blues_gans.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Dataloader executed\n",
      "Networks defined\n",
      "Loop starting now\n",
      "[0/75][0/2] Loss_D: 1.9494 Loss_G: 5.7653\n",
      "[0/75][1/2] Loss_D: 1.5667 Loss_G: 21.2722\n",
      "[1/75][0/2] Loss_D: 0.0000 Loss_G: 21.1985\n",
      "[1/75][1/2] Loss_D: 0.1045 Loss_G: 18.1663\n",
      "[2/75][0/2] Loss_D: 0.0018 Loss_G: 10.8344\n",
      "[2/75][1/2] Loss_D: 1.9049 Loss_G: 28.5911\n",
      "[3/75][0/2] Loss_D: 1.8375 Loss_G: 24.7300\n",
      "[3/75][1/2] Loss_D: 0.0000 Loss_G: 18.8437\n",
      "[4/75][0/2] Loss_D: 0.0297 Loss_G: 10.5248\n",
      "[4/75][1/2] Loss_D: 5.3280 Loss_G: 31.8827\n",
      "[5/75][0/2] Loss_D: 29.5119 Loss_G: 27.2819\n",
      "[5/75][1/2] Loss_D: 9.9463 Loss_G: 15.2354\n",
      "[6/75][0/2] Loss_D: 0.0005 Loss_G: 7.4452\n",
      "[6/75][1/2] Loss_D: 2.7934 Loss_G: 16.1866\n",
      "[7/75][0/2] Loss_D: 0.4147 Loss_G: 15.6007\n",
      "[7/75][1/2] Loss_D: 0.0015 Loss_G: 14.9349\n",
      "[8/75][0/2] Loss_D: 0.0270 Loss_G: 11.4063\n",
      "[8/75][1/2] Loss_D: 0.1223 Loss_G: 7.7063\n",
      "[9/75][0/2] Loss_D: 0.0731 Loss_G: 8.1336\n",
      "[9/75][1/2] Loss_D: 0.0680 Loss_G: 9.2292\n",
      "[10/75][0/2] Loss_D: 0.0105 Loss_G: 9.3465\n",
      "[10/75][1/2] Loss_D: 0.0278 Loss_G: 8.5770\n",
      "[11/75][0/2] Loss_D: 0.2036 Loss_G: 20.1728\n",
      "[11/75][1/2] Loss_D: 0.1032 Loss_G: 18.6710\n",
      "[12/75][0/2] Loss_D: 0.0003 Loss_G: 15.0413\n",
      "[12/75][1/2] Loss_D: 0.0007 Loss_G: 14.9834\n",
      "[13/75][0/2] Loss_D: 0.0022 Loss_G: 13.0323\n",
      "[13/75][1/2] Loss_D: 0.0070 Loss_G: 11.7643\n",
      "[14/75][0/2] Loss_D: 0.0175 Loss_G: 10.1487\n",
      "[14/75][1/2] Loss_D: 0.0128 Loss_G: 10.6663\n",
      "[15/75][0/2] Loss_D: 0.0269 Loss_G: 9.5200\n",
      "[15/75][1/2] Loss_D: 0.0212 Loss_G: 10.0208\n",
      "[16/75][0/2] Loss_D: 0.0531 Loss_G: 9.2614\n",
      "[16/75][1/2] Loss_D: 0.0310 Loss_G: 8.7239\n",
      "[17/75][0/2] Loss_D: 0.0136 Loss_G: 9.3896\n",
      "[17/75][1/2] Loss_D: 0.0079 Loss_G: 9.3686\n",
      "[18/75][0/2] Loss_D: 0.0134 Loss_G: 7.5979\n",
      "[18/75][1/2] Loss_D: 0.0147 Loss_G: 8.5046\n",
      "[19/75][0/2] Loss_D: 0.0104 Loss_G: 9.2660\n",
      "[19/75][1/2] Loss_D: 0.0224 Loss_G: 8.4864\n",
      "[20/75][0/2] Loss_D: 0.0759 Loss_G: 9.8397\n",
      "[20/75][1/2] Loss_D: 0.0158 Loss_G: 10.0928\n",
      "[21/75][0/2] Loss_D: 0.0025 Loss_G: 9.6060\n",
      "[21/75][1/2] Loss_D: 0.0028 Loss_G: 9.6288\n",
      "[22/75][0/2] Loss_D: 0.0078 Loss_G: 7.7763\n",
      "[22/75][1/2] Loss_D: 0.0081 Loss_G: 7.8689\n",
      "[23/75][0/2] Loss_D: 0.0123 Loss_G: 8.3045\n",
      "[23/75][1/2] Loss_D: 0.0125 Loss_G: 8.7587\n",
      "[24/75][0/2] Loss_D: 0.0733 Loss_G: 10.1575\n",
      "[24/75][1/2] Loss_D: 0.0021 Loss_G: 10.9487\n",
      "[25/75][0/2] Loss_D: 0.0099 Loss_G: 10.3432\n",
