Week 11/01 - Pytorch.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "# conda install pytorch torchvision cudatoolkit=10.1 -c pytorch"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch\n",
    "import numpy as np \n",
    "import pandas as pd"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Intro Pytorch"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<img src=\"https://www.researchgate.net/publication/335742056/figure/fig1/AS:801925004005376@1568205228035/Mode-i-tensors-of-a-3-way-tensor-For-a-3-way-tensor-X-R-n-1-n-2-n-3-it-has-3-types.png\">"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[1., 1.],\n",
       "        [1., 1.]])"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Tensor of ones \n",
    "ones = torch.ones(2,2)\n",
    "ones"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[ 0.0000e+00,  1.8750e+00],\n",
       "        [-8.6639e-26,  4.5908e-41]])"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# 2x2 Random Tensor \n",
    "ran = torch.Tensor(2,2)\n",
    "ran"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[0.8604, 0.8176],\n",
       "        [0.1798, 0.5157]])"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Uniform values\n",
    "ran.uniform_(0,1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[0.4788, 0.8232],\n",
       "        [0.2921, 0.5677]])"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Random tensor \n",
    "ran2 = torch.rand(2,2)\n",
    "ran2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[1.3392, 1.6408],\n",
       "        [0.4720, 1.0834]])"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Operations\n",
    "result = ran + ran2\n",
    "result"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "torch.Size([2, 2])"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Shape of our tensor\n",
    "result.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(tensor([[[ 0.9280, -0.2825],\n",
       "          [-1.0039, -0.9079],\n",
       "          [-0.0081,  0.3408]],\n",
       " \n",
       "         [[ 0.7514,  0.0672],\n",
       "          [ 1.3958, -0.1282],\n",
       "          [ 0.2167,  0.2324]],\n",
       " \n",
       "         [[-0.7716, -1.2923],\n",
       "          [ 0.8282, -0.3049],\n",
       "          [ 0.7549,  1.8177]],\n",
       " \n",
       "         [[ 2.1639,  0.0411],\n",
       "          [ 0.1020, -0.0212],\n",
       "          [-0.9488,  1.7438]],\n",
       " \n",
       "         [[ 0.2883,  1.0290],\n",
       "          [-0.1566, -0.0819],\n",
       "          [-0.3464,  0.9773]],\n",
       " \n",
       "         [[ 0.3368,  0.8061],\n",
       "          [ 1.4546,  0.2340],\n",
       "          [-0.8690, -1.3185]],\n",
       " \n",
       "         [[-1.2758, -0.6351],\n",
       "          [-0.6969, -0.1927],\n",
       "          [ 1.1757, -1.9515]],\n",
       " \n",
       "         [[-0.4388,  0.4235],\n",
       "          [ 0.6944,  0.6487],\n",
       "          [-0.3601,  0.5507]]]), torch.Size([8, 3, 2]))"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "A = torch.randn((8,3,2))\n",
    "A, A.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor(0.9280)"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "A[0][0][0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "tensor([[1.3392, 1.6408, 0.4720, 1.0834]])\n",
      "torch.Size([1, 4])\n"
     ]
    }
   ],
   "source": [
    "# Reshape\n",
    "reshaped = result.view(1,4)\n",
    "print(reshaped)\n",
    "print(reshaped.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "tensor([[1.3392],\n",
      "        [1.6408],\n",
      "        [0.4720],\n",
      "        [1.0834]])\n",
      "torch.Size([4, 1])\n"
     ]
    }
   ],
   "source": [
    "reshaped = result.view(4,1)\n",
    "print(reshaped)\n",
    "print(reshaped.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[1., 2.],\n",
       "        [3., 4.]])"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Create tensor with my own values\n",
    "ten = torch.tensor([[1.0,2.0],[3.0,4.0]])\n",
    "ten"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor(2.)"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ten[0][1]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       " 1.0\n",
       " 2.0\n",
       " 3.0\n",
       " 4.0\n",
       "[torch.FloatStorage of size 4]"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# storage\n",
