6_resnext50_resize.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Kaggle Competition:  Iceberg or Ship"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# resnext50 architecture (rest is Grover's options)\n",
    "### see if log_loss=0.29 from resnet18 can be improved"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "total 2136\r\n",
      "-rw-rw-r-- 1 ubuntu ubuntu 625023 Jan 14 18:47 1_getdata.ipynb\r\n",
      "-rw-rw-r-- 1 ubuntu ubuntu 245619 Jan 14 18:36 2_model_resnet34.ipynb\r\n",
      "-rw-rw-r-- 1 ubuntu ubuntu  13445 Jan 14 18:36 3_model_resnext50.ipynb\r\n",
      "-rw-rw-r-- 1 ubuntu ubuntu  14220 Jan 13 21:14 4_model_nasnet.ipynb\r\n",
      "-rw-rw-r-- 1 ubuntu ubuntu  64965 Jan 14 19:40 4_resnext_resize.ipynb\r\n",
      "-rw-rw-r-- 1 ubuntu ubuntu 147077 Jan 14 19:42 5_resnet18_resize.ipynb\r\n",
      "-rw-rw-r-- 1 ubuntu ubuntu 147134 Jan 14 19:44 6_resnext50_resize.ipynb\r\n",
      "-rw-rw-r-- 1 ubuntu ubuntu 137540 Jan 14 18:36 baseline.ipynb\r\n",
      "-rw-rw-r-- 1 ubuntu ubuntu    536 Jan 14 18:36 helper.py\r\n",
      "-rw-rw-r-- 1 ubuntu ubuntu 767392 Jan 14 18:36 other_groverpr.ipynb\r\n",
      "drwxrwxr-x 2 ubuntu ubuntu   4096 Jan 14 18:51 __pycache__\r\n",
      "-rw-rw-r-- 1 ubuntu ubuntu    461 Jan 14 19:39 README.md\r\n"
     ]
    }
   ],
   "source": [
    "!ls -l"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Put these at the top of every notebook, to get automatic reloading and inline plotting\n",
    "%reload_ext autoreload\n",
    "%autoreload 2\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "from fastai.imports import *"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "# This file contains all the main external libs we'll use\n",
    "from fastai.imports import *\n",
    "from fastai.transforms import *\n",
    "from fastai.conv_learner import *\n",
    "from fastai.model import *\n",
    "from fastai.dataset import *\n",
    "from fastai.sgdr import *\n",
    "from fastai.plots import *\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import shutil\n",
    "from IPython.display import FileLink, FileLinks\n",
    "\n",
    "from sklearn.model_selection import train_test_split\n",
    "import helper"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "#PATH = \"/home/nidhin/data/iceberg/\"\n",
    "PATH = os.path.expanduser(\"~/data/iceberg/\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "total 1973260\r\n",
      "drwxrwxr-x  2 ubuntu ubuntu       4096 Jan  9 00:38 models\r\n",
      "-rw-rw-r--  1 ubuntu ubuntu     117951 Oct 23 17:27 sample_submission.csv\r\n",
      "-rw-rw-r--  1 ubuntu ubuntu      38566 Jan  7 20:36 sample_submission.csv.7z\r\n",
      "drwxrwxr-x  2 ubuntu ubuntu     274432 Jan 14 18:44 test\r\n",
      "-rw-rw-r--  1 ubuntu ubuntu 1521771850 Oct 23 17:27 test.json\r\n",
      "-rw-rw-r--  1 ubuntu ubuntu  257127394 Jan  7 20:36 test.json.7z\r\n",
      "drwxrwxr-x 45 ubuntu ubuntu       4096 Jan 14 18:51 tmp\r\n",
      "drwxrwxr-x  4 ubuntu ubuntu       4096 Jan 14 18:43 train\r\n",
      "-rw-rw-r--  1 ubuntu ubuntu  196313674 Oct 23 17:23 train.json\r\n",
      "-rw-rw-r--  1 ubuntu ubuntu   44932785 Jan  7 20:36 train.json.7z\r\n",
      "drwxrwxr-x  4 ubuntu ubuntu       4096 Jan 14 18:43 valid\r\n"
     ]
    }
   ],
   "source": [
    "!ls -l {PATH}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "np.random.seed(0)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Set Architecture"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "arch=resnext50\n",
    "sz=75\n",
    "bs=32\n",
    "lr = 1e-2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "68bcd4eb4e5b43fbb94cfe24afc1319c",
       "version_major": 2,
       "version_minor": 0
      },
      "text/html": [
       "<p>Failed to display Jupyter Widget of type <code>HBox</code>.</p>\n",
       "<p>\n",
       "  If you're reading this message in the Jupyter Notebook or JupyterLab Notebook, it may mean\n",
       "  that the widgets JavaScript is still loading. If this message persists, it\n",
       "  likely means that the widgets JavaScript library is either not installed or\n",
       "  not enabled. See the <a href=\"https://ipywidgets.readthedocs.io/en/stable/user_install.html\">Jupyter\n",
       "  Widgets Documentation</a> for setup instructions.\n",
       "</p>\n",
       "<p>\n",
       "  If you're reading this message in another frontend (for example, a static\n",
       "  rendering on GitHub or <a href=\"https://nbviewer.jupyter.org/\">NBViewer</a>),\n",
       "  it may mean that your frontend doesn't currently support widgets.\n",
       "</p>\n"
      ],
      "text/plain": [
       "HBox(children=(IntProgress(value=0, max=6), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "def get_data(sz, bs):\n",
    "    tfms = tfms_from_model(arch, sz, aug_tfms=transforms_top_down, max_zoom=1.00)\n",
    "    data = ImageClassifierData.from_paths(PATH, tfms=tfms, bs=bs,test_name='test')\n",
    "    return data\n",
    "\n",