      "[25/75][1/2] Loss_D: 0.0024 Loss_G: 10.6465\n",
      "[26/75][0/2] Loss_D: 0.0047 Loss_G: 8.8827\n",
      "[26/75][1/2] Loss_D: 0.0035 Loss_G: 9.4455\n",
      "[27/75][0/2] Loss_D: 0.0066 Loss_G: 7.9764\n",
      "[27/75][1/2] Loss_D: 0.0038 Loss_G: 8.3305\n",
      "[28/75][0/2] Loss_D: 0.0255 Loss_G: 11.5345\n",
      "[28/75][1/2] Loss_D: 0.0016 Loss_G: 12.0575\n",
      "[29/75][0/2] Loss_D: 0.0105 Loss_G: 9.0407\n",
      "[29/75][1/2] Loss_D: 0.0065 Loss_G: 9.2560\n",
      "[30/75][0/2] Loss_D: 0.4530 Loss_G: 40.1751\n",
      "[30/75][1/2] Loss_D: 13.8591 Loss_G: 25.3553\n",
      "[31/75][0/2] Loss_D: 0.0002 Loss_G: 3.6577\n",
      "[31/75][1/2] Loss_D: 10.3443 Loss_G: 27.8971\n",
      "[32/75][0/2] Loss_D: 13.9570 Loss_G: 17.3257\n",
      "[32/75][1/2] Loss_D: 0.0332 Loss_G: 8.6970\n",
      "[33/75][0/2] Loss_D: 0.0014 Loss_G: 5.0648\n",
      "[33/75][1/2] Loss_D: 0.0424 Loss_G: 4.2509\n",
      "[34/75][0/2] Loss_D: 0.0526 Loss_G: 4.8580\n",
      "[34/75][1/2] Loss_D: 0.0235 Loss_G: 5.2449\n",
      "[35/75][0/2] Loss_D: 0.0412 Loss_G: 5.3612\n",
      "[35/75][1/2] Loss_D: 0.0882 Loss_G: 5.4708\n",
      "[36/75][0/2] Loss_D: 0.0795 Loss_G: 6.2556\n",
      "[36/75][1/2] Loss_D: 0.2597 Loss_G: 9.1575\n",
      "[37/75][0/2] Loss_D: 0.0048 Loss_G: 9.6133\n",
      "[37/75][1/2] Loss_D: 0.0442 Loss_G: 7.2670\n",
      "[38/75][0/2] Loss_D: 0.0070 Loss_G: 12.2125\n",
      "[38/75][1/2] Loss_D: 0.1066 Loss_G: 6.3369\n",
      "[39/75][0/2] Loss_D: 0.0269 Loss_G: 4.7516\n",
      "[39/75][1/2] Loss_D: 0.4883 Loss_G: 16.4498\n",
      "[40/75][0/2] Loss_D: 1.2479 Loss_G: 5.1544\n",
      "[40/75][1/2] Loss_D: 1.2017 Loss_G: 10.3667\n",
      "[41/75][0/2] Loss_D: 0.0044 Loss_G: 10.0587\n",
      "[41/75][1/2] Loss_D: 0.0021 Loss_G: 9.4421\n",
      "[42/75][0/2] Loss_D: 0.0063 Loss_G: 7.5363\n",
      "[42/75][1/2] Loss_D: 0.0087 Loss_G: 7.1152\n",
      "[43/75][0/2] Loss_D: 0.0491 Loss_G: 5.4597\n",
      "[43/75][1/2] Loss_D: 1.9284 Loss_G: 28.5787\n",
      "[44/75][0/2] Loss_D: 3.1184 Loss_G: 12.2443\n",
      "[44/75][1/2] Loss_D: 0.0012 Loss_G: 8.7249\n",
      "[45/75][0/2] Loss_D: 0.0084 Loss_G: 6.9057\n",
      "[45/75][1/2] Loss_D: 0.0113 Loss_G: 6.5686\n",
      "[46/75][0/2] Loss_D: 0.0434 Loss_G: 5.1101\n",
      "[46/75][1/2] Loss_D: 0.0538 Loss_G: 4.7695\n",
      "[47/75][0/2] Loss_D: 0.0642 Loss_G: 4.6699\n",
      "[47/75][1/2] Loss_D: 0.1573 Loss_G: 4.4753\n",
      "[48/75][0/2] Loss_D: 0.0890 Loss_G: 5.9735\n",
      "[48/75][1/2] Loss_D: 0.0410 Loss_G: 5.8284\n",
      "[49/75][0/2] Loss_D: 0.0593 Loss_G: 5.9402\n",
      "[49/75][1/2] Loss_D: 0.0735 Loss_G: 6.8799\n",
      "[50/75][0/2] Loss_D: 0.1331 Loss_G: 10.4489\n",
      "[50/75][1/2] Loss_D: 0.9938 Loss_G: 1.1475\n",
      "[51/75][0/2] Loss_D: 2.2784 Loss_G: 6.0770\n",
      "[51/75][1/2] Loss_D: 0.0154 Loss_G: 7.4993\n",
