Tutorial.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [],
   "source": [
    "import torch \n",
    "from torch import nn\n",
    "from torch import FloatTensor\n",
    "from torch.autograd import Variable\n",
    "from torch import optim\n",
    "\n",
    "from numpy.linalg import matrix_rank\n",
    "from sklearn.datasets import make_spd_matrix\n",
    "import os\n",
    "from networkx import nx\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "\n",
    "from sklearn.decomposition import PCA\n",
    "from sklearn.manifold import TSNE\n",
    "from sklearn.model_selection import cross_val_score, KFold\n",
    "from sklearn.metrics import classification_report\n",
    "\n",
    "from random import shuffle\n",
    "\n",
    "import seaborn as sns\n",
    "\n",
    "%matplotlib inline\n",
    "%load_ext autoreload\n",
    "%autoreload 2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [],
   "source": [
    "import SPD\n",
    "import dataset_tools as dt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "DetNet(\n",
       "  (fc1): Linear(in_features=100, out_features=64, bias=True)\n",
       "  (fc2): Linear(in_features=64, out_features=32, bias=True)\n",
       "  (fc3): Linear(in_features=32, out_features=1, bias=True)\n",
       ")"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "SPD.DetNet(10)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "# Что хотим тут делать: \n",
    "\n",
    "Краткое описание задачи. Имеем [статью](https://aaai.org/ocs/index.php/AAAI/AAAI17/paper/download/14825/14321), в которой описан метод, который позволяет работать с положительными симметричными матрицами, а именно отображать их в матрица меньшего размера (что  круто). Сама статья не очень сложная, поэтому попробуем с помошью идей из нее решить некоторую другую задачу.\n",
    "\n",
    "Например, известно, что кроме матрицы смежностей для графов определяют еще [матрицу Лапласа](https://en.wikipedia.org/wiki/Laplacian_matrix), которая в качестве одного из своих свойств является неотрицательно определенной "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "## Маломерное представление для матриц смежности\n",
    "\n",
    "В качестве первой задачи попробуем сделать некоторое подобие энкодера для графов, у которого в качестве слоев возьмем слои из статьи выше. Эксперименты будем проводить на [датасете](https://figshare.com/articles/MUTAG_and_ENZYMES_DataSet/899875), в котором есть ~200  графов. Обучим самый простой энкодер всего с 4-мя слоями: по два линейных и два SymmetricallyCleanLayer из статьи. "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "### Датасет из положительно определённых матриц\n",
    "\n",
    "Сначала просто для теста сгенерим датасет из положительно определенных матриц и посмотрим на его результаты. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 187,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true,
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "  0%|          | 0/999 [00:00<?, ?it/s]\u001b[A\n",
      "  0%|          | 2/999 [00:00<01:18, 12.63it/s]\u001b[A\n",
      "  0%|          | 4/999 [00:00<01:18, 12.61it/s]\u001b[A\n",
      "  1%|          | 6/999 [00:00<01:19, 12.43it/s]\u001b[A\n",
      "  1%|          | 8/999 [00:00<01:21, 12.18it/s]\u001b[A\n",
      "  1%|          | 9/999 [00:00<01:23, 11.79it/s]\u001b[A\n",
      "  1%|          | 10/999 [00:00<01:25, 11.57it/s]\u001b[A\n",
      "  1%|          | 12/999 [00:01<01:24, 11.62it/s]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 10, TRAIN LOSS: 6.968207359313965, TEST LOSS 8.690054893493652\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "  1%|▏         | 14/999 [00:01<01:24, 11.61it/s]\u001b[A\n",
      "  2%|▏         | 16/999 [00:01<01:26, 11.35it/s]\u001b[A\n",
      "  2%|▏         | 18/999 [00:01<01:25, 11.44it/s]\u001b[A\n",
      "  2%|▏         | 20/999 [00:01<01:25, 11.47it/s]\u001b[A\n",
      "  2%|▏         | 22/999 [00:01<01:24, 11.55it/s]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 20, TRAIN LOSS: 2.341021776199341, TEST LOSS 2.6160192489624023\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "  2%|▏         | 24/999 [00:02<01:24, 11.53it/s]\u001b[A\n",
      "  3%|▎         | 26/999 [00:02<01:24, 11.53it/s]\u001b[A\n",
      "  3%|▎         | 28/999 [00:02<01:24, 11.48it/s]\u001b[A\n",
      "  3%|▎         | 30/999 [00:02<01:24, 11.45it/s]\u001b[A\n",
      "  3%|▎         | 32/999 [00:02<01:24, 11.44it/s]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 30, TRAIN LOSS: 1.4044148921966553, TEST LOSS 1.2379708290100098\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "  3%|▎         | 34/999 [00:03<01:27, 11.06it/s]\u001b[A\n",
      "  4%|▎         | 36/999 [00:03<01:27, 11.05it/s]\u001b[A\n",
      "  4%|▍         | 38/999 [00:03<01:26, 11.14it/s]\u001b[A\n",
      "  4%|▍         | 40/999 [00:03<01:25, 11.19it/s]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 40, TRAIN LOSS: 1.2449212074279785, TEST LOSS 1.1252540349960327\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "  4%|▍         | 42/999 [00:03<01:27, 10.94it/s]\u001b[A\n",
      "  4%|▍         | 44/999 [00:04<01:29, 10.65it/s]\u001b[A\n",
      "  5%|▍         | 45/999 [00:04<01:29, 10.62it/s]\u001b[A\n",
      "  5%|▍         | 47/999 [00:04<01:29, 10.65it/s]\u001b[A\n",
      "  5%|▍         | 49/999 [00:04<01:28, 10.69it/s]\u001b[A\n",
      "  5%|▌         | 51/999 [00:04<01:28, 10.70it/s]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 50, TRAIN LOSS: 1.174118995666504, TEST LOSS 1.1086227893829346\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "  5%|▌         | 53/999 [00:04<01:28, 10.68it/s]\u001b[A\n",
      "  6%|▌         | 55/999 [00:05<01:28, 10.68it/s]\u001b[A\n",
      "  6%|▌         | 57/999 [00:05<01:28, 10.68it/s]\u001b[A\n",
      "  6%|▌         | 59/999 [00:05<01:29, 10.45it/s]\u001b[A\n",
      "  6%|▌         | 60/999 [00:05<01:30, 10.36it/s]\u001b[A\n",
      "  6%|▌         | 61/999 [00:05<01:31, 10.26it/s]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 60, TRAIN LOSS: 1.140807032585144, TEST LOSS 1.1053504943847656\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "  6%|▌         | 62/999 [00:06<01:32, 10.17it/s]\u001b[A\n",
      "  6%|▋         | 63/999 [00:06<01:32, 10.12it/s]\u001b[A\n",
      "  6%|▋         | 64/999 [00:06<01:32, 10.09it/s]\u001b[A\n",
      "  7%|▋         | 66/999 [00:06<01:32, 10.14it/s]\u001b[A\n",
      "  7%|▋         | 67/999 [00:06<01:31, 10.13it/s]\u001b[A\n",
      "  7%|▋         | 69/999 [00:06<01:31, 10.18it/s]\u001b[A\n",
      "  7%|▋         | 71/999 [00:06<01:30, 10.21it/s]\u001b[A"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 70, TRAIN LOSS: 1.119438648223877, TEST LOSS 1.1050883531570435\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "  7%|▋         | 73/999 [00:07<01:32, 10.06it/s]\u001b[A\n",
      "Exception in thread Thread-17:\n",
      "Traceback (most recent call last):\n",
      "  File \"/Users/Macbook/anaconda/envs/py35/lib/python3.5/threading.py\", line 914, in _bootstrap_inner\n",
      "    self.run()\n",
      "  File \"/Users/Macbook/anaconda/envs/py35/lib/python3.5/site-packages/tqdm/_tqdm.py\", line 148, in run\n",
      "    for instance in self.tqdm_cls._instances:\n",
      "  File \"/Users/Macbook/anaconda/envs/py35/lib/python3.5/_weakrefset.py\", line 60, in __iter__\n",
      "    for itemref in self.data:\n",
      "RuntimeError: Set changed size during iteration\n",
      "\n",
      "  8%|▊         | 82/999 [00:08<01:29, 10.23it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 80, TRAIN LOSS: 1.097057819366455, TEST LOSS 1.1052597761154175\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  9%|▉         | 90/999 [00:08<01:29, 10.19it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 90, TRAIN LOSS: 1.0706404447555542, TEST LOSS 1.0951968431472778\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 10%|█         | 101/999 [00:09<01:28, 10.16it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 100, TRAIN LOSS: 1.0402038097381592, TEST LOSS 1.0721750259399414\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 11%|█         | 111/999 [00:10<01:26, 10.22it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 110, TRAIN LOSS: 1.0018846988677979, TEST LOSS 1.0364058017730713\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 12%|█▏        | 121/999 [00:11<01:24, 10.36it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 120, TRAIN LOSS: 0.9751952886581421, TEST LOSS 1.0089508295059204\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 13%|█▎        | 131/999 [00:12<01:23, 10.42it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 130, TRAIN LOSS: 0.9579085111618042, TEST LOSS 0.9946497082710266\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 14%|█▍        | 141/999 [00:13<01:22, 10.35it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 140, TRAIN LOSS: 0.9389818906784058, TEST LOSS 0.9835097789764404\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 15%|█▌        | 151/999 [00:14<01:22, 10.34it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 150, TRAIN LOSS: 0.9207599759101868, TEST LOSS 0.9750453233718872\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 16%|█▌        | 161/999 [00:15<01:21, 10.23it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 160, TRAIN LOSS: 0.9022831916809082, TEST LOSS 0.9627546668052673\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 17%|█▋        | 171/999 [00:16<01:20, 10.24it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 170, TRAIN LOSS: 0.8818556070327759, TEST LOSS 0.9468567967414856\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 18%|█▊        | 181/999 [00:17<01:19, 10.33it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 180, TRAIN LOSS: 0.8600527048110962, TEST LOSS 0.9284488558769226\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 19%|█▉        | 191/999 [00:18<01:18, 10.35it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 190, TRAIN LOSS: 0.8371096849441528, TEST LOSS 0.9050159454345703\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 20%|██        | 201/999 [00:19<01:16, 10.42it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 200, TRAIN LOSS: 0.8155864477157593, TEST LOSS 0.8808035850524902\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 21%|██        | 211/999 [00:20<01:15, 10.48it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 210, TRAIN LOSS: 0.7945561408996582, TEST LOSS 0.8602314591407776\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 22%|██▏       | 221/999 [00:20<01:13, 10.53it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 220, TRAIN LOSS: 0.77338045835495, TEST LOSS 0.841893196105957\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 23%|██▎       | 231/999 [00:22<01:13, 10.49it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 230, TRAIN LOSS: 0.7513979077339172, TEST LOSS 0.8237003087997437\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 24%|██▍       | 242/999 [00:23<01:12, 10.41it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 240, TRAIN LOSS: 0.7281006574630737, TEST LOSS 0.8054870963096619\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 25%|██▌       | 252/999 [00:24<01:11, 10.46it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 250, TRAIN LOSS: 0.7027376890182495, TEST LOSS 0.7831583619117737\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 26%|██▌       | 260/999 [00:24<01:10, 10.43it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 260, TRAIN LOSS: 0.6744909286499023, TEST LOSS 0.7549208998680115\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 27%|██▋       | 272/999 [00:26<01:10, 10.36it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 270, TRAIN LOSS: 0.644681453704834, TEST LOSS 0.7194063067436218\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 28%|██▊       | 282/999 [00:27<01:09, 10.37it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 280, TRAIN LOSS: 0.6157381534576416, TEST LOSS 0.6807851791381836\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 29%|██▉       | 292/999 [00:28<01:07, 10.41it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 290, TRAIN LOSS: 0.5910767316818237, TEST LOSS 0.6493476033210754\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 30%|███       | 302/999 [00:28<01:06, 10.42it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 300, TRAIN LOSS: 0.5682643055915833, TEST LOSS 0.6242988109588623\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 31%|███       | 312/999 [00:29<01:05, 10.46it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 310, TRAIN LOSS: 0.5459584593772888, TEST LOSS 0.603882908821106\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 32%|███▏      | 322/999 [00:30<01:04, 10.50it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 320, TRAIN LOSS: 0.5239473581314087, TEST LOSS 0.5831701755523682\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 33%|███▎      | 332/999 [00:31<01:03, 10.52it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 330, TRAIN LOSS: 0.5008547902107239, TEST LOSS 0.5630534887313843\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 34%|███▍      | 342/999 [00:32<01:02, 10.52it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 340, TRAIN LOSS: 0.47753703594207764, TEST LOSS 0.542799711227417\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 35%|███▌      | 352/999 [00:33<01:01, 10.54it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 350, TRAIN LOSS: 0.4532957971096039, TEST LOSS 0.5200625658035278\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 36%|███▌      | 362/999 [00:34<01:00, 10.58it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 360, TRAIN LOSS: 0.43020886182785034, TEST LOSS 0.4977532923221588\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 37%|███▋      | 370/999 [00:34<00:59, 10.59it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 370, TRAIN LOSS: 0.40932437777519226, TEST LOSS 0.47675567865371704\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 38%|███▊      | 382/999 [00:36<00:58, 10.58it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 380, TRAIN LOSS: 0.3919260799884796, TEST LOSS 0.46117103099823\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 39%|███▉      | 392/999 [00:37<00:57, 10.59it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 390, TRAIN LOSS: 0.376848042011261, TEST LOSS 0.4479031562805176\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 40%|████      | 402/999 [00:37<00:56, 10.63it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 400, TRAIN LOSS: 0.36309459805488586, TEST LOSS 0.43634364008903503\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 41%|████      | 412/999 [00:38<00:55, 10.60it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 410, TRAIN LOSS: 0.35072094202041626, TEST LOSS 0.4279869496822357\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 42%|████▏     | 422/999 [00:39<00:54, 10.61it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 420, TRAIN LOSS: 0.33966511487960815, TEST LOSS 0.4205702543258667\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 43%|████▎     | 432/999 [00:40<00:53, 10.63it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 430, TRAIN LOSS: 0.32970428466796875, TEST LOSS 0.4144282639026642\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 44%|████▍     | 442/999 [00:41<00:52, 10.67it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 440, TRAIN LOSS: 0.3210178315639496, TEST LOSS 0.40911248326301575\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 45%|████▌     | 452/999 [00:42<00:51, 10.68it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 450, TRAIN LOSS: 0.313498318195343, TEST LOSS 0.4042074978351593\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 46%|████▌     | 462/999 [00:43<00:50, 10.72it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 460, TRAIN LOSS: 0.3064330518245697, TEST LOSS 0.4002317786216736\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 47%|████▋     | 472/999 [00:43<00:48, 10.76it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 470, TRAIN LOSS: 0.30021706223487854, TEST LOSS 0.3970071077346802\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 48%|████▊     | 482/999 [00:44<00:47, 10.80it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 480, TRAIN LOSS: 0.2947506606578827, TEST LOSS 0.3945746421813965\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 49%|████▉     | 492/999 [00:45<00:46, 10.82it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 490, TRAIN LOSS: 0.290019690990448, TEST LOSS 0.39277124404907227\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 50%|█████     | 500/999 [00:46<00:46, 10.84it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 500, TRAIN LOSS: 0.28606343269348145, TEST LOSS 0.3914687931537628\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 51%|█████▏    | 512/999 [00:47<00:44, 10.86it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 510, TRAIN LOSS: 0.2828051745891571, TEST LOSS 0.3904089033603668\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 52%|█████▏    | 520/999 [00:47<00:44, 10.88it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 520, TRAIN LOSS: 0.2801343500614166, TEST LOSS 0.38960394263267517\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 53%|█████▎    | 532/999 [00:48<00:42, 10.89it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 530, TRAIN LOSS: 0.27786290645599365, TEST LOSS 0.38878029584884644\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 54%|█████▍    | 542/999 [00:49<00:41, 10.91it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 540, TRAIN LOSS: 0.27581286430358887, TEST LOSS 0.38791623711586\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 55%|█████▌    | 552/999 [00:50<00:40, 10.93it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 550, TRAIN LOSS: 0.2740107476711273, TEST LOSS 0.38689884543418884\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 56%|█████▋    | 562/999 [00:51<00:40, 10.92it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 560, TRAIN LOSS: 0.27241870760917664, TEST LOSS 0.38578611612319946\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 57%|█████▋    | 572/999 [00:52<00:39, 10.93it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 570, TRAIN LOSS: 0.27096718549728394, TEST LOSS 0.38464629650115967\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 58%|█████▊    | 582/999 [00:53<00:38, 10.94it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 580, TRAIN LOSS: 0.26960596442222595, TEST LOSS 0.3834735155105591\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 59%|█████▉    | 592/999 [00:54<00:37, 10.95it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 590, TRAIN LOSS: 0.2683123052120209, TEST LOSS 0.38238564133644104\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 60%|██████    | 602/999 [00:55<00:36, 