    "data = get_data(sz, bs)\n",
    "data = data.resize(int(sz*1.3), 'tmp')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "100%|██████████| 41/41 [00:03<00:00, 12.67it/s]\n",
      "100%|██████████| 11/11 [00:00<00:00, 16.08it/s]\n",
      "100%|██████████| 264/264 [00:17<00:00, 15.28it/s]\n"
     ]
    }
   ],
   "source": [
    "#tfms = tfms_from_model(arch, sz, aug_tfms=transforms_side_on, max_zoom=1.1)\n",
    "#data = ImageClassifierData.from_paths(PATH, tfms=tfms, bs=bs,test_name='test',)\n",
    "\n",
    "learn = ConvLearner.pretrained(arch, data, precompute=True, ps=0.5)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Learning Rate Finder"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "dcadf58b80df4deba81d57297ce7c14f",
       "version_major": 2,
       "version_minor": 0
      },
      "text/html": [
       "<p>Failed to display Jupyter Widget of type <code>HBox</code>.</p>\n",
       "<p>\n",
       "  If you're reading this message in the Jupyter Notebook or JupyterLab Notebook, it may mean\n",
       "  that the widgets JavaScript is still loading. If this message persists, it\n",
       "  likely means that the widgets JavaScript library is either not installed or\n",
       "  not enabled. See the <a href=\"https://ipywidgets.readthedocs.io/en/stable/user_install.html\">Jupyter\n",
       "  Widgets Documentation</a> for setup instructions.\n",
       "</p>\n",
       "<p>\n",
       "  If you're reading this message in another frontend (for example, a static\n",
       "  rendering on GitHub or <a href=\"https://nbviewer.jupyter.org/\">NBViewer</a>),\n",
       "  it may mean that your frontend doesn't currently support widgets.\n",
       "</p>\n"
      ],
      "text/plain": [
       "HBox(children=(IntProgress(value=0, description='Epoch', max=1), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 61%|██████    | 25/41 [00:00<00:00, 57.98it/s, loss=3.54] \n",
      "                                                          \r"
     ]
    }
   ],
   "source": [
    "lrf = learn.lr_find()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fa400d3e4a8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.sched.plot_lr()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fa3c02201d0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.sched.plot() "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "24b4ce844f0f4aa7ab6c6bfb295ea024",
       "version_major": 2,
       "version_minor": 0
      },
      "text/html": [
       "<p>Failed to display Jupyter Widget of type <code>HBox</code>.</p>\n",
       "<p>\n",
       "  If you're reading this message in the Jupyter Notebook or JupyterLab Notebook, it may mean\n",
       "  that the widgets JavaScript is still loading. If this message persists, it\n",
       "  likely means that the widgets JavaScript library is either not installed or\n",
       "  not enabled. See the <a href=\"https://ipywidgets.readthedocs.io/en/stable/user_install.html\">Jupyter\n",
       "  Widgets Documentation</a> for setup instructions.\n",
       "</p>\n",
       "<p>\n",
       "  If you're reading this message in another frontend (for example, a static\n",
       "  rendering on GitHub or <a href=\"https://nbviewer.jupyter.org/\">NBViewer</a>),\n",
       "  it may mean that your frontend doesn't currently support widgets.\n",
       "</p>\n"
      ],
      "text/plain": [
       "HBox(children=(IntProgress(value=0, description='Epoch', max=8), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0.      0.63148 0.49687 0.71591]                          \n",
      "[1.      0.52316 0.46574 0.73579]                          \n",
      "[2.      0.50881 0.47736 0.8125 ]                          \n",
      "[3.      0.44346 0.52311 0.73864]                          \n",
      "[4.      0.40702 0.49164 0.73579]                          \n",
      "[5.      0.41032 0.41383 0.83523]                          \n",
      "[6.      0.39129 0.4498  0.84091]                          \n",
      "[7.      0.37984 0.48039 0.82671]                          \n",
      "\n"
     ]
    }
   ],
   "source": [
    "learn.fit(lr, 8)\n",
    "learn.precompute=False"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Stochastic Gradient Descent with Restart"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "37414a243798415ea55ea9611f0b8dea",
       "version_major": 2,
       "version_minor": 0
      },
      "text/html": [
       "<p>Failed to display Jupyter Widget of type <code>HBox</code>.</p>\n",
       "<p>\n",
       "  If you're reading this message in the Jupyter Notebook or JupyterLab Notebook, it may mean\n",
       "  that the widgets JavaScript is still loading. If this message persists, it\n",
       "  likely means that the widgets JavaScript library is either not installed or\n",
       "  not enabled. See the <a href=\"https://ipywidgets.readthedocs.io/en/stable/user_install.html\">Jupyter\n",
       "  Widgets Documentation</a> for setup instructions.\n",
       "</p>\n",
