Algonauts_Challenge_Codes.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# Loading the required packages\n",
    "%reload_ext autoreload\n",
    "%autoreload 2\n",
    "%matplotlib inline\n",
    "from fastai import *\n",
    "from fastai.vision import *\n",
    "from fastai.callbacks import *\n",
    "from torchvision.models import *\n",
    "import pretrainedmodels\n",
    "from fastai.vision import *\n",
    "from fastai.vision.models import *\n",
    "from fastai.vision.learner import model_meta\n",
    "from utils import *\n",
    "import sys"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "import scipy.io\n",
    "import numpy as np\n",
    "import csv\n",
    "import pandas as pd\n",
    "import os,sys\n",
    "#import tensorflow as tf\n",
    "from numpy import linalg as LA\n",
    "from typing import Callable"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Main Path\n",
    "path_main=Path('/home/visionlab/Georgin/algonautsChallenge_codes/')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# All paths relative to main paths\n",
    "path_training_image_pairs=Path(path_main/'Training_Data/210_imagepairs/')\n",
    "path_training_csv_set1=Path(path_main/'Training_Data/92_Image_Set/')\n",
    "path_training_csv_set2=Path(path_main/'Training_Data/118_Image_Set/')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# select early or late response using this flag_EL\n",
    "# Early=1, Late=2\n",
    "flag_EL=1\n",
    "# select fmRI or MEG using this flag\n",
    "flag_FM=1 # 1 for MEG and 2 for Fmri "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "scrolled": false
   },
   "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>ImagePair</th>\n",
       "      <th>earlyDistance</th>\n",
       "      <th>lateDistance</th>\n",
       "      <th>fmri_evc</th>\n",
       "      <th>fmri_it</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>set92_image_01_image_02.jpg</td>\n",
       "      <td>0.374229</td>\n",
       "      <td>0.557752</td>\n",
       "      <td>0.651895</td>\n",
       "      <td>0.438132</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>set92_image_01_image_03.jpg</td>\n",
       "      <td>0.368932</td>\n",
       "      <td>0.439496</td>\n",
       "      <td>0.626384</td>\n",
       "      <td>0.395722</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>set92_image_01_image_04.jpg</td>\n",
       "      <td>0.463916</td>\n",
       "      <td>0.524331</td>\n",
       "      <td>0.627330</td>\n",
       "      <td>0.416473</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>set92_image_01_image_05.jpg</td>\n",
       "      <td>0.666419</td>\n",
       "      <td>0.518694</td>\n",
       "      <td>0.736833</td>\n",
       "      <td>0.454702</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>set92_image_01_image_06.jpg</td>\n",
       "      <td>0.421931</td>\n",
       "      <td>0.490629</td>\n",
       "      <td>0.644696</td>\n",
       "      <td>0.415026</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                     ImagePair  earlyDistance  lateDistance  fmri_evc  \\\n",
       "0  set92_image_01_image_02.jpg       0.374229      0.557752  0.651895   \n",
       "1  set92_image_01_image_03.jpg       0.368932      0.439496  0.626384   \n",
       "2  set92_image_01_image_04.jpg       0.463916      0.524331  0.627330   \n",
       "3  set92_image_01_image_05.jpg       0.666419      0.518694  0.736833   \n",
       "4  set92_image_01_image_06.jpg       0.421931      0.490629  0.644696   \n",
       "\n",
       "    fmri_it  \n",
       "0  0.438132  \n",
       "1  0.395722  \n",
       "2  0.416473  \n",
       "3  0.454702  \n",
       "4  0.415026  "
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Training the DATA\n",
    "df_set1 = pd.read_csv(path_training_csv_set1/'set1_training_distance_file.csv'); \n",
    "df_set2=pd.read_csv(path_training_csv_set2/'set2_training_distance_file.csv')\n",
    "df=pd.concat([df_set1.loc[0:92*91-1],df_set2.loc[0:118*117-1]], ignore_index=True)# removed the identical pairs\n",
    "#df=pd.concat([df_set1,df_set2], ignore_index=True)# removed the identical pairs\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# normlization values of mean and std taken from ImageNet\n",
    "mean, std = torch.tensor(imagenet_stats)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# Custom Data bunch\n",
    "# Creating the data bunch by concatenating two image together after normalizaing the images separately. \n",
    "# Size of the image is fixed to 175x175\n",
    "class SiamImage(ItemBase):\n",
    "    def __init__(self, img1, img2): ## These should of Image type\n",
    "        self.img1, self.img2 = img1, img2\n",
    "        self.obj, self.data = (img1, img2), [(img1.data-mean[...,None,None])/std[...,None,None], (img2.data-mean[...,None,None])/std[...,None,None]]\n",
    "    def apply_tfms(self, tfms,*args, **kwargs):\n",
    "        self.img1 = self.img1.apply_tfms(tfms, *args, **kwargs)\n",
    "        self.img2 = self.img2.apply_tfms(tfms, *args, **kwargs)\n",
    "        self.data = [(self.img1.data-mean[...,None,None])/std[...,None,None], (self.img2.data-mean[...,None,None])/std[...,None,None]]\n",
    "        return self\n",
    "    def __repr__(self): return f'{self.__class__.__name__} {self.img1.shape, self.img2.shape}'\n",
