Cora.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "from models import GAT, GCN\n",
    "from utils import Cora\n",
    "import torch\n",
    "import torch.nn as nn\n",
    "from torch.optim import SGD, Adam\n",
    "from torch.utils.data import DataLoader\n",
    "from sklearn.metrics import accuracy_score\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "CLASS = 7\n",
    "EPOCH = 1000\n",
    "TEST = 100\n",
    "LR = 0.001"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "adj, features, labels, idx_train, idx_val, idx_test = Cora.load_data()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x7f61683b4910>"
      ]
     },
     "execution_count": 3,
     "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": [
    "plt.imshow(adj[0][idx_train, :len(idx_train)].cpu(), cmap ='gray')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = GCN(features.shape[-1], 1024, CLASS, 0.3, True).cuda()\n",
    "# model = GAT(features.shape[-1], 16, CLASS, 0.3, True).cuda()\n",
    "loss_fn = nn.CrossEntropyLoss()\n",
    "optim = Adam(model.parameters(), lr=LR)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[  1/1000]  loss: 0.1833   precision: 97.14%\n",
      "[  2/1000]  loss: 0.1774   precision: 99.29%\n",
      "[  3/1000]  loss: 0.1726   precision: 99.29%\n",
      "[  4/1000]  loss: 0.1520   precision: 99.29%\n",
      "[  5/1000]  loss: 0.1507   precision: 100.00%\n",
      "[  6/1000]  loss: 0.1471   precision: 98.57%\n",
      "[  7/1000]  loss: 0.1423   precision: 99.29%\n",
      "[  8/1000]  loss: 0.1406   precision: 99.29%\n",
      "[  9/1000]  loss: 0.1286   precision: 99.29%\n",
      "[ 10/1000]  loss: 0.1342   precision: 99.29%\n",
      "[ 11/1000]  loss: 0.1235   precision: 99.29%\n",
      "[ 12/1000]  loss: 0.1145   precision: 100.00%\n",
      "[ 13/1000]  loss: 0.1237   precision: 100.00%\n",
      "[ 14/1000]  loss: 0.1155   precision: 100.00%\n",
      "[ 15/1000]  loss: 0.1099   precision: 100.00%\n",
      "[ 16/1000]  loss: 0.1098   precision: 100.00%\n",
      "[ 17/1000]  loss: 0.1092   precision: 100.00%\n",
      "[ 18/1000]  loss: 0.1053   precision: 100.00%\n",
      "[ 19/1000]  loss: 0.1049   precision: 100.00%\n",
      "[ 20/1000]  loss: 0.1008   precision: 100.00%\n",
      "[ 21/1000]  loss: 0.1005   precision: 100.00%\n",
      "[ 22/1000]  loss: 0.1013   precision: 100.00%\n",
      "[ 23/1000]  loss: 0.0984   precision: 100.00%\n",
      "[ 24/1000]  loss: 0.0954   precision: 100.00%\n",
      "[ 25/1000]  loss: 0.0987   precision: 100.00%\n",
      "[ 26/1000]  loss: 0.0950   precision: 100.00%\n",
      "[ 27/1000]  loss: 0.0949   precision: 100.00%\n",
      "[ 28/1000]  loss: 0.0934   precision: 100.00%\n",
      "[ 29/1000]  loss: 0.0939   precision: 100.00%\n",
      "[ 30/1000]  loss: 0.0933   precision: 100.00%\n",
      "[ 31/1000]  loss: 0.0921   precision: 100.00%\n",
      "[ 32/1000]  loss: 0.0917   precision: 100.00%\n",
      "[ 33/1000]  loss: 0.0889   precision: 100.00%\n",
      "[ 34/1000]  loss: 0.0881   precision: 100.00%\n",
      "[ 35/1000]  loss: 0.0893   precision: 100.00%\n",
      "[ 36/1000]  loss: 0.0873   precision: 100.00%\n",
      "[ 37/1000]  loss: 0.0870   precision: 100.00%\n",
      "[ 38/1000]  loss: 0.0875   precision: 100.00%\n",
      "[ 39/1000]  loss: 0.0859   precision: 100.00%\n",
      "[ 40/1000]  loss: 0.0862   precision: 100.00%\n",
      "[ 41/1000]  loss: 0.0854   precision: 100.00%\n",
      "[ 42/1000]  loss: 0.0846   precision: 100.00%\n",
      "[ 43/1000]  loss: 0.0854   precision: 100.00%\n",
      "[ 44/1000]  loss: 0.0837   precision: 100.00%\n",
      "[ 45/1000]  loss: 0.0818   precision: 100.00%\n",
      "[ 46/1000]  loss: 0.0825   precision: 100.00%\n",
      "[ 47/1000]  loss: 0.0823   precision: 100.00%\n",
      "[ 48/1000]  loss: 0.0821   precision: 100.00%\n",
      "[ 49/1000]  loss: 0.0830   precision: 100.00%\n",
      "[ 50/1000]  loss: 0.0810   precision: 100.00%\n",
      "[ 51/1000]  loss: 0.0794   precision: 100.00%\n",
      "[ 52/1000]  loss: 0.0810   precision: 100.00%\n",
      "[ 53/1000]  loss: 0.0795   precision: 100.00%\n",
      "[ 54/1000]  loss: 0.0810   precision: 100.00%\n",
      "[ 55/1000]  loss: 0.0800   precision: 100.00%\n",
      "[ 56/1000]  loss: 0.0785   precision: 100.00%\n",
      "[ 57/1000]  loss: 0.0794   precision: 100.00%\n",
      "[ 58/1000]  loss: 0.0791   precision: 100.00%\n",
      "[ 59/1000]  loss: 0.0799   precision: 100.00%\n",
      "[ 60/1000]  loss: 0.0785   precision: 100.00%\n",
      "[ 61/1000]  loss: 0.0794   precision: 100.00%\n",
      "[ 62/1000]  loss: 0.0785   precision: 100.00%\n",
      "[ 63/1000]  loss: 0.0774   precision: 100.00%\n",
      "[ 64/1000]  loss: 0.0771   precision: 100.00%\n",
      "[ 65/1000]  loss: 0.0777   precision: 100.00%\n",
      "[ 66/1000]  loss: 0.0764   precision: 100.00%\n",
      "[ 67/1000]  loss: 0.0757   precision: 100.00%\n",
      "[ 68/1000]  loss: 0.0757   precision: 100.00%\n",
      "[ 69/1000]  loss: 0.0766   precision: 100.00%\n",
      "[ 70/1000]  loss: 0.0750   precision: 100.00%\n",
      "[ 71/1000]  loss: 0.0755   precision: 100.00%\n",
      "[ 72/1000]  loss: 0.0760   precision: 100.00%\n",
      "[ 73/1000]  loss: 0.0744   precision: 100.00%\n",
      "[ 74/1000]  loss: 0.0744   precision: 100.00%\n",
      "[ 75/1000]  loss: 0.0740   precision: 100.00%\n",
      "[ 76/1000]  loss: 0.0751   precision: 100.00%\n",
      "[ 77/1000]  loss: 0.0742   precision: 100.00%\n",
      "[ 78/1000]  loss: 0.0747   precision: 100.00%\n",
      "[ 79/1000]  loss: 0.0738   precision: 100.00%\n",
      "[ 80/1000]  loss: 0.0748   precision: 100.00%\n",
      "[ 81/1000]  loss: 0.0731   precision: 100.00%\n",
      "[ 82/1000]  loss: 0.0739   precision: 100.00%\n",
      "[ 83/1000]  loss: 0.0737   precision: 100.00%\n",
      "[ 84/1000]  loss: 0.0736   precision: 100.00%\n",
      "[ 85/1000]  loss: 0.0719   precision: 100.00%\n",
      "[ 86/1000]  loss: 0.0737   precision: 100.00%\n",
      "[ 87/1000]  loss: 0.0727   precision: 100.00%\n",
      "[ 88/1000]  loss: 0.0724   precision: 100.00%\n",
      "[ 89/1000]  loss: 0.0730   precision: 100.00%\n",
      "[ 90/1000]  loss: 0.0718   precision: 100.00%\n",
      "[ 91/1000]  loss: 0.0723   precision: 100.00%\n",
      "[ 92/1000]  loss: 0.0722   precision: 100.00%\n",
      "[ 93/1000]  loss: 0.0717   precision: 100.00%\n",
      "[ 94/1000]  loss: 0.0728   precision: 100.00%\n",
      "[ 95/1000]  loss: 0.0719   precision: 100.00%\n",
      "[ 96/1000]  loss: 0.0715   precision: 100.00%\n",
      "[ 97/1000]  loss: 0.0715   precision: 100.00%\n",
      "[ 98/1000]  loss: 0.0704   precision: 100.00%\n",
      "[ 99/1000]  loss: 0.0698   precision: 100.00%\n",
      "[100/1000]  loss: 0.0702   precision: 100.00%\n",
      "=============================================\n",
      "[Testing]  loss: 0.6586   precision: 79.60%\n",
      "=============================================\n",
      "[101/1000]  loss: 0.0707   precision: 100.00%\n",
      "[102/1000]  loss: 0.0720   precision: 100.00%\n",
      "[103/1000]  loss: 0.0699   precision: 100.00%\n",
      "[104/1000]  loss: 0.0696   precision: 100.00%\n",
      "[105/1000]  loss: 0.0699   precision: 100.00%\n",
      "[106/1000]  loss: 0.0685   precision: 100.00%\n",
      "[107/1000]  loss: 0.0704   precision: 100.00%\n",
      "[108/1000]  loss: 0.0690   precision: 100.00%\n",
      "[109/1000]  loss: 0.0691   precision: 100.00%\n",
      "[110/1000]  loss: 0.0687   precision: 100.00%\n",
      "[111/1000]  loss: 0.0696   precision: 100.00%\n",
      "[112/1000]  loss: 0.0700   precision: 100.00%\n",
