test_save_features.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch\n",
    "from torch import nn\n",
    "from torch.utils.data import DataLoader\n",
    "import torch.optim as optim\n",
    "from torch.optim import lr_scheduler\n",
    "from torchvision import transforms, models\n",
    "from sklearn.model_selection import train_test_split\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import os\n",
    "import logging\n",
    "import math\n",
    "import random\n",
    "import pickle\n",
    "import argparse\n",
    "import traceback\n",
    "from torchsummary import summary\n",
    "\n",
    "from dataset import PEDataset\n",
    "from utils import Timer"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Hyperparameters"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "load_feature_extractor = 'new/r3d_18_chunk10_batch16_new_splits/epoch9.pth'\n",
    "feature_dir = 'test/features/new/r3d_18_chunk10_batch16_epoch10_new_splits'\n",
    "gpu = 0\n",
    "batch_size = 1\n",
    "chunk_size = 10"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Data Splits"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "test 1000\n"
     ]
    }
   ],
   "source": [
    "with open('data/data_splits.pkl', 'rb') as f:\n",
    "    data_splits = pickle.load(f)\n",
    "\n",
    "test_paths = data_splits['test_paths']\n",
    "print(\"test {}\".format(len(test_paths)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Feature Extraction"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1000/1000 [00:00<00:00, 2033.46it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "step 1 test exams: 1000\n",
      "step 1 test chunks: 26836\n",
      "step 1 test loader iters: 26836\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "transform_test = transforms.Compose([\n",
    "        transforms.ToTensor(),\n",
    "        transforms.Resize((192, 192)),\n",
    "    ])\n",
    "\n",
    "test_set_s1 = PEDataset(paths=test_paths, mode='test', chunk_size=chunk_size,\n",
    "                        transform=transform_test)\n",
    "\n",
    "print(\"step 1 test exams: {}\".format(test_set_s1.get_num_exams()))\n",
    "print(\"step 1 test chunks: {}\".format(len(test_set_s1)))\n",
    "\n",
    "test_loader_s1 = DataLoader(test_set_s1, batch_size=1, shuffle=False)\n",
    "print(\"step 1 test loader iters: {}\".format(len(test_loader_s1)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Use cuda 0\n",
      "\n",
      "Feature Extractor:\n",
      "----------------------------------------------------------------\n",
      "        Layer (type)               Output Shape         Param #\n",
      "================================================================\n",
      "            Conv3d-1        [1, 64, 10, 96, 96]          28,224\n",
      "       BatchNorm3d-2        [1, 64, 10, 96, 96]             128\n",
      "              ReLU-3        [1, 64, 10, 96, 96]               0\n",
      "      Conv3DSimple-4        [1, 64, 10, 96, 96]         110,592\n",
      "       BatchNorm3d-5        [1, 64, 10, 96, 96]             128\n",
      "              ReLU-6        [1, 64, 10, 96, 96]               0\n",
      "      Conv3DSimple-7        [1, 64, 10, 96, 96]         110,592\n",
      "       BatchNorm3d-8        [1, 64, 10, 96, 96]             128\n",
      "              ReLU-9        [1, 64, 10, 96, 96]               0\n",