    "    def to_one(self):\n",
    "        return Image(mean[...,None,None]+torch.cat(self.data,2)*std[...,None,None])\n",
    "    \n",
    "    \n",
    "    \n",
    "class SiamImageItemList(ImageList):\n",
    "    def __init__(self, *args, **kwargs):\n",
    "        super().__init__(*args, **kwargs)\n",
    "#         self._label_cls=FloatList\n",
    "    \n",
    "    def __len__(self)->int: return len(self.items) or 1 \n",
    "    \n",
    "    def get(self, i):\n",
    "        img1=super().get(i)\n",
    "        crop(img1,175,*[0.0,0.0])\n",
    "        img2=super().get(i) \n",
    "        crop(img2,175,*[0.0,1.0])\n",
    "        return SiamImage(img1, img2)\n",
    "    def reconstruct(self, t): return SiamImage(mean[...,None,None]+t[0]*std[...,None,None], mean[...,None,None]+t[1]*std[...,None,None])\n",
    "    \n",
    "    def show_xys(self, xs, ys, figsize:Tuple[int,int]=(9,10), **kwargs):\n",
    "        rows = int(math.sqrt(len(xs)))\n",
    "        fig, axs = plt.subplots(rows,rows,figsize=figsize)\n",
    "        for i, ax in enumerate(axs.flatten() if rows > 1 else [axs]):\n",
    "            xs[i].to_one().show(ax=ax, y=ys[i], **kwargs)\n",
    "        plt.tight_layout()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# Preparing the data\n",
    "whl_tfms=get_transforms(do_flip=False,max_rotate=0.0,max_warp=0.0)\n",
    "bs =64\n",
    "# early data/Late\n",
    "# flag_EL decides by selecting diffrent columns\n",
    "data = (SiamImageItemList.from_df(df=df, path=path_training_image_pairs, cols=0)\n",
    "         .split_by_rand_pct(valid_pct=0.005, seed=34)\n",
    "         .label_from_df(cols=flag_EL, label_cls=FloatList)\n",
    "         .transform(whl_tfms, size=224)\n",
    "         .databunch(bs=bs))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x7fa17a3f1290>"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Checking if the normalization done above is correct\n",
    "x = next(iter(data.train_dl))\n",
    "t=x[0][0][0].cpu()\n",
    "to = mean[...,None,None] + t* std[...,None,None]\n",
    "import torchvision\n",
    "ti = torchvision.transforms.ToPILImage(to)\n",
    "plt.imshow(to.numpy().transpose(1,2,0))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "def alexnet_ITmodel(pretrained=True):\n",
    "    pretrained = 'imagenet' if pretrained else None\n",
    "    model = pretrainedmodels.alexnet(pretrained=pretrained)\n",
    "    all_layers = list(model.children())\n",
    "    return nn.Sequential(*all_layers[1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# alternate model\n",
    "# DNN training Model\n",
    "from fastai.vision import learner\n",
    "from numpy import linalg as LA\n",
    "class SiameseNet(nn.Module):\n",
    "    def __init__(self):\n",
    "        super(SiameseNet, self).__init__()\n",
    "        self.body = alexnet_ITmodel()\n",
    "        self.groupconv=nn.Sequential(\n",
    "                       nn. Conv2d(512, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1),groups=256),\n",
    "                       nn.ReLU(),\n",
    "                       nn.AvgPool2d(kernel_size=3, stride=2, padding=0, ceil_mode=False     ))\n",
    "        self.distance=nn.Linear(1024,1)\n",
    "    def forward(self, x1, x2):\n",
    "       # pdb.set_trace();\n",
    "        output1 = self.body(x1) \n",
    "        output2 = self.body(x2)\n",
    "        out=torch.stack((output1,output2),dim=2)\n",
    "        out=out.reshape([ output1.shape[0],output1.shape[1]*2,output1.shape[2],output1.shape[3]])\n",
    "        out=self.groupconv(out)\n",
    "        out = out.view(out.size()[0], -1)\n",
    "        dist=self.distance(out)\n",
    "        dist=torch.relu(dist)\n",
    "       # output  = torch.sigmoid(dist)\n",
    "        return dist\n",
    "\n",
    "    def get_embedding(self, x1,x2):\n",
    "       # pdb.set_trace();\n",
    "        output1 = self.body(x1) \n",
    "        output2 = self.body(x2)\n",
    "        out=torch.stack((output1,output2),dim=2)\n",
    "        out=out.reshape([ output1.shape[0],output1.shape[1]*2,output1.shape[2],output1.shape[3]])\n",
    "        out=self.groupconv(out)\n",
    "        out = out.view(out.size()[0], -1)\n",
    "        dist=self.distance(out)\n",
    "        dist=torch.relu(dist)\n",
    "       # output  = torch.sigmoid(dist)\n",
    "        return dist"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "import pdb;"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# Learner\n",
    "model = SiameseNet().cuda()\n",
    "#apply_init(model.head, nn.init.kaiming_normal_)\n",
    "#loss_func=ContrastiveLoss().cuda()\n",
    "siam_learner = Learner(data, model,metrics=explained_variance)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "SiameseNet(\n",
       "  (body): Sequential(\n",
       "    (0): Conv2d(3, 64, kernel_size=(11, 11), stride=(4, 4), padding=(2, 2))\n",
       "    (1): ReLU(inplace=True)\n",
       "    (2): MaxPool2d(kernel_size=3, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
       "    (3): Conv2d(64, 192, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2))\n",
       "    (4): ReLU(inplace=True)\n",
       "    (5): MaxPool2d(kernel_size=3, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