      "[113/1000]  loss: 0.0688   precision: 100.00%\n",
      "[114/1000]  loss: 0.0681   precision: 100.00%\n",
      "[115/1000]  loss: 0.0683   precision: 100.00%\n",
      "[116/1000]  loss: 0.0680   precision: 100.00%\n",
      "[117/1000]  loss: 0.0676   precision: 100.00%\n",
      "[118/1000]  loss: 0.0690   precision: 100.00%\n",
      "[119/1000]  loss: 0.0680   precision: 100.00%\n",
      "[120/1000]  loss: 0.0675   precision: 100.00%\n",
      "[121/1000]  loss: 0.0674   precision: 100.00%\n",
      "[122/1000]  loss: 0.0681   precision: 100.00%\n",
      "[123/1000]  loss: 0.0679   precision: 100.00%\n",
      "[124/1000]  loss: 0.0666   precision: 100.00%\n",
      "[125/1000]  loss: 0.0671   precision: 100.00%\n",
      "[126/1000]  loss: 0.0673   precision: 100.00%\n",
      "[127/1000]  loss: 0.0662   precision: 100.00%\n",
      "[128/1000]  loss: 0.0660   precision: 100.00%\n",
      "[129/1000]  loss: 0.0665   precision: 100.00%\n",
      "[130/1000]  loss: 0.0654   precision: 100.00%\n",
      "[131/1000]  loss: 0.0655   precision: 100.00%\n",
      "[132/1000]  loss: 0.0664   precision: 100.00%\n",
      "[133/1000]  loss: 0.0676   precision: 100.00%\n",
      "[134/1000]  loss: 0.0657   precision: 100.00%\n",
      "[135/1000]  loss: 0.0652   precision: 100.00%\n",
      "[136/1000]  loss: 0.0654   precision: 100.00%\n",
      "[137/1000]  loss: 0.0647   precision: 100.00%\n",
      "[138/1000]  loss: 0.0650   precision: 100.00%\n",
      "[139/1000]  loss: 0.0649   precision: 100.00%\n",
      "[140/1000]  loss: 0.0655   precision: 100.00%\n",
      "[141/1000]  loss: 0.0650   precision: 100.00%\n",
      "[142/1000]  loss: 0.0645   precision: 100.00%\n",
      "[143/1000]  loss: 0.0645   precision: 100.00%\n",
      "[144/1000]  loss: 0.0637   precision: 100.00%\n",
      "[145/1000]  loss: 0.0650   precision: 100.00%\n",
      "[146/1000]  loss: 0.0648   precision: 100.00%\n",
      "[147/1000]  loss: 0.0639   precision: 100.00%\n",
      "[148/1000]  loss: 0.0654   precision: 100.00%\n",
      "[149/1000]  loss: 0.0639   precision: 100.00%\n",
      "[150/1000]  loss: 0.0648   precision: 100.00%\n",
      "[151/1000]  loss: 0.0638   precision: 100.00%\n",
      "[152/1000]  loss: 0.0633   precision: 100.00%\n",
      "[153/1000]  loss: 0.0636   precision: 100.00%\n",
      "[154/1000]  loss: 0.0637   precision: 100.00%\n",
      "[155/1000]  loss: 0.0626   precision: 100.00%\n",
      "[156/1000]  loss: 0.0634   precision: 100.00%\n",
      "[157/1000]  loss: 0.0622   precision: 100.00%\n",
      "[158/1000]  loss: 0.0628   precision: 100.00%\n",
      "[159/1000]  loss: 0.0635   precision: 100.00%\n",
      "[160/1000]  loss: 0.0631   precision: 100.00%\n",
      "[161/1000]  loss: 0.0633   precision: 100.00%\n",
      "[162/1000]  loss: 0.0619   precision: 100.00%\n",
      "[163/1000]  loss: 0.0628   precision: 100.00%\n",
      "[164/1000]  loss: 0.0622   precision: 100.00%\n",
      "[165/1000]  loss: 0.0618   precision: 100.00%\n",
      "[166/1000]  loss: 0.0627   precision: 100.00%\n",
      "[167/1000]  loss: 0.0611   precision: 100.00%\n",
      "[168/1000]  loss: 0.0615   precision: 100.00%\n",
      "[169/1000]  loss: 0.0617   precision: 100.00%\n",
      "[170/1000]  loss: 0.0618   precision: 100.00%\n",
      "[171/1000]  loss: 0.0611   precision: 100.00%\n",
      "[172/1000]  loss: 0.0620   precision: 100.00%\n",
      "[173/1000]  loss: 0.0614   precision: 100.00%\n",
      "[174/1000]  loss: 0.0609   precision: 100.00%\n",
      "[175/1000]  loss: 0.0611   precision: 100.00%\n",
      "[176/1000]  loss: 0.0614   precision: 100.00%\n",
      "[177/1000]  loss: 0.0602   precision: 100.00%\n",
      "[178/1000]  loss: 0.0609   precision: 100.00%\n",
      "[179/1000]  loss: 0.0606   precision: 100.00%\n",
      "[180/1000]  loss: 0.0599   precision: 100.00%\n",
      "[181/1000]  loss: 0.0602   precision: 100.00%\n",
      "[182/1000]  loss: 0.0608   precision: 100.00%\n",
      "[183/1000]  loss: 0.0600   precision: 100.00%\n",
      "[184/1000]  loss: 0.0599   precision: 100.00%\n",
      "[185/1000]  loss: 0.0598   precision: 100.00%\n",
      "[186/1000]  loss: 0.0604   precision: 100.00%\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[187/1000]  loss: 0.0591   precision: 100.00%\n",
      "[188/1000]  loss: 0.0599   precision: 100.00%\n",
      "[189/1000]  loss: 0.0593   precision: 100.00%\n",
      "[190/1000]  loss: 0.0588   precision: 100.00%\n",
      "[191/1000]  loss: 0.0591   precision: 100.00%\n",
      "[192/1000]  loss: 0.0596   precision: 100.00%\n",
      "[193/1000]  loss: 0.0592   precision: 100.00%\n",
      "[194/1000]  loss: 0.0593   precision: 100.00%\n",
      "[195/1000]  loss: 0.0597   precision: 100.00%\n",
      "[196/1000]  loss: 0.0593   precision: 100.00%\n",
      "[197/1000]  loss: 0.0588   precision: 100.00%\n",
      "[198/1000]  loss: 0.0588   precision: 100.00%\n",
      "[199/1000]  loss: 0.0584   precision: 100.00%\n",
      "[200/1000]  loss: 0.0582   precision: 100.00%\n",
      "=============================================\n",
      "[Testing]  loss: 0.6709   precision: 78.60%\n",
      "=============================================\n",
      "[201/1000]  loss: 0.0590   precision: 100.00%\n",
      "[202/1000]  loss: 0.0584   precision: 100.00%\n",
      "[203/1000]  loss: 0.0580   precision: 100.00%\n",
      "[204/1000]  loss: 0.0593   precision: 100.00%\n",
      "[205/1000]  loss: 0.0580   precision: 100.00%\n",
      "[206/1000]  loss: 0.0585   precision: 100.00%\n",
      "[207/1000]  loss: 0.0581   precision: 100.00%\n",
      "[208/1000]  loss: 0.0580   precision: 100.00%\n",
      "[209/1000]  loss: 0.0580   precision: 100.00%\n",
      "[210/1000]  loss: 0.0577   precision: 100.00%\n",
      "[211/1000]  loss: 0.0573   precision: 100.00%\n",
      "[212/1000]  loss: 0.0578   precision: 100.00%\n",
      "[213/1000]  loss: 0.0575   precision: 100.00%\n",
      "[214/1000]  loss: 0.0579   precision: 100.00%\n",
      "[215/1000]  loss: 0.0577   precision: 100.00%\n",
      "[216/1000]  loss: 0.0571   precision: 100.00%\n",
      "[217/1000]  loss: 0.0574   precision: 100.00%\n",
      "[218/1000]  loss: 0.0567   precision: 100.00%\n",
      "[219/1000]  loss: 0.0573   precision: 100.00%\n",
      "[220/1000]  loss: 0.0564   precision: 100.00%\n",
      "[221/1000]  loss: 0.0568   precision: 100.00%\n",
      "[222/1000]  loss: 0.0568   precision: 100.00%\n",
      "[223/1000]  loss: 0.0568   precision: 100.00%\n",
      "[224/1000]  loss: 0.0563   precision: 100.00%\n",
      "[225/1000]  loss: 0.0569   precision: 100.00%\n",
      "[226/1000]  loss: 0.0570   precision: 100.00%\n",
      "[227/1000]  loss: 0.0563   precision: 100.00%\n",
      "[228/1000]  loss: 0.0563   precision: 100.00%\n",
      "[229/1000]  loss: 0.0563   precision: 100.00%\n",
      "[230/1000]  loss: 0.0565   precision: 100.00%\n",
      "[231/1000]  loss: 0.0565   precision: 100.00%\n",
      "[232/1000]  loss: 0.0558   precision: 100.00%\n",
      "[233/1000]  loss: 0.0555   precision: 100.00%\n",
      "[234/1000]  loss: 0.0561   precision: 100.00%\n",
      "[235/1000]  loss: 0.0557   precision: 100.00%\n",
      "[236/1000]  loss: 0.0556   precision: 100.00%\n",
      "[237/1000]  loss: 0.0559   precision: 100.00%\n",
      "[238/1000]  loss: 0.0555   precision: 100.00%\n",
      "[239/1000]  loss: 0.0554   precision: 100.00%\n",
      "[240/1000]  loss: 0.0555   precision: 100.00%\n",
      "[241/1000]  loss: 0.0551   precision: 100.00%\n",
      "[242/1000]  loss: 0.0560   precision: 100.00%\n",
      "[243/1000]  loss: 0.0552   precision: 100.00%\n",