      "       BasicBlock-10        [1, 64, 10, 96, 96]               0\n",
      "     Conv3DSimple-11        [1, 64, 10, 96, 96]         110,592\n",
      "      BatchNorm3d-12        [1, 64, 10, 96, 96]             128\n",
      "             ReLU-13        [1, 64, 10, 96, 96]               0\n",
      "     Conv3DSimple-14        [1, 64, 10, 96, 96]         110,592\n",
      "      BatchNorm3d-15        [1, 64, 10, 96, 96]             128\n",
      "             ReLU-16        [1, 64, 10, 96, 96]               0\n",
      "       BasicBlock-17        [1, 64, 10, 96, 96]               0\n",
      "     Conv3DSimple-18        [1, 128, 5, 48, 48]         221,184\n",
      "      BatchNorm3d-19        [1, 128, 5, 48, 48]             256\n",
      "             ReLU-20        [1, 128, 5, 48, 48]               0\n",
      "     Conv3DSimple-21        [1, 128, 5, 48, 48]         442,368\n",
      "      BatchNorm3d-22        [1, 128, 5, 48, 48]             256\n",
      "           Conv3d-23        [1, 128, 5, 48, 48]           8,192\n",
      "      BatchNorm3d-24        [1, 128, 5, 48, 48]             256\n",
      "             ReLU-25        [1, 128, 5, 48, 48]               0\n",
      "       BasicBlock-26        [1, 128, 5, 48, 48]               0\n",
      "     Conv3DSimple-27        [1, 128, 5, 48, 48]         442,368\n",
      "      BatchNorm3d-28        [1, 128, 5, 48, 48]             256\n",
      "             ReLU-29        [1, 128, 5, 48, 48]               0\n",
      "     Conv3DSimple-30        [1, 128, 5, 48, 48]         442,368\n",
      "      BatchNorm3d-31        [1, 128, 5, 48, 48]             256\n",
      "             ReLU-32        [1, 128, 5, 48, 48]               0\n",
      "       BasicBlock-33        [1, 128, 5, 48, 48]               0\n",
      "     Conv3DSimple-34        [1, 256, 3, 24, 24]         884,736\n",
      "      BatchNorm3d-35        [1, 256, 3, 24, 24]             512\n",
      "             ReLU-36        [1, 256, 3, 24, 24]               0\n",
      "     Conv3DSimple-37        [1, 256, 3, 24, 24]       1,769,472\n",
      "      BatchNorm3d-38        [1, 256, 3, 24, 24]             512\n",
      "           Conv3d-39        [1, 256, 3, 24, 24]          32,768\n",
      "      BatchNorm3d-40        [1, 256, 3, 24, 24]             512\n",
      "             ReLU-41        [1, 256, 3, 24, 24]               0\n",
      "       BasicBlock-42        [1, 256, 3, 24, 24]               0\n",
      "     Conv3DSimple-43        [1, 256, 3, 24, 24]       1,769,472\n",
      "      BatchNorm3d-44        [1, 256, 3, 24, 24]             512\n",
      "             ReLU-45        [1, 256, 3, 24, 24]               0\n",
      "     Conv3DSimple-46        [1, 256, 3, 24, 24]       1,769,472\n",
      "      BatchNorm3d-47        [1, 256, 3, 24, 24]             512\n",
      "             ReLU-48        [1, 256, 3, 24, 24]               0\n",
      "       BasicBlock-49        [1, 256, 3, 24, 24]               0\n",
      "     Conv3DSimple-50        [1, 512, 2, 12, 12]       3,538,944\n",
      "      BatchNorm3d-51        [1, 512, 2, 12, 12]           1,024\n",
      "             ReLU-52        [1, 512, 2, 12, 12]               0\n",
      "     Conv3DSimple-53        [1, 512, 2, 12, 12]       7,077,888\n",