       "    (6): Conv2d(192, 384, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
       "    (7): ReLU(inplace=True)\n",
       "    (8): Conv2d(384, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
       "    (9): ReLU(inplace=True)\n",
       "    (10): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
       "    (11): ReLU(inplace=True)\n",
       "    (12): MaxPool2d(kernel_size=3, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
       "  )\n",
       "  (groupconv): Sequential(\n",
       "    (0): Conv2d(512, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=256)\n",
       "    (1): ReLU()\n",
       "    (2): AvgPool2d(kernel_size=3, stride=2, padding=0)\n",
       "  )\n",
       "  (distance): Linear(in_features=1024, out_features=1, bias=True)\n",
       ")"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "LR Finder is complete, type {learner_name}.recorder.plot() to see the graph.\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": [
    "siam_learner.lr_find()\n",
    "siam_learner.recorder.plot()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "if(flag_EL==1):\n",
    "    path_model=Path(path_main/'train_models/alexnet/evc')\n",
    "if(flag_EL==2):\n",
    "    path_model=Path(path_main/'train_models/alexnet/it')\n",
    "epoch_count=0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0.00072, 0.00064, 0.00056, 0.00048, 0.0004, 0.00032, 0.00024, 0.00016, 8e-05]\n"
     ]
    }
   ],
   "source": [
    "EpochMax=10\n",
    "Lmax=8*1e-4\n",
    "cycle_per_lr=1\n",
    "LR=[round(Lmax*(1-(epoch/EpochMax)),5) for epoch in range(1,EpochMax)]\n",
    "print(LR)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.001753</td>\n",
       "      <td>0.001835</td>\n",
       "      <td>0.773585</td>\n",
       "      <td>00:35</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.001283</td>\n",
       "      <td>0.000923</td>\n",
       "      <td>0.882365</td>\n",
       "      <td>00:35</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.001045</td>\n",
       "      <td>0.000787</td>\n",
       "      <td>0.904613</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000868</td>\n",
       "      <td>0.000688</td>\n",
       "      <td>0.913821</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000760</td>\n",
       "      <td>0.000601</td>\n",
       "      <td>0.927966</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000635</td>\n",
       "      <td>0.000537</td>\n",
       "      <td>0.934594</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000546</td>\n",
       "      <td>0.000482</td>\n",
       "      <td>0.940771</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000499</td>\n",
       "      <td>0.000451</td>\n",
       "      <td>0.945839</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000442</td>\n",
       "      <td>0.000377</td>\n",
       "      <td>0.953469</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "for lr in LR:\n",
    "    siam_learner.fit_one_cycle(cycle_per_lr,lr)\n",
    "    epoch_count+=cycle_per_lr\n",
    "    name='evc_epoch='+str(epoch_count)+'learning_rate='+str(lr)\n",
    "    siam_learner.save(path_model/name)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "LR Finder is complete, type {learner_name}.recorder.plot() to see the graph.\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": [
    "siam_learner.lr_find()\n",
    "siam_learner.recorder.plot()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "LR Finder is complete, type {learner_name}.recorder.plot() to see the graph.\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": [
    "# Unfreezing last convolutional lay\n",
    "siam_learner.unfreeze\n",
    "siam_learner.lr_find()\n",
    "siam_learner.recorder.plot()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#EpochMax=500\n",
    "EpochMax=200\n",
    "Lmax=2*1e-5\n",
    "cycle_per_lr=1\n",
    "LR=[round(Lmax*(1-(epoch/EpochMax)),10) for epoch in range(1,EpochMax)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000405</td>\n",
       "      <td>0.000392</td>\n",
       "      <td>0.952966</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000391</td>\n",
       "      <td>0.000401</td>\n",
       "      <td>0.951813</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000371</td>\n",
       "      <td>0.000386</td>\n",
       "      <td>0.953704</td>\n",
       "      <td>00:35</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000379</td>\n",
       "      <td>0.000383</td>\n",
       "      <td>0.954024</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000372</td>\n",
       "      <td>0.000386</td>\n",
       "      <td>0.953307</td>\n",
       "      <td>00:35</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000364</td>\n",
       "      <td>0.000393</td>\n",
       "      <td>0.952522</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000358</td>\n",
       "      <td>0.000369</td>\n",
       "      <td>0.955390</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000360</td>\n",
       "      <td>0.000351</td>\n",
       "      <td>0.957345</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000341</td>\n",
       "      <td>0.000351</td>\n",
       "      <td>0.957052</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000340</td>\n",
