WeightWatcher-SVDSmoothing-VGG16.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "\n",
    "# WeightWatcher - SVDSmoothing\n",
    "\n",
    "### Example of SVDSmoothing using the VGG16 model \n",
    "\n",
    "Compare the original (green) vs. smoothed (red) ESDs\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-03-05T23:51:16.544327Z",
     "start_time": "2019-03-05T23:51:16.531457Z"
    },
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# Suppress the powerlaw package warnings\n",
    "# \"powerlaw.py:700: RuntimeWarning: divide by zero encountered in true_divide\"\n",
    "# \"powerlaw.py:700: RuntimeWarning: invalid value encountered in true_divide\"\n",
    "import warnings\n",
    "warnings.simplefilter(action='ignore', category=RuntimeWarning)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "\n",
    "import tensorflow\n",
    "import torchvision.models as models\n",
    "from tqdm import tqdm\n",
    "\n",
    "import matplotlib\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-03-06T00:16:43.249725Z",
     "start_time": "2019-03-06T00:16:43.240611Z"
    }
   },
   "source": [
    "### Import WeightWatcher\n",
    "\n",
    "set custom Logging at WARN Level"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "import logging\n",
    "logging.basicConfig(level=logging.INFO)\n",
    "logger = logging.getLogger(__name__)\n",
    "logger.setLevel(logging.INFO)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'0.5'"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import weightwatcher as ww\n",
    "logger = logging.getLogger(ww.__name__)\n",
    "logger.setLevel(logging.INFO)\n",
    "\n",
    "ww.__version__"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### PyTorch Example\n",
    "\n",
    "VGG16 has 16 layers\n",
    "\n",
    "- The last 3 are dense"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:weightwatcher:\n",
      "\n",
      "python      version 3.9.5 (default, May 18 2021, 12:31:01) \n",
      "[Clang 10.0.0 ]\n",
      "numpy       version 1.19.5\n",
      "tensforflow version 2.5.0\n",
      "keras       version 2.5.0\n",
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': []}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n",
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': []}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n"
     ]
    }
   ],
   "source": [
    "vgg16_pt = models.vgg16(pretrained=True)\n",
    "watcher = ww.WeightWatcher(model=vgg16_pt)\n",
    "details = watcher.describe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>layer_id</th>\n",
       "      <th>name</th>\n",
       "      <th>M</th>\n",
       "      <th>N</th>\n",
       "      <th>layer_type</th>\n",
       "      <th>num_evals</th>\n",
       "      <th>rf</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>2</td>\n",
       "      <td>None</td>\n",
       "      <td>3.0</td>\n",
       "      <td>64.0</td>\n",
       "      <td>LAYER_TYPE.CONV2D</td>\n",
       "      <td>27.0</td>\n",
       "      <td>9.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>4</td>\n",
       "      <td>None</td>\n",
       "      <td>64.0</td>\n",
       "      <td>64.0</td>\n",
       "      <td>LAYER_TYPE.CONV2D</td>\n",
       "      <td>576.0</td>\n",
       "      <td>9.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>7</td>\n",
       "      <td>None</td>\n",
       "      <td>64.0</td>\n",
       "      <td>128.0</td>\n",
       "      <td>LAYER_TYPE.CONV2D</td>\n",
       "      <td>576.0</td>\n",
       "      <td>9.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>9</td>\n",
       "      <td>None</td>\n",
       "      <td>128.0</td>\n",
       "      <td>128.0</td>\n",
       "      <td>LAYER_TYPE.CONV2D</td>\n",
       "      <td>1152.0</td>\n",
       "      <td>9.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>12</td>\n",
       "      <td>None</td>\n",
       "      <td>128.0</td>\n",
       "      <td>256.0</td>\n",
       "      <td>LAYER_TYPE.CONV2D</td>\n",
       "      <td>1152.0</td>\n",
       "      <td>9.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>14</td>\n",
       "      <td>None</td>\n",
       "      <td>256.0</td>\n",
       "      <td>256.0</td>\n",
       "      <td>LAYER_TYPE.CONV2D</td>\n",
       "      <td>2304.0</td>\n",
       "      <td>9.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>16</td>\n",
       "      <td>None</td>\n",
       "      <td>256.0</td>\n",
       "      <td>256.0</td>\n",
       "      <td>LAYER_TYPE.CONV2D</td>\n",
       "      <td>2304.0</td>\n",
       "      <td>9.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>19</td>\n",
       "      <td>None</td>\n",
       "      <td>256.0</td>\n",
       "      <td>512.0</td>\n",
       "      <td>LAYER_TYPE.CONV2D</td>\n",
       "      <td>2304.0</td>\n",
       "      <td>9.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>21</td>\n",
       "      <td>None</td>\n",
       "      <td>512.0</td>\n",
       "      <td>512.0</td>\n",
       "      <td>LAYER_TYPE.CONV2D</td>\n",
       "      <td>4608.0</td>\n",
       "      <td>9.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>23</td>\n",
       "      <td>None</td>\n",
       "      <td>512.0</td>\n",
       "      <td>512.0</td>\n",
       "      <td>LAYER_TYPE.CONV2D</td>\n",
       "      <td>4608.0</td>\n",
       "      <td>9.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>26</td>\n",
       "      <td>None</td>\n",
       "      <td>512.0</td>\n",
       "      <td>512.0</td>\n",
       "      <td>LAYER_TYPE.CONV2D</td>\n",
       "      <td>4608.0</td>\n",
       "      <td>9.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>28</td>\n",
       "      <td>None</td>\n",
       "      <td>512.0</td>\n",
       "      <td>512.0</td>\n",
       "      <td>LAYER_TYPE.CONV2D</td>\n",
       "      <td>4608.0</td>\n",
       "      <td>9.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>30</td>\n",
       "      <td>None</td>\n",
       "      <td>512.0</td>\n",
       "      <td>512.0</td>\n",
       "      <td>LAYER_TYPE.CONV2D</td>\n",
       "      <td>4608.0</td>\n",
       "      <td>9.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>35</td>\n",
       "      <td>None</td>\n",
       "      <td>4096.0</td>\n",
       "      <td>25088.0</td>\n",
       "      <td>LAYER_TYPE.DENSE</td>\n",
       "      <td>4096.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>38</td>\n",
       "      <td>None</td>\n",
       "      <td>4096.0</td>\n",
       "      <td>4096.0</td>\n",
       "      <td>LAYER_TYPE.DENSE</td>\n",
       "      <td>4096.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>41</td>\n",
       "      <td>None</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>4096.0</td>\n",
       "      <td>LAYER_TYPE.DENSE</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   layer_id  name       M        N         layer_type  num_evals   rf\n",
       "0         2  None     3.0     64.0  LAYER_TYPE.CONV2D       27.0  9.0\n",
       "1         4  None    64.0     64.0  LAYER_TYPE.CONV2D      576.0  9.0\n",
       "2         7  None    64.0    128.0  LAYER_TYPE.CONV2D      576.0  9.0\n",
       "3         9  None   128.0    128.0  LAYER_TYPE.CONV2D     1152.0  9.0\n",
       "4        12  None   128.0    256.0  LAYER_TYPE.CONV2D     1152.0  9.0\n",
       "5        14  None   256.0    256.0  LAYER_TYPE.CONV2D     2304.0  9.0\n",
       "6        16  None   256.0    256.0  LAYER_TYPE.CONV2D     2304.0  9.0\n",
       "7        19  None   256.0    512.0  LAYER_TYPE.CONV2D     2304.0  9.0\n",
       "8        21  None   512.0    512.0  LAYER_TYPE.CONV2D     4608.0  9.0\n",
       "9        23  None   512.0    512.0  LAYER_TYPE.CONV2D     4608.0  9.0\n",
       "10       26  None   512.0    512.0  LAYER_TYPE.CONV2D     4608.0  9.0\n",
       "11       28  None   512.0    512.0  LAYER_TYPE.CONV2D     4608.0  9.0\n",
       "12       30  None   512.0    512.0  LAYER_TYPE.CONV2D     4608.0  9.0\n",
       "13       35  None  4096.0  25088.0   LAYER_TYPE.DENSE     4096.0  1.0\n",
       "14       38  None  4096.0   4096.0   LAYER_TYPE.DENSE     4096.0  1.0\n",
       "15       41  None  1000.0   4096.0   LAYER_TYPE.DENSE     1000.0  1.0"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "details"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': []}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n",
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': []}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:weightwatcher:Filtering layer by id 2\n",
      "INFO:weightwatcher:skipping layer 0 None by id\n",
      "INFO:weightwatcher:skipping layer 1 None by id\n",
      "INFO:weightwatcher:keeping layer 2 None by id\n",
      "INFO:weightwatcher:Getting ESD for layer 2 \n",
      "INFO:weightwatcher:Found 27 eiganvalues for 2 None\n",
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': []}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n",
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': []}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "4\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:weightwatcher:Filtering layer by id 4\n",
      "INFO:weightwatcher:skipping layer 0 None by id\n",
      "INFO:weightwatcher:skipping layer 1 None by id\n",
      "INFO:weightwatcher:skipping layer 2 None by id\n",
      "INFO:weightwatcher:skipping layer 3 None by id\n",
      "INFO:weightwatcher:keeping layer 4 None by id\n",
      "INFO:weightwatcher:Getting ESD for layer 4 \n",
      "INFO:weightwatcher:Found 576 eiganvalues for 4 None\n",
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': []}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n",
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': []}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:weightwatcher:Filtering layer by id 7\n",
      "INFO:weightwatcher:skipping layer 0 None by id\n",
      "INFO:weightwatcher:skipping layer 1 None by id\n",