      "      BatchNorm3d-54        [1, 512, 2, 12, 12]           1,024\n",
      "           Conv3d-55        [1, 512, 2, 12, 12]         131,072\n",
      "      BatchNorm3d-56        [1, 512, 2, 12, 12]           1,024\n",
      "             ReLU-57        [1, 512, 2, 12, 12]               0\n",
      "       BasicBlock-58        [1, 512, 2, 12, 12]               0\n",
      "     Conv3DSimple-59        [1, 512, 2, 12, 12]       7,077,888\n",
      "      BatchNorm3d-60        [1, 512, 2, 12, 12]           1,024\n",
      "             ReLU-61        [1, 512, 2, 12, 12]               0\n",
      "     Conv3DSimple-62        [1, 512, 2, 12, 12]       7,077,888\n",
      "      BatchNorm3d-63        [1, 512, 2, 12, 12]           1,024\n",
      "             ReLU-64        [1, 512, 2, 12, 12]               0\n",
      "       BasicBlock-65        [1, 512, 2, 12, 12]               0\n",
      "AdaptiveAvgPool3d-66          [1, 512, 1, 1, 1]               0\n",
      "================================================================\n",
      "Total params: 33,166,272\n",
      "Trainable params: 33,166,272\n",
      "Non-trainable params: 0\n",
      "----------------------------------------------------------------\n",
      "Input size (MB): 4.22\n",
      "Forward/backward pass size (MB): 1017.00\n",
      "Params size (MB): 126.52\n",
      "Estimated Total Size (MB): 1147.74\n",
      "----------------------------------------------------------------\n",
      "\n",
      "\n",
      "100 iters | time: 3.20\n",
      "200 iters | time: 2.69\n",
      "300 iters | time: 2.80\n",
      "400 iters | time: 2.80\n",
      "500 iters | time: 2.78\n",
      "600 iters | time: 2.71\n",
      "700 iters | time: 3.13\n",
      "800 iters | time: 3.79\n",
      "900 iters | time: 3.67\n",
      "1000 iters | time: 2.91\n",
      "1100 iters | time: 2.76\n",
      "1200 iters | time: 3.16\n",
      "1300 iters | time: 3.39\n",
      "1400 iters | time: 2.93\n",
      "1500 iters | time: 2.61\n",
      "1600 iters | time: 2.79\n",
      "1700 iters | time: 2.73\n",
      "1800 iters | time: 2.72\n",
      "1900 iters | time: 3.78\n",
      "2000 iters | time: 3.59\n",
      "2100 iters | time: 2.87\n",
      "2200 iters | time: 2.73\n",
      "2300 iters | time: 2.74\n",
      "2400 iters | time: 2.47\n",
      "2500 iters | time: 2.61\n",
      "2600 iters | time: 2.66\n",
      "2700 iters | time: 2.55\n",
      "2800 iters | time: 2.66\n",
      "2900 iters | time: 2.71\n",
      "3000 iters | time: 2.61\n",
      "3100 iters | time: 3.21\n",
      "3200 iters | time: 2.95\n",
      "3300 iters | time: 2.82\n",
      "3400 iters | time: 2.93\n",
      "3500 iters | time: 2.88\n",
      "3600 iters | time: 2.82\n",
      "3700 iters | time: 3.24\n",
      "3800 iters | time: 2.94\n",
      "3900 iters | time: 2.93\n",
      "4000 iters | time: 3.43\n",
      "4100 iters | time: 3.45\n",
      "4200 iters | time: 2.88\n",
      "4300 iters | time: 3.11\n",
      "4400 iters | time: 2.65\n",
      "4500 iters | time: 2.97\n",
      "4600 iters | time: 2.86\n",
      "4700 iters | time: 3.02\n",
      "4800 iters | time: 2.82\n",
      "4900 iters | time: 3.12\n",
      "5000 iters | time: 2.84\n",
      "5100 iters | time: 3.09\n",
      "5200 iters | time: 2.83\n",
      "5300 iters | time: 3.64\n",
      "5400 iters | time: 3.03\n",