       "      <td>0.000364</td>\n",
       "      <td>0.956092</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000342</td>\n",
       "      <td>0.000360</td>\n",
       "      <td>0.956410</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000338</td>\n",
       "      <td>0.000371</td>\n",
       "      <td>0.955645</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000330</td>\n",
       "      <td>0.000367</td>\n",
       "      <td>0.956036</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000328</td>\n",
       "      <td>0.000348</td>\n",
       "      <td>0.958283</td>\n",
       "      <td>00:35</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000330</td>\n",
       "      <td>0.000337</td>\n",
       "      <td>0.959747</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000330</td>\n",
       "      <td>0.000347</td>\n",
       "      <td>0.958645</td>\n",
       "      <td>00:35</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000317</td>\n",
       "      <td>0.000339</td>\n",
       "      <td>0.959598</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000316</td>\n",
       "      <td>0.000341</td>\n",
       "      <td>0.959582</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000312</td>\n",
       "      <td>0.000355</td>\n",
       "      <td>0.958284</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000305</td>\n",
       "      <td>0.000323</td>\n",
       "      <td>0.961349</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000307</td>\n",
       "      <td>0.000320</td>\n",
       "      <td>0.962279</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000304</td>\n",
       "      <td>0.000331</td>\n",
       "      <td>0.960909</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000303</td>\n",
       "      <td>0.000337</td>\n",
       "      <td>0.960503</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000298</td>\n",
       "      <td>0.000321</td>\n",
       "      <td>0.962162</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000290</td>\n",
       "      <td>0.000313</td>\n",
       "      <td>0.962605</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000283</td>\n",
       "      <td>0.000320</td>\n",
       "      <td>0.962812</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000286</td>\n",
       "      <td>0.000328</td>\n",
       "      <td>0.961317</td>\n",
       "      <td>00:35</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000285</td>\n",
       "      <td>0.000309</td>\n",
       "      <td>0.964574</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000286</td>\n",
       "      <td>0.000334</td>\n",
       "      <td>0.961319</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000285</td>\n",
       "      <td>0.000323</td>\n",
       "      <td>0.962917</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000281</td>\n",
       "      <td>0.000322</td>\n",
       "      <td>0.962870</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000279</td>\n",
       "      <td>0.000318</td>\n",
       "      <td>0.962981</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000271</td>\n",
       "      <td>0.000320</td>\n",
       "      <td>0.963061</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000265</td>\n",
       "      <td>0.000327</td>\n",
       "      <td>0.962230</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000273</td>\n",
       "      <td>0.000313</td>\n",
       "      <td>0.963787</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000263</td>\n",
       "      <td>0.000318</td>\n",
       "      <td>0.963618</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000260</td>\n",
       "      <td>0.000310</td>\n",
       "      <td>0.963338</td>\n",
       "      <td>00:35</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000277</td>\n",
       "      <td>0.000324</td>\n",
       "      <td>0.962815</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000261</td>\n",
       "      <td>0.000302</td>\n",
       "      <td>0.965021</td>\n",
       "      <td>00:35</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000256</td>\n",
       "      <td>0.000307</td>\n",
       "      <td>0.964141</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000261</td>\n",
       "      <td>0.000308</td>\n",
       "      <td>0.964306</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000261</td>\n",
       "      <td>0.000303</td>\n",
       "      <td>0.965047</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000253</td>\n",
       "      <td>0.000305</td>\n",
       "      <td>0.964649</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000253</td>\n",
       "      <td>0.000300</td>\n",
       "      <td>0.965081</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000252</td>\n",
       "      <td>0.000299</td>\n",
       "      <td>0.965248</td>\n",
       "      <td>00:35</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000247</td>\n",
       "      <td>0.000291</td>\n",
       "      <td>0.966062</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000245</td>\n",
       "      <td>0.000314</td>\n",
       "      <td>0.963485</td>\n",
       "      <td>00:35</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000238</td>\n",
       "      <td>0.000304</td>\n",