    "ten.storage()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(2, 1)"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Stride\n",
    "ten.stride()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(tensor([[1., 3.],\n",
       "         [2., 4.]]), (1, 2))"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tent = ten.t()\n",
    "tent, tent.stride()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(tensor([1, 2, 3, 4, 5]), tensor([[1, 2, 3, 4, 5]]), tensor([[1],\n",
       "         [2],\n",
       "         [3],\n",
       "         [4],\n",
       "         [5]]))"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Add dimensions\n",
    "\n",
    "ten2 = torch.tensor([1,2,3,4,5])\n",
    "ten2, torch.unsqueeze(ten2, 0), torch.unsqueeze(ten2, 1)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Numpy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "na = np.random.randn(2,2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "numpy_tensor = torch.from_numpy(na)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "tensor([1.5000, 3.5000]) tensor([2., 3.]) tensor(2.5000)\n"
     ]
    }
   ],
   "source": [
    "new_tensor = torch.tensor([[1.0, 2.0],[3.0, 4.0]])\n",
    "print(torch.mean(new_tensor, dim=1),\n",
    "      torch.mean(new_tensor, dim=0),\n",
    "      torch.mean(new_tensor))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "tensor([[1],\n",
      "        [2],\n",
      "        [3],\n",
      "        [4]])\n",
      "tensor([[[1],\n",
      "         [2],\n",
      "         [3],\n",
      "         [4]]])\n"
     ]
    }
   ],
   "source": [
    "tensorX = torch.tensor([1, 2, 3, 4])\n",
    "print(torch.unsqueeze(tensorX, 1))\n",
    "tensorX = torch.tensor([[1],\n",
    "        [2],\n",
    "        [3],\n",
    "        [4]])\n",
    "print(torch.unsqueeze(tensorX,0))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor(-0.0567, dtype=torch.float64)"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Mean\n",
    "torch.mean(numpy_tensor)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "tensor([-0.3030,  0.1896], dtype=torch.float64)\n",
      "tensor([ 0.7738, -0.8872], dtype=torch.float64)\n"
     ]
    }
   ],
   "source": [
    "# Dimensions of or tensor\n",
    "\n",
    "print(torch.mean(numpy_tensor, dim=0))\n",
    "print(torch.mean(numpy_tensor, dim=1))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "# save tensor\n",
    "torch.save(numpy_tensor, 'nt.t')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[ 0.8291,  0.7184],\n",
       "        [-1.4352, -0.3393]], dtype=torch.float64)"
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# load tensor\n",
    "tensor_loaded = torch.load('nt.t')\n",
    "tensor_loaded"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Datasets"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Unnamed: 0</th>\n",
       "      <th>ID</th>\n",
       "      <th>Name</th>\n",
       "      <th>Age</th>\n",
       "      <th>Photo</th>\n",
       "      <th>Nationality</th>\n",
       "      <th>Flag</th>\n",
       "      <th>Overall</th>\n",
       "      <th>Potential</th>\n",
       "      <th>Club</th>\n",
       "      <th>...</th>\n",
       "      <th>Composure</th>\n",
       "      <th>Marking</th>\n",
       "      <th>StandingTackle</th>\n",
       "      <th>SlidingTackle</th>\n",
       "      <th>GKDiving</th>\n",
       "      <th>GKHandling</th>\n",
       "      <th>GKKicking</th>\n",
       "      <th>GKPositioning</th>\n",
       "      <th>GKReflexes</th>\n",
       "      <th>Release Clause</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>158023</td>\n",
       "      <td>L. Messi</td>\n",
       "      <td>31</td>\n",
       "      <td>https://cdn.sofifa.org/players/4/19/158023.png</td>\n",
       "      <td>Argentina</td>\n",
       "      <td>https://cdn.sofifa.org/flags/52.png</td>\n",
       "      <td>94</td>\n",
       "      <td>94</td>\n",
       "      <td>FC Barcelona</td>\n",
       "      <td>...</td>\n",
       "      <td>96.0</td>\n",
       "      <td>33.0</td>\n",
       "      <td>28.0</td>\n",
       "      <td>26.0</td>\n",
       "      <td>6.0</td>\n",
       "      <td>11.0</td>\n",
       "      <td>15.0</td>\n",
       "      <td>14.0</td>\n",
       "      <td>8.0</td>\n",
       "      <td>€226.5M</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>20801</td>\n",
       "      <td>Cristiano Ronaldo</td>\n",
       "      <td>33</td>\n",