       "<p>\n",
       "  If you're reading this message in another frontend (for example, a static\n",
       "  rendering on GitHub or <a href=\"https://nbviewer.jupyter.org/\">NBViewer</a>),\n",
       "  it may mean that your frontend doesn't currently support widgets.\n",
       "</p>\n"
      ],
      "text/plain": [
       "HBox(children=(IntProgress(value=0, description='Epoch', max=7), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0.      0.52736 0.4316  0.80398]                          \n",
      "[1.      0.50568 0.40859 0.83807]                          \n",
      "[2.      0.47793 0.41686 0.80682]                          \n",
      "[3.      0.49381 0.4621  0.78125]                          \n",
      "[4.      0.4685  0.41905 0.81534]                          \n",
      "[5.      0.46717 0.40806 0.81818]                          \n",
      "[6.      0.44333 0.43367 0.80114]                          \n",
      "\n"
     ]
    }
   ],
   "source": [
    "#learn.fit(lr, 4, cycle_len=2)\n",
    "learn.fit(lr, 3, cycle_len=1, cycle_mult=2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fa3c01396d8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.sched.plot_lr()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYkAAAD8CAYAAACCRVh7AAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMS4xLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvAOZPmwAAIABJREFUeJzt3Xd42+W58PHvLVmS9x5xbCd2dpydmCSQMMsIK1DooVDaQhddHDre9i2ctnBK23PantPTySkvUMooZRRaCBBWgEBYIU7IThyPLMd2vPe2nvcP/azIjhQLx44s5/5cl65Iz2/oeaJEt54txhiUUkopf2yhzoBSSqmxS4OEUkqpgDRIKKWUCkiDhFJKqYA0SCillApIg4RSSqmANEgopZQKSIOEUkqpgDRIKKWUCigi1Bn4OFJTU01ubm6os6GUUmFl8+bNtcaYtOFcG1ZBIjc3l8LCwlBnQymlwoqIHBzutdrcpJRSKiANEkoppQLSIKGUUiogDRJKKaUC0iChlFIqIA0SSimlAtIgoZRSKqDTKkgcbe7k1V1Voc6GUkqFjdMqSDz2wUG+9tfN9PS5Q50VpZQKC6dVkGjs6MFtoLmjJ9RZUUqpsHBaBYmWzl4Amqwg4XabUGZHKaXGvNMsSHiCQ3NnLxWNHcy+82W2HW4Mca6UUmrsOs2CxLGaxIG6Nrp63ew40hTiXCml1NgVVJAQkVUiUiQiJSJye4BzrhOR3SKyS0T+5pN+k4gUW4+bfNKXiMgO656/FxE5+eKcmG+QaGr31CoqGjtG+22VUipsDblUuIjYgXuAi4ByYJOIrDHG7PY5ZzpwB7DCGNMgIulWejJwF1AAGGCzdW0D8CfgFuADYC2wCnhpJAs3WEuXJzA0dfTgsHliUmVT52i+pVJKhbVgahJLgRJjTJkxpht4Arhq0DlfAe6xvvwxxlRb6ZcArxlj6q1jrwGrRCQTiDfGvG+MMcAjwNUjUJ4T6q9JNHf00Gh1Xh/RmoRSSgUUTJDIAg77vC630nzNAGaIyLsi8oGIrBri2izr+YnuOaKMMQObmzq0uUkppYYSzM50/voKBo8djQCmA+cB2cAGEZl7gmuDuafnzUVuwdMsxaRJk4LIrn+dPW76rCGvzR092KzmpqPNnfS5DXbbqHeJKKVU2AmmJlEO5Pi8zgYq/JzznDGmxxizHyjCEzQCXVtuPT/RPQEwxtxnjCkwxhSkpQ1ri1bg2PBXGNhx3dNnqG3tGvZ9lVJqPAsmSGwCpotInog4geuBNYPOeRY4H0BEUvE0P5UBrwAXi0iSiCQBFwOvGGMqgRYRWW6Navo88NyIlCiAZqupCQY2N4E2OSmlVCBDBgljTC9wK54v/D3AU8aYXSJyt4istk57BagTkd3Am8D3jTF1xph64Kd4As0m4G4rDeDrwANACVDKaI9ssmoSETahqaOHxo5uJiZEAlDRqCOclFLKn2D6JDDGrMUzTNU37U6f5wb4rvUYfO2DwIN+0guBuR8zv8PW2uWpSWQmRtLU0YMITMuIo6Kpk5oWDRJKKeVPUEEi3P342Z08sekQAFmJUeyuaAYgOykKgLbuvpDlTSmlxrLTYlkOtzH09HlGNmUnRdPc2UtzZy9psS6cdpt3aKxSSqmBTosgsXxKivd5VmKU93lClIMYl522Lg0SSinlz2kXJBbmJHqfJ0Q5iI2M8PZXKKWUGui0CBJpcS7v83NmHJtrkRjtIMYZoc1NSikVwGkRJACumJ9JYrQDu024YFY6AJEOO3GREdrcpJRSAZwWo5sA/nDDIu/z33x6IQ+9e4ClecnEuiKobe0OYc6UUmrsOm1qEiJC/5YVCVEOvnXhdBx2GzEurUkopVQgp02QCCQuMoIWDRJKKeXXaR8kYpxak1BKqUBO+yARGxlBe3efdxlxpZRSx2iQcHn67tu6tTahlFKDaZCwgkSrzpVQSqnjaJCItIKE9ksopdRxTvsgEePSIKGUUoGc9kEiTpublFIqoNM+SPTXJHQYrFJKHe+0DxL9Hdc6oU4ppY532geJuEitSSilVCCnfZDob27S5cKVUup4p32QcNhtRDvtNHX0hDorSik15gQVJERklYgUiUiJiNzu5/jNIlIjIlutx5et9PN90raKSKeIXG0de0hE9vscWziyRQteSqyT+jZdLlwppQYbcj8JEbED9wAXAeXAJhFZY4zZPejUJ40xt/omGGPeBBZa90kGSoBXfU75vjHm6ZPI/4hIiXFR29oV6mwopdSYE0xNYilQYowpM8Z0A08AVw3jvT4FvGSMaR/GtaMqNdZJnW48pJRSxwkmSGQBh31el1tpg10rIttF5GkRyfFz/Hrg8UFpP7eu+Y2IuPxcc0okxzipa9OahFJKDRZMkBA/aYPX1X4eyDXGzAfWAQ8PuIFIJjAPeMUn+Q5gFnAGkAz8wO+bi9wiIoUiUlhTUxNEdj++lFgX9W3dGKPLhSullK9ggkQ54FszyAYqfE8wxtQZY/p/it8PLBl0j+uAfxpjenyuqTQeXcBf8DRrHccYc58xpsAYU5CWlhZEdj++lBgnPX2GZh0Gq5RSAwQTJDYB00UkT0SceJqN1vieYNUU+q0G9gy6xw0Mamrqv0Y8G09fDez8eFkfOamxnpauOu28VkqpAYYc3WSM6RWRW/E0FdmBB40xu0TkbqDQGLMGuE1EVgO9QD1wc//1IpKLpyby1qBbPyYiaXias7YCXzvp0gxTSqwTgLq2bqaMTmVFKaXC0pBBAsAYsxZYOyjtTp/nd+DpY/B37QH8dHQbYy74OBkdTckxVpDQmoRSSg1w2s+4hmPNTbU6DFYppQbQIAEkRXtqEjrrWimlBtIgATgjbCREObS5SSmlBtEgYUmJdVKrNQmllBpAg4QlJcapNQmllBpEg4QlJcal6zcppdQgGiQsKbFO6rS5SSmlBtAgYUmJddHQ3k2fW9dvUkqpfhokLKmxToyBhnatTSilVD8NEpZjs641SCilVD8NEpaUGF3kTymlBtMgYUm1FvnTuRJKKXWMBglLirV+U73WJJRSykuDhCUxyoFN0GGwSinlQ4OExWYTkmOcuhKsUkr50CDhIyXGRa02NymllJcGCR8ZCZEcbe4MdTaUUmrM0CDhIysxiiMNHaHOhlJKjRkaJHxkJ0VR19ZNe3dvqLOilFJjggYJH9lJUQBUNJ64NvHXDw6yvqj6VGRJKaVCSoOEj6xET5A4PEST0x/eKObBdw+cghwppVRoaZDwkZ0UDTBkv0Rjew8HattORZaUUiqkggoSIrJKRIpEpEREbvdz/GYRqRGRrdbjyz7H+nzS1/ik54nIRhEpFpEnRcQ5MkUavvQ4Fw67cOQEzU2dPX109bopb2inu9d9CnOnlFKn3pBBQkTswD3ApUA+cIOI5Ps59UljzELr8YBPeodP+mqf9F8CvzHGTAcagC8Nvxgjw2YTMhOiKD9BTaKxvQcAt4HDDe2nKmtKKRUSwdQklgIlxpgyY0w38ARw1cm8qYgIcAHwtJX0MHD1ydxzpGQnRXHkBF/+vvtN7K/RJiel1PgWTJDIAg77vC630ga7VkS2i8jTIpLjkx4pIoUi8oGI9AeCFKDRGNM/1jTQPRGRW6zrC2tqaoLI7slJj3NRc4JZ1/01CYADdRoklFLjWzBBQvykDd7j83kg1xgzH1iHp2bQb5IxpgD4DPBbEZka5D09icbcZ4wpMMYUpKWlBZHdk5MY7aSxrSfg8aYOn5qEdl4rpca5YIJEOeBbM8gGKnxPMMbUGWP6f37fDyzxOVZh/