      "[244/1000]  loss: 0.0557   precision: 100.00%\n",
      "[245/1000]  loss: 0.0553   precision: 100.00%\n",
      "[246/1000]  loss: 0.0547   precision: 100.00%\n",
      "[247/1000]  loss: 0.0549   precision: 100.00%\n",
      "[248/1000]  loss: 0.0545   precision: 100.00%\n",
      "[249/1000]  loss: 0.0548   precision: 100.00%\n",
      "[250/1000]  loss: 0.0543   precision: 100.00%\n",
      "[251/1000]  loss: 0.0549   precision: 100.00%\n",
      "[252/1000]  loss: 0.0547   precision: 100.00%\n",
      "[253/1000]  loss: 0.0539   precision: 100.00%\n",
      "[254/1000]  loss: 0.0544   precision: 100.00%\n",
      "[255/1000]  loss: 0.0542   precision: 100.00%\n",
      "[256/1000]  loss: 0.0544   precision: 100.00%\n",
      "[257/1000]  loss: 0.0539   precision: 100.00%\n",
      "[258/1000]  loss: 0.0540   precision: 100.00%\n",
      "[259/1000]  loss: 0.0539   precision: 100.00%\n",
      "[260/1000]  loss: 0.0537   precision: 100.00%\n",
      "[261/1000]  loss: 0.0537   precision: 100.00%\n",
      "[262/1000]  loss: 0.0535   precision: 100.00%\n",
      "[263/1000]  loss: 0.0544   precision: 100.00%\n",
      "[264/1000]  loss: 0.0534   precision: 100.00%\n",
      "[265/1000]  loss: 0.0536   precision: 100.00%\n",
      "[266/1000]  loss: 0.0531   precision: 100.00%\n",
      "[267/1000]  loss: 0.0532   precision: 100.00%\n",
      "[268/1000]  loss: 0.0528   precision: 100.00%\n",
      "[269/1000]  loss: 0.0534   precision: 100.00%\n",
      "[270/1000]  loss: 0.0529   precision: 100.00%\n",
      "[271/1000]  loss: 0.0529   precision: 100.00%\n",
      "[272/1000]  loss: 0.0534   precision: 100.00%\n",
      "[273/1000]  loss: 0.0523   precision: 100.00%\n",
      "[274/1000]  loss: 0.0530   precision: 100.00%\n",
      "[275/1000]  loss: 0.0530   precision: 100.00%\n",
      "[276/1000]  loss: 0.0524   precision: 100.00%\n",
      "[277/1000]  loss: 0.0523   precision: 100.00%\n",
      "[278/1000]  loss: 0.0526   precision: 100.00%\n",
      "[279/1000]  loss: 0.0524   precision: 100.00%\n",
      "[280/1000]  loss: 0.0528   precision: 100.00%\n",
      "[281/1000]  loss: 0.0526   precision: 100.00%\n",
      "[282/1000]  loss: 0.0522   precision: 100.00%\n",
      "[283/1000]  loss: 0.0521   precision: 100.00%\n",
      "[284/1000]  loss: 0.0525   precision: 100.00%\n",
      "[285/1000]  loss: 0.0520   precision: 100.00%\n",
      "[286/1000]  loss: 0.0526   precision: 100.00%\n",
      "[287/1000]  loss: 0.0520   precision: 100.00%\n",
      "[288/1000]  loss: 0.0519   precision: 100.00%\n",
      "[289/1000]  loss: 0.0514   precision: 100.00%\n",
      "[290/1000]  loss: 0.0518   precision: 100.00%\n",
      "[291/1000]  loss: 0.0514   precision: 100.00%\n",
      "[292/1000]  loss: 0.0518   precision: 100.00%\n",
      "[293/1000]  loss: 0.0512   precision: 100.00%\n",
      "[294/1000]  loss: 0.0515   precision: 100.00%\n",
      "[295/1000]  loss: 0.0513   precision: 100.00%\n",
      "[296/1000]  loss: 0.0513   precision: 100.00%\n",
      "[297/1000]  loss: 0.0513   precision: 100.00%\n",
      "[298/1000]  loss: 0.0510   precision: 100.00%\n",
      "[299/1000]  loss: 0.0514   precision: 100.00%\n",
      "[300/1000]  loss: 0.0515   precision: 100.00%\n",
      "=============================================\n",
      "[Testing]  loss: 0.6776   precision: 79.00%\n",
      "=============================================\n",
      "[301/1000]  loss: 0.0510   precision: 100.00%\n",
      "[302/1000]  loss: 0.0505   precision: 100.00%\n",
      "[303/1000]  loss: 0.0507   precision: 100.00%\n",
      "[304/1000]  loss: 0.0509   precision: 100.00%\n",
      "[305/1000]  loss: 0.0511   precision: 100.00%\n",
      "[306/1000]  loss: 0.0505   precision: 100.00%\n",
      "[307/1000]  loss: 0.0506   precision: 100.00%\n",
      "[308/1000]  loss: 0.0506   precision: 100.00%\n",
      "[309/1000]  loss: 0.0506   precision: 100.00%\n",
      "[310/1000]  loss: 0.0505   precision: 100.00%\n",
      "[311/1000]  loss: 0.0503   precision: 100.00%\n",
      "[312/1000]  loss: 0.0502   precision: 100.00%\n",
      "[313/1000]  loss: 0.0501   precision: 100.00%\n",
      "[314/1000]  loss: 0.0505   precision: 100.00%\n",
      "[315/1000]  loss: 0.0503   precision: 100.00%\n",
      "[316/1000]  loss: 0.0507   precision: 100.00%\n",
      "[317/1000]  loss: 0.0497   precision: 100.00%\n",
      "[318/1000]  loss: 0.0499   precision: 100.00%\n",
      "[319/1000]  loss: 0.0502   precision: 100.00%\n",
      "[320/1000]  loss: 0.0503   precision: 100.00%\n",
      "[321/1000]  loss: 0.0494   precision: 100.00%\n",
      "[322/1000]  loss: 0.0498   precision: 100.00%\n",
      "[323/1000]  loss: 0.0494   precision: 100.00%\n",
      "[324/1000]  loss: 0.0497   precision: 100.00%\n",
      "[325/1000]  loss: 0.0498   precision: 100.00%\n",
      "[326/1000]  loss: 0.0495   precision: 100.00%\n",
      "[327/1000]  loss: 0.0496   precision: 100.00%\n",
      "[328/1000]  loss: 0.0494   precision: 100.00%\n",
      "[329/1000]  loss: 0.0491   precision: 100.00%\n",
      "[330/1000]  loss: 0.0491   precision: 100.00%\n",
      "[331/1000]  loss: 0.0494   precision: 100.00%\n",
      "[332/1000]  loss: 0.0490   precision: 100.00%\n",
      "[333/1000]  loss: 0.0489   precision: 100.00%\n",
      "[334/1000]  loss: 0.0484   precision: 100.00%\n",
      "[335/1000]  loss: 0.0486   precision: 100.00%\n",
      "[336/1000]  loss: 0.0492   precision: 100.00%\n",
      "[337/1000]  loss: 0.0487   precision: 100.00%\n",
      "[338/1000]  loss: 0.0486   precision: 100.00%\n",
      "[339/1000]  loss: 0.0486   precision: 100.00%\n",
      "[340/1000]  loss: 0.0486   precision: 100.00%\n",
      "[341/1000]  loss: 0.0484   precision: 100.00%\n",
      "[342/1000]  loss: 0.0487   precision: 100.00%\n",
      "[343/1000]  loss: 0.0485   precision: 100.00%\n",
      "[344/1000]  loss: 0.0485   precision: 100.00%\n",
      "[345/1000]  loss: 0.0480   precision: 100.00%\n",
      "[346/1000]  loss: 0.0485   precision: 100.00%\n",
      "[347/1000]  loss: 0.0484   precision: 100.00%\n",
      "[348/1000]  loss: 0.0482   precision: 100.00%\n",
      "[349/1000]  loss: 0.0478   precision: 100.00%\n",
      "[350/1000]  loss: 0.0477   precision: 100.00%\n",
      "[351/1000]  loss: 0.0477   precision: 100.00%\n",
      "[352/1000]  loss: 0.0480   precision: 100.00%\n",
      "[353/1000]  loss: 0.0477   precision: 100.00%\n",
      "[354/1000]  loss: 0.0476   precision: 100.00%\n",
      "[355/1000]  loss: 0.0476   precision: 100.00%\n",
      "[356/1000]  loss: 0.0474   precision: 100.00%\n",
      "[357/1000]  loss: 0.0479   precision: 100.00%\n",
      "[358/1000]  loss: 0.0475   precision: 100.00%\n",
      "[359/1000]  loss: 0.0470   precision: 100.00%\n",
      "[360/1000]  loss: 0.0473   precision: 100.00%\n",
      "[361/1000]  loss: 0.0473   precision: 100.00%\n",
      "[362/1000]  loss: 0.0473   precision: 100.00%\n",
      "[363/1000]  loss: 0.0478   precision: 100.00%\n",
      "[364/1000]  loss: 0.0473   precision: 100.00%\n",
      "[365/1000]  loss: 0.0475   precision: 100.00%\n",
      "[366/1000]  loss: 0.0474   precision: 100.00%\n",
      "[367/1000]  loss: 0.0472   precision: 100.00%\n",
      "[368/1000]  loss: 0.0471   precision: 100.00%\n",
      "[369/1000]  loss: 0.0467   precision: 100.00%\n",
      "[370/1000]  loss: 0.0469   precision: 100.00%\n",
      "[371/1000]  loss: 0.0472   precision: 100.00%\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[372/1000]  loss: 0.0471   precision: 100.00%\n",
      "[373/1000]  loss: 0.0469   precision: 100.00%\n",
      "[374/1000]  loss: 0.0467   precision: 100.00%\n",