      "INFO:weightwatcher:skipping layer 2 None by id\n",
      "INFO:weightwatcher:skipping layer 3 None by id\n",
      "INFO:weightwatcher:skipping layer 4 None by id\n",
      "INFO:weightwatcher:skipping layer 5 None by id\n",
      "INFO:weightwatcher:skipping layer 6 None by id\n",
      "INFO:weightwatcher:keeping layer 7 None by id\n",
      "INFO:weightwatcher:Getting ESD for layer 7 \n",
      "INFO:weightwatcher:Found 576 eiganvalues for 7 None\n",
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': []}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n",
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': []}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "9\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:weightwatcher:Filtering layer by id 9\n",
      "INFO:weightwatcher:skipping layer 0 None by id\n",
      "INFO:weightwatcher:skipping layer 1 None by id\n",
      "INFO:weightwatcher:skipping layer 2 None by id\n",
      "INFO:weightwatcher:skipping layer 3 None by id\n",
      "INFO:weightwatcher:skipping layer 4 None by id\n",
      "INFO:weightwatcher:skipping layer 5 None by id\n",
      "INFO:weightwatcher:skipping layer 6 None by id\n",
      "INFO:weightwatcher:skipping layer 7 None by id\n",
      "INFO:weightwatcher:skipping layer 8 None by id\n",
      "INFO:weightwatcher:keeping layer 9 None by id\n",
      "INFO:weightwatcher:Getting ESD for layer 9 \n",
      "INFO:weightwatcher:Found 1152 eiganvalues for 9 None\n",
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': []}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n",
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': []}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "12\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:weightwatcher:Filtering layer by id 12\n",
      "INFO:weightwatcher:skipping layer 0 None by id\n",
      "INFO:weightwatcher:skipping layer 1 None by id\n",
      "INFO:weightwatcher:skipping layer 2 None by id\n",
      "INFO:weightwatcher:skipping layer 3 None by id\n",
      "INFO:weightwatcher:skipping layer 4 None by id\n",
      "INFO:weightwatcher:skipping layer 5 None by id\n",
      "INFO:weightwatcher:skipping layer 6 None by id\n",
      "INFO:weightwatcher:skipping layer 7 None by id\n",
      "INFO:weightwatcher:skipping layer 8 None by id\n",
      "INFO:weightwatcher:skipping layer 9 None by id\n",
      "INFO:weightwatcher:skipping layer 10 None by id\n",
      "INFO:weightwatcher:skipping layer 11 None by id\n",
      "INFO:weightwatcher:keeping layer 12 None by id\n",
      "INFO:weightwatcher:Getting ESD for layer 12 \n",
      "INFO:weightwatcher:Found 1152 eiganvalues for 12 None\n",
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': []}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n",
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': []}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "14\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:weightwatcher:Filtering layer by id 14\n",
      "INFO:weightwatcher:skipping layer 0 None by id\n",
      "INFO:weightwatcher:skipping layer 1 None by id\n",
      "INFO:weightwatcher:skipping layer 2 None by id\n",
      "INFO:weightwatcher:skipping layer 3 None by id\n",
      "INFO:weightwatcher:skipping layer 4 None by id\n",
      "INFO:weightwatcher:skipping layer 5 None by id\n",
      "INFO:weightwatcher:skipping layer 6 None by id\n",
      "INFO:weightwatcher:skipping layer 7 None by id\n",
      "INFO:weightwatcher:skipping layer 8 None by id\n",
      "INFO:weightwatcher:skipping layer 9 None by id\n",
      "INFO:weightwatcher:skipping layer 10 None by id\n",
      "INFO:weightwatcher:skipping layer 11 None by id\n",
      "INFO:weightwatcher:skipping layer 12 None by id\n",
      "INFO:weightwatcher:skipping layer 13 None by id\n",
      "INFO:weightwatcher:keeping layer 14 None by id\n",
      "INFO:weightwatcher:Getting ESD for layer 14 \n",
      "INFO:weightwatcher:Found 2304 eiganvalues for 14 None\n",
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': []}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n",
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': []}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "16\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:weightwatcher:Filtering layer by id 16\n",
      "INFO:weightwatcher:skipping layer 0 None by id\n",
      "INFO:weightwatcher:skipping layer 1 None by id\n",
      "INFO:weightwatcher:skipping layer 2 None by id\n",
      "INFO:weightwatcher:skipping layer 3 None by id\n",
      "INFO:weightwatcher:skipping layer 4 None by id\n",
      "INFO:weightwatcher:skipping layer 5 None by id\n",
      "INFO:weightwatcher:skipping layer 6 None by id\n",
      "INFO:weightwatcher:skipping layer 7 None by id\n",
      "INFO:weightwatcher:skipping layer 8 None by id\n",
      "INFO:weightwatcher:skipping layer 9 None by id\n",
      "INFO:weightwatcher:skipping layer 10 None by id\n",
      "INFO:weightwatcher:skipping layer 11 None by id\n",
      "INFO:weightwatcher:skipping layer 12 None by id\n",
      "INFO:weightwatcher:skipping layer 13 None by id\n",
      "INFO:weightwatcher:skipping layer 14 None by id\n",
      "INFO:weightwatcher:skipping layer 15 None by id\n",
      "INFO:weightwatcher:keeping layer 16 None by id\n",
      "INFO:weightwatcher:Getting ESD for layer 16 \n",
      "INFO:weightwatcher:Found 2304 eiganvalues for 16 None\n",
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': []}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n",
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': []}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "19\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:weightwatcher:Filtering layer by id 19\n",
      "INFO:weightwatcher:skipping layer 0 None by id\n",
      "INFO:weightwatcher:skipping layer 1 None by id\n",
      "INFO:weightwatcher:skipping layer 2 None by id\n",
      "INFO:weightwatcher:skipping layer 3 None by id\n",
      "INFO:weightwatcher:skipping layer 4 None by id\n",
      "INFO:weightwatcher:skipping layer 5 None by id\n",
      "INFO:weightwatcher:skipping layer 6 None by id\n",
      "INFO:weightwatcher:skipping layer 7 None by id\n",
      "INFO:weightwatcher:skipping layer 8 None by id\n",
      "INFO:weightwatcher:skipping layer 9 None by id\n",
      "INFO:weightwatcher:skipping layer 10 None by id\n",
      "INFO:weightwatcher:skipping layer 11 None by id\n",
      "INFO:weightwatcher:skipping layer 12 None by id\n",
      "INFO:weightwatcher:skipping layer 13 None by id\n",
      "INFO:weightwatcher:skipping layer 14 None by id\n",
      "INFO:weightwatcher:skipping layer 15 None by id\n",
      "INFO:weightwatcher:skipping layer 16 None by id\n",
      "INFO:weightwatcher:skipping layer 17 None by id\n",
      "INFO:weightwatcher:skipping layer 18 None by id\n",
      "INFO:weightwatcher:keeping layer 19 None by id\n",
      "INFO:weightwatcher:Getting ESD for layer 19 \n",
      "INFO:weightwatcher:Found 2304 eiganvalues for 19 None\n",
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': []}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n",
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': []}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "21\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:weightwatcher:Filtering layer by id 21\n",
      "INFO:weightwatcher:skipping layer 0 None by id\n",
      "INFO:weightwatcher:skipping layer 1 None by id\n",
      "INFO:weightwatcher:skipping layer 2 None by id\n",
      "INFO:weightwatcher:skipping layer 3 None by id\n",
      "INFO:weightwatcher:skipping layer 4 None by id\n",
      "INFO:weightwatcher:skipping layer 5 None by id\n",
      "INFO:weightwatcher:skipping layer 6 None by id\n",
      "INFO:weightwatcher:skipping layer 7 None by id\n",
      "INFO:weightwatcher:skipping layer 8 None by id\n",
      "INFO:weightwatcher:skipping layer 9 None by id\n",
      "INFO:weightwatcher:skipping layer 10 None by id\n",
      "INFO:weightwatcher:skipping layer 11 None by id\n",
      "INFO:weightwatcher:skipping layer 12 None by id\n",
      "INFO:weightwatcher:skipping layer 13 None by id\n",
      "INFO:weightwatcher:skipping layer 14 None by id\n",
      "INFO:weightwatcher:skipping layer 15 None by id\n",
      "INFO:weightwatcher:skipping layer 16 None by id\n",
      "INFO:weightwatcher:skipping layer 17 None by id\n",
      "INFO:weightwatcher:skipping layer 18 None by id\n",
      "INFO:weightwatcher:skipping layer 19 None by id\n",
      "INFO:weightwatcher:skipping layer 20 None by id\n",
      "INFO:weightwatcher:keeping layer 21 None by id\n",
      "INFO:weightwatcher:Getting ESD for layer 21 \n",
      "INFO:weightwatcher:Found 4608 eiganvalues for 21 None\n",
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': []}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n",
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': []}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "23\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:weightwatcher:Filtering layer by id 23\n",
      "INFO:weightwatcher:skipping layer 0 None by id\n",
      "INFO:weightwatcher:skipping layer 1 None by id\n",
      "INFO:weightwatcher:skipping layer 2 None by id\n",
      "INFO:weightwatcher:skipping layer 3 None by id\n",
      "INFO:weightwatcher:skipping layer 4 None by id\n",
      "INFO:weightwatcher:skipping layer 5 None by id\n",
      "INFO:weightwatcher:skipping layer 6 None by id\n",
      "INFO:weightwatcher:skipping layer 7 None by id\n",
      "INFO:weightwatcher:skipping layer 8 None by id\n",
      "INFO:weightwatcher:skipping layer 9 None by id\n",
      "INFO:weightwatcher:skipping layer 10 None by id\n",
      "INFO:weightwatcher:skipping layer 11 None by id\n",
      "INFO:weightwatcher:skipping layer 12 None by id\n",
      "INFO:weightwatcher:skipping layer 13 None by id\n",
      "INFO:weightwatcher:skipping layer 14 None by id\n",