10.94it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 600, TRAIN LOSS: 0.26707711815834045, TEST LOSS 0.3811727464199066\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 61%|██████▏   | 612/999 [00:55<00:35, 10.97it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 610, TRAIN LOSS: 0.26590651273727417, TEST LOSS 0.3799596130847931\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 62%|██████▏   | 622/999 [00:56<00:34, 11.00it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 620, TRAIN LOSS: 0.2647686004638672, TEST LOSS 0.37864622473716736\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 63%|██████▎   | 632/999 [00:57<00:33, 11.02it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 630, TRAIN LOSS: 0.26363107562065125, TEST LOSS 0.37736016511917114\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 64%|██████▍   | 642/999 [00:58<00:32, 11.03it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 640, TRAIN LOSS: 0.26247379183769226, TEST LOSS 0.37601345777511597\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 65%|██████▌   | 652/999 [00:58<00:31, 11.06it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 650, TRAIN LOSS: 0.2612477242946625, TEST LOSS 0.374494343996048\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 66%|██████▋   | 662/999 [00:59<00:30, 11.08it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 660, TRAIN LOSS: 0.2600168287754059, TEST LOSS 0.37287259101867676\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 67%|██████▋   | 670/999 [01:00<00:29, 11.10it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 670, TRAIN LOSS: 0.25879156589508057, TEST LOSS 0.3712191581726074\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 68%|██████▊   | 682/999 [01:01<00:28, 11.11it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 680, TRAIN LOSS: 0.25753113627433777, TEST LOSS 0.3695200979709625\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 69%|██████▉   | 692/999 [01:02<00:27, 11.13it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 690, TRAIN LOSS: 0.25619906187057495, TEST LOSS 0.3678280711174011\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 70%|███████   | 702/999 [01:03<00:26, 11.13it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 700, TRAIN LOSS: 0.25483667850494385, TEST LOSS 0.366073340177536\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 71%|███████▏  | 712/999 [01:04<00:25, 11.12it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 710, TRAIN LOSS: 0.2533820867538452, TEST LOSS 0.3641358017921448\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 72%|███████▏  | 721/999 [01:04<00:25, 11.10it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 720, TRAIN LOSS: 0.25186705589294434, TEST LOSS 0.3622240424156189\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 73%|███████▎  | 731/999 [01:05<00:24, 11.10it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 730, TRAIN LOSS: 0.25027212500572205, TEST LOSS 0.36020535230636597\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 74%|███████▍  | 741/999 [01:06<00:23, 11.09it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 740, TRAIN LOSS: 0.2486053705215454, TEST LOSS 0.3577384948730469\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 75%|███████▌  | 751/999 [01:07<00:22, 11.07it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 750, TRAIN LOSS: 0.24686041474342346, TEST LOSS 0.3550771474838257\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 76%|███████▌  | 761/999 [01:08<00:21, 11.07it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 760, TRAIN LOSS: 0.24501803517341614, TEST LOSS 0.3521961271762848\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 77%|███████▋  | 771/999 [01:09<00:20, 11.05it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 770, TRAIN LOSS: 0.24306116998195648, TEST LOSS 0.34909093379974365\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 78%|███████▊  | 781/999 [01:10<00:19, 11.03it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 780, TRAIN LOSS: 0.24097806215286255, TEST LOSS 0.3455047607421875\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 79%|███████▉  | 792/999 [01:12<00:18, 10.99it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 790, TRAIN LOSS: 0.23875373601913452, TEST LOSS 0.3413948118686676\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 80%|████████  | 801/999 [01:12<00:18, 10.98it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 800, TRAIN LOSS: 0.23637869954109192, TEST LOSS 0.3368360698223114\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 81%|████████  | 811/999 [01:13<00:17, 10.97it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 810, TRAIN LOSS: 0.23384152352809906, TEST LOSS 0.3317544162273407\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 82%|████████▏ | 821/999 [01:14<00:16, 10.96it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 820, TRAIN LOSS: 0.23115096986293793, TEST LOSS 0.3262110650539398\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 83%|████████▎ | 832/999 [01:16<00:15, 10.93it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 830, TRAIN LOSS: 0.22831961512565613, TEST LOSS 0.3201711177825928\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 84%|████████▍ | 841/999 [01:17<00:14, 10.91it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 840, TRAIN LOSS: 0.22535930573940277, TEST LOSS 0.3135351538658142\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 85%|████████▌ | 851/999 [01:17<00:13, 10.92it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 850, TRAIN LOSS: 0.22225700318813324, TEST LOSS 0.30645814538002014\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 86%|████████▌ | 861/999 [01:18<00:12, 10.90it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 860, TRAIN LOSS: 0.2190544307231903, TEST LOSS 0.2989732325077057\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 87%|████████▋ | 872/999 [01:20<00:11, 10.89it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 870, TRAIN LOSS: 0.21580001711845398, TEST LOSS 0.29105526208877563\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 88%|████████▊ | 881/999 [01:21<00:10, 10.86it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 880, TRAIN LOSS: 0.21252332627773285, TEST LOSS 0.2828589081764221\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 89%|████████▉ | 891/999 [01:22<00:09, 10.86it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 890, TRAIN LOSS: 0.20925837755203247, TEST LOSS 0.274495393037796\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 90%|█████████ | 900/999 [01:23<00:09, 10.84it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 900, TRAIN LOSS: 0.20605264604091644, TEST LOSS 0.26626259088516235\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 91%|█████████ | 911/999 [01:24<00:08, 10.83it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 910, TRAIN LOSS: 0.20291344821453094, TEST LOSS 0.25824427604675293\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 92%|█████████▏| 921/999 [01:25<00:07, 10.82it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 920, TRAIN LOSS: 0.19982409477233887, TEST LOSS 0.25048592686653137\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 93%|█████████▎| 932/999 [01:26<00:06, 10.80it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 930, TRAIN LOSS: 0.19680476188659668, TEST LOSS 0.24297714233398438\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 94%|█████████▍| 942/999 [01:27<00:05, 10.78it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 940, TRAIN LOSS: 0.19389823079109192, TEST LOSS 0.23559673130512238\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 95%|█████████▌| 952/999 [01:28<00:04, 10.78it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 950, TRAIN LOSS: 0.1911197006702423, TEST LOSS 0.22859683632850647\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 96%|█████████▋| 962/999 [01:29<00:03, 10.78it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 960, TRAIN LOSS: 0.1884855180978775, TEST LOSS 0.22202636301517487\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 97%|█████████▋| 971/999 [01:30<00:02, 10.77it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 970, TRAIN LOSS: 0.18602560460567474, TEST LOSS 0.21582242846488953\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 98%|█████████▊| 981/999 [01:31<00:01, 10.74it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 980, TRAIN LOSS: 0.18375475704669952, TEST LOSS 0.20998935401439667\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 99%|█████████▉| 991/999 [01:32<00:00, 10.73it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 990, TRAIN LOSS: 0.18164803087711334, TEST LOSS 0.2044777125120163\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 999/999 [01:33<00:00, 10.72it/s]\n"
     ]
    }
   ],
   "source": [
    "# учим сетку, возвращаем ее и датасет\n",
    "coder, dataset = SPD.run_test(n_iterations=1000)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 190,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Результат применения нашей модели к произвольной матрице\n",
      " Variable containing:\n",
      " 1.1597 -1.2545  0.2238 -1.3226 -0.8728 -0.5074\n",
      "-1.2545  1.8764 -0.3922  1.6513  0.9479  0.4915\n",
      " 0.2238 -0.3922  0.3590 -0.5085 -0.3040 -0.0251\n",
      "-1.3226  1.6513 -0.5085  2.5638  1.1053  0.3887\n",
      "-0.8728  0.9479 -0.3040  1.1053  1.4248  0.0865\n",
      "-0.5074  0.4915 -0.0251  0.3887  0.0865  0.3594\n",
      "[torch.FloatTensor of size 6x6]\n",
      "\n",
      "Сама матрица\n",
      " Variable containing:\n",
      " 1.6538 -1.6044  0.2423 -1.5757 -0.8835 -0.0873\n",
      "-1.6044  2.3821 -0.3803  2.0151  1.1325  0.0097\n",
      " 0.2423 -0.3803  0.3766 -0.4845 -0.2825 -0.0868\n",
      "-1.5757  2.0151 -0.4845  2.5606  1.0514  0.2524\n",
      "-0.8835  1.1325 -0.2825  1.0514  1.3400  0.0003\n",
      "-0.0873  0.0097 -0.0868  0.2524  0.0003  0.3175\n",
      "[torch.FloatTensor of size 6x6]\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(\"Результат применения нашей модели к произвольной матрице\\n\", coder(dataset)[0])\n",
    "print(\"Сама матрица\\n\", dataset[0])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "Работает вполне неплохо, что и ожидалось "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "### Mutag датасет\n",
    "\n",
    "Тут пробуем нашу модель на графовом датасете. Для каждого графа считаем его матрицу лапласа и кладем ее в выборку, потом на этом деле снова начинаем учить сетку. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 193,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true,
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  2%|▏         | 11/499 [00:02<01:28,  5.49it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 10, TRAIN LOSS: 0.19712954759597778, TEST LOSS 0.14197520911693573\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  4%|▍         | 21/499 [00:03<01:29,  5.32it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 20, TRAIN LOSS: 0.14037972688674927, TEST LOSS 0.12351789325475693\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  6%|▌         | 30/499 [00:05<01:27,  5.35it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 30, TRAIN LOSS: 0.127322718501091, TEST LOSS 0.11167646944522858\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  8%|▊         | 41/499 [00:07<01:27,  5.25it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 40, TRAIN LOSS: 0.11849711835384369, TEST LOSS 0.1044565960764885\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 10%|█         | 51/499 [00:09<01:24,  5.27it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 50, TRAIN LOSS: 0.11534447222948074, TEST LOSS 0.10198447108268738\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 12%|█▏        | 61/499 [00:11<01:23,  5.28it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 60, TRAIN LOSS: 0.11363948881626129, TEST LOSS 0.10063354671001434\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 14%|█▍        | 71/499 [00:13<01:20,  5.33it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 70, TRAIN LOSS: 0.11213558167219162, TEST LOSS 0.09911224246025085\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 16%|█▌        | 81/499 [00:15<01:18,  5.35it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 80, TRAIN LOSS: 0.11067698895931244, TEST LOSS 0.09766237437725067\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 18%|█▊        | 91/499 [00:17<01:16,  5.35it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 90, TRAIN LOSS: 0.10889769345521927, TEST LOSS 0.0962556004524231\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 20%|██        | 101/499 [00:18<01:14,  5.36it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 100, TRAIN LOSS: 0.10659370571374893, TEST LOSS 0.09432914108037949\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 22%|██▏       | 111/499 [00:20<01:12,  5.37it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 110, TRAIN LOSS: 0.10393054038286209, TEST LOSS 0.0922497883439064\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 24%|██▍       | 121/499 [00:22<01:10,  5.39it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 120, TRAIN LOSS: 0.10133963823318481, TEST LOSS 0.08909841626882553\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 26%|██▋       | 131/499 [00:24<01:07,  5.41it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 130, TRAIN LOSS: 0.09909091144800186, TEST LOSS 0.08707249909639359\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 28%|██▊       | 141/499 [00:26<01:06,  5.42it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 140, TRAIN LOSS: 0.0968422219157219, TEST LOSS 0.0854298546910286\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 30%|███       | 151/499 [00:27<01:04,  5.43it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 150, TRAIN LOSS: 0.09379082173109055, TEST LOSS 0.0823960155248642\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 32%|███▏      | 161/499 [00:29<01:02,  5.45it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 160, TRAIN LOSS: 0.0889708399772644, TEST LOSS 0.07758042961359024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 34%|███▍      | 171/499 [00:31<01:00,  5.45it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 170, TRAIN LOSS: 0.0828920304775238, TEST LOSS 0.07190632075071335\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 36%|███▋      | 181/499 [00:33<00:58,  5.42it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 180, TRAIN LOSS: 0.07788261026144028, TEST LOSS 0.06751977652311325\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 38%|███▊      | 191/499 [00:35<00:56,  5.44it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 190, TRAIN LOSS: 0.07396620512008667, TEST LOSS 0.06443893909454346\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 40%|████      | 201/499 [00:36<00:54,  5.44it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 200, TRAIN LOSS: 0.0702691301703453, TEST LOSS 0.06176358088850975\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 42%|████▏     | 211/499 [00:38<00:53,  5.43it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 210, TRAIN LOSS: 0.06590130925178528, TEST LOSS 0.05924762412905693\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 44%|████▍     | 220/499 [00:41<00:52,  5.29it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 220, TRAIN LOSS: 0.0626073107123375, TEST LOSS 0.05728299170732498\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 46%|████▋     | 231/499 [00:43<00:50,  5.27it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 230, TRAIN LOSS: 0.06048348546028137, TEST LOSS 0.055967047810554504\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 48%|████▊     | 240/499 [00:45<00:49,  5.25it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 240, TRAIN LOSS: 0.0583941787481308, TEST LOSS 0.05473392456769943\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 50%|█████     | 250/499 [00:47<00:47,  5.24it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 250, TRAIN LOSS: 0.05624337121844292, TEST LOSS 0.05337804555892944\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 52%|█████▏    | 261/499 [00:50<00:45,  5.22it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 260, TRAIN LOSS: 0.053436532616615295, TEST LOSS 0.051445819437503815\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 54%|█████▍    | 270/499 [00:52<00:44,  5.19it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 270, TRAIN LOSS: 0.051017891615629196, TEST LOSS 0.04970152676105499\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 56%|█████▌    | 280/499 [00:54<00:42,  5.16it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 280, TRAIN LOSS: 0.04926129803061485, TEST LOSS 0.048522159457206726\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 58%|█████▊    | 291/499 [00:56<00:40,  5.16it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 290, TRAIN LOSS: 0.04794570803642273, TEST LOSS 0.047715023159980774\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 60%|██████    | 301/499 [00:58<00:38,  5.16it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 300, TRAIN LOSS: 0.046789806336164474, TEST LOSS 0.04697312042117119\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 62%|██████▏   | 311/499 [01:00<00:36,  5.11it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 310, TRAIN LOSS: 0.04569118842482567, TEST LOSS 0.046245116740465164\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 64%|██████▍   | 321/499 [01:02<00:34,  5.12it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 320, TRAIN LOSS: 0.04464269056916237, TEST LOSS 0.045528557151556015\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 66%|██████▋   | 331/499 [01:04<00:32,  5.14it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 330, TRAIN LOSS: 0.043658193200826645, TEST LOSS 0.0448172353208065\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 68%|██████▊   | 341/499 [01:06<00:30,  5.15it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 340, TRAIN LOSS: 0.04271497204899788, TEST LOSS 0.04410538449883461\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 70%|███████   | 351/499 [01:07<00:28,  5.16it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 350, TRAIN LOSS: 0.04161064699292183, TEST LOSS 0.04310958832502365\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 72%|███████▏  | 361/499 [01:09<00:26,  5.17it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 360, TRAIN LOSS: 0.04047416150569916, TEST LOSS 0.04213237389922142\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 74%|███████▍  | 371/499 [01:11<00:24,  5.18it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 370, TRAIN LOSS: 0.03930344432592392, TEST LOSS 0.04111156240105629\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 76%|███████▋  | 381/499 [01:13<00:22,  5.20it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 380, TRAIN LOSS: 0.0379781648516655, TEST LOSS 0.039856504648923874\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 78%|███████▊  | 391/499 [01:15<00:20,  5.21it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 390, TRAIN LOSS: 0.03662174195051193, TEST LOSS 0.03844829648733139\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 80%|████████  | 400/499 [01:16<00:19,  5.20it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 400, TRAIN LOSS: 0.034548625349998474, TEST LOSS 0.03631008788943291\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 82%|████████▏ | 411/499 [01:19<00:16,  5.20it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 410, TRAIN LOSS: 0.032528914511203766, TEST LOSS 0.034469470381736755\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 84%|████████▍ | 421/499 [01:20<00:14,  5.21it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 420, TRAIN LOSS: 0.030917072668671608, TEST LOSS 0.03305266797542572\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 86%|████████▋ | 431/499 [01:22<00:13,  5.21it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 430, TRAIN LOSS: 0.02959536574780941, TEST LOSS 0.03154238313436508\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 88%|████████▊ | 441/499 [01:24<00:11,  5.22it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 440, TRAIN LOSS: 0.028693348169326782, TEST LOSS 0.03057944029569626\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 90%|█████████ | 451/499 [01:26<00:09,  5.23it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 450, TRAIN LOSS: 0.028074275702238083, TEST LOSS 0.029863033443689346\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 92%|█████████▏| 461/499 [01:28<00:07,  5.24it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 460, TRAIN LOSS: 0.027620680630207062, TEST LOSS 0.02927342988550663\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 94%|█████████▍| 470/499 [01:29<00:05,  5.25it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 470, TRAIN LOSS: 0.027228673920035362, TEST LOSS 0.02875402756035328\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 96%|█████████▋| 481/499 [01:31<00:03,  5.26it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 480, TRAIN LOSS: 0.026817919686436653, TEST LOSS 0.02820870466530323\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 98%|█████████▊| 491/499 [01:33<00:01,  5.27it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 490, TRAIN LOSS: 0.026432575657963753, TEST LOSS 0.027644818648695946\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 499/499 [01:34<00:00,  5.28it/s]\n"