      "[375/1000]  loss: 0.0469   precision: 100.00%\n",
      "[376/1000]  loss: 0.0464   precision: 100.00%\n",
      "[377/1000]  loss: 0.0463   precision: 100.00%\n",
      "[378/1000]  loss: 0.0468   precision: 100.00%\n",
      "[379/1000]  loss: 0.0466   precision: 100.00%\n",
      "[380/1000]  loss: 0.0466   precision: 100.00%\n",
      "[381/1000]  loss: 0.0464   precision: 100.00%\n",
      "[382/1000]  loss: 0.0462   precision: 100.00%\n",
      "[383/1000]  loss: 0.0462   precision: 100.00%\n",
      "[384/1000]  loss: 0.0458   precision: 100.00%\n",
      "[385/1000]  loss: 0.0459   precision: 100.00%\n",
      "[386/1000]  loss: 0.0463   precision: 100.00%\n",
      "[387/1000]  loss: 0.0459   precision: 100.00%\n",
      "[388/1000]  loss: 0.0461   precision: 100.00%\n",
      "[389/1000]  loss: 0.0460   precision: 100.00%\n",
      "[390/1000]  loss: 0.0458   precision: 100.00%\n",
      "[391/1000]  loss: 0.0456   precision: 100.00%\n",
      "[392/1000]  loss: 0.0459   precision: 100.00%\n",
      "[393/1000]  loss: 0.0454   precision: 100.00%\n",
      "[394/1000]  loss: 0.0458   precision: 100.00%\n",
      "[395/1000]  loss: 0.0457   precision: 100.00%\n",
      "[396/1000]  loss: 0.0457   precision: 100.00%\n",
      "[397/1000]  loss: 0.0455   precision: 100.00%\n",
      "[398/1000]  loss: 0.0455   precision: 100.00%\n",
      "[399/1000]  loss: 0.0455   precision: 100.00%\n",
      "[400/1000]  loss: 0.0455   precision: 100.00%\n",
      "=============================================\n",
      "[Testing]  loss: 0.6966   precision: 78.30%\n",
      "=============================================\n",
      "[401/1000]  loss: 0.0452   precision: 100.00%\n",
      "[402/1000]  loss: 0.0451   precision: 100.00%\n",
      "[403/1000]  loss: 0.0449   precision: 100.00%\n",
      "[404/1000]  loss: 0.0448   precision: 100.00%\n",
      "[405/1000]  loss: 0.0451   precision: 100.00%\n",
      "[406/1000]  loss: 0.0448   precision: 100.00%\n",
      "[407/1000]  loss: 0.0448   precision: 100.00%\n",
      "[408/1000]  loss: 0.0447   precision: 100.00%\n",
      "[409/1000]  loss: 0.0445   precision: 100.00%\n",
      "[410/1000]  loss: 0.0445   precision: 100.00%\n",
      "[411/1000]  loss: 0.0445   precision: 100.00%\n",
      "[412/1000]  loss: 0.0447   precision: 100.00%\n",
      "[413/1000]  loss: 0.0444   precision: 100.00%\n",
      "[414/1000]  loss: 0.0445   precision: 100.00%\n",
      "[415/1000]  loss: 0.0450   precision: 100.00%\n",
      "[416/1000]  loss: 0.0444   precision: 100.00%\n",
      "[417/1000]  loss: 0.0442   precision: 100.00%\n",
      "[418/1000]  loss: 0.0444   precision: 100.00%\n",
      "[419/1000]  loss: 0.0441   precision: 100.00%\n",
      "[420/1000]  loss: 0.0443   precision: 100.00%\n",
      "[421/1000]  loss: 0.0443   precision: 100.00%\n",
      "[422/1000]  loss: 0.0443   precision: 100.00%\n",
      "[423/1000]  loss: 0.0444   precision: 100.00%\n",
      "[424/1000]  loss: 0.0440   precision: 100.00%\n",
      "[425/1000]  loss: 0.0440   precision: 100.00%\n",
      "[426/1000]  loss: 0.0441   precision: 100.00%\n",
      "[427/1000]  loss: 0.0440   precision: 100.00%\n",
      "[428/1000]  loss: 0.0438   precision: 100.00%\n",
      "[429/1000]  loss: 0.0438   precision: 100.00%\n",
      "[430/1000]  loss: 0.0438   precision: 100.00%\n",
      "[431/1000]  loss: 0.0435   precision: 100.00%\n",
      "[432/1000]  loss: 0.0438   precision: 100.00%\n",
      "[433/1000]  loss: 0.0436   precision: 100.00%\n",
      "[434/1000]  loss: 0.0431   precision: 100.00%\n",
      "[435/1000]  loss: 0.0434   precision: 100.00%\n",
      "[436/1000]  loss: 0.0437   precision: 100.00%\n",
      "[437/1000]  loss: 0.0436   precision: 100.00%\n",
      "[438/1000]  loss: 0.0434   precision: 100.00%\n",
      "[439/1000]  loss: 0.0431   precision: 100.00%\n",
      "[440/1000]  loss: 0.0432   precision: 100.00%\n",
      "[441/1000]  loss: 0.0432   precision: 100.00%\n",
      "[442/1000]  loss: 0.0432   precision: 100.00%\n",
      "[443/1000]  loss: 0.0433   precision: 100.00%\n",
      "[444/1000]  loss: 0.0433   precision: 100.00%\n",
      "[445/1000]  loss: 0.0432   precision: 100.00%\n",
      "[446/1000]  loss: 0.0429   precision: 100.00%\n",
      "[447/1000]  loss: 0.0429   precision: 100.00%\n",
      "[448/1000]  loss: 0.0429   precision: 100.00%\n",
      "[449/1000]  loss: 0.0431   precision: 100.00%\n",
      "[450/1000]  loss: 0.0426   precision: 100.00%\n",
      "[451/1000]  loss: 0.0427   precision: 100.00%\n",
      "[452/1000]  loss: 0.0425   precision: 100.00%\n",
      "[453/1000]  loss: 0.0425   precision: 100.00%\n",
      "[454/1000]  loss: 0.0426   precision: 100.00%\n",
      "[455/1000]  loss: 0.0429   precision: 100.00%\n",
      "[456/1000]  loss: 0.0429   precision: 100.00%\n",
      "[457/1000]  loss: 0.0424   precision: 100.00%\n",
      "[458/1000]  loss: 0.0429   precision: 100.00%\n",
      "[459/1000]  loss: 0.0423   precision: 100.00%\n",
      "[460/1000]  loss: 0.0424   precision: 100.00%\n",
      "[461/1000]  loss: 0.0421   precision: 100.00%\n",
      "[462/1000]  loss: 0.0422   precision: 100.00%\n",
      "[463/1000]  loss: 0.0421   precision: 100.00%\n",
      "[464/1000]  loss: 0.0423   precision: 100.00%\n",
      "[465/1000]  loss: 0.0423   precision: 100.00%\n",
      "[466/1000]  loss: 0.0420   precision: 100.00%\n",
      "[467/1000]  loss: 0.0419   precision: 100.00%\n",
      "[468/1000]  loss: 0.0421   precision: 100.00%\n",
      "[469/1000]  loss: 0.0423   precision: 100.00%\n",
      "[470/1000]  loss: 0.0420   precision: 100.00%\n",
      "[471/1000]  loss: 0.0419   precision: 100.00%\n",
      "[472/1000]  loss: 0.0419   precision: 100.00%\n",
      "[473/1000]  loss: 0.0419   precision: 100.00%\n",
      "[474/1000]  loss: 0.0429   precision: 100.00%\n",
      "[475/1000]  loss: 0.0418   precision: 100.00%\n",
      "[476/1000]  loss: 0.0416   precision: 100.00%\n",
      "[477/1000]  loss: 0.0418   precision: 100.00%\n",
      "[478/1000]  loss: 0.0414   precision: 100.00%\n",
      "[479/1000]  loss: 0.0414   precision: 100.00%\n",
      "[480/1000]  loss: 0.0417   precision: 100.00%\n",
      "[481/1000]  loss: 0.0415   precision: 100.00%\n",
      "[482/1000]  loss: 0.0412   precision: 100.00%\n",
      "[483/1000]  loss: 0.0413   precision: 100.00%\n",
      "[484/1000]  loss: 0.0414   precision: 100.00%\n",
      "[485/1000]  loss: 0.0411   precision: 100.00%\n",
      "[486/1000]  loss: 0.0412   precision: 100.00%\n",
      "[487/1000]  loss: 0.0412   precision: 100.00%\n",
      "[488/1000]  loss: 0.0410   precision: 100.00%\n",
      "[489/1000]  loss: 0.0411   precision: 100.00%\n",
      "[490/1000]  loss: 0.0411   precision: 100.00%\n",
      "[491/1000]  loss: 0.0411   precision: 100.00%\n",
      "[492/1000]  loss: 0.0409   precision: 100.00%\n",
      "[493/1000]  loss: 0.0409   precision: 100.00%\n",
      "[494/1000]  loss: 0.0409   precision: 100.00%\n",
      "[495/1000]  loss: 0.0407   precision: 100.00%\n",
      "[496/1000]  loss: 0.0406   precision: 100.00%\n",
      "[497/1000]  loss: 0.0406   precision: 100.00%\n",
      "[498/1000]  loss: 0.0407   precision: 100.00%\n",
      "[499/1000]  loss: 0.0408   precision: 100.00%\n",
      "[500/1000]  loss: 0.0404   precision: 100.00%\n",
      "=============================================\n",
      "[Testing]  loss: 0.7012   precision: 78.10%\n",
      "=============================================\n",
      "[501/1000]  loss: 0.0403   precision: 100.00%\n",
      "[502/1000]  loss: 0.0408   precision: 100.00%\n",
      "[503/1000]  loss: 0.0408   precision: 100.00%\n",
