgUdlaCwWW8s8I6ug+wN9HLwbRsxZwK0PRmGxTROOn3VC7F3MktsTwLhgFFw8CatRlnx+VguS650WuzunaRLy+BIiZZTMOdqiNLFCdZaQ6066HmBZSm0eXI4XgdhGHAJOb3StsGRyFIJpEP8UDu5QhTl+jy3rTvJgZ5thDvacQpFnHx+Re1fUwIsAn4y631kNsL/WPdP7DIUwIeDQ8mxyUvBkVobFUfndt8kG1SDz/V/i3xvnTeJX2EmEKbTbkG5EgTY60EJ+Bvie5YAy0oVxXGQtnN1VCZdBlQhiKkeYipsjB1hJrSPP3v7vTw/30t+KUpAURCl0lUvwPaxE8SOnTjntQs9soWmg89LV1+SD7c/Tp9mYQOetMB7agknm7vs9l4O1LKFJ+OlZCl0GynwOuDxIKpVZPUifXopQdo4xSKUSkQfO0UkGmLsBg9/2Plt7uvfgt2dQJ1Vr24BBji9FasuT6WqtJixbf5o+icZy8WYno2CpXDfzkPcFhrheu8kCVXwhp6TPCEGSaXCtM/0I9NZ7OVg+5rgdInkrs3kB/yUWhSqMRo5mOrLt41WEniXZCNt5ngWJV+mPj51TVZ/zdtVRqwwi/UQJ6pt9BpLvMM/gUdoqGc1lvErgpuDJ/CrVR68bzfe62K0Px3AO12A8ekVv2muNIRuoAz1UeqL0B8aoUvLkXIgZXtRqoBVf7lM9wpDyZYITplk16vEtAKJtiyZe9YRmKqgPHvo0h/CUiJLZYSqUl6M8lD+OqQmyQ75idTqMDW9ovavPQE+CweHvFPnYLWTvd/ZQmTEYfPeRUS5yqMf3cmLW3v4jfWP8GZzgj9qfwSr7WFuTf0G/vlWAseNM7udawGh6QhDZ+quIOE3zvG+zsM8EN6LXxGEFePMdS9UVb6UvIU9E0OUvxslOB3ENzuPU7kyb5DXw4gV5ivJGxlOtzM51kJbT5o7tn6emGqjnjV1w4rBB4KzvCcwxW+97QmOWSE+HPgIkeEobcXyNS/Ais9L5ro4uX6FD0RG2KAbPFXROVjpQS9IZKXSWIZficwvEbDqeHZ206uleHv3Yb7wzhvJHjTp3Wf8WKsgO5dDlEr4x/r5QueNSN0huV3Dkw3gObCy5q9pAa7IOgdqcZ4prMM/IwlOVCCbR9YsokcgX2zhC+bNWJ0qP+GdoE+T9PcuMn1jJx12BN9JT6PhRrOXokIgtqyj3BWgEpcEpOAHi+vZn+vGr9UI62VMpUZUL7JQCzFTDmF6LJI7HCotOt2L61HTReT0HLJWa/73WSXyjo/pUoSp+SiRgzqpZILfDryb9cFF3h3eS0yt0K3qZ7xhXagEFejU8iS2LjIXiqFanQT742i5KkqpBnNLa2tldLlRVITfJDuoUBq06DGSVKXFFxbv5Pm5HiJLNrJau2I9YFmrQaFIYELyb4c/iN+osaVrjoOlbmq3bMZYKsGJiUvenJWOxJuUJKdDYDjUu6tUIwaeFbZ/TQtw1rH5RnoXT04P0Lk3hzxwFHt5KR774l5a/D5Oia38xU8E+c2ND7M5OM8n1n2Ng109/Hn9fgafCyNLZWSzPSChMHd7lOzuKpq3SL7sZemlVsInwDYEdR/U/VCLOjimjTdaoTOa5VfveYjvp7ZwyN5EYNpP/Ac1ZDZ3zZTfzllhxlIxfMNe2v/uIKK7nfHxIQ4NDFF/k8J2/xTv9I9hipezZHSh0qepfHvbF1jcIvlXPT/HyYkY5lgYb1LS+owKmcw1MX4IgdA1ZDSEdlOaXxrcx3ZjibwDjz2zjZa9gsDhucZ8ukJxKhWoVEj8YJriQhtTd2l8+f1/wb9Er+Nz5TsJjEXpyRQaIclLuW+kQ3i0hlQMMndYvGPLQR49cCOhFbZ/TQuwAviUGrpqNzaq5Msp1dKq4RQhdEqSMaJ8PbqLLv1RVCSbPLNUuy2KN/Zjjufg0EjTsgoUvx9hmtTCEI4WycyEUBY0wpOS4KSFoys4hsDyCWohge3VqQV1Rjt87DE3kan6KA5ZVGMa0I03aWM+P4qTzTc26YRAbU00dn41tVGeeiGsesODLq3dZfnL9ViNjUkpQciGp6Okc4TGIyA1vnl8O8Pt7dyzbpTAeQ52CSgeoMpNreMc1C3GgzFKWQMjHyFqDcFCctXq/ZuF4vHApkHyg0E2tpxgs3eGoqOQcrx4kgr++RqiWG62mSuCLBTxzRYJnQzzsWMfRAJmV4G8x2T2rd3opW70ooNak+j55YraV36GriAVgVQFUkAtrKFWQdUc1nkX+Z4Jimmu6Cp0TQuwIQT93iRtgXbq3lZUjwenWj0jpNKqEf/GEVoCAY7YQ/zxzX5+vveHPBCc5oFdz/FP/p34H4vSdlxH1uurv3QXAqUtQb0lSKmnzt2dp3jy4evp+PxBsCykvTwJlIboCFU983v2jiGezW5G9pb5T3d8DYDv3bGNvdM99CY7UUdncDJZhKZR3dZDJa5j+QRSvbA5npyDkanjnczCsTUowAJU0RBTZ7m/w9nPk/r8Ap5H0rS3thCc6mJyWz+Tv2rSoZ7/wRpQPHyi7UnsNkl1o0PekdwX/DdU4gnanvXC1S7A8Rjjb45QHKjzf7c/ye3eLE9WorxQGiA06uB7aRLnKqkitJMpSGVoGw/BEy3M3p3gox99iB4jiXqPJOd4OVFpZ6oa5ZHRDVgVjVfWXvv8VTx6HZ9hoQpJ6qkOoscdfL4ad/qP8eext6Ik4shcYcXCWCsiwIppgqoiK9VGMP/H9TYtC0/WQS+qWFJFwSKmFQgZFZKaaHjBQjmn+YpTLCNsB/8UjJ9o5WC8mweC02zyzbC9s5VD7RtQOtuRuTz2UnIlvu7FIxSsjgjFLi8oNmPFGHpBXtD7PHvEtMU8oVMmBcvHlztupM2Xp99Mkmv1MrNjAH/bOtSyjdQUlq7TqUYkUlvOqHBEw2tUJfIs51ArKmhFD/6uBOFEAD1dhpkFZLXaqCBbI9jLK51OPc1gPMlwNIQSCiHzeZxKBZkv4J0tEYgE+dOJt3NH/AQ/F95Hi/qjLRUDivfMZwZEnY1d8wxv6cab8hNLrsFsmZVAUVEDfpyWMMX+Om29KTq1LDaS72e38sxiP96M3QjPXS37CadTyUollIUkockonx25g9ZAgW5/Bk2xcaRC1vJiGHWEkBhGHVVIvIaFIiTpgkmh6CWbM5F1hVAajJxNzlGIKHWkIZGGDtrK+a2v/5MUFdHZhvR7UedTOPnCj12N5OTyhF5apO5tJed48QqLTcYsxwPtPOEZRDd0qJzb90FaNWTdov170yReDPHP8W38QdtTvM0/zlv84/zkDT/P0okOwqMRxHJt+GohdI25G0xyWy1wBEcO99K7cHHj4pyaoPWLi7THoxSf6WLvJp0bfumb3NI1wtc/nCdrNYTFr9X4dy3P06+lKUoNS6pkHJOKY9CjJ4krL49XRapUpMo3crv47vRmUvsT9H/Lh7aYh5HR5sdFZUMoneVH0Z2+JLsHHuSnyj+Hva4DbcGPc2ocO59HHBohOhWmPN/LlzcNMPSxee73/6iQWtJGQeAgUYXgr9Z9heyAyntbPkK5pZPEixGUJ68uAVb8Js7GPtKbAvzq7Q/z7tB++jSD8XqNrz9+I9HDgsjwLPWrcC9BVqvYtRrm918i8EIEJx5heFs3lYggPwi26aDEqng8Fh3BPFFvievDjeZPn9lzL8FRlfBoHe9CFS2bh5pFtRKmQ/UhfTZOwIdaXrneM69bgIWuUdrYQqFdw5sJoZUdPKkqar6KyOQb3cpq1kUVFUirjpIvolUkDgoKChGlRkCt4KiN1ntCEWeHgpd/USKzeVTbwVro4qFCDzu9U2zWdToDOU62J/Cl9BXfwbwYHA8oZh0WPBg5Ba1oXdTvyXodmc8jVAVPMoKRa/xXRZQSN4dOUpGNjSevsOjRMpjC5qnyOiaqcTKWSdXR6PMladVzqDjook6XnqZHy7HVN0W2zcc3ukIUerwEAHHi3JVFszgtvg4OXqHhVTW6Almm+lsJqArKxFRjbKwasljEmC/ia9GZsaKknVkC4uWsiNPNx52z1hYJ1UNMkWxpn2P/kImR9dEyP4jIF7HTmbWRNfPjIASKz4fS2oIdD5LcFiDfL9jsnaZb1dFQ0ZE4EYtyq4fi5gTeqB9larFxj1r1qyfXXEqcSgVndg61UiUQ8mDkDKSmUzdVqgUfFY+X4zkvqsdmPBpFERJzSsU/62DOlFGXclC3QRHYdYWCUwVLICx7RTNGXrcAK8EA2Y/k+S/bvoolNSpS57+NvonJ8Rbiz8ZIvJBBnU9dVKMcadWoLyxhZLupyUYHrG7NQ5+xhGOIhusvfnTDBcDOZBD5PB1PdPPHxfdx/R3H+EL/97ktfoKju1vJVIO0idUXGamAUCSBMYXosRqe8SSXcntL20Ep1VArfrJ1ExXJOwMn8YpGsLckbR4tdfNSqYev/9PtxIZttJKDYjkcbtGp+QW2R+DokL++wnu37WN3YJTfa/0B0Z0lPu/cQnifSftzKnKN3oBvjh/mk+/sJnvQT+9+zxmBlLaNWijhyQZ5OjNIm57hLt8MYWG86uepQvBnfV9jrtvkdze+lxMb24kOQ8tTs80JVb1eTp9+PNDD2P1xyl02H7ztKa4zJ7nTl8RUGqGZbs3Hg3d9hunbI3xx4WZGUi3oXx4k/sNZnKXUmt2YfT3YmQzqi2V8qor5rGf59BCtsbmgLm9aa417KZw/BjULp1rFtm2UcAhhmsiixvPVMHpKQ0zP45RXbuPy9YcghEJXOMtdXguwqFNgX/tx9gBzS+1olTC+jgCexRaUXBlSGWSl2ij7Ox+OjWJLSo6HqkzhETpeYVH3KYiAicjnkedzIqVE1ut4lywCUx6m8hEUBD16isFEkrFQ8HV/1UtFCEHdlERDJWrCRC/UG1VHF4Oiohg6SiRMfiBEqV0QVCvYCEbrBgYOYcXCQpCxTTJ1E9/ickf/fAVh1TGyQaygQTWqUQso5NMGw7l21nkXiCo+uo0UoWgJy+85sxHYTNSyYF/NoV2t0qa+vF6JawXaYzlmQ2Yj26NcaXhrjoRKFbViM18OMlWLU/FOE76Iv9WhGrSoVXa3jPNQf5h82cRcTOCd9SHyhbXtCS/PDWEYCNPXcEwMnVJPiFJfnWhnlruCw/TrGUqOpCqLFJ1GW8Y2FRLqAnOxo7R7O/neQAuebBv+o8rywZxXkScMZ7xhAC6kORfitDhLyDgmitUIcZzZPF8BVmYTTpxVSovKr8ee40OR55gZCrJgB5m3IszWwnxp3420f7+V0KkSPHvwgvEnYTkcKnczaCywTbeJqCVy/Qp6sQ1/qQyv0unIe2SKjtkQw7tisB