      "[52/75][0/2] Loss_D: 0.0055 Loss_G: 6.9157\n",
      "[52/75][1/2] Loss_D: 0.0032 Loss_G: 6.9727\n",
      "[53/75][0/2] Loss_D: 0.0338 Loss_G: 5.5191\n",
      "[53/75][1/2] Loss_D: 0.3124 Loss_G: 12.7276\n",
      "[54/75][0/2] Loss_D: 0.0012 Loss_G: 18.1048\n",
      "[54/75][1/2] Loss_D: 1.7755 Loss_G: 3.5058\n",
      "[55/75][0/2] Loss_D: 6.6960 Loss_G: 31.7848\n",
      "[55/75][1/2] Loss_D: 18.9838 Loss_G: 16.1080\n",
      "[56/75][0/2] Loss_D: 0.0021 Loss_G: 7.0690\n",
      "[56/75][1/2] Loss_D: 0.6986 Loss_G: 7.0583\n",
      "[57/75][0/2] Loss_D: 0.0758 Loss_G: 7.6259\n",
      "[57/75][1/2] Loss_D: 0.2324 Loss_G: 7.0716\n",
      "[58/75][0/2] Loss_D: 0.2346 Loss_G: 6.2340\n",
      "[58/75][1/2] Loss_D: 0.0429 Loss_G: 6.3695\n",
      "[59/75][0/2] Loss_D: 0.0432 Loss_G: 5.4571\n",
      "[59/75][1/2] Loss_D: 0.1628 Loss_G: 7.9146\n",
      "[60/75][0/2] Loss_D: 0.0603 Loss_G: 7.1199\n",
      "[60/75][1/2] Loss_D: 0.1340 Loss_G: 5.0654\n",
      "[61/75][0/2] Loss_D: 0.5416 Loss_G: 14.0178\n",
      "[61/75][1/2] Loss_D: 4.0870 Loss_G: 0.0036\n",
      "[62/75][0/2] Loss_D: 6.7083 Loss_G: 12.0839\n",
      "[62/75][1/2] Loss_D: 0.1545 Loss_G: 11.3103\n",
      "[63/75][0/2] Loss_D: 1.3606 Loss_G: 1.3780\n",
      "[63/75][1/2] Loss_D: 6.0183 Loss_G: 9.9882\n",
      "[64/75][0/2] Loss_D: 0.0362 Loss_G: 12.7421\n",
      "[64/75][1/2] Loss_D: 1.6408 Loss_G: 5.2799\n",
      "[65/75][0/2] Loss_D: 0.4114 Loss_G: 5.0603\n",
      "[65/75][1/2] Loss_D: 1.1497 Loss_G: 19.0258\n",
      "[66/75][0/2] Loss_D: 15.6521 Loss_G: 6.0043\n",
      "[66/75][1/2] Loss_D: 0.6027 Loss_G: 6.9809\n",
      "[67/75][0/2] Loss_D: 0.2413 Loss_G: 7.5670\n",
      "[67/75][1/2] Loss_D: 1.4106 Loss_G: 0.0652\n",
      "[68/75][0/2] Loss_D: 3.3100 Loss_G: 16.0231\n",
      "[68/75][1/2] Loss_D: 3.6684 Loss_G: 2.5323\n",
      "[69/75][0/2] Loss_D: 5.5051 Loss_G: 4.9219\n",
      "[69/75][1/2] Loss_D: 0.2440 Loss_G: 6.8913\n",
      "[70/75][0/2] Loss_D: 1.4336 Loss_G: 0.1685\n",
      "[70/75][1/2] Loss_D: 3.5070 Loss_G: 7.4245\n",
      "[71/75][0/2] Loss_D: 0.5845 Loss_G: 3.8849\n",
      "[71/75][1/2] Loss_D: 0.6311 Loss_G: 4.3075\n",
      "[72/75][0/2] Loss_D: 0.5824 Loss_G: 2.6633\n",
      "[72/75][1/2] Loss_D: 0.9456 Loss_G: 8.4061\n",
      "[73/75][0/2] Loss_D: 5.5061 Loss_G: 0.1226\n",
      "[73/75][1/2] Loss_D: 2.6715 Loss_G: 8.3256\n",
      "[74/75][0/2] Loss_D: 6.3456 Loss_G: 1.9697\n",
      "[74/75][1/2] Loss_D: 0.5651 Loss_G: 1.2654\n",
      "Training Completed\n"
     ]
    }
   ],
   "source": [
    "import torch\n",
    "import torch.nn as nn\n",
    "import torch.nn.parallel\n",
    "import torch.optim as optim\n",
    "import torch.utils.data\n",
    "import torchvision.datasets as dsets \n",
    "import torchvision.transforms as transforms\n",
    "import torchvision.utils as vutils\n",