       "      <td>https://cdn.sofifa.org/players/4/19/20801.png</td>\n",
       "      <td>Portugal</td>\n",
       "      <td>https://cdn.sofifa.org/flags/38.png</td>\n",
       "      <td>94</td>\n",
       "      <td>94</td>\n",
       "      <td>Juventus</td>\n",
       "      <td>...</td>\n",
       "      <td>95.0</td>\n",
       "      <td>28.0</td>\n",
       "      <td>31.0</td>\n",
       "      <td>23.0</td>\n",
       "      <td>7.0</td>\n",
       "      <td>11.0</td>\n",
       "      <td>15.0</td>\n",
       "      <td>14.0</td>\n",
       "      <td>11.0</td>\n",
       "      <td>€127.1M</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>190871</td>\n",
       "      <td>Neymar Jr</td>\n",
       "      <td>26</td>\n",
       "      <td>https://cdn.sofifa.org/players/4/19/190871.png</td>\n",
       "      <td>Brazil</td>\n",
       "      <td>https://cdn.sofifa.org/flags/54.png</td>\n",
       "      <td>92</td>\n",
       "      <td>93</td>\n",
       "      <td>Paris Saint-Germain</td>\n",
       "      <td>...</td>\n",
       "      <td>94.0</td>\n",
       "      <td>27.0</td>\n",
       "      <td>24.0</td>\n",
       "      <td>33.0</td>\n",
       "      <td>9.0</td>\n",
       "      <td>9.0</td>\n",
       "      <td>15.0</td>\n",
       "      <td>15.0</td>\n",
       "      <td>11.0</td>\n",
       "      <td>€228.1M</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>3</td>\n",
       "      <td>193080</td>\n",
       "      <td>De Gea</td>\n",
       "      <td>27</td>\n",
       "      <td>https://cdn.sofifa.org/players/4/19/193080.png</td>\n",
       "      <td>Spain</td>\n",
       "      <td>https://cdn.sofifa.org/flags/45.png</td>\n",
       "      <td>91</td>\n",
       "      <td>93</td>\n",
       "      <td>Manchester United</td>\n",
       "      <td>...</td>\n",
       "      <td>68.0</td>\n",
       "      <td>15.0</td>\n",
       "      <td>21.0</td>\n",
       "      <td>13.0</td>\n",
       "      <td>90.0</td>\n",
       "      <td>85.0</td>\n",
       "      <td>87.0</td>\n",
       "      <td>88.0</td>\n",
       "      <td>94.0</td>\n",
       "      <td>€138.6M</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>4</td>\n",
       "      <td>192985</td>\n",
       "      <td>K. De Bruyne</td>\n",
       "      <td>27</td>\n",
       "      <td>https://cdn.sofifa.org/players/4/19/192985.png</td>\n",
       "      <td>Belgium</td>\n",
       "      <td>https://cdn.sofifa.org/flags/7.png</td>\n",
       "      <td>91</td>\n",
       "      <td>92</td>\n",
       "      <td>Manchester City</td>\n",
       "      <td>...</td>\n",
       "      <td>88.0</td>\n",
       "      <td>68.0</td>\n",
       "      <td>58.0</td>\n",
       "      <td>51.0</td>\n",
       "      <td>15.0</td>\n",
       "      <td>13.0</td>\n",
       "      <td>5.0</td>\n",
       "      <td>10.0</td>\n",
       "      <td>13.0</td>\n",
       "      <td>€196.4M</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 89 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   Unnamed: 0      ID               Name  Age  \\\n",
       "0           0  158023           L. Messi   31   \n",
       "1           1   20801  Cristiano Ronaldo   33   \n",
       "2           2  190871          Neymar Jr   26   \n",
       "3           3  193080             De Gea   27   \n",
       "4           4  192985       K. De Bruyne   27   \n",
       "\n",
       "                                            Photo Nationality  \\\n",
       "0  https://cdn.sofifa.org/players/4/19/158023.png   Argentina   \n",
       "1   https://cdn.sofifa.org/players/4/19/20801.png    Portugal   \n",
       "2  https://cdn.sofifa.org/players/4/19/190871.png      Brazil   \n",
       "3  https://cdn.sofifa.org/players/4/19/193080.png       Spain   \n",
       "4  https://cdn.sofifa.org/players/4/19/192985.png     Belgium   \n",
       "\n",
       "                                  Flag  Overall  Potential  \\\n",
       "0  https://cdn.sofifa.org/flags/52.png       94         94   \n",
       "1  https://cdn.sofifa.org/flags/38.png       94         94   \n",
       "2  https://cdn.sofifa.org/flags/54.png       92         93   \n",
       "3  https://cdn.sofifa.org/flags/45.png       91         93   \n",
       "4   https://cdn.sofifa.org/flags/7.png       91         92   \n",
       "\n",
       "                  Club  ... Composure Marking StandingTackle  SlidingTackle  \\\n",