3e6p/hpsGvcM+p30Coq+zl6Trq1hx/vfULvP/Qb1zSr6oBP7K/k6WdUfo+cpy3hye41z/MZD3Mbx54P9WKwaauORLeAi1GgbKtY+Qk6lwaJ5vDqVQRcxqGqmLfsYl8r4qREQzv70PZJXkgdIJ2PcOOtmmejkYQQjS9HWH0KPz007/MW9YP8187nzzzeruaZWd8islEHOIRVBqejbRq2EtJjNkgw+OtfFds4R7/MG2vkglyGlUIVDQ+lnicD8SeZc/2LTx81yZGn+5m3d91Qa6APb9w+b7s60CNhqE9QaUrSGqjgaOD7YFSX53/ce/fM6il6NM0so7kfxc2M12NciTXAcAH259jvTHPW/zHeYf/ONs/NMWRd3fy8D/cTNe3FFhMXX2bkj8mwufDCftBQMb2o1RFw/tdwbj5ighw0TJIOmU8QkFHJap4iSuCdVodSJNzZkk5Dnv7epnq70UqJrHcerDqiJoFioLUNUS1hj3TCFXMV0PM1cNs1JMY2NRCknJMJfCK1pSvxMkXEFIiKjEKshEsj6kqQlu5p9alEPRV2aAL6n4HK6Ch668+5EI3UEIBaIlRHAyR7xH8ZMsB+o0lxutRXiz1U5kIohUFw047E8EKN3VMYCh1LL/AiQURhQKOddYx4xJsAxxdInWJoTQ8u6LjIVX1I9aIo+dN2ygTPkbaEtgd8kws11QsujwZvKEqte4ohq41mu1Y9eVOVg7UBVVbwz4rt8iSNrN2jaLUiCh1vEJgLjd3P02b6qFNBcd/FKXN4TM9CcqDMTxJExUaS9JCEaTTdI9Y8ftRQkGcthjF/gDlqEo1CmfSoD02XWqWuCqZtWtM1kPsSW5kthhiPtlYF3zfs5XJQIxbzRHa1RI3+06x1TPNQ103Uh6IYtbta6dQ5TWQphcr4gUBS1YQxWbFx+X1C3C9zuhYK5+I3MVNwVH69UUGtRKtqnnmElPR8QjJn617kIPdXcxbYSaqMY7nWjk63Ybpr7KrfYqnTq5j45+a1HWFpyYHKNYNtnd/k5hawbclw5I/THxfEF6l2Zms16FURk8pfK3QR9HxUHIMlAWDH929Wx1UIYj2p5m7uYWuahz11PgFr1XW9zN/R5zcANxy52HeEZzmTeYYP6x08u+/8wD+KZX1j2RRsiWcaIByW5Cx34zxro79PPW2QY7tDLPuy16UJ/ef+UzHENgeiO1c5I82PES/niEgfPxzcgfj3x6gbWRtLLcD+2cYWIwxEuhiaX2NoFAwFZ0+TfJAeB/RbUX++6+/EftEC0N/60A6B1YNJ+TDE60wEEpiikb7UoCsU+NDh3+e+eko1286xa3RUe7xD7P5PCHijXqVztABOm/O8M2hHRyY6ULZN4Q5K0k8s4TIl6jPzDVneb7cyKn8hi1M36Uh+oq8e+NzPLvYT/pAB+acoHVvhcyQh08MvR2vWmffbBelJZPOR1R8izXWpysgJccHt3IwvJ1P3/hGAh0Ffnvzw7zNf4qffdPjPLlrHfNf6ibxxYVrquT9vCgq+a1xkptVhLfK9+c3YWRW/qH0+gXYttGXdJ6eG6BY9zBgJrD8x7HJEVRUTGE08hkEbNZVNusLVOU0JWnzYjjGP/pupNeX4v3hF/gkb2aibyOWX0FKi5qj4gA6kp5Ihj5vD4AAABRrSURBVOGCF+nRG7GZVz6JTmfsOxJsG7UmmKlFKdgeMnUTtdqcGKdlK+SdGiFvlUzEwfaqnG+FLDQN4fNhJfwUekH0lfjpxLPE1ALzts6+Uh/BUZXQuI1ycgo7m0MxTfy5DuaKJpZU2dI6x4Q3SqmzhXAigSwWcSpVFEuiVsFQbbYbaYKKhioUUlU/5qzEm6ohneZ7PE4mi1avY6QGGLHCdKp5+oQ8kw2x0zvOG/pO8Ki1kVpXGN30IGp1qgmTkD9Di1HAEC8/ZC1gfjaCOapzMNSJIwXdRpI2dRpTUfGKl6e/qeiYwK3eceJtBb5h7OJfstuwPQbB6TBG0kDJ5hq57qvcJlQxTUTAT7FDwxjKsaN9hvtCB5ksRZmtdqAXJMZsDn8gxsGZThRFYo378S8phA8nYSGFsxxWCOW7saN+KvEwxWqI8cEW8J/i9sAxOowM/1+8B2HojZXFtSzAQCWsUGmzkRJm0yHClyFV/nULsF0oMvT3SeoRH6OhzRz3KXzu7jvp3TTPT/c8x/sCJ/AI7Zxln0fo6ELlBk+Knvbv4RU2LarKx9of5q8+buFR6twTHiah5mhTNSzp8EDHczzm28RI6xbMcAinUETW643G5qrS2NFUlMbmiW0j6pC1fUyXI0zko2gFsfoiY9ukpiN8qnc3uYoHx29je87/IFA2rmPxxhiZzfDmu1+kz7eEV1h8NX0j3/j2zZgzgq5HFiCTW14SS5xyBTWTo/jSOj5TfgO/svVJ/nXHHj76wQdYuH6I7j11fM+fxH9wFnPEYIIOHhA/w/2dL/FbsVHKdR2j4KAWrabHfwGcYglZqZLY38eH236R7Zsm+ezAg2fEcr1m8eute7gtNMJnf+cNpGo6lq0R9iX5w6FvM6ilaFNfntJ5RyXyokHH9+aoPhdlITrI77y1j7u2HuOtsZe4w/ujQpxQNfxKip7EHu678xBHyl185+YtnJpoofehzfjmy4iXjq/qQbDV2zYzf4OBemOaT23/R76UvIkP7/kF/CMGg99KoeSLOItJ/Kks/QvtAKjFNNQs5Mz8OXn5zswcypJBT7qIHQvw5e7r2bxjmg36Ahv9x/nPLRLaEijJ9DV91qJQBPkB2LlzlAMvrCP4hEF0uLDi98nr94AdG/vIcQTg9XoRPh/hzs2M+xI8Gxpkp3ecoLAIKg2vQQV0IQgqBkHFIHxWM/LNus3vt38fAK8Q6ELBIwwUbNYbc8z7wxwKbccfDCIqy5NKaaSTCL2RoiZ5ucKw6mikqiZLeT96k0re1YLCwVwnlq0iPDZSfUUMW1ERqkot4SffLxC9BX4yuh8Fh6Qd4HC2g/hBSWCqjByfOveoFceGahXfvCAfMPFst9hm5Lmte5TntV7yx2P4TsUaHeRSabzJNmbTISbiMQrOEUqWjrfsIGp1ZJPCM+fg2EjHxjdfJXDSZCSWIN8vUKWDVzS81I0KRJQxnH6FouOh6uiE1RK7Pclz2noC2Ah8SQd7ZBRPOo43GCC1qYu98W5aPXn69SXaqeI9S7R1oRIWKmEFhvQ0O4w5uo0kX/HsZrp3AKn68B/RVlWAKzGN0oDFPe2T3OVz+IqQ+MYNwqcc5NET1E97qqUSLC4uf/fzc7p5DbkcypxJJb2V8VoLW405OlQTx3RwQj7UwtqpjFx1llPVrKBkR3iaQ9UhwqeqqMn8JaWQXgwr2gvCqVmIep2O782ReDHI4cFt/FLndVRaJPUWC6E5KJpkW/cMv9XzXdrVIgOa98zv60IltnwTnd1wWxcq67UqlnmCT910H9VwD+0P63DiVMPjheWuLQIlEkb6G3mPx3OtDA93k3hOJXyyvOoxYGnbBEcV9pqD+NuLDHQtkQ92YZ51jdbfQ3FjgtnbNO5/29NsNafZZiT5Um4Hn3nsXvzjKj17ZxunvJ7npncKRTofTVI9HOC/tdzL/vW97A6d4t2xF/kz880M357AmErgWxCUby/wuzu+w0wtyn+cvZPFIwk2Hp9D5tZWRZR2dIKeVJxxNcHXNu1km2+Su725M3Mipmjc6RvFoSGyOpLAa+T+Otk8olSm7+s+rKdMvn3DrXxl/W7uuW6YT3R+B+9yvPmVxFSVO3xj9Pcu8a1f3sk3T24juD924TTKy0AlIujsSbLB39ignipFCExKfAuvM3TkOOgpjR8sbWC7d4pNehlPa4ml6wK0AEzPrIj9VxRCoLW1IqMhSFTZZY7xj8W7MA5NXpb/85VtxuPYSAfskVHECMRneogkwhT6/eS7G+kyjg4HZSdHEl2UjCVMkTzj7aoIPMutBZ1XHCYdUDx0qiWMriL5SoDWoPly3HdZhIUQoGtI04NwYL4QwDunETlWRJ/PUl9tkXEk3rSDb1aDdugNpDng60LojQ1B6UjseJB8j0a9r8Ivx5/EFBJDCCYrMcJHVULjdeTU7AUPGZT1Os6JMbzJCNr4IM8FerkrcpRbvBl+tvsZXoz0sSeyntx0kJu7J7nXHOVL9R08O9+HZ0lBptI45bV1HpidTEEyRWB3CwfzXXgUi92eJB4aWTYKCp3ahesabSkpSYuM4+d0SFiezgo5dBQVSKjXo1ge9rV3MdNmEFNq6MJGQTnn4e8VGp2aRqdmcVPr8zhScNgzeJlH4FwcQ9Bm5mnRGoUS5bqOkXdQSxdXVXlBFAWlBumKj6LjAcqY3iq1SBDb1M7TY+7aQAZM6lETw1MnpFRQLLCXVxYrzWXthuYsJlEKRcJLAYJHTVAFUlGo7DP568fvp24KamGottpct22MbjPDG8JHCSkVgkoZv7AY0u0zohxTVf7dtu/zTO86Dhy9jpbSOphfxM7mGiIsFFhMIrJ5er4L5YNhEvNZlImFpjSbkbZN6FQJ1fKR/QnJT8YO8NjOzXjf8xPoJQetZDN7q4euuya5v+UkQSH5bmmQvxx9A0vDLax/MoOSzFGvvfqNJq06TiZL/0MFKk8H+f37foq/WJfi7s4Rbg8dZ8emSUobPOiizneLQ3xu+FbMhwO0n6jiFMvI+uu8kS8TsQMZhj+/med7NvO3229hIJriZzueJqHlGNQKeJdDWaexpM0JSzBWj/N7B99JaTLIhtHzx+18h6fpnIuQzMT4wLGP4RvK8jNDz7PDN8HdvsLqfckmoZgmSjhEra/Kz/Y+y3bPLOAjk/XTMW6jJcusgaDUqiNUlcLWBNl+jWqxyv8z9nbMucvnuF1eAS6VGnGpV7T983m9mMEgwu/DjgfJbghywOxmPBbFr1WJaUVatDwRtUSbOgNKIwzhFRrvCYyy3TvJL3TsINwaxMjm4XRLvdMH6xWLkE7j2d/oMNa0uh7poC7lMYUgC2wy5ol2ZslsiKPnFfS8SmV9hf/Y/20iSuNY8OFyJ0vDLYSPCzg6Sv1ijtd27MaRRc8dxKcbxFuvJ1VtYTKyxNvD+9ntmaZNNXi4HOHR3BZqk3569xXQFnPU1/DBn2J8hraKRWBDjDklwkvdPvaF+xjwLBJRxvCLOt5lr9XBoSRtJuoJDpR6sV8K0zrioC5mzxu3q8/OwewcLc4GvOkI87UwT8XWocQdbvEeQkc9xxNuKhLqUsVe9kkVJFIVP371ohAI04cMmESiRe40R0gojYZFTknDm7RQitemAKOqlFpUSl0SWVU4OZ2gK3/5RqIp/YCdmoWSzyMrFZRimWi+gicTw/aG2RO+BUeDuq+