     ]
    }
   ],
   "source": [
    "coder, dataset = SPD.run_test_mutag(n_iterations=500)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 197,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Результат применения нашей модели к матрице построенной по графу\n",
      " Variable containing:\n",
      " 0  1 -0  0 -0  0\n",
      " 1 -0  1  0  0 -0\n",
      "-0  1 -0  0  0  1\n",
      " 0  0  0 -0 -0  0\n",
      "-0  0  0 -0 -0  1\n",
      " 0 -0  1  0  1  0\n",
      "[torch.FloatTensor of size 6x6]\n",
      "\n",
      "Сама матрица\n",
      " Variable containing:\n",
      " 0  1  0  0  0  0\n",
      " 1  0  1  0  0  0\n",
      " 0  1  0  0  0  1\n",
      " 0  0  0  0  0  0\n",
      " 0  0  0  0  0  1\n",
      " 0  0  1  0  1  0\n",
      "[torch.FloatTensor of size 6x6]\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(\"Результат применения нашей модели к матрице построенной по графу\\n\", coder(dataset)[9].round())\n",
    "print(\"Сама матрица\\n\", dataset[9])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "Отрабатывает тоже вполне неплохо. Пробовал изначально сворачивать матрицу лапласа, которую считали для графа, но качество получилось плохим. Поэтому сейчас сетка работает сразу на матрицах смежностей, не особо задумываясь о том, что она может быть не положительно определенной, но блоки в сети оставляеем те же. "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "Отрисуем пару получившихся графов. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 241,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAU4AAAGrCAYAAACv2n5YAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nrkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x117e47978>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "dt.draw_graphs(10, dataset, coder)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 246,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAU4AAAGrCAYAAACv2n5YAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\ngpw4zcwKcuI0MyvIidPMrKD/H6Io7hLEg9EaAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x11832c390>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "dt.draw_graphs(15, dataset, coder)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 245,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAUgAAAGrCAYAAACixA4fAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x11811f400>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "dt.draw_graphs(100, dataset, coder)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 318,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [],
   "source": [
    "# свернем пару графов и посмотрим на их эмбеддинги \n",
    "A = coder.encoder(dataset[100].unsqueeze(0))\n",
    "B = coder.encoder(dataset[4].unsqueeze(0))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 343,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYYAAAEICAYAAABbOlNNAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\ngg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1192e8d68>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots(ncols=2, nrows=1)\n",
    "dt.draw_graph(coder.decoder(A).round(), ax=ax[0])\n",
    "ax[0].set_title(\"First Graph\")\n",
    "dt.draw_graph(coder.decoder(B).round(), ax=ax[1])\n",
    "ax[1].set_title(\"Second Graph\")\n",
    "plt.legend()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 375,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true,
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA7QAAAEICAYAAABvULiJAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nBFoAAABaEmgBAABoSaAFAACgpR/nXBlL3O57tgAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x119db27b8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "n_points = 5\n",
    "fig = plt.figure(figsize=(16, 4))\n",
    "for i, t in enumerate(np.linspace(0, 1, n_points)):\n",
    "    C_t = A - (A-B)*t\n",
    "    GC_t = coder.decoder(C_t).round()\n",
    "    ax = fig.add_subplot(1,n_points,i+1)\n",
    "    dt.draw_graph(GC_t, ax)\n",
    "    ax.set_title(round(t, 2))\n",
    "    ax.plot()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 629,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true,
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Reading graphs ...\n",
      "---:  mutag_1.graphml\n",
      "---:  mutag_2.graphml\n",
      "---:  mutag_3.graphml\n",
      "---:  mutag_4.graphml\n",
      "---:  mutag_5.graphml\n",
      "---:  mutag_6.graphml\n",
      "---:  mutag_7.graphml\n",
      "---:  mutag_8.graphml\n",
      "---:  mutag_9.graphml\n",
      "---:  mutag_10.graphml\n",
      "---:  mutag_11.graphml\n",
      "---:  mutag_12.graphml\n",
      "---:  mutag_13.graphml\n",
      "---:  mutag_14.graphml\n",
      "---:  mutag_15.graphml\n",
      "---:  mutag_16.graphml\n",
      "---:  mutag_17.graphml\n",
      "---:  mutag_18.graphml\n",
      "---:  mutag_19.graphml\n",
      "---:  mutag_20.graphml\n",
      "---:  mutag_21.graphml\n",
      "---:  mutag_22.graphml\n",
      "---:  mutag_23.graphml\n",
      "---:  mutag_24.graphml\n",
      "---:  mutag_25.graphml\n",
      "---:  mutag_26.graphml\n",
      "---:  mutag_27.graphml\n",
      "---:  mutag_28.graphml\n",
      "---:  mutag_29.graphml\n",
      "---:  mutag_30.graphml\n",
      "---:  mutag_31.graphml\n",
      "---:  mutag_32.graphml\n",
      "---:  mutag_33.graphml\n",
      "---:  mutag_34.graphml\n",
      "---:  mutag_35.graphml\n",
      "---:  mutag_36.graphml\n",
      "---:  mutag_37.graphml\n",
      "---:  mutag_38.graphml\n",
      "---:  mutag_39.graphml\n",
      "---:  mutag_40.graphml\n",
      "---:  mutag_41.graphml\n",
      "---:  mutag_42.graphml\n",
      "---:  mutag_43.graphml\n",
      "---:  mutag_44.graphml\n",
      "---:  "
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\r",
      "  0%|          | 0/4999 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "mutag_45.graphml\n",
      "---:  mutag_46.graphml\n",
      "---:  mutag_47.graphml\n",
      "---:  mutag_48.graphml\n",
      "---:  mutag_49.graphml\n",
      "---:  mutag_50.graphml\n",
      "---:  mutag_51.graphml\n",
      "---:  mutag_52.graphml\n",
      "---:  mutag_53.graphml\n",
      "---:  mutag_54.graphml\n",
      "---:  mutag_55.graphml\n",
      "---:  mutag_56.graphml\n",
      "---:  mutag_57.graphml\n",
      "---:  mutag_58.graphml\n",
      "---:  mutag_59.graphml\n",
      "---:  mutag_60.graphml\n",
      "---:  mutag_61.graphml\n",
      "---:  mutag_62.graphml\n",
      "---:  mutag_63.graphml\n",
      "---:  mutag_64.graphml\n",
      "---:  mutag_65.graphml\n",
      "---:  mutag_66.graphml\n",
      "---:  mutag_67.graphml\n",
      "---:  mutag_68.graphml\n",
      "---:  mutag_69.graphml\n",
      "---:  mutag_70.graphml\n",
      "---:  mutag_71.graphml\n",
      "---:  mutag_72.graphml\n",
      "---:  mutag_73.graphml\n",
      "---:  mutag_74.graphml\n",
      "---:  mutag_75.graphml\n",
      "---:  mutag_76.graphml\n",
      "---:  mutag_77.graphml\n",
      "---:  mutag_78.graphml\n",
      "---:  mutag_79.graphml\n",
      "---:  mutag_80.graphml\n",
      "---:  mutag_81.graphml\n",
      "---:  mutag_82.graphml\n",
      "---:  mutag_83.graphml\n",
      "---:  mutag_84.graphml\n",
      "---:  mutag_85.graphml\n",
      "---:  mutag_86.graphml\n",
      "---:  mutag_87.graphml\n",
      "---:  mutag_88.graphml\n",
      "---:  mutag_89.graphml\n",
      "---:  mutag_90.graphml\n",
      "---:  mutag_91.graphml\n",
      "---:  mutag_92.graphml\n",
      "---:  mutag_93.graphml\n",
      "---:  mutag_94.graphml\n",
      "---:  mutag_95.graphml\n",
      "---:  mutag_96.graphml\n",
      "---:  mutag_97.graphml\n",
      "---:  mutag_98.graphml\n",
      "---:  mutag_99.graphml\n",
      "---:  mutag_100.graphml\n",
      "---:  mutag_101.graphml\n",
      "---:  mutag_102.graphml\n",
      "---:  mutag_103.graphml\n",
      "---:  mutag_104.graphml\n",
      "---:  mutag_105.graphml\n",
      "---:  mutag_106.graphml\n",
      "---:  mutag_107.graphml\n",
      "---:  mutag_108.graphml\n",
      "---:  mutag_109.graphml\n",
      "---:  mutag_110.graphml\n",
      "---:  mutag_111.graphml\n",
      "---:  mutag_112.graphml\n",
      "---:  mutag_113.graphml\n",
      "---:  mutag_114.graphml\n",
      "---:  mutag_115.graphml\n",
      "---:  mutag_116.graphml\n",
      "---:  mutag_117.graphml\n",
      "---:  mutag_118.graphml\n",
      "---:  mutag_119.graphml\n",
      "---:  mutag_120.graphml\n",
      "---:  mutag_121.graphml\n",
      "---:  mutag_122.graphml\n",
      "---:  mutag_123.graphml\n",
      "---:  mutag_124.graphml\n",
      "---:  mutag_125.graphml\n",
      "---:  mutag_126.graphml\n",
      "---:  mutag_127.graphml\n",
      "---:  mutag_128.graphml\n",
      "---:  mutag_129.graphml\n",
      "---:  mutag_130.graphml\n",
      "---:  mutag_131.graphml\n",
      "---:  mutag_132.graphml\n",
      "---:  mutag_133.graphml\n",
      "---:  mutag_134.graphml\n",
      "---:  mutag_135.graphml\n",
      "---:  mutag_136.graphml\n",
      "---:  mutag_137.graphml\n",
      "---:  mutag_138.graphml\n",
      "---:  mutag_139.graphml\n",
      "---:  mutag_140.graphml\n",
      "---:  mutag_141.graphml\n",
      "---:  mutag_142.graphml\n",
      "---:  mutag_143.graphml\n",
      "---:  mutag_144.graphml\n",
      "---:  mutag_145.graphml\n",
      "---:  mutag_146.graphml\n",
      "---:  mutag_147.graphml\n",
      "---:  mutag_148.graphml\n",
      "---:  mutag_149.graphml\n",
      "---:  mutag_150.graphml\n",
      "---:  mutag_151.graphml\n",
      "---:  mutag_152.graphml\n",
      "---:  mutag_153.graphml\n",
      "---:  mutag_154.graphml\n",
      "---:  mutag_155.graphml\n",
      "---:  mutag_156.graphml\n",
      "---:  mutag_157.graphml\n",
      "---:  mutag_158.graphml\n",
      "---:  mutag_159.graphml\n",
      "---:  mutag_160.graphml\n",
      "---:  mutag_161.graphml\n",
      "---:  mutag_162.graphml\n",
      "---:  mutag_163.graphml\n",
      "---:  mutag_164.graphml\n",
      "---:  mutag_165.graphml\n",
      "---:  mutag_166.graphml\n",
      "---:  mutag_167.graphml\n",
      "---:  mutag_168.graphml\n",
      "---:  mutag_169.graphml\n",
      "---:  mutag_170.graphml\n",
      "---:  mutag_171.graphml\n",
      "---:  mutag_172.graphml\n",
      "---:  mutag_173.graphml\n",
      "---:  mutag_174.graphml\n",
      "---:  mutag_175.graphml\n",
      "---:  mutag_176.graphml\n",
      "---:  mutag_177.graphml\n",
      "---:  mutag_178.graphml\n",
      "---:  mutag_179.graphml\n",
      "---:  mutag_180.graphml\n",
      "---:  mutag_181.graphml\n",
      "---:  mutag_182.graphml\n",
      "---:  mutag_183.graphml\n",
      "---:  mutag_184.graphml\n",
      "---:  mutag_185.graphml\n",
      "---:  mutag_186.graphml\n",
      "---:  mutag_187.graphml\n",
      "---:  mutag_188.graphml\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/Macbook/Desktop/MIPT/Optimization_course/SPD_reduce/SPD.py:420: UserWarning: invalid index of a 0-dim tensor. This will be an error in PyTorch 0.5. Use tensor.item() to convert a 0-dim tensor to a Python number\n",
      "/Users/Macbook/Desktop/MIPT/Optimization_course/SPD_reduce/SPD.py:415: UserWarning: invalid index of a 0-dim tensor. This will be an error in PyTorch 0.5. Use tensor.item() to convert a 0-dim tensor to a Python number\n",
      "  1%|          | 35/4999 [00:00<00:28, 171.35it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 10, TRAIN LOSS: 12025.7685546875, TEST LOSS 194.8460693359375\n",
      "EPOCH: 20, TRAIN LOSS: 10893.1064453125, TEST LOSS 168.12779235839844\n",
      "EPOCH: 30, TRAIN LOSS: 9602.6767578125, TEST LOSS 137.49652099609375\n",
      "EPOCH: 40, TRAIN LOSS: 8409.6689453125, TEST LOSS 117.1836929321289\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  1%|▏         | 71/4999 [00:00<00:28, 173.36it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 50, TRAIN LOSS: 7694.2724609375, TEST LOSS 107.41378784179688\n",
      "EPOCH: 60, TRAIN LOSS: 7388.806640625, TEST LOSS 114.82440948486328\n",
      "EPOCH: 70, TRAIN LOSS: 7180.66650390625, TEST LOSS 117.98662567138672\n",
      "EPOCH: 80, TRAIN LOSS: 7044.8994140625, TEST LOSS 118.58479309082031\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  2%|▏         | 123/4999 [00:00<00:28, 170.08it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 90, TRAIN LOSS: 6901.1337890625, TEST LOSS 100.3840103149414\n",
      "EPOCH: 100, TRAIN LOSS: 6828.9716796875, TEST LOSS 96.12066650390625\n",
      "EPOCH: 110, TRAIN LOSS: 6858.3212890625, TEST LOSS 95.39215087890625\n",
      "EPOCH: 120, TRAIN LOSS: 6664.2431640625, TEST LOSS 96.34513854980469\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  3%|▎         | 161/4999 [00:00<00:27, 173.17it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 130, TRAIN LOSS: 6567.32177734375, TEST LOSS 98.2985610961914\n",
      "EPOCH: 140, TRAIN LOSS: 6487.99462890625, TEST LOSS 85.55653381347656\n",
      "EPOCH: 150, TRAIN LOSS: 6330.880859375, TEST LOSS 89.89641571044922\n",
      "EPOCH: 160, TRAIN LOSS: 6210.84814453125, TEST LOSS 93.99952697753906\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  4%|▍         | 198/4999 [00:01<00:27, 174.03it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 170, TRAIN LOSS: 6084.890625, TEST LOSS 95.74514770507812\n",
      "EPOCH: 180, TRAIN LOSS: 5966.3857421875, TEST LOSS 83.70118713378906\n",
      "EPOCH: 190, TRAIN LOSS: 5819.076171875, TEST LOSS 91.65496063232422\n",
      "EPOCH: 200, TRAIN LOSS: 5692.30224609375, TEST LOSS 91.95552062988281\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  5%|▍         | 234/4999 [00:01<00:27, 173.33it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 210, TRAIN LOSS: 5588.3046875, TEST LOSS 81.29997253417969\n",
      "EPOCH: 220, TRAIN LOSS: 5398.16650390625, TEST LOSS 81.06413269042969\n",
      "EPOCH: 230, TRAIN LOSS: 5372.1181640625, TEST LOSS 96.39537048339844\n",
      "EPOCH: 240, TRAIN LOSS: 5184.00146484375, TEST LOSS 84.15594482421875\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  5%|▌         | 271/4999 [00:01<00:27, 172.74it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 250, TRAIN LOSS: 5062.4677734375, TEST LOSS 96.05667114257812\n",
      "EPOCH: 260, TRAIN LOSS: 5002.328125, TEST LOSS 80.81526184082031\n",
      "EPOCH: 270, TRAIN LOSS: 4920.01611328125, TEST LOSS 89.84008026123047\n",
      "EPOCH: 280, TRAIN LOSS: 4810.1015625, TEST LOSS 92.42985534667969\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  6%|▌         | 308/4999 [00:01<00:27, 167.76it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 290, TRAIN LOSS: 4788.240234375, TEST LOSS 111.89091491699219\n",
      "EPOCH: 300, TRAIN LOSS: 4697.6318359375, TEST LOSS 95.19720458984375\n",
      "EPOCH: 310, TRAIN LOSS: 4562.81689453125, TEST LOSS 81.44309997558594\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  7%|▋         | 339/4999 [00:02<00:29, 159.14it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 320, TRAIN LOSS: 4516.25, TEST LOSS 82.58027648925781\n",
      "EPOCH: 330, TRAIN LOSS: 4613.041015625, TEST LOSS 92.15776062011719\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\r",
      "  7%|▋         | 354/4999 [00:02<00:29, 158.36it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 340, TRAIN LOSS: 4444.650390625, TEST LOSS 78.31021881103516\n",
      "EPOCH: 350, TRAIN LOSS: 4413.224609375, TEST LOSS 86.90116882324219\n",
      "EPOCH: 360, TRAIN LOSS: 4412.8671875, TEST LOSS 75.87663269042969\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  8%|▊         | 381/4999 [00:02<00:29, 154.08it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 370, TRAIN LOSS: 4365.0185546875, TEST LOSS 86.58741760253906\n",
      "EPOCH: 380, TRAIN LOSS: 4198.580078125, TEST LOSS 85.86817169189453\n",
      "EPOCH: 390, TRAIN LOSS: 4244.111328125, TEST LOSS 92.91175842285156\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  8%|▊         | 407/4999 [00:02<00:30, 150.04it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 400, TRAIN LOSS: 4088.23828125, TEST LOSS 78.73849487304688\n",
      "EPOCH: 410, TRAIN LOSS: 4190.05908203125, TEST LOSS 83.45155334472656\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  9%|▉         | 443/4999 [00:03<00:31, 144.45it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 420, TRAIN LOSS: 3961.6318359375, TEST LOSS 77.25059509277344\n",