       "0         FC Barcelona  ...      96.0    33.0           28.0           26.0   \n",
       "1             Juventus  ...      95.0    28.0           31.0           23.0   \n",
       "2  Paris Saint-Germain  ...      94.0    27.0           24.0           33.0   \n",
       "3    Manchester United  ...      68.0    15.0           21.0           13.0   \n",
       "4      Manchester City  ...      88.0    68.0           58.0           51.0   \n",
       "\n",
       "  GKDiving  GKHandling  GKKicking  GKPositioning GKReflexes Release Clause  \n",
       "0      6.0        11.0       15.0           14.0        8.0        €226.5M  \n",
       "1      7.0        11.0       15.0           14.0       11.0        €127.1M  \n",
       "2      9.0         9.0       15.0           15.0       11.0        €228.1M  \n",
       "3     90.0        85.0       87.0           88.0       94.0        €138.6M  \n",
       "4     15.0        13.0        5.0           10.0       13.0        €196.4M  \n",
       "\n",
       "[5 rows x 89 columns]"
      ]
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "## football dataset\n",
    "url = 'https://raw.githubusercontent.com/amanthedorkknight/fifa18-all-player-statistics/master/2019/data.csv'\n",
    "df = pd.read_csv(url)\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Index(['Unnamed: 0', 'ID', 'Name', 'Age', 'Photo', 'Nationality', 'Flag',\n",
       "       'Overall', 'Potential', 'Club', 'Club Logo', 'Value', 'Wage', 'Special',\n",
       "       'Preferred Foot', 'International Reputation', 'Weak Foot',\n",
       "       'Skill Moves', 'Work Rate', 'Body Type', 'Real Face', 'Position',\n",
       "       'Jersey Number', 'Joined', 'Loaned From', 'Contract Valid Until',\n",
       "       'Height', 'Weight', 'LS', 'ST', 'RS', 'LW', 'LF', 'CF', 'RF', 'RW',\n",
       "       'LAM', 'CAM', 'RAM', 'LM', 'LCM', 'CM', 'RCM', 'RM', 'LWB', 'LDM',\n",
       "       'CDM', 'RDM', 'RWB', 'LB', 'LCB', 'CB', 'RCB', 'RB', 'Crossing',\n",
       "       'Finishing', 'HeadingAccuracy', 'ShortPassing', 'Volleys', 'Dribbling',\n",
       "       'Curve', 'FKAccuracy', 'LongPassing', 'BallControl', 'Acceleration',\n",
       "       'SprintSpeed', 'Agility', 'Reactions', 'Balance', 'ShotPower',\n",
       "       'Jumping', 'Stamina', 'Strength', 'LongShots', 'Aggression',\n",
       "       'Interceptions', 'Positioning', 'Vision', 'Penalties', 'Composure',\n",
       "       'Marking', 'StandingTackle', 'SlidingTackle', 'GKDiving', 'GKHandling',\n",
       "       'GKKicking', 'GKPositioning', 'GKReflexes', 'Release Clause'],\n",
       "      dtype='object')"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(torch.Size([18159, 5]), 'torch.FloatTensor')"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "subset = df[['Overall', 'Age','International Reputation', 'Weak Foot', 'Skill Moves' ]].dropna(axis=0, how='any')\n",
    "\n",
    "columns = subset.columns[1:]\n",
    "players = torch.tensor(subset.values).float()\n",
    "players.shape, players.type()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "torch.Size([18159, 4])"
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data = players[:, 1:]\n",
    "data.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "torch.Size([18159])"
      ]
     },
     "execution_count": 37,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "target = players[:,0]\n",
    "target.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([25.1225,  1.1132,  2.9473,  2.3613])"
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mean = torch.mean(data, dim=0)\n",
    "mean"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([4.6706, 0.3940, 0.6605, 0.7562])"
      ]
     },
     "execution_count": 39,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "std = torch.std(data, dim=0)\n",
    "std"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[ 2.7196,  6.1919,  1.2953,  1.8845],\n",
       "        [ 3.6450,  6.1919,  1.2953,  3.0345],\n",
       "        [ 0.4060,  6.1919,  2.5258,  3.0345],\n",
       "        ...,\n",
       "        [-4.2211, -0.1804,  0.0648, -0.4155],\n",
       "        [-3.7584, -0.1804,  0.0648, -0.4155],\n",
       "        [-4.2211, -0.1804,  0.0648, -0.4155]])"
      ]
     },
     "execution_count": 40,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "norm = (data-mean)/torch.sqrt(std)\n",
    "norm"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(tensor([28.3455,  3.3818,  3.4455,  3.0364]),\n",