    "from torch.autograd import Variable\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "#hyperparameters\n",
    "batchsize=64 #batch-size of 64\n",
    "imgsize=64 #image of size (64,64)\n",
    "\n",
    "#creating transformations so that the image is ready to be fed to the neural network\n",
    "transform=transforms.Compose([transforms.Resize(imgsize),transforms.ToTensor(),transforms.Normalize((0.5,0.5,0.5),(0.5,0.5,0.5))]) # Applying transformation(resizing,tensor conversion and normalization) to the image  \n",
    "\n",
    "#loading the dataset\n",
    "dataset=dsets.ImageFolder(root='mydata',transform=transform)\n",
    "dataloader=torch.utils.data.DataLoader(dataset,batch_size=batchsize,shuffle=True,num_workers=3) #Dataloader helps is to load data in specific batch size and whether we want to shuffle it or not and num_workers=-1 means it uses all the parallel processes available\n",
    "\n",
    "print('Dataloader executed')\n",
    "#GANs use a different type of weight initialization called Xavier's Initialization\n",
    "def weights(network):\n",
    "\tclassname=network.__class__.__name__ # this line searches for the name of the class in the network defination\n",
    "\tif classname.find('Conv') !=-1: #this line finds the occurance of the strin g 'Conv' in the network function's defination and if the string 'Conv' is present then it activates the if loop\n",
    "\t\twith torch.no_grad():\n",
    "\t\t\tnetwork.weight.normal_(0.0,0.02)\n",
    "\telif classname.find('BatchNorm')!=-1: #similar operation as above\n",
    "\t\twith torch.no_grad():\n",
    "\t\t\tnetwork.weight.data.normal_(1.0,0.02)\n",
    "\t\t# network.bias.data.fill_(0)\n",
    "# Defining the generator network\n",
    "\n",
    "class generator(nn.Module):\n",
    "\tdef __init__(self):\n",
    "\t\tsuper(generator,self).__init__() # this line passes all the goodness of nn.module in the generator class that we defined \n",
    "\t\tself.main=nn.Sequential(nn.ConvTranspose2d(100,512,4,1,0,bias=False), #100 is the input noise vector, 512 are the new feature maps,4 is the size of kernel(4,4),1 is the stride of the kernel,0 is the padding and bias is set to false\n",
    "\t\t\tnn.ReLU(True), #to introduce non-linearity\n",
    "\t\t\tnn.BatchNorm2d(512), #normalizing all the new feature maps\n",
    "\t\t\tnn.ConvTranspose2d(512,256,4,2,1,bias=False),\n",
    "\t\t\tnn.ReLU(True),\n",
    "\t\t\tnn.BatchNorm2d(256),\n",
    "\t\t\tnn.ConvTranspose2d(256,128,4,2,1,bias=False),\n",
    "\t\t\tnn.ReLU(True),\n",
    "\t\t\tnn.BatchNorm2d(128),\n",
    "\t\t\tnn.ConvTranspose2d(128,64,4,2,1,bias=False),\n",
    "\t\t\tnn.ReLU(True),\n",
    "\t\t\tnn.BatchNorm2d(64),\n",
    "\t\t\tnn.ConvTranspose2d(64,3,4,2,1,bias=False),\n",
    "\t\t\tnn.Tanh() #Tanh is selected here because the dataset has values between -1 and 1 , also Tanh is bounded, this bounded and symmetrical nature of the function helps the model to learn faster \n",