      "5500 iters | time: 2.70\n",
      "5600 iters | time: 2.83\n",
      "5700 iters | time: 2.89\n",
      "5800 iters | time: 3.37\n",
      "5900 iters | time: 2.47\n",
      "6000 iters | time: 2.95\n",
      "6100 iters | time: 3.25\n",
      "6200 iters | time: 2.76\n",
      "6300 iters | time: 2.70\n",
      "6400 iters | time: 2.72\n",
      "6500 iters | time: 2.79\n",
      "6600 iters | time: 2.91\n",
      "6700 iters | time: 2.83\n",
      "6800 iters | time: 2.62\n",
      "6900 iters | time: 3.02\n",
      "7000 iters | time: 2.87\n",
      "7100 iters | time: 2.74\n",
      "7200 iters | time: 2.55\n",
      "7300 iters | time: 2.62\n",
      "7400 iters | time: 2.50\n",
      "7500 iters | time: 2.61\n",
      "7600 iters | time: 2.70\n",
      "7700 iters | time: 2.53\n",
      "7800 iters | time: 2.58\n",
      "7900 iters | time: 3.06\n",
      "8000 iters | time: 2.93\n",
      "8100 iters | time: 3.32\n",
      "8200 iters | time: 2.69\n",
      "8300 iters | time: 2.75\n",
      "8400 iters | time: 2.82\n",
      "8500 iters | time: 2.71\n",
      "8600 iters | time: 2.82\n",
      "8700 iters | time: 2.83\n",
      "8800 iters | time: 2.86\n",
      "8900 iters | time: 2.83\n",
      "9000 iters | time: 2.86\n",
      "9100 iters | time: 2.60\n",
      "9200 iters | time: 2.57\n",
      "9300 iters | time: 2.82\n",
      "9400 iters | time: 3.06\n",
      "9500 iters | time: 3.36\n",
      "9600 iters | time: 2.82\n",
      "9700 iters | time: 2.88\n",
      "9800 iters | time: 2.71\n",
      "9900 iters | time: 2.65\n",
      "10000 iters | time: 2.87\n",
      "10100 iters | time: 2.76\n",
      "10200 iters | time: 2.64\n",
      "10300 iters | time: 2.71\n",
      "10400 iters | time: 2.63\n",
      "10500 iters | time: 2.65\n",
      "10600 iters | time: 2.77\n",
      "10700 iters | time: 2.89\n",
      "10800 iters | time: 2.69\n",
      "10900 iters | time: 2.71\n",
      "11000 iters | time: 2.77\n",
      "11100 iters | time: 2.88\n",
      "11200 iters | time: 2.78\n",
      "11300 iters | time: 2.89\n",
      "11400 iters | time: 2.93\n",
      "11500 iters | time: 3.10\n",
      "11600 iters | time: 2.93\n",
      "11700 iters | time: 3.13\n",
      "11800 iters | time: 2.98\n",
      "11900 iters | time: 2.84\n",
      "12000 iters | time: 2.67\n",
      "12100 iters | time: 3.22\n",
      "12200 iters | time: 3.26\n",
      "12300 iters | time: 3.01\n",
      "12400 iters | time: 3.14\n",
      "12500 iters | time: 2.81\n",
      "12600 iters | time: 3.00\n",
      "12700 iters | time: 2.99\n",
      "12800 iters | time: 3.06\n",
      "12900 iters | time: 3.25\n",
      "13000 iters | time: 3.00\n",
      "13100 iters | time: 3.86\n",
      "13200 iters | time: 3.08\n",
      "13300 iters | time: 2.56\n",
      "13400 iters | time: 2.55\n",
      "13500 iters | time: 2.96\n",
      "13600 iters | time: 3.12\n",
      "13700 iters | time: 2.63\n",
      "13800 iters | time: 3.10\n",
      "13900 iters | time: 2.41\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "14000 iters | time: 3.05\n",
      "14100 iters | time: 3.38\n",
      "14200 iters | time: 3.49\n",
      "14300 iters | time: 2.76\n",
      "14400 iters | time: 2.65\n",
      "14500 iters | time: 2.97\n",
      "14600 iters | time: 3.08\n",
      "14700 iters | time: 2.81\n",
      "14800 iters | time: 2.92\n",