VkGrAcWAbVAoohEBLpnqCRFO2np6qWnz3+ndH9NYkJ8JGU+zU3GGJraAwcXpZQKR8EEiU3AdGs0khO4Hljje4KIZPq8XA3ssdKTRMRlPU8FVgC7jTEGeBP4lHXNTcBzJ1OQkZIU4wAGNiv5arDSz8hLZt/RFm/6uj3VLLj7Vd4vrRv9TCql1CkyZJCw+g1uBV7B8+X/lDFml4jcLSL9o5VuE5FdIrINuA242UqfDRRa6W8CvzDG7LaO/QD4roiU4Omj+PNIFepkJEZ7RuI2tvtfMryxoxtnhI2FOYnUtXVT0+KpQO080gTAL17ac2oyqpRSp0DE0KeAMWYtsHZQ2p0+z+8A7vBz3XvAvAD3LMMzcmpMSYr21CQaAtQkmtp7SIxyMHtCHABbDjUwIyOOlk5PH/y28iZ2HmliblbCqcmwUkqNoqCCxOkkyapJNASqSbT3kBjtYKYVJL7z5FYibMJZU1O95xxp7NAgoZQaF3RZjkESo/v7JLrZcqiBWx4ppLOnz3u8saObxCgnKbEu0uJctHf30dzZy0eHG8hJ9qz91NyhHdhKqfFBg8Qg/TWJA3XtXPO/7/Hq7qPsrTrWQd1fkwCYZdUmAI42dzEzIx6A5k5dalwpNT5okBgk2mnHabfx53f2e9MO+kya8w0S37loBnddeWyFkukZsQA0aU1CKTVOaJAYRERIjHbQ3ev2Nh8dqvMs09Hb56a+rZukGE9tY/GkJG4+K5cYpx3wLDUe54rQ5ial1LihQcKP/ian5XkppMe52F3ZzK1/28L6ohq6+9zMzDjWzCQiTEnz1CAmxEcSH+WguVODhFJqfNDRTX70NyfNz0nkQF0br+4+Sp/bsMOaCzFv0MilKWkx7DjSxIQEK0hoTUIpNU5oTcKP/prEguwEJiXH0Of2LCt1sK6daKfdW3PoNz3d83piYhQJURE0d2jHtVJqfNCahB/JsU6cdhuzJsQzOSV6wLH8zHjstoHrE35ueS6zM+NJjnESH+ngUL3uM6GUGh+0JuHHl1bm8cfPLMIZYfMGidULJgL4nSSXEO3gE7MzAIiPcujoJqXUuKE1CT+mpsUy1WpSOmtqKqvmTOCHl8+m1+1m9cKJJ7w2QfsklFLjiAaJIaTFubj3c56Vz//3xiVDnA3xkQ7auvvo7XMTYdeKmlIqvOm32AiLj/LEXZ11rZQaDzRIjLCEKM/wWW1yUkqNBxokRlh8pCdIaOe1Umo80CAxwuL7axLDnHXd0NbN0ebOkcySUkoNmwaJEdbf3DTcmsQtjxay7D9ep661a+iTlVJqlGmQGGETEiIBOFzfMazrtx32LP1x53O7RixPSik1XBokRlhClIOsxCj2VjUP6/qUWM+SIOuLqkcyW0opNSwaJEbBrAlx7K1sGfrEQYwx1LZ2YbcJbd19tHbpMFqlVGhpkBgFszLjKK1ppav32LannT19NLWfuJ+iqaOHnj7DnImeHe6qtQNbKRViQQUJEVklIkUiUiIit/s5frOI1IjIVuvxZSt9oYi8LyK7RGS7iHza55qHRGS/zzULR65YoTVrQjy9bkNp9bEd7W55dDML7n4VY0zA62qtzur8TE+QONqsnddKqdAaclkOEbED9wAXAeXAJhFZY4zZPejUJ40xtw5Kawc+b4wpFpGJwGYRecUY02gd/74x5umTLMOYMzvTsynRnspm8q1awdv7agAorm5lhs+mRb5qWroBvNdUt2hNQikVWsHUJJYCJcaYMmNMN/AEcFUwNzfG7DPGFFvPK4BqIG24mQ0XuSkxRNiEkppWb9o0a8+J/mDhT41Vk+hvbqpp0ZqEUiq0ggkSWcBhn9flVtpg11pNSk+LSM7ggyKyFHACpT7JP7eu+Y2IuPy9uYjcIiKFIlJYUxP4C3YsibDbyE6K8u6NDZBozZ946wRBotYKCnmpsUQ6bDqpTikVcsEECfGTNrhh/Xkg1xgzH1gHPDzgBiKZwKPAF4wxbiv5DmAWcAaQDPzA35sbY+4zxhQYYwrS0sKnEjI5JYYDdcf6JNq6PZ3YH+6vx+323y9R29pFhE1IjHKQHhdJtdYklFIhFkyQKAd8awbZQIXvCcaYOmNM/zfa/YB3TW0RiQdeBH5kjPnA55pK49EF/AVPs9a4kZsSzaG6dm9HdZs1nLWr1019e7ffa2pbu0iJdWKzCRnxLq1JKKVCLpggsQmYLiJ5IuIErgfW+J5g1RT6rQb2WOlO4J/AI8aYv/u7RkQEuBrYOdxCjEWTUmJo6eqlvs0TENq7e0m1JspVNfn/8q9t7SY11tPqpjUJpdRYMGSQMMb0ArcCr+D58n/KGLNLRO4WkdXWabdZw1y3AbcBN1vp1wHnADf7Ger6mIjsAHYAqcDPRqxUY0Cute3pAatfoq2rjympns7rQDWEmpauY0Ei3kWNDoFVSoVYUDvTGWPWAmsHpd3p8/wOPH0Mg6/7K/DXAPe84GPlNMxMTokB4FB9GwtzEuno6WNKWgwfHqinKkCQqGru9I5smpwcTUtXL+UN7WQnRZ+yfCullC+dcT1KcpKjEIE9lS20d3v6IyanxGATOOqnuam7101ta5d3gcAzp6YC8H5p3anLtFJKDaJBYpS4IuxcMDOdBzaU8exHRwDP1qapsS6/NYmjzZ0YA5lWkJiREUtKjFODhFIqpDRIjKI/fGYRWUlRPPjuAQBinBFMSIjkzaIaLvvdBu8yHIA3cGQmRAEgIpw5NYX3SutOuJTHx/X2vhpW//Edb+0mkEN17d7gppQ6fWmQGEXRzgimpMZS3tBuvbaTER9JTUsXuyubKTxQ7z23sqk/SER605ZNSaGquZPyhuHtTeHPW/tq2F7exBt7T7wU+QW/Xs+3n9xKb5/7uGO/eGkvL+2oHLE8KaXGLg0Soyw11kVPn6cmEOuKICXG6T22u+LYnhOVjZ5AMMEnSMya4FnjyXd5j+EwxvDKripe33OU/bWeCX5rT/Al73Ybeq0Jf42Ddtjr6O7jvrdLufftspPKk1IqPGiQGGX9cyMAol0RtHQea+bZXdlCW1cvxhgqmzqJc0UQF+nwHp+a5hkyW1ZzbOb2cNz9wm6++uhmvv3kVkqqPQHnjb3VtHX14nYb/vhGMaU+gWhnRZP3eUPbwIl/uyubcBvYXt6oW6wqdRoIagisGr4UnyAR47TznYtmkBbnoqaliw3FNSz+6WtkJUbR2NEzoBYBkBzjJCnaMeALfDjWF9VgE2jp7KWls5cFOYlsO9zI1sONHK5v579f3UdxdSu/u36R9/x+DYP2wNhe7gkgxsCG4lquXuRvGS+l1HihNYlR1j85Djw1iWnpsfz76jnMy06gubMXu03odRvq27pxOY7/OKakxVJaPfwg0dvn5nB9O5fMmeBN++TCiQC8V1rLL1/eC4BNji3R9dGhBu/z+kE1iR3lTaTFuUiJcZ5wsUKl1PigQWKUpfgEiRin3fu8f9LcF1fkce9nPUtdZcQNrEkATE2LofQkmpsqGjvpdRvOm5lGUrSnKWvhpCTyUmN46N0D3ppCZdOxzvF9R1spmJwEQKPPOlPGGLaWN7IgO4FFk5LYcaQJpdT4pkFilA3ok3Aea907a2oq/3nNPL55/jTyJ8bzwr+u5D+vnXfc9VPTYqlt7WLLoYZhDYXtX4k2LzWWxZM8X/xT0mJYkJ1AW3cfaXEuLp+f6R1d1drVy5HGDpZPSQGONTd19vRx3f97n7KaNs6amsrMCbHsr20bsEWrUmr80SAxyvqbm5x2G86IY3/ddptww9JJRFm1i7lZCaT7qUnMsrYyveZ/3xvQV9DV2xfUF/RBK0jkpkTzLwU5XL1wIvGRDuZnJwJwyZwMshOjqGzqxBhD8dEWAOZnJ+CKsNFg1STeLall04EG7rh0FjedlcvMCfH0DdqiVSk1/miQGGXJ1pDXaJd9iDP9O3taKk/cshybwNbDjd70Wx7ZzFcf3Tzk9Qfq2oly2EmLc7Fq7gR+a3VOnzk1hQibcPXCLDITIunudVPX1s0+K0jMnBBHcozTO7pp3Z6jxLoiuHlFLnabeIfnFh1t9v/GSqlxQUc3jTKH3UZitIMY5/D+qm02YfmUFHJTY9hT6flC7ulz80FZHV29bnZVNJGTHE28z9BZXwfr2picEo3IwL2jZmfGs/3fLybaGUGdFQgqGzspqmol0mEjJymaxGgnDe3duN2G1/dUc+6MNFwRnmCXlxqDwy4UVZ3cyCul1NimNYlTIDXWRbRzeDWJfrMnxLO3yvMrv6iqha5ez0zoq+95l5W/eMMbQAY7WNfO5BT/q8j295FMtJYCqWjqoLi6hRkZcdhsQlK0g4b2HjYdqKe6pYsL89O91zrsNqamxVJUpTWJwUZyGRWlQk2DxCmQEe8iPsr/L/1gzc6M41B9O61dvWwr9zQ7XTE/k9mZ8UQ7I/jCXzbR3Xv8EhpVzZ3e9aACyUz09IVUNnZQVNXC9HRPU1KS1dx0/4YykqIdrJqTOeC6aemxJz0bfLwpqmrhvP9erwszqnFDg8QpcOcVc/jJ6jkndY9ZEzwd2EVVzWw73EhStIM/3LCINbeu5EdXzKaquZNdFQOHpHb29NHS2UtanMvfLb1SYpw4I2zsrmymuqWLmRM8M72Toh2U1baxbk81N52V6+1k75ebEsORhg6/wel08E5xLa/sqhqQ9vvXizlY184d/9hOZ4+O/FLhT4PEKTBzQhxzsxJO6h6zrXkVHx1qZPPBBhbkJHr7GZbmJgNQeKCBzQcbcFvrLtVY258OFSREhNkT4nhxu2c9pxkZnppEglX7iY+M4KYzc4+7bnJKNG4DRxpHbgHCcHLXmp18/a+bedUKFKU1razdWcmKaSkcqGvn6c3lIc6hUidPg0SYyEqMYm5WPH9aX0ppTRufmJ3hPZYeH0lOchS/e72Ya//0Hg+/fwCAmtbgggTAyumptHV7fvn2B4lIq5P69ktnk+SzMGG/3FTP7nv9w2xPJ9XNnZTWtOGw2/j2k1spq2nluY+OIMBvP72IySnRvL7naKizqdRJ0yARRj5dkENdWzdxrgiuGbRmUsHkZFq7PIsH/ua1fWwsq6Pa2qMiLTaIIDEtDYA4V4R3ufLPn5XLvZ9dwg1Lc/xe098hftDax/t08n6Zp8/h9zcswmG3cdsTH/HijkqW5iWTFufi/JnpvFdaR0e3Njmp8KZBIoysXphFnCuCT5+RQ4xr4JDa5VM8TU53XZmP28Cn7/uA364rBiA9iJrE4smJRDnszJgQ523GSohysGruhOOGz/ZLs0ZtnZZBorSOuMgILpydwa8+NZ+dR5oprWljlbVG1gWz0unqdfN+WW2Ic6rUydF5EmEkIcrB6987l8So45t+rlmczcwJ8SzMSeTaJdl84tdvsbeqBZsMXD8qEFeEne9fMpP0+KHP7SciTEqOPu2am9q7e1m3p5ozp6RgtwmXzJnA58+czOMfHuKSuZ4gsWxKMnGuCF7YXskFszKGuKNSY1dQNQkRWSUiRSJSIiK3+zl+s4jUiMhW6/Fln2M3iUix9bjJJ32JiOyw7vl7CfRzVQ2QHhc5YHmPfg67jYU5nqU24iMdzLM6ypNjXNhtwf3VfnFlHlfMn/ix8pObEuNdH+p08ecN+6lt7eKr507xpv37lXNY//3zvcONXRF2rlgwkZd2VNHS2RPoVkqNeUMGCRGxA/cAlwL5wA0iku/n1CeNMQutxwPWtcnAXcAyYClwl4gkWef/CbgFmG49Vp1sYdQxc63RUMF0Wp+MGRmxHKhrp63rxHtmjxe9fW7u31DGRfkZLJmc7E232YSsxIHzUa4ryKajp887akypcBRMTWIpUGKMKTPGdANPAFcFef9LgNeMMfXGmAbgNWCViGQC8caY941neuojwNXDyL8KIH+ipyYx2kFi8eQk+tyGrYcbKalu4az/fJ1fv1rkd2/s8WDHkSaaO3u5auHQNa6FOYlMTYvhua0VpyBnSo2OYIJEFnDY53W5lTbYtSKyXUSeFpH+4TCBrs2yng91TzVMc7M8NYlgOq1PxuLJSYh45mi8X1pHRVMnf3ijhN+/UTKq7xsq71kzqfuXUj8REeHy+RPZuL/OO2dFqXATTJDw16A9eHGa54FcY8x8YB3w8BDXBnNPzw1EbhGRQhEprKnRndCClZUYRX5mPAusforREh/pYGZGHIUH6ympbiXWFcHqBRO5d30pK37xBk9tOjz0TcLI+6V1zMyIG7Dj4IlcNm8CbgMvD5qZrVS4CCZIlAO+A+WzgQH1Z2NMnTGm/6fS/cCSIa4tt54HvKfPve8zxhQYYwrS0tKCyK4Cz6/Ytd86m88tnzzq71WQm8RHhxrZU9nCtPRYfnxFPrmp0TS0d7N25/hpj+/udVN4sJ4zpw5di+g3MyOOaemx/HlDGc3aga3CUDBBYhMwXUTyRMQJXA+s8T3B6mPotxrYYz1/BbhYRJKsDuuLgVeMMZVAi4gst0Y1fR547iTLokJk5bRUWrt6+fBAPdPTY0mLc/Hqd87lqoVZbPFZJiTclVS30tnjZtGk4GtnIsJ/fHIehxs6uOu5XYBnD/Ef/nMHTe0aNNTYN+Q8CWNMr4jciucL3w48aIzZJSJ3A4XGmDXAbSKyGugF6oGbrWvrReSneAINwN3GmHrr+deBh4Ao4CXrocLQyulpOO02uvvcTEuP9aYvmZzE4x8eYl91C1NSY/0O3Q0nu63l2Pv3Jw/W0rxkPrd8Mo9tPMi1i7P56qOFtHX3sfVwI+fNTOOmM3NJjz9+V0KlxoKg/tcaY9YaY2YYY6YaY35upd1pBQiMMXcYY+YYYxYYY843xuz1ufZBY8w06/EXn/RCY8xc6563Gl2EP2zFuiJYZs34np4xMEgArPrtBr771NaQ5G0k7alsJtJhIy81duiTB1m9cCI9fYavPFJIXKSDn39yLlVNndz7VhnX/Ok9Dtcfm7W+ZlsF971dOpJZV2rYwvunnRozLp2baW1reuxXdm5KNBnWDO4XtleGfbPT7opmZmbEBT050dfC7EQyEyLp6OnjOxdN58Zlk9n844v45zfOoqm9h3/75w7vZkUPbCjjv1/dR3v36TH3RI1tGiTUiLj+jBzWffdcJvpMKBMRHv/Kcn6wahZAWG9QZIxhd2Uz+R+zqamfzSbcuGySZ9mUxcfGbMzPTuT/XDyDDcW1/OXdA3T19rGnspnuXjfvFOu6Tyr0NEioEWGzCXnW0uG+pqTFcskcz9pFWw42nOpsjZgdR5po6ughP3N4QQLg1gum8+w3VxBhH/jf7rPLJ3P29FTufmE333xsCz19nhrF/7y2j688UjigKUqJ28wZAAAcxUlEQVSpU02DhBp1eakxJEU72HIoPINER3cf335yK+lxLi7/mGtbBSPCbuPhLyzl4vwM1u2pBjyztfdWtbC+qJrVf3yH2ladjKdCQ4OEGnUiwqJJSRRaNYm2rl7WbKsgXMYqvFlUTVlNG7+4dh7JfjZfGgk2m/CVczwLBiZGO7jvc0v425eX8c9vrKChvYdnPzoyKu+r1FA0SKhTYvmUZMpq2qhq6uS/Xy3itsc/YldFc6izFZQP99cT5bBz9vTRncxZMDmJeVkJLM1NJj0+krOmpTI3K4EF2Qm6FaoKGQ0S6pRYMS0VgGe2lPPYB4eAY/MOQm1HeROdPYF3kNt0oJ5FkxJx2Ef3v4uI8LevLOO31y8ckH7tkmz2VrWwq6JpVN9fKX80SKhTYvaEeJJjnPz61SIQcEbY2DMGgsQ7xbVc+cd3uPet4+clbC9v5Ht/38aeymYKcpP9XD3y4iIdRDsHznG9cv5EHHbhmc3a5KROPQ0S6pSw2YQzp6bgNvDzq+cyOzM+5EGis6eP2/+xHYDntg7sI3G7DXf8YwdPby7HbWDpKQoS/iTFOLlwdgbPbT1Czzhdgl2NXRok1Cnz3Ytm8LvrF/IvBTnkZ8axp7IlpJ3XG4prKW/o4NK5E9hf28Ylv32bP77h2Rf8hR2V7Kpo5tbzp/GtT0z3zigPlWsXZ1PX1s36Il0JWZ1aGiTUKTM1LZarFnq2DZmdGU9TRw+VTZ1BX9/dO7K/ot/eV0O0087dV83FFWFj39FWHnn/IK1dvfzn2j3MzoznOxfN4DsXzRj1/oihnDszjZQYJ89oB7Y6xTRIqJDoXyRv55GBnbGFB+r9zjSub+tmwU9e5eWdI7cvw9vFNSyfkkJanIvn/3Uld12ZT3VLF7c8UkhlUyc/u3rOsJbgGA0Ou42rFmbx+t6jNLR1+z2nq7ePN/YeDZuhxSo8aJBQITFnYgIRNmHr4UZvWmdPH1/762Y+++eNPPrBwQHnl9a00tHTN2LzBQ7WtXGwrp1zpntGXc3IiOOaxdlE2IT3Suu46czJA/awHguuXZJFT58JuEfHcx9V8MWHCnll19FTnDM1nmmQUCER6bAzOzOezQcb+OMbxRRVtfDMlnJqW7uZmhbDL1/aS0+fm+pmT3PUkYYOwPPr/x9bytl3tOWk3v91a2bzeTPTvWkJUQ6uWZzFRfkZ/OiK/JO6/2jIz4wnMyGS960tVAf7yAq4f1pf4q1NbD5Yz9Hm4Jv0lBpsyP0klBotiyYl8sj7B9m4v56/by6nvbuP+dkJfGllHt96Yiu3P7ODF3dUsPGOCznS6AkS7d19fPepbaTGOnn2myvITooe1nu/uruKGRmx5A5ab+pXn1pw0uUaLSLC0rxk3i+twxiDZ7+uY7aXN+KMsLGtvInCgw3MyIjjhvs3cuaUFB7+4tIQ5VqFO61JqJDp3+EtNdbFwbp23G7Df31qAWdYw02f2VJOZ4+bfdUtHGnsID4yggU5iXx2+SS6etz859q9J7q9X5sP1vONxzaz6UADF+dPGNHynApL85KpbuniYN3ARf86e/ooqmrhhjNycEbYeHlnFa/sqqK7181b+2ooqjq5mpc6fWlNQoXMsrwUnBE27royn7jICKakxjIpxVMzyEmO4nC9p/ZQfLSVisYOJqVE89w3VwDQ1NE7rFVln9x0mLU7PJ3fl8wJwyBhBdAP99cPqAXtqWym1204c2oKh+rbeXV3FfuOtpCZEEljew8PbCjjv/5l7NaS1NilQUKFzMTEKLbfdTGRDvtxx5bmpnC4vhyn3ca+oy0caehgStqxL8X8zHie31ZBY3s3idHBL7q3u7KZuVnx/Ntls5mXnTAi5TiVpqXHkhHv4uVdVVx3Ro43/cP9nl2B52cn0tDew5v/2MHh+g5uu2AajR09PP7hIVJiXTR39vAfn5w35PvsPNJEaU2rd8iyOn1pc5MKKX8BAuBfL5jG729YxKzMOIqt5ibfDY36h9B+nPWfuno9TTIrpqVy1tTUk8t4iIgI1xXk8GZRNT/85w4e2FBGb5+bRz84SMHkJCYmRnFxfgYzM+L46jlT+Mb50/jSyjx63YZ73yrliQ8PBbXj3Y+e3cn3n95Or87wPu1pkFBjUm5qDKsXTGR6ehyFBxpo7+4jyydI9O8Qt/tjrCRbfLSVnj7DvKzwq0H4un7pJAR4bOMhfv3qPp7eXE55Qwe3WEuNp8S6eOU753DHZbOJdNiZnBLD55dPZnp6LG7jWdDQlzGG/3ltH+t2H+VwfTvvFNey9XAj3b1uymrbjjv3a49u5qUd/ofhqvFHm5vUmDY9I5Yua6Z1dtKxIJEa6yIj3nVckOjpc3Pf22V8+owcUmNdA47tsCbuzZ0Y3kEiKzGKH16eT3lDO3959wA/fm4nMzJiuXB2RsBrfnLVXOpau1jys3VsPdzIsikp3mNbDjXy+9eL/V63u6KZGRlx3tf1bd28vKuK+rZuLp2XOXKFUmNWUDUJEVklIkUiUiIit5/gvE+JiBGRAuv1jSKy1efhFpGF1rH11j37j6UHuq86fZ01NYWUGCdXzM/kzEFNRHMnJrC1vHFA2paDDfzXK0Ws/OUbdHQPXP57Q3ENCVEOJqcMb9jsWPKllXn8+PJ8shKj6Okz3HXlHGxDzA5PiXWRkxzF5oMN1PnsdPfMlnIiHTa+f8lMbr90Fl89Zwo/unw2zgjbcc15+62aReHB+oC75XX3uvnZC7t5cXslbrfO/g53Q9YkRMQO3ANcBJQDm0RkjTFm96Dz4oDbgI39acaYx4DHrOPzgOeMMVt9LrvRGFN40qVQ49b87EQ2//giv8eWTUnm9b3VFB9toamjh4LcZMqtSXedPW4e/eAAt5wzFYDioy28tLOKr5879bj5BeHKZhNuv3QWJdWt3v06hrIwJ4nnt1Xwxt5qHvrCUqZnxPL8tgoum5vJN8+fNuDcNdsqjqup9Tc/uQ28safa23ne0d3H/to28ifGs628kQfe2Q/s5z8+OY/PLJt08oVVIRNMTWIpUGKMKTPGdANPAFf5Oe+nwK+AQNM7bwAeH1YulfKjv/P5hvs/4NP3fUBTe4930t2MjNgB6zz97/pSoh12vnz2lJDkdbRcuWAi37loRtDn37A0h8vmTSA3NYavPFLIpb/bgNtt+OLKvOPOzc+MZ8uhBn7+4m7vpkxlNW047EJanIv3So+tsfX7N4q5/A8b2HmkiZLqVgBs4tn69eH3DvD6Hl0qJFwF0yeRBRz2eV0OLPM9QUQWATnGmBdE5HsB7vNpjg8ufxGRPuAZ4GdGVyZTH8PszHgSohzUtnoWvHu/rI4jDR2kxbm4cv5Efv3aPqqbO3FF2HlxRyXXn5EzantUh4uzpnpGdh1p7OA3r+2juaOH2z4xnbl+OvOvWpjFjiNNPPDOfg7WtfPLa+ezv7aVScnRpMa6vLW2Prfhn1uOYAz8/MU95E+MJ9Jh4/J5E3l1dxXri6rJz4znEyfoM1FjVzBBwl/d3PtlLiI24DfAzQFvILIMaDfG7PRJvtEYc8RqpnoG+BzwiJ9rbwFuAZg0Saut6hi7TThzSgpvFFVjF+HdklrKG9vJSozikrkT+PVr+7jzuV1MSIiku9fNdQU5Q9/0NJGVGMV/DzG57sypKbx429n85d39/OT53awvep3uPjcXzs4gPiqCD6w1pD4oq6OquZOzpqbwXmkd+2vbmJIay7IpyTyzxbO0+Y4jTTR39hAf6Rj1sqmRFUxzUzng+78rG6jweR0HzAXWi8gBYDmwpr/z2nI9g5qajDFHrD9bgL/hadY6jjHmPmNMgTGmIC1tdDeiV+Hnh5fP5m9fXsbyKcm8W1LLkYYOspKimJ4eyxdX5LF+XzUPvXeAORPj/f5aVkP7woo8XrxtJVPTYwFIi3OSnRhFVXMnPX1uXt5ZRbTTzq+vW4BNoKq5k6npsSzL88wOd0XYcBvYtL+etq5e/rS+lO88uZW39ukGSuEgmCCxCZguInki4sTzhb+m/6AxpskYk2qMyTXG5AIfAKv7O6Stmsa/4OnLwEqLEJFU67kDuALwrWUoFZSc5GgKcpNZOT2Nsto2DtS1k50YhYhw55X5bP7RRTz25WXc+9kloc5qWJszMYH/vXEx6XEuLsrPICspCreBqqZOtpc3Mj87gcyEKO9e4NPSYpmUHM3KaancceksnBE23i+t41cv7+WXL+9l3Z6j3PTgh8fN2VBjz5BBwhjTC9wKvALsAZ4yxuwSkbtFZHUQ73EOUG6MKfNJcwGviMh2YCtwBLj/Y+deKcuVC46N2c/ymU8R44pgxbRUcpLDf9hrqOWlxvDhDy/kglkZ3tnvB+va2VPVwvxsz2KNF+d7+h2mpcciIvz1y8u4eUUeBZOTWLfnKC/vquKSORm8/f3zsduEl3fppLyxLqjJdMaYtcDaQWl3Bjj3vEGv1+NpgvJNawP0p50aMelxkWQlRnGksYP0ONfQF6iT0j/7/c2iarp73d5Z7FcvymLnkSZWTEsZcP51BTl8+0nP6PeL8ieQFOOkYHISr+0+yvcvmXVqM68+Fl2WQ40bv/n0QjLiXSyelBTqrIx7/TWJ/mHG/UEiNdbFb69fdNyii5fOm0BKjBObwAWzPPNmL8rPYN/RVvZWNdPZ08dThYe9w2fV2KHLcqhxY2leMhv/7cJQZ+O0EOmw47TbONLYQVxkxJCz2F0Rdn5w6SxKa1q9w5CvXDCRe94s4cb7N+I2hob2HmZnxvPiv64ccva4OnW0JqGUGpaF1qZR3zhvWlCz2K8ryOGOS2d7X2fER/L4LcuZkBDJ2dPT+MZ5U9lT2cyabRUnuIs61SSc5q8VFBSYwkJdxUOpsaC6uZMetxmwOu/JcLsNV/7xHepau1n3f84l1jXyDR37jrawvbyJTy3JHvF7j2UistkYUzD0mcfTmoRSaljS4yNHLECAZy2qu6+aS1VzJ38IsCrtybrxgY187+/baO0aek8N5aFBQik1ZiyZnMTl8zN5qvAwfaOwgmx/y8n2QasHq8A0SCilxpRVcybQ0N7DR4caaBvhX/xT0zyzxj86pEEiWBoklFJjyjkz0rDbhC89XMi5/7XeuwLtSIiwezrYPzrUMGL3HO80SCilxpSEKAcFk5No6uihtrWLwgMj94Xe2uUJOFsONeqGSEHSIKGUGnN+fEU+P//kXBx2YUPJyC0E2N98Vd/WzWMfHhpw7KF393trGIUH6rnv7VJqWvzvvnc60SChlBpz5mYlcOOyySyelMSGfbVDXxCk1s5e/mVJNmdPT+UXa/d4t3Ht7Onj7hd282//3ElXbx/femIr/7F2Lxf+z1un/SxwDRJKqTHrnBlp7K5sDrif9sfV1tVLbGQEP1g1i7buPl7fUw3Agbo23Ab2VDZz698+4khjB3dfNQeHXfjSw5tobO/m168W0dLZ4/e+Te09FB6oH5E8jjUaJJRSY9ZKa+/ud0tOvjZhjKG1u5c4VwRzJsYzMSGSdda2qsVHPbWFSIeN13Yf5ROz0vnc8sn84pr5HKxr518f/4g/vFHCSz5b4vr6yQu7PFvodvgPIuFM125SSo1Zc7MSSIx2sKG4lqsWZgHwVOFh8jM//iZS7d19GONZPl5EuDA/g6cKD/PNv20hwibYBNbedja9bsO0NM9S5+fNTCM11smGYk+Q2nKwwbvD4UeHGoh1RZAQ7eD5bRX0uQ1bDjZwvrWA4XihQUIpNWbZbcKKqam8U1yLMYZNBxr4v09vx2m38fsbFrFq7oSg79XfaR1jLfdx6dxMHnn/IC9u9+xpMTklminWPIp+EXYbV8yfyEPvHSDCJmw+6OnY3nywnhvu24gzwsbiyUn0ug12m7DpQL13CO94oc1NSqkx7ezpqVQ1d7L6j+/yo2d3kBrrJDc1mt+9Xkyf29DVG9w8iv6lOPrXhDpzagpvf/98rl44EfDspufPTWflclF+BjeflUtxdSs1LV1887GPyEyMJCHKwQeldXzv4pnMzUrg8Q8PMfeuVyirGT+d3VqTUEqNadcuyaa2tYt1e6rZd7SVu67Mxxi4+4XdXPmHd3Abw3O3rqC8oYN73iyhrauXX1wzn6SYgXtaDA4SAJNSornujBye3VrBtAz/QSIvNYb7P1/AeyW1PPDOfn707A6qmjt55ItLmZ+dQHefm/S4SBrbu9l22DOT+/U91cfVSsKVBgml1JjmsNu49YLp3HrBdGpbu0iJcVLT0sVPX9zN7spmAL739+28X1pHR3cvbd19nDezihuWThpwn9ZBzU39luelcNsF01ht1SgCWTw5iSmpMbyy6yizJsRx9vTUAUukf3JRNhWNnWw/0sg7JbV85ZwpQZXvvZJaluYlE2Efmw07YzNXSinlR2qsCxEhPT6SqxZM5NrF2XxyURbPb6vAFWHjuVtXMDEhkvVF1cdd22bNth68BLnNJnz34plMS4874XtHOuw8cctyLpiVzh2XzT5uD438ifHcc+NiLpiZzsb9dUE1g2051MBnHtjI45sOD3luqGhNQikVln57/SIAevvc/PDy2aTEOBERzp2ZzvPbKujudeOMsPG7dcVMS4+lu88KEpHD/9pLj4/kwZvPOOE5K6en8fD7B9lY5unEPpG3ijyzyV/YVsHnlk8edr5Gk9YklFJhLcJu89YwAM6bmUZrVy9v7K2m+GgLv1m3j5+9uJvmjv7mJvuo5ufs6amkxDj5y7v7hzx3Q7EnSHx4oJ7q5s5RzddwBRUkRGSViBSJSImI3H6C8z4lIkZECqzXuSLSISJbrce9PucuEZEd1j1/L8Hsf6iUUkM4d0YaMzPi+L9Pb+OuNbsAqGzq9G6LOho73vmKdNj5wopc3iyqYY/VZ9LR3cdzW4/wTvGxSYHNnT1sK2/ikjkZGAPPW0Nxx5ohg4SI2IF7gEuBfOAGEcn3c14ccBuwcdChUmPMQuvxNZ/0PwG3ANOtx6rhFUEppY6JdNh54KYCEqOdvFdax2eWTWJCfCSbDzZgE4hyjG5NAuBzy3OJcdq5961Smjt7WP3Hd/jWE1v5xmObvX0V7xbX0uc2fHFFHgtzEnnsg4NjcmXaYGoSS4ESY0yZMaYbeAK4ys95PwV+BQxZZxKRTCDeGPO+8WwV9QhwdfDZVkqpwHKSo3nr++fx5vfO464r870jnfpnW4+2hGgHNy6fzPPbKrjpwQ/ZX9vGV8+ZQnNnr7cfYt2eahKiHCyZnMRNZ02mrLaNd0tHbjHDkRJMkMgCfLvey600LxFZBOQYY17wc32eiHwkIm+JyNk+9yw/0T2VUupkiAh5qTG4IuxcvzSHCJuMelOTry+tzMNht1FS3covr53P9y6ZSUqMk9+9Xszv1hXzZlE1589MI8Ju47J5maTEOHniw7E3yimYvzF/YddbJxIRG/Ab4GY/51UCk4wxdSKyBHhWROYMdc8Bby5yC55mKSZNmuTvFKWUOqGM+EiuXpRFeUP7KX3Pdd89l8RoB3GRDsAzMfC+t8vYVeHpq7gwPwMAV4SdKxdM5G8fHqKpo4dnNpfz6AcHyc+M54+fWXRKaj+BBBMkyoEcn9fZQIXP6zhgLrDeKsgEYI2IrDbGFAJdAMaYzSJSCsyw7pl9gnt6GWPuA+4DKCgoGHsNdkqpsPCLa+ad8i/bnOToAa9vXzWL2z4xnXeKa3n2oyOcP/PYYoCfXJTFQ+8dYO2OSu7fUEZbVy8v7qjkuuIczh1iKO1oCqa5aRMwXUTyRMQJXA