Ru1qNS6yIzXtufIHdgVPc6pukR62y8a0jHNrRQeKf+ok8ZzRq2VcxLndRSAkLSfRimeqxAX4/fj87WmcIvnOUYt1Dvu7hfeEJIkqZL6Zu5uvDO9BP+uj7QRVjoYjzGp7vef9k3SLxbIrwqJ/jRzfxrxObsT2NIgwjI/BkJP2naqjj8ziltd2M2ymWUeYXCdQsetNRrKDOdx+7nVpYkF9no8RqvH/LXky1xpF8BzPFMFMHOvAmBV3PVjBmcjhLr9Hzd3aRYLmKnm9hZmyAvx4Y4EvbbmBH6wy/1vYoxitkaLocaRR+rCKhiTrDzwzg3CT4SHiGe9uG+fRb2gnvD9C+99KrOxWfj7n3bSC7XvIL/Y/RrsKLtSBHq534JnQ8YwvIdObyfJk1zOlCl+wG6No5y8RIG76jOt6Fy7d6bk5D9tMeW6UC+TwsLuI5fhKgsUGlqKjhEMLrwe6MU+r2s6drPXTBDs80g7rOx3u/yWhHC7//4s8RGgkiiqVLr/VeBexcDgpFzOlBDkx085EdT/I78ZEzPW+TTpk5W+XFVA/hp7yET1loj+3H+XET/qXEPnwMFWg9EEUE/I3caVVF5Ao4ucaZevYa9nxPI60adqYGmSzKqXE8gAfQ2tvI3dJPvtvHnth6TN3i1HQLImXQ81gd33QRTk5e1FE7djoN6TTGwhJtx8IEdnQxZ8V5cr2PN0cPYSrnbnwmK360VU5n9M1XCI/4OTUUw5YOu32nuHdrO3tS2xtHDl3K/6UQCJ+X9A6bu3YOc1/wICHFy0i1nSfS6/EuSZzZ+WuvCGN5XGTAxO6ocl/7MH9zpB3/tETLVS7bamBtnoghHZxSCVGroUpJoFCl8O0E/9x1E8rbJe+I7MOSXoqOB6UGolxb20nj0qHt2Tz+eT9/Wbibg9d1MV0Ms1TwU0ia6As6wXFo3ZtvpLqsULaGUyoh6vUzedJOpbrcTOTKbrbi5PIEDy9hTptk0wnyqqAr34ip+04sQb5RgHIpyGoVJ5PFf9SguxSjctDDn7zwwDkN7QF8S5KW1JEV/DavjTaVpMWWjGwJ8tSuhkHvjb/AyHUJFj+0A2/aITBWRMmVcUbHf0Q8hccDW4awol7SGwwqccFt1x3hp1qeJygspu0qnzp2J/aLETqPVRrtPtfIHFG8XtB15FAvtYQP71gaOTW7YpV6QtNQBnqxo37G3xik3GXTEk/xQqaX4KhCfF8KMXv5uuOtUQGWDaGoNrqmiaUkHVadWleUJ28YZJ13gYhaouLoKHUQNQtnBTsUrThSwvOHCO7TSZjX85RYj2dewzcn6J2oEzg8g8wXsJcurVXla/7Z5TG82nBKJVheMYWfO/e9H1c2Th9Z5YwWUUfH8AP+C1y72tJkz80jUmnM2Z0cKPex3TvJm3xlxrqf55O3vgV9zsBRA5gLBtq4+iPOiOLxkBsMUOxQKd5eYFPHAh9tf4TdHsG07TBT91E9Gqb/0RL6+OLLecVrgNMd37KDAXJ9Kol6BE+qER5ZES9dVbE6wxS6PCTunuGne57jO4tbGcvECE7ZOC8dff1/41VYmwL8CqRtI9NZDCmpfLOLP297J7ZXgoS+wxVkplEiuqaREmnbxA7l0UsBjFwdPVdDSxYb9l+FQumyMkjbhlqN1herfNr/dpzNBf5g57ewUXjPdfuY3hBheHMbiyUPzv27EK/wRRxdYnYXiPlLvKv9GH2eJXKOlycqNp9beBvDyXYix0CfSjZywtcSqgq6Tma9Sm1ngbFuE+9NGwlOOIROldEWcjjj08tNmS7u0Sg8HpRImMzdg5QSCtnNddRwlXDV4G/HbmHpQCv+GUFgNHPZNyKvCAFGykasLpOh4x/SYOiIUBCpazC70IizXgk4NvKFQwReePmltbHQc1nTLPe61p88xMB+Pwvv2cjXOndxe+wE/2/7840y/4HGcU4FWcV5Rb67IgSmMM60A7CkzbdKYZ4rdfPEwY34R3W6D+epj08249u9OqoCukZxwOJj2x9ndmOEpWqARw9sphY0iY5oGPNLUKshqxdxNwnROC26JcrsPTaDgzP8VvcPadey/OaB95MZidHzRB3/obnX3sBdAa4MAT6NlDjVKqJebyzrFbVxSrKLy7WAbSPLFaLDZY5+awP72od4cNMCGyML/GLr48SUCj2aglc597a2pM1xq0bG8fBwfhsT5Rg/OLkeJ2kQf0khOGWhLmRWvMx2JZDLRUKJpzv4VPqt2B1VorECkfY89bcIRncGMW/ajpEF/7yNcECxJKIu0co2wpaIuoPjVcl3e6j7oBIXWEFJe88CPs3ik4ffTLngIfiil/ZZB3Msh8wVkD9GFtKlcmUJMMtxTWjkF7u4XEOcjlMrTx2g64cK6voBkje18cSmNtremmOnf5xBfelM4dJpLGmzv9rNcLmTLx7YjTrnoeMZB/9EAWVsFjuZWpPiC5zJ649/eR8tPh+L797E0o0ab73+JT7d9QzHrSL7q538S+o6Hh8ZwqmpUFNQyiqeVGOPSK1ALQyB3Uv0hdK8q3UffqXGcKWTQ/lOPA+H6BiuYhwewV5cXNVV6RUnwC4u1zxSAg4imyd8MoBa8/JV9Ta+7L+F/ytWPeeEGgDHUXBSBmpZIXIKPFmJf6KAupRbc2XoF0LW61AuEzlZQQov38/uYvuGQXTVxmdYVCwNj8/C8dSp11Vsj0ZNaji6RERr+Pw1drdNoAmbr8ztJl3xMTPWgp5S6T5pYczmkE1w6lwBdnG5EpGS+tw8Yn6BIBB6cLnd6wUOrT3diOp08x4pndXvjfI6OOP9P76f+BMKrdEwIhigtKmN1CadQp9D15Z5VKXxPQtVDxmvSUcsxx8OfZO42qgReL48wLcfux7/lMKmh1OIuUWcbL5pefGuALu4XMksi+g1UzghJUgbWWx4q94Zk4gWQs+rLOTbkcvV2Yol8JYhGTD5taWfQVUbwmzlDVoOC8zFOko6h1MqNzXn2RVgFxeXK44zfY1TaXxHVHxCoeUCvTGEeu4ZNNKqg3So23bT27C6Auzi4nLlspyi96qXrJIpPw7XastPFxcXl6bjCrCLi4tLk3AF2MXFxaVJuALs4uLi0iRcAXZxcXFpEq4Au7i4uDQJIS8hD04IsQiMXz5z1gR9UsrExV58jYwJXMK4uGNyfq6RcXHH5Pycd1wuSYBdXFxcXFYONwTh4uLi0iRcAXZxcXFpEq4Au7i4uDQJV4BdXFxcmoQrwC4uLi5NwhVgFxcXlybhCrCLi4tLk3AF2MXFxaVJuALs4uLi0iT+D4NIqwDfb66lAAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<Figure size 432x288 with 5 Axes>"
            ]
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "A0AH7VSN7Vn0",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "1b795cc4-4133-4eaa-9d44-63c0a6025db3"
      },
      "source": [
        "h1=net.module_.hidden.weight\n",
        "print(h1.detach().numpy().shape)\n",
        "np.savetxt(\"111.csv\",h1.detach().numpy(), delimiter=',')\n",
        "h2= net.module_.hidden1.weight\n",
        "np.savetxt(\"222.csv\",h2.detach().numpy(), delimiter=',')\n",
        "h3= net.module_.hidden2.weight\n",
        "np.savetxt(\"333.csv\",h3.detach().numpy(), delimiter=',')\n",
        "h4= net.module_.hidden3.weight\n",
        "np.savetxt(\"444.csv\",h4.detach().numpy(), delimiter=',')\n",
        "h5= net.module_.output4.weight\n",
        "np.savetxt(\"555.csv\",h5.detach().numpy(), delimiter=',')\n",
        "\n"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(8, 784)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "NPkcwuQkSthI",
        "outputId": "8e7ddc17-4bad-4edf-8562-e0b50b192177"
      },
      "source": [
        "b1 = net.module_.hidden.bias;\n",
        "print(b1)\n",
        "\n",
        "b2 = net.module_.hidden1.bias;\n",
        "print(b2)\n",
        "\n",
        "b3 = net.module_.hidden2.bias;\n",
        "print(b3)\n",
        "\n",
        "b4 = net.module_.hidden3.bias;\n",
        "print(b4)\n",
        "\n",
        "\n",
        "b5 = net.module_.output4.bias;\n",
        "print(b5)\n"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Parameter containing:\n",
            "tensor([2.9851, 0.8467, 1.8964, 2.9366, 3.7874, 0.6813, 1.9752, 0.5708],\n",
            "       requires_grad=True)\n",
            "Parameter containing:\n",
            "tensor([-0.8909,  3.1375, -0.9986,  1.4742,  6.6393,  5.6268,  2.5455, -2.2740,\n",
            "         5.2660,  0.1251,  2.4599,  7.3509, -2.3175, -4.1341,  2.2511, -3.4067],\n",
            "       requires_grad=True)\n",