       "      <td>0.965363</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000240</td>\n",
       "      <td>0.000301</td>\n",
       "      <td>0.965052</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000236</td>\n",
       "      <td>0.000295</td>\n",
       "      <td>0.966015</td>\n",
       "      <td>00:35</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000239</td>\n",
       "      <td>0.000308</td>\n",
       "      <td>0.964589</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000240</td>\n",
       "      <td>0.000310</td>\n",
       "      <td>0.963935</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000232</td>\n",
       "      <td>0.000284</td>\n",
       "      <td>0.967208</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000239</td>\n",
       "      <td>0.000287</td>\n",
       "      <td>0.967072</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000239</td>\n",
       "      <td>0.000282</td>\n",
       "      <td>0.967030</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000232</td>\n",
       "      <td>0.000296</td>\n",
       "      <td>0.965766</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000235</td>\n",
       "      <td>0.000303</td>\n",
       "      <td>0.964681</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000226</td>\n",
       "      <td>0.000291</td>\n",
       "      <td>0.965732</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000228</td>\n",
       "      <td>0.000285</td>\n",
       "      <td>0.967252</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000228</td>\n",
       "      <td>0.000298</td>\n",
       "      <td>0.965723</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000222</td>\n",
       "      <td>0.000293</td>\n",
       "      <td>0.966756</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000217</td>\n",
       "      <td>0.000297</td>\n",
       "      <td>0.966235</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000221</td>\n",
       "      <td>0.000284</td>\n",
       "      <td>0.967094</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000224</td>\n",
       "      <td>0.000275</td>\n",
       "      <td>0.968236</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000219</td>\n",
       "      <td>0.000277</td>\n",
       "      <td>0.967155</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000215</td>\n",
       "      <td>0.000275</td>\n",
       "      <td>0.967885</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000216</td>\n",
       "      <td>0.000281</td>\n",
       "      <td>0.967318</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000211</td>\n",
       "      <td>0.000272</td>\n",
       "      <td>0.968207</td>\n",
       "      <td>00:35</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000218</td>\n",
       "      <td>0.000276</td>\n",
       "      <td>0.968190</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000213</td>\n",
       "      <td>0.000276</td>\n",
       "      <td>0.967905</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000213</td>\n",
       "      <td>0.000282</td>\n",
       "      <td>0.967462</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000212</td>\n",
       "      <td>0.000268</td>\n",
       "      <td>0.968582</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000217</td>\n",
       "      <td>0.000271</td>\n",
       "      <td>0.968381</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000212</td>\n",
       "      <td>0.000261</td>\n",
       "      <td>0.969347</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000205</td>\n",
       "      <td>0.000262</td>\n",
       "      <td>0.969977</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000205</td>\n",
       "      <td>0.000277</td>\n",
       "      <td>0.968714</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000213</td>\n",
       "      <td>0.000269</td>\n",
       "      <td>0.968994</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000202</td>\n",
       "      <td>0.000271</td>\n",
       "      <td>0.968593</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000197</td>\n",
       "      <td>0.000263</td>\n",
       "      <td>0.969221</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000200</td>\n",
       "      <td>0.000258</td>\n",
       "      <td>0.969726</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000199</td>\n",
       "      <td>0.000269</td>\n",
       "      <td>0.968746</td>\n",
       "      <td>00:35</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000197</td>\n",
       "      <td>0.000266</td>\n",
       "      <td>0.969030</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000198</td>\n",
       "      <td>0.000268</td>\n",
       "      <td>0.969180</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000200</td>\n",
       "      <td>0.000269</td>\n",
       "      <td>0.968862</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000198</td>\n",
       "      <td>0.000268</td>\n",
       "      <td>0.968788</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000193</td>\n",
       "      <td>0.000264</td>\n",
       "      <td>0.969160</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000199</td>\n",
       "      <td>0.000267</td>\n",