      "INFO:weightwatcher:skipping layer 15 None by id\n",
      "INFO:weightwatcher:skipping layer 16 None by id\n",
      "INFO:weightwatcher:skipping layer 17 None by id\n",
      "INFO:weightwatcher:skipping layer 18 None by id\n",
      "INFO:weightwatcher:skipping layer 19 None by id\n",
      "INFO:weightwatcher:skipping layer 20 None by id\n",
      "INFO:weightwatcher:skipping layer 21 None by id\n",
      "INFO:weightwatcher:skipping layer 22 None by id\n",
      "INFO:weightwatcher:keeping layer 23 None by id\n",
      "INFO:weightwatcher:Getting ESD for layer 23 \n",
      "INFO:weightwatcher:Found 4608 eiganvalues for 23 None\n",
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': []}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n",
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': []}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "26\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:weightwatcher:Filtering layer by id 26\n",
      "INFO:weightwatcher:skipping layer 0 None by id\n",
      "INFO:weightwatcher:skipping layer 1 None by id\n",
      "INFO:weightwatcher:skipping layer 2 None by id\n",
      "INFO:weightwatcher:skipping layer 3 None by id\n",
      "INFO:weightwatcher:skipping layer 4 None by id\n",
      "INFO:weightwatcher:skipping layer 5 None by id\n",
      "INFO:weightwatcher:skipping layer 6 None by id\n",
      "INFO:weightwatcher:skipping layer 7 None by id\n",
      "INFO:weightwatcher:skipping layer 8 None by id\n",
      "INFO:weightwatcher:skipping layer 9 None by id\n",
      "INFO:weightwatcher:skipping layer 10 None by id\n",
      "INFO:weightwatcher:skipping layer 11 None by id\n",
      "INFO:weightwatcher:skipping layer 12 None by id\n",
      "INFO:weightwatcher:skipping layer 13 None by id\n",
      "INFO:weightwatcher:skipping layer 14 None by id\n",
      "INFO:weightwatcher:skipping layer 15 None by id\n",
      "INFO:weightwatcher:skipping layer 16 None by id\n",
      "INFO:weightwatcher:skipping layer 17 None by id\n",
      "INFO:weightwatcher:skipping layer 18 None by id\n",
      "INFO:weightwatcher:skipping layer 19 None by id\n",
      "INFO:weightwatcher:skipping layer 20 None by id\n",
      "INFO:weightwatcher:skipping layer 21 None by id\n",
      "INFO:weightwatcher:skipping layer 22 None by id\n",
      "INFO:weightwatcher:skipping layer 23 None by id\n",
      "INFO:weightwatcher:skipping layer 24 None by id\n",
      "INFO:weightwatcher:skipping layer 25 None by id\n",
      "INFO:weightwatcher:keeping layer 26 None by id\n",
      "INFO:weightwatcher:Getting ESD for layer 26 \n",
      "INFO:weightwatcher:Found 4608 eiganvalues for 26 None\n",
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': []}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n",
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': []}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "28\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:weightwatcher:Filtering layer by id 28\n",
      "INFO:weightwatcher:skipping layer 0 None by id\n",
      "INFO:weightwatcher:skipping layer 1 None by id\n",
      "INFO:weightwatcher:skipping layer 2 None by id\n",
      "INFO:weightwatcher:skipping layer 3 None by id\n",
      "INFO:weightwatcher:skipping layer 4 None by id\n",
      "INFO:weightwatcher:skipping layer 5 None by id\n",
      "INFO:weightwatcher:skipping layer 6 None by id\n",
      "INFO:weightwatcher:skipping layer 7 None by id\n",
      "INFO:weightwatcher:skipping layer 8 None by id\n",
      "INFO:weightwatcher:skipping layer 9 None by id\n",
      "INFO:weightwatcher:skipping layer 10 None by id\n",
      "INFO:weightwatcher:skipping layer 11 None by id\n",
      "INFO:weightwatcher:skipping layer 12 None by id\n",
      "INFO:weightwatcher:skipping layer 13 None by id\n",
      "INFO:weightwatcher:skipping layer 14 None by id\n",
      "INFO:weightwatcher:skipping layer 15 None by id\n",
      "INFO:weightwatcher:skipping layer 16 None by id\n",
      "INFO:weightwatcher:skipping layer 17 None by id\n",
      "INFO:weightwatcher:skipping layer 18 None by id\n",
      "INFO:weightwatcher:skipping layer 19 None by id\n",
      "INFO:weightwatcher:skipping layer 20 None by id\n",
      "INFO:weightwatcher:skipping layer 21 None by id\n",
      "INFO:weightwatcher:skipping layer 22 None by id\n",
      "INFO:weightwatcher:skipping layer 23 None by id\n",
      "INFO:weightwatcher:skipping layer 24 None by id\n",
      "INFO:weightwatcher:skipping layer 25 None by id\n",
      "INFO:weightwatcher:skipping layer 26 None by id\n",
      "INFO:weightwatcher:skipping layer 27 None by id\n",
      "INFO:weightwatcher:keeping layer 28 None by id\n",
      "INFO:weightwatcher:Getting ESD for layer 28 \n",
      "INFO:weightwatcher:Found 4608 eiganvalues for 28 None\n",
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': []}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n",
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': []}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "30\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:weightwatcher:Filtering layer by id 30\n",
      "INFO:weightwatcher:skipping layer 0 None by id\n",
      "INFO:weightwatcher:skipping layer 1 None by id\n",
      "INFO:weightwatcher:skipping layer 2 None by id\n",
      "INFO:weightwatcher:skipping layer 3 None by id\n",
      "INFO:weightwatcher:skipping layer 4 None by id\n",
      "INFO:weightwatcher:skipping layer 5 None by id\n",
      "INFO:weightwatcher:skipping layer 6 None by id\n",
      "INFO:weightwatcher:skipping layer 7 None by id\n",
      "INFO:weightwatcher:skipping layer 8 None by id\n",
      "INFO:weightwatcher:skipping layer 9 None by id\n",
      "INFO:weightwatcher:skipping layer 10 None by id\n",
      "INFO:weightwatcher:skipping layer 11 None by id\n",
      "INFO:weightwatcher:skipping layer 12 None by id\n",
      "INFO:weightwatcher:skipping layer 13 None by id\n",
      "INFO:weightwatcher:skipping layer 14 None by id\n",
      "INFO:weightwatcher:skipping layer 15 None by id\n",
      "INFO:weightwatcher:skipping layer 16 None by id\n",
      "INFO:weightwatcher:skipping layer 17 None by id\n",
      "INFO:weightwatcher:skipping layer 18 None by id\n",
      "INFO:weightwatcher:skipping layer 19 None by id\n",
      "INFO:weightwatcher:skipping layer 20 None by id\n",
      "INFO:weightwatcher:skipping layer 21 None by id\n",
      "INFO:weightwatcher:skipping layer 22 None by id\n",
      "INFO:weightwatcher:skipping layer 23 None by id\n",
      "INFO:weightwatcher:skipping layer 24 None by id\n",
      "INFO:weightwatcher:skipping layer 25 None by id\n",
      "INFO:weightwatcher:skipping layer 26 None by id\n",
      "INFO:weightwatcher:skipping layer 27 None by id\n",
      "INFO:weightwatcher:skipping layer 28 None by id\n",
      "INFO:weightwatcher:skipping layer 29 None by id\n",
      "INFO:weightwatcher:keeping layer 30 None by id\n",
      "INFO:weightwatcher:Getting ESD for layer 30 \n",
      "INFO:weightwatcher:Found 4608 eiganvalues for 30 None\n",
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': []}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n",
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': []}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "35\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:weightwatcher:Filtering layer by id 35\n",
      "INFO:weightwatcher:skipping layer 0 None by id\n",
      "INFO:weightwatcher:skipping layer 1 None by id\n",
      "INFO:weightwatcher:skipping layer 2 None by id\n",
      "INFO:weightwatcher:skipping layer 3 None by id\n",
      "INFO:weightwatcher:skipping layer 4 None by id\n",
      "INFO:weightwatcher:skipping layer 5 None by id\n",
      "INFO:weightwatcher:skipping layer 6 None by id\n",
      "INFO:weightwatcher:skipping layer 7 None by id\n",
      "INFO:weightwatcher:skipping layer 8 None by id\n",
      "INFO:weightwatcher:skipping layer 9 None by id\n",
      "INFO:weightwatcher:skipping layer 10 None by id\n",
      "INFO:weightwatcher:skipping layer 11 None by id\n",
      "INFO:weightwatcher:skipping layer 12 None by id\n",
      "INFO:weightwatcher:skipping layer 13 None by id\n",
      "INFO:weightwatcher:skipping layer 14 None by id\n",
      "INFO:weightwatcher:skipping layer 15 None by id\n",
      "INFO:weightwatcher:skipping layer 16 None by id\n",
      "INFO:weightwatcher:skipping layer 17 None by id\n",
      "INFO:weightwatcher:skipping layer 18 None by id\n",
      "INFO:weightwatcher:skipping layer 19 None by id\n",
      "INFO:weightwatcher:skipping layer 20 None by id\n",
      "INFO:weightwatcher:skipping layer 21 None by id\n",
      "INFO:weightwatcher:skipping layer 22 None by id\n",
      "INFO:weightwatcher:skipping layer 23 None by id\n",
      "INFO:weightwatcher:skipping layer 24 None by id\n",
      "INFO:weightwatcher:skipping layer 25 None by id\n",
      "INFO:weightwatcher:skipping layer 26 None by id\n",
      "INFO:weightwatcher:skipping layer 27 None by id\n",
      "INFO:weightwatcher:skipping layer 28 None by id\n",
      "INFO:weightwatcher:skipping layer 29 None by id\n",
      "INFO:weightwatcher:skipping layer 30 None by id\n",
      "INFO:weightwatcher:skipping layer 31 None by id\n",
      "INFO:weightwatcher:skipping layer 32 None by id\n",
      "INFO:weightwatcher:skipping layer 33 None by id\n",
      "INFO:weightwatcher:skipping layer 34 None by id\n",
      "INFO:weightwatcher:keeping layer 35 None by id\n",
      "INFO:weightwatcher:Getting ESD for layer 35 \n",
      "INFO:weightwatcher:Found 4096 eiganvalues for 35 None\n",
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': []}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n",
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': []}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "38\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:weightwatcher:Filtering layer by id 38\n",
      "INFO:weightwatcher:skipping layer 0 None by id\n",