      "EPOCH: 430, TRAIN LOSS: 4036.84130859375, TEST LOSS 82.53994750976562\n",
      "EPOCH: 440, TRAIN LOSS: 3950.26171875, TEST LOSS 74.07083892822266\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  9%|▉         | 468/4999 [00:03<00:31, 142.42it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 450, TRAIN LOSS: 3878.01611328125, TEST LOSS 77.6578369140625\n",
      "EPOCH: 460, TRAIN LOSS: 3837.8046875, TEST LOSS 84.20787048339844\n",
      "EPOCH: 470, TRAIN LOSS: 3884.576904296875, TEST LOSS 72.25751495361328\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 10%|▉         | 492/4999 [00:03<00:32, 140.52it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 480, TRAIN LOSS: 3820.401123046875, TEST LOSS 75.83910369873047\n",
      "EPOCH: 490, TRAIN LOSS: 3884.9560546875, TEST LOSS 68.76390838623047\n",
      "EPOCH: 500, TRAIN LOSS: 3792.28076171875, TEST LOSS 72.09223937988281\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 10%|█         | 523/4999 [00:03<00:31, 141.07it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 510, TRAIN LOSS: 3775.084716796875, TEST LOSS 66.59706115722656\n",
      "EPOCH: 520, TRAIN LOSS: 3722.83251953125, TEST LOSS 83.16352844238281\n",
      "EPOCH: 530, TRAIN LOSS: 3656.442626953125, TEST LOSS 66.83773803710938\n",
      "EPOCH: 540, TRAIN LOSS: 3617.1884765625, TEST LOSS"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 11%|█▏        | 569/4999 [00:04<00:31, 140.29it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 66.8174819946289\n",
      "EPOCH: 550, TRAIN LOSS: 3711.983154296875, TEST LOSS 71.2048568725586\n",
      "EPOCH: 560, TRAIN LOSS: 3656.72998046875, TEST LOSS 66.71329498291016\n",
      "EPOCH: 570, TRAIN LOSS: 3550.33056640625, TEST LOSS 83.2120361328125\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 12%|█▏        | 594/4999 [00:04<00:31, 138.28it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 580, TRAIN LOSS: 3586.3203125, TEST LOSS 77.603759765625\n",
      "EPOCH: 590, TRAIN LOSS: 3551.81640625, TEST LOSS 69.16986083984375\n",
      "EPOCH: 600, TRAIN LOSS: 3593.026611328125, TEST LOSS 76.73757934570312\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 13%|█▎        | 629/4999 [00:04<00:31, 139.75it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 610, TRAIN LOSS: 3493.11083984375, TEST LOSS 89.3577880859375\n",
      "EPOCH: 620, TRAIN LOSS: 3517.9677734375, TEST LOSS 58.01588821411133\n",
      "EPOCH: 630, TRAIN LOSS: 3412.65185546875, TEST LOSS 74.90330505371094\n",
      "EPOCH: 640, TRAIN LOSS: 3455.36572265625, TEST LOSS 64.90219116210938\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 13%|█▎        | 665/4999 [00:04<00:30, 140.96it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 650, TRAIN LOSS: 3474.36328125, TEST LOSS 67.44979858398438\n",
      "EPOCH: 660, TRAIN LOSS: 3384.404296875, TEST LOSS 71.23445892333984\n",
      "EPOCH: 670, TRAIN LOSS: 3444.09765625, TEST LOSS 71.02660369873047\n",
      "EPOCH: 680, TRAIN LOSS: 3329.8544921875, TEST LOSS 69.75025939941406\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 14%|█▍        | 718/4999 [00:05<00:30, 142.64it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 690, TRAIN LOSS: 3356.751708984375, TEST LOSS 69.91500854492188\n",
      "EPOCH: 700, TRAIN LOSS: 3319.431884765625, TEST LOSS 89.01901245117188\n",
      "EPOCH: 710, TRAIN LOSS: 3275.4580078125, TEST LOSS 71.53062438964844\n",
      "EPOCH: 720, TRAIN LOSS: 3320.1064453125, TEST LOSS 74.80609130859375\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 15%|█▌        | 753/4999 [00:05<00:29, 143.20it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 730, TRAIN LOSS: 3308.431884765625, TEST LOSS 67.58133697509766\n",
      "EPOCH: 740, TRAIN LOSS: 3235.91015625, TEST LOSS 62.74290466308594\n",
      "EPOCH: 750, TRAIN LOSS: 3209.30615234375, TEST LOSS 74.01911926269531\n",
      "EPOCH: 760, TRAIN LOSS: 3278.373046875, TEST LOSS 57.29388427734375\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 16%|█▌        | 789/4999 [00:05<00:29, 144.44it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 770, TRAIN LOSS: 3260.966796875, TEST LOSS 64.07685089111328\n",
      "EPOCH: 780, TRAIN LOSS: 3217.020263671875, TEST LOSS 80.58541870117188\n",
      "EPOCH: 790, TRAIN LOSS: 3063.1376953125, TEST LOSS 65.81718444824219\n",
      "EPOCH: 800, TRAIN LOSS: 3233.11474609375, TEST LOSS 72.3906478881836\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 17%|█▋        | 846/4999 [00:05<00:28, 146.65it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 810, TRAIN LOSS: 3207.6513671875, TEST LOSS 56.09724426269531\n",
      "EPOCH: 820, TRAIN LOSS: 3196.59814453125, TEST LOSS 59.24530792236328\n",
      "EPOCH: 830, TRAIN LOSS: 3274.859619140625, TEST LOSS 57.282081604003906\n",
      "EPOCH: 840, TRAIN LOSS: 3151.57666015625, TEST LOSS 73.91874694824219\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 18%|█▊        | 885/4999 [00:05<00:27, 147.96it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 850, TRAIN LOSS: 3055.0791015625, TEST LOSS 67.73583984375\n",
      "EPOCH: 860, TRAIN LOSS: 3109.736083984375, TEST LOSS 70.66268920898438\n",
      "EPOCH: 870, TRAIN LOSS: 3150.23193359375, TEST LOSS 53.21601867675781\n",
      "EPOCH: 880, TRAIN LOSS: 3048.3857421875, TEST LOSS 77.92149353027344\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 18%|█▊        | 923/4999 [00:06<00:27, 149.06it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 890, TRAIN LOSS: 3177.87060546875, TEST LOSS 55.773433685302734\n",
      "EPOCH: 900, TRAIN LOSS: 3066.2939453125, TEST LOSS 72.76176452636719\n",
      "EPOCH: 910, TRAIN LOSS: 3054.53955078125, TEST LOSS 66.72488403320312\n",
      "EPOCH: 920, TRAIN LOSS: 3084.76953125, TEST LOSS 58.337825775146484\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 19%|█▉        | 960/4999 [00:06<00:26, 149.86it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 930, TRAIN LOSS: 3029.522705078125, TEST LOSS 50.566925048828125\n",
      "EPOCH: 940, TRAIN LOSS: 3053.6826171875, TEST LOSS 53.486454010009766\n",
      "EPOCH: 950, TRAIN LOSS: 3040.95751953125, TEST LOSS 70.64351654052734\n",
      "EPOCH: 960, TRAIN LOSS: 2982.65869140625, TEST LOSS 76.02529907226562\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 20%|█▉        | 996/4999 [00:06<00:26, 150.51it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 970, TRAIN LOSS: 3051.890869140625, TEST LOSS 85.64360809326172\n",
      "EPOCH: 980, TRAIN LOSS: 2977.1953125, TEST LOSS 63.44834899902344\n",
      "EPOCH: 990, TRAIN LOSS: 2935.55615234375, TEST LOSS 58.7186279296875\n",
      "EPOCH: 1000, TRAIN LOSS: 2987.29931640625, TEST LOSS 79.14607238769531\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 21%|██        | 1031/4999 [00:06<00:26, 149.64it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 1010, TRAIN LOSS: 2938.29541015625, TEST LOSS 72.02265167236328\n",
      "EPOCH: 1020, TRAIN LOSS: 2959.841796875, TEST LOSS 69.29051208496094\n",
      "EPOCH: 1030, TRAIN LOSS: 2980.97802734375, TEST LOSS 71.57372283935547\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 21%|██▏       | 1064/4999 [00:07<00:26, 150.00it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 1040, TRAIN LOSS: 2889.775390625, TEST LOSS 59.285728454589844\n",
      "EPOCH: 1050, TRAIN LOSS: 2956.826416015625, TEST LOSS 81.85728454589844\n",
      "EPOCH: 1060, TRAIN LOSS: 2950.002685546875, TEST LOSS 49.577762603759766\n",
      "EPOCH: 1070, TRAIN LOSS: 2875.78955078125, TEST LOSS 61.222904205322266\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 22%|██▏       | 1098/4999 [00:07<00:25, 150.40it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 1080, TRAIN LOSS: 2876.2431640625, TEST LOSS 70.08877563476562\n",
      "EPOCH: 1090, TRAIN LOSS: 2805.68896484375, TEST LOSS 64.97931671142578\n",
      "EPOCH: 1100, TRAIN LOSS: 2842.661865234375, TEST LOSS 71.17755889892578\n",
      "EPOCH: 1110, TRAIN LOSS: 2856.61181640625, TEST LOSS 69.4129638671875\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 23%|██▎       | 1153/4999 [00:07<00:25, 151.50it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 1120, TRAIN LOSS: 2790.339599609375, TEST LOSS 77.78715515136719\n",
      "EPOCH: 1130, TRAIN LOSS: 2823.873291015625, TEST LOSS 56.965633392333984\n",
      "EPOCH: 1140, TRAIN LOSS: 2837.1318359375, TEST LOSS 58.937957763671875\n",
      "EPOCH: 1150, TRAIN LOSS: 2814.826171875, TEST LOSS 54.596466064453125\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 24%|██▍       | 1189/4999 [00:07<00:25, 151.91it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 1160, TRAIN LOSS: 2807.57470703125, TEST LOSS 55.93389892578125\n",
      "EPOCH: 1170, TRAIN LOSS: 2849.25634765625, TEST LOSS 68.0046157836914\n",
      "EPOCH: 1180, TRAIN LOSS: 2793.771728515625, TEST LOSS 62.51194763183594\n",
      "EPOCH: 1190, TRAIN LOSS: 2721.64111328125, TEST LOSS 72.43515014648438\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 25%|██▍       | 1226/4999 [00:08<00:24, 152.65it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 1200, TRAIN LOSS: 2769.234375, TEST LOSS 82.4151382446289\n",
      "EPOCH: 1210, TRAIN LOSS: 2780.145263671875, TEST LOSS 67.13475036621094\n",
      "EPOCH: 1220, TRAIN LOSS: 2838.96484375, TEST LOSS 64.35566711425781\n",
      "EPOCH: 1230, TRAIN LOSS: 2708.16259765625, TEST LOSS 87.67457580566406\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 25%|██▌       | 1262/4999 [00:08<00:24, 153.01it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 1240, TRAIN LOSS: 2705.57080078125, TEST LOSS 54.88541030883789\n",
      "EPOCH: 1250, TRAIN LOSS: 2818.457275390625, TEST LOSS 75.8180160522461\n",
      "EPOCH: 1260, TRAIN LOSS: 2712.7822265625, TEST LOSS 74.4408950805664\n",
      "EPOCH: 1270, TRAIN LOSS: 2736.861328125, TEST LOSS 69.08475494384766\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 26%|██▌       | 1297/4999 [00:08<00:24, 153.27it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 1280, TRAIN LOSS: 2780.311279296875, TEST LOSS 57.037662506103516\n",
      "EPOCH: 1290, TRAIN LOSS: 2729.058349609375, TEST LOSS 70.69198608398438\n",
      "EPOCH: 1300, TRAIN LOSS: 2740.250732421875, TEST LOSS 80.64201354980469\n",
      "EPOCH: 1310, TRAIN LOSS: 2750.93359375, TEST LOSS 73.54669952392578\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 27%|██▋       | 1353/4999 [00:08<00:23, 154.31it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 1320, TRAIN LOSS: 2664.760498046875, TEST LOSS 63.194976806640625\n",
      "EPOCH: 1330, TRAIN LOSS: 2662.845703125, TEST LOSS 70.92374420166016\n",
      "EPOCH: 1340, TRAIN LOSS: 2604.604248046875, TEST LOSS 49.66624450683594\n",
      "EPOCH: 1350, TRAIN LOSS: 2683.56982421875, TEST LOSS 66.76299285888672\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 28%|██▊       | 1389/4999 [00:08<00:23, 154.42it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 1360, TRAIN LOSS: 2644.715576171875, TEST LOSS 43.79309844970703\n",
      "EPOCH: 1370, TRAIN LOSS: 2684.5654296875, TEST LOSS 53.13848114013672\n",
      "EPOCH: 1380, TRAIN LOSS: 2628.61865234375, TEST LOSS 60.60378646850586\n",
      "EPOCH: 1390, TRAIN LOSS: 2723.364501953125, TEST LOSS 68.39921569824219\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 28%|██▊       | 1424/4999 [00:09<00:23, 154.39it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 1400, TRAIN LOSS: 2667.638916015625, TEST LOSS 69.94725036621094\n",
      "EPOCH: 1410, TRAIN LOSS: 2683.42041015625, TEST LOSS 75.67213439941406\n",
      "EPOCH: 1420, TRAIN LOSS: 2595.01611328125, TEST LOSS 64.51237487792969\n",
      "EPOCH: 1430, TRAIN LOSS: 2596.36865234375, TEST LOSS 46.81591796875\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 29%|██▉       | 1458/4999 [00:09<00:22, 154.49it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 1440, TRAIN LOSS: 2672.15234375, TEST LOSS 69.0604019165039\n",
      "EPOCH: 1450, TRAIN LOSS: 2582.65673828125, TEST LOSS 95.16786193847656\n",
      "EPOCH: 1460, TRAIN LOSS: 2585.650634765625, TEST LOSS 65.59832763671875\n",
      "EPOCH: 1470, TRAIN LOSS: 2588.060546875, TEST LOSS 60.99700164794922\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 30%|██▉       | 1493/4999 [00:09<00:22, 154.84it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 1480, TRAIN LOSS: 2591.533203125, TEST LOSS 66.62162780761719\n",
      "EPOCH: 1490, TRAIN LOSS: 2711.818359375, TEST LOSS 54.27493667602539\n",
      "EPOCH: 1500, TRAIN LOSS: 2534.212646484375, TEST LOSS 56.11487579345703\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 31%|███       | 1529/4999 [00:09<00:22, 154.36it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 1510, TRAIN LOSS: 2570.4736328125, TEST LOSS 65.36408996582031\n",
      "EPOCH: 1520, TRAIN LOSS: 2517.669189453125, TEST LOSS 54.8070068359375\n",
      "EPOCH: 1530, TRAIN LOSS: 2604.7392578125, TEST LOSS 69.01535034179688\n",
      "EPOCH: 1540, TRAIN LOSS: 2593.766845703125, TEST LOSS 52.87445068359375\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 31%|███▏      | 1566/4999 [00:10<00:22, 154.90it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 1550, TRAIN LOSS: 2510.48779296875, TEST LOSS 55.703941345214844\n",
      "EPOCH: 1560, TRAIN LOSS: 2620.7177734375, TEST LOSS 41.364349365234375\n",
      "EPOCH: 1570, TRAIN LOSS: 2534.7265625, TEST LOSS 51.28819274902344\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 32%|███▏      | 1600/4999 [00:10<00:22, 154.39it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 1580, TRAIN LOSS: 2585.599609375, TEST LOSS 58.31455993652344\n",
      "EPOCH: 1590, TRAIN LOSS: 2575.40087890625, TEST LOSS 66.46778869628906\n",
      "EPOCH: 1600, TRAIN LOSS: 2576.458251953125, TEST LOSS 47.03121566772461\n",
      "EPOCH: 1610, TRAIN LOSS: 2604.273193359375, TEST LOSS 74.57105255126953\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 33%|███▎      | 1650/4999 [00:10<00:21, 154.48it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 1620, TRAIN LOSS: 2569.724609375, TEST LOSS 57.43207550048828\n",
      "EPOCH: 1630, TRAIN LOSS: 2583.63671875, TEST LOSS 66.36903381347656\n",
      "EPOCH: 1640, TRAIN LOSS: 2528.447021484375, TEST LOSS 50.43230056762695\n",
      "EPOCH: 1650, TRAIN LOSS: 2532.707763671875, TEST LOSS 55.875267028808594\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 34%|███▎      | 1687/4999 [00:10<00:21, 154.92it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 1660, TRAIN LOSS: 2554.895751953125, TEST LOSS 56.77922058105469\n",
      "EPOCH: 1670, TRAIN LOSS: 2501.601318359375, TEST LOSS 62.21728515625\n",
      "EPOCH: 1680, TRAIN LOSS: 2464.556640625, TEST LOSS 56.25313949584961\n",
      "EPOCH: 1690, TRAIN LOSS: 2453.8798828125, TEST LOSS 51.025779724121094\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 34%|███▍      | 1720/4999 [00:11<00:21, 154.59it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 1700, TRAIN LOSS: 2416.80126953125, TEST LOSS 78.55474090576172\n",
      "EPOCH: 1710, TRAIN LOSS: 2496.32763671875, TEST LOSS 66.54532623291016\n",
      "EPOCH: 1720, TRAIN LOSS: 2426.94091796875, TEST LOSS 65.0513687133789\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 35%|███▌      | 1755/4999 [00:11<00:20, 154.88it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 1730, TRAIN LOSS: 2424.28125, TEST LOSS 60.408546447753906\n",
      "EPOCH: 1740, TRAIN LOSS: 2525.494873046875, TEST LOSS 83.38349914550781\n",
      "EPOCH: 1750, TRAIN LOSS: 2506.705078125, TEST LOSS 53.913063049316406\n",
      "EPOCH: 1760, TRAIN LOSS: 2488.469970703125, TEST LOSS 59.253639221191406\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 36%|███▌      | 1792/4999 [00:11<00:20, 155.30it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 1770, TRAIN LOSS: 2403.087890625, TEST LOSS 73.9764175415039\n",
      "EPOCH: 1780, TRAIN LOSS: 2472.500244140625, TEST LOSS 77.71036529541016\n",
      "EPOCH: 1790, TRAIN LOSS: 2504.004150390625, TEST LOSS 63.91400909423828\n",
      "EPOCH: 1800, TRAIN LOSS: 2469.027099609375, TEST LOSS 49.897457122802734\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 37%|███▋      | 1829/4999 [00:11<00:20, 155.72it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 1810, TRAIN LOSS: 2382.94970703125, TEST LOSS 52.38528060913086\n",
      "EPOCH: 1820, TRAIN LOSS: 2456.559814453125, TEST LOSS 53.74964904785156\n",
      "EPOCH: 1830, TRAIN LOSS: 2514.697021484375, TEST LOSS 52.426265716552734\n",
      "EPOCH: 1840, TRAIN LOSS: 2361.2763671875, TEST LOSS 74.8202896118164\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 38%|███▊      | 1885/4999 [00:12<00:19, 156.27it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 1850, TRAIN LOSS: 2419.27783203125, TEST LOSS 64.7063217163086\n",
      "EPOCH: 1860, TRAIN LOSS: 2461.103271484375, TEST LOSS 53.31297302246094\n",
      "EPOCH: 1870, TRAIN LOSS: 2430.021728515625, TEST LOSS 69.30891418457031\n",
      "EPOCH: 1880, TRAIN LOSS: 2516.24072265625, TEST LOSS 56.6904296875\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 38%|███▊      | 1923/4999 [00:12<00:19, 156.66it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 1890, TRAIN LOSS: 2394.31005859375, TEST LOSS 53.39031219482422\n",
      "EPOCH: 1900, TRAIN LOSS: 2448.7216796875, TEST LOSS 55.824737548828125\n",
      "EPOCH: 1910, TRAIN LOSS: 2468.293701171875, TEST LOSS 53.43516159057617\n",
      "EPOCH: 1920, TRAIN LOSS: 2460.40576171875, TEST LOSS 58.199771881103516\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 39%|███▉      | 1962/4999 [00:12<00:19, 157.21it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 1930, TRAIN LOSS: 2441.18994140625, TEST LOSS 62.41214370727539\n",
      "EPOCH: 1940, TRAIN LOSS: 2451.64501953125, TEST LOSS 62.088539123535156\n",
      "EPOCH: 1950, TRAIN LOSS: 2385.407470703125, TEST LOSS 70.517822265625\n",
      "EPOCH: 1960, TRAIN LOSS: 2456.801513671875, TEST LOSS 58.90534210205078\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 40%|████      | 2001/4999 [00:12<00:19, 157.73it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 1970, TRAIN LOSS: 2377.04833984375, TEST LOSS 50.017578125\n",
      "EPOCH: 1980, TRAIN LOSS: 2365.581298828125, TEST LOSS 57.639774322509766\n",
      "EPOCH: 1990, TRAIN LOSS: 2353.47607421875, TEST LOSS 73.58148193359375\n",
      "EPOCH: 2000, TRAIN LOSS: 2306.3564453125, TEST LOSS 63.363285064697266\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 41%|████      | 2039/4999 [00:12<00:18, 158.16it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 2010, TRAIN LOSS: 2312.174072265625, TEST LOSS 49.28786087036133\n",
      "EPOCH: 2020, TRAIN LOSS: 2358.52880859375, TEST LOSS 49.00982666015625\n",
      "EPOCH: 2030, TRAIN LOSS: 2368.22900390625, TEST LOSS 57.90654754638672\n",