      "[504/1000]  loss: 0.0405   precision: 100.00%\n",
      "[505/1000]  loss: 0.0404   precision: 100.00%\n",
      "[506/1000]  loss: 0.0404   precision: 100.00%\n",
      "[507/1000]  loss: 0.0401   precision: 100.00%\n",
      "[508/1000]  loss: 0.0402   precision: 100.00%\n",
      "[509/1000]  loss: 0.0402   precision: 100.00%\n",
      "[510/1000]  loss: 0.0402   precision: 100.00%\n",
      "[511/1000]  loss: 0.0401   precision: 100.00%\n",
      "[512/1000]  loss: 0.0399   precision: 100.00%\n",
      "[513/1000]  loss: 0.0401   precision: 100.00%\n",
      "[514/1000]  loss: 0.0401   precision: 100.00%\n",
      "[515/1000]  loss: 0.0400   precision: 100.00%\n",
      "[516/1000]  loss: 0.0399   precision: 100.00%\n",
      "[517/1000]  loss: 0.0401   precision: 100.00%\n",
      "[518/1000]  loss: 0.0398   precision: 100.00%\n",
      "[519/1000]  loss: 0.0398   precision: 100.00%\n",
      "[520/1000]  loss: 0.0394   precision: 100.00%\n",
      "[521/1000]  loss: 0.0396   precision: 100.00%\n",
      "[522/1000]  loss: 0.0399   precision: 100.00%\n",
      "[523/1000]  loss: 0.0398   precision: 100.00%\n",
      "[524/1000]  loss: 0.0397   precision: 100.00%\n",
      "[525/1000]  loss: 0.0398   precision: 100.00%\n",
      "[526/1000]  loss: 0.0394   precision: 100.00%\n",
      "[527/1000]  loss: 0.0394   precision: 100.00%\n",
      "[528/1000]  loss: 0.0395   precision: 100.00%\n",
      "[529/1000]  loss: 0.0395   precision: 100.00%\n",
      "[530/1000]  loss: 0.0392   precision: 100.00%\n",
      "[531/1000]  loss: 0.0391   precision: 100.00%\n",
      "[532/1000]  loss: 0.0394   precision: 100.00%\n",
      "[533/1000]  loss: 0.0393   precision: 100.00%\n",
      "[534/1000]  loss: 0.0394   precision: 100.00%\n",
      "[535/1000]  loss: 0.0391   precision: 100.00%\n",
      "[536/1000]  loss: 0.0394   precision: 100.00%\n",
      "[537/1000]  loss: 0.0388   precision: 100.00%\n",
      "[538/1000]  loss: 0.0394   precision: 100.00%\n",
      "[539/1000]  loss: 0.0389   precision: 100.00%\n",
      "[540/1000]  loss: 0.0389   precision: 100.00%\n",
      "[541/1000]  loss: 0.0387   precision: 100.00%\n",
      "[542/1000]  loss: 0.0387   precision: 100.00%\n",
      "[543/1000]  loss: 0.0391   precision: 100.00%\n",
      "[544/1000]  loss: 0.0386   precision: 100.00%\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[545/1000]  loss: 0.0387   precision: 100.00%\n",
      "[546/1000]  loss: 0.0387   precision: 100.00%\n",
      "[547/1000]  loss: 0.0386   precision: 100.00%\n",
      "[548/1000]  loss: 0.0384   precision: 100.00%\n",
      "[549/1000]  loss: 0.0387   precision: 100.00%\n",
      "[550/1000]  loss: 0.0384   precision: 100.00%\n",
      "[551/1000]  loss: 0.0388   precision: 100.00%\n",
      "[552/1000]  loss: 0.0384   precision: 100.00%\n",
      "[553/1000]  loss: 0.0383   precision: 100.00%\n",
      "[554/1000]  loss: 0.0386   precision: 100.00%\n",
      "[555/1000]  loss: 0.0386   precision: 100.00%\n",
      "[556/1000]  loss: 0.0387   precision: 100.00%\n",
      "[557/1000]  loss: 0.0381   precision: 100.00%\n",
      "[558/1000]  loss: 0.0385   precision: 100.00%\n",
      "[559/1000]  loss: 0.0382   precision: 100.00%\n",
      "[560/1000]  loss: 0.0383   precision: 100.00%\n",
      "[561/1000]  loss: 0.0383   precision: 100.00%\n",
      "[562/1000]  loss: 0.0387   precision: 100.00%\n",
      "[563/1000]  loss: 0.0384   precision: 100.00%\n",
      "[564/1000]  loss: 0.0384   precision: 100.00%\n",
      "[565/1000]  loss: 0.0381   precision: 100.00%\n",
      "[566/1000]  loss: 0.0379   precision: 100.00%\n",
      "[567/1000]  loss: 0.0378   precision: 100.00%\n",
      "[568/1000]  loss: 0.0377   precision: 100.00%\n",
      "[569/1000]  loss: 0.0377   precision: 100.00%\n",
      "[570/1000]  loss: 0.0377   precision: 100.00%\n",
      "[571/1000]  loss: 0.0377   precision: 100.00%\n",
      "[572/1000]  loss: 0.0376   precision: 100.00%\n",
      "[573/1000]  loss: 0.0377   precision: 100.00%\n",
      "[574/1000]  loss: 0.0376   precision: 100.00%\n",
      "[575/1000]  loss: 0.0375   precision: 100.00%\n",
      "[576/1000]  loss: 0.0374   precision: 100.00%\n",
      "[577/1000]  loss: 0.0376   precision: 100.00%\n",
      "[578/1000]  loss: 0.0373   precision: 100.00%\n",
      "[579/1000]  loss: 0.0376   precision: 100.00%\n",
      "[580/1000]  loss: 0.0377   precision: 100.00%\n",
      "[581/1000]  loss: 0.0374   precision: 100.00%\n",
      "[582/1000]  loss: 0.0372   precision: 100.00%\n",
      "[583/1000]  loss: 0.0371   precision: 100.00%\n",
      "[584/1000]  loss: 0.0371   precision: 100.00%\n",
      "[585/1000]  loss: 0.0370   precision: 100.00%\n",
      "[586/1000]  loss: 0.0370   precision: 100.00%\n",
      "[587/1000]  loss: 0.0370   precision: 100.00%\n",
      "[588/1000]  loss: 0.0373   precision: 100.00%\n",
      "[589/1000]  loss: 0.0370   precision: 100.00%\n",
      "[590/1000]  loss: 0.0373   precision: 100.00%\n",
      "[591/1000]  loss: 0.0368   precision: 100.00%\n",
      "[592/1000]  loss: 0.0373   precision: 100.00%\n",
      "[593/1000]  loss: 0.0368   precision: 100.00%\n",
      "[594/1000]  loss: 0.0371   precision: 100.00%\n",
      "[595/1000]  loss: 0.0370   precision: 100.00%\n",
      "[596/1000]  loss: 0.0370   precision: 100.00%\n",
      "[597/1000]  loss: 0.0365   precision: 100.00%\n",
      "[598/1000]  loss: 0.0368   precision: 100.00%\n",
      "[599/1000]  loss: 0.0364   precision: 100.00%\n",
      "[600/1000]  loss: 0.0364   precision: 100.00%\n",
      "=============================================\n",
      "[Testing]  loss: 0.7087   precision: 78.10%\n",
      "=============================================\n",
      "[601/1000]  loss: 0.0366   precision: 100.00%\n",
      "[602/1000]  loss: 0.0365   precision: 100.00%\n",
      "[603/1000]  loss: 0.0364   precision: 100.00%\n",
      "[604/1000]  loss: 0.0364   precision: 100.00%\n",
      "[605/1000]  loss: 0.0366   precision: 100.00%\n",
      "[606/1000]  loss: 0.0363   precision: 100.00%\n",
      "[607/1000]  loss: 0.0363   precision: 100.00%\n",
      "[608/1000]  loss: 0.0362   precision: 100.00%\n",
      "[609/1000]  loss: 0.0364   precision: 100.00%\n",
      "[610/1000]  loss: 0.0363   precision: 100.00%\n",
      "[611/1000]  loss: 0.0362   precision: 100.00%\n",
      "[612/1000]  loss: 0.0364   precision: 100.00%\n",
      "[613/1000]  loss: 0.0359   precision: 100.00%\n",
      "[614/1000]  loss: 0.0361   precision: 100.00%\n",
      "[615/1000]  loss: 0.0361   precision: 100.00%\n",
      "[616/1000]  loss: 0.0360   precision: 100.00%\n",
      "[617/1000]  loss: 0.0360   precision: 100.00%\n",
      "[618/1000]  loss: 0.0359   precision: 100.00%\n",
      "[619/1000]  loss: 0.0361   precision: 100.00%\n",
      "[620/1000]  loss: 0.0359   precision: 100.00%\n",
      "[621/1000]  loss: 0.0360   precision: 100.00%\n",
      "[622/1000]  loss: 0.0358   precision: 100.00%\n",
      "[623/1000]  loss: 0.0359   precision: 100.00%\n",
      "[624/1000]  loss: 0.0360   precision: 100.00%\n",
      "[625/1000]  loss: 0.0364   precision: 100.00%\n",
      "[626/1000]  loss: 0.0358   precision: 100.00%\n",
      "[627/1000]  loss: 0.0358   precision: 100.00%\n",
      "[628/1000]  loss: 0.0358   precision: 100.00%\n",
      "[629/1000]  loss: 0.0355   precision: 100.00%\n",
      "[630/1000]  loss: 0.0355   precision: 100.00%\n",
      "[631/1000]  loss: 0.0356   precision: 100.00%\n",
      "[632/1000]  loss: 0.0353   precision: 100.00%\n",
      "[633/1000]  loss: 0.0354   precision: 100.00%\n",
      "[634/1000]  loss: 0.0355   precision: 100.00%\n",