       " tensor([27.2285,  1.3561,  3.1114,  2.7515]),\n",
       " tensor([24.3676,  1.0106,  2.8864,  2.2208]))"
      ]
     },
     "execution_count": 41,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "good = data[torch.ge(target , 85)] # Grater or equal\n",
    "average = data[torch.gt(target, 70) & torch.lt(target, 85)] # Greater Than and less than\n",
    "not_good = data[torch.le(target, 70)] # Less or equal\n",
    "\n",
    "\n",
    "mean_g = torch.mean(good, dim=0)\n",
    "mean_avg = torch.mean(average, dim=0)\n",
    "mean_bad = torch.mean(not_good, dim=0)\n",
    "\n",
    "mean_g, mean_avg, mean_bad"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " Age                        28.35  27.23  24.37\n",
      " International Reputation    3.38   1.36   1.01\n",
      " Weak Foot                   3.45   3.11   2.89\n",
      " Skill Moves                 3.04   2.75   2.22\n"
     ]
    }
   ],
   "source": [
    "for i, args in enumerate(zip(columns, mean_g, mean_avg, mean_bad)):\n",
    "    print(' {:25} {:6.2f} {:6.2f} {:6.2f}'.format(*args))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Device"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "device(type='cpu')"
      ]
     },
     "execution_count": 43,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# CPU or GPU\n",
    "\n",
    "good.device"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "True"
      ]
     },
     "execution_count": 44,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "torch.cuda.is_available() # GPU is available "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [],
   "source": [
    "# set up device \n",
    "device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [],
   "source": [
    "x = torch.randn(3,5) # CPU\n",
    "y = torch.ones_like(x, device=device) # GPU"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [],
   "source": [
    "x = x.to(device) # to gpu"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[2.7553, 0.8537, 2.4414, 2.2105, 1.4826],\n",
       "        [1.7138, 0.3800, 1.9334, 0.6455, 1.4251],\n",
       "        [1.2834, 1.0221, 1.0574, 2.4910, 0.4829]], device='cuda:0')"
      ]
     },
     "execution_count": 48,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "r = x + y \n",
    "r          # r.cpu() cpu "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### CIFAR"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [],
   "source": [
    "from torchvision import datasets"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['CIFAR10',\n",
       " 'CIFAR100',\n",
       " 'Caltech101',\n",
       " 'Caltech256',\n",
       " 'CelebA',\n",
       " 'Cityscapes',\n",
       " 'CocoCaptions',\n",
       " 'CocoDetection',\n",
       " 'DatasetFolder',\n",
       " 'EMNIST',\n",
       " 'FakeData',\n",
       " 'FashionMNIST',\n",
       " 'Flickr30k',\n",
       " 'Flickr8k',\n",
       " 'HMDB51',\n",
       " 'ImageFolder',\n",
       " 'ImageNet',\n",
       " 'KMNIST',\n",
       " 'Kinetics400',\n",
       " 'LSUN',\n",
       " 'LSUNClass',\n",
       " 'MNIST',\n",
       " 'Omniglot',\n",
       " 'PhotoTour',\n",
       " 'QMNIST',\n",
       " 'SBDataset',\n",
       " 'SBU',\n",
       " 'SEMEION',\n",
       " 'STL10',\n",
       " 'SVHN',\n",
       " 'UCF101',\n",
       " 'USPS',\n",
       " 'VOCDetection',\n",
       " 'VOCSegmentation',\n",
       " 'VisionDataset',\n",
       " '__all__',\n",
       " '__builtins__',\n",
       " '__cached__',\n",
       " '__doc__',\n",
       " '__file__',\n",
       " '__loader__',\n",
       " '__name__',\n",
       " '__package__',\n",
       " '__path__',\n",
       " '__spec__',\n",
       " 'caltech',\n",
       " 'celeba',\n",
       " 'cifar',\n",
       " 'cityscapes',\n",
       " 'coco',\n",
       " 'fakedata',\n",
       " 'flickr',\n",
       " 'folder',\n",
       " 'hmdb51',\n",
       " 'imagenet',\n",
       " 'kinetics',\n",
       " 'lsun',\n",
       " 'mnist',\n",
       " 'omniglot',\n",
       " 'phototour',\n",
       " 'sbd',\n",
       " 'sbu',\n",
       " 'semeion',\n",
       " 'stl10',\n",
       " 'svhn',\n",
       " 'ucf101',\n",
       " 'usps',\n",
       " 'utils',\n",
       " 'video_utils',\n",
       " 'vision',\n",
       " 'voc']"
      ]
     },
     "execution_count": 50,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dir(datasets)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Files already downloaded and verified\n"