    "\t\t\t)\n",
    "\tdef forward(self,x):\n",
    "\t\toutput=self.main(x)\n",
    "\t\treturn output\n",
    "netG=generator()\n",
    "netG.cuda()\n",
    "netG.apply(weights) #applying weights initizlization function on the generator function\n",
    "\n",
    "\n",
    "class discriminator(nn.Module):\n",
    "\tdef __init__(self):\n",
    "\t\tsuper(discriminator,self).__init__() #to activate inheritance\n",
    "\t\tself.main=nn.Sequential(\n",
    "\t\t\tnn.Conv2d(3,64,4,2,1,bias=False),\n",
    "\t\t\tnn.LeakyReLU(0.02,inplace=True),\n",
    "\t\t\tnn.Conv2d(64,128,4,2,1,bias=False),\n",
    "\t\t\tnn.BatchNorm2d(128),\n",
    "\t\t\tnn.LeakyReLU(0.02,inplace=True),\n",
    "\t\t\tnn.Conv2d(128,256,4,2,1,bias=False),\n",
    "\t\t\tnn.BatchNorm2d(256),\n",
    "\t\t\tnn.LeakyReLU(0.02,inplace=True),\n",
    "\t\t\tnn.Conv2d(256,512,4,2,1,bias=False),\n",
    "\t\t\tnn.BatchNorm2d(512),\n",
    "\t\t\tnn.LeakyReLU(0.02,inplace=True),\n",
    "\t\t\tnn.Conv2d(512,1,4,1,0,bias=False),\n",
    "\t\t\tnn.Sigmoid()\t\t\t\n",
    "\t\t\t)\n",
    "\tdef forward(self,x):\n",
    "\t\toutput=self.main(x)\n",
    "\t\treturn output.view(-1) # .view(-1) is used to reshape the convouled image and flatten it along a single axis.\n",
    "\n",
    "print('Networks defined')\n",
    "netD=discriminator()\n",
    "netD.cuda()\n",
    "netD.apply(weights) #applying weights initizlization function on the generator function\n",
    "\n",
    "run=75\n",
    "###### TRAINING THE GANs ########\n",
    "criterion = nn.BCELoss()\n",
    "optimizerD=optim.Adam(netD.parameters(),lr=0.001,betas=(0.5,0.999)) #the value of betas is selected from experimentation\n",
    "optimizerG=optim.Adam(netG.parameters(),lr=0.001,betas=(0.5,0.999))\n",
    "err_dis=[]\n",
    "err_gen=[]\n",
    "print('Loop starting now')\n",
    "for epoch in range(run):\t\n",
    "\tfor i,data in enumerate(dataloader,0): #0 in the enumerate parenthesis is just the starting value we want 'i' to take.\n",
    "\t\tnetD.zero_grad() #clearing gradient buffers and setting it to zero for every new batch\n",
    "\n",
    "\n",
    "\n",
    "\t\t###Training the discriminator network on real images\n",
    "\n",
    "\t\timages,_=data #_ here holds the labels of the images. As we dont want the labels here we just throw it away\n",
    "\t\treal=Variable(images.cuda()) #to compute gradients, we need to wrap the image with a Variable class\n",
    "\t\treal_targets=Variable(torch.ones(real.size()[0]).cuda()) #setting real tagerts to 1s and deciding the size on basis of how many images are present in real variable\n",
    "\t\toutput=netD(real) #computing the outputs through discriminator which will be a number between 0 and 1 giving probablity of image being real or fake\n",
    "\t\terror_real=criterion(output,real_targets) #calculating the error for real images and their output given by our discriminator network\n",