      "INFO:weightwatcher:skipping layer 1 None by id\n",
      "INFO:weightwatcher:skipping layer 2 None by id\n",
      "INFO:weightwatcher:skipping layer 3 None by id\n",
      "INFO:weightwatcher:skipping layer 4 None by id\n",
      "INFO:weightwatcher:skipping layer 5 None by id\n",
      "INFO:weightwatcher:skipping layer 6 None by id\n",
      "INFO:weightwatcher:skipping layer 7 None by id\n",
      "INFO:weightwatcher:skipping layer 8 None by id\n",
      "INFO:weightwatcher:skipping layer 9 None by id\n",
      "INFO:weightwatcher:skipping layer 10 None by id\n",
      "INFO:weightwatcher:skipping layer 11 None by id\n",
      "INFO:weightwatcher:skipping layer 12 None by id\n",
      "INFO:weightwatcher:skipping layer 13 None by id\n",
      "INFO:weightwatcher:skipping layer 14 None by id\n",
      "INFO:weightwatcher:skipping layer 15 None by id\n",
      "INFO:weightwatcher:skipping layer 16 None by id\n",
      "INFO:weightwatcher:skipping layer 17 None by id\n",
      "INFO:weightwatcher:skipping layer 18 None by id\n",
      "INFO:weightwatcher:skipping layer 19 None by id\n",
      "INFO:weightwatcher:skipping layer 20 None by id\n",
      "INFO:weightwatcher:skipping layer 21 None by id\n",
      "INFO:weightwatcher:skipping layer 22 None by id\n",
      "INFO:weightwatcher:skipping layer 23 None by id\n",
      "INFO:weightwatcher:skipping layer 24 None by id\n",
      "INFO:weightwatcher:skipping layer 25 None by id\n",
      "INFO:weightwatcher:skipping layer 26 None by id\n",
      "INFO:weightwatcher:skipping layer 27 None by id\n",
      "INFO:weightwatcher:skipping layer 28 None by id\n",
      "INFO:weightwatcher:skipping layer 29 None by id\n",
      "INFO:weightwatcher:skipping layer 30 None by id\n",
      "INFO:weightwatcher:skipping layer 31 None by id\n",
      "INFO:weightwatcher:skipping layer 32 None by id\n",
      "INFO:weightwatcher:skipping layer 33 None by id\n",
      "INFO:weightwatcher:skipping layer 34 None by id\n",
      "INFO:weightwatcher:skipping layer 35 None by id\n",
      "INFO:weightwatcher:skipping layer 36 None by id\n",
      "INFO:weightwatcher:skipping layer 37 None by id\n",
      "INFO:weightwatcher:keeping layer 38 None by id\n",
      "INFO:weightwatcher:Getting ESD for layer 38 \n",
      "INFO:weightwatcher:Found 4096 eiganvalues for 38 None\n",
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': []}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n",
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': []}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "41\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:weightwatcher:Filtering layer by id 41\n",
      "INFO:weightwatcher:skipping layer 0 None by id\n",
      "INFO:weightwatcher:skipping layer 1 None by id\n",
      "INFO:weightwatcher:skipping layer 2 None by id\n",
      "INFO:weightwatcher:skipping layer 3 None by id\n",
      "INFO:weightwatcher:skipping layer 4 None by id\n",
      "INFO:weightwatcher:skipping layer 5 None by id\n",
      "INFO:weightwatcher:skipping layer 6 None by id\n",
      "INFO:weightwatcher:skipping layer 7 None by id\n",
      "INFO:weightwatcher:skipping layer 8 None by id\n",
      "INFO:weightwatcher:skipping layer 9 None by id\n",
      "INFO:weightwatcher:skipping layer 10 None by id\n",
      "INFO:weightwatcher:skipping layer 11 None by id\n",
      "INFO:weightwatcher:skipping layer 12 None by id\n",
      "INFO:weightwatcher:skipping layer 13 None by id\n",
      "INFO:weightwatcher:skipping layer 14 None by id\n",
      "INFO:weightwatcher:skipping layer 15 None by id\n",
      "INFO:weightwatcher:skipping layer 16 None by id\n",
      "INFO:weightwatcher:skipping layer 17 None by id\n",
      "INFO:weightwatcher:skipping layer 18 None by id\n",
      "INFO:weightwatcher:skipping layer 19 None by id\n",
      "INFO:weightwatcher:skipping layer 20 None by id\n",
      "INFO:weightwatcher:skipping layer 21 None by id\n",
      "INFO:weightwatcher:skipping layer 22 None by id\n",
      "INFO:weightwatcher:skipping layer 23 None by id\n",
      "INFO:weightwatcher:skipping layer 24 None by id\n",
      "INFO:weightwatcher:skipping layer 25 None by id\n",
      "INFO:weightwatcher:skipping layer 26 None by id\n",
      "INFO:weightwatcher:skipping layer 27 None by id\n",
      "INFO:weightwatcher:skipping layer 28 None by id\n",
      "INFO:weightwatcher:skipping layer 29 None by id\n",
      "INFO:weightwatcher:skipping layer 30 None by id\n",
      "INFO:weightwatcher:skipping layer 31 None by id\n",
      "INFO:weightwatcher:skipping layer 32 None by id\n",
      "INFO:weightwatcher:skipping layer 33 None by id\n",
      "INFO:weightwatcher:skipping layer 34 None by id\n",
      "INFO:weightwatcher:skipping layer 35 None by id\n",
      "INFO:weightwatcher:skipping layer 36 None by id\n",
      "INFO:weightwatcher:skipping layer 37 None by id\n",
      "INFO:weightwatcher:skipping layer 38 None by id\n",
      "INFO:weightwatcher:skipping layer 39 None by id\n",
      "INFO:weightwatcher:skipping layer 40 None by id\n",
      "INFO:weightwatcher:keeping layer 41 None by id\n",
      "INFO:weightwatcher:Getting ESD for layer 41 \n",
      "INFO:weightwatcher:Found 1000 eiganvalues for 41 None\n"
     ]
    }
   ],
   "source": [
    "all_esds = []\n",
    "for layer_id in details.layer_id.to_numpy():\n",
    "    print(layer_id)\n",
    "    esd = watcher.get_ESD(layer=layer_id) \n",
    "    all_esds.append(esd)\n",
    "    #plt.hist(np.log10(esd[esd>0.0]), bins=100, color='green', label='vgg16', alpha=0.5);\n",
    "\n",
    "    #plt.title(\"Layer {}, VGG16\".format(layer_id))\n",
    "    #plt.xlabel(r\"$log_{10} \\lambda $ Log 10 Eigenvalues\")\n",
    "    #plt.legend()\n",
    "    #plt.show()\n",
    "    \n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': []}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n",
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': []}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n",
      "INFO:weightwatcher:LAYER: 2 16  : <class 'torch.nn.modules.conv.Conv2d'>\n",
      "INFO:weightwatcher:apply truncated SVD on Layer 2 None, with nsmooth=5,  keeping ncomp=5 out of 3. of the singular vectors\n",
      "INFO:weightwatcher:LAYER TYPE  16 out of 4 8 64 \n",
      "INFO:weightwatcher:LAYER: 4 16  : <class 'torch.nn.modules.conv.Conv2d'>\n",
      "INFO:weightwatcher:apply truncated SVD on Layer 4 None, with nsmooth=115,  keeping ncomp=115 out of 64. of the singular vectors\n",
      "INFO:weightwatcher:LAYER TYPE  16 out of 4 8 64 \n",
      "INFO:weightwatcher:LAYER: 7 16  : <class 'torch.nn.modules.conv.Conv2d'>\n",
      "INFO:weightwatcher:apply truncated SVD on Layer 7 None, with nsmooth=115,  keeping ncomp=115 out of 64. of the singular vectors\n",
      "INFO:weightwatcher:LAYER TYPE  16 out of 4 8 64 \n",
      "INFO:weightwatcher:LAYER: 9 16  : <class 'torch.nn.modules.conv.Conv2d'>\n",
      "INFO:weightwatcher:apply truncated SVD on Layer 9 None, with nsmooth=230,  keeping ncomp=230 out of 128. of the singular vectors\n",
      "INFO:weightwatcher:LAYER TYPE  16 out of 4 8 64 \n",
      "INFO:weightwatcher:LAYER: 12 16  : <class 'torch.nn.modules.conv.Conv2d'>\n",
      "INFO:weightwatcher:apply truncated SVD on Layer 12 None, with nsmooth=230,  keeping ncomp=230 out of 128. of the singular vectors\n",
      "INFO:weightwatcher:LAYER TYPE  16 out of 4 8 64 \n",
      "INFO:weightwatcher:LAYER: 14 16  : <class 'torch.nn.modules.conv.Conv2d'>\n",
      "INFO:weightwatcher:apply truncated SVD on Layer 14 None, with nsmooth=460,  keeping ncomp=460 out of 256. of the singular vectors\n",
      "INFO:weightwatcher:LAYER TYPE  16 out of 4 8 64 \n",
      "INFO:weightwatcher:LAYER: 16 16  : <class 'torch.nn.modules.conv.Conv2d'>\n",
      "INFO:weightwatcher:apply truncated SVD on Layer 16 None, with nsmooth=460,  keeping ncomp=460 out of 256. of the singular vectors\n",
      "INFO:weightwatcher:LAYER TYPE  16 out of 4 8 64 \n",
      "INFO:weightwatcher:LAYER: 19 16  : <class 'torch.nn.modules.conv.Conv2d'>\n",
      "INFO:weightwatcher:apply truncated SVD on Layer 19 None, with nsmooth=460,  keeping ncomp=460 out of 256. of the singular vectors\n",
      "INFO:weightwatcher:LAYER TYPE  16 out of 4 8 64 \n",
      "INFO:weightwatcher:LAYER: 21 16  : <class 'torch.nn.modules.conv.Conv2d'>\n",
      "INFO:weightwatcher:apply truncated SVD on Layer 21 None, with nsmooth=921,  keeping ncomp=921 out of 512. of the singular vectors\n",
      "INFO:weightwatcher:LAYER TYPE  16 out of 4 8 64 \n",
      "INFO:weightwatcher:LAYER: 23 16  : <class 'torch.nn.modules.conv.Conv2d'>\n",
      "INFO:weightwatcher:apply truncated SVD on Layer 23 None, with nsmooth=921,  keeping ncomp=921 out of 512. of the singular vectors\n",
      "INFO:weightwatcher:LAYER TYPE  16 out of 4 8 64 \n",
      "INFO:weightwatcher:LAYER: 26 16  : <class 'torch.nn.modules.conv.Conv2d'>\n",
      "INFO:weightwatcher:apply truncated SVD on Layer 26 None, with nsmooth=921,  keeping ncomp=921 out of 512. of the singular vectors\n",
      "INFO:weightwatcher:LAYER TYPE  16 out of 4 8 64 \n",
      "INFO:weightwatcher:LAYER: 28 16  : <class 'torch.nn.modules.conv.Conv2d'>\n",
      "INFO:weightwatcher:apply truncated SVD on Layer 28 None, with nsmooth=921,  keeping ncomp=921 out of 512. of the singular vectors\n",
      "INFO:weightwatcher:LAYER TYPE  16 out of 4 8 64 \n",
      "INFO:weightwatcher:LAYER: 30 16  : <class 'torch.nn.modules.conv.Conv2d'>\n",
      "INFO:weightwatcher:apply truncated SVD on Layer 30 None, with nsmooth=921,  keeping ncomp=921 out of 512. of the singular vectors\n",
      "INFO:weightwatcher:LAYER TYPE  16 out of 4 8 64 \n",
      "INFO:weightwatcher:LAYER: 35 4  : <class 'torch.nn.modules.linear.Linear'>\n",
      "INFO:weightwatcher:apply truncated SVD on Layer 35 None, with nsmooth=819,  keeping ncomp=819 out of 4096. of the singular vectors\n",
      "INFO:weightwatcher:LAYER TYPE  4 out of 4 8 64 \n",