+s6T9ojGkyxqQaY3KNMbnAB8BqY0yhiKRZHd+IyBQ8HdRlxphKoEVEllujmj4PPDeyRVNKqWMi7LaQL7xns5q8Vs2dwL2fWzJg9vf87ARmZMTyX68UUdnUyR2XzSY9zsUDG8pCmOMggoQxphe4FXgF2AM8ZYzZJSJ3i8jqIS4/B9guItuAp4GvGWP6d+b4OvAAUAKUAi8NswxKKRX2RITvXjSD+rZu7Dbh0rkT+OzyyWworuVoCOdQBNWLY4xZC6wdlHZngHPP83n+DPBMgPMK8TRTKaWUAi6ZM4HlU5KJi3SQGO3kvJlp/M9r+9i4v57VC068ttRo0WU5lFJqjBAR/vqlZd6+k/zMeGJdEXy4v06DhFJKKQasBhtht7FkchIby0K3f7au3aSUUmPYsinJFFe3UtfaFZL315qEUkqNYSunpbJrXjNtXX2khGAfIw0SSik1hs3PTuSeGxeH7P21uUkppVRAGiSUUkoFpEFCKaVUQBoklFJKBaRBQimlVEAaJJRSSgWkQUIppVRAGiSUUkoFJJ4tpsODiNQAB62XqcDY2xB2ZGjZwpOWLXyN5/KlAjHGmGHtXBRWQcKXiBQaYwpCnY/RoGULT1q28DWey3eyZdPmJqWUUgFpkFBKKRVQOAeJ+0KdgVGkZQtPWrbwNZ7Ld1JlC9s+CaWUUqMvnGsSSimlRlnYBQkRWSUiRSJSIiK3hzo/I0FEDojIDhHZKiKFVlqyiLwmIsXWn0mhzmcwRORBEakWkZ0+aX7LIh6/tz7L7SISukXzgxCgbP8uIkesz26riFzmc+wOq2xFInJJaHIdHBHJEZE3RWSPiOwSkW9Z6WH/2Z2gbGH/2YlIpIh8KCLbrLL9xErPE5GN1uf2pIg4rXSX9brEOp475JsYY8LmAdiBUmAK4AS2AfmhztcIlOsAkDoo7VfA7dbz24FfhjqfQZblHGAxsHOosgCXAS8BAiwHNoY6/8Mo278D3/Nzbr7179MF5Fn/bu2hLsMJypYJLLaexwH7rDKE/Wd3grKF/Wdn/f3HWs8dwEbr83gKuN5Kvxf4uvX8G8C91vPrgSeHeo9wq0ksBUqMMWXGmG7gCeCqEOdptFwFPGw9fxi4OoR5CZox5m1g8K7tgcpyFfCI8fgASBSRzFOT048vQNkCuQp4whjTZYzZD5Tg+fc7JhljKo0xW6znLcAeIItx8NmdoGyBhM1nZ/39t1ovHdbDABcAT1vpgz+3/s/zaeATIiIneo9wCxJZwGGf1+Wc+MMOFwZ4VUQ2i8gtVlqGMaYSPP/IgfSQ5e7kBSrLePk8b7WaXB70aRYM27JZTRCL8PwqHVef3aCywTj47ETELiJbgWrgNTw1n0ZjTK91im/+vWWzjjcBKSe6f7gFCX8RbzwMz1phjFkMXAp8U0TOCXWGTpHx8Hn+CZgKLAQqgV9b6WFZNhGJBZ4Bvm2MaT7RqX7SxnT5/JRtXHx2xpg+Y8xCIBtPjWe2v9OsPz922cItSJQDOT6vs4GKEOVlxBhjKqw/q4F/4vmgj/ZX360/q0OXw5MWqCxh/3kaY45a/0ndwP0ca5YIu7KJiAPPl+hjxph/WMnj4rPzV7bx9NkBGGMagfV4+iQSRSTCOuSbf2/ZrOMJDNGEGm5BYhMw3eq5d+LpeFkT4jydFBGJEZG4/ufAxcBOPOW6yTrtJuC50ORwRAQqyxrg89ZImeVAU3/TRrgY1A7/STyfHXjKdr01miQPmA58eKrzFyyrXfrPwB5jzP/4HAr7zy5Q2cbDZyciaSKSaD2PAi7E0+fyJvAp67TBn1v/5/kp4A1j9WIHFOre+WH05l+GZ3RCKfDDUOdnBMozBc9Iim3Arv4y4WknfB0otv5MDnVegyzP43iq7j14frV8KVBZ8FR977E+yx1AQajzP4yyPWrlfbv1HzDT5/wfWmUrAi4Ndf6HKNtKPM0O24Gt1uOy8fDZnaBsYf/ZAfOBj6wy7ATutNKn4AlsJcDfAZeVHmm9LrGOTxnqPXTGtVJKqYDCrblJKaXUKaRBQimlVEAaJJRSSgWkQUIppVRAGiSUUkoFpEFCKaVUQBoklFJKBaRBQimlVED/H2gnSn+CNJllAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fa3b59a2080>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.sched.plot_loss()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "lr/10"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "lr/9"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "lr/3"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "lr/4"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Unfreeze Layers"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "# unfreeze layers\n",
    "learn.unfreeze()\n",
    "#lrs = np.array([lr/15,lr/12,lr/9])\n",
    "lrs = np.array([lr/9,lr/3,lr/1])\n",
    "\n",
    "#wd = 0.025\n",
    "#wds=[wd/100, wd/10, wd]\n",
    "#wds=[0.0025,0.025,0.25]\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "0128d73a58914b83ad632c0af648abee",
       "version_major": 2,
       "version_minor": 0
      },
      "text/html": [
       "<p>Failed to display Jupyter Widget of type <code>HBox</code>.</p>\n",
       "<p>\n",
       "  If you're reading this message in the Jupyter Notebook or JupyterLab Notebook, it may mean\n",
       "  that the widgets JavaScript is still loading. If this message persists, it\n",
       "  likely means that the widgets JavaScript library is either not installed or\n",
       "  not enabled. See the <a href=\"https://ipywidgets.readthedocs.io/en/stable/user_install.html\">Jupyter\n",
       "  Widgets Documentation</a> for setup instructions.\n",
       "</p>\n",
       "<p>\n",
       "  If you're reading this message in another frontend (for example, a static\n",
       "  rendering on GitHub or <a href=\"https://nbviewer.jupyter.org/\">NBViewer</a>),\n",
       "  it may mean that your frontend doesn't currently support widgets.\n",
       "</p>\n"
      ],
      "text/plain": [
       "HBox(children=(IntProgress(value=0, description='Epoch', max=7), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0.      0.57173 0.37544 0.8125 ]                          \n",
      "[1.      0.46852 0.319   0.84375]                          \n",
      "[2.      0.4216  0.37089 0.82954]                          \n",
      "[3.      0.44327 0.35393 0.81818]                          \n",
      "[4.      0.39818 0.27858 0.87784]                          \n",
      "[5.      0.37309 0.34977 0.87784]                          \n",
      "[6.      0.32137 0.28579 0.87216]                          \n",
      "\n"
     ]
    }
   ],
   "source": [
    "learn.fit(lrs, 3, cycle_len=1, cycle_mult=2)\n",
    "#learn.fit(lrs, 5, cycle_len=1, cycle_mult=2,use_wd_sched=True,wds=wds)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "learn.save('resnext50_sz75')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "learn.load('resnext50_sz75')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fa3b5970b70>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.sched.plot_loss()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "                                             \r"