       "      <td>0.969072</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000194</td>\n",
       "      <td>0.000258</td>\n",
       "      <td>0.970455</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000188</td>\n",
       "      <td>0.000261</td>\n",
       "      <td>0.970018</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000191</td>\n",
       "      <td>0.000261</td>\n",
       "      <td>0.969442</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000195</td>\n",
       "      <td>0.000265</td>\n",
       "      <td>0.969310</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000186</td>\n",
       "      <td>0.000252</td>\n",
       "      <td>0.970633</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000186</td>\n",
       "      <td>0.000265</td>\n",
       "      <td>0.969585</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000186</td>\n",
       "      <td>0.000260</td>\n",
       "      <td>0.970600</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000191</td>\n",
       "      <td>0.000255</td>\n",
       "      <td>0.970310</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000184</td>\n",
       "      <td>0.000244</td>\n",
       "      <td>0.971608</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000185</td>\n",
       "      <td>0.000257</td>\n",
       "      <td>0.970234</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000186</td>\n",
       "      <td>0.000255</td>\n",
       "      <td>0.970690</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000180</td>\n",
       "      <td>0.000241</td>\n",
       "      <td>0.971758</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000185</td>\n",
       "      <td>0.000248</td>\n",
       "      <td>0.970831</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000178</td>\n",
       "      <td>0.000253</td>\n",
       "      <td>0.970296</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000184</td>\n",
       "      <td>0.000261</td>\n",
       "      <td>0.968994</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000178</td>\n",
       "      <td>0.000250</td>\n",
       "      <td>0.970419</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000183</td>\n",
       "      <td>0.000243</td>\n",
       "      <td>0.970931</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000179</td>\n",
       "      <td>0.000238</td>\n",
       "      <td>0.971647</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000180</td>\n",
       "      <td>0.000240</td>\n",
       "      <td>0.971434</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000177</td>\n",
       "      <td>0.000251</td>\n",
       "      <td>0.970442</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000178</td>\n",
       "      <td>0.000246</td>\n",
       "      <td>0.970903</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000180</td>\n",
       "      <td>0.000252</td>\n",
       "      <td>0.970278</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000177</td>\n",
       "      <td>0.000246</td>\n",
       "      <td>0.970648</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000174</td>\n",
       "      <td>0.000242</td>\n",
       "      <td>0.971857</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000178</td>\n",
       "      <td>0.000242</td>\n",
       "      <td>0.971059</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000178</td>\n",
       "      <td>0.000250</td>\n",
       "      <td>0.970332</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000174</td>\n",
       "      <td>0.000255</td>\n",
       "      <td>0.970666</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000170</td>\n",
       "      <td>0.000243</td>\n",
       "      <td>0.971457</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000175</td>\n",
       "      <td>0.000236</td>\n",
       "      <td>0.972160</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000171</td>\n",
       "      <td>0.000241</td>\n",
       "      <td>0.971768</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000170</td>\n",
       "      <td>0.000237</td>\n",
       "      <td>0.972036</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000170</td>\n",
       "      <td>0.000237</td>\n",
       "      <td>0.972005</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000165</td>\n",
       "      <td>0.000245</td>\n",
       "      <td>0.971548</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000168</td>\n",
       "      <td>0.000238</td>\n",
       "      <td>0.972095</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000167</td>\n",
       "      <td>0.000234</td>\n",
       "      <td>0.972554</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000167</td>\n",
       "      <td>0.000242</td>\n",
       "      <td>0.971705</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000167</td>\n",
       "      <td>0.000228</td>\n",
       "      <td>0.973514</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000168</td>\n",
       "      <td>0.000231</td>\n",
       "      <td>0.972823</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000166</td>\n",
       "      <td>0.000229</td>\n",
       "      <td>0.973171</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000163</td>\n",
       "      <td>0.000234</td>\n",