      "INFO:weightwatcher:LAYER: 38 4  : <class 'torch.nn.modules.linear.Linear'>\n",
      "INFO:weightwatcher:apply truncated SVD on Layer 38 None, with nsmooth=819,  keeping ncomp=819 out of 4096. of the singular vectors\n",
      "INFO:weightwatcher:LAYER TYPE  4 out of 4 8 64 \n",
      "INFO:weightwatcher:LAYER: 41 4  : <class 'torch.nn.modules.linear.Linear'>\n",
      "INFO:weightwatcher:apply truncated SVD on Layer 41 None, with nsmooth=200,  keeping ncomp=200 out of 1000. of the singular vectors\n",
      "INFO:weightwatcher:LAYER TYPE  4 out of 4 8 64 \n",
      "INFO:weightwatcher:Returning smoothed model\n"
     ]
    }
   ],
   "source": [
    "smoother_vgg16 = watcher.SVDSmoothing(percent=0.2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': [], 'num_smooth': 200}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n",
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': [], 'num_smooth': 200}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:weightwatcher:Filtering layer by id 2\n",
      "INFO:weightwatcher:skipping layer 0 None by id\n",
      "INFO:weightwatcher:skipping layer 1 None by id\n",
      "INFO:weightwatcher:keeping layer 2 None by id\n",
      "INFO:weightwatcher:Getting ESD for layer 2 \n",
      "INFO:weightwatcher:Found 27 eiganvalues for 2 None\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': [], 'num_smooth': 200}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n",
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': [], 'num_smooth': 200}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "4\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:weightwatcher:Filtering layer by id 4\n",
      "INFO:weightwatcher:skipping layer 0 None by id\n",
      "INFO:weightwatcher:skipping layer 1 None by id\n",
      "INFO:weightwatcher:skipping layer 2 None by id\n",
      "INFO:weightwatcher:skipping layer 3 None by id\n",
      "INFO:weightwatcher:keeping layer 4 None by id\n",
      "INFO:weightwatcher:Getting ESD for layer 4 \n",
      "INFO:weightwatcher:Found 576 eiganvalues for 4 None\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': [], 'num_smooth': 200}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n",
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': [], 'num_smooth': 200}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:weightwatcher:Filtering layer by id 7\n",
      "INFO:weightwatcher:skipping layer 0 None by id\n",
      "INFO:weightwatcher:skipping layer 1 None by id\n",
      "INFO:weightwatcher:skipping layer 2 None by id\n",
      "INFO:weightwatcher:skipping layer 3 None by id\n",
      "INFO:weightwatcher:skipping layer 4 None by id\n",
      "INFO:weightwatcher:skipping layer 5 None by id\n",
      "INFO:weightwatcher:skipping layer 6 None by id\n",
      "INFO:weightwatcher:keeping layer 7 None by id\n",
      "INFO:weightwatcher:Getting ESD for layer 7 \n",
      "INFO:weightwatcher:Found 576 eiganvalues for 7 None\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': [], 'num_smooth': 200}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n",
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': [], 'num_smooth': 200}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "9\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:weightwatcher:Filtering layer by id 9\n",
      "INFO:weightwatcher:skipping layer 0 None by id\n",
      "INFO:weightwatcher:skipping layer 1 None by id\n",
      "INFO:weightwatcher:skipping layer 2 None by id\n",
      "INFO:weightwatcher:skipping layer 3 None by id\n",
      "INFO:weightwatcher:skipping layer 4 None by id\n",
      "INFO:weightwatcher:skipping layer 5 None by id\n",
      "INFO:weightwatcher:skipping layer 6 None by id\n",
      "INFO:weightwatcher:skipping layer 7 None by id\n",
      "INFO:weightwatcher:skipping layer 8 None by id\n",
      "INFO:weightwatcher:keeping layer 9 None by id\n",
      "INFO:weightwatcher:Getting ESD for layer 9 \n",
      "INFO:weightwatcher:Found 1152 eiganvalues for 9 None\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': [], 'num_smooth': 200}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n",
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': [], 'num_smooth': 200}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "12\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:weightwatcher:Filtering layer by id 12\n",
      "INFO:weightwatcher:skipping layer 0 None by id\n",
      "INFO:weightwatcher:skipping layer 1 None by id\n",
      "INFO:weightwatcher:skipping layer 2 None by id\n",
      "INFO:weightwatcher:skipping layer 3 None by id\n",
      "INFO:weightwatcher:skipping layer 4 None by id\n",
      "INFO:weightwatcher:skipping layer 5 None by id\n",
      "INFO:weightwatcher:skipping layer 6 None by id\n",
      "INFO:weightwatcher:skipping layer 7 None by id\n",
      "INFO:weightwatcher:skipping layer 8 None by id\n",
      "INFO:weightwatcher:skipping layer 9 None by id\n",
      "INFO:weightwatcher:skipping layer 10 None by id\n",
      "INFO:weightwatcher:skipping layer 11 None by id\n",
      "INFO:weightwatcher:keeping layer 12 None by id\n",
      "INFO:weightwatcher:Getting ESD for layer 12 \n",
      "INFO:weightwatcher:Found 1152 eiganvalues for 12 None\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': [], 'num_smooth': 200}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n",
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': [], 'num_smooth': 200}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "14\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:weightwatcher:Filtering layer by id 14\n",
      "INFO:weightwatcher:skipping layer 0 None by id\n",
      "INFO:weightwatcher:skipping layer 1 None by id\n",
      "INFO:weightwatcher:skipping layer 2 None by id\n",
      "INFO:weightwatcher:skipping layer 3 None by id\n",
      "INFO:weightwatcher:skipping layer 4 None by id\n",
      "INFO:weightwatcher:skipping layer 5 None by id\n",
      "INFO:weightwatcher:skipping layer 6 None by id\n",
      "INFO:weightwatcher:skipping layer 7 None by id\n",
      "INFO:weightwatcher:skipping layer 8 None by id\n",
      "INFO:weightwatcher:skipping layer 9 None by id\n",
      "INFO:weightwatcher:skipping layer 10 None by id\n",
      "INFO:weightwatcher:skipping layer 11 None by id\n",
      "INFO:weightwatcher:skipping layer 12 None by id\n",
      "INFO:weightwatcher:skipping layer 13 None by id\n",
      "INFO:weightwatcher:keeping layer 14 None by id\n",
      "INFO:weightwatcher:Getting ESD for layer 14 \n",
      "INFO:weightwatcher:Found 2304 eiganvalues for 14 None\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': [], 'num_smooth': 200}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n",
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': [], 'num_smooth': 200}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "16\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:weightwatcher:Filtering layer by id 16\n",
      "INFO:weightwatcher:skipping layer 0 None by id\n",
      "INFO:weightwatcher:skipping layer 1 None by id\n",
      "INFO:weightwatcher:skipping layer 2 None by id\n",
      "INFO:weightwatcher:skipping layer 3 None by id\n",
      "INFO:weightwatcher:skipping layer 4 None by id\n",
      "INFO:weightwatcher:skipping layer 5 None by id\n",
      "INFO:weightwatcher:skipping layer 6 None by id\n",
      "INFO:weightwatcher:skipping layer 7 None by id\n",
      "INFO:weightwatcher:skipping layer 8 None by id\n",
      "INFO:weightwatcher:skipping layer 9 None by id\n",
      "INFO:weightwatcher:skipping layer 10 None by id\n",
      "INFO:weightwatcher:skipping layer 11 None by id\n",
      "INFO:weightwatcher:skipping layer 12 None by id\n",
      "INFO:weightwatcher:skipping layer 13 None by id\n",
      "INFO:weightwatcher:skipping layer 14 None by id\n",
      "INFO:weightwatcher:skipping layer 15 None by id\n",
      "INFO:weightwatcher:keeping layer 16 None by id\n",
      "INFO:weightwatcher:Getting ESD for layer 16 \n",
      "INFO:weightwatcher:Found 2304 eiganvalues for 16 None\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': [], 'num_smooth': 200}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n",
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': [], 'num_smooth': 200}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "19\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:weightwatcher:Filtering layer by id 19\n",
      "INFO:weightwatcher:skipping layer 0 None by id\n",
      "INFO:weightwatcher:skipping layer 1 None by id\n",
      "INFO:weightwatcher:skipping layer 2 None by id\n",
      "INFO:weightwatcher:skipping layer 3 None by id\n",
      "INFO:weightwatcher:skipping layer 4 None by id\n",
      "INFO:weightwatcher:skipping layer 5 None by id\n",
      "INFO:weightwatcher:skipping layer 6 None by id\n",
      "INFO:weightwatcher:skipping layer 7 None by id\n",
      "INFO:weightwatcher:skipping layer 8 None by id\n",
      "INFO:weightwatcher:skipping layer 9 None by id\n",
      "INFO:weightwatcher:skipping layer 10 None by id\n",
      "INFO:weightwatcher:skipping layer 11 None by id\n",
      "INFO:weightwatcher:skipping layer 12 None by id\n",
      "INFO:weightwatcher:skipping layer 13 None by id\n",
      "INFO:weightwatcher:skipping layer 14 None by id\n",
      "INFO:weightwatcher:skipping layer 15 None by id\n",
      "INFO:weightwatcher:skipping layer 16 None by id\n",
      "INFO:weightwatcher:skipping layer 17 None by id\n",
      "INFO:weightwatcher:skipping layer 18 None by id\n",
      "INFO:weightwatcher:keeping layer 19 None by id\n",
      "INFO:weightwatcher:Getting ESD for layer 19 \n",
      "INFO:weightwatcher:Found 2304 eiganvalues for 19 None\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': [], 'num_smooth': 200}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n",
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': [], 'num_smooth': 200}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "21\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:weightwatcher:Filtering layer by id 21\n",