     ]
    },
    {
     "data": {
      "text/plain": [
       "(0.2727280071497432, 0.8691588785046729)"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "log_preds,y = learn.TTA()\n",
    "preds = np.mean(np.exp(log_preds),axis=0)\n",
    "(metrics.log_loss(y,preds), accuracy(preds,y))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "8035683047154abd86ee76bf11741be7",
       "version_major": 2,
       "version_minor": 0
      },
      "text/html": [
       "<p>Failed to display Jupyter Widget of type <code>HBox</code>.</p>\n",
       "<p>\n",
       "  If you're reading this message in the Jupyter Notebook or JupyterLab Notebook, it may mean\n",
       "  that the widgets JavaScript is still loading. If this message persists, it\n",
       "  likely means that the widgets JavaScript library is either not installed or\n",
       "  not enabled. See the <a href=\"https://ipywidgets.readthedocs.io/en/stable/user_install.html\">Jupyter\n",
       "  Widgets Documentation</a> for setup instructions.\n",
       "</p>\n",
       "<p>\n",
       "  If you're reading this message in another frontend (for example, a static\n",
       "  rendering on GitHub or <a href=\"https://nbviewer.jupyter.org/\">NBViewer</a>),\n",
       "  it may mean that your frontend doesn't currently support widgets.\n",
       "</p>\n"
      ],
      "text/plain": [
       "HBox(children=(IntProgress(value=0, description='Epoch', max=7), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0.      0.29739 0.25388 0.89773]                          \n",
      "[1.      0.31191 0.28796 0.89204]                          \n",
      "[2.      0.28412 0.26385 0.89204]                          \n",
      "[3.      0.27975 0.26224 0.88068]                          \n",
      "[4.      0.27783 0.27686 0.88921]                          \n",
      "[5.      0.28858 0.25292 0.88352]                          \n",
      "[6.      0.25265 0.27511 0.87784]                          \n",
      "\n"
     ]
    }
   ],
   "source": [
    "learn.fit(lrs, 3, cycle_len=1, cycle_mult=2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "7f5e341d3d6c4777a705ac316d7c6b1c",
       "version_major": 2,
       "version_minor": 0
      },
      "text/html": [
       "<p>Failed to display Jupyter Widget of type <code>HBox</code>.</p>\n",
       "<p>\n",
       "  If you're reading this message in the Jupyter Notebook or JupyterLab Notebook, it may mean\n",
       "  that the widgets JavaScript is still loading. If this message persists, it\n",
       "  likely means that the widgets JavaScript library is either not installed or\n",
       "  not enabled. See the <a href=\"https://ipywidgets.readthedocs.io/en/stable/user_install.html\">Jupyter\n",
       "  Widgets Documentation</a> for setup instructions.\n",
       "</p>\n",
       "<p>\n",
       "  If you're reading this message in another frontend (for example, a static\n",
       "  rendering on GitHub or <a href=\"https://nbviewer.jupyter.org/\">NBViewer</a>),\n",
       "  it may mean that your frontend doesn't currently support widgets.\n",
       "</p>\n"
      ],
      "text/plain": [
       "HBox(children=(IntProgress(value=0, description='Epoch', max=7), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0.      0.23223 0.23889 0.90057]                          \n",
      "[1.      0.27451 0.25853 0.88636]                          \n",
      "[2.      0.24057 0.276   0.88636]                          \n",
      "[3.      0.24242 0.28228 0.875  ]                          \n",
      "[4.      0.23523 0.30291 0.87784]                          \n",
      "[5.      0.2119  0.25228 0.89204]                          \n",
      "[6.      0.18577 0.24105 0.89204]                          \n",
      "\n"
     ]
    }
   ],
   "source": [
    "learn.fit(lrs, 3, cycle_len=1, cycle_mult=2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fa3c0b14f98>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.sched.plot_loss()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Log Loss; Accuracy (final model here)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "                                             \r"
     ]
    },
    {
     "data": {
      "text/plain": [
       "(0.24804256874349564, 0.8940809968847352)"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "log_preds,y = learn.TTA()\n",
    "preds = np.mean(np.exp(log_preds),axis=0)\n",
    "(metrics.log_loss(y,preds), accuracy(preds,y))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Save predictions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_df = pd.read_json(PATH+\"/test.json\")\n",
    "helper.make_test_file(log_preds,test_df,model_name='resnext50_sz75_zm13_bs32_comp_band',\n",
    "                      classes=data.classes,\n",
    "                      submission_folder =f'{PATH}sub'\n",
    "                     )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}