      "14900 iters | time: 3.00\n",
      "15000 iters | time: 2.99\n",
      "15100 iters | time: 2.77\n",
      "15200 iters | time: 3.45\n",
      "15300 iters | time: 3.00\n",
      "15400 iters | time: 2.94\n",
      "15500 iters | time: 2.94\n",
      "15600 iters | time: 2.61\n",
      "15700 iters | time: 2.83\n",
      "15800 iters | time: 3.06\n",
      "15900 iters | time: 3.16\n",
      "16000 iters | time: 2.82\n",
      "16100 iters | time: 2.87\n",
      "16200 iters | time: 2.64\n",
      "16300 iters | time: 2.81\n",
      "16400 iters | time: 2.84\n",
      "16500 iters | time: 3.15\n",
      "16600 iters | time: 2.57\n",
      "16700 iters | time: 2.64\n",
      "16800 iters | time: 3.83\n",
      "16900 iters | time: 2.52\n",
      "17000 iters | time: 2.57\n",
      "17100 iters | time: 2.81\n",
      "17200 iters | time: 3.40\n",
      "17300 iters | time: 2.92\n",
      "17400 iters | time: 2.93\n",
      "17500 iters | time: 2.94\n",
      "17600 iters | time: 2.64\n",
      "17700 iters | time: 2.66\n",
      "17800 iters | time: 2.75\n",
      "17900 iters | time: 2.66\n",
      "18000 iters | time: 2.58\n",
      "18100 iters | time: 2.63\n",
      "18200 iters | time: 3.19\n",
      "18300 iters | time: 2.82\n",
      "18400 iters | time: 2.52\n",
      "18500 iters | time: 2.83\n",
      "18600 iters | time: 2.83\n",
      "18700 iters | time: 2.73\n",
      "18800 iters | time: 2.70\n",
      "18900 iters | time: 2.61\n",
      "19000 iters | time: 3.01\n",
      "19100 iters | time: 3.02\n",
      "19200 iters | time: 2.89\n",
      "19300 iters | time: 3.45\n",
      "19400 iters | time: 2.96\n",
      "19500 iters | time: 2.91\n",
      "19600 iters | time: 2.70\n",
      "19700 iters | time: 2.68\n",
      "19800 iters | time: 3.02\n",
      "19900 iters | time: 2.87\n",
      "20000 iters | time: 3.45\n",
      "20100 iters | time: 3.25\n",
      "20200 iters | time: 2.99\n",
      "20300 iters | time: 3.69\n",
      "20400 iters | time: 3.40\n",
      "20500 iters | time: 3.69\n",
      "20600 iters | time: 3.18\n",
      "20700 iters | time: 3.19\n",
      "20800 iters | time: 2.76\n",
      "20900 iters | time: 2.74\n",
      "21000 iters | time: 3.12\n",
      "21100 iters | time: 2.90\n",
      "21200 iters | time: 2.76\n",
      "21300 iters | time: 2.76\n",
      "21400 iters | time: 2.76\n",
      "21500 iters | time: 2.67\n",
      "21600 iters | time: 2.75\n",
      "21700 iters | time: 2.85\n",
      "21800 iters | time: 3.38\n",
      "21900 iters | time: 3.10\n",
      "22000 iters | time: 3.14\n",
      "22100 iters | time: 2.93\n",
      "22200 iters | time: 3.93\n",
      "22300 iters | time: 4.42\n",
      "22400 iters | time: 3.38\n",
      "22500 iters | time: 2.86\n",
      "22600 iters | time: 3.38\n",
      "22700 iters | time: 2.81\n",
      "22800 iters | time: 3.19\n",
      "22900 iters | time: 2.83\n",
      "23000 iters | time: 2.81\n",
      "23100 iters | time: 2.88\n",
      "23200 iters | time: 3.11\n",
      "23300 iters | time: 4.62\n",
      "23400 iters | time: 2.92\n",
      "23500 iters | time: 2.85\n",
      "23600 iters | time: 2.90\n",
      "23700 iters | time: 2.83\n",
      "23800 iters | time: 3.32\n",
      "23900 iters | time: 2.95\n",
      "24000 iters | time: 2.91\n",
      "24100 iters | time: 2.75\n",