     ]
    }
   ],
   "source": [
    "cifar = datasets.CIFAR10('./data/(datasets)', download=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "50000"
      ]
     },
     "execution_count": 52,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(cifar)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "torch.Size([50000, 32, 32, 3])"
      ]
     },
     "execution_count": 53,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data = torch.Tensor(cifar.data)\n",
    "data.size()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "50000 - Imgs\n",
    "32x32 pixels\n",
    "3 ch (RGB)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "dog\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x2434dda3288>"
      ]
     },
     "execution_count": 55,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "img, label = cifar[40]\n",
    "classes = ['plane','car','bird','cat','deer','dog','frog','horse','ship','truck']\n",
    "print(classes[label])\n",
    "plt.imshow(img)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "deer\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x2434e29fd08>"
      ]
     },
     "execution_count": 56,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAPsAAAD5CAYAAADhukOtAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjEsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy8QZhcZAAAdO0lEQVR4nO2dXYxlWXXf/+uce+5H3arq6p7+oGlGGYwmkrEVD6g0QiKyiJ1YE2RpQIoteEDzgNxWZKQgOQ8jIgUi5QFHAcQTURNGHkeEjxgQowjFRiNHyC9jGjwMg8cxmIxh6KY/6/vW/TwrD/eO0zPe/1XVXVW3Bvb/J7X61tl3n7POPnvdc+/+n7WWuTuEED//FMdtgBBiPsjZhcgEObsQmSBnFyIT5OxCZIKcXYhMaByks5k9AuCTAEoA/9XdPxoerNnwVruVbgwUQCPbI9WwKPnnWFmWvGOw00ldJ7cz+wDAjLc62d9e/YqgzcipFcbHo675OUdt7tx+RhGMfXRekUQctVmRPu/xaEL7jMdj2obAxmgmhPOA2B+N73ictn8yHqOu6+TB7F51djMrAfwNgH8B4CUA3wTwXnf/K9anu9z1X1r9pfT+gklVTNInHXRBp9ulbSdOnKBtdeCAW1tbye2FcUPazYq29Xd6tK3TbNO2ZpM7bqub/vxuVXx//T6f3P3+kLcNdmmbFenJvdhdpH1abW7jeDyibcMht7HV6iS337q5Tvtcu3aDtpUNcrMCYCW/1tENZjRKn1t0Xmtra8ntN69dx2g4TA7+Qb7GPwzgB+7+Q3cfAvg8gEcPsD8hxBFyEGe/AODHd/z90mybEOI1yEF+s6e+KvyD77NmdhHARQBotpoHOJwQ4iAc5M7+EoD77/j7DQCuvPpN7n7J3VfdfbXRPNB6oBDiABzE2b8J4EEze6OZNQG8B8BTh2OWEOKwuedbrbuPzewDAP4EU+ntCXf/XtinrjEYbifbWiU3pSaKQRmsfjq4tLLTS6+qA0BV8Z8anYX0SuwgWpVucMll8QRfmW4WwaWp+Spts0irCcuLfKV7d5uvPhfOx7HT4SvTTNMYjrntCJoWFtKr6gBgRSDLEPlqcWmBdrl5k1+zUSDLlcG9M1K92Gp8pAw1Gun5EUl8B/pe7e5fA/C1g+xDCDEf9ASdEJkgZxciE+TsQmSCnF2ITJCzC5EJc37KxakkRmJdAADjwSC5vd3m8klZc1mu0+GS1/LyMm3b3tlJbh+O+7RPa4FLXp2KS1dloCYNdrkcxoJyNtZv0z71hAeZVBUfx1EQAFaSqMMoIKTR4G2DIR/jyP56kjYyULXQCp70HO9y6S2SyiJYlF20v0hiY+jOLkQmyNmFyAQ5uxCZIGcXIhPk7EJkwlxX460o0CEr6KN+esUdAAoSFBKvSPKVzLIR5GMLAj+MrHR3unzFPQr8aFZB8E+Qc2tphafVapTpld0rP/kp7dNqcVWjCIKNLBgrlOlrU1Z87EfBWO1spwOoAKBZ8FX8iige0RxYDgKUhmNux2DI51ykarCglgFRoQBgaWkpuf1GlOOPtgghfq6QswuRCXJ2ITJBzi5EJsjZhcgEObsQmTBf6c0KVI10LrE6+NjpLqf77O6mA1MAYLfPAye2tjZpmwV1qGqSz2xc8+CIbpfnTovy5HWCAJoykOwm5PN76fRZ2ieaBlubXGpyku8OACoSCDNyPlaTQMo7fe40bWuCy001qyYUTLjRMLBxEgXCcCk4KinFpLeoIszCQlouLUi5K0B3diGyQc4uRCbI2YXIBDm7EJkgZxciE+TsQmTCgaQ3M3sRwBaACYCxu6/u0QOwdPTP4iLPx9ZupPuE+dHqHm2rgoin4YhHGoFE2UWRcu0OjyiLIv12dnlJqZ0+l3gWFtMRW3VQTmpnmx+rs8wj7Ho7PK8dSNTe0nI6WgsABoHUFMlQ7nw8mk1SsiuQZttRWauaX+uoHFkk2TEbWy1uBysZFZWZOgyd/Z+5+81D2I8Q4gjR13ghMuGgzu4A/tTMvmVmFw/DICHE0XDQr/Fvd/crZnYWwNfN7K/d/Rt3vmH2IXARAFpt/htECHG0HOjO7u5XZv9fB/AVAA8n3nPJ3VfdfbXR5AsYQoij5Z6d3cy6Zrb08msAvwHg+cMyTAhxuBzka/w5AF+ZJX1sAPjv7v6/og7uwIhEIQXKEPqkvFLhQdmfEZdWBiR6DQCqFo9SK5vpskCLRO4CAAsisiaT4KQDOS8qk7SxvpW2Y8Jlvn6QzHFpiZ/bqUUuy1mdlsrKKDIsyF/Z6/HruRNElK2cSI9VESW+JLYDQCeQiHvbfD5acfcRcUHOUQTDSLlnZ3f3HwL4lXvtL4SYL5LehMgEObsQmSBnFyIT5OxCZIKcXYhMmGvCScBpVM5gyKWhhVb6YZzuApfJJhXXLaL6ZQ1Siw4AfnojHe/TG/DEl92FZdrWrnhSyfGIR6K1g4STIMkvLZAbOxXXcSaBhLkYRPQNd9Py1TCI9CsDSbHdCa51IL2xs17octv7A37Oy8tcitzZ5vFgnXaXtjlJfjkJtLea1B2M0J1diEyQswuRCXJ2ITJBzi5EJsjZhciEua7GF0WBDllVnQz5CmhZpldp2XYA6ATBKQ2S8wsARkGEAct55xMewbG1ts7tcK4KNAu+z+4yt7+09CXdHfAgjbOneUBLP1gRHk/4PhtkrKKV7k6LqxMNuq4OFCQ3IACMx2kbNzZ4sEs/yE9XVelgKAAog9yGCFbPGyQop/QoWIfMjyBARnd2ITJBzi5EJsjZhcgEObsQmSBnFyIT5OxCZMLcpbeFhXQAwnqfB5OMx2nZwp2bH8lyQYUc9Ho8AIXtsx1IeRhxyWgy5CWqrOL9zp14PW37v1euJLefXuEBOSdPnqRtm7tcAuztcultRCSvKMMwP2NgUvPWOmjbJWW0otJKUVmxesLvj41AegvLRpEEjOMxlwdrprEFc1t3diEyQc4uRCbI2YXIBDm7EJkgZxciE+TsQmTCntKbmT0B4DcBXHf3X55tOwXgCwAeAPAigN9297W99uXutNSNBZFLo2Fagtjc5NJEucxzjFkQURZpFyxib9TjEtrpU1zWKhs8d1o14fscbqZLPAHA7lZaauqCS003rtygbes9Lq8VQZRa1U5Hh9VBLrwJkesAYDeIlmsWXGZlpbm6XZ4TbjMY32bFc+H1driNGxu8xBaLzKtIuTEAGA/53GHs587+hwAeedW2xwE87e4PAnh69rcQ4jXMns4+q7d++1WbHwXw5Oz1kwDedch2CSEOmXv9zX7O3a8CwOz/s4dnkhDiKDjyx2XN7CKAiwDQ7vDfeEKIo+Ve7+zXzOw8AMz+v87e6O6X3H3V3VerFl9wEEIcLffq7E8BeGz2+jEAXz0cc4QQR8V+pLfPAXgHgNNm9hKADwP4KIAvmtn7AfwIwG8d1JBIChn00rLFeMyljuGIy3KBUoMggAoo05+NJ5Z5wsZRUO6oHRjifS69/fRHP6ZtKyvnk9v72zzx5cbGJm3bHnEpcvkcnz7jIj2Qw6BUUyP45tcM2vqbPGJyeTkd7dcL5NIqKK9VkjkAAC1SpgwAalKWCwAKojo3gwjBCUlGGUnYezq7u7+XNP36Xn2FEK8d9ASdEJkgZxciE+TsQmSCnF2ITJCzC5EJc004CQATIkFEZbLKKi1RFWVQsy2QjDpkfwDQbgayC5FkPEgqubXDo53qkh/rRItH7fV2ueS49uN0wslGzSPK2h0+jgtt3rZy+gxtu3brWnK7RxkRRzwaMVCU0AiuZ6+XluUagbzWafMnPbe3NrgdkSwXRLANh+n5Mxhw+bjVTEffGdPxoDu7ENkgZxciE+TsQmSCnF2ITJCzC5EJcnYhMmGu0pt7jfEwLRt5GWgr5COp9iBqzPjn2G4gaZw5waPvFpfSbT/5SVpmAoBJxc9rEiUU7HDprdnhUXa3X/h+cnsRJHM8t8CTKC6eSidsBIBJMHuapKbfKBh7TAJZLqgE113k9m9tpZNHNio+9qMxj1ScjHibTfh8LIP5OBqmr814wseqapBzVq03IYScXYhMkLMLkQlydiEyQc4uRCbMdzW+rjHpp8sToeQrmVWwcsqog2Ry9YSvTO9sB2WXyErsOEpcF5zX2PjS6U6QQ+/0SR6A0m6lFQMvyLgD8GClu6y4jYMBD/IZDdPH80mQgy5KDujcjmEQGNQmikcjWB2PgnXGkZpQc/sLBLnhWEBUMB79XTK+wVzUnV2ITJCzC5EJcnYhMkHOLkQmyNmFyAQ5uxCZsJ/yT08A+E0A1939l2fbPgLgdwDcmL3tQ+7+tT2P5g4jARnjAZfDmJXNFje/6gRBCQ1eVidKdmZI73Nl5RTtc+Pmq0vb/38WloJgl8CO7hIP/DhFbNlZp7U3MR5x6Wp78xZtWznHJcB1Isu1grx7VZA/rR5zSWlnh9t/4fUXaBvj5o0btK3Z4DJwq+LXs9/nuevM03N/EpxzEeTdo3328Z4/BPBIYvsn3P2h2b+9HV0Icazs6ezu/g0A/PYkhPiZ4CC/2T9gZs+Z2RNmdvLQLBJCHAn36uyfAvAmAA8BuArgY+yNZnbRzC6b2eXxiD8qKYQ4Wu7J2d39mrtP3L0G8GkADwfvveTuq+6+GiXmF0IcLffk7GZ2/o4/3w3g+cMxRwhxVOxHevscgHcAOG1mLwH4MIB3mNlDmGa8ehHA7+7nYIUZmiSCrS54pJGTiKealJICgKoZyGsB4zEvQdRmJZmCCKrTZ07TtgLc/mabSyuTmkdeNcg43ndyhfZZ2+Gy3PoajwJcPLFM24pJehwXF5donwnJxQYAQYAguhWXInfW0znoWi1e1gpjfrBWyefV1sY6bRv2+TVjefkmzudVSSTMKIvfns7u7u9NbP7MXv2EEK8t9ASdEJkgZxciE+TsQmSCnF2ITJCzC5EJc33KxYoSVTtdTigIhkK/v5PcPhrzJIq7u1xCKwoun9S8G3Z7aYmkvcwlqPMXXkfbBrs8EqrX58kcF9tcNmq309u3bm3SPkG+SVhQ42njVlrWAoBhLy0rbo55n06QWLQRXLPednp+AMBGPy2HnTzJn/BuFXx819d4mMit22u0baEbHI+cd38UTMZQZEujO7sQmSBnFyIT5OxCZIKcXYhMkLMLkQlydiEyYb4B5kWBsp2Oetru8SR/RTMt47Q7gflBsr5mEFc/CSLYdknk0u01LrlYxZMoLrT5sTY2ucRz/ux9tO3Bf/z65Pbnv8X319viY9UfcYlnNObyYIvUuNsKZLIxuc4AYM7HcafHI/OKIj3GVvOxryou842iyLygnlsZ1G1jAZrDIPoOwbEYurMLkQlydiEyQc4uRCbI2YXIBDm7EJkw53SvhglZlWwt8Dxi7W565bFT8c+qtSt8pRhRSusg9qBBFlSHQ55fbLDFA1A6ZZe2jUleMgDY2eHndmIxvbTb7vAgE9vkAUXjAR+rosHbuifS+fpuXOWBMCcWeUDR7g63cTQMchG20ue9tcPtWOjyMk7jYBW8DpQcDzytaenG8XY0h8mxSL5GQHd2IbJBzi5EJsjZhcgEObsQmSBnFyIT5OxCZMJ+yj/dD+CPALwO02xll9z9k2Z2CsAXADyAaQmo33Z3HhECAAY0SGDI7jaXT0qih7UaPGCh2+ayVjEMkq4FSeiKKq29LS1wySgqQ9Uqg7JRK6do20KbS0O9fj+5fafHpatGMI4NHveBhQUu59135kRy+/ptHpDjQTksK7nkNZzw6+mevp6l8ets4CddR0EyRSDLFfx4TuS8shHsj5QpY6XSgP3d2ccAft/dfxHA2wD8npm9GcDjAJ529wcBPD37WwjxGmVPZ3f3q+7+7dnrLQAvALgA4FEAT87e9iSAdx2VkUKIg3NXv9nN7AEAbwHwDIBz7n4VmH4gADh72MYJIQ6PfTu7mS0C+BKAD7p7kIT8H/S7aGaXzezysD+4FxuFEIfAvpzdzCpMHf2z7v7l2eZrZnZ+1n4eQLLIt7tfcvdVd19tBsUNhBBHy57ObmaGaT32F9z943c0PQXgsdnrxwB89fDNE0IcFvuJens7gPcB+K6ZPTvb9iEAHwXwRTN7P4AfAfitvXZk7ijHaWmoHUQMjTfTMkN/xCPDxiMux3SCWlMelNVh4kmzySWo5eV0zj0AQCD/nFzhcl4zsL+3lS4pVTsfj0aD769RcTlsEuRx29xIy1dFUFrpzNkz3I4GH+Mrt/+StlXNdD2sssMltKEF0XzL6fJlANANouWGI54nr7eVbmsF34T7vUA+Juzp7O7+5+DZ7X79ro8ohDgW9ASdEJkgZxciE+TsQmSCnF2ITJCzC5EJ8004WU/gu+kkesWIRxo5iWra2eVP5JWBHNZp8+SWk0Ci2hykI8caQTmpuub7qydcOrwdJKpcCWS5wtLCyalTJ2mf4ZDLjcPgWcntPpeoNsv0tekscHlqfXOdtk2CaK4ySKZZEIltEETYRTRq3s/HQdSecfsXF9Pzce1WWqae7TFoS6M7uxCZIGcXIhPk7EJkgpxdiEyQswuRCXJ2ITJhvtKbOzBOSzJVkKyvu5CWjSaB+jBwLmv1dnnyxShBZLebTmJZlKQIHOIouk4ziABb5vJau8P73b6dzvlZBgkbo8SRbwii9v76xb+jbe2FdLTZaMDrl+0O+XWZ8GEEokSPRPIKcn2itkAuJQks99pnpJSx+dNq87m4s50eq4MmnBRC/BwgZxciE+TsQmSCnF2ITJCzC5EJc12Nd3eMRulgge4yD04ZjdIr+HXBV8EHQZBJx3i/yYSvtk5IXrvBhAfxLC/wMlQngpXuVnBuTsYQAMakLFCrxVfw2+30yjkAbJGxB4BRzVfPrZm2cTkIhBn2+LF6m3wVf3mJ77Nqp5WGshWVk+JzZ3s7neMPAC6cfR3v1+NBPkNSsivKbXgv6M4uRCbI2YXIBDm7EJkgZxciE+TsQmSCnF2ITNhTejOz+wH8EYDXAagBXHL3T5rZRwD8DoAbs7d+yN2/tsfOgEb64f664A/wj+u0tOXggQKNIDilGZQSGgYlpViutuGES2FVUNaqcXKFtk0Cea1s8HNrtdIymhVcHuwucult/dYWbbv/AV6uqSjTY9UNgm4Q5P/rX+flkxaXT9C2FhmrosGvS7vFx3fc4vOj2eLn1q75GA/66TGOZGBWsstIDkJgfzr7GMDvu/u3zWwJwLfM7Ouztk+4+3/exz6EEMfMfmq9XQVwdfZ6y8xeAHDhqA0TQhwud/Wb3cweAPAWAM/MNn3AzJ4zsyfMjOcqFkIcO/t2djNbBPAlAB90900AnwLwJgAPYXrn/xjpd9HMLpvZ5eGQ/24UQhwt+3J2M6swdfTPuvuXAcDdr7n7xN1rAJ8G8HCqr7tfcvdVd19tBllghBBHy57ObtPlvc8AeMHdP37H9vN3vO3dAJ4/fPOEEIfFflbj3w7gfQC+a2bPzrZ9CMB7zewhAA7gRQC/u9eOHMCQqCtFyaPeWq30N4LhgMsg7SDKq9MJorxu8egqq9KSTDvKgdbnkWFjko8PAMqKfw6Phrws0Eo7HQG2FuR32wmi15bOLtK2asClJlYlaTDkEpoXXGq67+wp2jYK5gHqtAQ4CkqHVW1+Pc24jVXFv7kO1risCL/74NOykT6vQHnb12