      "EPOCH: 2040, TRAIN LOSS: 2405.524169921875, TEST LOSS 51.9130744934082\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 42%|████▏     | 2077/4999 [00:13<00:18, 158.54it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 2050, TRAIN LOSS: 2286.830810546875, TEST LOSS 56.57178497314453\n",
      "EPOCH: 2060, TRAIN LOSS: 2404.41552734375, TEST LOSS 47.20049285888672\n",
      "EPOCH: 2070, TRAIN LOSS: 2381.53955078125, TEST LOSS 47.08343505859375\n",
      "EPOCH: 2080, TRAIN LOSS: 2297.135498046875, TEST LOSS 70.07039642333984\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 42%|████▏     | 2115/4999 [00:13<00:18, 158.74it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 2090, TRAIN LOSS: 2369.41748046875, TEST LOSS 68.46885681152344\n",
      "EPOCH: 2100, TRAIN LOSS: 2395.10888671875, TEST LOSS 70.37513732910156\n",
      "EPOCH: 2110, TRAIN LOSS: 2380.2998046875, TEST LOSS 57.15585708618164\n",
      "EPOCH: 2120, TRAIN LOSS: 2363.1533203125, TEST LOSS 61.048789978027344\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 43%|████▎     | 2155/4999 [00:13<00:17, 159.29it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 2130, TRAIN LOSS: 2293.86376953125, TEST LOSS 66.46609497070312\n",
      "EPOCH: 2140, TRAIN LOSS: 2306.349609375, TEST LOSS 69.13490295410156\n",
      "EPOCH: 2150, TRAIN LOSS: 2400.99658203125, TEST LOSS 76.39861297607422\n",
      "EPOCH: 2160, TRAIN LOSS: 2305.29296875, TEST LOSS 43.06189727783203\n",
      "EPOCH: 2170, TRAIN LOSS: 2269.136474609375, TEST LOSS"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 44%|████▍     | 2193/4999 [00:13<00:17, 158.94it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 57.61634063720703\n",
      "EPOCH: 2180, TRAIN LOSS: 2283.1962890625, TEST LOSS 44.98870849609375\n",
      "EPOCH: 2190, TRAIN LOSS: 2333.6552734375, TEST LOSS 50.65382385253906\n",
      "EPOCH: 2200, TRAIN LOSS: 2365.16015625, TEST LOSS 51.931800842285156\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 45%|████▍     | 2229/4999 [00:14<00:17, 159.18it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 2210, TRAIN LOSS: 2311.96044921875, TEST LOSS 51.424129486083984\n",
      "EPOCH: 2220, TRAIN LOSS: 2337.6640625, TEST LOSS 80.87760925292969\n",
      "EPOCH: 2230, TRAIN LOSS: 2229.75341796875, TEST LOSS 56.67884063720703\n",
      "EPOCH: 2240, TRAIN LOSS: 2218.75, TEST LOSS 64.34320068359375\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 46%|████▌     | 2287/4999 [00:14<00:16, 159.83it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 2250, TRAIN LOSS: 2344.846923828125, TEST LOSS 74.02134704589844\n",
      "EPOCH: 2260, TRAIN LOSS: 2365.67529296875, TEST LOSS 59.953121185302734\n",
      "EPOCH: 2270, TRAIN LOSS: 2239.79736328125, TEST LOSS 66.49934387207031\n",
      "EPOCH: 2280, TRAIN LOSS: 2282.697021484375, TEST LOSS 38.789485931396484\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 47%|████▋     | 2327/4999 [00:14<00:16, 160.31it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 2290, TRAIN LOSS: 2325.85693359375, TEST LOSS 42.65321350097656\n",
      "EPOCH: 2300, TRAIN LOSS: 2319.970947265625, TEST LOSS 53.45872497558594\n",
      "EPOCH: 2310, TRAIN LOSS: 2342.53369140625, TEST LOSS 48.152557373046875\n",
      "EPOCH: 2320, TRAIN LOSS: 2225.166259765625, TEST LOSS 64.04635620117188\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 47%|████▋     | 2365/4999 [00:14<00:16, 160.68it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 2330, TRAIN LOSS: 2250.759033203125, TEST LOSS 57.12000274658203\n",
      "EPOCH: 2340, TRAIN LOSS: 2334.5888671875, TEST LOSS 53.5019645690918\n",
      "EPOCH: 2350, TRAIN LOSS: 2354.213134765625, TEST LOSS 49.25639343261719\n",
      "EPOCH: 2360, TRAIN LOSS: 2186.153564453125, TEST LOSS 53.94501495361328\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 48%|████▊     | 2404/4999 [00:14<00:16, 161.01it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 2370, TRAIN LOSS: 2386.520751953125, TEST LOSS 59.31389236450195\n",
      "EPOCH: 2380, TRAIN LOSS: 2277.744140625, TEST LOSS 76.68576049804688\n",
      "EPOCH: 2390, TRAIN LOSS: 2330.19873046875, TEST LOSS 64.73509216308594\n",
      "EPOCH: 2400, TRAIN LOSS: 2162.71337890625, TEST LOSS 53.87666320800781\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 49%|████▉     | 2442/4999 [00:15<00:15, 161.34it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 2410, TRAIN LOSS: 2286.52001953125, TEST LOSS 55.56943130493164\n",
      "EPOCH: 2420, TRAIN LOSS: 2183.7724609375, TEST LOSS 50.6483154296875\n",
      "EPOCH: 2430, TRAIN LOSS: 2299.85595703125, TEST LOSS 44.1198616027832\n",
      "EPOCH: 2440, TRAIN LOSS: 2211.9599609375, TEST LOSS 46.71657180786133\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 50%|████▉     | 2480/4999 [00:15<00:15, 161.51it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 2450, TRAIN LOSS: 2267.412353515625, TEST LOSS 57.69116973876953\n",
      "EPOCH: 2460, TRAIN LOSS: 2255.66064453125, TEST LOSS 52.82275390625\n",
      "EPOCH: 2470, TRAIN LOSS: 2364.537353515625, TEST LOSS 71.06201171875\n",
      "EPOCH: 2480, TRAIN LOSS: 2255.00927734375, TEST LOSS 54.877193450927734\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 50%|█████     | 2518/4999 [00:15<00:15, 161.58it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 2490, TRAIN LOSS: 2249.087158203125, TEST LOSS 47.07874298095703\n",
      "EPOCH: 2500, TRAIN LOSS: 2255.5654296875, TEST LOSS 61.26422119140625\n",
      "EPOCH: 2510, TRAIN LOSS: 2361.43798828125, TEST LOSS 57.5023193359375\n",
      "EPOCH: 2520, TRAIN LOSS: 2269.196044921875, TEST LOSS 49.207515716552734\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 51%|█████     | 2555/4999 [00:15<00:15, 161.86it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 2530, TRAIN LOSS: 2218.83056640625, TEST LOSS 56.84020233154297\n",
      "EPOCH: 2540, TRAIN LOSS: 2202.7431640625, TEST LOSS 65.96517944335938\n",
      "EPOCH: 2550, TRAIN LOSS: 2244.2958984375, TEST LOSS 50.90547180175781\n",
      "EPOCH: 2560, TRAIN LOSS: 2184.870849609375, TEST LOSS 71.04054260253906\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 52%|█████▏    | 2593/4999 [00:15<00:14, 162.17it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 2570, TRAIN LOSS: 2232.095703125, TEST LOSS 65.2898178100586\n",
      "EPOCH: 2580, TRAIN LOSS: 2248.326904296875, TEST LOSS 79.25486755371094\n",
      "EPOCH: 2590, TRAIN LOSS: 2266.335205078125, TEST LOSS 51.435508728027344\n",
      "EPOCH: 2600, TRAIN LOSS: 2239.49072265625, TEST LOSS 85.74420166015625\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 53%|█████▎    | 2631/4999 [00:16<00:14, 162.34it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 2610, TRAIN LOSS: 2257.0048828125, TEST LOSS 62.785552978515625\n",
      "EPOCH: 2620, TRAIN LOSS: 2132.30712890625, TEST LOSS 67.6385269165039\n",
      "EPOCH: 2630, TRAIN LOSS: 2209.573974609375, TEST LOSS 62.82867431640625\n",
      "EPOCH: 2640, TRAIN LOSS: 2229.543701171875, TEST LOSS 53.842567443847656\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 54%|█████▍    | 2687/4999 [00:16<00:14, 162.72it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 2650, TRAIN LOSS: 2184.91552734375, TEST LOSS 49.112545013427734\n",
      "EPOCH: 2660, TRAIN LOSS: 2182.184326171875, TEST LOSS 44.10333251953125\n",
      "EPOCH: 2670, TRAIN LOSS: 2192.7490234375, TEST LOSS 65.175048828125\n",
      "EPOCH: 2680, TRAIN LOSS: 2234.89013671875, TEST LOSS 58.10491943359375\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\r",
      " 54%|█████▍    | 2706/4999 [00:16<00:14, 162.78it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 2690, TRAIN LOSS: 2169.61865234375, TEST LOSS 62.681331634521484\n",
      "EPOCH: 2700, TRAIN LOSS: 2229.4951171875, TEST LOSS 54.34508514404297\n",
      "EPOCH: 2710, TRAIN LOSS: 2146.830810546875, TEST LOSS 44.44953536987305\n",
      "EPOCH: 2720, TRAIN LOSS: 2176.5859375, TEST LOSS 68.16629028320312\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 55%|█████▌    | 2760/4999 [00:16<00:13, 162.82it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 2730, TRAIN LOSS: 2197.687255859375, TEST LOSS 48.589576721191406\n",
      "EPOCH: 2740, TRAIN LOSS: 2100.8779296875, TEST LOSS 59.81328201293945\n",
      "EPOCH: 2750, TRAIN LOSS: 2222.338134765625, TEST LOSS 58.784568786621094\n",
      "EPOCH: 2760, TRAIN LOSS: 2140.097900390625, TEST LOSS 52.46299743652344\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 56%|█████▌    | 2798/4999 [00:17<00:13, 163.07it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 2770, TRAIN LOSS: 2134.81884765625, TEST LOSS 52.991397857666016\n",
      "EPOCH: 2780, TRAIN LOSS: 2176.019287109375, TEST LOSS 51.132598876953125\n",
      "EPOCH: 2790, TRAIN LOSS: 2169.18408203125, TEST LOSS 63.30195236206055\n",
      "EPOCH: 2800, TRAIN LOSS: 2144.59423828125, TEST LOSS 86.16069030761719\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 57%|█████▋    | 2838/4999 [00:17<00:13, 163.43it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 2810, TRAIN LOSS: 2183.259521484375, TEST LOSS 63.744956970214844\n",
      "EPOCH: 2820, TRAIN LOSS: 2166.8642578125, TEST LOSS 52.051082611083984\n",
      "EPOCH: 2830, TRAIN LOSS: 2166.78125, TEST LOSS 71.36126708984375\n",
      "EPOCH: 2840, TRAIN LOSS: 2089.744140625, TEST LOSS 67.10308074951172\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 58%|█████▊    | 2876/4999 [00:17<00:12, 163.57it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 2850, TRAIN LOSS: 2145.28759765625, TEST LOSS 75.1706771850586\n",
      "EPOCH: 2860, TRAIN LOSS: 2110.6171875, TEST LOSS 51.05527877807617\n",
      "EPOCH: 2870, TRAIN LOSS: 2183.945556640625, TEST LOSS 70.55304718017578\n",
      "EPOCH: 2880, TRAIN LOSS: 2197.7109375, TEST LOSS 62.951141357421875\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 58%|█████▊    | 2915/4999 [00:17<00:12, 163.76it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 2890, TRAIN LOSS: 2121.62451171875, TEST LOSS 55.40375518798828\n",
      "EPOCH: 2900, TRAIN LOSS: 2205.50830078125, TEST LOSS 57.96319580078125\n",
      "EPOCH: 2910, TRAIN LOSS: 2269.4814453125, TEST LOSS 48.47742462158203\n",
      "EPOCH: 2920, TRAIN LOSS: 2164.5439453125, TEST LOSS 53.96907424926758\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 59%|█████▉    | 2953/4999 [00:18<00:12, 163.98it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 2930, TRAIN LOSS: 2169.8017578125, TEST LOSS 49.070865631103516\n",
      "EPOCH: 2940, TRAIN LOSS: 2098.277587890625, TEST LOSS 75.76344299316406\n",
      "EPOCH: 2950, TRAIN LOSS: 2112.38427734375, TEST LOSS 80.20883178710938\n",
      "EPOCH: 2960, TRAIN LOSS: 2145.849365234375, TEST LOSS 63.734153747558594\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 60%|█████▉    | 2992/4999 [00:18<00:12, 164.26it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 2970, TRAIN LOSS: 2185.881591796875, TEST LOSS 78.10728454589844\n",
      "EPOCH: 2980, TRAIN LOSS: 2163.3515625, TEST LOSS 55.064430236816406\n",
      "EPOCH: 2990, TRAIN LOSS: 2193.623291015625, TEST LOSS 48.747154235839844\n",
      "EPOCH: 3000, TRAIN LOSS: 2172.346923828125, TEST LOSS 53.64409637451172\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 61%|██████    | 3030/4999 [00:18<00:11, 164.45it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 3010, TRAIN LOSS: 2082.13720703125, TEST LOSS 62.66688919067383\n",
      "EPOCH: 3020, TRAIN LOSS: 2114.4580078125, TEST LOSS 75.96720886230469\n",
      "EPOCH: 3030, TRAIN LOSS: 2181.58642578125, TEST LOSS 60.8073616027832\n",
      "EPOCH: 3040, TRAIN LOSS: 2031.0791015625, TEST LOSS 60.50005340576172\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 62%|██████▏   | 3088/4999 [00:18<00:11, 164.85it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 3050, TRAIN LOSS: 2107.611328125, TEST LOSS 55.39567184448242\n",
      "EPOCH: 3060, TRAIN LOSS: 2091.650634765625, TEST LOSS 75.97535705566406\n",
      "EPOCH: 3070, TRAIN LOSS: 2115.836669921875, TEST LOSS 50.09797668457031\n",
      "EPOCH: 3080, TRAIN LOSS: 2110.676025390625, TEST LOSS 57.411888122558594\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 63%|██████▎   | 3126/4999 [00:18<00:11, 165.08it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 3090, TRAIN LOSS: 2119.937255859375, TEST LOSS 62.05194854736328\n",
      "EPOCH: 3100, TRAIN LOSS: 2104.256591796875, TEST LOSS 56.166202545166016\n",
      "EPOCH: 3110, TRAIN LOSS: 2094.99072265625, TEST LOSS 59.94209289550781\n",
      "EPOCH: 3120, TRAIN LOSS: 2087.612060546875, TEST LOSS 46.192626953125\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 63%|██████▎   | 3164/4999 [00:19<00:11, 165.26it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 3130, TRAIN LOSS: 2081.023681640625, TEST LOSS 37.23591613769531\n",
      "EPOCH: 3140, TRAIN LOSS: 2179.16650390625, TEST LOSS 53.343833923339844\n",
      "EPOCH: 3150, TRAIN LOSS: 2085.25146484375, TEST LOSS 81.90038299560547\n",
      "EPOCH: 3160, TRAIN LOSS: 2155.7587890625, TEST LOSS 58.16061782836914\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 64%|██████▍   | 3202/4999 [00:19<00:10, 165.41it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 3170, TRAIN LOSS: 2102.088623046875, TEST LOSS 48.20731735229492\n",
      "EPOCH: 3180, TRAIN LOSS: 2108.41650390625, TEST LOSS 62.19799041748047\n",
      "EPOCH: 3190, TRAIN LOSS: 2136.376953125, TEST LOSS 67.4444580078125\n",
      "EPOCH: 3200, TRAIN LOSS: 2068.18017578125, TEST LOSS 49.97957229614258\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 65%|██████▍   | 3240/4999 [00:19<00:10, 165.57it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 3210, TRAIN LOSS: 2126.14013671875, TEST LOSS 52.028953552246094\n",
      "EPOCH: 3220, TRAIN LOSS: 2059.247802734375, TEST LOSS 55.50469970703125\n",
      "EPOCH: 3230, TRAIN LOSS: 2033.5035400390625, TEST LOSS 68.07865905761719\n",
      "EPOCH: 3240, TRAIN LOSS: 2052.8359375, TEST LOSS 39.56742858886719\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 66%|██████▌   | 3278/4999 [00:19<00:10, 165.70it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 3250, TRAIN LOSS: 2081.48095703125, TEST LOSS 47.80378723144531\n",
      "EPOCH: 3260, TRAIN LOSS: 2195.697265625, TEST LOSS 53.35551071166992\n",
      "EPOCH: 3270, TRAIN LOSS: 2100.371826171875, TEST LOSS 56.21418380737305\n",
      "EPOCH: 3280, TRAIN LOSS: 2156.45654296875, TEST LOSS 44.70524597167969\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 66%|██████▋   | 3315/4999 [00:19<00:10, 165.85it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 3290, TRAIN LOSS: 2096.382568359375, TEST LOSS 62.128211975097656\n",
      "EPOCH: 3300, TRAIN LOSS: 2154.525390625, TEST LOSS 50.368377685546875\n",
      "EPOCH: 3310, TRAIN LOSS: 2094.91357421875, TEST LOSS 66.93871307373047\n",
      "EPOCH: 3320, TRAIN LOSS: 2110.754638671875, TEST LOSS 41.02085876464844\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 67%|██████▋   | 3353/4999 [00:20<00:09, 166.01it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 3330, TRAIN LOSS: 2061.50732421875, TEST LOSS 58.30023193359375\n",
      "EPOCH: 3340, TRAIN LOSS: 2091.41357421875, TEST LOSS 59.506595611572266\n",
      "EPOCH: 3350, TRAIN LOSS: 2092.95458984375, TEST LOSS 54.20311737060547\n",
      "EPOCH: 3360, TRAIN LOSS: 2050.531005859375, TEST LOSS 82.00444030761719\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 68%|██████▊   | 3391/4999 [00:20<00:09, 166.15it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 3370, TRAIN LOSS: 2078.578125, TEST LOSS 55.72557067871094\n",
      "EPOCH: 3380, TRAIN LOSS: 2162.8330078125, TEST LOSS 54.71666717529297\n",
      "EPOCH: 3390, TRAIN LOSS: 2148.669677734375, TEST LOSS 65.3474349975586\n",
      "EPOCH: 3400, TRAIN LOSS: 2133.510986328125, TEST LOSS 81.24114990234375\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 69%|██████▊   | 3429/4999 [00:20<00:09, 166.31it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 3410, TRAIN LOSS: 2064.98046875, TEST LOSS 66.49407196044922\n",
      "EPOCH: 3420, TRAIN LOSS: 2050.37744140625, TEST LOSS 53.004173278808594\n",
      "EPOCH: 3430, TRAIN LOSS: 2006.4368896484375, TEST LOSS 62.682838439941406\n",
      "EPOCH: 3440, TRAIN LOSS: 2057.18798828125, TEST LOSS 71.1197280883789\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 70%|██████▉   | 3484/4999 [00:20<00:09, 166.44it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 3450, TRAIN LOSS: 2094.03076171875, TEST LOSS 42.03459548950195\n",
      "EPOCH: 3460, TRAIN LOSS: 2041.04052734375, TEST LOSS 75.49488830566406\n",
      "EPOCH: 3470, TRAIN LOSS: 2084.0458984375, TEST LOSS 56.731529235839844\n",
      "EPOCH: 3480, TRAIN LOSS: 2090.161865234375, TEST LOSS 64.83951568603516\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 70%|███████   | 3520/4999 [00:21<00:08, 166.41it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 3490, TRAIN LOSS: 2070.45166015625, TEST LOSS 55.637237548828125\n",
      "EPOCH: 3500, TRAIN LOSS: 2068.068359375, TEST LOSS 47.706573486328125\n",
      "EPOCH: 3510, TRAIN LOSS: 2090.1982421875, TEST LOSS 52.175899505615234\n",
      "EPOCH: 3520, TRAIN LOSS: 1996.982421875, TEST LOSS 66.49296569824219\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 71%|███████   | 3557/4999 [00:21<00:08, 166.48it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 3530, TRAIN LOSS: 2020.826416015625, TEST LOSS 56.065185546875\n",
      "EPOCH: 3540, TRAIN LOSS: 2086.2822265625, TEST LOSS 58.98755645751953\n",
      "EPOCH: 3550, TRAIN LOSS: 2095.6494140625, TEST LOSS 47.872802734375\n",
      "EPOCH: 3560, TRAIN LOSS: 2024.3363037109375, TEST LOSS 54.63603210449219\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 72%|███████▏  | 3594/4999 [00:21<00:08, 166.53it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 3570, TRAIN LOSS: 2032.8748779296875, TEST LOSS 49.87078094482422\n",
      "EPOCH: 3580, TRAIN LOSS: 2054.438232421875, TEST LOSS 46.629791259765625\n",
      "EPOCH: 3590, TRAIN LOSS: 2101.260009765625, TEST LOSS 59.90672302246094\n",
      "EPOCH: 3600, TRAIN LOSS: 2063.22216796875, TEST LOSS 65.0179443359375\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 73%|███████▎  | 3632/4999 [00:21<00:08, 166.69it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 3610, TRAIN LOSS: 2096.489990234375, TEST LOSS 47.36033630371094\n",
      "EPOCH: 3620, TRAIN LOSS: 2081.634521484375, TEST LOSS 57.52577209472656\n",
      "EPOCH: 3630, TRAIN LOSS: 1999.2291259765625, TEST LOSS 59.80122756958008\n",
      "EPOCH: 3640, TRAIN LOSS: 2079.44970703125, TEST LOSS 78.31433868408203\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 73%|███████▎  | 3668/4999 [00:21<00:07, 166.75it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 3650, TRAIN LOSS: 2079.126220703125, TEST LOSS 58.04450607299805\n",