      "[635/1000]  loss: 0.0352   precision: 100.00%\n",
      "[636/1000]  loss: 0.0356   precision: 100.00%\n",
      "[637/1000]  loss: 0.0354   precision: 100.00%\n",
      "[638/1000]  loss: 0.0357   precision: 100.00%\n",
      "[639/1000]  loss: 0.0354   precision: 100.00%\n",
      "[640/1000]  loss: 0.0352   precision: 100.00%\n",
      "[641/1000]  loss: 0.0353   precision: 100.00%\n",
      "[642/1000]  loss: 0.0353   precision: 100.00%\n",
      "[643/1000]  loss: 0.0351   precision: 100.00%\n",
      "[644/1000]  loss: 0.0348   precision: 100.00%\n",
      "[645/1000]  loss: 0.0353   precision: 100.00%\n",
      "[646/1000]  loss: 0.0355   precision: 100.00%\n",
      "[647/1000]  loss: 0.0349   precision: 100.00%\n",
      "[648/1000]  loss: 0.0348   precision: 100.00%\n",
      "[649/1000]  loss: 0.0350   precision: 100.00%\n",
      "[650/1000]  loss: 0.0348   precision: 100.00%\n",
      "[651/1000]  loss: 0.0349   precision: 100.00%\n",
      "[652/1000]  loss: 0.0348   precision: 100.00%\n",
      "[653/1000]  loss: 0.0348   precision: 100.00%\n",
      "[654/1000]  loss: 0.0347   precision: 100.00%\n",
      "[655/1000]  loss: 0.0347   precision: 100.00%\n",
      "[656/1000]  loss: 0.0347   precision: 100.00%\n",
      "[657/1000]  loss: 0.0348   precision: 100.00%\n",
      "[658/1000]  loss: 0.0345   precision: 100.00%\n",
      "[659/1000]  loss: 0.0345   precision: 100.00%\n",
      "[660/1000]  loss: 0.0344   precision: 100.00%\n",
      "[661/1000]  loss: 0.0345   precision: 100.00%\n",
      "[662/1000]  loss: 0.0348   precision: 100.00%\n",
      "[663/1000]  loss: 0.0345   precision: 100.00%\n",
      "[664/1000]  loss: 0.0345   precision: 100.00%\n",
      "[665/1000]  loss: 0.0346   precision: 100.00%\n",
      "[666/1000]  loss: 0.0342   precision: 100.00%\n",
      "[667/1000]  loss: 0.0343   precision: 100.00%\n",
      "[668/1000]  loss: 0.0343   precision: 100.00%\n",
      "[669/1000]  loss: 0.0346   precision: 100.00%\n",
      "[670/1000]  loss: 0.0342   precision: 100.00%\n",
      "[671/1000]  loss: 0.0340   precision: 100.00%\n",
      "[672/1000]  loss: 0.0345   precision: 100.00%\n",
      "[673/1000]  loss: 0.0342   precision: 100.00%\n",
      "[674/1000]  loss: 0.0343   precision: 100.00%\n",
      "[675/1000]  loss: 0.0343   precision: 100.00%\n",
      "[676/1000]  loss: 0.0341   precision: 100.00%\n",
      "[677/1000]  loss: 0.0341   precision: 100.00%\n",
      "[678/1000]  loss: 0.0340   precision: 100.00%\n",
      "[679/1000]  loss: 0.0340   precision: 100.00%\n",
      "[680/1000]  loss: 0.0342   precision: 100.00%\n",
      "[681/1000]  loss: 0.0339   precision: 100.00%\n",
      "[682/1000]  loss: 0.0340   precision: 100.00%\n",
      "[683/1000]  loss: 0.0339   precision: 100.00%\n",
      "[684/1000]  loss: 0.0338   precision: 100.00%\n",
      "[685/1000]  loss: 0.0338   precision: 100.00%\n",
      "[686/1000]  loss: 0.0338   precision: 100.00%\n",
      "[687/1000]  loss: 0.0339   precision: 100.00%\n",
      "[688/1000]  loss: 0.0334   precision: 100.00%\n",
      "[689/1000]  loss: 0.0334   precision: 100.00%\n",
      "[690/1000]  loss: 0.0335   precision: 100.00%\n",
      "[691/1000]  loss: 0.0339   precision: 100.00%\n",
      "[692/1000]  loss: 0.0334   precision: 100.00%\n",
      "[693/1000]  loss: 0.0337   precision: 100.00%\n",
      "[694/1000]  loss: 0.0336   precision: 100.00%\n",
      "[695/1000]  loss: 0.0336   precision: 100.00%\n",
      "[696/1000]  loss: 0.0334   precision: 100.00%\n",
      "[697/1000]  loss: 0.0334   precision: 100.00%\n",
      "[698/1000]  loss: 0.0333   precision: 100.00%\n",
      "[699/1000]  loss: 0.0334   precision: 100.00%\n",
      "[700/1000]  loss: 0.0336   precision: 100.00%\n",
      "=============================================\n",
      "[Testing]  loss: 0.7192   precision: 77.70%\n",
      "=============================================\n",
      "[701/1000]  loss: 0.0333   precision: 100.00%\n",
      "[702/1000]  loss: 0.0333   precision: 100.00%\n",
      "[703/1000]  loss: 0.0334   precision: 100.00%\n",
      "[704/1000]  loss: 0.0330   precision: 100.00%\n",
      "[705/1000]  loss: 0.0332   precision: 100.00%\n",
      "[706/1000]  loss: 0.0333   precision: 100.00%\n",
      "[707/1000]  loss: 0.0331   precision: 100.00%\n",
      "[708/1000]  loss: 0.0330   precision: 100.00%\n",
      "[709/1000]  loss: 0.0330   precision: 100.00%\n",
      "[710/1000]  loss: 0.0331   precision: 100.00%\n",
      "[711/1000]  loss: 0.0328   precision: 100.00%\n",
      "[712/1000]  loss: 0.0329   precision: 100.00%\n",
      "[713/1000]  loss: 0.0329   precision: 100.00%\n",
      "[714/1000]  loss: 0.0329   precision: 100.00%\n",
      "[715/1000]  loss: 0.0327   precision: 100.00%\n",
      "[716/1000]  loss: 0.0329   precision: 100.00%\n",
      "[717/1000]  loss: 0.0328   precision: 100.00%\n",
      "[718/1000]  loss: 0.0328   precision: 100.00%\n",
      "[719/1000]  loss: 0.0328   precision: 100.00%\n",
      "[720/1000]  loss: 0.0327   precision: 100.00%\n",
      "[721/1000]  loss: 0.0328   precision: 100.00%\n",
      "[722/1000]  loss: 0.0325   precision: 100.00%\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[723/1000]  loss: 0.0330   precision: 100.00%\n",
      "[724/1000]  loss: 0.0327   precision: 100.00%\n",
      "[725/1000]  loss: 0.0324   precision: 100.00%\n",
      "[726/1000]  loss: 0.0325   precision: 100.00%\n",
      "[727/1000]  loss: 0.0326   precision: 100.00%\n",
      "[728/1000]  loss: 0.0325   precision: 100.00%\n",
      "[729/1000]  loss: 0.0323   precision: 100.00%\n",
      "[730/1000]  loss: 0.0324   precision: 100.00%\n",
      "[731/1000]  loss: 0.0327   precision: 100.00%\n",
      "[732/1000]  loss: 0.0321   precision: 100.00%\n",
      "[733/1000]  loss: 0.0323   precision: 100.00%\n",
      "[734/1000]  loss: 0.0324   precision: 100.00%\n",
      "[735/1000]  loss: 0.0322   precision: 100.00%\n",
      "[736/1000]  loss: 0.0326   precision: 100.00%\n",
      "[737/1000]  loss: 0.0323   precision: 100.00%\n",
      "[738/1000]  loss: 0.0322   precision: 100.00%\n",
      "[739/1000]  loss: 0.0320   precision: 100.00%\n",
      "[740/1000]  loss: 0.0320   precision: 100.00%\n",
      "[741/1000]  loss: 0.0320   precision: 100.00%\n",
      "[742/1000]  loss: 0.0319   precision: 100.00%\n",
      "[743/1000]  loss: 0.0322   precision: 100.00%\n",
      "[744/1000]  loss: 0.0319   precision: 100.00%\n",
      "[745/1000]  loss: 0.0318   precision: 100.00%\n",
      "[746/1000]  loss: 0.0321   precision: 100.00%\n",
      "[747/1000]  loss: 0.0320   precision: 100.00%\n",
      "[748/1000]  loss: 0.0319   precision: 100.00%\n",
      "[749/1000]  loss: 0.0319   precision: 100.00%\n",
      "[750/1000]  loss: 0.0319   precision: 100.00%\n",
      "[751/1000]  loss: 0.0317   precision: 100.00%\n",
      "[752/1000]  loss: 0.0318   precision: 100.00%\n",
      "[753/1000]  loss: 0.0317   precision: 100.00%\n",
      "[754/1000]  loss: 0.0315   precision: 100.00%\n",
      "[755/1000]  loss: 0.0317   precision: 100.00%\n",
      "[756/1000]  loss: 0.0316   precision: 100.00%\n",
      "[757/1000]  loss: 0.0317   precision: 100.00%\n",
      "[758/1000]  loss: 0.0316   precision: 100.00%\n",
      "[759/1000]  loss: 0.0317   precision: 100.00%\n",
      "[760/1000]  loss: 0.0316   precision: 100.00%\n",
      "[761/1000]  loss: 0.0314   precision: 100.00%\n",
      "[762/1000]  loss: 0.0315   precision: 100.00%\n",
      "[763/1000]  loss: 0.0315   precision: 100.00%\n",
      "[764/1000]  loss: 0.0314   precision: 100.00%\n",
      "[765/1000]  loss: 0.0314   precision: 100.00%\n",
      "[766/1000]  loss: 0.0314   precision: 100.00%\n",