r8nyOdLi/W1IUQryn0BJ0QmSBnFyIT5OxCZIKcXYhMkLMLkQlzTThZu2NANJmiweWwBtJ9IsnFgjI4ozGPKGu2uWTHygw1g9o+neBBojKoF+SB9La9wSPRqkla4qmdn/OPfnqTtp18/WnaNuxzGWqwk5bYrBEk9AxqPDWCSD+r+ViNybUejvnc8UBKHQy4dLi7y2XbKAqTJQmtmtwnat9Jbo/KjenOLkQmyNmFyAQ5uxCZIGcXIhPk7EJkgpxdiEyYq/RWFAXaC+nots1eWkoAeFRZk+wLAMyiBJY8AqlFosYAYDBKJ9+oA5mv1eW13oI4qLDuWZSIsLa0jaNAalpe4okvfcynyCBItDlA2saTHX7NVoLrub3B58dGUI9uOEy3DQP5tdXldpw6yaPv+qRmGzCtc8hgNo5IbUGAS3lB0Jvu7ELkgpxdiEyQswuRCXJ2ITJBzi5EJsjZhciEuUpvZoaK1KjiwgQwIXpCL5BcFpo8GWJ3aYm27Q65JMOiqyakFh0A9Aa8rQpqg0W13qJ6Xq1uOmqvGnM7ag8iyiZ8ivT6d1/3zEkCSABot3mE4E4gN5akrty0LT1WkwGXtSLJq9vhUZG9bZ6404PIvJpEgo5GwTkXxI5gbujOLkQmyNmFyAQ5uxCZIGcXIhPk7EJkwp6r8WbWBvANAK3Z+//Y3T9sZm8E8HkApwB8G8D73D2owzN9SL/h6dXCRpCPzUjISJRvyxp8f0GqM7jxIWFBPA5+2v0gZxm2eHAHosCVBb4ivEUCaGoy7gDQ7welkIIp4kFAUc0GOcitx/LFAcCY1ZMCcPoMD07pDtJKw+Cla7RPzRfBQxuHQYmtqsGDaxa66VxzdMUdwPoav2aM/dzZBwB+zd1/BdPyzI+Y2dsA/AGAT7j7gwDWALz/ro8uhJgbezq7T3k5bWY1++cAfg3AH8+2PwngXUdioRDiUNhvffZyVsH1OoCvA/hbAOvuf5+f+CUAF47GRCHEYbAvZ3f3ibs/BOANAB4G8Iupt6X6mtlFM7tsZpeHwdNkQoij5a5W4919HcD/BvA2ACtmf7+a9QYAV0ifS+6+6u6rTVJnXQhx9Ozp7GZ2xsxWZq87AP45gBcA/BmAfzV722MAvnpURgohDs5+AmHOA3jSzEpMPxy+6O7/08z+CsDnzew/AvhLAJ/Za0cFDAtM2grkMCM56LzigSR1kIMuyj82qfmQFEVaxnHjgSRFk8snVcWPVZa8rSYlngBgfT2dB62ouI2ddpDLL7gdNKNrRqQ3C7KkDQLNy5p8PDpBcMqttY3k9oUOzw3YCqTNyYRLqVGJKliUcZC18T5RrjnGns7u7s8BeEti+w8x/f0uhPgZQE/QCZEJcnYhMkHOLkQmyNmFyAQ5uxCZYFFZmkM/mNkNAH83+/M0gJtzOzhHdrwS2fFKftbs+EfufibVMFdnf8WBzS67++qxHFx2yI4M7dDXeCEyQc4uRCYcp7NfOsZj34nseCWy45X83NhxbL/ZhRDzRV/jhciEY3F2M3vEzP6Pmf3AzB4/DhtmdrxoZt81s2fN7PIcj/uEmV03s+fv2HbKzL5uZt+f/X/ymOz4iJn9ZDYmz5rZO+dgx/1m9mdm9oKZfc/M/s1s+1zHJLBjrmNiZm0z+wsz+87Mjv8w2/5GM3tmNh5fMDMe9pnC3ef6D0CJaVqrXwDQBPAdAG+etx0zW14EcPoYjvurAN4K4Pk7tv0nAI/PXj8O4A+OyY6PAPi3cx6P8wDeOnu9BOBvALx53mMS2DHXMcE0gnVx9roC8AymCWO+COA9s+3/BcC/vpv9Hsed/WEAP3D3H/o09fTnATx6DHYcG+7+DQC3X7X5UUwTdwJzSuBJ7Jg77n7V3b89e72FaXKUC5jzmAR2zBWfcuhJXo/D2S8A+PEdfx9nskoH8Kdm9i0zu3hMNrzMOXe/CkwnHYCzx2jLB8zsudnX/CP/OXEnZvYApvkTnsExjsmr7ADmPCZHkeT1OJw9lWTjuCSBt7v7WwH8SwC/Z2a/ekx2vJb4FIA3YVoj4CqAj83rwGa2COBLAD7o7pvzOu4+7Jj7mPgBkrwyjsPZXwJw/x1/02SVR427X5n9fx3AV3C8mXeumdl5AJj9f/04jHD3a7OJVgP4NOY0JmZWYepgn3X3L882z31MUnYc15jMjn3XSV4Zx+Hs3wTw4GxlsQngPQCemrcRZtY1s6WXXwP4DQDPx72OlKcwTdwJHGMCz5eda8a7MYcxMTPDNIfhC+7+8Tua5jomzI55j8mRJXmd1wrjq1Yb34npSuffAvh3x2TDL2CqBHwHwPfmaQeAz2H6dXCE6Ted9wO4D8DTAL4/+//UMdnx3wB8F8BzmDrb+TnY8U8x/Ur6HIBnZ//eOe8xCeyY65gA+CeYJnF9DtMPln9/x5z9CwA/APA/ALTuZr96gk6ITNATdEJkgpxdiEyQswuRCXJ2ITJBzi5EJsjZhcgEObsQmSBnFyIT/h+HMxDVscHrzwAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "img, label = cifar[10]\n",
    "classes = ['plane','car','bird','cat','deer','dog','frog','horse','ship','truck']\n",
    "print(classes[label])\n",
    "plt.imshow(img)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "car\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x2434e3161c8>"
      ]
     },
     "execution_count": 57,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "img, label = cifar[5]\n",
    "classes = ['plane','car','bird','cat','deer','dog','frog','horse','ship','truck']\n",
    "print(classes[label])\n",
    "plt.imshow(img)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Sample Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch.nn as nn "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [],
   "source": [
    "linear = nn.Linear(in_features=4096, out_features=10) # 1 dim tensor \n",
    "conv = nn.Conv2d(in_channels=3, out_channels=9, kernel_size=3, stride=1, padding=1) # 2D to img 3D to video\n",
    "relu = nn.ReLU(False)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "torch.Size([10])"
      ]
     },
     "execution_count": 60,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "x = torch.randn(4096)\n",
    "y = linear(x)\n",
    "y.size()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(torch.Size([1, 9, 7, 7]), torch.Size([1, 3, 7, 7]))"
      ]
     },
     "execution_count": 61,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Image\n",
    "\n",
    "x = torch.randn(1,3,7,7) # 1 img - 3 ch - 7x7 px \n",
    "y = conv(x)\n",
    "z = relu(x)\n",
    "y.size(), z.size()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "weight\n",
      "torch.Size([10, 4096])\n",
      "bias\n",
      "torch.Size([10])\n"
     ]
    }
   ],
   "source": [
    "for name, param in linear.named_parameters():\n",
    "    print(name)\n",
    "    print(param.size())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}