      "EPOCH: 3660, TRAIN LOSS: 2044.543212890625, TEST LOSS 54.13657760620117\n",
      "EPOCH: 3670, TRAIN LOSS: 2089.654052734375, TEST LOSS 76.7908935546875\n",
      "EPOCH: 3680, TRAIN LOSS: 2067.1728515625, TEST LOSS 46.70761489868164\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 74%|███████▍  | 3722/4999 [00:22<00:07, 166.82it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 3690, TRAIN LOSS: 2210.531005859375, TEST LOSS 79.89225769042969\n",
      "EPOCH: 3700, TRAIN LOSS: 2089.263671875, TEST LOSS 61.80740737915039\n",
      "EPOCH: 3710, TRAIN LOSS: 2073.63232421875, TEST LOSS 84.80342864990234\n",
      "EPOCH: 3720, TRAIN LOSS: 2044.89013671875, TEST LOSS 73.69059753417969\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 75%|███████▌  | 3759/4999 [00:22<00:07, 166.96it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 3730, TRAIN LOSS: 2074.962158203125, TEST LOSS 64.26992797851562\n",
      "EPOCH: 3740, TRAIN LOSS: 2065.273193359375, TEST LOSS 48.06717300415039\n",
      "EPOCH: 3750, TRAIN LOSS: 2023.5230712890625, TEST LOSS 78.64096069335938\n",
      "EPOCH: 3760, TRAIN LOSS: 2051.577392578125, TEST LOSS 62.594078063964844\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 76%|███████▌  | 3795/4999 [00:22<00:07, 167.05it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 3770, TRAIN LOSS: 2029.703369140625, TEST LOSS 66.04579162597656\n",
      "EPOCH: 3780, TRAIN LOSS: 2010.7080078125, TEST LOSS 57.06034851074219\n",
      "EPOCH: 3790, TRAIN LOSS: 1984.08154296875, TEST LOSS 63.37679672241211\n",
      "EPOCH: 3800, TRAIN LOSS: 2133.033935546875, TEST LOSS 47.567413330078125\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 77%|███████▋  | 3831/4999 [00:22<00:06, 167.08it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 3810, TRAIN LOSS: 2035.8265380859375, TEST LOSS 59.23109436035156\n",
      "EPOCH: 3820, TRAIN LOSS: 2034.372802734375, TEST LOSS 50.82428741455078\n",
      "EPOCH: 3830, TRAIN LOSS: 2061.92724609375, TEST LOSS 75.53189086914062\n",
      "EPOCH: 3840, TRAIN LOSS: 2071.634765625, TEST LOSS 52.128822326660156\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 77%|███████▋  | 3867/4999 [00:23<00:06, 166.94it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 3850, TRAIN LOSS: 2080.47509765625, TEST LOSS 57.82633972167969\n",
      "EPOCH: 3860, TRAIN LOSS: 2098.607666015625, TEST LOSS 75.45792388916016\n",
      "EPOCH: 3870, TRAIN LOSS: 2040.411376953125, TEST LOSS 83.59524536132812\n",
      "EPOCH: 3880, TRAIN LOSS: 2019.3094482421875, TEST LOSS 60.1749267578125\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 78%|███████▊  | 3921/4999 [00:23<00:06, 167.05it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 3890, TRAIN LOSS: 2018.2308349609375, TEST LOSS 89.86080932617188\n",
      "EPOCH: 3900, TRAIN LOSS: 2100.78955078125, TEST LOSS 64.58430480957031\n",
      "EPOCH: 3910, TRAIN LOSS: 2038.279541015625, TEST LOSS 46.94025421142578\n",
      "EPOCH: 3920, TRAIN LOSS: 2075.481689453125, TEST LOSS 69.04391479492188\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 79%|███████▉  | 3958/4999 [00:23<00:06, 167.16it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 3930, TRAIN LOSS: 2162.881103515625, TEST LOSS 79.59993743896484\n",
      "EPOCH: 3940, TRAIN LOSS: 2070.5146484375, TEST LOSS 64.14996337890625\n",
      "EPOCH: 3950, TRAIN LOSS: 2008.936767578125, TEST LOSS 84.05148315429688\n",
      "EPOCH: 3960, TRAIN LOSS: 1950.317138671875, TEST LOSS 49.2220458984375\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 80%|███████▉  | 3995/4999 [00:23<00:06, 167.28it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 3970, TRAIN LOSS: 2023.3775634765625, TEST LOSS 64.27300262451172\n",
      "EPOCH: 3980, TRAIN LOSS: 2018.728515625, TEST LOSS 61.02723693847656\n",
      "EPOCH: 3990, TRAIN LOSS: 2025.6278076171875, TEST LOSS 68.65361022949219\n",
      "EPOCH: 4000, TRAIN LOSS: 2049.760986328125, TEST LOSS 55.3555793762207\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 81%|████████  | 4032/4999 [00:24<00:05, 167.39it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 4010, TRAIN LOSS: 2002.0247802734375, TEST LOSS 67.84590911865234\n",
      "EPOCH: 4020, TRAIN LOSS: 2010.962646484375, TEST LOSS 64.33390045166016\n",
      "EPOCH: 4030, TRAIN LOSS: 2076.392822265625, TEST LOSS 70.49857330322266\n",
      "EPOCH: 4040, TRAIN LOSS: 1932.4632568359375, TEST LOSS 52.01563262939453\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 81%|████████▏ | 4069/4999 [00:24<00:05, 167.48it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 4050, TRAIN LOSS: 2062.039794921875, TEST LOSS 62.664764404296875\n",
      "EPOCH: 4060, TRAIN LOSS: 2085.5380859375, TEST LOSS 67.8922348022461\n",
      "EPOCH: 4070, TRAIN LOSS: 2021.173583984375, TEST LOSS 52.04295349121094\n",
      "EPOCH: 4080, TRAIN LOSS: 1950.44580078125, TEST LOSS 53.781646728515625\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 82%|████████▏ | 4124/4999 [00:24<00:05, 167.57it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 4090, TRAIN LOSS: 1930.5750732421875, TEST LOSS 100.15518951416016\n",
      "EPOCH: 4100, TRAIN LOSS: 2079.396728515625, TEST LOSS 48.641998291015625\n",
      "EPOCH: 4110, TRAIN LOSS: 2032.712890625, TEST LOSS 61.629981994628906\n",
      "EPOCH: 4120, TRAIN LOSS: 2032.999267578125, TEST LOSS 68.70059204101562\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 83%|████████▎ | 4162/4999 [00:24<00:04, 167.70it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 4130, TRAIN LOSS: 1988.38671875, TEST LOSS 48.47610855102539\n",
      "EPOCH: 4140, TRAIN LOSS: 2097.38818359375, TEST LOSS 80.03334045410156\n",
      "EPOCH: 4150, TRAIN LOSS: 1976.88330078125, TEST LOSS 64.63805389404297\n",
      "EPOCH: 4160, TRAIN LOSS: 1962.266845703125, TEST LOSS 67.07939147949219\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 84%|████████▍ | 4200/4999 [00:25<00:04, 167.81it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 4170, TRAIN LOSS: 2026.234619140625, TEST LOSS 55.50738525390625\n",
      "EPOCH: 4180, TRAIN LOSS: 2011.510986328125, TEST LOSS 50.21141052246094\n",
      "EPOCH: 4190, TRAIN LOSS: 2029.5546875, TEST LOSS 48.798824310302734\n",
      "EPOCH: 4200, TRAIN LOSS: 1955.37255859375, TEST LOSS 57.597251892089844\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 85%|████████▍ | 4237/4999 [00:25<00:04, 167.88it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 4210, TRAIN LOSS: 2005.5916748046875, TEST LOSS 65.23552703857422\n",
      "EPOCH: 4220, TRAIN LOSS: 2037.419189453125, TEST LOSS 61.96764373779297\n",
      "EPOCH: 4230, TRAIN LOSS: 2016.22412109375, TEST LOSS 50.647254943847656\n",
      "EPOCH: 4240, TRAIN LOSS: 2009.8077392578125, TEST LOSS 41.94554138183594\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 85%|████████▌ | 4273/4999 [00:25<00:04, 167.94it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 4250, TRAIN LOSS: 2087.611328125, TEST LOSS 90.00689697265625\n",
      "EPOCH: 4260, TRAIN LOSS: 2013.22265625, TEST LOSS 74.66850280761719\n",
      "EPOCH: 4270, TRAIN LOSS: 2028.1171875, TEST LOSS 67.60108947753906\n",
      "EPOCH: 4280, TRAIN LOSS: 1958.239501953125, TEST LOSS 60.627830505371094\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 86%|████████▌ | 4309/4999 [00:25<00:04, 167.99it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 4290, TRAIN LOSS: 2014.0751953125, TEST LOSS 51.522850036621094\n",
      "EPOCH: 4300, TRAIN LOSS: 2058.372314453125, TEST LOSS 64.41620635986328\n",
      "EPOCH: 4310, TRAIN LOSS: 2054.189453125, TEST LOSS 68.8840103149414\n",
      "EPOCH: 4320, TRAIN LOSS: 2074.41943359375, TEST LOSS 71.77559661865234\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 87%|████████▋ | 4345/4999 [00:25<00:03, 167.64it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 4330, TRAIN LOSS: 2051.1357421875, TEST LOSS 55.96656799316406\n",
      "EPOCH: 4340, TRAIN LOSS: 2027.197998046875, TEST LOSS 44.156654357910156\n",
      "EPOCH: 4350, TRAIN LOSS: 2060.15966796875, TEST LOSS 61.47119140625\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 88%|████████▊ | 4375/4999 [00:26<00:03, 166.52it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 4360, TRAIN LOSS: 2023.4556884765625, TEST LOSS 66.83734893798828\n",
      "EPOCH: 4370, TRAIN LOSS: 1980.7528076171875, TEST LOSS 73.89335632324219\n",
      "EPOCH: 4380, TRAIN LOSS: 1993.7125244140625, TEST LOSS 63.004859924316406\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 88%|████████▊ | 4416/4999 [00:26<00:03, 165.99it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 4390, TRAIN LOSS: 1967.4564208984375, TEST LOSS 58.805686950683594\n",
      "EPOCH: 4400, TRAIN LOSS: 2009.71484375, TEST LOSS 51.778804779052734\n",
      "EPOCH: 4410, TRAIN LOSS: 1972.0302734375, TEST LOSS 63.03126525878906\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 89%|████████▉ | 4443/4999 [00:26<00:03, 165.72it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 4420, TRAIN LOSS: 1949.0465087890625, TEST LOSS 68.14706420898438\n",
      "EPOCH: 4430, TRAIN LOSS: 1993.089111328125, TEST LOSS 51.05876922607422\n",
      "EPOCH: 4440, TRAIN LOSS: 1995.3466796875, TEST LOSS 74.32038116455078\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 89%|████████▉ | 4470/4999 [00:27<00:03, 165.28it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 4450, TRAIN LOSS: 1940.4207763671875, TEST LOSS 55.40403747558594\n",
      "EPOCH: 4460, TRAIN LOSS: 2006.31494140625, TEST LOSS 69.57514953613281\n",
      "EPOCH: 4470, TRAIN LOSS: 1982.18310546875, TEST LOSS 56.02783966064453\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 90%|████████▉ | 4499/4999 [00:27<00:03, 165.08it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 4480, TRAIN LOSS: 2019.35205078125, TEST LOSS 67.00525665283203\n",
      "EPOCH: 4490, TRAIN LOSS: 2023.64111328125, TEST LOSS 48.033653259277344\n",
      "EPOCH: 4500, TRAIN LOSS: 2014.5911865234375, TEST LOSS 85.41758728027344\n",
      "EPOCH: 4510, TRAIN LOSS: 1986.32763671875, TEST LOSS 67.39469909667969\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 91%|█████████ | 4551/4999 [00:27<00:02, 165.12it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 4520, TRAIN LOSS: 2004.1812744140625, TEST LOSS 63.279388427734375\n",
      "EPOCH: 4530, TRAIN LOSS: 1933.5887451171875, TEST LOSS 60.39237976074219\n",
      "EPOCH: 4540, TRAIN LOSS: 1940.748779296875, TEST LOSS 77.77096557617188\n",
      "EPOCH: 4550, TRAIN LOSS: 2113.788330078125, TEST LOSS 66.91285705566406\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 92%|█████████▏| 4586/4999 [00:27<00:02, 165.14it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 4560, TRAIN LOSS: 1938.536376953125, TEST LOSS 74.17314910888672\n",
      "EPOCH: 4570, TRAIN LOSS: 1925.666015625, TEST LOSS 60.204444885253906\n",
      "EPOCH: 4580, TRAIN LOSS: 2016.96923828125, TEST LOSS 81.44805145263672\n",
      "EPOCH: 4590, TRAIN LOSS: 1962.658935546875, TEST LOSS 75.7470474243164\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 92%|█████████▏| 4621/4999 [00:27<00:02, 165.18it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 4600, TRAIN LOSS: 1918.2611083984375, TEST LOSS 57.692962646484375\n",
      "EPOCH: 4610, TRAIN LOSS: 1921.6295166015625, TEST LOSS 60.271827697753906\n",
      "EPOCH: 4620, TRAIN LOSS: 2019.153564453125, TEST LOSS 56.796669006347656\n",
      "EPOCH: 4630, TRAIN LOSS: 1919.2506103515625, TEST LOSS 71.33833312988281\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 94%|█████████▎| 4675/4999 [00:28<00:01, 165.28it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 4640, TRAIN LOSS: 1969.38427734375, TEST LOSS 86.21183013916016\n",
      "EPOCH: 4650, TRAIN LOSS: 2026.1976318359375, TEST LOSS 62.20585632324219\n",
      "EPOCH: 4660, TRAIN LOSS: 1990.6668701171875, TEST LOSS 45.7756462097168\n",
      "EPOCH: 4670, TRAIN LOSS: 1994.09423828125, TEST LOSS 65.01725769042969\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 94%|█████████▍| 4711/4999 [00:28<00:01, 165.31it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 4680, TRAIN LOSS: 2028.240234375, TEST LOSS 48.52566146850586\n",
      "EPOCH: 4690, TRAIN LOSS: 1974.2017822265625, TEST LOSS 56.39065170288086\n",
      "EPOCH: 4700, TRAIN LOSS: 2020.93212890625, TEST LOSS 60.27377700805664\n",
      "EPOCH: 4710, TRAIN LOSS: 1936.83349609375, TEST LOSS 57.56073760986328\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 95%|█████████▍| 4748/4999 [00:28<00:01, 165.39it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 4720, TRAIN LOSS: 2059.619384765625, TEST LOSS 77.10453796386719\n",
      "EPOCH: 4730, TRAIN LOSS: 1960.9599609375, TEST LOSS 64.07734680175781\n",
      "EPOCH: 4740, TRAIN LOSS: 1988.709716796875, TEST LOSS 79.66734313964844\n",
      "EPOCH: 4750, TRAIN LOSS: 1937.730712890625, TEST LOSS 79.78736114501953\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 96%|█████████▌| 4784/4999 [00:28<00:01, 165.42it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 4760, TRAIN LOSS: 1986.7227783203125, TEST LOSS 47.66364288330078\n",
      "EPOCH: 4770, TRAIN LOSS: 1973.2042236328125, TEST LOSS 60.223716735839844\n",
      "EPOCH: 4780, TRAIN LOSS: 2001.4471435546875, TEST LOSS 73.53474426269531\n",
      "EPOCH: 4790, TRAIN LOSS: 2132.02099609375, TEST LOSS 65.30496215820312\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 96%|█████████▋| 4820/4999 [00:29<00:01, 165.48it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 4800, TRAIN LOSS: 1959.133544921875, TEST LOSS 45.72225570678711\n",
      "EPOCH: 4810, TRAIN LOSS: 1953.0880126953125, TEST LOSS 61.34743118286133\n",
      "EPOCH: 4820, TRAIN LOSS: 1954.361572265625, TEST LOSS 94.60649871826172\n",
      "EPOCH: 4830, TRAIN LOSS: 1974.38037109375, TEST LOSS 77.5087890625\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 98%|█████████▊| 4875/4999 [00:29<00:00, 165.59it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 4840, TRAIN LOSS: 1944.800537109375, TEST LOSS 71.80673217773438\n",
      "EPOCH: 4850, TRAIN LOSS: 1983.8226318359375, TEST LOSS 72.70068359375\n",
      "EPOCH: 4860, TRAIN LOSS: 1955.3212890625, TEST LOSS 75.5531005859375\n",
      "EPOCH: 4870, TRAIN LOSS: 1949.5206298828125, TEST LOSS 61.70474624633789\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 98%|█████████▊| 4911/4999 [00:29<00:00, 165.63it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 4880, TRAIN LOSS: 1983.311279296875, TEST LOSS 56.20811462402344\n",
      "EPOCH: 4890, TRAIN LOSS: 1952.534423828125, TEST LOSS 70.2987060546875\n",
      "EPOCH: 4900, TRAIN LOSS: 1924.532470703125, TEST LOSS 69.97403717041016\n",
      "EPOCH: 4910, TRAIN LOSS: 1921.8443603515625, TEST LOSS 80.82601165771484\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 99%|█████████▉| 4947/4999 [00:29<00:00, 165.62it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 4920, TRAIN LOSS: 2004.56787109375, TEST LOSS 59.87004089355469\n",
      "EPOCH: 4930, TRAIN LOSS: 2068.346923828125, TEST LOSS 64.72772979736328\n",
      "EPOCH: 4940, TRAIN LOSS: 1990.1883544921875, TEST LOSS 104.54320526123047\n",
      "EPOCH: 4950, TRAIN LOSS: 1893.645751953125, TEST LOSS 77.3018798828125\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|█████████▉| 4982/4999 [00:30<00:00, 165.67it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 4960, TRAIN LOSS: 1955.8157958984375, TEST LOSS 51.73723602294922\n",
      "EPOCH: 4970, TRAIN LOSS: 2003.1929931640625, TEST LOSS 70.36030578613281\n",
      "EPOCH: 4980, TRAIN LOSS: 2016.199462890625, TEST LOSS 63.51121520996094\n",
      "EPOCH: 4990, TRAIN LOSS: 2016.95947265625, TEST LOSS 53.41533660888672\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4999/4999 [00:30<00:00, 165.67it/s]\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZUAAAEWCAYAAACufwpNAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nFRERyZj/BxDarL+PxDCJAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1a23876d30>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "(VAE(\n",
       "   (fc1): Linear(in_features=100, out_features=66, bias=True)\n",
       "   (fc21): Linear(in_features=66, out_features=25, bias=True)\n",
       "   (fc22): Linear(in_features=66, out_features=25, bias=True)\n",
       "   (fc3): Linear(in_features=25, out_features=66, bias=True)\n",
       "   (fc4): Linear(in_features=66, out_features=100, bias=True)\n",
       " ), tensor([[[ 0.,  0.,  0.,  ...,  0.,  0.,  0.],\n",
       "          [ 0.,  0.,  0.,  ...,  0.,  0.,  0.],\n",
       "          [ 0.,  0.,  0.,  ...,  0.,  0.,  0.],\n",
       "          ...,\n",
       "          [ 0.,  0.,  0.,  ...,  0.,  0.,  1.],\n",
       "          [ 0.,  0.,  0.,  ...,  0.,  0.,  0.],\n",
       "          [ 0.,  0.,  0.,  ...,  1.,  0.,  0.]],\n",
       " \n",
       "         [[ 0.,  0.,  0.,  ...,  0.,  0.,  0.],\n",
       "          [ 0.,  0.,  0.,  ...,  1.,  0.,  0.],\n",
       "          [ 0.,  0.,  0.,  ...,  0.,  1.,  0.],\n",
       "          ...,\n",
       "          [ 0.,  1.,  0.,  ...,  0.,  0.,  0.],\n",
       "          [ 0.,  0.,  1.,  ...,  0.,  0.,  0.],\n",
       "          [ 0.,  0.,  0.,  ...,  0.,  0.,  0.]],\n",
       " \n",
       "         [[ 0.,  0.,  0.,  ...,  0.,  0.,  0.],\n",
       "          [ 0.,  0.,  0.,  ...,  0.,  1.,  0.],\n",
       "          [ 0.,  0.,  0.,  ...,  0.,  0.,  0.],\n",
       "          ...,\n",
       "          [ 0.,  0.,  0.,  ...,  0.,  0.,  0.],\n",
       "          [ 0.,  1.,  0.,  ...,  0.,  0.,  0.],\n",
       "          [ 0.,  0.,  0.,  ...,  0.,  0.,  0.]],\n",
       " \n",
       "         ...,\n",
       " \n",
       "         [[ 0.,  0.,  0.,  ...,  0.,  0.,  0.],\n",
       "          [ 0.,  0.,  0.,  ...,  0.,  0.,  1.],\n",
       "          [ 0.,  0.,  0.,  ...,  0.,  1.,  0.],\n",
       "          ...,\n",
       "          [ 0.,  0.,  0.,  ...,  0.,  0.,  0.],\n",
       "          [ 0.,  0.,  1.,  ...,  0.,  0.,  0.],\n",
       "          [ 0.,  1.,  0.,  ...,  0.,  0.,  0.]],\n",
       " \n",
       "         [[ 0.,  0.,  0.,  ...,  0.,  0.,  0.],\n",
       "          [ 0.,  0.,  0.,  ...,  0.,  0.,  0.],\n",
       "          [ 0.,  0.,  0.,  ...,  0.,  0.,  0.],\n",
       "          ...,\n",
       "          [ 0.,  0.,  0.,  ...,  0.,  0.,  0.],\n",
       "          [ 0.,  0.,  0.,  ...,  0.,  0.,  0.],\n",
       "          [ 0.,  0.,  0.,  ...,  0.,  0.,  0.]],\n",
       " \n",
       "         [[ 0.,  0.,  1.,  ...,  0.,  0.,  0.],\n",
       "          [ 0.,  0.,  0.,  ...,  0.,  0.,  1.],\n",
       "          [ 1.,  0.,  0.,  ...,  0.,  0.,  0.],\n",
       "          ...,\n",
       "          [ 0.,  0.,  0.,  ...,  0.,  1.,  0.],\n",
       "          [ 0.,  0.,  0.,  ...,  1.,  0.,  0.],\n",
       "          [ 0.,  1.,  0.,  ...,  0.,  0.,  0.]]]))"