            "Parameter containing:\n",
            "tensor([ 1.2774, -1.9519, -0.7424, -0.5910, -0.1406,  1.4695,  1.7393,  2.1252,\n",
            "         4.3853, -1.1400,  1.0052, -2.1643], requires_grad=True)\n",
            "Parameter containing:\n",
            "tensor([ 1.5017,  0.1662,  2.6586,  2.5365,  1.3033,  3.9124, -0.1954, -0.8800],\n",
            "       requires_grad=True)\n",
            "Parameter containing:\n",
            "tensor([ 1.5805e+00,  1.5317e+00, -2.7760e+00,  1.0533e+00, -9.8251e-02,\n",
            "         4.9737e-01, -1.3128e+00,  2.0145e+00,  1.5262e-01, -1.8041e+00,\n",
            "         8.4457e-02, -3.3549e+00,  3.7815e+00,  4.7976e+00, -6.8797e-01,\n",
            "        -5.7070e-01, -3.4836e+00, -1.6651e+00, -7.1152e-01, -6.2390e-01,\n",
            "         4.9149e+00,  4.3996e+00,  4.0420e+00,  4.8088e+00, -3.6460e+00,\n",
            "         5.0570e-01,  6.5232e-01, -6.7006e-01, -2.6455e+00, -2.2698e+00,\n",
            "        -5.2573e+00, -1.6270e+00, -8.5105e-01,  8.9991e-02,  1.7505e+00,\n",
            "         3.1373e+00, -3.5241e+00, -2.7846e+00, -7.9315e-01,  4.0354e-03],\n",
            "       requires_grad=True)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "z9DoBnKBL5yj"
      },
      "source": [
        "# New Section"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "fjiAzfUvGy_g",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "75069095-5afe-4a14-d416-22f0e55a820f"
      },
      "source": [
        "output_dim = T*len(np.unique(mnist.target))\n",
        "print(max(y_train))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "39\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "WTac4C-D_z40"
      },
      "source": [
        ""
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "9FrSvolJ_0Sy",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "e239e6d6-eadc-421d-d664-72af944fa74f"
      },
      "source": [
        "print(accuracy_score(y_test, y_pred))\n"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "0.8076607142857143\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "UUZ2XX6kMN4W"
      },
      "source": [
        "\n",
        "\n"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "F5KX4i3wRTiZ"
      },
      "source": [
        "Y_classify_test=np.zeros(y_test.shape)\n",
        "Y_classify_pred=np.zeros(y_test.shape)\n",
        "Y_ssl_test=np.zeros(y_test.shape)\n",
        "Y_ssl_pred=np.zeros(y_test.shape)\n",
        "\n",
        "for count in range(10):\n",
        "  Y_classify_test[(y_test>=count*T) & (y_test<(count+1)*T)]=count\n",
        "  Y_classify_pred[(y_pred>=count*T) & (y_pred<(count+1)*T)]=count\n",