       "      <td>0.972918</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000169</td>\n",
       "      <td>0.000225</td>\n",
       "      <td>0.973499</td>\n",
       "      <td>00:34</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000171</td>\n",
       "      <td>0.000234</td>\n",
       "      <td>0.972679</td>\n",
       "      <td>00:35</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000161</td>\n",
       "      <td>0.000232</td>\n",
       "      <td>0.972922</td>\n",
       "      <td>00:35</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000160</td>\n",
       "      <td>0.000230</td>\n",
       "      <td>0.973064</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000164</td>\n",
       "      <td>0.000233</td>\n",
       "      <td>0.972648</td>\n",
       "      <td>00:35</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.000162</td>\n",
       "      <td>0.000236</td>\n",
       "      <td>0.972265</td>\n",
       "      <td>00:35</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "\n",
       "    <div>\n",
       "        <style>\n",
       "            /* Turns off some styling */\n",
       "            progress {\n",
       "                /* gets rid of default border in Firefox and Opera. */\n",
       "                border: none;\n",
       "                /* Needs to be in here for Safari polyfill so background images work as expected. */\n",
       "                background-size: auto;\n",
       "            }\n",
       "            .progress-bar-interrupted, .progress-bar-interrupted::-webkit-progress-bar {\n",
       "                background: #F44336;\n",
       "            }\n",
       "        </style>\n",
       "      <progress value='0' class='' max='1', style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
       "      0.00% [0/1 00:00<00:00]\n",
       "    </div>\n",
       "    \n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>explained_variance</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "  </tbody>\n",
       "</table><p>\n",
       "\n",
       "    <div>\n",
       "        <style>\n",
       "            /* Turns off some styling */\n",
       "            progress {\n",
       "                /* gets rid of default border in Firefox and Opera. */\n",
       "                border: none;\n",
       "                /* Needs to be in here for Safari polyfill so background images work as expected. */\n",
       "                background-size: auto;\n",
       "            }\n",
       "            .progress-bar-interrupted, .progress-bar-interrupted::-webkit-progress-bar {\n",
       "                background: #F44336;\n",
       "            }\n",
       "        </style>\n",
       "      <progress value='0' class='' max='2', style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
       "      0.00% [0/2 00:00<00:00]\n",
       "    </div>\n",
       "    "
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "for lr in LR:\n",
    "    siam_learner.fit_one_cycle(cycle_per_lr,lr)\n",
    "    epoch_count+=cycle_per_lr\n",
    "    name='evc_epoch='+str(epoch_count)+'learning_rate='+str(lr)\n",
    "    siam_learner.save(path_model/name)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Initializing the Matlibplot\n",
    "import matplotlib\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "matplotlib.style.use('ggplot')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Finding the correlation observed and predicted correlational distance\n",
    "dl=data.train_dl\n",
    "ds=data.train_ds\n",
    "dlv=data.valid_dl\n",
    "dsv=data.valid_ds\n",
    "x=siam_learner.get_preds(dlv)\n",
    "preds=x[0].numpy()\n",
    "areds=x[1].numpy()\n",
    "\n",
    "plt.scatter(preds,areds)\n",
    "plt.xlabel('predicted distance')\n",
    "plt.ylabel('observed distance')\n",
    "plt.title('Training Data, set1 and set2')\n",
    "plt.show()\n",
    "model_correlation=np.corrcoef(preds.transpose(),areds)[0][1]\n",
    "print('Correlation between observed and Predicted =',model_correlation)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Extracting the distance from last layer\n",
    "# This is done by creating image pair using test and and finding the distance from the last layer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "path_test_file=Path(path_main/'Test_Data')\n",
    "path_test_image_pairs=Path(path_main/'Test_Data/78images/')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "image_file_names=os.listdir(path_test_file/'78images_imagepairs')\n",
    "image_file_names=np.sort(image_file_names)\n",
    "image_file_names=image_file_names.tolist()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_test = pd.read_csv(path_test_file/'testset_testing_distance_file.csv')\n",
    "df_test.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "data_test = (SiamImageItemList.from_df(df=df_test, path=path_test_image_pairs, cols=0)\n",
    "         .random_split_by_pct(valid_pct=1, seed=34)\n",
    "         .label_from_df(cols=1, label_cls=FloatList)\n",
    "         .transform(whl_tfms, size=224)\n",
    "         .databunch(bs=bs))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def extract_embedding_test(dl, ds, mdl):\n",
    "    mdl.eval()\n",
    "    with torch.no_grad():\n",
    "        #preds = torch.zeros((len(ds), 1))\n",