      "24200 iters | time: 2.69\n",
      "24300 iters | time: 2.98\n",
      "24400 iters | time: 2.96\n",
      "24500 iters | time: 3.15\n",
      "24600 iters | time: 3.35\n",
      "24700 iters | time: 2.66\n",
      "24800 iters | time: 2.82\n",
      "24900 iters | time: 2.59\n",
      "25000 iters | time: 2.70\n",
      "25100 iters | time: 3.37\n",
      "25200 iters | time: 4.85\n",
      "25300 iters | time: 3.34\n",
      "25400 iters | time: 3.24\n",
      "25500 iters | time: 3.44\n",
      "25600 iters | time: 3.84\n",
      "25700 iters | time: 2.97\n",
      "25800 iters | time: 3.43\n",
      "25900 iters | time: 3.31\n",
      "26000 iters | time: 3.43\n",
      "26100 iters | time: 2.70\n",
      "26200 iters | time: 3.54\n",
      "26300 iters | time: 2.80\n",
      "26400 iters | time: 2.99\n",
      "26500 iters | time: 3.08\n",
      "26600 iters | time: 2.99\n",
      "26700 iters | time: 2.60\n",
      "26800 iters | time: 2.81\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "1.1151161193847656"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "if load_feature_extractor == '':\n",
    "    model = models.video.r3d_18(pretrained=True)\n",
    "    num_features = model.fc.in_features\n",
    "    model.fc = nn.Linear(num_features, 1)\n",
    "else:\n",
    "    model = models.video.r3d_18(pretrained=False)\n",
    "    num_features = model.fc.in_features\n",
    "    model.fc = nn.Linear(num_features, 1)\n",
    "    model.load_state_dict(torch.load(\"weights/\"+load_feature_extractor))\n",
    "\n",
    "feature_extractor = torch.nn.Sequential(*(list(model.children())[:-1]))\n",
    "\n",
    "print('')\n",
    "print(\"Use cuda \" + str(gpu))\n",
    "print('')\n",
    "device = torch.device(\"cuda:\"+str(gpu) if torch.cuda.is_available() else \"cpu\")\n",
    "torch.cuda.empty_cache()\n",
    "feature_extractor.to(device)\n",
    "print(\"Feature Extractor:\")\n",
    "summary(feature_extractor, input_size=(3, 10, 192, 192), batch_size=batch_size, device=device)\n",
    "print('')\n",
    "\n",
    "test_size = len(test_set_s1)\n",
    "\n",
    "timer = Timer()\n",
    "\n",
    "feature_extractor.eval()\n",
    "\n",
    "test_features = np.zeros((test_size, 512))\n",
    "test_labels = np.zeros((test_size, ))\n",
    "\n",
    "timer.start()\n",
    "\n",
    "for batch, (inputs, labels) in enumerate(test_loader_s1):\n",
    "    inputs = inputs.to(device)\n",
    "    with torch.no_grad():\n",
    "        feature = feature_extractor(inputs)\n",
    "        feature = feature.view(inputs.size(0), -1)\n",
    "        start= batch * batch_size\n",
    "        end = start + inputs.size(0)\n",
    "        test_features[start:end] = feature.cpu().numpy()\n",
    "        test_labels[start:end] = labels.cpu().numpy()\n",
    "\n",
    "    if batch % 100 == 99:\n",
    "        print('{} iters | time: {:.2f}'.format(batch+1, timer.stop()))\n",
    "        timer.start()\n",
    "\n",
    "timer.stop()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "if not os.path.exists(feature_dir):\n",
    "    os.makedirs(feature_dir)\n",
    "np.save(feature_dir+'/test_features', test_features)\n",
    "np.save(feature_dir+'/test_labels', test_labels)"
   ]
  },
  {
   "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": "3.6.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}