      "[767/1000]  loss: 0.0316   precision: 100.00%\n",
      "[768/1000]  loss: 0.0315   precision: 100.00%\n",
      "[769/1000]  loss: 0.0313   precision: 100.00%\n",
      "[770/1000]  loss: 0.0312   precision: 100.00%\n",
      "[771/1000]  loss: 0.0312   precision: 100.00%\n",
      "[772/1000]  loss: 0.0313   precision: 100.00%\n",
      "[773/1000]  loss: 0.0310   precision: 100.00%\n",
      "[774/1000]  loss: 0.0312   precision: 100.00%\n",
      "[775/1000]  loss: 0.0312   precision: 100.00%\n",
      "[776/1000]  loss: 0.0315   precision: 100.00%\n",
      "[777/1000]  loss: 0.0311   precision: 100.00%\n",
      "[778/1000]  loss: 0.0312   precision: 100.00%\n",
      "[779/1000]  loss: 0.0311   precision: 100.00%\n",
      "[780/1000]  loss: 0.0309   precision: 100.00%\n",
      "[781/1000]  loss: 0.0310   precision: 100.00%\n",
      "[782/1000]  loss: 0.0313   precision: 100.00%\n",
      "[783/1000]  loss: 0.0312   precision: 100.00%\n",
      "[784/1000]  loss: 0.0307   precision: 100.00%\n",
      "[785/1000]  loss: 0.0306   precision: 100.00%\n",
      "[786/1000]  loss: 0.0307   precision: 100.00%\n",
      "[787/1000]  loss: 0.0311   precision: 100.00%\n",
      "[788/1000]  loss: 0.0306   precision: 100.00%\n",
      "[789/1000]  loss: 0.0308   precision: 100.00%\n",
      "[790/1000]  loss: 0.0307   precision: 100.00%\n",
      "[791/1000]  loss: 0.0305   precision: 100.00%\n",
      "[792/1000]  loss: 0.0309   precision: 100.00%\n",
      "[793/1000]  loss: 0.0307   precision: 100.00%\n",
      "[794/1000]  loss: 0.0305   precision: 100.00%\n",
      "[795/1000]  loss: 0.0306   precision: 100.00%\n",
      "[796/1000]  loss: 0.0305   precision: 100.00%\n",
      "[797/1000]  loss: 0.0305   precision: 100.00%\n",
      "[798/1000]  loss: 0.0303   precision: 100.00%\n",
      "[799/1000]  loss: 0.0304   precision: 100.00%\n",
      "[800/1000]  loss: 0.0308   precision: 100.00%\n",
      "=============================================\n",
      "[Testing]  loss: 0.7185   precision: 78.10%\n",
      "=============================================\n",
      "[801/1000]  loss: 0.0304   precision: 100.00%\n",
      "[802/1000]  loss: 0.0303   precision: 100.00%\n",
      "[803/1000]  loss: 0.0304   precision: 100.00%\n",
      "[804/1000]  loss: 0.0303   precision: 100.00%\n",
      "[805/1000]  loss: 0.0305   precision: 100.00%\n",
      "[806/1000]  loss: 0.0302   precision: 100.00%\n",
      "[807/1000]  loss: 0.0301   precision: 100.00%\n",
      "[808/1000]  loss: 0.0303   precision: 100.00%\n",
      "[809/1000]  loss: 0.0302   precision: 100.00%\n",
      "[810/1000]  loss: 0.0303   precision: 100.00%\n",
      "[811/1000]  loss: 0.0301   precision: 100.00%\n",
      "[812/1000]  loss: 0.0301   precision: 100.00%\n",
      "[813/1000]  loss: 0.0299   precision: 100.00%\n",
      "[814/1000]  loss: 0.0300   precision: 100.00%\n",
      "[815/1000]  loss: 0.0300   precision: 100.00%\n",
      "[816/1000]  loss: 0.0301   precision: 100.00%\n",
      "[817/1000]  loss: 0.0302   precision: 100.00%\n",
      "[818/1000]  loss: 0.0300   precision: 100.00%\n",
      "[819/1000]  loss: 0.0301   precision: 100.00%\n",
      "[820/1000]  loss: 0.0298   precision: 100.00%\n",
      "[821/1000]  loss: 0.0299   precision: 100.00%\n",
      "[822/1000]  loss: 0.0297   precision: 100.00%\n",
      "[823/1000]  loss: 0.0299   precision: 100.00%\n",
      "[824/1000]  loss: 0.0297   precision: 100.00%\n",
      "[825/1000]  loss: 0.0297   precision: 100.00%\n",
      "[826/1000]  loss: 0.0298   precision: 100.00%\n",
      "[827/1000]  loss: 0.0296   precision: 100.00%\n",
      "[828/1000]  loss: 0.0295   precision: 100.00%\n",
      "[829/1000]  loss: 0.0295   precision: 100.00%\n",
      "[830/1000]  loss: 0.0296   precision: 100.00%\n",
      "[831/1000]  loss: 0.0296   precision: 100.00%\n",
      "[832/1000]  loss: 0.0298   precision: 100.00%\n",
      "[833/1000]  loss: 0.0294   precision: 100.00%\n",
      "[834/1000]  loss: 0.0299   precision: 100.00%\n",
      "[835/1000]  loss: 0.0293   precision: 100.00%\n",
      "[836/1000]  loss: 0.0296   precision: 100.00%\n",
      "[837/1000]  loss: 0.0295   precision: 100.00%\n",
      "[838/1000]  loss: 0.0296   precision: 100.00%\n",
      "[839/1000]  loss: 0.0294   precision: 100.00%\n",
      "[840/1000]  loss: 0.0293   precision: 100.00%\n",
      "[841/1000]  loss: 0.0292   precision: 100.00%\n",
      "[842/1000]  loss: 0.0292   precision: 100.00%\n",
      "[843/1000]  loss: 0.0297   precision: 100.00%\n",
      "[844/1000]  loss: 0.0291   precision: 100.00%\n",
      "[845/1000]  loss: 0.0292   precision: 100.00%\n",
      "[846/1000]  loss: 0.0291   precision: 100.00%\n",
      "[847/1000]  loss: 0.0291   precision: 100.00%\n",
      "[848/1000]  loss: 0.0289   precision: 100.00%\n",
      "[849/1000]  loss: 0.0292   precision: 100.00%\n",
      "[850/1000]  loss: 0.0294   precision: 100.00%\n",
      "[851/1000]  loss: 0.0293   precision: 100.00%\n",
      "[852/1000]  loss: 0.0291   precision: 100.00%\n",
      "[853/1000]  loss: 0.0293   precision: 100.00%\n",
      "[854/1000]  loss: 0.0289   precision: 100.00%\n",
      "[855/1000]  loss: 0.0292   precision: 100.00%\n",
      "[856/1000]  loss: 0.0290   precision: 100.00%\n",
      "[857/1000]  loss: 0.0291   precision: 100.00%\n",
      "[858/1000]  loss: 0.0290   precision: 100.00%\n",
      "[859/1000]  loss: 0.0288   precision: 100.00%\n",
      "[860/1000]  loss: 0.0288   precision: 100.00%\n",
      "[861/1000]  loss: 0.0289   precision: 100.00%\n",
      "[862/1000]  loss: 0.0287   precision: 100.00%\n",
      "[863/1000]  loss: 0.0288   precision: 100.00%\n",
      "[864/1000]  loss: 0.0288   precision: 100.00%\n",
      "[865/1000]  loss: 0.0287   precision: 100.00%\n",
      "[866/1000]  loss: 0.0287   precision: 100.00%\n",
      "[867/1000]  loss: 0.0290   precision: 100.00%\n",
      "[868/1000]  loss: 0.0286   precision: 100.00%\n",
      "[869/1000]  loss: 0.0289   precision: 100.00%\n",
      "[870/1000]  loss: 0.0285   precision: 100.00%\n",
      "[871/1000]  loss: 0.0286   precision: 100.00%\n",
      "[872/1000]  loss: 0.0285   precision: 100.00%\n",
      "[873/1000]  loss: 0.0285   precision: 100.00%\n",
      "[874/1000]  loss: 0.0284   precision: 100.00%\n",
      "[875/1000]  loss: 0.0283   precision: 100.00%\n",
      "[876/1000]  loss: 0.0285   precision: 100.00%\n",
      "[877/1000]  loss: 0.0284   precision: 100.00%\n",
      "[878/1000]  loss: 0.0284   precision: 100.00%\n",
      "[879/1000]  loss: 0.0283   precision: 100.00%\n",
      "[880/1000]  loss: 0.0284   precision: 100.00%\n",
      "[881/1000]  loss: 0.0284   precision: 100.00%\n",
      "[882/1000]  loss: 0.0282   precision: 100.00%\n",
      "[883/1000]  loss: 0.0282   precision: 100.00%\n",
      "[884/1000]  loss: 0.0283   precision: 100.00%\n",
      "[885/1000]  loss: 0.0282   precision: 100.00%\n",
      "[886/1000]  loss: 0.0284   precision: 100.00%\n",
      "[887/1000]  loss: 0.0281   precision: 100.00%\n",
      "[888/1000]  loss: 0.0281   precision: 100.00%\n",
      "[889/1000]  loss: 0.0281   precision: 100.00%\n",
      "[890/1000]  loss: 0.0280   precision: 100.00%\n",
      "[891/1000]  loss: 0.0280   precision: 100.00%\n",
      "[892/1000]  loss: 0.0280   precision: 100.00%\n",
      "[893/1000]  loss: 0.0281   precision: 100.00%\n",
      "[894/1000]  loss: 0.0280   precision: 100.00%\n",
      "[895/1000]  loss: 0.0280   precision: 100.00%\n",
      "[896/1000]  loss: 0.0280   precision: 100.00%\n",
      "[897/1000]  loss: 0.0281   precision: 100.00%\n",