        "\n",
        "for c in range(T):\n",
        "  Y_ssl_test[y_test % T==c]=np.int(c)\n",
        "  Y_ssl_pred[y_pred % T==c]=np.int(c)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "SIyJ15eWLrZ9",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 160
        },
        "outputId": "feaf00da-feeb-498f-8d8e-0f7d7645d19e"
      },
      "source": [
        "print(\"Total accuracy:\", accuracy_score(y_test, y_pred))\n",
        "print(\"Classifcation Accuravy:\", accuracy_score(Y_classify_test, Y_classify_pred))\n",
        "print(\"SSL Task accuracy:\", accuracy_score(Y_ssl_test,Y_ssl_pred))\n",
        "\n",
        "\n",
        "error_mask = y_pred != y_test\n",
        "plot_example(X_test[error_mask], y_pred[error_mask]%T)\n",
        "\n"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Total accuracy: 0.8076607142857143\n",
            "Classifcation Accuravy: 0.8111428571428572\n",
            "SSL Task accuracy: 0.9840535714285714\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 5 Axes>"
            ]
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "FSigmZ3XshJc",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 108
        },
        "outputId": "5027d027-9386-407e-dd2c-dd822cd44b13"
      },
      "source": [
        "error_mask = Y_classify_test != Y_classify_pred\n",
        "plot_example(X_test[error_mask], Y_classify_pred[error_mask])"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 5 Axes>"
            ]
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "J2jI5XLUtBjd",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 108
        },
        "outputId": "d5952aff-35b9-4068-91cb-e5a1ee032e5b"
      },
      "source": [
        "error_mask = Y_ssl_test != Y_ssl_pred\n",
        "plot_example(X_test[error_mask], Y_ssl_pred[error_mask])"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 5 Axes>"
            ]
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "rz3CCWOsyc5k",
        "outputId": "7285d8cc-44cb-42cc-9e79-4a7a35cde38b"
      },
      "source": [
        "a0 = [1 ,0.5, .1, -1 ,0]\n",
        "a = torch.tensor(np.asarray(a0))\n",
        "b = F.softmax(a)\n",
        "print(b)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "tensor([0.3974, 0.2410, 0.1616, 0.0538, 0.1462], dtype=torch.float64)\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/ipykernel_launcher.py:3: UserWarning: Implicit dimension choice for softmax has been deprecated. Change the call to include dim=X as an argument.\n",
            "  This is separate from the ipykernel package so we can avoid doing imports until\n"
          ],
          "name": "stderr"
        }
      ]
    }
  ]
}