      "INFO:weightwatcher:skipping layer 0 None by id\n",
      "INFO:weightwatcher:skipping layer 1 None by id\n",
      "INFO:weightwatcher:skipping layer 2 None by id\n",
      "INFO:weightwatcher:skipping layer 3 None by id\n",
      "INFO:weightwatcher:skipping layer 4 None by id\n",
      "INFO:weightwatcher:skipping layer 5 None by id\n",
      "INFO:weightwatcher:skipping layer 6 None by id\n",
      "INFO:weightwatcher:skipping layer 7 None by id\n",
      "INFO:weightwatcher:skipping layer 8 None by id\n",
      "INFO:weightwatcher:skipping layer 9 None by id\n",
      "INFO:weightwatcher:skipping layer 10 None by id\n",
      "INFO:weightwatcher:skipping layer 11 None by id\n",
      "INFO:weightwatcher:skipping layer 12 None by id\n",
      "INFO:weightwatcher:skipping layer 13 None by id\n",
      "INFO:weightwatcher:skipping layer 14 None by id\n",
      "INFO:weightwatcher:skipping layer 15 None by id\n",
      "INFO:weightwatcher:skipping layer 16 None by id\n",
      "INFO:weightwatcher:skipping layer 17 None by id\n",
      "INFO:weightwatcher:skipping layer 18 None by id\n",
      "INFO:weightwatcher:skipping layer 19 None by id\n",
      "INFO:weightwatcher:skipping layer 20 None by id\n",
      "INFO:weightwatcher:keeping layer 21 None by id\n",
      "INFO:weightwatcher:Getting ESD for layer 21 \n",
      "INFO:weightwatcher:Found 4608 eiganvalues for 21 None\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': [], 'num_smooth': 200}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n",
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': [], 'num_smooth': 200}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "23\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:weightwatcher:Filtering layer by id 23\n",
      "INFO:weightwatcher:skipping layer 0 None by id\n",
      "INFO:weightwatcher:skipping layer 1 None by id\n",
      "INFO:weightwatcher:skipping layer 2 None by id\n",
      "INFO:weightwatcher:skipping layer 3 None by id\n",
      "INFO:weightwatcher:skipping layer 4 None by id\n",
      "INFO:weightwatcher:skipping layer 5 None by id\n",
      "INFO:weightwatcher:skipping layer 6 None by id\n",
      "INFO:weightwatcher:skipping layer 7 None by id\n",
      "INFO:weightwatcher:skipping layer 8 None by id\n",
      "INFO:weightwatcher:skipping layer 9 None by id\n",
      "INFO:weightwatcher:skipping layer 10 None by id\n",
      "INFO:weightwatcher:skipping layer 11 None by id\n",
      "INFO:weightwatcher:skipping layer 12 None by id\n",
      "INFO:weightwatcher:skipping layer 13 None by id\n",
      "INFO:weightwatcher:skipping layer 14 None by id\n",
      "INFO:weightwatcher:skipping layer 15 None by id\n",
      "INFO:weightwatcher:skipping layer 16 None by id\n",
      "INFO:weightwatcher:skipping layer 17 None by id\n",
      "INFO:weightwatcher:skipping layer 18 None by id\n",
      "INFO:weightwatcher:skipping layer 19 None by id\n",
      "INFO:weightwatcher:skipping layer 20 None by id\n",
      "INFO:weightwatcher:skipping layer 21 None by id\n",
      "INFO:weightwatcher:skipping layer 22 None by id\n",
      "INFO:weightwatcher:keeping layer 23 None by id\n",
      "INFO:weightwatcher:Getting ESD for layer 23 \n",
      "INFO:weightwatcher:Found 4608 eiganvalues for 23 None\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': [], 'num_smooth': 200}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n",
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': [], 'num_smooth': 200}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "26\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:weightwatcher:Filtering layer by id 26\n",
      "INFO:weightwatcher:skipping layer 0 None by id\n",
      "INFO:weightwatcher:skipping layer 1 None by id\n",
      "INFO:weightwatcher:skipping layer 2 None by id\n",
      "INFO:weightwatcher:skipping layer 3 None by id\n",
      "INFO:weightwatcher:skipping layer 4 None by id\n",
      "INFO:weightwatcher:skipping layer 5 None by id\n",
      "INFO:weightwatcher:skipping layer 6 None by id\n",
      "INFO:weightwatcher:skipping layer 7 None by id\n",
      "INFO:weightwatcher:skipping layer 8 None by id\n",
      "INFO:weightwatcher:skipping layer 9 None by id\n",
      "INFO:weightwatcher:skipping layer 10 None by id\n",
      "INFO:weightwatcher:skipping layer 11 None by id\n",
      "INFO:weightwatcher:skipping layer 12 None by id\n",
      "INFO:weightwatcher:skipping layer 13 None by id\n",
      "INFO:weightwatcher:skipping layer 14 None by id\n",
      "INFO:weightwatcher:skipping layer 15 None by id\n",
      "INFO:weightwatcher:skipping layer 16 None by id\n",
      "INFO:weightwatcher:skipping layer 17 None by id\n",
      "INFO:weightwatcher:skipping layer 18 None by id\n",
      "INFO:weightwatcher:skipping layer 19 None by id\n",
      "INFO:weightwatcher:skipping layer 20 None by id\n",
      "INFO:weightwatcher:skipping layer 21 None by id\n",
      "INFO:weightwatcher:skipping layer 22 None by id\n",
      "INFO:weightwatcher:skipping layer 23 None by id\n",
      "INFO:weightwatcher:skipping layer 24 None by id\n",
      "INFO:weightwatcher:skipping layer 25 None by id\n",
      "INFO:weightwatcher:keeping layer 26 None by id\n",
      "INFO:weightwatcher:Getting ESD for layer 26 \n",
      "INFO:weightwatcher:Found 4608 eiganvalues for 26 None\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': [], 'num_smooth': 200}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n",
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': [], 'num_smooth': 200}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "28\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:weightwatcher:Filtering layer by id 28\n",
      "INFO:weightwatcher:skipping layer 0 None by id\n",
      "INFO:weightwatcher:skipping layer 1 None by id\n",
      "INFO:weightwatcher:skipping layer 2 None by id\n",
      "INFO:weightwatcher:skipping layer 3 None by id\n",
      "INFO:weightwatcher:skipping layer 4 None by id\n",
      "INFO:weightwatcher:skipping layer 5 None by id\n",
      "INFO:weightwatcher:skipping layer 6 None by id\n",
      "INFO:weightwatcher:skipping layer 7 None by id\n",
      "INFO:weightwatcher:skipping layer 8 None by id\n",
      "INFO:weightwatcher:skipping layer 9 None by id\n",
      "INFO:weightwatcher:skipping layer 10 None by id\n",
      "INFO:weightwatcher:skipping layer 11 None by id\n",
      "INFO:weightwatcher:skipping layer 12 None by id\n",
      "INFO:weightwatcher:skipping layer 13 None by id\n",
      "INFO:weightwatcher:skipping layer 14 None by id\n",
      "INFO:weightwatcher:skipping layer 15 None by id\n",
      "INFO:weightwatcher:skipping layer 16 None by id\n",
      "INFO:weightwatcher:skipping layer 17 None by id\n",
      "INFO:weightwatcher:skipping layer 18 None by id\n",
      "INFO:weightwatcher:skipping layer 19 None by id\n",
      "INFO:weightwatcher:skipping layer 20 None by id\n",
      "INFO:weightwatcher:skipping layer 21 None by id\n",
      "INFO:weightwatcher:skipping layer 22 None by id\n",
      "INFO:weightwatcher:skipping layer 23 None by id\n",
      "INFO:weightwatcher:skipping layer 24 None by id\n",
      "INFO:weightwatcher:skipping layer 25 None by id\n",
      "INFO:weightwatcher:skipping layer 26 None by id\n",
      "INFO:weightwatcher:skipping layer 27 None by id\n",
      "INFO:weightwatcher:keeping layer 28 None by id\n",
      "INFO:weightwatcher:Getting ESD for layer 28 \n",
      "INFO:weightwatcher:Found 4608 eiganvalues for 28 None\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXQAAAEaCAYAAAABnax5AAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjQuMiwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8rg+JYAAAACXBIWXMAAAsTAAALEwEAmpwYAAAev0lEQVR4nO3de5wU5Z3v8c8X5CKIRHEMxoGAWbwghktG0OVgZNGI0aMxwiKJOVFj0EROPBqJuDE6uuYco7smJ1lN1KxKTJRhiTGoeDccL6sJF4egIAZxXMYrkqCiIqK/80fVQDv2zDTQM91T/X2/XvWyq+rp6l8X43eeebr6KUUEZmbW+XUpdQFmZlYcDnQzs4xwoJuZZYQD3cwsIxzoZmYZ4UA3M8sIB7qZWUY40K1oJDVIOqLUdeSSdIykRyWtl/SKpF9K6pOzf3dJdZLWSXpd0m8k7VrAcQ+R9LakXfLse1LS9PRxd0kXSVqZtn9R0t2SvtDsOSdJ+mPa5rX08bclKd0/XtIfJL0hqaGFms6W9Hx6jBWS9t3G02WdnAPdMkPSTnk29wUuAz4FHADsDVyZs/8yYDdgMPAZ4JNAbVuvFRFPAI3ApGY1DAOGAremm+YCxwP/I+d1/i9wTM5zvptuuxLon9ZwJjAW6J42exu4AZiRrx5JpwPfSI+7C3As8Hpb78MyJiK8eCnKAjQAR+TZvhtwJ7AW+Fv6uDrdNxlY3Kz9ucDv08c9gH8B/gt4FfgFsHO673CSUD0feAW4uYAavwwsy1m/G/h2zvpZwL0Fvt9/Ah5qtu0K4Hfp4yOAd5veawvH6EsS1icW+JpHAA3NtnUB1gATSv0z4KW0i3vo1hG6ADcCnwYGkoTcv6X75gGDJR2Q0/5rwK/Sx5cD+wIjgL8j6WFflNO2P7B7euxpBdRyGPB0zvrVwLGSdpO0G3AiScgX4mbgMEkDACR1Ab4CzEr3HwH8MSIaWznGoSS/tH5f4GvmU50uwyStSYddLknrsQrif3BrdxGxLiJ+GxHvRMRbwA+Bz6f73gPqgJMBJB0IDALuTMePpwHnRMRf0+f+b+CknMN/CFwcEe9FxLut1SHpSODrfPQXwhKSYY116fIBcE2B72sNsIDkFxDABJJwvitd34PkL4em1989Hct/Q9LGnDavR8TmnHb/mbZ7V9JhBZRSnf73C8BBwHhgKskQjFUQB7q1O0m9JF0r6QVJbwIPA5+Q1DVtMgv4ShrgXwPmpEFfBfQCFqcBtx64J93eZG1EbKQNkg4BbgEmRcSzObvmAM8CfYBdgeeAX2/D25vF1kD/GjA7It5P19cBezU1TH8pfQL4HEnwN7XZI3f8PyL+Pm23jsL+H236RXZFRKyPiAbgWuCL2/A+LAMc6NYRvgvsB4yJiF1Jhj0ABFs+YNwEjCMZsrg53f86SVgdGBGfSJe+EZF7ZUmb04VKGkkytHNaRDzYbPcI4NqIeDsiNpCM0W9LEN4GVEsaTzI+Pytn34PAwZKq8z4z8TjwHskHp9trJcn5yz0Xnka1AjnQrdi6SeqZs+xE0vt9F1gvaXfg4jzP+xXJuPr7EfEoQER8CFwP/FjSngCS9pZ0VKHFpFed3AP8z4i4I0+ThcDpknaWtDPJEM+fc56/QFJtS8ePiLdJrmS5EXghIhbl7LsP+ANwu6Qx6SWM3YBDctqsBy4BrpE0SVIfSV0kjQB659TRRVJPoFuyqp6SuqfHeIdk2Op76fOr0/dxZ6HnyTKi1J/KesnOQnKVSzRbmi4ZXABsIBneOCPdt1POcweSjIdf0uyYPUnGzVcDbwIrgO+k+w4HGtuo6cb0uBtylqdz9g8G7iAZ3vgrSfgPydn/HHBkG69xePp+zs+zrzvJZZB/Ad4huSrnbuALzdp9FfhT2mYt8EeSUO7e7DVylwU5z98VmA28RXLFy0WASv0z4aVjF6U/DGYllfaOXwNGRcRfSl0PQNrTnRMRf1/qWswK4UC3siDpXODYiPiHUtdi1lnl+2adWYdKv8ou4EulrcSsc3MP3cwsI3yVi5lZRpRsyGWPPfaIQYMGlerlzcw6pcWLF78eEVX59pUs0AcNGsSiRYvabmhmZltIeqGlfR5yMTPLCAe6mVlGONDNzDKirK5Df//992lsbGTjxjYnz7MC9ezZk+rqarp161bqUsysnZVVoDc2NtKnTx8GDRpEeitF2wERwbp162hsbGTw4MGlLsfM2llZDbls3LiRfv36OcyLRBL9+vXzXzxmFaKsAh1wmBeZz6dZ5Si7QDczs+1TVmPozdUuqC3u8Q4v7vFa8swzz3DqqaeyZMkSfvjDH3Leeedt2bd+/XpOP/10nnrqKSRxww03cOihh3ZIXWaWbWUd6J3V7rvvzk9/+lNuv/32j+07++yzmThxInPnzmXTpk288847HV+gmbWpqUPZUR3BYnCg55g5cyYDBgzgrLPOAqC2tpZevXrR0NDAQw89xIABA+jWrRunnXYakyZNYv78+Zx77rn07t2bsWPHsnr1au6880723HNP9txzT+66666PHP+NN97g4Ycf5qabbgKge/fudO/evaPfpplllMfQc0yZMoU5c+ZsWZ8zZw577bUXDQ0NLF++nJtvvpnHH38cSK7IOeOMM7j77rtZvHgxa9eubfP4zz//PFVVVZx66qmMHDmS008/nbfffrvd3o+Z7bhiD/22Jwd6jpEjR/Laa6/x0ksvsXTpUnbbbTcWL17M5MmT6dKlC/3792f8+PFAMk6+zz77bLm+e+rUqW0ef/PmzSxZsoRvfetbPPnkk/Tu3ZvLL7+8Xd+TmVUOB3ozkydPZu7cudTV1TFlypSiHru6uprq6mrGjBkDwKRJk1iyZElRX8PMKpcDvZkpU6Ywe/Zs5s6dy+TJkxk7diy//e1v+fDDD3n11VdZsGABAPvttx+rV6+moaEBgLq6ujaP3b9/fwYMGMDKlSsBePDBBxk6dGh7vRUzqzBl/aFoKT5dPvDAA3nrrbfYe++92WuvvTjxxBO3BO+AAQMYNWoUffv2Zeedd+aaa65h4sSJ9O7dm4MPPnjLMV555RVqamp488036dKlCz/5yU9Yvnw5u+66Kz/72c/46le/yqZNm9hnn3248cYbO/w9mlk2leyeojU1NdH8BhcrVqzggAMOKEk9rdmwYQO77LIL69atY/To0Tz22GP0799/y/aI4KyzzmLIkCGcc845pS73Y8r1vJqVq5Y+CC2HSxglLY6Imnz7yrqHXi6OPfZY1q9fz6ZNm/jBD35A//79Abj++uuZNWsWmzZtYuTIkZxxxhklrtTMKpkDvQBN4+bNnXPOOWXZIzezyuQPRc3MMsKBbmaWEQ50M7OMcKCbmWVEeQe6VNyljDQ0NHDLLbdsWb/pppuYPn36dh9vwYIFHHvsscUozayidaa5W5or70DPsOaBbma2owoKdEkTJa2UtErSzDz7T5G0VlJ9upxe/FI7xttvv80xxxzD8OHDGTZsGHV1dQwaNIgLLriAESNGUFNTw5IlSzjqqKP4zGc+wy9+8QsguSHzjBkzGDZsGAcddNCWqQBa2j5z5kweeeQRRowYwY9//GMAXnrpJSZOnMiQIUP43ve+t6Wm++67j0MPPZRRo0YxefJkNmzYAMA999zD/vvvz6hRo7jttts68jSZWRlqM9AldQWuBo4GhgJTJeWbgKQuIkakyy+LXGeHueeee/jUpz7F0qVLeeqpp5g4cSIAAwcOpL6+nnHjxnHKKacwd+5cnnjiCS6++GIAbrvtNurr61m6dCkPPPAAM2bM4OWXX25x++WXX864ceOor6/fci17fX09dXV1LFu2jLq6OtasWcPrr7/OZZddxgMPPMCSJUuoqanhqquuYuPGjXzzm9/kjjvuYPHixbzyyislO2dmVh4K6aGPBlZFxOqI2ATMBo5v37JK56CDDuL+++/n/PPP55FHHqFv374AHHfccVv2jxkzhj59+lBVVUWPHj1Yv349jz76KFOnTqVr16588pOf5POf/zwLFy5scXs+EyZMoG/fvvTs2ZOhQ4fywgsv8MQTT7B8+XLGjh3LiBEjmDVrFi+88ALPPPMMgwcPZsiQIUji5JNP7rBzZFapyn18vZBviu4NrMlZbwTG5Gl3oqTDgGeBcyJiTfMGkqYB0yDp8ZajfffdlyVLljB//nwuvPBCJkyYAECPHj0A6NKly5bHTeubN28uymvnHrdr165s3ryZiODII4/k1ltv/Ujb+vr6orymmWVHsT4UvQMYFBGfBe4HZuVrFBHXRURNRNRUVVUV6aWL66WXXqJXr16cfPLJzJgxo+D5yseNG0ddXR0ffPABa9eu5eGHH2b06NEtbu/Tpw9vvfVWm8c95JBDeOyxx1i1ahWQjPE/++yz7L///jQ0NPDcc88BfCzwzazyFNJDfxEYkLNenW7bIiLW5az+Erhix0sDSjAT5LJly5gxYwZdunShW7du/PznP2fSpEltPu+EE07g8ccfZ/jw4UjiiiuuoH///i1u79evH127dmX48OGccsop7LbbbnmPW1VVxU033cTUqVN57733ALjsssvYd999ue666zjmmGPo1asX48aNK+gXRLvJvSy0RDN4mlW6NqfPlbQTyTDKBJIgXwh8JSKezmmzV0S8nD4+ATg/Ig5p7bidafrczq5DzqsD3TKirXHyUk+hu0PT50bEZknTgXuBrsANEfG0pEuBRRExD/iOpOOAzcBfgVOKVr2ZmRWkoOlzI2I+ML/ZtotyHl8AXFDc0szMbFuU3TdFS3UHpazy+TQrXCGXJZbzpYtlFeg9e/Zk3bp1DqEiiQjWrVtHz549S12KmXWAsrpjUXV1NY2Njaxdu7bUpWRGz549qa6uLnUZZtYByirQu3XrxuDBg0tdhplZp1RWQy5mZrb9HOhmZhnhQDczywgHuplZRjjQzcwo7+vLC1VWV7lYJ9PSfVo9r4tZSbiHbmaWEQ50M7OMcKCbmWWEA93MLCMc6GZmGeFANzPLCAe6mVlG+Dp0M6toWfhCURP30M3MMsKBbmaWEQ50M7OMcKCbmWWEA93MLCMc6GZmGeFANzPLCAe6mVlGONDNzDLCgW5mlhEFBbqkiZJWSlolaWYr7U6UFJJqileimVl5KdfpAtoMdEldgauBo4GhwFRJQ/O06wOcDfyx2EWamVnbCumhjwZWRcTqiNgEzAaOz9Pun4EfARuLWJ+ZmRWokEDfG1iTs96YbttC0ihgQETc1dqBJE2TtEjSorVr125zsWZm1rId/lBUUhfgKuC7bbWNiOsioiYiaqqqqnb0pc3MLEchgf4iMCBnvTrd1qQPMAxYIKkBOASY5w9GzazcleuHm9urkBtcLASGSBpMEuQnAV9p2hkRbwB7NK1LWgCcFxGLiluqlQWp1BWYlUTt+Eu2Pv7DxSWspGVt9tAjYjMwHbgXWAHMiYinJV0q6bj2LtDMzApT0C3oImI+ML/ZtotaaHv4jpdlZmbbyt8UNTPLCAe6mVlGONDNzDLCgW5mlhEOdDOrSFm7Bh0c6GZmmeFANzPLCAe6mVlGONDNzDLCgW5mlhEOdDOzjHCgm5llhAPdzCpOFq9BBwe6mVlmONDNzDLCgW5mth3KcdjGgW5mlhEOdDOzjHCgm5llhAPdzCwjHOhmZhnhQDczywgHuplZRjjQzayilOP148WyU6kLsE5AKnUFZlYA99DNzDLCgW5mlhEecjEz20a14y9JH10CESWtJZd76GZmGVFQoEuaKGmlpFWSZubZf6akZZLqJT0qaWjxSzUzs9a0GeiSugJXA0cDQ4GpeQL7log4KCJGAFcAVxW7UDMza10hPfTRwKqIWB0Rm4DZwPG5DSLizZzV3kD5DCqZmVWIQj4U3RtYk7PeCIxp3kjSWcC5QHfgH/IdSNI0YBrAwIEDt7VWMzNrRdE+FI2IqyPiM8D5wIUttLkuImoioqaqqqpYL21mZhQW6C8CA3LWq9NtLZkNfGkHajIzs+1QSKAvBIZIGiypO3ASMC+3gaQhOavHAH8pXolmZlaINsfQI2KzpOnAvUBX4IaIeFrSpcCiiJgHTJd0BPA+8Dfg6+1ZtJmZfVxB3xSNiPnA/GbbLsp5fHaR6zIzK7osz7QI/qaomVlmONDNzDLCgW5mlhEOdDOzjHCgm1lFyPoHouBANzPLDAe6mVlGONDNzDLCgW5mlhEOdDOzjHCgm5llhAPdzDKvEi5ZhAIn57IKJJW6AjPbRu6hm5llhAPdzCwjHOhmZhnhQDczywgHuplZRjjQzcwywoFuZpYRDnQzs4xwoJuZZYQD3cwsIxzoZmYZ4blcrH3lzgkTUbo6zCqAe+hmZhnhHrqZZValTJvbxD10M8ukSgtzKDDQJU2UtFLSKkkz8+w/V9JySX+W9KCkTxe/VGt30tbFzDqdNgNdUlfgauBoYCgwVdLQZs2eBGoi4rPAXOCKYhdqZmatK6SHPhpYFRGrI2ITMBs4PrdBRPwhIt5JV58AqotbppmZtaWQQN8bWJOz3phua8k3gLt3pCgzM9t2Rb3KRdLJQA3w+Rb2TwOmAQwcOLCYL21mVvEK6aG/CAzIWa9Ot32EpCOA7wPHRcR7+Q4UEddFRE1E1FRVVW1PvWZm1oJCAn0hMETSYEndgZOAebkNJI0EriUJ89eKX6aZmbWlzUCPiM3AdOBeYAUwJyKelnSppOPSZlcCuwD/Iale0rwWDmdmlinldL17QWPoETEfmN9s20U5j48ocl1mZtutnEK2I/mbomZmGeFANzPLCAe6mVlGONDNzDLCgW5mlhEOdDOzjHCgm5llhAPdzCwjfAu6SuYbWVgGVeqXisA9dDOzzHCgm5llhAPdzDKjkodbwIFuZpYZDnQzs4xwoJuZZYQD3cwsIxzoZpYJlf6BKDjQzcwyw4FuZpYRDnQzsx1QO/6SZBqNpqWEHOhmZhnhQDczywgHuplZRnj63ErjKXMtQ5ouVaw9vLakdZQL99DNrFPydecf50A3s07P4Z5woJuZZYQD3cwsIxzoZmYZUVCgS5ooaaWkVZJm5tl/mKQlkjZLmlT8Ms3MrC1tBrqkrsDVwNHAUGCqpKHNmv0XcApwS7ELNDOzwhRyHfpoYFVErAaQNBs4Hlje1CAiGtJ9H7ZDjWZmVoBChlz2BtbkrDem27aZpGmSFklatHbt2u05hJmZtaBDPxSNiOsioiYiaqqqqjrypc3MMq+QIZcXgQE569XpNuss/HV/s4pQSKAvBIZIGkwS5CcBX2nXqszMWuBvhbaszSGXiNgMTAfuBVYAcyLiaUmXSjoOQNLBkhqBycC1kp5uz6LNzOzjCpptMSLmA/Obbbso5/FCkqEYMzMrEX9T1MwsIxzoZmYZ4UA3s07BH4a2zYFuZpYRvgVdVvnac7OK4x66mVlGONDNrKzljp17HL11DnQzK3sO8sI40M3MiqiUv3z8oWiW+INQs4rmHrqZWUY40M2srHi8fPs50M3MMsKBbmaWEf5Q1MzKjoddto976GZmGeEeemfnSxXNLOUeupmVBQ+z7DgHupmVTPMQd6jvGAd6ZyRtXcysrNSOv6Rk/396DN3MOpx74u3Dgd5ZuDduGZEvzB3wxeEhFzOzjHCgm1lJuXdePA50M+sQDu7250AvZ76axTqxpgB3kHccB7qZtTuHesfwVS7lwD1wyygHecdyoJeKQ9w6mdoFtdQeXtvqfqCgNpUm9323dn52VEFDLpImSlopaZWkmXn295BUl+7/o6RBRa+0s8odB/eYuJVAoSH6kdApYPy7dkHtRxZrXe34S7Z+i7SdtNlDl9QVuBo4EmgEFkqaFxHLc5p9A/hbRPydpJOAHwFT2qPgTsGBnV/ueYkoXR22zfKFfSFtK176M1/bQS9XSA99NLAqIlZHxCZgNnB8szbHA7PSx3OBCVI7ptq29nRb6iW312LWjtqa0Kql8M3Xk26td+1g3tqrrh1/SalLKYiijZ6SpEnAxIg4PV3/GjAmIqbntHkqbdOYrj+Xtnm92bGmAdPS1f2AlcV6I+1kD+D1NltVDp+PrXwutvK52KojzsWnI6Iq344O/VA0Iq4DruvI19wRkhZFRE2p6ygXPh9b+Vxs5XOxVanPRSFDLi8CA3LWq9NtedtI2gnoC6wrRoFmZlaYQgJ9ITBE0mBJ3YGTgHnN2swDvp4+ngQ8FG2N5ZiZWVG1OeQSEZslTQfuBboCN0TE05IuBRZFxDzg34GbJa0C/koS+lnQaYaHOojPx1Y+F1v5XGxV0nPR5oeiZmbWOXguFzOzjHCgm5llhAM9D0lXSnpG0p8l/U7SJ3L2XZBOcbBS0lElLLNDSJos6WlJH0qqydk+SNK7kurT5RelrLOjtHQ+0n0V9bORS1KtpBdzfh6+WOqaOlpbU6R0BAd6fvcDwyLis8CzwAUAkoaSfOB7IDARuCadGiHLngK+DDycZ99zETEiXc7s4LpKJe/5qNCfjeZ+nPPzML/UxXSknClSjgaGAlPTn4kO5UDPIyLui4jN6eoTJNfeQzLFweyIeC8ingdWkUyNkFkRsSIiyv0bvR2mlfNRcT8b9hGFTJHS7hzobTsNuDt9vDewJmdfY7qtUg2W9KSk/ydpXKmLKTH/bMD0dJjyBkm7lbqYDlYW//4VOx+6pAeA/nl2fT8ifp+2+T6wGfhNR9bW0Qo5F3m8DAyMiHWSPgfcLunAiHiz3QrtINt5PjKvtfMC/Bz4ZyDS//4rSWfIOlDFBnpEHNHafkmnAMcCE3K+9VrINAidTlvnooXnvAe8lz5enE7Iti+wqMjldbjtOR9k9GcjV6HnRdL1wJ3tXE65KYt/fw+55CFpIvA94LiIeCdn1zzgpPSGHoOBIcCfSlFjqUmqavrQT9I+JOdidWmrKqmK/tmQtFfO6gkkHx5XkkKmSGl3FdtDb8O/AT2A+9Np3Z+IiDPTKQ/mAMtJhmLOiogPSlhnu5N0AvAzoAq4S1J9RBwFHAZcKul94EPgzIj4awlL7RAtnY9K/Nlo5gpJI0iGXBqAM0paTQdraYqUjq7DX/03M8sID7mYmWWEA93MLCMc6GZmGeFANzPLCAe6mVlGONDNzDLCgW5mlhEOdPsYSRMk/bodj3+QpBckfauVNhva4XVvkPSapI99i7GQuawlfZAz33d9UztJ/1nsWrdHe5wz61z8xSL7GEnnAl0i4l/a8TUOBa6KiENb2L8hInYp8mseBmwAfhURw3K2dyWZ9/5IklnyFgJTI2J5e9dUTOVen7U/99Atn+HAUkn7S3oo7Y0+IGkPAEkHSHo4nSp1hqRV2/Ear5HcDGKbSDpX0lPp8r9ytv8g7WE/KulWSec1f25EPAzkm55gh+ayzu0Zt1aHpJMl/Sk9n9dK6pre+WmFpOvTOyHdJ2lnSZdLOivnubVNx5J0u6TFaftpeeoZlPtXiKTzJNW2UkNvSXdJWpqe1ymFvncrLw50y2c48Gfgt8C5ETGC5C5O50jaiWQ64bPTOzrtw/ZNxHQ50EPSpwt9QjpN76nAGOAQ4JuSRko6GDgxrftooKblo+RV6FzWOzcbcvlI8LVWh6QDgCnA2PR8fgB8Nd09BLg6Ig4E1qfHqAP+Mefw/5huAzgtIj6XHv87kvoV8iZbqWEi8FJEDE//crmnkONZ+fHkXPYRkroBfYHDgUcjoj7dtRw4juT2a0sj4smc7a+lz92HZG7svhExSVJv4BpgE7AgIn6Ttjsa6A3cRdJLf6HA8v4b8LuIeDs9zm3AOJKOye8jYiOwUdId2/fu2/RuGoQtGdtKHROAzwEL0wnfdiY5bw8Dz+ec58XAoIj4taQ9JX2KZCKwv0VE0y+d76SThEEyZesQYF0B9bdUwy3Av0r6EXBnRDxSwLGsDLmHbs0dAKwguS/ispztB5GE92eB+pztw5rW0yGLb+Ts+zIwNyK+SfLLAEk9gR8B306PP4zS64i5rAXMyrnn5n4RUZvuey+n3Qds7Wj9BzCJpFddByDpcOAI4NCIGA48CfRs9lqb+ej/203789YQEc8Co0j+PS6TdNEOv1srCQe6NTccWEoSaENhS8/7a8CvSHqC+6bbRwAnp+3zqWbrUEbTVLIXknwo2cC2B/ojwJck9Up7/yek2x4D/ruknpJ2IbkxybYo1lzWrdXxIDBJ0p4AknYvYLipLq1lEkm4Q/LX098i4h1J+5MMPTX3KrCnpH6SeuTUkbeG9K+AdyLi18CVJOFunZCHXKy54SQBNw/4oqRlwLsk47brJN1MMg/4MmAB0BARLd3YopEk1OuBLpL2I7mSZGy6fxnwTy08t5ekxpz1qyLiKkk3sfXGEb9sGvqRNI9k3P/V9LhvND+gpFtJhpL2SI99cUT8+zbMZb2zpPqc9XsiYssljhGxsKU6ImK5pAuB+yR1Ad4HzgJeaeH9k86x3gd4MSJebnpN4ExJK4CVJDcxb/689yVdmp6nF4Fn2qihL3ClpA/TbS1eTmrlzZct2jaRtEtEbEgfzyAZL78wXe8H/JAktH8J/JTkZiEbScbj2+3erE11SepFMi49LSKWtNfrlXsdVpkc6LZNJP2AZBjgfZIhhnPT+4uWlKRbSIaIepKME/+fSq7DKpMD3cwsI/yhqJlZRjjQzcwywoFuZpYRDnQzs4xwoJuZZYQD3cwsIxzoZmYZ8f8BrAUh+50HCkcAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': [], 'num_smooth': 200}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n",
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': [], 'num_smooth': 200}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "30\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:weightwatcher:Filtering layer by id 30\n",
      "INFO:weightwatcher:skipping layer 0 None by id\n",
      "INFO:weightwatcher:skipping layer 1 None by id\n",
      "INFO:weightwatcher:skipping layer 2 None by id\n",
      "INFO:weightwatcher:skipping layer 3 None by id\n",
      "INFO:weightwatcher:skipping layer 4 None by id\n",
      "INFO:weightwatcher:skipping layer 5 None by id\n",
      "INFO:weightwatcher:skipping layer 6 None by id\n",
      "INFO:weightwatcher:skipping layer 7 None by id\n",
      "INFO:weightwatcher:skipping layer 8 None by id\n",
      "INFO:weightwatcher:skipping layer 9 None by id\n",
      "INFO:weightwatcher:skipping layer 10 None by id\n",
      "INFO:weightwatcher:skipping layer 11 None by id\n",
      "INFO:weightwatcher:skipping layer 12 None by id\n",
      "INFO:weightwatcher:skipping layer 13 None by id\n",
      "INFO:weightwatcher:skipping layer 14 None by id\n",
      "INFO:weightwatcher:skipping layer 15 None by id\n",
      "INFO:weightwatcher:skipping layer 16 None by id\n",
      "INFO:weightwatcher:skipping layer 17 None by id\n",
      "INFO:weightwatcher:skipping layer 18 None by id\n",
      "INFO:weightwatcher:skipping layer 19 None by id\n",
      "INFO:weightwatcher:skipping layer 20 None by id\n",
      "INFO:weightwatcher:skipping layer 21 None by id\n",
      "INFO:weightwatcher:skipping layer 22 None by id\n",