    "\n",
    "\t\t###Training the images on fake network\n",
    "\t\tnoise=Variable(torch.randn(real.size()[0],100,1,1).cuda()) #this line can be interpreted like a matrix which has real.size()[0] number of rows and 100 number of columns(technically feature maps) and each column has a 1x1 array\n",
    "\t\tfake_output=netG(noise)  #generator uses this noise vector to generate a mini-batch of fakeimages\n",
    "\t\tfake_target=Variable(torch.zeros(real.size()[0]).cuda())#instantiating the fake target as 0s\n",
    "\t\toutput=netD(fake_output.detach()) #the fake_output is a torch variable and it contains the data as well as the gradients, but we dont want this gradients so we will only use the data and for that we use .detach() method\n",
    "\t\terror_fake=criterion(output,fake_target) #calculating the error of fake images with wrt their output\n",
    "\n",
    "\t\terror_total=error_real+error_fake\n",
    "\t\terror_total.backward() #this line back-propagates the error \n",
    "\t\toptimizerD.step()      #this line updates the parameters of netD\n",
    "\n",
    "\t\t##Training the generator\n",
    "\t\tnetG.zero_grad() #initializing the gradients of generaotr neural net\n",
    "\t\toutput=netD(fake_output) #notice here we didnt use .detach() because we want to update the weights of generator unlike previously where we were updating the weights of discriminator, because the fake_output is generated by generator and not discriminator\n",
    "\t\ttarget=Variable(torch.ones(real.size()[0]).cuda()) #here we defined targets as 1s because we want the generator to generate near perfect images which the discriminator should predict as real( or target =1) \n",
    "\t\terror_gen=criterion(output,target)\n",
    "\t\terror_gen.backward()\n",
    "\t\toptimizerG.step()\n",
    "\t\tprint('[%d/%d][%d/%d] Loss_D: %.4f Loss_G: %.4f' % (epoch, run, i, len(dataloader), error_total.data, error_gen.data))\n",
    "\t\tif i%10000==0:\n",
    "\t\t\tvutils.save_image(real, '%s/real_samples.png' % \"./results\", normalize = True)\n",
    "\t\t\tfake = netG(noise)\n",
    "\t\t\tvutils.save_image(fake.data, '%s/fake_samples_epoch_%03d.png' % (\"./results\", epoch), normalize = True)\n",
    "\terr_dis.append(torch.sum(error_total))\n",
    "\terr_gen.append(torch.sum(error_gen))\n",
    "\n",
    "print('Training Completed')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0, 0.5, 'Loss')"
      ]
     },
     "execution_count": 24,
     "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.plot(range(run),err_dis,label='Discriminator')\n",
    "plt.plot(range(run),err_gen,label='Generaotr')\n",
    "plt.legend()\n",
    "plt.xlabel('Epochs')\n",
    "plt.ylabel('Loss')"
   ]
  },
  {
   "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": "3.7.1"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}