    "        #areds = torch.zeros(1,(len(ds)))\n",
    "        preds=np.zeros((len(ds),1))\n",
    "        areds=np.zeros((len(ds),1))\n",
    "        start=0\n",
    "        for cnt, (x, y) in enumerate(dl, start=0):\n",
    "            size = x[0].shape[0]\n",
    "            index=y.cpu()\n",
    "            index=index.numpy().astype(int)\n",
    "           # pdb.set_trace();\n",
    "            value=model.get_embedding(x[0],x[1])\n",
    "            #index=list(range(cnt+))\n",
    "            preds[index] =value.cpu().numpy()\n",
    "            start+= size\n",
    "        return preds"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "dlt=data_test.valid_dl\n",
    "dst=data_test.valid_ds\n",
    "preds = extract_embedding_test(dlt,dst, model)\n",
    "preds=preds.squeeze()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Creating the RDMs of 78 image test set \n",
    "from itertools import combinations\n",
    "N=78\n",
    "RDM=np.zeros([N,N])\n",
    "combAB = combinations(range(0,N),2)\n",
    "combAB=list(combAB)\n",
    "Nc=len(combAB)\n",
    "count=0\n",
    "for ij in combAB:\n",
    "    i=ij[0]\n",
    "    j=ij[1]\n",
    "    RDM[i][j] = preds[count]\n",
    "    RDM[j][i]=  RDM[i][j]\n",
    "    count=count+1;"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import scipy.io as sio\n",
    "if(flag_EL==1):\n",
    "    sio.savemat(path_test_image_pairs/'../RDM_EVS_groupConv_res_18_ver8.mat', {'evs':RDM})\n",
    "if(flag_EL==2):\n",
    "    sio.savemat(path_test_image_pairs/'../RDM_IT_groupConv_res_18_ver8mat', {'it':RDM}) \n",
    "plt.matshow(RDM);    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# After submitting the models we found that the last layers were not giving the best solutions\n",
    "# Hence we took the intermediate layer features and got the distance from it. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# this set has 78 images identical\n",
    "path_test_image_pairs_combined=Path(path_main/'Test_Data/78images_combined')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "image_file_names=os.listdir(path_test_image_pairs_combined)\n",
    "image_file_names=np.sort(image_file_names)\n",
    "image_file_names=image_file_names.tolist()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_cond=[]\n",
    "count=0\n",
    "for i in image_file_names:\n",
    "    one=['','']\n",
    "    one[0]=i\n",
    "    one[1]=count\n",
    "    count=count+1\n",
    "    test_cond.append(one)\n",
    "df_test_identical=DataFrame(test_cond,columns=['Image Pairs','Count'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "data_test_identical = (SiamImageItemList.from_df(df=df_test_identical, path=path_test_image_pairs_combined, cols=0)\n",
    "         .random_split_by_pct(valid_pct=1, seed=34)\n",
    "         .label_from_df(cols=1, label_cls=FloatList)\n",
    "         .transform(whl_tfms, size=224)\n",
    "         .databunch(bs=bs))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Feature Extraction from the trained model\n",
    "class SaveFeatures:\n",
    "    def __init__(self, m):\n",
    "        self.handle = m.register_forward_hook(self.hook_fn)\n",
    "    def hook_fn(self, m, inp, outp):\n",
    "        self.features = outp\n",
    "    def remove(self):\n",
    "        self.handle.remove()\n",
    "mdl=siam_learner.model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "mdl.body"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# defining the layers \n",
    "sf = [SaveFeatures(m) for m in [mdl.body[12]]]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "Predicted_Features=np.zeros((len( data_test_identical.valid_ds.x),9216))\n",
    "for i in range(0,len(data_test_identical.valid_ds.x)):\n",
    "    #pdb.set_trace()\n",
    "    img1 = data_test_identical.valid_ds.x[i]\n",
    "    ind=data_test_identical.valid_ds.y[i]\n",
    "    ind=int(float(str(ind)))\n",
    "    siam_learner.predict(img1)\n",
    "    #\n",
    "    Feat_oneimage=[o.features.cpu().numpy() for o in sf]\n",
    "    Feat_oneimage=Feat_oneimage[0][0]\n",
    "    Feat_oneimage=np.reshape(Feat_oneimage,(np.size(Feat_oneimage),1))\n",
    "    #pdb.set_trace()\n",
    "    Predicted_Features[ind][:]=Feat_oneimage.squeeze()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Creating the RDMS of the test set of 78images \n",
    "from itertools import combinations\n",
    "N=78\n",
    "RDM_corr=np.zeros([N,N])\n",
    "combAB = combinations(range(0,N),2)\n",
    "combAB=list(combAB)\n",
    "Nc=len(combAB)\n",
    "count=0\n",
    "for ij in combAB:\n",
    "    i=ij[0]\n",
    "    j=ij[1]\n",
    "    RDM_corr[i][j] = 1-np.corrcoef(Predicted_Features[i][:],Predicted_Features[j][:])[0][1]\n",
    "    RDM_corr[j][i]=  RDM_corr[i][j]\n",
    "    count=count+1;"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "plt.matshow(RDM_corr);  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import scipy.io as sio\n",
    "if(flag_EL==1):\n",
    "    sio.savemat(path_test_image_pairs/'../RDM_EVS_layer_12_res_18_ver5.mat', {'evs':RDM_corr})\n",
    "if(flag_EL==2):\n",
    "    sio.savemat(path_test_image_pairs/'../RDM_IT_layer_12_res_18_ver5.mat', {'it':RDM_corr}) \n",
    "plt.matshow(RDM_corr);    "
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}