      ]
     },
     "execution_count": 629,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# тест c VAE моделькой\n",
    "SPD.run_test_vae_mutag(10, n_iterations=5000)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 661,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Reading graphs ...\n",
      "---:  mutag_1.graphml\n",
      "---:  mutag_2.graphml\n",
      "---:  mutag_3.graphml\n",
      "---:  mutag_4.graphml\n",
      "---:  mutag_5.graphml\n",
      "---:  mutag_6.graphml\n",
      "---:  mutag_7.graphml\n",
      "---:  mutag_8.graphml\n",
      "---:  mutag_9.graphml\n",
      "---:  mutag_10.graphml\n",
      "---:  mutag_11.graphml\n",
      "---:  mutag_12.graphml\n",
      "---:  mutag_13.graphml\n",
      "---:  mutag_14.graphml\n",
      "---:  mutag_15.graphml\n",
      "---:  mutag_16.graphml\n",
      "---:  mutag_17.graphml\n",
      "---:  mutag_18.graphml\n",
      "---:  mutag_19.graphml\n",
      "---:  mutag_20.graphml\n",
      "---:  mutag_21.graphml\n",
      "---:  mutag_22.graphml\n",
      "---:  mutag_23.graphml\n",
      "---:  mutag_24.graphml\n",
      "---:  mutag_25.graphml\n",
      "---:  mutag_26.graphml\n",
      "---:  mutag_27.graphml\n",
      "---:  mutag_28.graphml\n",
      "---:  mutag_29.graphml\n",
      "---:  mutag_30.graphml\n",
      "---:  mutag_31.graphml\n",
      "---:  mutag_32.graphml\n",
      "---:  mutag_33.graphml\n",
      "---:  mutag_34.graphml\n",
      "---:  mutag_35.graphml\n",
      "---:  mutag_36.graphml\n",
      "---:  mutag_37.graphml\n",
      "---:  mutag_38.graphml\n",
      "---:  mutag_39.graphml\n",
      "---:  mutag_40.graphml\n",
      "---:  mutag_41.graphml\n",
      "---:  mutag_42.graphml\n",
      "---:  mutag_43.graphml\n",
      "---:  mutag_44.graphml\n",
      "---:  mutag_45.graphml\n",
      "---:  mutag_46.graphml\n",
      "---:  mutag_47.graphml\n",
      "---:  mutag_48.graphml\n",
      "---:  mutag_49.graphml\n",
      "---:  mutag_50.graphml\n",
      "---:  mutag_51.graphml\n",
      "---:  mutag_52.graphml\n",
      "---:  mutag_53.graphml\n",
      "---:  mutag_54.graphml\n",
      "---:  mutag_55.graphml\n",
      "---:  mutag_56.graphml\n",
      "---:  mutag_57.graphml\n",
      "---:  mutag_58.graphml\n",
      "---:  mutag_59.graphml\n",
      "---:  mutag_60.graphml\n",
      "---:  mutag_61.graphml\n",
      "---:  mutag_62.graphml\n",
      "---:  mutag_63.graphml\n",
      "---:  mutag_64.graphml\n",
      "---:  mutag_65.graphml\n",
      "---:  mutag_66.graphml\n",
      "---:  mutag_67.graphml\n",
      "---:  mutag_68.graphml\n",
      "---:  mutag_69.graphml\n",
      "---:  mutag_70.graphml\n",
      "---:  mutag_71.graphml\n",
      "---:  mutag_72.graphml\n",
      "---:  mutag_73.graphml\n",
      "---:  mutag_74.graphml\n",
      "---:  mutag_75.graphml\n",
      "---:  mutag_76.graphml\n",
      "---:  mutag_77.graphml\n",
      "---:  mutag_78.graphml\n",
      "---:  mutag_79.graphml\n",
      "---:  mutag_80.graphml\n",
      "---:  mutag_81.graphml\n",
      "---:  mutag_82.graphml\n",
      "---:  mutag_83.graphml\n",
      "---:  mutag_84.graphml\n",
      "---:  mutag_85.graphml\n",
      "---:  mutag_86.graphml\n",
      "---:  mutag_87.graphml\n",
      "---:  mutag_88.graphml\n",
      "---:  mutag_89.graphml\n",
      "---:  mutag_90.graphml\n",
      "---:  mutag_91.graphml\n",
      "---:  mutag_92.graphml\n",
      "---:  mutag_93.graphml\n",
      "---:  mutag_94.graphml\n",
      "---:  mutag_95.graphml\n",
      "---:  mutag_96.graphml\n",
      "---:  mutag_97.graphml\n",
      "---:  mutag_98.graphml\n",
      "---:  mutag_99.graphml\n",
      "---:  mutag_100.graphml\n",
      "---:  mutag_101.graphml\n",
      "---:  mutag_102.graphml\n",
      "---:  mutag_103.graphml\n",
      "---:  mutag_104.graphml\n",
      "---:  mutag_105.graphml\n",
      "---:  mutag_106.graphml\n",
      "---:  mutag_107.graphml\n",
      "---:  mutag_108.graphml\n",
      "---:  mutag_109.graphml\n",
      "---:  mutag_110.graphml\n",
      "---:  mutag_111.graphml\n",
      "---:  mutag_112.graphml\n",
      "---:  mutag_113.graphml\n",
      "---:  mutag_114.graphml\n",
      "---:  mutag_115.graphml\n",
      "---:  mutag_116.graphml\n",
      "---:  mutag_117.graphml\n",
      "---:  mutag_118.graphml\n",
      "---:  mutag_119.graphml\n",
      "---:  mutag_120.graphml\n",
      "---:  mutag_121.graphml\n",
      "---:  mutag_122.graphml\n",
      "---:  mutag_123.graphml\n",
      "---:  mutag_124.graphml\n",
      "---:  mutag_125.graphml\n",
      "---:  mutag_126.graphml\n",
      "---:  mutag_127.graphml\n",
      "---:  mutag_128.graphml\n",
      "---:  mutag_129.graphml\n",
      "---:  mutag_130.graphml\n",
      "---:  mutag_131.graphml\n",
      "---:  mutag_132.graphml\n",
      "---:  mutag_133.graphml\n",
      "---:  mutag_134.graphml\n",
      "---:  mutag_135.graphml\n",
      "---:  mutag_136.graphml\n",
      "---:  mutag_137.graphml\n",
      "---:  mutag_138.graphml\n",
      "---:  mutag_139.graphml\n",
      "---:  mutag_140.graphml\n",
      "---:  mutag_141.graphml\n",
      "---:  mutag_142.graphml\n",
      "---:  mutag_143.graphml\n",
      "---:  mutag_144.graphml\n",
      "---:  mutag_145.graphml\n",
      "---:  mutag_146.graphml\n",
      "---:  mutag_147.graphml\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  0%|          | 0/199 [00:00<?, ?it/s]/Users/Macbook/Desktop/MIPT/Optimization_course/SPD_reduce/SPD.py:317: UserWarning: invalid index of a 0-dim tensor. This will be an error in PyTorch 0.5. Use tensor.item() to convert a 0-dim tensor to a Python number\n",
      "  loss_data.append((loss_train.data[0], float(loss_test.data[0])))\n",
      "  5%|▍         | 9/199 [00:00<00:08, 22.51it/s]/Users/Macbook/Desktop/MIPT/Optimization_course/SPD_reduce/SPD.py:312: UserWarning: invalid index of a 0-dim tensor. This will be an error in PyTorch 0.5. Use tensor.item() to convert a 0-dim tensor to a Python number\n",
      "  print(\"EPOCH: {0}, TRAIN LOSS: {1}, TEST LOSS\".format(epoch, loss_train.data[0]), float(loss_test.data[0]))\n",
      "  6%|▌         | 12/199 [00:00<00:08, 22.28it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "---:  mutag_148.graphml\n",
      "---:  mutag_149.graphml\n",
      "---:  mutag_150.graphml\n",
      "---:  mutag_151.graphml\n",
      "---:  mutag_152.graphml\n",
      "---:  mutag_153.graphml\n",
      "---:  mutag_154.graphml\n",
      "---:  mutag_155.graphml\n",
      "---:  mutag_156.graphml\n",
      "---:  mutag_157.graphml\n",
      "---:  mutag_158.graphml\n",
      "---:  mutag_159.graphml\n",
      "---:  mutag_160.graphml\n",
      "---:  mutag_161.graphml\n",
      "---:  mutag_162.graphml\n",
      "---:  mutag_163.graphml\n",
      "---:  mutag_164.graphml\n",
      "---:  mutag_165.graphml\n",
      "---:  mutag_166.graphml\n",
      "---:  mutag_167.graphml\n",
      "---:  mutag_168.graphml\n",
      "---:  mutag_169.graphml\n",
      "---:  mutag_170.graphml\n",
      "---:  mutag_171.graphml\n",
      "---:  mutag_172.graphml\n",
      "---:  mutag_173.graphml\n",
      "---:  mutag_174.graphml\n",
      "---:  mutag_175.graphml\n",
      "---:  mutag_176.graphml\n",
      "---:  mutag_177.graphml\n",
      "---:  mutag_178.graphml\n",
      "---:  mutag_179.graphml\n",
      "---:  mutag_180.graphml\n",
      "---:  mutag_181.graphml\n",
      "---:  mutag_182.graphml\n",
      "---:  mutag_183.graphml\n",
      "---:  mutag_184.graphml\n",
      "---:  mutag_185.graphml\n",
      "---:  mutag_186.graphml\n",
      "---:  mutag_187.graphml\n",
      "---:  mutag_188.graphml\n",
      "EPOCH: 10, TRAIN LOSS: 0.3239057660102844, TEST LOSS 0.3026762902736664\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 12%|█▏        | 23/199 [00:01<00:08, 21.76it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 20, TRAIN LOSS: 0.14611342549324036, TEST LOSS 0.14572131633758545\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 16%|█▌        | 32/199 [00:01<00:07, 21.70it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 30, TRAIN LOSS: 0.13143573701381683, TEST LOSS 0.13874071836471558\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 22%|██▏       | 44/199 [00:02<00:07, 19.87it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 40, TRAIN LOSS: 0.12582707405090332, TEST LOSS 0.12487766891717911\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 26%|██▌       | 52/199 [00:02<00:07, 20.04it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 50, TRAIN LOSS: 0.12440881878137589, TEST LOSS 0.12449484318494797\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 31%|███       | 62/199 [00:03<00:07, 19.27it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 60, TRAIN LOSS: 0.12250521779060364, TEST LOSS 0.12260595709085464\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 37%|███▋      | 73/199 [00:03<00:06, 19.48it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 70, TRAIN LOSS: 0.12054916471242905, TEST LOSS 0.12064839899539948\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 42%|████▏     | 83/199 [00:04<00:05, 19.40it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 80, TRAIN LOSS: 0.12023970484733582, TEST LOSS 0.12040434777736664\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 47%|████▋     | 93/199 [00:04<00:05, 19.20it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 90, TRAIN LOSS: 0.11967432498931885, TEST LOSS 0.11981910467147827\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 51%|█████     | 101/199 [00:05<00:05, 19.05it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 100, TRAIN LOSS: 0.11915827542543411, TEST LOSS 0.11912261694669724\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 57%|█████▋    | 113/199 [00:05<00:04, 19.02it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 110, TRAIN LOSS: 0.11862742900848389, TEST LOSS 0.11916164308786392\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 62%|██████▏   | 123/199 [00:06<00:03, 19.12it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 120, TRAIN LOSS: 0.11836675554513931, TEST LOSS 0.1192394271492958\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 66%|██████▋   | 132/199 [00:06<00:03, 19.24it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 130, TRAIN LOSS: 0.11813563108444214, TEST LOSS 0.11918319016695023\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 72%|███████▏  | 144/199 [00:07<00:02, 19.42it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 140, TRAIN LOSS: 0.11795154213905334, TEST LOSS 0.11920249462127686\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 77%|███████▋  | 153/199 [00:07<00:02, 19.38it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 150, TRAIN LOSS: 0.11750530451536179, TEST LOSS 0.11928265541791916\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 81%|████████  | 161/199 [00:08<00:01, 19.42it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 160, TRAIN LOSS: 0.11699243634939194, TEST LOSS 0.1189749464392662\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 86%|████████▋ | 172/199 [00:08<00:01, 19.35it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 170, TRAIN LOSS: 0.11672215908765793, TEST LOSS 0.11859949678182602\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 92%|█████████▏| 184/199 [00:09<00:00, 19.45it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 180, TRAIN LOSS: 0.1164478287100792, TEST LOSS 0.11834581196308136\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 97%|█████████▋| 193/199 [00:09<00:00, 19.51it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "EPOCH: 190, TRAIN LOSS: 0.11623617261648178, TEST LOSS 0.1183684766292572\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 199/199 [00:10<00:00, 19.46it/s]\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZIAAAEWCAYAAABMoxE0AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\njDHGVIslEmOMMdViicQYY0y1/H/80lWzSXtRGgAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1a23639240>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "(MatrixEncoder(\n",
       "   (encoder): Sequential(\n",
       "     (0): Linear2D()\n",
       "     (1): Tanh()\n",
       "   )\n",
       "   (decoder): Sequential(\n",
       "     (0): Linear2D()\n",
       "     (1): ReLU()\n",
       "   )\n",
       " ), tensor([[[ 0.,  0.,  0.,  ...,  0.,  0.,  0.],\n",
       "          [ 0.,  0.,  0.,  ...,  0.,  0.,  1.],\n",
       "          [ 0.,  0.,  0.,  ...,  0.,  0.,  0.],\n",
       "          ...,\n",
       "          [ 0.,  0.,  0.,  ...,  0.,  0.,  0.],\n",
       "          [ 0.,  0.,  0.,  ...,  0.,  0.,  0.],\n",
       "          [ 0.,  1.,  0.,  ...,  0.,  0.,  0.]],\n",
       " \n",
       "         [[ 0.,  0.,  1.,  ...,  0.,  1.,  0.],\n",
       "          [ 0.,  0.,  0.,  ...,  0.,  0.,  0.],\n",
       "          [ 1.,  0.,  0.,  ...,  0.,  0.,  0.],\n",
       "          ...,\n",
       "          [ 0.,  0.,  0.,  ...,  0.,  0.,  1.],\n",
       "          [ 1.,  0.,  0.,  ...,  0.,  0.,  0.],\n",
       "          [ 0.,  0.,  0.,  ...,  1.,  0.,  0.]],\n",
       " \n",
       "         [[ 0.,  0.,  1.,  ...,  0.,  0.,  0.],\n",
       "          [ 0.,  0.,  0.,  ...,  0.,  0.,  0.],\n",
       "          [ 1.,  0.,  0.,  ...,  0.,  0.,  0.],\n",
       "          ...,\n",
       "          [ 0.,  0.,  0.,  ...,  0.,  0.,  0.],\n",
       "          [ 0.,  0.,  0.,  ...,  0.,  0.,  1.],\n",
       "          [ 0.,  0.,  0.,  ...,  0.,  1.,  0.]],\n",
       " \n",
       "         ...,\n",
       " \n",
       "         [[ 0.,  0.,  1.,  ...,  0.,  0.,  0.],\n",
       "          [ 0.,  0.,  0.,  ...,  0.,  1.,  0.],\n",
       "          [ 1.,  0.,  0.,  ...,  0.,  0.,  0.],\n",
       "          ...,\n",
       "          [ 0.,  0.,  0.,  ...,  0.,  0.,  1.],\n",
       "          [ 0.,  1.,  0.,  ...,  0.,  0.,  0.],\n",
       "          [ 0.,  0.,  0.,  ...,  1.,  0.,  0.]],\n",
       " \n",
       "         [[ 0.,  0.,  0.,  ...,  1.,  0.,  0.],\n",
       "          [ 0.,  0.,  0.,  ...,  0.,  0.,  1.],\n",
       "          [ 0.,  0.,  0.,  ...,  0.,  0.,  0.],\n",
       "          ...,\n",
       "          [ 1.,  0.,  0.,  ...,  0.,  0.,  0.],\n",
       "          [ 0.,  0.,  0.,  ...,  0.,  0.,  0.],\n",
       "          [ 0.,  1.,  0.,  ...,  0.,  0.,  0.]],\n",
       " \n",
       "         [[ 0.,  0.,  1.,  ...,  0.,  0.,  0.],\n",
       "          [ 0.,  0.,  0.,  ...,  0.,  1.,  0.],\n",
       "          [ 1.,  0.,  0.,  ...,  0.,  1.,  0.],\n",
       "          ...,\n",
       "          [ 0.,  0.,  0.,  ...,  0.,  0.,  0.],\n",
       "          [ 0.,  1.,  1.,  ...,  0.,  0.,  0.],\n",
       "          [ 0.,  0.,  0.,  ...,  0.,  0.,  0.]]]), array([[ 6.13026285,  6.39704275],\n",
       "        [ 1.18494475,  1.14114141],\n",
       "        [ 0.52672094,  0.56396168],\n",
       "        [ 0.42545369,  0.51564878],\n",
       "        [ 0.51299453,  0.57797444],\n",
       "        [ 0.53975666,  0.60462165],\n",
       "        [ 0.45994422,  0.53902322],\n",
       "        [ 0.39491993,  0.41928688],\n",
       "        [ 0.35809532,  0.34611624],\n",
       "        [ 0.32390577,  0.30267629],\n",
       "        [ 0.29161528,  0.27037421],\n",
       "        [ 0.25802118,  0.24827296],\n",
       "        [ 0.22928415,  0.23537602],\n",
       "        [ 0.20603406,  0.21261111],\n",
       "        [ 0.18677348,  0.18593352],\n",
       "        [ 0.16945612,  0.16674004],\n",
       "        [ 0.15709038,  0.1541324 ],\n",
       "        [ 0.15081377,  0.14996712],\n",
       "        [ 0.1474499 ,  0.14668356],\n",
       "        [ 0.14611343,  0.14572132],\n",
       "        [ 0.14423111,  0.14487815],\n",
       "        [ 0.14231898,  0.14452007],\n",
       "        [ 0.1407828 ,  0.14377841],\n",
       "        [ 0.13950352,  0.1426696 ],\n",
       "        [ 0.13879241,  0.14220619],\n",
       "        [ 0.13824768,  0.14255962],\n",
       "        [ 0.13686679,  0.14278156],\n",
       "        [ 0.13451532,  0.14236781],\n",
       "        [ 0.13288577,  0.14119385],\n",
       "        [ 0.13143574,  0.13874072],\n",
       "        [ 0.13021578,  0.13586359],\n",
       "        [ 0.12975368,  0.13374215],\n",
       "        [ 0.12990965,  0.13248417],\n",
       "        [ 0.12986441,  0.13131873],\n",
       "        [ 0.12905243,  0.12962936],\n",
       "        [ 0.12778531,  0.12773839],\n",
       "        [ 0.12662108,  0.1261715 ],\n",
       "        [ 0.12596186,  0.12533522],\n",
       "        [ 0.1257699 ,  0.12493421],\n",
       "        [ 0.12582707,  0.12487767],\n",
       "        [ 0.12588646,  0.12488746],\n",
       "        [ 0.12584642,  0.12493401],\n",
       "        [ 0.12567988,  0.12496018],\n",
       "        [ 0.12546067,  0.12499721],\n",
       "        [ 0.12525927,  0.12505148],\n",
       "        [ 0.12516077,  0.1251844 ],\n",
       "        [ 0.1251151 ,  0.12532574],\n",
       "        [ 0.12501344,  0.12529349],\n",
       "        [ 0.12476985,  0.12500679],\n",
       "        [ 0.12440882,  0.12449484],\n",
       "        [ 0.1240607 ,  0.12395877],\n",
       "        [ 0.12382657,  0.12355525],\n",
       "        [ 0.12372829,  0.12334643],\n",
       "        [ 0.12369171,  0.12328021],\n",
       "        [ 0.12364873,  0.12327645],\n",
       "        [ 0.12354356,  0.12327498],\n",
       "        [ 0.12333354,  0.12316426],\n",
       "        [ 0.12306325,  0.12298045],\n",
       "        [ 0.12278291,  0.12279073],\n",
       "        [ 0.12250522,  0.12260596],\n",
       "        [ 0.12223644,  0.12242166],\n",
       "        [ 0.12198205,  0.12222959],\n",
       "        [ 0.12174879,  0.12203069],\n",
       "        [ 0.12154   ,  0.12183072],\n",
       "        [ 0.12135644,  0.12162368],\n",
       "        [ 0.12118591,  0.12140898],\n",
       "        [ 0.12100998,  0.12119447],\n",
       "        [ 0.12083936,  0.12098643],\n",
       "        [ 0.12068244,  0.12079929],\n",
       "        [ 0.12054916,  0.1206484 ],\n",
       "        [ 0.12045639,  0.12055928],\n",
       "        [ 0.12040459,  0.1205488 ],\n",
       "        [ 0.12038296,  0.12058695],\n",
       "        [ 0.12036622,  0.12062582],\n",
       "        [ 0.1203445 ,  0.12063607],\n",
       "        [ 0.12032192,  0.12061668],\n",
       "        [ 0.12030614,  0.12058252],\n",
       "        [ 0.12029774,  0.12053839],\n",
       "        [ 0.12028069,  0.12047863],\n",
       "        [ 0.1202397 ,  0.12040435],\n",
       "        [ 0.12017706,  0.12031959],\n",
       "        [ 0.12010679,  0.12023836],\n",
       "        [ 0.12004248,  0.12017576],\n",
       "        [ 0.1199903 ,  0.1201325 ],\n",
       "        [ 0.11994356,  0.12009903],\n",
       "        [ 0.1198944 ,  0.12006027],\n",
       "        [ 0.11983957,  0.12000779],\n",
       "        [ 0.11977621,  0.11994481],\n",
       "        [ 0.11972052,  0.11988001],\n",
       "        [ 0.11967432,  0.1198191 ],\n",
       "        [ 0.11963515,  0.11975916],\n",
       "        [ 0.1195943 ,  0.1196917 ],\n",
       "        [ 0.11955088,  0.1196155 ],\n",
       "        [ 0.11950316,  0.11953387],\n",
       "        [ 0.11945585,  0.11945218],\n",
       "        [ 0.11940067,  0.11937653],\n",