      "INFO:weightwatcher:skipping layer 23 None by id\n",
      "INFO:weightwatcher:skipping layer 24 None by id\n",
      "INFO:weightwatcher:skipping layer 25 None by id\n",
      "INFO:weightwatcher:skipping layer 26 None by id\n",
      "INFO:weightwatcher:skipping layer 27 None by id\n",
      "INFO:weightwatcher:skipping layer 28 None by id\n",
      "INFO:weightwatcher:skipping layer 29 None by id\n",
      "INFO:weightwatcher:keeping layer 30 None by id\n",
      "INFO:weightwatcher:Getting ESD for layer 30 \n",
      "INFO:weightwatcher:Found 4608 eiganvalues for 30 None\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': [], 'num_smooth': 200}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n",
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': [], 'num_smooth': 200}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "35\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:weightwatcher:Filtering layer by id 35\n",
      "INFO:weightwatcher:skipping layer 0 None by id\n",
      "INFO:weightwatcher:skipping layer 1 None by id\n",
      "INFO:weightwatcher:skipping layer 2 None by id\n",
      "INFO:weightwatcher:skipping layer 3 None by id\n",
      "INFO:weightwatcher:skipping layer 4 None by id\n",
      "INFO:weightwatcher:skipping layer 5 None by id\n",
      "INFO:weightwatcher:skipping layer 6 None by id\n",
      "INFO:weightwatcher:skipping layer 7 None by id\n",
      "INFO:weightwatcher:skipping layer 8 None by id\n",
      "INFO:weightwatcher:skipping layer 9 None by id\n",
      "INFO:weightwatcher:skipping layer 10 None by id\n",
      "INFO:weightwatcher:skipping layer 11 None by id\n",
      "INFO:weightwatcher:skipping layer 12 None by id\n",
      "INFO:weightwatcher:skipping layer 13 None by id\n",
      "INFO:weightwatcher:skipping layer 14 None by id\n",
      "INFO:weightwatcher:skipping layer 15 None by id\n",
      "INFO:weightwatcher:skipping layer 16 None by id\n",
      "INFO:weightwatcher:skipping layer 17 None by id\n",
      "INFO:weightwatcher:skipping layer 18 None by id\n",
      "INFO:weightwatcher:skipping layer 19 None by id\n",
      "INFO:weightwatcher:skipping layer 20 None by id\n",
      "INFO:weightwatcher:skipping layer 21 None by id\n",
      "INFO:weightwatcher:skipping layer 22 None by id\n",
      "INFO:weightwatcher:skipping layer 23 None by id\n",
      "INFO:weightwatcher:skipping layer 24 None by id\n",
      "INFO:weightwatcher:skipping layer 25 None by id\n",
      "INFO:weightwatcher:skipping layer 26 None by id\n",
      "INFO:weightwatcher:skipping layer 27 None by id\n",
      "INFO:weightwatcher:skipping layer 28 None by id\n",
      "INFO:weightwatcher:skipping layer 29 None by id\n",
      "INFO:weightwatcher:skipping layer 30 None by id\n",
      "INFO:weightwatcher:skipping layer 31 None by id\n",
      "INFO:weightwatcher:skipping layer 32 None by id\n",
      "INFO:weightwatcher:skipping layer 33 None by id\n",
      "INFO:weightwatcher:skipping layer 34 None by id\n",
      "INFO:weightwatcher:keeping layer 35 None by id\n",
      "INFO:weightwatcher:Getting ESD for layer 35 \n",
      "INFO:weightwatcher:Found 4096 eiganvalues for 35 None\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': [], 'num_smooth': 200}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n",
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': [], 'num_smooth': 200}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "38\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:weightwatcher:Filtering layer by id 38\n",
      "INFO:weightwatcher:skipping layer 0 None by id\n",
      "INFO:weightwatcher:skipping layer 1 None by id\n",
      "INFO:weightwatcher:skipping layer 2 None by id\n",
      "INFO:weightwatcher:skipping layer 3 None by id\n",
      "INFO:weightwatcher:skipping layer 4 None by id\n",
      "INFO:weightwatcher:skipping layer 5 None by id\n",
      "INFO:weightwatcher:skipping layer 6 None by id\n",
      "INFO:weightwatcher:skipping layer 7 None by id\n",
      "INFO:weightwatcher:skipping layer 8 None by id\n",
      "INFO:weightwatcher:skipping layer 9 None by id\n",
      "INFO:weightwatcher:skipping layer 10 None by id\n",
      "INFO:weightwatcher:skipping layer 11 None by id\n",
      "INFO:weightwatcher:skipping layer 12 None by id\n",
      "INFO:weightwatcher:skipping layer 13 None by id\n",
      "INFO:weightwatcher:skipping layer 14 None by id\n",
      "INFO:weightwatcher:skipping layer 15 None by id\n",
      "INFO:weightwatcher:skipping layer 16 None by id\n",
      "INFO:weightwatcher:skipping layer 17 None by id\n",
      "INFO:weightwatcher:skipping layer 18 None by id\n",
      "INFO:weightwatcher:skipping layer 19 None by id\n",
      "INFO:weightwatcher:skipping layer 20 None by id\n",
      "INFO:weightwatcher:skipping layer 21 None by id\n",
      "INFO:weightwatcher:skipping layer 22 None by id\n",
      "INFO:weightwatcher:skipping layer 23 None by id\n",
      "INFO:weightwatcher:skipping layer 24 None by id\n",
      "INFO:weightwatcher:skipping layer 25 None by id\n",
      "INFO:weightwatcher:skipping layer 26 None by id\n",
      "INFO:weightwatcher:skipping layer 27 None by id\n",
      "INFO:weightwatcher:skipping layer 28 None by id\n",
      "INFO:weightwatcher:skipping layer 29 None by id\n",
      "INFO:weightwatcher:skipping layer 30 None by id\n",
      "INFO:weightwatcher:skipping layer 31 None by id\n",
      "INFO:weightwatcher:skipping layer 32 None by id\n",
      "INFO:weightwatcher:skipping layer 33 None by id\n",
      "INFO:weightwatcher:skipping layer 34 None by id\n",
      "INFO:weightwatcher:skipping layer 35 None by id\n",
      "INFO:weightwatcher:skipping layer 36 None by id\n",
      "INFO:weightwatcher:skipping layer 37 None by id\n",
      "INFO:weightwatcher:keeping layer 38 None by id\n",
      "INFO:weightwatcher:Getting ESD for layer 38 \n",
      "INFO:weightwatcher:Found 4096 eiganvalues for 38 None\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': [], 'num_smooth': 200}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n",
      "INFO:weightwatcher:params {'glorot_fix': False, 'normalize': False, 'conv2d_norm': True, 'randomize': False, 'savedir': 'ww-img', 'savefig': 'ww-img', 'rescale': True, 'deltaEs': False, 'intra': False, 'channels': None, 'conv2d_fft': False, 'ww2x': False, 'min_evals': 0, 'max_evals': None, 'plot': False, 'mp_fit': False, 'layers': [], 'num_smooth': 200}\n",
      "INFO:weightwatcher:Saving all images to ww-img\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "41\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:weightwatcher:Filtering layer by id 41\n",
      "INFO:weightwatcher:skipping layer 0 None by id\n",
      "INFO:weightwatcher:skipping layer 1 None by id\n",
      "INFO:weightwatcher:skipping layer 2 None by id\n",
      "INFO:weightwatcher:skipping layer 3 None by id\n",
      "INFO:weightwatcher:skipping layer 4 None by id\n",
      "INFO:weightwatcher:skipping layer 5 None by id\n",
      "INFO:weightwatcher:skipping layer 6 None by id\n",
      "INFO:weightwatcher:skipping layer 7 None by id\n",
      "INFO:weightwatcher:skipping layer 8 None by id\n",
      "INFO:weightwatcher:skipping layer 9 None by id\n",
      "INFO:weightwatcher:skipping layer 10 None by id\n",
      "INFO:weightwatcher:skipping layer 11 None by id\n",
      "INFO:weightwatcher:skipping layer 12 None by id\n",
      "INFO:weightwatcher:skipping layer 13 None by id\n",
      "INFO:weightwatcher:skipping layer 14 None by id\n",
      "INFO:weightwatcher:skipping layer 15 None by id\n",
      "INFO:weightwatcher:skipping layer 16 None by id\n",
      "INFO:weightwatcher:skipping layer 17 None by id\n",
      "INFO:weightwatcher:skipping layer 18 None by id\n",
      "INFO:weightwatcher:skipping layer 19 None by id\n",
      "INFO:weightwatcher:skipping layer 20 None by id\n",
      "INFO:weightwatcher:skipping layer 21 None by id\n",
      "INFO:weightwatcher:skipping layer 22 None by id\n",
      "INFO:weightwatcher:skipping layer 23 None by id\n",
      "INFO:weightwatcher:skipping layer 24 None by id\n",
      "INFO:weightwatcher:skipping layer 25 None by id\n",
      "INFO:weightwatcher:skipping layer 26 None by id\n",
      "INFO:weightwatcher:skipping layer 27 None by id\n",
      "INFO:weightwatcher:skipping layer 28 None by id\n",
      "INFO:weightwatcher:skipping layer 29 None by id\n",
      "INFO:weightwatcher:skipping layer 30 None by id\n",
      "INFO:weightwatcher:skipping layer 31 None by id\n",
      "INFO:weightwatcher:skipping layer 32 None by id\n",
      "INFO:weightwatcher:skipping layer 33 None by id\n",
      "INFO:weightwatcher:skipping layer 34 None by id\n",
      "INFO:weightwatcher:skipping layer 35 None by id\n",
      "INFO:weightwatcher:skipping layer 36 None by id\n",
      "INFO:weightwatcher:skipping layer 37 None by id\n",
      "INFO:weightwatcher:skipping layer 38 None by id\n",
      "INFO:weightwatcher:skipping layer 39 None by id\n",
      "INFO:weightwatcher:skipping layer 40 None by id\n",
      "INFO:weightwatcher:keeping layer 41 None by id\n",
      "INFO:weightwatcher:Getting ESD for layer 41 \n",
      "INFO:weightwatcher:Found 1000 eiganvalues for 41 None\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "\n",
    "for iesd, layer_id in enumerate(details.layer_id.to_numpy()):\n",
    "    print(layer_id)\n",
    "    esd = all_esds[iesd]\n",
    "    smoothed_esd = watcher.get_ESD(model=smoother_vgg16, layer=layer_id) \n",
    "    plt.hist(np.log10(esd[esd>0.0]), bins=100, color='green', label='vgg16', alpha=0.5, density=True);\n",
    "    plt.hist(np.log10(smoothed_esd[smoothed_esd>0.0]), bins=100, color='red', label='smoothed', density=True);\n",
    "\n",
    "    plt.title(\"Layer {}, VGG16\".format(layer_id))\n",
    "    plt.xlabel(r\"$log_{10} \\lambda $ Log 10 Eigenvalues\")\n",
    "    plt.legend()\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.9.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}