      "[898/1000]  loss: 0.0278   precision: 100.00%\n",
      "[899/1000]  loss: 0.0280   precision: 100.00%\n",
      "[900/1000]  loss: 0.0278   precision: 100.00%\n",
      "=============================================\n",
      "[Testing]  loss: 0.7341   precision: 77.70%\n",
      "=============================================\n",
      "[901/1000]  loss: 0.0277   precision: 100.00%\n",
      "[902/1000]  loss: 0.0279   precision: 100.00%\n",
      "[903/1000]  loss: 0.0278   precision: 100.00%\n",
      "[904/1000]  loss: 0.0278   precision: 100.00%\n",
      "[905/1000]  loss: 0.0280   precision: 100.00%\n",
      "[906/1000]  loss: 0.0278   precision: 100.00%\n",
      "[907/1000]  loss: 0.0276   precision: 100.00%\n",
      "[908/1000]  loss: 0.0274   precision: 100.00%\n",
      "[909/1000]  loss: 0.0276   precision: 100.00%\n",
      "[910/1000]  loss: 0.0277   precision: 100.00%\n",
      "[911/1000]  loss: 0.0275   precision: 100.00%\n",
      "[912/1000]  loss: 0.0274   precision: 100.00%\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[913/1000]  loss: 0.0274   precision: 100.00%\n",
      "[914/1000]  loss: 0.0279   precision: 100.00%\n",
      "[915/1000]  loss: 0.0277   precision: 100.00%\n",
      "[916/1000]  loss: 0.0276   precision: 100.00%\n",
      "[917/1000]  loss: 0.0274   precision: 100.00%\n",
      "[918/1000]  loss: 0.0274   precision: 100.00%\n",
      "[919/1000]  loss: 0.0274   precision: 100.00%\n",
      "[920/1000]  loss: 0.0275   precision: 100.00%\n",
      "[921/1000]  loss: 0.0274   precision: 100.00%\n",
      "[922/1000]  loss: 0.0273   precision: 100.00%\n",
      "[923/1000]  loss: 0.0273   precision: 100.00%\n",
      "[924/1000]  loss: 0.0273   precision: 100.00%\n",
      "[925/1000]  loss: 0.0273   precision: 100.00%\n",
      "[926/1000]  loss: 0.0272   precision: 100.00%\n",
      "[927/1000]  loss: 0.0272   precision: 100.00%\n",
      "[928/1000]  loss: 0.0273   precision: 100.00%\n",
      "[929/1000]  loss: 0.0273   precision: 100.00%\n",
      "[930/1000]  loss: 0.0272   precision: 100.00%\n",
      "[931/1000]  loss: 0.0272   precision: 100.00%\n",
      "[932/1000]  loss: 0.0271   precision: 100.00%\n",
      "[933/1000]  loss: 0.0272   precision: 100.00%\n",
      "[934/1000]  loss: 0.0271   precision: 100.00%\n",
      "[935/1000]  loss: 0.0269   precision: 100.00%\n",
      "[936/1000]  loss: 0.0270   precision: 100.00%\n",
      "[937/1000]  loss: 0.0270   precision: 100.00%\n",
      "[938/1000]  loss: 0.0269   precision: 100.00%\n",
      "[939/1000]  loss: 0.0268   precision: 100.00%\n",
      "[940/1000]  loss: 0.0269   precision: 100.00%\n",
      "[941/1000]  loss: 0.0270   precision: 100.00%\n",
      "[942/1000]  loss: 0.0269   precision: 100.00%\n",
      "[943/1000]  loss: 0.0269   precision: 100.00%\n",
      "[944/1000]  loss: 0.0269   precision: 100.00%\n",
      "[945/1000]  loss: 0.0268   precision: 100.00%\n",
      "[946/1000]  loss: 0.0269   precision: 100.00%\n",
      "[947/1000]  loss: 0.0267   precision: 100.00%\n",
      "[948/1000]  loss: 0.0269   precision: 100.00%\n",
      "[949/1000]  loss: 0.0266   precision: 100.00%\n",
      "[950/1000]  loss: 0.0268   precision: 100.00%\n",
      "[951/1000]  loss: 0.0268   precision: 100.00%\n",
      "[952/1000]  loss: 0.0265   precision: 100.00%\n",
      "[953/1000]  loss: 0.0265   precision: 100.00%\n",
      "[954/1000]  loss: 0.0267   precision: 100.00%\n",
      "[955/1000]  loss: 0.0265   precision: 100.00%\n",
      "[956/1000]  loss: 0.0267   precision: 100.00%\n",
      "[957/1000]  loss: 0.0266   precision: 100.00%\n",
      "[958/1000]  loss: 0.0265   precision: 100.00%\n",
      "[959/1000]  loss: 0.0264   precision: 100.00%\n",
      "[960/1000]  loss: 0.0265   precision: 100.00%\n",
      "[961/1000]  loss: 0.0266   precision: 100.00%\n",
      "[962/1000]  loss: 0.0265   precision: 100.00%\n",
      "[963/1000]  loss: 0.0264   precision: 100.00%\n",
      "[964/1000]  loss: 0.0265   precision: 100.00%\n",
      "[965/1000]  loss: 0.0267   precision: 100.00%\n",
      "[966/1000]  loss: 0.0265   precision: 100.00%\n",
      "[967/1000]  loss: 0.0263   precision: 100.00%\n",
      "[968/1000]  loss: 0.0265   precision: 100.00%\n",
      "[969/1000]  loss: 0.0266   precision: 100.00%\n",
      "[970/1000]  loss: 0.0262   precision: 100.00%\n",
      "[971/1000]  loss: 0.0262   precision: 100.00%\n",
      "[972/1000]  loss: 0.0264   precision: 100.00%\n",
      "[973/1000]  loss: 0.0261   precision: 100.00%\n",
      "[974/1000]  loss: 0.0262   precision: 100.00%\n",
      "[975/1000]  loss: 0.0260   precision: 100.00%\n",
      "[976/1000]  loss: 0.0260   precision: 100.00%\n",
      "[977/1000]  loss: 0.0262   precision: 100.00%\n",
      "[978/1000]  loss: 0.0261   precision: 100.00%\n",
      "[979/1000]  loss: 0.0262   precision: 100.00%\n",
      "[980/1000]  loss: 0.0259   precision: 100.00%\n",
      "[981/1000]  loss: 0.0259   precision: 100.00%\n",
      "[982/1000]  loss: 0.0264   precision: 100.00%\n",
      "[983/1000]  loss: 0.0260   precision: 100.00%\n",
      "[984/1000]  loss: 0.0258   precision: 100.00%\n",
      "[985/1000]  loss: 0.0260   precision: 100.00%\n",
      "[986/1000]  loss: 0.0260   precision: 100.00%\n",
      "[987/1000]  loss: 0.0260   precision: 100.00%\n",
      "[988/1000]  loss: 0.0258   precision: 100.00%\n",
      "[989/1000]  loss: 0.0259   precision: 100.00%\n",
      "[990/1000]  loss: 0.0260   precision: 100.00%\n",
      "[991/1000]  loss: 0.0257   precision: 100.00%\n",
      "[992/1000]  loss: 0.0258   precision: 100.00%\n",
      "[993/1000]  loss: 0.0256   precision: 100.00%\n",
      "[994/1000]  loss: 0.0257   precision: 100.00%\n",
      "[995/1000]  loss: 0.0259   precision: 100.00%\n",
      "[996/1000]  loss: 0.0255   precision: 100.00%\n",
      "[997/1000]  loss: 0.0256   precision: 100.00%\n",
      "[998/1000]  loss: 0.0258   precision: 100.00%\n",
      "[999/1000]  loss: 0.0256   precision: 100.00%\n",
      "[1000/1000]  loss: 0.0256   precision: 100.00%\n",
      "=============================================\n",
      "[Testing]  loss: 0.7301   precision: 78.10%\n",
      "=============================================\n"
     ]
    }
   ],
   "source": [
    "features = features.cuda()\n",
    "adj = adj.cuda()\n",
    "labels = labels.cuda()\n",
    "\n",
    "for i in range(EPOCH):\n",
    "    optim.zero_grad()\n",
    "    outputs = model(features, adj)\n",
    "    loss = loss_fn(outputs[0, idx_train], labels[0, idx_train])\n",
    "    loss.backward()\n",
    "    optim.step()\n",
    "\n",
    "    # precision\n",
    "    predicts = outputs.argmax(dim=2)\n",
    "    precision = accuracy_score(labels[0, idx_train].cpu(), predicts[0, idx_train].cpu())\n",
    "    print(\"[{:3d}/{:3d}]  loss: {:.4f}   precision: {:5.2%}\".format(i+1, EPOCH, loss, precision))\n",
    "    \n",
    "    if(i % TEST == TEST-1):\n",
    "        model.eval()\n",
    "        outputs = model(features, adj)\n",
    "        loss = loss_fn(outputs[0, idx_test], labels[0, idx_test])\n",
    "        \n",
    "        predicts = outputs.argmax(dim=2)\n",
    "        precision = accuracy_score(labels[0, idx_test].cpu(), predicts[0, idx_test].cpu())\n",
    "        \n",
    "        print(\"=============================================\")\n",
    "        print(\"[Testing]  loss: {:.4f}   precision: {:5.2%}\".format(loss, precision))\n",
    "        print(\"=============================================\")\n",
    "        model.train()"
   ]
  }
 ],
 "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": "3.8.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}