       "        [ 0.11934179,  0.11930633],\n",
       "        [ 0.1192742 ,  0.11924078],\n",
       "        [ 0.11921196,  0.11917911],\n",
       "        [ 0.11915828,  0.11912262],\n",
       "        [ 0.11910931,  0.1190707 ],\n",
       "        [ 0.11906254,  0.11903576],\n",
       "        [ 0.11901788,  0.11902011],\n",
       "        [ 0.11897095,  0.11902294],\n",
       "        [ 0.11892273,  0.11904083],\n",
       "        [ 0.11887215,  0.11906849],\n",
       "        [ 0.11881669,  0.11909989],\n",
       "        [ 0.11874383,  0.11912891],\n",
       "        [ 0.11867935,  0.11915089],\n",
       "        [ 0.11862743,  0.11916164],\n",
       "        [ 0.11858418,  0.11915939],\n",
       "        [ 0.11855154,  0.11914563],\n",
       "        [ 0.11852325,  0.11912851],\n",
       "        [ 0.11850133,  0.11911582],\n",
       "        [ 0.11848186,  0.1191139 ],\n",
       "        [ 0.11846235,  0.119124  ],\n",
       "        [ 0.11844053,  0.119146  ],\n",
       "        [ 0.11841615,  0.11917602],\n",
       "        [ 0.11839258,  0.11920921],\n",
       "        [ 0.11836676,  0.11923943],\n",
       "        [ 0.1183423 ,  0.11926114],\n",
       "        [ 0.11831742,  0.11927161],\n",
       "        [ 0.11829181,  0.11926968],\n",
       "        [ 0.11826946,  0.11925704],\n",
       "        [ 0.11824702,  0.11923819],\n",
       "        [ 0.11822753,  0.11921737],\n",
       "        [ 0.11820272,  0.11919896],\n",
       "        [ 0.11817728,  0.11918678],\n",
       "        [ 0.11815537,  0.11918171],\n",
       "        [ 0.11813563,  0.11918319],\n",
       "        [ 0.11811731,  0.11918908],\n",
       "        [ 0.11810143,  0.11919723],\n",
       "        [ 0.1180849 ,  0.11920522],\n",
       "        [ 0.11806887,  0.11921219],\n",
       "        [ 0.11805379,  0.11921681],\n",
       "        [ 0.11803769,  0.11922003],\n",
       "        [ 0.11801974,  0.1192159 ],\n",
       "        [ 0.11800082,  0.11921004],\n",
       "        [ 0.11797744,  0.11920486],\n",
       "        [ 0.11795154,  0.11920249],\n",
       "        [ 0.1179226 ,  0.1191991 ],\n",
       "        [ 0.11788926,  0.11920272],\n",
       "        [ 0.11784489,  0.11921333],\n",
       "        [ 0.11779418,  0.11923032],\n",
       "        [ 0.11773947,  0.11925226],\n",
       "        [ 0.11768799,  0.1192759 ],\n",
       "        [ 0.11764372,  0.11928963],\n",
       "        [ 0.11760077,  0.1192926 ],\n",
       "        [ 0.11755673,  0.11929013],\n",
       "        [ 0.1175053 ,  0.11928266],\n",
       "        [ 0.11744665,  0.11926972],\n",
       "        [ 0.11737897,  0.11925304],\n",
       "        [ 0.11730596,  0.11923174],\n",
       "        [ 0.11723716,  0.11920654],\n",
       "        [ 0.11717968,  0.11917792],\n",
       "        [ 0.1171324 ,  0.11914376],\n",
       "        [ 0.11709379,  0.11910427],\n",
       "        [ 0.11705861,  0.11906146],\n",
       "        [ 0.11702517,  0.11901765],\n",
       "        [ 0.11699244,  0.11897495],\n",
       "        [ 0.11696077,  0.11893452],\n",
       "        [ 0.11693066,  0.11889718],\n",
       "        [ 0.11690198,  0.11886159],\n",
       "        [ 0.11687682,  0.11882682],\n",
       "        [ 0.1168502 ,  0.11879247],\n",
       "        [ 0.11682488,  0.11875612],\n",
       "        [ 0.11679932,  0.11871827],\n",
       "        [ 0.11677403,  0.11867905],\n",
       "        [ 0.11674879,  0.11863939],\n",
       "        [ 0.11672216,  0.1185995 ],\n",
       "        [ 0.11669564,  0.11856101],\n",
       "        [ 0.11666846,  0.11852367],\n",
       "        [ 0.11664013,  0.11848832],\n",
       "        [ 0.11661182,  0.11845598],\n",
       "        [ 0.1165844 ,  0.11842674],\n",
       "        [ 0.11655511,  0.1184022 ],\n",
       "        [ 0.11652759,  0.11838137],\n",
       "        [ 0.11649906,  0.11836493],\n",
       "        [ 0.11647256,  0.11835299],\n",
       "        [ 0.11644783,  0.11834581],\n",
       "        [ 0.11642291,  0.11834262],\n",
       "        [ 0.11640033,  0.1183425 ],\n",
       "        [ 0.11637796,  0.11834545],\n",
       "        [ 0.11635533,  0.11834945],\n",
       "        [ 0.11633623,  0.11835431],\n",
       "        [ 0.11631574,  0.11835869],\n",
       "        [ 0.11629542,  0.11836237],\n",
       "        [ 0.11627671,  0.11836515],\n",
       "        [ 0.11625641,  0.1183673 ],\n",
       "        [ 0.11623617,  0.11836848],\n",
       "        [ 0.11621457,  0.11836991],\n",
       "        [ 0.1161933 ,  0.11837099],\n",
       "        [ 0.11617097,  0.11837181],\n",
       "        [ 0.11614837,  0.1183719 ],\n",
       "        [ 0.11612633,  0.11837133],\n",
       "        [ 0.11610442,  0.11837053],\n",
       "        [ 0.11608425,  0.11836813],\n",
       "        [ 0.11606525,  0.11836586],\n",
       "        [ 0.11604004,  0.11836277]]))"
      ]
     },
     "execution_count": 661,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# еще один тест с блоками из статьи, но с другими параметрами\n",
    "SPD.run_test_mutag(10, 200)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 684,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAa4AAAEICAYAAAAHsBBpAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1a271c6c18>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "n_points = 6\n",
    "fig, axes = plt.subplots(nrows=2, ncols=3, figsize=(7, 4))\n",
    "cmap = sns.cubehelix_palette(as_cmap=True)\n",
    "for i, t in enumerate(np.linspace(0, 1, n_points)):\n",
    "    C_t = A - (A-B)*t\n",
    "    GC_t = coder.decoder(C_t).round()\n",
    "    dt.draw_graph(GC_t, ax=axes[i//3, i%3], node_color='r')\n",
    "    axes[i//3, i%3].set_title(round(t, 2))\n",
    "    axes[i//3, i%3].plot()\n",
    "fig.savefig('homotopy.pdf', format='pdf')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Здесь попробуем порешать задачу классификации на датасете mutag"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 534,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true,
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Reading graphs ...\n",
      "---:  mutag_1.graphml\n",
      "---:  mutag_2.graphml\n",
      "---:  mutag_3.graphml\n",
      "---:  mutag_4.graphml\n",
      "---:  mutag_5.graphml\n",
      "---:  mutag_6.graphml\n",
      "---:  mutag_7.graphml\n",
      "---:  mutag_8.graphml\n",
      "---:  mutag_9.graphml\n",
      "---:  mutag_10.graphml\n",
      "---:  mutag_11.graphml\n",
      "---:  mutag_12.graphml\n",
      "---:  mutag_13.graphml\n",
      "---:  mutag_14.graphml\n",
      "---:  mutag_15.graphml\n",
      "---:  mutag_16.graphml\n",
      "---:  mutag_17.graphml\n",
      "---:  mutag_18.graphml\n",
      "---:  mutag_19.graphml\n",
      "---:  mutag_20.graphml\n",
      "---:  mutag_21.graphml\n",
      "---:  mutag_22.graphml\n",
      "---:  mutag_23.graphml\n",
      "---:  mutag_24.graphml\n",
      "---:  mutag_25.graphml\n",
      "---:  mutag_26.graphml\n",
      "---:  mutag_27.graphml\n",
      "---:  mutag_28.graphml\n",
      "---:  mutag_29.graphml\n",
      "---:  mutag_30.graphml\n",
      "---:  mutag_31.graphml\n",
      "---:  mutag_32.graphml\n",
      "---:  mutag_33.graphml\n",
      "---:  mutag_34.graphml\n",
      "---:  mutag_35.graphml\n",
      "---:  mutag_36.graphml\n",
      "---:  mutag_37.graphml\n",
      "---:  mutag_38.graphml\n",
      "---:  mutag_39.graphml\n",
      "---:  mutag_40.graphml\n",
      "---:  mutag_41.graphml\n",
      "---:  mutag_42.graphml\n",
      "---:  mutag_43.graphml\n",
      "---:  mutag_44.graphml\n",
      "---:  mutag_45.graphml\n",
      "---:  mutag_46.graphml\n",
      "---:  mutag_47.graphml\n",
      "---:  mutag_48.graphml\n",
      "---:  mutag_49.graphml\n",
      "---:  mutag_50.graphml\n",
      "---:  mutag_51.graphml\n",
      "---:  mutag_52.graphml\n",
      "---:  mutag_53.graphml\n",
      "---:  mutag_54.graphml\n",
      "---:  mutag_55.graphml\n",
      "---:  mutag_56.graphml\n",
      "---:  mutag_57.graphml\n",
      "---:  mutag_58.graphml\n",
      "---:  mutag_59.graphml\n",
      "---:  mutag_60.graphml\n",
      "---:  mutag_61.graphml\n",
      "---:  mutag_62.graphml\n",
      "---:  mutag_63.graphml\n",
      "---:  mutag_64.graphml\n",
      "---:  mutag_65.graphml\n",
      "---:  mutag_66.graphml\n",
      "---:  mutag_67.graphml\n",
      "---:  mutag_68.graphml\n",
      "---:  mutag_69.graphml\n",
      "---:  mutag_70.graphml\n",
      "---:  mutag_71.graphml\n",
      "---:  mutag_72.graphml\n",
      "---:  mutag_73.graphml\n",
      "---:  mutag_74.graphml\n",
      "---:  mutag_75.graphml\n",
      "---:  mutag_76.graphml\n",
      "---:  mutag_77.graphml\n",
      "---:  mutag_78.graphml\n",
      "---:  mutag_79.graphml\n",
      "---:  mutag_80.graphml\n",
      "---:  mutag_81.graphml\n",
      "---:  mutag_82.graphml\n",
      "---:  mutag_83.graphml\n",
      "---:  mutag_84.graphml\n",
      "---:  mutag_85.graphml\n",
      "---:  mutag_86.graphml\n",
      "---:  mutag_87.graphml\n",
      "---:  mutag_88.graphml\n",
      "---:  mutag_89.graphml\n",
      "---:  mutag_90.graphml\n",
      "---:  mutag_91.graphml\n",
      "---:  mutag_92.graphml\n",
      "---:  mutag_93.graphml\n",
      "---:  mutag_94.graphml\n",
      "---:  mutag_95.graphml\n",
      "---:  mutag_96.graphml\n",
      "---:  mutag_97.graphml\n",
      "---:  mutag_98.graphml\n",
      "---:  mutag_99.graphml\n",
      "---:  mutag_100.graphml\n",
      "---:  mutag_101.graphml\n",
      "---:  mutag_102.graphml\n",
      "---:  mutag_103.graphml\n",
      "---:  mutag_104.graphml\n",
      "---:  mutag_105.graphml\n",
      "---:  mutag_106.graphml\n",
      "---:  mutag_107.graphml\n",
      "---:  mutag_108.graphml\n",
      "---:  mutag_109.graphml\n",
      "---:  mutag_110.graphml\n",
      "---:  mutag_111.graphml\n",
      "---:  mutag_112.graphml\n",
      "---:  mutag_113.graphml\n",
      "---:  mutag_114.graphml\n",
      "---:  mutag_115.graphml\n",
      "---:  mutag_116.graphml\n",
      "---:  mutag_117.graphml\n",
      "---:  mutag_118.graphml\n",
      "---:  mutag_119.graphml\n",
      "---:  mutag_120.graphml\n",
      "---:  mutag_121.graphml\n",
      "---:  mutag_122.graphml\n",
      "---:  mutag_123.graphml\n",
      "---:  mutag_124.graphml\n",
      "---:  mutag_125.graphml\n",
      "---:  mutag_126.graphml\n",
      "---:  mutag_127.graphml\n",
      "---:  mutag_128.graphml\n",
      "---:  mutag_129.graphml\n",
      "---:  mutag_130.graphml\n",
      "---:  mutag_131.graphml\n",
      "---:  mutag_132.graphml\n",
      "---:  mutag_133.graphml\n",
      "---:  mutag_134.graphml\n",
      "---:  mutag_135.graphml\n",
      "---:  mutag_136.graphml\n",
      "---:  mutag_137.graphml\n",
      "---:  mutag_138.graphml\n",
      "---:  mutag_139.graphml\n",
      "---:  mutag_140.graphml\n",
      "---:  mutag_141.graphml\n",
      "---:  mutag_142.graphml\n",
      "---:  mutag_143.graphml\n",
      "---:  mutag_144.graphml\n",
      "---:  mutag_145.graphml\n",
      "---:  mutag_146.graphml\n",
      "---:  mutag_147.graphml\n",
      "---:  mutag_148.graphml\n",
      "---:  mutag_149.graphml\n",
      "---:  mutag_150.graphml\n",
      "---:  mutag_151.graphml\n",
      "---:  mutag_152.graphml\n",
      "---:  mutag_153.graphml\n",
      "---:  mutag_154.graphml\n",
      "---:  mutag_155.graphml\n",
      "---:  mutag_156.graphml\n",
      "---:  mutag_157.graphml\n",
      "---:  mutag_158.graphml\n",
      "---:  mutag_159.graphml\n",
      "---:  mutag_160.graphml\n",
      "---:  mutag_161.graphml\n",
      "---:  mutag_162.graphml\n",
      "---:  mutag_163.graphml\n",
      "---:  mutag_164.graphml\n",
      "---:  mutag_165.graphml\n",
      "---:  mutag_166.graphml\n",
      "---:  mutag_167.graphml\n",
      "---:  mutag_168.graphml\n",
      "---:  mutag_169.graphml\n",
      "---:  mutag_170.graphml\n",
      "---:  mutag_171.graphml\n",
      "---:  mutag_172.graphml\n",
      "---:  mutag_173.graphml\n",
      "---:  mutag_174.graphml\n",
      "---:  mutag_175.graphml\n",
      "---:  mutag_176.graphml\n",
      "---:  mutag_177.graphml\n",
      "---:  mutag_178.graphml\n",
      "---:  mutag_179.graphml\n",
      "---:  mutag_180.graphml\n",
      "---:  mutag_181.graphml\n",
      "---:  mutag_182.graphml\n",
      "---:  mutag_183.graphml\n",
      "---:  mutag_184.graphml\n",
      "---:  mutag_185.graphml\n",
      "---:  mutag_186.graphml\n",
      "---:  mutag_187.graphml\n",
      "---:  mutag_188.graphml\n"
     ]
    }
   ],
   "source": [
    "# создаем сам датасет\n",
    "X, y = dt.build_dataset(dt.read_test_graphs(), 10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 549,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [],
   "source": [
    "# строим эмбеддинги для графов \n",
    "temp = np.array([coder.encoder(x.unsqueeze(0)).view(-1,).data.numpy() for x in X])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 577,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [],
   "source": [
    "# пробуем визуализировать датасет\n",
    "tsne = TSNE(n_components=2)\n",
    "q = tsne.fit_transform(temp)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 562,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXwAAAEICAYAAABcVE8dAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nkt9GAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1a23858470>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# отрисовываем все это дело\n",
    "plt.scatter(q[np.where(y == 0)][:, 0], q[np.where(y == 0)][:, 1], c='r', label='mutag')\n",
    "plt.scatter(q[np.where(y == 1)][:, 0], q[np.where(y == 1)][:, 1], c='b', label='none')\n",
    "plt.title(\"t-SNE for Graph embeddings\")\n",
    "plt.grid()\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 659,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXwAAAEICAYAAABcVE8dAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nQjDBNwzDSAgm+IZhGAnh/wMCxh438aXy+QAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1a265fcb70>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# и еще раз\n",
    "fig = plt.figure()\n",
    "plt.scatter(q[np.where(y == 0)][:, 0], q[np.where(y == 0)][:, 1], c='r', label='mutag')\n",
    "plt.scatter(q[np.where(y == 1)][:, 0], q[np.where(y == 1)][:, 1], c='b', label='none')\n",
    "plt.title(\"t-SNE for Graph embeddings\")\n",
    "plt.grid()\n",
    "plt.legend()\n",
    "plt.show()\n",
    "fig.savefig('t-sne.pdf', format='pdf')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 639,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 3 folds for each of 30 candidates, totalling 90 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/Macbook/anaconda/envs/py35/lib/python3.5/site-packages/sklearn/model_selection/_split.py:2026: FutureWarning: From version 0.21, test_size will always complement train_size unless both are specified.\n",
      "  FutureWarning)\n",
      "[Parallel(n_jobs=1)]: Done  90 out of  90 | elapsed:    0.3s finished\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "RandomizedSearchCV(cv=None, error_score='raise',\n",
       "          estimator=SVC(C=1.0, cache_size=200, class_weight=None, coef0=0.0,\n",
       "  decision_function_shape='ovr', degree=3, gamma='auto', kernel='rbf',\n",
       "  max_iter=-1, probability=False, random_state=None, shrinking=True,\n",
       "  tol=0.001, verbose=False),\n",
       "          fit_params=None, iid=True, n_iter=30, n_jobs=1,\n",
       "          param_distributions={'kernel': ('linear', 'rbf'), 'C': array([  0.01   ,   0.26615,   0.52231,   0.77846,   1.03462,   1.29077,\n",
       "         1.54692,   1.80308,   2.05923,   2.31538,   2.57154,   2.82769,\n",
       "         3.08385,   3.34   ,   3.59615,   3.85231,   4.10846,   4.36462,\n",
       "         4.62077,   4.8769...      8.57286,   8.77673,   8.98061,   9.18449,   9.38837,   9.59224,\n",
       "         9.79612,  10.     ])},\n",
       "          pre_dispatch='2*n_jobs', random_state=None, refit=True,\n",
       "          return_train_score='warn', scoring='accuracy', verbose=True)"
      ]
     },
     "execution_count": 639,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# ищем какие-нибудь параметры, которые могут подойти для модели\n",
    "\n",
    "from sklearn.svm import SVC\n",
    "from sklearn.model_selection import train_test_split, RandomizedSearchCV\n",
    "\n",
    "cls = SVC()\n",
    "X_train, X_test, y_train, y_test = train_test_split(temp, \n",
    "                                                    y, stratify=y, train_size=0.7)\n",
    "\n",
    "srch = RandomizedSearchCV(cls, {'C': np.linspace(0.01, 10, 40), \n",
    "                                'kernel': ('linear', 'rbf'),\n",
    "                                'gamma': np.linspace(0.01, 10, 50)}, \n",
    "                          verbose=True, \n",
    "                          scoring='accuracy', \n",
    "                          n_iter=30)\n",
    "\n",
    "srch.fit(X_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 640,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "mean:  0.797554347826\n",
      "std:  0.0849749208557\n"
     ]
    }
   ],
   "source": [
    "# смотрим на качество по валидации\n",
    "cls = SVC(**srch.best_params_)\n",
    "cv=KFold(n_splits=8, shuffle=True)\n",
    "res = cross_val_score(cls, temp, y, scoring='accuracy', cv=cv)\n",
    "print('mean: ', res.mean())\n",
    "print('std: ', res.std())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 657,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "             precision    recall  f1-score   support\n",
      "\n",
      "          0       0.82      0.91      0.86        34\n",
      "          1       0.75      0.56      0.64        16\n",
      "\n",
      "avg / total       0.79      0.80      0.79        50\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# смотрим на качество по отложенной\n",
    "indexes =list(range(len(y)))\n",
    "shuffle(indexes)\n",
    "size = 50\n",
    "cls.fit(temp[indexes][:-size], y[indexes][:-size])\n",
    "print(classification_report(y[indexes][-size:], cls.predict(temp[indexes][-size:])))"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "kernel35",
   "language": "python",
   "name": "kernel35"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.5.4"
  },
  "toc": {
   "nav_menu": {},
   "number_sections": true,
   "sideBar": true,
   "skip_h1_title": false,
   "title_cell": "Table of Contents",
   "title_sidebar": "Contents",
   "toc_cell": false,
   "toc_position": {},
   "toc_section_display": true,
   "toc_window_display": false
  },
  "varInspector": {
   "cols": {
    "lenName": 16,
    "lenType": 16,
    "lenVar": 40
   },
   "kernels_config": {
    "python": {
     "delete_cmd_postfix": "",
     "delete_cmd_prefix": "del ",
     "library": "var_list.py",
     "varRefreshCmd": "print(var_dic_list())"
    },
    "r": {
     "delete_cmd_postfix": ") ",
     "delete_cmd_prefix": "rm(",
     "library": "var_list.r",
     "varRefreshCmd": "cat(var_dic_list()) "
    }
   },
   "types_to_exclude": [
    "module",
    "function",
    "builtin_function_or_method",
    "instance",
    "_Feature"
   ],
   "window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}